From 501ab32387533924b211cacff36d19296414ec0b Mon Sep 17 00:00:00 2001 From: Dave Chinner Date: Mon, 23 Feb 2015 21:19:28 +1100 Subject: xfs: use generic percpu counters for inode counter XFS has hand-rolled per-cpu counters for the superblock since before there was any generic implementation. There are some warts around the use of them for the inode counter as the hand rolled counter is designed to be accurate at zero, but has no specific accurracy at any other value. This design causes problems for the maximum inode count threshold enforcement, as there is no trigger that balances the counters as they get close tothe maximum threshold. Instead of designing new triggers for balancing, just replace the handrolled per-cpu counter with a generic counter. This enables us to update the counter through the normal superblock modification funtions, but rather than do that we add a xfs_mod_icount() helper function (from Christoph Hellwig) and keep the percpu counter outside the superblock in the struct xfs_mount. This means we still need to initialise the per-cpu counter specifically when we read the superblock, and vice versa when we log/write it, but it does mean that we don't need to change any other code. Signed-off-by: Dave Chinner Reviewed-by: Brian Foster Signed-off-by: Dave Chinner --- fs/xfs/libxfs/xfs_ialloc.c | 6 ++-- fs/xfs/libxfs/xfs_sb.c | 2 ++ fs/xfs/xfs_fsops.c | 3 +- fs/xfs/xfs_mount.c | 76 +++++++++++++++++++++------------------------- fs/xfs/xfs_mount.h | 7 +++-- fs/xfs/xfs_super.c | 7 +++-- fs/xfs/xfs_trans.c | 5 ++- 7 files changed, 54 insertions(+), 52 deletions(-) diff --git a/fs/xfs/libxfs/xfs_ialloc.c b/fs/xfs/libxfs/xfs_ialloc.c index 116ef1ddb3e3..5b4ba9f6b37d 100644 --- a/fs/xfs/libxfs/xfs_ialloc.c +++ b/fs/xfs/libxfs/xfs_ialloc.c @@ -376,7 +376,8 @@ xfs_ialloc_ag_alloc( */ newlen = args.mp->m_ialloc_inos; if (args.mp->m_maxicount && - args.mp->m_sb.sb_icount + newlen > args.mp->m_maxicount) + percpu_counter_read(&args.mp->m_icount) + newlen > + args.mp->m_maxicount) return -ENOSPC; args.minlen = args.maxlen = args.mp->m_ialloc_blks; /* @@ -1340,7 +1341,8 @@ xfs_dialloc( * inode. */ if (mp->m_maxicount && - mp->m_sb.sb_icount + mp->m_ialloc_inos > mp->m_maxicount) { + percpu_counter_read(&mp->m_icount) + mp->m_ialloc_inos > + mp->m_maxicount) { noroom = 1; okalloc = 0; } diff --git a/fs/xfs/libxfs/xfs_sb.c b/fs/xfs/libxfs/xfs_sb.c index b0a5fe95a3e2..017cb2fc53eb 100644 --- a/fs/xfs/libxfs/xfs_sb.c +++ b/fs/xfs/libxfs/xfs_sb.c @@ -771,6 +771,8 @@ xfs_log_sb( struct xfs_mount *mp = tp->t_mountp; struct xfs_buf *bp = xfs_trans_getsb(tp, mp, 0); + mp->m_sb.sb_icount = percpu_counter_sum(&mp->m_icount); + xfs_sb_to_disk(XFS_BUF_TO_SBP(bp), &mp->m_sb); xfs_trans_buf_set_type(tp, bp, XFS_BLFT_SB_BUF); xfs_trans_log_buf(tp, bp, 0, sizeof(struct xfs_dsb)); diff --git a/fs/xfs/xfs_fsops.c b/fs/xfs/xfs_fsops.c index 74efe5b760dc..b87a6f92263b 100644 --- a/fs/xfs/xfs_fsops.c +++ b/fs/xfs/xfs_fsops.c @@ -638,11 +638,12 @@ xfs_fs_counts( xfs_fsop_counts_t *cnt) { xfs_icsb_sync_counters(mp, XFS_ICSB_LAZY_COUNT); + cnt->allocino = percpu_counter_read_positive(&mp->m_icount); + spin_lock(&mp->m_sb_lock); cnt->freedata = mp->m_sb.sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp); cnt->freertx = mp->m_sb.sb_frextents; cnt->freeino = mp->m_sb.sb_ifree; - cnt->allocino = mp->m_sb.sb_icount; spin_unlock(&mp->m_sb_lock); return 0; } diff --git a/fs/xfs/xfs_mount.c b/fs/xfs/xfs_mount.c index 4fa80e63eea2..702ea6a7e648 100644 --- a/fs/xfs/xfs_mount.c +++ b/fs/xfs/xfs_mount.c @@ -1099,6 +1099,21 @@ xfs_log_sbcount(xfs_mount_t *mp) return xfs_sync_sb(mp, true); } +int +xfs_mod_icount( + struct xfs_mount *mp, + int64_t delta) +{ + /* deltas are +/-64, hence the large batch size of 128. */ + __percpu_counter_add(&mp->m_icount, delta, 128); + if (percpu_counter_compare(&mp->m_icount, 0) < 0) { + ASSERT(0); + percpu_counter_add(&mp->m_icount, -delta); + return -EINVAL; + } + return 0; +} + /* * xfs_mod_incore_sb_unlocked() is a utility routine commonly used to apply * a delta to a specified field in the in-core superblock. Simply @@ -1127,14 +1142,8 @@ xfs_mod_incore_sb_unlocked( */ switch (field) { case XFS_SBS_ICOUNT: - lcounter = (long long)mp->m_sb.sb_icount; - lcounter += delta; - if (lcounter < 0) { - ASSERT(0); - return -EINVAL; - } - mp->m_sb.sb_icount = lcounter; - return 0; + ASSERT(0); + return -ENOSPC; case XFS_SBS_IFREE: lcounter = (long long)mp->m_sb.sb_ifree; lcounter += delta; @@ -1288,8 +1297,9 @@ xfs_mod_incore_sb( int status; #ifdef HAVE_PERCPU_SB - ASSERT(field < XFS_SBS_ICOUNT || field > XFS_SBS_FDBLOCKS); + ASSERT(field < XFS_SBS_IFREE || field > XFS_SBS_FDBLOCKS); #endif + spin_lock(&mp->m_sb_lock); status = xfs_mod_incore_sb_unlocked(mp, field, delta, rsvd); spin_unlock(&mp->m_sb_lock); @@ -1492,7 +1502,6 @@ xfs_icsb_cpu_notify( case CPU_ONLINE: case CPU_ONLINE_FROZEN: xfs_icsb_lock(mp); - xfs_icsb_balance_counter(mp, XFS_SBS_ICOUNT, 0); xfs_icsb_balance_counter(mp, XFS_SBS_IFREE, 0); xfs_icsb_balance_counter(mp, XFS_SBS_FDBLOCKS, 0); xfs_icsb_unlock(mp); @@ -1504,17 +1513,14 @@ xfs_icsb_cpu_notify( * re-enable the counters. */ xfs_icsb_lock(mp); spin_lock(&mp->m_sb_lock); - xfs_icsb_disable_counter(mp, XFS_SBS_ICOUNT); xfs_icsb_disable_counter(mp, XFS_SBS_IFREE); xfs_icsb_disable_counter(mp, XFS_SBS_FDBLOCKS); - mp->m_sb.sb_icount += cntp->icsb_icount; mp->m_sb.sb_ifree += cntp->icsb_ifree; mp->m_sb.sb_fdblocks += cntp->icsb_fdblocks; memset(cntp, 0, sizeof(xfs_icsb_cnts_t)); - xfs_icsb_balance_counter_locked(mp, XFS_SBS_ICOUNT, 0); xfs_icsb_balance_counter_locked(mp, XFS_SBS_IFREE, 0); xfs_icsb_balance_counter_locked(mp, XFS_SBS_FDBLOCKS, 0); spin_unlock(&mp->m_sb_lock); @@ -1531,11 +1537,18 @@ xfs_icsb_init_counters( xfs_mount_t *mp) { xfs_icsb_cnts_t *cntp; + int error; int i; + error = percpu_counter_init(&mp->m_icount, 0, GFP_KERNEL); + if (error) + return error; + mp->m_sb_cnts = alloc_percpu(xfs_icsb_cnts_t); - if (mp->m_sb_cnts == NULL) + if (!mp->m_sb_cnts) { + percpu_counter_destroy(&mp->m_icount); return -ENOMEM; + } for_each_online_cpu(i) { cntp = (xfs_icsb_cnts_t *)per_cpu_ptr(mp->m_sb_cnts, i); @@ -1563,13 +1576,14 @@ void xfs_icsb_reinit_counters( xfs_mount_t *mp) { + percpu_counter_set(&mp->m_icount, mp->m_sb.sb_icount); + xfs_icsb_lock(mp); /* * start with all counters disabled so that the * initial balance kicks us off correctly */ mp->m_icsb_counters = -1; - xfs_icsb_balance_counter(mp, XFS_SBS_ICOUNT, 0); xfs_icsb_balance_counter(mp, XFS_SBS_IFREE, 0); xfs_icsb_balance_counter(mp, XFS_SBS_FDBLOCKS, 0); xfs_icsb_unlock(mp); @@ -1583,6 +1597,9 @@ xfs_icsb_destroy_counters( unregister_hotcpu_notifier(&mp->m_icsb_notifier); free_percpu(mp->m_sb_cnts); } + + percpu_counter_destroy(&mp->m_icount); + mutex_destroy(&mp->m_icsb_mutex); } @@ -1645,7 +1662,6 @@ xfs_icsb_count( for_each_online_cpu(i) { cntp = (xfs_icsb_cnts_t *)per_cpu_ptr(mp->m_sb_cnts, i); - cnt->icsb_icount += cntp->icsb_icount; cnt->icsb_ifree += cntp->icsb_ifree; cnt->icsb_fdblocks += cntp->icsb_fdblocks; } @@ -1659,7 +1675,7 @@ xfs_icsb_counter_disabled( xfs_mount_t *mp, xfs_sb_field_t field) { - ASSERT((field >= XFS_SBS_ICOUNT) && (field <= XFS_SBS_FDBLOCKS)); + ASSERT((field >= XFS_SBS_IFREE) && (field <= XFS_SBS_FDBLOCKS)); return test_bit(field, &mp->m_icsb_counters); } @@ -1670,7 +1686,7 @@ xfs_icsb_disable_counter( { xfs_icsb_cnts_t cnt; - ASSERT((field >= XFS_SBS_ICOUNT) && (field <= XFS_SBS_FDBLOCKS)); + ASSERT((field >= XFS_SBS_IFREE) && (field <= XFS_SBS_FDBLOCKS)); /* * If we are already disabled, then there is nothing to do @@ -1689,9 +1705,6 @@ xfs_icsb_disable_counter( xfs_icsb_count(mp, &cnt, XFS_ICSB_LAZY_COUNT); switch(field) { - case XFS_SBS_ICOUNT: - mp->m_sb.sb_icount = cnt.icsb_icount; - break; case XFS_SBS_IFREE: mp->m_sb.sb_ifree = cnt.icsb_ifree; break; @@ -1716,15 +1729,12 @@ xfs_icsb_enable_counter( xfs_icsb_cnts_t *cntp; int i; - ASSERT((field >= XFS_SBS_ICOUNT) && (field <= XFS_SBS_FDBLOCKS)); + ASSERT((field >= XFS_SBS_IFREE) && (field <= XFS_SBS_FDBLOCKS)); xfs_icsb_lock_all_counters(mp); for_each_online_cpu(i) { cntp = per_cpu_ptr(mp->m_sb_cnts, i); switch (field) { - case XFS_SBS_ICOUNT: - cntp->icsb_icount = count + resid; - break; case XFS_SBS_IFREE: cntp->icsb_ifree = count + resid; break; @@ -1750,8 +1760,6 @@ xfs_icsb_sync_counters_locked( xfs_icsb_count(mp, &cnt, flags); - if (!xfs_icsb_counter_disabled(mp, XFS_SBS_ICOUNT)) - mp->m_sb.sb_icount = cnt.icsb_icount; if (!xfs_icsb_counter_disabled(mp, XFS_SBS_IFREE)) mp->m_sb.sb_ifree = cnt.icsb_ifree; if (!xfs_icsb_counter_disabled(mp, XFS_SBS_FDBLOCKS)) @@ -1805,12 +1813,6 @@ xfs_icsb_balance_counter_locked( /* update counters - first CPU gets residual*/ switch (field) { - case XFS_SBS_ICOUNT: - count = mp->m_sb.sb_icount; - resid = do_div(count, weight); - if (count < max(min, XFS_ICSB_INO_CNTR_REENABLE)) - return; - break; case XFS_SBS_IFREE: count = mp->m_sb.sb_ifree; resid = do_div(count, weight); @@ -1871,14 +1873,6 @@ again: } switch (field) { - case XFS_SBS_ICOUNT: - lcounter = icsbp->icsb_icount; - lcounter += delta; - if (unlikely(lcounter < 0)) - goto balance_counter; - icsbp->icsb_icount = lcounter; - break; - case XFS_SBS_IFREE: lcounter = icsbp->icsb_ifree; lcounter += delta; diff --git a/fs/xfs/xfs_mount.h b/fs/xfs/xfs_mount.h index 0d8abd6364d9..76b18c8c58c5 100644 --- a/fs/xfs/xfs_mount.h +++ b/fs/xfs/xfs_mount.h @@ -41,7 +41,6 @@ struct xfs_da_geometry; typedef struct xfs_icsb_cnts { uint64_t icsb_fdblocks; uint64_t icsb_ifree; - uint64_t icsb_icount; unsigned long icsb_flags; } xfs_icsb_cnts_t; @@ -81,8 +80,11 @@ typedef struct xfs_mount { struct super_block *m_super; xfs_tid_t m_tid; /* next unused tid for fs */ struct xfs_ail *m_ail; /* fs active log item list */ - xfs_sb_t m_sb; /* copy of fs superblock */ + + struct xfs_sb m_sb; /* copy of fs superblock */ spinlock_t m_sb_lock; /* sb counter lock */ + struct percpu_counter m_icount; /* allocated inodes counter */ + struct xfs_buf *m_sb_bp; /* buffer for superblock */ char *m_fsname; /* filesystem name */ int m_fsname_len; /* strlen of fs name */ @@ -388,6 +390,7 @@ extern void xfs_unmountfs(xfs_mount_t *); extern int xfs_mod_incore_sb(xfs_mount_t *, xfs_sb_field_t, int64_t, int); extern int xfs_mod_incore_sb_batch(xfs_mount_t *, xfs_mod_sb_t *, uint, int); +extern int xfs_mod_icount(struct xfs_mount *mp, int64_t delta); extern int xfs_mount_log_sb(xfs_mount_t *); extern struct xfs_buf *xfs_getsb(xfs_mount_t *, int); extern int xfs_readsb(xfs_mount_t *, int); diff --git a/fs/xfs/xfs_super.c b/fs/xfs/xfs_super.c index 8fcc4ccc5c79..0aa4428bfa31 100644 --- a/fs/xfs/xfs_super.c +++ b/fs/xfs/xfs_super.c @@ -1085,6 +1085,7 @@ xfs_fs_statfs( xfs_sb_t *sbp = &mp->m_sb; struct xfs_inode *ip = XFS_I(dentry->d_inode); __uint64_t fakeinos, id; + __uint64_t icount; xfs_extlen_t lsize; __int64_t ffree; @@ -1096,6 +1097,7 @@ xfs_fs_statfs( statp->f_fsid.val[1] = (u32)(id >> 32); xfs_icsb_sync_counters(mp, XFS_ICSB_LAZY_COUNT); + icount = percpu_counter_sum(&mp->m_icount); spin_lock(&mp->m_sb_lock); statp->f_bsize = sbp->sb_blocksize; @@ -1104,8 +1106,7 @@ xfs_fs_statfs( statp->f_bfree = statp->f_bavail = sbp->sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp); fakeinos = statp->f_bfree << sbp->sb_inopblog; - statp->f_files = - MIN(sbp->sb_icount + fakeinos, (__uint64_t)XFS_MAXINUMBER); + statp->f_files = MIN(icount + fakeinos, (__uint64_t)XFS_MAXINUMBER); if (mp->m_maxicount) statp->f_files = min_t(typeof(statp->f_files), statp->f_files, @@ -1117,7 +1118,7 @@ xfs_fs_statfs( sbp->sb_icount); /* make sure statp->f_ffree does not underflow */ - ffree = statp->f_files - (sbp->sb_icount - sbp->sb_ifree); + ffree = statp->f_files - (icount - sbp->sb_ifree); statp->f_ffree = max_t(__int64_t, ffree, 0); spin_unlock(&mp->m_sb_lock); diff --git a/fs/xfs/xfs_trans.c b/fs/xfs/xfs_trans.c index eb90cd59a0ec..9bc742b65f24 100644 --- a/fs/xfs/xfs_trans.c +++ b/fs/xfs/xfs_trans.c @@ -554,8 +554,7 @@ xfs_trans_unreserve_and_mod_sb( } if (idelta) { - error = xfs_icsb_modify_counters(mp, XFS_SBS_ICOUNT, - idelta, rsvd); + error = xfs_mod_icount(mp, idelta); if (error) goto out_undo_fdblocks; } @@ -634,7 +633,7 @@ out_undo_ifreecount: xfs_icsb_modify_counters(mp, XFS_SBS_IFREE, -ifreedelta, rsvd); out_undo_icount: if (idelta) - xfs_icsb_modify_counters(mp, XFS_SBS_ICOUNT, -idelta, rsvd); + xfs_mod_icount(mp, -idelta); out_undo_fdblocks: if (blkdelta) xfs_icsb_modify_counters(mp, XFS_SBS_FDBLOCKS, -blkdelta, rsvd); -- cgit v1.2.3-59-g8ed1b From e88b64ea1f3da64dbb52636377be295c90367377 Mon Sep 17 00:00:00 2001 From: Dave Chinner Date: Mon, 23 Feb 2015 21:19:53 +1100 Subject: xfs: use generic percpu counters for free inode counter XFS has hand-rolled per-cpu counters for the superblock since before there was any generic implementation. The free inode counter is not used for any limit enforcement - the per-AG free inode counters are used during allocation to determine if there are inode available for allocation. Hence we don't need any of the complexity of the hand-rolled counters and we can simply replace them with generic per-cpu counters similar to the inode counter. This version introduces a xfs_mod_ifree() helper function from Christoph Hellwig. Signed-off-by: Dave Chinner Reviewed-by: Brian Foster Signed-off-by: Dave Chinner --- fs/xfs/libxfs/xfs_sb.c | 1 + fs/xfs/xfs_fsops.c | 2 +- fs/xfs/xfs_mount.c | 76 ++++++++++++++++++++++---------------------------- fs/xfs/xfs_mount.h | 2 ++ fs/xfs/xfs_super.c | 4 ++- fs/xfs/xfs_trans.c | 5 ++-- 6 files changed, 42 insertions(+), 48 deletions(-) diff --git a/fs/xfs/libxfs/xfs_sb.c b/fs/xfs/libxfs/xfs_sb.c index 017cb2fc53eb..b66aeab99cfb 100644 --- a/fs/xfs/libxfs/xfs_sb.c +++ b/fs/xfs/libxfs/xfs_sb.c @@ -772,6 +772,7 @@ xfs_log_sb( struct xfs_buf *bp = xfs_trans_getsb(tp, mp, 0); mp->m_sb.sb_icount = percpu_counter_sum(&mp->m_icount); + mp->m_sb.sb_ifree = percpu_counter_sum(&mp->m_ifree); xfs_sb_to_disk(XFS_BUF_TO_SBP(bp), &mp->m_sb); xfs_trans_buf_set_type(tp, bp, XFS_BLFT_SB_BUF); diff --git a/fs/xfs/xfs_fsops.c b/fs/xfs/xfs_fsops.c index b87a6f92263b..a1ca9c2b8c00 100644 --- a/fs/xfs/xfs_fsops.c +++ b/fs/xfs/xfs_fsops.c @@ -639,11 +639,11 @@ xfs_fs_counts( { xfs_icsb_sync_counters(mp, XFS_ICSB_LAZY_COUNT); cnt->allocino = percpu_counter_read_positive(&mp->m_icount); + cnt->freeino = percpu_counter_read_positive(&mp->m_ifree); spin_lock(&mp->m_sb_lock); cnt->freedata = mp->m_sb.sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp); cnt->freertx = mp->m_sb.sb_frextents; - cnt->freeino = mp->m_sb.sb_ifree; spin_unlock(&mp->m_sb_lock); return 0; } diff --git a/fs/xfs/xfs_mount.c b/fs/xfs/xfs_mount.c index 702ea6a7e648..650e8f18cd2a 100644 --- a/fs/xfs/xfs_mount.c +++ b/fs/xfs/xfs_mount.c @@ -1114,6 +1114,20 @@ xfs_mod_icount( 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; +} /* * xfs_mod_incore_sb_unlocked() is a utility routine commonly used to apply * a delta to a specified field in the in-core superblock. Simply @@ -1142,17 +1156,9 @@ xfs_mod_incore_sb_unlocked( */ switch (field) { case XFS_SBS_ICOUNT: - ASSERT(0); - return -ENOSPC; case XFS_SBS_IFREE: - lcounter = (long long)mp->m_sb.sb_ifree; - lcounter += delta; - if (lcounter < 0) { - ASSERT(0); - return -EINVAL; - } - mp->m_sb.sb_ifree = lcounter; - return 0; + ASSERT(0); + return -EINVAL; case XFS_SBS_FDBLOCKS: lcounter = (long long) mp->m_sb.sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp); @@ -1502,7 +1508,6 @@ xfs_icsb_cpu_notify( case CPU_ONLINE: case CPU_ONLINE_FROZEN: xfs_icsb_lock(mp); - xfs_icsb_balance_counter(mp, XFS_SBS_IFREE, 0); xfs_icsb_balance_counter(mp, XFS_SBS_FDBLOCKS, 0); xfs_icsb_unlock(mp); break; @@ -1513,15 +1518,12 @@ xfs_icsb_cpu_notify( * re-enable the counters. */ xfs_icsb_lock(mp); spin_lock(&mp->m_sb_lock); - xfs_icsb_disable_counter(mp, XFS_SBS_IFREE); xfs_icsb_disable_counter(mp, XFS_SBS_FDBLOCKS); - mp->m_sb.sb_ifree += cntp->icsb_ifree; mp->m_sb.sb_fdblocks += cntp->icsb_fdblocks; memset(cntp, 0, sizeof(xfs_icsb_cnts_t)); - xfs_icsb_balance_counter_locked(mp, XFS_SBS_IFREE, 0); xfs_icsb_balance_counter_locked(mp, XFS_SBS_FDBLOCKS, 0); spin_unlock(&mp->m_sb_lock); xfs_icsb_unlock(mp); @@ -1544,10 +1546,14 @@ xfs_icsb_init_counters( if (error) return error; + error = percpu_counter_init(&mp->m_ifree, 0, GFP_KERNEL); + if (error) + goto free_icount; + mp->m_sb_cnts = alloc_percpu(xfs_icsb_cnts_t); if (!mp->m_sb_cnts) { - percpu_counter_destroy(&mp->m_icount); - return -ENOMEM; + error = -ENOMEM; + goto free_ifree; } for_each_online_cpu(i) { @@ -1570,6 +1576,12 @@ xfs_icsb_init_counters( #endif /* CONFIG_HOTPLUG_CPU */ return 0; + +free_ifree: + percpu_counter_destroy(&mp->m_ifree); +free_icount: + percpu_counter_destroy(&mp->m_icount); + return error; } void @@ -1577,6 +1589,7 @@ xfs_icsb_reinit_counters( xfs_mount_t *mp) { percpu_counter_set(&mp->m_icount, mp->m_sb.sb_icount); + percpu_counter_set(&mp->m_ifree, mp->m_sb.sb_ifree); xfs_icsb_lock(mp); /* @@ -1584,7 +1597,6 @@ xfs_icsb_reinit_counters( * initial balance kicks us off correctly */ mp->m_icsb_counters = -1; - xfs_icsb_balance_counter(mp, XFS_SBS_IFREE, 0); xfs_icsb_balance_counter(mp, XFS_SBS_FDBLOCKS, 0); xfs_icsb_unlock(mp); } @@ -1599,6 +1611,7 @@ xfs_icsb_destroy_counters( } percpu_counter_destroy(&mp->m_icount); + percpu_counter_destroy(&mp->m_ifree); mutex_destroy(&mp->m_icsb_mutex); } @@ -1662,7 +1675,6 @@ xfs_icsb_count( for_each_online_cpu(i) { cntp = (xfs_icsb_cnts_t *)per_cpu_ptr(mp->m_sb_cnts, i); - cnt->icsb_ifree += cntp->icsb_ifree; cnt->icsb_fdblocks += cntp->icsb_fdblocks; } @@ -1675,7 +1687,7 @@ xfs_icsb_counter_disabled( xfs_mount_t *mp, xfs_sb_field_t field) { - ASSERT((field >= XFS_SBS_IFREE) && (field <= XFS_SBS_FDBLOCKS)); + ASSERT(field == XFS_SBS_FDBLOCKS); return test_bit(field, &mp->m_icsb_counters); } @@ -1686,7 +1698,7 @@ xfs_icsb_disable_counter( { xfs_icsb_cnts_t cnt; - ASSERT((field >= XFS_SBS_IFREE) && (field <= XFS_SBS_FDBLOCKS)); + ASSERT(field == XFS_SBS_FDBLOCKS); /* * If we are already disabled, then there is nothing to do @@ -1705,9 +1717,6 @@ xfs_icsb_disable_counter( xfs_icsb_count(mp, &cnt, XFS_ICSB_LAZY_COUNT); switch(field) { - case XFS_SBS_IFREE: - mp->m_sb.sb_ifree = cnt.icsb_ifree; - break; case XFS_SBS_FDBLOCKS: mp->m_sb.sb_fdblocks = cnt.icsb_fdblocks; break; @@ -1729,15 +1738,12 @@ xfs_icsb_enable_counter( xfs_icsb_cnts_t *cntp; int i; - ASSERT((field >= XFS_SBS_IFREE) && (field <= XFS_SBS_FDBLOCKS)); + ASSERT(field == XFS_SBS_FDBLOCKS); xfs_icsb_lock_all_counters(mp); for_each_online_cpu(i) { cntp = per_cpu_ptr(mp->m_sb_cnts, i); switch (field) { - case XFS_SBS_IFREE: - cntp->icsb_ifree = count + resid; - break; case XFS_SBS_FDBLOCKS: cntp->icsb_fdblocks = count + resid; break; @@ -1760,8 +1766,6 @@ xfs_icsb_sync_counters_locked( xfs_icsb_count(mp, &cnt, flags); - if (!xfs_icsb_counter_disabled(mp, XFS_SBS_IFREE)) - mp->m_sb.sb_ifree = cnt.icsb_ifree; if (!xfs_icsb_counter_disabled(mp, XFS_SBS_FDBLOCKS)) mp->m_sb.sb_fdblocks = cnt.icsb_fdblocks; } @@ -1813,12 +1817,6 @@ xfs_icsb_balance_counter_locked( /* update counters - first CPU gets residual*/ switch (field) { - case XFS_SBS_IFREE: - count = mp->m_sb.sb_ifree; - resid = do_div(count, weight); - if (count < max(min, XFS_ICSB_INO_CNTR_REENABLE)) - return; - break; case XFS_SBS_FDBLOCKS: count = mp->m_sb.sb_fdblocks; resid = do_div(count, weight); @@ -1873,14 +1871,6 @@ again: } switch (field) { - case XFS_SBS_IFREE: - lcounter = icsbp->icsb_ifree; - lcounter += delta; - if (unlikely(lcounter < 0)) - goto balance_counter; - icsbp->icsb_ifree = lcounter; - break; - case XFS_SBS_FDBLOCKS: BUG_ON((mp->m_resblks - mp->m_resblks_avail) != 0); diff --git a/fs/xfs/xfs_mount.h b/fs/xfs/xfs_mount.h index 76b18c8c58c5..7ce997d43d81 100644 --- a/fs/xfs/xfs_mount.h +++ b/fs/xfs/xfs_mount.h @@ -84,6 +84,7 @@ typedef struct xfs_mount { struct xfs_sb m_sb; /* copy of fs superblock */ spinlock_t m_sb_lock; /* sb counter lock */ struct percpu_counter m_icount; /* allocated inodes counter */ + struct percpu_counter m_ifree; /* free inodes counter */ struct xfs_buf *m_sb_bp; /* buffer for superblock */ char *m_fsname; /* filesystem name */ @@ -391,6 +392,7 @@ extern int xfs_mod_incore_sb(xfs_mount_t *, xfs_sb_field_t, int64_t, int); extern int xfs_mod_incore_sb_batch(xfs_mount_t *, xfs_mod_sb_t *, uint, int); 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_mount_log_sb(xfs_mount_t *); extern struct xfs_buf *xfs_getsb(xfs_mount_t *, int); extern int xfs_readsb(xfs_mount_t *, int); diff --git a/fs/xfs/xfs_super.c b/fs/xfs/xfs_super.c index 0aa4428bfa31..049147776ee1 100644 --- a/fs/xfs/xfs_super.c +++ b/fs/xfs/xfs_super.c @@ -1086,6 +1086,7 @@ xfs_fs_statfs( struct xfs_inode *ip = XFS_I(dentry->d_inode); __uint64_t fakeinos, id; __uint64_t icount; + __uint64_t ifree; xfs_extlen_t lsize; __int64_t ffree; @@ -1098,6 +1099,7 @@ xfs_fs_statfs( xfs_icsb_sync_counters(mp, XFS_ICSB_LAZY_COUNT); icount = percpu_counter_sum(&mp->m_icount); + ifree = percpu_counter_sum(&mp->m_ifree); spin_lock(&mp->m_sb_lock); statp->f_bsize = sbp->sb_blocksize; @@ -1118,7 +1120,7 @@ xfs_fs_statfs( sbp->sb_icount); /* make sure statp->f_ffree does not underflow */ - ffree = statp->f_files - (icount - sbp->sb_ifree); + ffree = statp->f_files - (icount - ifree); statp->f_ffree = max_t(__int64_t, ffree, 0); spin_unlock(&mp->m_sb_lock); diff --git a/fs/xfs/xfs_trans.c b/fs/xfs/xfs_trans.c index 9bc742b65f24..68680ce67547 100644 --- a/fs/xfs/xfs_trans.c +++ b/fs/xfs/xfs_trans.c @@ -560,8 +560,7 @@ xfs_trans_unreserve_and_mod_sb( } if (ifreedelta) { - error = xfs_icsb_modify_counters(mp, XFS_SBS_IFREE, - ifreedelta, rsvd); + error = xfs_mod_ifree(mp, ifreedelta); if (error) goto out_undo_icount; } @@ -630,7 +629,7 @@ xfs_trans_unreserve_and_mod_sb( out_undo_ifreecount: if (ifreedelta) - xfs_icsb_modify_counters(mp, XFS_SBS_IFREE, -ifreedelta, rsvd); + xfs_mod_ifree(mp, -ifreedelta); out_undo_icount: if (idelta) xfs_mod_icount(mp, -idelta); -- cgit v1.2.3-59-g8ed1b From 0d485ada404b3614b045e574bec26aaf5d9b3c5b Mon Sep 17 00:00:00 2001 From: Dave Chinner Date: Mon, 23 Feb 2015 21:22:03 +1100 Subject: xfs: use generic percpu counters for free block counter XFS has hand-rolled per-cpu counters for the superblock since before there was any generic implementation. The free block counter is special in that it is used for ENOSPC detection outside transaction contexts for for delayed allocation. This means that the counter needs to be accurate at zero. The current per-cpu counter code jumps through lots of hoops to ensure we never run past zero, but we don't need to make all those jumps with the generic counter implementation. The generic counter implementation allows us to pass a "batch" threshold at which the addition/subtraction to the counter value will be folded back into global value under lock. We can use this feature to reduce the batch size as we approach 0 in a very similar manner to the existing counters and their rebalance algorithm. If we use a batch size of 1 as we approach 0, then every addition and subtraction will be done against the global value and hence allow accurate detection of zero threshold crossing. Hence we can replace the handrolled, accurate-at-zero counters with generic percpu counters. Note: this removes just enough of the icsb infrastructure to compile without warnings. The rest will go in subsequent commits. Signed-off-by: Dave Chinner Reviewed-by: Brian Foster Signed-off-by: Dave Chinner --- fs/xfs/libxfs/xfs_bmap.c | 32 ++++---- fs/xfs/libxfs/xfs_sb.c | 1 + fs/xfs/xfs_fsops.c | 9 ++- fs/xfs/xfs_iomap.c | 2 +- fs/xfs/xfs_mount.c | 192 ++++++++++++++++++++++++----------------------- fs/xfs/xfs_mount.h | 3 + fs/xfs/xfs_super.c | 10 ++- fs/xfs/xfs_trans.c | 16 ++-- 8 files changed, 134 insertions(+), 131 deletions(-) diff --git a/fs/xfs/libxfs/xfs_bmap.c b/fs/xfs/libxfs/xfs_bmap.c index 61ec015dca16..e39c9e83670e 100644 --- a/fs/xfs/libxfs/xfs_bmap.c +++ b/fs/xfs/libxfs/xfs_bmap.c @@ -2212,9 +2212,8 @@ xfs_bmap_add_extent_delay_real( diff = (int)(temp + temp2 - startblockval(PREV.br_startblock) - (bma->cur ? bma->cur->bc_private.b.allocated : 0)); if (diff > 0) { - error = xfs_icsb_modify_counters(bma->ip->i_mount, - XFS_SBS_FDBLOCKS, - -((int64_t)diff), 0); + error = xfs_mod_fdblocks(bma->ip->i_mount, + -((int64_t)diff), false); ASSERT(!error); if (error) goto done; @@ -2265,9 +2264,8 @@ xfs_bmap_add_extent_delay_real( temp += bma->cur->bc_private.b.allocated; ASSERT(temp <= da_old); if (temp < da_old) - xfs_icsb_modify_counters(bma->ip->i_mount, - XFS_SBS_FDBLOCKS, - (int64_t)(da_old - temp), 0); + xfs_mod_fdblocks(bma->ip->i_mount, + (int64_t)(da_old - temp), false); } /* clear out the allocated field, done with it now in any case. */ @@ -2944,8 +2942,8 @@ xfs_bmap_add_extent_hole_delay( } if (oldlen != newlen) { ASSERT(oldlen > newlen); - xfs_icsb_modify_counters(ip->i_mount, XFS_SBS_FDBLOCKS, - (int64_t)(oldlen - newlen), 0); + xfs_mod_fdblocks(ip->i_mount, (int64_t)(oldlen - newlen), + false); /* * Nothing to do for disk quota accounting here. */ @@ -4163,15 +4161,13 @@ xfs_bmapi_reserve_delalloc( error = xfs_mod_incore_sb(mp, XFS_SBS_FREXTENTS, -((int64_t)extsz), 0); } else { - error = xfs_icsb_modify_counters(mp, XFS_SBS_FDBLOCKS, - -((int64_t)alen), 0); + error = xfs_mod_fdblocks(mp, -((int64_t)alen), false); } if (error) goto out_unreserve_quota; - error = xfs_icsb_modify_counters(mp, XFS_SBS_FDBLOCKS, - -((int64_t)indlen), 0); + error = xfs_mod_fdblocks(mp, -((int64_t)indlen), false); if (error) goto out_unreserve_blocks; @@ -4200,7 +4196,7 @@ out_unreserve_blocks: if (rt) xfs_mod_incore_sb(mp, XFS_SBS_FREXTENTS, extsz, 0); else - xfs_icsb_modify_counters(mp, XFS_SBS_FDBLOCKS, alen, 0); + xfs_mod_fdblocks(mp, alen, false); out_unreserve_quota: if (XFS_IS_QUOTA_ON(mp)) xfs_trans_unreserve_quota_nblks(NULL, ip, (long)alen, 0, rt ? @@ -5012,10 +5008,8 @@ xfs_bmap_del_extent( * Nothing to do for disk quota accounting here. */ ASSERT(da_old >= da_new); - if (da_old > da_new) { - xfs_icsb_modify_counters(mp, XFS_SBS_FDBLOCKS, - (int64_t)(da_old - da_new), 0); - } + if (da_old > da_new) + xfs_mod_fdblocks(mp, (int64_t)(da_old - da_new), false); done: *logflagsp = flags; return error; @@ -5290,8 +5284,8 @@ xfs_bunmapi( ip, -((long)del.br_blockcount), 0, XFS_QMOPT_RES_RTBLKS); } else { - xfs_icsb_modify_counters(mp, XFS_SBS_FDBLOCKS, - (int64_t)del.br_blockcount, 0); + xfs_mod_fdblocks(mp, (int64_t)del.br_blockcount, + false); (void)xfs_trans_reserve_quota_nblks(NULL, ip, -((long)del.br_blockcount), 0, XFS_QMOPT_RES_REGBLKS); diff --git a/fs/xfs/libxfs/xfs_sb.c b/fs/xfs/libxfs/xfs_sb.c index b66aeab99cfb..31a3e972f86f 100644 --- a/fs/xfs/libxfs/xfs_sb.c +++ b/fs/xfs/libxfs/xfs_sb.c @@ -773,6 +773,7 @@ xfs_log_sb( mp->m_sb.sb_icount = percpu_counter_sum(&mp->m_icount); mp->m_sb.sb_ifree = percpu_counter_sum(&mp->m_ifree); + mp->m_sb.sb_fdblocks = percpu_counter_sum(&mp->m_fdblocks); xfs_sb_to_disk(XFS_BUF_TO_SBP(bp), &mp->m_sb); xfs_trans_buf_set_type(tp, bp, XFS_BLFT_SB_BUF); diff --git a/fs/xfs/xfs_fsops.c b/fs/xfs/xfs_fsops.c index a1ca9c2b8c00..7ef25588062f 100644 --- a/fs/xfs/xfs_fsops.c +++ b/fs/xfs/xfs_fsops.c @@ -640,9 +640,10 @@ xfs_fs_counts( xfs_icsb_sync_counters(mp, XFS_ICSB_LAZY_COUNT); cnt->allocino = percpu_counter_read_positive(&mp->m_icount); cnt->freeino = percpu_counter_read_positive(&mp->m_ifree); + cnt->freedata = percpu_counter_read_positive(&mp->m_fdblocks) - + XFS_ALLOC_SET_ASIDE(mp); spin_lock(&mp->m_sb_lock); - cnt->freedata = mp->m_sb.sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp); cnt->freertx = mp->m_sb.sb_frextents; spin_unlock(&mp->m_sb_lock); return 0; @@ -717,7 +718,8 @@ retry: } else { __int64_t free; - free = mp->m_sb.sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp); + free = percpu_counter_sum(&mp->m_fdblocks) - + XFS_ALLOC_SET_ASIDE(mp); if (!free) goto out; /* ENOSPC and fdblks_delta = 0 */ @@ -756,8 +758,7 @@ out: * the extra reserve blocks from the reserve..... */ int error; - error = xfs_icsb_modify_counters(mp, XFS_SBS_FDBLOCKS, - fdblks_delta, 0); + error = xfs_mod_fdblocks(mp, fdblks_delta, 0); if (error == -ENOSPC) goto retry; } diff --git a/fs/xfs/xfs_iomap.c b/fs/xfs/xfs_iomap.c index ccb1dd0d509e..205b948c9d20 100644 --- a/fs/xfs/xfs_iomap.c +++ b/fs/xfs/xfs_iomap.c @@ -461,7 +461,7 @@ xfs_iomap_prealloc_size( alloc_blocks); xfs_icsb_sync_counters(mp, XFS_ICSB_LAZY_COUNT); - freesp = mp->m_sb.sb_fdblocks; + freesp = percpu_counter_read_positive(&mp->m_fdblocks); if (freesp < mp->m_low_space[XFS_LOWSP_5_PCNT]) { shift = 2; if (freesp < mp->m_low_space[XFS_LOWSP_4_PCNT]) diff --git a/fs/xfs/xfs_mount.c b/fs/xfs/xfs_mount.c index 650e8f18cd2a..767c09a5d3ff 100644 --- a/fs/xfs/xfs_mount.c +++ b/fs/xfs/xfs_mount.c @@ -1114,7 +1114,6 @@ xfs_mod_icount( return 0; } - int xfs_mod_ifree( struct xfs_mount *mp, @@ -1128,6 +1127,92 @@ xfs_mod_ifree( } return 0; } + +int +xfs_mod_fdblocks( + struct xfs_mount *mp, + int64_t delta, + bool rsvd) +{ + int64_t lcounter; + long long res_used; + s32 batch; + + if (delta > 0) { + /* + * If the reserve pool is depleted, put blocks back into it + * first. Most of the time the pool is full. + */ + if (likely(mp->m_resblks == mp->m_resblks_avail)) { + percpu_counter_add(&mp->m_fdblocks, delta); + return 0; + } + + spin_lock(&mp->m_sb_lock); + res_used = (long long)(mp->m_resblks - mp->m_resblks_avail); + + if (res_used > delta) { + mp->m_resblks_avail += delta; + } else { + delta -= res_used; + mp->m_resblks_avail = mp->m_resblks; + percpu_counter_add(&mp->m_fdblocks, delta); + } + spin_unlock(&mp->m_sb_lock); + return 0; + } + + /* + * Taking blocks away, need to be more accurate the closer we + * are to zero. + * + * batch size is set to a maximum of 1024 blocks - if we are + * allocating of freeing extents larger than this then we aren't + * going to be hammering the counter lock so a lock per update + * is not a problem. + * + * If the counter has a value of less than 2 * max batch size, + * then make everything serialise as we are real close to + * ENOSPC. + */ +#define __BATCH 1024 + if (percpu_counter_compare(&mp->m_fdblocks, 2 * __BATCH) < 0) + batch = 1; + else + batch = __BATCH; +#undef __BATCH + + __percpu_counter_add(&mp->m_fdblocks, delta, batch); + if (percpu_counter_compare(&mp->m_fdblocks, + XFS_ALLOC_SET_ASIDE(mp)) >= 0) { + /* we had space! */ + return 0; + } + + /* + * lock up the sb for dipping into reserves before releasing the space + * that took us to ENOSPC. + */ + spin_lock(&mp->m_sb_lock); + percpu_counter_add(&mp->m_fdblocks, -delta); + if (!rsvd) + goto fdblocks_enospc; + + lcounter = (long long)mp->m_resblks_avail + delta; + if (lcounter >= 0) { + mp->m_resblks_avail = lcounter; + spin_unlock(&mp->m_sb_lock); + return 0; + } + printk_once(KERN_WARNING + "Filesystem \"%s\": reserve blocks depleted! " + "Consider increasing reserve pool size.", + mp->m_fsname); +fdblocks_enospc: + spin_unlock(&mp->m_sb_lock); + return -ENOSPC; +} + /* * xfs_mod_incore_sb_unlocked() is a utility routine commonly used to apply * a delta to a specified field in the in-core superblock. Simply @@ -1146,7 +1231,6 @@ xfs_mod_incore_sb_unlocked( { int scounter; /* short counter for 32 bit fields */ long long lcounter; /* long counter for 64 bit fields */ - long long res_used, rem; /* * With the in-core superblock spin lock held, switch @@ -1157,50 +1241,9 @@ xfs_mod_incore_sb_unlocked( switch (field) { case XFS_SBS_ICOUNT: case XFS_SBS_IFREE: + case XFS_SBS_FDBLOCKS: ASSERT(0); return -EINVAL; - case XFS_SBS_FDBLOCKS: - lcounter = (long long) - mp->m_sb.sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp); - res_used = (long long)(mp->m_resblks - mp->m_resblks_avail); - - if (delta > 0) { /* Putting blocks back */ - if (res_used > delta) { - mp->m_resblks_avail += delta; - } else { - rem = delta - res_used; - mp->m_resblks_avail = mp->m_resblks; - lcounter += rem; - } - } else { /* Taking blocks away */ - lcounter += delta; - if (lcounter >= 0) { - mp->m_sb.sb_fdblocks = lcounter + - XFS_ALLOC_SET_ASIDE(mp); - return 0; - } - - /* - * We are out of blocks, use any available reserved - * blocks if were allowed to. - */ - if (!rsvd) - return -ENOSPC; - - lcounter = (long long)mp->m_resblks_avail + delta; - if (lcounter >= 0) { - mp->m_resblks_avail = lcounter; - return 0; - } - printk_once(KERN_WARNING - "Filesystem \"%s\": reserve blocks depleted! " - "Consider increasing reserve pool size.", - mp->m_fsname); - return -ENOSPC; - } - - mp->m_sb.sb_fdblocks = lcounter + XFS_ALLOC_SET_ASIDE(mp); - return 0; case XFS_SBS_FREXTENTS: lcounter = (long long)mp->m_sb.sb_frextents; lcounter += delta; @@ -1323,7 +1366,7 @@ xfs_mod_incore_sb( * * Note that this function may not be used for the superblock values that * are tracked with the in-memory per-cpu counters - a direct call to - * xfs_icsb_modify_counters is required for these. + * xfs_mod_incore_sb is required for these. */ int xfs_mod_incore_sb_batch( @@ -1508,7 +1551,6 @@ xfs_icsb_cpu_notify( case CPU_ONLINE: case CPU_ONLINE_FROZEN: xfs_icsb_lock(mp); - xfs_icsb_balance_counter(mp, XFS_SBS_FDBLOCKS, 0); xfs_icsb_unlock(mp); break; case CPU_DEAD: @@ -1518,13 +1560,9 @@ xfs_icsb_cpu_notify( * re-enable the counters. */ xfs_icsb_lock(mp); spin_lock(&mp->m_sb_lock); - xfs_icsb_disable_counter(mp, XFS_SBS_FDBLOCKS); - - mp->m_sb.sb_fdblocks += cntp->icsb_fdblocks; memset(cntp, 0, sizeof(xfs_icsb_cnts_t)); - xfs_icsb_balance_counter_locked(mp, XFS_SBS_FDBLOCKS, 0); spin_unlock(&mp->m_sb_lock); xfs_icsb_unlock(mp); break; @@ -1550,10 +1588,14 @@ xfs_icsb_init_counters( if (error) goto free_icount; + error = percpu_counter_init(&mp->m_fdblocks, 0, GFP_KERNEL); + if (error) + goto free_ifree; + mp->m_sb_cnts = alloc_percpu(xfs_icsb_cnts_t); if (!mp->m_sb_cnts) { error = -ENOMEM; - goto free_ifree; + goto free_fdblocks; } for_each_online_cpu(i) { @@ -1577,6 +1619,8 @@ xfs_icsb_init_counters( return 0; +free_fdblocks: + percpu_counter_destroy(&mp->m_fdblocks); free_ifree: percpu_counter_destroy(&mp->m_ifree); free_icount: @@ -1590,6 +1634,7 @@ xfs_icsb_reinit_counters( { percpu_counter_set(&mp->m_icount, mp->m_sb.sb_icount); percpu_counter_set(&mp->m_ifree, mp->m_sb.sb_ifree); + percpu_counter_set(&mp->m_fdblocks, mp->m_sb.sb_fdblocks); xfs_icsb_lock(mp); /* @@ -1597,7 +1642,6 @@ xfs_icsb_reinit_counters( * initial balance kicks us off correctly */ mp->m_icsb_counters = -1; - xfs_icsb_balance_counter(mp, XFS_SBS_FDBLOCKS, 0); xfs_icsb_unlock(mp); } @@ -1612,6 +1656,7 @@ xfs_icsb_destroy_counters( percpu_counter_destroy(&mp->m_icount); percpu_counter_destroy(&mp->m_ifree); + percpu_counter_destroy(&mp->m_fdblocks); mutex_destroy(&mp->m_icsb_mutex); } @@ -1665,18 +1710,11 @@ xfs_icsb_count( xfs_icsb_cnts_t *cnt, int flags) { - xfs_icsb_cnts_t *cntp; - int i; - memset(cnt, 0, sizeof(xfs_icsb_cnts_t)); if (!(flags & XFS_ICSB_LAZY_COUNT)) xfs_icsb_lock_all_counters(mp); - for_each_online_cpu(i) { - cntp = (xfs_icsb_cnts_t *)per_cpu_ptr(mp->m_sb_cnts, i); - cnt->icsb_fdblocks += cntp->icsb_fdblocks; - } if (!(flags & XFS_ICSB_LAZY_COUNT)) xfs_icsb_unlock_all_counters(mp); @@ -1687,7 +1725,6 @@ xfs_icsb_counter_disabled( xfs_mount_t *mp, xfs_sb_field_t field) { - ASSERT(field == XFS_SBS_FDBLOCKS); return test_bit(field, &mp->m_icsb_counters); } @@ -1698,8 +1735,6 @@ xfs_icsb_disable_counter( { xfs_icsb_cnts_t cnt; - ASSERT(field == XFS_SBS_FDBLOCKS); - /* * If we are already disabled, then there is nothing to do * here. We check before locking all the counters to avoid @@ -1717,9 +1752,6 @@ xfs_icsb_disable_counter( xfs_icsb_count(mp, &cnt, XFS_ICSB_LAZY_COUNT); switch(field) { - case XFS_SBS_FDBLOCKS: - mp->m_sb.sb_fdblocks = cnt.icsb_fdblocks; - break; default: BUG(); } @@ -1735,18 +1767,11 @@ xfs_icsb_enable_counter( uint64_t count, uint64_t resid) { - xfs_icsb_cnts_t *cntp; int i; - ASSERT(field == XFS_SBS_FDBLOCKS); - xfs_icsb_lock_all_counters(mp); for_each_online_cpu(i) { - cntp = per_cpu_ptr(mp->m_sb_cnts, i); switch (field) { - case XFS_SBS_FDBLOCKS: - cntp->icsb_fdblocks = count + resid; - break; default: BUG(); break; @@ -1765,9 +1790,6 @@ xfs_icsb_sync_counters_locked( xfs_icsb_cnts_t cnt; xfs_icsb_count(mp, &cnt, flags); - - if (!xfs_icsb_counter_disabled(mp, XFS_SBS_FDBLOCKS)) - mp->m_sb.sb_fdblocks = cnt.icsb_fdblocks; } /* @@ -1809,20 +1831,12 @@ xfs_icsb_balance_counter_locked( int min_per_cpu) { uint64_t count, resid; - int weight = num_online_cpus(); - uint64_t min = (uint64_t)min_per_cpu; /* disable counter and sync counter */ xfs_icsb_disable_counter(mp, field); /* update counters - first CPU gets residual*/ switch (field) { - case XFS_SBS_FDBLOCKS: - count = mp->m_sb.sb_fdblocks; - resid = do_div(count, weight); - if (count < max(min, XFS_ICSB_FDBLK_CNTR_REENABLE(mp))) - return; - break; default: BUG(); count = resid = 0; /* quiet, gcc */ @@ -1851,7 +1865,6 @@ xfs_icsb_modify_counters( int rsvd) { xfs_icsb_cnts_t *icsbp; - long long lcounter; /* long counter for 64 bit fields */ int ret = 0; might_sleep(); @@ -1871,18 +1884,9 @@ again: } switch (field) { - case XFS_SBS_FDBLOCKS: - BUG_ON((mp->m_resblks - mp->m_resblks_avail) != 0); - - lcounter = icsbp->icsb_fdblocks - XFS_ALLOC_SET_ASIDE(mp); - lcounter += delta; - if (unlikely(lcounter < 0)) - goto balance_counter; - icsbp->icsb_fdblocks = lcounter + XFS_ALLOC_SET_ASIDE(mp); - break; default: BUG(); - break; + goto balance_counter; /* be still, gcc */ } xfs_icsb_unlock_cntr(icsbp); preempt_enable(); diff --git a/fs/xfs/xfs_mount.h b/fs/xfs/xfs_mount.h index 7ce997d43d81..84b745fbc787 100644 --- a/fs/xfs/xfs_mount.h +++ b/fs/xfs/xfs_mount.h @@ -85,6 +85,7 @@ typedef struct xfs_mount { spinlock_t m_sb_lock; /* sb counter lock */ struct percpu_counter m_icount; /* allocated inodes counter */ struct percpu_counter m_ifree; /* free inodes counter */ + struct percpu_counter m_fdblocks; /* free block counter */ struct xfs_buf *m_sb_bp; /* buffer for superblock */ char *m_fsname; /* filesystem name */ @@ -393,6 +394,8 @@ extern int xfs_mod_incore_sb_batch(xfs_mount_t *, xfs_mod_sb_t *, uint, int); 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_mount_log_sb(xfs_mount_t *); extern struct xfs_buf *xfs_getsb(xfs_mount_t *, int); extern int xfs_readsb(xfs_mount_t *, int); diff --git a/fs/xfs/xfs_super.c b/fs/xfs/xfs_super.c index 049147776ee1..9ec75074026d 100644 --- a/fs/xfs/xfs_super.c +++ b/fs/xfs/xfs_super.c @@ -1087,6 +1087,7 @@ xfs_fs_statfs( __uint64_t fakeinos, id; __uint64_t icount; __uint64_t ifree; + __uint64_t fdblocks; xfs_extlen_t lsize; __int64_t ffree; @@ -1100,13 +1101,17 @@ xfs_fs_statfs( xfs_icsb_sync_counters(mp, XFS_ICSB_LAZY_COUNT); icount = percpu_counter_sum(&mp->m_icount); ifree = percpu_counter_sum(&mp->m_ifree); + fdblocks = percpu_counter_sum(&mp->m_fdblocks); spin_lock(&mp->m_sb_lock); statp->f_bsize = sbp->sb_blocksize; lsize = sbp->sb_logstart ? sbp->sb_logblocks : 0; statp->f_blocks = sbp->sb_dblocks - lsize; - statp->f_bfree = statp->f_bavail = - sbp->sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp); + spin_unlock(&mp->m_sb_lock); + + statp->f_bfree = fdblocks - XFS_ALLOC_SET_ASIDE(mp); + statp->f_bavail = statp->f_bfree; + fakeinos = statp->f_bfree << sbp->sb_inopblog; statp->f_files = MIN(icount + fakeinos, (__uint64_t)XFS_MAXINUMBER); if (mp->m_maxicount) @@ -1123,7 +1128,6 @@ xfs_fs_statfs( ffree = statp->f_files - (icount - ifree); statp->f_ffree = max_t(__int64_t, ffree, 0); - spin_unlock(&mp->m_sb_lock); if ((ip->i_d.di_flags & XFS_DIFLAG_PROJINHERIT) && ((mp->m_qflags & (XFS_PQUOTA_ACCT|XFS_PQUOTA_ENFD))) == diff --git a/fs/xfs/xfs_trans.c b/fs/xfs/xfs_trans.c index 68680ce67547..e99f5e552c64 100644 --- a/fs/xfs/xfs_trans.c +++ b/fs/xfs/xfs_trans.c @@ -173,7 +173,7 @@ xfs_trans_reserve( uint rtextents) { int error = 0; - int rsvd = (tp->t_flags & XFS_TRANS_RESERVE) != 0; + bool rsvd = (tp->t_flags & XFS_TRANS_RESERVE) != 0; /* Mark this thread as being in a transaction */ current_set_flags_nested(&tp->t_pflags, PF_FSTRANS); @@ -184,8 +184,7 @@ xfs_trans_reserve( * fail if the count would go below zero. */ if (blocks > 0) { - error = xfs_icsb_modify_counters(tp->t_mountp, XFS_SBS_FDBLOCKS, - -((int64_t)blocks), rsvd); + error = xfs_mod_fdblocks(tp->t_mountp, -((int64_t)blocks), rsvd); if (error != 0) { current_restore_flags_nested(&tp->t_pflags, PF_FSTRANS); return -ENOSPC; @@ -268,8 +267,7 @@ undo_log: undo_blocks: if (blocks > 0) { - xfs_icsb_modify_counters(tp->t_mountp, XFS_SBS_FDBLOCKS, - (int64_t)blocks, rsvd); + xfs_mod_fdblocks(tp->t_mountp, -((int64_t)blocks), rsvd); tp->t_blk_res = 0; } @@ -516,14 +514,13 @@ xfs_trans_unreserve_and_mod_sb( xfs_mount_t *mp = tp->t_mountp; /* REFERENCED */ int error; - int rsvd; + bool rsvd = (tp->t_flags & XFS_TRANS_RESERVE) != 0; int64_t blkdelta = 0; int64_t rtxdelta = 0; int64_t idelta = 0; int64_t ifreedelta = 0; msbp = msb; - rsvd = (tp->t_flags & XFS_TRANS_RESERVE) != 0; /* calculate deltas */ if (tp->t_blk_res > 0) @@ -547,8 +544,7 @@ xfs_trans_unreserve_and_mod_sb( /* apply the per-cpu counters */ if (blkdelta) { - error = xfs_icsb_modify_counters(mp, XFS_SBS_FDBLOCKS, - blkdelta, rsvd); + error = xfs_mod_fdblocks(mp, blkdelta, rsvd); if (error) goto out; } @@ -635,7 +631,7 @@ out_undo_icount: xfs_mod_icount(mp, -idelta); out_undo_fdblocks: if (blkdelta) - xfs_icsb_modify_counters(mp, XFS_SBS_FDBLOCKS, -blkdelta, rsvd); + xfs_mod_fdblocks(mp, -blkdelta, rsvd); out: ASSERT(error == 0); return; -- cgit v1.2.3-59-g8ed1b From 5681ca40064fdb3efe477a604d690ab0425708b3 Mon Sep 17 00:00:00 2001 From: Dave Chinner Date: Mon, 23 Feb 2015 21:22:31 +1100 Subject: xfs: Remove icsb infrastructure Now that the in-core superblock infrastructure has been replaced with generic per-cpu counters, we don't need it anymore. Nuke it from orbit so we are sure that it won't haunt us again... Signed-off-by: Dave Chinner Reviewed-by: Brian Foster Signed-off-by: Dave Chinner --- fs/xfs/libxfs/xfs_sb.c | 8 +- fs/xfs/xfs_fsops.c | 2 - fs/xfs/xfs_iomap.c | 1 - fs/xfs/xfs_linux.h | 9 - fs/xfs/xfs_log_recover.c | 4 +- fs/xfs/xfs_mount.c | 516 +---------------------------------------------- fs/xfs/xfs_mount.h | 67 ------ fs/xfs/xfs_super.c | 84 ++++++-- fs/xfs/xfs_super.h | 2 + 9 files changed, 72 insertions(+), 621 deletions(-) diff --git a/fs/xfs/libxfs/xfs_sb.c b/fs/xfs/libxfs/xfs_sb.c index 31a3e972f86f..a270095ec3c0 100644 --- a/fs/xfs/libxfs/xfs_sb.c +++ b/fs/xfs/libxfs/xfs_sb.c @@ -743,17 +743,15 @@ xfs_initialize_perag_data( btree += pag->pagf_btreeblks; xfs_perag_put(pag); } - /* - * Overwrite incore superblock counters with just-read data - */ + + /* Overwrite incore superblock counters with just-read data */ spin_lock(&mp->m_sb_lock); sbp->sb_ifree = ifree; sbp->sb_icount = ialloc; sbp->sb_fdblocks = bfree + bfreelst + btree; spin_unlock(&mp->m_sb_lock); - /* Fixup the per-cpu counters as well. */ - xfs_icsb_reinit_counters(mp); + xfs_reinit_percpu_counters(mp); return 0; } diff --git a/fs/xfs/xfs_fsops.c b/fs/xfs/xfs_fsops.c index 7ef25588062f..16e62edc9dae 100644 --- a/fs/xfs/xfs_fsops.c +++ b/fs/xfs/xfs_fsops.c @@ -637,7 +637,6 @@ xfs_fs_counts( xfs_mount_t *mp, xfs_fsop_counts_t *cnt) { - xfs_icsb_sync_counters(mp, XFS_ICSB_LAZY_COUNT); cnt->allocino = percpu_counter_read_positive(&mp->m_icount); cnt->freeino = percpu_counter_read_positive(&mp->m_ifree); cnt->freedata = percpu_counter_read_positive(&mp->m_fdblocks) - @@ -701,7 +700,6 @@ xfs_reserve_blocks( */ retry: spin_lock(&mp->m_sb_lock); - xfs_icsb_sync_counters_locked(mp, 0); /* * If our previous reservation was larger than the current value, diff --git a/fs/xfs/xfs_iomap.c b/fs/xfs/xfs_iomap.c index 205b948c9d20..38e633bad8c2 100644 --- a/fs/xfs/xfs_iomap.c +++ b/fs/xfs/xfs_iomap.c @@ -460,7 +460,6 @@ xfs_iomap_prealloc_size( alloc_blocks = XFS_FILEOFF_MIN(roundup_pow_of_two(MAXEXTLEN), alloc_blocks); - xfs_icsb_sync_counters(mp, XFS_ICSB_LAZY_COUNT); freesp = percpu_counter_read_positive(&mp->m_fdblocks); if (freesp < mp->m_low_space[XFS_LOWSP_5_PCNT]) { shift = 2; diff --git a/fs/xfs/xfs_linux.h b/fs/xfs/xfs_linux.h index c31d2c2eadc4..7c7842c85a08 100644 --- a/fs/xfs/xfs_linux.h +++ b/fs/xfs/xfs_linux.h @@ -116,15 +116,6 @@ typedef __uint64_t __psunsigned_t; #undef XFS_NATIVE_HOST #endif -/* - * Feature macros (disable/enable) - */ -#ifdef CONFIG_SMP -#define HAVE_PERCPU_SB /* per cpu superblock counters are a 2.6 feature */ -#else -#undef HAVE_PERCPU_SB /* per cpu superblock counters are a 2.6 feature */ -#endif - #define irix_sgid_inherit xfs_params.sgid_inherit.val #define irix_symlink_mode xfs_params.symlink_mode.val #define xfs_panic_mask xfs_params.panic_mask.val diff --git a/fs/xfs/xfs_log_recover.c b/fs/xfs/xfs_log_recover.c index a5a945fc3bdc..4f5784f85a5b 100644 --- a/fs/xfs/xfs_log_recover.c +++ b/fs/xfs/xfs_log_recover.c @@ -4463,10 +4463,10 @@ xlog_do_recover( xfs_sb_from_disk(sbp, XFS_BUF_TO_SBP(bp)); ASSERT(sbp->sb_magicnum == XFS_SB_MAGIC); ASSERT(xfs_sb_good_version(sbp)); + xfs_reinit_percpu_counters(log->l_mp); + xfs_buf_relse(bp); - /* We've re-read the superblock so re-initialize per-cpu counters */ - xfs_icsb_reinit_counters(log->l_mp); xlog_recover_check_summary(log); diff --git a/fs/xfs/xfs_mount.c b/fs/xfs/xfs_mount.c index 767c09a5d3ff..05b392e35e35 100644 --- a/fs/xfs/xfs_mount.c +++ b/fs/xfs/xfs_mount.c @@ -43,18 +43,6 @@ #include "xfs_sysfs.h" -#ifdef HAVE_PERCPU_SB -STATIC void xfs_icsb_balance_counter(xfs_mount_t *, xfs_sb_field_t, - int); -STATIC void xfs_icsb_balance_counter_locked(xfs_mount_t *, xfs_sb_field_t, - int); -STATIC void xfs_icsb_disable_counter(xfs_mount_t *, xfs_sb_field_t); -#else - -#define xfs_icsb_balance_counter(mp, a, b) do { } while (0) -#define xfs_icsb_balance_counter_locked(mp, a, b) do { } while (0) -#endif - static DEFINE_MUTEX(xfs_uuid_table_mutex); static int xfs_uuid_table_size; static uuid_t *xfs_uuid_table; @@ -347,8 +335,7 @@ reread: goto reread; } - /* Initialize per-cpu counters */ - xfs_icsb_reinit_counters(mp); + xfs_reinit_percpu_counters(mp); /* no need to be quiet anymore, so reset the buf ops */ bp->b_ops = &xfs_sb_buf_ops; @@ -1087,8 +1074,6 @@ xfs_log_sbcount(xfs_mount_t *mp) if (!xfs_fs_writable(mp, SB_FREEZE_COMPLETE)) return 0; - xfs_icsb_sync_counters(mp, 0); - /* * we don't need to do this if we are updating the superblock * counters on every modification. @@ -1466,502 +1451,3 @@ xfs_dev_is_read_only( } return 0; } - -#ifdef HAVE_PERCPU_SB -/* - * Per-cpu incore superblock counters - * - * Simple concept, difficult implementation - * - * Basically, replace the incore superblock counters with a distributed per cpu - * counter for contended fields (e.g. free block count). - * - * Difficulties arise in that the incore sb is used for ENOSPC checking, and - * hence needs to be accurately read when we are running low on space. Hence - * there is a method to enable and disable the per-cpu counters based on how - * much "stuff" is available in them. - * - * Basically, a counter is enabled if there is enough free resource to justify - * running a per-cpu fast-path. If the per-cpu counter runs out (i.e. a local - * ENOSPC), then we disable the counters to synchronise all callers and - * re-distribute the available resources. - * - * If, once we redistributed the available resources, we still get a failure, - * we disable the per-cpu counter and go through the slow path. - * - * The slow path is the current xfs_mod_incore_sb() function. This means that - * when we disable a per-cpu counter, we need to drain its resources back to - * the global superblock. We do this after disabling the counter to prevent - * more threads from queueing up on the counter. - * - * Essentially, this means that we still need a lock in the fast path to enable - * synchronisation between the global counters and the per-cpu counters. This - * is not a problem because the lock will be local to a CPU almost all the time - * and have little contention except when we get to ENOSPC conditions. - * - * Basically, this lock becomes a barrier that enables us to lock out the fast - * path while we do things like enabling and disabling counters and - * synchronising the counters. - * - * Locking rules: - * - * 1. m_sb_lock before picking up per-cpu locks - * 2. per-cpu locks always picked up via for_each_online_cpu() order - * 3. accurate counter sync requires m_sb_lock + per cpu locks - * 4. modifying per-cpu counters requires holding per-cpu lock - * 5. modifying global counters requires holding m_sb_lock - * 6. enabling or disabling a counter requires holding the m_sb_lock - * and _none_ of the per-cpu locks. - * - * Disabled counters are only ever re-enabled by a balance operation - * that results in more free resources per CPU than a given threshold. - * To ensure counters don't remain disabled, they are rebalanced when - * the global resource goes above a higher threshold (i.e. some hysteresis - * is present to prevent thrashing). - */ - -#ifdef CONFIG_HOTPLUG_CPU -/* - * hot-plug CPU notifier support. - * - * We need a notifier per filesystem as we need to be able to identify - * the filesystem to balance the counters out. This is achieved by - * having a notifier block embedded in the xfs_mount_t and doing pointer - * magic to get the mount pointer from the notifier block address. - */ -STATIC int -xfs_icsb_cpu_notify( - struct notifier_block *nfb, - unsigned long action, - void *hcpu) -{ - xfs_icsb_cnts_t *cntp; - xfs_mount_t *mp; - - mp = (xfs_mount_t *)container_of(nfb, xfs_mount_t, m_icsb_notifier); - cntp = (xfs_icsb_cnts_t *) - per_cpu_ptr(mp->m_sb_cnts, (unsigned long)hcpu); - switch (action) { - case CPU_UP_PREPARE: - case CPU_UP_PREPARE_FROZEN: - /* Easy Case - initialize the area and locks, and - * then rebalance when online does everything else for us. */ - memset(cntp, 0, sizeof(xfs_icsb_cnts_t)); - break; - case CPU_ONLINE: - case CPU_ONLINE_FROZEN: - xfs_icsb_lock(mp); - xfs_icsb_unlock(mp); - break; - case CPU_DEAD: - case CPU_DEAD_FROZEN: - /* Disable all the counters, then fold the dead cpu's - * count into the total on the global superblock and - * re-enable the counters. */ - xfs_icsb_lock(mp); - spin_lock(&mp->m_sb_lock); - - memset(cntp, 0, sizeof(xfs_icsb_cnts_t)); - - spin_unlock(&mp->m_sb_lock); - xfs_icsb_unlock(mp); - break; - } - - return NOTIFY_OK; -} -#endif /* CONFIG_HOTPLUG_CPU */ - -int -xfs_icsb_init_counters( - xfs_mount_t *mp) -{ - xfs_icsb_cnts_t *cntp; - int error; - int i; - - error = percpu_counter_init(&mp->m_icount, 0, GFP_KERNEL); - if (error) - return error; - - error = percpu_counter_init(&mp->m_ifree, 0, GFP_KERNEL); - if (error) - goto free_icount; - - error = percpu_counter_init(&mp->m_fdblocks, 0, GFP_KERNEL); - if (error) - goto free_ifree; - - mp->m_sb_cnts = alloc_percpu(xfs_icsb_cnts_t); - if (!mp->m_sb_cnts) { - error = -ENOMEM; - goto free_fdblocks; - } - - for_each_online_cpu(i) { - cntp = (xfs_icsb_cnts_t *)per_cpu_ptr(mp->m_sb_cnts, i); - memset(cntp, 0, sizeof(xfs_icsb_cnts_t)); - } - - mutex_init(&mp->m_icsb_mutex); - - /* - * start with all counters disabled so that the - * initial balance kicks us off correctly - */ - mp->m_icsb_counters = -1; - -#ifdef CONFIG_HOTPLUG_CPU - mp->m_icsb_notifier.notifier_call = xfs_icsb_cpu_notify; - mp->m_icsb_notifier.priority = 0; - register_hotcpu_notifier(&mp->m_icsb_notifier); -#endif /* CONFIG_HOTPLUG_CPU */ - - return 0; - -free_fdblocks: - percpu_counter_destroy(&mp->m_fdblocks); -free_ifree: - percpu_counter_destroy(&mp->m_ifree); -free_icount: - percpu_counter_destroy(&mp->m_icount); - return error; -} - -void -xfs_icsb_reinit_counters( - xfs_mount_t *mp) -{ - percpu_counter_set(&mp->m_icount, mp->m_sb.sb_icount); - percpu_counter_set(&mp->m_ifree, mp->m_sb.sb_ifree); - percpu_counter_set(&mp->m_fdblocks, mp->m_sb.sb_fdblocks); - - xfs_icsb_lock(mp); - /* - * start with all counters disabled so that the - * initial balance kicks us off correctly - */ - mp->m_icsb_counters = -1; - xfs_icsb_unlock(mp); -} - -void -xfs_icsb_destroy_counters( - xfs_mount_t *mp) -{ - if (mp->m_sb_cnts) { - unregister_hotcpu_notifier(&mp->m_icsb_notifier); - free_percpu(mp->m_sb_cnts); - } - - percpu_counter_destroy(&mp->m_icount); - percpu_counter_destroy(&mp->m_ifree); - percpu_counter_destroy(&mp->m_fdblocks); - - mutex_destroy(&mp->m_icsb_mutex); -} - -STATIC void -xfs_icsb_lock_cntr( - xfs_icsb_cnts_t *icsbp) -{ - while (test_and_set_bit(XFS_ICSB_FLAG_LOCK, &icsbp->icsb_flags)) { - ndelay(1000); - } -} - -STATIC void -xfs_icsb_unlock_cntr( - xfs_icsb_cnts_t *icsbp) -{ - clear_bit(XFS_ICSB_FLAG_LOCK, &icsbp->icsb_flags); -} - - -STATIC void -xfs_icsb_lock_all_counters( - xfs_mount_t *mp) -{ - xfs_icsb_cnts_t *cntp; - int i; - - for_each_online_cpu(i) { - cntp = (xfs_icsb_cnts_t *)per_cpu_ptr(mp->m_sb_cnts, i); - xfs_icsb_lock_cntr(cntp); - } -} - -STATIC void -xfs_icsb_unlock_all_counters( - xfs_mount_t *mp) -{ - xfs_icsb_cnts_t *cntp; - int i; - - for_each_online_cpu(i) { - cntp = (xfs_icsb_cnts_t *)per_cpu_ptr(mp->m_sb_cnts, i); - xfs_icsb_unlock_cntr(cntp); - } -} - -STATIC void -xfs_icsb_count( - xfs_mount_t *mp, - xfs_icsb_cnts_t *cnt, - int flags) -{ - memset(cnt, 0, sizeof(xfs_icsb_cnts_t)); - - if (!(flags & XFS_ICSB_LAZY_COUNT)) - xfs_icsb_lock_all_counters(mp); - - - if (!(flags & XFS_ICSB_LAZY_COUNT)) - xfs_icsb_unlock_all_counters(mp); -} - -STATIC int -xfs_icsb_counter_disabled( - xfs_mount_t *mp, - xfs_sb_field_t field) -{ - return test_bit(field, &mp->m_icsb_counters); -} - -STATIC void -xfs_icsb_disable_counter( - xfs_mount_t *mp, - xfs_sb_field_t field) -{ - xfs_icsb_cnts_t cnt; - - /* - * If we are already disabled, then there is nothing to do - * here. We check before locking all the counters to avoid - * the expensive lock operation when being called in the - * slow path and the counter is already disabled. This is - * safe because the only time we set or clear this state is under - * the m_icsb_mutex. - */ - if (xfs_icsb_counter_disabled(mp, field)) - return; - - xfs_icsb_lock_all_counters(mp); - if (!test_and_set_bit(field, &mp->m_icsb_counters)) { - /* drain back to superblock */ - - xfs_icsb_count(mp, &cnt, XFS_ICSB_LAZY_COUNT); - switch(field) { - default: - BUG(); - } - } - - xfs_icsb_unlock_all_counters(mp); -} - -STATIC void -xfs_icsb_enable_counter( - xfs_mount_t *mp, - xfs_sb_field_t field, - uint64_t count, - uint64_t resid) -{ - int i; - - xfs_icsb_lock_all_counters(mp); - for_each_online_cpu(i) { - switch (field) { - default: - BUG(); - break; - } - resid = 0; - } - clear_bit(field, &mp->m_icsb_counters); - xfs_icsb_unlock_all_counters(mp); -} - -void -xfs_icsb_sync_counters_locked( - xfs_mount_t *mp, - int flags) -{ - xfs_icsb_cnts_t cnt; - - xfs_icsb_count(mp, &cnt, flags); -} - -/* - * Accurate update of per-cpu counters to incore superblock - */ -void -xfs_icsb_sync_counters( - xfs_mount_t *mp, - int flags) -{ - spin_lock(&mp->m_sb_lock); - xfs_icsb_sync_counters_locked(mp, flags); - spin_unlock(&mp->m_sb_lock); -} - -/* - * Balance and enable/disable counters as necessary. - * - * Thresholds for re-enabling counters are somewhat magic. inode counts are - * chosen to be the same number as single on disk allocation chunk per CPU, and - * free blocks is something far enough zero that we aren't going thrash when we - * get near ENOSPC. We also need to supply a minimum we require per cpu to - * prevent looping endlessly when xfs_alloc_space asks for more than will - * be distributed to a single CPU but each CPU has enough blocks to be - * reenabled. - * - * Note that we can be called when counters are already disabled. - * xfs_icsb_disable_counter() optimises the counter locking in this case to - * prevent locking every per-cpu counter needlessly. - */ - -#define XFS_ICSB_INO_CNTR_REENABLE (uint64_t)64 -#define XFS_ICSB_FDBLK_CNTR_REENABLE(mp) \ - (uint64_t)(512 + XFS_ALLOC_SET_ASIDE(mp)) -STATIC void -xfs_icsb_balance_counter_locked( - xfs_mount_t *mp, - xfs_sb_field_t field, - int min_per_cpu) -{ - uint64_t count, resid; - - /* disable counter and sync counter */ - xfs_icsb_disable_counter(mp, field); - - /* update counters - first CPU gets residual*/ - switch (field) { - default: - BUG(); - count = resid = 0; /* quiet, gcc */ - break; - } - - xfs_icsb_enable_counter(mp, field, count, resid); -} - -STATIC void -xfs_icsb_balance_counter( - xfs_mount_t *mp, - xfs_sb_field_t fields, - int min_per_cpu) -{ - spin_lock(&mp->m_sb_lock); - xfs_icsb_balance_counter_locked(mp, fields, min_per_cpu); - spin_unlock(&mp->m_sb_lock); -} - -int -xfs_icsb_modify_counters( - xfs_mount_t *mp, - xfs_sb_field_t field, - int64_t delta, - int rsvd) -{ - xfs_icsb_cnts_t *icsbp; - int ret = 0; - - might_sleep(); -again: - preempt_disable(); - icsbp = this_cpu_ptr(mp->m_sb_cnts); - - /* - * if the counter is disabled, go to slow path - */ - if (unlikely(xfs_icsb_counter_disabled(mp, field))) - goto slow_path; - xfs_icsb_lock_cntr(icsbp); - if (unlikely(xfs_icsb_counter_disabled(mp, field))) { - xfs_icsb_unlock_cntr(icsbp); - goto slow_path; - } - - switch (field) { - default: - BUG(); - goto balance_counter; /* be still, gcc */ - } - xfs_icsb_unlock_cntr(icsbp); - preempt_enable(); - return 0; - -slow_path: - preempt_enable(); - - /* - * serialise with a mutex so we don't burn lots of cpu on - * the superblock lock. We still need to hold the superblock - * lock, however, when we modify the global structures. - */ - xfs_icsb_lock(mp); - - /* - * Now running atomically. - * - * If the counter is enabled, someone has beaten us to rebalancing. - * Drop the lock and try again in the fast path.... - */ - if (!(xfs_icsb_counter_disabled(mp, field))) { - xfs_icsb_unlock(mp); - goto again; - } - - /* - * The counter is currently disabled. Because we are - * running atomically here, we know a rebalance cannot - * be in progress. Hence we can go straight to operating - * on the global superblock. We do not call xfs_mod_incore_sb() - * here even though we need to get the m_sb_lock. Doing so - * will cause us to re-enter this function and deadlock. - * Hence we get the m_sb_lock ourselves and then call - * xfs_mod_incore_sb_unlocked() as the unlocked path operates - * directly on the global counters. - */ - spin_lock(&mp->m_sb_lock); - ret = xfs_mod_incore_sb_unlocked(mp, field, delta, rsvd); - spin_unlock(&mp->m_sb_lock); - - /* - * Now that we've modified the global superblock, we - * may be able to re-enable the distributed counters - * (e.g. lots of space just got freed). After that - * we are done. - */ - if (ret != -ENOSPC) - xfs_icsb_balance_counter(mp, field, 0); - xfs_icsb_unlock(mp); - return ret; - -balance_counter: - xfs_icsb_unlock_cntr(icsbp); - preempt_enable(); - - /* - * We may have multiple threads here if multiple per-cpu - * counters run dry at the same time. This will mean we can - * do more balances than strictly necessary but it is not - * the common slowpath case. - */ - xfs_icsb_lock(mp); - - /* - * running atomically. - * - * This will leave the counter in the correct state for future - * accesses. After the rebalance, we simply try again and our retry - * will either succeed through the fast path or slow path without - * another balance operation being required. - */ - xfs_icsb_balance_counter(mp, field, delta); - xfs_icsb_unlock(mp); - goto again; -} - -#endif diff --git a/fs/xfs/xfs_mount.h b/fs/xfs/xfs_mount.h index 84b745fbc787..205f23a240a7 100644 --- a/fs/xfs/xfs_mount.h +++ b/fs/xfs/xfs_mount.h @@ -18,8 +18,6 @@ #ifndef __XFS_MOUNT_H__ #define __XFS_MOUNT_H__ -#ifdef __KERNEL__ - struct xlog; struct xfs_inode; struct xfs_mru_cache; @@ -29,43 +27,6 @@ struct xfs_quotainfo; struct xfs_dir_ops; struct xfs_da_geometry; -#ifdef HAVE_PERCPU_SB - -/* - * Valid per-cpu incore superblock counters. Note that if you add new counters, - * you may need to define new counter disabled bit field descriptors as there - * are more possible fields in the superblock that can fit in a bitfield on a - * 32 bit platform. The XFS_SBS_* values for the current current counters just - * fit. - */ -typedef struct xfs_icsb_cnts { - uint64_t icsb_fdblocks; - uint64_t icsb_ifree; - unsigned long icsb_flags; -} xfs_icsb_cnts_t; - -#define XFS_ICSB_FLAG_LOCK (1 << 0) /* counter lock bit */ - -#define XFS_ICSB_LAZY_COUNT (1 << 1) /* accuracy not needed */ - -extern int xfs_icsb_init_counters(struct xfs_mount *); -extern void xfs_icsb_reinit_counters(struct xfs_mount *); -extern void xfs_icsb_destroy_counters(struct xfs_mount *); -extern void xfs_icsb_sync_counters(struct xfs_mount *, int); -extern void xfs_icsb_sync_counters_locked(struct xfs_mount *, int); -extern int xfs_icsb_modify_counters(struct xfs_mount *, xfs_sb_field_t, - int64_t, int); - -#else -#define xfs_icsb_init_counters(mp) (0) -#define xfs_icsb_destroy_counters(mp) do { } while (0) -#define xfs_icsb_reinit_counters(mp) do { } while (0) -#define xfs_icsb_sync_counters(mp, flags) do { } while (0) -#define xfs_icsb_sync_counters_locked(mp, flags) do { } while (0) -#define xfs_icsb_modify_counters(mp, field, delta, rsvd) \ - xfs_mod_incore_sb(mp, field, delta, rsvd) -#endif - /* dynamic preallocation free space thresholds, 5% down to 1% */ enum { XFS_LOWSP_1_PCNT = 0, @@ -156,12 +117,6 @@ typedef struct xfs_mount { const struct xfs_dir_ops *m_nondir_inode_ops; /* !dir inode ops */ uint m_chsize; /* size of next field */ atomic_t m_active_trans; /* number trans frozen */ -#ifdef HAVE_PERCPU_SB - xfs_icsb_cnts_t __percpu *m_sb_cnts; /* per-cpu superblock counters */ - unsigned long m_icsb_counters; /* disabled per-cpu counters */ - struct notifier_block m_icsb_notifier; /* hotplug cpu notifier */ - struct mutex m_icsb_mutex; /* balancer sync lock */ -#endif struct xfs_mru_cache *m_filestream; /* per-mount filestream data */ struct delayed_work m_reclaim_work; /* background inode reclaim */ struct delayed_work m_eofblocks_work; /* background eof blocks @@ -304,26 +259,6 @@ xfs_daddr_to_agbno(struct xfs_mount *mp, xfs_daddr_t d) return (xfs_agblock_t) do_div(ld, mp->m_sb.sb_agblocks); } -/* - * Per-cpu superblock locking functions - */ -#ifdef HAVE_PERCPU_SB -static inline void -xfs_icsb_lock(xfs_mount_t *mp) -{ - mutex_lock(&mp->m_icsb_mutex); -} - -static inline void -xfs_icsb_unlock(xfs_mount_t *mp) -{ - mutex_unlock(&mp->m_icsb_mutex); -} -#else -#define xfs_icsb_lock(mp) -#define xfs_icsb_unlock(mp) -#endif - /* * This structure is for use by the xfs_mod_incore_sb_batch() routine. * xfs_growfs can specify a few fields which are more than int limit @@ -407,6 +342,4 @@ extern int xfs_dev_is_read_only(struct xfs_mount *, char *); extern void xfs_set_low_space_thresholds(struct xfs_mount *); -#endif /* __KERNEL__ */ - #endif /* __XFS_MOUNT_H__ */ diff --git a/fs/xfs/xfs_super.c b/fs/xfs/xfs_super.c index 9ec75074026d..53c56a913778 100644 --- a/fs/xfs/xfs_super.c +++ b/fs/xfs/xfs_super.c @@ -1033,23 +1033,6 @@ xfs_free_fsname( kfree(mp->m_logname); } -STATIC void -xfs_fs_put_super( - struct super_block *sb) -{ - struct xfs_mount *mp = XFS_M(sb); - - xfs_filestream_unmount(mp); - xfs_unmountfs(mp); - - xfs_freesb(mp); - xfs_icsb_destroy_counters(mp); - xfs_destroy_mount_workqueues(mp); - xfs_close_devices(mp); - xfs_free_fsname(mp); - kfree(mp); -} - STATIC int xfs_fs_sync_fs( struct super_block *sb, @@ -1098,7 +1081,6 @@ xfs_fs_statfs( statp->f_fsid.val[0] = (u32)id; statp->f_fsid.val[1] = (u32)(id >> 32); - xfs_icsb_sync_counters(mp, XFS_ICSB_LAZY_COUNT); icount = percpu_counter_sum(&mp->m_icount); ifree = percpu_counter_sum(&mp->m_ifree); fdblocks = percpu_counter_sum(&mp->m_fdblocks); @@ -1408,6 +1390,51 @@ xfs_finish_flags( return 0; } +static int +xfs_init_percpu_counters( + struct xfs_mount *mp) +{ + int error; + + error = percpu_counter_init(&mp->m_icount, 0, GFP_KERNEL); + if (error) + return ENOMEM; + + error = percpu_counter_init(&mp->m_ifree, 0, GFP_KERNEL); + if (error) + goto free_icount; + + error = percpu_counter_init(&mp->m_fdblocks, 0, GFP_KERNEL); + if (error) + goto free_ifree; + + return 0; + +free_ifree: + percpu_counter_destroy(&mp->m_ifree); +free_icount: + percpu_counter_destroy(&mp->m_icount); + return -ENOMEM; +} + +void +xfs_reinit_percpu_counters( + struct xfs_mount *mp) +{ + percpu_counter_set(&mp->m_icount, mp->m_sb.sb_icount); + percpu_counter_set(&mp->m_ifree, mp->m_sb.sb_ifree); + percpu_counter_set(&mp->m_fdblocks, mp->m_sb.sb_fdblocks); +} + +static void +xfs_destroy_percpu_counters( + struct xfs_mount *mp) +{ + percpu_counter_destroy(&mp->m_icount); + percpu_counter_destroy(&mp->m_ifree); + percpu_counter_destroy(&mp->m_fdblocks); +} + STATIC int xfs_fs_fill_super( struct super_block *sb, @@ -1456,7 +1483,7 @@ xfs_fs_fill_super( if (error) goto out_close_devices; - error = xfs_icsb_init_counters(mp); + error = xfs_init_percpu_counters(mp); if (error) goto out_destroy_workqueues; @@ -1514,7 +1541,7 @@ xfs_fs_fill_super( out_free_sb: xfs_freesb(mp); out_destroy_counters: - xfs_icsb_destroy_counters(mp); + xfs_destroy_percpu_counters(mp); out_destroy_workqueues: xfs_destroy_mount_workqueues(mp); out_close_devices: @@ -1531,6 +1558,23 @@ out_destroy_workqueues: goto out_free_sb; } +STATIC void +xfs_fs_put_super( + struct super_block *sb) +{ + struct xfs_mount *mp = XFS_M(sb); + + xfs_filestream_unmount(mp); + xfs_unmountfs(mp); + + xfs_freesb(mp); + xfs_destroy_percpu_counters(mp); + xfs_destroy_mount_workqueues(mp); + xfs_close_devices(mp); + xfs_free_fsname(mp); + kfree(mp); +} + STATIC struct dentry * xfs_fs_mount( struct file_system_type *fs_type, diff --git a/fs/xfs/xfs_super.h b/fs/xfs/xfs_super.h index 2b830c2f322e..499058fea303 100644 --- a/fs/xfs/xfs_super.h +++ b/fs/xfs/xfs_super.h @@ -72,6 +72,8 @@ extern const struct export_operations xfs_export_operations; extern const struct xattr_handler *xfs_xattr_handlers[]; extern const struct quotactl_ops xfs_quotactl_operations; +extern void xfs_reinit_percpu_counters(struct xfs_mount *mp); + #define XFS_M(sb) ((struct xfs_mount *)((sb)->s_fs_info)) #endif /* __XFS_SUPER_H__ */ -- cgit v1.2.3-59-g8ed1b From bab98bbe6e1e38bf2fa5018a195608553095f51b Mon Sep 17 00:00:00 2001 From: Dave Chinner Date: Mon, 23 Feb 2015 21:22:54 +1100 Subject: xfs: introduce xfs_mod_frextents Add a new helper to modify the incore counter of free realtime extents. This matches the helpers used for inode and data block counters, and removes a significant users of the xfs_mod_incore_sb() interface. Based on a patch originally from Christoph Hellwig. Signed-off-by: Dave Chinner Reviewed-by: Brian Foster Signed-off-by: Dave Chinner --- fs/xfs/libxfs/xfs_bmap.c | 8 +++----- fs/xfs/xfs_mount.c | 27 +++++++++++++++++++-------- fs/xfs/xfs_mount.h | 2 ++ fs/xfs/xfs_trans.c | 18 ++++++++++-------- 4 files changed, 34 insertions(+), 21 deletions(-) diff --git a/fs/xfs/libxfs/xfs_bmap.c b/fs/xfs/libxfs/xfs_bmap.c index e39c9e83670e..b8e97fd0bac1 100644 --- a/fs/xfs/libxfs/xfs_bmap.c +++ b/fs/xfs/libxfs/xfs_bmap.c @@ -4158,8 +4158,7 @@ xfs_bmapi_reserve_delalloc( ASSERT(indlen > 0); if (rt) { - error = xfs_mod_incore_sb(mp, XFS_SBS_FREXTENTS, - -((int64_t)extsz), 0); + error = xfs_mod_frextents(mp, -((int64_t)extsz)); } else { error = xfs_mod_fdblocks(mp, -((int64_t)alen), false); } @@ -4194,7 +4193,7 @@ xfs_bmapi_reserve_delalloc( out_unreserve_blocks: if (rt) - xfs_mod_incore_sb(mp, XFS_SBS_FREXTENTS, extsz, 0); + xfs_mod_frextents(mp, extsz); else xfs_mod_fdblocks(mp, alen, false); out_unreserve_quota: @@ -5278,8 +5277,7 @@ xfs_bunmapi( rtexts = XFS_FSB_TO_B(mp, del.br_blockcount); do_div(rtexts, mp->m_sb.sb_rextsize); - xfs_mod_incore_sb(mp, XFS_SBS_FREXTENTS, - (int64_t)rtexts, 0); + xfs_mod_frextents(mp, (int64_t)rtexts); (void)xfs_trans_reserve_quota_nblks(NULL, ip, -((long)del.br_blockcount), 0, XFS_QMOPT_RES_RTBLKS); diff --git a/fs/xfs/xfs_mount.c b/fs/xfs/xfs_mount.c index 05b392e35e35..df4c32fdc706 100644 --- a/fs/xfs/xfs_mount.c +++ b/fs/xfs/xfs_mount.c @@ -1198,6 +1198,24 @@ fdblocks_enospc: return -ENOSPC; } +int +xfs_mod_frextents( + struct xfs_mount *mp, + int64_t delta) +{ + int64_t lcounter; + int ret = 0; + + spin_lock(&mp->m_sb_lock); + lcounter = mp->m_sb.sb_frextents + delta; + if (lcounter < 0) + ret = -ENOSPC; + else + mp->m_sb.sb_frextents = lcounter; + spin_unlock(&mp->m_sb_lock); + return ret; +} + /* * xfs_mod_incore_sb_unlocked() is a utility routine commonly used to apply * a delta to a specified field in the in-core superblock. Simply @@ -1227,16 +1245,9 @@ xfs_mod_incore_sb_unlocked( case XFS_SBS_ICOUNT: case XFS_SBS_IFREE: case XFS_SBS_FDBLOCKS: + case XFS_SBS_FREXTENTS: ASSERT(0); return -EINVAL; - case XFS_SBS_FREXTENTS: - lcounter = (long long)mp->m_sb.sb_frextents; - lcounter += delta; - if (lcounter < 0) { - return -ENOSPC; - } - mp->m_sb.sb_frextents = lcounter; - return 0; case XFS_SBS_DBLOCKS: lcounter = (long long)mp->m_sb.sb_dblocks; lcounter += delta; diff --git a/fs/xfs/xfs_mount.h b/fs/xfs/xfs_mount.h index 205f23a240a7..d65e0f23b847 100644 --- a/fs/xfs/xfs_mount.h +++ b/fs/xfs/xfs_mount.h @@ -331,6 +331,8 @@ 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); + extern int xfs_mount_log_sb(xfs_mount_t *); extern struct xfs_buf *xfs_getsb(xfs_mount_t *, int); extern int xfs_readsb(xfs_mount_t *, int); diff --git a/fs/xfs/xfs_trans.c b/fs/xfs/xfs_trans.c index e99f5e552c64..4e4bc5aed6b6 100644 --- a/fs/xfs/xfs_trans.c +++ b/fs/xfs/xfs_trans.c @@ -235,8 +235,7 @@ xfs_trans_reserve( * fail if the count would go below zero. */ if (rtextents > 0) { - error = xfs_mod_incore_sb(tp->t_mountp, XFS_SBS_FREXTENTS, - -((int64_t)rtextents), rsvd); + error = xfs_mod_frextents(tp->t_mountp, -((int64_t)rtextents)); if (error) { error = -ENOSPC; goto undo_log; @@ -562,10 +561,10 @@ xfs_trans_unreserve_and_mod_sb( } /* apply remaining deltas */ - if (rtxdelta != 0) { - msbp->msb_field = XFS_SBS_FREXTENTS; - msbp->msb_delta = rtxdelta; - msbp++; + if (rtxdelta) { + error = xfs_mod_frextents(mp, rtxdelta); + if (error) + goto out_undo_ifree; } if (tp->t_flags & XFS_TRANS_SB_DIRTY) { @@ -618,12 +617,15 @@ xfs_trans_unreserve_and_mod_sb( error = xfs_mod_incore_sb_batch(tp->t_mountp, msb, (uint)(msbp - msb), rsvd); if (error) - goto out_undo_ifreecount; + goto out_undo_frextents; } return; -out_undo_ifreecount: +out_undo_frextents: + if (rtxdelta) + xfs_mod_frextents(mp, -rtxdelta); +out_undo_ifree: if (ifreedelta) xfs_mod_ifree(mp, -ifreedelta); out_undo_icount: -- cgit v1.2.3-59-g8ed1b From 0bd5ddedccca4451ac2390d1155b4ab74b990eff Mon Sep 17 00:00:00 2001 From: Dave Chinner Date: Mon, 23 Feb 2015 21:24:11 +1100 Subject: xfs: replace xfs_mod_incore_sb_batched Introduce helper functions for modifying fields in the superblock into xfs_trans.c, the only caller of xfs_mod_incore_sb_batch(). We can then use these directly in xfs_trans_unreserve_and_mod_sb() and so remove another user of the xfs_mode_incore_sb() API without losing any functionality or scalability of the transaction commit code.. Based on a patch from Christoph Hellwig. Signed-off-by: Dave Chinner Reviewed-by: Brian Foster Signed-off-by: Dave Chinner --- fs/xfs/xfs_mount.c | 51 -------------- fs/xfs/xfs_mount.h | 11 --- fs/xfs/xfs_trans.c | 198 ++++++++++++++++++++++++++++++++++------------------- 3 files changed, 126 insertions(+), 134 deletions(-) diff --git a/fs/xfs/xfs_mount.c b/fs/xfs/xfs_mount.c index df4c32fdc706..d748aa73003b 100644 --- a/fs/xfs/xfs_mount.c +++ b/fs/xfs/xfs_mount.c @@ -1352,57 +1352,6 @@ xfs_mod_incore_sb( return status; } -/* - * Change more than one field in the in-core superblock structure at a time. - * - * The fields and changes to those fields are specified in the array of - * xfs_mod_sb structures passed in. Either all of the specified deltas - * will be applied or none of them will. If any modified field dips below 0, - * then all modifications will be backed out and EINVAL will be returned. - * - * Note that this function may not be used for the superblock values that - * are tracked with the in-memory per-cpu counters - a direct call to - * xfs_mod_incore_sb is required for these. - */ -int -xfs_mod_incore_sb_batch( - struct xfs_mount *mp, - xfs_mod_sb_t *msb, - uint nmsb, - int rsvd) -{ - xfs_mod_sb_t *msbp; - int error = 0; - - /* - * Loop through the array of mod structures and apply each individually. - * If any fail, then back out all those which have already been applied. - * Do all of this within the scope of the m_sb_lock so that all of the - * changes will be atomic. - */ - spin_lock(&mp->m_sb_lock); - for (msbp = msb; msbp < (msb + nmsb); msbp++) { - ASSERT(msbp->msb_field < XFS_SBS_ICOUNT || - msbp->msb_field > XFS_SBS_FDBLOCKS); - - error = xfs_mod_incore_sb_unlocked(mp, msbp->msb_field, - msbp->msb_delta, rsvd); - if (error) - goto unwind; - } - spin_unlock(&mp->m_sb_lock); - return 0; - -unwind: - while (--msbp >= msb) { - error = xfs_mod_incore_sb_unlocked(mp, msbp->msb_field, - -msbp->msb_delta, rsvd); - ASSERT(error == 0); - } - spin_unlock(&mp->m_sb_lock); - return error; -} - /* * xfs_getsb() is called to obtain the buffer for the superblock. * The buffer is returned locked and read in from disk. diff --git a/fs/xfs/xfs_mount.h b/fs/xfs/xfs_mount.h index d65e0f23b847..1c11512bab83 100644 --- a/fs/xfs/xfs_mount.h +++ b/fs/xfs/xfs_mount.h @@ -259,15 +259,6 @@ xfs_daddr_to_agbno(struct xfs_mount *mp, xfs_daddr_t d) return (xfs_agblock_t) do_div(ld, mp->m_sb.sb_agblocks); } -/* - * This structure is for use by the xfs_mod_incore_sb_batch() routine. - * xfs_growfs can specify a few fields which are more than int limit - */ -typedef struct xfs_mod_sb { - xfs_sb_field_t msb_field; /* Field to modify, see below */ - int64_t msb_delta; /* Change to make to specified field */ -} xfs_mod_sb_t; - /* * Per-ag incore structure, copies of information in agf and agi, to improve the * performance of allocation group selection. @@ -325,8 +316,6 @@ extern int xfs_initialize_perag(xfs_mount_t *mp, xfs_agnumber_t agcount, extern void xfs_unmountfs(xfs_mount_t *); extern int xfs_mod_incore_sb(xfs_mount_t *, xfs_sb_field_t, int64_t, int); -extern int xfs_mod_incore_sb_batch(xfs_mount_t *, xfs_mod_sb_t *, - uint, int); 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, diff --git a/fs/xfs/xfs_trans.c b/fs/xfs/xfs_trans.c index 4e4bc5aed6b6..220ef2c906b2 100644 --- a/fs/xfs/xfs_trans.c +++ b/fs/xfs/xfs_trans.c @@ -485,6 +485,54 @@ xfs_trans_apply_sb_deltas( sizeof(sbp->sb_frextents) - 1); } +STATIC int +xfs_sb_mod8( + uint8_t *field, + int8_t delta) +{ + int8_t counter = *field; + + counter += delta; + if (counter < 0) { + ASSERT(0); + return -EINVAL; + } + *field = counter; + return 0; +} + +STATIC int +xfs_sb_mod32( + uint32_t *field, + int32_t delta) +{ + int32_t counter = *field; + + counter += delta; + if (counter < 0) { + ASSERT(0); + return -EINVAL; + } + *field = counter; + return 0; +} + +STATIC int +xfs_sb_mod64( + uint64_t *field, + int64_t delta) +{ + int64_t counter = *field; + + counter += delta; + if (counter < 0) { + ASSERT(0); + return -EINVAL; + } + *field = counter; + return 0; +} + /* * xfs_trans_unreserve_and_mod_sb() is called to release unused reservations * and apply superblock counter changes to the in-core superblock. The @@ -492,13 +540,6 @@ xfs_trans_apply_sb_deltas( * applied to the in-core superblock. The idea is that that has already been * done. * - * This is done efficiently with a single call to xfs_mod_incore_sb_batch(). - * However, we have to ensure that we only modify each superblock field only - * once because the application of the delta values may not be atomic. That can - * lead to ENOSPC races occurring if we have two separate modifcations of the - * free space counter to put back the entire reservation and then take away - * what we used. - * * If we are not logging superblock counters, then the inode allocated/free and * 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 @@ -506,20 +547,15 @@ xfs_trans_apply_sb_deltas( */ void xfs_trans_unreserve_and_mod_sb( - xfs_trans_t *tp) + struct xfs_trans *tp) { - xfs_mod_sb_t msb[9]; /* If you add cases, add entries */ - xfs_mod_sb_t *msbp; - xfs_mount_t *mp = tp->t_mountp; - /* REFERENCED */ - int error; - bool rsvd = (tp->t_flags & XFS_TRANS_RESERVE) != 0; - int64_t blkdelta = 0; - int64_t rtxdelta = 0; - int64_t idelta = 0; - int64_t ifreedelta = 0; - - msbp = msb; + struct xfs_mount *mp = tp->t_mountp; + bool rsvd = (tp->t_flags & XFS_TRANS_RESERVE) != 0; + int64_t blkdelta = 0; + int64_t rtxdelta = 0; + int64_t idelta = 0; + int64_t ifreedelta = 0; + int error; /* calculate deltas */ if (tp->t_blk_res > 0) @@ -560,72 +596,90 @@ xfs_trans_unreserve_and_mod_sb( goto out_undo_icount; } + if (rtxdelta == 0 && !(tp->t_flags & XFS_TRANS_SB_DIRTY)) + return; + /* apply remaining deltas */ + spin_lock(&mp->m_sb_lock); if (rtxdelta) { - error = xfs_mod_frextents(mp, rtxdelta); + error = xfs_sb_mod64(&mp->m_sb.sb_frextents, rtxdelta); if (error) goto out_undo_ifree; } - if (tp->t_flags & XFS_TRANS_SB_DIRTY) { - if (tp->t_dblocks_delta != 0) { - msbp->msb_field = XFS_SBS_DBLOCKS; - msbp->msb_delta = tp->t_dblocks_delta; - msbp++; - } - if (tp->t_agcount_delta != 0) { - msbp->msb_field = XFS_SBS_AGCOUNT; - msbp->msb_delta = tp->t_agcount_delta; - msbp++; - } - if (tp->t_imaxpct_delta != 0) { - msbp->msb_field = XFS_SBS_IMAX_PCT; - msbp->msb_delta = tp->t_imaxpct_delta; - msbp++; - } - if (tp->t_rextsize_delta != 0) { - msbp->msb_field = XFS_SBS_REXTSIZE; - msbp->msb_delta = tp->t_rextsize_delta; - msbp++; - } - if (tp->t_rbmblocks_delta != 0) { - msbp->msb_field = XFS_SBS_RBMBLOCKS; - msbp->msb_delta = tp->t_rbmblocks_delta; - msbp++; - } - if (tp->t_rblocks_delta != 0) { - msbp->msb_field = XFS_SBS_RBLOCKS; - msbp->msb_delta = tp->t_rblocks_delta; - msbp++; - } - if (tp->t_rextents_delta != 0) { - msbp->msb_field = XFS_SBS_REXTENTS; - msbp->msb_delta = tp->t_rextents_delta; - msbp++; - } - if (tp->t_rextslog_delta != 0) { - msbp->msb_field = XFS_SBS_REXTSLOG; - msbp->msb_delta = tp->t_rextslog_delta; - msbp++; - } - } - - /* - * If we need to change anything, do it. - */ - if (msbp > msb) { - error = xfs_mod_incore_sb_batch(tp->t_mountp, msb, - (uint)(msbp - msb), rsvd); + if (tp->t_dblocks_delta != 0) { + error = xfs_sb_mod64(&mp->m_sb.sb_dblocks, tp->t_dblocks_delta); if (error) goto out_undo_frextents; } - + if (tp->t_agcount_delta != 0) { + error = xfs_sb_mod32(&mp->m_sb.sb_agcount, tp->t_agcount_delta); + if (error) + goto out_undo_dblocks; + } + if (tp->t_imaxpct_delta != 0) { + error = xfs_sb_mod8(&mp->m_sb.sb_imax_pct, tp->t_imaxpct_delta); + if (error) + goto out_undo_agcount; + } + if (tp->t_rextsize_delta != 0) { + error = xfs_sb_mod32(&mp->m_sb.sb_rextsize, + tp->t_rextsize_delta); + if (error) + goto out_undo_imaxpct; + } + if (tp->t_rbmblocks_delta != 0) { + error = xfs_sb_mod32(&mp->m_sb.sb_rbmblocks, + tp->t_rbmblocks_delta); + if (error) + goto out_undo_rextsize; + } + if (tp->t_rblocks_delta != 0) { + error = xfs_sb_mod64(&mp->m_sb.sb_rblocks, tp->t_rblocks_delta); + if (error) + goto out_undo_rbmblocks; + } + if (tp->t_rextents_delta != 0) { + error = xfs_sb_mod64(&mp->m_sb.sb_rextents, + tp->t_rextents_delta); + if (error) + goto out_undo_rblocks; + } + if (tp->t_rextslog_delta != 0) { + error = xfs_sb_mod8(&mp->m_sb.sb_rextslog, + tp->t_rextslog_delta); + if (error) + goto out_undo_rextents; + } + spin_unlock(&mp->m_sb_lock); return; +out_undo_rextents: + if (tp->t_rextents_delta) + xfs_sb_mod64(&mp->m_sb.sb_rextents, -tp->t_rextents_delta); +out_undo_rblocks: + if (tp->t_rblocks_delta) + xfs_sb_mod64(&mp->m_sb.sb_rblocks, -tp->t_rblocks_delta); +out_undo_rbmblocks: + if (tp->t_rbmblocks_delta) + xfs_sb_mod32(&mp->m_sb.sb_rbmblocks, -tp->t_rbmblocks_delta); +out_undo_rextsize: + if (tp->t_rextsize_delta) + xfs_sb_mod32(&mp->m_sb.sb_rextsize, -tp->t_rextsize_delta); +out_undo_imaxpct: + if (tp->t_rextsize_delta) + xfs_sb_mod8(&mp->m_sb.sb_imax_pct, -tp->t_imaxpct_delta); +out_undo_agcount: + if (tp->t_agcount_delta) + xfs_sb_mod32(&mp->m_sb.sb_agcount, -tp->t_agcount_delta); +out_undo_dblocks: + if (tp->t_dblocks_delta) + xfs_sb_mod64(&mp->m_sb.sb_dblocks, -tp->t_dblocks_delta); out_undo_frextents: if (rtxdelta) - xfs_mod_frextents(mp, -rtxdelta); + xfs_sb_mod64(&mp->m_sb.sb_frextents, -rtxdelta); out_undo_ifree: + spin_unlock(&mp->m_sb_lock); if (ifreedelta) xfs_mod_ifree(mp, -ifreedelta); out_undo_icount: -- cgit v1.2.3-59-g8ed1b From 964aa8d9e4d36e0e54a88683d14c7d5b9d02aed8 Mon Sep 17 00:00:00 2001 From: Dave Chinner Date: Mon, 23 Feb 2015 21:24:37 +1100 Subject: xfs: remove xfs_mod_incore_sb API Now that there are no users of the bitfield based incore superblock modification API, just remove the whole damn lot of it, including all the bitfield definitions. This finally removes a lot of cruft that has been around for a long time. Credit goes to Christoph Hellwig for providing a great patch connecting all the dots to enale us to do this. This patch is derived from that work. Signed-off-by: Dave Chinner Reviewed-by: Brian Foster Signed-off-by: Dave Chinner --- fs/xfs/libxfs/xfs_format.h | 62 --------------------- fs/xfs/xfs_fsops.c | 4 -- fs/xfs/xfs_mount.c | 136 --------------------------------------------- fs/xfs/xfs_mount.h | 3 +- 4 files changed, 1 insertion(+), 204 deletions(-) diff --git a/fs/xfs/libxfs/xfs_format.h b/fs/xfs/libxfs/xfs_format.h index 8eb718979383..4daaa662337b 100644 --- a/fs/xfs/libxfs/xfs_format.h +++ b/fs/xfs/libxfs/xfs_format.h @@ -264,68 +264,6 @@ typedef struct xfs_dsb { /* must be padded to 64 bit alignment */ } xfs_dsb_t; -/* - * Sequence number values for the fields. - */ -typedef enum { - XFS_SBS_MAGICNUM, XFS_SBS_BLOCKSIZE, XFS_SBS_DBLOCKS, XFS_SBS_RBLOCKS, - XFS_SBS_REXTENTS, XFS_SBS_UUID, XFS_SBS_LOGSTART, XFS_SBS_ROOTINO, - XFS_SBS_RBMINO, XFS_SBS_RSUMINO, XFS_SBS_REXTSIZE, XFS_SBS_AGBLOCKS, - XFS_SBS_AGCOUNT, XFS_SBS_RBMBLOCKS, XFS_SBS_LOGBLOCKS, - XFS_SBS_VERSIONNUM, XFS_SBS_SECTSIZE, XFS_SBS_INODESIZE, - XFS_SBS_INOPBLOCK, XFS_SBS_FNAME, XFS_SBS_BLOCKLOG, - XFS_SBS_SECTLOG, XFS_SBS_INODELOG, XFS_SBS_INOPBLOG, XFS_SBS_AGBLKLOG, - XFS_SBS_REXTSLOG, XFS_SBS_INPROGRESS, XFS_SBS_IMAX_PCT, XFS_SBS_ICOUNT, - XFS_SBS_IFREE, XFS_SBS_FDBLOCKS, XFS_SBS_FREXTENTS, XFS_SBS_UQUOTINO, - XFS_SBS_GQUOTINO, XFS_SBS_QFLAGS, XFS_SBS_FLAGS, XFS_SBS_SHARED_VN, - XFS_SBS_INOALIGNMT, XFS_SBS_UNIT, XFS_SBS_WIDTH, XFS_SBS_DIRBLKLOG, - XFS_SBS_LOGSECTLOG, XFS_SBS_LOGSECTSIZE, XFS_SBS_LOGSUNIT, - XFS_SBS_FEATURES2, XFS_SBS_BAD_FEATURES2, XFS_SBS_FEATURES_COMPAT, - XFS_SBS_FEATURES_RO_COMPAT, XFS_SBS_FEATURES_INCOMPAT, - XFS_SBS_FEATURES_LOG_INCOMPAT, XFS_SBS_CRC, XFS_SBS_PAD, - XFS_SBS_PQUOTINO, XFS_SBS_LSN, - XFS_SBS_FIELDCOUNT -} xfs_sb_field_t; - -/* - * Mask values, defined based on the xfs_sb_field_t values. - * Only define the ones we're using. - */ -#define XFS_SB_MVAL(x) (1LL << XFS_SBS_ ## x) -#define XFS_SB_UUID XFS_SB_MVAL(UUID) -#define XFS_SB_FNAME XFS_SB_MVAL(FNAME) -#define XFS_SB_ROOTINO XFS_SB_MVAL(ROOTINO) -#define XFS_SB_RBMINO XFS_SB_MVAL(RBMINO) -#define XFS_SB_RSUMINO XFS_SB_MVAL(RSUMINO) -#define XFS_SB_VERSIONNUM XFS_SB_MVAL(VERSIONNUM) -#define XFS_SB_UQUOTINO XFS_SB_MVAL(UQUOTINO) -#define XFS_SB_GQUOTINO XFS_SB_MVAL(GQUOTINO) -#define XFS_SB_QFLAGS XFS_SB_MVAL(QFLAGS) -#define XFS_SB_SHARED_VN XFS_SB_MVAL(SHARED_VN) -#define XFS_SB_UNIT XFS_SB_MVAL(UNIT) -#define XFS_SB_WIDTH XFS_SB_MVAL(WIDTH) -#define XFS_SB_ICOUNT XFS_SB_MVAL(ICOUNT) -#define XFS_SB_IFREE XFS_SB_MVAL(IFREE) -#define XFS_SB_FDBLOCKS XFS_SB_MVAL(FDBLOCKS) -#define XFS_SB_FEATURES2 (XFS_SB_MVAL(FEATURES2) | \ - XFS_SB_MVAL(BAD_FEATURES2)) -#define XFS_SB_FEATURES_COMPAT XFS_SB_MVAL(FEATURES_COMPAT) -#define XFS_SB_FEATURES_RO_COMPAT XFS_SB_MVAL(FEATURES_RO_COMPAT) -#define XFS_SB_FEATURES_INCOMPAT XFS_SB_MVAL(FEATURES_INCOMPAT) -#define XFS_SB_FEATURES_LOG_INCOMPAT XFS_SB_MVAL(FEATURES_LOG_INCOMPAT) -#define XFS_SB_CRC XFS_SB_MVAL(CRC) -#define XFS_SB_PQUOTINO XFS_SB_MVAL(PQUOTINO) -#define XFS_SB_NUM_BITS ((int)XFS_SBS_FIELDCOUNT) -#define XFS_SB_ALL_BITS ((1LL << XFS_SB_NUM_BITS) - 1) -#define XFS_SB_MOD_BITS \ - (XFS_SB_UUID | XFS_SB_ROOTINO | XFS_SB_RBMINO | XFS_SB_RSUMINO | \ - XFS_SB_VERSIONNUM | XFS_SB_UQUOTINO | XFS_SB_GQUOTINO | \ - XFS_SB_QFLAGS | XFS_SB_SHARED_VN | XFS_SB_UNIT | XFS_SB_WIDTH | \ - XFS_SB_ICOUNT | XFS_SB_IFREE | XFS_SB_FDBLOCKS | XFS_SB_FEATURES2 | \ - XFS_SB_FEATURES_COMPAT | XFS_SB_FEATURES_RO_COMPAT | \ - XFS_SB_FEATURES_INCOMPAT | XFS_SB_FEATURES_LOG_INCOMPAT | \ - XFS_SB_PQUOTINO) - /* * Misc. Flags - warning - these will be cleared by xfs_repair unless diff --git a/fs/xfs/xfs_fsops.c b/fs/xfs/xfs_fsops.c index 16e62edc9dae..cb7e8a29dfb6 100644 --- a/fs/xfs/xfs_fsops.c +++ b/fs/xfs/xfs_fsops.c @@ -693,10 +693,6 @@ xfs_reserve_blocks( * what to do. This means that the amount of free space can * change while we do this, so we need to retry if we end up * trying to reserve more space than is available. - * - * We also use the xfs_mod_incore_sb() interface so that we - * don't have to care about whether per cpu counter are - * enabled, disabled or even compiled in.... */ retry: spin_lock(&mp->m_sb_lock); diff --git a/fs/xfs/xfs_mount.c b/fs/xfs/xfs_mount.c index d748aa73003b..2ce7ee3b4ec1 100644 --- a/fs/xfs/xfs_mount.c +++ b/fs/xfs/xfs_mount.c @@ -1216,142 +1216,6 @@ xfs_mod_frextents( return ret; } -/* - * xfs_mod_incore_sb_unlocked() is a utility routine commonly used to apply - * a delta to a specified field in the in-core superblock. Simply - * switch on the field indicated and apply the delta to that field. - * Fields are not allowed to dip below zero, so if the delta would - * do this do not apply it and return EINVAL. - * - * The m_sb_lock must be held when this routine is called. - */ -STATIC int -xfs_mod_incore_sb_unlocked( - xfs_mount_t *mp, - xfs_sb_field_t field, - int64_t delta, - int rsvd) -{ - int scounter; /* short counter for 32 bit fields */ - long long lcounter; /* long counter for 64 bit fields */ - - /* - * With the in-core superblock spin lock held, switch - * on the indicated field. Apply the delta to the - * proper field. If the fields value would dip below - * 0, then do not apply the delta and return EINVAL. - */ - switch (field) { - case XFS_SBS_ICOUNT: - case XFS_SBS_IFREE: - case XFS_SBS_FDBLOCKS: - case XFS_SBS_FREXTENTS: - ASSERT(0); - return -EINVAL; - case XFS_SBS_DBLOCKS: - lcounter = (long long)mp->m_sb.sb_dblocks; - lcounter += delta; - if (lcounter < 0) { - ASSERT(0); - return -EINVAL; - } - mp->m_sb.sb_dblocks = lcounter; - return 0; - case XFS_SBS_AGCOUNT: - scounter = mp->m_sb.sb_agcount; - scounter += delta; - if (scounter < 0) { - ASSERT(0); - return -EINVAL; - } - mp->m_sb.sb_agcount = scounter; - return 0; - case XFS_SBS_IMAX_PCT: - scounter = mp->m_sb.sb_imax_pct; - scounter += delta; - if (scounter < 0) { - ASSERT(0); - return -EINVAL; - } - mp->m_sb.sb_imax_pct = scounter; - return 0; - case XFS_SBS_REXTSIZE: - scounter = mp->m_sb.sb_rextsize; - scounter += delta; - if (scounter < 0) { - ASSERT(0); - return -EINVAL; - } - mp->m_sb.sb_rextsize = scounter; - return 0; - case XFS_SBS_RBMBLOCKS: - scounter = mp->m_sb.sb_rbmblocks; - scounter += delta; - if (scounter < 0) { - ASSERT(0); - return -EINVAL; - } - mp->m_sb.sb_rbmblocks = scounter; - return 0; - case XFS_SBS_RBLOCKS: - lcounter = (long long)mp->m_sb.sb_rblocks; - lcounter += delta; - if (lcounter < 0) { - ASSERT(0); - return -EINVAL; - } - mp->m_sb.sb_rblocks = lcounter; - return 0; - case XFS_SBS_REXTENTS: - lcounter = (long long)mp->m_sb.sb_rextents; - lcounter += delta; - if (lcounter < 0) { - ASSERT(0); - return -EINVAL; - } - mp->m_sb.sb_rextents = lcounter; - return 0; - case XFS_SBS_REXTSLOG: - scounter = mp->m_sb.sb_rextslog; - scounter += delta; - if (scounter < 0) { - ASSERT(0); - return -EINVAL; - } - mp->m_sb.sb_rextslog = scounter; - return 0; - default: - ASSERT(0); - return -EINVAL; - } -} - -/* - * xfs_mod_incore_sb() is used to change a field in the in-core - * superblock structure by the specified delta. This modification - * is protected by the m_sb_lock. Just use the xfs_mod_incore_sb_unlocked() - * routine to do the work. - */ -int -xfs_mod_incore_sb( - struct xfs_mount *mp, - xfs_sb_field_t field, - int64_t delta, - int rsvd) -{ - int status; - -#ifdef HAVE_PERCPU_SB - ASSERT(field < XFS_SBS_IFREE || field > XFS_SBS_FDBLOCKS); -#endif - - spin_lock(&mp->m_sb_lock); - status = xfs_mod_incore_sb_unlocked(mp, field, delta, rsvd); - spin_unlock(&mp->m_sb_lock); - - return status; -} - /* * xfs_getsb() is called to obtain the buffer for the superblock. * The buffer is returned locked and read in from disk. diff --git a/fs/xfs/xfs_mount.h b/fs/xfs/xfs_mount.h index 1c11512bab83..8c995a2ccb6f 100644 --- a/fs/xfs/xfs_mount.h +++ b/fs/xfs/xfs_mount.h @@ -313,9 +313,8 @@ extern __uint64_t xfs_default_resblks(xfs_mount_t *mp); extern int xfs_mountfs(xfs_mount_t *mp); 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_incore_sb(xfs_mount_t *, xfs_sb_field_t, int64_t, int); + 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, -- cgit v1.2.3-59-g8ed1b From 653c60b633a9019a54a80d64b5ed33ecb214823c Mon Sep 17 00:00:00 2001 From: Dave Chinner Date: Mon, 23 Feb 2015 21:43:37 +1100 Subject: xfs: introduce mmap/truncate lock Right now we cannot serialise mmap against truncate or hole punch sanely. ->page_mkwrite is not able to take locks that the read IO path normally takes (i.e. the inode iolock) because that could result in lock inversions (read - iolock - page fault - page_mkwrite - iolock) and so we cannot use an IO path lock to serialise page write faults against truncate operations. Instead, introduce a new lock that is used *only* in the ->page_mkwrite path that is the equivalent of the iolock. The lock ordering in a page fault is i_mmaplock -> page lock -> i_ilock, and so in truncate we can i_iolock -> i_mmaplock and so lock out new write faults during the process of truncation. Because i_mmap_lock is outside the page lock, we can hold it across all the same operations we hold the i_iolock for. The only difference is that we never hold the i_mmaplock in the normal IO path and so do not ever have the possibility that we can page fault inside it. Hence there are no recursion issues on the i_mmap_lock and so we can use it to serialise page fault IO against inode modification operations that affect the IO path. This patch introduces the i_mmaplock infrastructure, lockdep annotations and initialisation/destruction code. Use of the new lock will be in subsequent patches. Signed-off-by: Dave Chinner Reviewed-by: Brian Foster Signed-off-by: Dave Chinner --- fs/xfs/xfs_inode.c | 128 ++++++++++++++++++++++++++++++++++++++++------------- fs/xfs/xfs_inode.h | 29 +++++++++--- fs/xfs/xfs_super.c | 2 + 3 files changed, 121 insertions(+), 38 deletions(-) diff --git a/fs/xfs/xfs_inode.c b/fs/xfs/xfs_inode.c index daafa1f6d260..ac24818f7b2d 100644 --- a/fs/xfs/xfs_inode.c +++ b/fs/xfs/xfs_inode.c @@ -117,24 +117,34 @@ xfs_ilock_attr_map_shared( } /* - * The xfs inode contains 2 locks: a multi-reader lock called the - * i_iolock and a multi-reader lock called the i_lock. This routine - * allows either or both of the locks to be obtained. + * The xfs inode contains 3 multi-reader locks: the i_iolock the i_mmap_lock and + * the i_lock. This routine allows various combinations of the locks to be + * obtained. * - * The 2 locks should always be ordered so that the IO lock is - * obtained first in order to prevent deadlock. + * The 3 locks should always be ordered so that the IO lock is obtained first, + * the mmap lock second and the ilock last in order to prevent deadlock. * - * ip -- the inode being locked - * lock_flags -- this parameter indicates the inode's locks - * to be locked. It can be: - * XFS_IOLOCK_SHARED, - * XFS_IOLOCK_EXCL, - * XFS_ILOCK_SHARED, - * XFS_ILOCK_EXCL, - * XFS_IOLOCK_SHARED | XFS_ILOCK_SHARED, - * XFS_IOLOCK_SHARED | XFS_ILOCK_EXCL, - * XFS_IOLOCK_EXCL | XFS_ILOCK_SHARED, - * XFS_IOLOCK_EXCL | XFS_ILOCK_EXCL + * Basic locking order: + * + * i_iolock -> i_mmap_lock -> page_lock -> i_ilock + * + * mmap_sem locking order: + * + * i_iolock -> page lock -> mmap_sem + * mmap_sem -> i_mmap_lock -> page_lock + * + * The difference in mmap_sem locking order mean that we cannot hold the + * i_mmap_lock over syscall based read(2)/write(2) based IO. These IO paths can + * fault in pages during copy in/out (for buffered IO) or require the mmap_sem + * in get_user_pages() to map the user pages into the kernel address space for + * direct IO. Similarly the i_iolock cannot be taken inside a page fault because + * page faults already hold the mmap_sem. + * + * Hence to serialise fully against both syscall and mmap based IO, we need to + * take both the i_iolock and the i_mmap_lock. These locks should *only* be both + * taken in places where we need to invalidate the page cache in a race + * free manner (e.g. truncate, hole punch and other extent manipulation + * functions). */ void xfs_ilock( @@ -150,6 +160,8 @@ xfs_ilock( */ ASSERT((lock_flags & (XFS_IOLOCK_SHARED | XFS_IOLOCK_EXCL)) != (XFS_IOLOCK_SHARED | XFS_IOLOCK_EXCL)); + ASSERT((lock_flags & (XFS_MMAPLOCK_SHARED | XFS_MMAPLOCK_EXCL)) != + (XFS_MMAPLOCK_SHARED | XFS_MMAPLOCK_EXCL)); ASSERT((lock_flags & (XFS_ILOCK_SHARED | XFS_ILOCK_EXCL)) != (XFS_ILOCK_SHARED | XFS_ILOCK_EXCL)); ASSERT((lock_flags & ~(XFS_LOCK_MASK | XFS_LOCK_DEP_MASK)) == 0); @@ -159,6 +171,11 @@ xfs_ilock( else if (lock_flags & XFS_IOLOCK_SHARED) mraccess_nested(&ip->i_iolock, XFS_IOLOCK_DEP(lock_flags)); + if (lock_flags & XFS_MMAPLOCK_EXCL) + mrupdate_nested(&ip->i_mmaplock, XFS_MMAPLOCK_DEP(lock_flags)); + else if (lock_flags & XFS_MMAPLOCK_SHARED) + mraccess_nested(&ip->i_mmaplock, XFS_MMAPLOCK_DEP(lock_flags)); + if (lock_flags & XFS_ILOCK_EXCL) mrupdate_nested(&ip->i_lock, XFS_ILOCK_DEP(lock_flags)); else if (lock_flags & XFS_ILOCK_SHARED) @@ -191,6 +208,8 @@ xfs_ilock_nowait( */ ASSERT((lock_flags & (XFS_IOLOCK_SHARED | XFS_IOLOCK_EXCL)) != (XFS_IOLOCK_SHARED | XFS_IOLOCK_EXCL)); + ASSERT((lock_flags & (XFS_MMAPLOCK_SHARED | XFS_MMAPLOCK_EXCL)) != + (XFS_MMAPLOCK_SHARED | XFS_MMAPLOCK_EXCL)); ASSERT((lock_flags & (XFS_ILOCK_SHARED | XFS_ILOCK_EXCL)) != (XFS_ILOCK_SHARED | XFS_ILOCK_EXCL)); ASSERT((lock_flags & ~(XFS_LOCK_MASK | XFS_LOCK_DEP_MASK)) == 0); @@ -202,21 +221,35 @@ xfs_ilock_nowait( if (!mrtryaccess(&ip->i_iolock)) goto out; } + + if (lock_flags & XFS_MMAPLOCK_EXCL) { + if (!mrtryupdate(&ip->i_mmaplock)) + goto out_undo_iolock; + } else if (lock_flags & XFS_MMAPLOCK_SHARED) { + if (!mrtryaccess(&ip->i_mmaplock)) + goto out_undo_iolock; + } + if (lock_flags & XFS_ILOCK_EXCL) { if (!mrtryupdate(&ip->i_lock)) - goto out_undo_iolock; + goto out_undo_mmaplock; } else if (lock_flags & XFS_ILOCK_SHARED) { if (!mrtryaccess(&ip->i_lock)) - goto out_undo_iolock; + goto out_undo_mmaplock; } return 1; - out_undo_iolock: +out_undo_mmaplock: + if (lock_flags & XFS_MMAPLOCK_EXCL) + mrunlock_excl(&ip->i_mmaplock); + else if (lock_flags & XFS_MMAPLOCK_SHARED) + mrunlock_shared(&ip->i_mmaplock); +out_undo_iolock: if (lock_flags & XFS_IOLOCK_EXCL) mrunlock_excl(&ip->i_iolock); else if (lock_flags & XFS_IOLOCK_SHARED) mrunlock_shared(&ip->i_iolock); - out: +out: return 0; } @@ -244,6 +277,8 @@ xfs_iunlock( */ ASSERT((lock_flags & (XFS_IOLOCK_SHARED | XFS_IOLOCK_EXCL)) != (XFS_IOLOCK_SHARED | XFS_IOLOCK_EXCL)); + ASSERT((lock_flags & (XFS_MMAPLOCK_SHARED | XFS_MMAPLOCK_EXCL)) != + (XFS_MMAPLOCK_SHARED | XFS_MMAPLOCK_EXCL)); ASSERT((lock_flags & (XFS_ILOCK_SHARED | XFS_ILOCK_EXCL)) != (XFS_ILOCK_SHARED | XFS_ILOCK_EXCL)); ASSERT((lock_flags & ~(XFS_LOCK_MASK | XFS_LOCK_DEP_MASK)) == 0); @@ -254,6 +289,11 @@ xfs_iunlock( else if (lock_flags & XFS_IOLOCK_SHARED) mrunlock_shared(&ip->i_iolock); + if (lock_flags & XFS_MMAPLOCK_EXCL) + mrunlock_excl(&ip->i_mmaplock); + else if (lock_flags & XFS_MMAPLOCK_SHARED) + mrunlock_shared(&ip->i_mmaplock); + if (lock_flags & XFS_ILOCK_EXCL) mrunlock_excl(&ip->i_lock); else if (lock_flags & XFS_ILOCK_SHARED) @@ -271,11 +311,14 @@ xfs_ilock_demote( xfs_inode_t *ip, uint lock_flags) { - ASSERT(lock_flags & (XFS_IOLOCK_EXCL|XFS_ILOCK_EXCL)); - ASSERT((lock_flags & ~(XFS_IOLOCK_EXCL|XFS_ILOCK_EXCL)) == 0); + ASSERT(lock_flags & (XFS_IOLOCK_EXCL|XFS_MMAPLOCK_EXCL|XFS_ILOCK_EXCL)); + ASSERT((lock_flags & + ~(XFS_IOLOCK_EXCL|XFS_MMAPLOCK_EXCL|XFS_ILOCK_EXCL)) == 0); if (lock_flags & XFS_ILOCK_EXCL) mrdemote(&ip->i_lock); + if (lock_flags & XFS_MMAPLOCK_EXCL) + mrdemote(&ip->i_mmaplock); if (lock_flags & XFS_IOLOCK_EXCL) mrdemote(&ip->i_iolock); @@ -294,6 +337,12 @@ xfs_isilocked( return rwsem_is_locked(&ip->i_lock.mr_lock); } + if (lock_flags & (XFS_MMAPLOCK_EXCL|XFS_MMAPLOCK_SHARED)) { + if (!(lock_flags & XFS_MMAPLOCK_SHARED)) + return !!ip->i_mmaplock.mr_writer; + return rwsem_is_locked(&ip->i_mmaplock.mr_lock); + } + if (lock_flags & (XFS_IOLOCK_EXCL|XFS_IOLOCK_SHARED)) { if (!(lock_flags & XFS_IOLOCK_SHARED)) return !!ip->i_iolock.mr_writer; @@ -314,14 +363,27 @@ int xfs_lock_delays; #endif /* - * Bump the subclass so xfs_lock_inodes() acquires each lock with - * a different value + * Bump the subclass so xfs_lock_inodes() acquires each lock with a different + * value. This shouldn't be called for page fault locking, but we also need to + * ensure we don't overrun the number of lockdep subclasses for the iolock or + * mmaplock as that is limited to 12 by the mmap lock lockdep annotations. */ static inline int xfs_lock_inumorder(int lock_mode, int subclass) { - if (lock_mode & (XFS_IOLOCK_SHARED|XFS_IOLOCK_EXCL)) + if (lock_mode & (XFS_IOLOCK_SHARED|XFS_IOLOCK_EXCL)) { + ASSERT(subclass + XFS_LOCK_INUMORDER < + (1 << (XFS_MMAPLOCK_SHIFT - XFS_IOLOCK_SHIFT))); lock_mode |= (subclass + XFS_LOCK_INUMORDER) << XFS_IOLOCK_SHIFT; + } + + if (lock_mode & (XFS_MMAPLOCK_SHARED|XFS_MMAPLOCK_EXCL)) { + ASSERT(subclass + XFS_LOCK_INUMORDER < + (1 << (XFS_ILOCK_SHIFT - XFS_MMAPLOCK_SHIFT))); + lock_mode |= (subclass + XFS_LOCK_INUMORDER) << + XFS_MMAPLOCK_SHIFT; + } + if (lock_mode & (XFS_ILOCK_SHARED|XFS_ILOCK_EXCL)) lock_mode |= (subclass + XFS_LOCK_INUMORDER) << XFS_ILOCK_SHIFT; @@ -440,10 +502,10 @@ again: } /* - * xfs_lock_two_inodes() can only be used to lock one type of lock - * at a time - the iolock or the ilock, but not both at once. If - * we lock both at once, lockdep will report false positives saying - * we have violated locking orders. + * xfs_lock_two_inodes() can only be used to lock one type of lock at a time - + * the iolock, the mmaplock or the ilock, but not more than one at a time. If we + * lock more than one at a time, lockdep will report false positives saying we + * have violated locking orders. */ void xfs_lock_two_inodes( @@ -455,8 +517,12 @@ xfs_lock_two_inodes( int attempts = 0; xfs_log_item_t *lp; - if (lock_mode & (XFS_IOLOCK_SHARED|XFS_IOLOCK_EXCL)) - ASSERT((lock_mode & (XFS_ILOCK_SHARED|XFS_ILOCK_EXCL)) == 0); + if (lock_mode & (XFS_IOLOCK_SHARED|XFS_IOLOCK_EXCL)) { + ASSERT(!(lock_mode & (XFS_MMAPLOCK_SHARED|XFS_MMAPLOCK_EXCL))); + ASSERT(!(lock_mode & (XFS_ILOCK_SHARED|XFS_ILOCK_EXCL))); + } else if (lock_mode & (XFS_MMAPLOCK_SHARED|XFS_MMAPLOCK_EXCL)) + ASSERT(!(lock_mode & (XFS_ILOCK_SHARED|XFS_ILOCK_EXCL))); + ASSERT(ip0->i_ino != ip1->i_ino); if (ip0->i_ino > ip1->i_ino) { diff --git a/fs/xfs/xfs_inode.h b/fs/xfs/xfs_inode.h index 86cd6b39bed7..f65ab8176c2f 100644 --- a/fs/xfs/xfs_inode.h +++ b/fs/xfs/xfs_inode.h @@ -56,6 +56,7 @@ typedef struct xfs_inode { struct xfs_inode_log_item *i_itemp; /* logging information */ mrlock_t i_lock; /* inode lock */ mrlock_t i_iolock; /* inode IO lock */ + mrlock_t i_mmaplock; /* inode mmap IO lock */ atomic_t i_pincount; /* inode pin count */ spinlock_t i_flags_lock; /* inode i_flags lock */ /* Miscellaneous state. */ @@ -263,15 +264,20 @@ static inline int xfs_isiflocked(struct xfs_inode *ip) #define XFS_IOLOCK_SHARED (1<<1) #define XFS_ILOCK_EXCL (1<<2) #define XFS_ILOCK_SHARED (1<<3) +#define XFS_MMAPLOCK_EXCL (1<<4) +#define XFS_MMAPLOCK_SHARED (1<<5) #define XFS_LOCK_MASK (XFS_IOLOCK_EXCL | XFS_IOLOCK_SHARED \ - | XFS_ILOCK_EXCL | XFS_ILOCK_SHARED) + | XFS_ILOCK_EXCL | XFS_ILOCK_SHARED \ + | XFS_MMAPLOCK_EXCL | XFS_MMAPLOCK_SHARED) #define XFS_LOCK_FLAGS \ { XFS_IOLOCK_EXCL, "IOLOCK_EXCL" }, \ { XFS_IOLOCK_SHARED, "IOLOCK_SHARED" }, \ { XFS_ILOCK_EXCL, "ILOCK_EXCL" }, \ - { XFS_ILOCK_SHARED, "ILOCK_SHARED" } + { XFS_ILOCK_SHARED, "ILOCK_SHARED" }, \ + { XFS_MMAPLOCK_EXCL, "MMAPLOCK_EXCL" }, \ + { XFS_MMAPLOCK_SHARED, "MMAPLOCK_SHARED" } /* @@ -302,17 +308,26 @@ static inline int xfs_isiflocked(struct xfs_inode *ip) #define XFS_IOLOCK_SHIFT 16 #define XFS_IOLOCK_PARENT (XFS_LOCK_PARENT << XFS_IOLOCK_SHIFT) +#define XFS_MMAPLOCK_SHIFT 20 + #define XFS_ILOCK_SHIFT 24 #define XFS_ILOCK_PARENT (XFS_LOCK_PARENT << XFS_ILOCK_SHIFT) #define XFS_ILOCK_RTBITMAP (XFS_LOCK_RTBITMAP << XFS_ILOCK_SHIFT) #define XFS_ILOCK_RTSUM (XFS_LOCK_RTSUM << XFS_ILOCK_SHIFT) -#define XFS_IOLOCK_DEP_MASK 0x00ff0000 +#define XFS_IOLOCK_DEP_MASK 0x000f0000 +#define XFS_MMAPLOCK_DEP_MASK 0x00f00000 #define XFS_ILOCK_DEP_MASK 0xff000000 -#define XFS_LOCK_DEP_MASK (XFS_IOLOCK_DEP_MASK | XFS_ILOCK_DEP_MASK) - -#define XFS_IOLOCK_DEP(flags) (((flags) & XFS_IOLOCK_DEP_MASK) >> XFS_IOLOCK_SHIFT) -#define XFS_ILOCK_DEP(flags) (((flags) & XFS_ILOCK_DEP_MASK) >> XFS_ILOCK_SHIFT) +#define XFS_LOCK_DEP_MASK (XFS_IOLOCK_DEP_MASK | \ + XFS_MMAPLOCK_DEP_MASK | \ + XFS_ILOCK_DEP_MASK) + +#define XFS_IOLOCK_DEP(flags) (((flags) & XFS_IOLOCK_DEP_MASK) \ + >> XFS_IOLOCK_SHIFT) +#define XFS_MMAPLOCK_DEP(flags) (((flags) & XFS_MMAPLOCK_DEP_MASK) \ + >> XFS_MMAPLOCK_SHIFT) +#define XFS_ILOCK_DEP(flags) (((flags) & XFS_ILOCK_DEP_MASK) \ + >> XFS_ILOCK_SHIFT) /* * For multiple groups support: if S_ISGID bit is set in the parent diff --git a/fs/xfs/xfs_super.c b/fs/xfs/xfs_super.c index 8fcc4ccc5c79..50c6fd3bd362 100644 --- a/fs/xfs/xfs_super.c +++ b/fs/xfs/xfs_super.c @@ -986,6 +986,8 @@ xfs_fs_inode_init_once( atomic_set(&ip->i_pincount, 0); spin_lock_init(&ip->i_flags_lock); + mrlock_init(&ip->i_mmaplock, MRLOCK_ALLOW_EQUAL_PRI|MRLOCK_BARRIER, + "xfsino", ip->i_ino); mrlock_init(&ip->i_lock, MRLOCK_ALLOW_EQUAL_PRI|MRLOCK_BARRIER, "xfsino", ip->i_ino); } -- cgit v1.2.3-59-g8ed1b From de0e8c20ba3a65b0f15040aabbefdc1999876e6b Mon Sep 17 00:00:00 2001 From: Dave Chinner Date: Mon, 23 Feb 2015 21:44:19 +1100 Subject: xfs: use i_mmaplock on read faults Take the i_mmaplock over read page faults. These come through the ->fault callout, so we need to wrap the generic implementation with the i_mmaplock. While there, add tracepoints for the read fault as it passes through XFS. This gives us a lock order of mmap_sem -> i_mmaplock -> page_lock -> i_lock. Signed-off-by: Dave Chinner Reviewed-by: Brian Foster Signed-off-by: Dave Chinner --- fs/xfs/xfs_file.c | 28 +++++++++++++++++++++++++++- fs/xfs/xfs_trace.h | 2 ++ 2 files changed, 29 insertions(+), 1 deletion(-) diff --git a/fs/xfs/xfs_file.c b/fs/xfs/xfs_file.c index ce615d12fb44..ac174226244a 100644 --- a/fs/xfs/xfs_file.c +++ b/fs/xfs/xfs_file.c @@ -1379,6 +1379,32 @@ xfs_file_llseek( } } +/* + * Locking for serialisation of IO during page faults. This results in a lock + * ordering of: + * + * mmap_sem (MM) + * i_mmap_lock (XFS - truncate serialisation) + * page_lock (MM) + * i_lock (XFS - extent map serialisation) + */ +STATIC int +xfs_filemap_fault( + struct vm_area_struct *vma, + struct vm_fault *vmf) +{ + struct xfs_inode *ip = XFS_I(vma->vm_file->f_mapping->host); + int error; + + trace_xfs_filemap_fault(ip); + + xfs_ilock(ip, XFS_MMAPLOCK_SHARED); + error = filemap_fault(vma, vmf); + xfs_iunlock(ip, XFS_MMAPLOCK_SHARED); + + return error; +} + const struct file_operations xfs_file_operations = { .llseek = xfs_file_llseek, .read = new_sync_read, @@ -1411,7 +1437,7 @@ const struct file_operations xfs_dir_file_operations = { }; static const struct vm_operations_struct xfs_file_vm_ops = { - .fault = filemap_fault, + .fault = xfs_filemap_fault, .map_pages = filemap_map_pages, .page_mkwrite = xfs_vm_page_mkwrite, }; diff --git a/fs/xfs/xfs_trace.h b/fs/xfs/xfs_trace.h index 51372e34d988..c496153b976b 100644 --- a/fs/xfs/xfs_trace.h +++ b/fs/xfs/xfs_trace.h @@ -685,6 +685,8 @@ DEFINE_INODE_EVENT(xfs_inode_set_eofblocks_tag); DEFINE_INODE_EVENT(xfs_inode_clear_eofblocks_tag); DEFINE_INODE_EVENT(xfs_inode_free_eofblocks_invalid); +DEFINE_INODE_EVENT(xfs_filemap_fault); + DECLARE_EVENT_CLASS(xfs_iref_class, TP_PROTO(struct xfs_inode *ip, unsigned long caller_ip), TP_ARGS(ip, caller_ip), -- cgit v1.2.3-59-g8ed1b From 075a924d45cc69c75a35f20b4912b85aa98b180a Mon Sep 17 00:00:00 2001 From: Dave Chinner Date: Mon, 23 Feb 2015 21:44:54 +1100 Subject: xfs: use i_mmaplock on write faults Take the i_mmaplock over write page faults. These come through the ->page_mkwrite callout, so we need to wrap that calls with the i_mmaplock. This gives us a lock order of mmap_sem -> i_mmaplock -> page_lock -> i_lock. Also, move the page_mkwrite wrapper to the same region of xfs_file.c as the read fault wrappers and add a tracepoint. Signed-off-by: Dave Chinner Reviewed-by: Brian Foster Signed-off-by: Dave Chinner --- fs/xfs/xfs_file.c | 39 ++++++++++++++++++++++++--------------- fs/xfs/xfs_trace.h | 1 + 2 files changed, 25 insertions(+), 15 deletions(-) diff --git a/fs/xfs/xfs_file.c b/fs/xfs/xfs_file.c index ac174226244a..d55f011401bf 100644 --- a/fs/xfs/xfs_file.c +++ b/fs/xfs/xfs_file.c @@ -990,20 +990,6 @@ xfs_file_mmap( return 0; } -/* - * mmap()d file has taken write protection fault and is being made - * writable. We can set the page state up correctly for a writable - * page, which means we can do correct delalloc accounting (ENOSPC - * checking!) and unwritten extent mapping. - */ -STATIC int -xfs_vm_page_mkwrite( - struct vm_area_struct *vma, - struct vm_fault *vmf) -{ - return block_page_mkwrite(vma, vmf, xfs_get_blocks); -} - /* * This type is designed to indicate the type of offset we would like * to search from page cache for xfs_seek_hole_data(). @@ -1405,6 +1391,29 @@ xfs_filemap_fault( return error; } +/* + * mmap()d file has taken write protection fault and is being made writable. We + * can set the page state up correctly for a writable page, which means we can + * do correct delalloc accounting (ENOSPC checking!) and unwritten extent + * mapping. + */ +STATIC int +xfs_filemap_page_mkwrite( + struct vm_area_struct *vma, + struct vm_fault *vmf) +{ + struct xfs_inode *ip = XFS_I(vma->vm_file->f_mapping->host); + int error; + + trace_xfs_filemap_page_mkwrite(ip); + + xfs_ilock(ip, XFS_MMAPLOCK_SHARED); + error = block_page_mkwrite(vma, vmf, xfs_get_blocks); + xfs_iunlock(ip, XFS_MMAPLOCK_SHARED); + + return error; +} + const struct file_operations xfs_file_operations = { .llseek = xfs_file_llseek, .read = new_sync_read, @@ -1439,5 +1448,5 @@ const struct file_operations xfs_dir_file_operations = { static const struct vm_operations_struct xfs_file_vm_ops = { .fault = xfs_filemap_fault, .map_pages = filemap_map_pages, - .page_mkwrite = xfs_vm_page_mkwrite, + .page_mkwrite = xfs_filemap_page_mkwrite, }; diff --git a/fs/xfs/xfs_trace.h b/fs/xfs/xfs_trace.h index c496153b976b..b1e059b398c0 100644 --- a/fs/xfs/xfs_trace.h +++ b/fs/xfs/xfs_trace.h @@ -686,6 +686,7 @@ DEFINE_INODE_EVENT(xfs_inode_clear_eofblocks_tag); DEFINE_INODE_EVENT(xfs_inode_free_eofblocks_invalid); DEFINE_INODE_EVENT(xfs_filemap_fault); +DEFINE_INODE_EVENT(xfs_filemap_page_mkwrite); DECLARE_EVENT_CLASS(xfs_iref_class, TP_PROTO(struct xfs_inode *ip, unsigned long caller_ip), -- cgit v1.2.3-59-g8ed1b From e8e9ad42c1f1e1bfbe0e8c32c8cac02e9ebfb7ef Mon Sep 17 00:00:00 2001 From: Dave Chinner Date: Mon, 23 Feb 2015 21:45:32 +1100 Subject: xfs: take i_mmap_lock on extent manipulation operations Now we have the i_mmap_lock being held across the page fault IO path, we now add extent manipulation operation exclusion by adding the lock to the paths that directly modify extent maps. This includes truncate, hole punching and other fallocate based operations. The operations will now take both the i_iolock and the i_mmaplock in exclusive mode, thereby ensuring that all IO and page faults block without holding any page locks while the extent manipulation is in progress. This gives us the lock order during truncate of i_iolock -> i_mmaplock -> page_lock -> i_lock, hence providing the same lock order as the iolock provides the normal IO path without involving the mmap_sem. Signed-off-by: Dave Chinner Reviewed-by: Brian Foster Signed-off-by: Dave Chinner --- fs/xfs/xfs_file.c | 3 +++ fs/xfs/xfs_ioctl.c | 5 ++++- fs/xfs/xfs_iops.c | 7 ++++++- 3 files changed, 13 insertions(+), 2 deletions(-) diff --git a/fs/xfs/xfs_file.c b/fs/xfs/xfs_file.c index d55f011401bf..609b5aaddd8e 100644 --- a/fs/xfs/xfs_file.c +++ b/fs/xfs/xfs_file.c @@ -841,6 +841,9 @@ xfs_file_fallocate( if (error) goto out_unlock; + xfs_ilock(ip, XFS_MMAPLOCK_EXCL); + iolock |= XFS_MMAPLOCK_EXCL; + if (mode & FALLOC_FL_PUNCH_HOLE) { error = xfs_free_file_space(ip, offset, len); if (error) diff --git a/fs/xfs/xfs_ioctl.c b/fs/xfs/xfs_ioctl.c index ac4feae45eb3..4ee44ddfdfb7 100644 --- a/fs/xfs/xfs_ioctl.c +++ b/fs/xfs/xfs_ioctl.c @@ -631,7 +631,7 @@ xfs_ioc_space( if (filp->f_flags & O_DSYNC) flags |= XFS_PREALLOC_SYNC; - if (ioflags & XFS_IO_INVIS) + if (ioflags & XFS_IO_INVIS) flags |= XFS_PREALLOC_INVISIBLE; error = mnt_want_write_file(filp); @@ -643,6 +643,9 @@ xfs_ioc_space( if (error) goto out_unlock; + xfs_ilock(ip, XFS_MMAPLOCK_EXCL); + iolock |= XFS_MMAPLOCK_EXCL; + switch (bf->l_whence) { case 0: /*SEEK_SET*/ break; diff --git a/fs/xfs/xfs_iops.c b/fs/xfs/xfs_iops.c index d919ad7b16bf..7f59ad34b5c5 100644 --- a/fs/xfs/xfs_iops.c +++ b/fs/xfs/xfs_iops.c @@ -765,6 +765,7 @@ xfs_setattr_size( return error; ASSERT(xfs_isilocked(ip, XFS_IOLOCK_EXCL)); + ASSERT(xfs_isilocked(ip, XFS_MMAPLOCK_EXCL)); ASSERT(S_ISREG(ip->i_d.di_mode)); ASSERT((iattr->ia_valid & (ATTR_UID|ATTR_GID|ATTR_ATIME|ATTR_ATIME_SET| ATTR_MTIME_SET|ATTR_KILL_PRIV|ATTR_TIMES_SET)) == 0); @@ -984,8 +985,12 @@ xfs_vn_setattr( xfs_ilock(ip, iolock); error = xfs_break_layouts(dentry->d_inode, &iolock); - if (!error) + if (!error) { + xfs_ilock(ip, XFS_MMAPLOCK_EXCL); + iolock |= XFS_MMAPLOCK_EXCL; + error = xfs_setattr_size(ip, iattr); + } xfs_iunlock(ip, iolock); } else { error = xfs_setattr_nonsize(ip, iattr, 0); -- cgit v1.2.3-59-g8ed1b From 0f9160b444e4de33b65dfcd3b901358a3129461a Mon Sep 17 00:00:00 2001 From: Dave Chinner Date: Mon, 23 Feb 2015 21:46:58 +1100 Subject: xfs: xfs_setattr_size no longer races with page faults Now that truncate locks out new page faults, we no longer need to do special writeback hacks in truncate to work around potential races between page faults, page cache truncation and file size updates to ensure we get write page faults for extending truncates on sub-page block size filesystems. Hence we can remove the code in xfs_setattr_size() that handles this and update the comments around the code tha thandles page cache truncate and size updates to reflect the new reality. Signed-off-by: Dave Chinner Reviewed-by: Brian Foster Signed-off-by: Dave Chinner --- fs/xfs/xfs_iops.c | 56 ++++++++++++++----------------------------------------- 1 file changed, 14 insertions(+), 42 deletions(-) diff --git a/fs/xfs/xfs_iops.c b/fs/xfs/xfs_iops.c index 7f59ad34b5c5..5c0c27c6fb75 100644 --- a/fs/xfs/xfs_iops.c +++ b/fs/xfs/xfs_iops.c @@ -838,55 +838,27 @@ xfs_setattr_size( inode_dio_wait(inode); /* - * Do all the page cache truncate work outside the transaction context - * as the "lock" order is page lock->log space reservation. i.e. - * locking pages inside the transaction can ABBA deadlock with - * writeback. We have to do the VFS inode size update before we truncate - * the pagecache, however, to avoid racing with page faults beyond the - * new EOF they are not serialised against truncate operations except by - * page locks and size updates. + * We've already locked out new page faults, so now we can safely remove + * pages from the page cache knowing they won't get refaulted until we + * drop the XFS_MMAP_EXCL lock after the extent manipulations are + * complete. The truncate_setsize() call also cleans partial EOF page + * PTEs on extending truncates and hence ensures sub-page block size + * filesystems are correctly handled, too. * - * Hence we are in a situation where a truncate can fail with ENOMEM - * from xfs_trans_reserve(), but having already truncated the in-memory - * version of the file (i.e. made user visible changes). There's not - * much we can do about this, except to hope that the caller sees ENOMEM - * and retries the truncate operation. + * We have to do all the page cache truncate work outside the + * transaction context as the "lock" order is page lock->log space + * reservation as defined by extent allocation in the writeback path. + * Hence a truncate can fail with ENOMEM from xfs_trans_reserve(), but + * having already truncated the in-memory version of the file (i.e. made + * user visible changes). There's not much we can do about this, except + * to hope that the caller sees ENOMEM and retries the truncate + * operation. */ error = block_truncate_page(inode->i_mapping, newsize, xfs_get_blocks); if (error) return error; truncate_setsize(inode, newsize); - /* - * The "we can't serialise against page faults" pain gets worse. - * - * If the file is mapped then we have to clean the page at the old EOF - * when extending the file. Extending the file can expose changes the - * underlying page mapping (e.g. from beyond EOF to a hole or - * unwritten), and so on the next attempt to write to that page we need - * to remap it for write. i.e. we need .page_mkwrite() to be called. - * Hence we need to clean the page to clean the pte and so a new write - * fault will be triggered appropriately. - * - * If we do it before we change the inode size, then we can race with a - * page fault that maps the page with exactly the same problem. If we do - * it after we change the file size, then a new page fault can come in - * and allocate space before we've run the rest of the truncate - * transaction. That's kinda grotesque, but it's better than have data - * over a hole, and so that's the lesser evil that has been chosen here. - * - * The real solution, however, is to have some mechanism for locking out - * page faults while a truncate is in progress. - */ - if (newsize > oldsize && mapping_mapped(VFS_I(ip)->i_mapping)) { - error = filemap_write_and_wait_range( - VFS_I(ip)->i_mapping, - round_down(oldsize, PAGE_CACHE_SIZE), - round_up(oldsize, PAGE_CACHE_SIZE) - 1); - if (error) - return error; - } - tp = xfs_trans_alloc(mp, XFS_TRANS_SETATTR_SIZE); error = xfs_trans_reserve(tp, &M_RES(mp)->tr_itruncate, 0, 0); if (error) -- cgit v1.2.3-59-g8ed1b From 723cac48473358939759885a18e8df113ea96138 Mon Sep 17 00:00:00 2001 From: Dave Chinner Date: Mon, 23 Feb 2015 21:47:29 +1100 Subject: xfs: lock out page faults from extent swap operations Extent swap operations are another extent manipulation operation that we need to ensure does not race against mmap page faults. The current code returns if the file is mapped prior to the swap being done, but it could potentially race against new page faults while the swap is in progress. Hence we should use the XFS_MMAPLOCK_EXCL for this operation, too. While there, fix the error path handling that can result in double unlocks of the inodes when cancelling the swapext transaction. Signed-off-by: Dave Chinner Reviewed-by: Brian Foster Signed-off-by: Dave Chinner --- fs/xfs/xfs_bmap_util.c | 31 +++++++++++++++---------------- 1 file changed, 15 insertions(+), 16 deletions(-) diff --git a/fs/xfs/xfs_bmap_util.c b/fs/xfs/xfs_bmap_util.c index 22a5dcb70b32..7efa23e72a90 100644 --- a/fs/xfs/xfs_bmap_util.c +++ b/fs/xfs/xfs_bmap_util.c @@ -1599,13 +1599,6 @@ xfs_swap_extent_flush( /* Verify O_DIRECT for ftmp */ if (VFS_I(ip)->i_mapping->nrpages) return -EINVAL; - - /* - * Don't try to swap extents on mmap()d files because we can't lock - * out races against page faults safely. - */ - if (mapping_mapped(VFS_I(ip)->i_mapping)) - return -EBUSY; return 0; } @@ -1633,13 +1626,14 @@ xfs_swap_extents( } /* - * Lock up the inodes against other IO and truncate to begin with. - * Then we can ensure the inodes are flushed and have no page cache - * safely. Once we have done this we can take the ilocks and do the rest - * of the checks. + * Lock the inodes against other IO, page faults and truncate to + * begin with. Then we can ensure the inodes are flushed and have no + * page cache safely. Once we have done this we can take the ilocks and + * do the rest of the checks. */ - lock_flags = XFS_IOLOCK_EXCL; + lock_flags = XFS_IOLOCK_EXCL | XFS_MMAPLOCK_EXCL; xfs_lock_two_inodes(ip, tip, XFS_IOLOCK_EXCL); + xfs_lock_two_inodes(ip, tip, XFS_MMAPLOCK_EXCL); /* Verify that both files have the same format */ if ((ip->i_d.di_mode & S_IFMT) != (tip->i_d.di_mode & S_IFMT)) { @@ -1666,8 +1660,16 @@ xfs_swap_extents( xfs_trans_cancel(tp, 0); goto out_unlock; } + + /* + * Lock and join the inodes to the tansaction so that transaction commit + * or cancel will unlock the inodes from this point onwards. + */ xfs_lock_two_inodes(ip, tip, XFS_ILOCK_EXCL); lock_flags |= XFS_ILOCK_EXCL; + xfs_trans_ijoin(tp, ip, lock_flags); + xfs_trans_ijoin(tp, tip, lock_flags); + /* Verify all data are being swapped */ if (sxp->sx_offset != 0 || @@ -1720,9 +1722,6 @@ xfs_swap_extents( goto out_trans_cancel; } - xfs_trans_ijoin(tp, ip, lock_flags); - xfs_trans_ijoin(tp, tip, lock_flags); - /* * Before we've swapped the forks, lets set the owners of the forks * appropriately. We have to do this as we are demand paging the btree @@ -1856,5 +1855,5 @@ out_unlock: out_trans_cancel: xfs_trans_cancel(tp, 0); - goto out_unlock; + goto out; } -- cgit v1.2.3-59-g8ed1b From 58c904734cd0917cd0953067dd68003572407c7b Mon Sep 17 00:00:00 2001 From: Dave Chinner Date: Mon, 23 Feb 2015 22:38:08 +1100 Subject: xfs: inodes are new until the dentry cache is set up Al Viro noticed a generic set of issues to do with filehandle lookup racing with dentry cache setup. They involve a filehandle lookup occurring while an inode is being created and the filehandle lookup racing with the dentry creation for the real file. This can lead to multiple dentries for the one path being instantiated. There are a host of other issues around this same set of paths. The underlying cause is that file handle lookup only waits on inode cache instantiation rather than full dentry cache instantiation. XFS is mostly immune to the problems discovered due to it's own internal inode cache, but there are a couple of corner cases where races can happen. We currently clear the XFS_INEW flag when the inode is fully set up after insertion into the cache. Newly allocated inodes are inserted locked and so aren't usable until the allocation transaction commits. This, however, occurs before the dentry and security information is fully initialised and hence the inode is unlocked and available for lookups to find too early. To solve the problem, only clear the XFS_INEW flag for newly created inodes once the dentry is fully instantiated. This means lookups will retry until the XFS_INEW flag is removed from the inode and hence avoids the race conditions in questions. THis also means that xfs_create(), xfs_create_tmpfile() and xfs_symlink() need to finish the setup of the inode in their error paths if we had allocated the inode but failed later in the creation process. xfs_symlink(), in particular, needed a lot of help to make it's error handling match that of xfs_create(). Signed-off-by: Dave Chinner Reviewed-by: Brian Foster Signed-off-by: Dave Chinner --- fs/xfs/xfs_icache.c | 4 ++-- fs/xfs/xfs_inode.c | 22 ++++++++++++-------- fs/xfs/xfs_inode.h | 22 ++++++++++++++++++++ fs/xfs/xfs_iops.c | 24 +++++++++------------- fs/xfs/xfs_iops.h | 2 -- fs/xfs/xfs_qm.c | 13 ++++++++---- fs/xfs/xfs_symlink.c | 58 ++++++++++++++++++++++++++++++---------------------- 7 files changed, 90 insertions(+), 55 deletions(-) diff --git a/fs/xfs/xfs_icache.c b/fs/xfs/xfs_icache.c index 9771b7ef62ed..76a9f2783282 100644 --- a/fs/xfs/xfs_icache.c +++ b/fs/xfs/xfs_icache.c @@ -439,11 +439,11 @@ again: *ipp = ip; /* - * If we have a real type for an on-disk inode, we can set ops(&unlock) + * If we have a real type for an on-disk inode, we can setup the inode * now. If it's a new inode being created, xfs_ialloc will handle it. */ if (xfs_iflags_test(ip, XFS_INEW) && ip->i_d.di_mode != 0) - xfs_setup_inode(ip); + xfs_setup_existing_inode(ip); return 0; out_error_or_again: diff --git a/fs/xfs/xfs_inode.c b/fs/xfs/xfs_inode.c index daafa1f6d260..d0414f305967 100644 --- a/fs/xfs/xfs_inode.c +++ b/fs/xfs/xfs_inode.c @@ -818,7 +818,7 @@ xfs_ialloc( xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL); xfs_trans_log_inode(tp, ip, flags); - /* now that we have an i_mode we can setup inode ops and unlock */ + /* now that we have an i_mode we can setup the inode structure */ xfs_setup_inode(ip); *ipp = ip; @@ -1235,12 +1235,14 @@ xfs_create( xfs_trans_cancel(tp, cancel_flags); out_release_inode: /* - * Wait until after the current transaction is aborted to - * release the inode. This prevents recursive transactions - * and deadlocks from xfs_inactive. + * Wait until after the current transaction is aborted to finish the + * setup of the inode and release the inode. This prevents recursive + * transactions and deadlocks from xfs_inactive. */ - if (ip) + if (ip) { + xfs_finish_inode_setup(ip); IRELE(ip); + } xfs_qm_dqrele(udqp); xfs_qm_dqrele(gdqp); @@ -1345,12 +1347,14 @@ xfs_create_tmpfile( xfs_trans_cancel(tp, cancel_flags); out_release_inode: /* - * Wait until after the current transaction is aborted to - * release the inode. This prevents recursive transactions - * and deadlocks from xfs_inactive. + * Wait until after the current transaction is aborted to finish the + * setup of the inode and release the inode. This prevents recursive + * transactions and deadlocks from xfs_inactive. */ - if (ip) + if (ip) { + xfs_finish_inode_setup(ip); IRELE(ip); + } xfs_qm_dqrele(udqp); xfs_qm_dqrele(gdqp); diff --git a/fs/xfs/xfs_inode.h b/fs/xfs/xfs_inode.h index 86cd6b39bed7..8e82b41d2050 100644 --- a/fs/xfs/xfs_inode.h +++ b/fs/xfs/xfs_inode.h @@ -390,6 +390,28 @@ int xfs_zero_eof(struct xfs_inode *, xfs_off_t, xfs_fsize_t); int xfs_iozero(struct xfs_inode *, loff_t, size_t); +/* from xfs_iops.c */ +/* + * When setting up a newly allocated inode, we need to call + * xfs_finish_inode_setup() once the inode is fully instantiated at + * the VFS level to prevent the rest of the world seeing the inode + * before we've completed instantiation. Otherwise we can do it + * the moment the inode lookup is complete. + */ +extern void xfs_setup_inode(struct xfs_inode *ip); +static inline void xfs_finish_inode_setup(struct xfs_inode *ip) +{ + xfs_iflags_clear(ip, XFS_INEW); + barrier(); + unlock_new_inode(VFS_I(ip)); +} + +static inline void xfs_setup_existing_inode(struct xfs_inode *ip) +{ + xfs_setup_inode(ip); + xfs_finish_inode_setup(ip); +} + #define IHOLD(ip) \ do { \ ASSERT(atomic_read(&VFS_I(ip)->i_count) > 0) ; \ diff --git a/fs/xfs/xfs_iops.c b/fs/xfs/xfs_iops.c index d919ad7b16bf..d7782ae1af3c 100644 --- a/fs/xfs/xfs_iops.c +++ b/fs/xfs/xfs_iops.c @@ -187,6 +187,8 @@ xfs_generic_create( else d_instantiate(dentry, inode); + xfs_finish_inode_setup(ip); + out_free_acl: if (default_acl) posix_acl_release(default_acl); @@ -195,6 +197,7 @@ xfs_generic_create( return error; out_cleanup_inode: + xfs_finish_inode_setup(ip); if (!tmpfile) xfs_cleanup_inode(dir, inode, dentry); iput(inode); @@ -367,9 +370,11 @@ xfs_vn_symlink( goto out_cleanup_inode; d_instantiate(dentry, inode); + xfs_finish_inode_setup(cip); return 0; out_cleanup_inode: + xfs_finish_inode_setup(cip); xfs_cleanup_inode(dir, inode, dentry); iput(inode); out: @@ -1236,16 +1241,12 @@ xfs_diflags_to_iflags( } /* - * Initialize the Linux inode, set up the operation vectors and - * unlock the inode. - * - * When reading existing inodes from disk this is called directly - * from xfs_iget, when creating a new inode it is called from - * xfs_ialloc after setting up the inode. + * Initialize the Linux inode and set up the operation vectors. * - * We are always called with an uninitialised linux inode here. - * We need to initialise the necessary fields and take a reference - * on it. + * When reading existing inodes from disk this is called directly from xfs_iget, + * when creating a new inode it is called from xfs_ialloc after setting up the + * inode. These callers have different criteria for clearing XFS_INEW, so leave + * it up to the caller to deal with unlocking the inode appropriately. */ void xfs_setup_inode( @@ -1332,9 +1333,4 @@ xfs_setup_inode( inode_has_no_xattr(inode); cache_no_acl(inode); } - - xfs_iflags_clear(ip, XFS_INEW); - barrier(); - - unlock_new_inode(inode); } diff --git a/fs/xfs/xfs_iops.h b/fs/xfs/xfs_iops.h index ea7a98e9cb70..a0f84abb0d09 100644 --- a/fs/xfs/xfs_iops.h +++ b/fs/xfs/xfs_iops.h @@ -25,8 +25,6 @@ extern const struct file_operations xfs_dir_file_operations; extern ssize_t xfs_vn_listxattr(struct dentry *, char *data, size_t size); -extern void xfs_setup_inode(struct xfs_inode *); - /* * Internal setattr interfaces. */ diff --git a/fs/xfs/xfs_qm.c b/fs/xfs/xfs_qm.c index 53cc2aaf8d2b..c6b22e1e77ed 100644 --- a/fs/xfs/xfs_qm.c +++ b/fs/xfs/xfs_qm.c @@ -719,6 +719,7 @@ xfs_qm_qino_alloc( xfs_trans_t *tp; int error; int committed; + bool need_alloc = true; *ip = NULL; /* @@ -747,6 +748,7 @@ xfs_qm_qino_alloc( return error; mp->m_sb.sb_gquotino = NULLFSINO; mp->m_sb.sb_pquotino = NULLFSINO; + need_alloc = false; } } @@ -758,7 +760,7 @@ xfs_qm_qino_alloc( return error; } - if (!*ip) { + if (need_alloc) { error = xfs_dir_ialloc(&tp, NULL, S_IFREG, 1, 0, 0, 1, ip, &committed); if (error) { @@ -794,11 +796,14 @@ xfs_qm_qino_alloc( spin_unlock(&mp->m_sb_lock); xfs_log_sb(tp); - if ((error = xfs_trans_commit(tp, XFS_TRANS_RELEASE_LOG_RES))) { + error = xfs_trans_commit(tp, XFS_TRANS_RELEASE_LOG_RES); + if (error) { + ASSERT(XFS_FORCED_SHUTDOWN(mp)); xfs_alert(mp, "%s failed (error %d)!", __func__, error); - return error; } - return 0; + if (need_alloc) + xfs_finish_inode_setup(*ip); + return error; } diff --git a/fs/xfs/xfs_symlink.c b/fs/xfs/xfs_symlink.c index 25791df6f638..3df411eadb86 100644 --- a/fs/xfs/xfs_symlink.c +++ b/fs/xfs/xfs_symlink.c @@ -177,7 +177,7 @@ xfs_symlink( int pathlen; struct xfs_bmap_free free_list; xfs_fsblock_t first_block; - bool unlock_dp_on_error = false; + bool unlock_dp_on_error = false; uint cancel_flags; int committed; xfs_fileoff_t first_fsb; @@ -221,7 +221,7 @@ xfs_symlink( XFS_QMOPT_QUOTALL | XFS_QMOPT_INHERIT, &udqp, &gdqp, &pdqp); if (error) - goto std_return; + return error; tp = xfs_trans_alloc(mp, XFS_TRANS_SYMLINK); cancel_flags = XFS_TRANS_RELEASE_LOG_RES; @@ -241,7 +241,7 @@ xfs_symlink( } if (error) { cancel_flags = 0; - goto error_return; + goto out_trans_cancel; } xfs_ilock(dp, XFS_ILOCK_EXCL | XFS_ILOCK_PARENT); @@ -252,7 +252,7 @@ xfs_symlink( */ if (dp->i_d.di_flags & XFS_DIFLAG_NOSYMLINKS) { error = -EPERM; - goto error_return; + goto out_trans_cancel; } /* @@ -261,7 +261,7 @@ xfs_symlink( error = xfs_trans_reserve_quota(tp, mp, udqp, gdqp, pdqp, resblks, 1, 0); if (error) - goto error_return; + goto out_trans_cancel; /* * Check for ability to enter directory entry, if no space reserved. @@ -269,7 +269,7 @@ xfs_symlink( if (!resblks) { error = xfs_dir_canenter(tp, dp, link_name); if (error) - goto error_return; + goto out_trans_cancel; } /* * Initialize the bmap freelist prior to calling either @@ -282,15 +282,14 @@ xfs_symlink( */ error = xfs_dir_ialloc(&tp, dp, S_IFLNK | (mode & ~S_IFMT), 1, 0, prid, resblks > 0, &ip, NULL); - if (error) { - if (error == -ENOSPC) - goto error_return; - goto error1; - } + if (error) + goto out_trans_cancel; /* - * An error after we've joined dp to the transaction will result in the - * transaction cancel unlocking dp so don't do it explicitly in the + * Now we join the directory inode to the transaction. We do not do it + * earlier because xfs_dir_ialloc might commit the previous transaction + * (and release all the locks). An error from here on will result in + * the transaction cancel unlocking dp so don't do it explicitly in the * error path. */ xfs_trans_ijoin(tp, dp, XFS_ILOCK_EXCL); @@ -330,7 +329,7 @@ xfs_symlink( XFS_BMAPI_METADATA, &first_block, resblks, mval, &nmaps, &free_list); if (error) - goto error2; + goto out_bmap_cancel; if (resblks) resblks -= fs_blocks; @@ -348,7 +347,7 @@ xfs_symlink( BTOBB(byte_cnt), 0); if (!bp) { error = -ENOMEM; - goto error2; + goto out_bmap_cancel; } bp->b_ops = &xfs_symlink_buf_ops; @@ -378,7 +377,7 @@ xfs_symlink( error = xfs_dir_createname(tp, dp, link_name, ip->i_ino, &first_block, &free_list, resblks); if (error) - goto error2; + goto out_bmap_cancel; xfs_trans_ichgtime(tp, dp, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG); xfs_trans_log_inode(tp, dp, XFS_ILOG_CORE); @@ -392,10 +391,13 @@ xfs_symlink( } error = xfs_bmap_finish(&tp, &free_list, &committed); - if (error) { - goto error2; - } + if (error) + goto out_bmap_cancel; + error = xfs_trans_commit(tp, XFS_TRANS_RELEASE_LOG_RES); + if (error) + goto out_release_inode; + xfs_qm_dqrele(udqp); xfs_qm_dqrele(gdqp); xfs_qm_dqrele(pdqp); @@ -403,20 +405,28 @@ xfs_symlink( *ipp = ip; return 0; - error2: - IRELE(ip); - error1: +out_bmap_cancel: xfs_bmap_cancel(&free_list); cancel_flags |= XFS_TRANS_ABORT; - error_return: +out_trans_cancel: xfs_trans_cancel(tp, cancel_flags); +out_release_inode: + /* + * Wait until after the current transaction is aborted to finish the + * setup of the inode and release the inode. This prevents recursive + * transactions and deadlocks from xfs_inactive. + */ + if (ip) { + xfs_finish_inode_setup(ip); + IRELE(ip); + } + xfs_qm_dqrele(udqp); xfs_qm_dqrele(gdqp); xfs_qm_dqrele(pdqp); if (unlock_dp_on_error) xfs_iunlock(dp, XFS_ILOCK_EXCL); - std_return: return error; } -- cgit v1.2.3-59-g8ed1b From c29aad41159829c0e1dcbf06c8e02284993b7901 Mon Sep 17 00:00:00 2001 From: Eric Sandeen Date: Mon, 23 Feb 2015 22:39:08 +1100 Subject: xfs: pass mp to XFS_WANT_CORRUPTED_GOTO Today, if we hit an XFS_WANT_CORRUPTED_GOTO we don't print any information about which filesystem hit it. Passing in the mp allows us to print the filesystem (device) name, which is a pretty critical piece of information. Tested by running fsfuzzer 'til I hit some. Signed-off-by: Eric Sandeen Reviewed-by: Dave Chinner Signed-off-by: Dave Chinner --- fs/xfs/libxfs/xfs_alloc.c | 71 +++++++++++++------------- fs/xfs/libxfs/xfs_bmap.c | 121 ++++++++++++++++++++++++--------------------- fs/xfs/libxfs/xfs_btree.c | 18 +++---- fs/xfs/libxfs/xfs_ialloc.c | 24 ++++----- fs/xfs/xfs_discard.c | 2 +- fs/xfs/xfs_error.h | 4 +- 6 files changed, 124 insertions(+), 116 deletions(-) diff --git a/fs/xfs/libxfs/xfs_alloc.c b/fs/xfs/libxfs/xfs_alloc.c index a6fbf4472017..d38d69a64af4 100644 --- a/fs/xfs/libxfs/xfs_alloc.c +++ b/fs/xfs/libxfs/xfs_alloc.c @@ -682,7 +682,7 @@ xfs_alloc_ag_vextent_exact( error = xfs_alloc_get_rec(bno_cur, &fbno, &flen, &i); if (error) goto error0; - XFS_WANT_CORRUPTED_GOTO(i == 1, error0); + XFS_WANT_CORRUPTED_GOTO(args->mp, i == 1, error0); ASSERT(fbno <= args->agbno); /* @@ -783,7 +783,7 @@ xfs_alloc_find_best_extent( error = xfs_alloc_get_rec(*scur, sbno, slen, &i); if (error) goto error0; - XFS_WANT_CORRUPTED_GOTO(i == 1, error0); + XFS_WANT_CORRUPTED_GOTO(args->mp, i == 1, error0); xfs_alloc_compute_aligned(args, *sbno, *slen, sbnoa, slena); /* @@ -946,7 +946,7 @@ restart: if ((error = xfs_alloc_get_rec(cnt_cur, <bno, <len, &i))) goto error0; - XFS_WANT_CORRUPTED_GOTO(i == 1, error0); + XFS_WANT_CORRUPTED_GOTO(args->mp, i == 1, error0); if (ltlen >= args->minlen) break; if ((error = xfs_btree_increment(cnt_cur, 0, &i))) @@ -966,7 +966,7 @@ restart: */ if ((error = xfs_alloc_get_rec(cnt_cur, <bno, <len, &i))) goto error0; - XFS_WANT_CORRUPTED_GOTO(i == 1, error0); + XFS_WANT_CORRUPTED_GOTO(args->mp, i == 1, error0); xfs_alloc_compute_aligned(args, ltbno, ltlen, <bnoa, <lena); if (ltlena < args->minlen) @@ -999,7 +999,7 @@ restart: cnt_cur->bc_ptrs[0] = besti; if ((error = xfs_alloc_get_rec(cnt_cur, <bno, <len, &i))) goto error0; - XFS_WANT_CORRUPTED_GOTO(i == 1, error0); + XFS_WANT_CORRUPTED_GOTO(args->mp, i == 1, error0); ASSERT(ltbno + ltlen <= be32_to_cpu(XFS_BUF_TO_AGF(args->agbp)->agf_length)); args->len = blen; if (!xfs_alloc_fix_minleft(args)) { @@ -1088,7 +1088,7 @@ restart: if (bno_cur_lt) { if ((error = xfs_alloc_get_rec(bno_cur_lt, <bno, <len, &i))) goto error0; - XFS_WANT_CORRUPTED_GOTO(i == 1, error0); + XFS_WANT_CORRUPTED_GOTO(args->mp, i == 1, error0); xfs_alloc_compute_aligned(args, ltbno, ltlen, <bnoa, <lena); if (ltlena >= args->minlen) @@ -1104,7 +1104,7 @@ restart: if (bno_cur_gt) { if ((error = xfs_alloc_get_rec(bno_cur_gt, >bno, >len, &i))) goto error0; - XFS_WANT_CORRUPTED_GOTO(i == 1, error0); + XFS_WANT_CORRUPTED_GOTO(args->mp, i == 1, error0); xfs_alloc_compute_aligned(args, gtbno, gtlen, >bnoa, >lena); if (gtlena >= args->minlen) @@ -1303,7 +1303,7 @@ restart: error = xfs_alloc_get_rec(cnt_cur, &fbno, &flen, &i); if (error) goto error0; - XFS_WANT_CORRUPTED_GOTO(i == 1, error0); + XFS_WANT_CORRUPTED_GOTO(args->mp, i == 1, error0); xfs_alloc_compute_aligned(args, fbno, flen, &rbno, &rlen); @@ -1342,7 +1342,7 @@ restart: * This can't happen in the second case above. */ rlen = XFS_EXTLEN_MIN(args->maxlen, rlen); - XFS_WANT_CORRUPTED_GOTO(rlen == 0 || + XFS_WANT_CORRUPTED_GOTO(args->mp, rlen == 0 || (rlen <= flen && rbno + rlen <= fbno + flen), error0); if (rlen < args->maxlen) { xfs_agblock_t bestfbno; @@ -1362,13 +1362,13 @@ restart: if ((error = xfs_alloc_get_rec(cnt_cur, &fbno, &flen, &i))) goto error0; - XFS_WANT_CORRUPTED_GOTO(i == 1, error0); + XFS_WANT_CORRUPTED_GOTO(args->mp, i == 1, error0); if (flen < bestrlen) break; xfs_alloc_compute_aligned(args, fbno, flen, &rbno, &rlen); rlen = XFS_EXTLEN_MIN(args->maxlen, rlen); - XFS_WANT_CORRUPTED_GOTO(rlen == 0 || + XFS_WANT_CORRUPTED_GOTO(args->mp, rlen == 0 || (rlen <= flen && rbno + rlen <= fbno + flen), error0); if (rlen > bestrlen) { @@ -1383,7 +1383,7 @@ restart: if ((error = xfs_alloc_lookup_eq(cnt_cur, bestfbno, bestflen, &i))) goto error0; - XFS_WANT_CORRUPTED_GOTO(i == 1, error0); + XFS_WANT_CORRUPTED_GOTO(args->mp, i == 1, error0); rlen = bestrlen; rbno = bestrbno; flen = bestflen; @@ -1408,7 +1408,7 @@ restart: if (!xfs_alloc_fix_minleft(args)) goto out_nominleft; rlen = args->len; - XFS_WANT_CORRUPTED_GOTO(rlen <= flen, error0); + XFS_WANT_CORRUPTED_GOTO(args->mp, rlen <= flen, error0); /* * Allocate and initialize a cursor for the by-block tree. */ @@ -1422,7 +1422,7 @@ restart: cnt_cur = bno_cur = NULL; args->len = rlen; args->agbno = rbno; - XFS_WANT_CORRUPTED_GOTO( + XFS_WANT_CORRUPTED_GOTO(args->mp, args->agbno + args->len <= be32_to_cpu(XFS_BUF_TO_AGF(args->agbp)->agf_length), error0); @@ -1467,7 +1467,7 @@ xfs_alloc_ag_vextent_small( if (i) { if ((error = xfs_alloc_get_rec(ccur, &fbno, &flen, &i))) goto error0; - XFS_WANT_CORRUPTED_GOTO(i == 1, error0); + XFS_WANT_CORRUPTED_GOTO(args->mp, i == 1, error0); } /* * Nothing in the btree, try the freelist. Make sure @@ -1493,7 +1493,7 @@ xfs_alloc_ag_vextent_small( } args->len = 1; args->agbno = fbno; - XFS_WANT_CORRUPTED_GOTO( + XFS_WANT_CORRUPTED_GOTO(args->mp, args->agbno + args->len <= be32_to_cpu(XFS_BUF_TO_AGF(args->agbp)->agf_length), error0); @@ -1579,7 +1579,7 @@ xfs_free_ag_extent( */ if ((error = xfs_alloc_get_rec(bno_cur, <bno, <len, &i))) goto error0; - XFS_WANT_CORRUPTED_GOTO(i == 1, error0); + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, error0); /* * It's not contiguous, though. */ @@ -1591,7 +1591,8 @@ xfs_free_ag_extent( * space was invalid, it's (partly) already free. * Very bad. */ - XFS_WANT_CORRUPTED_GOTO(ltbno + ltlen <= bno, error0); + XFS_WANT_CORRUPTED_GOTO(mp, + ltbno + ltlen <= bno, error0); } } /* @@ -1606,7 +1607,7 @@ xfs_free_ag_extent( */ if ((error = xfs_alloc_get_rec(bno_cur, >bno, >len, &i))) goto error0; - XFS_WANT_CORRUPTED_GOTO(i == 1, error0); + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, error0); /* * It's not contiguous, though. */ @@ -1618,7 +1619,7 @@ xfs_free_ag_extent( * space was invalid, it's (partly) already free. * Very bad. */ - XFS_WANT_CORRUPTED_GOTO(gtbno >= bno + len, error0); + XFS_WANT_CORRUPTED_GOTO(mp, gtbno >= bno + len, error0); } } /* @@ -1635,31 +1636,31 @@ xfs_free_ag_extent( */ if ((error = xfs_alloc_lookup_eq(cnt_cur, ltbno, ltlen, &i))) goto error0; - XFS_WANT_CORRUPTED_GOTO(i == 1, error0); + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, error0); if ((error = xfs_btree_delete(cnt_cur, &i))) goto error0; - XFS_WANT_CORRUPTED_GOTO(i == 1, error0); + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, error0); /* * Delete the old by-size entry on the right. */ if ((error = xfs_alloc_lookup_eq(cnt_cur, gtbno, gtlen, &i))) goto error0; - XFS_WANT_CORRUPTED_GOTO(i == 1, error0); + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, error0); if ((error = xfs_btree_delete(cnt_cur, &i))) goto error0; - XFS_WANT_CORRUPTED_GOTO(i == 1, error0); + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, error0); /* * Delete the old by-block entry for the right block. */ if ((error = xfs_btree_delete(bno_cur, &i))) goto error0; - XFS_WANT_CORRUPTED_GOTO(i == 1, error0); + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, error0); /* * Move the by-block cursor back to the left neighbor. */ if ((error = xfs_btree_decrement(bno_cur, 0, &i))) goto error0; - XFS_WANT_CORRUPTED_GOTO(i == 1, error0); + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, error0); #ifdef DEBUG /* * Check that this is the right record: delete didn't @@ -1672,7 +1673,7 @@ xfs_free_ag_extent( if ((error = xfs_alloc_get_rec(bno_cur, &xxbno, &xxlen, &i))) goto error0; - XFS_WANT_CORRUPTED_GOTO( + XFS_WANT_CORRUPTED_GOTO(mp, i == 1 && xxbno == ltbno && xxlen == ltlen, error0); } @@ -1695,17 +1696,17 @@ xfs_free_ag_extent( */ if ((error = xfs_alloc_lookup_eq(cnt_cur, ltbno, ltlen, &i))) goto error0; - XFS_WANT_CORRUPTED_GOTO(i == 1, error0); + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, error0); if ((error = xfs_btree_delete(cnt_cur, &i))) goto error0; - XFS_WANT_CORRUPTED_GOTO(i == 1, error0); + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, error0); /* * Back up the by-block cursor to the left neighbor, and * update its length. */ if ((error = xfs_btree_decrement(bno_cur, 0, &i))) goto error0; - XFS_WANT_CORRUPTED_GOTO(i == 1, error0); + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, error0); nbno = ltbno; nlen = len + ltlen; if ((error = xfs_alloc_update(bno_cur, nbno, nlen))) @@ -1721,10 +1722,10 @@ xfs_free_ag_extent( */ if ((error = xfs_alloc_lookup_eq(cnt_cur, gtbno, gtlen, &i))) goto error0; - XFS_WANT_CORRUPTED_GOTO(i == 1, error0); + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, error0); if ((error = xfs_btree_delete(cnt_cur, &i))) goto error0; - XFS_WANT_CORRUPTED_GOTO(i == 1, error0); + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, error0); /* * Update the starting block and length of the right * neighbor in the by-block tree. @@ -1743,7 +1744,7 @@ xfs_free_ag_extent( nlen = len; if ((error = xfs_btree_insert(bno_cur, &i))) goto error0; - XFS_WANT_CORRUPTED_GOTO(i == 1, error0); + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, error0); } xfs_btree_del_cursor(bno_cur, XFS_BTREE_NOERROR); bno_cur = NULL; @@ -1752,10 +1753,10 @@ xfs_free_ag_extent( */ if ((error = xfs_alloc_lookup_eq(cnt_cur, nbno, nlen, &i))) goto error0; - XFS_WANT_CORRUPTED_GOTO(i == 0, error0); + XFS_WANT_CORRUPTED_GOTO(mp, i == 0, error0); if ((error = xfs_btree_insert(cnt_cur, &i))) goto error0; - XFS_WANT_CORRUPTED_GOTO(i == 1, error0); + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, error0); xfs_btree_del_cursor(cnt_cur, XFS_BTREE_NOERROR); cnt_cur = NULL; diff --git a/fs/xfs/libxfs/xfs_bmap.c b/fs/xfs/libxfs/xfs_bmap.c index 61ec015dca16..c52ef579b86c 100644 --- a/fs/xfs/libxfs/xfs_bmap.c +++ b/fs/xfs/libxfs/xfs_bmap.c @@ -410,7 +410,7 @@ xfs_bmap_check_leaf_extents( goto error_norelse; } block = XFS_BUF_TO_BLOCK(bp); - XFS_WANT_CORRUPTED_GOTO( + XFS_WANT_CORRUPTED_GOTO(mp, xfs_bmap_sanity_check(mp, bp, level), error0); if (level == 0) @@ -424,7 +424,8 @@ xfs_bmap_check_leaf_extents( xfs_check_block(block, mp, 0, 0); pp = XFS_BMBT_PTR_ADDR(mp, block, 1, mp->m_bmap_dmxr[1]); bno = be64_to_cpu(*pp); - XFS_WANT_CORRUPTED_GOTO(XFS_FSB_SANITY_CHECK(mp, bno), error0); + XFS_WANT_CORRUPTED_GOTO(mp, + XFS_FSB_SANITY_CHECK(mp, bno), error0); if (bp_release) { bp_release = 0; xfs_trans_brelse(NULL, bp); @@ -1029,7 +1030,7 @@ xfs_bmap_add_attrfork_btree( if ((error = xfs_bmbt_lookup_ge(cur, 0, 0, 0, &stat))) goto error0; /* must be at least one entry */ - XFS_WANT_CORRUPTED_GOTO(stat == 1, error0); + XFS_WANT_CORRUPTED_GOTO(mp, stat == 1, error0); if ((error = xfs_btree_new_iroot(cur, flags, &stat))) goto error0; if (stat == 0) { @@ -1311,14 +1312,14 @@ xfs_bmap_read_extents( if (error) return error; block = XFS_BUF_TO_BLOCK(bp); - XFS_WANT_CORRUPTED_GOTO( - xfs_bmap_sanity_check(mp, bp, level), - error0); + XFS_WANT_CORRUPTED_GOTO(mp, + xfs_bmap_sanity_check(mp, bp, level), error0); if (level == 0) break; pp = XFS_BMBT_PTR_ADDR(mp, block, 1, mp->m_bmap_dmxr[1]); bno = be64_to_cpu(*pp); - XFS_WANT_CORRUPTED_GOTO(XFS_FSB_SANITY_CHECK(mp, bno), error0); + XFS_WANT_CORRUPTED_GOTO(mp, + XFS_FSB_SANITY_CHECK(mp, bno), error0); xfs_trans_brelse(tp, bp); } /* @@ -1345,7 +1346,7 @@ xfs_bmap_read_extents( XFS_ERRLEVEL_LOW, ip->i_mount, block); goto error0; } - XFS_WANT_CORRUPTED_GOTO( + XFS_WANT_CORRUPTED_GOTO(mp, xfs_bmap_sanity_check(mp, bp, 0), error0); /* @@ -1755,7 +1756,9 @@ xfs_bmap_add_extent_delay_real( xfs_filblks_t temp=0; /* value for da_new calculations */ xfs_filblks_t temp2=0;/* value for da_new calculations */ int tmp_rval; /* partial logging flags */ + struct xfs_mount *mp; + mp = bma->tp ? bma->tp->t_mountp : NULL; ifp = XFS_IFORK_PTR(bma->ip, XFS_DATA_FORK); ASSERT(bma->idx >= 0); @@ -1866,15 +1869,15 @@ xfs_bmap_add_extent_delay_real( RIGHT.br_blockcount, &i); if (error) goto done; - XFS_WANT_CORRUPTED_GOTO(i == 1, done); + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, done); error = xfs_btree_delete(bma->cur, &i); if (error) goto done; - XFS_WANT_CORRUPTED_GOTO(i == 1, done); + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, done); error = xfs_btree_decrement(bma->cur, 0, &i); if (error) goto done; - XFS_WANT_CORRUPTED_GOTO(i == 1, done); + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, done); error = xfs_bmbt_update(bma->cur, LEFT.br_startoff, LEFT.br_startblock, LEFT.br_blockcount + @@ -1907,7 +1910,7 @@ xfs_bmap_add_extent_delay_real( &i); if (error) goto done; - XFS_WANT_CORRUPTED_GOTO(i == 1, done); + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, done); error = xfs_bmbt_update(bma->cur, LEFT.br_startoff, LEFT.br_startblock, LEFT.br_blockcount + @@ -1938,7 +1941,7 @@ xfs_bmap_add_extent_delay_real( RIGHT.br_blockcount, &i); if (error) goto done; - XFS_WANT_CORRUPTED_GOTO(i == 1, done); + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, done); error = xfs_bmbt_update(bma->cur, PREV.br_startoff, new->br_startblock, PREV.br_blockcount + @@ -1968,12 +1971,12 @@ xfs_bmap_add_extent_delay_real( &i); if (error) goto done; - XFS_WANT_CORRUPTED_GOTO(i == 0, done); + XFS_WANT_CORRUPTED_GOTO(mp, i == 0, done); bma->cur->bc_rec.b.br_state = XFS_EXT_NORM; error = xfs_btree_insert(bma->cur, &i); if (error) goto done; - XFS_WANT_CORRUPTED_GOTO(i == 1, done); + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, done); } break; @@ -2001,7 +2004,7 @@ xfs_bmap_add_extent_delay_real( &i); if (error) goto done; - XFS_WANT_CORRUPTED_GOTO(i == 1, done); + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, done); error = xfs_bmbt_update(bma->cur, LEFT.br_startoff, LEFT.br_startblock, LEFT.br_blockcount + @@ -2038,12 +2041,12 @@ xfs_bmap_add_extent_delay_real( &i); if (error) goto done; - XFS_WANT_CORRUPTED_GOTO(i == 0, done); + XFS_WANT_CORRUPTED_GOTO(mp, i == 0, done); bma->cur->bc_rec.b.br_state = XFS_EXT_NORM; error = xfs_btree_insert(bma->cur, &i); if (error) goto done; - XFS_WANT_CORRUPTED_GOTO(i == 1, done); + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, done); } if (xfs_bmap_needs_btree(bma->ip, XFS_DATA_FORK)) { @@ -2084,7 +2087,7 @@ xfs_bmap_add_extent_delay_real( RIGHT.br_blockcount, &i); if (error) goto done; - XFS_WANT_CORRUPTED_GOTO(i == 1, done); + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, done); error = xfs_bmbt_update(bma->cur, new->br_startoff, new->br_startblock, new->br_blockcount + @@ -2122,12 +2125,12 @@ xfs_bmap_add_extent_delay_real( &i); if (error) goto done; - XFS_WANT_CORRUPTED_GOTO(i == 0, done); + XFS_WANT_CORRUPTED_GOTO(mp, i == 0, done); bma->cur->bc_rec.b.br_state = XFS_EXT_NORM; error = xfs_btree_insert(bma->cur, &i); if (error) goto done; - XFS_WANT_CORRUPTED_GOTO(i == 1, done); + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, done); } if (xfs_bmap_needs_btree(bma->ip, XFS_DATA_FORK)) { @@ -2191,12 +2194,12 @@ xfs_bmap_add_extent_delay_real( &i); if (error) goto done; - XFS_WANT_CORRUPTED_GOTO(i == 0, done); + XFS_WANT_CORRUPTED_GOTO(mp, i == 0, done); bma->cur->bc_rec.b.br_state = XFS_EXT_NORM; error = xfs_btree_insert(bma->cur, &i); if (error) goto done; - XFS_WANT_CORRUPTED_GOTO(i == 1, done); + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, done); } if (xfs_bmap_needs_btree(bma->ip, XFS_DATA_FORK)) { @@ -2309,6 +2312,7 @@ xfs_bmap_add_extent_unwritten_real( /* left is 0, right is 1, prev is 2 */ int rval=0; /* return value (logging flags) */ int state = 0;/* state bits, accessed thru macros */ + struct xfs_mount *mp = tp->t_mountp; *logflagsp = 0; @@ -2421,19 +2425,19 @@ xfs_bmap_add_extent_unwritten_real( RIGHT.br_startblock, RIGHT.br_blockcount, &i))) goto done; - XFS_WANT_CORRUPTED_GOTO(i == 1, done); + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, done); if ((error = xfs_btree_delete(cur, &i))) goto done; - XFS_WANT_CORRUPTED_GOTO(i == 1, done); + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, done); if ((error = xfs_btree_decrement(cur, 0, &i))) goto done; - XFS_WANT_CORRUPTED_GOTO(i == 1, done); + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, done); if ((error = xfs_btree_delete(cur, &i))) goto done; - XFS_WANT_CORRUPTED_GOTO(i == 1, done); + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, done); if ((error = xfs_btree_decrement(cur, 0, &i))) goto done; - XFS_WANT_CORRUPTED_GOTO(i == 1, done); + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, done); if ((error = xfs_bmbt_update(cur, LEFT.br_startoff, LEFT.br_startblock, LEFT.br_blockcount + PREV.br_blockcount + @@ -2464,13 +2468,13 @@ xfs_bmap_add_extent_unwritten_real( PREV.br_startblock, PREV.br_blockcount, &i))) goto done; - XFS_WANT_CORRUPTED_GOTO(i == 1, done); + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, done); if ((error = xfs_btree_delete(cur, &i))) goto done; - XFS_WANT_CORRUPTED_GOTO(i == 1, done); + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, done); if ((error = xfs_btree_decrement(cur, 0, &i))) goto done; - XFS_WANT_CORRUPTED_GOTO(i == 1, done); + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, done); if ((error = xfs_bmbt_update(cur, LEFT.br_startoff, LEFT.br_startblock, LEFT.br_blockcount + PREV.br_blockcount, @@ -2499,13 +2503,13 @@ xfs_bmap_add_extent_unwritten_real( RIGHT.br_startblock, RIGHT.br_blockcount, &i))) goto done; - XFS_WANT_CORRUPTED_GOTO(i == 1, done); + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, done); if ((error = xfs_btree_delete(cur, &i))) goto done; - XFS_WANT_CORRUPTED_GOTO(i == 1, done); + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, done); if ((error = xfs_btree_decrement(cur, 0, &i))) goto done; - XFS_WANT_CORRUPTED_GOTO(i == 1, done); + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, done); if ((error = xfs_bmbt_update(cur, new->br_startoff, new->br_startblock, new->br_blockcount + RIGHT.br_blockcount, @@ -2532,7 +2536,7 @@ xfs_bmap_add_extent_unwritten_real( new->br_startblock, new->br_blockcount, &i))) goto done; - XFS_WANT_CORRUPTED_GOTO(i == 1, done); + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, done); if ((error = xfs_bmbt_update(cur, new->br_startoff, new->br_startblock, new->br_blockcount, newext))) @@ -2569,7 +2573,7 @@ xfs_bmap_add_extent_unwritten_real( PREV.br_startblock, PREV.br_blockcount, &i))) goto done; - XFS_WANT_CORRUPTED_GOTO(i == 1, done); + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, done); if ((error = xfs_bmbt_update(cur, PREV.br_startoff + new->br_blockcount, PREV.br_startblock + new->br_blockcount, @@ -2611,7 +2615,7 @@ xfs_bmap_add_extent_unwritten_real( PREV.br_startblock, PREV.br_blockcount, &i))) goto done; - XFS_WANT_CORRUPTED_GOTO(i == 1, done); + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, done); if ((error = xfs_bmbt_update(cur, PREV.br_startoff + new->br_blockcount, PREV.br_startblock + new->br_blockcount, @@ -2621,7 +2625,7 @@ xfs_bmap_add_extent_unwritten_real( cur->bc_rec.b = *new; if ((error = xfs_btree_insert(cur, &i))) goto done; - XFS_WANT_CORRUPTED_GOTO(i == 1, done); + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, done); } break; @@ -2651,7 +2655,7 @@ xfs_bmap_add_extent_unwritten_real( PREV.br_startblock, PREV.br_blockcount, &i))) goto done; - XFS_WANT_CORRUPTED_GOTO(i == 1, done); + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, done); if ((error = xfs_bmbt_update(cur, PREV.br_startoff, PREV.br_startblock, PREV.br_blockcount - new->br_blockcount, @@ -2689,7 +2693,7 @@ xfs_bmap_add_extent_unwritten_real( PREV.br_startblock, PREV.br_blockcount, &i))) goto done; - XFS_WANT_CORRUPTED_GOTO(i == 1, done); + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, done); if ((error = xfs_bmbt_update(cur, PREV.br_startoff, PREV.br_startblock, PREV.br_blockcount - new->br_blockcount, @@ -2699,11 +2703,11 @@ xfs_bmap_add_extent_unwritten_real( new->br_startblock, new->br_blockcount, &i))) goto done; - XFS_WANT_CORRUPTED_GOTO(i == 0, done); + XFS_WANT_CORRUPTED_GOTO(mp, i == 0, done); cur->bc_rec.b.br_state = XFS_EXT_NORM; if ((error = xfs_btree_insert(cur, &i))) goto done; - XFS_WANT_CORRUPTED_GOTO(i == 1, done); + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, done); } break; @@ -2737,7 +2741,7 @@ xfs_bmap_add_extent_unwritten_real( PREV.br_startblock, PREV.br_blockcount, &i))) goto done; - XFS_WANT_CORRUPTED_GOTO(i == 1, done); + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, done); /* new right extent - oldext */ if ((error = xfs_bmbt_update(cur, r[1].br_startoff, r[1].br_startblock, r[1].br_blockcount, @@ -2749,7 +2753,7 @@ xfs_bmap_add_extent_unwritten_real( new->br_startoff - PREV.br_startoff; if ((error = xfs_btree_insert(cur, &i))) goto done; - XFS_WANT_CORRUPTED_GOTO(i == 1, done); + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, done); /* * Reset the cursor to the position of the new extent * we are about to insert as we can't trust it after @@ -2759,12 +2763,12 @@ xfs_bmap_add_extent_unwritten_real( new->br_startblock, new->br_blockcount, &i))) goto done; - XFS_WANT_CORRUPTED_GOTO(i == 0, done); + XFS_WANT_CORRUPTED_GOTO(mp, i == 0, done); /* new middle extent - newext */ cur->bc_rec.b.br_state = new->br_state; if ((error = xfs_btree_insert(cur, &i))) goto done; - XFS_WANT_CORRUPTED_GOTO(i == 1, done); + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, done); } break; @@ -2968,7 +2972,9 @@ xfs_bmap_add_extent_hole_real( xfs_bmbt_irec_t right; /* right neighbor extent entry */ int rval=0; /* return value (logging flags) */ int state; /* state bits, accessed thru macros */ + struct xfs_mount *mp; + mp = bma->tp ? bma->tp->t_mountp : NULL; ifp = XFS_IFORK_PTR(bma->ip, whichfork); ASSERT(bma->idx >= 0); @@ -3056,15 +3062,15 @@ xfs_bmap_add_extent_hole_real( &i); if (error) goto done; - XFS_WANT_CORRUPTED_GOTO(i == 1, done); + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, done); error = xfs_btree_delete(bma->cur, &i); if (error) goto done; - XFS_WANT_CORRUPTED_GOTO(i == 1, done); + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, done); error = xfs_btree_decrement(bma->cur, 0, &i); if (error) goto done; - XFS_WANT_CORRUPTED_GOTO(i == 1, done); + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, done); error = xfs_bmbt_update(bma->cur, left.br_startoff, left.br_startblock, left.br_blockcount + @@ -3097,7 +3103,7 @@ xfs_bmap_add_extent_hole_real( &i); if (error) goto done; - XFS_WANT_CORRUPTED_GOTO(i == 1, done); + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, done); error = xfs_bmbt_update(bma->cur, left.br_startoff, left.br_startblock, left.br_blockcount + @@ -3131,7 +3137,7 @@ xfs_bmap_add_extent_hole_real( right.br_blockcount, &i); if (error) goto done; - XFS_WANT_CORRUPTED_GOTO(i == 1, done); + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, done); error = xfs_bmbt_update(bma->cur, new->br_startoff, new->br_startblock, new->br_blockcount + @@ -3161,12 +3167,12 @@ xfs_bmap_add_extent_hole_real( new->br_blockcount, &i); if (error) goto done; - XFS_WANT_CORRUPTED_GOTO(i == 0, done); + XFS_WANT_CORRUPTED_GOTO(mp, i == 0, done); bma->cur->bc_rec.b.br_state = new->br_state; error = xfs_btree_insert(bma->cur, &i); if (error) goto done; - XFS_WANT_CORRUPTED_GOTO(i == 1, done); + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, done); } break; } @@ -4801,7 +4807,7 @@ xfs_bmap_del_extent( got.br_startblock, got.br_blockcount, &i))) goto done; - XFS_WANT_CORRUPTED_GOTO(i == 1, done); + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, done); } da_old = da_new = 0; } else { @@ -4835,7 +4841,7 @@ xfs_bmap_del_extent( } if ((error = xfs_btree_delete(cur, &i))) goto done; - XFS_WANT_CORRUPTED_GOTO(i == 1, done); + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, done); break; case 2: @@ -4935,7 +4941,8 @@ xfs_bmap_del_extent( got.br_startblock, temp, &i))) goto done; - XFS_WANT_CORRUPTED_GOTO(i == 1, done); + XFS_WANT_CORRUPTED_GOTO(mp, + i == 1, done); /* * Update the btree record back * to the original value. @@ -4956,7 +4963,7 @@ xfs_bmap_del_extent( error = -ENOSPC; goto done; } - XFS_WANT_CORRUPTED_GOTO(i == 1, done); + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, done); } else flags |= xfs_ilog_fext(whichfork); XFS_IFORK_NEXT_SET(ip, whichfork, diff --git a/fs/xfs/libxfs/xfs_btree.c b/fs/xfs/libxfs/xfs_btree.c index 81cad433df85..8c5ce10e65b3 100644 --- a/fs/xfs/libxfs/xfs_btree.c +++ b/fs/xfs/libxfs/xfs_btree.c @@ -2285,7 +2285,7 @@ xfs_btree_rshift( if (error) goto error0; i = xfs_btree_lastrec(tcur, level); - XFS_WANT_CORRUPTED_GOTO(i == 1, error0); + XFS_WANT_CORRUPTED_GOTO(cur->bc_mp, i == 1, error0); error = xfs_btree_increment(tcur, level, &i); if (error) @@ -3138,7 +3138,7 @@ xfs_btree_insert( goto error0; } - XFS_WANT_CORRUPTED_GOTO(i == 1, error0); + XFS_WANT_CORRUPTED_GOTO(cur->bc_mp, i == 1, error0); level++; /* @@ -3582,15 +3582,15 @@ xfs_btree_delrec( * Actually any entry but the first would suffice. */ i = xfs_btree_lastrec(tcur, level); - XFS_WANT_CORRUPTED_GOTO(i == 1, error0); + XFS_WANT_CORRUPTED_GOTO(cur->bc_mp, i == 1, error0); error = xfs_btree_increment(tcur, level, &i); if (error) goto error0; - XFS_WANT_CORRUPTED_GOTO(i == 1, error0); + XFS_WANT_CORRUPTED_GOTO(cur->bc_mp, i == 1, error0); i = xfs_btree_lastrec(tcur, level); - XFS_WANT_CORRUPTED_GOTO(i == 1, error0); + XFS_WANT_CORRUPTED_GOTO(cur->bc_mp, i == 1, error0); /* Grab a pointer to the block. */ right = xfs_btree_get_block(tcur, level, &rbp); @@ -3634,12 +3634,12 @@ xfs_btree_delrec( rrecs = xfs_btree_get_numrecs(right); if (!xfs_btree_ptr_is_null(cur, &lptr)) { i = xfs_btree_firstrec(tcur, level); - XFS_WANT_CORRUPTED_GOTO(i == 1, error0); + XFS_WANT_CORRUPTED_GOTO(cur->bc_mp, i == 1, error0); error = xfs_btree_decrement(tcur, level, &i); if (error) goto error0; - XFS_WANT_CORRUPTED_GOTO(i == 1, error0); + XFS_WANT_CORRUPTED_GOTO(cur->bc_mp, i == 1, error0); } } @@ -3653,13 +3653,13 @@ xfs_btree_delrec( * previous block. */ i = xfs_btree_firstrec(tcur, level); - XFS_WANT_CORRUPTED_GOTO(i == 1, error0); + XFS_WANT_CORRUPTED_GOTO(cur->bc_mp, i == 1, error0); error = xfs_btree_decrement(tcur, level, &i); if (error) goto error0; i = xfs_btree_firstrec(tcur, level); - XFS_WANT_CORRUPTED_GOTO(i == 1, error0); + XFS_WANT_CORRUPTED_GOTO(cur->bc_mp, i == 1, error0); /* Grab a pointer to the block. */ left = xfs_btree_get_block(tcur, level, &lbp); diff --git a/fs/xfs/libxfs/xfs_ialloc.c b/fs/xfs/libxfs/xfs_ialloc.c index 116ef1ddb3e3..f95c4ae46e28 100644 --- a/fs/xfs/libxfs/xfs_ialloc.c +++ b/fs/xfs/libxfs/xfs_ialloc.c @@ -783,12 +783,12 @@ xfs_dialloc_ag_inobt( error = xfs_inobt_lookup(cur, pagino, XFS_LOOKUP_LE, &i); if (error) goto error0; - XFS_WANT_CORRUPTED_GOTO(i == 1, error0); + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, error0); error = xfs_inobt_get_rec(cur, &rec, &j); if (error) goto error0; - XFS_WANT_CORRUPTED_GOTO(j == 1, error0); + XFS_WANT_CORRUPTED_GOTO(mp, j == 1, error0); if (rec.ir_freecount > 0) { /* @@ -944,19 +944,19 @@ newino: error = xfs_inobt_lookup(cur, 0, XFS_LOOKUP_GE, &i); if (error) goto error0; - XFS_WANT_CORRUPTED_GOTO(i == 1, error0); + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, error0); for (;;) { error = xfs_inobt_get_rec(cur, &rec, &i); if (error) goto error0; - XFS_WANT_CORRUPTED_GOTO(i == 1, error0); + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, error0); if (rec.ir_freecount > 0) break; error = xfs_btree_increment(cur, 0, &i); if (error) goto error0; - XFS_WANT_CORRUPTED_GOTO(i == 1, error0); + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, error0); } alloc_inode: @@ -1039,10 +1039,10 @@ xfs_dialloc_ag_finobt_near( error = xfs_inobt_get_rec(rcur, &rrec, &j); if (error) goto error_rcur; - XFS_WANT_CORRUPTED_GOTO(j == 1, error_rcur); + XFS_WANT_CORRUPTED_GOTO(lcur->bc_mp, j == 1, error_rcur); } - XFS_WANT_CORRUPTED_GOTO(i == 1 || j == 1, error_rcur); + XFS_WANT_CORRUPTED_GOTO(lcur->bc_mp, i == 1 || j == 1, error_rcur); if (i == 1 && j == 1) { /* * Both the left and right records are valid. Choose the closer @@ -1475,14 +1475,14 @@ xfs_difree_inobt( __func__, error); goto error0; } - XFS_WANT_CORRUPTED_GOTO(i == 1, error0); + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, error0); error = xfs_inobt_get_rec(cur, &rec, &i); if (error) { xfs_warn(mp, "%s: xfs_inobt_get_rec() returned error %d.", __func__, error); goto error0; } - XFS_WANT_CORRUPTED_GOTO(i == 1, error0); + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, error0); /* * Get the offset in the inode chunk. */ @@ -1592,7 +1592,7 @@ xfs_difree_finobt( * freed an inode in a previously fully allocated chunk. If not, * something is out of sync. */ - XFS_WANT_CORRUPTED_GOTO(ibtrec->ir_freecount == 1, error); + XFS_WANT_CORRUPTED_GOTO(mp, ibtrec->ir_freecount == 1, error); error = xfs_inobt_insert_rec(cur, ibtrec->ir_freecount, ibtrec->ir_free, &i); @@ -1613,12 +1613,12 @@ xfs_difree_finobt( error = xfs_inobt_get_rec(cur, &rec, &i); if (error) goto error; - XFS_WANT_CORRUPTED_GOTO(i == 1, error); + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, error); rec.ir_free |= XFS_INOBT_MASK(offset); rec.ir_freecount++; - XFS_WANT_CORRUPTED_GOTO((rec.ir_free == ibtrec->ir_free) && + XFS_WANT_CORRUPTED_GOTO(mp, (rec.ir_free == ibtrec->ir_free) && (rec.ir_freecount == ibtrec->ir_freecount), error); diff --git a/fs/xfs/xfs_discard.c b/fs/xfs/xfs_discard.c index 799e5a2d334d..e85a9519a5ae 100644 --- a/fs/xfs/xfs_discard.c +++ b/fs/xfs/xfs_discard.c @@ -84,7 +84,7 @@ xfs_trim_extents( error = xfs_alloc_get_rec(cur, &fbno, &flen, &i); if (error) goto out_del_cursor; - XFS_WANT_CORRUPTED_GOTO(i == 1, out_del_cursor); + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, out_del_cursor); ASSERT(flen <= be32_to_cpu(XFS_BUF_TO_AGF(agbp)->agf_longest)); /* diff --git a/fs/xfs/xfs_error.h b/fs/xfs/xfs_error.h index 279a76e52791..13eeca300281 100644 --- a/fs/xfs/xfs_error.h +++ b/fs/xfs/xfs_error.h @@ -40,13 +40,13 @@ extern void xfs_verifier_error(struct xfs_buf *bp); /* * Macros to set EFSCORRUPTED & return/branch. */ -#define XFS_WANT_CORRUPTED_GOTO(x,l) \ +#define XFS_WANT_CORRUPTED_GOTO(mp, x, l) \ { \ int fs_is_ok = (x); \ ASSERT(fs_is_ok); \ if (unlikely(!fs_is_ok)) { \ XFS_ERROR_REPORT("XFS_WANT_CORRUPTED_GOTO", \ - XFS_ERRLEVEL_LOW, NULL); \ + XFS_ERRLEVEL_LOW, mp); \ error = -EFSCORRUPTED; \ goto l; \ } \ -- cgit v1.2.3-59-g8ed1b From 5fb5aeeeb65726c62dc39986d7a080309259e29c Mon Sep 17 00:00:00 2001 From: Eric Sandeen Date: Mon, 23 Feb 2015 22:39:13 +1100 Subject: xfs: pass mp to XFS_WANT_CORRUPTED_RETURN Today, if we hit an XFS_WANT_CORRUPTED_RETURN we don't print any information about which filesystem hit it. Passing in the mp allows us to print the filesystem (device) name, which is a pretty critical piece of information. Tested by running fsfuzzer 'til I hit some. Signed-off-by: Eric Sandeen Reviewed-by: Dave Chinner Signed-off-by: Dave Chinner --- fs/xfs/libxfs/xfs_alloc.c | 29 ++++++++++++++++------------- fs/xfs/libxfs/xfs_bmap.c | 13 ++++++++----- fs/xfs/libxfs/xfs_btree.c | 6 +++--- fs/xfs/libxfs/xfs_dir2_data.c | 39 ++++++++++++++++++++------------------- fs/xfs/libxfs/xfs_ialloc.c | 18 +++++++++--------- fs/xfs/xfs_error.h | 4 ++-- fs/xfs/xfs_itable.c | 2 +- 7 files changed, 59 insertions(+), 52 deletions(-) diff --git a/fs/xfs/libxfs/xfs_alloc.c b/fs/xfs/libxfs/xfs_alloc.c index d38d69a64af4..14a222fdc665 100644 --- a/fs/xfs/libxfs/xfs_alloc.c +++ b/fs/xfs/libxfs/xfs_alloc.c @@ -315,6 +315,9 @@ xfs_alloc_fixup_trees( xfs_agblock_t nfbno2; /* second new free startblock */ xfs_extlen_t nflen1=0; /* first new free length */ xfs_extlen_t nflen2=0; /* second new free length */ + struct xfs_mount *mp; + + mp = cnt_cur->bc_mp; /* * Look up the record in the by-size tree if necessary. @@ -323,13 +326,13 @@ xfs_alloc_fixup_trees( #ifdef DEBUG if ((error = xfs_alloc_get_rec(cnt_cur, &nfbno1, &nflen1, &i))) return error; - XFS_WANT_CORRUPTED_RETURN( + XFS_WANT_CORRUPTED_RETURN(mp, i == 1 && nfbno1 == fbno && nflen1 == flen); #endif } else { if ((error = xfs_alloc_lookup_eq(cnt_cur, fbno, flen, &i))) return error; - XFS_WANT_CORRUPTED_RETURN(i == 1); + XFS_WANT_CORRUPTED_RETURN(mp, i == 1); } /* * Look up the record in the by-block tree if necessary. @@ -338,13 +341,13 @@ xfs_alloc_fixup_trees( #ifdef DEBUG if ((error = xfs_alloc_get_rec(bno_cur, &nfbno1, &nflen1, &i))) return error; - XFS_WANT_CORRUPTED_RETURN( + XFS_WANT_CORRUPTED_RETURN(mp, i == 1 && nfbno1 == fbno && nflen1 == flen); #endif } else { if ((error = xfs_alloc_lookup_eq(bno_cur, fbno, flen, &i))) return error; - XFS_WANT_CORRUPTED_RETURN(i == 1); + XFS_WANT_CORRUPTED_RETURN(mp, i == 1); } #ifdef DEBUG @@ -355,7 +358,7 @@ xfs_alloc_fixup_trees( bnoblock = XFS_BUF_TO_BLOCK(bno_cur->bc_bufs[0]); cntblock = XFS_BUF_TO_BLOCK(cnt_cur->bc_bufs[0]); - XFS_WANT_CORRUPTED_RETURN( + XFS_WANT_CORRUPTED_RETURN(mp, bnoblock->bb_numrecs == cntblock->bb_numrecs); } #endif @@ -386,25 +389,25 @@ xfs_alloc_fixup_trees( */ if ((error = xfs_btree_delete(cnt_cur, &i))) return error; - XFS_WANT_CORRUPTED_RETURN(i == 1); + XFS_WANT_CORRUPTED_RETURN(mp, i == 1); /* * Add new by-size btree entry(s). */ if (nfbno1 != NULLAGBLOCK) { if ((error = xfs_alloc_lookup_eq(cnt_cur, nfbno1, nflen1, &i))) return error; - XFS_WANT_CORRUPTED_RETURN(i == 0); + XFS_WANT_CORRUPTED_RETURN(mp, i == 0); if ((error = xfs_btree_insert(cnt_cur, &i))) return error; - XFS_WANT_CORRUPTED_RETURN(i == 1); + XFS_WANT_CORRUPTED_RETURN(mp, i == 1); } if (nfbno2 != NULLAGBLOCK) { if ((error = xfs_alloc_lookup_eq(cnt_cur, nfbno2, nflen2, &i))) return error; - XFS_WANT_CORRUPTED_RETURN(i == 0); + XFS_WANT_CORRUPTED_RETURN(mp, i == 0); if ((error = xfs_btree_insert(cnt_cur, &i))) return error; - XFS_WANT_CORRUPTED_RETURN(i == 1); + XFS_WANT_CORRUPTED_RETURN(mp, i == 1); } /* * Fix up the by-block btree entry(s). @@ -415,7 +418,7 @@ xfs_alloc_fixup_trees( */ if ((error = xfs_btree_delete(bno_cur, &i))) return error; - XFS_WANT_CORRUPTED_RETURN(i == 1); + XFS_WANT_CORRUPTED_RETURN(mp, i == 1); } else { /* * Update the by-block entry to start later|be shorter. @@ -429,10 +432,10 @@ xfs_alloc_fixup_trees( */ if ((error = xfs_alloc_lookup_eq(bno_cur, nfbno2, nflen2, &i))) return error; - XFS_WANT_CORRUPTED_RETURN(i == 0); + XFS_WANT_CORRUPTED_RETURN(mp, i == 0); if ((error = xfs_btree_insert(bno_cur, &i))) return error; - XFS_WANT_CORRUPTED_RETURN(i == 1); + XFS_WANT_CORRUPTED_RETURN(mp, i == 1); } return 0; } diff --git a/fs/xfs/libxfs/xfs_bmap.c b/fs/xfs/libxfs/xfs_bmap.c index c52ef579b86c..60cfa90163b8 100644 --- a/fs/xfs/libxfs/xfs_bmap.c +++ b/fs/xfs/libxfs/xfs_bmap.c @@ -5460,6 +5460,7 @@ xfs_bmse_merge( struct xfs_bmbt_irec left; xfs_filblks_t blockcount; int error, i; + struct xfs_mount *mp = ip->i_mount; xfs_bmbt_get_all(gotp, &got); xfs_bmbt_get_all(leftp, &left); @@ -5494,19 +5495,19 @@ xfs_bmse_merge( got.br_blockcount, &i); if (error) return error; - XFS_WANT_CORRUPTED_RETURN(i == 1); + XFS_WANT_CORRUPTED_RETURN(mp, i == 1); error = xfs_btree_delete(cur, &i); if (error) return error; - XFS_WANT_CORRUPTED_RETURN(i == 1); + XFS_WANT_CORRUPTED_RETURN(mp, i == 1); /* lookup and update size of the previous extent */ error = xfs_bmbt_lookup_eq(cur, left.br_startoff, left.br_startblock, left.br_blockcount, &i); if (error) return error; - XFS_WANT_CORRUPTED_RETURN(i == 1); + XFS_WANT_CORRUPTED_RETURN(mp, i == 1); left.br_blockcount = blockcount; @@ -5528,6 +5529,7 @@ xfs_bmse_shift_one( int *logflags) { struct xfs_ifork *ifp; + struct xfs_mount *mp; xfs_fileoff_t startoff; struct xfs_bmbt_rec_host *leftp; struct xfs_bmbt_irec got; @@ -5535,13 +5537,14 @@ xfs_bmse_shift_one( int error; int i; + mp = ip->i_mount; ifp = XFS_IFORK_PTR(ip, whichfork); xfs_bmbt_get_all(gotp, &got); startoff = got.br_startoff - offset_shift_fsb; /* delalloc extents should be prevented by caller */ - XFS_WANT_CORRUPTED_RETURN(!isnullstartblock(got.br_startblock)); + XFS_WANT_CORRUPTED_RETURN(mp, !isnullstartblock(got.br_startblock)); /* * Check for merge if we've got an extent to the left, otherwise make @@ -5580,7 +5583,7 @@ xfs_bmse_shift_one( got.br_blockcount, &i); if (error) return error; - XFS_WANT_CORRUPTED_RETURN(i == 1); + XFS_WANT_CORRUPTED_RETURN(mp, i == 1); got.br_startoff = startoff; return xfs_bmbt_update(cur, got.br_startoff, got.br_startblock, diff --git a/fs/xfs/libxfs/xfs_btree.c b/fs/xfs/libxfs/xfs_btree.c index 8c5ce10e65b3..c72283dd8d44 100644 --- a/fs/xfs/libxfs/xfs_btree.c +++ b/fs/xfs/libxfs/xfs_btree.c @@ -168,7 +168,7 @@ xfs_btree_check_lptr( xfs_fsblock_t bno, /* btree block disk address */ int level) /* btree block level */ { - XFS_WANT_CORRUPTED_RETURN( + XFS_WANT_CORRUPTED_RETURN(cur->bc_mp, level > 0 && bno != NULLFSBLOCK && XFS_FSB_SANITY_CHECK(cur->bc_mp, bno)); @@ -187,7 +187,7 @@ xfs_btree_check_sptr( { xfs_agblock_t agblocks = cur->bc_mp->m_sb.sb_agblocks; - XFS_WANT_CORRUPTED_RETURN( + XFS_WANT_CORRUPTED_RETURN(cur->bc_mp, level > 0 && bno != NULLAGBLOCK && bno != 0 && @@ -1825,7 +1825,7 @@ xfs_btree_lookup( error = xfs_btree_increment(cur, 0, &i); if (error) goto error0; - XFS_WANT_CORRUPTED_RETURN(i == 1); + XFS_WANT_CORRUPTED_RETURN(cur->bc_mp, i == 1); XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT); *stat = 1; return 0; diff --git a/fs/xfs/libxfs/xfs_dir2_data.c b/fs/xfs/libxfs/xfs_dir2_data.c index 5ff31be9b1cd..de1ea16f5748 100644 --- a/fs/xfs/libxfs/xfs_dir2_data.c +++ b/fs/xfs/libxfs/xfs_dir2_data.c @@ -89,7 +89,7 @@ __xfs_dir3_data_check( * so just ensure that the count falls somewhere inside the * block right now. */ - XFS_WANT_CORRUPTED_RETURN(be32_to_cpu(btp->count) < + XFS_WANT_CORRUPTED_RETURN(mp, be32_to_cpu(btp->count) < ((char *)btp - p) / sizeof(struct xfs_dir2_leaf_entry)); break; case cpu_to_be32(XFS_DIR3_DATA_MAGIC): @@ -107,21 +107,21 @@ __xfs_dir3_data_check( bf = ops->data_bestfree_p(hdr); count = lastfree = freeseen = 0; if (!bf[0].length) { - XFS_WANT_CORRUPTED_RETURN(!bf[0].offset); + XFS_WANT_CORRUPTED_RETURN(mp, !bf[0].offset); freeseen |= 1 << 0; } if (!bf[1].length) { - XFS_WANT_CORRUPTED_RETURN(!bf[1].offset); + XFS_WANT_CORRUPTED_RETURN(mp, !bf[1].offset); freeseen |= 1 << 1; } if (!bf[2].length) { - XFS_WANT_CORRUPTED_RETURN(!bf[2].offset); + XFS_WANT_CORRUPTED_RETURN(mp, !bf[2].offset); freeseen |= 1 << 2; } - XFS_WANT_CORRUPTED_RETURN(be16_to_cpu(bf[0].length) >= + XFS_WANT_CORRUPTED_RETURN(mp, be16_to_cpu(bf[0].length) >= be16_to_cpu(bf[1].length)); - XFS_WANT_CORRUPTED_RETURN(be16_to_cpu(bf[1].length) >= + XFS_WANT_CORRUPTED_RETURN(mp, be16_to_cpu(bf[1].length) >= be16_to_cpu(bf[2].length)); /* * Loop over the data/unused entries. @@ -134,18 +134,18 @@ __xfs_dir3_data_check( * doesn't need to be there. */ if (be16_to_cpu(dup->freetag) == XFS_DIR2_DATA_FREE_TAG) { - XFS_WANT_CORRUPTED_RETURN(lastfree == 0); - XFS_WANT_CORRUPTED_RETURN( + XFS_WANT_CORRUPTED_RETURN(mp, lastfree == 0); + XFS_WANT_CORRUPTED_RETURN(mp, be16_to_cpu(*xfs_dir2_data_unused_tag_p(dup)) == (char *)dup - (char *)hdr); dfp = xfs_dir2_data_freefind(hdr, bf, dup); if (dfp) { i = (int)(dfp - bf); - XFS_WANT_CORRUPTED_RETURN( + XFS_WANT_CORRUPTED_RETURN(mp, (freeseen & (1 << i)) == 0); freeseen |= 1 << i; } else { - XFS_WANT_CORRUPTED_RETURN( + XFS_WANT_CORRUPTED_RETURN(mp, be16_to_cpu(dup->length) <= be16_to_cpu(bf[2].length)); } @@ -160,13 +160,13 @@ __xfs_dir3_data_check( * The linear search is crude but this is DEBUG code. */ dep = (xfs_dir2_data_entry_t *)p; - XFS_WANT_CORRUPTED_RETURN(dep->namelen != 0); - XFS_WANT_CORRUPTED_RETURN( + XFS_WANT_CORRUPTED_RETURN(mp, dep->namelen != 0); + XFS_WANT_CORRUPTED_RETURN(mp, !xfs_dir_ino_validate(mp, be64_to_cpu(dep->inumber))); - XFS_WANT_CORRUPTED_RETURN( + XFS_WANT_CORRUPTED_RETURN(mp, be16_to_cpu(*ops->data_entry_tag_p(dep)) == (char *)dep - (char *)hdr); - XFS_WANT_CORRUPTED_RETURN( + XFS_WANT_CORRUPTED_RETURN(mp, ops->data_get_ftype(dep) < XFS_DIR3_FT_MAX); count++; lastfree = 0; @@ -183,14 +183,15 @@ __xfs_dir3_data_check( be32_to_cpu(lep[i].hashval) == hash) break; } - XFS_WANT_CORRUPTED_RETURN(i < be32_to_cpu(btp->count)); + XFS_WANT_CORRUPTED_RETURN(mp, + i < be32_to_cpu(btp->count)); } p += ops->data_entsize(dep->namelen); } /* * Need to have seen all the entries and all the bestfree slots. */ - XFS_WANT_CORRUPTED_RETURN(freeseen == 7); + XFS_WANT_CORRUPTED_RETURN(mp, freeseen == 7); if (hdr->magic == cpu_to_be32(XFS_DIR2_BLOCK_MAGIC) || hdr->magic == cpu_to_be32(XFS_DIR3_BLOCK_MAGIC)) { for (i = stale = 0; i < be32_to_cpu(btp->count); i++) { @@ -198,13 +199,13 @@ __xfs_dir3_data_check( cpu_to_be32(XFS_DIR2_NULL_DATAPTR)) stale++; if (i > 0) - XFS_WANT_CORRUPTED_RETURN( + XFS_WANT_CORRUPTED_RETURN(mp, be32_to_cpu(lep[i].hashval) >= be32_to_cpu(lep[i - 1].hashval)); } - XFS_WANT_CORRUPTED_RETURN(count == + XFS_WANT_CORRUPTED_RETURN(mp, count == be32_to_cpu(btp->count) - be32_to_cpu(btp->stale)); - XFS_WANT_CORRUPTED_RETURN(stale == be32_to_cpu(btp->stale)); + XFS_WANT_CORRUPTED_RETURN(mp, stale == be32_to_cpu(btp->stale)); } return 0; } diff --git a/fs/xfs/libxfs/xfs_ialloc.c b/fs/xfs/libxfs/xfs_ialloc.c index f95c4ae46e28..db0444893e96 100644 --- a/fs/xfs/libxfs/xfs_ialloc.c +++ b/fs/xfs/libxfs/xfs_ialloc.c @@ -700,7 +700,7 @@ xfs_ialloc_next_rec( error = xfs_inobt_get_rec(cur, rec, &i); if (error) return error; - XFS_WANT_CORRUPTED_RETURN(i == 1); + XFS_WANT_CORRUPTED_RETURN(cur->bc_mp, i == 1); } return 0; @@ -724,7 +724,7 @@ xfs_ialloc_get_rec( error = xfs_inobt_get_rec(cur, rec, &i); if (error) return error; - XFS_WANT_CORRUPTED_RETURN(i == 1); + XFS_WANT_CORRUPTED_RETURN(cur->bc_mp, i == 1); } return 0; @@ -1016,7 +1016,7 @@ xfs_dialloc_ag_finobt_near( error = xfs_inobt_get_rec(lcur, rec, &i); if (error) return error; - XFS_WANT_CORRUPTED_RETURN(i == 1); + XFS_WANT_CORRUPTED_RETURN(lcur->bc_mp, i == 1); /* * See if we've landed in the parent inode record. The finobt @@ -1095,7 +1095,7 @@ xfs_dialloc_ag_finobt_newino( error = xfs_inobt_get_rec(cur, rec, &i); if (error) return error; - XFS_WANT_CORRUPTED_RETURN(i == 1); + XFS_WANT_CORRUPTED_RETURN(cur->bc_mp, i == 1); return 0; } } @@ -1106,12 +1106,12 @@ xfs_dialloc_ag_finobt_newino( error = xfs_inobt_lookup(cur, 0, XFS_LOOKUP_GE, &i); if (error) return error; - XFS_WANT_CORRUPTED_RETURN(i == 1); + XFS_WANT_CORRUPTED_RETURN(cur->bc_mp, i == 1); error = xfs_inobt_get_rec(cur, rec, &i); if (error) return error; - XFS_WANT_CORRUPTED_RETURN(i == 1); + XFS_WANT_CORRUPTED_RETURN(cur->bc_mp, i == 1); return 0; } @@ -1133,19 +1133,19 @@ xfs_dialloc_ag_update_inobt( error = xfs_inobt_lookup(cur, frec->ir_startino, XFS_LOOKUP_EQ, &i); if (error) return error; - XFS_WANT_CORRUPTED_RETURN(i == 1); + XFS_WANT_CORRUPTED_RETURN(cur->bc_mp, i == 1); error = xfs_inobt_get_rec(cur, &rec, &i); if (error) return error; - XFS_WANT_CORRUPTED_RETURN(i == 1); + XFS_WANT_CORRUPTED_RETURN(cur->bc_mp, i == 1); ASSERT((XFS_AGINO_TO_OFFSET(cur->bc_mp, rec.ir_startino) % XFS_INODES_PER_CHUNK) == 0); rec.ir_free &= ~XFS_INOBT_MASK(offset); rec.ir_freecount--; - XFS_WANT_CORRUPTED_RETURN((rec.ir_free == frec->ir_free) && + XFS_WANT_CORRUPTED_RETURN(cur->bc_mp, (rec.ir_free == frec->ir_free) && (rec.ir_freecount == frec->ir_freecount)); return xfs_inobt_update(cur, &rec); diff --git a/fs/xfs/xfs_error.h b/fs/xfs/xfs_error.h index 13eeca300281..c0394ed126fc 100644 --- a/fs/xfs/xfs_error.h +++ b/fs/xfs/xfs_error.h @@ -52,13 +52,13 @@ extern void xfs_verifier_error(struct xfs_buf *bp); } \ } -#define XFS_WANT_CORRUPTED_RETURN(x) \ +#define XFS_WANT_CORRUPTED_RETURN(mp, x) \ { \ int fs_is_ok = (x); \ ASSERT(fs_is_ok); \ if (unlikely(!fs_is_ok)) { \ XFS_ERROR_REPORT("XFS_WANT_CORRUPTED_RETURN", \ - XFS_ERRLEVEL_LOW, NULL); \ + XFS_ERRLEVEL_LOW, mp); \ return -EFSCORRUPTED; \ } \ } diff --git a/fs/xfs/xfs_itable.c b/fs/xfs/xfs_itable.c index 82e314258f73..80429891dc9b 100644 --- a/fs/xfs/xfs_itable.c +++ b/fs/xfs/xfs_itable.c @@ -229,7 +229,7 @@ xfs_bulkstat_grab_ichunk( error = xfs_inobt_get_rec(cur, irec, &stat); if (error) return error; - XFS_WANT_CORRUPTED_RETURN(stat == 1); + XFS_WANT_CORRUPTED_RETURN(cur->bc_mp, stat == 1); /* Check if the record contains the inode in request */ if (irec->ir_startino + XFS_INODES_PER_CHUNK <= agino) { -- cgit v1.2.3-59-g8ed1b From 3b9ce795fa895e192991150aac03bc3f1614f360 Mon Sep 17 00:00:00 2001 From: Eric Sandeen Date: Tue, 24 Feb 2015 10:13:37 +1100 Subject: xfs: log unmount events on console There are times, when doing triage and forensics, that we would like to know whether a filesystem was unmounted, or if the plug was pulled without a clean unmount. Log unmounts at the same level (NOTICE) as we log mounts. Signed-off-by: Eric Sandeen Reviewed-by: Dave Chinner Signed-off-by: Dave Chinner --- fs/xfs/xfs_super.c | 1 + 1 file changed, 1 insertion(+) diff --git a/fs/xfs/xfs_super.c b/fs/xfs/xfs_super.c index 8fcc4ccc5c79..dd40d255536a 100644 --- a/fs/xfs/xfs_super.c +++ b/fs/xfs/xfs_super.c @@ -1039,6 +1039,7 @@ xfs_fs_put_super( { struct xfs_mount *mp = XFS_M(sb); + xfs_notice(mp, "Unmounting Filesystem"); xfs_filestream_unmount(mp); xfs_unmountfs(mp); -- cgit v1.2.3-59-g8ed1b From fdadf2676aeebbbfb7a28790471fb982fdf388fa Mon Sep 17 00:00:00 2001 From: Eric Sandeen Date: Tue, 24 Feb 2015 10:14:04 +1100 Subject: xfs: clarify async write failure ratelimit message Today, when the "failing async writes" get ratelimited, we see: XFS:: 62836 callbacks suppressed Aside from the extra ":" it's not entirely clear which message is being suppressed, especially if other messages or ratelimits are happening at the same time. Clarify this as i.e.: XFS (dm-11): Failing async write on buffer block 0x140090. Retrying async write. XFS: Failing async write: 62836 callbacks suppressed Signed-off-by: Eric Sandeen Reviewed-by: Christoph Hellwig Signed-off-by: Dave Chinner --- fs/xfs/xfs_buf_item.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/fs/xfs/xfs_buf_item.c b/fs/xfs/xfs_buf_item.c index 507d96a57ac7..092d652bc03d 100644 --- a/fs/xfs/xfs_buf_item.c +++ b/fs/xfs/xfs_buf_item.c @@ -537,9 +537,9 @@ xfs_buf_item_push( /* has a previous flush failed due to IO errors? */ if ((bp->b_flags & XBF_WRITE_FAIL) && - ___ratelimit(&xfs_buf_write_fail_rl_state, "XFS:")) { + ___ratelimit(&xfs_buf_write_fail_rl_state, "XFS: Failing async write")) { xfs_warn(bp->b_target->bt_mount, -"Detected failing async write on buffer block 0x%llx. Retrying async write.", +"Failing async write on buffer block 0x%llx. Retrying async write.", (long long)bp->b_bn); } -- cgit v1.2.3-59-g8ed1b From dd5e71274a6707c638d76b62361ae5d521ba6fb6 Mon Sep 17 00:00:00 2001 From: Wang Sheng-Hui Date: Tue, 24 Feb 2015 10:15:04 +1100 Subject: xfs: remove old and redundant comment in xfs_mount_validate_sb The error messages document the reason for the checks better than the comment and the comments about volume mounts date back to Irix and so aren't relevant any more. So just remove the old and redundant comment. Signed-off-by: Wang Sheng-Hui Reviewed-by: Dave Chinner Signed-off-by: Dave Chinner --- fs/xfs/libxfs/xfs_sb.c | 8 -------- 1 file changed, 8 deletions(-) diff --git a/fs/xfs/libxfs/xfs_sb.c b/fs/xfs/libxfs/xfs_sb.c index b0a5fe95a3e2..f3ea02bf893e 100644 --- a/fs/xfs/libxfs/xfs_sb.c +++ b/fs/xfs/libxfs/xfs_sb.c @@ -111,14 +111,6 @@ xfs_mount_validate_sb( bool check_inprogress, bool check_version) { - - /* - * If the log device and data device have the - * same device number, the log is internal. - * Consequently, the sb_logstart should be non-zero. If - * we have a zero sb_logstart in this case, we may be trying to mount - * a volume filesystem in a non-volume manner. - */ if (sbp->sb_magicnum != XFS_SB_MAGIC) { xfs_warn(mp, "bad magic number"); return -EWRONGFS; -- cgit v1.2.3-59-g8ed1b From 3790a8cd8a145f14f773c7f42cd11fcfb62a6975 Mon Sep 17 00:00:00 2001 From: Dave Chinner Date: Tue, 24 Feb 2015 10:16:04 +1100 Subject: xfs: xfs_alloc_fix_minleft can underflow near ENOSPC Test generic/224 is failing with a corruption being detected on one of Michael's test boxes. Debug that Michael added is indicating that the minleft trimming is resulting in an underflow: ..... before fixup: rlen 1 args->len 0 after xfs_alloc_fix_len : rlen 1 args->len 1 before goto out_nominleft: rlen 1 args->len 0 before fixup: rlen 1 args->len 0 after xfs_alloc_fix_len : rlen 1 args->len 1 after fixup: rlen 1 args->len 1 before fixup: rlen 1 args->len 0 after xfs_alloc_fix_len : rlen 1 args->len 1 after fixup: rlen 4294967295 args->len 4294967295 XFS: Assertion failed: fs_is_ok, file: fs/xfs/libxfs/xfs_alloc.c, line: 1424 The "goto out_nominleft:" indicates that we are getting close to ENOSPC in the AG, and a couple of allocations later we underflow and the corruption check fires in xfs_alloc_ag_vextent_size(). The issue is that the extent length fixups comaprisons are done with variables of xfs_extlen_t types. These are unsigned so an underflow looks like a really big value and hence is not detected as being smaller than the minimum length allowed for the extent. Hence the corruption check fires as it is noticing that the returned length is longer than the original extent length passed in. This can be easily fixed by ensuring we do the underflow test on signed values, the same way xfs_alloc_fix_len() prevents underflow. So we realise in future that these casts prevent underflows from going undetected, add comments to the code indicating this. Reported-by: Michael L. Semon Tested-by: Michael L. Semon Signed-off-by: Dave Chinner Reviewed-by: Brian Foster Signed-off-by: Dave Chinner --- fs/xfs/libxfs/xfs_alloc.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) diff --git a/fs/xfs/libxfs/xfs_alloc.c b/fs/xfs/libxfs/xfs_alloc.c index 14a222fdc665..516162be1398 100644 --- a/fs/xfs/libxfs/xfs_alloc.c +++ b/fs/xfs/libxfs/xfs_alloc.c @@ -260,6 +260,7 @@ xfs_alloc_fix_len( rlen = rlen - (k - args->mod); else rlen = rlen - args->prod + (args->mod - k); + /* casts to (int) catch length underflows */ if ((int)rlen < (int)args->minlen) return; ASSERT(rlen >= args->minlen && rlen <= args->maxlen); @@ -286,7 +287,8 @@ xfs_alloc_fix_minleft( if (diff >= 0) return 1; args->len += diff; /* shrink the allocated space */ - if (args->len >= args->minlen) + /* casts to (int) catch length underflows */ + if ((int)args->len >= (int)args->minlen) return 1; args->agbno = NULLAGBLOCK; return 0; -- cgit v1.2.3-59-g8ed1b From 444a702231412e82fb1c09679adc159301e9242c Mon Sep 17 00:00:00 2001 From: Eric Sandeen Date: Tue, 24 Feb 2015 10:17:04 +1100 Subject: xfs: remove deprecated mount options We recently removed deprecated sysctls; may as well remove deprecated mount options as well, we've stated that they'd be gone by now in the docs. Signed-off-by: Eric Sandeen Reviewed-by: Carlos Maiolino Signed-off-by: Dave Chinner --- Documentation/filesystems/xfs.txt | 29 +++++++++-------------------- fs/xfs/xfs_super.c | 20 -------------------- 2 files changed, 9 insertions(+), 40 deletions(-) diff --git a/Documentation/filesystems/xfs.txt b/Documentation/filesystems/xfs.txt index 0bfafe108357..5a5a05582b58 100644 --- a/Documentation/filesystems/xfs.txt +++ b/Documentation/filesystems/xfs.txt @@ -228,30 +228,19 @@ default behaviour. Deprecated Mount Options ======================== - delaylog/nodelaylog - Delayed logging is the only logging method that XFS supports - now, so these mount options are now ignored. - - Due for removal in 3.12. - - ihashsize=value - In memory inode hashes have been removed, so this option has - no function as of August 2007. Option is deprecated. - - Due for removal in 3.12. +None at present. - irixsgid - This behaviour is now controlled by a sysctl, so the mount - option is ignored. - Due for removal in 3.12. +Removed Mount Options +===================== - osyncisdsync - osyncisosync - O_SYNC and O_DSYNC are fully supported, so there is no need - for these options any more. + Name Removed + ---- ------- + delaylog/nodelaylog v3.20 + ihashsize v3.20 + irixsgid v3.20 + osyncisdsync/osyncisosync v3.20 - Due for removal in 3.12. sysctls ======= diff --git a/fs/xfs/xfs_super.c b/fs/xfs/xfs_super.c index dd40d255536a..58453e3255f8 100644 --- a/fs/xfs/xfs_super.c +++ b/fs/xfs/xfs_super.c @@ -109,8 +109,6 @@ static struct xfs_kobj xfs_dbg_kobj; /* global debug sysfs attrs */ #define MNTOPT_GQUOTANOENF "gqnoenforce"/* group quota limit enforcement */ #define MNTOPT_PQUOTANOENF "pqnoenforce"/* project quota limit enforcement */ #define MNTOPT_QUOTANOENF "qnoenforce" /* same as uqnoenforce */ -#define MNTOPT_DELAYLOG "delaylog" /* Delayed logging enabled */ -#define MNTOPT_NODELAYLOG "nodelaylog" /* Delayed logging disabled */ #define MNTOPT_DISCARD "discard" /* Discard unused blocks */ #define MNTOPT_NODISCARD "nodiscard" /* Do not discard unused blocks */ @@ -361,28 +359,10 @@ xfs_parseargs( } else if (!strcmp(this_char, MNTOPT_GQUOTANOENF)) { mp->m_qflags |= (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE); mp->m_qflags &= ~XFS_GQUOTA_ENFD; - } else if (!strcmp(this_char, MNTOPT_DELAYLOG)) { - xfs_warn(mp, - "delaylog is the default now, option is deprecated."); - } else if (!strcmp(this_char, MNTOPT_NODELAYLOG)) { - xfs_warn(mp, - "nodelaylog support has been removed, option is deprecated."); } else if (!strcmp(this_char, MNTOPT_DISCARD)) { mp->m_flags |= XFS_MOUNT_DISCARD; } else if (!strcmp(this_char, MNTOPT_NODISCARD)) { mp->m_flags &= ~XFS_MOUNT_DISCARD; - } else if (!strcmp(this_char, "ihashsize")) { - xfs_warn(mp, - "ihashsize no longer used, option is deprecated."); - } else if (!strcmp(this_char, "osyncisdsync")) { - xfs_warn(mp, - "osyncisdsync has no effect, option is deprecated."); - } else if (!strcmp(this_char, "osyncisosync")) { - xfs_warn(mp, - "osyncisosync has no effect, option is deprecated."); - } else if (!strcmp(this_char, "irixsgid")) { - xfs_warn(mp, - "irixsgid is now a sysctl(2) variable, option is deprecated."); } else { xfs_warn(mp, "unknown mount option [%s].", this_char); return -EINVAL; -- cgit v1.2.3-59-g8ed1b From 95afcf5c7bca93fb84d260f70c304f35ef4c3114 Mon Sep 17 00:00:00 2001 From: Dave Chinner Date: Wed, 25 Mar 2015 14:03:32 +1100 Subject: xfs: clean up inode locking for RENAME_WHITEOUT When doing RENAME_WHITEOUT, we now have to lock 5 inodes into the rename transaction. This means we need to update xfs_sort_for_rename() and xfs_lock_inodes() to handle up to 5 inodes. Because of the vagaries of rename, this means we could have anywhere between 3 and 5 inodes locked into the transaction.... While xfs_lock_inodes() does not need anything other than an assert telling us we are passing more inodes that we ever thought we should see, it could do with a logic rework to remove all the indenting. This is not a functional change - it just makes the code a lot easier to read. Signed-off-by: Dave Chinner Reviewed-by: Eric Sandeen Reviewed-by: Brian Foster Signed-off-by: Dave Chinner --- fs/xfs/xfs_inode.c | 145 +++++++++++++++++++++++++---------------------------- 1 file changed, 67 insertions(+), 78 deletions(-) diff --git a/fs/xfs/xfs_inode.c b/fs/xfs/xfs_inode.c index d0414f305967..d0a98bafcbac 100644 --- a/fs/xfs/xfs_inode.c +++ b/fs/xfs/xfs_inode.c @@ -329,15 +329,14 @@ xfs_lock_inumorder(int lock_mode, int subclass) } /* - * The following routine will lock n inodes in exclusive mode. - * We assume the caller calls us with the inodes in i_ino order. + * The following routine will lock n inodes in exclusive mode. We assume the + * caller calls us with the inodes in i_ino order. * - * We need to detect deadlock where an inode that we lock - * is in the AIL and we start waiting for another inode that is locked - * by a thread in a long running transaction (such as truncate). This can - * result in deadlock since the long running trans might need to wait - * for the inode we just locked in order to push the tail and free space - * in the log. + * We need to detect deadlock where an inode that we lock is in the AIL and we + * start waiting for another inode that is locked by a thread in a long running + * transaction (such as truncate). This can result in deadlock since the long + * running trans might need to wait for the inode we just locked in order to + * push the tail and free space in the log. */ void xfs_lock_inodes( @@ -348,30 +347,27 @@ xfs_lock_inodes( int attempts = 0, i, j, try_lock; xfs_log_item_t *lp; - ASSERT(ips && (inodes >= 2)); /* we need at least two */ + /* currently supports between 2 and 5 inodes */ + ASSERT(ips && inodes >= 2 && inodes <= 5); try_lock = 0; i = 0; - again: for (; i < inodes; i++) { ASSERT(ips[i]); - if (i && (ips[i] == ips[i-1])) /* Already locked */ + if (i && (ips[i] == ips[i - 1])) /* Already locked */ continue; /* - * If try_lock is not set yet, make sure all locked inodes - * are not in the AIL. - * If any are, set try_lock to be used later. + * If try_lock is not set yet, make sure all locked inodes are + * not in the AIL. If any are, set try_lock to be used later. */ - if (!try_lock) { for (j = (i - 1); j >= 0 && !try_lock; j--) { lp = (xfs_log_item_t *)ips[j]->i_itemp; - if (lp && (lp->li_flags & XFS_LI_IN_AIL)) { + if (lp && (lp->li_flags & XFS_LI_IN_AIL)) try_lock++; - } } } @@ -381,51 +377,42 @@ again: * we can't get any, we must release all we have * and try again. */ + if (!try_lock) { + xfs_ilock(ips[i], xfs_lock_inumorder(lock_mode, i)); + continue; + } + + /* try_lock means we have an inode locked that is in the AIL. */ + ASSERT(i != 0); + if (xfs_ilock_nowait(ips[i], xfs_lock_inumorder(lock_mode, i))) + continue; - if (try_lock) { - /* try_lock must be 0 if i is 0. */ + /* + * Unlock all previous guys and try again. xfs_iunlock will try + * to push the tail if the inode is in the AIL. + */ + attempts++; + for (j = i - 1; j >= 0; j--) { /* - * try_lock means we have an inode locked - * that is in the AIL. + * Check to see if we've already unlocked this one. Not + * the first one going back, and the inode ptr is the + * same. */ - ASSERT(i != 0); - if (!xfs_ilock_nowait(ips[i], xfs_lock_inumorder(lock_mode, i))) { - attempts++; - - /* - * Unlock all previous guys and try again. - * xfs_iunlock will try to push the tail - * if the inode is in the AIL. - */ - - for(j = i - 1; j >= 0; j--) { - - /* - * Check to see if we've already - * unlocked this one. - * Not the first one going back, - * and the inode ptr is the same. - */ - if ((j != (i - 1)) && ips[j] == - ips[j+1]) - continue; - - xfs_iunlock(ips[j], lock_mode); - } + if (j != (i - 1) && ips[j] == ips[j + 1]) + continue; + + xfs_iunlock(ips[j], lock_mode); + } - if ((attempts % 5) == 0) { - delay(1); /* Don't just spin the CPU */ + if ((attempts % 5) == 0) { + delay(1); /* Don't just spin the CPU */ #ifdef DEBUG - xfs_lock_delays++; + xfs_lock_delays++; #endif - } - i = 0; - try_lock = 0; - goto again; - } - } else { - xfs_ilock(ips[i], xfs_lock_inumorder(lock_mode, i)); } + i = 0; + try_lock = 0; + goto again; } #ifdef DEBUG @@ -2615,19 +2602,22 @@ xfs_remove( /* * Enter all inodes for a rename transaction into a sorted array. */ +#define __XFS_SORT_INODES 5 STATIC void xfs_sort_for_rename( - xfs_inode_t *dp1, /* in: old (source) directory inode */ - xfs_inode_t *dp2, /* in: new (target) directory inode */ - xfs_inode_t *ip1, /* in: inode of old entry */ - xfs_inode_t *ip2, /* in: inode of new entry, if it - already exists, NULL otherwise. */ - xfs_inode_t **i_tab,/* out: array of inode returned, sorted */ - int *num_inodes) /* out: number of inodes in array */ + struct xfs_inode *dp1, /* in: old (source) directory inode */ + struct xfs_inode *dp2, /* in: new (target) directory inode */ + struct xfs_inode *ip1, /* in: inode of old entry */ + struct xfs_inode *ip2, /* in: inode of new entry */ + struct xfs_inode *wip, /* in: whiteout inode */ + struct xfs_inode **i_tab,/* out: sorted array of inodes */ + int *num_inodes) /* in/out: inodes in array */ { - xfs_inode_t *temp; int i, j; + ASSERT(*num_inodes == __XFS_SORT_INODES); + memset(i_tab, 0, *num_inodes * sizeof(struct xfs_inode *)); + /* * i_tab contains a list of pointers to inodes. We initialize * the table here & we'll sort it. We will then use it to @@ -2635,25 +2625,24 @@ xfs_sort_for_rename( * * Note that the table may contain duplicates. e.g., dp1 == dp2. */ - i_tab[0] = dp1; - i_tab[1] = dp2; - i_tab[2] = ip1; - if (ip2) { - *num_inodes = 4; - i_tab[3] = ip2; - } else { - *num_inodes = 3; - i_tab[3] = NULL; - } + i = 0; + i_tab[i++] = dp1; + i_tab[i++] = dp2; + i_tab[i++] = ip1; + if (ip2) + i_tab[i++] = ip2; + if (wip) + i_tab[i++] = wip; + *num_inodes = i; /* * Sort the elements via bubble sort. (Remember, there are at - * most 4 elements to sort, so this is adequate.) + * most 5 elements to sort, so this is adequate.) */ for (i = 0; i < *num_inodes; i++) { for (j = 1; j < *num_inodes; j++) { if (i_tab[j]->i_ino < i_tab[j-1]->i_ino) { - temp = i_tab[j]; + struct xfs_inode *temp = i_tab[j]; i_tab[j] = i_tab[j-1]; i_tab[j-1] = temp; } @@ -2801,16 +2790,16 @@ xfs_rename( xfs_fsblock_t first_block; int cancel_flags; int committed; - xfs_inode_t *inodes[4]; + xfs_inode_t *inodes[__XFS_SORT_INODES]; + int num_inodes = __XFS_SORT_INODES; int spaceres; - int num_inodes; trace_xfs_rename(src_dp, target_dp, src_name, target_name); new_parent = (src_dp != target_dp); src_is_directory = S_ISDIR(src_ip->i_d.di_mode); - xfs_sort_for_rename(src_dp, target_dp, src_ip, target_ip, + xfs_sort_for_rename(src_dp, target_dp, src_ip, target_ip, NULL, inodes, &num_inodes); xfs_bmap_init(&free_list, &first_block); -- cgit v1.2.3-59-g8ed1b From 445883e8133975fe8d555610c70c1c4c501f0bf8 Mon Sep 17 00:00:00 2001 From: Dave Chinner Date: Wed, 25 Mar 2015 14:05:43 +1100 Subject: xfs: cleanup xfs_rename error handling The jump labels are ambiguous and unclear and some of the error paths are used inconsistently. Rules for error jumps are: - use out_trans_cancel for unmodified transaction context - use out_bmap_cancel on ENOSPC errors - use out_trans_abort when transaction is likely to be dirty. Signed-off-by: Dave Chinner Reviewed-by: Eric Sandeen Reviewed-by: Brian Foster Signed-off-by: Dave Chinner --- fs/xfs/xfs_inode.c | 59 ++++++++++++++++++++++++------------------------------ 1 file changed, 26 insertions(+), 33 deletions(-) diff --git a/fs/xfs/xfs_inode.c b/fs/xfs/xfs_inode.c index d0a98bafcbac..426dbf7d094a 100644 --- a/fs/xfs/xfs_inode.c +++ b/fs/xfs/xfs_inode.c @@ -2788,7 +2788,7 @@ xfs_rename( int error; xfs_bmap_free_t free_list; xfs_fsblock_t first_block; - int cancel_flags; + int cancel_flags = 0; int committed; xfs_inode_t *inodes[__XFS_SORT_INODES]; int num_inodes = __XFS_SORT_INODES; @@ -2802,28 +2802,23 @@ xfs_rename( xfs_sort_for_rename(src_dp, target_dp, src_ip, target_ip, NULL, inodes, &num_inodes); - xfs_bmap_init(&free_list, &first_block); tp = xfs_trans_alloc(mp, XFS_TRANS_RENAME); - cancel_flags = XFS_TRANS_RELEASE_LOG_RES; spaceres = XFS_RENAME_SPACE_RES(mp, target_name->len); error = xfs_trans_reserve(tp, &M_RES(mp)->tr_rename, spaceres, 0); if (error == -ENOSPC) { spaceres = 0; error = xfs_trans_reserve(tp, &M_RES(mp)->tr_rename, 0, 0); } - if (error) { - xfs_trans_cancel(tp, 0); - goto std_return; - } + if (error) + goto out_trans_cancel; + cancel_flags = XFS_TRANS_RELEASE_LOG_RES; /* * Attach the dquots to the inodes */ error = xfs_qm_vop_rename_dqattach(inodes); - if (error) { - xfs_trans_cancel(tp, cancel_flags); - goto std_return; - } + if (error) + goto out_trans_cancel; /* * Lock all the participating inodes. Depending upon whether @@ -2853,9 +2848,11 @@ xfs_rename( if (unlikely((target_dp->i_d.di_flags & XFS_DIFLAG_PROJINHERIT) && (xfs_get_projid(target_dp) != xfs_get_projid(src_ip)))) { error = -EXDEV; - goto error_return; + goto out_trans_cancel; } + xfs_bmap_init(&free_list, &first_block); + /* * Handle RENAME_EXCHANGE flags */ @@ -2864,7 +2861,7 @@ xfs_rename( target_dp, target_name, target_ip, &free_list, &first_block, spaceres); if (error) - goto abort_return; + goto out_trans_abort; goto finish_rename; } @@ -2879,7 +2876,7 @@ xfs_rename( if (!spaceres) { error = xfs_dir_canenter(tp, target_dp, target_name); if (error) - goto error_return; + goto out_trans_cancel; } /* * If target does not exist and the rename crosses @@ -2890,9 +2887,9 @@ xfs_rename( src_ip->i_ino, &first_block, &free_list, spaceres); if (error == -ENOSPC) - goto error_return; + goto out_bmap_cancel; if (error) - goto abort_return; + goto out_trans_abort; xfs_trans_ichgtime(tp, target_dp, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG); @@ -2900,7 +2897,7 @@ xfs_rename( if (new_parent && src_is_directory) { error = xfs_bumplink(tp, target_dp); if (error) - goto abort_return; + goto out_trans_abort; } } else { /* target_ip != NULL */ /* @@ -2915,7 +2912,7 @@ xfs_rename( if (!(xfs_dir_isempty(target_ip)) || (target_ip->i_d.di_nlink > 2)) { error = -EEXIST; - goto error_return; + goto out_trans_cancel; } } @@ -2932,7 +2929,7 @@ xfs_rename( src_ip->i_ino, &first_block, &free_list, spaceres); if (error) - goto abort_return; + goto out_trans_abort; xfs_trans_ichgtime(tp, target_dp, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG); @@ -2943,7 +2940,7 @@ xfs_rename( */ error = xfs_droplink(tp, target_ip); if (error) - goto abort_return; + goto out_trans_abort; if (src_is_directory) { /* @@ -2951,7 +2948,7 @@ xfs_rename( */ error = xfs_droplink(tp, target_ip); if (error) - goto abort_return; + goto out_trans_abort; } } /* target_ip != NULL */ @@ -2968,7 +2965,7 @@ xfs_rename( &first_block, &free_list, spaceres); ASSERT(error != -EEXIST); if (error) - goto abort_return; + goto out_trans_abort; } /* @@ -2994,13 +2991,13 @@ xfs_rename( */ error = xfs_droplink(tp, src_dp); if (error) - goto abort_return; + goto out_trans_abort; } error = xfs_dir_removename(tp, src_dp, src_name, src_ip->i_ino, &first_block, &free_list, spaceres); if (error) - goto abort_return; + goto out_trans_abort; xfs_trans_ichgtime(tp, src_dp, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG); xfs_trans_log_inode(tp, src_dp, XFS_ILOG_CORE); @@ -3018,12 +3015,8 @@ finish_rename: } error = xfs_bmap_finish(&tp, &free_list, &committed); - if (error) { - xfs_bmap_cancel(&free_list); - xfs_trans_cancel(tp, (XFS_TRANS_RELEASE_LOG_RES | - XFS_TRANS_ABORT)); - goto std_return; - } + if (error) + goto out_trans_abort; /* * trans_commit will unlock src_ip, target_ip & decrement @@ -3031,12 +3024,12 @@ finish_rename: */ return xfs_trans_commit(tp, XFS_TRANS_RELEASE_LOG_RES); - abort_return: +out_trans_abort: cancel_flags |= XFS_TRANS_ABORT; - error_return: +out_bmap_cancel: xfs_bmap_cancel(&free_list); +out_trans_cancel: xfs_trans_cancel(tp, cancel_flags); - std_return: return error; } -- cgit v1.2.3-59-g8ed1b From 310606b0c7e385e9dd3533d168413ad2c579d961 Mon Sep 17 00:00:00 2001 From: Dave Chinner Date: Wed, 25 Mar 2015 14:06:07 +1100 Subject: xfs: factor out xfs_finish_rename() Rather than use a jump label for the final transaction commit in the rename, factor it into a simple helper function and call it appropriately. This slightly reduces the spaghetti nature of xfs_rename. Signed-off-by: Dave Chinner Reviewed-by: Eric Sandeen Reviewed-by: Brian Foster Signed-off-by: Dave Chinner --- fs/xfs/xfs_inode.c | 48 +++++++++++++++++++++++++++--------------------- 1 file changed, 27 insertions(+), 21 deletions(-) diff --git a/fs/xfs/xfs_inode.c b/fs/xfs/xfs_inode.c index 426dbf7d094a..c3fe00cb19e4 100644 --- a/fs/xfs/xfs_inode.c +++ b/fs/xfs/xfs_inode.c @@ -2650,6 +2650,31 @@ xfs_sort_for_rename( } } +static int +xfs_finish_rename( + struct xfs_trans *tp, + struct xfs_bmap_free *free_list) +{ + int committed = 0; + int error; + + /* + * If this is a synchronous mount, make sure that the rename transaction + * goes to disk before returning to the user. + */ + if (tp->t_mountp->m_flags & (XFS_MOUNT_WSYNC|XFS_MOUNT_DIRSYNC)) + xfs_trans_set_sync(tp); + + error = xfs_bmap_finish(&tp, free_list, &committed); + if (error) { + xfs_bmap_cancel(free_list); + xfs_trans_cancel(tp, XFS_TRANS_RELEASE_LOG_RES|XFS_TRANS_ABORT); + return error; + } + + return xfs_trans_commit(tp, XFS_TRANS_RELEASE_LOG_RES); +} + /* * xfs_cross_rename() * @@ -2789,7 +2814,6 @@ xfs_rename( xfs_bmap_free_t free_list; xfs_fsblock_t first_block; int cancel_flags = 0; - int committed; xfs_inode_t *inodes[__XFS_SORT_INODES]; int num_inodes = __XFS_SORT_INODES; int spaceres; @@ -2862,7 +2886,7 @@ xfs_rename( &free_list, &first_block, spaceres); if (error) goto out_trans_abort; - goto finish_rename; + return xfs_finish_rename(tp, &free_list); } /* @@ -3004,25 +3028,7 @@ xfs_rename( if (new_parent) xfs_trans_log_inode(tp, target_dp, XFS_ILOG_CORE); -finish_rename: - /* - * If this is a synchronous mount, make sure that the - * rename transaction goes to disk before returning to - * the user. - */ - if (mp->m_flags & (XFS_MOUNT_WSYNC|XFS_MOUNT_DIRSYNC)) { - xfs_trans_set_sync(tp); - } - - error = xfs_bmap_finish(&tp, &free_list, &committed); - if (error) - goto out_trans_abort; - - /* - * trans_commit will unlock src_ip, target_ip & decrement - * the vnode references. - */ - return xfs_trans_commit(tp, XFS_TRANS_RELEASE_LOG_RES); + return xfs_finish_rename(tp, &free_list); out_trans_abort: cancel_flags |= XFS_TRANS_ABORT; -- cgit v1.2.3-59-g8ed1b From eeacd3217b8fa8143f5dc27ded405790c74f01e8 Mon Sep 17 00:00:00 2001 From: Dave Chinner Date: Wed, 25 Mar 2015 14:08:07 +1100 Subject: xfs: make xfs_cross_rename() complete fully Now that xfs_finish_rename() exists, there is no reason for xfs_cross_rename() to return to xfs_rename() to finish off the rename transaction. Drive the completion code into xfs_cross_rename() and handle all errors there so as to simplify the xfs_rename() code. Further, push the rename exchange target_ip check to early in the rename code so as to make the error handling easy and obviously correct. Signed-off-by: Dave Chinner Reviewed-by: Eric Sandeen Reviewed-by: Brian Foster Signed-off-by: Dave Chinner --- fs/xfs/xfs_inode.c | 41 +++++++++++++++++++++-------------------- 1 file changed, 21 insertions(+), 20 deletions(-) diff --git a/fs/xfs/xfs_inode.c b/fs/xfs/xfs_inode.c index c3fe00cb19e4..b376ebe5fad9 100644 --- a/fs/xfs/xfs_inode.c +++ b/fs/xfs/xfs_inode.c @@ -2703,14 +2703,14 @@ xfs_cross_rename( ip2->i_ino, first_block, free_list, spaceres); if (error) - goto out; + goto out_trans_abort; /* Swap inode number for dirent in second parent */ error = xfs_dir_replace(tp, dp2, name2, ip1->i_ino, first_block, free_list, spaceres); if (error) - goto out; + goto out_trans_abort; /* * If we're renaming one or more directories across different parents, @@ -2725,16 +2725,16 @@ xfs_cross_rename( dp1->i_ino, first_block, free_list, spaceres); if (error) - goto out; + goto out_trans_abort; /* transfer ip2 ".." reference to dp1 */ if (!S_ISDIR(ip1->i_d.di_mode)) { error = xfs_droplink(tp, dp2); if (error) - goto out; + goto out_trans_abort; error = xfs_bumplink(tp, dp1); if (error) - goto out; + goto out_trans_abort; } /* @@ -2752,16 +2752,16 @@ xfs_cross_rename( dp2->i_ino, first_block, free_list, spaceres); if (error) - goto out; + goto out_trans_abort; /* transfer ip1 ".." reference to dp2 */ if (!S_ISDIR(ip2->i_d.di_mode)) { error = xfs_droplink(tp, dp1); if (error) - goto out; + goto out_trans_abort; error = xfs_bumplink(tp, dp2); if (error) - goto out; + goto out_trans_abort; } /* @@ -2789,7 +2789,11 @@ xfs_cross_rename( } xfs_trans_ichgtime(tp, dp1, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG); xfs_trans_log_inode(tp, dp1, XFS_ILOG_CORE); -out: + return xfs_finish_rename(tp, free_list); + +out_trans_abort: + xfs_bmap_cancel(free_list); + xfs_trans_cancel(tp, XFS_TRANS_RELEASE_LOG_RES|XFS_TRANS_ABORT); return error; } @@ -2820,6 +2824,9 @@ xfs_rename( trace_xfs_rename(src_dp, target_dp, src_name, target_name); + if ((flags & RENAME_EXCHANGE) && !target_ip) + return -EINVAL; + new_parent = (src_dp != target_dp); src_is_directory = S_ISDIR(src_ip->i_d.di_mode); @@ -2877,17 +2884,11 @@ xfs_rename( xfs_bmap_init(&free_list, &first_block); - /* - * Handle RENAME_EXCHANGE flags - */ - if (flags & RENAME_EXCHANGE) { - error = xfs_cross_rename(tp, src_dp, src_name, src_ip, - target_dp, target_name, target_ip, - &free_list, &first_block, spaceres); - if (error) - goto out_trans_abort; - return xfs_finish_rename(tp, &free_list); - } + /* RENAME_EXCHANGE is unique from here on. */ + if (flags & RENAME_EXCHANGE) + return xfs_cross_rename(tp, src_dp, src_name, src_ip, + target_dp, target_name, target_ip, + &free_list, &first_block, spaceres); /* * Set up the target. -- cgit v1.2.3-59-g8ed1b From 7dcf5c3e4527cfa2807567b00387cf2ed5e07f00 Mon Sep 17 00:00:00 2001 From: Dave Chinner Date: Wed, 25 Mar 2015 14:08:08 +1100 Subject: xfs: add RENAME_WHITEOUT support Whiteouts are used by overlayfs - it has a crazy convention that a whiteout is a character device inode with a major:minor of 0:0. Because it's not documented anywhere, here's an example of what RENAME_WHITEOUT does on ext4: # echo foo > /mnt/scratch/foo # echo bar > /mnt/scratch/bar # ls -l /mnt/scratch total 24 -rw-r--r-- 1 root root 4 Feb 11 20:22 bar -rw-r--r-- 1 root root 4 Feb 11 20:22 foo drwx------ 2 root root 16384 Feb 11 20:18 lost+found # src/renameat2 -w /mnt/scratch/foo /mnt/scratch/bar # ls -l /mnt/scratch total 20 -rw-r--r-- 1 root root 4 Feb 11 20:22 bar c--------- 1 root root 0, 0 Feb 11 20:23 foo drwx------ 2 root root 16384 Feb 11 20:18 lost+found # cat /mnt/scratch/bar foo # In XFS rename terms, the operation that has been done is that source (foo) has been moved to the target (bar), which is like a nomal rename operation, but rather than the source being removed, it have been replaced with a whiteout. We can't allocate whiteout inodes within the rename transaction due to allocation being a multi-commit transaction: rename needs to be a single, atomic commit. Hence we have several options here, form most efficient to least efficient: - use DT_WHT in the target dirent and do no whiteout inode allocation. The main issue with this approach is that we need hooks in lookup to create a virtual chardev inode to present to userspace and in places where we might need to modify the dirent e.g. unlink. Overlayfs also needs to be taught about DT_WHT. Most invasive change, lowest overhead. - create a special whiteout inode in the root directory (e.g. a ".wino" dirent) and then hardlink every new whiteout to it. This means we only need to create a single whiteout inode, and rename simply creates a hardlink to it. We can use DT_WHT for these, though using DT_CHR means we won't have to modify overlayfs, nor anything in userspace. Downside is we have to look up the whiteout inode on every operation and create it if it doesn't exist. - copy ext4: create a special whiteout chardev inode for every whiteout. This is more complex than the above options because of the lack of atomicity between inode creation and the rename operation, requiring us to create a tmpfile inode and then linking it into the directory structure during the rename. At least with a tmpfile inode crashes between the create and rename doesn't leave unreferenced inodes or directory pollution around. By far the simplest thing to do in the short term is to copy ext4. While it is the most inefficient way of supporting whiteouts, but as an initial implementation we can simply reuse existing functions and add a small amount of extra code the the rename operation. When we get full whiteout support in the VFS (via the dentry cache) we can then look to supporting DT_WHT method outlined as the first method of supporting whiteouts. But until then, we'll stick with what overlayfs expects us to be: dumb and stupid. Signed-off-by: Dave Chinner --- fs/xfs/xfs_inode.c | 129 +++++++++++++++++++++++++++++++++++++++++++---------- fs/xfs/xfs_iops.c | 2 +- 2 files changed, 107 insertions(+), 24 deletions(-) diff --git a/fs/xfs/xfs_inode.c b/fs/xfs/xfs_inode.c index b376ebe5fad9..ded129d5ec0d 100644 --- a/fs/xfs/xfs_inode.c +++ b/fs/xfs/xfs_inode.c @@ -2797,40 +2797,81 @@ out_trans_abort: return error; } +/* + * xfs_rename_alloc_whiteout() + * + * Return a referenced, unlinked, unlocked inode that that can be used as a + * whiteout in a rename transaction. We use a tmpfile inode here so that if we + * crash between allocating the inode and linking it into the rename transaction + * recovery will free the inode and we won't leak it. + */ +static int +xfs_rename_alloc_whiteout( + struct xfs_inode *dp, + struct xfs_inode **wip) +{ + struct xfs_inode *tmpfile; + int error; + + error = xfs_create_tmpfile(dp, NULL, S_IFCHR | WHITEOUT_MODE, &tmpfile); + if (error) + return error; + + /* Satisfy xfs_bumplink that this is a real tmpfile */ + xfs_finish_inode_setup(tmpfile); + VFS_I(tmpfile)->i_state |= I_LINKABLE; + + *wip = tmpfile; + return 0; +} + /* * xfs_rename */ int xfs_rename( - xfs_inode_t *src_dp, - struct xfs_name *src_name, - xfs_inode_t *src_ip, - xfs_inode_t *target_dp, - struct xfs_name *target_name, - xfs_inode_t *target_ip, - unsigned int flags) + struct xfs_inode *src_dp, + struct xfs_name *src_name, + struct xfs_inode *src_ip, + struct xfs_inode *target_dp, + struct xfs_name *target_name, + struct xfs_inode *target_ip, + unsigned int flags) { - xfs_trans_t *tp = NULL; - xfs_mount_t *mp = src_dp->i_mount; - int new_parent; /* moving to a new dir */ - int src_is_directory; /* src_name is a directory */ - int error; - xfs_bmap_free_t free_list; - xfs_fsblock_t first_block; - int cancel_flags = 0; - xfs_inode_t *inodes[__XFS_SORT_INODES]; - int num_inodes = __XFS_SORT_INODES; - int spaceres; + struct xfs_mount *mp = src_dp->i_mount; + struct xfs_trans *tp; + struct xfs_bmap_free free_list; + xfs_fsblock_t first_block; + struct xfs_inode *wip = NULL; /* whiteout inode */ + struct xfs_inode *inodes[__XFS_SORT_INODES]; + int num_inodes = __XFS_SORT_INODES; + int new_parent = (src_dp != target_dp); + int src_is_directory = S_ISDIR(src_ip->i_d.di_mode); + int cancel_flags = 0; + int spaceres; + int error; trace_xfs_rename(src_dp, target_dp, src_name, target_name); if ((flags & RENAME_EXCHANGE) && !target_ip) return -EINVAL; - new_parent = (src_dp != target_dp); - src_is_directory = S_ISDIR(src_ip->i_d.di_mode); + /* + * If we are doing a whiteout operation, allocate the whiteout inode + * we will be placing at the target and ensure the type is set + * appropriately. + */ + if (flags & RENAME_WHITEOUT) { + ASSERT(!(flags & (RENAME_NOREPLACE | RENAME_EXCHANGE))); + error = xfs_rename_alloc_whiteout(target_dp, &wip); + if (error) + return error; + + /* setup target dirent info as whiteout */ + src_name->type = XFS_DIR3_FT_CHRDEV; + } - xfs_sort_for_rename(src_dp, target_dp, src_ip, target_ip, NULL, + xfs_sort_for_rename(src_dp, target_dp, src_ip, target_ip, wip, inodes, &num_inodes); tp = xfs_trans_alloc(mp, XFS_TRANS_RENAME); @@ -2870,6 +2911,8 @@ xfs_rename( xfs_trans_ijoin(tp, src_ip, XFS_ILOCK_EXCL); if (target_ip) xfs_trans_ijoin(tp, target_ip, XFS_ILOCK_EXCL); + if (wip) + xfs_trans_ijoin(tp, wip, XFS_ILOCK_EXCL); /* * If we are using project inheritance, we only allow renames @@ -3019,17 +3062,55 @@ xfs_rename( goto out_trans_abort; } - error = xfs_dir_removename(tp, src_dp, src_name, src_ip->i_ino, + /* + * For whiteouts, we only need to update the source dirent with the + * inode number of the whiteout inode rather than removing it + * altogether. + */ + if (wip) { + error = xfs_dir_replace(tp, src_dp, src_name, wip->i_ino, &first_block, &free_list, spaceres); + } else + error = xfs_dir_removename(tp, src_dp, src_name, src_ip->i_ino, + &first_block, &free_list, spaceres); if (error) goto out_trans_abort; + /* + * For whiteouts, we need to bump the link count on the whiteout inode. + * This means that failures all the way up to this point leave the inode + * on the unlinked list and so cleanup is a simple matter of dropping + * the remaining reference to it. If we fail here after bumping the link + * count, we're shutting down the filesystem so we'll never see the + * intermediate state on disk. + */ + if (wip) { + ASSERT(wip->i_d.di_nlink == 0); + error = xfs_bumplink(tp, wip); + if (error) + goto out_trans_abort; + error = xfs_iunlink_remove(tp, wip); + if (error) + goto out_trans_abort; + xfs_trans_log_inode(tp, wip, XFS_ILOG_CORE); + + /* + * Now we have a real link, clear the "I'm a tmpfile" state + * flag from the inode so it doesn't accidentally get misused in + * future. + */ + VFS_I(wip)->i_state &= ~I_LINKABLE; + } + xfs_trans_ichgtime(tp, src_dp, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG); xfs_trans_log_inode(tp, src_dp, XFS_ILOG_CORE); if (new_parent) xfs_trans_log_inode(tp, target_dp, XFS_ILOG_CORE); - return xfs_finish_rename(tp, &free_list); + error = xfs_finish_rename(tp, &free_list); + if (wip) + IRELE(wip); + return error; out_trans_abort: cancel_flags |= XFS_TRANS_ABORT; @@ -3037,6 +3118,8 @@ out_bmap_cancel: xfs_bmap_cancel(&free_list); out_trans_cancel: xfs_trans_cancel(tp, cancel_flags); + if (wip) + IRELE(wip); return error; } diff --git a/fs/xfs/xfs_iops.c b/fs/xfs/xfs_iops.c index d7782ae1af3c..9bcad7132f75 100644 --- a/fs/xfs/xfs_iops.c +++ b/fs/xfs/xfs_iops.c @@ -394,7 +394,7 @@ xfs_vn_rename( struct xfs_name oname; struct xfs_name nname; - if (flags & ~(RENAME_NOREPLACE | RENAME_EXCHANGE)) + if (flags & ~(RENAME_NOREPLACE | RENAME_EXCHANGE | RENAME_WHITEOUT)) return -EINVAL; /* if we are exchanging files, we need to set i_mode of both files */ -- cgit v1.2.3-59-g8ed1b From d64588ca28bcd58f100f7933d0c18c8b504162bb Mon Sep 17 00:00:00 2001 From: Dave Chinner Date: Wed, 25 Mar 2015 14:53:48 +1100 Subject: xfs: remove xfs_bmap_sanity_check() This code is redundant now that we have verifiers that sanity check the buffers as they are read from disk. Signed-off-by: Dave Chinner Reviewed-by: Brian Foster Reviewed-by: Carlos Maiolino Signed-off-by: Dave Chinner --- fs/xfs/libxfs/xfs_bmap.c | 33 --------------------------------- 1 file changed, 33 deletions(-) diff --git a/fs/xfs/libxfs/xfs_bmap.c b/fs/xfs/libxfs/xfs_bmap.c index b8e97fd0bac1..990595548958 100644 --- a/fs/xfs/libxfs/xfs_bmap.c +++ b/fs/xfs/libxfs/xfs_bmap.c @@ -244,30 +244,6 @@ xfs_bmap_forkoff_reset( } } -/* - * Debug/sanity checking code - */ - -STATIC int -xfs_bmap_sanity_check( - struct xfs_mount *mp, - struct xfs_buf *bp, - int level) -{ - struct xfs_btree_block *block = XFS_BUF_TO_BLOCK(bp); - - if (block->bb_magic != cpu_to_be32(XFS_BMAP_CRC_MAGIC) && - block->bb_magic != cpu_to_be32(XFS_BMAP_MAGIC)) - return 0; - - if (be16_to_cpu(block->bb_level) != level || - be16_to_cpu(block->bb_numrecs) == 0 || - be16_to_cpu(block->bb_numrecs) > mp->m_bmap_dmxr[level != 0]) - return 0; - - return 1; -} - #ifdef DEBUG STATIC struct xfs_buf * xfs_bmap_get_bp( @@ -410,9 +386,6 @@ xfs_bmap_check_leaf_extents( goto error_norelse; } block = XFS_BUF_TO_BLOCK(bp); - XFS_WANT_CORRUPTED_GOTO( - xfs_bmap_sanity_check(mp, bp, level), - error0); if (level == 0) break; @@ -1311,9 +1284,6 @@ xfs_bmap_read_extents( if (error) return error; block = XFS_BUF_TO_BLOCK(bp); - XFS_WANT_CORRUPTED_GOTO( - xfs_bmap_sanity_check(mp, bp, level), - error0); if (level == 0) break; pp = XFS_BMBT_PTR_ADDR(mp, block, 1, mp->m_bmap_dmxr[1]); @@ -1345,9 +1315,6 @@ xfs_bmap_read_extents( XFS_ERRLEVEL_LOW, ip->i_mount, block); goto error0; } - XFS_WANT_CORRUPTED_GOTO( - xfs_bmap_sanity_check(mp, bp, 0), - error0); /* * Read-ahead the next leaf block, if any. */ -- cgit v1.2.3-59-g8ed1b From b26384dc52edba2f5fcc2b38eccc98e1f44bb379 Mon Sep 17 00:00:00 2001 From: Eric Sandeen Date: Wed, 25 Mar 2015 14:54:25 +1100 Subject: xfs: fix NULL pointer dereference in xfs_filestream_lookup_ag() If xfs_filestream_get_parent() fails, we have a null pip, goto out, and attempt to IRELE(NULL). This causes a null pointer dereference and BUG(). Fix this by directly returning NULLAGNUMBER in this case. Reported-by: Adrien Nader Signed-off-by: Eric Sandeen Reviewed-by: Dave Chinner Signed-off-by: Dave Chinner --- fs/xfs/xfs_filestream.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/fs/xfs/xfs_filestream.c b/fs/xfs/xfs_filestream.c index a2e86e8a0fea..8f9f854376c6 100644 --- a/fs/xfs/xfs_filestream.c +++ b/fs/xfs/xfs_filestream.c @@ -322,7 +322,7 @@ xfs_filestream_lookup_ag( pip = xfs_filestream_get_parent(ip); if (!pip) - goto out; + return NULLAGNUMBER; mru = xfs_mru_cache_lookup(mp->m_filestream, pip->i_ino); if (mru) { -- cgit v1.2.3-59-g8ed1b From 86aaf02e57d25a3b4094ca76ff805ee2d7aa30f8 Mon Sep 17 00:00:00 2001 From: Fabian Frederick Date: Wed, 25 Mar 2015 14:54:53 +1100 Subject: xfs: use bool instead of int in xfs_rename() new_parent and src_is_directory are only used in 0/1 context. Signed-off-by: Fabian Frederick Reviewed-by: Dave Chinner Signed-off-by: Dave Chinner --- fs/xfs/xfs_inode.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/fs/xfs/xfs_inode.c b/fs/xfs/xfs_inode.c index daafa1f6d260..b7064f20de7f 100644 --- a/fs/xfs/xfs_inode.c +++ b/fs/xfs/xfs_inode.c @@ -2790,8 +2790,8 @@ xfs_rename( { xfs_trans_t *tp = NULL; xfs_mount_t *mp = src_dp->i_mount; - int new_parent; /* moving to a new dir */ - int src_is_directory; /* src_name is a directory */ + bool new_parent; /* moving to a new dir */ + bool src_is_directory; /* src_name is a directory */ int error; xfs_bmap_free_t free_list; xfs_fsblock_t first_block; -- cgit v1.2.3-59-g8ed1b From 29916df08db4726c92bdfe72ce524b4b6a9e3c54 Mon Sep 17 00:00:00 2001 From: Fabian Frederick Date: Wed, 25 Mar 2015 14:55:25 +1100 Subject: xfs: fix shadow warning in xfs_da3_root_split() Use icnodehdr for struct xfs_da3_icnode_hdr instead of nodehdr (already declared above). Signed-off-by: Fabian Frederick Reviewed-by: Dave Chinner Signed-off-by: Dave Chinner --- fs/xfs/libxfs/xfs_da_btree.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) diff --git a/fs/xfs/libxfs/xfs_da_btree.c b/fs/xfs/libxfs/xfs_da_btree.c index 9cb0115c6bd1..2385f8cd08ab 100644 --- a/fs/xfs/libxfs/xfs_da_btree.c +++ b/fs/xfs/libxfs/xfs_da_btree.c @@ -538,12 +538,12 @@ xfs_da3_root_split( oldroot = blk1->bp->b_addr; if (oldroot->hdr.info.magic == cpu_to_be16(XFS_DA_NODE_MAGIC) || oldroot->hdr.info.magic == cpu_to_be16(XFS_DA3_NODE_MAGIC)) { - struct xfs_da3_icnode_hdr nodehdr; + struct xfs_da3_icnode_hdr icnodehdr; - dp->d_ops->node_hdr_from_disk(&nodehdr, oldroot); + dp->d_ops->node_hdr_from_disk(&icnodehdr, oldroot); btree = dp->d_ops->node_tree_p(oldroot); - size = (int)((char *)&btree[nodehdr.count] - (char *)oldroot); - level = nodehdr.level; + size = (int)((char *)&btree[icnodehdr.count] - (char *)oldroot); + level = icnodehdr.level; /* * we are about to copy oldroot to bp, so set up the type -- cgit v1.2.3-59-g8ed1b From 65dd297ac25565701fead5e4ee69b9ca62729f0e Mon Sep 17 00:00:00 2001 From: Scott Wood Date: Wed, 25 Mar 2015 14:56:21 +1100 Subject: xfs: %pF is only for function pointers Use %pS for actual addresses, otherwise you'll get bad output on arches like ppc64 where %pF expects a function descriptor. Signed-off-by: Scott Wood Reviewed-by: Dave Chinner Signed-off-by: Dave Chinner --- fs/xfs/xfs_error.c | 2 +- fs/xfs/xfs_trace.h | 20 ++++++++++---------- 2 files changed, 11 insertions(+), 11 deletions(-) diff --git a/fs/xfs/xfs_error.c b/fs/xfs/xfs_error.c index 3ee186ac1093..338e50bbfd1e 100644 --- a/fs/xfs/xfs_error.c +++ b/fs/xfs/xfs_error.c @@ -131,7 +131,7 @@ xfs_error_report( { if (level <= xfs_error_level) { xfs_alert_tag(mp, XFS_PTAG_ERROR_REPORT, - "Internal error %s at line %d of file %s. Caller %pF", + "Internal error %s at line %d of file %s. Caller %pS", tag, linenum, filename, ra); xfs_stack_trace(); diff --git a/fs/xfs/xfs_trace.h b/fs/xfs/xfs_trace.h index 51372e34d988..b5ac81eeb061 100644 --- a/fs/xfs/xfs_trace.h +++ b/fs/xfs/xfs_trace.h @@ -115,7 +115,7 @@ DECLARE_EVENT_CLASS(xfs_perag_class, __entry->refcount = refcount; __entry->caller_ip = caller_ip; ), - TP_printk("dev %d:%d agno %u refcount %d caller %pf", + TP_printk("dev %d:%d agno %u refcount %d caller %ps", MAJOR(__entry->dev), MINOR(__entry->dev), __entry->agno, __entry->refcount, @@ -239,7 +239,7 @@ TRACE_EVENT(xfs_iext_insert, __entry->caller_ip = caller_ip; ), TP_printk("dev %d:%d ino 0x%llx state %s idx %ld " - "offset %lld block %lld count %lld flag %d caller %pf", + "offset %lld block %lld count %lld flag %d caller %ps", MAJOR(__entry->dev), MINOR(__entry->dev), __entry->ino, __print_flags(__entry->bmap_state, "|", XFS_BMAP_EXT_FLAGS), @@ -283,7 +283,7 @@ DECLARE_EVENT_CLASS(xfs_bmap_class, __entry->caller_ip = caller_ip; ), TP_printk("dev %d:%d ino 0x%llx state %s idx %ld " - "offset %lld block %lld count %lld flag %d caller %pf", + "offset %lld block %lld count %lld flag %d caller %ps", MAJOR(__entry->dev), MINOR(__entry->dev), __entry->ino, __print_flags(__entry->bmap_state, "|", XFS_BMAP_EXT_FLAGS), @@ -329,7 +329,7 @@ DECLARE_EVENT_CLASS(xfs_buf_class, __entry->caller_ip = caller_ip; ), TP_printk("dev %d:%d bno 0x%llx nblks 0x%x hold %d pincount %d " - "lock %d flags %s caller %pf", + "lock %d flags %s caller %ps", MAJOR(__entry->dev), MINOR(__entry->dev), (unsigned long long)__entry->bno, __entry->nblks, @@ -402,7 +402,7 @@ DECLARE_EVENT_CLASS(xfs_buf_flags_class, __entry->caller_ip = caller_ip; ), TP_printk("dev %d:%d bno 0x%llx len 0x%zx hold %d pincount %d " - "lock %d flags %s caller %pf", + "lock %d flags %s caller %ps", MAJOR(__entry->dev), MINOR(__entry->dev), (unsigned long long)__entry->bno, __entry->buffer_length, @@ -447,7 +447,7 @@ TRACE_EVENT(xfs_buf_ioerror, __entry->caller_ip = caller_ip; ), TP_printk("dev %d:%d bno 0x%llx len 0x%zx hold %d pincount %d " - "lock %d error %d flags %s caller %pf", + "lock %d error %d flags %s caller %ps", MAJOR(__entry->dev), MINOR(__entry->dev), (unsigned long long)__entry->bno, __entry->buffer_length, @@ -613,7 +613,7 @@ DECLARE_EVENT_CLASS(xfs_lock_class, __entry->lock_flags = lock_flags; __entry->caller_ip = caller_ip; ), - TP_printk("dev %d:%d ino 0x%llx flags %s caller %pf", + TP_printk("dev %d:%d ino 0x%llx flags %s caller %ps", MAJOR(__entry->dev), MINOR(__entry->dev), __entry->ino, __print_flags(__entry->lock_flags, "|", XFS_LOCK_FLAGS), @@ -702,7 +702,7 @@ DECLARE_EVENT_CLASS(xfs_iref_class, __entry->pincount = atomic_read(&ip->i_pincount); __entry->caller_ip = caller_ip; ), - TP_printk("dev %d:%d ino 0x%llx count %d pincount %d caller %pf", + TP_printk("dev %d:%d ino 0x%llx count %d pincount %d caller %ps", MAJOR(__entry->dev), MINOR(__entry->dev), __entry->ino, __entry->count, @@ -1333,7 +1333,7 @@ TRACE_EVENT(xfs_bunmap, __entry->flags = flags; ), TP_printk("dev %d:%d ino 0x%llx size 0x%llx bno 0x%llx len 0x%llx" - "flags %s caller %pf", + "flags %s caller %ps", MAJOR(__entry->dev), MINOR(__entry->dev), __entry->ino, __entry->size, @@ -1466,7 +1466,7 @@ TRACE_EVENT(xfs_agf, ), TP_printk("dev %d:%d agno %u flags %s length %u roots b %u c %u " "levels b %u c %u flfirst %u fllast %u flcount %u " - "freeblks %u longest %u caller %pf", + "freeblks %u longest %u caller %ps", MAJOR(__entry->dev), MINOR(__entry->dev), __entry->agno, __print_flags(__entry->flags, "|", XFS_AGF_FLAGS), -- cgit v1.2.3-59-g8ed1b From 20dafeefac97d1a690b113f2a954dc84fdf8f290 Mon Sep 17 00:00:00 2001 From: Byoungyoung Lee Date: Wed, 25 Mar 2015 14:57:53 +1100 Subject: xfs: xfs_mru_cache_insert() should use GFP_NOFS xfs_mru_cache_insert() can be called from within transaction context during block allocation like so: write(2) .... xfs_get_blocks xfs_iomap_write_direct start transaction xfs_bmapi_write xfs_bmapi_allocate xfs_bmap_btalloc xfs_bmap_btalloc_filestreams xfs_filestream_new_ag xfs_filestream_pick_ag xfs_mru_cache_insert radix_tree_preload(GFP_KERNEL) In this case, GFP_KERNEL is incorrect and can potentially lead to deadlocks in memory reclaim. It should use GFP_NOFS allocations to avoid lock recursion problems. [dchinner: rewrote commit message] Signed-off-by: Byoungyoung Lee Signed-off-by: Sanidhya Kashyap Reviewed-by: Dave Chinner Signed-off-by: Dave Chinner --- fs/xfs/xfs_mru_cache.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/fs/xfs/xfs_mru_cache.c b/fs/xfs/xfs_mru_cache.c index 30ecca3037e3..f8a674d7f092 100644 --- a/fs/xfs/xfs_mru_cache.c +++ b/fs/xfs/xfs_mru_cache.c @@ -437,7 +437,7 @@ xfs_mru_cache_insert( if (!mru || !mru->lists) return -EINVAL; - if (radix_tree_preload(GFP_KERNEL)) + if (radix_tree_preload(GFP_NOFS)) return -ENOMEM; INIT_LIST_HEAD(&elem->list_node); -- cgit v1.2.3-59-g8ed1b From 5e9383f97e773e9a5385144ef5561f2ac0ee1349 Mon Sep 17 00:00:00 2001 From: Joe Perches Date: Wed, 25 Mar 2015 15:00:24 +1100 Subject: xfs: Fix incorrect positive ENOMEM return added a positive error return value. This value filters up through the return layers and should be negative as the other return values are in the same function. Signed-off-by: Joe Perches Reviewed-by: Dave Chinner Signed-off-by: Dave Chinner --- fs/xfs/xfs_super.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/fs/xfs/xfs_super.c b/fs/xfs/xfs_super.c index 53c56a913778..194291381252 100644 --- a/fs/xfs/xfs_super.c +++ b/fs/xfs/xfs_super.c @@ -1398,7 +1398,7 @@ xfs_init_percpu_counters( error = percpu_counter_init(&mp->m_icount, 0, GFP_KERNEL); if (error) - return ENOMEM; + return -ENOMEM; error = percpu_counter_init(&mp->m_ifree, 0, GFP_KERNEL); if (error) -- cgit v1.2.3-59-g8ed1b From dd46c787788d5bf5b974729d43e4c405814a4c7d Mon Sep 17 00:00:00 2001 From: Namjae Jeon Date: Wed, 25 Mar 2015 15:07:05 +1100 Subject: fs: Add support FALLOC_FL_INSERT_RANGE for fallocate FALLOC_FL_INSERT_RANGE command is the opposite command of FALLOC_FL_COLLAPSE_RANGE that is needed for someone who wants to add some data in the middle of file. FALLOC_FL_INSERT_RANGE will create space for writing new data within a file after shifting extents to right as given length. This command also has same limitations as FALLOC_FL_COLLAPSE_RANGE in that operations need to be filesystem block boundary aligned and cannot cross the current EOF. Signed-off-by: Namjae Jeon Signed-off-by: Ashish Sangwan Reviewed-by: Dave Chinner Signed-off-by: Dave Chinner --- fs/open.c | 8 ++++++-- include/linux/falloc.h | 6 ++++++ include/uapi/linux/falloc.h | 17 +++++++++++++++++ 3 files changed, 29 insertions(+), 2 deletions(-) diff --git a/fs/open.c b/fs/open.c index 33f9cbf2610b..b724cc0e0228 100644 --- a/fs/open.c +++ b/fs/open.c @@ -231,8 +231,7 @@ int vfs_fallocate(struct file *file, int mode, loff_t offset, loff_t len) return -EINVAL; /* Return error if mode is not supported */ - if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE | - FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE)) + if (mode & ~FALLOC_FL_SUPPORTED_MASK) return -EOPNOTSUPP; /* Punch hole and zero range are mutually exclusive */ @@ -250,6 +249,11 @@ int vfs_fallocate(struct file *file, int mode, loff_t offset, loff_t len) (mode & ~FALLOC_FL_COLLAPSE_RANGE)) return -EINVAL; + /* Insert range should only be used exclusively. */ + if ((mode & FALLOC_FL_INSERT_RANGE) && + (mode & ~FALLOC_FL_INSERT_RANGE)) + return -EINVAL; + if (!(file->f_mode & FMODE_WRITE)) return -EBADF; diff --git a/include/linux/falloc.h b/include/linux/falloc.h index 31591686ac2d..996111000a8c 100644 --- a/include/linux/falloc.h +++ b/include/linux/falloc.h @@ -21,4 +21,10 @@ struct space_resv { #define FS_IOC_RESVSP _IOW('X', 40, struct space_resv) #define FS_IOC_RESVSP64 _IOW('X', 42, struct space_resv) +#define FALLOC_FL_SUPPORTED_MASK (FALLOC_FL_KEEP_SIZE | \ + FALLOC_FL_PUNCH_HOLE | \ + FALLOC_FL_COLLAPSE_RANGE | \ + FALLOC_FL_ZERO_RANGE | \ + FALLOC_FL_INSERT_RANGE) + #endif /* _FALLOC_H_ */ diff --git a/include/uapi/linux/falloc.h b/include/uapi/linux/falloc.h index d1197ae3723c..3e445a760f14 100644 --- a/include/uapi/linux/falloc.h +++ b/include/uapi/linux/falloc.h @@ -41,4 +41,21 @@ */ #define FALLOC_FL_ZERO_RANGE 0x10 +/* + * FALLOC_FL_INSERT_RANGE is use to insert space within the file size without + * overwriting any existing data. The contents of the file beyond offset are + * shifted towards right by len bytes to create a hole. As such, this + * operation will increase the size of the file by len bytes. + * + * Different filesystems may implement different limitations on the granularity + * of the operation. Most will limit operations to filesystem block size + * boundaries, but this boundary may be larger or smaller depending on + * the filesystem and/or the configuration of the filesystem or file. + * + * Attempting to insert space using this flag at OR beyond the end of + * the file is considered an illegal operation - just use ftruncate(2) or + * fallocate(2) with mode 0 for such type of operations. + */ +#define FALLOC_FL_INSERT_RANGE 0x20 + #endif /* _UAPI_FALLOC_H_ */ -- cgit v1.2.3-59-g8ed1b From a904b1ca5751faf5ece8600e18cd3b674afcca1b Mon Sep 17 00:00:00 2001 From: Namjae Jeon Date: Wed, 25 Mar 2015 15:08:56 +1100 Subject: xfs: Add support FALLOC_FL_INSERT_RANGE for fallocate This patch implements fallocate's FALLOC_FL_INSERT_RANGE for XFS. 1) Make sure that both offset and len are block size aligned. 2) Update the i_size of inode by len bytes. 3) Compute the file's logical block number against offset. If the computed block number is not the starting block of the extent, split the extent such that the block number is the starting block of the extent. 4) Shift all the extents which are lying bewteen [offset, last allocated extent] towards right by len bytes. This step will make a hole of len bytes at offset. Signed-off-by: Namjae Jeon Signed-off-by: Ashish Sangwan Reviewed-by: Brian Foster Reviewed-by: Dave Chinner Signed-off-by: Dave Chinner --- fs/xfs/libxfs/xfs_bmap.c | 356 +++++++++++++++++++++++++++++++++++++++++------ fs/xfs/libxfs/xfs_bmap.h | 13 +- fs/xfs/xfs_bmap_util.c | 131 ++++++++++++----- fs/xfs/xfs_bmap_util.h | 2 + fs/xfs/xfs_file.c | 41 +++++- fs/xfs/xfs_trace.h | 1 + 6 files changed, 461 insertions(+), 83 deletions(-) diff --git a/fs/xfs/libxfs/xfs_bmap.c b/fs/xfs/libxfs/xfs_bmap.c index 60cfa90163b8..e8696f5a8041 100644 --- a/fs/xfs/libxfs/xfs_bmap.c +++ b/fs/xfs/libxfs/xfs_bmap.c @@ -5526,52 +5526,92 @@ xfs_bmse_shift_one( int *current_ext, struct xfs_bmbt_rec_host *gotp, struct xfs_btree_cur *cur, - int *logflags) + int *logflags, + enum shift_direction direction) { struct xfs_ifork *ifp; struct xfs_mount *mp; xfs_fileoff_t startoff; - struct xfs_bmbt_rec_host *leftp; + struct xfs_bmbt_rec_host *adj_irecp; struct xfs_bmbt_irec got; - struct xfs_bmbt_irec left; + struct xfs_bmbt_irec adj_irec; int error; int i; + int total_extents; mp = ip->i_mount; ifp = XFS_IFORK_PTR(ip, whichfork); + total_extents = ifp->if_bytes / sizeof(xfs_bmbt_rec_t); xfs_bmbt_get_all(gotp, &got); - startoff = got.br_startoff - offset_shift_fsb; /* delalloc extents should be prevented by caller */ XFS_WANT_CORRUPTED_RETURN(mp, !isnullstartblock(got.br_startblock)); - /* - * Check for merge if we've got an extent to the left, otherwise make - * sure there's enough room at the start of the file for the shift. - */ - if (*current_ext) { - /* grab the left extent and check for a large enough hole */ - leftp = xfs_iext_get_ext(ifp, *current_ext - 1); - xfs_bmbt_get_all(leftp, &left); + if (direction == SHIFT_LEFT) { + startoff = got.br_startoff - offset_shift_fsb; + + /* + * Check for merge if we've got an extent to the left, + * otherwise make sure there's enough room at the start + * of the file for the shift. + */ + if (!*current_ext) { + if (got.br_startoff < offset_shift_fsb) + return -EINVAL; + goto update_current_ext; + } + /* + * grab the left extent and check for a large + * enough hole. + */ + adj_irecp = xfs_iext_get_ext(ifp, *current_ext - 1); + xfs_bmbt_get_all(adj_irecp, &adj_irec); - if (startoff < left.br_startoff + left.br_blockcount) + if (startoff < + adj_irec.br_startoff + adj_irec.br_blockcount) return -EINVAL; /* check whether to merge the extent or shift it down */ - if (xfs_bmse_can_merge(&left, &got, offset_shift_fsb)) { + if (xfs_bmse_can_merge(&adj_irec, &got, + offset_shift_fsb)) { return xfs_bmse_merge(ip, whichfork, offset_shift_fsb, - *current_ext, gotp, leftp, cur, - logflags); + *current_ext, gotp, adj_irecp, + cur, logflags); } - } else if (got.br_startoff < offset_shift_fsb) - return -EINVAL; - + } else { + startoff = got.br_startoff + offset_shift_fsb; + /* nothing to move if this is the last extent */ + if (*current_ext >= (total_extents - 1)) + goto update_current_ext; + /* + * If this is not the last extent in the file, make sure there + * is enough room between current extent and next extent for + * accommodating the shift. + */ + adj_irecp = xfs_iext_get_ext(ifp, *current_ext + 1); + xfs_bmbt_get_all(adj_irecp, &adj_irec); + if (startoff + got.br_blockcount > adj_irec.br_startoff) + return -EINVAL; + /* + * Unlike a left shift (which involves a hole punch), + * a right shift does not modify extent neighbors + * in any way. We should never find mergeable extents + * in this scenario. Check anyways and warn if we + * encounter two extents that could be one. + */ + if (xfs_bmse_can_merge(&got, &adj_irec, offset_shift_fsb)) + WARN_ON_ONCE(1); + } /* * Increment the extent index for the next iteration, update the start * offset of the in-core extent and update the btree if applicable. */ - (*current_ext)++; +update_current_ext: + if (direction == SHIFT_LEFT) + (*current_ext)++; + else + (*current_ext)--; xfs_bmbt_set_startoff(gotp, startoff); *logflags |= XFS_ILOG_CORE; if (!cur) { @@ -5587,14 +5627,14 @@ xfs_bmse_shift_one( got.br_startoff = startoff; return xfs_bmbt_update(cur, got.br_startoff, got.br_startblock, - got.br_blockcount, got.br_state); + got.br_blockcount, got.br_state); } /* - * Shift extent records to the left to cover a hole. + * Shift extent records to the left/right to cover/create a hole. * * The maximum number of extents to be shifted in a single operation is - * @num_exts. @start_fsb specifies the file offset to start the shift and the + * @num_exts. @stop_fsb specifies the file offset at which to stop shift and the * file offset where we've left off is returned in @next_fsb. @offset_shift_fsb * is the length by which each extent is shifted. If there is no hole to shift * the extents into, this will be considered invalid operation and we abort @@ -5604,12 +5644,13 @@ int xfs_bmap_shift_extents( struct xfs_trans *tp, struct xfs_inode *ip, - xfs_fileoff_t start_fsb, + xfs_fileoff_t *next_fsb, xfs_fileoff_t offset_shift_fsb, int *done, - xfs_fileoff_t *next_fsb, + xfs_fileoff_t stop_fsb, xfs_fsblock_t *firstblock, struct xfs_bmap_free *flist, + enum shift_direction direction, int num_exts) { struct xfs_btree_cur *cur = NULL; @@ -5619,10 +5660,11 @@ xfs_bmap_shift_extents( struct xfs_ifork *ifp; xfs_extnum_t nexts = 0; xfs_extnum_t current_ext; + xfs_extnum_t total_extents; + xfs_extnum_t stop_extent; int error = 0; int whichfork = XFS_DATA_FORK; int logflags = 0; - int total_extents; if (unlikely(XFS_TEST_ERROR( (XFS_IFORK_FORMAT(ip, whichfork) != XFS_DINODE_FMT_EXTENTS && @@ -5638,6 +5680,8 @@ xfs_bmap_shift_extents( ASSERT(xfs_isilocked(ip, XFS_IOLOCK_EXCL)); ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL)); + ASSERT(direction == SHIFT_LEFT || direction == SHIFT_RIGHT); + ASSERT(*next_fsb != NULLFSBLOCK || direction == SHIFT_RIGHT); ifp = XFS_IFORK_PTR(ip, whichfork); if (!(ifp->if_flags & XFS_IFEXTENTS)) { @@ -5654,44 +5698,84 @@ xfs_bmap_shift_extents( cur->bc_private.b.flags = 0; } + /* + * There may be delalloc extents in the data fork before the range we + * are collapsing out, so we cannot use the count of real extents here. + * Instead we have to calculate it from the incore fork. + */ + total_extents = ifp->if_bytes / sizeof(xfs_bmbt_rec_t); + if (total_extents == 0) { + *done = 1; + goto del_cursor; + } + + /* + * In case of first right shift, we need to initialize next_fsb + */ + if (*next_fsb == NULLFSBLOCK) { + gotp = xfs_iext_get_ext(ifp, total_extents - 1); + xfs_bmbt_get_all(gotp, &got); + *next_fsb = got.br_startoff; + if (stop_fsb > *next_fsb) { + *done = 1; + goto del_cursor; + } + } + + /* Lookup the extent index at which we have to stop */ + if (direction == SHIFT_RIGHT) { + gotp = xfs_iext_bno_to_ext(ifp, stop_fsb, &stop_extent); + /* Make stop_extent exclusive of shift range */ + stop_extent--; + } else + stop_extent = total_extents; + /* * Look up the extent index for the fsb where we start shifting. We can * henceforth iterate with current_ext as extent list changes are locked * out via ilock. * * gotp can be null in 2 cases: 1) if there are no extents or 2) - * start_fsb lies in a hole beyond which there are no extents. Either + * *next_fsb lies in a hole beyond which there are no extents. Either * way, we are done. */ - gotp = xfs_iext_bno_to_ext(ifp, start_fsb, ¤t_ext); + gotp = xfs_iext_bno_to_ext(ifp, *next_fsb, ¤t_ext); if (!gotp) { *done = 1; goto del_cursor; } - /* - * There may be delalloc extents in the data fork before the range we - * are collapsing out, so we cannot use the count of real extents here. - * Instead we have to calculate it from the incore fork. - */ - total_extents = ifp->if_bytes / sizeof(xfs_bmbt_rec_t); - while (nexts++ < num_exts && current_ext < total_extents) { + /* some sanity checking before we finally start shifting extents */ + if ((direction == SHIFT_LEFT && current_ext >= stop_extent) || + (direction == SHIFT_RIGHT && current_ext <= stop_extent)) { + error = -EIO; + goto del_cursor; + } + + while (nexts++ < num_exts) { error = xfs_bmse_shift_one(ip, whichfork, offset_shift_fsb, - ¤t_ext, gotp, cur, &logflags); + ¤t_ext, gotp, cur, &logflags, + direction); if (error) goto del_cursor; + /* + * If there was an extent merge during the shift, the extent + * count can change. Update the total and grade the next record. + */ + if (direction == SHIFT_LEFT) { + total_extents = ifp->if_bytes / sizeof(xfs_bmbt_rec_t); + stop_extent = total_extents; + } - /* update total extent count and grab the next record */ - total_extents = ifp->if_bytes / sizeof(xfs_bmbt_rec_t); - if (current_ext >= total_extents) + if (current_ext == stop_extent) { + *done = 1; + *next_fsb = NULLFSBLOCK; break; + } gotp = xfs_iext_get_ext(ifp, current_ext); } - /* Check if we are done */ - if (current_ext == total_extents) { - *done = 1; - } else if (next_fsb) { + if (!*done) { xfs_bmbt_get_all(gotp, &got); *next_fsb = got.br_startoff; } @@ -5706,3 +5790,189 @@ del_cursor: return error; } + +/* + * Splits an extent into two extents at split_fsb block such that it is + * the first block of the current_ext. @current_ext is a target extent + * to be split. @split_fsb is a block where the extents is split. + * If split_fsb lies in a hole or the first block of extents, just return 0. + */ +STATIC int +xfs_bmap_split_extent_at( + struct xfs_trans *tp, + struct xfs_inode *ip, + xfs_fileoff_t split_fsb, + xfs_fsblock_t *firstfsb, + struct xfs_bmap_free *free_list) +{ + int whichfork = XFS_DATA_FORK; + struct xfs_btree_cur *cur = NULL; + struct xfs_bmbt_rec_host *gotp; + struct xfs_bmbt_irec got; + struct xfs_bmbt_irec new; /* split extent */ + struct xfs_mount *mp = ip->i_mount; + struct xfs_ifork *ifp; + xfs_fsblock_t gotblkcnt; /* new block count for got */ + xfs_extnum_t current_ext; + int error = 0; + int logflags = 0; + int i = 0; + + if (unlikely(XFS_TEST_ERROR( + (XFS_IFORK_FORMAT(ip, whichfork) != XFS_DINODE_FMT_EXTENTS && + XFS_IFORK_FORMAT(ip, whichfork) != XFS_DINODE_FMT_BTREE), + mp, XFS_ERRTAG_BMAPIFORMAT, XFS_RANDOM_BMAPIFORMAT))) { + XFS_ERROR_REPORT("xfs_bmap_split_extent_at", + XFS_ERRLEVEL_LOW, mp); + return -EFSCORRUPTED; + } + + if (XFS_FORCED_SHUTDOWN(mp)) + return -EIO; + + ifp = XFS_IFORK_PTR(ip, whichfork); + if (!(ifp->if_flags & XFS_IFEXTENTS)) { + /* Read in all the extents */ + error = xfs_iread_extents(tp, ip, whichfork); + if (error) + return error; + } + + /* + * gotp can be null in 2 cases: 1) if there are no extents + * or 2) split_fsb lies in a hole beyond which there are + * no extents. Either way, we are done. + */ + gotp = xfs_iext_bno_to_ext(ifp, split_fsb, ¤t_ext); + if (!gotp) + return 0; + + xfs_bmbt_get_all(gotp, &got); + + /* + * Check split_fsb lies in a hole or the start boundary offset + * of the extent. + */ + if (got.br_startoff >= split_fsb) + return 0; + + gotblkcnt = split_fsb - got.br_startoff; + new.br_startoff = split_fsb; + new.br_startblock = got.br_startblock + gotblkcnt; + new.br_blockcount = got.br_blockcount - gotblkcnt; + new.br_state = got.br_state; + + if (ifp->if_flags & XFS_IFBROOT) { + cur = xfs_bmbt_init_cursor(mp, tp, ip, whichfork); + cur->bc_private.b.firstblock = *firstfsb; + cur->bc_private.b.flist = free_list; + cur->bc_private.b.flags = 0; + error = xfs_bmbt_lookup_eq(cur, got.br_startoff, + got.br_startblock, + got.br_blockcount, + &i); + if (error) + goto del_cursor; + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, del_cursor); + } + + xfs_bmbt_set_blockcount(gotp, gotblkcnt); + got.br_blockcount = gotblkcnt; + + logflags = XFS_ILOG_CORE; + if (cur) { + error = xfs_bmbt_update(cur, got.br_startoff, + got.br_startblock, + got.br_blockcount, + got.br_state); + if (error) + goto del_cursor; + } else + logflags |= XFS_ILOG_DEXT; + + /* Add new extent */ + current_ext++; + xfs_iext_insert(ip, current_ext, 1, &new, 0); + XFS_IFORK_NEXT_SET(ip, whichfork, + XFS_IFORK_NEXTENTS(ip, whichfork) + 1); + + if (cur) { + error = xfs_bmbt_lookup_eq(cur, new.br_startoff, + new.br_startblock, new.br_blockcount, + &i); + if (error) + goto del_cursor; + XFS_WANT_CORRUPTED_GOTO(mp, i == 0, del_cursor); + cur->bc_rec.b.br_state = new.br_state; + + error = xfs_btree_insert(cur, &i); + if (error) + goto del_cursor; + XFS_WANT_CORRUPTED_GOTO(mp, i == 1, del_cursor); + } + + /* + * Convert to a btree if necessary. + */ + if (xfs_bmap_needs_btree(ip, whichfork)) { + int tmp_logflags; /* partial log flag return val */ + + ASSERT(cur == NULL); + error = xfs_bmap_extents_to_btree(tp, ip, firstfsb, free_list, + &cur, 0, &tmp_logflags, whichfork); + logflags |= tmp_logflags; + } + +del_cursor: + if (cur) { + cur->bc_private.b.allocated = 0; + xfs_btree_del_cursor(cur, + error ? XFS_BTREE_ERROR : XFS_BTREE_NOERROR); + } + + if (logflags) + xfs_trans_log_inode(tp, ip, logflags); + return error; +} + +int +xfs_bmap_split_extent( + struct xfs_inode *ip, + xfs_fileoff_t split_fsb) +{ + struct xfs_mount *mp = ip->i_mount; + struct xfs_trans *tp; + struct xfs_bmap_free free_list; + xfs_fsblock_t firstfsb; + int committed; + int error; + + tp = xfs_trans_alloc(mp, XFS_TRANS_DIOSTRAT); + error = xfs_trans_reserve(tp, &M_RES(mp)->tr_write, + XFS_DIOSTRAT_SPACE_RES(mp, 0), 0); + if (error) { + xfs_trans_cancel(tp, 0); + return error; + } + + xfs_ilock(ip, XFS_ILOCK_EXCL); + xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL); + + xfs_bmap_init(&free_list, &firstfsb); + + error = xfs_bmap_split_extent_at(tp, ip, split_fsb, + &firstfsb, &free_list); + if (error) + goto out; + + error = xfs_bmap_finish(&tp, &free_list, &committed); + if (error) + goto out; + + return xfs_trans_commit(tp, XFS_TRANS_RELEASE_LOG_RES); + + +out: + xfs_trans_cancel(tp, XFS_TRANS_RELEASE_LOG_RES | XFS_TRANS_ABORT); + return error; +} diff --git a/fs/xfs/libxfs/xfs_bmap.h b/fs/xfs/libxfs/xfs_bmap.h index b9d8a499d2c4..6aaa0c1c7200 100644 --- a/fs/xfs/libxfs/xfs_bmap.h +++ b/fs/xfs/libxfs/xfs_bmap.h @@ -166,6 +166,11 @@ static inline void xfs_bmap_init(xfs_bmap_free_t *flp, xfs_fsblock_t *fbp) */ #define XFS_BMAP_MAX_SHIFT_EXTENTS 1 +enum shift_direction { + SHIFT_LEFT = 0, + SHIFT_RIGHT, +}; + #ifdef DEBUG void xfs_bmap_trace_exlist(struct xfs_inode *ip, xfs_extnum_t cnt, int whichfork, unsigned long caller_ip); @@ -211,8 +216,10 @@ int xfs_check_nostate_extents(struct xfs_ifork *ifp, xfs_extnum_t idx, xfs_extnum_t num); uint xfs_default_attroffset(struct xfs_inode *ip); int xfs_bmap_shift_extents(struct xfs_trans *tp, struct xfs_inode *ip, - xfs_fileoff_t start_fsb, xfs_fileoff_t offset_shift_fsb, - int *done, xfs_fileoff_t *next_fsb, xfs_fsblock_t *firstblock, - struct xfs_bmap_free *flist, int num_exts); + xfs_fileoff_t *next_fsb, xfs_fileoff_t offset_shift_fsb, + int *done, xfs_fileoff_t stop_fsb, xfs_fsblock_t *firstblock, + struct xfs_bmap_free *flist, enum shift_direction direction, + int num_exts); +int xfs_bmap_split_extent(struct xfs_inode *ip, xfs_fileoff_t split_offset); #endif /* __XFS_BMAP_H__ */ diff --git a/fs/xfs/xfs_bmap_util.c b/fs/xfs/xfs_bmap_util.c index 22a5dcb70b32..fe1f11b96d0d 100644 --- a/fs/xfs/xfs_bmap_util.c +++ b/fs/xfs/xfs_bmap_util.c @@ -1376,22 +1376,19 @@ out: } /* - * xfs_collapse_file_space() - * This routine frees disk space and shift extent for the given file. - * The first thing we do is to free data blocks in the specified range - * by calling xfs_free_file_space(). It would also sync dirty data - * and invalidate page cache over the region on which collapse range - * is working. And Shift extent records to the left to cover a hole. - * RETURNS: - * 0 on success - * errno on error - * + * @next_fsb will keep track of the extent currently undergoing shift. + * @stop_fsb will keep track of the extent at which we have to stop. + * If we are shifting left, we will start with block (offset + len) and + * shift each extent till last extent. + * If we are shifting right, we will start with last extent inside file space + * and continue until we reach the block corresponding to offset. */ int -xfs_collapse_file_space( - struct xfs_inode *ip, - xfs_off_t offset, - xfs_off_t len) +xfs_shift_file_space( + struct xfs_inode *ip, + xfs_off_t offset, + xfs_off_t len, + enum shift_direction direction) { int done = 0; struct xfs_mount *mp = ip->i_mount; @@ -1400,21 +1397,26 @@ xfs_collapse_file_space( struct xfs_bmap_free free_list; xfs_fsblock_t first_block; int committed; - xfs_fileoff_t start_fsb; + xfs_fileoff_t stop_fsb; xfs_fileoff_t next_fsb; xfs_fileoff_t shift_fsb; - ASSERT(xfs_isilocked(ip, XFS_IOLOCK_EXCL)); + ASSERT(direction == SHIFT_LEFT || direction == SHIFT_RIGHT); - trace_xfs_collapse_file_space(ip); + if (direction == SHIFT_LEFT) { + next_fsb = XFS_B_TO_FSB(mp, offset + len); + stop_fsb = XFS_B_TO_FSB(mp, VFS_I(ip)->i_size); + } else { + /* + * If right shift, delegate the work of initialization of + * next_fsb to xfs_bmap_shift_extent as it has ilock held. + */ + next_fsb = NULLFSBLOCK; + stop_fsb = XFS_B_TO_FSB(mp, offset); + } - next_fsb = XFS_B_TO_FSB(mp, offset + len); shift_fsb = XFS_B_TO_FSB(mp, len); - error = xfs_free_file_space(ip, offset, len); - if (error) - return error; - /* * Trim eofblocks to avoid shifting uninitialized post-eof preallocation * into the accessible region of the file. @@ -1427,20 +1429,28 @@ xfs_collapse_file_space( /* * Writeback and invalidate cache for the remainder of the file as we're - * about to shift down every extent from the collapse range to EOF. The - * free of the collapse range above might have already done some of - * this, but we shouldn't rely on it to do anything outside of the range - * that was freed. + * about to shift down every extent from offset to EOF. */ error = filemap_write_and_wait_range(VFS_I(ip)->i_mapping, - offset + len, -1); + offset, -1); if (error) return error; error = invalidate_inode_pages2_range(VFS_I(ip)->i_mapping, - (offset + len) >> PAGE_CACHE_SHIFT, -1); + offset >> PAGE_CACHE_SHIFT, -1); if (error) return error; + /* + * The extent shiting code works on extent granularity. So, if + * stop_fsb is not the starting block of extent, we need to split + * the extent at stop_fsb. + */ + if (direction == SHIFT_RIGHT) { + error = xfs_bmap_split_extent(ip, stop_fsb); + if (error) + return error; + } + while (!error && !done) { tp = xfs_trans_alloc(mp, XFS_TRANS_DIOSTRAT); /* @@ -1464,7 +1474,7 @@ xfs_collapse_file_space( if (error) goto out; - xfs_trans_ijoin(tp, ip, 0); + xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL); xfs_bmap_init(&free_list, &first_block); @@ -1472,10 +1482,9 @@ xfs_collapse_file_space( * We are using the write transaction in which max 2 bmbt * updates are allowed */ - start_fsb = next_fsb; - error = xfs_bmap_shift_extents(tp, ip, start_fsb, shift_fsb, - &done, &next_fsb, &first_block, &free_list, - XFS_BMAP_MAX_SHIFT_EXTENTS); + error = xfs_bmap_shift_extents(tp, ip, &next_fsb, shift_fsb, + &done, stop_fsb, &first_block, &free_list, + direction, XFS_BMAP_MAX_SHIFT_EXTENTS); if (error) goto out; @@ -1484,17 +1493,69 @@ xfs_collapse_file_space( goto out; error = xfs_trans_commit(tp, XFS_TRANS_RELEASE_LOG_RES); - xfs_iunlock(ip, XFS_ILOCK_EXCL); } return error; out: xfs_trans_cancel(tp, XFS_TRANS_RELEASE_LOG_RES | XFS_TRANS_ABORT); - xfs_iunlock(ip, XFS_ILOCK_EXCL); return error; } +/* + * xfs_collapse_file_space() + * This routine frees disk space and shift extent for the given file. + * The first thing we do is to free data blocks in the specified range + * by calling xfs_free_file_space(). It would also sync dirty data + * and invalidate page cache over the region on which collapse range + * is working. And Shift extent records to the left to cover a hole. + * RETURNS: + * 0 on success + * errno on error + * + */ +int +xfs_collapse_file_space( + struct xfs_inode *ip, + xfs_off_t offset, + xfs_off_t len) +{ + int error; + + ASSERT(xfs_isilocked(ip, XFS_IOLOCK_EXCL)); + trace_xfs_collapse_file_space(ip); + + error = xfs_free_file_space(ip, offset, len); + if (error) + return error; + + return xfs_shift_file_space(ip, offset, len, SHIFT_LEFT); +} + +/* + * xfs_insert_file_space() + * This routine create hole space by shifting extents for the given file. + * The first thing we do is to sync dirty data and invalidate page cache + * over the region on which insert range is working. And split an extent + * to two extents at given offset by calling xfs_bmap_split_extent. + * And shift all extent records which are laying between [offset, + * last allocated extent] to the right to reserve hole range. + * RETURNS: + * 0 on success + * errno on error + */ +int +xfs_insert_file_space( + struct xfs_inode *ip, + loff_t offset, + loff_t len) +{ + ASSERT(xfs_isilocked(ip, XFS_IOLOCK_EXCL)); + trace_xfs_insert_file_space(ip); + + return xfs_shift_file_space(ip, offset, len, SHIFT_RIGHT); +} + /* * We need to check that the format of the data fork in the temporary inode is * valid for the target inode before doing the swap. This is not a problem with diff --git a/fs/xfs/xfs_bmap_util.h b/fs/xfs/xfs_bmap_util.h index 736429a72a12..af97d9a1dfb4 100644 --- a/fs/xfs/xfs_bmap_util.h +++ b/fs/xfs/xfs_bmap_util.h @@ -63,6 +63,8 @@ int xfs_zero_file_space(struct xfs_inode *ip, xfs_off_t offset, xfs_off_t len); int xfs_collapse_file_space(struct xfs_inode *, xfs_off_t offset, xfs_off_t len); +int xfs_insert_file_space(struct xfs_inode *, xfs_off_t offset, + xfs_off_t len); /* EOF block manipulation functions */ bool xfs_can_free_eofblocks(struct xfs_inode *ip, bool force); diff --git a/fs/xfs/xfs_file.c b/fs/xfs/xfs_file.c index ce615d12fb44..edeaccc7961a 100644 --- a/fs/xfs/xfs_file.c +++ b/fs/xfs/xfs_file.c @@ -816,6 +816,11 @@ xfs_file_write_iter( return ret; } +#define XFS_FALLOC_FL_SUPPORTED \ + (FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE | \ + FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE | \ + FALLOC_FL_INSERT_RANGE) + STATIC long xfs_file_fallocate( struct file *file, @@ -829,11 +834,11 @@ xfs_file_fallocate( enum xfs_prealloc_flags flags = 0; uint iolock = XFS_IOLOCK_EXCL; loff_t new_size = 0; + bool do_file_insert = 0; if (!S_ISREG(inode->i_mode)) return -EINVAL; - if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE | - FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE)) + if (mode & ~XFS_FALLOC_FL_SUPPORTED) return -EOPNOTSUPP; xfs_ilock(ip, iolock); @@ -867,6 +872,27 @@ xfs_file_fallocate( error = xfs_collapse_file_space(ip, offset, len); if (error) goto out_unlock; + } else if (mode & FALLOC_FL_INSERT_RANGE) { + unsigned blksize_mask = (1 << inode->i_blkbits) - 1; + + new_size = i_size_read(inode) + len; + if (offset & blksize_mask || len & blksize_mask) { + error = -EINVAL; + goto out_unlock; + } + + /* check the new inode size does not wrap through zero */ + if (new_size > inode->i_sb->s_maxbytes) { + error = -EFBIG; + goto out_unlock; + } + + /* Offset should be less than i_size */ + if (offset >= i_size_read(inode)) { + error = -EINVAL; + goto out_unlock; + } + do_file_insert = 1; } else { flags |= XFS_PREALLOC_SET; @@ -901,8 +927,19 @@ xfs_file_fallocate( iattr.ia_valid = ATTR_SIZE; iattr.ia_size = new_size; error = xfs_setattr_size(ip, &iattr); + if (error) + goto out_unlock; } + /* + * Perform hole insertion now that the file size has been + * updated so that if we crash during the operation we don't + * leave shifted extents past EOF and hence losing access to + * the data that is contained within them. + */ + if (do_file_insert) + error = xfs_insert_file_space(ip, offset, len); + out_unlock: xfs_iunlock(ip, iolock); return error; diff --git a/fs/xfs/xfs_trace.h b/fs/xfs/xfs_trace.h index 51372e34d988..7e45fa155ea8 100644 --- a/fs/xfs/xfs_trace.h +++ b/fs/xfs/xfs_trace.h @@ -664,6 +664,7 @@ DEFINE_INODE_EVENT(xfs_alloc_file_space); DEFINE_INODE_EVENT(xfs_free_file_space); DEFINE_INODE_EVENT(xfs_zero_file_space); DEFINE_INODE_EVENT(xfs_collapse_file_space); +DEFINE_INODE_EVENT(xfs_insert_file_space); DEFINE_INODE_EVENT(xfs_readdir); #ifdef CONFIG_XFS_POSIX_ACL DEFINE_INODE_EVENT(xfs_get_acl); -- cgit v1.2.3-59-g8ed1b From 72c1a73993cfa3572e45e1a878ff7acf31d14fc8 Mon Sep 17 00:00:00 2001 From: kbuild test robot Date: Mon, 13 Apr 2015 11:25:04 +1000 Subject: xfs: xfs_shift_file_space can be static Signed-off-by: Fengguang Wu Reviewed-by: Dave Chinner Signed-off-by: Dave Chinner --- fs/xfs/xfs_bmap_util.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/fs/xfs/xfs_bmap_util.c b/fs/xfs/xfs_bmap_util.c index fe1f11b96d0d..e86757358d5b 100644 --- a/fs/xfs/xfs_bmap_util.c +++ b/fs/xfs/xfs_bmap_util.c @@ -1383,7 +1383,7 @@ out: * If we are shifting right, we will start with last extent inside file space * and continue until we reach the block corresponding to offset. */ -int +static int xfs_shift_file_space( struct xfs_inode *ip, xfs_off_t offset, -- cgit v1.2.3-59-g8ed1b From bbe051c841d522bf2aaa1d362b57fe47457187bf Mon Sep 17 00:00:00 2001 From: Eric Sandeen Date: Mon, 13 Apr 2015 11:25:41 +1000 Subject: xfs: disallow ro->rw remount on norecovery mount There's a bit of a loophole in norecovery mount handling right now: an initial mount must be readonly, but nothing prevents a mount -o remount,rw from producing a writable, unrecovered xfs filesystem. It might be possible to try to perform a log recovery when this is requested, but I'm not sure it's worth the effort. For now, simply disallow this sort of transition. Signed-off-by: Eric Sandeen Reviewed-by: Dave Chinner Signed-off-by: Dave Chinner --- fs/xfs/xfs_super.c | 6 ++++++ 1 file changed, 6 insertions(+) diff --git a/fs/xfs/xfs_super.c b/fs/xfs/xfs_super.c index 58453e3255f8..02718638dc12 100644 --- a/fs/xfs/xfs_super.c +++ b/fs/xfs/xfs_super.c @@ -1237,6 +1237,12 @@ xfs_fs_remount( /* ro -> rw */ if ((mp->m_flags & XFS_MOUNT_RDONLY) && !(*flags & MS_RDONLY)) { + if (mp->m_flags & XFS_MOUNT_NORECOVERY) { + xfs_warn(mp, + "ro->rw transition prohibited on norecovery mount"); + return -EINVAL; + } + mp->m_flags &= ~XFS_MOUNT_RDONLY; /* -- cgit v1.2.3-59-g8ed1b From 2f6612415436bab4d923a61d8df399b134247810 Mon Sep 17 00:00:00 2001 From: Brian Foster Date: Mon, 13 Apr 2015 11:26:02 +1000 Subject: xfs: pass attr geometry to attr leaf header conversion functions The firstused field of the xfs_attr3_leaf_hdr structure is subject to an overflow when fs blocksize is 64k. In preparation to handle this overflow in the header conversion functions, pass the attribute geometry to the functions that convert the in-core structure to and from the on-disk structure. Signed-off-by: Brian Foster Reviewed-by: Dave Chinner Signed-off-by: Dave Chinner --- fs/xfs/libxfs/xfs_attr_leaf.c | 63 +++++++++++++++++++++++-------------------- fs/xfs/libxfs/xfs_attr_leaf.h | 6 +++-- fs/xfs/xfs_attr_inactive.c | 3 ++- fs/xfs/xfs_attr_list.c | 9 ++++--- 4 files changed, 46 insertions(+), 35 deletions(-) diff --git a/fs/xfs/libxfs/xfs_attr_leaf.c b/fs/xfs/libxfs/xfs_attr_leaf.c index 15105dbc9e28..3337516feeb0 100644 --- a/fs/xfs/libxfs/xfs_attr_leaf.c +++ b/fs/xfs/libxfs/xfs_attr_leaf.c @@ -88,6 +88,7 @@ STATIC int xfs_attr_leaf_entsize(xfs_attr_leafblock_t *leaf, int index); void xfs_attr3_leaf_hdr_from_disk( + struct xfs_da_geometry *geo, struct xfs_attr3_icleaf_hdr *to, struct xfs_attr_leafblock *from) { @@ -129,6 +130,7 @@ xfs_attr3_leaf_hdr_from_disk( void xfs_attr3_leaf_hdr_to_disk( + struct xfs_da_geometry *geo, struct xfs_attr_leafblock *to, struct xfs_attr3_icleaf_hdr *from) { @@ -178,7 +180,7 @@ xfs_attr3_leaf_verify( struct xfs_attr_leafblock *leaf = bp->b_addr; struct xfs_attr3_icleaf_hdr ichdr; - xfs_attr3_leaf_hdr_from_disk(&ichdr, leaf); + xfs_attr3_leaf_hdr_from_disk(mp->m_attr_geo, &ichdr, leaf); if (xfs_sb_version_hascrc(&mp->m_sb)) { struct xfs_da3_node_hdr *hdr3 = bp->b_addr; @@ -757,9 +759,10 @@ xfs_attr_shortform_allfit( struct xfs_attr3_icleaf_hdr leafhdr; int bytes; int i; + struct xfs_mount *mp = bp->b_target->bt_mount; leaf = bp->b_addr; - xfs_attr3_leaf_hdr_from_disk(&leafhdr, leaf); + xfs_attr3_leaf_hdr_from_disk(mp->m_attr_geo, &leafhdr, leaf); entry = xfs_attr3_leaf_entryp(leaf); bytes = sizeof(struct xfs_attr_sf_hdr); @@ -812,7 +815,7 @@ xfs_attr3_leaf_to_shortform( memcpy(tmpbuffer, bp->b_addr, args->geo->blksize); leaf = (xfs_attr_leafblock_t *)tmpbuffer; - xfs_attr3_leaf_hdr_from_disk(&ichdr, leaf); + xfs_attr3_leaf_hdr_from_disk(args->geo, &ichdr, leaf); entry = xfs_attr3_leaf_entryp(leaf); /* XXX (dgc): buffer is about to be marked stale - why zero it? */ @@ -923,7 +926,7 @@ xfs_attr3_leaf_to_node( btree = dp->d_ops->node_tree_p(node); leaf = bp2->b_addr; - xfs_attr3_leaf_hdr_from_disk(&icleafhdr, leaf); + xfs_attr3_leaf_hdr_from_disk(args->geo, &icleafhdr, leaf); entries = xfs_attr3_leaf_entryp(leaf); /* both on-disk, don't endian-flip twice */ @@ -988,7 +991,7 @@ xfs_attr3_leaf_create( } ichdr.freemap[0].size = ichdr.firstused - ichdr.freemap[0].base; - xfs_attr3_leaf_hdr_to_disk(leaf, &ichdr); + xfs_attr3_leaf_hdr_to_disk(args->geo, leaf, &ichdr); xfs_trans_log_buf(args->trans, bp, 0, args->geo->blksize - 1); *bpp = bp; @@ -1073,7 +1076,7 @@ xfs_attr3_leaf_add( trace_xfs_attr_leaf_add(args); leaf = bp->b_addr; - xfs_attr3_leaf_hdr_from_disk(&ichdr, leaf); + xfs_attr3_leaf_hdr_from_disk(args->geo, &ichdr, leaf); ASSERT(args->index >= 0 && args->index <= ichdr.count); entsize = xfs_attr_leaf_newentsize(args, NULL); @@ -1126,7 +1129,7 @@ xfs_attr3_leaf_add( tmp = xfs_attr3_leaf_add_work(bp, &ichdr, args, 0); out_log_hdr: - xfs_attr3_leaf_hdr_to_disk(leaf, &ichdr); + xfs_attr3_leaf_hdr_to_disk(args->geo, leaf, &ichdr); xfs_trans_log_buf(args->trans, bp, XFS_DA_LOGRANGE(leaf, &leaf->hdr, xfs_attr3_leaf_hdr_size(leaf))); @@ -1294,7 +1297,7 @@ xfs_attr3_leaf_compact( ichdr_dst->freemap[0].base; /* write the header back to initialise the underlying buffer */ - xfs_attr3_leaf_hdr_to_disk(leaf_dst, ichdr_dst); + xfs_attr3_leaf_hdr_to_disk(args->geo, leaf_dst, ichdr_dst); /* * Copy all entry's in the same (sorted) order, @@ -1344,9 +1347,10 @@ xfs_attr_leaf_order( { struct xfs_attr3_icleaf_hdr ichdr1; struct xfs_attr3_icleaf_hdr ichdr2; + struct xfs_mount *mp = leaf1_bp->b_target->bt_mount; - xfs_attr3_leaf_hdr_from_disk(&ichdr1, leaf1_bp->b_addr); - xfs_attr3_leaf_hdr_from_disk(&ichdr2, leaf2_bp->b_addr); + xfs_attr3_leaf_hdr_from_disk(mp->m_attr_geo, &ichdr1, leaf1_bp->b_addr); + xfs_attr3_leaf_hdr_from_disk(mp->m_attr_geo, &ichdr2, leaf2_bp->b_addr); return xfs_attr3_leaf_order(leaf1_bp, &ichdr1, leaf2_bp, &ichdr2); } @@ -1388,8 +1392,8 @@ xfs_attr3_leaf_rebalance( ASSERT(blk2->magic == XFS_ATTR_LEAF_MAGIC); leaf1 = blk1->bp->b_addr; leaf2 = blk2->bp->b_addr; - xfs_attr3_leaf_hdr_from_disk(&ichdr1, leaf1); - xfs_attr3_leaf_hdr_from_disk(&ichdr2, leaf2); + xfs_attr3_leaf_hdr_from_disk(state->args->geo, &ichdr1, leaf1); + xfs_attr3_leaf_hdr_from_disk(state->args->geo, &ichdr2, leaf2); ASSERT(ichdr2.count == 0); args = state->args; @@ -1490,8 +1494,8 @@ xfs_attr3_leaf_rebalance( ichdr1.count, count); } - xfs_attr3_leaf_hdr_to_disk(leaf1, &ichdr1); - xfs_attr3_leaf_hdr_to_disk(leaf2, &ichdr2); + xfs_attr3_leaf_hdr_to_disk(state->args->geo, leaf1, &ichdr1); + xfs_attr3_leaf_hdr_to_disk(state->args->geo, leaf2, &ichdr2); xfs_trans_log_buf(args->trans, blk1->bp, 0, args->geo->blksize - 1); xfs_trans_log_buf(args->trans, blk2->bp, 0, args->geo->blksize - 1); @@ -1684,7 +1688,7 @@ xfs_attr3_leaf_toosmall( */ blk = &state->path.blk[ state->path.active-1 ]; leaf = blk->bp->b_addr; - xfs_attr3_leaf_hdr_from_disk(&ichdr, leaf); + xfs_attr3_leaf_hdr_from_disk(state->args->geo, &ichdr, leaf); bytes = xfs_attr3_leaf_hdr_size(leaf) + ichdr.count * sizeof(xfs_attr_leaf_entry_t) + ichdr.usedbytes; @@ -1740,7 +1744,7 @@ xfs_attr3_leaf_toosmall( if (error) return error; - xfs_attr3_leaf_hdr_from_disk(&ichdr2, bp->b_addr); + xfs_attr3_leaf_hdr_from_disk(state->args->geo, &ichdr2, bp->b_addr); bytes = state->args->geo->blksize - (state->args->geo->blksize >> 2) - @@ -1805,7 +1809,7 @@ xfs_attr3_leaf_remove( trace_xfs_attr_leaf_remove(args); leaf = bp->b_addr; - xfs_attr3_leaf_hdr_from_disk(&ichdr, leaf); + xfs_attr3_leaf_hdr_from_disk(args->geo, &ichdr, leaf); ASSERT(ichdr.count > 0 && ichdr.count < args->geo->blksize / 8); ASSERT(args->index >= 0 && args->index < ichdr.count); @@ -1923,7 +1927,7 @@ xfs_attr3_leaf_remove( } else { ichdr.holes = 1; /* mark as needing compaction */ } - xfs_attr3_leaf_hdr_to_disk(leaf, &ichdr); + xfs_attr3_leaf_hdr_to_disk(args->geo, leaf, &ichdr); xfs_trans_log_buf(args->trans, bp, XFS_DA_LOGRANGE(leaf, &leaf->hdr, xfs_attr3_leaf_hdr_size(leaf))); @@ -1957,8 +1961,8 @@ xfs_attr3_leaf_unbalance( drop_leaf = drop_blk->bp->b_addr; save_leaf = save_blk->bp->b_addr; - xfs_attr3_leaf_hdr_from_disk(&drophdr, drop_leaf); - xfs_attr3_leaf_hdr_from_disk(&savehdr, save_leaf); + xfs_attr3_leaf_hdr_from_disk(state->args->geo, &drophdr, drop_leaf); + xfs_attr3_leaf_hdr_from_disk(state->args->geo, &savehdr, save_leaf); entry = xfs_attr3_leaf_entryp(drop_leaf); /* @@ -2012,7 +2016,7 @@ xfs_attr3_leaf_unbalance( tmphdr.firstused = state->args->geo->blksize; /* write the header to the temp buffer to initialise it */ - xfs_attr3_leaf_hdr_to_disk(tmp_leaf, &tmphdr); + xfs_attr3_leaf_hdr_to_disk(state->args->geo, tmp_leaf, &tmphdr); if (xfs_attr3_leaf_order(save_blk->bp, &savehdr, drop_blk->bp, &drophdr)) { @@ -2039,7 +2043,7 @@ xfs_attr3_leaf_unbalance( kmem_free(tmp_leaf); } - xfs_attr3_leaf_hdr_to_disk(save_leaf, &savehdr); + xfs_attr3_leaf_hdr_to_disk(state->args->geo, save_leaf, &savehdr); xfs_trans_log_buf(state->args->trans, save_blk->bp, 0, state->args->geo->blksize - 1); @@ -2085,7 +2089,7 @@ xfs_attr3_leaf_lookup_int( trace_xfs_attr_leaf_lookup(args); leaf = bp->b_addr; - xfs_attr3_leaf_hdr_from_disk(&ichdr, leaf); + xfs_attr3_leaf_hdr_from_disk(args->geo, &ichdr, leaf); entries = xfs_attr3_leaf_entryp(leaf); ASSERT(ichdr.count < args->geo->blksize / 8); @@ -2190,7 +2194,7 @@ xfs_attr3_leaf_getvalue( int valuelen; leaf = bp->b_addr; - xfs_attr3_leaf_hdr_from_disk(&ichdr, leaf); + xfs_attr3_leaf_hdr_from_disk(args->geo, &ichdr, leaf); ASSERT(ichdr.count < args->geo->blksize / 8); ASSERT(args->index < ichdr.count); @@ -2391,8 +2395,9 @@ xfs_attr_leaf_lasthash( { struct xfs_attr3_icleaf_hdr ichdr; struct xfs_attr_leaf_entry *entries; + struct xfs_mount *mp = bp->b_target->bt_mount; - xfs_attr3_leaf_hdr_from_disk(&ichdr, bp->b_addr); + xfs_attr3_leaf_hdr_from_disk(mp->m_attr_geo, &ichdr, bp->b_addr); entries = xfs_attr3_leaf_entryp(bp->b_addr); if (count) *count = ichdr.count; @@ -2486,7 +2491,7 @@ xfs_attr3_leaf_clearflag( ASSERT(entry->flags & XFS_ATTR_INCOMPLETE); #ifdef DEBUG - xfs_attr3_leaf_hdr_from_disk(&ichdr, leaf); + xfs_attr3_leaf_hdr_from_disk(args->geo, &ichdr, leaf); ASSERT(args->index < ichdr.count); ASSERT(args->index >= 0); @@ -2550,7 +2555,7 @@ xfs_attr3_leaf_setflag( leaf = bp->b_addr; #ifdef DEBUG - xfs_attr3_leaf_hdr_from_disk(&ichdr, leaf); + xfs_attr3_leaf_hdr_from_disk(args->geo, &ichdr, leaf); ASSERT(args->index < ichdr.count); ASSERT(args->index >= 0); #endif @@ -2629,11 +2634,11 @@ xfs_attr3_leaf_flipflags( entry2 = &xfs_attr3_leaf_entryp(leaf2)[args->index2]; #ifdef DEBUG - xfs_attr3_leaf_hdr_from_disk(&ichdr1, leaf1); + xfs_attr3_leaf_hdr_from_disk(args->geo, &ichdr1, leaf1); ASSERT(args->index < ichdr1.count); ASSERT(args->index >= 0); - xfs_attr3_leaf_hdr_from_disk(&ichdr2, leaf2); + xfs_attr3_leaf_hdr_from_disk(args->geo, &ichdr2, leaf2); ASSERT(args->index2 < ichdr2.count); ASSERT(args->index2 >= 0); diff --git a/fs/xfs/libxfs/xfs_attr_leaf.h b/fs/xfs/libxfs/xfs_attr_leaf.h index e2929da7c3ba..025c4b820c03 100644 --- a/fs/xfs/libxfs/xfs_attr_leaf.h +++ b/fs/xfs/libxfs/xfs_attr_leaf.h @@ -100,9 +100,11 @@ int xfs_attr_leaf_newentsize(struct xfs_da_args *args, int *local); int xfs_attr3_leaf_read(struct xfs_trans *tp, struct xfs_inode *dp, xfs_dablk_t bno, xfs_daddr_t mappedbno, struct xfs_buf **bpp); -void xfs_attr3_leaf_hdr_from_disk(struct xfs_attr3_icleaf_hdr *to, +void xfs_attr3_leaf_hdr_from_disk(struct xfs_da_geometry *geo, + struct xfs_attr3_icleaf_hdr *to, struct xfs_attr_leafblock *from); -void xfs_attr3_leaf_hdr_to_disk(struct xfs_attr_leafblock *to, +void xfs_attr3_leaf_hdr_to_disk(struct xfs_da_geometry *geo, + struct xfs_attr_leafblock *to, struct xfs_attr3_icleaf_hdr *from); #endif /* __XFS_ATTR_LEAF_H__ */ diff --git a/fs/xfs/xfs_attr_inactive.c b/fs/xfs/xfs_attr_inactive.c index 83af4c149635..f9c1c64782d3 100644 --- a/fs/xfs/xfs_attr_inactive.c +++ b/fs/xfs/xfs_attr_inactive.c @@ -132,9 +132,10 @@ xfs_attr3_leaf_inactive( int size; int tmp; int i; + struct xfs_mount *mp = bp->b_target->bt_mount; leaf = bp->b_addr; - xfs_attr3_leaf_hdr_from_disk(&ichdr, leaf); + xfs_attr3_leaf_hdr_from_disk(mp->m_attr_geo, &ichdr, leaf); /* * Count the number of "remote" value extents. diff --git a/fs/xfs/xfs_attr_list.c b/fs/xfs/xfs_attr_list.c index a43d370d2c58..65fb37a18e92 100644 --- a/fs/xfs/xfs_attr_list.c +++ b/fs/xfs/xfs_attr_list.c @@ -225,6 +225,7 @@ xfs_attr_node_list(xfs_attr_list_context_t *context) int error, i; struct xfs_buf *bp; struct xfs_inode *dp = context->dp; + struct xfs_mount *mp = dp->i_mount; trace_xfs_attr_node_list(context); @@ -256,7 +257,8 @@ xfs_attr_node_list(xfs_attr_list_context_t *context) case XFS_ATTR_LEAF_MAGIC: case XFS_ATTR3_LEAF_MAGIC: leaf = bp->b_addr; - xfs_attr3_leaf_hdr_from_disk(&leafhdr, leaf); + xfs_attr3_leaf_hdr_from_disk(mp->m_attr_geo, + &leafhdr, leaf); entries = xfs_attr3_leaf_entryp(leaf); if (cursor->hashval > be32_to_cpu( entries[leafhdr.count - 1].hashval)) { @@ -340,7 +342,7 @@ xfs_attr_node_list(xfs_attr_list_context_t *context) xfs_trans_brelse(NULL, bp); return error; } - xfs_attr3_leaf_hdr_from_disk(&leafhdr, leaf); + xfs_attr3_leaf_hdr_from_disk(mp->m_attr_geo, &leafhdr, leaf); if (context->seen_enough || leafhdr.forw == 0) break; cursor->blkno = leafhdr.forw; @@ -368,11 +370,12 @@ xfs_attr3_leaf_list_int( struct xfs_attr_leaf_entry *entry; int retval; int i; + struct xfs_mount *mp = context->dp->i_mount; trace_xfs_attr_list_leaf(context); leaf = bp->b_addr; - xfs_attr3_leaf_hdr_from_disk(&ichdr, leaf); + xfs_attr3_leaf_hdr_from_disk(mp->m_attr_geo, &ichdr, leaf); entries = xfs_attr3_leaf_entryp(leaf); cursor = context->cursor; -- cgit v1.2.3-59-g8ed1b From e87021a2bc100d330ae859c2cc0614be3e335c5a Mon Sep 17 00:00:00 2001 From: Brian Foster Date: Mon, 13 Apr 2015 11:27:10 +1000 Subject: xfs: use larger in-core attr firstused field and detect overflow The on-disk xfs_attr3_leaf_hdr structure firstused field is 16-bit and subject to overflow when fs block size is 64k. The field is typically initialized to block size when an attr leaf block is initialized. This problem is demonstrated by assert failures when running xfstests generic/117 on an fs with 64k blocks. To support the existing attr leaf block algorithms for insertion, rebalance and entry movement, increase the size of the in-core firstused field to 32-bit and handle the potential overflow on conversion to/from the on-disk structure. If the overflow condition occurs, set a special value in the firstused field that is translated back on header read. The special value is only required in the case of an empty 64k attr block. A value of zero is used because firstused is initialized to the block size and grows backwards from there. Furthermore, the attribute block header occupies the first bytes of the block. Thus, a value of zero has no other legitimate meaning for this structure. Two new conversion helpers are created to manage the conversion of firstused to and from disk. Signed-off-by: Brian Foster Reviewed-by: Dave Chinner Signed-off-by: Dave Chinner --- fs/xfs/libxfs/xfs_attr_leaf.c | 84 ++++++++++++++++++++++++++++++++++++++++--- fs/xfs/libxfs/xfs_da_format.h | 14 +++++++- 2 files changed, 92 insertions(+), 6 deletions(-) diff --git a/fs/xfs/libxfs/xfs_attr_leaf.c b/fs/xfs/libxfs/xfs_attr_leaf.c index 3337516feeb0..149ef3f66735 100644 --- a/fs/xfs/libxfs/xfs_attr_leaf.c +++ b/fs/xfs/libxfs/xfs_attr_leaf.c @@ -86,6 +86,80 @@ STATIC void xfs_attr3_leaf_moveents(struct xfs_da_args *args, int move_count); STATIC int xfs_attr_leaf_entsize(xfs_attr_leafblock_t *leaf, int index); +/* + * attr3 block 'firstused' conversion helpers. + * + * firstused refers to the offset of the first used byte of the nameval region + * of an attr leaf block. The region starts at the tail of the block and expands + * backwards towards the middle. As such, firstused is initialized to the block + * size for an empty leaf block and is reduced from there. + * + * The attr3 block size is pegged to the fsb size and the maximum fsb is 64k. + * The in-core firstused field is 32-bit and thus supports the maximum fsb size. + * The on-disk field is only 16-bit, however, and overflows at 64k. Since this + * only occurs at exactly 64k, we use zero as a magic on-disk value to represent + * the attr block size. The following helpers manage the conversion between the + * in-core and on-disk formats. + */ + +static void +xfs_attr3_leaf_firstused_from_disk( + struct xfs_da_geometry *geo, + struct xfs_attr3_icleaf_hdr *to, + struct xfs_attr_leafblock *from) +{ + struct xfs_attr3_leaf_hdr *hdr3; + + if (from->hdr.info.magic == cpu_to_be16(XFS_ATTR3_LEAF_MAGIC)) { + hdr3 = (struct xfs_attr3_leaf_hdr *) from; + to->firstused = be16_to_cpu(hdr3->firstused); + } else { + to->firstused = be16_to_cpu(from->hdr.firstused); + } + + /* + * Convert from the magic fsb size value to actual blocksize. This + * should only occur for empty blocks when the block size overflows + * 16-bits. + */ + if (to->firstused == XFS_ATTR3_LEAF_NULLOFF) { + ASSERT(!to->count && !to->usedbytes); + ASSERT(geo->blksize > USHRT_MAX); + to->firstused = geo->blksize; + } +} + +static void +xfs_attr3_leaf_firstused_to_disk( + struct xfs_da_geometry *geo, + struct xfs_attr_leafblock *to, + struct xfs_attr3_icleaf_hdr *from) +{ + struct xfs_attr3_leaf_hdr *hdr3; + uint32_t firstused; + + /* magic value should only be seen on disk */ + ASSERT(from->firstused != XFS_ATTR3_LEAF_NULLOFF); + + /* + * Scale down the 32-bit in-core firstused value to the 16-bit on-disk + * value. This only overflows at the max supported value of 64k. Use the + * magic on-disk value to represent block size in this case. + */ + firstused = from->firstused; + if (firstused > USHRT_MAX) { + ASSERT(from->firstused == geo->blksize); + firstused = XFS_ATTR3_LEAF_NULLOFF; + } + + if (from->magic == XFS_ATTR3_LEAF_MAGIC) { + hdr3 = (struct xfs_attr3_leaf_hdr *) to; + hdr3->firstused = cpu_to_be16(firstused); + } else { + to->hdr.firstused = cpu_to_be16(firstused); + } +} + void xfs_attr3_leaf_hdr_from_disk( struct xfs_da_geometry *geo, @@ -105,7 +179,7 @@ xfs_attr3_leaf_hdr_from_disk( to->magic = be16_to_cpu(hdr3->info.hdr.magic); to->count = be16_to_cpu(hdr3->count); to->usedbytes = be16_to_cpu(hdr3->usedbytes); - to->firstused = be16_to_cpu(hdr3->firstused); + xfs_attr3_leaf_firstused_from_disk(geo, to, from); to->holes = hdr3->holes; for (i = 0; i < XFS_ATTR_LEAF_MAPSIZE; i++) { @@ -119,7 +193,7 @@ xfs_attr3_leaf_hdr_from_disk( to->magic = be16_to_cpu(from->hdr.info.magic); to->count = be16_to_cpu(from->hdr.count); to->usedbytes = be16_to_cpu(from->hdr.usedbytes); - to->firstused = be16_to_cpu(from->hdr.firstused); + xfs_attr3_leaf_firstused_from_disk(geo, to, from); to->holes = from->hdr.holes; for (i = 0; i < XFS_ATTR_LEAF_MAPSIZE; i++) { @@ -134,7 +208,7 @@ xfs_attr3_leaf_hdr_to_disk( struct xfs_attr_leafblock *to, struct xfs_attr3_icleaf_hdr *from) { - int i; + int i; ASSERT(from->magic == XFS_ATTR_LEAF_MAGIC || from->magic == XFS_ATTR3_LEAF_MAGIC); @@ -147,7 +221,7 @@ xfs_attr3_leaf_hdr_to_disk( hdr3->info.hdr.magic = cpu_to_be16(from->magic); hdr3->count = cpu_to_be16(from->count); hdr3->usedbytes = cpu_to_be16(from->usedbytes); - hdr3->firstused = cpu_to_be16(from->firstused); + xfs_attr3_leaf_firstused_to_disk(geo, to, from); hdr3->holes = from->holes; hdr3->pad1 = 0; @@ -162,7 +236,7 @@ xfs_attr3_leaf_hdr_to_disk( to->hdr.info.magic = cpu_to_be16(from->magic); to->hdr.count = cpu_to_be16(from->count); to->hdr.usedbytes = cpu_to_be16(from->usedbytes); - to->hdr.firstused = cpu_to_be16(from->firstused); + xfs_attr3_leaf_firstused_to_disk(geo, to, from); to->hdr.holes = from->holes; to->hdr.pad1 = 0; diff --git a/fs/xfs/libxfs/xfs_da_format.h b/fs/xfs/libxfs/xfs_da_format.h index 0a49b0286372..74bcbabfa523 100644 --- a/fs/xfs/libxfs/xfs_da_format.h +++ b/fs/xfs/libxfs/xfs_da_format.h @@ -725,7 +725,13 @@ struct xfs_attr3_icleaf_hdr { __uint16_t magic; __uint16_t count; __uint16_t usedbytes; - __uint16_t firstused; + /* + * firstused is 32-bit here instead of 16-bit like the on-disk variant + * to support maximum fsb size of 64k without overflow issues throughout + * the attr code. Instead, the overflow condition is handled on + * conversion to/from disk. + */ + __uint32_t firstused; __u8 holes; struct { __uint16_t base; @@ -733,6 +739,12 @@ struct xfs_attr3_icleaf_hdr { } freemap[XFS_ATTR_LEAF_MAPSIZE]; }; +/* + * Special value to represent fs block size in the leaf header firstused field. + * Only used when block size overflows the 2-bytes available on disk. + */ +#define XFS_ATTR3_LEAF_NULLOFF 0 + /* * Flags used in the leaf_entry[i].flags field. * NOTE: the INCOMPLETE bit must not collide with the flags bits specified -- cgit v1.2.3-59-g8ed1b From 66db8104968ad8c0bf5a45a100ae586ddfadc1e1 Mon Sep 17 00:00:00 2001 From: Brian Foster Date: Mon, 13 Apr 2015 11:27:59 +1000 Subject: xfs: kill unnecessary firstused overflow check on attr3 leaf removal xfs_attr3_leaf_remove() removes an attribute from an attr leaf block. If the attribute nameval data happens to be at the start of the nameval region, a new start offset (firstused) for the region is calculated (since the region grows from the tail of the block to the start). Once the new firstused is calculated, it is checked for zero in an apparent overflow check. Now that the in-core firstused is 32-bit, overflow is not possible and this check can be removed. Since the purpose for this check is not documented and appears to exist since the port to Linux, be conservative and replace it with an assert. Signed-off-by: Brian Foster Reviewed-by: Dave Chinner Signed-off-by: Dave Chinner --- fs/xfs/libxfs/xfs_attr_leaf.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) diff --git a/fs/xfs/libxfs/xfs_attr_leaf.c b/fs/xfs/libxfs/xfs_attr_leaf.c index 149ef3f66735..04e79d57bca6 100644 --- a/fs/xfs/libxfs/xfs_attr_leaf.c +++ b/fs/xfs/libxfs/xfs_attr_leaf.c @@ -1996,8 +1996,7 @@ xfs_attr3_leaf_remove( tmp = be16_to_cpu(entry->nameidx); } ichdr.firstused = tmp; - if (!ichdr.firstused) - ichdr.firstused = tmp - XFS_ATTR_LEAF_NAME_ALIGN; + ASSERT(ichdr.firstused != 0); } else { ichdr.holes = 1; /* mark as needing compaction */ } -- cgit v1.2.3-59-g8ed1b From 21c3ea18819b5f650c75f59a0457415bc05d2b17 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Mon, 13 Apr 2015 11:38:29 +1000 Subject: xfs: unlock i_mutex in xfs_break_layouts We want to drop all I/O path locks when recalling layouts, and that includes i_mutex for the write path. Without this we get stuck processe when recalls take too long. [dchinner: fix build with !CONFIG_PNFS] Signed-off-by: Christoph Hellwig Reviewed-by: Dave Chinner Signed-off-by: Dave Chinner --- fs/xfs/xfs_file.c | 4 ++-- fs/xfs/xfs_ioctl.c | 2 +- fs/xfs/xfs_iops.c | 2 +- fs/xfs/xfs_pnfs.c | 7 ++++++- fs/xfs/xfs_pnfs.h | 5 +++-- 5 files changed, 13 insertions(+), 7 deletions(-) diff --git a/fs/xfs/xfs_file.c b/fs/xfs/xfs_file.c index edeaccc7961a..f63aeddd31d5 100644 --- a/fs/xfs/xfs_file.c +++ b/fs/xfs/xfs_file.c @@ -555,7 +555,7 @@ restart: if (error) return error; - error = xfs_break_layouts(inode, iolock); + error = xfs_break_layouts(inode, iolock, true); if (error) return error; @@ -842,7 +842,7 @@ xfs_file_fallocate( return -EOPNOTSUPP; xfs_ilock(ip, iolock); - error = xfs_break_layouts(inode, &iolock); + error = xfs_break_layouts(inode, &iolock, false); if (error) goto out_unlock; diff --git a/fs/xfs/xfs_ioctl.c b/fs/xfs/xfs_ioctl.c index ac4feae45eb3..3a21cc71fda0 100644 --- a/fs/xfs/xfs_ioctl.c +++ b/fs/xfs/xfs_ioctl.c @@ -639,7 +639,7 @@ xfs_ioc_space( return error; xfs_ilock(ip, iolock); - error = xfs_break_layouts(inode, &iolock); + error = xfs_break_layouts(inode, &iolock, false); if (error) goto out_unlock; diff --git a/fs/xfs/xfs_iops.c b/fs/xfs/xfs_iops.c index d7782ae1af3c..1d4efee4be17 100644 --- a/fs/xfs/xfs_iops.c +++ b/fs/xfs/xfs_iops.c @@ -988,7 +988,7 @@ xfs_vn_setattr( uint iolock = XFS_IOLOCK_EXCL; xfs_ilock(ip, iolock); - error = xfs_break_layouts(dentry->d_inode, &iolock); + error = xfs_break_layouts(dentry->d_inode, &iolock, true); if (!error) error = xfs_setattr_size(ip, iattr); xfs_iunlock(ip, iolock); diff --git a/fs/xfs/xfs_pnfs.c b/fs/xfs/xfs_pnfs.c index 4b33ef112400..cbb424f4d93a 100644 --- a/fs/xfs/xfs_pnfs.c +++ b/fs/xfs/xfs_pnfs.c @@ -31,7 +31,8 @@ int xfs_break_layouts( struct inode *inode, - uint *iolock) + uint *iolock, + bool with_imutex) { struct xfs_inode *ip = XFS_I(inode); int error; @@ -40,8 +41,12 @@ xfs_break_layouts( while ((error = break_layout(inode, false) == -EWOULDBLOCK)) { xfs_iunlock(ip, *iolock); + if (with_imutex && (*iolock & XFS_IOLOCK_EXCL)) + mutex_unlock(&inode->i_mutex); error = break_layout(inode, true); *iolock = XFS_IOLOCK_EXCL; + if (with_imutex) + mutex_lock(&inode->i_mutex); xfs_ilock(ip, *iolock); } diff --git a/fs/xfs/xfs_pnfs.h b/fs/xfs/xfs_pnfs.h index b7fbfce660f6..8147ac108820 100644 --- a/fs/xfs/xfs_pnfs.h +++ b/fs/xfs/xfs_pnfs.h @@ -8,9 +8,10 @@ int xfs_fs_map_blocks(struct inode *inode, loff_t offset, u64 length, int xfs_fs_commit_blocks(struct inode *inode, struct iomap *maps, int nr_maps, struct iattr *iattr); -int xfs_break_layouts(struct inode *inode, uint *iolock); +int xfs_break_layouts(struct inode *inode, uint *iolock, bool with_imutex); #else -static inline int xfs_break_layouts(struct inode *inode, uint *iolock) +static inline int +xfs_break_layouts(struct inode *inode, uint *iolock, bool with_imutex) { return 0; } -- cgit v1.2.3-59-g8ed1b From a719370be52d1152a5f6e19c4af3b73280e25475 Mon Sep 17 00:00:00 2001 From: Dave Chinner Date: Thu, 16 Apr 2015 21:57:48 +1000 Subject: xfs: factor DIO write mapping from get_blocks Clarify and separate the buffer mapping logic so that the direct IO mapping is not tangled up in propagating the extent status to teh mapping buffer. This makes it easier to extend the direct IO mapping to use an ioend in future. Signed-off-by: Dave Chinner Reviewed-by: Brian Foster Signed-off-by: Dave Chinner --- fs/xfs/xfs_aops.c | 40 +++++++++++++++++++++++++++------------- 1 file changed, 27 insertions(+), 13 deletions(-) diff --git a/fs/xfs/xfs_aops.c b/fs/xfs/xfs_aops.c index 3a9b7a1b8704..489ed200bbbb 100644 --- a/fs/xfs/xfs_aops.c +++ b/fs/xfs/xfs_aops.c @@ -1233,6 +1233,22 @@ xfs_vm_releasepage( return try_to_free_buffers(page); } +/* + * do all the direct IO specific mapping buffer manipulation here. + */ +static void +xfs_map_direct( + struct inode *inode, + struct buffer_head *bh_result, + struct xfs_bmbt_irec *imap, + xfs_off_t offset) +{ + if (ISUNWRITTEN(imap)) { + bh_result->b_private = inode; + set_buffer_defer_completion(bh_result); + } +} + STATIC int __xfs_get_blocks( struct inode *inode, @@ -1331,21 +1347,19 @@ __xfs_get_blocks( goto out_unlock; } + /* + * For unwritten extents do not report a disk address in the buffered + * read case (treat as if we're reading into a hole). + */ if (imap.br_startblock != HOLESTARTBLOCK && - imap.br_startblock != DELAYSTARTBLOCK) { - /* - * For unwritten extents do not report a disk address on - * the read case (treat as if we're reading into a hole). - */ - if (create || !ISUNWRITTEN(&imap)) - xfs_map_buffer(inode, bh_result, &imap, offset); - if (create && ISUNWRITTEN(&imap)) { - if (direct) { - bh_result->b_private = inode; - set_buffer_defer_completion(bh_result); - } + imap.br_startblock != DELAYSTARTBLOCK && + (create || !ISUNWRITTEN(&imap))) { + xfs_map_buffer(inode, bh_result, &imap, offset); + if (ISUNWRITTEN(&imap)) set_buffer_unwritten(bh_result); - } + /* direct IO needs special help */ + if (create && direct) + xfs_map_direct(inode, bh_result, &imap, offset); } /* -- cgit v1.2.3-59-g8ed1b From 1fdca9c21198b2c2869086ac3629612492476f28 Mon Sep 17 00:00:00 2001 From: Dave Chinner Date: Thu, 16 Apr 2015 21:58:21 +1000 Subject: xfs: move DIO mapping size calculation The mapping size calculation is done last in __xfs_get_blocks(), but we are going to need the actual mapping size we will use to map the direct IO correctly in xfs_map_direct(). Factor out the calculation for code clarity, and move the call to be the first operation in mapping the extent to the returned buffer. Signed-off-by: Dave Chinner Reviewed-by: Brian Foster Signed-off-by: Dave Chinner --- fs/xfs/xfs_aops.c | 79 ++++++++++++++++++++++++++++++++----------------------- 1 file changed, 46 insertions(+), 33 deletions(-) diff --git a/fs/xfs/xfs_aops.c b/fs/xfs/xfs_aops.c index 489ed200bbbb..4a29399ed549 100644 --- a/fs/xfs/xfs_aops.c +++ b/fs/xfs/xfs_aops.c @@ -1249,6 +1249,47 @@ xfs_map_direct( } } + +/* + * If this is O_DIRECT or the mpage code calling tell them how large the mapping + * is, so that we can avoid repeated get_blocks calls. + * + * If the mapping spans EOF, then we have to break the mapping up as the mapping + * for blocks beyond EOF must be marked new so that sub block regions can be + * correctly zeroed. We can't do this for mappings within EOF unless the mapping + * was just allocated or is unwritten, otherwise the callers would overwrite + * existing data with zeros. Hence we have to split the mapping into a range up + * to and including EOF, and a second mapping for beyond EOF. + */ +static void +xfs_map_trim_size( + struct inode *inode, + sector_t iblock, + struct buffer_head *bh_result, + struct xfs_bmbt_irec *imap, + xfs_off_t offset, + ssize_t size) +{ + xfs_off_t mapping_size; + + mapping_size = imap->br_startoff + imap->br_blockcount - iblock; + mapping_size <<= inode->i_blkbits; + + ASSERT(mapping_size > 0); + if (mapping_size > size) + mapping_size = size; + if (offset < i_size_read(inode) && + offset + mapping_size >= i_size_read(inode)) { + /* limit mapping to block that spans EOF */ + mapping_size = roundup_64(i_size_read(inode) - offset, + 1 << inode->i_blkbits); + } + if (mapping_size > LONG_MAX) + mapping_size = LONG_MAX; + + bh_result->b_size = mapping_size; +} + STATIC int __xfs_get_blocks( struct inode *inode, @@ -1347,6 +1388,11 @@ __xfs_get_blocks( goto out_unlock; } + /* trim mapping down to size requested */ + if (direct || size > (1 << inode->i_blkbits)) + xfs_map_trim_size(inode, iblock, bh_result, + &imap, offset, size); + /* * For unwritten extents do not report a disk address in the buffered * read case (treat as if we're reading into a hole). @@ -1392,39 +1438,6 @@ __xfs_get_blocks( } } - /* - * If this is O_DIRECT or the mpage code calling tell them how large - * the mapping is, so that we can avoid repeated get_blocks calls. - * - * If the mapping spans EOF, then we have to break the mapping up as the - * mapping for blocks beyond EOF must be marked new so that sub block - * regions can be correctly zeroed. We can't do this for mappings within - * EOF unless the mapping was just allocated or is unwritten, otherwise - * the callers would overwrite existing data with zeros. Hence we have - * to split the mapping into a range up to and including EOF, and a - * second mapping for beyond EOF. - */ - if (direct || size > (1 << inode->i_blkbits)) { - xfs_off_t mapping_size; - - mapping_size = imap.br_startoff + imap.br_blockcount - iblock; - mapping_size <<= inode->i_blkbits; - - ASSERT(mapping_size > 0); - if (mapping_size > size) - mapping_size = size; - if (offset < i_size_read(inode) && - offset + mapping_size >= i_size_read(inode)) { - /* limit mapping to block that spans EOF */ - mapping_size = roundup_64(i_size_read(inode) - offset, - 1 << inode->i_blkbits); - } - if (mapping_size > LONG_MAX) - mapping_size = LONG_MAX; - - bh_result->b_size = mapping_size; - } - return 0; out_unlock: -- cgit v1.2.3-59-g8ed1b From d5cc2e3f968ff60f247fdef15b04fac788ef46d2 Mon Sep 17 00:00:00 2001 From: Dave Chinner Date: Thu, 16 Apr 2015 21:59:07 +1000 Subject: xfs: DIO needs an ioend for writes Currently we can only tell DIO completion that an IO requires unwritten extent completion. This is done by a hacky non-null private pointer passed to Io completion, but the private pointer does not actually contain any information that is used. We also need to pass to IO completion the fact that the IO may be beyond EOF and so a size update transaction needs to be done. This is currently determined by checks in the io completion, but we need to determine if this is necessary at block mapping time as we need to defer the size update transactions to a completion workqueue, just like unwritten extent conversion. To do this, first we need to allocate and pass an ioend to to IO completion. Add this for unwritten extent conversion; we'll do the EOF updates in the next commit. Signed-off-by: Dave Chinner Reviewed-by: Brian Foster Signed-off-by: Dave Chinner --- fs/xfs/xfs_aops.c | 92 ++++++++++++++++++++++++++++++++++++++++++++++++------ fs/xfs/xfs_trace.h | 3 ++ 2 files changed, 85 insertions(+), 10 deletions(-) diff --git a/fs/xfs/xfs_aops.c b/fs/xfs/xfs_aops.c index 4a29399ed549..60d6466d72f6 100644 --- a/fs/xfs/xfs_aops.c +++ b/fs/xfs/xfs_aops.c @@ -1234,7 +1234,23 @@ xfs_vm_releasepage( } /* - * do all the direct IO specific mapping buffer manipulation here. + * When we map a DIO buffer, we need to attach an ioend that describes the type + * of write IO we are doing. This passes to the completion function the + * operations it needs to perform. + * + * If we get multiple mappings in a single IO, we might be mapping different + * types. But because the direct IO can only have a single private pointer, we + * need to ensure that: + * + * a) the ioend spans the entire region of the IO; and + * b) if it contains unwritten extents, it is *permanently* marked as such + * + * We could do this by chaining ioends like buffered IO does, but we only + * actually get one IO completion callback from the direct IO, and that spans + * the entire IO regardless of how many mappings and IOs are needed to complete + * the DIO. There is only going to be one reference to the ioend and its life + * cycle is constrained by the DIO completion code. hence we don't need + * reference counting here. */ static void xfs_map_direct( @@ -1243,10 +1259,42 @@ xfs_map_direct( struct xfs_bmbt_irec *imap, xfs_off_t offset) { - if (ISUNWRITTEN(imap)) { - bh_result->b_private = inode; - set_buffer_defer_completion(bh_result); + struct xfs_ioend *ioend; + xfs_off_t size = bh_result->b_size; + int type; + + if (ISUNWRITTEN(imap)) + type = XFS_IO_UNWRITTEN; + else + type = XFS_IO_OVERWRITE; + + trace_xfs_gbmap_direct(XFS_I(inode), offset, size, type, imap); + + if (bh_result->b_private) { + ioend = bh_result->b_private; + ASSERT(ioend->io_size > 0); + ASSERT(offset >= ioend->io_offset); + if (offset + size > ioend->io_offset + ioend->io_size) + ioend->io_size = offset - ioend->io_offset + size; + + if (type == XFS_IO_UNWRITTEN && type != ioend->io_type) + ioend->io_type = XFS_IO_UNWRITTEN; + + trace_xfs_gbmap_direct_update(XFS_I(inode), ioend->io_offset, + ioend->io_size, ioend->io_type, + imap); + } else { + ioend = xfs_alloc_ioend(inode, type); + ioend->io_offset = offset; + ioend->io_size = size; + bh_result->b_private = ioend; + + trace_xfs_gbmap_direct_new(XFS_I(inode), offset, size, type, + imap); } + + if (ioend->io_type == XFS_IO_UNWRITTEN) + set_buffer_defer_completion(bh_result); } @@ -1378,10 +1426,13 @@ __xfs_get_blocks( xfs_iunlock(ip, lockmode); } - - trace_xfs_get_blocks_alloc(ip, offset, size, 0, &imap); + trace_xfs_get_blocks_alloc(ip, offset, size, + ISUNWRITTEN(&imap) ? XFS_IO_UNWRITTEN + : XFS_IO_DELALLOC, &imap); } else if (nimaps) { - trace_xfs_get_blocks_found(ip, offset, size, 0, &imap); + trace_xfs_get_blocks_found(ip, offset, size, + ISUNWRITTEN(&imap) ? XFS_IO_UNWRITTEN + : XFS_IO_OVERWRITE, &imap); xfs_iunlock(ip, lockmode); } else { trace_xfs_get_blocks_notfound(ip, offset, size); @@ -1482,9 +1533,28 @@ xfs_end_io_direct_write( struct inode *inode = file_inode(iocb->ki_filp); struct xfs_inode *ip = XFS_I(inode); struct xfs_mount *mp = ip->i_mount; + struct xfs_ioend *ioend = private; if (XFS_FORCED_SHUTDOWN(mp)) - return; + goto out_destroy_ioend; + + /* + * dio completion end_io functions are only called on writes if more + * than 0 bytes was written. + */ + ASSERT(size > 0); + + /* + * The ioend only maps whole blocks, while the IO may be sector aligned. + * Hence the ioend offset/size may not match the IO offset/size exactly, + * but should span it completely. Write the IO sizes into the ioend so + * that completion processing does the right thing. + */ + ASSERT(size <= ioend->io_size); + ASSERT(offset >= ioend->io_offset); + ASSERT(offset + size <= ioend->io_offset + ioend->io_size); + ioend->io_size = size; + ioend->io_offset = offset; /* * While the generic direct I/O code updates the inode size, it does @@ -1504,7 +1574,7 @@ xfs_end_io_direct_write( * we can pass the ioend to the direct IO allocation callbacks and * avoid nesting that way. */ - if (private && size > 0) { + if (ioend->io_type == XFS_IO_UNWRITTEN) { xfs_iomap_write_unwritten(ip, offset, size); } else if (offset + size > ip->i_d.di_size) { struct xfs_trans *tp; @@ -1514,11 +1584,13 @@ xfs_end_io_direct_write( error = xfs_trans_reserve(tp, &M_RES(mp)->tr_fsyncts, 0, 0); if (error) { xfs_trans_cancel(tp, 0); - return; + goto out_destroy_ioend; } xfs_setfilesize(ip, tp, offset, size); } +out_destroy_ioend: + xfs_destroy_ioend(ioend); } STATIC ssize_t diff --git a/fs/xfs/xfs_trace.h b/fs/xfs/xfs_trace.h index 51372e34d988..2de8556ffac2 100644 --- a/fs/xfs/xfs_trace.h +++ b/fs/xfs/xfs_trace.h @@ -1217,6 +1217,9 @@ DEFINE_IOMAP_EVENT(xfs_map_blocks_found); DEFINE_IOMAP_EVENT(xfs_map_blocks_alloc); DEFINE_IOMAP_EVENT(xfs_get_blocks_found); DEFINE_IOMAP_EVENT(xfs_get_blocks_alloc); +DEFINE_IOMAP_EVENT(xfs_gbmap_direct); +DEFINE_IOMAP_EVENT(xfs_gbmap_direct_new); +DEFINE_IOMAP_EVENT(xfs_gbmap_direct_update); DECLARE_EVENT_CLASS(xfs_simple_io_class, TP_PROTO(struct xfs_inode *ip, xfs_off_t offset, ssize_t count), -- cgit v1.2.3-59-g8ed1b From 6dfa1b67e3b3a9bf536e2fb9ed99001c219822a5 Mon Sep 17 00:00:00 2001 From: Dave Chinner Date: Thu, 16 Apr 2015 21:59:34 +1000 Subject: xfs: handle DIO overwrite EOF update completion correctly Currently a DIO overwrite that extends the EOF (e.g sub-block IO or write into allocated blocks beyond EOF) requires a transaction for the EOF update. Thi is done in IO completion context, but we aren't explicitly handling this situation properly and so it can run in interrupt context. Ensure that we defer IO that spans EOF correctly to the DIO completion workqueue, and now that we have an ioend in IO completion we can use the common ioend completion path to do all the work. Note: we do not preallocate the append transaction as we can have multiple mapping and allocation calls per direct IO. hence preallocating can still leave us with nested transactions by attempting to map and allocate more blocks after we've preallocated an append transaction. Signed-off-by: Dave Chinner Reviewed-by: Brian Foster Signed-off-by: Dave Chinner --- fs/xfs/xfs_aops.c | 61 +++++++++++++++++++++++++++--------------------------- fs/xfs/xfs_trace.h | 1 + 2 files changed, 31 insertions(+), 31 deletions(-) diff --git a/fs/xfs/xfs_aops.c b/fs/xfs/xfs_aops.c index 60d6466d72f6..a59443db1de9 100644 --- a/fs/xfs/xfs_aops.c +++ b/fs/xfs/xfs_aops.c @@ -1293,7 +1293,7 @@ xfs_map_direct( imap); } - if (ioend->io_type == XFS_IO_UNWRITTEN) + if (ioend->io_type == XFS_IO_UNWRITTEN || xfs_ioend_is_append(ioend)) set_buffer_defer_completion(bh_result); } @@ -1535,8 +1535,10 @@ xfs_end_io_direct_write( struct xfs_mount *mp = ip->i_mount; struct xfs_ioend *ioend = private; + trace_xfs_gbmap_direct_endio(ip, offset, size, ioend->io_type, NULL); + if (XFS_FORCED_SHUTDOWN(mp)) - goto out_destroy_ioend; + goto out_end_io; /* * dio completion end_io functions are only called on writes if more @@ -1557,40 +1559,37 @@ xfs_end_io_direct_write( ioend->io_offset = offset; /* - * While the generic direct I/O code updates the inode size, it does - * so only after the end_io handler is called, which means our - * end_io handler thinks the on-disk size is outside the in-core - * size. To prevent this just update it a little bit earlier here. + * The ioend tells us whether we are doing unwritten extent conversion + * or an append transaction that updates the on-disk file size. These + * cases are the only cases where we should *potentially* be needing + * to update the VFS inode size. When the ioend indicates this, we + * are *guaranteed* to be running in non-interrupt context. + * + * We need to update the in-core inode size here so that we don't end up + * with the on-disk inode size being outside the in-core inode size. + * While we can do this in the process context after the IO has + * completed, this does not work for AIO and hence we always update + * the in-core inode size here if necessary. */ - if (offset + size > i_size_read(inode)) - i_size_write(inode, offset + size); + if (ioend->io_type == XFS_IO_UNWRITTEN || xfs_ioend_is_append(ioend)) { + if (offset + size > i_size_read(inode)) + i_size_write(inode, offset + size); + } else + ASSERT(offset + size <= i_size_read(inode)); /* - * For direct I/O we do not know if we need to allocate blocks or not, - * so we can't preallocate an append transaction, as that results in - * nested reservations and log space deadlocks. Hence allocate the - * transaction here. While this is sub-optimal and can block IO - * completion for some time, we're stuck with doing it this way until - * we can pass the ioend to the direct IO allocation callbacks and - * avoid nesting that way. + * If we are doing an append IO that needs to update the EOF on disk, + * do the transaction reserve now so we can use common end io + * processing. Stashing the error (if there is one) in the ioend will + * result in the ioend processing passing on the error if it is + * possible as we can't return it from here. */ - if (ioend->io_type == XFS_IO_UNWRITTEN) { - xfs_iomap_write_unwritten(ip, offset, size); - } else if (offset + size > ip->i_d.di_size) { - struct xfs_trans *tp; - int error; - - tp = xfs_trans_alloc(mp, XFS_TRANS_FSYNC_TS); - error = xfs_trans_reserve(tp, &M_RES(mp)->tr_fsyncts, 0, 0); - if (error) { - xfs_trans_cancel(tp, 0); - goto out_destroy_ioend; - } + if (ioend->io_type == XFS_IO_OVERWRITE && xfs_ioend_is_append(ioend)) + ioend->io_error = xfs_setfilesize_trans_alloc(ioend); - xfs_setfilesize(ip, tp, offset, size); - } -out_destroy_ioend: - xfs_destroy_ioend(ioend); +out_end_io: + xfs_end_io(&ioend->io_work); + return; } STATIC ssize_t diff --git a/fs/xfs/xfs_trace.h b/fs/xfs/xfs_trace.h index 2de8556ffac2..0ae50e9847bb 100644 --- a/fs/xfs/xfs_trace.h +++ b/fs/xfs/xfs_trace.h @@ -1220,6 +1220,7 @@ DEFINE_IOMAP_EVENT(xfs_get_blocks_alloc); DEFINE_IOMAP_EVENT(xfs_gbmap_direct); DEFINE_IOMAP_EVENT(xfs_gbmap_direct_new); DEFINE_IOMAP_EVENT(xfs_gbmap_direct_update); +DEFINE_IOMAP_EVENT(xfs_gbmap_direct_endio); DECLARE_EVENT_CLASS(xfs_simple_io_class, TP_PROTO(struct xfs_inode *ip, xfs_off_t offset, ssize_t count), -- cgit v1.2.3-59-g8ed1b From a06c277a13c3620c8ee9304891758f2fcff9c4a4 Mon Sep 17 00:00:00 2001 From: Dave Chinner Date: Thu, 16 Apr 2015 22:00:00 +1000 Subject: xfs: DIO writes within EOF don't need an ioend DIO writes that lie entirely within EOF have nothing to do in IO completion. In this case, we don't need no steekin' ioend, and so we can avoid allocating an ioend until we have a mapping that spans EOF. This means that IO completion has two contexts - deferred completion to the dio workqueue that uses an ioend, and interrupt completion that does nothing because there is nothing that can be done in this context. Signed-off-by: Dave Chinner Reviewed-by: Brian Foster Signed-off-by: Dave Chinner --- fs/xfs/xfs_aops.c | 69 ++++++++++++++++++++++++++++++------------------------ fs/xfs/xfs_trace.h | 1 + 2 files changed, 40 insertions(+), 30 deletions(-) diff --git a/fs/xfs/xfs_aops.c b/fs/xfs/xfs_aops.c index a59443db1de9..c02a47453137 100644 --- a/fs/xfs/xfs_aops.c +++ b/fs/xfs/xfs_aops.c @@ -1234,15 +1234,19 @@ xfs_vm_releasepage( } /* - * When we map a DIO buffer, we need to attach an ioend that describes the type - * of write IO we are doing. This passes to the completion function the - * operations it needs to perform. + * When we map a DIO buffer, we may need to attach an ioend that describes the + * type of write IO we are doing. This passes to the completion function the + * operations it needs to perform. If the mapping is for an overwrite wholly + * within the EOF then we don't need an ioend and so we don't allocate one. + * This avoids the unnecessary overhead of allocating and freeing ioends for + * workloads that don't require transactions on IO completion. * * If we get multiple mappings in a single IO, we might be mapping different * types. But because the direct IO can only have a single private pointer, we * need to ensure that: * - * a) the ioend spans the entire region of the IO; and + * a) i) the ioend spans the entire region of unwritten mappings; or + * ii) the ioend spans all the mappings that cross or are beyond EOF; and * b) if it contains unwritten extents, it is *permanently* marked as such * * We could do this by chaining ioends like buffered IO does, but we only @@ -1283,21 +1287,23 @@ xfs_map_direct( trace_xfs_gbmap_direct_update(XFS_I(inode), ioend->io_offset, ioend->io_size, ioend->io_type, imap); - } else { + } else if (type == XFS_IO_UNWRITTEN || + offset + size > i_size_read(inode)) { ioend = xfs_alloc_ioend(inode, type); ioend->io_offset = offset; ioend->io_size = size; + bh_result->b_private = ioend; + set_buffer_defer_completion(bh_result); trace_xfs_gbmap_direct_new(XFS_I(inode), offset, size, type, imap); + } else { + trace_xfs_gbmap_direct_none(XFS_I(inode), offset, size, type, + imap); } - - if (ioend->io_type == XFS_IO_UNWRITTEN || xfs_ioend_is_append(ioend)) - set_buffer_defer_completion(bh_result); } - /* * If this is O_DIRECT or the mpage code calling tell them how large the mapping * is, so that we can avoid repeated get_blocks calls. @@ -1519,9 +1525,11 @@ xfs_get_blocks_direct( /* * Complete a direct I/O write request. * - * If the private argument is non-NULL __xfs_get_blocks signals us that we - * need to issue a transaction to convert the range from unwritten to written - * extents. + * The ioend structure is passed from __xfs_get_blocks() to tell us what to do. + * If no ioend exists (i.e. @private == NULL) then the write IO is an overwrite + * wholly within the EOF and so there is nothing for us to do. Note that in this + * case the completion can be called in interrupt context, whereas if we have an + * ioend we will always be called in task context (i.e. from a workqueue). */ STATIC void xfs_end_io_direct_write( @@ -1535,7 +1543,13 @@ xfs_end_io_direct_write( struct xfs_mount *mp = ip->i_mount; struct xfs_ioend *ioend = private; - trace_xfs_gbmap_direct_endio(ip, offset, size, ioend->io_type, NULL); + trace_xfs_gbmap_direct_endio(ip, offset, size, + ioend ? ioend->io_type : 0, NULL); + + if (!ioend) { + ASSERT(offset + size <= i_size_read(inode)); + return; + } if (XFS_FORCED_SHUTDOWN(mp)) goto out_end_io; @@ -1548,12 +1562,12 @@ xfs_end_io_direct_write( /* * The ioend only maps whole blocks, while the IO may be sector aligned. - * Hence the ioend offset/size may not match the IO offset/size exactly, - * but should span it completely. Write the IO sizes into the ioend so - * that completion processing does the right thing. + * Hence the ioend offset/size may not match the IO offset/size exactly. + * Because we don't map overwrites within EOF into the ioend, the offset + * may not match, but only if the endio spans EOF. Either way, write + * the IO sizes into the ioend so that completion processing does the + * right thing. */ - ASSERT(size <= ioend->io_size); - ASSERT(offset >= ioend->io_offset); ASSERT(offset + size <= ioend->io_offset + ioend->io_size); ioend->io_size = size; ioend->io_offset = offset; @@ -1562,20 +1576,15 @@ xfs_end_io_direct_write( * The ioend tells us whether we are doing unwritten extent conversion * or an append transaction that updates the on-disk file size. These * cases are the only cases where we should *potentially* be needing - * to update the VFS inode size. When the ioend indicates this, we - * are *guaranteed* to be running in non-interrupt context. + * to update the VFS inode size. * * We need to update the in-core inode size here so that we don't end up - * with the on-disk inode size being outside the in-core inode size. - * While we can do this in the process context after the IO has - * completed, this does not work for AIO and hence we always update - * the in-core inode size here if necessary. + * with the on-disk inode size being outside the in-core inode size. We + * have no other method of updating EOF for AIO, so always do it here + * if necessary. */ - if (ioend->io_type == XFS_IO_UNWRITTEN || xfs_ioend_is_append(ioend)) { - if (offset + size > i_size_read(inode)) - i_size_write(inode, offset + size); - } else - ASSERT(offset + size <= i_size_read(inode)); + if (offset + size > i_size_read(inode)) + i_size_write(inode, offset + size); /* * If we are doing an append IO that needs to update the EOF on disk, @@ -1584,7 +1593,7 @@ xfs_end_io_direct_write( * result in the ioend processing passing on the error if it is * possible as we can't return it from here. */ - if (ioend->io_type == XFS_IO_OVERWRITE && xfs_ioend_is_append(ioend)) + if (ioend->io_type == XFS_IO_OVERWRITE) ioend->io_error = xfs_setfilesize_trans_alloc(ioend); out_end_io: diff --git a/fs/xfs/xfs_trace.h b/fs/xfs/xfs_trace.h index 0ae50e9847bb..4e0a5773eee4 100644 --- a/fs/xfs/xfs_trace.h +++ b/fs/xfs/xfs_trace.h @@ -1220,6 +1220,7 @@ DEFINE_IOMAP_EVENT(xfs_get_blocks_alloc); DEFINE_IOMAP_EVENT(xfs_gbmap_direct); DEFINE_IOMAP_EVENT(xfs_gbmap_direct_new); DEFINE_IOMAP_EVENT(xfs_gbmap_direct_update); +DEFINE_IOMAP_EVENT(xfs_gbmap_direct_none); DEFINE_IOMAP_EVENT(xfs_gbmap_direct_endio); DECLARE_EVENT_CLASS(xfs_simple_io_class, -- cgit v1.2.3-59-g8ed1b From b9d59846f73713d77f0f3fb784c7f84249fc2b93 Mon Sep 17 00:00:00 2001 From: Dave Chinner Date: Thu, 16 Apr 2015 22:03:07 +1000 Subject: xfs: DIO write completion size updates race xfs_end_io_direct_write() can race with other IO completions when updating the in-core inode size. The IO completion processing is not serialised for direct IO - they are done either under the IOLOCK_SHARED for non-AIO DIO, and without any IOLOCK held at all during AIO DIO completion. Hence the non-atomic test-and-set update of the in-core inode size is racy and can result in the in-core inode size going backwards if the race if hit just right. If the inode size goes backwards, this can trigger the EOF zeroing code to run incorrectly on the next IO, which then will zero data that has successfully been written to disk by a previous DIO. To fix this bug, we need to serialise the test/set updates of the in-core inode size. This first patch introduces locking around the relevant updates and checks in the DIO path. Because we now have an ioend in xfs_end_io_direct_write(), we know exactly then we are doing an IO that requires an in-core EOF update, and we know that they are not running in interrupt context. As such, we do not need to use irqsave() spinlock variants to protect against interrupts while the lock is held. Hence we can use an existing spinlock in the inode to do this serialisation and so not need to grow the struct xfs_inode just to work around this problem. This patch does not address the test/set EOF update in generic_file_write_direct() for various reasons - that will be done as a followup with separate explanation. Signed-off-by: Dave Chinner Reviewed-by: Brian Foster Signed-off-by: Dave Chinner --- fs/xfs/xfs_aops.c | 7 +++++++ fs/xfs/xfs_file.c | 13 ++++++++++++- 2 files changed, 19 insertions(+), 1 deletion(-) diff --git a/fs/xfs/xfs_aops.c b/fs/xfs/xfs_aops.c index c02a47453137..598b259fda04 100644 --- a/fs/xfs/xfs_aops.c +++ b/fs/xfs/xfs_aops.c @@ -1582,9 +1582,16 @@ xfs_end_io_direct_write( * with the on-disk inode size being outside the in-core inode size. We * have no other method of updating EOF for AIO, so always do it here * if necessary. + * + * We need to lock the test/set EOF update as we can be racing with + * other IO completions here to update the EOF. Failing to serialise + * here can result in EOF moving backwards and Bad Things Happen when + * that occurs. */ + spin_lock(&ip->i_flags_lock); if (offset + size > i_size_read(inode)) i_size_write(inode, offset + size); + spin_unlock(&ip->i_flags_lock); /* * If we are doing an append IO that needs to update the EOF on disk, diff --git a/fs/xfs/xfs_file.c b/fs/xfs/xfs_file.c index ce615d12fb44..2323b8b63183 100644 --- a/fs/xfs/xfs_file.c +++ b/fs/xfs/xfs_file.c @@ -565,8 +565,18 @@ restart: * write. If zeroing is needed and we are currently holding the * iolock shared, we need to update it to exclusive which implies * having to redo all checks before. + * + * We need to serialise against EOF updates that occur in IO + * completions here. We want to make sure that nobody is changing the + * size while we do this check until we have placed an IO barrier (i.e. + * hold the XFS_IOLOCK_EXCL) that prevents new IO from being dispatched. + * The spinlock effectively forms a memory barrier once we have the + * XFS_IOLOCK_EXCL so we are guaranteed to see the latest EOF value + * and hence be able to correctly determine if we need to run zeroing. */ + spin_lock(&ip->i_flags_lock); if (*pos > i_size_read(inode)) { + spin_unlock(&ip->i_flags_lock); if (*iolock == XFS_IOLOCK_SHARED) { xfs_rw_iunlock(ip, *iolock); *iolock = XFS_IOLOCK_EXCL; @@ -576,7 +586,8 @@ restart: error = xfs_zero_eof(ip, *pos, i_size_read(inode)); if (error) return error; - } + } else + spin_unlock(&ip->i_flags_lock); /* * Updating the timestamps will grab the ilock again from -- cgit v1.2.3-59-g8ed1b From 40c63fbc55a968383b8bb5cacad81585e80cd323 Mon Sep 17 00:00:00 2001 From: Dave Chinner Date: Thu, 16 Apr 2015 22:03:17 +1000 Subject: xfs: direct IO EOF zeroing needs to drain AIO When we are doing AIO DIO writes, the IOLOCK only provides an IO submission barrier. When we need to do EOF zeroing, we need to ensure that no other IO is in progress and all pending in-core EOF updates have been completed. This requires us to wait for all outstanding AIO DIO writes to the inode to complete and, if necessary, run their EOF updates. Once all the EOF updates are complete, we can then restart xfs_file_aio_write_checks() while holding the IOLOCK_EXCL, knowing that EOF is up to date and we have exclusive IO access to the file so we can run EOF block zeroing if we need to without interference. This gives EOF zeroing the same exclusivity against other IO as we provide truncate operations. Signed-off-by: Dave Chinner Reviewed-by: Brian Foster Signed-off-by: Dave Chinner --- fs/xfs/xfs_file.c | 10 ++++++++++ 1 file changed, 10 insertions(+) diff --git a/fs/xfs/xfs_file.c b/fs/xfs/xfs_file.c index 2323b8b63183..f6f0e9671919 100644 --- a/fs/xfs/xfs_file.c +++ b/fs/xfs/xfs_file.c @@ -581,6 +581,16 @@ restart: xfs_rw_iunlock(ip, *iolock); *iolock = XFS_IOLOCK_EXCL; xfs_rw_ilock(ip, *iolock); + + /* + * We now have an IO submission barrier in place, but + * AIO can do EOF updates during IO completion and hence + * we now need to wait for all of them to drain. Non-AIO + * DIO will have drained before we are given the + * XFS_IOLOCK_EXCL, and so for most cases this wait is a + * no-op. + */ + inode_dio_wait(inode); goto restart; } error = xfs_zero_eof(ip, *pos, i_size_read(inode)); -- cgit v1.2.3-59-g8ed1b From 0cefb29e6a63727bc7606c47fc538467594ef112 Mon Sep 17 00:00:00 2001 From: Dave Chinner Date: Thu, 16 Apr 2015 22:03:27 +1000 Subject: xfs: using generic_file_direct_write() is unnecessary generic_file_direct_write() does all sorts of things to make DIO work "sorta ok" with mixed buffered IO workloads. We already do most of this work in xfs_file_aio_dio_write() because of the locking requirements, so there's only a couple of things it does for us. The first thing is that it does a page cache invalidation after the ->direct_IO callout. This can easily be added to the XFS code. The second thing it does is that if data was written, it updates the iov_iter structure to reflect the data written, and then does EOF size updates if necessary. For XFS, these EOF size updates are now not necessary, as we do them safely and race-free in IO completion context. That leaves just the iov_iter update, and that's also moved to the XFS code. Therefore we don't need to call generic_file_direct_write() and in doing so remove redundant buffered writeback and page cache invalidation calls from the DIO submission path. We also remove a racy EOF size update, and make the DIO submission code in XFS much easier to follow. Wins all round, really. Signed-off-by: Dave Chinner Reviewed-by: Brian Foster Signed-off-by: Dave Chinner --- fs/xfs/xfs_file.c | 23 ++++++++++++++++++++--- 1 file changed, 20 insertions(+), 3 deletions(-) diff --git a/fs/xfs/xfs_file.c b/fs/xfs/xfs_file.c index f6f0e9671919..79ffb3e74f49 100644 --- a/fs/xfs/xfs_file.c +++ b/fs/xfs/xfs_file.c @@ -659,6 +659,8 @@ xfs_file_dio_aio_write( int iolock; size_t count = iov_iter_count(from); loff_t pos = iocb->ki_pos; + loff_t end; + struct iov_iter data; struct xfs_buftarg *target = XFS_IS_REALTIME_INODE(ip) ? mp->m_rtdev_targp : mp->m_ddev_targp; @@ -698,10 +700,11 @@ xfs_file_dio_aio_write( if (ret) goto out; iov_iter_truncate(from, count); + end = pos + count - 1; if (mapping->nrpages) { ret = filemap_write_and_wait_range(VFS_I(ip)->i_mapping, - pos, pos + count - 1); + pos, end); if (ret) goto out; /* @@ -711,7 +714,7 @@ xfs_file_dio_aio_write( */ ret = invalidate_inode_pages2_range(VFS_I(ip)->i_mapping, pos >> PAGE_CACHE_SHIFT, - (pos + count - 1) >> PAGE_CACHE_SHIFT); + end >> PAGE_CACHE_SHIFT); WARN_ON_ONCE(ret); ret = 0; } @@ -728,8 +731,22 @@ xfs_file_dio_aio_write( } trace_xfs_file_direct_write(ip, count, iocb->ki_pos, 0); - ret = generic_file_direct_write(iocb, from, pos); + data = *from; + ret = mapping->a_ops->direct_IO(WRITE, iocb, &data, pos); + + /* see generic_file_direct_write() for why this is necessary */ + if (mapping->nrpages) { + invalidate_inode_pages2_range(mapping, + pos >> PAGE_CACHE_SHIFT, + end >> PAGE_CACHE_SHIFT); + } + + if (ret > 0) { + pos += ret; + iov_iter_advance(from, ret); + iocb->ki_pos = pos; + } out: xfs_rw_iunlock(ip, iolock); -- cgit v1.2.3-59-g8ed1b