06:53sooda: mupuf: cyber! https://pbs.twimg.com/media/CtWb47tWAAEhjhV.jpg
06:57pmoreau: sooda: Hey! Care to translate for those of us not speaking Finnish? :-D
06:57sooda: it's "i will never learn to understand cyberspace... everything is so confusing"
06:58sooda: referring to the yellow CYBER stickers in XDC :)
06:58pmoreau: :-D Thanks!
06:58pmoreau: I missed those stickers…
06:58sooda: i wish i still had those. could print this comic and use one of the stickers to hang that print on our door at work
06:59pmoreau: Could you display it as a Gif on the NVIDIA metro tech display?
07:38mupuf: sooda: I still have some :D
07:38mupuf: as for the translation ... my take on this was quite bad :D
11:57karolherbst: pmoreau: how could you miss those stickers? :D
11:58karolherbst: sooda: I still have like 20 of them
11:58karolherbst: you can come and pick them up if you want to :)
12:31sooda: thanks, i'm good, mikko already gave one from his stash and that's enough :)
13:05karolherbst: airlied: is there something holding the nouveau merge besides time?
13:07karolherbst: sooda: one? …. you can't have enough of those! :D
14:03imirkin: karolherbst: pretty sure ben hasn't sent a pull request out
14:03imirkin: [or put a tree together]
14:07karolherbst: I see
14:08karolherbst: well, we made all kinds of promises for 4.9 :D
14:21pmoreau: karolherbst: Well, dunno…
14:40pmoreau: CUDA 8.0 was finally released! With clang 3.8 support (and GCC 5.4 I think)…
14:41Tom^: karolherbst: everybody lies.
14:56karolherbst: Tom^: not me!
15:53karolherbst: why are there only shitty sites to design t-shirts or stuff ...
16:14jrun_: chromium (with ignore-blacklist) doesn't do webgl here, any idea?
16:45zeq: jrun_: You need to look at chrome://gpu see if the GPU successfully initialised
16:46zeq:has quite a lot of experience with Chromium hwaccel since he maintains a vaapi support patch for Gentoo
17:06waltercool: Guys, question, if I would like to test the nouveau kernel module from the skeggsb, I do need to also use libdrm at git version?
17:06waltercool: I had some problems trying to compile&load last time
17:07waltercool: Following the out-of-tree steps at freedesktop website
17:12karolherbst: waltercool: you need a compatible drm-next kernel
17:13karolherbst: the last drm-next: $hash commit from nouveau tells you a compatible commit
17:13karolherbst: or just use my repository and the mastet_4.7 branch
17:14karolherbst: there are like a handful commits missing though
17:20imirkin: waltercool: libdrm has nothing to do with it
18:02jrun_: zeq: i only have 'Initialisation time' which is 0
18:03jrun_: that's from chrome://gpu
18:04imirkin: jrun_: webgl should work on chrome... what gpu do you have again?
18:05imirkin: if it says "hardware accelerated" there, it should be enabled
18:05imirkin: that said, i don't think WebGL 2.0 is enabled by default yet
18:10waltercool: karolherbst: yeah, with 4.8-rc wasn't working, with 4.7 compiled fine, but I got some errors trying to load correctly the module
18:11waltercool: imirkin: My bad, I meant drm kernel module
18:11karolherbst: waltercool: ohh wait, my master_4.7 branch is up to date
18:11karolherbst: what errors do you have loading it?
18:12waltercool: let me load back to 4.7, give me a sec. I'm at 4.8-rc right now
18:15jrun_: imirkin: it does say hw accelerated but this, for example, fails:
18:15jrun_: and this:
18:15jrun_: ^that actually worked with nvidia blob on the same system
18:16imirkin: Uncaught ReferenceError: Wasm is not defined
18:16imirkin: the latter works for me, i think
18:16imirkin: [still loading]
18:17waltercool: karolherbst: http://pastebin.com/wKdSW4Sy
18:17imirkin: jrun_: the second one definitely works. not gonna restart chrome to test that first one
18:18waltercool: Is a nasty maxwell2, but works fine using 4.7 nouveau from tag
18:18karolherbst: waltercool: is this, this crazy mobile 980 with 8GB vram?
18:18waltercool: karolherbst: yeah...
18:19waltercool: why so crazy? Lawl
18:19karolherbst: mhh, well I know the issue anyway, or I guess it is the one I have in mind
18:19karolherbst: could you upload the vbios somewhere?
18:19waltercool: yeah, give me a sec
18:19imirkin: karolherbst: i doubt it's the partition issue
18:19karolherbst: imirkin: it ain't
18:19karolherbst: imirkin: it is the 32bit table offset one
18:20imirkin: that one.
18:21waltercool: well, give me another second, because this module doesn't work, I need to switch back to 4.8 and get the vbios
18:21karolherbst: shouldn't matter I think
18:21karolherbst: ohh wait
18:21karolherbst: the probe fails
18:21karolherbst: maybe it does
18:21karolherbst: doesn't matter from where you get the vbios though
18:22jrun_: imirkin: driver version?
18:22waltercool: what do you mean? I don't have the dri/1/
18:22karolherbst: I see
18:22waltercool: only the intel ones
18:22imirkin: jrun_: mesa 12.0.3
18:22jrun_: 12.0.1 here
18:22imirkin: shouldn't matter
18:22imirkin: jrun_: which GPU do you have again?
18:23jrun_: 01:00.0 VGA compatible controller: NVIDIA Corporation G92 [GeForce 9800 GT] (rev a2)
18:24jrun_: should matter, it can do webgl (whatever version) needed for the second one.
18:24jrun_: well, unless driver doesn't support it...
18:24jrun_: i gotta shutdown, there is gonna be a power outage; brb.
18:39waltercool: got it now
18:41waltercool: karolherbst: https://www.slash.cl/vbios.rom
18:42karolherbst: yep, as I thought
18:42karolherbst: waltercool: well the fix is rather trivial, if you have some experience with programing, you could even do it yourself
18:43waltercool: karolherbst: depends of the programming skill, but yeah
18:43karolherbst: waltercool: it is basically replacing uint16_t with uint32_t ;)
18:43waltercool: karolherbst: some specific file?
18:44karolherbst: waltercool: a few
18:44waltercool: how did you get that? Just knowing the architecture?
18:44karolherbst: nope, just looking into the vbios
18:44karolherbst: we know this issue
18:44karolherbst: but nobody bothered enough for now
18:44karolherbst: it is super rare
18:45karolherbst: and maxwell2 is pretty much useless anyway
18:45karolherbst: waltercool: drm/nouveau/nvkm/subdev/bios and drm/nouveau/include/nvkm/subdev/bios
18:45karolherbst: for example drm/nouveau/include/nvkm/subdev/bios/vmap.h
18:45karolherbst: the functions return u16
18:45karolherbst: they should return u32
18:45waltercool: should I do it in master or 4.7 branch?
18:45karolherbst: doesn't really matter
18:46karolherbst: it is just _a_lot_ of work and you have to be a bit careful
18:46karolherbst: but basically you get the idea
18:46waltercool: yeah, grep for u16 and sed for u32
18:46karolherbst: waltercool: nope
18:46karolherbst: you have to check every occurance
18:46karolherbst: otherwise, we would have fixed it already :D
18:46karolherbst: waltercool: start with ramcfg
18:48waltercool: oh, I see now the issue
18:48waltercool: let me start in a few hours and I will let you know as soon I can
18:48karolherbst: ohh and another thing
18:49karolherbst: inside the bit parts, you have to change it from u8 to u16
18:49karolherbst: like here: https://github.com/karolherbst/nouveau/blob/master_4.7/drm/nouveau/nvkm/subdev/bios/ramcfg.c#L40-L44
18:49karolherbst: mhh wait a second
18:49karolherbst: that should be fine
18:49Jaga: karolherbst: re: shitty sites to design t-shirts - what's the issue?
18:49karolherbst: Jaga: I want to make awesome nouveau shirts!
18:50Jaga: karolherbst: do you want someone to design them for you or just make them according to your design?
18:50karolherbst: waltercool: anyway, if you would start that until it works on your card, enough is done already, sadly I don't own any gpu with that issue...
18:50karolherbst: Jaga: I want to design myself
18:50Jaga: karolherbst: and what kind of tshirt would you like? inked or plasticy iron-on image?
18:50Jaga: karolherbst: how many would you be ordering/producing?
18:50karolherbst: mhh good question
18:50karolherbst: I would go for individual shirts
18:50waltercool: I hope something like ramcfg should be a final solution for a patch, isn't?
18:51karolherbst: Jaga: basically nouveau logo + name + "area of development"
18:51karolherbst: waltercool: there are many
18:51karolherbst: waltercool: you can install envytools and put your vbios through nvbios
18:51karolherbst: waltercool: https://gist.github.com/karolherbst/97e92a11453c19bef207d291e75e26f5
18:51karolherbst: ;) obvious, isn't it?
18:52Jaga: karolherbst: I'm asking cause you could ink them at home if you want to (as in, it's doable if you're ok with DIY)
18:52karolherbst: mhhhh, good idea actually, just no clue how I would have to start
18:52Jaga: karolherbst: inking is harder, print and iron on is very easy but breaks easier
18:52karolherbst: and it should hold on for a bit
18:52karolherbst: yeah, it shouldn't break
18:52waltercool: karolherbst: lol yeah, even yugi's puzzle seems a piece of cake
18:53Jaga: karolherbst: I've done iron on, I've done acrylic painting with stencils, and I'm now thinking of doing inks (either screen printing or stamping)
18:53waltercool: so, let me try in a couple of hours to play with it
18:53karolherbst: waltercool: well, the performance table is at 0x14c35 actually ;)
18:53karolherbst: but we just parse it wrongly
18:53Jaga: karolherbst: I can help you if you want - as much as I can over the internets :p
18:54karolherbst: Jaga: well, what I had in mind is printing this on the back: http://www.phoronix.net/image.php?id=0x2014&image=nouveau_new_logo
18:54karolherbst: + some more infor about the person wearing it
18:54Jaga: karolherbst: what tshirt color? white or black?
18:54Jaga: or something else
18:54karolherbst: Jaga: no idea yet, but I would got for black and turn the font white or something
18:55karolherbst: maybe even green, allthough that could be too much green
18:55Jaga: karolherbst: when you do it on black, are you ok with the white areas of the logo being black too or not?
18:56karolherbst: Jaga: nope, should stay white :)
18:56waltercool: well, yeah, I'm very used of the whole nouveau and nvidia relationship, doing hardware analysis on that way is madness
18:57Jaga: ok, so you'd need 3 colors for it at the same time, and the logo is pretty detailed - I'd say that means you have 2 options: 1. easy one is to do iron-on prints 2. hard one is to build a frame (to hold the tshirt in place) for screen-printing and do 3 screens (one for each color)
18:57waltercool: well, just in case you want to play with my videocard, I can lend a hand of it, and let me check the u16->u32 later
18:58karolherbst: Jaga: ohh, I think I actually saw the second option once and was pretty amazed on how well it works and looks :)
18:58karolherbst: waltercool: nope, most of the things are fine, just the table thing ain't
18:58Jaga: karolherbst: screen printing makes sense when you want 3+ copies of the same, so text for each person would probably be a waste - you could do ink + stamps tho, like the gutenber's first printers
18:59karolherbst: yeah, the name ain't important anyway :D
18:59Jaga: karolherbst: I saw it too in action, and it is cool but also lots of work - if you have one color, you can actually just make one screen and tape it with the painter's tape on the tshirt
18:59karolherbst: no idea who would want one anyway
19:00Jaga: well, I'd guess if the team is tight-knit they'll think it's cool
19:07karolherbst: I hope so, anyway, I really lack everything I would need for this to do it myself anyway, but I am sure I should find something in hamburg, never done it though and I am a bit scared to screw up
19:10Jaga: karolherbst: I saw at 32c3 people with a frame doing screen prints, no idea what hacker association they were from... you could try them
19:10karolherbst: Jaga: the plan was to have a shirt before 33c3 :D
19:10Jaga: karolherbst: if you were to find them, you can ask for help/advice
19:11karolherbst: ohh I see
19:11Jaga: karolherbst: there's also one group.. something-raum, that has a machine for sewing images on clothing, they do pre-orders before c3 events, but maybe they also do it also at other times
19:12Jaga: I have a hoodie with full back logo made by them, really nice stuff <3
19:12Jaga: and since they can load up what they want into it, it's full custom :D
19:16Jaga: found 'em https://raumzeitlabor.de/
19:16karolherbst: so far into the south :O
19:17Jaga: example of their work https://66.media.tumblr.com/8783fd09b209f1fe622fdf00ddd0a383/tumblr_o3bcxw2vsM1rsbdmgo1_1280.jpg
19:17Jaga: you get thick lines with one setting and thin filling with another, both opaque colors
19:18karolherbst: mhh no idea if that would fit with the nouveau logo though
19:18Jaga: you can also have glow in the dark thread and other stuff
19:18karolherbst: ohh of course, the glowing is important! how could I have forgotten about this
19:18Jaga: uh, told ya I have full back image
19:18karolherbst: and LEDs! always LEDs!
19:19karolherbst: glowing with LEDs :D how awesome would that be
19:19Jaga: 36 cm x 22 cm is what my hoodie had
19:20Jaga: iirc they had just these frames in which the material is secured and those just limit the size - if clothing can fit in a larger without problem to print, then it will be done
19:20dugivug: we want dx11 support at last
19:21karolherbst: Jaga: I see, mhh I think I would rather go for a print then, cause I don't think sewing would really look good with that logo
19:26karolherbst: Jaga: or is it also possible to sew this without those thick borders?
19:27Jaga: karolherbst: might be possible, or boarders can be made thin, my hoodie had no such borders
19:27karolherbst: Jaga: mind showing a photo of that?
19:27Jaga: I don't have a pic
19:27Jaga: idk if it's a good example tho, it's a line-art pretty much, no filling
19:27karolherbst: I see
19:28Jaga: https://66.media.tumblr.com/3b82f77ccec0ac4d682d340e81345b15/tumblr_natvymezMo1rsbdmgo4_1280.jpg here's a borderless image
19:28Jaga: just filling
19:29karolherbst: that doesn't look too bad indeed
19:32Jaga: there's also multi-color vinyl cutting and iron-on (I haven't done it, didn't have access to vinyl cutter): http://www.thepinningmama.com/how-to-layer-heat-transfer-vinyl/
19:44karolherbst: Jaga: mhh, I am troubled now. The sewing thing indeed looks so much more awesome, but the printing may fit better with the logo... mhh anything you would prefer for that logo?
19:45Jaga: karolherbst: I would assume a good silk screen print would be awesome, but I'd afraid it's a lot of work
19:46Jaga: karolherbst: you'd need a frame + shelf to stabilize the tshirts when putting on colors so they absolutely align
19:46karolherbst: well, I don't exactly have to do that myself though, even if that would be super awesome :D
19:46karolherbst: yeah, I am kind of aware
19:47Jaga: pity I don't remember the hacker association that did the screen prints at 32c3... they were opposite the c-base
19:58Jaga: bed time for me, good night
19:58karolherbst: good night and thanks for the suggestions
19:59Jaga: glad to help :)
20:35dugivug: make dx11 support
20:35RSpliet: bash: make: command not found...
20:35dugivug: wanna play
20:35karolherbst: dugivug: ask the #d3d9 devs ;)
20:36karolherbst: or #winehq
20:36imirkin: dugivug: what does that have to do with nouveau?
22:17gsgf: can someone tell me how to read out the fan speed of the gpu card with nouveau?
22:18gsgf: i try for example to set the fan speed to auto using this howto but this does not seem to work
22:18imirkin: run 'sensors'
22:19imirkin: nouveau should be set to auto-control fan speed nowadays
22:19imirkin: it wasn't back in 2013 though, i think
22:19gsgf: lm-sensors ?
22:20imirkin: you can also read it out directly from sysfs...
22:20imirkin: not all fans have tachometers though
22:21gsgf: thats the reason why i liked to first try to read the fan speed: http://pastebin.com/0670SD4k
22:22imirkin: right. no tachometer.
22:22gsgf: oh, its a two wire cable i see now... so speed can not be read out. fan then always run on full speed?
22:22imirkin: otherwise it'd show up on there.
22:22imirkin: no, it's adjusted based on temperature
22:22gsgf: but it does not adjust
22:22imirkin: it should.
22:23mupuf: gsgf: what temperature is your gpu at now?
22:23gsgf: thats why i am here. tested with 4 different nv40 cards. no adjustment. always 100% speed
22:23mupuf: ah, old gpus may not have fan management support
22:23imirkin: pastebin the output of 'grep . /sys/class/drm/card0/device/hwmon/hwmon*/pwm*'
22:23gsgf: mupuf: hi, i read the FAQ that the people should ping you about FAN things :)
22:24imirkin: nv40 should have fan control iirc
22:24mupuf: imirkin: not all GPUs can
22:24imirkin: i'm almsot sure my NV42 had it
22:25gsgf: mupuf: how can i find out if its based on I2C?
22:25imirkin: gsgf: well, for whatever reason, we're driving it at 100%
22:25imirkin: you can flip it into manual mode
22:25gsgf: imirkin: i tried 7 different NV40 cards with this machine. Kernel 4.7. Always 100%
22:25imirkin: and then echo other values into that 'pwm1' file
22:25gsgf: can i help debug this?
22:25mupuf: yes :)
22:26gsgf: i have 11 different NV4x series cards here. Should be more then enough for helping out nouveau ;)
22:28mupuf: first try what imirkin said
22:29gsgf_: mupuf: i lost your messages after "yes"
22:29gsgf_: how can i help?
22:29mupuf: <imirkin> gsgf: well, for whatever reason, we're driving it at 100%
22:29mupuf: <imirkin> you can flip it into manual mode
22:29mupuf: <imirkin> and then echo other values into that 'pwm1' file
22:29mupuf: echo 1 > pwm1_enable
22:29mupuf: echo 30 > pwm1
22:30gsgf_: i get the message i were not allowed to set values into pwm - as root
22:31imirkin: did you first echo 1 into pwm_enable?
22:31gsgf_: to find pwm_enable i ran find /sys -name pwm1_enable
22:31imirkin: it's in the same directory
22:31gsgf_: and then i echo 1 into it
22:31imirkin: as the other file.
22:31imirkin: look in the grep output above
22:32gsgf_: could you post the grep output again?
22:32imirkin: you're the one who posted it...
22:32imirkin: anyways, i'll bbiab
22:32gsgf_: ah, that you mean. yes, this /sys/devices/pci0000:00/0000:00:02.0/0000:01:00.0/hwmon/hwmon2/pwm
22:33gsgf_: i get message, that i were not allowed
22:33imirkin: i mean /sys/class/drm/card0/device/hwmon/hwmon*/pwm*
22:34gsgf_: great. works: echo 30 > /sys/class/drm/card0/device/hwmon/hwmon2/pwm1
22:34imirkin: does fan speed go down?
22:35gsgf_: defenetly yes
22:35mupuf: ok, then I screwed up, badly...
22:35imirkin: cool. so we can control the fan just fine, we just decide to put rpm at 100%
22:35mupuf: re-enable auto fan management
22:35mupuf: echo 2 > pwm1
22:35gsgf_: echo 2 > /sys/class/drm/card0/device/hwmon/hwmon2/pwm1_enable
22:36gsgf_: seems to stay at 30
22:36gsgf_: i cant hear any difference
22:36imirkin: mupuf: btw, any update on that crash that i and a bunch of other people have with GK208's and weirdly-controlled fans?
22:36mupuf: imirkin: oh dear....
22:36imirkin: [might not be GK208-specific]
22:37mupuf: you mean the alarm issue?
22:37gsgf_: temp is with 30% fan at auto now at 78°
22:37imirkin: although i think you said it was brought on by the lack of fan controllability
22:37gsgf_: before it was with 100% fan at 50°
22:37gsgf_: now 80° and rising
22:37imirkin: and indeed iirc my fan was not controllable even though nouveau thought it was
22:38imirkin: gsgf_: temp1: +50.0°C (high = +95.0°C, hyst = +3.0°C)
22:38imirkin: i guess nouveau will start freaking out around 98 C?
22:38gsgf_: (crit = +145.0°C, hyst = +2.0°C)
22:39imirkin: that's when the gpu will freak out ;)
22:39gsgf_: (emerg = +135.0°C, hyst = +5.0°C)
22:39imirkin: i doubt those are correct though
22:39gsgf_: why is emerg lower then crit ?
22:39gsgf_: we are at 87° now
22:40gsgf_: lets see what happen at 95° ...
22:40gsgf_: fan is spinning really really slow
22:40gsgf_: with 30% at "auto"
22:40gsgf_: but this auto is really broken
22:40imirkin: normally there are ramps
22:40imirkin: that will cause the fan speed to ramp up
22:40imirkin: perhaps we don't parse those out properly
22:41imirkin: and/or don't have proper defaults for
22:41gsgf_: fan spinned up. hurray
22:42gsgf_: cat /sys/class/drm/card0/device/hwmon/hwmon2/pwm* -> its at 100% again
22:43gsgf_: how can i help fix this?
22:44gsgf_: mupuf: any idea?
22:46gsgf_: i see why the temp ramp up in dmesg
22:46gsgf_: nouveau 0000:01:00.0: therm: temperature (90 C) hit the 'fanboost' threshold
22:46gsgf_: nouveau 0000:01:00.0: therm: temperature (88 C) went below the 'fanboost' threshold
22:47gsgf_: sensors report now again 50 C. Fan Speed stay at 100%
22:47gsgf_: kernel 4.7.5
22:51gsgf_: sensors does not report any lower temp then 50 C. Thats normal?
22:55imirkin: depends on thr gpu... is this a higher end one?
22:56gsgf_: mid end
22:56gsgf_: no external power connector
22:56gsgf_: about 70*70mm square fan size
22:56gsgf_: how can i help fixing this?
22:57mupuf: gsgf_: I will plug an nv40
22:57mupuf: and check
22:57gsgf_: mupuf: great :)
22:58gsgf_: i plugged as reported many different ones and all of them had this problem. all nv4x.
22:58gsgf_: the fan speed you hear at powering up the computer stay the whole time.
22:59mupuf: that really sucks
23:00gsgf_: mupuf: you have a second machine to test the nv40 at the moment?
23:00gsgf_: booting up now? :)
23:01mupuf: nope, first I needed to do some archeology
23:02gsgf_: hopefully you did not just find a passive cooled one :p
23:02mupuf: only found one so far, will check somewhere else
23:02mupuf: they are all passively cooled, but one exposes a fan in the bios, so it is the same for me
23:02mupuf: in the mean time, I will need your vbios
23:06gsgf_: i love the "cat /sys/kernel/debug/dri/0/vbios.rom > vbios.rom" . i already reported once a vbios with this. this should stay at first place and the people should try something else only when this dont work :)
23:07gsgf_: mupuf: http://www.file-upload.net/download-11973040/nvidia_quadro_fx_550.html
23:08gsgf_: i have a second card with same chip, same design but other vbios if thats interesting. Probably more interesting would be a totaly different card with also same problem i think
23:10mupuf: ok, it is not a toggle fan
23:14mupuf: yeah, I can reproduce
23:14gsgf_: with your nv40 card you probably already know really good?
23:15mupuf: well, knowing the hw is one thing, not screwing up the code is apparently another one
23:18mupuf: nouveau 0000:01:00.0: therm: parsing the fan table failed
23:18gsgf_: in the mean time i found the perfect speed. its 40%. I cant hear the card and it dont get up to 90 C
23:18mupuf: but then, it says therm: fan management: automatic
23:19gsgf_: mupuf: lets grep my dmesg for this
23:19mupuf: you will need to load nouveau with debug=debug
23:20gsgf_: i see. because i cant find it right now in the dmesg
23:23mupuf: hmm, this code could definitely see a rewrite
23:23gsgf_: if i could i would like to vote for "parsing the fan table failed - message without enabled debugging" . I think its way more usefull for the users then something like this: 10.323815] nouveau 0000:01:00.0: DRM: Saving VGA fonts
23:24mupuf: it is not an issue
23:24mupuf: the second table actually got parsed without issues
23:25mupuf: nvidia's PM used to be so messed up...
23:25mupuf: I do not miss this time!
23:25gsgf_: by the way: kwin crashed now the second time in 30 minutes. KDE developers say thats in most cases a driver problem. but lets fix fist the not working fan speed :)
23:27imirkin: gsgf_: i agree with kde developers. make sure to disable any use of GL in kde.
23:27gsgf_: mupuf: did you plan to rewrite this part? I would love to test with some different hardware then. It would be great if you could provide precompiled debian kernels that could be installed on the way to test quick :)
23:28mupuf: lol, me providing packages for debian
23:28mupuf: on my dead body :D
23:28gsgf_: mupuf: what distro did you prefere?
23:28mupuf: seriously, if you cannot recompile your kernel, then you are not fit to test, sorry :s
23:28mupuf: it is ok though, I have the issue locally
23:28gsgf_: i can recompile ;)
23:28mupuf: updates will come in due time
23:29Tom^: he doesnt necessery have to recompile the entire kernel tho, could just compile nouveau out of tree and move it into proper place run depmod -A etc
23:30gsgf_: i didnt know if you maybe also use deb based OS. Then only one compilation were needed ;)
23:30mupuf: Tom^: well, right... maybe not? Nouveau OoT is based on drm-next commits :s
23:30imirkin: i suspect kernel developers that use debian tend not to make packages out of their work-in-progress stuff.
23:31Tom^: mupuf: mhm ok
23:32gsgf_: imirkin: the one i know do that to quickly install and remove testing versions. But thats offtopic
23:32mupuf: I wonder if alarm is working at all on nv40
23:33gsgf_: imirkin: should i collect the logs of those kwin crashes and report them to nouveau?
23:33imirkin: huh ok.
23:33imirkin: gsgf_: nope
23:33Tom^: mupuf: are there an actualy arm device with such old gpu?
23:33imirkin: gsgf_: there are a dozen separate issues
23:34gsgf_: mupuf: alarm mean getting higher then 90 C and then ramp up? it worked here: [ 1860.470225] nouveau 0000:01:00.0: therm: temperature (90 C) hit the 'fanboost' threshold
23:34mupuf: yeah, because we use the CPU to poll on the temperature sensor
23:34Tom^: mupuf: oh lol, thought you meant the arch arm port, usually called alarm xD
23:35gsgf_: mupuf: did you also have always lowest reported temperature at 50 C like i have?
23:36mupuf: this is not an interesting point, who cares about the temperature of my board ;)
23:36gsgf_: lower then 50 C is 50 C. Only when it gets above 50 C it starts to show some usefull information
23:37gsgf_: i mean GPU temp reported by nouveau
23:37mupuf: lol, idle, it already hit 95°C
23:38imirkin: mupuf: maybe that fan was on there for a reason?
23:38mupuf: imirkin: there are no fans, it is a fanless gpu
23:38gsgf_: mupuf: sounds like a nvidia geforce 6600 passive cooled
23:38imirkin: mupuf: now it is. but perhaps not originally? :)
23:38mupuf: GeForce 7600 GT
23:38gsgf_: yes, the 600 series. they get extremely hot
23:40gsgf_: imirkin: a second reason for such high temperature is the worse quality of cooling paste between the GPU and the cooler that was used at this times
23:40gsgf_: imirkin: passive cooled GPUs always gets the recoomendation in this time to replace it with the old arctic silver 5 (these days there are much better ones)
23:42imirkin: duly noted.
23:42imirkin: skeggsb: welcome back!
23:49gsgf_: mupuf: you told "updates will come in due time". i dont know what this mean in usual in the nouveau development. Does it mean "in most cases its fixed in 2 weeks and pushed into kernel in 4"? Or does it mean "maybe 2017, more likely 2018"?
23:50mupuf: gsgf_: I am trying to fix it tonight
23:50mupuf: and send the patches when I have slept
23:50imirkin: mupuf: isn't it already tomorrow?
23:50gsgf_: mupuf: ok, thats really awesome.
23:50mupuf: it is...
23:50mupuf: imirkin: yeaaaahhh!!
23:50mupuf: I know what it is .... more locking issues!
23:52imirkin: that whole alarm thing seems overly complicated...
23:53mupuf: yeah, all of this because we did not want to depend on linux's internal
23:54mupuf: I need to just add support for it in the userspace and be done with it
23:54mupuf: and get rid of all these racy alarms
23:54mupuf: and I can just make it a separate process that just runs all the time
23:54mupuf: with sleeps inside, that would make it super simple to read
23:55imirkin: you mean a kworker?
23:57imirkin: i think normally one uses a wq for this? i forget
23:57imirkin: afaik this is a pretty standard problem that's easily addressed
23:57mupuf: wq have no guarantees in when they will be serviced :s
23:57imirkin: does anything?
23:57imirkin: you can set priorities on your things
23:58mupuf: surely, but for toggle fan management, you really want to be accurate
23:58mupuf: and sleep_range would help, rather than just hoping
23:58imirkin: i think it's generally pretty frowned upon to use a kworker
23:58mupuf: and having a resolution of 1 tick
23:58imirkin: it's tricky to get right wrt hibernation & co
23:59mupuf: right, quite likely too, but ... I am toggling a GPIO
23:59mupuf: I need some time constraints
23:59imirkin: right, i mean you can set a priority on the wq task
23:59mupuf: the frequency is 10 Hz, so that is 20 wake ups per second
23:59mupuf: yeah, but I cannot schedule it to arrive at exactly the right ms