A reasonably common problem with some devices is that they will
disable MSI generation when the INTX_DISABLE bit is set in the
PCI_COMMAND register.
Quirk this explicitly, guarding the pci_intx() calls in msi.c with
this quirk indication.
The first entries for this quirk are for 5714 and 5780 Tigon3 chips,
and thus we can remove the workaround code from the tg3.c driver.
Signed-off-by: David S. Miller <davem@davemloft.net>
Acked-by: Michael Chan <mchan@broadcom.com>
Acked-by: Jeff Garzik <jgarzik@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
When devices are under a p2p bridge, upstream transactions get replaced by the
device id of the bridge as it owns the PCIE transaction. Hence its necessary
to setup translations on behalf of the bridge as well. Due to this limitation
all devices under a p2p share the same domain in a DMAR.
We just cache the type of device, if its a native PCIe device
or not for later use.
[akpm@linux-foundation.org: BUG_ON -> WARN_ON+recover]
Signed-off-by: Anil S Keshavamurthy <anil.s.keshavamurthy@intel.com>
Cc: Andi Kleen <ak@suse.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Muli Ben-Yehuda <muli@il.ibm.com>
Cc: "Siddha, Suresh B" <suresh.b.siddha@intel.com>
Cc: Arjan van de Ven <arjan@infradead.org>
Cc: Ashok Raj <ashok.raj@intel.com>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Christoph Lameter <clameter@sgi.com>
Cc: Greg KH <greg@kroah.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
* Introduce pci_domains_supported global, hardcoded to zero if
!CONFIG_PCI_DOMAINS.
* Introduce 'nodomains' boot option, which clears pci_domains_supported
on platforms that enable it by default (x86, x86-64, and others when
they are converted to use this).
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Cc: Andi Kleen <ak@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
[PATCH] PCI: make pcie_get_readrq visible in pci.h
pcie_get_readrq() is EXPORT_SYMBOL'ed, but its prototype is not
visible in pci.h, add it there.
This is needed by some network drivers.
Signed-off-by: Brice Goglin <brice@myri.com>
Acked-by: Peter Oruba <peter.oruba@amd.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
We need pci_bus_find_capability() in some arch/powerpc code so move
the prototype into a header accessible to it.
Also kill the duplicate prototype for pci_bus_alloc_resource().
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Some odd ACPI implementations choke if certain controller is disabled
when ACPI suspend is invoked but we still need to make sure the PCI
device is enabled during resume. Simply using pci_enable_device()
unbalances device enable count. Export __pci_reenable_device().
Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
Before calling init_hwif_default, ide_unregister gets lock ide_lock and
disables irq. init_hwif_default calls ide_default_io_base which calls
pci_get_device and later pci_get_subsys tries to apply for semaphore
pci_bus_sem and goes to sleep.
Mostly, pci_get_device should be called when irq is turned on.
ide_default_io_base just needs find if list pci_devices is empty.
Signed-off-by: Zhang Yanmin <yanmin.zhang@intel.com>
Cc: Greg KH <greg@kroah.com>
Cc: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
* 'release' of git://git.kernel.org/pub/scm/linux/kernel/git/aegl/linux-2.6:
[IA64] Support multiple CPUs going through OS_MCA
[IA64] silence GCC ia64 unused variable warnings
[IA64] prevent MCA when performing MMIO mmap to PCI config space
[IA64] add sn_register_pmi_handler oemcall
[IA64] Stop bit for brl instruction
[IA64] SN: Correct ROM resource length for BIOS copy
[IA64] Don't set psr.ic and psr.i simultaneously
As suggested by Andrew, add pci_try_set_mwi(), which does not require
return-value checking.
- add pci_try_set_mwi() without __must_check
- make it return 0 on success, errno if the "try" failed or error
- review callers
Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Currently there are 97 occurrences where drivers need the pci
revision ID. We can do this once for all devices. Even the pci
subsystem needs the revision several times for quirks. The extra
u8 member pads out nicely in the pci_dev struct.
Signed-off-by: Auke Kok <auke-jan.h.kok@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Remove pointless and never-called enable_wake() hook from pci_driver and
from documentation. Evidently this was introduced in the 2.4.6 kernel,
but there's no evidence it was ever called; and it was rarely implemented.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
We've now fixed up most users of pci_find_slot, and the remainder are either
hard and need someone with the hardware and info to work on it, or patches
exist but are not yet merged.
Time therefore for some gentle encouragement
Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Currently pcibios_add_platform_entries() returns void, but could fail,
so instead have it return an int and propagate errors up to
pci_create_sysfs_dev_files().
Fixes:
arch/powerpc/kernel/pci_64.c: In function 'pcibios_add_platform_entries':
arch/powerpc/kernel/pci_64.c:878: warning: ignoring return value of
'device_create_file', declared with attribute warn_unused_result
arch/powerpc/kernel/pci_32.c: In function 'pcibios_add_platform_entries':
arch/powerpc/kernel/pci_32.c:1043: warning: ignoring return value of
'device_create_file', declared with attribute warn_unused_result
Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
I'm not sure if this is going to fly, weak symbols work on the compilers I'm
using, but whether they work for all of the affected architectures I can't say.
I've cc'ed as many arch maintainers/lists as I could find.
But assuming they do, we can use a weak empty definition of
pcibios_add_platform_entries() to avoid having an empty definition on every
arch.
Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
This patch introduces an interface to read and write PCI-X / PCI-Express
maximum read byte count values from PCI config space. There is a second
function that returns the maximum _designed_ read byte count, which marks the
maximum value for a device, since some drivers try to set MMRBC to the
highest allowed value and rely on such a function.
Based on patch set by Stephen Hemminger <shemminger@linux-foundation.org>
Cc: Stephen Hemminger <shemminger@linux-foundation.org>
Signed-off-by: Peter Oruba <peter.oruba@amd.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
On SN systems, when setting the IORESOURCE_ROM_BIOS_COPY resource flag,
the resource length should be set to the actual size of the ROM image
so that a call to pci_map_rom() returns the correct size.
Signed-off-by: John Keller <jpk@sgi.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Tony Luck <tony.luck@intel.com>
Provide stubs for more PCI bus/slot functions when CONFIG_PCI=n.
Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Cc: Matt Domsch <Matt_Domsch@dell.com>
Cc: Greg KH <greg@kroah.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Unless we finally completely remove it, people will always add new users.
Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Now that we keep a list of msi descriptors, we don't need first_msi_irq
in the pci dev.
If we somehow have zero MSIs configured list_entry() will give us weird
oopes or nice memory corruption bugs. So be paranoid. Add BUG_ONs and also
a check in pci_msi_check_device() to make sure nvec > 0.
Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The msi descriptors are linked together with what looks a lot like
a linked list, but isn't a struct list_head list. Make it one.
The only complication is that previously we walked a list of irqs, and
got the descriptor for each with get_irq_msi(). Now we have a list of
descriptors and need to get the irq out of it, so it needs to be in the
actual struct msi_desc. We use 0 to indicate no irq is setup.
Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
There are currently several places in the kernel where we kmalloc()
a struct pci_dev and start initialising it. It'd be preferable to
have an allocator so we can ensure the pci_dev is correctly initialised
in one place.
Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Balance declarations of pci_request_regions() and pci_release_regions() with
empty inline definitions for the CONFIG_PCI=n case -- otherwise my patch to
drivers/net/3c59x.c in the -mm tree doesn't compile. :-)
Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Adds a new API which can be used to issue various types
of PCI-E reset, including PCI-E warm reset and PCI-E hot reset.
This is needed for an ipr PCI-E adapter which does not properly
implement BIST. Running BIST on this adapter results in PCI-E
errors. The only reliable reset mechanism that exists on this
hardware is PCI Fundamental reset (warm reset). Since driving
this type of reset is architecture unique, this provides the
necessary hooks for architectures to add this support.
Signed-off-by: Brian King <brking@linux.vnet.ibm.com>
Acked-by: Linas Vepstas <linas@austin.ibm.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Implement pcim_iounmap_regions() - the opposite of
pcim_iomap_regions().
Signed-off-by: Tejun heo <htejun@gmail.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
Make multithreaded probing work per subsystem instead of per driver.
It doesn't make much sense to probe the same device for multiple drivers in
parallel (after all, only one driver can bind to the device). Instead, create
a probing thread for each device that probes the drivers one after another.
Also make the decision to use multi-threaded probe per bus instead of per
device and adapt the pci code.
Signed-off-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Because we do not reserve space for the pci-x and pci-e state in struct
pci dev we need to dynamically allocate it. However because we need
to support restore being called multiple times after a single save
it is never safe to free the buffers we have allocated to hold the
state.
So this patch modifies the save routines to first check to see
if we have already allocated a state buffer before allocating
a new one. Then the restore routines are modified to not free
the state after restoring it. Simple and it fixes some subtle
error path handling bugs, that are hard to test for.
Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Acked-by: Auke Kok <auke-jan.h.kok@intel.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
In some cases when we are not using msi we need a way to ensure that the
hardware does not have an msi capability enabled. Currently the code has been
calling disable_msi_mode to try and achieve that. However disable_msi_mode
has several other side effects and is only available when msi support is
compiled in so it isn't really appropriate.
Instead this patch implements pci_msi_off which disables all msi and msix
capabilities unconditionally with no additional side effects.
pci_disable_device was redundantly clearing the bus master enable flag and
clearing the msi enable bit. A device that is not allowed to perform bus
mastering operations cannot generate intx or msi interrupt messages as those
are essentially a special case of dma, and require bus mastering. So the call
in pci_disable_device to disable msi capabilities was redundant.
quirk_pcie_pxh also called disable_msi_mode and is updated to use pci_msi_off.
Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Cc: Michael Ellerman <michael@ellerman.id.au>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Greg KH <greg@kroah.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
CARDBUS_MEM_SIZE was increased to 64MB on 2.6.20-rc2, but larger size might
result in allocation failure for the reserving itself on some platforms
(for example typical 32bit MIPS). Make it (and CARDBUS_IO_SIZE too)
customizable by "pci=" option for such platforms.
Signed-off-by: Atsushi Nemoto <anemo@mba.ocn.ne.jp>
Cc: Daniel Ritz <daniel.ritz@gmx.ch>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: Dominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
* Split the implementation-agnostic stuff in separate files.
* Make sure that targets using non-default request_irq() pull
kernel/irq/devres.o
* Introduce new symbols (HAS_IOPORT and HAS_IOMEM) defaulting to positive;
allow architectures to turn them off (we needed these symbols anyway for
dependencies of quite a few drivers).
* protect the ioport-related parts of lib/devres.o with CONFIG_HAS_IOPORT.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Implement device resource management, in short, devres. A device
driver can allocate arbirary size of devres data which is associated
with a release function. On driver detach, release function is
invoked on the devres data, then, devres data is freed.
devreses are typed by associated release functions. Some devreses are
better represented by single instance of the type while others need
multiple instances sharing the same release function. Both usages are
supported.
devreses can be grouped using devres group such that a device driver
can easily release acquired resources halfway through initialization
or selectively release resources (e.g. resources for port 1 out of 4
ports).
This patch adds devres core including documentation and the following
managed interfaces.
* alloc/free : devm_kzalloc(), devm_kzfree()
* IO region : devm_request_region(), devm_release_region()
* IRQ : devm_request_irq(), devm_free_irq()
* DMA : dmam_alloc_coherent(), dmam_free_coherent(),
dmam_declare_coherent_memory(), dmam_pool_create(),
dmam_pool_destroy()
* PCI : pcim_enable_device(), pcim_pin_device(), pci_is_managed()
* iomap : devm_ioport_map(), devm_ioport_unmap(), devm_ioremap(),
devm_ioremap_nocache(), devm_iounmap(), pcim_iomap_table(),
pcim_iomap(), pcim_iounmap()
Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Jeff Garzik <jeff@garzik.org>
The function msi_lookup_irq was horrible. As a side effect of running
it changed dev->irq, and then the callers would need to change it
back. In addition it does a global scan through all of the irqs,
which seems to be the sole justification of the msi_lock.
To remove the neede for msi_lookup_irq I added first_msi_irq to struct
pci_dev. Then depending on the context I replaced msi_lookup_irq with
dev->first_msi_irq, dev->msi_enabled, or dev->msix_enabled.
msi_enabled and msix_enabled were already present in pci_dev for other
reasons.
Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Acked-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
pci_scan_msi_device() doesn't do anything anymore, so remove it.
Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Since isa_bridge is neither assigned any value !NULL nor used on !Alpha,
there's no reason for providing it.
Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
On Fri, Nov 17, 2006 at 09:32:36AM -0500, Alan Cox wrote:
>
> Soon we should deprecate pci_find_device as well
So let's mark it as __deprecated now, which also has the side effect
that noone can later whine that removing it might break some shiny
external modules.
Oh, and if anything starts complaining "But this adds some warnings to
my kernel build!", he should either first fix the 200 kB (sic) of
warnings I'm getting in 2.6.19-rc5-mm2 starting at MODPOST or go to hell.
Signed-off-by: Adrian Bunk <bunk@stusta.de>
Acked-by: Alan Cox <alan@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
This patch removes the no longer used pci_find_device_reverse().
Signed-off-by: Adrian Bunk <bunk@stusta.de>
Acked-by: Alan Cox <alan@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
This patch adds the following changes into generic PCI code especially
for PCI legacy I/O port free drivers.
- Added new pci_request_selected_regions() and
pci_release_selected_regions() for PCI legacy I/O port free
drivers in order to request/release only the selected regions.
- Added helper routine pci_select_bars() which makes proper mask
of BARs from the specified resource type. This would be very
helpful for users of pci_enable_device_bars().
Signed-off-by: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
Signed-off-by: Hidetoshi Seto <seto.hidetoshi@jp.fujitsu.com>
Cc: Inaky Perez-Gonzalez <inaky@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
This adds the module name to all PCI drivers, if they are built into the
kernel or not. It will show up in /sys/modules/MODULE_NAME/drivers/
It also fixes up the IDE core, which was calling __pci_register_driver()
directly.
Cc: Kay Sievers <kay.sievers@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
I don't see any good reason for exporting device IDs to userspace.
Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
This patch is designed to fix:
- Disk eating corruptor on KT7 after resume from RAM
- VIA IRQ handling
- VIA fixups for bus lockups after resume from RAM
The core of this is to add a table of resume fixups run at resume time.
We need to do this for a variety of boards and features, but particularly
we need to do this to get various critical VIA fixups done on resume.
The second part of the problem is to handle VIA IRQ number rules which
are a bit odd and need special handling for PIC interrupts. Various
patches broke various boxes and while this one may not be perfect
(hopefully it is) it ensures the workaround is applied to the right
devices only.
From: Jean Delvare <khali@linux-fr.org>
Now that PCI quirks are replayed on software resume, we can safely
re-enable the Asus SMBus unhiding quirk even when software suspend support
is enabled.
[akpm@osdl.org: fix const warning]
Signed-off-by: Alan Cox <alan@redhat.com>
Cc: Jean Delvare <khali@linux-fr.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
There are already several places in the kernel that want to search a PCI
device for a given Hypertransport capability. Although this is possible
using pci_find_capability() etc., it makes sense to encapsulate that
logic in a helper - pci_find_ht_capability().
To cater for searching exhaustively for a capability, we also provide
pci_find_next_ht_capability().
We also need to cater for the fact that the HT capability fields may be
either 3 or 5 bits wide. pci_find_ht_capability() deals with this for you,
but callers using the #defines directly must handle that themselves.
Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
This works like pci_dev_present but instead of returning boolean returns
the matching pci_device_id entry. This makes it much more useful. Code
bloat is basically nil as the old boolean function is rewritten in terms of
the new one.
This will be used by the updated VIA PCI quirks for one
Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Changes the pci_{enable,disable}_device() functions to work in a
nested basis, so that eg, three calls to enable_device() require three
calls to disable_device().
The reason for this is to simplify PCI drivers for
multi-interface/capability devices. These are devices that cram more
than one interface in a single function. A relevant example of that is
the Wireless [USB] Host Controller Interface (similar to EHCI) [see
http://www.intel.com/technology/comms/wusb/whci.htm].
In these kind of devices, multiple interfaces are accessed through a
single bar and IRQ line. For that, the drivers map only the smallest
area of the bar to access their register banks and use shared IRQ
handlers.
However, because the order at which those drivers load cannot be known
ahead of time, the sequence in which the calls to pci_enable_device()
and pci_disable_device() cannot be predicted. Thus:
1. driverA starts pci_enable_device()
2. driverB starts pci_enable_device()
3. driverA shutdown pci_disable_device()
4. driverB shutdown pci_disable_device()
between steps 3 and 4, driver B would loose access to it's device,
even if it didn't intend to.
By using this modification, the device won't be disabled until all the
callers to enable() have called disable().
This is implemented by replacing 'struct pci_dev->is_enabled' from a
bitfield to an atomic use count. Each caller to enable increments it,
each caller to disable decrements it. When the count increments from 0
to 1, __pci_enable_device() is called to actually enable the
device. When it drops to zero, pci_disable_device() actually does the
disabling.
We keep the backend __pci_enable_device() for pci_default_resume() to
use and also change the sysfs method implementation, so that userspace
enabling/disabling the device doesn't disable it one time too much.
Signed-off-by: Inaky Perez-Gonzalez <inaky@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
We seem to have lost the declaration of pci_get_device_reverse(), if we ever
had one.
Add a CONFIG_PCI=0 stub too.
Acked-by: Alan Cox <alan@lxorguk.ukuu.org.uk>
Cc: Greg KH <greg@kroah.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Problem:
New Dell PowerEdge servers have 2 embedded ethernet ports, which are
labeled NIC1 and NIC2 on the chassis, in the BIOS setup screens, and
in the printed documentation. Assuming no other add-in ethernet ports
in the system, Linux 2.4 kernels name these eth0 and eth1
respectively. Many people have come to expect this naming. Linux 2.6
kernels name these eth1 and eth0 respectively (backwards from
expectations). I also have reports that various Sun and HP servers
have similar behavior.
Root cause:
Linux 2.4 kernels walk the pci_devices list, which happens to be
sorted in breadth-first order (or pcbios_find_device order on i386,
which most often is breadth-first also). 2.6 kernels have both the
pci_devices list and the pci_bus_type.klist_devices list, the latter
is what is walked at driver load time to match the pci_id tables; this
klist happens to be in depth-first order.
On systems where, for physical routing reasons, NIC1 appears on a
lower bus number than NIC2, but NIC2's bridge is discovered first in
the depth-first ordering, NIC2 will be discovered before NIC1. If the
list were sorted breadth-first, NIC1 would be discovered before NIC2.
A PowerEdge 1955 system has the following topology which easily
exhibits the difference between depth-first and breadth-first device
lists.
-[0000:00]-+-00.0 Intel Corporation 5000P Chipset Memory Controller Hub
+-02.0-[0000:03-08]--+-00.0-[0000:04-07]--+-00.0-[0000:05-06]----00.0-[0000:06]----00.0 Broadcom Corporation NetXtreme II BCM5708S Gigabit Ethernet (labeled NIC2, 2.4 kernel name eth1, 2.6 kernel name eth0)
+-1c.0-[0000:01-02]----00.0-[0000:02]----00.0 Broadcom Corporation NetXtreme II BCM5708S Gigabit Ethernet (labeled NIC1, 2.4 kernel name eth0, 2.6 kernel name eth1)
Other factors, such as device driver load order and the presence of
PCI slots at various points in the bus hierarchy further complicate
this problem; I'm not trying to solve those here, just restore the
device order, and thus basic behavior, that 2.4 kernels had.
Solution:
The solution can come in multiple steps.
Suggested fix#1: kernel
Patch below optionally sorts the two device lists into breadth-first
ordering to maintain compatibility with 2.4 kernels. It adds two new
command line options:
pci=bfsort
pci=nobfsort
to force the sort order, or not, as you wish. It also adds DMI checks
for the specific Dell systems which exhibit "backwards" ordering, to
make them "right".
Suggested fix#2: udev rules from userland
Many people also have the expectation that embedded NICs are always
discovered before add-in NICs (which this patch does not try to do).
Using the PCI IRQ Routing Table provided by system BIOS, it's easy to
determine which PCI devices are embedded, or if add-in, which PCI slot
they're in. I'm working on a tool that would allow udev to name
ethernet devices in ascending embedded, slot 1 .. slot N order,
subsort by PCI bus/dev/fn breadth-first. It'll be possible to use it
independent of udev as well for those distributions that don't use
udev in their installers.
Suggested fix#3: system board routing rules
One can constrain the system board layout to put NIC1 ahead of NIC2
regardless of breadth-first or depth-first discovery order. This adds
a significant level of complexity to board routing, and may not be
possible in all instances (witness the above systems from several
major manufacturers). I don't want to encourage this particular train
of thought too far, at the expense of not doing #1 or #2 above.
Feedback appreciated. Patch tested on a Dell PowerEdge 1955 blade
with 2.6.18.
You'll also note I took some liberty and temporarily break the klist
abstraction to simplify and speed up the sort algorithm. I think
that's both safe and appropriate in this instance.
Signed-off-by: Matt Domsch <Matt_Domsch@dell.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Jeff Garzik