Re: [PATCH 5/5] drm/i915: Cancel non-persistent contexts on close

[Chris Wilson <chris@chris-wilson.co.uk>]

Quoting Bloomfield, Jon (2019-08-06 15:26:07)
> > -----Original Message-----
> > From: Chris Wilson <chris@chris-wilson.co.uk>
> > +
> > +     /*
> > +      * If the user has disabled hangchecking, we can not be sure that
> > +      * the batches will ever complete and let the context be freed.
> > +      * Forcibly kill off any remaining requests in this case.
> > +      */
> 
> Persistence is independent of hangcheck which makes the above comment unrepresentative of the below code.
> Do we even need a comment here, it looks self-explanatory? Instead maybe comment the defaulting state in __createContext() to make the connection between hangcheck and persistence?

Indeed, it can be moved almost verbatim to explain the default behaviour
and expand upon the cgroup wish. (Rather than using hangcheck modparam,
let the sysadmin specify defaults for groups of processes -- planning
for when we remove hangcheck entirely.)
-Chris
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx