07:31PaulePanter: karolherbst, Lyude: Sorry, I just now read your replies.
07:32PaulePanter: karolherbst, Lyude: The thing is after a reboot everything is fine.
07:32PaulePanter: But I’ll relay that with the broken cable.
07:33PaulePanter: But should the system become unresponsive?
07:35PaulePanter: I was wrong, the error/warning *message* is still present after a reboot. I guess the message were a red herring, and the reason for the hang lies somewhere else.
07:36karolherbst: PaulePanter: yeah, probably
07:37PaulePanter: I am going to check later and ask my colleague about the details.
07:37karolherbst: okay, thanks
07:38karolherbst: usually it the machine hangs, it's worth trying to ssh into it
07:38karolherbst: and retrieve logs
07:38karolherbst: a froce shutdown always runs into the problem of logs getting lost
07:38PaulePanter: Yeah, my colleague knows about these debug methods, so I’ll ask him.
17:34Lyude: karolherbst: does the version of the runtime PM fix patch series for pascal that you have on the mailing list have the extra fixes you needed to get secboot to work after runtime resume?
17:35Lyude: karolherbst: think you could link me to them? Wonder if they might fix some of the issues that my friend has been having with that patch series
17:36karolherbst: Lyude: I have everything on that branch: https://github.com/karolherbst/nouveau/commits/secboot_fixes
17:36Lyude: karolherbst:cool, thanks!
17:36karolherbst: I removed the unrelated commit
17:38Lyude: karolherbst: didn't glisse mention something about doing a posting read that you said worked better then that fix?
17:38karolherbst: the point of the read was to flush out the register writes
17:39karolherbst: to ensure the engine is really going to get turned of
17:39karolherbst: I think we do that anyway... not sure
17:39karolherbst: maybe inserting a barrier would solve it as well?
17:39Lyude: probably not. karolherbst: looking at how i915 does mmio reads is probably a good way to figure that out
17:40Lyude: since they have a separate macro for POSTING_READ()
17:43karolherbst: Lyude: heh.. it's a normal read..
17:45karolherbst: Lyude: intel_uncore_funcs.mmio_readl
17:49Lyude: karolherbst: check what each gen specifies for that, iirc the uncore funcs are just to help report unmapped mmio accesses when doing read/writes
18:42airlied: karolherbst: posting reads are just normal reads
18:43airlied: but it's nice to indicate that the purpose of the read is for posting write not fore reading