[PATCH] btrfs: remove broken memory barrier

[PATCH] btrfs: remove broken memory barrier

[Nikolay Borisov]

Commit 38851cc19adb ("Btrfs: implement unlocked dio write") implemented
unlocked dio write, allowing multiple dio writers to write to non-overlapping,
and non-eof-extending regions. In doing so it also introduced a broken memory
barrier. It is broken due to 2 things:

1. Memory barriers _MUST_ always be paired, this is clearly not the case here

2. Checkpatch actually produces a warning if a memory barrier is introduced that
doesn't have a comment explaining how it's being paired.

Signed-off-by: Nikolay Borisov <nborisov@suse.com>
---
 fs/btrfs/inode.c | 1 -
 1 file changed, 1 deletion(-)

diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 95c212037095..5e48d2c10152 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -8731,7 +8731,6 @@ static ssize_t btrfs_direct_IO(struct kiocb *iocb, struct iov_iter *iter)
 		return 0;
 
 	inode_dio_begin(inode);
-	smp_mb__after_atomic();
 
 	/*
 	 * The generic stuff only does filemap_write_and_wait_range, which
-- 
2.7.4

Re: [PATCH] btrfs: remove broken memory barrier

[David Sterba]

On Tue, Aug 01, 2017 at 06:25:56PM +0300, Nikolay Borisov wrote:
> Commit 38851cc19adb ("Btrfs: implement unlocked dio write") implemented
> unlocked dio write, allowing multiple dio writers to write to non-overlapping,
> and non-eof-extending regions. In doing so it also introduced a broken memory
> barrier. It is broken due to 2 things:
> 
> 1. Memory barriers _MUST_ always be paired, this is clearly not the case here
> 
> 2. Checkpatch actually produces a warning if a memory barrier is introduced that
> doesn't have a comment explaining how it's being paired.

Specifically for inode::i_dio_count that's wrapped inside
inode_dio_begin, there is no explicit barrier semantics attached, so
removing is fine as the atomic is used in common the waiter/wakeup pattern.

Reviewed-by: David Sterba <dsterba@suse.com>