Quoting Mika Kuoppala (2018-02-05 09:25:14)
> Chris Wilson <[email protected]> writes:
> 
> > During testing, we trigger a lot of resets on an unbannable context
> > leading to massive amounts of irrelevant debug spam. Remove the
> > ban_score accounting and message for the unbannable context so that we
> > improve the signal:noise in the log messages for when the unexpected
> > occurs.
> >
> > Signed-off-by: Chris Wilson <[email protected]>
> > Cc: Mika Kuoppala <[email protected]>
> > ---
> >  drivers/gpu/drm/i915/i915_gem.c | 23 +++++++++++------------
> >  1 file changed, 11 insertions(+), 12 deletions(-)
> >
> > diff --git a/drivers/gpu/drm/i915/i915_gem.c 
> > b/drivers/gpu/drm/i915/i915_gem.c
> > index 63308ec016a3..088579324c14 100644
> > --- a/drivers/gpu/drm/i915/i915_gem.c
> > +++ b/drivers/gpu/drm/i915/i915_gem.c
> > @@ -2823,24 +2823,23 @@ i915_gem_object_pwrite_gtt(struct 
> > drm_i915_gem_object *obj,
> >       return 0;
> >  }
> >  
> > -static bool ban_context(const struct i915_gem_context *ctx,
> > -                     unsigned int score)
> > -{
> > -     return (i915_gem_context_is_bannable(ctx) &&
> > -             score >= CONTEXT_SCORE_BAN_THRESHOLD);
> > -}
> > -
> >  static void i915_gem_context_mark_guilty(struct i915_gem_context *ctx)
> >  {
> > -     unsigned int score;
> >       bool banned;
> >  
> >       atomic_inc(&ctx->guilty_count);
> >  
> > -     score = atomic_add_return(CONTEXT_SCORE_GUILTY, &ctx->ban_score);
> > -     banned = ban_context(ctx, score);
> > -     DRM_DEBUG_DRIVER("context %s marked guilty (score %d) banned? %s\n",
> > -                      ctx->name, score, yesno(banned));
> > +     banned = false;
> > +     if (i915_gem_context_is_bannable(ctx)) {
> > +             unsigned int score;
> > +
> > +             score = atomic_add_return(CONTEXT_SCORE_GUILTY,
> > +                                       &ctx->ban_score);
> 
> First I thought we should let the guilty score increment even if
> context is not bannable. To be able to easily identify unbannables
> causing trouble in error states.
> 
> But score zero with guilty count should be even more clear.

We can make that even clearer by shouting in the error state if the hang
occurs in one of the special unbannable contexts.
-Chris
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

Reply via email to