00:52 karolherbst: RSpliet: okay, so the nvac should work perfectly in general?
00:53 karolherbst: imirkin: any idea how I can boot from a gentoo boot disc without a keyboard? :D
01:30 pq: karolherbst, don't wait for me to review the mmiotrace patch.. can't say when or if I might get to it
01:31 karolherbst: pq: it got already merged
01:32 karolherbst: meh, installing an EFI linux on this mac without a keyboard will be really painfull :/
01:34 pq: karolherbst, oh? That's cool. I was on CC when you sent it, right? I didn't get a CC on any ack or review.
01:34 pq: karolherbst, or was it a note from a bot?
01:35 karolherbst: bot note
01:35 pq: are you sure that not just kernel CI?
01:35 karolherbst: yes
01:36 pq: i.e. a bot that gathers all patches from LKML, and queues them build & boot testing
01:36 karolherbst: pq: https://git.kernel.org/cgit/linux/kernel/git/tip/tip.git/commit/?id=cfa52c0cfa4d727aa3e457bf29aeff296c528a08
01:36 pq: ooh, Ingo, nice
01:37 pq: karolherbst, and congrats :-)
01:39 karolherbst: thanks
01:39 karolherbst: maybe it would make sense to add it to stable kernels as well
01:39 karolherbst: but meh
01:40 pq: it certainly would, if there are nvidia blobs not working on the tip kernel
01:40 pq: and that would need it
01:41 karolherbst: well I had the same issue more than a year ago
01:42 karolherbst: but then it just disappeared
02:03 pmoreau: karolherbst: Why not use a keyboard? :-D
02:03 karolherbst: because I have no external one
02:03 pmoreau: But otherwise, the MCP79 should work nicely, reclocking included.
02:04 karolherbst: yeah, the efi stuff is just a bit messy on a mac
02:04 karolherbst: or does efibootmgr work?
02:04 pmoreau: Gummiboot/systemd-boot always worked perflectly on mine
02:04 pmoreau: efibootmgr works
02:04 karolherbst: okay
02:05 karolherbst: well still without keyboard it is a bit messy :/
02:05 pmoreau: :-D
02:05 karolherbst: because the only boot cd I found, needs a keyboard otherwise it boots from disc...
02:05 karolherbst: and somehow burning the ubuntu iso failed (is use usually ubuntu to bootstrap my gentoos :D)
02:07 pmoreau: Hum… by default the mac will boot from disc, unless you press alt
02:07 pmoreau: Not sure if you can avoid that
02:08 pmoreau: Maybe you should ask for keyboard donation, rather than GPUs :-p
02:08 karolherbst: well I have a working mac os x installation
02:08 karolherbst: so there I can do pretty much everything
02:08 karolherbst: I even setup ssh
02:08 karolherbst: and there is this startup disc entry inside systemsettings
02:09 pmoreau: Oh, right
02:09 karolherbst: I already setup refind and added the linux boot entry
03:53 Sascha_: Hi all. Can someone help me with issues on a NVE7? I only get blank screens whenever I plug in two displays. Tested on kernel 4.3.5 and 4.4.3
03:56 pmoreau: Could you link to your dmesg and Xorg.0.log when this happens please?
03:57 Sascha_: sure. give me a sec
03:59 sascha_: wohooo hotplugin produces even a soft lockup in nv50_dac_detect
03:59 pmoreau: Yeah! … But, that should make it easier to debug, hopefully :-)
04:00 sascha_: http://hastebin.com/kuvejesuga.xml
04:01 sascha_: can't get the Xorg log as ssh hangs... :)
04:02 sascha_: But after hotplug, it shows no new information.
04:06 pmoreau: Try to get the whole dmesg and Xorg.0.log (by cat'ing them for example)
04:07 pmoreau: And open a bug report. I might try to reproduce on my GK107, but with the bug report, other people could help investigating.
04:08 pmoreau: Apparently one of the monitors provides a bad EDID, and Nouveau didn't like that.
04:08 pmoreau: s/like/liked
04:11 martm: liked or did not like, not did not liked..but the point understood non the less
04:11 martm: first time you had it correct in terms of english:)
04:15 martm: cause did is allready reference to past tense:)
04:18 pmoreau: True, and if I wanted to use the past perfect, "had" would have been needed, instead of "did".
04:21 martm: yes, i think in that case it would be, "he had not liked that" indeed.
04:22 martm: i would use would not had liked it, which sounds kinda better though
04:25 martm: in other words actually your first sentence was gramatically absolutely correct
04:28 martm: hehe, enough of this and logically correct too, fixup was wrong, it's exactly how you should had said it, how you typed it the first time
04:29 martm: some say that french is that much a big nation, and their language is that complex, that they all do not speak english, but seems to improving a lot
04:32 sascha_: The funny thing is, the monitors are identical...one is plugged in with DVI the other via DP
04:32 sascha_: I tried to hotplug the DP one
04:35 sascha_: haha :) bringing all the french to speak english is probably harder than providing hardware accelaration on new NVIDIA hardware :)
04:38 martm: well segfault usually means memory fault right, pointer pointing to wrong address or something, or memory corruption in a worst case etc. pretty harch stuff mostly
04:39 pmoreau: sascha_: What happened if you hotplug the other one?
04:40 karolherbst: meh stupid bootloader
04:41 karolherbst: pmoreau: now I got a usb stick working to be recognized by refind, but guess what, I disabled the internal discs and now refind falls back to recovery allthough it finds the boot stuff on the usb stick .
04:41 pmoreau: karolherbst: :-D
04:42 pmoreau: I never used refind, so I sadly can't really help you…
04:45 karolherbst: ohh seriously...
04:45 karolherbst: I hate all those boot images
04:46 karolherbst: is it too much to expect an efi binary on a boot disc?
04:47 sascha_: the DP Monitor is half a year newer and says its EDID Version 1.4, the DVI one is 1.3
04:47 martm: right
04:47 sascha_: but DP alone works also fine...lets plug the dvi one in
04:47 sascha_: and see the kernel go crazy...
04:48 sascha_: same [ 221.188078] [drm:drm_edid_block_valid [drm]] *ERROR* EDID checksum is invalid, remainder is 108
04:48 sascha_: and a completely scrambled DP display output
04:49 sascha_: should i boot with nouveau.debug something when opening the bug?
04:49 sascha_: i've seen something like edid trace
04:50 pmoreau: Well, it won't hurt.
04:50 pmoreau: So no matter which one you hotplug, it always goes berserk?
04:51 pmoreau: karolherbst: You should get yourself an external keyboard, it would be soooo much easier. :-)
04:51 karolherbst: ahh well, it will be fine
04:51 karolherbst: I will just tell the efi to boot from the usb stick directly
04:58 sascha_: yes no matter which one i hotplug the other one stalls and then gots completely scrambled after 10 sec...
04:58 sascha_: after hotplug and stalling nothing happens, then I see this in dmesg: nouveau 0000:05:00.0: disp: chid 2 mthd 0080 data 00000000 00005080 0000000d
04:58 sascha_: or disp: outp 02:0006:0f42: link training failed
04:59 pmoreau: I guess it's simply unhappy by what happened previously
04:59 sascha_: in the DVI case and then i looks like someone confuses the framebuffer layout completely and writes chunks of the image to random places
05:01 sascha_: and after unplug of the DP I get the nv04_timer_read soft lockups (i guess it still tries to reread something from that display)
05:02 sascha_: what would be the nouveau.debug line for this kind of stuff? PDISP=debug, anything else?
05:04 pmoreau: nouveau.debug=debug should be enough, and you might want to add drm.debug=0x3
05:54 karolherbst: pmoreau: yay, I managed to load grub from my usb stick :D
05:54 pmoreau: \o/
05:54 karolherbst: mhh, now guess what
05:54 karolherbst: :D
05:54 pmoreau: You need a keyboard for the installation?
05:54 karolherbst: yes :D
05:54 pmoreau: xD
05:54 karolherbst: well not for isntallation, I just want a bash actually
05:54 karolherbst: to setup ssh
05:55 pmoreau: I don't remember if the Arch ISO boots with SSH enabled or not
05:55 karlmag: karolherbst: why don't you just connect a keyboard to it then? Even borrow one from a different computer while you do it?
05:56 karolherbst: karlmag: guess what
05:56 karlmag: you don't have a usb keyboard at all?
05:57 karolherbst: right
05:57 karolherbst: well not here
05:57 karlmag: I can see your problem then
05:58 karlmag: remote usb keyboarding can be quite tricky ;-)
05:59 karolherbst: and now I have a keyboard :D
06:00 karlmag: I guess also serial/network console might not be as easy either.
06:00 karolherbst: of course not
06:00 karolherbst: it is a mac mini
06:01 karlmag: *nods* Yeah, I have no idea how to go about to be able to do it on that.
06:02 karolherbst: wtf...
06:02 karolherbst: that live cd is soo anyoing
06:04 karolherbst: there is no eject command
06:04 karlmag: I do have some hp microservers with builtin remote connnection (separate network) that is actually a bit handy.
06:04 karlmag: heh.. that sounds like not too helpful.
06:04 karlmag: especially on those machines that doesn't have a physical eject button.
06:05 karolherbst: right
06:05 karolherbst: the mac mini hasn't either a force eject thing inside the driver
06:06 karlmag: no? I was pretty sure they did have some kind of eject, but..
06:06 karlmag: Then again.. nearly nothing surprises me anymore :-P
06:07 karolherbst: press mouse on boot
06:07 RSpliet: let me guess... no mouse?
06:08 karolherbst: that I have
06:08 karolherbst: I am not that crazy to use the touchpad on my laptop for daily stuff
06:08 karolherbst: :D
06:09 karlmag: *shrugs* many does...
06:09 karolherbst: maybe I even got network now! :O
06:09 RSpliet: we should have someone hack up an on-screen keyboard for VTs
06:09 karolherbst: and how do you use them?
06:09 karlmag: I don't though.. But I don't use laptops as primary machines either. Only occationally really.
06:10 RSpliet: karolherbst: mouse?
06:10 karolherbst: is it that simply on a VT?
06:10 karolherbst: *simple
06:10 karlmag: HUD? With touch interface?
06:10 karlmag: :-D
06:10 RSpliet: idk, there's GPM for mouse input
06:14 karolherbst: okay, now how do I get ssh started
06:19 karolherbst: meh, no links
06:32 prg: karolherbst: which branch of yours should i be using if i want the voltage fixes but no dynamic reclocking on my gk106 with a 4.4 kernel?
06:35 karolherbst: prg: stable_reclocking_kepler_v2
06:35 prg: thanks
07:02 john_cephalopoda: Hey, short question: When I open, close or resize a window, there are sometimes fragments of the video memory displayed. Like snippets of firefox windows I closed a few minutes before. Is that related to nouveau, mesa, Xorg or something else?
07:37 sascha_: it looks like I get a random EDID very time I hotplug my second monitor...could this be a clock issue?
07:37 prg: karolherbst: this branch was supposed to fix the voltage issues? i'm getting really interesting lockups again, like http://paste.debian.net/hidden/127ad8dd/ and http://paste.debian.net/hidden/8f4359f9/
07:40 sascha_: I get this one on 4.4.3 "nouveau 0000:05:00.0: fifo: write fault at 000010b000 engine 04 [BAR1] client 08 [HOST_CPU_NB] reason 00 [PDE] on channel -1 [003fe1f000 unknown]" anybody knows what that means?
07:41 sascha_: whats BAR1 and what is a PDE? :)
07:45 karolherbst: mhhh
07:45 karolherbst: maybe I messed with the scheduler
07:46 karolherbst: ohh wait
07:46 karolherbst: can't be
07:46 karolherbst: prg: but it is always this error, right?
07:46 karolherbst: and the gpu still runs somewhat?
07:47 prg: i didn't try that often
07:47 karolherbst: mhh
07:47 prg: mouse pointer was still moving, rest of the desktop was frozen
07:47 karolherbst: okay, so just the channel was bricked
07:48 karolherbst: prg: does it only happens with mplayer?
07:48 karolherbst: ehhh
07:48 karolherbst: mpv
07:48 prg: the first paste if from some game
07:48 prg: *is
07:49 karolherbst: ohhh
07:49 karolherbst: the same paste opened twice here
07:50 karolherbst: no idea really
07:51 karolherbst: prg: is it more stable than stock nouveau by the way?
07:52 prg: hard to say, haven't done much testing recently
07:53 prg: just boot with nouveau, run some game, still locks up, back to blob, wait another few months
07:53 karolherbst: I see
07:54 karolherbst: prg: could you boot with nvidia and run a tool inside my branch?
07:54 prg: the nv_cmp_volt thing?
07:54 karolherbst: yeah
07:55 prg: we already did that, looked fine
07:55 karolherbst: mhh we couild do something else then too
07:57 karolherbst: prg: can you checkout this? https://github.com/mupuf/pdaemon_trace
07:57 prg: sure
07:58 karolherbst: mhh I hope this works though, compiling the tool is a bit messy
07:58 karolherbst: inside the nouveau folder run make
07:59 prg: no Makefile. maybe autogen.sh first?
07:59 prg: compiling...
08:00 prg: done
08:00 karolherbst: k
08:00 karolherbst: do you have envytools compiled somewhere?
08:01 prg: yes
08:01 prg: (already needed envyas right now)
08:04 karolherbst: okay
08:04 karolherbst: then inside pwr_read
08:05 karolherbst: compile pwr_read.c
08:05 karolherbst: gcc -o pwr_read pwr_read.c -lpciaccess -lpthread -lnvhw -lnva -Lpath_to_envytools/build/nvhw -Lpath_to_envytools/build/nva
08:10 karolherbst: k
08:10 karolherbst: RSpliet: nouveau doesn't like it when I remove all the displays
08:13 prg: hah, -lnvhw needs to come after -lnva or i get "undefined reference to `parse_pmc_id'"
08:13 prg: got it compiled now
08:13 karolherbst: nice
08:13 karolherbst: run it :)
08:14 prg: with blob?
08:14 karolherbst: ohh wait
08:14 karolherbst: one thing
08:14 karolherbst: you need to adjust this path: https://github.com/mupuf/pdaemon_trace/blob/master/pwr_read/pwr_read.c#L458
08:14 karolherbst: it has to point to the
08:14 karolherbst: well I have set it to ../nouveau/bin/nv_init
08:14 karolherbst: but then you can only run it inside this directory
08:16 prg: ok, path adjusted. now with blob or nouveau?
08:17 karolherbst: prg: with both
08:17 karolherbst: once with nouveau at max clocks
08:17 karolherbst: then with nvidia at full load
08:18 prg: so first echo 0f > pstate then just run this program?
08:19 karolherbst: yeah
08:19 prg: which one of the three pstate files do i use? /sys/kernel/debug/dri/{0,64,128}/pstate
08:19 karolherbst: doesn't matter
08:21 prg: ok, it's outputting stuff
08:21 prg: ^C and pastebin?
08:22 prg: http://paste.debian.net/hidden/622608dd/
08:23 karolherbst: this is with nouveau?
08:23 prg: yes
08:23 karolherbst: okay, and then once with blob at full load
08:23 karolherbst: it has to clock to max
08:24 prg: brb
08:29 prg: karolherbst: http://paste.debian.net/hidden/e6318eb7/
08:29 karolherbst: mhh, seems okay though
08:29 karolherbst: ohhhh wait
08:29 karolherbst: I have a stupid idea
08:30 karolherbst: mupuf: I saw that for core clocks sometimes there are also 2 PLLs like for memory, could it be, that we may have to do something similiar as I do with the memory plls?
08:38 mupuf: karolherbst: possibly, but I doubt that the core would be as senstive
08:38 karolherbst: well
08:38 karolherbst: it seems there are strange crashes still
08:39 karolherbst: I saw simliar patterns on the maxwell gpu
08:39 mupuf: karolherbst: I take pull requests on the nouveau_daemon thingy
08:39 karolherbst: mupuf: I didn't change anything there :D
08:39 karolherbst: just the path stuff
08:42 mupuf: yes, but that is likely due to us not halting hte gpu in the proper way
08:42 karolherbst: mhhh
08:42 karolherbst: but why does it happen when we put some load on it ?
08:42 karolherbst: and not immediatly
08:44 mupuf: well, it does not happen when there is no load because ... the gpu is not doing anything
08:44 mupuf: so there is no data to corrupt
08:44 mupuf: oh! I may have a hack for us to try!
08:45 karolherbst: nice!
08:45 karolherbst: I love hacks
08:45 mupuf: that is extreme, but we could context switch to another channel, do the reclocking and then reload the channel
08:45 mupuf: that should be the ultimate way for pausing hte gpu
08:46 mupuf: but maybe just checking the mmiotraces for anything that looks like gpu-pausing code would be good
08:46 mupuf: I checked it on tesla, never on later gens
08:50 karolherbst: well at least on the tegras, we have some slow up/downclocking thing
08:50 karolherbst: well "slow"
08:50 karolherbst: I meant like stepping up/down
08:52 mupuf: well, there is either the gpu problem OR we are completely ignoring dependencies on clock domains
08:53 mupuf: could you check in which order the clock domains are reclocked?
08:53 mupuf: there may be some treasures in there
08:59 karolherbst: yeah I guess I will spend some time with that later this week or at the weekend
09:00 karolherbst: but it would help if some of your gpus would actually crash with my patches, but they don't :D
09:35 mupuf: karolherbst: I know... it is so annoying when other machines do not have a bug
09:35 mupuf: sorry, when other machines have bugs you do not have
09:51 karolherbst: mupuf: yeah, I just wanted to get my mac mini working today and I may spend the next days to get some stuff working on it so that I might can do some tesla stuff (it has a nvac)
09:51 karolherbst: those load things as one example
09:52 karolherbst: funny thing though, because I have that one for around 3 years now...
09:53 mupuf: dude, nice!
09:53 mupuf: I have been trying to find one of those for cheap and I failed
09:53 karolherbst: :)
09:53 karolherbst: it was sold as bricked
09:53 karolherbst: because the seller said dvd driver might be bricked
09:53 karolherbst: and fan is broken
09:54 karolherbst: but for some reasons both works :D
09:54 mupuf: who cares about dvds :D
09:54 karolherbst: well
09:54 karolherbst: but yeah
09:54 karolherbst: it is even one with 2.53GHz core 2 duo
09:56 karolherbst: Macmini3,1 Late 2009 I think
09:56 karolherbst: if you need it for something, I could also setup some ssh stuff and so on
09:56 karolherbst: I don't plan to actually use it for anything anymore yet
09:57 karolherbst: maybe later some os x developement again, but meh
12:41 hakzsam: mwk, imirkin any objections if I merge this PR https://github.com/envytools/envytools/pull/47/commits ?
14:15 pmoreau: imirkin: :-( I can't seem to convince the blob to generate `cvt u64 u32`, nor `cvt u32 u64`…
14:49 pmoreau: Since a truncation is taking place when going from u64 (or s64) to u32 (resp. s32) with no sat, I can simply replace the cvt by a mov.
15:26 jayhost: Looks like my keyboard was spamming char 'r' while I was away. It's turned off now when I'm not using.
16:19 discipline: hi guys, I'm trying to understand how does the nvidia drivers for NV110 change the laptop LCD backlight PWM. I understand it gets the current value, the max and the min values through the ACPI method \_SB.PCI0.WMI1.WMMX at start up
16:21 discipline: but I cannot figure out how it actually changes it. My checked build of Windows 10 unfortunately doesn't install, so I cannot use the AML debugger to see if it is done through an ACPI call or not.
16:21 discipline: if anyone could give me some directions on how to tackle this problem, it is greatly appreciated. Thanks