xfs: remove the per-filesystem list of dquots
Instead of keeping a separate per-filesystem list of dquots we can walk
the radix tree for the two places where we need to iterate all quota
structures.
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Ben Myers <bpm@sgi.com>
diff --git a/fs/xfs/xfs_dquot.c b/fs/xfs/xfs_dquot.c
index 49456e5..2896ac9 100644
--- a/fs/xfs/xfs_dquot.c
+++ b/fs/xfs/xfs_dquot.c
@@ -44,10 +44,9 @@
*
* ip->i_lock
* qi->qi_tree_lock
- * qi->qi_dqlist_lock
- * dquot->q_qlock (xfs_dqlock() and friends)
- * dquot->q_flush (xfs_dqflock() and friends)
- * qi->qi_lru_lock
+ * dquot->q_qlock (xfs_dqlock() and friends)
+ * dquot->q_flush (xfs_dqflock() and friends)
+ * qi->qi_lru_lock
*
* If two dquots need to be locked the order is user before group/project,
* otherwise by the lowest id first, see xfs_dqlock2.
@@ -729,20 +728,12 @@
}
/*
- * Attach this dquot to this filesystem's list of all dquots,
- * kept inside the mount structure in m_quotainfo field
- */
- mutex_lock(&qi->qi_dqlist_lock);
-
- /*
* We return a locked dquot to the caller, with a reference taken
*/
xfs_dqlock(dqp);
dqp->q_nrefs = 1;
- list_add(&dqp->q_mplist, &qi->qi_dqlist);
qi->qi_dquots++;
- mutex_unlock(&qi->qi_dqlist_lock);
mutex_unlock(&qi->qi_tree_lock);
dqret:
@@ -1018,86 +1009,6 @@
}
/*
- * Take a dquot out of the mount's dqlist as well as the hashlist. This is
- * called via unmount as well as quotaoff, and the purge will always succeed.
- */
-void
-xfs_qm_dqpurge(
- struct xfs_dquot *dqp)
-{
- struct xfs_mount *mp = dqp->q_mount;
- struct xfs_quotainfo *qi = mp->m_quotainfo;
-
- xfs_dqlock(dqp);
-
- /*
- * If we're turning off quotas, we have to make sure that, for
- * example, we don't delete quota disk blocks while dquots are
- * in the process of getting written to those disk blocks.
- * This dquot might well be on AIL, and we can't leave it there
- * if we're turning off quotas. Basically, we need this flush
- * lock, and are willing to block on it.
- */
- if (!xfs_dqflock_nowait(dqp)) {
- /*
- * Block on the flush lock after nudging dquot buffer,
- * if it is incore.
- */
- xfs_dqflock_pushbuf_wait(dqp);
- }
-
- /*
- * If we are turning this type of quotas off, we don't care
- * about the dirty metadata sitting in this dquot. OTOH, if
- * we're unmounting, we do care, so we flush it and wait.
- */
- if (XFS_DQ_IS_DIRTY(dqp)) {
- int error;
-
- /*
- * We don't care about getting disk errors here. We need
- * to purge this dquot anyway, so we go ahead regardless.
- */
- error = xfs_qm_dqflush(dqp, SYNC_WAIT);
- if (error)
- xfs_warn(mp, "%s: dquot %p flush failed",
- __func__, dqp);
- xfs_dqflock(dqp);
- }
-
- ASSERT(atomic_read(&dqp->q_pincount) == 0);
- ASSERT(XFS_FORCED_SHUTDOWN(mp) ||
- !(dqp->q_logitem.qli_item.li_flags & XFS_LI_IN_AIL));
-
- xfs_dqfunlock(dqp);
- xfs_dqunlock(dqp);
-
- mutex_lock(&qi->qi_tree_lock);
- radix_tree_delete(XFS_DQUOT_TREE(qi, dqp->q_core.d_flags),
- be32_to_cpu(dqp->q_core.d_id));
- mutex_unlock(&qi->qi_tree_lock);
-
- mutex_lock(&qi->qi_dqlist_lock);
- list_del_init(&dqp->q_mplist);
- qi->qi_dqreclaims++;
- qi->qi_dquots--;
- mutex_unlock(&qi->qi_dqlist_lock);
-
- /*
- * We move dquots to the freelist as soon as their reference count
- * hits zero, so it really should be on the freelist here.
- */
- mutex_lock(&qi->qi_lru_lock);
- ASSERT(!list_empty(&dqp->q_lru));
- list_del_init(&dqp->q_lru);
- qi->qi_lru_count--;
- XFS_STATS_DEC(xs_qm_dquot_unused);
- mutex_unlock(&qi->qi_lru_lock);
-
- xfs_qm_dqdestroy(dqp);
-}
-
-/*
* Give the buffer a little push if it is incore and
* wait on the flush lock.
*/
diff --git a/fs/xfs/xfs_dquot.h b/fs/xfs/xfs_dquot.h
index 4061f17..ef9190b 100644
--- a/fs/xfs/xfs_dquot.h
+++ b/fs/xfs/xfs_dquot.h
@@ -38,7 +38,6 @@
typedef struct xfs_dquot {
uint dq_flags; /* various flags (XFS_DQ_*) */
struct list_head q_lru; /* global free list of dquots */
- struct list_head q_mplist; /* mount's list of dquots */
struct xfs_mount*q_mount; /* filesystem this relates to */
struct xfs_trans*q_transp; /* trans this belongs to currently */
uint q_nrefs; /* # active refs from inodes */
@@ -143,7 +142,6 @@
uint, struct xfs_dquot **);
extern void xfs_qm_dqdestroy(xfs_dquot_t *);
extern int xfs_qm_dqflush(xfs_dquot_t *, uint);
-extern void xfs_qm_dqpurge(xfs_dquot_t *);
extern void xfs_qm_dqunpin_wait(xfs_dquot_t *);
extern void xfs_qm_adjust_dqtimers(xfs_mount_t *,
xfs_disk_dquot_t *);
diff --git a/fs/xfs/xfs_qm.c b/fs/xfs/xfs_qm.c
index bb884e7..2f92d3b 100644
--- a/fs/xfs/xfs_qm.c
+++ b/fs/xfs/xfs_qm.c
@@ -169,6 +169,187 @@
}
/*
+ * We use the batch lookup interface to iterate over the dquots as it
+ * currently is the only interface into the radix tree code that allows
+ * fuzzy lookups instead of exact matches. Holding the lock over multiple
+ * operations is fine as all callers are used either during mount/umount
+ * or quotaoff.
+ */
+#define XFS_DQ_LOOKUP_BATCH 32
+
+STATIC int
+xfs_qm_dquot_walk(
+ struct xfs_mount *mp,
+ int type,
+ int (*execute)(struct xfs_dquot *dqp))
+{
+ struct xfs_quotainfo *qi = mp->m_quotainfo;
+ struct radix_tree_root *tree = XFS_DQUOT_TREE(qi, type);
+ uint32_t next_index;
+ int last_error = 0;
+ int skipped;
+ int nr_found;
+
+restart:
+ skipped = 0;
+ next_index = 0;
+ nr_found = 0;
+
+ while (1) {
+ struct xfs_dquot *batch[XFS_DQ_LOOKUP_BATCH];
+ int error = 0;
+ int i;
+
+ mutex_lock(&qi->qi_tree_lock);
+ nr_found = radix_tree_gang_lookup(tree, (void **)batch,
+ next_index, XFS_DQ_LOOKUP_BATCH);
+ if (!nr_found) {
+ mutex_unlock(&qi->qi_tree_lock);
+ break;
+ }
+
+ for (i = 0; i < nr_found; i++) {
+ struct xfs_dquot *dqp = batch[i];
+
+ next_index = be32_to_cpu(dqp->q_core.d_id) + 1;
+
+ error = execute(batch[i]);
+ if (error == EAGAIN) {
+ skipped++;
+ continue;
+ }
+ if (error && last_error != EFSCORRUPTED)
+ last_error = error;
+ }
+
+ mutex_unlock(&qi->qi_tree_lock);
+
+ /* bail out if the filesystem is corrupted. */
+ if (last_error == EFSCORRUPTED) {
+ skipped = 0;
+ break;
+ }
+ }
+
+ if (skipped) {
+ delay(1);
+ goto restart;
+ }
+
+ return last_error;
+}
+
+
+/*
+ * Purge a dquot from all tracking data structures and free it.
+ */
+STATIC int
+xfs_qm_dqpurge(
+ struct xfs_dquot *dqp)
+{
+ struct xfs_mount *mp = dqp->q_mount;
+ struct xfs_quotainfo *qi = mp->m_quotainfo;
+ struct xfs_dquot *gdqp = NULL;
+
+ xfs_dqlock(dqp);
+ if ((dqp->dq_flags & XFS_DQ_FREEING) || dqp->q_nrefs != 0) {
+ xfs_dqunlock(dqp);
+ return EAGAIN;
+ }
+
+ /*
+ * If this quota has a group hint attached, prepare for releasing it
+ * now.
+ */
+ gdqp = dqp->q_gdquot;
+ if (gdqp) {
+ xfs_dqlock(gdqp);
+ dqp->q_gdquot = NULL;
+ }
+
+ dqp->dq_flags |= XFS_DQ_FREEING;
+
+ /*
+ * If we're turning off quotas, we have to make sure that, for
+ * example, we don't delete quota disk blocks while dquots are
+ * in the process of getting written to those disk blocks.
+ * This dquot might well be on AIL, and we can't leave it there
+ * if we're turning off quotas. Basically, we need this flush
+ * lock, and are willing to block on it.
+ */
+ if (!xfs_dqflock_nowait(dqp)) {
+ /*
+ * Block on the flush lock after nudging dquot buffer,
+ * if it is incore.
+ */
+ xfs_dqflock_pushbuf_wait(dqp);
+ }
+
+ /*
+ * If we are turning this type of quotas off, we don't care
+ * about the dirty metadata sitting in this dquot. OTOH, if
+ * we're unmounting, we do care, so we flush it and wait.
+ */
+ if (XFS_DQ_IS_DIRTY(dqp)) {
+ int error;
+
+ /*
+ * We don't care about getting disk errors here. We need
+ * to purge this dquot anyway, so we go ahead regardless.
+ */
+ error = xfs_qm_dqflush(dqp, SYNC_WAIT);
+ if (error)
+ xfs_warn(mp, "%s: dquot %p flush failed",
+ __func__, dqp);
+ xfs_dqflock(dqp);
+ }
+
+ ASSERT(atomic_read(&dqp->q_pincount) == 0);
+ ASSERT(XFS_FORCED_SHUTDOWN(mp) ||
+ !(dqp->q_logitem.qli_item.li_flags & XFS_LI_IN_AIL));
+
+ xfs_dqfunlock(dqp);
+ xfs_dqunlock(dqp);
+
+ radix_tree_delete(XFS_DQUOT_TREE(qi, dqp->q_core.d_flags),
+ be32_to_cpu(dqp->q_core.d_id));
+ qi->qi_dquots--;
+
+ /*
+ * We move dquots to the freelist as soon as their reference count
+ * hits zero, so it really should be on the freelist here.
+ */
+ mutex_lock(&qi->qi_lru_lock);
+ ASSERT(!list_empty(&dqp->q_lru));
+ list_del_init(&dqp->q_lru);
+ qi->qi_lru_count--;
+ XFS_STATS_DEC(xs_qm_dquot_unused);
+ mutex_unlock(&qi->qi_lru_lock);
+
+ xfs_qm_dqdestroy(dqp);
+
+ if (gdqp)
+ xfs_qm_dqput(gdqp);
+ return 0;
+}
+
+/*
+ * Purge the dquot cache.
+ */
+void
+xfs_qm_dqpurge_all(
+ struct xfs_mount *mp,
+ uint flags)
+{
+ if (flags & XFS_QMOPT_UQUOTA)
+ xfs_qm_dquot_walk(mp, XFS_DQ_USER, xfs_qm_dqpurge);
+ if (flags & XFS_QMOPT_GQUOTA)
+ xfs_qm_dquot_walk(mp, XFS_DQ_GROUP, xfs_qm_dqpurge);
+ if (flags & XFS_QMOPT_PQUOTA)
+ xfs_qm_dquot_walk(mp, XFS_DQ_PROJ, xfs_qm_dqpurge);
+}
+
+/*
* Just destroy the quotainfo structure.
*/
void
@@ -306,175 +487,6 @@
}
}
-/*
- * Flush all dquots of the given file system to disk. The dquots are
- * _not_ purged from memory here, just their data written to disk.
- */
-STATIC int
-xfs_qm_dqflush_all(
- struct xfs_mount *mp)
-{
- struct xfs_quotainfo *q = mp->m_quotainfo;
- int recl;
- struct xfs_dquot *dqp;
- int error;
-
- if (!q)
- return 0;
-again:
- mutex_lock(&q->qi_dqlist_lock);
- list_for_each_entry(dqp, &q->qi_dqlist, q_mplist) {
- xfs_dqlock(dqp);
- if ((dqp->dq_flags & XFS_DQ_FREEING) ||
- !XFS_DQ_IS_DIRTY(dqp)) {
- xfs_dqunlock(dqp);
- continue;
- }
-
- /* XXX a sentinel would be better */
- recl = q->qi_dqreclaims;
- if (!xfs_dqflock_nowait(dqp)) {
- /*
- * If we can't grab the flush lock then check
- * to see if the dquot has been flushed delayed
- * write. If so, grab its buffer and send it
- * out immediately. We'll be able to acquire
- * the flush lock when the I/O completes.
- */
- xfs_dqflock_pushbuf_wait(dqp);
- }
- /*
- * Let go of the mplist lock. We don't want to hold it
- * across a disk write.
- */
- mutex_unlock(&q->qi_dqlist_lock);
- error = xfs_qm_dqflush(dqp, 0);
- xfs_dqunlock(dqp);
- if (error)
- return error;
-
- mutex_lock(&q->qi_dqlist_lock);
- if (recl != q->qi_dqreclaims) {
- mutex_unlock(&q->qi_dqlist_lock);
- /* XXX restart limit */
- goto again;
- }
- }
-
- mutex_unlock(&q->qi_dqlist_lock);
- /* return ! busy */
- return 0;
-}
-
-/*
- * Release the group dquot pointers the user dquots may be
- * carrying around as a hint. mplist is locked on entry and exit.
- */
-STATIC void
-xfs_qm_detach_gdquots(
- struct xfs_mount *mp)
-{
- struct xfs_quotainfo *q = mp->m_quotainfo;
- struct xfs_dquot *dqp, *gdqp;
-
- again:
- ASSERT(mutex_is_locked(&q->qi_dqlist_lock));
- list_for_each_entry(dqp, &q->qi_dqlist, q_mplist) {
- xfs_dqlock(dqp);
- if (dqp->dq_flags & XFS_DQ_FREEING) {
- xfs_dqunlock(dqp);
- mutex_unlock(&q->qi_dqlist_lock);
- delay(1);
- mutex_lock(&q->qi_dqlist_lock);
- goto again;
- }
-
- gdqp = dqp->q_gdquot;
- if (gdqp)
- dqp->q_gdquot = NULL;
- xfs_dqunlock(dqp);
-
- if (gdqp)
- xfs_qm_dqrele(gdqp);
- }
-}
-
-/*
- * Go through all the incore dquots of this file system and take them
- * off the mplist and hashlist, if the dquot type matches the dqtype
- * parameter. This is used when turning off quota accounting for
- * users and/or groups, as well as when the filesystem is unmounting.
- */
-STATIC int
-xfs_qm_dqpurge_int(
- struct xfs_mount *mp,
- uint flags)
-{
- struct xfs_quotainfo *q = mp->m_quotainfo;
- struct xfs_dquot *dqp, *n;
- uint dqtype;
- int nmisses = 0;
- LIST_HEAD (dispose_list);
-
- if (!q)
- return 0;
-
- dqtype = (flags & XFS_QMOPT_UQUOTA) ? XFS_DQ_USER : 0;
- dqtype |= (flags & XFS_QMOPT_PQUOTA) ? XFS_DQ_PROJ : 0;
- dqtype |= (flags & XFS_QMOPT_GQUOTA) ? XFS_DQ_GROUP : 0;
-
- mutex_lock(&q->qi_dqlist_lock);
-
- /*
- * In the first pass through all incore dquots of this filesystem,
- * we release the group dquot pointers the user dquots may be
- * carrying around as a hint. We need to do this irrespective of
- * what's being turned off.
- */
- xfs_qm_detach_gdquots(mp);
-
- /*
- * Try to get rid of all of the unwanted dquots.
- */
- list_for_each_entry_safe(dqp, n, &q->qi_dqlist, q_mplist) {
- xfs_dqlock(dqp);
- if ((dqp->dq_flags & dqtype) != 0 &&
- !(dqp->dq_flags & XFS_DQ_FREEING)) {
- if (dqp->q_nrefs == 0) {
- dqp->dq_flags |= XFS_DQ_FREEING;
- list_move_tail(&dqp->q_mplist, &dispose_list);
- } else
- nmisses++;
- }
- xfs_dqunlock(dqp);
- }
- mutex_unlock(&q->qi_dqlist_lock);
-
- list_for_each_entry_safe(dqp, n, &dispose_list, q_mplist)
- xfs_qm_dqpurge(dqp);
-
- return nmisses;
-}
-
-int
-xfs_qm_dqpurge_all(
- xfs_mount_t *mp,
- uint flags)
-{
- int ndquots;
-
- /*
- * Purge the dquot cache.
- * None of the dquots should really be busy at this point.
- */
- if (mp->m_quotainfo) {
- while ((ndquots = xfs_qm_dqpurge_int(mp, flags))) {
- delay(ndquots * 10);
- }
- }
- return 0;
-}
-
STATIC int
xfs_qm_dqattach_one(
xfs_inode_t *ip,
@@ -749,15 +761,10 @@
INIT_RADIX_TREE(&qinf->qi_gquota_tree, GFP_NOFS);
mutex_init(&qinf->qi_tree_lock);
- INIT_LIST_HEAD(&qinf->qi_dqlist);
- mutex_init(&qinf->qi_dqlist_lock);
-
INIT_LIST_HEAD(&qinf->qi_lru_list);
qinf->qi_lru_count = 0;
mutex_init(&qinf->qi_lru_lock);
- qinf->qi_dqreclaims = 0;
-
/* mutex used to serialize quotaoffs */
mutex_init(&qinf->qi_quotaofflock);
@@ -854,9 +861,6 @@
*/
xfs_qm_rele_quotafs_ref(mp);
- ASSERT(list_empty(&qi->qi_dqlist));
- mutex_destroy(&qi->qi_dqlist_lock);
-
if (qi->qi_uquotaip) {
IRELE(qi->qi_uquotaip);
qi->qi_uquotaip = NULL; /* paranoia */
@@ -1307,6 +1311,28 @@
return error;
}
+STATIC int
+xfs_qm_flush_one(
+ struct xfs_dquot *dqp)
+{
+ int error = 0;
+
+ xfs_dqlock(dqp);
+ if (dqp->dq_flags & XFS_DQ_FREEING)
+ goto out_unlock;
+ if (!XFS_DQ_IS_DIRTY(dqp))
+ goto out_unlock;
+
+ if (!xfs_dqflock_nowait(dqp))
+ xfs_dqflock_pushbuf_wait(dqp);
+
+ error = xfs_qm_dqflush(dqp, 0);
+
+out_unlock:
+ xfs_dqunlock(dqp);
+ return error;
+}
+
/*
* Walk thru all the filesystem inodes and construct a consistent view
* of the disk quota world. If the quotacheck fails, disable quotas.
@@ -1315,7 +1341,7 @@
xfs_qm_quotacheck(
xfs_mount_t *mp)
{
- int done, count, error;
+ int done, count, error, error2;
xfs_ino_t lastino;
size_t structsz;
xfs_inode_t *uip, *gip;
@@ -1329,12 +1355,6 @@
ASSERT(mp->m_quotainfo->qi_uquotaip || mp->m_quotainfo->qi_gquotaip);
ASSERT(XFS_IS_QUOTA_RUNNING(mp));
- /*
- * There should be no cached dquots. The (simplistic) quotacheck
- * algorithm doesn't like that.
- */
- ASSERT(list_empty(&mp->m_quotainfo->qi_dqlist));
-
xfs_notice(mp, "Quotacheck needed: Please wait.");
/*
@@ -1373,12 +1393,21 @@
} while (!done);
/*
- * We've made all the changes that we need to make incore.
- * Flush them down to disk buffers if everything was updated
- * successfully.
+ * We've made all the changes that we need to make incore. Flush them
+ * down to disk buffers if everything was updated successfully.
*/
- if (!error)
- error = xfs_qm_dqflush_all(mp);
+ if (XFS_IS_UQUOTA_ON(mp))
+ error = xfs_qm_dquot_walk(mp, XFS_DQ_USER, xfs_qm_flush_one);
+ if (XFS_IS_GQUOTA_ON(mp)) {
+ error2 = xfs_qm_dquot_walk(mp, XFS_DQ_GROUP, xfs_qm_flush_one);
+ if (!error)
+ error = error2;
+ }
+ if (XFS_IS_PQUOTA_ON(mp)) {
+ error2 = xfs_qm_dquot_walk(mp, XFS_DQ_PROJ, xfs_qm_flush_one);
+ if (!error)
+ error = error2;
+ }
/*
* We can get this error if we couldn't do a dquot allocation inside
@@ -1517,13 +1546,9 @@
mutex_lock(&qi->qi_tree_lock);
radix_tree_delete(XFS_DQUOT_TREE(qi, dqp->q_core.d_flags),
be32_to_cpu(dqp->q_core.d_id));
- mutex_unlock(&qi->qi_tree_lock);
- mutex_lock(&qi->qi_dqlist_lock);
- list_del_init(&dqp->q_mplist);
qi->qi_dquots--;
- qi->qi_dqreclaims++;
- mutex_unlock(&qi->qi_dqlist_lock);
+ mutex_unlock(&qi->qi_tree_lock);
xfs_qm_dqdestroy(dqp);
}
@@ -1556,8 +1581,6 @@
return;
}
- ASSERT(!list_empty(&dqp->q_mplist));
-
/*
* Try to grab the flush lock. If this dquot is in the process of
* getting flushed to disk, we don't want to reclaim it.
diff --git a/fs/xfs/xfs_qm.h b/fs/xfs/xfs_qm.h
index 8f4b117..7644706 100644
--- a/fs/xfs/xfs_qm.h
+++ b/fs/xfs/xfs_qm.h
@@ -63,11 +63,7 @@
struct list_head qi_lru_list;
struct mutex qi_lru_lock;
int qi_lru_count;
- struct list_head qi_dqlist; /* all dquots in filesys */
- struct mutex qi_dqlist_lock;
int qi_dquots;
- int qi_dqreclaims; /* a change here indicates
- a removal in the dqlist */
time_t qi_btimelimit; /* limit for blks timer */
time_t qi_itimelimit; /* limit for inodes timer */
time_t qi_rtbtimelimit;/* limit for rt blks timer */
@@ -126,7 +122,7 @@
extern int xfs_qm_write_sb_changes(xfs_mount_t *, __int64_t);
/* dquot stuff */
-extern int xfs_qm_dqpurge_all(xfs_mount_t *, uint);
+extern void xfs_qm_dqpurge_all(xfs_mount_t *, uint);
extern void xfs_qm_dqrele_all_inodes(xfs_mount_t *, uint);
/* quota ops */
diff --git a/fs/xfs/xfs_qm_syscalls.c b/fs/xfs/xfs_qm_syscalls.c
index b9ac268..c4f396e 100644
--- a/fs/xfs/xfs_qm_syscalls.c
+++ b/fs/xfs/xfs_qm_syscalls.c
@@ -66,7 +66,6 @@
int error;
uint inactivate_flags;
xfs_qoff_logitem_t *qoffstart;
- int nculprits;
/*
* No file system can have quotas enabled on disk but not in core.
@@ -172,18 +171,13 @@
* This isn't protected by a particular lock directly, because we
* don't want to take a mrlock every time we depend on quotas being on.
*/
- mp->m_qflags &= ~(flags);
+ mp->m_qflags &= ~flags;
/*
* Go through all the dquots of this file system and purge them,
- * according to what was turned off. We may not be able to get rid
- * of all dquots, because dquots can have temporary references that
- * are not attached to inodes. eg. xfs_setattr, xfs_create.
- * So, if we couldn't purge all the dquots from the filesystem,
- * we can't get rid of the incore data structures.
+ * according to what was turned off.
*/
- while ((nculprits = xfs_qm_dqpurge_all(mp, dqtype)))
- delay(10 * nculprits);
+ xfs_qm_dqpurge_all(mp, dqtype);
/*
* Transactions that had started before ACTIVE state bit was cleared
