08:56 karolherbst: magic_rb[d]: your statement is kinda confusing, because none of the GPUs where the old pstate interface worked are supported by GSP
08:57 magic_rb[d]: Well exactly my point
08:58 magic_rb[d]: I just tried to force it back up, because it got stuck in d3hot after resume. But the pstate interface not working is entirely unsurprising. Wanted to confirm its actually not working and its not just my laptop being weird
08:58 karolherbst[d]: but d3hot has nothing to do with pstate?
08:59 magic_rb[d]: Does it not?
08:59 karolherbst[d]: d states are a PCI level tihng
08:59 karolherbst[d]: where pstates are nvidia specific performance levels, as in core/memory clocks
08:59 magic_rb[d]: Well yeah, but generally id expect that to be somewhat linked
08:59 karolherbst[d]: not really
08:59 magic_rb[d]: But yeah youve a point
08:59 karolherbst[d]: d3 just means the PCI device is in a low power state as in "disabled"
09:00 karolherbst[d]: like a TV on standby
09:00 magic_rb[d]: Can i then force dstates up?
09:00 karolherbst[d]: kinda
09:00 karolherbst[d]: d3hot/d3cold states are set by the kernel when the device is runtime suspended
09:00 magic_rb[d]: Because it definitely shouldnt be d3hot while im trying to play a game, (sidenote, my laptop doesnt do S3, only s2idle
09:00 karolherbst[d]: there is a `power/` node in the sysfs device
09:01 karolherbst[d]: `/sys/bus/pci/devices/0000:01:00.0/power/runtime_status` should state "suspended"
09:01 karolherbst[d]: control should be "auto"
09:01 karolherbst[d]: you could write "on" to `control` to force the kernel to put it into d0 state
09:01 karolherbst[d]:but
09:02 karolherbst[d]: if the device is stuck it could also be some firmware stuff going wrong (tm)
09:02 karolherbst[d]: this is all very low level and usually managed by the pci part of the kernel
09:02 karolherbst[d]: and there is also ACPI involved
09:02 magic_rb[d]: I did check the locks, it didnt spew any errors, kernel side or gsp side, but yes my first thought was also that the gsp firmware doesnt quite like s0ix
09:04 magic_rb[d]: \*logs
09:05 magic_rb[d]: Sorry just woke up, like 20 mins ago
09:09 magic_rb[d]: Ill try to force the bus back up with the power sysfs node next time i gets stuck, thanks karol
10:43 f_: Morning all! I just hit into some issues with Fermi again. My second monitor is unresponsive now.
10:44 f_: It had a slight glitch where it became green-tinted, then came back, and now it's just stuck. I have that display connected via DP.
10:45 f_: Will send dmesg shortly.
10:45 karolherbst[d]: magic_rb[d]: mhhh yeah.. `sOix` is kinda... well.. poorly tested, but some modern laptops default to it. However, there are some chipsets we technically need to apply quirks on in order to make it work. The nvidia driver has quite a lot of those.
10:45 f_: I also enabled sshd just in case my laptop's internal screen dies in a similar way.
10:45 magic_rb[d]: for me its not about defaulting, my laptop does NOT support s3 lol :)
10:45 karolherbst[d]: 💀
10:45 magic_rb[d]: it works reasonably well,
10:46 karolherbst[d]: yeah, it's not a bad idea, just uhm...
10:46 magic_rb[d]: it sometimes gets stuck and i have to reboot, but hey
10:46 karolherbst[d]: s0ix is tough
10:46 karolherbst[d]: yeah...
10:46 karolherbst[d]: it relies on runtime power management to actually work 🙂
10:46 karolherbst[d]: which normally does, but sometimes random things happen
10:46 f_: ring a bell? :p been a long while since I didn't encounter that issue
10:47 f_: running kernel 6.10.3
10:47 f_: logs:
10:47 magic_rb[d]: im a bit behind on mesa and kernel, since i also use zfs
10:47 karolherbst: f_: yeah, but the issue with issues happening like once a month are, that they are impossible to debug 🙃
10:47 magic_rb[d]: so im on 6.6 and mesa stable
10:47 karolherbst[d]: ah yeah.. newer kernel might fix random things
10:48 f_: karolherbst: Yep.. though I suspect it's coming from that same source
10:48 f_: that is, nouveau being unable to bring the GPU back into a known state (as I understand it)
10:49 karolherbst: f_: also, you didn't post any logs
10:50 f_: as I said they're coming soon
10:50 karolherbst[d]: ahh
10:52 f_: karolherbst[d]: https://bouncy.vitali64.duckdns.org/uploads/funderscore/e7dc54f4-nouveau-dmesg-fail.log
10:53 f_: I also still get those usual `script needs OR link` and `training failed` errors when exiting suspend.
10:53 f_: well
10:54 f_: "usual"
10:54 f_: (I get those everytime laptop comes back from suspend)
10:55 f_: The Joys of Nvidia & Fermi...
10:56 karolherbst[d]: mhhh, yeah, this is the display block was wrongly configured
10:57 karolherbst[d]: I don't know how to read those 🙃
10:57 f_: What display block?
10:57 karolherbst: Lyude, skeggsb9778[d]: ^^ any idea on what might be wrong there?
10:57 karolherbst: f_: of the GPU
10:58 f_: right
10:58 karolherbst: anyway, somebody familiar with the area might be able to tell what's going wrong there
11:01 f_: Ping me if you need anything then ^^
11:03 f_[x]: okay I unplugged the DP cable and my laptop's internal display also died
11:03 f_[x]: good thing I enabled ssh shortly before...
11:04 f_[x]: Will send more logs shortly.
11:12 f_: https://bouncy.vitali64.duckdns.org/uploads/funderscore/01634634-nouveau-fail.log
11:12 f_: Now I get kernel oopses.
11:13 f_: (Plugged back in DP, by the way)
11:14 karolherbst: ah yeah, the driver just times out
11:14 f_: so the GPU is dead? :^)
11:15 f_: Wrong timings or something?
11:15 karolherbst: no, the GPU works alright, just the display block is in a broken state so it doesn't do anything
11:16 f_: didn't mean dead dead, but rather nouveau asks it to do things and it doesn't do it
11:16 karolherbst: it's more like nouveau asked for the impossible and the GPU rightfully is upset
11:16 f_: very much so
11:17 f_: If that matters, the second display is a 3440x1440@165Hz display, but I only do 60Hz because apparently the GPU (or nouveau) doesn't like 165.
11:18 f_: I think I remember before nouveau was upset at 3440x1440 resolution but that seems to have gotten fixed a while ago.
11:18 karolherbst: mhhhh
11:18 karolherbst: it _might_ matter
11:20 f_: yeah before then getting that second display working at 3440x1440 was slightly painful
11:20 f_: and I'd notice slight lag .. or sometimes the monitor not liking whatever the GPU was outputting.
11:21 karolherbst: yeah, can certainly be something being wrongly configured due to that, maybe some undefined values even make it to the GPU so it only sometimes doesn't work? oh well...
11:22 f_: *didn't
11:22 f_: Been a long while since I didn't get those issues.
11:22 f_: Now the resolution is properly recognised
11:23 f_: But could be that it still has a few remaining issues about it...
11:23 f_: (I think it coincided with me moving to wayland)
11:30 f_: Before I potentially reboot - do you need anything else?
11:38 karolherbst: not really
11:39 f_: karolherbst: Gotcha. Rebooting.