01:05pmoreau: skeggsb_: Thanks for having a look! :-)
02:40pmoreau: mupuf: I would love PFB documentation!!
02:41pmoreau: I will probably add a few unknown regs to Envytools, once I checked their existence on my other boards.
03:07RSpliet: pmoreau: PFB documentation would make life boring :p
03:16skeggsb_: pmoreau: i dunno if it works, i couldn't manage to reproduce the issue, the fix is a likely guess
05:01pmoreau: RSpliet: I second mupuf :D
05:58pmoreau: Is there a way to power down one card under the blob? Would vgaswitcheroo work?
06:00mlankhorst: bbswitch perhaps?
06:04pmoreau: It bypasses the driver and disables directly at ACPI or PCI level it seems
07:04pmoreau: mlankhorst: Maybe you'll have some ideas: when I powerdown the discrete card (IGD: MCP79, DIS: G96), it hangs right during a nouveau_fence_wait ( nouveau_channel_idle(drm->channel) )
07:05mlankhorst: pmoreau: is it already powered down at that point? :P
07:07pmoreau: Hum... vgaswitcheroo reports that they are both on, and both were initialised and the discrete wasn't powered down automatically as it isn't detected as an Optimus laptop
07:08pmoreau: And I forgot to precise: if I do a PCI reset before loading Nouveau, it doesn't hang and suspend completes.
07:08mlankhorst: do you have working interrupts? :p
07:08pmoreau: What do you mean by "working"? :)
07:09mlankhorst: hm no it busy waits..
07:09mlankhorst: what if you simply remove the channel idle call, what breaks?
07:10pmoreau: And it's not infinite looping, I added some printk and it never exists the loop, just does two runs and then nothing...
07:10mlankhorst: in that case it shouldn't hang..
07:10pmoreau: The next call, which is nouveau_fence(drm)->suspend(drm)
07:11mlankhorst: hm.. dmesg from the idle stuff?
07:11pmoreau: With channel idling commented out or not?
07:12mlankhorst: without, lets see what breaks normally
07:13pmoreau: Here you are: http://pastebin.com/F2BcyBzU
07:19mlankhorst: and it's not doing an infinite loop in nouveau_fence_wait_busy ?
07:20pmoreau: Let me check
07:21mlankhorst: even so.. it should timeout after 15 seconds or so
07:31pmoreau: I still get two runs, then nothing more. It never printed the message placed just before the return statement.
07:33mlankhorst: that's weird... try with CONFIG_PROVE_LOCKING ?
07:33mlankhorst: and maybe PROVE_RCU
07:33mlankhorst: if feeling particularly paranodi
07:35pmoreau: Which engines / regs influence nouveau_fence behaviour?
07:35mlankhorst: in case of busy wait that won't be a worry..
07:36mlankhorst: it reads some vram, that should be all
07:36pmoreau: Ok. Then that it works when pci reset'ing the card is even weirder
08:00pmoreau: mlankhorst: Do I need to add some flags or does activating those options will automatically add messages?
08:35pmoreau: mlankhorst: Eh eh, don't know if I'll be able to post the results, and updated my kernel repo and fell into a new bug.
08:35pmoreau: I'll post it after tea. :-)
09:02pmoreau: So, here is the issue I'm now hitting: http://pastebin.com/x5pmNeh4
09:31mlankhorst: pmoreau: seems to me you have bigger issues then :P
09:37Tom^: so uhm how do i build and use dartamas repo? it only seems to contain some bits of nouveau?
10:06xexaxo: Tom^: doing an out of three build ought to work. not sure if it holds true since the latest changes/movement
10:07Tom^: hm ok
11:26RSpliet: who;'s coming to FOSDEM?
11:31hakzsam: pmoreau, ^
13:47pmoreau: RSpliet: If I take some time to book somewhere to sleep and transportation, I should come to FOSDEM. :D
13:47pmoreau: mlankhorst: Meh... It wasn't there before, it just appeared suddendly after the update. :/
13:48pmoreau: But, I do have bigger issues with it. I was trying to see if solving this issue might help solving/understanding another one.
13:48RSpliet: pmoreau: I usually just take a random intercity train to Brussles
13:48RSpliet: from my city there's one every one or two hours
13:50RSpliet: I'm staying with some ultimaker folks in a hotel they booked, but I'm sure there's some hostel beds free somewhere
13:50RSpliet: I stayed in 2go4 before, that was quite decent
13:51pmoreau: It's a bit more from mine :-) About 5h
13:51RSpliet: pmoreau: I was talking about the frequency, not the distance ;-)
13:51pmoreau: Oh, right!
13:52pmoreau: Same for me then, but the price can double so...
13:53pmoreau: Do you know who else is coming?
13:54pmoreau: I know Martin, Ilia and Samuel don't
13:54RSpliet: oh I see Alexandre is giving a talk!
13:54pmoreau: Yes, and Thierry too iirc
13:55RSpliet: interesting that the only nouveau-ish talks this year are done by NVIDIA folks
13:55RSpliet: who would've thought a year ago
13:55pmoreau: Combined with the pull request from this morning (or from yesterday)... :-)
13:56RSpliet: I'm... disappointed I'm hardly part of it right now
13:57RSpliet: but heck
13:57RSpliet: it's going to be a good weekend
13:57pmoreau: Yeah! I wonder if Andy and Lauri are coming too.
13:58RSpliet: curro: are you planning to visit Fosdem?
13:58tobijk: heh, whish you much fun there :)
13:58pmoreau: tobijk: Thanks!
13:58pmoreau: tobijk: Btw, the blob does autoreclocking :-)
13:58tobijk: i miagined, but i was not sure :D
13:59tobijk: way too much typing today, sorry
13:59pmoreau: No problem
14:11curro: RSpliet: nope, i don't think i can make it :(
14:12RSpliet: awh, that's unfortunate!
22:23mupuf: pmoreau: I saw Lauri this week end for my house warming
22:23mupuf: I don't think he is going to FOSDEM