On Mon, 10 Mar 2014, Jiri Kosina wrote: > > >> > *ERROR* render ring initialization failed ctl 0001f001 head 00003004 tail 00000000 start 00003000 > > >> > > > >> > as it doesn't seem to be there in case of resumption that works properly. > > >> > > > >> > Please see the dmesg from the broken case below. > > >> > > >> I encountered this again with -rc5. > > >> > > >> If there is anything I can do to help debug this, please let me know. > > > > > > I hate to be doing this, but ... ping? :) > > > > gm45 and rendering ring init failures. We've seen this occasionally > > crop up due to rather unrelated changes. Some even hit stable > > backports and had to be backed out again. We essentially have no clue > > what's amiss, but it seems to /mostly/ work. Thus far I've only hear > > reports of this for gm45 and not yet really for upstream. Until > > someone digs up more evidence I think we need to classify this as a > > rare heisenbug and not really a regression :( > > Thanks for getting back to me with this, Daniel. > > Well, this started to be rather regular sometime during the 3.14 > development cycle on my system. > > I am able to reproduce it after ~20 suspend-resume cycles. Which is rather > inconvenient for bisect, but enough for making the kernel unusable on that > system. > > > One thing we could try is to simply repeat the ring init setup, maybe > > after a gpu reset or something like that. > > Will be happy to test any patches. So I am still seeing this with current Linus' tree rather regularly (but it's not deterministic enough for a reliable bisect). Unfortunately I haven't received any patches to test; what do you propose? Would reporting this on bugs.freedesktop.org instead of ranting here change anything? Thanks, -- Jiri Kosina SUSE Labs _______________________________________________ Intel-gfx mailing list Intel-gfx@xxxxxxxxxxxxxxxxxxxxx http://lists.freedesktop.org/mailman/listinfo/intel-gfx
