On Fri, Nov 19, 2021 at 2:00 PM dmanye <dmanye@xxxxxxx> wrote: > > On 19/11/21 13:31, Karol Herbst wrote: > > yeah.. not quite sure yet. I tried it out with my gk208b gpus, but > > couldn't hit anything. Maybe using VGA makes the difference here. Or > > something else is different. Might also already be fixed in 5.16... > > too many unknowns still. I will do another round of testing with VGA > > and see how that ends. > > hello again, > > thanks for your effort Karol. i've just found another time the bug and: > > 1) kept the vga cable in the same place. > > 2) connected a dvi cable between the same computer and the same monitor. > > 3) told xrandr to dup the signal through the dvi cable: > > DISPLAY=:0.0 xrandr --output DVI-D-1 --mode "1920x1080" --same-as VGA-1 > > 4) voilà! the monitor, all alone, switched to dvi input and there is my > x session up and running. > > 5) if i force the monitor to change to vga input it refuses it and > returns to dvi > > so i think the driver does its job but without actually sending anything > through the vga cable. > yeah.. not quite sure yet what's going on here. Maybe the cable are just a bit broken, or.... other signals interfering or whatever. Maybe a driver bug. VGA being an analog signal and stuff makes it a bit painful to tell what's wrong. If dmesg is clean, then no idea how to dig into this even. I couldn't reproduce it with my Fedora system running the latest 5.10 kernel either with VGA. Could also be the displays being weird... Anyway, if using DVI is an option for you I highly recommend using that. But I'd also like to figure out what's wrong here. What I think could help is to enable drm debugging once you hit the error. For that you can write "0x1ff" into /sys/module/drm/paramters/debug as root, restart X, change the resolution via xrandr to force X to set a new mode on the connector and pastebin the output of dmesg somewhere. Maybe that helps, maybe not. Let's see what information that gives us. > > thanks. > >
