Quoting Chris Wilson (2019-05-01 14:57:52) > The counter goes to zero at the start of the parking cycle, but the > wakeref itself is held until the end. Likewise, the counter becomes one > at the end of the unparking, but the wakeref is taken first. If we check > the wakeref instead of the counter, we include the unpark/unparking time > as intel_wakeref_is_active(), and do not spuriously declare inactive if > we fail to park (i.e. the parking and wakeref drop is postponed). > > The premature inactive deactivation may result us in randomly stopping > the retire worker too early with a potential for a livelock if that was > the only means by which we were retiring at the time (e.g. in handling > i915_drop_caches). Forget this paragraph, this was not the solution. I need the longer active boundary for the next patch. -Chris _______________________________________________ Intel-gfx mailing list Intel-gfx@xxxxxxxxxxxxxxxxxxxxx https://lists.freedesktop.org/mailman/listinfo/intel-gfx
