[Intel-gfx] [PATCH 20/30] drm/i915: Cancel non-persistent contexts on close

[Bloomfield, Jon]

> -----Original Message-----
> From: Chris Wilson <chris at chris-wilson.co.uk>
> Sent: Wednesday, October 2, 2019 7:24 AM
> To: Bloomfield, Jon <jon.bloomfield at intel.com>; intel-
> gfx at lists.freedesktop.org
> Cc: Joonas Lahtinen <joonas.lahtinen at linux.intel.com>; Winiarski, Michal
> <michal.winiarski at intel.com>
> Subject: RE: [PATCH 20/30] drm/i915: Cancel non-persistent contexts on close
> 
> Quoting Bloomfield, Jon (2019-10-02 14:52:32)
> >
> >
> >
> > Hmmn. Given that disabling hangcheck is an explicit operation, and we
> already change the default setting, can't we make it a hard requirement that
> persistence requires hangcheck? You should not really be able to opt back in to
> persistence if hangcheck is disabled. In fact you could just test for hangcheck
> when deciding whether to kill the context, and force-kill if it is off - that way if
> hangcheck is disabled after a context starts it will get cleaned up.
> 
> Just great, now I got to update the igt to treat i915.enable_hangcheck
> as API!
> -Chris
Don't blame me ;-)
I'm in damage limitation mode. I'd prefer we didn't have persistence at all.