[Intel-gfx] [PATCH 1/3] drm/i915: Rename engines with to match their user interface
Chris Wilson
chris at chris-wilson.co.uk
Wed Aug 7 10:44:23 UTC 2019
Quoting Mika Kuoppala (2019-08-07 11:31:14)
> Chris Wilson <chris at chris-wilson.co.uk> writes:
>
> > During engine setup, we may find that some engines are fused off causing
> > a misalignment between internal names and the instances seen by users,
> > e.g. (I915_ENGINE_CLASS_VIDEO_DECODE, 1) may be vcs2 in hardware.
> > Normally this is invisible to the user, but a few debug interfaces (and
> > our own internal tracing) use the original HW name not the name the user
> > would expect as formed their class:instance tuple. Replace our internal
> > name with the uabi name for consistency, for example in error states.
> >
> > Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=111311
> > Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
> > ---
> > drivers/gpu/drm/i915/gt/intel_engine_cs.c | 37 ++++----------------
> > drivers/gpu/drm/i915/gt/intel_engine_user.c | 21 +++++++++++
> > drivers/gpu/drm/i915/gt/intel_engine_user.h | 2 ++
> > drivers/gpu/drm/i915/gt/selftest_engine_cs.c | 19 +++++-----
> > 4 files changed, 37 insertions(+), 42 deletions(-)
> >
> > diff --git a/drivers/gpu/drm/i915/gt/intel_engine_cs.c b/drivers/gpu/drm/i915/gt/intel_engine_cs.c
> > index d0befd6c023a..d38c114b0964 100644
> > --- a/drivers/gpu/drm/i915/gt/intel_engine_cs.c
> > +++ b/drivers/gpu/drm/i915/gt/intel_engine_cs.c
> > @@ -55,30 +55,6 @@
> >
> > #define GEN8_LR_CONTEXT_OTHER_SIZE ( 2 * PAGE_SIZE)
> >
> > -struct engine_class_info {
> > - const char *name;
> > - u8 uabi_class;
> > -};
> > -
> > -static const struct engine_class_info intel_engine_classes[] = {
> > - [RENDER_CLASS] = {
> > - .name = "rcs",
> > - .uabi_class = I915_ENGINE_CLASS_RENDER,
> > - },
> > - [COPY_ENGINE_CLASS] = {
> > - .name = "bcs",
> > - .uabi_class = I915_ENGINE_CLASS_COPY,
> > - },
> > - [VIDEO_DECODE_CLASS] = {
> > - .name = "vcs",
> > - .uabi_class = I915_ENGINE_CLASS_VIDEO,
> > - },
> > - [VIDEO_ENHANCEMENT_CLASS] = {
> > - .name = "vecs",
> > - .uabi_class = I915_ENGINE_CLASS_VIDEO_ENHANCE,
> > - },
> > -};
> > -
> > #define MAX_MMIO_BASES 3
> > struct engine_info {
> > unsigned int hw_id;
> > @@ -259,11 +235,11 @@ static u32 __engine_mmio_base(struct drm_i915_private *i915,
> > return bases[i].base;
> > }
> >
> > -static void __sprint_engine_name(char *name, const struct engine_info *info)
> > +static void __sprint_engine_name(struct intel_engine_cs *engine)
> > {
> > - WARN_ON(snprintf(name, INTEL_ENGINE_CS_MAX_NAME, "%s%u",
> > - intel_engine_classes[info->class].name,
> > - info->instance) >= INTEL_ENGINE_CS_MAX_NAME);
> > + GEM_WARN_ON(snprintf(engine->name, sizeof(engine->name), "%s'%u",
> ^ ?
>
> unexpected
I needed some short way of identifying the proto-name from the final
name for debugging. (It has to be kept short to fit inside the small
embedded name[].) After some deliberation, I went with class' as being
easy to distinguish and not likely to match other shorthand.
I thought about keeping engine->name and engine->uabi_name, but I think
that leads to longer term confusion. Our debug traces will say one thing,
but uAPI another, with no clear correlation.
It's only a name and I expect we will revisit the topic again.
> > + intel_engine_class_repr(engine),
> > + engine->instance) >= sizeof(engine->name));
> > }
> >
> > void intel_engine_set_hwsp_writemask(struct intel_engine_cs *engine, u32 mask)
> > @@ -292,8 +268,6 @@ static int intel_engine_setup(struct intel_gt *gt, enum intel_engine_id id)
> > const struct engine_info *info = &intel_engines[id];
> > struct intel_engine_cs *engine;
> >
> > - GEM_BUG_ON(info->class >= ARRAY_SIZE(intel_engine_classes));
> > -
> > BUILD_BUG_ON(MAX_ENGINE_CLASS >= BIT(GEN11_ENGINE_CLASS_WIDTH));
> > BUILD_BUG_ON(MAX_ENGINE_INSTANCE >= BIT(GEN11_ENGINE_INSTANCE_WIDTH));
> >
> > @@ -317,11 +291,12 @@ static int intel_engine_setup(struct intel_gt *gt, enum intel_engine_id id)
> > engine->i915 = gt->i915;
> > engine->gt = gt;
> > engine->uncore = gt->uncore;
> > - __sprint_engine_name(engine->name, info);
> > engine->hw_id = engine->guc_id = info->hw_id;
> > engine->mmio_base = __engine_mmio_base(gt->i915, info->mmio_bases);
> > +
> > engine->class = info->class;
> > engine->instance = info->instance;
> > + __sprint_engine_name(engine);
> >
> > /*
> > * To be overridden by the backend on setup. However to facilitate
> > diff --git a/drivers/gpu/drm/i915/gt/intel_engine_user.c b/drivers/gpu/drm/i915/gt/intel_engine_user.c
> > index 68fda1ac3c60..03b4ace578d7 100644
> > --- a/drivers/gpu/drm/i915/gt/intel_engine_user.c
> > +++ b/drivers/gpu/drm/i915/gt/intel_engine_user.c
> > @@ -127,6 +127,22 @@ static void set_scheduler_caps(struct drm_i915_private *i915)
> > i915->caps.scheduler = 0;
> > }
> >
> > +const char *intel_engine_class_repr(struct intel_engine_cs *engine)
>
> I dont mind the repr, but we have been using _str.
>
> > +{
> > + static const char *uabi_names[] = {
> > + [RENDER_CLASS] = "rcs",
> > + [COPY_ENGINE_CLASS] = "bcs",
> > + [VIDEO_DECODE_CLASS] = "vcs",
> > + [VIDEO_ENHANCEMENT_CLASS] = "vecs",
> > + };
> > +
> > + if (engine->class >= ARRAY_SIZE(uabi_names) ||
> > + !uabi_names[engine->class])
> > + return "xxx";
>
> or "unknown" shrug.
We're a bit tight on characters.
> > +
> > + return uabi_names[engine->class];
> > +}
> > +
> > void intel_engines_driver_register(struct drm_i915_private *i915)
> > {
> > u8 uabi_instances[4] = {};
> > @@ -149,6 +165,11 @@ void intel_engines_driver_register(struct drm_i915_private *i915)
> > GEM_BUG_ON(engine->uabi_class >= ARRAY_SIZE(uabi_instances));
> > engine->uabi_instance = uabi_instances[engine->uabi_class]++;
> >
> > + /* Replace the internal name with the final user facing name */
> > + scnprintf(engine->name, sizeof(engine->name), "%s%u",
> > + intel_engine_class_repr(engine),
> > + engine->uabi_instance);
>
> Hmm, this was the reason for ' ?
Yup.
-Chris
More information about the Intel-gfx
mailing list