Quoting Tvrtko Ursulin (2018-10-05 09:37:37)
>
> On 05/10/2018 09:02, Chris Wilson wrote:
> > Add a warning for an allocation stall to make it painfully more obvious
> > while debugging.
> >
> > Signed-off-by: Chris Wilson <chris@xxxxxxxxxxxxxxxxxx>
> > ---
> > drivers/gpu/drm/i915/i915_gem.h | 2 ++
> > drivers/gpu/drm/i915/i915_request.c | 6 ++++++
> > 2 files changed, 8 insertions(+)
> >
> > diff --git a/drivers/gpu/drm/i915/i915_gem.h b/drivers/gpu/drm/i915/i915_gem.h
> > index 599c4f6eb1ea..e2db030fd75a 100644
> > --- a/drivers/gpu/drm/i915/i915_gem.h
> > +++ b/drivers/gpu/drm/i915/i915_gem.h
> > @@ -47,6 +47,7 @@ struct drm_i915_private;
> > #define GEM_DEBUG_DECL(var) var
> > #define GEM_DEBUG_EXEC(expr) expr
> > #define GEM_DEBUG_BUG_ON(expr) GEM_BUG_ON(expr)
> > +#define GEM_DEBUG_WARN(expr...) WARN(expr)
>
> Reminds me of one patch which would look so good preceding this one!
>
> >
> > #else
> >
> > @@ -58,6 +59,7 @@ struct drm_i915_private;
> > #define GEM_DEBUG_DECL(var)
> > #define GEM_DEBUG_EXEC(expr) do { } while (0)
> > #define GEM_DEBUG_BUG_ON(expr)
> > +#define GEM_DEBUG_WARN(expr...) do { } while (0)
> > #endif
> >
> > #if IS_ENABLED(CONFIG_DRM_I915_TRACE_GEM)
> > diff --git a/drivers/gpu/drm/i915/i915_request.c b/drivers/gpu/drm/i915/i915_request.c
> > index abd4dacbab8e..32bf2c9868bf 100644
> > --- a/drivers/gpu/drm/i915/i915_request.c
> > +++ b/drivers/gpu/drm/i915/i915_request.c
> > @@ -647,6 +647,8 @@ i915_request_alloc(struct intel_engine_cs *engine, struct i915_gem_context *ctx)
> > rq = kmem_cache_alloc(i915->requests,
> > GFP_KERNEL | __GFP_RETRY_MAYFAIL | __GFP_NOWARN);
> > if (unlikely(!rq)) {
> > + GEM_DEBUG_DECL(ktime_t start = ktime_get());
> > +
> > i915_retire_requests(i915);
> >
> > /* Ratelimit ourselves to prevent oom from malicious clients */
> > @@ -671,6 +673,10 @@ i915_request_alloc(struct intel_engine_cs *engine, struct i915_gem_context *ctx)
> > ret = -ENOMEM;
> > goto err_unreserve;
> > }
> > +
> > + GEM_DEBUG_WARN(ktime_ms_delta(ktime_get(), start) > 500,
> > + "request allocation delayed for %lldms\n",
> > + ktime_ms_delta(ktime_get(), start));
>
> Usefulness of the arbitrary threshold vs source code pollution?
I'm just trying to prove that the hang is here, so purely for debugging
purposes.
-Chris
_______________________________________________
Intel-gfx mailing list
Intel-gfx@xxxxxxxxxxxxxxxxxxxxx
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
