pinchartl: kbingham: good morning! reviving the topic of https://patchwork.kernel.org/patch/10491117/
do you think we could either go with this or switch uvc to use kmalloc()
?
tfiga: Kieran will be back from holidays tomorrow
Kwiboo: Hi
any reason for not grouping the rockchip_unregister_mc() and video_unregister_device() calls in a single if() block here https://github.com/Kwiboo/linux-rockchip/commit/0729b53c825a630c9040d28541d2635a42177361#diff-997d7f50e608e02f2515ac8345e4ebd5R818 ?
bbrezillon, hello :-)
not really, it was only the last test I found video_device_release the issue, so I just tried to follow same pattern the error path in rockchip_vpu_probe
Kwiboo: ok, if that works for you, I'd like to group them
so I reordered all unregister/release to happen in reverse order from alloc/register call order, and I did not like that that order in error path was different from remove, ocd thoughts ;)
ezequielg: any reason for not using devm_kasprinf() to allocate entity->name?
video_device_release and pm_runtime_dont_use_autosuspend was the important changes, the error path in register/probe seemd out of order when I was reviewing code in search for crash culprit
Kwiboo: got it
it's just that it looks weird to have the check done twice
Kwiboo: also not a big fan of the asymmetry we have because of rockchip_register_media_controller() (there's no unregister counterpart)
bbrezillon, hehe, I have tried to kept my fingers away from other parts then mpeg-2 in the driver, at least there is less duplicated code with v3 :-)
bbrezillon: nope, devm_kasprintf is fine I think.
bbrezillon: also, if you dislike rockchip_register_media_controller, you can kill it.