From 3136e8bb3054d3bb68942f8f1ee6c26c05f798b0 Mon Sep 17 00:00:00 2001
From: Brian Foster <bfoster@redhat.com>
Date: Mon, 12 Oct 2015 16:02:05 +1100
Subject: xfs: always drain dio before extending aio write submission

XFS supports and typically allows concurrent asynchronous direct I/O
submission to a single file. One exception to the rule is that file
extending dio writes that start beyond the current EOF (e.g.,
potentially create a hole at EOF) require exclusive I/O access to the
file. This is because such writes must zero any pre-existing blocks
beyond EOF that are exposed by virtue of now residing within EOF as a
result of the write about to be submitted.

Before EOF zeroing can occur, the current file i_size must be stabilized
to avoid data corruption. In this scenario, XFS upgrades the iolock to
exclude any further I/O submission, waits on in-flight I/O to complete
to ensure i_size is up to date (i_size is updated on dio write
completion) and restarts the various checks against the state of the
file. The problem is that this protection sequence is triggered only
when the iolock is currently held shared. While this is true for async
dio in most cases, the caller may upgrade the lock in advance based on
arbitrary circumstances with respect to EOF zeroing. For example, the
iolock is always acquired exclusively if the start offset is not block
aligned. This means that even though the iolock is already held
exclusive for such I/Os, pending I/O is not drained and thus EOF zeroing
can occur based on an unstable i_size.

This problem has been reproduced as guest data corruption in virtual
machines with file-backed qcow2 virtual disks hosted on an XFS
filesystem. The virtual disks must be configured with aio=native mode
and the must not be truncated out to the maximum file size (as some virt
managers will do).

Update xfs_file_aio_write_checks() to unconditionally drain in-flight
dio before EOF zeroing can occur. Rather than trigger the wait based on
iolock state, use a new flag and upgrade the iolock when necessary. Note
that this results in a full restart of the inode checks even when the
iolock was already held exclusive when technically it is only required
to recheck i_size. This should be a rare enough occurrence that it is
preferable to keep the code simple rather than create an alternate
restart jump target.

Signed-off-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
---
 fs/xfs/xfs_file.c | 15 +++++++++------
 1 file changed, 9 insertions(+), 6 deletions(-)

(limited to 'fs/xfs/xfs_file.c')

diff --git a/fs/xfs/xfs_file.c b/fs/xfs/xfs_file.c
index e78feb400e22..347b3e07ec2b 100644
--- a/fs/xfs/xfs_file.c
+++ b/fs/xfs/xfs_file.c
@@ -574,6 +574,7 @@ xfs_file_aio_write_checks(
 	struct xfs_inode	*ip = XFS_I(inode);
 	ssize_t			error = 0;
 	size_t			count = iov_iter_count(from);
+	bool			drained_dio = false;
 
 restart:
 	error = generic_write_checks(iocb, from);
@@ -611,12 +612,13 @@ restart:
 		bool	zero = false;
 
 		spin_unlock(&ip->i_flags_lock);
-		if (*iolock == XFS_IOLOCK_SHARED) {
-			xfs_rw_iunlock(ip, *iolock);
-			*iolock = XFS_IOLOCK_EXCL;
-			xfs_rw_ilock(ip, *iolock);
-			iov_iter_reexpand(from, count);
-
+		if (!drained_dio) {
+			if (*iolock == XFS_IOLOCK_SHARED) {
+				xfs_rw_iunlock(ip, *iolock);
+				*iolock = XFS_IOLOCK_EXCL;
+				xfs_rw_ilock(ip, *iolock);
+				iov_iter_reexpand(from, count);
+			}
 			/*
 			 * We now have an IO submission barrier in place, but
 			 * AIO can do EOF updates during IO completion and hence
@@ -626,6 +628,7 @@ restart:
 			 * no-op.
 			 */
 			inode_dio_wait(inode);
+			drained_dio = true;
 			goto restart;
 		}
 		error = xfs_zero_eof(ip, iocb->ki_pos, i_size_read(inode), &zero);
-- 
cgit v1.2.3-59-g8ed1b


From 0a50f162af6ddc2db02c9edc5bbb823c336100a4 Mon Sep 17 00:00:00 2001
From: Brian Foster <bfoster@redhat.com>
Date: Mon, 12 Oct 2015 16:02:08 +1100
Subject: xfs: add an xfs_zero_eof() tracepoint

Add a tracepoint in xfs_zero_eof() to facilitate tracking and debugging
EOF zeroing events. This has proven useful in the context of other
direct I/O tracepoints to ensure EOF zeroing occurs within appropriate
file ranges.

Signed-off-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
---
 fs/xfs/xfs_file.c  | 2 ++
 fs/xfs/xfs_trace.h | 1 +
 2 files changed, 3 insertions(+)

(limited to 'fs/xfs/xfs_file.c')

diff --git a/fs/xfs/xfs_file.c b/fs/xfs/xfs_file.c
index 347b3e07ec2b..541dcfbc6f49 100644
--- a/fs/xfs/xfs_file.c
+++ b/fs/xfs/xfs_file.c
@@ -482,6 +482,8 @@ xfs_zero_eof(
 	ASSERT(xfs_isilocked(ip, XFS_IOLOCK_EXCL));
 	ASSERT(offset > isize);
 
+	trace_xfs_zero_eof(ip, isize, offset - isize);
+
 	/*
 	 * First handle zeroing the block on which isize resides.
 	 *
diff --git a/fs/xfs/xfs_trace.h b/fs/xfs/xfs_trace.h
index 5ed36b1e04c1..957f5ccdd84f 100644
--- a/fs/xfs/xfs_trace.h
+++ b/fs/xfs/xfs_trace.h
@@ -1312,6 +1312,7 @@ DEFINE_SIMPLE_IO_EVENT(xfs_delalloc_enospc);
 DEFINE_SIMPLE_IO_EVENT(xfs_unwritten_convert);
 DEFINE_SIMPLE_IO_EVENT(xfs_get_blocks_notfound);
 DEFINE_SIMPLE_IO_EVENT(xfs_setfilesize);
+DEFINE_SIMPLE_IO_EVENT(xfs_zero_eof);
 
 DECLARE_EVENT_CLASS(xfs_itrunc_class,
 	TP_PROTO(struct xfs_inode *ip, xfs_fsize_t new_size),
-- 
cgit v1.2.3-59-g8ed1b


From ff6d6af2351caea7db681f4539d0d893e400557a Mon Sep 17 00:00:00 2001
From: Bill O'Donnell <billodo@redhat.com>
Date: Mon, 12 Oct 2015 18:21:22 +1100
Subject: xfs: per-filesystem stats counter implementation

This patch modifies the stats counting macros and the callers
to those macros to properly increment, decrement, and add-to
the xfs stats counts. The counts for global and per-fs stats
are correctly advanced, and cleared by writing a "1" to the
corresponding clear file.

global counts: /sys/fs/xfs/stats/stats
per-fs counts: /sys/fs/xfs/sda*/stats/stats

global clear:  /sys/fs/xfs/stats/stats_clear
per-fs clear:  /sys/fs/xfs/sda*/stats/stats_clear

[dchinner: cleaned up macro variables, removed CONFIG_FS_PROC around
 stats structures and macros. ]

Signed-off-by: Bill O'Donnell <billodo@redhat.com>
Reviewed-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
---
 fs/xfs/libxfs/xfs_alloc.c |  8 ++++----
 fs/xfs/libxfs/xfs_attr.c  |  6 +++---
 fs/xfs/libxfs/xfs_bmap.c  | 20 ++++++++++----------
 fs/xfs/libxfs/xfs_btree.h | 39 +++++++++++++++++++++++----------------
 fs/xfs/libxfs/xfs_dir2.c  |  6 +++---
 fs/xfs/xfs_attr_list.c    |  2 +-
 fs/xfs/xfs_buf.c          | 18 +++++++++---------
 fs/xfs/xfs_dir2_readdir.c |  2 +-
 fs/xfs/xfs_dquot.c        | 14 +++++++-------
 fs/xfs/xfs_file.c         | 12 ++++++------
 fs/xfs/xfs_icache.c       | 18 +++++++++---------
 fs/xfs/xfs_inode.c        |  6 +++---
 fs/xfs/xfs_ioctl.c        |  2 +-
 fs/xfs/xfs_iomap.c        |  4 ++--
 fs/xfs/xfs_iops.c         |  4 ++--
 fs/xfs/xfs_log.c          | 22 +++++++++++-----------
 fs/xfs/xfs_qm.c           | 14 +++++++-------
 fs/xfs/xfs_stats.h        | 28 +++++++++++++++++-----------
 fs/xfs/xfs_super.c        |  6 +++---
 fs/xfs/xfs_trans.c        |  6 +++---
 fs/xfs/xfs_trans_ail.c    | 12 ++++++------
 21 files changed, 131 insertions(+), 118 deletions(-)

(limited to 'fs/xfs/xfs_file.c')

diff --git a/fs/xfs/libxfs/xfs_alloc.c b/fs/xfs/libxfs/xfs_alloc.c
index ffad7f20342f..9b5da7e3b5c2 100644
--- a/fs/xfs/libxfs/xfs_alloc.c
+++ b/fs/xfs/libxfs/xfs_alloc.c
@@ -651,8 +651,8 @@ xfs_alloc_ag_vextent(
 				 -((long)(args->len)));
 	}
 
-	XFS_STATS_INC(xs_allocx);
-	XFS_STATS_ADD(xs_allocb, args->len);
+	XFS_STATS_INC(args->mp, xs_allocx);
+	XFS_STATS_ADD(args->mp, xs_allocb, args->len);
 	return error;
 }
 
@@ -1808,8 +1808,8 @@ xfs_free_ag_extent(
 
 	if (!isfl)
 		xfs_trans_mod_sb(tp, XFS_TRANS_SB_FDBLOCKS, (long)len);
-	XFS_STATS_INC(xs_freex);
-	XFS_STATS_ADD(xs_freeb, len);
+	XFS_STATS_INC(mp, xs_freex);
+	XFS_STATS_ADD(mp, xs_freeb, len);
 
 	trace_xfs_free_extent(mp, agno, bno, len, isfl, haveleft, haveright);
 
diff --git a/fs/xfs/libxfs/xfs_attr.c b/fs/xfs/libxfs/xfs_attr.c
index ff065578969f..f949818fa1c7 100644
--- a/fs/xfs/libxfs/xfs_attr.c
+++ b/fs/xfs/libxfs/xfs_attr.c
@@ -125,7 +125,7 @@ xfs_attr_get(
 	uint			lock_mode;
 	int			error;
 
-	XFS_STATS_INC(xs_attr_get);
+	XFS_STATS_INC(ip->i_mount, xs_attr_get);
 
 	if (XFS_FORCED_SHUTDOWN(ip->i_mount))
 		return -EIO;
@@ -209,7 +209,7 @@ xfs_attr_set(
 	int			rsvd = (flags & ATTR_ROOT) != 0;
 	int			error, err2, committed, local;
 
-	XFS_STATS_INC(xs_attr_set);
+	XFS_STATS_INC(mp, xs_attr_set);
 
 	if (XFS_FORCED_SHUTDOWN(dp->i_mount))
 		return -EIO;
@@ -412,7 +412,7 @@ xfs_attr_remove(
 	xfs_fsblock_t		firstblock;
 	int			error;
 
-	XFS_STATS_INC(xs_attr_remove);
+	XFS_STATS_INC(mp, xs_attr_remove);
 
 	if (XFS_FORCED_SHUTDOWN(dp->i_mount))
 		return -EIO;
diff --git a/fs/xfs/libxfs/xfs_bmap.c b/fs/xfs/libxfs/xfs_bmap.c
index 8e2010d53b07..5256fe59623b 100644
--- a/fs/xfs/libxfs/xfs_bmap.c
+++ b/fs/xfs/libxfs/xfs_bmap.c
@@ -1435,7 +1435,7 @@ xfs_bmap_search_extents(
 	xfs_ifork_t	*ifp;		/* inode fork pointer */
 	xfs_bmbt_rec_host_t  *ep;            /* extent record pointer */
 
-	XFS_STATS_INC(xs_look_exlist);
+	XFS_STATS_INC(ip->i_mount, xs_look_exlist);
 	ifp = XFS_IFORK_PTR(ip, fork);
 
 	ep = xfs_bmap_search_multi_extents(ifp, bno, eofp, lastxp, gotp, prevp);
@@ -1732,7 +1732,7 @@ xfs_bmap_add_extent_delay_real(
 	ASSERT(!bma->cur ||
 	       (bma->cur->bc_private.b.flags & XFS_BTCUR_BPRV_WASDEL));
 
-	XFS_STATS_INC(xs_add_exlist);
+	XFS_STATS_INC(mp, xs_add_exlist);
 
 #define	LEFT		r[0]
 #define	RIGHT		r[1]
@@ -2286,7 +2286,7 @@ xfs_bmap_add_extent_unwritten_real(
 	ASSERT(*idx <= ifp->if_bytes / sizeof(struct xfs_bmbt_rec));
 	ASSERT(!isnullstartblock(new->br_startblock));
 
-	XFS_STATS_INC(xs_add_exlist);
+	XFS_STATS_INC(mp, xs_add_exlist);
 
 #define	LEFT		r[0]
 #define	RIGHT		r[1]
@@ -2946,7 +2946,7 @@ xfs_bmap_add_extent_hole_real(
 	ASSERT(!bma->cur ||
 	       !(bma->cur->bc_private.b.flags & XFS_BTCUR_BPRV_WASDEL));
 
-	XFS_STATS_INC(xs_add_exlist);
+	XFS_STATS_INC(mp, xs_add_exlist);
 
 	state = 0;
 	if (whichfork == XFS_ATTR_FORK)
@@ -4036,7 +4036,7 @@ xfs_bmapi_read(
 	if (XFS_FORCED_SHUTDOWN(mp))
 		return -EIO;
 
-	XFS_STATS_INC(xs_blk_mapr);
+	XFS_STATS_INC(mp, xs_blk_mapr);
 
 	ifp = XFS_IFORK_PTR(ip, whichfork);
 
@@ -4221,7 +4221,7 @@ xfs_bmapi_delay(
 	if (XFS_FORCED_SHUTDOWN(mp))
 		return -EIO;
 
-	XFS_STATS_INC(xs_blk_mapw);
+	XFS_STATS_INC(mp, xs_blk_mapw);
 
 	if (!(ifp->if_flags & XFS_IFEXTENTS)) {
 		error = xfs_iread_extents(NULL, ip, XFS_DATA_FORK);
@@ -4525,7 +4525,7 @@ xfs_bmapi_write(
 
 	ifp = XFS_IFORK_PTR(ip, whichfork);
 
-	XFS_STATS_INC(xs_blk_mapw);
+	XFS_STATS_INC(mp, xs_blk_mapw);
 
 	if (*firstblock == NULLFSBLOCK) {
 		if (XFS_IFORK_FORMAT(ip, whichfork) == XFS_DINODE_FMT_BTREE)
@@ -4718,12 +4718,12 @@ xfs_bmap_del_extent(
 	xfs_filblks_t		temp2;	/* for indirect length calculations */
 	int			state = 0;
 
-	XFS_STATS_INC(xs_del_exlist);
+	mp = ip->i_mount;
+	XFS_STATS_INC(mp, xs_del_exlist);
 
 	if (whichfork == XFS_ATTR_FORK)
 		state |= BMAP_ATTRFORK;
 
-	mp = ip->i_mount;
 	ifp = XFS_IFORK_PTR(ip, whichfork);
 	ASSERT((*idx >= 0) && (*idx < ifp->if_bytes /
 		(uint)sizeof(xfs_bmbt_rec_t)));
@@ -5070,7 +5070,7 @@ xfs_bunmapi(
 		*done = 1;
 		return 0;
 	}
-	XFS_STATS_INC(xs_blk_unmap);
+	XFS_STATS_INC(mp, xs_blk_unmap);
 	isrt = (whichfork == XFS_DATA_FORK) && XFS_IS_REALTIME_INODE(ip);
 	start = bno;
 	bno = start + len - 1;
diff --git a/fs/xfs/libxfs/xfs_btree.h b/fs/xfs/libxfs/xfs_btree.h
index 8f18bab73ea5..992dec0638f3 100644
--- a/fs/xfs/libxfs/xfs_btree.h
+++ b/fs/xfs/libxfs/xfs_btree.h
@@ -84,31 +84,38 @@ union xfs_btree_rec {
 /*
  * Generic stats interface
  */
-#define __XFS_BTREE_STATS_INC(type, stat) \
-	XFS_STATS_INC(xs_ ## type ## _2_ ## stat)
-#define XFS_BTREE_STATS_INC(cur, stat)  \
+#define __XFS_BTREE_STATS_INC(mp, type, stat) \
+	XFS_STATS_INC(mp, xs_ ## type ## _2_ ## stat)
+#define XFS_BTREE_STATS_INC(cur, stat)	\
 do {    \
+	struct xfs_mount *__mp = cur->bc_mp; \
 	switch (cur->bc_btnum) {  \
-	case XFS_BTNUM_BNO: __XFS_BTREE_STATS_INC(abtb, stat); break;	\
-	case XFS_BTNUM_CNT: __XFS_BTREE_STATS_INC(abtc, stat); break;	\
-	case XFS_BTNUM_BMAP: __XFS_BTREE_STATS_INC(bmbt, stat); break;	\
-	case XFS_BTNUM_INO: __XFS_BTREE_STATS_INC(ibt, stat); break;	\
-	case XFS_BTNUM_FINO: __XFS_BTREE_STATS_INC(fibt, stat); break;	\
+	case XFS_BTNUM_BNO: __XFS_BTREE_STATS_INC(__mp, abtb, stat); break; \
+	case XFS_BTNUM_CNT: __XFS_BTREE_STATS_INC(__mp, abtc, stat); break; \
+	case XFS_BTNUM_BMAP: __XFS_BTREE_STATS_INC(__mp, bmbt, stat); break; \
+	case XFS_BTNUM_INO: __XFS_BTREE_STATS_INC(__mp, ibt, stat); break; \
+	case XFS_BTNUM_FINO: __XFS_BTREE_STATS_INC(__mp, fibt, stat); break; \
 	case XFS_BTNUM_MAX: ASSERT(0); /* fucking gcc */ ; break;	\
 	}       \
 } while (0)
 
-#define __XFS_BTREE_STATS_ADD(type, stat, val) \
-	XFS_STATS_ADD(xs_ ## type ## _2_ ## stat, val)
+#define __XFS_BTREE_STATS_ADD(mp, type, stat, val) \
+	XFS_STATS_ADD(mp, xs_ ## type ## _2_ ## stat, val)
 #define XFS_BTREE_STATS_ADD(cur, stat, val)  \
 do {    \
+	struct xfs_mount *__mp = cur->bc_mp; \
 	switch (cur->bc_btnum) {  \
-	case XFS_BTNUM_BNO: __XFS_BTREE_STATS_ADD(abtb, stat, val); break; \
-	case XFS_BTNUM_CNT: __XFS_BTREE_STATS_ADD(abtc, stat, val); break; \
-	case XFS_BTNUM_BMAP: __XFS_BTREE_STATS_ADD(bmbt, stat, val); break; \
-	case XFS_BTNUM_INO: __XFS_BTREE_STATS_ADD(ibt, stat, val); break; \
-	case XFS_BTNUM_FINO: __XFS_BTREE_STATS_ADD(fibt, stat, val); break; \
-	case XFS_BTNUM_MAX: ASSERT(0); /* fucking gcc */ ; break;	\
+	case XFS_BTNUM_BNO:	\
+		__XFS_BTREE_STATS_ADD(__mp, abtb, stat, val); break; \
+	case XFS_BTNUM_CNT:	\
+		__XFS_BTREE_STATS_ADD(__mp, abtc, stat, val); break; \
+	case XFS_BTNUM_BMAP:	\
+		__XFS_BTREE_STATS_ADD(__mp, bmbt, stat, val); break; \
+	case XFS_BTNUM_INO:	\
+		__XFS_BTREE_STATS_ADD(__mp, ibt, stat, val); break; \
+	case XFS_BTNUM_FINO:	\
+		__XFS_BTREE_STATS_ADD(__mp, fibt, stat, val); break; \
+	case XFS_BTNUM_MAX: ASSERT(0); /* fucking gcc */ ; break; \
 	}       \
 } while (0)
 
diff --git a/fs/xfs/libxfs/xfs_dir2.c b/fs/xfs/libxfs/xfs_dir2.c
index 9de401d297e5..2fb53a5c0a74 100644
--- a/fs/xfs/libxfs/xfs_dir2.c
+++ b/fs/xfs/libxfs/xfs_dir2.c
@@ -271,7 +271,7 @@ xfs_dir_createname(
 		rval = xfs_dir_ino_validate(tp->t_mountp, inum);
 		if (rval)
 			return rval;
-		XFS_STATS_INC(xs_dir_create);
+		XFS_STATS_INC(dp->i_mount, xs_dir_create);
 	}
 
 	args = kmem_zalloc(sizeof(*args), KM_SLEEP | KM_NOFS);
@@ -365,7 +365,7 @@ xfs_dir_lookup(
 	int		lock_mode;
 
 	ASSERT(S_ISDIR(dp->i_d.di_mode));
-	XFS_STATS_INC(xs_dir_lookup);
+	XFS_STATS_INC(dp->i_mount, xs_dir_lookup);
 
 	/*
 	 * We need to use KM_NOFS here so that lockdep will not throw false
@@ -444,7 +444,7 @@ xfs_dir_removename(
 	int		v;		/* type-checking value */
 
 	ASSERT(S_ISDIR(dp->i_d.di_mode));
-	XFS_STATS_INC(xs_dir_remove);
+	XFS_STATS_INC(dp->i_mount, xs_dir_remove);
 
 	args = kmem_zalloc(sizeof(*args), KM_SLEEP | KM_NOFS);
 	if (!args)
diff --git a/fs/xfs/xfs_attr_list.c b/fs/xfs/xfs_attr_list.c
index 65fb37a18e92..0ef7c2ed3f8a 100644
--- a/fs/xfs/xfs_attr_list.c
+++ b/fs/xfs/xfs_attr_list.c
@@ -511,7 +511,7 @@ xfs_attr_list_int(
 	xfs_inode_t *dp = context->dp;
 	uint		lock_mode;
 
-	XFS_STATS_INC(xs_attr_list);
+	XFS_STATS_INC(dp->i_mount, xs_attr_list);
 
 	if (XFS_FORCED_SHUTDOWN(dp->i_mount))
 		return -EIO;
diff --git a/fs/xfs/xfs_buf.c b/fs/xfs/xfs_buf.c
index 8ecffb35935b..90815c22b22d 100644
--- a/fs/xfs/xfs_buf.c
+++ b/fs/xfs/xfs_buf.c
@@ -201,7 +201,7 @@ _xfs_buf_alloc(
 	atomic_set(&bp->b_pin_count, 0);
 	init_waitqueue_head(&bp->b_waiters);
 
-	XFS_STATS_INC(xb_create);
+	XFS_STATS_INC(target->bt_mount, xb_create);
 	trace_xfs_buf_init(bp, _RET_IP_);
 
 	return bp;
@@ -357,12 +357,12 @@ retry:
 		"possible memory allocation deadlock in %s (mode:0x%x)",
 					__func__, gfp_mask);
 
-			XFS_STATS_INC(xb_page_retries);
+			XFS_STATS_INC(bp->b_target->bt_mount, xb_page_retries);
 			congestion_wait(BLK_RW_ASYNC, HZ/50);
 			goto retry;
 		}
 
-		XFS_STATS_INC(xb_page_found);
+		XFS_STATS_INC(bp->b_target->bt_mount, xb_page_found);
 
 		nbytes = min_t(size_t, size, PAGE_SIZE - offset);
 		size -= nbytes;
@@ -516,7 +516,7 @@ _xfs_buf_find(
 		new_bp->b_pag = pag;
 		spin_unlock(&pag->pag_buf_lock);
 	} else {
-		XFS_STATS_INC(xb_miss_locked);
+		XFS_STATS_INC(btp->bt_mount, xb_miss_locked);
 		spin_unlock(&pag->pag_buf_lock);
 		xfs_perag_put(pag);
 	}
@@ -529,11 +529,11 @@ found:
 	if (!xfs_buf_trylock(bp)) {
 		if (flags & XBF_TRYLOCK) {
 			xfs_buf_rele(bp);
-			XFS_STATS_INC(xb_busy_locked);
+			XFS_STATS_INC(btp->bt_mount, xb_busy_locked);
 			return NULL;
 		}
 		xfs_buf_lock(bp);
-		XFS_STATS_INC(xb_get_locked_waited);
+		XFS_STATS_INC(btp->bt_mount, xb_get_locked_waited);
 	}
 
 	/*
@@ -549,7 +549,7 @@ found:
 	}
 
 	trace_xfs_buf_find(bp, flags, _RET_IP_);
-	XFS_STATS_INC(xb_get_locked);
+	XFS_STATS_INC(btp->bt_mount, xb_get_locked);
 	return bp;
 }
 
@@ -603,7 +603,7 @@ found:
 		}
 	}
 
-	XFS_STATS_INC(xb_get);
+	XFS_STATS_INC(target->bt_mount, xb_get);
 	trace_xfs_buf_get(bp, flags, _RET_IP_);
 	return bp;
 }
@@ -643,7 +643,7 @@ xfs_buf_read_map(
 		trace_xfs_buf_read(bp, flags, _RET_IP_);
 
 		if (!XFS_BUF_ISDONE(bp)) {
-			XFS_STATS_INC(xb_get_read);
+			XFS_STATS_INC(target->bt_mount, xb_get_read);
 			bp->b_ops = ops;
 			_xfs_buf_read(bp, flags);
 		} else if (flags & XBF_ASYNC) {
diff --git a/fs/xfs/xfs_dir2_readdir.c b/fs/xfs/xfs_dir2_readdir.c
index a989a9c7edb7..642d55d10075 100644
--- a/fs/xfs/xfs_dir2_readdir.c
+++ b/fs/xfs/xfs_dir2_readdir.c
@@ -666,7 +666,7 @@ xfs_readdir(
 		return -EIO;
 
 	ASSERT(S_ISDIR(dp->i_d.di_mode));
-	XFS_STATS_INC(xs_dir_getdents);
+	XFS_STATS_INC(dp->i_mount, xs_dir_getdents);
 
 	args.dp = dp;
 	args.geo = dp->i_mount->m_dir_geo;
diff --git a/fs/xfs/xfs_dquot.c b/fs/xfs/xfs_dquot.c
index 30cb3afb67f0..7ac6c5c586cb 100644
--- a/fs/xfs/xfs_dquot.c
+++ b/fs/xfs/xfs_dquot.c
@@ -75,9 +75,9 @@ xfs_qm_dqdestroy(
 	ASSERT(list_empty(&dqp->q_lru));
 
 	mutex_destroy(&dqp->q_qlock);
-	kmem_zone_free(xfs_qm_dqzone, dqp);
 
-	XFS_STATS_DEC(xs_qm_dquot);
+	XFS_STATS_DEC(dqp->q_mount, xs_qm_dquot);
+	kmem_zone_free(xfs_qm_dqzone, dqp);
 }
 
 /*
@@ -605,7 +605,7 @@ xfs_qm_dqread(
 		break;
 	}
 
-	XFS_STATS_INC(xs_qm_dquot);
+	XFS_STATS_INC(mp, xs_qm_dquot);
 
 	trace_xfs_dqread(dqp);
 
@@ -747,12 +747,12 @@ restart:
 		mutex_unlock(&qi->qi_tree_lock);
 
 		trace_xfs_dqget_hit(dqp);
-		XFS_STATS_INC(xs_qm_dqcachehits);
+		XFS_STATS_INC(mp, xs_qm_dqcachehits);
 		*O_dqpp = dqp;
 		return 0;
 	}
 	mutex_unlock(&qi->qi_tree_lock);
-	XFS_STATS_INC(xs_qm_dqcachemisses);
+	XFS_STATS_INC(mp, xs_qm_dqcachemisses);
 
 	/*
 	 * Dquot cache miss. We don't want to keep the inode lock across
@@ -806,7 +806,7 @@ restart:
 		mutex_unlock(&qi->qi_tree_lock);
 		trace_xfs_dqget_dup(dqp);
 		xfs_qm_dqdestroy(dqp);
-		XFS_STATS_INC(xs_qm_dquot_dups);
+		XFS_STATS_INC(mp, xs_qm_dquot_dups);
 		goto restart;
 	}
 
@@ -846,7 +846,7 @@ xfs_qm_dqput(
 		trace_xfs_dqput_free(dqp);
 
 		if (list_lru_add(&qi->qi_lru, &dqp->q_lru))
-			XFS_STATS_INC(xs_qm_dquot_unused);
+			XFS_STATS_INC(dqp->q_mount, xs_qm_dquot_unused);
 	}
 	xfs_dqunlock(dqp);
 }
diff --git a/fs/xfs/xfs_file.c b/fs/xfs/xfs_file.c
index e78feb400e22..088e50923830 100644
--- a/fs/xfs/xfs_file.c
+++ b/fs/xfs/xfs_file.c
@@ -287,7 +287,7 @@ xfs_file_read_iter(
 	xfs_fsize_t		n;
 	loff_t			pos = iocb->ki_pos;
 
-	XFS_STATS_INC(xs_read_calls);
+	XFS_STATS_INC(mp, xs_read_calls);
 
 	if (unlikely(iocb->ki_flags & IOCB_DIRECT))
 		ioflags |= XFS_IO_ISDIRECT;
@@ -365,7 +365,7 @@ xfs_file_read_iter(
 
 	ret = generic_file_read_iter(iocb, to);
 	if (ret > 0)
-		XFS_STATS_ADD(xs_read_bytes, ret);
+		XFS_STATS_ADD(mp, xs_read_bytes, ret);
 
 	xfs_rw_iunlock(ip, XFS_IOLOCK_SHARED);
 	return ret;
@@ -383,7 +383,7 @@ xfs_file_splice_read(
 	int			ioflags = 0;
 	ssize_t			ret;
 
-	XFS_STATS_INC(xs_read_calls);
+	XFS_STATS_INC(ip->i_mount, xs_read_calls);
 
 	if (infilp->f_mode & FMODE_NOCMTIME)
 		ioflags |= XFS_IO_INVIS;
@@ -401,7 +401,7 @@ xfs_file_splice_read(
 	else
 		ret = generic_file_splice_read(infilp, ppos, pipe, count, flags);
 	if (ret > 0)
-		XFS_STATS_ADD(xs_read_bytes, ret);
+		XFS_STATS_ADD(ip->i_mount, xs_read_bytes, ret);
 
 	xfs_rw_iunlock(ip, XFS_IOLOCK_SHARED);
 	return ret;
@@ -867,7 +867,7 @@ xfs_file_write_iter(
 	ssize_t			ret;
 	size_t			ocount = iov_iter_count(from);
 
-	XFS_STATS_INC(xs_write_calls);
+	XFS_STATS_INC(ip->i_mount, xs_write_calls);
 
 	if (ocount == 0)
 		return 0;
@@ -883,7 +883,7 @@ xfs_file_write_iter(
 	if (ret > 0) {
 		ssize_t err;
 
-		XFS_STATS_ADD(xs_write_bytes, ret);
+		XFS_STATS_ADD(ip->i_mount, xs_write_bytes, ret);
 
 		/* Handle various SYNC-type writes */
 		err = generic_write_sync(file, iocb->ki_pos - ret, ret);
diff --git a/fs/xfs/xfs_icache.c b/fs/xfs/xfs_icache.c
index 0a326bd64d4e..d7a490f24ead 100644
--- a/fs/xfs/xfs_icache.c
+++ b/fs/xfs/xfs_icache.c
@@ -63,7 +63,7 @@ xfs_inode_alloc(
 		return NULL;
 	}
 
-	XFS_STATS_INC(vn_active);
+	XFS_STATS_INC(mp, vn_active);
 	ASSERT(atomic_read(&ip->i_pincount) == 0);
 	ASSERT(!spin_is_locked(&ip->i_flags_lock));
 	ASSERT(!xfs_isiflocked(ip));
@@ -129,7 +129,7 @@ xfs_inode_free(
 	/* asserts to verify all state is correct here */
 	ASSERT(atomic_read(&ip->i_pincount) == 0);
 	ASSERT(!xfs_isiflocked(ip));
-	XFS_STATS_DEC(vn_active);
+	XFS_STATS_DEC(ip->i_mount, vn_active);
 
 	call_rcu(&VFS_I(ip)->i_rcu, xfs_inode_free_callback);
 }
@@ -159,7 +159,7 @@ xfs_iget_cache_hit(
 	spin_lock(&ip->i_flags_lock);
 	if (ip->i_ino != ino) {
 		trace_xfs_iget_skip(ip);
-		XFS_STATS_INC(xs_ig_frecycle);
+		XFS_STATS_INC(mp, xs_ig_frecycle);
 		error = -EAGAIN;
 		goto out_error;
 	}
@@ -177,7 +177,7 @@ xfs_iget_cache_hit(
 	 */
 	if (ip->i_flags & (XFS_INEW|XFS_IRECLAIM)) {
 		trace_xfs_iget_skip(ip);
-		XFS_STATS_INC(xs_ig_frecycle);
+		XFS_STATS_INC(mp, xs_ig_frecycle);
 		error = -EAGAIN;
 		goto out_error;
 	}
@@ -259,7 +259,7 @@ xfs_iget_cache_hit(
 		xfs_ilock(ip, lock_flags);
 
 	xfs_iflags_clear(ip, XFS_ISTALE | XFS_IDONTCACHE);
-	XFS_STATS_INC(xs_ig_found);
+	XFS_STATS_INC(mp, xs_ig_found);
 
 	return 0;
 
@@ -342,7 +342,7 @@ xfs_iget_cache_miss(
 	error = radix_tree_insert(&pag->pag_ici_root, agino, ip);
 	if (unlikely(error)) {
 		WARN_ON(error != -EEXIST);
-		XFS_STATS_INC(xs_ig_dup);
+		XFS_STATS_INC(mp, xs_ig_dup);
 		error = -EAGAIN;
 		goto out_preload_end;
 	}
@@ -412,7 +412,7 @@ xfs_iget(
 	if (!ino || XFS_INO_TO_AGNO(mp, ino) >= mp->m_sb.sb_agcount)
 		return -EINVAL;
 
-	XFS_STATS_INC(xs_ig_attempts);
+	XFS_STATS_INC(mp, xs_ig_attempts);
 
 	/* get the perag structure and ensure that it's inode capable */
 	pag = xfs_perag_get(mp, XFS_INO_TO_AGNO(mp, ino));
@@ -429,7 +429,7 @@ again:
 			goto out_error_or_again;
 	} else {
 		rcu_read_unlock();
-		XFS_STATS_INC(xs_ig_missed);
+		XFS_STATS_INC(mp, xs_ig_missed);
 
 		error = xfs_iget_cache_miss(mp, pag, tp, ino, &ip,
 							flags, lock_flags);
@@ -965,7 +965,7 @@ reclaim:
 	xfs_ifunlock(ip);
 	xfs_iunlock(ip, XFS_ILOCK_EXCL);
 
-	XFS_STATS_INC(xs_ig_reclaims);
+	XFS_STATS_INC(ip->i_mount, xs_ig_reclaims);
 	/*
 	 * Remove the inode from the per-AG radix tree.
 	 *
diff --git a/fs/xfs/xfs_inode.c b/fs/xfs/xfs_inode.c
index dc40a6d5ae0d..a0f2bae6dc1e 100644
--- a/fs/xfs/xfs_inode.c
+++ b/fs/xfs/xfs_inode.c
@@ -3271,8 +3271,8 @@ xfs_iflush_cluster(
 	}
 
 	if (clcount) {
-		XFS_STATS_INC(xs_icluster_flushcnt);
-		XFS_STATS_ADD(xs_icluster_flushinode, clcount);
+		XFS_STATS_INC(mp, xs_icluster_flushcnt);
+		XFS_STATS_ADD(mp, xs_icluster_flushinode, clcount);
 	}
 
 out_free:
@@ -3345,7 +3345,7 @@ xfs_iflush(
 	struct xfs_dinode	*dip;
 	int			error;
 
-	XFS_STATS_INC(xs_iflush_count);
+	XFS_STATS_INC(mp, xs_iflush_count);
 
 	ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL|XFS_ILOCK_SHARED));
 	ASSERT(xfs_isiflocked(ip));
diff --git a/fs/xfs/xfs_ioctl.c b/fs/xfs/xfs_ioctl.c
index ea7d85af5310..b67a130134fb 100644
--- a/fs/xfs/xfs_ioctl.c
+++ b/fs/xfs/xfs_ioctl.c
@@ -1028,7 +1028,7 @@ xfs_ioctl_setattr_xflags(
 	xfs_diflags_to_linux(ip);
 	xfs_trans_ichgtime(tp, ip, XFS_ICHGTIME_CHG);
 	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
-	XFS_STATS_INC(xs_ig_attrchg);
+	XFS_STATS_INC(mp, xs_ig_attrchg);
 	return 0;
 }
 
diff --git a/fs/xfs/xfs_iomap.c b/fs/xfs/xfs_iomap.c
index 1f86033171c8..dca69c6711b2 100644
--- a/fs/xfs/xfs_iomap.c
+++ b/fs/xfs/xfs_iomap.c
@@ -670,7 +670,7 @@ xfs_iomap_write_allocate(
 	count_fsb = imap->br_blockcount;
 	map_start_fsb = imap->br_startoff;
 
-	XFS_STATS_ADD(xs_xstrat_bytes, XFS_FSB_TO_B(mp, count_fsb));
+	XFS_STATS_ADD(mp, xs_xstrat_bytes, XFS_FSB_TO_B(mp, count_fsb));
 
 	while (count_fsb != 0) {
 		/*
@@ -777,7 +777,7 @@ xfs_iomap_write_allocate(
 		if ((offset_fsb >= imap->br_startoff) &&
 		    (offset_fsb < (imap->br_startoff +
 				   imap->br_blockcount))) {
-			XFS_STATS_INC(xs_xstrat_quick);
+			XFS_STATS_INC(mp, xs_xstrat_quick);
 			return 0;
 		}
 
diff --git a/fs/xfs/xfs_iops.c b/fs/xfs/xfs_iops.c
index 8294132e6a3c..245268a0cdf0 100644
--- a/fs/xfs/xfs_iops.c
+++ b/fs/xfs/xfs_iops.c
@@ -695,7 +695,7 @@ xfs_setattr_nonsize(
 
 	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
 
-	XFS_STATS_INC(xs_ig_attrchg);
+	XFS_STATS_INC(mp, xs_ig_attrchg);
 
 	if (mp->m_flags & XFS_MOUNT_WSYNC)
 		xfs_trans_set_sync(tp);
@@ -922,7 +922,7 @@ xfs_setattr_size(
 
 	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
 
-	XFS_STATS_INC(xs_ig_attrchg);
+	XFS_STATS_INC(mp, xs_ig_attrchg);
 
 	if (mp->m_flags & XFS_MOUNT_WSYNC)
 		xfs_trans_set_sync(tp);
diff --git a/fs/xfs/xfs_log.c b/fs/xfs/xfs_log.c
index aaadee0969c9..401252360a62 100644
--- a/fs/xfs/xfs_log.c
+++ b/fs/xfs/xfs_log.c
@@ -268,7 +268,7 @@ xlog_grant_head_wait(
 		__set_current_state(TASK_UNINTERRUPTIBLE);
 		spin_unlock(&head->lock);
 
-		XFS_STATS_INC(xs_sleep_logspace);
+		XFS_STATS_INC(log->l_mp, xs_sleep_logspace);
 
 		trace_xfs_log_grant_sleep(log, tic);
 		schedule();
@@ -379,7 +379,7 @@ xfs_log_regrant(
 	if (XLOG_FORCED_SHUTDOWN(log))
 		return -EIO;
 
-	XFS_STATS_INC(xs_try_logspace);
+	XFS_STATS_INC(mp, xs_try_logspace);
 
 	/*
 	 * This is a new transaction on the ticket, so we need to change the
@@ -448,7 +448,7 @@ xfs_log_reserve(
 	if (XLOG_FORCED_SHUTDOWN(log))
 		return -EIO;
 
-	XFS_STATS_INC(xs_try_logspace);
+	XFS_STATS_INC(mp, xs_try_logspace);
 
 	ASSERT(*ticp == NULL);
 	tic = xlog_ticket_alloc(log, unit_bytes, cnt, client, permanent,
@@ -1768,7 +1768,7 @@ xlog_sync(
 	int		v2 = xfs_sb_version_haslogv2(&log->l_mp->m_sb);
 	int		size;
 
-	XFS_STATS_INC(xs_log_writes);
+	XFS_STATS_INC(log->l_mp, xs_log_writes);
 	ASSERT(atomic_read(&iclog->ic_refcnt) == 0);
 
 	/* Add for LR header */
@@ -1805,7 +1805,7 @@ xlog_sync(
 	bp = iclog->ic_bp;
 	XFS_BUF_SET_ADDR(bp, BLOCK_LSN(be64_to_cpu(iclog->ic_header.h_lsn)));
 
-	XFS_STATS_ADD(xs_log_blocks, BTOBB(count));
+	XFS_STATS_ADD(log->l_mp, xs_log_blocks, BTOBB(count));
 
 	/* Do we need to split this write into 2 parts? */
 	if (XFS_BUF_ADDR(bp) + BTOBB(count) > log->l_logBBsize) {
@@ -2913,7 +2913,7 @@ restart:
 
 	iclog = log->l_iclog;
 	if (iclog->ic_state != XLOG_STATE_ACTIVE) {
-		XFS_STATS_INC(xs_log_noiclogs);
+		XFS_STATS_INC(log->l_mp, xs_log_noiclogs);
 
 		/* Wait for log writes to have flushed */
 		xlog_wait(&log->l_flush_wait, &log->l_icloglock);
@@ -3212,7 +3212,7 @@ _xfs_log_force(
 	struct xlog_in_core	*iclog;
 	xfs_lsn_t		lsn;
 
-	XFS_STATS_INC(xs_log_force);
+	XFS_STATS_INC(mp, xs_log_force);
 
 	xlog_cil_force(log);
 
@@ -3297,7 +3297,7 @@ maybe_sleep:
 			spin_unlock(&log->l_icloglock);
 			return -EIO;
 		}
-		XFS_STATS_INC(xs_log_force_sleep);
+		XFS_STATS_INC(mp, xs_log_force_sleep);
 		xlog_wait(&iclog->ic_force_wait, &log->l_icloglock);
 		/*
 		 * No need to grab the log lock here since we're
@@ -3362,7 +3362,7 @@ _xfs_log_force_lsn(
 
 	ASSERT(lsn != 0);
 
-	XFS_STATS_INC(xs_log_force);
+	XFS_STATS_INC(mp, xs_log_force);
 
 	lsn = xlog_cil_force_lsn(log, lsn);
 	if (lsn == NULLCOMMITLSN)
@@ -3411,7 +3411,7 @@ try_again:
 			     (XLOG_STATE_WANT_SYNC | XLOG_STATE_SYNCING))) {
 				ASSERT(!(iclog->ic_state & XLOG_STATE_IOERROR));
 
-				XFS_STATS_INC(xs_log_force_sleep);
+				XFS_STATS_INC(mp, xs_log_force_sleep);
 
 				xlog_wait(&iclog->ic_prev->ic_write_wait,
 							&log->l_icloglock);
@@ -3441,7 +3441,7 @@ try_again:
 				spin_unlock(&log->l_icloglock);
 				return -EIO;
 			}
-			XFS_STATS_INC(xs_log_force_sleep);
+			XFS_STATS_INC(mp, xs_log_force_sleep);
 			xlog_wait(&iclog->ic_force_wait, &log->l_icloglock);
 			/*
 			 * No need to grab the log lock here since we're
diff --git a/fs/xfs/xfs_qm.c b/fs/xfs/xfs_qm.c
index eac9549efd52..7af7648c06c6 100644
--- a/fs/xfs/xfs_qm.c
+++ b/fs/xfs/xfs_qm.c
@@ -184,7 +184,7 @@ xfs_qm_dqpurge(
 	 */
 	ASSERT(!list_empty(&dqp->q_lru));
 	list_lru_del(&qi->qi_lru, &dqp->q_lru);
-	XFS_STATS_DEC(xs_qm_dquot_unused);
+	XFS_STATS_DEC(mp, xs_qm_dquot_unused);
 
 	xfs_qm_dqdestroy(dqp);
 	return 0;
@@ -448,11 +448,11 @@ xfs_qm_dquot_isolate(
 	 */
 	if (dqp->q_nrefs) {
 		xfs_dqunlock(dqp);
-		XFS_STATS_INC(xs_qm_dqwants);
+		XFS_STATS_INC(dqp->q_mount, xs_qm_dqwants);
 
 		trace_xfs_dqreclaim_want(dqp);
 		list_lru_isolate(lru, &dqp->q_lru);
-		XFS_STATS_DEC(xs_qm_dquot_unused);
+		XFS_STATS_DEC(dqp->q_mount, xs_qm_dquot_unused);
 		return LRU_REMOVED;
 	}
 
@@ -496,19 +496,19 @@ xfs_qm_dquot_isolate(
 
 	ASSERT(dqp->q_nrefs == 0);
 	list_lru_isolate_move(lru, &dqp->q_lru, &isol->dispose);
-	XFS_STATS_DEC(xs_qm_dquot_unused);
+	XFS_STATS_DEC(dqp->q_mount, xs_qm_dquot_unused);
 	trace_xfs_dqreclaim_done(dqp);
-	XFS_STATS_INC(xs_qm_dqreclaims);
+	XFS_STATS_INC(dqp->q_mount, xs_qm_dqreclaims);
 	return LRU_REMOVED;
 
 out_miss_busy:
 	trace_xfs_dqreclaim_busy(dqp);
-	XFS_STATS_INC(xs_qm_dqreclaim_misses);
+	XFS_STATS_INC(dqp->q_mount, xs_qm_dqreclaim_misses);
 	return LRU_SKIP;
 
 out_unlock_dirty:
 	trace_xfs_dqreclaim_busy(dqp);
-	XFS_STATS_INC(xs_qm_dqreclaim_misses);
+	XFS_STATS_INC(dqp->q_mount, xs_qm_dqreclaim_misses);
 	xfs_dqunlock(dqp);
 	spin_lock(lru_lock);
 	return LRU_RETRY;
diff --git a/fs/xfs/xfs_stats.h b/fs/xfs/xfs_stats.h
index 54f2260299d1..483b0eff1988 100644
--- a/fs/xfs/xfs_stats.h
+++ b/fs/xfs/xfs_stats.h
@@ -18,7 +18,6 @@
 #ifndef __XFS_STATS_H__
 #define __XFS_STATS_H__
 
-#if defined(CONFIG_PROC_FS) && !defined(XFS_STATS_OFF)
 
 #include <linux/percpu.h>
 
@@ -218,14 +217,25 @@ int xfs_stats_format(struct xfsstats __percpu *stats, char *buf);
 void xfs_stats_clearall(struct xfsstats __percpu *stats);
 extern struct xstats xfsstats;
 
-#define XFS_STATS_INC(v)	\
-	(per_cpu_ptr(xfsstats.xs_stats, current_cpu())->v++)
+#define XFS_STATS_INC(mp, v)					\
+do {								\
+	per_cpu_ptr(xfsstats.xs_stats, current_cpu())->v++;	\
+	per_cpu_ptr(mp->m_stats.xs_stats, current_cpu())->v++;	\
+} while (0)
 
-#define XFS_STATS_DEC(v)	\
-	(per_cpu_ptr(xfsstats.xs_stats, current_cpu())->v--)
+#define XFS_STATS_DEC(mp, v)					\
+do {								\
+	per_cpu_ptr(xfsstats.xs_stats, current_cpu())->v--;	\
+	per_cpu_ptr(mp->m_stats.xs_stats, current_cpu())->v--;	\
+} while (0)
 
-#define XFS_STATS_ADD(v, inc)	\
-	(per_cpu_ptr(xfsstats.xs_stats, current_cpu())->v += (inc))
+#define XFS_STATS_ADD(mp, v, inc)					\
+do {									\
+	per_cpu_ptr(xfsstats.xs_stats, current_cpu())->v += (inc);	\
+	per_cpu_ptr(mp->m_stats.xs_stats, current_cpu())->v += (inc);	\
+} while (0)
+
+#if defined(CONFIG_PROC_FS)
 
 extern int xfs_init_procfs(void);
 extern void xfs_cleanup_procfs(void);
@@ -233,10 +243,6 @@ extern void xfs_cleanup_procfs(void);
 
 #else	/* !CONFIG_PROC_FS */
 
-# define XFS_STATS_INC(count)
-# define XFS_STATS_DEC(count)
-# define XFS_STATS_ADD(count, inc)
-
 static inline int xfs_init_procfs(void)
 {
 	return 0;
diff --git a/fs/xfs/xfs_super.c b/fs/xfs/xfs_super.c
index 9d11d3ea8030..368c55adee9d 100644
--- a/fs/xfs/xfs_super.c
+++ b/fs/xfs/xfs_super.c
@@ -922,7 +922,7 @@ xfs_fs_destroy_inode(
 
 	trace_xfs_destroy_inode(ip);
 
-	XFS_STATS_INC(vn_reclaim);
+	XFS_STATS_INC(ip->i_mount, vn_reclaim);
 
 	ASSERT(XFS_FORCED_SHUTDOWN(ip->i_mount) || ip->i_delayed_blks == 0);
 
@@ -983,8 +983,8 @@ xfs_fs_evict_inode(
 
 	truncate_inode_pages_final(&inode->i_data);
 	clear_inode(inode);
-	XFS_STATS_INC(vn_rele);
-	XFS_STATS_INC(vn_remove);
+	XFS_STATS_INC(ip->i_mount, vn_rele);
+	XFS_STATS_INC(ip->i_mount, vn_remove);
 
 	xfs_inactive(ip);
 }
diff --git a/fs/xfs/xfs_trans.c b/fs/xfs/xfs_trans.c
index a0ab1dae9c31..748b16aff45a 100644
--- a/fs/xfs/xfs_trans.c
+++ b/fs/xfs/xfs_trans.c
@@ -930,9 +930,9 @@ __xfs_trans_commit(
 	 */
 	if (sync) {
 		error = _xfs_log_force_lsn(mp, commit_lsn, XFS_LOG_SYNC, NULL);
-		XFS_STATS_INC(xs_trans_sync);
+		XFS_STATS_INC(mp, xs_trans_sync);
 	} else {
-		XFS_STATS_INC(xs_trans_async);
+		XFS_STATS_INC(mp, xs_trans_async);
 	}
 
 	return error;
@@ -955,7 +955,7 @@ out_unreserve:
 	xfs_trans_free_items(tp, NULLCOMMITLSN, !!error);
 	xfs_trans_free(tp);
 
-	XFS_STATS_INC(xs_trans_empty);
+	XFS_STATS_INC(mp, xs_trans_empty);
 	return error;
 }
 
diff --git a/fs/xfs/xfs_trans_ail.c b/fs/xfs/xfs_trans_ail.c
index 1098cf490189..4f18fd92ca13 100644
--- a/fs/xfs/xfs_trans_ail.c
+++ b/fs/xfs/xfs_trans_ail.c
@@ -349,7 +349,7 @@ xfsaild_push(
 	     xfs_ail_min_lsn(ailp))) {
 		ailp->xa_log_flush = 0;
 
-		XFS_STATS_INC(xs_push_ail_flush);
+		XFS_STATS_INC(mp, xs_push_ail_flush);
 		xfs_log_force(mp, XFS_LOG_SYNC);
 	}
 
@@ -371,7 +371,7 @@ xfsaild_push(
 		goto out_done;
 	}
 
-	XFS_STATS_INC(xs_push_ail);
+	XFS_STATS_INC(mp, xs_push_ail);
 
 	lsn = lip->li_lsn;
 	while ((XFS_LSN_CMP(lip->li_lsn, target) <= 0)) {
@@ -385,7 +385,7 @@ xfsaild_push(
 		lock_result = lip->li_ops->iop_push(lip, &ailp->xa_buf_list);
 		switch (lock_result) {
 		case XFS_ITEM_SUCCESS:
-			XFS_STATS_INC(xs_push_ail_success);
+			XFS_STATS_INC(mp, xs_push_ail_success);
 			trace_xfs_ail_push(lip);
 
 			ailp->xa_last_pushed_lsn = lsn;
@@ -403,7 +403,7 @@ xfsaild_push(
 			 * re-try the flushing relatively soon if most of the
 			 * AIL is beeing flushed.
 			 */
-			XFS_STATS_INC(xs_push_ail_flushing);
+			XFS_STATS_INC(mp, xs_push_ail_flushing);
 			trace_xfs_ail_flushing(lip);
 
 			flushing++;
@@ -411,14 +411,14 @@ xfsaild_push(
 			break;
 
 		case XFS_ITEM_PINNED:
-			XFS_STATS_INC(xs_push_ail_pinned);
+			XFS_STATS_INC(mp, xs_push_ail_pinned);
 			trace_xfs_ail_pinned(lip);
 
 			stuck++;
 			ailp->xa_log_flush++;
 			break;
 		case XFS_ITEM_LOCKED:
-			XFS_STATS_INC(xs_push_ail_locked);
+			XFS_STATS_INC(mp, xs_push_ail_locked);
 			trace_xfs_ail_locked(lip);
 
 			stuck++;
-- 
cgit v1.2.3-59-g8ed1b


From 3e12dbbdbd8809f0455920e42fdbf9eddc002651 Mon Sep 17 00:00:00 2001
From: Dave Chinner <dchinner@redhat.com>
Date: Tue, 3 Nov 2015 12:27:22 +1100
Subject: xfs: fix inode size update overflow in xfs_map_direct()

Both direct IO and DAX pass an offset and count into get_blocks that
will overflow a s64 variable when an IO goes into the last supported
block in a file (i.e. at offset 2^63 - 1FSB bytes). This can be seen
from the tracing:

xfs_get_blocks_alloc: [...] offset 0x7ffffffffffff000 count 4096
xfs_gbmap_direct:     [...] offset 0x7ffffffffffff000 count 4096
xfs_gbmap_direct_none:[...] offset 0x7ffffffffffff000 count 4096

0x7ffffffffffff000 + 4096 = 0x8000000000000000, and hence that
overflows the s64 offset and we fail to detect the need for a
filesize update and an ioend is not allocated.

This is *mostly* avoided for direct IO because such extending IOs
occur with full block allocation, and so the "IS_UNWRITTEN()" check
still evaluates as true and we get an ioend that way. However, doing
single sector extending IOs to this last block will expose the fact
that file size updates will not occur after the first allocating
direct IO as the overflow will then be exposed.

There is one further complexity: the DAX page fault path also
exposes the same issue in block allocation. However, page faults
cannot extend the file size, so in this case we want to allocate the
block but do not want to allocate an ioend to enable file size
update at IO completion. Hence we now need to distinguish between
the direct IO patch allocation and dax fault path allocation to
avoid leaking ioend structures.

Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
---
 fs/xfs/xfs_aops.c | 50 ++++++++++++++++++++++++++++++++++++++++++++------
 fs/xfs/xfs_aops.h |  2 ++
 fs/xfs/xfs_file.c |  6 +++---
 3 files changed, 49 insertions(+), 9 deletions(-)

(limited to 'fs/xfs/xfs_file.c')

diff --git a/fs/xfs/xfs_aops.c b/fs/xfs/xfs_aops.c
index 50ab2879b9da..e747d6ad5d18 100644
--- a/fs/xfs/xfs_aops.c
+++ b/fs/xfs/xfs_aops.c
@@ -1250,13 +1250,28 @@ xfs_vm_releasepage(
  * 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.
+ *
+ * Note that for DIO, an IO to the highest supported file block offset (i.e.
+ * 2^63 - 1FSB bytes) will result in the offset + count overflowing a signed 64
+ * bit variable. Hence if we see this overflow, we have to assume that the IO is
+ * extending the file size. We won't know for sure until IO completion is run
+ * and the actual max write offset is communicated to the IO completion
+ * routine.
+ *
+ * For DAX page faults, we are preparing to never see unwritten extents here,
+ * nor should we ever extend the inode size. Hence we will soon have nothing to
+ * do here for this case, ensuring we don't have to provide an IO completion
+ * callback to free an ioend that we don't actually need for a fault into the
+ * page at offset (2^63 - 1FSB) bytes.
  */
+
 static void
 xfs_map_direct(
 	struct inode		*inode,
 	struct buffer_head	*bh_result,
 	struct xfs_bmbt_irec	*imap,
-	xfs_off_t		offset)
+	xfs_off_t		offset,
+	bool			dax_fault)
 {
 	struct xfs_ioend	*ioend;
 	xfs_off_t		size = bh_result->b_size;
@@ -1269,6 +1284,16 @@ xfs_map_direct(
 
 	trace_xfs_gbmap_direct(XFS_I(inode), offset, size, type, imap);
 
+	/* XXX: preparation for removing unwritten extents in DAX */
+#if 0
+	if (dax_fault) {
+		ASSERT(type == XFS_IO_OVERWRITE);
+		trace_xfs_gbmap_direct_none(XFS_I(inode), offset, size, type,
+					    imap);
+		return;
+	}
+#endif
+
 	if (bh_result->b_private) {
 		ioend = bh_result->b_private;
 		ASSERT(ioend->io_size > 0);
@@ -1283,7 +1308,8 @@ xfs_map_direct(
 					      ioend->io_size, ioend->io_type,
 					      imap);
 	} else if (type == XFS_IO_UNWRITTEN ||
-		   offset + size > i_size_read(inode)) {
+		   offset + size > i_size_read(inode) ||
+		   offset + size < 0) {
 		ioend = xfs_alloc_ioend(inode, type);
 		ioend->io_offset = offset;
 		ioend->io_size = size;
@@ -1345,7 +1371,8 @@ __xfs_get_blocks(
 	sector_t		iblock,
 	struct buffer_head	*bh_result,
 	int			create,
-	bool			direct)
+	bool			direct,
+	bool			dax_fault)
 {
 	struct xfs_inode	*ip = XFS_I(inode);
 	struct xfs_mount	*mp = ip->i_mount;
@@ -1458,7 +1485,8 @@ __xfs_get_blocks(
 			set_buffer_unwritten(bh_result);
 		/* direct IO needs special help */
 		if (create && direct)
-			xfs_map_direct(inode, bh_result, &imap, offset);
+			xfs_map_direct(inode, bh_result, &imap, offset,
+				       dax_fault);
 	}
 
 	/*
@@ -1505,7 +1533,7 @@ xfs_get_blocks(
 	struct buffer_head	*bh_result,
 	int			create)
 {
-	return __xfs_get_blocks(inode, iblock, bh_result, create, false);
+	return __xfs_get_blocks(inode, iblock, bh_result, create, false, false);
 }
 
 int
@@ -1515,7 +1543,17 @@ xfs_get_blocks_direct(
 	struct buffer_head	*bh_result,
 	int			create)
 {
-	return __xfs_get_blocks(inode, iblock, bh_result, create, true);
+	return __xfs_get_blocks(inode, iblock, bh_result, create, true, false);
+}
+
+int
+xfs_get_blocks_dax_fault(
+	struct inode		*inode,
+	sector_t		iblock,
+	struct buffer_head	*bh_result,
+	int			create)
+{
+	return __xfs_get_blocks(inode, iblock, bh_result, create, true, true);
 }
 
 static void
diff --git a/fs/xfs/xfs_aops.h b/fs/xfs/xfs_aops.h
index 86afd1ac7895..d39ba25ccc98 100644
--- a/fs/xfs/xfs_aops.h
+++ b/fs/xfs/xfs_aops.h
@@ -58,6 +58,8 @@ int	xfs_get_blocks(struct inode *inode, sector_t offset,
 		       struct buffer_head *map_bh, int create);
 int	xfs_get_blocks_direct(struct inode *inode, sector_t offset,
 			      struct buffer_head *map_bh, int create);
+int	xfs_get_blocks_dax_fault(struct inode *inode, sector_t offset,
+			         struct buffer_head *map_bh, int create);
 void	xfs_end_io_dax_write(struct buffer_head *bh, int uptodate);
 
 extern void xfs_count_page_state(struct page *, int *, int *);
diff --git a/fs/xfs/xfs_file.c b/fs/xfs/xfs_file.c
index e78feb400e22..27abe1c92184 100644
--- a/fs/xfs/xfs_file.c
+++ b/fs/xfs/xfs_file.c
@@ -1503,7 +1503,7 @@ xfs_filemap_page_mkwrite(
 	xfs_ilock(XFS_I(inode), XFS_MMAPLOCK_SHARED);
 
 	if (IS_DAX(inode)) {
-		ret = __dax_mkwrite(vma, vmf, xfs_get_blocks_direct,
+		ret = __dax_mkwrite(vma, vmf, xfs_get_blocks_dax_fault,
 				    xfs_end_io_dax_write);
 	} else {
 		ret = __block_page_mkwrite(vma, vmf, xfs_get_blocks);
@@ -1538,7 +1538,7 @@ xfs_filemap_fault(
 		 * changes to xfs_get_blocks_direct() to map unwritten extent
 		 * ioend for conversion on read-only mappings.
 		 */
-		ret = __dax_fault(vma, vmf, xfs_get_blocks_direct, NULL);
+		ret = __dax_fault(vma, vmf, xfs_get_blocks_dax_fault, NULL);
 	} else
 		ret = filemap_fault(vma, vmf);
 	xfs_iunlock(XFS_I(inode), XFS_MMAPLOCK_SHARED);
@@ -1565,7 +1565,7 @@ xfs_filemap_pmd_fault(
 	sb_start_pagefault(inode->i_sb);
 	file_update_time(vma->vm_file);
 	xfs_ilock(XFS_I(inode), XFS_MMAPLOCK_SHARED);
-	ret = __dax_pmd_fault(vma, addr, pmd, flags, xfs_get_blocks_direct,
+	ret = __dax_pmd_fault(vma, addr, pmd, flags, xfs_get_blocks_dax_fault,
 				    xfs_end_io_dax_write);
 	xfs_iunlock(XFS_I(inode), XFS_MMAPLOCK_SHARED);
 	sb_end_pagefault(inode->i_sb);
-- 
cgit v1.2.3-59-g8ed1b


From 01a155e6cf7db1a8ff2aa73162d7d9ec05ad298f Mon Sep 17 00:00:00 2001
From: Dave Chinner <dchinner@redhat.com>
Date: Tue, 3 Nov 2015 12:37:02 +1100
Subject: xfs: DAX does not use IO completion callbacks

For DAX, we are now doing block zeroing during allocation. This
means we no longer need a special DAX fault IO completion callback
to do unwritten extent conversion. Because mmap never extends the
file size (it SEGVs the process) we don't need a callback to update
the file size, either. Hence we can remove the completion callbacks
from the __dax_fault and __dax_mkwrite calls.

Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
---
 fs/xfs/xfs_aops.c | 39 ---------------------------------------
 fs/xfs/xfs_aops.h |  1 -
 fs/xfs/xfs_file.c |  5 ++---
 3 files changed, 2 insertions(+), 43 deletions(-)

(limited to 'fs/xfs/xfs_file.c')

diff --git a/fs/xfs/xfs_aops.c b/fs/xfs/xfs_aops.c
index df3dabd469b9..69c2dbc20836 100644
--- a/fs/xfs/xfs_aops.c
+++ b/fs/xfs/xfs_aops.c
@@ -1657,45 +1657,6 @@ xfs_end_io_direct_write(
 	__xfs_end_io_direct_write(inode, ioend, offset, size);
 }
 
-/*
- * For DAX we need a mapping buffer callback for unwritten extent conversion
- * when page faults allocate blocks and then zero them. Note that in this
- * case the mapping indicated by the ioend may extend beyond EOF. We most
- * definitely do not want to extend EOF here, so we trim back the ioend size to
- * EOF.
- */
-#ifdef CONFIG_FS_DAX
-void
-xfs_end_io_dax_write(
-	struct buffer_head	*bh,
-	int			uptodate)
-{
-	struct xfs_ioend	*ioend = bh->b_private;
-	struct inode		*inode = ioend->io_inode;
-	ssize_t			size = ioend->io_size;
-
-	ASSERT(IS_DAX(ioend->io_inode));
-
-	/* if there was an error zeroing, then don't convert it */
-	if (!uptodate)
-		ioend->io_error = -EIO;
-
-	/*
-	 * Trim update to EOF, so we don't extend EOF during unwritten extent
-	 * conversion of partial EOF blocks.
-	 */
-	spin_lock(&XFS_I(inode)->i_flags_lock);
-	if (ioend->io_offset + size > i_size_read(inode))
-		size = i_size_read(inode) - ioend->io_offset;
-	spin_unlock(&XFS_I(inode)->i_flags_lock);
-
-	__xfs_end_io_direct_write(inode, ioend, ioend->io_offset, size);
-
-}
-#else
-void xfs_end_io_dax_write(struct buffer_head *bh, int uptodate) { }
-#endif
-
 static inline ssize_t
 xfs_vm_do_dio(
 	struct inode		*inode,
diff --git a/fs/xfs/xfs_aops.h b/fs/xfs/xfs_aops.h
index d39ba25ccc98..f6ffc9ae5ceb 100644
--- a/fs/xfs/xfs_aops.h
+++ b/fs/xfs/xfs_aops.h
@@ -60,7 +60,6 @@ int	xfs_get_blocks_direct(struct inode *inode, sector_t offset,
 			      struct buffer_head *map_bh, int create);
 int	xfs_get_blocks_dax_fault(struct inode *inode, sector_t offset,
 			         struct buffer_head *map_bh, int create);
-void	xfs_end_io_dax_write(struct buffer_head *bh, int uptodate);
 
 extern void xfs_count_page_state(struct page *, int *, int *);
 
diff --git a/fs/xfs/xfs_file.c b/fs/xfs/xfs_file.c
index 27abe1c92184..9c8eef7c57b4 100644
--- a/fs/xfs/xfs_file.c
+++ b/fs/xfs/xfs_file.c
@@ -1503,8 +1503,7 @@ xfs_filemap_page_mkwrite(
 	xfs_ilock(XFS_I(inode), XFS_MMAPLOCK_SHARED);
 
 	if (IS_DAX(inode)) {
-		ret = __dax_mkwrite(vma, vmf, xfs_get_blocks_dax_fault,
-				    xfs_end_io_dax_write);
+		ret = __dax_mkwrite(vma, vmf, xfs_get_blocks_dax_fault, NULL);
 	} else {
 		ret = __block_page_mkwrite(vma, vmf, xfs_get_blocks);
 		ret = block_page_mkwrite_return(ret);
@@ -1566,7 +1565,7 @@ xfs_filemap_pmd_fault(
 	file_update_time(vma->vm_file);
 	xfs_ilock(XFS_I(inode), XFS_MMAPLOCK_SHARED);
 	ret = __dax_pmd_fault(vma, addr, pmd, flags, xfs_get_blocks_dax_fault,
-				    xfs_end_io_dax_write);
+			      NULL);
 	xfs_iunlock(XFS_I(inode), XFS_MMAPLOCK_SHARED);
 	sb_end_pagefault(inode->i_sb);
 
-- 
cgit v1.2.3-59-g8ed1b


From 3af49285854df66260a263198cc15abb07b95287 Mon Sep 17 00:00:00 2001
From: Dave Chinner <dchinner@redhat.com>
Date: Tue, 3 Nov 2015 12:37:02 +1100
Subject: xfs: add ->pfn_mkwrite support for DAX

->pfn_mkwrite support is needed so that when a page with allocated
backing store takes a write fault we can check that the fault has
not raced with a truncate and is pointing to a region beyond the
current end of file.

This also allows us to update the timestamp on the inode, too, which
fixes a generic/080 failure.

Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
---
 fs/xfs/xfs_file.c  | 35 +++++++++++++++++++++++++++++++++++
 fs/xfs/xfs_trace.h |  1 +
 2 files changed, 36 insertions(+)

(limited to 'fs/xfs/xfs_file.c')

diff --git a/fs/xfs/xfs_file.c b/fs/xfs/xfs_file.c
index 9c8eef7c57b4..f429662d7d42 100644
--- a/fs/xfs/xfs_file.c
+++ b/fs/xfs/xfs_file.c
@@ -1572,11 +1572,46 @@ xfs_filemap_pmd_fault(
 	return ret;
 }
 
+/*
+ * pfn_mkwrite was originally inteneded to ensure we capture time stamp
+ * updates on write faults. In reality, it's need to serialise against
+ * truncate similar to page_mkwrite. Hence we open-code dax_pfn_mkwrite()
+ * here and cycle the XFS_MMAPLOCK_SHARED to ensure we serialise the fault
+ * barrier in place.
+ */
+static int
+xfs_filemap_pfn_mkwrite(
+	struct vm_area_struct	*vma,
+	struct vm_fault		*vmf)
+{
+
+	struct inode		*inode = file_inode(vma->vm_file);
+	struct xfs_inode	*ip = XFS_I(inode);
+	int			ret = VM_FAULT_NOPAGE;
+	loff_t			size;
+
+	trace_xfs_filemap_pfn_mkwrite(ip);
+
+	sb_start_pagefault(inode->i_sb);
+	file_update_time(vma->vm_file);
+
+	/* check if the faulting page hasn't raced with truncate */
+	xfs_ilock(ip, XFS_MMAPLOCK_SHARED);
+	size = (i_size_read(inode) + PAGE_SIZE - 1) >> PAGE_SHIFT;
+	if (vmf->pgoff >= size)
+		ret = VM_FAULT_SIGBUS;
+	xfs_iunlock(ip, XFS_MMAPLOCK_SHARED);
+	sb_end_pagefault(inode->i_sb);
+	return ret;
+
+}
+
 static const struct vm_operations_struct xfs_file_vm_ops = {
 	.fault		= xfs_filemap_fault,
 	.pmd_fault	= xfs_filemap_pmd_fault,
 	.map_pages	= filemap_map_pages,
 	.page_mkwrite	= xfs_filemap_page_mkwrite,
+	.pfn_mkwrite	= xfs_filemap_pfn_mkwrite,
 };
 
 STATIC int
diff --git a/fs/xfs/xfs_trace.h b/fs/xfs/xfs_trace.h
index 5ed36b1e04c1..c53beda675d6 100644
--- a/fs/xfs/xfs_trace.h
+++ b/fs/xfs/xfs_trace.h
@@ -689,6 +689,7 @@ DEFINE_INODE_EVENT(xfs_inode_free_eofblocks_invalid);
 DEFINE_INODE_EVENT(xfs_filemap_fault);
 DEFINE_INODE_EVENT(xfs_filemap_pmd_fault);
 DEFINE_INODE_EVENT(xfs_filemap_page_mkwrite);
+DEFINE_INODE_EVENT(xfs_filemap_pfn_mkwrite);
 
 DECLARE_EVENT_CLASS(xfs_iref_class,
 	TP_PROTO(struct xfs_inode *ip, unsigned long caller_ip),
-- 
cgit v1.2.3-59-g8ed1b


From 13ad4fe3e087ab66a140f1e00d98f28aa4e3bb28 Mon Sep 17 00:00:00 2001
From: Dave Chinner <dchinner@redhat.com>
Date: Tue, 3 Nov 2015 12:37:02 +1100
Subject: xfs: xfs_filemap_pmd_fault treats read faults as write faults

The code initially committed didn't have the same checks for write
faults as the dax_pmd_fault code and hence treats all faults as
write faults. We can get read faults through this path because they
is no pmd_mkwrite path for write faults similar to the normal page
fault path. Hence we need to ensure that we only do c/mtime updates
on write faults, and freeze protection is unnecessary for read
faults.

Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
---
 fs/xfs/xfs_file.c | 20 ++++++++++++++++----
 1 file changed, 16 insertions(+), 4 deletions(-)

(limited to 'fs/xfs/xfs_file.c')

diff --git a/fs/xfs/xfs_file.c b/fs/xfs/xfs_file.c
index f429662d7d42..ce208e3896aa 100644
--- a/fs/xfs/xfs_file.c
+++ b/fs/xfs/xfs_file.c
@@ -1477,7 +1477,7 @@ xfs_file_llseek(
  *
  * mmap_sem (MM)
  *   sb_start_pagefault(vfs, freeze)
- *     i_mmap_lock (XFS - truncate serialisation)
+ *     i_mmaplock (XFS - truncate serialisation)
  *       page_lock (MM)
  *         i_lock (XFS - extent map serialisation)
  */
@@ -1545,6 +1545,13 @@ xfs_filemap_fault(
 	return ret;
 }
 
+/*
+ * Similar to xfs_filemap_fault(), the DAX fault path can call into here on
+ * both read and write faults. Hence we need to handle both cases. There is no
+ * ->pmd_mkwrite callout for huge pages, so we have a single function here to
+ * handle both cases here. @flags carries the information on the type of fault
+ * occuring.
+ */
 STATIC int
 xfs_filemap_pmd_fault(
 	struct vm_area_struct	*vma,
@@ -1561,13 +1568,18 @@ xfs_filemap_pmd_fault(
 
 	trace_xfs_filemap_pmd_fault(ip);
 
-	sb_start_pagefault(inode->i_sb);
-	file_update_time(vma->vm_file);
+	if (flags & FAULT_FLAG_WRITE) {
+		sb_start_pagefault(inode->i_sb);
+		file_update_time(vma->vm_file);
+	}
+
 	xfs_ilock(XFS_I(inode), XFS_MMAPLOCK_SHARED);
 	ret = __dax_pmd_fault(vma, addr, pmd, flags, xfs_get_blocks_dax_fault,
 			      NULL);
 	xfs_iunlock(XFS_I(inode), XFS_MMAPLOCK_SHARED);
-	sb_end_pagefault(inode->i_sb);
+
+	if (flags & FAULT_FLAG_WRITE)
+		sb_end_pagefault(inode->i_sb);
 
 	return ret;
 }
-- 
cgit v1.2.3-59-g8ed1b


From fc0561cefc04e7803c0f6501ca4f310a502f65b8 Mon Sep 17 00:00:00 2001
From: Dave Chinner <dchinner@redhat.com>
Date: Tue, 3 Nov 2015 13:14:59 +1100
Subject: xfs: optimise away log forces on timestamp updates for fdatasync

xfs: timestamp updates cause excessive fdatasync log traffic

Sage Weil reported that a ceph test workload was writing to the
log on every fdatasync during an overwrite workload. Event tracing
showed that the only metadata modification being made was the
timestamp updates during the write(2) syscall, but fdatasync(2)
is supposed to ignore them. The key observation was that the
transactions in the log all looked like this:

INODE: #regs: 4   ino: 0x8b  flags: 0x45   dsize: 32

And contained a flags field of 0x45 or 0x85, and had data and
attribute forks following the inode core. This means that the
timestamp updates were triggering dirty relogging of previously
logged parts of the inode that hadn't yet been flushed back to
disk.

There are two parts to this problem. The first is that XFS relogs
dirty regions in subsequent transactions, so it carries around the
fields that have been dirtied since the last time the inode was
written back to disk, not since the last time the inode was forced
into the log.

The second part is that on v5 filesystems, the inode change count
update during inode dirtying also sets the XFS_ILOG_CORE flag, so
on v5 filesystems this makes a timestamp update dirty the entire
inode.

As a result when fdatasync is run, it looks at the dirty fields in
the inode, and sees more than just the timestamp flag, even though
the only metadata change since the last fdatasync was just the
timestamps. Hence we force the log on every subsequent fdatasync
even though it is not needed.

To fix this, add a new field to the inode log item that tracks
changes since the last time fsync/fdatasync forced the log to flush
the changes to the journal. This flag is updated when we dirty the
inode, but we do it before updating the change count so it does not
carry the "core dirty" flag from timestamp updates. The fields are
zeroed when the inode is marked clean (due to writeback/freeing) or
when an fsync/datasync forces the log. Hence if we only dirty the
timestamps on the inode between fsync/fdatasync calls, the fdatasync
will not trigger another log force.

Over 100 runs of the test program:

Ext4 baseline:
	runtime: 1.63s +/- 0.24s
	avg lat: 1.59ms +/- 0.24ms
	iops: ~2000

XFS, vanilla kernel:
        runtime: 2.45s +/- 0.18s
	avg lat: 2.39ms +/- 0.18ms
	log forces: ~400/s
	iops: ~1000

XFS, patched kernel:
        runtime: 1.49s +/- 0.26s
	avg lat: 1.46ms +/- 0.25ms
	log forces: ~30/s
	iops: ~1500

Reported-by: Sage Weil <sage@redhat.com>
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
---
 fs/xfs/xfs_file.c        | 21 ++++++++++++++++-----
 fs/xfs/xfs_inode.c       |  2 ++
 fs/xfs/xfs_inode_item.c  |  1 +
 fs/xfs/xfs_inode_item.h  |  1 +
 fs/xfs/xfs_trans_inode.c |  9 +++++++++
 5 files changed, 29 insertions(+), 5 deletions(-)

(limited to 'fs/xfs/xfs_file.c')

diff --git a/fs/xfs/xfs_file.c b/fs/xfs/xfs_file.c
index e78feb400e22..c94699cbc667 100644
--- a/fs/xfs/xfs_file.c
+++ b/fs/xfs/xfs_file.c
@@ -242,19 +242,30 @@ xfs_file_fsync(
 	}
 
 	/*
-	 * All metadata updates are logged, which means that we just have
-	 * to flush the log up to the latest LSN that touched the inode.
+	 * All metadata updates are logged, which means that we just have to
+	 * flush the log up to the latest LSN that touched the inode. If we have
+	 * concurrent fsync/fdatasync() calls, we need them to all block on the
+	 * log force before we clear the ili_fsync_fields field. This ensures
+	 * that we don't get a racing sync operation that does not wait for the
+	 * metadata to hit the journal before returning. If we race with
+	 * clearing the ili_fsync_fields, then all that will happen is the log
+	 * force will do nothing as the lsn will already be on disk. We can't
+	 * race with setting ili_fsync_fields because that is done under
+	 * XFS_ILOCK_EXCL, and that can't happen because we hold the lock shared
+	 * until after the ili_fsync_fields is cleared.
 	 */
 	xfs_ilock(ip, XFS_ILOCK_SHARED);
 	if (xfs_ipincount(ip)) {
 		if (!datasync ||
-		    (ip->i_itemp->ili_fields & ~XFS_ILOG_TIMESTAMP))
+		    (ip->i_itemp->ili_fsync_fields & ~XFS_ILOG_TIMESTAMP))
 			lsn = ip->i_itemp->ili_last_lsn;
 	}
-	xfs_iunlock(ip, XFS_ILOCK_SHARED);
 
-	if (lsn)
+	if (lsn) {
 		error = _xfs_log_force_lsn(mp, lsn, XFS_LOG_SYNC, &log_flushed);
+		ip->i_itemp->ili_fsync_fields = 0;
+	}
+	xfs_iunlock(ip, XFS_ILOCK_SHARED);
 
 	/*
 	 * If we only have a single device, and the log force about was
diff --git a/fs/xfs/xfs_inode.c b/fs/xfs/xfs_inode.c
index dc40a6d5ae0d..ff629d544706 100644
--- a/fs/xfs/xfs_inode.c
+++ b/fs/xfs/xfs_inode.c
@@ -2365,6 +2365,7 @@ retry:
 
 			iip->ili_last_fields = iip->ili_fields;
 			iip->ili_fields = 0;
+			iip->ili_fsync_fields = 0;
 			iip->ili_logged = 1;
 			xfs_trans_ail_copy_lsn(mp->m_ail, &iip->ili_flush_lsn,
 						&iip->ili_item.li_lsn);
@@ -3560,6 +3561,7 @@ xfs_iflush_int(
 	 */
 	iip->ili_last_fields = iip->ili_fields;
 	iip->ili_fields = 0;
+	iip->ili_fsync_fields = 0;
 	iip->ili_logged = 1;
 
 	xfs_trans_ail_copy_lsn(mp->m_ail, &iip->ili_flush_lsn,
diff --git a/fs/xfs/xfs_inode_item.c b/fs/xfs/xfs_inode_item.c
index 62bd80f4edd9..d14b12b8cfef 100644
--- a/fs/xfs/xfs_inode_item.c
+++ b/fs/xfs/xfs_inode_item.c
@@ -719,6 +719,7 @@ xfs_iflush_abort(
 		 * attempted.
 		 */
 		iip->ili_fields = 0;
+		iip->ili_fsync_fields = 0;
 	}
 	/*
 	 * Release the inode's flush lock since we're done with it.
diff --git a/fs/xfs/xfs_inode_item.h b/fs/xfs/xfs_inode_item.h
index 488d81254e28..4c7722e325b3 100644
--- a/fs/xfs/xfs_inode_item.h
+++ b/fs/xfs/xfs_inode_item.h
@@ -34,6 +34,7 @@ typedef struct xfs_inode_log_item {
 	unsigned short		ili_logged;	   /* flushed logged data */
 	unsigned int		ili_last_fields;   /* fields when flushed */
 	unsigned int		ili_fields;	   /* fields to be logged */
+	unsigned int		ili_fsync_fields;  /* logged since last fsync */
 } xfs_inode_log_item_t;
 
 static inline int xfs_inode_clean(xfs_inode_t *ip)
diff --git a/fs/xfs/xfs_trans_inode.c b/fs/xfs/xfs_trans_inode.c
index 17280cd71934..b97f1df910ab 100644
--- a/fs/xfs/xfs_trans_inode.c
+++ b/fs/xfs/xfs_trans_inode.c
@@ -107,6 +107,15 @@ xfs_trans_log_inode(
 	ASSERT(ip->i_itemp != NULL);
 	ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL));
 
+	/*
+	 * Record the specific change for fdatasync optimisation. This
+	 * allows fdatasync to skip log forces for inodes that are only
+	 * timestamp dirty. We do this before the change count so that
+	 * the core being logged in this case does not impact on fdatasync
+	 * behaviour.
+	 */
+	ip->i_itemp->ili_fsync_fields |= flags;
+
 	/*
 	 * First time we log the inode in a transaction, bump the inode change
 	 * counter if it is configured for this to occur. We don't use
-- 
cgit v1.2.3-59-g8ed1b