00:12 steel01[d]: Guess it'd help if I pulled the nvk ahb mr too.
00:47 gfxstrand[d]: Now that you've got everything rebased, I can work out the texture import issues.
00:49 steel01[d]: I've got a couple more things to push once the thrashing finishes.
00:49 gfxstrand[d]: 👍🏻
00:50 steel01[d]: gfxstrand[d]: Beyond the tegra format thing you linked yesterday and the nvk ahb mr, is there anything else pending floating that's relevant to this effort? Wanna make sure I've actually got all of it.
00:50 gfxstrand[d]: Nope. Not that I've written yet, anyway. 🙃
00:51 steel01[d]: Heh, yeah. That's the fun part.
01:00 gfxstrand[d]: steel01[d]: NVK is probably rejecting the 0x15 modifier
01:48 gfxstrand[d]: steel01[d]: Pull the latest minigbm. I haven't tested it yet but it should choose modifiers NVK doesn't mind.
01:50 gfxstrand[d]: I should probably also make NVK less picky but <a:shrug_anim:1096500513106841673>
01:50 steel01[d]: Make something work first, then make it more robust. I'll see what it does.
03:01 gfxstrand[d]: gfxstrand[d]: For 128x128, it should give you a 0x14 modifier
03:29 steel01[d]: Alright, that got the bootani to render with skiavk, the vulkan ui renderer. But the normal ui is failing, causing renderthread stack traces through the vulkan stack. There's a few instances of this error around, but it's not matching every crash loop.
03:29 steel01[d]: 10-26 02:19:23.228 507 507 E [minigbm:nouveau_bo_create_for_modifier(681)]: Allocating new BO: 48x48, fourcc = AB24, size = 12288 domain = 0xc, pte_kind = 0xfe, tile_mode = 0x0, modifier = 0x0
03:29 steel01[d]: 10-26 02:19:23.230 5413 5436 E [minigbm:nouveau_bo_map(782)]: DRM_NOUVEAU_GEM_INFO failed with No such file or directory
03:29 steel01[d]: 10-26 02:19:23.230 5413 5436 E system_server: Mapping failed.
03:29 steel01[d]: 10-26 02:19:23.230 5413 5436 W Gralloc4: lock(0xb4000024894642e0, ...) failed: 3
03:29 steel01[d]: 10-26 02:19:23.230 5413 5436 W Surface : failed locking buffer (handle = 0xb4000024894642e0)
03:29 steel01[d]: 10-26 02:19:23.230 5413 5436 E system_server: Error -38 locking sprite surface before drawing.
03:52 gfxstrand[d]: Oh, damn... I know what's happening. That's gonna be a little annoying. Probably not going to get that one fixed until Monday.
03:52 gfxstrand[d]: I wonder if there's somewhere in the BO that I can tuck the nouveau handle. 🤔
03:53 gfxstrand[d]: Or I could maybe map it from Tegra.
03:53 gfxstrand[d]: But the caching is going to be a mess
03:53 gfxstrand[d]: Ugh
03:56 gfxstrand[d]: Yeah, there's extra fields I can tuck it in. This shouldn't be that hard. Just annoying.
03:57 gfxstrand[d]: Or maybe I'll add some just for nouveau. 🤔
14:18 steel01[d]: 01-01 00:00:49.985 529 565 F libc : Fatal signal 7 (SIGBUS), code 1 (BUS_ADRALN), fault addr 0x27ee380008 in tid 565 (RenderEngine), pid 529 (surfaceflinger)
14:18 steel01[d]: Mmm. nvk+angle might be hitting drm/nouveau#452 now.
14:27 mangodev[d]: uh oh
14:28 mangodev[d]: fencing is borked maybe?
14:28 mangodev[d]: discord crashes in both electron and firefox when viewing animated things
14:29 mangodev[d]: https://cdn.discordapp.com/attachments/1034184951790305330/1432013325109035179/image.png?ex=68ff81d6&is=68fe3056&hm=63c361cfa657d18652218f9fcc25dbdf22a9adc3e6511182a5703d639b8977fa&
14:29 mangodev[d]: firefox actually gives a different error
14:30 mangodev[d]: https://cdn.discordapp.com/attachments/1034184951790305330/1432013520492433499/image.png?ex=68ff8204&is=68fe3084&hm=ae401b6305ad01c228cf28ab4f3b80ec6a5981383c79300b458fae9fd11bce36&
14:30 mangodev[d]: meanwhile electron:
14:33 mangodev[d]: mesa git from today, no custom patches
14:34 mangodev[d]: could be from latest `linux-firmware-nvidia`?
14:36 chikuwad[d]: doubt it, we're still on 570.144 GSP
14:37 chikuwad[d]: GSP hasn't changed
14:37 mangodev[d]: the downgrade did something
14:37 mangodev[d]: less cursor stutter
14:37 mangodev[d]: strange
14:38 mangodev[d]: gonna hope i don't crash again
14:38 mangodev[d]: if not, would i make an issue for mesa, or for the arch package?
14:39 mangodev[d]: wth
14:39 mangodev[d]: my cursor stutter wasn't from plasma 6.5?
14:39 mangodev[d]: :|
14:40 mangodev[d]: new nvidia firmware package update caused my cursor to stutter when switching windows (even in software cursor mode)
14:40 mangodev[d]: gone with a rollback to last month's version
14:46 mangodev[d]: wait
14:46 mangodev[d]: nevermind????
14:46 mangodev[d]: this is all confusing me
15:23 snowycoder[d]: chikuwad[d]: arch updated nvidia-open at `580.95.05`
15:29 chikuwad[d]: snowycoder[d]: that doesn't power nouveau though
15:30 chikuwad[d]: the GSP for nouveau is separate from the nvidia-provided kernel module
15:30 chikuwad[d]: well, it's the same GSP, but you know
16:29 gfxstrand[d]: steel01[d]: Untested but I pushed again
16:30 gfxstrand[d]: steel01[d]: Yeah, fixing that is top priority for Tegra.
16:35 gfxstrand[d]: Which doesn't mean I've actually looked at it or plan to for a bit.
17:17 gfxstrand[d]: Though I've thought about playing similar tricks to what I'm doing on minigbm to hack around it.
17:19 mangodev[d]: https://cdn.discordapp.com/attachments/1034184951790305330/1432056061270954095/image.png?ex=68ffa9a3&is=68fe5823&hm=fd64e5b46fad1fe0c00d699b463c5a5c2ff993ceaef5798b4eed8d9ed3325a22&
17:19 mangodev[d]: funky stuff going on right now
17:26 gfxstrand[d]: 502? I don't think I've seen that one before.
17:26 gfxstrand[d]: Probably doing maintenance over the weekend.
18:51 lassebq: I'm unable to create an issue at https://gitlab.freedesktop.org/drm/nouveau
18:51 lassebq: is something going on with freedesktop.org gitlab?
18:56 lassebq: Nvm I guess that could've been because I dumped part of log as part of the issue backticked code block
18:57 lassebq: https://gitlab.freedesktop.org/drm/nouveau/-/issues/456
18:58 lassebq: I encounter a similar freeze with firefox and also epiphany
18:58 lassebq: All with a different frequency. Firefox being the most probably to cause GPU hang
18:59 lassebq: probable*
23:12 steel01[d]: gfxstrand[d]: ```
23:12 steel01[d]: 01-01 00:00:51.820 656 656 E android.hardware.graphics.allocator@4.0-service.minigbm_nouveau: Failed to initialize driver.
23:12 steel01[d]: Something got unhappy. I'll see if I can trace why init started failing. It's currently not logging anything else.
23:14 steel01[d]: nvdev = nv_device_alloc(NV_DRIVER_TEGRA);
23:14 steel01[d]: if (nvdev != NULL)
23:14 steel01[d]: return -ENOMEM;
23:14 steel01[d]: Maybe this.
23:16 gfxstrand[d]: Yup. Fixed and pushed