lkml.org 
[lkml]   [2020]   [Mar]   [7]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
SubjectRe: [PATCHv4 04/11] videobuf2: add queue memory consistency parameter
From
Date
On 07/03/2020 08:50, Sergey Senozhatsky wrote:
> On (20/03/06 15:04), Hans Verkuil wrote:
> [..]
>>> +static bool verify_consistency_attr(struct vb2_queue *q, bool consistent_mem)
>>> +{
>>> + bool queue_attr = q->dma_attrs & DMA_ATTR_NON_CONSISTENT;
>>> +
>>> + if (consistent_mem != queue_attr) {
>>
>> This is the wrong way around!
>>
>> It's much better to write it like this:
>>
>> bool queue_is_consistent = !(q->dma_attrs & DMA_ATTR_NON_CONSISTENT);
>>
>> if (consistent_mem != queue_is_consistent) {
>
> Hmm... That's a great catch. Thanks for spotting this.
> Puzzled, how come I've never seen problems.
>
>> What concerns me more is that this means that this series has not been
>> tested properly. I found this when testing with v4l2-compliance and vivid.
>
> I fully understand your concerns. Give me a moment to figure
> out what's going on...
>
>
> OK.
>
> Apparently, the user-space I'm using for tests, utilizes different
> call path. vb2_core_create_bufs() is never even invoked. Hence queue
> consistency vs. request consistency checks are not performed.
>
> What happens, instead, is v4l_reqbufs()->vb2_core_reqbufs() path.
> It orphans existing buffers (if any), sets queue memory model, sets
> queue consistency model (DMA attr), then allocates buffers.
>
> On my test environment, I see that vb2_core_reqbufs() orphans the
> buffers, but it's always due to "*count == 0 || q->num_buffers != 0"
> conditions. The user-space I'm using does not twist queue ->memory
> or consistency attr, so the tests I'm running are limited in scenarios.

That's why v4l2-compliance is so important: it tests 'twisty code' for
correct handling.

>
> verify_consistency_attr() is not on the list of reasons to orphan
> allocated buffer. It probably should be, tho.
>
> ===
> diff --git a/drivers/media/common/videobuf2/videobuf2-core.c b/drivers/media/common/videobuf2/videobuf2-core.c
> index afb3c21a5902..d6b1d32bef3f 100644
> --- a/drivers/media/common/videobuf2/videobuf2-core.c
> +++ b/drivers/media/common/videobuf2/videobuf2-core.c
> @@ -730,7 +730,8 @@ int vb2_core_reqbufs(struct vb2_queue *q, enum vb2_memory memory,
> }
>
> if (*count == 0 || q->num_buffers != 0 ||
> - (q->memory != VB2_MEMORY_UNKNOWN && q->memory != memory)) {
> + (q->memory != VB2_MEMORY_UNKNOWN && q->memory != memory) ||
> + !verify_consistency_attr(q, consistent_mem)) {
> /*
> * We already have buffers allocated, so first check if they
> * are not in use and can be freed.
> ===
>
>>> + dprintk(1, "memory consistency model mismatch\n");
>>> + return false;
>>> + }
>>> + return true;
>>> +}
>>> +
>>> int vb2_core_create_bufs(struct vb2_queue *q, enum vb2_memory memory,
>>> - unsigned int *count, unsigned requested_planes,
>>> - const unsigned requested_sizes[])
>>> + bool consistent_mem, unsigned int *count,
>>> + unsigned requested_planes,
>>> + const unsigned requested_sizes[])
>>
>> Use 'unsigned int' in the two lines above, as per checkpatch suggestion.
>
> OK, will do.
>
> This comes from the original code. There are 'unsigned'-s in the
> existing code, I saw it and didn't want to modify, in order to keep
> diffstats shorter.

Yeah, but the prototype was already inconsistent (count is an unsigned int *),
so it makes sense to fix this.

Regards,

Hans

>
> -ss
>

\
 
 \ /
  Last update: 2020-03-07 11:05    [W:0.056 / U:0.356 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site