05:03 Un1corn: Hi, I want to stop using the Nvidia proprietary drivers but I need 2 screen support, is it possible in Nouveau?
05:10 RSpliet: Un1corn: you mean two monitors? sure*...
05:10 RSpliet: did you check the status matrix to see how well your particular graphics card is supported?
05:11 Un1corn: RSpliet: yeah two monitors, is it working by default? When I installed Ubuntu the second monitor showed the same thing
05:11 Un1corn: no, how do I check that?
05:11 Un1corn: I have GTX 970
05:11 RSpliet: hmm... that complicates matters slightly :-P
05:11 RSpliet: but a couple of hints:
05:13 RSpliet: 1) a GTX 970 is of the second maxwell generation. They are only very poorly supported by nouveau (no firmware, so no accel)
05:15 RSpliet: 2) Ubuntu generally ships with old software. Your luck wrt. modesetting (setting resolutions) is likely to increase with newer software (and we're only able to help with the very latest, eg. a 4.3 kernel, X.org 1.18 etc...)
05:15 Un1corn: RSpliet: I will also install Fedora, so I hope it will help
05:16 Un1corn: so I won't have two monitors with nouveau?
05:18 RSpliet: you can use two monitors, but expect things to be extremely slow
05:19 RSpliet: because nouveau can't really do OpenGL or even accelerated 2D for your graphics card yet
05:19 RSpliet: Fedora tends to be a better bet than Ubuntu wrt. running the latest greatest software, but if you prefer there are third party repositories for Ubuntu to achieve something similar. Don't ask me for details though
05:27 Un1corn: RSpliet: when will my card have better support? I want to stop use proprietary programs and the drivers are the last one
05:29 pmoreau: Un1corn: Once NVIDIA releases the signed firmware, probably first quarter 2016.
05:29 pmoreau: Or maybe later, but it's a WIP
05:35 karlmag:crosses fingers for that..
06:49 jarnos: imirkin, oh, now system detects the pci video card 0000:02:08.0 also with the MX 420. Maybe the video card was not well connected before or something.
10:33 karolherbst: does somebody want to look over my pcie patches again? I changed some bits, but overall it is the same though
10:33 Tom^: ive ran them for quite some time now without hickups tho :p
10:34 karolherbst: :D
10:34 karolherbst: yeah I have quite a good understanding how that kind of stuff works, so there shouldn't be any technical issues
10:34 karolherbst: just code issues
11:48 joi: karolherbst: you should use your name and email in copyright headers - "Nouveau community" is not a legal entity
12:06 karolherbst: joi: well I saw it in other files :/
12:07 joi: then they are wrong
12:08 karolherbst: k
12:26 josla972: any nv50 vbioses rom dumps available? I want to compare my g84 bios with a >= g94 bios
12:31 josla972: nvm, I found one on bz
12:43 mupuf: karolherbst: yep, my fault, I should never have used this as a copyright header
12:50 Tom^: karolherbst: is it just my game that segfaults on exit?
12:53 mupuf: karolherbst: no, I do not have the algo
13:12 karolherbst: mupuf: k
13:27 karolherbst: mupuf: do you have any clue? It looks to be above (min + max) / 2 in general, maybe using this might be enough for now until this gets properly REed
13:31 ifwe: i have huge problems with most recent linux kernel and nvidia 7000 series (onboard gpu)
13:32 karolherbst: ifwe: then tell us :p
13:32 ifwe: http://pastebin.com/MpSFbCXi
13:33 karolherbst: ifwe: does your machine freezes after that?
13:33 karolherbst: or do you notice anything else being off after that happens?
13:33 ifwe: this is at the plasma5 login screen. when i enter password and press enter, the machine freeze, kernel halt/panic and capslock is flashing
13:33 karolherbst: k, the flashing means the kernel paniced, right
13:34 ifwe: so the only way to use the machine now is to go to tty2 and use command line
13:34 karolherbst: that indeed sounds serious then
13:34 karolherbst: ifwe: do you know if it worked with a 4.2 kernel?
13:35 ifwe: karolherbst: i dont have tried any other os
13:35 karolherbst: there is a change this is a regression from the rework landed in 4.3
13:35 karolherbst: *chance
13:35 ifwe: karolherbst: you are a nouveau developer?
13:35 karolherbst: maybe it is already fixed upstream
13:36 karolherbst: ifwe: either way I have no clue about teslas or desktop systems ;)
13:36 karolherbst: I can just ask you to check out a 4.2 kernel or the newest nouveau module
13:36 karolherbst: pmoreau: is your live cd fixed by the way?
13:37 ifwe: karolherbst: what linux iso would you like to use for testing 4.2?
13:38 karolherbst: ifwe: are you on debian experimental?
13:38 ifwe: karolherbst: its the debian testing iso from today with a fully basic install by selecting kde in the gui installation
13:38 karolherbst: or jessie?
13:39 karolherbst: then you could simply try out jessie
13:39 karolherbst: jessie has a backport package for 4.2
13:40 ifwe: karolherbst: cant i just say "downgrade kernel" or use any linux iso you prefere with kernel 4.2?
13:40 karolherbst: ifwe: or
13:40 karolherbst: ifwe: try out 4.4-rc6 first
13:40 karolherbst: because this you can simply upgrade to on your installation
13:40 ifwe: where did i get 4.4?
13:40 karolherbst: linux-image-4.4.0-rc6-amd64 is the package name
13:40 karolherbst: ifwe: https://packages.debian.org/experimental/linux-image-4.4.0-rc6-amd64
13:41 karolherbst: you should be able to install it through apt
13:41 karolherbst: apt-get install linux-image-4.4.0-rc6-amd64 might be enough?
13:46 ifwe: karolherbst: still not working
13:46 ifwe: with 4.4.0
13:46 karolherbst: ifwe: okay, the chances were bad from the beginning anyway
13:47 karolherbst: well trying out 4.2 would be the next step then
13:47 ifwe: karolherbst: can i help the devs here to debug that?
13:47 ifwe: are you a dev?
13:47 karolherbst: yes by trying out 4.2 ;)
13:47 karolherbst: ifwe: well more or less I would say so, I am still a bit new to nouveau
13:48 karolherbst: thing is, there was a big rework of nouveau in 4.3
13:48 karolherbst: and stuff might be broken because of that
13:48 Tom^: hes not only a dev, he is our kepler reclocking hero.
13:48 karolherbst: yeah lol
13:48 ifwe: ah, nice. thanks to him there are now the phoronix benchmarks with propper speed :)
13:52 karolherbst: I have no clue about all that vblanc stuff though :/
13:55 ifwe: karolherbst: is that fine for you for testing? http://distrowatch.com/?newsid=09236
13:55 karolherbst: no clue
13:55 ifwe: based on debian stable with 64bit 4.2 kernel
13:57 karolherbst: well you could try it out, but it might be, that it only triggers when special software starts or something
13:57 karolherbst: sadly, debian experimental doesn't have a 4.2 kernel anymore :/
14:01 ifwe: i also download plasma-wayland-201512181217-amd64.iso .maybe that have 4.2 kernel and you are interested into knowing if it crashes
14:02 karolherbst: yeah it it has a kernel older than 4.3 it should be fine
14:02 ifwe: its nowhere explained what is inside
14:02 ifwe: https://dot.kde.org/2015/12/18/first-plasma-wayland-live-image
14:04 karolherbst: mhh
14:04 karolherbst: ifwe: I think downloading debian jessi and install kde through that is the surest way
14:04 karolherbst: maybe even faster, because the iso isn't that big I think
14:05 ifwe: i am already dd'ing the 4.2 based iso i listed first
14:05 pmoreau: karolherbst: Unfortunately not… I tried a couple of times to fix it, but it's still complaining about X ABI version between the X server and the Radeon DDX… --"
14:07 Tom^: isnt your livecd sort of a nouveau testing ground, isnt it simpler to only build nouveau and remove any other drivers? :P
14:08 karolherbst: Tom^: what about those radeon+nvidia setups :O
14:08 karolherbst: :D
14:08 karolherbst: I doubt there are any
14:08 karolherbst: but who knows
14:08 pmoreau: Tom^: That was the case initially, and then came people asking for other drivers as well
14:08 Tom^: heh
14:08 Tom^: i can see the point of having intel around for those with optimus laptops.
14:09 Tom^: but radeon sounds far fetched
14:09 karolherbst: pmoreau: weird
14:09 l1k: pmoreau: thanks a lot for trying my patches. about this issue that the GPU locks up when woken... you write that it worked with 4.3.3. do you mean with older patches from me?
14:09 karolherbst: pmoreau: maybe the module path is just polluted?
14:10 karolherbst: pmoreau: something strange inside /usr/lib64/xorg/modules/drivers/ ?
14:10 pmoreau: ifwe: Do you have a full dmesg somewhere? I only glanced quickly on the logs, but which card do you have?
14:10 pmoreau: l1k: No, stock 4.3.3
14:10 karlmag: radeon+nvidia setups? Oh, there will possibly be some at least ;-)
14:11 l1k: pmoreau: hm, but how could switching work without my patches? the inactive GPU should not have been able to probe the panel resolution without my patches.
14:11 pmoreau: karolherbst: Dunno, it's when pacman tries to install the DDX on the live image, it looks like the installed X server reports ABI 20 but the DDX requires ABI 19
14:11 karolherbst: okay
14:11 karolherbst: pmoreau: then it is compiled against older headers files
14:12 pmoreau: l1k: Well, it's just that inactive card doesn't see the screen and defaults to a fb of 1024x768 or whatever resolution, so this is what you get when you switch without your patches.
14:13 l1k: pmoreau: right. okay so the GPU didn't lock up back then, now it does. hm... I'll see if my patches can also be applied to 4.3.3.
14:13 karolherbst: pmoreau: ohh wait
14:13 karolherbst: pmoreau: 1.17.4 is ABI 19
14:15 pmoreau: l1k: Exactly. This is also something I reported on earlier version of your patches (and said I would investigate, and didn't… --"), but it was after ~10 switches before it lockup.
14:15 pmoreau: karolherbst: 1.17.4 of which package? xorg-server?
14:15 l1k: pmoreau: my patches don't touch GPU sleep/wakeup code. and it seems implausible that the GPU locks up just because it has the proper resolution instead of 1024x768. so maybe it's a regression between 4.3.3 and 4.4-rc7
14:15 karolherbst: pmoreau: also check /usr/include/xorg/xf86Module.h:#define ABI_VIDEODRV_VERSION
14:15 karolherbst: pmoreau: yes
14:16 pmoreau: l1k: Could be, that's why I want to test on 4.4-rc7 without your patches
14:16 l1k: pmoreau: ok, understood now. thanks. :)
14:16 karolherbst: pmoreau: I have 1.17.4 installed and it is 19 for me
14:16 pmoreau: karolherbst: I have 1.18.0-3 installed
14:16 karolherbst: okay
14:16 ifwe: pmoreau: http://pastebin.com/z4MemknD
14:16 karolherbst: then the header file should contain 20
14:17 pmoreau: l1k: You're welcome. :-) Thanks for the hard workd on it!
14:17 ifwe: karolherbst: kernel 4.2 with xfce works: http://pastebin.com/z4MemknD
14:17 karolherbst: ifwe: mhhh
14:17 l1k: pmoreau: don't mention it. ;)
14:17 ifwe: karolherbst: but there is already [ 19.245415] nouveau E[ PBUS][0000:00:0d.0] MMIO write of 0x00820001 FAULT at 0x00b000
14:17 karolherbst: yeah, something like that happens :/
14:17 karolherbst: usually it doesn't do much
14:18 ifwe: and other problems like [ 8.107862] nouveau W[ DRM] DCB type 4 not known
14:18 ifwe: [ 8.107865] nouveau W[ DRM] Unknown-1 has no encoders, removing
14:18 karolherbst: ifwe: could you try to install kde on it then?
14:18 karolherbst: maybe kwin triggers it
14:18 pmoreau: ifwe: Could "simply" be that your card isn't powerful enough to drive all the composition stuff of KDE, since you have an NV40
14:19 pmoreau: Well, Nouveau could still fail somewhat better than a lockup
14:20 ifwe: pmoreau: the login screen(please enter password) already on 4.3.3 kernel already creates that: http://pastebin.com/pUK26XHK
14:20 karolherbst: I could imagine that the tearing prevention of kwin really locks up the screen
14:21 ifwe: karolherbst: i will try some kde on wayland based iso in few moments
14:21 ifwe: hopefully its kernel 4.2 or maybe its 4.3 but works and the problem is somehwere in debian. lets find out ;)
14:21 pmoreau: If you could get a whole dmesg when you get the lockup, that could be nice. Since we currently miss all previous error messages that could have been thrown before.
14:22 karolherbst: ifwe: do you have a path /sys/fs/pstore/ on your system?
14:22 karolherbst: and with bunch of files in it?
14:22 ifwe: karolherbst: on the installed kernel 4.4.0 system now?
14:22 karolherbst: whatever is running
14:23 ifwe: booting 4.4.0...
14:23 karolherbst: this is a path to some firmware storage
14:23 karolherbst: and the kernel writes dmesg on crashing in there
14:23 karolherbst: well
14:23 karolherbst: kmsg
14:24 ifwe: karolherbst: there is nothing inside pstore
14:24 karolherbst: ifwe: can you check with mount if its mounted?
14:24 karolherbst: mount | grep pstore
14:25 ifwe: karolherbst: yes, its mounted rw
14:25 karolherbst: k
14:25 karolherbst: mhhh
14:25 karolherbst: it is hard to get the kernel log when the kernel is crashing
14:25 karolherbst: because usually stuff doesn't get written to the disc in time
14:25 imirkin: ifwe: unfortunately kernel 4.3 contains a nasty regression for pre-nv50 gpu's
14:26 imirkin: i believe this should be fixed in the latest 4.4-rcN
14:26 karolherbst: imirkin: rc7?
14:26 ifwe: hi imirkin rc6 is still broken
14:27 karolherbst: ifwe: mhh could you test with uname if 4.4 is indeed running?
14:27 karolherbst: ifwe: uname -a
14:27 ifwe: karolherbst: yes, its 4.4
14:27 karolherbst: k
14:27 imirkin: ifwe: ok, give me a minute, i need to check then
14:29 imirkin: hm, the commit i had in mind made it into 4.4-rc4
14:29 ifwe: so rc6 should not make such a kernel panic and kill whole system
14:29 ifwe: i am glad i can help now testing :)
14:29 imirkin: ifwe: your log at http://pastebin.com/pUK26XHK appears to be from 4.3 though
14:30 imirkin: could you supply a log from 4.4-rc4 or later?
14:30 ifwe: imirkin: i create now a 4.4
14:30 pmoreau: karolherbst: Meh… looks like the ATI DDX package is to blame, which is great since easier to debug. :-D
14:30 karolherbst: :D
14:30 karolherbst: pmoreau: did you check /usr/include/xorg/xf86Module.h ?
14:31 karolherbst: or aren't you compiling the readeon ddx?
14:31 pmoreau: karolherbst: I did check it, and it reports 20. Maybe for some weird reason the ddx wasn't recompiled? But I think I already checked that
14:32 karolherbst: or you have another file with such an aBI decleration
14:32 karolherbst: would be weird though
14:33 ifwe: imirkin: you are correct. its much more clean. still kernel panic: http://pastebin.com/N5ussLH2
14:33 pmoreau: karolherbst: It should be an error somewhere in the PKGBUILD, since pacman is the one complaining when trying to install.
14:33 ifwe: imirkin: i cant post a logfile from a kernel panic because i have to press reset button and cant get into a terminal
14:33 karolherbst: ifwe: yeah, that's why I wanted you to check pstate :/
14:34 karolherbst: it preserves logs across boots
14:34 karolherbst: after the kernel crashed real hard
14:34 imirkin: ifwe: i assume you have a second computer?
14:34 imirkin: karolherbst: pstate is a thing of the future. he has an old box.
14:35 karolherbst: yeah I already thought so much :/
14:35 ifwe: imirkin: yes, i have enought computer here
14:35 imirkin: ifwe: if so, you could set up netconsole
14:35 karolherbst: imirkin: I guess the old BIOS doesn't have such storage and it can only really be used with uefi in the avarage case
14:36 imirkin: karolherbst: exactly.
14:36 ifwe: karolherbst: i let it crash again (enter kde password and press enter) and there is again nothing inside the pstore
14:36 imirkin: ifwe: ok, see if you can use netconsole to stream the logs over
14:37 karolherbst: ifwe: yeah, I think it only works on newer hardware
14:37 imirkin: ifwe: sometimes you can even just ssh in, and run "dmesg -w" in the ssh, and get a lot of the info
14:37 imirkin: depends a lot on what's hanging and how
14:37 karolherbst: imirkin: well the LEDs are flashing so the kernel crashes real hard
14:37 imirkin: karolherbst: there are different levels of hard :)
14:37 karolherbst: don't think ssh will get the dmesg -w output in time
14:37 karolherbst: mhh worth a try though
14:39 ifwe: i sometimes also get the screen full of lines and cant also do nothing. no flashing of capslock led
14:39 ifwe: for example now i have such a situation again with 4.4.0 kernel
14:40 karolherbst: is it a kernel trace?
14:40 karolherbst: well you could make a photo and upload it
14:41 ifwe: there are just colored lines on the screen. i cant do anything. computer is frozen
14:42 ifwe: i got kde testing with kernek 4.2 running
14:42 ifwe: running in wayland
14:42 ifwe: i have so much screen glitches, i cannot see much and when i move the mouse i got more glitches
14:43 ifwe: maybe i can get some logfiles
14:43 imirkin: karolherbst: definitely worked for me in the past :)
14:45 pmoreau: karolherbst: Seems to be fixed :-)
14:45 ifwe: dmesg form plasma on kernel 4.2 running wayland: http://pastebin.com/NM1H2rmi
14:47 karolherbst: pmoreau: nice
14:47 ifwe: when playing around and checking again dmesg i see PBUS MMIO write FAULT
14:48 karolherbst: pmoreau: so the error was radeon wasn't recompiled against the new x server ;)
14:48 pmoreau: karolherbst: No, some issue in the PKGBUILD writing. I probably made some error when bumping the version and couldn't spot it, so I just copy-pasted from the official PKGBUILD. :-D
14:48 ifwe: karolherbst: any idea how i can get this logfile other then netconsole?
14:50 pmoreau: Why not try netconsole?
14:54 ifwe: imirkin: this is how dmesg -w looks like: http://pastebin.com/eTxbbnts
14:54 imirkin: hmm ok. those aren't unexpected
14:54 ifwe: this is the situation where i got now this lines all over the screen and everything freezed up
14:54 imirkin: for some dumb reason we try to initialize the video engine
14:54 ifwe: ssh also freezed
14:54 imirkin: and for some even dumber reason we get those errors
14:57 ifwe: imirkin: and again freezed with lines on screen: http://pastebin.com/QRp7MBvS
14:57 Tom^: what card did you have?
14:59 imirkin: ifwe: yeah, all those 00bxxx addresses are the video engine
14:59 imirkin: i wouldn't worry about them
14:59 Tom^: onboard 7xxx series, imirkin by shere coincidence there is a guy having trouble with 4.3 on his 7600 in #archlinux too
14:59 ifwe: Tom^: nouveau 0000:00:0d.0: NVIDIA C61 (04c000a2)
14:59 Tom^: and he claims rolling back to 4.2 fixes it
15:00 imirkin: Tom^: yeah, 4.3 has a nasty issue for pre-nv50
15:00 imirkin: Tom^: have him try 4.4-rc4 or later
15:00 Tom^: ah mk
15:00 ifwe: Tom^: GeForce 7025 / nForce 630a
15:00 imirkin: Tom^: specifically commit bbc8764f80eb
15:00 Tom^: imirkin: hm ok
15:01 ifwe: Tom^: i am using 4.4.0-rc6 and giving the logfiles from this kernel version
15:02 pmoreau: Why did I had Linux headers from 3.13 still installed on my laptop?! O.O Especially since I installed Arch way after 3.13…
15:02 karolherbst: :D
15:02 karolherbst: pmoreau: I love this kind of issues
15:02 karolherbst: you never expect them
15:02 karolherbst: but they are there :O
15:02 pmoreau: Tom^: I'll have a look at adding the headers for the drm-next version of Linux on the image
15:03 ifwe: karolherbst, imirkin : can i help somehow to fix this error?
15:03 Tom^: pmoreau: well yea now that i got bored of windows im actually on arch myself, so i dont use your isos any more.
15:03 Tom^: pmoreau: they were lovely when i did use them tho <3
15:03 ifwe: imirkin: can this video-engine be disabled with some kernel boot config?
15:03 imirkin: ifwe: set up netconsole and see if you get more info
15:04 imirkin: ifwe: you could also try removing libvdpau_nouveau.so but it's highly unlikely to be connected
15:04 imirkin: ifwe: i doubt the video engine has anything to do with anything
15:04 pmoreau: Tom^: Oh, you could write the PKGBUILD yourself then and send it to me for inclusion. O:-D
15:04 Tom^: =D
15:04 Tom^: yea maybe later.
15:05 pmoreau: Just kidding, I'm going to write it right now :-)
15:08 ifwe: karolherbst: can you help me setup neconsole on debian?
15:08 karolherbst: I never used it
15:08 ifwe: me either
15:09 karolherbst: ifwe: maybe this helps? https://www.kernel.org/doc/Documentation/networking/netconsole.txt
15:14 ifwe: some new information: i seem to always get this colored lines freeze without flashing capslock when i dont get to tty before. when i go to tty before and back to f7 and login then, i get kernel panic
15:15 imirkin: ifwe: quite frankly, pre-nv50 won't run plasma well if they have the GL compositor enabled.
15:15 ifwe: imirkin: but the system should not die...
15:16 imirkin: ifwe: yeah that's a nice thought
15:16 imirkin: ifwe: the reality is that nouveau often takes down the system :(
15:17 imirkin: i'm not saying that this is right, but it's definitely the reality
15:20 ifwe: i try now to somehow enable kdump
15:21 imirkin: ifwe: start with netconsole :)
15:22 ifwe: i have really no idea how to enable it :(
15:22 karolherbst: imirkin: is netconsole=some_port enough to get it working?
15:22 imirkin: the arg you have to give it is totally nuts
15:22 imirkin: sec
15:22 imirkin: i use this: netconsole=@192.168.3.101/eth0,@192.168.3.1/
15:22 imirkin: that means send packets from 192.168.3.101 to 192.168.3.1
15:22 karolherbst: ohhh :O
15:22 karolherbst: insane
15:22 imirkin: and then on the target machine, you just do 'nc -l -u -p 6666'
15:23 imirkin: and wait
15:23 imirkin: all the ports/etc stuff is configurable but defaults are ok
15:23 karolherbst: I think you can actually leave out the source thing
15:23 karolherbst: netconsole=@/,@10.0.0.2
15:23 imirkin: you're missing a trailing /
15:23 karolherbst: right
15:23 karolherbst: netconsole=@/,@10.0.0.2/
15:24 karolherbst: :D
15:24 imirkin: and yes, you can, but that relies on some stuff to be set up which might not be
15:24 ifwe: karolherbst: this as a boot option at the linux kernel line?
15:24 karolherbst: ifwe: yeah, but you need to figure out your ip addresses yourself
15:24 karolherbst: from the machine running the nvidia card
15:24 ifwe: karolherbst: thats not the problem
15:24 karolherbst: and the machine where you want to read it
15:41 ifwe: imirkin: i have now netconsole running (enabled after boot) but how can i test if its really on?
15:41 ifwe: i have not seen anything on the second computer
15:41 imirkin: on the computer it's sending to, you need to be running 'nc -l -u -p 6666'
15:41 ifwe: imirkin: it is
15:41 imirkin: for reasons beyond me some distros ship a netcat with slightly different arguments, so that might not work as-is
15:42 karolherbst: it could be that netconsole isn't built into the kernel?
15:42 ifwe: i logged in some tty. i should see a message in the netconsole, right?
15:43 imirkin: not necessarily
15:43 imirkin: but the boot messages should have showed up
15:43 imirkin: try running 'dmesg -n 8"
15:44 ifwe: i enabled it after boot. so no boot messages
15:44 ifwe: yes, its set to 8
15:44 imirkin: oh, then you should only see messages when things show up
15:44 ifwe: how can i make such "things"
15:44 ifwe: when it crashes i have to do everything once again
15:45 karolherbst: echo 5,bla > /dev/kmsg
15:45 karolherbst: ohh wait
15:45 karolherbst: it has to be 8
15:45 karolherbst: anyway, the message should show up in dmesg
15:46 ifwe: yes, i see 8,bla
15:47 imirkin: did you see it show up in netcat?
15:47 ifwe: on ssh, on screen of crashing computer and on remote machine running nc ...
15:47 imirkin: ok cool
15:47 imirkin: so all's well :)
15:47 ifwe: so i can tell it to crash now again
15:48 ifwe: THERE IT IS!!!
15:48 imirkin: share with the class? :)
15:48 ifwe: http://pastebin.com/uVBg84YX
15:49 karolherbst: null pointer how nice
15:49 imirkin: gyah
15:50 ifwe: you now have all you need?
15:50 ifwe: can i help some more?
15:51 RSpliet: ifwe: I think this is already quite helpful, although if you have debugging symbols, you could use gdb to find out which line this nullptr exception corresponds to
15:51 karolherbst: RSpliet: I don't think the kernel does that
15:52 karolherbst: read2addr magic needed
15:52 ifwe: RSpliet: i dont know what a gdb is
15:52 karolherbst: and fetching load address of modules and such
15:52 RSpliet: karolherbst: I *have* done that many times before
15:52 karolherbst: ohhh
15:53 ifwe: RSpliet: this are two debian machines. just tell me what i should type in console
15:53 imirkin: ifwe: could you file a bug at bugs.freedesktop.org, xorg -> Driver/nouveau ?
15:53 karolherbst: RSpliet: I bet it is this: http://lxr.free-electrons.com/source/drivers/gpu/drm/drm_irq.c#L1783
15:53 karolherbst: wake_up ...
15:54 RSpliet: karolherbst: how is that nv40_gr_intr+0x1fd ?
15:54 RSpliet: I think something along the lines of
15:54 karolherbst: ohhh
15:54 RSpliet: gdb nouveau.ko
15:54 RSpliet: l *nv40_gr_intr+0x1fd
15:54 karolherbst: right
15:56 RSpliet: but your module might not be equal to ifwe's, so if you think nv40_gr_init is not very transparent (inlining could make it difficult to judge where in the function things go wrong), talk ifwe through it and keep your fingers crossed :-P
15:57 RSpliet: alternatively you could look at the structs used, and find out which access is offset 0x20 into one of those structs.
15:58 RSpliet:&
16:00 ifwe: imirkin_: https://bugs.freedesktop.org/show_bug.cgi?id=93557
16:02 ifwe: imirkin_: this is like you asked to be reported?
16:02 RSpliet: ifwe: thanks, that's very helpful.
16:03 ifwe: RSpliet: are there often people with such problems? because maybe a distribution that delivers netconsole logs to some server owned by you devs could be usefull
16:04 RSpliet: ifwe: hahaha, wow, well, I haven't thought of such a set-up before
16:04 Tom^: you mean big brother spying by default?
16:04 ifwe: so the user just have to dd a iso and boot from that
16:04 karolherbst: RSpliet: I land before if (show)
16:04 RSpliet: could be useful in some occasions, but infrastructure is not our strong suit I believe :-P
16:04 ifwe: Tom^: yes, complete spying. of course running trough tor to encrypt everything from ISP and so on
16:04 karolherbst: RSpliet: maybe chid = pfifo->chid(pfifo, engctx); ?
16:05 RSpliet: karolherbst: I
16:05 RSpliet: 'll leave this bug for you for now if you don't mind :-)
16:05 karolherbst: :D
16:06 karolherbst: ifwe: okay, now we could do some crazy stuff
16:06 RSpliet: karolherbst: dealing with users 101 tip 1, don't call stuff crazy unless there's a chance of frying a card or if it requires plugging in a 2GHz logic analyser :-P
16:07 karolherbst: :D
16:07 ifwe: karolherbst: just tell me the lines i should enter in the console
16:08 karolherbst: on the faulty machine: git clone git://people.freedesktop.org/~darktama/nouveau
16:08 karolherbst: cd nouveau/drm
16:08 karolherbst: make
16:09 imirkin: ifwe: can you find the nouveau.ko on the relevant system
16:09 imirkin: ifwe: and then do "gdb /path/to/nouveau.ko" and then "disassemble nv40_gr_intr"
16:10 imirkin: RSpliet: yeah, 2ghz logic analyzer is crazy. 4ghz is much more reasonable :)
16:14 ifwe: karolherbst: something kernek related/build is missing and build fails
16:14 karolherbst: yeah
16:14 karolherbst: you need to install your kernel headers
16:14 karolherbst: or kernel dev package or however that is called on debian
16:17 karolherbst: ifwe: but before doing make again modify one file
16:17 pmoreau: Tom^: I have the package compiling :-) You can leave Arch and reuse the ISO now. :-D
16:17 ifwe: imirkin: you need the assembler dump?
16:18 Tom^: pmoreau: even if i said your isos was lovely, thats not gonna happend.
16:18 pmoreau: Tom^: :-p
16:18 imirkin: ifwe: that should tell us where the real error is
16:18 imirkin: ifwe: although i guess there's a bunch of that assembly in the dump you already provided, perhaps i can figure it out from that
16:19 imirkin: mmmm... no. a bit hard to tell
16:20 ifwe: imirkin: https://bugs.freedesktop.org/show_bug.cgi?id=93557#c1
16:23 imirkin: awesome thanks
16:23 ifwe: karolherbst: missing: scripts/Makefile.modpost:42: include/config/auto.conf:
16:24 karolherbst: imirkin: fyi it is nv40_gr_intr+0x1f2 in my binary
16:25 imirkin: that's super.
16:25 imirkin: chan ? chan->fifo->object.client->name : "unknown",
16:25 imirkin: i bet client == null or something
16:25 karolherbst: l points me to the if (show) { thingy
16:25 imirkin: :)
16:25 imirkin: my analysis as well
16:27 karolherbst: ifwe: mhh
16:27 karolherbst: ifwe: what did you install?
16:27 karolherbst: there should be something like linux-headers-4.4 or something
16:28 ifwe: karolherbst: apt-get install build-essential , kernel-headers-4.4.0-rc6-amd64, svn
16:28 karolherbst: well you won't need svn
16:28 karolherbst: :D
16:28 imirkin: hmmm.... otoh ->name is at a higher offset than 0x20
16:28 ifwe: i mean git
16:28 ifwe: not svn
16:28 karolherbst: k
16:28 karolherbst: mhh
16:29 ifwe: karolherbst: should i just setup ssh root for you on a linux distribution prefered by you?
16:29 karolherbst: not needed
16:29 karolherbst: there is just a package missing
16:30 karolherbst: no idea which one though
16:30 karolherbst: ahhh
16:30 karolherbst: linux-source
16:30 karolherbst: ifwe: there should be a linux-source package for 4.4.0-rc6
16:30 karolherbst: you need that too
16:31 ifwe: apt-get install linux-source-4.4.0-rc6-amd64 - not existing
16:31 karolherbst: linux-source-4.4
16:31 ifwe: https://packages.debian.org/search?keywords=linux-source-4.4&searchon=names&suite=all&section=all
16:32 ifwe: yes, installing...
16:32 ifwe: scripts/Makefile.modpost:42: include/config/auto.conf missing
16:34 karolherbst: why is this always such a pain :/
16:34 ifwe: karolherbst: thats why i asked if you would like to get your os on the machine here. you can do then what you want with it
16:35 karolherbst: k, if there is nothing important with that, it is fine by me
16:36 karolherbst: but then I could do all the other stuff as well too, so :D
16:36 karolherbst: imirkin: well I guess I will add some printks
16:36 karolherbst: imirkin: what do you want?
16:36 Tom^: all your base are belong to karol.
16:37 ifwe: you can use it for compiling 24/7 for any stuff you like. just dont do illegal stuff with it
16:37 imirkin: i think chan->fifo is null
16:37 karolherbst: k
16:38 imirkin: or i dunno. this is a pain. sticking some printk's in should be pretty easy
16:50 ifwe: you need me any more for anything?
16:50 imirkin: ifwe: if you can build your own nouveau kernel module, it'd be useful if you could throw a few printk's in to figure out wtf is going wrong
16:51 karolherbst: imirkin: maybe show is just above 2
16:52 ifwe: imirkin: i asked karolherbst about a linux distro he would like to have so that he can test as long as he like on it
16:52 karolherbst: ifwe: well the current state would be fine enough already
16:52 ifwe: imirkin: if you would like to do that, i can set that up for you
16:53 imirkin: nah... i bet ben will know exactly what's wrong from looking at it..
16:53 ifwe: karolherbst = ben?
16:53 imirkin: no
16:53 imirkin: karolherbst = karol
16:53 karolherbst: :D
16:54 imirkin: skeggsb = ben
16:54 ifwe: ok :D
16:54 karolherbst: we could still build nouveau and check out what is NULL
16:55 karolherbst: I would add printks for everything used in the if case and then we would know
16:56 ifwe: i know what distro karolherbst would prefere now. its gentoo https://github.com/karolherbst
16:56 karolherbst: :D
16:57 ifwe: i can install sabayon linux if you like that ;)
16:57 karolherbst: nah
16:57 karolherbst: just leave the machine as it is :D
16:57 karolherbst: its faster than actually reinstall stuff
16:57 karolherbst: also the bug triffers now
16:57 karolherbst: maybe it won't later anymore
16:58 ifwe: karolherbst: i should mail you the address to reach the machine to git@ ... ?
16:58 karolherbst: you can also pm me the stuff through irc
17:00 ifwe: ok. i set this up in some hours/days and give you access to the machine
17:01 karolherbst: yeah well.. then it would be faster to just fix that compile issue
17:01 karolherbst: also you have to be there too cause we have to netconsole the output agian
17:03 ifwe: karolherbst: ok. have to go now. thanks so far. lets see if we can fix this bug for the masses. this board/chipset/gpu is still on sale as new
17:04 karolherbst: what? :D
17:04 karolherbst: new... well
17:04 Tom^: huh you can actually but his old chipset still?
17:04 karolherbst: ohhhh
17:05 karolherbst: ifwe: you installed linux-headers-4.4.0-rc6-amd64 right?
17:05 ifwe: http://www.amazon.com/ASRock-N68-GS4-FX-Socket-GeForce/dp/B00O0PAALU
17:05 ifwe: and many other boards
17:05 Tom^: hm i be damn :p
17:06 karolherbst: ifwe: try to install linux-headers-4.4.0-rc6-all
17:06 karolherbst: ohh wait
17:06 karolherbst: wrong
17:06 karolherbst: ifwe: linux-headers-4.4.0-rc6-common
17:06 karolherbst: you need the common header package
17:07 ifwe: karolherbst: i have to go now. i will try to fix that btw. i will set this up so that you can do what you like with the machine
17:07 karolherbst: okay
17:27 imirkin: gah! i need ONE MORE REGISTER in the macro thing. why did they have to wire $r0 to 0 :(
17:28 imirkin: i guess i can use those SCRATCH "methods" to spill :(
18:10 imirkin: i should probably write this in pseudocode before trying to go straight to asm =/
18:32 imirkin: gr. what a ridiculous isa. no way to compare 2 numbers :(
18:32 imirkin: and can only branch on zero/not-zero
20:51 Kazune: how bad of an idea it is to enable reclocking on a NVC3 card?
20:51 Kazune: also, Happy New Year :D
20:51 imirkin: Kazune: there's no support for it...
20:52 Kazune: imirkin, but the status matrix shows it as 'Mostly' like all the others, so I guessed it was available just not enabled
20:53 imirkin: nope, the status matrix doesn't have "reclocking", it has "power management"
20:53 imirkin: which is a ton of various stuff
20:53 Kazune: Oh, I see
20:54 Tom^: imirkin: the feature matrix should support clicking on the "mostly" text which brings up a little info on what is working and whats being worked on. :P
20:54 Tom^: but thats just my humble opinion.
20:55 imirkin: Tom^: go for it
20:55 Tom^: but i dont know what is supported
20:58 Tom^: if i did i would gladly help with that
20:58 orbea: imirkin: in case you're interested, there is now a slackbuild for the nvidia firmware for video acceleration on nouveau. http://slackbuilds.org/repository/14.1/system/nvidia-firmware/?search=nvidia-firmware
20:59 imirkin: orbea: cool. might want to fix it to a specific checkout of extract_firmware.py
20:59 imirkin: in case i update it
20:59 orbea: yea, that might be a good idea
21:09 karlmag: orbea: that's neat :-)
21:10 karlmag: obviously makes me a bit sad at the same time (that it's needed/needed to do), but alas..
21:13 imirkin: for kepler+ it shouldn't be _too_ difficult to make open firmware
21:13 orbea: that'd be cool to see
21:14 imirkin: go for it :)
21:14 karlmag: imirkin: full feature?
21:14 imirkin: i think so, i haven't really investigated
21:14 karlmag: I seem to remember reading something about hardware lockdown.
21:14 imirkin: but on fermi, there was user-supplied firmware to do the decoding
21:14 imirkin: while on kepler, that's no longer the case
21:15 karlmag: which makes parts of the features needing a signed firmware to work
21:16 karlmag: I could be (and guess I hope to be) wrong
21:23 imirkin: starting with maxwell2 that's the case
21:23 imirkin: on kepler the only related thing is the encrypted video stuff which ... is not used at all on linux
21:23 imirkin: (at least not afaik)
21:24 karlmag: 'k