linux_dsm_epyc7002/fs/xfs/xfs_dquot_item.c

386 lines
9.3 KiB
C
Raw Normal View History

// SPDX-License-Identifier: GPL-2.0
/*
* Copyright (c) 2000-2003 Silicon Graphics, Inc.
* All Rights Reserved.
*/
#include "xfs.h"
#include "xfs_fs.h"
#include "xfs_shared.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_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);
}
/*
* 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));
}
/*
* Callback used to mark a buffer with XFS_LI_FAILED when items in the buffer
* have been failed during writeback
*
* this informs the AIL that the dquot is already flush locked on the next push,
* and acquires a hold on the buffer to ensure that it isn't reclaimed before
* dirty data makes it to disk.
*/
STATIC void
xfs_dquot_item_error(
struct xfs_log_item *lip,
struct xfs_buf *bp)
{
ASSERT(!completion_done(&DQUOT_ITEM(lip)->qli_dquot->q_flush));
xfs_set_li_failed(lip, bp);
}
STATIC uint
xfs: on-stack delayed write buffer lists Queue delwri buffers on a local on-stack list instead of a per-buftarg one, and write back the buffers per-process instead of by waking up xfsbufd. This is now easily doable given that we have very few places left that write delwri buffers: - log recovery: Only done at mount time, and already forcing out the buffers synchronously using xfs_flush_buftarg - quotacheck: Same story. - dquot reclaim: Writes out dirty dquots on the LRU under memory pressure. We might want to look into doing more of this via xfsaild, but it's already more optimal than the synchronous inode reclaim that writes each buffer synchronously. - xfsaild: This is the main beneficiary of the change. By keeping a local list of buffers to write we reduce latency of writing out buffers, and more importably we can remove all the delwri list promotions which were hitting the buffer cache hard under sustained metadata loads. The implementation is very straight forward - xfs_buf_delwri_queue now gets a new list_head pointer that it adds the delwri buffers to, and all callers need to eventually submit the list using xfs_buf_delwi_submit or xfs_buf_delwi_submit_nowait. Buffers that already are on a delwri list are skipped in xfs_buf_delwri_queue, assuming they already are on another delwri list. The biggest change to pass down the buffer list was done to the AIL pushing. Now that we operate on buffers the trylock, push and pushbuf log item methods are merged into a single push routine, which tries to lock the item, and if possible add the buffer that needs writeback to the buffer list. This leads to much simpler code than the previous split but requires the individual IOP_PUSH instances to unlock and reacquire the AIL around calls to blocking routines. Given that xfsailds now also handle writing out buffers, the conditions for log forcing and the sleep times needed some small changes. The most important one is that we consider an AIL busy as long we still have buffers to push, and the other one is that we do increment the pushed LSN for buffers that are under flushing at this moment, but still count them towards the stuck items for restart purposes. Without this we could hammer on stuck items without ever forcing the log and not make progress under heavy random delete workloads on fast flash storage devices. [ Dave Chinner: - rebase on previous patches. - improved comments for XBF_DELWRI_Q handling - fix XBF_ASYNC handling in queue submission (test 106 failure) - rename delwri submit function buffer list parameters for clarity - xfs_efd_item_push() should return XFS_ITEM_PINNED ] Signed-off-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Dave Chinner <dchinner@redhat.com> Reviewed-by: Mark Tinguely <tinguely@sgi.com> Signed-off-by: Ben Myers <bpm@sgi.com>
2012-04-23 12:58:39 +07:00
xfs_qm_dquot_logitem_push(
struct xfs_log_item *lip,
struct list_head *buffer_list)
__releases(&lip->li_ailp->ail_lock)
__acquires(&lip->li_ailp->ail_lock)
{
struct xfs_dquot *dqp = DQUOT_ITEM(lip)->qli_dquot;
struct xfs_buf *bp = lip->li_buf;
xfs: on-stack delayed write buffer lists Queue delwri buffers on a local on-stack list instead of a per-buftarg one, and write back the buffers per-process instead of by waking up xfsbufd. This is now easily doable given that we have very few places left that write delwri buffers: - log recovery: Only done at mount time, and already forcing out the buffers synchronously using xfs_flush_buftarg - quotacheck: Same story. - dquot reclaim: Writes out dirty dquots on the LRU under memory pressure. We might want to look into doing more of this via xfsaild, but it's already more optimal than the synchronous inode reclaim that writes each buffer synchronously. - xfsaild: This is the main beneficiary of the change. By keeping a local list of buffers to write we reduce latency of writing out buffers, and more importably we can remove all the delwri list promotions which were hitting the buffer cache hard under sustained metadata loads. The implementation is very straight forward - xfs_buf_delwri_queue now gets a new list_head pointer that it adds the delwri buffers to, and all callers need to eventually submit the list using xfs_buf_delwi_submit or xfs_buf_delwi_submit_nowait. Buffers that already are on a delwri list are skipped in xfs_buf_delwri_queue, assuming they already are on another delwri list. The biggest change to pass down the buffer list was done to the AIL pushing. Now that we operate on buffers the trylock, push and pushbuf log item methods are merged into a single push routine, which tries to lock the item, and if possible add the buffer that needs writeback to the buffer list. This leads to much simpler code than the previous split but requires the individual IOP_PUSH instances to unlock and reacquire the AIL around calls to blocking routines. Given that xfsailds now also handle writing out buffers, the conditions for log forcing and the sleep times needed some small changes. The most important one is that we consider an AIL busy as long we still have buffers to push, and the other one is that we do increment the pushed LSN for buffers that are under flushing at this moment, but still count them towards the stuck items for restart purposes. Without this we could hammer on stuck items without ever forcing the log and not make progress under heavy random delete workloads on fast flash storage devices. [ Dave Chinner: - rebase on previous patches. - improved comments for XBF_DELWRI_Q handling - fix XBF_ASYNC handling in queue submission (test 106 failure) - rename delwri submit function buffer list parameters for clarity - xfs_efd_item_push() should return XFS_ITEM_PINNED ] Signed-off-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Dave Chinner <dchinner@redhat.com> Reviewed-by: Mark Tinguely <tinguely@sgi.com> Signed-off-by: Ben Myers <bpm@sgi.com>
2012-04-23 12:58:39 +07:00
uint rval = XFS_ITEM_SUCCESS;
int error;
if (atomic_read(&dqp->q_pincount) > 0)
return XFS_ITEM_PINNED;
/*
* The buffer containing this item failed to be written back
* previously. Resubmit the buffer for IO
*/
if (test_bit(XFS_LI_FAILED, &lip->li_flags)) {
if (!xfs_buf_trylock(bp))
return XFS_ITEM_LOCKED;
if (!xfs_buf_resubmit_failed_buffers(bp, buffer_list))
rval = XFS_ITEM_FLUSHING;
xfs_buf_unlock(bp);
return rval;
}
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) {
xfs: on-stack delayed write buffer lists Queue delwri buffers on a local on-stack list instead of a per-buftarg one, and write back the buffers per-process instead of by waking up xfsbufd. This is now easily doable given that we have very few places left that write delwri buffers: - log recovery: Only done at mount time, and already forcing out the buffers synchronously using xfs_flush_buftarg - quotacheck: Same story. - dquot reclaim: Writes out dirty dquots on the LRU under memory pressure. We might want to look into doing more of this via xfsaild, but it's already more optimal than the synchronous inode reclaim that writes each buffer synchronously. - xfsaild: This is the main beneficiary of the change. By keeping a local list of buffers to write we reduce latency of writing out buffers, and more importably we can remove all the delwri list promotions which were hitting the buffer cache hard under sustained metadata loads. The implementation is very straight forward - xfs_buf_delwri_queue now gets a new list_head pointer that it adds the delwri buffers to, and all callers need to eventually submit the list using xfs_buf_delwi_submit or xfs_buf_delwi_submit_nowait. Buffers that already are on a delwri list are skipped in xfs_buf_delwri_queue, assuming they already are on another delwri list. The biggest change to pass down the buffer list was done to the AIL pushing. Now that we operate on buffers the trylock, push and pushbuf log item methods are merged into a single push routine, which tries to lock the item, and if possible add the buffer that needs writeback to the buffer list. This leads to much simpler code than the previous split but requires the individual IOP_PUSH instances to unlock and reacquire the AIL around calls to blocking routines. Given that xfsailds now also handle writing out buffers, the conditions for log forcing and the sleep times needed some small changes. The most important one is that we consider an AIL busy as long we still have buffers to push, and the other one is that we do increment the pushed LSN for buffers that are under flushing at this moment, but still count them towards the stuck items for restart purposes. Without this we could hammer on stuck items without ever forcing the log and not make progress under heavy random delete workloads on fast flash storage devices. [ Dave Chinner: - rebase on previous patches. - improved comments for XBF_DELWRI_Q handling - fix XBF_ASYNC handling in queue submission (test 106 failure) - rename delwri submit function buffer list parameters for clarity - xfs_efd_item_push() should return XFS_ITEM_PINNED ] Signed-off-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Dave Chinner <dchinner@redhat.com> Reviewed-by: Mark Tinguely <tinguely@sgi.com> Signed-off-by: Ben Myers <bpm@sgi.com>
2012-04-23 12:58:39 +07:00
rval = XFS_ITEM_PINNED;
goto out_unlock;
}
xfs: on-stack delayed write buffer lists Queue delwri buffers on a local on-stack list instead of a per-buftarg one, and write back the buffers per-process instead of by waking up xfsbufd. This is now easily doable given that we have very few places left that write delwri buffers: - log recovery: Only done at mount time, and already forcing out the buffers synchronously using xfs_flush_buftarg - quotacheck: Same story. - dquot reclaim: Writes out dirty dquots on the LRU under memory pressure. We might want to look into doing more of this via xfsaild, but it's already more optimal than the synchronous inode reclaim that writes each buffer synchronously. - xfsaild: This is the main beneficiary of the change. By keeping a local list of buffers to write we reduce latency of writing out buffers, and more importably we can remove all the delwri list promotions which were hitting the buffer cache hard under sustained metadata loads. The implementation is very straight forward - xfs_buf_delwri_queue now gets a new list_head pointer that it adds the delwri buffers to, and all callers need to eventually submit the list using xfs_buf_delwi_submit or xfs_buf_delwi_submit_nowait. Buffers that already are on a delwri list are skipped in xfs_buf_delwri_queue, assuming they already are on another delwri list. The biggest change to pass down the buffer list was done to the AIL pushing. Now that we operate on buffers the trylock, push and pushbuf log item methods are merged into a single push routine, which tries to lock the item, and if possible add the buffer that needs writeback to the buffer list. This leads to much simpler code than the previous split but requires the individual IOP_PUSH instances to unlock and reacquire the AIL around calls to blocking routines. Given that xfsailds now also handle writing out buffers, the conditions for log forcing and the sleep times needed some small changes. The most important one is that we consider an AIL busy as long we still have buffers to push, and the other one is that we do increment the pushed LSN for buffers that are under flushing at this moment, but still count them towards the stuck items for restart purposes. Without this we could hammer on stuck items without ever forcing the log and not make progress under heavy random delete workloads on fast flash storage devices. [ Dave Chinner: - rebase on previous patches. - improved comments for XBF_DELWRI_Q handling - fix XBF_ASYNC handling in queue submission (test 106 failure) - rename delwri submit function buffer list parameters for clarity - xfs_efd_item_push() should return XFS_ITEM_PINNED ] Signed-off-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Dave Chinner <dchinner@redhat.com> Reviewed-by: Mark Tinguely <tinguely@sgi.com> Signed-off-by: Ben Myers <bpm@sgi.com>
2012-04-23 12:58:39 +07:00
/*
* 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)) {
xfs: on-stack delayed write buffer lists Queue delwri buffers on a local on-stack list instead of a per-buftarg one, and write back the buffers per-process instead of by waking up xfsbufd. This is now easily doable given that we have very few places left that write delwri buffers: - log recovery: Only done at mount time, and already forcing out the buffers synchronously using xfs_flush_buftarg - quotacheck: Same story. - dquot reclaim: Writes out dirty dquots on the LRU under memory pressure. We might want to look into doing more of this via xfsaild, but it's already more optimal than the synchronous inode reclaim that writes each buffer synchronously. - xfsaild: This is the main beneficiary of the change. By keeping a local list of buffers to write we reduce latency of writing out buffers, and more importably we can remove all the delwri list promotions which were hitting the buffer cache hard under sustained metadata loads. The implementation is very straight forward - xfs_buf_delwri_queue now gets a new list_head pointer that it adds the delwri buffers to, and all callers need to eventually submit the list using xfs_buf_delwi_submit or xfs_buf_delwi_submit_nowait. Buffers that already are on a delwri list are skipped in xfs_buf_delwri_queue, assuming they already are on another delwri list. The biggest change to pass down the buffer list was done to the AIL pushing. Now that we operate on buffers the trylock, push and pushbuf log item methods are merged into a single push routine, which tries to lock the item, and if possible add the buffer that needs writeback to the buffer list. This leads to much simpler code than the previous split but requires the individual IOP_PUSH instances to unlock and reacquire the AIL around calls to blocking routines. Given that xfsailds now also handle writing out buffers, the conditions for log forcing and the sleep times needed some small changes. The most important one is that we consider an AIL busy as long we still have buffers to push, and the other one is that we do increment the pushed LSN for buffers that are under flushing at this moment, but still count them towards the stuck items for restart purposes. Without this we could hammer on stuck items without ever forcing the log and not make progress under heavy random delete workloads on fast flash storage devices. [ Dave Chinner: - rebase on previous patches. - improved comments for XBF_DELWRI_Q handling - fix XBF_ASYNC handling in queue submission (test 106 failure) - rename delwri submit function buffer list parameters for clarity - xfs_efd_item_push() should return XFS_ITEM_PINNED ] Signed-off-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Dave Chinner <dchinner@redhat.com> Reviewed-by: Mark Tinguely <tinguely@sgi.com> Signed-off-by: Ben Myers <bpm@sgi.com>
2012-04-23 12:58:39 +07:00
rval = XFS_ITEM_FLUSHING;
goto out_unlock;
}
spin_unlock(&lip->li_ailp->ail_lock);
xfs: on-stack delayed write buffer lists Queue delwri buffers on a local on-stack list instead of a per-buftarg one, and write back the buffers per-process instead of by waking up xfsbufd. This is now easily doable given that we have very few places left that write delwri buffers: - log recovery: Only done at mount time, and already forcing out the buffers synchronously using xfs_flush_buftarg - quotacheck: Same story. - dquot reclaim: Writes out dirty dquots on the LRU under memory pressure. We might want to look into doing more of this via xfsaild, but it's already more optimal than the synchronous inode reclaim that writes each buffer synchronously. - xfsaild: This is the main beneficiary of the change. By keeping a local list of buffers to write we reduce latency of writing out buffers, and more importably we can remove all the delwri list promotions which were hitting the buffer cache hard under sustained metadata loads. The implementation is very straight forward - xfs_buf_delwri_queue now gets a new list_head pointer that it adds the delwri buffers to, and all callers need to eventually submit the list using xfs_buf_delwi_submit or xfs_buf_delwi_submit_nowait. Buffers that already are on a delwri list are skipped in xfs_buf_delwri_queue, assuming they already are on another delwri list. The biggest change to pass down the buffer list was done to the AIL pushing. Now that we operate on buffers the trylock, push and pushbuf log item methods are merged into a single push routine, which tries to lock the item, and if possible add the buffer that needs writeback to the buffer list. This leads to much simpler code than the previous split but requires the individual IOP_PUSH instances to unlock and reacquire the AIL around calls to blocking routines. Given that xfsailds now also handle writing out buffers, the conditions for log forcing and the sleep times needed some small changes. The most important one is that we consider an AIL busy as long we still have buffers to push, and the other one is that we do increment the pushed LSN for buffers that are under flushing at this moment, but still count them towards the stuck items for restart purposes. Without this we could hammer on stuck items without ever forcing the log and not make progress under heavy random delete workloads on fast flash storage devices. [ Dave Chinner: - rebase on previous patches. - improved comments for XBF_DELWRI_Q handling - fix XBF_ASYNC handling in queue submission (test 106 failure) - rename delwri submit function buffer list parameters for clarity - xfs_efd_item_push() should return XFS_ITEM_PINNED ] Signed-off-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Dave Chinner <dchinner@redhat.com> Reviewed-by: Mark Tinguely <tinguely@sgi.com> Signed-off-by: Ben Myers <bpm@sgi.com>
2012-04-23 12:58:39 +07:00
error = xfs_qm_dqflush(dqp, &bp);
if (!error) {
xfs: on-stack delayed write buffer lists Queue delwri buffers on a local on-stack list instead of a per-buftarg one, and write back the buffers per-process instead of by waking up xfsbufd. This is now easily doable given that we have very few places left that write delwri buffers: - log recovery: Only done at mount time, and already forcing out the buffers synchronously using xfs_flush_buftarg - quotacheck: Same story. - dquot reclaim: Writes out dirty dquots on the LRU under memory pressure. We might want to look into doing more of this via xfsaild, but it's already more optimal than the synchronous inode reclaim that writes each buffer synchronously. - xfsaild: This is the main beneficiary of the change. By keeping a local list of buffers to write we reduce latency of writing out buffers, and more importably we can remove all the delwri list promotions which were hitting the buffer cache hard under sustained metadata loads. The implementation is very straight forward - xfs_buf_delwri_queue now gets a new list_head pointer that it adds the delwri buffers to, and all callers need to eventually submit the list using xfs_buf_delwi_submit or xfs_buf_delwi_submit_nowait. Buffers that already are on a delwri list are skipped in xfs_buf_delwri_queue, assuming they already are on another delwri list. The biggest change to pass down the buffer list was done to the AIL pushing. Now that we operate on buffers the trylock, push and pushbuf log item methods are merged into a single push routine, which tries to lock the item, and if possible add the buffer that needs writeback to the buffer list. This leads to much simpler code than the previous split but requires the individual IOP_PUSH instances to unlock and reacquire the AIL around calls to blocking routines. Given that xfsailds now also handle writing out buffers, the conditions for log forcing and the sleep times needed some small changes. The most important one is that we consider an AIL busy as long we still have buffers to push, and the other one is that we do increment the pushed LSN for buffers that are under flushing at this moment, but still count them towards the stuck items for restart purposes. Without this we could hammer on stuck items without ever forcing the log and not make progress under heavy random delete workloads on fast flash storage devices. [ Dave Chinner: - rebase on previous patches. - improved comments for XBF_DELWRI_Q handling - fix XBF_ASYNC handling in queue submission (test 106 failure) - rename delwri submit function buffer list parameters for clarity - xfs_efd_item_push() should return XFS_ITEM_PINNED ] Signed-off-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Dave Chinner <dchinner@redhat.com> Reviewed-by: Mark Tinguely <tinguely@sgi.com> Signed-off-by: Ben Myers <bpm@sgi.com>
2012-04-23 12:58:39 +07:00
if (!xfs_buf_delwri_queue(bp, buffer_list))
rval = XFS_ITEM_FLUSHING;
xfs_buf_relse(bp);
} else if (error == -EAGAIN)
rval = XFS_ITEM_LOCKED;
spin_lock(&lip->li_ailp->ail_lock);
xfs: on-stack delayed write buffer lists Queue delwri buffers on a local on-stack list instead of a per-buftarg one, and write back the buffers per-process instead of by waking up xfsbufd. This is now easily doable given that we have very few places left that write delwri buffers: - log recovery: Only done at mount time, and already forcing out the buffers synchronously using xfs_flush_buftarg - quotacheck: Same story. - dquot reclaim: Writes out dirty dquots on the LRU under memory pressure. We might want to look into doing more of this via xfsaild, but it's already more optimal than the synchronous inode reclaim that writes each buffer synchronously. - xfsaild: This is the main beneficiary of the change. By keeping a local list of buffers to write we reduce latency of writing out buffers, and more importably we can remove all the delwri list promotions which were hitting the buffer cache hard under sustained metadata loads. The implementation is very straight forward - xfs_buf_delwri_queue now gets a new list_head pointer that it adds the delwri buffers to, and all callers need to eventually submit the list using xfs_buf_delwi_submit or xfs_buf_delwi_submit_nowait. Buffers that already are on a delwri list are skipped in xfs_buf_delwri_queue, assuming they already are on another delwri list. The biggest change to pass down the buffer list was done to the AIL pushing. Now that we operate on buffers the trylock, push and pushbuf log item methods are merged into a single push routine, which tries to lock the item, and if possible add the buffer that needs writeback to the buffer list. This leads to much simpler code than the previous split but requires the individual IOP_PUSH instances to unlock and reacquire the AIL around calls to blocking routines. Given that xfsailds now also handle writing out buffers, the conditions for log forcing and the sleep times needed some small changes. The most important one is that we consider an AIL busy as long we still have buffers to push, and the other one is that we do increment the pushed LSN for buffers that are under flushing at this moment, but still count them towards the stuck items for restart purposes. Without this we could hammer on stuck items without ever forcing the log and not make progress under heavy random delete workloads on fast flash storage devices. [ Dave Chinner: - rebase on previous patches. - improved comments for XBF_DELWRI_Q handling - fix XBF_ASYNC handling in queue submission (test 106 failure) - rename delwri submit function buffer list parameters for clarity - xfs_efd_item_push() should return XFS_ITEM_PINNED ] Signed-off-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Dave Chinner <dchinner@redhat.com> Reviewed-by: Mark Tinguely <tinguely@sgi.com> Signed-off-by: Ben Myers <bpm@sgi.com>
2012-04-23 12:58:39 +07:00
out_unlock:
xfs_dqunlock(dqp);
return rval;
}
STATIC void
xfs: split iop_unlock The iop_unlock method is called when comitting or cancelling a transaction. In the latter case, the transaction may or may not be aborted. While there is no known problem with the current code in practice, this implementation is limited in that any log item implementation that might want to differentiate between a commit and a cancellation must rely on the aborted state. The aborted bit is only set when the cancelled transaction is dirty, however. This means that there is no way to distinguish between a commit and a clean transaction cancellation. For example, intent log items currently rely on this distinction. The log item is either transferred to the CIL on commit or released on transaction cancel. There is currently no possibility for a clean intent log item in a transaction, but if that state is ever introduced a cancel of such a transaction will immediately result in memory leaks of the associated log item(s). This is an interface deficiency and landmine. To clean this up, replace the iop_unlock method with an iop_release method that is specific to transaction cancel. The existing iop_committing method occurs at the same time as iop_unlock in the commit path and there is no need for two separate callbacks here. Overload the iop_committing method with the current commit time iop_unlock implementations to eliminate the need for the latter and further simplify the interface. Signed-off-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Brian Foster <bfoster@redhat.com> Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com> Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
2019-06-29 09:27:32 +07:00
xfs_qm_dquot_logitem_release(
struct xfs_log_item *lip)
{
struct xfs_dquot *dqp = DQUOT_ITEM(lip)->qli_dquot;
ASSERT(XFS_DQ_IS_LOCKED(dqp));
/*
* 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);
}
xfs: split iop_unlock The iop_unlock method is called when comitting or cancelling a transaction. In the latter case, the transaction may or may not be aborted. While there is no known problem with the current code in practice, this implementation is limited in that any log item implementation that might want to differentiate between a commit and a cancellation must rely on the aborted state. The aborted bit is only set when the cancelled transaction is dirty, however. This means that there is no way to distinguish between a commit and a clean transaction cancellation. For example, intent log items currently rely on this distinction. The log item is either transferred to the CIL on commit or released on transaction cancel. There is currently no possibility for a clean intent log item in a transaction, but if that state is ever introduced a cancel of such a transaction will immediately result in memory leaks of the associated log item(s). This is an interface deficiency and landmine. To clean this up, replace the iop_unlock method with an iop_release method that is specific to transaction cancel. The existing iop_committing method occurs at the same time as iop_unlock in the commit path and there is no need for two separate callbacks here. Overload the iop_committing method with the current commit time iop_unlock implementations to eliminate the need for the latter and further simplify the interface. Signed-off-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Brian Foster <bfoster@redhat.com> Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com> Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
2019-06-29 09:27:32 +07:00
STATIC void
xfs_qm_dquot_logitem_committing(
struct xfs_log_item *lip,
xfs_lsn_t commit_lsn)
{
return xfs_qm_dquot_logitem_release(lip);
}
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,
xfs: split iop_unlock The iop_unlock method is called when comitting or cancelling a transaction. In the latter case, the transaction may or may not be aborted. While there is no known problem with the current code in practice, this implementation is limited in that any log item implementation that might want to differentiate between a commit and a cancellation must rely on the aborted state. The aborted bit is only set when the cancelled transaction is dirty, however. This means that there is no way to distinguish between a commit and a clean transaction cancellation. For example, intent log items currently rely on this distinction. The log item is either transferred to the CIL on commit or released on transaction cancel. There is currently no possibility for a clean intent log item in a transaction, but if that state is ever introduced a cancel of such a transaction will immediately result in memory leaks of the associated log item(s). This is an interface deficiency and landmine. To clean this up, replace the iop_unlock method with an iop_release method that is specific to transaction cancel. The existing iop_committing method occurs at the same time as iop_unlock in the commit path and there is no need for two separate callbacks here. Overload the iop_committing method with the current commit time iop_unlock implementations to eliminate the need for the latter and further simplify the interface. Signed-off-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Brian Foster <bfoster@redhat.com> Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com> Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
2019-06-29 09:27:32 +07:00
.iop_release = xfs_qm_dquot_logitem_release,
.iop_committing = xfs_qm_dquot_logitem_committing,
.iop_push = xfs_qm_dquot_logitem_push,
.iop_error = xfs_dquot_item_error
};
/*
* 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));
}
/*
xfs: on-stack delayed write buffer lists Queue delwri buffers on a local on-stack list instead of a per-buftarg one, and write back the buffers per-process instead of by waking up xfsbufd. This is now easily doable given that we have very few places left that write delwri buffers: - log recovery: Only done at mount time, and already forcing out the buffers synchronously using xfs_flush_buftarg - quotacheck: Same story. - dquot reclaim: Writes out dirty dquots on the LRU under memory pressure. We might want to look into doing more of this via xfsaild, but it's already more optimal than the synchronous inode reclaim that writes each buffer synchronously. - xfsaild: This is the main beneficiary of the change. By keeping a local list of buffers to write we reduce latency of writing out buffers, and more importably we can remove all the delwri list promotions which were hitting the buffer cache hard under sustained metadata loads. The implementation is very straight forward - xfs_buf_delwri_queue now gets a new list_head pointer that it adds the delwri buffers to, and all callers need to eventually submit the list using xfs_buf_delwi_submit or xfs_buf_delwi_submit_nowait. Buffers that already are on a delwri list are skipped in xfs_buf_delwri_queue, assuming they already are on another delwri list. The biggest change to pass down the buffer list was done to the AIL pushing. Now that we operate on buffers the trylock, push and pushbuf log item methods are merged into a single push routine, which tries to lock the item, and if possible add the buffer that needs writeback to the buffer list. This leads to much simpler code than the previous split but requires the individual IOP_PUSH instances to unlock and reacquire the AIL around calls to blocking routines. Given that xfsailds now also handle writing out buffers, the conditions for log forcing and the sleep times needed some small changes. The most important one is that we consider an AIL busy as long we still have buffers to push, and the other one is that we do increment the pushed LSN for buffers that are under flushing at this moment, but still count them towards the stuck items for restart purposes. Without this we could hammer on stuck items without ever forcing the log and not make progress under heavy random delete workloads on fast flash storage devices. [ Dave Chinner: - rebase on previous patches. - improved comments for XBF_DELWRI_Q handling - fix XBF_ASYNC handling in queue submission (test 106 failure) - rename delwri submit function buffer list parameters for clarity - xfs_efd_item_push() should return XFS_ITEM_PINNED ] Signed-off-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Dave Chinner <dchinner@redhat.com> Reviewed-by: Mark Tinguely <tinguely@sgi.com> Signed-off-by: Ben Myers <bpm@sgi.com>
2012-04-23 12:58:39 +07:00
* 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: on-stack delayed write buffer lists Queue delwri buffers on a local on-stack list instead of a per-buftarg one, and write back the buffers per-process instead of by waking up xfsbufd. This is now easily doable given that we have very few places left that write delwri buffers: - log recovery: Only done at mount time, and already forcing out the buffers synchronously using xfs_flush_buftarg - quotacheck: Same story. - dquot reclaim: Writes out dirty dquots on the LRU under memory pressure. We might want to look into doing more of this via xfsaild, but it's already more optimal than the synchronous inode reclaim that writes each buffer synchronously. - xfsaild: This is the main beneficiary of the change. By keeping a local list of buffers to write we reduce latency of writing out buffers, and more importably we can remove all the delwri list promotions which were hitting the buffer cache hard under sustained metadata loads. The implementation is very straight forward - xfs_buf_delwri_queue now gets a new list_head pointer that it adds the delwri buffers to, and all callers need to eventually submit the list using xfs_buf_delwi_submit or xfs_buf_delwi_submit_nowait. Buffers that already are on a delwri list are skipped in xfs_buf_delwri_queue, assuming they already are on another delwri list. The biggest change to pass down the buffer list was done to the AIL pushing. Now that we operate on buffers the trylock, push and pushbuf log item methods are merged into a single push routine, which tries to lock the item, and if possible add the buffer that needs writeback to the buffer list. This leads to much simpler code than the previous split but requires the individual IOP_PUSH instances to unlock and reacquire the AIL around calls to blocking routines. Given that xfsailds now also handle writing out buffers, the conditions for log forcing and the sleep times needed some small changes. The most important one is that we consider an AIL busy as long we still have buffers to push, and the other one is that we do increment the pushed LSN for buffers that are under flushing at this moment, but still count them towards the stuck items for restart purposes. Without this we could hammer on stuck items without ever forcing the log and not make progress under heavy random delete workloads on fast flash storage devices. [ Dave Chinner: - rebase on previous patches. - improved comments for XBF_DELWRI_Q handling - fix XBF_ASYNC handling in queue submission (test 106 failure) - rename delwri submit function buffer list parameters for clarity - xfs_efd_item_push() should return XFS_ITEM_PINNED ] Signed-off-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Dave Chinner <dchinner@redhat.com> Reviewed-by: Mark Tinguely <tinguely@sgi.com> Signed-off-by: Ben Myers <bpm@sgi.com>
2012-04-23 12:58:39 +07:00
xfs_qm_qoff_logitem_push(
struct xfs_log_item *lip,
struct list_head *buffer_list)
{
return XFS_ITEM_LOCKED;
}
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;
xfs_qm_qoff_logitem_relse(qfs);
xfs: allocate log vector buffers outside CIL context lock 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>
2016-07-22 06:52:35 +07:00
kmem_free(lip->li_lv_shadow);
kmem_free(qfe);
return (xfs_lsn_t)-1;
}
xfs: fix unmount hang and memory leak on shutdown during quotaoff AIL removal of the quotaoff start intent and free of both quotaoff intents is currently limited to the ->iop_committed() handler of the end intent. This executes when the end intent is committed to the on-disk log and marks the completion of the operation. The problem with this is it assumes the success of the operation. If a shutdown or other error occurs during the quotaoff, it's possible for the quotaoff task to exit without removing the start intent from the AIL. This results in an unmount hang as the AIL cannot be emptied. Further, no other codepath frees the intents and so this is also a memory leak vector. First, update the high level quotaoff error path to directly remove and free the quotaoff start intent if it still exists in the AIL at the time of the error. Next, update both of the start and end quotaoff intents with an ->iop_release() callback to properly handle transaction abort. This means that If the quotaoff start transaction aborts, it frees the start intent in the transaction commit path. If the filesystem shuts down before the end transaction allocates, the quotaoff sequence removes and frees the start intent. If the end transaction aborts, it removes the start intent and frees both. This ensures that a shutdown does not result in a hung unmount and that memory is not leaked regardless of when a quotaoff error occurs. Signed-off-by: Brian Foster <bfoster@redhat.com> Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com> Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Christoph Hellwig <hch@lst.de>
2020-03-17 04:26:09 +07:00
STATIC void
xfs_qm_qoff_logitem_release(
struct xfs_log_item *lip)
{
struct xfs_qoff_logitem *qoff = QOFF_ITEM(lip);
if (test_bit(XFS_LI_ABORTED, &lip->li_flags)) {
if (qoff->qql_start_lip)
xfs_qm_qoff_logitem_relse(qoff->qql_start_lip);
xfs_qm_qoff_logitem_relse(qoff);
}
}
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_committed = xfs_qm_qoffend_logitem_committed,
.iop_push = xfs_qm_qoff_logitem_push,
xfs: fix unmount hang and memory leak on shutdown during quotaoff AIL removal of the quotaoff start intent and free of both quotaoff intents is currently limited to the ->iop_committed() handler of the end intent. This executes when the end intent is committed to the on-disk log and marks the completion of the operation. The problem with this is it assumes the success of the operation. If a shutdown or other error occurs during the quotaoff, it's possible for the quotaoff task to exit without removing the start intent from the AIL. This results in an unmount hang as the AIL cannot be emptied. Further, no other codepath frees the intents and so this is also a memory leak vector. First, update the high level quotaoff error path to directly remove and free the quotaoff start intent if it still exists in the AIL at the time of the error. Next, update both of the start and end quotaoff intents with an ->iop_release() callback to properly handle transaction abort. This means that If the quotaoff start transaction aborts, it frees the start intent in the transaction commit path. If the filesystem shuts down before the end transaction allocates, the quotaoff sequence removes and frees the start intent. If the end transaction aborts, it removes the start intent and frees both. This ensures that a shutdown does not result in a hung unmount and that memory is not leaked regardless of when a quotaoff error occurs. Signed-off-by: Brian Foster <bfoster@redhat.com> Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com> Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Christoph Hellwig <hch@lst.de>
2020-03-17 04:26:09 +07:00
.iop_release = xfs_qm_qoff_logitem_release,
};
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_push = xfs_qm_qoff_logitem_push,
xfs: fix unmount hang and memory leak on shutdown during quotaoff AIL removal of the quotaoff start intent and free of both quotaoff intents is currently limited to the ->iop_committed() handler of the end intent. This executes when the end intent is committed to the on-disk log and marks the completion of the operation. The problem with this is it assumes the success of the operation. If a shutdown or other error occurs during the quotaoff, it's possible for the quotaoff task to exit without removing the start intent from the AIL. This results in an unmount hang as the AIL cannot be emptied. Further, no other codepath frees the intents and so this is also a memory leak vector. First, update the high level quotaoff error path to directly remove and free the quotaoff start intent if it still exists in the AIL at the time of the error. Next, update both of the start and end quotaoff intents with an ->iop_release() callback to properly handle transaction abort. This means that If the quotaoff start transaction aborts, it frees the start intent in the transaction commit path. If the filesystem shuts down before the end transaction allocates, the quotaoff sequence removes and frees the start intent. If the end transaction aborts, it removes the start intent and frees both. This ensures that a shutdown does not result in a hung unmount and that memory is not leaked regardless of when a quotaoff error occurs. Signed-off-by: Brian Foster <bfoster@redhat.com> Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com> Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Christoph Hellwig <hch@lst.de>
2020-03-17 04:26:09 +07:00
.iop_release = xfs_qm_qoff_logitem_release,
};
/*
* Delete the quotaoff intent from the AIL and free it. On success,
* this should only be called for the start item. It can be used for
* either on shutdown or abort.
*/
void
xfs_qm_qoff_logitem_relse(
struct xfs_qoff_logitem *qoff)
{
struct xfs_log_item *lip = &qoff->qql_item;
ASSERT(test_bit(XFS_LI_IN_AIL, &lip->li_flags) ||
test_bit(XFS_LI_ABORTED, &lip->li_flags) ||
XFS_FORCED_SHUTDOWN(lip->li_mountp));
xfs_trans_ail_remove(lip, SHUTDOWN_LOG_IO_ERROR);
kmem_free(lip->li_lv_shadow);
kmem_free(qoff);
}
/*
* 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), 0);
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;
}