00:25 rhyskidd: huh, so envytool's nvapy Python module has been broken for a little bit
00:31 rhyskidd: hard to test via script, so if anyone else used it, do you mind verifying its fixed for you as well? https://github.com/Echelon9/envytools/tree/fix/nvapy-compile
00:33 mwk: rhyskidd: yeah, it works for me
00:33 mwk: as in, it fixes nvapy for me
00:33 rhyskidd: awesome
00:34 rhyskidd: i thought about how to test that the module is actually importable, but it breaks unless the module can grab a handle to a card (so is going to fail in the cloud, Travis etc)
01:36 mwk: hmm
01:36 mwk: check for ImportError?
01:37 mwk: the "raise Exception" should probably be changed into some better exception class
01:37 mwk: then we could count raising the NoCardsException or whatever as well as no exception as successes for CI purposes
06:25 damo22: hey my friend here has 02:00.0 VGA compatible controller [0300]: NVIDIA Corporation G86 [Quadro NVS 290] [10de:042f] (rev a1) but the driver on his machine doesnt kick in
06:26 damo22: he is getting an Xorg.0.log that looks like this : https://pastebin.com/8ed8zA4M
06:26 damo22: where should we be looking for ideas?
06:31 damo22: (EE) [drm] Failed to open DRM device for pci:0000:02:00.0: -19
06:34 fleeky: hey there , for some reason the fan is really loud , it seems like the gpu is under load but nothing is using it ?
06:36 fleeky: is there any known issue with nouveau vs the nvidia drivers about this?
06:39 damo22: Subsystem: NVIDIA Corporation G86 [Quadro NVS 290]
06:39 damo22: Kernel modules: nvidiafb, nouveau
06:39 damo22: i assume he needs to purge all nvidia related crap from his system right?
11:15 mupuf: rhyskidd: good catch, ship it!
19:20 karolherbst: pmoreau: I think your for_nouveau branch of spvtools doesn't built anymore
19:21 karolherbst: some include issue with the spv-headers
19:46 karolherbst: anybody any idea what this is all about? "nouveau 0000:01:00.0: swiotlb buffer is full (sz: 2097152 bytes)"
19:47 imirkin: see dri-devel
19:47 imirkin: allegedly harmless.
19:47 karolherbst: uhm
19:48 karolherbst: well
19:48 karolherbst: processes got stuck
19:49 karolherbst: I wanted to do a piglit run, but now I have 8 processes being stuck, which is kind of annoying
19:49 karolherbst: and not something I would consider harmless
19:52 karolherbst: in this line: "ret = __ttm_bo_reserve(bo, interruptible, no_wait, ticket);" in include/drm/ttm/ttm_bo_driver.h:905
19:56 imirkin: the warning itself is harmless
19:56 imirkin: it indicates a fairly normal condition and shouldn't be a warning
19:56 imirkin: what you're seeing is different
19:56 imirkin: and unrelated to the warning
19:56 imirkin: i have like 10000000 of those warnings in my logs
19:56 karolherbst: mhh
19:57 karolherbst: I saw that today for the first time
19:57 karolherbst: allthough today is my first day on 4.15
22:30 rhyskidd: any interested reviewers for -> https://github.com/envytools/envytools/pull/118
22:49 karolherbst: rhyskidd: looks good, I don't know if there is any distribution without a cython3 binary though
22:49 karolherbst: and I totally forgot I had to do this docker stuff...
22:49 nwm: is nvkm supposed to have around 100 wakeups/s?
22:50 karolherbst: nwm: depends on what you do
22:50 nwm: ssh
22:50 karolherbst: mhh
22:51 nwm: debian autoloads it so i'm guessing I could just blacklist it
22:51 karolherbst: maybe some fan stuff?
22:51 karolherbst: on my laptop nvkm doesn't really get interrupts when idling
22:58 imirkin: nwm: laptop? might have a VGA port with a floating hpd pin. (and in fact, totally disconnected, but still existent.)
22:58 nwm: yep laptop w/ vga port
22:58 nwm: how do I fix the wakeups?
22:59 nwm: do I need to blacklist nouveau?
22:59 imirkin: well, first it'd be good to properly understand why they're happening
22:59 imirkin: mmm ... blacklisting nouveau will mean that the GPU will stay awake and eat up more power.
22:59 imirkin: but it'll avoid the wakeups.
23:00 karolherbst: imirkin: by the way, all bindless_texture piglit tests are passing with my pascal card
23:01 imirkin: except one skip presumably
23:01 karolherbst: but now I remember we already knew that...
23:01 imirkin: i pushed the patches btw
23:01 karolherbst: might be, it just shows "131/131"
23:01 imirkin: oh. my piglit might be out of date. i have skip: 1, pass: 28
23:01 imirkin: for -t bindless.
23:02 imirkin: oh, perhaps you're running compiler tests too?
23:02 karolherbst: I ran everything
23:03 karolherbst: *sigh* now how I get those tessellation things to pass again with my nir stuff after somebody "broke" it
23:04 imirkin: wasn't me!
23:04 imirkin: (hopefully)
23:05 karolherbst: it was the AMD guys
23:06 imirkin: ah yeah. that happens a lot. nouveau's in a pretty broken state atm with all the sampler changes.
23:06 imirkin: haven't had time to investigate properly
23:06 karolherbst: :) I really should get the CI stuff running
23:07 imirkin: if anyone used nouveau, i'd just revert those changes
23:07 nwm: would any module params be useful runpm?
23:07 imirkin: nwm: runpm defaults to on
23:08 imirkin: nwm: to figure out wakeups, boot with nouveau.debug=trace drm.debug=0x1e
23:08 imirkin: in some instances, it's actually the desktop manager that's constantly polling for stuff =/
23:08 imirkin: although probably not 100x/sec :)
23:11 karolherbst: nwm: do you monitor your system like temperature and stuff?
23:12 karolherbst: allthough that shouldn't really cause wakeups
23:12 nwm: imirkin, it's a headless machine w/ ssh
23:12 karolherbst: maybe with a power sensor it does
23:12 imirkin: nwm: irrelevant.
23:12 nwm: well no desktop stuff installed just default debian 9 and openssh-server
23:12 nwm: oh
23:13 imirkin: if there's a floating hpd pin
23:13 imirkin: you'll get interrupts up the wazoo
23:14 imirkin: could also be something equally dumb like the kernel polling for edid all the time