Re: [PATCH] dma-buf/reservation: should keep the new fence in add_shared_inplace

2018-03-06 Thread Liu, Monk 
why? is there a design doc mentioned for this on reservation ?


From: Christian K?nig 
Sent: Tuesday, March 6, 2018 4:03:39 PM
To: Liu, Monk; dri-devel@lists.freedesktop.org
Subject: Re: [PATCH] dma-buf/reservation: should keep the new fence in 
add_shared_inplace

NAK, the newly added fence must always be newer than the existing one.

Christian.

Am 06.03.2018 um 04:09 schrieb Monk Liu:
> Change-Id: If6a979ba9fd6c923b82212f35f07a9ff31c86767
> Signed-off-by: Monk Liu 
> ---
>   drivers/dma-buf/reservation.c | 2 +-
>   1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/dma-buf/reservation.c b/drivers/dma-buf/reservation.c
> index 314eb10..29b7e45 100644
> --- a/drivers/dma-buf/reservation.c
> +++ b/drivers/dma-buf/reservation.c
> @@ -118,7 +118,7 @@ reservation_object_add_shared_inplace(struct 
> reservation_object *obj,
>old_fence = rcu_dereference_protected(fobj->shared[i],
>reservation_object_held(obj));
>
> - if (old_fence->context == fence->context) {
> + if (dma_fence_is_later(fence, old_fence)) {
>/* memory barrier is added by write_seqcount_begin */
>RCU_INIT_POINTER(fobj->shared[i], fence);
>write_seqcount_end(&obj->seq);

___
dri-devel mailing list
dri-devel@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/dri-devel

Re: [PATCH] dma-buf/reservation: should keep the new fence in add_shared_inplace

2018-03-06 Thread Christian König 

NAK, the newly added fence must always be newer than the existing one.

Christian.

Am 06.03.2018 um 04:09 schrieb Monk Liu:

Change-Id: If6a979ba9fd6c923b82212f35f07a9ff31c86767
Signed-off-by: Monk Liu 
---
  drivers/dma-buf/reservation.c | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/dma-buf/reservation.c b/drivers/dma-buf/reservation.c
index 314eb10..29b7e45 100644
--- a/drivers/dma-buf/reservation.c
+++ b/drivers/dma-buf/reservation.c
@@ -118,7 +118,7 @@ reservation_object_add_shared_inplace(struct 
reservation_object *obj,
old_fence = rcu_dereference_protected(fobj->shared[i],
reservation_object_held(obj));
  
-		if (old_fence->context == fence->context) {

+   if (dma_fence_is_later(fence, old_fence)) {
/* memory barrier is added by write_seqcount_begin */
RCU_INIT_POINTER(fobj->shared[i], fence);
write_seqcount_end(&obj->seq);


___
dri-devel mailing list
dri-devel@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/dri-devel

RE: [PATCH] dma-buf/reservation: should keep the new fence in add_shared_inplace

2018-03-05 Thread Liu, Monk 
Yeah, right 

-Original Message-
From: Zhou, David(ChunMing) 
Sent: 2018年3月6日 11:38
To: Liu, Monk ; dri-devel@lists.freedesktop.org
Subject: Re: [PATCH] dma-buf/reservation: should keep the new fence in 
add_shared_inplace



On 2018年03月06日 11:09, Monk Liu wrote:
> Change-Id: If6a979ba9fd6c923b82212f35f07a9ff31c86767
> Signed-off-by: Monk Liu 
> ---
>   drivers/dma-buf/reservation.c | 2 +-
>   1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/dma-buf/reservation.c 
> b/drivers/dma-buf/reservation.c index 314eb10..29b7e45 100644
> --- a/drivers/dma-buf/reservation.c
> +++ b/drivers/dma-buf/reservation.c
> @@ -118,7 +118,7 @@ reservation_object_add_shared_inplace(struct 
> reservation_object *obj,
>   old_fence = rcu_dereference_protected(fobj->shared[i],
>   reservation_object_held(obj));
>   
> - if (old_fence->context == fence->context) {
> + if (dma_fence_is_later(fence, old_fence)) {
OK, good catch, to avoid warning of different context, which should be "if 
((old_fence->context == fence->context) && dma_fence_is_later(fence,
old_fence)) { "
and reservation_object_add_shared_replace need this fix as well.

Regards,
David Zhou
>   /* memory barrier is added by write_seqcount_begin */
>   RCU_INIT_POINTER(fobj->shared[i], fence);
>   write_seqcount_end(&obj->seq);

___
dri-devel mailing list
dri-devel@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/dri-devel

Re: [PATCH] dma-buf/reservation: should keep the new fence in add_shared_inplace

2018-03-05 Thread zhoucm1 



On 2018年03月06日 11:09, Monk Liu wrote:

Change-Id: If6a979ba9fd6c923b82212f35f07a9ff31c86767
Signed-off-by: Monk Liu 
---
  drivers/dma-buf/reservation.c | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/dma-buf/reservation.c b/drivers/dma-buf/reservation.c
index 314eb10..29b7e45 100644
--- a/drivers/dma-buf/reservation.c
+++ b/drivers/dma-buf/reservation.c
@@ -118,7 +118,7 @@ reservation_object_add_shared_inplace(struct 
reservation_object *obj,
old_fence = rcu_dereference_protected(fobj->shared[i],
reservation_object_held(obj));
  
-		if (old_fence->context == fence->context) {

+   if (dma_fence_is_later(fence, old_fence)) {
OK, good catch, to avoid warning of different context, which should be 
"if ((old_fence->context == fence->context) && dma_fence_is_later(fence, 
old_fence)) { "

and reservation_object_add_shared_replace need this fix as well.

Regards,
David Zhou

/* memory barrier is added by write_seqcount_begin */
RCU_INIT_POINTER(fobj->shared[i], fence);
write_seqcount_end(&obj->seq);


___
dri-devel mailing list
dri-devel@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/dri-devel