08:16 <vknecht[m]> robclark: thanks, Jessica's patches seem to have helped with bringing panel back up reliabily after suspend/screen-off, even if I still get many mdp5_pipe_release warns eg. after 09:56:17 in this log : https://sebsauvage.net/paste/?68d9d18335dafad5#OkjIWcdaBhSA9Xo7hFSE5gOqXivzDAYZUknWmg7HfN0=

14:58 <robclark> hmm, if I had to take a guess, in the system suspend path the atomic helper tries to disable already disabled planes? So maybe that WARN_ON() should be removed? Perhaps you could `echo 0x1f > /sys/modules/drm/parameters/debug` and then trigger suspend, hopefully that would confirm or disprove the theory?

14:58 <robclark> vknecht[m]: ^^^

15:54 <vknecht[m]> robclark: couldn't say if warn is undue, here's 0x1f logs when pressing button for panel off, then for back at ~ 17:41:04.272 : https://sebsauvage.net/paste/?a331dbb600455501#xXyziK0iDvb2UyIL7s8xd1ViN2tgcGNwdGEcI6OSMI0=

15:57 <vknecht[m]> (it's drm_hwc with " drm_hwcomposer: Add non-blocking commit support" cherry-picked, if that matters)

16:12 <robclark> hmm, ok so plane is enabled.. so I guess before jessica_24's patch you'd see the warn_on splat once or occasionally.. but now it causes the atomic-check to fail so we never recover from the bad state.. this would kinda paper over the problem and make things less spammy, I think. But not sure how we get into that state in the first place

16:12 <robclark> https://www.irccloud.com/pastebin/Yz08em7k/

17:37 <vknecht[m]> still getting one warn on panel back on : https://pastebin.com/dkCh6ae7