[XFS] Map unwritten extents correctly for I/o completion processing
authorDavid Chinner <dgc@sgi.com>
Mon, 18 Jun 2007 06:49:58 +0000 (16:49 +1000)
committerTim Shimmin <tes@chook.melbourne.sgi.com>
Sat, 14 Jul 2007 05:32:49 +0000 (15:32 +1000)
If we have multiple unwritten extents within a single page, we fail to
tell the I/o completion construction handlers we need a new handle for the
second and subsequent blocks in the page. While we still issue the I/O
correctly, we do not have the correct ranges recorded in the ioend
structures and hence when we go to convert the unwritten extents we screw
it up.

Make sure we start a new ioend every time the mapping changes so that we
convert the correct ranges on I/O completion.

SGI-PV: 964647
SGI-Modid: xfs-linux-melb:xfs-kern:28797a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Tim Shimmin <tes@sgi.com>
fs/xfs/linux-2.6/xfs_aops.c

index c097e4e69768df096e751ee62ea26b61fe2a97eb..fd4105d662e036b11821f15fb7780a91f0929b00 100644 (file)
@@ -1010,6 +1010,8 @@ xfs_page_state_convert(
                if (buffer_unwritten(bh) || buffer_delay(bh) ||
                    ((buffer_uptodate(bh) || PageUptodate(page)) &&
                     !buffer_mapped(bh) && (unmapped || startio))) {
+                       int new_ioend = 0;
+
                        /*
                         * Make sure we don't use a read-only iomap
                         */
@@ -1028,6 +1030,15 @@ xfs_page_state_convert(
                        }
 
                        if (!iomap_valid) {
+                               /*
+                                * if we didn't have a valid mapping then we
+                                * need to ensure that we put the new mapping
+                                * in a new ioend structure. This needs to be
+                                * done to ensure that the ioends correctly
+                                * reflect the block mappings at io completion
+                                * for unwritten extent conversion.
+                                */
+                               new_ioend = 1;
                                if (type == IOMAP_NEW) {
                                        size = xfs_probe_cluster(inode,
                                                        page, bh, head, 0);
@@ -1047,7 +1058,7 @@ xfs_page_state_convert(
                                if (startio) {
                                        xfs_add_to_ioend(inode, bh, offset,
                                                        type, &ioend,
-                                                       !iomap_valid);
+                                                       new_ioend);
                                } else {
                                        set_buffer_dirty(bh);
                                        unlock_buffer(bh);