2008-10-30 07:06:08 +01:00
|
|
|
/*
|
|
|
|
* Copyright (c) 2000-2005 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_types.h"
|
|
|
|
#include "xfs_log.h"
|
2012-10-08 12:56:02 +02:00
|
|
|
#include "xfs_log_priv.h"
|
2008-10-30 07:06:08 +01:00
|
|
|
#include "xfs_inum.h"
|
|
|
|
#include "xfs_trans.h"
|
2011-04-08 04:45:07 +02:00
|
|
|
#include "xfs_trans_priv.h"
|
2008-10-30 07:06:08 +01:00
|
|
|
#include "xfs_sb.h"
|
|
|
|
#include "xfs_ag.h"
|
|
|
|
#include "xfs_mount.h"
|
|
|
|
#include "xfs_bmap_btree.h"
|
|
|
|
#include "xfs_inode.h"
|
|
|
|
#include "xfs_dinode.h"
|
|
|
|
#include "xfs_error.h"
|
|
|
|
#include "xfs_filestream.h"
|
|
|
|
#include "xfs_vnodeops.h"
|
|
|
|
#include "xfs_inode_item.h"
|
2009-06-08 15:33:32 +02:00
|
|
|
#include "xfs_quota.h"
|
2009-12-15 00:14:59 +01:00
|
|
|
#include "xfs_trace.h"
|
2010-08-24 03:46:31 +02:00
|
|
|
#include "xfs_fsops.h"
|
2008-10-30 07:06:08 +01:00
|
|
|
|
2008-10-30 07:06:18 +01:00
|
|
|
#include <linux/kthread.h>
|
|
|
|
#include <linux/freezer.h>
|
|
|
|
|
2011-04-08 04:45:07 +02:00
|
|
|
struct workqueue_struct *xfs_syncd_wq; /* sync workqueue */
|
|
|
|
|
2010-09-28 04:28:19 +02:00
|
|
|
/*
|
|
|
|
* The inode lookup is done in batches to keep the amount of lock traffic and
|
|
|
|
* radix tree lookups to a minimum. The batch size is a trade off between
|
|
|
|
* lookup reduction and stack usage. This is in the reclaim path, so we can't
|
|
|
|
* be too greedy.
|
|
|
|
*/
|
|
|
|
#define XFS_LOOKUP_BATCH 32
|
|
|
|
|
2010-09-28 04:28:06 +02:00
|
|
|
STATIC int
|
|
|
|
xfs_inode_ag_walk_grab(
|
|
|
|
struct xfs_inode *ip)
|
|
|
|
{
|
|
|
|
struct inode *inode = VFS_I(ip);
|
|
|
|
|
2010-12-17 07:29:43 +01:00
|
|
|
ASSERT(rcu_read_lock_held());
|
|
|
|
|
|
|
|
/*
|
|
|
|
* check for stale RCU freed inode
|
|
|
|
*
|
|
|
|
* If the inode has been reallocated, it doesn't matter if it's not in
|
|
|
|
* the AG we are walking - we are walking for writeback, so if it
|
|
|
|
* passes all the "valid inode" checks and is dirty, then we'll write
|
|
|
|
* it back anyway. If it has been reallocated and still being
|
|
|
|
* initialised, the XFS_INEW check below will catch it.
|
|
|
|
*/
|
|
|
|
spin_lock(&ip->i_flags_lock);
|
|
|
|
if (!ip->i_ino)
|
|
|
|
goto out_unlock_noent;
|
|
|
|
|
|
|
|
/* avoid new or reclaimable inodes. Leave for reclaim code to flush */
|
|
|
|
if (__xfs_iflags_test(ip, XFS_INEW | XFS_IRECLAIMABLE | XFS_IRECLAIM))
|
|
|
|
goto out_unlock_noent;
|
|
|
|
spin_unlock(&ip->i_flags_lock);
|
|
|
|
|
2010-09-28 04:28:06 +02:00
|
|
|
/* nothing to sync during shutdown */
|
|
|
|
if (XFS_FORCED_SHUTDOWN(ip->i_mount))
|
|
|
|
return EFSCORRUPTED;
|
|
|
|
|
|
|
|
/* If we can't grab the inode, it must on it's way to reclaim. */
|
|
|
|
if (!igrab(inode))
|
|
|
|
return ENOENT;
|
|
|
|
|
|
|
|
if (is_bad_inode(inode)) {
|
|
|
|
IRELE(ip);
|
|
|
|
return ENOENT;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* inode is valid */
|
|
|
|
return 0;
|
2010-12-17 07:29:43 +01:00
|
|
|
|
|
|
|
out_unlock_noent:
|
|
|
|
spin_unlock(&ip->i_flags_lock);
|
|
|
|
return ENOENT;
|
2010-09-28 04:28:06 +02:00
|
|
|
}
|
|
|
|
|
2009-06-08 15:35:14 +02:00
|
|
|
STATIC int
|
|
|
|
xfs_inode_ag_walk(
|
|
|
|
struct xfs_mount *mp,
|
2010-01-11 12:47:40 +01:00
|
|
|
struct xfs_perag *pag,
|
2009-06-08 15:35:14 +02:00
|
|
|
int (*execute)(struct xfs_inode *ip,
|
|
|
|
struct xfs_perag *pag, int flags),
|
2010-09-24 10:40:15 +02:00
|
|
|
int flags)
|
2009-06-08 15:35:14 +02:00
|
|
|
{
|
|
|
|
uint32_t first_index;
|
|
|
|
int last_error = 0;
|
|
|
|
int skipped;
|
2010-09-24 10:40:15 +02:00
|
|
|
int done;
|
2010-09-28 04:28:19 +02:00
|
|
|
int nr_found;
|
2009-06-08 15:35:14 +02:00
|
|
|
|
|
|
|
restart:
|
2010-09-24 10:40:15 +02:00
|
|
|
done = 0;
|
2009-06-08 15:35:14 +02:00
|
|
|
skipped = 0;
|
|
|
|
first_index = 0;
|
2010-09-28 04:28:19 +02:00
|
|
|
nr_found = 0;
|
2009-06-08 15:35:14 +02:00
|
|
|
do {
|
2010-09-28 04:28:19 +02:00
|
|
|
struct xfs_inode *batch[XFS_LOOKUP_BATCH];
|
2009-06-08 15:35:14 +02:00
|
|
|
int error = 0;
|
2010-09-28 04:28:19 +02:00
|
|
|
int i;
|
2009-06-08 15:35:14 +02:00
|
|
|
|
2010-12-17 07:29:43 +01:00
|
|
|
rcu_read_lock();
|
2010-09-24 10:40:15 +02:00
|
|
|
nr_found = radix_tree_gang_lookup(&pag->pag_ici_root,
|
2010-09-28 04:28:19 +02:00
|
|
|
(void **)batch, first_index,
|
|
|
|
XFS_LOOKUP_BATCH);
|
2010-09-24 10:40:15 +02:00
|
|
|
if (!nr_found) {
|
2010-12-17 07:29:43 +01:00
|
|
|
rcu_read_unlock();
|
2009-06-08 15:35:14 +02:00
|
|
|
break;
|
2010-01-11 00:51:45 +01:00
|
|
|
}
|
2009-06-08 15:35:14 +02:00
|
|
|
|
2010-09-24 10:40:15 +02:00
|
|
|
/*
|
2010-09-28 04:28:19 +02:00
|
|
|
* Grab the inodes before we drop the lock. if we found
|
|
|
|
* nothing, nr == 0 and the loop will be skipped.
|
2010-09-24 10:40:15 +02:00
|
|
|
*/
|
2010-09-28 04:28:19 +02:00
|
|
|
for (i = 0; i < nr_found; i++) {
|
|
|
|
struct xfs_inode *ip = batch[i];
|
|
|
|
|
|
|
|
if (done || xfs_inode_ag_walk_grab(ip))
|
|
|
|
batch[i] = NULL;
|
|
|
|
|
|
|
|
/*
|
2010-12-17 07:29:43 +01:00
|
|
|
* Update the index for the next lookup. Catch
|
|
|
|
* overflows into the next AG range which can occur if
|
|
|
|
* we have inodes in the last block of the AG and we
|
|
|
|
* are currently pointing to the last inode.
|
|
|
|
*
|
|
|
|
* Because we may see inodes that are from the wrong AG
|
|
|
|
* due to RCU freeing and reallocation, only update the
|
|
|
|
* index if it lies in this AG. It was a race that lead
|
|
|
|
* us to see this inode, so another lookup from the
|
|
|
|
* same index will not find it again.
|
2010-09-28 04:28:19 +02:00
|
|
|
*/
|
2010-12-17 07:29:43 +01:00
|
|
|
if (XFS_INO_TO_AGNO(mp, ip->i_ino) != pag->pag_agno)
|
|
|
|
continue;
|
2010-09-28 04:28:19 +02:00
|
|
|
first_index = XFS_INO_TO_AGINO(mp, ip->i_ino + 1);
|
|
|
|
if (first_index < XFS_INO_TO_AGINO(mp, ip->i_ino))
|
|
|
|
done = 1;
|
2010-09-28 04:28:06 +02:00
|
|
|
}
|
2010-09-28 04:28:19 +02:00
|
|
|
|
|
|
|
/* unlock now we've grabbed the inodes. */
|
2010-12-17 07:29:43 +01:00
|
|
|
rcu_read_unlock();
|
2010-09-28 04:28:06 +02:00
|
|
|
|
2010-09-28 04:28:19 +02:00
|
|
|
for (i = 0; i < nr_found; i++) {
|
|
|
|
if (!batch[i])
|
|
|
|
continue;
|
|
|
|
error = execute(batch[i], pag, flags);
|
|
|
|
IRELE(batch[i]);
|
|
|
|
if (error == EAGAIN) {
|
|
|
|
skipped++;
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
if (error && last_error != EFSCORRUPTED)
|
|
|
|
last_error = error;
|
2009-06-08 15:35:14 +02:00
|
|
|
}
|
2010-01-11 00:51:45 +01:00
|
|
|
|
|
|
|
/* bail out if the filesystem is corrupted. */
|
2009-06-08 15:35:14 +02:00
|
|
|
if (error == EFSCORRUPTED)
|
|
|
|
break;
|
|
|
|
|
2011-07-08 06:14:46 +02:00
|
|
|
cond_resched();
|
|
|
|
|
2010-09-28 04:28:19 +02:00
|
|
|
} while (nr_found && !done);
|
2009-06-08 15:35:14 +02:00
|
|
|
|
|
|
|
if (skipped) {
|
|
|
|
delay(1);
|
|
|
|
goto restart;
|
|
|
|
}
|
|
|
|
return last_error;
|
|
|
|
}
|
|
|
|
|
2009-06-08 15:35:27 +02:00
|
|
|
int
|
2009-06-08 15:35:14 +02:00
|
|
|
xfs_inode_ag_iterator(
|
|
|
|
struct xfs_mount *mp,
|
|
|
|
int (*execute)(struct xfs_inode *ip,
|
|
|
|
struct xfs_perag *pag, int flags),
|
2010-09-24 10:40:15 +02:00
|
|
|
int flags)
|
2009-06-08 15:35:14 +02:00
|
|
|
{
|
2010-07-20 01:43:39 +02:00
|
|
|
struct xfs_perag *pag;
|
2009-06-08 15:35:14 +02:00
|
|
|
int error = 0;
|
|
|
|
int last_error = 0;
|
|
|
|
xfs_agnumber_t ag;
|
|
|
|
|
2010-07-20 01:43:39 +02:00
|
|
|
ag = 0;
|
2010-09-24 10:40:15 +02:00
|
|
|
while ((pag = xfs_perag_get(mp, ag))) {
|
|
|
|
ag = pag->pag_agno + 1;
|
|
|
|
error = xfs_inode_ag_walk(mp, pag, execute, flags);
|
2010-01-11 12:47:40 +01:00
|
|
|
xfs_perag_put(pag);
|
2009-06-08 15:35:14 +02:00
|
|
|
if (error) {
|
|
|
|
last_error = error;
|
|
|
|
if (error == EFSCORRUPTED)
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return XFS_ERROR(last_error);
|
|
|
|
}
|
|
|
|
|
2009-11-19 16:52:00 +01:00
|
|
|
STATIC int
|
2008-10-30 07:14:53 +01:00
|
|
|
xfs_sync_fsdata(
|
2010-03-12 11:59:16 +01:00
|
|
|
struct xfs_mount *mp)
|
2008-10-30 07:14:53 +01:00
|
|
|
{
|
|
|
|
struct xfs_buf *bp;
|
2011-08-23 10:28:07 +02:00
|
|
|
int error;
|
2008-10-30 07:14:53 +01:00
|
|
|
|
|
|
|
/*
|
2010-03-12 11:59:16 +01:00
|
|
|
* If the buffer is pinned then push on the log so we won't get stuck
|
|
|
|
* waiting in the write for someone, maybe ourselves, to flush the log.
|
|
|
|
*
|
|
|
|
* Even though we just pushed the log above, we did not have the
|
|
|
|
* superblock buffer locked at that point so it can become pinned in
|
|
|
|
* between there and here.
|
2008-10-30 07:14:53 +01:00
|
|
|
*/
|
2010-03-12 11:59:16 +01:00
|
|
|
bp = xfs_getsb(mp, 0);
|
2011-07-23 01:40:27 +02:00
|
|
|
if (xfs_buf_ispinned(bp))
|
2010-03-12 11:59:16 +01:00
|
|
|
xfs_log_force(mp, 0);
|
2011-08-23 10:28:07 +02:00
|
|
|
error = xfs_bwrite(bp);
|
|
|
|
xfs_buf_relse(bp);
|
|
|
|
return error;
|
2008-10-30 07:15:50 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
2008-10-30 07:16:11 +01:00
|
|
|
* When remounting a filesystem read-only or freezing the filesystem, we have
|
|
|
|
* two phases to execute. This first phase is syncing the data before we
|
|
|
|
* quiesce the filesystem, and the second is flushing all the inodes out after
|
|
|
|
* we've waited for all the transactions created by the first phase to
|
|
|
|
* complete. The second phase ensures that the inodes are written to their
|
|
|
|
* location on disk rather than just existing in transactions in the log. This
|
|
|
|
* means after a quiesce there is no log replay required to write the inodes to
|
|
|
|
* disk (this is the main difference between a sync and a quiesce).
|
|
|
|
*/
|
|
|
|
/*
|
|
|
|
* First stage of freeze - no writers will make progress now we are here,
|
2008-10-30 07:15:50 +01:00
|
|
|
* so we flush delwri and delalloc buffers here, then wait for all I/O to
|
|
|
|
* complete. Data is frozen at that point. Metadata is not frozen,
|
2012-04-23 07:58:34 +02:00
|
|
|
* transactions can still occur here so don't bother emptying the AIL
|
2008-10-30 07:16:11 +01:00
|
|
|
* because it'll just get dirty again.
|
2008-10-30 07:15:50 +01:00
|
|
|
*/
|
|
|
|
int
|
|
|
|
xfs_quiesce_data(
|
|
|
|
struct xfs_mount *mp)
|
|
|
|
{
|
2010-03-12 11:59:16 +01:00
|
|
|
int error, error2 = 0;
|
2008-10-30 07:15:50 +01:00
|
|
|
|
2011-12-06 22:58:12 +01:00
|
|
|
/* force out the log */
|
2011-07-08 14:34:39 +02:00
|
|
|
xfs_log_force(mp, XFS_LOG_SYNC);
|
|
|
|
|
2008-10-30 07:16:11 +01:00
|
|
|
/* write superblock and hoover up shutdown errors */
|
2010-03-12 11:59:16 +01:00
|
|
|
error = xfs_sync_fsdata(mp);
|
|
|
|
|
|
|
|
/* mark the log as covered if needed */
|
|
|
|
if (xfs_log_need_covered(mp))
|
2011-01-04 05:49:29 +01:00
|
|
|
error2 = xfs_fs_log_dummy(mp);
|
2008-10-30 07:15:50 +01:00
|
|
|
|
2010-03-12 11:59:16 +01:00
|
|
|
return error ? error : error2;
|
2008-10-30 07:14:53 +01:00
|
|
|
}
|
|
|
|
|
2008-10-30 07:16:21 +01:00
|
|
|
/*
|
|
|
|
* Second stage of a quiesce. The data is already synced, now we have to take
|
|
|
|
* care of the metadata. New transactions are already blocked, so we need to
|
2011-03-31 03:57:33 +02:00
|
|
|
* wait for any remaining transactions to drain out before proceeding.
|
2012-10-08 12:56:01 +02:00
|
|
|
*
|
|
|
|
* Note: this stops background sync work - the callers must ensure it is started
|
|
|
|
* again when appropriate.
|
2008-10-30 07:16:21 +01:00
|
|
|
*/
|
|
|
|
void
|
|
|
|
xfs_quiesce_attr(
|
|
|
|
struct xfs_mount *mp)
|
|
|
|
{
|
|
|
|
int error = 0;
|
|
|
|
|
|
|
|
/* wait for all modifications to complete */
|
|
|
|
while (atomic_read(&mp->m_active_trans) > 0)
|
|
|
|
delay(100);
|
|
|
|
|
2012-04-23 07:58:34 +02:00
|
|
|
/* reclaim inodes to do any IO before the freeze completes */
|
|
|
|
xfs_reclaim_inodes(mp, 0);
|
|
|
|
xfs_reclaim_inodes(mp, SYNC_WAIT);
|
|
|
|
|
|
|
|
/* flush all pending changes from the AIL */
|
|
|
|
xfs_ail_push_all_sync(mp->m_ail);
|
2008-10-30 07:16:21 +01:00
|
|
|
|
2012-10-08 12:56:02 +02:00
|
|
|
/* stop background log work */
|
|
|
|
cancel_delayed_work_sync(&mp->m_log->l_work);
|
2012-10-08 12:56:01 +02:00
|
|
|
|
2009-01-23 04:34:05 +01:00
|
|
|
/*
|
|
|
|
* Just warn here till VFS can correctly support
|
|
|
|
* read-only remount without racing.
|
|
|
|
*/
|
|
|
|
WARN_ON(atomic_read(&mp->m_active_trans) != 0);
|
2008-10-30 07:16:21 +01:00
|
|
|
|
|
|
|
/* Push the superblock and write an unmount record */
|
2011-06-30 00:10:14 +02:00
|
|
|
error = xfs_log_sbcount(mp);
|
2008-10-30 07:16:21 +01:00
|
|
|
if (error)
|
2011-03-07 00:00:35 +01:00
|
|
|
xfs_warn(mp, "xfs_attr_quiesce: failed to log sb changes. "
|
2008-10-30 07:16:21 +01:00
|
|
|
"Frozen image may not be consistent.");
|
|
|
|
xfs_log_unmount_write(mp);
|
2012-04-23 07:58:34 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* At this point we might have modified the superblock again and thus
|
|
|
|
* added an item to the AIL, thus flush it again.
|
|
|
|
*/
|
|
|
|
xfs_ail_push_all_sync(mp->m_ail);
|
2012-07-24 17:59:19 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* The superblock buffer is uncached and xfsaild_push() will lock and
|
|
|
|
* set the XBF_ASYNC flag on the buffer. We cannot do xfs_buf_iowait()
|
|
|
|
* here but a lock on the superblock buffer will block until iodone()
|
|
|
|
* has completed.
|
|
|
|
*/
|
|
|
|
xfs_buf_lock(mp->m_sb_bp);
|
|
|
|
xfs_buf_unlock(mp->m_sb_bp);
|
2008-10-30 07:16:21 +01:00
|
|
|
}
|
|
|
|
|
2011-04-08 04:45:07 +02:00
|
|
|
/*
|
|
|
|
* Queue a new inode reclaim pass if there are reclaimable inodes and there
|
|
|
|
* isn't a reclaim pass already in progress. By default it runs every 5s based
|
|
|
|
* on the xfs syncd work default of 30s. Perhaps this should have it's own
|
|
|
|
* tunable, but that can be done if this method proves to be ineffective or too
|
|
|
|
* aggressive.
|
|
|
|
*/
|
|
|
|
static void
|
|
|
|
xfs_syncd_queue_reclaim(
|
|
|
|
struct xfs_mount *mp)
|
2008-10-30 07:06:18 +01:00
|
|
|
{
|
|
|
|
|
2011-04-08 04:45:07 +02:00
|
|
|
rcu_read_lock();
|
|
|
|
if (radix_tree_tagged(&mp->m_perag_tree, XFS_ICI_RECLAIM_TAG)) {
|
|
|
|
queue_delayed_work(xfs_syncd_wq, &mp->m_reclaim_work,
|
|
|
|
msecs_to_jiffies(xfs_syncd_centisecs / 6 * 10));
|
2008-10-30 07:06:18 +01:00
|
|
|
}
|
2011-04-08 04:45:07 +02:00
|
|
|
rcu_read_unlock();
|
|
|
|
}
|
2008-10-30 07:06:18 +01:00
|
|
|
|
2011-04-08 04:45:07 +02:00
|
|
|
/*
|
|
|
|
* This is a fast pass over the inode cache to try to get reclaim moving on as
|
|
|
|
* many inodes as possible in a short period of time. It kicks itself every few
|
|
|
|
* seconds, as well as being kicked by the inode cache shrinker when memory
|
|
|
|
* goes low. It scans as quickly as possible avoiding locked inodes or those
|
|
|
|
* already being flushed, and once done schedules a future pass.
|
|
|
|
*/
|
2012-10-08 12:55:59 +02:00
|
|
|
void
|
2011-04-08 04:45:07 +02:00
|
|
|
xfs_reclaim_worker(
|
|
|
|
struct work_struct *work)
|
|
|
|
{
|
|
|
|
struct xfs_mount *mp = container_of(to_delayed_work(work),
|
|
|
|
struct xfs_mount, m_reclaim_work);
|
|
|
|
|
|
|
|
xfs_reclaim_inodes(mp, SYNC_TRYLOCK);
|
|
|
|
xfs_syncd_queue_reclaim(mp);
|
|
|
|
}
|
|
|
|
|
2009-08-17 02:36:34 +02:00
|
|
|
void
|
|
|
|
__xfs_inode_set_reclaim_tag(
|
|
|
|
struct xfs_perag *pag,
|
|
|
|
struct xfs_inode *ip)
|
|
|
|
{
|
|
|
|
radix_tree_tag_set(&pag->pag_ici_root,
|
|
|
|
XFS_INO_TO_AGINO(ip->i_mount, ip->i_ino),
|
|
|
|
XFS_ICI_RECLAIM_TAG);
|
2010-07-20 01:43:39 +02:00
|
|
|
|
|
|
|
if (!pag->pag_ici_reclaimable) {
|
|
|
|
/* propagate the reclaim tag up into the perag radix tree */
|
|
|
|
spin_lock(&ip->i_mount->m_perag_lock);
|
|
|
|
radix_tree_tag_set(&ip->i_mount->m_perag_tree,
|
|
|
|
XFS_INO_TO_AGNO(ip->i_mount, ip->i_ino),
|
|
|
|
XFS_ICI_RECLAIM_TAG);
|
|
|
|
spin_unlock(&ip->i_mount->m_perag_lock);
|
2011-04-08 04:45:07 +02:00
|
|
|
|
|
|
|
/* schedule periodic background inode reclaim */
|
|
|
|
xfs_syncd_queue_reclaim(ip->i_mount);
|
|
|
|
|
2010-07-20 01:43:39 +02:00
|
|
|
trace_xfs_perag_set_reclaim(ip->i_mount, pag->pag_agno,
|
|
|
|
-1, _RET_IP_);
|
|
|
|
}
|
2010-04-29 01:55:50 +02:00
|
|
|
pag->pag_ici_reclaimable++;
|
2009-08-17 02:36:34 +02:00
|
|
|
}
|
|
|
|
|
2008-10-30 07:37:49 +01:00
|
|
|
/*
|
|
|
|
* We set the inode flag atomically with the radix tree tag.
|
|
|
|
* Once we get tag lookups on the radix tree, this inode flag
|
|
|
|
* can go away.
|
|
|
|
*/
|
2008-10-30 07:37:26 +01:00
|
|
|
void
|
|
|
|
xfs_inode_set_reclaim_tag(
|
|
|
|
xfs_inode_t *ip)
|
|
|
|
{
|
2010-01-11 12:47:40 +01:00
|
|
|
struct xfs_mount *mp = ip->i_mount;
|
|
|
|
struct xfs_perag *pag;
|
2008-10-30 07:37:26 +01:00
|
|
|
|
2010-01-11 12:47:40 +01:00
|
|
|
pag = xfs_perag_get(mp, XFS_INO_TO_AGNO(mp, ip->i_ino));
|
2010-12-16 07:08:41 +01:00
|
|
|
spin_lock(&pag->pag_ici_lock);
|
2008-10-30 07:37:26 +01:00
|
|
|
spin_lock(&ip->i_flags_lock);
|
2009-08-17 02:36:34 +02:00
|
|
|
__xfs_inode_set_reclaim_tag(pag, ip);
|
2008-10-30 07:37:49 +01:00
|
|
|
__xfs_iflags_set(ip, XFS_IRECLAIMABLE);
|
2008-10-30 07:37:26 +01:00
|
|
|
spin_unlock(&ip->i_flags_lock);
|
2010-12-16 07:08:41 +01:00
|
|
|
spin_unlock(&pag->pag_ici_lock);
|
2010-01-11 12:47:40 +01:00
|
|
|
xfs_perag_put(pag);
|
2008-10-30 07:37:26 +01:00
|
|
|
}
|
|
|
|
|
xfs: properly account for reclaimed inodes
When marking an inode reclaimable, a per-AG counter is increased, the
inode is tagged reclaimable in its per-AG tree, and, when this is the
first reclaimable inode in the AG, the AG entry in the per-mount tree
is also tagged.
When an inode is finally reclaimed, however, it is only deleted from
the per-AG tree. Neither the counter is decreased, nor is the parent
tree's AG entry untagged properly.
Since the tags in the per-mount tree are not cleared, the inode
shrinker iterates over all AGs that have had reclaimable inodes at one
point in time.
The counters on the other hand signal an increasing amount of slab
objects to reclaim. Since "70e60ce xfs: convert inode shrinker to
per-filesystem context" this is not a real issue anymore because the
shrinker bails out after one iteration.
But the problem was observable on a machine running v2.6.34, where the
reclaimable work increased and each process going into direct reclaim
eventually got stuck on the xfs inode shrinking path, trying to scan
several million objects.
Fix this by properly unwinding the reclaimable-state tracking of an
inode when it is reclaimed.
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: stable@kernel.org
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Alex Elder <aelder@sgi.com>
2010-10-01 09:43:54 +02:00
|
|
|
STATIC void
|
|
|
|
__xfs_inode_clear_reclaim(
|
2008-10-30 07:37:26 +01:00
|
|
|
xfs_perag_t *pag,
|
|
|
|
xfs_inode_t *ip)
|
|
|
|
{
|
2010-04-29 01:55:50 +02:00
|
|
|
pag->pag_ici_reclaimable--;
|
2010-07-20 01:43:39 +02:00
|
|
|
if (!pag->pag_ici_reclaimable) {
|
|
|
|
/* clear the reclaim tag from the perag radix tree */
|
|
|
|
spin_lock(&ip->i_mount->m_perag_lock);
|
|
|
|
radix_tree_tag_clear(&ip->i_mount->m_perag_tree,
|
|
|
|
XFS_INO_TO_AGNO(ip->i_mount, ip->i_ino),
|
|
|
|
XFS_ICI_RECLAIM_TAG);
|
|
|
|
spin_unlock(&ip->i_mount->m_perag_lock);
|
|
|
|
trace_xfs_perag_clear_reclaim(ip->i_mount, pag->pag_agno,
|
|
|
|
-1, _RET_IP_);
|
|
|
|
}
|
2008-10-30 07:37:26 +01:00
|
|
|
}
|
|
|
|
|
xfs: properly account for reclaimed inodes
When marking an inode reclaimable, a per-AG counter is increased, the
inode is tagged reclaimable in its per-AG tree, and, when this is the
first reclaimable inode in the AG, the AG entry in the per-mount tree
is also tagged.
When an inode is finally reclaimed, however, it is only deleted from
the per-AG tree. Neither the counter is decreased, nor is the parent
tree's AG entry untagged properly.
Since the tags in the per-mount tree are not cleared, the inode
shrinker iterates over all AGs that have had reclaimable inodes at one
point in time.
The counters on the other hand signal an increasing amount of slab
objects to reclaim. Since "70e60ce xfs: convert inode shrinker to
per-filesystem context" this is not a real issue anymore because the
shrinker bails out after one iteration.
But the problem was observable on a machine running v2.6.34, where the
reclaimable work increased and each process going into direct reclaim
eventually got stuck on the xfs inode shrinking path, trying to scan
several million objects.
Fix this by properly unwinding the reclaimable-state tracking of an
inode when it is reclaimed.
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: stable@kernel.org
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Alex Elder <aelder@sgi.com>
2010-10-01 09:43:54 +02:00
|
|
|
void
|
|
|
|
__xfs_inode_clear_reclaim_tag(
|
|
|
|
xfs_mount_t *mp,
|
|
|
|
xfs_perag_t *pag,
|
|
|
|
xfs_inode_t *ip)
|
|
|
|
{
|
|
|
|
radix_tree_tag_clear(&pag->pag_ici_root,
|
|
|
|
XFS_INO_TO_AGINO(mp, ip->i_ino), XFS_ICI_RECLAIM_TAG);
|
|
|
|
__xfs_inode_clear_reclaim(pag, ip);
|
|
|
|
}
|
|
|
|
|
2010-09-24 11:51:50 +02:00
|
|
|
/*
|
|
|
|
* Grab the inode for reclaim exclusively.
|
|
|
|
* Return 0 if we grabbed it, non-zero otherwise.
|
|
|
|
*/
|
|
|
|
STATIC int
|
|
|
|
xfs_reclaim_inode_grab(
|
|
|
|
struct xfs_inode *ip,
|
|
|
|
int flags)
|
|
|
|
{
|
2010-12-17 07:29:43 +01:00
|
|
|
ASSERT(rcu_read_lock_held());
|
|
|
|
|
|
|
|
/* quick check for stale RCU freed inode */
|
|
|
|
if (!ip->i_ino)
|
|
|
|
return 1;
|
2010-09-24 11:51:50 +02:00
|
|
|
|
|
|
|
/*
|
2011-12-18 21:00:09 +01:00
|
|
|
* If we are asked for non-blocking operation, do unlocked checks to
|
|
|
|
* see if the inode already is being flushed or in reclaim to avoid
|
|
|
|
* lock traffic.
|
2010-09-24 11:51:50 +02:00
|
|
|
*/
|
|
|
|
if ((flags & SYNC_TRYLOCK) &&
|
2011-12-18 21:00:09 +01:00
|
|
|
__xfs_iflags_test(ip, XFS_IFLOCK | XFS_IRECLAIM))
|
2010-09-24 11:51:50 +02:00
|
|
|
return 1;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* The radix tree lock here protects a thread in xfs_iget from racing
|
|
|
|
* with us starting reclaim on the inode. Once we have the
|
|
|
|
* XFS_IRECLAIM flag set it will not touch us.
|
2010-12-17 07:29:43 +01:00
|
|
|
*
|
|
|
|
* Due to RCU lookup, we may find inodes that have been freed and only
|
|
|
|
* have XFS_IRECLAIM set. Indeed, we may see reallocated inodes that
|
|
|
|
* aren't candidates for reclaim at all, so we must check the
|
|
|
|
* XFS_IRECLAIMABLE is set first before proceeding to reclaim.
|
2010-09-24 11:51:50 +02:00
|
|
|
*/
|
|
|
|
spin_lock(&ip->i_flags_lock);
|
2010-12-17 07:29:43 +01:00
|
|
|
if (!__xfs_iflags_test(ip, XFS_IRECLAIMABLE) ||
|
|
|
|
__xfs_iflags_test(ip, XFS_IRECLAIM)) {
|
|
|
|
/* not a reclaim candidate. */
|
2010-09-24 11:51:50 +02:00
|
|
|
spin_unlock(&ip->i_flags_lock);
|
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
__xfs_iflags_set(ip, XFS_IRECLAIM);
|
|
|
|
spin_unlock(&ip->i_flags_lock);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2010-02-06 02:37:26 +01:00
|
|
|
/*
|
2012-04-23 07:58:35 +02:00
|
|
|
* Inodes in different states need to be treated differently. The following
|
|
|
|
* table lists the inode states and the reclaim actions necessary:
|
2010-02-06 02:37:26 +01:00
|
|
|
*
|
|
|
|
* inode state iflush ret required action
|
|
|
|
* --------------- ---------- ---------------
|
|
|
|
* bad - reclaim
|
|
|
|
* shutdown EIO unpin and reclaim
|
|
|
|
* clean, unpinned 0 reclaim
|
|
|
|
* stale, unpinned 0 reclaim
|
2010-02-06 02:39:36 +01:00
|
|
|
* clean, pinned(*) 0 requeue
|
|
|
|
* stale, pinned EAGAIN requeue
|
2012-04-23 07:58:35 +02:00
|
|
|
* dirty, async - requeue
|
|
|
|
* dirty, sync 0 reclaim
|
2010-02-06 02:37:26 +01:00
|
|
|
*
|
|
|
|
* (*) dgc: I don't think the clean, pinned state is possible but it gets
|
|
|
|
* handled anyway given the order of checks implemented.
|
|
|
|
*
|
2010-02-06 02:39:36 +01:00
|
|
|
* Also, because we get the flush lock first, we know that any inode that has
|
|
|
|
* been flushed delwri has had the flush completed by the time we check that
|
2012-04-23 07:58:35 +02:00
|
|
|
* the inode is clean.
|
2010-02-06 02:39:36 +01:00
|
|
|
*
|
2012-04-23 07:58:35 +02:00
|
|
|
* Note that because the inode is flushed delayed write by AIL pushing, the
|
|
|
|
* flush lock may already be held here and waiting on it can result in very
|
|
|
|
* long latencies. Hence for sync reclaims, where we wait on the flush lock,
|
|
|
|
* the caller should push the AIL first before trying to reclaim inodes to
|
|
|
|
* minimise the amount of time spent waiting. For background relaim, we only
|
|
|
|
* bother to reclaim clean inodes anyway.
|
2010-02-06 02:39:36 +01:00
|
|
|
*
|
2010-02-06 02:37:26 +01:00
|
|
|
* Hence the order of actions after gaining the locks should be:
|
|
|
|
* bad => reclaim
|
|
|
|
* shutdown => unpin and reclaim
|
2012-04-23 07:58:35 +02:00
|
|
|
* pinned, async => requeue
|
2010-02-06 02:39:36 +01:00
|
|
|
* pinned, sync => unpin
|
2010-02-06 02:37:26 +01:00
|
|
|
* stale => reclaim
|
|
|
|
* clean => reclaim
|
2012-04-23 07:58:35 +02:00
|
|
|
* dirty, async => requeue
|
2010-02-06 02:39:36 +01:00
|
|
|
* dirty, sync => flush, wait and reclaim
|
2010-02-06 02:37:26 +01:00
|
|
|
*/
|
2009-06-08 15:35:14 +02:00
|
|
|
STATIC int
|
2010-01-11 00:51:45 +01:00
|
|
|
xfs_reclaim_inode(
|
2009-06-08 15:35:14 +02:00
|
|
|
struct xfs_inode *ip,
|
|
|
|
struct xfs_perag *pag,
|
2010-01-11 00:51:45 +01:00
|
|
|
int sync_mode)
|
2008-10-30 07:37:03 +01:00
|
|
|
{
|
2012-04-23 07:58:36 +02:00
|
|
|
struct xfs_buf *bp = NULL;
|
|
|
|
int error;
|
2010-02-06 02:37:26 +01:00
|
|
|
|
2011-03-25 23:13:55 +01:00
|
|
|
restart:
|
|
|
|
error = 0;
|
2010-01-11 00:51:45 +01:00
|
|
|
xfs_ilock(ip, XFS_ILOCK_EXCL);
|
2010-02-06 02:39:36 +01:00
|
|
|
if (!xfs_iflock_nowait(ip)) {
|
|
|
|
if (!(sync_mode & SYNC_WAIT))
|
|
|
|
goto out;
|
|
|
|
xfs_iflock(ip);
|
|
|
|
}
|
2008-10-30 07:37:37 +01:00
|
|
|
|
2010-02-06 02:37:26 +01:00
|
|
|
if (is_bad_inode(VFS_I(ip)))
|
|
|
|
goto reclaim;
|
|
|
|
if (XFS_FORCED_SHUTDOWN(ip->i_mount)) {
|
|
|
|
xfs_iunpin_wait(ip);
|
2012-04-23 07:58:41 +02:00
|
|
|
xfs_iflush_abort(ip, false);
|
2010-02-06 02:37:26 +01:00
|
|
|
goto reclaim;
|
|
|
|
}
|
2010-02-06 02:39:36 +01:00
|
|
|
if (xfs_ipincount(ip)) {
|
2012-04-23 07:58:35 +02:00
|
|
|
if (!(sync_mode & SYNC_WAIT))
|
|
|
|
goto out_ifunlock;
|
2010-02-06 02:37:26 +01:00
|
|
|
xfs_iunpin_wait(ip);
|
2010-02-06 02:39:36 +01:00
|
|
|
}
|
2010-02-06 02:37:26 +01:00
|
|
|
if (xfs_iflags_test(ip, XFS_ISTALE))
|
|
|
|
goto reclaim;
|
|
|
|
if (xfs_inode_clean(ip))
|
|
|
|
goto reclaim;
|
|
|
|
|
2012-04-23 07:58:35 +02:00
|
|
|
/*
|
|
|
|
* Never flush out dirty data during non-blocking reclaim, as it would
|
|
|
|
* just contend with AIL pushing trying to do the same job.
|
|
|
|
*/
|
|
|
|
if (!(sync_mode & SYNC_WAIT))
|
|
|
|
goto out_ifunlock;
|
|
|
|
|
2011-03-25 23:13:55 +01:00
|
|
|
/*
|
|
|
|
* Now we have an inode that needs flushing.
|
|
|
|
*
|
2012-04-23 07:58:36 +02:00
|
|
|
* Note that xfs_iflush will never block on the inode buffer lock, as
|
2011-03-25 23:13:55 +01:00
|
|
|
* xfs_ifree_cluster() can lock the inode buffer before it locks the
|
2012-04-23 07:58:36 +02:00
|
|
|
* ip->i_lock, and we are doing the exact opposite here. As a result,
|
2012-07-03 18:21:22 +02:00
|
|
|
* doing a blocking xfs_imap_to_bp() to get the cluster buffer would
|
|
|
|
* result in an ABBA deadlock with xfs_ifree_cluster().
|
2011-03-25 23:13:55 +01:00
|
|
|
*
|
|
|
|
* As xfs_ifree_cluser() must gather all inodes that are active in the
|
|
|
|
* cache to mark them stale, if we hit this case we don't actually want
|
|
|
|
* to do IO here - we want the inode marked stale so we can simply
|
2012-04-23 07:58:36 +02:00
|
|
|
* reclaim it. Hence if we get an EAGAIN error here, just unlock the
|
|
|
|
* inode, back off and try again. Hopefully the next pass through will
|
|
|
|
* see the stale flag set on the inode.
|
2011-03-25 23:13:55 +01:00
|
|
|
*/
|
2012-04-23 07:58:36 +02:00
|
|
|
error = xfs_iflush(ip, &bp);
|
2012-04-23 07:58:35 +02:00
|
|
|
if (error == EAGAIN) {
|
|
|
|
xfs_iunlock(ip, XFS_ILOCK_EXCL);
|
|
|
|
/* backoff longer than in xfs_ifree_cluster */
|
|
|
|
delay(2);
|
|
|
|
goto restart;
|
2010-02-06 02:39:36 +01:00
|
|
|
}
|
|
|
|
|
2012-04-23 07:58:36 +02:00
|
|
|
if (!error) {
|
|
|
|
error = xfs_bwrite(bp);
|
|
|
|
xfs_buf_relse(bp);
|
|
|
|
}
|
|
|
|
|
|
|
|
xfs_iflock(ip);
|
2010-02-06 02:37:26 +01:00
|
|
|
reclaim:
|
|
|
|
xfs_ifunlock(ip);
|
2010-01-11 00:51:45 +01:00
|
|
|
xfs_iunlock(ip, XFS_ILOCK_EXCL);
|
2010-07-20 09:53:25 +02:00
|
|
|
|
|
|
|
XFS_STATS_INC(xs_ig_reclaims);
|
|
|
|
/*
|
|
|
|
* Remove the inode from the per-AG radix tree.
|
|
|
|
*
|
|
|
|
* Because radix_tree_delete won't complain even if the item was never
|
|
|
|
* added to the tree assert that it's been there before to catch
|
|
|
|
* problems with the inode life time early on.
|
|
|
|
*/
|
2010-12-16 07:08:41 +01:00
|
|
|
spin_lock(&pag->pag_ici_lock);
|
2010-07-20 09:53:25 +02:00
|
|
|
if (!radix_tree_delete(&pag->pag_ici_root,
|
|
|
|
XFS_INO_TO_AGINO(ip->i_mount, ip->i_ino)))
|
|
|
|
ASSERT(0);
|
xfs: properly account for reclaimed inodes
When marking an inode reclaimable, a per-AG counter is increased, the
inode is tagged reclaimable in its per-AG tree, and, when this is the
first reclaimable inode in the AG, the AG entry in the per-mount tree
is also tagged.
When an inode is finally reclaimed, however, it is only deleted from
the per-AG tree. Neither the counter is decreased, nor is the parent
tree's AG entry untagged properly.
Since the tags in the per-mount tree are not cleared, the inode
shrinker iterates over all AGs that have had reclaimable inodes at one
point in time.
The counters on the other hand signal an increasing amount of slab
objects to reclaim. Since "70e60ce xfs: convert inode shrinker to
per-filesystem context" this is not a real issue anymore because the
shrinker bails out after one iteration.
But the problem was observable on a machine running v2.6.34, where the
reclaimable work increased and each process going into direct reclaim
eventually got stuck on the xfs inode shrinking path, trying to scan
several million objects.
Fix this by properly unwinding the reclaimable-state tracking of an
inode when it is reclaimed.
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: stable@kernel.org
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Alex Elder <aelder@sgi.com>
2010-10-01 09:43:54 +02:00
|
|
|
__xfs_inode_clear_reclaim(pag, ip);
|
2010-12-16 07:08:41 +01:00
|
|
|
spin_unlock(&pag->pag_ici_lock);
|
2010-07-20 09:53:25 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Here we do an (almost) spurious inode lock in order to coordinate
|
|
|
|
* with inode cache radix tree lookups. This is because the lookup
|
|
|
|
* can reference the inodes in the cache without taking references.
|
|
|
|
*
|
|
|
|
* We make that OK here by ensuring that we wait until the inode is
|
2012-02-16 23:01:00 +01:00
|
|
|
* unlocked after the lookup before we go ahead and free it.
|
2010-07-20 09:53:25 +02:00
|
|
|
*/
|
2012-02-16 23:01:00 +01:00
|
|
|
xfs_ilock(ip, XFS_ILOCK_EXCL);
|
2010-07-20 09:53:25 +02:00
|
|
|
xfs_qm_dqdetach(ip);
|
2012-02-16 23:01:00 +01:00
|
|
|
xfs_iunlock(ip, XFS_ILOCK_EXCL);
|
2010-07-20 09:53:25 +02:00
|
|
|
|
|
|
|
xfs_inode_free(ip);
|
2012-02-16 23:01:00 +01:00
|
|
|
return error;
|
2012-04-23 07:58:35 +02:00
|
|
|
|
|
|
|
out_ifunlock:
|
|
|
|
xfs_ifunlock(ip);
|
|
|
|
out:
|
|
|
|
xfs_iflags_clear(ip, XFS_IRECLAIM);
|
|
|
|
xfs_iunlock(ip, XFS_ILOCK_EXCL);
|
|
|
|
/*
|
|
|
|
* We could return EAGAIN here to make reclaim rescan the inode tree in
|
|
|
|
* a short while. However, this just burns CPU time scanning the tree
|
|
|
|
* waiting for IO to complete and xfssyncd never goes back to the idle
|
|
|
|
* state. Instead, return 0 to let the next scheduled background reclaim
|
|
|
|
* attempt to reclaim the inode again.
|
|
|
|
*/
|
|
|
|
return 0;
|
2008-10-30 07:37:37 +01:00
|
|
|
}
|
|
|
|
|
2010-09-24 10:40:15 +02:00
|
|
|
/*
|
|
|
|
* Walk the AGs and reclaim the inodes in them. Even if the filesystem is
|
|
|
|
* corrupted, we still want to try to reclaim all the inodes. If we don't,
|
|
|
|
* then a shut down during filesystem unmount reclaim walk leak all the
|
|
|
|
* unreclaimed inodes.
|
|
|
|
*/
|
|
|
|
int
|
|
|
|
xfs_reclaim_inodes_ag(
|
|
|
|
struct xfs_mount *mp,
|
|
|
|
int flags,
|
|
|
|
int *nr_to_scan)
|
|
|
|
{
|
|
|
|
struct xfs_perag *pag;
|
|
|
|
int error = 0;
|
|
|
|
int last_error = 0;
|
|
|
|
xfs_agnumber_t ag;
|
2010-09-27 03:09:51 +02:00
|
|
|
int trylock = flags & SYNC_TRYLOCK;
|
|
|
|
int skipped;
|
2010-09-24 10:40:15 +02:00
|
|
|
|
2010-09-27 03:09:51 +02:00
|
|
|
restart:
|
2010-09-24 10:40:15 +02:00
|
|
|
ag = 0;
|
2010-09-27 03:09:51 +02:00
|
|
|
skipped = 0;
|
2010-09-24 10:40:15 +02:00
|
|
|
while ((pag = xfs_perag_get_tag(mp, ag, XFS_ICI_RECLAIM_TAG))) {
|
|
|
|
unsigned long first_index = 0;
|
|
|
|
int done = 0;
|
2010-09-24 11:51:50 +02:00
|
|
|
int nr_found = 0;
|
2010-09-24 10:40:15 +02:00
|
|
|
|
|
|
|
ag = pag->pag_agno + 1;
|
|
|
|
|
2010-09-27 03:09:51 +02:00
|
|
|
if (trylock) {
|
|
|
|
if (!mutex_trylock(&pag->pag_ici_reclaim_lock)) {
|
|
|
|
skipped++;
|
2010-11-08 09:55:04 +01:00
|
|
|
xfs_perag_put(pag);
|
2010-09-27 03:09:51 +02:00
|
|
|
continue;
|
|
|
|
}
|
|
|
|
first_index = pag->pag_ici_reclaim_cursor;
|
|
|
|
} else
|
|
|
|
mutex_lock(&pag->pag_ici_reclaim_lock);
|
|
|
|
|
2010-09-24 10:40:15 +02:00
|
|
|
do {
|
2010-09-24 11:51:50 +02:00
|
|
|
struct xfs_inode *batch[XFS_LOOKUP_BATCH];
|
|
|
|
int i;
|
2010-09-24 10:40:15 +02:00
|
|
|
|
2010-12-17 07:29:43 +01:00
|
|
|
rcu_read_lock();
|
2010-09-24 11:51:50 +02:00
|
|
|
nr_found = radix_tree_gang_lookup_tag(
|
|
|
|
&pag->pag_ici_root,
|
|
|
|
(void **)batch, first_index,
|
|
|
|
XFS_LOOKUP_BATCH,
|
2010-09-24 10:40:15 +02:00
|
|
|
XFS_ICI_RECLAIM_TAG);
|
|
|
|
if (!nr_found) {
|
2011-05-06 04:54:04 +02:00
|
|
|
done = 1;
|
2010-12-17 07:29:43 +01:00
|
|
|
rcu_read_unlock();
|
2010-09-24 10:40:15 +02:00
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
2010-09-24 11:51:50 +02:00
|
|
|
* Grab the inodes before we drop the lock. if we found
|
|
|
|
* nothing, nr == 0 and the loop will be skipped.
|
2010-09-24 10:40:15 +02:00
|
|
|
*/
|
2010-09-24 11:51:50 +02:00
|
|
|
for (i = 0; i < nr_found; i++) {
|
|
|
|
struct xfs_inode *ip = batch[i];
|
|
|
|
|
|
|
|
if (done || xfs_reclaim_inode_grab(ip, flags))
|
|
|
|
batch[i] = NULL;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Update the index for the next lookup. Catch
|
|
|
|
* overflows into the next AG range which can
|
|
|
|
* occur if we have inodes in the last block of
|
|
|
|
* the AG and we are currently pointing to the
|
|
|
|
* last inode.
|
2010-12-17 07:29:43 +01:00
|
|
|
*
|
|
|
|
* Because we may see inodes that are from the
|
|
|
|
* wrong AG due to RCU freeing and
|
|
|
|
* reallocation, only update the index if it
|
|
|
|
* lies in this AG. It was a race that lead us
|
|
|
|
* to see this inode, so another lookup from
|
|
|
|
* the same index will not find it again.
|
2010-09-24 11:51:50 +02:00
|
|
|
*/
|
2010-12-17 07:29:43 +01:00
|
|
|
if (XFS_INO_TO_AGNO(mp, ip->i_ino) !=
|
|
|
|
pag->pag_agno)
|
|
|
|
continue;
|
2010-09-24 11:51:50 +02:00
|
|
|
first_index = XFS_INO_TO_AGINO(mp, ip->i_ino + 1);
|
|
|
|
if (first_index < XFS_INO_TO_AGINO(mp, ip->i_ino))
|
|
|
|
done = 1;
|
|
|
|
}
|
2010-09-24 10:40:15 +02:00
|
|
|
|
2010-09-24 11:51:50 +02:00
|
|
|
/* unlock now we've grabbed the inodes. */
|
2010-12-17 07:29:43 +01:00
|
|
|
rcu_read_unlock();
|
2010-09-24 11:51:50 +02:00
|
|
|
|
|
|
|
for (i = 0; i < nr_found; i++) {
|
|
|
|
if (!batch[i])
|
|
|
|
continue;
|
|
|
|
error = xfs_reclaim_inode(batch[i], pag, flags);
|
|
|
|
if (error && last_error != EFSCORRUPTED)
|
|
|
|
last_error = error;
|
|
|
|
}
|
|
|
|
|
|
|
|
*nr_to_scan -= XFS_LOOKUP_BATCH;
|
2010-09-24 10:40:15 +02:00
|
|
|
|
2011-07-08 06:14:46 +02:00
|
|
|
cond_resched();
|
|
|
|
|
2010-09-24 11:51:50 +02:00
|
|
|
} while (nr_found && !done && *nr_to_scan > 0);
|
2010-09-24 10:40:15 +02:00
|
|
|
|
2010-09-27 03:09:51 +02:00
|
|
|
if (trylock && !done)
|
|
|
|
pag->pag_ici_reclaim_cursor = first_index;
|
|
|
|
else
|
|
|
|
pag->pag_ici_reclaim_cursor = 0;
|
|
|
|
mutex_unlock(&pag->pag_ici_reclaim_lock);
|
2010-09-24 10:40:15 +02:00
|
|
|
xfs_perag_put(pag);
|
|
|
|
}
|
2010-09-27 03:09:51 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* if we skipped any AG, and we still have scan count remaining, do
|
|
|
|
* another pass this time using blocking reclaim semantics (i.e
|
|
|
|
* waiting on the reclaim locks and ignoring the reclaim cursors). This
|
|
|
|
* ensure that when we get more reclaimers than AGs we block rather
|
|
|
|
* than spin trying to execute reclaim.
|
|
|
|
*/
|
2011-07-08 06:14:46 +02:00
|
|
|
if (skipped && (flags & SYNC_WAIT) && *nr_to_scan > 0) {
|
2010-09-27 03:09:51 +02:00
|
|
|
trylock = 0;
|
|
|
|
goto restart;
|
|
|
|
}
|
2010-09-24 10:40:15 +02:00
|
|
|
return XFS_ERROR(last_error);
|
|
|
|
}
|
|
|
|
|
2008-10-30 07:37:37 +01:00
|
|
|
int
|
|
|
|
xfs_reclaim_inodes(
|
|
|
|
xfs_mount_t *mp,
|
|
|
|
int mode)
|
|
|
|
{
|
2010-09-24 10:40:15 +02:00
|
|
|
int nr_to_scan = INT_MAX;
|
|
|
|
|
|
|
|
return xfs_reclaim_inodes_ag(mp, mode, &nr_to_scan);
|
2010-04-29 01:55:50 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
2011-07-08 06:14:46 +02:00
|
|
|
* Scan a certain number of inodes for reclaim.
|
2011-04-08 04:45:07 +02:00
|
|
|
*
|
|
|
|
* When called we make sure that there is a background (fast) inode reclaim in
|
2011-07-08 06:14:46 +02:00
|
|
|
* progress, while we will throttle the speed of reclaim via doing synchronous
|
2011-04-08 04:45:07 +02:00
|
|
|
* reclaim of inodes. That means if we come across dirty inodes, we wait for
|
|
|
|
* them to be cleaned, which we hope will not be very long due to the
|
|
|
|
* background walker having already kicked the IO off on those dirty inodes.
|
2010-04-29 01:55:50 +02:00
|
|
|
*/
|
2011-07-08 06:14:46 +02:00
|
|
|
void
|
|
|
|
xfs_reclaim_inodes_nr(
|
|
|
|
struct xfs_mount *mp,
|
|
|
|
int nr_to_scan)
|
2010-04-29 01:55:50 +02:00
|
|
|
{
|
2011-07-08 06:14:46 +02:00
|
|
|
/* kick background reclaimer and push the AIL */
|
|
|
|
xfs_syncd_queue_reclaim(mp);
|
|
|
|
xfs_ail_push_all(mp->m_ail);
|
2011-04-08 04:45:07 +02:00
|
|
|
|
2011-07-08 06:14:46 +02:00
|
|
|
xfs_reclaim_inodes_ag(mp, SYNC_TRYLOCK | SYNC_WAIT, &nr_to_scan);
|
|
|
|
}
|
2010-04-29 01:55:50 +02:00
|
|
|
|
2011-07-08 06:14:46 +02:00
|
|
|
/*
|
|
|
|
* Return the number of reclaimable inodes in the filesystem for
|
|
|
|
* the shrinker to determine how much to reclaim.
|
|
|
|
*/
|
|
|
|
int
|
|
|
|
xfs_reclaim_inodes_count(
|
|
|
|
struct xfs_mount *mp)
|
|
|
|
{
|
|
|
|
struct xfs_perag *pag;
|
|
|
|
xfs_agnumber_t ag = 0;
|
|
|
|
int reclaimable = 0;
|
2010-04-29 01:55:50 +02:00
|
|
|
|
2010-09-24 10:40:15 +02:00
|
|
|
while ((pag = xfs_perag_get_tag(mp, ag, XFS_ICI_RECLAIM_TAG))) {
|
|
|
|
ag = pag->pag_agno + 1;
|
2010-07-20 00:07:02 +02:00
|
|
|
reclaimable += pag->pag_ici_reclaimable;
|
|
|
|
xfs_perag_put(pag);
|
2010-04-29 01:55:50 +02:00
|
|
|
}
|
|
|
|
return reclaimable;
|
|
|
|
}
|
|
|
|
|