05:14imirkin: RSpliet: fyi i pushed your nv50 mad changes
08:51RSpliet: imirkin: cheers
08:59imirkin: hope you tested... coz i didn't :)
09:16RSpliet: imirkin: of course I did
09:38ki7rw: i was told that this is the place to ask about the linux nvidia drivers including nouveau - my mouse randomly stops responding but if i do this key sequence the mouse starts responding again: ctrl-alt-f1 then ctrl-alt-f7 - ubuntu 14.04 64 bit currently using nvidia-331 driver (the nouveau driver does the same thing)
09:39ki7rw: after the key sequence the desktop comes sometimes comes back quickly and sometimes very slowly
09:41imirkin: ki7rw: i gtg, but it'd be nice if you could supply some sort of information about your system. dmesg + xorg log would go a long way (under nouveau, for nvidia blob driver, try #nvidia)
09:41imirkin: ki7rw: it kinda sounds like you're having interrupt delivery issues... or something
09:41imirkin: er wait... nevermind. misread. ignore that.
09:48ki7rw: imirkin, http://pastebin.com/eykV5VcS
13:26Manoa: hi, im trying to get my nouveau boat to float, I have the 3.19 kernel and GT210 videocard, when I am loading the nouveau module I am getting: BUG: unable to handle kernel NULL pointer dereference at 00000040, and then the screen goes blank and system not responding
13:27Manoa: when I am loading the kernel with statically built nouveau driver and nomodeset, everything works and I can even run X with the VESA driver
13:28Manoa: hopefully this helps to indicate a problem with specifically with modeset ?
13:28RSpliet: Manoa: the full backtrace would be greatly helpful
13:28RSpliet: a GT210 should work just fine :)
13:28RSpliet: it's a kernel panic on boot; are you able to capture the contents using 1) netconsole, 2) serial console or 3) a photocamera?
13:29Manoa: I doubt question marks will be helpful to you xD, how to build a kernel with full debugging symbols ?
13:29RSpliet: there's a kernel option for that... somewhere
13:29RSpliet: good question, no idea where it's hidden :p
13:29Manoa: right, I will get right on it
13:29RSpliet: somewhere in kernel hacking I think
13:30RSpliet: anything else on this system that might be silly? is it a 32-bit build?
13:30Manoa: yhe 32-bit
13:30RSpliet: ok... yeah, I guess a lot of code is ill-tested on i686 nowadays on account of very little users still using it (and no devs)
13:31RSpliet: but it should work, and it is definitely a bug :)
13:32RSpliet: I'll probably be gone in a little, but I request you to fetch this backtrace using one of the methods I suggested (netconsole is surprisingly easy if it's on your home network! :))
13:32RSpliet: and file a bug in bugzilla
13:32Manoa: I should mention that nouveau did not have this problem in the last time I checked it (the 3.15 kernel)
13:32RSpliet: you may
13:32Manoa: right, I will report, thanks
13:33RSpliet: make sure to exclude build goof-ups
13:33RSpliet: and if you did, I think you have a good reason to mark it a high prio regression :p
13:34Manoa: there might be an other detail I did not mention because I don't think it has relevance,
13:34RSpliet: it probably does
13:34pmoreau: Manoa: Remember than having nomodeset will disable Nouveau
13:34Manoa: the card is connected to a DVI>HDMI cable
13:34Manoa: it converts the signal from DVI to HDMI
13:35RSpliet: hmm... if that somehow screws with the EDID it might explain why we haven't had more bug reports
13:36Manoa: ok so 2 things to do for me, try connecting to DVI and see what hapen and get the trace
13:36pmoreau: joi, RSpliet: How are EVO commands displayed in the mmt trace? I remember demmt doesn't handle EVO but it doesn't mean it removes them, does it?
13:37RSpliet: pmoreau: it doesn't
13:37pmoreau: Or maybe I need to add some options to valgrind-mmt / demmt to get them?
13:37RSpliet: it's been a while back, I can't remember much, sorry
13:37pmoreau: No problem :)
13:37RSpliet: all I recall is searching explicitly for parameters I expected
13:37RSpliet: like my resolution
13:38RSpliet: rather than actually trying to understand the full data stream
13:38pmoreau: I tried something like that, but I didn't find what I was looking for.
13:38pmoreau: Or maybe my grep wasn't totaly appropriated
13:39RSpliet: did you attach valgrind to xorg?
13:39RSpliet: then what could possibly go wrong :D
13:39pmoreau: I'm trying to get the initialisation of the EVO engine
13:40pmoreau: The whole world? :D
13:51Manoa: I assume trace (level 5) debugging is enough ?
13:51pmoreau: It should be
13:52Manoa: it appears to be taking significantly more time to compile a kernel with debugging enabled
13:54Manoa: ok I go boot the kernel with the cable and upload the trace to you, then try the same without the cable
14:00Manoa: ok im back, and with the kernel builded in debug mode everything is working, the module, modesetting, framebuffer console, X driver
14:02joi: pmoreau: demmt has 0 code for evo detection/decoding
14:04Manoa: yhe and GLX reports direct rendering enabled
14:05RSpliet: Manoa: damn, heisenbugs
14:07joi: pmoreau: if you tell dedma to decode proper map id you may see evo commands (but without names)
14:07Manoa: im going to try to load Doom 3, let's see what hapen :)
14:10pmoreau: joi: I don't care for names, just seeing W some_adress some_value would already be great. :-)
14:10pmoreau: So I'll need to try dedma
14:12pmoreau: joi: Would there be a way to tell demmt to at least keep commands he doesn't know about in the resulting file, so I could still use the sync between mmt trace and mmiotrace?
14:13Manoa: a small question, it appears that my normal users on this box have no direct rendering: libGL error: failed to open drm device: Permission denied, is there a standardized way to enable access to direct rendering to normal users ? I couldn't find this information in the website
14:16joi: pmoreau: demmt does not hide writes, however it does not print them immediately
14:17mlankhorst: Manoa: usually your distro sets it up..
14:17Manoa: they never had nouveau support in the distro, I added it manually, I compiled X, DRM, kernel and nouveau all alone
14:19pmoreau: joi: But they should appear at some point. I couldn't find any writes to 0x80 or 0x84 - which flush the engine - in the trace, or they were unrelated to that operation.
14:19mlankhorst: Manoa: in that case just chmod 777 /dev/dri/*
14:19Manoa: yhe it worked, thanks :)
14:21joi: pmoreau: demmt keeps writes in memory until next ioctl/write to another buffer/syscall/memory read
14:21joi: pmoreau: can I see the trace?
14:23pmoreau: joi: I don't have the binary, only the resulting sync with mmio as I used the mmt-app-demmt-mmiotrace script
14:24pmoreau: I can grab a new trace if you want
14:25joi: yeah, please do
14:26pmoreau: I'll go without the sync thing
14:43pmoreau: joi: Trace sent. And it looks like my mail wasn't quarantined by Google this time! My mail server is finally correctly configured! :D
14:45joi: pmoreau: where did you send it?
14:46joi: ah, it was in spam dir
14:46pmoreau: :/ But it was not marked as quarantined...
14:55joi: pmoreau: your lucky numbers are: 3 7 14 38
14:56joi: demmt -l log-file.bin.xz | grep -e "w 3:" -e "w 7:" -e "w 14:" -e "w 38:"
14:56pmoreau: I'll have a look
14:58joi: fyi, I got those numbers by looking at the output of: for c in `seq 1 100`; do xzcat log-file.bin.xz | mmt_bin2dedma | dedma -m nv96 -v $c -; echo $c; read; done
15:05pmoreau: So that was it. I saw them but I thougth it was something else as the address was quite always the previous address + 0x4, which seemed to represent something else
15:39Manoa: hey guys this is amazing, I actually played Doom 3 on nouveau, it worked ! you guys win 10,000 internets !!
15:40pmoreau: Manoa: So with debug, it works?!
15:40Manoa: I haven't disabled the debug
15:41Manoa: but yhe it works well
15:41Manoa: it crashed eventually but there were no graphical anomalies at all i nthe rendering
15:41pmoreau: Cool! :)
15:41Manoa: missing some soft shadows and anti aliasing but basically fully functional
15:41pmoreau: What was the crash about?
15:42Manoa: I can't tell, there are just so many lines that look like: nouveau: 0xbf1b69d3
15:42Manoa: and it exceeded my console buffer size
15:43Manoa: and the logs are not available because they filled the hard disk, I may have set the log level too high lol
15:44pmoreau: Which log level did you set? :D
15:44Manoa: it was set to 5
15:44Manoa: but I enabled some more aggressive kernel logging functionality
15:45Manoa: this is a long way apart from what I experienced in 3.15, a whole different world !
16:49martm1: prolly has happned before_ a death came finally being 31 years old..You know THE cause of death..Natural! its something that seems to happen to me
17:09martm1: during 31 years its around 9000days it happened so that none never understood what i was talking about
17:24martm1: i think its one of the reason why life is lived so quick, when you are regarded to be so retarded that dad just says 20first years that i do not understand a damn thing you say
17:36martm1: my understanding about it has been that this guy has feelings then, i would before i go would want to put together a group this bold guy can be found in cambodia, perhaps he better feels when dick is put into his ass
23:29imirkin: gnurou: your recent patch series reminds me... i never got around to reviewing/pushing your gk20a patches. i've completely lost all context about them... could you resend them to the list (assuming you still think it's the right approach)
23:45imirkin: are vm addresses 40 or 44 bits? i thought 40
23:45imirkin: but nv50_instobj_rd32 does u64 base = (node->mem->offset + offset) & 0xffffff00000ULL
23:46imirkin: that's 1 f too many, no?
23:46imirkin: skeggsb: --^