summaryrefslogtreecommitdiffstats
path: root/fs/xfs
diff options
context:
space:
mode:
authorDave Chinner <dchinner@redhat.com>2020-05-20 13:17:11 -0700
committerDarrick J. Wong <darrick.wong@oracle.com>2020-05-27 08:49:25 -0700
commitf18c9a9030972d892a244968c653aceb98e27c70 (patch)
treef3f8282052855b69a97214dc85a7e1e16d0a0d48 /fs/xfs
parentdc3ffbb14060c943469d5e12900db3a60bc3fa64 (diff)
downloadlinux-f18c9a9030972d892a244968c653aceb98e27c70.tar.bz2
xfs: reduce free inode accounting overhead
Shaokun Zhang reported that XFS was using substantial CPU time in percpu_count_sum() when running a single threaded benchmark on a high CPU count (128p) machine from xfs_mod_ifree(). The issue is that the filesystem is empty when the benchmark runs, so inode allocation is running with a very low inode free count. With the percpu counter batching, this means comparisons when the counter is less that 128 * 256 = 32768 use the slow path of adding up all the counters across the CPUs, and this is expensive on high CPU count machines. The summing in xfs_mod_ifree() is only used to fire an assert if an underrun occurs. The error is ignored by the higher level code. Hence this is really just debug code and we don't need to run it on production kernels, nor do we need such debug checks to return error values just to trigger an assert. Finally, xfs_mod_icount/xfs_mod_ifree are only called from xfs_trans_unreserve_and_mod_sb(), so get rid of them and just directly call the percpu_counter_add/percpu_counter_compare functions. The compare functions are now run only on debug builds as they are internal to ASSERT() checks and so only compiled in when ASSERTs are active (CONFIG_XFS_DEBUG=y or CONFIG_XFS_WARN=y). Reported-by: Shaokun Zhang <zhangshaokun@hisilicon.com> Signed-off-by: Dave Chinner <dchinner@redhat.com> Reviewed-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com> Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Diffstat (limited to 'fs/xfs')
-rw-r--r--fs/xfs/xfs_mount.c33
-rw-r--r--fs/xfs/xfs_mount.h2
-rw-r--r--fs/xfs/xfs_trans.c17
3 files changed, 13 insertions, 39 deletions
diff --git a/fs/xfs/xfs_mount.c b/fs/xfs/xfs_mount.c
index bb91f04266b9..d5dcf9869860 100644
--- a/fs/xfs/xfs_mount.c
+++ b/fs/xfs/xfs_mount.c
@@ -1190,39 +1190,6 @@ xfs_log_sbcount(xfs_mount_t *mp)
}
/*
- * Deltas for the inode count are +/-64, hence we use a large batch size
- * of 128 so we don't need to take the counter lock on every update.
- */
-#define XFS_ICOUNT_BATCH 128
-int
-xfs_mod_icount(
- struct xfs_mount *mp,
- int64_t delta)
-{
- percpu_counter_add_batch(&mp->m_icount, delta, XFS_ICOUNT_BATCH);
- if (__percpu_counter_compare(&mp->m_icount, 0, XFS_ICOUNT_BATCH) < 0) {
- ASSERT(0);
- percpu_counter_add(&mp->m_icount, -delta);
- return -EINVAL;
- }
- return 0;
-}
-
-int
-xfs_mod_ifree(
- struct xfs_mount *mp,
- int64_t delta)
-{
- percpu_counter_add(&mp->m_ifree, delta);
- if (percpu_counter_compare(&mp->m_ifree, 0) < 0) {
- ASSERT(0);
- percpu_counter_add(&mp->m_ifree, -delta);
- return -EINVAL;
- }
- return 0;
-}
-
-/*
* Deltas for the block count can vary from 1 to very large, but lock contention
* only occurs on frequent small block count updates such as in the delayed
* allocation path for buffered writes (page a time updates). Hence we set
diff --git a/fs/xfs/xfs_mount.h b/fs/xfs/xfs_mount.h
index aba5a1579279..4835581f3eb0 100644
--- a/fs/xfs/xfs_mount.h
+++ b/fs/xfs/xfs_mount.h
@@ -392,8 +392,6 @@ extern int xfs_initialize_perag(xfs_mount_t *mp, xfs_agnumber_t agcount,
xfs_agnumber_t *maxagi);
extern void xfs_unmountfs(xfs_mount_t *);
-extern int xfs_mod_icount(struct xfs_mount *mp, int64_t delta);
-extern int xfs_mod_ifree(struct xfs_mount *mp, int64_t delta);
extern int xfs_mod_fdblocks(struct xfs_mount *mp, int64_t delta,
bool reserved);
extern int xfs_mod_frextents(struct xfs_mount *mp, int64_t delta);
diff --git a/fs/xfs/xfs_trans.c b/fs/xfs/xfs_trans.c
index e4e29135ad1b..2222a0ed3155 100644
--- a/fs/xfs/xfs_trans.c
+++ b/fs/xfs/xfs_trans.c
@@ -545,7 +545,12 @@ xfs_trans_apply_sb_deltas(
* used block counts are not updated in the on disk superblock. In this case,
* XFS_TRANS_SB_DIRTY will not be set when the transaction is updated but we
* still need to update the incore superblock with the changes.
+ *
+ * Deltas for the inode count are +/-64, hence we use a large batch size of 128
+ * so we don't need to take the counter lock on every update.
*/
+#define XFS_ICOUNT_BATCH 128
+
void
xfs_trans_unreserve_and_mod_sb(
struct xfs_trans *tp)
@@ -585,13 +590,17 @@ xfs_trans_unreserve_and_mod_sb(
}
if (idelta) {
- error = xfs_mod_icount(mp, idelta);
- ASSERT(!error);
+ percpu_counter_add_batch(&mp->m_icount, idelta,
+ XFS_ICOUNT_BATCH);
+ if (idelta < 0)
+ ASSERT(__percpu_counter_compare(&mp->m_icount, 0,
+ XFS_ICOUNT_BATCH) >= 0);
}
if (ifreedelta) {
- error = xfs_mod_ifree(mp, ifreedelta);
- ASSERT(!error);
+ percpu_counter_add(&mp->m_ifree, ifreedelta);
+ if (ifreedelta < 0)
+ ASSERT(percpu_counter_compare(&mp->m_ifree, 0) >= 0);
}
if (rtxdelta == 0 && !(tp->t_flags & XFS_TRANS_SB_DIRTY))