drm/i915/guc: Clear terminated attribute bit on GuC preemption context
authorJeff McGee <jeff.mcgee@intel.com>
Wed, 1 Nov 2017 22:16:30 +0000 (15:16 -0700)
committerChris Wilson <chris@chris-wilson.co.uk>
Thu, 2 Nov 2017 08:07:22 +0000 (08:07 +0000)
If GuC firmware performs an engine reset while that engine had a
preemption pending, it will set the terminated attribute bit on our
preemption stage descriptor. GuC firmware retains all pending work
items for a high-priority GuC client, unlike the normal-priority GuC
client where work items are dropped. It wants to make sure the preempt-
to-idle work doesn't run when scheduling resumes, and uses this bit to
inform its scheduler and presumably us as well. Our job is to clear it
for the next preemption after reset, otherwise that and future
preemptions will never complete. We'll just clear it every time.

Signed-off-by: Jeff McGee <jeff.mcgee@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171101221630.25086-1-jeff.mcgee@intel.com
Reviewed-by: Michel Thierry <michel.thierry@intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
drivers/gpu/drm/i915/i915_guc_submission.c

index 3049a07..d14c134 100644 (file)
@@ -590,6 +590,7 @@ static void inject_preempt_context(struct work_struct *work)
        struct intel_guc *guc = container_of(preempt_work, typeof(*guc),
                                             preempt_work[engine->id]);
        struct i915_guc_client *client = guc->preempt_client;
+       struct guc_stage_desc *stage_desc = __get_stage_desc(client);
        struct intel_ring *ring = client->owner->engine[engine->id].ring;
        u32 ctx_desc = lower_32_bits(intel_lr_context_descriptor(client->owner,
                                                                 engine));
@@ -623,6 +624,20 @@ static void inject_preempt_context(struct work_struct *work)
                           ring->tail / sizeof(u64), 0);
        spin_unlock_irq(&client->wq_lock);
 
+       /*
+        * If GuC firmware performs an engine reset while that engine had
+        * a preemption pending, it will set the terminated attribute bit
+        * on our preemption stage descriptor. GuC firmware retains all
+        * pending work items for a high-priority GuC client, unlike the
+        * normal-priority GuC client where work items are dropped. It
+        * wants to make sure the preempt-to-idle work doesn't run when
+        * scheduling resumes, and uses this bit to inform its scheduler
+        * and presumably us as well. Our job is to clear it for the next
+        * preemption after reset, otherwise that and future preemptions
+        * will never complete. We'll just clear it every time.
+        */
+       stage_desc->attribute &= ~GUC_STAGE_DESC_ATTR_TERMINATED;
+
        data[0] = INTEL_GUC_ACTION_REQUEST_PREEMPTION;
        data[1] = client->stage_id;
        data[2] = INTEL_GUC_PREEMPT_OPTION_DROP_WORK_Q |