03:22 rmbeer: hello
03:22 rmbeer: https://ptpb.pw/7HZO
03:22 rmbeer: have problem with webkit2
03:24 rmbeer: i can't open the browser for post this in bugtrack
03:40 rmbeer: https://ptpb.pw/AVlY
03:41 rmbeer: i can't open the browser for post this in bugtrack. The browser is break.
04:57 AndrewR: imirkin, may be I found problem for nv43 xvmc and new kernels... see in https://bugs.freedesktop.org/attachment.cgi?id=129209 for example line nouveau: xvmc_bench[626]:00000000:00000080: ioctl: new vers 0 handle 00000000 class 0000506f route 00 token dbd048c4 object 00000000dbd048c4
04:59 AndrewR: imirkin, if I looked at correct file http://lxr.free-electrons.com/source/drivers/gpu/drm/nouveau/core/include/core/class.h?v=3.9 those classes are from nv50! (#define NV50_CHANNEL_IND_CLASS 0x0000506f)
04:59 imirkin: that's expected
04:59 imirkin: don't worry about that
04:59 imirkin: nouveau: xvmc_bench[626]:00000000:0000406e: ioctl: new vers 0 handle beef3174 class 00003174 route 01 token dd348880 object 00000000082dbc38
04:59 imirkin: that's the line that creates the NV31_MPEG class (3174)
05:00 imirkin: it goes through the init
05:00 AndrewR: imirkin, so, after trying them on nv4x hw we got fail from kernel ...but ..why XvMCnouveau.so tries to drive nv4x as nv5x ....
05:00 imirkin: and then returns a -19
05:00 imirkin: coz there's something with fallback logic that tries all the classes until it finds one that works.
05:00 imirkin: so the question is ...
05:01 AndrewR: imirkin, but can this logic at least temporarily fooled into trying nv4x classes first?
05:01 imirkin: why does it return a ENODEV??
05:01 imirkin: AndrewR: don't worry about that bit
05:01 imirkin: look at the line i pasted
05:01 imirkin: that's the one to worry about.
05:02 AndrewR: imirkin, well, but those -22 surprisingly close to this -19 ...
05:02 AndrewR: imirkin, and I can't find anything indicating mpeg unit per se failed
05:02 imirkin: that's a totally diff error
05:05 AndrewR: imirkin, well, but it may explain why my nv92 worked in this case..because those methods succesed there, kernel not complained, and xvmc lib continued happiely
05:05 imirkin: nope. not the reason. trust me.
05:05 AndrewR: imirkin, ok, ok ...
05:06 imirkin: gr. this code is all too roundabout, i don't know it well enough =/
05:07 imirkin: really just need skeggsb to glance at that log, i'm sure it'll be obvious to him
05:12 imirkin: i dunno what all the abstractions are, and they're much too opaque to figure out on my own
05:41 AndrewR: imirkin, http://marc.info/?l=dri-patches&m=145075455113616&w=2 - commit f6b1b5b7c9cf6667d169bad3b33a73e4fe2bc14c ..(don't worry, I just read all this in hope some understanding will finally come closer to me..)
05:42 imirkin: yep, well aware of that commit
06:00 imirkin: skeggsb: feel free to jump in anytime :p
06:02 skeggsb: imirkin: i haven't been entirely following, been heads-down on another matter :P
06:02 imirkin: skeggsb: nv4x pmpeg channel creation appears to fail with a ENODEV
06:03 imirkin: specifically on a nv43 (so not the nv44 variant of it)
06:03 imirkin: i have a vague recollection of you being unsure about the mpeg conversion
06:07 skeggsb: added a note to plug one in tomorrow and take a quick look
06:08 imirkin: thanks :)
10:14 hakzsam: imirkin: it has to be after ConstantFolding because we fold some ops to SHL IIRC
10:14 hakzsam: and this increases the use of SHLADD
10:17 imirkin: hakzsam: yes, but could we do it just a bit later?
10:18 hakzsam: imirkin: yeah, I think
10:18 imirkin: hakzsam: basically i want to let IndirectPropagation do its work
10:19 hakzsam: that should work
10:38 rmbeer: hello
10:38 rmbeer: for report in the bugtrack, also need all info from the system? like ram, gpu,etc? i unknown what or where get this info.
10:55 rmbeer: !topic
11:21 karolherbst: rmbeer: well, gpu makes sense and most of the time also the output of dmesg
11:23 rmbeer: karolherbst, i trying to report to bugreport, but with problem of nouveau and webkit i can't use any browser, also with not use the webkit never work good and lost all data of login...
11:23 karolherbst: ohh I see
11:23 karolherbst: this issue
11:23 karolherbst: afaik somebody is working on it to improve it
11:24 karolherbst: no idea when the stuff will be published though
11:24 rmbeer: humm, then not need my log of kernel and xorg?
11:25 rmbeer: i take error like -22 and -16 from nouveau driver
11:25 rmbeer: also with webkit appear error more strong and can break all the system
11:26 rmbeer: can past to a site from terminal and past with all links here...
12:20 RSpliet: oh god, just when I thought DRAM wouldn't get much worse than the tFAW latency ruining performance... GDDR5 turns out to have a t32AW as well
13:56 ladis: Hi there. I hit this bug: https://bugzilla.kernel.org/show_bug.cgi?id=156341 but cannot attach acpidump (timeout from server)
13:57 ladis: Still present on 4.10-rc6. Any progress? Happily test any patch and do any debugging
16:10 karolherbst: RSpliet: what do you mean?
16:58 imirkin: ladis: did you see the workaround at the end of that patch?
16:58 imirkin: er, bug
17:06 ladis: imirkin: sure, it works, but I's like to see fix, not a workaround
17:07 imirkin: ah ok. there is none atm. (also it's unclear that it's a nouveau problem...)
17:14 ladis: Also selecting "Discrete Graphics" instead of "Hybrid Graphics" in BIOS helps (no acpi_osi parameter needed)
17:14 imirkin: makes sense - no more runtime pm to mess things up, as the discrete gpu has to always be on
17:14 ladis: This is on HP ZBook. There's also Intel graphics on chip.
17:14 ladis: Yes
17:41 ladis: imirkin: bugreport is in NEEDINFO state. What info is missing? Seems related to documentation rather that logs from failing machines..
18:01 imirkin: sounds right
23:33 MarcinWieczorek: Hello guys
23:37 MarcinWieczorek: It might be a bad place to look for help, but I have no xorg now. imirkin?
23:38 imirkin_: is there a follow-up question?
23:39 MarcinWieczorek: I don't understand what you mean
23:39 MarcinWieczorek: Oh I think I actually do
23:39 ladis: MarcinWieczorek: Just ask...
23:39 imirkin_: your comment made it sound like you had a question but you didn't actually ask one.
23:39 MarcinWieczorek: Sooo xorg does not find screens on 4.10
23:39 imirkin_: xorg is not in the business of finding screens with KMS
23:40 imirkin_: do you get a high-resolution console?
23:40 MarcinWieczorek: Is that some known issue? Just asking, I'll look for help somewhere else
23:40 MarcinWieczorek: nope
23:40 MarcinWieczorek: low resolution
23:40 imirkin_: ok, so nouveau just didn't load i think
23:40 imirkin_: can you pastebin your dmesg?
23:40 MarcinWieczorek: NVIDIA drivers
23:40 imirkin_: ok
23:40 imirkin_: well i'm still going to need to see a dmesg from an affected boot.
23:41 MarcinWieczorek: yeah sprunge is unavailable at the moment. I need to work on it
23:41 MarcinWieczorek: Do you know any paste that I can use from the shell?
23:42 imirkin_: dmesg > foo.txt is a great way to save data.
23:42 MarcinWieczorek: yeah I know
23:42 MarcinWieczorek: I actually can use sftp to put it on my server
23:43 MarcinWieczorek: http://kosz.marcin.co/dmesg.txt
23:43 MarcinWieczorek: does this work?
23:43 imirkin_: there's no mention of nouveau.
23:43 MarcinWieczorek: I told you I'm using NVIDIA drivers
23:43 MarcinWieczorek: I'll just install them and reboot.
23:44 imirkin_: wait, are you trying to get help using nvidia drivers in here?
23:44 imirkin_: if so, i can't help you.
23:45 MarcinWieczorek: I don't think that's related. As you probably remember you guys told me to try the mainline kernel in order to reclock my GPU
23:46 imirkin_: correct.
23:46 MarcinWieczorek: Does nouveau blacklist NVIDIA driver or the other way?
23:46 MarcinWieczorek: I don't remember...
23:46 imirkin_: neither does any such thing. blacklisting is done in modprobe config files, which are not part of ... well, drivers.
23:47 MarcinWieczorek: It was arch specific then.
23:47 MarcinWieczorek: ok brb
23:47 imirkin_: and you can always modprobe directly...
23:47 MarcinWieczorek: It's easier to just remove the package
23:48 MarcinWieczorek: Interesting
23:49 MarcinWieczorek: It works. Any idea why it didn't with NVIDIA drivers?
23:49 imirkin_: fwiw i saw no reference of nvidia drivers loading in your first dmesg either.
23:50 imirkin_: but i don't do distro support, and i don't do nvidia proprietary driver support.
23:50 MarcinWieczorek: Yeah it seems weird that xorg broke the NVIDIA driver after update to 4.10-rc6
23:51 MarcinWieczorek: I should investigate that and report
23:51 imirkin_: i'd start assuming PEBKAC if i were you
23:52 MarcinWieczorek: Nevermind, it works now.
23:53 MarcinWieczorek: Could you please remind me the path for reclocking?
23:54 MarcinWieczorek: Oh I have some more 'files' there now
23:54 MarcinWieczorek: pstate, right?
23:55 imirkin_: right.
23:55 MarcinWieczorek: I have 3 options, 07, 0f, AC. Where can I find some information about that?
23:56 imirkin_: AC isn't an option - it's the current state
23:56 imirkin_: 07 and 0f are the two levels. you can echo them into the pstate file to have the gpu switch between the two perf states.
23:56 MarcinWieczorek: That was my guess
23:57 MarcinWieczorek: 0f: 270-1241MHz memory 5400MHz
23:57 MarcinWieczorek: So I turn this on to play some games and then I turn it back.
23:57 MarcinWieczorek: There's a chance that it hangs, what do I do then?
23:57 imirkin_: press the reset button.
23:58 MarcinWieczorek: Is it harmful?
23:58 imirkin_: pressing the reset button? HD's tend to not like that too much
23:58 MarcinWieczorek: for the GPU?
23:58 MarcinWieczorek: I mean it hanging.
23:58 imirkin_: nope
23:58 MarcinWieczorek: Cool.
23:58 orbea: would it hang so much that you cant ssh in and reboot?
23:58 imirkin_: at least not in our experience with it
23:58 imirkin_: orbea: depends. can lock the PCI bus...
23:59 orbea: never seen it hang personally
23:59 orbea: ah
23:59 MarcinWieczorek: If I don't lower the frequency when the load is low it's only about the power it consumes right?