On ke, 2015-11-04 at 20:57 +0000, Chris Wilson wrote: > On Wed, Nov 04, 2015 at 09:25:32PM +0200, Imre Deak wrote: > > After Damien's D3 fix I started to get runtime suspend residency for the > > first time and that revealed a breakage on the set_caching IOCTL path > > that accesses the HW but doesn't take an RPM ref. Fix this up. > > Why here (and in so many random places) and not around the PTE write > itself?
Imo we should take the RPM ref outside of any of our locks. Otherwise we need hacks like we have currently in the runtime suspend handler to work around lock inversions. It works now, but we couldn't do the same trick if we needed to take struct_mutex for example in the resume handler too for some reason. > -Chris > _______________________________________________ Intel-gfx mailing list Intel-gfx@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/intel-gfx
