On (20/01/23 12:02), Hans Verkuil wrote: > On 1/22/20 3:05 AM, Sergey Senozhatsky wrote: > > On (20/01/10 10:47), Hans Verkuil wrote: > >> On 12/17/19 4:20 AM, Sergey Senozhatsky wrote: > >>> Preparations for future V4L2_FLAG_MEMORY_NON_CONSISTENT support. > >>> > >>> Extend vb2_core_reqbufs() with queue memory consistency flag. > >>> API permits queue's consistency attribute adjustment only if > >>> the queue has no allocated buffers, not busy, and does not have > >>> buffers waiting to be de-queued. > >> > >> Actually, you can call vb2_core_reqbufs() when buffers are allocated: > >> it will free the old buffers, then allocate the new ones. > >> So drop the 'has no allocated buffers' bit. > > > > Well, the wording, basically, follows the existing vb2_core_reqbufs() > > behavior "queue memory type"-wise. What I'm trying to say: > > How about this commit log replacement of the first paragraph: > > "Extend vb2_core_reqbufs() with queue memory consistency flag that is > applied to the newly allocated buffers." Looks good. > The bits about 'only if the queue has no allocated buffers, not busy, and does > not have buffers waiting to be de-queued.' is really irrelevant and confusing > (at least to me!). Agreed, those bits describe implementation details which can change. Better get rid of them. -ss
