03:18 gnurou: imirkin_: devinit scripts should definitely set bit 1 of 0x2240c, at least according to the doc I have
03:28 imirkin: gnurou: well, half the GF104 scripts we have don't do it (at least not that i found), and even one GK107 one
03:28 imirkin: we don't exactly have an enormous collection either, so i dunno how common it is
03:29 imirkin: now, in theory, running devinit a second time *shouldnt* hurt... but ...
03:31 gnurou: imirkin_: yeah I saw that on the bug report, this is quite confusing - the doc I have is pretty clear about this bit
03:32 gnurou: imirkin_: IIUC the regression is precisely caused by running devinit twice, isn't it?
03:32 imirkin: yes
03:33 imirkin: at least as far as i understood
03:33 imirkin: if you have any GF104's lying around the office, you could give it a shot
03:34 imirkin: dunno how deep your pile-o-hardware goes
03:34 gnurou: not very deep sadly, I'm a Tegra guy in a remote office...
03:34 imirkin: i see
03:38 imirkin: skeggsb: gotten anywhere with the GTX 970 4GB thing?
03:38 gnurou: in any case this is pretty annoying - I will try to ping someone about this
03:38 imirkin: [or he might be traveling already]
03:38 imirkin: gnurou: well, i still think that running devinit a second time shouldn't break anything
03:39 imirkin: and for one of the GF104 vbios's in our repo, we have a mmiotrace which doesn't seem to indicate like the presumably-bad script is getting run at all in the first place
03:39 imirkin: so i need to do a bit more sleuthing
03:40 gnurou: in this case it looks like it does though... it may be useful if the reporter just reverted that precise commit on top of 4.6.4 to see if that improves things
03:40 imirkin: well i'm sure it will
03:41 gnurou: or try booting with NvForcePost on 4.5.1 and see if things break as we would expect them to
03:43 imirkin: very interesting.
03:43 imirkin: on the mmiotrace i'm looking at
03:43 imirkin: it very clearly does *NOT* run the vbios
03:43 imirkin: [0] 637.271550 MMIO32 R 0x02240c 0x00000000 PUNITS.UNK00C => 0
03:43 imirkin: so... there's some other condition in place
03:43 imirkin: which lets it figure out that the vbios doesn't need running
04:00 gnurou: uh
04:04 imirkin: so just looking at stuff adjacent to that read, it reads out
04:04 imirkin: [0] 637.272153 MMIO32 R 0x619488 0x00000000 PDISPLAY.VGA.CR+0x88 => 0
04:05 imirkin: [0] 637.270788 MMIO32 R 0x619f04 0x003ffe09 PDISPLAY.VGA.ROM_WINDOW => { BASE = 0x3ffe0000 | ENABLE | TARGET = VRAM }
04:05 imirkin: [could look at the fact that the window's already enabled]
04:06 imirkin: also the PMC.ENABLE state...
05:50 pmoreau: (imirkin: FYI, since he has already landed, he was already travelling. :-) )
06:37 pmoreau: Let’s try to add GL_ARB_spirv…
07:56 pmoreau: Meh, I spent half an hour trying to understand why glapi_gen would not recognise GL_SHADER_BINARY_FORMAT_SPIR_V_ARB, only to realise I had forgotten to close a tag in the XML specification…
07:59 pmoreau: Apparently it's still broken… --"
08:19 mupuf: pmoreau: how is it going? :D
08:20 pmoreau: Better thanks to chrisf
08:20 pmoreau: But now I need to pack everything to get going :-D
08:20 mupuf: oh, cool!
08:21 pmoreau: Bus is in <25min, need to pack everything
08:24 mupuf: pmoreau: stop checking IRC!
08:28 pmoreau: :-D
08:44 karolherbst: meh and I could be at the airport already :(
08:56 mupuf: karolherbst: better to be home than learning at the airport that your flight is canceled
09:00 hakzsam: mupuf, are you with ben already?
09:01 RSpliet: hakzsam: we're here about to munch some pasta
09:05 karolherbst: mupuf: true
09:06 karolherbst: mupuf: well the airport is like 25 minutes away from the center of hamburg, so it ain't that bad
09:12 gnurou: mupuf: I am running "./piglit run -v gpu results/gpu" under X on Jetson TK1, no freeze - and it still has the suspicious patch that karolherbst wants to revert. I guess I an running it wrong and tests are done in software?
09:13 hakzsam: RSpliet, okay, enjoy :)
09:14 mupuf: gnurou: ok, I will update the kernel and see if it helps, OK?
09:14 mupuf: I am stuck on a 4.5
10:07 karolherbst: gnurou: piglit uses GLX usually
10:07 karolherbst: gnurou: run shader-db directly
10:07 karolherbst: the issue only appears with EGL for me
10:38 karolherbst: mhh, I could get 500MB data for 7 days for 5€, is it cheaper to get a finnish prepaid card for the week?
10:39 karolherbst: well I don't care anywaym just the limited data is shit
20:55 idl0r: imirkin: hm, either it still needs to be reviewed or something went wrong with my answer re patchwork
20:56 imirkin_: no, it went through
20:56 imirkin_: at least i got it
20:56 idl0r: hm
20:56 imirkin_: maybe it's not in patchwork, but nobody looks at that anyways
20:56 imirkin_: i mean it's nice to reference patches, but people aren't obsessingly checking anything in there
20:56 idl0r: :D
22:19 pauloalvarez: hello guys
22:20 pauloalvarez: my computer sometimes hangs
22:20 pauloalvarez: and I would like help to confirm if this is an issue with nouveau
22:20 pauloalvarez: I have some messages in dmesg about multiple warp errors
22:20 pauloalvarez: can somebody give a help, please?
22:21 imirkin: pastebin messages?
22:21 imirkin: [if you're using some fancy desktop, chances are that yes - the issue is with nouveau.]
22:22 pauloalvarez: just wait a minute please
22:24 pauloalvarez: http://pastebin.com/sVLPCwQM
22:24 pauloalvarez: i have a single monitor desktop
22:24 Hussam: linux 4.4 also has the hibernate issue. nouveau tries to push vram to physical ram instead of swap on hibernate.
22:24 pauloalvarez: not suer if it's fancy
22:24 Hussam: I should file a kernel bug
22:25 imirkin: GPR_OUT_OF_BOUNDS? interesting. that shouldn't happen.
22:25 imirkin: pauloalvarez: what version of mesa?
22:25 imirkin: and what GPU
22:25 pauloalvarez: wait a minute
22:26 pauloalvarez: GK208 [GeForce GT 640 Rev. 2]
22:27 pauloalvarez: mesa version: 12.0.1
22:29 imirkin: hmmmm
22:29 imirkin: well i don't see anything that could fix it in 12.0.3
22:29 imirkin: do those messages coincide with the hangs?
22:29 imirkin: or are they unrelated?
22:29 pauloalvarez: imirkin, not always
22:29 pauloalvarez: sometimes i get flickers on screen
22:29 pauloalvarez: and then the dmesg messages appear
22:30 pauloalvarez: usually happens with chromium
22:30 pauloalvarez: i then have to restart chromium
22:30 pauloalvarez: but the system does not hang
22:30 imirkin: yeah, that sounds like a different issue
22:31 pauloalvarez: but then the flicker happens again
22:31 pauloalvarez: and the system hangs
22:31 imirkin: hm
22:31 imirkin: sorry, doesn't sound familiar =/
22:31 pauloalvarez: ok
22:31 pauloalvarez: zsnes is one application where it usually just hangs
22:33 pauloalvarez: imirkin, no problem, thanks for trying to help
22:34 Hussam: imirkin, not sure I used the correct bug component but https://bugzilla.kernel.org/show_bug.cgi?id=172381
22:37 imirkin: Hussam: correct bug component, wrong bug tracker
22:38 Hussam: but isn't the kernel module in the linux kernel source tree?
22:38 imirkin: Hussam: if i could tell you with a straight face that someone was going to look into that bug, i'd get you to re-file it at bugs.freedesktop.org
22:38 imirkin: but ... i can't, so no need to bother
22:38 Hussam: ok, I see.
22:39 imirkin: i'm basically the only one who looks/responds to these bugs, and i haven't the faintest clue about anything related to suspend-to-disk
22:40 Hussam: I understand :)
22:40 Hussam: I am curious if it happens with non NVIDIA GPUs.
22:41 imirkin: could be that it's hooked up wrong in nouveau
22:41 Hussam: yes.
22:41 imirkin: you're probably the first and only person to have ever tested that
22:41 Hussam: heh
22:42 imirkin: i dunno. maybe it's part of the things skeggsb tests
22:49 pauloalvarez: imirkin, I have some dmesg messages from when the computer hang just now
22:49 pauloalvarez: want to see them?
22:50 imirkin: sure
22:51 pauloalvarez: wait a minute
22:54 pauloalvarez: http://pastebin.com/FPZ2upHe
22:54 pauloalvarez: there you go
22:55 imirkin: yeah that's much worse :)
22:55 pauloalvarez: hahahahahaha
22:55 imirkin: anything before the first one of those?
22:56 pauloalvarez: there was a message about MULTIPLE_WARP_ERROS, GPR_OUT_OF_BOUNDS
22:59 imirkin: right before?
22:59 pauloalvarez: no
22:59 imirkin: i guess bad things could happen if it executes random instructions
22:59 imirkin: could hang the pgraph firmware perhaps? a breakpoint of sorts? hm...
22:59 imirkin: but then we'd get a breakpoint interrupt...
23:02 pauloalvarez: well, I have configured and compiled my own kernel
23:02 pauloalvarez: could I have missed something?
23:02 imirkin: nope
23:02 imirkin: nouveau's just not very battle-tested. you hit a slightly unexpected condition, and it kinda goes boom
23:03 imirkin: you could try using blob pgraph firmware, perhaps it'd fare better
23:03 imirkin: but i don't remember if my extractor script retrieves GK208 fw or not
23:04 pauloalvarez: hmmmm, so maybe my card is not supported?
23:04 imirkin: i haven't updated it in quite a while
23:05 imirkin: supported is a pretty relative term
23:05 imirkin: it's as supported as all the other cards
23:05 pauloalvarez: ok, sounds fair to me
23:05 imirkin: if you want real support, get an AMD gpu
23:06 Hussam: I miss the 3dfx days in the 90s.
23:06 Hussam: those were massive graphics cards
23:07 pauloalvarez: understood
23:08 pauloalvarez: i have another computer with an nvidia card (much older though) and it worked fine
23:08 pauloalvarez: thought maybe this one would work as well
23:08 pauloalvarez: but well, limitations, nouveau probably is not a trivial job
23:09 imirkin: pauloalvarez: chances are there's something much more coincidental going on
23:10 imirkin: fwiw i also use a GK208 as my main gpu
23:10 imirkin: and haven't really had these ctxsw timeout issues
23:10 pauloalvarez: hummm, what could it be imirkin?
23:10 pauloalvarez: I see
23:12 pauloalvarez: that's strange
23:12 pauloalvarez: I may be able to reproduce a hang in a deterministic manner using wine + warcraft 3
23:13 imirkin: my strategy tends to be to avoid doing things that are likely to cause hangs :)
23:21 pauloalvarez: guess I just got unlucky then
23:21 pauloalvarez: maybe I can use the intel driver
23:43 pauloalvarez: imirkin, do you use kde?
23:44 imirkin: oh god no
23:44 imirkin: i don't go in for any of this new-fangled hotness
23:45 imirkin: i have the same setup i did 15+ years ago