Support paired capture/output devices

[umlaeute]

Support paired capture/output devices

[umlaeute]

i think that the 2 device files approach is a very ugly hack, and honestly i would not like to include it unless forced by some kernel policy or the like (the device-pair approach was the one found in the v4l1 "vloopback" driver and i always hated it).
looking at the chrome patch rev151076, it seems that the chromiums devs explicitely prevent v4l2loopback devices (and mem2mem devices, or any other devices that are both input and output) to be enumerated for capture.

before i will accept any changes, i want to know what is wrong with a device being both input and output (the discussion on the chromium bug mentions something that was "requested by kernel people" but no further information is provided.)

[umlaeute]

i think a more appropriate fix would be to only announce OUTPUT capabilities if the device is not currently locked by a producer.
so as soon as the the device emitting data, it behaves as an ordinary fixed imagesize webcam (with only CAPTURE support)

[umlaeute]

i also wonder whether the device should announce the CAPTURE capability while it does not provide any image yet.

[umlaeute]

check todays experimental branch (3929af6) to see whether this helps

[hamishcampbell]

Works beautifully! 👍

[umlaeute]

cool, now that we now it works, i would like to make it optional. suggestions for the module parameter name are welcome

[ROBERT-MCDOWELL]

Hi, so what option does it need to allow chrome list v4l2loopback today?
thanks