linux-stable/drivers/scsi/lpfc/lpfc_hbadisc.c

7272 lines
211 KiB
C
Raw Normal View History

/*******************************************************************
* This file is part of the Emulex Linux Device Driver for *
* Fibre Channel Host Bus Adapters. *
* Copyright (C) 2017-2023 Broadcom. All Rights Reserved. The term *
* Broadcom refers to Broadcom Inc. and/or its subsidiaries. *
* Copyright (C) 2004-2016 Emulex. All rights reserved. *
* EMULEX and SLI are trademarks of Emulex. *
* www.broadcom.com *
* Portions Copyright (C) 2004-2005 Christoph Hellwig *
* *
* This program is free software; you can redistribute it and/or *
* modify it under the terms of version 2 of the GNU General *
* Public License as published by the Free Software Foundation. *
* This program is distributed in the hope that it will be useful. *
* ALL EXPRESS OR IMPLIED CONDITIONS, REPRESENTATIONS AND *
* WARRANTIES, INCLUDING ANY IMPLIED WARRANTY OF MERCHANTABILITY, *
* FITNESS FOR A PARTICULAR PURPOSE, OR NON-INFRINGEMENT, ARE *
* DISCLAIMED, EXCEPT TO THE EXTENT THAT SUCH DISCLAIMERS ARE HELD *
* TO BE LEGALLY INVALID. See the GNU General Public License for *
* more details, a copy of which can be found in the file COPYING *
* included with this package. *
*******************************************************************/
#include <linux/blkdev.h>
#include <linux/delay.h>
include cleanup: Update gfp.h and slab.h includes to prepare for breaking implicit slab.h inclusion from percpu.h percpu.h is included by sched.h and module.h and thus ends up being included when building most .c files. percpu.h includes slab.h which in turn includes gfp.h making everything defined by the two files universally available and complicating inclusion dependencies. percpu.h -> slab.h dependency is about to be removed. Prepare for this change by updating users of gfp and slab facilities include those headers directly instead of assuming availability. As this conversion needs to touch large number of source files, the following script is used as the basis of conversion. http://userweb.kernel.org/~tj/misc/slabh-sweep.py The script does the followings. * Scan files for gfp and slab usages and update includes such that only the necessary includes are there. ie. if only gfp is used, gfp.h, if slab is used, slab.h. * When the script inserts a new include, it looks at the include blocks and try to put the new include such that its order conforms to its surrounding. It's put in the include block which contains core kernel includes, in the same order that the rest are ordered - alphabetical, Christmas tree, rev-Xmas-tree or at the end if there doesn't seem to be any matching order. * If the script can't find a place to put a new include (mostly because the file doesn't have fitting include block), it prints out an error message indicating which .h file needs to be added to the file. The conversion was done in the following steps. 1. The initial automatic conversion of all .c files updated slightly over 4000 files, deleting around 700 includes and adding ~480 gfp.h and ~3000 slab.h inclusions. The script emitted errors for ~400 files. 2. Each error was manually checked. Some didn't need the inclusion, some needed manual addition while adding it to implementation .h or embedding .c file was more appropriate for others. This step added inclusions to around 150 files. 3. The script was run again and the output was compared to the edits from #2 to make sure no file was left behind. 4. Several build tests were done and a couple of problems were fixed. e.g. lib/decompress_*.c used malloc/free() wrappers around slab APIs requiring slab.h to be added manually. 5. The script was run on all .h files but without automatically editing them as sprinkling gfp.h and slab.h inclusions around .h files could easily lead to inclusion dependency hell. Most gfp.h inclusion directives were ignored as stuff from gfp.h was usually wildly available and often used in preprocessor macros. Each slab.h inclusion directive was examined and added manually as necessary. 6. percpu.h was updated not to include slab.h. 7. Build test were done on the following configurations and failures were fixed. CONFIG_GCOV_KERNEL was turned off for all tests (as my distributed build env didn't work with gcov compiles) and a few more options had to be turned off depending on archs to make things build (like ipr on powerpc/64 which failed due to missing writeq). * x86 and x86_64 UP and SMP allmodconfig and a custom test config. * powerpc and powerpc64 SMP allmodconfig * sparc and sparc64 SMP allmodconfig * ia64 SMP allmodconfig * s390 SMP allmodconfig * alpha SMP allmodconfig * um on x86_64 SMP allmodconfig 8. percpu.h modifications were reverted so that it could be applied as a separate patch and serve as bisection point. Given the fact that I had only a couple of failures from tests on step 6, I'm fairly confident about the coverage of this conversion patch. If there is a breakage, it's likely to be something in one of the arch headers which should be easily discoverable easily on most builds of the specific arch. Signed-off-by: Tejun Heo <tj@kernel.org> Guess-its-ok-by: Christoph Lameter <cl@linux-foundation.org> Cc: Ingo Molnar <mingo@redhat.com> Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
2010-03-24 08:04:11 +00:00
#include <linux/slab.h>
#include <linux/pci.h>
#include <linux/kthread.h>
#include <linux/interrupt.h>
#include <linux/lockdep.h>
#include <linux/utsname.h>
#include <scsi/scsi.h>
#include <scsi/scsi_device.h>
#include <scsi/scsi_host.h>
#include <scsi/scsi_transport_fc.h>
#include <scsi/fc/fc_fs.h>
#include "lpfc_hw4.h"
#include "lpfc_hw.h"
#include "lpfc_nl.h"
#include "lpfc_disc.h"
#include "lpfc_sli.h"
#include "lpfc_sli4.h"
#include "lpfc.h"
#include "lpfc_scsi.h"
#include "lpfc_nvme.h"
#include "lpfc_logmsg.h"
#include "lpfc_crtn.h"
#include "lpfc_vport.h"
#include "lpfc_debugfs.h"
/* AlpaArray for assignment of scsid for scan-down and bind_method */
static uint8_t lpfcAlpaArray[] = {
0xEF, 0xE8, 0xE4, 0xE2, 0xE1, 0xE0, 0xDC, 0xDA, 0xD9, 0xD6,
0xD5, 0xD4, 0xD3, 0xD2, 0xD1, 0xCE, 0xCD, 0xCC, 0xCB, 0xCA,
0xC9, 0xC7, 0xC6, 0xC5, 0xC3, 0xBC, 0xBA, 0xB9, 0xB6, 0xB5,
0xB4, 0xB3, 0xB2, 0xB1, 0xAE, 0xAD, 0xAC, 0xAB, 0xAA, 0xA9,
0xA7, 0xA6, 0xA5, 0xA3, 0x9F, 0x9E, 0x9D, 0x9B, 0x98, 0x97,
0x90, 0x8F, 0x88, 0x84, 0x82, 0x81, 0x80, 0x7C, 0x7A, 0x79,
0x76, 0x75, 0x74, 0x73, 0x72, 0x71, 0x6E, 0x6D, 0x6C, 0x6B,
0x6A, 0x69, 0x67, 0x66, 0x65, 0x63, 0x5C, 0x5A, 0x59, 0x56,
0x55, 0x54, 0x53, 0x52, 0x51, 0x4E, 0x4D, 0x4C, 0x4B, 0x4A,
0x49, 0x47, 0x46, 0x45, 0x43, 0x3C, 0x3A, 0x39, 0x36, 0x35,
0x34, 0x33, 0x32, 0x31, 0x2E, 0x2D, 0x2C, 0x2B, 0x2A, 0x29,
0x27, 0x26, 0x25, 0x23, 0x1F, 0x1E, 0x1D, 0x1B, 0x18, 0x17,
0x10, 0x0F, 0x08, 0x04, 0x02, 0x01
};
static void lpfc_disc_timeout_handler(struct lpfc_vport *);
static void lpfc_disc_flush_list(struct lpfc_vport *vport);
static void lpfc_unregister_fcfi_cmpl(struct lpfc_hba *, LPFC_MBOXQ_t *);
static int lpfc_fcf_inuse(struct lpfc_hba *);
static void lpfc_mbx_cmpl_read_sparam(struct lpfc_hba *, LPFC_MBOXQ_t *);
static void lpfc_check_inactive_vmid(struct lpfc_hba *phba);
static void lpfc_check_vmid_qfpa_issue(struct lpfc_hba *phba);
static int
lpfc_valid_xpt_node(struct lpfc_nodelist *ndlp)
{
if (ndlp->nlp_fc4_type ||
ndlp->nlp_type & NLP_FABRIC)
return 1;
return 0;
}
/* The source of a terminate rport I/O is either a dev_loss_tmo
* event or a call to fc_remove_host. While the rport should be
* valid during these downcalls, the transport can call twice
* in a single event. This routine provides somoe protection
* as the NDLP isn't really free, just released to the pool.
*/
static int
lpfc_rport_invalid(struct fc_rport *rport)
{
struct lpfc_rport_data *rdata;
struct lpfc_nodelist *ndlp;
if (!rport) {
pr_err("**** %s: NULL rport, exit.\n", __func__);
return -EINVAL;
}
rdata = rport->dd_data;
if (!rdata) {
pr_err("**** %s: NULL dd_data on rport x%px SID x%x\n",
__func__, rport, rport->scsi_target_id);
return -EINVAL;
}
ndlp = rdata->pnode;
if (!rdata->pnode) {
scsi: lpfc: Improve PCI EEH Error and Recovery Handling Following EEH errors, the driver can crash or hang when deleting the localport or when attempting to unload. The EEH handlers in the driver did not notify the NVMe-FC transport before tearing the driver down. This was delayed until the resume steps. This worked for SCSI because lpfc_block_scsi() would notify the scsi_fc_transport that the target was not available but it would not clean up all the references to the ndlp. The SLI3 prep for dev reset handler did the lpfc_offline_prep() and lpfc_offline() calls to get the port stopped before restarting. The SLI4 version of the prep for dev reset just destroyed the queues and did not stop NVMe from continuing. Also because the port was not really stopped the localport destroy would hang because the transport was still waiting for I/O. Additionally, a devloss tmo can fire and post events to a stopped worker thread creating another hang condition. lpfc_sli4_prep_dev_for_reset() is modified to call lpfc_offline_prep() and lpfc_offline() rather than just lpfc_scsi_dev_block() to ensure both SCSI and NVMe transports are notified to block I/O to the driver. Logic is added to devloss handler and worker thread to clean up ndlp references and quiesce appropriately. Link: https://lore.kernel.org/r/20220317032737.45308-2-jsmart2021@gmail.com Co-developed-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2022-03-17 03:27:34 +00:00
pr_info("**** %s: NULL ndlp on rport x%px SID x%x\n",
__func__, rport, rport->scsi_target_id);
return -EINVAL;
}
if (!ndlp->vport) {
pr_err("**** %s: Null vport on ndlp x%px, DID x%x rport x%px "
"SID x%x\n", __func__, ndlp, ndlp->nlp_DID, rport,
rport->scsi_target_id);
return -EINVAL;
}
return 0;
}
void
lpfc_terminate_rport_io(struct fc_rport *rport)
{
struct lpfc_rport_data *rdata;
struct lpfc_nodelist *ndlp;
struct lpfc_vport *vport;
if (lpfc_rport_invalid(rport))
return;
rdata = rport->dd_data;
ndlp = rdata->pnode;
vport = ndlp->vport;
lpfc_debugfs_disc_trc(vport, LPFC_DISC_TRC_RPORT,
"rport terminate: sid:x%x did:x%x flg:x%x",
ndlp->nlp_sid, ndlp->nlp_DID, ndlp->nlp_flag);
if (ndlp->nlp_sid != NLP_NO_SID)
lpfc_sli_abort_iocb(vport, ndlp->nlp_sid, 0, LPFC_CTX_TGT);
}
/*
* This function will be called when dev_loss_tmo fire.
*/
void
lpfc_dev_loss_tmo_callbk(struct fc_rport *rport)
{
struct lpfc_nodelist *ndlp;
struct lpfc_vport *vport;
struct lpfc_hba *phba;
struct lpfc_work_evt *evtp;
unsigned long iflags;
ndlp = ((struct lpfc_rport_data *)rport->dd_data)->pnode;
scsi: lpfc: Rework remote port ref counting and node freeing When a remote port is disconnected and disappears, its node structure (ndlp) stays allocated and on a vport node list. While on the list it can be matched, thus requires validation checks on state to be added in numerous code paths. If the node comes back, its possible for there to be multiple node structures for the same device on the vport node list. There is no reason to keep the node structure around after it is no longer in existence, and the current implementation creates problems for itself (multiple nodes) and lots of unnecessary code for state validation. Additionally, the reference taking on the node structure didn't follow the normal model used by the kernel kref api. It included lots of odd logic to match state with reference count. The combination of this odd logic plus the way it was implicitly used in the discovery engine made its reference taking implementation suspect and extremely hard to follow. Change the driver such that the reference taking routines are now normal ref increments/decrements and callout on refcount=0. With this in place, the rework can be done such that the node structure is fully removed and deallocated when the remote port no longer exists and all references are removed. This removal logic, and the basic ref counting are intrically tied, thus in a single patch. Link: https://lore.kernel.org/r/20201115192646.12977-2-james.smart@broadcom.com Co-developed-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <james.smart@broadcom.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-11-15 19:26:30 +00:00
if (!ndlp)
return;
vport = ndlp->vport;
phba = vport->phba;
lpfc_debugfs_disc_trc(vport, LPFC_DISC_TRC_RPORT,
"rport devlosscb: sid:x%x did:x%x flg:x%x",
ndlp->nlp_sid, ndlp->nlp_DID, ndlp->nlp_flag);
lpfc_printf_vlog(ndlp->vport, KERN_INFO, LOG_NODE,
"3181 dev_loss_callbk x%06x, rport x%px flg x%x "
scsi: lpfc: Improve PCI EEH Error and Recovery Handling Following EEH errors, the driver can crash or hang when deleting the localport or when attempting to unload. The EEH handlers in the driver did not notify the NVMe-FC transport before tearing the driver down. This was delayed until the resume steps. This worked for SCSI because lpfc_block_scsi() would notify the scsi_fc_transport that the target was not available but it would not clean up all the references to the ndlp. The SLI3 prep for dev reset handler did the lpfc_offline_prep() and lpfc_offline() calls to get the port stopped before restarting. The SLI4 version of the prep for dev reset just destroyed the queues and did not stop NVMe from continuing. Also because the port was not really stopped the localport destroy would hang because the transport was still waiting for I/O. Additionally, a devloss tmo can fire and post events to a stopped worker thread creating another hang condition. lpfc_sli4_prep_dev_for_reset() is modified to call lpfc_offline_prep() and lpfc_offline() rather than just lpfc_scsi_dev_block() to ensure both SCSI and NVMe transports are notified to block I/O to the driver. Logic is added to devloss handler and worker thread to clean up ndlp references and quiesce appropriately. Link: https://lore.kernel.org/r/20220317032737.45308-2-jsmart2021@gmail.com Co-developed-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2022-03-17 03:27:34 +00:00
"load_flag x%x refcnt %d state %d xpt x%x\n",
ndlp->nlp_DID, ndlp->rport, ndlp->nlp_flag,
scsi: lpfc: Improve PCI EEH Error and Recovery Handling Following EEH errors, the driver can crash or hang when deleting the localport or when attempting to unload. The EEH handlers in the driver did not notify the NVMe-FC transport before tearing the driver down. This was delayed until the resume steps. This worked for SCSI because lpfc_block_scsi() would notify the scsi_fc_transport that the target was not available but it would not clean up all the references to the ndlp. The SLI3 prep for dev reset handler did the lpfc_offline_prep() and lpfc_offline() calls to get the port stopped before restarting. The SLI4 version of the prep for dev reset just destroyed the queues and did not stop NVMe from continuing. Also because the port was not really stopped the localport destroy would hang because the transport was still waiting for I/O. Additionally, a devloss tmo can fire and post events to a stopped worker thread creating another hang condition. lpfc_sli4_prep_dev_for_reset() is modified to call lpfc_offline_prep() and lpfc_offline() rather than just lpfc_scsi_dev_block() to ensure both SCSI and NVMe transports are notified to block I/O to the driver. Logic is added to devloss handler and worker thread to clean up ndlp references and quiesce appropriately. Link: https://lore.kernel.org/r/20220317032737.45308-2-jsmart2021@gmail.com Co-developed-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2022-03-17 03:27:34 +00:00
vport->load_flag, kref_read(&ndlp->kref),
ndlp->nlp_state, ndlp->fc4_xpt_flags);
/* Don't schedule a worker thread event if the vport is going down.
* The teardown process cleans up the node via lpfc_drop_node.
*/
if (vport->load_flag & FC_UNLOADING) {
((struct lpfc_rport_data *)rport->dd_data)->pnode = NULL;
ndlp->rport = NULL;
ndlp->fc4_xpt_flags &= ~SCSI_XPT_REGD;
scsi: lpfc: Improve PCI EEH Error and Recovery Handling Following EEH errors, the driver can crash or hang when deleting the localport or when attempting to unload. The EEH handlers in the driver did not notify the NVMe-FC transport before tearing the driver down. This was delayed until the resume steps. This worked for SCSI because lpfc_block_scsi() would notify the scsi_fc_transport that the target was not available but it would not clean up all the references to the ndlp. The SLI3 prep for dev reset handler did the lpfc_offline_prep() and lpfc_offline() calls to get the port stopped before restarting. The SLI4 version of the prep for dev reset just destroyed the queues and did not stop NVMe from continuing. Also because the port was not really stopped the localport destroy would hang because the transport was still waiting for I/O. Additionally, a devloss tmo can fire and post events to a stopped worker thread creating another hang condition. lpfc_sli4_prep_dev_for_reset() is modified to call lpfc_offline_prep() and lpfc_offline() rather than just lpfc_scsi_dev_block() to ensure both SCSI and NVMe transports are notified to block I/O to the driver. Logic is added to devloss handler and worker thread to clean up ndlp references and quiesce appropriately. Link: https://lore.kernel.org/r/20220317032737.45308-2-jsmart2021@gmail.com Co-developed-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2022-03-17 03:27:34 +00:00
/* clear the NLP_XPT_REGD if the node is not registered
* with nvme-fc
*/
if (ndlp->fc4_xpt_flags == NLP_XPT_REGD)
ndlp->fc4_xpt_flags &= ~NLP_XPT_REGD;
/* Remove the node reference from remote_port_add now.
* The driver will not call remote_port_delete.
*/
lpfc_nlp_put(ndlp);
return;
}
if (ndlp->nlp_state == NLP_STE_MAPPED_NODE)
return;
if (rport->port_name != wwn_to_u64(ndlp->nlp_portname.u.wwn))
scsi: lpfc: Add an internal trace log buffer The current logging methods typically end up requesting a reproduction with a different logging level set to figure out what happened. This was mainly by design to not clutter the kernel log messages with things that were typically not interesting and the messages themselves could cause other issues. When looking to make a better system, it was seen that in many cases when more data was wanted was when another message, usually at KERN_ERR level, was logged. And in most cases, what the additional logging that was then enabled was typically. Most of these areas fell into the discovery machine. Based on this summary, the following design has been put in place: The driver will maintain an internal log (256 elements of 256 bytes). The "additional logging" messages that are usually enabled in a reproduction will be changed to now log all the time to the internal log. A new logging level is defined - LOG_TRACE_EVENT. When this level is set (it is not by default) and a message marked as KERN_ERR is logged, all the messages in the internal log will be dumped to the kernel log before the KERN_ERR message is logged. There is a timestamp on each message added to the internal log. However, this timestamp is not converted to wall time when logged. The value of the timestamp is solely to give a crude time reference for the messages. Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-06-30 21:50:00 +00:00
lpfc_printf_vlog(vport, KERN_ERR, LOG_TRACE_EVENT,
"6789 rport name %llx != node port name %llx",
rport->port_name,
wwn_to_u64(ndlp->nlp_portname.u.wwn));
evtp = &ndlp->dev_loss_evt;
if (!list_empty(&evtp->evt_listp)) {
scsi: lpfc: Add an internal trace log buffer The current logging methods typically end up requesting a reproduction with a different logging level set to figure out what happened. This was mainly by design to not clutter the kernel log messages with things that were typically not interesting and the messages themselves could cause other issues. When looking to make a better system, it was seen that in many cases when more data was wanted was when another message, usually at KERN_ERR level, was logged. And in most cases, what the additional logging that was then enabled was typically. Most of these areas fell into the discovery machine. Based on this summary, the following design has been put in place: The driver will maintain an internal log (256 elements of 256 bytes). The "additional logging" messages that are usually enabled in a reproduction will be changed to now log all the time to the internal log. A new logging level is defined - LOG_TRACE_EVENT. When this level is set (it is not by default) and a message marked as KERN_ERR is logged, all the messages in the internal log will be dumped to the kernel log before the KERN_ERR message is logged. There is a timestamp on each message added to the internal log. However, this timestamp is not converted to wall time when logged. The value of the timestamp is solely to give a crude time reference for the messages. Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-06-30 21:50:00 +00:00
lpfc_printf_vlog(vport, KERN_ERR, LOG_TRACE_EVENT,
"6790 rport name %llx dev_loss_evt pending\n",
scsi: lpfc: Add an internal trace log buffer The current logging methods typically end up requesting a reproduction with a different logging level set to figure out what happened. This was mainly by design to not clutter the kernel log messages with things that were typically not interesting and the messages themselves could cause other issues. When looking to make a better system, it was seen that in many cases when more data was wanted was when another message, usually at KERN_ERR level, was logged. And in most cases, what the additional logging that was then enabled was typically. Most of these areas fell into the discovery machine. Based on this summary, the following design has been put in place: The driver will maintain an internal log (256 elements of 256 bytes). The "additional logging" messages that are usually enabled in a reproduction will be changed to now log all the time to the internal log. A new logging level is defined - LOG_TRACE_EVENT. When this level is set (it is not by default) and a message marked as KERN_ERR is logged, all the messages in the internal log will be dumped to the kernel log before the KERN_ERR message is logged. There is a timestamp on each message added to the internal log. However, this timestamp is not converted to wall time when logged. The value of the timestamp is solely to give a crude time reference for the messages. Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-06-30 21:50:00 +00:00
rport->port_name);
return;
}
spin_lock_irqsave(&ndlp->lock, iflags);
ndlp->nlp_flag |= NLP_IN_DEV_LOSS;
/* If there is a PLOGI in progress, and we are in a
* NLP_NPR_2B_DISC state, don't turn off the flag.
*/
if (ndlp->nlp_state != NLP_STE_PLOGI_ISSUE)
ndlp->nlp_flag &= ~NLP_NPR_2B_DISC;
/*
* The backend does not expect any more calls associated with this
* rport. Remove the association between rport and ndlp.
*/
ndlp->fc4_xpt_flags &= ~SCSI_XPT_REGD;
((struct lpfc_rport_data *)rport->dd_data)->pnode = NULL;
ndlp->rport = NULL;
spin_unlock_irqrestore(&ndlp->lock, iflags);
scsi: lpfc: Improve PCI EEH Error and Recovery Handling Following EEH errors, the driver can crash or hang when deleting the localport or when attempting to unload. The EEH handlers in the driver did not notify the NVMe-FC transport before tearing the driver down. This was delayed until the resume steps. This worked for SCSI because lpfc_block_scsi() would notify the scsi_fc_transport that the target was not available but it would not clean up all the references to the ndlp. The SLI3 prep for dev reset handler did the lpfc_offline_prep() and lpfc_offline() calls to get the port stopped before restarting. The SLI4 version of the prep for dev reset just destroyed the queues and did not stop NVMe from continuing. Also because the port was not really stopped the localport destroy would hang because the transport was still waiting for I/O. Additionally, a devloss tmo can fire and post events to a stopped worker thread creating another hang condition. lpfc_sli4_prep_dev_for_reset() is modified to call lpfc_offline_prep() and lpfc_offline() rather than just lpfc_scsi_dev_block() to ensure both SCSI and NVMe transports are notified to block I/O to the driver. Logic is added to devloss handler and worker thread to clean up ndlp references and quiesce appropriately. Link: https://lore.kernel.org/r/20220317032737.45308-2-jsmart2021@gmail.com Co-developed-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2022-03-17 03:27:34 +00:00
if (phba->worker_thread) {
/* We need to hold the node by incrementing the reference
* count until this queued work is done
*/
evtp->evt_arg1 = lpfc_nlp_get(ndlp);
spin_lock_irqsave(&phba->hbalock, iflags);
if (evtp->evt_arg1) {
evtp->evt = LPFC_EVT_DEV_LOSS;
list_add_tail(&evtp->evt_listp, &phba->work_list);
lpfc_worker_wake_up(phba);
}
spin_unlock_irqrestore(&phba->hbalock, iflags);
} else {
lpfc_printf_vlog(ndlp->vport, KERN_INFO, LOG_NODE,
"3188 worker thread is stopped %s x%06x, "
" rport x%px flg x%x load_flag x%x refcnt "
"%d\n", __func__, ndlp->nlp_DID,
ndlp->rport, ndlp->nlp_flag,
vport->load_flag, kref_read(&ndlp->kref));
if (!(ndlp->fc4_xpt_flags & NVME_XPT_REGD)) {
spin_lock_irqsave(&ndlp->lock, iflags);
/* Node is in dev loss. No further transaction. */
ndlp->nlp_flag &= ~NLP_IN_DEV_LOSS;
spin_unlock_irqrestore(&ndlp->lock, iflags);
lpfc_disc_state_machine(vport, ndlp, NULL,
NLP_EVT_DEVICE_RM);
}
}
return;
}
/**
* lpfc_check_inactive_vmid_one - VMID inactivity checker for a vport
* @vport: Pointer to vport context object.
*
* This function checks for idle VMID entries related to a particular vport. If
* found unused/idle, free them accordingly.
**/
static void lpfc_check_inactive_vmid_one(struct lpfc_vport *vport)
{
u16 keep;
u32 difftime = 0, r, bucket;
u64 *lta;
int cpu;
struct lpfc_vmid *vmp;
write_lock(&vport->vmid_lock);
if (!vport->cur_vmid_cnt)
goto out;
/* iterate through the table */
hash_for_each(vport->hash_table, bucket, vmp, hnode) {
keep = 0;
if (vmp->flag & LPFC_VMID_REGISTERED) {
/* check if the particular VMID is in use */
/* for all available per cpu variable */
for_each_possible_cpu(cpu) {
/* if last access time is less than timeout */
lta = per_cpu_ptr(vmp->last_io_time, cpu);
if (!lta)
continue;
difftime = (jiffies) - (*lta);
if ((vport->vmid_inactivity_timeout *
JIFFIES_PER_HR) > difftime) {
keep = 1;
break;
}
}
/* if none of the cpus have been used by the vm, */
/* remove the entry if already registered */
if (!keep) {
/* mark the entry for deregistration */
vmp->flag = LPFC_VMID_DE_REGISTER;
write_unlock(&vport->vmid_lock);
if (vport->vmid_priority_tagging)
r = lpfc_vmid_uvem(vport, vmp, false);
else
r = lpfc_vmid_cmd(vport,
SLI_CTAS_DAPP_IDENT,
vmp);
/* decrement number of active vms and mark */
/* entry in slot as free */
write_lock(&vport->vmid_lock);
if (!r) {
struct lpfc_vmid *ht = vmp;
vport->cur_vmid_cnt--;
ht->flag = LPFC_VMID_SLOT_FREE;
free_percpu(ht->last_io_time);
ht->last_io_time = NULL;
hash_del(&ht->hnode);
}
}
}
}
out:
write_unlock(&vport->vmid_lock);
}
/**
* lpfc_check_inactive_vmid - VMID inactivity checker
* @phba: Pointer to hba context object.
*
* This function is called from the worker thread to determine if an entry in
* the VMID table can be released since there was no I/O activity seen from that
* particular VM for the specified time. When this happens, the entry in the
* table is released and also the resources on the switch cleared.
**/
static void lpfc_check_inactive_vmid(struct lpfc_hba *phba)
{
struct lpfc_vport *vport;
struct lpfc_vport **vports;
int i;
vports = lpfc_create_vport_work_array(phba);
if (!vports)
return;
for (i = 0; i <= phba->max_vports; i++) {
if ((!vports[i]) && (i == 0))
vport = phba->pport;
else
vport = vports[i];
if (!vport)
break;
lpfc_check_inactive_vmid_one(vport);
}
lpfc_destroy_vport_work_array(phba, vports);
}
/**
* lpfc_check_nlp_post_devloss - Check to restore ndlp refcnt after devloss
* @vport: Pointer to vport object.
* @ndlp: Pointer to remote node object.
*
* If NLP_IN_RECOV_POST_DEV_LOSS flag was set due to outstanding recovery of
* node during dev_loss_tmo processing, then this function restores the nlp_put
* kref decrement from lpfc_dev_loss_tmo_handler.
**/
void
lpfc_check_nlp_post_devloss(struct lpfc_vport *vport,
struct lpfc_nodelist *ndlp)
{
unsigned long iflags;
spin_lock_irqsave(&ndlp->lock, iflags);
if (ndlp->save_flags & NLP_IN_RECOV_POST_DEV_LOSS) {
ndlp->save_flags &= ~NLP_IN_RECOV_POST_DEV_LOSS;
spin_unlock_irqrestore(&ndlp->lock, iflags);
lpfc_nlp_get(ndlp);
lpfc_printf_vlog(vport, KERN_INFO, LOG_DISCOVERY | LOG_NODE,
"8438 Devloss timeout reversed on DID x%x "
"refcnt %d ndlp %p flag x%x "
"port_state = x%x\n",
ndlp->nlp_DID, kref_read(&ndlp->kref), ndlp,
ndlp->nlp_flag, vport->port_state);
spin_lock_irqsave(&ndlp->lock, iflags);
}
spin_unlock_irqrestore(&ndlp->lock, iflags);
}
/**
* lpfc_dev_loss_tmo_handler - Remote node devloss timeout handler
* @ndlp: Pointer to remote node object.
*
* This function is called from the worker thread when devloss timeout timer
* expires. For SLI4 host, this routine shall return 1 when at lease one
* remote node, including this @ndlp, is still in use of FCF; otherwise, this
* routine shall return 0 when there is no remote node is still in use of FCF
* when devloss timeout happened to this @ndlp.
**/
static int
lpfc_dev_loss_tmo_handler(struct lpfc_nodelist *ndlp)
{
struct lpfc_vport *vport;
struct lpfc_hba *phba;
uint8_t *name;
int warn_on = 0;
int fcf_inuse = 0;
bool recovering = false;
struct fc_vport *fc_vport = NULL;
unsigned long iflags;
vport = ndlp->vport;
name = (uint8_t *)&ndlp->nlp_portname;
phba = vport->phba;
if (phba->sli_rev == LPFC_SLI_REV4)
fcf_inuse = lpfc_fcf_inuse(phba);
lpfc_debugfs_disc_trc(vport, LPFC_DISC_TRC_RPORT,
"rport devlosstmo:did:x%x type:x%x id:x%x",
ndlp->nlp_DID, ndlp->nlp_type, ndlp->nlp_sid);
lpfc_printf_vlog(ndlp->vport, KERN_INFO, LOG_NODE,
"3182 %s x%06x, nflag x%x xflags x%x refcnt %d\n",
__func__, ndlp->nlp_DID, ndlp->nlp_flag,
ndlp->fc4_xpt_flags, kref_read(&ndlp->kref));
/* If the driver is recovering the rport, ignore devloss. */
if (ndlp->nlp_state == NLP_STE_MAPPED_NODE) {
lpfc_printf_vlog(vport, KERN_INFO, LOG_DISCOVERY,
"0284 Devloss timeout Ignored on "
"WWPN %x:%x:%x:%x:%x:%x:%x:%x "
"NPort x%x\n",
*name, *(name+1), *(name+2), *(name+3),
*(name+4), *(name+5), *(name+6), *(name+7),
ndlp->nlp_DID);
spin_lock_irqsave(&ndlp->lock, iflags);
ndlp->nlp_flag &= ~NLP_IN_DEV_LOSS;
spin_unlock_irqrestore(&ndlp->lock, iflags);
return fcf_inuse;
}
/* Fabric nodes are done. */
if (ndlp->nlp_type & NLP_FABRIC) {
spin_lock_irqsave(&ndlp->lock, iflags);
/* In massive vport configuration settings or when the FLOGI
* completes with a sequence timeout, it's possible
* dev_loss_tmo fired during node recovery. The driver has to
* account for this race to allow for recovery and keep
* the reference counting correct.
*/
switch (ndlp->nlp_DID) {
case Fabric_DID:
fc_vport = vport->fc_vport;
if (fc_vport) {
/* NPIV path. */
if (fc_vport->vport_state ==
FC_VPORT_INITIALIZING)
recovering = true;
} else {
/* Physical port path. */
if (phba->hba_flag & HBA_FLOGI_OUTSTANDING)
recovering = true;
}
break;
case Fabric_Cntl_DID:
if (ndlp->nlp_flag & NLP_REG_LOGIN_SEND)
recovering = true;
break;
case FDMI_DID:
fallthrough;
case NameServer_DID:
if (ndlp->nlp_state >= NLP_STE_PLOGI_ISSUE &&
ndlp->nlp_state <= NLP_STE_REG_LOGIN_ISSUE)
recovering = true;
break;
}
spin_unlock_irqrestore(&ndlp->lock, iflags);
/* Mark an NLP_IN_RECOV_POST_DEV_LOSS flag to know if reversing
* the following lpfc_nlp_put is necessary after fabric node is
* recovered.
*/
if (recovering) {
lpfc_printf_vlog(vport, KERN_INFO,
LOG_DISCOVERY | LOG_NODE,
"8436 Devloss timeout marked on "
"DID x%x refcnt %d ndlp %p "
"flag x%x port_state = x%x\n",
ndlp->nlp_DID, kref_read(&ndlp->kref),
ndlp, ndlp->nlp_flag,
vport->port_state);
spin_lock_irqsave(&ndlp->lock, iflags);
ndlp->save_flags |= NLP_IN_RECOV_POST_DEV_LOSS;
spin_unlock_irqrestore(&ndlp->lock, iflags);
} else if (ndlp->nlp_state == NLP_STE_UNMAPPED_NODE) {
/* Fabric node fully recovered before this dev_loss_tmo
* queue work is processed. Thus, ignore the
* dev_loss_tmo event.
*/
lpfc_printf_vlog(vport, KERN_INFO,
LOG_DISCOVERY | LOG_NODE,
"8437 Devloss timeout ignored on "
"DID x%x refcnt %d ndlp %p "
"flag x%x port_state = x%x\n",
ndlp->nlp_DID, kref_read(&ndlp->kref),
ndlp, ndlp->nlp_flag,
vport->port_state);
return fcf_inuse;
}
spin_lock_irqsave(&ndlp->lock, iflags);
ndlp->nlp_flag &= ~NLP_IN_DEV_LOSS;
spin_unlock_irqrestore(&ndlp->lock, iflags);
lpfc_nlp_put(ndlp);
return fcf_inuse;
}
if (ndlp->nlp_sid != NLP_NO_SID) {
warn_on = 1;
lpfc_sli_abort_iocb(vport, ndlp->nlp_sid, 0, LPFC_CTX_TGT);
}
if (warn_on) {
scsi: lpfc: Add an internal trace log buffer The current logging methods typically end up requesting a reproduction with a different logging level set to figure out what happened. This was mainly by design to not clutter the kernel log messages with things that were typically not interesting and the messages themselves could cause other issues. When looking to make a better system, it was seen that in many cases when more data was wanted was when another message, usually at KERN_ERR level, was logged. And in most cases, what the additional logging that was then enabled was typically. Most of these areas fell into the discovery machine. Based on this summary, the following design has been put in place: The driver will maintain an internal log (256 elements of 256 bytes). The "additional logging" messages that are usually enabled in a reproduction will be changed to now log all the time to the internal log. A new logging level is defined - LOG_TRACE_EVENT. When this level is set (it is not by default) and a message marked as KERN_ERR is logged, all the messages in the internal log will be dumped to the kernel log before the KERN_ERR message is logged. There is a timestamp on each message added to the internal log. However, this timestamp is not converted to wall time when logged. The value of the timestamp is solely to give a crude time reference for the messages. Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-06-30 21:50:00 +00:00
lpfc_printf_vlog(vport, KERN_ERR, LOG_TRACE_EVENT,
"0203 Devloss timeout on "
[SCSI] lpfc 8.2.6 : Multiple discovery fixes Multiple Discovery Fixes: - Fix race on discovery due to link events coinciding with vport_delete. - Use NLP_FABRIC state to filter out switch-based pseudo initiators that reuse the same WWNs. - Correct erroneous setting of DID=0 in lpfc_matchdid() - Correct extra reference count that was in the lookup path for the remoteid from an unsolicited ELS. - Correct double-free bug in els abort path. - Correct FDMI server discovery logic for switch that return a WWN of 0. - Fix bugs in ndlp mgmt when a node changes address - Correct bug that did not delete RSCNs for vports upon link transitions - Fix "0216 Link event during NS query" error which pops up when vports are swapped to different switch ports. - Add sanity checks on ndlp structures - Fix devloss log message to dump WWN correctly - Hold off mgmt commands that were interferring with discovery mailbox cmds - Remove unnecessary FC_ESTABLISH_LINK logic. - Correct some race conditions in the worker thread, resulting in devloss: - Clear the work_port_events field before handling the work port events - Clear the deferred ring event before handling a deferred ring event - Hold the hba lock when waking up the work thread - Send an acc for the rscn even when we aren't going to handle it - Fix locking behavior that was not properly protecting the ACTIVE flag, thus allowing mailbox command order to shift. Signed-off-by: James Smart <james.smart@emulex.com> Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
2008-04-07 14:15:56 +00:00
"WWPN %02x:%02x:%02x:%02x:%02x:%02x:%02x:%02x "
scsi: lpfc: Improve PCI EEH Error and Recovery Handling Following EEH errors, the driver can crash or hang when deleting the localport or when attempting to unload. The EEH handlers in the driver did not notify the NVMe-FC transport before tearing the driver down. This was delayed until the resume steps. This worked for SCSI because lpfc_block_scsi() would notify the scsi_fc_transport that the target was not available but it would not clean up all the references to the ndlp. The SLI3 prep for dev reset handler did the lpfc_offline_prep() and lpfc_offline() calls to get the port stopped before restarting. The SLI4 version of the prep for dev reset just destroyed the queues and did not stop NVMe from continuing. Also because the port was not really stopped the localport destroy would hang because the transport was still waiting for I/O. Additionally, a devloss tmo can fire and post events to a stopped worker thread creating another hang condition. lpfc_sli4_prep_dev_for_reset() is modified to call lpfc_offline_prep() and lpfc_offline() rather than just lpfc_scsi_dev_block() to ensure both SCSI and NVMe transports are notified to block I/O to the driver. Logic is added to devloss handler and worker thread to clean up ndlp references and quiesce appropriately. Link: https://lore.kernel.org/r/20220317032737.45308-2-jsmart2021@gmail.com Co-developed-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2022-03-17 03:27:34 +00:00
"NPort x%06x Data: x%x x%x x%x refcnt %d\n",
*name, *(name+1), *(name+2), *(name+3),
*(name+4), *(name+5), *(name+6), *(name+7),
ndlp->nlp_DID, ndlp->nlp_flag,
scsi: lpfc: Improve PCI EEH Error and Recovery Handling Following EEH errors, the driver can crash or hang when deleting the localport or when attempting to unload. The EEH handlers in the driver did not notify the NVMe-FC transport before tearing the driver down. This was delayed until the resume steps. This worked for SCSI because lpfc_block_scsi() would notify the scsi_fc_transport that the target was not available but it would not clean up all the references to the ndlp. The SLI3 prep for dev reset handler did the lpfc_offline_prep() and lpfc_offline() calls to get the port stopped before restarting. The SLI4 version of the prep for dev reset just destroyed the queues and did not stop NVMe from continuing. Also because the port was not really stopped the localport destroy would hang because the transport was still waiting for I/O. Additionally, a devloss tmo can fire and post events to a stopped worker thread creating another hang condition. lpfc_sli4_prep_dev_for_reset() is modified to call lpfc_offline_prep() and lpfc_offline() rather than just lpfc_scsi_dev_block() to ensure both SCSI and NVMe transports are notified to block I/O to the driver. Logic is added to devloss handler and worker thread to clean up ndlp references and quiesce appropriately. Link: https://lore.kernel.org/r/20220317032737.45308-2-jsmart2021@gmail.com Co-developed-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2022-03-17 03:27:34 +00:00
ndlp->nlp_state, ndlp->nlp_rpi,
kref_read(&ndlp->kref));
} else {
scsi: lpfc: Add an internal trace log buffer The current logging methods typically end up requesting a reproduction with a different logging level set to figure out what happened. This was mainly by design to not clutter the kernel log messages with things that were typically not interesting and the messages themselves could cause other issues. When looking to make a better system, it was seen that in many cases when more data was wanted was when another message, usually at KERN_ERR level, was logged. And in most cases, what the additional logging that was then enabled was typically. Most of these areas fell into the discovery machine. Based on this summary, the following design has been put in place: The driver will maintain an internal log (256 elements of 256 bytes). The "additional logging" messages that are usually enabled in a reproduction will be changed to now log all the time to the internal log. A new logging level is defined - LOG_TRACE_EVENT. When this level is set (it is not by default) and a message marked as KERN_ERR is logged, all the messages in the internal log will be dumped to the kernel log before the KERN_ERR message is logged. There is a timestamp on each message added to the internal log. However, this timestamp is not converted to wall time when logged. The value of the timestamp is solely to give a crude time reference for the messages. Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-06-30 21:50:00 +00:00
lpfc_printf_vlog(vport, KERN_INFO, LOG_TRACE_EVENT,
"0204 Devloss timeout on "
[SCSI] lpfc 8.2.6 : Multiple discovery fixes Multiple Discovery Fixes: - Fix race on discovery due to link events coinciding with vport_delete. - Use NLP_FABRIC state to filter out switch-based pseudo initiators that reuse the same WWNs. - Correct erroneous setting of DID=0 in lpfc_matchdid() - Correct extra reference count that was in the lookup path for the remoteid from an unsolicited ELS. - Correct double-free bug in els abort path. - Correct FDMI server discovery logic for switch that return a WWN of 0. - Fix bugs in ndlp mgmt when a node changes address - Correct bug that did not delete RSCNs for vports upon link transitions - Fix "0216 Link event during NS query" error which pops up when vports are swapped to different switch ports. - Add sanity checks on ndlp structures - Fix devloss log message to dump WWN correctly - Hold off mgmt commands that were interferring with discovery mailbox cmds - Remove unnecessary FC_ESTABLISH_LINK logic. - Correct some race conditions in the worker thread, resulting in devloss: - Clear the work_port_events field before handling the work port events - Clear the deferred ring event before handling a deferred ring event - Hold the hba lock when waking up the work thread - Send an acc for the rscn even when we aren't going to handle it - Fix locking behavior that was not properly protecting the ACTIVE flag, thus allowing mailbox command order to shift. Signed-off-by: James Smart <james.smart@emulex.com> Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
2008-04-07 14:15:56 +00:00
"WWPN %02x:%02x:%02x:%02x:%02x:%02x:%02x:%02x "
"NPort x%06x Data: x%x x%x x%x\n",
*name, *(name+1), *(name+2), *(name+3),
*(name+4), *(name+5), *(name+6), *(name+7),
ndlp->nlp_DID, ndlp->nlp_flag,
ndlp->nlp_state, ndlp->nlp_rpi);
}
/* If we are devloss, but we are in the process of rediscovering the
* ndlp, don't issue a NLP_EVT_DEVICE_RM event.
*/
if (ndlp->nlp_state >= NLP_STE_PLOGI_ISSUE &&
ndlp->nlp_state <= NLP_STE_PRLI_ISSUE) {
return fcf_inuse;
}
spin_lock_irqsave(&ndlp->lock, iflags);
ndlp->nlp_flag &= ~NLP_IN_DEV_LOSS;
spin_unlock_irqrestore(&ndlp->lock, iflags);
if (!(ndlp->fc4_xpt_flags & NVME_XPT_REGD))
lpfc_disc_state_machine(vport, ndlp, NULL, NLP_EVT_DEVICE_RM);
return fcf_inuse;
}
static void lpfc_check_vmid_qfpa_issue(struct lpfc_hba *phba)
{
struct lpfc_vport *vport;
struct lpfc_vport **vports;
int i;
vports = lpfc_create_vport_work_array(phba);
if (!vports)
return;
for (i = 0; i <= phba->max_vports; i++) {
if ((!vports[i]) && (i == 0))
vport = phba->pport;
else
vport = vports[i];
if (!vport)
break;
if (vport->vmid_flag & LPFC_VMID_ISSUE_QFPA) {
if (!lpfc_issue_els_qfpa(vport))
vport->vmid_flag &= ~LPFC_VMID_ISSUE_QFPA;
}
}
lpfc_destroy_vport_work_array(phba, vports);
}
/**
* lpfc_sli4_post_dev_loss_tmo_handler - SLI4 post devloss timeout handler
* @phba: Pointer to hba context object.
* @fcf_inuse: SLI4 FCF in-use state reported from devloss timeout handler.
* @nlp_did: remote node identifer with devloss timeout.
*
* This function is called from the worker thread after invoking devloss
* timeout handler and releasing the reference count for the ndlp with
* which the devloss timeout was handled for SLI4 host. For the devloss
* timeout of the last remote node which had been in use of FCF, when this
* routine is invoked, it shall be guaranteed that none of the remote are
* in-use of FCF. When devloss timeout to the last remote using the FCF,
* if the FIP engine is neither in FCF table scan process nor roundrobin
* failover process, the in-use FCF shall be unregistered. If the FIP
* engine is in FCF discovery process, the devloss timeout state shall
* be set for either the FCF table scan process or roundrobin failover
* process to unregister the in-use FCF.
**/
static void
lpfc_sli4_post_dev_loss_tmo_handler(struct lpfc_hba *phba, int fcf_inuse,
uint32_t nlp_did)
{
/* If devloss timeout happened to a remote node when FCF had no
* longer been in-use, do nothing.
*/
if (!fcf_inuse)
return;
if ((phba->hba_flag & HBA_FIP_SUPPORT) && !lpfc_fcf_inuse(phba)) {
spin_lock_irq(&phba->hbalock);
if (phba->fcf.fcf_flag & FCF_DISCOVERY) {
if (phba->hba_flag & HBA_DEVLOSS_TMO) {
spin_unlock_irq(&phba->hbalock);
return;
}
phba->hba_flag |= HBA_DEVLOSS_TMO;
lpfc_printf_log(phba, KERN_INFO, LOG_FIP,
"2847 Last remote node (x%x) using "
"FCF devloss tmo\n", nlp_did);
}
if (phba->fcf.fcf_flag & FCF_REDISC_PROG) {
spin_unlock_irq(&phba->hbalock);
lpfc_printf_log(phba, KERN_INFO, LOG_FIP,
"2868 Devloss tmo to FCF rediscovery "
"in progress\n");
return;
}
if (!(phba->hba_flag & (FCF_TS_INPROG | FCF_RR_INPROG))) {
spin_unlock_irq(&phba->hbalock);
lpfc_printf_log(phba, KERN_INFO, LOG_FIP,
"2869 Devloss tmo to idle FIP engine, "
"unreg in-use FCF and rescan.\n");
/* Unregister in-use FCF and rescan */
lpfc_unregister_fcf_rescan(phba);
return;
}
spin_unlock_irq(&phba->hbalock);
if (phba->hba_flag & FCF_TS_INPROG)
lpfc_printf_log(phba, KERN_INFO, LOG_FIP,
"2870 FCF table scan in progress\n");
if (phba->hba_flag & FCF_RR_INPROG)
lpfc_printf_log(phba, KERN_INFO, LOG_FIP,
"2871 FLOGI roundrobin FCF failover "
"in progress\n");
}
lpfc_unregister_unused_fcf(phba);
}
/**
* lpfc_alloc_fast_evt - Allocates data structure for posting event
* @phba: Pointer to hba context object.
*
* This function is called from the functions which need to post
* events from interrupt context. This function allocates data
* structure required for posting event. It also keeps track of
* number of events pending and prevent event storm when there are
* too many events.
**/
struct lpfc_fast_path_event *
lpfc_alloc_fast_evt(struct lpfc_hba *phba) {
struct lpfc_fast_path_event *ret;
/* If there are lot of fast event do not exhaust memory due to this */
if (atomic_read(&phba->fast_event_count) > LPFC_MAX_EVT_COUNT)
return NULL;
ret = kzalloc(sizeof(struct lpfc_fast_path_event),
GFP_ATOMIC);
if (ret) {
atomic_inc(&phba->fast_event_count);
INIT_LIST_HEAD(&ret->work_evt.evt_listp);
ret->work_evt.evt = LPFC_EVT_FASTPATH_MGMT_EVT;
}
return ret;
}
/**
* lpfc_free_fast_evt - Frees event data structure
* @phba: Pointer to hba context object.
* @evt: Event object which need to be freed.
*
* This function frees the data structure required for posting
* events.
**/
void
lpfc_free_fast_evt(struct lpfc_hba *phba,
struct lpfc_fast_path_event *evt) {
atomic_dec(&phba->fast_event_count);
kfree(evt);
}
/**
* lpfc_send_fastpath_evt - Posts events generated from fast path
* @phba: Pointer to hba context object.
* @evtp: Event data structure.
*
* This function is called from worker thread, when the interrupt
* context need to post an event. This function posts the event
* to fc transport netlink interface.
**/
static void
lpfc_send_fastpath_evt(struct lpfc_hba *phba,
struct lpfc_work_evt *evtp)
{
unsigned long evt_category, evt_sub_category;
struct lpfc_fast_path_event *fast_evt_data;
char *evt_data;
uint32_t evt_data_size;
struct Scsi_Host *shost;
fast_evt_data = container_of(evtp, struct lpfc_fast_path_event,
work_evt);
evt_category = (unsigned long) fast_evt_data->un.fabric_evt.event_type;
evt_sub_category = (unsigned long) fast_evt_data->un.
fabric_evt.subcategory;
shost = lpfc_shost_from_vport(fast_evt_data->vport);
if (evt_category == FC_REG_FABRIC_EVENT) {
if (evt_sub_category == LPFC_EVENT_FCPRDCHKERR) {
evt_data = (char *) &fast_evt_data->un.read_check_error;
evt_data_size = sizeof(fast_evt_data->un.
read_check_error);
} else if ((evt_sub_category == LPFC_EVENT_FABRIC_BUSY) ||
(evt_sub_category == LPFC_EVENT_PORT_BUSY)) {
evt_data = (char *) &fast_evt_data->un.fabric_evt;
evt_data_size = sizeof(fast_evt_data->un.fabric_evt);
} else {
lpfc_free_fast_evt(phba, fast_evt_data);
return;
}
} else if (evt_category == FC_REG_SCSI_EVENT) {
switch (evt_sub_category) {
case LPFC_EVENT_QFULL:
case LPFC_EVENT_DEVBSY:
evt_data = (char *) &fast_evt_data->un.scsi_evt;
evt_data_size = sizeof(fast_evt_data->un.scsi_evt);
break;
case LPFC_EVENT_CHECK_COND:
evt_data = (char *) &fast_evt_data->un.check_cond_evt;
evt_data_size = sizeof(fast_evt_data->un.
check_cond_evt);
break;
case LPFC_EVENT_VARQUEDEPTH:
evt_data = (char *) &fast_evt_data->un.queue_depth_evt;
evt_data_size = sizeof(fast_evt_data->un.
queue_depth_evt);
break;
default:
lpfc_free_fast_evt(phba, fast_evt_data);
return;
}
} else {
lpfc_free_fast_evt(phba, fast_evt_data);
return;
}
scsi: lpfc: NVME Initiator: Base modifications NVME Initiator: Base modifications This patch adds base modifications for NVME initiator support. The base modifications consist of: - Formal split of SLI3 rings from SLI-4 WQs (sometimes referred to as rings as well) as implementation now widely varies between the two. - Addition of configuration modes: SCSI initiator only; NVME initiator only; NVME target only; and SCSI and NVME initiator. The configuration mode drives overall adapter configuration, offloads enabled, and resource splits. NVME support is only available on SLI-4 devices and newer fw. - Implements the following based on configuration mode: - Exchange resources are split by protocol; Obviously, if only 1 mode, then no split occurs. Default is 50/50. module attribute allows tuning. - Pools and config parameters are separated per-protocol - Each protocol has it's own set of queues, but share interrupt vectors. SCSI: SLI3 devices have few queues and the original style of queue allocation remains. SLI4 devices piggy back on an "io-channel" concept that eventually needs to merge with scsi-mq/blk-mq support (it is underway). For now, the paradigm continues as it existed prior. io channel allocates N msix and N WQs (N=4 default) and either round robins or uses cpu # modulo N for scheduling. A bunch of module parameters allow the configuration to be tuned. NVME (initiator): Allocates an msix per cpu (or whatever pci_alloc_irq_vectors gets) Allocates a WQ per cpu, and maps the WQs to msix on a WQ # modulo msix vector count basis. Module parameters exist to cap/control the config if desired. - Each protocol has its own buffer and dma pools. I apologize for the size of the patch. Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <james.smart@broadcom.com> ---- Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2017-02-12 21:52:30 +00:00
if (phba->cfg_enable_fc4_type != LPFC_ENABLE_NVME)
fc_host_post_vendor_event(shost,
fc_get_event_number(),
evt_data_size,
evt_data,
LPFC_NL_VENDOR_ID);
lpfc_free_fast_evt(phba, fast_evt_data);
return;
}
static void
lpfc_work_list_done(struct lpfc_hba *phba)
{
struct lpfc_work_evt *evtp = NULL;
struct lpfc_nodelist *ndlp;
int free_evt;
int fcf_inuse;
uint32_t nlp_did;
scsi: lpfc: Improve PCI EEH Error and Recovery Handling Following EEH errors, the driver can crash or hang when deleting the localport or when attempting to unload. The EEH handlers in the driver did not notify the NVMe-FC transport before tearing the driver down. This was delayed until the resume steps. This worked for SCSI because lpfc_block_scsi() would notify the scsi_fc_transport that the target was not available but it would not clean up all the references to the ndlp. The SLI3 prep for dev reset handler did the lpfc_offline_prep() and lpfc_offline() calls to get the port stopped before restarting. The SLI4 version of the prep for dev reset just destroyed the queues and did not stop NVMe from continuing. Also because the port was not really stopped the localport destroy would hang because the transport was still waiting for I/O. Additionally, a devloss tmo can fire and post events to a stopped worker thread creating another hang condition. lpfc_sli4_prep_dev_for_reset() is modified to call lpfc_offline_prep() and lpfc_offline() rather than just lpfc_scsi_dev_block() to ensure both SCSI and NVMe transports are notified to block I/O to the driver. Logic is added to devloss handler and worker thread to clean up ndlp references and quiesce appropriately. Link: https://lore.kernel.org/r/20220317032737.45308-2-jsmart2021@gmail.com Co-developed-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2022-03-17 03:27:34 +00:00
bool hba_pci_err;
spin_lock_irq(&phba->hbalock);
while (!list_empty(&phba->work_list)) {
list_remove_head((&phba->work_list), evtp, typeof(*evtp),
evt_listp);
spin_unlock_irq(&phba->hbalock);
scsi: lpfc: Improve PCI EEH Error and Recovery Handling Following EEH errors, the driver can crash or hang when deleting the localport or when attempting to unload. The EEH handlers in the driver did not notify the NVMe-FC transport before tearing the driver down. This was delayed until the resume steps. This worked for SCSI because lpfc_block_scsi() would notify the scsi_fc_transport that the target was not available but it would not clean up all the references to the ndlp. The SLI3 prep for dev reset handler did the lpfc_offline_prep() and lpfc_offline() calls to get the port stopped before restarting. The SLI4 version of the prep for dev reset just destroyed the queues and did not stop NVMe from continuing. Also because the port was not really stopped the localport destroy would hang because the transport was still waiting for I/O. Additionally, a devloss tmo can fire and post events to a stopped worker thread creating another hang condition. lpfc_sli4_prep_dev_for_reset() is modified to call lpfc_offline_prep() and lpfc_offline() rather than just lpfc_scsi_dev_block() to ensure both SCSI and NVMe transports are notified to block I/O to the driver. Logic is added to devloss handler and worker thread to clean up ndlp references and quiesce appropriately. Link: https://lore.kernel.org/r/20220317032737.45308-2-jsmart2021@gmail.com Co-developed-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2022-03-17 03:27:34 +00:00
hba_pci_err = test_bit(HBA_PCI_ERR, &phba->bit_flags);
free_evt = 1;
switch (evtp->evt) {
case LPFC_EVT_ELS_RETRY:
ndlp = (struct lpfc_nodelist *) (evtp->evt_arg1);
scsi: lpfc: Improve PCI EEH Error and Recovery Handling Following EEH errors, the driver can crash or hang when deleting the localport or when attempting to unload. The EEH handlers in the driver did not notify the NVMe-FC transport before tearing the driver down. This was delayed until the resume steps. This worked for SCSI because lpfc_block_scsi() would notify the scsi_fc_transport that the target was not available but it would not clean up all the references to the ndlp. The SLI3 prep for dev reset handler did the lpfc_offline_prep() and lpfc_offline() calls to get the port stopped before restarting. The SLI4 version of the prep for dev reset just destroyed the queues and did not stop NVMe from continuing. Also because the port was not really stopped the localport destroy would hang because the transport was still waiting for I/O. Additionally, a devloss tmo can fire and post events to a stopped worker thread creating another hang condition. lpfc_sli4_prep_dev_for_reset() is modified to call lpfc_offline_prep() and lpfc_offline() rather than just lpfc_scsi_dev_block() to ensure both SCSI and NVMe transports are notified to block I/O to the driver. Logic is added to devloss handler and worker thread to clean up ndlp references and quiesce appropriately. Link: https://lore.kernel.org/r/20220317032737.45308-2-jsmart2021@gmail.com Co-developed-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2022-03-17 03:27:34 +00:00
if (!hba_pci_err) {
lpfc_els_retry_delay_handler(ndlp);
free_evt = 0; /* evt is part of ndlp */
}
/* decrement the node reference count held
* for this queued work
*/
lpfc_nlp_put(ndlp);
break;
case LPFC_EVT_DEV_LOSS:
ndlp = (struct lpfc_nodelist *)(evtp->evt_arg1);
fcf_inuse = lpfc_dev_loss_tmo_handler(ndlp);
free_evt = 0;
/* decrement the node reference count held for
* this queued work
*/
nlp_did = ndlp->nlp_DID;
lpfc_nlp_put(ndlp);
if (phba->sli_rev == LPFC_SLI_REV4)
lpfc_sli4_post_dev_loss_tmo_handler(phba,
fcf_inuse,
nlp_did);
break;
scsi: lpfc: Fix scheduling call while in softirq context in lpfc_unreg_rpi The following call trace was seen during HBA reset testing: BUG: scheduling while atomic: swapper/2/0/0x10000100 ... Call Trace: dump_stack+0x19/0x1b __schedule_bug+0x64/0x72 __schedule+0x782/0x840 __cond_resched+0x26/0x30 _cond_resched+0x3a/0x50 mempool_alloc+0xa0/0x170 lpfc_unreg_rpi+0x151/0x630 [lpfc] lpfc_sli_abts_recover_port+0x171/0x190 [lpfc] lpfc_sli4_abts_err_handler+0xb2/0x1f0 [lpfc] lpfc_sli4_io_xri_aborted+0x256/0x300 [lpfc] lpfc_sli4_sp_handle_abort_xri_wcqe.isra.51+0xa3/0x190 [lpfc] lpfc_sli4_fp_handle_cqe+0x89/0x4d0 [lpfc] __lpfc_sli4_process_cq+0xdb/0x2e0 [lpfc] __lpfc_sli4_hba_process_cq+0x41/0x100 [lpfc] lpfc_cq_poll_hdler+0x1a/0x30 [lpfc] irq_poll_softirq+0xc7/0x100 __do_softirq+0xf5/0x280 call_softirq+0x1c/0x30 do_softirq+0x65/0xa0 irq_exit+0x105/0x110 do_IRQ+0x56/0xf0 common_interrupt+0x16a/0x16a With the conversion to blk_io_poll for better interrupt latency in normal cases, it introduced this code path, executed when I/O aborts or logouts are seen, which attempts to allocate memory for a mailbox command to be issued. The allocation is GFP_KERNEL, thus it could attempt to sleep. Fix by creating a work element that performs the event handling for the remote port. This will have the mailbox commands and other items performed in the work element, not the irq. A much better method as the "irq" routine does not stall while performing all this deep handling code. Ensure that allocation failures are handled and send LOGO on failure. Additionally, enlarge the mailbox memory pool to reduce the possibility of additional allocation in this path. Link: https://lore.kernel.org/r/20201020202719.54726-3-james.smart@broadcom.com Fixes: 317aeb83c92b ("scsi: lpfc: Add blk_io_poll support for latency improvment") Cc: <stable@vger.kernel.org> # v5.9+ Co-developed-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <james.smart@broadcom.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-10-20 20:27:12 +00:00
case LPFC_EVT_RECOVER_PORT:
ndlp = (struct lpfc_nodelist *)(evtp->evt_arg1);
scsi: lpfc: Improve PCI EEH Error and Recovery Handling Following EEH errors, the driver can crash or hang when deleting the localport or when attempting to unload. The EEH handlers in the driver did not notify the NVMe-FC transport before tearing the driver down. This was delayed until the resume steps. This worked for SCSI because lpfc_block_scsi() would notify the scsi_fc_transport that the target was not available but it would not clean up all the references to the ndlp. The SLI3 prep for dev reset handler did the lpfc_offline_prep() and lpfc_offline() calls to get the port stopped before restarting. The SLI4 version of the prep for dev reset just destroyed the queues and did not stop NVMe from continuing. Also because the port was not really stopped the localport destroy would hang because the transport was still waiting for I/O. Additionally, a devloss tmo can fire and post events to a stopped worker thread creating another hang condition. lpfc_sli4_prep_dev_for_reset() is modified to call lpfc_offline_prep() and lpfc_offline() rather than just lpfc_scsi_dev_block() to ensure both SCSI and NVMe transports are notified to block I/O to the driver. Logic is added to devloss handler and worker thread to clean up ndlp references and quiesce appropriately. Link: https://lore.kernel.org/r/20220317032737.45308-2-jsmart2021@gmail.com Co-developed-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2022-03-17 03:27:34 +00:00
if (!hba_pci_err) {
lpfc_sli_abts_recover_port(ndlp->vport, ndlp);
free_evt = 0;
}
scsi: lpfc: Fix scheduling call while in softirq context in lpfc_unreg_rpi The following call trace was seen during HBA reset testing: BUG: scheduling while atomic: swapper/2/0/0x10000100 ... Call Trace: dump_stack+0x19/0x1b __schedule_bug+0x64/0x72 __schedule+0x782/0x840 __cond_resched+0x26/0x30 _cond_resched+0x3a/0x50 mempool_alloc+0xa0/0x170 lpfc_unreg_rpi+0x151/0x630 [lpfc] lpfc_sli_abts_recover_port+0x171/0x190 [lpfc] lpfc_sli4_abts_err_handler+0xb2/0x1f0 [lpfc] lpfc_sli4_io_xri_aborted+0x256/0x300 [lpfc] lpfc_sli4_sp_handle_abort_xri_wcqe.isra.51+0xa3/0x190 [lpfc] lpfc_sli4_fp_handle_cqe+0x89/0x4d0 [lpfc] __lpfc_sli4_process_cq+0xdb/0x2e0 [lpfc] __lpfc_sli4_hba_process_cq+0x41/0x100 [lpfc] lpfc_cq_poll_hdler+0x1a/0x30 [lpfc] irq_poll_softirq+0xc7/0x100 __do_softirq+0xf5/0x280 call_softirq+0x1c/0x30 do_softirq+0x65/0xa0 irq_exit+0x105/0x110 do_IRQ+0x56/0xf0 common_interrupt+0x16a/0x16a With the conversion to blk_io_poll for better interrupt latency in normal cases, it introduced this code path, executed when I/O aborts or logouts are seen, which attempts to allocate memory for a mailbox command to be issued. The allocation is GFP_KERNEL, thus it could attempt to sleep. Fix by creating a work element that performs the event handling for the remote port. This will have the mailbox commands and other items performed in the work element, not the irq. A much better method as the "irq" routine does not stall while performing all this deep handling code. Ensure that allocation failures are handled and send LOGO on failure. Additionally, enlarge the mailbox memory pool to reduce the possibility of additional allocation in this path. Link: https://lore.kernel.org/r/20201020202719.54726-3-james.smart@broadcom.com Fixes: 317aeb83c92b ("scsi: lpfc: Add blk_io_poll support for latency improvment") Cc: <stable@vger.kernel.org> # v5.9+ Co-developed-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <james.smart@broadcom.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-10-20 20:27:12 +00:00
/* decrement the node reference count held for
* this queued work
*/
lpfc_nlp_put(ndlp);
break;
case LPFC_EVT_ONLINE:
if (phba->link_state < LPFC_LINK_DOWN)
*(int *) (evtp->evt_arg1) = lpfc_online(phba);
else
*(int *) (evtp->evt_arg1) = 0;
complete((struct completion *)(evtp->evt_arg2));
break;
case LPFC_EVT_OFFLINE_PREP:
if (phba->link_state >= LPFC_LINK_DOWN)
lpfc_offline_prep(phba, LPFC_MBX_WAIT);
*(int *)(evtp->evt_arg1) = 0;
complete((struct completion *)(evtp->evt_arg2));
break;
case LPFC_EVT_OFFLINE:
lpfc_offline(phba);
lpfc_sli_brdrestart(phba);
*(int *)(evtp->evt_arg1) =
lpfc_sli_brdready(phba, HS_FFRDY | HS_MBRDY);
lpfc_unblock_mgmt_io(phba);
complete((struct completion *)(evtp->evt_arg2));
break;
case LPFC_EVT_WARM_START:
lpfc_offline(phba);
lpfc_reset_barrier(phba);
lpfc_sli_brdreset(phba);
lpfc_hba_down_post(phba);
*(int *)(evtp->evt_arg1) =
lpfc_sli_brdready(phba, HS_MBRDY);
lpfc_unblock_mgmt_io(phba);
complete((struct completion *)(evtp->evt_arg2));
break;
case LPFC_EVT_KILL:
lpfc_offline(phba);
*(int *)(evtp->evt_arg1)
= (phba->pport->stopped)
? 0 : lpfc_sli_brdkill(phba);
lpfc_unblock_mgmt_io(phba);
complete((struct completion *)(evtp->evt_arg2));
break;
case LPFC_EVT_FASTPATH_MGMT_EVT:
lpfc_send_fastpath_evt(phba, evtp);
free_evt = 0;
break;
case LPFC_EVT_RESET_HBA:
if (!(phba->pport->load_flag & FC_UNLOADING))
lpfc_reset_hba(phba);
break;
}
if (free_evt)
kfree(evtp);
spin_lock_irq(&phba->hbalock);
}
spin_unlock_irq(&phba->hbalock);
}
static void
lpfc_work_done(struct lpfc_hba *phba)
{
struct lpfc_sli_ring *pring;
uint32_t ha_copy, status, control, work_port_events;
struct lpfc_vport **vports;
struct lpfc_vport *vport;
int i;
scsi: lpfc: Improve PCI EEH Error and Recovery Handling Following EEH errors, the driver can crash or hang when deleting the localport or when attempting to unload. The EEH handlers in the driver did not notify the NVMe-FC transport before tearing the driver down. This was delayed until the resume steps. This worked for SCSI because lpfc_block_scsi() would notify the scsi_fc_transport that the target was not available but it would not clean up all the references to the ndlp. The SLI3 prep for dev reset handler did the lpfc_offline_prep() and lpfc_offline() calls to get the port stopped before restarting. The SLI4 version of the prep for dev reset just destroyed the queues and did not stop NVMe from continuing. Also because the port was not really stopped the localport destroy would hang because the transport was still waiting for I/O. Additionally, a devloss tmo can fire and post events to a stopped worker thread creating another hang condition. lpfc_sli4_prep_dev_for_reset() is modified to call lpfc_offline_prep() and lpfc_offline() rather than just lpfc_scsi_dev_block() to ensure both SCSI and NVMe transports are notified to block I/O to the driver. Logic is added to devloss handler and worker thread to clean up ndlp references and quiesce appropriately. Link: https://lore.kernel.org/r/20220317032737.45308-2-jsmart2021@gmail.com Co-developed-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2022-03-17 03:27:34 +00:00
bool hba_pci_err;
scsi: lpfc: Improve PCI EEH Error and Recovery Handling Following EEH errors, the driver can crash or hang when deleting the localport or when attempting to unload. The EEH handlers in the driver did not notify the NVMe-FC transport before tearing the driver down. This was delayed until the resume steps. This worked for SCSI because lpfc_block_scsi() would notify the scsi_fc_transport that the target was not available but it would not clean up all the references to the ndlp. The SLI3 prep for dev reset handler did the lpfc_offline_prep() and lpfc_offline() calls to get the port stopped before restarting. The SLI4 version of the prep for dev reset just destroyed the queues and did not stop NVMe from continuing. Also because the port was not really stopped the localport destroy would hang because the transport was still waiting for I/O. Additionally, a devloss tmo can fire and post events to a stopped worker thread creating another hang condition. lpfc_sli4_prep_dev_for_reset() is modified to call lpfc_offline_prep() and lpfc_offline() rather than just lpfc_scsi_dev_block() to ensure both SCSI and NVMe transports are notified to block I/O to the driver. Logic is added to devloss handler and worker thread to clean up ndlp references and quiesce appropriately. Link: https://lore.kernel.org/r/20220317032737.45308-2-jsmart2021@gmail.com Co-developed-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2022-03-17 03:27:34 +00:00
hba_pci_err = test_bit(HBA_PCI_ERR, &phba->bit_flags);
spin_lock_irq(&phba->hbalock);
ha_copy = phba->work_ha;
phba->work_ha = 0;
spin_unlock_irq(&phba->hbalock);
scsi: lpfc: Improve PCI EEH Error and Recovery Handling Following EEH errors, the driver can crash or hang when deleting the localport or when attempting to unload. The EEH handlers in the driver did not notify the NVMe-FC transport before tearing the driver down. This was delayed until the resume steps. This worked for SCSI because lpfc_block_scsi() would notify the scsi_fc_transport that the target was not available but it would not clean up all the references to the ndlp. The SLI3 prep for dev reset handler did the lpfc_offline_prep() and lpfc_offline() calls to get the port stopped before restarting. The SLI4 version of the prep for dev reset just destroyed the queues and did not stop NVMe from continuing. Also because the port was not really stopped the localport destroy would hang because the transport was still waiting for I/O. Additionally, a devloss tmo can fire and post events to a stopped worker thread creating another hang condition. lpfc_sli4_prep_dev_for_reset() is modified to call lpfc_offline_prep() and lpfc_offline() rather than just lpfc_scsi_dev_block() to ensure both SCSI and NVMe transports are notified to block I/O to the driver. Logic is added to devloss handler and worker thread to clean up ndlp references and quiesce appropriately. Link: https://lore.kernel.org/r/20220317032737.45308-2-jsmart2021@gmail.com Co-developed-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2022-03-17 03:27:34 +00:00
if (hba_pci_err)
ha_copy = 0;
/* First, try to post the next mailbox command to SLI4 device */
scsi: lpfc: Improve PCI EEH Error and Recovery Handling Following EEH errors, the driver can crash or hang when deleting the localport or when attempting to unload. The EEH handlers in the driver did not notify the NVMe-FC transport before tearing the driver down. This was delayed until the resume steps. This worked for SCSI because lpfc_block_scsi() would notify the scsi_fc_transport that the target was not available but it would not clean up all the references to the ndlp. The SLI3 prep for dev reset handler did the lpfc_offline_prep() and lpfc_offline() calls to get the port stopped before restarting. The SLI4 version of the prep for dev reset just destroyed the queues and did not stop NVMe from continuing. Also because the port was not really stopped the localport destroy would hang because the transport was still waiting for I/O. Additionally, a devloss tmo can fire and post events to a stopped worker thread creating another hang condition. lpfc_sli4_prep_dev_for_reset() is modified to call lpfc_offline_prep() and lpfc_offline() rather than just lpfc_scsi_dev_block() to ensure both SCSI and NVMe transports are notified to block I/O to the driver. Logic is added to devloss handler and worker thread to clean up ndlp references and quiesce appropriately. Link: https://lore.kernel.org/r/20220317032737.45308-2-jsmart2021@gmail.com Co-developed-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2022-03-17 03:27:34 +00:00
if (phba->pci_dev_grp == LPFC_PCI_DEV_OC && !hba_pci_err)
lpfc_sli4_post_async_mbox(phba);
if (ha_copy & HA_ERATT) {
/* Handle the error attention event */
lpfc_handle_eratt(phba);
if (phba->fw_dump_cmpl) {
complete(phba->fw_dump_cmpl);
phba->fw_dump_cmpl = NULL;
}
}
if (ha_copy & HA_MBATT)
lpfc_sli_handle_mb_event(phba);
if (ha_copy & HA_LATT)
lpfc_handle_latt(phba);
/* Handle VMID Events */
scsi: lpfc: Improve PCI EEH Error and Recovery Handling Following EEH errors, the driver can crash or hang when deleting the localport or when attempting to unload. The EEH handlers in the driver did not notify the NVMe-FC transport before tearing the driver down. This was delayed until the resume steps. This worked for SCSI because lpfc_block_scsi() would notify the scsi_fc_transport that the target was not available but it would not clean up all the references to the ndlp. The SLI3 prep for dev reset handler did the lpfc_offline_prep() and lpfc_offline() calls to get the port stopped before restarting. The SLI4 version of the prep for dev reset just destroyed the queues and did not stop NVMe from continuing. Also because the port was not really stopped the localport destroy would hang because the transport was still waiting for I/O. Additionally, a devloss tmo can fire and post events to a stopped worker thread creating another hang condition. lpfc_sli4_prep_dev_for_reset() is modified to call lpfc_offline_prep() and lpfc_offline() rather than just lpfc_scsi_dev_block() to ensure both SCSI and NVMe transports are notified to block I/O to the driver. Logic is added to devloss handler and worker thread to clean up ndlp references and quiesce appropriately. Link: https://lore.kernel.org/r/20220317032737.45308-2-jsmart2021@gmail.com Co-developed-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2022-03-17 03:27:34 +00:00
if (lpfc_is_vmid_enabled(phba) && !hba_pci_err) {
if (phba->pport->work_port_events &
WORKER_CHECK_VMID_ISSUE_QFPA) {
lpfc_check_vmid_qfpa_issue(phba);
phba->pport->work_port_events &=
~WORKER_CHECK_VMID_ISSUE_QFPA;
}
if (phba->pport->work_port_events &
WORKER_CHECK_INACTIVE_VMID) {
lpfc_check_inactive_vmid(phba);
phba->pport->work_port_events &=
~WORKER_CHECK_INACTIVE_VMID;
}
}
/* Process SLI4 events */
if (phba->pci_dev_grp == LPFC_PCI_DEV_OC) {
if (phba->hba_flag & HBA_RRQ_ACTIVE)
lpfc_handle_rrq_active(phba);
if (phba->hba_flag & ELS_XRI_ABORT_EVENT)
lpfc_sli4_els_xri_abort_event_proc(phba);
if (phba->hba_flag & ASYNC_EVENT)
lpfc_sli4_async_event_proc(phba);
if (phba->hba_flag & HBA_POST_RECEIVE_BUFFER) {
spin_lock_irq(&phba->hbalock);
phba->hba_flag &= ~HBA_POST_RECEIVE_BUFFER;
spin_unlock_irq(&phba->hbalock);
lpfc_sli_hbqbuf_add_hbqs(phba, LPFC_ELS_HBQ);
}
if (phba->fcf.fcf_flag & FCF_REDISC_EVT)
lpfc_sli4_fcf_redisc_event_proc(phba);
}
vports = lpfc_create_vport_work_array(phba);
if (vports != NULL)
for (i = 0; i <= phba->max_vports; i++) {
/*
* We could have no vports in array if unloading, so if
* this happens then just use the pport
*/
if (vports[i] == NULL && i == 0)
vport = phba->pport;
else
vport = vports[i];
if (vport == NULL)
break;
[SCSI] lpfc 8.2.6 : Multiple discovery fixes Multiple Discovery Fixes: - Fix race on discovery due to link events coinciding with vport_delete. - Use NLP_FABRIC state to filter out switch-based pseudo initiators that reuse the same WWNs. - Correct erroneous setting of DID=0 in lpfc_matchdid() - Correct extra reference count that was in the lookup path for the remoteid from an unsolicited ELS. - Correct double-free bug in els abort path. - Correct FDMI server discovery logic for switch that return a WWN of 0. - Fix bugs in ndlp mgmt when a node changes address - Correct bug that did not delete RSCNs for vports upon link transitions - Fix "0216 Link event during NS query" error which pops up when vports are swapped to different switch ports. - Add sanity checks on ndlp structures - Fix devloss log message to dump WWN correctly - Hold off mgmt commands that were interferring with discovery mailbox cmds - Remove unnecessary FC_ESTABLISH_LINK logic. - Correct some race conditions in the worker thread, resulting in devloss: - Clear the work_port_events field before handling the work port events - Clear the deferred ring event before handling a deferred ring event - Hold the hba lock when waking up the work thread - Send an acc for the rscn even when we aren't going to handle it - Fix locking behavior that was not properly protecting the ACTIVE flag, thus allowing mailbox command order to shift. Signed-off-by: James Smart <james.smart@emulex.com> Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
2008-04-07 14:15:56 +00:00
spin_lock_irq(&vport->work_port_lock);
work_port_events = vport->work_port_events;
[SCSI] lpfc 8.2.6 : Multiple discovery fixes Multiple Discovery Fixes: - Fix race on discovery due to link events coinciding with vport_delete. - Use NLP_FABRIC state to filter out switch-based pseudo initiators that reuse the same WWNs. - Correct erroneous setting of DID=0 in lpfc_matchdid() - Correct extra reference count that was in the lookup path for the remoteid from an unsolicited ELS. - Correct double-free bug in els abort path. - Correct FDMI server discovery logic for switch that return a WWN of 0. - Fix bugs in ndlp mgmt when a node changes address - Correct bug that did not delete RSCNs for vports upon link transitions - Fix "0216 Link event during NS query" error which pops up when vports are swapped to different switch ports. - Add sanity checks on ndlp structures - Fix devloss log message to dump WWN correctly - Hold off mgmt commands that were interferring with discovery mailbox cmds - Remove unnecessary FC_ESTABLISH_LINK logic. - Correct some race conditions in the worker thread, resulting in devloss: - Clear the work_port_events field before handling the work port events - Clear the deferred ring event before handling a deferred ring event - Hold the hba lock when waking up the work thread - Send an acc for the rscn even when we aren't going to handle it - Fix locking behavior that was not properly protecting the ACTIVE flag, thus allowing mailbox command order to shift. Signed-off-by: James Smart <james.smart@emulex.com> Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
2008-04-07 14:15:56 +00:00
vport->work_port_events &= ~work_port_events;
spin_unlock_irq(&vport->work_port_lock);
scsi: lpfc: Improve PCI EEH Error and Recovery Handling Following EEH errors, the driver can crash or hang when deleting the localport or when attempting to unload. The EEH handlers in the driver did not notify the NVMe-FC transport before tearing the driver down. This was delayed until the resume steps. This worked for SCSI because lpfc_block_scsi() would notify the scsi_fc_transport that the target was not available but it would not clean up all the references to the ndlp. The SLI3 prep for dev reset handler did the lpfc_offline_prep() and lpfc_offline() calls to get the port stopped before restarting. The SLI4 version of the prep for dev reset just destroyed the queues and did not stop NVMe from continuing. Also because the port was not really stopped the localport destroy would hang because the transport was still waiting for I/O. Additionally, a devloss tmo can fire and post events to a stopped worker thread creating another hang condition. lpfc_sli4_prep_dev_for_reset() is modified to call lpfc_offline_prep() and lpfc_offline() rather than just lpfc_scsi_dev_block() to ensure both SCSI and NVMe transports are notified to block I/O to the driver. Logic is added to devloss handler and worker thread to clean up ndlp references and quiesce appropriately. Link: https://lore.kernel.org/r/20220317032737.45308-2-jsmart2021@gmail.com Co-developed-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2022-03-17 03:27:34 +00:00
if (hba_pci_err)
continue;
if (work_port_events & WORKER_DISC_TMO)
lpfc_disc_timeout_handler(vport);
if (work_port_events & WORKER_ELS_TMO)
lpfc_els_timeout_handler(vport);
if (work_port_events & WORKER_HB_TMO)
lpfc_hb_timeout_handler(phba);
if (work_port_events & WORKER_MBOX_TMO)
lpfc_mbox_timeout_handler(phba);
if (work_port_events & WORKER_FABRIC_BLOCK_TMO)
lpfc_unblock_fabric_iocbs(phba);
if (work_port_events & WORKER_RAMP_DOWN_QUEUE)
lpfc_ramp_down_queue_handler(phba);
if (work_port_events & WORKER_DELAYED_DISC_TMO)
lpfc_delayed_disc_timeout_handler(vport);
}
[SCSI] lpfc 8.2.4 : Miscellaneous Fixes Miscellaneous Fixes: - Fix a couple of sparse complaints - Reset the FCP recovery flag when the node is not a FCP2 device. - Speed up offline prep delays - Fixed a memory leak in lpfc_mem_alloc failure path - Fixed external loopback test. - Fixed error code returned from the driver when HBA is over heated. - Correct Max NPIV vport to limits read from adapter - Add missing locks around fc_flag and FC_NEEDS_REG_VPI - Add missing hba ids for device identification - Added support for SET_VARIABLE and MBX_WRITE_WWN mailbox commands - Changed all temperature event messages from warning to error - Fix reporting of link speed when link is down - Added support for MBX_WRITE_WWN mailbox command - Change del_timer_sync() in ISR to del_timer() in interrupt handler - Correct instances of beXX_to_cpu() that should be cpu_to_beXX() - Perform target flush before releasing node references on module unload - Avoid bogus devloss_tmo messages when driver unloads - Fix panic when HBA generates ERATT interupt - Fix mbox race condition and a workaround on back-to-back mailbox commands - Force NPIV off for pt2pt mode between 2 NPorts - Stop worker thread before removing fc_host. - Fix up discovery timeout error case due to missing clear_la - Tighten mailbox polling code to speed up detection of fast completions - Only allow DUMP_MEMORY if adapter offline due to overtemp errors - Added extended error information to the log messages in chip init. Signed-off-by: James Smart <James.Smart@emulex.com> Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
2008-01-11 06:52:54 +00:00
lpfc_destroy_vport_work_array(phba, vports);
scsi: lpfc: NVME Initiator: Base modifications NVME Initiator: Base modifications This patch adds base modifications for NVME initiator support. The base modifications consist of: - Formal split of SLI3 rings from SLI-4 WQs (sometimes referred to as rings as well) as implementation now widely varies between the two. - Addition of configuration modes: SCSI initiator only; NVME initiator only; NVME target only; and SCSI and NVME initiator. The configuration mode drives overall adapter configuration, offloads enabled, and resource splits. NVME support is only available on SLI-4 devices and newer fw. - Implements the following based on configuration mode: - Exchange resources are split by protocol; Obviously, if only 1 mode, then no split occurs. Default is 50/50. module attribute allows tuning. - Pools and config parameters are separated per-protocol - Each protocol has it's own set of queues, but share interrupt vectors. SCSI: SLI3 devices have few queues and the original style of queue allocation remains. SLI4 devices piggy back on an "io-channel" concept that eventually needs to merge with scsi-mq/blk-mq support (it is underway). For now, the paradigm continues as it existed prior. io channel allocates N msix and N WQs (N=4 default) and either round robins or uses cpu # modulo N for scheduling. A bunch of module parameters allow the configuration to be tuned. NVME (initiator): Allocates an msix per cpu (or whatever pci_alloc_irq_vectors gets) Allocates a WQ per cpu, and maps the WQs to msix on a WQ # modulo msix vector count basis. Module parameters exist to cap/control the config if desired. - Each protocol has its own buffer and dma pools. I apologize for the size of the patch. Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <james.smart@broadcom.com> ---- Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2017-02-12 21:52:30 +00:00
pring = lpfc_phba_elsring(phba);
status = (ha_copy & (HA_RXMASK << (4*LPFC_ELS_RING)));
status >>= (4*LPFC_ELS_RING);
if (pring && (status & HA_RXMASK ||
pring->flag & LPFC_DEFERRED_RING_EVENT ||
phba->hba_flag & HBA_SP_QUEUE_EVT)) {
if (pring->flag & LPFC_STOP_IOCB_EVENT) {
pring->flag |= LPFC_DEFERRED_RING_EVENT;
/* Preserve legacy behavior. */
if (!(phba->hba_flag & HBA_SP_QUEUE_EVT))
set_bit(LPFC_DATA_READY, &phba->data_flags);
} else {
scsi: lpfc: Fix list corruption in lpfc_sli_get_iocbq After study, it was determined there was a double free of a CT iocb during execution of lpfc_offline_prep and lpfc_offline. The prep routine issued an abort for some CT iocbs, but the aborts did not complete fast enough for a subsequent routine that waits for completion. Thus the driver proceeded to lpfc_offline, which releases any pending iocbs. Unfortunately, the completions for the aborts were then received which re-released the ct iocbs. Turns out the issue for why the aborts didn't complete fast enough was not their time on the wire/in the adapter. It was the lpfc_work_done routine, which requires the adapter state to be UP before it calls lpfc_sli_handle_slow_ring_event() to process the completions. The issue is the prep routine takes the link down as part of it's processing. To fix, the following was performed: - Prevent the offline routine from releasing iocbs that have had aborts issued on them. Defer to the abort completions. Also means the driver fully waits for the completions. Given this change, the recognition of "driver-generated" status which then releases the iocb is no longer valid. As such, the change made in the commit 296012285c90 is reverted. As recognition of "driver-generated" status is no longer valid, this patch reverts the changes made in commit 296012285c90 ("scsi: lpfc: Fix leak of ELS completions on adapter reset") - Modify lpfc_work_done to allow slow path completions so that the abort completions aren't ignored. - Updated the fdmi path to recognize a CT request that fails due to the port being unusable. This stops FDMI retries. FDMI will be restarted on next link up. Link: https://lore.kernel.org/r/20190922035906.10977-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2019-09-22 03:58:59 +00:00
/* Driver could have abort request completed in queue
* when link goes down. Allow for this transition.
*/
if (phba->link_state >= LPFC_LINK_DOWN ||
phba->link_flag & LS_MDS_LOOPBACK) {
pring->flag &= ~LPFC_DEFERRED_RING_EVENT;
lpfc_sli_handle_slow_ring_event(phba, pring,
(status &
HA_RXMASK));
}
}
if (phba->sli_rev == LPFC_SLI_REV4)
lpfc_drain_txq(phba);
/*
* Turn on Ring interrupts
*/
if (phba->sli_rev <= LPFC_SLI_REV3) {
spin_lock_irq(&phba->hbalock);
control = readl(phba->HCregaddr);
if (!(control & (HC_R0INT_ENA << LPFC_ELS_RING))) {
lpfc_debugfs_slow_ring_trc(phba,
"WRK Enable ring: cntl:x%x hacopy:x%x",
control, ha_copy, 0);
control |= (HC_R0INT_ENA << LPFC_ELS_RING);
writel(control, phba->HCregaddr);
readl(phba->HCregaddr); /* flush */
} else {
lpfc_debugfs_slow_ring_trc(phba,
"WRK Ring ok: cntl:x%x hacopy:x%x",
control, ha_copy, 0);
}
spin_unlock_irq(&phba->hbalock);
}
}
lpfc_work_list_done(phba);
}
int
lpfc_do_work(void *p)
{
struct lpfc_hba *phba = p;
int rc;
set_user_nice(current, MIN_NICE);
current->flags |= PF_NOFREEZE;
phba->data_flags = 0;
while (!kthread_should_stop()) {
/* wait and check worker queue activities */
rc = wait_event_interruptible(phba->work_waitq,
(test_and_clear_bit(LPFC_DATA_READY,
&phba->data_flags)
|| kthread_should_stop()));
/* Signal wakeup shall terminate the worker thread */
if (rc) {
scsi: lpfc: Add an internal trace log buffer The current logging methods typically end up requesting a reproduction with a different logging level set to figure out what happened. This was mainly by design to not clutter the kernel log messages with things that were typically not interesting and the messages themselves could cause other issues. When looking to make a better system, it was seen that in many cases when more data was wanted was when another message, usually at KERN_ERR level, was logged. And in most cases, what the additional logging that was then enabled was typically. Most of these areas fell into the discovery machine. Based on this summary, the following design has been put in place: The driver will maintain an internal log (256 elements of 256 bytes). The "additional logging" messages that are usually enabled in a reproduction will be changed to now log all the time to the internal log. A new logging level is defined - LOG_TRACE_EVENT. When this level is set (it is not by default) and a message marked as KERN_ERR is logged, all the messages in the internal log will be dumped to the kernel log before the KERN_ERR message is logged. There is a timestamp on each message added to the internal log. However, this timestamp is not converted to wall time when logged. The value of the timestamp is solely to give a crude time reference for the messages. Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-06-30 21:50:00 +00:00
lpfc_printf_log(phba, KERN_ERR, LOG_TRACE_EVENT,
"0433 Wakeup on signal: rc=x%x\n", rc);
break;
}
/* Attend pending lpfc data processing */
lpfc_work_done(phba);
}
phba->worker_thread = NULL;
lpfc_printf_log(phba, KERN_INFO, LOG_ELS,
"0432 Worker thread stopped.\n");
return 0;
}
/*
* This is only called to handle FC worker events. Since this a rare
* occurrence, we allocate a struct lpfc_work_evt structure here instead of
* embedding it in the IOCB.
*/
int
lpfc_workq_post_event(struct lpfc_hba *phba, void *arg1, void *arg2,
uint32_t evt)
{
struct lpfc_work_evt *evtp;
unsigned long flags;
/*
* All Mailbox completions and LPFC_ELS_RING rcv ring IOCB events will
* be queued to worker thread for processing
*/
evtp = kmalloc(sizeof(struct lpfc_work_evt), GFP_ATOMIC);
if (!evtp)
return 0;
evtp->evt_arg1 = arg1;
evtp->evt_arg2 = arg2;
evtp->evt = evt;
spin_lock_irqsave(&phba->hbalock, flags);
list_add_tail(&evtp->evt_listp, &phba->work_list);
spin_unlock_irqrestore(&phba->hbalock, flags);
lpfc_worker_wake_up(phba);
return 1;
}
void
lpfc_cleanup_rpis(struct lpfc_vport *vport, int remove)
{
[SCSI] lpfc 8.2.4 : Miscellaneous Fixes Miscellaneous Fixes: - Fix a couple of sparse complaints - Reset the FCP recovery flag when the node is not a FCP2 device. - Speed up offline prep delays - Fixed a memory leak in lpfc_mem_alloc failure path - Fixed external loopback test. - Fixed error code returned from the driver when HBA is over heated. - Correct Max NPIV vport to limits read from adapter - Add missing locks around fc_flag and FC_NEEDS_REG_VPI - Add missing hba ids for device identification - Added support for SET_VARIABLE and MBX_WRITE_WWN mailbox commands - Changed all temperature event messages from warning to error - Fix reporting of link speed when link is down - Added support for MBX_WRITE_WWN mailbox command - Change del_timer_sync() in ISR to del_timer() in interrupt handler - Correct instances of beXX_to_cpu() that should be cpu_to_beXX() - Perform target flush before releasing node references on module unload - Avoid bogus devloss_tmo messages when driver unloads - Fix panic when HBA generates ERATT interupt - Fix mbox race condition and a workaround on back-to-back mailbox commands - Force NPIV off for pt2pt mode between 2 NPorts - Stop worker thread before removing fc_host. - Fix up discovery timeout error case due to missing clear_la - Tighten mailbox polling code to speed up detection of fast completions - Only allow DUMP_MEMORY if adapter offline due to overtemp errors - Added extended error information to the log messages in chip init. Signed-off-by: James Smart <James.Smart@emulex.com> Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
2008-01-11 06:52:54 +00:00
struct Scsi_Host *shost = lpfc_shost_from_vport(vport);
struct lpfc_hba *phba = vport->phba;
struct lpfc_nodelist *ndlp, *next_ndlp;
list_for_each_entry_safe(ndlp, next_ndlp, &vport->fc_nodes, nlp_listp) {
if ((phba->sli3_options & LPFC_SLI3_VPORT_TEARDOWN) ||
((vport->port_type == LPFC_NPIV_PORT) &&
((ndlp->nlp_DID == NameServer_DID) ||
(ndlp->nlp_DID == FDMI_DID) ||
(ndlp->nlp_DID == Fabric_Cntl_DID))))
lpfc_unreg_rpi(vport, ndlp);
/* Leave Fabric nodes alone on link down */
if ((phba->sli_rev < LPFC_SLI_REV4) &&
(!remove && ndlp->nlp_type & NLP_FABRIC))
continue;
/* Notify transport of connectivity loss to trigger cleanup. */
if (phba->nvmet_support &&
ndlp->nlp_state == NLP_STE_UNMAPPED_NODE)
lpfc_nvmet_invalidate_host(phba, ndlp);
lpfc_disc_state_machine(vport, ndlp, NULL,
remove
? NLP_EVT_DEVICE_RM
: NLP_EVT_DEVICE_RECOVERY);
}
if (phba->sli3_options & LPFC_SLI3_VPORT_TEARDOWN) {
if (phba->sli_rev == LPFC_SLI_REV4)
lpfc_sli4_unreg_all_rpis(vport);
lpfc_mbx_unreg_vpi(vport);
[SCSI] lpfc 8.2.4 : Miscellaneous Fixes Miscellaneous Fixes: - Fix a couple of sparse complaints - Reset the FCP recovery flag when the node is not a FCP2 device. - Speed up offline prep delays - Fixed a memory leak in lpfc_mem_alloc failure path - Fixed external loopback test. - Fixed error code returned from the driver when HBA is over heated. - Correct Max NPIV vport to limits read from adapter - Add missing locks around fc_flag and FC_NEEDS_REG_VPI - Add missing hba ids for device identification - Added support for SET_VARIABLE and MBX_WRITE_WWN mailbox commands - Changed all temperature event messages from warning to error - Fix reporting of link speed when link is down - Added support for MBX_WRITE_WWN mailbox command - Change del_timer_sync() in ISR to del_timer() in interrupt handler - Correct instances of beXX_to_cpu() that should be cpu_to_beXX() - Perform target flush before releasing node references on module unload - Avoid bogus devloss_tmo messages when driver unloads - Fix panic when HBA generates ERATT interupt - Fix mbox race condition and a workaround on back-to-back mailbox commands - Force NPIV off for pt2pt mode between 2 NPorts - Stop worker thread before removing fc_host. - Fix up discovery timeout error case due to missing clear_la - Tighten mailbox polling code to speed up detection of fast completions - Only allow DUMP_MEMORY if adapter offline due to overtemp errors - Added extended error information to the log messages in chip init. Signed-off-by: James Smart <James.Smart@emulex.com> Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
2008-01-11 06:52:54 +00:00
spin_lock_irq(shost->host_lock);
vport->fc_flag |= FC_VPORT_NEEDS_REG_VPI;
[SCSI] lpfc 8.2.4 : Miscellaneous Fixes Miscellaneous Fixes: - Fix a couple of sparse complaints - Reset the FCP recovery flag when the node is not a FCP2 device. - Speed up offline prep delays - Fixed a memory leak in lpfc_mem_alloc failure path - Fixed external loopback test. - Fixed error code returned from the driver when HBA is over heated. - Correct Max NPIV vport to limits read from adapter - Add missing locks around fc_flag and FC_NEEDS_REG_VPI - Add missing hba ids for device identification - Added support for SET_VARIABLE and MBX_WRITE_WWN mailbox commands - Changed all temperature event messages from warning to error - Fix reporting of link speed when link is down - Added support for MBX_WRITE_WWN mailbox command - Change del_timer_sync() in ISR to del_timer() in interrupt handler - Correct instances of beXX_to_cpu() that should be cpu_to_beXX() - Perform target flush before releasing node references on module unload - Avoid bogus devloss_tmo messages when driver unloads - Fix panic when HBA generates ERATT interupt - Fix mbox race condition and a workaround on back-to-back mailbox commands - Force NPIV off for pt2pt mode between 2 NPorts - Stop worker thread before removing fc_host. - Fix up discovery timeout error case due to missing clear_la - Tighten mailbox polling code to speed up detection of fast completions - Only allow DUMP_MEMORY if adapter offline due to overtemp errors - Added extended error information to the log messages in chip init. Signed-off-by: James Smart <James.Smart@emulex.com> Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
2008-01-11 06:52:54 +00:00
spin_unlock_irq(shost->host_lock);
}
}
void
lpfc_port_link_failure(struct lpfc_vport *vport)
{
lpfc_vport_set_state(vport, FC_VPORT_LINKDOWN);
/* Cleanup any outstanding received buffers */
lpfc_cleanup_rcv_buffers(vport);
/* Cleanup any outstanding RSCN activity */
lpfc_els_flush_rscn(vport);
/* Cleanup any outstanding ELS commands */
lpfc_els_flush_cmd(vport);
lpfc_cleanup_rpis(vport, 0);
/* Turn off discovery timer if its running */
lpfc_can_disctmo(vport);
}
void
lpfc_linkdown_port(struct lpfc_vport *vport)
{
struct lpfc_hba *phba = vport->phba;
struct Scsi_Host *shost = lpfc_shost_from_vport(vport);
if (vport->cfg_enable_fc4_type != LPFC_ENABLE_NVME)
fc_host_post_event(shost, fc_get_event_number(),
FCH_EVT_LINKDOWN, 0);
lpfc_debugfs_disc_trc(vport, LPFC_DISC_TRC_ELS_CMD,
"Link Down: state:x%x rtry:x%x flg:x%x",
vport->port_state, vport->fc_ns_retry, vport->fc_flag);
lpfc_port_link_failure(vport);
/* Stop delayed Nport discovery */
spin_lock_irq(shost->host_lock);
vport->fc_flag &= ~FC_DISC_DELAYED;
spin_unlock_irq(shost->host_lock);
del_timer_sync(&vport->delayed_disc_tmo);
if (phba->sli_rev == LPFC_SLI_REV4 &&
vport->port_type == LPFC_PHYSICAL_PORT &&
phba->sli4_hba.fawwpn_flag & LPFC_FAWWPN_CONFIG) {
/* Assume success on link up */
phba->sli4_hba.fawwpn_flag |= LPFC_FAWWPN_FABRIC;
}
}
int
lpfc_linkdown(struct lpfc_hba *phba)
{
struct lpfc_vport *vport = phba->pport;
struct Scsi_Host *shost = lpfc_shost_from_vport(vport);
struct lpfc_vport **vports;
LPFC_MBOXQ_t *mb;
int i;
scsi: lpfc: Improve PCI EEH Error and Recovery Handling Following EEH errors, the driver can crash or hang when deleting the localport or when attempting to unload. The EEH handlers in the driver did not notify the NVMe-FC transport before tearing the driver down. This was delayed until the resume steps. This worked for SCSI because lpfc_block_scsi() would notify the scsi_fc_transport that the target was not available but it would not clean up all the references to the ndlp. The SLI3 prep for dev reset handler did the lpfc_offline_prep() and lpfc_offline() calls to get the port stopped before restarting. The SLI4 version of the prep for dev reset just destroyed the queues and did not stop NVMe from continuing. Also because the port was not really stopped the localport destroy would hang because the transport was still waiting for I/O. Additionally, a devloss tmo can fire and post events to a stopped worker thread creating another hang condition. lpfc_sli4_prep_dev_for_reset() is modified to call lpfc_offline_prep() and lpfc_offline() rather than just lpfc_scsi_dev_block() to ensure both SCSI and NVMe transports are notified to block I/O to the driver. Logic is added to devloss handler and worker thread to clean up ndlp references and quiesce appropriately. Link: https://lore.kernel.org/r/20220317032737.45308-2-jsmart2021@gmail.com Co-developed-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2022-03-17 03:27:34 +00:00
int offline;
if (phba->link_state == LPFC_LINK_DOWN)
return 0;
/* Block all SCSI stack I/Os */
lpfc_scsi_dev_block(phba);
scsi: lpfc: Improve PCI EEH Error and Recovery Handling Following EEH errors, the driver can crash or hang when deleting the localport or when attempting to unload. The EEH handlers in the driver did not notify the NVMe-FC transport before tearing the driver down. This was delayed until the resume steps. This worked for SCSI because lpfc_block_scsi() would notify the scsi_fc_transport that the target was not available but it would not clean up all the references to the ndlp. The SLI3 prep for dev reset handler did the lpfc_offline_prep() and lpfc_offline() calls to get the port stopped before restarting. The SLI4 version of the prep for dev reset just destroyed the queues and did not stop NVMe from continuing. Also because the port was not really stopped the localport destroy would hang because the transport was still waiting for I/O. Additionally, a devloss tmo can fire and post events to a stopped worker thread creating another hang condition. lpfc_sli4_prep_dev_for_reset() is modified to call lpfc_offline_prep() and lpfc_offline() rather than just lpfc_scsi_dev_block() to ensure both SCSI and NVMe transports are notified to block I/O to the driver. Logic is added to devloss handler and worker thread to clean up ndlp references and quiesce appropriately. Link: https://lore.kernel.org/r/20220317032737.45308-2-jsmart2021@gmail.com Co-developed-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2022-03-17 03:27:34 +00:00
offline = pci_channel_offline(phba->pcidev);
scsi: lpfc: Defer LS_ACC to FLOGI on point to point logins The current discovery state machine the driver treated FLOGI oddly. When point to point, an FLOGI is to be exchanged by the two ports, with the port with the most significant WWN then proceeding with PLOGI. The implementation in the driver was keyed to closely with "what have I sent", not with what has happened between the two endpoints. Thus, it blatantly would ACC an FLOGI, but reject PLOGI's until it had its FLOGI ACC'd. The problem is - the sending of FLOGI may be delayed for some reason, or the response to FLOGI held off by the other side. In the failing situation the other side sent an FLOGI, which was ACC'd, then sent PLOGIs which were then rjt'd until the retry count for the PLOGIs were exceeded and the port gave up. The FLOGI may have been very late in transmit, or the response held off until the PLOGIs failed. Given the other port had the higher WWN, no PLOGIs would occur and communication stopped. Correct the situation by changing the FLOGI handling. Defer any response to an FLOGI until the driver has sent its FLOGI as well. Then, upon either completion of the sent FLOGI, or upon sending an ACC to a received FLOGI (which may be received before or just after FLOGI was sent). the driver will act on who has the higher WWN. if the other port does, the driver will noop any handling of an FLOGI response (if outstanding) and wait for PLOGI. If the local port does, the driver will transition to sending PLOGI and will noop any action on responding to an FLOGI (if not yet received). Fortunately, to implement this, it only took another state flag and deferring any FLOGI response if the FLOGI has yet to be transmit. All subsequent actions were already in place. Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2018-11-30 00:09:36 +00:00
phba->defer_flogi_acc_flag = false;
/* Clear external loopback plug detected flag */
phba->link_flag &= ~LS_EXTERNAL_LOOPBACK;
spin_lock_irq(&phba->hbalock);
phba->fcf.fcf_flag &= ~(FCF_AVAILABLE | FCF_SCAN_DONE);
spin_unlock_irq(&phba->hbalock);
if (phba->link_state > LPFC_LINK_DOWN) {
phba->link_state = LPFC_LINK_DOWN;
if (phba->sli4_hba.conf_trunk) {
phba->trunk_link.link0.state = 0;
phba->trunk_link.link1.state = 0;
phba->trunk_link.link2.state = 0;
phba->trunk_link.link3.state = 0;
phba->trunk_link.phy_lnk_speed =
LPFC_LINK_SPEED_UNKNOWN;
phba->sli4_hba.link_state.logical_speed =
LPFC_LINK_SPEED_UNKNOWN;
}
spin_lock_irq(shost->host_lock);
phba->pport->fc_flag &= ~FC_LBIT;
spin_unlock_irq(shost->host_lock);
}
vports = lpfc_create_vport_work_array(phba);
scsi: lpfc: NVME Initiator: Base modifications NVME Initiator: Base modifications This patch adds base modifications for NVME initiator support. The base modifications consist of: - Formal split of SLI3 rings from SLI-4 WQs (sometimes referred to as rings as well) as implementation now widely varies between the two. - Addition of configuration modes: SCSI initiator only; NVME initiator only; NVME target only; and SCSI and NVME initiator. The configuration mode drives overall adapter configuration, offloads enabled, and resource splits. NVME support is only available on SLI-4 devices and newer fw. - Implements the following based on configuration mode: - Exchange resources are split by protocol; Obviously, if only 1 mode, then no split occurs. Default is 50/50. module attribute allows tuning. - Pools and config parameters are separated per-protocol - Each protocol has it's own set of queues, but share interrupt vectors. SCSI: SLI3 devices have few queues and the original style of queue allocation remains. SLI4 devices piggy back on an "io-channel" concept that eventually needs to merge with scsi-mq/blk-mq support (it is underway). For now, the paradigm continues as it existed prior. io channel allocates N msix and N WQs (N=4 default) and either round robins or uses cpu # modulo N for scheduling. A bunch of module parameters allow the configuration to be tuned. NVME (initiator): Allocates an msix per cpu (or whatever pci_alloc_irq_vectors gets) Allocates a WQ per cpu, and maps the WQs to msix on a WQ # modulo msix vector count basis. Module parameters exist to cap/control the config if desired. - Each protocol has its own buffer and dma pools. I apologize for the size of the patch. Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <james.smart@broadcom.com> ---- Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2017-02-12 21:52:30 +00:00
if (vports != NULL) {
for (i = 0; i <= phba->max_vports && vports[i] != NULL; i++) {
/* Issue a LINK DOWN event to all nodes */
lpfc_linkdown_port(vports[i]);
scsi: lpfc: NVME Initiator: Base modifications NVME Initiator: Base modifications This patch adds base modifications for NVME initiator support. The base modifications consist of: - Formal split of SLI3 rings from SLI-4 WQs (sometimes referred to as rings as well) as implementation now widely varies between the two. - Addition of configuration modes: SCSI initiator only; NVME initiator only; NVME target only; and SCSI and NVME initiator. The configuration mode drives overall adapter configuration, offloads enabled, and resource splits. NVME support is only available on SLI-4 devices and newer fw. - Implements the following based on configuration mode: - Exchange resources are split by protocol; Obviously, if only 1 mode, then no split occurs. Default is 50/50. module attribute allows tuning. - Pools and config parameters are separated per-protocol - Each protocol has it's own set of queues, but share interrupt vectors. SCSI: SLI3 devices have few queues and the original style of queue allocation remains. SLI4 devices piggy back on an "io-channel" concept that eventually needs to merge with scsi-mq/blk-mq support (it is underway). For now, the paradigm continues as it existed prior. io channel allocates N msix and N WQs (N=4 default) and either round robins or uses cpu # modulo N for scheduling. A bunch of module parameters allow the configuration to be tuned. NVME (initiator): Allocates an msix per cpu (or whatever pci_alloc_irq_vectors gets) Allocates a WQ per cpu, and maps the WQs to msix on a WQ # modulo msix vector count basis. Module parameters exist to cap/control the config if desired. - Each protocol has its own buffer and dma pools. I apologize for the size of the patch. Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <james.smart@broadcom.com> ---- Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2017-02-12 21:52:30 +00:00
vports[i]->fc_myDID = 0;
if ((vport->cfg_enable_fc4_type == LPFC_ENABLE_BOTH) ||
(vport->cfg_enable_fc4_type == LPFC_ENABLE_NVME)) {
if (phba->nvmet_support)
lpfc_nvmet_update_targetport(phba);
else
lpfc_nvme_update_localport(vports[i]);
}
}
scsi: lpfc: NVME Initiator: Base modifications NVME Initiator: Base modifications This patch adds base modifications for NVME initiator support. The base modifications consist of: - Formal split of SLI3 rings from SLI-4 WQs (sometimes referred to as rings as well) as implementation now widely varies between the two. - Addition of configuration modes: SCSI initiator only; NVME initiator only; NVME target only; and SCSI and NVME initiator. The configuration mode drives overall adapter configuration, offloads enabled, and resource splits. NVME support is only available on SLI-4 devices and newer fw. - Implements the following based on configuration mode: - Exchange resources are split by protocol; Obviously, if only 1 mode, then no split occurs. Default is 50/50. module attribute allows tuning. - Pools and config parameters are separated per-protocol - Each protocol has it's own set of queues, but share interrupt vectors. SCSI: SLI3 devices have few queues and the original style of queue allocation remains. SLI4 devices piggy back on an "io-channel" concept that eventually needs to merge with scsi-mq/blk-mq support (it is underway). For now, the paradigm continues as it existed prior. io channel allocates N msix and N WQs (N=4 default) and either round robins or uses cpu # modulo N for scheduling. A bunch of module parameters allow the configuration to be tuned. NVME (initiator): Allocates an msix per cpu (or whatever pci_alloc_irq_vectors gets) Allocates a WQ per cpu, and maps the WQs to msix on a WQ # modulo msix vector count basis. Module parameters exist to cap/control the config if desired. - Each protocol has its own buffer and dma pools. I apologize for the size of the patch. Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <james.smart@broadcom.com> ---- Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2017-02-12 21:52:30 +00:00
}
[SCSI] lpfc 8.2.4 : Miscellaneous Fixes Miscellaneous Fixes: - Fix a couple of sparse complaints - Reset the FCP recovery flag when the node is not a FCP2 device. - Speed up offline prep delays - Fixed a memory leak in lpfc_mem_alloc failure path - Fixed external loopback test. - Fixed error code returned from the driver when HBA is over heated. - Correct Max NPIV vport to limits read from adapter - Add missing locks around fc_flag and FC_NEEDS_REG_VPI - Add missing hba ids for device identification - Added support for SET_VARIABLE and MBX_WRITE_WWN mailbox commands - Changed all temperature event messages from warning to error - Fix reporting of link speed when link is down - Added support for MBX_WRITE_WWN mailbox command - Change del_timer_sync() in ISR to del_timer() in interrupt handler - Correct instances of beXX_to_cpu() that should be cpu_to_beXX() - Perform target flush before releasing node references on module unload - Avoid bogus devloss_tmo messages when driver unloads - Fix panic when HBA generates ERATT interupt - Fix mbox race condition and a workaround on back-to-back mailbox commands - Force NPIV off for pt2pt mode between 2 NPorts - Stop worker thread before removing fc_host. - Fix up discovery timeout error case due to missing clear_la - Tighten mailbox polling code to speed up detection of fast completions - Only allow DUMP_MEMORY if adapter offline due to overtemp errors - Added extended error information to the log messages in chip init. Signed-off-by: James Smart <James.Smart@emulex.com> Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
2008-01-11 06:52:54 +00:00
lpfc_destroy_vport_work_array(phba, vports);
/* Clean up any SLI3 firmware default rpi's */
scsi: lpfc: Improve PCI EEH Error and Recovery Handling Following EEH errors, the driver can crash or hang when deleting the localport or when attempting to unload. The EEH handlers in the driver did not notify the NVMe-FC transport before tearing the driver down. This was delayed until the resume steps. This worked for SCSI because lpfc_block_scsi() would notify the scsi_fc_transport that the target was not available but it would not clean up all the references to the ndlp. The SLI3 prep for dev reset handler did the lpfc_offline_prep() and lpfc_offline() calls to get the port stopped before restarting. The SLI4 version of the prep for dev reset just destroyed the queues and did not stop NVMe from continuing. Also because the port was not really stopped the localport destroy would hang because the transport was still waiting for I/O. Additionally, a devloss tmo can fire and post events to a stopped worker thread creating another hang condition. lpfc_sli4_prep_dev_for_reset() is modified to call lpfc_offline_prep() and lpfc_offline() rather than just lpfc_scsi_dev_block() to ensure both SCSI and NVMe transports are notified to block I/O to the driver. Logic is added to devloss handler and worker thread to clean up ndlp references and quiesce appropriately. Link: https://lore.kernel.org/r/20220317032737.45308-2-jsmart2021@gmail.com Co-developed-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2022-03-17 03:27:34 +00:00
if (phba->sli_rev > LPFC_SLI_REV3 || offline)
goto skip_unreg_did;
mb = mempool_alloc(phba->mbox_mem_pool, GFP_KERNEL);
if (mb) {
lpfc_unreg_did(phba, 0xffff, LPFC_UNREG_ALL_DFLT_RPIS, mb);
mb->vport = vport;
mb->mbox_cmpl = lpfc_sli_def_mbox_cmpl;
if (lpfc_sli_issue_mbox(phba, mb, MBX_NOWAIT)
== MBX_NOT_FINISHED) {
mempool_free(mb, phba->mbox_mem_pool);
}
}
skip_unreg_did:
/* Setup myDID for link up if we are in pt2pt mode */
if (phba->pport->fc_flag & FC_PT2PT) {
mb = mempool_alloc(phba->mbox_mem_pool, GFP_KERNEL);
if (mb) {
lpfc_config_link(phba, mb);
mb->mbox_cmpl = lpfc_sli_def_mbox_cmpl;
mb->vport = vport;
if (lpfc_sli_issue_mbox(phba, mb, MBX_NOWAIT)
== MBX_NOT_FINISHED) {
mempool_free(mb, phba->mbox_mem_pool);
}
}
spin_lock_irq(shost->host_lock);
phba->pport->fc_flag &= ~(FC_PT2PT | FC_PT2PT_PLOGI);
phba->pport->rcv_flogi_cnt = 0;
spin_unlock_irq(shost->host_lock);
}
return 0;
}
static void
lpfc_linkup_cleanup_nodes(struct lpfc_vport *vport)
{
struct lpfc_nodelist *ndlp;
list_for_each_entry(ndlp, &vport->fc_nodes, nlp_listp) {
ndlp->nlp_fc4_type &= ~(NLP_FC4_FCP | NLP_FC4_NVME);
scsi: lpfc: Rework remote port ref counting and node freeing When a remote port is disconnected and disappears, its node structure (ndlp) stays allocated and on a vport node list. While on the list it can be matched, thus requires validation checks on state to be added in numerous code paths. If the node comes back, its possible for there to be multiple node structures for the same device on the vport node list. There is no reason to keep the node structure around after it is no longer in existence, and the current implementation creates problems for itself (multiple nodes) and lots of unnecessary code for state validation. Additionally, the reference taking on the node structure didn't follow the normal model used by the kernel kref api. It included lots of odd logic to match state with reference count. The combination of this odd logic plus the way it was implicitly used in the discovery engine made its reference taking implementation suspect and extremely hard to follow. Change the driver such that the reference taking routines are now normal ref increments/decrements and callout on refcount=0. With this in place, the rework can be done such that the node structure is fully removed and deallocated when the remote port no longer exists and all references are removed. This removal logic, and the basic ref counting are intrically tied, thus in a single patch. Link: https://lore.kernel.org/r/20201115192646.12977-2-james.smart@broadcom.com Co-developed-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <james.smart@broadcom.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-11-15 19:26:30 +00:00
if (ndlp->nlp_state == NLP_STE_UNUSED_NODE)
continue;
if (ndlp->nlp_type & NLP_FABRIC) {
/* On Linkup its safe to clean up the ndlp
* from Fabric connections.
*/
if (ndlp->nlp_DID != Fabric_DID)
lpfc_unreg_rpi(vport, ndlp);
lpfc_nlp_set_state(vport, ndlp, NLP_STE_NPR_NODE);
} else if (!(ndlp->nlp_flag & NLP_NPR_ADISC)) {
/* Fail outstanding IO now since device is
* marked for PLOGI.
*/
lpfc_unreg_rpi(vport, ndlp);
}
}
}
static void
lpfc_linkup_port(struct lpfc_vport *vport)
{
struct Scsi_Host *shost = lpfc_shost_from_vport(vport);
struct lpfc_hba *phba = vport->phba;
if ((vport->load_flag & FC_UNLOADING) != 0)
return;
lpfc_debugfs_disc_trc(vport, LPFC_DISC_TRC_ELS_CMD,
"Link Up: top:x%x speed:x%x flg:x%x",
phba->fc_topology, phba->fc_linkspeed, phba->link_flag);
/* If NPIV is not enabled, only bring the physical port up */
if (!(phba->sli3_options & LPFC_SLI3_NPIV_ENABLED) &&
(vport != phba->pport))
return;
if (vport->cfg_enable_fc4_type != LPFC_ENABLE_NVME)
fc_host_post_event(shost, fc_get_event_number(),
FCH_EVT_LINKUP, 0);
spin_lock_irq(shost->host_lock);
if (phba->defer_flogi_acc_flag)
vport->fc_flag &= ~(FC_ABORT_DISCOVERY | FC_RSCN_MODE |
FC_NLP_MORE | FC_RSCN_DISCOVERY);
else
vport->fc_flag &= ~(FC_PT2PT | FC_PT2PT_PLOGI |
FC_ABORT_DISCOVERY | FC_RSCN_MODE |
FC_NLP_MORE | FC_RSCN_DISCOVERY);
vport->fc_flag |= FC_NDISC_ACTIVE;
vport->fc_ns_retry = 0;
spin_unlock_irq(shost->host_lock);
lpfc_setup_fdmi_mask(vport);
lpfc_linkup_cleanup_nodes(vport);
}
static int
lpfc_linkup(struct lpfc_hba *phba)
{
struct lpfc_vport **vports;
int i;
struct Scsi_Host *shost = lpfc_shost_from_vport(phba->pport);
phba->link_state = LPFC_LINK_UP;
/* Unblock fabric iocbs if they are blocked */
clear_bit(FABRIC_COMANDS_BLOCKED, &phba->bit_flags);
del_timer_sync(&phba->fabric_block_timer);
vports = lpfc_create_vport_work_array(phba);
if (vports != NULL)
for (i = 0; i <= phba->max_vports && vports[i] != NULL; i++)
lpfc_linkup_port(vports[i]);
[SCSI] lpfc 8.2.4 : Miscellaneous Fixes Miscellaneous Fixes: - Fix a couple of sparse complaints - Reset the FCP recovery flag when the node is not a FCP2 device. - Speed up offline prep delays - Fixed a memory leak in lpfc_mem_alloc failure path - Fixed external loopback test. - Fixed error code returned from the driver when HBA is over heated. - Correct Max NPIV vport to limits read from adapter - Add missing locks around fc_flag and FC_NEEDS_REG_VPI - Add missing hba ids for device identification - Added support for SET_VARIABLE and MBX_WRITE_WWN mailbox commands - Changed all temperature event messages from warning to error - Fix reporting of link speed when link is down - Added support for MBX_WRITE_WWN mailbox command - Change del_timer_sync() in ISR to del_timer() in interrupt handler - Correct instances of beXX_to_cpu() that should be cpu_to_beXX() - Perform target flush before releasing node references on module unload - Avoid bogus devloss_tmo messages when driver unloads - Fix panic when HBA generates ERATT interupt - Fix mbox race condition and a workaround on back-to-back mailbox commands - Force NPIV off for pt2pt mode between 2 NPorts - Stop worker thread before removing fc_host. - Fix up discovery timeout error case due to missing clear_la - Tighten mailbox polling code to speed up detection of fast completions - Only allow DUMP_MEMORY if adapter offline due to overtemp errors - Added extended error information to the log messages in chip init. Signed-off-by: James Smart <James.Smart@emulex.com> Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
2008-01-11 06:52:54 +00:00
lpfc_destroy_vport_work_array(phba, vports);
/* Clear the pport flogi counter in case the link down was
* absorbed without an ACQE. No lock here - in worker thread
* and discovery is synchronized.
*/
spin_lock_irq(shost->host_lock);
phba->pport->rcv_flogi_cnt = 0;
spin_unlock_irq(shost->host_lock);
scsi: lpfc: Defer LS_ACC to FLOGI on point to point logins The current discovery state machine the driver treated FLOGI oddly. When point to point, an FLOGI is to be exchanged by the two ports, with the port with the most significant WWN then proceeding with PLOGI. The implementation in the driver was keyed to closely with "what have I sent", not with what has happened between the two endpoints. Thus, it blatantly would ACC an FLOGI, but reject PLOGI's until it had its FLOGI ACC'd. The problem is - the sending of FLOGI may be delayed for some reason, or the response to FLOGI held off by the other side. In the failing situation the other side sent an FLOGI, which was ACC'd, then sent PLOGIs which were then rjt'd until the retry count for the PLOGIs were exceeded and the port gave up. The FLOGI may have been very late in transmit, or the response held off until the PLOGIs failed. Given the other port had the higher WWN, no PLOGIs would occur and communication stopped. Correct the situation by changing the FLOGI handling. Defer any response to an FLOGI until the driver has sent its FLOGI as well. Then, upon either completion of the sent FLOGI, or upon sending an ACC to a received FLOGI (which may be received before or just after FLOGI was sent). the driver will act on who has the higher WWN. if the other port does, the driver will noop any handling of an FLOGI response (if outstanding) and wait for PLOGI. If the local port does, the driver will transition to sending PLOGI and will noop any action on responding to an FLOGI (if not yet received). Fortunately, to implement this, it only took another state flag and deferring any FLOGI response if the FLOGI has yet to be transmit. All subsequent actions were already in place. Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2018-11-30 00:09:36 +00:00
/* reinitialize initial HBA flag */
phba->hba_flag &= ~(HBA_FLOGI_ISSUED | HBA_RHBA_CMPL);
scsi: lpfc: Defer LS_ACC to FLOGI on point to point logins The current discovery state machine the driver treated FLOGI oddly. When point to point, an FLOGI is to be exchanged by the two ports, with the port with the most significant WWN then proceeding with PLOGI. The implementation in the driver was keyed to closely with "what have I sent", not with what has happened between the two endpoints. Thus, it blatantly would ACC an FLOGI, but reject PLOGI's until it had its FLOGI ACC'd. The problem is - the sending of FLOGI may be delayed for some reason, or the response to FLOGI held off by the other side. In the failing situation the other side sent an FLOGI, which was ACC'd, then sent PLOGIs which were then rjt'd until the retry count for the PLOGIs were exceeded and the port gave up. The FLOGI may have been very late in transmit, or the response held off until the PLOGIs failed. Given the other port had the higher WWN, no PLOGIs would occur and communication stopped. Correct the situation by changing the FLOGI handling. Defer any response to an FLOGI until the driver has sent its FLOGI as well. Then, upon either completion of the sent FLOGI, or upon sending an ACC to a received FLOGI (which may be received before or just after FLOGI was sent). the driver will act on who has the higher WWN. if the other port does, the driver will noop any handling of an FLOGI response (if outstanding) and wait for PLOGI. If the local port does, the driver will transition to sending PLOGI and will noop any action on responding to an FLOGI (if not yet received). Fortunately, to implement this, it only took another state flag and deferring any FLOGI response if the FLOGI has yet to be transmit. All subsequent actions were already in place. Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2018-11-30 00:09:36 +00:00
return 0;
}
/*
* This routine handles processing a CLEAR_LA mailbox
* command upon completion. It is setup in the LPFC_MBOXQ
* as the completion routine when the command is
scsi: lpfc: NVME Initiator: Base modifications NVME Initiator: Base modifications This patch adds base modifications for NVME initiator support. The base modifications consist of: - Formal split of SLI3 rings from SLI-4 WQs (sometimes referred to as rings as well) as implementation now widely varies between the two. - Addition of configuration modes: SCSI initiator only; NVME initiator only; NVME target only; and SCSI and NVME initiator. The configuration mode drives overall adapter configuration, offloads enabled, and resource splits. NVME support is only available on SLI-4 devices and newer fw. - Implements the following based on configuration mode: - Exchange resources are split by protocol; Obviously, if only 1 mode, then no split occurs. Default is 50/50. module attribute allows tuning. - Pools and config parameters are separated per-protocol - Each protocol has it's own set of queues, but share interrupt vectors. SCSI: SLI3 devices have few queues and the original style of queue allocation remains. SLI4 devices piggy back on an "io-channel" concept that eventually needs to merge with scsi-mq/blk-mq support (it is underway). For now, the paradigm continues as it existed prior. io channel allocates N msix and N WQs (N=4 default) and either round robins or uses cpu # modulo N for scheduling. A bunch of module parameters allow the configuration to be tuned. NVME (initiator): Allocates an msix per cpu (or whatever pci_alloc_irq_vectors gets) Allocates a WQ per cpu, and maps the WQs to msix on a WQ # modulo msix vector count basis. Module parameters exist to cap/control the config if desired. - Each protocol has its own buffer and dma pools. I apologize for the size of the patch. Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <james.smart@broadcom.com> ---- Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2017-02-12 21:52:30 +00:00
* handed off to the SLI layer. SLI3 only.
*/
static void
lpfc_mbx_cmpl_clear_la(struct lpfc_hba *phba, LPFC_MBOXQ_t *pmb)
{
struct lpfc_vport *vport = pmb->vport;
struct Scsi_Host *shost = lpfc_shost_from_vport(vport);
struct lpfc_sli *psli = &phba->sli;
MAILBOX_t *mb = &pmb->u.mb;
uint32_t control;
/* Since we don't do discovery right now, turn these off here */
scsi: lpfc: NVME Initiator: Base modifications NVME Initiator: Base modifications This patch adds base modifications for NVME initiator support. The base modifications consist of: - Formal split of SLI3 rings from SLI-4 WQs (sometimes referred to as rings as well) as implementation now widely varies between the two. - Addition of configuration modes: SCSI initiator only; NVME initiator only; NVME target only; and SCSI and NVME initiator. The configuration mode drives overall adapter configuration, offloads enabled, and resource splits. NVME support is only available on SLI-4 devices and newer fw. - Implements the following based on configuration mode: - Exchange resources are split by protocol; Obviously, if only 1 mode, then no split occurs. Default is 50/50. module attribute allows tuning. - Pools and config parameters are separated per-protocol - Each protocol has it's own set of queues, but share interrupt vectors. SCSI: SLI3 devices have few queues and the original style of queue allocation remains. SLI4 devices piggy back on an "io-channel" concept that eventually needs to merge with scsi-mq/blk-mq support (it is underway). For now, the paradigm continues as it existed prior. io channel allocates N msix and N WQs (N=4 default) and either round robins or uses cpu # modulo N for scheduling. A bunch of module parameters allow the configuration to be tuned. NVME (initiator): Allocates an msix per cpu (or whatever pci_alloc_irq_vectors gets) Allocates a WQ per cpu, and maps the WQs to msix on a WQ # modulo msix vector count basis. Module parameters exist to cap/control the config if desired. - Each protocol has its own buffer and dma pools. I apologize for the size of the patch. Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <james.smart@broadcom.com> ---- Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2017-02-12 21:52:30 +00:00
psli->sli3_ring[LPFC_EXTRA_RING].flag &= ~LPFC_STOP_IOCB_EVENT;
psli->sli3_ring[LPFC_FCP_RING].flag &= ~LPFC_STOP_IOCB_EVENT;
/* Check for error */
if ((mb->mbxStatus) && (mb->mbxStatus != 0x1601)) {
/* CLEAR_LA mbox error <mbxStatus> state <hba_state> */
scsi: lpfc: Add an internal trace log buffer The current logging methods typically end up requesting a reproduction with a different logging level set to figure out what happened. This was mainly by design to not clutter the kernel log messages with things that were typically not interesting and the messages themselves could cause other issues. When looking to make a better system, it was seen that in many cases when more data was wanted was when another message, usually at KERN_ERR level, was logged. And in most cases, what the additional logging that was then enabled was typically. Most of these areas fell into the discovery machine. Based on this summary, the following design has been put in place: The driver will maintain an internal log (256 elements of 256 bytes). The "additional logging" messages that are usually enabled in a reproduction will be changed to now log all the time to the internal log. A new logging level is defined - LOG_TRACE_EVENT. When this level is set (it is not by default) and a message marked as KERN_ERR is logged, all the messages in the internal log will be dumped to the kernel log before the KERN_ERR message is logged. There is a timestamp on each message added to the internal log. However, this timestamp is not converted to wall time when logged. The value of the timestamp is solely to give a crude time reference for the messages. Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-06-30 21:50:00 +00:00
lpfc_printf_vlog(vport, KERN_ERR, LOG_TRACE_EVENT,
"0320 CLEAR_LA mbxStatus error x%x hba "
"state x%x\n",
mb->mbxStatus, vport->port_state);
phba->link_state = LPFC_HBA_ERROR;
goto out;
}
if (vport->port_type == LPFC_PHYSICAL_PORT)
phba->link_state = LPFC_HBA_READY;
spin_lock_irq(&phba->hbalock);
psli->sli_flag |= LPFC_PROCESS_LA;
control = readl(phba->HCregaddr);
control |= HC_LAINT_ENA;
writel(control, phba->HCregaddr);
readl(phba->HCregaddr); /* flush */
spin_unlock_irq(&phba->hbalock);
mempool_free(pmb, phba->mbox_mem_pool);
return;
out:
/* Device Discovery completes */
lpfc_printf_vlog(vport, KERN_INFO, LOG_DISCOVERY,
"0225 Device Discovery completes\n");
mempool_free(pmb, phba->mbox_mem_pool);
spin_lock_irq(shost->host_lock);
[SCSI] lpfc 8.2.6 : Multiple discovery fixes Multiple Discovery Fixes: - Fix race on discovery due to link events coinciding with vport_delete. - Use NLP_FABRIC state to filter out switch-based pseudo initiators that reuse the same WWNs. - Correct erroneous setting of DID=0 in lpfc_matchdid() - Correct extra reference count that was in the lookup path for the remoteid from an unsolicited ELS. - Correct double-free bug in els abort path. - Correct FDMI server discovery logic for switch that return a WWN of 0. - Fix bugs in ndlp mgmt when a node changes address - Correct bug that did not delete RSCNs for vports upon link transitions - Fix "0216 Link event during NS query" error which pops up when vports are swapped to different switch ports. - Add sanity checks on ndlp structures - Fix devloss log message to dump WWN correctly - Hold off mgmt commands that were interferring with discovery mailbox cmds - Remove unnecessary FC_ESTABLISH_LINK logic. - Correct some race conditions in the worker thread, resulting in devloss: - Clear the work_port_events field before handling the work port events - Clear the deferred ring event before handling a deferred ring event - Hold the hba lock when waking up the work thread - Send an acc for the rscn even when we aren't going to handle it - Fix locking behavior that was not properly protecting the ACTIVE flag, thus allowing mailbox command order to shift. Signed-off-by: James Smart <james.smart@emulex.com> Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
2008-04-07 14:15:56 +00:00
vport->fc_flag &= ~FC_ABORT_DISCOVERY;
spin_unlock_irq(shost->host_lock);
lpfc_can_disctmo(vport);
/* turn on Link Attention interrupts */
spin_lock_irq(&phba->hbalock);
psli->sli_flag |= LPFC_PROCESS_LA;
control = readl(phba->HCregaddr);
control |= HC_LAINT_ENA;
writel(control, phba->HCregaddr);
readl(phba->HCregaddr); /* flush */
spin_unlock_irq(&phba->hbalock);
return;
}
void
lpfc_mbx_cmpl_local_config_link(struct lpfc_hba *phba, LPFC_MBOXQ_t *pmb)
{
struct lpfc_vport *vport = pmb->vport;
LPFC_MBOXQ_t *sparam_mb;
u16 status = pmb->u.mb.mbxStatus;
int rc;
mempool_free(pmb, phba->mbox_mem_pool);
if (status)
goto out;
/* don't perform discovery for SLI4 loopback diagnostic test */
if ((phba->sli_rev == LPFC_SLI_REV4) &&
!(phba->hba_flag & HBA_FCOE_MODE) &&
(phba->link_flag & LS_LOOPBACK_MODE))
return;
2010-11-21 04:11:48 +00:00
if (phba->fc_topology == LPFC_TOPOLOGY_LOOP &&
vport->fc_flag & FC_PUBLIC_LOOP &&
!(vport->fc_flag & FC_LBIT)) {
/* Need to wait for FAN - use discovery timer
* for timeout. port_state is identically
* LPFC_LOCAL_CFG_LINK while waiting for FAN
*/
lpfc_set_disctmo(vport);
return;
}
/* Start discovery by sending a FLOGI. port_state is identically
* LPFC_FLOGI while waiting for FLOGI cmpl.
*/
if (vport->port_state != LPFC_FLOGI) {
/* Issue MBX_READ_SPARAM to update CSPs before FLOGI if
* bb-credit recovery is in place.
*/
if (phba->bbcredit_support && phba->cfg_enable_bbcr &&
!(phba->link_flag & LS_LOOPBACK_MODE)) {
sparam_mb = mempool_alloc(phba->mbox_mem_pool,
GFP_KERNEL);
if (!sparam_mb)
goto sparam_out;
rc = lpfc_read_sparam(phba, sparam_mb, 0);
if (rc) {
mempool_free(sparam_mb, phba->mbox_mem_pool);
goto sparam_out;
}
sparam_mb->vport = vport;
sparam_mb->mbox_cmpl = lpfc_mbx_cmpl_read_sparam;
rc = lpfc_sli_issue_mbox(phba, sparam_mb, MBX_NOWAIT);
if (rc == MBX_NOT_FINISHED) {
lpfc_mbox_rsrc_cleanup(phba, sparam_mb,
MBOX_THD_UNLOCKED);
goto sparam_out;
}
phba->hba_flag |= HBA_DEFER_FLOGI;
} else {
lpfc_initial_flogi(vport);
}
} else {
if (vport->fc_flag & FC_PT2PT)
lpfc_disc_start(vport);
}
return;
out:
scsi: lpfc: Add an internal trace log buffer The current logging methods typically end up requesting a reproduction with a different logging level set to figure out what happened. This was mainly by design to not clutter the kernel log messages with things that were typically not interesting and the messages themselves could cause other issues. When looking to make a better system, it was seen that in many cases when more data was wanted was when another message, usually at KERN_ERR level, was logged. And in most cases, what the additional logging that was then enabled was typically. Most of these areas fell into the discovery machine. Based on this summary, the following design has been put in place: The driver will maintain an internal log (256 elements of 256 bytes). The "additional logging" messages that are usually enabled in a reproduction will be changed to now log all the time to the internal log. A new logging level is defined - LOG_TRACE_EVENT. When this level is set (it is not by default) and a message marked as KERN_ERR is logged, all the messages in the internal log will be dumped to the kernel log before the KERN_ERR message is logged. There is a timestamp on each message added to the internal log. However, this timestamp is not converted to wall time when logged. The value of the timestamp is solely to give a crude time reference for the messages. Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-06-30 21:50:00 +00:00
lpfc_printf_vlog(vport, KERN_ERR, LOG_TRACE_EVENT,
"0306 CONFIG_LINK mbxStatus error x%x HBA state x%x\n",
status, vport->port_state);
sparam_out:
lpfc_linkdown(phba);
scsi: lpfc: Add an internal trace log buffer The current logging methods typically end up requesting a reproduction with a different logging level set to figure out what happened. This was mainly by design to not clutter the kernel log messages with things that were typically not interesting and the messages themselves could cause other issues. When looking to make a better system, it was seen that in many cases when more data was wanted was when another message, usually at KERN_ERR level, was logged. And in most cases, what the additional logging that was then enabled was typically. Most of these areas fell into the discovery machine. Based on this summary, the following design has been put in place: The driver will maintain an internal log (256 elements of 256 bytes). The "additional logging" messages that are usually enabled in a reproduction will be changed to now log all the time to the internal log. A new logging level is defined - LOG_TRACE_EVENT. When this level is set (it is not by default) and a message marked as KERN_ERR is logged, all the messages in the internal log will be dumped to the kernel log before the KERN_ERR message is logged. There is a timestamp on each message added to the internal log. However, this timestamp is not converted to wall time when logged. The value of the timestamp is solely to give a crude time reference for the messages. Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-06-30 21:50:00 +00:00
lpfc_printf_vlog(vport, KERN_ERR, LOG_TRACE_EVENT,
"0200 CONFIG_LINK bad hba state x%x\n",
vport->port_state);
lpfc_issue_clear_la(phba, vport);
return;
}
/**
* lpfc_sli4_clear_fcf_rr_bmask
scsi: lpfc: Fix kerneldoc parameter formatting/misnaming/missing issues Fixes the following W=1 kernel build warning(s): drivers/scsi/lpfc/lpfc_hbadisc.c:1209: warning: Function parameter or member 'phba' not described in 'lpfc_sli4_clear_fcf_rr_bmask' drivers/scsi/lpfc/lpfc_hbadisc.c:1309: warning: Function parameter or member 'sw_name' not described in 'lpfc_sw_name_match' drivers/scsi/lpfc/lpfc_hbadisc.c:1309: warning: Excess function parameter 'fab_name' description in 'lpfc_sw_name_match' drivers/scsi/lpfc/lpfc_hbadisc.c:1397: warning: Function parameter or member 'fcf_rec' not described in 'lpfc_copy_fcf_record' drivers/scsi/lpfc/lpfc_hbadisc.c:1397: warning: Excess function parameter 'fcf' description in 'lpfc_copy_fcf_record' drivers/scsi/lpfc/lpfc_hbadisc.c:1956: warning: Cannot understand lpfc_sli4_fcf_record_match - testing new FCF record for matching existing FCF on line 1956 - I thought it was a doc line drivers/scsi/lpfc/lpfc_hbadisc.c:2078: warning: Function parameter or member 'fcf_index' not described in 'lpfc_sli4_fcf_pri_list_del' drivers/scsi/lpfc/lpfc_hbadisc.c:2109: warning: Function parameter or member 'fcf_index' not described in 'lpfc_sli4_set_fcf_flogi_fail' drivers/scsi/lpfc/lpfc_hbadisc.c:2135: warning: Function parameter or member 'fcf_index' not described in 'lpfc_sli4_fcf_pri_list_add' drivers/scsi/lpfc/lpfc_hbadisc.c:2135: warning: Function parameter or member 'new_fcf_record' not described in 'lpfc_sli4_fcf_pri_list_add' Link: https://lore.kernel.org/r/20200723122446.1329773-3-lee.jones@linaro.org Cc: James Smart <james.smart@broadcom.com> Cc: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: Lee Jones <lee.jones@linaro.org> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-07-23 12:24:08 +00:00
* @phba: pointer to the struct lpfc_hba for this port.
* This fucnction resets the round robin bit mask and clears the
* fcf priority list. The list deletions are done while holding the
* hbalock. The ON_LIST flag and the FLOGI_FAILED flags are cleared
* from the lpfc_fcf_pri record.
**/
void
lpfc_sli4_clear_fcf_rr_bmask(struct lpfc_hba *phba)
{
struct lpfc_fcf_pri *fcf_pri;
struct lpfc_fcf_pri *next_fcf_pri;
memset(phba->fcf.fcf_rr_bmask, 0, sizeof(*phba->fcf.fcf_rr_bmask));
spin_lock_irq(&phba->hbalock);
list_for_each_entry_safe(fcf_pri, next_fcf_pri,
&phba->fcf.fcf_pri_list, list) {
list_del_init(&fcf_pri->list);
fcf_pri->fcf_rec.flag = 0;
}
spin_unlock_irq(&phba->hbalock);
}
static void
lpfc_mbx_cmpl_reg_fcfi(struct lpfc_hba *phba, LPFC_MBOXQ_t *mboxq)
{
struct lpfc_vport *vport = mboxq->vport;
if (mboxq->u.mb.mbxStatus) {
scsi: lpfc: Add an internal trace log buffer The current logging methods typically end up requesting a reproduction with a different logging level set to figure out what happened. This was mainly by design to not clutter the kernel log messages with things that were typically not interesting and the messages themselves could cause other issues. When looking to make a better system, it was seen that in many cases when more data was wanted was when another message, usually at KERN_ERR level, was logged. And in most cases, what the additional logging that was then enabled was typically. Most of these areas fell into the discovery machine. Based on this summary, the following design has been put in place: The driver will maintain an internal log (256 elements of 256 bytes). The "additional logging" messages that are usually enabled in a reproduction will be changed to now log all the time to the internal log. A new logging level is defined - LOG_TRACE_EVENT. When this level is set (it is not by default) and a message marked as KERN_ERR is logged, all the messages in the internal log will be dumped to the kernel log before the KERN_ERR message is logged. There is a timestamp on each message added to the internal log. However, this timestamp is not converted to wall time when logged. The value of the timestamp is solely to give a crude time reference for the messages. Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-06-30 21:50:00 +00:00
lpfc_printf_vlog(vport, KERN_ERR, LOG_TRACE_EVENT,
"2017 REG_FCFI mbxStatus error x%x "
"HBA state x%x\n", mboxq->u.mb.mbxStatus,
vport->port_state);
goto fail_out;
}
/* Start FCoE discovery by sending a FLOGI. */
phba->fcf.fcfi = bf_get(lpfc_reg_fcfi_fcfi, &mboxq->u.mqe.un.reg_fcfi);
/* Set the FCFI registered flag */
spin_lock_irq(&phba->hbalock);
phba->fcf.fcf_flag |= FCF_REGISTERED;
spin_unlock_irq(&phba->hbalock);
/* If there is a pending FCoE event, restart FCF table scan. */
if ((!(phba->hba_flag & FCF_RR_INPROG)) &&
lpfc_check_pending_fcoe_event(phba, LPFC_UNREG_FCF))
goto fail_out;
/* Mark successful completion of FCF table scan */
spin_lock_irq(&phba->hbalock);
phba->fcf.fcf_flag |= (FCF_SCAN_DONE | FCF_IN_USE);
phba->hba_flag &= ~FCF_TS_INPROG;
if (vport->port_state != LPFC_FLOGI) {
phba->hba_flag |= FCF_RR_INPROG;
spin_unlock_irq(&phba->hbalock);
2010-11-21 04:11:48 +00:00
lpfc_issue_init_vfi(vport);
goto out;
}
spin_unlock_irq(&phba->hbalock);
goto out;
fail_out:
spin_lock_irq(&phba->hbalock);
phba->hba_flag &= ~FCF_RR_INPROG;
spin_unlock_irq(&phba->hbalock);
out:
mempool_free(mboxq, phba->mbox_mem_pool);
}
/**
* lpfc_fab_name_match - Check if the fcf fabric name match.
* @fab_name: pointer to fabric name.
* @new_fcf_record: pointer to fcf record.
*
* This routine compare the fcf record's fabric name with provided
* fabric name. If the fabric name are identical this function
* returns 1 else return 0.
**/
static uint32_t
lpfc_fab_name_match(uint8_t *fab_name, struct fcf_record *new_fcf_record)
{
if (fab_name[0] != bf_get(lpfc_fcf_record_fab_name_0, new_fcf_record))
return 0;
if (fab_name[1] != bf_get(lpfc_fcf_record_fab_name_1, new_fcf_record))
return 0;
if (fab_name[2] != bf_get(lpfc_fcf_record_fab_name_2, new_fcf_record))
return 0;
if (fab_name[3] != bf_get(lpfc_fcf_record_fab_name_3, new_fcf_record))
return 0;
if (fab_name[4] != bf_get(lpfc_fcf_record_fab_name_4, new_fcf_record))
return 0;
if (fab_name[5] != bf_get(lpfc_fcf_record_fab_name_5, new_fcf_record))
return 0;
if (fab_name[6] != bf_get(lpfc_fcf_record_fab_name_6, new_fcf_record))
return 0;
if (fab_name[7] != bf_get(lpfc_fcf_record_fab_name_7, new_fcf_record))
return 0;
return 1;
}
/**
* lpfc_sw_name_match - Check if the fcf switch name match.
scsi: lpfc: Fix kerneldoc parameter formatting/misnaming/missing issues Fixes the following W=1 kernel build warning(s): drivers/scsi/lpfc/lpfc_hbadisc.c:1209: warning: Function parameter or member 'phba' not described in 'lpfc_sli4_clear_fcf_rr_bmask' drivers/scsi/lpfc/lpfc_hbadisc.c:1309: warning: Function parameter or member 'sw_name' not described in 'lpfc_sw_name_match' drivers/scsi/lpfc/lpfc_hbadisc.c:1309: warning: Excess function parameter 'fab_name' description in 'lpfc_sw_name_match' drivers/scsi/lpfc/lpfc_hbadisc.c:1397: warning: Function parameter or member 'fcf_rec' not described in 'lpfc_copy_fcf_record' drivers/scsi/lpfc/lpfc_hbadisc.c:1397: warning: Excess function parameter 'fcf' description in 'lpfc_copy_fcf_record' drivers/scsi/lpfc/lpfc_hbadisc.c:1956: warning: Cannot understand lpfc_sli4_fcf_record_match - testing new FCF record for matching existing FCF on line 1956 - I thought it was a doc line drivers/scsi/lpfc/lpfc_hbadisc.c:2078: warning: Function parameter or member 'fcf_index' not described in 'lpfc_sli4_fcf_pri_list_del' drivers/scsi/lpfc/lpfc_hbadisc.c:2109: warning: Function parameter or member 'fcf_index' not described in 'lpfc_sli4_set_fcf_flogi_fail' drivers/scsi/lpfc/lpfc_hbadisc.c:2135: warning: Function parameter or member 'fcf_index' not described in 'lpfc_sli4_fcf_pri_list_add' drivers/scsi/lpfc/lpfc_hbadisc.c:2135: warning: Function parameter or member 'new_fcf_record' not described in 'lpfc_sli4_fcf_pri_list_add' Link: https://lore.kernel.org/r/20200723122446.1329773-3-lee.jones@linaro.org Cc: James Smart <james.smart@broadcom.com> Cc: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: Lee Jones <lee.jones@linaro.org> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-07-23 12:24:08 +00:00
* @sw_name: pointer to switch name.
* @new_fcf_record: pointer to fcf record.
*
* This routine compare the fcf record's switch name with provided
* switch name. If the switch name are identical this function
* returns 1 else return 0.
**/
static uint32_t
lpfc_sw_name_match(uint8_t *sw_name, struct fcf_record *new_fcf_record)
{
if (sw_name[0] != bf_get(lpfc_fcf_record_switch_name_0, new_fcf_record))
return 0;
if (sw_name[1] != bf_get(lpfc_fcf_record_switch_name_1, new_fcf_record))
return 0;
if (sw_name[2] != bf_get(lpfc_fcf_record_switch_name_2, new_fcf_record))
return 0;
if (sw_name[3] != bf_get(lpfc_fcf_record_switch_name_3, new_fcf_record))
return 0;
if (sw_name[4] != bf_get(lpfc_fcf_record_switch_name_4, new_fcf_record))
return 0;
if (sw_name[5] != bf_get(lpfc_fcf_record_switch_name_5, new_fcf_record))
return 0;
if (sw_name[6] != bf_get(lpfc_fcf_record_switch_name_6, new_fcf_record))
return 0;
if (sw_name[7] != bf_get(lpfc_fcf_record_switch_name_7, new_fcf_record))
return 0;
return 1;
}
/**
* lpfc_mac_addr_match - Check if the fcf mac address match.
* @mac_addr: pointer to mac address.
* @new_fcf_record: pointer to fcf record.
*
* This routine compare the fcf record's mac address with HBA's
* FCF mac address. If the mac addresses are identical this function
* returns 1 else return 0.
**/
static uint32_t
lpfc_mac_addr_match(uint8_t *mac_addr, struct fcf_record *new_fcf_record)
{
if (mac_addr[0] != bf_get(lpfc_fcf_record_mac_0, new_fcf_record))
return 0;
if (mac_addr[1] != bf_get(lpfc_fcf_record_mac_1, new_fcf_record))
return 0;
if (mac_addr[2] != bf_get(lpfc_fcf_record_mac_2, new_fcf_record))
return 0;
if (mac_addr[3] != bf_get(lpfc_fcf_record_mac_3, new_fcf_record))
return 0;
if (mac_addr[4] != bf_get(lpfc_fcf_record_mac_4, new_fcf_record))
return 0;
if (mac_addr[5] != bf_get(lpfc_fcf_record_mac_5, new_fcf_record))
return 0;
return 1;
}
static bool
lpfc_vlan_id_match(uint16_t curr_vlan_id, uint16_t new_vlan_id)
{
return (curr_vlan_id == new_vlan_id);
}
/**
* __lpfc_update_fcf_record_pri - update the lpfc_fcf_pri record.
* @phba: pointer to lpfc hba data structure.
* @fcf_index: Index for the lpfc_fcf_record.
* @new_fcf_record: pointer to hba fcf record.
*
* This routine updates the driver FCF priority record from the new HBA FCF
* record. The hbalock is asserted held in the code path calling this
* routine.
**/
static void
__lpfc_update_fcf_record_pri(struct lpfc_hba *phba, uint16_t fcf_index,
struct fcf_record *new_fcf_record
)
{
struct lpfc_fcf_pri *fcf_pri;
fcf_pri = &phba->fcf.fcf_pri[fcf_index];
fcf_pri->fcf_rec.fcf_index = fcf_index;
/* FCF record priority */
fcf_pri->fcf_rec.priority = new_fcf_record->fip_priority;
}
/**
* lpfc_copy_fcf_record - Copy fcf information to lpfc_hba.
scsi: lpfc: Fix kerneldoc parameter formatting/misnaming/missing issues Fixes the following W=1 kernel build warning(s): drivers/scsi/lpfc/lpfc_hbadisc.c:1209: warning: Function parameter or member 'phba' not described in 'lpfc_sli4_clear_fcf_rr_bmask' drivers/scsi/lpfc/lpfc_hbadisc.c:1309: warning: Function parameter or member 'sw_name' not described in 'lpfc_sw_name_match' drivers/scsi/lpfc/lpfc_hbadisc.c:1309: warning: Excess function parameter 'fab_name' description in 'lpfc_sw_name_match' drivers/scsi/lpfc/lpfc_hbadisc.c:1397: warning: Function parameter or member 'fcf_rec' not described in 'lpfc_copy_fcf_record' drivers/scsi/lpfc/lpfc_hbadisc.c:1397: warning: Excess function parameter 'fcf' description in 'lpfc_copy_fcf_record' drivers/scsi/lpfc/lpfc_hbadisc.c:1956: warning: Cannot understand lpfc_sli4_fcf_record_match - testing new FCF record for matching existing FCF on line 1956 - I thought it was a doc line drivers/scsi/lpfc/lpfc_hbadisc.c:2078: warning: Function parameter or member 'fcf_index' not described in 'lpfc_sli4_fcf_pri_list_del' drivers/scsi/lpfc/lpfc_hbadisc.c:2109: warning: Function parameter or member 'fcf_index' not described in 'lpfc_sli4_set_fcf_flogi_fail' drivers/scsi/lpfc/lpfc_hbadisc.c:2135: warning: Function parameter or member 'fcf_index' not described in 'lpfc_sli4_fcf_pri_list_add' drivers/scsi/lpfc/lpfc_hbadisc.c:2135: warning: Function parameter or member 'new_fcf_record' not described in 'lpfc_sli4_fcf_pri_list_add' Link: https://lore.kernel.org/r/20200723122446.1329773-3-lee.jones@linaro.org Cc: James Smart <james.smart@broadcom.com> Cc: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: Lee Jones <lee.jones@linaro.org> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-07-23 12:24:08 +00:00
* @fcf_rec: pointer to driver fcf record.
* @new_fcf_record: pointer to fcf record.
*
* This routine copies the FCF information from the FCF
* record to lpfc_hba data structure.
**/
static void
lpfc_copy_fcf_record(struct lpfc_fcf_rec *fcf_rec,
struct fcf_record *new_fcf_record)
{
/* Fabric name */
fcf_rec->fabric_name[0] =
bf_get(lpfc_fcf_record_fab_name_0, new_fcf_record);
fcf_rec->fabric_name[1] =
bf_get(lpfc_fcf_record_fab_name_1, new_fcf_record);
fcf_rec->fabric_name[2] =
bf_get(lpfc_fcf_record_fab_name_2, new_fcf_record);
fcf_rec->fabric_name[3] =
bf_get(lpfc_fcf_record_fab_name_3, new_fcf_record);
fcf_rec->fabric_name[4] =
bf_get(lpfc_fcf_record_fab_name_4, new_fcf_record);
fcf_rec->fabric_name[5] =
bf_get(lpfc_fcf_record_fab_name_5, new_fcf_record);
fcf_rec->fabric_name[6] =
bf_get(lpfc_fcf_record_fab_name_6, new_fcf_record);
fcf_rec->fabric_name[7] =
bf_get(lpfc_fcf_record_fab_name_7, new_fcf_record);
/* Mac address */
fcf_rec->mac_addr[0] = bf_get(lpfc_fcf_record_mac_0, new_fcf_record);
fcf_rec->mac_addr[1] = bf_get(lpfc_fcf_record_mac_1, new_fcf_record);
fcf_rec->mac_addr[2] = bf_get(lpfc_fcf_record_mac_2, new_fcf_record);
fcf_rec->mac_addr[3] = bf_get(lpfc_fcf_record_mac_3, new_fcf_record);
fcf_rec->mac_addr[4] = bf_get(lpfc_fcf_record_mac_4, new_fcf_record);
fcf_rec->mac_addr[5] = bf_get(lpfc_fcf_record_mac_5, new_fcf_record);
/* FCF record index */
fcf_rec->fcf_indx = bf_get(lpfc_fcf_record_fcf_index, new_fcf_record);
/* FCF record priority */
fcf_rec->priority = new_fcf_record->fip_priority;
/* Switch name */
fcf_rec->switch_name[0] =
bf_get(lpfc_fcf_record_switch_name_0, new_fcf_record);
fcf_rec->switch_name[1] =
bf_get(lpfc_fcf_record_switch_name_1, new_fcf_record);
fcf_rec->switch_name[2] =
bf_get(lpfc_fcf_record_switch_name_2, new_fcf_record);
fcf_rec->switch_name[3] =
bf_get(lpfc_fcf_record_switch_name_3, new_fcf_record);
fcf_rec->switch_name[4] =
bf_get(lpfc_fcf_record_switch_name_4, new_fcf_record);
fcf_rec->switch_name[5] =
bf_get(lpfc_fcf_record_switch_name_5, new_fcf_record);
fcf_rec->switch_name[6] =
bf_get(lpfc_fcf_record_switch_name_6, new_fcf_record);
fcf_rec->switch_name[7] =
bf_get(lpfc_fcf_record_switch_name_7, new_fcf_record);
}
/**
* __lpfc_update_fcf_record - Update driver fcf record
* @phba: pointer to lpfc hba data structure.
* @fcf_rec: pointer to driver fcf record.
* @new_fcf_record: pointer to hba fcf record.
* @addr_mode: address mode to be set to the driver fcf record.
* @vlan_id: vlan tag to be set to the driver fcf record.
* @flag: flag bits to be set to the driver fcf record.
*
* This routine updates the driver FCF record from the new HBA FCF record
* together with the address mode, vlan_id, and other informations. This
* routine is called with the hbalock held.
**/
static void
__lpfc_update_fcf_record(struct lpfc_hba *phba, struct lpfc_fcf_rec *fcf_rec,
struct fcf_record *new_fcf_record, uint32_t addr_mode,
uint16_t vlan_id, uint32_t flag)
{
lockdep_assert_held(&phba->hbalock);
/* Copy the fields from the HBA's FCF record */
lpfc_copy_fcf_record(fcf_rec, new_fcf_record);
/* Update other fields of driver FCF record */
fcf_rec->addr_mode = addr_mode;
fcf_rec->vlan_id = vlan_id;
fcf_rec->flag |= (flag | RECORD_VALID);
__lpfc_update_fcf_record_pri(phba,
bf_get(lpfc_fcf_record_fcf_index, new_fcf_record),
new_fcf_record);
}
/**
* lpfc_register_fcf - Register the FCF with hba.
* @phba: pointer to lpfc hba data structure.
*
* This routine issues a register fcfi mailbox command to register
* the fcf with HBA.
**/
static void
lpfc_register_fcf(struct lpfc_hba *phba)
{
LPFC_MBOXQ_t *fcf_mbxq;
int rc;
spin_lock_irq(&phba->hbalock);
/* If the FCF is not available do nothing. */
if (!(phba->fcf.fcf_flag & FCF_AVAILABLE)) {
phba->hba_flag &= ~(FCF_TS_INPROG | FCF_RR_INPROG);
spin_unlock_irq(&phba->hbalock);
return;
}
/* The FCF is already registered, start discovery */
if (phba->fcf.fcf_flag & FCF_REGISTERED) {
phba->fcf.fcf_flag |= (FCF_SCAN_DONE | FCF_IN_USE);
phba->hba_flag &= ~FCF_TS_INPROG;
if (phba->pport->port_state != LPFC_FLOGI &&
phba->pport->fc_flag & FC_FABRIC) {
phba->hba_flag |= FCF_RR_INPROG;
spin_unlock_irq(&phba->hbalock);
lpfc_initial_flogi(phba->pport);
return;
}
spin_unlock_irq(&phba->hbalock);
return;
}
spin_unlock_irq(&phba->hbalock);
fcf_mbxq = mempool_alloc(phba->mbox_mem_pool, GFP_KERNEL);
if (!fcf_mbxq) {
spin_lock_irq(&phba->hbalock);
phba->hba_flag &= ~(FCF_TS_INPROG | FCF_RR_INPROG);
spin_unlock_irq(&phba->hbalock);
return;
}
lpfc_reg_fcfi(phba, fcf_mbxq);
fcf_mbxq->vport = phba->pport;
fcf_mbxq->mbox_cmpl = lpfc_mbx_cmpl_reg_fcfi;
rc = lpfc_sli_issue_mbox(phba, fcf_mbxq, MBX_NOWAIT);
if (rc == MBX_NOT_FINISHED) {
spin_lock_irq(&phba->hbalock);
phba->hba_flag &= ~(FCF_TS_INPROG | FCF_RR_INPROG);
spin_unlock_irq(&phba->hbalock);
mempool_free(fcf_mbxq, phba->mbox_mem_pool);
}
return;
}
/**
* lpfc_match_fcf_conn_list - Check if the FCF record can be used for discovery.
* @phba: pointer to lpfc hba data structure.
* @new_fcf_record: pointer to fcf record.
* @boot_flag: Indicates if this record used by boot bios.
* @addr_mode: The address mode to be used by this FCF
* @vlan_id: The vlan id to be used as vlan tagging by this FCF.
*
* This routine compare the fcf record with connect list obtained from the
* config region to decide if this FCF can be used for SAN discovery. It returns
* 1 if this record can be used for SAN discovery else return zero. If this FCF
* record can be used for SAN discovery, the boot_flag will indicate if this FCF
* is used by boot bios and addr_mode will indicate the addressing mode to be
* used for this FCF when the function returns.
* If the FCF record need to be used with a particular vlan id, the vlan is
* set in the vlan_id on return of the function. If not VLAN tagging need to
* be used with the FCF vlan_id will be set to LPFC_FCOE_NULL_VID;
**/
static int
lpfc_match_fcf_conn_list(struct lpfc_hba *phba,
struct fcf_record *new_fcf_record,
uint32_t *boot_flag, uint32_t *addr_mode,
uint16_t *vlan_id)
{
struct lpfc_fcf_conn_entry *conn_entry;
int i, j, fcf_vlan_id = 0;
/* Find the lowest VLAN id in the FCF record */
for (i = 0; i < 512; i++) {
if (new_fcf_record->vlan_bitmap[i]) {
fcf_vlan_id = i * 8;
j = 0;
while (!((new_fcf_record->vlan_bitmap[i] >> j) & 1)) {
j++;
fcf_vlan_id++;
}
break;
}
}
/* FCF not valid/available or solicitation in progress */
if (!bf_get(lpfc_fcf_record_fcf_avail, new_fcf_record) ||
!bf_get(lpfc_fcf_record_fcf_valid, new_fcf_record) ||
bf_get(lpfc_fcf_record_fcf_sol, new_fcf_record))
return 0;
if (!(phba->hba_flag & HBA_FIP_SUPPORT)) {
*boot_flag = 0;
*addr_mode = bf_get(lpfc_fcf_record_mac_addr_prov,
new_fcf_record);
if (phba->valid_vlan)
*vlan_id = phba->vlan_id;
else
*vlan_id = LPFC_FCOE_NULL_VID;
return 1;
}
/*
* If there are no FCF connection table entry, driver connect to all
* FCFs.
*/
if (list_empty(&phba->fcf_conn_rec_list)) {
*boot_flag = 0;
*addr_mode = bf_get(lpfc_fcf_record_mac_addr_prov,
new_fcf_record);
/*
* When there are no FCF connect entries, use driver's default
* addressing mode - FPMA.
*/
if (*addr_mode & LPFC_FCF_FPMA)
*addr_mode = LPFC_FCF_FPMA;
/* If FCF record report a vlan id use that vlan id */
if (fcf_vlan_id)
*vlan_id = fcf_vlan_id;
else
*vlan_id = LPFC_FCOE_NULL_VID;
return 1;
}
list_for_each_entry(conn_entry,
&phba->fcf_conn_rec_list, list) {
if (!(conn_entry->conn_rec.flags & FCFCNCT_VALID))
continue;
if ((conn_entry->conn_rec.flags & FCFCNCT_FBNM_VALID) &&
!lpfc_fab_name_match(conn_entry->conn_rec.fabric_name,
new_fcf_record))
continue;
if ((conn_entry->conn_rec.flags & FCFCNCT_SWNM_VALID) &&
!lpfc_sw_name_match(conn_entry->conn_rec.switch_name,
new_fcf_record))
continue;
if (conn_entry->conn_rec.flags & FCFCNCT_VLAN_VALID) {
/*
* If the vlan bit map does not have the bit set for the
* vlan id to be used, then it is not a match.
*/
if (!(new_fcf_record->vlan_bitmap
[conn_entry->conn_rec.vlan_tag / 8] &
(1 << (conn_entry->conn_rec.vlan_tag % 8))))
continue;
}
/*
* If connection record does not support any addressing mode,
* skip the FCF record.
*/
if (!(bf_get(lpfc_fcf_record_mac_addr_prov, new_fcf_record)
& (LPFC_FCF_FPMA | LPFC_FCF_SPMA)))
continue;
/*
* Check if the connection record specifies a required
* addressing mode.
*/
if ((conn_entry->conn_rec.flags & FCFCNCT_AM_VALID) &&
!(conn_entry->conn_rec.flags & FCFCNCT_AM_PREFERRED)) {
/*
* If SPMA required but FCF not support this continue.
*/
if ((conn_entry->conn_rec.flags & FCFCNCT_AM_SPMA) &&
!(bf_get(lpfc_fcf_record_mac_addr_prov,
new_fcf_record) & LPFC_FCF_SPMA))
continue;
/*
* If FPMA required but FCF not support this continue.
*/
if (!(conn_entry->conn_rec.flags & FCFCNCT_AM_SPMA) &&
!(bf_get(lpfc_fcf_record_mac_addr_prov,
new_fcf_record) & LPFC_FCF_FPMA))
continue;
}
/*
* This fcf record matches filtering criteria.
*/
if (conn_entry->conn_rec.flags & FCFCNCT_BOOT)
*boot_flag = 1;
else
*boot_flag = 0;
/*
* If user did not specify any addressing mode, or if the
* preferred addressing mode specified by user is not supported
* by FCF, allow fabric to pick the addressing mode.
*/
*addr_mode = bf_get(lpfc_fcf_record_mac_addr_prov,
new_fcf_record);
/*
* If the user specified a required address mode, assign that
* address mode
*/
if ((conn_entry->conn_rec.flags & FCFCNCT_AM_VALID) &&
(!(conn_entry->conn_rec.flags & FCFCNCT_AM_PREFERRED)))
*addr_mode = (conn_entry->conn_rec.flags &
FCFCNCT_AM_SPMA) ?
LPFC_FCF_SPMA : LPFC_FCF_FPMA;
/*
* If the user specified a preferred address mode, use the
* addr mode only if FCF support the addr_mode.
*/
else if ((conn_entry->conn_rec.flags & FCFCNCT_AM_VALID) &&
(conn_entry->conn_rec.flags & FCFCNCT_AM_PREFERRED) &&
(conn_entry->conn_rec.flags & FCFCNCT_AM_SPMA) &&
(*addr_mode & LPFC_FCF_SPMA))
*addr_mode = LPFC_FCF_SPMA;
else if ((conn_entry->conn_rec.flags & FCFCNCT_AM_VALID) &&
(conn_entry->conn_rec.flags & FCFCNCT_AM_PREFERRED) &&
!(conn_entry->conn_rec.flags & FCFCNCT_AM_SPMA) &&
(*addr_mode & LPFC_FCF_FPMA))
*addr_mode = LPFC_FCF_FPMA;
/* If matching connect list has a vlan id, use it */
if (conn_entry->conn_rec.flags & FCFCNCT_VLAN_VALID)
*vlan_id = conn_entry->conn_rec.vlan_tag;
/*
* If no vlan id is specified in connect list, use the vlan id
* in the FCF record
*/
else if (fcf_vlan_id)
*vlan_id = fcf_vlan_id;
else
*vlan_id = LPFC_FCOE_NULL_VID;
return 1;
}
return 0;
}
/**
* lpfc_check_pending_fcoe_event - Check if there is pending fcoe event.
* @phba: pointer to lpfc hba data structure.
* @unreg_fcf: Unregister FCF if FCF table need to be re-scaned.
*
* This function check if there is any fcoe event pending while driver
* scan FCF entries. If there is any pending event, it will restart the
* FCF saning and return 1 else return 0.
*/
int
lpfc_check_pending_fcoe_event(struct lpfc_hba *phba, uint8_t unreg_fcf)
{
/*
* If the Link is up and no FCoE events while in the
* FCF discovery, no need to restart FCF discovery.
*/
if ((phba->link_state >= LPFC_LINK_UP) &&
(phba->fcoe_eventtag == phba->fcoe_eventtag_at_fcf_scan))
return 0;
lpfc_printf_log(phba, KERN_INFO, LOG_FIP,
"2768 Pending link or FCF event during current "
"handling of the previous event: link_state:x%x, "
"evt_tag_at_scan:x%x, evt_tag_current:x%x\n",
phba->link_state, phba->fcoe_eventtag_at_fcf_scan,
phba->fcoe_eventtag);
spin_lock_irq(&phba->hbalock);
phba->fcf.fcf_flag &= ~FCF_AVAILABLE;
spin_unlock_irq(&phba->hbalock);
if (phba->link_state >= LPFC_LINK_UP) {
lpfc_printf_log(phba, KERN_INFO, LOG_FIP | LOG_DISCOVERY,
"2780 Restart FCF table scan due to "
"pending FCF event:evt_tag_at_scan:x%x, "
"evt_tag_current:x%x\n",
phba->fcoe_eventtag_at_fcf_scan,
phba->fcoe_eventtag);
lpfc_sli4_fcf_scan_read_fcf_rec(phba, LPFC_FCOE_FCF_GET_FIRST);
} else {
/*
* Do not continue FCF discovery and clear FCF_TS_INPROG
* flag
*/
lpfc_printf_log(phba, KERN_INFO, LOG_FIP | LOG_DISCOVERY,
"2833 Stop FCF discovery process due to link "
"state change (x%x)\n", phba->link_state);
spin_lock_irq(&phba->hbalock);
phba->hba_flag &= ~(FCF_TS_INPROG | FCF_RR_INPROG);
phba->fcf.fcf_flag &= ~(FCF_REDISC_FOV | FCF_DISCOVERY);
spin_unlock_irq(&phba->hbalock);
}
/* Unregister the currently registered FCF if required */
if (unreg_fcf) {
spin_lock_irq(&phba->hbalock);
phba->fcf.fcf_flag &= ~FCF_REGISTERED;
spin_unlock_irq(&phba->hbalock);
lpfc_sli4_unregister_fcf(phba);
}
return 1;
}
/**
* lpfc_sli4_new_fcf_random_select - Randomly select an eligible new fcf record
* @phba: pointer to lpfc hba data structure.
* @fcf_cnt: number of eligible fcf record seen so far.
*
* This function makes an running random selection decision on FCF record to
* use through a sequence of @fcf_cnt eligible FCF records with equal
* probability. To perform integer manunipulation of random numbers with
treewide: use get_random_{u8,u16}() when possible, part 1 Rather than truncate a 32-bit value to a 16-bit value or an 8-bit value, simply use the get_random_{u8,u16}() functions, which are faster than wasting the additional bytes from a 32-bit value. This was done mechanically with this coccinelle script: @@ expression E; identifier get_random_u32 =~ "get_random_int|prandom_u32|get_random_u32"; typedef u16; typedef __be16; typedef __le16; typedef u8; @@ ( - (get_random_u32() & 0xffff) + get_random_u16() | - (get_random_u32() & 0xff) + get_random_u8() | - (get_random_u32() % 65536) + get_random_u16() | - (get_random_u32() % 256) + get_random_u8() | - (get_random_u32() >> 16) + get_random_u16() | - (get_random_u32() >> 24) + get_random_u8() | - (u16)get_random_u32() + get_random_u16() | - (u8)get_random_u32() + get_random_u8() | - (__be16)get_random_u32() + (__be16)get_random_u16() | - (__le16)get_random_u32() + (__le16)get_random_u16() | - prandom_u32_max(65536) + get_random_u16() | - prandom_u32_max(256) + get_random_u8() | - E->inet_id = get_random_u32() + E->inet_id = get_random_u16() ) @@ identifier get_random_u32 =~ "get_random_int|prandom_u32|get_random_u32"; typedef u16; identifier v; @@ - u16 v = get_random_u32(); + u16 v = get_random_u16(); @@ identifier get_random_u32 =~ "get_random_int|prandom_u32|get_random_u32"; typedef u8; identifier v; @@ - u8 v = get_random_u32(); + u8 v = get_random_u8(); @@ identifier get_random_u32 =~ "get_random_int|prandom_u32|get_random_u32"; typedef u16; u16 v; @@ - v = get_random_u32(); + v = get_random_u16(); @@ identifier get_random_u32 =~ "get_random_int|prandom_u32|get_random_u32"; typedef u8; u8 v; @@ - v = get_random_u32(); + v = get_random_u8(); // Find a potential literal @literal_mask@ expression LITERAL; type T; identifier get_random_u32 =~ "get_random_int|prandom_u32|get_random_u32"; position p; @@ ((T)get_random_u32()@p & (LITERAL)) // Examine limits @script:python add_one@ literal << literal_mask.LITERAL; RESULT; @@ value = None if literal.startswith('0x'): value = int(literal, 16) elif literal[0] in '123456789': value = int(literal, 10) if value is None: print("I don't know how to handle %s" % (literal)) cocci.include_match(False) elif value < 256: coccinelle.RESULT = cocci.make_ident("get_random_u8") elif value < 65536: coccinelle.RESULT = cocci.make_ident("get_random_u16") else: print("Skipping large mask of %s" % (literal)) cocci.include_match(False) // Replace the literal mask with the calculated result. @plus_one@ expression literal_mask.LITERAL; position literal_mask.p; identifier add_one.RESULT; identifier FUNC; @@ - (FUNC()@p & (LITERAL)) + (RESULT() & LITERAL) Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Kees Cook <keescook@chromium.org> Reviewed-by: Yury Norov <yury.norov@gmail.com> Acked-by: Jakub Kicinski <kuba@kernel.org> Acked-by: Toke Høiland-Jørgensen <toke@toke.dk> # for sch_cake Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
2022-10-05 15:23:53 +00:00
* size unit32_t, a 16-bit random number returned from get_random_u16() is
* taken as the random random number generated.
*
* Returns true when outcome is for the newly read FCF record should be
* chosen; otherwise, return false when outcome is for keeping the previously
* chosen FCF record.
**/
static bool
lpfc_sli4_new_fcf_random_select(struct lpfc_hba *phba, uint32_t fcf_cnt)
{
uint32_t rand_num;
/* Get 16-bit uniform random number */
treewide: use get_random_{u8,u16}() when possible, part 1 Rather than truncate a 32-bit value to a 16-bit value or an 8-bit value, simply use the get_random_{u8,u16}() functions, which are faster than wasting the additional bytes from a 32-bit value. This was done mechanically with this coccinelle script: @@ expression E; identifier get_random_u32 =~ "get_random_int|prandom_u32|get_random_u32"; typedef u16; typedef __be16; typedef __le16; typedef u8; @@ ( - (get_random_u32() & 0xffff) + get_random_u16() | - (get_random_u32() & 0xff) + get_random_u8() | - (get_random_u32() % 65536) + get_random_u16() | - (get_random_u32() % 256) + get_random_u8() | - (get_random_u32() >> 16) + get_random_u16() | - (get_random_u32() >> 24) + get_random_u8() | - (u16)get_random_u32() + get_random_u16() | - (u8)get_random_u32() + get_random_u8() | - (__be16)get_random_u32() + (__be16)get_random_u16() | - (__le16)get_random_u32() + (__le16)get_random_u16() | - prandom_u32_max(65536) + get_random_u16() | - prandom_u32_max(256) + get_random_u8() | - E->inet_id = get_random_u32() + E->inet_id = get_random_u16() ) @@ identifier get_random_u32 =~ "get_random_int|prandom_u32|get_random_u32"; typedef u16; identifier v; @@ - u16 v = get_random_u32(); + u16 v = get_random_u16(); @@ identifier get_random_u32 =~ "get_random_int|prandom_u32|get_random_u32"; typedef u8; identifier v; @@ - u8 v = get_random_u32(); + u8 v = get_random_u8(); @@ identifier get_random_u32 =~ "get_random_int|prandom_u32|get_random_u32"; typedef u16; u16 v; @@ - v = get_random_u32(); + v = get_random_u16(); @@ identifier get_random_u32 =~ "get_random_int|prandom_u32|get_random_u32"; typedef u8; u8 v; @@ - v = get_random_u32(); + v = get_random_u8(); // Find a potential literal @literal_mask@ expression LITERAL; type T; identifier get_random_u32 =~ "get_random_int|prandom_u32|get_random_u32"; position p; @@ ((T)get_random_u32()@p & (LITERAL)) // Examine limits @script:python add_one@ literal << literal_mask.LITERAL; RESULT; @@ value = None if literal.startswith('0x'): value = int(literal, 16) elif literal[0] in '123456789': value = int(literal, 10) if value is None: print("I don't know how to handle %s" % (literal)) cocci.include_match(False) elif value < 256: coccinelle.RESULT = cocci.make_ident("get_random_u8") elif value < 65536: coccinelle.RESULT = cocci.make_ident("get_random_u16") else: print("Skipping large mask of %s" % (literal)) cocci.include_match(False) // Replace the literal mask with the calculated result. @plus_one@ expression literal_mask.LITERAL; position literal_mask.p; identifier add_one.RESULT; identifier FUNC; @@ - (FUNC()@p & (LITERAL)) + (RESULT() & LITERAL) Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Reviewed-by: Kees Cook <keescook@chromium.org> Reviewed-by: Yury Norov <yury.norov@gmail.com> Acked-by: Jakub Kicinski <kuba@kernel.org> Acked-by: Toke Høiland-Jørgensen <toke@toke.dk> # for sch_cake Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
2022-10-05 15:23:53 +00:00
rand_num = get_random_u16();
/* Decision with probability 1/fcf_cnt */
if ((fcf_cnt * rand_num) < 0xFFFF)
return true;
else
return false;
}
/**
* lpfc_sli4_fcf_rec_mbox_parse - Parse read_fcf mbox command.
* @phba: pointer to lpfc hba data structure.
* @mboxq: pointer to mailbox object.
* @next_fcf_index: pointer to holder of next fcf index.
*
* This routine parses the non-embedded fcf mailbox command by performing the
* necessarily error checking, non-embedded read FCF record mailbox command
* SGE parsing, and endianness swapping.
*
* Returns the pointer to the new FCF record in the non-embedded mailbox
* command DMA memory if successfully, other NULL.
*/
static struct fcf_record *
lpfc_sli4_fcf_rec_mbox_parse(struct lpfc_hba *phba, LPFC_MBOXQ_t *mboxq,
uint16_t *next_fcf_index)
{
void *virt_addr;
struct lpfc_mbx_sge sge;
struct lpfc_mbx_read_fcf_tbl *read_fcf;
uint32_t shdr_status, shdr_add_status, if_type;
union lpfc_sli4_cfg_shdr *shdr;
struct fcf_record *new_fcf_record;
/* Get the first SGE entry from the non-embedded DMA memory. This
* routine only uses a single SGE.
*/
lpfc_sli4_mbx_sge_get(mboxq, 0, &sge);
if (unlikely(!mboxq->sge_array)) {
scsi: lpfc: Add an internal trace log buffer The current logging methods typically end up requesting a reproduction with a different logging level set to figure out what happened. This was mainly by design to not clutter the kernel log messages with things that were typically not interesting and the messages themselves could cause other issues. When looking to make a better system, it was seen that in many cases when more data was wanted was when another message, usually at KERN_ERR level, was logged. And in most cases, what the additional logging that was then enabled was typically. Most of these areas fell into the discovery machine. Based on this summary, the following design has been put in place: The driver will maintain an internal log (256 elements of 256 bytes). The "additional logging" messages that are usually enabled in a reproduction will be changed to now log all the time to the internal log. A new logging level is defined - LOG_TRACE_EVENT. When this level is set (it is not by default) and a message marked as KERN_ERR is logged, all the messages in the internal log will be dumped to the kernel log before the KERN_ERR message is logged. There is a timestamp on each message added to the internal log. However, this timestamp is not converted to wall time when logged. The value of the timestamp is solely to give a crude time reference for the messages. Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-06-30 21:50:00 +00:00
lpfc_printf_log(phba, KERN_ERR, LOG_TRACE_EVENT,
"2524 Failed to get the non-embedded SGE "
"virtual address\n");
return NULL;
}
virt_addr = mboxq->sge_array->addr[0];
shdr = (union lpfc_sli4_cfg_shdr *)virt_addr;
lpfc_sli_pcimem_bcopy(shdr, shdr,
sizeof(union lpfc_sli4_cfg_shdr));
shdr_status = bf_get(lpfc_mbox_hdr_status, &shdr->response);
if_type = bf_get(lpfc_sli_intf_if_type, &phba->sli4_hba.sli_intf);
shdr_add_status = bf_get(lpfc_mbox_hdr_add_status, &shdr->response);
if (shdr_status || shdr_add_status) {
if (shdr_status == STATUS_FCF_TABLE_EMPTY ||
if_type == LPFC_SLI_INTF_IF_TYPE_2)
scsi: lpfc: Add an internal trace log buffer The current logging methods typically end up requesting a reproduction with a different logging level set to figure out what happened. This was mainly by design to not clutter the kernel log messages with things that were typically not interesting and the messages themselves could cause other issues. When looking to make a better system, it was seen that in many cases when more data was wanted was when another message, usually at KERN_ERR level, was logged. And in most cases, what the additional logging that was then enabled was typically. Most of these areas fell into the discovery machine. Based on this summary, the following design has been put in place: The driver will maintain an internal log (256 elements of 256 bytes). The "additional logging" messages that are usually enabled in a reproduction will be changed to now log all the time to the internal log. A new logging level is defined - LOG_TRACE_EVENT. When this level is set (it is not by default) and a message marked as KERN_ERR is logged, all the messages in the internal log will be dumped to the kernel log before the KERN_ERR message is logged. There is a timestamp on each message added to the internal log. However, this timestamp is not converted to wall time when logged. The value of the timestamp is solely to give a crude time reference for the messages. Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-06-30 21:50:00 +00:00
lpfc_printf_log(phba, KERN_ERR,
LOG_TRACE_EVENT,
"2726 READ_FCF_RECORD Indicates empty "
"FCF table.\n");
else
scsi: lpfc: Add an internal trace log buffer The current logging methods typically end up requesting a reproduction with a different logging level set to figure out what happened. This was mainly by design to not clutter the kernel log messages with things that were typically not interesting and the messages themselves could cause other issues. When looking to make a better system, it was seen that in many cases when more data was wanted was when another message, usually at KERN_ERR level, was logged. And in most cases, what the additional logging that was then enabled was typically. Most of these areas fell into the discovery machine. Based on this summary, the following design has been put in place: The driver will maintain an internal log (256 elements of 256 bytes). The "additional logging" messages that are usually enabled in a reproduction will be changed to now log all the time to the internal log. A new logging level is defined - LOG_TRACE_EVENT. When this level is set (it is not by default) and a message marked as KERN_ERR is logged, all the messages in the internal log will be dumped to the kernel log before the KERN_ERR message is logged. There is a timestamp on each message added to the internal log. However, this timestamp is not converted to wall time when logged. The value of the timestamp is solely to give a crude time reference for the messages. Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-06-30 21:50:00 +00:00
lpfc_printf_log(phba, KERN_ERR, LOG_TRACE_EVENT,
"2521 READ_FCF_RECORD mailbox failed "
"with status x%x add_status x%x, "
"mbx\n", shdr_status, shdr_add_status);
return NULL;
}
/* Interpreting the returned information of the FCF record */
read_fcf = (struct lpfc_mbx_read_fcf_tbl *)virt_addr;
lpfc_sli_pcimem_bcopy(read_fcf, read_fcf,
sizeof(struct lpfc_mbx_read_fcf_tbl));
*next_fcf_index = bf_get(lpfc_mbx_read_fcf_tbl_nxt_vindx, read_fcf);
new_fcf_record = (struct fcf_record *)(virt_addr +
sizeof(struct lpfc_mbx_read_fcf_tbl));
lpfc_sli_pcimem_bcopy(new_fcf_record, new_fcf_record,
offsetof(struct fcf_record, vlan_bitmap));
new_fcf_record->word137 = le32_to_cpu(new_fcf_record->word137);
new_fcf_record->word138 = le32_to_cpu(new_fcf_record->word138);
return new_fcf_record;
}
/**
* lpfc_sli4_log_fcf_record_info - Log the information of a fcf record
* @phba: pointer to lpfc hba data structure.
* @fcf_record: pointer to the fcf record.
* @vlan_id: the lowest vlan identifier associated to this fcf record.
* @next_fcf_index: the index to the next fcf record in hba's fcf table.
*
* This routine logs the detailed FCF record if the LOG_FIP loggin is
* enabled.
**/
static void
lpfc_sli4_log_fcf_record_info(struct lpfc_hba *phba,
struct fcf_record *fcf_record,
uint16_t vlan_id,
uint16_t next_fcf_index)
{
lpfc_printf_log(phba, KERN_INFO, LOG_FIP,
"2764 READ_FCF_RECORD:\n"
"\tFCF_Index : x%x\n"
"\tFCF_Avail : x%x\n"
"\tFCF_Valid : x%x\n"
"\tFCF_SOL : x%x\n"
"\tFIP_Priority : x%x\n"
"\tMAC_Provider : x%x\n"
"\tLowest VLANID : x%x\n"
"\tFCF_MAC Addr : x%x:%x:%x:%x:%x:%x\n"
"\tFabric_Name : x%x:%x:%x:%x:%x:%x:%x:%x\n"
"\tSwitch_Name : x%x:%x:%x:%x:%x:%x:%x:%x\n"
"\tNext_FCF_Index: x%x\n",
bf_get(lpfc_fcf_record_fcf_index, fcf_record),
bf_get(lpfc_fcf_record_fcf_avail, fcf_record),
bf_get(lpfc_fcf_record_fcf_valid, fcf_record),
bf_get(lpfc_fcf_record_fcf_sol, fcf_record),
fcf_record->fip_priority,
bf_get(lpfc_fcf_record_mac_addr_prov, fcf_record),
vlan_id,
bf_get(lpfc_fcf_record_mac_0, fcf_record),
bf_get(lpfc_fcf_record_mac_1, fcf_record),
bf_get(lpfc_fcf_record_mac_2, fcf_record),
bf_get(lpfc_fcf_record_mac_3, fcf_record),
bf_get(lpfc_fcf_record_mac_4, fcf_record),
bf_get(lpfc_fcf_record_mac_5, fcf_record),
bf_get(lpfc_fcf_record_fab_name_0, fcf_record),
bf_get(lpfc_fcf_record_fab_name_1, fcf_record),
bf_get(lpfc_fcf_record_fab_name_2, fcf_record),
bf_get(lpfc_fcf_record_fab_name_3, fcf_record),
bf_get(lpfc_fcf_record_fab_name_4, fcf_record),
bf_get(lpfc_fcf_record_fab_name_5, fcf_record),
bf_get(lpfc_fcf_record_fab_name_6, fcf_record),
bf_get(lpfc_fcf_record_fab_name_7, fcf_record),
bf_get(lpfc_fcf_record_switch_name_0, fcf_record),
bf_get(lpfc_fcf_record_switch_name_1, fcf_record),
bf_get(lpfc_fcf_record_switch_name_2, fcf_record),
bf_get(lpfc_fcf_record_switch_name_3, fcf_record),
bf_get(lpfc_fcf_record_switch_name_4, fcf_record),
bf_get(lpfc_fcf_record_switch_name_5, fcf_record),
bf_get(lpfc_fcf_record_switch_name_6, fcf_record),
bf_get(lpfc_fcf_record_switch_name_7, fcf_record),
next_fcf_index);
}
/**
scsi: lpfc: Fix kerneldoc parameter formatting/misnaming/missing issues Fixes the following W=1 kernel build warning(s): drivers/scsi/lpfc/lpfc_hbadisc.c:1209: warning: Function parameter or member 'phba' not described in 'lpfc_sli4_clear_fcf_rr_bmask' drivers/scsi/lpfc/lpfc_hbadisc.c:1309: warning: Function parameter or member 'sw_name' not described in 'lpfc_sw_name_match' drivers/scsi/lpfc/lpfc_hbadisc.c:1309: warning: Excess function parameter 'fab_name' description in 'lpfc_sw_name_match' drivers/scsi/lpfc/lpfc_hbadisc.c:1397: warning: Function parameter or member 'fcf_rec' not described in 'lpfc_copy_fcf_record' drivers/scsi/lpfc/lpfc_hbadisc.c:1397: warning: Excess function parameter 'fcf' description in 'lpfc_copy_fcf_record' drivers/scsi/lpfc/lpfc_hbadisc.c:1956: warning: Cannot understand lpfc_sli4_fcf_record_match - testing new FCF record for matching existing FCF on line 1956 - I thought it was a doc line drivers/scsi/lpfc/lpfc_hbadisc.c:2078: warning: Function parameter or member 'fcf_index' not described in 'lpfc_sli4_fcf_pri_list_del' drivers/scsi/lpfc/lpfc_hbadisc.c:2109: warning: Function parameter or member 'fcf_index' not described in 'lpfc_sli4_set_fcf_flogi_fail' drivers/scsi/lpfc/lpfc_hbadisc.c:2135: warning: Function parameter or member 'fcf_index' not described in 'lpfc_sli4_fcf_pri_list_add' drivers/scsi/lpfc/lpfc_hbadisc.c:2135: warning: Function parameter or member 'new_fcf_record' not described in 'lpfc_sli4_fcf_pri_list_add' Link: https://lore.kernel.org/r/20200723122446.1329773-3-lee.jones@linaro.org Cc: James Smart <james.smart@broadcom.com> Cc: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: Lee Jones <lee.jones@linaro.org> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-07-23 12:24:08 +00:00
* lpfc_sli4_fcf_record_match - testing new FCF record for matching existing FCF
* @phba: pointer to lpfc hba data structure.
* @fcf_rec: pointer to an existing FCF record.
* @new_fcf_record: pointer to a new FCF record.
* @new_vlan_id: vlan id from the new FCF record.
*
* This function performs matching test of a new FCF record against an existing
* FCF record. If the new_vlan_id passed in is LPFC_FCOE_IGNORE_VID, vlan id
* will not be used as part of the FCF record matching criteria.
*
* Returns true if all the fields matching, otherwise returns false.
*/
static bool
lpfc_sli4_fcf_record_match(struct lpfc_hba *phba,
struct lpfc_fcf_rec *fcf_rec,
struct fcf_record *new_fcf_record,
uint16_t new_vlan_id)
{
if (new_vlan_id != LPFC_FCOE_IGNORE_VID)
if (!lpfc_vlan_id_match(fcf_rec->vlan_id, new_vlan_id))
return false;
if (!lpfc_mac_addr_match(fcf_rec->mac_addr, new_fcf_record))
return false;
if (!lpfc_sw_name_match(fcf_rec->switch_name, new_fcf_record))
return false;
if (!lpfc_fab_name_match(fcf_rec->fabric_name, new_fcf_record))
return false;
if (fcf_rec->priority != new_fcf_record->fip_priority)
return false;
return true;
}
/**
* lpfc_sli4_fcf_rr_next_proc - processing next roundrobin fcf
* @vport: Pointer to vport object.
* @fcf_index: index to next fcf.
*
* This function processing the roundrobin fcf failover to next fcf index.
* When this function is invoked, there will be a current fcf registered
* for flogi.
* Return: 0 for continue retrying flogi on currently registered fcf;
* 1 for stop flogi on currently registered fcf;
*/
int lpfc_sli4_fcf_rr_next_proc(struct lpfc_vport *vport, uint16_t fcf_index)
{
struct lpfc_hba *phba = vport->phba;
int rc;
if (fcf_index == LPFC_FCOE_FCF_NEXT_NONE) {
spin_lock_irq(&phba->hbalock);
if (phba->hba_flag & HBA_DEVLOSS_TMO) {
spin_unlock_irq(&phba->hbalock);
lpfc_printf_log(phba, KERN_INFO, LOG_FIP,
"2872 Devloss tmo with no eligible "
"FCF, unregister in-use FCF (x%x) "
"and rescan FCF table\n",
phba->fcf.current_rec.fcf_indx);
lpfc_unregister_fcf_rescan(phba);
goto stop_flogi_current_fcf;
}
/* Mark the end to FLOGI roundrobin failover */
phba->hba_flag &= ~FCF_RR_INPROG;
/* Allow action to new fcf asynchronous event */
phba->fcf.fcf_flag &= ~(FCF_AVAILABLE | FCF_SCAN_DONE);
spin_unlock_irq(&phba->hbalock);
lpfc_printf_log(phba, KERN_INFO, LOG_FIP,
"2865 No FCF available, stop roundrobin FCF "
"failover and change port state:x%x/x%x\n",
phba->pport->port_state, LPFC_VPORT_UNKNOWN);
phba->pport->port_state = LPFC_VPORT_UNKNOWN;
if (!phba->fcf.fcf_redisc_attempted) {
lpfc_unregister_fcf(phba);
rc = lpfc_sli4_redisc_fcf_table(phba);
if (!rc) {
lpfc_printf_log(phba, KERN_INFO, LOG_FIP,
"3195 Rediscover FCF table\n");
phba->fcf.fcf_redisc_attempted = 1;
lpfc_sli4_clear_fcf_rr_bmask(phba);
} else {
lpfc_printf_log(phba, KERN_WARNING, LOG_FIP,
"3196 Rediscover FCF table "
"failed. Status:x%x\n", rc);
}
} else {
lpfc_printf_log(phba, KERN_WARNING, LOG_FIP,
"3197 Already rediscover FCF table "
"attempted. No more retry\n");
}
goto stop_flogi_current_fcf;
} else {
lpfc_printf_log(phba, KERN_INFO, LOG_FIP | LOG_ELS,
"2794 Try FLOGI roundrobin FCF failover to "
"(x%x)\n", fcf_index);
rc = lpfc_sli4_fcf_rr_read_fcf_rec(phba, fcf_index);
if (rc)
lpfc_printf_log(phba, KERN_WARNING, LOG_FIP | LOG_ELS,
"2761 FLOGI roundrobin FCF failover "
"failed (rc:x%x) to read FCF (x%x)\n",
rc, phba->fcf.current_rec.fcf_indx);
else
goto stop_flogi_current_fcf;
}
return 0;
stop_flogi_current_fcf:
lpfc_can_disctmo(vport);
return 1;
}
/**
* lpfc_sli4_fcf_pri_list_del
* @phba: pointer to lpfc hba data structure.
scsi: lpfc: Fix kerneldoc parameter formatting/misnaming/missing issues Fixes the following W=1 kernel build warning(s): drivers/scsi/lpfc/lpfc_hbadisc.c:1209: warning: Function parameter or member 'phba' not described in 'lpfc_sli4_clear_fcf_rr_bmask' drivers/scsi/lpfc/lpfc_hbadisc.c:1309: warning: Function parameter or member 'sw_name' not described in 'lpfc_sw_name_match' drivers/scsi/lpfc/lpfc_hbadisc.c:1309: warning: Excess function parameter 'fab_name' description in 'lpfc_sw_name_match' drivers/scsi/lpfc/lpfc_hbadisc.c:1397: warning: Function parameter or member 'fcf_rec' not described in 'lpfc_copy_fcf_record' drivers/scsi/lpfc/lpfc_hbadisc.c:1397: warning: Excess function parameter 'fcf' description in 'lpfc_copy_fcf_record' drivers/scsi/lpfc/lpfc_hbadisc.c:1956: warning: Cannot understand lpfc_sli4_fcf_record_match - testing new FCF record for matching existing FCF on line 1956 - I thought it was a doc line drivers/scsi/lpfc/lpfc_hbadisc.c:2078: warning: Function parameter or member 'fcf_index' not described in 'lpfc_sli4_fcf_pri_list_del' drivers/scsi/lpfc/lpfc_hbadisc.c:2109: warning: Function parameter or member 'fcf_index' not described in 'lpfc_sli4_set_fcf_flogi_fail' drivers/scsi/lpfc/lpfc_hbadisc.c:2135: warning: Function parameter or member 'fcf_index' not described in 'lpfc_sli4_fcf_pri_list_add' drivers/scsi/lpfc/lpfc_hbadisc.c:2135: warning: Function parameter or member 'new_fcf_record' not described in 'lpfc_sli4_fcf_pri_list_add' Link: https://lore.kernel.org/r/20200723122446.1329773-3-lee.jones@linaro.org Cc: James Smart <james.smart@broadcom.com> Cc: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: Lee Jones <lee.jones@linaro.org> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-07-23 12:24:08 +00:00
* @fcf_index: the index of the fcf record to delete
* This routine checks the on list flag of the fcf_index to be deleted.
* If it is one the list then it is removed from the list, and the flag
* is cleared. This routine grab the hbalock before removing the fcf
* record from the list.
**/
static void lpfc_sli4_fcf_pri_list_del(struct lpfc_hba *phba,
uint16_t fcf_index)
{
struct lpfc_fcf_pri *new_fcf_pri;
new_fcf_pri = &phba->fcf.fcf_pri[fcf_index];
lpfc_printf_log(phba, KERN_INFO, LOG_FIP,
"3058 deleting idx x%x pri x%x flg x%x\n",
fcf_index, new_fcf_pri->fcf_rec.priority,
new_fcf_pri->fcf_rec.flag);
spin_lock_irq(&phba->hbalock);
if (new_fcf_pri->fcf_rec.flag & LPFC_FCF_ON_PRI_LIST) {
if (phba->fcf.current_rec.priority ==
new_fcf_pri->fcf_rec.priority)
phba->fcf.eligible_fcf_cnt--;
list_del_init(&new_fcf_pri->list);
new_fcf_pri->fcf_rec.flag &= ~LPFC_FCF_ON_PRI_LIST;
}
spin_unlock_irq(&phba->hbalock);
}
/**
* lpfc_sli4_set_fcf_flogi_fail
* @phba: pointer to lpfc hba data structure.
scsi: lpfc: Fix kerneldoc parameter formatting/misnaming/missing issues Fixes the following W=1 kernel build warning(s): drivers/scsi/lpfc/lpfc_hbadisc.c:1209: warning: Function parameter or member 'phba' not described in 'lpfc_sli4_clear_fcf_rr_bmask' drivers/scsi/lpfc/lpfc_hbadisc.c:1309: warning: Function parameter or member 'sw_name' not described in 'lpfc_sw_name_match' drivers/scsi/lpfc/lpfc_hbadisc.c:1309: warning: Excess function parameter 'fab_name' description in 'lpfc_sw_name_match' drivers/scsi/lpfc/lpfc_hbadisc.c:1397: warning: Function parameter or member 'fcf_rec' not described in 'lpfc_copy_fcf_record' drivers/scsi/lpfc/lpfc_hbadisc.c:1397: warning: Excess function parameter 'fcf' description in 'lpfc_copy_fcf_record' drivers/scsi/lpfc/lpfc_hbadisc.c:1956: warning: Cannot understand lpfc_sli4_fcf_record_match - testing new FCF record for matching existing FCF on line 1956 - I thought it was a doc line drivers/scsi/lpfc/lpfc_hbadisc.c:2078: warning: Function parameter or member 'fcf_index' not described in 'lpfc_sli4_fcf_pri_list_del' drivers/scsi/lpfc/lpfc_hbadisc.c:2109: warning: Function parameter or member 'fcf_index' not described in 'lpfc_sli4_set_fcf_flogi_fail' drivers/scsi/lpfc/lpfc_hbadisc.c:2135: warning: Function parameter or member 'fcf_index' not described in 'lpfc_sli4_fcf_pri_list_add' drivers/scsi/lpfc/lpfc_hbadisc.c:2135: warning: Function parameter or member 'new_fcf_record' not described in 'lpfc_sli4_fcf_pri_list_add' Link: https://lore.kernel.org/r/20200723122446.1329773-3-lee.jones@linaro.org Cc: James Smart <james.smart@broadcom.com> Cc: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: Lee Jones <lee.jones@linaro.org> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-07-23 12:24:08 +00:00
* @fcf_index: the index of the fcf record to update
* This routine acquires the hbalock and then set the LPFC_FCF_FLOGI_FAILED
* flag so the round robin selection for the particular priority level
* will try a different fcf record that does not have this bit set.
* If the fcf record is re-read for any reason this flag is cleared brfore
* adding it to the priority list.
**/
void
lpfc_sli4_set_fcf_flogi_fail(struct lpfc_hba *phba, uint16_t fcf_index)
{
struct lpfc_fcf_pri *new_fcf_pri;
new_fcf_pri = &phba->fcf.fcf_pri[fcf_index];
spin_lock_irq(&phba->hbalock);
new_fcf_pri->fcf_rec.flag |= LPFC_FCF_FLOGI_FAILED;
spin_unlock_irq(&phba->hbalock);
}
/**
* lpfc_sli4_fcf_pri_list_add
* @phba: pointer to lpfc hba data structure.
scsi: lpfc: Fix kerneldoc parameter formatting/misnaming/missing issues Fixes the following W=1 kernel build warning(s): drivers/scsi/lpfc/lpfc_hbadisc.c:1209: warning: Function parameter or member 'phba' not described in 'lpfc_sli4_clear_fcf_rr_bmask' drivers/scsi/lpfc/lpfc_hbadisc.c:1309: warning: Function parameter or member 'sw_name' not described in 'lpfc_sw_name_match' drivers/scsi/lpfc/lpfc_hbadisc.c:1309: warning: Excess function parameter 'fab_name' description in 'lpfc_sw_name_match' drivers/scsi/lpfc/lpfc_hbadisc.c:1397: warning: Function parameter or member 'fcf_rec' not described in 'lpfc_copy_fcf_record' drivers/scsi/lpfc/lpfc_hbadisc.c:1397: warning: Excess function parameter 'fcf' description in 'lpfc_copy_fcf_record' drivers/scsi/lpfc/lpfc_hbadisc.c:1956: warning: Cannot understand lpfc_sli4_fcf_record_match - testing new FCF record for matching existing FCF on line 1956 - I thought it was a doc line drivers/scsi/lpfc/lpfc_hbadisc.c:2078: warning: Function parameter or member 'fcf_index' not described in 'lpfc_sli4_fcf_pri_list_del' drivers/scsi/lpfc/lpfc_hbadisc.c:2109: warning: Function parameter or member 'fcf_index' not described in 'lpfc_sli4_set_fcf_flogi_fail' drivers/scsi/lpfc/lpfc_hbadisc.c:2135: warning: Function parameter or member 'fcf_index' not described in 'lpfc_sli4_fcf_pri_list_add' drivers/scsi/lpfc/lpfc_hbadisc.c:2135: warning: Function parameter or member 'new_fcf_record' not described in 'lpfc_sli4_fcf_pri_list_add' Link: https://lore.kernel.org/r/20200723122446.1329773-3-lee.jones@linaro.org Cc: James Smart <james.smart@broadcom.com> Cc: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: Lee Jones <lee.jones@linaro.org> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-07-23 12:24:08 +00:00
* @fcf_index: the index of the fcf record to add
* @new_fcf_record: pointer to a new FCF record.
* This routine checks the priority of the fcf_index to be added.
* If it is a lower priority than the current head of the fcf_pri list
* then it is added to the list in the right order.
* If it is the same priority as the current head of the list then it
* is added to the head of the list and its bit in the rr_bmask is set.
* If the fcf_index to be added is of a higher priority than the current
* head of the list then the rr_bmask is cleared, its bit is set in the
* rr_bmask and it is added to the head of the list.
* returns:
* 0=success 1=failure
**/
static int lpfc_sli4_fcf_pri_list_add(struct lpfc_hba *phba,
uint16_t fcf_index,
struct fcf_record *new_fcf_record)
{
uint16_t current_fcf_pri;
uint16_t last_index;
struct lpfc_fcf_pri *fcf_pri;
struct lpfc_fcf_pri *next_fcf_pri;
struct lpfc_fcf_pri *new_fcf_pri;
int ret;
new_fcf_pri = &phba->fcf.fcf_pri[fcf_index];
lpfc_printf_log(phba, KERN_INFO, LOG_FIP,
"3059 adding idx x%x pri x%x flg x%x\n",
fcf_index, new_fcf_record->fip_priority,
new_fcf_pri->fcf_rec.flag);
spin_lock_irq(&phba->hbalock);
if (new_fcf_pri->fcf_rec.flag & LPFC_FCF_ON_PRI_LIST)
list_del_init(&new_fcf_pri->list);
new_fcf_pri->fcf_rec.fcf_index = fcf_index;
new_fcf_pri->fcf_rec.priority = new_fcf_record->fip_priority;
if (list_empty(&phba->fcf.fcf_pri_list)) {
list_add(&new_fcf_pri->list, &phba->fcf.fcf_pri_list);
ret = lpfc_sli4_fcf_rr_index_set(phba,
new_fcf_pri->fcf_rec.fcf_index);
goto out;
}
last_index = find_first_bit(phba->fcf.fcf_rr_bmask,
LPFC_SLI4_FCF_TBL_INDX_MAX);
if (last_index >= LPFC_SLI4_FCF_TBL_INDX_MAX) {
ret = 0; /* Empty rr list */
goto out;
}
current_fcf_pri = phba->fcf.fcf_pri[last_index].fcf_rec.priority;
if (new_fcf_pri->fcf_rec.priority <= current_fcf_pri) {
list_add(&new_fcf_pri->list, &phba->fcf.fcf_pri_list);
if (new_fcf_pri->fcf_rec.priority < current_fcf_pri) {
memset(phba->fcf.fcf_rr_bmask, 0,
sizeof(*phba->fcf.fcf_rr_bmask));
/* fcfs_at_this_priority_level = 1; */
phba->fcf.eligible_fcf_cnt = 1;
} else
/* fcfs_at_this_priority_level++; */
phba->fcf.eligible_fcf_cnt++;
ret = lpfc_sli4_fcf_rr_index_set(phba,
new_fcf_pri->fcf_rec.fcf_index);
goto out;
}
list_for_each_entry_safe(fcf_pri, next_fcf_pri,
&phba->fcf.fcf_pri_list, list) {
if (new_fcf_pri->fcf_rec.priority <=
fcf_pri->fcf_rec.priority) {
if (fcf_pri->list.prev == &phba->fcf.fcf_pri_list)
list_add(&new_fcf_pri->list,
&phba->fcf.fcf_pri_list);
else
list_add(&new_fcf_pri->list,
&((struct lpfc_fcf_pri *)
fcf_pri->list.prev)->list);
ret = 0;
goto out;
} else if (fcf_pri->list.next == &phba->fcf.fcf_pri_list
|| new_fcf_pri->fcf_rec.priority <
next_fcf_pri->fcf_rec.priority) {
list_add(&new_fcf_pri->list, &fcf_pri->list);
ret = 0;
goto out;
}
if (new_fcf_pri->fcf_rec.priority > fcf_pri->fcf_rec.priority)
continue;
}
ret = 1;
out:
/* we use = instead of |= to clear the FLOGI_FAILED flag. */
new_fcf_pri->fcf_rec.flag = LPFC_FCF_ON_PRI_LIST;
spin_unlock_irq(&phba->hbalock);
return ret;
}
/**
* lpfc_mbx_cmpl_fcf_scan_read_fcf_rec - fcf scan read_fcf mbox cmpl handler.
* @phba: pointer to lpfc hba data structure.
* @mboxq: pointer to mailbox object.
*
* This function iterates through all the fcf records available in
* HBA and chooses the optimal FCF record for discovery. After finding
* the FCF for discovery it registers the FCF record and kicks start
* discovery.
* If FCF_IN_USE flag is set in currently used FCF, the routine tries to
* use an FCF record which matches fabric name and mac address of the
* currently used FCF record.
* If the driver supports only one FCF, it will try to use the FCF record
* used by BOOT_BIOS.
*/
void
lpfc_mbx_cmpl_fcf_scan_read_fcf_rec(struct lpfc_hba *phba, LPFC_MBOXQ_t *mboxq)
{
struct fcf_record *new_fcf_record;
uint32_t boot_flag, addr_mode;
uint16_t fcf_index, next_fcf_index;
struct lpfc_fcf_rec *fcf_rec = NULL;
uint16_t vlan_id = LPFC_FCOE_NULL_VID;
bool select_new_fcf;
int rc;
/* If there is pending FCoE event restart FCF table scan */
if (lpfc_check_pending_fcoe_event(phba, LPFC_SKIP_UNREG_FCF)) {
lpfc_sli4_mbox_cmd_free(phba, mboxq);
return;
}
/* Parse the FCF record from the non-embedded mailbox command */
new_fcf_record = lpfc_sli4_fcf_rec_mbox_parse(phba, mboxq,
&next_fcf_index);
if (!new_fcf_record) {
scsi: lpfc: Add an internal trace log buffer The current logging methods typically end up requesting a reproduction with a different logging level set to figure out what happened. This was mainly by design to not clutter the kernel log messages with things that were typically not interesting and the messages themselves could cause other issues. When looking to make a better system, it was seen that in many cases when more data was wanted was when another message, usually at KERN_ERR level, was logged. And in most cases, what the additional logging that was then enabled was typically. Most of these areas fell into the discovery machine. Based on this summary, the following design has been put in place: The driver will maintain an internal log (256 elements of 256 bytes). The "additional logging" messages that are usually enabled in a reproduction will be changed to now log all the time to the internal log. A new logging level is defined - LOG_TRACE_EVENT. When this level is set (it is not by default) and a message marked as KERN_ERR is logged, all the messages in the internal log will be dumped to the kernel log before the KERN_ERR message is logged. There is a timestamp on each message added to the internal log. However, this timestamp is not converted to wall time when logged. The value of the timestamp is solely to give a crude time reference for the messages. Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-06-30 21:50:00 +00:00
lpfc_printf_log(phba, KERN_ERR, LOG_TRACE_EVENT,
"2765 Mailbox command READ_FCF_RECORD "
"failed to retrieve a FCF record.\n");
/* Let next new FCF event trigger fast failover */
spin_lock_irq(&phba->hbalock);
phba->hba_flag &= ~FCF_TS_INPROG;
spin_unlock_irq(&phba->hbalock);
lpfc_sli4_mbox_cmd_free(phba, mboxq);
return;
}
/* Check the FCF record against the connection list */
rc = lpfc_match_fcf_conn_list(phba, new_fcf_record, &boot_flag,
&addr_mode, &vlan_id);
/* Log the FCF record information if turned on */
lpfc_sli4_log_fcf_record_info(phba, new_fcf_record, vlan_id,
next_fcf_index);
/*
* If the fcf record does not match with connect list entries
* read the next entry; otherwise, this is an eligible FCF
* record for roundrobin FCF failover.
*/
if (!rc) {
lpfc_sli4_fcf_pri_list_del(phba,
bf_get(lpfc_fcf_record_fcf_index,
new_fcf_record));
lpfc_printf_log(phba, KERN_WARNING, LOG_FIP,
"2781 FCF (x%x) failed connection "
"list check: (x%x/x%x/%x)\n",
bf_get(lpfc_fcf_record_fcf_index,
new_fcf_record),
bf_get(lpfc_fcf_record_fcf_avail,
new_fcf_record),
bf_get(lpfc_fcf_record_fcf_valid,
new_fcf_record),
bf_get(lpfc_fcf_record_fcf_sol,
new_fcf_record));
if ((phba->fcf.fcf_flag & FCF_IN_USE) &&
lpfc_sli4_fcf_record_match(phba, &phba->fcf.current_rec,
new_fcf_record, LPFC_FCOE_IGNORE_VID)) {
if (bf_get(lpfc_fcf_record_fcf_index, new_fcf_record) !=
phba->fcf.current_rec.fcf_indx) {
scsi: lpfc: Add an internal trace log buffer The current logging methods typically end up requesting a reproduction with a different logging level set to figure out what happened. This was mainly by design to not clutter the kernel log messages with things that were typically not interesting and the messages themselves could cause other issues. When looking to make a better system, it was seen that in many cases when more data was wanted was when another message, usually at KERN_ERR level, was logged. And in most cases, what the additional logging that was then enabled was typically. Most of these areas fell into the discovery machine. Based on this summary, the following design has been put in place: The driver will maintain an internal log (256 elements of 256 bytes). The "additional logging" messages that are usually enabled in a reproduction will be changed to now log all the time to the internal log. A new logging level is defined - LOG_TRACE_EVENT. When this level is set (it is not by default) and a message marked as KERN_ERR is logged, all the messages in the internal log will be dumped to the kernel log before the KERN_ERR message is logged. There is a timestamp on each message added to the internal log. However, this timestamp is not converted to wall time when logged. The value of the timestamp is solely to give a crude time reference for the messages. Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-06-30 21:50:00 +00:00
lpfc_printf_log(phba, KERN_ERR,
LOG_TRACE_EVENT,
"2862 FCF (x%x) matches property "
"of in-use FCF (x%x)\n",
bf_get(lpfc_fcf_record_fcf_index,
new_fcf_record),
phba->fcf.current_rec.fcf_indx);
goto read_next_fcf;
}
/*
* In case the current in-use FCF record becomes
* invalid/unavailable during FCF discovery that
* was not triggered by fast FCF failover process,
* treat it as fast FCF failover.
*/
if (!(phba->fcf.fcf_flag & FCF_REDISC_PEND) &&
!(phba->fcf.fcf_flag & FCF_REDISC_FOV)) {
lpfc_printf_log(phba, KERN_WARNING, LOG_FIP,
"2835 Invalid in-use FCF "
"(x%x), enter FCF failover "
"table scan.\n",
phba->fcf.current_rec.fcf_indx);
spin_lock_irq(&phba->hbalock);
phba->fcf.fcf_flag |= FCF_REDISC_FOV;
spin_unlock_irq(&phba->hbalock);
lpfc_sli4_mbox_cmd_free(phba, mboxq);
lpfc_sli4_fcf_scan_read_fcf_rec(phba,
LPFC_FCOE_FCF_GET_FIRST);
return;
}
}
goto read_next_fcf;
} else {
fcf_index = bf_get(lpfc_fcf_record_fcf_index, new_fcf_record);
rc = lpfc_sli4_fcf_pri_list_add(phba, fcf_index,
new_fcf_record);
if (rc)
goto read_next_fcf;
}
/*
* If this is not the first FCF discovery of the HBA, use last
* FCF record for the discovery. The condition that a rescan
* matches the in-use FCF record: fabric name, switch name, mac
* address, and vlan_id.
*/
spin_lock_irq(&phba->hbalock);
if (phba->fcf.fcf_flag & FCF_IN_USE) {
if (phba->cfg_fcf_failover_policy == LPFC_FCF_FOV &&
lpfc_sli4_fcf_record_match(phba, &phba->fcf.current_rec,
new_fcf_record, vlan_id)) {
if (bf_get(lpfc_fcf_record_fcf_index, new_fcf_record) ==
phba->fcf.current_rec.fcf_indx) {
phba->fcf.fcf_flag |= FCF_AVAILABLE;
if (phba->fcf.fcf_flag & FCF_REDISC_PEND)
/* Stop FCF redisc wait timer */
__lpfc_sli4_stop_fcf_redisc_wait_timer(
phba);
else if (phba->fcf.fcf_flag & FCF_REDISC_FOV)
/* Fast failover, mark completed */
phba->fcf.fcf_flag &= ~FCF_REDISC_FOV;
spin_unlock_irq(&phba->hbalock);
lpfc_printf_log(phba, KERN_INFO, LOG_FIP,
"2836 New FCF matches in-use "
"FCF (x%x), port_state:x%x, "
"fc_flag:x%x\n",
phba->fcf.current_rec.fcf_indx,
phba->pport->port_state,
phba->pport->fc_flag);
goto out;
} else
scsi: lpfc: Add an internal trace log buffer The current logging methods typically end up requesting a reproduction with a different logging level set to figure out what happened. This was mainly by design to not clutter the kernel log messages with things that were typically not interesting and the messages themselves could cause other issues. When looking to make a better system, it was seen that in many cases when more data was wanted was when another message, usually at KERN_ERR level, was logged. And in most cases, what the additional logging that was then enabled was typically. Most of these areas fell into the discovery machine. Based on this summary, the following design has been put in place: The driver will maintain an internal log (256 elements of 256 bytes). The "additional logging" messages that are usually enabled in a reproduction will be changed to now log all the time to the internal log. A new logging level is defined - LOG_TRACE_EVENT. When this level is set (it is not by default) and a message marked as KERN_ERR is logged, all the messages in the internal log will be dumped to the kernel log before the KERN_ERR message is logged. There is a timestamp on each message added to the internal log. However, this timestamp is not converted to wall time when logged. The value of the timestamp is solely to give a crude time reference for the messages. Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-06-30 21:50:00 +00:00
lpfc_printf_log(phba, KERN_ERR, LOG_TRACE_EVENT,
"2863 New FCF (x%x) matches "
"property of in-use FCF (x%x)\n",
bf_get(lpfc_fcf_record_fcf_index,
new_fcf_record),
phba->fcf.current_rec.fcf_indx);
}
/*
* Read next FCF record from HBA searching for the matching
* with in-use record only if not during the fast failover
* period. In case of fast failover period, it shall try to
* determine whether the FCF record just read should be the
* next candidate.
*/
if (!(phba->fcf.fcf_flag & FCF_REDISC_FOV)) {
spin_unlock_irq(&phba->hbalock);
goto read_next_fcf;
}
}
/*
* Update on failover FCF record only if it's in FCF fast-failover
* period; otherwise, update on current FCF record.
*/
if (phba->fcf.fcf_flag & FCF_REDISC_FOV)
fcf_rec = &phba->fcf.failover_rec;
else
fcf_rec = &phba->fcf.current_rec;
if (phba->fcf.fcf_flag & FCF_AVAILABLE) {
/*
* If the driver FCF record does not have boot flag
* set and new hba fcf record has boot flag set, use
* the new hba fcf record.
*/
if (boot_flag && !(fcf_rec->flag & BOOT_ENABLE)) {
/* Choose this FCF record */
lpfc_printf_log(phba, KERN_INFO, LOG_FIP,
"2837 Update current FCF record "
"(x%x) with new FCF record (x%x)\n",
fcf_rec->fcf_indx,
bf_get(lpfc_fcf_record_fcf_index,
new_fcf_record));
__lpfc_update_fcf_record(phba, fcf_rec, new_fcf_record,
addr_mode, vlan_id, BOOT_ENABLE);
spin_unlock_irq(&phba->hbalock);
goto read_next_fcf;
}
/*
* If the driver FCF record has boot flag set and the
* new hba FCF record does not have boot flag, read
* the next FCF record.
*/
if (!boot_flag && (fcf_rec->flag & BOOT_ENABLE)) {
spin_unlock_irq(&phba->hbalock);
goto read_next_fcf;
}
/*
* If the new hba FCF record has lower priority value
* than the driver FCF record, use the new record.
*/
if (new_fcf_record->fip_priority < fcf_rec->priority) {
/* Choose the new FCF record with lower priority */
lpfc_printf_log(phba, KERN_INFO, LOG_FIP,
"2838 Update current FCF record "
"(x%x) with new FCF record (x%x)\n",
fcf_rec->fcf_indx,
bf_get(lpfc_fcf_record_fcf_index,
new_fcf_record));
__lpfc_update_fcf_record(phba, fcf_rec, new_fcf_record,
addr_mode, vlan_id, 0);
/* Reset running random FCF selection count */
phba->fcf.eligible_fcf_cnt = 1;
} else if (new_fcf_record->fip_priority == fcf_rec->priority) {
/* Update running random FCF selection count */
phba->fcf.eligible_fcf_cnt++;
select_new_fcf = lpfc_sli4_new_fcf_random_select(phba,
phba->fcf.eligible_fcf_cnt);
if (select_new_fcf) {
lpfc_printf_log(phba, KERN_INFO, LOG_FIP,
"2839 Update current FCF record "
"(x%x) with new FCF record (x%x)\n",
fcf_rec->fcf_indx,
bf_get(lpfc_fcf_record_fcf_index,
new_fcf_record));
/* Choose the new FCF by random selection */
__lpfc_update_fcf_record(phba, fcf_rec,
new_fcf_record,
addr_mode, vlan_id, 0);
}
}
spin_unlock_irq(&phba->hbalock);
goto read_next_fcf;
}
/*
* This is the first suitable FCF record, choose this record for
* initial best-fit FCF.
*/
if (fcf_rec) {
lpfc_printf_log(phba, KERN_INFO, LOG_FIP,
"2840 Update initial FCF candidate "
"with FCF (x%x)\n",
bf_get(lpfc_fcf_record_fcf_index,
new_fcf_record));
__lpfc_update_fcf_record(phba, fcf_rec, new_fcf_record,
addr_mode, vlan_id, (boot_flag ?
BOOT_ENABLE : 0));
phba->fcf.fcf_flag |= FCF_AVAILABLE;
/* Setup initial running random FCF selection count */
phba->fcf.eligible_fcf_cnt = 1;
}
spin_unlock_irq(&phba->hbalock);
goto read_next_fcf;
read_next_fcf:
lpfc_sli4_mbox_cmd_free(phba, mboxq);
if (next_fcf_index == LPFC_FCOE_FCF_NEXT_NONE || next_fcf_index == 0) {
if (phba->fcf.fcf_flag & FCF_REDISC_FOV) {
/*
* Case of FCF fast failover scan
*/
/*
* It has not found any suitable FCF record, cancel
* FCF scan inprogress, and do nothing
*/
if (!(phba->fcf.failover_rec.flag & RECORD_VALID)) {
lpfc_printf_log(phba, KERN_WARNING, LOG_FIP,
"2782 No suitable FCF found: "
"(x%x/x%x)\n",
phba->fcoe_eventtag_at_fcf_scan,
bf_get(lpfc_fcf_record_fcf_index,
new_fcf_record));
spin_lock_irq(&phba->hbalock);
if (phba->hba_flag & HBA_DEVLOSS_TMO) {
phba->hba_flag &= ~FCF_TS_INPROG;
spin_unlock_irq(&phba->hbalock);
/* Unregister in-use FCF and rescan */
lpfc_printf_log(phba, KERN_INFO,
LOG_FIP,
"2864 On devloss tmo "
"unreg in-use FCF and "
"rescan FCF table\n");
lpfc_unregister_fcf_rescan(phba);
return;
}
/*
* Let next new FCF event trigger fast failover
*/
phba->hba_flag &= ~FCF_TS_INPROG;
spin_unlock_irq(&phba->hbalock);
return;
}
/*
* It has found a suitable FCF record that is not
* the same as in-use FCF record, unregister the
* in-use FCF record, replace the in-use FCF record
* with the new FCF record, mark FCF fast failover
* completed, and then start register the new FCF
* record.
*/
/* Unregister the current in-use FCF record */
lpfc_unregister_fcf(phba);
/* Replace in-use record with the new record */
lpfc_printf_log(phba, KERN_INFO, LOG_FIP,
"2842 Replace in-use FCF (x%x) "
"with failover FCF (x%x)\n",
phba->fcf.current_rec.fcf_indx,
phba->fcf.failover_rec.fcf_indx);
memcpy(&phba->fcf.current_rec,
&phba->fcf.failover_rec,
sizeof(struct lpfc_fcf_rec));
/*
* Mark the fast FCF failover rediscovery completed
* and the start of the first round of the roundrobin
* FCF failover.
*/
spin_lock_irq(&phba->hbalock);
phba->fcf.fcf_flag &= ~FCF_REDISC_FOV;
spin_unlock_irq(&phba->hbalock);
/* Register to the new FCF record */
lpfc_register_fcf(phba);
} else {
/*
* In case of transaction period to fast FCF failover,
* do nothing when search to the end of the FCF table.
*/
if ((phba->fcf.fcf_flag & FCF_REDISC_EVT) ||
(phba->fcf.fcf_flag & FCF_REDISC_PEND))
return;
if (phba->cfg_fcf_failover_policy == LPFC_FCF_FOV &&
phba->fcf.fcf_flag & FCF_IN_USE) {
/*
* In case the current in-use FCF record no
* longer existed during FCF discovery that
* was not triggered by fast FCF failover
* process, treat it as fast FCF failover.
*/
lpfc_printf_log(phba, KERN_INFO, LOG_FIP,
"2841 In-use FCF record (x%x) "
"not reported, entering fast "
"FCF failover mode scanning.\n",
phba->fcf.current_rec.fcf_indx);
spin_lock_irq(&phba->hbalock);
phba->fcf.fcf_flag |= FCF_REDISC_FOV;
spin_unlock_irq(&phba->hbalock);
lpfc_sli4_fcf_scan_read_fcf_rec(phba,
LPFC_FCOE_FCF_GET_FIRST);
return;
}
/* Register to the new FCF record */
lpfc_register_fcf(phba);
}
} else
lpfc_sli4_fcf_scan_read_fcf_rec(phba, next_fcf_index);
return;
out:
lpfc_sli4_mbox_cmd_free(phba, mboxq);
lpfc_register_fcf(phba);
return;
}
/**
* lpfc_mbx_cmpl_fcf_rr_read_fcf_rec - fcf roundrobin read_fcf mbox cmpl hdler
* @phba: pointer to lpfc hba data structure.
* @mboxq: pointer to mailbox object.
*
* This is the callback function for FLOGI failure roundrobin FCF failover
* read FCF record mailbox command from the eligible FCF record bmask for
* performing the failover. If the FCF read back is not valid/available, it
* fails through to retrying FLOGI to the currently registered FCF again.
* Otherwise, if the FCF read back is valid and available, it will set the
* newly read FCF record to the failover FCF record, unregister currently
* registered FCF record, copy the failover FCF record to the current
* FCF record, and then register the current FCF record before proceeding
* to trying FLOGI on the new failover FCF.
*/
void
lpfc_mbx_cmpl_fcf_rr_read_fcf_rec(struct lpfc_hba *phba, LPFC_MBOXQ_t *mboxq)
{
struct fcf_record *new_fcf_record;
uint32_t boot_flag, addr_mode;
uint16_t next_fcf_index, fcf_index;
uint16_t current_fcf_index;
scsi: lpfc: Fix various issues reported by tools This patch fixes below Smatch reported issues: 1. lpfc_hbadisc.c:3020 lpfc_mbx_cmpl_fcf_rr_read_fcf_rec() error: uninitialized symbol 'vlan_id'. 2. lpfc_hbadisc.c:3121 lpfc_mbx_cmpl_read_fcf_rec() error: uninitialized symbol 'vlan_id'. 3. lpfc_init.c:335 lpfc_dump_wakeup_param_cmpl() warn: always true condition '(prg->dist < 4) => (0-3 < 4)' 4. lpfc_init.c:2419 lpfc_parse_vpd() warn: inconsistent indenting. 5. lpfc_init.c:13248 lpfc_sli4_enable_msi() warn: 'phba->pcidev->irq' 2147483648 can't fit into 65535 'eqhdl->irq' 6. lpfc_debugfs.c:5300 lpfc_idiag_extacc_avail_get() error: uninitialized symbol 'ext_cnt' 7. lpfc_debugfs.c:5300 lpfc_idiag_extacc_avail_get() error: uninitialized symbol 'ext_size' 8. lpfc_vmid.c:248 lpfc_vmid_get_appid() warn: sleeping in atomic context. 9. lpfc_init.c:8342 lpfc_sli4_driver_resource_setup() warn: missing error code 'rc'. 10. lpfc_init.c:13573 lpfc_sli4_hba_unset() warn: variable dereferenced before check 'phba->pport' (see line 13546) 11. lpfc_auth.c:1923 lpfc_auth_handle_dhchap_reply() error: double free of 'hash_value' Fixes: 1. Initialize vlan_id to LPFC_FCOE_NULL_VID. 2. Initialize vlan_id to LPFC_FCOE_NULL_VID. 3. prg->dist is a 2 bit field. Its value can only be between 0-3. Remove redundent check 'if (prg->dist < 4)'. 4. Fix inconsistent indenting. Moved logic into helper function lpfc_fill_vpd(). 5. Define 'eqhdl->irq' as int value as pci_irq_vector() returns int. Also, check for return value of pci_irq_vector() and log message in case of failure. 6. Initialize 'ext_cnt' to 0. 7. Initialize 'ext_size' to 0. 8. Use alloc_percpu_gfp() with GFP_ATOMIC flag. 9. 'rc' was not updated when dma_pool_create() fails. Update 'rc = -ENOMEM' when dma_pool_create() fails before calling goto statement. 10. Add check for 'phba->pport' in lpfc_cpuhp_remove(). 11. Initialize 'hash_value' to NULL, same like 'aug_chal' variable. Link: https://lore.kernel.org/r/20220911221505.117655-13-jsmart2021@gmail.com Co-developed-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2022-09-11 22:15:04 +00:00
uint16_t vlan_id = LPFC_FCOE_NULL_VID;
int rc;
/* If link state is not up, stop the roundrobin failover process */
if (phba->link_state < LPFC_LINK_UP) {
spin_lock_irq(&phba->hbalock);
phba->fcf.fcf_flag &= ~FCF_DISCOVERY;
phba->hba_flag &= ~FCF_RR_INPROG;
spin_unlock_irq(&phba->hbalock);
goto out;
}
/* Parse the FCF record from the non-embedded mailbox command */
new_fcf_record = lpfc_sli4_fcf_rec_mbox_parse(phba, mboxq,
&next_fcf_index);
if (!new_fcf_record) {
lpfc_printf_log(phba, KERN_WARNING, LOG_FIP,
"2766 Mailbox command READ_FCF_RECORD "
"failed to retrieve a FCF record. "
"hba_flg x%x fcf_flg x%x\n", phba->hba_flag,
phba->fcf.fcf_flag);
lpfc_unregister_fcf_rescan(phba);
goto out;
}
/* Get the needed parameters from FCF record */
rc = lpfc_match_fcf_conn_list(phba, new_fcf_record, &boot_flag,
&addr_mode, &vlan_id);
/* Log the FCF record information if turned on */
lpfc_sli4_log_fcf_record_info(phba, new_fcf_record, vlan_id,
next_fcf_index);
fcf_index = bf_get(lpfc_fcf_record_fcf_index, new_fcf_record);
if (!rc) {
lpfc_printf_log(phba, KERN_INFO, LOG_FIP,
"2848 Remove ineligible FCF (x%x) from "
"from roundrobin bmask\n", fcf_index);
/* Clear roundrobin bmask bit for ineligible FCF */
lpfc_sli4_fcf_rr_index_clear(phba, fcf_index);
/* Perform next round of roundrobin FCF failover */
fcf_index = lpfc_sli4_fcf_rr_next_index_get(phba);
rc = lpfc_sli4_fcf_rr_next_proc(phba->pport, fcf_index);
if (rc)
goto out;
goto error_out;
}
if (fcf_index == phba->fcf.current_rec.fcf_indx) {
lpfc_printf_log(phba, KERN_INFO, LOG_FIP,
"2760 Perform FLOGI roundrobin FCF failover: "
"FCF (x%x) back to FCF (x%x)\n",
phba->fcf.current_rec.fcf_indx, fcf_index);
/* Wait 500 ms before retrying FLOGI to current FCF */
msleep(500);
2010-11-21 04:11:48 +00:00
lpfc_issue_init_vfi(phba->pport);
goto out;
}
/* Upload new FCF record to the failover FCF record */
lpfc_printf_log(phba, KERN_INFO, LOG_FIP,
"2834 Update current FCF (x%x) with new FCF (x%x)\n",
phba->fcf.failover_rec.fcf_indx, fcf_index);
spin_lock_irq(&phba->hbalock);
__lpfc_update_fcf_record(phba, &phba->fcf.failover_rec,
new_fcf_record, addr_mode, vlan_id,
(boot_flag ? BOOT_ENABLE : 0));
spin_unlock_irq(&phba->hbalock);
current_fcf_index = phba->fcf.current_rec.fcf_indx;
/* Unregister the current in-use FCF record */
lpfc_unregister_fcf(phba);
/* Replace in-use record with the new record */
memcpy(&phba->fcf.current_rec, &phba->fcf.failover_rec,
sizeof(struct lpfc_fcf_rec));
lpfc_printf_log(phba, KERN_INFO, LOG_FIP,
"2783 Perform FLOGI roundrobin FCF failover: FCF "
"(x%x) to FCF (x%x)\n", current_fcf_index, fcf_index);
error_out:
lpfc_register_fcf(phba);
out:
lpfc_sli4_mbox_cmd_free(phba, mboxq);
}
/**
* lpfc_mbx_cmpl_read_fcf_rec - read fcf completion handler.
* @phba: pointer to lpfc hba data structure.
* @mboxq: pointer to mailbox object.
*
* This is the callback function of read FCF record mailbox command for
* updating the eligible FCF bmask for FLOGI failure roundrobin FCF
* failover when a new FCF event happened. If the FCF read back is
* valid/available and it passes the connection list check, it updates
* the bmask for the eligible FCF record for roundrobin failover.
*/
void
lpfc_mbx_cmpl_read_fcf_rec(struct lpfc_hba *phba, LPFC_MBOXQ_t *mboxq)
{
struct fcf_record *new_fcf_record;
uint32_t boot_flag, addr_mode;
uint16_t fcf_index, next_fcf_index;
scsi: lpfc: Fix various issues reported by tools This patch fixes below Smatch reported issues: 1. lpfc_hbadisc.c:3020 lpfc_mbx_cmpl_fcf_rr_read_fcf_rec() error: uninitialized symbol 'vlan_id'. 2. lpfc_hbadisc.c:3121 lpfc_mbx_cmpl_read_fcf_rec() error: uninitialized symbol 'vlan_id'. 3. lpfc_init.c:335 lpfc_dump_wakeup_param_cmpl() warn: always true condition '(prg->dist < 4) => (0-3 < 4)' 4. lpfc_init.c:2419 lpfc_parse_vpd() warn: inconsistent indenting. 5. lpfc_init.c:13248 lpfc_sli4_enable_msi() warn: 'phba->pcidev->irq' 2147483648 can't fit into 65535 'eqhdl->irq' 6. lpfc_debugfs.c:5300 lpfc_idiag_extacc_avail_get() error: uninitialized symbol 'ext_cnt' 7. lpfc_debugfs.c:5300 lpfc_idiag_extacc_avail_get() error: uninitialized symbol 'ext_size' 8. lpfc_vmid.c:248 lpfc_vmid_get_appid() warn: sleeping in atomic context. 9. lpfc_init.c:8342 lpfc_sli4_driver_resource_setup() warn: missing error code 'rc'. 10. lpfc_init.c:13573 lpfc_sli4_hba_unset() warn: variable dereferenced before check 'phba->pport' (see line 13546) 11. lpfc_auth.c:1923 lpfc_auth_handle_dhchap_reply() error: double free of 'hash_value' Fixes: 1. Initialize vlan_id to LPFC_FCOE_NULL_VID. 2. Initialize vlan_id to LPFC_FCOE_NULL_VID. 3. prg->dist is a 2 bit field. Its value can only be between 0-3. Remove redundent check 'if (prg->dist < 4)'. 4. Fix inconsistent indenting. Moved logic into helper function lpfc_fill_vpd(). 5. Define 'eqhdl->irq' as int value as pci_irq_vector() returns int. Also, check for return value of pci_irq_vector() and log message in case of failure. 6. Initialize 'ext_cnt' to 0. 7. Initialize 'ext_size' to 0. 8. Use alloc_percpu_gfp() with GFP_ATOMIC flag. 9. 'rc' was not updated when dma_pool_create() fails. Update 'rc = -ENOMEM' when dma_pool_create() fails before calling goto statement. 10. Add check for 'phba->pport' in lpfc_cpuhp_remove(). 11. Initialize 'hash_value' to NULL, same like 'aug_chal' variable. Link: https://lore.kernel.org/r/20220911221505.117655-13-jsmart2021@gmail.com Co-developed-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2022-09-11 22:15:04 +00:00
uint16_t vlan_id = LPFC_FCOE_NULL_VID;
int rc;
/* If link state is not up, no need to proceed */
if (phba->link_state < LPFC_LINK_UP)
goto out;
/* If FCF discovery period is over, no need to proceed */
if (!(phba->fcf.fcf_flag & FCF_DISCOVERY))
goto out;
/* Parse the FCF record from the non-embedded mailbox command */
new_fcf_record = lpfc_sli4_fcf_rec_mbox_parse(phba, mboxq,
&next_fcf_index);
if (!new_fcf_record) {
lpfc_printf_log(phba, KERN_INFO, LOG_FIP,
"2767 Mailbox command READ_FCF_RECORD "
"failed to retrieve a FCF record.\n");
goto out;
}
/* Check the connection list for eligibility */
rc = lpfc_match_fcf_conn_list(phba, new_fcf_record, &boot_flag,
&addr_mode, &vlan_id);
/* Log the FCF record information if turned on */
lpfc_sli4_log_fcf_record_info(phba, new_fcf_record, vlan_id,
next_fcf_index);
if (!rc)
goto out;
/* Update the eligible FCF record index bmask */
fcf_index = bf_get(lpfc_fcf_record_fcf_index, new_fcf_record);
rc = lpfc_sli4_fcf_pri_list_add(phba, fcf_index, new_fcf_record);
out:
lpfc_sli4_mbox_cmd_free(phba, mboxq);
}
2010-11-21 04:11:48 +00:00
/**
* lpfc_init_vfi_cmpl - Completion handler for init_vfi mbox command.
* @phba: pointer to lpfc hba data structure.
* @mboxq: pointer to mailbox data structure.
*
* This function handles completion of init vfi mailbox command.
*/
static void
2010-11-21 04:11:48 +00:00
lpfc_init_vfi_cmpl(struct lpfc_hba *phba, LPFC_MBOXQ_t *mboxq)
{
struct lpfc_vport *vport = mboxq->vport;
/*
* VFI not supported on interface type 0, just do the flogi
* Also continue if the VFI is in use - just use the same one.
*/
if (mboxq->u.mb.mbxStatus &&
(bf_get(lpfc_sli_intf_if_type, &phba->sli4_hba.sli_intf) !=
LPFC_SLI_INTF_IF_TYPE_0) &&
mboxq->u.mb.mbxStatus != MBX_VFI_IN_USE) {
scsi: lpfc: Add an internal trace log buffer The current logging methods typically end up requesting a reproduction with a different logging level set to figure out what happened. This was mainly by design to not clutter the kernel log messages with things that were typically not interesting and the messages themselves could cause other issues. When looking to make a better system, it was seen that in many cases when more data was wanted was when another message, usually at KERN_ERR level, was logged. And in most cases, what the additional logging that was then enabled was typically. Most of these areas fell into the discovery machine. Based on this summary, the following design has been put in place: The driver will maintain an internal log (256 elements of 256 bytes). The "additional logging" messages that are usually enabled in a reproduction will be changed to now log all the time to the internal log. A new logging level is defined - LOG_TRACE_EVENT. When this level is set (it is not by default) and a message marked as KERN_ERR is logged, all the messages in the internal log will be dumped to the kernel log before the KERN_ERR message is logged. There is a timestamp on each message added to the internal log. However, this timestamp is not converted to wall time when logged. The value of the timestamp is solely to give a crude time reference for the messages. Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-06-30 21:50:00 +00:00
lpfc_printf_vlog(vport, KERN_ERR, LOG_TRACE_EVENT,
"2891 Init VFI mailbox failed 0x%x\n",
mboxq->u.mb.mbxStatus);
2010-11-21 04:11:48 +00:00
mempool_free(mboxq, phba->mbox_mem_pool);
lpfc_vport_set_state(vport, FC_VPORT_FAILED);
return;
}
2011-10-11 01:32:10 +00:00
2010-11-21 04:11:48 +00:00
lpfc_initial_flogi(vport);
mempool_free(mboxq, phba->mbox_mem_pool);
return;
}
/**
* lpfc_issue_init_vfi - Issue init_vfi mailbox command.
* @vport: pointer to lpfc_vport data structure.
*
* This function issue a init_vfi mailbox command to initialize the VFI and
* VPI for the physical port.
*/
void
lpfc_issue_init_vfi(struct lpfc_vport *vport)
{
LPFC_MBOXQ_t *mboxq;
int rc;
struct lpfc_hba *phba = vport->phba;
mboxq = mempool_alloc(phba->mbox_mem_pool, GFP_KERNEL);
if (!mboxq) {
lpfc_printf_vlog(vport, KERN_ERR,
scsi: lpfc: Add an internal trace log buffer The current logging methods typically end up requesting a reproduction with a different logging level set to figure out what happened. This was mainly by design to not clutter the kernel log messages with things that were typically not interesting and the messages themselves could cause other issues. When looking to make a better system, it was seen that in many cases when more data was wanted was when another message, usually at KERN_ERR level, was logged. And in most cases, what the additional logging that was then enabled was typically. Most of these areas fell into the discovery machine. Based on this summary, the following design has been put in place: The driver will maintain an internal log (256 elements of 256 bytes). The "additional logging" messages that are usually enabled in a reproduction will be changed to now log all the time to the internal log. A new logging level is defined - LOG_TRACE_EVENT. When this level is set (it is not by default) and a message marked as KERN_ERR is logged, all the messages in the internal log will be dumped to the kernel log before the KERN_ERR message is logged. There is a timestamp on each message added to the internal log. However, this timestamp is not converted to wall time when logged. The value of the timestamp is solely to give a crude time reference for the messages. Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-06-30 21:50:00 +00:00
LOG_TRACE_EVENT, "2892 Failed to allocate "
2010-11-21 04:11:48 +00:00
"init_vfi mailbox\n");
return;
}
lpfc_init_vfi(mboxq, vport);
mboxq->mbox_cmpl = lpfc_init_vfi_cmpl;
rc = lpfc_sli_issue_mbox(phba, mboxq, MBX_NOWAIT);
if (rc == MBX_NOT_FINISHED) {
scsi: lpfc: Add an internal trace log buffer The current logging methods typically end up requesting a reproduction with a different logging level set to figure out what happened. This was mainly by design to not clutter the kernel log messages with things that were typically not interesting and the messages themselves could cause other issues. When looking to make a better system, it was seen that in many cases when more data was wanted was when another message, usually at KERN_ERR level, was logged. And in most cases, what the additional logging that was then enabled was typically. Most of these areas fell into the discovery machine. Based on this summary, the following design has been put in place: The driver will maintain an internal log (256 elements of 256 bytes). The "additional logging" messages that are usually enabled in a reproduction will be changed to now log all the time to the internal log. A new logging level is defined - LOG_TRACE_EVENT. When this level is set (it is not by default) and a message marked as KERN_ERR is logged, all the messages in the internal log will be dumped to the kernel log before the KERN_ERR message is logged. There is a timestamp on each message added to the internal log. However, this timestamp is not converted to wall time when logged. The value of the timestamp is solely to give a crude time reference for the messages. Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-06-30 21:50:00 +00:00
lpfc_printf_vlog(vport, KERN_ERR, LOG_TRACE_EVENT,
"2893 Failed to issue init_vfi mailbox\n");
2010-11-21 04:11:48 +00:00
mempool_free(mboxq, vport->phba->mbox_mem_pool);
}
}
/**
* lpfc_init_vpi_cmpl - Completion handler for init_vpi mbox command.
* @phba: pointer to lpfc hba data structure.
* @mboxq: pointer to mailbox data structure.
*
* This function handles completion of init vpi mailbox command.
*/
void
lpfc_init_vpi_cmpl(struct lpfc_hba *phba, LPFC_MBOXQ_t *mboxq)
{
struct lpfc_vport *vport = mboxq->vport;
struct lpfc_nodelist *ndlp;
struct Scsi_Host *shost = lpfc_shost_from_vport(vport);
if (mboxq->u.mb.mbxStatus) {
scsi: lpfc: Add an internal trace log buffer The current logging methods typically end up requesting a reproduction with a different logging level set to figure out what happened. This was mainly by design to not clutter the kernel log messages with things that were typically not interesting and the messages themselves could cause other issues. When looking to make a better system, it was seen that in many cases when more data was wanted was when another message, usually at KERN_ERR level, was logged. And in most cases, what the additional logging that was then enabled was typically. Most of these areas fell into the discovery machine. Based on this summary, the following design has been put in place: The driver will maintain an internal log (256 elements of 256 bytes). The "additional logging" messages that are usually enabled in a reproduction will be changed to now log all the time to the internal log. A new logging level is defined - LOG_TRACE_EVENT. When this level is set (it is not by default) and a message marked as KERN_ERR is logged, all the messages in the internal log will be dumped to the kernel log before the KERN_ERR message is logged. There is a timestamp on each message added to the internal log. However, this timestamp is not converted to wall time when logged. The value of the timestamp is solely to give a crude time reference for the messages. Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-06-30 21:50:00 +00:00
lpfc_printf_vlog(vport, KERN_ERR, LOG_TRACE_EVENT,
"2609 Init VPI mailbox failed 0x%x\n",
mboxq->u.mb.mbxStatus);
mempool_free(mboxq, phba->mbox_mem_pool);
lpfc_vport_set_state(vport, FC_VPORT_FAILED);
return;
}
spin_lock_irq(shost->host_lock);
vport->fc_flag &= ~FC_VPORT_NEEDS_INIT_VPI;
spin_unlock_irq(shost->host_lock);
/* If this port is physical port or FDISC is done, do reg_vpi */
if ((phba->pport == vport) || (vport->port_state == LPFC_FDISC)) {
ndlp = lpfc_findnode_did(vport, Fabric_DID);
if (!ndlp)
lpfc_printf_vlog(vport, KERN_ERR,
scsi: lpfc: Add an internal trace log buffer The current logging methods typically end up requesting a reproduction with a different logging level set to figure out what happened. This was mainly by design to not clutter the kernel log messages with things that were typically not interesting and the messages themselves could cause other issues. When looking to make a better system, it was seen that in many cases when more data was wanted was when another message, usually at KERN_ERR level, was logged. And in most cases, what the additional logging that was then enabled was typically. Most of these areas fell into the discovery machine. Based on this summary, the following design has been put in place: The driver will maintain an internal log (256 elements of 256 bytes). The "additional logging" messages that are usually enabled in a reproduction will be changed to now log all the time to the internal log. A new logging level is defined - LOG_TRACE_EVENT. When this level is set (it is not by default) and a message marked as KERN_ERR is logged, all the messages in the internal log will be dumped to the kernel log before the KERN_ERR message is logged. There is a timestamp on each message added to the internal log. However, this timestamp is not converted to wall time when logged. The value of the timestamp is solely to give a crude time reference for the messages. Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-06-30 21:50:00 +00:00
LOG_TRACE_EVENT,
"2731 Cannot find fabric "
"controller node\n");
else
lpfc_register_new_vport(phba, vport, ndlp);
mempool_free(mboxq, phba->mbox_mem_pool);
return;
}
if (phba->link_flag & LS_NPIV_FAB_SUPPORTED)
lpfc_initial_fdisc(vport);
else {
lpfc_vport_set_state(vport, FC_VPORT_NO_FABRIC_SUPP);
scsi: lpfc: Add an internal trace log buffer The current logging methods typically end up requesting a reproduction with a different logging level set to figure out what happened. This was mainly by design to not clutter the kernel log messages with things that were typically not interesting and the messages themselves could cause other issues. When looking to make a better system, it was seen that in many cases when more data was wanted was when another message, usually at KERN_ERR level, was logged. And in most cases, what the additional logging that was then enabled was typically. Most of these areas fell into the discovery machine. Based on this summary, the following design has been put in place: The driver will maintain an internal log (256 elements of 256 bytes). The "additional logging" messages that are usually enabled in a reproduction will be changed to now log all the time to the internal log. A new logging level is defined - LOG_TRACE_EVENT. When this level is set (it is not by default) and a message marked as KERN_ERR is logged, all the messages in the internal log will be dumped to the kernel log before the KERN_ERR message is logged. There is a timestamp on each message added to the internal log. However, this timestamp is not converted to wall time when logged. The value of the timestamp is solely to give a crude time reference for the messages. Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-06-30 21:50:00 +00:00
lpfc_printf_vlog(vport, KERN_ERR, LOG_TRACE_EVENT,
"2606 No NPIV Fabric support\n");
}
mempool_free(mboxq, phba->mbox_mem_pool);
return;
}
/**
* lpfc_issue_init_vpi - Issue init_vpi mailbox command.
* @vport: pointer to lpfc_vport data structure.
*
* This function issue a init_vpi mailbox command to initialize
* VPI for the vport.
*/
void
lpfc_issue_init_vpi(struct lpfc_vport *vport)
{
LPFC_MBOXQ_t *mboxq;
int rc, vpi;
if ((vport->port_type != LPFC_PHYSICAL_PORT) && (!vport->vpi)) {
vpi = lpfc_alloc_vpi(vport->phba);
if (!vpi) {
scsi: lpfc: Add an internal trace log buffer The current logging methods typically end up requesting a reproduction with a different logging level set to figure out what happened. This was mainly by design to not clutter the kernel log messages with things that were typically not interesting and the messages themselves could cause other issues. When looking to make a better system, it was seen that in many cases when more data was wanted was when another message, usually at KERN_ERR level, was logged. And in most cases, what the additional logging that was then enabled was typically. Most of these areas fell into the discovery machine. Based on this summary, the following design has been put in place: The driver will maintain an internal log (256 elements of 256 bytes). The "additional logging" messages that are usually enabled in a reproduction will be changed to now log all the time to the internal log. A new logging level is defined - LOG_TRACE_EVENT. When this level is set (it is not by default) and a message marked as KERN_ERR is logged, all the messages in the internal log will be dumped to the kernel log before the KERN_ERR message is logged. There is a timestamp on each message added to the internal log. However, this timestamp is not converted to wall time when logged. The value of the timestamp is solely to give a crude time reference for the messages. Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-06-30 21:50:00 +00:00
lpfc_printf_vlog(vport, KERN_ERR, LOG_TRACE_EVENT,
"3303 Failed to obtain vport vpi\n");
lpfc_vport_set_state(vport, FC_VPORT_FAILED);
return;
}
vport->vpi = vpi;
}
mboxq = mempool_alloc(vport->phba->mbox_mem_pool, GFP_KERNEL);
if (!mboxq) {
lpfc_printf_vlog(vport, KERN_ERR,
scsi: lpfc: Add an internal trace log buffer The current logging methods typically end up requesting a reproduction with a different logging level set to figure out what happened. This was mainly by design to not clutter the kernel log messages with things that were typically not interesting and the messages themselves could cause other issues. When looking to make a better system, it was seen that in many cases when more data was wanted was when another message, usually at KERN_ERR level, was logged. And in most cases, what the additional logging that was then enabled was typically. Most of these areas fell into the discovery machine. Based on this summary, the following design has been put in place: The driver will maintain an internal log (256 elements of 256 bytes). The "additional logging" messages that are usually enabled in a reproduction will be changed to now log all the time to the internal log. A new logging level is defined - LOG_TRACE_EVENT. When this level is set (it is not by default) and a message marked as KERN_ERR is logged, all the messages in the internal log will be dumped to the kernel log before the KERN_ERR message is logged. There is a timestamp on each message added to the internal log. However, this timestamp is not converted to wall time when logged. The value of the timestamp is solely to give a crude time reference for the messages. Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-06-30 21:50:00 +00:00
LOG_TRACE_EVENT, "2607 Failed to allocate "
"init_vpi mailbox\n");
return;
}
lpfc_init_vpi(vport->phba, mboxq, vport->vpi);
mboxq->vport = vport;
mboxq->mbox_cmpl = lpfc_init_vpi_cmpl;
rc = lpfc_sli_issue_mbox(vport->phba, mboxq, MBX_NOWAIT);
if (rc == MBX_NOT_FINISHED) {
scsi: lpfc: Add an internal trace log buffer The current logging methods typically end up requesting a reproduction with a different logging level set to figure out what happened. This was mainly by design to not clutter the kernel log messages with things that were typically not interesting and the messages themselves could cause other issues. When looking to make a better system, it was seen that in many cases when more data was wanted was when another message, usually at KERN_ERR level, was logged. And in most cases, what the additional logging that was then enabled was typically. Most of these areas fell into the discovery machine. Based on this summary, the following design has been put in place: The driver will maintain an internal log (256 elements of 256 bytes). The "additional logging" messages that are usually enabled in a reproduction will be changed to now log all the time to the internal log. A new logging level is defined - LOG_TRACE_EVENT. When this level is set (it is not by default) and a message marked as KERN_ERR is logged, all the messages in the internal log will be dumped to the kernel log before the KERN_ERR message is logged. There is a timestamp on each message added to the internal log. However, this timestamp is not converted to wall time when logged. The value of the timestamp is solely to give a crude time reference for the messages. Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-06-30 21:50:00 +00:00
lpfc_printf_vlog(vport, KERN_ERR, LOG_TRACE_EVENT,
"2608 Failed to issue init_vpi mailbox\n");
mempool_free(mboxq, vport->phba->mbox_mem_pool);
}
}
/**
* lpfc_start_fdiscs - send fdiscs for each vports on this port.
* @phba: pointer to lpfc hba data structure.
*
* This function loops through the list of vports on the @phba and issues an
* FDISC if possible.
*/
void
lpfc_start_fdiscs(struct lpfc_hba *phba)
{
struct lpfc_vport **vports;
int i;
vports = lpfc_create_vport_work_array(phba);
if (vports != NULL) {
for (i = 0; i <= phba->max_vports && vports[i] != NULL; i++) {
if (vports[i]->port_type == LPFC_PHYSICAL_PORT)
continue;
/* There are no vpi for this vport */
if (vports[i]->vpi > phba->max_vpi) {
lpfc_vport_set_state(vports[i],
FC_VPORT_FAILED);
continue;
}
2010-11-21 04:11:48 +00:00
if (phba->fc_topology == LPFC_TOPOLOGY_LOOP) {
lpfc_vport_set_state(vports[i],
FC_VPORT_LINKDOWN);
continue;
}
if (vports[i]->fc_flag & FC_VPORT_NEEDS_INIT_VPI) {
lpfc_issue_init_vpi(vports[i]);
continue;
}
if (phba->link_flag & LS_NPIV_FAB_SUPPORTED)
lpfc_initial_fdisc(vports[i]);
else {
lpfc_vport_set_state(vports[i],
FC_VPORT_NO_FABRIC_SUPP);
lpfc_printf_vlog(vports[i], KERN_ERR,
scsi: lpfc: Add an internal trace log buffer The current logging methods typically end up requesting a reproduction with a different logging level set to figure out what happened. This was mainly by design to not clutter the kernel log messages with things that were typically not interesting and the messages themselves could cause other issues. When looking to make a better system, it was seen that in many cases when more data was wanted was when another message, usually at KERN_ERR level, was logged. And in most cases, what the additional logging that was then enabled was typically. Most of these areas fell into the discovery machine. Based on this summary, the following design has been put in place: The driver will maintain an internal log (256 elements of 256 bytes). The "additional logging" messages that are usually enabled in a reproduction will be changed to now log all the time to the internal log. A new logging level is defined - LOG_TRACE_EVENT. When this level is set (it is not by default) and a message marked as KERN_ERR is logged, all the messages in the internal log will be dumped to the kernel log before the KERN_ERR message is logged. There is a timestamp on each message added to the internal log. However, this timestamp is not converted to wall time when logged. The value of the timestamp is solely to give a crude time reference for the messages. Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-06-30 21:50:00 +00:00
LOG_TRACE_EVENT,
"0259 No NPIV "
"Fabric support\n");
}
}
}
lpfc_destroy_vport_work_array(phba, vports);
}
void
lpfc_mbx_cmpl_reg_vfi(struct lpfc_hba *phba, LPFC_MBOXQ_t *mboxq)
{
struct lpfc_vport *vport = mboxq->vport;
struct Scsi_Host *shost = lpfc_shost_from_vport(vport);
/*
* VFI not supported for interface type 0, so ignore any mailbox
* error (except VFI in use) and continue with the discovery.
*/
if (mboxq->u.mb.mbxStatus &&
(bf_get(lpfc_sli_intf_if_type, &phba->sli4_hba.sli_intf) !=
LPFC_SLI_INTF_IF_TYPE_0) &&
mboxq->u.mb.mbxStatus != MBX_VFI_IN_USE) {
scsi: lpfc: Add an internal trace log buffer The current logging methods typically end up requesting a reproduction with a different logging level set to figure out what happened. This was mainly by design to not clutter the kernel log messages with things that were typically not interesting and the messages themselves could cause other issues. When looking to make a better system, it was seen that in many cases when more data was wanted was when another message, usually at KERN_ERR level, was logged. And in most cases, what the additional logging that was then enabled was typically. Most of these areas fell into the discovery machine. Based on this summary, the following design has been put in place: The driver will maintain an internal log (256 elements of 256 bytes). The "additional logging" messages that are usually enabled in a reproduction will be changed to now log all the time to the internal log. A new logging level is defined - LOG_TRACE_EVENT. When this level is set (it is not by default) and a message marked as KERN_ERR is logged, all the messages in the internal log will be dumped to the kernel log before the KERN_ERR message is logged. There is a timestamp on each message added to the internal log. However, this timestamp is not converted to wall time when logged. The value of the timestamp is solely to give a crude time reference for the messages. Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-06-30 21:50:00 +00:00
lpfc_printf_vlog(vport, KERN_ERR, LOG_TRACE_EVENT,
"2018 REG_VFI mbxStatus error x%x "
"HBA state x%x\n",
mboxq->u.mb.mbxStatus, vport->port_state);
2010-11-21 04:11:48 +00:00
if (phba->fc_topology == LPFC_TOPOLOGY_LOOP) {
/* FLOGI failed, use loop map to make discovery list */
lpfc_disc_list_loopmap(vport);
/* Start discovery */
lpfc_disc_start(vport);
goto out_free_mem;
}
lpfc_vport_set_state(vport, FC_VPORT_FAILED);
goto out_free_mem;
}
/* If the VFI is already registered, there is nothing else to do
* Unless this was a VFI update and we are in PT2PT mode, then
* we should drop through to set the port state to ready.
*/
if (vport->fc_flag & FC_VFI_REGISTERED)
if (!(phba->sli_rev == LPFC_SLI_REV4 &&
vport->fc_flag & FC_PT2PT))
goto out_free_mem;
/* The VPI is implicitly registered when the VFI is registered */
spin_lock_irq(shost->host_lock);
vport->vpi_state |= LPFC_VPI_REGISTERED;
vport->fc_flag |= FC_VFI_REGISTERED;
vport->fc_flag &= ~FC_VPORT_NEEDS_REG_VPI;
vport->fc_flag &= ~FC_VPORT_NEEDS_INIT_VPI;
spin_unlock_irq(shost->host_lock);
/* In case SLI4 FC loopback test, we are ready */
if ((phba->sli_rev == LPFC_SLI_REV4) &&
(phba->link_flag & LS_LOOPBACK_MODE)) {
phba->link_state = LPFC_HBA_READY;
goto out_free_mem;
}
lpfc_printf_vlog(vport, KERN_INFO, LOG_SLI,
"3313 cmpl reg vfi port_state:%x fc_flag:%x myDid:%x "
"alpacnt:%d LinkState:%x topology:%x\n",
vport->port_state, vport->fc_flag, vport->fc_myDID,
vport->phba->alpa_map[0],
phba->link_state, phba->fc_topology);
if (vport->port_state == LPFC_FABRIC_CFG_LINK) {
/*
* For private loop or for NPort pt2pt,
* just start discovery and we are done.
*/
if ((vport->fc_flag & FC_PT2PT) ||
((phba->fc_topology == LPFC_TOPOLOGY_LOOP) &&
!(vport->fc_flag & FC_PUBLIC_LOOP))) {
2010-11-21 04:11:48 +00:00
/* Use loop map to make discovery list */
lpfc_disc_list_loopmap(vport);
/* Start discovery */
if (vport->fc_flag & FC_PT2PT)
vport->port_state = LPFC_VPORT_READY;
else
lpfc_disc_start(vport);
2010-11-21 04:11:48 +00:00
} else {
lpfc_start_fdiscs(phba);
lpfc_do_scr_ns_plogi(phba, vport);
}
}
out_free_mem:
lpfc_mbox_rsrc_cleanup(phba, mboxq, MBOX_THD_UNLOCKED);
}
static void
lpfc_mbx_cmpl_read_sparam(struct lpfc_hba *phba, LPFC_MBOXQ_t *pmb)
{
MAILBOX_t *mb = &pmb->u.mb;
struct lpfc_dmabuf *mp = (struct lpfc_dmabuf *)pmb->ctx_buf;
struct lpfc_vport *vport = pmb->vport;
struct Scsi_Host *shost = lpfc_shost_from_vport(vport);
struct serv_parm *sp = &vport->fc_sparam;
uint32_t ed_tov;
/* Check for error */
if (mb->mbxStatus) {
/* READ_SPARAM mbox error <mbxStatus> state <hba_state> */
scsi: lpfc: Add an internal trace log buffer The current logging methods typically end up requesting a reproduction with a different logging level set to figure out what happened. This was mainly by design to not clutter the kernel log messages with things that were typically not interesting and the messages themselves could cause other issues. When looking to make a better system, it was seen that in many cases when more data was wanted was when another message, usually at KERN_ERR level, was logged. And in most cases, what the additional logging that was then enabled was typically. Most of these areas fell into the discovery machine. Based on this summary, the following design has been put in place: The driver will maintain an internal log (256 elements of 256 bytes). The "additional logging" messages that are usually enabled in a reproduction will be changed to now log all the time to the internal log. A new logging level is defined - LOG_TRACE_EVENT. When this level is set (it is not by default) and a message marked as KERN_ERR is logged, all the messages in the internal log will be dumped to the kernel log before the KERN_ERR message is logged. There is a timestamp on each message added to the internal log. However, this timestamp is not converted to wall time when logged. The value of the timestamp is solely to give a crude time reference for the messages. Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-06-30 21:50:00 +00:00
lpfc_printf_vlog(vport, KERN_ERR, LOG_TRACE_EVENT,
"0319 READ_SPARAM mbxStatus error x%x "
"hba state x%x>\n",
mb->mbxStatus, vport->port_state);
lpfc_linkdown(phba);
goto out;
}
memcpy((uint8_t *) &vport->fc_sparam, (uint8_t *) mp->virt,
sizeof (struct serv_parm));
ed_tov = be32_to_cpu(sp->cmn.e_d_tov);
if (sp->cmn.edtovResolution) /* E_D_TOV ticks are in nanoseconds */
ed_tov = (ed_tov + 999999) / 1000000;
phba->fc_edtov = ed_tov;
phba->fc_ratov = (2 * ed_tov) / 1000;
if (phba->fc_ratov < FF_DEF_RATOV) {
/* RA_TOV should be atleast 10sec for initial flogi */
phba->fc_ratov = FF_DEF_RATOV;
}
lpfc_update_vport_wwn(vport);
fc_host_port_name(shost) = wwn_to_u64(vport->fc_portname.u.wwn);
if (vport->port_type == LPFC_PHYSICAL_PORT) {
memcpy(&phba->wwnn, &vport->fc_nodename, sizeof(phba->wwnn));
memcpy(&phba->wwpn, &vport->fc_portname, sizeof(phba->wwnn));
}
lpfc_mbox_rsrc_cleanup(phba, pmb, MBOX_THD_UNLOCKED);
/* Check if sending the FLOGI is being deferred to after we get
* up to date CSPs from MBX_READ_SPARAM.
*/
if (phba->hba_flag & HBA_DEFER_FLOGI) {
lpfc_initial_flogi(vport);
phba->hba_flag &= ~HBA_DEFER_FLOGI;
}
return;
out:
lpfc_mbox_rsrc_cleanup(phba, pmb, MBOX_THD_UNLOCKED);
lpfc_issue_clear_la(phba, vport);
}
static void
2010-11-21 04:11:48 +00:00
lpfc_mbx_process_link_up(struct lpfc_hba *phba, struct lpfc_mbx_read_top *la)
{
struct lpfc_vport *vport = phba->pport;
LPFC_MBOXQ_t *sparam_mbox, *cfglink_mbox = NULL;
struct Scsi_Host *shost;
int i;
int rc;
struct fcf_record *fcf_record;
uint32_t fc_flags = 0;
unsigned long iflags;
spin_lock_irqsave(&phba->hbalock, iflags);
phba->fc_linkspeed = bf_get(lpfc_mbx_read_top_link_spd, la);
if (!(phba->hba_flag & HBA_FCOE_MODE)) {
switch (bf_get(lpfc_mbx_read_top_link_spd, la)) {
case LPFC_LINK_SPEED_1GHZ:
case LPFC_LINK_SPEED_2GHZ:
case LPFC_LINK_SPEED_4GHZ:
case LPFC_LINK_SPEED_8GHZ:
case LPFC_LINK_SPEED_10GHZ:
case LPFC_LINK_SPEED_16GHZ:
case LPFC_LINK_SPEED_32GHZ:
case LPFC_LINK_SPEED_64GHZ:
case LPFC_LINK_SPEED_128GHZ:
case LPFC_LINK_SPEED_256GHZ:
break;
default:
phba->fc_linkspeed = LPFC_LINK_SPEED_UNKNOWN;
break;
}
}
if (phba->fc_topology &&
phba->fc_topology != bf_get(lpfc_mbx_read_top_topology, la)) {
lpfc_printf_log(phba, KERN_WARNING, LOG_SLI,
"3314 Toplogy changed was 0x%x is 0x%x\n",
phba->fc_topology,
bf_get(lpfc_mbx_read_top_topology, la));
phba->fc_topology_changed = 1;
}
2010-11-21 04:11:48 +00:00
phba->fc_topology = bf_get(lpfc_mbx_read_top_topology, la);
phba->link_flag &= ~(LS_NPIV_FAB_SUPPORTED | LS_CT_VEN_RPA);
shost = lpfc_shost_from_vport(vport);
2010-11-21 04:11:48 +00:00
if (phba->fc_topology == LPFC_TOPOLOGY_LOOP) {
phba->sli3_options &= ~LPFC_SLI3_NPIV_ENABLED;
/* if npiv is enabled and this adapter supports npiv log
* a message that npiv is not supported in this topology
*/
if (phba->cfg_enable_npiv && phba->max_vpi)
lpfc_printf_log(phba, KERN_ERR, LOG_LINK_EVENT,
"1309 Link Up Event npiv not supported in loop "
"topology\n");
/* Get Loop Map information */
if (bf_get(lpfc_mbx_read_top_il, la))
fc_flags |= FC_LBIT;
2010-11-21 04:11:48 +00:00
vport->fc_myDID = bf_get(lpfc_mbx_read_top_alpa_granted, la);
i = la->lilpBde64.tus.f.bdeSize;
if (i == 0) {
phba->alpa_map[0] = 0;
} else {
if (vport->cfg_log_verbose & LOG_LINK_EVENT) {
int numalpa, j, k;
union {
uint8_t pamap[16];
struct {
uint32_t wd1;
uint32_t wd2;
uint32_t wd3;
uint32_t wd4;
} pa;
} un;
numalpa = phba->alpa_map[0];
j = 0;
while (j < numalpa) {
memset(un.pamap, 0, 16);
for (k = 1; j < numalpa; k++) {
un.pamap[k - 1] =
phba->alpa_map[j + 1];
j++;
if (k == 16)
break;
}
/* Link Up Event ALPA map */
lpfc_printf_log(phba,
KERN_WARNING,
LOG_LINK_EVENT,
"1304 Link Up Event "
"ALPA map Data: x%x "
"x%x x%x x%x\n",
un.pa.wd1, un.pa.wd2,
un.pa.wd3, un.pa.wd4);
}
}
}
} else {
if (!(phba->sli3_options & LPFC_SLI3_NPIV_ENABLED)) {
if (phba->max_vpi && phba->cfg_enable_npiv &&
(phba->sli_rev >= LPFC_SLI_REV3))
phba->sli3_options |= LPFC_SLI3_NPIV_ENABLED;
}
vport->fc_myDID = phba->fc_pref_DID;
fc_flags |= FC_LBIT;
}
spin_unlock_irqrestore(&phba->hbalock, iflags);
if (fc_flags) {
spin_lock_irqsave(shost->host_lock, iflags);
vport->fc_flag |= fc_flags;
spin_unlock_irqrestore(shost->host_lock, iflags);
}
lpfc_linkup(phba);
sparam_mbox = NULL;
sparam_mbox = mempool_alloc(phba->mbox_mem_pool, GFP_KERNEL);
if (!sparam_mbox)
goto out;
rc = lpfc_read_sparam(phba, sparam_mbox, 0);
if (rc) {
mempool_free(sparam_mbox, phba->mbox_mem_pool);
goto out;
}
sparam_mbox->vport = vport;
sparam_mbox->mbox_cmpl = lpfc_mbx_cmpl_read_sparam;
rc = lpfc_sli_issue_mbox(phba, sparam_mbox, MBX_NOWAIT);
if (rc == MBX_NOT_FINISHED) {
lpfc_mbox_rsrc_cleanup(phba, sparam_mbox, MBOX_THD_UNLOCKED);
goto out;
}
if (!(phba->hba_flag & HBA_FCOE_MODE)) {
cfglink_mbox = mempool_alloc(phba->mbox_mem_pool, GFP_KERNEL);
if (!cfglink_mbox)
goto out;
vport->port_state = LPFC_LOCAL_CFG_LINK;
lpfc_config_link(phba, cfglink_mbox);
cfglink_mbox->vport = vport;
cfglink_mbox->mbox_cmpl = lpfc_mbx_cmpl_local_config_link;
rc = lpfc_sli_issue_mbox(phba, cfglink_mbox, MBX_NOWAIT);
if (rc == MBX_NOT_FINISHED) {
mempool_free(cfglink_mbox, phba->mbox_mem_pool);
goto out;
}
} else {
vport->port_state = LPFC_VPORT_UNKNOWN;
/*
* Add the driver's default FCF record at FCF index 0 now. This
* is phase 1 implementation that support FCF index 0 and driver
* defaults.
*/
if (!(phba->hba_flag & HBA_FIP_SUPPORT)) {
fcf_record = kzalloc(sizeof(struct fcf_record),
GFP_KERNEL);
if (unlikely(!fcf_record)) {
lpfc_printf_log(phba, KERN_ERR,
scsi: lpfc: Add an internal trace log buffer The current logging methods typically end up requesting a reproduction with a different logging level set to figure out what happened. This was mainly by design to not clutter the kernel log messages with things that were typically not interesting and the messages themselves could cause other issues. When looking to make a better system, it was seen that in many cases when more data was wanted was when another message, usually at KERN_ERR level, was logged. And in most cases, what the additional logging that was then enabled was typically. Most of these areas fell into the discovery machine. Based on this summary, the following design has been put in place: The driver will maintain an internal log (256 elements of 256 bytes). The "additional logging" messages that are usually enabled in a reproduction will be changed to now log all the time to the internal log. A new logging level is defined - LOG_TRACE_EVENT. When this level is set (it is not by default) and a message marked as KERN_ERR is logged, all the messages in the internal log will be dumped to the kernel log before the KERN_ERR message is logged. There is a timestamp on each message added to the internal log. However, this timestamp is not converted to wall time when logged. The value of the timestamp is solely to give a crude time reference for the messages. Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-06-30 21:50:00 +00:00
LOG_TRACE_EVENT,
"2554 Could not allocate memory for "
"fcf record\n");
rc = -ENODEV;
goto out;
}
lpfc_sli4_build_dflt_fcf_record(phba, fcf_record,
LPFC_FCOE_FCF_DEF_INDEX);
rc = lpfc_sli4_add_fcf_record(phba, fcf_record);
if (unlikely(rc)) {
lpfc_printf_log(phba, KERN_ERR,
scsi: lpfc: Add an internal trace log buffer The current logging methods typically end up requesting a reproduction with a different logging level set to figure out what happened. This was mainly by design to not clutter the kernel log messages with things that were typically not interesting and the messages themselves could cause other issues. When looking to make a better system, it was seen that in many cases when more data was wanted was when another message, usually at KERN_ERR level, was logged. And in most cases, what the additional logging that was then enabled was typically. Most of these areas fell into the discovery machine. Based on this summary, the following design has been put in place: The driver will maintain an internal log (256 elements of 256 bytes). The "additional logging" messages that are usually enabled in a reproduction will be changed to now log all the time to the internal log. A new logging level is defined - LOG_TRACE_EVENT. When this level is set (it is not by default) and a message marked as KERN_ERR is logged, all the messages in the internal log will be dumped to the kernel log before the KERN_ERR message is logged. There is a timestamp on each message added to the internal log. However, this timestamp is not converted to wall time when logged. The value of the timestamp is solely to give a crude time reference for the messages. Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-06-30 21:50:00 +00:00
LOG_TRACE_EVENT,
"2013 Could not manually add FCF "
"record 0, status %d\n", rc);
rc = -ENODEV;
kfree(fcf_record);
goto out;
}
kfree(fcf_record);
}
/*
* The driver is expected to do FIP/FCF. Call the port
* and get the FCF Table.
*/
spin_lock_irqsave(&phba->hbalock, iflags);
if (phba->hba_flag & FCF_TS_INPROG) {
spin_unlock_irqrestore(&phba->hbalock, iflags);
return;
}
/* This is the initial FCF discovery scan */
phba->fcf.fcf_flag |= FCF_INIT_DISC;
spin_unlock_irqrestore(&phba->hbalock, iflags);
lpfc_printf_log(phba, KERN_INFO, LOG_FIP | LOG_DISCOVERY,
"2778 Start FCF table scan at linkup\n");
rc = lpfc_sli4_fcf_scan_read_fcf_rec(phba,
LPFC_FCOE_FCF_GET_FIRST);
if (rc) {
spin_lock_irqsave(&phba->hbalock, iflags);
phba->fcf.fcf_flag &= ~FCF_INIT_DISC;
spin_unlock_irqrestore(&phba->hbalock, iflags);
goto out;
}
/* Reset FCF roundrobin bmask for new discovery */
lpfc_sli4_clear_fcf_rr_bmask(phba);
}
/* Prepare for LINK up registrations */
memset(phba->os_host_name, 0, sizeof(phba->os_host_name));
scnprintf(phba->os_host_name, sizeof(phba->os_host_name), "%s",
init_utsname()->nodename);
return;
out:
lpfc_vport_set_state(vport, FC_VPORT_FAILED);
scsi: lpfc: Add an internal trace log buffer The current logging methods typically end up requesting a reproduction with a different logging level set to figure out what happened. This was mainly by design to not clutter the kernel log messages with things that were typically not interesting and the messages themselves could cause other issues. When looking to make a better system, it was seen that in many cases when more data was wanted was when another message, usually at KERN_ERR level, was logged. And in most cases, what the additional logging that was then enabled was typically. Most of these areas fell into the discovery machine. Based on this summary, the following design has been put in place: The driver will maintain an internal log (256 elements of 256 bytes). The "additional logging" messages that are usually enabled in a reproduction will be changed to now log all the time to the internal log. A new logging level is defined - LOG_TRACE_EVENT. When this level is set (it is not by default) and a message marked as KERN_ERR is logged, all the messages in the internal log will be dumped to the kernel log before the KERN_ERR message is logged. There is a timestamp on each message added to the internal log. However, this timestamp is not converted to wall time when logged. The value of the timestamp is solely to give a crude time reference for the messages. Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-06-30 21:50:00 +00:00
lpfc_printf_vlog(vport, KERN_ERR, LOG_TRACE_EVENT,
"0263 Discovery Mailbox error: state: 0x%x : x%px x%px\n",
vport->port_state, sparam_mbox, cfglink_mbox);
lpfc_issue_clear_la(phba, vport);
return;
}
static void
lpfc_enable_la(struct lpfc_hba *phba)
{
uint32_t control;
struct lpfc_sli *psli = &phba->sli;
spin_lock_irq(&phba->hbalock);
psli->sli_flag |= LPFC_PROCESS_LA;
if (phba->sli_rev <= LPFC_SLI_REV3) {
control = readl(phba->HCregaddr);
control |= HC_LAINT_ENA;
writel(control, phba->HCregaddr);
readl(phba->HCregaddr); /* flush */
}
spin_unlock_irq(&phba->hbalock);
}
static void
lpfc_mbx_issue_link_down(struct lpfc_hba *phba)
{
lpfc_linkdown(phba);
lpfc_enable_la(phba);
lpfc_unregister_unused_fcf(phba);
/* turn on Link Attention interrupts - no CLEAR_LA needed */
}
/*
2010-11-21 04:11:48 +00:00
* This routine handles processing a READ_TOPOLOGY mailbox
* command upon completion. It is setup in the LPFC_MBOXQ
* as the completion routine when the command is
scsi: lpfc: NVME Initiator: Base modifications NVME Initiator: Base modifications This patch adds base modifications for NVME initiator support. The base modifications consist of: - Formal split of SLI3 rings from SLI-4 WQs (sometimes referred to as rings as well) as implementation now widely varies between the two. - Addition of configuration modes: SCSI initiator only; NVME initiator only; NVME target only; and SCSI and NVME initiator. The configuration mode drives overall adapter configuration, offloads enabled, and resource splits. NVME support is only available on SLI-4 devices and newer fw. - Implements the following based on configuration mode: - Exchange resources are split by protocol; Obviously, if only 1 mode, then no split occurs. Default is 50/50. module attribute allows tuning. - Pools and config parameters are separated per-protocol - Each protocol has it's own set of queues, but share interrupt vectors. SCSI: SLI3 devices have few queues and the original style of queue allocation remains. SLI4 devices piggy back on an "io-channel" concept that eventually needs to merge with scsi-mq/blk-mq support (it is underway). For now, the paradigm continues as it existed prior. io channel allocates N msix and N WQs (N=4 default) and either round robins or uses cpu # modulo N for scheduling. A bunch of module parameters allow the configuration to be tuned. NVME (initiator): Allocates an msix per cpu (or whatever pci_alloc_irq_vectors gets) Allocates a WQ per cpu, and maps the WQs to msix on a WQ # modulo msix vector count basis. Module parameters exist to cap/control the config if desired. - Each protocol has its own buffer and dma pools. I apologize for the size of the patch. Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <james.smart@broadcom.com> ---- Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2017-02-12 21:52:30 +00:00
* handed off to the SLI layer. SLI4 only.
*/
void
2010-11-21 04:11:48 +00:00
lpfc_mbx_cmpl_read_topology(struct lpfc_hba *phba, LPFC_MBOXQ_t *pmb)
{
struct lpfc_vport *vport = pmb->vport;
struct Scsi_Host *shost = lpfc_shost_from_vport(vport);
2010-11-21 04:11:48 +00:00
struct lpfc_mbx_read_top *la;
scsi: lpfc: NVME Initiator: Base modifications NVME Initiator: Base modifications This patch adds base modifications for NVME initiator support. The base modifications consist of: - Formal split of SLI3 rings from SLI-4 WQs (sometimes referred to as rings as well) as implementation now widely varies between the two. - Addition of configuration modes: SCSI initiator only; NVME initiator only; NVME target only; and SCSI and NVME initiator. The configuration mode drives overall adapter configuration, offloads enabled, and resource splits. NVME support is only available on SLI-4 devices and newer fw. - Implements the following based on configuration mode: - Exchange resources are split by protocol; Obviously, if only 1 mode, then no split occurs. Default is 50/50. module attribute allows tuning. - Pools and config parameters are separated per-protocol - Each protocol has it's own set of queues, but share interrupt vectors. SCSI: SLI3 devices have few queues and the original style of queue allocation remains. SLI4 devices piggy back on an "io-channel" concept that eventually needs to merge with scsi-mq/blk-mq support (it is underway). For now, the paradigm continues as it existed prior. io channel allocates N msix and N WQs (N=4 default) and either round robins or uses cpu # modulo N for scheduling. A bunch of module parameters allow the configuration to be tuned. NVME (initiator): Allocates an msix per cpu (or whatever pci_alloc_irq_vectors gets) Allocates a WQ per cpu, and maps the WQs to msix on a WQ # modulo msix vector count basis. Module parameters exist to cap/control the config if desired. - Each protocol has its own buffer and dma pools. I apologize for the size of the patch. Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <james.smart@broadcom.com> ---- Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2017-02-12 21:52:30 +00:00
struct lpfc_sli_ring *pring;
MAILBOX_t *mb = &pmb->u.mb;
struct lpfc_dmabuf *mp = (struct lpfc_dmabuf *)(pmb->ctx_buf);
uint8_t attn_type;
unsigned long iflags;
/* Unblock ELS traffic */
scsi: lpfc: NVME Initiator: Base modifications NVME Initiator: Base modifications This patch adds base modifications for NVME initiator support. The base modifications consist of: - Formal split of SLI3 rings from SLI-4 WQs (sometimes referred to as rings as well) as implementation now widely varies between the two. - Addition of configuration modes: SCSI initiator only; NVME initiator only; NVME target only; and SCSI and NVME initiator. The configuration mode drives overall adapter configuration, offloads enabled, and resource splits. NVME support is only available on SLI-4 devices and newer fw. - Implements the following based on configuration mode: - Exchange resources are split by protocol; Obviously, if only 1 mode, then no split occurs. Default is 50/50. module attribute allows tuning. - Pools and config parameters are separated per-protocol - Each protocol has it's own set of queues, but share interrupt vectors. SCSI: SLI3 devices have few queues and the original style of queue allocation remains. SLI4 devices piggy back on an "io-channel" concept that eventually needs to merge with scsi-mq/blk-mq support (it is underway). For now, the paradigm continues as it existed prior. io channel allocates N msix and N WQs (N=4 default) and either round robins or uses cpu # modulo N for scheduling. A bunch of module parameters allow the configuration to be tuned. NVME (initiator): Allocates an msix per cpu (or whatever pci_alloc_irq_vectors gets) Allocates a WQ per cpu, and maps the WQs to msix on a WQ # modulo msix vector count basis. Module parameters exist to cap/control the config if desired. - Each protocol has its own buffer and dma pools. I apologize for the size of the patch. Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <james.smart@broadcom.com> ---- Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2017-02-12 21:52:30 +00:00
pring = lpfc_phba_elsring(phba);
if (pring)
pring->flag &= ~LPFC_STOP_IOCB_EVENT;
scsi: lpfc: NVME Initiator: Base modifications NVME Initiator: Base modifications This patch adds base modifications for NVME initiator support. The base modifications consist of: - Formal split of SLI3 rings from SLI-4 WQs (sometimes referred to as rings as well) as implementation now widely varies between the two. - Addition of configuration modes: SCSI initiator only; NVME initiator only; NVME target only; and SCSI and NVME initiator. The configuration mode drives overall adapter configuration, offloads enabled, and resource splits. NVME support is only available on SLI-4 devices and newer fw. - Implements the following based on configuration mode: - Exchange resources are split by protocol; Obviously, if only 1 mode, then no split occurs. Default is 50/50. module attribute allows tuning. - Pools and config parameters are separated per-protocol - Each protocol has it's own set of queues, but share interrupt vectors. SCSI: SLI3 devices have few queues and the original style of queue allocation remains. SLI4 devices piggy back on an "io-channel" concept that eventually needs to merge with scsi-mq/blk-mq support (it is underway). For now, the paradigm continues as it existed prior. io channel allocates N msix and N WQs (N=4 default) and either round robins or uses cpu # modulo N for scheduling. A bunch of module parameters allow the configuration to be tuned. NVME (initiator): Allocates an msix per cpu (or whatever pci_alloc_irq_vectors gets) Allocates a WQ per cpu, and maps the WQs to msix on a WQ # modulo msix vector count basis. Module parameters exist to cap/control the config if desired. - Each protocol has its own buffer and dma pools. I apologize for the size of the patch. Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <james.smart@broadcom.com> ---- Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2017-02-12 21:52:30 +00:00
/* Check for error */
if (mb->mbxStatus) {
lpfc_printf_log(phba, KERN_INFO, LOG_LINK_EVENT,
"1307 READ_LA mbox error x%x state x%x\n",
mb->mbxStatus, vport->port_state);
lpfc_mbx_issue_link_down(phba);
phba->link_state = LPFC_HBA_ERROR;
2010-11-21 04:11:48 +00:00
goto lpfc_mbx_cmpl_read_topology_free_mbuf;
}
2010-11-21 04:11:48 +00:00
la = (struct lpfc_mbx_read_top *) &pmb->u.mb.un.varReadTop;
attn_type = bf_get(lpfc_mbx_read_top_att_type, la);
memcpy(&phba->alpa_map[0], mp->virt, 128);
spin_lock_irqsave(shost->host_lock, iflags);
2010-11-21 04:11:48 +00:00
if (bf_get(lpfc_mbx_read_top_pb, la))
vport->fc_flag |= FC_BYPASSED_MODE;
else
vport->fc_flag &= ~FC_BYPASSED_MODE;
spin_unlock_irqrestore(shost->host_lock, iflags);
if (phba->fc_eventTag <= la->eventTag) {
phba->fc_stat.LinkMultiEvent++;
if (attn_type == LPFC_ATT_LINK_UP)
if (phba->fc_eventTag != 0)
lpfc_linkdown(phba);
}
phba->fc_eventTag = la->eventTag;
phba->link_events++;
if (attn_type == LPFC_ATT_LINK_UP) {
phba->fc_stat.LinkUp++;
if (phba->link_flag & LS_LOOPBACK_MODE) {
lpfc_printf_log(phba, KERN_ERR, LOG_LINK_EVENT,
"1306 Link Up Event in loop back mode "
"x%x received Data: x%x x%x x%x x%x\n",
la->eventTag, phba->fc_eventTag,
2010-11-21 04:11:48 +00:00
bf_get(lpfc_mbx_read_top_alpa_granted,
la),
bf_get(lpfc_mbx_read_top_link_spd, la),
phba->alpa_map[0]);
} else {
lpfc_printf_log(phba, KERN_ERR, LOG_LINK_EVENT,
"1303 Link Up Event x%x received "
"Data: x%x x%x x%x x%x x%x\n",
la->eventTag, phba->fc_eventTag,
2010-11-21 04:11:48 +00:00
bf_get(lpfc_mbx_read_top_alpa_granted,
la),
bf_get(lpfc_mbx_read_top_link_spd, la),
phba->alpa_map[0],
bf_get(lpfc_mbx_read_top_fa, la));
}
lpfc_mbx_process_link_up(phba, la);
if (phba->cmf_active_mode != LPFC_CFG_OFF)
lpfc_cmf_signal_init(phba);
if (phba->lmt & LMT_64Gb)
lpfc_read_lds_params(phba);
} else if (attn_type == LPFC_ATT_LINK_DOWN ||
attn_type == LPFC_ATT_UNEXP_WWPN) {
phba->fc_stat.LinkDown++;
if (phba->link_flag & LS_LOOPBACK_MODE)
lpfc_printf_log(phba, KERN_ERR, LOG_LINK_EVENT,
"1308 Link Down Event in loop back mode "
"x%x received "
"Data: x%x x%x x%x\n",
la->eventTag, phba->fc_eventTag,
phba->pport->port_state, vport->fc_flag);
else if (attn_type == LPFC_ATT_UNEXP_WWPN)
lpfc_printf_log(phba, KERN_ERR, LOG_LINK_EVENT,
"1313 Link Down Unexpected FA WWPN Event x%x "
"received Data: x%x x%x x%x x%x\n",
la->eventTag, phba->fc_eventTag,
phba->pport->port_state, vport->fc_flag,
bf_get(lpfc_mbx_read_top_fa, la));
else
lpfc_printf_log(phba, KERN_ERR, LOG_LINK_EVENT,
"1305 Link Down Event x%x received "
"Data: x%x x%x x%x x%x\n",
la->eventTag, phba->fc_eventTag,
phba->pport->port_state, vport->fc_flag,
2010-11-21 04:11:48 +00:00
bf_get(lpfc_mbx_read_top_fa, la));
lpfc_mbx_issue_link_down(phba);
}
if ((phba->sli_rev < LPFC_SLI_REV4) &&
bf_get(lpfc_mbx_read_top_fa, la))
lpfc_printf_log(phba, KERN_INFO, LOG_LINK_EVENT,
2010-11-21 04:11:48 +00:00
"1311 fa %d\n",
bf_get(lpfc_mbx_read_top_fa, la));
2010-11-21 04:11:48 +00:00
lpfc_mbx_cmpl_read_topology_free_mbuf:
lpfc_mbox_rsrc_cleanup(phba, pmb, MBOX_THD_UNLOCKED);
}
/*
* This routine handles processing a REG_LOGIN mailbox
* command upon completion. It is setup in the LPFC_MBOXQ
* as the completion routine when the command is
* handed off to the SLI layer.
*/
void
lpfc_mbx_cmpl_reg_login(struct lpfc_hba *phba, LPFC_MBOXQ_t *pmb)
{
struct lpfc_vport *vport = pmb->vport;
struct lpfc_dmabuf *mp = (struct lpfc_dmabuf *)pmb->ctx_buf;
struct lpfc_nodelist *ndlp = (struct lpfc_nodelist *)pmb->ctx_ndlp;
/* The driver calls the state machine with the pmb pointer
* but wants to make sure a stale ctx_buf isn't acted on.
* The ctx_buf is restored later and cleaned up.
*/
pmb->ctx_buf = NULL;
pmb->ctx_ndlp = NULL;
lpfc_printf_vlog(vport, KERN_INFO, LOG_SLI | LOG_NODE | LOG_DISCOVERY,
scsi: lpfc: Rework remote port ref counting and node freeing When a remote port is disconnected and disappears, its node structure (ndlp) stays allocated and on a vport node list. While on the list it can be matched, thus requires validation checks on state to be added in numerous code paths. If the node comes back, its possible for there to be multiple node structures for the same device on the vport node list. There is no reason to keep the node structure around after it is no longer in existence, and the current implementation creates problems for itself (multiple nodes) and lots of unnecessary code for state validation. Additionally, the reference taking on the node structure didn't follow the normal model used by the kernel kref api. It included lots of odd logic to match state with reference count. The combination of this odd logic plus the way it was implicitly used in the discovery engine made its reference taking implementation suspect and extremely hard to follow. Change the driver such that the reference taking routines are now normal ref increments/decrements and callout on refcount=0. With this in place, the rework can be done such that the node structure is fully removed and deallocated when the remote port no longer exists and all references are removed. This removal logic, and the basic ref counting are intrically tied, thus in a single patch. Link: https://lore.kernel.org/r/20201115192646.12977-2-james.smart@broadcom.com Co-developed-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <james.smart@broadcom.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-11-15 19:26:30 +00:00
"0002 rpi:%x DID:%x flg:%x %d x%px\n",
ndlp->nlp_rpi, ndlp->nlp_DID, ndlp->nlp_flag,
kref_read(&ndlp->kref),
scsi: lpfc: Rework remote port ref counting and node freeing When a remote port is disconnected and disappears, its node structure (ndlp) stays allocated and on a vport node list. While on the list it can be matched, thus requires validation checks on state to be added in numerous code paths. If the node comes back, its possible for there to be multiple node structures for the same device on the vport node list. There is no reason to keep the node structure around after it is no longer in existence, and the current implementation creates problems for itself (multiple nodes) and lots of unnecessary code for state validation. Additionally, the reference taking on the node structure didn't follow the normal model used by the kernel kref api. It included lots of odd logic to match state with reference count. The combination of this odd logic plus the way it was implicitly used in the discovery engine made its reference taking implementation suspect and extremely hard to follow. Change the driver such that the reference taking routines are now normal ref increments/decrements and callout on refcount=0. With this in place, the rework can be done such that the node structure is fully removed and deallocated when the remote port no longer exists and all references are removed. This removal logic, and the basic ref counting are intrically tied, thus in a single patch. Link: https://lore.kernel.org/r/20201115192646.12977-2-james.smart@broadcom.com Co-developed-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <james.smart@broadcom.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-11-15 19:26:30 +00:00
ndlp);
if (ndlp->nlp_flag & NLP_REG_LOGIN_SEND)
ndlp->nlp_flag &= ~NLP_REG_LOGIN_SEND;
if (ndlp->nlp_flag & NLP_IGNR_REG_CMPL ||
ndlp->nlp_state != NLP_STE_REG_LOGIN_ISSUE) {
/* We rcvd a rscn after issuing this
* mbox reg login, we may have cycled
* back through the state and be
* back at reg login state so this
* mbox needs to be ignored becase
* there is another reg login in
* process.
*/
spin_lock_irq(&ndlp->lock);
ndlp->nlp_flag &= ~NLP_IGNR_REG_CMPL;
spin_unlock_irq(&ndlp->lock);
scsi: lpfc: NVME Initiator: Base modifications NVME Initiator: Base modifications This patch adds base modifications for NVME initiator support. The base modifications consist of: - Formal split of SLI3 rings from SLI-4 WQs (sometimes referred to as rings as well) as implementation now widely varies between the two. - Addition of configuration modes: SCSI initiator only; NVME initiator only; NVME target only; and SCSI and NVME initiator. The configuration mode drives overall adapter configuration, offloads enabled, and resource splits. NVME support is only available on SLI-4 devices and newer fw. - Implements the following based on configuration mode: - Exchange resources are split by protocol; Obviously, if only 1 mode, then no split occurs. Default is 50/50. module attribute allows tuning. - Pools and config parameters are separated per-protocol - Each protocol has it's own set of queues, but share interrupt vectors. SCSI: SLI3 devices have few queues and the original style of queue allocation remains. SLI4 devices piggy back on an "io-channel" concept that eventually needs to merge with scsi-mq/blk-mq support (it is underway). For now, the paradigm continues as it existed prior. io channel allocates N msix and N WQs (N=4 default) and either round robins or uses cpu # modulo N for scheduling. A bunch of module parameters allow the configuration to be tuned. NVME (initiator): Allocates an msix per cpu (or whatever pci_alloc_irq_vectors gets) Allocates a WQ per cpu, and maps the WQs to msix on a WQ # modulo msix vector count basis. Module parameters exist to cap/control the config if desired. - Each protocol has its own buffer and dma pools. I apologize for the size of the patch. Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <james.smart@broadcom.com> ---- Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2017-02-12 21:52:30 +00:00
/*
* We cannot leave the RPI registered because
* if we go thru discovery again for this ndlp
* a subsequent REG_RPI will fail.
*/
ndlp->nlp_flag |= NLP_RPI_REGISTERED;
lpfc_unreg_rpi(vport, ndlp);
}
/* Call state machine */
lpfc_disc_state_machine(vport, ndlp, pmb, NLP_EVT_CMPL_REG_LOGIN);
pmb->ctx_buf = mp;
lpfc_mbox_rsrc_cleanup(phba, pmb, MBOX_THD_UNLOCKED);
/* decrement the node reference count held for this callback
* function.
*/
lpfc_nlp_put(ndlp);
return;
}
static void
lpfc_mbx_cmpl_unreg_vpi(struct lpfc_hba *phba, LPFC_MBOXQ_t *pmb)
{
MAILBOX_t *mb = &pmb->u.mb;
struct lpfc_vport *vport = pmb->vport;
struct Scsi_Host *shost = lpfc_shost_from_vport(vport);
switch (mb->mbxStatus) {
case 0x0011:
case 0x0020:
lpfc_printf_vlog(vport, KERN_INFO, LOG_NODE,
"0911 cmpl_unreg_vpi, mb status = 0x%x\n",
mb->mbxStatus);
break;
/* If VPI is busy, reset the HBA */
case 0x9700:
scsi: lpfc: Add an internal trace log buffer The current logging methods typically end up requesting a reproduction with a different logging level set to figure out what happened. This was mainly by design to not clutter the kernel log messages with things that were typically not interesting and the messages themselves could cause other issues. When looking to make a better system, it was seen that in many cases when more data was wanted was when another message, usually at KERN_ERR level, was logged. And in most cases, what the additional logging that was then enabled was typically. Most of these areas fell into the discovery machine. Based on this summary, the following design has been put in place: The driver will maintain an internal log (256 elements of 256 bytes). The "additional logging" messages that are usually enabled in a reproduction will be changed to now log all the time to the internal log. A new logging level is defined - LOG_TRACE_EVENT. When this level is set (it is not by default) and a message marked as KERN_ERR is logged, all the messages in the internal log will be dumped to the kernel log before the KERN_ERR message is logged. There is a timestamp on each message added to the internal log. However, this timestamp is not converted to wall time when logged. The value of the timestamp is solely to give a crude time reference for the messages. Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-06-30 21:50:00 +00:00
lpfc_printf_vlog(vport, KERN_ERR, LOG_TRACE_EVENT,
"2798 Unreg_vpi failed vpi 0x%x, mb status = 0x%x\n",
vport->vpi, mb->mbxStatus);
if (!(phba->pport->load_flag & FC_UNLOADING))
lpfc_workq_post_event(phba, NULL, NULL,
LPFC_EVT_RESET_HBA);
}
spin_lock_irq(shost->host_lock);
vport->vpi_state &= ~LPFC_VPI_REGISTERED;
vport->fc_flag |= FC_VPORT_NEEDS_REG_VPI;
spin_unlock_irq(shost->host_lock);
mempool_free(pmb, phba->mbox_mem_pool);
lpfc_cleanup_vports_rrqs(vport, NULL);
/*
* This shost reference might have been taken at the beginning of
* lpfc_vport_delete()
*/
if ((vport->load_flag & FC_UNLOADING) && (vport != phba->pport))
scsi_host_put(shost);
}
int
lpfc_mbx_unreg_vpi(struct lpfc_vport *vport)
{
struct lpfc_hba *phba = vport->phba;
LPFC_MBOXQ_t *mbox;
int rc;
mbox = mempool_alloc(phba->mbox_mem_pool, GFP_KERNEL);
if (!mbox)
return 1;
lpfc_unreg_vpi(phba, vport->vpi, mbox);
mbox->vport = vport;
mbox->mbox_cmpl = lpfc_mbx_cmpl_unreg_vpi;
rc = lpfc_sli_issue_mbox(phba, mbox, MBX_NOWAIT);
if (rc == MBX_NOT_FINISHED) {
scsi: lpfc: Add an internal trace log buffer The current logging methods typically end up requesting a reproduction with a different logging level set to figure out what happened. This was mainly by design to not clutter the kernel log messages with things that were typically not interesting and the messages themselves could cause other issues. When looking to make a better system, it was seen that in many cases when more data was wanted was when another message, usually at KERN_ERR level, was logged. And in most cases, what the additional logging that was then enabled was typically. Most of these areas fell into the discovery machine. Based on this summary, the following design has been put in place: The driver will maintain an internal log (256 elements of 256 bytes). The "additional logging" messages that are usually enabled in a reproduction will be changed to now log all the time to the internal log. A new logging level is defined - LOG_TRACE_EVENT. When this level is set (it is not by default) and a message marked as KERN_ERR is logged, all the messages in the internal log will be dumped to the kernel log before the KERN_ERR message is logged. There is a timestamp on each message added to the internal log. However, this timestamp is not converted to wall time when logged. The value of the timestamp is solely to give a crude time reference for the messages. Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-06-30 21:50:00 +00:00
lpfc_printf_vlog(vport, KERN_ERR, LOG_TRACE_EVENT,
"1800 Could not issue unreg_vpi\n");
mempool_free(mbox, phba->mbox_mem_pool);
return rc;
}
return 0;
}
static void
lpfc_mbx_cmpl_reg_vpi(struct lpfc_hba *phba, LPFC_MBOXQ_t *pmb)
{
struct lpfc_vport *vport = pmb->vport;
struct Scsi_Host *shost = lpfc_shost_from_vport(vport);
MAILBOX_t *mb = &pmb->u.mb;
switch (mb->mbxStatus) {
case 0x0011:
case 0x9601:
case 0x9602:
lpfc_printf_vlog(vport, KERN_INFO, LOG_NODE,
"0912 cmpl_reg_vpi, mb status = 0x%x\n",
mb->mbxStatus);
lpfc_vport_set_state(vport, FC_VPORT_FAILED);
spin_lock_irq(shost->host_lock);
vport->fc_flag &= ~(FC_FABRIC | FC_PUBLIC_LOOP);
spin_unlock_irq(shost->host_lock);
vport->fc_myDID = 0;
if ((vport->cfg_enable_fc4_type == LPFC_ENABLE_BOTH) ||
(vport->cfg_enable_fc4_type == LPFC_ENABLE_NVME)) {
if (phba->nvmet_support)
lpfc_nvmet_update_targetport(phba);
else
lpfc_nvme_update_localport(vport);
}
goto out;
}
spin_lock_irq(shost->host_lock);
vport->vpi_state |= LPFC_VPI_REGISTERED;
vport->fc_flag &= ~FC_VPORT_NEEDS_REG_VPI;
spin_unlock_irq(shost->host_lock);
vport->num_disc_nodes = 0;
/* go thru NPR list and issue ELS PLOGIs */
if (vport->fc_npr_cnt)
lpfc_els_disc_plogi(vport);
if (!vport->num_disc_nodes) {
spin_lock_irq(shost->host_lock);
vport->fc_flag &= ~FC_NDISC_ACTIVE;
spin_unlock_irq(shost->host_lock);
lpfc_can_disctmo(vport);
}
vport->port_state = LPFC_VPORT_READY;
out:
mempool_free(pmb, phba->mbox_mem_pool);
return;
}
/**
* lpfc_create_static_vport - Read HBA config region to create static vports.
* @phba: pointer to lpfc hba data structure.
*
* This routine issue a DUMP mailbox command for config region 22 to get
* the list of static vports to be created. The function create vports
* based on the information returned from the HBA.
**/
void
lpfc_create_static_vport(struct lpfc_hba *phba)
{
LPFC_MBOXQ_t *pmb = NULL;
MAILBOX_t *mb;
struct static_vport_info *vport_info;
int mbx_wait_rc = 0, i;
struct fc_vport_identifiers vport_id;
struct fc_vport *new_fc_vport;
struct Scsi_Host *shost;
struct lpfc_vport *vport;
uint16_t offset = 0;
uint8_t *vport_buff;
struct lpfc_dmabuf *mp;
uint32_t byte_count = 0;
pmb = mempool_alloc(phba->mbox_mem_pool, GFP_KERNEL);
if (!pmb) {
scsi: lpfc: Add an internal trace log buffer The current logging methods typically end up requesting a reproduction with a different logging level set to figure out what happened. This was mainly by design to not clutter the kernel log messages with things that were typically not interesting and the messages themselves could cause other issues. When looking to make a better system, it was seen that in many cases when more data was wanted was when another message, usually at KERN_ERR level, was logged. And in most cases, what the additional logging that was then enabled was typically. Most of these areas fell into the discovery machine. Based on this summary, the following design has been put in place: The driver will maintain an internal log (256 elements of 256 bytes). The "additional logging" messages that are usually enabled in a reproduction will be changed to now log all the time to the internal log. A new logging level is defined - LOG_TRACE_EVENT. When this level is set (it is not by default) and a message marked as KERN_ERR is logged, all the messages in the internal log will be dumped to the kernel log before the KERN_ERR message is logged. There is a timestamp on each message added to the internal log. However, this timestamp is not converted to wall time when logged. The value of the timestamp is solely to give a crude time reference for the messages. Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-06-30 21:50:00 +00:00
lpfc_printf_log(phba, KERN_ERR, LOG_TRACE_EVENT,
"0542 lpfc_create_static_vport failed to"
" allocate mailbox memory\n");
return;
}
memset(pmb, 0, sizeof(LPFC_MBOXQ_t));
mb = &pmb->u.mb;
vport_info = kzalloc(sizeof(struct static_vport_info), GFP_KERNEL);
if (!vport_info) {
scsi: lpfc: Add an internal trace log buffer The current logging methods typically end up requesting a reproduction with a different logging level set to figure out what happened. This was mainly by design to not clutter the kernel log messages with things that were typically not interesting and the messages themselves could cause other issues. When looking to make a better system, it was seen that in many cases when more data was wanted was when another message, usually at KERN_ERR level, was logged. And in most cases, what the additional logging that was then enabled was typically. Most of these areas fell into the discovery machine. Based on this summary, the following design has been put in place: The driver will maintain an internal log (256 elements of 256 bytes). The "additional logging" messages that are usually enabled in a reproduction will be changed to now log all the time to the internal log. A new logging level is defined - LOG_TRACE_EVENT. When this level is set (it is not by default) and a message marked as KERN_ERR is logged, all the messages in the internal log will be dumped to the kernel log before the KERN_ERR message is logged. There is a timestamp on each message added to the internal log. However, this timestamp is not converted to wall time when logged. The value of the timestamp is solely to give a crude time reference for the messages. Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-06-30 21:50:00 +00:00
lpfc_printf_log(phba, KERN_ERR, LOG_TRACE_EVENT,
"0543 lpfc_create_static_vport failed to"
" allocate vport_info\n");
mempool_free(pmb, phba->mbox_mem_pool);
return;
}
vport_buff = (uint8_t *) vport_info;
do {
/* While loop iteration forces a free dma buffer from
* the previous loop because the mbox is reused and
* the dump routine is a single-use construct.
*/
if (pmb->ctx_buf) {
mp = (struct lpfc_dmabuf *)pmb->ctx_buf;
lpfc_mbuf_free(phba, mp->virt, mp->phys);
kfree(mp);
pmb->ctx_buf = NULL;
}
if (lpfc_dump_static_vport(phba, pmb, offset))
goto out;
pmb->vport = phba->pport;
mbx_wait_rc = lpfc_sli_issue_mbox_wait(phba, pmb,
LPFC_MBOX_TMO);
if ((mbx_wait_rc != MBX_SUCCESS) || mb->mbxStatus) {
lpfc_printf_log(phba, KERN_WARNING, LOG_INIT,
"0544 lpfc_create_static_vport failed to"
" issue dump mailbox command ret 0x%x "
"status 0x%x\n",
mbx_wait_rc, mb->mbxStatus);
goto out;
}
if (phba->sli_rev == LPFC_SLI_REV4) {
byte_count = pmb->u.mqe.un.mb_words[5];
mp = (struct lpfc_dmabuf *)pmb->ctx_buf;
if (byte_count > sizeof(struct static_vport_info) -
offset)
byte_count = sizeof(struct static_vport_info)
- offset;
memcpy(vport_buff + offset, mp->virt, byte_count);
offset += byte_count;
} else {
if (mb->un.varDmp.word_cnt >
sizeof(struct static_vport_info) - offset)
mb->un.varDmp.word_cnt =
sizeof(struct static_vport_info)
- offset;
byte_count = mb->un.varDmp.word_cnt;
lpfc_sli_pcimem_bcopy(((uint8_t *)mb) + DMP_RSP_OFFSET,
vport_buff + offset,
byte_count);
offset += byte_count;
}
} while (byte_count &&
offset < sizeof(struct static_vport_info));
if ((le32_to_cpu(vport_info->signature) != VPORT_INFO_SIG) ||
((le32_to_cpu(vport_info->rev) & VPORT_INFO_REV_MASK)
!= VPORT_INFO_REV)) {
scsi: lpfc: Add an internal trace log buffer The current logging methods typically end up requesting a reproduction with a different logging level set to figure out what happened. This was mainly by design to not clutter the kernel log messages with things that were typically not interesting and the messages themselves could cause other issues. When looking to make a better system, it was seen that in many cases when more data was wanted was when another message, usually at KERN_ERR level, was logged. And in most cases, what the additional logging that was then enabled was typically. Most of these areas fell into the discovery machine. Based on this summary, the following design has been put in place: The driver will maintain an internal log (256 elements of 256 bytes). The "additional logging" messages that are usually enabled in a reproduction will be changed to now log all the time to the internal log. A new logging level is defined - LOG_TRACE_EVENT. When this level is set (it is not by default) and a message marked as KERN_ERR is logged, all the messages in the internal log will be dumped to the kernel log before the KERN_ERR message is logged. There is a timestamp on each message added to the internal log. However, this timestamp is not converted to wall time when logged. The value of the timestamp is solely to give a crude time reference for the messages. Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-06-30 21:50:00 +00:00
lpfc_printf_log(phba, KERN_ERR, LOG_TRACE_EVENT,
"0545 lpfc_create_static_vport bad"
" information header 0x%x 0x%x\n",
le32_to_cpu(vport_info->signature),
le32_to_cpu(vport_info->rev) &
VPORT_INFO_REV_MASK);
goto out;
}
shost = lpfc_shost_from_vport(phba->pport);
for (i = 0; i < MAX_STATIC_VPORT_COUNT; i++) {
memset(&vport_id, 0, sizeof(vport_id));
vport_id.port_name = wwn_to_u64(vport_info->vport_list[i].wwpn);
vport_id.node_name = wwn_to_u64(vport_info->vport_list[i].wwnn);
if (!vport_id.port_name || !vport_id.node_name)
continue;
vport_id.roles = FC_PORT_ROLE_FCP_INITIATOR;
vport_id.vport_type = FC_PORTTYPE_NPIV;
vport_id.disable = false;
new_fc_vport = fc_vport_create(shost, 0, &vport_id);
if (!new_fc_vport) {
lpfc_printf_log(phba, KERN_WARNING, LOG_INIT,
"0546 lpfc_create_static_vport failed to"
" create vport\n");
continue;
}
vport = *(struct lpfc_vport **)new_fc_vport->dd_data;
vport->vport_flag |= STATIC_VPORT;
}
out:
kfree(vport_info);
if (mbx_wait_rc != MBX_TIMEOUT)
lpfc_mbox_rsrc_cleanup(phba, pmb, MBOX_THD_UNLOCKED);
}
/*
* This routine handles processing a Fabric REG_LOGIN mailbox
* command upon completion. It is setup in the LPFC_MBOXQ
* as the completion routine when the command is
* handed off to the SLI layer.
*/
void
lpfc_mbx_cmpl_fabric_reg_login(struct lpfc_hba *phba, LPFC_MBOXQ_t *pmb)
{
struct lpfc_vport *vport = pmb->vport;
MAILBOX_t *mb = &pmb->u.mb;
struct lpfc_nodelist *ndlp = (struct lpfc_nodelist *)pmb->ctx_ndlp;
struct Scsi_Host *shost;
pmb->ctx_ndlp = NULL;
if (mb->mbxStatus) {
scsi: lpfc: Add an internal trace log buffer The current logging methods typically end up requesting a reproduction with a different logging level set to figure out what happened. This was mainly by design to not clutter the kernel log messages with things that were typically not interesting and the messages themselves could cause other issues. When looking to make a better system, it was seen that in many cases when more data was wanted was when another message, usually at KERN_ERR level, was logged. And in most cases, what the additional logging that was then enabled was typically. Most of these areas fell into the discovery machine. Based on this summary, the following design has been put in place: The driver will maintain an internal log (256 elements of 256 bytes). The "additional logging" messages that are usually enabled in a reproduction will be changed to now log all the time to the internal log. A new logging level is defined - LOG_TRACE_EVENT. When this level is set (it is not by default) and a message marked as KERN_ERR is logged, all the messages in the internal log will be dumped to the kernel log before the KERN_ERR message is logged. There is a timestamp on each message added to the internal log. However, this timestamp is not converted to wall time when logged. The value of the timestamp is solely to give a crude time reference for the messages. Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-06-30 21:50:00 +00:00
lpfc_printf_vlog(vport, KERN_ERR, LOG_TRACE_EVENT,
"0258 Register Fabric login error: 0x%x\n",
mb->mbxStatus);
lpfc_mbox_rsrc_cleanup(phba, pmb, MBOX_THD_UNLOCKED);
2010-11-21 04:11:48 +00:00
if (phba->fc_topology == LPFC_TOPOLOGY_LOOP) {
/* FLOGI failed, use loop map to make discovery list */
lpfc_disc_list_loopmap(vport);
/* Start discovery */
lpfc_disc_start(vport);
/* Decrement the reference count to ndlp after the
* reference to the ndlp are done.
*/
lpfc_nlp_put(ndlp);
return;
}
lpfc_vport_set_state(vport, FC_VPORT_FAILED);
/* Decrement the reference count to ndlp after the reference
* to the ndlp are done.
*/
lpfc_nlp_put(ndlp);
return;
}
if (phba->sli_rev < LPFC_SLI_REV4)
ndlp->nlp_rpi = mb->un.varWords[0];
ndlp->nlp_flag |= NLP_RPI_REGISTERED;
ndlp->nlp_type |= NLP_FABRIC;
lpfc_nlp_set_state(vport, ndlp, NLP_STE_UNMAPPED_NODE);
if (vport->port_state == LPFC_FABRIC_CFG_LINK) {
/* when physical port receive logo donot start
* vport discovery */
if (!(vport->fc_flag & FC_LOGO_RCVD_DID_CHNG))
lpfc_start_fdiscs(phba);
else {
shost = lpfc_shost_from_vport(vport);
spin_lock_irq(shost->host_lock);
vport->fc_flag &= ~FC_LOGO_RCVD_DID_CHNG ;
spin_unlock_irq(shost->host_lock);
}
lpfc_do_scr_ns_plogi(phba, vport);
}
lpfc_mbox_rsrc_cleanup(phba, pmb, MBOX_THD_UNLOCKED);
/* Drop the reference count from the mbox at the end after
* all the current reference to the ndlp have been done.
*/
lpfc_nlp_put(ndlp);
return;
}
/*
* This routine will issue a GID_FT for each FC4 Type supported
* by the driver. ALL GID_FTs must complete before discovery is started.
*/
int
lpfc_issue_gidft(struct lpfc_vport *vport)
{
/* Good status, issue CT Request to NameServer */
if ((vport->cfg_enable_fc4_type == LPFC_ENABLE_BOTH) ||
(vport->cfg_enable_fc4_type == LPFC_ENABLE_FCP)) {
if (lpfc_ns_cmd(vport, SLI_CTNS_GID_FT, 0, SLI_CTPT_FCP)) {
/* Cannot issue NameServer FCP Query, so finish up
* discovery
*/
scsi: lpfc: Add an internal trace log buffer The current logging methods typically end up requesting a reproduction with a different logging level set to figure out what happened. This was mainly by design to not clutter the kernel log messages with things that were typically not interesting and the messages themselves could cause other issues. When looking to make a better system, it was seen that in many cases when more data was wanted was when another message, usually at KERN_ERR level, was logged. And in most cases, what the additional logging that was then enabled was typically. Most of these areas fell into the discovery machine. Based on this summary, the following design has been put in place: The driver will maintain an internal log (256 elements of 256 bytes). The "additional logging" messages that are usually enabled in a reproduction will be changed to now log all the time to the internal log. A new logging level is defined - LOG_TRACE_EVENT. When this level is set (it is not by default) and a message marked as KERN_ERR is logged, all the messages in the internal log will be dumped to the kernel log before the KERN_ERR message is logged. There is a timestamp on each message added to the internal log. However, this timestamp is not converted to wall time when logged. The value of the timestamp is solely to give a crude time reference for the messages. Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-06-30 21:50:00 +00:00
lpfc_printf_vlog(vport, KERN_ERR,
LOG_TRACE_EVENT,
"0604 %s FC TYPE %x %s\n",
"Failed to issue GID_FT to ",
FC_TYPE_FCP,
"Finishing discovery.");
return 0;
}
vport->gidft_inp++;
}
if ((vport->cfg_enable_fc4_type == LPFC_ENABLE_BOTH) ||
(vport->cfg_enable_fc4_type == LPFC_ENABLE_NVME)) {
if (lpfc_ns_cmd(vport, SLI_CTNS_GID_FT, 0, SLI_CTPT_NVME)) {
/* Cannot issue NameServer NVME Query, so finish up
* discovery
*/
scsi: lpfc: Add an internal trace log buffer The current logging methods typically end up requesting a reproduction with a different logging level set to figure out what happened. This was mainly by design to not clutter the kernel log messages with things that were typically not interesting and the messages themselves could cause other issues. When looking to make a better system, it was seen that in many cases when more data was wanted was when another message, usually at KERN_ERR level, was logged. And in most cases, what the additional logging that was then enabled was typically. Most of these areas fell into the discovery machine. Based on this summary, the following design has been put in place: The driver will maintain an internal log (256 elements of 256 bytes). The "additional logging" messages that are usually enabled in a reproduction will be changed to now log all the time to the internal log. A new logging level is defined - LOG_TRACE_EVENT. When this level is set (it is not by default) and a message marked as KERN_ERR is logged, all the messages in the internal log will be dumped to the kernel log before the KERN_ERR message is logged. There is a timestamp on each message added to the internal log. However, this timestamp is not converted to wall time when logged. The value of the timestamp is solely to give a crude time reference for the messages. Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-06-30 21:50:00 +00:00
lpfc_printf_vlog(vport, KERN_ERR,
LOG_TRACE_EVENT,
"0605 %s FC_TYPE %x %s %d\n",
"Failed to issue GID_FT to ",
FC_TYPE_NVME,
"Finishing discovery: gidftinp ",
vport->gidft_inp);
if (vport->gidft_inp == 0)
return 0;
} else
vport->gidft_inp++;
}
return vport->gidft_inp;
}
/**
* lpfc_issue_gidpt - issue a GID_PT for all N_Ports
* @vport: The virtual port for which this call is being executed.
*
* This routine will issue a GID_PT to get a list of all N_Ports
*
* Return value :
* 0 - Failure to issue a GID_PT
* 1 - GID_PT issued
**/
int
lpfc_issue_gidpt(struct lpfc_vport *vport)
{
/* Good status, issue CT Request to NameServer */
if (lpfc_ns_cmd(vport, SLI_CTNS_GID_PT, 0, GID_PT_N_PORT)) {
/* Cannot issue NameServer FCP Query, so finish up
* discovery
*/
scsi: lpfc: Add an internal trace log buffer The current logging methods typically end up requesting a reproduction with a different logging level set to figure out what happened. This was mainly by design to not clutter the kernel log messages with things that were typically not interesting and the messages themselves could cause other issues. When looking to make a better system, it was seen that in many cases when more data was wanted was when another message, usually at KERN_ERR level, was logged. And in most cases, what the additional logging that was then enabled was typically. Most of these areas fell into the discovery machine. Based on this summary, the following design has been put in place: The driver will maintain an internal log (256 elements of 256 bytes). The "additional logging" messages that are usually enabled in a reproduction will be changed to now log all the time to the internal log. A new logging level is defined - LOG_TRACE_EVENT. When this level is set (it is not by default) and a message marked as KERN_ERR is logged, all the messages in the internal log will be dumped to the kernel log before the KERN_ERR message is logged. There is a timestamp on each message added to the internal log. However, this timestamp is not converted to wall time when logged. The value of the timestamp is solely to give a crude time reference for the messages. Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-06-30 21:50:00 +00:00
lpfc_printf_vlog(vport, KERN_ERR, LOG_TRACE_EVENT,
"0606 %s Port TYPE %x %s\n",
"Failed to issue GID_PT to ",
GID_PT_N_PORT,
"Finishing discovery.");
return 0;
}
vport->gidft_inp++;
return 1;
}
/*
* This routine handles processing a NameServer REG_LOGIN mailbox
* command upon completion. It is setup in the LPFC_MBOXQ
* as the completion routine when the command is
* handed off to the SLI layer.
*/
void
lpfc_mbx_cmpl_ns_reg_login(struct lpfc_hba *phba, LPFC_MBOXQ_t *pmb)
{
MAILBOX_t *mb = &pmb->u.mb;
struct lpfc_nodelist *ndlp = (struct lpfc_nodelist *)pmb->ctx_ndlp;
struct lpfc_vport *vport = pmb->vport;
int rc;
pmb->ctx_ndlp = NULL;
vport->gidft_inp = 0;
if (mb->mbxStatus) {
scsi: lpfc: Add an internal trace log buffer The current logging methods typically end up requesting a reproduction with a different logging level set to figure out what happened. This was mainly by design to not clutter the kernel log messages with things that were typically not interesting and the messages themselves could cause other issues. When looking to make a better system, it was seen that in many cases when more data was wanted was when another message, usually at KERN_ERR level, was logged. And in most cases, what the additional logging that was then enabled was typically. Most of these areas fell into the discovery machine. Based on this summary, the following design has been put in place: The driver will maintain an internal log (256 elements of 256 bytes). The "additional logging" messages that are usually enabled in a reproduction will be changed to now log all the time to the internal log. A new logging level is defined - LOG_TRACE_EVENT. When this level is set (it is not by default) and a message marked as KERN_ERR is logged, all the messages in the internal log will be dumped to the kernel log before the KERN_ERR message is logged. There is a timestamp on each message added to the internal log. However, this timestamp is not converted to wall time when logged. The value of the timestamp is solely to give a crude time reference for the messages. Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-06-30 21:50:00 +00:00
lpfc_printf_vlog(vport, KERN_ERR, LOG_TRACE_EVENT,
"0260 Register NameServer error: 0x%x\n",
mb->mbxStatus);
out:
/* decrement the node reference count held for this
* callback function.
*/
lpfc_nlp_put(ndlp);
lpfc_mbox_rsrc_cleanup(phba, pmb, MBOX_THD_UNLOCKED);
/* If the node is not registered with the scsi or nvme
* transport, remove the fabric node. The failed reg_login
* is terminal.
*/
if (!(ndlp->fc4_xpt_flags & (SCSI_XPT_REGD | NVME_XPT_REGD))) {
spin_lock_irq(&ndlp->lock);
ndlp->nlp_flag &= ~NLP_NPR_2B_DISC;
spin_unlock_irq(&ndlp->lock);
lpfc_nlp_not_used(ndlp);
}
2010-11-21 04:11:48 +00:00
if (phba->fc_topology == LPFC_TOPOLOGY_LOOP) {
/*
* RegLogin failed, use loop map to make discovery
* list
*/
lpfc_disc_list_loopmap(vport);
/* Start discovery */
lpfc_disc_start(vport);
return;
}
lpfc_vport_set_state(vport, FC_VPORT_FAILED);
return;
}
if (phba->sli_rev < LPFC_SLI_REV4)
ndlp->nlp_rpi = mb->un.varWords[0];
ndlp->nlp_flag |= NLP_RPI_REGISTERED;
ndlp->nlp_type |= NLP_FABRIC;
lpfc_nlp_set_state(vport, ndlp, NLP_STE_UNMAPPED_NODE);
scsi: lpfc: Fix device recovery errors after PLOGI failures When target-side fault injections are made, the driver isn't reconnecting to the remote port. The driver is logging "2753" error messages which state: "PLOGI failure DID:1B2400 Status:x3/xf0240008" The failures status is indicating a Illegal field error, which points to the Temporary RPI field being used for the ELS. This error typically means the driver used an RPI that was already registered (shouldn't be registered if using it in this context). Study has found that if the driver were in discovery attempts and encountered an error, it wouldn't flag the temporary rpi in error. Yet the rpi was released for reallocation in these error paths and another ELS could allocate the rpi. In the failure situation a retry was done on an ELS that had encountered an error, and as the rpi wasn't marked in error, the ELS reused the rpi it originally allocated. But that rpi had been allocated by a different ELS issued after the original error and before the retry attempt. The different ELS had succeeded and the RPI was registered. Fix by marking the rpi state for the node to be in error, aka as needing reallocation, upon an error in the els processing. Error state marking is always done prior to release back to the internal rpi free list, which the driver wasn't doing in cases prior. Also enhanced some of the logging to help in the next case of problem troubleshooting. Link: https://lore.kernel.org/r/20190922035906.10977-7-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2019-09-22 03:58:52 +00:00
lpfc_printf_vlog(vport, KERN_INFO, LOG_NODE | LOG_DISCOVERY,
scsi: lpfc: Rework remote port ref counting and node freeing When a remote port is disconnected and disappears, its node structure (ndlp) stays allocated and on a vport node list. While on the list it can be matched, thus requires validation checks on state to be added in numerous code paths. If the node comes back, its possible for there to be multiple node structures for the same device on the vport node list. There is no reason to keep the node structure around after it is no longer in existence, and the current implementation creates problems for itself (multiple nodes) and lots of unnecessary code for state validation. Additionally, the reference taking on the node structure didn't follow the normal model used by the kernel kref api. It included lots of odd logic to match state with reference count. The combination of this odd logic plus the way it was implicitly used in the discovery engine made its reference taking implementation suspect and extremely hard to follow. Change the driver such that the reference taking routines are now normal ref increments/decrements and callout on refcount=0. With this in place, the rework can be done such that the node structure is fully removed and deallocated when the remote port no longer exists and all references are removed. This removal logic, and the basic ref counting are intrically tied, thus in a single patch. Link: https://lore.kernel.org/r/20201115192646.12977-2-james.smart@broadcom.com Co-developed-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <james.smart@broadcom.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-11-15 19:26:30 +00:00
"0003 rpi:%x DID:%x flg:%x %d x%px\n",
ndlp->nlp_rpi, ndlp->nlp_DID, ndlp->nlp_flag,
kref_read(&ndlp->kref),
scsi: lpfc: Rework remote port ref counting and node freeing When a remote port is disconnected and disappears, its node structure (ndlp) stays allocated and on a vport node list. While on the list it can be matched, thus requires validation checks on state to be added in numerous code paths. If the node comes back, its possible for there to be multiple node structures for the same device on the vport node list. There is no reason to keep the node structure around after it is no longer in existence, and the current implementation creates problems for itself (multiple nodes) and lots of unnecessary code for state validation. Additionally, the reference taking on the node structure didn't follow the normal model used by the kernel kref api. It included lots of odd logic to match state with reference count. The combination of this odd logic plus the way it was implicitly used in the discovery engine made its reference taking implementation suspect and extremely hard to follow. Change the driver such that the reference taking routines are now normal ref increments/decrements and callout on refcount=0. With this in place, the rework can be done such that the node structure is fully removed and deallocated when the remote port no longer exists and all references are removed. This removal logic, and the basic ref counting are intrically tied, thus in a single patch. Link: https://lore.kernel.org/r/20201115192646.12977-2-james.smart@broadcom.com Co-developed-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <james.smart@broadcom.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-11-15 19:26:30 +00:00
ndlp);
if (vport->port_state < LPFC_VPORT_READY) {
/* Link up discovery requires Fabric registration. */
lpfc_ns_cmd(vport, SLI_CTNS_RNN_ID, 0, 0);
lpfc_ns_cmd(vport, SLI_CTNS_RSNN_NN, 0, 0);
lpfc_ns_cmd(vport, SLI_CTNS_RSPN_ID, 0, 0);
lpfc_ns_cmd(vport, SLI_CTNS_RFT_ID, 0, 0);
if ((vport->cfg_enable_fc4_type == LPFC_ENABLE_BOTH) ||
(vport->cfg_enable_fc4_type == LPFC_ENABLE_FCP))
lpfc_ns_cmd(vport, SLI_CTNS_RFF_ID, 0, FC_TYPE_FCP);
if ((vport->cfg_enable_fc4_type == LPFC_ENABLE_BOTH) ||
(vport->cfg_enable_fc4_type == LPFC_ENABLE_NVME))
lpfc_ns_cmd(vport, SLI_CTNS_RFF_ID, 0,
FC_TYPE_NVME);
/* Issue SCR just before NameServer GID_FT Query */
lpfc_issue_els_scr(vport, 0);
/* Link was bounced or a Fabric LOGO occurred. Start EDC
* with initial FW values provided the congestion mode is
* not off. Note that signals may or may not be supported
* by the adapter but FPIN is provided by default for 1
* or both missing signals support.
*/
if (phba->cmf_active_mode != LPFC_CFG_OFF) {
phba->cgn_reg_fpin = phba->cgn_init_reg_fpin;
phba->cgn_reg_signal = phba->cgn_init_reg_signal;
rc = lpfc_issue_els_edc(vport, 0);
lpfc_printf_log(phba, KERN_INFO,
LOG_INIT | LOG_ELS | LOG_DISCOVERY,
"4220 Issue EDC status x%x Data x%x\n",
rc, phba->cgn_init_reg_signal);
} else if (phba->lmt & LMT_64Gb) {
/* may send link fault capability descriptor */
lpfc_issue_els_edc(vport, 0);
} else {
lpfc_issue_els_rdf(vport, 0);
}
}
vport->fc_ns_retry = 0;
if (lpfc_issue_gidft(vport) == 0)
goto out;
/*
* At this point in time we may need to wait for multiple
* SLI_CTNS_GID_FT CT commands to complete before we start discovery.
*
* decrement the node reference count held for this
* callback function.
*/
lpfc_nlp_put(ndlp);
lpfc_mbox_rsrc_cleanup(phba, pmb, MBOX_THD_UNLOCKED);
return;
}
/*
* This routine handles processing a Fabric Controller REG_LOGIN mailbox
* command upon completion. It is setup in the LPFC_MBOXQ
* as the completion routine when the command is handed off to the SLI layer.
*/
void
lpfc_mbx_cmpl_fc_reg_login(struct lpfc_hba *phba, LPFC_MBOXQ_t *pmb)
{
struct lpfc_vport *vport = pmb->vport;
MAILBOX_t *mb = &pmb->u.mb;
struct lpfc_nodelist *ndlp = (struct lpfc_nodelist *)pmb->ctx_ndlp;
pmb->ctx_ndlp = NULL;
if (mb->mbxStatus) {
lpfc_printf_vlog(vport, KERN_ERR, LOG_TRACE_EVENT,
"0933 %s: Register FC login error: 0x%x\n",
__func__, mb->mbxStatus);
goto out;
}
lpfc_check_nlp_post_devloss(vport, ndlp);
if (phba->sli_rev < LPFC_SLI_REV4)
ndlp->nlp_rpi = mb->un.varWords[0];
lpfc_printf_vlog(vport, KERN_INFO, LOG_NODE,
"0934 %s: Complete FC x%x RegLogin rpi x%x ste x%x\n",
__func__, ndlp->nlp_DID, ndlp->nlp_rpi,
ndlp->nlp_state);
ndlp->nlp_flag |= NLP_RPI_REGISTERED;
ndlp->nlp_flag &= ~NLP_REG_LOGIN_SEND;
ndlp->nlp_type |= NLP_FABRIC;
lpfc_nlp_set_state(vport, ndlp, NLP_STE_UNMAPPED_NODE);
out:
lpfc_mbox_rsrc_cleanup(phba, pmb, MBOX_THD_UNLOCKED);
/* Drop the reference count from the mbox at the end after
* all the current reference to the ndlp have been done.
*/
lpfc_nlp_put(ndlp);
}
static void
lpfc_register_remote_port(struct lpfc_vport *vport, struct lpfc_nodelist *ndlp)
{
struct Scsi_Host *shost = lpfc_shost_from_vport(vport);
struct fc_rport *rport;
struct lpfc_rport_data *rdata;
struct fc_rport_identifiers rport_ids;
struct lpfc_hba *phba = vport->phba;
unsigned long flags;
if (vport->cfg_enable_fc4_type == LPFC_ENABLE_NVME)
scsi: lpfc: NVME Initiator: Base modifications NVME Initiator: Base modifications This patch adds base modifications for NVME initiator support. The base modifications consist of: - Formal split of SLI3 rings from SLI-4 WQs (sometimes referred to as rings as well) as implementation now widely varies between the two. - Addition of configuration modes: SCSI initiator only; NVME initiator only; NVME target only; and SCSI and NVME initiator. The configuration mode drives overall adapter configuration, offloads enabled, and resource splits. NVME support is only available on SLI-4 devices and newer fw. - Implements the following based on configuration mode: - Exchange resources are split by protocol; Obviously, if only 1 mode, then no split occurs. Default is 50/50. module attribute allows tuning. - Pools and config parameters are separated per-protocol - Each protocol has it's own set of queues, but share interrupt vectors. SCSI: SLI3 devices have few queues and the original style of queue allocation remains. SLI4 devices piggy back on an "io-channel" concept that eventually needs to merge with scsi-mq/blk-mq support (it is underway). For now, the paradigm continues as it existed prior. io channel allocates N msix and N WQs (N=4 default) and either round robins or uses cpu # modulo N for scheduling. A bunch of module parameters allow the configuration to be tuned. NVME (initiator): Allocates an msix per cpu (or whatever pci_alloc_irq_vectors gets) Allocates a WQ per cpu, and maps the WQs to msix on a WQ # modulo msix vector count basis. Module parameters exist to cap/control the config if desired. - Each protocol has its own buffer and dma pools. I apologize for the size of the patch. Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <james.smart@broadcom.com> ---- Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2017-02-12 21:52:30 +00:00
return;
/* Remote port has reappeared. Re-register w/ FC transport */
rport_ids.node_name = wwn_to_u64(ndlp->nlp_nodename.u.wwn);
rport_ids.port_name = wwn_to_u64(ndlp->nlp_portname.u.wwn);
rport_ids.port_id = ndlp->nlp_DID;
rport_ids.roles = FC_RPORT_ROLE_UNKNOWN;
lpfc_debugfs_disc_trc(vport, LPFC_DISC_TRC_RPORT,
"rport add: did:x%x flg:x%x type x%x",
ndlp->nlp_DID, ndlp->nlp_flag, ndlp->nlp_type);
/* Don't add the remote port if unloading. */
if (vport->load_flag & FC_UNLOADING)
return;
/*
* Disassociate any older association between this ndlp and rport
*/
if (ndlp->rport) {
rdata = ndlp->rport->dd_data;
rdata->pnode = NULL;
}
ndlp->rport = rport = fc_remote_port_add(shost, 0, &rport_ids);
if (!rport) {
dev_printk(KERN_WARNING, &phba->pcidev->dev,
"Warning: fc_remote_port_add failed\n");
return;
}
/* Successful port add. Complete initializing node data */
rport->maxframe_size = ndlp->nlp_maxframe;
rport->supported_classes = ndlp->nlp_class_sup;
rdata = rport->dd_data;
rdata->pnode = lpfc_nlp_get(ndlp);
scsi: lpfc: Rework locations of ndlp reference taking Now that the driver has gone to a normal ref interface (with no odd logic) the discovery logic needs to be updated to reworked so that it properly takes references when it should and give them up when it should. Rework the driver for the following get/put model: - Move gets to just before an I/O is issued. Add gets for places where an I/O was issued without one. - Ensure that failures from lpfc_nlp_get() are handled by the driver. - Check and fix the placement of lpfc_nlp_puts relative to io completions. Note: some of these paths may not release the reference on the exact io completion as the reference is held as the code takes another step in the discovery thread and which may cause another io to be issued. - Rearrange some code for error processing and calling lpfc_nlp_put. - Fix some places of incorrect reference freeing that was causing the premature releasing of the structure. - Nvmet plogi handling performs unreg_rpi's. The reference counts were unbalanced resulting in premature node removal. In some cases this caused loss of node discovery. Corrected the reftaking around nvmet plogis. Nodes that experience devloss now get released from the node list now that there is a proper reference taking. Link: https://lore.kernel.org/r/20201115192646.12977-3-james.smart@broadcom.com Co-developed-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <james.smart@broadcom.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-11-15 19:26:31 +00:00
if (!rdata->pnode) {
dev_warn(&phba->pcidev->dev,
"Warning - node ref failed. Unreg rport\n");
fc_remote_port_delete(rport);
ndlp->rport = NULL;
return;
}
spin_lock_irqsave(&ndlp->lock, flags);
ndlp->fc4_xpt_flags |= SCSI_XPT_REGD;
spin_unlock_irqrestore(&ndlp->lock, flags);
if (ndlp->nlp_type & NLP_FCP_TARGET)
rport_ids.roles |= FC_PORT_ROLE_FCP_TARGET;
if (ndlp->nlp_type & NLP_FCP_INITIATOR)
rport_ids.roles |= FC_PORT_ROLE_FCP_INITIATOR;
if (ndlp->nlp_type & NLP_NVME_INITIATOR)
rport_ids.roles |= FC_PORT_ROLE_NVME_INITIATOR;
if (ndlp->nlp_type & NLP_NVME_TARGET)
rport_ids.roles |= FC_PORT_ROLE_NVME_TARGET;
if (ndlp->nlp_type & NLP_NVME_DISCOVERY)
rport_ids.roles |= FC_PORT_ROLE_NVME_DISCOVERY;
if (rport_ids.roles != FC_RPORT_ROLE_UNKNOWN)
fc_remote_port_rolechg(rport, rport_ids.roles);
lpfc_printf_vlog(ndlp->vport, KERN_INFO, LOG_NODE,
scsi: lpfc: Don't release final kref on Fport node while ABTS outstanding In a rarely executed path, FLOGI failure, there is a refcounting error. If FLOGI completed with an error, typically a timeout, the initial completion handler would remove the job reference. However, the job completion isn't the actual end of the job/exchange as the timeout usually initiates an ABTS, and upon that ABTS completion, a final completion is sent. The driver removes the reference again in the final completion. Thus the imbalance. In the buggy cases, if there was a link bounce while the delayed response is outstanding, the fport node may be referenced again but there was no additional reference as it is already present. The delayed completion then occurs and removes the last reference freeing the node and causing issues in the link up processed that is using the node. Fix this scenario by removing the snippet that removed the reference in the initial FLOGI completion. The bad snippet was poorly trying to identify the FLOGI as OK to do so by realizing the node was not registered with either SCSI or NVMe transport. Link: https://lore.kernel.org/r/20210910233159.115896-3-jsmart2021@gmail.com Fixes: 618e2ee146d4 ("scsi: lpfc: Fix FLOGI failure due to accessing a freed node") Cc: <stable@vger.kernel.org> # v5.13+ Co-developed-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2021-09-10 23:31:47 +00:00
"3183 %s rport x%px DID x%x, role x%x refcnt %d\n",
__func__, rport, rport->port_id, rport->roles,
kref_read(&ndlp->kref));
if ((rport->scsi_target_id != -1) &&
(rport->scsi_target_id < LPFC_MAX_TARGET)) {
ndlp->nlp_sid = rport->scsi_target_id;
}
[SCSI] update fc_transport for removal of block/unblock functions We recently went back to implement a board reset. When we perform the reset, we wanted to tear down the internal data structures and rebuild them. Unfortunately, when it came to the rport structure, things were odd. If we deleted them, the scsi targets and sdevs would be torn down. Not a good thing for a temporary reset. We could block the rports, but we either maintain the internal structures to keep the rport reference (perhaps even replicating what's in the transport), or we have to fatten the fc transport with new search routines to find the rport (and deal with a case of a dangling rport that the driver forgets). It dawned on me that we had actually reached this state incorrectly. When the fc transport first started, we did the block/unblock first, then added the rport interface. The purpose of block/unblock is to hide the temporary disappearance of the rport (e.g. being deleted, then readded). Why are we making the driver do the block/unblock ? We should be making the transport have only an rport add/delete, and the let the transport handle the block/unblock. So... This patch removes the existing fc_remote_port_block/unblock functions. It moves the block/unblock functionality into the fc_remote_port_add/delete functions. Updates for the lpfc driver are included. Qlogic driver updates are also enclosed, thanks to the contributions of Andrew Vasquez. [Note: the qla2xxx changes are relative to the scsi-misc-2.6 tree as of this morning - which does not include the recent patches sent by Andrew]. The zfcp driver does not use the block/unblock functions. One last comment: The resulting behavior feels very clean. The LLDD is concerned only with add/delete, which corresponds to the physical disappearance. However, the fact that the scsi target and sdevs are not immediately torn down after the LLDD calls delete causes an interesting scenario... the midlayer can call the xxx_slave_alloc and xxx_queuecommand functions with a sdev that is at the location the rport used to be. The driver must validate the device exists when it first enters these functions. In thinking about it, this has always been the case for the LLDD and these routines. The existing drivers already check for existence. However, this highlights that simple validation via data structure dereferencing needs to be watched. To deal with this, a new transport function, fc_remote_port_chkready() was created that LLDDs should call when they first enter these two routines. It validates the rport state, and returns a scsi result which could be returned. In addition to solving the above, it also creates consistent behavior from the LLDD's when the block and deletes are occuring. Rejections fixed up and Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
2005-10-18 16:03:35 +00:00
return;
}
static void
lpfc_unregister_remote_port(struct lpfc_nodelist *ndlp)
[SCSI] update fc_transport for removal of block/unblock functions We recently went back to implement a board reset. When we perform the reset, we wanted to tear down the internal data structures and rebuild them. Unfortunately, when it came to the rport structure, things were odd. If we deleted them, the scsi targets and sdevs would be torn down. Not a good thing for a temporary reset. We could block the rports, but we either maintain the internal structures to keep the rport reference (perhaps even replicating what's in the transport), or we have to fatten the fc transport with new search routines to find the rport (and deal with a case of a dangling rport that the driver forgets). It dawned on me that we had actually reached this state incorrectly. When the fc transport first started, we did the block/unblock first, then added the rport interface. The purpose of block/unblock is to hide the temporary disappearance of the rport (e.g. being deleted, then readded). Why are we making the driver do the block/unblock ? We should be making the transport have only an rport add/delete, and the let the transport handle the block/unblock. So... This patch removes the existing fc_remote_port_block/unblock functions. It moves the block/unblock functionality into the fc_remote_port_add/delete functions. Updates for the lpfc driver are included. Qlogic driver updates are also enclosed, thanks to the contributions of Andrew Vasquez. [Note: the qla2xxx changes are relative to the scsi-misc-2.6 tree as of this morning - which does not include the recent patches sent by Andrew]. The zfcp driver does not use the block/unblock functions. One last comment: The resulting behavior feels very clean. The LLDD is concerned only with add/delete, which corresponds to the physical disappearance. However, the fact that the scsi target and sdevs are not immediately torn down after the LLDD calls delete causes an interesting scenario... the midlayer can call the xxx_slave_alloc and xxx_queuecommand functions with a sdev that is at the location the rport used to be. The driver must validate the device exists when it first enters these functions. In thinking about it, this has always been the case for the LLDD and these routines. The existing drivers already check for existence. However, this highlights that simple validation via data structure dereferencing needs to be watched. To deal with this, a new transport function, fc_remote_port_chkready() was created that LLDDs should call when they first enter these two routines. It validates the rport state, and returns a scsi result which could be returned. In addition to solving the above, it also creates consistent behavior from the LLDD's when the block and deletes are occuring. Rejections fixed up and Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
2005-10-18 16:03:35 +00:00
{
struct fc_rport *rport = ndlp->rport;
struct lpfc_vport *vport = ndlp->vport;
if (vport->cfg_enable_fc4_type == LPFC_ENABLE_NVME)
return;
lpfc_debugfs_disc_trc(vport, LPFC_DISC_TRC_RPORT,
"rport delete: did:x%x flg:x%x type x%x",
ndlp->nlp_DID, ndlp->nlp_flag, ndlp->nlp_type);
lpfc_printf_vlog(vport, KERN_INFO, LOG_NODE,
"3184 rport unregister x%06x, rport x%px "
scsi: lpfc: Don't release final kref on Fport node while ABTS outstanding In a rarely executed path, FLOGI failure, there is a refcounting error. If FLOGI completed with an error, typically a timeout, the initial completion handler would remove the job reference. However, the job completion isn't the actual end of the job/exchange as the timeout usually initiates an ABTS, and upon that ABTS completion, a final completion is sent. The driver removes the reference again in the final completion. Thus the imbalance. In the buggy cases, if there was a link bounce while the delayed response is outstanding, the fport node may be referenced again but there was no additional reference as it is already present. The delayed completion then occurs and removes the last reference freeing the node and causing issues in the link up processed that is using the node. Fix this scenario by removing the snippet that removed the reference in the initial FLOGI completion. The bad snippet was poorly trying to identify the FLOGI as OK to do so by realizing the node was not registered with either SCSI or NVMe transport. Link: https://lore.kernel.org/r/20210910233159.115896-3-jsmart2021@gmail.com Fixes: 618e2ee146d4 ("scsi: lpfc: Fix FLOGI failure due to accessing a freed node") Cc: <stable@vger.kernel.org> # v5.13+ Co-developed-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2021-09-10 23:31:47 +00:00
"xptflg x%x refcnt %d\n",
ndlp->nlp_DID, rport, ndlp->fc4_xpt_flags,
kref_read(&ndlp->kref));
[SCSI] update fc_transport for removal of block/unblock functions We recently went back to implement a board reset. When we perform the reset, we wanted to tear down the internal data structures and rebuild them. Unfortunately, when it came to the rport structure, things were odd. If we deleted them, the scsi targets and sdevs would be torn down. Not a good thing for a temporary reset. We could block the rports, but we either maintain the internal structures to keep the rport reference (perhaps even replicating what's in the transport), or we have to fatten the fc transport with new search routines to find the rport (and deal with a case of a dangling rport that the driver forgets). It dawned on me that we had actually reached this state incorrectly. When the fc transport first started, we did the block/unblock first, then added the rport interface. The purpose of block/unblock is to hide the temporary disappearance of the rport (e.g. being deleted, then readded). Why are we making the driver do the block/unblock ? We should be making the transport have only an rport add/delete, and the let the transport handle the block/unblock. So... This patch removes the existing fc_remote_port_block/unblock functions. It moves the block/unblock functionality into the fc_remote_port_add/delete functions. Updates for the lpfc driver are included. Qlogic driver updates are also enclosed, thanks to the contributions of Andrew Vasquez. [Note: the qla2xxx changes are relative to the scsi-misc-2.6 tree as of this morning - which does not include the recent patches sent by Andrew]. The zfcp driver does not use the block/unblock functions. One last comment: The resulting behavior feels very clean. The LLDD is concerned only with add/delete, which corresponds to the physical disappearance. However, the fact that the scsi target and sdevs are not immediately torn down after the LLDD calls delete causes an interesting scenario... the midlayer can call the xxx_slave_alloc and xxx_queuecommand functions with a sdev that is at the location the rport used to be. The driver must validate the device exists when it first enters these functions. In thinking about it, this has always been the case for the LLDD and these routines. The existing drivers already check for existence. However, this highlights that simple validation via data structure dereferencing needs to be watched. To deal with this, a new transport function, fc_remote_port_chkready() was created that LLDDs should call when they first enter these two routines. It validates the rport state, and returns a scsi result which could be returned. In addition to solving the above, it also creates consistent behavior from the LLDD's when the block and deletes are occuring. Rejections fixed up and Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
2005-10-18 16:03:35 +00:00
fc_remote_port_delete(rport);
lpfc_nlp_put(ndlp);
}
static void
lpfc_nlp_counters(struct lpfc_vport *vport, int state, int count)
{
struct Scsi_Host *shost = lpfc_shost_from_vport(vport);
unsigned long iflags;
spin_lock_irqsave(shost->host_lock, iflags);
switch (state) {
case NLP_STE_UNUSED_NODE:
vport->fc_unused_cnt += count;
break;
case NLP_STE_PLOGI_ISSUE:
vport->fc_plogi_cnt += count;
break;
case NLP_STE_ADISC_ISSUE:
vport->fc_adisc_cnt += count;
break;
case NLP_STE_REG_LOGIN_ISSUE:
vport->fc_reglogin_cnt += count;
break;
case NLP_STE_PRLI_ISSUE:
vport->fc_prli_cnt += count;
break;
case NLP_STE_UNMAPPED_NODE:
vport->fc_unmap_cnt += count;
break;
case NLP_STE_MAPPED_NODE:
vport->fc_map_cnt += count;
break;
case NLP_STE_NPR_NODE:
if (vport->fc_npr_cnt == 0 && count == -1)
vport->fc_npr_cnt = 0;
else
vport->fc_npr_cnt += count;
break;
}
spin_unlock_irqrestore(shost->host_lock, iflags);
}
scsi: lpfc: Delay unregistering from transport until GIDFT or ADISC completes On an RSCN event, the nodes specified in RSCN payload and in MAPPED state are moved to NPR state in order to revalidate the login. This triggers an immediate unregister from SCSI/NVMe backend. The assumption is that the node may be missing. The re-registration with the backend happens after either relogin (PLOGI/PRLI; if ADISC is disabled or login truly lost) or when ADISC completes successfully (rediscover with ADISC enabled). However, the NVMe-FC standard provides for an RSCN to be triggered when the remote port supports a discovery controller and there was a change of discovery log content. As the remote port typically also supports storage subsystems, this unregister causes all storage controller connections to fail and require reconnect. Correct by reworking the code to ensure that the unregistration only occurs when a login state is truly terminated, thereby leaving the NVMe storage controllers in place. The changes made are: - Retain node state in ADISC_ISSUE when scheduling ADISC ELS retry. - Do not clear wwpn/wwnn values upon ADISC failure. - Move MAPPED nodes to NPR during RSCN processing, but do not unregister with transport. On GIDFT completion, identify missing nodes (not marked NLP_NPR_2B_DISC) and unregister them. - Perform unregistration for nodes that will go through ADISC processing if ADISC completion fails. - Successful ADISC completion will move node back to MAPPED state. Link: https://lore.kernel.org/r/20210707184351.67872-16-jsmart2021@gmail.com Co-developed-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2021-07-07 18:43:46 +00:00
/* Register a node with backend if not already done */
void
lpfc_nlp_reg_node(struct lpfc_vport *vport, struct lpfc_nodelist *ndlp)
{
unsigned long iflags;
lpfc_check_nlp_post_devloss(vport, ndlp);
scsi: lpfc: Delay unregistering from transport until GIDFT or ADISC completes On an RSCN event, the nodes specified in RSCN payload and in MAPPED state are moved to NPR state in order to revalidate the login. This triggers an immediate unregister from SCSI/NVMe backend. The assumption is that the node may be missing. The re-registration with the backend happens after either relogin (PLOGI/PRLI; if ADISC is disabled or login truly lost) or when ADISC completes successfully (rediscover with ADISC enabled). However, the NVMe-FC standard provides for an RSCN to be triggered when the remote port supports a discovery controller and there was a change of discovery log content. As the remote port typically also supports storage subsystems, this unregister causes all storage controller connections to fail and require reconnect. Correct by reworking the code to ensure that the unregistration only occurs when a login state is truly terminated, thereby leaving the NVMe storage controllers in place. The changes made are: - Retain node state in ADISC_ISSUE when scheduling ADISC ELS retry. - Do not clear wwpn/wwnn values upon ADISC failure. - Move MAPPED nodes to NPR during RSCN processing, but do not unregister with transport. On GIDFT completion, identify missing nodes (not marked NLP_NPR_2B_DISC) and unregister them. - Perform unregistration for nodes that will go through ADISC processing if ADISC completion fails. - Successful ADISC completion will move node back to MAPPED state. Link: https://lore.kernel.org/r/20210707184351.67872-16-jsmart2021@gmail.com Co-developed-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2021-07-07 18:43:46 +00:00
spin_lock_irqsave(&ndlp->lock, iflags);
if (ndlp->fc4_xpt_flags & NLP_XPT_REGD) {
/* Already registered with backend, trigger rescan */
spin_unlock_irqrestore(&ndlp->lock, iflags);
if (ndlp->fc4_xpt_flags & NVME_XPT_REGD &&
ndlp->nlp_type & (NLP_NVME_TARGET | NLP_NVME_DISCOVERY)) {
lpfc_nvme_rescan_port(vport, ndlp);
}
return;
}
ndlp->fc4_xpt_flags |= NLP_XPT_REGD;
spin_unlock_irqrestore(&ndlp->lock, iflags);
if (lpfc_valid_xpt_node(ndlp)) {
vport->phba->nport_event_cnt++;
/*
* Tell the fc transport about the port, if we haven't
* already. If we have, and it's a scsi entity, be
*/
lpfc_register_remote_port(vport, ndlp);
}
/* We are done if we do not have any NVME remote node */
if (!(ndlp->nlp_fc4_type & NLP_FC4_NVME))
return;
/* Notify the NVME transport of this new rport. */
if (vport->phba->sli_rev >= LPFC_SLI_REV4 &&
ndlp->nlp_fc4_type & NLP_FC4_NVME) {
if (vport->phba->nvmet_support == 0) {
/* Register this rport with the transport.
* Only NVME Target Rports are registered with
* the transport.
*/
if (ndlp->nlp_type & NLP_NVME_TARGET) {
vport->phba->nport_event_cnt++;
lpfc_nvme_register_port(vport, ndlp);
}
} else {
/* Just take an NDLP ref count since the
* target does not register rports.
*/
lpfc_nlp_get(ndlp);
}
}
}
/* Unregister a node with backend if not already done */
void
lpfc_nlp_unreg_node(struct lpfc_vport *vport, struct lpfc_nodelist *ndlp)
{
unsigned long iflags;
spin_lock_irqsave(&ndlp->lock, iflags);
if (!(ndlp->fc4_xpt_flags & NLP_XPT_REGD)) {
spin_unlock_irqrestore(&ndlp->lock, iflags);
scsi: lpfc: Improve PCI EEH Error and Recovery Handling Following EEH errors, the driver can crash or hang when deleting the localport or when attempting to unload. The EEH handlers in the driver did not notify the NVMe-FC transport before tearing the driver down. This was delayed until the resume steps. This worked for SCSI because lpfc_block_scsi() would notify the scsi_fc_transport that the target was not available but it would not clean up all the references to the ndlp. The SLI3 prep for dev reset handler did the lpfc_offline_prep() and lpfc_offline() calls to get the port stopped before restarting. The SLI4 version of the prep for dev reset just destroyed the queues and did not stop NVMe from continuing. Also because the port was not really stopped the localport destroy would hang because the transport was still waiting for I/O. Additionally, a devloss tmo can fire and post events to a stopped worker thread creating another hang condition. lpfc_sli4_prep_dev_for_reset() is modified to call lpfc_offline_prep() and lpfc_offline() rather than just lpfc_scsi_dev_block() to ensure both SCSI and NVMe transports are notified to block I/O to the driver. Logic is added to devloss handler and worker thread to clean up ndlp references and quiesce appropriately. Link: https://lore.kernel.org/r/20220317032737.45308-2-jsmart2021@gmail.com Co-developed-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2022-03-17 03:27:34 +00:00
lpfc_printf_vlog(vport, KERN_INFO, LOG_SLI,
"0999 %s Not regd: ndlp x%px rport x%px DID "
"x%x FLG x%x XPT x%x\n",
__func__, ndlp, ndlp->rport, ndlp->nlp_DID,
ndlp->nlp_flag, ndlp->fc4_xpt_flags);
scsi: lpfc: Delay unregistering from transport until GIDFT or ADISC completes On an RSCN event, the nodes specified in RSCN payload and in MAPPED state are moved to NPR state in order to revalidate the login. This triggers an immediate unregister from SCSI/NVMe backend. The assumption is that the node may be missing. The re-registration with the backend happens after either relogin (PLOGI/PRLI; if ADISC is disabled or login truly lost) or when ADISC completes successfully (rediscover with ADISC enabled). However, the NVMe-FC standard provides for an RSCN to be triggered when the remote port supports a discovery controller and there was a change of discovery log content. As the remote port typically also supports storage subsystems, this unregister causes all storage controller connections to fail and require reconnect. Correct by reworking the code to ensure that the unregistration only occurs when a login state is truly terminated, thereby leaving the NVMe storage controllers in place. The changes made are: - Retain node state in ADISC_ISSUE when scheduling ADISC ELS retry. - Do not clear wwpn/wwnn values upon ADISC failure. - Move MAPPED nodes to NPR during RSCN processing, but do not unregister with transport. On GIDFT completion, identify missing nodes (not marked NLP_NPR_2B_DISC) and unregister them. - Perform unregistration for nodes that will go through ADISC processing if ADISC completion fails. - Successful ADISC completion will move node back to MAPPED state. Link: https://lore.kernel.org/r/20210707184351.67872-16-jsmart2021@gmail.com Co-developed-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2021-07-07 18:43:46 +00:00
return;
}
ndlp->fc4_xpt_flags &= ~NLP_XPT_REGD;
spin_unlock_irqrestore(&ndlp->lock, iflags);
if (ndlp->rport &&
ndlp->fc4_xpt_flags & SCSI_XPT_REGD) {
vport->phba->nport_event_cnt++;
lpfc_unregister_remote_port(ndlp);
scsi: lpfc: Improve PCI EEH Error and Recovery Handling Following EEH errors, the driver can crash or hang when deleting the localport or when attempting to unload. The EEH handlers in the driver did not notify the NVMe-FC transport before tearing the driver down. This was delayed until the resume steps. This worked for SCSI because lpfc_block_scsi() would notify the scsi_fc_transport that the target was not available but it would not clean up all the references to the ndlp. The SLI3 prep for dev reset handler did the lpfc_offline_prep() and lpfc_offline() calls to get the port stopped before restarting. The SLI4 version of the prep for dev reset just destroyed the queues and did not stop NVMe from continuing. Also because the port was not really stopped the localport destroy would hang because the transport was still waiting for I/O. Additionally, a devloss tmo can fire and post events to a stopped worker thread creating another hang condition. lpfc_sli4_prep_dev_for_reset() is modified to call lpfc_offline_prep() and lpfc_offline() rather than just lpfc_scsi_dev_block() to ensure both SCSI and NVMe transports are notified to block I/O to the driver. Logic is added to devloss handler and worker thread to clean up ndlp references and quiesce appropriately. Link: https://lore.kernel.org/r/20220317032737.45308-2-jsmart2021@gmail.com Co-developed-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2022-03-17 03:27:34 +00:00
} else if (!ndlp->rport) {
lpfc_printf_vlog(vport, KERN_INFO, LOG_SLI,
"1999 %s NDLP in devloss x%px DID x%x FLG x%x"
" XPT x%x refcnt %d\n",
__func__, ndlp, ndlp->nlp_DID, ndlp->nlp_flag,
ndlp->fc4_xpt_flags,
kref_read(&ndlp->kref));
scsi: lpfc: Delay unregistering from transport until GIDFT or ADISC completes On an RSCN event, the nodes specified in RSCN payload and in MAPPED state are moved to NPR state in order to revalidate the login. This triggers an immediate unregister from SCSI/NVMe backend. The assumption is that the node may be missing. The re-registration with the backend happens after either relogin (PLOGI/PRLI; if ADISC is disabled or login truly lost) or when ADISC completes successfully (rediscover with ADISC enabled). However, the NVMe-FC standard provides for an RSCN to be triggered when the remote port supports a discovery controller and there was a change of discovery log content. As the remote port typically also supports storage subsystems, this unregister causes all storage controller connections to fail and require reconnect. Correct by reworking the code to ensure that the unregistration only occurs when a login state is truly terminated, thereby leaving the NVMe storage controllers in place. The changes made are: - Retain node state in ADISC_ISSUE when scheduling ADISC ELS retry. - Do not clear wwpn/wwnn values upon ADISC failure. - Move MAPPED nodes to NPR during RSCN processing, but do not unregister with transport. On GIDFT completion, identify missing nodes (not marked NLP_NPR_2B_DISC) and unregister them. - Perform unregistration for nodes that will go through ADISC processing if ADISC completion fails. - Successful ADISC completion will move node back to MAPPED state. Link: https://lore.kernel.org/r/20210707184351.67872-16-jsmart2021@gmail.com Co-developed-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2021-07-07 18:43:46 +00:00
}
if (ndlp->fc4_xpt_flags & NVME_XPT_REGD) {
vport->phba->nport_event_cnt++;
if (vport->phba->nvmet_support == 0) {
/* Start devloss if target. */
if (ndlp->nlp_type & NLP_NVME_TARGET)
lpfc_nvme_unregister_port(vport, ndlp);
} else {
/* NVMET has no upcall. */
lpfc_nlp_put(ndlp);
}
}
}
/*
* Adisc state change handling
*/
static void
lpfc_handle_adisc_state(struct lpfc_vport *vport, struct lpfc_nodelist *ndlp,
int new_state)
{
switch (new_state) {
/*
* Any state to ADISC_ISSUE
* Do nothing, adisc cmpl handling will trigger state changes
*/
case NLP_STE_ADISC_ISSUE:
break;
/*
* ADISC_ISSUE to mapped states
* Trigger a registration with backend, it will be nop if
* already registered
*/
case NLP_STE_UNMAPPED_NODE:
ndlp->nlp_type |= NLP_FC_NODE;
fallthrough;
case NLP_STE_MAPPED_NODE:
ndlp->nlp_flag &= ~NLP_NODEV_REMOVE;
lpfc_nlp_reg_node(vport, ndlp);
break;
/*
* ADISC_ISSUE to non-mapped states
* We are moving from ADISC_ISSUE to a non-mapped state because
* ADISC failed, we would have skipped unregistering with
* backend, attempt it now
*/
case NLP_STE_NPR_NODE:
ndlp->nlp_flag &= ~NLP_RCV_PLOGI;
fallthrough;
default:
lpfc_nlp_unreg_node(vport, ndlp);
break;
}
}
static void
lpfc_nlp_state_cleanup(struct lpfc_vport *vport, struct lpfc_nodelist *ndlp,
int old_state, int new_state)
{
scsi: lpfc: Delay unregistering from transport until GIDFT or ADISC completes On an RSCN event, the nodes specified in RSCN payload and in MAPPED state are moved to NPR state in order to revalidate the login. This triggers an immediate unregister from SCSI/NVMe backend. The assumption is that the node may be missing. The re-registration with the backend happens after either relogin (PLOGI/PRLI; if ADISC is disabled or login truly lost) or when ADISC completes successfully (rediscover with ADISC enabled). However, the NVMe-FC standard provides for an RSCN to be triggered when the remote port supports a discovery controller and there was a change of discovery log content. As the remote port typically also supports storage subsystems, this unregister causes all storage controller connections to fail and require reconnect. Correct by reworking the code to ensure that the unregistration only occurs when a login state is truly terminated, thereby leaving the NVMe storage controllers in place. The changes made are: - Retain node state in ADISC_ISSUE when scheduling ADISC ELS retry. - Do not clear wwpn/wwnn values upon ADISC failure. - Move MAPPED nodes to NPR during RSCN processing, but do not unregister with transport. On GIDFT completion, identify missing nodes (not marked NLP_NPR_2B_DISC) and unregister them. - Perform unregistration for nodes that will go through ADISC processing if ADISC completion fails. - Successful ADISC completion will move node back to MAPPED state. Link: https://lore.kernel.org/r/20210707184351.67872-16-jsmart2021@gmail.com Co-developed-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2021-07-07 18:43:46 +00:00
/* Trap ADISC changes here */
if (new_state == NLP_STE_ADISC_ISSUE ||
old_state == NLP_STE_ADISC_ISSUE) {
lpfc_handle_adisc_state(vport, ndlp, new_state);
return;
}
if (new_state == NLP_STE_UNMAPPED_NODE) {
ndlp->nlp_flag &= ~NLP_NODEV_REMOVE;
ndlp->nlp_type |= NLP_FC_NODE;
}
if (new_state == NLP_STE_MAPPED_NODE)
ndlp->nlp_flag &= ~NLP_NODEV_REMOVE;
if (new_state == NLP_STE_NPR_NODE)
ndlp->nlp_flag &= ~NLP_RCV_PLOGI;
scsi: lpfc: Delay unregistering from transport until GIDFT or ADISC completes On an RSCN event, the nodes specified in RSCN payload and in MAPPED state are moved to NPR state in order to revalidate the login. This triggers an immediate unregister from SCSI/NVMe backend. The assumption is that the node may be missing. The re-registration with the backend happens after either relogin (PLOGI/PRLI; if ADISC is disabled or login truly lost) or when ADISC completes successfully (rediscover with ADISC enabled). However, the NVMe-FC standard provides for an RSCN to be triggered when the remote port supports a discovery controller and there was a change of discovery log content. As the remote port typically also supports storage subsystems, this unregister causes all storage controller connections to fail and require reconnect. Correct by reworking the code to ensure that the unregistration only occurs when a login state is truly terminated, thereby leaving the NVMe storage controllers in place. The changes made are: - Retain node state in ADISC_ISSUE when scheduling ADISC ELS retry. - Do not clear wwpn/wwnn values upon ADISC failure. - Move MAPPED nodes to NPR during RSCN processing, but do not unregister with transport. On GIDFT completion, identify missing nodes (not marked NLP_NPR_2B_DISC) and unregister them. - Perform unregistration for nodes that will go through ADISC processing if ADISC completion fails. - Successful ADISC completion will move node back to MAPPED state. Link: https://lore.kernel.org/r/20210707184351.67872-16-jsmart2021@gmail.com Co-developed-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2021-07-07 18:43:46 +00:00
/* Reg/Unreg for FCP and NVME Transport interface */
if ((old_state == NLP_STE_MAPPED_NODE ||
old_state == NLP_STE_UNMAPPED_NODE)) {
/* For nodes marked for ADISC, Handle unreg in ADISC cmpl
* if linkup. In linkdown do unreg_node
*/
if (!(ndlp->nlp_flag & NLP_NPR_ADISC) ||
!lpfc_is_link_up(vport->phba))
scsi: lpfc: Delay unregistering from transport until GIDFT or ADISC completes On an RSCN event, the nodes specified in RSCN payload and in MAPPED state are moved to NPR state in order to revalidate the login. This triggers an immediate unregister from SCSI/NVMe backend. The assumption is that the node may be missing. The re-registration with the backend happens after either relogin (PLOGI/PRLI; if ADISC is disabled or login truly lost) or when ADISC completes successfully (rediscover with ADISC enabled). However, the NVMe-FC standard provides for an RSCN to be triggered when the remote port supports a discovery controller and there was a change of discovery log content. As the remote port typically also supports storage subsystems, this unregister causes all storage controller connections to fail and require reconnect. Correct by reworking the code to ensure that the unregistration only occurs when a login state is truly terminated, thereby leaving the NVMe storage controllers in place. The changes made are: - Retain node state in ADISC_ISSUE when scheduling ADISC ELS retry. - Do not clear wwpn/wwnn values upon ADISC failure. - Move MAPPED nodes to NPR during RSCN processing, but do not unregister with transport. On GIDFT completion, identify missing nodes (not marked NLP_NPR_2B_DISC) and unregister them. - Perform unregistration for nodes that will go through ADISC processing if ADISC completion fails. - Successful ADISC completion will move node back to MAPPED state. Link: https://lore.kernel.org/r/20210707184351.67872-16-jsmart2021@gmail.com Co-developed-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2021-07-07 18:43:46 +00:00
lpfc_nlp_unreg_node(vport, ndlp);
}
if (new_state == NLP_STE_MAPPED_NODE ||
scsi: lpfc: Delay unregistering from transport until GIDFT or ADISC completes On an RSCN event, the nodes specified in RSCN payload and in MAPPED state are moved to NPR state in order to revalidate the login. This triggers an immediate unregister from SCSI/NVMe backend. The assumption is that the node may be missing. The re-registration with the backend happens after either relogin (PLOGI/PRLI; if ADISC is disabled or login truly lost) or when ADISC completes successfully (rediscover with ADISC enabled). However, the NVMe-FC standard provides for an RSCN to be triggered when the remote port supports a discovery controller and there was a change of discovery log content. As the remote port typically also supports storage subsystems, this unregister causes all storage controller connections to fail and require reconnect. Correct by reworking the code to ensure that the unregistration only occurs when a login state is truly terminated, thereby leaving the NVMe storage controllers in place. The changes made are: - Retain node state in ADISC_ISSUE when scheduling ADISC ELS retry. - Do not clear wwpn/wwnn values upon ADISC failure. - Move MAPPED nodes to NPR during RSCN processing, but do not unregister with transport. On GIDFT completion, identify missing nodes (not marked NLP_NPR_2B_DISC) and unregister them. - Perform unregistration for nodes that will go through ADISC processing if ADISC completion fails. - Successful ADISC completion will move node back to MAPPED state. Link: https://lore.kernel.org/r/20210707184351.67872-16-jsmart2021@gmail.com Co-developed-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2021-07-07 18:43:46 +00:00
new_state == NLP_STE_UNMAPPED_NODE)
lpfc_nlp_reg_node(vport, ndlp);
/*
* If the node just added to Mapped list was an FCP target,
* but the remote port registration failed or assigned a target
* id outside the presentable range - move the node to the
* Unmapped List.
*/
if ((new_state == NLP_STE_MAPPED_NODE) &&
(ndlp->nlp_type & NLP_FCP_TARGET) &&
(!ndlp->rport ||
ndlp->rport->scsi_target_id == -1 ||
ndlp->rport->scsi_target_id >= LPFC_MAX_TARGET)) {
spin_lock_irq(&ndlp->lock);
ndlp->nlp_flag |= NLP_TGT_NO_SCSIID;
spin_unlock_irq(&ndlp->lock);
lpfc_nlp_set_state(vport, ndlp, NLP_STE_UNMAPPED_NODE);
}
}
static char *
lpfc_nlp_state_name(char *buffer, size_t size, int state)
{
static char *states[] = {
[NLP_STE_UNUSED_NODE] = "UNUSED",
[NLP_STE_PLOGI_ISSUE] = "PLOGI",
[NLP_STE_ADISC_ISSUE] = "ADISC",
[NLP_STE_REG_LOGIN_ISSUE] = "REGLOGIN",
[NLP_STE_PRLI_ISSUE] = "PRLI",
[NLP_STE_LOGO_ISSUE] = "LOGO",
[NLP_STE_UNMAPPED_NODE] = "UNMAPPED",
[NLP_STE_MAPPED_NODE] = "MAPPED",
[NLP_STE_NPR_NODE] = "NPR",
};
if (state < NLP_STE_MAX_STATE && states[state])
strlcpy(buffer, states[state], size);
else
snprintf(buffer, size, "unknown (%d)", state);
return buffer;
}
void
lpfc_nlp_set_state(struct lpfc_vport *vport, struct lpfc_nodelist *ndlp,
int state)
{
struct Scsi_Host *shost = lpfc_shost_from_vport(vport);
int old_state = ndlp->nlp_state;
scsi: lpfc: Rework remote port ref counting and node freeing When a remote port is disconnected and disappears, its node structure (ndlp) stays allocated and on a vport node list. While on the list it can be matched, thus requires validation checks on state to be added in numerous code paths. If the node comes back, its possible for there to be multiple node structures for the same device on the vport node list. There is no reason to keep the node structure around after it is no longer in existence, and the current implementation creates problems for itself (multiple nodes) and lots of unnecessary code for state validation. Additionally, the reference taking on the node structure didn't follow the normal model used by the kernel kref api. It included lots of odd logic to match state with reference count. The combination of this odd logic plus the way it was implicitly used in the discovery engine made its reference taking implementation suspect and extremely hard to follow. Change the driver such that the reference taking routines are now normal ref increments/decrements and callout on refcount=0. With this in place, the rework can be done such that the node structure is fully removed and deallocated when the remote port no longer exists and all references are removed. This removal logic, and the basic ref counting are intrically tied, thus in a single patch. Link: https://lore.kernel.org/r/20201115192646.12977-2-james.smart@broadcom.com Co-developed-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <james.smart@broadcom.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-11-15 19:26:30 +00:00
int node_dropped = ndlp->nlp_flag & NLP_DROPPED;
char name1[16], name2[16];
lpfc_printf_vlog(vport, KERN_INFO, LOG_NODE,
"0904 NPort state transition x%06x, %s -> %s\n",
ndlp->nlp_DID,
lpfc_nlp_state_name(name1, sizeof(name1), old_state),
lpfc_nlp_state_name(name2, sizeof(name2), state));
lpfc_debugfs_disc_trc(vport, LPFC_DISC_TRC_NODE,
"node statechg did:x%x old:%d ste:%d",
ndlp->nlp_DID, old_state, state);
scsi: lpfc: Rework remote port ref counting and node freeing When a remote port is disconnected and disappears, its node structure (ndlp) stays allocated and on a vport node list. While on the list it can be matched, thus requires validation checks on state to be added in numerous code paths. If the node comes back, its possible for there to be multiple node structures for the same device on the vport node list. There is no reason to keep the node structure around after it is no longer in existence, and the current implementation creates problems for itself (multiple nodes) and lots of unnecessary code for state validation. Additionally, the reference taking on the node structure didn't follow the normal model used by the kernel kref api. It included lots of odd logic to match state with reference count. The combination of this odd logic plus the way it was implicitly used in the discovery engine made its reference taking implementation suspect and extremely hard to follow. Change the driver such that the reference taking routines are now normal ref increments/decrements and callout on refcount=0. With this in place, the rework can be done such that the node structure is fully removed and deallocated when the remote port no longer exists and all references are removed. This removal logic, and the basic ref counting are intrically tied, thus in a single patch. Link: https://lore.kernel.org/r/20201115192646.12977-2-james.smart@broadcom.com Co-developed-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <james.smart@broadcom.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-11-15 19:26:30 +00:00
if (node_dropped && old_state == NLP_STE_UNUSED_NODE &&
state != NLP_STE_UNUSED_NODE) {
ndlp->nlp_flag &= ~NLP_DROPPED;
lpfc_nlp_get(ndlp);
}
if (old_state == NLP_STE_NPR_NODE &&
state != NLP_STE_NPR_NODE)
lpfc_cancel_retry_delay_tmo(vport, ndlp);
if (old_state == NLP_STE_UNMAPPED_NODE) {
ndlp->nlp_flag &= ~NLP_TGT_NO_SCSIID;
ndlp->nlp_type &= ~NLP_FC_NODE;
}
if (list_empty(&ndlp->nlp_listp)) {
spin_lock_irq(shost->host_lock);
list_add_tail(&ndlp->nlp_listp, &vport->fc_nodes);
spin_unlock_irq(shost->host_lock);
} else if (old_state)
lpfc_nlp_counters(vport, old_state, -1);
ndlp->nlp_state = state;
lpfc_nlp_counters(vport, state, 1);
lpfc_nlp_state_cleanup(vport, ndlp, old_state, state);
}
void
lpfc_enqueue_node(struct lpfc_vport *vport, struct lpfc_nodelist *ndlp)
{
struct Scsi_Host *shost = lpfc_shost_from_vport(vport);
if (list_empty(&ndlp->nlp_listp)) {
spin_lock_irq(shost->host_lock);
list_add_tail(&ndlp->nlp_listp, &vport->fc_nodes);
spin_unlock_irq(shost->host_lock);
}
}
void
lpfc_dequeue_node(struct lpfc_vport *vport, struct lpfc_nodelist *ndlp)
{
struct Scsi_Host *shost = lpfc_shost_from_vport(vport);
lpfc_cancel_retry_delay_tmo(vport, ndlp);
if (ndlp->nlp_state && !list_empty(&ndlp->nlp_listp))
lpfc_nlp_counters(vport, ndlp->nlp_state, -1);
spin_lock_irq(shost->host_lock);
list_del_init(&ndlp->nlp_listp);
spin_unlock_irq(shost->host_lock);
lpfc_nlp_state_cleanup(vport, ndlp, ndlp->nlp_state,
NLP_STE_UNUSED_NODE);
}
/**
* lpfc_initialize_node - Initialize all fields of node object
* @vport: Pointer to Virtual Port object.
* @ndlp: Pointer to FC node object.
* @did: FC_ID of the node.
[SCSI] lpfc 8.3.1: misc fixes/changes 8.3.1 Fixes/Changes : - Fix incorrect byte-swapping on word 4 of IOCB (data length) which caused LUNs to not be discovered on big-endian (e.g. PPC) - Remove a bad cast of MBslimaddr which loses the __iomem (sparse) - Make lpfc_debugfs_mask_disc_trc static (sparse) - Correct misspelled word BlockGuard in lpfc_logmsg.h comment - Replaced repeated code segment for canceling IOCBs from a list with a function call, lpfc_sli_cancel_iocbs(). - Increased HBQ buffers to support 40KB SSC sequences. - Added sysfs interface to update speed and topology parameter without link bounce. - Fixed bug with sysfs fc_host WWNs not being updated after changing the WWNs. - Check if the active mailbox is NULL in the beginning of the mailbox timeout handler - fixes panic in the mailbox timeout handler while running IO stress test - Fixed system panic in lpfc_pci_remove_one() due to ndlp indirect reference to phba through vport - Removed de-reference of scsi device after call to scsi_done() to fix panic in scsi completion path while accessing scsi device after scsi_done is called. - Fixed "Nodelist not empty" message when unloading the driver after target reboot test - Added LP2105 HBA model description - Added code to print all 16 words of unrecognized ASYNC events - Fixed memory leak in vport create + delete loop - Added support for handling dual error bit from HBA - Fixed a driver NULL pointer dereference in lpfc_sli_process_sol_iocb - Fixed a discovery bug with FC switch reboot in lpfc_setup_disc_node - Take NULL termintator into account when calculating available buffer space Signed-off-by: James Smart <james.smart@emulex.com> Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
2009-04-06 22:48:10 +00:00
*
* This function is always called when node object need to be initialized.
* It initializes all the fields of the node object. Although the reference
* to phba from @ndlp can be obtained indirectly through it's reference to
* @vport, a direct reference to phba is taken here by @ndlp. This is due
* to the life-span of the @ndlp might go beyond the existence of @vport as
* the final release of ndlp is determined by its reference count. And, the
* operation on @ndlp needs the reference to phba.
**/
static inline void
lpfc_initialize_node(struct lpfc_vport *vport, struct lpfc_nodelist *ndlp,
uint32_t did)
{
INIT_LIST_HEAD(&ndlp->els_retry_evt.evt_listp);
INIT_LIST_HEAD(&ndlp->dev_loss_evt.evt_listp);
timer_setup(&ndlp->nlp_delayfunc, lpfc_els_retry_delay, 0);
scsi: lpfc: Fix scheduling call while in softirq context in lpfc_unreg_rpi The following call trace was seen during HBA reset testing: BUG: scheduling while atomic: swapper/2/0/0x10000100 ... Call Trace: dump_stack+0x19/0x1b __schedule_bug+0x64/0x72 __schedule+0x782/0x840 __cond_resched+0x26/0x30 _cond_resched+0x3a/0x50 mempool_alloc+0xa0/0x170 lpfc_unreg_rpi+0x151/0x630 [lpfc] lpfc_sli_abts_recover_port+0x171/0x190 [lpfc] lpfc_sli4_abts_err_handler+0xb2/0x1f0 [lpfc] lpfc_sli4_io_xri_aborted+0x256/0x300 [lpfc] lpfc_sli4_sp_handle_abort_xri_wcqe.isra.51+0xa3/0x190 [lpfc] lpfc_sli4_fp_handle_cqe+0x89/0x4d0 [lpfc] __lpfc_sli4_process_cq+0xdb/0x2e0 [lpfc] __lpfc_sli4_hba_process_cq+0x41/0x100 [lpfc] lpfc_cq_poll_hdler+0x1a/0x30 [lpfc] irq_poll_softirq+0xc7/0x100 __do_softirq+0xf5/0x280 call_softirq+0x1c/0x30 do_softirq+0x65/0xa0 irq_exit+0x105/0x110 do_IRQ+0x56/0xf0 common_interrupt+0x16a/0x16a With the conversion to blk_io_poll for better interrupt latency in normal cases, it introduced this code path, executed when I/O aborts or logouts are seen, which attempts to allocate memory for a mailbox command to be issued. The allocation is GFP_KERNEL, thus it could attempt to sleep. Fix by creating a work element that performs the event handling for the remote port. This will have the mailbox commands and other items performed in the work element, not the irq. A much better method as the "irq" routine does not stall while performing all this deep handling code. Ensure that allocation failures are handled and send LOGO on failure. Additionally, enlarge the mailbox memory pool to reduce the possibility of additional allocation in this path. Link: https://lore.kernel.org/r/20201020202719.54726-3-james.smart@broadcom.com Fixes: 317aeb83c92b ("scsi: lpfc: Add blk_io_poll support for latency improvment") Cc: <stable@vger.kernel.org> # v5.9+ Co-developed-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <james.smart@broadcom.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-10-20 20:27:12 +00:00
INIT_LIST_HEAD(&ndlp->recovery_evt.evt_listp);
ndlp->nlp_DID = did;
ndlp->vport = vport;
[SCSI] lpfc 8.3.1: misc fixes/changes 8.3.1 Fixes/Changes : - Fix incorrect byte-swapping on word 4 of IOCB (data length) which caused LUNs to not be discovered on big-endian (e.g. PPC) - Remove a bad cast of MBslimaddr which loses the __iomem (sparse) - Make lpfc_debugfs_mask_disc_trc static (sparse) - Correct misspelled word BlockGuard in lpfc_logmsg.h comment - Replaced repeated code segment for canceling IOCBs from a list with a function call, lpfc_sli_cancel_iocbs(). - Increased HBQ buffers to support 40KB SSC sequences. - Added sysfs interface to update speed and topology parameter without link bounce. - Fixed bug with sysfs fc_host WWNs not being updated after changing the WWNs. - Check if the active mailbox is NULL in the beginning of the mailbox timeout handler - fixes panic in the mailbox timeout handler while running IO stress test - Fixed system panic in lpfc_pci_remove_one() due to ndlp indirect reference to phba through vport - Removed de-reference of scsi device after call to scsi_done() to fix panic in scsi completion path while accessing scsi device after scsi_done is called. - Fixed "Nodelist not empty" message when unloading the driver after target reboot test - Added LP2105 HBA model description - Added code to print all 16 words of unrecognized ASYNC events - Fixed memory leak in vport create + delete loop - Added support for handling dual error bit from HBA - Fixed a driver NULL pointer dereference in lpfc_sli_process_sol_iocb - Fixed a discovery bug with FC switch reboot in lpfc_setup_disc_node - Take NULL termintator into account when calculating available buffer space Signed-off-by: James Smart <james.smart@emulex.com> Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
2009-04-06 22:48:10 +00:00
ndlp->phba = vport->phba;
ndlp->nlp_sid = NLP_NO_SID;
ndlp->nlp_fc4_type = NLP_FC4_NONE;
kref_init(&ndlp->kref);
atomic_set(&ndlp->cmd_pending, 0);
ndlp->cmd_qdepth = vport->cfg_tgt_queue_depth;
ndlp->nlp_defer_did = NLP_EVT_NOTHING_PENDING;
}
void
lpfc_drop_node(struct lpfc_vport *vport, struct lpfc_nodelist *ndlp)
{
/*
* Use of lpfc_drop_node and UNUSED list: lpfc_drop_node should
* be used if we wish to issue the "last" lpfc_nlp_put() to remove
* the ndlp from the vport. The ndlp marked as UNUSED on the list
* until ALL other outstanding threads have completed. We check
* that the ndlp not already in the UNUSED state before we proceed.
*/
if (ndlp->nlp_state == NLP_STE_UNUSED_NODE)
return;
lpfc_nlp_set_state(vport, ndlp, NLP_STE_UNUSED_NODE);
scsi: lpfc: Rework remote port ref counting and node freeing When a remote port is disconnected and disappears, its node structure (ndlp) stays allocated and on a vport node list. While on the list it can be matched, thus requires validation checks on state to be added in numerous code paths. If the node comes back, its possible for there to be multiple node structures for the same device on the vport node list. There is no reason to keep the node structure around after it is no longer in existence, and the current implementation creates problems for itself (multiple nodes) and lots of unnecessary code for state validation. Additionally, the reference taking on the node structure didn't follow the normal model used by the kernel kref api. It included lots of odd logic to match state with reference count. The combination of this odd logic plus the way it was implicitly used in the discovery engine made its reference taking implementation suspect and extremely hard to follow. Change the driver such that the reference taking routines are now normal ref increments/decrements and callout on refcount=0. With this in place, the rework can be done such that the node structure is fully removed and deallocated when the remote port no longer exists and all references are removed. This removal logic, and the basic ref counting are intrically tied, thus in a single patch. Link: https://lore.kernel.org/r/20201115192646.12977-2-james.smart@broadcom.com Co-developed-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <james.smart@broadcom.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-11-15 19:26:30 +00:00
ndlp->nlp_flag |= NLP_DROPPED;
if (vport->phba->sli_rev == LPFC_SLI_REV4) {
lpfc_cleanup_vports_rrqs(vport, ndlp);
lpfc_unreg_rpi(vport, ndlp);
}
lpfc_nlp_put(ndlp);
return;
}
/*
* Start / ReStart rescue timer for Discovery / RSCN handling
*/
void
lpfc_set_disctmo(struct lpfc_vport *vport)
{
struct Scsi_Host *shost = lpfc_shost_from_vport(vport);
struct lpfc_hba *phba = vport->phba;
uint32_t tmo;
if (vport->port_state == LPFC_LOCAL_CFG_LINK) {
/* For FAN, timeout should be greater than edtov */
tmo = (((phba->fc_edtov + 999) / 1000) + 1);
} else {
/* Normal discovery timeout should be > than ELS/CT timeout
* FC spec states we need 3 * ratov for CT requests
*/
tmo = ((phba->fc_ratov * 3) + 3);
}
if (!timer_pending(&vport->fc_disctmo)) {
lpfc_debugfs_disc_trc(vport, LPFC_DISC_TRC_ELS_CMD,
"set disc timer: tmo:x%x state:x%x flg:x%x",
tmo, vport->port_state, vport->fc_flag);
}
mod_timer(&vport->fc_disctmo, jiffies + msecs_to_jiffies(1000 * tmo));
spin_lock_irq(shost->host_lock);
vport->fc_flag |= FC_DISC_TMO;
spin_unlock_irq(shost->host_lock);
/* Start Discovery Timer state <hba_state> */
lpfc_printf_vlog(vport, KERN_INFO, LOG_DISCOVERY,
"0247 Start Discovery Timer state x%x "
"Data: x%x x%lx x%x x%x\n",
vport->port_state, tmo,
(unsigned long)&vport->fc_disctmo, vport->fc_plogi_cnt,
vport->fc_adisc_cnt);
return;
}
/*
* Cancel rescue timer for Discovery / RSCN handling
*/
int
lpfc_can_disctmo(struct lpfc_vport *vport)
{
struct Scsi_Host *shost = lpfc_shost_from_vport(vport);
unsigned long iflags;
lpfc_debugfs_disc_trc(vport, LPFC_DISC_TRC_ELS_CMD,
"can disc timer: state:x%x rtry:x%x flg:x%x",
vport->port_state, vport->fc_ns_retry, vport->fc_flag);
/* Turn off discovery timer if its running */
if (vport->fc_flag & FC_DISC_TMO ||
timer_pending(&vport->fc_disctmo)) {
spin_lock_irqsave(shost->host_lock, iflags);
vport->fc_flag &= ~FC_DISC_TMO;
spin_unlock_irqrestore(shost->host_lock, iflags);
del_timer_sync(&vport->fc_disctmo);
spin_lock_irqsave(&vport->work_port_lock, iflags);
vport->work_port_events &= ~WORKER_DISC_TMO;
spin_unlock_irqrestore(&vport->work_port_lock, iflags);
}
/* Cancel Discovery Timer state <hba_state> */
lpfc_printf_vlog(vport, KERN_INFO, LOG_DISCOVERY,
"0248 Cancel Discovery Timer state x%x "
"Data: x%x x%x x%x\n",
vport->port_state, vport->fc_flag,
vport->fc_plogi_cnt, vport->fc_adisc_cnt);
return 0;
}
/*
* Check specified ring for outstanding IOCB on the SLI queue
* Return true if iocb matches the specified nport
*/
int
lpfc_check_sli_ndlp(struct lpfc_hba *phba,
struct lpfc_sli_ring *pring,
struct lpfc_iocbq *iocb,
struct lpfc_nodelist *ndlp)
{
struct lpfc_vport *vport = ndlp->vport;
u8 ulp_command;
u16 ulp_context;
u32 remote_id;
if (iocb->vport != vport)
return 0;
ulp_command = get_job_cmnd(phba, iocb);
ulp_context = get_job_ulpcontext(phba, iocb);
remote_id = get_job_els_rsp64_did(phba, iocb);
if (pring->ringno == LPFC_ELS_RING) {
switch (ulp_command) {
case CMD_GEN_REQUEST64_CR:
if (iocb->ndlp == ndlp)
return 1;
fallthrough;
case CMD_ELS_REQUEST64_CR:
if (remote_id == ndlp->nlp_DID)
return 1;
fallthrough;
case CMD_XMIT_ELS_RSP64_CX:
if (iocb->ndlp == ndlp)
return 1;
}
scsi: lpfc: NVME Initiator: Base modifications NVME Initiator: Base modifications This patch adds base modifications for NVME initiator support. The base modifications consist of: - Formal split of SLI3 rings from SLI-4 WQs (sometimes referred to as rings as well) as implementation now widely varies between the two. - Addition of configuration modes: SCSI initiator only; NVME initiator only; NVME target only; and SCSI and NVME initiator. The configuration mode drives overall adapter configuration, offloads enabled, and resource splits. NVME support is only available on SLI-4 devices and newer fw. - Implements the following based on configuration mode: - Exchange resources are split by protocol; Obviously, if only 1 mode, then no split occurs. Default is 50/50. module attribute allows tuning. - Pools and config parameters are separated per-protocol - Each protocol has it's own set of queues, but share interrupt vectors. SCSI: SLI3 devices have few queues and the original style of queue allocation remains. SLI4 devices piggy back on an "io-channel" concept that eventually needs to merge with scsi-mq/blk-mq support (it is underway). For now, the paradigm continues as it existed prior. io channel allocates N msix and N WQs (N=4 default) and either round robins or uses cpu # modulo N for scheduling. A bunch of module parameters allow the configuration to be tuned. NVME (initiator): Allocates an msix per cpu (or whatever pci_alloc_irq_vectors gets) Allocates a WQ per cpu, and maps the WQs to msix on a WQ # modulo msix vector count basis. Module parameters exist to cap/control the config if desired. - Each protocol has its own buffer and dma pools. I apologize for the size of the patch. Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <james.smart@broadcom.com> ---- Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2017-02-12 21:52:30 +00:00
} else if (pring->ringno == LPFC_FCP_RING) {
/* Skip match check if waiting to relogin to FCP target */
if ((ndlp->nlp_type & NLP_FCP_TARGET) &&
(ndlp->nlp_flag & NLP_DELAY_TMO)) {
return 0;
}
if (ulp_context == ndlp->nlp_rpi)
return 1;
}
return 0;
}
scsi: lpfc: NVME Initiator: Base modifications NVME Initiator: Base modifications This patch adds base modifications for NVME initiator support. The base modifications consist of: - Formal split of SLI3 rings from SLI-4 WQs (sometimes referred to as rings as well) as implementation now widely varies between the two. - Addition of configuration modes: SCSI initiator only; NVME initiator only; NVME target only; and SCSI and NVME initiator. The configuration mode drives overall adapter configuration, offloads enabled, and resource splits. NVME support is only available on SLI-4 devices and newer fw. - Implements the following based on configuration mode: - Exchange resources are split by protocol; Obviously, if only 1 mode, then no split occurs. Default is 50/50. module attribute allows tuning. - Pools and config parameters are separated per-protocol - Each protocol has it's own set of queues, but share interrupt vectors. SCSI: SLI3 devices have few queues and the original style of queue allocation remains. SLI4 devices piggy back on an "io-channel" concept that eventually needs to merge with scsi-mq/blk-mq support (it is underway). For now, the paradigm continues as it existed prior. io channel allocates N msix and N WQs (N=4 default) and either round robins or uses cpu # modulo N for scheduling. A bunch of module parameters allow the configuration to be tuned. NVME (initiator): Allocates an msix per cpu (or whatever pci_alloc_irq_vectors gets) Allocates a WQ per cpu, and maps the WQs to msix on a WQ # modulo msix vector count basis. Module parameters exist to cap/control the config if desired. - Each protocol has its own buffer and dma pools. I apologize for the size of the patch. Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <james.smart@broadcom.com> ---- Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2017-02-12 21:52:30 +00:00
static void
__lpfc_dequeue_nport_iocbs(struct lpfc_hba *phba,
struct lpfc_nodelist *ndlp, struct lpfc_sli_ring *pring,
struct list_head *dequeue_list)
{
struct lpfc_iocbq *iocb, *next_iocb;
list_for_each_entry_safe(iocb, next_iocb, &pring->txq, list) {
/* Check to see if iocb matches the nport */
if (lpfc_check_sli_ndlp(phba, pring, iocb, ndlp))
/* match, dequeue */
list_move_tail(&iocb->list, dequeue_list);
}
}
static void
lpfc_sli3_dequeue_nport_iocbs(struct lpfc_hba *phba,
struct lpfc_nodelist *ndlp, struct list_head *dequeue_list)
{
struct lpfc_sli *psli = &phba->sli;
uint32_t i;
spin_lock_irq(&phba->hbalock);
for (i = 0; i < psli->num_rings; i++)
__lpfc_dequeue_nport_iocbs(phba, ndlp, &psli->sli3_ring[i],
dequeue_list);
spin_unlock_irq(&phba->hbalock);
}
static void
lpfc_sli4_dequeue_nport_iocbs(struct lpfc_hba *phba,
struct lpfc_nodelist *ndlp, struct list_head *dequeue_list)
{
struct lpfc_sli_ring *pring;
struct lpfc_queue *qp = NULL;
spin_lock_irq(&phba->hbalock);
list_for_each_entry(qp, &phba->sli4_hba.lpfc_wq_list, wq_list) {
pring = qp->pring;
if (!pring)
continue;
spin_lock(&pring->ring_lock);
scsi: lpfc: NVME Initiator: Base modifications NVME Initiator: Base modifications This patch adds base modifications for NVME initiator support. The base modifications consist of: - Formal split of SLI3 rings from SLI-4 WQs (sometimes referred to as rings as well) as implementation now widely varies between the two. - Addition of configuration modes: SCSI initiator only; NVME initiator only; NVME target only; and SCSI and NVME initiator. The configuration mode drives overall adapter configuration, offloads enabled, and resource splits. NVME support is only available on SLI-4 devices and newer fw. - Implements the following based on configuration mode: - Exchange resources are split by protocol; Obviously, if only 1 mode, then no split occurs. Default is 50/50. module attribute allows tuning. - Pools and config parameters are separated per-protocol - Each protocol has it's own set of queues, but share interrupt vectors. SCSI: SLI3 devices have few queues and the original style of queue allocation remains. SLI4 devices piggy back on an "io-channel" concept that eventually needs to merge with scsi-mq/blk-mq support (it is underway). For now, the paradigm continues as it existed prior. io channel allocates N msix and N WQs (N=4 default) and either round robins or uses cpu # modulo N for scheduling. A bunch of module parameters allow the configuration to be tuned. NVME (initiator): Allocates an msix per cpu (or whatever pci_alloc_irq_vectors gets) Allocates a WQ per cpu, and maps the WQs to msix on a WQ # modulo msix vector count basis. Module parameters exist to cap/control the config if desired. - Each protocol has its own buffer and dma pools. I apologize for the size of the patch. Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <james.smart@broadcom.com> ---- Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2017-02-12 21:52:30 +00:00
__lpfc_dequeue_nport_iocbs(phba, ndlp, pring, dequeue_list);
spin_unlock(&pring->ring_lock);
scsi: lpfc: NVME Initiator: Base modifications NVME Initiator: Base modifications This patch adds base modifications for NVME initiator support. The base modifications consist of: - Formal split of SLI3 rings from SLI-4 WQs (sometimes referred to as rings as well) as implementation now widely varies between the two. - Addition of configuration modes: SCSI initiator only; NVME initiator only; NVME target only; and SCSI and NVME initiator. The configuration mode drives overall adapter configuration, offloads enabled, and resource splits. NVME support is only available on SLI-4 devices and newer fw. - Implements the following based on configuration mode: - Exchange resources are split by protocol; Obviously, if only 1 mode, then no split occurs. Default is 50/50. module attribute allows tuning. - Pools and config parameters are separated per-protocol - Each protocol has it's own set of queues, but share interrupt vectors. SCSI: SLI3 devices have few queues and the original style of queue allocation remains. SLI4 devices piggy back on an "io-channel" concept that eventually needs to merge with scsi-mq/blk-mq support (it is underway). For now, the paradigm continues as it existed prior. io channel allocates N msix and N WQs (N=4 default) and either round robins or uses cpu # modulo N for scheduling. A bunch of module parameters allow the configuration to be tuned. NVME (initiator): Allocates an msix per cpu (or whatever pci_alloc_irq_vectors gets) Allocates a WQ per cpu, and maps the WQs to msix on a WQ # modulo msix vector count basis. Module parameters exist to cap/control the config if desired. - Each protocol has its own buffer and dma pools. I apologize for the size of the patch. Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <james.smart@broadcom.com> ---- Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2017-02-12 21:52:30 +00:00
}
spin_unlock_irq(&phba->hbalock);
}
/*
* Free resources / clean up outstanding I/Os
* associated with nlp_rpi in the LPFC_NODELIST entry.
*/
static int
lpfc_no_rpi(struct lpfc_hba *phba, struct lpfc_nodelist *ndlp)
{
LIST_HEAD(completions);
lpfc_fabric_abort_nport(ndlp);
/*
* Everything that matches on txcmplq will be returned
* by firmware with a no rpi error.
*/
if (ndlp->nlp_flag & NLP_RPI_REGISTERED) {
scsi: lpfc: NVME Initiator: Base modifications NVME Initiator: Base modifications This patch adds base modifications for NVME initiator support. The base modifications consist of: - Formal split of SLI3 rings from SLI-4 WQs (sometimes referred to as rings as well) as implementation now widely varies between the two. - Addition of configuration modes: SCSI initiator only; NVME initiator only; NVME target only; and SCSI and NVME initiator. The configuration mode drives overall adapter configuration, offloads enabled, and resource splits. NVME support is only available on SLI-4 devices and newer fw. - Implements the following based on configuration mode: - Exchange resources are split by protocol; Obviously, if only 1 mode, then no split occurs. Default is 50/50. module attribute allows tuning. - Pools and config parameters are separated per-protocol - Each protocol has it's own set of queues, but share interrupt vectors. SCSI: SLI3 devices have few queues and the original style of queue allocation remains. SLI4 devices piggy back on an "io-channel" concept that eventually needs to merge with scsi-mq/blk-mq support (it is underway). For now, the paradigm continues as it existed prior. io channel allocates N msix and N WQs (N=4 default) and either round robins or uses cpu # modulo N for scheduling. A bunch of module parameters allow the configuration to be tuned. NVME (initiator): Allocates an msix per cpu (or whatever pci_alloc_irq_vectors gets) Allocates a WQ per cpu, and maps the WQs to msix on a WQ # modulo msix vector count basis. Module parameters exist to cap/control the config if desired. - Each protocol has its own buffer and dma pools. I apologize for the size of the patch. Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <james.smart@broadcom.com> ---- Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2017-02-12 21:52:30 +00:00
if (phba->sli_rev != LPFC_SLI_REV4)
lpfc_sli3_dequeue_nport_iocbs(phba, ndlp, &completions);
else
lpfc_sli4_dequeue_nport_iocbs(phba, ndlp, &completions);
}
[SCSI] lpfc 8.3.1: misc fixes/changes 8.3.1 Fixes/Changes : - Fix incorrect byte-swapping on word 4 of IOCB (data length) which caused LUNs to not be discovered on big-endian (e.g. PPC) - Remove a bad cast of MBslimaddr which loses the __iomem (sparse) - Make lpfc_debugfs_mask_disc_trc static (sparse) - Correct misspelled word BlockGuard in lpfc_logmsg.h comment - Replaced repeated code segment for canceling IOCBs from a list with a function call, lpfc_sli_cancel_iocbs(). - Increased HBQ buffers to support 40KB SSC sequences. - Added sysfs interface to update speed and topology parameter without link bounce. - Fixed bug with sysfs fc_host WWNs not being updated after changing the WWNs. - Check if the active mailbox is NULL in the beginning of the mailbox timeout handler - fixes panic in the mailbox timeout handler while running IO stress test - Fixed system panic in lpfc_pci_remove_one() due to ndlp indirect reference to phba through vport - Removed de-reference of scsi device after call to scsi_done() to fix panic in scsi completion path while accessing scsi device after scsi_done is called. - Fixed "Nodelist not empty" message when unloading the driver after target reboot test - Added LP2105 HBA model description - Added code to print all 16 words of unrecognized ASYNC events - Fixed memory leak in vport create + delete loop - Added support for handling dual error bit from HBA - Fixed a driver NULL pointer dereference in lpfc_sli_process_sol_iocb - Fixed a discovery bug with FC switch reboot in lpfc_setup_disc_node - Take NULL termintator into account when calculating available buffer space Signed-off-by: James Smart <james.smart@emulex.com> Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
2009-04-06 22:48:10 +00:00
/* Cancel all the IOCBs from the completions list */
lpfc_sli_cancel_iocbs(phba, &completions, IOSTAT_LOCAL_REJECT,
IOERR_SLI_ABORTED);
return 0;
}
/**
* lpfc_nlp_logo_unreg - Unreg mailbox completion handler before LOGO
* @phba: Pointer to HBA context object.
* @pmb: Pointer to mailbox object.
*
* This function will issue an ELS LOGO command after completing
* the UNREG_RPI.
**/
static void
lpfc_nlp_logo_unreg(struct lpfc_hba *phba, LPFC_MBOXQ_t *pmb)
{
struct lpfc_vport *vport = pmb->vport;
struct lpfc_nodelist *ndlp;
ndlp = (struct lpfc_nodelist *)(pmb->ctx_ndlp);
if (!ndlp)
return;
lpfc_issue_els_logo(vport, ndlp, 0);
/* Check to see if there are any deferred events to process */
if ((ndlp->nlp_flag & NLP_UNREG_INP) &&
(ndlp->nlp_defer_did != NLP_EVT_NOTHING_PENDING)) {
lpfc_printf_vlog(vport, KERN_INFO, LOG_DISCOVERY,
"1434 UNREG cmpl deferred logo x%x "
"on NPort x%x Data: x%x x%px\n",
ndlp->nlp_rpi, ndlp->nlp_DID,
ndlp->nlp_defer_did, ndlp);
ndlp->nlp_flag &= ~NLP_UNREG_INP;
ndlp->nlp_defer_did = NLP_EVT_NOTHING_PENDING;
lpfc_issue_els_plogi(vport, ndlp->nlp_DID, 0);
} else {
scsi: lpfc: Fix unreleased RPIs when NPIV ports are created While testing NPIV and watching logins and used RPI levels, it was seen the used RPI count was much higher than the number of remote ports discovered. Code inspection showed that remote port removals on any NPIV instance are releasing the RPI, but not performing an UNREG_RPI with the adapter thus the reference counting never fully drops and the RPI is never fully released. This was happening on NPIV nodes due to a log of fabric ELS's to fabric addresses. This lack of UNREG_RPI was introduced by a prior node rework patch that performed the UNREG_RPI as part of node cleanup. To resolve the issue, do the following: - Restore the RPI release code, but move the location to so that it is in line with the new node cleanup design. - NPIV ports now release the RPI and drop the node when the caller sets the NLP_RELEASE_RPI flag. - Set the NLP_RELEASE_RPI flag in node cleanup which will trigger a release of RPI to free pool. - Ensure there's an UNREG_RPI at LOGO completion so that RPI release is completed. - Stop offline_prep from skipping nodes that are UNUSED. The RPI may not have been released. - Stop the default RPI handling in lpfc_cmpl_els_rsp() for SLI4. - Fixed up debugfs RPI displays for better debugging. Fixes: a70e63eee1c1 ("scsi: lpfc: Fix NPIV Fabric Node reference counting") Link: https://lore.kernel.org/r/20210514195559.119853-2-jsmart2021@gmail.com Cc: <stable@vger.kernel.org> # v5.11+ Co-developed-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2021-05-14 19:55:49 +00:00
/* NLP_RELEASE_RPI is only set for SLI4 ports. */
if (ndlp->nlp_flag & NLP_RELEASE_RPI) {
lpfc_sli4_free_rpi(vport->phba, ndlp->nlp_rpi);
scsi: lpfc: Fix unreleased RPIs when NPIV ports are created While testing NPIV and watching logins and used RPI levels, it was seen the used RPI count was much higher than the number of remote ports discovered. Code inspection showed that remote port removals on any NPIV instance are releasing the RPI, but not performing an UNREG_RPI with the adapter thus the reference counting never fully drops and the RPI is never fully released. This was happening on NPIV nodes due to a log of fabric ELS's to fabric addresses. This lack of UNREG_RPI was introduced by a prior node rework patch that performed the UNREG_RPI as part of node cleanup. To resolve the issue, do the following: - Restore the RPI release code, but move the location to so that it is in line with the new node cleanup design. - NPIV ports now release the RPI and drop the node when the caller sets the NLP_RELEASE_RPI flag. - Set the NLP_RELEASE_RPI flag in node cleanup which will trigger a release of RPI to free pool. - Ensure there's an UNREG_RPI at LOGO completion so that RPI release is completed. - Stop offline_prep from skipping nodes that are UNUSED. The RPI may not have been released. - Stop the default RPI handling in lpfc_cmpl_els_rsp() for SLI4. - Fixed up debugfs RPI displays for better debugging. Fixes: a70e63eee1c1 ("scsi: lpfc: Fix NPIV Fabric Node reference counting") Link: https://lore.kernel.org/r/20210514195559.119853-2-jsmart2021@gmail.com Cc: <stable@vger.kernel.org> # v5.11+ Co-developed-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2021-05-14 19:55:49 +00:00
spin_lock_irq(&ndlp->lock);
ndlp->nlp_flag &= ~NLP_RELEASE_RPI;
scsi: lpfc: Fix device recovery errors after PLOGI failures When target-side fault injections are made, the driver isn't reconnecting to the remote port. The driver is logging "2753" error messages which state: "PLOGI failure DID:1B2400 Status:x3/xf0240008" The failures status is indicating a Illegal field error, which points to the Temporary RPI field being used for the ELS. This error typically means the driver used an RPI that was already registered (shouldn't be registered if using it in this context). Study has found that if the driver were in discovery attempts and encountered an error, it wouldn't flag the temporary rpi in error. Yet the rpi was released for reallocation in these error paths and another ELS could allocate the rpi. In the failure situation a retry was done on an ELS that had encountered an error, and as the rpi wasn't marked in error, the ELS reused the rpi it originally allocated. But that rpi had been allocated by a different ELS issued after the original error and before the retry attempt. The different ELS had succeeded and the RPI was registered. Fix by marking the rpi state for the node to be in error, aka as needing reallocation, upon an error in the els processing. Error state marking is always done prior to release back to the internal rpi free list, which the driver wasn't doing in cases prior. Also enhanced some of the logging to help in the next case of problem troubleshooting. Link: https://lore.kernel.org/r/20190922035906.10977-7-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2019-09-22 03:58:52 +00:00
ndlp->nlp_rpi = LPFC_RPI_ALLOC_ERROR;
scsi: lpfc: Fix unreleased RPIs when NPIV ports are created While testing NPIV and watching logins and used RPI levels, it was seen the used RPI count was much higher than the number of remote ports discovered. Code inspection showed that remote port removals on any NPIV instance are releasing the RPI, but not performing an UNREG_RPI with the adapter thus the reference counting never fully drops and the RPI is never fully released. This was happening on NPIV nodes due to a log of fabric ELS's to fabric addresses. This lack of UNREG_RPI was introduced by a prior node rework patch that performed the UNREG_RPI as part of node cleanup. To resolve the issue, do the following: - Restore the RPI release code, but move the location to so that it is in line with the new node cleanup design. - NPIV ports now release the RPI and drop the node when the caller sets the NLP_RELEASE_RPI flag. - Set the NLP_RELEASE_RPI flag in node cleanup which will trigger a release of RPI to free pool. - Ensure there's an UNREG_RPI at LOGO completion so that RPI release is completed. - Stop offline_prep from skipping nodes that are UNUSED. The RPI may not have been released. - Stop the default RPI handling in lpfc_cmpl_els_rsp() for SLI4. - Fixed up debugfs RPI displays for better debugging. Fixes: a70e63eee1c1 ("scsi: lpfc: Fix NPIV Fabric Node reference counting") Link: https://lore.kernel.org/r/20210514195559.119853-2-jsmart2021@gmail.com Cc: <stable@vger.kernel.org> # v5.11+ Co-developed-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2021-05-14 19:55:49 +00:00
spin_unlock_irq(&ndlp->lock);
}
scsi: lpfc: Fix unreleased RPIs when NPIV ports are created While testing NPIV and watching logins and used RPI levels, it was seen the used RPI count was much higher than the number of remote ports discovered. Code inspection showed that remote port removals on any NPIV instance are releasing the RPI, but not performing an UNREG_RPI with the adapter thus the reference counting never fully drops and the RPI is never fully released. This was happening on NPIV nodes due to a log of fabric ELS's to fabric addresses. This lack of UNREG_RPI was introduced by a prior node rework patch that performed the UNREG_RPI as part of node cleanup. To resolve the issue, do the following: - Restore the RPI release code, but move the location to so that it is in line with the new node cleanup design. - NPIV ports now release the RPI and drop the node when the caller sets the NLP_RELEASE_RPI flag. - Set the NLP_RELEASE_RPI flag in node cleanup which will trigger a release of RPI to free pool. - Ensure there's an UNREG_RPI at LOGO completion so that RPI release is completed. - Stop offline_prep from skipping nodes that are UNUSED. The RPI may not have been released. - Stop the default RPI handling in lpfc_cmpl_els_rsp() for SLI4. - Fixed up debugfs RPI displays for better debugging. Fixes: a70e63eee1c1 ("scsi: lpfc: Fix NPIV Fabric Node reference counting") Link: https://lore.kernel.org/r/20210514195559.119853-2-jsmart2021@gmail.com Cc: <stable@vger.kernel.org> # v5.11+ Co-developed-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2021-05-14 19:55:49 +00:00
spin_lock_irq(&ndlp->lock);
ndlp->nlp_flag &= ~NLP_UNREG_INP;
scsi: lpfc: Fix unreleased RPIs when NPIV ports are created While testing NPIV and watching logins and used RPI levels, it was seen the used RPI count was much higher than the number of remote ports discovered. Code inspection showed that remote port removals on any NPIV instance are releasing the RPI, but not performing an UNREG_RPI with the adapter thus the reference counting never fully drops and the RPI is never fully released. This was happening on NPIV nodes due to a log of fabric ELS's to fabric addresses. This lack of UNREG_RPI was introduced by a prior node rework patch that performed the UNREG_RPI as part of node cleanup. To resolve the issue, do the following: - Restore the RPI release code, but move the location to so that it is in line with the new node cleanup design. - NPIV ports now release the RPI and drop the node when the caller sets the NLP_RELEASE_RPI flag. - Set the NLP_RELEASE_RPI flag in node cleanup which will trigger a release of RPI to free pool. - Ensure there's an UNREG_RPI at LOGO completion so that RPI release is completed. - Stop offline_prep from skipping nodes that are UNUSED. The RPI may not have been released. - Stop the default RPI handling in lpfc_cmpl_els_rsp() for SLI4. - Fixed up debugfs RPI displays for better debugging. Fixes: a70e63eee1c1 ("scsi: lpfc: Fix NPIV Fabric Node reference counting") Link: https://lore.kernel.org/r/20210514195559.119853-2-jsmart2021@gmail.com Cc: <stable@vger.kernel.org> # v5.11+ Co-developed-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2021-05-14 19:55:49 +00:00
spin_unlock_irq(&ndlp->lock);
}
/* The node has an outstanding reference for the unreg. Now
* that the LOGO action and cleanup are finished, release
* resources.
*/
lpfc_nlp_put(ndlp);
mempool_free(pmb, phba->mbox_mem_pool);
}
/*
* Sets the mailbox completion handler to be used for the
* unreg_rpi command. The handler varies based on the state of
* the port and what will be happening to the rpi next.
*/
static void
lpfc_set_unreg_login_mbx_cmpl(struct lpfc_hba *phba, struct lpfc_vport *vport,
struct lpfc_nodelist *ndlp, LPFC_MBOXQ_t *mbox)
{
unsigned long iflags;
/* Driver always gets a reference on the mailbox job
* in support of async jobs.
*/
mbox->ctx_ndlp = lpfc_nlp_get(ndlp);
if (!mbox->ctx_ndlp)
return;
if (ndlp->nlp_flag & NLP_ISSUE_LOGO) {
mbox->mbox_cmpl = lpfc_nlp_logo_unreg;
} else if (phba->sli_rev == LPFC_SLI_REV4 &&
(!(vport->load_flag & FC_UNLOADING)) &&
(bf_get(lpfc_sli_intf_if_type, &phba->sli4_hba.sli_intf) >=
LPFC_SLI_INTF_IF_TYPE_2) &&
(kref_read(&ndlp->kref) > 0)) {
mbox->mbox_cmpl = lpfc_sli4_unreg_rpi_cmpl_clr;
} else {
if (vport->load_flag & FC_UNLOADING) {
if (phba->sli_rev == LPFC_SLI_REV4) {
spin_lock_irqsave(&ndlp->lock, iflags);
ndlp->nlp_flag |= NLP_RELEASE_RPI;
spin_unlock_irqrestore(&ndlp->lock, iflags);
}
}
mbox->mbox_cmpl = lpfc_sli_def_mbox_cmpl;
}
}
/*
* Free rpi associated with LPFC_NODELIST entry.
* This routine is called from lpfc_freenode(), when we are removing
* a LPFC_NODELIST entry. It is also called if the driver initiates a
* LOGO that completes successfully, and we are waiting to PLOGI back
* to the remote NPort. In addition, it is called after we receive
* and unsolicated ELS cmd, send back a rsp, the rsp completes and
* we are waiting to PLOGI back to the remote NPort.
*/
int
lpfc_unreg_rpi(struct lpfc_vport *vport, struct lpfc_nodelist *ndlp)
{
struct lpfc_hba *phba = vport->phba;
LPFC_MBOXQ_t *mbox;
int rc, acc_plogi = 1;
uint16_t rpi;
if (ndlp->nlp_flag & NLP_RPI_REGISTERED ||
ndlp->nlp_flag & NLP_REG_LOGIN_SEND) {
if (ndlp->nlp_flag & NLP_REG_LOGIN_SEND)
scsi: lpfc: Fix device recovery errors after PLOGI failures When target-side fault injections are made, the driver isn't reconnecting to the remote port. The driver is logging "2753" error messages which state: "PLOGI failure DID:1B2400 Status:x3/xf0240008" The failures status is indicating a Illegal field error, which points to the Temporary RPI field being used for the ELS. This error typically means the driver used an RPI that was already registered (shouldn't be registered if using it in this context). Study has found that if the driver were in discovery attempts and encountered an error, it wouldn't flag the temporary rpi in error. Yet the rpi was released for reallocation in these error paths and another ELS could allocate the rpi. In the failure situation a retry was done on an ELS that had encountered an error, and as the rpi wasn't marked in error, the ELS reused the rpi it originally allocated. But that rpi had been allocated by a different ELS issued after the original error and before the retry attempt. The different ELS had succeeded and the RPI was registered. Fix by marking the rpi state for the node to be in error, aka as needing reallocation, upon an error in the els processing. Error state marking is always done prior to release back to the internal rpi free list, which the driver wasn't doing in cases prior. Also enhanced some of the logging to help in the next case of problem troubleshooting. Link: https://lore.kernel.org/r/20190922035906.10977-7-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2019-09-22 03:58:52 +00:00
lpfc_printf_vlog(vport, KERN_INFO,
LOG_NODE | LOG_DISCOVERY,
"3366 RPI x%x needs to be "
"unregistered nlp_flag x%x "
"did x%x\n",
ndlp->nlp_rpi, ndlp->nlp_flag,
ndlp->nlp_DID);
/* If there is already an UNREG in progress for this ndlp,
* no need to queue up another one.
*/
if (ndlp->nlp_flag & NLP_UNREG_INP) {
scsi: lpfc: Fix device recovery errors after PLOGI failures When target-side fault injections are made, the driver isn't reconnecting to the remote port. The driver is logging "2753" error messages which state: "PLOGI failure DID:1B2400 Status:x3/xf0240008" The failures status is indicating a Illegal field error, which points to the Temporary RPI field being used for the ELS. This error typically means the driver used an RPI that was already registered (shouldn't be registered if using it in this context). Study has found that if the driver were in discovery attempts and encountered an error, it wouldn't flag the temporary rpi in error. Yet the rpi was released for reallocation in these error paths and another ELS could allocate the rpi. In the failure situation a retry was done on an ELS that had encountered an error, and as the rpi wasn't marked in error, the ELS reused the rpi it originally allocated. But that rpi had been allocated by a different ELS issued after the original error and before the retry attempt. The different ELS had succeeded and the RPI was registered. Fix by marking the rpi state for the node to be in error, aka as needing reallocation, upon an error in the els processing. Error state marking is always done prior to release back to the internal rpi free list, which the driver wasn't doing in cases prior. Also enhanced some of the logging to help in the next case of problem troubleshooting. Link: https://lore.kernel.org/r/20190922035906.10977-7-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2019-09-22 03:58:52 +00:00
lpfc_printf_vlog(vport, KERN_INFO,
LOG_NODE | LOG_DISCOVERY,
"1436 unreg_rpi SKIP UNREG x%x on "
"NPort x%x deferred x%x flg x%x "
"Data: x%px\n",
ndlp->nlp_rpi, ndlp->nlp_DID,
ndlp->nlp_defer_did,
ndlp->nlp_flag, ndlp);
goto out;
}
mbox = mempool_alloc(phba->mbox_mem_pool, GFP_KERNEL);
if (mbox) {
/* SLI4 ports require the physical rpi value. */
rpi = ndlp->nlp_rpi;
if (phba->sli_rev == LPFC_SLI_REV4)
rpi = phba->sli4_hba.rpi_ids[ndlp->nlp_rpi];
lpfc_unreg_login(phba, vport->vpi, rpi, mbox);
mbox->vport = vport;
lpfc_set_unreg_login_mbx_cmpl(phba, vport, ndlp, mbox);
if (!mbox->ctx_ndlp) {
mempool_free(mbox, phba->mbox_mem_pool);
return 1;
}
if (mbox->mbox_cmpl == lpfc_sli4_unreg_rpi_cmpl_clr)
/*
* accept PLOGIs after unreg_rpi_cmpl
*/
acc_plogi = 0;
if (((ndlp->nlp_DID & Fabric_DID_MASK) !=
Fabric_DID_MASK) &&
(!(vport->fc_flag & FC_OFFLINE_MODE)))
ndlp->nlp_flag |= NLP_UNREG_INP;
scsi: lpfc: Fix device recovery errors after PLOGI failures When target-side fault injections are made, the driver isn't reconnecting to the remote port. The driver is logging "2753" error messages which state: "PLOGI failure DID:1B2400 Status:x3/xf0240008" The failures status is indicating a Illegal field error, which points to the Temporary RPI field being used for the ELS. This error typically means the driver used an RPI that was already registered (shouldn't be registered if using it in this context). Study has found that if the driver were in discovery attempts and encountered an error, it wouldn't flag the temporary rpi in error. Yet the rpi was released for reallocation in these error paths and another ELS could allocate the rpi. In the failure situation a retry was done on an ELS that had encountered an error, and as the rpi wasn't marked in error, the ELS reused the rpi it originally allocated. But that rpi had been allocated by a different ELS issued after the original error and before the retry attempt. The different ELS had succeeded and the RPI was registered. Fix by marking the rpi state for the node to be in error, aka as needing reallocation, upon an error in the els processing. Error state marking is always done prior to release back to the internal rpi free list, which the driver wasn't doing in cases prior. Also enhanced some of the logging to help in the next case of problem troubleshooting. Link: https://lore.kernel.org/r/20190922035906.10977-7-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2019-09-22 03:58:52 +00:00
lpfc_printf_vlog(vport, KERN_INFO,
LOG_NODE | LOG_DISCOVERY,
"1433 unreg_rpi UNREG x%x on "
"NPort x%x deferred flg x%x "
"Data:x%px\n",
ndlp->nlp_rpi, ndlp->nlp_DID,
ndlp->nlp_flag, ndlp);
rc = lpfc_sli_issue_mbox(phba, mbox, MBX_NOWAIT);
if (rc == MBX_NOT_FINISHED) {
scsi: lpfc: Fix EEH support for NVMe I/O Injecting errors on the PCI slot while the driver is handling NVMe I/O will cause crashes and hangs. There are several rather difficult scenarios occurring. The main issue is that the adapter can report a PCI error before or simultaneously to the PCI subsystem reporting the error. Both paths have different entry points and currently there is no interlock between them. Thus multiple teardown paths are competing and all heck breaks loose. Complicating things is the NVMs path. To a large degree, I/O was able to be shutdown for a full FC port on the SCSI stack. But on NVMe, there isn't a similar call. At best, it works on a per-controller basis, but even at the controller level, it's a controller "reset" call. All of which means I/O is still flowing on different CPUs with reset paths expecting hw access (mailbox commands) to execute properly. The following modifications are made: - A new flag is set in PCI error entrypoints so the driver can track being called by that path. - An interlock is added in the SLI hw error path and the PCI error path such that only one of the paths proceeds with the teardown logic. - RPI cleanup is patched such that RPIs are marked unregistered w/o mbx cmds in cases of hw error. - If entering the SLI port re-init calls, a case where SLI error teardown was quick and beat the PCI calls now reporting error, check whether the SLI port is still live on the PCI bus. - In the PCI reset code to bring the adapter back, recheck the IRQ settings. Different checks for SLI3 vs SLI4. - In I/O completions, that may be called as part of the cleanup or underway just before the hw error, check the state of the adapter. If in error, shortcut handling that would expect further adapter completions as the hw error won't be sending them. - In routines waiting on I/O completions, which may have been in progress prior to the hw error, detect the device is being torn down and abort from their waits and just give up. This points to a larger issue in the driver on ref-counting for data structures, as it doesn't have ref-counting on q and port structures. We'll do this fix for now as it would be a major rework to be done differently. - Fix the NVMe cleanup to simulate NVMe I/O completions if I/O is being failed back due to hw error. - In I/O buf allocation, done at the start of new I/Os, check hw state and fail if hw error. Link: https://lore.kernel.org/r/20210910233159.115896-10-jsmart2021@gmail.com Co-developed-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2021-09-10 23:31:54 +00:00
ndlp->nlp_flag &= ~NLP_UNREG_INP;
mempool_free(mbox, phba->mbox_mem_pool);
acc_plogi = 1;
lpfc_nlp_put(ndlp);
}
scsi: lpfc: Fix scheduling call while in softirq context in lpfc_unreg_rpi The following call trace was seen during HBA reset testing: BUG: scheduling while atomic: swapper/2/0/0x10000100 ... Call Trace: dump_stack+0x19/0x1b __schedule_bug+0x64/0x72 __schedule+0x782/0x840 __cond_resched+0x26/0x30 _cond_resched+0x3a/0x50 mempool_alloc+0xa0/0x170 lpfc_unreg_rpi+0x151/0x630 [lpfc] lpfc_sli_abts_recover_port+0x171/0x190 [lpfc] lpfc_sli4_abts_err_handler+0xb2/0x1f0 [lpfc] lpfc_sli4_io_xri_aborted+0x256/0x300 [lpfc] lpfc_sli4_sp_handle_abort_xri_wcqe.isra.51+0xa3/0x190 [lpfc] lpfc_sli4_fp_handle_cqe+0x89/0x4d0 [lpfc] __lpfc_sli4_process_cq+0xdb/0x2e0 [lpfc] __lpfc_sli4_hba_process_cq+0x41/0x100 [lpfc] lpfc_cq_poll_hdler+0x1a/0x30 [lpfc] irq_poll_softirq+0xc7/0x100 __do_softirq+0xf5/0x280 call_softirq+0x1c/0x30 do_softirq+0x65/0xa0 irq_exit+0x105/0x110 do_IRQ+0x56/0xf0 common_interrupt+0x16a/0x16a With the conversion to blk_io_poll for better interrupt latency in normal cases, it introduced this code path, executed when I/O aborts or logouts are seen, which attempts to allocate memory for a mailbox command to be issued. The allocation is GFP_KERNEL, thus it could attempt to sleep. Fix by creating a work element that performs the event handling for the remote port. This will have the mailbox commands and other items performed in the work element, not the irq. A much better method as the "irq" routine does not stall while performing all this deep handling code. Ensure that allocation failures are handled and send LOGO on failure. Additionally, enlarge the mailbox memory pool to reduce the possibility of additional allocation in this path. Link: https://lore.kernel.org/r/20201020202719.54726-3-james.smart@broadcom.com Fixes: 317aeb83c92b ("scsi: lpfc: Add blk_io_poll support for latency improvment") Cc: <stable@vger.kernel.org> # v5.9+ Co-developed-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <james.smart@broadcom.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-10-20 20:27:12 +00:00
} else {
lpfc_printf_vlog(vport, KERN_INFO,
LOG_NODE | LOG_DISCOVERY,
"1444 Failed to allocate mempool "
"unreg_rpi UNREG x%x, "
"DID x%x, flag x%x, "
"ndlp x%px\n",
ndlp->nlp_rpi, ndlp->nlp_DID,
ndlp->nlp_flag, ndlp);
/* Because mempool_alloc failed, we
* will issue a LOGO here and keep the rpi alive if
* not unloading.
*/
if (!(vport->load_flag & FC_UNLOADING)) {
ndlp->nlp_flag &= ~NLP_UNREG_INP;
lpfc_issue_els_logo(vport, ndlp, 0);
ndlp->nlp_prev_state = ndlp->nlp_state;
lpfc_nlp_set_state(vport, ndlp,
NLP_STE_NPR_NODE);
}
return 1;
}
lpfc_no_rpi(phba, ndlp);
out:
if (phba->sli_rev != LPFC_SLI_REV4)
ndlp->nlp_rpi = 0;
ndlp->nlp_flag &= ~NLP_RPI_REGISTERED;
ndlp->nlp_flag &= ~NLP_NPR_ADISC;
if (acc_plogi)
ndlp->nlp_flag &= ~NLP_LOGO_ACC;
return 1;
}
ndlp->nlp_flag &= ~NLP_LOGO_ACC;
return 0;
}
/**
* lpfc_unreg_hba_rpis - Unregister rpis registered to the hba.
* @phba: pointer to lpfc hba data structure.
*
* This routine is invoked to unregister all the currently registered RPIs
* to the HBA.
**/
void
lpfc_unreg_hba_rpis(struct lpfc_hba *phba)
{
struct lpfc_vport **vports;
struct lpfc_nodelist *ndlp;
struct Scsi_Host *shost;
int i;
vports = lpfc_create_vport_work_array(phba);
if (!vports) {
scsi: lpfc: Add an internal trace log buffer The current logging methods typically end up requesting a reproduction with a different logging level set to figure out what happened. This was mainly by design to not clutter the kernel log messages with things that were typically not interesting and the messages themselves could cause other issues. When looking to make a better system, it was seen that in many cases when more data was wanted was when another message, usually at KERN_ERR level, was logged. And in most cases, what the additional logging that was then enabled was typically. Most of these areas fell into the discovery machine. Based on this summary, the following design has been put in place: The driver will maintain an internal log (256 elements of 256 bytes). The "additional logging" messages that are usually enabled in a reproduction will be changed to now log all the time to the internal log. A new logging level is defined - LOG_TRACE_EVENT. When this level is set (it is not by default) and a message marked as KERN_ERR is logged, all the messages in the internal log will be dumped to the kernel log before the KERN_ERR message is logged. There is a timestamp on each message added to the internal log. However, this timestamp is not converted to wall time when logged. The value of the timestamp is solely to give a crude time reference for the messages. Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-06-30 21:50:00 +00:00
lpfc_printf_log(phba, KERN_ERR, LOG_TRACE_EVENT,
"2884 Vport array allocation failed \n");
return;
}
for (i = 0; i <= phba->max_vports && vports[i] != NULL; i++) {
shost = lpfc_shost_from_vport(vports[i]);
spin_lock_irq(shost->host_lock);
list_for_each_entry(ndlp, &vports[i]->fc_nodes, nlp_listp) {
if (ndlp->nlp_flag & NLP_RPI_REGISTERED) {
/* The mempool_alloc might sleep */
spin_unlock_irq(shost->host_lock);
lpfc_unreg_rpi(vports[i], ndlp);
spin_lock_irq(shost->host_lock);
}
}
spin_unlock_irq(shost->host_lock);
}
lpfc_destroy_vport_work_array(phba, vports);
}
void
lpfc_unreg_all_rpis(struct lpfc_vport *vport)
{
struct lpfc_hba *phba = vport->phba;
LPFC_MBOXQ_t *mbox;
int rc;
if (phba->sli_rev == LPFC_SLI_REV4) {
lpfc_sli4_unreg_all_rpis(vport);
return;
}
mbox = mempool_alloc(phba->mbox_mem_pool, GFP_KERNEL);
if (mbox) {
lpfc_unreg_login(phba, vport->vpi, LPFC_UNREG_ALL_RPIS_VPORT,
mbox);
mbox->vport = vport;
mbox->mbox_cmpl = lpfc_sli_def_mbox_cmpl;
mbox->ctx_ndlp = NULL;
[SCSI] lpfc 8.2.4 : Miscellaneous Fixes Miscellaneous Fixes: - Fix a couple of sparse complaints - Reset the FCP recovery flag when the node is not a FCP2 device. - Speed up offline prep delays - Fixed a memory leak in lpfc_mem_alloc failure path - Fixed external loopback test. - Fixed error code returned from the driver when HBA is over heated. - Correct Max NPIV vport to limits read from adapter - Add missing locks around fc_flag and FC_NEEDS_REG_VPI - Add missing hba ids for device identification - Added support for SET_VARIABLE and MBX_WRITE_WWN mailbox commands - Changed all temperature event messages from warning to error - Fix reporting of link speed when link is down - Added support for MBX_WRITE_WWN mailbox command - Change del_timer_sync() in ISR to del_timer() in interrupt handler - Correct instances of beXX_to_cpu() that should be cpu_to_beXX() - Perform target flush before releasing node references on module unload - Avoid bogus devloss_tmo messages when driver unloads - Fix panic when HBA generates ERATT interupt - Fix mbox race condition and a workaround on back-to-back mailbox commands - Force NPIV off for pt2pt mode between 2 NPorts - Stop worker thread before removing fc_host. - Fix up discovery timeout error case due to missing clear_la - Tighten mailbox polling code to speed up detection of fast completions - Only allow DUMP_MEMORY if adapter offline due to overtemp errors - Added extended error information to the log messages in chip init. Signed-off-by: James Smart <James.Smart@emulex.com> Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
2008-01-11 06:52:54 +00:00
rc = lpfc_sli_issue_mbox_wait(phba, mbox, LPFC_MBOX_TMO);
[SCSI] lpfc 8.3.1: misc fixes/changes 8.3.1 Fixes/Changes : - Fix incorrect byte-swapping on word 4 of IOCB (data length) which caused LUNs to not be discovered on big-endian (e.g. PPC) - Remove a bad cast of MBslimaddr which loses the __iomem (sparse) - Make lpfc_debugfs_mask_disc_trc static (sparse) - Correct misspelled word BlockGuard in lpfc_logmsg.h comment - Replaced repeated code segment for canceling IOCBs from a list with a function call, lpfc_sli_cancel_iocbs(). - Increased HBQ buffers to support 40KB SSC sequences. - Added sysfs interface to update speed and topology parameter without link bounce. - Fixed bug with sysfs fc_host WWNs not being updated after changing the WWNs. - Check if the active mailbox is NULL in the beginning of the mailbox timeout handler - fixes panic in the mailbox timeout handler while running IO stress test - Fixed system panic in lpfc_pci_remove_one() due to ndlp indirect reference to phba through vport - Removed de-reference of scsi device after call to scsi_done() to fix panic in scsi completion path while accessing scsi device after scsi_done is called. - Fixed "Nodelist not empty" message when unloading the driver after target reboot test - Added LP2105 HBA model description - Added code to print all 16 words of unrecognized ASYNC events - Fixed memory leak in vport create + delete loop - Added support for handling dual error bit from HBA - Fixed a driver NULL pointer dereference in lpfc_sli_process_sol_iocb - Fixed a discovery bug with FC switch reboot in lpfc_setup_disc_node - Take NULL termintator into account when calculating available buffer space Signed-off-by: James Smart <james.smart@emulex.com> Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
2009-04-06 22:48:10 +00:00
if (rc != MBX_TIMEOUT)
mempool_free(mbox, phba->mbox_mem_pool);
[SCSI] lpfc 8.3.1: misc fixes/changes 8.3.1 Fixes/Changes : - Fix incorrect byte-swapping on word 4 of IOCB (data length) which caused LUNs to not be discovered on big-endian (e.g. PPC) - Remove a bad cast of MBslimaddr which loses the __iomem (sparse) - Make lpfc_debugfs_mask_disc_trc static (sparse) - Correct misspelled word BlockGuard in lpfc_logmsg.h comment - Replaced repeated code segment for canceling IOCBs from a list with a function call, lpfc_sli_cancel_iocbs(). - Increased HBQ buffers to support 40KB SSC sequences. - Added sysfs interface to update speed and topology parameter without link bounce. - Fixed bug with sysfs fc_host WWNs not being updated after changing the WWNs. - Check if the active mailbox is NULL in the beginning of the mailbox timeout handler - fixes panic in the mailbox timeout handler while running IO stress test - Fixed system panic in lpfc_pci_remove_one() due to ndlp indirect reference to phba through vport - Removed de-reference of scsi device after call to scsi_done() to fix panic in scsi completion path while accessing scsi device after scsi_done is called. - Fixed "Nodelist not empty" message when unloading the driver after target reboot test - Added LP2105 HBA model description - Added code to print all 16 words of unrecognized ASYNC events - Fixed memory leak in vport create + delete loop - Added support for handling dual error bit from HBA - Fixed a driver NULL pointer dereference in lpfc_sli_process_sol_iocb - Fixed a discovery bug with FC switch reboot in lpfc_setup_disc_node - Take NULL termintator into account when calculating available buffer space Signed-off-by: James Smart <james.smart@emulex.com> Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
2009-04-06 22:48:10 +00:00
if ((rc == MBX_TIMEOUT) || (rc == MBX_NOT_FINISHED))
scsi: lpfc: Add an internal trace log buffer The current logging methods typically end up requesting a reproduction with a different logging level set to figure out what happened. This was mainly by design to not clutter the kernel log messages with things that were typically not interesting and the messages themselves could cause other issues. When looking to make a better system, it was seen that in many cases when more data was wanted was when another message, usually at KERN_ERR level, was logged. And in most cases, what the additional logging that was then enabled was typically. Most of these areas fell into the discovery machine. Based on this summary, the following design has been put in place: The driver will maintain an internal log (256 elements of 256 bytes). The "additional logging" messages that are usually enabled in a reproduction will be changed to now log all the time to the internal log. A new logging level is defined - LOG_TRACE_EVENT. When this level is set (it is not by default) and a message marked as KERN_ERR is logged, all the messages in the internal log will be dumped to the kernel log before the KERN_ERR message is logged. There is a timestamp on each message added to the internal log. However, this timestamp is not converted to wall time when logged. The value of the timestamp is solely to give a crude time reference for the messages. Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-06-30 21:50:00 +00:00
lpfc_printf_vlog(vport, KERN_ERR, LOG_TRACE_EVENT,
"1836 Could not issue "
"unreg_login(all_rpis) status %d\n",
rc);
}
}
void
lpfc_unreg_default_rpis(struct lpfc_vport *vport)
{
struct lpfc_hba *phba = vport->phba;
LPFC_MBOXQ_t *mbox;
int rc;
/* Unreg DID is an SLI3 operation. */
if (phba->sli_rev > LPFC_SLI_REV3)
return;
mbox = mempool_alloc(phba->mbox_mem_pool, GFP_KERNEL);
if (mbox) {
lpfc_unreg_did(phba, vport->vpi, LPFC_UNREG_ALL_DFLT_RPIS,
mbox);
mbox->vport = vport;
mbox->mbox_cmpl = lpfc_sli_def_mbox_cmpl;
mbox->ctx_ndlp = NULL;
[SCSI] lpfc 8.2.4 : Miscellaneous Fixes Miscellaneous Fixes: - Fix a couple of sparse complaints - Reset the FCP recovery flag when the node is not a FCP2 device. - Speed up offline prep delays - Fixed a memory leak in lpfc_mem_alloc failure path - Fixed external loopback test. - Fixed error code returned from the driver when HBA is over heated. - Correct Max NPIV vport to limits read from adapter - Add missing locks around fc_flag and FC_NEEDS_REG_VPI - Add missing hba ids for device identification - Added support for SET_VARIABLE and MBX_WRITE_WWN mailbox commands - Changed all temperature event messages from warning to error - Fix reporting of link speed when link is down - Added support for MBX_WRITE_WWN mailbox command - Change del_timer_sync() in ISR to del_timer() in interrupt handler - Correct instances of beXX_to_cpu() that should be cpu_to_beXX() - Perform target flush before releasing node references on module unload - Avoid bogus devloss_tmo messages when driver unloads - Fix panic when HBA generates ERATT interupt - Fix mbox race condition and a workaround on back-to-back mailbox commands - Force NPIV off for pt2pt mode between 2 NPorts - Stop worker thread before removing fc_host. - Fix up discovery timeout error case due to missing clear_la - Tighten mailbox polling code to speed up detection of fast completions - Only allow DUMP_MEMORY if adapter offline due to overtemp errors - Added extended error information to the log messages in chip init. Signed-off-by: James Smart <James.Smart@emulex.com> Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
2008-01-11 06:52:54 +00:00
rc = lpfc_sli_issue_mbox_wait(phba, mbox, LPFC_MBOX_TMO);
[SCSI] lpfc 8.3.1: misc fixes/changes 8.3.1 Fixes/Changes : - Fix incorrect byte-swapping on word 4 of IOCB (data length) which caused LUNs to not be discovered on big-endian (e.g. PPC) - Remove a bad cast of MBslimaddr which loses the __iomem (sparse) - Make lpfc_debugfs_mask_disc_trc static (sparse) - Correct misspelled word BlockGuard in lpfc_logmsg.h comment - Replaced repeated code segment for canceling IOCBs from a list with a function call, lpfc_sli_cancel_iocbs(). - Increased HBQ buffers to support 40KB SSC sequences. - Added sysfs interface to update speed and topology parameter without link bounce. - Fixed bug with sysfs fc_host WWNs not being updated after changing the WWNs. - Check if the active mailbox is NULL in the beginning of the mailbox timeout handler - fixes panic in the mailbox timeout handler while running IO stress test - Fixed system panic in lpfc_pci_remove_one() due to ndlp indirect reference to phba through vport - Removed de-reference of scsi device after call to scsi_done() to fix panic in scsi completion path while accessing scsi device after scsi_done is called. - Fixed "Nodelist not empty" message when unloading the driver after target reboot test - Added LP2105 HBA model description - Added code to print all 16 words of unrecognized ASYNC events - Fixed memory leak in vport create + delete loop - Added support for handling dual error bit from HBA - Fixed a driver NULL pointer dereference in lpfc_sli_process_sol_iocb - Fixed a discovery bug with FC switch reboot in lpfc_setup_disc_node - Take NULL termintator into account when calculating available buffer space Signed-off-by: James Smart <james.smart@emulex.com> Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
2009-04-06 22:48:10 +00:00
if (rc != MBX_TIMEOUT)
mempool_free(mbox, phba->mbox_mem_pool);
if ((rc == MBX_TIMEOUT) || (rc == MBX_NOT_FINISHED))
scsi: lpfc: Add an internal trace log buffer The current logging methods typically end up requesting a reproduction with a different logging level set to figure out what happened. This was mainly by design to not clutter the kernel log messages with things that were typically not interesting and the messages themselves could cause other issues. When looking to make a better system, it was seen that in many cases when more data was wanted was when another message, usually at KERN_ERR level, was logged. And in most cases, what the additional logging that was then enabled was typically. Most of these areas fell into the discovery machine. Based on this summary, the following design has been put in place: The driver will maintain an internal log (256 elements of 256 bytes). The "additional logging" messages that are usually enabled in a reproduction will be changed to now log all the time to the internal log. A new logging level is defined - LOG_TRACE_EVENT. When this level is set (it is not by default) and a message marked as KERN_ERR is logged, all the messages in the internal log will be dumped to the kernel log before the KERN_ERR message is logged. There is a timestamp on each message added to the internal log. However, this timestamp is not converted to wall time when logged. The value of the timestamp is solely to give a crude time reference for the messages. Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-06-30 21:50:00 +00:00
lpfc_printf_vlog(vport, KERN_ERR, LOG_TRACE_EVENT,
"1815 Could not issue "
[SCSI] lpfc 8.3.1: misc fixes/changes 8.3.1 Fixes/Changes : - Fix incorrect byte-swapping on word 4 of IOCB (data length) which caused LUNs to not be discovered on big-endian (e.g. PPC) - Remove a bad cast of MBslimaddr which loses the __iomem (sparse) - Make lpfc_debugfs_mask_disc_trc static (sparse) - Correct misspelled word BlockGuard in lpfc_logmsg.h comment - Replaced repeated code segment for canceling IOCBs from a list with a function call, lpfc_sli_cancel_iocbs(). - Increased HBQ buffers to support 40KB SSC sequences. - Added sysfs interface to update speed and topology parameter without link bounce. - Fixed bug with sysfs fc_host WWNs not being updated after changing the WWNs. - Check if the active mailbox is NULL in the beginning of the mailbox timeout handler - fixes panic in the mailbox timeout handler while running IO stress test - Fixed system panic in lpfc_pci_remove_one() due to ndlp indirect reference to phba through vport - Removed de-reference of scsi device after call to scsi_done() to fix panic in scsi completion path while accessing scsi device after scsi_done is called. - Fixed "Nodelist not empty" message when unloading the driver after target reboot test - Added LP2105 HBA model description - Added code to print all 16 words of unrecognized ASYNC events - Fixed memory leak in vport create + delete loop - Added support for handling dual error bit from HBA - Fixed a driver NULL pointer dereference in lpfc_sli_process_sol_iocb - Fixed a discovery bug with FC switch reboot in lpfc_setup_disc_node - Take NULL termintator into account when calculating available buffer space Signed-off-by: James Smart <james.smart@emulex.com> Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
2009-04-06 22:48:10 +00:00
"unreg_did (default rpis) status %d\n",
rc);
}
}
/*
* Free resources associated with LPFC_NODELIST entry
* so it can be freed.
*/
static int
lpfc_cleanup_node(struct lpfc_vport *vport, struct lpfc_nodelist *ndlp)
{
struct lpfc_hba *phba = vport->phba;
LPFC_MBOXQ_t *mb, *nextmb;
/* Cleanup node for NPort <nlp_DID> */
lpfc_printf_vlog(vport, KERN_INFO, LOG_NODE,
"0900 Cleanup node for NPort x%x "
"Data: x%x x%x x%x\n",
ndlp->nlp_DID, ndlp->nlp_flag,
ndlp->nlp_state, ndlp->nlp_rpi);
scsi: lpfc: Rework remote port ref counting and node freeing When a remote port is disconnected and disappears, its node structure (ndlp) stays allocated and on a vport node list. While on the list it can be matched, thus requires validation checks on state to be added in numerous code paths. If the node comes back, its possible for there to be multiple node structures for the same device on the vport node list. There is no reason to keep the node structure around after it is no longer in existence, and the current implementation creates problems for itself (multiple nodes) and lots of unnecessary code for state validation. Additionally, the reference taking on the node structure didn't follow the normal model used by the kernel kref api. It included lots of odd logic to match state with reference count. The combination of this odd logic plus the way it was implicitly used in the discovery engine made its reference taking implementation suspect and extremely hard to follow. Change the driver such that the reference taking routines are now normal ref increments/decrements and callout on refcount=0. With this in place, the rework can be done such that the node structure is fully removed and deallocated when the remote port no longer exists and all references are removed. This removal logic, and the basic ref counting are intrically tied, thus in a single patch. Link: https://lore.kernel.org/r/20201115192646.12977-2-james.smart@broadcom.com Co-developed-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <james.smart@broadcom.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-11-15 19:26:30 +00:00
lpfc_dequeue_node(vport, ndlp);
/* Don't need to clean up REG_LOGIN64 cmds for Default RPI cleanup */
/* cleanup any ndlp on mbox q waiting for reglogin cmpl */
if ((mb = phba->sli.mbox_active)) {
if ((mb->u.mb.mbxCommand == MBX_REG_LOGIN64) &&
!(mb->mbox_flag & LPFC_MBX_IMED_UNREG) &&
(ndlp == (struct lpfc_nodelist *)mb->ctx_ndlp)) {
mb->ctx_ndlp = NULL;
mb->mbox_cmpl = lpfc_sli_def_mbox_cmpl;
}
}
spin_lock_irq(&phba->hbalock);
/* Cleanup REG_LOGIN completions which are not yet processed */
list_for_each_entry(mb, &phba->sli.mboxq_cmpl, list) {
if ((mb->u.mb.mbxCommand != MBX_REG_LOGIN64) ||
(mb->mbox_flag & LPFC_MBX_IMED_UNREG) ||
(ndlp != (struct lpfc_nodelist *)mb->ctx_ndlp))
continue;
mb->ctx_ndlp = NULL;
mb->mbox_cmpl = lpfc_sli_def_mbox_cmpl;
}
list_for_each_entry_safe(mb, nextmb, &phba->sli.mboxq, list) {
if ((mb->u.mb.mbxCommand == MBX_REG_LOGIN64) &&
!(mb->mbox_flag & LPFC_MBX_IMED_UNREG) &&
(ndlp == (struct lpfc_nodelist *)mb->ctx_ndlp)) {
list_del(&mb->list);
lpfc_mbox_rsrc_cleanup(phba, mb, MBOX_THD_LOCKED);
/* Don't invoke lpfc_nlp_put. The driver is in
* lpfc_nlp_release context.
*/
}
}
spin_unlock_irq(&phba->hbalock);
lpfc_els_abort(phba, ndlp);
spin_lock_irq(&ndlp->lock);
ndlp->nlp_flag &= ~NLP_DELAY_TMO;
spin_unlock_irq(&ndlp->lock);
ndlp->nlp_last_elscmd = 0;
del_timer_sync(&ndlp->nlp_delayfunc);
list_del_init(&ndlp->els_retry_evt.evt_listp);
list_del_init(&ndlp->dev_loss_evt.evt_listp);
scsi: lpfc: Fix scheduling call while in softirq context in lpfc_unreg_rpi The following call trace was seen during HBA reset testing: BUG: scheduling while atomic: swapper/2/0/0x10000100 ... Call Trace: dump_stack+0x19/0x1b __schedule_bug+0x64/0x72 __schedule+0x782/0x840 __cond_resched+0x26/0x30 _cond_resched+0x3a/0x50 mempool_alloc+0xa0/0x170 lpfc_unreg_rpi+0x151/0x630 [lpfc] lpfc_sli_abts_recover_port+0x171/0x190 [lpfc] lpfc_sli4_abts_err_handler+0xb2/0x1f0 [lpfc] lpfc_sli4_io_xri_aborted+0x256/0x300 [lpfc] lpfc_sli4_sp_handle_abort_xri_wcqe.isra.51+0xa3/0x190 [lpfc] lpfc_sli4_fp_handle_cqe+0x89/0x4d0 [lpfc] __lpfc_sli4_process_cq+0xdb/0x2e0 [lpfc] __lpfc_sli4_hba_process_cq+0x41/0x100 [lpfc] lpfc_cq_poll_hdler+0x1a/0x30 [lpfc] irq_poll_softirq+0xc7/0x100 __do_softirq+0xf5/0x280 call_softirq+0x1c/0x30 do_softirq+0x65/0xa0 irq_exit+0x105/0x110 do_IRQ+0x56/0xf0 common_interrupt+0x16a/0x16a With the conversion to blk_io_poll for better interrupt latency in normal cases, it introduced this code path, executed when I/O aborts or logouts are seen, which attempts to allocate memory for a mailbox command to be issued. The allocation is GFP_KERNEL, thus it could attempt to sleep. Fix by creating a work element that performs the event handling for the remote port. This will have the mailbox commands and other items performed in the work element, not the irq. A much better method as the "irq" routine does not stall while performing all this deep handling code. Ensure that allocation failures are handled and send LOGO on failure. Additionally, enlarge the mailbox memory pool to reduce the possibility of additional allocation in this path. Link: https://lore.kernel.org/r/20201020202719.54726-3-james.smart@broadcom.com Fixes: 317aeb83c92b ("scsi: lpfc: Add blk_io_poll support for latency improvment") Cc: <stable@vger.kernel.org> # v5.9+ Co-developed-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <james.smart@broadcom.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-10-20 20:27:12 +00:00
list_del_init(&ndlp->recovery_evt.evt_listp);
lpfc_cleanup_vports_rrqs(vport, ndlp);
scsi: lpfc: Fix unreleased RPIs when NPIV ports are created While testing NPIV and watching logins and used RPI levels, it was seen the used RPI count was much higher than the number of remote ports discovered. Code inspection showed that remote port removals on any NPIV instance are releasing the RPI, but not performing an UNREG_RPI with the adapter thus the reference counting never fully drops and the RPI is never fully released. This was happening on NPIV nodes due to a log of fabric ELS's to fabric addresses. This lack of UNREG_RPI was introduced by a prior node rework patch that performed the UNREG_RPI as part of node cleanup. To resolve the issue, do the following: - Restore the RPI release code, but move the location to so that it is in line with the new node cleanup design. - NPIV ports now release the RPI and drop the node when the caller sets the NLP_RELEASE_RPI flag. - Set the NLP_RELEASE_RPI flag in node cleanup which will trigger a release of RPI to free pool. - Ensure there's an UNREG_RPI at LOGO completion so that RPI release is completed. - Stop offline_prep from skipping nodes that are UNUSED. The RPI may not have been released. - Stop the default RPI handling in lpfc_cmpl_els_rsp() for SLI4. - Fixed up debugfs RPI displays for better debugging. Fixes: a70e63eee1c1 ("scsi: lpfc: Fix NPIV Fabric Node reference counting") Link: https://lore.kernel.org/r/20210514195559.119853-2-jsmart2021@gmail.com Cc: <stable@vger.kernel.org> # v5.11+ Co-developed-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2021-05-14 19:55:49 +00:00
if (phba->sli_rev == LPFC_SLI_REV4)
ndlp->nlp_flag |= NLP_RELEASE_RPI;
scsi: lpfc: Fix unreleased RPIs when NPIV ports are created While testing NPIV and watching logins and used RPI levels, it was seen the used RPI count was much higher than the number of remote ports discovered. Code inspection showed that remote port removals on any NPIV instance are releasing the RPI, but not performing an UNREG_RPI with the adapter thus the reference counting never fully drops and the RPI is never fully released. This was happening on NPIV nodes due to a log of fabric ELS's to fabric addresses. This lack of UNREG_RPI was introduced by a prior node rework patch that performed the UNREG_RPI as part of node cleanup. To resolve the issue, do the following: - Restore the RPI release code, but move the location to so that it is in line with the new node cleanup design. - NPIV ports now release the RPI and drop the node when the caller sets the NLP_RELEASE_RPI flag. - Set the NLP_RELEASE_RPI flag in node cleanup which will trigger a release of RPI to free pool. - Ensure there's an UNREG_RPI at LOGO completion so that RPI release is completed. - Stop offline_prep from skipping nodes that are UNUSED. The RPI may not have been released. - Stop the default RPI handling in lpfc_cmpl_els_rsp() for SLI4. - Fixed up debugfs RPI displays for better debugging. Fixes: a70e63eee1c1 ("scsi: lpfc: Fix NPIV Fabric Node reference counting") Link: https://lore.kernel.org/r/20210514195559.119853-2-jsmart2021@gmail.com Cc: <stable@vger.kernel.org> # v5.11+ Co-developed-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2021-05-14 19:55:49 +00:00
return 0;
}
static int
lpfc_matchdid(struct lpfc_vport *vport, struct lpfc_nodelist *ndlp,
uint32_t did)
{
D_ID mydid, ndlpdid, matchdid;
if (did == Bcast_DID)
return 0;
/* First check for Direct match */
if (ndlp->nlp_DID == did)
return 1;
/* Next check for area/domain identically equals 0 match */
mydid.un.word = vport->fc_myDID;
if ((mydid.un.b.domain == 0) && (mydid.un.b.area == 0)) {
return 0;
}
matchdid.un.word = did;
ndlpdid.un.word = ndlp->nlp_DID;
if (matchdid.un.b.id == ndlpdid.un.b.id) {
if ((mydid.un.b.domain == matchdid.un.b.domain) &&
(mydid.un.b.area == matchdid.un.b.area)) {
/* This code is supposed to match the ID
* for a private loop device that is
* connect to fl_port. But we need to
* check that the port did not just go
* from pt2pt to fabric or we could end
* up matching ndlp->nlp_DID 000001 to
* fabric DID 0x20101
*/
if ((ndlpdid.un.b.domain == 0) &&
(ndlpdid.un.b.area == 0)) {
if (ndlpdid.un.b.id &&
vport->phba->fc_topology ==
LPFC_TOPOLOGY_LOOP)
return 1;
}
return 0;
}
matchdid.un.word = ndlp->nlp_DID;
if ((mydid.un.b.domain == ndlpdid.un.b.domain) &&
(mydid.un.b.area == ndlpdid.un.b.area)) {
if ((matchdid.un.b.domain == 0) &&
(matchdid.un.b.area == 0)) {
if (matchdid.un.b.id)
return 1;
}
}
}
return 0;
}
/* Search for a nodelist entry */
static struct lpfc_nodelist *
__lpfc_findnode_did(struct lpfc_vport *vport, uint32_t did)
{
struct lpfc_nodelist *ndlp;
uint32_t data1;
list_for_each_entry(ndlp, &vport->fc_nodes, nlp_listp) {
if (lpfc_matchdid(vport, ndlp, did)) {
scsi: lpfc: Fix loss of remote port after devloss due to lack of RPIs In tests with remote ports contantly logging out/logging coupled with occassional local link bounce, if a remote port is disocnnected for longer than devloss_tmo and then subsequently reconnected, eventually the test will fail to login with the remote port and remote port connectivity is lost. When devloss_tmo expires, the driver does not free the node struct until the port or npiv instances is being deleted. The node is left allocated but the state set to UNUSED. If the node was in the process of logging in when the local link drop occurred, meaning the RPI was allocated for the node in order to send the ELS, but not yet registered which comes after successful login, the node is moved to the NPR state, and if devloss expires, to UNUSED state. If the remote port comes back, the node associated with it is restarted and this path happens to allocate a new RPI and overwrites the prior RPI value. In the cases where the port was logged in and loggs out, the path did release the RPI but did not set the node rpi value. In the cases where the remote port never finished logging in, the path never did the call to release the rpi. In this latter case, when the node is subsequently restore, the new rpi allocation overwrites the rpi that was not released, and the rpi is now leaked. Eventually the port will run out of RPI resources to log into new remote ports. Fix by following changes: - When an rpi is released, do so under locks and ensure the node rpi value is set to a non-allocated value (LPFC_RPI_ALLOC_ERROR). Note: refactored to a small service routine to avoid indentation issues. - When re-enabling a node, check the rpi value to determine if a new allocation is necessary. If already set, use the prior rpi. Enhanced logging to help in the future. Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2019-08-14 23:56:47 +00:00
data1 = (((uint32_t)ndlp->nlp_state << 24) |
((uint32_t)ndlp->nlp_xri << 16) |
scsi: lpfc: Rework remote port ref counting and node freeing When a remote port is disconnected and disappears, its node structure (ndlp) stays allocated and on a vport node list. While on the list it can be matched, thus requires validation checks on state to be added in numerous code paths. If the node comes back, its possible for there to be multiple node structures for the same device on the vport node list. There is no reason to keep the node structure around after it is no longer in existence, and the current implementation creates problems for itself (multiple nodes) and lots of unnecessary code for state validation. Additionally, the reference taking on the node structure didn't follow the normal model used by the kernel kref api. It included lots of odd logic to match state with reference count. The combination of this odd logic plus the way it was implicitly used in the discovery engine made its reference taking implementation suspect and extremely hard to follow. Change the driver such that the reference taking routines are now normal ref increments/decrements and callout on refcount=0. With this in place, the rework can be done such that the node structure is fully removed and deallocated when the remote port no longer exists and all references are removed. This removal logic, and the basic ref counting are intrically tied, thus in a single patch. Link: https://lore.kernel.org/r/20201115192646.12977-2-james.smart@broadcom.com Co-developed-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <james.smart@broadcom.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-11-15 19:26:30 +00:00
((uint32_t)ndlp->nlp_type << 8)
);
lpfc_printf_vlog(vport, KERN_INFO, LOG_NODE,
"0929 FIND node DID "
scsi: lpfc: Fix loss of remote port after devloss due to lack of RPIs In tests with remote ports contantly logging out/logging coupled with occassional local link bounce, if a remote port is disocnnected for longer than devloss_tmo and then subsequently reconnected, eventually the test will fail to login with the remote port and remote port connectivity is lost. When devloss_tmo expires, the driver does not free the node struct until the port or npiv instances is being deleted. The node is left allocated but the state set to UNUSED. If the node was in the process of logging in when the local link drop occurred, meaning the RPI was allocated for the node in order to send the ELS, but not yet registered which comes after successful login, the node is moved to the NPR state, and if devloss expires, to UNUSED state. If the remote port comes back, the node associated with it is restarted and this path happens to allocate a new RPI and overwrites the prior RPI value. In the cases where the port was logged in and loggs out, the path did release the RPI but did not set the node rpi value. In the cases where the remote port never finished logging in, the path never did the call to release the rpi. In this latter case, when the node is subsequently restore, the new rpi allocation overwrites the rpi that was not released, and the rpi is now leaked. Eventually the port will run out of RPI resources to log into new remote ports. Fix by following changes: - When an rpi is released, do so under locks and ensure the node rpi value is set to a non-allocated value (LPFC_RPI_ALLOC_ERROR). Note: refactored to a small service routine to avoid indentation issues. - When re-enabling a node, check the rpi value to determine if a new allocation is necessary. If already set, use the prior rpi. Enhanced logging to help in the future. Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2019-08-14 23:56:47 +00:00
"Data: x%px x%x x%x x%x x%x x%px\n",
ndlp, ndlp->nlp_DID,
scsi: lpfc: Fix loss of remote port after devloss due to lack of RPIs In tests with remote ports contantly logging out/logging coupled with occassional local link bounce, if a remote port is disocnnected for longer than devloss_tmo and then subsequently reconnected, eventually the test will fail to login with the remote port and remote port connectivity is lost. When devloss_tmo expires, the driver does not free the node struct until the port or npiv instances is being deleted. The node is left allocated but the state set to UNUSED. If the node was in the process of logging in when the local link drop occurred, meaning the RPI was allocated for the node in order to send the ELS, but not yet registered which comes after successful login, the node is moved to the NPR state, and if devloss expires, to UNUSED state. If the remote port comes back, the node associated with it is restarted and this path happens to allocate a new RPI and overwrites the prior RPI value. In the cases where the port was logged in and loggs out, the path did release the RPI but did not set the node rpi value. In the cases where the remote port never finished logging in, the path never did the call to release the rpi. In this latter case, when the node is subsequently restore, the new rpi allocation overwrites the rpi that was not released, and the rpi is now leaked. Eventually the port will run out of RPI resources to log into new remote ports. Fix by following changes: - When an rpi is released, do so under locks and ensure the node rpi value is set to a non-allocated value (LPFC_RPI_ALLOC_ERROR). Note: refactored to a small service routine to avoid indentation issues. - When re-enabling a node, check the rpi value to determine if a new allocation is necessary. If already set, use the prior rpi. Enhanced logging to help in the future. Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2019-08-14 23:56:47 +00:00
ndlp->nlp_flag, data1, ndlp->nlp_rpi,
ndlp->active_rrqs_xri_bitmap);
return ndlp;
}
}
/* FIND node did <did> NOT FOUND */
lpfc_printf_vlog(vport, KERN_INFO, LOG_NODE,
"0932 FIND node did x%x NOT FOUND.\n", did);
return NULL;
}
struct lpfc_nodelist *
lpfc_findnode_did(struct lpfc_vport *vport, uint32_t did)
{
struct Scsi_Host *shost = lpfc_shost_from_vport(vport);
struct lpfc_nodelist *ndlp;
unsigned long iflags;
spin_lock_irqsave(shost->host_lock, iflags);
ndlp = __lpfc_findnode_did(vport, did);
spin_unlock_irqrestore(shost->host_lock, iflags);
return ndlp;
}
struct lpfc_nodelist *
lpfc_findnode_mapped(struct lpfc_vport *vport)
{
struct Scsi_Host *shost = lpfc_shost_from_vport(vport);
struct lpfc_nodelist *ndlp;
uint32_t data1;
unsigned long iflags;
spin_lock_irqsave(shost->host_lock, iflags);
list_for_each_entry(ndlp, &vport->fc_nodes, nlp_listp) {
if (ndlp->nlp_state == NLP_STE_UNMAPPED_NODE ||
ndlp->nlp_state == NLP_STE_MAPPED_NODE) {
data1 = (((uint32_t)ndlp->nlp_state << 24) |
((uint32_t)ndlp->nlp_xri << 16) |
((uint32_t)ndlp->nlp_type << 8) |
((uint32_t)ndlp->nlp_rpi & 0xff));
spin_unlock_irqrestore(shost->host_lock, iflags);
lpfc_printf_vlog(vport, KERN_INFO, LOG_NODE,
"2025 FIND node DID "
"Data: x%px x%x x%x x%x x%px\n",
ndlp, ndlp->nlp_DID,
ndlp->nlp_flag, data1,
ndlp->active_rrqs_xri_bitmap);
return ndlp;
}
}
spin_unlock_irqrestore(shost->host_lock, iflags);
/* FIND node did <did> NOT FOUND */
lpfc_printf_vlog(vport, KERN_INFO, LOG_NODE,
"2026 FIND mapped did NOT FOUND.\n");
return NULL;
}
struct lpfc_nodelist *
lpfc_setup_disc_node(struct lpfc_vport *vport, uint32_t did)
{
struct lpfc_nodelist *ndlp;
ndlp = lpfc_findnode_did(vport, did);
if (!ndlp) {
if (vport->phba->nvmet_support)
return NULL;
if ((vport->fc_flag & FC_RSCN_MODE) != 0 &&
lpfc_rscn_payload_check(vport, did) == 0)
return NULL;
ndlp = lpfc_nlp_init(vport, did);
if (!ndlp)
return NULL;
lpfc_nlp_set_state(vport, ndlp, NLP_STE_NPR_NODE);
lpfc_printf_vlog(vport, KERN_INFO, LOG_DISCOVERY,
"6453 Setup New Node 2B_DISC x%x "
"Data:x%x x%x x%x\n",
ndlp->nlp_DID, ndlp->nlp_flag,
ndlp->nlp_state, vport->fc_flag);
spin_lock_irq(&ndlp->lock);
ndlp->nlp_flag |= NLP_NPR_2B_DISC;
spin_unlock_irq(&ndlp->lock);
return ndlp;
}
/* The NVME Target does not want to actively manage an rport.
* The goal is to allow the target to reset its state and clear
* pending IO in preparation for the initiator to recover.
*/
[SCSI] lpfc 8.2.6 : Multiple discovery fixes Multiple Discovery Fixes: - Fix race on discovery due to link events coinciding with vport_delete. - Use NLP_FABRIC state to filter out switch-based pseudo initiators that reuse the same WWNs. - Correct erroneous setting of DID=0 in lpfc_matchdid() - Correct extra reference count that was in the lookup path for the remoteid from an unsolicited ELS. - Correct double-free bug in els abort path. - Correct FDMI server discovery logic for switch that return a WWN of 0. - Fix bugs in ndlp mgmt when a node changes address - Correct bug that did not delete RSCNs for vports upon link transitions - Fix "0216 Link event during NS query" error which pops up when vports are swapped to different switch ports. - Add sanity checks on ndlp structures - Fix devloss log message to dump WWN correctly - Hold off mgmt commands that were interferring with discovery mailbox cmds - Remove unnecessary FC_ESTABLISH_LINK logic. - Correct some race conditions in the worker thread, resulting in devloss: - Clear the work_port_events field before handling the work port events - Clear the deferred ring event before handling a deferred ring event - Hold the hba lock when waking up the work thread - Send an acc for the rscn even when we aren't going to handle it - Fix locking behavior that was not properly protecting the ACTIVE flag, thus allowing mailbox command order to shift. Signed-off-by: James Smart <james.smart@emulex.com> Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
2008-04-07 14:15:56 +00:00
if ((vport->fc_flag & FC_RSCN_MODE) &&
!(vport->fc_flag & FC_NDISC_ACTIVE)) {
if (lpfc_rscn_payload_check(vport, did)) {
/* Since this node is marked for discovery,
* delay timeout is not needed.
*/
lpfc_cancel_retry_delay_tmo(vport, ndlp);
lpfc_printf_vlog(vport, KERN_INFO, LOG_DISCOVERY,
"6455 Setup RSCN Node 2B_DISC x%x "
"Data:x%x x%x x%x\n",
ndlp->nlp_DID, ndlp->nlp_flag,
ndlp->nlp_state, vport->fc_flag);
/* NVME Target mode waits until rport is known to be
* impacted by the RSCN before it transitions. No
* active management - just go to NPR provided the
* node had a valid login.
*/
if (vport->phba->nvmet_support)
return ndlp;
/* If we've already received a PLOGI from this NPort
* we don't need to try to discover it again.
*/
if (ndlp->nlp_flag & NLP_RCV_PLOGI &&
!(ndlp->nlp_type &
(NLP_FCP_TARGET | NLP_NVME_TARGET)))
return NULL;
ndlp->nlp_prev_state = ndlp->nlp_state;
lpfc_nlp_set_state(vport, ndlp, NLP_STE_NPR_NODE);
spin_lock_irq(&ndlp->lock);
[SCSI] lpfc 8.3.1: misc fixes/changes 8.3.1 Fixes/Changes : - Fix incorrect byte-swapping on word 4 of IOCB (data length) which caused LUNs to not be discovered on big-endian (e.g. PPC) - Remove a bad cast of MBslimaddr which loses the __iomem (sparse) - Make lpfc_debugfs_mask_disc_trc static (sparse) - Correct misspelled word BlockGuard in lpfc_logmsg.h comment - Replaced repeated code segment for canceling IOCBs from a list with a function call, lpfc_sli_cancel_iocbs(). - Increased HBQ buffers to support 40KB SSC sequences. - Added sysfs interface to update speed and topology parameter without link bounce. - Fixed bug with sysfs fc_host WWNs not being updated after changing the WWNs. - Check if the active mailbox is NULL in the beginning of the mailbox timeout handler - fixes panic in the mailbox timeout handler while running IO stress test - Fixed system panic in lpfc_pci_remove_one() due to ndlp indirect reference to phba through vport - Removed de-reference of scsi device after call to scsi_done() to fix panic in scsi completion path while accessing scsi device after scsi_done is called. - Fixed "Nodelist not empty" message when unloading the driver after target reboot test - Added LP2105 HBA model description - Added code to print all 16 words of unrecognized ASYNC events - Fixed memory leak in vport create + delete loop - Added support for handling dual error bit from HBA - Fixed a driver NULL pointer dereference in lpfc_sli_process_sol_iocb - Fixed a discovery bug with FC switch reboot in lpfc_setup_disc_node - Take NULL termintator into account when calculating available buffer space Signed-off-by: James Smart <james.smart@emulex.com> Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
2009-04-06 22:48:10 +00:00
ndlp->nlp_flag |= NLP_NPR_2B_DISC;
spin_unlock_irq(&ndlp->lock);
} else {
lpfc_printf_vlog(vport, KERN_INFO, LOG_DISCOVERY,
"6456 Skip Setup RSCN Node x%x "
"Data:x%x x%x x%x\n",
ndlp->nlp_DID, ndlp->nlp_flag,
ndlp->nlp_state, vport->fc_flag);
ndlp = NULL;
}
} else {
lpfc_printf_vlog(vport, KERN_INFO, LOG_DISCOVERY,
"6457 Setup Active Node 2B_DISC x%x "
"Data:x%x x%x x%x\n",
ndlp->nlp_DID, ndlp->nlp_flag,
ndlp->nlp_state, vport->fc_flag);
/* If the initiator received a PLOGI from this NPort or if the
* initiator is already in the process of discovery on it,
* there's no need to try to discover it again.
*/
if (ndlp->nlp_state == NLP_STE_ADISC_ISSUE ||
ndlp->nlp_state == NLP_STE_PLOGI_ISSUE ||
(!vport->phba->nvmet_support &&
ndlp->nlp_flag & NLP_RCV_PLOGI))
return NULL;
if (vport->phba->nvmet_support)
return ndlp;
/* Moving to NPR state clears unsolicited flags and
* allows for rediscovery
*/
lpfc_nlp_set_state(vport, ndlp, NLP_STE_NPR_NODE);
spin_lock_irq(&ndlp->lock);
ndlp->nlp_flag |= NLP_NPR_2B_DISC;
spin_unlock_irq(&ndlp->lock);
}
return ndlp;
}
/* Build a list of nodes to discover based on the loopmap */
void
lpfc_disc_list_loopmap(struct lpfc_vport *vport)
{
struct lpfc_hba *phba = vport->phba;
int j;
uint32_t alpa, index;
if (!lpfc_is_link_up(phba))
return;
2010-11-21 04:11:48 +00:00
if (phba->fc_topology != LPFC_TOPOLOGY_LOOP)
return;
/* Check for loop map present or not */
if (phba->alpa_map[0]) {
for (j = 1; j <= phba->alpa_map[0]; j++) {
alpa = phba->alpa_map[j];
if (((vport->fc_myDID & 0xff) == alpa) || (alpa == 0))
continue;
lpfc_setup_disc_node(vport, alpa);
}
} else {
/* No alpamap, so try all alpa's */
for (j = 0; j < FC_MAXLOOP; j++) {
/* If cfg_scan_down is set, start from highest
* ALPA (0xef) to lowest (0x1).
*/
if (vport->cfg_scan_down)
index = j;
else
index = FC_MAXLOOP - j - 1;
alpa = lpfcAlpaArray[index];
if ((vport->fc_myDID & 0xff) == alpa)
continue;
lpfc_setup_disc_node(vport, alpa);
}
}
return;
}
scsi: lpfc: NVME Initiator: Base modifications NVME Initiator: Base modifications This patch adds base modifications for NVME initiator support. The base modifications consist of: - Formal split of SLI3 rings from SLI-4 WQs (sometimes referred to as rings as well) as implementation now widely varies between the two. - Addition of configuration modes: SCSI initiator only; NVME initiator only; NVME target only; and SCSI and NVME initiator. The configuration mode drives overall adapter configuration, offloads enabled, and resource splits. NVME support is only available on SLI-4 devices and newer fw. - Implements the following based on configuration mode: - Exchange resources are split by protocol; Obviously, if only 1 mode, then no split occurs. Default is 50/50. module attribute allows tuning. - Pools and config parameters are separated per-protocol - Each protocol has it's own set of queues, but share interrupt vectors. SCSI: SLI3 devices have few queues and the original style of queue allocation remains. SLI4 devices piggy back on an "io-channel" concept that eventually needs to merge with scsi-mq/blk-mq support (it is underway). For now, the paradigm continues as it existed prior. io channel allocates N msix and N WQs (N=4 default) and either round robins or uses cpu # modulo N for scheduling. A bunch of module parameters allow the configuration to be tuned. NVME (initiator): Allocates an msix per cpu (or whatever pci_alloc_irq_vectors gets) Allocates a WQ per cpu, and maps the WQs to msix on a WQ # modulo msix vector count basis. Module parameters exist to cap/control the config if desired. - Each protocol has its own buffer and dma pools. I apologize for the size of the patch. Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <james.smart@broadcom.com> ---- Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2017-02-12 21:52:30 +00:00
/* SLI3 only */
void
lpfc_issue_clear_la(struct lpfc_hba *phba, struct lpfc_vport *vport)
{
LPFC_MBOXQ_t *mbox;
struct lpfc_sli *psli = &phba->sli;
scsi: lpfc: NVME Initiator: Base modifications NVME Initiator: Base modifications This patch adds base modifications for NVME initiator support. The base modifications consist of: - Formal split of SLI3 rings from SLI-4 WQs (sometimes referred to as rings as well) as implementation now widely varies between the two. - Addition of configuration modes: SCSI initiator only; NVME initiator only; NVME target only; and SCSI and NVME initiator. The configuration mode drives overall adapter configuration, offloads enabled, and resource splits. NVME support is only available on SLI-4 devices and newer fw. - Implements the following based on configuration mode: - Exchange resources are split by protocol; Obviously, if only 1 mode, then no split occurs. Default is 50/50. module attribute allows tuning. - Pools and config parameters are separated per-protocol - Each protocol has it's own set of queues, but share interrupt vectors. SCSI: SLI3 devices have few queues and the original style of queue allocation remains. SLI4 devices piggy back on an "io-channel" concept that eventually needs to merge with scsi-mq/blk-mq support (it is underway). For now, the paradigm continues as it existed prior. io channel allocates N msix and N WQs (N=4 default) and either round robins or uses cpu # modulo N for scheduling. A bunch of module parameters allow the configuration to be tuned. NVME (initiator): Allocates an msix per cpu (or whatever pci_alloc_irq_vectors gets) Allocates a WQ per cpu, and maps the WQs to msix on a WQ # modulo msix vector count basis. Module parameters exist to cap/control the config if desired. - Each protocol has its own buffer and dma pools. I apologize for the size of the patch. Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <james.smart@broadcom.com> ---- Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2017-02-12 21:52:30 +00:00
struct lpfc_sli_ring *extra_ring = &psli->sli3_ring[LPFC_EXTRA_RING];
struct lpfc_sli_ring *fcp_ring = &psli->sli3_ring[LPFC_FCP_RING];
int rc;
/*
* if it's not a physical port or if we already send
* clear_la then don't send it.
*/
if ((phba->link_state >= LPFC_CLEAR_LA) ||
(vport->port_type != LPFC_PHYSICAL_PORT) ||
(phba->sli_rev == LPFC_SLI_REV4))
return;
/* Link up discovery */
if ((mbox = mempool_alloc(phba->mbox_mem_pool, GFP_KERNEL)) != NULL) {
phba->link_state = LPFC_CLEAR_LA;
lpfc_clear_la(phba, mbox);
mbox->mbox_cmpl = lpfc_mbx_cmpl_clear_la;
mbox->vport = vport;
rc = lpfc_sli_issue_mbox(phba, mbox, MBX_NOWAIT);
if (rc == MBX_NOT_FINISHED) {
mempool_free(mbox, phba->mbox_mem_pool);
lpfc_disc_flush_list(vport);
extra_ring->flag &= ~LPFC_STOP_IOCB_EVENT;
fcp_ring->flag &= ~LPFC_STOP_IOCB_EVENT;
phba->link_state = LPFC_HBA_ERROR;
}
}
}
/* Reg_vpi to tell firmware to resume normal operations */
void
lpfc_issue_reg_vpi(struct lpfc_hba *phba, struct lpfc_vport *vport)
{
LPFC_MBOXQ_t *regvpimbox;
regvpimbox = mempool_alloc(phba->mbox_mem_pool, GFP_KERNEL);
if (regvpimbox) {
lpfc_reg_vpi(vport, regvpimbox);
regvpimbox->mbox_cmpl = lpfc_mbx_cmpl_reg_vpi;
regvpimbox->vport = vport;
if (lpfc_sli_issue_mbox(phba, regvpimbox, MBX_NOWAIT)
== MBX_NOT_FINISHED) {
mempool_free(regvpimbox, phba->mbox_mem_pool);
}
}
}
/* Start Link up / RSCN discovery on NPR nodes */
void
lpfc_disc_start(struct lpfc_vport *vport)
{
struct Scsi_Host *shost = lpfc_shost_from_vport(vport);
struct lpfc_hba *phba = vport->phba;
uint32_t num_sent;
uint32_t clear_la_pending;
if (!lpfc_is_link_up(phba)) {
lpfc_printf_vlog(vport, KERN_INFO, LOG_SLI,
"3315 Link is not up %x\n",
phba->link_state);
return;
}
if (phba->link_state == LPFC_CLEAR_LA)
clear_la_pending = 1;
else
clear_la_pending = 0;
if (vport->port_state < LPFC_VPORT_READY)
vport->port_state = LPFC_DISC_AUTH;
lpfc_set_disctmo(vport);
vport->fc_prevDID = vport->fc_myDID;
vport->num_disc_nodes = 0;
/* Start Discovery state <hba_state> */
lpfc_printf_vlog(vport, KERN_INFO, LOG_DISCOVERY,
"0202 Start Discovery port state x%x "
"flg x%x Data: x%x x%x x%x\n",
vport->port_state, vport->fc_flag, vport->fc_plogi_cnt,
vport->fc_adisc_cnt, vport->fc_npr_cnt);
/* First do ADISCs - if any */
num_sent = lpfc_els_disc_adisc(vport);
if (num_sent)
return;
/* Register the VPI for SLI3, NPIV only. */
if ((phba->sli3_options & LPFC_SLI3_NPIV_ENABLED) &&
!(vport->fc_flag & FC_PT2PT) &&
!(vport->fc_flag & FC_RSCN_MODE) &&
(phba->sli_rev < LPFC_SLI_REV4)) {
lpfc_issue_clear_la(phba, vport);
lpfc_issue_reg_vpi(phba, vport);
return;
}
/*
* For SLI2, we need to set port_state to READY and continue
* discovery.
*/
if (vport->port_state < LPFC_VPORT_READY && !clear_la_pending) {
/* If we get here, there is nothing to ADISC */
lpfc_issue_clear_la(phba, vport);
if (!(vport->fc_flag & FC_ABORT_DISCOVERY)) {
vport->num_disc_nodes = 0;
/* go thru NPR nodes and issue ELS PLOGIs */
if (vport->fc_npr_cnt)
lpfc_els_disc_plogi(vport);
if (!vport->num_disc_nodes) {
spin_lock_irq(shost->host_lock);
vport->fc_flag &= ~FC_NDISC_ACTIVE;
spin_unlock_irq(shost->host_lock);
lpfc_can_disctmo(vport);
}
}
vport->port_state = LPFC_VPORT_READY;
} else {
/* Next do PLOGIs - if any */
num_sent = lpfc_els_disc_plogi(vport);
if (num_sent)
return;
if (vport->fc_flag & FC_RSCN_MODE) {
/* Check to see if more RSCNs came in while we
* were processing this one.
*/
if ((vport->fc_rscn_id_cnt == 0) &&
(!(vport->fc_flag & FC_RSCN_DISCOVERY))) {
spin_lock_irq(shost->host_lock);
vport->fc_flag &= ~FC_RSCN_MODE;
spin_unlock_irq(shost->host_lock);
lpfc_can_disctmo(vport);
} else
lpfc_els_handle_rscn(vport);
}
}
return;
}
/*
* Ignore completion for all IOCBs on tx and txcmpl queue for ELS
* ring the match the sppecified nodelist.
*/
static void
lpfc_free_tx(struct lpfc_hba *phba, struct lpfc_nodelist *ndlp)
{
LIST_HEAD(completions);
struct lpfc_iocbq *iocb, *next_iocb;
struct lpfc_sli_ring *pring;
u32 ulp_command;
scsi: lpfc: NVME Initiator: Base modifications NVME Initiator: Base modifications This patch adds base modifications for NVME initiator support. The base modifications consist of: - Formal split of SLI3 rings from SLI-4 WQs (sometimes referred to as rings as well) as implementation now widely varies between the two. - Addition of configuration modes: SCSI initiator only; NVME initiator only; NVME target only; and SCSI and NVME initiator. The configuration mode drives overall adapter configuration, offloads enabled, and resource splits. NVME support is only available on SLI-4 devices and newer fw. - Implements the following based on configuration mode: - Exchange resources are split by protocol; Obviously, if only 1 mode, then no split occurs. Default is 50/50. module attribute allows tuning. - Pools and config parameters are separated per-protocol - Each protocol has it's own set of queues, but share interrupt vectors. SCSI: SLI3 devices have few queues and the original style of queue allocation remains. SLI4 devices piggy back on an "io-channel" concept that eventually needs to merge with scsi-mq/blk-mq support (it is underway). For now, the paradigm continues as it existed prior. io channel allocates N msix and N WQs (N=4 default) and either round robins or uses cpu # modulo N for scheduling. A bunch of module parameters allow the configuration to be tuned. NVME (initiator): Allocates an msix per cpu (or whatever pci_alloc_irq_vectors gets) Allocates a WQ per cpu, and maps the WQs to msix on a WQ # modulo msix vector count basis. Module parameters exist to cap/control the config if desired. - Each protocol has its own buffer and dma pools. I apologize for the size of the patch. Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <james.smart@broadcom.com> ---- Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2017-02-12 21:52:30 +00:00
pring = lpfc_phba_elsring(phba);
if (unlikely(!pring))
return;
/* Error matching iocb on txq or txcmplq
* First check the txq.
*/
spin_lock_irq(&phba->hbalock);
list_for_each_entry_safe(iocb, next_iocb, &pring->txq, list) {
if (iocb->ndlp != ndlp)
continue;
ulp_command = get_job_cmnd(phba, iocb);
if (ulp_command == CMD_ELS_REQUEST64_CR ||
ulp_command == CMD_XMIT_ELS_RSP64_CX) {
list_move_tail(&iocb->list, &completions);
}
}
/* Next check the txcmplq */
list_for_each_entry_safe(iocb, next_iocb, &pring->txcmplq, list) {
if (iocb->ndlp != ndlp)
continue;
ulp_command = get_job_cmnd(phba, iocb);
if (ulp_command == CMD_ELS_REQUEST64_CR ||
ulp_command == CMD_XMIT_ELS_RSP64_CX) {
scsi: lpfc: Convert abort handling to SLI-3 and SLI-4 handlers This patch reworks the abort interfaces such that SLI-3 retains the iocb-based formatting and completions and SLI-4 now uses native WQEs and completion routines. The following changes are made: - The code is refactored from a confusing 2 routine sequence of xx_abort_iotag_issue(), which creates/formats and abort cmd, and xx_issue_abort_tag(), which then issues and handles the completion of the abort cmd - into a single interface of xx_issue_abort_iotag(). The new interface will determine whether SLI-3 or SLI-4 and then call the appropriate handler. A completion handler can now be specified to address the differences in completion handling. Note: original code is all iocb based, with SLI-4 converting to SLI-3 for the SCSI/ELS path, and NVMe natively using wqes. - The SLI-3 side is refactored: The older iocb-base lpfc_sli_issue_abort_iotag() routine is combined with the logic of lpfc_sli_abort_iotag_issue() as well as the iocb-specific code in lpfc_abort_handler() and lpfc_sli_abort_iocb() to create the new single SLI-3 abort routine that formats and issues the iocb. - The SLI-4 side is refactored and added to: The native WQE abort code in NVMe is moved to the new SLI-4 issue_abort_iotag() routine. Items in SCSI that set fields not set by NVMe is migrated into the new routine. Thus the routine supports NVMe and SCSI initiators. The nvmet block (target) formats the abort slightly different (like the old NVMe initiator) thus it has its own prep routine stolen from NVMe initiator and it retains the current code it has for issuing the WQE (does not use the commonized routine the initiators do). SLI-4 completion handlers were also added. - lpfc_abort_handler now becomes a wrapper that determines whether SLI-3 or SLI-4 and calls the proper abort handler. Link: https://lore.kernel.org/r/20201115192646.12977-16-james.smart@broadcom.com Co-developed-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <james.smart@broadcom.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-11-15 19:26:44 +00:00
lpfc_sli_issue_abort_iotag(phba, pring, iocb, NULL);
}
}
spin_unlock_irq(&phba->hbalock);
scsi: lpfc: Implement health checking when aborting I/O Several errors have occurred where the adapter stops or fails but does not raise the register values for the driver to detect failure. Thus driver is unaware of the failure. The failure typically results in I/O timeouts, the I/O timeout handler failing (after several seconds), and the error handler escalating recovery policy and resulting in more errors. Eventually, the driver is in a position where things have spiraled and it can't do recovery because other recovery ops are still outstanding and it becomes unusable. Resolve the situation by having the I/O timeout handler (actually a els, SCSI I/O, NVMe ls, or NVMe I/O timeout), in addition to aborting the I/O, perform a mailbox command and look for a response from the hardware. If the mailbox command fails, it will mark the adapter offline and then invoke the adapter reset handler to clean up. The new I/O timeout test will be limited to a test every 5s. If there are multiple I/O timeouts concurrently, only the 1st I/O timeout will generate the mailbox command. Further testing will only occur once a timeout occurs after a 5s delay from the last mailbox command has expired. Link: https://lore.kernel.org/r/20210104180240.46824-14-jsmart2021@gmail.com Co-developed-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2021-01-04 18:02:38 +00:00
/* Make sure HBA is alive */
lpfc_issue_hb_tmo(phba);
[SCSI] lpfc 8.3.1: misc fixes/changes 8.3.1 Fixes/Changes : - Fix incorrect byte-swapping on word 4 of IOCB (data length) which caused LUNs to not be discovered on big-endian (e.g. PPC) - Remove a bad cast of MBslimaddr which loses the __iomem (sparse) - Make lpfc_debugfs_mask_disc_trc static (sparse) - Correct misspelled word BlockGuard in lpfc_logmsg.h comment - Replaced repeated code segment for canceling IOCBs from a list with a function call, lpfc_sli_cancel_iocbs(). - Increased HBQ buffers to support 40KB SSC sequences. - Added sysfs interface to update speed and topology parameter without link bounce. - Fixed bug with sysfs fc_host WWNs not being updated after changing the WWNs. - Check if the active mailbox is NULL in the beginning of the mailbox timeout handler - fixes panic in the mailbox timeout handler while running IO stress test - Fixed system panic in lpfc_pci_remove_one() due to ndlp indirect reference to phba through vport - Removed de-reference of scsi device after call to scsi_done() to fix panic in scsi completion path while accessing scsi device after scsi_done is called. - Fixed "Nodelist not empty" message when unloading the driver after target reboot test - Added LP2105 HBA model description - Added code to print all 16 words of unrecognized ASYNC events - Fixed memory leak in vport create + delete loop - Added support for handling dual error bit from HBA - Fixed a driver NULL pointer dereference in lpfc_sli_process_sol_iocb - Fixed a discovery bug with FC switch reboot in lpfc_setup_disc_node - Take NULL termintator into account when calculating available buffer space Signed-off-by: James Smart <james.smart@emulex.com> Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
2009-04-06 22:48:10 +00:00
/* Cancel all the IOCBs from the completions list */
lpfc_sli_cancel_iocbs(phba, &completions, IOSTAT_LOCAL_REJECT,
IOERR_SLI_ABORTED);
}
static void
lpfc_disc_flush_list(struct lpfc_vport *vport)
{
struct lpfc_nodelist *ndlp, *next_ndlp;
struct lpfc_hba *phba = vport->phba;
if (vport->fc_plogi_cnt || vport->fc_adisc_cnt) {
list_for_each_entry_safe(ndlp, next_ndlp, &vport->fc_nodes,
nlp_listp) {
if (ndlp->nlp_state == NLP_STE_PLOGI_ISSUE ||
ndlp->nlp_state == NLP_STE_ADISC_ISSUE) {
lpfc_free_tx(phba, ndlp);
}
}
}
}
scsi: lpfc: Improve PCI EEH Error and Recovery Handling Following EEH errors, the driver can crash or hang when deleting the localport or when attempting to unload. The EEH handlers in the driver did not notify the NVMe-FC transport before tearing the driver down. This was delayed until the resume steps. This worked for SCSI because lpfc_block_scsi() would notify the scsi_fc_transport that the target was not available but it would not clean up all the references to the ndlp. The SLI3 prep for dev reset handler did the lpfc_offline_prep() and lpfc_offline() calls to get the port stopped before restarting. The SLI4 version of the prep for dev reset just destroyed the queues and did not stop NVMe from continuing. Also because the port was not really stopped the localport destroy would hang because the transport was still waiting for I/O. Additionally, a devloss tmo can fire and post events to a stopped worker thread creating another hang condition. lpfc_sli4_prep_dev_for_reset() is modified to call lpfc_offline_prep() and lpfc_offline() rather than just lpfc_scsi_dev_block() to ensure both SCSI and NVMe transports are notified to block I/O to the driver. Logic is added to devloss handler and worker thread to clean up ndlp references and quiesce appropriately. Link: https://lore.kernel.org/r/20220317032737.45308-2-jsmart2021@gmail.com Co-developed-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2022-03-17 03:27:34 +00:00
/*
* lpfc_notify_xport_npr - notifies xport of node disappearance
* @vport: Pointer to Virtual Port object.
*
* Transitions all ndlps to NPR state. When lpfc_nlp_set_state
* calls lpfc_nlp_state_cleanup, the ndlp->rport is unregistered
* and transport notified that the node is gone.
* Return Code:
* none
*/
static void
lpfc_notify_xport_npr(struct lpfc_vport *vport)
{
struct lpfc_nodelist *ndlp, *next_ndlp;
list_for_each_entry_safe(ndlp, next_ndlp, &vport->fc_nodes,
nlp_listp) {
lpfc_nlp_set_state(vport, ndlp, NLP_STE_NPR_NODE);
}
}
void
lpfc_cleanup_discovery_resources(struct lpfc_vport *vport)
{
lpfc_els_flush_rscn(vport);
lpfc_els_flush_cmd(vport);
lpfc_disc_flush_list(vport);
scsi: lpfc: Improve PCI EEH Error and Recovery Handling Following EEH errors, the driver can crash or hang when deleting the localport or when attempting to unload. The EEH handlers in the driver did not notify the NVMe-FC transport before tearing the driver down. This was delayed until the resume steps. This worked for SCSI because lpfc_block_scsi() would notify the scsi_fc_transport that the target was not available but it would not clean up all the references to the ndlp. The SLI3 prep for dev reset handler did the lpfc_offline_prep() and lpfc_offline() calls to get the port stopped before restarting. The SLI4 version of the prep for dev reset just destroyed the queues and did not stop NVMe from continuing. Also because the port was not really stopped the localport destroy would hang because the transport was still waiting for I/O. Additionally, a devloss tmo can fire and post events to a stopped worker thread creating another hang condition. lpfc_sli4_prep_dev_for_reset() is modified to call lpfc_offline_prep() and lpfc_offline() rather than just lpfc_scsi_dev_block() to ensure both SCSI and NVMe transports are notified to block I/O to the driver. Logic is added to devloss handler and worker thread to clean up ndlp references and quiesce appropriately. Link: https://lore.kernel.org/r/20220317032737.45308-2-jsmart2021@gmail.com Co-developed-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2022-03-17 03:27:34 +00:00
if (pci_channel_offline(vport->phba->pcidev))
lpfc_notify_xport_npr(vport);
}
/*****************************************************************************/
/*
* NAME: lpfc_disc_timeout
*
* FUNCTION: Fibre Channel driver discovery timeout routine.
*
* EXECUTION ENVIRONMENT: interrupt only
*
* CALLED FROM:
* Timer function
*
* RETURNS:
* none
*/
/*****************************************************************************/
void
lpfc_disc_timeout(struct timer_list *t)
{
struct lpfc_vport *vport = from_timer(vport, t, fc_disctmo);
struct lpfc_hba *phba = vport->phba;
uint32_t tmo_posted;
unsigned long flags = 0;
if (unlikely(!phba))
return;
spin_lock_irqsave(&vport->work_port_lock, flags);
tmo_posted = vport->work_port_events & WORKER_DISC_TMO;
if (!tmo_posted)
vport->work_port_events |= WORKER_DISC_TMO;
spin_unlock_irqrestore(&vport->work_port_lock, flags);
if (!tmo_posted)
lpfc_worker_wake_up(phba);
return;
}
static void
lpfc_disc_timeout_handler(struct lpfc_vport *vport)
{
struct Scsi_Host *shost = lpfc_shost_from_vport(vport);
struct lpfc_hba *phba = vport->phba;
struct lpfc_sli *psli = &phba->sli;
struct lpfc_nodelist *ndlp, *next_ndlp;
LPFC_MBOXQ_t *initlinkmbox;
int rc, clrlaerr = 0;
if (!(vport->fc_flag & FC_DISC_TMO))
return;
spin_lock_irq(shost->host_lock);
vport->fc_flag &= ~FC_DISC_TMO;
spin_unlock_irq(shost->host_lock);
lpfc_debugfs_disc_trc(vport, LPFC_DISC_TRC_ELS_CMD,
"disc timeout: state:x%x rtry:x%x flg:x%x",
vport->port_state, vport->fc_ns_retry, vport->fc_flag);
switch (vport->port_state) {
case LPFC_LOCAL_CFG_LINK:
/*
* port_state is identically LPFC_LOCAL_CFG_LINK while
* waiting for FAN timeout
*/
lpfc_printf_vlog(vport, KERN_WARNING, LOG_DISCOVERY,
"0221 FAN timeout\n");
/* Start discovery by sending FLOGI, clean up old rpis */
list_for_each_entry_safe(ndlp, next_ndlp, &vport->fc_nodes,
nlp_listp) {
if (ndlp->nlp_state != NLP_STE_NPR_NODE)
continue;
if (ndlp->nlp_type & NLP_FABRIC) {
/* Clean up the ndlp on Fabric connections */
lpfc_drop_node(vport, ndlp);
} else if (!(ndlp->nlp_flag & NLP_NPR_ADISC)) {
/* Fail outstanding IO now since device
* is marked for PLOGI.
*/
lpfc_unreg_rpi(vport, ndlp);
}
}
if (vport->port_state != LPFC_FLOGI) {
2010-11-21 04:11:48 +00:00
if (phba->sli_rev <= LPFC_SLI_REV3)
lpfc_initial_flogi(vport);
else
lpfc_issue_init_vfi(vport);
return;
}
break;
case LPFC_FDISC:
case LPFC_FLOGI:
/* port_state is identically LPFC_FLOGI while waiting for FLOGI cmpl */
/* Initial FLOGI timeout */
scsi: lpfc: Add an internal trace log buffer The current logging methods typically end up requesting a reproduction with a different logging level set to figure out what happened. This was mainly by design to not clutter the kernel log messages with things that were typically not interesting and the messages themselves could cause other issues. When looking to make a better system, it was seen that in many cases when more data was wanted was when another message, usually at KERN_ERR level, was logged. And in most cases, what the additional logging that was then enabled was typically. Most of these areas fell into the discovery machine. Based on this summary, the following design has been put in place: The driver will maintain an internal log (256 elements of 256 bytes). The "additional logging" messages that are usually enabled in a reproduction will be changed to now log all the time to the internal log. A new logging level is defined - LOG_TRACE_EVENT. When this level is set (it is not by default) and a message marked as KERN_ERR is logged, all the messages in the internal log will be dumped to the kernel log before the KERN_ERR message is logged. There is a timestamp on each message added to the internal log. However, this timestamp is not converted to wall time when logged. The value of the timestamp is solely to give a crude time reference for the messages. Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-06-30 21:50:00 +00:00
lpfc_printf_vlog(vport, KERN_ERR,
LOG_TRACE_EVENT,
"0222 Initial %s timeout\n",
vport->vpi ? "FDISC" : "FLOGI");
/* Assume no Fabric and go on with discovery.
* Check for outstanding ELS FLOGI to abort.
*/
/* FLOGI failed, so just use loop map to make discovery list */
lpfc_disc_list_loopmap(vport);
/* Start discovery */
lpfc_disc_start(vport);
break;
case LPFC_FABRIC_CFG_LINK:
/* hba_state is identically LPFC_FABRIC_CFG_LINK while waiting for
NameServer login */
scsi: lpfc: Add an internal trace log buffer The current logging methods typically end up requesting a reproduction with a different logging level set to figure out what happened. This was mainly by design to not clutter the kernel log messages with things that were typically not interesting and the messages themselves could cause other issues. When looking to make a better system, it was seen that in many cases when more data was wanted was when another message, usually at KERN_ERR level, was logged. And in most cases, what the additional logging that was then enabled was typically. Most of these areas fell into the discovery machine. Based on this summary, the following design has been put in place: The driver will maintain an internal log (256 elements of 256 bytes). The "additional logging" messages that are usually enabled in a reproduction will be changed to now log all the time to the internal log. A new logging level is defined - LOG_TRACE_EVENT. When this level is set (it is not by default) and a message marked as KERN_ERR is logged, all the messages in the internal log will be dumped to the kernel log before the KERN_ERR message is logged. There is a timestamp on each message added to the internal log. However, this timestamp is not converted to wall time when logged. The value of the timestamp is solely to give a crude time reference for the messages. Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-06-30 21:50:00 +00:00
lpfc_printf_vlog(vport, KERN_ERR,
LOG_TRACE_EVENT,
"0223 Timeout while waiting for "
"NameServer login\n");
/* Next look for NameServer ndlp */
ndlp = lpfc_findnode_did(vport, NameServer_DID);
scsi: lpfc: Rework remote port ref counting and node freeing When a remote port is disconnected and disappears, its node structure (ndlp) stays allocated and on a vport node list. While on the list it can be matched, thus requires validation checks on state to be added in numerous code paths. If the node comes back, its possible for there to be multiple node structures for the same device on the vport node list. There is no reason to keep the node structure around after it is no longer in existence, and the current implementation creates problems for itself (multiple nodes) and lots of unnecessary code for state validation. Additionally, the reference taking on the node structure didn't follow the normal model used by the kernel kref api. It included lots of odd logic to match state with reference count. The combination of this odd logic plus the way it was implicitly used in the discovery engine made its reference taking implementation suspect and extremely hard to follow. Change the driver such that the reference taking routines are now normal ref increments/decrements and callout on refcount=0. With this in place, the rework can be done such that the node structure is fully removed and deallocated when the remote port no longer exists and all references are removed. This removal logic, and the basic ref counting are intrically tied, thus in a single patch. Link: https://lore.kernel.org/r/20201115192646.12977-2-james.smart@broadcom.com Co-developed-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <james.smart@broadcom.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-11-15 19:26:30 +00:00
if (ndlp)
lpfc_els_abort(phba, ndlp);
/* ReStart discovery */
goto restart_disc;
case LPFC_NS_QRY:
/* Check for wait for NameServer Rsp timeout */
scsi: lpfc: Add an internal trace log buffer The current logging methods typically end up requesting a reproduction with a different logging level set to figure out what happened. This was mainly by design to not clutter the kernel log messages with things that were typically not interesting and the messages themselves could cause other issues. When looking to make a better system, it was seen that in many cases when more data was wanted was when another message, usually at KERN_ERR level, was logged. And in most cases, what the additional logging that was then enabled was typically. Most of these areas fell into the discovery machine. Based on this summary, the following design has been put in place: The driver will maintain an internal log (256 elements of 256 bytes). The "additional logging" messages that are usually enabled in a reproduction will be changed to now log all the time to the internal log. A new logging level is defined - LOG_TRACE_EVENT. When this level is set (it is not by default) and a message marked as KERN_ERR is logged, all the messages in the internal log will be dumped to the kernel log before the KERN_ERR message is logged. There is a timestamp on each message added to the internal log. However, this timestamp is not converted to wall time when logged. The value of the timestamp is solely to give a crude time reference for the messages. Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-06-30 21:50:00 +00:00
lpfc_printf_vlog(vport, KERN_ERR,
LOG_TRACE_EVENT,
"0224 NameServer Query timeout "
"Data: x%x x%x\n",
vport->fc_ns_retry, LPFC_MAX_NS_RETRY);
if (vport->fc_ns_retry < LPFC_MAX_NS_RETRY) {
/* Try it one more time */
vport->fc_ns_retry++;
vport->gidft_inp = 0;
rc = lpfc_issue_gidft(vport);
if (rc == 0)
break;
}
vport->fc_ns_retry = 0;
restart_disc:
/*
* Discovery is over.
* set port_state to PORT_READY if SLI2.
* cmpl_reg_vpi will set port_state to READY for SLI3.
*/
if (phba->sli_rev < LPFC_SLI_REV4) {
if (phba->sli3_options & LPFC_SLI3_NPIV_ENABLED)
lpfc_issue_reg_vpi(phba, vport);
else {
lpfc_issue_clear_la(phba, vport);
vport->port_state = LPFC_VPORT_READY;
}
}
/* Setup and issue mailbox INITIALIZE LINK command */
initlinkmbox = mempool_alloc(phba->mbox_mem_pool, GFP_KERNEL);
if (!initlinkmbox) {
scsi: lpfc: Add an internal trace log buffer The current logging methods typically end up requesting a reproduction with a different logging level set to figure out what happened. This was mainly by design to not clutter the kernel log messages with things that were typically not interesting and the messages themselves could cause other issues. When looking to make a better system, it was seen that in many cases when more data was wanted was when another message, usually at KERN_ERR level, was logged. And in most cases, what the additional logging that was then enabled was typically. Most of these areas fell into the discovery machine. Based on this summary, the following design has been put in place: The driver will maintain an internal log (256 elements of 256 bytes). The "additional logging" messages that are usually enabled in a reproduction will be changed to now log all the time to the internal log. A new logging level is defined - LOG_TRACE_EVENT. When this level is set (it is not by default) and a message marked as KERN_ERR is logged, all the messages in the internal log will be dumped to the kernel log before the KERN_ERR message is logged. There is a timestamp on each message added to the internal log. However, this timestamp is not converted to wall time when logged. The value of the timestamp is solely to give a crude time reference for the messages. Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-06-30 21:50:00 +00:00
lpfc_printf_vlog(vport, KERN_ERR,
LOG_TRACE_EVENT,
"0206 Device Discovery "
"completion error\n");
phba->link_state = LPFC_HBA_ERROR;
break;
}
lpfc_linkdown(phba);
lpfc_init_link(phba, initlinkmbox, phba->cfg_topology,
phba->cfg_link_speed);
initlinkmbox->u.mb.un.varInitLnk.lipsr_AL_PA = 0;
initlinkmbox->vport = vport;
initlinkmbox->mbox_cmpl = lpfc_sli_def_mbox_cmpl;
rc = lpfc_sli_issue_mbox(phba, initlinkmbox, MBX_NOWAIT);
lpfc_set_loopback_flag(phba);
if (rc == MBX_NOT_FINISHED)
mempool_free(initlinkmbox, phba->mbox_mem_pool);
break;
case LPFC_DISC_AUTH:
/* Node Authentication timeout */
scsi: lpfc: Add an internal trace log buffer The current logging methods typically end up requesting a reproduction with a different logging level set to figure out what happened. This was mainly by design to not clutter the kernel log messages with things that were typically not interesting and the messages themselves could cause other issues. When looking to make a better system, it was seen that in many cases when more data was wanted was when another message, usually at KERN_ERR level, was logged. And in most cases, what the additional logging that was then enabled was typically. Most of these areas fell into the discovery machine. Based on this summary, the following design has been put in place: The driver will maintain an internal log (256 elements of 256 bytes). The "additional logging" messages that are usually enabled in a reproduction will be changed to now log all the time to the internal log. A new logging level is defined - LOG_TRACE_EVENT. When this level is set (it is not by default) and a message marked as KERN_ERR is logged, all the messages in the internal log will be dumped to the kernel log before the KERN_ERR message is logged. There is a timestamp on each message added to the internal log. However, this timestamp is not converted to wall time when logged. The value of the timestamp is solely to give a crude time reference for the messages. Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-06-30 21:50:00 +00:00
lpfc_printf_vlog(vport, KERN_ERR,
LOG_TRACE_EVENT,
"0227 Node Authentication timeout\n");
lpfc_disc_flush_list(vport);
/*
* set port_state to PORT_READY if SLI2.
* cmpl_reg_vpi will set port_state to READY for SLI3.
*/
if (phba->sli_rev < LPFC_SLI_REV4) {
if (phba->sli3_options & LPFC_SLI3_NPIV_ENABLED)
lpfc_issue_reg_vpi(phba, vport);
else { /* NPIV Not enabled */
lpfc_issue_clear_la(phba, vport);
vport->port_state = LPFC_VPORT_READY;
}
}
break;
case LPFC_VPORT_READY:
if (vport->fc_flag & FC_RSCN_MODE) {
scsi: lpfc: Add an internal trace log buffer The current logging methods typically end up requesting a reproduction with a different logging level set to figure out what happened. This was mainly by design to not clutter the kernel log messages with things that were typically not interesting and the messages themselves could cause other issues. When looking to make a better system, it was seen that in many cases when more data was wanted was when another message, usually at KERN_ERR level, was logged. And in most cases, what the additional logging that was then enabled was typically. Most of these areas fell into the discovery machine. Based on this summary, the following design has been put in place: The driver will maintain an internal log (256 elements of 256 bytes). The "additional logging" messages that are usually enabled in a reproduction will be changed to now log all the time to the internal log. A new logging level is defined - LOG_TRACE_EVENT. When this level is set (it is not by default) and a message marked as KERN_ERR is logged, all the messages in the internal log will be dumped to the kernel log before the KERN_ERR message is logged. There is a timestamp on each message added to the internal log. However, this timestamp is not converted to wall time when logged. The value of the timestamp is solely to give a crude time reference for the messages. Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-06-30 21:50:00 +00:00
lpfc_printf_vlog(vport, KERN_ERR,
LOG_TRACE_EVENT,
"0231 RSCN timeout Data: x%x "
"x%x x%x x%x\n",
vport->fc_ns_retry, LPFC_MAX_NS_RETRY,
vport->port_state, vport->gidft_inp);
/* Cleanup any outstanding ELS commands */
lpfc_els_flush_cmd(vport);
lpfc_els_flush_rscn(vport);
lpfc_disc_flush_list(vport);
}
break;
default:
scsi: lpfc: Add an internal trace log buffer The current logging methods typically end up requesting a reproduction with a different logging level set to figure out what happened. This was mainly by design to not clutter the kernel log messages with things that were typically not interesting and the messages themselves could cause other issues. When looking to make a better system, it was seen that in many cases when more data was wanted was when another message, usually at KERN_ERR level, was logged. And in most cases, what the additional logging that was then enabled was typically. Most of these areas fell into the discovery machine. Based on this summary, the following design has been put in place: The driver will maintain an internal log (256 elements of 256 bytes). The "additional logging" messages that are usually enabled in a reproduction will be changed to now log all the time to the internal log. A new logging level is defined - LOG_TRACE_EVENT. When this level is set (it is not by default) and a message marked as KERN_ERR is logged, all the messages in the internal log will be dumped to the kernel log before the KERN_ERR message is logged. There is a timestamp on each message added to the internal log. However, this timestamp is not converted to wall time when logged. The value of the timestamp is solely to give a crude time reference for the messages. Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-06-30 21:50:00 +00:00
lpfc_printf_vlog(vport, KERN_ERR,
LOG_TRACE_EVENT,
"0273 Unexpected discovery timeout, "
"vport State x%x\n", vport->port_state);
break;
}
switch (phba->link_state) {
case LPFC_CLEAR_LA:
/* CLEAR LA timeout */
scsi: lpfc: Add an internal trace log buffer The current logging methods typically end up requesting a reproduction with a different logging level set to figure out what happened. This was mainly by design to not clutter the kernel log messages with things that were typically not interesting and the messages themselves could cause other issues. When looking to make a better system, it was seen that in many cases when more data was wanted was when another message, usually at KERN_ERR level, was logged. And in most cases, what the additional logging that was then enabled was typically. Most of these areas fell into the discovery machine. Based on this summary, the following design has been put in place: The driver will maintain an internal log (256 elements of 256 bytes). The "additional logging" messages that are usually enabled in a reproduction will be changed to now log all the time to the internal log. A new logging level is defined - LOG_TRACE_EVENT. When this level is set (it is not by default) and a message marked as KERN_ERR is logged, all the messages in the internal log will be dumped to the kernel log before the KERN_ERR message is logged. There is a timestamp on each message added to the internal log. However, this timestamp is not converted to wall time when logged. The value of the timestamp is solely to give a crude time reference for the messages. Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-06-30 21:50:00 +00:00
lpfc_printf_vlog(vport, KERN_ERR,
LOG_TRACE_EVENT,
"0228 CLEAR LA timeout\n");
clrlaerr = 1;
break;
[SCSI] lpfc 8.2.4 : Miscellaneous Fixes Miscellaneous Fixes: - Fix a couple of sparse complaints - Reset the FCP recovery flag when the node is not a FCP2 device. - Speed up offline prep delays - Fixed a memory leak in lpfc_mem_alloc failure path - Fixed external loopback test. - Fixed error code returned from the driver when HBA is over heated. - Correct Max NPIV vport to limits read from adapter - Add missing locks around fc_flag and FC_NEEDS_REG_VPI - Add missing hba ids for device identification - Added support for SET_VARIABLE and MBX_WRITE_WWN mailbox commands - Changed all temperature event messages from warning to error - Fix reporting of link speed when link is down - Added support for MBX_WRITE_WWN mailbox command - Change del_timer_sync() in ISR to del_timer() in interrupt handler - Correct instances of beXX_to_cpu() that should be cpu_to_beXX() - Perform target flush before releasing node references on module unload - Avoid bogus devloss_tmo messages when driver unloads - Fix panic when HBA generates ERATT interupt - Fix mbox race condition and a workaround on back-to-back mailbox commands - Force NPIV off for pt2pt mode between 2 NPorts - Stop worker thread before removing fc_host. - Fix up discovery timeout error case due to missing clear_la - Tighten mailbox polling code to speed up detection of fast completions - Only allow DUMP_MEMORY if adapter offline due to overtemp errors - Added extended error information to the log messages in chip init. Signed-off-by: James Smart <James.Smart@emulex.com> Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
2008-01-11 06:52:54 +00:00
case LPFC_LINK_UP:
lpfc_issue_clear_la(phba, vport);
fallthrough;
case LPFC_LINK_UNKNOWN:
case LPFC_WARM_START:
case LPFC_INIT_START:
case LPFC_INIT_MBX_CMDS:
case LPFC_LINK_DOWN:
case LPFC_HBA_ERROR:
scsi: lpfc: Add an internal trace log buffer The current logging methods typically end up requesting a reproduction with a different logging level set to figure out what happened. This was mainly by design to not clutter the kernel log messages with things that were typically not interesting and the messages themselves could cause other issues. When looking to make a better system, it was seen that in many cases when more data was wanted was when another message, usually at KERN_ERR level, was logged. And in most cases, what the additional logging that was then enabled was typically. Most of these areas fell into the discovery machine. Based on this summary, the following design has been put in place: The driver will maintain an internal log (256 elements of 256 bytes). The "additional logging" messages that are usually enabled in a reproduction will be changed to now log all the time to the internal log. A new logging level is defined - LOG_TRACE_EVENT. When this level is set (it is not by default) and a message marked as KERN_ERR is logged, all the messages in the internal log will be dumped to the kernel log before the KERN_ERR message is logged. There is a timestamp on each message added to the internal log. However, this timestamp is not converted to wall time when logged. The value of the timestamp is solely to give a crude time reference for the messages. Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-06-30 21:50:00 +00:00
lpfc_printf_vlog(vport, KERN_ERR,
LOG_TRACE_EVENT,
"0230 Unexpected timeout, hba link "
"state x%x\n", phba->link_state);
clrlaerr = 1;
break;
case LPFC_HBA_READY:
break;
}
if (clrlaerr) {
lpfc_disc_flush_list(vport);
scsi: lpfc: NVME Initiator: Base modifications NVME Initiator: Base modifications This patch adds base modifications for NVME initiator support. The base modifications consist of: - Formal split of SLI3 rings from SLI-4 WQs (sometimes referred to as rings as well) as implementation now widely varies between the two. - Addition of configuration modes: SCSI initiator only; NVME initiator only; NVME target only; and SCSI and NVME initiator. The configuration mode drives overall adapter configuration, offloads enabled, and resource splits. NVME support is only available on SLI-4 devices and newer fw. - Implements the following based on configuration mode: - Exchange resources are split by protocol; Obviously, if only 1 mode, then no split occurs. Default is 50/50. module attribute allows tuning. - Pools and config parameters are separated per-protocol - Each protocol has it's own set of queues, but share interrupt vectors. SCSI: SLI3 devices have few queues and the original style of queue allocation remains. SLI4 devices piggy back on an "io-channel" concept that eventually needs to merge with scsi-mq/blk-mq support (it is underway). For now, the paradigm continues as it existed prior. io channel allocates N msix and N WQs (N=4 default) and either round robins or uses cpu # modulo N for scheduling. A bunch of module parameters allow the configuration to be tuned. NVME (initiator): Allocates an msix per cpu (or whatever pci_alloc_irq_vectors gets) Allocates a WQ per cpu, and maps the WQs to msix on a WQ # modulo msix vector count basis. Module parameters exist to cap/control the config if desired. - Each protocol has its own buffer and dma pools. I apologize for the size of the patch. Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <james.smart@broadcom.com> ---- Reviewed-by: Hannes Reinecke <hare@suse.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2017-02-12 21:52:30 +00:00
if (phba->sli_rev != LPFC_SLI_REV4) {
psli->sli3_ring[(LPFC_EXTRA_RING)].flag &=
~LPFC_STOP_IOCB_EVENT;
psli->sli3_ring[LPFC_FCP_RING].flag &=
~LPFC_STOP_IOCB_EVENT;
}
vport->port_state = LPFC_VPORT_READY;
}
return;
}
/*
* This routine handles processing a NameServer REG_LOGIN mailbox
* command upon completion. It is setup in the LPFC_MBOXQ
* as the completion routine when the command is
* handed off to the SLI layer.
*/
void
lpfc_mbx_cmpl_fdmi_reg_login(struct lpfc_hba *phba, LPFC_MBOXQ_t *pmb)
{
MAILBOX_t *mb = &pmb->u.mb;
struct lpfc_nodelist *ndlp = (struct lpfc_nodelist *)pmb->ctx_ndlp;
struct lpfc_vport *vport = pmb->vport;
pmb->ctx_ndlp = NULL;
if (phba->sli_rev < LPFC_SLI_REV4)
ndlp->nlp_rpi = mb->un.varWords[0];
ndlp->nlp_flag |= NLP_RPI_REGISTERED;
ndlp->nlp_type |= NLP_FABRIC;
lpfc_nlp_set_state(vport, ndlp, NLP_STE_UNMAPPED_NODE);
scsi: lpfc: Fix device recovery errors after PLOGI failures When target-side fault injections are made, the driver isn't reconnecting to the remote port. The driver is logging "2753" error messages which state: "PLOGI failure DID:1B2400 Status:x3/xf0240008" The failures status is indicating a Illegal field error, which points to the Temporary RPI field being used for the ELS. This error typically means the driver used an RPI that was already registered (shouldn't be registered if using it in this context). Study has found that if the driver were in discovery attempts and encountered an error, it wouldn't flag the temporary rpi in error. Yet the rpi was released for reallocation in these error paths and another ELS could allocate the rpi. In the failure situation a retry was done on an ELS that had encountered an error, and as the rpi wasn't marked in error, the ELS reused the rpi it originally allocated. But that rpi had been allocated by a different ELS issued after the original error and before the retry attempt. The different ELS had succeeded and the RPI was registered. Fix by marking the rpi state for the node to be in error, aka as needing reallocation, upon an error in the els processing. Error state marking is always done prior to release back to the internal rpi free list, which the driver wasn't doing in cases prior. Also enhanced some of the logging to help in the next case of problem troubleshooting. Link: https://lore.kernel.org/r/20190922035906.10977-7-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2019-09-22 03:58:52 +00:00
lpfc_printf_vlog(vport, KERN_INFO, LOG_NODE | LOG_DISCOVERY,
scsi: lpfc: Rework remote port ref counting and node freeing When a remote port is disconnected and disappears, its node structure (ndlp) stays allocated and on a vport node list. While on the list it can be matched, thus requires validation checks on state to be added in numerous code paths. If the node comes back, its possible for there to be multiple node structures for the same device on the vport node list. There is no reason to keep the node structure around after it is no longer in existence, and the current implementation creates problems for itself (multiple nodes) and lots of unnecessary code for state validation. Additionally, the reference taking on the node structure didn't follow the normal model used by the kernel kref api. It included lots of odd logic to match state with reference count. The combination of this odd logic plus the way it was implicitly used in the discovery engine made its reference taking implementation suspect and extremely hard to follow. Change the driver such that the reference taking routines are now normal ref increments/decrements and callout on refcount=0. With this in place, the rework can be done such that the node structure is fully removed and deallocated when the remote port no longer exists and all references are removed. This removal logic, and the basic ref counting are intrically tied, thus in a single patch. Link: https://lore.kernel.org/r/20201115192646.12977-2-james.smart@broadcom.com Co-developed-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <james.smart@broadcom.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-11-15 19:26:30 +00:00
"0004 rpi:%x DID:%x flg:%x %d x%px\n",
ndlp->nlp_rpi, ndlp->nlp_DID, ndlp->nlp_flag,
kref_read(&ndlp->kref),
scsi: lpfc: Rework remote port ref counting and node freeing When a remote port is disconnected and disappears, its node structure (ndlp) stays allocated and on a vport node list. While on the list it can be matched, thus requires validation checks on state to be added in numerous code paths. If the node comes back, its possible for there to be multiple node structures for the same device on the vport node list. There is no reason to keep the node structure around after it is no longer in existence, and the current implementation creates problems for itself (multiple nodes) and lots of unnecessary code for state validation. Additionally, the reference taking on the node structure didn't follow the normal model used by the kernel kref api. It included lots of odd logic to match state with reference count. The combination of this odd logic plus the way it was implicitly used in the discovery engine made its reference taking implementation suspect and extremely hard to follow. Change the driver such that the reference taking routines are now normal ref increments/decrements and callout on refcount=0. With this in place, the rework can be done such that the node structure is fully removed and deallocated when the remote port no longer exists and all references are removed. This removal logic, and the basic ref counting are intrically tied, thus in a single patch. Link: https://lore.kernel.org/r/20201115192646.12977-2-james.smart@broadcom.com Co-developed-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <james.smart@broadcom.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-11-15 19:26:30 +00:00
ndlp);
/*
* Start issuing Fabric-Device Management Interface (FDMI) command to
* 0xfffffa (FDMI well known port).
* DHBA -> DPRT -> RHBA -> RPA (physical port)
* DPRT -> RPRT (vports)
*/
if (vport->port_type == LPFC_PHYSICAL_PORT) {
phba->link_flag &= ~LS_CT_VEN_RPA; /* For extra Vendor RPA */
lpfc_fdmi_cmd(vport, ndlp, SLI_MGMT_DHBA, 0);
} else {
lpfc_fdmi_cmd(vport, ndlp, SLI_MGMT_DPRT, 0);
}
/* decrement the node reference count held for this callback
* function.
*/
lpfc_nlp_put(ndlp);
lpfc_mbox_rsrc_cleanup(phba, pmb, MBOX_THD_UNLOCKED);
return;
}
static int
lpfc_filter_by_rpi(struct lpfc_nodelist *ndlp, void *param)
{
uint16_t *rpi = param;
return ndlp->nlp_rpi == *rpi;
}
static int
lpfc_filter_by_wwpn(struct lpfc_nodelist *ndlp, void *param)
{
return memcmp(&ndlp->nlp_portname, param,
sizeof(ndlp->nlp_portname)) == 0;
}
static struct lpfc_nodelist *
__lpfc_find_node(struct lpfc_vport *vport, node_filter filter, void *param)
{
struct lpfc_nodelist *ndlp;
list_for_each_entry(ndlp, &vport->fc_nodes, nlp_listp) {
if (filter(ndlp, param)) {
lpfc_printf_vlog(vport, KERN_INFO, LOG_NODE,
"3185 FIND node filter %ps DID "
"ndlp x%px did x%x flg x%x st x%x "
"xri x%x type x%x rpi x%x\n",
filter, ndlp, ndlp->nlp_DID,
ndlp->nlp_flag, ndlp->nlp_state,
ndlp->nlp_xri, ndlp->nlp_type,
ndlp->nlp_rpi);
return ndlp;
}
}
lpfc_printf_vlog(vport, KERN_INFO, LOG_NODE,
"3186 FIND node filter %ps NOT FOUND.\n", filter);
return NULL;
}
/*
* This routine looks up the ndlp lists for the given RPI. If rpi found it
* returns the node list element pointer else return NULL.
*/
struct lpfc_nodelist *
__lpfc_findnode_rpi(struct lpfc_vport *vport, uint16_t rpi)
{
return __lpfc_find_node(vport, lpfc_filter_by_rpi, &rpi);
}
/*
* This routine looks up the ndlp lists for the given WWPN. If WWPN found it
* returns the node element list pointer else return NULL.
*/
struct lpfc_nodelist *
lpfc_findnode_wwpn(struct lpfc_vport *vport, struct lpfc_name *wwpn)
{
struct Scsi_Host *shost = lpfc_shost_from_vport(vport);
struct lpfc_nodelist *ndlp;
spin_lock_irq(shost->host_lock);
ndlp = __lpfc_find_node(vport, lpfc_filter_by_wwpn, wwpn);
spin_unlock_irq(shost->host_lock);
return ndlp;
}
/*
* This routine looks up the ndlp lists for the given RPI. If the rpi
* is found, the routine returns the node element list pointer else
* return NULL.
*/
struct lpfc_nodelist *
lpfc_findnode_rpi(struct lpfc_vport *vport, uint16_t rpi)
{
struct Scsi_Host *shost = lpfc_shost_from_vport(vport);
struct lpfc_nodelist *ndlp;
unsigned long flags;
spin_lock_irqsave(shost->host_lock, flags);
ndlp = __lpfc_findnode_rpi(vport, rpi);
spin_unlock_irqrestore(shost->host_lock, flags);
return ndlp;
}
/**
* lpfc_find_vport_by_vpid - Find a vport on a HBA through vport identifier
* @phba: pointer to lpfc hba data structure.
* @vpi: the physical host virtual N_Port identifier.
*
* This routine finds a vport on a HBA (referred by @phba) through a
* @vpi. The function walks the HBA's vport list and returns the address
* of the vport with the matching @vpi.
*
* Return code
* NULL - No vport with the matching @vpi found
* Otherwise - Address to the vport with the matching @vpi.
**/
struct lpfc_vport *
lpfc_find_vport_by_vpid(struct lpfc_hba *phba, uint16_t vpi)
{
struct lpfc_vport *vport;
unsigned long flags;
int i = 0;
/* The physical ports are always vpi 0 - translate is unnecessary. */
if (vpi > 0) {
/*
* Translate the physical vpi to the logical vpi. The
* vport stores the logical vpi.
*/
for (i = 0; i <= phba->max_vpi; i++) {
if (vpi == phba->vpi_ids[i])
break;
}
if (i > phba->max_vpi) {
scsi: lpfc: Add an internal trace log buffer The current logging methods typically end up requesting a reproduction with a different logging level set to figure out what happened. This was mainly by design to not clutter the kernel log messages with things that were typically not interesting and the messages themselves could cause other issues. When looking to make a better system, it was seen that in many cases when more data was wanted was when another message, usually at KERN_ERR level, was logged. And in most cases, what the additional logging that was then enabled was typically. Most of these areas fell into the discovery machine. Based on this summary, the following design has been put in place: The driver will maintain an internal log (256 elements of 256 bytes). The "additional logging" messages that are usually enabled in a reproduction will be changed to now log all the time to the internal log. A new logging level is defined - LOG_TRACE_EVENT. When this level is set (it is not by default) and a message marked as KERN_ERR is logged, all the messages in the internal log will be dumped to the kernel log before the KERN_ERR message is logged. There is a timestamp on each message added to the internal log. However, this timestamp is not converted to wall time when logged. The value of the timestamp is solely to give a crude time reference for the messages. Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-06-30 21:50:00 +00:00
lpfc_printf_log(phba, KERN_ERR, LOG_TRACE_EVENT,
"2936 Could not find Vport mapped "
"to vpi %d\n", vpi);
return NULL;
}
}
spin_lock_irqsave(&phba->port_list_lock, flags);
list_for_each_entry(vport, &phba->port_list, listentry) {
if (vport->vpi == i) {
spin_unlock_irqrestore(&phba->port_list_lock, flags);
return vport;
}
}
spin_unlock_irqrestore(&phba->port_list_lock, flags);
return NULL;
}
struct lpfc_nodelist *
lpfc_nlp_init(struct lpfc_vport *vport, uint32_t did)
{
struct lpfc_nodelist *ndlp;
int rpi = LPFC_RPI_ALLOC_ERROR;
if (vport->phba->sli_rev == LPFC_SLI_REV4) {
rpi = lpfc_sli4_alloc_rpi(vport->phba);
if (rpi == LPFC_RPI_ALLOC_ERROR)
return NULL;
}
ndlp = mempool_alloc(vport->phba->nlp_mem_pool, GFP_KERNEL);
if (!ndlp) {
if (vport->phba->sli_rev == LPFC_SLI_REV4)
lpfc_sli4_free_rpi(vport->phba, rpi);
return NULL;
}
memset(ndlp, 0, sizeof (struct lpfc_nodelist));
spin_lock_init(&ndlp->lock);
lpfc_initialize_node(vport, ndlp, did);
INIT_LIST_HEAD(&ndlp->nlp_listp);
if (vport->phba->sli_rev == LPFC_SLI_REV4) {
ndlp->nlp_rpi = rpi;
scsi: lpfc: Fix device recovery errors after PLOGI failures When target-side fault injections are made, the driver isn't reconnecting to the remote port. The driver is logging "2753" error messages which state: "PLOGI failure DID:1B2400 Status:x3/xf0240008" The failures status is indicating a Illegal field error, which points to the Temporary RPI field being used for the ELS. This error typically means the driver used an RPI that was already registered (shouldn't be registered if using it in this context). Study has found that if the driver were in discovery attempts and encountered an error, it wouldn't flag the temporary rpi in error. Yet the rpi was released for reallocation in these error paths and another ELS could allocate the rpi. In the failure situation a retry was done on an ELS that had encountered an error, and as the rpi wasn't marked in error, the ELS reused the rpi it originally allocated. But that rpi had been allocated by a different ELS issued after the original error and before the retry attempt. The different ELS had succeeded and the RPI was registered. Fix by marking the rpi state for the node to be in error, aka as needing reallocation, upon an error in the els processing. Error state marking is always done prior to release back to the internal rpi free list, which the driver wasn't doing in cases prior. Also enhanced some of the logging to help in the next case of problem troubleshooting. Link: https://lore.kernel.org/r/20190922035906.10977-7-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2019-09-22 03:58:52 +00:00
lpfc_printf_vlog(vport, KERN_INFO, LOG_NODE | LOG_DISCOVERY,
"0007 Init New ndlp x%px, rpi:x%x DID:%x "
scsi: lpfc: Rework remote port ref counting and node freeing When a remote port is disconnected and disappears, its node structure (ndlp) stays allocated and on a vport node list. While on the list it can be matched, thus requires validation checks on state to be added in numerous code paths. If the node comes back, its possible for there to be multiple node structures for the same device on the vport node list. There is no reason to keep the node structure around after it is no longer in existence, and the current implementation creates problems for itself (multiple nodes) and lots of unnecessary code for state validation. Additionally, the reference taking on the node structure didn't follow the normal model used by the kernel kref api. It included lots of odd logic to match state with reference count. The combination of this odd logic plus the way it was implicitly used in the discovery engine made its reference taking implementation suspect and extremely hard to follow. Change the driver such that the reference taking routines are now normal ref increments/decrements and callout on refcount=0. With this in place, the rework can be done such that the node structure is fully removed and deallocated when the remote port no longer exists and all references are removed. This removal logic, and the basic ref counting are intrically tied, thus in a single patch. Link: https://lore.kernel.org/r/20201115192646.12977-2-james.smart@broadcom.com Co-developed-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <james.smart@broadcom.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-11-15 19:26:30 +00:00
"flg:x%x refcnt:%d\n",
scsi: lpfc: Fix device recovery errors after PLOGI failures When target-side fault injections are made, the driver isn't reconnecting to the remote port. The driver is logging "2753" error messages which state: "PLOGI failure DID:1B2400 Status:x3/xf0240008" The failures status is indicating a Illegal field error, which points to the Temporary RPI field being used for the ELS. This error typically means the driver used an RPI that was already registered (shouldn't be registered if using it in this context). Study has found that if the driver were in discovery attempts and encountered an error, it wouldn't flag the temporary rpi in error. Yet the rpi was released for reallocation in these error paths and another ELS could allocate the rpi. In the failure situation a retry was done on an ELS that had encountered an error, and as the rpi wasn't marked in error, the ELS reused the rpi it originally allocated. But that rpi had been allocated by a different ELS issued after the original error and before the retry attempt. The different ELS had succeeded and the RPI was registered. Fix by marking the rpi state for the node to be in error, aka as needing reallocation, upon an error in the els processing. Error state marking is always done prior to release back to the internal rpi free list, which the driver wasn't doing in cases prior. Also enhanced some of the logging to help in the next case of problem troubleshooting. Link: https://lore.kernel.org/r/20190922035906.10977-7-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2019-09-22 03:58:52 +00:00
ndlp, ndlp->nlp_rpi, ndlp->nlp_DID,
scsi: lpfc: Rework remote port ref counting and node freeing When a remote port is disconnected and disappears, its node structure (ndlp) stays allocated and on a vport node list. While on the list it can be matched, thus requires validation checks on state to be added in numerous code paths. If the node comes back, its possible for there to be multiple node structures for the same device on the vport node list. There is no reason to keep the node structure around after it is no longer in existence, and the current implementation creates problems for itself (multiple nodes) and lots of unnecessary code for state validation. Additionally, the reference taking on the node structure didn't follow the normal model used by the kernel kref api. It included lots of odd logic to match state with reference count. The combination of this odd logic plus the way it was implicitly used in the discovery engine made its reference taking implementation suspect and extremely hard to follow. Change the driver such that the reference taking routines are now normal ref increments/decrements and callout on refcount=0. With this in place, the rework can be done such that the node structure is fully removed and deallocated when the remote port no longer exists and all references are removed. This removal logic, and the basic ref counting are intrically tied, thus in a single patch. Link: https://lore.kernel.org/r/20201115192646.12977-2-james.smart@broadcom.com Co-developed-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <james.smart@broadcom.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-11-15 19:26:30 +00:00
ndlp->nlp_flag, kref_read(&ndlp->kref));
ndlp->active_rrqs_xri_bitmap =
mempool_alloc(vport->phba->active_rrq_pool,
GFP_KERNEL);
if (ndlp->active_rrqs_xri_bitmap)
memset(ndlp->active_rrqs_xri_bitmap, 0,
ndlp->phba->cfg_rrq_xri_bitmap_sz);
}
lpfc_debugfs_disc_trc(vport, LPFC_DISC_TRC_NODE,
"node init: did:x%x",
ndlp->nlp_DID, 0, 0);
return ndlp;
}
/* This routine releases all resources associated with a specifc NPort's ndlp
* and mempool_free's the nodelist.
*/
static void
lpfc_nlp_release(struct kref *kref)
{
struct lpfc_nodelist *ndlp = container_of(kref, struct lpfc_nodelist,
kref);
struct lpfc_vport *vport = ndlp->vport;
lpfc_debugfs_disc_trc(ndlp->vport, LPFC_DISC_TRC_NODE,
"node release: did:x%x flg:x%x type:x%x",
ndlp->nlp_DID, ndlp->nlp_flag, ndlp->nlp_type);
lpfc_printf_vlog(vport, KERN_INFO, LOG_NODE,
"0279 %s: ndlp: x%px did %x refcnt:%d rpi:%x\n",
__func__, ndlp, ndlp->nlp_DID,
kref_read(&ndlp->kref), ndlp->nlp_rpi);
/* remove ndlp from action. */
lpfc_cancel_retry_delay_tmo(vport, ndlp);
lpfc_cleanup_node(vport, ndlp);
scsi: lpfc: Fix unreleased RPIs when NPIV ports are created While testing NPIV and watching logins and used RPI levels, it was seen the used RPI count was much higher than the number of remote ports discovered. Code inspection showed that remote port removals on any NPIV instance are releasing the RPI, but not performing an UNREG_RPI with the adapter thus the reference counting never fully drops and the RPI is never fully released. This was happening on NPIV nodes due to a log of fabric ELS's to fabric addresses. This lack of UNREG_RPI was introduced by a prior node rework patch that performed the UNREG_RPI as part of node cleanup. To resolve the issue, do the following: - Restore the RPI release code, but move the location to so that it is in line with the new node cleanup design. - NPIV ports now release the RPI and drop the node when the caller sets the NLP_RELEASE_RPI flag. - Set the NLP_RELEASE_RPI flag in node cleanup which will trigger a release of RPI to free pool. - Ensure there's an UNREG_RPI at LOGO completion so that RPI release is completed. - Stop offline_prep from skipping nodes that are UNUSED. The RPI may not have been released. - Stop the default RPI handling in lpfc_cmpl_els_rsp() for SLI4. - Fixed up debugfs RPI displays for better debugging. Fixes: a70e63eee1c1 ("scsi: lpfc: Fix NPIV Fabric Node reference counting") Link: https://lore.kernel.org/r/20210514195559.119853-2-jsmart2021@gmail.com Cc: <stable@vger.kernel.org> # v5.11+ Co-developed-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2021-05-14 19:55:49 +00:00
/* Not all ELS transactions have registered the RPI with the port.
* In these cases the rpi usage is temporary and the node is
* released when the WQE is completed. Catch this case to free the
* RPI to the pool. Because this node is in the release path, a lock
* is unnecessary. All references are gone and the node has been
* dequeued.
*/
if (ndlp->nlp_flag & NLP_RELEASE_RPI) {
if (ndlp->nlp_rpi != LPFC_RPI_ALLOC_ERROR &&
!(ndlp->nlp_flag & (NLP_RPI_REGISTERED | NLP_UNREG_INP))) {
lpfc_sli4_free_rpi(vport->phba, ndlp->nlp_rpi);
ndlp->nlp_rpi = LPFC_RPI_ALLOC_ERROR;
}
}
/* The node is not freed back to memory, it is released to a pool so
* the node fields need to be cleaned up.
*/
ndlp->vport = NULL;
ndlp->nlp_state = NLP_STE_FREED_NODE;
ndlp->nlp_flag = 0;
ndlp->fc4_xpt_flags = 0;
/* free ndlp memory for final ndlp release */
scsi: lpfc: Rework remote port ref counting and node freeing When a remote port is disconnected and disappears, its node structure (ndlp) stays allocated and on a vport node list. While on the list it can be matched, thus requires validation checks on state to be added in numerous code paths. If the node comes back, its possible for there to be multiple node structures for the same device on the vport node list. There is no reason to keep the node structure around after it is no longer in existence, and the current implementation creates problems for itself (multiple nodes) and lots of unnecessary code for state validation. Additionally, the reference taking on the node structure didn't follow the normal model used by the kernel kref api. It included lots of odd logic to match state with reference count. The combination of this odd logic plus the way it was implicitly used in the discovery engine made its reference taking implementation suspect and extremely hard to follow. Change the driver such that the reference taking routines are now normal ref increments/decrements and callout on refcount=0. With this in place, the rework can be done such that the node structure is fully removed and deallocated when the remote port no longer exists and all references are removed. This removal logic, and the basic ref counting are intrically tied, thus in a single patch. Link: https://lore.kernel.org/r/20201115192646.12977-2-james.smart@broadcom.com Co-developed-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <james.smart@broadcom.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-11-15 19:26:30 +00:00
if (ndlp->phba->sli_rev == LPFC_SLI_REV4)
mempool_free(ndlp->active_rrqs_xri_bitmap,
ndlp->phba->active_rrq_pool);
mempool_free(ndlp, ndlp->phba->nlp_mem_pool);
}
/* This routine bumps the reference count for a ndlp structure to ensure
* that one discovery thread won't free a ndlp while another discovery thread
* is using it.
*/
struct lpfc_nodelist *
lpfc_nlp_get(struct lpfc_nodelist *ndlp)
{
unsigned long flags;
if (ndlp) {
lpfc_debugfs_disc_trc(ndlp->vport, LPFC_DISC_TRC_NODE,
"node get: did:x%x flg:x%x refcnt:x%x",
ndlp->nlp_DID, ndlp->nlp_flag,
kref_read(&ndlp->kref));
scsi: lpfc: Rework remote port ref counting and node freeing When a remote port is disconnected and disappears, its node structure (ndlp) stays allocated and on a vport node list. While on the list it can be matched, thus requires validation checks on state to be added in numerous code paths. If the node comes back, its possible for there to be multiple node structures for the same device on the vport node list. There is no reason to keep the node structure around after it is no longer in existence, and the current implementation creates problems for itself (multiple nodes) and lots of unnecessary code for state validation. Additionally, the reference taking on the node structure didn't follow the normal model used by the kernel kref api. It included lots of odd logic to match state with reference count. The combination of this odd logic plus the way it was implicitly used in the discovery engine made its reference taking implementation suspect and extremely hard to follow. Change the driver such that the reference taking routines are now normal ref increments/decrements and callout on refcount=0. With this in place, the rework can be done such that the node structure is fully removed and deallocated when the remote port no longer exists and all references are removed. This removal logic, and the basic ref counting are intrically tied, thus in a single patch. Link: https://lore.kernel.org/r/20201115192646.12977-2-james.smart@broadcom.com Co-developed-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <james.smart@broadcom.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-11-15 19:26:30 +00:00
/* The check of ndlp usage to prevent incrementing the
* ndlp reference count that is in the process of being
* released.
*/
spin_lock_irqsave(&ndlp->lock, flags);
scsi: lpfc: Rework remote port ref counting and node freeing When a remote port is disconnected and disappears, its node structure (ndlp) stays allocated and on a vport node list. While on the list it can be matched, thus requires validation checks on state to be added in numerous code paths. If the node comes back, its possible for there to be multiple node structures for the same device on the vport node list. There is no reason to keep the node structure around after it is no longer in existence, and the current implementation creates problems for itself (multiple nodes) and lots of unnecessary code for state validation. Additionally, the reference taking on the node structure didn't follow the normal model used by the kernel kref api. It included lots of odd logic to match state with reference count. The combination of this odd logic plus the way it was implicitly used in the discovery engine made its reference taking implementation suspect and extremely hard to follow. Change the driver such that the reference taking routines are now normal ref increments/decrements and callout on refcount=0. With this in place, the rework can be done such that the node structure is fully removed and deallocated when the remote port no longer exists and all references are removed. This removal logic, and the basic ref counting are intrically tied, thus in a single patch. Link: https://lore.kernel.org/r/20201115192646.12977-2-james.smart@broadcom.com Co-developed-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <james.smart@broadcom.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-11-15 19:26:30 +00:00
if (!kref_get_unless_zero(&ndlp->kref)) {
spin_unlock_irqrestore(&ndlp->lock, flags);
lpfc_printf_vlog(ndlp->vport, KERN_WARNING, LOG_NODE,
scsi: lpfc: Rework remote port ref counting and node freeing When a remote port is disconnected and disappears, its node structure (ndlp) stays allocated and on a vport node list. While on the list it can be matched, thus requires validation checks on state to be added in numerous code paths. If the node comes back, its possible for there to be multiple node structures for the same device on the vport node list. There is no reason to keep the node structure around after it is no longer in existence, and the current implementation creates problems for itself (multiple nodes) and lots of unnecessary code for state validation. Additionally, the reference taking on the node structure didn't follow the normal model used by the kernel kref api. It included lots of odd logic to match state with reference count. The combination of this odd logic plus the way it was implicitly used in the discovery engine made its reference taking implementation suspect and extremely hard to follow. Change the driver such that the reference taking routines are now normal ref increments/decrements and callout on refcount=0. With this in place, the rework can be done such that the node structure is fully removed and deallocated when the remote port no longer exists and all references are removed. This removal logic, and the basic ref counting are intrically tied, thus in a single patch. Link: https://lore.kernel.org/r/20201115192646.12977-2-james.smart@broadcom.com Co-developed-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <james.smart@broadcom.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-11-15 19:26:30 +00:00
"0276 %s: ndlp:x%px refcnt:%d\n",
__func__, (void *)ndlp, kref_read(&ndlp->kref));
return NULL;
scsi: lpfc: Rework remote port ref counting and node freeing When a remote port is disconnected and disappears, its node structure (ndlp) stays allocated and on a vport node list. While on the list it can be matched, thus requires validation checks on state to be added in numerous code paths. If the node comes back, its possible for there to be multiple node structures for the same device on the vport node list. There is no reason to keep the node structure around after it is no longer in existence, and the current implementation creates problems for itself (multiple nodes) and lots of unnecessary code for state validation. Additionally, the reference taking on the node structure didn't follow the normal model used by the kernel kref api. It included lots of odd logic to match state with reference count. The combination of this odd logic plus the way it was implicitly used in the discovery engine made its reference taking implementation suspect and extremely hard to follow. Change the driver such that the reference taking routines are now normal ref increments/decrements and callout on refcount=0. With this in place, the rework can be done such that the node structure is fully removed and deallocated when the remote port no longer exists and all references are removed. This removal logic, and the basic ref counting are intrically tied, thus in a single patch. Link: https://lore.kernel.org/r/20201115192646.12977-2-james.smart@broadcom.com Co-developed-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <james.smart@broadcom.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-11-15 19:26:30 +00:00
}
spin_unlock_irqrestore(&ndlp->lock, flags);
scsi: lpfc: Rework remote port ref counting and node freeing When a remote port is disconnected and disappears, its node structure (ndlp) stays allocated and on a vport node list. While on the list it can be matched, thus requires validation checks on state to be added in numerous code paths. If the node comes back, its possible for there to be multiple node structures for the same device on the vport node list. There is no reason to keep the node structure around after it is no longer in existence, and the current implementation creates problems for itself (multiple nodes) and lots of unnecessary code for state validation. Additionally, the reference taking on the node structure didn't follow the normal model used by the kernel kref api. It included lots of odd logic to match state with reference count. The combination of this odd logic plus the way it was implicitly used in the discovery engine made its reference taking implementation suspect and extremely hard to follow. Change the driver such that the reference taking routines are now normal ref increments/decrements and callout on refcount=0. With this in place, the rework can be done such that the node structure is fully removed and deallocated when the remote port no longer exists and all references are removed. This removal logic, and the basic ref counting are intrically tied, thus in a single patch. Link: https://lore.kernel.org/r/20201115192646.12977-2-james.smart@broadcom.com Co-developed-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <james.smart@broadcom.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-11-15 19:26:30 +00:00
} else {
WARN_ONCE(!ndlp, "**** %s, get ref on NULL ndlp!", __func__);
}
scsi: lpfc: Rework remote port ref counting and node freeing When a remote port is disconnected and disappears, its node structure (ndlp) stays allocated and on a vport node list. While on the list it can be matched, thus requires validation checks on state to be added in numerous code paths. If the node comes back, its possible for there to be multiple node structures for the same device on the vport node list. There is no reason to keep the node structure around after it is no longer in existence, and the current implementation creates problems for itself (multiple nodes) and lots of unnecessary code for state validation. Additionally, the reference taking on the node structure didn't follow the normal model used by the kernel kref api. It included lots of odd logic to match state with reference count. The combination of this odd logic plus the way it was implicitly used in the discovery engine made its reference taking implementation suspect and extremely hard to follow. Change the driver such that the reference taking routines are now normal ref increments/decrements and callout on refcount=0. With this in place, the rework can be done such that the node structure is fully removed and deallocated when the remote port no longer exists and all references are removed. This removal logic, and the basic ref counting are intrically tied, thus in a single patch. Link: https://lore.kernel.org/r/20201115192646.12977-2-james.smart@broadcom.com Co-developed-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <james.smart@broadcom.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-11-15 19:26:30 +00:00
return ndlp;
}
/* This routine decrements the reference count for a ndlp structure. If the
scsi: lpfc: Rework remote port ref counting and node freeing When a remote port is disconnected and disappears, its node structure (ndlp) stays allocated and on a vport node list. While on the list it can be matched, thus requires validation checks on state to be added in numerous code paths. If the node comes back, its possible for there to be multiple node structures for the same device on the vport node list. There is no reason to keep the node structure around after it is no longer in existence, and the current implementation creates problems for itself (multiple nodes) and lots of unnecessary code for state validation. Additionally, the reference taking on the node structure didn't follow the normal model used by the kernel kref api. It included lots of odd logic to match state with reference count. The combination of this odd logic plus the way it was implicitly used in the discovery engine made its reference taking implementation suspect and extremely hard to follow. Change the driver such that the reference taking routines are now normal ref increments/decrements and callout on refcount=0. With this in place, the rework can be done such that the node structure is fully removed and deallocated when the remote port no longer exists and all references are removed. This removal logic, and the basic ref counting are intrically tied, thus in a single patch. Link: https://lore.kernel.org/r/20201115192646.12977-2-james.smart@broadcom.com Co-developed-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <james.smart@broadcom.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-11-15 19:26:30 +00:00
* count goes to 0, this indicates the associated nodelist should be freed.
*/
int
lpfc_nlp_put(struct lpfc_nodelist *ndlp)
{
scsi: lpfc: Rework remote port ref counting and node freeing When a remote port is disconnected and disappears, its node structure (ndlp) stays allocated and on a vport node list. While on the list it can be matched, thus requires validation checks on state to be added in numerous code paths. If the node comes back, its possible for there to be multiple node structures for the same device on the vport node list. There is no reason to keep the node structure around after it is no longer in existence, and the current implementation creates problems for itself (multiple nodes) and lots of unnecessary code for state validation. Additionally, the reference taking on the node structure didn't follow the normal model used by the kernel kref api. It included lots of odd logic to match state with reference count. The combination of this odd logic plus the way it was implicitly used in the discovery engine made its reference taking implementation suspect and extremely hard to follow. Change the driver such that the reference taking routines are now normal ref increments/decrements and callout on refcount=0. With this in place, the rework can be done such that the node structure is fully removed and deallocated when the remote port no longer exists and all references are removed. This removal logic, and the basic ref counting are intrically tied, thus in a single patch. Link: https://lore.kernel.org/r/20201115192646.12977-2-james.smart@broadcom.com Co-developed-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <james.smart@broadcom.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-11-15 19:26:30 +00:00
if (ndlp) {
lpfc_debugfs_disc_trc(ndlp->vport, LPFC_DISC_TRC_NODE,
"node put: did:x%x flg:x%x refcnt:x%x",
ndlp->nlp_DID, ndlp->nlp_flag,
kref_read(&ndlp->kref));
scsi: lpfc: Rework remote port ref counting and node freeing When a remote port is disconnected and disappears, its node structure (ndlp) stays allocated and on a vport node list. While on the list it can be matched, thus requires validation checks on state to be added in numerous code paths. If the node comes back, its possible for there to be multiple node structures for the same device on the vport node list. There is no reason to keep the node structure around after it is no longer in existence, and the current implementation creates problems for itself (multiple nodes) and lots of unnecessary code for state validation. Additionally, the reference taking on the node structure didn't follow the normal model used by the kernel kref api. It included lots of odd logic to match state with reference count. The combination of this odd logic plus the way it was implicitly used in the discovery engine made its reference taking implementation suspect and extremely hard to follow. Change the driver such that the reference taking routines are now normal ref increments/decrements and callout on refcount=0. With this in place, the rework can be done such that the node structure is fully removed and deallocated when the remote port no longer exists and all references are removed. This removal logic, and the basic ref counting are intrically tied, thus in a single patch. Link: https://lore.kernel.org/r/20201115192646.12977-2-james.smart@broadcom.com Co-developed-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <james.smart@broadcom.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-11-15 19:26:30 +00:00
} else {
WARN_ONCE(!ndlp, "**** %s, put ref on NULL ndlp!", __func__);
}
scsi: lpfc: Rework remote port ref counting and node freeing When a remote port is disconnected and disappears, its node structure (ndlp) stays allocated and on a vport node list. While on the list it can be matched, thus requires validation checks on state to be added in numerous code paths. If the node comes back, its possible for there to be multiple node structures for the same device on the vport node list. There is no reason to keep the node structure around after it is no longer in existence, and the current implementation creates problems for itself (multiple nodes) and lots of unnecessary code for state validation. Additionally, the reference taking on the node structure didn't follow the normal model used by the kernel kref api. It included lots of odd logic to match state with reference count. The combination of this odd logic plus the way it was implicitly used in the discovery engine made its reference taking implementation suspect and extremely hard to follow. Change the driver such that the reference taking routines are now normal ref increments/decrements and callout on refcount=0. With this in place, the rework can be done such that the node structure is fully removed and deallocated when the remote port no longer exists and all references are removed. This removal logic, and the basic ref counting are intrically tied, thus in a single patch. Link: https://lore.kernel.org/r/20201115192646.12977-2-james.smart@broadcom.com Co-developed-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <james.smart@broadcom.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-11-15 19:26:30 +00:00
return ndlp ? kref_put(&ndlp->kref, lpfc_nlp_release) : 0;
}
/* This routine free's the specified nodelist if it is not in use
* by any other discovery thread. This routine returns 1 if the
* ndlp has been freed. A return value of 0 indicates the ndlp is
* not yet been released.
*/
int
lpfc_nlp_not_used(struct lpfc_nodelist *ndlp)
{
lpfc_debugfs_disc_trc(ndlp->vport, LPFC_DISC_TRC_NODE,
"node not used: did:x%x flg:x%x refcnt:x%x",
ndlp->nlp_DID, ndlp->nlp_flag,
kref_read(&ndlp->kref));
scsi: lpfc: Fix unreleased RPIs when NPIV ports are created While testing NPIV and watching logins and used RPI levels, it was seen the used RPI count was much higher than the number of remote ports discovered. Code inspection showed that remote port removals on any NPIV instance are releasing the RPI, but not performing an UNREG_RPI with the adapter thus the reference counting never fully drops and the RPI is never fully released. This was happening on NPIV nodes due to a log of fabric ELS's to fabric addresses. This lack of UNREG_RPI was introduced by a prior node rework patch that performed the UNREG_RPI as part of node cleanup. To resolve the issue, do the following: - Restore the RPI release code, but move the location to so that it is in line with the new node cleanup design. - NPIV ports now release the RPI and drop the node when the caller sets the NLP_RELEASE_RPI flag. - Set the NLP_RELEASE_RPI flag in node cleanup which will trigger a release of RPI to free pool. - Ensure there's an UNREG_RPI at LOGO completion so that RPI release is completed. - Stop offline_prep from skipping nodes that are UNUSED. The RPI may not have been released. - Stop the default RPI handling in lpfc_cmpl_els_rsp() for SLI4. - Fixed up debugfs RPI displays for better debugging. Fixes: a70e63eee1c1 ("scsi: lpfc: Fix NPIV Fabric Node reference counting") Link: https://lore.kernel.org/r/20210514195559.119853-2-jsmart2021@gmail.com Cc: <stable@vger.kernel.org> # v5.11+ Co-developed-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: Justin Tee <justin.tee@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2021-05-14 19:55:49 +00:00
if (kref_read(&ndlp->kref) == 1)
if (lpfc_nlp_put(ndlp))
return 1;
return 0;
}
/**
* lpfc_fcf_inuse - Check if FCF can be unregistered.
* @phba: Pointer to hba context object.
*
* This function iterate through all FC nodes associated
* will all vports to check if there is any node with
* fc_rports associated with it. If there is an fc_rport
* associated with the node, then the node is either in
* discovered state or its devloss_timer is pending.
*/
static int
lpfc_fcf_inuse(struct lpfc_hba *phba)
{
struct lpfc_vport **vports;
int i, ret = 0;
struct lpfc_nodelist *ndlp;
struct Scsi_Host *shost;
vports = lpfc_create_vport_work_array(phba);
/* If driver cannot allocate memory, indicate fcf is in use */
if (!vports)
return 1;
for (i = 0; i <= phba->max_vports && vports[i] != NULL; i++) {
shost = lpfc_shost_from_vport(vports[i]);
spin_lock_irq(shost->host_lock);
/*
* IF the CVL_RCVD bit is not set then we have sent the
* flogi.
* If dev_loss fires while we are waiting we do not want to
* unreg the fcf.
*/
if (!(vports[i]->fc_flag & FC_VPORT_CVL_RCVD)) {
spin_unlock_irq(shost->host_lock);
ret = 1;
goto out;
}
list_for_each_entry(ndlp, &vports[i]->fc_nodes, nlp_listp) {
scsi: lpfc: Rework remote port ref counting and node freeing When a remote port is disconnected and disappears, its node structure (ndlp) stays allocated and on a vport node list. While on the list it can be matched, thus requires validation checks on state to be added in numerous code paths. If the node comes back, its possible for there to be multiple node structures for the same device on the vport node list. There is no reason to keep the node structure around after it is no longer in existence, and the current implementation creates problems for itself (multiple nodes) and lots of unnecessary code for state validation. Additionally, the reference taking on the node structure didn't follow the normal model used by the kernel kref api. It included lots of odd logic to match state with reference count. The combination of this odd logic plus the way it was implicitly used in the discovery engine made its reference taking implementation suspect and extremely hard to follow. Change the driver such that the reference taking routines are now normal ref increments/decrements and callout on refcount=0. With this in place, the rework can be done such that the node structure is fully removed and deallocated when the remote port no longer exists and all references are removed. This removal logic, and the basic ref counting are intrically tied, thus in a single patch. Link: https://lore.kernel.org/r/20201115192646.12977-2-james.smart@broadcom.com Co-developed-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <james.smart@broadcom.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-11-15 19:26:30 +00:00
if (ndlp->rport &&
(ndlp->rport->roles & FC_RPORT_ROLE_FCP_TARGET)) {
ret = 1;
spin_unlock_irq(shost->host_lock);
goto out;
} else if (ndlp->nlp_flag & NLP_RPI_REGISTERED) {
ret = 1;
scsi: lpfc: Fix device recovery errors after PLOGI failures When target-side fault injections are made, the driver isn't reconnecting to the remote port. The driver is logging "2753" error messages which state: "PLOGI failure DID:1B2400 Status:x3/xf0240008" The failures status is indicating a Illegal field error, which points to the Temporary RPI field being used for the ELS. This error typically means the driver used an RPI that was already registered (shouldn't be registered if using it in this context). Study has found that if the driver were in discovery attempts and encountered an error, it wouldn't flag the temporary rpi in error. Yet the rpi was released for reallocation in these error paths and another ELS could allocate the rpi. In the failure situation a retry was done on an ELS that had encountered an error, and as the rpi wasn't marked in error, the ELS reused the rpi it originally allocated. But that rpi had been allocated by a different ELS issued after the original error and before the retry attempt. The different ELS had succeeded and the RPI was registered. Fix by marking the rpi state for the node to be in error, aka as needing reallocation, upon an error in the els processing. Error state marking is always done prior to release back to the internal rpi free list, which the driver wasn't doing in cases prior. Also enhanced some of the logging to help in the next case of problem troubleshooting. Link: https://lore.kernel.org/r/20190922035906.10977-7-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2019-09-22 03:58:52 +00:00
lpfc_printf_log(phba, KERN_INFO,
LOG_NODE | LOG_DISCOVERY,
"2624 RPI %x DID %x flag %x "
"still logged in\n",
ndlp->nlp_rpi, ndlp->nlp_DID,
ndlp->nlp_flag);
}
}
spin_unlock_irq(shost->host_lock);
}
out:
lpfc_destroy_vport_work_array(phba, vports);
return ret;
}
/**
* lpfc_unregister_vfi_cmpl - Completion handler for unreg vfi.
* @phba: Pointer to hba context object.
* @mboxq: Pointer to mailbox object.
*
* This function frees memory associated with the mailbox command.
*/
void
lpfc_unregister_vfi_cmpl(struct lpfc_hba *phba, LPFC_MBOXQ_t *mboxq)
{
struct lpfc_vport *vport = mboxq->vport;
struct Scsi_Host *shost = lpfc_shost_from_vport(vport);
if (mboxq->u.mb.mbxStatus) {
scsi: lpfc: Add an internal trace log buffer The current logging methods typically end up requesting a reproduction with a different logging level set to figure out what happened. This was mainly by design to not clutter the kernel log messages with things that were typically not interesting and the messages themselves could cause other issues. When looking to make a better system, it was seen that in many cases when more data was wanted was when another message, usually at KERN_ERR level, was logged. And in most cases, what the additional logging that was then enabled was typically. Most of these areas fell into the discovery machine. Based on this summary, the following design has been put in place: The driver will maintain an internal log (256 elements of 256 bytes). The "additional logging" messages that are usually enabled in a reproduction will be changed to now log all the time to the internal log. A new logging level is defined - LOG_TRACE_EVENT. When this level is set (it is not by default) and a message marked as KERN_ERR is logged, all the messages in the internal log will be dumped to the kernel log before the KERN_ERR message is logged. There is a timestamp on each message added to the internal log. However, this timestamp is not converted to wall time when logged. The value of the timestamp is solely to give a crude time reference for the messages. Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-06-30 21:50:00 +00:00
lpfc_printf_log(phba, KERN_ERR, LOG_TRACE_EVENT,
"2555 UNREG_VFI mbxStatus error x%x "
"HBA state x%x\n",
mboxq->u.mb.mbxStatus, vport->port_state);
}
spin_lock_irq(shost->host_lock);
phba->pport->fc_flag &= ~FC_VFI_REGISTERED;
spin_unlock_irq(shost->host_lock);
mempool_free(mboxq, phba->mbox_mem_pool);
return;
}
/**
* lpfc_unregister_fcfi_cmpl - Completion handler for unreg fcfi.
* @phba: Pointer to hba context object.
* @mboxq: Pointer to mailbox object.
*
* This function frees memory associated with the mailbox command.
*/
static void
lpfc_unregister_fcfi_cmpl(struct lpfc_hba *phba, LPFC_MBOXQ_t *mboxq)
{
struct lpfc_vport *vport = mboxq->vport;
if (mboxq->u.mb.mbxStatus) {
scsi: lpfc: Add an internal trace log buffer The current logging methods typically end up requesting a reproduction with a different logging level set to figure out what happened. This was mainly by design to not clutter the kernel log messages with things that were typically not interesting and the messages themselves could cause other issues. When looking to make a better system, it was seen that in many cases when more data was wanted was when another message, usually at KERN_ERR level, was logged. And in most cases, what the additional logging that was then enabled was typically. Most of these areas fell into the discovery machine. Based on this summary, the following design has been put in place: The driver will maintain an internal log (256 elements of 256 bytes). The "additional logging" messages that are usually enabled in a reproduction will be changed to now log all the time to the internal log. A new logging level is defined - LOG_TRACE_EVENT. When this level is set (it is not by default) and a message marked as KERN_ERR is logged, all the messages in the internal log will be dumped to the kernel log before the KERN_ERR message is logged. There is a timestamp on each message added to the internal log. However, this timestamp is not converted to wall time when logged. The value of the timestamp is solely to give a crude time reference for the messages. Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-06-30 21:50:00 +00:00
lpfc_printf_log(phba, KERN_ERR, LOG_TRACE_EVENT,
"2550 UNREG_FCFI mbxStatus error x%x "
"HBA state x%x\n",
mboxq->u.mb.mbxStatus, vport->port_state);
}
mempool_free(mboxq, phba->mbox_mem_pool);
return;
}
/**
* lpfc_unregister_fcf_prep - Unregister fcf record preparation
* @phba: Pointer to hba context object.
*
* This function prepare the HBA for unregistering the currently registered
* FCF from the HBA. It performs unregistering, in order, RPIs, VPIs, and
* VFIs.
*/
int
lpfc_unregister_fcf_prep(struct lpfc_hba *phba)
{
struct lpfc_vport **vports;
struct lpfc_nodelist *ndlp;
struct Scsi_Host *shost;
int i = 0, rc;
/* Unregister RPIs */
if (lpfc_fcf_inuse(phba))
lpfc_unreg_hba_rpis(phba);
/* At this point, all discovery is aborted */
phba->pport->port_state = LPFC_VPORT_UNKNOWN;
/* Unregister VPIs */
vports = lpfc_create_vport_work_array(phba);
if (vports && (phba->sli3_options & LPFC_SLI3_NPIV_ENABLED))
for (i = 0; i <= phba->max_vports && vports[i] != NULL; i++) {
/* Stop FLOGI/FDISC retries */
ndlp = lpfc_findnode_did(vports[i], Fabric_DID);
if (ndlp)
lpfc_cancel_retry_delay_tmo(vports[i], ndlp);
lpfc_cleanup_pending_mbox(vports[i]);
if (phba->sli_rev == LPFC_SLI_REV4)
lpfc_sli4_unreg_all_rpis(vports[i]);
lpfc_mbx_unreg_vpi(vports[i]);
shost = lpfc_shost_from_vport(vports[i]);
spin_lock_irq(shost->host_lock);
vports[i]->fc_flag |= FC_VPORT_NEEDS_INIT_VPI;
vports[i]->vpi_state &= ~LPFC_VPI_REGISTERED;
spin_unlock_irq(shost->host_lock);
}
lpfc_destroy_vport_work_array(phba, vports);
if (i == 0 && (!(phba->sli3_options & LPFC_SLI3_NPIV_ENABLED))) {
ndlp = lpfc_findnode_did(phba->pport, Fabric_DID);
if (ndlp)
lpfc_cancel_retry_delay_tmo(phba->pport, ndlp);
lpfc_cleanup_pending_mbox(phba->pport);
if (phba->sli_rev == LPFC_SLI_REV4)
lpfc_sli4_unreg_all_rpis(phba->pport);
lpfc_mbx_unreg_vpi(phba->pport);
shost = lpfc_shost_from_vport(phba->pport);
spin_lock_irq(shost->host_lock);
phba->pport->fc_flag |= FC_VPORT_NEEDS_INIT_VPI;
phba->pport->vpi_state &= ~LPFC_VPI_REGISTERED;
spin_unlock_irq(shost->host_lock);
}
/* Cleanup any outstanding ELS commands */
lpfc_els_flush_all_cmd(phba);
/* Unregister the physical port VFI */
rc = lpfc_issue_unreg_vfi(phba->pport);
return rc;
}
/**
* lpfc_sli4_unregister_fcf - Unregister currently registered FCF record
* @phba: Pointer to hba context object.
*
* This function issues synchronous unregister FCF mailbox command to HBA to
* unregister the currently registered FCF record. The driver does not reset
* the driver FCF usage state flags.
*
* Return 0 if successfully issued, none-zero otherwise.
*/
int
lpfc_sli4_unregister_fcf(struct lpfc_hba *phba)
{
LPFC_MBOXQ_t *mbox;
int rc;
mbox = mempool_alloc(phba->mbox_mem_pool, GFP_KERNEL);
if (!mbox) {
scsi: lpfc: Add an internal trace log buffer The current logging methods typically end up requesting a reproduction with a different logging level set to figure out what happened. This was mainly by design to not clutter the kernel log messages with things that were typically not interesting and the messages themselves could cause other issues. When looking to make a better system, it was seen that in many cases when more data was wanted was when another message, usually at KERN_ERR level, was logged. And in most cases, what the additional logging that was then enabled was typically. Most of these areas fell into the discovery machine. Based on this summary, the following design has been put in place: The driver will maintain an internal log (256 elements of 256 bytes). The "additional logging" messages that are usually enabled in a reproduction will be changed to now log all the time to the internal log. A new logging level is defined - LOG_TRACE_EVENT. When this level is set (it is not by default) and a message marked as KERN_ERR is logged, all the messages in the internal log will be dumped to the kernel log before the KERN_ERR message is logged. There is a timestamp on each message added to the internal log. However, this timestamp is not converted to wall time when logged. The value of the timestamp is solely to give a crude time reference for the messages. Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-06-30 21:50:00 +00:00
lpfc_printf_log(phba, KERN_ERR, LOG_TRACE_EVENT,
"2551 UNREG_FCFI mbox allocation failed"
"HBA state x%x\n", phba->pport->port_state);
return -ENOMEM;
}
lpfc_unreg_fcfi(mbox, phba->fcf.fcfi);
mbox->vport = phba->pport;
mbox->mbox_cmpl = lpfc_unregister_fcfi_cmpl;
rc = lpfc_sli_issue_mbox(phba, mbox, MBX_NOWAIT);
if (rc == MBX_NOT_FINISHED) {
scsi: lpfc: Add an internal trace log buffer The current logging methods typically end up requesting a reproduction with a different logging level set to figure out what happened. This was mainly by design to not clutter the kernel log messages with things that were typically not interesting and the messages themselves could cause other issues. When looking to make a better system, it was seen that in many cases when more data was wanted was when another message, usually at KERN_ERR level, was logged. And in most cases, what the additional logging that was then enabled was typically. Most of these areas fell into the discovery machine. Based on this summary, the following design has been put in place: The driver will maintain an internal log (256 elements of 256 bytes). The "additional logging" messages that are usually enabled in a reproduction will be changed to now log all the time to the internal log. A new logging level is defined - LOG_TRACE_EVENT. When this level is set (it is not by default) and a message marked as KERN_ERR is logged, all the messages in the internal log will be dumped to the kernel log before the KERN_ERR message is logged. There is a timestamp on each message added to the internal log. However, this timestamp is not converted to wall time when logged. The value of the timestamp is solely to give a crude time reference for the messages. Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-06-30 21:50:00 +00:00
lpfc_printf_log(phba, KERN_ERR, LOG_TRACE_EVENT,
"2552 Unregister FCFI command failed rc x%x "
"HBA state x%x\n",
rc, phba->pport->port_state);
return -EINVAL;
}
return 0;
}
/**
* lpfc_unregister_fcf_rescan - Unregister currently registered fcf and rescan
* @phba: Pointer to hba context object.
*
* This function unregisters the currently reigstered FCF. This function
* also tries to find another FCF for discovery by rescan the HBA FCF table.
*/
void
lpfc_unregister_fcf_rescan(struct lpfc_hba *phba)
{
int rc;
/* Preparation for unregistering fcf */
rc = lpfc_unregister_fcf_prep(phba);
if (rc) {
scsi: lpfc: Add an internal trace log buffer The current logging methods typically end up requesting a reproduction with a different logging level set to figure out what happened. This was mainly by design to not clutter the kernel log messages with things that were typically not interesting and the messages themselves could cause other issues. When looking to make a better system, it was seen that in many cases when more data was wanted was when another message, usually at KERN_ERR level, was logged. And in most cases, what the additional logging that was then enabled was typically. Most of these areas fell into the discovery machine. Based on this summary, the following design has been put in place: The driver will maintain an internal log (256 elements of 256 bytes). The "additional logging" messages that are usually enabled in a reproduction will be changed to now log all the time to the internal log. A new logging level is defined - LOG_TRACE_EVENT. When this level is set (it is not by default) and a message marked as KERN_ERR is logged, all the messages in the internal log will be dumped to the kernel log before the KERN_ERR message is logged. There is a timestamp on each message added to the internal log. However, this timestamp is not converted to wall time when logged. The value of the timestamp is solely to give a crude time reference for the messages. Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-06-30 21:50:00 +00:00
lpfc_printf_log(phba, KERN_ERR, LOG_TRACE_EVENT,
"2748 Failed to prepare for unregistering "
"HBA's FCF record: rc=%d\n", rc);
return;
}
/* Now, unregister FCF record and reset HBA FCF state */
rc = lpfc_sli4_unregister_fcf(phba);
if (rc)
return;
/* Reset HBA FCF states after successful unregister FCF */
phba->fcf.fcf_flag = 0;
phba->fcf.current_rec.flag = 0;
/*
* If driver is not unloading, check if there is any other
* FCF record that can be used for discovery.
*/
if ((phba->pport->load_flag & FC_UNLOADING) ||
(phba->link_state < LPFC_LINK_UP))
return;
/* This is considered as the initial FCF discovery scan */
spin_lock_irq(&phba->hbalock);
phba->fcf.fcf_flag |= FCF_INIT_DISC;
spin_unlock_irq(&phba->hbalock);
/* Reset FCF roundrobin bmask for new discovery */
lpfc_sli4_clear_fcf_rr_bmask(phba);
rc = lpfc_sli4_fcf_scan_read_fcf_rec(phba, LPFC_FCOE_FCF_GET_FIRST);
if (rc) {
spin_lock_irq(&phba->hbalock);
phba->fcf.fcf_flag &= ~FCF_INIT_DISC;
spin_unlock_irq(&phba->hbalock);
scsi: lpfc: Add an internal trace log buffer The current logging methods typically end up requesting a reproduction with a different logging level set to figure out what happened. This was mainly by design to not clutter the kernel log messages with things that were typically not interesting and the messages themselves could cause other issues. When looking to make a better system, it was seen that in many cases when more data was wanted was when another message, usually at KERN_ERR level, was logged. And in most cases, what the additional logging that was then enabled was typically. Most of these areas fell into the discovery machine. Based on this summary, the following design has been put in place: The driver will maintain an internal log (256 elements of 256 bytes). The "additional logging" messages that are usually enabled in a reproduction will be changed to now log all the time to the internal log. A new logging level is defined - LOG_TRACE_EVENT. When this level is set (it is not by default) and a message marked as KERN_ERR is logged, all the messages in the internal log will be dumped to the kernel log before the KERN_ERR message is logged. There is a timestamp on each message added to the internal log. However, this timestamp is not converted to wall time when logged. The value of the timestamp is solely to give a crude time reference for the messages. Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-06-30 21:50:00 +00:00
lpfc_printf_log(phba, KERN_ERR, LOG_TRACE_EVENT,
"2553 lpfc_unregister_unused_fcf failed "
"to read FCF record HBA state x%x\n",
phba->pport->port_state);
}
}
/**
* lpfc_unregister_fcf - Unregister the currently registered fcf record
* @phba: Pointer to hba context object.
*
* This function just unregisters the currently reigstered FCF. It does not
* try to find another FCF for discovery.
*/
void
lpfc_unregister_fcf(struct lpfc_hba *phba)
{
int rc;
/* Preparation for unregistering fcf */
rc = lpfc_unregister_fcf_prep(phba);
if (rc) {
scsi: lpfc: Add an internal trace log buffer The current logging methods typically end up requesting a reproduction with a different logging level set to figure out what happened. This was mainly by design to not clutter the kernel log messages with things that were typically not interesting and the messages themselves could cause other issues. When looking to make a better system, it was seen that in many cases when more data was wanted was when another message, usually at KERN_ERR level, was logged. And in most cases, what the additional logging that was then enabled was typically. Most of these areas fell into the discovery machine. Based on this summary, the following design has been put in place: The driver will maintain an internal log (256 elements of 256 bytes). The "additional logging" messages that are usually enabled in a reproduction will be changed to now log all the time to the internal log. A new logging level is defined - LOG_TRACE_EVENT. When this level is set (it is not by default) and a message marked as KERN_ERR is logged, all the messages in the internal log will be dumped to the kernel log before the KERN_ERR message is logged. There is a timestamp on each message added to the internal log. However, this timestamp is not converted to wall time when logged. The value of the timestamp is solely to give a crude time reference for the messages. Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-06-30 21:50:00 +00:00
lpfc_printf_log(phba, KERN_ERR, LOG_TRACE_EVENT,
"2749 Failed to prepare for unregistering "
"HBA's FCF record: rc=%d\n", rc);
return;
}
/* Now, unregister FCF record and reset HBA FCF state */
rc = lpfc_sli4_unregister_fcf(phba);
if (rc)
return;
/* Set proper HBA FCF states after successful unregister FCF */
spin_lock_irq(&phba->hbalock);
phba->fcf.fcf_flag &= ~FCF_REGISTERED;
spin_unlock_irq(&phba->hbalock);
}
/**
* lpfc_unregister_unused_fcf - Unregister FCF if all devices are disconnected.
* @phba: Pointer to hba context object.
*
* This function check if there are any connected remote port for the FCF and
* if all the devices are disconnected, this function unregister FCFI.
* This function also tries to use another FCF for discovery.
*/
void
lpfc_unregister_unused_fcf(struct lpfc_hba *phba)
{
/*
* If HBA is not running in FIP mode, if HBA does not support
* FCoE, if FCF discovery is ongoing, or if FCF has not been
* registered, do nothing.
*/
spin_lock_irq(&phba->hbalock);
2010-11-21 04:11:48 +00:00
if (!(phba->hba_flag & HBA_FCOE_MODE) ||
!(phba->fcf.fcf_flag & FCF_REGISTERED) ||
!(phba->hba_flag & HBA_FIP_SUPPORT) ||
(phba->fcf.fcf_flag & FCF_DISCOVERY) ||
(phba->pport->port_state == LPFC_FLOGI)) {
spin_unlock_irq(&phba->hbalock);
return;
}
spin_unlock_irq(&phba->hbalock);
if (lpfc_fcf_inuse(phba))
return;
lpfc_unregister_fcf_rescan(phba);
}
/**
* lpfc_read_fcf_conn_tbl - Create driver FCF connection table.
* @phba: Pointer to hba context object.
* @buff: Buffer containing the FCF connection table as in the config
* region.
* This function create driver data structure for the FCF connection
* record table read from config region 23.
*/
static void
lpfc_read_fcf_conn_tbl(struct lpfc_hba *phba,
uint8_t *buff)
{
struct lpfc_fcf_conn_entry *conn_entry, *next_conn_entry;
struct lpfc_fcf_conn_hdr *conn_hdr;
struct lpfc_fcf_conn_rec *conn_rec;
uint32_t record_count;
int i;
/* Free the current connect table */
list_for_each_entry_safe(conn_entry, next_conn_entry,
&phba->fcf_conn_rec_list, list) {
list_del_init(&conn_entry->list);
kfree(conn_entry);
}
conn_hdr = (struct lpfc_fcf_conn_hdr *) buff;
record_count = conn_hdr->length * sizeof(uint32_t)/
sizeof(struct lpfc_fcf_conn_rec);
conn_rec = (struct lpfc_fcf_conn_rec *)
(buff + sizeof(struct lpfc_fcf_conn_hdr));
for (i = 0; i < record_count; i++) {
if (!(conn_rec[i].flags & FCFCNCT_VALID))
continue;
conn_entry = kzalloc(sizeof(struct lpfc_fcf_conn_entry),
GFP_KERNEL);
if (!conn_entry) {
scsi: lpfc: Add an internal trace log buffer The current logging methods typically end up requesting a reproduction with a different logging level set to figure out what happened. This was mainly by design to not clutter the kernel log messages with things that were typically not interesting and the messages themselves could cause other issues. When looking to make a better system, it was seen that in many cases when more data was wanted was when another message, usually at KERN_ERR level, was logged. And in most cases, what the additional logging that was then enabled was typically. Most of these areas fell into the discovery machine. Based on this summary, the following design has been put in place: The driver will maintain an internal log (256 elements of 256 bytes). The "additional logging" messages that are usually enabled in a reproduction will be changed to now log all the time to the internal log. A new logging level is defined - LOG_TRACE_EVENT. When this level is set (it is not by default) and a message marked as KERN_ERR is logged, all the messages in the internal log will be dumped to the kernel log before the KERN_ERR message is logged. There is a timestamp on each message added to the internal log. However, this timestamp is not converted to wall time when logged. The value of the timestamp is solely to give a crude time reference for the messages. Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-06-30 21:50:00 +00:00
lpfc_printf_log(phba, KERN_ERR, LOG_TRACE_EVENT,
"2566 Failed to allocate connection"
" table entry\n");
return;
}
memcpy(&conn_entry->conn_rec, &conn_rec[i],
sizeof(struct lpfc_fcf_conn_rec));
list_add_tail(&conn_entry->list,
&phba->fcf_conn_rec_list);
}
if (!list_empty(&phba->fcf_conn_rec_list)) {
i = 0;
list_for_each_entry(conn_entry, &phba->fcf_conn_rec_list,
list) {
conn_rec = &conn_entry->conn_rec;
lpfc_printf_log(phba, KERN_INFO, LOG_INIT,
"3345 FCF connection list rec[%02d]: "
"flags:x%04x, vtag:x%04x, "
"fabric_name:x%02x:%02x:%02x:%02x:"
"%02x:%02x:%02x:%02x, "
"switch_name:x%02x:%02x:%02x:%02x:"
"%02x:%02x:%02x:%02x\n", i++,
conn_rec->flags, conn_rec->vlan_tag,
conn_rec->fabric_name[0],
conn_rec->fabric_name[1],
conn_rec->fabric_name[2],
conn_rec->fabric_name[3],
conn_rec->fabric_name[4],
conn_rec->fabric_name[5],
conn_rec->fabric_name[6],
conn_rec->fabric_name[7],
conn_rec->switch_name[0],
conn_rec->switch_name[1],
conn_rec->switch_name[2],
conn_rec->switch_name[3],
conn_rec->switch_name[4],
conn_rec->switch_name[5],
conn_rec->switch_name[6],
conn_rec->switch_name[7]);
}
}
}
/**
* lpfc_read_fcoe_param - Read FCoe parameters from conf region..
* @phba: Pointer to hba context object.
* @buff: Buffer containing the FCoE parameter data structure.
*
* This function update driver data structure with config
* parameters read from config region 23.
*/
static void
lpfc_read_fcoe_param(struct lpfc_hba *phba,
uint8_t *buff)
{
struct lpfc_fip_param_hdr *fcoe_param_hdr;
struct lpfc_fcoe_params *fcoe_param;
fcoe_param_hdr = (struct lpfc_fip_param_hdr *)
buff;
fcoe_param = (struct lpfc_fcoe_params *)
(buff + sizeof(struct lpfc_fip_param_hdr));
if ((fcoe_param_hdr->parm_version != FIPP_VERSION) ||
(fcoe_param_hdr->length != FCOE_PARAM_LENGTH))
return;
if (fcoe_param_hdr->parm_flags & FIPP_VLAN_VALID) {
phba->valid_vlan = 1;
phba->vlan_id = le16_to_cpu(fcoe_param->vlan_tag) &
0xFFF;
}
phba->fc_map[0] = fcoe_param->fc_map[0];
phba->fc_map[1] = fcoe_param->fc_map[1];
phba->fc_map[2] = fcoe_param->fc_map[2];
return;
}
/**
* lpfc_get_rec_conf23 - Get a record type in config region data.
* @buff: Buffer containing config region 23 data.
* @size: Size of the data buffer.
* @rec_type: Record type to be searched.
*
* This function searches config region data to find the beginning
* of the record specified by record_type. If record found, this
* function return pointer to the record else return NULL.
*/
static uint8_t *
lpfc_get_rec_conf23(uint8_t *buff, uint32_t size, uint8_t rec_type)
{
uint32_t offset = 0, rec_length;
if ((buff[0] == LPFC_REGION23_LAST_REC) ||
(size < sizeof(uint32_t)))
return NULL;
rec_length = buff[offset + 1];
/*
* One TLV record has one word header and number of data words
* specified in the rec_length field of the record header.
*/
while ((offset + rec_length * sizeof(uint32_t) + sizeof(uint32_t))
<= size) {
if (buff[offset] == rec_type)
return &buff[offset];
if (buff[offset] == LPFC_REGION23_LAST_REC)
return NULL;
offset += rec_length * sizeof(uint32_t) + sizeof(uint32_t);
rec_length = buff[offset + 1];
}
return NULL;
}
/**
* lpfc_parse_fcoe_conf - Parse FCoE config data read from config region 23.
* @phba: Pointer to lpfc_hba data structure.
* @buff: Buffer containing config region 23 data.
* @size: Size of the data buffer.
*
* This function parses the FCoE config parameters in config region 23 and
* populate driver data structure with the parameters.
*/
void
lpfc_parse_fcoe_conf(struct lpfc_hba *phba,
uint8_t *buff,
uint32_t size)
{
uint32_t offset = 0;
uint8_t *rec_ptr;
/*
* If data size is less than 2 words signature and version cannot be
* verified.
*/
if (size < 2*sizeof(uint32_t))
return;
/* Check the region signature first */
if (memcmp(buff, LPFC_REGION23_SIGNATURE, 4)) {
scsi: lpfc: Add an internal trace log buffer The current logging methods typically end up requesting a reproduction with a different logging level set to figure out what happened. This was mainly by design to not clutter the kernel log messages with things that were typically not interesting and the messages themselves could cause other issues. When looking to make a better system, it was seen that in many cases when more data was wanted was when another message, usually at KERN_ERR level, was logged. And in most cases, what the additional logging that was then enabled was typically. Most of these areas fell into the discovery machine. Based on this summary, the following design has been put in place: The driver will maintain an internal log (256 elements of 256 bytes). The "additional logging" messages that are usually enabled in a reproduction will be changed to now log all the time to the internal log. A new logging level is defined - LOG_TRACE_EVENT. When this level is set (it is not by default) and a message marked as KERN_ERR is logged, all the messages in the internal log will be dumped to the kernel log before the KERN_ERR message is logged. There is a timestamp on each message added to the internal log. However, this timestamp is not converted to wall time when logged. The value of the timestamp is solely to give a crude time reference for the messages. Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-06-30 21:50:00 +00:00
lpfc_printf_log(phba, KERN_ERR, LOG_TRACE_EVENT,
"2567 Config region 23 has bad signature\n");
return;
}
offset += 4;
/* Check the data structure version */
if (buff[offset] != LPFC_REGION23_VERSION) {
scsi: lpfc: Add an internal trace log buffer The current logging methods typically end up requesting a reproduction with a different logging level set to figure out what happened. This was mainly by design to not clutter the kernel log messages with things that were typically not interesting and the messages themselves could cause other issues. When looking to make a better system, it was seen that in many cases when more data was wanted was when another message, usually at KERN_ERR level, was logged. And in most cases, what the additional logging that was then enabled was typically. Most of these areas fell into the discovery machine. Based on this summary, the following design has been put in place: The driver will maintain an internal log (256 elements of 256 bytes). The "additional logging" messages that are usually enabled in a reproduction will be changed to now log all the time to the internal log. A new logging level is defined - LOG_TRACE_EVENT. When this level is set (it is not by default) and a message marked as KERN_ERR is logged, all the messages in the internal log will be dumped to the kernel log before the KERN_ERR message is logged. There is a timestamp on each message added to the internal log. However, this timestamp is not converted to wall time when logged. The value of the timestamp is solely to give a crude time reference for the messages. Link: https://lore.kernel.org/r/20200630215001.70793-14-jsmart2021@gmail.com Signed-off-by: Dick Kennedy <dick.kennedy@broadcom.com> Signed-off-by: James Smart <jsmart2021@gmail.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2020-06-30 21:50:00 +00:00
lpfc_printf_log(phba, KERN_ERR, LOG_TRACE_EVENT,
"2568 Config region 23 has bad version\n");
return;
}
offset += 4;
/* Read FCoE param record */
rec_ptr = lpfc_get_rec_conf23(&buff[offset],
size - offset, FCOE_PARAM_TYPE);
if (rec_ptr)
lpfc_read_fcoe_param(phba, rec_ptr);
/* Read FCF connection table */
rec_ptr = lpfc_get_rec_conf23(&buff[offset],
size - offset, FCOE_CONN_TBL_TYPE);
if (rec_ptr)
lpfc_read_fcf_conn_tbl(phba, rec_ptr);
}