02:57 pmoreau: imirkin: Looks like my version of LLVM is too recent to get Gallium from Mesa 10.1 to build
04:06 necro: ah, a channel for nouveau, great! I'm having some issues using nouveau with gtx 970m
04:07 necro: aural is me also in case I lose connection here
04:07 RSpliet: necro: so do the devs unfortunately
04:07 necro: well, my issues are firstly kernel mode setting concerns
04:07 necro: which according to http://nouveau.freedesktop.org/wiki/FeatureMatrix/ is done
04:09 necro: particularly, when I boot up to user login prompt (with nvidia proprietary drivers removed), i cannot type, or if i can, the screen does not update. And if i close the lid of the laptop, and open it again, the screen stays off
04:10 necro: I am using kernel 3.19.3-3-ARCH (archlinux)
04:10 RSpliet: but it does so in the native resolution of your monitor, hence KMS works
04:11 necro: hmm
04:11 RSpliet: what login prompt are you talking about? is that a VT or a GUI?
04:11 necro: I do not understand what that means
04:11 necro: vt
04:12 RSpliet: ok, that's slightly unexpected, but please try a kernel 4.0 first
04:13 RSpliet: if it doesn't... it might help trying something even newer; your GM204 is still a work in progress (see the patches here: http://cgit.freedesktop.org/~darktama/nouveau/log/ )
04:18 pmoreau: Does someone know a way to dump the whole content of BAR0? When I try using nvapeek, my laptop will hang up at some point.
04:44 pmoreau: Nevermind, if I stay away from USER_DMA, everything works fine.
04:45 pmoreau: Now let's see which bits flippings solve everything... :)
07:22 bonbons: can someone explain this message: nouveau E[ PGR][0000:01:00.0] NOTIFY nsource: PROTECTION_ERROR
07:23 bonbons: that's the last sign of life from a NV04 sometime while starting Xorg, running linux-4.0 (system frozen after that, issue not seen with linux-3.19)
07:49 imirkin: bonbons: fairly sure it's not an issue... expected on NV04
07:49 imirkin: there's a bug about it... it doesn't like the NvNull binding that xf86-video-nouveau does
07:49 bonbons: ok, but still now anytime I try to run Xorg on that box the box freezes
07:50 imirkin: bisect? :)
07:50 bonbons: well, my attempts for today with 3.19 fail just as well :(
07:50 imirkin: oh
07:50 bonbons: so no good reference for bisect :/
07:51 bonbons: last output I get from Xorg over SSH is: "(II) NOUVEAU(0): NVEnterVT is called." followed by "(II) NOUVEAU(0): Setting screen physical size to 338 x 270
07:51 bonbons: resize called 1280 1024"
07:52 imirkin: yeah that's all happy
07:53 bonbons: don't know what I did special two weeks ago when I could start X on 3.19.3 and enjoy a slow XFCE (render accel is not really fast for whatever reason)
07:54 imirkin: hmmm
07:54 imirkin: well, the NV05 i have is plenty fast for the basics... doesn't go up to 1920x1200 though, had to run at 1650x1080 or whatever
07:54 imirkin: don't have a nv04, sorry. but people have indicated they work fine
07:56 bonbons: here smaller screen than that, 1280x1024, 16MiB VRAM, 192MiB RAM, PentiumII.
07:57 imirkin: right... i guess i was doing it on my Core i7 ;)
08:03 bonbons: might be it's XFCE that's not really moving windows around in the most efficient way while compositing...
08:04 imirkin: oh, compositing. heh.
08:04 imirkin: yeah, you don't want nothing to do with no compositing on a NV04 :)
08:06 imirkin: on the bright side, you get a YUV KMS plane. on the downside, no userspace support for using it.
08:11 imirkin: pmoreau: just don't enable the gallium swrast, you should be fine
08:32 pmoreau: imirkin: Hum, I thought I had disabled it… I'll retry a bit later
10:28 pmoreau: Is there a way to check with version of Mesa Steam is using?
10:28 pmoreau: s/with/which
10:28 imirkin: probably not the one you want
10:29 imirkin: it uses 32-bit meas
10:29 imirkin: mesa*
10:29 pmoreau: Ah, damn it! I forgot it was 32bit...
11:14 tobijk: pmoreau: apitrace the bit you want if possible and replay it on the 64 bit mesa maybe
12:12 RSpliet: pmoreau: do you have your G98 loaded?
12:13 RSpliet: sorry
12:13 RSpliet: G96
12:13 pmoreau: Always, I'm not going to remove it from my laptop
12:13 pmoreau: Especially as it is a Mac ;)
12:14 RSpliet: hahah, yeah, I'm trying to figure out which bits in VBIOS control the mem voltage
12:14 RSpliet: there's one big bad bit marked with a red flag shouting *ooh, pick me, pick me*
12:14 RSpliet: but there's two GPIO bits :-D
12:15 RSpliet: could you obtain reclocking trace with nvafakebios -e d26f:0a
12:16 pmoreau: I'll try that
12:16 pmoreau: So, do the nvafakebios before loading the blob?
12:17 pmoreau: tobijk: Ah, could try that! I had some issues compiling the 32bit version of Mesa
12:17 RSpliet: yes, in fact, double-check that the blob is really unloaded
12:18 RSpliet: I would thank you, but it's in your own interest :p
12:19 pmoreau: :D
12:29 pmoreau: RSpliet: How many up and down would you need?
12:31 RSpliet: well, just one from up to down, just to be sure
12:31 pmoreau: K
12:31 RSpliet: I already have a sneaking suspicion
12:31 pmoreau: Doing that right now
12:31 pmoreau: :)
12:31 RSpliet: thanks :-)
12:32 pmoreau: Well, thanks for having a look at it! ;)
13:17 pmoreau: RSpliet: https://phabricator.pmoreau.org/F5717
13:17 pmoreau: I hope it is correct
13:20 RSpliet: phabulous!
13:20 pmoreau: :p
13:37 RSpliet: pmoreau: perfect!
13:37 RSpliet: this seems to be about right: https://github.com/RSpliet/kernel-nouveau-nv50-pm/commit/01c28a3189735b0e8d723d4676bd74cd0b54111c
13:39 pmoreau: RSpliet: Awesome! Good job!
13:40 RSpliet: don't get your hopes up too high just yet :D
13:40 pmoreau: :D
13:40 RSpliet: there still is some timing bits and bobs I'm not confident about
13:40 pmoreau: If I apply it on Git, it won't work?
13:42 RSpliet: you're best off trying this tree: https://github.com/RSpliet/kernel-nouveau-nv50-pm
13:43 RSpliet: but... there's some timing things I didn't verify yet
13:43 pmoreau: Ok
13:43 RSpliet: nor did I test behaviour when the GPIO pins don't exist
13:43 pmoreau: :)
13:43 pmoreau: Undefined?
13:43 RSpliet: but I can check that last bit earily
13:43 RSpliet: *easily
13:46 RSpliet: I think GPIO should be fine... but not the most tidy piece in the world
13:46 RSpliet: (as it doesn't check for negative values of gpio->get... it checks later though on ->find)
13:59 pmoreau: Compiling...
14:03 RSpliet: oh yes, I might have accidentally left a compile-time error in there :s
14:03 RSpliet: eclipse warned about ram->base.pfb not existing
14:03 RSpliet: sorry, that's hwsq->base.pfb
14:04 pmoreau: Ok
14:09 RSpliet: that probably needs to change into something like hwsq->hwsq->pbus.base
14:10 RSpliet: but don't take my word on it, I already went into non-dev mode :p
14:11 pmoreau: :D
14:12 pmoreau: No problem
23:28 RSpliet: pmoreau: I take it you dozed off or everything worked? :-P