Message ID | 1518598416-13069-1-git-send-email-nborisov@suse.com (mailing list archive) |
---|---|
State | New, archived |
Headers | show |
On Wed, Feb 14, 2018 at 10:53:36AM +0200, Nikolay Borisov wrote: > Using any kind of memory barriers around atomic operations which have > a return value is redundant, since those operations themselves are > fully ordered. atomic_t.txt states: > > - RMW operations that have a return value are fully ordered; > > Fully ordered primitives are ordered against everything prior and > everything subsequent. Therefore a fully ordered primitive is like > having an smp_mb() before and an smp_mb() after the primitive. > > Given this let's replace the extra memory barriers with comments. > > Signed-off-by: Nikolay Borisov <nborisov@suse.com> Reviewed-by: David Sterba <dsterba@suse.com> -- To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 8997e39c8e37..b6e4288010fa 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -8321,13 +8321,13 @@ static void btrfs_end_dio_bio(struct bio *bio) err = dip->subio_endio(dip->inode, btrfs_io_bio(bio), err); if (err) { - dip->errors = 1; - /* - * before atomic variable goto zero, we must make sure - * dip->errors is perceived to be set. + * We want to perceive the errors flag being set before + * decrementing the reference count. We don't need a barrier + * since atomic operations with a return value are fully + * ordered as per atomic_t.txt */ - smp_mb__before_atomic(); + dip->errors = 1; } /* if there are more bios still pending for this dio, just exit */ @@ -8515,10 +8515,11 @@ static int btrfs_submit_direct_hook(struct btrfs_dio_private *dip) out_err: dip->errors = 1; /* - * before atomic variable goto zero, we must - * make sure dip->errors is perceived to be set. + * Before atomic variable goto zero, we must make sure dip->errors is + * perceived to be set. This ordering is ensured by the fact that an + * atomic operations with a return value are fully ordered as per + * atomic_t.txt */ - smp_mb__before_atomic(); if (atomic_dec_and_test(&dip->pending_bios)) bio_io_error(dip->orig_bio);
Using any kind of memory barriers around atomic operations which have a return value is redundant, since those operations themselves are fully ordered. atomic_t.txt states: - RMW operations that have a return value are fully ordered; Fully ordered primitives are ordered against everything prior and everything subsequent. Therefore a fully ordered primitive is like having an smp_mb() before and an smp_mb() after the primitive. Given this let's replace the extra memory barriers with comments. Signed-off-by: Nikolay Borisov <nborisov@suse.com> --- fs/btrfs/inode.c | 17 +++++++++-------- 1 file changed, 9 insertions(+), 8 deletions(-)