[PATCH net] tun: remove possible false sharing in tun_flow_update()

[PATCH net] tun: remove possible false sharing in tun_flow_update()

[Eric Dumazet]

As mentioned in https://github.com/google/ktsan/wiki/READ_ONCE-and-WRITE_ONCE#it-may-improve-performance
a C compiler can legally transform

if (e->queue_index != queue_index)
	e->queue_index = queue_index;

to :

	e->queue_index = queue_index;

Note that the code using jiffies has no issue, since jiffies
has volatile attribute.

if (e->updated != jiffies)
    e->updated = jiffies;

Fixes: 83b1bc122cab ("tun: align write-heavy flow entry members to a cache line")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Zhang Yu <zhangyu31@baidu.com>
Cc: Wang Li <wangli39@baidu.com>
Cc: Li RongQing <lirongqing@baidu.com>
Cc: Jason Wang <jasowang@redhat.com>
---
 drivers/net/tun.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/net/tun.c b/drivers/net/tun.c
index 812dc3a65efbb9d1ee2724e73978dbc4803ec171..a8d3141582a53caf407dc9aff61c452998de068f 100644
--- a/drivers/net/tun.c
+++ b/drivers/net/tun.c
@@ -526,8 +526,8 @@ static void tun_flow_update(struct tun_struct *tun, u32 rxhash,
 	e = tun_flow_find(head, rxhash);
 	if (likely(e)) {
 		/* TODO: keep queueing to old queue until it's empty? */
-		if (e->queue_index != queue_index)
-			e->queue_index = queue_index;
+		if (READ_ONCE(e->queue_index) != queue_index)
+			WRITE_ONCE(e->queue_index, queue_index);
 		if (e->updated != jiffies)
 			e->updated = jiffies;
 		sock_rps_record_flow_hash(e->rps_rxhash);
-- 
2.23.0.581.g78d2f28ef7-goog

Re: [PATCH net] tun: remove possible false sharing in tun_flow_update()

[Jason Wang]

On 2019/10/10 上午12:20, Eric Dumazet wrote:
> As mentioned in https://github.com/google/ktsan/wiki/READ_ONCE-and-WRITE_ONCE#it-may-improve-performance
> a C compiler can legally transform
>
> if (e->queue_index != queue_index)
> 	e->queue_index = queue_index;
>
> to :
>
> 	e->queue_index = queue_index;
>
> Note that the code using jiffies has no issue, since jiffies
> has volatile attribute.
>
> if (e->updated != jiffies)
>      e->updated = jiffies;
>
> Fixes: 83b1bc122cab ("tun: align write-heavy flow entry members to a cache line")
> Signed-off-by: Eric Dumazet <edumazet@google.com>
> Cc: Zhang Yu <zhangyu31@baidu.com>
> Cc: Wang Li <wangli39@baidu.com>
> Cc: Li RongQing <lirongqing@baidu.com>
> Cc: Jason Wang <jasowang@redhat.com>
> ---
>   drivers/net/tun.c | 4 ++--
>   1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/net/tun.c b/drivers/net/tun.c
> index 812dc3a65efbb9d1ee2724e73978dbc4803ec171..a8d3141582a53caf407dc9aff61c452998de068f 100644
> --- a/drivers/net/tun.c
> +++ b/drivers/net/tun.c
> @@ -526,8 +526,8 @@ static void tun_flow_update(struct tun_struct *tun, u32 rxhash,
>   	e = tun_flow_find(head, rxhash);
>   	if (likely(e)) {
>   		/* TODO: keep queueing to old queue until it's empty? */
> -		if (e->queue_index != queue_index)
> -			e->queue_index = queue_index;
> +		if (READ_ONCE(e->queue_index) != queue_index)
> +			WRITE_ONCE(e->queue_index, queue_index);
>   		if (e->updated != jiffies)
>   			e->updated = jiffies;
>   		sock_rps_record_flow_hash(e->rps_rxhash);


Acked-by: Jason Wang <jasowang@redhat.com>

Re: [PATCH net] tun: remove possible false sharing in tun_flow_update()

[Jakub Kicinski]

On Wed,  9 Oct 2019 09:20:02 -0700, Eric Dumazet wrote:
> As mentioned in https://github.com/google/ktsan/wiki/READ_ONCE-and-WRITE_ONCE#it-may-improve-performance
> a C compiler can legally transform
> 
> if (e->queue_index != queue_index)
> 	e->queue_index = queue_index;
> 
> to :
> 
> 	e->queue_index = queue_index;
> 
> Note that the code using jiffies has no issue, since jiffies
> has volatile attribute.
> 
> if (e->updated != jiffies)
>     e->updated = jiffies;
> 
> Fixes: 83b1bc122cab ("tun: align write-heavy flow entry members to a cache line")
> Signed-off-by: Eric Dumazet <edumazet@google.com>
> Cc: Zhang Yu <zhangyu31@baidu.com>
> Cc: Wang Li <wangli39@baidu.com>
> Cc: Li RongQing <lirongqing@baidu.com>
> Cc: Jason Wang <jasowang@redhat.com>

Applied, same story with stable, thanks!