Op 28-05-2019 om 11:27 schreef Chris Wilson: > Quoting Maarten Lankhorst (2019-05-28 10:24:40) >> Op 27-05-2019 om 13:51 schreef Chris Wilson: >>> Do not allow runtime pm autosuspend to remove userspace GGTT mmaps too >>> quickly. For example, igt sets the autosuspend delay to 0, and so we >>> immediately attempt to perform runtime suspend upon releasing the >>> wakeref. Unfortunately, that involves tearing down GGTT mmaps as they >>> require an active device. >>> >>> Override the autosuspend for GGTT mmaps, by keeping the wakeref around >>> for 250ms after populating the PTE for a fresh mmap. >>> >>> v2: Prefer refcount_t for its under/overflow error detection >>> v3: Flush the user runtime autosuspend prior to system system. >>> >>> Signed-off-by: Chris Wilson <chris@xxxxxxxxxxxxxxxxxx> >>> Cc: Mika Kuoppala <mika.kuoppala@xxxxxxxxxxxxxxx> >>> Reviewed-by: Mika Kuoppala <mika.kuoppala@xxxxxxxxxxxxxxx> >> Can't this extra delay be added to the kernel core? Feels like we're just duplicating autosuspend behavior here.. > Yes. In the end, we would need to adjust the device pm to clamp the user > sysfs value. But really, we want different floors for different > operations as we would expect different wakeup latencies / penalties. > E.g. reaping GTT mmaps is expensive, but if none are in use... > -Chris Seems a bit like the PM QoS management then, but for autosuspend. _______________________________________________ Intel-gfx mailing list Intel-gfx@xxxxxxxxxxxxxxxxxxxxx https://lists.freedesktop.org/mailman/listinfo/intel-gfx
