[PATCH 4/5] xfs: fix splice_write locking

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



Call splice_from_pipe directly from XFS so that we can do our normal
I/O path locking.  This fixes a rare to hit deadlock vs direct I/O
as reported by Dave Chinner long time ago.

Signed-off-by: Christoph Hellwig <hch@xxxxxx>
---
 fs/xfs/xfs_file.c |   26 +++++++++++++-------------
 1 file changed, 13 insertions(+), 13 deletions(-)

diff --git a/fs/xfs/xfs_file.c b/fs/xfs/xfs_file.c
index 52c91e1..9d0da98 100644
--- a/fs/xfs/xfs_file.c
+++ b/fs/xfs/xfs_file.c
@@ -43,6 +43,7 @@
 #include <linux/dcache.h>
 #include <linux/falloc.h>
 #include <linux/pagevec.h>
+#include <linux/splice.h>
 
 static const struct vm_operations_struct xfs_file_vm_ops;
 
@@ -348,14 +349,6 @@ xfs_file_splice_read(
 	return ret;
 }
 
-/*
- * xfs_file_splice_write() does not use xfs_rw_ilock() because
- * generic_file_splice_write() takes the i_mutex itself. This, in theory,
- * couuld cause lock inversions between the aio_write path and the splice path
- * if someone is doing concurrent splice(2) based writes and write(2) based
- * writes to the same inode. The only real way to fix this is to re-implement
- * the generic code here with correct locking orders.
- */
 STATIC ssize_t
 xfs_file_splice_write(
 	struct pipe_inode_info	*pipe,
@@ -377,15 +370,22 @@ xfs_file_splice_write(
 	if (XFS_FORCED_SHUTDOWN(ip->i_mount))
 		return -EIO;
 
-	xfs_ilock(ip, XFS_IOLOCK_EXCL);
-
+	xfs_rw_ilock(ip, XFS_IOLOCK_EXCL);
 	trace_xfs_file_splice_write(ip, count, *ppos, ioflags);
+	ret = splice_from_pipe(pipe, outfilp, ppos, count, flags, pipe_to_file);
+	xfs_rw_iunlock(ip, XFS_IOLOCK_EXCL);
+
+	if (ret > 0) {
+		int err;
 
-	ret = generic_file_splice_write(pipe, outfilp, ppos, count, flags);
-	if (ret > 0)
 		XFS_STATS_ADD(xs_write_bytes, ret);
 
-	xfs_iunlock(ip, XFS_IOLOCK_EXCL);
+		err = generic_write_sync(outfilp, *ppos, ret);
+		if (err)
+			return err;
+		*ppos += ret;
+	}
+
 	return ret;
 }
 
-- 
1.7.10.4


_______________________________________________
xfs mailing list
xfs@xxxxxxxxxxx
http://oss.sgi.com/mailman/listinfo/xfs




[Index of Archives]     [Linux XFS Devel]     [Linux Filesystem Development]     [Filesystem Testing]     [Linux USB Devel]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]

  Powered by Linux