[Intel-gfx] [PATCH 6/8] drm/i915/gt: Pull ring submission resume under its caller forcewake
Take advantage of calling xcs_resume under a forcewake by using direct mmio access. In particular, we can avoid the sleeping variants to allow resume to be called from softirq context, required for engine resets. Signed-off-by: Chris Wilson --- .../gpu/drm/i915/gt/intel_ring_submission.c | 96 --- 1 file changed, 42 insertions(+), 54 deletions(-) diff --git a/drivers/gpu/drm/i915/gt/intel_ring_submission.c b/drivers/gpu/drm/i915/gt/intel_ring_submission.c index 7c31126a1b6d..77aec0a94541 100644 --- a/drivers/gpu/drm/i915/gt/intel_ring_submission.c +++ b/drivers/gpu/drm/i915/gt/intel_ring_submission.c @@ -121,31 +121,27 @@ static void set_hwsp(struct intel_engine_cs *engine, u32 offset) hwsp = RING_HWS_PGA(engine->mmio_base); } - intel_uncore_write(engine->uncore, hwsp, offset); - intel_uncore_posting_read(engine->uncore, hwsp); + intel_uncore_write_fw(engine->uncore, hwsp, offset); + intel_uncore_posting_read_fw(engine->uncore, hwsp); } static void flush_cs_tlb(struct intel_engine_cs *engine) { - struct drm_i915_private *dev_priv = engine->i915; - - if (!IS_GEN_RANGE(dev_priv, 6, 7)) + if (!IS_GEN_RANGE(engine->i915, 6, 7)) return; /* ring should be idle before issuing a sync flush*/ - drm_WARN_ON(_priv->drm, - (ENGINE_READ(engine, RING_MI_MODE) & MODE_IDLE) == 0); + GEM_DEBUG_WARN_ON((ENGINE_READ(engine, RING_MI_MODE) & MODE_IDLE) == 0); - ENGINE_WRITE(engine, RING_INSTPM, -_MASKED_BIT_ENABLE(INSTPM_TLB_INVALIDATE | - INSTPM_SYNC_FLUSH)); - if (intel_wait_for_register(engine->uncore, - RING_INSTPM(engine->mmio_base), - INSTPM_SYNC_FLUSH, 0, - 1000)) - drm_err(_priv->drm, - "%s: wait for SyncFlush to complete for TLB invalidation timed out\n", - engine->name); + ENGINE_WRITE_FW(engine, RING_INSTPM, + _MASKED_BIT_ENABLE(INSTPM_TLB_INVALIDATE | + INSTPM_SYNC_FLUSH)); + if (__intel_wait_for_register_fw(engine->uncore, +RING_INSTPM(engine->mmio_base), +INSTPM_SYNC_FLUSH, 0, +2000, 0, NULL)) + ENGINE_TRACE(engine, +"wait for SyncFlush to complete for TLB invalidation timed out\n"); } static void ring_setup_status_page(struct intel_engine_cs *engine) @@ -176,13 +172,13 @@ static void set_pp_dir(struct intel_engine_cs *engine) if (!vm) return; - ENGINE_WRITE(engine, RING_PP_DIR_DCLV, PP_DIR_DCLV_2G); - ENGINE_WRITE(engine, RING_PP_DIR_BASE, pp_dir(vm)); + ENGINE_WRITE_FW(engine, RING_PP_DIR_DCLV, PP_DIR_DCLV_2G); + ENGINE_WRITE_FW(engine, RING_PP_DIR_BASE, pp_dir(vm)); if (INTEL_GEN(engine->i915) >= 7) { - ENGINE_WRITE(engine, -RING_MODE_GEN7, -_MASKED_BIT_ENABLE(GFX_PPGTT_ENABLE)); + ENGINE_WRITE_FW(engine, + RING_MODE_GEN7, + _MASKED_BIT_ENABLE(GFX_PPGTT_ENABLE)); } } @@ -190,13 +186,10 @@ static int xcs_resume(struct intel_engine_cs *engine) { struct drm_i915_private *dev_priv = engine->i915; struct intel_ring *ring = engine->legacy.ring; - int ret = 0; ENGINE_TRACE(engine, "ring:{HEAD:%04x, TAIL:%04x}\n", ring->head, ring->tail); - intel_uncore_forcewake_get(engine->uncore, FORCEWAKE_ALL); - if (HWS_NEEDS_PHYSICAL(dev_priv)) ring_setup_phys_status_page(engine); else @@ -204,16 +197,13 @@ static int xcs_resume(struct intel_engine_cs *engine) intel_breadcrumbs_reset(engine->breadcrumbs); - /* Enforce ordering by reading HEAD register back */ - ENGINE_POSTING_READ(engine, RING_HEAD); - /* * Initialize the ring. This must happen _after_ we've cleared the ring * registers with the above sequence (the readback of the HEAD registers * also enforces ordering), otherwise the hw might lose the new ring * register values. */ - ENGINE_WRITE(engine, RING_START, i915_ggtt_offset(ring->vma)); + ENGINE_WRITE_FW(engine, RING_START, i915_ggtt_offset(ring->vma)); /* Check that the ring offsets point within the ring! */ GEM_BUG_ON(!intel_ring_offset_valid(ring, ring->head)); @@ -223,46 +213,44 @@ static int xcs_resume(struct intel_engine_cs *engine) set_pp_dir(engine); /* First wake the ring up to an empty/idle ring */ - ENGINE_WRITE(engine, RING_HEAD, ring->head); -
[Intel-gfx] [PATCH 6/8] drm/i915/gt: Pull ring submission resume under its caller forcewake
Take advantage of calling xcs_resume under a forcewake by using direct mmio access. In particular, we can avoid the sleeping variants to allow resume to be called from softirq context, required for engine resets. Signed-off-by: Chris Wilson --- .../gpu/drm/i915/gt/intel_ring_submission.c | 98 --- 1 file changed, 43 insertions(+), 55 deletions(-) diff --git a/drivers/gpu/drm/i915/gt/intel_ring_submission.c b/drivers/gpu/drm/i915/gt/intel_ring_submission.c index 7c31126a1b6d..77aec0a94541 100644 --- a/drivers/gpu/drm/i915/gt/intel_ring_submission.c +++ b/drivers/gpu/drm/i915/gt/intel_ring_submission.c @@ -121,31 +121,27 @@ static void set_hwsp(struct intel_engine_cs *engine, u32 offset) hwsp = RING_HWS_PGA(engine->mmio_base); } - intel_uncore_write(engine->uncore, hwsp, offset); - intel_uncore_posting_read(engine->uncore, hwsp); + intel_uncore_write_fw(engine->uncore, hwsp, offset); + intel_uncore_posting_read_fw(engine->uncore, hwsp); } static void flush_cs_tlb(struct intel_engine_cs *engine) { - struct drm_i915_private *dev_priv = engine->i915; - - if (!IS_GEN_RANGE(dev_priv, 6, 7)) + if (!IS_GEN_RANGE(engine->i915, 6, 7)) return; /* ring should be idle before issuing a sync flush*/ - drm_WARN_ON(_priv->drm, - (ENGINE_READ(engine, RING_MI_MODE) & MODE_IDLE) == 0); - - ENGINE_WRITE(engine, RING_INSTPM, -_MASKED_BIT_ENABLE(INSTPM_TLB_INVALIDATE | - INSTPM_SYNC_FLUSH)); - if (intel_wait_for_register(engine->uncore, - RING_INSTPM(engine->mmio_base), - INSTPM_SYNC_FLUSH, 0, - 1000)) - drm_err(_priv->drm, - "%s: wait for SyncFlush to complete for TLB invalidation timed out\n", - engine->name); + GEM_DEBUG_WARN_ON((ENGINE_READ(engine, RING_MI_MODE) & MODE_IDLE) == 0); + + ENGINE_WRITE_FW(engine, RING_INSTPM, + _MASKED_BIT_ENABLE(INSTPM_TLB_INVALIDATE | + INSTPM_SYNC_FLUSH)); + if (__intel_wait_for_register_fw(engine->uncore, +RING_INSTPM(engine->mmio_base), +INSTPM_SYNC_FLUSH, 0, +2000, 0, NULL)) + ENGINE_TRACE(engine, +"wait for SyncFlush to complete for TLB invalidation timed out\n"); } static void ring_setup_status_page(struct intel_engine_cs *engine) @@ -176,13 +172,13 @@ static void set_pp_dir(struct intel_engine_cs *engine) if (!vm) return; - ENGINE_WRITE(engine, RING_PP_DIR_DCLV, PP_DIR_DCLV_2G); - ENGINE_WRITE(engine, RING_PP_DIR_BASE, pp_dir(vm)); + ENGINE_WRITE_FW(engine, RING_PP_DIR_DCLV, PP_DIR_DCLV_2G); + ENGINE_WRITE_FW(engine, RING_PP_DIR_BASE, pp_dir(vm)); if (INTEL_GEN(engine->i915) >= 7) { - ENGINE_WRITE(engine, -RING_MODE_GEN7, -_MASKED_BIT_ENABLE(GFX_PPGTT_ENABLE)); + ENGINE_WRITE_FW(engine, + RING_MODE_GEN7, + _MASKED_BIT_ENABLE(GFX_PPGTT_ENABLE)); } } @@ -190,13 +186,10 @@ static int xcs_resume(struct intel_engine_cs *engine) { struct drm_i915_private *dev_priv = engine->i915; struct intel_ring *ring = engine->legacy.ring; - int ret = 0; ENGINE_TRACE(engine, "ring:{HEAD:%04x, TAIL:%04x}\n", ring->head, ring->tail); - intel_uncore_forcewake_get(engine->uncore, FORCEWAKE_ALL); - if (HWS_NEEDS_PHYSICAL(dev_priv)) ring_setup_phys_status_page(engine); else @@ -204,16 +197,13 @@ static int xcs_resume(struct intel_engine_cs *engine) intel_breadcrumbs_reset(engine->breadcrumbs); - /* Enforce ordering by reading HEAD register back */ - ENGINE_POSTING_READ(engine, RING_HEAD); - /* * Initialize the ring. This must happen _after_ we've cleared the ring * registers with the above sequence (the readback of the HEAD registers * also enforces ordering), otherwise the hw might lose the new ring * register values. */ - ENGINE_WRITE(engine, RING_START, i915_ggtt_offset(ring->vma)); + ENGINE_WRITE_FW(engine, RING_START, i915_ggtt_offset(ring->vma)); /* Check that the ring offsets point within the ring! */ GEM_BUG_ON(!intel_ring_offset_valid(ring, ring->head)); @@ -223,46 +213,44 @@ static int xcs_resume(struct intel_engine_cs *engine) set_pp_dir(engine); /* First wake the ring up to an empty/idle ring */ - ENGINE_WRITE(engine, RING_HEAD, ring->head); -