[v1] Fix missing memory barriers on ARM

[PATCH 6/8] aio-wait: switch to smp_mb__after_rmw()

Posted by Paolo Bonzini 2 years, 11 months ago

The barrier comes after an atomic increment, so it is enough to use
smp_mb__after_rmw(); this avoids a double barrier on x86 systems.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
---
 include/block/aio-wait.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/include/block/aio-wait.h b/include/block/aio-wait.h
index dd9a7f6461ef..da13357bb8cf 100644
--- a/include/block/aio-wait.h
+++ b/include/block/aio-wait.h
@@ -85,7 +85,7 @@ extern AioWait global_aio_wait;
     /* Increment wait_->num_waiters before evaluating cond. */     \
     qatomic_inc(&wait_->num_waiters);                              \
     /* Paired with smp_mb in aio_wait_kick(). */                   \
-    smp_mb();                                                      \
+    smp_mb__after_rmw();                                           \
     if (ctx_ && in_aio_context_home_thread(ctx_)) {                \
         while ((cond)) {                                           \
             aio_poll(ctx_, true);                                  \
-- 
2.39.1

Re: [PATCH 6/8] aio-wait: switch to smp_mb__after_rmw()

Posted by Stefan Hajnoczi 2 years, 11 months ago

On Fri, Mar 03, 2023 at 06:19:37PM +0100, Paolo Bonzini wrote:
> The barrier comes after an atomic increment, so it is enough to use
> smp_mb__after_rmw(); this avoids a double barrier on x86 systems.
> 
> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
> ---
>  include/block/aio-wait.h | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)

Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>

Re: [PATCH 6/8] aio-wait: switch to smp_mb__after_rmw()

Posted by David Hildenbrand 2 years, 11 months ago

On 03.03.23 18:19, Paolo Bonzini wrote:
> The barrier comes after an atomic increment, so it is enough to use
> smp_mb__after_rmw(); this avoids a double barrier on x86 systems.
> 
> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
> ---
>   include/block/aio-wait.h | 2 +-
>   1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/include/block/aio-wait.h b/include/block/aio-wait.h
> index dd9a7f6461ef..da13357bb8cf 100644
> --- a/include/block/aio-wait.h
> +++ b/include/block/aio-wait.h
> @@ -85,7 +85,7 @@ extern AioWait global_aio_wait;
>       /* Increment wait_->num_waiters before evaluating cond. */     \
>       qatomic_inc(&wait_->num_waiters);                              \
>       /* Paired with smp_mb in aio_wait_kick(). */                   \
> -    smp_mb();                                                      \
> +    smp_mb__after_rmw();                                           \
>       if (ctx_ && in_aio_context_home_thread(ctx_)) {                \
>           while ((cond)) {                                           \
>               aio_poll(ctx_, true);                                  \

Reviewed-by: David Hildenbrand <david@redhat.com>

-- 
Thanks,

David / dhildenb

Re: [PATCH 6/8] aio-wait: switch to smp_mb__after_rmw()

Posted by Richard Henderson 2 years, 11 months ago

On 3/3/23 09:19, Paolo Bonzini wrote:
> The barrier comes after an atomic increment, so it is enough to use
> smp_mb__after_rmw(); this avoids a double barrier on x86 systems.
> 
> Signed-off-by: Paolo Bonzini<pbonzini@redhat.com>
> ---
>   include/block/aio-wait.h | 2 +-
>   1 file changed, 1 insertion(+), 1 deletion(-)

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>

r~