2022-08-09 16:41:17

by Uros Bizjak

[permalink] [raw]
Subject: [PATCH] fs/btrfs: Use atomic_try_cmpxchg in free_extent_buffer

Use `atomic_try_cmpxchg(ptr, &old, new)` instead of
`atomic_cmpxchg(ptr, old, new) == old` in free_extent_buffer. This
has two benefits:

- The x86 cmpxchg instruction returns success in the ZF flag, so this
change saves a compare after cmpxchg, as well as a related move
instruction in the front of cmpxchg.

- atomic_try_cmpxchg implicitly assigns the *ptr value to &old when
cmpxchg fails, enabling further code simplifications.

This patch has no functional change.

Cc: Chris Mason <[email protected]>
Cc: Josef Bacik <[email protected]>
Cc: David Sterba <[email protected]>
Signed-off-by: Uros Bizjak <[email protected]>
---
fs/btrfs/extent_io.c | 6 ++----
1 file changed, 2 insertions(+), 4 deletions(-)

diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c
index bfae67c593c5..15ff196cbd6d 100644
--- a/fs/btrfs/extent_io.c
+++ b/fs/btrfs/extent_io.c
@@ -6328,18 +6328,16 @@ static int release_extent_buffer(struct extent_buffer *eb)
void free_extent_buffer(struct extent_buffer *eb)
{
int refs;
- int old;
if (!eb)
return;

+ refs = atomic_read(&eb->refs);
while (1) {
- refs = atomic_read(&eb->refs);
if ((!test_bit(EXTENT_BUFFER_UNMAPPED, &eb->bflags) && refs <= 3)
|| (test_bit(EXTENT_BUFFER_UNMAPPED, &eb->bflags) &&
refs == 1))
break;
- old = atomic_cmpxchg(&eb->refs, refs, refs - 1);
- if (old == refs)
+ if (atomic_try_cmpxchg(&eb->refs, &refs, refs - 1))
return;
}

--
2.37.1


2022-08-16 02:04:00

by Boris Burkov

[permalink] [raw]
Subject: Re: [PATCH] fs/btrfs: Use atomic_try_cmpxchg in free_extent_buffer

On Tue, Aug 09, 2022 at 06:36:33PM +0200, Uros Bizjak wrote:
> Use `atomic_try_cmpxchg(ptr, &old, new)` instead of
> `atomic_cmpxchg(ptr, old, new) == old` in free_extent_buffer. This
> has two benefits:
>
> - The x86 cmpxchg instruction returns success in the ZF flag, so this
> change saves a compare after cmpxchg, as well as a related move
> instruction in the front of cmpxchg.
>
> - atomic_try_cmpxchg implicitly assigns the *ptr value to &old when
> cmpxchg fails, enabling further code simplifications.
>
> This patch has no functional change.
>
> Cc: Chris Mason <[email protected]>
> Cc: Josef Bacik <[email protected]>
> Cc: David Sterba <[email protected]>
> Signed-off-by: Uros Bizjak <[email protected]>
Reviewed-by: Boris Burkov <[email protected]>
> ---
> fs/btrfs/extent_io.c | 6 ++----
> 1 file changed, 2 insertions(+), 4 deletions(-)
>
> diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c
> index bfae67c593c5..15ff196cbd6d 100644
> --- a/fs/btrfs/extent_io.c
> +++ b/fs/btrfs/extent_io.c
> @@ -6328,18 +6328,16 @@ static int release_extent_buffer(struct extent_buffer *eb)
> void free_extent_buffer(struct extent_buffer *eb)
> {
> int refs;
> - int old;
> if (!eb)
> return;
>
> + refs = atomic_read(&eb->refs);
> while (1) {
> - refs = atomic_read(&eb->refs);
> if ((!test_bit(EXTENT_BUFFER_UNMAPPED, &eb->bflags) && refs <= 3)
> || (test_bit(EXTENT_BUFFER_UNMAPPED, &eb->bflags) &&
> refs == 1))
> break;
> - old = atomic_cmpxchg(&eb->refs, refs, refs - 1);
> - if (old == refs)
> + if (atomic_try_cmpxchg(&eb->refs, &refs, refs - 1))
> return;
> }
>
> --
> 2.37.1
>

2022-08-18 17:29:50

by David Sterba

[permalink] [raw]
Subject: Re: [PATCH] fs/btrfs: Use atomic_try_cmpxchg in free_extent_buffer

On Tue, Aug 09, 2022 at 06:36:33PM +0200, Uros Bizjak wrote:
> Use `atomic_try_cmpxchg(ptr, &old, new)` instead of
> `atomic_cmpxchg(ptr, old, new) == old` in free_extent_buffer. This
> has two benefits:
>
> - The x86 cmpxchg instruction returns success in the ZF flag, so this
> change saves a compare after cmpxchg, as well as a related move
> instruction in the front of cmpxchg.
>
> - atomic_try_cmpxchg implicitly assigns the *ptr value to &old when
> cmpxchg fails, enabling further code simplifications.
>
> This patch has no functional change.
>
> Cc: Chris Mason <[email protected]>
> Cc: Josef Bacik <[email protected]>
> Cc: David Sterba <[email protected]>
> Signed-off-by: Uros Bizjak <[email protected]>

It's not in a performance critical code but resulting code is slightly
more efficient. Added to misc-next, thanks.