From 22b7a426bbe1ebe1520f92da4cd1617d1e1b5fc4 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Thu, 20 Jun 2019 15:20:51 +0100 Subject: drm/i915/execlists: Preempt-to-busy When using a global seqno, we required a precise stop-the-workd event to handle preemption and unwind the global seqno counter. To accomplish this, we would preempt to a special out-of-band context and wait for the machine to report that it was idle. Given an idle machine, we could very precisely see which requests had completed and which we needed to feed back into the run queue. However, now that we have scrapped the global seqno, we no longer need to precisely unwind the global counter and only track requests by their per-context seqno. This allows us to loosely unwind inflight requests while scheduling a preemption, with the enormous caveat that the requests we put back on the run queue are still _inflight_ (until the preemption request is complete). This makes request tracking much more messy, as at any point then we can see a completed request that we believe is not currently scheduled for execution. We also have to be careful not to rewind RING_TAIL past RING_HEAD on preempting to the running context, and for this we use a semaphore to prevent completion of the request before continuing. To accomplish this feat, we change how we track requests scheduled to the HW. Instead of appending our requests onto a single list as we submit, we track each submission to ELSP as its own block. Then upon receiving the CS preemption event, we promote the pending block to the inflight block (discarding what was previously being tracked). As normal CS completion events arrive, we then remove stale entries from the inflight tracker. v2: Be a tinge paranoid and ensure we flush the write into the HWS page for the GPU semaphore to pick in a timely fashion. Signed-off-by: Chris Wilson Reviewed-by: Mika Kuoppala Link: https://patchwork.freedesktop.org/patch/msgid/20190620142052.19311-1-chris@chris-wilson.co.uk --- drivers/gpu/drm/i915/gt/intel_engine.h | 61 +++++----------------------------- 1 file changed, 8 insertions(+), 53 deletions(-) (limited to 'drivers/gpu/drm/i915/gt/intel_engine.h') diff --git a/drivers/gpu/drm/i915/gt/intel_engine.h b/drivers/gpu/drm/i915/gt/intel_engine.h index 2f1c6871ee95..9bb6ff76680e 100644 --- a/drivers/gpu/drm/i915/gt/intel_engine.h +++ b/drivers/gpu/drm/i915/gt/intel_engine.h @@ -125,71 +125,26 @@ hangcheck_action_to_str(const enum intel_engine_hangcheck_action a) void intel_engines_set_scheduler_caps(struct drm_i915_private *i915); -static inline void -execlists_set_active(struct intel_engine_execlists *execlists, - unsigned int bit) -{ - __set_bit(bit, (unsigned long *)&execlists->active); -} - -static inline bool -execlists_set_active_once(struct intel_engine_execlists *execlists, - unsigned int bit) -{ - return !__test_and_set_bit(bit, (unsigned long *)&execlists->active); -} - -static inline void -execlists_clear_active(struct intel_engine_execlists *execlists, - unsigned int bit) -{ - __clear_bit(bit, (unsigned long *)&execlists->active); -} - -static inline void -execlists_clear_all_active(struct intel_engine_execlists *execlists) +static inline unsigned int +execlists_num_ports(const struct intel_engine_execlists * const execlists) { - execlists->active = 0; + return execlists->port_mask + 1; } -static inline bool -execlists_is_active(const struct intel_engine_execlists *execlists, - unsigned int bit) +static inline struct i915_request * +execlists_active(const struct intel_engine_execlists *execlists) { - return test_bit(bit, (unsigned long *)&execlists->active); + GEM_BUG_ON(execlists->active - execlists->inflight > + execlists_num_ports(execlists)); + return READ_ONCE(*execlists->active); } -void execlists_user_begin(struct intel_engine_execlists *execlists, - const struct execlist_port *port); -void execlists_user_end(struct intel_engine_execlists *execlists); - void execlists_cancel_port_requests(struct intel_engine_execlists * const execlists); struct i915_request * execlists_unwind_incomplete_requests(struct intel_engine_execlists *execlists); -static inline unsigned int -execlists_num_ports(const struct intel_engine_execlists * const execlists) -{ - return execlists->port_mask + 1; -} - -static inline struct execlist_port * -execlists_port_complete(struct intel_engine_execlists * const execlists, - struct execlist_port * const port) -{ - const unsigned int m = execlists->port_mask; - - GEM_BUG_ON(port_index(port, execlists) != 0); - GEM_BUG_ON(!execlists_is_active(execlists, EXECLISTS_ACTIVE_USER)); - - memmove(port, port + 1, m * sizeof(struct execlist_port)); - memset(port + m, 0, sizeof(struct execlist_port)); - - return port; -} - static inline u32 intel_read_status_page(const struct intel_engine_cs *engine, int reg) { -- cgit v1.2.3 From f0c02c1b91888ccac539388eacb0659bf263a557 Mon Sep 17 00:00:00 2001 From: Tvrtko Ursulin Date: Fri, 21 Jun 2019 08:08:10 +0100 Subject: drm/i915: Rename i915_timeline to intel_timeline and move under gt Move all timeline code under gt and rename to intel_gt prefix. Signed-off-by: Tvrtko Ursulin Suggested-by: Chris Wilson Reviewed-by: Chris Wilson Link: https://patchwork.freedesktop.org/patch/msgid/20190621070811.7006-32-tvrtko.ursulin@linux.intel.com --- drivers/gpu/drm/i915/Makefile | 2 +- drivers/gpu/drm/i915/Makefile.header-test | 1 - drivers/gpu/drm/i915/gem/i915_gem_context.c | 10 +- drivers/gpu/drm/i915/gem/i915_gem_context_types.h | 4 +- drivers/gpu/drm/i915/gem/i915_gem_pm.c | 2 +- drivers/gpu/drm/i915/gt/intel_engine.h | 4 +- drivers/gpu/drm/i915/gt/intel_engine_cs.c | 14 +- drivers/gpu/drm/i915/gt/intel_engine_types.h | 4 +- drivers/gpu/drm/i915/gt/intel_lrc.c | 10 +- drivers/gpu/drm/i915/gt/intel_reset.c | 2 +- drivers/gpu/drm/i915/gt/intel_ringbuffer.c | 18 +- drivers/gpu/drm/i915/gt/intel_timeline.c | 591 ++++++++++++++ drivers/gpu/drm/i915/gt/intel_timeline.h | 93 +++ drivers/gpu/drm/i915/gt/intel_timeline_types.h | 67 ++ drivers/gpu/drm/i915/gt/mock_engine.c | 10 +- drivers/gpu/drm/i915/gt/selftest_timeline.c | 845 +++++++++++++++++++++ drivers/gpu/drm/i915/gt/selftests/mock_timeline.c | 29 + drivers/gpu/drm/i915/gt/selftests/mock_timeline.h | 15 + drivers/gpu/drm/i915/i915_drv.h | 2 +- drivers/gpu/drm/i915/i915_gem.c | 8 +- drivers/gpu/drm/i915/i915_gem_gtt.h | 2 +- drivers/gpu/drm/i915/i915_request.c | 14 +- drivers/gpu/drm/i915/i915_request.h | 8 +- drivers/gpu/drm/i915/i915_timeline.c | 591 -------------- drivers/gpu/drm/i915/i915_timeline.h | 93 --- drivers/gpu/drm/i915/i915_timeline_types.h | 67 -- .../gpu/drm/i915/selftests/i915_live_selftests.h | 2 +- .../gpu/drm/i915/selftests/i915_mock_selftests.h | 2 +- drivers/gpu/drm/i915/selftests/i915_timeline.c | 845 --------------------- drivers/gpu/drm/i915/selftests/mock_gem_device.c | 6 +- drivers/gpu/drm/i915/selftests/mock_timeline.c | 29 - drivers/gpu/drm/i915/selftests/mock_timeline.h | 15 - 32 files changed, 1702 insertions(+), 1703 deletions(-) create mode 100644 drivers/gpu/drm/i915/gt/intel_timeline.c create mode 100644 drivers/gpu/drm/i915/gt/intel_timeline.h create mode 100644 drivers/gpu/drm/i915/gt/intel_timeline_types.h create mode 100644 drivers/gpu/drm/i915/gt/selftest_timeline.c create mode 100644 drivers/gpu/drm/i915/gt/selftests/mock_timeline.c create mode 100644 drivers/gpu/drm/i915/gt/selftests/mock_timeline.h delete mode 100644 drivers/gpu/drm/i915/i915_timeline.c delete mode 100644 drivers/gpu/drm/i915/i915_timeline.h delete mode 100644 drivers/gpu/drm/i915/i915_timeline_types.h delete mode 100644 drivers/gpu/drm/i915/selftests/i915_timeline.c delete mode 100644 drivers/gpu/drm/i915/selftests/mock_timeline.c delete mode 100644 drivers/gpu/drm/i915/selftests/mock_timeline.h (limited to 'drivers/gpu/drm/i915/gt/intel_engine.h') diff --git a/drivers/gpu/drm/i915/Makefile b/drivers/gpu/drm/i915/Makefile index 19f8b6745772..84ac0fd1b8d0 100644 --- a/drivers/gpu/drm/i915/Makefile +++ b/drivers/gpu/drm/i915/Makefile @@ -82,6 +82,7 @@ gt-y += \ gt/intel_ringbuffer.o \ gt/intel_mocs.o \ gt/intel_sseu.o \ + gt/intel_timeline.o \ gt/intel_workarounds.o gt-$(CONFIG_DRM_I915_SELFTEST) += \ gt/mock_engine.o @@ -127,7 +128,6 @@ i915-y += \ i915_query.o \ i915_request.o \ i915_scheduler.o \ - i915_timeline.o \ i915_trace_points.o \ i915_vma.o \ intel_wopcm.o diff --git a/drivers/gpu/drm/i915/Makefile.header-test b/drivers/gpu/drm/i915/Makefile.header-test index cb74242f9c3b..b1c3e642f621 100644 --- a/drivers/gpu/drm/i915/Makefile.header-test +++ b/drivers/gpu/drm/i915/Makefile.header-test @@ -12,7 +12,6 @@ header_test := \ i915_priolist_types.h \ i915_reg.h \ i915_scheduler_types.h \ - i915_timeline_types.h \ i915_utils.h \ intel_csr.h \ intel_drv.h \ diff --git a/drivers/gpu/drm/i915/gem/i915_gem_context.c b/drivers/gpu/drm/i915/gem/i915_gem_context.c index fb691535fbf2..628673d1d7f8 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_context.c +++ b/drivers/gpu/drm/i915/gem/i915_gem_context.c @@ -316,7 +316,7 @@ static void i915_gem_context_free(struct i915_gem_context *ctx) mutex_destroy(&ctx->engines_mutex); if (ctx->timeline) - i915_timeline_put(ctx->timeline); + intel_timeline_put(ctx->timeline); kfree(ctx->name); put_pid(ctx->pid); @@ -528,9 +528,9 @@ i915_gem_create_context(struct drm_i915_private *dev_priv, unsigned int flags) } if (flags & I915_CONTEXT_CREATE_FLAGS_SINGLE_TIMELINE) { - struct i915_timeline *timeline; + struct intel_timeline *timeline; - timeline = i915_timeline_create(&dev_priv->gt, NULL); + timeline = intel_timeline_create(&dev_priv->gt, NULL); if (IS_ERR(timeline)) { context_close(ctx); return ERR_CAST(timeline); @@ -2015,8 +2015,8 @@ static int clone_timeline(struct i915_gem_context *dst, GEM_BUG_ON(src->timeline == dst->timeline); if (dst->timeline) - i915_timeline_put(dst->timeline); - dst->timeline = i915_timeline_get(src->timeline); + intel_timeline_put(dst->timeline); + dst->timeline = intel_timeline_get(src->timeline); } return 0; diff --git a/drivers/gpu/drm/i915/gem/i915_gem_context_types.h b/drivers/gpu/drm/i915/gem/i915_gem_context_types.h index cc513410eeef..0ee61482ef94 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_context_types.h +++ b/drivers/gpu/drm/i915/gem/i915_gem_context_types.h @@ -26,7 +26,7 @@ struct pid; struct drm_i915_private; struct drm_i915_file_private; struct i915_address_space; -struct i915_timeline; +struct intel_timeline; struct intel_ring; struct i915_gem_engines { @@ -77,7 +77,7 @@ struct i915_gem_context { struct i915_gem_engines __rcu *engines; struct mutex engines_mutex; /* guards writes to engines */ - struct i915_timeline *timeline; + struct intel_timeline *timeline; /** * @vm: unique address space (GTT) diff --git a/drivers/gpu/drm/i915/gem/i915_gem_pm.c b/drivers/gpu/drm/i915/gem/i915_gem_pm.c index 05011d4a3b88..8f721cf0ab99 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_pm.c +++ b/drivers/gpu/drm/i915/gem/i915_gem_pm.c @@ -38,7 +38,7 @@ static void i915_gem_park(struct drm_i915_private *i915) i915_gem_batch_pool_fini(&engine->batch_pool); } - i915_timelines_park(i915); + intel_timelines_park(i915); i915_vma_parked(i915); i915_globals_park(); diff --git a/drivers/gpu/drm/i915/gt/intel_engine.h b/drivers/gpu/drm/i915/gt/intel_engine.h index 9bb6ff76680e..557b08b13feb 100644 --- a/drivers/gpu/drm/i915/gt/intel_engine.h +++ b/drivers/gpu/drm/i915/gt/intel_engine.h @@ -14,7 +14,7 @@ #include "i915_reg.h" #include "i915_request.h" #include "i915_selftest.h" -#include "i915_timeline.h" +#include "gt/intel_timeline.h" #include "intel_engine_types.h" #include "intel_gpu_commands.h" #include "intel_workarounds.h" @@ -200,7 +200,7 @@ intel_write_status_page(struct intel_engine_cs *engine, int reg, u32 value) struct intel_ring * intel_engine_create_ring(struct intel_engine_cs *engine, - struct i915_timeline *timeline, + struct intel_timeline *timeline, int size); int intel_ring_pin(struct intel_ring *ring); void intel_ring_reset(struct intel_ring *ring, u32 tail); diff --git a/drivers/gpu/drm/i915/gt/intel_engine_cs.c b/drivers/gpu/drm/i915/gt/intel_engine_cs.c index 89edf97d8ad7..e30212e219ec 100644 --- a/drivers/gpu/drm/i915/gt/intel_engine_cs.c +++ b/drivers/gpu/drm/i915/gt/intel_engine_cs.c @@ -724,7 +724,7 @@ void intel_engines_set_scheduler_caps(struct drm_i915_private *i915) struct measure_breadcrumb { struct i915_request rq; - struct i915_timeline timeline; + struct intel_timeline timeline; struct intel_ring ring; u32 cs[1024]; }; @@ -740,9 +740,9 @@ static int measure_breadcrumb_dw(struct intel_engine_cs *engine) if (!frame) return -ENOMEM; - if (i915_timeline_init(&frame->timeline, - engine->gt, - engine->status_page.vma)) + if (intel_timeline_init(&frame->timeline, + engine->gt, + engine->status_page.vma)) goto out_frame; INIT_LIST_HEAD(&frame->ring.request_list); @@ -757,17 +757,17 @@ static int measure_breadcrumb_dw(struct intel_engine_cs *engine) frame->rq.ring = &frame->ring; frame->rq.timeline = &frame->timeline; - dw = i915_timeline_pin(&frame->timeline); + dw = intel_timeline_pin(&frame->timeline); if (dw < 0) goto out_timeline; dw = engine->emit_fini_breadcrumb(&frame->rq, frame->cs) - frame->cs; GEM_BUG_ON(dw & 1); /* RING_TAIL must be qword aligned */ - i915_timeline_unpin(&frame->timeline); + intel_timeline_unpin(&frame->timeline); out_timeline: - i915_timeline_fini(&frame->timeline); + intel_timeline_fini(&frame->timeline); out_frame: kfree(frame); return dw; diff --git a/drivers/gpu/drm/i915/gt/intel_engine_types.h b/drivers/gpu/drm/i915/gt/intel_engine_types.h index fb65e96fa36b..7e056114344e 100644 --- a/drivers/gpu/drm/i915/gt/intel_engine_types.h +++ b/drivers/gpu/drm/i915/gt/intel_engine_types.h @@ -20,7 +20,7 @@ #include "i915_pmu.h" #include "i915_priolist_types.h" #include "i915_selftest.h" -#include "i915_timeline_types.h" +#include "gt/intel_timeline_types.h" #include "intel_sseu.h" #include "intel_wakeref.h" #include "intel_workarounds_types.h" @@ -68,7 +68,7 @@ struct intel_ring { struct i915_vma *vma; void *vaddr; - struct i915_timeline *timeline; + struct intel_timeline *timeline; struct list_head request_list; struct list_head active_link; diff --git a/drivers/gpu/drm/i915/gt/intel_lrc.c b/drivers/gpu/drm/i915/gt/intel_lrc.c index 415fdf2eb997..3abcec3e4e0e 100644 --- a/drivers/gpu/drm/i915/gt/intel_lrc.c +++ b/drivers/gpu/drm/i915/gt/intel_lrc.c @@ -3005,13 +3005,13 @@ err_unpin_ctx: return ret; } -static struct i915_timeline * +static struct intel_timeline * get_timeline(struct i915_gem_context *ctx, struct intel_gt *gt) { if (ctx->timeline) - return i915_timeline_get(ctx->timeline); + return intel_timeline_get(ctx->timeline); else - return i915_timeline_create(gt, NULL); + return intel_timeline_create(gt, NULL); } static int execlists_context_deferred_alloc(struct intel_context *ce, @@ -3021,7 +3021,7 @@ static int execlists_context_deferred_alloc(struct intel_context *ce, struct i915_vma *vma; u32 context_size; struct intel_ring *ring; - struct i915_timeline *timeline; + struct intel_timeline *timeline; int ret; if (ce->state) @@ -3054,7 +3054,7 @@ static int execlists_context_deferred_alloc(struct intel_context *ce, ring = intel_engine_create_ring(engine, timeline, ce->gem_context->ring_size); - i915_timeline_put(timeline); + intel_timeline_put(timeline); if (IS_ERR(ring)) { ret = PTR_ERR(ring); goto error_deref_obj; diff --git a/drivers/gpu/drm/i915/gt/intel_reset.c b/drivers/gpu/drm/i915/gt/intel_reset.c index 5297b3acb56d..3c925af64793 100644 --- a/drivers/gpu/drm/i915/gt/intel_reset.c +++ b/drivers/gpu/drm/i915/gt/intel_reset.c @@ -851,7 +851,7 @@ void i915_gem_set_wedged(struct drm_i915_private *i915) static bool __i915_gem_unset_wedged(struct drm_i915_private *i915) { struct i915_gpu_error *error = &i915->gpu_error; - struct i915_timeline *tl; + struct intel_timeline *tl; if (!test_bit(I915_WEDGED, &error->flags)) return true; diff --git a/drivers/gpu/drm/i915/gt/intel_ringbuffer.c b/drivers/gpu/drm/i915/gt/intel_ringbuffer.c index 9a748be0ce0c..aa483bba04bf 100644 --- a/drivers/gpu/drm/i915/gt/intel_ringbuffer.c +++ b/drivers/gpu/drm/i915/gt/intel_ringbuffer.c @@ -1156,7 +1156,7 @@ int intel_ring_pin(struct intel_ring *ring) if (atomic_fetch_inc(&ring->pin_count)) return 0; - ret = i915_timeline_pin(ring->timeline); + ret = intel_timeline_pin(ring->timeline); if (ret) goto err_unpin; @@ -1194,7 +1194,7 @@ int intel_ring_pin(struct intel_ring *ring) err_ring: i915_vma_unpin(vma); err_timeline: - i915_timeline_unpin(ring->timeline); + intel_timeline_unpin(ring->timeline); err_unpin: atomic_dec(&ring->pin_count); return ret; @@ -1231,7 +1231,7 @@ void intel_ring_unpin(struct intel_ring *ring) ring->vma->obj->pin_global--; i915_vma_unpin(ring->vma); - i915_timeline_unpin(ring->timeline); + intel_timeline_unpin(ring->timeline); } static struct i915_vma *create_ring_vma(struct i915_ggtt *ggtt, int size) @@ -1267,7 +1267,7 @@ err: struct intel_ring * intel_engine_create_ring(struct intel_engine_cs *engine, - struct i915_timeline *timeline, + struct intel_timeline *timeline, int size) { struct drm_i915_private *i915 = engine->i915; @@ -1283,7 +1283,7 @@ intel_engine_create_ring(struct intel_engine_cs *engine, kref_init(&ring->ref); INIT_LIST_HEAD(&ring->request_list); - ring->timeline = i915_timeline_get(timeline); + ring->timeline = intel_timeline_get(timeline); ring->size = size; /* Workaround an erratum on the i830 which causes a hang if @@ -1313,7 +1313,7 @@ void intel_ring_free(struct kref *ref) i915_vma_close(ring->vma); i915_vma_put(ring->vma); - i915_timeline_put(ring->timeline); + intel_timeline_put(ring->timeline); kfree(ring); } @@ -2269,11 +2269,11 @@ int intel_ring_submission_setup(struct intel_engine_cs *engine) int intel_ring_submission_init(struct intel_engine_cs *engine) { - struct i915_timeline *timeline; + struct intel_timeline *timeline; struct intel_ring *ring; int err; - timeline = i915_timeline_create(engine->gt, engine->status_page.vma); + timeline = intel_timeline_create(engine->gt, engine->status_page.vma); if (IS_ERR(timeline)) { err = PTR_ERR(timeline); goto err; @@ -2281,7 +2281,7 @@ int intel_ring_submission_init(struct intel_engine_cs *engine) GEM_BUG_ON(timeline->has_initial_breadcrumb); ring = intel_engine_create_ring(engine, timeline, 32 * PAGE_SIZE); - i915_timeline_put(timeline); + intel_timeline_put(timeline); if (IS_ERR(ring)) { err = PTR_ERR(ring); goto err; diff --git a/drivers/gpu/drm/i915/gt/intel_timeline.c b/drivers/gpu/drm/i915/gt/intel_timeline.c new file mode 100644 index 000000000000..1a3f04458730 --- /dev/null +++ b/drivers/gpu/drm/i915/gt/intel_timeline.c @@ -0,0 +1,591 @@ +/* + * SPDX-License-Identifier: MIT + * + * Copyright © 2016-2018 Intel Corporation + */ + +#include "gt/intel_gt_types.h" + +#include "i915_drv.h" + +#include "i915_active.h" +#include "i915_syncmap.h" +#include "gt/intel_timeline.h" + +#define ptr_set_bit(ptr, bit) ((typeof(ptr))((unsigned long)(ptr) | BIT(bit))) +#define ptr_test_bit(ptr, bit) ((unsigned long)(ptr) & BIT(bit)) + +struct intel_timeline_hwsp { + struct intel_gt *gt; + struct i915_gt_timelines *gt_timelines; + struct list_head free_link; + struct i915_vma *vma; + u64 free_bitmap; +}; + +struct intel_timeline_cacheline { + struct i915_active active; + struct intel_timeline_hwsp *hwsp; + void *vaddr; +#define CACHELINE_BITS 6 +#define CACHELINE_FREE CACHELINE_BITS +}; + +static struct i915_vma *__hwsp_alloc(struct intel_gt *gt) +{ + struct drm_i915_private *i915 = gt->i915; + struct drm_i915_gem_object *obj; + struct i915_vma *vma; + + obj = i915_gem_object_create_internal(i915, PAGE_SIZE); + if (IS_ERR(obj)) + return ERR_CAST(obj); + + i915_gem_object_set_cache_coherency(obj, I915_CACHE_LLC); + + vma = i915_vma_instance(obj, >->ggtt->vm, NULL); + if (IS_ERR(vma)) + i915_gem_object_put(obj); + + return vma; +} + +static struct i915_vma * +hwsp_alloc(struct intel_timeline *timeline, unsigned int *cacheline) +{ + struct i915_gt_timelines *gt = &timeline->gt->timelines; + struct intel_timeline_hwsp *hwsp; + + BUILD_BUG_ON(BITS_PER_TYPE(u64) * CACHELINE_BYTES > PAGE_SIZE); + + spin_lock_irq(>->hwsp_lock); + + /* hwsp_free_list only contains HWSP that have available cachelines */ + hwsp = list_first_entry_or_null(>->hwsp_free_list, + typeof(*hwsp), free_link); + if (!hwsp) { + struct i915_vma *vma; + + spin_unlock_irq(>->hwsp_lock); + + hwsp = kmalloc(sizeof(*hwsp), GFP_KERNEL); + if (!hwsp) + return ERR_PTR(-ENOMEM); + + vma = __hwsp_alloc(timeline->gt); + if (IS_ERR(vma)) { + kfree(hwsp); + return vma; + } + + vma->private = hwsp; + hwsp->gt = timeline->gt; + hwsp->vma = vma; + hwsp->free_bitmap = ~0ull; + hwsp->gt_timelines = gt; + + spin_lock_irq(>->hwsp_lock); + list_add(&hwsp->free_link, >->hwsp_free_list); + } + + GEM_BUG_ON(!hwsp->free_bitmap); + *cacheline = __ffs64(hwsp->free_bitmap); + hwsp->free_bitmap &= ~BIT_ULL(*cacheline); + if (!hwsp->free_bitmap) + list_del(&hwsp->free_link); + + spin_unlock_irq(>->hwsp_lock); + + GEM_BUG_ON(hwsp->vma->private != hwsp); + return hwsp->vma; +} + +static void __idle_hwsp_free(struct intel_timeline_hwsp *hwsp, int cacheline) +{ + struct i915_gt_timelines *gt = hwsp->gt_timelines; + unsigned long flags; + + spin_lock_irqsave(>->hwsp_lock, flags); + + /* As a cacheline becomes available, publish the HWSP on the freelist */ + if (!hwsp->free_bitmap) + list_add_tail(&hwsp->free_link, >->hwsp_free_list); + + GEM_BUG_ON(cacheline >= BITS_PER_TYPE(hwsp->free_bitmap)); + hwsp->free_bitmap |= BIT_ULL(cacheline); + + /* And if no one is left using it, give the page back to the system */ + if (hwsp->free_bitmap == ~0ull) { + i915_vma_put(hwsp->vma); + list_del(&hwsp->free_link); + kfree(hwsp); + } + + spin_unlock_irqrestore(>->hwsp_lock, flags); +} + +static void __idle_cacheline_free(struct intel_timeline_cacheline *cl) +{ + GEM_BUG_ON(!i915_active_is_idle(&cl->active)); + + i915_gem_object_unpin_map(cl->hwsp->vma->obj); + i915_vma_put(cl->hwsp->vma); + __idle_hwsp_free(cl->hwsp, ptr_unmask_bits(cl->vaddr, CACHELINE_BITS)); + + i915_active_fini(&cl->active); + kfree(cl); +} + +static void __cacheline_retire(struct i915_active *active) +{ + struct intel_timeline_cacheline *cl = + container_of(active, typeof(*cl), active); + + i915_vma_unpin(cl->hwsp->vma); + if (ptr_test_bit(cl->vaddr, CACHELINE_FREE)) + __idle_cacheline_free(cl); +} + +static struct intel_timeline_cacheline * +cacheline_alloc(struct intel_timeline_hwsp *hwsp, unsigned int cacheline) +{ + struct intel_timeline_cacheline *cl; + void *vaddr; + + GEM_BUG_ON(cacheline >= BIT(CACHELINE_BITS)); + + cl = kmalloc(sizeof(*cl), GFP_KERNEL); + if (!cl) + return ERR_PTR(-ENOMEM); + + vaddr = i915_gem_object_pin_map(hwsp->vma->obj, I915_MAP_WB); + if (IS_ERR(vaddr)) { + kfree(cl); + return ERR_CAST(vaddr); + } + + i915_vma_get(hwsp->vma); + cl->hwsp = hwsp; + cl->vaddr = page_pack_bits(vaddr, cacheline); + + i915_active_init(hwsp->gt->i915, &cl->active, __cacheline_retire); + + return cl; +} + +static void cacheline_acquire(struct intel_timeline_cacheline *cl) +{ + if (cl && i915_active_acquire(&cl->active)) + __i915_vma_pin(cl->hwsp->vma); +} + +static void cacheline_release(struct intel_timeline_cacheline *cl) +{ + if (cl) + i915_active_release(&cl->active); +} + +static void cacheline_free(struct intel_timeline_cacheline *cl) +{ + GEM_BUG_ON(ptr_test_bit(cl->vaddr, CACHELINE_FREE)); + cl->vaddr = ptr_set_bit(cl->vaddr, CACHELINE_FREE); + + if (i915_active_is_idle(&cl->active)) + __idle_cacheline_free(cl); +} + +int intel_timeline_init(struct intel_timeline *timeline, + struct intel_gt *gt, + struct i915_vma *hwsp) +{ + void *vaddr; + + /* + * Ideally we want a set of engines on a single leaf as we expect + * to mostly be tracking synchronisation between engines. It is not + * a huge issue if this is not the case, but we may want to mitigate + * any page crossing penalties if they become an issue. + * + * Called during early_init before we know how many engines there are. + */ + BUILD_BUG_ON(KSYNCMAP < I915_NUM_ENGINES); + + timeline->gt = gt; + timeline->pin_count = 0; + timeline->has_initial_breadcrumb = !hwsp; + timeline->hwsp_cacheline = NULL; + + if (!hwsp) { + struct intel_timeline_cacheline *cl; + unsigned int cacheline; + + hwsp = hwsp_alloc(timeline, &cacheline); + if (IS_ERR(hwsp)) + return PTR_ERR(hwsp); + + cl = cacheline_alloc(hwsp->private, cacheline); + if (IS_ERR(cl)) { + __idle_hwsp_free(hwsp->private, cacheline); + return PTR_ERR(cl); + } + + timeline->hwsp_cacheline = cl; + timeline->hwsp_offset = cacheline * CACHELINE_BYTES; + + vaddr = page_mask_bits(cl->vaddr); + } else { + timeline->hwsp_offset = I915_GEM_HWS_SEQNO_ADDR; + + vaddr = i915_gem_object_pin_map(hwsp->obj, I915_MAP_WB); + if (IS_ERR(vaddr)) + return PTR_ERR(vaddr); + } + + timeline->hwsp_seqno = + memset(vaddr + timeline->hwsp_offset, 0, CACHELINE_BYTES); + + timeline->hwsp_ggtt = i915_vma_get(hwsp); + GEM_BUG_ON(timeline->hwsp_offset >= hwsp->size); + + timeline->fence_context = dma_fence_context_alloc(1); + + mutex_init(&timeline->mutex); + + INIT_ACTIVE_REQUEST(&timeline->last_request); + INIT_LIST_HEAD(&timeline->requests); + + i915_syncmap_init(&timeline->sync); + + return 0; +} + +static void timelines_init(struct intel_gt *gt) +{ + struct i915_gt_timelines *timelines = >->timelines; + + mutex_init(&timelines->mutex); + INIT_LIST_HEAD(&timelines->active_list); + + spin_lock_init(&timelines->hwsp_lock); + INIT_LIST_HEAD(&timelines->hwsp_free_list); + + /* via i915_gem_wait_for_idle() */ + i915_gem_shrinker_taints_mutex(gt->i915, &timelines->mutex); +} + +void intel_timelines_init(struct drm_i915_private *i915) +{ + timelines_init(&i915->gt); +} + +static void timeline_add_to_active(struct intel_timeline *tl) +{ + struct i915_gt_timelines *gt = &tl->gt->timelines; + + mutex_lock(>->mutex); + list_add(&tl->link, >->active_list); + mutex_unlock(>->mutex); +} + +static void timeline_remove_from_active(struct intel_timeline *tl) +{ + struct i915_gt_timelines *gt = &tl->gt->timelines; + + mutex_lock(>->mutex); + list_del(&tl->link); + mutex_unlock(>->mutex); +} + +static void timelines_park(struct intel_gt *gt) +{ + struct i915_gt_timelines *timelines = >->timelines; + struct intel_timeline *timeline; + + mutex_lock(&timelines->mutex); + list_for_each_entry(timeline, &timelines->active_list, link) { + /* + * All known fences are completed so we can scrap + * the current sync point tracking and start afresh, + * any attempt to wait upon a previous sync point + * will be skipped as the fence was signaled. + */ + i915_syncmap_free(&timeline->sync); + } + mutex_unlock(&timelines->mutex); +} + +/** + * intel_timelines_park - called when the driver idles + * @i915: the drm_i915_private device + * + * When the driver is completely idle, we know that all of our sync points + * have been signaled and our tracking is then entirely redundant. Any request + * to wait upon an older sync point will be completed instantly as we know + * the fence is signaled and therefore we will not even look them up in the + * sync point map. + */ +void intel_timelines_park(struct drm_i915_private *i915) +{ + timelines_park(&i915->gt); +} + +void intel_timeline_fini(struct intel_timeline *timeline) +{ + GEM_BUG_ON(timeline->pin_count); + GEM_BUG_ON(!list_empty(&timeline->requests)); + + i915_syncmap_free(&timeline->sync); + + if (timeline->hwsp_cacheline) + cacheline_free(timeline->hwsp_cacheline); + else + i915_gem_object_unpin_map(timeline->hwsp_ggtt->obj); + + i915_vma_put(timeline->hwsp_ggtt); +} + +struct intel_timeline * +intel_timeline_create(struct intel_gt *gt, struct i915_vma *global_hwsp) +{ + struct intel_timeline *timeline; + int err; + + timeline = kzalloc(sizeof(*timeline), GFP_KERNEL); + if (!timeline) + return ERR_PTR(-ENOMEM); + + err = intel_timeline_init(timeline, gt, global_hwsp); + if (err) { + kfree(timeline); + return ERR_PTR(err); + } + + kref_init(&timeline->kref); + + return timeline; +} + +int intel_timeline_pin(struct intel_timeline *tl) +{ + int err; + + if (tl->pin_count++) + return 0; + GEM_BUG_ON(!tl->pin_count); + + err = i915_vma_pin(tl->hwsp_ggtt, 0, 0, PIN_GLOBAL | PIN_HIGH); + if (err) + goto unpin; + + tl->hwsp_offset = + i915_ggtt_offset(tl->hwsp_ggtt) + + offset_in_page(tl->hwsp_offset); + + cacheline_acquire(tl->hwsp_cacheline); + timeline_add_to_active(tl); + + return 0; + +unpin: + tl->pin_count = 0; + return err; +} + +static u32 timeline_advance(struct intel_timeline *tl) +{ + GEM_BUG_ON(!tl->pin_count); + GEM_BUG_ON(tl->seqno & tl->has_initial_breadcrumb); + + return tl->seqno += 1 + tl->has_initial_breadcrumb; +} + +static void timeline_rollback(struct intel_timeline *tl) +{ + tl->seqno -= 1 + tl->has_initial_breadcrumb; +} + +static noinline int +__intel_timeline_get_seqno(struct intel_timeline *tl, + struct i915_request *rq, + u32 *seqno) +{ + struct intel_timeline_cacheline *cl; + unsigned int cacheline; + struct i915_vma *vma; + void *vaddr; + int err; + + /* + * If there is an outstanding GPU reference to this cacheline, + * such as it being sampled by a HW semaphore on another timeline, + * we cannot wraparound our seqno value (the HW semaphore does + * a strict greater-than-or-equals compare, not i915_seqno_passed). + * So if the cacheline is still busy, we must detach ourselves + * from it and leave it inflight alongside its users. + * + * However, if nobody is watching and we can guarantee that nobody + * will, we could simply reuse the same cacheline. + * + * if (i915_active_request_is_signaled(&tl->last_request) && + * i915_active_is_signaled(&tl->hwsp_cacheline->active)) + * return 0; + * + * That seems unlikely for a busy timeline that needed to wrap in + * the first place, so just replace the cacheline. + */ + + vma = hwsp_alloc(tl, &cacheline); + if (IS_ERR(vma)) { + err = PTR_ERR(vma); + goto err_rollback; + } + + err = i915_vma_pin(vma, 0, 0, PIN_GLOBAL | PIN_HIGH); + if (err) { + __idle_hwsp_free(vma->private, cacheline); + goto err_rollback; + } + + cl = cacheline_alloc(vma->private, cacheline); + if (IS_ERR(cl)) { + err = PTR_ERR(cl); + __idle_hwsp_free(vma->private, cacheline); + goto err_unpin; + } + GEM_BUG_ON(cl->hwsp->vma != vma); + + /* + * Attach the old cacheline to the current request, so that we only + * free it after the current request is retired, which ensures that + * all writes into the cacheline from previous requests are complete. + */ + err = i915_active_ref(&tl->hwsp_cacheline->active, + tl->fence_context, rq); + if (err) + goto err_cacheline; + + cacheline_release(tl->hwsp_cacheline); /* ownership now xfered to rq */ + cacheline_free(tl->hwsp_cacheline); + + i915_vma_unpin(tl->hwsp_ggtt); /* binding kept alive by old cacheline */ + i915_vma_put(tl->hwsp_ggtt); + + tl->hwsp_ggtt = i915_vma_get(vma); + + vaddr = page_mask_bits(cl->vaddr); + tl->hwsp_offset = cacheline * CACHELINE_BYTES; + tl->hwsp_seqno = + memset(vaddr + tl->hwsp_offset, 0, CACHELINE_BYTES); + + tl->hwsp_offset += i915_ggtt_offset(vma); + + cacheline_acquire(cl); + tl->hwsp_cacheline = cl; + + *seqno = timeline_advance(tl); + GEM_BUG_ON(i915_seqno_passed(*tl->hwsp_seqno, *seqno)); + return 0; + +err_cacheline: + cacheline_free(cl); +err_unpin: + i915_vma_unpin(vma); +err_rollback: + timeline_rollback(tl); + return err; +} + +int intel_timeline_get_seqno(struct intel_timeline *tl, + struct i915_request *rq, + u32 *seqno) +{ + *seqno = timeline_advance(tl); + + /* Replace the HWSP on wraparound for HW semaphores */ + if (unlikely(!*seqno && tl->hwsp_cacheline)) + return __intel_timeline_get_seqno(tl, rq, seqno); + + return 0; +} + +static int cacheline_ref(struct intel_timeline_cacheline *cl, + struct i915_request *rq) +{ + return i915_active_ref(&cl->active, rq->fence.context, rq); +} + +int intel_timeline_read_hwsp(struct i915_request *from, + struct i915_request *to, + u32 *hwsp) +{ + struct intel_timeline_cacheline *cl = from->hwsp_cacheline; + struct intel_timeline *tl = from->timeline; + int err; + + GEM_BUG_ON(to->timeline == tl); + + mutex_lock_nested(&tl->mutex, SINGLE_DEPTH_NESTING); + err = i915_request_completed(from); + if (!err) + err = cacheline_ref(cl, to); + if (!err) { + if (likely(cl == tl->hwsp_cacheline)) { + *hwsp = tl->hwsp_offset; + } else { /* across a seqno wrap, recover the original offset */ + *hwsp = i915_ggtt_offset(cl->hwsp->vma) + + ptr_unmask_bits(cl->vaddr, CACHELINE_BITS) * + CACHELINE_BYTES; + } + } + mutex_unlock(&tl->mutex); + + return err; +} + +void intel_timeline_unpin(struct intel_timeline *tl) +{ + GEM_BUG_ON(!tl->pin_count); + if (--tl->pin_count) + return; + + timeline_remove_from_active(tl); + cacheline_release(tl->hwsp_cacheline); + + /* + * Since this timeline is idle, all bariers upon which we were waiting + * must also be complete and so we can discard the last used barriers + * without loss of information. + */ + i915_syncmap_free(&tl->sync); + + __i915_vma_unpin(tl->hwsp_ggtt); +} + +void __intel_timeline_free(struct kref *kref) +{ + struct intel_timeline *timeline = + container_of(kref, typeof(*timeline), kref); + + intel_timeline_fini(timeline); + kfree(timeline); +} + +static void timelines_fini(struct intel_gt *gt) +{ + struct i915_gt_timelines *timelines = >->timelines; + + GEM_BUG_ON(!list_empty(&timelines->active_list)); + GEM_BUG_ON(!list_empty(&timelines->hwsp_free_list)); + + mutex_destroy(&timelines->mutex); +} + +void intel_timelines_fini(struct drm_i915_private *i915) +{ + timelines_fini(&i915->gt); +} + +#if IS_ENABLED(CONFIG_DRM_I915_SELFTEST) +#include "gt/selftests/mock_timeline.c" +#include "gt/selftest_timeline.c" +#endif diff --git a/drivers/gpu/drm/i915/gt/intel_timeline.h b/drivers/gpu/drm/i915/gt/intel_timeline.h new file mode 100644 index 000000000000..e08cebf64833 --- /dev/null +++ b/drivers/gpu/drm/i915/gt/intel_timeline.h @@ -0,0 +1,93 @@ +/* + * Copyright © 2016 Intel Corporation + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS + * IN THE SOFTWARE. + * + */ + +#ifndef I915_TIMELINE_H +#define I915_TIMELINE_H + +#include + +#include "i915_active.h" +#include "i915_syncmap.h" +#include "gt/intel_timeline_types.h" + +int intel_timeline_init(struct intel_timeline *tl, + struct intel_gt *gt, + struct i915_vma *hwsp); +void intel_timeline_fini(struct intel_timeline *tl); + +struct intel_timeline * +intel_timeline_create(struct intel_gt *gt, struct i915_vma *global_hwsp); + +static inline struct intel_timeline * +intel_timeline_get(struct intel_timeline *timeline) +{ + kref_get(&timeline->kref); + return timeline; +} + +void __intel_timeline_free(struct kref *kref); +static inline void intel_timeline_put(struct intel_timeline *timeline) +{ + kref_put(&timeline->kref, __intel_timeline_free); +} + +static inline int __intel_timeline_sync_set(struct intel_timeline *tl, + u64 context, u32 seqno) +{ + return i915_syncmap_set(&tl->sync, context, seqno); +} + +static inline int intel_timeline_sync_set(struct intel_timeline *tl, + const struct dma_fence *fence) +{ + return __intel_timeline_sync_set(tl, fence->context, fence->seqno); +} + +static inline bool __intel_timeline_sync_is_later(struct intel_timeline *tl, + u64 context, u32 seqno) +{ + return i915_syncmap_is_later(&tl->sync, context, seqno); +} + +static inline bool intel_timeline_sync_is_later(struct intel_timeline *tl, + const struct dma_fence *fence) +{ + return __intel_timeline_sync_is_later(tl, fence->context, fence->seqno); +} + +int intel_timeline_pin(struct intel_timeline *tl); +int intel_timeline_get_seqno(struct intel_timeline *tl, + struct i915_request *rq, + u32 *seqno); +void intel_timeline_unpin(struct intel_timeline *tl); + +int intel_timeline_read_hwsp(struct i915_request *from, + struct i915_request *until, + u32 *hwsp_offset); + +void intel_timelines_init(struct drm_i915_private *i915); +void intel_timelines_park(struct drm_i915_private *i915); +void intel_timelines_fini(struct drm_i915_private *i915); + +#endif diff --git a/drivers/gpu/drm/i915/gt/intel_timeline_types.h b/drivers/gpu/drm/i915/gt/intel_timeline_types.h new file mode 100644 index 000000000000..9a71aea7a338 --- /dev/null +++ b/drivers/gpu/drm/i915/gt/intel_timeline_types.h @@ -0,0 +1,67 @@ +/* + * SPDX-License-Identifier: MIT + * + * Copyright © 2016 Intel Corporation + */ + +#ifndef __I915_TIMELINE_TYPES_H__ +#define __I915_TIMELINE_TYPES_H__ + +#include +#include +#include +#include + +#include "i915_active_types.h" + +struct drm_i915_private; +struct i915_vma; +struct intel_timeline_cacheline; +struct i915_syncmap; + +struct intel_timeline { + u64 fence_context; + u32 seqno; + + struct mutex mutex; /* protects the flow of requests */ + + unsigned int pin_count; + const u32 *hwsp_seqno; + struct i915_vma *hwsp_ggtt; + u32 hwsp_offset; + + struct intel_timeline_cacheline *hwsp_cacheline; + + bool has_initial_breadcrumb; + + /** + * List of breadcrumbs associated with GPU requests currently + * outstanding. + */ + struct list_head requests; + + /* Contains an RCU guarded pointer to the last request. No reference is + * held to the request, users must carefully acquire a reference to + * the request using i915_active_request_get_request_rcu(), or hold the + * struct_mutex. + */ + struct i915_active_request last_request; + + /** + * We track the most recent seqno that we wait on in every context so + * that we only have to emit a new await and dependency on a more + * recent sync point. As the contexts may be executed out-of-order, we + * have to track each individually and can not rely on an absolute + * global_seqno. When we know that all tracked fences are completed + * (i.e. when the driver is idle), we know that the syncmap is + * redundant and we can discard it without loss of generality. + */ + struct i915_syncmap *sync; + + struct list_head link; + struct intel_gt *gt; + + struct kref kref; +}; + +#endif /* __I915_TIMELINE_TYPES_H__ */ diff --git a/drivers/gpu/drm/i915/gt/mock_engine.c b/drivers/gpu/drm/i915/gt/mock_engine.c index 423027aa71cd..bf0974b12f3d 100644 --- a/drivers/gpu/drm/i915/gt/mock_engine.c +++ b/drivers/gpu/drm/i915/gt/mock_engine.c @@ -33,15 +33,15 @@ struct mock_ring { struct intel_ring base; - struct i915_timeline timeline; + struct intel_timeline timeline; }; -static void mock_timeline_pin(struct i915_timeline *tl) +static void mock_timeline_pin(struct intel_timeline *tl) { tl->pin_count++; } -static void mock_timeline_unpin(struct i915_timeline *tl) +static void mock_timeline_unpin(struct intel_timeline *tl) { GEM_BUG_ON(!tl->pin_count); tl->pin_count--; @@ -56,7 +56,7 @@ static struct intel_ring *mock_ring(struct intel_engine_cs *engine) if (!ring) return NULL; - if (i915_timeline_init(&ring->timeline, engine->gt, NULL)) { + if (intel_timeline_init(&ring->timeline, engine->gt, NULL)) { kfree(ring); return NULL; } @@ -78,7 +78,7 @@ static void mock_ring_free(struct intel_ring *base) { struct mock_ring *ring = container_of(base, typeof(*ring), base); - i915_timeline_fini(&ring->timeline); + intel_timeline_fini(&ring->timeline); kfree(ring); } diff --git a/drivers/gpu/drm/i915/gt/selftest_timeline.c b/drivers/gpu/drm/i915/gt/selftest_timeline.c new file mode 100644 index 000000000000..193cc564ade2 --- /dev/null +++ b/drivers/gpu/drm/i915/gt/selftest_timeline.c @@ -0,0 +1,845 @@ +/* + * SPDX-License-Identifier: MIT + * + * Copyright © 2017-2018 Intel Corporation + */ + +#include + +#include "gem/i915_gem_pm.h" + +#include "../selftests/i915_random.h" +#include "../i915_selftest.h" + +#include "../selftests/igt_flush_test.h" +#include "../selftests/mock_gem_device.h" +#include "selftests/mock_timeline.h" + +static struct page *hwsp_page(struct intel_timeline *tl) +{ + struct drm_i915_gem_object *obj = tl->hwsp_ggtt->obj; + + GEM_BUG_ON(!i915_gem_object_has_pinned_pages(obj)); + return sg_page(obj->mm.pages->sgl); +} + +static unsigned long hwsp_cacheline(struct intel_timeline *tl) +{ + unsigned long address = (unsigned long)page_address(hwsp_page(tl)); + + return (address + tl->hwsp_offset) / CACHELINE_BYTES; +} + +#define CACHELINES_PER_PAGE (PAGE_SIZE / CACHELINE_BYTES) + +struct mock_hwsp_freelist { + struct drm_i915_private *i915; + struct radix_tree_root cachelines; + struct intel_timeline **history; + unsigned long count, max; + struct rnd_state prng; +}; + +enum { + SHUFFLE = BIT(0), +}; + +static void __mock_hwsp_record(struct mock_hwsp_freelist *state, + unsigned int idx, + struct intel_timeline *tl) +{ + tl = xchg(&state->history[idx], tl); + if (tl) { + radix_tree_delete(&state->cachelines, hwsp_cacheline(tl)); + intel_timeline_put(tl); + } +} + +static int __mock_hwsp_timeline(struct mock_hwsp_freelist *state, + unsigned int count, + unsigned int flags) +{ + struct intel_timeline *tl; + unsigned int idx; + + while (count--) { + unsigned long cacheline; + int err; + + tl = intel_timeline_create(&state->i915->gt, NULL); + if (IS_ERR(tl)) + return PTR_ERR(tl); + + cacheline = hwsp_cacheline(tl); + err = radix_tree_insert(&state->cachelines, cacheline, tl); + if (err) { + if (err == -EEXIST) { + pr_err("HWSP cacheline %lu already used; duplicate allocation!\n", + cacheline); + } + intel_timeline_put(tl); + return err; + } + + idx = state->count++ % state->max; + __mock_hwsp_record(state, idx, tl); + } + + if (flags & SHUFFLE) + i915_prandom_shuffle(state->history, + sizeof(*state->history), + min(state->count, state->max), + &state->prng); + + count = i915_prandom_u32_max_state(min(state->count, state->max), + &state->prng); + while (count--) { + idx = --state->count % state->max; + __mock_hwsp_record(state, idx, NULL); + } + + return 0; +} + +static int mock_hwsp_freelist(void *arg) +{ + struct mock_hwsp_freelist state; + const struct { + const char *name; + unsigned int flags; + } phases[] = { + { "linear", 0 }, + { "shuffled", SHUFFLE }, + { }, + }, *p; + unsigned int na; + int err = 0; + + INIT_RADIX_TREE(&state.cachelines, GFP_KERNEL); + state.prng = I915_RND_STATE_INITIALIZER(i915_selftest.random_seed); + + state.i915 = mock_gem_device(); + if (!state.i915) + return -ENOMEM; + + /* + * Create a bunch of timelines and check that their HWSP do not overlap. + * Free some, and try again. + */ + + state.max = PAGE_SIZE / sizeof(*state.history); + state.count = 0; + state.history = kcalloc(state.max, sizeof(*state.history), GFP_KERNEL); + if (!state.history) { + err = -ENOMEM; + goto err_put; + } + + mutex_lock(&state.i915->drm.struct_mutex); + for (p = phases; p->name; p++) { + pr_debug("%s(%s)\n", __func__, p->name); + for_each_prime_number_from(na, 1, 2 * CACHELINES_PER_PAGE) { + err = __mock_hwsp_timeline(&state, na, p->flags); + if (err) + goto out; + } + } + +out: + for (na = 0; na < state.max; na++) + __mock_hwsp_record(&state, na, NULL); + mutex_unlock(&state.i915->drm.struct_mutex); + kfree(state.history); +err_put: + drm_dev_put(&state.i915->drm); + return err; +} + +struct __igt_sync { + const char *name; + u32 seqno; + bool expected; + bool set; +}; + +static int __igt_sync(struct intel_timeline *tl, + u64 ctx, + const struct __igt_sync *p, + const char *name) +{ + int ret; + + if (__intel_timeline_sync_is_later(tl, ctx, p->seqno) != p->expected) { + pr_err("%s: %s(ctx=%llu, seqno=%u) expected passed %s but failed\n", + name, p->name, ctx, p->seqno, yesno(p->expected)); + return -EINVAL; + } + + if (p->set) { + ret = __intel_timeline_sync_set(tl, ctx, p->seqno); + if (ret) + return ret; + } + + return 0; +} + +static int igt_sync(void *arg) +{ + const struct __igt_sync pass[] = { + { "unset", 0, false, false }, + { "new", 0, false, true }, + { "0a", 0, true, true }, + { "1a", 1, false, true }, + { "1b", 1, true, true }, + { "0b", 0, true, false }, + { "2a", 2, false, true }, + { "4", 4, false, true }, + { "INT_MAX", INT_MAX, false, true }, + { "INT_MAX-1", INT_MAX-1, true, false }, + { "INT_MAX+1", (u32)INT_MAX+1, false, true }, + { "INT_MAX", INT_MAX, true, false }, + { "UINT_MAX", UINT_MAX, false, true }, + { "wrap", 0, false, true }, + { "unwrap", UINT_MAX, true, false }, + {}, + }, *p; + struct intel_timeline tl; + int order, offset; + int ret = -ENODEV; + + mock_timeline_init(&tl, 0); + for (p = pass; p->name; p++) { + for (order = 1; order < 64; order++) { + for (offset = -1; offset <= (order > 1); offset++) { + u64 ctx = BIT_ULL(order) + offset; + + ret = __igt_sync(&tl, ctx, p, "1"); + if (ret) + goto out; + } + } + } + mock_timeline_fini(&tl); + + mock_timeline_init(&tl, 0); + for (order = 1; order < 64; order++) { + for (offset = -1; offset <= (order > 1); offset++) { + u64 ctx = BIT_ULL(order) + offset; + + for (p = pass; p->name; p++) { + ret = __igt_sync(&tl, ctx, p, "2"); + if (ret) + goto out; + } + } + } + +out: + mock_timeline_fini(&tl); + return ret; +} + +static unsigned int random_engine(struct rnd_state *rnd) +{ + return i915_prandom_u32_max_state(I915_NUM_ENGINES, rnd); +} + +static int bench_sync(void *arg) +{ + struct rnd_state prng; + struct intel_timeline tl; + unsigned long end_time, count; + u64 prng32_1M; + ktime_t kt; + int order, last_order; + + mock_timeline_init(&tl, 0); + + /* Lookups from cache are very fast and so the random number generation + * and the loop itself becomes a significant factor in the per-iteration + * timings. We try to compensate the results by measuring the overhead + * of the prng and subtract it from the reported results. + */ + prandom_seed_state(&prng, i915_selftest.random_seed); + count = 0; + kt = ktime_get(); + end_time = jiffies + HZ/10; + do { + u32 x; + + /* Make sure the compiler doesn't optimise away the prng call */ + WRITE_ONCE(x, prandom_u32_state(&prng)); + + count++; + } while (!time_after(jiffies, end_time)); + kt = ktime_sub(ktime_get(), kt); + pr_debug("%s: %lu random evaluations, %lluns/prng\n", + __func__, count, (long long)div64_ul(ktime_to_ns(kt), count)); + prng32_1M = div64_ul(ktime_to_ns(kt) << 20, count); + + /* Benchmark (only) setting random context ids */ + prandom_seed_state(&prng, i915_selftest.random_seed); + count = 0; + kt = ktime_get(); + end_time = jiffies + HZ/10; + do { + u64 id = i915_prandom_u64_state(&prng); + + __intel_timeline_sync_set(&tl, id, 0); + count++; + } while (!time_after(jiffies, end_time)); + kt = ktime_sub(ktime_get(), kt); + kt = ktime_sub_ns(kt, (count * prng32_1M * 2) >> 20); + pr_info("%s: %lu random insertions, %lluns/insert\n", + __func__, count, (long long)div64_ul(ktime_to_ns(kt), count)); + + /* Benchmark looking up the exact same context ids as we just set */ + prandom_seed_state(&prng, i915_selftest.random_seed); + end_time = count; + kt = ktime_get(); + while (end_time--) { + u64 id = i915_prandom_u64_state(&prng); + + if (!__intel_timeline_sync_is_later(&tl, id, 0)) { + mock_timeline_fini(&tl); + pr_err("Lookup of %llu failed\n", id); + return -EINVAL; + } + } + kt = ktime_sub(ktime_get(), kt); + kt = ktime_sub_ns(kt, (count * prng32_1M * 2) >> 20); + pr_info("%s: %lu random lookups, %lluns/lookup\n", + __func__, count, (long long)div64_ul(ktime_to_ns(kt), count)); + + mock_timeline_fini(&tl); + cond_resched(); + + mock_timeline_init(&tl, 0); + + /* Benchmark setting the first N (in order) contexts */ + count = 0; + kt = ktime_get(); + end_time = jiffies + HZ/10; + do { + __intel_timeline_sync_set(&tl, count++, 0); + } while (!time_after(jiffies, end_time)); + kt = ktime_sub(ktime_get(), kt); + pr_info("%s: %lu in-order insertions, %lluns/insert\n", + __func__, count, (long long)div64_ul(ktime_to_ns(kt), count)); + + /* Benchmark looking up the exact same context ids as we just set */ + end_time = count; + kt = ktime_get(); + while (end_time--) { + if (!__intel_timeline_sync_is_later(&tl, end_time, 0)) { + pr_err("Lookup of %lu failed\n", end_time); + mock_timeline_fini(&tl); + return -EINVAL; + } + } + kt = ktime_sub(ktime_get(), kt); + pr_info("%s: %lu in-order lookups, %lluns/lookup\n", + __func__, count, (long long)div64_ul(ktime_to_ns(kt), count)); + + mock_timeline_fini(&tl); + cond_resched(); + + mock_timeline_init(&tl, 0); + + /* Benchmark searching for a random context id and maybe changing it */ + prandom_seed_state(&prng, i915_selftest.random_seed); + count = 0; + kt = ktime_get(); + end_time = jiffies + HZ/10; + do { + u32 id = random_engine(&prng); + u32 seqno = prandom_u32_state(&prng); + + if (!__intel_timeline_sync_is_later(&tl, id, seqno)) + __intel_timeline_sync_set(&tl, id, seqno); + + count++; + } while (!time_after(jiffies, end_time)); + kt = ktime_sub(ktime_get(), kt); + kt = ktime_sub_ns(kt, (count * prng32_1M * 2) >> 20); + pr_info("%s: %lu repeated insert/lookups, %lluns/op\n", + __func__, count, (long long)div64_ul(ktime_to_ns(kt), count)); + mock_timeline_fini(&tl); + cond_resched(); + + /* Benchmark searching for a known context id and changing the seqno */ + for (last_order = 1, order = 1; order < 32; + ({ int tmp = last_order; last_order = order; order += tmp; })) { + unsigned int mask = BIT(order) - 1; + + mock_timeline_init(&tl, 0); + + count = 0; + kt = ktime_get(); + end_time = jiffies + HZ/10; + do { + /* Without assuming too many details of the underlying + * implementation, try to identify its phase-changes + * (if any)! + */ + u64 id = (u64)(count & mask) << order; + + __intel_timeline_sync_is_later(&tl, id, 0); + __intel_timeline_sync_set(&tl, id, 0); + + count++; + } while (!time_after(jiffies, end_time)); + kt = ktime_sub(ktime_get(), kt); + pr_info("%s: %lu cyclic/%d insert/lookups, %lluns/op\n", + __func__, count, order, + (long long)div64_ul(ktime_to_ns(kt), count)); + mock_timeline_fini(&tl); + cond_resched(); + } + + return 0; +} + +int intel_timeline_mock_selftests(void) +{ + static const struct i915_subtest tests[] = { + SUBTEST(mock_hwsp_freelist), + SUBTEST(igt_sync), + SUBTEST(bench_sync), + }; + + return i915_subtests(tests, NULL); +} + +static int emit_ggtt_store_dw(struct i915_request *rq, u32 addr, u32 value) +{ + u32 *cs; + + cs = intel_ring_begin(rq, 4); + if (IS_ERR(cs)) + return PTR_ERR(cs); + + if (INTEL_GEN(rq->i915) >= 8) { + *cs++ = MI_STORE_DWORD_IMM_GEN4 | MI_USE_GGTT; + *cs++ = addr; + *cs++ = 0; + *cs++ = value; + } else if (INTEL_GEN(rq->i915) >= 4) { + *cs++ = MI_STORE_DWORD_IMM_GEN4 | MI_USE_GGTT; + *cs++ = 0; + *cs++ = addr; + *cs++ = value; + } else { + *cs++ = MI_STORE_DWORD_IMM | MI_MEM_VIRTUAL; + *cs++ = addr; + *cs++ = value; + *cs++ = MI_NOOP; + } + + intel_ring_advance(rq, cs); + + return 0; +} + +static struct i915_request * +tl_write(struct intel_timeline *tl, struct intel_engine_cs *engine, u32 value) +{ + struct i915_request *rq; + int err; + + lockdep_assert_held(&tl->gt->i915->drm.struct_mutex); /* lazy rq refs */ + + err = intel_timeline_pin(tl); + if (err) { + rq = ERR_PTR(err); + goto out; + } + + rq = i915_request_create(engine->kernel_context); + if (IS_ERR(rq)) + goto out_unpin; + + err = emit_ggtt_store_dw(rq, tl->hwsp_offset, value); + i915_request_add(rq); + if (err) + rq = ERR_PTR(err); + +out_unpin: + intel_timeline_unpin(tl); +out: + if (IS_ERR(rq)) + pr_err("Failed to write to timeline!\n"); + return rq; +} + +static struct intel_timeline * +checked_intel_timeline_create(struct drm_i915_private *i915) +{ + struct intel_timeline *tl; + + tl = intel_timeline_create(&i915->gt, NULL); + if (IS_ERR(tl)) + return tl; + + if (*tl->hwsp_seqno != tl->seqno) { + pr_err("Timeline created with incorrect breadcrumb, found %x, expected %x\n", + *tl->hwsp_seqno, tl->seqno); + intel_timeline_put(tl); + return ERR_PTR(-EINVAL); + } + + return tl; +} + +static int live_hwsp_engine(void *arg) +{ +#define NUM_TIMELINES 4096 + struct drm_i915_private *i915 = arg; + struct intel_timeline **timelines; + struct intel_engine_cs *engine; + enum intel_engine_id id; + intel_wakeref_t wakeref; + unsigned long count, n; + int err = 0; + + /* + * Create a bunch of timelines and check we can write + * independently to each of their breadcrumb slots. + */ + + timelines = kvmalloc_array(NUM_TIMELINES * I915_NUM_ENGINES, + sizeof(*timelines), + GFP_KERNEL); + if (!timelines) + return -ENOMEM; + + mutex_lock(&i915->drm.struct_mutex); + wakeref = intel_runtime_pm_get(&i915->runtime_pm); + + count = 0; + for_each_engine(engine, i915, id) { + if (!intel_engine_can_store_dword(engine)) + continue; + + for (n = 0; n < NUM_TIMELINES; n++) { + struct intel_timeline *tl; + struct i915_request *rq; + + tl = checked_intel_timeline_create(i915); + if (IS_ERR(tl)) { + err = PTR_ERR(tl); + goto out; + } + + rq = tl_write(tl, engine, count); + if (IS_ERR(rq)) { + intel_timeline_put(tl); + err = PTR_ERR(rq); + goto out; + } + + timelines[count++] = tl; + } + } + +out: + if (igt_flush_test(i915, I915_WAIT_LOCKED)) + err = -EIO; + + for (n = 0; n < count; n++) { + struct intel_timeline *tl = timelines[n]; + + if (!err && *tl->hwsp_seqno != n) { + pr_err("Invalid seqno stored in timeline %lu, found 0x%x\n", + n, *tl->hwsp_seqno); + err = -EINVAL; + } + intel_timeline_put(tl); + } + + intel_runtime_pm_put(&i915->runtime_pm, wakeref); + mutex_unlock(&i915->drm.struct_mutex); + + kvfree(timelines); + + return err; +#undef NUM_TIMELINES +} + +static int live_hwsp_alternate(void *arg) +{ +#define NUM_TIMELINES 4096 + struct drm_i915_private *i915 = arg; + struct intel_timeline **timelines; + struct intel_engine_cs *engine; + enum intel_engine_id id; + intel_wakeref_t wakeref; + unsigned long count, n; + int err = 0; + + /* + * Create a bunch of timelines and check we can write + * independently to each of their breadcrumb slots with adjacent + * engines. + */ + + timelines = kvmalloc_array(NUM_TIMELINES * I915_NUM_ENGINES, + sizeof(*timelines), + GFP_KERNEL); + if (!timelines) + return -ENOMEM; + + mutex_lock(&i915->drm.struct_mutex); + wakeref = intel_runtime_pm_get(&i915->runtime_pm); + + count = 0; + for (n = 0; n < NUM_TIMELINES; n++) { + for_each_engine(engine, i915, id) { + struct intel_timeline *tl; + struct i915_request *rq; + + if (!intel_engine_can_store_dword(engine)) + continue; + + tl = checked_intel_timeline_create(i915); + if (IS_ERR(tl)) { + err = PTR_ERR(tl); + goto out; + } + + rq = tl_write(tl, engine, count); + if (IS_ERR(rq)) { + intel_timeline_put(tl); + err = PTR_ERR(rq); + goto out; + } + + timelines[count++] = tl; + } + } + +out: + if (igt_flush_test(i915, I915_WAIT_LOCKED)) + err = -EIO; + + for (n = 0; n < count; n++) { + struct intel_timeline *tl = timelines[n]; + + if (!err && *tl->hwsp_seqno != n) { + pr_err("Invalid seqno stored in timeline %lu, found 0x%x\n", + n, *tl->hwsp_seqno); + err = -EINVAL; + } + intel_timeline_put(tl); + } + + intel_runtime_pm_put(&i915->runtime_pm, wakeref); + mutex_unlock(&i915->drm.struct_mutex); + + kvfree(timelines); + + return err; +#undef NUM_TIMELINES +} + +static int live_hwsp_wrap(void *arg) +{ + struct drm_i915_private *i915 = arg; + struct intel_engine_cs *engine; + struct intel_timeline *tl; + enum intel_engine_id id; + intel_wakeref_t wakeref; + int err = 0; + + /* + * Across a seqno wrap, we need to keep the old cacheline alive for + * foreign GPU references. + */ + + mutex_lock(&i915->drm.struct_mutex); + wakeref = intel_runtime_pm_get(&i915->runtime_pm); + + tl = intel_timeline_create(&i915->gt, NULL); + if (IS_ERR(tl)) { + err = PTR_ERR(tl); + goto out_rpm; + } + if (!tl->has_initial_breadcrumb || !tl->hwsp_cacheline) + goto out_free; + + err = intel_timeline_pin(tl); + if (err) + goto out_free; + + for_each_engine(engine, i915, id) { + const u32 *hwsp_seqno[2]; + struct i915_request *rq; + u32 seqno[2]; + + if (!intel_engine_can_store_dword(engine)) + continue; + + rq = i915_request_create(engine->kernel_context); + if (IS_ERR(rq)) { + err = PTR_ERR(rq); + goto out; + } + + tl->seqno = -4u; + + err = intel_timeline_get_seqno(tl, rq, &seqno[0]); + if (err) { + i915_request_add(rq); + goto out; + } + pr_debug("seqno[0]:%08x, hwsp_offset:%08x\n", + seqno[0], tl->hwsp_offset); + + err = emit_ggtt_store_dw(rq, tl->hwsp_offset, seqno[0]); + if (err) { + i915_request_add(rq); + goto out; + } + hwsp_seqno[0] = tl->hwsp_seqno; + + err = intel_timeline_get_seqno(tl, rq, &seqno[1]); + if (err) { + i915_request_add(rq); + goto out; + } + pr_debug("seqno[1]:%08x, hwsp_offset:%08x\n", + seqno[1], tl->hwsp_offset); + + err = emit_ggtt_store_dw(rq, tl->hwsp_offset, seqno[1]); + if (err) { + i915_request_add(rq); + goto out; + } + hwsp_seqno[1] = tl->hwsp_seqno; + + /* With wrap should come a new hwsp */ + GEM_BUG_ON(seqno[1] >= seqno[0]); + GEM_BUG_ON(hwsp_seqno[0] == hwsp_seqno[1]); + + i915_request_add(rq); + + if (i915_request_wait(rq, 0, HZ / 5) < 0) { + pr_err("Wait for timeline writes timed out!\n"); + err = -EIO; + goto out; + } + + if (*hwsp_seqno[0] != seqno[0] || *hwsp_seqno[1] != seqno[1]) { + pr_err("Bad timeline values: found (%x, %x), expected (%x, %x)\n", + *hwsp_seqno[0], *hwsp_seqno[1], + seqno[0], seqno[1]); + err = -EINVAL; + goto out; + } + + i915_retire_requests(i915); /* recycle HWSP */ + } + +out: + if (igt_flush_test(i915, I915_WAIT_LOCKED)) + err = -EIO; + + intel_timeline_unpin(tl); +out_free: + intel_timeline_put(tl); +out_rpm: + intel_runtime_pm_put(&i915->runtime_pm, wakeref); + mutex_unlock(&i915->drm.struct_mutex); + + return err; +} + +static int live_hwsp_recycle(void *arg) +{ + struct drm_i915_private *i915 = arg; + struct intel_engine_cs *engine; + enum intel_engine_id id; + intel_wakeref_t wakeref; + unsigned long count; + int err = 0; + + /* + * Check seqno writes into one timeline at a time. We expect to + * recycle the breadcrumb slot between iterations and neither + * want to confuse ourselves or the GPU. + */ + + mutex_lock(&i915->drm.struct_mutex); + wakeref = intel_runtime_pm_get(&i915->runtime_pm); + + count = 0; + for_each_engine(engine, i915, id) { + IGT_TIMEOUT(end_time); + + if (!intel_engine_can_store_dword(engine)) + continue; + + do { + struct intel_timeline *tl; + struct i915_request *rq; + + tl = checked_intel_timeline_create(i915); + if (IS_ERR(tl)) { + err = PTR_ERR(tl); + goto out; + } + + rq = tl_write(tl, engine, count); + if (IS_ERR(rq)) { + intel_timeline_put(tl); + err = PTR_ERR(rq); + goto out; + } + + if (i915_request_wait(rq, 0, HZ / 5) < 0) { + pr_err("Wait for timeline writes timed out!\n"); + intel_timeline_put(tl); + err = -EIO; + goto out; + } + + if (*tl->hwsp_seqno != count) { + pr_err("Invalid seqno stored in timeline %lu, found 0x%x\n", + count, *tl->hwsp_seqno); + err = -EINVAL; + } + + intel_timeline_put(tl); + count++; + + if (err) + goto out; + + intel_timelines_park(i915); /* Encourage recycling! */ + } while (!__igt_timeout(end_time, NULL)); + } + +out: + if (igt_flush_test(i915, I915_WAIT_LOCKED)) + err = -EIO; + intel_runtime_pm_put(&i915->runtime_pm, wakeref); + mutex_unlock(&i915->drm.struct_mutex); + + return err; +} + +int intel_timeline_live_selftests(struct drm_i915_private *i915) +{ + static const struct i915_subtest tests[] = { + SUBTEST(live_hwsp_recycle), + SUBTEST(live_hwsp_engine), + SUBTEST(live_hwsp_alternate), + SUBTEST(live_hwsp_wrap), + }; + + if (i915_terminally_wedged(i915)) + return 0; + + return i915_subtests(tests, i915); +} diff --git a/drivers/gpu/drm/i915/gt/selftests/mock_timeline.c b/drivers/gpu/drm/i915/gt/selftests/mock_timeline.c new file mode 100644 index 000000000000..5c549205828a --- /dev/null +++ b/drivers/gpu/drm/i915/gt/selftests/mock_timeline.c @@ -0,0 +1,29 @@ +/* + * SPDX-License-Identifier: MIT + * + * Copyright © 2017-2018 Intel Corporation + */ + +#include "../intel_timeline.h" + +#include "mock_timeline.h" + +void mock_timeline_init(struct intel_timeline *timeline, u64 context) +{ + timeline->gt = NULL; + timeline->fence_context = context; + + mutex_init(&timeline->mutex); + + INIT_ACTIVE_REQUEST(&timeline->last_request); + INIT_LIST_HEAD(&timeline->requests); + + i915_syncmap_init(&timeline->sync); + + INIT_LIST_HEAD(&timeline->link); +} + +void mock_timeline_fini(struct intel_timeline *timeline) +{ + i915_syncmap_free(&timeline->sync); +} diff --git a/drivers/gpu/drm/i915/gt/selftests/mock_timeline.h b/drivers/gpu/drm/i915/gt/selftests/mock_timeline.h new file mode 100644 index 000000000000..689efc66c908 --- /dev/null +++ b/drivers/gpu/drm/i915/gt/selftests/mock_timeline.h @@ -0,0 +1,15 @@ +/* + * SPDX-License-Identifier: MIT + * + * Copyright © 2017-2018 Intel Corporation + */ + +#ifndef __MOCK_TIMELINE__ +#define __MOCK_TIMELINE__ + +struct intel_timeline; + +void mock_timeline_init(struct intel_timeline *timeline, u64 context); +void mock_timeline_fini(struct intel_timeline *timeline); + +#endif /* !__MOCK_TIMELINE__ */ diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index b574aea23581..89a21fa4eac2 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -89,7 +89,7 @@ #include "i915_gpu_error.h" #include "i915_request.h" #include "i915_scheduler.h" -#include "i915_timeline.h" +#include "gt/intel_timeline.h" #include "i915_vma.h" #include "intel_gvt.h" diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index 50d7e1e8d8ad..6e07127242d9 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -909,7 +909,7 @@ wait_for_timelines(struct drm_i915_private *i915, unsigned int flags, long timeout) { struct i915_gt_timelines *gt = &i915->gt.timelines; - struct i915_timeline *tl; + struct intel_timeline *tl; mutex_lock(>->mutex); list_for_each_entry(tl, >->active_list, link) { @@ -1487,7 +1487,7 @@ int i915_gem_init(struct drm_i915_private *dev_priv) dev_priv->mm.unordered_timeline = dma_fence_context_alloc(1); - i915_timelines_init(dev_priv); + intel_timelines_init(dev_priv); ret = i915_gem_init_userptr(dev_priv); if (ret) @@ -1624,7 +1624,7 @@ err_uc_misc: if (ret != -EIO) { i915_gem_cleanup_userptr(dev_priv); - i915_timelines_fini(dev_priv); + intel_timelines_fini(dev_priv); } if (ret == -EIO) { @@ -1688,7 +1688,7 @@ void i915_gem_fini(struct drm_i915_private *dev_priv) intel_uc_fini_misc(dev_priv); i915_gem_cleanup_userptr(dev_priv); - i915_timelines_fini(dev_priv); + intel_timelines_fini(dev_priv); i915_gem_drain_freed_objects(dev_priv); diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.h b/drivers/gpu/drm/i915/i915_gem_gtt.h index e8b9ebe50c4e..028be3b44d07 100644 --- a/drivers/gpu/drm/i915/i915_gem_gtt.h +++ b/drivers/gpu/drm/i915/i915_gem_gtt.h @@ -47,7 +47,7 @@ #include "i915_request.h" #include "i915_scatterlist.h" #include "i915_selftest.h" -#include "i915_timeline.h" +#include "gt/intel_timeline.h" #define I915_GTT_PAGE_SIZE_4K BIT_ULL(12) #define I915_GTT_PAGE_SIZE_64K BIT_ULL(16) diff --git a/drivers/gpu/drm/i915/i915_request.c b/drivers/gpu/drm/i915/i915_request.c index 0c99694faab7..5ff87c4a0cd5 100644 --- a/drivers/gpu/drm/i915/i915_request.c +++ b/drivers/gpu/drm/i915/i915_request.c @@ -607,7 +607,7 @@ out: struct i915_request * __i915_request_create(struct intel_context *ce, gfp_t gfp) { - struct i915_timeline *tl = ce->ring->timeline; + struct intel_timeline *tl = ce->ring->timeline; struct i915_request *rq; u32 seqno; int ret; @@ -656,7 +656,7 @@ __i915_request_create(struct intel_context *ce, gfp_t gfp) } } - ret = i915_timeline_get_seqno(tl, rq, &seqno); + ret = intel_timeline_get_seqno(tl, rq, &seqno); if (ret) goto err_free; @@ -775,7 +775,7 @@ i915_request_await_start(struct i915_request *rq, struct i915_request *signal) return 0; signal = list_prev_entry(signal, ring_link); - if (i915_timeline_sync_is_later(rq->timeline, &signal->fence)) + if (intel_timeline_sync_is_later(rq->timeline, &signal->fence)) return 0; return i915_sw_fence_await_dma_fence(&rq->submit, @@ -829,7 +829,7 @@ emit_semaphore_wait(struct i915_request *to, return err; /* We need to pin the signaler's HWSP until we are finished reading. */ - err = i915_timeline_read_hwsp(from, to, &hwsp_offset); + err = intel_timeline_read_hwsp(from, to, &hwsp_offset); if (err) return err; @@ -940,7 +940,7 @@ i915_request_await_dma_fence(struct i915_request *rq, struct dma_fence *fence) /* Squash repeated waits to the same timelines */ if (fence->context != rq->i915->mm.unordered_timeline && - i915_timeline_sync_is_later(rq->timeline, fence)) + intel_timeline_sync_is_later(rq->timeline, fence)) continue; if (dma_fence_is_i915(fence)) @@ -954,7 +954,7 @@ i915_request_await_dma_fence(struct i915_request *rq, struct dma_fence *fence) /* Record the latest fence used against each timeline */ if (fence->context != rq->i915->mm.unordered_timeline) - i915_timeline_sync_set(rq->timeline, fence); + intel_timeline_sync_set(rq->timeline, fence); } while (--nchild); return 0; @@ -1092,7 +1092,7 @@ void i915_request_skip(struct i915_request *rq, int error) static struct i915_request * __i915_request_add_to_timeline(struct i915_request *rq) { - struct i915_timeline *timeline = rq->timeline; + struct intel_timeline *timeline = rq->timeline; struct i915_request *prev; /* diff --git a/drivers/gpu/drm/i915/i915_request.h b/drivers/gpu/drm/i915/i915_request.h index bebc1e9b4a5e..b58ceef92e20 100644 --- a/drivers/gpu/drm/i915/i915_request.h +++ b/drivers/gpu/drm/i915/i915_request.h @@ -41,8 +41,8 @@ struct drm_file; struct drm_i915_gem_object; struct i915_request; -struct i915_timeline; -struct i915_timeline_cacheline; +struct intel_timeline; +struct intel_timeline_cacheline; struct i915_capture_list { struct i915_capture_list *next; @@ -113,7 +113,7 @@ struct i915_request { struct intel_engine_cs *engine; struct intel_context *hw_context; struct intel_ring *ring; - struct i915_timeline *timeline; + struct intel_timeline *timeline; struct list_head signal_link; /* @@ -176,7 +176,7 @@ struct i915_request { * inside the timeline's HWSP vma, but it is only valid while this * request has not completed and guarded by the timeline mutex. */ - struct i915_timeline_cacheline *hwsp_cacheline; + struct intel_timeline_cacheline *hwsp_cacheline; /** Position in the ring of the start of the request */ u32 head; diff --git a/drivers/gpu/drm/i915/i915_timeline.c b/drivers/gpu/drm/i915/i915_timeline.c deleted file mode 100644 index 3e2c3169dc69..000000000000 --- a/drivers/gpu/drm/i915/i915_timeline.c +++ /dev/null @@ -1,591 +0,0 @@ -/* - * SPDX-License-Identifier: MIT - * - * Copyright © 2016-2018 Intel Corporation - */ - -#include "gt/intel_gt_types.h" - -#include "i915_drv.h" - -#include "i915_active.h" -#include "i915_syncmap.h" -#include "i915_timeline.h" - -#define ptr_set_bit(ptr, bit) ((typeof(ptr))((unsigned long)(ptr) | BIT(bit))) -#define ptr_test_bit(ptr, bit) ((unsigned long)(ptr) & BIT(bit)) - -struct i915_timeline_hwsp { - struct intel_gt *gt; - struct i915_gt_timelines *gt_timelines; - struct list_head free_link; - struct i915_vma *vma; - u64 free_bitmap; -}; - -struct i915_timeline_cacheline { - struct i915_active active; - struct i915_timeline_hwsp *hwsp; - void *vaddr; -#define CACHELINE_BITS 6 -#define CACHELINE_FREE CACHELINE_BITS -}; - -static struct i915_vma *__hwsp_alloc(struct intel_gt *gt) -{ - struct drm_i915_private *i915 = gt->i915; - struct drm_i915_gem_object *obj; - struct i915_vma *vma; - - obj = i915_gem_object_create_internal(i915, PAGE_SIZE); - if (IS_ERR(obj)) - return ERR_CAST(obj); - - i915_gem_object_set_cache_coherency(obj, I915_CACHE_LLC); - - vma = i915_vma_instance(obj, >->ggtt->vm, NULL); - if (IS_ERR(vma)) - i915_gem_object_put(obj); - - return vma; -} - -static struct i915_vma * -hwsp_alloc(struct i915_timeline *timeline, unsigned int *cacheline) -{ - struct i915_gt_timelines *gt = &timeline->gt->timelines; - struct i915_timeline_hwsp *hwsp; - - BUILD_BUG_ON(BITS_PER_TYPE(u64) * CACHELINE_BYTES > PAGE_SIZE); - - spin_lock_irq(>->hwsp_lock); - - /* hwsp_free_list only contains HWSP that have available cachelines */ - hwsp = list_first_entry_or_null(>->hwsp_free_list, - typeof(*hwsp), free_link); - if (!hwsp) { - struct i915_vma *vma; - - spin_unlock_irq(>->hwsp_lock); - - hwsp = kmalloc(sizeof(*hwsp), GFP_KERNEL); - if (!hwsp) - return ERR_PTR(-ENOMEM); - - vma = __hwsp_alloc(timeline->gt); - if (IS_ERR(vma)) { - kfree(hwsp); - return vma; - } - - vma->private = hwsp; - hwsp->gt = timeline->gt; - hwsp->vma = vma; - hwsp->free_bitmap = ~0ull; - hwsp->gt_timelines = gt; - - spin_lock_irq(>->hwsp_lock); - list_add(&hwsp->free_link, >->hwsp_free_list); - } - - GEM_BUG_ON(!hwsp->free_bitmap); - *cacheline = __ffs64(hwsp->free_bitmap); - hwsp->free_bitmap &= ~BIT_ULL(*cacheline); - if (!hwsp->free_bitmap) - list_del(&hwsp->free_link); - - spin_unlock_irq(>->hwsp_lock); - - GEM_BUG_ON(hwsp->vma->private != hwsp); - return hwsp->vma; -} - -static void __idle_hwsp_free(struct i915_timeline_hwsp *hwsp, int cacheline) -{ - struct i915_gt_timelines *gt = hwsp->gt_timelines; - unsigned long flags; - - spin_lock_irqsave(>->hwsp_lock, flags); - - /* As a cacheline becomes available, publish the HWSP on the freelist */ - if (!hwsp->free_bitmap) - list_add_tail(&hwsp->free_link, >->hwsp_free_list); - - GEM_BUG_ON(cacheline >= BITS_PER_TYPE(hwsp->free_bitmap)); - hwsp->free_bitmap |= BIT_ULL(cacheline); - - /* And if no one is left using it, give the page back to the system */ - if (hwsp->free_bitmap == ~0ull) { - i915_vma_put(hwsp->vma); - list_del(&hwsp->free_link); - kfree(hwsp); - } - - spin_unlock_irqrestore(>->hwsp_lock, flags); -} - -static void __idle_cacheline_free(struct i915_timeline_cacheline *cl) -{ - GEM_BUG_ON(!i915_active_is_idle(&cl->active)); - - i915_gem_object_unpin_map(cl->hwsp->vma->obj); - i915_vma_put(cl->hwsp->vma); - __idle_hwsp_free(cl->hwsp, ptr_unmask_bits(cl->vaddr, CACHELINE_BITS)); - - i915_active_fini(&cl->active); - kfree(cl); -} - -static void __cacheline_retire(struct i915_active *active) -{ - struct i915_timeline_cacheline *cl = - container_of(active, typeof(*cl), active); - - i915_vma_unpin(cl->hwsp->vma); - if (ptr_test_bit(cl->vaddr, CACHELINE_FREE)) - __idle_cacheline_free(cl); -} - -static struct i915_timeline_cacheline * -cacheline_alloc(struct i915_timeline_hwsp *hwsp, unsigned int cacheline) -{ - struct i915_timeline_cacheline *cl; - void *vaddr; - - GEM_BUG_ON(cacheline >= BIT(CACHELINE_BITS)); - - cl = kmalloc(sizeof(*cl), GFP_KERNEL); - if (!cl) - return ERR_PTR(-ENOMEM); - - vaddr = i915_gem_object_pin_map(hwsp->vma->obj, I915_MAP_WB); - if (IS_ERR(vaddr)) { - kfree(cl); - return ERR_CAST(vaddr); - } - - i915_vma_get(hwsp->vma); - cl->hwsp = hwsp; - cl->vaddr = page_pack_bits(vaddr, cacheline); - - i915_active_init(hwsp->gt->i915, &cl->active, __cacheline_retire); - - return cl; -} - -static void cacheline_acquire(struct i915_timeline_cacheline *cl) -{ - if (cl && i915_active_acquire(&cl->active)) - __i915_vma_pin(cl->hwsp->vma); -} - -static void cacheline_release(struct i915_timeline_cacheline *cl) -{ - if (cl) - i915_active_release(&cl->active); -} - -static void cacheline_free(struct i915_timeline_cacheline *cl) -{ - GEM_BUG_ON(ptr_test_bit(cl->vaddr, CACHELINE_FREE)); - cl->vaddr = ptr_set_bit(cl->vaddr, CACHELINE_FREE); - - if (i915_active_is_idle(&cl->active)) - __idle_cacheline_free(cl); -} - -int i915_timeline_init(struct i915_timeline *timeline, - struct intel_gt *gt, - struct i915_vma *hwsp) -{ - void *vaddr; - - /* - * Ideally we want a set of engines on a single leaf as we expect - * to mostly be tracking synchronisation between engines. It is not - * a huge issue if this is not the case, but we may want to mitigate - * any page crossing penalties if they become an issue. - * - * Called during early_init before we know how many engines there are. - */ - BUILD_BUG_ON(KSYNCMAP < I915_NUM_ENGINES); - - timeline->gt = gt; - timeline->pin_count = 0; - timeline->has_initial_breadcrumb = !hwsp; - timeline->hwsp_cacheline = NULL; - - if (!hwsp) { - struct i915_timeline_cacheline *cl; - unsigned int cacheline; - - hwsp = hwsp_alloc(timeline, &cacheline); - if (IS_ERR(hwsp)) - return PTR_ERR(hwsp); - - cl = cacheline_alloc(hwsp->private, cacheline); - if (IS_ERR(cl)) { - __idle_hwsp_free(hwsp->private, cacheline); - return PTR_ERR(cl); - } - - timeline->hwsp_cacheline = cl; - timeline->hwsp_offset = cacheline * CACHELINE_BYTES; - - vaddr = page_mask_bits(cl->vaddr); - } else { - timeline->hwsp_offset = I915_GEM_HWS_SEQNO_ADDR; - - vaddr = i915_gem_object_pin_map(hwsp->obj, I915_MAP_WB); - if (IS_ERR(vaddr)) - return PTR_ERR(vaddr); - } - - timeline->hwsp_seqno = - memset(vaddr + timeline->hwsp_offset, 0, CACHELINE_BYTES); - - timeline->hwsp_ggtt = i915_vma_get(hwsp); - GEM_BUG_ON(timeline->hwsp_offset >= hwsp->size); - - timeline->fence_context = dma_fence_context_alloc(1); - - mutex_init(&timeline->mutex); - - INIT_ACTIVE_REQUEST(&timeline->last_request); - INIT_LIST_HEAD(&timeline->requests); - - i915_syncmap_init(&timeline->sync); - - return 0; -} - -static void timelines_init(struct intel_gt *gt) -{ - struct i915_gt_timelines *timelines = >->timelines; - - mutex_init(&timelines->mutex); - INIT_LIST_HEAD(&timelines->active_list); - - spin_lock_init(&timelines->hwsp_lock); - INIT_LIST_HEAD(&timelines->hwsp_free_list); - - /* via i915_gem_wait_for_idle() */ - i915_gem_shrinker_taints_mutex(gt->i915, &timelines->mutex); -} - -void i915_timelines_init(struct drm_i915_private *i915) -{ - timelines_init(&i915->gt); -} - -static void timeline_add_to_active(struct i915_timeline *tl) -{ - struct i915_gt_timelines *gt = &tl->gt->timelines; - - mutex_lock(>->mutex); - list_add(&tl->link, >->active_list); - mutex_unlock(>->mutex); -} - -static void timeline_remove_from_active(struct i915_timeline *tl) -{ - struct i915_gt_timelines *gt = &tl->gt->timelines; - - mutex_lock(>->mutex); - list_del(&tl->link); - mutex_unlock(>->mutex); -} - -static void timelines_park(struct intel_gt *gt) -{ - struct i915_gt_timelines *timelines = >->timelines; - struct i915_timeline *timeline; - - mutex_lock(&timelines->mutex); - list_for_each_entry(timeline, &timelines->active_list, link) { - /* - * All known fences are completed so we can scrap - * the current sync point tracking and start afresh, - * any attempt to wait upon a previous sync point - * will be skipped as the fence was signaled. - */ - i915_syncmap_free(&timeline->sync); - } - mutex_unlock(&timelines->mutex); -} - -/** - * i915_timelines_park - called when the driver idles - * @i915: the drm_i915_private device - * - * When the driver is completely idle, we know that all of our sync points - * have been signaled and our tracking is then entirely redundant. Any request - * to wait upon an older sync point will be completed instantly as we know - * the fence is signaled and therefore we will not even look them up in the - * sync point map. - */ -void i915_timelines_park(struct drm_i915_private *i915) -{ - timelines_park(&i915->gt); -} - -void i915_timeline_fini(struct i915_timeline *timeline) -{ - GEM_BUG_ON(timeline->pin_count); - GEM_BUG_ON(!list_empty(&timeline->requests)); - - i915_syncmap_free(&timeline->sync); - - if (timeline->hwsp_cacheline) - cacheline_free(timeline->hwsp_cacheline); - else - i915_gem_object_unpin_map(timeline->hwsp_ggtt->obj); - - i915_vma_put(timeline->hwsp_ggtt); -} - -struct i915_timeline * -i915_timeline_create(struct intel_gt *gt, struct i915_vma *global_hwsp) -{ - struct i915_timeline *timeline; - int err; - - timeline = kzalloc(sizeof(*timeline), GFP_KERNEL); - if (!timeline) - return ERR_PTR(-ENOMEM); - - err = i915_timeline_init(timeline, gt, global_hwsp); - if (err) { - kfree(timeline); - return ERR_PTR(err); - } - - kref_init(&timeline->kref); - - return timeline; -} - -int i915_timeline_pin(struct i915_timeline *tl) -{ - int err; - - if (tl->pin_count++) - return 0; - GEM_BUG_ON(!tl->pin_count); - - err = i915_vma_pin(tl->hwsp_ggtt, 0, 0, PIN_GLOBAL | PIN_HIGH); - if (err) - goto unpin; - - tl->hwsp_offset = - i915_ggtt_offset(tl->hwsp_ggtt) + - offset_in_page(tl->hwsp_offset); - - cacheline_acquire(tl->hwsp_cacheline); - timeline_add_to_active(tl); - - return 0; - -unpin: - tl->pin_count = 0; - return err; -} - -static u32 timeline_advance(struct i915_timeline *tl) -{ - GEM_BUG_ON(!tl->pin_count); - GEM_BUG_ON(tl->seqno & tl->has_initial_breadcrumb); - - return tl->seqno += 1 + tl->has_initial_breadcrumb; -} - -static void timeline_rollback(struct i915_timeline *tl) -{ - tl->seqno -= 1 + tl->has_initial_breadcrumb; -} - -static noinline int -__i915_timeline_get_seqno(struct i915_timeline *tl, - struct i915_request *rq, - u32 *seqno) -{ - struct i915_timeline_cacheline *cl; - unsigned int cacheline; - struct i915_vma *vma; - void *vaddr; - int err; - - /* - * If there is an outstanding GPU reference to this cacheline, - * such as it being sampled by a HW semaphore on another timeline, - * we cannot wraparound our seqno value (the HW semaphore does - * a strict greater-than-or-equals compare, not i915_seqno_passed). - * So if the cacheline is still busy, we must detach ourselves - * from it and leave it inflight alongside its users. - * - * However, if nobody is watching and we can guarantee that nobody - * will, we could simply reuse the same cacheline. - * - * if (i915_active_request_is_signaled(&tl->last_request) && - * i915_active_is_signaled(&tl->hwsp_cacheline->active)) - * return 0; - * - * That seems unlikely for a busy timeline that needed to wrap in - * the first place, so just replace the cacheline. - */ - - vma = hwsp_alloc(tl, &cacheline); - if (IS_ERR(vma)) { - err = PTR_ERR(vma); - goto err_rollback; - } - - err = i915_vma_pin(vma, 0, 0, PIN_GLOBAL | PIN_HIGH); - if (err) { - __idle_hwsp_free(vma->private, cacheline); - goto err_rollback; - } - - cl = cacheline_alloc(vma->private, cacheline); - if (IS_ERR(cl)) { - err = PTR_ERR(cl); - __idle_hwsp_free(vma->private, cacheline); - goto err_unpin; - } - GEM_BUG_ON(cl->hwsp->vma != vma); - - /* - * Attach the old cacheline to the current request, so that we only - * free it after the current request is retired, which ensures that - * all writes into the cacheline from previous requests are complete. - */ - err = i915_active_ref(&tl->hwsp_cacheline->active, - tl->fence_context, rq); - if (err) - goto err_cacheline; - - cacheline_release(tl->hwsp_cacheline); /* ownership now xfered to rq */ - cacheline_free(tl->hwsp_cacheline); - - i915_vma_unpin(tl->hwsp_ggtt); /* binding kept alive by old cacheline */ - i915_vma_put(tl->hwsp_ggtt); - - tl->hwsp_ggtt = i915_vma_get(vma); - - vaddr = page_mask_bits(cl->vaddr); - tl->hwsp_offset = cacheline * CACHELINE_BYTES; - tl->hwsp_seqno = - memset(vaddr + tl->hwsp_offset, 0, CACHELINE_BYTES); - - tl->hwsp_offset += i915_ggtt_offset(vma); - - cacheline_acquire(cl); - tl->hwsp_cacheline = cl; - - *seqno = timeline_advance(tl); - GEM_BUG_ON(i915_seqno_passed(*tl->hwsp_seqno, *seqno)); - return 0; - -err_cacheline: - cacheline_free(cl); -err_unpin: - i915_vma_unpin(vma); -err_rollback: - timeline_rollback(tl); - return err; -} - -int i915_timeline_get_seqno(struct i915_timeline *tl, - struct i915_request *rq, - u32 *seqno) -{ - *seqno = timeline_advance(tl); - - /* Replace the HWSP on wraparound for HW semaphores */ - if (unlikely(!*seqno && tl->hwsp_cacheline)) - return __i915_timeline_get_seqno(tl, rq, seqno); - - return 0; -} - -static int cacheline_ref(struct i915_timeline_cacheline *cl, - struct i915_request *rq) -{ - return i915_active_ref(&cl->active, rq->fence.context, rq); -} - -int i915_timeline_read_hwsp(struct i915_request *from, - struct i915_request *to, - u32 *hwsp) -{ - struct i915_timeline_cacheline *cl = from->hwsp_cacheline; - struct i915_timeline *tl = from->timeline; - int err; - - GEM_BUG_ON(to->timeline == tl); - - mutex_lock_nested(&tl->mutex, SINGLE_DEPTH_NESTING); - err = i915_request_completed(from); - if (!err) - err = cacheline_ref(cl, to); - if (!err) { - if (likely(cl == tl->hwsp_cacheline)) { - *hwsp = tl->hwsp_offset; - } else { /* across a seqno wrap, recover the original offset */ - *hwsp = i915_ggtt_offset(cl->hwsp->vma) + - ptr_unmask_bits(cl->vaddr, CACHELINE_BITS) * - CACHELINE_BYTES; - } - } - mutex_unlock(&tl->mutex); - - return err; -} - -void i915_timeline_unpin(struct i915_timeline *tl) -{ - GEM_BUG_ON(!tl->pin_count); - if (--tl->pin_count) - return; - - timeline_remove_from_active(tl); - cacheline_release(tl->hwsp_cacheline); - - /* - * Since this timeline is idle, all bariers upon which we were waiting - * must also be complete and so we can discard the last used barriers - * without loss of information. - */ - i915_syncmap_free(&tl->sync); - - __i915_vma_unpin(tl->hwsp_ggtt); -} - -void __i915_timeline_free(struct kref *kref) -{ - struct i915_timeline *timeline = - container_of(kref, typeof(*timeline), kref); - - i915_timeline_fini(timeline); - kfree(timeline); -} - -static void timelines_fini(struct intel_gt *gt) -{ - struct i915_gt_timelines *timelines = >->timelines; - - GEM_BUG_ON(!list_empty(&timelines->active_list)); - GEM_BUG_ON(!list_empty(&timelines->hwsp_free_list)); - - mutex_destroy(&timelines->mutex); -} - -void i915_timelines_fini(struct drm_i915_private *i915) -{ - timelines_fini(&i915->gt); -} - -#if IS_ENABLED(CONFIG_DRM_I915_SELFTEST) -#include "selftests/mock_timeline.c" -#include "selftests/i915_timeline.c" -#endif diff --git a/drivers/gpu/drm/i915/i915_timeline.h b/drivers/gpu/drm/i915/i915_timeline.h deleted file mode 100644 index a454d49f229f..000000000000 --- a/drivers/gpu/drm/i915/i915_timeline.h +++ /dev/null @@ -1,93 +0,0 @@ -/* - * Copyright © 2016 Intel Corporation - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER - * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING - * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS - * IN THE SOFTWARE. - * - */ - -#ifndef I915_TIMELINE_H -#define I915_TIMELINE_H - -#include - -#include "i915_active.h" -#include "i915_syncmap.h" -#include "i915_timeline_types.h" - -int i915_timeline_init(struct i915_timeline *tl, - struct intel_gt *gt, - struct i915_vma *hwsp); -void i915_timeline_fini(struct i915_timeline *tl); - -struct i915_timeline * -i915_timeline_create(struct intel_gt *gt, struct i915_vma *global_hwsp); - -static inline struct i915_timeline * -i915_timeline_get(struct i915_timeline *timeline) -{ - kref_get(&timeline->kref); - return timeline; -} - -void __i915_timeline_free(struct kref *kref); -static inline void i915_timeline_put(struct i915_timeline *timeline) -{ - kref_put(&timeline->kref, __i915_timeline_free); -} - -static inline int __i915_timeline_sync_set(struct i915_timeline *tl, - u64 context, u32 seqno) -{ - return i915_syncmap_set(&tl->sync, context, seqno); -} - -static inline int i915_timeline_sync_set(struct i915_timeline *tl, - const struct dma_fence *fence) -{ - return __i915_timeline_sync_set(tl, fence->context, fence->seqno); -} - -static inline bool __i915_timeline_sync_is_later(struct i915_timeline *tl, - u64 context, u32 seqno) -{ - return i915_syncmap_is_later(&tl->sync, context, seqno); -} - -static inline bool i915_timeline_sync_is_later(struct i915_timeline *tl, - const struct dma_fence *fence) -{ - return __i915_timeline_sync_is_later(tl, fence->context, fence->seqno); -} - -int i915_timeline_pin(struct i915_timeline *tl); -int i915_timeline_get_seqno(struct i915_timeline *tl, - struct i915_request *rq, - u32 *seqno); -void i915_timeline_unpin(struct i915_timeline *tl); - -int i915_timeline_read_hwsp(struct i915_request *from, - struct i915_request *until, - u32 *hwsp_offset); - -void i915_timelines_init(struct drm_i915_private *i915); -void i915_timelines_park(struct drm_i915_private *i915); -void i915_timelines_fini(struct drm_i915_private *i915); - -#endif diff --git a/drivers/gpu/drm/i915/i915_timeline_types.h b/drivers/gpu/drm/i915/i915_timeline_types.h deleted file mode 100644 index 931585e12d41..000000000000 --- a/drivers/gpu/drm/i915/i915_timeline_types.h +++ /dev/null @@ -1,67 +0,0 @@ -/* - * SPDX-License-Identifier: MIT - * - * Copyright © 2016 Intel Corporation - */ - -#ifndef __I915_TIMELINE_TYPES_H__ -#define __I915_TIMELINE_TYPES_H__ - -#include -#include -#include -#include - -#include "i915_active_types.h" - -struct drm_i915_private; -struct i915_vma; -struct i915_timeline_cacheline; -struct i915_syncmap; - -struct i915_timeline { - u64 fence_context; - u32 seqno; - - struct mutex mutex; /* protects the flow of requests */ - - unsigned int pin_count; - const u32 *hwsp_seqno; - struct i915_vma *hwsp_ggtt; - u32 hwsp_offset; - - struct i915_timeline_cacheline *hwsp_cacheline; - - bool has_initial_breadcrumb; - - /** - * List of breadcrumbs associated with GPU requests currently - * outstanding. - */ - struct list_head requests; - - /* Contains an RCU guarded pointer to the last request. No reference is - * held to the request, users must carefully acquire a reference to - * the request using i915_active_request_get_request_rcu(), or hold the - * struct_mutex. - */ - struct i915_active_request last_request; - - /** - * We track the most recent seqno that we wait on in every context so - * that we only have to emit a new await and dependency on a more - * recent sync point. As the contexts may be executed out-of-order, we - * have to track each individually and can not rely on an absolute - * global_seqno. When we know that all tracked fences are completed - * (i.e. when the driver is idle), we know that the syncmap is - * redundant and we can discard it without loss of generality. - */ - struct i915_syncmap *sync; - - struct list_head link; - struct intel_gt *gt; - - struct kref kref; -}; - -#endif /* __I915_TIMELINE_TYPES_H__ */ diff --git a/drivers/gpu/drm/i915/selftests/i915_live_selftests.h b/drivers/gpu/drm/i915/selftests/i915_live_selftests.h index d5dc4427d664..2b31a4ee0b4c 100644 --- a/drivers/gpu/drm/i915/selftests/i915_live_selftests.h +++ b/drivers/gpu/drm/i915/selftests/i915_live_selftests.h @@ -12,7 +12,7 @@ selftest(sanitycheck, i915_live_sanitycheck) /* keep first (igt selfcheck) */ selftest(uncore, intel_uncore_live_selftests) selftest(workarounds, intel_workarounds_live_selftests) -selftest(timelines, i915_timeline_live_selftests) +selftest(timelines, intel_timeline_live_selftests) selftest(requests, i915_request_live_selftests) selftest(active, i915_active_live_selftests) selftest(objects, i915_gem_object_live_selftests) diff --git a/drivers/gpu/drm/i915/selftests/i915_mock_selftests.h b/drivers/gpu/drm/i915/selftests/i915_mock_selftests.h index 510eb176bb2c..b55da4d9ccba 100644 --- a/drivers/gpu/drm/i915/selftests/i915_mock_selftests.h +++ b/drivers/gpu/drm/i915/selftests/i915_mock_selftests.h @@ -15,7 +15,7 @@ selftest(scatterlist, scatterlist_mock_selftests) selftest(syncmap, i915_syncmap_mock_selftests) selftest(uncore, intel_uncore_mock_selftests) selftest(engine, intel_engine_cs_mock_selftests) -selftest(timelines, i915_timeline_mock_selftests) +selftest(timelines, intel_timeline_mock_selftests) selftest(requests, i915_request_mock_selftests) selftest(objects, i915_gem_object_mock_selftests) selftest(phys, i915_gem_phys_mock_selftests) diff --git a/drivers/gpu/drm/i915/selftests/i915_timeline.c b/drivers/gpu/drm/i915/selftests/i915_timeline.c deleted file mode 100644 index 44d031446f08..000000000000 --- a/drivers/gpu/drm/i915/selftests/i915_timeline.c +++ /dev/null @@ -1,845 +0,0 @@ -/* - * SPDX-License-Identifier: MIT - * - * Copyright © 2017-2018 Intel Corporation - */ - -#include - -#include "gem/i915_gem_pm.h" - -#include "i915_random.h" -#include "i915_selftest.h" - -#include "igt_flush_test.h" -#include "mock_gem_device.h" -#include "mock_timeline.h" - -static struct page *hwsp_page(struct i915_timeline *tl) -{ - struct drm_i915_gem_object *obj = tl->hwsp_ggtt->obj; - - GEM_BUG_ON(!i915_gem_object_has_pinned_pages(obj)); - return sg_page(obj->mm.pages->sgl); -} - -static unsigned long hwsp_cacheline(struct i915_timeline *tl) -{ - unsigned long address = (unsigned long)page_address(hwsp_page(tl)); - - return (address + tl->hwsp_offset) / CACHELINE_BYTES; -} - -#define CACHELINES_PER_PAGE (PAGE_SIZE / CACHELINE_BYTES) - -struct mock_hwsp_freelist { - struct drm_i915_private *i915; - struct radix_tree_root cachelines; - struct i915_timeline **history; - unsigned long count, max; - struct rnd_state prng; -}; - -enum { - SHUFFLE = BIT(0), -}; - -static void __mock_hwsp_record(struct mock_hwsp_freelist *state, - unsigned int idx, - struct i915_timeline *tl) -{ - tl = xchg(&state->history[idx], tl); - if (tl) { - radix_tree_delete(&state->cachelines, hwsp_cacheline(tl)); - i915_timeline_put(tl); - } -} - -static int __mock_hwsp_timeline(struct mock_hwsp_freelist *state, - unsigned int count, - unsigned int flags) -{ - struct i915_timeline *tl; - unsigned int idx; - - while (count--) { - unsigned long cacheline; - int err; - - tl = i915_timeline_create(&state->i915->gt, NULL); - if (IS_ERR(tl)) - return PTR_ERR(tl); - - cacheline = hwsp_cacheline(tl); - err = radix_tree_insert(&state->cachelines, cacheline, tl); - if (err) { - if (err == -EEXIST) { - pr_err("HWSP cacheline %lu already used; duplicate allocation!\n", - cacheline); - } - i915_timeline_put(tl); - return err; - } - - idx = state->count++ % state->max; - __mock_hwsp_record(state, idx, tl); - } - - if (flags & SHUFFLE) - i915_prandom_shuffle(state->history, - sizeof(*state->history), - min(state->count, state->max), - &state->prng); - - count = i915_prandom_u32_max_state(min(state->count, state->max), - &state->prng); - while (count--) { - idx = --state->count % state->max; - __mock_hwsp_record(state, idx, NULL); - } - - return 0; -} - -static int mock_hwsp_freelist(void *arg) -{ - struct mock_hwsp_freelist state; - const struct { - const char *name; - unsigned int flags; - } phases[] = { - { "linear", 0 }, - { "shuffled", SHUFFLE }, - { }, - }, *p; - unsigned int na; - int err = 0; - - INIT_RADIX_TREE(&state.cachelines, GFP_KERNEL); - state.prng = I915_RND_STATE_INITIALIZER(i915_selftest.random_seed); - - state.i915 = mock_gem_device(); - if (!state.i915) - return -ENOMEM; - - /* - * Create a bunch of timelines and check that their HWSP do not overlap. - * Free some, and try again. - */ - - state.max = PAGE_SIZE / sizeof(*state.history); - state.count = 0; - state.history = kcalloc(state.max, sizeof(*state.history), GFP_KERNEL); - if (!state.history) { - err = -ENOMEM; - goto err_put; - } - - mutex_lock(&state.i915->drm.struct_mutex); - for (p = phases; p->name; p++) { - pr_debug("%s(%s)\n", __func__, p->name); - for_each_prime_number_from(na, 1, 2 * CACHELINES_PER_PAGE) { - err = __mock_hwsp_timeline(&state, na, p->flags); - if (err) - goto out; - } - } - -out: - for (na = 0; na < state.max; na++) - __mock_hwsp_record(&state, na, NULL); - mutex_unlock(&state.i915->drm.struct_mutex); - kfree(state.history); -err_put: - drm_dev_put(&state.i915->drm); - return err; -} - -struct __igt_sync { - const char *name; - u32 seqno; - bool expected; - bool set; -}; - -static int __igt_sync(struct i915_timeline *tl, - u64 ctx, - const struct __igt_sync *p, - const char *name) -{ - int ret; - - if (__i915_timeline_sync_is_later(tl, ctx, p->seqno) != p->expected) { - pr_err("%s: %s(ctx=%llu, seqno=%u) expected passed %s but failed\n", - name, p->name, ctx, p->seqno, yesno(p->expected)); - return -EINVAL; - } - - if (p->set) { - ret = __i915_timeline_sync_set(tl, ctx, p->seqno); - if (ret) - return ret; - } - - return 0; -} - -static int igt_sync(void *arg) -{ - const struct __igt_sync pass[] = { - { "unset", 0, false, false }, - { "new", 0, false, true }, - { "0a", 0, true, true }, - { "1a", 1, false, true }, - { "1b", 1, true, true }, - { "0b", 0, true, false }, - { "2a", 2, false, true }, - { "4", 4, false, true }, - { "INT_MAX", INT_MAX, false, true }, - { "INT_MAX-1", INT_MAX-1, true, false }, - { "INT_MAX+1", (u32)INT_MAX+1, false, true }, - { "INT_MAX", INT_MAX, true, false }, - { "UINT_MAX", UINT_MAX, false, true }, - { "wrap", 0, false, true }, - { "unwrap", UINT_MAX, true, false }, - {}, - }, *p; - struct i915_timeline tl; - int order, offset; - int ret = -ENODEV; - - mock_timeline_init(&tl, 0); - for (p = pass; p->name; p++) { - for (order = 1; order < 64; order++) { - for (offset = -1; offset <= (order > 1); offset++) { - u64 ctx = BIT_ULL(order) + offset; - - ret = __igt_sync(&tl, ctx, p, "1"); - if (ret) - goto out; - } - } - } - mock_timeline_fini(&tl); - - mock_timeline_init(&tl, 0); - for (order = 1; order < 64; order++) { - for (offset = -1; offset <= (order > 1); offset++) { - u64 ctx = BIT_ULL(order) + offset; - - for (p = pass; p->name; p++) { - ret = __igt_sync(&tl, ctx, p, "2"); - if (ret) - goto out; - } - } - } - -out: - mock_timeline_fini(&tl); - return ret; -} - -static unsigned int random_engine(struct rnd_state *rnd) -{ - return i915_prandom_u32_max_state(I915_NUM_ENGINES, rnd); -} - -static int bench_sync(void *arg) -{ - struct rnd_state prng; - struct i915_timeline tl; - unsigned long end_time, count; - u64 prng32_1M; - ktime_t kt; - int order, last_order; - - mock_timeline_init(&tl, 0); - - /* Lookups from cache are very fast and so the random number generation - * and the loop itself becomes a significant factor in the per-iteration - * timings. We try to compensate the results by measuring the overhead - * of the prng and subtract it from the reported results. - */ - prandom_seed_state(&prng, i915_selftest.random_seed); - count = 0; - kt = ktime_get(); - end_time = jiffies + HZ/10; - do { - u32 x; - - /* Make sure the compiler doesn't optimise away the prng call */ - WRITE_ONCE(x, prandom_u32_state(&prng)); - - count++; - } while (!time_after(jiffies, end_time)); - kt = ktime_sub(ktime_get(), kt); - pr_debug("%s: %lu random evaluations, %lluns/prng\n", - __func__, count, (long long)div64_ul(ktime_to_ns(kt), count)); - prng32_1M = div64_ul(ktime_to_ns(kt) << 20, count); - - /* Benchmark (only) setting random context ids */ - prandom_seed_state(&prng, i915_selftest.random_seed); - count = 0; - kt = ktime_get(); - end_time = jiffies + HZ/10; - do { - u64 id = i915_prandom_u64_state(&prng); - - __i915_timeline_sync_set(&tl, id, 0); - count++; - } while (!time_after(jiffies, end_time)); - kt = ktime_sub(ktime_get(), kt); - kt = ktime_sub_ns(kt, (count * prng32_1M * 2) >> 20); - pr_info("%s: %lu random insertions, %lluns/insert\n", - __func__, count, (long long)div64_ul(ktime_to_ns(kt), count)); - - /* Benchmark looking up the exact same context ids as we just set */ - prandom_seed_state(&prng, i915_selftest.random_seed); - end_time = count; - kt = ktime_get(); - while (end_time--) { - u64 id = i915_prandom_u64_state(&prng); - - if (!__i915_timeline_sync_is_later(&tl, id, 0)) { - mock_timeline_fini(&tl); - pr_err("Lookup of %llu failed\n", id); - return -EINVAL; - } - } - kt = ktime_sub(ktime_get(), kt); - kt = ktime_sub_ns(kt, (count * prng32_1M * 2) >> 20); - pr_info("%s: %lu random lookups, %lluns/lookup\n", - __func__, count, (long long)div64_ul(ktime_to_ns(kt), count)); - - mock_timeline_fini(&tl); - cond_resched(); - - mock_timeline_init(&tl, 0); - - /* Benchmark setting the first N (in order) contexts */ - count = 0; - kt = ktime_get(); - end_time = jiffies + HZ/10; - do { - __i915_timeline_sync_set(&tl, count++, 0); - } while (!time_after(jiffies, end_time)); - kt = ktime_sub(ktime_get(), kt); - pr_info("%s: %lu in-order insertions, %lluns/insert\n", - __func__, count, (long long)div64_ul(ktime_to_ns(kt), count)); - - /* Benchmark looking up the exact same context ids as we just set */ - end_time = count; - kt = ktime_get(); - while (end_time--) { - if (!__i915_timeline_sync_is_later(&tl, end_time, 0)) { - pr_err("Lookup of %lu failed\n", end_time); - mock_timeline_fini(&tl); - return -EINVAL; - } - } - kt = ktime_sub(ktime_get(), kt); - pr_info("%s: %lu in-order lookups, %lluns/lookup\n", - __func__, count, (long long)div64_ul(ktime_to_ns(kt), count)); - - mock_timeline_fini(&tl); - cond_resched(); - - mock_timeline_init(&tl, 0); - - /* Benchmark searching for a random context id and maybe changing it */ - prandom_seed_state(&prng, i915_selftest.random_seed); - count = 0; - kt = ktime_get(); - end_time = jiffies + HZ/10; - do { - u32 id = random_engine(&prng); - u32 seqno = prandom_u32_state(&prng); - - if (!__i915_timeline_sync_is_later(&tl, id, seqno)) - __i915_timeline_sync_set(&tl, id, seqno); - - count++; - } while (!time_after(jiffies, end_time)); - kt = ktime_sub(ktime_get(), kt); - kt = ktime_sub_ns(kt, (count * prng32_1M * 2) >> 20); - pr_info("%s: %lu repeated insert/lookups, %lluns/op\n", - __func__, count, (long long)div64_ul(ktime_to_ns(kt), count)); - mock_timeline_fini(&tl); - cond_resched(); - - /* Benchmark searching for a known context id and changing the seqno */ - for (last_order = 1, order = 1; order < 32; - ({ int tmp = last_order; last_order = order; order += tmp; })) { - unsigned int mask = BIT(order) - 1; - - mock_timeline_init(&tl, 0); - - count = 0; - kt = ktime_get(); - end_time = jiffies + HZ/10; - do { - /* Without assuming too many details of the underlying - * implementation, try to identify its phase-changes - * (if any)! - */ - u64 id = (u64)(count & mask) << order; - - __i915_timeline_sync_is_later(&tl, id, 0); - __i915_timeline_sync_set(&tl, id, 0); - - count++; - } while (!time_after(jiffies, end_time)); - kt = ktime_sub(ktime_get(), kt); - pr_info("%s: %lu cyclic/%d insert/lookups, %lluns/op\n", - __func__, count, order, - (long long)div64_ul(ktime_to_ns(kt), count)); - mock_timeline_fini(&tl); - cond_resched(); - } - - return 0; -} - -int i915_timeline_mock_selftests(void) -{ - static const struct i915_subtest tests[] = { - SUBTEST(mock_hwsp_freelist), - SUBTEST(igt_sync), - SUBTEST(bench_sync), - }; - - return i915_subtests(tests, NULL); -} - -static int emit_ggtt_store_dw(struct i915_request *rq, u32 addr, u32 value) -{ - u32 *cs; - - cs = intel_ring_begin(rq, 4); - if (IS_ERR(cs)) - return PTR_ERR(cs); - - if (INTEL_GEN(rq->i915) >= 8) { - *cs++ = MI_STORE_DWORD_IMM_GEN4 | MI_USE_GGTT; - *cs++ = addr; - *cs++ = 0; - *cs++ = value; - } else if (INTEL_GEN(rq->i915) >= 4) { - *cs++ = MI_STORE_DWORD_IMM_GEN4 | MI_USE_GGTT; - *cs++ = 0; - *cs++ = addr; - *cs++ = value; - } else { - *cs++ = MI_STORE_DWORD_IMM | MI_MEM_VIRTUAL; - *cs++ = addr; - *cs++ = value; - *cs++ = MI_NOOP; - } - - intel_ring_advance(rq, cs); - - return 0; -} - -static struct i915_request * -tl_write(struct i915_timeline *tl, struct intel_engine_cs *engine, u32 value) -{ - struct i915_request *rq; - int err; - - lockdep_assert_held(&tl->gt->i915->drm.struct_mutex); /* lazy rq refs */ - - err = i915_timeline_pin(tl); - if (err) { - rq = ERR_PTR(err); - goto out; - } - - rq = i915_request_create(engine->kernel_context); - if (IS_ERR(rq)) - goto out_unpin; - - err = emit_ggtt_store_dw(rq, tl->hwsp_offset, value); - i915_request_add(rq); - if (err) - rq = ERR_PTR(err); - -out_unpin: - i915_timeline_unpin(tl); -out: - if (IS_ERR(rq)) - pr_err("Failed to write to timeline!\n"); - return rq; -} - -static struct i915_timeline * -checked_i915_timeline_create(struct drm_i915_private *i915) -{ - struct i915_timeline *tl; - - tl = i915_timeline_create(&i915->gt, NULL); - if (IS_ERR(tl)) - return tl; - - if (*tl->hwsp_seqno != tl->seqno) { - pr_err("Timeline created with incorrect breadcrumb, found %x, expected %x\n", - *tl->hwsp_seqno, tl->seqno); - i915_timeline_put(tl); - return ERR_PTR(-EINVAL); - } - - return tl; -} - -static int live_hwsp_engine(void *arg) -{ -#define NUM_TIMELINES 4096 - struct drm_i915_private *i915 = arg; - struct i915_timeline **timelines; - struct intel_engine_cs *engine; - enum intel_engine_id id; - intel_wakeref_t wakeref; - unsigned long count, n; - int err = 0; - - /* - * Create a bunch of timelines and check we can write - * independently to each of their breadcrumb slots. - */ - - timelines = kvmalloc_array(NUM_TIMELINES * I915_NUM_ENGINES, - sizeof(*timelines), - GFP_KERNEL); - if (!timelines) - return -ENOMEM; - - mutex_lock(&i915->drm.struct_mutex); - wakeref = intel_runtime_pm_get(&i915->runtime_pm); - - count = 0; - for_each_engine(engine, i915, id) { - if (!intel_engine_can_store_dword(engine)) - continue; - - for (n = 0; n < NUM_TIMELINES; n++) { - struct i915_timeline *tl; - struct i915_request *rq; - - tl = checked_i915_timeline_create(i915); - if (IS_ERR(tl)) { - err = PTR_ERR(tl); - goto out; - } - - rq = tl_write(tl, engine, count); - if (IS_ERR(rq)) { - i915_timeline_put(tl); - err = PTR_ERR(rq); - goto out; - } - - timelines[count++] = tl; - } - } - -out: - if (igt_flush_test(i915, I915_WAIT_LOCKED)) - err = -EIO; - - for (n = 0; n < count; n++) { - struct i915_timeline *tl = timelines[n]; - - if (!err && *tl->hwsp_seqno != n) { - pr_err("Invalid seqno stored in timeline %lu, found 0x%x\n", - n, *tl->hwsp_seqno); - err = -EINVAL; - } - i915_timeline_put(tl); - } - - intel_runtime_pm_put(&i915->runtime_pm, wakeref); - mutex_unlock(&i915->drm.struct_mutex); - - kvfree(timelines); - - return err; -#undef NUM_TIMELINES -} - -static int live_hwsp_alternate(void *arg) -{ -#define NUM_TIMELINES 4096 - struct drm_i915_private *i915 = arg; - struct i915_timeline **timelines; - struct intel_engine_cs *engine; - enum intel_engine_id id; - intel_wakeref_t wakeref; - unsigned long count, n; - int err = 0; - - /* - * Create a bunch of timelines and check we can write - * independently to each of their breadcrumb slots with adjacent - * engines. - */ - - timelines = kvmalloc_array(NUM_TIMELINES * I915_NUM_ENGINES, - sizeof(*timelines), - GFP_KERNEL); - if (!timelines) - return -ENOMEM; - - mutex_lock(&i915->drm.struct_mutex); - wakeref = intel_runtime_pm_get(&i915->runtime_pm); - - count = 0; - for (n = 0; n < NUM_TIMELINES; n++) { - for_each_engine(engine, i915, id) { - struct i915_timeline *tl; - struct i915_request *rq; - - if (!intel_engine_can_store_dword(engine)) - continue; - - tl = checked_i915_timeline_create(i915); - if (IS_ERR(tl)) { - err = PTR_ERR(tl); - goto out; - } - - rq = tl_write(tl, engine, count); - if (IS_ERR(rq)) { - i915_timeline_put(tl); - err = PTR_ERR(rq); - goto out; - } - - timelines[count++] = tl; - } - } - -out: - if (igt_flush_test(i915, I915_WAIT_LOCKED)) - err = -EIO; - - for (n = 0; n < count; n++) { - struct i915_timeline *tl = timelines[n]; - - if (!err && *tl->hwsp_seqno != n) { - pr_err("Invalid seqno stored in timeline %lu, found 0x%x\n", - n, *tl->hwsp_seqno); - err = -EINVAL; - } - i915_timeline_put(tl); - } - - intel_runtime_pm_put(&i915->runtime_pm, wakeref); - mutex_unlock(&i915->drm.struct_mutex); - - kvfree(timelines); - - return err; -#undef NUM_TIMELINES -} - -static int live_hwsp_wrap(void *arg) -{ - struct drm_i915_private *i915 = arg; - struct intel_engine_cs *engine; - struct i915_timeline *tl; - enum intel_engine_id id; - intel_wakeref_t wakeref; - int err = 0; - - /* - * Across a seqno wrap, we need to keep the old cacheline alive for - * foreign GPU references. - */ - - mutex_lock(&i915->drm.struct_mutex); - wakeref = intel_runtime_pm_get(&i915->runtime_pm); - - tl = i915_timeline_create(&i915->gt, NULL); - if (IS_ERR(tl)) { - err = PTR_ERR(tl); - goto out_rpm; - } - if (!tl->has_initial_breadcrumb || !tl->hwsp_cacheline) - goto out_free; - - err = i915_timeline_pin(tl); - if (err) - goto out_free; - - for_each_engine(engine, i915, id) { - const u32 *hwsp_seqno[2]; - struct i915_request *rq; - u32 seqno[2]; - - if (!intel_engine_can_store_dword(engine)) - continue; - - rq = i915_request_create(engine->kernel_context); - if (IS_ERR(rq)) { - err = PTR_ERR(rq); - goto out; - } - - tl->seqno = -4u; - - err = i915_timeline_get_seqno(tl, rq, &seqno[0]); - if (err) { - i915_request_add(rq); - goto out; - } - pr_debug("seqno[0]:%08x, hwsp_offset:%08x\n", - seqno[0], tl->hwsp_offset); - - err = emit_ggtt_store_dw(rq, tl->hwsp_offset, seqno[0]); - if (err) { - i915_request_add(rq); - goto out; - } - hwsp_seqno[0] = tl->hwsp_seqno; - - err = i915_timeline_get_seqno(tl, rq, &seqno[1]); - if (err) { - i915_request_add(rq); - goto out; - } - pr_debug("seqno[1]:%08x, hwsp_offset:%08x\n", - seqno[1], tl->hwsp_offset); - - err = emit_ggtt_store_dw(rq, tl->hwsp_offset, seqno[1]); - if (err) { - i915_request_add(rq); - goto out; - } - hwsp_seqno[1] = tl->hwsp_seqno; - - /* With wrap should come a new hwsp */ - GEM_BUG_ON(seqno[1] >= seqno[0]); - GEM_BUG_ON(hwsp_seqno[0] == hwsp_seqno[1]); - - i915_request_add(rq); - - if (i915_request_wait(rq, 0, HZ / 5) < 0) { - pr_err("Wait for timeline writes timed out!\n"); - err = -EIO; - goto out; - } - - if (*hwsp_seqno[0] != seqno[0] || *hwsp_seqno[1] != seqno[1]) { - pr_err("Bad timeline values: found (%x, %x), expected (%x, %x)\n", - *hwsp_seqno[0], *hwsp_seqno[1], - seqno[0], seqno[1]); - err = -EINVAL; - goto out; - } - - i915_retire_requests(i915); /* recycle HWSP */ - } - -out: - if (igt_flush_test(i915, I915_WAIT_LOCKED)) - err = -EIO; - - i915_timeline_unpin(tl); -out_free: - i915_timeline_put(tl); -out_rpm: - intel_runtime_pm_put(&i915->runtime_pm, wakeref); - mutex_unlock(&i915->drm.struct_mutex); - - return err; -} - -static int live_hwsp_recycle(void *arg) -{ - struct drm_i915_private *i915 = arg; - struct intel_engine_cs *engine; - enum intel_engine_id id; - intel_wakeref_t wakeref; - unsigned long count; - int err = 0; - - /* - * Check seqno writes into one timeline at a time. We expect to - * recycle the breadcrumb slot between iterations and neither - * want to confuse ourselves or the GPU. - */ - - mutex_lock(&i915->drm.struct_mutex); - wakeref = intel_runtime_pm_get(&i915->runtime_pm); - - count = 0; - for_each_engine(engine, i915, id) { - IGT_TIMEOUT(end_time); - - if (!intel_engine_can_store_dword(engine)) - continue; - - do { - struct i915_timeline *tl; - struct i915_request *rq; - - tl = checked_i915_timeline_create(i915); - if (IS_ERR(tl)) { - err = PTR_ERR(tl); - goto out; - } - - rq = tl_write(tl, engine, count); - if (IS_ERR(rq)) { - i915_timeline_put(tl); - err = PTR_ERR(rq); - goto out; - } - - if (i915_request_wait(rq, 0, HZ / 5) < 0) { - pr_err("Wait for timeline writes timed out!\n"); - i915_timeline_put(tl); - err = -EIO; - goto out; - } - - if (*tl->hwsp_seqno != count) { - pr_err("Invalid seqno stored in timeline %lu, found 0x%x\n", - count, *tl->hwsp_seqno); - err = -EINVAL; - } - - i915_timeline_put(tl); - count++; - - if (err) - goto out; - - i915_timelines_park(i915); /* Encourage recycling! */ - } while (!__igt_timeout(end_time, NULL)); - } - -out: - if (igt_flush_test(i915, I915_WAIT_LOCKED)) - err = -EIO; - intel_runtime_pm_put(&i915->runtime_pm, wakeref); - mutex_unlock(&i915->drm.struct_mutex); - - return err; -} - -int i915_timeline_live_selftests(struct drm_i915_private *i915) -{ - static const struct i915_subtest tests[] = { - SUBTEST(live_hwsp_recycle), - SUBTEST(live_hwsp_engine), - SUBTEST(live_hwsp_alternate), - SUBTEST(live_hwsp_wrap), - }; - - if (i915_terminally_wedged(i915)) - return 0; - - return i915_subtests(tests, i915); -} diff --git a/drivers/gpu/drm/i915/selftests/mock_gem_device.c b/drivers/gpu/drm/i915/selftests/mock_gem_device.c index 021ba42a3a00..2741805b56c2 100644 --- a/drivers/gpu/drm/i915/selftests/mock_gem_device.c +++ b/drivers/gpu/drm/i915/selftests/mock_gem_device.c @@ -68,7 +68,7 @@ static void mock_device_release(struct drm_device *dev) i915_gem_contexts_fini(i915); mutex_unlock(&i915->drm.struct_mutex); - i915_timelines_fini(i915); + intel_timelines_fini(i915); drain_workqueue(i915->wq); i915_gem_drain_freed_objects(i915); @@ -199,7 +199,7 @@ struct drm_i915_private *mock_gem_device(void) i915->gt.awake = true; - i915_timelines_init(i915); + intel_timelines_init(i915); mutex_lock(&i915->drm.struct_mutex); @@ -230,7 +230,7 @@ err_engine: mock_engine_free(i915->engine[RCS0]); err_unlock: mutex_unlock(&i915->drm.struct_mutex); - i915_timelines_fini(i915); + intel_timelines_fini(i915); destroy_workqueue(i915->wq); err_drv: drm_mode_config_cleanup(&i915->drm); diff --git a/drivers/gpu/drm/i915/selftests/mock_timeline.c b/drivers/gpu/drm/i915/selftests/mock_timeline.c deleted file mode 100644 index c80ac0fbdd3b..000000000000 --- a/drivers/gpu/drm/i915/selftests/mock_timeline.c +++ /dev/null @@ -1,29 +0,0 @@ -/* - * SPDX-License-Identifier: MIT - * - * Copyright © 2017-2018 Intel Corporation - */ - -#include "../i915_timeline.h" - -#include "mock_timeline.h" - -void mock_timeline_init(struct i915_timeline *timeline, u64 context) -{ - timeline->gt = NULL; - timeline->fence_context = context; - - mutex_init(&timeline->mutex); - - INIT_ACTIVE_REQUEST(&timeline->last_request); - INIT_LIST_HEAD(&timeline->requests); - - i915_syncmap_init(&timeline->sync); - - INIT_LIST_HEAD(&timeline->link); -} - -void mock_timeline_fini(struct i915_timeline *timeline) -{ - i915_syncmap_free(&timeline->sync); -} diff --git a/drivers/gpu/drm/i915/selftests/mock_timeline.h b/drivers/gpu/drm/i915/selftests/mock_timeline.h deleted file mode 100644 index b6deaa61110d..000000000000 --- a/drivers/gpu/drm/i915/selftests/mock_timeline.h +++ /dev/null @@ -1,15 +0,0 @@ -/* - * SPDX-License-Identifier: MIT - * - * Copyright © 2017-2018 Intel Corporation - */ - -#ifndef __MOCK_TIMELINE__ -#define __MOCK_TIMELINE__ - -struct i915_timeline; - -void mock_timeline_init(struct i915_timeline *timeline, u64 context); -void mock_timeline_fini(struct i915_timeline *timeline); - -#endif /* !__MOCK_TIMELINE__ */ -- cgit v1.2.3 From 21de5a9e34fb0f3efa31df5712cedc0ae97f7b4d Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Wed, 3 Jul 2019 16:52:25 +0100 Subject: drm/i915/gt: Ignore forcewake acquisition for posting_reads We don't care about the result of the read, so it may be garbage, we only care that the mmio is flushed. As such, we can forgo using an individual forcewake and lock around any posting-read for an engine. Signed-off-by: Chris Wilson Reviewed-by: Tvrtko Ursulin Link: https://patchwork.freedesktop.org/patch/msgid/20190703155225.9501-4-chris@chris-wilson.co.uk --- drivers/gpu/drm/i915/gt/intel_engine.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/gpu/drm/i915/gt/intel_engine.h') diff --git a/drivers/gpu/drm/i915/gt/intel_engine.h b/drivers/gpu/drm/i915/gt/intel_engine.h index 557b08b13feb..0331e9ac2485 100644 --- a/drivers/gpu/drm/i915/gt/intel_engine.h +++ b/drivers/gpu/drm/i915/gt/intel_engine.h @@ -51,7 +51,7 @@ struct drm_printer; #define ENGINE_READ16(...) __ENGINE_READ_OP(read16, __VA_ARGS__) #define ENGINE_READ(...) __ENGINE_READ_OP(read, __VA_ARGS__) #define ENGINE_READ_FW(...) __ENGINE_READ_OP(read_fw, __VA_ARGS__) -#define ENGINE_POSTING_READ(...) __ENGINE_READ_OP(posting_read, __VA_ARGS__) +#define ENGINE_POSTING_READ(...) __ENGINE_READ_OP(posting_read_fw, __VA_ARGS__) #define ENGINE_POSTING_READ16(...) __ENGINE_READ_OP(posting_read16, __VA_ARGS__) #define ENGINE_READ64(engine__, lower_reg__, upper_reg__) \ -- cgit v1.2.3 From cb823ed9915b0d4064f3f51e936fbe13c089948a Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Fri, 12 Jul 2019 20:29:53 +0100 Subject: drm/i915/gt: Use intel_gt as the primary object for handling resets Having taken the first step in encapsulating the functionality by moving the related files under gt/, the next step is to start encapsulating by passing around the relevant structs rather than the global drm_i915_private. In this step, we pass intel_gt to intel_reset.c Signed-off-by: Chris Wilson Cc: Tvrtko Ursulin Cc: Daniele Ceraolo Spurio Reviewed-by: Daniele Ceraolo Spurio Link: https://patchwork.freedesktop.org/patch/msgid/20190712192953.9187-1-chris@chris-wilson.co.uk --- drivers/gpu/drm/i915/display/intel_display.c | 22 +- drivers/gpu/drm/i915/gem/i915_gem_context.c | 2 +- drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c | 2 +- drivers/gpu/drm/i915/gem/i915_gem_mman.c | 8 +- drivers/gpu/drm/i915/gem/i915_gem_pm.c | 25 +- drivers/gpu/drm/i915/gem/i915_gem_throttle.c | 2 +- drivers/gpu/drm/i915/gem/selftests/huge_pages.c | 20 +- .../drm/i915/gem/selftests/i915_gem_client_blt.c | 4 +- .../drm/i915/gem/selftests/i915_gem_coherency.c | 6 +- .../gpu/drm/i915/gem/selftests/i915_gem_context.c | 17 +- drivers/gpu/drm/i915/gem/selftests/i915_gem_mman.c | 2 +- .../drm/i915/gem/selftests/i915_gem_object_blt.c | 4 +- drivers/gpu/drm/i915/gt/intel_engine.h | 8 +- drivers/gpu/drm/i915/gt/intel_engine_cs.c | 16 +- drivers/gpu/drm/i915/gt/intel_engine_pm.c | 3 +- drivers/gpu/drm/i915/gt/intel_gt.c | 7 + drivers/gpu/drm/i915/gt/intel_gt.h | 12 + drivers/gpu/drm/i915/gt/intel_gt_pm.c | 22 +- drivers/gpu/drm/i915/gt/intel_gt_types.h | 12 + drivers/gpu/drm/i915/gt/intel_hangcheck.c | 68 +-- drivers/gpu/drm/i915/gt/intel_lrc.c | 2 +- drivers/gpu/drm/i915/gt/intel_reset.c | 440 ++++++++++---------- drivers/gpu/drm/i915/gt/intel_reset.h | 73 ++-- drivers/gpu/drm/i915/gt/intel_reset_types.h | 50 +++ drivers/gpu/drm/i915/gt/intel_ringbuffer.c | 2 +- drivers/gpu/drm/i915/gt/selftest_hangcheck.c | 461 +++++++++++---------- drivers/gpu/drm/i915/gt/selftest_lrc.c | 38 +- drivers/gpu/drm/i915/gt/selftest_reset.c | 97 +++-- drivers/gpu/drm/i915/gt/selftest_timeline.c | 3 +- drivers/gpu/drm/i915/gt/selftest_workarounds.c | 33 +- drivers/gpu/drm/i915/i915_debugfs.c | 63 ++- drivers/gpu/drm/i915/i915_drv.c | 5 +- drivers/gpu/drm/i915/i915_drv.h | 35 +- drivers/gpu/drm/i915/i915_gem.c | 31 +- drivers/gpu/drm/i915/i915_gpu_error.h | 52 +-- drivers/gpu/drm/i915/i915_request.c | 5 +- drivers/gpu/drm/i915/i915_selftest.h | 9 + drivers/gpu/drm/i915/intel_guc_submission.c | 2 +- drivers/gpu/drm/i915/intel_uc.c | 2 +- drivers/gpu/drm/i915/selftests/i915_active.c | 3 +- drivers/gpu/drm/i915/selftests/i915_gem.c | 3 +- drivers/gpu/drm/i915/selftests/i915_gem_evict.c | 3 +- drivers/gpu/drm/i915/selftests/i915_request.c | 4 +- drivers/gpu/drm/i915/selftests/i915_selftest.c | 23 +- drivers/gpu/drm/i915/selftests/igt_flush_test.c | 5 +- drivers/gpu/drm/i915/selftests/igt_reset.c | 38 +- drivers/gpu/drm/i915/selftests/igt_reset.h | 10 +- drivers/gpu/drm/i915/selftests/igt_wedge_me.h | 58 --- drivers/gpu/drm/i915/selftests/mock_gem_device.c | 5 - 49 files changed, 903 insertions(+), 914 deletions(-) create mode 100644 drivers/gpu/drm/i915/gt/intel_reset_types.h delete mode 100644 drivers/gpu/drm/i915/selftests/igt_wedge_me.h (limited to 'drivers/gpu/drm/i915/gt/intel_engine.h') diff --git a/drivers/gpu/drm/i915/display/intel_display.c b/drivers/gpu/drm/i915/display/intel_display.c index 9758523f0837..e25b82d07d4f 100644 --- a/drivers/gpu/drm/i915/display/intel_display.c +++ b/drivers/gpu/drm/i915/display/intel_display.c @@ -4271,12 +4271,13 @@ void intel_prepare_reset(struct drm_i915_private *dev_priv) return; /* We have a modeset vs reset deadlock, defensively unbreak it. */ - set_bit(I915_RESET_MODESET, &dev_priv->gpu_error.flags); - wake_up_all(&dev_priv->gpu_error.wait_queue); + set_bit(I915_RESET_MODESET, &dev_priv->gt.reset.flags); + smp_mb__after_atomic(); + wake_up_bit(&dev_priv->gt.reset.flags, I915_RESET_MODESET); if (atomic_read(&dev_priv->gpu_error.pending_fb_pin)) { DRM_DEBUG_KMS("Modeset potentially stuck, unbreaking through wedging\n"); - i915_gem_set_wedged(dev_priv); + intel_gt_set_wedged(&dev_priv->gt); } /* @@ -4322,7 +4323,7 @@ void intel_finish_reset(struct drm_i915_private *dev_priv) int ret; /* reset doesn't touch the display */ - if (!test_bit(I915_RESET_MODESET, &dev_priv->gpu_error.flags)) + if (!test_bit(I915_RESET_MODESET, &dev_priv->gt.reset.flags)) return; state = fetch_and_zero(&dev_priv->modeset_restore_state); @@ -4362,7 +4363,7 @@ unlock: drm_modeset_acquire_fini(ctx); mutex_unlock(&dev->mode_config.mutex); - clear_bit(I915_RESET_MODESET, &dev_priv->gpu_error.flags); + clear_bit_unlock(I915_RESET_MODESET, &dev_priv->gt.reset.flags); } static void icl_set_pipe_chicken(struct intel_crtc *crtc) @@ -13873,18 +13874,21 @@ static void intel_atomic_commit_fence_wait(struct intel_atomic_state *intel_stat for (;;) { prepare_to_wait(&intel_state->commit_ready.wait, &wait_fence, TASK_UNINTERRUPTIBLE); - prepare_to_wait(&dev_priv->gpu_error.wait_queue, + prepare_to_wait(bit_waitqueue(&dev_priv->gt.reset.flags, + I915_RESET_MODESET), &wait_reset, TASK_UNINTERRUPTIBLE); - if (i915_sw_fence_done(&intel_state->commit_ready) - || test_bit(I915_RESET_MODESET, &dev_priv->gpu_error.flags)) + if (i915_sw_fence_done(&intel_state->commit_ready) || + test_bit(I915_RESET_MODESET, &dev_priv->gt.reset.flags)) break; schedule(); } finish_wait(&intel_state->commit_ready.wait, &wait_fence); - finish_wait(&dev_priv->gpu_error.wait_queue, &wait_reset); + finish_wait(bit_waitqueue(&dev_priv->gt.reset.flags, + I915_RESET_MODESET), + &wait_reset); } static void intel_atomic_cleanup_work(struct work_struct *work) diff --git a/drivers/gpu/drm/i915/gem/i915_gem_context.c b/drivers/gpu/drm/i915/gem/i915_gem_context.c index 078592912d97..c5f8bfa3f7b0 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_context.c +++ b/drivers/gpu/drm/i915/gem/i915_gem_context.c @@ -2127,7 +2127,7 @@ int i915_gem_context_create_ioctl(struct drm_device *dev, void *data, if (args->flags & I915_CONTEXT_CREATE_FLAGS_UNKNOWN) return -EINVAL; - ret = i915_terminally_wedged(i915); + ret = intel_gt_terminally_wedged(&i915->gt); if (ret) return ret; diff --git a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c index 1c5dfbfad71b..8a2047c4e7c3 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c +++ b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c @@ -2130,7 +2130,7 @@ static int eb_pin_context(struct i915_execbuffer *eb, struct intel_context *ce) * ABI: Before userspace accesses the GPU (e.g. execbuffer), report * EIO if the GPU is already wedged. */ - err = i915_terminally_wedged(eb->i915); + err = intel_gt_terminally_wedged(ce->engine->gt); if (err) return err; diff --git a/drivers/gpu/drm/i915/gem/i915_gem_mman.c b/drivers/gpu/drm/i915/gem/i915_gem_mman.c index 391621ee3cbb..a564c1e4231b 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_mman.c +++ b/drivers/gpu/drm/i915/gem/i915_gem_mman.c @@ -7,6 +7,8 @@ #include #include +#include "gt/intel_gt.h" + #include "i915_drv.h" #include "i915_gem_gtt.h" #include "i915_gem_ioctls.h" @@ -246,7 +248,7 @@ vm_fault_t i915_gem_fault(struct vm_fault *vmf) wakeref = intel_runtime_pm_get(rpm); - srcu = i915_reset_trylock(i915); + srcu = intel_gt_reset_trylock(ggtt->vm.gt); if (srcu < 0) { ret = srcu; goto err_rpm; @@ -326,7 +328,7 @@ err_unpin: err_unlock: mutex_unlock(&dev->struct_mutex); err_reset: - i915_reset_unlock(i915, srcu); + intel_gt_reset_unlock(ggtt->vm.gt, srcu); err_rpm: intel_runtime_pm_put(rpm, wakeref); i915_gem_object_unpin_pages(obj); @@ -339,7 +341,7 @@ err: * fail). But any other -EIO isn't ours (e.g. swap in failure) * and so needs to be reported. */ - if (!i915_terminally_wedged(i915)) + if (!intel_gt_is_wedged(ggtt->vm.gt)) return VM_FAULT_SIGBUS; /* else: fall through */ case -EAGAIN: diff --git a/drivers/gpu/drm/i915/gem/i915_gem_pm.c b/drivers/gpu/drm/i915/gem/i915_gem_pm.c index 4d774376f5b8..9ee6edbad4c5 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_pm.c +++ b/drivers/gpu/drm/i915/gem/i915_gem_pm.c @@ -5,6 +5,7 @@ */ #include "gem/i915_gem_pm.h" +#include "gt/intel_gt.h" #include "gt/intel_gt_pm.h" #include "i915_drv.h" @@ -106,18 +107,18 @@ static int pm_notifier(struct notifier_block *nb, return NOTIFY_OK; } -static bool switch_to_kernel_context_sync(struct drm_i915_private *i915) +static bool switch_to_kernel_context_sync(struct intel_gt *gt) { - bool result = !i915_terminally_wedged(i915); + bool result = !intel_gt_is_wedged(gt); do { - if (i915_gem_wait_for_idle(i915, + if (i915_gem_wait_for_idle(gt->i915, I915_WAIT_LOCKED | I915_WAIT_FOR_IDLE_BOOST, I915_GEM_IDLE_TIMEOUT) == -ETIME) { /* XXX hide warning from gem_eio */ if (i915_modparams.reset) { - dev_err(i915->drm.dev, + dev_err(gt->i915->drm.dev, "Failed to idle engines, declaring wedged!\n"); GEM_TRACE_DUMP(); } @@ -126,18 +127,18 @@ static bool switch_to_kernel_context_sync(struct drm_i915_private *i915) * Forcibly cancel outstanding work and leave * the gpu quiet. */ - i915_gem_set_wedged(i915); + intel_gt_set_wedged(gt); result = false; } - } while (i915_retire_requests(i915) && result); + } while (i915_retire_requests(gt->i915) && result); - GEM_BUG_ON(i915->gt.awake); + GEM_BUG_ON(gt->awake); return result; } bool i915_gem_load_power_context(struct drm_i915_private *i915) { - return switch_to_kernel_context_sync(i915); + return switch_to_kernel_context_sync(&i915->gt); } void i915_gem_suspend(struct drm_i915_private *i915) @@ -158,7 +159,7 @@ void i915_gem_suspend(struct drm_i915_private *i915) * state. Fortunately, the kernel_context is disposable and we do * not rely on its state. */ - switch_to_kernel_context_sync(i915); + switch_to_kernel_context_sync(&i915->gt); mutex_unlock(&i915->drm.struct_mutex); @@ -169,7 +170,7 @@ void i915_gem_suspend(struct drm_i915_private *i915) GEM_BUG_ON(i915->gt.awake); flush_work(&i915->gem.idle_work); - cancel_delayed_work_sync(&i915->gpu_error.hangcheck_work); + cancel_delayed_work_sync(&i915->gt.hangcheck.work); i915_gem_drain_freed_objects(i915); @@ -277,10 +278,10 @@ out_unlock: return; err_wedged: - if (!i915_reset_failed(i915)) { + if (!intel_gt_is_wedged(&i915->gt)) { dev_err(i915->drm.dev, "Failed to re-initialize GPU, declaring it wedged!\n"); - i915_gem_set_wedged(i915); + intel_gt_set_wedged(&i915->gt); } goto out_unlock; } diff --git a/drivers/gpu/drm/i915/gem/i915_gem_throttle.c b/drivers/gpu/drm/i915/gem/i915_gem_throttle.c index adb3074d9ce2..1e372420771b 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_throttle.c +++ b/drivers/gpu/drm/i915/gem/i915_gem_throttle.c @@ -41,7 +41,7 @@ i915_gem_throttle_ioctl(struct drm_device *dev, void *data, long ret; /* ABI: return -EIO if already wedged */ - ret = i915_terminally_wedged(to_i915(dev)); + ret = intel_gt_terminally_wedged(&to_i915(dev)->gt); if (ret) return ret; diff --git a/drivers/gpu/drm/i915/gem/selftests/huge_pages.c b/drivers/gpu/drm/i915/gem/selftests/huge_pages.c index 86eed4c3ae2b..6cbd4a668c9a 100644 --- a/drivers/gpu/drm/i915/gem/selftests/huge_pages.c +++ b/drivers/gpu/drm/i915/gem/selftests/huge_pages.c @@ -1753,7 +1753,7 @@ out_unlock: return err; } -int i915_gem_huge_page_live_selftests(struct drm_i915_private *dev_priv) +int i915_gem_huge_page_live_selftests(struct drm_i915_private *i915) { static const struct i915_subtest tests[] = { SUBTEST(igt_shrink_thp), @@ -1768,22 +1768,22 @@ int i915_gem_huge_page_live_selftests(struct drm_i915_private *dev_priv) intel_wakeref_t wakeref; int err; - if (!HAS_PPGTT(dev_priv)) { + if (!HAS_PPGTT(i915)) { pr_info("PPGTT not supported, skipping live-selftests\n"); return 0; } - if (i915_terminally_wedged(dev_priv)) + if (intel_gt_is_wedged(&i915->gt)) return 0; - file = mock_file(dev_priv); + file = mock_file(i915); if (IS_ERR(file)) return PTR_ERR(file); - mutex_lock(&dev_priv->drm.struct_mutex); - wakeref = intel_runtime_pm_get(&dev_priv->runtime_pm); + mutex_lock(&i915->drm.struct_mutex); + wakeref = intel_runtime_pm_get(&i915->runtime_pm); - ctx = live_context(dev_priv, file); + ctx = live_context(i915, file); if (IS_ERR(ctx)) { err = PTR_ERR(ctx); goto out_unlock; @@ -1795,10 +1795,10 @@ int i915_gem_huge_page_live_selftests(struct drm_i915_private *dev_priv) err = i915_subtests(tests, ctx); out_unlock: - intel_runtime_pm_put(&dev_priv->runtime_pm, wakeref); - mutex_unlock(&dev_priv->drm.struct_mutex); + intel_runtime_pm_put(&i915->runtime_pm, wakeref); + mutex_unlock(&i915->drm.struct_mutex); - mock_file_free(dev_priv, file); + mock_file_free(i915, file); return err; } diff --git a/drivers/gpu/drm/i915/gem/selftests/i915_gem_client_blt.c b/drivers/gpu/drm/i915/gem/selftests/i915_gem_client_blt.c index fa79233093eb..275c28926067 100644 --- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_client_blt.c +++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_client_blt.c @@ -5,6 +5,8 @@ #include "i915_selftest.h" +#include "gt/intel_gt.h" + #include "selftests/igt_flush_test.h" #include "selftests/mock_drm.h" #include "mock_context.h" @@ -101,7 +103,7 @@ int i915_gem_client_blt_live_selftests(struct drm_i915_private *i915) SUBTEST(igt_client_fill), }; - if (i915_terminally_wedged(i915)) + if (intel_gt_is_wedged(&i915->gt)) return 0; if (!HAS_ENGINE(i915, BCS0)) diff --git a/drivers/gpu/drm/i915/gem/selftests/i915_gem_coherency.c b/drivers/gpu/drm/i915/gem/selftests/i915_gem_coherency.c index 861f32be7d46..a1a4b53cdc4a 100644 --- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_coherency.c +++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_coherency.c @@ -6,6 +6,8 @@ #include +#include "gt/intel_gt.h" + #include "i915_selftest.h" #include "selftests/i915_random.h" @@ -242,12 +244,12 @@ static bool always_valid(struct drm_i915_private *i915) static bool needs_fence_registers(struct drm_i915_private *i915) { - return !i915_terminally_wedged(i915); + return !intel_gt_is_wedged(&i915->gt); } static bool needs_mi_store_dword(struct drm_i915_private *i915) { - if (i915_terminally_wedged(i915)) + if (intel_gt_is_wedged(&i915->gt)) return false; if (!HAS_ENGINE(i915, RCS0)) diff --git a/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c b/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c index 695bfb18b0d4..db7856f0f31e 100644 --- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c +++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c @@ -7,6 +7,7 @@ #include #include "gem/i915_gem_pm.h" +#include "gt/intel_gt.h" #include "gt/intel_reset.h" #include "i915_selftest.h" @@ -83,7 +84,7 @@ static int live_nop_switch(void *arg) } if (i915_request_wait(rq, 0, HZ / 5) < 0) { pr_err("Failed to populated %d contexts\n", nctx); - i915_gem_set_wedged(i915); + intel_gt_set_wedged(&i915->gt); err = -EIO; goto out_unlock; } @@ -127,7 +128,7 @@ static int live_nop_switch(void *arg) if (i915_request_wait(rq, 0, HZ / 5) < 0) { pr_err("Switching between %ld contexts timed out\n", prime); - i915_gem_set_wedged(i915); + intel_gt_set_wedged(&i915->gt); break; } @@ -956,7 +957,7 @@ __sseu_finish(struct drm_i915_private *i915, int ret = 0; if (flags & TEST_RESET) { - ret = i915_reset_engine(ce->engine, "sseu"); + ret = intel_engine_reset(ce->engine, "sseu"); if (ret) goto out; } @@ -1059,7 +1060,7 @@ __igt_ctx_sseu(struct drm_i915_private *i915, return PTR_ERR(file); if (flags & TEST_RESET) - igt_global_reset_lock(i915); + igt_global_reset_lock(&i915->gt); mutex_lock(&i915->drm.struct_mutex); @@ -1120,7 +1121,7 @@ out_unlock: mutex_unlock(&i915->drm.struct_mutex); if (flags & TEST_RESET) - igt_global_reset_unlock(i915); + igt_global_reset_unlock(&i915->gt); mock_file_free(i915, file); @@ -1722,7 +1723,7 @@ int i915_gem_context_mock_selftests(void) return err; } -int i915_gem_context_live_selftests(struct drm_i915_private *dev_priv) +int i915_gem_context_live_selftests(struct drm_i915_private *i915) { static const struct i915_subtest tests[] = { SUBTEST(live_nop_switch), @@ -1733,8 +1734,8 @@ int i915_gem_context_live_selftests(struct drm_i915_private *dev_priv) SUBTEST(igt_vm_isolation), }; - if (i915_terminally_wedged(dev_priv)) + if (intel_gt_is_wedged(&i915->gt)) return 0; - return i915_live_subtests(tests, dev_priv); + return i915_live_subtests(tests, i915); } diff --git a/drivers/gpu/drm/i915/gem/selftests/i915_gem_mman.c b/drivers/gpu/drm/i915/gem/selftests/i915_gem_mman.c index 5635cbb4af22..01857c12f12f 100644 --- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_mman.c +++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_mman.c @@ -478,7 +478,7 @@ static int igt_mmap_offset_exhaustion(void *arg) /* Now fill with busy dead objects that we expect to reap */ for (loop = 0; loop < 3; loop++) { - if (i915_terminally_wedged(i915)) + if (intel_gt_is_wedged(&i915->gt)) break; obj = i915_gem_object_create_internal(i915, PAGE_SIZE); diff --git a/drivers/gpu/drm/i915/gem/selftests/i915_gem_object_blt.c b/drivers/gpu/drm/i915/gem/selftests/i915_gem_object_blt.c index 11d37238c62c..19843acc84d3 100644 --- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_object_blt.c +++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_object_blt.c @@ -3,6 +3,8 @@ * Copyright © 2019 Intel Corporation */ +#include "gt/intel_gt.h" + #include "i915_selftest.h" #include "selftests/igt_flush_test.h" @@ -95,7 +97,7 @@ int i915_gem_object_blt_live_selftests(struct drm_i915_private *i915) SUBTEST(igt_fill_blt), }; - if (i915_terminally_wedged(i915)) + if (intel_gt_is_wedged(&i915->gt)) return 0; if (!HAS_ENGINE(i915, BCS0)) diff --git a/drivers/gpu/drm/i915/gt/intel_engine.h b/drivers/gpu/drm/i915/gt/intel_engine.h index 0331e9ac2485..db5c73ce86ee 100644 --- a/drivers/gpu/drm/i915/gt/intel_engine.h +++ b/drivers/gpu/drm/i915/gt/intel_engine.h @@ -411,8 +411,8 @@ gen8_emit_ggtt_write(u32 *cs, u32 value, u32 gtt_offset, u32 flags) return cs; } -static inline void intel_engine_reset(struct intel_engine_cs *engine, - bool stalled) +static inline void __intel_engine_reset(struct intel_engine_cs *engine, + bool stalled) { if (engine->reset.reset) engine->reset.reset(engine, stalled); @@ -420,9 +420,9 @@ static inline void intel_engine_reset(struct intel_engine_cs *engine, } bool intel_engine_is_idle(struct intel_engine_cs *engine); -bool intel_engines_are_idle(struct drm_i915_private *dev_priv); +bool intel_engines_are_idle(struct intel_gt *gt); -void intel_engines_reset_default_submission(struct drm_i915_private *i915); +void intel_engines_reset_default_submission(struct intel_gt *gt); unsigned int intel_engines_has_context_isolation(struct drm_i915_private *i915); bool intel_engine_can_store_dword(struct intel_engine_cs *engine); diff --git a/drivers/gpu/drm/i915/gt/intel_engine_cs.c b/drivers/gpu/drm/i915/gt/intel_engine_cs.c index 2dc1917b9d30..022e00eb79ad 100644 --- a/drivers/gpu/drm/i915/gt/intel_engine_cs.c +++ b/drivers/gpu/drm/i915/gt/intel_engine_cs.c @@ -1138,7 +1138,7 @@ static bool ring_is_idle(struct intel_engine_cs *engine) bool intel_engine_is_idle(struct intel_engine_cs *engine) { /* More white lies, if wedged, hw state is inconsistent */ - if (i915_reset_failed(engine->i915)) + if (intel_gt_is_wedged(engine->gt)) return true; if (!intel_engine_pm_is_awake(engine)) @@ -1174,7 +1174,7 @@ bool intel_engine_is_idle(struct intel_engine_cs *engine) return ring_is_idle(engine); } -bool intel_engines_are_idle(struct drm_i915_private *i915) +bool intel_engines_are_idle(struct intel_gt *gt) { struct intel_engine_cs *engine; enum intel_engine_id id; @@ -1183,14 +1183,14 @@ bool intel_engines_are_idle(struct drm_i915_private *i915) * If the driver is wedged, HW state may be very inconsistent and * report that it is still busy, even though we have stopped using it. */ - if (i915_reset_failed(i915)) + if (intel_gt_is_wedged(gt)) return true; /* Already parked (and passed an idleness test); must still be idle */ - if (!READ_ONCE(i915->gt.awake)) + if (!READ_ONCE(gt->awake)) return true; - for_each_engine(engine, i915, id) { + for_each_engine(engine, gt->i915, id) { if (!intel_engine_is_idle(engine)) return false; } @@ -1198,12 +1198,12 @@ bool intel_engines_are_idle(struct drm_i915_private *i915) return true; } -void intel_engines_reset_default_submission(struct drm_i915_private *i915) +void intel_engines_reset_default_submission(struct intel_gt *gt) { struct intel_engine_cs *engine; enum intel_engine_id id; - for_each_engine(engine, i915, id) + for_each_engine(engine, gt->i915, id) engine->set_default_submission(engine); } @@ -1480,7 +1480,7 @@ void intel_engine_dump(struct intel_engine_cs *engine, va_end(ap); } - if (i915_reset_failed(engine->i915)) + if (intel_gt_is_wedged(engine->gt)) drm_printf(m, "*** WEDGED ***\n"); drm_printf(m, "\tAwake? %d\n", atomic_read(&engine->wakeref.count)); diff --git a/drivers/gpu/drm/i915/gt/intel_engine_pm.c b/drivers/gpu/drm/i915/gt/intel_engine_pm.c index 84e432abe8e0..e74fbf04a68d 100644 --- a/drivers/gpu/drm/i915/gt/intel_engine_pm.c +++ b/drivers/gpu/drm/i915/gt/intel_engine_pm.c @@ -8,6 +8,7 @@ #include "intel_engine.h" #include "intel_engine_pm.h" +#include "intel_gt.h" #include "intel_gt_pm.h" static int __engine_unpark(struct intel_wakeref *wf) @@ -66,7 +67,7 @@ static bool switch_to_kernel_context(struct intel_engine_cs *engine) return true; /* GPU is pointing to the void, as good as in the kernel context. */ - if (i915_reset_failed(engine->i915)) + if (intel_gt_is_wedged(engine->gt)) return true; /* diff --git a/drivers/gpu/drm/i915/gt/intel_gt.c b/drivers/gpu/drm/i915/gt/intel_gt.c index 8cca6b22b386..f7e69db4019d 100644 --- a/drivers/gpu/drm/i915/gt/intel_gt.c +++ b/drivers/gpu/drm/i915/gt/intel_gt.c @@ -19,6 +19,8 @@ void intel_gt_init_early(struct intel_gt *gt, struct drm_i915_private *i915) spin_lock_init(>->closed_lock); + intel_gt_init_hangcheck(gt); + intel_gt_init_reset(gt); intel_gt_pm_init_early(gt); } @@ -241,3 +243,8 @@ void intel_gt_fini_scratch(struct intel_gt *gt) { i915_vma_unpin_and_release(>->scratch, 0); } + +void intel_gt_cleanup_early(struct intel_gt *gt) +{ + intel_gt_fini_reset(gt); +} diff --git a/drivers/gpu/drm/i915/gt/intel_gt.h b/drivers/gpu/drm/i915/gt/intel_gt.h index 1093dcf36f63..49c0085385a0 100644 --- a/drivers/gpu/drm/i915/gt/intel_gt.h +++ b/drivers/gpu/drm/i915/gt/intel_gt.h @@ -8,12 +8,15 @@ #include "intel_engine_types.h" #include "intel_gt_types.h" +#include "intel_reset.h" struct drm_i915_private; void intel_gt_init_early(struct intel_gt *gt, struct drm_i915_private *i915); void intel_gt_init_hw(struct drm_i915_private *i915); +void intel_gt_cleanup_early(struct intel_gt *gt); + void intel_gt_check_and_clear_faults(struct intel_gt *gt); void intel_gt_clear_error_registers(struct intel_gt *gt, intel_engine_mask_t engine_mask); @@ -21,6 +24,8 @@ void intel_gt_clear_error_registers(struct intel_gt *gt, void intel_gt_flush_ggtt_writes(struct intel_gt *gt); void intel_gt_chipset_flush(struct intel_gt *gt); +void intel_gt_init_hangcheck(struct intel_gt *gt); + int intel_gt_init_scratch(struct intel_gt *gt, unsigned int size); void intel_gt_fini_scratch(struct intel_gt *gt); @@ -30,4 +35,11 @@ static inline u32 intel_gt_scratch_offset(const struct intel_gt *gt, return i915_ggtt_offset(gt->scratch) + field; } +static inline bool intel_gt_is_wedged(struct intel_gt *gt) +{ + return __intel_reset_failed(>->reset); +} + +void intel_gt_queue_hangcheck(struct intel_gt *gt); + #endif /* __INTEL_GT_H__ */ diff --git a/drivers/gpu/drm/i915/gt/intel_gt_pm.c b/drivers/gpu/drm/i915/gt/intel_gt_pm.c index da81b3a92d16..61ed912341f1 100644 --- a/drivers/gpu/drm/i915/gt/intel_gt_pm.c +++ b/drivers/gpu/drm/i915/gt/intel_gt_pm.c @@ -5,7 +5,9 @@ */ #include "i915_drv.h" +#include "i915_params.h" #include "intel_engine_pm.h" +#include "intel_gt.h" #include "intel_gt_pm.h" #include "intel_pm.h" #include "intel_wakeref.h" @@ -17,8 +19,8 @@ static void pm_notify(struct drm_i915_private *i915, int state) static int intel_gt_unpark(struct intel_wakeref *wf) { - struct drm_i915_private *i915 = - container_of(wf, typeof(*i915), gt.wakeref); + struct intel_gt *gt = container_of(wf, typeof(*gt), wakeref); + struct drm_i915_private *i915 = gt->i915; GEM_TRACE("\n"); @@ -33,8 +35,8 @@ static int intel_gt_unpark(struct intel_wakeref *wf) * Work around it by grabbing a GT IRQ power domain whilst there is any * GT activity, preventing any DC state transitions. */ - i915->gt.awake = intel_display_power_get(i915, POWER_DOMAIN_GT_IRQ); - GEM_BUG_ON(!i915->gt.awake); + gt->awake = intel_display_power_get(i915, POWER_DOMAIN_GT_IRQ); + GEM_BUG_ON(!gt->awake); intel_enable_gt_powersave(i915); @@ -44,7 +46,7 @@ static int intel_gt_unpark(struct intel_wakeref *wf) i915_pmu_gt_unparked(i915); - i915_queue_hangcheck(i915); + intel_gt_queue_hangcheck(gt); pm_notify(i915, INTEL_GT_UNPARK); @@ -91,12 +93,12 @@ void intel_gt_pm_init_early(struct intel_gt *gt) BLOCKING_INIT_NOTIFIER_HEAD(>->pm_notifications); } -static bool reset_engines(struct drm_i915_private *i915) +static bool reset_engines(struct intel_gt *gt) { - if (INTEL_INFO(i915)->gpu_reset_clobbers_display) + if (INTEL_INFO(gt->i915)->gpu_reset_clobbers_display) return false; - return intel_gpu_reset(i915, ALL_ENGINES) == 0; + return __intel_gt_reset(gt, ALL_ENGINES) == 0; } /** @@ -116,11 +118,11 @@ void intel_gt_sanitize(struct intel_gt *gt, bool force) GEM_TRACE("\n"); - if (!reset_engines(gt->i915) && !force) + if (!reset_engines(gt) && !force) return; for_each_engine(engine, gt->i915, id) - intel_engine_reset(engine, false); + __intel_engine_reset(engine, false); } int intel_gt_resume(struct intel_gt *gt) diff --git a/drivers/gpu/drm/i915/gt/intel_gt_types.h b/drivers/gpu/drm/i915/gt/intel_gt_types.h index 3563ce970102..1ffbc3ec6ef3 100644 --- a/drivers/gpu/drm/i915/gt/intel_gt_types.h +++ b/drivers/gpu/drm/i915/gt/intel_gt_types.h @@ -14,12 +14,21 @@ #include #include "i915_vma.h" +#include "intel_reset_types.h" #include "intel_wakeref.h" struct drm_i915_private; struct i915_ggtt; struct intel_uncore; +struct intel_hangcheck { + /* For hangcheck timer */ +#define DRM_I915_HANGCHECK_PERIOD 1500 /* in ms */ +#define DRM_I915_HANGCHECK_JIFFIES msecs_to_jiffies(DRM_I915_HANGCHECK_PERIOD) + + struct delayed_work work; +}; + struct intel_gt { struct drm_i915_private *i915; struct intel_uncore *uncore; @@ -41,6 +50,9 @@ struct intel_gt { struct list_head closed_vma; spinlock_t closed_lock; /* guards the list of closed_vma */ + struct intel_hangcheck hangcheck; + struct intel_reset reset; + /** * Is the GPU currently considered idle, or busy executing * userspace requests? Whilst idle, we allow runtime power diff --git a/drivers/gpu/drm/i915/gt/intel_hangcheck.c b/drivers/gpu/drm/i915/gt/intel_hangcheck.c index 797d8ef0969c..05d042cdefe2 100644 --- a/drivers/gpu/drm/i915/gt/intel_hangcheck.c +++ b/drivers/gpu/drm/i915/gt/intel_hangcheck.c @@ -22,8 +22,10 @@ * */ -#include "intel_reset.h" #include "i915_drv.h" +#include "intel_engine.h" +#include "intel_gt.h" +#include "intel_reset.h" struct hangcheck { u64 acthd; @@ -100,7 +102,6 @@ head_stuck(struct intel_engine_cs *engine, u64 acthd) static enum intel_engine_hangcheck_action engine_stuck(struct intel_engine_cs *engine, u64 acthd) { - struct drm_i915_private *dev_priv = engine->i915; enum intel_engine_hangcheck_action ha; u32 tmp; @@ -108,7 +109,7 @@ engine_stuck(struct intel_engine_cs *engine, u64 acthd) if (ha != ENGINE_DEAD) return ha; - if (IS_GEN(dev_priv, 2)) + if (IS_GEN(engine->i915, 2)) return ENGINE_DEAD; /* Is the chip hanging on a WAIT_FOR_EVENT? @@ -118,8 +119,8 @@ engine_stuck(struct intel_engine_cs *engine, u64 acthd) */ tmp = ENGINE_READ(engine, RING_CTL); if (tmp & RING_WAIT) { - i915_handle_error(dev_priv, engine->mask, 0, - "stuck wait on %s", engine->name); + intel_gt_handle_error(engine->gt, engine->mask, 0, + "stuck wait on %s", engine->name); ENGINE_WRITE(engine, RING_CTL, tmp); return ENGINE_WAIT_KICK; } @@ -219,7 +220,7 @@ static void hangcheck_accumulate_sample(struct intel_engine_cs *engine, I915_ENGINE_WEDGED_TIMEOUT); } -static void hangcheck_declare_hang(struct drm_i915_private *i915, +static void hangcheck_declare_hang(struct intel_gt *gt, intel_engine_mask_t hung, intel_engine_mask_t stuck) { @@ -235,12 +236,12 @@ static void hangcheck_declare_hang(struct drm_i915_private *i915, hung &= ~stuck; len = scnprintf(msg, sizeof(msg), "%s on ", stuck == hung ? "no progress" : "hang"); - for_each_engine_masked(engine, i915, hung, tmp) + for_each_engine_masked(engine, gt->i915, hung, tmp) len += scnprintf(msg + len, sizeof(msg) - len, "%s, ", engine->name); msg[len-2] = '\0'; - return i915_handle_error(i915, hung, I915_ERROR_CAPTURE, "%s", msg); + return intel_gt_handle_error(gt, hung, I915_ERROR_CAPTURE, "%s", msg); } /* @@ -251,11 +252,10 @@ static void hangcheck_declare_hang(struct drm_i915_private *i915, * we kick the ring. If we see no progress on three subsequent calls * we assume chip is wedged and try to fix it by resetting the chip. */ -static void i915_hangcheck_elapsed(struct work_struct *work) +static void hangcheck_elapsed(struct work_struct *work) { - struct drm_i915_private *dev_priv = - container_of(work, typeof(*dev_priv), - gpu_error.hangcheck_work.work); + struct intel_gt *gt = + container_of(work, typeof(*gt), hangcheck.work.work); intel_engine_mask_t hung = 0, stuck = 0, wedged = 0; struct intel_engine_cs *engine; enum intel_engine_id id; @@ -264,13 +264,13 @@ static void i915_hangcheck_elapsed(struct work_struct *work) if (!i915_modparams.enable_hangcheck) return; - if (!READ_ONCE(dev_priv->gt.awake)) + if (!READ_ONCE(gt->awake)) return; - if (i915_terminally_wedged(dev_priv)) + if (intel_gt_is_wedged(gt)) return; - wakeref = intel_runtime_pm_get_if_in_use(&dev_priv->runtime_pm); + wakeref = intel_runtime_pm_get_if_in_use(>->i915->runtime_pm); if (!wakeref) return; @@ -278,9 +278,9 @@ static void i915_hangcheck_elapsed(struct work_struct *work) * periodically arm the mmio checker to see if we are triggering * any invalid access. */ - intel_uncore_arm_unclaimed_mmio_detection(&dev_priv->uncore); + intel_uncore_arm_unclaimed_mmio_detection(gt->uncore); - for_each_engine(engine, dev_priv, id) { + for_each_engine(engine, gt->i915, id) { struct hangcheck hc; intel_engine_signal_breadcrumbs(engine); @@ -302,7 +302,7 @@ static void i915_hangcheck_elapsed(struct work_struct *work) if (GEM_SHOW_DEBUG() && (hung | stuck)) { struct drm_printer p = drm_debug_printer("hangcheck"); - for_each_engine(engine, dev_priv, id) { + for_each_engine(engine, gt->i915, id) { if (intel_engine_is_idle(engine)) continue; @@ -311,20 +311,37 @@ static void i915_hangcheck_elapsed(struct work_struct *work) } if (wedged) { - dev_err(dev_priv->drm.dev, + dev_err(gt->i915->drm.dev, "GPU recovery timed out," " cancelling all in-flight rendering.\n"); GEM_TRACE_DUMP(); - i915_gem_set_wedged(dev_priv); + intel_gt_set_wedged(gt); } if (hung) - hangcheck_declare_hang(dev_priv, hung, stuck); + hangcheck_declare_hang(gt, hung, stuck); - intel_runtime_pm_put(&dev_priv->runtime_pm, wakeref); + intel_runtime_pm_put(>->i915->runtime_pm, wakeref); /* Reset timer in case GPU hangs without another request being added */ - i915_queue_hangcheck(dev_priv); + intel_gt_queue_hangcheck(gt); +} + +void intel_gt_queue_hangcheck(struct intel_gt *gt) +{ + unsigned long delay; + + if (unlikely(!i915_modparams.enable_hangcheck)) + return; + + /* + * Don't continually defer the hangcheck so that it is always run at + * least once after work has been scheduled on any ring. Otherwise, + * we will ignore a hung ring if a second ring is kept busy. + */ + + delay = round_jiffies_up_relative(DRM_I915_HANGCHECK_JIFFIES); + queue_delayed_work(system_long_wq, >->hangcheck.work, delay); } void intel_engine_init_hangcheck(struct intel_engine_cs *engine) @@ -333,10 +350,9 @@ void intel_engine_init_hangcheck(struct intel_engine_cs *engine) engine->hangcheck.action_timestamp = jiffies; } -void intel_hangcheck_init(struct drm_i915_private *i915) +void intel_gt_init_hangcheck(struct intel_gt *gt) { - INIT_DELAYED_WORK(&i915->gpu_error.hangcheck_work, - i915_hangcheck_elapsed); + INIT_DELAYED_WORK(>->hangcheck.work, hangcheck_elapsed); } #if IS_ENABLED(CONFIG_DRM_I915_SELFTEST) diff --git a/drivers/gpu/drm/i915/gt/intel_lrc.c b/drivers/gpu/drm/i915/gt/intel_lrc.c index 270ef417dd1a..9e0992498087 100644 --- a/drivers/gpu/drm/i915/gt/intel_lrc.c +++ b/drivers/gpu/drm/i915/gt/intel_lrc.c @@ -2293,7 +2293,7 @@ static void __execlists_reset(struct intel_engine_cs *engine, bool stalled) * and have to at least restore the RING register in the context * image back to the expected values to skip over the guilty request. */ - i915_reset_request(rq, stalled); + __i915_request_reset(rq, stalled); if (!stalled) goto out_replay; diff --git a/drivers/gpu/drm/i915/gt/intel_reset.c b/drivers/gpu/drm/i915/gt/intel_reset.c index adfdb908587f..7b150ec3d10a 100644 --- a/drivers/gpu/drm/i915/gt/intel_reset.c +++ b/drivers/gpu/drm/i915/gt/intel_reset.c @@ -114,7 +114,7 @@ static void context_mark_innocent(struct i915_gem_context *ctx) atomic_inc(&ctx->active_count); } -void i915_reset_request(struct i915_request *rq, bool guilty) +void __i915_request_reset(struct i915_request *rq, bool guilty) { GEM_TRACE("%s rq=%llx:%lld, guilty? %s\n", rq->engine->name, @@ -164,16 +164,15 @@ static void gen3_stop_engine(struct intel_engine_cs *engine) intel_uncore_read_fw(uncore, RING_HEAD(base))); } -static void i915_stop_engines(struct drm_i915_private *i915, - intel_engine_mask_t engine_mask) +static void stop_engines(struct intel_gt *gt, intel_engine_mask_t engine_mask) { struct intel_engine_cs *engine; intel_engine_mask_t tmp; - if (INTEL_GEN(i915) < 3) + if (INTEL_GEN(gt->i915) < 3) return; - for_each_engine_masked(engine, i915, engine_mask, tmp) + for_each_engine_masked(engine, gt->i915, engine_mask, tmp) gen3_stop_engine(engine); } @@ -185,11 +184,11 @@ static bool i915_in_reset(struct pci_dev *pdev) return gdrst & GRDOM_RESET_STATUS; } -static int i915_do_reset(struct drm_i915_private *i915, +static int i915_do_reset(struct intel_gt *gt, intel_engine_mask_t engine_mask, unsigned int retry) { - struct pci_dev *pdev = i915->drm.pdev; + struct pci_dev *pdev = gt->i915->drm.pdev; int err; /* Assert reset for at least 20 usec, and wait for acknowledgement. */ @@ -214,22 +213,22 @@ static bool g4x_reset_complete(struct pci_dev *pdev) return (gdrst & GRDOM_RESET_ENABLE) == 0; } -static int g33_do_reset(struct drm_i915_private *i915, +static int g33_do_reset(struct intel_gt *gt, intel_engine_mask_t engine_mask, unsigned int retry) { - struct pci_dev *pdev = i915->drm.pdev; + struct pci_dev *pdev = gt->i915->drm.pdev; pci_write_config_byte(pdev, I915_GDRST, GRDOM_RESET_ENABLE); return wait_for_atomic(g4x_reset_complete(pdev), 50); } -static int g4x_do_reset(struct drm_i915_private *i915, +static int g4x_do_reset(struct intel_gt *gt, intel_engine_mask_t engine_mask, unsigned int retry) { - struct pci_dev *pdev = i915->drm.pdev; - struct intel_uncore *uncore = &i915->uncore; + struct pci_dev *pdev = gt->i915->drm.pdev; + struct intel_uncore *uncore = gt->uncore; int ret; /* WaVcpClkGateDisableForMediaReset:ctg,elk */ @@ -261,11 +260,11 @@ out: return ret; } -static int ironlake_do_reset(struct drm_i915_private *i915, +static int ironlake_do_reset(struct intel_gt *gt, intel_engine_mask_t engine_mask, unsigned int retry) { - struct intel_uncore *uncore = &i915->uncore; + struct intel_uncore *uncore = gt->uncore; int ret; intel_uncore_write_fw(uncore, ILK_GDSR, @@ -297,10 +296,9 @@ out: } /* Reset the hardware domains (GENX_GRDOM_*) specified by mask */ -static int gen6_hw_domain_reset(struct drm_i915_private *i915, - u32 hw_domain_mask) +static int gen6_hw_domain_reset(struct intel_gt *gt, u32 hw_domain_mask) { - struct intel_uncore *uncore = &i915->uncore; + struct intel_uncore *uncore = gt->uncore; int err; /* @@ -322,7 +320,7 @@ static int gen6_hw_domain_reset(struct drm_i915_private *i915, return err; } -static int gen6_reset_engines(struct drm_i915_private *i915, +static int gen6_reset_engines(struct intel_gt *gt, intel_engine_mask_t engine_mask, unsigned int retry) { @@ -342,13 +340,13 @@ static int gen6_reset_engines(struct drm_i915_private *i915, intel_engine_mask_t tmp; hw_mask = 0; - for_each_engine_masked(engine, i915, engine_mask, tmp) { + for_each_engine_masked(engine, gt->i915, engine_mask, tmp) { GEM_BUG_ON(engine->id >= ARRAY_SIZE(hw_engine_mask)); hw_mask |= hw_engine_mask[engine->id]; } } - return gen6_hw_domain_reset(i915, hw_mask); + return gen6_hw_domain_reset(gt, hw_mask); } static u32 gen11_lock_sfc(struct intel_engine_cs *engine) @@ -446,7 +444,7 @@ static void gen11_unlock_sfc(struct intel_engine_cs *engine) rmw_clear_fw(uncore, sfc_forced_lock, sfc_forced_lock_bit); } -static int gen11_reset_engines(struct drm_i915_private *i915, +static int gen11_reset_engines(struct intel_gt *gt, intel_engine_mask_t engine_mask, unsigned int retry) { @@ -469,17 +467,17 @@ static int gen11_reset_engines(struct drm_i915_private *i915, hw_mask = GEN11_GRDOM_FULL; } else { hw_mask = 0; - for_each_engine_masked(engine, i915, engine_mask, tmp) { + for_each_engine_masked(engine, gt->i915, engine_mask, tmp) { GEM_BUG_ON(engine->id >= ARRAY_SIZE(hw_engine_mask)); hw_mask |= hw_engine_mask[engine->id]; hw_mask |= gen11_lock_sfc(engine); } } - ret = gen6_hw_domain_reset(i915, hw_mask); + ret = gen6_hw_domain_reset(gt, hw_mask); if (engine_mask != ALL_ENGINES) - for_each_engine_masked(engine, i915, engine_mask, tmp) + for_each_engine_masked(engine, gt->i915, engine_mask, tmp) gen11_unlock_sfc(engine); return ret; @@ -529,7 +527,7 @@ static void gen8_engine_reset_cancel(struct intel_engine_cs *engine) _MASKED_BIT_DISABLE(RESET_CTL_REQUEST_RESET)); } -static int gen8_reset_engines(struct drm_i915_private *i915, +static int gen8_reset_engines(struct intel_gt *gt, intel_engine_mask_t engine_mask, unsigned int retry) { @@ -538,7 +536,7 @@ static int gen8_reset_engines(struct drm_i915_private *i915, intel_engine_mask_t tmp; int ret; - for_each_engine_masked(engine, i915, engine_mask, tmp) { + for_each_engine_masked(engine, gt->i915, engine_mask, tmp) { ret = gen8_engine_reset_prepare(engine); if (ret && !reset_non_ready) goto skip_reset; @@ -554,23 +552,23 @@ static int gen8_reset_engines(struct drm_i915_private *i915, * We rather take context corruption instead of * failed reset with a wedged driver/gpu. And * active bb execution case should be covered by - * i915_stop_engines we have before the reset. + * stop_engines() we have before the reset. */ } - if (INTEL_GEN(i915) >= 11) - ret = gen11_reset_engines(i915, engine_mask, retry); + if (INTEL_GEN(gt->i915) >= 11) + ret = gen11_reset_engines(gt, engine_mask, retry); else - ret = gen6_reset_engines(i915, engine_mask, retry); + ret = gen6_reset_engines(gt, engine_mask, retry); skip_reset: - for_each_engine_masked(engine, i915, engine_mask, tmp) + for_each_engine_masked(engine, gt->i915, engine_mask, tmp) gen8_engine_reset_cancel(engine); return ret; } -typedef int (*reset_func)(struct drm_i915_private *, +typedef int (*reset_func)(struct intel_gt *, intel_engine_mask_t engine_mask, unsigned int retry); @@ -592,15 +590,14 @@ static reset_func intel_get_gpu_reset(struct drm_i915_private *i915) return NULL; } -int intel_gpu_reset(struct drm_i915_private *i915, - intel_engine_mask_t engine_mask) +int __intel_gt_reset(struct intel_gt *gt, intel_engine_mask_t engine_mask) { const int retries = engine_mask == ALL_ENGINES ? RESET_MAX_RETRIES : 1; reset_func reset; int ret = -ETIMEDOUT; int retry; - reset = intel_get_gpu_reset(i915); + reset = intel_get_gpu_reset(gt->i915); if (!reset) return -ENODEV; @@ -608,7 +605,7 @@ int intel_gpu_reset(struct drm_i915_private *i915, * If the power well sleeps during the reset, the reset * request may be dropped and never completes (causing -EIO). */ - intel_uncore_forcewake_get(&i915->uncore, FORCEWAKE_ALL); + intel_uncore_forcewake_get(gt->uncore, FORCEWAKE_ALL); for (retry = 0; ret == -ETIMEDOUT && retry < retries; retry++) { /* * We stop engines, otherwise we might get failed reset and a @@ -625,14 +622,14 @@ int intel_gpu_reset(struct drm_i915_private *i915, * FIXME: Wa for more modern gens needs to be validated */ if (retry) - i915_stop_engines(i915, engine_mask); + stop_engines(gt, engine_mask); GEM_TRACE("engine_mask=%x\n", engine_mask); preempt_disable(); - ret = reset(i915, engine_mask, retry); + ret = reset(gt, engine_mask, retry); preempt_enable(); } - intel_uncore_forcewake_put(&i915->uncore, FORCEWAKE_ALL); + intel_uncore_forcewake_put(gt->uncore, FORCEWAKE_ALL); return ret; } @@ -650,17 +647,17 @@ bool intel_has_reset_engine(struct drm_i915_private *i915) return INTEL_INFO(i915)->has_reset_engine && i915_modparams.reset >= 2; } -int intel_reset_guc(struct drm_i915_private *i915) +int intel_reset_guc(struct intel_gt *gt) { u32 guc_domain = - INTEL_GEN(i915) >= 11 ? GEN11_GRDOM_GUC : GEN9_GRDOM_GUC; + INTEL_GEN(gt->i915) >= 11 ? GEN11_GRDOM_GUC : GEN9_GRDOM_GUC; int ret; - GEM_BUG_ON(!HAS_GUC(i915)); + GEM_BUG_ON(!HAS_GUC(gt->i915)); - intel_uncore_forcewake_get(&i915->uncore, FORCEWAKE_ALL); - ret = gen6_hw_domain_reset(i915, guc_domain); - intel_uncore_forcewake_put(&i915->uncore, FORCEWAKE_ALL); + intel_uncore_forcewake_get(gt->uncore, FORCEWAKE_ALL); + ret = gen6_hw_domain_reset(gt, guc_domain); + intel_uncore_forcewake_put(gt->uncore, FORCEWAKE_ALL); return ret; } @@ -682,56 +679,55 @@ static void reset_prepare_engine(struct intel_engine_cs *engine) engine->reset.prepare(engine); } -static void revoke_mmaps(struct drm_i915_private *i915) +static void revoke_mmaps(struct intel_gt *gt) { int i; - for (i = 0; i < i915->ggtt.num_fences; i++) { + for (i = 0; i < gt->ggtt->num_fences; i++) { struct drm_vma_offset_node *node; struct i915_vma *vma; u64 vma_offset; - vma = READ_ONCE(i915->ggtt.fence_regs[i].vma); + vma = READ_ONCE(gt->ggtt->fence_regs[i].vma); if (!vma) continue; if (!i915_vma_has_userfault(vma)) continue; - GEM_BUG_ON(vma->fence != &i915->ggtt.fence_regs[i]); + GEM_BUG_ON(vma->fence != >->ggtt->fence_regs[i]); node = &vma->obj->base.vma_node; vma_offset = vma->ggtt_view.partial.offset << PAGE_SHIFT; - unmap_mapping_range(i915->drm.anon_inode->i_mapping, + unmap_mapping_range(gt->i915->drm.anon_inode->i_mapping, drm_vma_node_offset_addr(node) + vma_offset, vma->size, 1); } } -static intel_engine_mask_t reset_prepare(struct drm_i915_private *i915) +static intel_engine_mask_t reset_prepare(struct intel_gt *gt) { struct intel_engine_cs *engine; intel_engine_mask_t awake = 0; enum intel_engine_id id; - for_each_engine(engine, i915, id) { + for_each_engine(engine, gt->i915, id) { if (intel_engine_pm_get_if_awake(engine)) awake |= engine->mask; reset_prepare_engine(engine); } - intel_uc_reset_prepare(i915); + intel_uc_reset_prepare(gt->i915); return awake; } -static void gt_revoke(struct drm_i915_private *i915) +static void gt_revoke(struct intel_gt *gt) { - revoke_mmaps(i915); + revoke_mmaps(gt); } -static int gt_reset(struct drm_i915_private *i915, - intel_engine_mask_t stalled_mask) +static int gt_reset(struct intel_gt *gt, intel_engine_mask_t stalled_mask) { struct intel_engine_cs *engine; enum intel_engine_id id; @@ -741,14 +737,14 @@ static int gt_reset(struct drm_i915_private *i915, * Everything depends on having the GTT running, so we need to start * there. */ - err = i915_ggtt_enable_hw(i915); + err = i915_ggtt_enable_hw(gt->i915); if (err) return err; - for_each_engine(engine, i915, id) - intel_engine_reset(engine, stalled_mask & engine->mask); + for_each_engine(engine, gt->i915, id) + __intel_engine_reset(engine, stalled_mask & engine->mask); - i915_gem_restore_fences(i915); + i915_gem_restore_fences(gt->i915); return err; } @@ -761,13 +757,12 @@ static void reset_finish_engine(struct intel_engine_cs *engine) intel_engine_signal_breadcrumbs(engine); } -static void reset_finish(struct drm_i915_private *i915, - intel_engine_mask_t awake) +static void reset_finish(struct intel_gt *gt, intel_engine_mask_t awake) { struct intel_engine_cs *engine; enum intel_engine_id id; - for_each_engine(engine, i915, id) { + for_each_engine(engine, gt->i915, id) { reset_finish_engine(engine); if (awake & engine->mask) intel_engine_pm_put(engine); @@ -791,20 +786,19 @@ static void nop_submit_request(struct i915_request *request) intel_engine_queue_breadcrumbs(engine); } -static void __i915_gem_set_wedged(struct drm_i915_private *i915) +static void __intel_gt_set_wedged(struct intel_gt *gt) { - struct i915_gpu_error *error = &i915->gpu_error; struct intel_engine_cs *engine; intel_engine_mask_t awake; enum intel_engine_id id; - if (test_bit(I915_WEDGED, &error->flags)) + if (test_bit(I915_WEDGED, >->reset.flags)) return; - if (GEM_SHOW_DEBUG() && !intel_engines_are_idle(i915)) { + if (GEM_SHOW_DEBUG() && !intel_engines_are_idle(gt)) { struct drm_printer p = drm_debug_printer(__func__); - for_each_engine(engine, i915, id) + for_each_engine(engine, gt->i915, id) intel_engine_dump(engine, &p, "%s\n", engine->name); } @@ -815,17 +809,17 @@ static void __i915_gem_set_wedged(struct drm_i915_private *i915) * rolling the global seqno forward (since this would complete requests * for which we haven't set the fence error to EIO yet). */ - awake = reset_prepare(i915); + awake = reset_prepare(gt); /* Even if the GPU reset fails, it should still stop the engines */ - if (!INTEL_INFO(i915)->gpu_reset_clobbers_display) - intel_gpu_reset(i915, ALL_ENGINES); + if (!INTEL_INFO(gt->i915)->gpu_reset_clobbers_display) + __intel_gt_reset(gt, ALL_ENGINES); - for_each_engine(engine, i915, id) { + for_each_engine(engine, gt->i915, id) { engine->submit_request = nop_submit_request; engine->schedule = NULL; } - i915->caps.scheduler = 0; + gt->i915->caps.scheduler = 0; /* * Make sure no request can slip through without getting completed by @@ -833,37 +827,36 @@ static void __i915_gem_set_wedged(struct drm_i915_private *i915) * in nop_submit_request. */ synchronize_rcu_expedited(); - set_bit(I915_WEDGED, &error->flags); + set_bit(I915_WEDGED, >->reset.flags); /* Mark all executing requests as skipped */ - for_each_engine(engine, i915, id) + for_each_engine(engine, gt->i915, id) engine->cancel_requests(engine); - reset_finish(i915, awake); + reset_finish(gt, awake); GEM_TRACE("end\n"); } -void i915_gem_set_wedged(struct drm_i915_private *i915) +void intel_gt_set_wedged(struct intel_gt *gt) { - struct i915_gpu_error *error = &i915->gpu_error; intel_wakeref_t wakeref; - mutex_lock(&error->wedge_mutex); - with_intel_runtime_pm(&i915->runtime_pm, wakeref) - __i915_gem_set_wedged(i915); - mutex_unlock(&error->wedge_mutex); + mutex_lock(>->reset.mutex); + with_intel_runtime_pm(>->i915->runtime_pm, wakeref) + __intel_gt_set_wedged(gt); + mutex_unlock(>->reset.mutex); } -static bool __i915_gem_unset_wedged(struct drm_i915_private *i915) +static bool __intel_gt_unset_wedged(struct intel_gt *gt) { - struct i915_gpu_error *error = &i915->gpu_error; + struct intel_gt_timelines *timelines = >->timelines; struct intel_timeline *tl; - if (!test_bit(I915_WEDGED, &error->flags)) + if (!test_bit(I915_WEDGED, >->reset.flags)) return true; - if (!i915->gt.scratch) /* Never full initialised, recovery impossible */ + if (!gt->scratch) /* Never full initialised, recovery impossible */ return false; GEM_TRACE("start\n"); @@ -878,8 +871,8 @@ static bool __i915_gem_unset_wedged(struct drm_i915_private *i915) * * No more can be submitted until we reset the wedged bit. */ - mutex_lock(&i915->gt.timelines.mutex); - list_for_each_entry(tl, &i915->gt.timelines.active_list, link) { + mutex_lock(&timelines->mutex); + list_for_each_entry(tl, &timelines->active_list, link) { struct i915_request *rq; rq = i915_active_request_get_unlocked(&tl->last_request); @@ -896,9 +889,9 @@ static bool __i915_gem_unset_wedged(struct drm_i915_private *i915) dma_fence_default_wait(&rq->fence, false, MAX_SCHEDULE_TIMEOUT); i915_request_put(rq); } - mutex_unlock(&i915->gt.timelines.mutex); + mutex_unlock(&timelines->mutex); - intel_gt_sanitize(&i915->gt, false); + intel_gt_sanitize(gt, false); /* * Undo nop_submit_request. We prevent all new i915 requests from @@ -909,53 +902,51 @@ static bool __i915_gem_unset_wedged(struct drm_i915_private *i915) * the nop_submit_request on reset, we can do this from normal * context and do not require stop_machine(). */ - intel_engines_reset_default_submission(i915); + intel_engines_reset_default_submission(gt); GEM_TRACE("end\n"); smp_mb__before_atomic(); /* complete takeover before enabling execbuf */ - clear_bit(I915_WEDGED, &i915->gpu_error.flags); + clear_bit(I915_WEDGED, >->reset.flags); return true; } -bool i915_gem_unset_wedged(struct drm_i915_private *i915) +bool intel_gt_unset_wedged(struct intel_gt *gt) { - struct i915_gpu_error *error = &i915->gpu_error; bool result; - mutex_lock(&error->wedge_mutex); - result = __i915_gem_unset_wedged(i915); - mutex_unlock(&error->wedge_mutex); + mutex_lock(>->reset.mutex); + result = __intel_gt_unset_wedged(gt); + mutex_unlock(>->reset.mutex); return result; } -static int do_reset(struct drm_i915_private *i915, - intel_engine_mask_t stalled_mask) +static int do_reset(struct intel_gt *gt, intel_engine_mask_t stalled_mask) { int err, i; - gt_revoke(i915); + gt_revoke(gt); - err = intel_gpu_reset(i915, ALL_ENGINES); + err = __intel_gt_reset(gt, ALL_ENGINES); for (i = 0; err && i < RESET_MAX_RETRIES; i++) { msleep(10 * (i + 1)); - err = intel_gpu_reset(i915, ALL_ENGINES); + err = __intel_gt_reset(gt, ALL_ENGINES); } if (err) return err; - return gt_reset(i915, stalled_mask); + return gt_reset(gt, stalled_mask); } -static int resume(struct drm_i915_private *i915) +static int resume(struct intel_gt *gt) { struct intel_engine_cs *engine; enum intel_engine_id id; int ret; - for_each_engine(engine, i915, id) { + for_each_engine(engine, gt->i915, id) { ret = engine->resume(engine); if (ret) return ret; @@ -965,8 +956,8 @@ static int resume(struct drm_i915_private *i915) } /** - * i915_reset - reset chip after a hang - * @i915: #drm_i915_private to reset + * intel_gt_reset - reset chip after a hang + * @gt: #intel_gt to reset * @stalled_mask: mask of the stalled engines with the guilty requests * @reason: user error message for why we are resetting * @@ -981,50 +972,50 @@ static int resume(struct drm_i915_private *i915) * - re-init interrupt state * - re-init display */ -void i915_reset(struct drm_i915_private *i915, - intel_engine_mask_t stalled_mask, - const char *reason) +void intel_gt_reset(struct intel_gt *gt, + intel_engine_mask_t stalled_mask, + const char *reason) { - struct i915_gpu_error *error = &i915->gpu_error; intel_engine_mask_t awake; int ret; - GEM_TRACE("flags=%lx\n", error->flags); + GEM_TRACE("flags=%lx\n", gt->reset.flags); might_sleep(); - GEM_BUG_ON(!test_bit(I915_RESET_BACKOFF, &error->flags)); - mutex_lock(&error->wedge_mutex); + GEM_BUG_ON(!test_bit(I915_RESET_BACKOFF, >->reset.flags)); + mutex_lock(>->reset.mutex); /* Clear any previous failed attempts at recovery. Time to try again. */ - if (!__i915_gem_unset_wedged(i915)) + if (!__intel_gt_unset_wedged(gt)) goto unlock; if (reason) - dev_notice(i915->drm.dev, "Resetting chip for %s\n", reason); - error->reset_count++; + dev_notice(gt->i915->drm.dev, + "Resetting chip for %s\n", reason); + atomic_inc(>->i915->gpu_error.reset_count); - awake = reset_prepare(i915); + awake = reset_prepare(gt); - if (!intel_has_gpu_reset(i915)) { + if (!intel_has_gpu_reset(gt->i915)) { if (i915_modparams.reset) - dev_err(i915->drm.dev, "GPU reset not supported\n"); + dev_err(gt->i915->drm.dev, "GPU reset not supported\n"); else DRM_DEBUG_DRIVER("GPU reset disabled\n"); goto error; } - if (INTEL_INFO(i915)->gpu_reset_clobbers_display) - intel_runtime_pm_disable_interrupts(i915); + if (INTEL_INFO(gt->i915)->gpu_reset_clobbers_display) + intel_runtime_pm_disable_interrupts(gt->i915); - if (do_reset(i915, stalled_mask)) { - dev_err(i915->drm.dev, "Failed to reset chip\n"); + if (do_reset(gt, stalled_mask)) { + dev_err(gt->i915->drm.dev, "Failed to reset chip\n"); goto taint; } - if (INTEL_INFO(i915)->gpu_reset_clobbers_display) - intel_runtime_pm_enable_interrupts(i915); + if (INTEL_INFO(gt->i915)->gpu_reset_clobbers_display) + intel_runtime_pm_enable_interrupts(gt->i915); - intel_overlay_reset(i915); + intel_overlay_reset(gt->i915); /* * Next we need to restore the context, but we don't use those @@ -1034,23 +1025,23 @@ void i915_reset(struct drm_i915_private *i915, * was running at the time of the reset (i.e. we weren't VT * switched away). */ - ret = i915_gem_init_hw(i915); + ret = i915_gem_init_hw(gt->i915); if (ret) { DRM_ERROR("Failed to initialise HW following reset (%d)\n", ret); goto taint; } - ret = resume(i915); + ret = resume(gt); if (ret) goto taint; - i915_queue_hangcheck(i915); + intel_gt_queue_hangcheck(gt); finish: - reset_finish(i915, awake); + reset_finish(gt, awake); unlock: - mutex_unlock(&error->wedge_mutex); + mutex_unlock(>->reset.mutex); return; taint: @@ -1068,18 +1059,17 @@ taint: */ add_taint_for_CI(TAINT_WARN); error: - __i915_gem_set_wedged(i915); + __intel_gt_set_wedged(gt); goto finish; } -static inline int intel_gt_reset_engine(struct drm_i915_private *i915, - struct intel_engine_cs *engine) +static inline int intel_gt_reset_engine(struct intel_engine_cs *engine) { - return intel_gpu_reset(i915, engine->mask); + return __intel_gt_reset(engine->gt, engine->mask); } /** - * i915_reset_engine - reset GPU engine to recover from a hang + * intel_engine_reset - reset GPU engine to recover from a hang * @engine: engine to reset * @msg: reason for GPU reset; or NULL for no dev_notice() * @@ -1091,13 +1081,13 @@ static inline int intel_gt_reset_engine(struct drm_i915_private *i915, * - reset engine (which will force the engine to idle) * - re-init/configure engine */ -int i915_reset_engine(struct intel_engine_cs *engine, const char *msg) +int intel_engine_reset(struct intel_engine_cs *engine, const char *msg) { - struct i915_gpu_error *error = &engine->i915->gpu_error; + struct intel_gt *gt = engine->gt; int ret; - GEM_TRACE("%s flags=%lx\n", engine->name, error->flags); - GEM_BUG_ON(!test_bit(I915_RESET_ENGINE + engine->id, &error->flags)); + GEM_TRACE("%s flags=%lx\n", engine->name, gt->reset.flags); + GEM_BUG_ON(!test_bit(I915_RESET_ENGINE + engine->id, >->reset.flags)); if (!intel_engine_pm_get_if_awake(engine)) return 0; @@ -1107,10 +1097,10 @@ int i915_reset_engine(struct intel_engine_cs *engine, const char *msg) if (msg) dev_notice(engine->i915->drm.dev, "Resetting %s for %s\n", engine->name, msg); - error->reset_engine_count[engine->id]++; + atomic_inc(&engine->i915->gpu_error.reset_engine_count[engine->uabi_class]); if (!engine->i915->guc.execbuf_client) - ret = intel_gt_reset_engine(engine->i915, engine); + ret = intel_gt_reset_engine(engine); else ret = intel_guc_reset_engine(&engine->i915->guc, engine); if (ret) { @@ -1126,7 +1116,7 @@ int i915_reset_engine(struct intel_engine_cs *engine, const char *msg) * active request and can drop it, adjust head to skip the offending * request to resume executing remaining requests in the queue. */ - intel_engine_reset(engine, true); + __intel_engine_reset(engine, true); /* * The engine and its registers (and workarounds in case of render) @@ -1142,16 +1132,15 @@ out: return ret; } -static void i915_reset_device(struct drm_i915_private *i915, - u32 engine_mask, - const char *reason) +static void intel_gt_reset_global(struct intel_gt *gt, + u32 engine_mask, + const char *reason) { - struct i915_gpu_error *error = &i915->gpu_error; - struct kobject *kobj = &i915->drm.primary->kdev->kobj; + struct kobject *kobj = >->i915->drm.primary->kdev->kobj; char *error_event[] = { I915_ERROR_UEVENT "=1", NULL }; char *reset_event[] = { I915_RESET_UEVENT "=1", NULL }; char *reset_done_event[] = { I915_ERROR_UEVENT "=0", NULL }; - struct i915_wedge_me w; + struct intel_wedge_me w; kobject_uevent_env(kobj, KOBJ_CHANGE, error_event); @@ -1159,24 +1148,24 @@ static void i915_reset_device(struct drm_i915_private *i915, kobject_uevent_env(kobj, KOBJ_CHANGE, reset_event); /* Use a watchdog to ensure that our reset completes */ - i915_wedge_on_timeout(&w, i915, 5 * HZ) { - intel_prepare_reset(i915); + intel_wedge_on_timeout(&w, gt, 5 * HZ) { + intel_prepare_reset(gt->i915); /* Flush everyone using a resource about to be clobbered */ - synchronize_srcu_expedited(&error->reset_backoff_srcu); + synchronize_srcu_expedited(>->reset.backoff_srcu); - i915_reset(i915, engine_mask, reason); + intel_gt_reset(gt, engine_mask, reason); - intel_finish_reset(i915); + intel_finish_reset(gt->i915); } - if (!test_bit(I915_WEDGED, &error->flags)) + if (!test_bit(I915_WEDGED, >->reset.flags)) kobject_uevent_env(kobj, KOBJ_CHANGE, reset_done_event); } /** - * i915_handle_error - handle a gpu error - * @i915: i915 device private + * intel_gt_handle_error - handle a gpu error + * @gt: the intel_gt * @engine_mask: mask representing engines that are hung * @flags: control flags * @fmt: Error message format string @@ -1187,12 +1176,11 @@ static void i915_reset_device(struct drm_i915_private *i915, * so userspace knows something bad happened (should trigger collection * of a ring dump etc.). */ -void i915_handle_error(struct drm_i915_private *i915, - intel_engine_mask_t engine_mask, - unsigned long flags, - const char *fmt, ...) +void intel_gt_handle_error(struct intel_gt *gt, + intel_engine_mask_t engine_mask, + unsigned long flags, + const char *fmt, ...) { - struct i915_gpu_error *error = &i915->gpu_error; struct intel_engine_cs *engine; intel_wakeref_t wakeref; intel_engine_mask_t tmp; @@ -1216,33 +1204,31 @@ void i915_handle_error(struct drm_i915_private *i915, * isn't the case at least when we get here by doing a * simulated reset via debugfs, so get an RPM reference. */ - wakeref = intel_runtime_pm_get(&i915->runtime_pm); + wakeref = intel_runtime_pm_get(>->i915->runtime_pm); - engine_mask &= INTEL_INFO(i915)->engine_mask; + engine_mask &= INTEL_INFO(gt->i915)->engine_mask; if (flags & I915_ERROR_CAPTURE) { - i915_capture_error_state(i915, engine_mask, msg); - intel_gt_clear_error_registers(&i915->gt, engine_mask); + i915_capture_error_state(gt->i915, engine_mask, msg); + intel_gt_clear_error_registers(gt, engine_mask); } /* * Try engine reset when available. We fall back to full reset if * single reset fails. */ - if (intel_has_reset_engine(i915) && !__i915_wedged(error)) { - for_each_engine_masked(engine, i915, engine_mask, tmp) { + if (intel_has_reset_engine(gt->i915) && !intel_gt_is_wedged(gt)) { + for_each_engine_masked(engine, gt->i915, engine_mask, tmp) { BUILD_BUG_ON(I915_RESET_MODESET >= I915_RESET_ENGINE); if (test_and_set_bit(I915_RESET_ENGINE + engine->id, - &error->flags)) + >->reset.flags)) continue; - if (i915_reset_engine(engine, msg) == 0) + if (intel_engine_reset(engine, msg) == 0) engine_mask &= ~engine->mask; - clear_bit(I915_RESET_ENGINE + engine->id, - &error->flags); - wake_up_bit(&error->flags, - I915_RESET_ENGINE + engine->id); + clear_and_wake_up_bit(I915_RESET_ENGINE + engine->id, + >->reset.flags); } } @@ -1250,9 +1236,9 @@ void i915_handle_error(struct drm_i915_private *i915, goto out; /* Full reset needs the mutex, stop any other user trying to do so. */ - if (test_and_set_bit(I915_RESET_BACKOFF, &error->flags)) { - wait_event(error->reset_queue, - !test_bit(I915_RESET_BACKOFF, &error->flags)); + if (test_and_set_bit(I915_RESET_BACKOFF, >->reset.flags)) { + wait_event(gt->reset.queue, + !test_bit(I915_RESET_BACKOFF, >->reset.flags)); goto out; /* piggy-back on the other reset */ } @@ -1260,113 +1246,119 @@ void i915_handle_error(struct drm_i915_private *i915, synchronize_rcu_expedited(); /* Prevent any other reset-engine attempt. */ - for_each_engine(engine, i915, tmp) { + for_each_engine(engine, gt->i915, tmp) { while (test_and_set_bit(I915_RESET_ENGINE + engine->id, - &error->flags)) - wait_on_bit(&error->flags, + >->reset.flags)) + wait_on_bit(>->reset.flags, I915_RESET_ENGINE + engine->id, TASK_UNINTERRUPTIBLE); } - i915_reset_device(i915, engine_mask, msg); + intel_gt_reset_global(gt, engine_mask, msg); - for_each_engine(engine, i915, tmp) { - clear_bit(I915_RESET_ENGINE + engine->id, - &error->flags); - } - - clear_bit(I915_RESET_BACKOFF, &error->flags); - wake_up_all(&error->reset_queue); + for_each_engine(engine, gt->i915, tmp) + clear_bit_unlock(I915_RESET_ENGINE + engine->id, + >->reset.flags); + clear_bit_unlock(I915_RESET_BACKOFF, >->reset.flags); + smp_mb__after_atomic(); + wake_up_all(>->reset.queue); out: - intel_runtime_pm_put(&i915->runtime_pm, wakeref); + intel_runtime_pm_put(>->i915->runtime_pm, wakeref); } -int i915_reset_trylock(struct drm_i915_private *i915) +int intel_gt_reset_trylock(struct intel_gt *gt) { - struct i915_gpu_error *error = &i915->gpu_error; int srcu; - might_lock(&error->reset_backoff_srcu); + might_lock(>->reset.backoff_srcu); might_sleep(); rcu_read_lock(); - while (test_bit(I915_RESET_BACKOFF, &error->flags)) { + while (test_bit(I915_RESET_BACKOFF, >->reset.flags)) { rcu_read_unlock(); - if (wait_event_interruptible(error->reset_queue, + if (wait_event_interruptible(gt->reset.queue, !test_bit(I915_RESET_BACKOFF, - &error->flags))) + >->reset.flags))) return -EINTR; rcu_read_lock(); } - srcu = srcu_read_lock(&error->reset_backoff_srcu); + srcu = srcu_read_lock(>->reset.backoff_srcu); rcu_read_unlock(); return srcu; } -void i915_reset_unlock(struct drm_i915_private *i915, int tag) -__releases(&i915->gpu_error.reset_backoff_srcu) +void intel_gt_reset_unlock(struct intel_gt *gt, int tag) +__releases(>->reset.backoff_srcu) { - struct i915_gpu_error *error = &i915->gpu_error; - - srcu_read_unlock(&error->reset_backoff_srcu, tag); + srcu_read_unlock(>->reset.backoff_srcu, tag); } -int i915_terminally_wedged(struct drm_i915_private *i915) +int intel_gt_terminally_wedged(struct intel_gt *gt) { - struct i915_gpu_error *error = &i915->gpu_error; - might_sleep(); - if (!__i915_wedged(error)) + if (!intel_gt_is_wedged(gt)) return 0; /* Reset still in progress? Maybe we will recover? */ - if (!test_bit(I915_RESET_BACKOFF, &error->flags)) + if (!test_bit(I915_RESET_BACKOFF, >->reset.flags)) return -EIO; /* XXX intel_reset_finish() still takes struct_mutex!!! */ - if (mutex_is_locked(&i915->drm.struct_mutex)) + if (mutex_is_locked(>->i915->drm.struct_mutex)) return -EAGAIN; - if (wait_event_interruptible(error->reset_queue, + if (wait_event_interruptible(gt->reset.queue, !test_bit(I915_RESET_BACKOFF, - &error->flags))) + >->reset.flags))) return -EINTR; - return __i915_wedged(error) ? -EIO : 0; + return intel_gt_is_wedged(gt) ? -EIO : 0; +} + +void intel_gt_init_reset(struct intel_gt *gt) +{ + init_waitqueue_head(>->reset.queue); + mutex_init(>->reset.mutex); + init_srcu_struct(>->reset.backoff_srcu); +} + +void intel_gt_fini_reset(struct intel_gt *gt) +{ + cleanup_srcu_struct(>->reset.backoff_srcu); } -static void i915_wedge_me(struct work_struct *work) +static void intel_wedge_me(struct work_struct *work) { - struct i915_wedge_me *w = container_of(work, typeof(*w), work.work); + struct intel_wedge_me *w = container_of(work, typeof(*w), work.work); - dev_err(w->i915->drm.dev, + dev_err(w->gt->i915->drm.dev, "%s timed out, cancelling all in-flight rendering.\n", w->name); - i915_gem_set_wedged(w->i915); + intel_gt_set_wedged(w->gt); } -void __i915_init_wedge(struct i915_wedge_me *w, - struct drm_i915_private *i915, - long timeout, - const char *name) +void __intel_init_wedge(struct intel_wedge_me *w, + struct intel_gt *gt, + long timeout, + const char *name) { - w->i915 = i915; + w->gt = gt; w->name = name; - INIT_DELAYED_WORK_ONSTACK(&w->work, i915_wedge_me); + INIT_DELAYED_WORK_ONSTACK(&w->work, intel_wedge_me); schedule_delayed_work(&w->work, timeout); } -void __i915_fini_wedge(struct i915_wedge_me *w) +void __intel_fini_wedge(struct intel_wedge_me *w) { cancel_delayed_work_sync(&w->work); destroy_delayed_work_on_stack(&w->work); - w->i915 = NULL; + w->gt = NULL; } #if IS_ENABLED(CONFIG_DRM_I915_SELFTEST) diff --git a/drivers/gpu/drm/i915/gt/intel_reset.h b/drivers/gpu/drm/i915/gt/intel_reset.h index 03fba0ab3868..37a987b17108 100644 --- a/drivers/gpu/drm/i915/gt/intel_reset.h +++ b/drivers/gpu/drm/i915/gt/intel_reset.h @@ -11,56 +11,67 @@ #include #include -#include "gt/intel_engine_types.h" +#include "intel_engine_types.h" +#include "intel_reset_types.h" struct drm_i915_private; struct i915_request; struct intel_engine_cs; +struct intel_gt; struct intel_guc; +void intel_gt_init_reset(struct intel_gt *gt); +void intel_gt_fini_reset(struct intel_gt *gt); + __printf(4, 5) -void i915_handle_error(struct drm_i915_private *i915, - intel_engine_mask_t engine_mask, - unsigned long flags, - const char *fmt, ...); +void intel_gt_handle_error(struct intel_gt *gt, + intel_engine_mask_t engine_mask, + unsigned long flags, + const char *fmt, ...); #define I915_ERROR_CAPTURE BIT(0) -void i915_reset(struct drm_i915_private *i915, - intel_engine_mask_t stalled_mask, - const char *reason); -int i915_reset_engine(struct intel_engine_cs *engine, - const char *reason); - -void i915_reset_request(struct i915_request *rq, bool guilty); +void intel_gt_reset(struct intel_gt *gt, + intel_engine_mask_t stalled_mask, + const char *reason); +int intel_engine_reset(struct intel_engine_cs *engine, + const char *reason); -int __must_check i915_reset_trylock(struct drm_i915_private *i915); -void i915_reset_unlock(struct drm_i915_private *i915, int tag); +void __i915_request_reset(struct i915_request *rq, bool guilty); -int i915_terminally_wedged(struct drm_i915_private *i915); +int __must_check intel_gt_reset_trylock(struct intel_gt *gt); +void intel_gt_reset_unlock(struct intel_gt *gt, int tag); -bool intel_has_gpu_reset(struct drm_i915_private *i915); -bool intel_has_reset_engine(struct drm_i915_private *i915); +void intel_gt_set_wedged(struct intel_gt *gt); +bool intel_gt_unset_wedged(struct intel_gt *gt); +int intel_gt_terminally_wedged(struct intel_gt *gt); -int intel_gpu_reset(struct drm_i915_private *i915, - intel_engine_mask_t engine_mask); +int __intel_gt_reset(struct intel_gt *gt, intel_engine_mask_t engine_mask); -int intel_reset_guc(struct drm_i915_private *i915); +int intel_reset_guc(struct intel_gt *gt); -struct i915_wedge_me { +struct intel_wedge_me { struct delayed_work work; - struct drm_i915_private *i915; + struct intel_gt *gt; const char *name; }; -void __i915_init_wedge(struct i915_wedge_me *w, - struct drm_i915_private *i915, - long timeout, - const char *name); -void __i915_fini_wedge(struct i915_wedge_me *w); +void __intel_init_wedge(struct intel_wedge_me *w, + struct intel_gt *gt, + long timeout, + const char *name); +void __intel_fini_wedge(struct intel_wedge_me *w); -#define i915_wedge_on_timeout(W, DEV, TIMEOUT) \ - for (__i915_init_wedge((W), (DEV), (TIMEOUT), __func__); \ - (W)->i915; \ - __i915_fini_wedge((W))) +#define intel_wedge_on_timeout(W, GT, TIMEOUT) \ + for (__intel_init_wedge((W), (GT), (TIMEOUT), __func__); \ + (W)->gt; \ + __intel_fini_wedge((W))) + +static inline bool __intel_reset_failed(const struct intel_reset *reset) +{ + return unlikely(test_bit(I915_WEDGED, &reset->flags)); +} + +bool intel_has_gpu_reset(struct drm_i915_private *i915); +bool intel_has_reset_engine(struct drm_i915_private *i915); #endif /* I915_RESET_H */ diff --git a/drivers/gpu/drm/i915/gt/intel_reset_types.h b/drivers/gpu/drm/i915/gt/intel_reset_types.h new file mode 100644 index 000000000000..31968356e0c0 --- /dev/null +++ b/drivers/gpu/drm/i915/gt/intel_reset_types.h @@ -0,0 +1,50 @@ +/* SPDX-License-Identifier: MIT */ +/* + * Copyright © 2019 Intel Corporation + */ + +#ifndef __INTEL_RESET_TYPES_H_ +#define __INTEL_RESET_TYPES_H_ + +#include +#include +#include + +struct intel_reset { + /** + * flags: Control various stages of the GPU reset + * + * #I915_RESET_BACKOFF - When we start a global reset, we need to + * serialise with any other users attempting to do the same, and + * any global resources that may be clobber by the reset (such as + * FENCE registers). + * + * #I915_RESET_ENGINE[num_engines] - Since the driver doesn't need to + * acquire the struct_mutex to reset an engine, we need an explicit + * flag to prevent two concurrent reset attempts in the same engine. + * As the number of engines continues to grow, allocate the flags from + * the most significant bits. + * + * #I915_WEDGED - If reset fails and we can no longer use the GPU, + * we set the #I915_WEDGED bit. Prior to command submission, e.g. + * i915_request_alloc(), this bit is checked and the sequence + * aborted (with -EIO reported to userspace) if set. + */ + unsigned long flags; +#define I915_RESET_BACKOFF 0 +#define I915_RESET_MODESET 1 +#define I915_RESET_ENGINE 2 +#define I915_WEDGED (BITS_PER_LONG - 1) + + struct mutex mutex; /* serialises wedging/unwedging */ + + /** + * Waitqueue to signal when the reset has completed. Used by clients + * that wait for dev_priv->mm.wedged to settle. + */ + wait_queue_head_t queue; + + struct srcu_struct backoff_srcu; +}; + +#endif /* _INTEL_RESET_TYPES_H_ */ diff --git a/drivers/gpu/drm/i915/gt/intel_ringbuffer.c b/drivers/gpu/drm/i915/gt/intel_ringbuffer.c index 2873276879cb..f1e571fa2e6d 100644 --- a/drivers/gpu/drm/i915/gt/intel_ringbuffer.c +++ b/drivers/gpu/drm/i915/gt/intel_ringbuffer.c @@ -795,7 +795,7 @@ static void reset_ring(struct intel_engine_cs *engine, bool stalled) * If the request was innocent, we try to replay the request * with the restored context. */ - i915_reset_request(rq, stalled); + __i915_request_reset(rq, stalled); GEM_BUG_ON(rq->ring != engine->buffer); head = rq->head; diff --git a/drivers/gpu/drm/i915/gt/selftest_hangcheck.c b/drivers/gpu/drm/i915/gt/selftest_hangcheck.c index 2d9cc3cd1f27..e2fa38a1ff0f 100644 --- a/drivers/gpu/drm/i915/gt/selftest_hangcheck.c +++ b/drivers/gpu/drm/i915/gt/selftest_hangcheck.c @@ -32,7 +32,6 @@ #include "selftests/i915_random.h" #include "selftests/igt_flush_test.h" #include "selftests/igt_reset.h" -#include "selftests/igt_wedge_me.h" #include "selftests/igt_atomic.h" #include "selftests/mock_drm.h" @@ -43,7 +42,6 @@ #define IGT_IDLE_TIMEOUT 50 /* ms; time to wait after flushing between tests */ struct hang { - struct drm_i915_private *i915; struct intel_gt *gt; struct drm_i915_gem_object *hws; struct drm_i915_gem_object *obj; @@ -52,27 +50,27 @@ struct hang { u32 *batch; }; -static int hang_init(struct hang *h, struct drm_i915_private *i915) +static int hang_init(struct hang *h, struct intel_gt *gt) { void *vaddr; int err; memset(h, 0, sizeof(*h)); - h->i915 = i915; + h->gt = gt; - h->ctx = kernel_context(i915); + h->ctx = kernel_context(gt->i915); if (IS_ERR(h->ctx)) return PTR_ERR(h->ctx); GEM_BUG_ON(i915_gem_context_is_bannable(h->ctx)); - h->hws = i915_gem_object_create_internal(i915, PAGE_SIZE); + h->hws = i915_gem_object_create_internal(gt->i915, PAGE_SIZE); if (IS_ERR(h->hws)) { err = PTR_ERR(h->hws); goto err_ctx; } - h->obj = i915_gem_object_create_internal(i915, PAGE_SIZE); + h->obj = i915_gem_object_create_internal(gt->i915, PAGE_SIZE); if (IS_ERR(h->obj)) { err = PTR_ERR(h->obj); goto err_hws; @@ -87,7 +85,7 @@ static int hang_init(struct hang *h, struct drm_i915_private *i915) h->seqno = memset(vaddr, 0xff, PAGE_SIZE); vaddr = i915_gem_object_pin_map(h->obj, - i915_coherent_map_type(i915)); + i915_coherent_map_type(gt->i915)); if (IS_ERR(vaddr)) { err = PTR_ERR(vaddr); goto err_unpin_hws; @@ -129,7 +127,7 @@ static int move_to_active(struct i915_vma *vma, static struct i915_request * hang_create_request(struct hang *h, struct intel_engine_cs *engine) { - struct drm_i915_private *i915 = h->i915; + struct intel_gt *gt = h->gt; struct i915_address_space *vm = h->ctx->vm ?: &engine->gt->ggtt->vm; struct drm_i915_gem_object *obj; struct i915_request *rq = NULL; @@ -139,11 +137,11 @@ hang_create_request(struct hang *h, struct intel_engine_cs *engine) u32 *batch; int err; - obj = i915_gem_object_create_internal(i915, PAGE_SIZE); + obj = i915_gem_object_create_internal(gt->i915, PAGE_SIZE); if (IS_ERR(obj)) return ERR_CAST(obj); - vaddr = i915_gem_object_pin_map(obj, i915_coherent_map_type(i915)); + vaddr = i915_gem_object_pin_map(obj, i915_coherent_map_type(gt->i915)); if (IS_ERR(vaddr)) { i915_gem_object_put(obj); return ERR_CAST(vaddr); @@ -186,7 +184,7 @@ hang_create_request(struct hang *h, struct intel_engine_cs *engine) goto cancel_rq; batch = h->batch; - if (INTEL_GEN(i915) >= 8) { + if (INTEL_GEN(gt->i915) >= 8) { *batch++ = MI_STORE_DWORD_IMM_GEN4; *batch++ = lower_32_bits(hws_address(hws, rq)); *batch++ = upper_32_bits(hws_address(hws, rq)); @@ -200,7 +198,7 @@ hang_create_request(struct hang *h, struct intel_engine_cs *engine) *batch++ = MI_BATCH_BUFFER_START | 1 << 8 | 1; *batch++ = lower_32_bits(vma->node.start); *batch++ = upper_32_bits(vma->node.start); - } else if (INTEL_GEN(i915) >= 6) { + } else if (INTEL_GEN(gt->i915) >= 6) { *batch++ = MI_STORE_DWORD_IMM_GEN4; *batch++ = 0; *batch++ = lower_32_bits(hws_address(hws, rq)); @@ -213,7 +211,7 @@ hang_create_request(struct hang *h, struct intel_engine_cs *engine) *batch++ = MI_ARB_CHECK; *batch++ = MI_BATCH_BUFFER_START | 1 << 8; *batch++ = lower_32_bits(vma->node.start); - } else if (INTEL_GEN(i915) >= 4) { + } else if (INTEL_GEN(gt->i915) >= 4) { *batch++ = MI_STORE_DWORD_IMM_GEN4 | MI_USE_GGTT; *batch++ = 0; *batch++ = lower_32_bits(hws_address(hws, rq)); @@ -249,7 +247,7 @@ hang_create_request(struct hang *h, struct intel_engine_cs *engine) } flags = 0; - if (INTEL_GEN(i915) <= 5) + if (INTEL_GEN(gt->i915) <= 5) flags |= I915_DISPATCH_SECURE; err = rq->engine->emit_bb_start(rq, vma->node.start, PAGE_SIZE, flags); @@ -286,7 +284,7 @@ static void hang_fini(struct hang *h) kernel_context_close(h->ctx); - igt_flush_test(h->i915, I915_WAIT_LOCKED); + igt_flush_test(h->gt->i915, I915_WAIT_LOCKED); } static bool wait_until_running(struct hang *h, struct i915_request *rq) @@ -301,7 +299,7 @@ static bool wait_until_running(struct hang *h, struct i915_request *rq) static int igt_hang_sanitycheck(void *arg) { - struct drm_i915_private *i915 = arg; + struct intel_gt *gt = arg; struct i915_request *rq; struct intel_engine_cs *engine; enum intel_engine_id id; @@ -310,13 +308,13 @@ static int igt_hang_sanitycheck(void *arg) /* Basic check that we can execute our hanging batch */ - mutex_lock(&i915->drm.struct_mutex); - err = hang_init(&h, i915); + mutex_lock(>->i915->drm.struct_mutex); + err = hang_init(&h, gt); if (err) goto unlock; - for_each_engine(engine, i915, id) { - struct igt_wedge_me w; + for_each_engine(engine, gt->i915, id) { + struct intel_wedge_me w; long timeout; if (!intel_engine_can_store_dword(engine)) @@ -338,10 +336,10 @@ static int igt_hang_sanitycheck(void *arg) i915_request_add(rq); timeout = 0; - igt_wedge_on_timeout(&w, i915, HZ / 10 /* 100ms timeout*/) + intel_wedge_on_timeout(&w, gt, HZ / 10 /* 100ms */) timeout = i915_request_wait(rq, 0, MAX_SCHEDULE_TIMEOUT); - if (i915_reset_failed(i915)) + if (intel_gt_is_wedged(gt)) timeout = -EIO; i915_request_put(rq); @@ -357,7 +355,7 @@ static int igt_hang_sanitycheck(void *arg) fini: hang_fini(&h); unlock: - mutex_unlock(&i915->drm.struct_mutex); + mutex_unlock(>->i915->drm.struct_mutex); return err; } @@ -368,7 +366,8 @@ static bool wait_for_idle(struct intel_engine_cs *engine) static int igt_reset_nop(void *arg) { - struct drm_i915_private *i915 = arg; + struct intel_gt *gt = arg; + struct i915_gpu_error *global = >->i915->gpu_error; struct intel_engine_cs *engine; struct i915_gem_context *ctx; unsigned int reset_count, count; @@ -379,25 +378,25 @@ static int igt_reset_nop(void *arg) /* Check that we can reset during non-user portions of requests */ - file = mock_file(i915); + file = mock_file(gt->i915); if (IS_ERR(file)) return PTR_ERR(file); - mutex_lock(&i915->drm.struct_mutex); - ctx = live_context(i915, file); - mutex_unlock(&i915->drm.struct_mutex); + mutex_lock(>->i915->drm.struct_mutex); + ctx = live_context(gt->i915, file); + mutex_unlock(>->i915->drm.struct_mutex); if (IS_ERR(ctx)) { err = PTR_ERR(ctx); goto out; } i915_gem_context_clear_bannable(ctx); - reset_count = i915_reset_count(&i915->gpu_error); + reset_count = i915_reset_count(global); count = 0; do { - mutex_lock(&i915->drm.struct_mutex); + mutex_lock(>->i915->drm.struct_mutex); - for_each_engine(engine, i915, id) { + for_each_engine(engine, gt->i915, id) { int i; for (i = 0; i < 16; i++) { @@ -413,43 +412,43 @@ static int igt_reset_nop(void *arg) } } - igt_global_reset_lock(i915); - i915_reset(i915, ALL_ENGINES, NULL); - igt_global_reset_unlock(i915); + igt_global_reset_lock(gt); + intel_gt_reset(gt, ALL_ENGINES, NULL); + igt_global_reset_unlock(gt); - mutex_unlock(&i915->drm.struct_mutex); - if (i915_reset_failed(i915)) { + mutex_unlock(>->i915->drm.struct_mutex); + if (intel_gt_is_wedged(gt)) { err = -EIO; break; } - if (i915_reset_count(&i915->gpu_error) != - reset_count + ++count) { + if (i915_reset_count(global) != reset_count + ++count) { pr_err("Full GPU reset not recorded!\n"); err = -EINVAL; break; } - err = igt_flush_test(i915, 0); + err = igt_flush_test(gt->i915, 0); if (err) break; } while (time_before(jiffies, end_time)); pr_info("%s: %d resets\n", __func__, count); - mutex_lock(&i915->drm.struct_mutex); - err = igt_flush_test(i915, I915_WAIT_LOCKED); - mutex_unlock(&i915->drm.struct_mutex); + mutex_lock(>->i915->drm.struct_mutex); + err = igt_flush_test(gt->i915, I915_WAIT_LOCKED); + mutex_unlock(>->i915->drm.struct_mutex); out: - mock_file_free(i915, file); - if (i915_reset_failed(i915)) + mock_file_free(gt->i915, file); + if (intel_gt_is_wedged(gt)) err = -EIO; return err; } static int igt_reset_nop_engine(void *arg) { - struct drm_i915_private *i915 = arg; + struct intel_gt *gt = arg; + struct i915_gpu_error *global = >->i915->gpu_error; struct intel_engine_cs *engine; struct i915_gem_context *ctx; enum intel_engine_id id; @@ -458,33 +457,32 @@ static int igt_reset_nop_engine(void *arg) /* Check that we can engine-reset during non-user portions */ - if (!intel_has_reset_engine(i915)) + if (!intel_has_reset_engine(gt->i915)) return 0; - file = mock_file(i915); + file = mock_file(gt->i915); if (IS_ERR(file)) return PTR_ERR(file); - mutex_lock(&i915->drm.struct_mutex); - ctx = live_context(i915, file); - mutex_unlock(&i915->drm.struct_mutex); + mutex_lock(>->i915->drm.struct_mutex); + ctx = live_context(gt->i915, file); + mutex_unlock(>->i915->drm.struct_mutex); if (IS_ERR(ctx)) { err = PTR_ERR(ctx); goto out; } i915_gem_context_clear_bannable(ctx); - for_each_engine(engine, i915, id) { + for_each_engine(engine, gt->i915, id) { unsigned int reset_count, reset_engine_count; unsigned int count; IGT_TIMEOUT(end_time); - reset_count = i915_reset_count(&i915->gpu_error); - reset_engine_count = i915_reset_engine_count(&i915->gpu_error, - engine); + reset_count = i915_reset_count(global); + reset_engine_count = i915_reset_engine_count(global, engine); count = 0; - set_bit(I915_RESET_ENGINE + id, &i915->gpu_error.flags); + set_bit(I915_RESET_ENGINE + id, >->reset.flags); do { int i; @@ -495,7 +493,7 @@ static int igt_reset_nop_engine(void *arg) break; } - mutex_lock(&i915->drm.struct_mutex); + mutex_lock(>->i915->drm.struct_mutex); for (i = 0; i < 16; i++) { struct i915_request *rq; @@ -507,20 +505,20 @@ static int igt_reset_nop_engine(void *arg) i915_request_add(rq); } - err = i915_reset_engine(engine, NULL); - mutex_unlock(&i915->drm.struct_mutex); + err = intel_engine_reset(engine, NULL); + mutex_unlock(>->i915->drm.struct_mutex); if (err) { pr_err("i915_reset_engine failed\n"); break; } - if (i915_reset_count(&i915->gpu_error) != reset_count) { + if (i915_reset_count(global) != reset_count) { pr_err("Full GPU reset recorded! (engine reset expected)\n"); err = -EINVAL; break; } - if (i915_reset_engine_count(&i915->gpu_error, engine) != + if (i915_reset_engine_count(global, engine) != reset_engine_count + ++count) { pr_err("%s engine reset not recorded!\n", engine->name); @@ -528,30 +526,31 @@ static int igt_reset_nop_engine(void *arg) break; } } while (time_before(jiffies, end_time)); - clear_bit(I915_RESET_ENGINE + id, &i915->gpu_error.flags); + clear_bit(I915_RESET_ENGINE + id, >->reset.flags); pr_info("%s(%s): %d resets\n", __func__, engine->name, count); if (err) break; - err = igt_flush_test(i915, 0); + err = igt_flush_test(gt->i915, 0); if (err) break; } - mutex_lock(&i915->drm.struct_mutex); - err = igt_flush_test(i915, I915_WAIT_LOCKED); - mutex_unlock(&i915->drm.struct_mutex); + mutex_lock(>->i915->drm.struct_mutex); + err = igt_flush_test(gt->i915, I915_WAIT_LOCKED); + mutex_unlock(>->i915->drm.struct_mutex); out: - mock_file_free(i915, file); - if (i915_reset_failed(i915)) + mock_file_free(gt->i915, file); + if (intel_gt_is_wedged(gt)) err = -EIO; return err; } -static int __igt_reset_engine(struct drm_i915_private *i915, bool active) +static int __igt_reset_engine(struct intel_gt *gt, bool active) { + struct i915_gpu_error *global = >->i915->gpu_error; struct intel_engine_cs *engine; enum intel_engine_id id; struct hang h; @@ -559,18 +558,18 @@ static int __igt_reset_engine(struct drm_i915_private *i915, bool active) /* Check that we can issue an engine reset on an idle engine (no-op) */ - if (!intel_has_reset_engine(i915)) + if (!intel_has_reset_engine(gt->i915)) return 0; if (active) { - mutex_lock(&i915->drm.struct_mutex); - err = hang_init(&h, i915); - mutex_unlock(&i915->drm.struct_mutex); + mutex_lock(>->i915->drm.struct_mutex); + err = hang_init(&h, gt); + mutex_unlock(>->i915->drm.struct_mutex); if (err) return err; } - for_each_engine(engine, i915, id) { + for_each_engine(engine, gt->i915, id) { unsigned int reset_count, reset_engine_count; IGT_TIMEOUT(end_time); @@ -584,30 +583,29 @@ static int __igt_reset_engine(struct drm_i915_private *i915, bool active) break; } - reset_count = i915_reset_count(&i915->gpu_error); - reset_engine_count = i915_reset_engine_count(&i915->gpu_error, - engine); + reset_count = i915_reset_count(global); + reset_engine_count = i915_reset_engine_count(global, engine); intel_engine_pm_get(engine); - set_bit(I915_RESET_ENGINE + id, &i915->gpu_error.flags); + set_bit(I915_RESET_ENGINE + id, >->reset.flags); do { if (active) { struct i915_request *rq; - mutex_lock(&i915->drm.struct_mutex); + mutex_lock(>->i915->drm.struct_mutex); rq = hang_create_request(&h, engine); if (IS_ERR(rq)) { err = PTR_ERR(rq); - mutex_unlock(&i915->drm.struct_mutex); + mutex_unlock(>->i915->drm.struct_mutex); break; } i915_request_get(rq); i915_request_add(rq); - mutex_unlock(&i915->drm.struct_mutex); + mutex_unlock(>->i915->drm.struct_mutex); if (!wait_until_running(&h, rq)) { - struct drm_printer p = drm_info_printer(i915->drm.dev); + struct drm_printer p = drm_info_printer(gt->i915->drm.dev); pr_err("%s: Failed to start request %llx, at %x\n", __func__, rq->fence.seqno, hws_seqno(&h, rq)); @@ -622,19 +620,19 @@ static int __igt_reset_engine(struct drm_i915_private *i915, bool active) i915_request_put(rq); } - err = i915_reset_engine(engine, NULL); + err = intel_engine_reset(engine, NULL); if (err) { pr_err("i915_reset_engine failed\n"); break; } - if (i915_reset_count(&i915->gpu_error) != reset_count) { + if (i915_reset_count(global) != reset_count) { pr_err("Full GPU reset recorded! (engine reset expected)\n"); err = -EINVAL; break; } - if (i915_reset_engine_count(&i915->gpu_error, engine) != + if (i915_reset_engine_count(global, engine) != ++reset_engine_count) { pr_err("%s engine reset not recorded!\n", engine->name); @@ -642,24 +640,24 @@ static int __igt_reset_engine(struct drm_i915_private *i915, bool active) break; } } while (time_before(jiffies, end_time)); - clear_bit(I915_RESET_ENGINE + id, &i915->gpu_error.flags); + clear_bit(I915_RESET_ENGINE + id, >->reset.flags); intel_engine_pm_put(engine); if (err) break; - err = igt_flush_test(i915, 0); + err = igt_flush_test(gt->i915, 0); if (err) break; } - if (i915_reset_failed(i915)) + if (intel_gt_is_wedged(gt)) err = -EIO; if (active) { - mutex_lock(&i915->drm.struct_mutex); + mutex_lock(>->i915->drm.struct_mutex); hang_fini(&h); - mutex_unlock(&i915->drm.struct_mutex); + mutex_unlock(>->i915->drm.struct_mutex); } return err; @@ -701,7 +699,7 @@ static int active_request_put(struct i915_request *rq) rq->fence.seqno); GEM_TRACE_DUMP(); - i915_gem_set_wedged(rq->i915); + intel_gt_set_wedged(rq->engine->gt); err = -EIO; } @@ -778,10 +776,11 @@ err_file: return err; } -static int __igt_reset_engines(struct drm_i915_private *i915, +static int __igt_reset_engines(struct intel_gt *gt, const char *test_name, unsigned int flags) { + struct i915_gpu_error *global = >->i915->gpu_error; struct intel_engine_cs *engine, *other; enum intel_engine_id id, tmp; struct hang h; @@ -791,13 +790,13 @@ static int __igt_reset_engines(struct drm_i915_private *i915, * with any other engine. */ - if (!intel_has_reset_engine(i915)) + if (!intel_has_reset_engine(gt->i915)) return 0; if (flags & TEST_ACTIVE) { - mutex_lock(&i915->drm.struct_mutex); - err = hang_init(&h, i915); - mutex_unlock(&i915->drm.struct_mutex); + mutex_lock(>->i915->drm.struct_mutex); + err = hang_init(&h, gt); + mutex_unlock(>->i915->drm.struct_mutex); if (err) return err; @@ -805,9 +804,9 @@ static int __igt_reset_engines(struct drm_i915_private *i915, h.ctx->sched.priority = 1024; } - for_each_engine(engine, i915, id) { + for_each_engine(engine, gt->i915, id) { struct active_engine threads[I915_NUM_ENGINES] = {}; - unsigned long global = i915_reset_count(&i915->gpu_error); + unsigned long device = i915_reset_count(global); unsigned long count = 0, reported; IGT_TIMEOUT(end_time); @@ -823,12 +822,11 @@ static int __igt_reset_engines(struct drm_i915_private *i915, } memset(threads, 0, sizeof(threads)); - for_each_engine(other, i915, tmp) { + for_each_engine(other, gt->i915, tmp) { struct task_struct *tsk; threads[tmp].resets = - i915_reset_engine_count(&i915->gpu_error, - other); + i915_reset_engine_count(global, other); if (!(flags & TEST_OTHERS)) continue; @@ -851,25 +849,25 @@ static int __igt_reset_engines(struct drm_i915_private *i915, } intel_engine_pm_get(engine); - set_bit(I915_RESET_ENGINE + id, &i915->gpu_error.flags); + set_bit(I915_RESET_ENGINE + id, >->reset.flags); do { struct i915_request *rq = NULL; if (flags & TEST_ACTIVE) { - mutex_lock(&i915->drm.struct_mutex); + mutex_lock(>->i915->drm.struct_mutex); rq = hang_create_request(&h, engine); if (IS_ERR(rq)) { err = PTR_ERR(rq); - mutex_unlock(&i915->drm.struct_mutex); + mutex_unlock(>->i915->drm.struct_mutex); break; } i915_request_get(rq); i915_request_add(rq); - mutex_unlock(&i915->drm.struct_mutex); + mutex_unlock(>->i915->drm.struct_mutex); if (!wait_until_running(&h, rq)) { - struct drm_printer p = drm_info_printer(i915->drm.dev); + struct drm_printer p = drm_info_printer(gt->i915->drm.dev); pr_err("%s: Failed to start request %llx, at %x\n", __func__, rq->fence.seqno, hws_seqno(&h, rq)); @@ -882,7 +880,7 @@ static int __igt_reset_engines(struct drm_i915_private *i915, } } - err = i915_reset_engine(engine, NULL); + err = intel_engine_reset(engine, NULL); if (err) { pr_err("i915_reset_engine(%s:%s): failed, err=%d\n", engine->name, test_name, err); @@ -894,7 +892,7 @@ static int __igt_reset_engines(struct drm_i915_private *i915, if (rq) { if (i915_request_wait(rq, 0, HZ / 5) < 0) { struct drm_printer p = - drm_info_printer(i915->drm.dev); + drm_info_printer(gt->i915->drm.dev); pr_err("i915_reset_engine(%s:%s):" " failed to complete request after reset\n", @@ -904,7 +902,7 @@ static int __igt_reset_engines(struct drm_i915_private *i915, i915_request_put(rq); GEM_TRACE_DUMP(); - i915_gem_set_wedged(i915); + intel_gt_set_wedged(gt); err = -EIO; break; } @@ -914,7 +912,7 @@ static int __igt_reset_engines(struct drm_i915_private *i915, if (!(flags & TEST_SELF) && !wait_for_idle(engine)) { struct drm_printer p = - drm_info_printer(i915->drm.dev); + drm_info_printer(gt->i915->drm.dev); pr_err("i915_reset_engine(%s:%s):" " failed to idle after reset\n", @@ -926,12 +924,12 @@ static int __igt_reset_engines(struct drm_i915_private *i915, break; } } while (time_before(jiffies, end_time)); - clear_bit(I915_RESET_ENGINE + id, &i915->gpu_error.flags); + clear_bit(I915_RESET_ENGINE + id, >->reset.flags); intel_engine_pm_put(engine); pr_info("i915_reset_engine(%s:%s): %lu resets\n", engine->name, test_name, count); - reported = i915_reset_engine_count(&i915->gpu_error, engine); + reported = i915_reset_engine_count(global, engine); reported -= threads[engine->id].resets; if (reported != count) { pr_err("i915_reset_engine(%s:%s): reset %lu times, but reported %lu\n", @@ -941,7 +939,7 @@ static int __igt_reset_engines(struct drm_i915_private *i915, } unwind: - for_each_engine(other, i915, tmp) { + for_each_engine(other, gt->i915, tmp) { int ret; if (!threads[tmp].task) @@ -956,22 +954,21 @@ unwind: } put_task_struct(threads[tmp].task); - if (other != engine && + if (other->uabi_class != engine->uabi_class && threads[tmp].resets != - i915_reset_engine_count(&i915->gpu_error, other)) { + i915_reset_engine_count(global, other)) { pr_err("Innocent engine %s was reset (count=%ld)\n", other->name, - i915_reset_engine_count(&i915->gpu_error, - other) - + i915_reset_engine_count(global, other) - threads[tmp].resets); if (!err) err = -EINVAL; } } - if (global != i915_reset_count(&i915->gpu_error)) { + if (device != i915_reset_count(global)) { pr_err("Global reset (count=%ld)!\n", - i915_reset_count(&i915->gpu_error) - global); + i915_reset_count(global) - device); if (!err) err = -EINVAL; } @@ -979,20 +976,20 @@ unwind: if (err) break; - mutex_lock(&i915->drm.struct_mutex); - err = igt_flush_test(i915, I915_WAIT_LOCKED); - mutex_unlock(&i915->drm.struct_mutex); + mutex_lock(>->i915->drm.struct_mutex); + err = igt_flush_test(gt->i915, I915_WAIT_LOCKED); + mutex_unlock(>->i915->drm.struct_mutex); if (err) break; } - if (i915_reset_failed(i915)) + if (intel_gt_is_wedged(gt)) err = -EIO; if (flags & TEST_ACTIVE) { - mutex_lock(&i915->drm.struct_mutex); + mutex_lock(>->i915->drm.struct_mutex); hang_fini(&h); - mutex_unlock(&i915->drm.struct_mutex); + mutex_unlock(>->i915->drm.struct_mutex); } return err; @@ -1018,13 +1015,13 @@ static int igt_reset_engines(void *arg) }, { } }; - struct drm_i915_private *i915 = arg; + struct intel_gt *gt = arg; typeof(*phases) *p; int err; for (p = phases; p->name; p++) { if (p->flags & TEST_PRIORITY) { - if (!(i915->caps.scheduler & I915_SCHEDULER_CAP_PRIORITY)) + if (!(gt->i915->caps.scheduler & I915_SCHEDULER_CAP_PRIORITY)) continue; } @@ -1036,38 +1033,39 @@ static int igt_reset_engines(void *arg) return 0; } -static u32 fake_hangcheck(struct drm_i915_private *i915, - intel_engine_mask_t mask) +static u32 fake_hangcheck(struct intel_gt *gt, intel_engine_mask_t mask) { - u32 count = i915_reset_count(&i915->gpu_error); + u32 count = i915_reset_count(>->i915->gpu_error); - i915_reset(i915, mask, NULL); + intel_gt_reset(gt, mask, NULL); return count; } static int igt_reset_wait(void *arg) { - struct drm_i915_private *i915 = arg; + struct intel_gt *gt = arg; + struct i915_gpu_error *global = >->i915->gpu_error; + struct intel_engine_cs *engine = gt->i915->engine[RCS0]; struct i915_request *rq; unsigned int reset_count; struct hang h; long timeout; int err; - if (!intel_engine_can_store_dword(i915->engine[RCS0])) + if (!engine || !intel_engine_can_store_dword(engine)) return 0; /* Check that we detect a stuck waiter and issue a reset */ - igt_global_reset_lock(i915); + igt_global_reset_lock(gt); - mutex_lock(&i915->drm.struct_mutex); - err = hang_init(&h, i915); + mutex_lock(>->i915->drm.struct_mutex); + err = hang_init(&h, gt); if (err) goto unlock; - rq = hang_create_request(&h, i915->engine[RCS0]); + rq = hang_create_request(&h, engine); if (IS_ERR(rq)) { err = PTR_ERR(rq); goto fini; @@ -1077,19 +1075,19 @@ static int igt_reset_wait(void *arg) i915_request_add(rq); if (!wait_until_running(&h, rq)) { - struct drm_printer p = drm_info_printer(i915->drm.dev); + struct drm_printer p = drm_info_printer(gt->i915->drm.dev); pr_err("%s: Failed to start request %llx, at %x\n", __func__, rq->fence.seqno, hws_seqno(&h, rq)); intel_engine_dump(rq->engine, &p, "%s\n", rq->engine->name); - i915_gem_set_wedged(i915); + intel_gt_set_wedged(gt); err = -EIO; goto out_rq; } - reset_count = fake_hangcheck(i915, ALL_ENGINES); + reset_count = fake_hangcheck(gt, ALL_ENGINES); timeout = i915_request_wait(rq, 0, 10); if (timeout < 0) { @@ -1099,7 +1097,7 @@ static int igt_reset_wait(void *arg) goto out_rq; } - if (i915_reset_count(&i915->gpu_error) == reset_count) { + if (i915_reset_count(global) == reset_count) { pr_err("No GPU reset recorded!\n"); err = -EINVAL; goto out_rq; @@ -1110,10 +1108,10 @@ out_rq: fini: hang_fini(&h); unlock: - mutex_unlock(&i915->drm.struct_mutex); - igt_global_reset_unlock(i915); + mutex_unlock(>->i915->drm.struct_mutex); + igt_global_reset_unlock(gt); - if (i915_reset_failed(i915)) + if (intel_gt_is_wedged(gt)) return -EIO; return err; @@ -1172,11 +1170,12 @@ out_unlock: return err; } -static int __igt_reset_evict_vma(struct drm_i915_private *i915, +static int __igt_reset_evict_vma(struct intel_gt *gt, struct i915_address_space *vm, int (*fn)(void *), unsigned int flags) { + struct intel_engine_cs *engine = gt->i915->engine[RCS0]; struct drm_i915_gem_object *obj; struct task_struct *tsk = NULL; struct i915_request *rq; @@ -1184,17 +1183,17 @@ static int __igt_reset_evict_vma(struct drm_i915_private *i915, struct hang h; int err; - if (!intel_engine_can_store_dword(i915->engine[RCS0])) + if (!engine || !intel_engine_can_store_dword(engine)) return 0; /* Check that we can recover an unbind stuck on a hanging request */ - mutex_lock(&i915->drm.struct_mutex); - err = hang_init(&h, i915); + mutex_lock(>->i915->drm.struct_mutex); + err = hang_init(&h, gt); if (err) goto unlock; - obj = i915_gem_object_create_internal(i915, SZ_1M); + obj = i915_gem_object_create_internal(gt->i915, SZ_1M); if (IS_ERR(obj)) { err = PTR_ERR(obj); goto fini; @@ -1214,7 +1213,7 @@ static int __igt_reset_evict_vma(struct drm_i915_private *i915, goto out_obj; } - rq = hang_create_request(&h, i915->engine[RCS0]); + rq = hang_create_request(&h, engine); if (IS_ERR(rq)) { err = PTR_ERR(rq); goto out_obj; @@ -1252,16 +1251,16 @@ static int __igt_reset_evict_vma(struct drm_i915_private *i915, if (err) goto out_rq; - mutex_unlock(&i915->drm.struct_mutex); + mutex_unlock(>->i915->drm.struct_mutex); if (!wait_until_running(&h, rq)) { - struct drm_printer p = drm_info_printer(i915->drm.dev); + struct drm_printer p = drm_info_printer(gt->i915->drm.dev); pr_err("%s: Failed to start request %llx, at %x\n", __func__, rq->fence.seqno, hws_seqno(&h, rq)); intel_engine_dump(rq->engine, &p, "%s\n", rq->engine->name); - i915_gem_set_wedged(i915); + intel_gt_set_wedged(gt); goto out_reset; } @@ -1278,31 +1277,31 @@ static int __igt_reset_evict_vma(struct drm_i915_private *i915, wait_for_completion(&arg.completion); if (wait_for(!list_empty(&rq->fence.cb_list), 10)) { - struct drm_printer p = drm_info_printer(i915->drm.dev); + struct drm_printer p = drm_info_printer(gt->i915->drm.dev); pr_err("igt/evict_vma kthread did not wait\n"); intel_engine_dump(rq->engine, &p, "%s\n", rq->engine->name); - i915_gem_set_wedged(i915); + intel_gt_set_wedged(gt); goto out_reset; } out_reset: - igt_global_reset_lock(i915); - fake_hangcheck(rq->i915, rq->engine->mask); - igt_global_reset_unlock(i915); + igt_global_reset_lock(gt); + fake_hangcheck(gt, rq->engine->mask); + igt_global_reset_unlock(gt); if (tsk) { - struct igt_wedge_me w; + struct intel_wedge_me w; /* The reset, even indirectly, should take less than 10ms. */ - igt_wedge_on_timeout(&w, i915, HZ / 10 /* 100ms timeout*/) + intel_wedge_on_timeout(&w, gt, HZ / 10 /* 100ms */) err = kthread_stop(tsk); put_task_struct(tsk); } - mutex_lock(&i915->drm.struct_mutex); + mutex_lock(>->i915->drm.struct_mutex); out_rq: i915_request_put(rq); out_obj: @@ -1310,9 +1309,9 @@ out_obj: fini: hang_fini(&h); unlock: - mutex_unlock(&i915->drm.struct_mutex); + mutex_unlock(>->i915->drm.struct_mutex); - if (i915_reset_failed(i915)) + if (intel_gt_is_wedged(gt)) return -EIO; return err; @@ -1320,26 +1319,26 @@ unlock: static int igt_reset_evict_ggtt(void *arg) { - struct drm_i915_private *i915 = arg; + struct intel_gt *gt = arg; - return __igt_reset_evict_vma(i915, &i915->ggtt.vm, + return __igt_reset_evict_vma(gt, >->ggtt->vm, evict_vma, EXEC_OBJECT_WRITE); } static int igt_reset_evict_ppgtt(void *arg) { - struct drm_i915_private *i915 = arg; + struct intel_gt *gt = arg; struct i915_gem_context *ctx; struct drm_file *file; int err; - file = mock_file(i915); + file = mock_file(gt->i915); if (IS_ERR(file)) return PTR_ERR(file); - mutex_lock(&i915->drm.struct_mutex); - ctx = live_context(i915, file); - mutex_unlock(&i915->drm.struct_mutex); + mutex_lock(>->i915->drm.struct_mutex); + ctx = live_context(gt->i915, file); + mutex_unlock(>->i915->drm.struct_mutex); if (IS_ERR(ctx)) { err = PTR_ERR(ctx); goto out; @@ -1347,29 +1346,29 @@ static int igt_reset_evict_ppgtt(void *arg) err = 0; if (ctx->vm) /* aliasing == global gtt locking, covered above */ - err = __igt_reset_evict_vma(i915, ctx->vm, + err = __igt_reset_evict_vma(gt, ctx->vm, evict_vma, EXEC_OBJECT_WRITE); out: - mock_file_free(i915, file); + mock_file_free(gt->i915, file); return err; } static int igt_reset_evict_fence(void *arg) { - struct drm_i915_private *i915 = arg; + struct intel_gt *gt = arg; - return __igt_reset_evict_vma(i915, &i915->ggtt.vm, + return __igt_reset_evict_vma(gt, >->ggtt->vm, evict_fence, EXEC_OBJECT_NEEDS_FENCE); } -static int wait_for_others(struct drm_i915_private *i915, +static int wait_for_others(struct intel_gt *gt, struct intel_engine_cs *exclude) { struct intel_engine_cs *engine; enum intel_engine_id id; - for_each_engine(engine, i915, id) { + for_each_engine(engine, gt->i915, id) { if (engine == exclude) continue; @@ -1382,7 +1381,8 @@ static int wait_for_others(struct drm_i915_private *i915, static int igt_reset_queue(void *arg) { - struct drm_i915_private *i915 = arg; + struct intel_gt *gt = arg; + struct i915_gpu_error *global = >->i915->gpu_error; struct intel_engine_cs *engine; enum intel_engine_id id; struct hang h; @@ -1390,14 +1390,14 @@ static int igt_reset_queue(void *arg) /* Check that we replay pending requests following a hang */ - igt_global_reset_lock(i915); + igt_global_reset_lock(gt); - mutex_lock(&i915->drm.struct_mutex); - err = hang_init(&h, i915); + mutex_lock(>->i915->drm.struct_mutex); + err = hang_init(&h, gt); if (err) goto unlock; - for_each_engine(engine, i915, id) { + for_each_engine(engine, gt->i915, id) { struct i915_request *prev; IGT_TIMEOUT(end_time); unsigned int count; @@ -1438,7 +1438,7 @@ static int igt_reset_queue(void *arg) * (hangcheck), or we focus on resetting just one * engine and so avoid repeatedly resetting innocents. */ - err = wait_for_others(i915, engine); + err = wait_for_others(gt, engine); if (err) { pr_err("%s(%s): Failed to idle other inactive engines after device reset\n", __func__, engine->name); @@ -1446,12 +1446,12 @@ static int igt_reset_queue(void *arg) i915_request_put(prev); GEM_TRACE_DUMP(); - i915_gem_set_wedged(i915); + intel_gt_set_wedged(gt); goto fini; } if (!wait_until_running(&h, prev)) { - struct drm_printer p = drm_info_printer(i915->drm.dev); + struct drm_printer p = drm_info_printer(gt->i915->drm.dev); pr_err("%s(%s): Failed to start request %llx, at %x\n", __func__, engine->name, @@ -1462,13 +1462,13 @@ static int igt_reset_queue(void *arg) i915_request_put(rq); i915_request_put(prev); - i915_gem_set_wedged(i915); + intel_gt_set_wedged(gt); err = -EIO; goto fini; } - reset_count = fake_hangcheck(i915, BIT(id)); + reset_count = fake_hangcheck(gt, BIT(id)); if (prev->fence.error != -EIO) { pr_err("GPU reset not recorded on hanging request [fence.error=%d]!\n", @@ -1488,7 +1488,7 @@ static int igt_reset_queue(void *arg) goto fini; } - if (i915_reset_count(&i915->gpu_error) == reset_count) { + if (i915_reset_count(global) == reset_count) { pr_err("No GPU reset recorded!\n"); i915_request_put(rq); i915_request_put(prev); @@ -1507,7 +1507,7 @@ static int igt_reset_queue(void *arg) i915_request_put(prev); - err = igt_flush_test(i915, I915_WAIT_LOCKED); + err = igt_flush_test(gt->i915, I915_WAIT_LOCKED); if (err) break; } @@ -1515,10 +1515,10 @@ static int igt_reset_queue(void *arg) fini: hang_fini(&h); unlock: - mutex_unlock(&i915->drm.struct_mutex); - igt_global_reset_unlock(i915); + mutex_unlock(>->i915->drm.struct_mutex); + igt_global_reset_unlock(gt); - if (i915_reset_failed(i915)) + if (intel_gt_is_wedged(gt)) return -EIO; return err; @@ -1526,8 +1526,9 @@ unlock: static int igt_handle_error(void *arg) { - struct drm_i915_private *i915 = arg; - struct intel_engine_cs *engine = i915->engine[RCS0]; + struct intel_gt *gt = arg; + struct i915_gpu_error *global = >->i915->gpu_error; + struct intel_engine_cs *engine = gt->i915->engine[RCS0]; struct hang h; struct i915_request *rq; struct i915_gpu_state *error; @@ -1535,15 +1536,15 @@ static int igt_handle_error(void *arg) /* Check that we can issue a global GPU and engine reset */ - if (!intel_has_reset_engine(i915)) + if (!intel_has_reset_engine(gt->i915)) return 0; if (!engine || !intel_engine_can_store_dword(engine)) return 0; - mutex_lock(&i915->drm.struct_mutex); + mutex_lock(>->i915->drm.struct_mutex); - err = hang_init(&h, i915); + err = hang_init(&h, gt); if (err) goto err_unlock; @@ -1557,28 +1558,28 @@ static int igt_handle_error(void *arg) i915_request_add(rq); if (!wait_until_running(&h, rq)) { - struct drm_printer p = drm_info_printer(i915->drm.dev); + struct drm_printer p = drm_info_printer(gt->i915->drm.dev); pr_err("%s: Failed to start request %llx, at %x\n", __func__, rq->fence.seqno, hws_seqno(&h, rq)); intel_engine_dump(rq->engine, &p, "%s\n", rq->engine->name); - i915_gem_set_wedged(i915); + intel_gt_set_wedged(gt); err = -EIO; goto err_request; } - mutex_unlock(&i915->drm.struct_mutex); + mutex_unlock(>->i915->drm.struct_mutex); /* Temporarily disable error capture */ - error = xchg(&i915->gpu_error.first_error, (void *)-1); + error = xchg(&global->first_error, (void *)-1); - i915_handle_error(i915, engine->mask, 0, NULL); + intel_gt_handle_error(gt, engine->mask, 0, NULL); - xchg(&i915->gpu_error.first_error, error); + xchg(&global->first_error, error); - mutex_lock(&i915->drm.struct_mutex); + mutex_lock(>->i915->drm.struct_mutex); if (rq->fence.error != -EIO) { pr_err("Guilty request not identified!\n"); @@ -1591,7 +1592,7 @@ err_request: err_fini: hang_fini(&h); err_unlock: - mutex_unlock(&i915->drm.struct_mutex); + mutex_unlock(>->i915->drm.struct_mutex); return err; } @@ -1608,7 +1609,7 @@ static int __igt_atomic_reset_engine(struct intel_engine_cs *engine, tasklet_disable_nosync(t); p->critical_section_begin(); - err = i915_reset_engine(engine, NULL); + err = intel_engine_reset(engine, NULL); p->critical_section_end(); tasklet_enable(t); @@ -1623,7 +1624,6 @@ static int __igt_atomic_reset_engine(struct intel_engine_cs *engine, static int igt_atomic_reset_engine(struct intel_engine_cs *engine, const struct igt_atomic_section *p) { - struct drm_i915_private *i915 = engine->i915; struct i915_request *rq; struct hang h; int err; @@ -1632,7 +1632,7 @@ static int igt_atomic_reset_engine(struct intel_engine_cs *engine, if (err) return err; - err = hang_init(&h, i915); + err = hang_init(&h, engine->gt); if (err) return err; @@ -1651,16 +1651,16 @@ static int igt_atomic_reset_engine(struct intel_engine_cs *engine, pr_err("%s(%s): Failed to start request %llx, at %x\n", __func__, engine->name, rq->fence.seqno, hws_seqno(&h, rq)); - i915_gem_set_wedged(i915); + intel_gt_set_wedged(engine->gt); err = -EIO; } if (err == 0) { - struct igt_wedge_me w; + struct intel_wedge_me w; - igt_wedge_on_timeout(&w, i915, HZ / 20 /* 50ms timeout*/) + intel_wedge_on_timeout(&w, engine->gt, HZ / 20 /* 50ms */) i915_request_wait(rq, 0, MAX_SCHEDULE_TIMEOUT); - if (i915_reset_failed(i915)) + if (intel_gt_is_wedged(engine->gt)) err = -EIO; } @@ -1672,30 +1672,30 @@ out: static int igt_reset_engines_atomic(void *arg) { - struct drm_i915_private *i915 = arg; + struct intel_gt *gt = arg; const typeof(*igt_atomic_phases) *p; int err = 0; /* Check that the engines resets are usable from atomic context */ - if (!intel_has_reset_engine(i915)) + if (!intel_has_reset_engine(gt->i915)) return 0; - if (USES_GUC_SUBMISSION(i915)) + if (USES_GUC_SUBMISSION(gt->i915)) return 0; - igt_global_reset_lock(i915); - mutex_lock(&i915->drm.struct_mutex); + igt_global_reset_lock(gt); + mutex_lock(>->i915->drm.struct_mutex); /* Flush any requests before we get started and check basics */ - if (!igt_force_reset(i915)) + if (!igt_force_reset(gt)) goto unlock; for (p = igt_atomic_phases; p->name; p++) { struct intel_engine_cs *engine; enum intel_engine_id id; - for_each_engine(engine, i915, id) { + for_each_engine(engine, gt->i915, id) { err = igt_atomic_reset_engine(engine, p); if (err) goto out; @@ -1704,11 +1704,11 @@ static int igt_reset_engines_atomic(void *arg) out: /* As we poke around the guts, do a full reset before continuing. */ - igt_force_reset(i915); + igt_force_reset(gt); unlock: - mutex_unlock(&i915->drm.struct_mutex); - igt_global_reset_unlock(i915); + mutex_unlock(>->i915->drm.struct_mutex); + igt_global_reset_unlock(gt); return err; } @@ -1730,28 +1730,29 @@ int intel_hangcheck_live_selftests(struct drm_i915_private *i915) SUBTEST(igt_reset_evict_fence), SUBTEST(igt_handle_error), }; + struct intel_gt *gt = &i915->gt; intel_wakeref_t wakeref; bool saved_hangcheck; int err; - if (!intel_has_gpu_reset(i915)) + if (!intel_has_gpu_reset(gt->i915)) return 0; - if (i915_terminally_wedged(i915)) + if (intel_gt_is_wedged(gt)) return -EIO; /* we're long past hope of a successful reset */ - wakeref = intel_runtime_pm_get(&i915->runtime_pm); + wakeref = intel_runtime_pm_get(>->i915->runtime_pm); saved_hangcheck = fetch_and_zero(&i915_modparams.enable_hangcheck); - drain_delayed_work(&i915->gpu_error.hangcheck_work); /* flush param */ + drain_delayed_work(>->hangcheck.work); /* flush param */ - err = i915_live_subtests(tests, i915); + err = intel_gt_live_subtests(tests, gt); - mutex_lock(&i915->drm.struct_mutex); - igt_flush_test(i915, I915_WAIT_LOCKED); - mutex_unlock(&i915->drm.struct_mutex); + mutex_lock(>->i915->drm.struct_mutex); + igt_flush_test(gt->i915, I915_WAIT_LOCKED); + mutex_unlock(>->i915->drm.struct_mutex); i915_modparams.enable_hangcheck = saved_hangcheck; - intel_runtime_pm_put(&i915->runtime_pm, wakeref); + intel_runtime_pm_put(>->i915->runtime_pm, wakeref); return err; } diff --git a/drivers/gpu/drm/i915/gt/selftest_lrc.c b/drivers/gpu/drm/i915/gt/selftest_lrc.c index b9b881ab8e7c..678e9b2edf8d 100644 --- a/drivers/gpu/drm/i915/gt/selftest_lrc.c +++ b/drivers/gpu/drm/i915/gt/selftest_lrc.c @@ -55,7 +55,7 @@ static int live_sanitycheck(void *arg) if (!igt_wait_for_spinner(&spin, rq)) { GEM_TRACE("spinner failed to start\n"); GEM_TRACE_DUMP(); - i915_gem_set_wedged(i915); + intel_gt_set_wedged(&i915->gt); err = -EIO; goto err_ctx; } @@ -211,7 +211,7 @@ slice_semaphore_queue(struct intel_engine_cs *outer, pr_err("Failed to slice along semaphore chain of length (%d, %d)!\n", count, n); GEM_TRACE_DUMP(); - i915_gem_set_wedged(outer->i915); + intel_gt_set_wedged(outer->gt); err = -EIO; } @@ -445,7 +445,7 @@ static int live_busywait_preempt(void *arg) intel_engine_dump(engine, &p, "%s\n", engine->name); GEM_TRACE_DUMP(); - i915_gem_set_wedged(i915); + intel_gt_set_wedged(&i915->gt); err = -EIO; goto err_vma; } @@ -534,7 +534,7 @@ static int live_preempt(void *arg) if (!igt_wait_for_spinner(&spin_lo, rq)) { GEM_TRACE("lo spinner failed to start\n"); GEM_TRACE_DUMP(); - i915_gem_set_wedged(i915); + intel_gt_set_wedged(&i915->gt); err = -EIO; goto err_ctx_lo; } @@ -551,7 +551,7 @@ static int live_preempt(void *arg) if (!igt_wait_for_spinner(&spin_hi, rq)) { GEM_TRACE("hi spinner failed to start\n"); GEM_TRACE_DUMP(); - i915_gem_set_wedged(i915); + intel_gt_set_wedged(&i915->gt); err = -EIO; goto err_ctx_lo; } @@ -688,7 +688,7 @@ err_unlock: err_wedged: igt_spinner_end(&spin_hi); igt_spinner_end(&spin_lo); - i915_gem_set_wedged(i915); + intel_gt_set_wedged(&i915->gt); err = -EIO; goto err_ctx_lo; } @@ -824,7 +824,7 @@ err_unlock: err_wedged: igt_spinner_end(&b.spin); igt_spinner_end(&a.spin); - i915_gem_set_wedged(i915); + intel_gt_set_wedged(&i915->gt); err = -EIO; goto err_client_b; } @@ -934,7 +934,7 @@ err_unlock: err_wedged: igt_spinner_end(&b.spin); igt_spinner_end(&a.spin); - i915_gem_set_wedged(i915); + intel_gt_set_wedged(&i915->gt); err = -EIO; goto err_client_b; } @@ -1105,7 +1105,7 @@ err_unlock: err_wedged: for (i = 0; i < ARRAY_SIZE(client); i++) igt_spinner_end(&client[i].spin); - i915_gem_set_wedged(i915); + intel_gt_set_wedged(&i915->gt); err = -EIO; goto err_client_3; } @@ -1251,7 +1251,7 @@ err_unlock: err_wedged: igt_spinner_end(&hi.spin); igt_spinner_end(&lo.spin); - i915_gem_set_wedged(i915); + intel_gt_set_wedged(&i915->gt); err = -EIO; goto err_client_lo; } @@ -1310,7 +1310,7 @@ static int live_preempt_hang(void *arg) if (!igt_wait_for_spinner(&spin_lo, rq)) { GEM_TRACE("lo spinner failed to start\n"); GEM_TRACE_DUMP(); - i915_gem_set_wedged(i915); + intel_gt_set_wedged(&i915->gt); err = -EIO; goto err_ctx_lo; } @@ -1332,21 +1332,21 @@ static int live_preempt_hang(void *arg) HZ / 10)) { pr_err("Preemption did not occur within timeout!"); GEM_TRACE_DUMP(); - i915_gem_set_wedged(i915); + intel_gt_set_wedged(&i915->gt); err = -EIO; goto err_ctx_lo; } - set_bit(I915_RESET_ENGINE + id, &i915->gpu_error.flags); - i915_reset_engine(engine, NULL); - clear_bit(I915_RESET_ENGINE + id, &i915->gpu_error.flags); + set_bit(I915_RESET_ENGINE + id, &i915->gt.reset.flags); + intel_engine_reset(engine, NULL); + clear_bit(I915_RESET_ENGINE + id, &i915->gt.reset.flags); engine->execlists.preempt_hang.inject_hang = false; if (!igt_wait_for_spinner(&spin_hi, rq)) { GEM_TRACE("hi spinner failed to start\n"); GEM_TRACE_DUMP(); - i915_gem_set_wedged(i915); + intel_gt_set_wedged(&i915->gt); err = -EIO; goto err_ctx_lo; } @@ -1726,7 +1726,7 @@ static int nop_virtual_engine(struct drm_i915_private *i915, request[nc]->fence.context, request[nc]->fence.seqno); GEM_TRACE_DUMP(); - i915_gem_set_wedged(i915); + intel_gt_set_wedged(&i915->gt); break; } } @@ -1873,7 +1873,7 @@ static int mask_virtual_engine(struct drm_i915_private *i915, request[n]->fence.context, request[n]->fence.seqno); GEM_TRACE_DUMP(); - i915_gem_set_wedged(i915); + intel_gt_set_wedged(&i915->gt); err = -EIO; goto out; } @@ -2150,7 +2150,7 @@ int intel_execlists_live_selftests(struct drm_i915_private *i915) if (!HAS_EXECLISTS(i915)) return 0; - if (i915_terminally_wedged(i915)) + if (intel_gt_is_wedged(&i915->gt)) return 0; return i915_live_subtests(tests, i915); diff --git a/drivers/gpu/drm/i915/gt/selftest_reset.c b/drivers/gpu/drm/i915/gt/selftest_reset.c index 672e32e1ef95..00a4f60cdfd5 100644 --- a/drivers/gpu/drm/i915/gt/selftest_reset.c +++ b/drivers/gpu/drm/i915/gt/selftest_reset.c @@ -9,26 +9,29 @@ static int igt_global_reset(void *arg) { - struct drm_i915_private *i915 = arg; + struct intel_gt *gt = arg; unsigned int reset_count; + intel_wakeref_t wakeref; int err = 0; /* Check that we can issue a global GPU reset */ - igt_global_reset_lock(i915); + igt_global_reset_lock(gt); + wakeref = intel_runtime_pm_get(>->i915->runtime_pm); - reset_count = i915_reset_count(&i915->gpu_error); + reset_count = i915_reset_count(>->i915->gpu_error); - i915_reset(i915, ALL_ENGINES, NULL); + intel_gt_reset(gt, ALL_ENGINES, NULL); - if (i915_reset_count(&i915->gpu_error) == reset_count) { + if (i915_reset_count(>->i915->gpu_error) == reset_count) { pr_err("No GPU reset recorded!\n"); err = -EINVAL; } - igt_global_reset_unlock(i915); + intel_runtime_pm_put(>->i915->runtime_pm, wakeref); + igt_global_reset_unlock(gt); - if (i915_reset_failed(i915)) + if (intel_gt_is_wedged(gt)) err = -EIO; return err; @@ -36,72 +39,72 @@ static int igt_global_reset(void *arg) static int igt_wedged_reset(void *arg) { - struct drm_i915_private *i915 = arg; + struct intel_gt *gt = arg; intel_wakeref_t wakeref; /* Check that we can recover a wedged device with a GPU reset */ - igt_global_reset_lock(i915); - wakeref = intel_runtime_pm_get(&i915->runtime_pm); + igt_global_reset_lock(gt); + wakeref = intel_runtime_pm_get(>->i915->runtime_pm); - i915_gem_set_wedged(i915); + intel_gt_set_wedged(gt); - GEM_BUG_ON(!i915_reset_failed(i915)); - i915_reset(i915, ALL_ENGINES, NULL); + GEM_BUG_ON(!intel_gt_is_wedged(gt)); + intel_gt_reset(gt, ALL_ENGINES, NULL); - intel_runtime_pm_put(&i915->runtime_pm, wakeref); - igt_global_reset_unlock(i915); + intel_runtime_pm_put(>->i915->runtime_pm, wakeref); + igt_global_reset_unlock(gt); - return i915_reset_failed(i915) ? -EIO : 0; + return intel_gt_is_wedged(gt) ? -EIO : 0; } static int igt_atomic_reset(void *arg) { - struct drm_i915_private *i915 = arg; + struct intel_gt *gt = arg; const typeof(*igt_atomic_phases) *p; int err = 0; /* Check that the resets are usable from atomic context */ - intel_gt_pm_get(&i915->gt); - igt_global_reset_lock(i915); + intel_gt_pm_get(gt); + igt_global_reset_lock(gt); /* Flush any requests before we get started and check basics */ - if (!igt_force_reset(i915)) + if (!igt_force_reset(gt)) goto unlock; for (p = igt_atomic_phases; p->name; p++) { intel_engine_mask_t awake; - GEM_TRACE("intel_gpu_reset under %s\n", p->name); + GEM_TRACE("__intel_gt_reset under %s\n", p->name); - awake = reset_prepare(i915); + awake = reset_prepare(gt); p->critical_section_begin(); - err = intel_gpu_reset(i915, ALL_ENGINES); + err = __intel_gt_reset(gt, ALL_ENGINES); p->critical_section_end(); - reset_finish(i915, awake); + reset_finish(gt, awake); if (err) { - pr_err("intel_gpu_reset failed under %s\n", p->name); + pr_err("__intel_gt_reset failed under %s\n", p->name); break; } } /* As we poke around the guts, do a full reset before continuing. */ - igt_force_reset(i915); + igt_force_reset(gt); unlock: - igt_global_reset_unlock(i915); - intel_gt_pm_put(&i915->gt); + igt_global_reset_unlock(gt); + intel_gt_pm_put(gt); return err; } static int igt_atomic_engine_reset(void *arg) { - struct drm_i915_private *i915 = arg; + struct intel_gt *gt = arg; const typeof(*igt_atomic_phases) *p; struct intel_engine_cs *engine; enum intel_engine_id id; @@ -109,33 +112,33 @@ static int igt_atomic_engine_reset(void *arg) /* Check that the resets are usable from atomic context */ - if (!intel_has_reset_engine(i915)) + if (!intel_has_reset_engine(gt->i915)) return 0; - if (USES_GUC_SUBMISSION(i915)) + if (USES_GUC_SUBMISSION(gt->i915)) return 0; - intel_gt_pm_get(&i915->gt); - igt_global_reset_lock(i915); + intel_gt_pm_get(gt); + igt_global_reset_lock(gt); /* Flush any requests before we get started and check basics */ - if (!igt_force_reset(i915)) + if (!igt_force_reset(gt)) goto out_unlock; - for_each_engine(engine, i915, id) { + for_each_engine(engine, gt->i915, id) { tasklet_disable_nosync(&engine->execlists.tasklet); intel_engine_pm_get(engine); for (p = igt_atomic_phases; p->name; p++) { - GEM_TRACE("i915_reset_engine(%s) under %s\n", + GEM_TRACE("intel_engine_reset(%s) under %s\n", engine->name, p->name); p->critical_section_begin(); - err = i915_reset_engine(engine, NULL); + err = intel_engine_reset(engine, NULL); p->critical_section_end(); if (err) { - pr_err("i915_reset_engine(%s) failed under %s\n", + pr_err("intel_engine_reset(%s) failed under %s\n", engine->name, p->name); break; } @@ -148,11 +151,11 @@ static int igt_atomic_engine_reset(void *arg) } /* As we poke around the guts, do a full reset before continuing. */ - igt_force_reset(i915); + igt_force_reset(gt); out_unlock: - igt_global_reset_unlock(i915); - intel_gt_pm_put(&i915->gt); + igt_global_reset_unlock(gt); + intel_gt_pm_put(gt); return err; } @@ -165,17 +168,13 @@ int intel_reset_live_selftests(struct drm_i915_private *i915) SUBTEST(igt_atomic_reset), SUBTEST(igt_atomic_engine_reset), }; - intel_wakeref_t wakeref; - int err = 0; + struct intel_gt *gt = &i915->gt; - if (!intel_has_gpu_reset(i915)) + if (!intel_has_gpu_reset(gt->i915)) return 0; - if (i915_terminally_wedged(i915)) + if (intel_gt_is_wedged(gt)) return -EIO; /* we're long past hope of a successful reset */ - with_intel_runtime_pm(&i915->runtime_pm, wakeref) - err = i915_subtests(tests, i915); - - return err; + return intel_gt_live_subtests(tests, gt); } diff --git a/drivers/gpu/drm/i915/gt/selftest_timeline.c b/drivers/gpu/drm/i915/gt/selftest_timeline.c index eae3b1963bf7..f0a840030382 100644 --- a/drivers/gpu/drm/i915/gt/selftest_timeline.c +++ b/drivers/gpu/drm/i915/gt/selftest_timeline.c @@ -7,6 +7,7 @@ #include #include "gem/i915_gem_pm.h" +#include "intel_gt.h" #include "../selftests/i915_random.h" #include "../i915_selftest.h" @@ -836,7 +837,7 @@ int intel_timeline_live_selftests(struct drm_i915_private *i915) SUBTEST(live_hwsp_wrap), }; - if (i915_terminally_wedged(i915)) + if (intel_gt_is_wedged(&i915->gt)) return 0; return i915_live_subtests(tests, i915); diff --git a/drivers/gpu/drm/i915/gt/selftest_workarounds.c b/drivers/gpu/drm/i915/gt/selftest_workarounds.c index a0d8f1bfe0ad..ab147985fa74 100644 --- a/drivers/gpu/drm/i915/gt/selftest_workarounds.c +++ b/drivers/gpu/drm/i915/gt/selftest_workarounds.c @@ -12,7 +12,6 @@ #include "selftests/igt_flush_test.h" #include "selftests/igt_reset.h" #include "selftests/igt_spinner.h" -#include "selftests/igt_wedge_me.h" #include "selftests/mock_drm.h" #include "gem/selftests/igt_gem_utils.h" @@ -178,7 +177,7 @@ static int check_whitelist(struct i915_gem_context *ctx, struct intel_engine_cs *engine) { struct drm_i915_gem_object *results; - struct igt_wedge_me wedge; + struct intel_wedge_me wedge; u32 *vaddr; int err; int i; @@ -189,10 +188,10 @@ static int check_whitelist(struct i915_gem_context *ctx, err = 0; i915_gem_object_lock(results); - igt_wedge_on_timeout(&wedge, ctx->i915, HZ / 5) /* a safety net! */ + intel_wedge_on_timeout(&wedge, &ctx->i915->gt, HZ / 5) /* safety net! */ err = i915_gem_object_set_to_cpu_domain(results, false); i915_gem_object_unlock(results); - if (i915_terminally_wedged(ctx->i915)) + if (intel_gt_is_wedged(&ctx->i915->gt)) err = -EIO; if (err) goto out_put; @@ -225,13 +224,13 @@ out_put: static int do_device_reset(struct intel_engine_cs *engine) { - i915_reset(engine->i915, engine->mask, "live_workarounds"); + intel_gt_reset(engine->gt, engine->mask, "live_workarounds"); return 0; } static int do_engine_reset(struct intel_engine_cs *engine) { - return i915_reset_engine(engine, "live_workarounds"); + return intel_engine_reset(engine, "live_workarounds"); } static int @@ -572,7 +571,7 @@ err_request: if (i915_request_wait(rq, 0, HZ / 5) < 0) { pr_err("%s: Futzing %x timedout; cancelling test\n", engine->name, reg); - i915_gem_set_wedged(ctx->i915); + intel_gt_set_wedged(&ctx->i915->gt); err = -EIO; goto out_batch; } @@ -730,7 +729,7 @@ static int live_reset_whitelist(void *arg) if (!engine || engine->whitelist.count == 0) return 0; - igt_global_reset_lock(i915); + igt_global_reset_lock(&i915->gt); if (intel_has_reset_engine(i915)) { err = check_whitelist_across_reset(engine, @@ -749,7 +748,7 @@ static int live_reset_whitelist(void *arg) } out: - igt_global_reset_unlock(i915); + igt_global_reset_unlock(&i915->gt); return err; } @@ -1118,7 +1117,7 @@ live_gpu_reset_workarounds(void *arg) pr_info("Verifying after GPU reset...\n"); - igt_global_reset_lock(i915); + igt_global_reset_lock(&i915->gt); wakeref = intel_runtime_pm_get(&i915->runtime_pm); reference_lists_init(i915, &lists); @@ -1127,7 +1126,7 @@ live_gpu_reset_workarounds(void *arg) if (!ok) goto out; - i915_reset(i915, ALL_ENGINES, "live_workarounds"); + intel_gt_reset(&i915->gt, ALL_ENGINES, "live_workarounds"); ok = verify_wa_lists(ctx, &lists, "after reset"); @@ -1135,7 +1134,7 @@ out: kernel_context_close(ctx); reference_lists_fini(i915, &lists); intel_runtime_pm_put(&i915->runtime_pm, wakeref); - igt_global_reset_unlock(i915); + igt_global_reset_unlock(&i915->gt); return ok ? 0 : -ESRCH; } @@ -1160,7 +1159,7 @@ live_engine_reset_workarounds(void *arg) if (IS_ERR(ctx)) return PTR_ERR(ctx); - igt_global_reset_lock(i915); + igt_global_reset_lock(&i915->gt); wakeref = intel_runtime_pm_get(&i915->runtime_pm); reference_lists_init(i915, &lists); @@ -1176,7 +1175,7 @@ live_engine_reset_workarounds(void *arg) goto err; } - i915_reset_engine(engine, "live_workarounds"); + intel_engine_reset(engine, "live_workarounds"); ok = verify_wa_lists(ctx, &lists, "after idle reset"); if (!ok) { @@ -1204,7 +1203,7 @@ live_engine_reset_workarounds(void *arg) goto err; } - i915_reset_engine(engine, "live_workarounds"); + intel_engine_reset(engine, "live_workarounds"); igt_spinner_end(&spin); igt_spinner_fini(&spin); @@ -1219,7 +1218,7 @@ live_engine_reset_workarounds(void *arg) err: reference_lists_fini(i915, &lists); intel_runtime_pm_put(&i915->runtime_pm, wakeref); - igt_global_reset_unlock(i915); + igt_global_reset_unlock(&i915->gt); kernel_context_close(ctx); igt_flush_test(i915, I915_WAIT_LOCKED); @@ -1238,7 +1237,7 @@ int intel_workarounds_live_selftests(struct drm_i915_private *i915) }; int err; - if (i915_terminally_wedged(i915)) + if (intel_gt_is_wedged(&i915->gt)) return 0; mutex_lock(&i915->drm.struct_mutex); diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c index 41245acb0a0f..a1843e3de6d7 100644 --- a/drivers/gpu/drm/i915/i915_debugfs.c +++ b/drivers/gpu/drm/i915/i915_debugfs.c @@ -1074,15 +1074,16 @@ static void i915_instdone_info(struct drm_i915_private *dev_priv, static int i915_hangcheck_info(struct seq_file *m, void *unused) { - struct drm_i915_private *dev_priv = node_to_i915(m->private); + struct drm_i915_private *i915 = node_to_i915(m->private); + struct intel_gt *gt = &i915->gt; struct intel_engine_cs *engine; intel_wakeref_t wakeref; enum intel_engine_id id; - seq_printf(m, "Reset flags: %lx\n", dev_priv->gpu_error.flags); - if (test_bit(I915_WEDGED, &dev_priv->gpu_error.flags)) + seq_printf(m, "Reset flags: %lx\n", gt->reset.flags); + if (test_bit(I915_WEDGED, >->reset.flags)) seq_puts(m, "\tWedged\n"); - if (test_bit(I915_RESET_BACKOFF, &dev_priv->gpu_error.flags)) + if (test_bit(I915_RESET_BACKOFF, >->reset.flags)) seq_puts(m, "\tDevice (global) reset in progress\n"); if (!i915_modparams.enable_hangcheck) { @@ -1090,19 +1091,19 @@ static int i915_hangcheck_info(struct seq_file *m, void *unused) return 0; } - if (timer_pending(&dev_priv->gpu_error.hangcheck_work.timer)) + if (timer_pending(>->hangcheck.work.timer)) seq_printf(m, "Hangcheck active, timer fires in %dms\n", - jiffies_to_msecs(dev_priv->gpu_error.hangcheck_work.timer.expires - + jiffies_to_msecs(gt->hangcheck.work.timer.expires - jiffies)); - else if (delayed_work_pending(&dev_priv->gpu_error.hangcheck_work)) + else if (delayed_work_pending(>->hangcheck.work)) seq_puts(m, "Hangcheck active, work pending\n"); else seq_puts(m, "Hangcheck inactive\n"); - seq_printf(m, "GT active? %s\n", yesno(dev_priv->gt.awake)); + seq_printf(m, "GT active? %s\n", yesno(gt->awake)); - with_intel_runtime_pm(&dev_priv->runtime_pm, wakeref) { - for_each_engine(engine, dev_priv, id) { + with_intel_runtime_pm(&i915->runtime_pm, wakeref) { + for_each_engine(engine, i915, id) { struct intel_instdone instdone; seq_printf(m, "%s: %d ms ago\n", @@ -1117,10 +1118,10 @@ static int i915_hangcheck_info(struct seq_file *m, void *unused) intel_engine_get_instdone(engine, &instdone); seq_puts(m, "\tinstdone read =\n"); - i915_instdone_info(dev_priv, m, &instdone); + i915_instdone_info(i915, m, &instdone); seq_puts(m, "\tinstdone accu =\n"); - i915_instdone_info(dev_priv, m, + i915_instdone_info(i915, m, &engine->hangcheck.instdone); } } @@ -1128,23 +1129,6 @@ static int i915_hangcheck_info(struct seq_file *m, void *unused) return 0; } -static int i915_reset_info(struct seq_file *m, void *unused) -{ - struct drm_i915_private *dev_priv = node_to_i915(m->private); - struct i915_gpu_error *error = &dev_priv->gpu_error; - struct intel_engine_cs *engine; - enum intel_engine_id id; - - seq_printf(m, "full gpu reset = %u\n", i915_reset_count(error)); - - for_each_engine(engine, dev_priv, id) { - seq_printf(m, "%s = %u\n", engine->name, - i915_reset_engine_count(error, engine)); - } - - return 0; -} - static int ironlake_drpc_info(struct seq_file *m) { struct drm_i915_private *i915 = node_to_i915(m->private); @@ -3616,7 +3600,8 @@ static const struct file_operations i915_cur_wm_latency_fops = { static int i915_wedged_get(void *data, u64 *val) { - int ret = i915_terminally_wedged(data); + struct drm_i915_private *i915 = data; + int ret = intel_gt_terminally_wedged(&i915->gt); switch (ret) { case -EIO: @@ -3636,11 +3621,11 @@ i915_wedged_set(void *data, u64 val) struct drm_i915_private *i915 = data; /* Flush any previous reset before applying for a new one */ - wait_event(i915->gpu_error.reset_queue, - !test_bit(I915_RESET_BACKOFF, &i915->gpu_error.flags)); + wait_event(i915->gt.reset.queue, + !test_bit(I915_RESET_BACKOFF, &i915->gt.reset.flags)); - i915_handle_error(i915, val, I915_ERROR_CAPTURE, - "Manually set wedged engine mask = %llx", val); + intel_gt_handle_error(&i915->gt, val, I915_ERROR_CAPTURE, + "Manually set wedged engine mask = %llx", val); return 0; } @@ -3683,8 +3668,9 @@ i915_drop_caches_set(void *data, u64 val) val, val & DROP_ALL); if (val & DROP_RESET_ACTIVE && - wait_for(intel_engines_are_idle(i915), I915_IDLE_ENGINES_TIMEOUT)) - i915_gem_set_wedged(i915); + wait_for(intel_engines_are_idle(&i915->gt), + I915_IDLE_ENGINES_TIMEOUT)) + intel_gt_set_wedged(&i915->gt); /* No need to check and wait for gpu resets, only libdrm auto-restarts * on ioctls on -EAGAIN. */ @@ -3719,8 +3705,8 @@ i915_drop_caches_set(void *data, u64 val) mutex_unlock(&i915->drm.struct_mutex); } - if (val & DROP_RESET_ACTIVE && i915_terminally_wedged(i915)) - i915_handle_error(i915, ALL_ENGINES, 0, NULL); + if (val & DROP_RESET_ACTIVE && intel_gt_terminally_wedged(&i915->gt)) + intel_gt_handle_error(&i915->gt, ALL_ENGINES, 0, NULL); fs_reclaim_acquire(GFP_KERNEL); if (val & DROP_BOUND) @@ -4375,7 +4361,6 @@ static const struct drm_info_list i915_debugfs_list[] = { {"i915_huc_load_status", i915_huc_load_status_info, 0}, {"i915_frequency_info", i915_frequency_info, 0}, {"i915_hangcheck_info", i915_hangcheck_info, 0}, - {"i915_reset_info", i915_reset_info, 0}, {"i915_drpc_info", i915_drpc_info, 0}, {"i915_emon_status", i915_emon_status, 0}, {"i915_ring_freq_table", i915_ring_freq_table, 0}, diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index e2d1bed9454c..6953732719e2 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -947,7 +947,6 @@ static int i915_driver_early_probe(struct drm_i915_private *dev_priv) if (ret < 0) goto err_uc; intel_irq_init(dev_priv); - intel_hangcheck_init(dev_priv); intel_init_display_hooks(dev_priv); intel_init_clock_gating_hooks(dev_priv); intel_init_audio_hooks(dev_priv); @@ -1967,7 +1966,7 @@ void i915_driver_remove(struct drm_device *dev) * all in-flight requests so that we can quickly unbind the active * resources. */ - i915_gem_set_wedged(dev_priv); + intel_gt_set_wedged(&dev_priv->gt); /* Flush any external code that still may be under the RCU lock */ synchronize_rcu(); @@ -1988,7 +1987,7 @@ void i915_driver_remove(struct drm_device *dev) intel_csr_ucode_fini(dev_priv); /* Free error state after interrupts are fully disabled. */ - cancel_delayed_work_sync(&dev_priv->gpu_error.hangcheck_work); + cancel_delayed_work_sync(&dev_priv->gt.hangcheck.work); i915_reset_error_state(dev_priv); i915_gem_driver_remove(dev_priv); diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 7e1ee30d74a9..1a0b114da77e 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -2400,28 +2400,10 @@ int i915_driver_probe(struct pci_dev *pdev, const struct pci_device_id *ent); void i915_driver_remove(struct drm_device *dev); void intel_engine_init_hangcheck(struct intel_engine_cs *engine); -void intel_hangcheck_init(struct drm_i915_private *dev_priv); int vlv_force_gfx_clock(struct drm_i915_private *dev_priv, bool on); u32 intel_calculate_mcr_s_ss_select(struct drm_i915_private *dev_priv); -static inline void i915_queue_hangcheck(struct drm_i915_private *dev_priv) -{ - unsigned long delay; - - if (unlikely(!i915_modparams.enable_hangcheck)) - return; - - /* Don't continually defer the hangcheck so that it is always run at - * least once after work has been scheduled on any ring. Otherwise, - * we will ignore a hung ring if a second ring is kept busy. - */ - - delay = round_jiffies_up_relative(DRM_I915_HANGCHECK_JIFFIES); - queue_delayed_work(system_long_wq, - &dev_priv->gpu_error.hangcheck_work, delay); -} - static inline bool intel_gvt_active(struct drm_i915_private *dev_priv) { return dev_priv->gvt; @@ -2510,30 +2492,17 @@ void i915_gem_track_fb(struct drm_i915_gem_object *old, int __must_check i915_gem_set_global_seqno(struct drm_device *dev, u32 seqno); -static inline bool __i915_wedged(struct i915_gpu_error *error) -{ - return unlikely(test_bit(I915_WEDGED, &error->flags)); -} - -static inline bool i915_reset_failed(struct drm_i915_private *i915) -{ - return __i915_wedged(&i915->gpu_error); -} - static inline u32 i915_reset_count(struct i915_gpu_error *error) { - return READ_ONCE(error->reset_count); + return atomic_read(&error->reset_count); } static inline u32 i915_reset_engine_count(struct i915_gpu_error *error, struct intel_engine_cs *engine) { - return READ_ONCE(error->reset_engine_count[engine->id]); + return atomic_read(&error->reset_engine_count[engine->uabi_class]); } -void i915_gem_set_wedged(struct drm_i915_private *dev_priv); -bool i915_gem_unset_wedged(struct drm_i915_private *dev_priv); - void i915_gem_init_mmio(struct drm_i915_private *i915); int __must_check i915_gem_init(struct drm_i915_private *dev_priv); int __must_check i915_gem_init_hw(struct drm_i915_private *dev_priv); diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index 37fe2ed2f582..e24955b5ebc2 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -894,13 +894,13 @@ void i915_gem_runtime_suspend(struct drm_i915_private *i915) } } -static int wait_for_engines(struct drm_i915_private *i915) +static int wait_for_engines(struct intel_gt *gt) { - if (wait_for(intel_engines_are_idle(i915), I915_IDLE_ENGINES_TIMEOUT)) { - dev_err(i915->drm.dev, + if (wait_for(intel_engines_are_idle(gt), I915_IDLE_ENGINES_TIMEOUT)) { + dev_err(gt->i915->drm.dev, "Failed to idle engines, declaring wedged!\n"); GEM_TRACE_DUMP(); - i915_gem_set_wedged(i915); + intel_gt_set_wedged(gt); return -EIO; } @@ -971,7 +971,7 @@ int i915_gem_wait_for_idle(struct drm_i915_private *i915, lockdep_assert_held(&i915->drm.struct_mutex); - err = wait_for_engines(i915); + err = wait_for_engines(&i915->gt); if (err) return err; @@ -1149,8 +1149,8 @@ void i915_gem_sanitize(struct drm_i915_private *i915) * back to defaults, recovering from whatever wedged state we left it * in and so worth trying to use the device once more. */ - if (i915_terminally_wedged(i915)) - i915_gem_unset_wedged(i915); + if (intel_gt_is_wedged(&i915->gt)) + intel_gt_unset_wedged(&i915->gt); /* * If we inherit context state from the BIOS or earlier occupants @@ -1202,7 +1202,7 @@ int i915_gem_init_hw(struct drm_i915_private *i915) int ret; BUG_ON(!i915->kernel_context); - ret = i915_terminally_wedged(i915); + ret = intel_gt_terminally_wedged(gt); if (ret) return ret; @@ -1384,7 +1384,7 @@ err_active: * and ready to be torn-down. The quickest way we can accomplish * this is by declaring ourselves wedged. */ - i915_gem_set_wedged(i915); + intel_gt_set_wedged(&i915->gt); goto out_ctx; } @@ -1539,7 +1539,7 @@ int i915_gem_init(struct drm_i915_private *dev_priv) err_gt: mutex_unlock(&dev_priv->drm.struct_mutex); - i915_gem_set_wedged(dev_priv); + intel_gt_set_wedged(&dev_priv->gt); i915_gem_suspend(dev_priv); i915_gem_suspend_late(dev_priv); @@ -1581,10 +1581,10 @@ err_uc_misc: * wedged. But we only want to do this where the GPU is angry, * for all other failure, such as an allocation failure, bail. */ - if (!i915_reset_failed(dev_priv)) { + if (!intel_gt_is_wedged(&dev_priv->gt)) { i915_probe_error(dev_priv, "Failed to initialize GPU, declaring it wedged!\n"); - i915_gem_set_wedged(dev_priv); + intel_gt_set_wedged(&dev_priv->gt); } /* Minimal basic recovery for KMS */ @@ -1666,11 +1666,6 @@ int i915_gem_init_early(struct drm_i915_private *dev_priv) i915_gem_init__mm(dev_priv); i915_gem_init__pm(dev_priv); - init_waitqueue_head(&dev_priv->gpu_error.wait_queue); - init_waitqueue_head(&dev_priv->gpu_error.reset_queue); - mutex_init(&dev_priv->gpu_error.wedge_mutex); - init_srcu_struct(&dev_priv->gpu_error.reset_backoff_srcu); - atomic_set(&dev_priv->mm.bsd_engine_dispatch_index, 0); spin_lock_init(&dev_priv->fb_tracking.lock); @@ -1689,7 +1684,7 @@ void i915_gem_cleanup_early(struct drm_i915_private *dev_priv) GEM_BUG_ON(atomic_read(&dev_priv->mm.free_count)); WARN_ON(dev_priv->mm.shrink_count); - cleanup_srcu_struct(&dev_priv->gpu_error.reset_backoff_srcu); + intel_gt_cleanup_early(&dev_priv->gt); i915_gemfs_fini(dev_priv); } diff --git a/drivers/gpu/drm/i915/i915_gpu_error.h b/drivers/gpu/drm/i915/i915_gpu_error.h index 2ecd0c6a1c94..7dfbfda48733 100644 --- a/drivers/gpu/drm/i915/i915_gpu_error.h +++ b/drivers/gpu/drm/i915/i915_gpu_error.h @@ -7,6 +7,7 @@ #ifndef _I915_GPU_ERROR_H_ #define _I915_GPU_ERROR_H_ +#include #include #include #include @@ -180,12 +181,6 @@ struct i915_gpu_state { }; struct i915_gpu_error { - /* For hangcheck timer */ -#define DRM_I915_HANGCHECK_PERIOD 1500 /* in ms */ -#define DRM_I915_HANGCHECK_JIFFIES msecs_to_jiffies(DRM_I915_HANGCHECK_PERIOD) - - struct delayed_work hangcheck_work; - /* For reset and error_state handling. */ spinlock_t lock; /* Protected by the above dev->gpu_error.lock. */ @@ -193,52 +188,11 @@ struct i915_gpu_error { atomic_t pending_fb_pin; - /** - * flags: Control various stages of the GPU reset - * - * #I915_RESET_BACKOFF - When we start a global reset, we need to - * serialise with any other users attempting to do the same, and - * any global resources that may be clobber by the reset (such as - * FENCE registers). - * - * #I915_RESET_ENGINE[num_engines] - Since the driver doesn't need to - * acquire the struct_mutex to reset an engine, we need an explicit - * flag to prevent two concurrent reset attempts in the same engine. - * As the number of engines continues to grow, allocate the flags from - * the most significant bits. - * - * #I915_WEDGED - If reset fails and we can no longer use the GPU, - * we set the #I915_WEDGED bit. Prior to command submission, e.g. - * i915_request_alloc(), this bit is checked and the sequence - * aborted (with -EIO reported to userspace) if set. - */ - unsigned long flags; -#define I915_RESET_BACKOFF 0 -#define I915_RESET_MODESET 1 -#define I915_RESET_ENGINE 2 -#define I915_WEDGED (BITS_PER_LONG - 1) - /** Number of times the device has been reset (global) */ - u32 reset_count; + atomic_t reset_count; /** Number of times an engine has been reset */ - u32 reset_engine_count[I915_NUM_ENGINES]; - - struct mutex wedge_mutex; /* serialises wedging/unwedging */ - - /** - * Waitqueue to signal when a hang is detected. Used to for waiters - * to release the struct_mutex for the reset to procede. - */ - wait_queue_head_t wait_queue; - - /** - * Waitqueue to signal when the reset has completed. Used by clients - * that wait for dev_priv->mm.wedged to settle. - */ - wait_queue_head_t reset_queue; - - struct srcu_struct reset_backoff_srcu; + atomic_t reset_engine_count[I915_NUM_ENGINES]; }; struct drm_i915_error_state_buf { diff --git a/drivers/gpu/drm/i915/i915_request.c b/drivers/gpu/drm/i915/i915_request.c index 222c9c56e9de..8ac7d14ec8c9 100644 --- a/drivers/gpu/drm/i915/i915_request.c +++ b/drivers/gpu/drm/i915/i915_request.c @@ -1401,8 +1401,7 @@ long i915_request_wait(struct i915_request *rq, * serialise wait/reset with an explicit lock, we do want * lockdep to detect potential dependency cycles. */ - mutex_acquire(&rq->i915->gpu_error.wedge_mutex.dep_map, - 0, 0, _THIS_IP_); + mutex_acquire(&rq->engine->gt->reset.mutex.dep_map, 0, 0, _THIS_IP_); /* * Optimistic spin before touching IRQs. @@ -1480,7 +1479,7 @@ long i915_request_wait(struct i915_request *rq, dma_fence_remove_callback(&rq->fence, &wait.cb); out: - mutex_release(&rq->i915->gpu_error.wedge_mutex.dep_map, 0, _THIS_IP_); + mutex_release(&rq->engine->gt->reset.mutex.dep_map, 0, _THIS_IP_); trace_i915_request_wait_end(rq); return timeout; } diff --git a/drivers/gpu/drm/i915/i915_selftest.h b/drivers/gpu/drm/i915/i915_selftest.h index d9b17b9e6993..acdf6eb9e262 100644 --- a/drivers/gpu/drm/i915/i915_selftest.h +++ b/drivers/gpu/drm/i915/i915_selftest.h @@ -72,6 +72,9 @@ int __i915_nop_teardown(int err, void *data); int __i915_live_setup(void *data); int __i915_live_teardown(int err, void *data); +int __intel_gt_live_setup(void *data); +int __intel_gt_live_teardown(int err, void *data); + int __i915_subtests(const char *caller, int (*setup)(void *data), int (*teardown)(int err, void *data), @@ -88,6 +91,12 @@ int __i915_subtests(const char *caller, __i915_live_setup, __i915_live_teardown, \ T, ARRAY_SIZE(T), data); \ }) +#define intel_gt_live_subtests(T, data) ({ \ + typecheck(struct intel_gt *, data); \ + __i915_subtests(__func__, \ + __intel_gt_live_setup, __intel_gt_live_teardown, \ + T, ARRAY_SIZE(T), data); \ +}) #define SUBTEST(x) { x, #x } diff --git a/drivers/gpu/drm/i915/intel_guc_submission.c b/drivers/gpu/drm/i915/intel_guc_submission.c index b663b5fe51a8..d86336aa4178 100644 --- a/drivers/gpu/drm/i915/intel_guc_submission.c +++ b/drivers/gpu/drm/i915/intel_guc_submission.c @@ -667,7 +667,7 @@ static void guc_reset(struct intel_engine_cs *engine, bool stalled) if (!i915_request_started(rq)) stalled = false; - i915_reset_request(rq, stalled); + __i915_request_reset(rq, stalled); intel_lr_context_reset(engine, rq->hw_context, rq->head, stalled); out_unlock: diff --git a/drivers/gpu/drm/i915/intel_uc.c b/drivers/gpu/drm/i915/intel_uc.c index 72bbcb2597fd..bc589efd3a6d 100644 --- a/drivers/gpu/drm/i915/intel_uc.c +++ b/drivers/gpu/drm/i915/intel_uc.c @@ -38,7 +38,7 @@ static int __intel_uc_reset_hw(struct drm_i915_private *dev_priv) int ret; u32 guc_status; - ret = intel_reset_guc(dev_priv); + ret = intel_reset_guc(&dev_priv->gt); if (ret) { DRM_ERROR("Failed to reset GuC, ret = %d\n", ret); return ret; diff --git a/drivers/gpu/drm/i915/selftests/i915_active.c b/drivers/gpu/drm/i915/selftests/i915_active.c index 84fce379c0de..e5cd5d47e380 100644 --- a/drivers/gpu/drm/i915/selftests/i915_active.c +++ b/drivers/gpu/drm/i915/selftests/i915_active.c @@ -7,6 +7,7 @@ #include #include "gem/i915_gem_pm.h" +#include "gt/intel_gt.h" #include "i915_selftest.h" @@ -221,7 +222,7 @@ int i915_active_live_selftests(struct drm_i915_private *i915) SUBTEST(live_active_retire), }; - if (i915_terminally_wedged(i915)) + if (intel_gt_is_wedged(&i915->gt)) return 0; return i915_subtests(tests, i915); diff --git a/drivers/gpu/drm/i915/selftests/i915_gem.c b/drivers/gpu/drm/i915/selftests/i915_gem.c index b8ffae481730..bb6dd54a6ff3 100644 --- a/drivers/gpu/drm/i915/selftests/i915_gem.c +++ b/drivers/gpu/drm/i915/selftests/i915_gem.c @@ -8,6 +8,7 @@ #include "gem/selftests/igt_gem_utils.h" #include "gem/selftests/mock_context.h" +#include "gt/intel_gt.h" #include "i915_selftest.h" @@ -206,7 +207,7 @@ int i915_gem_live_selftests(struct drm_i915_private *i915) SUBTEST(igt_gem_hibernate), }; - if (i915_terminally_wedged(i915)) + if (intel_gt_is_wedged(&i915->gt)) return 0; return i915_live_subtests(tests, i915); diff --git a/drivers/gpu/drm/i915/selftests/i915_gem_evict.c b/drivers/gpu/drm/i915/selftests/i915_gem_evict.c index a3cb0aade6f1..b6449d0a8c17 100644 --- a/drivers/gpu/drm/i915/selftests/i915_gem_evict.c +++ b/drivers/gpu/drm/i915/selftests/i915_gem_evict.c @@ -25,6 +25,7 @@ #include "gem/i915_gem_pm.h" #include "gem/selftests/igt_gem_utils.h" #include "gem/selftests/mock_context.h" +#include "gt/intel_gt.h" #include "i915_selftest.h" @@ -557,7 +558,7 @@ int i915_gem_evict_live_selftests(struct drm_i915_private *i915) SUBTEST(igt_evict_contexts), }; - if (i915_terminally_wedged(i915)) + if (intel_gt_is_wedged(&i915->gt)) return 0; return i915_subtests(tests, i915); diff --git a/drivers/gpu/drm/i915/selftests/i915_request.c b/drivers/gpu/drm/i915/selftests/i915_request.c index 1bbfc43d4a9e..86c299663934 100644 --- a/drivers/gpu/drm/i915/selftests/i915_request.c +++ b/drivers/gpu/drm/i915/selftests/i915_request.c @@ -380,7 +380,7 @@ static int __igt_breadcrumbs_smoketest(void *arg) t->engine->name); GEM_TRACE_DUMP(); - i915_gem_set_wedged(t->engine->i915); + intel_gt_set_wedged(t->engine->gt); GEM_BUG_ON(!i915_request_completed(rq)); i915_sw_fence_wait(wait); err = -EIO; @@ -1234,7 +1234,7 @@ int i915_request_live_selftests(struct drm_i915_private *i915) SUBTEST(live_breadcrumbs_smoketest), }; - if (i915_terminally_wedged(i915)) + if (intel_gt_is_wedged(&i915->gt)) return 0; return i915_subtests(tests, i915); diff --git a/drivers/gpu/drm/i915/selftests/i915_selftest.c b/drivers/gpu/drm/i915/selftests/i915_selftest.c index f46ccf817ad5..db9c645bbdfe 100644 --- a/drivers/gpu/drm/i915/selftests/i915_selftest.c +++ b/drivers/gpu/drm/i915/selftests/i915_selftest.c @@ -256,7 +256,7 @@ int __i915_live_setup(void *data) { struct drm_i915_private *i915 = data; - return i915_terminally_wedged(i915); + return intel_gt_terminally_wedged(&i915->gt); } int __i915_live_teardown(int err, void *data) @@ -273,6 +273,27 @@ int __i915_live_teardown(int err, void *data) return err; } +int __intel_gt_live_setup(void *data) +{ + struct intel_gt *gt = data; + + return intel_gt_terminally_wedged(gt); +} + +int __intel_gt_live_teardown(int err, void *data) +{ + struct intel_gt *gt = data; + + mutex_lock(>->i915->drm.struct_mutex); + if (igt_flush_test(gt->i915, I915_WAIT_LOCKED)) + err = -EIO; + mutex_unlock(>->i915->drm.struct_mutex); + + i915_gem_drain_freed_objects(gt->i915); + + return err; +} + int __i915_subtests(const char *caller, int (*setup)(void *data), int (*teardown)(int err, void *data), diff --git a/drivers/gpu/drm/i915/selftests/igt_flush_test.c b/drivers/gpu/drm/i915/selftests/igt_flush_test.c index 5bfd1b2626a2..d3b5eb402d33 100644 --- a/drivers/gpu/drm/i915/selftests/igt_flush_test.c +++ b/drivers/gpu/drm/i915/selftests/igt_flush_test.c @@ -5,6 +5,7 @@ */ #include "gem/i915_gem_context.h" +#include "gt/intel_gt.h" #include "i915_drv.h" #include "i915_selftest.h" @@ -13,7 +14,7 @@ int igt_flush_test(struct drm_i915_private *i915, unsigned int flags) { - int ret = i915_terminally_wedged(i915) ? -EIO : 0; + int ret = intel_gt_is_wedged(&i915->gt) ? -EIO : 0; int repeat = !!(flags & I915_WAIT_LOCKED); cond_resched(); @@ -27,7 +28,7 @@ int igt_flush_test(struct drm_i915_private *i915, unsigned int flags) __builtin_return_address(0)); GEM_TRACE_DUMP(); - i915_gem_set_wedged(i915); + intel_gt_set_wedged(&i915->gt); repeat = 0; ret = -EIO; } diff --git a/drivers/gpu/drm/i915/selftests/igt_reset.c b/drivers/gpu/drm/i915/selftests/igt_reset.c index 587df6fd4ffe..7ec8f8b049c6 100644 --- a/drivers/gpu/drm/i915/selftests/igt_reset.c +++ b/drivers/gpu/drm/i915/selftests/igt_reset.c @@ -7,47 +7,45 @@ #include "igt_reset.h" #include "gt/intel_engine.h" +#include "gt/intel_gt.h" #include "../i915_drv.h" -void igt_global_reset_lock(struct drm_i915_private *i915) +void igt_global_reset_lock(struct intel_gt *gt) { struct intel_engine_cs *engine; enum intel_engine_id id; - pr_debug("%s: current gpu_error=%08lx\n", - __func__, i915->gpu_error.flags); + pr_debug("%s: current gpu_error=%08lx\n", __func__, gt->reset.flags); - while (test_and_set_bit(I915_RESET_BACKOFF, &i915->gpu_error.flags)) - wait_event(i915->gpu_error.reset_queue, - !test_bit(I915_RESET_BACKOFF, - &i915->gpu_error.flags)); + while (test_and_set_bit(I915_RESET_BACKOFF, >->reset.flags)) + wait_event(gt->reset.queue, + !test_bit(I915_RESET_BACKOFF, >->reset.flags)); - for_each_engine(engine, i915, id) { + for_each_engine(engine, gt->i915, id) { while (test_and_set_bit(I915_RESET_ENGINE + id, - &i915->gpu_error.flags)) - wait_on_bit(&i915->gpu_error.flags, - I915_RESET_ENGINE + id, + >->reset.flags)) + wait_on_bit(>->reset.flags, I915_RESET_ENGINE + id, TASK_UNINTERRUPTIBLE); } } -void igt_global_reset_unlock(struct drm_i915_private *i915) +void igt_global_reset_unlock(struct intel_gt *gt) { struct intel_engine_cs *engine; enum intel_engine_id id; - for_each_engine(engine, i915, id) - clear_bit(I915_RESET_ENGINE + id, &i915->gpu_error.flags); + for_each_engine(engine, gt->i915, id) + clear_bit(I915_RESET_ENGINE + id, >->reset.flags); - clear_bit(I915_RESET_BACKOFF, &i915->gpu_error.flags); - wake_up_all(&i915->gpu_error.reset_queue); + clear_bit(I915_RESET_BACKOFF, >->reset.flags); + wake_up_all(>->reset.queue); } -bool igt_force_reset(struct drm_i915_private *i915) +bool igt_force_reset(struct intel_gt *gt) { - i915_gem_set_wedged(i915); - i915_reset(i915, 0, NULL); + intel_gt_set_wedged(gt); + intel_gt_reset(gt, 0, NULL); - return !i915_reset_failed(i915); + return !intel_gt_is_wedged(gt); } diff --git a/drivers/gpu/drm/i915/selftests/igt_reset.h b/drivers/gpu/drm/i915/selftests/igt_reset.h index 363bd853e50f..851873b67ab3 100644 --- a/drivers/gpu/drm/i915/selftests/igt_reset.h +++ b/drivers/gpu/drm/i915/selftests/igt_reset.h @@ -7,10 +7,12 @@ #ifndef __I915_SELFTESTS_IGT_RESET_H__ #define __I915_SELFTESTS_IGT_RESET_H__ -#include "../i915_drv.h" +#include -void igt_global_reset_lock(struct drm_i915_private *i915); -void igt_global_reset_unlock(struct drm_i915_private *i915); -bool igt_force_reset(struct drm_i915_private *i915); +struct intel_gt; + +void igt_global_reset_lock(struct intel_gt *gt); +void igt_global_reset_unlock(struct intel_gt *gt); +bool igt_force_reset(struct intel_gt *gt); #endif diff --git a/drivers/gpu/drm/i915/selftests/igt_wedge_me.h b/drivers/gpu/drm/i915/selftests/igt_wedge_me.h deleted file mode 100644 index 08e5ff11bbd9..000000000000 --- a/drivers/gpu/drm/i915/selftests/igt_wedge_me.h +++ /dev/null @@ -1,58 +0,0 @@ -/* - * SPDX-License-Identifier: MIT - * - * Copyright © 2018 Intel Corporation - */ - -#ifndef IGT_WEDGE_ME_H -#define IGT_WEDGE_ME_H - -#include - -#include "../i915_gem.h" - -struct drm_i915_private; - -struct igt_wedge_me { - struct delayed_work work; - struct drm_i915_private *i915; - const char *name; -}; - -static void __igt_wedge_me(struct work_struct *work) -{ - struct igt_wedge_me *w = container_of(work, typeof(*w), work.work); - - pr_err("%s timed out, cancelling test.\n", w->name); - - GEM_TRACE("%s timed out.\n", w->name); - GEM_TRACE_DUMP(); - - i915_gem_set_wedged(w->i915); -} - -static void __igt_init_wedge(struct igt_wedge_me *w, - struct drm_i915_private *i915, - long timeout, - const char *name) -{ - w->i915 = i915; - w->name = name; - - INIT_DELAYED_WORK_ONSTACK(&w->work, __igt_wedge_me); - schedule_delayed_work(&w->work, timeout); -} - -static void __igt_fini_wedge(struct igt_wedge_me *w) -{ - cancel_delayed_work_sync(&w->work); - destroy_delayed_work_on_stack(&w->work); - w->i915 = NULL; -} - -#define igt_wedge_on_timeout(W, DEV, TIMEOUT) \ - for (__igt_init_wedge((W), (DEV), (TIMEOUT), __func__); \ - (W)->i915; \ - __igt_fini_wedge((W))) - -#endif /* IGT_WEDGE_ME_H */ diff --git a/drivers/gpu/drm/i915/selftests/mock_gem_device.c b/drivers/gpu/drm/i915/selftests/mock_gem_device.c index 2741805b56c2..fd4cc4809eb8 100644 --- a/drivers/gpu/drm/i915/selftests/mock_gem_device.c +++ b/drivers/gpu/drm/i915/selftests/mock_gem_device.c @@ -183,11 +183,6 @@ struct drm_i915_private *mock_gem_device(void) intel_gt_init_early(&i915->gt, i915); atomic_inc(&i915->gt.wakeref.count); /* disable; no hw support */ - init_waitqueue_head(&i915->gpu_error.wait_queue); - init_waitqueue_head(&i915->gpu_error.reset_queue); - init_srcu_struct(&i915->gpu_error.reset_backoff_srcu); - mutex_init(&i915->gpu_error.wedge_mutex); - i915->wq = alloc_ordered_workqueue("mock", 0); if (!i915->wq) goto err_drv; -- cgit v1.2.3 From b40d73784ffc33f3c6431e7ceec3b20fffcd95c3 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Sun, 4 Aug 2019 13:48:26 +0100 Subject: drm/i915: Replace struct_mutex for batch pool serialisation Switch to tracking activity via i915_active on individual nodes, only keeping a list of retired objects in the cache, and reaping the cache when the engine itself idles. Signed-off-by: Chris Wilson Reviewed-by: Matthew Auld Link: https://patchwork.freedesktop.org/patch/msgid/20190804124826.30272-2-chris@chris-wilson.co.uk --- drivers/gpu/drm/i915/Makefile | 2 +- drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c | 58 ++++--- drivers/gpu/drm/i915/gem/i915_gem_object.c | 1 - drivers/gpu/drm/i915/gem/i915_gem_object_types.h | 1 - drivers/gpu/drm/i915/gem/i915_gem_pm.c | 4 +- drivers/gpu/drm/i915/gt/intel_engine.h | 1 - drivers/gpu/drm/i915/gt/intel_engine_cs.c | 11 +- drivers/gpu/drm/i915/gt/intel_engine_pm.c | 2 + drivers/gpu/drm/i915/gt/intel_engine_pool.c | 177 ++++++++++++++++++++++ drivers/gpu/drm/i915/gt/intel_engine_pool.h | 34 +++++ drivers/gpu/drm/i915/gt/intel_engine_pool_types.h | 29 ++++ drivers/gpu/drm/i915/gt/intel_engine_types.h | 6 +- drivers/gpu/drm/i915/gt/mock_engine.c | 2 + drivers/gpu/drm/i915/i915_debugfs.c | 68 --------- drivers/gpu/drm/i915/i915_gem_batch_pool.c | 132 ---------------- drivers/gpu/drm/i915/i915_gem_batch_pool.h | 26 ---- 16 files changed, 289 insertions(+), 265 deletions(-) create mode 100644 drivers/gpu/drm/i915/gt/intel_engine_pool.c create mode 100644 drivers/gpu/drm/i915/gt/intel_engine_pool.h create mode 100644 drivers/gpu/drm/i915/gt/intel_engine_pool_types.h delete mode 100644 drivers/gpu/drm/i915/i915_gem_batch_pool.c delete mode 100644 drivers/gpu/drm/i915/i915_gem_batch_pool.h (limited to 'drivers/gpu/drm/i915/gt/intel_engine.h') diff --git a/drivers/gpu/drm/i915/Makefile b/drivers/gpu/drm/i915/Makefile index 654bdcbd23c7..53a55326106a 100644 --- a/drivers/gpu/drm/i915/Makefile +++ b/drivers/gpu/drm/i915/Makefile @@ -72,6 +72,7 @@ obj-y += gt/ gt-y += \ gt/intel_breadcrumbs.o \ gt/intel_context.o \ + gt/intel_engine_pool.o \ gt/intel_engine_cs.o \ gt/intel_engine_pm.o \ gt/intel_gt.o \ @@ -125,7 +126,6 @@ i915-y += \ $(gem-y) \ i915_active.o \ i915_cmd_parser.o \ - i915_gem_batch_pool.o \ i915_gem_evict.o \ i915_gem_fence_reg.o \ i915_gem_gtt.o \ diff --git a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c index 8d90498eaf46..0c8c909743f1 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c +++ b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c @@ -16,6 +16,7 @@ #include "gem/i915_gem_ioctls.h" #include "gt/intel_context.h" +#include "gt/intel_engine_pool.h" #include "gt/intel_gt.h" #include "gt/intel_gt_pm.h" @@ -1198,25 +1199,26 @@ static int __reloc_gpu_alloc(struct i915_execbuffer *eb, unsigned int len) { struct reloc_cache *cache = &eb->reloc_cache; - struct drm_i915_gem_object *obj; + struct intel_engine_pool_node *pool; struct i915_request *rq; struct i915_vma *batch; u32 *cmd; int err; - obj = i915_gem_batch_pool_get(&eb->engine->batch_pool, PAGE_SIZE); - if (IS_ERR(obj)) - return PTR_ERR(obj); + pool = intel_engine_pool_get(&eb->engine->pool, PAGE_SIZE); + if (IS_ERR(pool)) + return PTR_ERR(pool); - cmd = i915_gem_object_pin_map(obj, + cmd = i915_gem_object_pin_map(pool->obj, cache->has_llc ? I915_MAP_FORCE_WB : I915_MAP_FORCE_WC); - i915_gem_object_unpin_pages(obj); - if (IS_ERR(cmd)) - return PTR_ERR(cmd); + if (IS_ERR(cmd)) { + err = PTR_ERR(cmd); + goto out_pool; + } - batch = i915_vma_instance(obj, vma->vm, NULL); + batch = i915_vma_instance(pool->obj, vma->vm, NULL); if (IS_ERR(batch)) { err = PTR_ERR(batch); goto err_unmap; @@ -1232,6 +1234,10 @@ static int __reloc_gpu_alloc(struct i915_execbuffer *eb, goto err_unpin; } + err = intel_engine_pool_mark_active(pool, rq); + if (err) + goto err_request; + err = reloc_move_to_gpu(rq, vma); if (err) goto err_request; @@ -1257,7 +1263,7 @@ static int __reloc_gpu_alloc(struct i915_execbuffer *eb, cache->rq_size = 0; /* Return with batch mapping (cmd) still pinned */ - return 0; + goto out_pool; skip_request: i915_request_skip(rq, err); @@ -1266,7 +1272,9 @@ err_request: err_unpin: i915_vma_unpin(batch); err_unmap: - i915_gem_object_unpin_map(obj); + i915_gem_object_unpin_map(pool->obj); +out_pool: + intel_engine_pool_put(pool); return err; } @@ -2010,18 +2018,17 @@ static int i915_reset_gen7_sol_offsets(struct i915_request *rq) static struct i915_vma *eb_parse(struct i915_execbuffer *eb, bool is_master) { - struct drm_i915_gem_object *shadow_batch_obj; + struct intel_engine_pool_node *pool; struct i915_vma *vma; int err; - shadow_batch_obj = i915_gem_batch_pool_get(&eb->engine->batch_pool, - PAGE_ALIGN(eb->batch_len)); - if (IS_ERR(shadow_batch_obj)) - return ERR_CAST(shadow_batch_obj); + pool = intel_engine_pool_get(&eb->engine->pool, eb->batch_len); + if (IS_ERR(pool)) + return ERR_CAST(pool); err = intel_engine_cmd_parser(eb->engine, eb->batch->obj, - shadow_batch_obj, + pool->obj, eb->batch_start_offset, eb->batch_len, is_master); @@ -2030,12 +2037,12 @@ static struct i915_vma *eb_parse(struct i915_execbuffer *eb, bool is_master) vma = NULL; else vma = ERR_PTR(err); - goto out; + goto err; } - vma = i915_gem_object_ggtt_pin(shadow_batch_obj, NULL, 0, 0, 0); + vma = i915_gem_object_ggtt_pin(pool->obj, NULL, 0, 0, 0); if (IS_ERR(vma)) - goto out; + goto err; eb->vma[eb->buffer_count] = i915_vma_get(vma); eb->flags[eb->buffer_count] = @@ -2043,8 +2050,11 @@ static struct i915_vma *eb_parse(struct i915_execbuffer *eb, bool is_master) vma->exec_flags = &eb->flags[eb->buffer_count]; eb->buffer_count++; -out: - i915_gem_object_unpin_pages(shadow_batch_obj); + vma->private = pool; + return vma; + +err: + intel_engine_pool_put(pool); return vma; } @@ -2588,6 +2598,8 @@ i915_gem_do_execbuffer(struct drm_device *dev, * to explicitly hold another reference here. */ eb.request->batch = eb.batch; + if (eb.batch->private) + intel_engine_pool_mark_active(eb.batch->private, eb.request); trace_i915_request_queue(eb.request, eb.batch_flags); err = eb_submit(&eb); @@ -2612,6 +2624,8 @@ err_request: err_batch_unpin: if (eb.batch_flags & I915_DISPATCH_SECURE) i915_vma_unpin(eb.batch); + if (eb.batch->private) + intel_engine_pool_put(eb.batch->private); err_vma: if (eb.exec) eb_release_vmas(&eb); diff --git a/drivers/gpu/drm/i915/gem/i915_gem_object.c b/drivers/gpu/drm/i915/gem/i915_gem_object.c index 19d55115747c..09f1843f9274 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_object.c +++ b/drivers/gpu/drm/i915/gem/i915_gem_object.c @@ -66,7 +66,6 @@ void i915_gem_object_init(struct drm_i915_gem_object *obj, INIT_LIST_HEAD(&obj->mm.link); INIT_LIST_HEAD(&obj->lut_list); - INIT_LIST_HEAD(&obj->batch_pool_link); init_rcu_head(&obj->rcu); diff --git a/drivers/gpu/drm/i915/gem/i915_gem_object_types.h b/drivers/gpu/drm/i915/gem/i915_gem_object_types.h index 34b51fad02de..d474c6ac4100 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_object_types.h +++ b/drivers/gpu/drm/i915/gem/i915_gem_object_types.h @@ -114,7 +114,6 @@ struct drm_i915_gem_object { unsigned int userfault_count; struct list_head userfault_link; - struct list_head batch_pool_link; I915_SELFTEST_DECLARE(struct list_head st_link); /* diff --git a/drivers/gpu/drm/i915/gem/i915_gem_pm.c b/drivers/gpu/drm/i915/gem/i915_gem_pm.c index b5561cbdc5ea..72922703af49 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_pm.c +++ b/drivers/gpu/drm/i915/gem/i915_gem_pm.c @@ -34,10 +34,8 @@ static void i915_gem_park(struct drm_i915_private *i915) lockdep_assert_held(&i915->drm.struct_mutex); - for_each_engine(engine, i915, id) { + for_each_engine(engine, i915, id) call_idle_barriers(engine); /* cleanup after wedging */ - i915_gem_batch_pool_fini(&engine->batch_pool); - } intel_timelines_park(i915); i915_vma_parked(i915); diff --git a/drivers/gpu/drm/i915/gt/intel_engine.h b/drivers/gpu/drm/i915/gt/intel_engine.h index db5c73ce86ee..a4db4dd22b4f 100644 --- a/drivers/gpu/drm/i915/gt/intel_engine.h +++ b/drivers/gpu/drm/i915/gt/intel_engine.h @@ -9,7 +9,6 @@ #include #include -#include "i915_gem_batch_pool.h" #include "i915_pmu.h" #include "i915_reg.h" #include "i915_request.h" diff --git a/drivers/gpu/drm/i915/gt/intel_engine_cs.c b/drivers/gpu/drm/i915/gt/intel_engine_cs.c index 8bd9a9adf4a5..a91f15717cc1 100644 --- a/drivers/gpu/drm/i915/gt/intel_engine_cs.c +++ b/drivers/gpu/drm/i915/gt/intel_engine_cs.c @@ -32,6 +32,7 @@ #include "intel_engine.h" #include "intel_engine_pm.h" +#include "intel_engine_pool.h" #include "intel_context.h" #include "intel_lrc.h" #include "intel_reset.h" @@ -492,11 +493,6 @@ cleanup: return err; } -static void intel_engine_init_batch_pool(struct intel_engine_cs *engine) -{ - i915_gem_batch_pool_init(&engine->batch_pool, engine); -} - void intel_engine_init_execlists(struct intel_engine_cs *engine) { struct intel_engine_execlists * const execlists = &engine->execlists; @@ -622,10 +618,11 @@ static int intel_engine_setup_common(struct intel_engine_cs *engine) intel_engine_init_breadcrumbs(engine); intel_engine_init_execlists(engine); intel_engine_init_hangcheck(engine); - intel_engine_init_batch_pool(engine); intel_engine_init_cmd_parser(engine); intel_engine_init__pm(engine); + intel_engine_pool_init(&engine->pool); + /* Use the whole device by default */ engine->sseu = intel_sseu_from_device_info(&RUNTIME_INFO(engine->i915)->sseu); @@ -869,9 +866,9 @@ void intel_engine_cleanup_common(struct intel_engine_cs *engine) cleanup_status_page(engine); + intel_engine_pool_fini(&engine->pool); intel_engine_fini_breadcrumbs(engine); intel_engine_cleanup_cmd_parser(engine); - i915_gem_batch_pool_fini(&engine->batch_pool); if (engine->default_state) i915_gem_object_put(engine->default_state); diff --git a/drivers/gpu/drm/i915/gt/intel_engine_pm.c b/drivers/gpu/drm/i915/gt/intel_engine_pm.c index ce54092475da..0336204988d6 100644 --- a/drivers/gpu/drm/i915/gt/intel_engine_pm.c +++ b/drivers/gpu/drm/i915/gt/intel_engine_pm.c @@ -8,6 +8,7 @@ #include "intel_engine.h" #include "intel_engine_pm.h" +#include "intel_engine_pool.h" #include "intel_gt.h" #include "intel_gt_pm.h" @@ -116,6 +117,7 @@ static int __engine_park(struct intel_wakeref *wf) GEM_TRACE("%s\n", engine->name); intel_engine_disarm_breadcrumbs(engine); + intel_engine_pool_park(&engine->pool); /* Must be reset upon idling, or we may miss the busy wakeup. */ GEM_BUG_ON(engine->execlists.queue_priority_hint != INT_MIN); diff --git a/drivers/gpu/drm/i915/gt/intel_engine_pool.c b/drivers/gpu/drm/i915/gt/intel_engine_pool.c new file mode 100644 index 000000000000..03d90b49584a --- /dev/null +++ b/drivers/gpu/drm/i915/gt/intel_engine_pool.c @@ -0,0 +1,177 @@ +/* + * SPDX-License-Identifier: MIT + * + * Copyright © 2014-2018 Intel Corporation + */ + +#include "gem/i915_gem_object.h" + +#include "i915_drv.h" +#include "intel_engine_pm.h" +#include "intel_engine_pool.h" + +static struct intel_engine_cs *to_engine(struct intel_engine_pool *pool) +{ + return container_of(pool, struct intel_engine_cs, pool); +} + +static struct list_head * +bucket_for_size(struct intel_engine_pool *pool, size_t sz) +{ + int n; + + /* + * Compute a power-of-two bucket, but throw everything greater than + * 16KiB into the same bucket: i.e. the buckets hold objects of + * (1 page, 2 pages, 4 pages, 8+ pages). + */ + n = fls(sz >> PAGE_SHIFT) - 1; + if (n >= ARRAY_SIZE(pool->cache_list)) + n = ARRAY_SIZE(pool->cache_list) - 1; + + return &pool->cache_list[n]; +} + +static void node_free(struct intel_engine_pool_node *node) +{ + i915_gem_object_put(node->obj); + i915_active_fini(&node->active); + kfree(node); +} + +static int pool_active(struct i915_active *ref) +{ + struct intel_engine_pool_node *node = + container_of(ref, typeof(*node), active); + struct reservation_object *resv = node->obj->base.resv; + int err; + + if (reservation_object_trylock(resv)) { + reservation_object_add_excl_fence(resv, NULL); + reservation_object_unlock(resv); + } + + err = i915_gem_object_pin_pages(node->obj); + if (err) + return err; + + /* Hide this pinned object from the shrinker until retired */ + i915_gem_object_make_unshrinkable(node->obj); + + return 0; +} + +static void pool_retire(struct i915_active *ref) +{ + struct intel_engine_pool_node *node = + container_of(ref, typeof(*node), active); + struct intel_engine_pool *pool = node->pool; + struct list_head *list = bucket_for_size(pool, node->obj->base.size); + unsigned long flags; + + GEM_BUG_ON(!intel_engine_pm_is_awake(to_engine(pool))); + + i915_gem_object_unpin_pages(node->obj); + + /* Return this object to the shrinker pool */ + i915_gem_object_make_purgeable(node->obj); + + spin_lock_irqsave(&pool->lock, flags); + list_add(&node->link, list); + spin_unlock_irqrestore(&pool->lock, flags); +} + +static struct intel_engine_pool_node * +node_create(struct intel_engine_pool *pool, size_t sz) +{ + struct intel_engine_cs *engine = to_engine(pool); + struct intel_engine_pool_node *node; + struct drm_i915_gem_object *obj; + + node = kmalloc(sizeof(*node), + GFP_KERNEL | __GFP_RETRY_MAYFAIL | __GFP_NOWARN); + if (!node) + return ERR_PTR(-ENOMEM); + + node->pool = pool; + i915_active_init(engine->i915, &node->active, pool_active, pool_retire); + + obj = i915_gem_object_create_internal(engine->i915, sz); + if (IS_ERR(obj)) { + i915_active_fini(&node->active); + kfree(node); + return ERR_CAST(obj); + } + + node->obj = obj; + return node; +} + +struct intel_engine_pool_node * +intel_engine_pool_get(struct intel_engine_pool *pool, size_t size) +{ + struct intel_engine_pool_node *node; + struct list_head *list; + unsigned long flags; + int ret; + + GEM_BUG_ON(!intel_engine_pm_is_awake(to_engine(pool))); + + size = PAGE_ALIGN(size); + list = bucket_for_size(pool, size); + + spin_lock_irqsave(&pool->lock, flags); + list_for_each_entry(node, list, link) { + if (node->obj->base.size < size) + continue; + list_del(&node->link); + break; + } + spin_unlock_irqrestore(&pool->lock, flags); + + if (&node->link == list) { + node = node_create(pool, size); + if (IS_ERR(node)) + return node; + } + + ret = i915_active_acquire(&node->active); + if (ret) { + node_free(node); + return ERR_PTR(ret); + } + + return node; +} + +void intel_engine_pool_init(struct intel_engine_pool *pool) +{ + int n; + + spin_lock_init(&pool->lock); + for (n = 0; n < ARRAY_SIZE(pool->cache_list); n++) + INIT_LIST_HEAD(&pool->cache_list[n]); +} + +void intel_engine_pool_park(struct intel_engine_pool *pool) +{ + int n; + + for (n = 0; n < ARRAY_SIZE(pool->cache_list); n++) { + struct list_head *list = &pool->cache_list[n]; + struct intel_engine_pool_node *node, *nn; + + list_for_each_entry_safe(node, nn, list, link) + node_free(node); + + INIT_LIST_HEAD(list); + } +} + +void intel_engine_pool_fini(struct intel_engine_pool *pool) +{ + int n; + + for (n = 0; n < ARRAY_SIZE(pool->cache_list); n++) + GEM_BUG_ON(!list_empty(&pool->cache_list[n])); +} diff --git a/drivers/gpu/drm/i915/gt/intel_engine_pool.h b/drivers/gpu/drm/i915/gt/intel_engine_pool.h new file mode 100644 index 000000000000..f7a0a660c1c9 --- /dev/null +++ b/drivers/gpu/drm/i915/gt/intel_engine_pool.h @@ -0,0 +1,34 @@ +/* + * SPDX-License-Identifier: MIT + * + * Copyright © 2014-2018 Intel Corporation + */ + +#ifndef INTEL_ENGINE_POOL_H +#define INTEL_ENGINE_POOL_H + +#include "intel_engine_pool_types.h" +#include "i915_active.h" +#include "i915_request.h" + +struct intel_engine_pool_node * +intel_engine_pool_get(struct intel_engine_pool *pool, size_t size); + +static inline int +intel_engine_pool_mark_active(struct intel_engine_pool_node *node, + struct i915_request *rq) +{ + return i915_active_ref(&node->active, rq->fence.context, rq); +} + +static inline void +intel_engine_pool_put(struct intel_engine_pool_node *node) +{ + i915_active_release(&node->active); +} + +void intel_engine_pool_init(struct intel_engine_pool *pool); +void intel_engine_pool_park(struct intel_engine_pool *pool); +void intel_engine_pool_fini(struct intel_engine_pool *pool); + +#endif /* INTEL_ENGINE_POOL_H */ diff --git a/drivers/gpu/drm/i915/gt/intel_engine_pool_types.h b/drivers/gpu/drm/i915/gt/intel_engine_pool_types.h new file mode 100644 index 000000000000..e31ee361b76f --- /dev/null +++ b/drivers/gpu/drm/i915/gt/intel_engine_pool_types.h @@ -0,0 +1,29 @@ +/* + * SPDX-License-Identifier: MIT + * + * Copyright © 2014-2018 Intel Corporation + */ + +#ifndef INTEL_ENGINE_POOL_TYPES_H +#define INTEL_ENGINE_POOL_TYPES_H + +#include +#include + +#include "i915_active_types.h" + +struct drm_i915_gem_object; + +struct intel_engine_pool { + spinlock_t lock; + struct list_head cache_list[4]; +}; + +struct intel_engine_pool_node { + struct i915_active active; + struct drm_i915_gem_object *obj; + struct list_head link; + struct intel_engine_pool *pool; +}; + +#endif /* INTEL_ENGINE_POOL_TYPES_H */ diff --git a/drivers/gpu/drm/i915/gt/intel_engine_types.h b/drivers/gpu/drm/i915/gt/intel_engine_types.h index da61dd329210..2e273e8f5064 100644 --- a/drivers/gpu/drm/i915/gt/intel_engine_types.h +++ b/drivers/gpu/drm/i915/gt/intel_engine_types.h @@ -16,12 +16,12 @@ #include #include "i915_gem.h" -#include "i915_gem_batch_pool.h" #include "i915_pmu.h" #include "i915_priolist_types.h" #include "i915_selftest.h" -#include "gt/intel_timeline_types.h" +#include "intel_engine_pool_types.h" #include "intel_sseu.h" +#include "intel_timeline_types.h" #include "intel_wakeref.h" #include "intel_workarounds_types.h" @@ -354,7 +354,7 @@ struct intel_engine_cs { * when the command parser is enabled. Prevents the client from * modifying the batch contents after software parsing. */ - struct i915_gem_batch_pool batch_pool; + struct intel_engine_pool pool; struct intel_hw_status_page status_page; struct i915_ctx_workarounds wa_ctx; diff --git a/drivers/gpu/drm/i915/gt/mock_engine.c b/drivers/gpu/drm/i915/gt/mock_engine.c index 10cb312462e5..8a5f07935b84 100644 --- a/drivers/gpu/drm/i915/gt/mock_engine.c +++ b/drivers/gpu/drm/i915/gt/mock_engine.c @@ -27,6 +27,7 @@ #include "i915_drv.h" #include "intel_context.h" #include "intel_engine_pm.h" +#include "intel_engine_pool.h" #include "mock_engine.h" #include "selftests/mock_request.h" @@ -292,6 +293,7 @@ int mock_engine_init(struct intel_engine_cs *engine) intel_engine_init_breadcrumbs(engine); intel_engine_init_execlists(engine); intel_engine_init__pm(engine); + intel_engine_pool_init(&engine->pool); engine->kernel_context = i915_gem_context_get_engine(i915->kernel_context, engine->id); diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c index 19f156a7f501..8953336f2ae5 100644 --- a/drivers/gpu/drm/i915/i915_debugfs.c +++ b/drivers/gpu/drm/i915/i915_debugfs.c @@ -295,27 +295,6 @@ static int per_file_stats(int id, void *ptr, void *data) stats.closed); \ } while (0) -static void print_batch_pool_stats(struct seq_file *m, - struct drm_i915_private *dev_priv) -{ - struct drm_i915_gem_object *obj; - struct intel_engine_cs *engine; - struct file_stats stats = {}; - enum intel_engine_id id; - int j; - - for_each_engine(engine, dev_priv, id) { - for (j = 0; j < ARRAY_SIZE(engine->batch_pool.cache_list); j++) { - list_for_each_entry(obj, - &engine->batch_pool.cache_list[j], - batch_pool_link) - per_file_stats(0, obj, &stats); - } - } - - print_file_stats(m, "[k]batch pool", stats); -} - static void print_context_stats(struct seq_file *m, struct drm_i915_private *i915) { @@ -374,58 +353,12 @@ static int i915_gem_object_info(struct seq_file *m, void *data) if (ret) return ret; - print_batch_pool_stats(m, i915); print_context_stats(m, i915); mutex_unlock(&i915->drm.struct_mutex); return 0; } -static int i915_gem_batch_pool_info(struct seq_file *m, void *data) -{ - struct drm_i915_private *dev_priv = node_to_i915(m->private); - struct drm_device *dev = &dev_priv->drm; - struct drm_i915_gem_object *obj; - struct intel_engine_cs *engine; - enum intel_engine_id id; - int total = 0; - int ret, j; - - ret = mutex_lock_interruptible(&dev->struct_mutex); - if (ret) - return ret; - - for_each_engine(engine, dev_priv, id) { - for (j = 0; j < ARRAY_SIZE(engine->batch_pool.cache_list); j++) { - int count; - - count = 0; - list_for_each_entry(obj, - &engine->batch_pool.cache_list[j], - batch_pool_link) - count++; - seq_printf(m, "%s cache[%d]: %d objects\n", - engine->name, j, count); - - list_for_each_entry(obj, - &engine->batch_pool.cache_list[j], - batch_pool_link) { - seq_puts(m, " "); - describe_obj(m, obj); - seq_putc(m, '\n'); - } - - total += count; - } - } - - seq_printf(m, "total: %d\n", total); - - mutex_unlock(&dev->struct_mutex); - - return 0; -} - static void gen8_display_interrupt_info(struct seq_file *m) { struct drm_i915_private *dev_priv = node_to_i915(m->private); @@ -4385,7 +4318,6 @@ static const struct drm_info_list i915_debugfs_list[] = { {"i915_gem_objects", i915_gem_object_info, 0}, {"i915_gem_fence_regs", i915_gem_fence_regs_info, 0}, {"i915_gem_interrupt", i915_interrupt_info, 0}, - {"i915_gem_batch_pool", i915_gem_batch_pool_info, 0}, {"i915_guc_info", i915_guc_info, 0}, {"i915_guc_load_status", i915_guc_load_status_info, 0}, {"i915_guc_log_dump", i915_guc_log_dump, 0}, diff --git a/drivers/gpu/drm/i915/i915_gem_batch_pool.c b/drivers/gpu/drm/i915/i915_gem_batch_pool.c deleted file mode 100644 index b17f23991253..000000000000 --- a/drivers/gpu/drm/i915/i915_gem_batch_pool.c +++ /dev/null @@ -1,132 +0,0 @@ -/* - * SPDX-License-Identifier: MIT - * - * Copyright © 2014-2018 Intel Corporation - */ - -#include "i915_gem_batch_pool.h" -#include "i915_drv.h" - -/** - * DOC: batch pool - * - * In order to submit batch buffers as 'secure', the software command parser - * must ensure that a batch buffer cannot be modified after parsing. It does - * this by copying the user provided batch buffer contents to a kernel owned - * buffer from which the hardware will actually execute, and by carefully - * managing the address space bindings for such buffers. - * - * The batch pool framework provides a mechanism for the driver to manage a - * set of scratch buffers to use for this purpose. The framework can be - * extended to support other uses cases should they arise. - */ - -/** - * i915_gem_batch_pool_init() - initialize a batch buffer pool - * @pool: the batch buffer pool - * @engine: the associated request submission engine - */ -void i915_gem_batch_pool_init(struct i915_gem_batch_pool *pool, - struct intel_engine_cs *engine) -{ - int n; - - pool->engine = engine; - - for (n = 0; n < ARRAY_SIZE(pool->cache_list); n++) - INIT_LIST_HEAD(&pool->cache_list[n]); -} - -/** - * i915_gem_batch_pool_fini() - clean up a batch buffer pool - * @pool: the pool to clean up - * - * Note: Callers must hold the struct_mutex. - */ -void i915_gem_batch_pool_fini(struct i915_gem_batch_pool *pool) -{ - int n; - - lockdep_assert_held(&pool->engine->i915->drm.struct_mutex); - - for (n = 0; n < ARRAY_SIZE(pool->cache_list); n++) { - struct drm_i915_gem_object *obj, *next; - - list_for_each_entry_safe(obj, next, - &pool->cache_list[n], - batch_pool_link) - i915_gem_object_put(obj); - - INIT_LIST_HEAD(&pool->cache_list[n]); - } -} - -/** - * i915_gem_batch_pool_get() - allocate a buffer from the pool - * @pool: the batch buffer pool - * @size: the minimum desired size of the returned buffer - * - * Returns an inactive buffer from @pool with at least @size bytes, - * with the pages pinned. The caller must i915_gem_object_unpin_pages() - * on the returned object. - * - * Note: Callers must hold the struct_mutex - * - * Return: the buffer object or an error pointer - */ -struct drm_i915_gem_object * -i915_gem_batch_pool_get(struct i915_gem_batch_pool *pool, - size_t size) -{ - struct drm_i915_gem_object *obj; - struct list_head *list; - int n, ret; - - lockdep_assert_held(&pool->engine->i915->drm.struct_mutex); - - /* Compute a power-of-two bucket, but throw everything greater than - * 16KiB into the same bucket: i.e. the the buckets hold objects of - * (1 page, 2 pages, 4 pages, 8+ pages). - */ - n = fls(size >> PAGE_SHIFT) - 1; - if (n >= ARRAY_SIZE(pool->cache_list)) - n = ARRAY_SIZE(pool->cache_list) - 1; - list = &pool->cache_list[n]; - - list_for_each_entry(obj, list, batch_pool_link) { - struct reservation_object *resv = obj->base.resv; - - /* The batches are strictly LRU ordered */ - if (!reservation_object_test_signaled_rcu(resv, true)) - break; - - /* - * The object is now idle, clear the array of shared - * fences before we add a new request. Although, we - * remain on the same engine, we may be on a different - * timeline and so may continually grow the array, - * trapping a reference to all the old fences, rather - * than replace the existing fence. - */ - if (rcu_access_pointer(resv->fence)) { - reservation_object_lock(resv, NULL); - reservation_object_add_excl_fence(resv, NULL); - reservation_object_unlock(resv); - } - - if (obj->base.size >= size) - goto found; - } - - obj = i915_gem_object_create_internal(pool->engine->i915, size); - if (IS_ERR(obj)) - return obj; - -found: - ret = i915_gem_object_pin_pages(obj); - if (ret) - return ERR_PTR(ret); - - list_move_tail(&obj->batch_pool_link, list); - return obj; -} diff --git a/drivers/gpu/drm/i915/i915_gem_batch_pool.h b/drivers/gpu/drm/i915/i915_gem_batch_pool.h deleted file mode 100644 index feeeeeaa54d8..000000000000 --- a/drivers/gpu/drm/i915/i915_gem_batch_pool.h +++ /dev/null @@ -1,26 +0,0 @@ -/* - * SPDX-License-Identifier: MIT - * - * Copyright © 2014-2018 Intel Corporation - */ - -#ifndef I915_GEM_BATCH_POOL_H -#define I915_GEM_BATCH_POOL_H - -#include - -struct drm_i915_gem_object; -struct intel_engine_cs; - -struct i915_gem_batch_pool { - struct intel_engine_cs *engine; - struct list_head cache_list[4]; -}; - -void i915_gem_batch_pool_init(struct i915_gem_batch_pool *pool, - struct intel_engine_cs *engine); -void i915_gem_batch_pool_fini(struct i915_gem_batch_pool *pool); -struct drm_i915_gem_object * -i915_gem_batch_pool_get(struct i915_gem_batch_pool *pool, size_t size); - -#endif /* I915_GEM_BATCH_POOL_H */ -- cgit v1.2.3 From 750e76b4f9f63c95bf4c283ccf8e2af0e258d3bb Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Tue, 6 Aug 2019 13:43:00 +0100 Subject: drm/i915/gt: Move the [class][inst] lookup for engines onto the GT To maintain a fast lookup from a GT centric irq handler, we want the engine lookup tables on the intel_gt. To avoid having multiple copies of the same multi-dimension lookup table, move the generic user engine lookup into an rbtree (for fast and flexible indexing). v2: Split uabi_instance cf uabi_class v3: Set uabi_class/uabi_instance after collating all engines to provide a stable uabi across parallel unordered construction. Signed-off-by: Chris Wilson Cc: Tvrtko Ursulin Cc: Daniele Ceraolo Spurio Reviewed-by: Daniele Ceraolo Spurio #v2 Link: https://patchwork.freedesktop.org/patch/msgid/20190806124300.24945-2-chris@chris-wilson.co.uk --- drivers/gpu/drm/i915/Makefile | 3 +- drivers/gpu/drm/i915/gem/i915_gem_context.c | 3 +- drivers/gpu/drm/i915/gt/intel_engine.h | 6 - drivers/gpu/drm/i915/gt/intel_engine_cs.c | 109 ++--------- drivers/gpu/drm/i915/gt/intel_engine_types.h | 9 +- drivers/gpu/drm/i915/gt/intel_engine_user.c | 228 ++++++++++++++++++++++ drivers/gpu/drm/i915/gt/intel_engine_user.h | 23 +++ drivers/gpu/drm/i915/gt/intel_gt_types.h | 4 + drivers/gpu/drm/i915/gt/selftest_lrc.c | 15 +- drivers/gpu/drm/i915/gt/selftest_workarounds.c | 1 + drivers/gpu/drm/i915/gt/uc/intel_guc_submission.c | 2 +- drivers/gpu/drm/i915/i915_cmd_parser.c | 3 +- drivers/gpu/drm/i915/i915_debugfs.c | 17 +- drivers/gpu/drm/i915/i915_drv.c | 1 + drivers/gpu/drm/i915/i915_drv.h | 15 +- drivers/gpu/drm/i915/i915_gem.c | 22 +-- drivers/gpu/drm/i915/i915_irq.c | 2 +- drivers/gpu/drm/i915/i915_perf.c | 3 +- drivers/gpu/drm/i915/i915_pmu.c | 8 +- drivers/gpu/drm/i915/i915_query.c | 5 +- drivers/gpu/drm/i915/i915_trace.h | 10 +- 21 files changed, 328 insertions(+), 161 deletions(-) create mode 100644 drivers/gpu/drm/i915/gt/intel_engine_user.c create mode 100644 drivers/gpu/drm/i915/gt/intel_engine_user.h (limited to 'drivers/gpu/drm/i915/gt/intel_engine.h') diff --git a/drivers/gpu/drm/i915/Makefile b/drivers/gpu/drm/i915/Makefile index 53a55326106a..327fc448b2d5 100644 --- a/drivers/gpu/drm/i915/Makefile +++ b/drivers/gpu/drm/i915/Makefile @@ -72,9 +72,10 @@ obj-y += gt/ gt-y += \ gt/intel_breadcrumbs.o \ gt/intel_context.o \ - gt/intel_engine_pool.o \ gt/intel_engine_cs.o \ + gt/intel_engine_pool.o \ gt/intel_engine_pm.o \ + gt/intel_engine_user.o \ gt/intel_gt.o \ gt/intel_gt_pm.o \ gt/intel_hangcheck.o \ diff --git a/drivers/gpu/drm/i915/gem/i915_gem_context.c b/drivers/gpu/drm/i915/gem/i915_gem_context.c index 64f7a533e886..1c5bc21a80ff 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_context.c +++ b/drivers/gpu/drm/i915/gem/i915_gem_context.c @@ -70,6 +70,7 @@ #include #include "gt/intel_lrc_reg.h" +#include "gt/intel_engine_user.h" #include "i915_gem_context.h" #include "i915_globals.h" @@ -1729,7 +1730,7 @@ get_engines(struct i915_gem_context *ctx, if (e->engines[n]) { ci.engine_class = e->engines[n]->engine->uabi_class; - ci.engine_instance = e->engines[n]->engine->instance; + ci.engine_instance = e->engines[n]->engine->uabi_instance; } if (copy_to_user(&user->engines[n], &ci, sizeof(ci))) { diff --git a/drivers/gpu/drm/i915/gt/intel_engine.h b/drivers/gpu/drm/i915/gt/intel_engine.h index a4db4dd22b4f..37c391cee441 100644 --- a/drivers/gpu/drm/i915/gt/intel_engine.h +++ b/drivers/gpu/drm/i915/gt/intel_engine.h @@ -122,8 +122,6 @@ hangcheck_action_to_str(const enum intel_engine_hangcheck_action a) return "unknown"; } -void intel_engines_set_scheduler_caps(struct drm_i915_private *i915); - static inline unsigned int execlists_num_ports(const struct intel_engine_execlists * const execlists) { @@ -422,7 +420,6 @@ bool intel_engine_is_idle(struct intel_engine_cs *engine); bool intel_engines_are_idle(struct intel_gt *gt); void intel_engines_reset_default_submission(struct intel_gt *gt); -unsigned int intel_engines_has_context_isolation(struct drm_i915_private *i915); bool intel_engine_can_store_dword(struct intel_engine_cs *engine); @@ -431,9 +428,6 @@ void intel_engine_dump(struct intel_engine_cs *engine, struct drm_printer *m, const char *header, ...); -struct intel_engine_cs * -intel_engine_lookup_user(struct drm_i915_private *i915, u8 class, u8 instance); - static inline void intel_engine_context_in(struct intel_engine_cs *engine) { unsigned long flags; diff --git a/drivers/gpu/drm/i915/gt/intel_engine_cs.c b/drivers/gpu/drm/i915/gt/intel_engine_cs.c index a91f15717cc1..d0befd6c023a 100644 --- a/drivers/gpu/drm/i915/gt/intel_engine_cs.c +++ b/drivers/gpu/drm/i915/gt/intel_engine_cs.c @@ -33,6 +33,7 @@ #include "intel_engine.h" #include "intel_engine_pm.h" #include "intel_engine_pool.h" +#include "intel_engine_user.h" #include "intel_context.h" #include "intel_lrc.h" #include "intel_reset.h" @@ -286,9 +287,7 @@ static void intel_engine_sanitize_mmio(struct intel_engine_cs *engine) intel_engine_set_hwsp_writemask(engine, ~0u); } -static int -intel_engine_setup(struct drm_i915_private *dev_priv, - enum intel_engine_id id) +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; @@ -304,10 +303,9 @@ intel_engine_setup(struct drm_i915_private *dev_priv, if (GEM_DEBUG_WARN_ON(info->instance > MAX_ENGINE_INSTANCE)) return -EINVAL; - if (GEM_DEBUG_WARN_ON(dev_priv->engine_class[info->class][info->instance])) + if (GEM_DEBUG_WARN_ON(gt->engine_class[info->class][info->instance])) return -EINVAL; - GEM_BUG_ON(dev_priv->engine[id]); engine = kzalloc(sizeof(*engine), GFP_KERNEL); if (!engine) return -ENOMEM; @@ -316,12 +314,12 @@ intel_engine_setup(struct drm_i915_private *dev_priv, engine->id = id; engine->mask = BIT(id); - engine->i915 = dev_priv; - engine->gt = &dev_priv->gt; - engine->uncore = &dev_priv->uncore; + 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(dev_priv, info->mmio_bases); + engine->mmio_base = __engine_mmio_base(gt->i915, info->mmio_bases); engine->class = info->class; engine->instance = info->instance; @@ -331,14 +329,12 @@ intel_engine_setup(struct drm_i915_private *dev_priv, */ engine->destroy = (typeof(engine->destroy))kfree; - engine->uabi_class = intel_engine_classes[info->class].uabi_class; - - engine->context_size = intel_engine_context_size(dev_priv, + engine->context_size = intel_engine_context_size(gt->i915, engine->class); if (WARN_ON(engine->context_size > BIT(20))) engine->context_size = 0; if (engine->context_size) - DRIVER_CAPS(dev_priv)->has_logical_contexts = true; + DRIVER_CAPS(gt->i915)->has_logical_contexts = true; /* Nothing to do here, execute in order of dependencies */ engine->schedule = NULL; @@ -350,8 +346,11 @@ intel_engine_setup(struct drm_i915_private *dev_priv, /* Scrub mmio state on takeover */ intel_engine_sanitize_mmio(engine); - dev_priv->engine_class[info->class][info->instance] = engine; - dev_priv->engine[id] = engine; + gt->engine_class[info->class][info->instance] = engine; + + intel_engine_add_user(engine); + gt->i915->engine[id] = engine; + return 0; } @@ -434,7 +433,7 @@ int intel_engines_init_mmio(struct drm_i915_private *i915) if (!HAS_ENGINE(i915, i)) continue; - err = intel_engine_setup(i915, i); + err = intel_engine_setup(&i915->gt, i); if (err) goto cleanup; @@ -677,47 +676,6 @@ cleanup: return err; } -void intel_engines_set_scheduler_caps(struct drm_i915_private *i915) -{ - static const struct { - u8 engine; - u8 sched; - } map[] = { -#define MAP(x, y) { ilog2(I915_ENGINE_##x), ilog2(I915_SCHEDULER_CAP_##y) } - MAP(HAS_PREEMPTION, PREEMPTION), - MAP(HAS_SEMAPHORES, SEMAPHORES), - MAP(SUPPORTS_STATS, ENGINE_BUSY_STATS), -#undef MAP - }; - struct intel_engine_cs *engine; - enum intel_engine_id id; - u32 enabled, disabled; - - enabled = 0; - disabled = 0; - for_each_engine(engine, i915, id) { /* all engines must agree! */ - int i; - - if (engine->schedule) - enabled |= (I915_SCHEDULER_CAP_ENABLED | - I915_SCHEDULER_CAP_PRIORITY); - else - disabled |= (I915_SCHEDULER_CAP_ENABLED | - I915_SCHEDULER_CAP_PRIORITY); - - for (i = 0; i < ARRAY_SIZE(map); i++) { - if (engine->flags & BIT(map[i].engine)) - enabled |= BIT(map[i].sched); - else - disabled |= BIT(map[i].sched); - } - } - - i915->caps.scheduler = enabled & ~disabled; - if (!(i915->caps.scheduler & I915_SCHEDULER_CAP_ENABLED)) - i915->caps.scheduler = 0; -} - struct measure_breadcrumb { struct i915_request rq; struct intel_timeline timeline; @@ -1187,20 +1145,6 @@ bool intel_engine_can_store_dword(struct intel_engine_cs *engine) } } -unsigned int intel_engines_has_context_isolation(struct drm_i915_private *i915) -{ - struct intel_engine_cs *engine; - enum intel_engine_id id; - unsigned int which; - - which = 0; - for_each_engine(engine, i915, id) - if (engine->default_state) - which |= BIT(engine->uabi_class); - - return which; -} - static int print_sched_attr(struct drm_i915_private *i915, const struct i915_sched_attr *attr, char *buf, int x, int len) @@ -1498,29 +1442,6 @@ void intel_engine_dump(struct intel_engine_cs *engine, intel_engine_print_breadcrumbs(engine, m); } -static u8 user_class_map[] = { - [I915_ENGINE_CLASS_RENDER] = RENDER_CLASS, - [I915_ENGINE_CLASS_COPY] = COPY_ENGINE_CLASS, - [I915_ENGINE_CLASS_VIDEO] = VIDEO_DECODE_CLASS, - [I915_ENGINE_CLASS_VIDEO_ENHANCE] = VIDEO_ENHANCEMENT_CLASS, -}; - -struct intel_engine_cs * -intel_engine_lookup_user(struct drm_i915_private *i915, u8 class, u8 instance) -{ - if (class >= ARRAY_SIZE(user_class_map)) - return NULL; - - class = user_class_map[class]; - - GEM_BUG_ON(class > MAX_ENGINE_CLASS); - - if (instance > MAX_ENGINE_INSTANCE) - return NULL; - - return i915->engine_class[class][instance]; -} - /** * intel_enable_engine_stats() - Enable engine busy tracking on engine * @engine: engine to enable stats collection diff --git a/drivers/gpu/drm/i915/gt/intel_engine_types.h b/drivers/gpu/drm/i915/gt/intel_engine_types.h index 2e273e8f5064..dacaa707c797 100644 --- a/drivers/gpu/drm/i915/gt/intel_engine_types.h +++ b/drivers/gpu/drm/i915/gt/intel_engine_types.h @@ -12,6 +12,7 @@ #include #include #include +#include #include #include @@ -267,15 +268,19 @@ struct intel_engine_cs { unsigned int guc_id; intel_engine_mask_t mask; - u8 uabi_class; - u8 class; u8 instance; + + u8 uabi_class; + u8 uabi_instance; + u32 context_size; u32 mmio_base; u32 uabi_capabilities; + struct rb_node uabi_node; + struct intel_sseu sseu; struct intel_ring *buffer; diff --git a/drivers/gpu/drm/i915/gt/intel_engine_user.c b/drivers/gpu/drm/i915/gt/intel_engine_user.c new file mode 100644 index 000000000000..68fda1ac3c60 --- /dev/null +++ b/drivers/gpu/drm/i915/gt/intel_engine_user.c @@ -0,0 +1,228 @@ +/* + * SPDX-License-Identifier: MIT + * + * Copyright © 2019 Intel Corporation + */ + +#include +#include +#include + +#include "i915_drv.h" +#include "intel_engine.h" +#include "intel_engine_user.h" + +struct intel_engine_cs * +intel_engine_lookup_user(struct drm_i915_private *i915, u8 class, u8 instance) +{ + struct rb_node *p = i915->uabi_engines.rb_node; + + while (p) { + struct intel_engine_cs *it = + rb_entry(p, typeof(*it), uabi_node); + + if (class < it->uabi_class) + p = p->rb_left; + else if (class > it->uabi_class || + instance > it->uabi_instance) + p = p->rb_right; + else if (instance < it->uabi_instance) + p = p->rb_left; + else + return it; + } + + return NULL; +} + +void intel_engine_add_user(struct intel_engine_cs *engine) +{ + llist_add((struct llist_node *)&engine->uabi_node, + (struct llist_head *)&engine->i915->uabi_engines); +} + +static const u8 uabi_classes[] = { + [RENDER_CLASS] = I915_ENGINE_CLASS_RENDER, + [COPY_ENGINE_CLASS] = I915_ENGINE_CLASS_COPY, + [VIDEO_DECODE_CLASS] = I915_ENGINE_CLASS_VIDEO, + [VIDEO_ENHANCEMENT_CLASS] = I915_ENGINE_CLASS_VIDEO_ENHANCE, +}; + +static int engine_cmp(void *priv, struct list_head *A, struct list_head *B) +{ + const struct intel_engine_cs *a = + container_of((struct rb_node *)A, typeof(*a), uabi_node); + const struct intel_engine_cs *b = + container_of((struct rb_node *)B, typeof(*b), uabi_node); + + if (uabi_classes[a->class] < uabi_classes[b->class]) + return -1; + if (uabi_classes[a->class] > uabi_classes[b->class]) + return 1; + + if (a->instance < b->instance) + return -1; + if (a->instance > b->instance) + return 1; + + return 0; +} + +static struct llist_node *get_engines(struct drm_i915_private *i915) +{ + return llist_del_all((struct llist_head *)&i915->uabi_engines); +} + +static void sort_engines(struct drm_i915_private *i915, + struct list_head *engines) +{ + struct llist_node *pos, *next; + + llist_for_each_safe(pos, next, get_engines(i915)) { + struct intel_engine_cs *engine = + container_of((struct rb_node *)pos, typeof(*engine), + uabi_node); + list_add((struct list_head *)&engine->uabi_node, engines); + } + list_sort(NULL, engines, engine_cmp); +} + +static void set_scheduler_caps(struct drm_i915_private *i915) +{ + static const struct { + u8 engine; + u8 sched; + } map[] = { +#define MAP(x, y) { ilog2(I915_ENGINE_##x), ilog2(I915_SCHEDULER_CAP_##y) } + MAP(HAS_PREEMPTION, PREEMPTION), + MAP(HAS_SEMAPHORES, SEMAPHORES), + MAP(SUPPORTS_STATS, ENGINE_BUSY_STATS), +#undef MAP + }; + struct intel_engine_cs *engine; + u32 enabled, disabled; + + enabled = 0; + disabled = 0; + for_each_uabi_engine(engine, i915) { /* all engines must agree! */ + int i; + + if (engine->schedule) + enabled |= (I915_SCHEDULER_CAP_ENABLED | + I915_SCHEDULER_CAP_PRIORITY); + else + disabled |= (I915_SCHEDULER_CAP_ENABLED | + I915_SCHEDULER_CAP_PRIORITY); + + for (i = 0; i < ARRAY_SIZE(map); i++) { + if (engine->flags & BIT(map[i].engine)) + enabled |= BIT(map[i].sched); + else + disabled |= BIT(map[i].sched); + } + } + + i915->caps.scheduler = enabled & ~disabled; + if (!(i915->caps.scheduler & I915_SCHEDULER_CAP_ENABLED)) + i915->caps.scheduler = 0; +} + +void intel_engines_driver_register(struct drm_i915_private *i915) +{ + u8 uabi_instances[4] = {}; + struct list_head *it, *next; + struct rb_node **p, *prev; + LIST_HEAD(engines); + + sort_engines(i915, &engines); + + prev = NULL; + p = &i915->uabi_engines.rb_node; + list_for_each_safe(it, next, &engines) { + struct intel_engine_cs *engine = + container_of((struct rb_node *)it, typeof(*engine), + uabi_node); + + GEM_BUG_ON(engine->class >= ARRAY_SIZE(uabi_classes)); + engine->uabi_class = uabi_classes[engine->class]; + + GEM_BUG_ON(engine->uabi_class >= ARRAY_SIZE(uabi_instances)); + engine->uabi_instance = uabi_instances[engine->uabi_class]++; + + rb_link_node(&engine->uabi_node, prev, p); + rb_insert_color(&engine->uabi_node, &i915->uabi_engines); + + GEM_BUG_ON(intel_engine_lookup_user(i915, + engine->uabi_class, + engine->uabi_instance) != engine); + + prev = &engine->uabi_node; + p = &prev->rb_right; + } + + if (IS_ENABLED(CONFIG_DRM_I915_SELFTESTS) && + IS_ENABLED(CONFIG_DRM_I915_DEBUG_GEM)) { + struct intel_engine_cs *engine; + unsigned int isolation; + int class, inst; + int errors = 0; + + for (class = 0; class < ARRAY_SIZE(uabi_instances); class++) { + for (inst = 0; inst < uabi_instances[class]; inst++) { + engine = intel_engine_lookup_user(i915, + class, inst); + if (!engine) { + pr_err("UABI engine not found for { class:%d, instance:%d }\n", + class, inst); + errors++; + continue; + } + + if (engine->uabi_class != class || + engine->uabi_instance != inst) { + pr_err("Wrong UABI engine:%s { class:%d, instance:%d } found for { class:%d, instance:%d }\n", + engine->name, + engine->uabi_class, + engine->uabi_instance, + class, inst); + errors++; + continue; + } + } + } + + /* + * Make sure that classes with multiple engine instances all + * share the same basic configuration. + */ + isolation = intel_engines_has_context_isolation(i915); + for_each_uabi_engine(engine, i915) { + unsigned int bit = BIT(engine->uabi_class); + unsigned int expected = engine->default_state ? bit : 0; + + if ((isolation & bit) != expected) { + pr_err("mismatching default context state for class %d on engine %s\n", + engine->uabi_class, engine->name); + errors++; + } + } + + if (WARN(errors, "Invalid UABI engine mapping found")) + i915->uabi_engines = RB_ROOT; + } + + set_scheduler_caps(i915); +} + +unsigned int intel_engines_has_context_isolation(struct drm_i915_private *i915) +{ + struct intel_engine_cs *engine; + unsigned int which; + + which = 0; + for_each_uabi_engine(engine, i915) + if (engine->default_state) + which |= BIT(engine->uabi_class); + + return which; +} diff --git a/drivers/gpu/drm/i915/gt/intel_engine_user.h b/drivers/gpu/drm/i915/gt/intel_engine_user.h new file mode 100644 index 000000000000..9e5f113e5027 --- /dev/null +++ b/drivers/gpu/drm/i915/gt/intel_engine_user.h @@ -0,0 +1,23 @@ +/* + * SPDX-License-Identifier: MIT + * + * Copyright © 2019 Intel Corporation + */ + +#ifndef INTEL_ENGINE_USER_H +#define INTEL_ENGINE_USER_H + +#include + +struct drm_i915_private; +struct intel_engine_cs; + +struct intel_engine_cs * +intel_engine_lookup_user(struct drm_i915_private *i915, u8 class, u8 instance); + +unsigned int intel_engines_has_context_isolation(struct drm_i915_private *i915); + +void intel_engine_add_user(struct intel_engine_cs *engine); +void intel_engines_driver_register(struct drm_i915_private *i915); + +#endif /* INTEL_ENGINE_USER_H */ diff --git a/drivers/gpu/drm/i915/gt/intel_gt_types.h b/drivers/gpu/drm/i915/gt/intel_gt_types.h index 34d4a868e4f1..5fd11e361d03 100644 --- a/drivers/gpu/drm/i915/gt/intel_gt_types.h +++ b/drivers/gpu/drm/i915/gt/intel_gt_types.h @@ -21,6 +21,7 @@ struct drm_i915_private; struct i915_ggtt; +struct intel_engine_cs; struct intel_uncore; struct intel_hangcheck { @@ -76,6 +77,9 @@ struct intel_gt { u32 pm_ier; u32 pm_guc_events; + + struct intel_engine_cs *engine_class[MAX_ENGINE_CLASS + 1] + [MAX_ENGINE_INSTANCE + 1]; }; enum intel_gt_scratch_field { diff --git a/drivers/gpu/drm/i915/gt/selftest_lrc.c b/drivers/gpu/drm/i915/gt/selftest_lrc.c index b40b57d2daae..b26225751a54 100644 --- a/drivers/gpu/drm/i915/gt/selftest_lrc.c +++ b/drivers/gpu/drm/i915/gt/selftest_lrc.c @@ -1792,6 +1792,7 @@ static int live_virtual_engine(void *arg) struct drm_i915_private *i915 = arg; struct intel_engine_cs *siblings[MAX_ENGINE_INSTANCE + 1]; struct intel_engine_cs *engine; + struct intel_gt *gt = &i915->gt; enum intel_engine_id id; unsigned int class, inst; int err = -ENODEV; @@ -1815,10 +1816,10 @@ static int live_virtual_engine(void *arg) nsibling = 0; for (inst = 0; inst <= MAX_ENGINE_INSTANCE; inst++) { - if (!i915->engine_class[class][inst]) + if (!gt->engine_class[class][inst]) continue; - siblings[nsibling++] = i915->engine_class[class][inst]; + siblings[nsibling++] = gt->engine_class[class][inst]; } if (nsibling < 2) continue; @@ -1939,6 +1940,7 @@ static int live_virtual_mask(void *arg) { struct drm_i915_private *i915 = arg; struct intel_engine_cs *siblings[MAX_ENGINE_INSTANCE + 1]; + struct intel_gt *gt = &i915->gt; unsigned int class, inst; int err = 0; @@ -1952,10 +1954,10 @@ static int live_virtual_mask(void *arg) nsibling = 0; for (inst = 0; inst <= MAX_ENGINE_INSTANCE; inst++) { - if (!i915->engine_class[class][inst]) + if (!gt->engine_class[class][inst]) break; - siblings[nsibling++] = i915->engine_class[class][inst]; + siblings[nsibling++] = gt->engine_class[class][inst]; } if (nsibling < 2) continue; @@ -2116,6 +2118,7 @@ static int live_virtual_bond(void *arg) }; struct drm_i915_private *i915 = arg; struct intel_engine_cs *siblings[MAX_ENGINE_INSTANCE + 1]; + struct intel_gt *gt = &i915->gt; unsigned int class, inst; int err = 0; @@ -2130,11 +2133,11 @@ static int live_virtual_bond(void *arg) nsibling = 0; for (inst = 0; inst <= MAX_ENGINE_INSTANCE; inst++) { - if (!i915->engine_class[class][inst]) + if (!gt->engine_class[class][inst]) break; GEM_BUG_ON(nsibling == ARRAY_SIZE(siblings)); - siblings[nsibling++] = i915->engine_class[class][inst]; + siblings[nsibling++] = gt->engine_class[class][inst]; } if (nsibling < 2) continue; diff --git a/drivers/gpu/drm/i915/gt/selftest_workarounds.c b/drivers/gpu/drm/i915/gt/selftest_workarounds.c index 997da94821d9..5efd2bb89d5f 100644 --- a/drivers/gpu/drm/i915/gt/selftest_workarounds.c +++ b/drivers/gpu/drm/i915/gt/selftest_workarounds.c @@ -5,6 +5,7 @@ */ #include "gem/i915_gem_pm.h" +#include "gt/intel_engine_user.h" #include "gt/intel_gt.h" #include "i915_selftest.h" #include "intel_reset.h" 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 412892096daa..a78bd99bc6cb 100644 --- a/drivers/gpu/drm/i915/gt/uc/intel_guc_submission.c +++ b/drivers/gpu/drm/i915/gt/uc/intel_guc_submission.c @@ -488,7 +488,7 @@ static void guc_add_request(struct intel_guc *guc, struct i915_request *rq) ring_tail, rq->fence.seqno); guc_ring_doorbell(client); - client->submissions[engine->id] += 1; + client->submissions[engine->guc_id] += 1; } /* diff --git a/drivers/gpu/drm/i915/i915_cmd_parser.c b/drivers/gpu/drm/i915/i915_cmd_parser.c index a28bcd2d7c09..cb4a0ca72efe 100644 --- a/drivers/gpu/drm/i915/i915_cmd_parser.c +++ b/drivers/gpu/drm/i915/i915_cmd_parser.c @@ -1352,11 +1352,10 @@ int intel_engine_cmd_parser(struct intel_engine_cs *engine, int i915_cmd_parser_get_version(struct drm_i915_private *dev_priv) { struct intel_engine_cs *engine; - enum intel_engine_id id; bool active = false; /* If the command parser is not enabled, report 0 - unsupported */ - for_each_engine(engine, dev_priv, id) { + for_each_uabi_engine(engine, dev_priv) { if (intel_engine_needs_cmd_parser(engine)) { active = true; break; diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c index 8953336f2ae5..63b599f584db 100644 --- a/drivers/gpu/drm/i915/i915_debugfs.c +++ b/drivers/gpu/drm/i915/i915_debugfs.c @@ -415,7 +415,6 @@ static int i915_interrupt_info(struct seq_file *m, void *data) { struct drm_i915_private *dev_priv = node_to_i915(m->private); struct intel_engine_cs *engine; - enum intel_engine_id id; intel_wakeref_t wakeref; int i, pipe; @@ -618,7 +617,7 @@ static int i915_interrupt_info(struct seq_file *m, void *data) I915_READ(GEN11_GUNIT_CSME_INTR_MASK)); } else if (INTEL_GEN(dev_priv) >= 6) { - for_each_engine(engine, dev_priv, id) { + for_each_uabi_engine(engine, dev_priv) { seq_printf(m, "Graphics Interrupt mask (%s): %08x\n", engine->name, ENGINE_READ(engine, RING_IMR)); @@ -1890,7 +1889,6 @@ static void i915_guc_client_info(struct seq_file *m, struct intel_guc_client *client) { struct intel_engine_cs *engine; - enum intel_engine_id id; u64 tot = 0; seq_printf(m, "\tPriority %d, GuC stage index: %u, PD offset 0x%x\n", @@ -1898,8 +1896,8 @@ static void i915_guc_client_info(struct seq_file *m, seq_printf(m, "\tDoorbell id %d, offset: 0x%lx\n", client->doorbell_id, client->doorbell_offset); - for_each_engine(engine, dev_priv, id) { - u64 submissions = client->submissions[id]; + for_each_uabi_engine(engine, dev_priv) { + u64 submissions = client->submissions[engine->guc_id]; tot += submissions; seq_printf(m, "\tSubmissions: %llu %s\n", submissions, engine->name); @@ -1939,7 +1937,6 @@ static int i915_guc_stage_pool(struct seq_file *m, void *data) struct drm_i915_private *dev_priv = node_to_i915(m->private); const struct intel_guc *guc = &dev_priv->gt.uc.guc; struct guc_stage_desc *desc = guc->stage_desc_pool_vaddr; - intel_engine_mask_t tmp; int index; if (!USES_GUC_SUBMISSION(dev_priv)) @@ -1968,7 +1965,7 @@ static int i915_guc_stage_pool(struct seq_file *m, void *data) desc->wq_addr, desc->wq_size); seq_putc(m, '\n'); - for_each_engine(engine, dev_priv, tmp) { + for_each_uabi_engine(engine, dev_priv) { u32 guc_engine_id = engine->guc_id; struct guc_execlist_context *lrc = &desc->lrc[guc_engine_id]; @@ -2806,7 +2803,6 @@ static int i915_engine_info(struct seq_file *m, void *unused) struct drm_i915_private *dev_priv = node_to_i915(m->private); struct intel_engine_cs *engine; intel_wakeref_t wakeref; - enum intel_engine_id id; struct drm_printer p; wakeref = intel_runtime_pm_get(&dev_priv->runtime_pm); @@ -2818,7 +2814,7 @@ static int i915_engine_info(struct seq_file *m, void *unused) RUNTIME_INFO(dev_priv)->cs_timestamp_frequency_khz); p = drm_seq_file_printer(m); - for_each_engine(engine, dev_priv, id) + for_each_uabi_engine(engine, dev_priv) intel_engine_dump(engine, &p, "%s\n", engine->name); intel_runtime_pm_put(&dev_priv->runtime_pm, wakeref); @@ -2899,9 +2895,8 @@ static int i915_wa_registers(struct seq_file *m, void *unused) { struct drm_i915_private *i915 = node_to_i915(m->private); struct intel_engine_cs *engine; - enum intel_engine_id id; - for_each_engine(engine, i915, id) { + for_each_uabi_engine(engine, i915) { const struct i915_wa_list *wal = &engine->ctx_wa_list; const struct i915_wa *wa; unsigned int count; diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index ec61e8a7c818..bea440ac8b0f 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -61,6 +61,7 @@ #include "gem/i915_gem_context.h" #include "gem/i915_gem_ioctls.h" +#include "gt/intel_engine_user.h" #include "gt/intel_gt.h" #include "gt/intel_gt_pm.h" #include "gt/intel_reset.h" diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index cb64dd9b6e64..0c4db51cdf6e 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -1376,11 +1376,12 @@ struct drm_i915_private { wait_queue_head_t gmbus_wait_queue; struct pci_dev *bridge_dev; - struct intel_engine_cs *engine[I915_NUM_ENGINES]; + /* Context used internally to idle the GPU and setup initial state */ struct i915_gem_context *kernel_context; - struct intel_engine_cs *engine_class[MAX_ENGINE_CLASS + 1] - [MAX_ENGINE_INSTANCE + 1]; + + struct intel_engine_cs *engine[I915_NUM_ENGINES]; + struct rb_root uabi_engines; struct resource mch_res; @@ -1924,6 +1925,14 @@ static inline struct drm_i915_private *wopcm_to_i915(struct intel_wopcm *wopcm) ((engine__) = (dev_priv__)->engine[__mask_next_bit(tmp__)]), 1 : \ 0;) +#define rb_to_uabi_engine(rb) \ + rb_entry_safe(rb, struct intel_engine_cs, uabi_node) + +#define for_each_uabi_engine(engine__, i915__) \ + for ((engine__) = rb_to_uabi_engine(rb_first(&(i915__)->uabi_engines));\ + (engine__); \ + (engine__) = rb_to_uabi_engine(rb_next(&(engine__)->uabi_node))) + enum hdmi_force_audio { HDMI_AUDIO_OFF_DVI = -2, /* no aux data for HDMI-DVI converter */ HDMI_AUDIO_OFF, /* force turn off HDMI audio */ diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index 5ab1ddfef23c..7efff14b9137 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -46,6 +46,7 @@ #include "gem/i915_gem_ioctls.h" #include "gem/i915_gem_pm.h" #include "gem/i915_gemfs.h" +#include "gt/intel_engine_user.h" #include "gt/intel_gt.h" #include "gt/intel_gt_pm.h" #include "gt/intel_mocs.h" @@ -1359,24 +1360,6 @@ err_rq: i915_gem_object_unpin_map(engine->default_state); } - if (IS_ENABLED(CONFIG_DRM_I915_DEBUG_GEM)) { - unsigned int found = intel_engines_has_context_isolation(i915); - - /* - * Make sure that classes with multiple engine instances all - * share the same basic configuration. - */ - for_each_engine(engine, i915, id) { - unsigned int bit = BIT(engine->uabi_class); - unsigned int expected = engine->default_state ? bit : 0; - - if ((found & bit) != expected) { - DRM_ERROR("mismatching default context state for class %d on engine %s\n", - engine->uabi_class, engine->name); - } - } - } - out_ctx: i915_gem_context_unlock_engines(ctx); i915_gem_context_set_closed(ctx); @@ -1600,7 +1583,8 @@ err_unlock: void i915_gem_driver_register(struct drm_i915_private *i915) { i915_gem_driver_register__shrinker(i915); - intel_engines_set_scheduler_caps(i915); + + intel_engines_driver_register(i915); } void i915_gem_driver_unregister(struct drm_i915_private *i915) diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c index 33ac5d7e1e9e..9094c4811f54 100644 --- a/drivers/gpu/drm/i915/i915_irq.c +++ b/drivers/gpu/drm/i915/i915_irq.c @@ -3251,7 +3251,7 @@ gen11_engine_irq_handler(struct intel_gt *gt, const u8 class, struct intel_engine_cs *engine; if (instance <= MAX_ENGINE_INSTANCE) - engine = gt->i915->engine_class[class][instance]; + engine = gt->engine_class[class][instance]; else engine = NULL; diff --git a/drivers/gpu/drm/i915/i915_perf.c b/drivers/gpu/drm/i915/i915_perf.c index f2cc69ccb635..2bcf0d6e2ee2 100644 --- a/drivers/gpu/drm/i915/i915_perf.c +++ b/drivers/gpu/drm/i915/i915_perf.c @@ -1875,7 +1875,6 @@ static int gen8_configure_all_contexts(struct drm_i915_private *i915, #undef ctx_flexeuN struct intel_engine_cs *engine; struct i915_gem_context *ctx; - enum intel_engine_id id; int i; for (i = 2; i < ARRAY_SIZE(regs); i++) @@ -1915,7 +1914,7 @@ static int gen8_configure_all_contexts(struct drm_i915_private *i915, * If we don't modify the kernel_context, we do not get events while * idle. */ - for_each_engine(engine, i915, id) { + for_each_uabi_engine(engine, i915) { struct intel_context *ce = engine->kernel_context; int err; diff --git a/drivers/gpu/drm/i915/i915_pmu.c b/drivers/gpu/drm/i915/i915_pmu.c index c2e5f6d5c1e0..c7ee0ab180e8 100644 --- a/drivers/gpu/drm/i915/i915_pmu.c +++ b/drivers/gpu/drm/i915/i915_pmu.c @@ -9,6 +9,7 @@ #include "gt/intel_engine.h" #include "gt/intel_engine_pm.h" +#include "gt/intel_engine_user.h" #include "gt/intel_gt_pm.h" #include "i915_drv.h" @@ -860,7 +861,6 @@ create_event_attributes(struct i915_pmu *pmu) struct i915_ext_attribute *i915_attr = NULL, *i915_iter; struct attribute **attr = NULL, **attr_iter; struct intel_engine_cs *engine; - enum intel_engine_id id; unsigned int i; /* Count how many counters we will be exposing. */ @@ -869,7 +869,7 @@ create_event_attributes(struct i915_pmu *pmu) count++; } - for_each_engine(engine, i915, id) { + for_each_uabi_engine(engine, i915) { for (i = 0; i < ARRAY_SIZE(engine_events); i++) { if (!engine_event_status(engine, engine_events[i].sample)) @@ -920,7 +920,7 @@ create_event_attributes(struct i915_pmu *pmu) } /* Initialize supported engine counters. */ - for_each_engine(engine, i915, id) { + for_each_uabi_engine(engine, i915) { for (i = 0; i < ARRAY_SIZE(engine_events); i++) { char *str; @@ -937,7 +937,7 @@ create_event_attributes(struct i915_pmu *pmu) i915_iter = add_i915_attr(i915_iter, str, __I915_PMU_ENGINE(engine->uabi_class, - engine->instance, + engine->uabi_instance, engine_events[i].sample)); str = kasprintf(GFP_KERNEL, "%s-%s.unit", diff --git a/drivers/gpu/drm/i915/i915_query.c b/drivers/gpu/drm/i915/i915_query.c index 7b7016171057..ad9240a0817a 100644 --- a/drivers/gpu/drm/i915/i915_query.c +++ b/drivers/gpu/drm/i915/i915_query.c @@ -105,7 +105,6 @@ query_engine_info(struct drm_i915_private *i915, struct drm_i915_query_engine_info query; struct drm_i915_engine_info info = { }; struct intel_engine_cs *engine; - enum intel_engine_id id; int len, ret; if (query_item->flags) @@ -125,9 +124,9 @@ query_engine_info(struct drm_i915_private *i915, info_ptr = &query_ptr->engines[0]; - for_each_engine(engine, i915, id) { + for_each_uabi_engine(engine, i915) { info.engine.engine_class = engine->uabi_class; - info.engine.engine_instance = engine->instance; + info.engine.engine_instance = engine->uabi_instance; info.capabilities = engine->uabi_capabilities; if (__copy_to_user(info_ptr, &info, sizeof(info))) diff --git a/drivers/gpu/drm/i915/i915_trace.h b/drivers/gpu/drm/i915/i915_trace.h index da18b8d6b80c..1d11245c4c87 100644 --- a/drivers/gpu/drm/i915/i915_trace.h +++ b/drivers/gpu/drm/i915/i915_trace.h @@ -677,7 +677,7 @@ TRACE_EVENT(i915_request_queue, __entry->dev = rq->i915->drm.primary->index; __entry->hw_id = rq->gem_context->hw_id; __entry->class = rq->engine->uabi_class; - __entry->instance = rq->engine->instance; + __entry->instance = rq->engine->uabi_instance; __entry->ctx = rq->fence.context; __entry->seqno = rq->fence.seqno; __entry->flags = flags; @@ -706,7 +706,7 @@ DECLARE_EVENT_CLASS(i915_request, __entry->dev = rq->i915->drm.primary->index; __entry->hw_id = rq->gem_context->hw_id; __entry->class = rq->engine->uabi_class; - __entry->instance = rq->engine->instance; + __entry->instance = rq->engine->uabi_instance; __entry->ctx = rq->fence.context; __entry->seqno = rq->fence.seqno; ), @@ -751,7 +751,7 @@ TRACE_EVENT(i915_request_in, __entry->dev = rq->i915->drm.primary->index; __entry->hw_id = rq->gem_context->hw_id; __entry->class = rq->engine->uabi_class; - __entry->instance = rq->engine->instance; + __entry->instance = rq->engine->uabi_instance; __entry->ctx = rq->fence.context; __entry->seqno = rq->fence.seqno; __entry->prio = rq->sched.attr.priority; @@ -782,7 +782,7 @@ TRACE_EVENT(i915_request_out, __entry->dev = rq->i915->drm.primary->index; __entry->hw_id = rq->gem_context->hw_id; __entry->class = rq->engine->uabi_class; - __entry->instance = rq->engine->instance; + __entry->instance = rq->engine->uabi_instance; __entry->ctx = rq->fence.context; __entry->seqno = rq->fence.seqno; __entry->completed = i915_request_completed(rq); @@ -847,7 +847,7 @@ TRACE_EVENT(i915_request_wait_begin, __entry->dev = rq->i915->drm.primary->index; __entry->hw_id = rq->gem_context->hw_id; __entry->class = rq->engine->uabi_class; - __entry->instance = rq->engine->instance; + __entry->instance = rq->engine->uabi_instance; __entry->ctx = rq->fence.context; __entry->seqno = rq->fence.seqno; __entry->flags = flags; -- cgit v1.2.3 From 75d0a7f31eec8ec4a53b4485905800e09dc5091f Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Fri, 9 Aug 2019 19:25:18 +0100 Subject: drm/i915: Lift timeline into intel_context Move the timeline from being inside the intel_ring to intel_context itself. This saves much pointer dancing and makes the relations of the context to its timeline much clearer. Signed-off-by: Chris Wilson Reviewed-by: Matthew Auld Link: https://patchwork.freedesktop.org/patch/msgid/20190809182518.20486-4-chris@chris-wilson.co.uk --- drivers/gpu/drm/i915/gem/i915_gem_context.c | 35 ++++++++++---- drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c | 2 +- drivers/gpu/drm/i915/gt/intel_context.c | 21 +++++++-- drivers/gpu/drm/i915/gt/intel_context.h | 8 ++-- drivers/gpu/drm/i915/gt/intel_context_types.h | 1 + drivers/gpu/drm/i915/gt/intel_engine.h | 4 +- drivers/gpu/drm/i915/gt/intel_engine_cs.c | 1 - drivers/gpu/drm/i915/gt/intel_engine_types.h | 8 ++-- drivers/gpu/drm/i915/gt/intel_lrc.c | 33 +++++-------- drivers/gpu/drm/i915/gt/intel_ringbuffer.c | 63 ++++++++++++------------- drivers/gpu/drm/i915/gt/mock_engine.c | 62 +++++++++--------------- drivers/gpu/drm/i915/gt/selftest_context.c | 2 +- drivers/gpu/drm/i915/i915_active.c | 6 +-- drivers/gpu/drm/i915/i915_request.c | 10 ++-- drivers/gpu/drm/i915/selftests/i915_gem_evict.c | 19 ++++---- 15 files changed, 140 insertions(+), 135 deletions(-) (limited to 'drivers/gpu/drm/i915/gt/intel_engine.h') diff --git a/drivers/gpu/drm/i915/gem/i915_gem_context.c b/drivers/gpu/drm/i915/gem/i915_gem_context.c index d1952637cd91..a6b0cb714292 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_context.c +++ b/drivers/gpu/drm/i915/gem/i915_gem_context.c @@ -489,6 +489,29 @@ static void __assign_ppgtt(struct i915_gem_context *ctx, i915_vm_put(vm); } +static void __set_timeline(struct intel_timeline **dst, + struct intel_timeline *src) +{ + struct intel_timeline *old = *dst; + + *dst = src ? intel_timeline_get(src) : NULL; + + if (old) + intel_timeline_put(old); +} + +static void __apply_timeline(struct intel_context *ce, void *timeline) +{ + __set_timeline(&ce->timeline, timeline); +} + +static void __assign_timeline(struct i915_gem_context *ctx, + struct intel_timeline *timeline) +{ + __set_timeline(&ctx->timeline, timeline); + context_apply_all(ctx, __apply_timeline, timeline); +} + static struct i915_gem_context * i915_gem_create_context(struct drm_i915_private *dev_priv, unsigned int flags) { @@ -531,7 +554,8 @@ i915_gem_create_context(struct drm_i915_private *dev_priv, unsigned int flags) return ERR_CAST(timeline); } - ctx->timeline = timeline; + __assign_timeline(ctx, timeline); + intel_timeline_put(timeline); } trace_i915_context_create(ctx); @@ -1931,13 +1955,8 @@ unlock: static int clone_timeline(struct i915_gem_context *dst, struct i915_gem_context *src) { - if (src->timeline) { - GEM_BUG_ON(src->timeline == dst->timeline); - - if (dst->timeline) - intel_timeline_put(dst->timeline); - dst->timeline = intel_timeline_get(src->timeline); - } + if (src->timeline) + __assign_timeline(dst, src->timeline); return 0; } diff --git a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c index 2fa08357944e..1bd2187ac8d6 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c +++ b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c @@ -2182,7 +2182,7 @@ err_unpin: static void eb_unpin_context(struct i915_execbuffer *eb) { struct intel_context *ce = eb->context; - struct intel_timeline *tl = ce->ring->timeline; + struct intel_timeline *tl = ce->timeline; mutex_lock(&tl->mutex); intel_context_exit(ce); diff --git a/drivers/gpu/drm/i915/gt/intel_context.c b/drivers/gpu/drm/i915/gt/intel_context.c index 6d1d4e8dbfc9..77833f1558a9 100644 --- a/drivers/gpu/drm/i915/gt/intel_context.c +++ b/drivers/gpu/drm/i915/gt/intel_context.c @@ -68,7 +68,7 @@ int __intel_context_do_pin(struct intel_context *ce) goto err; GEM_TRACE("%s context:%llx pin ring:{head:%04x, tail:%04x}\n", - ce->engine->name, ce->ring->timeline->fence_context, + ce->engine->name, ce->timeline->fence_context, ce->ring->head, ce->ring->tail); i915_gem_context_get(ce->gem_context); /* for ctx->ppgtt */ @@ -98,7 +98,7 @@ void intel_context_unpin(struct intel_context *ce) if (likely(atomic_dec_and_test(&ce->pin_count))) { GEM_TRACE("%s context:%llx retire\n", - ce->engine->name, ce->ring->timeline->fence_context); + ce->engine->name, ce->timeline->fence_context); ce->ops->unpin(ce); @@ -143,11 +143,12 @@ static void __intel_context_retire(struct i915_active *active) struct intel_context *ce = container_of(active, typeof(*ce), active); GEM_TRACE("%s context:%llx retire\n", - ce->engine->name, ce->ring->timeline->fence_context); + ce->engine->name, ce->timeline->fence_context); if (ce->state) __context_unpin_state(ce->state); + intel_timeline_unpin(ce->timeline); intel_ring_unpin(ce->ring); intel_context_put(ce); } @@ -163,15 +164,21 @@ static int __intel_context_active(struct i915_active *active) if (err) goto err_put; + err = intel_timeline_pin(ce->timeline); + if (err) + goto err_ring; + if (!ce->state) return 0; err = __context_pin_state(ce->state); if (err) - goto err_ring; + goto err_timeline; return 0; +err_timeline: + intel_timeline_unpin(ce->timeline); err_ring: intel_ring_unpin(ce->ring); err_put: @@ -218,6 +225,8 @@ intel_context_init(struct intel_context *ce, ce->gem_context = ctx; ce->vm = i915_vm_get(ctx->vm ?: &engine->gt->ggtt->vm); + if (ctx->timeline) + ce->timeline = intel_timeline_get(ctx->timeline); ce->engine = engine; ce->ops = engine->cops; @@ -235,6 +244,8 @@ intel_context_init(struct intel_context *ce, void intel_context_fini(struct intel_context *ce) { + if (ce->timeline) + intel_timeline_put(ce->timeline); i915_vm_put(ce->vm); mutex_destroy(&ce->pin_mutex); @@ -279,7 +290,7 @@ void intel_context_exit_engine(struct intel_context *ce) int intel_context_prepare_remote_request(struct intel_context *ce, struct i915_request *rq) { - struct intel_timeline *tl = ce->ring->timeline; + struct intel_timeline *tl = ce->timeline; int err; /* Only suitable for use in remotely modifying this context */ diff --git a/drivers/gpu/drm/i915/gt/intel_context.h b/drivers/gpu/drm/i915/gt/intel_context.h index 13f28dd316bc..9fa8b588f18e 100644 --- a/drivers/gpu/drm/i915/gt/intel_context.h +++ b/drivers/gpu/drm/i915/gt/intel_context.h @@ -120,15 +120,15 @@ static inline void intel_context_put(struct intel_context *ce) static inline int __must_check intel_context_timeline_lock(struct intel_context *ce) - __acquires(&ce->ring->timeline->mutex) + __acquires(&ce->timeline->mutex) { - return mutex_lock_interruptible(&ce->ring->timeline->mutex); + return mutex_lock_interruptible(&ce->timeline->mutex); } static inline void intel_context_timeline_unlock(struct intel_context *ce) - __releases(&ce->ring->timeline->mutex) + __releases(&ce->timeline->mutex) { - mutex_unlock(&ce->ring->timeline->mutex); + mutex_unlock(&ce->timeline->mutex); } int intel_context_prepare_remote_request(struct intel_context *ce, diff --git a/drivers/gpu/drm/i915/gt/intel_context_types.h b/drivers/gpu/drm/i915/gt/intel_context_types.h index cff6238c213a..a632b20ec4d8 100644 --- a/drivers/gpu/drm/i915/gt/intel_context_types.h +++ b/drivers/gpu/drm/i915/gt/intel_context_types.h @@ -53,6 +53,7 @@ struct intel_context { struct i915_vma *state; struct intel_ring *ring; + struct intel_timeline *timeline; unsigned long flags; #define CONTEXT_ALLOC_BIT 0 diff --git a/drivers/gpu/drm/i915/gt/intel_engine.h b/drivers/gpu/drm/i915/gt/intel_engine.h index 37c391cee441..e1228b0e577f 100644 --- a/drivers/gpu/drm/i915/gt/intel_engine.h +++ b/drivers/gpu/drm/i915/gt/intel_engine.h @@ -196,9 +196,7 @@ intel_write_status_page(struct intel_engine_cs *engine, int reg, u32 value) #define CNL_HWS_CSB_WRITE_INDEX 0x2f struct intel_ring * -intel_engine_create_ring(struct intel_engine_cs *engine, - struct intel_timeline *timeline, - int size); +intel_engine_create_ring(struct intel_engine_cs *engine, int size); int intel_ring_pin(struct intel_ring *ring); void intel_ring_reset(struct intel_ring *ring, u32 tail); unsigned int intel_ring_update_space(struct intel_ring *ring); diff --git a/drivers/gpu/drm/i915/gt/intel_engine_cs.c b/drivers/gpu/drm/i915/gt/intel_engine_cs.c index ece731ab7b3a..7d174af30f8c 100644 --- a/drivers/gpu/drm/i915/gt/intel_engine_cs.c +++ b/drivers/gpu/drm/i915/gt/intel_engine_cs.c @@ -680,7 +680,6 @@ static int measure_breadcrumb_dw(struct intel_engine_cs *engine) goto out_frame; INIT_LIST_HEAD(&frame->ring.request_list); - frame->ring.timeline = &frame->timeline; frame->ring.vaddr = frame->cs; frame->ring.size = sizeof(frame->cs); frame->ring.effective_size = frame->ring.size; diff --git a/drivers/gpu/drm/i915/gt/intel_engine_types.h b/drivers/gpu/drm/i915/gt/intel_engine_types.h index c79fd1dafa88..a0f372807dd4 100644 --- a/drivers/gpu/drm/i915/gt/intel_engine_types.h +++ b/drivers/gpu/drm/i915/gt/intel_engine_types.h @@ -69,7 +69,6 @@ struct intel_ring { struct i915_vma *vma; void *vaddr; - struct intel_timeline *timeline; struct list_head request_list; struct list_head active_link; @@ -286,8 +285,6 @@ struct intel_engine_cs { struct intel_sseu sseu; - struct intel_ring *buffer; - struct { spinlock_t lock; struct list_head requests; @@ -306,6 +303,11 @@ struct intel_engine_cs { struct drm_i915_gem_object *default_state; void *pinned_default_state; + struct { + struct intel_ring *ring; + struct intel_timeline *timeline; + } legacy; + /* Rather than have every client wait upon all user interrupts, * with the herd waking after every interrupt and each doing the * heavyweight seqno dance, we delegate the task (of being the diff --git a/drivers/gpu/drm/i915/gt/intel_lrc.c b/drivers/gpu/drm/i915/gt/intel_lrc.c index fc5fcc2d03ac..bb74954889dd 100644 --- a/drivers/gpu/drm/i915/gt/intel_lrc.c +++ b/drivers/gpu/drm/i915/gt/intel_lrc.c @@ -2821,9 +2821,6 @@ logical_ring_default_irqs(struct intel_engine_cs *engine) int intel_execlists_submission_setup(struct intel_engine_cs *engine) { - /* Intentionally left blank. */ - engine->buffer = NULL; - tasklet_init(&engine->execlists.tasklet, execlists_submission_tasklet, (unsigned long)engine); timer_setup(&engine->execlists.timer, execlists_submission_timer, 0); @@ -3071,23 +3068,13 @@ err_unpin_ctx: return ret; } -static struct intel_timeline * -get_timeline(struct i915_gem_context *ctx, struct intel_gt *gt) -{ - if (ctx->timeline) - return intel_timeline_get(ctx->timeline); - else - return intel_timeline_create(gt, NULL); -} - static int __execlists_context_alloc(struct intel_context *ce, struct intel_engine_cs *engine) { struct drm_i915_gem_object *ctx_obj; + struct intel_ring *ring; struct i915_vma *vma; u32 context_size; - struct intel_ring *ring; - struct intel_timeline *timeline; int ret; GEM_BUG_ON(ce->state); @@ -3109,15 +3096,19 @@ static int __execlists_context_alloc(struct intel_context *ce, goto error_deref_obj; } - timeline = get_timeline(ce->gem_context, engine->gt); - if (IS_ERR(timeline)) { - ret = PTR_ERR(timeline); - goto error_deref_obj; + if (!ce->timeline) { + struct intel_timeline *tl; + + tl = intel_timeline_create(engine->gt, NULL); + if (IS_ERR(tl)) { + ret = PTR_ERR(tl); + goto error_deref_obj; + } + + ce->timeline = tl; } - ring = intel_engine_create_ring(engine, timeline, - (unsigned long)ce->ring); - intel_timeline_put(timeline); + ring = intel_engine_create_ring(engine, (unsigned long)ce->ring); if (IS_ERR(ring)) { ret = PTR_ERR(ring); goto error_deref_obj; diff --git a/drivers/gpu/drm/i915/gt/intel_ringbuffer.c b/drivers/gpu/drm/i915/gt/intel_ringbuffer.c index d5b7e4dde762..dfacb265a995 100644 --- a/drivers/gpu/drm/i915/gt/intel_ringbuffer.c +++ b/drivers/gpu/drm/i915/gt/intel_ringbuffer.c @@ -636,7 +636,7 @@ static bool stop_ring(struct intel_engine_cs *engine) static int xcs_resume(struct intel_engine_cs *engine) { struct drm_i915_private *dev_priv = engine->i915; - struct intel_ring *ring = engine->buffer; + struct intel_ring *ring = engine->legacy.ring; int ret = 0; GEM_TRACE("%s: ring:{HEAD:%04x, TAIL:%04x}\n", @@ -832,12 +832,12 @@ static void reset_ring(struct intel_engine_cs *engine, bool stalled) */ __i915_request_reset(rq, stalled); - GEM_BUG_ON(rq->ring != engine->buffer); + GEM_BUG_ON(rq->ring != engine->legacy.ring); head = rq->head; } else { - head = engine->buffer->tail; + head = engine->legacy.ring->tail; } - engine->buffer->head = intel_ring_wrap(engine->buffer, head); + engine->legacy.ring->head = intel_ring_wrap(engine->legacy.ring, head); spin_unlock_irqrestore(&engine->active.lock, flags); } @@ -1192,10 +1192,6 @@ int intel_ring_pin(struct intel_ring *ring) if (atomic_fetch_inc(&ring->pin_count)) return 0; - ret = intel_timeline_pin(ring->timeline); - if (ret) - goto err_unpin; - flags = PIN_GLOBAL; /* Ring wraparound at offset 0 sometimes hangs. No idea why. */ @@ -1208,7 +1204,7 @@ int intel_ring_pin(struct intel_ring *ring) ret = i915_vma_pin(vma, 0, 0, flags); if (unlikely(ret)) - goto err_timeline; + goto err_unpin; if (i915_vma_is_map_and_fenceable(vma)) addr = (void __force *)i915_vma_pin_iomap(vma); @@ -1225,13 +1221,10 @@ int intel_ring_pin(struct intel_ring *ring) GEM_BUG_ON(ring->vaddr); ring->vaddr = addr; - GEM_TRACE("ring:%llx pin\n", ring->timeline->fence_context); return 0; err_ring: i915_vma_unpin(vma); -err_timeline: - intel_timeline_unpin(ring->timeline); err_unpin: atomic_dec(&ring->pin_count); return ret; @@ -1254,8 +1247,6 @@ void intel_ring_unpin(struct intel_ring *ring) if (!atomic_dec_and_test(&ring->pin_count)) return; - GEM_TRACE("ring:%llx unpin\n", ring->timeline->fence_context); - /* Discard any unused bytes beyond that submitted to hw. */ intel_ring_reset(ring, ring->tail); @@ -1270,8 +1261,6 @@ void intel_ring_unpin(struct intel_ring *ring) i915_vma_unpin(vma); i915_vma_make_purgeable(vma); - - intel_timeline_unpin(ring->timeline); } static struct i915_vma *create_ring_vma(struct i915_ggtt *ggtt, int size) @@ -1306,9 +1295,7 @@ err: } struct intel_ring * -intel_engine_create_ring(struct intel_engine_cs *engine, - struct intel_timeline *timeline, - int size) +intel_engine_create_ring(struct intel_engine_cs *engine, int size) { struct drm_i915_private *i915 = engine->i915; struct intel_ring *ring; @@ -1323,7 +1310,6 @@ intel_engine_create_ring(struct intel_engine_cs *engine, kref_init(&ring->ref); INIT_LIST_HEAD(&ring->request_list); - ring->timeline = intel_timeline_get(timeline); ring->size = size; /* Workaround an erratum on the i830 which causes a hang if @@ -1353,7 +1339,6 @@ void intel_ring_free(struct kref *ref) i915_vma_close(ring->vma); i915_vma_put(ring->vma); - intel_timeline_put(ring->timeline); kfree(ring); } @@ -1485,8 +1470,9 @@ static int ring_context_alloc(struct intel_context *ce) struct intel_engine_cs *engine = ce->engine; /* One ringbuffer to rule them all */ - GEM_BUG_ON(!engine->buffer); - ce->ring = engine->buffer; + GEM_BUG_ON(!engine->legacy.ring); + ce->ring = engine->legacy.ring; + ce->timeline = intel_timeline_get(engine->legacy.timeline); GEM_BUG_ON(ce->state); if (engine->context_size) { @@ -2165,8 +2151,11 @@ static void ring_destroy(struct intel_engine_cs *engine) intel_engine_cleanup_common(engine); - intel_ring_unpin(engine->buffer); - intel_ring_put(engine->buffer); + intel_ring_unpin(engine->legacy.ring); + intel_ring_put(engine->legacy.ring); + + intel_timeline_unpin(engine->legacy.timeline); + intel_timeline_put(engine->legacy.timeline); kfree(engine); } @@ -2350,32 +2339,40 @@ int intel_ring_submission_init(struct intel_engine_cs *engine) } GEM_BUG_ON(timeline->has_initial_breadcrumb); - ring = intel_engine_create_ring(engine, timeline, SZ_16K); - intel_timeline_put(timeline); + err = intel_timeline_pin(timeline); + if (err) + goto err_timeline; + + ring = intel_engine_create_ring(engine, SZ_16K); if (IS_ERR(ring)) { err = PTR_ERR(ring); - goto err; + goto err_timeline_unpin; } err = intel_ring_pin(ring); if (err) goto err_ring; - GEM_BUG_ON(engine->buffer); - engine->buffer = ring; + GEM_BUG_ON(engine->legacy.ring); + engine->legacy.ring = ring; + engine->legacy.timeline = timeline; err = intel_engine_init_common(engine); if (err) - goto err_unpin; + goto err_ring_unpin; - GEM_BUG_ON(ring->timeline->hwsp_ggtt != engine->status_page.vma); + GEM_BUG_ON(timeline->hwsp_ggtt != engine->status_page.vma); return 0; -err_unpin: +err_ring_unpin: intel_ring_unpin(ring); err_ring: intel_ring_put(ring); +err_timeline_unpin: + intel_timeline_unpin(timeline); +err_timeline: + intel_timeline_put(timeline); err: intel_engine_cleanup_common(engine); return err; diff --git a/drivers/gpu/drm/i915/gt/mock_engine.c b/drivers/gpu/drm/i915/gt/mock_engine.c index 79baed1aaf38..a63dd8a42cd4 100644 --- a/drivers/gpu/drm/i915/gt/mock_engine.c +++ b/drivers/gpu/drm/i915/gt/mock_engine.c @@ -32,11 +32,6 @@ #include "mock_engine.h" #include "selftests/mock_request.h" -struct mock_ring { - struct intel_ring base; - struct intel_timeline timeline; -}; - static void mock_timeline_pin(struct intel_timeline *tl) { tl->pin_count++; @@ -51,36 +46,22 @@ static void mock_timeline_unpin(struct intel_timeline *tl) static struct intel_ring *mock_ring(struct intel_engine_cs *engine) { const unsigned long sz = PAGE_SIZE / 2; - struct mock_ring *ring; + struct intel_ring *ring; ring = kzalloc(sizeof(*ring) + sz, GFP_KERNEL); if (!ring) return NULL; - if (intel_timeline_init(&ring->timeline, engine->gt, NULL)) { - kfree(ring); - return NULL; - } - - kref_init(&ring->base.ref); - ring->base.size = sz; - ring->base.effective_size = sz; - ring->base.vaddr = (void *)(ring + 1); - ring->base.timeline = &ring->timeline; - atomic_set(&ring->base.pin_count, 1); - - INIT_LIST_HEAD(&ring->base.request_list); - intel_ring_update_space(&ring->base); - - return &ring->base; -} + kref_init(&ring->ref); + ring->size = sz; + ring->effective_size = sz; + ring->vaddr = (void *)(ring + 1); + atomic_set(&ring->pin_count, 1); -static void mock_ring_free(struct intel_ring *base) -{ - struct mock_ring *ring = container_of(base, typeof(*ring), base); + INIT_LIST_HEAD(&ring->request_list); + intel_ring_update_space(ring); - intel_timeline_fini(&ring->timeline); - kfree(ring); + return ring; } static struct i915_request *first_request(struct mock_engine *engine) @@ -131,7 +112,6 @@ static void hw_delay_complete(struct timer_list *t) static void mock_context_unpin(struct intel_context *ce) { - mock_timeline_unpin(ce->ring->timeline); } static void mock_context_destroy(struct kref *ref) @@ -140,8 +120,10 @@ static void mock_context_destroy(struct kref *ref) GEM_BUG_ON(intel_context_is_pinned(ce)); - if (test_bit(CONTEXT_ALLOC_BIT, &ce->flags)) - mock_ring_free(ce->ring); + if (test_bit(CONTEXT_ALLOC_BIT, &ce->flags)) { + kfree(ce->ring); + mock_timeline_unpin(ce->timeline); + } intel_context_fini(ce); intel_context_free(ce); @@ -153,19 +135,21 @@ static int mock_context_alloc(struct intel_context *ce) if (!ce->ring) return -ENOMEM; + GEM_BUG_ON(ce->timeline); + ce->timeline = intel_timeline_create(ce->engine->gt, NULL); + if (IS_ERR(ce->timeline)) { + kfree(ce->engine); + return PTR_ERR(ce->timeline); + } + + mock_timeline_pin(ce->timeline); + return 0; } static int mock_context_pin(struct intel_context *ce) { - int ret; - - ret = intel_context_active_acquire(ce); - if (ret) - return ret; - - mock_timeline_pin(ce->ring->timeline); - return 0; + return intel_context_active_acquire(ce); } static const struct intel_context_ops mock_context_ops = { diff --git a/drivers/gpu/drm/i915/gt/selftest_context.c b/drivers/gpu/drm/i915/gt/selftest_context.c index 6e7e9a6fd235..da9c49e2adaf 100644 --- a/drivers/gpu/drm/i915/gt/selftest_context.c +++ b/drivers/gpu/drm/i915/gt/selftest_context.c @@ -32,7 +32,7 @@ static int request_sync(struct i915_request *rq) static int context_sync(struct intel_context *ce) { - struct intel_timeline *tl = ce->ring->timeline; + struct intel_timeline *tl = ce->timeline; int err = 0; do { diff --git a/drivers/gpu/drm/i915/i915_active.c b/drivers/gpu/drm/i915/i915_active.c index 1e09722b5317..7698fcaa648a 100644 --- a/drivers/gpu/drm/i915/i915_active.c +++ b/drivers/gpu/drm/i915/i915_active.c @@ -246,7 +246,7 @@ static bool __active_del_barrier(struct i915_active *ref, struct llist_node *head = NULL, *tail = NULL; struct llist_node *pos, *next; - GEM_BUG_ON(node->timeline != engine->kernel_context->ring->timeline->fence_context); + GEM_BUG_ON(node->timeline != engine->kernel_context->timeline->fence_context); /* * Rebuild the llist excluding our node. We may perform this @@ -568,7 +568,7 @@ int i915_active_acquire_preallocate_barrier(struct i915_active *ref, * i915_active_acquire_barrier() */ for_each_engine_masked(engine, i915, mask, tmp) { - u64 idx = engine->kernel_context->ring->timeline->fence_context; + u64 idx = engine->kernel_context->timeline->fence_context; struct active_node *node; node = reuse_idle_barrier(ref, idx); @@ -665,7 +665,7 @@ void i915_request_add_active_barriers(struct i915_request *rq) struct llist_node *node, *next; GEM_BUG_ON(intel_engine_is_virtual(engine)); - GEM_BUG_ON(rq->timeline != engine->kernel_context->ring->timeline); + GEM_BUG_ON(rq->timeline != engine->kernel_context->timeline); /* * Attach the list of proto-fences to the in-flight request such diff --git a/drivers/gpu/drm/i915/i915_request.c b/drivers/gpu/drm/i915/i915_request.c index 9b2b18f0196b..43175bada09e 100644 --- a/drivers/gpu/drm/i915/i915_request.c +++ b/drivers/gpu/drm/i915/i915_request.c @@ -306,12 +306,12 @@ static bool i915_request_retire(struct i915_request *rq) local_irq_enable(); - intel_context_exit(rq->hw_context); - intel_context_unpin(rq->hw_context); - i915_request_remove_from_client(rq); list_del(&rq->link); + intel_context_exit(rq->hw_context); + intel_context_unpin(rq->hw_context); + free_capture_list(rq); i915_sched_node_fini(&rq->sched); i915_request_put(rq); @@ -608,7 +608,7 @@ out: struct i915_request * __i915_request_create(struct intel_context *ce, gfp_t gfp) { - struct intel_timeline *tl = ce->ring->timeline; + struct intel_timeline *tl = ce->timeline; struct i915_request *rq; u32 seqno; int ret; @@ -760,7 +760,7 @@ i915_request_create(struct intel_context *ce) goto err_unlock; /* Check that we do not interrupt ourselves with a new request */ - rq->cookie = lockdep_pin_lock(&ce->ring->timeline->mutex); + rq->cookie = lockdep_pin_lock(&ce->timeline->mutex); return rq; diff --git a/drivers/gpu/drm/i915/selftests/i915_gem_evict.c b/drivers/gpu/drm/i915/selftests/i915_gem_evict.c index b6449d0a8c17..cb30c669b1b7 100644 --- a/drivers/gpu/drm/i915/selftests/i915_gem_evict.c +++ b/drivers/gpu/drm/i915/selftests/i915_gem_evict.c @@ -48,26 +48,29 @@ static int populate_ggtt(struct drm_i915_private *i915, { unsigned long unbound, bound, count; struct drm_i915_gem_object *obj; - u64 size; count = 0; - for (size = 0; - size + I915_GTT_PAGE_SIZE <= i915->ggtt.vm.total; - size += I915_GTT_PAGE_SIZE) { + do { struct i915_vma *vma; obj = i915_gem_object_create_internal(i915, I915_GTT_PAGE_SIZE); if (IS_ERR(obj)) return PTR_ERR(obj); - quirk_add(obj, objects); - vma = i915_gem_object_ggtt_pin(obj, NULL, 0, 0, 0); - if (IS_ERR(vma)) + if (IS_ERR(vma)) { + i915_gem_object_put(obj); + if (vma == ERR_PTR(-ENOSPC)) + break; + return PTR_ERR(vma); + } + quirk_add(obj, objects); count++; - } + } while (1); + pr_debug("Filled GGTT with %lu pages [%llu total]\n", + count, i915->ggtt.vm.total / PAGE_SIZE); bound = 0; unbound = 0; -- cgit v1.2.3 From 478ffad6d6902298802b5b8044e542739fcb0ed1 Mon Sep 17 00:00:00 2001 From: Daniele Ceraolo Spurio Date: Mon, 12 Aug 2019 16:31:51 -0700 Subject: drm/i915: drop engine_pin/unpin_breadcrumbs_irq The last user has been removed, so drop the functions. Signed-off-by: Daniele Ceraolo Spurio Cc: Chris Wilson Reviewed-by: Chris Wilson Signed-off-by: Chris Wilson Link: https://patchwork.freedesktop.org/patch/msgid/20190812233152.2172-2-daniele.ceraolospurio@intel.com --- drivers/gpu/drm/i915/gt/intel_breadcrumbs.c | 22 ---------------------- drivers/gpu/drm/i915/gt/intel_engine.h | 3 --- 2 files changed, 25 deletions(-) (limited to 'drivers/gpu/drm/i915/gt/intel_engine.h') diff --git a/drivers/gpu/drm/i915/gt/intel_breadcrumbs.c b/drivers/gpu/drm/i915/gt/intel_breadcrumbs.c index ceba1da61967..15bbdd8c7552 100644 --- a/drivers/gpu/drm/i915/gt/intel_breadcrumbs.c +++ b/drivers/gpu/drm/i915/gt/intel_breadcrumbs.c @@ -212,28 +212,6 @@ static void signal_irq_work(struct irq_work *work) intel_engine_breadcrumbs_irq(engine); } -void intel_engine_pin_breadcrumbs_irq(struct intel_engine_cs *engine) -{ - struct intel_breadcrumbs *b = &engine->breadcrumbs; - - spin_lock_irq(&b->irq_lock); - if (!b->irq_enabled++) - irq_enable(engine); - GEM_BUG_ON(!b->irq_enabled); /* no overflow! */ - spin_unlock_irq(&b->irq_lock); -} - -void intel_engine_unpin_breadcrumbs_irq(struct intel_engine_cs *engine) -{ - struct intel_breadcrumbs *b = &engine->breadcrumbs; - - spin_lock_irq(&b->irq_lock); - GEM_BUG_ON(!b->irq_enabled); /* no underflow! */ - if (!--b->irq_enabled) - irq_disable(engine); - spin_unlock_irq(&b->irq_lock); -} - static void __intel_breadcrumbs_arm_irq(struct intel_breadcrumbs *b) { struct intel_engine_cs *engine = diff --git a/drivers/gpu/drm/i915/gt/intel_engine.h b/drivers/gpu/drm/i915/gt/intel_engine.h index e1228b0e577f..bc694adcd9ea 100644 --- a/drivers/gpu/drm/i915/gt/intel_engine.h +++ b/drivers/gpu/drm/i915/gt/intel_engine.h @@ -338,9 +338,6 @@ void intel_engine_init_execlists(struct intel_engine_cs *engine); void intel_engine_init_breadcrumbs(struct intel_engine_cs *engine); void intel_engine_fini_breadcrumbs(struct intel_engine_cs *engine); -void intel_engine_pin_breadcrumbs_irq(struct intel_engine_cs *engine); -void intel_engine_unpin_breadcrumbs_irq(struct intel_engine_cs *engine); - void intel_engine_signal_breadcrumbs(struct intel_engine_cs *engine); void intel_engine_disarm_breadcrumbs(struct intel_engine_cs *engine); -- cgit v1.2.3 From 5f15c1e6e181636f4b2fe7a67d582356f57cfd2c Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Mon, 12 Aug 2019 21:36:26 +0100 Subject: drm/i915/guc: Use a local cancel_port_requests Since execlists and the guc have diverged in their port tracking, we cannot simply reuse the execlists cancellation code as it leads to unbalanced reference counting. Use a local, simpler routine for the guc. Signed-off-by: Chris Wilson Cc: Daniele Ceraolo Spurio Reviewed-by: Daniele Ceraolo Spurio Link: https://patchwork.freedesktop.org/patch/msgid/20190812203626.3948-1-chris@chris-wilson.co.uk --- drivers/gpu/drm/i915/gt/intel_engine.h | 3 -- drivers/gpu/drm/i915/gt/intel_lrc.c | 6 ++-- drivers/gpu/drm/i915/gt/uc/intel_guc_submission.c | 37 +++++++++++++++++------ 3 files changed, 30 insertions(+), 16 deletions(-) (limited to 'drivers/gpu/drm/i915/gt/intel_engine.h') diff --git a/drivers/gpu/drm/i915/gt/intel_engine.h b/drivers/gpu/drm/i915/gt/intel_engine.h index bc694adcd9ea..d3c6993f4f46 100644 --- a/drivers/gpu/drm/i915/gt/intel_engine.h +++ b/drivers/gpu/drm/i915/gt/intel_engine.h @@ -136,9 +136,6 @@ execlists_active(const struct intel_engine_execlists *execlists) return READ_ONCE(*execlists->active); } -void -execlists_cancel_port_requests(struct intel_engine_execlists * const execlists); - struct i915_request * execlists_unwind_incomplete_requests(struct intel_engine_execlists *execlists); diff --git a/drivers/gpu/drm/i915/gt/intel_lrc.c b/drivers/gpu/drm/i915/gt/intel_lrc.c index b97047d58d3d..5c26c4ae139b 100644 --- a/drivers/gpu/drm/i915/gt/intel_lrc.c +++ b/drivers/gpu/drm/i915/gt/intel_lrc.c @@ -1297,8 +1297,8 @@ done: } } -void -execlists_cancel_port_requests(struct intel_engine_execlists * const execlists) +static void +cancel_port_requests(struct intel_engine_execlists * const execlists) { struct i915_request * const *port, *rq; @@ -2355,7 +2355,7 @@ out_replay: unwind: /* Push back any incomplete requests for replay after the reset. */ - execlists_cancel_port_requests(execlists); + cancel_port_requests(execlists); __unwind_incomplete_requests(engine); } 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 deb054eeb37c..54c3ae2c49ff 100644 --- a/drivers/gpu/drm/i915/gt/uc/intel_guc_submission.c +++ b/drivers/gpu/drm/i915/gt/uc/intel_guc_submission.c @@ -517,11 +517,14 @@ static struct i915_request *schedule_in(struct i915_request *rq, int idx) { trace_i915_request_in(rq, idx); - if (!rq->hw_context->inflight) - rq->hw_context->inflight = rq->engine; - intel_context_inflight_inc(rq->hw_context); - intel_gt_pm_get(rq->engine->gt); + /* + * Currently we are not tracking the rq->context being inflight + * (ce->inflight = rq->engine). It is only used by the execlists + * backend at the moment, a similar counting strategy would be + * required if we generalise the inflight tracking. + */ + intel_gt_pm_get(rq->engine->gt); return i915_request_get(rq); } @@ -529,10 +532,6 @@ static void schedule_out(struct i915_request *rq) { trace_i915_request_out(rq); - intel_context_inflight_dec(rq->hw_context); - if (!intel_context_inflight_count(rq->hw_context)) - rq->hw_context->inflight = NULL; - intel_gt_pm_put(rq->engine->gt); i915_request_put(rq); } @@ -556,6 +555,11 @@ static void __guc_dequeue(struct intel_engine_cs *engine) last = NULL; } + /* + * We write directly into the execlists->inflight queue and don't use + * the execlists->pending queue, as we don't have a distinct switch + * event. + */ port = first; while ((rb = rb_first_cached(&execlists->queue))) { struct i915_priolist *p = to_priolist(rb); @@ -636,6 +640,19 @@ static void guc_reset_prepare(struct intel_engine_cs *engine) __tasklet_disable_sync_once(&execlists->tasklet); } +static void +cancel_port_requests(struct intel_engine_execlists * const execlists) +{ + struct i915_request * const *port, *rq; + + /* Note we are only using the inflight and not the pending queue */ + + for (port = execlists->active; (rq = *port); port++) + schedule_out(rq); + execlists->active = + memset(execlists->inflight, 0, sizeof(execlists->inflight)); +} + static void guc_reset(struct intel_engine_cs *engine, bool stalled) { struct intel_engine_execlists * const execlists = &engine->execlists; @@ -644,7 +661,7 @@ static void guc_reset(struct intel_engine_cs *engine, bool stalled) spin_lock_irqsave(&engine->active.lock, flags); - execlists_cancel_port_requests(execlists); + cancel_port_requests(execlists); /* Push back any incomplete requests for replay after the reset. */ rq = execlists_unwind_incomplete_requests(execlists); @@ -687,7 +704,7 @@ static void guc_cancel_requests(struct intel_engine_cs *engine) spin_lock_irqsave(&engine->active.lock, flags); /* Cancel the requests on the HW and clear the ELSP tracker. */ - execlists_cancel_port_requests(execlists); + cancel_port_requests(execlists); /* Mark all executing requests as skipped. */ list_for_each_entry(rq, &engine->active.requests, sched.link) { -- cgit v1.2.3