[Intel-gfx] [PATCH 3/5] drm/i915: Mark up "sentinel" requests

[Chris Wilson]

Quoting Mika Kuoppala (2019-08-06 15:29:58)
> Chris Wilson <chris at chris-wilson.co.uk> writes:
> 
> > Sometimes we want to emit a terminator request, a request that flushes
> > the pipeline and allows no request to come after it. This can be used
> > for a "preempt-to-idle" to ensure that upon processing the
> > context-switch to that request, all other active contexts have been
> > flushed.
> >
> > Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
> > ---
> >  drivers/gpu/drm/i915/gt/intel_lrc.c |  6 ++++++
> >  drivers/gpu/drm/i915/i915_request.h | 10 ++++++++--
> >  2 files changed, 14 insertions(+), 2 deletions(-)
> >
> > diff --git a/drivers/gpu/drm/i915/gt/intel_lrc.c b/drivers/gpu/drm/i915/gt/intel_lrc.c
> > index 4e45cd972267..59a7e4eb7e2a 100644
> > --- a/drivers/gpu/drm/i915/gt/intel_lrc.c
> > +++ b/drivers/gpu/drm/i915/gt/intel_lrc.c
> > @@ -759,6 +759,9 @@ static bool can_merge_rq(const struct i915_request *prev,
> >       GEM_BUG_ON(prev == next);
> >       GEM_BUG_ON(!assert_priority_queue(prev, next));
> >  
> > +     if (i915_request_has_sentinel(prev))
> > +             return false;
> > +
> >       if (!can_merge_ctx(prev->hw_context, next->hw_context))
> >               return false;
> >  
> > @@ -1250,6 +1253,9 @@ static void execlists_dequeue(struct intel_engine_cs *engine)
> >                               if (last->hw_context == rq->hw_context)
> >                                       goto done;
> >  
> > +                             if (i915_request_has_sentinel(last))
> > +                                     goto done;
> > +
> >                               /*
> >                                * If GVT overrides us we only ever submit
> >                                * port[0], leaving port[1] empty. Note that we
> > diff --git a/drivers/gpu/drm/i915/i915_request.h b/drivers/gpu/drm/i915/i915_request.h
> > index 313df3c37158..c41ecbe0bd0c 100644
> > --- a/drivers/gpu/drm/i915/i915_request.h
> > +++ b/drivers/gpu/drm/i915/i915_request.h
> > @@ -217,8 +217,9 @@ struct i915_request {
> >       unsigned long emitted_jiffies;
> >  
> >       unsigned long flags;
> > -#define I915_REQUEST_WAITBOOST BIT(0)
> > -#define I915_REQUEST_NOPREEMPT BIT(1)
> > +#define I915_REQUEST_WAITBOOST       BIT(0)
> > +#define I915_REQUEST_NOPREEMPT       BIT(1)
> > +#define I915_REQUEST_SENTINEL        BIT(2)
> 
> Would it be possible to use 'empty' ie non payloadable
> requests as a sentinel? (using the request->postfix).
> 
> Or is the advantage here that by attaching it as
> a property, you avoid submitting an extra (empty)?

To achieve the effect I want (effectively preempt-to-idle!), I need to
fill the ELSP with dummies (N different contexts) to prevent submitting
real requests after the preemption heartbeat. So I want an end-of-thread
marker.

I went through
	I915_REQUEST_EOT
	I915_REQUEST_TERMINATOR
	I915_REQEUST_SENTINEL
-Chris