diff options
Diffstat (limited to 'drivers/gpu/drm/i915/gt')
| -rw-r--r-- | drivers/gpu/drm/i915/gt/gen8_engine_cs.c | 11 | ||||
| -rw-r--r-- | drivers/gpu/drm/i915/gt/intel_engine_cs.c | 1 | ||||
| -rw-r--r-- | drivers/gpu/drm/i915/gt/intel_execlists_submission.c | 2 | ||||
| -rw-r--r-- | drivers/gpu/drm/i915/gt/intel_ggtt.c | 23 | ||||
| -rw-r--r-- | drivers/gpu/drm/i915/gt/intel_lrc.c | 5 | ||||
| -rw-r--r-- | drivers/gpu/drm/i915/gt/uc/intel_guc_submission.c | 38 |
6 files changed, 67 insertions, 13 deletions
diff --git a/drivers/gpu/drm/i915/gt/gen8_engine_cs.c b/drivers/gpu/drm/i915/gt/gen8_engine_cs.c index a4ff55aa5e55..7ad36198aab2 100644 --- a/drivers/gpu/drm/i915/gt/gen8_engine_cs.c +++ b/drivers/gpu/drm/i915/gt/gen8_engine_cs.c @@ -271,8 +271,17 @@ int gen12_emit_flush_rcs(struct i915_request *rq, u32 mode) if (GRAPHICS_VER_FULL(rq->i915) >= IP_VER(12, 70)) bit_group_0 |= PIPE_CONTROL_CCS_FLUSH; + /* + * L3 fabric flush is needed for AUX CCS invalidation + * which happens as part of pipe-control so we can + * ignore PIPE_CONTROL_FLUSH_L3. Also PIPE_CONTROL_FLUSH_L3 + * deals with Protected Memory which is not needed for + * AUX CCS invalidation and lead to unwanted side effects. + */ + if (mode & EMIT_FLUSH) + bit_group_1 |= PIPE_CONTROL_FLUSH_L3; + bit_group_1 |= PIPE_CONTROL_TILE_CACHE_FLUSH; - bit_group_1 |= PIPE_CONTROL_FLUSH_L3; bit_group_1 |= PIPE_CONTROL_RENDER_TARGET_CACHE_FLUSH; bit_group_1 |= PIPE_CONTROL_DEPTH_CACHE_FLUSH; /* Wa_1409600907:tgl,adl-p */ diff --git a/drivers/gpu/drm/i915/gt/intel_engine_cs.c b/drivers/gpu/drm/i915/gt/intel_engine_cs.c index ee15486fed0d..e85d70a62123 100644 --- a/drivers/gpu/drm/i915/gt/intel_engine_cs.c +++ b/drivers/gpu/drm/i915/gt/intel_engine_cs.c @@ -558,7 +558,6 @@ static int intel_engine_setup(struct intel_gt *gt, enum intel_engine_id id, DRIVER_CAPS(i915)->has_logical_contexts = true; ewma__engine_latency_init(&engine->latency); - seqcount_init(&engine->stats.execlists.lock); ATOMIC_INIT_NOTIFIER_HEAD(&engine->context_status_notifier); diff --git a/drivers/gpu/drm/i915/gt/intel_execlists_submission.c b/drivers/gpu/drm/i915/gt/intel_execlists_submission.c index 8a641bcf777c..3292524469d5 100644 --- a/drivers/gpu/drm/i915/gt/intel_execlists_submission.c +++ b/drivers/gpu/drm/i915/gt/intel_execlists_submission.c @@ -3550,6 +3550,8 @@ int intel_execlists_submission_setup(struct intel_engine_cs *engine) logical_ring_default_vfuncs(engine); logical_ring_default_irqs(engine); + seqcount_init(&engine->stats.execlists.lock); + if (engine->flags & I915_ENGINE_HAS_RCS_REG_STATE) rcs_submission_override(engine); diff --git a/drivers/gpu/drm/i915/gt/intel_ggtt.c b/drivers/gpu/drm/i915/gt/intel_ggtt.c index dd0ed941441a..da21f2786b5d 100644 --- a/drivers/gpu/drm/i915/gt/intel_ggtt.c +++ b/drivers/gpu/drm/i915/gt/intel_ggtt.c @@ -511,20 +511,31 @@ void intel_ggtt_unbind_vma(struct i915_address_space *vm, vm->clear_range(vm, vma_res->start, vma_res->vma_size); } +/* + * Reserve the top of the GuC address space for firmware images. Addresses + * beyond GUC_GGTT_TOP in the GuC address space are inaccessible by GuC, + * which makes for a suitable range to hold GuC/HuC firmware images if the + * size of the GGTT is 4G. However, on a 32-bit platform the size of the GGTT + * is limited to 2G, which is less than GUC_GGTT_TOP, but we reserve a chunk + * of the same size anyway, which is far more than needed, to keep the logic + * in uc_fw_ggtt_offset() simple. + */ +#define GUC_TOP_RESERVE_SIZE (SZ_4G - GUC_GGTT_TOP) + static int ggtt_reserve_guc_top(struct i915_ggtt *ggtt) { - u64 size; + u64 offset; int ret; if (!intel_uc_uses_guc(&ggtt->vm.gt->uc)) return 0; - GEM_BUG_ON(ggtt->vm.total <= GUC_GGTT_TOP); - size = ggtt->vm.total - GUC_GGTT_TOP; + GEM_BUG_ON(ggtt->vm.total <= GUC_TOP_RESERVE_SIZE); + offset = ggtt->vm.total - GUC_TOP_RESERVE_SIZE; - ret = i915_gem_gtt_reserve(&ggtt->vm, NULL, &ggtt->uc_fw, size, - GUC_GGTT_TOP, I915_COLOR_UNEVICTABLE, - PIN_NOEVICT); + ret = i915_gem_gtt_reserve(&ggtt->vm, NULL, &ggtt->uc_fw, + GUC_TOP_RESERVE_SIZE, offset, + I915_COLOR_UNEVICTABLE, PIN_NOEVICT); if (ret) drm_dbg(&ggtt->vm.i915->drm, "Failed to reserve top of GGTT for GuC\n"); diff --git a/drivers/gpu/drm/i915/gt/intel_lrc.c b/drivers/gpu/drm/i915/gt/intel_lrc.c index 957d0aeb0c02..c378cc7c953c 100644 --- a/drivers/gpu/drm/i915/gt/intel_lrc.c +++ b/drivers/gpu/drm/i915/gt/intel_lrc.c @@ -1094,6 +1094,9 @@ __lrc_alloc_state(struct intel_context *ce, struct intel_engine_cs *engine) I915_BO_ALLOC_PM_VOLATILE); if (IS_ERR(obj)) { obj = i915_gem_object_create_shmem(engine->i915, context_size); + if (IS_ERR(obj)) + return ERR_CAST(obj); + /* * Wa_22016122933: For Media version 13.0, all Media GT shared * memory needs to be mapped as WC on CPU side and UC (PAT @@ -1102,8 +1105,6 @@ __lrc_alloc_state(struct intel_context *ce, struct intel_engine_cs *engine) if (intel_gt_needs_wa_22016122933(engine->gt)) i915_gem_object_set_cache_coherency(obj, I915_CACHE_NONE); } - if (IS_ERR(obj)) - return ERR_CAST(obj); vma = i915_vma_instance(obj, &engine->gt->ggtt->vm, NULL); if (IS_ERR(vma)) { diff --git a/drivers/gpu/drm/i915/gt/uc/intel_guc_submission.c b/drivers/gpu/drm/i915/gt/uc/intel_guc_submission.c index b5b7f2fe8c78..dc7b40e06e38 100644 --- a/drivers/gpu/drm/i915/gt/uc/intel_guc_submission.c +++ b/drivers/gpu/drm/i915/gt/uc/intel_guc_submission.c @@ -1433,6 +1433,36 @@ static void guc_timestamp_ping(struct work_struct *wrk) int srcu, ret; /* + * Ideally the busyness worker should take a gt pm wakeref because the + * worker only needs to be active while gt is awake. However, the + * gt_park path cancels the worker synchronously and this complicates + * the flow if the worker is also running at the same time. The cancel + * waits for the worker and when the worker releases the wakeref, that + * would call gt_park and would lead to a deadlock. + * + * The resolution is to take the global pm wakeref if runtime pm is + * already active. If not, we don't need to update the busyness stats as + * the stats would already be updated when the gt was parked. + * + * Note: + * - We do not requeue the worker if we cannot take a reference to runtime + * pm since intel_guc_busyness_unpark would requeue the worker in the + * resume path. + * + * - If the gt was parked longer than time taken for GT timestamp to roll + * over, we ignore those rollovers since we don't care about tracking + * the exact GT time. We only care about roll overs when the gt is + * active and running workloads. + * + * - There is a window of time between gt_park and runtime suspend, + * where the worker may run. This is acceptable since the worker will + * not find any new data to update busyness. + */ + wakeref = intel_runtime_pm_get_if_active(>->i915->runtime_pm); + if (!wakeref) + return; + + /* * Synchronize with gt reset to make sure the worker does not * corrupt the engine/guc stats. NB: can't actually block waiting * for a reset to complete as the reset requires flushing out @@ -1440,10 +1470,9 @@ static void guc_timestamp_ping(struct work_struct *wrk) */ ret = intel_gt_reset_trylock(gt, &srcu); if (ret) - return; + goto err_trylock; - with_intel_runtime_pm(>->i915->runtime_pm, wakeref) - __update_guc_busyness_stats(guc); + __update_guc_busyness_stats(guc); /* adjust context stats for overflow */ xa_for_each(&guc->context_lookup, index, ce) @@ -1452,6 +1481,9 @@ static void guc_timestamp_ping(struct work_struct *wrk) intel_gt_reset_unlock(gt, srcu); guc_enable_busyness_worker(guc); + +err_trylock: + intel_runtime_pm_put(>->i915->runtime_pm, wakeref); } static int guc_action_enable_usage_stats(struct intel_guc *guc) |