[08:20] <tfiga> hverkuil: I'll try to take a look a bit later [08:27] <hverkuil> ezequielg: the rkvdec driver only needs a follow-up patch adding a MAINTAINERS entry, and then I can post a PR for it. Can you do that today? (PS: don't forget to also refer to the binding file in the MAINTAINERS entry, it's easy to forget about that one) [09:22] <hverkuil> ezequielg: see my follow-up mails, a v6 is needed after all. [09:28] <ezequielg> hverkuil: ok. I will try today. [10:53] <ezequielg> hverkuil: I'll post today after my daily meetings :) [18:09] <dafna2> hverkuil: hi, the rkisp1 driver fails in the compliance tests, in the case of using userptr when streaming [18:09] <dafna2> the error is because the vb2_dma_contig_memops expect the buffers allocated by the userspace to be alligned [18:10] <dafna2> I wonder if it should be fixed in v4l-utils, or either just drop the support for userptr? [18:12] <hverkuil> Drop support for userptr. userptr support makes no sense for drivers based on vb2_dma_contig. Support for userptr for dma_contig exists primarily for old drivers that abused userptr giving it pointers to known contiguous data (usually from a special memory pool). [18:12] <hverkuil> I'm not even sure if those drivers are still in use, I thought it was mostly soc-camera that did this. [18:16] <hverkuil> Tomorrow I'll check to see who still uses this and if support for this can be killed. [18:20] <dafna2> ok, and if it can you will send a patch removing support from all drivers with contig_dma? [18:20] <dafna2> in that case I don't have to write a patch just for rkisp1 [18:26] <hverkuil> That's putting the cart before the horse :-) [18:27] <hverkuil> It's a simple patch for rkisp1: just remove VB2_USERPTR for any dma_config queue. [18:37] <dafna2> :-) ok