07:46omeringen: @karolherbst hi, did you mention my issue(FS#72472) yesterday ? I see you said something about gf108. i wonder how it works on fedora, do they use a modified nouveau driver ?
08:06karolherbst: omeringen: not sure what you mena by FS#72472, but we did fix the nouveau regression on some Fermi GPUs
08:17omeringen: @karolherbst https://bugs.archlinux.org/task/72472 , interesting point is fedora works fine
08:21karolherbst: omeringen: ohh, I think we fixed this bug already
08:21omeringen: when ?
08:21karolherbst: fedora usually uses the latest or second latest kernel release
08:21karolherbst: omeringen: 1 month ago or so?
08:22karolherbst: 5.6 isn't supported upstream anymore
08:22karolherbst: so arch has to backport those fixes
08:22karolherbst: and I think 5.4 is a bit old for those patches.. let's see
08:23karolherbst: those three patches you'd need
08:23omeringen: so there is nothing special on fedora side, they just use upstream
08:23karolherbst: well.. the issue is a race condition in the kernel side
08:23karolherbst: so you might also just got lucky
08:24karolherbst: but I think the patches are in 5.14 already
08:24karolherbst: yeah, they are
08:24karolherbst: v5.14-rc7 even
08:24karolherbst: so I guess it's a bit older actually :D
08:25omeringen: there is 5.15.arch1-1 available, but you say that arch doesn't have that patch
08:25karolherbst: omeringen: it should be in all 5.14 and newer releases
08:25karolherbst: I just saw that you were testing with 5.6 and 5.4, correct?
08:26omeringen: linux-zen 5.14.15.zen1-1
08:26omeringen: linux-lqx 5.14.16.lqx1-1
08:26omeringen: i have also those , doesn't work
08:26karolherbst: then maybe you run into a different issue there
08:26omeringen: well, i created a new user and checked ,still same
08:26karolherbst: you said you had a gf108, right?
08:26karolherbst: ohh, do you have the logs from 5.14?
08:27karolherbst: we did had a regression with fermi starting with 5.14
08:27karolherbst: so you might have hit this one instead
08:27karolherbst: also breaks everything...
08:27karolherbst: I pushed the fix yesterday
08:28omeringen: 01:00.0 VGA compatible controller: NVIDIA Corporation GF108M [GeForce GT 525M] (rev a1)
08:28omeringen: Subsystem: Toshiba Corporation Device fc50
08:28omeringen: Kernel driver in use: nouveau
08:29karolherbst: but yeah.. I'd be interested in seeing the kernel log of 5.14, just to confirm you are hitting the issue I think you do or if it's something else
08:29omeringen: i am having this issue for at least one year :P
08:29omeringen: okay i'm gonna check 5.14
08:30karolherbst: omeringen: yeah... sorry about that, but we did some recent changes so fixes should arrive at users quicker now
08:30karolherbst: and regressions are painful :/
08:32omeringen: thanks for the fix, it's a nice coincidence to learn that it's already fixed, after i filled a bug :P
08:33omeringen: updating kernel now
08:35karolherbst: omeringen: yeah.. but that's actually quite common with those older kernels
09:06omeringen: @karolherbst failed again: linux 5.14.16.arch1-1 xf86-video-nouveau 1.0.17-1 kwin 5.23.2-1
09:07omeringen: gonna wait for 5.15
09:13karolherbst: omeringen: but do you have the log?
09:15omeringen: i will enable logging and boot again. rebooting...
09:44omeringen: @karolherbst https://dpaste.org/HUpQ also i have a folder about it https://drive.google.com/drive/folders/19kBULZ3OKNISdoFBkmCF66uqq4vLwVI5
09:45karolherbst: omeringen: yeah.. so it's the bug I pushed the fix for yesterday
09:45karolherbst: "Nov 04 12:17:15 arch kernel: nouveau 0000:01:00.0: ce: init failed, -22"
09:45omeringen: ah okay nice to know, thanks
09:45karolherbst: still a bit curious about the other errors though...
09:52karolherbst: omeringen: how difficult would it be for you to try out the patch?
09:53omeringen: i am just an ordinary n00b user :/
09:53karolherbst: omeringen: arch doesn't have a nice and easy way of patching the kernel?
09:54karolherbst: ehh.. doesn't seem like it
09:54omeringen: i have never tried, also compiling a kernel will take forever
09:55omeringen: but i want to help, what exactly do you want me to try ?
09:56karolherbst: omeringen: I am just wondering if the other issues I see are a consequence of the initial bug or if there is something else
09:58karolherbst: although probably.. dunno
10:18omeringen: there is a testing 5.15, going to check this now
10:19omeringen: could you check if your patch included in it ? -> https://github.com/archlinux/linux/commits/v5.15-arch1
10:20karolherbst: it's not
10:21karolherbst: sadly for whatever reason I never got a fermi GPU... mupuf btw, what's the status about sending some GPUs? I can also buy some if you are a bit busy atm
10:22mupuf: karolherbst: ah, let me go down to the basement
10:22mupuf: I was thinking I probably should send the lot of it, but first I should lay them down on the ground and catalog them
10:23karolherbst: mupuf: you also don't have to send all at once :D I mostly just need some fermi ones
10:24karolherbst: I've put my list here: https://gist.github.com/karolherbst/41a6613eaafd8ec0e3d53f408aa9bc37
10:26mupuf: karolherbst: yeah, sending all of this would be a little crazy: https://cloud.mupuf.org/s/NXA4Lokcr6BweTX
10:26karolherbst: so one or two PCIe cards pre nv40 and gf1 + a gf1 is probably enough for now
10:27karolherbst: I do have a quite a variety of GPUs already, just wanting to fill some gaps
10:27karolherbst: but yeah.. those are quite a lot :D
10:27omeringen: what about zen 5.15 kernel ? it might include your patches maybe -> https://archlinux.org/packages/testing/x86_64/linux-zen/
10:28karolherbst: omeringen: nope, I just pushed it into a drm tree yesterday
10:28omeringen: ah okay
10:28karolherbst: it will take a while until it reaches the kernel
10:28omeringen: i would send my laptop to you but i live in Turkey :P
10:28karolherbst: omeringen: not living _that_ far away :p
10:28karolherbst: but I guess custom rules are still a bit annoying shipping into the EU
10:29karolherbst: annyway, mupuf wants to send me some GPUs regardless
10:29karolherbst: so this will already help
10:31omeringen: okay. many thanks for your help :thumbs up:, time to sleep for me now. See ya.
10:34karolherbst: omeringen: good night!
10:46mupuf: karolherbst: that's the GPUs I have: https://cloud.mupuf.org/s/MRa3Xy8AGPnGFLq
10:46mupuf: let me compare to your list now
10:46mupuf: what will you want to run on them? Is it manual testing or automated testing?
10:47mupuf: I pretty much have every tesla, fermi, and kepler
10:48mupuf: then I have most gefore 7
10:48mupuf: for the nve4, I have two: one titan and the 690 (dual GPU IIRC)
10:49mupuf: looks like it is indeed fermi where you are lacking
10:52mupuf: for CI, you probably want the tk1 as it is self contained
10:55mupuf: I propose to send you these: nv35, 44, 50, c0, c4, cf, d9, e4, ea (Tegra K1)
10:57karolherbst: mupuf: sounds good. I do have a jetson nano myself, but that's... maxwell I think
10:57karolherbst: yep.. gm20b
10:57mupuf: karolherbst: so, what are you gonna do with these gpus? Because if you are planning on running piglit on it, I need to know
10:57mupuf: (some of my GPUs have damaged coolers)
10:58karolherbst: but yeah.. for now manual testing, because.. you know.. time and CI and stuff
10:58karolherbst: ohh.. damaged coolers?
10:58karolherbst: that sounds... bad
11:00mupuf: yeah, but I think all the ones I listed are in good shape
11:01karolherbst: mupuf: but I think you can skip the e4 in case it is already too much with the others
11:01mupuf:got a lot of GPUs as broken from ebay, and fixed them in a reflow over
11:01karolherbst: 44 would be good, because that's the different 3d class :)
11:01mupuf: I'm trying to pick the gpus that are the fastest of their gen, without being tooooooooo hot
11:01karolherbst: nv35 is pcie though, right?
11:02mupuf: only the nv1x is PCI
11:02mupuf: but this may be something you are still interested in
11:02mupuf: I did not want any AGP
11:02karolherbst: yeah... just I don't have any PCI slots
11:13karolherbst: mupuf: I am curious if there are some niche thingies around where you can connect a PCI or AGP GPU via TB
11:14karolherbst: and I still need to buy a label printer :D
11:18mupuf: yeah, get a dymo ASAP!
11:18mupuf: so freaking useful
11:18karolherbst: yeah... but I don't know which.. I actually wanted one with keyboard + internal battery, but those are like freaking expensive
11:19karolherbst: I refuse to touch anything cups related ever again :D
11:29mupuf: freaking expensive? Got one for ~50 euros
11:29karolherbst: mupuf: that's cheaper than I found
11:30karolherbst: which one do you have?
11:39mupuf: dymo letratag
11:39mupuf: it was on sale though
11:39karolherbst: although that needs 4 AA batteries
11:39mupuf: yep, looks like it
11:39mupuf: why do you care? you put them once and keep them forever
11:40mupuf: https://www.clasohlson.com/fi/Tarrakirjoitin-Dymo-Letra-Tag-LT-100H/p/Pr311846000 is selling for 40 euros now in not the cheapest store
11:40karolherbst: it has an integrated charger?
11:40mupuf: No. you are being unreasonable, Karol
11:40karolherbst: I know
11:40karolherbst: I was looking at ones with li-po included, because AA ones are just a hassle :p
11:41mupuf: So, the box I found was too big, so I threw in a couple more gpus
11:41mupuf: nvc1 and nva0
11:41karolherbst: but I also do have 4 AA rechargeables here and use them for nothing...
11:41mupuf: see, perfect!
11:41karolherbst: yeah.. I guess I will buy one
11:42karolherbst: seems like I can get one for 23€
11:42mupuf: sounds about right. As I said, clasohlson is not the cheapest store
11:43mupuf: get some rolls of "paper" along with it, you'll get through it fast if you get into labelling
11:43karolherbst: but I also need transparent rolls.. or wanted one
11:43karolherbst: yeah.. I don't have that much to label
11:44karolherbst: just wanted to label my shelfs
11:44karolherbst: well, the drawers on it actually
11:44mupuf: once you start, you won't stop :p
11:44karolherbst: I thought 16 drawers are "future proof"
11:45karolherbst: boi, how I was wrong
11:45karolherbst: *wrong I was
11:45mupuf: I bought it to label my nvidia gpus, and I ended up labelling a lot more... like all my tea collection
11:45karolherbst: but I will get another 4x4 kallax from my old apartment in january...
11:45karolherbst: my tea bags already have labels on them luckily
11:52mupuf: karolherbst: I'll send the parcel during the weekend
11:53karolherbst: awesome, thanks!
11:54mupuf: the nvc1 is not a nice gift, it is the gpu that has broken fan management
11:54mupuf: I asked nvidia for the doc, and I have not implemented what they sent
11:54mupuf: <brown bag>
11:54mupuf: but doing fixed point arithmetics in the kernel wasn't too appetizing
11:56karolherbst: I remember
11:56karolherbst: maybe I can get it then.. doesn't sound like a lot of work and I think we already have the lib for it, no?
15:10Lyude: karolherbst: sorry - got distracted by some other stuff, I'll start on it right now
15:10karolherbst: Lyude: do you think a CI pipeline report is more useful if done like this? https://gitlab.freedesktop.org/drm/nouveau/-/merge_requests/10#note_1147512
15:10karolherbst: Lyude: too late :D
15:11Lyude: karolherbst: ah ok lol, well thanks!
15:11karolherbst: well.. it was just adding brackets
15:11karolherbst: #define CRCC37D_FLIP_THRESHOLD CRCC37D_MAX_ENTRIES - 30
15:11karolherbst: #define CRCC37D_FLIP_THRESHOLD (CRCC37D_MAX_ENTRIES - 30)
15:12Lyude: also - maybe, tbh I think we should maybe just have some way to waive requiring a pass somewhere
15:12karolherbst: for... reasons
15:12karolherbst: Lyude: yeah... maybe
15:12karolherbst: I can just disable this requierment
15:12karolherbst: but I do want to rather disabling not critical stuff and keep a few ones we really don't want to miss
15:14karolherbst: not sure..
15:14karolherbst: I also want to enable ci-fairy
15:14karolherbst: and there are important things
15:14karolherbst: like the build testing
15:14karolherbst: don't want some compilation issue to slip through
15:15Lyude: could just block merge on errors and not block on warnings
15:15karolherbst: thta's what I am doing already
15:15Lyude: ahhh ok
15:15karolherbst: the first checkpatch stuff is the critical one failing CI
15:15karolherbst: the second is just a heads up
15:16karolherbst: you can see it here: https://gitlab.freedesktop.org/drm/nouveau/-/merge_requests/14
15:16karolherbst: but this was before I converted to the collapsable stuff
15:17karolherbst: checkpatch is just either super verbose or not all that useful :(
15:18karolherbst: but I think witht he collapsing stuff it's not that terrible.. oh well.. worst case we just patch checkpatch
15:31imirkin: i think the messages are nice to have. but should be informational rather than blocking.
15:31imirkin: (checkpatch messages, that is)
15:31imirkin: it's a very noisy tool, and hardly always right. but sometimes does point out obvious problems.
15:32karolherbst: imirkin: you can enable/disable certain checks
15:32imirkin: i think this applies to all checks.
15:32karolherbst: well.. I think requiring something like proper SPDX tags is I think valid enough
15:33karolherbst: or signed-off-by tags fromt he author
15:33karolherbst: but yeah.. I think most checks are actually just informational
15:34imirkin: ok, maybe 1 or 2 are ok
15:34imirkin: but absolute vast majority are basically for amusement purposes only
15:35karolherbst: there is actually also a complex macro one, which is quite useful though.. but not always correct :/
15:35karolherbst: or well..
15:35karolherbst: complains about headers from nvidia
15:36imirkin: like i said ... "for amusement purposes only"
15:36imirkin: not to say it's not useful
15:36imirkin: but it's not always actionable
15:36karolherbst: ohh, in this case it actually is and quite important so
15:37imirkin: sometimes it is
15:37imirkin: but like i said - not always - so can't be an error
15:37karolherbst: but yeah.. I think we should of course not block CI on any error, but just on a selected group of ones we are sure are almost always a valid concern
15:38karolherbst: yeah.. maybe that
15:53imirkin: the rest is definitely interesting to include
15:53imirkin: and the person operating the thing should use their discretion to determine what to care about / not.
17:39karolherbst: skeggsb: did you saw my comment on your MR with your old patches? https://gitlab.freedesktop.org/drm/nouveau/-/merge_requests/10/diffs?commit_id=df06ba4a9b3bd600749411ef786ba09dd878b1b6#55b095333ef8c361df45eea1300cc1854a3280b7_140_139