mirror of
https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git
synced 2024-08-26 19:00:25 +00:00
![Dave Chinner](/assets/img/avatar_default.png)
One of the problems we currently have with delayed logging is that under serious memory pressure we can deadlock memory reclaim. THis occurs when memory reclaim (such as run by kswapd) is reclaiming XFS inodes and issues a log force to unpin inodes that are dirty in the CIL. The CIL is pushed, but this will only occur once it gets the CIL context lock to ensure that all committing transactions are complete and no new transactions start being committed to the CIL while the push switches to a new context. The deadlock occurs when the CIL context lock is held by a committing process that is doing memory allocation for log vector buffers, and that allocation is then blocked on memory reclaim making progress. Memory reclaim, however, is blocked waiting for a log force to make progress, and so we effectively deadlock at this point. To solve this problem, we have to move the CIL log vector buffer allocation outside of the context lock so that memory reclaim can always make progress when it needs to force the log. The problem with doing this is that a CIL push can take place while we are determining if we need to allocate a new log vector buffer for an item and hence the current log vector may go away without warning. That means we canot rely on the existing log vector being present when we finally grab the context lock and so we must have a replacement buffer ready to go at all times. To ensure this, introduce a "shadow log vector" buffer that is always guaranteed to be present when we gain the CIL context lock and format the item. This shadow buffer may or may not be used during the formatting, but if the log item does not have an existing log vector buffer or that buffer is too small for the new modifications, we swap it for the new shadow buffer and format the modifications into that new log vector buffer. The result of this is that for any object we modify more than once in a given CIL checkpoint, we double the memory required to track dirty regions in the log. For single modifications then we consume the shadow log vectorwe allocate on commit, and that gets consumed by the checkpoint. However, if we make multiple modifications, then the second transaction commit will allocate a shadow log vector and hence we will end up with double the memory usage as only one of the log vectors is consumed by the CIL checkpoint. The remaining shadow vector will be freed when th elog item is freed. This can probably be optimised in future - access to the shadow log vector is serialised by the object lock (as opposited to the active log vector, which is controlled by the CIL context lock) and so we can probably free shadow log vector from some objects when the log item is marked clean on removal from the AIL. Signed-off-by: Dave Chinner <dchinner@redhat.com> Reviewed-by: Brian Foster <bfoster@redhat.com> Signed-off-by: Dave Chinner <david@fromorbit.com>
445 lines
11 KiB
C
445 lines
11 KiB
C
/*
|
|
* Copyright (c) 2000-2003 Silicon Graphics, Inc.
|
|
* All Rights Reserved.
|
|
*
|
|
* This program is free software; you can redistribute it and/or
|
|
* modify it under the terms of the GNU General Public License as
|
|
* published by the Free Software Foundation.
|
|
*
|
|
* This program is distributed in the hope that it would be useful,
|
|
* but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
|
* GNU General Public License for more details.
|
|
*
|
|
* You should have received a copy of the GNU General Public License
|
|
* along with this program; if not, write the Free Software Foundation,
|
|
* Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
|
|
*/
|
|
#include "xfs.h"
|
|
#include "xfs_fs.h"
|
|
#include "xfs_format.h"
|
|
#include "xfs_log_format.h"
|
|
#include "xfs_trans_resv.h"
|
|
#include "xfs_mount.h"
|
|
#include "xfs_inode.h"
|
|
#include "xfs_quota.h"
|
|
#include "xfs_error.h"
|
|
#include "xfs_trans.h"
|
|
#include "xfs_buf_item.h"
|
|
#include "xfs_trans_priv.h"
|
|
#include "xfs_qm.h"
|
|
#include "xfs_log.h"
|
|
|
|
static inline struct xfs_dq_logitem *DQUOT_ITEM(struct xfs_log_item *lip)
|
|
{
|
|
return container_of(lip, struct xfs_dq_logitem, qli_item);
|
|
}
|
|
|
|
/*
|
|
* returns the number of iovecs needed to log the given dquot item.
|
|
*/
|
|
STATIC void
|
|
xfs_qm_dquot_logitem_size(
|
|
struct xfs_log_item *lip,
|
|
int *nvecs,
|
|
int *nbytes)
|
|
{
|
|
*nvecs += 2;
|
|
*nbytes += sizeof(struct xfs_dq_logformat) +
|
|
sizeof(struct xfs_disk_dquot);
|
|
}
|
|
|
|
/*
|
|
* fills in the vector of log iovecs for the given dquot log item.
|
|
*/
|
|
STATIC void
|
|
xfs_qm_dquot_logitem_format(
|
|
struct xfs_log_item *lip,
|
|
struct xfs_log_vec *lv)
|
|
{
|
|
struct xfs_dq_logitem *qlip = DQUOT_ITEM(lip);
|
|
struct xfs_log_iovec *vecp = NULL;
|
|
struct xfs_dq_logformat *qlf;
|
|
|
|
qlf = xlog_prepare_iovec(lv, &vecp, XLOG_REG_TYPE_QFORMAT);
|
|
qlf->qlf_type = XFS_LI_DQUOT;
|
|
qlf->qlf_size = 2;
|
|
qlf->qlf_id = be32_to_cpu(qlip->qli_dquot->q_core.d_id);
|
|
qlf->qlf_blkno = qlip->qli_dquot->q_blkno;
|
|
qlf->qlf_len = 1;
|
|
qlf->qlf_boffset = qlip->qli_dquot->q_bufoffset;
|
|
xlog_finish_iovec(lv, vecp, sizeof(struct xfs_dq_logformat));
|
|
|
|
xlog_copy_iovec(lv, &vecp, XLOG_REG_TYPE_DQUOT,
|
|
&qlip->qli_dquot->q_core,
|
|
sizeof(struct xfs_disk_dquot));
|
|
}
|
|
|
|
/*
|
|
* Increment the pin count of the given dquot.
|
|
*/
|
|
STATIC void
|
|
xfs_qm_dquot_logitem_pin(
|
|
struct xfs_log_item *lip)
|
|
{
|
|
struct xfs_dquot *dqp = DQUOT_ITEM(lip)->qli_dquot;
|
|
|
|
ASSERT(XFS_DQ_IS_LOCKED(dqp));
|
|
atomic_inc(&dqp->q_pincount);
|
|
}
|
|
|
|
/*
|
|
* Decrement the pin count of the given dquot, and wake up
|
|
* anyone in xfs_dqwait_unpin() if the count goes to 0. The
|
|
* dquot must have been previously pinned with a call to
|
|
* xfs_qm_dquot_logitem_pin().
|
|
*/
|
|
STATIC void
|
|
xfs_qm_dquot_logitem_unpin(
|
|
struct xfs_log_item *lip,
|
|
int remove)
|
|
{
|
|
struct xfs_dquot *dqp = DQUOT_ITEM(lip)->qli_dquot;
|
|
|
|
ASSERT(atomic_read(&dqp->q_pincount) > 0);
|
|
if (atomic_dec_and_test(&dqp->q_pincount))
|
|
wake_up(&dqp->q_pinwait);
|
|
}
|
|
|
|
STATIC xfs_lsn_t
|
|
xfs_qm_dquot_logitem_committed(
|
|
struct xfs_log_item *lip,
|
|
xfs_lsn_t lsn)
|
|
{
|
|
/*
|
|
* We always re-log the entire dquot when it becomes dirty,
|
|
* so, the latest copy _is_ the only one that matters.
|
|
*/
|
|
return lsn;
|
|
}
|
|
|
|
/*
|
|
* This is called to wait for the given dquot to be unpinned.
|
|
* Most of these pin/unpin routines are plagiarized from inode code.
|
|
*/
|
|
void
|
|
xfs_qm_dqunpin_wait(
|
|
struct xfs_dquot *dqp)
|
|
{
|
|
ASSERT(XFS_DQ_IS_LOCKED(dqp));
|
|
if (atomic_read(&dqp->q_pincount) == 0)
|
|
return;
|
|
|
|
/*
|
|
* Give the log a push so we don't wait here too long.
|
|
*/
|
|
xfs_log_force(dqp->q_mount, 0);
|
|
wait_event(dqp->q_pinwait, (atomic_read(&dqp->q_pincount) == 0));
|
|
}
|
|
|
|
STATIC uint
|
|
xfs_qm_dquot_logitem_push(
|
|
struct xfs_log_item *lip,
|
|
struct list_head *buffer_list) __releases(&lip->li_ailp->xa_lock)
|
|
__acquires(&lip->li_ailp->xa_lock)
|
|
{
|
|
struct xfs_dquot *dqp = DQUOT_ITEM(lip)->qli_dquot;
|
|
struct xfs_buf *bp = NULL;
|
|
uint rval = XFS_ITEM_SUCCESS;
|
|
int error;
|
|
|
|
if (atomic_read(&dqp->q_pincount) > 0)
|
|
return XFS_ITEM_PINNED;
|
|
|
|
if (!xfs_dqlock_nowait(dqp))
|
|
return XFS_ITEM_LOCKED;
|
|
|
|
/*
|
|
* Re-check the pincount now that we stabilized the value by
|
|
* taking the quota lock.
|
|
*/
|
|
if (atomic_read(&dqp->q_pincount) > 0) {
|
|
rval = XFS_ITEM_PINNED;
|
|
goto out_unlock;
|
|
}
|
|
|
|
/*
|
|
* Someone else is already flushing the dquot. Nothing we can do
|
|
* here but wait for the flush to finish and remove the item from
|
|
* the AIL.
|
|
*/
|
|
if (!xfs_dqflock_nowait(dqp)) {
|
|
rval = XFS_ITEM_FLUSHING;
|
|
goto out_unlock;
|
|
}
|
|
|
|
spin_unlock(&lip->li_ailp->xa_lock);
|
|
|
|
error = xfs_qm_dqflush(dqp, &bp);
|
|
if (error) {
|
|
xfs_warn(dqp->q_mount, "%s: push error %d on dqp %p",
|
|
__func__, error, dqp);
|
|
} else {
|
|
if (!xfs_buf_delwri_queue(bp, buffer_list))
|
|
rval = XFS_ITEM_FLUSHING;
|
|
xfs_buf_relse(bp);
|
|
}
|
|
|
|
spin_lock(&lip->li_ailp->xa_lock);
|
|
out_unlock:
|
|
xfs_dqunlock(dqp);
|
|
return rval;
|
|
}
|
|
|
|
/*
|
|
* Unlock the dquot associated with the log item.
|
|
* Clear the fields of the dquot and dquot log item that
|
|
* are specific to the current transaction. If the
|
|
* hold flags is set, do not unlock the dquot.
|
|
*/
|
|
STATIC void
|
|
xfs_qm_dquot_logitem_unlock(
|
|
struct xfs_log_item *lip)
|
|
{
|
|
struct xfs_dquot *dqp = DQUOT_ITEM(lip)->qli_dquot;
|
|
|
|
ASSERT(XFS_DQ_IS_LOCKED(dqp));
|
|
|
|
/*
|
|
* Clear the transaction pointer in the dquot
|
|
*/
|
|
dqp->q_transp = NULL;
|
|
|
|
/*
|
|
* dquots are never 'held' from getting unlocked at the end of
|
|
* a transaction. Their locking and unlocking is hidden inside the
|
|
* transaction layer, within trans_commit. Hence, no LI_HOLD flag
|
|
* for the logitem.
|
|
*/
|
|
xfs_dqunlock(dqp);
|
|
}
|
|
|
|
/*
|
|
* this needs to stamp an lsn into the dquot, I think.
|
|
* rpc's that look at user dquot's would then have to
|
|
* push on the dependency recorded in the dquot
|
|
*/
|
|
STATIC void
|
|
xfs_qm_dquot_logitem_committing(
|
|
struct xfs_log_item *lip,
|
|
xfs_lsn_t lsn)
|
|
{
|
|
}
|
|
|
|
/*
|
|
* This is the ops vector for dquots
|
|
*/
|
|
static const struct xfs_item_ops xfs_dquot_item_ops = {
|
|
.iop_size = xfs_qm_dquot_logitem_size,
|
|
.iop_format = xfs_qm_dquot_logitem_format,
|
|
.iop_pin = xfs_qm_dquot_logitem_pin,
|
|
.iop_unpin = xfs_qm_dquot_logitem_unpin,
|
|
.iop_unlock = xfs_qm_dquot_logitem_unlock,
|
|
.iop_committed = xfs_qm_dquot_logitem_committed,
|
|
.iop_push = xfs_qm_dquot_logitem_push,
|
|
.iop_committing = xfs_qm_dquot_logitem_committing
|
|
};
|
|
|
|
/*
|
|
* Initialize the dquot log item for a newly allocated dquot.
|
|
* The dquot isn't locked at this point, but it isn't on any of the lists
|
|
* either, so we don't care.
|
|
*/
|
|
void
|
|
xfs_qm_dquot_logitem_init(
|
|
struct xfs_dquot *dqp)
|
|
{
|
|
struct xfs_dq_logitem *lp = &dqp->q_logitem;
|
|
|
|
xfs_log_item_init(dqp->q_mount, &lp->qli_item, XFS_LI_DQUOT,
|
|
&xfs_dquot_item_ops);
|
|
lp->qli_dquot = dqp;
|
|
}
|
|
|
|
/*------------------ QUOTAOFF LOG ITEMS -------------------*/
|
|
|
|
static inline struct xfs_qoff_logitem *QOFF_ITEM(struct xfs_log_item *lip)
|
|
{
|
|
return container_of(lip, struct xfs_qoff_logitem, qql_item);
|
|
}
|
|
|
|
|
|
/*
|
|
* This returns the number of iovecs needed to log the given quotaoff item.
|
|
* We only need 1 iovec for an quotaoff item. It just logs the
|
|
* quotaoff_log_format structure.
|
|
*/
|
|
STATIC void
|
|
xfs_qm_qoff_logitem_size(
|
|
struct xfs_log_item *lip,
|
|
int *nvecs,
|
|
int *nbytes)
|
|
{
|
|
*nvecs += 1;
|
|
*nbytes += sizeof(struct xfs_qoff_logitem);
|
|
}
|
|
|
|
STATIC void
|
|
xfs_qm_qoff_logitem_format(
|
|
struct xfs_log_item *lip,
|
|
struct xfs_log_vec *lv)
|
|
{
|
|
struct xfs_qoff_logitem *qflip = QOFF_ITEM(lip);
|
|
struct xfs_log_iovec *vecp = NULL;
|
|
struct xfs_qoff_logformat *qlf;
|
|
|
|
qlf = xlog_prepare_iovec(lv, &vecp, XLOG_REG_TYPE_QUOTAOFF);
|
|
qlf->qf_type = XFS_LI_QUOTAOFF;
|
|
qlf->qf_size = 1;
|
|
qlf->qf_flags = qflip->qql_flags;
|
|
xlog_finish_iovec(lv, vecp, sizeof(struct xfs_qoff_logitem));
|
|
}
|
|
|
|
/*
|
|
* Pinning has no meaning for an quotaoff item, so just return.
|
|
*/
|
|
STATIC void
|
|
xfs_qm_qoff_logitem_pin(
|
|
struct xfs_log_item *lip)
|
|
{
|
|
}
|
|
|
|
/*
|
|
* Since pinning has no meaning for an quotaoff item, unpinning does
|
|
* not either.
|
|
*/
|
|
STATIC void
|
|
xfs_qm_qoff_logitem_unpin(
|
|
struct xfs_log_item *lip,
|
|
int remove)
|
|
{
|
|
}
|
|
|
|
/*
|
|
* There isn't much you can do to push a quotaoff item. It is simply
|
|
* stuck waiting for the log to be flushed to disk.
|
|
*/
|
|
STATIC uint
|
|
xfs_qm_qoff_logitem_push(
|
|
struct xfs_log_item *lip,
|
|
struct list_head *buffer_list)
|
|
{
|
|
return XFS_ITEM_LOCKED;
|
|
}
|
|
|
|
/*
|
|
* Quotaoff items have no locking or pushing, so return failure
|
|
* so that the caller doesn't bother with us.
|
|
*/
|
|
STATIC void
|
|
xfs_qm_qoff_logitem_unlock(
|
|
struct xfs_log_item *lip)
|
|
{
|
|
}
|
|
|
|
/*
|
|
* The quotaoff-start-item is logged only once and cannot be moved in the log,
|
|
* so simply return the lsn at which it's been logged.
|
|
*/
|
|
STATIC xfs_lsn_t
|
|
xfs_qm_qoff_logitem_committed(
|
|
struct xfs_log_item *lip,
|
|
xfs_lsn_t lsn)
|
|
{
|
|
return lsn;
|
|
}
|
|
|
|
STATIC xfs_lsn_t
|
|
xfs_qm_qoffend_logitem_committed(
|
|
struct xfs_log_item *lip,
|
|
xfs_lsn_t lsn)
|
|
{
|
|
struct xfs_qoff_logitem *qfe = QOFF_ITEM(lip);
|
|
struct xfs_qoff_logitem *qfs = qfe->qql_start_lip;
|
|
struct xfs_ail *ailp = qfs->qql_item.li_ailp;
|
|
|
|
/*
|
|
* Delete the qoff-start logitem from the AIL.
|
|
* xfs_trans_ail_delete() drops the AIL lock.
|
|
*/
|
|
spin_lock(&ailp->xa_lock);
|
|
xfs_trans_ail_delete(ailp, &qfs->qql_item, SHUTDOWN_LOG_IO_ERROR);
|
|
|
|
kmem_free(qfs->qql_item.li_lv_shadow);
|
|
kmem_free(lip->li_lv_shadow);
|
|
kmem_free(qfs);
|
|
kmem_free(qfe);
|
|
return (xfs_lsn_t)-1;
|
|
}
|
|
|
|
/*
|
|
* XXX rcc - don't know quite what to do with this. I think we can
|
|
* just ignore it. The only time that isn't the case is if we allow
|
|
* the client to somehow see that quotas have been turned off in which
|
|
* we can't allow that to get back until the quotaoff hits the disk.
|
|
* So how would that happen? Also, do we need different routines for
|
|
* quotaoff start and quotaoff end? I suspect the answer is yes but
|
|
* to be sure, I need to look at the recovery code and see how quota off
|
|
* recovery is handled (do we roll forward or back or do something else).
|
|
* If we roll forwards or backwards, then we need two separate routines,
|
|
* one that does nothing and one that stamps in the lsn that matters
|
|
* (truly makes the quotaoff irrevocable). If we do something else,
|
|
* then maybe we don't need two.
|
|
*/
|
|
STATIC void
|
|
xfs_qm_qoff_logitem_committing(
|
|
struct xfs_log_item *lip,
|
|
xfs_lsn_t commit_lsn)
|
|
{
|
|
}
|
|
|
|
static const struct xfs_item_ops xfs_qm_qoffend_logitem_ops = {
|
|
.iop_size = xfs_qm_qoff_logitem_size,
|
|
.iop_format = xfs_qm_qoff_logitem_format,
|
|
.iop_pin = xfs_qm_qoff_logitem_pin,
|
|
.iop_unpin = xfs_qm_qoff_logitem_unpin,
|
|
.iop_unlock = xfs_qm_qoff_logitem_unlock,
|
|
.iop_committed = xfs_qm_qoffend_logitem_committed,
|
|
.iop_push = xfs_qm_qoff_logitem_push,
|
|
.iop_committing = xfs_qm_qoff_logitem_committing
|
|
};
|
|
|
|
/*
|
|
* This is the ops vector shared by all quotaoff-start log items.
|
|
*/
|
|
static const struct xfs_item_ops xfs_qm_qoff_logitem_ops = {
|
|
.iop_size = xfs_qm_qoff_logitem_size,
|
|
.iop_format = xfs_qm_qoff_logitem_format,
|
|
.iop_pin = xfs_qm_qoff_logitem_pin,
|
|
.iop_unpin = xfs_qm_qoff_logitem_unpin,
|
|
.iop_unlock = xfs_qm_qoff_logitem_unlock,
|
|
.iop_committed = xfs_qm_qoff_logitem_committed,
|
|
.iop_push = xfs_qm_qoff_logitem_push,
|
|
.iop_committing = xfs_qm_qoff_logitem_committing
|
|
};
|
|
|
|
/*
|
|
* Allocate and initialize an quotaoff item of the correct quota type(s).
|
|
*/
|
|
struct xfs_qoff_logitem *
|
|
xfs_qm_qoff_logitem_init(
|
|
struct xfs_mount *mp,
|
|
struct xfs_qoff_logitem *start,
|
|
uint flags)
|
|
{
|
|
struct xfs_qoff_logitem *qf;
|
|
|
|
qf = kmem_zalloc(sizeof(struct xfs_qoff_logitem), KM_SLEEP);
|
|
|
|
xfs_log_item_init(mp, &qf->qql_item, XFS_LI_QUOTAOFF, start ?
|
|
&xfs_qm_qoffend_logitem_ops : &xfs_qm_qoff_logitem_ops);
|
|
qf->qql_item.li_mountp = mp;
|
|
qf->qql_start_lip = start;
|
|
qf->qql_flags = flags;
|
|
return qf;
|
|
}
|