xfs: rename xfs_sync.[ch] to xfs_icache.[ch]
authorDave Chinner <dchinner@redhat.com>
Mon, 8 Oct 2012 10:56:09 +0000 (21:56 +1100)
committerBen Myers <bpm@sgi.com>
Wed, 17 Oct 2012 18:40:09 +0000 (13:40 -0500)
xfs_sync.c now only contains inode reclaim functions and inode cache
iteration functions. It is not related to sync operations anymore.
Rename to xfs_icache.c to reflect it's contents and prepare for
consolidation with the other inode cache file that exists
(xfs_iget.c).

Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Mark Tinguely <tinguely@sgi.com>
Signed-off-by: Ben Myers <bpm@sgi.com>
fs/xfs/Makefile
fs/xfs/xfs_icache.c [new file with mode: 0644]
fs/xfs/xfs_icache.h [new file with mode: 0644]
fs/xfs/xfs_iget.c
fs/xfs/xfs_mount.c
fs/xfs/xfs_mount.h
fs/xfs/xfs_qm_syscalls.c
fs/xfs/xfs_super.c
fs/xfs/xfs_sync.c [deleted file]
fs/xfs/xfs_sync.h [deleted file]

index d2bf974b1a2f68db6a47901cebd060bb49518bd1..442f256dbcac6a3a9b33fdcf65610e9233bf36f8 100644 (file)
@@ -39,6 +39,7 @@ xfs-y                         += xfs_aops.o \
                                   xfs_fsops.o \
                                   xfs_fs_subr.o \
                                   xfs_globals.o \
+                                  xfs_icache.o \
                                   xfs_iget.o \
                                   xfs_ioctl.o \
                                   xfs_iomap.o \
@@ -47,7 +48,6 @@ xfs-y                         += xfs_aops.o \
                                   xfs_message.o \
                                   xfs_mru_cache.o \
                                   xfs_super.o \
-                                  xfs_sync.o \
                                   xfs_xattr.o \
                                   xfs_rename.o \
                                   xfs_utils.o \
diff --git a/fs/xfs/xfs_icache.c b/fs/xfs/xfs_icache.c
new file mode 100644 (file)
index 0000000..eba216f
--- /dev/null
@@ -0,0 +1,715 @@
+/*
+ * 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"
+#include "xfs_log_priv.h"
+#include "xfs_inum.h"
+#include "xfs_trans.h"
+#include "xfs_trans_priv.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"
+#include "xfs_quota.h"
+#include "xfs_trace.h"
+#include "xfs_fsops.h"
+#include "xfs_icache.h"
+
+#include <linux/kthread.h>
+#include <linux/freezer.h>
+
+/*
+ * 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
+
+STATIC int
+xfs_inode_ag_walk_grab(
+       struct xfs_inode        *ip)
+{
+       struct inode            *inode = VFS_I(ip);
+
+       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);
+
+       /* 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;
+
+out_unlock_noent:
+       spin_unlock(&ip->i_flags_lock);
+       return ENOENT;
+}
+
+STATIC int
+xfs_inode_ag_walk(
+       struct xfs_mount        *mp,
+       struct xfs_perag        *pag,
+       int                     (*execute)(struct xfs_inode *ip,
+                                          struct xfs_perag *pag, int flags),
+       int                     flags)
+{
+       uint32_t                first_index;
+       int                     last_error = 0;
+       int                     skipped;
+       int                     done;
+       int                     nr_found;
+
+restart:
+       done = 0;
+       skipped = 0;
+       first_index = 0;
+       nr_found = 0;
+       do {
+               struct xfs_inode *batch[XFS_LOOKUP_BATCH];
+               int             error = 0;
+               int             i;
+
+               rcu_read_lock();
+               nr_found = radix_tree_gang_lookup(&pag->pag_ici_root,
+                                       (void **)batch, first_index,
+                                       XFS_LOOKUP_BATCH);
+               if (!nr_found) {
+                       rcu_read_unlock();
+                       break;
+               }
+
+               /*
+                * Grab the inodes before we drop the lock. if we found
+                * nothing, nr == 0 and the loop will be skipped.
+                */
+               for (i = 0; i < nr_found; i++) {
+                       struct xfs_inode *ip = batch[i];
+
+                       if (done || xfs_inode_ag_walk_grab(ip))
+                               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.
+                        *
+                        * 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.
+                        */
+                       if (XFS_INO_TO_AGNO(mp, ip->i_ino) != pag->pag_agno)
+                               continue;
+                       first_index = XFS_INO_TO_AGINO(mp, ip->i_ino + 1);
+                       if (first_index < XFS_INO_TO_AGINO(mp, ip->i_ino))
+                               done = 1;
+               }
+
+               /* unlock now we've grabbed the inodes. */
+               rcu_read_unlock();
+
+               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;
+               }
+
+               /* bail out if the filesystem is corrupted.  */
+               if (error == EFSCORRUPTED)
+                       break;
+
+               cond_resched();
+
+       } while (nr_found && !done);
+
+       if (skipped) {
+               delay(1);
+               goto restart;
+       }
+       return last_error;
+}
+
+int
+xfs_inode_ag_iterator(
+       struct xfs_mount        *mp,
+       int                     (*execute)(struct xfs_inode *ip,
+                                          struct xfs_perag *pag, int flags),
+       int                     flags)
+{
+       struct xfs_perag        *pag;
+       int                     error = 0;
+       int                     last_error = 0;
+       xfs_agnumber_t          ag;
+
+       ag = 0;
+       while ((pag = xfs_perag_get(mp, ag))) {
+               ag = pag->pag_agno + 1;
+               error = xfs_inode_ag_walk(mp, pag, execute, flags);
+               xfs_perag_put(pag);
+               if (error) {
+                       last_error = error;
+                       if (error == EFSCORRUPTED)
+                               break;
+               }
+       }
+       return XFS_ERROR(last_error);
+}
+
+/*
+ * 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 periodic sync 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_reclaim_work_queue(
+       struct xfs_mount        *mp)
+{
+
+       rcu_read_lock();
+       if (radix_tree_tagged(&mp->m_perag_tree, XFS_ICI_RECLAIM_TAG)) {
+               queue_delayed_work(mp->m_reclaim_workqueue, &mp->m_reclaim_work,
+                       msecs_to_jiffies(xfs_syncd_centisecs / 6 * 10));
+       }
+       rcu_read_unlock();
+}
+
+/*
+ * 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.
+ */
+void
+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_reclaim_work_queue(mp);
+}
+
+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);
+
+       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);
+
+               /* schedule periodic background inode reclaim */
+               xfs_reclaim_work_queue(ip->i_mount);
+
+               trace_xfs_perag_set_reclaim(ip->i_mount, pag->pag_agno,
+                                                       -1, _RET_IP_);
+       }
+       pag->pag_ici_reclaimable++;
+}
+
+/*
+ * 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.
+ */
+void
+xfs_inode_set_reclaim_tag(
+       xfs_inode_t     *ip)
+{
+       struct xfs_mount *mp = ip->i_mount;
+       struct xfs_perag *pag;
+
+       pag = xfs_perag_get(mp, XFS_INO_TO_AGNO(mp, ip->i_ino));
+       spin_lock(&pag->pag_ici_lock);
+       spin_lock(&ip->i_flags_lock);
+       __xfs_inode_set_reclaim_tag(pag, ip);
+       __xfs_iflags_set(ip, XFS_IRECLAIMABLE);
+       spin_unlock(&ip->i_flags_lock);
+       spin_unlock(&pag->pag_ici_lock);
+       xfs_perag_put(pag);
+}
+
+STATIC void
+__xfs_inode_clear_reclaim(
+       xfs_perag_t     *pag,
+       xfs_inode_t     *ip)
+{
+       pag->pag_ici_reclaimable--;
+       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_);
+       }
+}
+
+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);
+}
+
+/*
+ * 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)
+{
+       ASSERT(rcu_read_lock_held());
+
+       /* quick check for stale RCU freed inode */
+       if (!ip->i_ino)
+               return 1;
+
+       /*
+        * 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.
+        */
+       if ((flags & SYNC_TRYLOCK) &&
+           __xfs_iflags_test(ip, XFS_IFLOCK | 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.
+        *
+        * 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.
+        */
+       spin_lock(&ip->i_flags_lock);
+       if (!__xfs_iflags_test(ip, XFS_IRECLAIMABLE) ||
+           __xfs_iflags_test(ip, XFS_IRECLAIM)) {
+               /* not a reclaim candidate. */
+               spin_unlock(&ip->i_flags_lock);
+               return 1;
+       }
+       __xfs_iflags_set(ip, XFS_IRECLAIM);
+       spin_unlock(&ip->i_flags_lock);
+       return 0;
+}
+
+/*
+ * Inodes in different states need to be treated differently. The following
+ * table lists the inode states and the reclaim actions necessary:
+ *
+ *     inode state          iflush ret         required action
+ *      ---------------      ----------         ---------------
+ *     bad                     -               reclaim
+ *     shutdown                EIO             unpin and reclaim
+ *     clean, unpinned         0               reclaim
+ *     stale, unpinned         0               reclaim
+ *     clean, pinned(*)        0               requeue
+ *     stale, pinned           EAGAIN          requeue
+ *     dirty, async            -               requeue
+ *     dirty, sync             0               reclaim
+ *
+ * (*) dgc: I don't think the clean, pinned state is possible but it gets
+ * handled anyway given the order of checks implemented.
+ *
+ * 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.
+ *
+ * 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.
+ *
+ * Hence the order of actions after gaining the locks should be:
+ *     bad             => reclaim
+ *     shutdown        => unpin and reclaim
+ *     pinned, async   => requeue
+ *     pinned, sync    => unpin
+ *     stale           => reclaim
+ *     clean           => reclaim
+ *     dirty, async    => requeue
+ *     dirty, sync     => flush, wait and reclaim
+ */
+STATIC int
+xfs_reclaim_inode(
+       struct xfs_inode        *ip,
+       struct xfs_perag        *pag,
+       int                     sync_mode)
+{
+       struct xfs_buf          *bp = NULL;
+       int                     error;
+
+restart:
+       error = 0;
+       xfs_ilock(ip, XFS_ILOCK_EXCL);
+       if (!xfs_iflock_nowait(ip)) {
+               if (!(sync_mode & SYNC_WAIT))
+                       goto out;
+               xfs_iflock(ip);
+       }
+
+       if (is_bad_inode(VFS_I(ip)))
+               goto reclaim;
+       if (XFS_FORCED_SHUTDOWN(ip->i_mount)) {
+               xfs_iunpin_wait(ip);
+               xfs_iflush_abort(ip, false);
+               goto reclaim;
+       }
+       if (xfs_ipincount(ip)) {
+               if (!(sync_mode & SYNC_WAIT))
+                       goto out_ifunlock;
+               xfs_iunpin_wait(ip);
+       }
+       if (xfs_iflags_test(ip, XFS_ISTALE))
+               goto reclaim;
+       if (xfs_inode_clean(ip))
+               goto reclaim;
+
+       /*
+        * 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;
+
+       /*
+        * Now we have an inode that needs flushing.
+        *
+        * Note that xfs_iflush will never block on the inode buffer lock, as
+        * xfs_ifree_cluster() can lock the inode buffer before it locks the
+        * ip->i_lock, and we are doing the exact opposite here.  As a result,
+        * doing a blocking xfs_imap_to_bp() to get the cluster buffer would
+        * result in an ABBA deadlock with xfs_ifree_cluster().
+        *
+        * 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
+        * 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.
+        */
+       error = xfs_iflush(ip, &bp);
+       if (error == EAGAIN) {
+               xfs_iunlock(ip, XFS_ILOCK_EXCL);
+               /* backoff longer than in xfs_ifree_cluster */
+               delay(2);
+               goto restart;
+       }
+
+       if (!error) {
+               error = xfs_bwrite(bp);
+               xfs_buf_relse(bp);
+       }
+
+       xfs_iflock(ip);
+reclaim:
+       xfs_ifunlock(ip);
+       xfs_iunlock(ip, XFS_ILOCK_EXCL);
+
+       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.
+        */
+       spin_lock(&pag->pag_ici_lock);
+       if (!radix_tree_delete(&pag->pag_ici_root,
+                               XFS_INO_TO_AGINO(ip->i_mount, ip->i_ino)))
+               ASSERT(0);
+       __xfs_inode_clear_reclaim(pag, ip);
+       spin_unlock(&pag->pag_ici_lock);
+
+       /*
+        * 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.
+        */
+       xfs_ilock(ip, XFS_ILOCK_EXCL);
+       xfs_qm_dqdetach(ip);
+       xfs_iunlock(ip, XFS_ILOCK_EXCL);
+
+       xfs_inode_free(ip);
+       return error;
+
+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 the reclaim work 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;
+}
+
+/*
+ * 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;
+       int                     trylock = flags & SYNC_TRYLOCK;
+       int                     skipped;
+
+restart:
+       ag = 0;
+       skipped = 0;
+       while ((pag = xfs_perag_get_tag(mp, ag, XFS_ICI_RECLAIM_TAG))) {
+               unsigned long   first_index = 0;
+               int             done = 0;
+               int             nr_found = 0;
+
+               ag = pag->pag_agno + 1;
+
+               if (trylock) {
+                       if (!mutex_trylock(&pag->pag_ici_reclaim_lock)) {
+                               skipped++;
+                               xfs_perag_put(pag);
+                               continue;
+                       }
+                       first_index = pag->pag_ici_reclaim_cursor;
+               } else
+                       mutex_lock(&pag->pag_ici_reclaim_lock);
+
+               do {
+                       struct xfs_inode *batch[XFS_LOOKUP_BATCH];
+                       int     i;
+
+                       rcu_read_lock();
+                       nr_found = radix_tree_gang_lookup_tag(
+                                       &pag->pag_ici_root,
+                                       (void **)batch, first_index,
+                                       XFS_LOOKUP_BATCH,
+                                       XFS_ICI_RECLAIM_TAG);
+                       if (!nr_found) {
+                               done = 1;
+                               rcu_read_unlock();
+                               break;
+                       }
+
+                       /*
+                        * Grab the inodes before we drop the lock. if we found
+                        * nothing, nr == 0 and the loop will be skipped.
+                        */
+                       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.
+                                *
+                                * 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.
+                                */
+                               if (XFS_INO_TO_AGNO(mp, ip->i_ino) !=
+                                                               pag->pag_agno)
+                                       continue;
+                               first_index = XFS_INO_TO_AGINO(mp, ip->i_ino + 1);
+                               if (first_index < XFS_INO_TO_AGINO(mp, ip->i_ino))
+                                       done = 1;
+                       }
+
+                       /* unlock now we've grabbed the inodes. */
+                       rcu_read_unlock();
+
+                       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;
+
+                       cond_resched();
+
+               } while (nr_found && !done && *nr_to_scan > 0);
+
+               if (trylock && !done)
+                       pag->pag_ici_reclaim_cursor = first_index;
+               else
+                       pag->pag_ici_reclaim_cursor = 0;
+               mutex_unlock(&pag->pag_ici_reclaim_lock);
+               xfs_perag_put(pag);
+       }
+
+       /*
+        * 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 (skipped && (flags & SYNC_WAIT) && *nr_to_scan > 0) {
+               trylock = 0;
+               goto restart;
+       }
+       return XFS_ERROR(last_error);
+}
+
+int
+xfs_reclaim_inodes(
+       xfs_mount_t     *mp,
+       int             mode)
+{
+       int             nr_to_scan = INT_MAX;
+
+       return xfs_reclaim_inodes_ag(mp, mode, &nr_to_scan);
+}
+
+/*
+ * Scan a certain number of inodes for reclaim.
+ *
+ * When called we make sure that there is a background (fast) inode reclaim in
+ * progress, while we will throttle the speed of reclaim via doing synchronous
+ * 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.
+ */
+void
+xfs_reclaim_inodes_nr(
+       struct xfs_mount        *mp,
+       int                     nr_to_scan)
+{
+       /* kick background reclaimer and push the AIL */
+       xfs_reclaim_work_queue(mp);
+       xfs_ail_push_all(mp->m_ail);
+
+       xfs_reclaim_inodes_ag(mp, SYNC_TRYLOCK | SYNC_WAIT, &nr_to_scan);
+}
+
+/*
+ * 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;
+
+       while ((pag = xfs_perag_get_tag(mp, ag, XFS_ICI_RECLAIM_TAG))) {
+               ag = pag->pag_agno + 1;
+               reclaimable += pag->pag_ici_reclaimable;
+               xfs_perag_put(pag);
+       }
+       return reclaimable;
+}
+
diff --git a/fs/xfs/xfs_icache.h b/fs/xfs/xfs_icache.h
new file mode 100644 (file)
index 0000000..0ba9c89
--- /dev/null
@@ -0,0 +1,43 @@
+/*
+ * Copyright (c) 2000-2006 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
+ */
+#ifndef XFS_SYNC_H
+#define XFS_SYNC_H 1
+
+struct xfs_mount;
+struct xfs_perag;
+
+#define SYNC_WAIT              0x0001  /* wait for i/o to complete */
+#define SYNC_TRYLOCK           0x0002  /* only try to lock inodes */
+
+void xfs_reclaim_worker(struct work_struct *work);
+
+int xfs_reclaim_inodes(struct xfs_mount *mp, int mode);
+int xfs_reclaim_inodes_count(struct xfs_mount *mp);
+void xfs_reclaim_inodes_nr(struct xfs_mount *mp, int nr_to_scan);
+
+void xfs_inode_set_reclaim_tag(struct xfs_inode *ip);
+void __xfs_inode_set_reclaim_tag(struct xfs_perag *pag, struct xfs_inode *ip);
+void __xfs_inode_clear_reclaim_tag(struct xfs_mount *mp, struct xfs_perag *pag,
+                               struct xfs_inode *ip);
+
+int xfs_sync_inode_grab(struct xfs_inode *ip);
+int xfs_inode_ag_iterator(struct xfs_mount *mp,
+       int (*execute)(struct xfs_inode *ip, struct xfs_perag *pag, int flags),
+       int flags);
+
+#endif
index 784a803383ec01d670366fc6a334dcc9442176ca..069c5ceb9459d490362a22d1ac1bbdc52efa821d 100644 (file)
@@ -38,6 +38,7 @@
 #include "xfs_inode_item.h"
 #include "xfs_bmap.h"
 #include "xfs_trace.h"
+#include "xfs_icache.h"
 
 
 /*
index c195ec85c7253cd1efbcb49bdf58b39775a69f5a..6f1c997704cdc8c8065a907dc4e7096d256fe3b4 100644 (file)
@@ -42,6 +42,7 @@
 #include "xfs_fsops.h"
 #include "xfs_utils.h"
 #include "xfs_trace.h"
+#include "xfs_icache.h"
 
 
 #ifdef HAVE_PERCPU_SB
index 7c417b6b99ee24d44d6d9e0731b572fc9949f0a8..a631ca3b9065469d39a688da1f906bc6a6063bf2 100644 (file)
@@ -51,8 +51,6 @@ typedef struct xfs_trans_reservations {
 
 #else /* __KERNEL__ */
 
-#include "xfs_sync.h"
-
 struct xlog;
 struct xfs_inode;
 struct xfs_mru_cache;
index 858a3b186110ab74760ebebf0d6b8c1326a9ee5a..7a9071f8855fc85b09d324b4d9115eaad4e7da85 100644 (file)
@@ -40,6 +40,7 @@
 #include "xfs_utils.h"
 #include "xfs_qm.h"
 #include "xfs_trace.h"
+#include "xfs_icache.h"
 
 STATIC int     xfs_qm_log_quotaoff(xfs_mount_t *, xfs_qoff_logitem_t **, uint);
 STATIC int     xfs_qm_log_quotaoff_end(xfs_mount_t *, xfs_qoff_logitem_t *,
index fdedf2cabae3a3936de1c4c4fe1525ccdadda9bd..3d9ea947e9f80d347518c98081be7c645799a8fa 100644 (file)
@@ -49,7 +49,7 @@
 #include "xfs_extfree_item.h"
 #include "xfs_mru_cache.h"
 #include "xfs_inode_item.h"
-#include "xfs_sync.h"
+#include "xfs_icache.h"
 #include "xfs_trace.h"
 
 #include <linux/namei.h>
diff --git a/fs/xfs/xfs_sync.c b/fs/xfs/xfs_sync.c
deleted file mode 100644 (file)
index 7b63028..0000000
+++ /dev/null
@@ -1,714 +0,0 @@
-/*
- * 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"
-#include "xfs_log_priv.h"
-#include "xfs_inum.h"
-#include "xfs_trans.h"
-#include "xfs_trans_priv.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"
-#include "xfs_quota.h"
-#include "xfs_trace.h"
-#include "xfs_fsops.h"
-
-#include <linux/kthread.h>
-#include <linux/freezer.h>
-
-/*
- * 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
-
-STATIC int
-xfs_inode_ag_walk_grab(
-       struct xfs_inode        *ip)
-{
-       struct inode            *inode = VFS_I(ip);
-
-       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);
-
-       /* 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;
-
-out_unlock_noent:
-       spin_unlock(&ip->i_flags_lock);
-       return ENOENT;
-}
-
-STATIC int
-xfs_inode_ag_walk(
-       struct xfs_mount        *mp,
-       struct xfs_perag        *pag,
-       int                     (*execute)(struct xfs_inode *ip,
-                                          struct xfs_perag *pag, int flags),
-       int                     flags)
-{
-       uint32_t                first_index;
-       int                     last_error = 0;
-       int                     skipped;
-       int                     done;
-       int                     nr_found;
-
-restart:
-       done = 0;
-       skipped = 0;
-       first_index = 0;
-       nr_found = 0;
-       do {
-               struct xfs_inode *batch[XFS_LOOKUP_BATCH];
-               int             error = 0;
-               int             i;
-
-               rcu_read_lock();
-               nr_found = radix_tree_gang_lookup(&pag->pag_ici_root,
-                                       (void **)batch, first_index,
-                                       XFS_LOOKUP_BATCH);
-               if (!nr_found) {
-                       rcu_read_unlock();
-                       break;
-               }
-
-               /*
-                * Grab the inodes before we drop the lock. if we found
-                * nothing, nr == 0 and the loop will be skipped.
-                */
-               for (i = 0; i < nr_found; i++) {
-                       struct xfs_inode *ip = batch[i];
-
-                       if (done || xfs_inode_ag_walk_grab(ip))
-                               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.
-                        *
-                        * 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.
-                        */
-                       if (XFS_INO_TO_AGNO(mp, ip->i_ino) != pag->pag_agno)
-                               continue;
-                       first_index = XFS_INO_TO_AGINO(mp, ip->i_ino + 1);
-                       if (first_index < XFS_INO_TO_AGINO(mp, ip->i_ino))
-                               done = 1;
-               }
-
-               /* unlock now we've grabbed the inodes. */
-               rcu_read_unlock();
-
-               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;
-               }
-
-               /* bail out if the filesystem is corrupted.  */
-               if (error == EFSCORRUPTED)
-                       break;
-
-               cond_resched();
-
-       } while (nr_found && !done);
-
-       if (skipped) {
-               delay(1);
-               goto restart;
-       }
-       return last_error;
-}
-
-int
-xfs_inode_ag_iterator(
-       struct xfs_mount        *mp,
-       int                     (*execute)(struct xfs_inode *ip,
-                                          struct xfs_perag *pag, int flags),
-       int                     flags)
-{
-       struct xfs_perag        *pag;
-       int                     error = 0;
-       int                     last_error = 0;
-       xfs_agnumber_t          ag;
-
-       ag = 0;
-       while ((pag = xfs_perag_get(mp, ag))) {
-               ag = pag->pag_agno + 1;
-               error = xfs_inode_ag_walk(mp, pag, execute, flags);
-               xfs_perag_put(pag);
-               if (error) {
-                       last_error = error;
-                       if (error == EFSCORRUPTED)
-                               break;
-               }
-       }
-       return XFS_ERROR(last_error);
-}
-
-/*
- * 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 periodic sync 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_reclaim_work_queue(
-       struct xfs_mount        *mp)
-{
-
-       rcu_read_lock();
-       if (radix_tree_tagged(&mp->m_perag_tree, XFS_ICI_RECLAIM_TAG)) {
-               queue_delayed_work(mp->m_reclaim_workqueue, &mp->m_reclaim_work,
-                       msecs_to_jiffies(xfs_syncd_centisecs / 6 * 10));
-       }
-       rcu_read_unlock();
-}
-
-/*
- * 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.
- */
-void
-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_reclaim_work_queue(mp);
-}
-
-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);
-
-       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);
-
-               /* schedule periodic background inode reclaim */
-               xfs_reclaim_work_queue(ip->i_mount);
-
-               trace_xfs_perag_set_reclaim(ip->i_mount, pag->pag_agno,
-                                                       -1, _RET_IP_);
-       }
-       pag->pag_ici_reclaimable++;
-}
-
-/*
- * 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.
- */
-void
-xfs_inode_set_reclaim_tag(
-       xfs_inode_t     *ip)
-{
-       struct xfs_mount *mp = ip->i_mount;
-       struct xfs_perag *pag;
-
-       pag = xfs_perag_get(mp, XFS_INO_TO_AGNO(mp, ip->i_ino));
-       spin_lock(&pag->pag_ici_lock);
-       spin_lock(&ip->i_flags_lock);
-       __xfs_inode_set_reclaim_tag(pag, ip);
-       __xfs_iflags_set(ip, XFS_IRECLAIMABLE);
-       spin_unlock(&ip->i_flags_lock);
-       spin_unlock(&pag->pag_ici_lock);
-       xfs_perag_put(pag);
-}
-
-STATIC void
-__xfs_inode_clear_reclaim(
-       xfs_perag_t     *pag,
-       xfs_inode_t     *ip)
-{
-       pag->pag_ici_reclaimable--;
-       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_);
-       }
-}
-
-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);
-}
-
-/*
- * 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)
-{
-       ASSERT(rcu_read_lock_held());
-
-       /* quick check for stale RCU freed inode */
-       if (!ip->i_ino)
-               return 1;
-
-       /*
-        * 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.
-        */
-       if ((flags & SYNC_TRYLOCK) &&
-           __xfs_iflags_test(ip, XFS_IFLOCK | 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.
-        *
-        * 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.
-        */
-       spin_lock(&ip->i_flags_lock);
-       if (!__xfs_iflags_test(ip, XFS_IRECLAIMABLE) ||
-           __xfs_iflags_test(ip, XFS_IRECLAIM)) {
-               /* not a reclaim candidate. */
-               spin_unlock(&ip->i_flags_lock);
-               return 1;
-       }
-       __xfs_iflags_set(ip, XFS_IRECLAIM);
-       spin_unlock(&ip->i_flags_lock);
-       return 0;
-}
-
-/*
- * Inodes in different states need to be treated differently. The following
- * table lists the inode states and the reclaim actions necessary:
- *
- *     inode state          iflush ret         required action
- *      ---------------      ----------         ---------------
- *     bad                     -               reclaim
- *     shutdown                EIO             unpin and reclaim
- *     clean, unpinned         0               reclaim
- *     stale, unpinned         0               reclaim
- *     clean, pinned(*)        0               requeue
- *     stale, pinned           EAGAIN          requeue
- *     dirty, async            -               requeue
- *     dirty, sync             0               reclaim
- *
- * (*) dgc: I don't think the clean, pinned state is possible but it gets
- * handled anyway given the order of checks implemented.
- *
- * 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.
- *
- * 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.
- *
- * Hence the order of actions after gaining the locks should be:
- *     bad             => reclaim
- *     shutdown        => unpin and reclaim
- *     pinned, async   => requeue
- *     pinned, sync    => unpin
- *     stale           => reclaim
- *     clean           => reclaim
- *     dirty, async    => requeue
- *     dirty, sync     => flush, wait and reclaim
- */
-STATIC int
-xfs_reclaim_inode(
-       struct xfs_inode        *ip,
-       struct xfs_perag        *pag,
-       int                     sync_mode)
-{
-       struct xfs_buf          *bp = NULL;
-       int                     error;
-
-restart:
-       error = 0;
-       xfs_ilock(ip, XFS_ILOCK_EXCL);
-       if (!xfs_iflock_nowait(ip)) {
-               if (!(sync_mode & SYNC_WAIT))
-                       goto out;
-               xfs_iflock(ip);
-       }
-
-       if (is_bad_inode(VFS_I(ip)))
-               goto reclaim;
-       if (XFS_FORCED_SHUTDOWN(ip->i_mount)) {
-               xfs_iunpin_wait(ip);
-               xfs_iflush_abort(ip, false);
-               goto reclaim;
-       }
-       if (xfs_ipincount(ip)) {
-               if (!(sync_mode & SYNC_WAIT))
-                       goto out_ifunlock;
-               xfs_iunpin_wait(ip);
-       }
-       if (xfs_iflags_test(ip, XFS_ISTALE))
-               goto reclaim;
-       if (xfs_inode_clean(ip))
-               goto reclaim;
-
-       /*
-        * 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;
-
-       /*
-        * Now we have an inode that needs flushing.
-        *
-        * Note that xfs_iflush will never block on the inode buffer lock, as
-        * xfs_ifree_cluster() can lock the inode buffer before it locks the
-        * ip->i_lock, and we are doing the exact opposite here.  As a result,
-        * doing a blocking xfs_imap_to_bp() to get the cluster buffer would
-        * result in an ABBA deadlock with xfs_ifree_cluster().
-        *
-        * 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
-        * 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.
-        */
-       error = xfs_iflush(ip, &bp);
-       if (error == EAGAIN) {
-               xfs_iunlock(ip, XFS_ILOCK_EXCL);
-               /* backoff longer than in xfs_ifree_cluster */
-               delay(2);
-               goto restart;
-       }
-
-       if (!error) {
-               error = xfs_bwrite(bp);
-               xfs_buf_relse(bp);
-       }
-
-       xfs_iflock(ip);
-reclaim:
-       xfs_ifunlock(ip);
-       xfs_iunlock(ip, XFS_ILOCK_EXCL);
-
-       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.
-        */
-       spin_lock(&pag->pag_ici_lock);
-       if (!radix_tree_delete(&pag->pag_ici_root,
-                               XFS_INO_TO_AGINO(ip->i_mount, ip->i_ino)))
-               ASSERT(0);
-       __xfs_inode_clear_reclaim(pag, ip);
-       spin_unlock(&pag->pag_ici_lock);
-
-       /*
-        * 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.
-        */
-       xfs_ilock(ip, XFS_ILOCK_EXCL);
-       xfs_qm_dqdetach(ip);
-       xfs_iunlock(ip, XFS_ILOCK_EXCL);
-
-       xfs_inode_free(ip);
-       return error;
-
-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 the reclaim work 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;
-}
-
-/*
- * 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;
-       int                     trylock = flags & SYNC_TRYLOCK;
-       int                     skipped;
-
-restart:
-       ag = 0;
-       skipped = 0;
-       while ((pag = xfs_perag_get_tag(mp, ag, XFS_ICI_RECLAIM_TAG))) {
-               unsigned long   first_index = 0;
-               int             done = 0;
-               int             nr_found = 0;
-
-               ag = pag->pag_agno + 1;
-
-               if (trylock) {
-                       if (!mutex_trylock(&pag->pag_ici_reclaim_lock)) {
-                               skipped++;
-                               xfs_perag_put(pag);
-                               continue;
-                       }
-                       first_index = pag->pag_ici_reclaim_cursor;
-               } else
-                       mutex_lock(&pag->pag_ici_reclaim_lock);
-
-               do {
-                       struct xfs_inode *batch[XFS_LOOKUP_BATCH];
-                       int     i;
-
-                       rcu_read_lock();
-                       nr_found = radix_tree_gang_lookup_tag(
-                                       &pag->pag_ici_root,
-                                       (void **)batch, first_index,
-                                       XFS_LOOKUP_BATCH,
-                                       XFS_ICI_RECLAIM_TAG);
-                       if (!nr_found) {
-                               done = 1;
-                               rcu_read_unlock();
-                               break;
-                       }
-
-                       /*
-                        * Grab the inodes before we drop the lock. if we found
-                        * nothing, nr == 0 and the loop will be skipped.
-                        */
-                       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.
-                                *
-                                * 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.
-                                */
-                               if (XFS_INO_TO_AGNO(mp, ip->i_ino) !=
-                                                               pag->pag_agno)
-                                       continue;
-                               first_index = XFS_INO_TO_AGINO(mp, ip->i_ino + 1);
-                               if (first_index < XFS_INO_TO_AGINO(mp, ip->i_ino))
-                                       done = 1;
-                       }
-
-                       /* unlock now we've grabbed the inodes. */
-                       rcu_read_unlock();
-
-                       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;
-
-                       cond_resched();
-
-               } while (nr_found && !done && *nr_to_scan > 0);
-
-               if (trylock && !done)
-                       pag->pag_ici_reclaim_cursor = first_index;
-               else
-                       pag->pag_ici_reclaim_cursor = 0;
-               mutex_unlock(&pag->pag_ici_reclaim_lock);
-               xfs_perag_put(pag);
-       }
-
-       /*
-        * 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 (skipped && (flags & SYNC_WAIT) && *nr_to_scan > 0) {
-               trylock = 0;
-               goto restart;
-       }
-       return XFS_ERROR(last_error);
-}
-
-int
-xfs_reclaim_inodes(
-       xfs_mount_t     *mp,
-       int             mode)
-{
-       int             nr_to_scan = INT_MAX;
-
-       return xfs_reclaim_inodes_ag(mp, mode, &nr_to_scan);
-}
-
-/*
- * Scan a certain number of inodes for reclaim.
- *
- * When called we make sure that there is a background (fast) inode reclaim in
- * progress, while we will throttle the speed of reclaim via doing synchronous
- * 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.
- */
-void
-xfs_reclaim_inodes_nr(
-       struct xfs_mount        *mp,
-       int                     nr_to_scan)
-{
-       /* kick background reclaimer and push the AIL */
-       xfs_reclaim_work_queue(mp);
-       xfs_ail_push_all(mp->m_ail);
-
-       xfs_reclaim_inodes_ag(mp, SYNC_TRYLOCK | SYNC_WAIT, &nr_to_scan);
-}
-
-/*
- * 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;
-
-       while ((pag = xfs_perag_get_tag(mp, ag, XFS_ICI_RECLAIM_TAG))) {
-               ag = pag->pag_agno + 1;
-               reclaimable += pag->pag_ici_reclaimable;
-               xfs_perag_put(pag);
-       }
-       return reclaimable;
-}
-
diff --git a/fs/xfs/xfs_sync.h b/fs/xfs/xfs_sync.h
deleted file mode 100644 (file)
index 0ba9c89..0000000
+++ /dev/null
@@ -1,43 +0,0 @@
-/*
- * Copyright (c) 2000-2006 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
- */
-#ifndef XFS_SYNC_H
-#define XFS_SYNC_H 1
-
-struct xfs_mount;
-struct xfs_perag;
-
-#define SYNC_WAIT              0x0001  /* wait for i/o to complete */
-#define SYNC_TRYLOCK           0x0002  /* only try to lock inodes */
-
-void xfs_reclaim_worker(struct work_struct *work);
-
-int xfs_reclaim_inodes(struct xfs_mount *mp, int mode);
-int xfs_reclaim_inodes_count(struct xfs_mount *mp);
-void xfs_reclaim_inodes_nr(struct xfs_mount *mp, int nr_to_scan);
-
-void xfs_inode_set_reclaim_tag(struct xfs_inode *ip);
-void __xfs_inode_set_reclaim_tag(struct xfs_perag *pag, struct xfs_inode *ip);
-void __xfs_inode_clear_reclaim_tag(struct xfs_mount *mp, struct xfs_perag *pag,
-                               struct xfs_inode *ip);
-
-int xfs_sync_inode_grab(struct xfs_inode *ip);
-int xfs_inode_ag_iterator(struct xfs_mount *mp,
-       int (*execute)(struct xfs_inode *ip, struct xfs_perag *pag, int flags),
-       int flags);
-
-#endif