09:49erle: so i have some graphical corruption of X11
09:50erle: it happens sometimes when i access the gtk file selection from gajim
09:51erle: the interface locks up and the mouse cursor is garbled, even after restarting x11
09:53erle: https://mister-muffin.de/p/MwOY.txt
09:53erle: that is what dmesg has to say about it
09:54erle: lspci says card is: 01:00.0 VGA compatible controller: Advanced Micro Devices, Inc. [AMD/ATI] RV515/M54 [Mobility Radeon X1400]
09:54erle: anyone here had this issue before?
10:44Venemo: erle: it looks like you got a GPU hang, and it looks like that is a really old GPU, though I'm unsure if it uses the r300 or r600 driver. I recommend opening a bug report in mesa, although I wouldn't expect a quick resolution as most development effort these days go towards new GPUs
10:49erle: Venemo any idea how to reset the GPU properly?
10:49erle: like, cursor being normal
10:50Venemo: other than finding the reset button on your computer, I don't know
10:50erle: i have to reboot to get a proper non-garbled cursor
10:50erle: well, old … computers are fast. this is a thinkpad R60
10:50erle: can run debian, can watch youtube
10:51Venemo: no argument there
10:51erle: can play minetest (until they remove support for my hardware)
10:51Venemo: it's awesome that it still works
10:51Venemo: just I have 0 experience with it
10:51erle: i used to joke that old thinkpads like T60 last as long as russian tanks
10:51erle: but reality showed: a T60 can last far longer than a T64 hehehe
10:54erle: the only things that sometimes bite the dust are a) fan b) display backlight (needs new inverter)
10:54erle: Venemo i thought maybe there was a way to cleanly init hardware in linux or radeon drivers in general
10:54erle: like reload module with specific params
10:57Venemo: well you can try to unload and reload the module, but it's anyone's guess what really is happening
10:57Venemo: you can also try to detach and reattach the pcie bus
11:03Venemo: there was some way to do that from a terminal, but I never tried it myself
13:17erle: Venemo interesting, i did not know it was possible
13:17erle: any idea what the command is?
13:17Venemo: something in sysfs
17:23Venemo: agd5f: has there been any progress in the investigation of the SDMA issue?
17:24Venemo: agd5f: it wasn't really a priority before, but we want to eventually start using the SDMA in RADV, but I'm reluctant to enable that until the kernel itself can't use it without hanging it
17:56agd5f: Venemo, will check on the internal ticket
18:14pixelcluster: well that firmware issue has always been a priority for as long as we've known it :P
18:57Venemo: pixelcluster: I mean enabling the RADV transfer queue hasn't been a big priority
18:58Venemo: the problem is, if I enable it now, then we won't be able to tell the difference between whether an SDMA hang is caused by the kernel or RADV, so I hesitate to enable it until we know that kernel has issues with it
19:03agd5f: Venemo, well, the issue is that the doorbell seems to get missed in some cases, but if you update the wptr rather than the doorbell, missing the doorbell is not an issue, so the workaround should prevent that.
20:33Venemo: agd5f: is the workaround merged to recent kernels? and, can you say if other GPUs are also affected by the same issue?
20:34agd5f: Venemo, yes, it's been merged for a while. I don't think we've seen the same signature on any other asics yet
20:34Venemo: so what's up with the users who report that they still experience this problem?
20:35Venemo: and what's up with users who report SDMA hangs on other HW?
20:35agd5f: are there vangogh users that still experience the problem?
20:35agd5f: potentially different issues
20:36Venemo: according to issue 3440, yes there are
20:38Venemo: https://gitlab.freedesktop.org/drm/amd/-/issues/3440#note_2534834
20:39Venemo: he has full logs and everything at the link
20:44agd5f: I don't know
20:46agd5f: maybe it's rarer, but can still happen. They could try disabling CGPG or some of the other patches discussed in 3440 and see if that helps
21:39Venemo: well, they are a regular user who is tech savy enough to know how to use gitlab, but I'm not sure he can solve this on his own
21:40Venemo: I think some guidance would be appreciated as to what he should try