00:27Lyude: ergh, has anyone ever dealt with mmiotraces causing the system to completely hang while you try to load the nvidia module and get X running?
00:27Lyude: can't actually get any tracing done on this machine because it keeps hanging like that A:(
00:28karolherbst: Lyude: guess what I did today
00:28karolherbst: Lyude: https://github.com/karolherbst/linux/commit/1c071dd80b3aae339b5150eb9daabc51b73f64d6
00:28Lyude: is that the exact same bug as the one I'm hitting right now? lol
00:28karolherbst: might be
00:29karolherbst: if your last things with dmesg -w you see are some 0x200 mappings, then yeah
00:29karolherbst: ^please rephrase that yourself, I am too lazy for that now....
00:36karolherbst: Lyude: there is still an issue after unloading nvidia and disabling mmiotrace, some memory allocation in userspace failed (you know when you thought that malloc and new won't ever fail for silly reasons?), but it isn't as bad as a freezing machine
01:12karolherbst: Lyude: will you try my patch out? If it also helps for you, I will try to catch the last issue and send it to the ML
01:20Lyude: karolherbst: I will give it a shot tomorrow, I'm about to leave the office unfortunately and I'm going to have to update this machine. I've been trying to do as little messing with the 4.11 kernel on there as possible since it's the only setup i've been able to find that actually works with xwl eglstreams on nvidia's blob
01:44karolherbst: mupuf, pmoreau: I will just go ahead tomorrow and do the nouveau talk submission. We still have time to change stuff until the 15th of december
02:10imirkin: is there really anything to talk about?
03:40rhyskidd: karolherbst: you might find my mmiotraces for a GP107M in nvidia_bios repo helpful for what you're looking at now
03:40rhyskidd: same XPS 9560 etc
11:33karolherbst: imirkin: well, yeah, allthough not everything of that is technical or political
11:34karolherbst: rhyskidd_: that's not the issue here. I want to be able to create traces myself and fix the bugs in the mmiotracer :p
11:36karolherbst: I think I fully understood how the mmiotracer works now... well, hopefully there aren't many issues left. I know one upfront, but I think i will fix that one when it comes to it.
11:59karolherbst: in my trace: 1007943 lines with those UNKNOWN instructions... it is getting out of hand slowly
12:00karolherbst: that's 42%
12:01karolherbst: they start with "UNKNOWN 99.926290 389 0xd00040a8 f3,a4,c3 0x0 0" and end with "UNKNOWN 104.220165 389 0xd010af7f f3,a4,c3 0x0 0"
12:01karolherbst: so it is probably fine
12:01karolherbst: a few new instructions though? https://gist.github.com/karolherbst/b427275b945e09395a4214f820dbcb59
12:12karolherbst: mupuf: mind testing that mmiotrace patch as well? No idea if you had to trace GPUs/drivers where you hit that mess up as well. Just a good idea to have some more reviewers for that kind of stuff: https://github.com/karolherbst/linux/commit/fb0dd3ed935d19c50f61febdbe8ac70b5e686d51
12:12karolherbst: or tested bys
12:19mupuf: karolherbst: Will check that tonight :)
12:19karolherbst: mupuf: thanks!
12:21mupuf: oh, and are you still waiting for my R-b on the reclocking patches?
12:23pmoreau: karolherbst: I’ll try to give your MMIO patch a try this week, but I doubt I’ll review it, only test it.
12:23karolherbst: testing is fine by me as well
12:24karolherbst: I seriously doubt there are many people on earth who would be actually able to review it for real anyway :p
12:24karolherbst: even pq gave up at some point
12:24karolherbst: well not for this, but in general regarding mmiotrace
12:24karolherbst: I got his acked-by though....
12:25mupuf: karolherbst: what are you trying to fix there anyway?
12:26karolherbst: mmiotrace freezing/crashing the entire machine?
12:26mupuf: never happened to me
12:26mupuf: but... I have not mmiotraced in a while
12:26karolherbst: well, you never got those "mmiotrace: ioremap_*(0xcfdf0200, 0x200) = ffffa19241b06200", right?
12:26karolherbst: well it never happened on my machine either I think
12:26karolherbst: and it kind of depends on the driver/ GPU used?
12:27karolherbst: and also how lucky you get
12:28karolherbst: it was only triggered if you have something like ioremap((X & 0xfffffff000) + 0x200, ...)
12:28karolherbst: and then you get returned an offset to a internally mapped page
12:28karolherbst: so the returned address is not page alligned
12:28karolherbst: which basically messed up mmiotrace real bad
12:40karolherbst: Lyude: 0x40917c seems to be related to gating as well
12:40karolherbst: no idea if that's new in pascal or not
13:09clipcomp: New video created "https://www.youtube.com/watch?v=dnE838pivrE" has messages and graphics
13:22rhyskidd_: karolherbst, Lyude: 0x40917c (PGRAPH.CTXCTL.UNK17C) has been around since at least gm20b
13:22rhyskidd_: relates to SLCG for the gr
13:23rhyskidd_: probably should have renamed it in: https://github.com/envytools/envytools/pull/103
13:40karolherbst: rhyskidd_: might be, yes
13:40karolherbst: rhyskidd_: by the way: will you come to FOSDEM? would be quite a trip for you, but you'll manage :p
13:40rhyskidd_: i'm really considering it
13:41rhyskidd_: just trying to schedule around some other travel commitments
13:41rhyskidd_: as we discussed in the past, maybe even tacking on a short preso on the Pascal temp work would be more appropriate at FOSDEM, over say XDC
13:42karolherbst: mhh, I would be careful with technical stuff at Fosdem
13:42rhyskidd_: congrats on the RH role btw
13:42karolherbst: thanks :D
15:05imirkin: karolherbst: heh, i first read that as you having tested your own patch
15:06karolherbst: you know I never do that
15:07imirkin: i meant Tested-by :p
15:09karolherbst: yeah, I know
15:11pmoreau: imirkin: I read it wrongly as well :-D
15:15karolherbst: it hasn't even the same length...
15:16imirkin: that's not how people read
15:16imirkin: Kar... H....
15:17pmoreau: Exactly :-)
15:18karolherbst: well, I didn't do this mistake. I guess I am just different...
15:18imirkin: or you know your name and know the other dude
20:33mwk: Lyude: hey, you suggested using meson to build envytools some time ago
20:33mwk: do you have some experience in using that?
20:34imirkin_: meson's a non-starter atm
20:34imirkin_: it doesn't work with LANG=C
20:34mwk: ... what?
20:34imirkin_: it bitches and moans, and at least mesa build actually fails.
20:35imirkin_: that's what i said :)
20:39airlied: imirkin_: non starter for you, or do you know all the LANG=C users :-)
20:42imirkin_: non-starter for projects i'm involved wit
20:43pmoreau: I tried meson, and it worked rather nicely. Though, it didn’t build clover at the time, so I didn’t continue using it.
20:51imirkin_: airlied: it does appear as though all the LANG=C users tend to find each other though
20:52airlied: i thought there might be a club :-p