Not all filesystems are using i_mutex for serialization - reflect that
in the comment. Also expand the reasoning a bit. It is complex enough
that it deserves more details.
Reported-by: Jan Beulich <JBeulich@xxxxxxxx>
Signed-off-by: Jan Kara <jack@xxxxxxx>
---
mm/truncate.c | 11 +++++++----
1 file changed, 7 insertions(+), 4 deletions(-)
BTW Dave has queued patch which removes the
WARN_ON(!mutex_locked(inode->i_mutex)) from the function. That should go to
Linus ASAP.
diff --git a/mm/truncate.c b/mm/truncate.c
index 261eaf6e5a19..b248c0c8dcd1 100644
--- a/mm/truncate.c
+++ b/mm/truncate.c
@@ -743,10 +743,13 @@ EXPORT_SYMBOL(truncate_setsize);
* changed.
*
* The function must be called after i_size is updated so that page fault
- * coming after we unlock the page will already see the new i_size.
- * The function must be called while we still hold i_mutex - this not only
- * makes sure i_size is stable but also that userspace cannot observe new
- * i_size value before we are prepared to store mmap writes at new inode size.
+ * coming after we unlock the page will already see the new i_size. The caller
+ * must make sure (generally by holding i_mutex but e.g. XFS uses its private
+ * lock) i_size cannot change from the new value while we are called. It must
+ * also make sure userspace cannot observe new i_size value before we are
+ * prepared to store mmap writes upto new inode size (otherwise userspace could
+ * think it stored data via mmap within i_size but they would get zeroed due to
+ * writeback & reclaim because they have no backing blocks).
*/
void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to)
{
--
1.8.1.4
--
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
