10:32avph: Hi I'm having problems to understand what happens/should happen when a system has 2 pcie gpus (nvidia) using nouveau
10:33avph: Depending which card is in which pcie slot, it can show something on the display(s)
10:33avph: is this normal or should xorg automatically configure this?
10:38pmoreau: I would say that changing the pcie slot will change the other in which the cards are initialised, and maybe one card will fail to initialise in one of the configuration.
10:38pmoreau: Could you post the dmesg of a working configuration and a non-working one please?
10:38pmoreau: avph: -^
10:40avph: well its not that something is wrong in dmesg. it is just me that doesnt understand Xorg I think
10:41pmoreau: Xorg shouldn't care about the pcie address, unless you told him to do so
10:42pmoreau: Could you paste you xorg conf file if you use one?
10:42avph: Well Xorg find displays that are on the 'main' card, main=card that displays bios
10:43avph: if I do something like xrandr --setprovideroutput 1 0 it can see the other display
10:43avph: Xorg is empty now :)
10:45avph: Is it possible to have a config where one card does all the work and the other does just the output?
10:46tobijk: yeah that is called reverse prime
10:47tobijk: mostly what you did with xrandr --setprovideroutputsource a b
10:47avph: so thats with xrandr --setprovideroffloadsink?
10:48pmoreau: avph: You could have a look at http://nouveau.freedesktop.org/wiki/MultiMonitorDesktop/ for multi-monitor + multi-card setup (if you haven't yet)
10:51avph: I sure have :), but I'll try it again to be sure I did edit it right...
10:54pmoreau: tobijk will most likely have better suggestions than I would; I don't know that much about xrandr & co.
10:54avph: anyway xrandr --setprovideroffloadsink gives me errors: X Error of failed request: BadValue
10:55avph: and just xrand doesnt show any output on the second card before xrandr --setprovideroutputsource
11:01tobijk: avph: does xrandr recognize the two cards?
11:01tobijk: xrandr --listproviders
11:02tobijk: not sure if you have to do --setprovideroffloadsink at all
11:14avph: So then basically the question is whether or not I should see possible outputs of the second card in xrandr...
11:16tobijk: if you don't - something is wrong
11:25avph: ok when I do xrandr --setprovideroutputsource 0 1 does provide all the output in xrandr -q, Is it possbile now to change the device that is 'working'
11:25avph: such that in glxinfo renderer is the other card?
11:26dardevelin: hello everyone not sure anyone saw my question on the syslog that i detected while booting but nonetheless I leave the syslog link again in the hopes of having some hints on what might be the issue
11:27dardevelin: sometimes it happens during boot and I can actually see it for a while SCHED message
11:28dardevelin: others it just boots but I think it's still there
11:28dardevelin: the card is a nvidia gtx 850m
11:30pmoreau: Which kernel version are you using?
11:30tobijk: dardevelin: well we fail miserably there:
11:30tobijk: Jun 1 03:26:09 ycradnileved kernel: [ 60.826127] nouveau [ PFB][0000:01:00.0] RAM size: 1975398418 MiB
11:31tobijk: Jun 1 03:26:09 ycradnileved kernel: [ 60.826128] nouveau E[ PIBUS][0000:01:00.0] HUB0: 0x110050 0xff00000f (0x19400310)
11:31tobijk: i'd like to have that knid of ram :D
11:31pmoreau: I think I saw that error somewhere else
11:31pmoreau: Either bug report or fix
11:31tobijk: mh if i'm right this was fixed lately
11:31tobijk: maybe in darktamas tree?
11:32dardevelin: I am using 4.0.0-1-amd64 SMP debian 4.0.2-1 (2015-05-11) x86_64 GNU/Linux
11:33dardevelin: tobijk, well I have been 5+years with 3GB
11:34tobijk: dardevelin: the vram is reported completely wrong, thats what my sentence meant to say, look at dmesge part i have posted
11:34pmoreau: Could this bug https://bugs.freedesktop.org/show_bug.cgi?id=89558
11:35dardevelin: yeah I guess
11:35dardevelin: the VRAM is suppose to be 4GB
11:35tobijk: ah yes it seems to be that
11:35tobijk: fix for it
11:35pmoreau: And http://cgit.freedesktop.org/~darktama/nouveau/commit/?id=660c79568bd8885b28ca02153f4b4602476f8d24
11:36tobijk: right :)
11:36dardevelin: humm ideas ?
11:36dardevelin: should I be building a new kernel ?
11:37tobijk: yep with these fixes included
11:37dardevelin: were these merged upstream already ?
11:37pmoreau: I don't recall seeing them being pulled
11:38tobijk: they will likely be meged in the 4.2 cycle
11:38pmoreau: They aren't in Dave's tree
11:38dardevelin: also do you guys recommend installing bumblebee ?
11:38dardevelin: or prime or something in those lines or not at all ?
11:38tobijk: mh i'm happy with prime
11:38pmoreau: Given they are fixes, they could still make it into 4.1-rc6?
11:38dardevelin: tobijk, work wise which one is the simplest ?
11:39tobijk: prime with dri3
11:39tobijk: DRI_PRIME=1 app
11:39pmoreau: Hum... 4.1-rc6 was already released... So yeah, most likely 4.2
11:39pmoreau: Could be back ported to stable though
11:39tobijk: (after you have a working setup)
11:40tobijk: pmoreau: maybe we should tell darktama to get this into 4.1?
11:41pmoreau: Most likely
11:43pmoreau: skeggsb: Do you plan on submitting 5ecb79cc3794bcca08a38c7e7c408c3ec53714c2 and 660c79568bd8885b28ca02153f4b4602476f8d24 to stable?
11:43tobijk: oh he is here... :)
11:45pmoreau: tobijk: He changed his nickname like half a year ago, or maybe even more? :þ
11:46tobijk: i tend to forget those things :/
11:46dardevelin: so if I build the kernel test it and sign it should make it more likely that it gets merged on rc6 right ?
11:46pmoreau:has to head back to writing...
11:48pmoreau: rc6 was already released. I don't know if airlied would accept it for rc7. Otherwise it will be merged in 4.2, and could be also sent to stable for inclusion in 4.1.x
11:48dardevelin: let me just reconnect my client in the other pc so I can do this
11:48dardevelin:hasn't built a kernel for a while
11:48dardevelin: so i will have to check that
11:48dardevelin: make sure the patch is there
11:48dardevelin: and so on
11:48dardevelin: if that makes sense
11:49tobijk: for ease you can just build the tree above
11:49dardevelin: let me get to the other pc before the links
11:50pmoreau: Oh, you could also download one of our images
11:50tobijk: ah we have a 4.1-fix tree branch there: http://cgit.freedesktop.org/~darktama/nouveau/log/?h=linux-4.1
11:51pmoreau: I'm not sure you should consider them as fix branch
11:52tobijk: right, but it was merged into 4.1-rc6
11:52tobijk: i should have checked that
11:52pmoreau: Really? I missed that
11:53tobijk: dardevelin: i found the fix in 4.1-rc6
11:54tobijk: you can just grep that one
11:56pmoreau: Ah!! They were in drm-fixes, I only checked drm-next!
11:58dardevelin: tobijk, ah that's great
11:58dardevelin: so maybe I can actually just install it from the repo
11:59dardevelin: i think debian has that rc
11:59dardevelin: let me check
11:59dardevelin:uses debian unstable + experimental repo
12:00dardevelin: just curiosity anyone with a asus laptop with (the ones that bring a subwoofer) managed to make it work ? I know is not necessarily nouveau related but just checking
12:01dardevelin: just checked. I will have to build
12:01dardevelin: not a big deal though
12:02dardevelin: the machine is pretty fresh so I will have to set the build stuff first but isn't too bad
12:03pmoreau: dardevelin: If you just want to test, you could download an image from https://nouveau.pmoreau.org/ and boot the Nouveau entry (iirc the name)
12:03tobijk: mh i shoul create one as well, to get another flavor :D
12:06dardevelin:will have to setup build sets anyways and to have it properly following debian guidelines
12:06dardevelin: but thank you very much
12:07dardevelin: I will attempt to build and if I get too frustrated I will try the image.
12:50dardevelin: tobijk, btw gnome is suppose to report intel haswell mobile vs gallium correct
12:51dardevelin: because of having 'both' cards
12:52tobijk: dardevelin: was that a question or a statement? i dont have a clue about gnome ;-)
12:53dardevelin: sorry, question
12:53dardevelin: well it doesn't really matter if is gnome in this case. the reported card is suppose to be intel because the nvidia is connected via the intel one
12:53dardevelin: is what I want to know..
12:53tobijk: sorry cant help you there
12:53dardevelin: I know some lenovo laptops have actually the two cards super separate
12:54dardevelin: you can even enable/disable one or the other in bios
12:54dardevelin: but in this case (gtx 850m ) i think it always reports intel due to the way they are connected
12:55tobijk: thats the old version of optimus where only one card could driven at a time
12:56dardevelin: I see
12:56dardevelin: you have an optimus ?
12:56dardevelin: if so, what do you see reported in a 'system monitor or system details view' ?
12:57tobijk: huh? is that a gnome thing? :D
12:57dardevelin: tobijk, let me guess are you a kde user ?
12:57dardevelin: or i3 ?
12:57tobijk: right, kde
12:57dardevelin: ok. so you should also have some sort of tool that shows the info about your system
12:57tobijk: lspci -v ?!
12:58dardevelin: eh that's command line mate
12:58dardevelin: that always gets it right
12:58tobijk: yeah works everywhere ;-)
12:58dardevelin: yeah, that's not what I am trying to discover though. I am trying to see if is expected behavior on gnomes side
12:58dardevelin: or not
12:58dardevelin: but we can go with a lspci -v
13:00tobijk: well explain a bit more what you want to do
13:00pmoreau: Maybe in this case you're looking for the output of `glxinfo | grep "OpenGL vendor string"`
13:02dardevelin: 12 lines paste ?
13:02dardevelin: give or take ?
13:03dardevelin: there you go
13:03dardevelin: it's doing the right thing. correct (just making sure)
13:03tobijk: it is for the intel card
13:04tobijk: if you want the nouveau one, go wih DRI_PRIME=1 glxinfo
13:04pmoreau: Is accel working on Maxwell?
13:04dardevelin: pmoreau, how to test ?
13:05dardevelin: the full lspci -v
13:05dardevelin: glxgears ?
13:06pmoreau: You could look in the Xorg.log for some acceleratio fail messages
13:06tobijk: pmourea: ah right maxwells need the signed firmwares?!
13:07pmoreau: But I think it should be working for GM107 chipset
13:07pmoreau: tobijk: Only GM20x iirc, not GM10x
13:07tobijk: ah ok
13:08dardevelin: during the debian installed it asked if i wanted to use the proprietary firmware for wifi and realtek and I provided the files for those
13:08pmoreau: dardevelin: Did `DRI_PRIME=1 glxinfo` says Nouveau as vendor string or still Intel?
13:08dardevelin: still intel
13:08pmoreau: Remainds me of another bug report, one sec
13:09dardevelin: I haven't built the kernel yet
13:09dardevelin: let me do that run that kernel see if it happens and run this again
13:10pmoreau: Ah! :D
13:13pmoreau: It was for a GK106M, nevermind.
13:15dardevelin: brb reboot
13:25dardevelin: pmoreau, is there a git repo ? the links you gave me were bug and images right
13:25dardevelin: I guess I just get it from kernel.org
13:25pmoreau: A git repo for what?
13:25dardevelin: as it has like 4.0.4 now I think
13:26dardevelin: where it has the patch applied
13:26dardevelin: linux kernel
13:26pmoreau: There are multiple ones
13:27dardevelin: ok. any particular order I need to patch the kernel then ?
13:27pmoreau: You could use Linus' one https://www.kernel.org/, which contains the patch
15:25dardevelin: while building the kernel do i need nVidia Framebuffer Support ?
15:25dardevelin: recommended is 'No'
16:46dardevelin: ok built rc6 nothing solved, built 4.0.4 and nothing solved (regarding the errors)
16:47dardevelin: I guess I need to build something else
16:59dardevelin: So i built 4.0.4 from kernel.org and this is my dmesg http://paste.debian.net/195992/
18:57dardevelin: some assistance needed please
18:57dardevelin: how to obtain the repo ? http://cgit.freedesktop.org/~darktama/nouveau/
18:57dardevelin: I don't know how the cgit interface works to get the 'clone-enabled url'
18:59skeggsb: they're right down the bottom of the page
18:59skeggsb: under "Clone"
19:05dardevelin: I am so ridiculous
19:05dardevelin: does this get me the linux-4.1 too ?
19:05dardevelin: I guess not
19:05dardevelin: that's a branch
19:11dardevelin: skeggsb, ping
19:11dardevelin: skeggsb, uh I see these are not patched kernel
19:12dardevelin: earlier someone told me that there are some images/branches with kernels already patched
19:12dardevelin: skeggsb, not sure you saw my pasts http://paste.debian.net/195992/
19:12dardevelin: it's an nvidia gtx 850m
19:12dardevelin: I built 4.4
19:13dardevelin: the other question I had is do I have to put yes to 'nVidia Framebuffer Support' option in the kernel
19:14skeggsb: no, definitely not
19:15skeggsb: my tree isn't a normal kernel tree, you can build the drm from it with "cd drm && make".. but, if you want something more standard: http://cgit.freedesktop.org/nouveau/linux-2.6/log/?h=linux-4.1
19:16dardevelin: skeggsb, not normal in what sense ?
19:16skeggsb: it's not a linux tree
19:16dardevelin: so I build drm folder
19:16dardevelin: and then what ?
19:16skeggsb: it just has the nouveau drm, and an alternate userspace build of the driver that i do the vast majority of my development with
19:16dardevelin: move that folder and build my kernel with it ?
19:16skeggsb: it's not really meant for end-users
19:17skeggsb: just use the other link i gave you instead, it has the patches you want in it
19:17dardevelin: ok but ho to clone that specific one is my question then
19:17dardevelin: because I did got that link/tree
19:17dardevelin: as in got there but that's why i was asking about it
19:18dardevelin: git://people.freedesktop.org/nouveau/linux-4.1 ?
19:18dardevelin: this ^ does not work 'access denied'
19:19dardevelin: or do I have to do 2.6 and 4.1 will be a branch ?
19:19dardevelin: because 2.6 does work
19:19tobijk: dardevelin: google git fetch, git remote and git checkout...
19:19dardevelin: tobijk, I know basic git. my issue is just specifically on how this urls of cgit work
19:20tobijk: dardevelin: at the bottom of the pages are the git urls
19:20tobijk: just use them
19:21dardevelin: not really got to the link just above and you will see no links
19:21dardevelin: hence why I am asking
19:21dardevelin: try it yourself
19:21tobijk: its one click away...
19:21dardevelin: which implies a click in the right place
19:22tobijk: clone that: git://anongit.freedesktop.org/nouveau/linux-2.6
19:22tobijk: checkou the right branch
19:22dardevelin: well at that point it is what I asked which a yes would have done the trick
19:22dardevelin: <dardevelin> git://people.freedesktop.org/nouveau/linux-4.1 ?
19:22dardevelin: <dardevelin> this ^ does not work 'access denied'
19:22dardevelin: <dardevelin> or do I have to do 2.6 and 4.1 will be a branch ?
19:22dardevelin: <dardevelin> because 2.6 does work
19:22dardevelin: tobijk, ^ read this again ___
19:23tobijk: well i'm not gonna defend myself there :P
19:24tobijk: just try to seek around next time, the cgit interface isnt that bad :P
19:24dardevelin: I know probably you guys get a lot of folks with blinds in their faces and just want someone to hold their hands. I totally follow instructions when in such scenario
19:24dardevelin: but I am a dev. I just like to be sure I am following procedure when someone is 'helping' so if something f's up they know how to proceed
19:24dardevelin: tobijk, the home page shows it, but my issue was that those pages didn't had
19:25tobijk: fyi, its always at the overview page
19:26dardevelin: alright will keep that in mind
19:26tobijk: well its getting late, i'll go to bed now
19:30dardevelin: woops too late to tell him good night.
20:35dardevelin: so that was stupid on my part (I think)
20:35dardevelin: i got 2.6 checkout to linux-4.1
20:35dardevelin: built it
20:36dardevelin: indeed the SCHED message went away but the kernel is actually 3.18
20:36dardevelin: and some other things pop up
20:36dardevelin: dmesg ^
20:37dardevelin: that's the syslog of that tree