09:43tiredchiku[d]: https://cdn.discordapp.com/attachments/1034184951790305330/1261256649511796806/image.png?ex=66924c4a&is=6690faca&hm=359c2e96ea5dcfe0b9e8fd8204d695151dac3315fd3006520dc6738e93eb1a71&
09:43tiredchiku[d]: the trojan is ready:
09:43tiredchiku[d]: just waiting for my brother to come back from school so I can borrow his laptop to verify it works on GA107
09:44tiredchiku[d]: (it's an installation running off of a USB that I can boot from a UEFI boot menu)
09:45asdqueerfromeu[d]: tiredchiku[d]: How many computers has it infected? /s
09:45tiredchiku[d]: so far, 1
09:45tiredchiku[d]: my own
09:47tiredchiku[d]: tiredchiku[d]: I'm doing this only because he won't let me install linux on his laptop and I need an environment I can continuously modify easily to test multiple things on both my Turing (TU116M) and his Ampere (GA107)
09:49redsheep[d]: I will borrow your hacker wizard magic when I get the time so I can verify if dp audio worked on 535
09:49tiredchiku[d]: I'm confident the patch I applied will get it working on the current kernel too
09:49tiredchiku[d]: might swap out the lts kernel on this USB for current stable
09:50redsheep[d]: If that's the exact same gsp firmware nouveau loads and if it works then it's 100% a nouveau issue that the dp audio is busted, at least for some cards which would be good info
09:51redsheep[d]: I think it's pretty likely it won't work though which is probably better news on balance, don't need to debug, just get newer better gsp
09:52tiredchiku[d]: it is the exact same gsp firmware
09:52tiredchiku[d]: because we use their python script to extract it
09:52tiredchiku[d]: https://github.com/NVIDIA/open-gpu-kernel-modules/blob/main/nouveau/extract-firmware-nouveau.py
09:53tiredchiku[d]: `# Converts OpenRM binhex-encoded images to Nouveau-compatible binary blobs. See nouveau_firmware_layout.ods for documentation on the file format`
09:55redsheep[d]: Ah, knew version to version was the same firmware code wise but IIUC that conversion means it wouldn't hash the same, interesting
09:55tiredchiku[d]: mhm
09:55tiredchiku[d]: also, I like XFCE
09:56tiredchiku[d]: dkms built properly for kernel 6.9.8
09:56tiredchiku[d]: :yay:
09:56clangcat[d]: tiredchiku[d]: Me to I miss a lot of the X DE environments
09:56tiredchiku[d]: yeah, I'm using xfce on this USB because it's got a small install size and xorg works equally well on all hardware
09:57clangcat[d]: tiredchiku[d]: Yea fair. I mean is the one advantage X wms have over wayland.
09:57tiredchiku[d]: mhm
09:58clangcat[d]: They all use the Xserver as their input/output backend management. So there's no mismatch support for different hardware.
09:58tiredchiku[d]: yup
09:58clangcat[d]: Where as like wayland will depend if you use GNOME, COSMIC, KWIN, Wlroots, etc...
09:58tiredchiku[d]: mmhm
10:03clangcat[d]: tiredchiku[d]: Though I mean on a technicality.
10:04clangcat[d]: There are other Xserver implementations.
10:04clangcat[d]: Other than the one host by FDO so I mean technically same concept kinda applies to those other implementations.
10:04tiredchiku[d]: mhm
10:11tiredchiku[d]: yeah, 535 breaks on 6.9 for other reasons
10:11tiredchiku[d]: but I just need to check if GA107 works, so even LTS is fine
10:39tiredchiku[d]: https://cdn.discordapp.com/attachments/1034184951790305330/1261270708734918746/image.png?ex=66925962&is=669107e2&hm=f723910735ae34a6ee5c22d71686419a5363e1b1726b087c6295364ae24c2665&
10:39tiredchiku[d]: karolherbst[d]: clangcat[d]
10:39tiredchiku[d]: GSP Firmware is fine on openrm 535
10:39tiredchiku[d]: on GA107
10:40clangcat[d]: tiredchiku[d]: Hmmm must be something to with my exact GA107 then.
10:40clangcat[d]: :/
10:40tiredchiku[d]: no, because someone else also reported the issue you had
10:40tiredchiku[d]: it was demonkingofsalvation I think
10:40karolherbst[d]: it's still can be an issue with thoes two GA107
10:41tiredchiku[d]: I'll check nouveau, hang on
10:41karolherbst[d]: yeah
10:41karolherbst[d]: anyway, with nvidia you are best of considering that every GPU is unique and special 😄
10:41tiredchiku[d]: I don't need the nouveau.config thing for Ampere, right?
10:41karolherbst[d]: I think GSP is by default now?
10:42karolherbst[d]: or uhm..
10:42tiredchiku[d]: just making sure
10:42karolherbst[d]: I think it depends on the kernel config
10:42tiredchiku[d]: arch linux, hm
10:42clangcat[d]: karolherbst[d]: Yea you can disable it in the kernel config.
10:42clangcat[d]: I only know cause my current kernel is built with it disabled
10:42karolherbst[d]: `DRM_NOUVEAU_GSP_DEFAULT`
10:43karolherbst[d]: which is `n` by default unless distributions overwrite it
10:43clangcat[d]: tiredchiku[d]: Oh Nvidia prop has always worked for me
10:43clangcat[d]: It's just Nouveau that doesn't
10:43tiredchiku[d]: mhm
10:43tiredchiku[d]: I wanted to check if the specific GSP version nouveau uses was fine on nv prop
10:43tiredchiku[d]: and it is
10:44tiredchiku[d]: anyway, reboot time, brb
10:44clangcat[d]: tiredchiku[d]: Yea I could've told you that :p
10:44clangcat[d]: Nvidia prop drivers have been working perfectly for me. Including 535 when it was new.
10:45tiredchiku[d]: oh well 😅
10:45tiredchiku[d]: I mean, it wasn't much effort to verify it myself too c:
10:45asdqueerfromeu[d]: tiredchiku[d]: How about audio?
10:46clangcat[d]: tiredchiku[d]: Yea fair. would be interesting to see if Nouveau fails. But even then it might just be my and the other person's card.
10:46tiredchiku[d]: what audio
10:46clangcat[d]: And cards like ours.
10:46clangcat[d]: tiredchiku[d]: Probably HDMI audio thing they mean
10:46tiredchiku[d]: DP audio? redsheep[d] is going to check in his morning
10:46tiredchiku[d]: I don't have a DP display
10:47asdqueerfromeu[d]: tiredchiku[d]: Yes
10:49tiredchiku[d]: so I modprobed nouveau post boot cuz I forgot to remove the blacklist nvidia package puts in
10:49tiredchiku[d]: and: https://paste.fuelrats.com/xucigewifi.yaml
10:50clangcat[d]: tiredchiku[d]: I mean seems like it worked aside from
10:50clangcat[d]: [ 66.912870] nouveau 0000:01:00.0: [drm] Cannot find any crtc or sizes
10:50clangcat[d]: [ 66.917892] nouveau 0000:01:00.0: [drm] Cannot find any crtc or sizes
10:50clangcat[d]: [ 66.918064] nouveau 0000:01:00.0: [drm] Cannot find any crtc or sizes
10:50tiredchiku[d]: mhm
10:50tiredchiku[d]: vkcube on nvk also working fine
10:51tiredchiku[d]: :thonk:
10:52tiredchiku[d]: let me try removing the blacklist and booting with nouveau at init
10:53clangcat[d]: tiredchiku[d]: Possibly yea :p
10:54clangcat[d]: tiredchiku[d]: what kernel tho?
10:54tiredchiku[d]: 6.9.8
10:54tiredchiku[d]: arch linux default
10:55clangcat[d]: tiredchiku[d]: Mmmmm does arch apply any patches to their default kernel?
10:55tiredchiku[d]: yes, a couple I think
10:56clangcat[d]: Do you know which ones? trying to find them
10:56clangcat[d]: cause I could try building the kernel with arch patches
10:57tiredchiku[d]: yea just a sec
10:58clangcat[d]: Can't wait for the new RAM to arrive
10:58clangcat[d]: then I can finally turn GSP back on XD
11:00tiredchiku[d]: https://github.com/archlinux/linux/releases/tag/v6.9.9-arch1
11:00tiredchiku[d]: they have a patch file per kernel version
11:02clangcat[d]: https://cdn.discordapp.com/attachments/1034184951790305330/1261276451223634000/grafik.png?ex=66925ebb&is=66910d3b&hm=df8a533799544a52fdaf0eda6191e54f260d3ac3b00f1a15d77b7a5fcdb9334a&
11:02clangcat[d]: tiredchiku[d]:
11:02clangcat[d]: Started looking
11:02clangcat[d]: and I see this
11:02tiredchiku[d]: lol
11:02tiredchiku[d]: Hurr durr I'ma ninja sloth is upstream
11:03tiredchiku[d]: and torvalds just doesn't wanna change it xD
11:03clangcat[d]: Ahhhh
11:03clangcat[d]: Neat
11:04clangcat[d]: Anyway there isn't anything that stands out but will try when new ram gets here
11:04clangcat[d]: https://cdn.discordapp.com/attachments/1034184951790305330/1261276961146142750/grafik.png?ex=66925f34&is=66910db4&hm=202314be8036a92152f0951ddb022ad3dc38e08f424832231195dd260525b779&
11:04clangcat[d]: Cause I am not compiling with this
11:05tiredchiku[d]: removed the blacklist
11:05tiredchiku[d]: and added nouveau.config=NvGspRm=1 for good measure
11:05tiredchiku[d]: let's see what happens now
11:07tiredchiku[d]: still works fine
11:07tiredchiku[d]: clangcat[d]: demonkingofsalvation I'd like to know your distros... I know cat's is void linux
11:10tiredchiku[d]: hmm
11:11tiredchiku[d]: I wonder if I need to put some GPU load on this
11:11tiredchiku[d]: or if it's a userspace bug
11:11pixelcluster[d]: tiredchiku[d]: torvalds did that in the first place :D
11:11tiredchiku[d]: since I'm on mesa 24.1.3
11:11tiredchiku[d]: pixelcluster[d]: mh
11:11tiredchiku[d]: I should enable chaotic aur and grab mesa-tkg-git
11:12tiredchiku[d]: but the fucking keyserver is down?
11:13tiredchiku[d]: no nvm
11:14clangcat[d]: tiredchiku[d]: It's not userspace
11:14clangcat[d]: It happens on GPU init for me.
11:15tiredchiku[d]: still wouldn't hurt for me to move to mesa-git
11:15tiredchiku[d]: 😅
11:15tiredchiku[d]: or try to put some GPU load on this that's more than vkcube..
11:15clangcat[d]: tiredchiku[d]: It happens even when I don't have mesa installed :p
11:15tiredchiku[d]: yes darling I remember
11:16tiredchiku[d]: it happens for you even when headless, right?
11:17clangcat[d]: Yup
11:17tiredchiku[d]: funky..
11:18tiredchiku[d]: what's your rootfs filesystem?
11:19clangcat[d]: ext4
11:19clangcat[d]: was XFS for while
11:19clangcat[d]: but I swaped it hoping it would fix it but no
11:19tiredchiku[d]: ok so it's not filesystem dependent then
11:19tiredchiku[d]: cuz I'm on xfs right now
11:21tiredchiku[d]: ...hm
11:29tiredchiku[d]: okay, I'll look into these logs later
11:30tiredchiku[d]: brother needs his laptop back
11:30clangcat[d]: That's fine love
11:31tiredchiku[d]: extracted the logs, and they're on the USB
11:32tiredchiku[d]: so I can just boot it on my laptop and continue working
12:15tiredchiku[d]: clangcat[d]: do you have a log from one of your nouveau crashes?
12:16tiredchiku[d]: I'm spotting something funny
12:16clangcat[d]: tiredchiku[d]: Not rn one sec will install an unpatched kernel
12:19tiredchiku[d]: okay~
12:20clangcat[d]: https://cdn.discordapp.com/attachments/1034184951790305330/1261296083808686110/grafik.png?ex=66927104&is=66911f84&hm=07cb527ef8e136e8a5a73bcbd2241c105e68c6c9299cdbc15e2a20cbab75df1f&
12:20clangcat[d]: YOOOOOOO
12:20clangcat[d]: Nouveau seg faulted
12:20clangcat[d]: modprobe XD
12:21tiredchiku[d]: amazing
12:21clangcat[d]: tiredchiku[d]: https://paste.fuelrats.com/aqijecapun.yaml
12:21clangcat[d]: Nouveau is at the bottom
12:22tiredchiku[d]: thanks <3
12:22clangcat[d]: But yea just died in firmware ctors
12:22clangcat[d]: as per usual
12:22tiredchiku[d]: oh
12:23tiredchiku[d]: :notlikethishisui:
12:23tiredchiku[d]: you and the other GA107 user have different errors
12:23tiredchiku[d]: okay
12:23tiredchiku[d]: I'll poke at yours first
12:23clangcat[d]: tiredchiku[d]: Yes we do
12:23clangcat[d]: :3
12:23clangcat[d]: Mine crashes on boot
12:23clangcat[d]: and locks the entire kernel down
12:23tiredchiku[d]: gotcha
12:23clangcat[d]: Hence me starting it with modprobe
12:23tiredchiku[d]: but what about with that patch you were using
12:23marysaka[d]: the error looks very weird
12:24clangcat[d]: tiredchiku[d]: That allows me to actually boot the kernel and have the Nouveau card run.
12:24tiredchiku[d]: it's that SG_DEBUG error I think
12:24tiredchiku[d]: clangcat[d]: oh, the card works fully with the patch?
12:24clangcat[d]: marysaka[d]: Yea it is it's cause Nouveau has a NULL pointer in the ctors code and SG_DEBUG get's upsetting
12:24clangcat[d]: tiredchiku[d]: No
12:24clangcat[d]: Just doesn't crash
12:25tiredchiku[d]: so then, gimme logs for that too :D
12:25clangcat[d]: tiredchiku[d]: There aren't any really
12:25clangcat[d]: sway and other EGL/VK apps just segfault cause of NULL pointer derefence/bad alloc in the case of EGL
12:26clangcat[d]: And I just have sway set up to only open the AMDGPU
12:27clangcat[d]: marysaka[d]: Well it's litterally just because some function for Nouveau loading firmware ends up returning NULL and so the kernels SG_DEBUG catches it and throws a fuss. As for the userspace errors when I disable `SG_DEBUG`/Use Lyude's patch idk they might be related or might not be
12:27clangcat[d]: at this point I just use prop day to day.
12:28clangcat[d]: Especially cause all I've ever been told is to "wait for Nova" and "Nova will be better"
12:28clangcat[d]: and I just don't care at this point
12:44tiredchiku[d]: clangcat[d]: oh
12:44tiredchiku[d]: hmm, okay
13:34jfalempe: ahuillet: you mentioned that you may have some reference code/documentation regarding the blocklinear tiling layout.
13:36tiredchiku[d]: ahuillet
13:36tiredchiku[d]: (pinging so you don't miss it)
14:17demonkingofsalvation[d]: tiredchiku[d]: I use arch
14:17demonkingofsalvation[d]: tiredchiku[d]: yeah it was me
14:26demonkingofsalvation[d]: demonkingofsalvation[d]: also using hyprland if that matters...
14:36tiredchiku[d]: doesn't matter right now, no
22:29gfxstrand[d]: karolherbst[d]: Any idea why Maxwell doesn't like `shf.high`? I'm getting an illegal instruction encoding
22:30gfxstrand[d]: Or maybe it's just `shf.l.high`?
22:31dadschoorse[d]: what does .high do?
22:31gfxstrand[d]: I'm not sure, actually
22:32gfxstrand[d]: I really need to fuzz the shit out of some of these things and add a rust implementation for documentation
22:34gfxstrand[d]: I think it might just be for 32-bit shifts and tells me to take the high or the low
22:35gfxstrand[d]: Actually, I think it gives you the high part of the 64-bit shift
22:36gfxstrand[d]: The PTX docs are not, unfortunately, the hardware docs in this case