01:13rhyskidd: anyone done an MMIO trace successfully on an Optimus laptop?
01:29imirkin: karol's done it plenty of times
01:33rhyskidd: never quite seem to be on IRC at the same time as karolherbst unfortunately, given timezones
01:36imirkin: he's in GMT+1 or 2
01:37imirkin: (whatever germany's on, i always forget)
02:32imirkin: that sure is a good use of cpu power... http://timestamp.online/countdown/1500000000
02:41AndrewR: it doesn't eat much cpu (compared to facebook)
02:52rhyskidd: okay, so I appear to have a valid MMIO log from my GP107 (at least according to demmio)
02:52rhyskidd: how could I be useful with this, re: https://bugs.freedesktop.org/show_bug.cgi?id=100228
03:09mangix: what's the significance of an MMIO read fault?
03:13rhyskidd: I (think) it means that the high security mode wasn't bought up right, it's like reading from an area of memory that is read protected
03:13rhyskidd: (caveat emptor with that statement though, someone more knowledgeable here may correct me)
03:18mangix: i get three of those
03:18mangix: i don't notice any effects
03:19rhyskidd: hrmm -- and I recall you are on a GP107 as well (GTX 1050 Mobile)?
03:21rhyskidd: that MMIO read fault appears to be causing problems for at least some of the GP107's. See a possible duplicate: https://bugs.freedesktop.org/show_bug.cgi?id=101764
03:22rhyskidd: the line is effectively a: nvkm_rd32(device, 0x409800)
04:15coredump: soooo... Wasn't 1050 support landing on 4.12? I am getting an unknown chipset shenanigan here.
08:24RSpliet: imirkin: Germany is on CEST
08:26RSpliet: one hour ahead of Britain
11:49imirkin: coredump: lspci -nn -d 10de:
11:56imirkin: coredump: and what chipset does nouveau think you have?
12:38imirkin: pmoreau: https://github.com/google/clspv
14:15karolherbst: imirkin: you said I shall remind you, so I remind you to look over my gallium precise patches. As I said yesterday, they are already reviewed, so a final quick look shall be okay imho.
14:16imirkin_: yeah, i was dead tired last night. will try again tonight if i'm not dead tired. although it's not looking good.
14:16imirkin_: weekend's more likely.
14:22rhyskidd: hi karolherbst
14:22rhyskidd: imirkin mentioned you were the one with experience on mmiotracing optimus cards
14:23karolherbst: imirkin: yeah no problem, as long as it's before the feature freeze, it should be fine
14:23rhyskidd: I've recently got a set of MMIO's from my GP107M, which has been a card nouveau is having trouble bringing up properly
14:23karolherbst: tracing while nouveau is running?
14:23rhyskidd: trace of the blob
14:23karolherbst: okay, mhhh
14:24rhyskidd: when it brings up the card, basic glxgears and also when it probes the them, usage etc settings via nvidia-smi
14:24karolherbst: it should work (tm), but the mmiotraces isn't considered bug free
14:24rhyskidd: happy to do some of the debugging myself, with pointers here and there
14:25rhyskidd: is it helpful to diff the bringing of the blob on my card compared to a GP106?
14:26rhyskidd: as I know there's a specific nvkm_rd32() in nouveau that is causing a timeout fault, although I can see the blob reading from that location fine
14:34rhyskidd: I've got a working copy of demmio from envytools etc
14:38tobijk_: rhyskidd: yes a diff compared to a GP106 setup is helpful
15:25rhyskidd: tobijk_: okay, I'll work on preparing one. Had a quick skim through the two (GP107 vs GP106) last night
15:25rhyskidd: Will also have blocks of time this weekend to dig into it
15:28tobijk_: rhyskidd: thanks for the work, but no worries: take your time! :)
15:37rhyskidd: I'm motivated, I want to get my GPU booting and reasonably stable with nouveau :)
15:43imirkin_: rhyskidd: my guess is that the gpu is "more off" than usual due to some kind of PM thing, and we don't bring it back quite fully enough
15:43imirkin_: rhyskidd: we had a similar issue with GK106's a while back
15:47rhyskidd: imirkin_: any bugzillas or commits on the GK106 issue that you could point me to to start background reading?
15:49imirkin_: search for PGOB in the code
15:49rhyskidd: ok, thanks
15:50imirkin_: and see commit a0a60433f8e0f159077d59f5e4cbfc77f0f65606 in ben's repo
15:50imirkin_: some details at http://download.nvidia.com/open-gpu-doc/gk104-disable-graphics-power-gating/1/gk104-disable-graphics-power-gating.txt
16:37grift: since Linux 4.12/4.13rc0 it seems that nouveau freezes my system: the only message printed is something along the lines of "[DRM] nouveau ... DDC responded, but no EDID for DP1", Fedora, works fine when i boot in Linux 4.11
16:37grift: is this a known issue?
16:37karolherbst: grift: what gpu?
16:37imirkin_: grift: with DP, right?
16:38imirkin_: er, duh. "on DP1".
16:38imirkin_: i should learn to read
16:38imirkin_: yes, known issue
16:38imirkin_: but not yet fixed
16:38grift: ok thanks
16:38imirkin_: you can revert a patch
16:38imirkin_: [hold on]
16:38grift: naw ill stick with 4.11 for now thanks though
16:39imirkin_: hrmph, can't find it
16:39grift: its ok, patience is a virtue, just glad that its a known issue
16:39imirkin_: ah there ya go -- `nouveau "eDP-1: EDID is invalid" regression after 4.11 with HP ZBook 15 G3` on email@example.com
16:39imirkin_: surprised that 4.11 works for you - the patch went into 4.10
16:40grift: maybe redhat caries some patches that affect this?
16:40grift: or maybe the issue is not that same
16:41imirkin_: could be.
18:23karolherbst: why the heck does steam dos my dns server....
18:28karolherbst: tobijk_: for fun: remove your nameserver entry from /etc/resolv.conf -> steam stops downloading
18:28karolherbst: because it's stupid
18:29karolherbst: or whenever steam downloads are slow, write googles DNS into it and it gets faster again
18:30karolherbst: it's stupid
18:31karolherbst: tobijk_: https://github.com/ValveSoftware/steam-for-linux/issues/3401
18:32tobijk_: karolherbst: mh never had such a problem...guess i'm lucky :)
18:32karolherbst: I am running a dns cache locally
18:32karolherbst: for fun reaons it starts to throttle at some point
18:32karolherbst: cause why would you even want to do that many DNS queries...
18:33tobijk_: if its dos'ing thats even a good thing
18:33karolherbst: well I woulnd't call it dosing
18:33karolherbst: but it does several a second
18:33karolherbst: and I am sure for stupid reasons
18:34tobijk_: isnt the web browser chromium-based
18:34karolherbst: it has nothing to do with the web browser
18:35karolherbst: it's doing it while downloading stuff
18:35karolherbst: which makes even less sense
21:43mwk: xgfvrt cvxxzc cvbgfvtr5cv gfbtgrfcvcgftr cvxxccvbgftgrvvgfbftx cvbgfrtvrftdvgftr cvxvcvcftgrv cbfgdcex vdfcv vbgfgfbgffgvgtrhftgrfv bnmjk';
22:38aaronp: Huh. xf86-video-nouveau is trying to call nouveau_getparam(pNv->dev, NOUVEAU_GETPARAM_HAS_PAGEFLIP, &v) on an nvidia-drm device.
22:39karolherbst: aaronp: maybe messed up X config?
22:39aaronp: It's an Ubuntu X server, I think it's just loading all of the drivers.
22:40aaronp: Oh, I see. Our X driver failed because I built it with the wrong ABI, and it tried to fall back to nouveau.
22:42aaronp: But still, seems like a legitimate bug.
23:32imirkin_: aaronp: send patches :)