by "the format is validated when the stream start" I mean the validation along the pipeline to check if each subdev has a compatible format hverkuil: continuing the discussion, for mem2mem devices, we have two types: 1) devices that always have 1:1 output->capture (like color converters) and 2) codecs, when output is not 1:1 to capture. In case of 1), what should driver do on ERROR in OUTPUT, dqbuf CAPTURE buffer with ERROR as well, always? In case of 2) should we just not dqbuf CAPTURE at all (only OUTPUT with ERROR), and use the CAPTURE buffer for later frames? hverkuil, pinchartl, sailus: just post the announce for the Media Controller workshop in Finland we need to start building the agenda for the meeting http://linuxtv.org/news.php?entry=2015-07-08.mchehab mchehab: are you planning to make an agenda? If not, then I can do it tomorrow feel free to do it maybe the best is to do on Friday to give people some time to send us some proposals pinchartl, hverkuil, mchehab: any reasons why media-ctl tool can't reside under kernel git? It might make it easier to sync kernel API updates with the tool I have been using it to test kernel changes, hence the question shuah: IMHO, it doesn't make sense to merge all random userspace tools at the Kernel tree media-ctl is not a special case... almost all stuff at v4l-utils should be updated to match the features upstream... the very same applies to alsa libraries, rasdaemon, etc merging those stuff at the Kernel will just create some mess shuah: see Makefile.am, target sync-with-kernel So anything in v4l-utils will be synced whenever kernel headers change. Once a new feature is added to media_tree I generally update v4l-utils as well. You can also make a branch and do make sync-with-kernel and pointer KERNEL_DIR to your own test kernel. shuah: I'm in a telco, I'll be back in 15 minutes yes no issues with sync-with-kernel - I am using that just a thought back we can't have all userspace tools in the kernel tree I'm afraid :-) I've never thought about it actually I initially implemented media-ctl in a standalone git repo and then got asked to move it to v4l-utils we have a few tools like perf and usbip is another example. It makes sense in some cases as it is easier to merge patches in at the same time. However, I agree with the general principle that not all tools belong in the kernel git I need to go back home now I'll be back a bit later hverkuil: i remember you requested some time ago to update the certs at linuxtv.org... just did it today still fake certs but they're valid until 2025 hey guys: enuminput is the right ioctl to return signal status right? it seems you can query an input and get nice information such as e.g. horizontal lock status hverkuil: btw, tw686x supports interlaced in non-SG DMA mode, using vb2_dma_contig_memops. works like a charm.