01:28 TimurTabi: Man, I wish I had a gentle intro to our hardware.
08:30 Santurysim: Do I understand correctly that all cards with the same codename are equivalent in terms of programming them?
08:31 karolherbst: Santurysim: more or less yes
08:31 karolherbst: there is always the vbios which describes differences
08:32 karolherbst: but from an algorithm point of view it's the same as you have to check the vbios the same way
08:32 karolherbst: and you can also have hw differences, like amount of shader cores etc...
08:36 Santurysim: karolherbst: thank you!
15:06 andres: Hi! When using sway on a optimus setup with IOMMU enabled I get a lot of violations pointing to nouveau. This does not occur with intel_iommu=igfx_off.
15:08 karolherbst: andres: interesting
15:13 andres: Here's an excerpt: https://gist.github.com/anarazel/a23ed5f461743f4c53f538833f46e235
15:14 andres: The way this primarily presents is that attaching an external display (which on this laptop is routed to the TU117 card) makes the whole system almost unusably slow...
15:15 andres: karolherbst: any specific information you'd want?
15:16 karolherbst: It is probably easily reproduceable, so not sure.. I am just no expert if it comes to the iommu stuff
15:17 karolherbst: but I think we just have to register a few memory regions or something..
15:22 andres: Do you know how to map the addresses from the DMAR faults to what they do in the nvidia gpu?
15:26 andres: FWIW, using an external dispaly with X using xrandr --setprovideroutputsource also hits the issue, although the performance hit isn't *quite* as severe.
16:26 karolherbst: my GPU doesn't like me today :(
17:05 Santurysim: GPU, shame on you!
17:05 mrnuke: karolherbst: I was going to ask if it's an nvidia, then I saw I'm on #nouveau
17:06 karolherbst: Santurysim: yeah.. I can't be sure if that's my fault or nouveau just having a bad day
17:06 karolherbst: but I guess it's my fault to run 110k tests at once
18:38 spaethnl: After updating my system packages I am getting frequent Xorg crashes. Xorg Backtrace looks like this: https://pastebin.com/r2UAT1Vu and journalctl shows: https://pastebin.com/xc4gkSLB
18:41 Santurysim: Is out-of-tree version of nouveau enough to develop it? I failed to build skeggsb kernel
18:42 spaethnl: I had the impression that the crashes were related to `nouveau_pushbuf_data`. They seemed more frequent earlier in the day before I closed many firefox tabs.
19:06 RSpliet: Just found this in my dmesg, after suspend from resume on GK107 https://hastebin.com/ciroledido.yaml
19:13 RSpliet: This is on kernel 5.12.12-300.fc34.x86_64 , so I guess there is the slight chance that any of Christian Koenig's gem patches may fix this...?
19:18 karolherbst: RSpliet: that's a race condition in the kernel
19:18 karolherbst: skeggsb_ has patches for it
19:25 RSpliet: karolherbst: cool. Did they make it into 5.13?
19:25 RSpliet: Feel like Ben's kept his head low for a while, no idea what's in the pipeline anymore :-)
19:26 karolherbst: I doubt it
19:27 RSpliet: Ok... well weirdly it hasn't caused nouveau to stop functioning at all. And s2r breaks my audio now too, so I guess this issue isn't "Imma build my own kernel" critical to me right now. Thanks for the heads-up
19:29 Santurysim: Do I need skeggsb's tree to develop nouveau?
19:46 karolherbst: Santurysim: no