2020-01-29 18:24:07

by Davidlohr Bueso

Subject: [PATCH] rbd: optimize barrier usage for Rmw atomic bitops

For both set and clear_bit, we can avoid the unnecessary barrier
on non LL/SC architectures, such as x86. Instead, use the
smp_mb__{before,after}_atomic() calls.

Signed-off-by: Davidlohr Bueso <[email protected]>
---
drivers/block/rbd.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/block/rbd.c b/drivers/block/rbd.c
index 2b184563cd32..7bc79b2b8f65 100644
--- a/drivers/block/rbd.c
+++ b/drivers/block/rbd.c
@@ -1371,13 +1371,13 @@ static void rbd_osd_submit(struct ceph_osd_request *osd_req)
static void img_request_layered_set(struct rbd_img_request *img_request)
{
set_bit(IMG_REQ_LAYERED, &img_request->flags);
- smp_mb();
+ smp_mb__after_atomic();
}

static void img_request_layered_clear(struct rbd_img_request *img_request)
{
clear_bit(IMG_REQ_LAYERED, &img_request->flags);
- smp_mb();
+ smp_mb__after_atomic();
}

static bool img_request_layered_test(struct rbd_img_request *img_request)
--
2.16.4

2020-01-30 12:53:25

by Ilya Dryomov

Subject: Re: [PATCH] rbd: optimize barrier usage for Rmw atomic bitops

On Wed, Jan 29, 2020 at 7:23 PM Davidlohr Bueso <[email protected]> wrote:
>
> For both set and clear_bit, we can avoid the unnecessary barrier
> on non LL/SC architectures, such as x86. Instead, use the
> smp_mb__{before,after}_atomic() calls.
>
> Signed-off-by: Davidlohr Bueso <[email protected]>
> ---
> drivers/block/rbd.c | 4 ++--
> 1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/block/rbd.c b/drivers/block/rbd.c
> index 2b184563cd32..7bc79b2b8f65 100644
> --- a/drivers/block/rbd.c
> +++ b/drivers/block/rbd.c
> @@ -1371,13 +1371,13 @@ static void rbd_osd_submit(struct ceph_osd_request *osd_req)
> static void img_request_layered_set(struct rbd_img_request *img_request)
> {
> set_bit(IMG_REQ_LAYERED, &img_request->flags);
> - smp_mb();
> + smp_mb__after_atomic();
> }
>
> static void img_request_layered_clear(struct rbd_img_request *img_request)
> {
> clear_bit(IMG_REQ_LAYERED, &img_request->flags);
> - smp_mb();
> + smp_mb__after_atomic();
> }
>
> static bool img_request_layered_test(struct rbd_img_request *img_request)

Hi Davidlohr,

I don't think these barriers are needed at all. I'll remove them.

Thanks,

Ilya

2020-01-30 15:13:36

by Davidlohr Bueso

Subject: Re: [PATCH] rbd: optimize barrier usage for Rmw atomic bitops

On Thu, 30 Jan 2020, Ilya Dryomov wrote:

>I don't think these barriers are needed at all. I'll remove them.

Even better, thanks.