Message ID | 20180517074055.14638-7-chris@chris-wilson.co.uk (mailing list archive) |
---|---|
State | New, archived |
Headers | show |
On 17/05/2018 08:40, Chris Wilson wrote: > We want to be able to reset the GPU from inside a timer callback > (hardirq context). One step requires us to copy the default context > state over to the guilty context, which means we need to plan in advance > to have that object accessible from within an atomic context. The atomic > context prevents us from pinning the object or in peeking into the > shmemfs backing store (all may sleep), so we choose to pin the > default_state into memory when the engine becomes active. This > compromise allows us to swap out the default state when idle, if > required. > > References: 5692251c254a ("drm/i915/lrc: Scrub the GPU state of the guilty hanging request") > Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> > --- > drivers/gpu/drm/i915/intel_engine_cs.c | 15 +++++++++++++++ > drivers/gpu/drm/i915/intel_lrc.c | 15 ++++----------- > drivers/gpu/drm/i915/intel_ringbuffer.h | 1 + > 3 files changed, 20 insertions(+), 11 deletions(-) > > diff --git a/drivers/gpu/drm/i915/intel_engine_cs.c b/drivers/gpu/drm/i915/intel_engine_cs.c > index 333318b340e1..b1a1ca0758ce 100644 > --- a/drivers/gpu/drm/i915/intel_engine_cs.c > +++ b/drivers/gpu/drm/i915/intel_engine_cs.c > @@ -1082,6 +1082,11 @@ void intel_engines_park(struct drm_i915_private *i915) > if (engine->park) > engine->park(engine); > > + if (engine->pinned_default_state) { > + i915_gem_object_unpin_map(engine->default_state); > + engine->pinned_default_state = NULL; > + } > + > i915_gem_batch_pool_fini(&engine->batch_pool); > engine->execlists.no_priolist = false; > } > @@ -1099,6 +1104,16 @@ void intel_engines_unpark(struct drm_i915_private *i915) > enum intel_engine_id id; > > for_each_engine(engine, i915, id) { > + void *map; > + > + /* Pin the default state for fast resets from atomic context. */ > + map = NULL; > + if (engine->default_state) > + map = i915_gem_object_pin_map(engine->default_state, > + I915_MAP_WB); > + if (!IS_ERR_OR_NULL(map)) > + engine->pinned_default_state = map; > + > if (engine->unpark) > engine->unpark(engine); > > diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c > index f3470b95d64e..49283b3d3ebb 100644 > --- a/drivers/gpu/drm/i915/intel_lrc.c > +++ b/drivers/gpu/drm/i915/intel_lrc.c > @@ -1966,17 +1966,10 @@ static void execlists_reset(struct intel_engine_cs *engine, > * to recreate its own state. > */ > regs = request->hw_context->lrc_reg_state; > - if (engine->default_state) { > - void *defaults; > - > - defaults = i915_gem_object_pin_map(engine->default_state, > - I915_MAP_WB); > - if (!IS_ERR(defaults)) { > - memcpy(regs, /* skip restoring the vanilla PPHWSP */ > - defaults + LRC_STATE_PN * PAGE_SIZE, > - engine->context_size - PAGE_SIZE); > - i915_gem_object_unpin_map(engine->default_state); > - } > + if (engine->pinned_default_state) { > + memcpy(regs, /* skip restoring the vanilla PPHWSP */ > + engine->pinned_default_state + LRC_STATE_PN * PAGE_SIZE, > + engine->context_size - PAGE_SIZE); > } > execlists_init_reg_state(regs, > request->gem_context, engine, request->ring); > diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.h b/drivers/gpu/drm/i915/intel_ringbuffer.h > index 20c4e13efc0d..acef385c4c80 100644 > --- a/drivers/gpu/drm/i915/intel_ringbuffer.h > +++ b/drivers/gpu/drm/i915/intel_ringbuffer.h > @@ -342,6 +342,7 @@ struct intel_engine_cs { > struct i915_timeline timeline; > > struct drm_i915_gem_object *default_state; > + void *pinned_default_state; > > atomic_t irq_count; > unsigned long irq_posted; > Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Regards, Tvrtko
diff --git a/drivers/gpu/drm/i915/intel_engine_cs.c b/drivers/gpu/drm/i915/intel_engine_cs.c index 333318b340e1..b1a1ca0758ce 100644 --- a/drivers/gpu/drm/i915/intel_engine_cs.c +++ b/drivers/gpu/drm/i915/intel_engine_cs.c @@ -1082,6 +1082,11 @@ void intel_engines_park(struct drm_i915_private *i915) if (engine->park) engine->park(engine); + if (engine->pinned_default_state) { + i915_gem_object_unpin_map(engine->default_state); + engine->pinned_default_state = NULL; + } + i915_gem_batch_pool_fini(&engine->batch_pool); engine->execlists.no_priolist = false; } @@ -1099,6 +1104,16 @@ void intel_engines_unpark(struct drm_i915_private *i915) enum intel_engine_id id; for_each_engine(engine, i915, id) { + void *map; + + /* Pin the default state for fast resets from atomic context. */ + map = NULL; + if (engine->default_state) + map = i915_gem_object_pin_map(engine->default_state, + I915_MAP_WB); + if (!IS_ERR_OR_NULL(map)) + engine->pinned_default_state = map; + if (engine->unpark) engine->unpark(engine); diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c index f3470b95d64e..49283b3d3ebb 100644 --- a/drivers/gpu/drm/i915/intel_lrc.c +++ b/drivers/gpu/drm/i915/intel_lrc.c @@ -1966,17 +1966,10 @@ static void execlists_reset(struct intel_engine_cs *engine, * to recreate its own state. */ regs = request->hw_context->lrc_reg_state; - if (engine->default_state) { - void *defaults; - - defaults = i915_gem_object_pin_map(engine->default_state, - I915_MAP_WB); - if (!IS_ERR(defaults)) { - memcpy(regs, /* skip restoring the vanilla PPHWSP */ - defaults + LRC_STATE_PN * PAGE_SIZE, - engine->context_size - PAGE_SIZE); - i915_gem_object_unpin_map(engine->default_state); - } + if (engine->pinned_default_state) { + memcpy(regs, /* skip restoring the vanilla PPHWSP */ + engine->pinned_default_state + LRC_STATE_PN * PAGE_SIZE, + engine->context_size - PAGE_SIZE); } execlists_init_reg_state(regs, request->gem_context, engine, request->ring); diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.h b/drivers/gpu/drm/i915/intel_ringbuffer.h index 20c4e13efc0d..acef385c4c80 100644 --- a/drivers/gpu/drm/i915/intel_ringbuffer.h +++ b/drivers/gpu/drm/i915/intel_ringbuffer.h @@ -342,6 +342,7 @@ struct intel_engine_cs { struct i915_timeline timeline; struct drm_i915_gem_object *default_state; + void *pinned_default_state; atomic_t irq_count; unsigned long irq_posted;
We want to be able to reset the GPU from inside a timer callback (hardirq context). One step requires us to copy the default context state over to the guilty context, which means we need to plan in advance to have that object accessible from within an atomic context. The atomic context prevents us from pinning the object or in peeking into the shmemfs backing store (all may sleep), so we choose to pin the default_state into memory when the engine becomes active. This compromise allows us to swap out the default state when idle, if required. References: 5692251c254a ("drm/i915/lrc: Scrub the GPU state of the guilty hanging request") Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> --- drivers/gpu/drm/i915/intel_engine_cs.c | 15 +++++++++++++++ drivers/gpu/drm/i915/intel_lrc.c | 15 ++++----------- drivers/gpu/drm/i915/intel_ringbuffer.h | 1 + 3 files changed, 20 insertions(+), 11 deletions(-)