Btrfs: Fix the multi-bio code to save the original bio for completion
The multi-bio code is responsible for duplicating blocks in raid1 and
single spindle duplication. It has counters to make sure all of
the locations for a given extent are properly written before io completion
is returned to the higher layers.
But, it didn't always complete the same bio it was given, sometimes a
clone was completed instead. This lead to problems with the async
work queues because they saved a pointer to the bio in a struct off
bi_private.
The fix is to remember the original bio and only complete that one.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c
index 0de51e3..09311b3 100644
--- a/fs/btrfs/volumes.c
+++ b/fs/btrfs/volumes.c
@@ -2070,6 +2070,7 @@
#endif
{
struct btrfs_multi_bio *multi = bio->bi_private;
+ int is_orig_bio = 0;
#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23)
if (bio->bi_size)
@@ -2078,7 +2079,14 @@
if (err)
atomic_inc(&multi->error);
+ if (bio == multi->orig_bio)
+ is_orig_bio = 1;
+
if (atomic_dec_and_test(&multi->stripes_pending)) {
+ if (!is_orig_bio) {
+ bio_put(bio);
+ bio = multi->orig_bio;
+ }
bio->bi_private = multi->private;
bio->bi_end_io = multi->end_io;
/* only send an error to the higher layers if it is
@@ -2101,7 +2109,7 @@
#else
bio_endio(bio, err);
#endif
- } else {
+ } else if (!is_orig_bio) {
bio_put(bio);
}
#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23)
@@ -2196,6 +2204,7 @@
}
multi->end_io = first_bio->bi_end_io;
multi->private = first_bio->bi_private;
+ multi->orig_bio = first_bio;
atomic_set(&multi->stripes_pending, multi->num_stripes);
while(dev_nr < total_devs) {