01:55mlankhorst: voxadam: so just backport that patch then? :P
02:16voxadam: mlankhorst: Hey, I would but like I said earlier, the Nouveau codebase isn't exactly my jam.
02:18mlankhorst: it's the same as passing PCE1=0 though
02:25voxadam: Maybe I'll look at it tomorrow when I have some time.
02:55Calinou: hi, how do I disable V-Sync
02:55Calinou: it's forced :(
05:33AlbertP: just a short question, does anybody know which nvidia cards support UHD monitors, is that all Kepler and later hw?
06:31pmoreau: Let's go for some Apple GMUX <-> Nouveau relationship study!
07:03RSpliet: pmoreau: they need councelling?
08:15mlankhorst: oh, I'm trying to make present work for gpu screens :P
08:15mlankhorst: must get glxgears vblanked!
08:22pmoreau: RSpliet: Well, apparently it's already ok
08:22pmoreau: RSpliet: You didn't reply for the trace. Should I just brute-force something of my own? :D
08:24pmoreau: Is there an advantage at powering down the card with apple-gmux on top of powering it down with Nouveau?
08:25RSpliet: pmoreau: you asked me a question? whoops
08:25pmoreau: Even twice the same! :p
08:25RSpliet: anyway, just make it clock up and down a little
08:26RSpliet: make sure to capture "boot"
08:26RSpliet: yeah, make sure you capture the start of X.org
08:27RSpliet: loading of the nvidia driver, it could initialise some regs in PFB
08:27pmoreau: Ah, right
08:27pmoreau: Well, usual stuff
08:27RSpliet: and transitions to every clock freq...boot brings it up, let it go down, bring it back up, down once more and you should be complete
08:27RSpliet: (just play with nvidia-settings)
08:27RSpliet: I doubt there's many regs left to RE
08:28RSpliet: rather, I want to figure out which ones to scrap :-D
08:31Karlton: Okay, I made a 30 second apitrace of a game that has artifacts in it. Now how would I extract useful information from that?
08:31Karlton: just replay the trace?
08:34Karlton: "couldn't open libtxc_dxtn.so, software DXTn compression/decompression unavailable" could that be an issue?
08:36tobijk: Karlton: i asked you yesterday about that :-)
08:38Karlton: tobijk: yeah I remember you asked something about compression, but then I got busy with other stuff
08:38Karlton: is it a known issue?
08:38tobijk: its the extension for the compression is missing per default, its protected by a patent :/
08:39tobijk: causes many apps to have rendering issues
08:39Karlton: bloody patents!
08:41Karlton: okay and besides that error, there is the one: "4358595: warning: unsupported glXDestroyWindow call"
08:42tobijk: huh which app are you running? :O
08:42Karlton: terasology, a free game made with LWJGL
08:42Karlton: just replaying the apitrace
08:43tobijk: ah maybe that is creating its own windows and does not handle the glX calls, but im weak on that :/
08:45Karlton: okay, I had tried the game with an older card and the artifacts went away but the sky was black. So I must still be missing something
08:45tobijk: you installed teh s3tc lib?
08:46tobijk: they are in the wild after all
08:46tobijk: and free alternatives (s2tc)
08:46Karlton: I don't know what that is
08:47tobijk: look for libtxcn or s2tc
08:47tobijk: in your packagemanager
08:48tobijk: "libtxc_dxtn" thats the name, may fault
08:49Karlton: I am using gentoo
08:49Karlton: okay, I see it
08:52Karlton: it removed the compression error but still have the same issues with the graphics
08:53tobijk: mh i hoped for a now fixed sky
08:53Karlton: the was only for the old card
08:53Karlton: I got my kepler one installed atm
08:54Karlton: it renders the sky but there are a ton of artifacts and the ground is purple
08:54tobijk: so imirkin is right we mess up something :/
08:57tobijk: Karlton: is the trace available somewhere?
08:58Karlton: I haven't uploaded anywhere yet
08:59tobijk: i'd like to run the trace on my kepler as well and see what happens
09:00Karlton: where is a good place to upload it to?
09:00tobijk: first xz -9 it please :)
09:00Karlton: 30 seconds uncompressed is 100m+
09:03Karlton: that reduced it to "50M"
09:03tobijk: at least
09:03tobijk: can you sens it to firstname.lastname@example.org ?
09:04Karlton: I can try
09:30Karlton: It should be attached in the email
09:35Karlton: I have artifacts in flightgear also, so I made a trace and got this error:
09:35Karlton: "130706: message: major shader compiler issue 1: 0:9(12): warning: extension `GL_ATI_shader_texture_lod' unsupported in fragment shader"
09:36Karlton: only happens when I turn shader quality up and I do not have this issue when I use my older telsa card
09:40Karlton: and besides that there is this error: "2438347 @0 glDeleteTextures(n = 1, textures = &18) │2438347: warning: no current context
09:41Karlton: and another similar to it
09:43tobijk: the error happens only with the higher shaders?
09:44tobijk: where you get the warning about the GL_ATI_shader_texture_lod?
09:46Karlton: the artifacts disappear when I lower the shader quality
09:46Karlton: I'll try again with them down to see if I get the same error
09:50Karlton: yeah I don't get the shader compiler error when I turn shader quality all the way down
09:51tobijk: it resolves the error about glDeleteTextures as well i guess?
09:51Karlton: no, those still exist
09:53Karlton: oh and my email appeared to exceed the size limit, lol
09:54Karlton: it didn't send
09:54Karlton: I'll try to upload it elsewhere
10:21Karlton: try this link: https://www.sendspace.com/file/b0vlxy
10:23Karlton: file should be named java.trace.xz
10:29smekras: um... hi., could someone help me?
10:36tobijk: smekras: i can try, whats your problem?
10:38smekras: many, some might argue, but i am here for the nouveau related ones. i've gone through the faq, etc, but i am having trouble configuring nouveau the way it's supposed to be
10:39tobijk: which part of it, there is the kernel module, the xorg driver, mesa...
10:40smekras: i am not sure, to be honest. switching to glamor would be a good step though, i think
10:43tobijk: try to describe what you want to do and the outcome of it
10:47smekras: give me a few and i'll try to explain as best as i can
10:48tobijk: take your time
10:52tobijk: Karlton: running the trace i can reproduce the problem with my nvidia card, with the intel cards its allright
10:54smekras: on a sidenote, not sure how much it complicates things but i have a gtx295 (nv50/nvA0) and using two monitors, currently on kernel 4.0.0.rc3 with xorg 1.17.1 and mesa 10.6
10:54Karlton: tobijk: yeah I ran the game on my chromebook and it it ran just fine aswell
11:03Karlton: "imirkin> Karlton: fyi that terasology game appears to work fine... i didn't do much, but i started a new game... things displayed"
11:04Karlton: he was using gf108 fermi card
11:05nightfuri: Hi guys.. i find a little tearing in the texts when i enter in irc.(bottom the screen always).any settings should i look in to to fix it in the nvidia x server settings or any other settings ?
11:05tobijk: mh somewhere in our codepath we mess up something
11:06tobijk: nightfuri: this is the channel about the free nouveau driver not the nivida closed souce driver if you have problems with that, i'm afraid you need to go and ask nvidia for support
11:06nightfuri: tobijk: am sorry .. thank you
11:18smekras: wb tobijk
11:20smekras: as to what i am trying to do, i am sort of trying to switch to dri3 and glamor, but whenever i edited xorg.conf to do so, it doesn't even start X
11:20tobijk: so dri3 and glamor are two things
11:20tobijk: tackle one after the other
11:20tobijk: but you should not use glamor for now with nouveau
11:21smekras: it doesn't work?
11:21tobijk: it got some problems with the mesa driver
11:22tobijk: you got problems with exa?
11:22smekras: ok, so that's out the window
11:23smekras: not sure what the problem is really. it almost seems as if half the system knows i have 3d accel and the other half doesn't. it does not help that rawhide practically switched to wayland
11:24tobijk: im away for a while, ttyl
11:47pmoreau: RSpliet: Tracing done and xz'ed, just need to send it. :)
11:47RSpliet: could you add a note stating whether it's DDR2 or GDDR3?
11:47RSpliet: technically, I could figure it out from the trace
11:48RSpliet: but it's more convenient if you could tell me :p
11:48pmoreau: Such a lazy boy! :p
11:48RSpliet: I beg to disagree
11:49pmoreau: GDDR3 it is
11:49pmoreau: But I'll add a note in the mail
11:49RSpliet: great, thanks!
11:50RSpliet: you can also just push it to the nvbios repo if you prefer
11:50pmoreau: Might be easier :)
11:50RSpliet: absolutely :-)
11:50pmoreau: 8.1Mo, should be ok
11:51RSpliet: that's MB in English :p
11:51RSpliet: (does 'o' stem from "octo" somehow?)
11:51pmoreau: Right, it's the end of the week-end and I'm tired. :D
11:52pmoreau: It's for "octet", and yeah, comes from octo :)
11:54pmoreau: RSpliet: Pushed, and the note is in the commit message :D
12:29Apteryx: Hello, I'm a bit confused looking at the http://nouveau.freedesktop.org/wiki/FeatureMatrix/ page, and then the http://nouveau.freedesktop.org/wiki/VideoAcceleration/ one.
12:29Apteryx: For an old 7200GS card, it says the video decoding accel is "done" in the feature matrix.
12:31Apteryx: (under the NV40 column). But when I check the video engine support status, nothing there looks good for this card.
12:32Apteryx: except maybe XvMC support for MPEG1/2 (mostly).
12:33mlankhorst: erm xvmc is useless :p
12:34Apteryx: I'm a bit uneducated in the video format world. is MPEG1/2 still relevant on the web today?
12:35Apteryx: And which video players in the linux ecosystem can take advantage of XvMC?
12:36RSpliet: hardly... it's the only supported format for DVDs, and that's about it
12:36Apteryx: I see. Mostly the web has moved to h.264, I guess?
12:37Apteryx: *Most of the
12:37smekras: wow... i'm amazed. someone has an older card than i do
12:37RSpliet: or VP9, here and there a little Theora
12:38Apteryx: smekras: :D (my parents' PC actually) ^^
12:42imirkin: Apteryx: you have MPEG1/2 support with xvmc
12:42imirkin: contrary to what everyone says, xvmc is a *great* api for mpeg1/2, way better than vdpau for those cards
12:43imirkin: your card doesn't support accel for any other codecs though
12:44Apteryx: imirkin: I see. I guess that's why the support is deemed "done". There's nothing else to support ;)
12:44imirkin: vdpau support would be nice... it _ought_ to work, but it hangs the card hard for reasons unknown
12:44imirkin: but even then it'd still just be mpeg1/2 under vdpau
12:45mlankhorst: imirkin: but the api is terrible and requires manual configuration..
12:45imirkin: mlankhorst: yeah. but the actual api matches what the hw does
12:46imirkin: from a distro standpoint libXvMCW is horrid
12:46imirkin: but tbh i don't really care about the distro viewpoint :)
12:46mlankhorst: from a user standpoint too
12:46imirkin: set it up once and forget
12:46imirkin: i think that's a nice way to go
12:50mlankhorst: I dislike the setting up part, especially sinceo n different hw you would need diffferent configs
12:52tobijk: imirkin: did you run terasology somehow special yesterday when you tried it?
12:53imirkin: no, just downloaded the trace, will have a look at it
12:53smekras: wb tobijk
12:53imirkin: smekras: why are you trying to use glamor and dri3?
12:53imirkin: neither of those things work particularly well with nouveau
12:54smekras: i was told otherwise.. apparently they were wrong
12:54mlankhorst: imirkin: me needs dri3 for the present stuff :P
12:54smekras: what about wayland? does that work?
12:54imirkin: smekras: yeah, should be fine
12:55imirkin: smekras: not sure what you mean...
12:55imirkin: just run "wayland"? :)
12:55tobijk: smekras: i fear you have set your gl wrong
12:55tobijk: can you paste dmesg xorg.0.log and glxinfo output?
12:56smekras: i just left it at the default setup for rawhide
12:56smekras: should use pastebin or some such for it, right?
12:56tobijk: yep thats fine
12:58smekras: glxinfo output: http://pastebin.com/hpBn4Zuy
12:59imirkin: seems fine... what problem are you having again?
12:59smekras: wayland doesn't work and rawhide has switched to it, more or less
12:59imirkin: what does "doesn't work" mean?
13:00smekras: doesn't start at all
13:00imirkin: what error messages does it give?
13:00imirkin: or what does "doesn't start at all" mean?
13:02tobijk: imirkin: for the trace, i think its program 592 to look at...
13:02imirkin: tobijk: that trace renders just fine for me
13:02imirkin: tobijk: can you confirm it looks all purple and otherwise wrong on kepler?
13:03tobijk: yep its purple on kepler and allright with intel
13:03imirkin: well phooey
13:03tobijk: *starting with frame 320
13:04imirkin: can you run it with a debug build and do NV50_PROG_OPTIMIZE=0
13:04smekras: gdm is supposed to start in wayland by default now. it tries, falling back to x. trying to start a gnome wayland session throws me back to gdm. even in the x session of gnome, clutter is set to egl and half the apps won't even run. and for some reason, ever since mesa 10.6 wine forgot i have 3d accell
13:04tobijk: sure, sec
13:04imirkin: tobijk: also, are there any debug prints that complain of imminent death?
13:04tobijk: not that i have seen any, i'll recheck
13:05imirkin: smekras: presumably they log some messages in the process of falling back
13:06smekras: i would imagine so, but i have no clue where to find the logs
13:06imirkin: smekras: yeah, dunno. modern linux distros do a really good job of hiding it all.
13:06smekras: even xorg.0.log is not where i usually found it
13:06imirkin: i'm sure it's all been taken over by The One True Process
13:06imirkin: (aka systemd)
13:06smekras: ...yeah, most likely
13:07smekras: gotta love systemd, huh?
13:07imirkin: not something i have to worry about :)
13:07smekras: used to use gentoo
13:07smekras: i still type emerge sometimes...
13:08tobijk: imirkin: its purple without the optimizations as well
13:09imirkin: i guess that's comforting
13:09imirkin: although only in a limited way
13:09pmoreau: imirkin: I bet the X logs being in a different place is due to X not running as root any longer, rather than systemd
13:09tobijk: so something special to the nve+ path?
13:09imirkin: pmoreau: and why doesn't it run as root anymore?
13:10smekras: that could be it, actually
13:10imirkin: tobijk: yeah.... the only thing that comes to mind is texturing being all different
13:10imirkin: but if that were broken, i doubt this would be the lone application to hit the issue
13:10pmoreau: imirkin: Apparently thanks to systemd-logind! :D
13:10pmoreau: Ok, forget about what I said
13:11imirkin: pmoreau: ;)
13:11tobijk: i saw something similar with the new civilization, just with the color black
13:11tobijk: the right colors sparling through sometimes
13:12imirkin: btw, just to confirm -- you have libtxc_dxtn installed, right?
13:12imirkin: none of this s2tc crap?
13:12tobijk: i have s2tc
13:12tobijk: lib is called the same...
13:13imirkin: 0x60000000 WARP_CSTACK_SIZE = 1610612736
13:13imirkin: _somehow_ i think that might be a bit off
13:15imirkin: (looking at the GM204 trace)
13:15imirkin: good news is that they kept the same ISA
13:15imirkin: (as GM107)
13:15imirkin: although i'm SURE they took this minor revision as an opportunity to yet again reorder texture instruction arguments
13:15imirkin: (no, i'm not bitter _at all_)
13:16mlankhorst: why not :-D
13:18imirkin: i guess the good news is that once we can get pgraph to come up on there, most of the work for GM107 should be reusable
13:28mlankhorst: imirkin: but dont you already know the isa through cuda?
13:29imirkin: mlankhorst: well i wanted to see a trace to see what it was for real
13:29imirkin: mlankhorst: looks like it's the same as GM107
13:29imirkin: i.e. SM50
13:29imirkin: and yes, i probably could have used cuda tools to achieve the same conclusion
13:30mlankhorst: bleh my gf has a disk with errors >:(
13:32tobijk: imirkin: same color with the "real" libtxc_dxtn
13:32tobijk: to rule out that
13:32imirkin: tobijk: ok cool, thanks for checking
13:33imirkin: well, when i get the TK1 up and running [better than it is now], i'll see if i can repro there too
13:33imirkin: tobijk: until then, not much i can do unfortunately
13:33tobijk: heh its not for me, i was just checking up on that :>
13:34imirkin: yea i know
13:34mlankhorst: imirkin: i got glxgears running, vblank's still an issue though
13:35mlankhorst: I somehow need to hack it to use the present extension from the gpu screen, but the sync extension from the main screen
13:38mlankhorst: but that requires careful separation since crtc->screen potentially != pixmap->screen
13:38pmoreau: Is display_flip_stop, or maybe EVO base channel in general, dependent on something specific?
14:15smekras: guys, any of you familiar with systemd? cause all attempts to find the equivalent of xorg.0.log have failed
14:16RSpliet: /var/log/Xorg.0.log should still exist
14:16smekras: not on a clean rawhide install
14:16RSpliet: otherwise check out the log for gdm.service
14:17pmoreau: In ~/.local or something similar
14:17urjaman: says quick google
14:17smekras: yeah, neither exists
14:17pmoreau: Or ~/.local/xorg/Xorg.log I have both on my laptop for some reason
14:19pmoreau: You could try to run `find / -name "Xorg.*.log"` as root
14:19pmoreau: Erm, rather `find / -name "Xorg*.log"`
14:19urjaman: (maybe skip that first dot)
14:22smekras: it was in /var/log afterall... now why it didn't show up during the previous search... and yes, i was root
14:23smekras: i think my pc is trolling me
14:23pmoreau: Mine has been doing the same for some time, but it recently stopped
14:34Karlton: tobijk: after learning about libtxc_dxtn I ran some windows games, that previously had some blank textures, now display properly. :)
14:34tobijk: yay :)
14:34Karlton: I always thought it was an issue with nouveau
14:35tobijk: sadly thats a common mistake many do :/
14:36imirkin: Karlton: btw, i replayed your trace on my nvc0 and it all appears to render fine
14:36imirkin: so it's some issue that only appears on kepler :(
14:36imirkin: sadly i don't really have one of those on-hand
14:40Karlton: imirkin: as long as you know it's an issue.
14:42imirkin: yeah, tobijk confirmed he saw the same thing
14:42imirkin: so it's not just something isolated to your setup
14:42imirkin: and it works on nvc0, which eliminates a lot of things it might be
14:42imirkin: unfortunately what's left doesn't make a ton of sense =/
14:45airlied: smekras: journalctl -b should have it
14:45smekras: found it anyway, not that it shows anything odd
14:46Halfwit: How up to date is the feature matrix at http://nouveau.freedesktop.org/wiki/FeatureMatrix/ ?
15:48specing: imirkin: Hmm, about the glxinfo you linked when I asked about how software rendering performs
15:48specing: 1. Are there any performance comparisons?
15:49specing: 2. Can DRM _securely_ manage access to the card for several Xorgs? (I would like to move games into an LXC VM)
15:50imirkin: specing: software = slow. hardware = fast
15:50imirkin: on a scale from 0 to 1, software = 0, hardware = 1 :)
15:51imirkin: as for secure card access... sure, drm can manage it, but afaik no driver really implements anything that could be considered *secure*
15:52imirkin: and certainly not with nouveau -- userspace can easily hang the card... ideally it's just that one channel (which means that userspace just gets the axe) but often it will be more
19:17Apteryx: What is the state of OpenCL with the Nouveau driver?
19:18imirkin: i can give pointers if it's something you'd be interested in working on
19:25Apteryx: imirkin: thanks. I'm quite ignorant regarding the whole 3D stack. Not sure I'd be up to the task yet.
19:27Apteryx: imirkin: I understand OpenCL needs to be "accomodated" with recent hardware, right?
19:28imirkin: Apteryx: not sure what that means, but mainly the bits that are missing are glue between various components
19:30Apteryx: imirkin: I mean, that my 8800 GTS is probably too ancient hardware for OpenCL
19:30imirkin: nope, should support it just fine
19:31imirkin: at least up to OpenCL 1.1 i think
19:31imirkin: maybe 1.2? unlikely tho
19:31imirkin: G80 had a fully-functional compute shader...
19:34Apteryx: imirkin: ok! An OpenCL implementation would be part of the Mesa 3D stack?
22:50airlied: skeggsb, mlankhorst : my virgl test harness really makes a mess out of nouveau :)
22:51airlied: one list corruption and a lot of nv50cal_space: -16
22:52imirkin: list corruption is no good, but nv50cal_space means "you ran out of vram"
22:56imirkin: [or rather, there's a pushbuf submit request that requires a bunch of pushbufs to be in vram, and it can't fit it all in there]
22:57imirkin: so then the whole command fails, and the mesa driver really doesn't handle that very gracefully
23:26airlied: imirkin: also a bunch of failed to idle channel 0xcccc0000
23:26imirkin: yeah, that's ~expected given the other errors
23:31mlankhorst: yeah, usually means that the channel hung and idle fails
23:31airlied: granted the workload is very similiar to piglit
23:31airlied: since it is just piglit over virgl
23:32airlied: without qemu in between
23:32mlankhorst: do you disablel parallel piglit at least?
23:32airlied: nah, thats the whole point of runnig the virgl test harness :)
23:33mlankhorst: nouveau doesn't work with parallel piglit
23:33draven: Hello everyone! I have a NVE7 GK107 (nVidia GT740 2GB) card and io am confused as hell. I would want ( and expect ) to have tear-free videos but for some reason i do not understand how GLXVBlank works :|. If i disable it i have huge FPS in glxgears (2000++ in full screen 1920x1080) and if i enable GLXVBlank i only get 60 FPS. Both setups do not free me from tearing :C. Could i have a tip, advice or suggestion from you gu
23:33airlied: well parallel anything :)
23:33mlankhorst: most fun I've had was trying to figure out why max-texture-size stresstest fails, still uncertain
23:34mlankhorst: (running 4 max-texture-size repeatedly in parallel)
23:34imirkin: airlied: parallel mostly works on nouveau, but as soon as things render to screen, it all fails
23:34airlied: imirkin: didn't runs one EGL session per app
23:34airlied: this runs
23:34imirkin: draven: yeah, tearing is documented on kepler... doesn't tear on my fermi. no clue why.
23:35airlied: so for every piglit run, it has an egl rendernode user
23:35airlied: the only screen rendering is swrast putimage type stuff
23:35imirkin: draven: if you want to enable GLXVBlank and still see glxgears put out big numbers, you can run e.g. 'vblank_mode=0 glxgears'
23:35draven: imirkin: do i find this documentation on google or could you please give me a link to it?
23:35imirkin: draven: also make sure you have 3.17 or later... i think a bunch of tearing-related stuff got fixed then
23:36imirkin: draven: perhaps i overstated 'documented'. i've heard other people say it :)
23:36draven: imirkin: i thought glxvblank should have "fix" the tearing not make it worse x )
23:36draven: imirkin: oh ok that's another thing ^^
23:36imirkin: draven: it does. and it's on by default now.
23:36imirkin: airlied: ok. well i did a ton of parallel fp64 runs on my fermi -- all worked fine.
23:37imirkin: draven: by default, glxgears is sync'd to the screen, so if you have a 60hz refresh rate, it will display 60fps.
23:38draven: imirkin: btw i am the guy from a few days ago with the screen disturb on the white color, DVI cable solved everything :)
23:38draven: imirkin: yes indeed, 60 Hz
23:38imirkin: draven: ok. either a shitty cable or we're driving the vga pll's wrong somehow
23:39imirkin: (or everyone's favourite option -- both!)
23:39draven: imirkin: i think the VGA cable was sh(i18n)t
23:41draven: imirkin: i saw some people on google saying that compiz solved the tearing nightmare; my question is: it could be true? and xcompmgr would do the same?
23:41imirkin: draven: keplers have some sort of extra-special issue
23:41imirkin: presumably they're talking about a redirecting compositor
23:42imirkin: which would make some sense... i don't think that xcompmgr is one
23:43imirkin: but i'm very weak on the details of compositors...
23:43draven: imirkin: what kind of "extra-special issue" ? ^^
23:43imirkin: draven: i have a fermi, and no compositor, and i don't get tearing
23:44imirkin: draven: whereas you and i believe multiple other people have reported issues with kepler cards. and not just with nouveau -- also with nvidia's driver
23:44mlankhorst: haven't had tearing on compositing with fermi either
23:44imirkin: could be that we both have the same bug, or it could be that we're relying on the same bit of oddly-broken hw.
23:44draven: imirkin: sad me is sad :C
23:44imirkin: (or could be entirely unrelated bugs which exhibit the same end-user-visible issue)
23:47imirkin: anyways, their solution to the problem is an intel-style TearFree option
23:47imirkin: which basically does the minimum of what a redirecting compositor would, but in the driver
23:50draven: imirkin: lol... and it won't be fixed :|
23:58skeggsb: airlied: that's... not good.. do you have a backtrace for the list corruption?
23:58skeggsb: imirkin: actually, that error means "timed out waiting for space in the GPFIFO" == channel hung (most likely)
23:59imirkin: skeggsb: the nv50cal one?
23:59imirkin: oh right... yes
23:59imirkin: my bad :)