On Mon, 2015-11-09 at 23:24 +0200, Imre Deak wrote: > On Mon, 2015-11-09 at 21:11 +0000, Chris Wilson wrote: > > On Mon, Nov 09, 2015 at 09:14:19PM +0200, Imre Deak wrote: > > > The device should be on when updating the GGTT PTEs, so add an assert to > > > all relevant places. > > > > > > v2: > > > - use the existing dev_priv directly everywhere (Ville) > > > > > > Signed-off-by: Imre Deak <imre.deak@xxxxxxxxx> > > > > For completeness, add one to i915_ggtt_insert_entries(). Yes, I know > > that we never will support rpm on any of those devices, but we may as > > well be consistent in documenting that these functions will write > > through the magic PCI region. > > Ok, forgot about that one. > > > An issue I see here is dev_priv->pm.suspended doesn't actually say that > > rpm can't suspend during the GGTT PTEs writes. Would we not want > > something more like assert_rpm_wakelock() and !dev_priv->pm.can_suspend? > > Well, after this patchset we assert that the RPM refcount!=0 too, which > should guarantee that provided that pm.suspended=false as well. It's > possible to have a non-zero refcount with a suspended state (in case of > rpm_get_no_resume). But you are right that assert_device_not_suspended doesn't reflect that completely, so I could rename it. > > > -Chris > > > _______________________________________________ Intel-gfx mailing list Intel-gfx@xxxxxxxxxxxxxxxxxxxxx http://lists.freedesktop.org/mailman/listinfo/intel-gfx
