On Tue, Sep 28, 2021 at 2:02 AM Steve Cho <stevecho@xxxxxxxxxxxx> wrote: > > Hi Ezequiel, > > Thank you for reviewing these series from Yunfei! > This series is one of the main obstacles for us at the moment for MTK > so please continue to help & support reviewing this series. > > > > According to google's suggestion, it's better not to use v4l2 async > > > also. > > > > Hum? I haven't seen such objection on the mailing list. > Maybe coming from Tzung-Bi? > Yunfei, please let us know. I do object to using V4L2 async. It's designed for independent components of media pipelines, handled by multiple different drivers and also modelled as V4L2 subdevices. We don't have anything like that here. How about just open coding something trivial that only fits the needs of this specific driver? I think it would be as simple as a linked list and registering the V4L2 devices only after all the nodes probe. Best regards, Tomasz
