2021-12-14 04:53:53

by Jammy Huang

[permalink] [raw]
Subject: [PATCH] media: aspeed: Fix no complete irq for non-64-aligned width

In ast2500, engine will stop occasionally for 1360x768.

This is a bug which has been addressed, but the workaround is specific
for 1680 only. Here we make it more complete.

Signed-off-by: Jammy Huang <[email protected]>
---
drivers/media/platform/aspeed-video.c | 12 +++++++-----
1 file changed, 7 insertions(+), 5 deletions(-)

diff --git a/drivers/media/platform/aspeed-video.c b/drivers/media/platform/aspeed-video.c
index 793b2adaa0f5..4d3e6b105d44 100644
--- a/drivers/media/platform/aspeed-video.c
+++ b/drivers/media/platform/aspeed-video.c
@@ -1055,18 +1055,20 @@ static void aspeed_video_set_resolution(struct aspeed_video *video)
/* Set capture/compression frame sizes */
aspeed_video_calc_compressed_size(video, size);

- if (video->active_timings.width == 1680) {
+ if (!IS_ALIGNED(act->width, 64)) {
/*
* This is a workaround to fix a silicon bug on A1 and A2
* revisions. Since it doesn't break capturing operation of
* other revisions, use it for all revisions without checking
- * the revision ID. It picked 1728 which is a very next
- * 64-pixels aligned value to 1680 to minimize memory bandwidth
+ * the revision ID. It picked new width which is a very next
+ * 64-pixels aligned value to minimize memory bandwidth
* and to get better access speed from video engine.
*/
+ u32 width = ALIGN(act->width, 64);
+
aspeed_video_write(video, VE_CAP_WINDOW,
- 1728 << 16 | act->height);
- size += (1728 - 1680) * video->active_timings.height;
+ width << 16 | act->height);
+ size = width * act->height;
} else {
aspeed_video_write(video, VE_CAP_WINDOW,
act->width << 16 | act->height);
--
2.25.1



2022-01-12 00:52:09

by Joel Stanley

[permalink] [raw]
Subject: Re: [PATCH] media: aspeed: Fix no complete irq for non-64-aligned width

On Tue, 14 Dec 2021 at 04:53, Jammy Huang <[email protected]> wrote:
>
> In ast2500, engine will stop occasionally for 1360x768.
>
> This is a bug which has been addressed, but the workaround is specific
> for 1680 only. Here we make it more complete.
>
> Signed-off-by: Jammy Huang <[email protected]>
> ---
> drivers/media/platform/aspeed-video.c | 12 +++++++-----
> 1 file changed, 7 insertions(+), 5 deletions(-)
>
> diff --git a/drivers/media/platform/aspeed-video.c b/drivers/media/platform/aspeed-video.c
> index 793b2adaa0f5..4d3e6b105d44 100644
> --- a/drivers/media/platform/aspeed-video.c
> +++ b/drivers/media/platform/aspeed-video.c
> @@ -1055,18 +1055,20 @@ static void aspeed_video_set_resolution(struct aspeed_video *video)
> /* Set capture/compression frame sizes */
> aspeed_video_calc_compressed_size(video, size);
>
> - if (video->active_timings.width == 1680) {
> + if (!IS_ALIGNED(act->width, 64)) {
> /*
> * This is a workaround to fix a silicon bug on A1 and A2

Please add: "a AST2500 silicon bug" so we know which A1/A2 this is
referring to.

With that added, you can add:

Reviewed-by: Joel Stanley <[email protected]>

> * revisions. Since it doesn't break capturing operation of
> * other revisions, use it for all revisions without checking
> - * the revision ID. It picked 1728 which is a very next
> - * 64-pixels aligned value to 1680 to minimize memory bandwidth
> + * the revision ID. It picked new width which is a very next
> + * 64-pixels aligned value to minimize memory bandwidth
> * and to get better access speed from video engine.
> */
> + u32 width = ALIGN(act->width, 64);
> +
> aspeed_video_write(video, VE_CAP_WINDOW,
> - 1728 << 16 | act->height);
> - size += (1728 - 1680) * video->active_timings.height;
> + width << 16 | act->height);
> + size = width * act->height;

You could make it clearer by putting the write on one line:

aspeed_video_write(video, VE_CAP_WINDOW, width << 16 |
act->height);

2022-01-12 06:06:30

by Jammy Huang

[permalink] [raw]
Subject: Re: [PATCH] media: aspeed: Fix no complete irq for non-64-aligned width

Hi Joel,

OK, thanks for your review. I will send another updated patch per your
advice.

On 2022/1/12 上午 08:51, Joel Stanley wrote:
> On Tue, 14 Dec 2021 at 04:53, Jammy Huang <[email protected]> wrote:
>> In ast2500, engine will stop occasionally for 1360x768.
>>
>> This is a bug which has been addressed, but the workaround is specific
>> for 1680 only. Here we make it more complete.
>>
>> Signed-off-by: Jammy Huang <[email protected]>
>> ---
>> drivers/media/platform/aspeed-video.c | 12 +++++++-----
>> 1 file changed, 7 insertions(+), 5 deletions(-)
>>
>> diff --git a/drivers/media/platform/aspeed-video.c b/drivers/media/platform/aspeed-video.c
>> index 793b2adaa0f5..4d3e6b105d44 100644
>> --- a/drivers/media/platform/aspeed-video.c
>> +++ b/drivers/media/platform/aspeed-video.c
>> @@ -1055,18 +1055,20 @@ static void aspeed_video_set_resolution(struct aspeed_video *video)
>> /* Set capture/compression frame sizes */
>> aspeed_video_calc_compressed_size(video, size);
>>
>> - if (video->active_timings.width == 1680) {
>> + if (!IS_ALIGNED(act->width, 64)) {
>> /*
>> * This is a workaround to fix a silicon bug on A1 and A2
> Please add: "a AST2500 silicon bug" so we know which A1/A2 this is
> referring to.
>
> With that added, you can add:
>
> Reviewed-by: Joel Stanley <[email protected]>
>
>> * revisions. Since it doesn't break capturing operation of
>> * other revisions, use it for all revisions without checking
>> - * the revision ID. It picked 1728 which is a very next
>> - * 64-pixels aligned value to 1680 to minimize memory bandwidth
>> + * the revision ID. It picked new width which is a very next
>> + * 64-pixels aligned value to minimize memory bandwidth
>> * and to get better access speed from video engine.
>> */
>> + u32 width = ALIGN(act->width, 64);
>> +
>> aspeed_video_write(video, VE_CAP_WINDOW,
>> - 1728 << 16 | act->height);
>> - size += (1728 - 1680) * video->active_timings.height;
>> + width << 16 | act->height);
>> + size = width * act->height;
> You could make it clearer by putting the write on one line:
>
> aspeed_video_write(video, VE_CAP_WINDOW, width << 16 |
> act->height);

--
Best Regards
Jammy