[PATCH 47/58] xfs: handle directio copy-on-write for reflinked blocks

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

 



We hope that CoW writes will be rare and that directio CoW writes will
be even more rare.  Therefore, fall-back any such write to the
buffered path.

Signed-off-by: Darrick J. Wong <darrick.wong@xxxxxxxxxx>
---
 fs/xfs/xfs_aops.c    |    6 ++++++
 fs/xfs/xfs_file.c    |   12 ++++++++++--
 fs/xfs/xfs_reflink.c |   34 ++++++++++++++++++++++++++++++++++
 3 files changed, 50 insertions(+), 2 deletions(-)


diff --git a/fs/xfs/xfs_aops.c b/fs/xfs/xfs_aops.c
index 06a1d2f..4ab24fa 100644
--- a/fs/xfs/xfs_aops.c
+++ b/fs/xfs/xfs_aops.c
@@ -1491,6 +1491,12 @@ __xfs_get_blocks(
 	if (imap.br_startblock != HOLESTARTBLOCK &&
 	    imap.br_startblock != DELAYSTARTBLOCK &&
 	    (create || !ISUNWRITTEN(&imap))) {
+		if (create && direct) {
+			error = xfs_reflink_redirect_directio_write(ip, &imap,
+					offset);
+			if (error)
+				return error;
+		}
 		xfs_map_buffer(inode, bh_result, &imap, offset);
 		if (ISUNWRITTEN(&imap))
 			set_buffer_unwritten(bh_result);
diff --git a/fs/xfs/xfs_file.c b/fs/xfs/xfs_file.c
index 593223f..fc5b9ea 100644
--- a/fs/xfs/xfs_file.c
+++ b/fs/xfs/xfs_file.c
@@ -876,10 +876,18 @@ xfs_file_write_iter(
 	if (XFS_FORCED_SHUTDOWN(ip->i_mount))
 		return -EIO;
 
-	if ((iocb->ki_flags & IOCB_DIRECT) || IS_DAX(inode))
+	/*
+	 * Allow DIO to fall back to buffered *only* in the case that we're
+	 * doing a reflink CoW.
+	 */
+	if ((iocb->ki_flags & IOCB_DIRECT) || IS_DAX(inode)) {
 		ret = xfs_file_dio_aio_write(iocb, from);
-	else
+		if (ret == -EREMCHG)
+			goto buffered;
+	} else {
+buffered:
 		ret = xfs_file_buffered_aio_write(iocb, from);
+	}
 
 	if (ret > 0) {
 		ssize_t err;
diff --git a/fs/xfs/xfs_reflink.c b/fs/xfs/xfs_reflink.c
index 1e00be2..226e23f 100644
--- a/fs/xfs/xfs_reflink.c
+++ b/fs/xfs/xfs_reflink.c
@@ -407,6 +407,40 @@ advloop:
 }
 
 /**
+ * xfs_reflink_redirect_directio_write() - bounce a directio write to a
+ *					   reflinked region down to buffered
+ *					   write mode.
+ *
+ * @ip: XFS inode object
+ * @imap: the fileoff:fsblock mapping that we might fork
+ * @offset: the file byte offset of the block we're examining
+ */
+int
+xfs_reflink_redirect_directio_write(
+	struct xfs_inode	*ip,
+	struct xfs_bmbt_irec	*imap,
+	xfs_off_t		offset)
+{
+	bool			type = false;
+	int			error;
+
+	error = xfs_reflink_should_fork_block(ip, imap, offset, &type);
+	if (error)
+		return error;
+	if (!type)
+		return 0;
+
+	/*
+	 * Are we doing a DIO write to a reflinked block?  In the ideal world
+	 * we at least would fork full blocks, but for now just fall back to
+	 * buffered mode.  Yuck.  Use -EREMCHG ("remote address changed") to
+	 * signal this, since in general XFS doesn't do this sort of fallback.
+	 */
+	trace_xfs_reflink_bounce_direct_write(ip, imap);
+	return -EREMCHG;
+}
+
+/**
  * xfs_reflink_write_fork_block() -- find a remapping object and redirect the
  *				     write.
  *

--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html



[Index of Archives]     [Linux Ext4 Filesystem]     [Union Filesystem]     [Filesystem Testing]     [Ceph Users]     [Ecryptfs]     [AutoFS]     [Kernel Newbies]     [Share Photos]     [Security]     [Netfilter]     [Bugtraq]     [Yosemite News]     [MIPS Linux]     [ARM Linux]     [Linux Security]     [Linux Cachefs]     [Reiser Filesystem]     [Linux RAID]     [Samba]     [Device Mapper]     [CEPH Development]
  Powered by Linux