tmpfs: fix regression hang in fallocate undo

commit 7f556567036cb7f89aabe2f0954b08566b4efb53 upstream.

The well-spotted fallocate undo fix is good in most cases, but not when
fallocate failed on the very first page.  index 0 then passes lend -1
to shmem_undo_range(), and that has two bad effects: (a) that it will
undo every fallocation throughout the file, unrestricted by the current
range; but more importantly (b) it can cause the undo to hang, because
lend -1 is treated as truncation, which makes it keep on retrying until
every page has gone, but those already fully instantiated will never go
away.  Big thank you to xfstests generic/269 which demonstrates this.

Fixes: b9b4bb26af01 ("tmpfs: don't undo fallocate past its last page")
Signed-off-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
[bwh: Backported to 3.16: use PAGE_CACHE_SHIFT instead of PAGE_SHIFT]
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>

diff --git a/mm/shmem.c b/mm/shmem.c
index c4dc08ae240ffbd143c7c1b9092c409757fb0168..a2b87affdc23660a437703b3e1e75c3ea9d98fbb 100644 (file)
--- a/mm/shmem.c
+++ b/mm/shmem.c
@@ -1865,9 +1865,11 @@ static long shmem_fallocate(struct file *file, int mode, loff_t offset,
                                                                        NULL);
                if (error) {
                        /* Remove the !PageUptodate pages we added */
-                       shmem_undo_range(inode,
-                               (loff_t)start << PAGE_CACHE_SHIFT,
-                               ((loff_t)index << PAGE_CACHE_SHIFT) - 1, true);
+                       if (index > start) {
+                               shmem_undo_range(inode,
+                                   (loff_t)start << PAGE_CACHE_SHIFT,
+                                   ((loff_t)index << PAGE_CACHE_SHIFT) - 1, true);
+                       }
                        goto undone;
                }