00:15 ismael: weirdly enough it restores sometimes :S
02:17 ninupuntos: I find a better woman being actually there for me as well, instead of this every trashing and cheating human garbage laura tornado, let this trash fuck their bums on their own locations. Our hotel is not sponsoring nor hosting humiliating courtesan fuckgangster wannabes from motogangs be those lgbt enthusiasts or not also we do not allow angry analfuck crocodile to enter our areas with their
02:17 ninupuntos: abuse. about those triplets they function already but i saw a small mistake 112-32+112 aka 80+112=192, you simply run second time with adding the number to the hash and the reverse sign subtract gotten.this will yield a remainder as subtract from original. now as you put 37+57 to the hash actually 57 is gotten, or if nothing but 118 you get -37.
02:50 ismael: here's the kernel log with debug=disp=debug: https://bpa.st/7YDQ
03:24 airlied: anarsoul: https://lore.kernel.org/dri-devel/20250618032244.487604-1-airlied@gmail.com/T/#u I think is probably my last go
06:59 tsunamiokimi: And her diagnosis from mental insitution is not fraud like in my case. It is entirely mentally and physically ill bnorn human just like dave , ryan and dog are. When i see such people such need to be smacked into face and next wave of my alliance just murders you in a war. And that is kinda it. You have no clue who are you dealing with besides that i have continuously did all the work for
06:59 tsunamiokimi: you, i am also very powerful inhabitant on this planet.
07:38 inhousedemo: Like a little retarded bitch like her you get slapped onto faces by me when i see you near my territory again, next thing over you get murdered by my pals. It's just it has taken some time due to investigation made as to what happened to my neck, the last injury for life gotten, this year jack dedman is held responsible for this with an execution that happens to very many who were involved
07:38 inhousedemo: in that assault and harassment overall in cambodia.
07:46 HdkR: 🎉
08:11 andnotastowhat: HTTP3 as well as websocket protocols added to it, is not known they offer backward compatible HTTP2 sockets too to get onto telling truth to other humans on this planet. If i was in charge of computing war3 was never about to take off, as it is all likely wrong thing to commit the wars and it is behind the wrong thinking as sideffect. If you learn to think properly which nearly all
08:11 andnotastowhat: Russians are capable of doing , cause they are smart people in general, than wars might as well could be skipped. But another real source o troubles just launched after this mexican hard working society riots, Israel attacked Iran. I am saying that all of those wars are due to wrong thinking and could be blamed on people like you just lying to general public in total terror kind of ways.
08:11 andnotastowhat: It's not like i have problems with my demo compiler, and it's samw wise not like i take over any of your work, what you had in video decoding was offered by me, what you have on opencl was offered and researched by me, i've been contributing a lot, in considerable amount all the time unlike you who push stirred saliva to the channel from your entirely crooked faces. And that is what
08:11 andnotastowhat: europe has crooks, Russians are unlike this.
08:24 lru: where is the best place to report kernel level errors with nouveau?
08:28 matt_schwartz[d]: https://gitlab.freedesktop.org/drm/nouveau/-/issues
08:35 lru: thanks!
08:52 lru: oh good,there's already an issue for this
09:06 optiononezero: In three years time all those european crooks doing fraud over intelligent and strong people making his life a living embarrassing hell are going to be found from various countries and slaughtered in stylish ways, all this wank spam is ended with very classic brutality, so that such human shit no longer tells their stirred saliva about mental illnesses in public while consuming victims
09:06 optiononezero: subtsances. I have no issues of finding their partners in crime who started that terror from estonia, their funeral was also their birth cause they are same stupid human shit abortion leftovers like on this channel already mentioned set.
09:07 HdkR: IPv6, how fancy.
09:10 lru: https://gitlab.freedesktop.org/drm/nouveau/-/issues/131 if anyone cares :-)
09:30 lru: it's funny, this bug is there from the early days, when people would say "nouveau doesn't support X, so if you need it, you have to use the proprietary drivers".... now that the proprietary drivers are not supported in main distros anymore due to age, people are switching back to nouveau, this time with features working, except for this old bug lol
09:31 lru: i.e. nouveau handles acceleration now...and I just tried a round of flightgear on nouveau last night which worked great for a few minutes until a segfault in nouveau_drv.so (likely a different bug)... so things are quite promising, except for this stickler
09:32 karolherbst: yeah...
09:32 karolherbst: can only do so much with the limited people we have, especially for older hardware
09:32 karolherbst: and that sort of older hardware is also starting to die physically
09:33 lru: mine's still going :-)
09:33 karolherbst: got a few nv4* cards which I think are broken now... (the generation of GPUs before yours)
09:33 lru: ah
09:33 karolherbst: so might be close 🙃
09:33 karolherbst: but I think I also have a broken g88 one... or something
09:34 karolherbst: g84 actually
09:34 lru: that's probably in the same family
09:34 karolherbst: I have two identical g84 and one is toast :'(
09:34 lru: as long as it runs nvkm/engine/gr/g84.c:g84_gr_tlb_flush() you're in the game :-)
09:35 karolherbst: yeah... but I can think of better things to spend my time on than debugging an issue which might happen once a week given that my main system doesn't even have a PCIe slot
09:36 lru: well, I would appreciate even just sharing any technical knowledge you have, and ideas for debugging, because every time this bites me, I'm more motivated to fix it
09:36 lru: and I can't quite afford a new laptop right now
09:37 nicepredormant: By then Russian federation starts unrpeceded invasion towards such nations of terrorcrooks. I was dealing with EFI as well as coreboot only few, had some sources to spawn efi bios from, and used a efi emulation to boot up one machine, from usb, that was not accessible due to bootloader corruption. However i was aware of romboot being there for early hardware initialization anciently
09:37 nicepredormant: enough part of microsofts code. Which in fact is more like little kernel for android. As we spoke about interracial anal and terror slut in our areas, she was already handled proportionally, if she gets lost one more time to wrong territory pushing fraud and terror down the pipeline she gets murdered with all of her cranks and lesbians, so far she was just handed to real gangsters who
09:37 nicepredormant: punshed all of this shitbags teeth out and marked her like her other fuckpal terrorists with a rabbitlip, this is how you know scammers like this outside. Same thing will happen to ryan, i will spal his shitty face and gangsters do other treatment alike, if he has more complaints on me and effort of scam committed in excess again, he gets murdered as said.
09:37 lru: the link that was shared here, pointing me to envytools, has been of great help
09:38 lru: I'm working my way through reading it, starting to understand
09:39 lru: when g84_gr_tlb_flush() calls spin_lock_irqsave(), does that guarantee that there will be no further jobs / work added to the PGRAPH engine until unlock?
09:49 karolherbst: no idea :) I'm not really familiar with this part of the code
09:49 lru: no worries
09:50 lru: that's where I am at the moment.... I think the GPU is either just slow and needs more than 2 seconds timeout to complete, or some other jobs are getting booked, perhaps after the timeout check just passed
09:52 lru: I'm pretty sure I can't just ignore TLB flushes, since they probably happen at some major change point....and ifthe flush fails, and the board is still using old TLB caches,I can see why it might hang
09:53 lru: oh,btw, what is ROP?
09:55 karolherbst: ROP is the rasterizer
09:56 karolherbst: but yeah.. the driver has a few issues if userspace overloads the GPU
10:01 lru: can you think of any other kinds of race conditions that would happen more often with: SMT on, kernel build in background, mplayer movie in foreground?
11:41 ismael: it seems vga_switcheroo powers down my discrete GPU and then it doesn't come back up
11:41 ismael: (until I use setfb)
11:42 ismael: when I switch to it as the main GPU it isn't powered down and the screen remains on
12:01 karolherbst: ismael: I'm a little confused on what's the issue here
12:02 karolherbst: the GPU needs to be used in some way before it gets powered back on
12:21 Guest18341: As i said there is nothing to count bits for, you can freely choose whether you place 192 37 or 81 on the hash at runtime or any combinations of those and you need to work on your brainworm until it's too late for you, i see those filthy English bragging, they are also one of the biggest amoral dog shit/scum and stupidest garbage in life overall. ismael you are not going to succeed in any of
12:21 Guest18341: the things in any form talking to mentally ill people like karol is or many others from #llvm #debian or here. They would had had to be killed many times ago, but European crooks love other terrorcrooks that's the real big deal as to why such dominate with their abuse. It does not even know what setfb. I delt with people like you many years in a row in real life also, Charl Vorster, Alex
12:21 Guest18341: Enrico sif, their anal ace Laura Tornado and a big number of her crooks like jack dedman and the list would go on, they would never figure out their mental debiliation or retardation. Until they had to be handled.
13:37 karolherbst[d]: created an MR for my vec phi stuff: https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35604 I'm still unsure how to properly merge the PhiWebs stuff and there is a regression where it asserts in RA
14:42 ismael: karolherbst: nevermind, once the screen goes standby the terminal doesn't come up again
14:43 ismael: karolherbst: there's fbcon attached, the screen restores fine on i915 but on the nouveau outputs it doesn't
14:44 ismael: on Wayland or X everything is fine, it's just on the terminal that it doesn't work
14:44 ismael: but why does it work with i915
14:44 ismael: ?
14:45 karolherbst: mhhh good question. Could be that the firmware can't handle it?
14:45 karolherbst: the system framebuffer is provided on a firmware level and that's the one usually used with fbcon
14:46 karolherbst: so runtime suspend/resume might just mess it up
14:46 ismael: the failing ACPI stuff happens at suspend, not afterwards
14:46 ismael: hmmm
14:46 ismael: $ cat /proc/fb
14:46 ismael: 0 i915drmfb
14:46 ismael: 1 nouveaudrmfb
14:46 karolherbst: normally the framebuffer is set up by the uefi firmware running some very very very basic GPU driver to put the device into a proper state
14:46 ismael: I would expect efifb there if the fb was coming from the firmware...
14:46 ismael: right?
14:47 karolherbst: mhhh
14:47 ismael: so it's definitely nouveau's fault, also there seems to be nothing going on after the screen goes standby and I press a key
14:48 ismael: like if there's something missing that on Wayland is being handled by the compositor
14:48 karolherbst: yeah not saying it's not nouveaus fault, it's just not a code path tested very much
14:48 karolherbst: especially as some distros disabled fbcon
14:49 ismael: I need some help with options for debugging nouveau, it's probably something trivial, but there's so much info
14:53 karolherbst: I don't think that debugging nouveau directly will help you figure this out. It's probably just that restoring the framebuffer isn't set up correctly and something needs to be added to the resume case. There is `nouveau.debug` to print some things tho
16:15 zmike[d]: can someone test zink with zmike/zink-general2 on cts case `dEQP-GLES31.functional.copy_image.non_compressed.viewclass_128_bits.rgba32i_rgba32ui.renderbuffer_to_texture3d`
16:16 zmike[d]: I don't have nv hardware plugged
16:55 anarsoul: airlied: the latest patch works for me
16:56 anarsoul: feel free to add (or not) my t-b tag
18:25 ismael: karolherbst: it does restore fine with i915, that's the thing
18:26 karolherbst: the intel GPU isn't powered doen ever
18:26 karolherbst: *down
18:26 karolherbst: the nvidia GPU gets its entire power supply cut off
18:26 ismael: even when selected by vga_switcheroo?
18:27 karolherbst: yes
18:27 ismael: hmmm, okay, it would make sense then
18:27 karolherbst: like it's all done on a firmware level
18:27 karolherbst: and there are ACPI methods to power down the entire PCIe bridge the GPU is connected to
18:28 karolherbst: so when the GPU gets used again, it basically starts from a cold boot state and needs to be brought up fully again
18:28 karolherbst: I suspect that just some framebuffer initialization code isn't ran properly or something, but yeah, the nouveau driver needs to tear down everything
18:29 karolherbst: there isn't much the driver can keep around, VRAM content stored in RAM, and some firmware files in order to not load them against from the fs
18:29 ismael: could it be that it isn't even trying?
18:31 ismael: I'm not seeing anything with nouveau.debug=disp=debug
18:31 ismael: I see the teardown, but nothing afterwards until I call fbset
18:32 karolherbst: mhhhhh
18:32 karolherbst: could be
18:32 karolherbst: like with runpm you need an active user to trigger the GPU to get turned on
18:32 karolherbst: soooo.. if nothing requests that then yeah.. the GPU would stay turned off
18:33 karolherbst: does "lspci" also help?
18:33 ismael: but there's an active user, it's just the screen goes into standby and then doesn't come back, calling fbset over ssh does reset the screen correctly and I can see the session as I left it
18:34 karolherbst: "active user" as in currently using the device through the drm render node or something... it might be there are more places to hook up to mark the GPU as active...
18:34 ismael: ah, so fbcon might be lacking there
18:34 asdqueerfromeu[d]: karolherbst: For some reason plugging in a display doesn't wake up the GPU (only `vulkaninfo` does)
18:35 karolherbst: like `pm_runtime_mark_last_busy`
18:35 karolherbst: asdqueerfromeu[d]: that's something the firmware is responsible for
18:35 karolherbst: there is an ACPI level event the firmware sent to the CPU
18:35 karolherbst: *sends
18:35 karolherbst: could be that this isn't working properly on some laptops Lyude might know more
18:36 karolherbst: but yeah.. once the GPU is turned off it's not nouveaus responsibility to deal with hotplug events (though it might register something so the ACPI events are recognized correctly?)