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_bit.h"
|
|
|
|
#include "xfs_log.h"
|
|
|
|
#include "xfs_inum.h"
|
|
|
|
#include "xfs_trans.h"
|
|
|
|
#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>
|
|
|
|
|
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;
|
|
|
|
|
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);
|
|
|
|
}
|
|
|
|
|
2009-06-08 15:35:03 +02:00
|
|
|
STATIC int
|
|
|
|
xfs_sync_inode_data(
|
|
|
|
struct xfs_inode *ip,
|
2009-06-08 15:35:14 +02:00
|
|
|
struct xfs_perag *pag,
|
2009-06-08 15:35:03 +02:00
|
|
|
int flags)
|
|
|
|
{
|
|
|
|
struct inode *inode = VFS_I(ip);
|
|
|
|
struct address_space *mapping = inode->i_mapping;
|
|
|
|
int error = 0;
|
|
|
|
|
|
|
|
if (!mapping_tagged(mapping, PAGECACHE_TAG_DIRTY))
|
|
|
|
goto out_wait;
|
|
|
|
|
|
|
|
if (!xfs_ilock_nowait(ip, XFS_IOLOCK_SHARED)) {
|
|
|
|
if (flags & SYNC_TRYLOCK)
|
|
|
|
goto out_wait;
|
|
|
|
xfs_ilock(ip, XFS_IOLOCK_SHARED);
|
|
|
|
}
|
|
|
|
|
|
|
|
error = xfs_flush_pages(ip, 0, -1, (flags & SYNC_WAIT) ?
|
2010-01-19 10:56:44 +01:00
|
|
|
0 : XBF_ASYNC, FI_NONE);
|
2009-06-08 15:35:03 +02:00
|
|
|
xfs_iunlock(ip, XFS_IOLOCK_SHARED);
|
|
|
|
|
|
|
|
out_wait:
|
2009-06-08 15:37:11 +02:00
|
|
|
if (flags & SYNC_WAIT)
|
2009-06-08 15:35:03 +02:00
|
|
|
xfs_ioend_wait(ip);
|
|
|
|
return error;
|
|
|
|
}
|
|
|
|
|
2009-06-08 15:35:05 +02:00
|
|
|
STATIC int
|
|
|
|
xfs_sync_inode_attr(
|
|
|
|
struct xfs_inode *ip,
|
2009-06-08 15:35:14 +02:00
|
|
|
struct xfs_perag *pag,
|
2009-06-08 15:35:05 +02:00
|
|
|
int flags)
|
|
|
|
{
|
|
|
|
int error = 0;
|
|
|
|
|
|
|
|
xfs_ilock(ip, XFS_ILOCK_SHARED);
|
|
|
|
if (xfs_inode_clean(ip))
|
|
|
|
goto out_unlock;
|
|
|
|
if (!xfs_iflock_nowait(ip)) {
|
|
|
|
if (!(flags & SYNC_WAIT))
|
|
|
|
goto out_unlock;
|
|
|
|
xfs_iflock(ip);
|
|
|
|
}
|
|
|
|
|
|
|
|
if (xfs_inode_clean(ip)) {
|
|
|
|
xfs_ifunlock(ip);
|
|
|
|
goto out_unlock;
|
|
|
|
}
|
|
|
|
|
2010-02-06 02:39:36 +01:00
|
|
|
error = xfs_iflush(ip, flags);
|
2009-06-08 15:35:05 +02:00
|
|
|
|
|
|
|
out_unlock:
|
|
|
|
xfs_iunlock(ip, XFS_ILOCK_SHARED);
|
|
|
|
return error;
|
|
|
|
}
|
|
|
|
|
2009-06-08 15:35:48 +02:00
|
|
|
/*
|
|
|
|
* Write out pagecache data for the whole filesystem.
|
|
|
|
*/
|
2010-06-24 03:45:34 +02:00
|
|
|
STATIC int
|
2009-06-08 15:35:48 +02:00
|
|
|
xfs_sync_data(
|
|
|
|
struct xfs_mount *mp,
|
|
|
|
int flags)
|
2008-10-30 07:07:29 +01:00
|
|
|
{
|
2009-06-08 15:35:48 +02:00
|
|
|
int error;
|
2008-10-30 07:06:08 +01:00
|
|
|
|
2009-06-08 15:37:11 +02:00
|
|
|
ASSERT((flags & ~(SYNC_TRYLOCK|SYNC_WAIT)) == 0);
|
2008-10-30 07:06:08 +01:00
|
|
|
|
2010-09-24 10:40:15 +02:00
|
|
|
error = xfs_inode_ag_iterator(mp, xfs_sync_inode_data, flags);
|
2009-06-08 15:35:48 +02:00
|
|
|
if (error)
|
|
|
|
return XFS_ERROR(error);
|
2008-10-30 07:15:50 +01:00
|
|
|
|
2010-01-19 10:56:46 +01:00
|
|
|
xfs_log_force(mp, (flags & SYNC_WAIT) ? XFS_LOG_SYNC : 0);
|
2009-06-08 15:35:48 +02:00
|
|
|
return 0;
|
|
|
|
}
|
2008-10-30 07:15:50 +01:00
|
|
|
|
2009-06-08 15:35:48 +02:00
|
|
|
/*
|
|
|
|
* Write out inode metadata (attributes) for the whole filesystem.
|
|
|
|
*/
|
2010-06-24 03:45:34 +02:00
|
|
|
STATIC int
|
2009-06-08 15:35:48 +02:00
|
|
|
xfs_sync_attr(
|
|
|
|
struct xfs_mount *mp,
|
|
|
|
int flags)
|
|
|
|
{
|
|
|
|
ASSERT((flags & ~SYNC_WAIT) == 0);
|
2009-06-08 15:35:14 +02:00
|
|
|
|
2010-09-24 10:40:15 +02:00
|
|
|
return xfs_inode_ag_iterator(mp, xfs_sync_inode_attr, flags);
|
2008-10-30 07:06:08 +01:00
|
|
|
}
|
|
|
|
|
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;
|
|
|
|
|
|
|
|
/*
|
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);
|
|
|
|
if (XFS_BUF_ISPINNED(bp))
|
|
|
|
xfs_log_force(mp, 0);
|
2008-10-30 07:14:53 +01:00
|
|
|
|
2010-03-12 11:59:16 +01:00
|
|
|
return xfs_bwrite(mp, bp);
|
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,
|
2008-10-30 07:16:11 +01:00
|
|
|
* transactions can still occur here so don't bother flushing the buftarg
|
|
|
|
* 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
|
|
|
|
|
|
|
/* push non-blocking */
|
2009-06-08 15:35:48 +02:00
|
|
|
xfs_sync_data(mp, 0);
|
2009-06-08 15:37:16 +02:00
|
|
|
xfs_qm_sync(mp, SYNC_TRYLOCK);
|
2008-10-30 07:15:50 +01:00
|
|
|
|
2009-10-06 22:29:27 +02:00
|
|
|
/* push and block till complete */
|
2009-06-08 15:37:11 +02:00
|
|
|
xfs_sync_data(mp, SYNC_WAIT);
|
2009-06-08 15:33:32 +02:00
|
|
|
xfs_qm_sync(mp, SYNC_WAIT);
|
2008-10-30 07:15:50 +01:00
|
|
|
|
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);
|
|
|
|
|
|
|
|
/* make sure all delwri buffers are written out */
|
|
|
|
xfs_flush_buftarg(mp->m_ddev_targp, 1);
|
|
|
|
|
|
|
|
/* 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
|
|
|
|
2008-10-30 07:16:11 +01:00
|
|
|
/* flush data-only devices */
|
2008-10-30 07:15:50 +01:00
|
|
|
if (mp->m_rtdev_targp)
|
|
|
|
XFS_bflush(mp->m_rtdev_targp);
|
|
|
|
|
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
|
|
|
STATIC void
|
|
|
|
xfs_quiesce_fs(
|
|
|
|
struct xfs_mount *mp)
|
|
|
|
{
|
|
|
|
int count = 0, pincount;
|
|
|
|
|
2010-02-06 02:39:36 +01:00
|
|
|
xfs_reclaim_inodes(mp, 0);
|
2008-10-30 07:16:21 +01:00
|
|
|
xfs_flush_buftarg(mp->m_ddev_targp, 0);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* This loop must run at least twice. The first instance of the loop
|
|
|
|
* will flush most meta data but that will generate more meta data
|
|
|
|
* (typically directory updates). Which then must be flushed and
|
2010-02-06 02:39:36 +01:00
|
|
|
* logged before we can write the unmount record. We also so sync
|
|
|
|
* reclaim of inodes to catch any that the above delwri flush skipped.
|
2008-10-30 07:16:21 +01:00
|
|
|
*/
|
|
|
|
do {
|
2010-02-06 02:39:36 +01:00
|
|
|
xfs_reclaim_inodes(mp, SYNC_WAIT);
|
2009-06-08 15:35:48 +02:00
|
|
|
xfs_sync_attr(mp, SYNC_WAIT);
|
2008-10-30 07:16:21 +01:00
|
|
|
pincount = xfs_flush_buftarg(mp->m_ddev_targp, 1);
|
|
|
|
if (!pincount) {
|
|
|
|
delay(50);
|
|
|
|
count++;
|
|
|
|
}
|
|
|
|
} while (count < 2);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* 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
|
|
|
|
* wait for any remaining transactions to drain out before proceding.
|
|
|
|
*/
|
|
|
|
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);
|
|
|
|
|
|
|
|
/* flush inodes and push all remaining buffers out to disk */
|
|
|
|
xfs_quiesce_fs(mp);
|
|
|
|
|
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 */
|
|
|
|
error = xfs_log_sbcount(mp, 1);
|
|
|
|
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);
|
|
|
|
xfs_unmountfs_writesb(mp);
|
|
|
|
}
|
|
|
|
|
2008-10-30 07:06:18 +01:00
|
|
|
/*
|
|
|
|
* Enqueue a work item to be picked up by the vfs xfssyncd thread.
|
|
|
|
* Doing this has two advantages:
|
|
|
|
* - It saves on stack space, which is tight in certain situations
|
|
|
|
* - It can be used (with care) as a mechanism to avoid deadlocks.
|
|
|
|
* Flushing while allocating in a full filesystem requires both.
|
|
|
|
*/
|
|
|
|
STATIC void
|
|
|
|
xfs_syncd_queue_work(
|
|
|
|
struct xfs_mount *mp,
|
|
|
|
void *data,
|
2009-04-06 18:47:27 +02:00
|
|
|
void (*syncer)(struct xfs_mount *, void *),
|
|
|
|
struct completion *completion)
|
2008-10-30 07:06:18 +01:00
|
|
|
{
|
2009-04-06 18:44:54 +02:00
|
|
|
struct xfs_sync_work *work;
|
2008-10-30 07:06:18 +01:00
|
|
|
|
2009-04-06 18:44:54 +02:00
|
|
|
work = kmem_alloc(sizeof(struct xfs_sync_work), KM_SLEEP);
|
2008-10-30 07:06:18 +01:00
|
|
|
INIT_LIST_HEAD(&work->w_list);
|
|
|
|
work->w_syncer = syncer;
|
|
|
|
work->w_data = data;
|
|
|
|
work->w_mount = mp;
|
2009-04-06 18:47:27 +02:00
|
|
|
work->w_completion = completion;
|
2008-10-30 07:06:18 +01:00
|
|
|
spin_lock(&mp->m_sync_lock);
|
|
|
|
list_add_tail(&work->w_list, &mp->m_sync_list);
|
|
|
|
spin_unlock(&mp->m_sync_lock);
|
|
|
|
wake_up_process(mp->m_sync_task);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Flush delayed allocate data, attempting to free up reserved space
|
|
|
|
* from existing allocations. At this point a new allocation attempt
|
|
|
|
* has failed with ENOSPC and we are in the process of scratching our
|
|
|
|
* heads, looking about for more room...
|
|
|
|
*/
|
|
|
|
STATIC void
|
2009-04-06 18:44:54 +02:00
|
|
|
xfs_flush_inodes_work(
|
2008-10-30 07:06:18 +01:00
|
|
|
struct xfs_mount *mp,
|
|
|
|
void *arg)
|
|
|
|
{
|
|
|
|
struct inode *inode = arg;
|
2009-06-08 15:35:48 +02:00
|
|
|
xfs_sync_data(mp, SYNC_TRYLOCK);
|
2009-06-08 15:37:11 +02:00
|
|
|
xfs_sync_data(mp, SYNC_TRYLOCK | SYNC_WAIT);
|
2008-10-30 07:06:18 +01:00
|
|
|
iput(inode);
|
|
|
|
}
|
|
|
|
|
|
|
|
void
|
2009-04-06 18:44:54 +02:00
|
|
|
xfs_flush_inodes(
|
2008-10-30 07:06:18 +01:00
|
|
|
xfs_inode_t *ip)
|
|
|
|
{
|
|
|
|
struct inode *inode = VFS_I(ip);
|
2009-04-06 18:47:27 +02:00
|
|
|
DECLARE_COMPLETION_ONSTACK(completion);
|
2008-10-30 07:06:18 +01:00
|
|
|
|
|
|
|
igrab(inode);
|
2009-04-06 18:47:27 +02:00
|
|
|
xfs_syncd_queue_work(ip->i_mount, inode, xfs_flush_inodes_work, &completion);
|
|
|
|
wait_for_completion(&completion);
|
2010-01-19 10:56:46 +01:00
|
|
|
xfs_log_force(ip->i_mount, XFS_LOG_SYNC);
|
2008-10-30 07:06:18 +01:00
|
|
|
}
|
|
|
|
|
2008-10-30 07:15:29 +01:00
|
|
|
/*
|
2010-03-12 11:59:16 +01:00
|
|
|
* Every sync period we need to unpin all items, reclaim inodes and sync
|
|
|
|
* disk quotas. We might need to cover the log to indicate that the
|
2010-08-24 03:46:31 +02:00
|
|
|
* filesystem is idle and not frozen.
|
2008-10-30 07:15:29 +01:00
|
|
|
*/
|
2008-10-30 07:06:18 +01:00
|
|
|
STATIC void
|
|
|
|
xfs_sync_worker(
|
|
|
|
struct xfs_mount *mp,
|
|
|
|
void *unused)
|
|
|
|
{
|
|
|
|
int error;
|
|
|
|
|
2008-10-30 07:15:29 +01:00
|
|
|
if (!(mp->m_flags & XFS_MOUNT_RDONLY)) {
|
|
|
|
/* dgc: errors ignored here */
|
2010-08-24 03:46:31 +02:00
|
|
|
if (mp->m_super->s_frozen == SB_UNFROZEN &&
|
|
|
|
xfs_log_need_covered(mp))
|
2011-01-04 05:49:29 +01:00
|
|
|
error = xfs_fs_log_dummy(mp);
|
|
|
|
else
|
|
|
|
xfs_log_force(mp, 0);
|
|
|
|
xfs_reclaim_inodes(mp, 0);
|
|
|
|
error = xfs_qm_sync(mp, SYNC_TRYLOCK);
|
2008-10-30 07:15:29 +01:00
|
|
|
}
|
2008-10-30 07:06:18 +01:00
|
|
|
mp->m_sync_seq++;
|
|
|
|
wake_up(&mp->m_wait_single_sync_task);
|
|
|
|
}
|
|
|
|
|
|
|
|
STATIC int
|
|
|
|
xfssyncd(
|
|
|
|
void *arg)
|
|
|
|
{
|
|
|
|
struct xfs_mount *mp = arg;
|
|
|
|
long timeleft;
|
2009-04-06 18:44:54 +02:00
|
|
|
xfs_sync_work_t *work, *n;
|
2008-10-30 07:06:18 +01:00
|
|
|
LIST_HEAD (tmp);
|
|
|
|
|
|
|
|
set_freezable();
|
|
|
|
timeleft = xfs_syncd_centisecs * msecs_to_jiffies(10);
|
|
|
|
for (;;) {
|
2010-03-04 02:46:23 +01:00
|
|
|
if (list_empty(&mp->m_sync_list))
|
|
|
|
timeleft = schedule_timeout_interruptible(timeleft);
|
2008-10-30 07:06:18 +01:00
|
|
|
/* swsusp */
|
|
|
|
try_to_freeze();
|
|
|
|
if (kthread_should_stop() && list_empty(&mp->m_sync_list))
|
|
|
|
break;
|
|
|
|
|
|
|
|
spin_lock(&mp->m_sync_lock);
|
|
|
|
/*
|
|
|
|
* We can get woken by laptop mode, to do a sync -
|
|
|
|
* that's the (only!) case where the list would be
|
|
|
|
* empty with time remaining.
|
|
|
|
*/
|
|
|
|
if (!timeleft || list_empty(&mp->m_sync_list)) {
|
|
|
|
if (!timeleft)
|
|
|
|
timeleft = xfs_syncd_centisecs *
|
|
|
|
msecs_to_jiffies(10);
|
|
|
|
INIT_LIST_HEAD(&mp->m_sync_work.w_list);
|
|
|
|
list_add_tail(&mp->m_sync_work.w_list,
|
|
|
|
&mp->m_sync_list);
|
|
|
|
}
|
2010-03-04 02:46:23 +01:00
|
|
|
list_splice_init(&mp->m_sync_list, &tmp);
|
2008-10-30 07:06:18 +01:00
|
|
|
spin_unlock(&mp->m_sync_lock);
|
|
|
|
|
|
|
|
list_for_each_entry_safe(work, n, &tmp, w_list) {
|
|
|
|
(*work->w_syncer)(mp, work->w_data);
|
|
|
|
list_del(&work->w_list);
|
|
|
|
if (work == &mp->m_sync_work)
|
|
|
|
continue;
|
2009-04-06 18:47:27 +02:00
|
|
|
if (work->w_completion)
|
|
|
|
complete(work->w_completion);
|
2008-10-30 07:06:18 +01:00
|
|
|
kmem_free(work);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
int
|
|
|
|
xfs_syncd_init(
|
|
|
|
struct xfs_mount *mp)
|
|
|
|
{
|
|
|
|
mp->m_sync_work.w_syncer = xfs_sync_worker;
|
|
|
|
mp->m_sync_work.w_mount = mp;
|
2009-04-06 18:47:27 +02:00
|
|
|
mp->m_sync_work.w_completion = NULL;
|
2010-03-22 23:52:55 +01:00
|
|
|
mp->m_sync_task = kthread_run(xfssyncd, mp, "xfssyncd/%s", mp->m_fsname);
|
2008-10-30 07:06:18 +01:00
|
|
|
if (IS_ERR(mp->m_sync_task))
|
|
|
|
return -PTR_ERR(mp->m_sync_task);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
void
|
|
|
|
xfs_syncd_stop(
|
|
|
|
struct xfs_mount *mp)
|
|
|
|
{
|
|
|
|
kthread_stop(mp->m_sync_task);
|
|
|
|
}
|
|
|
|
|
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);
|
|
|
|
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
|
|
|
|
|
|
|
/*
|
2010-12-17 07:29:43 +01:00
|
|
|
* do some unlocked checks first to avoid unnecessary lock traffic.
|
2010-09-24 11:51:50 +02:00
|
|
|
* The first is a flush lock check, the second is a already in reclaim
|
|
|
|
* check. Only do these checks if we are not going to block on locks.
|
|
|
|
*/
|
|
|
|
if ((flags & SYNC_TRYLOCK) &&
|
|
|
|
(!ip->i_flush.done || __xfs_iflags_test(ip, XFS_IRECLAIM))) {
|
|
|
|
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
|
|
|
/*
|
|
|
|
* Inodes in different states need to be treated differently, and the return
|
|
|
|
* value of xfs_iflush is not sufficient to get this right. The following table
|
|
|
|
* lists the inode states and the reclaim actions necessary for non-blocking
|
|
|
|
* reclaim:
|
|
|
|
*
|
|
|
|
*
|
|
|
|
* 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
|
|
|
|
* dirty, delwri ok 0 requeue
|
|
|
|
* dirty, delwri blocked EAGAIN requeue
|
|
|
|
* dirty, sync flush 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
|
|
|
* As can be seen from the table, the return value of xfs_iflush() is not
|
|
|
|
* sufficient to correctly decide the reclaim action here. The checks in
|
|
|
|
* xfs_iflush() might look like duplicates, but they are not.
|
|
|
|
*
|
|
|
|
* 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
|
|
|
|
* the inode is clean. The clean inode check needs to be done before flushing
|
|
|
|
* the inode delwri otherwise we would loop forever requeuing clean inodes as
|
|
|
|
* we cannot tell apart a successful delwri flush and a clean inode from the
|
|
|
|
* return value of xfs_iflush().
|
|
|
|
*
|
|
|
|
* Note that because the inode is flushed delayed write by background
|
|
|
|
* writeback, 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 out delayed write inodes first before
|
|
|
|
* trying to reclaim them to minimise the amount of time spent waiting. For
|
|
|
|
* background relaim, we just requeue the inode for the next pass.
|
|
|
|
*
|
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
|
2010-02-06 02:39:36 +01:00
|
|
|
* pinned, delwri => requeue
|
|
|
|
* pinned, sync => unpin
|
2010-02-06 02:37:26 +01:00
|
|
|
* stale => reclaim
|
|
|
|
* clean => reclaim
|
2010-02-06 02:39:36 +01:00
|
|
|
* dirty, delwri => flush and requeue
|
|
|
|
* 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
|
|
|
{
|
2010-02-06 02:39:36 +01:00
|
|
|
int error = 0;
|
2010-02-06 02:37:26 +01:00
|
|
|
|
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);
|
|
|
|
goto reclaim;
|
|
|
|
}
|
2010-02-06 02:39:36 +01:00
|
|
|
if (xfs_ipincount(ip)) {
|
|
|
|
if (!(sync_mode & SYNC_WAIT)) {
|
|
|
|
xfs_ifunlock(ip);
|
|
|
|
goto out;
|
|
|
|
}
|
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;
|
|
|
|
|
|
|
|
/* Now we have an inode that needs flushing */
|
|
|
|
error = xfs_iflush(ip, sync_mode);
|
2010-02-06 02:39:36 +01:00
|
|
|
if (sync_mode & SYNC_WAIT) {
|
|
|
|
xfs_iflock(ip);
|
|
|
|
goto reclaim;
|
2010-01-11 00:51:45 +01:00
|
|
|
}
|
|
|
|
|
2010-02-06 02:39:36 +01:00
|
|
|
/*
|
|
|
|
* When we have to flush an inode but don't have SYNC_WAIT set, we
|
|
|
|
* flush the inode out using a delwri buffer and wait for the next
|
|
|
|
* call into reclaim to find it in a clean state instead of waiting for
|
|
|
|
* it now. We also don't return errors here - if the error is transient
|
|
|
|
* then the next reclaim pass will flush the inode, and if the error
|
2010-04-13 07:06:45 +02:00
|
|
|
* is permanent then the next sync reclaim will reclaim the inode and
|
2010-02-06 02:39:36 +01:00
|
|
|
* pass on the error.
|
|
|
|
*/
|
2010-04-13 07:06:45 +02:00
|
|
|
if (error && error != EAGAIN && !XFS_FORCED_SHUTDOWN(ip->i_mount)) {
|
2011-03-07 00:00:35 +01:00
|
|
|
xfs_warn(ip->i_mount,
|
2010-02-06 02:39:36 +01:00
|
|
|
"inode 0x%llx background reclaim flush failed with %d",
|
|
|
|
(long long)ip->i_ino, error);
|
|
|
|
}
|
|
|
|
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;
|
|
|
|
|
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
|
|
|
|
* unlocked after the lookup before we go ahead and free it. We get
|
|
|
|
* both the ilock and the iolock because the code may need to drop the
|
|
|
|
* ilock one but will still hold the iolock.
|
|
|
|
*/
|
|
|
|
xfs_ilock(ip, XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL);
|
|
|
|
xfs_qm_dqdetach(ip);
|
|
|
|
xfs_iunlock(ip, XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL);
|
|
|
|
|
|
|
|
xfs_inode_free(ip);
|
2010-02-06 02:39:36 +01:00
|
|
|
return error;
|
|
|
|
|
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) {
|
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
|
|
|
|
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.
|
|
|
|
*/
|
|
|
|
if (trylock && skipped && *nr_to_scan > 0) {
|
|
|
|
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
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Shrinker infrastructure.
|
|
|
|
*/
|
|
|
|
static int
|
|
|
|
xfs_reclaim_inode_shrink(
|
2010-07-19 06:56:17 +02:00
|
|
|
struct shrinker *shrink,
|
2010-04-29 01:55:50 +02:00
|
|
|
int nr_to_scan,
|
|
|
|
gfp_t gfp_mask)
|
|
|
|
{
|
|
|
|
struct xfs_mount *mp;
|
|
|
|
struct xfs_perag *pag;
|
|
|
|
xfs_agnumber_t ag;
|
2010-07-20 01:43:39 +02:00
|
|
|
int reclaimable;
|
2010-04-29 01:55:50 +02:00
|
|
|
|
2010-07-20 00:07:02 +02:00
|
|
|
mp = container_of(shrink, struct xfs_mount, m_inode_shrink);
|
2010-04-29 01:55:50 +02:00
|
|
|
if (nr_to_scan) {
|
|
|
|
if (!(gfp_mask & __GFP_FS))
|
|
|
|
return -1;
|
|
|
|
|
2010-09-24 11:51:50 +02:00
|
|
|
xfs_reclaim_inodes_ag(mp, SYNC_TRYLOCK, &nr_to_scan);
|
2010-09-24 10:40:15 +02:00
|
|
|
/* terminate if we don't exhaust the scan */
|
2010-07-20 00:07:02 +02:00
|
|
|
if (nr_to_scan > 0)
|
|
|
|
return -1;
|
|
|
|
}
|
2010-04-29 01:55:50 +02:00
|
|
|
|
2010-07-20 01:43:39 +02:00
|
|
|
reclaimable = 0;
|
|
|
|
ag = 0;
|
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;
|
|
|
|
}
|
|
|
|
|
|
|
|
void
|
|
|
|
xfs_inode_shrinker_register(
|
|
|
|
struct xfs_mount *mp)
|
|
|
|
{
|
2010-07-20 00:07:02 +02:00
|
|
|
mp->m_inode_shrink.shrink = xfs_reclaim_inode_shrink;
|
|
|
|
mp->m_inode_shrink.seeks = DEFAULT_SEEKS;
|
|
|
|
register_shrinker(&mp->m_inode_shrink);
|
2010-04-29 01:55:50 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
void
|
|
|
|
xfs_inode_shrinker_unregister(
|
|
|
|
struct xfs_mount *mp)
|
|
|
|
{
|
2010-07-20 00:07:02 +02:00
|
|
|
unregister_shrinker(&mp->m_inode_shrink);
|
2008-10-30 07:37:03 +01:00
|
|
|
}
|