[PATCH, RFC] xfs: fix multi-AG deadlock in xfs_bunmapi

* [PATCH, RFC] xfs: fix multi-AG deadlock in xfs_bunmapi
@ 2017-07-18 17:25 Christoph Hellwig
  2017-07-18 18:15 ` Darrick J. Wong
  2017-07-19 13:11 ` Nikolay Borisov
  0 siblings, 2 replies; 17+ messages in thread
From: Christoph Hellwig @ 2017-07-18 17:25 UTC (permalink / raw)
  To: n.borisov.lkml, linux-xfs

Just like in the allocator we must avoid touching multiple AGs out of
order when freeing blocks, as freeing still locks the AGF and can cause
the same AB-BA deadlocks as in the allocation path.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reported-by: Nikolay Borisov <n.borisov.lkml@gmail.com>
---
 fs/xfs/libxfs/xfs_bmap.c | 12 ++++++++++++
 1 file changed, 12 insertions(+)

diff --git a/fs/xfs/libxfs/xfs_bmap.c b/fs/xfs/libxfs/xfs_bmap.c
index 0a9880777c9c..935adde72a8b 100644
--- a/fs/xfs/libxfs/xfs_bmap.c
+++ b/fs/xfs/libxfs/xfs_bmap.c
@@ -5435,6 +5435,7 @@ __xfs_bunmapi(
 	xfs_fsblock_t		sum;
 	xfs_filblks_t		len = *rlen;	/* length to unmap in file */
 	xfs_fileoff_t		max_len;
+	xfs_agnumber_t		prev_agno = NULLAGNUMBER, agno;
 
 	trace_xfs_bunmap(ip, bno, len, flags, _RET_IP_);
 
@@ -5534,6 +5535,17 @@ __xfs_bunmapi(
 		 */
 		del = got;
 		wasdel = isnullstartblock(del.br_startblock);
+
+		/*
+		 * Make sure we don't touch multiple AGF headers out of order
+		 * in a single transaction, as that could cause AB-BA deadlocks.
+		 */
+		if (!wasdel) {
+			agno = XFS_FSB_TO_AGNO(mp, del.br_startblock);
+			if (prev_agno != NULLAGNUMBER && prev_agno > agno)
+				break;
+			prev_agno = agno;
+		}
 		if (got.br_startoff < start) {
 			del.br_startoff = start;
 			del.br_blockcount -= start - got.br_startoff;
-- 
2.11.0


^ permalink raw reply related	[flat|nested] 17+ messages in thread