02:29gnurou: karolherbst: mupuf: shader-db working fine on Jetson TK1
07:10mupuf: gnurou: ok, which kernel?
07:10gnurou: mupuf: -next with latest Nouveau module
07:11gnurou: (which doesn't include karol's revert yet"
07:11mupuf: ok, cool
07:11mupuf: I will update
07:11gnurou: if you can repro, please let me know
07:11mupuf: well, yeah, I am using an old kernel, so not cool!
07:12gnurou: but it seems Karol's issues was distinct from yours since the cpu_coherent removal is still here
07:12mupuf: yeah
07:29karolherbst_phon: gnurou: yeah no idea anyway, it helps me and somebody else. maybe it is also me just doing a bad rebase of nouveau
07:30gnurou: ¯\_(ツ)_/¯
07:32mupuf: karolherbst_phon: no, it helped reator too
07:32mupuf: gnurou: nice ascii art :)
07:33gnurou: (*^ω^)
07:35karolherbst_phon: ahh, k
07:35karolherbst_phon: thanks for verifying
08:33pmoreau: imirkin: I have the actual NV50 IR code, would that work? I’m not sure I can reproduce it properly with GLSL: it needs to be at least two blocks within the loop body for it to not work, if I merge the blocks, then it works.
15:41imirkin: pmoreau: hm ok. well, i guess it'll have to do.
16:33jrun: i have 2 gpus here. the dgpu is nvidia and i just moved to nouveau from nv blob but X refuses to start.
16:33imirkin_: jrun: pastebin logs?
16:33jrun: the setup is that i have to monitors for each card and i run kmscon (via intel) on one of them and X on the other which is hooked up to nvidia card.
16:34jrun: i would idealy want to run everything on nvidia since kmscon does work with nouveau.
16:34jrun: https://gist.github.com/9ba1ac59f126d4dbcba513ee42fcab3e
16:35jrun: Xorg.log^
16:37imirkin_: well that seems like a giant pile of fail...
16:37imirkin_: i'm guessing kmscon is grabbing the drm master on those devices and X doesn't know how to get it back?
16:39jrun: i'm gonna reboot
16:42imirkin_: -22 = EINVAL... if that were the case i think you'd be getting EACCES (13) or something. i forget.
16:47imirkin_: jrun: not sure if you saw my last comment...
16:47imirkin_: jrun: -22 = EINVAL... if that were the case i think you'd be getting EACCES (13) or something. i forget.
16:47jrun: imirkin: do i need CONFIG_VGA_SWICHEROO ?
16:47imirkin_: only if you want auto-suspend to work
16:48imirkin_: which it won't unless it's a laptop with ACPI methods to make that happen
16:48jrun: kernel msg:
16:48jrun: tail: invalid option -- 't'
16:48jrun: Try 'tail --help' for more information.
16:48jrun: https://gist.github.com/53ce73eeceb8d17d68ec562b752eb718
16:48jrun: sorry
16:49jrun: it's a desktop
16:49imirkin_: modprobe.blacklist=nouveau
16:49imirkin_: oops?
16:49imirkin_: looks like you're loading nv blob
16:49jrun: probably, let me check
16:49imirkin_: nouveau after nv blob iirc doesn't work very well
16:49imirkin_: they do something to the display which we don't properly undo
16:50imirkin_: and so everything works fine, but the screen never updates :)
16:50imirkin_: at least it started doing that around nvidia blob 340 or so
16:50imirkin_: i dunno if they've changed stuff "back"
16:52jrun: i will remove the blob shortly and building the nouveau kernel driver *in*.
16:52jrun: would nouveau fb conflict with kmscon?
16:52imirkin_: that's not a super-great move. i'd recommend leaving nouveau as a module
16:52imirkin_: but just don't blacklist it ;)
16:52imirkin_: yes, but kmscon can always unbind the nouveau fb.
16:52jrun: although i have uvesa and it doesn't :)
16:53imirkin_: iirc uvesa messes things up bigtime for nouveau
16:53imirkin_: but i don't have details, perhaps it can be used safely
16:53jrun: would that be FB_NVIDIA?
16:53imirkin_: no
16:53imirkin_: that's nvidiafb
16:53imirkin_: that won't even load on your gpu
16:54imirkin_: (you have a GF100 right?)
16:54jrun: *legacy*
16:54imirkin_: but nvidiafb is a totally separate driver from nouveau
16:54jrun: is that new?
16:54jrun: i mean nvidiafb?
16:54imirkin_: it's new as of like 2005 or so :p
16:55jrun: whwere is nouveaufb?
16:55imirkin_: it's created by nouveau
16:55jrun: is it hw accelerated?
16:55jrun: oh
16:55imirkin_: yes, unless you disable that explicitly
16:55imirkin_: (nouveau.nofbaccel=1 would turn that off)
16:56jrun: no i mean nvidiafb?
16:56imirkin_: nvidiafb is also hw-accelerated
16:56imirkin_: as is rivafb
16:56jrun: so all i need is DRM_NOUVEAU?
16:56imirkin_: and CONFIG_FBDEV_EMULATION
16:57imirkin_: [that might not be the exact name]
16:57jrun: VGEM?
16:57imirkin_: no
16:57imirkin_: that's for ... something.
16:57imirkin_: i forget
16:57imirkin_: either qemu or vmware
16:58imirkin_: CONFIG_DRM_FBDEV_EMULATION
16:58imirkin_: i had the right idea.
16:59jrun: i have DRM_BOCH for qemu
16:59imirkin_: that's a different one
16:59imirkin_: VGEM is to support acceleration
16:59imirkin_: (and it's also for the kernel *inside* the vm)
16:59imirkin_: or not? like i said - i don't super remember
17:00jrun: FB_VIRTUAL?
17:00imirkin_: read the help?
17:00imirkin_: all you need for nouveau are the two things i said
17:00imirkin_: everything else you don't need
17:04jrun: i'm still looking for FB_EMULATION
17:04imirkin_: CONFIG_DRM_FBDEV_EMULATION
17:04imirkin_: it's a new option
17:05imirkin_: it used to always be on, but it was recently made optionals
17:05jrun: legacy fb, found it
17:05imirkin_: turning it off destroys the universe.
17:05imirkin_: are you using menuconfig?
17:05jrun: nconfig
17:05imirkin_: if so, you can use / to search for stuff
17:05jrun: F8
17:05imirkin_: ah. dunno about nconfig.
17:06jrun: DRM_DP_AUX_CHARDEV; what is that?
17:06imirkin_: read the help for it.
17:06jrun: i did it just didn't make sense which probably means i don't need it?
17:06imirkin_: it's to give you access to the DP AUX channel as a character device
17:08jrun: it read/writes into some register that i have no idea about.
17:08jrun: i was black listing nouveau on bootloader; i should be fine on the next boot.
17:39jrun: hmm, kmscon is shotting off both cards now :)
17:39jrun: my monitors are mirrored; how is that even possible?
17:40imirkin_: i assume that's what kmscon does by default
17:43jrun: oh, it has a --gpu switch