[Regression] drm/scheduler: track GPU active time per entity

Daniel Vetter daniel at ffwll.ch
Thu Apr 6 17:46:34 UTC 2023


On Thu, Apr 06, 2023 at 07:42:44PM +0200, Daniel Vetter wrote:
> But I think there's tools to make sure we don't dig a complete hole with
> these it sounds like. I guess another topic for pestering the rust folks.

Or to put it very bluntly: Could we make Arc<T> at least runtime enforce
(with the usual lockdep annotation trick like fs_reclaim) that and Drop is
the final one?

If that's the rust Arc<T> linux semantics then I think my worries are 100%
covered. And we'll sort out the trickier type based enforcement for
special cases when they hit us.

The downside is that this is substantially stricter than kref on the C
side, but I think that's a Good Thing :-)
-Daniel
-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch


More information about the dri-devel mailing list