diff options
Diffstat (limited to 'drivers/gpu/drm/i915/gt')
46 files changed, 3138 insertions, 1029 deletions
diff --git a/drivers/gpu/drm/i915/gt/intel_breadcrumbs.c b/drivers/gpu/drm/i915/gt/intel_breadcrumbs.c index 09c68dda2098..55317081d48b 100644 --- a/drivers/gpu/drm/i915/gt/intel_breadcrumbs.c +++ b/drivers/gpu/drm/i915/gt/intel_breadcrumbs.c @@ -120,7 +120,6 @@ __dma_fence_signal__notify(struct dma_fence *fence, struct dma_fence_cb *cur, *tmp; lockdep_assert_held(fence->lock); - lockdep_assert_irqs_disabled(); list_for_each_entry_safe(cur, tmp, list, node) { INIT_LIST_HEAD(&cur->node); @@ -134,9 +133,10 @@ void intel_engine_breadcrumbs_irq(struct intel_engine_cs *engine) const ktime_t timestamp = ktime_get(); struct intel_context *ce, *cn; struct list_head *pos, *next; + unsigned long flags; LIST_HEAD(signal); - spin_lock(&b->irq_lock); + spin_lock_irqsave(&b->irq_lock, flags); if (b->irq_armed && list_empty(&b->signalers)) __intel_breadcrumbs_disarm_irq(b); @@ -182,30 +182,23 @@ void intel_engine_breadcrumbs_irq(struct intel_engine_cs *engine) } } - spin_unlock(&b->irq_lock); + spin_unlock_irqrestore(&b->irq_lock, flags); list_for_each_safe(pos, next, &signal) { struct i915_request *rq = list_entry(pos, typeof(*rq), signal_link); struct list_head cb_list; - spin_lock(&rq->lock); + spin_lock_irqsave(&rq->lock, flags); list_replace(&rq->fence.cb_list, &cb_list); __dma_fence_signal__timestamp(&rq->fence, timestamp); __dma_fence_signal__notify(&rq->fence, &cb_list); - spin_unlock(&rq->lock); + spin_unlock_irqrestore(&rq->lock, flags); i915_request_put(rq); } } -void intel_engine_signal_breadcrumbs(struct intel_engine_cs *engine) -{ - local_irq_disable(); - intel_engine_breadcrumbs_irq(engine); - local_irq_enable(); -} - static void signal_irq_work(struct irq_work *work) { struct intel_engine_cs *engine = @@ -275,7 +268,6 @@ void intel_engine_fini_breadcrumbs(struct intel_engine_cs *engine) bool i915_request_enable_breadcrumb(struct i915_request *rq) { lockdep_assert_held(&rq->lock); - lockdep_assert_irqs_disabled(); if (test_bit(I915_FENCE_FLAG_ACTIVE, &rq->fence.flags)) { struct intel_breadcrumbs *b = &rq->engine->breadcrumbs; @@ -325,7 +317,6 @@ void i915_request_cancel_breadcrumb(struct i915_request *rq) struct intel_breadcrumbs *b = &rq->engine->breadcrumbs; lockdep_assert_held(&rq->lock); - lockdep_assert_irqs_disabled(); /* * We must wait for b->irq_lock so that we know the interrupt handler diff --git a/drivers/gpu/drm/i915/gt/intel_context.c b/drivers/gpu/drm/i915/gt/intel_context.c index f55691d151ae..be34d97ac18f 100644 --- a/drivers/gpu/drm/i915/gt/intel_context.c +++ b/drivers/gpu/drm/i915/gt/intel_context.c @@ -134,10 +134,11 @@ static int __context_pin_state(struct i915_vma *vma) static void __context_unpin_state(struct i915_vma *vma) { - __i915_vma_unpin(vma); i915_vma_make_shrinkable(vma); + __i915_vma_unpin(vma); } +__i915_active_call static void __intel_context_retire(struct i915_active *active) { struct intel_context *ce = container_of(active, typeof(*ce), active); @@ -150,6 +151,7 @@ static void __intel_context_retire(struct i915_active *active) intel_timeline_unpin(ce->timeline); intel_ring_unpin(ce->ring); + intel_context_put(ce); } @@ -219,12 +221,20 @@ intel_context_init(struct intel_context *ce, struct i915_gem_context *ctx, struct intel_engine_cs *engine) { + struct i915_address_space *vm; + GEM_BUG_ON(!engine->cops); kref_init(&ce->ref); ce->gem_context = ctx; - ce->vm = i915_vm_get(ctx->vm ?: &engine->gt->ggtt->vm); + rcu_read_lock(); + vm = rcu_dereference(ctx->vm); + if (vm) + ce->vm = i915_vm_get(vm); + else + ce->vm = i915_vm_get(&engine->gt->ggtt->vm); + rcu_read_unlock(); if (ctx->timeline) ce->timeline = intel_timeline_get(ctx->timeline); @@ -238,7 +248,7 @@ intel_context_init(struct intel_context *ce, mutex_init(&ce->pin_mutex); - i915_active_init(ctx->i915, &ce->active, + i915_active_init(&ce->active, __intel_context_active, __intel_context_retire); } @@ -298,14 +308,14 @@ int intel_context_prepare_remote_request(struct intel_context *ce, /* Only suitable for use in remotely modifying this context */ GEM_BUG_ON(rq->hw_context == ce); - if (rq->timeline != tl) { /* beware timeline sharing */ + if (rcu_access_pointer(rq->timeline) != tl) { /* timeline sharing! */ err = mutex_lock_interruptible_nested(&tl->mutex, SINGLE_DEPTH_NESTING); if (err) return err; /* Queue this switch after current activity by this context. */ - err = i915_active_request_set(&tl->last_request, rq); + err = i915_active_fence_set(&tl->last_request, rq); mutex_unlock(&tl->mutex); if (err) return err; @@ -319,7 +329,7 @@ int intel_context_prepare_remote_request(struct intel_context *ce, * words transfer the pinned ce object to tracked active request. */ GEM_BUG_ON(i915_active_is_idle(&ce->active)); - return i915_active_ref(&ce->active, rq->timeline, rq); + return i915_active_add_request(&ce->active, rq); } struct i915_request *intel_context_create_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 bf9cedfccbf0..6959b05ae5f8 100644 --- a/drivers/gpu/drm/i915/gt/intel_context_types.h +++ b/drivers/gpu/drm/i915/gt/intel_context_types.h @@ -58,6 +58,7 @@ struct intel_context { u32 *lrc_reg_state; u64 lrc_desc; + u32 tag; /* cookie passed to HW to track this context on submission */ unsigned int active_count; /* protected by timeline->mutex */ diff --git a/drivers/gpu/drm/i915/gt/intel_engine.h b/drivers/gpu/drm/i915/gt/intel_engine.h index d3c6993f4f46..c9e8c8ccbd47 100644 --- a/drivers/gpu/drm/i915/gt/intel_engine.h +++ b/drivers/gpu/drm/i915/gt/intel_engine.h @@ -335,7 +335,6 @@ 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_signal_breadcrumbs(struct intel_engine_cs *engine); void intel_engine_disarm_breadcrumbs(struct intel_engine_cs *engine); static inline void diff --git a/drivers/gpu/drm/i915/gt/intel_engine_cs.c b/drivers/gpu/drm/i915/gt/intel_engine_cs.c index 82630db0394b..80fd072ac719 100644 --- a/drivers/gpu/drm/i915/gt/intel_engine_cs.c +++ b/drivers/gpu/drm/i915/gt/intel_engine_cs.c @@ -680,6 +680,8 @@ static int measure_breadcrumb_dw(struct intel_engine_cs *engine) engine->status_page.vma)) goto out_frame; + mutex_lock(&frame->timeline.mutex); + frame->ring.vaddr = frame->cs; frame->ring.size = sizeof(frame->cs); frame->ring.effective_size = frame->ring.size; @@ -688,18 +690,22 @@ static int measure_breadcrumb_dw(struct intel_engine_cs *engine) frame->rq.i915 = engine->i915; frame->rq.engine = engine; frame->rq.ring = &frame->ring; - frame->rq.timeline = &frame->timeline; + rcu_assign_pointer(frame->rq.timeline, &frame->timeline); dw = intel_timeline_pin(&frame->timeline); if (dw < 0) goto out_timeline; + spin_lock_irq(&engine->active.lock); dw = engine->emit_fini_breadcrumb(&frame->rq, frame->cs) - frame->cs; + spin_unlock_irq(&engine->active.lock); + GEM_BUG_ON(dw & 1); /* RING_TAIL must be qword aligned */ intel_timeline_unpin(&frame->timeline); out_timeline: + mutex_unlock(&frame->timeline.mutex); intel_timeline_fini(&frame->timeline); out_frame: kfree(frame); @@ -814,8 +820,10 @@ void intel_engine_cleanup_common(struct intel_engine_cs *engine) if (engine->default_state) i915_gem_object_put(engine->default_state); - intel_context_unpin(engine->kernel_context); - intel_context_put(engine->kernel_context); + if (engine->kernel_context) { + intel_context_unpin(engine->kernel_context); + intel_context_put(engine->kernel_context); + } GEM_BUG_ON(!llist_empty(&engine->barrier_tasks)); intel_wa_list_free(&engine->ctx_wa_list); @@ -948,6 +956,7 @@ void intel_engine_get_instdone(struct intel_engine_cs *engine, struct intel_instdone *instdone) { struct drm_i915_private *i915 = engine->i915; + const struct sseu_dev_info *sseu = &RUNTIME_INFO(i915)->sseu; struct intel_uncore *uncore = engine->uncore; u32 mmio_base = engine->mmio_base; int slice; @@ -965,7 +974,7 @@ void intel_engine_get_instdone(struct intel_engine_cs *engine, instdone->slice_common = intel_uncore_read(uncore, GEN7_SC_INSTDONE); - for_each_instdone_slice_subslice(i915, slice, subslice) { + for_each_instdone_slice_subslice(i915, sseu, slice, subslice) { instdone->sampler[slice][subslice] = read_subslice_reg(engine, slice, subslice, GEN7_SAMPLER_INSTDONE); @@ -1118,6 +1127,8 @@ bool intel_engine_can_store_dword(struct intel_engine_cs *engine) case 3: /* maybe only uses physical not virtual addresses */ return !(IS_I915G(engine->i915) || IS_I915GM(engine->i915)); + case 4: + return !IS_I965G(engine->i915); /* who knows! */ case 6: return engine->class != VIDEO_DECODE_CLASS; /* b0rked */ default: @@ -1193,6 +1204,27 @@ static void hexdump(struct drm_printer *m, const void *buf, size_t len) } } +static struct intel_timeline *get_timeline(struct i915_request *rq) +{ + struct intel_timeline *tl; + + /* + * Even though we are holding the engine->active.lock here, there + * is no control over the submission queue per-se and we are + * inspecting the active state at a random point in time, with an + * unknown queue. Play safe and make sure the timeline remains valid. + * (Only being used for pretty printing, one extra kref shouldn't + * cause a camel stampede!) + */ + rcu_read_lock(); + tl = rcu_dereference(rq->timeline); + if (!kref_get_unless_zero(&tl->kref)) + tl = NULL; + rcu_read_unlock(); + + return tl; +} + static void intel_engine_print_registers(struct intel_engine_cs *engine, struct drm_printer *m) { @@ -1287,27 +1319,37 @@ static void intel_engine_print_registers(struct intel_engine_cs *engine, int len; len = snprintf(hdr, sizeof(hdr), - "\t\tActive[%d: ", + "\t\tActive[%d]: ", (int)(port - execlists->active)); - if (!i915_request_signaled(rq)) + if (!i915_request_signaled(rq)) { + struct intel_timeline *tl = get_timeline(rq); + len += snprintf(hdr + len, sizeof(hdr) - len, "ring:{start:%08x, hwsp:%08x, seqno:%08x}, ", i915_ggtt_offset(rq->ring->vma), - rq->timeline->hwsp_offset, + tl ? tl->hwsp_offset : 0, hwsp_seqno(rq)); + + if (tl) + intel_timeline_put(tl); + } snprintf(hdr + len, sizeof(hdr) - len, "rq: "); print_request(m, rq, hdr); } for (port = execlists->pending; (rq = *port); port++) { + struct intel_timeline *tl = get_timeline(rq); char hdr[80]; snprintf(hdr, sizeof(hdr), "\t\tPending[%d] ring:{start:%08x, hwsp:%08x, seqno:%08x}, rq: ", (int)(port - execlists->pending), i915_ggtt_offset(rq->ring->vma), - rq->timeline->hwsp_offset, + tl ? tl->hwsp_offset : 0, hwsp_seqno(rq)); print_request(m, rq, hdr); + + if (tl) + intel_timeline_put(tl); } spin_unlock_irqrestore(&engine->active.lock, flags); } else if (INTEL_GEN(dev_priv) > 6) { @@ -1385,6 +1427,8 @@ void intel_engine_dump(struct intel_engine_cs *engine, spin_lock_irqsave(&engine->active.lock, flags); rq = intel_engine_find_active_request(engine); if (rq) { + struct intel_timeline *tl = get_timeline(rq); + print_request(m, rq, "\t\tactive "); drm_printf(m, "\t\tring->start: 0x%08x\n", @@ -1397,10 +1441,19 @@ void intel_engine_dump(struct intel_engine_cs *engine, rq->ring->emit); drm_printf(m, "\t\tring->space: 0x%08x\n", rq->ring->space); - drm_printf(m, "\t\tring->hwsp: 0x%08x\n", - rq->timeline->hwsp_offset); + + if (tl) { + drm_printf(m, "\t\tring->hwsp: 0x%08x\n", + tl->hwsp_offset); + intel_timeline_put(tl); + } print_request_ring(m, rq); + + if (rq->hw_context->lrc_reg_state) { + drm_printf(m, "Logical Ring Context:\n"); + hexdump(m, rq->hw_context->lrc_reg_state, PAGE_SIZE); + } } spin_unlock_irqrestore(&engine->active.lock, flags); diff --git a/drivers/gpu/drm/i915/gt/intel_engine_pm.c b/drivers/gpu/drm/i915/gt/intel_engine_pm.c index 65b5ca74b394..8e5e513eddc9 100644 --- a/drivers/gpu/drm/i915/gt/intel_engine_pm.c +++ b/drivers/gpu/drm/i915/gt/intel_engine_pm.c @@ -11,6 +11,7 @@ #include "intel_engine_pool.h" #include "intel_gt.h" #include "intel_gt_pm.h" +#include "intel_rc6.h" static int __engine_unpark(struct intel_wakeref *wf) { @@ -103,7 +104,7 @@ static bool switch_to_kernel_context(struct intel_engine_cs *engine) /* Context switch failed, hope for the best! Maybe reset? */ goto out_unlock; - intel_timeline_enter(rq->timeline); + intel_timeline_enter(i915_request_timeline(rq)); /* Check again on the next retirement. */ engine->wakeref_serial = engine->serial + 1; @@ -123,6 +124,19 @@ out_unlock: return result; } +static void call_idle_barriers(struct intel_engine_cs *engine) +{ + struct llist_node *node, *next; + + llist_for_each_safe(node, next, llist_del_all(&engine->barrier_tasks)) { + struct dma_fence_cb *cb = + container_of((struct list_head *)node, + typeof(*cb), node); + + cb->func(NULL, cb); + } +} + static int __engine_park(struct intel_wakeref *wf) { struct intel_engine_cs *engine = @@ -142,6 +156,8 @@ static int __engine_park(struct intel_wakeref *wf) GEM_TRACE("%s\n", engine->name); + call_idle_barriers(engine); /* cleanup after wedging */ + intel_engine_disarm_breadcrumbs(engine); intel_engine_pool_park(&engine->pool); diff --git a/drivers/gpu/drm/i915/gt/intel_engine_pool.c b/drivers/gpu/drm/i915/gt/intel_engine_pool.c index 4cd54c569911..3cdbd5f8b5be 100644 --- a/drivers/gpu/drm/i915/gt/intel_engine_pool.c +++ b/drivers/gpu/drm/i915/gt/intel_engine_pool.c @@ -61,6 +61,7 @@ static int pool_active(struct i915_active *ref) return 0; } +__i915_active_call static void pool_retire(struct i915_active *ref) { struct intel_engine_pool_node *node = @@ -94,7 +95,7 @@ node_create(struct intel_engine_pool *pool, size_t sz) return ERR_PTR(-ENOMEM); node->pool = pool; - i915_active_init(engine->i915, &node->active, pool_active, pool_retire); + i915_active_init(&node->active, pool_active, pool_retire); obj = i915_gem_object_create_internal(engine->i915, sz); if (IS_ERR(obj)) { @@ -107,9 +108,19 @@ node_create(struct intel_engine_pool *pool, size_t sz) return node; } +static struct intel_engine_pool *lookup_pool(struct intel_engine_cs *engine) +{ + if (intel_engine_is_virtual(engine)) + engine = intel_virtual_engine_get_sibling(engine, 0); + + GEM_BUG_ON(!engine); + return &engine->pool; +} + struct intel_engine_pool_node * -intel_engine_pool_get(struct intel_engine_pool *pool, size_t size) +intel_engine_get_pool(struct intel_engine_cs *engine, size_t size) { + struct intel_engine_pool *pool = lookup_pool(engine); struct intel_engine_pool_node *node; struct list_head *list; unsigned long flags; diff --git a/drivers/gpu/drm/i915/gt/intel_engine_pool.h b/drivers/gpu/drm/i915/gt/intel_engine_pool.h index 8d069efd9457..1bd89cadc3b7 100644 --- a/drivers/gpu/drm/i915/gt/intel_engine_pool.h +++ b/drivers/gpu/drm/i915/gt/intel_engine_pool.h @@ -12,13 +12,13 @@ #include "i915_request.h" struct intel_engine_pool_node * -intel_engine_pool_get(struct intel_engine_pool *pool, size_t size); +intel_engine_get_pool(struct intel_engine_cs *engine, 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->timeline, rq); + return i915_active_add_request(&node->active, rq); } static inline void diff --git a/drivers/gpu/drm/i915/gt/intel_engine_types.h b/drivers/gpu/drm/i915/gt/intel_engine_types.h index a82cea95c2f2..6199064f332b 100644 --- a/drivers/gpu/drm/i915/gt/intel_engine_types.h +++ b/drivers/gpu/drm/i915/gt/intel_engine_types.h @@ -303,10 +303,12 @@ struct intel_engine_cs { u8 uabi_class; u8 uabi_instance; + u32 uabi_capabilities; u32 context_size; u32 mmio_base; - u32 uabi_capabilities; + unsigned int context_tag; +#define NUM_CONTEXT_TAG roundup_pow_of_two(2 * EXECLIST_MAX_PORTS) struct rb_node uabi_node; @@ -481,6 +483,7 @@ struct intel_engine_cs { #define I915_ENGINE_HAS_SEMAPHORES BIT(3) #define I915_ENGINE_NEEDS_BREADCRUMB_TASKLET BIT(4) #define I915_ENGINE_IS_VIRTUAL BIT(5) +#define I915_ENGINE_HAS_RELATIVE_MMIO BIT(6) unsigned int flags; /* @@ -576,20 +579,24 @@ intel_engine_is_virtual(const struct intel_engine_cs *engine) return engine->flags & I915_ENGINE_IS_VIRTUAL; } -#define instdone_slice_mask(dev_priv__) \ - (IS_GEN(dev_priv__, 7) ? \ - 1 : RUNTIME_INFO(dev_priv__)->sseu.slice_mask) +static inline bool +intel_engine_has_relative_mmio(const struct intel_engine_cs * const engine) +{ + return engine->flags & I915_ENGINE_HAS_RELATIVE_MMIO; +} -#define instdone_subslice_mask(dev_priv__) \ - (IS_GEN(dev_priv__, 7) ? \ - 1 : RUNTIME_INFO(dev_priv__)->sseu.subslice_mask[0]) +#define instdone_has_slice(dev_priv___, sseu___, slice___) \ + ((IS_GEN(dev_priv___, 7) ? 1 : ((sseu___)->slice_mask)) & BIT(slice___)) -#define for_each_instdone_slice_subslice(dev_priv__, slice__, subslice__) \ - for ((slice__) = 0, (subslice__) = 0; \ - (slice__) < I915_MAX_SLICES; \ - (subslice__) = ((subslice__) + 1) < I915_MAX_SUBSLICES ? (subslice__) + 1 : 0, \ - (slice__) += ((subslice__) == 0)) \ - for_each_if((BIT(slice__) & instdone_slice_mask(dev_priv__)) && \ - (BIT(subslice__) & instdone_subslice_mask(dev_priv__))) +#define instdone_has_subslice(dev_priv__, sseu__, slice__, subslice__) \ + (IS_GEN(dev_priv__, 7) ? (1 & BIT(subslice__)) : \ + intel_sseu_has_subslice(sseu__, 0, subslice__)) +#define for_each_instdone_slice_subslice(dev_priv_, sseu_, slice_, subslice_) \ + for ((slice_) = 0, (subslice_) = 0; (slice_) < I915_MAX_SLICES; \ + (subslice_) = ((subslice_) + 1) % I915_MAX_SUBSLICES, \ + (slice_) += ((subslice_) == 0)) \ + for_each_if((instdone_has_slice(dev_priv_, sseu_, slice_)) && \ + (instdone_has_subslice(dev_priv_, sseu_, slice_, \ + subslice_))) #endif /* __INTEL_ENGINE_TYPES_H__ */ diff --git a/drivers/gpu/drm/i915/gt/intel_gpu_commands.h b/drivers/gpu/drm/i915/gt/intel_gpu_commands.h index 86e00a2db8a4..b0227ab2fe1b 100644 --- a/drivers/gpu/drm/i915/gt/intel_gpu_commands.h +++ b/drivers/gpu/drm/i915/gt/intel_gpu_commands.h @@ -112,6 +112,7 @@ #define MI_SEMAPHORE_SIGNAL MI_INSTR(0x1b, 0) /* GEN8+ */ #define MI_SEMAPHORE_TARGET(engine) ((engine)<<15) #define MI_SEMAPHORE_WAIT MI_INSTR(0x1c, 2) /* GEN8+ */ +#define MI_SEMAPHORE_WAIT_TOKEN MI_INSTR(0x1c, 3) /* GEN12+ */ #define MI_SEMAPHORE_POLL (1 << 15) #define MI_SEMAPHORE_SAD_GT_SDD (0 << 12) #define MI_SEMAPHORE_SAD_GTE_SDD (1 << 12) @@ -119,6 +120,8 @@ #define MI_SEMAPHORE_SAD_LTE_SDD (3 << 12) #define MI_SEMAPHORE_SAD_EQ_SDD (4 << 12) #define MI_SEMAPHORE_SAD_NEQ_SDD (5 << 12) +#define MI_SEMAPHORE_TOKEN_MASK REG_GENMASK(9, 5) +#define MI_SEMAPHORE_TOKEN_SHIFT 5 #define MI_STORE_DWORD_IMM MI_INSTR(0x20, 1) #define MI_STORE_DWORD_IMM_GEN4 MI_INSTR(0x20, 2) #define MI_MEM_VIRTUAL (1 << 22) /* 945,g33,965 */ @@ -132,6 +135,8 @@ * address/value pairs. Don't overdue it, though, x <= 2^4 must hold! */ #define MI_LOAD_REGISTER_IMM(x) MI_INSTR(0x22, 2*(x)-1) +/* Gen11+. addr = base + (ctx_restore ? offset & GENMASK(12,2) : offset) */ +#define MI_LRI_CS_MMIO (1<<19) #define MI_LRI_FORCE_POSTED (1<<12) #define MI_STORE_REGISTER_MEM MI_INSTR(0x24, 1) #define MI_STORE_REGISTER_MEM_GEN8 MI_INSTR(0x24, 2) @@ -147,6 +152,7 @@ #define MI_FLUSH_DW_USE_PPGTT (0<<2) #define MI_LOAD_REGISTER_MEM MI_INSTR(0x29, 1) #define MI_LOAD_REGISTER_MEM_GEN8 MI_INSTR(0x29, 2) +#define MI_LOAD_REGISTER_REG MI_INSTR(0x2A, 1) #define MI_BATCH_BUFFER MI_INSTR(0x30, 1) #define MI_BATCH_NON_SECURE (1) /* for snb/ivb/vlv this also means "batch in ppgtt" when ppgtt is enabled. */ @@ -235,6 +241,29 @@ #define PIPE_CONTROL_DEPTH_CACHE_FLUSH (1<<0) #define PIPE_CONTROL_GLOBAL_GTT (1<<2) /* in addr dword */ +#define MI_MATH(x) MI_INSTR(0x1a, (x) - 1) +#define MI_MATH_INSTR(opcode, op1, op2) ((opcode) << 20 | (op1) << 10 | (op2)) +/* Opcodes for MI_MATH_INSTR */ +#define MI_MATH_NOOP MI_MATH_INSTR(0x000, 0x0, 0x0) +#define MI_MATH_LOAD(op1, op2) MI_MATH_INSTR(0x080, op1, op2) +#define MI_MATH_LOADINV(op1, op2) MI_MATH_INSTR(0x480, op1, op2) +#define MI_MATH_LOAD0(op1) MI_MATH_INSTR(0x081, op1) +#define MI_MATH_LOAD1(op1) MI_MATH_INSTR(0x481, op1) +#define MI_MATH_ADD MI_MATH_INSTR(0x100, 0x0, 0x0) +#define MI_MATH_SUB MI_MATH_INSTR(0x101, 0x0, 0x0) +#define MI_MATH_AND MI_MATH_INSTR(0x102, 0x0, 0x0) +#define MI_MATH_OR MI_MATH_INSTR(0x103, 0x0, 0x0) +#define MI_MATH_XOR MI_MATH_INSTR(0x104, 0x0, 0x0) +#define MI_MATH_STORE(op1, op2) MI_MATH_INSTR(0x180, op1, op2) +#define MI_MATH_STOREINV(op1, op2) MI_MATH_INSTR(0x580, op1, op2) +/* Registers used as operands in MI_MATH_INSTR */ +#define MI_MATH_REG(x) (x) +#define MI_MATH_REG_SRCA 0x20 +#define MI_MATH_REG_SRCB 0x21 +#define MI_MATH_REG_ACCU 0x31 +#define MI_MATH_REG_ZF 0x32 +#define MI_MATH_REG_CF 0x33 + /* * Commands used only by the command parser */ @@ -251,7 +280,6 @@ #define MI_CLFLUSH MI_INSTR(0x27, 0) #define MI_REPORT_PERF_COUNT MI_INSTR(0x28, 0) #define MI_REPORT_PERF_COUNT_GGTT (1<<0) -#define MI_LOAD_REGISTER_REG MI_INSTR(0x2A, 0) #define MI_RS_STORE_DATA_IMM MI_INSTR(0x2B, 0) #define MI_LOAD_URB_MEM MI_INSTR(0x2C, 0) #define MI_STORE_URB_MEM MI_INSTR(0x2D, 0) diff --git a/drivers/gpu/drm/i915/gt/intel_gt.c b/drivers/gpu/drm/i915/gt/intel_gt.c index d48ec9a76ed1..8f44cf8c79b2 100644 --- a/drivers/gpu/drm/i915/gt/intel_gt.c +++ b/drivers/gpu/drm/i915/gt/intel_gt.c @@ -6,7 +6,11 @@ #include "i915_drv.h" #include "intel_gt.h" #include "intel_gt_pm.h" +#include "intel_gt_requests.h" +#include "intel_mocs.h" +#include "intel_rc6.h" #include "intel_uncore.h" +#include "intel_pm.h" void intel_gt_init_early(struct intel_gt *gt, struct drm_i915_private *i915) { @@ -20,13 +24,106 @@ void intel_gt_init_early(struct intel_gt *gt, struct drm_i915_private *i915) intel_gt_init_hangcheck(gt); intel_gt_init_reset(gt); + intel_gt_init_requests(gt); intel_gt_pm_init_early(gt); intel_uc_init_early(>->uc); } -void intel_gt_init_hw(struct drm_i915_private *i915) +void intel_gt_init_hw_early(struct drm_i915_private *i915) { i915->gt.ggtt = &i915->ggtt; + + /* BIOS often leaves RC6 enabled, but disable it for hw init */ + intel_gt_pm_disable(&i915->gt); +} + +static void init_unused_ring(struct intel_gt *gt, u32 base) +{ + struct intel_uncore *uncore = gt->uncore; + + intel_uncore_write(uncore, RING_CTL(base), 0); + intel_uncore_write(uncore, RING_HEAD(base), 0); + intel_uncore_write(uncore, RING_TAIL(base), 0); + intel_uncore_write(uncore, RING_START(base), 0); +} + +static void init_unused_rings(struct intel_gt *gt) +{ + struct drm_i915_private *i915 = gt->i915; + + if (IS_I830(i915)) { + init_unused_ring(gt, PRB1_BASE); + init_unused_ring(gt, SRB0_BASE); + init_unused_ring(gt, SRB1_BASE); + init_unused_ring(gt, SRB2_BASE); + init_unused_ring(gt, SRB3_BASE); + } else if (IS_GEN(i915, 2)) { + init_unused_ring(gt, SRB0_BASE); + init_unused_ring(gt, SRB1_BASE); + } else if (IS_GEN(i915, 3)) { + init_unused_ring(gt, PRB1_BASE); + init_unused_ring(gt, PRB2_BASE); + } +} + +int intel_gt_init_hw(struct intel_gt *gt) +{ + struct drm_i915_private *i915 = gt->i915; + struct intel_uncore *uncore = gt->uncore; + int ret; + + BUG_ON(!i915->kernel_context); + ret = intel_gt_terminally_wedged(gt); + if (ret) + return ret; + + gt->last_init_time = ktime_get(); + + /* Double layer security blanket, see i915_gem_init() */ + intel_uncore_forcewake_get(uncore, FORCEWAKE_ALL); + + if (HAS_EDRAM(i915) && INTEL_GEN(i915) < 9) + intel_uncore_rmw(uncore, HSW_IDICR, 0, IDIHASHMSK(0xf)); + + if (IS_HASWELL(i915)) + intel_uncore_write(uncore, + MI_PREDICATE_RESULT_2, + IS_HSW_GT3(i915) ? + LOWER_SLICE_ENABLED : LOWER_SLICE_DISABLED); + + /* Apply the GT workarounds... */ + intel_gt_apply_workarounds(gt); + /* ...and determine whether they are sticking. */ + intel_gt_verify_workarounds(gt, "init"); + + intel_gt_init_swizzling(gt); + + /* + * At least 830 can leave some of the unused rings + * "active" (ie. head != tail) after resume which + * will prevent c3 entry. Makes sure all unused rings + * are totally idle. + */ + init_unused_rings(gt); + + ret = i915_ppgtt_init_hw(gt); + if (ret) { + DRM_ERROR("Enabling PPGTT failed (%d)\n", ret); + goto out; + } + + /* We can't enable contexts until all firmware is loaded */ + ret = intel_uc_init_hw(>->uc); + if (ret) { + i915_probe_error(i915, "Enabling uc failed (%d)\n", ret); + goto out; + } + + intel_mocs_init(gt); + +out: + intel_uncore_forcewake_put(uncore, FORCEWAKE_ALL); + return ret; } static void rmw_set(struct intel_uncore *uncore, i915_reg_t reg, u32 set) @@ -207,11 +304,12 @@ void intel_gt_flush_ggtt_writes(struct intel_gt *gt) with_intel_runtime_pm(&i915->runtime_pm, wakeref) { struct intel_uncore *uncore = gt->uncore; + unsigned long flags; - spin_lock_irq(&uncore->lock); + spin_lock_irqsave(&uncore->lock, flags); intel_uncore_posting_read_fw(uncore, RING_HEAD(RENDER_RING_BASE)); - spin_unlock_irq(&uncore->lock); + spin_unlock_irqrestore(&uncore->lock, flags); } } @@ -222,7 +320,13 @@ void intel_gt_chipset_flush(struct intel_gt *gt) intel_gtt_chipset_flush(); } -int intel_gt_init_scratch(struct intel_gt *gt, unsigned int size) +void intel_gt_driver_register(struct intel_gt *gt) +{ + if (IS_GEN(gt->i915, 5)) + intel_gpu_ips_init(gt->i915); +} + +static int intel_gt_init_scratch(struct intel_gt *gt, unsigned int size) { struct drm_i915_private *i915 = gt->i915; struct drm_i915_gem_object *obj; @@ -230,7 +334,7 @@ int intel_gt_init_scratch(struct intel_gt *gt, unsigned int size) int ret; obj = i915_gem_object_create_stolen(i915, size); - if (!obj) + if (IS_ERR(obj)) obj = i915_gem_object_create_internal(i915, size); if (IS_ERR(obj)) { DRM_ERROR("Failed to allocate scratch page\n"); @@ -256,11 +360,44 @@ err_unref: return ret; } -void intel_gt_fini_scratch(struct intel_gt *gt) +static void intel_gt_fini_scratch(struct intel_gt *gt) { i915_vma_unpin_and_release(>->scratch, 0); } +int intel_gt_init(struct intel_gt *gt) +{ + int err; + + err = intel_gt_init_scratch(gt, IS_GEN(gt->i915, 2) ? SZ_256K : SZ_4K); + if (err) + return err; + + intel_gt_pm_init(gt); + + return 0; +} + +void intel_gt_driver_remove(struct intel_gt *gt) +{ + GEM_BUG_ON(gt->awake); + intel_gt_pm_disable(gt); +} + +void intel_gt_driver_unregister(struct intel_gt *gt) +{ + intel_gpu_ips_teardown(); +} + +void intel_gt_driver_release(struct intel_gt *gt) +{ + /* Paranoia: make sure we have disabled everything before we exit. */ + intel_gt_pm_disable(gt); + intel_gt_pm_fini(gt); + + intel_gt_fini_scratch(gt); +} + void intel_gt_driver_late_release(struct intel_gt *gt) { intel_uc_driver_late_release(>->uc); diff --git a/drivers/gpu/drm/i915/gt/intel_gt.h b/drivers/gpu/drm/i915/gt/intel_gt.h index 4920cb351f10..e6ab0bff0efb 100644 --- a/drivers/gpu/drm/i915/gt/intel_gt.h +++ b/drivers/gpu/drm/i915/gt/intel_gt.h @@ -28,7 +28,14 @@ static inline struct intel_gt *huc_to_gt(struct intel_huc *huc) } 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_init_hw_early(struct drm_i915_private *i915); +int __must_check intel_gt_init_hw(struct intel_gt *gt); +int intel_gt_init(struct intel_gt *gt); +void intel_gt_driver_register(struct intel_gt *gt); + +void intel_gt_driver_unregister(struct intel_gt *gt); +void intel_gt_driver_remove(struct intel_gt *gt); +void intel_gt_driver_release(struct intel_gt *gt); void intel_gt_driver_late_release(struct intel_gt *gt); @@ -41,9 +48,6 @@ 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); - static inline u32 intel_gt_scratch_offset(const struct intel_gt *gt, enum intel_gt_scratch_field field) { diff --git a/drivers/gpu/drm/i915/gt/intel_gt_pm.c b/drivers/gpu/drm/i915/gt/intel_gt_pm.c index 1363e069ec83..b52e2ba3d092 100644 --- a/drivers/gpu/drm/i915/gt/intel_gt_pm.c +++ b/drivers/gpu/drm/i915/gt/intel_gt_pm.c @@ -5,16 +5,20 @@ */ #include "i915_drv.h" +#include "i915_globals.h" #include "i915_params.h" +#include "intel_context.h" #include "intel_engine_pm.h" #include "intel_gt.h" #include "intel_gt_pm.h" +#include "intel_gt_requests.h" #include "intel_pm.h" +#include "intel_rc6.h" #include "intel_wakeref.h" -static void pm_notify(struct drm_i915_private *i915, int state) +static void pm_notify(struct intel_gt *gt, int state) { - blocking_notifier_call_chain(&i915->gt.pm_notifications, state, i915); + blocking_notifier_call_chain(>->pm_notifications, state, gt->i915); } static int __gt_unpark(struct intel_wakeref *wf) @@ -24,6 +28,8 @@ static int __gt_unpark(struct intel_wakeref *wf) GEM_TRACE("\n"); + i915_globals_unpark(); + /* * It seems that the DMC likes to transition between the DC states a lot * when there are no connected displays (no active power domains) during @@ -47,21 +53,23 @@ static int __gt_unpark(struct intel_wakeref *wf) i915_pmu_gt_unparked(i915); intel_gt_queue_hangcheck(gt); + intel_gt_unpark_requests(gt); - pm_notify(i915, INTEL_GT_UNPARK); + pm_notify(gt, INTEL_GT_UNPARK); return 0; } static int __gt_park(struct intel_wakeref *wf) { - struct drm_i915_private *i915 = - container_of(wf, typeof(*i915), gt.wakeref); - intel_wakeref_t wakeref = fetch_and_zero(&i915->gt.awake); + struct intel_gt *gt = container_of(wf, typeof(*gt), wakeref); + intel_wakeref_t wakeref = fetch_and_zero(>->awake); + struct drm_i915_private *i915 = gt->i915; GEM_TRACE("\n"); - pm_notify(i915, INTEL_GT_PARK); + pm_notify(gt, INTEL_GT_PARK); + intel_gt_park_requests(gt); i915_pmu_gt_parked(i915); if (INTEL_GEN(i915) >= 6) @@ -73,6 +81,8 @@ static int __gt_park(struct intel_wakeref *wf) GEM_BUG_ON(!wakeref); intel_display_power_put(i915, POWER_DOMAIN_GT_IRQ, wakeref); + i915_globals_park(); + return 0; } @@ -89,6 +99,16 @@ void intel_gt_pm_init_early(struct intel_gt *gt) BLOCKING_INIT_NOTIFIER_HEAD(>->pm_notifications); } +void intel_gt_pm_init(struct intel_gt *gt) +{ + /* + * Enabling power-management should be "self-healing". If we cannot + * enable a feature, simply leave it disabled with a notice to the + * user. + */ + intel_rc6_init(>->rc6); +} + static bool reset_engines(struct intel_gt *gt) { if (INTEL_INFO(gt->i915)->gpu_reset_clobbers_display) @@ -123,6 +143,17 @@ void intel_gt_sanitize(struct intel_gt *gt, bool force) __intel_engine_reset(engine, false); } +void intel_gt_pm_disable(struct intel_gt *gt) +{ + if (!is_mock_gt(gt)) + intel_sanitize_gt_powersave(gt->i915); +} + +void intel_gt_pm_fini(struct intel_gt *gt) +{ + intel_rc6_fini(>->rc6); +} + int intel_gt_resume(struct intel_gt *gt) { struct intel_engine_cs *engine; @@ -136,14 +167,21 @@ int intel_gt_resume(struct intel_gt *gt) * allowing us to fixup the user contexts on their first pin. */ intel_gt_pm_get(gt); + intel_uncore_forcewake_get(gt->uncore, FORCEWAKE_ALL); + intel_rc6_sanitize(>->rc6); + for_each_engine(engine, gt->i915, id) { struct intel_context *ce; intel_engine_pm_get(engine); ce = engine->kernel_context; - if (ce) + if (ce) { + GEM_BUG_ON(!intel_context_is_pinned(ce)); + mutex_acquire(&ce->pin_mutex.dep_map, 0, 0, _THIS_IP_); ce->ops->reset(ce); + mutex_release(&ce->pin_mutex.dep_map, 0, _THIS_IP_); + } engine->serial++; /* kernel context lost */ err = engine->resume(engine); @@ -156,11 +194,38 @@ int intel_gt_resume(struct intel_gt *gt) break; } } + + intel_rc6_enable(>->rc6); + intel_uncore_forcewake_put(gt->uncore, FORCEWAKE_ALL); intel_gt_pm_put(gt); return err; } +static void wait_for_idle(struct intel_gt *gt) +{ + if (intel_gt_wait_for_idle(gt, I915_GEM_IDLE_TIMEOUT) == -ETIME) { + /* + * Forcibly cancel outstanding work and leave + * the gpu quiet. + */ + intel_gt_set_wedged(gt); + } + + intel_gt_pm_wait_for_idle(gt); +} + +void intel_gt_suspend(struct intel_gt *gt) +{ + intel_wakeref_t wakeref; + + /* We expect to be idle already; but also want to be independent */ + wait_for_idle(gt); + + with_intel_runtime_pm(>->i915->runtime_pm, wakeref) + intel_rc6_disable(>->rc6); +} + void intel_gt_runtime_suspend(struct intel_gt *gt) { intel_uc_runtime_suspend(>->uc); @@ -172,3 +237,7 @@ int intel_gt_runtime_resume(struct intel_gt *gt) return intel_uc_runtime_resume(>->uc); } + +#if IS_ENABLED(CONFIG_DRM_I915_SELFTEST) +#include "selftest_gt_pm.c" +#endif diff --git a/drivers/gpu/drm/i915/gt/intel_gt_pm.h b/drivers/gpu/drm/i915/gt/intel_gt_pm.h index fb39d99cd6ee..997770d3a968 100644 --- a/drivers/gpu/drm/i915/gt/intel_gt_pm.h +++ b/drivers/gpu/drm/i915/gt/intel_gt_pm.h @@ -43,10 +43,21 @@ static inline int intel_gt_pm_wait_for_idle(struct intel_gt *gt) } void intel_gt_pm_init_early(struct intel_gt *gt); +void intel_gt_pm_init(struct intel_gt *gt); +void intel_gt_pm_disable(struct intel_gt *gt); +void intel_gt_pm_fini(struct intel_gt *gt); void intel_gt_sanitize(struct intel_gt *gt, bool force); + int intel_gt_resume(struct intel_gt *gt); +void intel_gt_suspend(struct intel_gt *gt); + void intel_gt_runtime_suspend(struct intel_gt *gt); int intel_gt_runtime_resume(struct intel_gt *gt); +static inline bool is_mock_gt(const struct intel_gt *gt) +{ + return I915_SELFTEST_ONLY(gt->awake == -ENODEV); +} + #endif /* INTEL_GT_PM_H */ diff --git a/drivers/gpu/drm/i915/gt/intel_gt_requests.c b/drivers/gpu/drm/i915/gt/intel_gt_requests.c new file mode 100644 index 000000000000..8aed89fd2cdc --- /dev/null +++ b/drivers/gpu/drm/i915/gt/intel_gt_requests.c @@ -0,0 +1,123 @@ +/* + * SPDX-License-Identifier: MIT + * + * Copyright © 2019 Intel Corporation + */ + +#include "i915_request.h" +#include "intel_gt.h" +#include "intel_gt_pm.h" +#include "intel_gt_requests.h" +#include "intel_timeline.h" + +static void retire_requests(struct intel_timeline *tl) +{ + struct i915_request *rq, *rn; + + list_for_each_entry_safe(rq, rn, &tl->requests, link) + if (!i915_request_retire(rq)) + break; +} + +long intel_gt_retire_requests_timeout(struct intel_gt *gt, long timeout) +{ + struct intel_gt_timelines *timelines = >->timelines; + struct intel_timeline *tl, *tn; + unsigned long active_count = 0; + unsigned long flags; + bool interruptible; + LIST_HEAD(free); + + interruptible = true; + if (unlikely(timeout < 0)) + timeout = -timeout, interruptible = false; + + spin_lock_irqsave(&timelines->lock, flags); + list_for_each_entry_safe(tl, tn, &timelines->active_list, link) { + if (!mutex_trylock(&tl->mutex)) + continue; + + intel_timeline_get(tl); + GEM_BUG_ON(!tl->active_count); + tl->active_count++; /* pin the list element */ + spin_unlock_irqrestore(&timelines->lock, flags); + + if (timeout > 0) { + struct dma_fence *fence; + + fence = i915_active_fence_get(&tl->last_request); + if (fence) { + timeout = dma_fence_wait_timeout(fence, + true, + timeout); + dma_fence_put(fence); + } + } + + retire_requests(tl); + + spin_lock_irqsave(&timelines->lock, flags); + + /* Resume iteration after dropping lock */ + list_safe_reset_next(tl, tn, link); + if (--tl->active_count) + active_count += !!rcu_access_pointer(tl->last_request.fence); + else + list_del(&tl->link); + + mutex_unlock(&tl->mutex); + + /* Defer the final release to after the spinlock */ + if (refcount_dec_and_test(&tl->kref.refcount)) { + GEM_BUG_ON(tl->active_count); + list_add(&tl->link, &free); + } + } + spin_unlock_irqrestore(&timelines->lock, flags); + + list_for_each_entry_safe(tl, tn, &free, link) + __intel_timeline_free(&tl->kref); + + return active_count ? timeout : 0; +} + +int intel_gt_wait_for_idle(struct intel_gt *gt, long timeout) +{ + /* If the device is asleep, we have no requests outstanding */ + if (!intel_gt_pm_is_awake(gt)) + return 0; + + while ((timeout = intel_gt_retire_requests_timeout(gt, timeout)) > 0) { + cond_resched(); + if (signal_pending(current)) + return -EINTR; + } + + return timeout; +} + +static void retire_work_handler(struct work_struct *work) +{ + struct intel_gt *gt = + container_of(work, typeof(*gt), requests.retire_work.work); + + intel_gt_retire_requests(gt); + schedule_delayed_work(>->requests.retire_work, + round_jiffies_up_relative(HZ)); +} + +void intel_gt_init_requests(struct intel_gt *gt) +{ + INIT_DELAYED_WORK(>->requests.retire_work, retire_work_handler); +} + +void intel_gt_park_requests(struct intel_gt *gt) +{ + cancel_delayed_work(>->requests.retire_work); +} + +void intel_gt_unpark_requests(struct intel_gt *gt) +{ + schedule_delayed_work(>->requests.retire_work, + round_jiffies_up_relative(HZ)); +} diff --git a/drivers/gpu/drm/i915/gt/intel_gt_requests.h b/drivers/gpu/drm/i915/gt/intel_gt_requests.h new file mode 100644 index 000000000000..bd31cbce47e0 --- /dev/null +++ b/drivers/gpu/drm/i915/gt/intel_gt_requests.h @@ -0,0 +1,24 @@ +/* + * SPDX-License-Identifier: MIT + * + * Copyright © 2019 Intel Corporation + */ + +#ifndef INTEL_GT_REQUESTS_H +#define INTEL_GT_REQUESTS_H + +struct intel_gt; + +long intel_gt_retire_requests_timeout(struct intel_gt *gt, long timeout); +static inline void intel_gt_retire_requests(struct intel_gt *gt) +{ + intel_gt_retire_requests_timeout(gt, 0); +} + +int intel_gt_wait_for_idle(struct intel_gt *gt, long timeout); + +void intel_gt_init_requests(struct intel_gt *gt); +void intel_gt_park_requests(struct intel_gt *gt); +void intel_gt_unpark_requests(struct intel_gt *gt); + +#endif /* INTEL_GT_REQUESTS_H */ diff --git a/drivers/gpu/drm/i915/gt/intel_gt_types.h b/drivers/gpu/drm/i915/gt/intel_gt_types.h index dc295c196d11..802f516a3430 100644 --- a/drivers/gpu/drm/i915/gt/intel_gt_types.h +++ b/drivers/gpu/drm/i915/gt/intel_gt_types.h @@ -18,6 +18,7 @@ #include "i915_vma.h" #include "intel_engine_types.h" #include "intel_reset_types.h" +#include "intel_rc6_types.h" #include "intel_wakeref.h" struct drm_i915_private; @@ -49,7 +50,19 @@ struct intel_gt { struct list_head hwsp_free_list; } timelines; + struct intel_gt_requests { + /** + * We leave the user IRQ off as much as possible, + * but this means that requests will finish and never + * be retired once the system goes idle. Set a timer to + * fire periodically while the ring is running. When it + * fires, go retire requests. + */ + struct delayed_work retire_work; + } requests; + struct intel_wakeref wakeref; + atomic_t user_wakeref; struct list_head closed_vma; spinlock_t closed_lock; /* guards the list of closed_vma */ @@ -66,6 +79,8 @@ struct intel_gt { */ intel_wakeref_t awake; + struct intel_rc6 rc6; + struct blocking_notifier_head pm_notifications; ktime_t last_init_time; @@ -89,9 +104,6 @@ enum intel_gt_scratch_field { INTEL_GT_SCRATCH_FIELD_DEFAULT = 0, /* 8 bytes */ - INTEL_GT_SCRATCH_FIELD_CLEAR_SLM_WA = 128, - - /* 8 bytes */ INTEL_GT_SCRATCH_FIELD_RENDER_FLUSH = 128, /* 8 bytes */ diff --git a/drivers/gpu/drm/i915/gt/intel_hangcheck.c b/drivers/gpu/drm/i915/gt/intel_hangcheck.c index 05d042cdefe2..9814b18b32ad 100644 --- a/drivers/gpu/drm/i915/gt/intel_hangcheck.c +++ b/drivers/gpu/drm/i915/gt/intel_hangcheck.c @@ -53,6 +53,7 @@ static bool instdone_unchanged(u32 current_instdone, u32 *old_instdone) static bool subunits_stuck(struct intel_engine_cs *engine) { struct drm_i915_private *dev_priv = engine->i915; + const struct sseu_dev_info *sseu = &RUNTIME_INFO(dev_priv)->sseu; struct intel_instdone instdone; struct intel_instdone *accu_instdone = &engine->hangcheck.instdone; bool stuck; @@ -71,7 +72,7 @@ static bool subunits_stuck(struct intel_engine_cs *engine) stuck &= instdone_unchanged(instdone.slice_common, &accu_instdone->slice_common); - for_each_instdone_slice_subslice(dev_priv, slice, subslice) { + for_each_instdone_slice_subslice(dev_priv, sseu, slice, subslice) { stuck &= instdone_unchanged(instdone.sampler[slice][subslice], &accu_instdone->sampler[slice][subslice]); stuck &= instdone_unchanged(instdone.row[slice][subslice], @@ -283,7 +284,7 @@ static void hangcheck_elapsed(struct work_struct *work) for_each_engine(engine, gt->i915, id) { struct hangcheck hc; - intel_engine_signal_breadcrumbs(engine); + intel_engine_breadcrumbs_irq(engine); hangcheck_load_sample(engine, &hc); hangcheck_accumulate_sample(engine, &hc); diff --git a/drivers/gpu/drm/i915/gt/intel_lrc.c b/drivers/gpu/drm/i915/gt/intel_lrc.c index d42584439f51..468438fb47af 100644 --- a/drivers/gpu/drm/i915/gt/intel_lrc.c +++ b/drivers/gpu/drm/i915/gt/intel_lrc.c @@ -230,9 +230,27 @@ static int __execlists_context_alloc(struct intel_context *ce, struct intel_engine_cs *engine); static void execlists_init_reg_state(u32 *reg_state, - struct intel_context *ce, - struct intel_engine_cs *engine, - struct intel_ring *ring); + const struct intel_context *ce, + const struct intel_engine_cs *engine, + const struct intel_ring *ring, + bool close); + +static void __context_pin_acquire(struct intel_context *ce) +{ + mutex_acquire(&ce->pin_mutex.dep_map, 2, 0, _RET_IP_); +} + +static void __context_pin_release(struct intel_context *ce) +{ + mutex_release(&ce->pin_mutex.dep_map, 0, _RET_IP_); +} + +static void mark_eio(struct i915_request *rq) +{ + if (!i915_request_signaled(rq)) + dma_fence_set_error(&rq->fence, -EIO); + i915_request_mark_complete(rq); +} static inline u32 intel_hws_preempt_address(struct intel_engine_cs *engine) { @@ -422,12 +440,8 @@ assert_priority_queue(const struct i915_request *prev, static u64 lrc_descriptor(struct intel_context *ce, struct intel_engine_cs *engine) { - struct i915_gem_context *ctx = ce->gem_context; u64 desc; - BUILD_BUG_ON(MAX_CONTEXT_HW_ID > (BIT(GEN8_CTX_ID_WIDTH))); - BUILD_BUG_ON(GEN11_MAX_CONTEXT_HW_ID > (BIT(GEN11_SW_CTX_ID_WIDTH))); - desc = INTEL_LEGACY_32B_CONTEXT; if (i915_vm_is_4lvl(ce->vm)) desc = INTEL_LEGACY_64B_CONTEXT; @@ -445,25 +459,421 @@ lrc_descriptor(struct intel_context *ce, struct intel_engine_cs *engine) * anything below. */ if (INTEL_GEN(engine->i915) >= 11) { - GEM_BUG_ON(ctx->hw_id >= BIT(GEN11_SW_CTX_ID_WIDTH)); - desc |= (u64)ctx->hw_id << GEN11_SW_CTX_ID_SHIFT; - /* bits 37-47 */ - desc |= (u64)engine->instance << GEN11_ENGINE_INSTANCE_SHIFT; /* bits 48-53 */ - /* TODO: decide what to do with SW counter (bits 55-60) */ - desc |= (u64)engine->class << GEN11_ENGINE_CLASS_SHIFT; /* bits 61-63 */ - } else { - GEM_BUG_ON(ctx->hw_id >= BIT(GEN8_CTX_ID_WIDTH)); - desc |= (u64)ctx->hw_id << GEN8_CTX_ID_SHIFT; /* bits 32-52 */ } return desc; } +static u32 *set_offsets(u32 *regs, + const u8 *data, + const struct intel_engine_cs *engine) +#define NOP(x) (BIT(7) | (x)) +#define LRI(count, flags) ((flags) << 6 | (count)) +#define POSTED BIT(0) +#define REG(x) (((x) >> 2) | BUILD_BUG_ON_ZERO(x >= 0x200)) +#define REG16(x) \ + (((x) >> 9) | BIT(7) | BUILD_BUG_ON_ZERO(x >= 0x10000)), \ + (((x) >> 2) & 0x7f) +#define END() 0 +{ + const u32 base = engine->mmio_base; + + while (*data) { + u8 count, flags; + + if (*data & BIT(7)) { /* skip */ + regs += *data++ & ~BIT(7); + continue; + } + + count = *data & 0x3f; + flags = *data >> 6; + data++; + + *regs = MI_LOAD_REGISTER_IMM(count); + if (flags & POSTED) + *regs |= MI_LRI_FORCE_POSTED; + if (INTEL_GEN(engine->i915) >= 11) + *regs |= MI_LRI_CS_MMIO; + regs++; + + GEM_BUG_ON(!count); + do { + u32 offset = 0; + u8 v; + + do { + v = *data++; + offset <<= 7; + offset |= v & ~BIT(7); + } while (v & BIT(7)); + + *regs = base + (offset << 2); + regs += 2; + } while (--count); + } + + return regs; +} + +static const u8 gen8_xcs_offsets[] = { + NOP(1), + LRI(11, 0), + REG16(0x244), + REG(0x034), + REG(0x030), + REG(0x038), + REG(0x03c), + REG(0x168), + REG(0x140), + REG(0x110), + REG(0x11c), + REG(0x114), + REG(0x118), + + NOP(9), + LRI(9, 0), + REG16(0x3a8), + REG16(0x28c), + REG16(0x288), + REG16(0x284), + REG16(0x280), + REG16(0x27c), + REG16(0x278), + REG16(0x274), + REG16(0x270), + + NOP(13), + LRI(2, 0), + REG16(0x200), + REG(0x028), + + END(), +}; + +static const u8 gen9_xcs_offsets[] = { + NOP(1), + LRI(14, POSTED), + REG16(0x244), + REG(0x034), + REG(0x030), + REG(0x038), + REG(0x03c), + REG(0x168), + REG(0x140), + REG(0x110), + REG(0x11c), + REG(0x114), + REG(0x118), + REG(0x1c0), + REG(0x1c4), + REG(0x1c8), + + NOP(3), + LRI(9, POSTED), + REG16(0x3a8), + REG16(0x28c), + REG16(0x288), + REG16(0x284), + REG16(0x280), + REG16(0x27c), + REG16(0x278), + REG16(0x274), + REG16(0x270), + + NOP(13), + LRI(1, POSTED), + REG16(0x200), + + NOP(13), + LRI(44, POSTED), + REG(0x028), + REG(0x09c), + REG(0x0c0), + REG(0x178), + REG(0x17c), + REG16(0x358), + REG(0x170), + REG(0x150), + REG(0x154), + REG(0x158), + REG16(0x41c), + REG16(0x600), + REG16(0x604), + REG16(0x608), + REG16(0x60c), + REG16(0x610), + REG16(0x614), + REG16(0x618), + REG16(0x61c), + REG16(0x620), + REG16(0x624), + REG16(0x628), + REG16(0x62c), + REG16(0x630), + REG16(0x634), + REG16(0x638), + REG16(0x63c), + REG16(0x640), + REG16(0x644), + REG16(0x648), + REG16(0x64c), + REG16(0x650), + REG16(0x654), + REG16(0x658), + REG16(0x65c), + REG16(0x660), + REG16(0x664), + REG16(0x668), + REG16(0x66c), + REG16(0x670), + REG16(0x674), + REG16(0x678), + REG16(0x67c), + REG(0x068), + + END(), +}; + +static const u8 gen12_xcs_offsets[] = { + NOP(1), + LRI(13, POSTED), + REG16(0x244), + REG(0x034), + REG(0x030), + REG(0x038), + REG(0x03c), + REG(0x168), + REG(0x140), + REG(0x110), + REG(0x1c0), + REG(0x1c4), + REG(0x1c8), + REG(0x180), + REG16(0x2b4), + + NOP(5), + LRI(9, POSTED), + REG16(0x3a8), + REG16(0x28c), + REG16(0x288), + REG16(0x284), + REG16(0x280), + REG16(0x27c), + REG16(0x278), + REG16(0x274), + REG16(0x270), + + NOP(13), + LRI(2, POSTED), + REG16(0x200), + REG16(0x204), + + NOP(11), + LRI(50, POSTED), + REG16(0x588), + REG16(0x588), + REG16(0x588), + REG16(0x588), + REG16(0x588), + REG16(0x588), + REG(0x028), + REG(0x09c), + REG(0x0c0), + REG(0x178), + REG(0x17c), + REG16(0x358), + REG(0x170), + REG(0x150), + REG(0x154), + REG(0x158), + REG16(0x41c), + REG16(0x600), + REG16(0x604), + REG16(0x608), + REG16(0x60c), + REG16(0x610), + REG16(0x614), + REG16(0x618), + REG16(0x61c), + REG16(0x620), + REG16(0x624), + REG16(0x628), + REG16(0x62c), + REG16(0x630), + REG16(0x634), + REG16(0x638), + REG16(0x63c), + REG16(0x640), + REG16(0x644), + REG16(0x648), + REG16(0x64c), + REG16(0x650), + REG16(0x654), + REG16(0x658), + REG16(0x65c), + REG16(0x660), + REG16(0x664), + REG16(0x668), + REG16(0x66c), + REG16(0x670), + REG16(0x674), + REG16(0x678), + REG16(0x67c), + REG(0x068), + + END(), +}; + +static const u8 gen8_rcs_offsets[] = { + NOP(1), + LRI(14, POSTED), + REG16(0x244), + REG(0x034), + REG(0x030), + REG(0x038), + REG(0x03c), + REG(0x168), + REG(0x140), + REG(0x110), + REG(0x11c), + REG(0x114), + REG(0x118), + REG(0x1c0), + REG(0x1c4), + REG(0x1c8), + + NOP(3), + LRI(9, POSTED), + REG16(0x3a8), + REG16(0x28c), + REG16(0x288), + REG16(0x284), + REG16(0x280), + REG16(0x27c), + REG16(0x278), + REG16(0x274), + REG16(0x270), + + NOP(13), + LRI(1, 0), + REG(0x0c8), + + END(), +}; + +static const u8 gen11_rcs_offsets[] = { + NOP(1), + LRI(15, POSTED), + REG16(0x244), + REG(0x034), + REG(0x030), + REG(0x038), + REG(0x03c), + REG(0x168), + REG(0x140), + REG(0x110), + REG(0x11c), + REG(0x114), + REG(0x118), + REG(0x1c0), + REG(0x1c4), + REG(0x1c8), + REG(0x180), + + NOP(1), + LRI(9, POSTED), + REG16(0x3a8), + REG16(0x28c), + REG16(0x288), + REG16(0x284), + REG16(0x280), + REG16(0x27c), + REG16(0x278), + REG16(0x274), + REG16(0x270), + + LRI(1, POSTED), + REG(0x1b0), + + NOP(10), + LRI(1, 0), + REG(0x0c8), + + END(), +}; + +static const u8 gen12_rcs_offsets[] = { + NOP(1), + LRI(13, POSTED), + REG16(0x244), + REG(0x034), + REG(0x030), + REG(0x038), + REG(0x03c), + REG(0x168), + REG(0x140), + REG(0x110), + REG(0x1c0), + REG(0x1c4), + REG(0x1c8), + REG(0x180), + REG16(0x2b4), + + NOP(5), + LRI(9, POSTED), + REG16(0x3a8), + REG16(0x28c), + REG16(0x288), + REG16(0x284), + REG16(0x280), + REG16(0x27c), + REG16(0x278), + REG16(0x274), + REG16(0x270), + + LRI(3, POSTED), + REG(0x1b0), + REG16(0x5a8), + REG16(0x5ac), + + NOP(6), + LRI(1, 0), + REG(0x0c8), + + END(), +}; + +#undef END +#undef REG16 +#undef REG +#undef LRI +#undef NOP + +static const u8 *reg_offsets(const struct intel_engine_cs *engine) +{ + if (engine->class == RENDER_CLASS) { + if (INTEL_GEN(engine->i915) >= 12) + return gen12_rcs_offsets; + else if (INTEL_GEN(engine->i915) >= 11) + return gen11_rcs_offsets; + else + return gen8_rcs_offsets; + } else { + if (INTEL_GEN(engine->i915) >= 12) + return gen12_xcs_offsets; + else if (INTEL_GEN(engine->i915) >= 9) + return gen9_xcs_offsets; + else + return gen8_xcs_offsets; + } +} + static void unwind_wa_tail(struct i915_request *rq) { rq->tail = intel_ring_wrap(rq->ring, rq->wa_tail - WA_TAIL_BYTES); @@ -562,6 +972,18 @@ __execlists_schedule_in(struct i915_request *rq) intel_context_get(ce); + if (ce->tag) { + /* Use a fixed tag for OA and friends */ + ce->lrc_desc |= (u64)ce->tag << 32; + } else { + /* We don't need a strict matching tag, just different values */ + ce->lrc_desc &= ~GENMASK_ULL(47, 37); + ce->lrc_desc |= + (u64)(engine->context_tag++ % NUM_CONTEXT_TAG) << + GEN11_SW_CTX_ID_SHIFT; + BUILD_BUG_ON(NUM_CONTEXT_TAG > GEN12_MAX_CONTEXT_HW_ID); + } + intel_gt_pm_get(engine->gt); execlists_context_status_change(rq, INTEL_CONTEXT_SCHEDULE_IN); intel_engine_context_in(engine); @@ -631,7 +1053,6 @@ execlists_schedule_out(struct i915_request *rq) struct intel_engine_cs *cur, *old; trace_i915_request_out(rq); - GEM_BUG_ON(intel_context_inflight(ce) != rq->engine); old = READ_ONCE(ce->inflight); do @@ -648,7 +1069,7 @@ static u64 execlists_update_context(const struct i915_request *rq) struct intel_context *ce = rq->hw_context; u64 desc; - ce->lrc_reg_state[CTX_RING_TAIL + 1] = + ce->lrc_reg_state[CTX_RING_TAIL] = intel_ring_set_tail(rq->ring, rq->tail); /* @@ -693,6 +1114,9 @@ trace_ports(const struct intel_engine_execlists *execlists, const struct intel_engine_cs *engine = container_of(execlists, typeof(*engine), execlists); + if (!ports[0]) + return; + GEM_TRACE("%s: %s { %llx:%lld%s, %llx:%lld }\n", engine->name, msg, ports[0]->fence.context, @@ -797,6 +1221,17 @@ static bool can_merge_rq(const struct i915_request *prev, GEM_BUG_ON(prev == next); GEM_BUG_ON(!assert_priority_queue(prev, next)); + /* + * We do not submit known completed requests. Therefore if the next + * request is already completed, we can pretend to merge it in + * with the previous context (and we will skip updating the ELSP + * and tracking). Thus hopefully keeping the ELSP full with active + * contexts, despite the best efforts of preempt-to-busy to confuse + * us. + */ + if (i915_request_completed(next)) + return true; + if (!can_merge_ctx(prev->hw_context, next->hw_context)) return false; @@ -806,47 +1241,7 @@ static bool can_merge_rq(const struct i915_request *prev, static void virtual_update_register_offsets(u32 *regs, struct intel_engine_cs *engine) { - u32 base = engine->mmio_base; - - /* Must match execlists_init_reg_state()! */ - - regs[CTX_CONTEXT_CONTROL] = - i915_mmio_reg_offset(RING_CONTEXT_CONTROL(base)); - regs[CTX_RING_HEAD] = i915_mmio_reg_offset(RING_HEAD(base)); - regs[CTX_RING_TAIL] = i915_mmio_reg_offset(RING_TAIL(base)); - regs[CTX_RING_BUFFER_START] = i915_mmio_reg_offset(RING_START(base)); - regs[CTX_RING_BUFFER_CONTROL] = i915_mmio_reg_offset(RING_CTL(base)); - - regs[CTX_BB_HEAD_U] = i915_mmio_reg_offset(RING_BBADDR_UDW(base)); - regs[CTX_BB_HEAD_L] = i915_mmio_reg_offset(RING_BBADDR(base)); - regs[CTX_BB_STATE] = i915_mmio_reg_offset(RING_BBSTATE(base)); - regs[CTX_SECOND_BB_HEAD_U] = - i915_mmio_reg_offset(RING_SBBADDR_UDW(base)); - regs[CTX_SECOND_BB_HEAD_L] = i915_mmio_reg_offset(RING_SBBADDR(base)); - regs[CTX_SECOND_BB_STATE] = i915_mmio_reg_offset(RING_SBBSTATE(base)); - - regs[CTX_CTX_TIMESTAMP] = - i915_mmio_reg_offset(RING_CTX_TIMESTAMP(base)); - regs[CTX_PDP3_UDW] = i915_mmio_reg_offset(GEN8_RING_PDP_UDW(base, 3)); - regs[CTX_PDP3_LDW] = i915_mmio_reg_offset(GEN8_RING_PDP_LDW(base, 3)); - regs[CTX_PDP2_UDW] = i915_mmio_reg_offset(GEN8_RING_PDP_UDW(base, 2)); - regs[CTX_PDP2_LDW] = i915_mmio_reg_offset(GEN8_RING_PDP_LDW(base, 2)); - regs[CTX_PDP1_UDW] = i915_mmio_reg_offset(GEN8_RING_PDP_UDW(base, 1)); - regs[CTX_PDP1_LDW] = i915_mmio_reg_offset(GEN8_RING_PDP_LDW(base, 1)); - regs[CTX_PDP0_UDW] = i915_mmio_reg_offset(GEN8_RING_PDP_UDW(base, 0)); - regs[CTX_PDP0_LDW] = i915_mmio_reg_offset(GEN8_RING_PDP_LDW(base, 0)); - - if (engine->class == RENDER_CLASS) { - regs[CTX_RCS_INDIRECT_CTX] = - i915_mmio_reg_offset(RING_INDIRECT_CTX(base)); - regs[CTX_RCS_INDIRECT_CTX_OFFSET] = - i915_mmio_reg_offset(RING_INDIRECT_CTX_OFFSET(base)); - regs[CTX_BB_PER_CTX_PTR] = - i915_mmio_reg_offset(RING_BB_PER_CTX_PTR(base)); - - regs[CTX_R_PWR_CLK_STATE] = - i915_mmio_reg_offset(GEN8_R_PWR_CLK_STATE); - } + set_offsets(regs, reg_offsets(engine), engine); } static bool virtual_matches(const struct virtual_engine *ve, @@ -1172,21 +1567,6 @@ static void execlists_dequeue(struct intel_engine_cs *engine) continue; } - if (i915_request_completed(rq)) { - ve->request = NULL; - ve->base.execlists.queue_priority_hint = INT_MIN; - rb_erase_cached(rb, &execlists->virtual); - RB_CLEAR_NODE(rb); - - rq->engine = engine; - __i915_request_submit(rq); - - spin_unlock(&ve->base.active.lock); - - rb = rb_first_cached(&execlists->virtual); - continue; - } - if (last && !can_merge_rq(last, rq)) { spin_unlock(&ve->base.active.lock); return; /* leave this for another */ @@ -1214,7 +1594,10 @@ static void execlists_dequeue(struct intel_engine_cs *engine) unsigned int n; GEM_BUG_ON(READ_ONCE(ve->context.inflight)); - virtual_update_register_offsets(regs, engine); + + if (!intel_engine_has_relative_mmio(engine)) + virtual_update_register_offsets(regs, + engine); if (!list_empty(&ve->context.signals)) virtual_xfer_breadcrumbs(ve, engine); @@ -1237,11 +1620,24 @@ static void execlists_dequeue(struct intel_engine_cs *engine) GEM_BUG_ON(ve->siblings[0] != engine); } - __i915_request_submit(rq); - if (!i915_request_completed(rq)) { + if (__i915_request_submit(rq)) { submit = true; last = rq; } + i915_request_put(rq); + + /* + * Hmm, we have a bunch of virtual engine requests, + * but the first one was already completed (thanks + * preempt-to-busy!). Keep looking at the veng queue + * until we have no more relevant requests (i.e. + * the normal submit queue has higher priority). + */ + if (!submit) { + spin_unlock(&ve->base.active.lock); + rb = rb_first_cached(&execlists->virtual); + continue; + } } spin_unlock(&ve->base.active.lock); @@ -1254,8 +1650,7 @@ static void execlists_dequeue(struct intel_engine_cs *engine) int i; priolist_for_each_request_consume(rq, rn, p, i) { - if (i915_request_completed(rq)) - goto skip; + bool merge = true; /* * Can we combine this request with the current port? @@ -1296,14 +1691,23 @@ static void execlists_dequeue(struct intel_engine_cs *engine) ctx_single_port_submission(rq->hw_context)) goto done; - *port = execlists_schedule_in(last, port - execlists->pending); - port++; + merge = false; } - last = rq; - submit = true; -skip: - __i915_request_submit(rq); + if (__i915_request_submit(rq)) { + if (!merge) { + *port = execlists_schedule_in(last, port - execlists->pending); + port++; + last = NULL; + } + + GEM_BUG_ON(last && + !can_merge_ctx(last->hw_context, + rq->hw_context)); + + submit = true; + last = rq; + } } rb_erase_cached(&p->node, &execlists->queue); @@ -1334,11 +1738,26 @@ done: if (submit) { *port = execlists_schedule_in(last, port - execlists->pending); - memset(port + 1, 0, (last_port - port) * sizeof(*port)); execlists->switch_priority_hint = switch_prio(engine, *execlists->pending); + + /* + * Skip if we ended up with exactly the same set of requests, + * e.g. trying to timeslice a pair of ordered contexts + */ + if (!memcmp(execlists->active, execlists->pending, + (port - execlists->pending + 1) * sizeof(*port))) { + do + execlists_schedule_out(fetch_and_zero(port)); + while (port-- != execlists->pending); + + goto skip_submit; + } + + memset(port + 1, 0, (last_port - port) * sizeof(*port)); execlists_submit_ports(engine); } else { +skip_submit: ring_set_paused(engine, 0); } } @@ -1371,13 +1790,6 @@ reset_in_progress(const struct intel_engine_execlists *execlists) return unlikely(!__tasklet_is_enabled(&execlists->tasklet)); } -enum csb_step { - CSB_NOP, - CSB_PROMOTE, - CSB_PREEMPT, - CSB_COMPLETE, -}; - /* * Starting with Gen12, the status has a new format: * @@ -1404,7 +1816,7 @@ enum csb_step { * bits 47-57: sw context id of the lrc the GT switched away from * bits 58-63: sw counter of the lrc the GT switched away from */ -static inline enum csb_step +static inline bool gen12_csb_parse(const struct intel_engine_execlists *execlists, const u32 *csb) { u32 lower_dw = csb[0]; @@ -1413,9 +1825,6 @@ gen12_csb_parse(const struct intel_engine_execlists *execlists, const u32 *csb) bool ctx_away_valid = GEN12_CSB_CTX_VALID(upper_dw); bool new_queue = lower_dw & GEN12_CTX_STATUS_SWITCHED_TO_NEW_QUEUE; - if (!ctx_away_valid && ctx_to_valid) - return CSB_PROMOTE; - /* * The context switch detail is not guaranteed to be 5 when a preemption * occurs, so we can't just check for that. The check below works for @@ -1423,8 +1832,10 @@ gen12_csb_parse(const struct intel_engine_execlists *execlists, const u32 *csb) * instructions and lite-restore. Preempt-to-idle via the CTRL register * would require some extra handling, but we don't support that. */ - if (new_queue && ctx_away_valid) - return CSB_PREEMPT; + if (!ctx_away_valid || new_queue) { + GEM_BUG_ON(!ctx_to_valid); + return true; + } /* * switch detail = 5 is covered by the case above and we do not expect a @@ -1432,30 +1843,13 @@ gen12_csb_parse(const struct intel_engine_execlists *execlists, const u32 *csb) * use polling mode. */ GEM_BUG_ON(GEN12_CTX_SWITCH_DETAIL(upper_dw)); - - if (*execlists->active) { - GEM_BUG_ON(!ctx_away_valid); - return CSB_COMPLETE; - } - - return CSB_NOP; + return false; } -static inline enum csb_step +static inline bool gen8_csb_parse(const struct intel_engine_execlists *execlists, const u32 *csb) { - unsigned int status = *csb; - - if (status & GEN8_CTX_STATUS_IDLE_ACTIVE) - return CSB_PROMOTE; - - if (status & GEN8_CTX_STATUS_PREEMPTED) - return CSB_PREEMPT; - - if (*execlists->active) - return CSB_COMPLETE; - - return CSB_NOP; + return *csb & (GEN8_CTX_STATUS_IDLE_ACTIVE | GEN8_CTX_STATUS_PREEMPTED); } static void process_csb(struct intel_engine_cs *engine) @@ -1494,7 +1888,7 @@ static void process_csb(struct intel_engine_cs *engine) rmb(); do { - enum csb_step csb_step; + bool promote; if (++head == num_entries) head = 0; @@ -1522,20 +1916,16 @@ static void process_csb(struct intel_engine_cs *engine) buf[2 * head + 0], buf[2 * head + 1]); if (INTEL_GEN(engine->i915) >= 12) - csb_step = gen12_csb_parse(execlists, buf + 2 * head); + promote = gen12_csb_parse(execlists, buf + 2 * head); else - csb_step = gen8_csb_parse(execlists, buf + 2 * head); - - switch (csb_step) { - case CSB_PREEMPT: /* cancel old inflight, prepare for switch */ + promote = gen8_csb_parse(execlists, buf + 2 * head); + if (promote) { + /* cancel old inflight, prepare for switch */ trace_ports(execlists, "preempted", execlists->active); - while (*execlists->active) execlists_schedule_out(*execlists->active++); - /* fallthrough */ - case CSB_PROMOTE: /* switch pending to inflight */ - GEM_BUG_ON(*execlists->active); + /* switch pending to inflight */ GEM_BUG_ON(!assert_pending_valid(execlists, "promote")); execlists->active = memcpy(execlists->inflight, @@ -1550,9 +1940,10 @@ static void process_csb(struct intel_engine_cs *engine) ring_set_paused(engine, 0); WRITE_ONCE(execlists->pending[0], NULL); - break; + } else { + GEM_BUG_ON(!*execlists->active); - case CSB_COMPLETE: /* port0 completed, advanced to port1 */ + /* port0 completed, advanced to port1 */ trace_ports(execlists, "completed", execlists->active); /* @@ -1567,10 +1958,6 @@ static void process_csb(struct intel_engine_cs *engine) GEM_BUG_ON(execlists->active - execlists->inflight > execlists_num_ports(execlists)); - break; - - case CSB_NOP: - break; } } while (head != tail); @@ -1726,14 +2113,13 @@ static void execlists_context_unpin(struct intel_context *ce) check_redzone((void *)ce->lrc_reg_state - LRC_STATE_PN * PAGE_SIZE, ce->engine); - i915_gem_context_unpin_hw_id(ce->gem_context); i915_gem_object_unpin_map(ce->state->obj); intel_ring_reset(ce->ring, ce->ring->tail); } static void -__execlists_update_reg_state(struct intel_context *ce, - struct intel_engine_cs *engine) +__execlists_update_reg_state(const struct intel_context *ce, + const struct intel_engine_cs *engine) { struct intel_ring *ring = ce->ring; u32 *regs = ce->lrc_reg_state; @@ -1741,16 +2127,16 @@ __execlists_update_reg_state(struct intel_context *ce, GEM_BUG_ON(!intel_ring_offset_valid(ring, ring->head)); GEM_BUG_ON(!intel_ring_offset_valid(ring, ring->tail)); - regs[CTX_RING_BUFFER_START + 1] = i915_ggtt_offset(ring->vma); - regs[CTX_RING_HEAD + 1] = ring->head; - regs[CTX_RING_TAIL + 1] = ring->tail; + regs[CTX_RING_BUFFER_START] = i915_ggtt_offset(ring->vma); + regs[CTX_RING_HEAD] = ring->head; + regs[CTX_RING_TAIL] = ring->tail; /* RPCS */ if (engine->class == RENDER_CLASS) { - regs[CTX_R_PWR_CLK_STATE + 1] = + regs[CTX_R_PWR_CLK_STATE] = intel_sseu_make_rpcs(engine->i915, &ce->sseu); - i915_oa_init_reg_state(engine, ce, regs); + i915_oa_init_reg_state(ce, engine); } } @@ -1776,18 +2162,12 @@ __execlists_context_pin(struct intel_context *ce, goto unpin_active; } - ret = i915_gem_context_pin_hw_id(ce->gem_context); - if (ret) - goto unpin_map; - ce->lrc_desc = lrc_descriptor(ce, engine); ce->lrc_reg_state = vaddr + LRC_STATE_PN * PAGE_SIZE; __execlists_update_reg_state(ce, engine); return 0; -unpin_map: - i915_gem_object_unpin_map(ce->state->obj); unpin_active: intel_context_active_release(ce); err: @@ -1843,7 +2223,7 @@ static int gen8_emit_init_breadcrumb(struct i915_request *rq) { u32 *cs; - GEM_BUG_ON(!rq->timeline->has_initial_breadcrumb); + GEM_BUG_ON(!i915_request_timeline(rq)->has_initial_breadcrumb); cs = intel_ring_begin(rq, 6); if (IS_ERR(cs)) @@ -1859,7 +2239,7 @@ static int gen8_emit_init_breadcrumb(struct i915_request *rq) *cs++ = MI_NOOP; *cs++ = MI_STORE_DWORD_IMM_GEN4 | MI_USE_GGTT; - *cs++ = rq->timeline->hwsp_offset; + *cs++ = i915_request_timeline(rq)->hwsp_offset; *cs++ = 0; *cs++ = rq->fence.seqno - 1; @@ -1871,60 +2251,6 @@ static int gen8_emit_init_breadcrumb(struct i915_request *rq) return 0; } -static int emit_pdps(struct i915_request *rq) -{ - const struct intel_engine_cs * const engine = rq->engine; - struct i915_ppgtt * const ppgtt = i915_vm_to_ppgtt(rq->hw_context->vm); - int err, i; - u32 *cs; - - GEM_BUG_ON(intel_vgpu_active(rq->i915)); - - /* - * Beware ye of the dragons, this sequence is magic! - * - * Small changes to this sequence can cause anything from - * GPU hangs to forcewake errors and machine lockups! - */ - - /* Flush any residual operations from the context load */ - err = engine->emit_flush(rq, EMIT_FLUSH); - if (err) - return err; - - /* Magic required to prevent forcewake errors! */ - err = engine->emit_flush(rq, EMIT_INVALIDATE); - if (err) - return err; - - cs = intel_ring_begin(rq, 4 * GEN8_3LVL_PDPES + 2); - if (IS_ERR(cs)) - return PTR_ERR(cs); - - /* Ensure the LRI have landed before we invalidate & continue */ - *cs++ = MI_LOAD_REGISTER_IMM(2 * GEN8_3LVL_PDPES) | MI_LRI_FORCE_POSTED; - for (i = GEN8_3LVL_PDPES; i--; ) { - const dma_addr_t pd_daddr = i915_page_dir_dma_addr(ppgtt, i); - u32 base = engine->mmio_base; - - *cs++ = i915_mmio_reg_offset(GEN8_RING_PDP_UDW(base, i)); - *cs++ = upper_32_bits(pd_daddr); - *cs++ = i915_mmio_reg_offset(GEN8_RING_PDP_LDW(base, i)); - *cs++ = lower_32_bits(pd_daddr); - } - *cs++ = MI_NOOP; - - intel_ring_advance(rq, cs); - - /* Be doubly sure the LRI have landed before proceeding */ - err = engine->emit_flush(rq, EMIT_FLUSH); - if (err) - return err; - - /* Re-invalidate the TLB for luck */ - return engine->emit_flush(rq, EMIT_INVALIDATE); -} - static int execlists_request_alloc(struct i915_request *request) { int ret; @@ -1947,10 +2273,7 @@ static int execlists_request_alloc(struct i915_request *request) */ /* Unconditionally invalidate GPU caches and TLBs. */ - if (i915_vm_is_4lvl(request->hw_context->vm)) - ret = request->engine->emit_flush(request, EMIT_INVALIDATE); - else - ret = emit_pdps(request); + ret = request->engine->emit_flush(request, EMIT_INVALIDATE); if (ret) return ret; @@ -2002,12 +2325,6 @@ gen8_emit_flush_coherentl3_wa(struct intel_engine_cs *engine, u32 *batch) return batch; } -static u32 slm_offset(struct intel_engine_cs *engine) -{ - return intel_gt_scratch_offset(engine->gt, - INTEL_GT_SCRATCH_FIELD_CLEAR_SLM_WA); -} - /* * Typically we only have one indirect_ctx and per_ctx batch buffer which are * initialized at the beginning and shared across all contexts but this field @@ -2036,10 +2353,10 @@ static u32 *gen8_init_indirectctx_bb(struct intel_engine_cs *engine, u32 *batch) /* Actual scratch location is at 128 bytes offset */ batch = gen8_emit_pipe_control(batch, PIPE_CONTROL_FLUSH_L3 | - PIPE_CONTROL_GLOBAL_GTT_IVB | + PIPE_CONTROL_STORE_DATA_INDEX | PIPE_CONTROL_CS_STALL | PIPE_CONTROL_QW_WRITE, - slm_offset(engine)); + LRC_PPHWSP_SCRATCH_ADDR); *batch++ = MI_ARB_ON_OFF | MI_ARB_ENABLE; @@ -2399,10 +2716,14 @@ static void reset_csb_pointers(struct intel_engine_cs *engine) static struct i915_request *active_request(struct i915_request *rq) { - const struct list_head * const list = &rq->timeline->requests; const struct intel_context * const ce = rq->hw_context; struct i915_request *active = NULL; + struct list_head *list; + + if (!i915_request_is_active(rq)) /* unwound, but incomplete! */ + return rq; + list = &i915_request_active_timeline(rq)->requests; list_for_each_entry_from_reverse(rq, list, link) { if (i915_request_completed(rq)) break; @@ -2416,6 +2737,17 @@ static struct i915_request *active_request(struct i915_request *rq) return active; } +static void __execlists_reset_reg_state(const struct intel_context *ce, + const struct intel_engine_cs *engine) +{ + u32 *regs = ce->lrc_reg_state; + + if (INTEL_GEN(engine->i915) >= 9) { + regs[GEN9_CTX_RING_MI_MODE + 1] &= ~STOP_RING; + regs[GEN9_CTX_RING_MI_MODE + 1] |= STOP_RING << 16; + } +} + static void __execlists_reset(struct intel_engine_cs *engine, bool stalled) { struct intel_engine_execlists * const execlists = &engine->execlists; @@ -2423,6 +2755,10 @@ static void __execlists_reset(struct intel_engine_cs *engine, bool stalled) struct i915_request *rq; u32 *regs; + mb(); /* paranoia: read the CSB pointers from after the reset */ + clflush(execlists->csb_write); + mb(); + process_csb(engine); /* drain preemption events */ /* Following the reset, we need to reload the CSB read/write pointers */ @@ -2440,6 +2776,10 @@ static void __execlists_reset(struct intel_engine_cs *engine, bool stalled) ce = rq->hw_context; GEM_BUG_ON(i915_active_is_idle(&ce->active)); GEM_BUG_ON(!i915_vma_is_pinned(ce->state)); + + /* Proclaim we have exclusive access to the context image! */ + __context_pin_acquire(ce); + rq = active_request(rq); if (!rq) { ce->ring->head = ce->ring->tail; @@ -2486,19 +2826,23 @@ static void __execlists_reset(struct intel_engine_cs *engine, bool stalled) * future request will be after userspace has had the opportunity * to recreate its own state. */ + GEM_BUG_ON(!intel_context_is_pinned(ce)); regs = ce->lrc_reg_state; if (engine->pinned_default_state) { memcpy(regs, /* skip restoring the vanilla PPHWSP */ engine->pinned_default_state + LRC_STATE_PN * PAGE_SIZE, engine->context_size - PAGE_SIZE); } - execlists_init_reg_state(regs, ce, engine, ce->ring); + execlists_init_reg_state(regs, ce, engine, ce->ring, false); out_replay: GEM_TRACE("%s replay {head:%04x, tail:%04x\n", engine->name, ce->ring->head, ce->ring->tail); intel_ring_update_space(ce->ring); + __execlists_reset_reg_state(ce, engine); __execlists_update_reg_state(ce, engine); + ce->lrc_desc |= CTX_DESC_FORCE_RESTORE; /* paranoid: GPU was reset! */ + __context_pin_release(ce); unwind: /* Push back any incomplete requests for replay after the reset. */ @@ -2552,12 +2896,8 @@ static void execlists_cancel_requests(struct intel_engine_cs *engine) __execlists_reset(engine, true); /* Mark all executing requests as skipped. */ - list_for_each_entry(rq, &engine->active.requests, sched.link) { - if (!i915_request_signaled(rq)) - dma_fence_set_error(&rq->fence, -EIO); - - i915_request_mark_complete(rq); - } + list_for_each_entry(rq, &engine->active.requests, sched.link) + mark_eio(rq); /* Flush the queued requests to the timeline list (for retiring). */ while ((rb = rb_first_cached(&execlists->queue))) { @@ -2565,10 +2905,8 @@ static void execlists_cancel_requests(struct intel_engine_cs *engine) int i; priolist_for_each_request_consume(rq, rn, p, i) { - list_del_init(&rq->sched.link); + mark_eio(rq); __i915_request_submit(rq); - dma_fence_set_error(&rq->fence, -EIO); - i915_request_mark_complete(rq); } rb_erase_cached(&p->node, &execlists->queue); @@ -2584,13 +2922,15 @@ static void execlists_cancel_requests(struct intel_engine_cs *engine) RB_CLEAR_NODE(rb); spin_lock(&ve->base.active.lock); - if (ve->request) { - ve->request->engine = engine; - __i915_request_submit(ve->request); - dma_fence_set_error(&ve->request->fence, -EIO); - i915_request_mark_complete(ve->request); + rq = fetch_and_zero(&ve->request); + if (rq) { + mark_eio(rq); + + rq->engine = engine; + __i915_request_submit(rq); + i915_request_put(rq); + ve->base.execlists.queue_priority_hint = INT_MIN; - ve->request = NULL; } spin_unlock(&ve->base.active.lock); } @@ -2723,7 +3063,7 @@ static int gen8_emit_flush(struct i915_request *request, u32 mode) } *cs++ = cmd; - *cs++ = I915_GEM_HWS_SCRATCH_ADDR | MI_FLUSH_DW_USE_GTT; + *cs++ = LRC_PPHWSP_SCRATCH_ADDR; *cs++ = 0; /* upper addr */ *cs++ = 0; /* value */ intel_ring_advance(request, cs); @@ -2734,10 +3074,6 @@ static int gen8_emit_flush(struct i915_request *request, u32 mode) static int gen8_emit_flush_render(struct i915_request *request, u32 mode) { - struct intel_engine_cs *engine = request->engine; - u32 scratch_addr = - intel_gt_scratch_offset(engine->gt, - INTEL_GT_SCRATCH_FIELD_RENDER_FLUSH); bool vf_flush_wa = false, dc_flush_wa = false; u32 *cs, flags = 0; int len; @@ -2759,7 +3095,7 @@ static int gen8_emit_flush_render(struct i915_request *request, flags |= PIPE_CONTROL_CONST_CACHE_INVALIDATE; flags |= PIPE_CONTROL_STATE_CACHE_INVALIDATE; flags |= PIPE_CONTROL_QW_WRITE; - flags |= PIPE_CONTROL_GLOBAL_GTT_IVB; + flags |= PIPE_CONTROL_STORE_DATA_INDEX; /* * On GEN9: before VF_CACHE_INVALIDATE we need to emit a NULL @@ -2792,7 +3128,7 @@ static int gen8_emit_flush_render(struct i915_request *request, cs = gen8_emit_pipe_control(cs, PIPE_CONTROL_DC_FLUSH_ENABLE, 0); - cs = gen8_emit_pipe_control(cs, flags, scratch_addr); + cs = gen8_emit_pipe_control(cs, flags, LRC_PPHWSP_SCRATCH_ADDR); if (dc_flush_wa) cs = gen8_emit_pipe_control(cs, PIPE_CONTROL_CS_STALL, 0); @@ -2805,11 +3141,6 @@ static int gen8_emit_flush_render(struct i915_request *request, static int gen11_emit_flush_render(struct i915_request *request, u32 mode) { - struct intel_engine_cs *engine = request->engine; - const u32 scratch_addr = - intel_gt_scratch_offset(engine->gt, - INTEL_GT_SCRATCH_FIELD_RENDER_FLUSH); - if (mode & EMIT_FLUSH) { u32 *cs; u32 flags = 0; @@ -2822,13 +3153,13 @@ static int gen11_emit_flush_render(struct i915_request *request, flags |= PIPE_CONTROL_DC_FLUSH_ENABLE; flags |= PIPE_CONTROL_FLUSH_ENABLE; flags |= PIPE_CONTROL_QW_WRITE; - flags |= PIPE_CONTROL_GLOBAL_GTT_IVB; + flags |= PIPE_CONTROL_STORE_DATA_INDEX; cs = intel_ring_begin(request, 6); if (IS_ERR(cs)) return PTR_ERR(cs); - cs = gen8_emit_pipe_control(cs, flags, scratch_addr); + cs = gen8_emit_pipe_control(cs, flags, LRC_PPHWSP_SCRATCH_ADDR); intel_ring_advance(request, cs); } @@ -2846,13 +3177,81 @@ static int gen11_emit_flush_render(struct i915_request *request, flags |= PIPE_CONTROL_CONST_CACHE_INVALIDATE; flags |= PIPE_CONTROL_STATE_CACHE_INVALIDATE; flags |= PIPE_CONTROL_QW_WRITE; - flags |= PIPE_CONTROL_GLOBAL_GTT_IVB; + flags |= PIPE_CONTROL_STORE_DATA_INDEX; cs = intel_ring_begin(request, 6); if (IS_ERR(cs)) return PTR_ERR(cs); - cs = gen8_emit_pipe_control(cs, flags, scratch_addr); + cs = gen8_emit_pipe_control(cs, flags, LRC_PPHWSP_SCRATCH_ADDR); + intel_ring_advance(request, cs); + } + + return 0; +} + +static u32 preparser_disable(bool state) +{ + return MI_ARB_CHECK | 1 << 8 | state; +} + +static int gen12_emit_flush_render(struct i915_request *request, + u32 mode) +{ + if (mode & EMIT_FLUSH) { + u32 flags = 0; + u32 *cs; + + flags |= PIPE_CONTROL_TILE_CACHE_FLUSH; + flags |= PIPE_CONTROL_RENDER_TARGET_CACHE_FLUSH; + flags |= PIPE_CONTROL_DEPTH_CACHE_FLUSH; + flags |= PIPE_CONTROL_DC_FLUSH_ENABLE; + flags |= PIPE_CONTROL_FLUSH_ENABLE; + + flags |= PIPE_CONTROL_STORE_DATA_INDEX; + flags |= PIPE_CONTROL_QW_WRITE; + + flags |= PIPE_CONTROL_CS_STALL; + + cs = intel_ring_begin(request, 6); + if (IS_ERR(cs)) + return PTR_ERR(cs); + + cs = gen8_emit_pipe_control(cs, flags, LRC_PPHWSP_SCRATCH_ADDR); + intel_ring_advance(request, cs); + } + + if (mode & EMIT_INVALIDATE) { + u32 flags = 0; + u32 *cs; + + flags |= PIPE_CONTROL_COMMAND_CACHE_INVALIDATE; + flags |= PIPE_CONTROL_TLB_INVALIDATE; + flags |= PIPE_CONTROL_INSTRUCTION_CACHE_INVALIDATE; + flags |= PIPE_CONTROL_TEXTURE_CACHE_INVALIDATE; + flags |= PIPE_CONTROL_VF_CACHE_INVALIDATE; + flags |= PIPE_CONTROL_CONST_CACHE_INVALIDATE; + flags |= PIPE_CONTROL_STATE_CACHE_INVALIDATE; + + flags |= PIPE_CONTROL_STORE_DATA_INDEX; + flags |= PIPE_CONTROL_QW_WRITE; + + flags |= PIPE_CONTROL_CS_STALL; + + cs = intel_ring_begin(request, 8); + if (IS_ERR(cs)) + return PTR_ERR(cs); + + /* + * Prevent the pre-parser from skipping past the TLB + * invalidate and loading a stale page for the batch + * buffer / request payload. + */ + *cs++ = preparser_disable(true); + + cs = gen8_emit_pipe_control(cs, flags, LRC_PPHWSP_SCRATCH_ADDR); + + *cs++ = preparser_disable(false); intel_ring_advance(request, cs); } @@ -2907,7 +3306,7 @@ static u32 *gen8_emit_fini_breadcrumb(struct i915_request *request, u32 *cs) { cs = gen8_emit_ggtt_write(cs, request->fence.seqno, - request->timeline->hwsp_offset, + i915_request_active_timeline(request)->hwsp_offset, 0); return gen8_emit_fini_breadcrumb_footer(request, cs); @@ -2915,28 +3314,103 @@ static u32 *gen8_emit_fini_breadcrumb(struct i915_request *request, u32 *cs) static u32 *gen8_emit_fini_breadcrumb_rcs(struct i915_request *request, u32 *cs) { + cs = gen8_emit_pipe_control(cs, + PIPE_CONTROL_RENDER_TARGET_CACHE_FLUSH | + PIPE_CONTROL_DEPTH_CACHE_FLUSH | + PIPE_CONTROL_DC_FLUSH_ENABLE, + 0); + + /* XXX flush+write+CS_STALL all in one upsets gem_concurrent_blt:kbl */ cs = gen8_emit_ggtt_write_rcs(cs, request->fence.seqno, - request->timeline->hwsp_offset, + i915_request_active_timeline(request)->hwsp_offset, + PIPE_CONTROL_FLUSH_ENABLE | + PIPE_CONTROL_CS_STALL); + + return gen8_emit_fini_breadcrumb_footer(request, cs); +} + +static u32 * +gen11_emit_fini_breadcrumb_rcs(struct i915_request *request, u32 *cs) +{ + cs = gen8_emit_ggtt_write_rcs(cs, + request->fence.seqno, + i915_request_active_timeline(request)->hwsp_offset, + PIPE_CONTROL_CS_STALL | + PIPE_CONTROL_TILE_CACHE_FLUSH | PIPE_CONTROL_RENDER_TARGET_CACHE_FLUSH | PIPE_CONTROL_DEPTH_CACHE_FLUSH | - PIPE_CONTROL_DC_FLUSH_ENABLE); - - /* XXX flush+write+CS_STALL all in one upsets gem_concurrent_blt:kbl */ - cs = gen8_emit_pipe_control(cs, - PIPE_CONTROL_FLUSH_ENABLE | - PIPE_CONTROL_CS_STALL, - 0); + PIPE_CONTROL_DC_FLUSH_ENABLE | + PIPE_CONTROL_FLUSH_ENABLE); return gen8_emit_fini_breadcrumb_footer(request, cs); } -static u32 *gen11_emit_fini_breadcrumb_rcs(struct i915_request *request, - u32 *cs) +/* + * Note that the CS instruction pre-parser will not stall on the breadcrumb + * flush and will continue pre-fetching the instructions after it before the + * memory sync is completed. On pre-gen12 HW, the pre-parser will stop at + * BB_START/END instructions, so, even though we might pre-fetch the pre-amble + * of the next request before the memory has been flushed, we're guaranteed that + * we won't access the batch itself too early. + * However, on gen12+ the parser can pre-fetch across the BB_START/END commands, + * so, if the current request is modifying an instruction in the next request on + * the same intel_context, we might pre-fetch and then execute the pre-update + * instruction. To avoid this, the users of self-modifying code should either + * disable the parser around the code emitting the memory writes, via a new flag + * added to MI_ARB_CHECK, or emit the writes from a different intel_context. For + * the in-kernel use-cases we've opted to use a separate context, see + * reloc_gpu() as an example. + * All the above applies only to the instructions themselves. Non-inline data + * used by the instructions is not pre-fetched. + */ + +static u32 *gen12_emit_preempt_busywait(struct i915_request *request, u32 *cs) +{ + *cs++ = MI_SEMAPHORE_WAIT_TOKEN | + MI_SEMAPHORE_GLOBAL_GTT | + MI_SEMAPHORE_POLL | + MI_SEMAPHORE_SAD_EQ_SDD; + *cs++ = 0; + *cs++ = intel_hws_preempt_address(request->engine); + *cs++ = 0; + *cs++ = 0; + *cs++ = MI_NOOP; + + return cs; +} + +static __always_inline u32* +gen12_emit_fini_breadcrumb_footer(struct i915_request *request, u32 *cs) +{ + *cs++ = MI_USER_INTERRUPT; + + *cs++ = MI_ARB_ON_OFF | MI_ARB_ENABLE; + if (intel_engine_has_semaphores(request->engine)) + cs = gen12_emit_preempt_busywait(request, cs); + + request->tail = intel_ring_offset(request, cs); + assert_ring_tail_valid(request->ring, request->tail); + + return gen8_emit_wa_tail(request, cs); +} + +static u32 *gen12_emit_fini_breadcrumb(struct i915_request *request, u32 *cs) +{ + cs = gen8_emit_ggtt_write(cs, + request->fence.seqno, + i915_request_active_timeline(request)->hwsp_offset, + 0); + + return gen12_emit_fini_breadcrumb_footer(request, cs); +} + +static u32 * +gen12_emit_fini_breadcrumb_rcs(struct i915_request *request, u32 *cs) { cs = gen8_emit_ggtt_write_rcs(cs, request->fence.seqno, - request->timeline->hwsp_offset, + i915_request_active_timeline(request)->hwsp_offset, PIPE_CONTROL_CS_STALL | PIPE_CONTROL_TILE_CACHE_FLUSH | PIPE_CONTROL_RENDER_TARGET_CACHE_FLUSH | @@ -2944,7 +3418,7 @@ static u32 *gen11_emit_fini_breadcrumb_rcs(struct i915_request *request, PIPE_CONTROL_DC_FLUSH_ENABLE | PIPE_CONTROL_FLUSH_ENABLE); - return gen8_emit_fini_breadcrumb_footer(request, cs); + return gen12_emit_fini_breadcrumb_footer(request, cs); } static void execlists_park(struct intel_engine_cs *engine) @@ -2972,6 +3446,9 @@ void intel_execlists_set_default_submission(struct intel_engine_cs *engine) if (HAS_LOGICAL_RING_PREEMPTION(engine->i915)) engine->flags |= I915_ENGINE_HAS_PREEMPTION; } + + if (engine->class != COPY_ENGINE_CLASS && INTEL_GEN(engine->i915) >= 12) + engine->flags |= I915_ENGINE_HAS_RELATIVE_MMIO; } static void execlists_destroy(struct intel_engine_cs *engine) @@ -2999,6 +3476,8 @@ logical_ring_default_vfuncs(struct intel_engine_cs *engine) engine->emit_flush = gen8_emit_flush; engine->emit_init_breadcrumb = gen8_emit_init_breadcrumb; engine->emit_fini_breadcrumb = gen8_emit_fini_breadcrumb; + if (INTEL_GEN(engine->i915) >= 12) + engine->emit_fini_breadcrumb = gen12_emit_fini_breadcrumb; engine->set_default_submission = intel_execlists_set_default_submission; @@ -3044,6 +3523,9 @@ static void rcs_submission_override(struct intel_engine_cs *engine) { switch (INTEL_GEN(engine->i915)) { case 12: + engine->emit_flush = gen12_emit_flush_render; + engine->emit_fini_breadcrumb = gen12_emit_fini_breadcrumb_rcs; + break; case 11: engine->emit_flush = gen11_emit_flush_render; engine->emit_fini_breadcrumb = gen11_emit_fini_breadcrumb_rcs; @@ -3116,7 +3598,7 @@ int intel_execlists_submission_init(struct intel_engine_cs *engine) return 0; } -static u32 intel_lr_indirect_ctx_offset(struct intel_engine_cs *engine) +static u32 intel_lr_indirect_ctx_offset(const struct intel_engine_cs *engine) { u32 indirect_ctx_offset; @@ -3149,86 +3631,50 @@ static u32 intel_lr_indirect_ctx_offset(struct intel_engine_cs *engine) return indirect_ctx_offset; } -static void execlists_init_reg_state(u32 *regs, - struct intel_context *ce, - struct intel_engine_cs *engine, - struct intel_ring *ring) -{ - struct i915_ppgtt *ppgtt = i915_vm_to_ppgtt(ce->vm); - bool rcs = engine->class == RENDER_CLASS; - u32 base = engine->mmio_base; - - /* - * A context is actually a big batch buffer with several - * MI_LOAD_REGISTER_IMM commands followed by (reg, value) pairs. The - * values we are setting here are only for the first context restore: - * on a subsequent save, the GPU will recreate this batchbuffer with new - * values (including all the missing MI_LOAD_REGISTER_IMM commands that - * we are not initializing here). - * - * Must keep consistent with virtual_update_register_offsets(). - */ - regs[CTX_LRI_HEADER_0] = MI_LOAD_REGISTER_IMM(rcs ? 14 : 11) | - MI_LRI_FORCE_POSTED; - CTX_REG(regs, CTX_CONTEXT_CONTROL, RING_CONTEXT_CONTROL(base), +static void init_common_reg_state(u32 * const regs, + const struct intel_engine_cs *engine, + const struct intel_ring *ring) +{ + regs[CTX_CONTEXT_CONTROL] = _MASKED_BIT_DISABLE(CTX_CTRL_ENGINE_CTX_RESTORE_INHIBIT) | - _MASKED_BIT_ENABLE(CTX_CTRL_INHIBIT_SYN_CTX_SWITCH)); - if (INTEL_GEN(engine->i915) < 11) { - regs[CTX_CONTEXT_CONTROL + 1] |= + _MASKED_BIT_ENABLE(CTX_CTRL_INHIBIT_SYN_CTX_SWITCH); + if (INTEL_GEN(engine->i915) < 11) + regs[CTX_CONTEXT_CONTROL] |= _MASKED_BIT_DISABLE(CTX_CTRL_ENGINE_CTX_SAVE_INHIBIT | CTX_CTRL_RS_CTX_ENABLE); - } - CTX_REG(regs, CTX_RING_HEAD, RING_HEAD(base), 0); - CTX_REG(regs, CTX_RING_TAIL, RING_TAIL(base), 0); - CTX_REG(regs, CTX_RING_BUFFER_START, RING_START(base), 0); - CTX_REG(regs, CTX_RING_BUFFER_CONTROL, RING_CTL(base), - RING_CTL_SIZE(ring->size) | RING_VALID); - CTX_REG(regs, CTX_BB_HEAD_U, RING_BBADDR_UDW(base), 0); - CTX_REG(regs, CTX_BB_HEAD_L, RING_BBADDR(base), 0); - CTX_REG(regs, CTX_BB_STATE, RING_BBSTATE(base), RING_BB_PPGTT); - CTX_REG(regs, CTX_SECOND_BB_HEAD_U, RING_SBBADDR_UDW(base), 0); - CTX_REG(regs, CTX_SECOND_BB_HEAD_L, RING_SBBADDR(base), 0); - CTX_REG(regs, CTX_SECOND_BB_STATE, RING_SBBSTATE(base), 0); - if (rcs) { - struct i915_ctx_workarounds *wa_ctx = &engine->wa_ctx; - - CTX_REG(regs, CTX_RCS_INDIRECT_CTX, RING_INDIRECT_CTX(base), 0); - CTX_REG(regs, CTX_RCS_INDIRECT_CTX_OFFSET, - RING_INDIRECT_CTX_OFFSET(base), 0); - if (wa_ctx->indirect_ctx.size) { - u32 ggtt_offset = i915_ggtt_offset(wa_ctx->vma); - - regs[CTX_RCS_INDIRECT_CTX + 1] = - (ggtt_offset + wa_ctx->indirect_ctx.offset) | - (wa_ctx->indirect_ctx.size / CACHELINE_BYTES); - - regs[CTX_RCS_INDIRECT_CTX_OFFSET + 1] = - intel_lr_indirect_ctx_offset(engine) << 6; - } - CTX_REG(regs, CTX_BB_PER_CTX_PTR, RING_BB_PER_CTX_PTR(base), 0); - if (wa_ctx->per_ctx.size) { - u32 ggtt_offset = i915_ggtt_offset(wa_ctx->vma); + regs[CTX_RING_BUFFER_CONTROL] = RING_CTL_SIZE(ring->size) | RING_VALID; + regs[CTX_BB_STATE] = RING_BB_PPGTT; +} - regs[CTX_BB_PER_CTX_PTR + 1] = - (ggtt_offset + wa_ctx->per_ctx.offset) | 0x01; - } +static void init_wa_bb_reg_state(u32 * const regs, + const struct intel_engine_cs *engine, + u32 pos_bb_per_ctx) +{ + const struct i915_ctx_workarounds * const wa_ctx = &engine->wa_ctx; + + if (wa_ctx->per_ctx.size) { + const u32 ggtt_offset = i915_ggtt_offset(wa_ctx->vma); + + regs[pos_bb_per_ctx] = + (ggtt_offset + wa_ctx->per_ctx.offset) | 0x01; } - regs[CTX_LRI_HEADER_1] = MI_LOAD_REGISTER_IMM(9) | MI_LRI_FORCE_POSTED; + if (wa_ctx->indirect_ctx.size) { + const u32 ggtt_offset = i915_ggtt_offset(wa_ctx->vma); - CTX_REG(regs, CTX_CTX_TIMESTAMP, RING_CTX_TIMESTAMP(base), 0); - /* PDP values well be assigned later if needed */ - CTX_REG(regs, CTX_PDP3_UDW, GEN8_RING_PDP_UDW(base, 3), 0); - CTX_REG(regs, CTX_PDP3_LDW, GEN8_RING_PDP_LDW(base, 3), 0); - CTX_REG(regs, CTX_PDP2_UDW, GEN8_RING_PDP_UDW(base, 2), 0); - CTX_REG(regs, CTX_PDP2_LDW, GEN8_RING_PDP_LDW(base, 2), 0); - CTX_REG(regs, CTX_PDP1_UDW, GEN8_RING_PDP_UDW(base, 1), 0); - CTX_REG(regs, CTX_PDP1_LDW, GEN8_RING_PDP_LDW(base, 1), 0); - CTX_REG(regs, CTX_PDP0_UDW, GEN8_RING_PDP_UDW(base, 0), 0); - CTX_REG(regs, CTX_PDP0_LDW, GEN8_RING_PDP_LDW(base, 0), 0); + regs[pos_bb_per_ctx + 2] = + (ggtt_offset + wa_ctx->indirect_ctx.offset) | + (wa_ctx->indirect_ctx.size / CACHELINE_BYTES); + regs[pos_bb_per_ctx + 4] = + intel_lr_indirect_ctx_offset(engine) << 6; + } +} + +static void init_ppgtt_reg_state(u32 *regs, const struct i915_ppgtt *ppgtt) +{ if (i915_vm_is_4lvl(&ppgtt->vm)) { /* 64b PPGTT (48bit canonical) * PDP0_DESCRIPTOR contains the base address to PML4 and @@ -3241,15 +3687,47 @@ static void execlists_init_reg_state(u32 *regs, ASSIGN_CTX_PDP(ppgtt, regs, 1); ASSIGN_CTX_PDP(ppgtt, regs, 0); } +} + +static struct i915_ppgtt *vm_alias(struct i915_address_space *vm) +{ + if (i915_is_ggtt(vm)) + return i915_vm_to_ggtt(vm)->alias; + else + return i915_vm_to_ppgtt(vm); +} - if (rcs) { - regs[CTX_LRI_HEADER_2] = MI_LOAD_REGISTER_IMM(1); - CTX_REG(regs, CTX_R_PWR_CLK_STATE, GEN8_R_PWR_CLK_STATE, 0); +static void execlists_init_reg_state(u32 *regs, + const struct intel_context *ce, + const struct intel_engine_cs *engine, + const struct intel_ring *ring, + bool close) +{ + /* + * A context is actually a big batch buffer with several + * MI_LOAD_REGISTER_IMM commands followed by (reg, value) pairs. The + * values we are setting here are only for the first context restore: + * on a subsequent save, the GPU will recreate this batchbuffer with new + * values (including all the missing MI_LOAD_REGISTER_IMM commands that + * we are not initializing here). + * + * Must keep consistent with virtual_update_register_offsets(). + */ + u32 *bbe = set_offsets(regs, reg_offsets(engine), engine); + + if (close) { /* Close the batch; used mainly by live_lrc_layout() */ + *bbe = MI_BATCH_BUFFER_END; + if (INTEL_GEN(engine->i915) >= 10) + *bbe |= BIT(0); } - regs[CTX_END] = MI_BATCH_BUFFER_END; - if (INTEL_GEN(engine->i915) >= 10) - regs[CTX_END] |= BIT(0); + init_common_reg_state(regs, engine, ring); + init_ppgtt_reg_state(regs, vm_alias(ce->vm)); + + init_wa_bb_reg_state(regs, engine, + INTEL_GEN(engine->i915) >= 12 ? + GEN12_CTX_BB_PER_CTX_PTR : + CTX_BB_PER_CTX_PTR); } static int @@ -3258,6 +3736,7 @@ populate_lr_context(struct intel_context *ce, struct intel_engine_cs *engine, struct intel_ring *ring) { + bool inhibit = true; void *vaddr; u32 *regs; int ret; @@ -3289,14 +3768,15 @@ populate_lr_context(struct intel_context *ce, memcpy(vaddr + start, defaults + start, engine->context_size); i915_gem_object_unpin_map(engine->default_state); + inhibit = false; } /* The second page of the context object contains some fields which must * be set up prior to the first execution. */ regs = vaddr + LRC_STATE_PN * PAGE_SIZE; - execlists_init_reg_state(regs, ce, engine, ring); - if (!engine->default_state) - regs[CTX_CONTEXT_CONTROL + 1] |= + execlists_init_reg_state(regs, ce, engine, ring, inhibit); + if (inhibit) + regs[CTX_CONTEXT_CONTROL] |= _MASKED_BIT_ENABLE(CTX_CTRL_ENGINE_CTX_RESTORE_INHIBIT); ret = 0; @@ -3436,8 +3916,9 @@ static void virtual_engine_initial_hint(struct virtual_engine *ve) return; swap(ve->siblings[swp], ve->siblings[0]); - virtual_update_register_offsets(ve->context.lrc_reg_state, - ve->siblings[0]); + if (!intel_engine_has_relative_mmio(ve->siblings[0])) + virtual_update_register_offsets(ve->context.lrc_reg_state, + ve->siblings[0]); } static int virtual_context_pin(struct intel_context *ce) @@ -3594,6 +4075,8 @@ submit_engine: static void virtual_submit_request(struct i915_request *rq) { struct virtual_engine *ve = to_virtual_engine(rq->engine); + struct i915_request *old; + unsigned long flags; GEM_TRACE("%s: rq=%llx:%lld\n", ve->base.name, @@ -3602,15 +4085,31 @@ static void virtual_submit_request(struct i915_request *rq) GEM_BUG_ON(ve->base.submit_request != virtual_submit_request); - GEM_BUG_ON(ve->request); - GEM_BUG_ON(!list_empty(virtual_queue(ve))); + spin_lock_irqsave(&ve->base.active.lock, flags); - ve->base.execlists.queue_priority_hint = rq_prio(rq); - WRITE_ONCE(ve->request, rq); + old = ve->request; + if (old) { /* background completion event from preempt-to-busy */ + GEM_BUG_ON(!i915_request_completed(old)); + __i915_request_submit(old); + i915_request_put(old); + } + + if (i915_request_completed(rq)) { + __i915_request_submit(rq); - list_move_tail(&rq->sched.link, virtual_queue(ve)); + ve->base.execlists.queue_priority_hint = INT_MIN; + ve->request = NULL; + } else { + ve->base.execlists.queue_priority_hint = rq_prio(rq); + ve->request = i915_request_get(rq); - tasklet_schedule(&ve->base.execlists.tasklet); + GEM_BUG_ON(!list_empty(virtual_queue(ve))); + list_move_tail(&rq->sched.link, virtual_queue(ve)); + + tasklet_schedule(&ve->base.execlists.tasklet); + } + + spin_unlock_irqrestore(&ve->base.active.lock, flags); } static struct ve_bond * @@ -3631,18 +4130,22 @@ static void virtual_bond_execute(struct i915_request *rq, struct dma_fence *signal) { struct virtual_engine *ve = to_virtual_engine(rq->engine); + intel_engine_mask_t allowed, exec; struct ve_bond *bond; + allowed = ~to_request(signal)->engine->mask; + bond = virtual_find_bond(ve, to_request(signal)->engine); - if (bond) { - intel_engine_mask_t old, new, cmp; + if (bond) + allowed &= bond->sibling_mask; - cmp = READ_ONCE(rq->execution_mask); - do { - old = cmp; - new = cmp & bond->sibling_mask; - } while ((cmp = cmpxchg(&rq->execution_mask, old, new)) != old); - } + /* Restrict the bonded request to run on only the available engines */ + exec = READ_ONCE(rq->execution_mask); + while (!try_cmpxchg(&rq->execution_mask, &exec, exec & allowed)) + ; + + /* Prevent the master from being re-run on the bonded engines */ + to_request(signal)->execution_mask &= ~allowed; } struct intel_context * @@ -3689,6 +4192,7 @@ intel_execlists_create_virtual(struct i915_gem_context *ctx, snprintf(ve->base.name, sizeof(ve->base.name), "virtual"); intel_engine_init_active(&ve->base, ENGINE_VIRTUAL); + intel_engine_init_breadcrumbs(&ve->base); intel_engine_init_execlists(&ve->base); @@ -3851,6 +4355,18 @@ int intel_virtual_engine_attach_bond(struct intel_engine_cs *engine, return 0; } +struct intel_engine_cs * +intel_virtual_engine_get_sibling(struct intel_engine_cs *engine, + unsigned int sibling) +{ + struct virtual_engine *ve = to_virtual_engine(engine); + + if (sibling >= ve->num_siblings) + return NULL; + + return ve->siblings[sibling]; +} + void intel_execlists_show_requests(struct intel_engine_cs *engine, struct drm_printer *m, void (*show_request)(struct drm_printer *m, @@ -3939,6 +4455,9 @@ void intel_lr_context_reset(struct intel_engine_cs *engine, u32 head, bool scrub) { + GEM_BUG_ON(!intel_context_is_pinned(ce)); + __context_pin_acquire(ce); + /* * We want a simple context + ring to execute the breadcrumb update. * We cannot rely on the context being intact across the GPU hang, @@ -3955,7 +4474,7 @@ void intel_lr_context_reset(struct intel_engine_cs *engine, engine->pinned_default_state + LRC_STATE_PN * PAGE_SIZE, engine->context_size - PAGE_SIZE); } - execlists_init_reg_state(regs, ce, engine, ce->ring); + execlists_init_reg_state(regs, ce, engine, ce->ring, false); } /* Rerun the request; its payload has been neutered (if guilty). */ @@ -3963,6 +4482,7 @@ void intel_lr_context_reset(struct intel_engine_cs *engine, intel_ring_update_space(ce->ring); __execlists_update_reg_state(ce, engine); + __context_pin_release(ce); } #if IS_ENABLED(CONFIG_DRM_I915_SELFTEST) diff --git a/drivers/gpu/drm/i915/gt/intel_lrc.h b/drivers/gpu/drm/i915/gt/intel_lrc.h index c2bba82bcc16..99dc576a4e25 100644 --- a/drivers/gpu/drm/i915/gt/intel_lrc.h +++ b/drivers/gpu/drm/i915/gt/intel_lrc.h @@ -66,6 +66,12 @@ struct intel_engine_cs; #define GEN11_CSB_READ_PTR_MASK (GEN11_CSB_PTR_MASK << 8) #define GEN11_CSB_WRITE_PTR_MASK (GEN11_CSB_PTR_MASK << 0) +#define MAX_CONTEXT_HW_ID (1<<21) /* exclusive */ +#define MAX_GUC_CONTEXT_HW_ID (1 << 20) /* exclusive */ +#define GEN11_MAX_CONTEXT_HW_ID (1<<11) /* exclusive */ +/* in Gen12 ID 0x7FF is reserved to indicate idle */ +#define GEN12_MAX_CONTEXT_HW_ID (GEN11_MAX_CONTEXT_HW_ID - 1) + enum { INTEL_CONTEXT_SCHEDULE_IN = 0, INTEL_CONTEXT_SCHEDULE_OUT, @@ -104,6 +110,10 @@ int intel_execlists_submission_init(struct intel_engine_cs *engine); */ #define LRC_HEADER_PAGES LRC_PPHWSP_PN +/* Space within PPHWSP reserved to be used as scratch */ +#define LRC_PPHWSP_SCRATCH 0x34 +#define LRC_PPHWSP_SCRATCH_ADDR (LRC_PPHWSP_SCRATCH * sizeof(u32)) + void intel_execlists_set_default_submission(struct intel_engine_cs *engine); void intel_lr_context_reset(struct intel_engine_cs *engine, @@ -131,4 +141,8 @@ int intel_virtual_engine_attach_bond(struct intel_engine_cs *engine, const struct intel_engine_cs *master, const struct intel_engine_cs *sibling); +struct intel_engine_cs * +intel_virtual_engine_get_sibling(struct intel_engine_cs *engine, + unsigned int sibling); + #endif /* _INTEL_LRC_H_ */ diff --git a/drivers/gpu/drm/i915/gt/intel_lrc_reg.h b/drivers/gpu/drm/i915/gt/intel_lrc_reg.h index b8f20ad71169..06ab0276e10e 100644 --- a/drivers/gpu/drm/i915/gt/intel_lrc_reg.h +++ b/drivers/gpu/drm/i915/gt/intel_lrc_reg.h @@ -9,55 +9,41 @@ #include <linux/types.h> -/* GEN8+ Reg State Context */ -#define CTX_LRI_HEADER_0 0x01 -#define CTX_CONTEXT_CONTROL 0x02 -#define CTX_RING_HEAD 0x04 -#define CTX_RING_TAIL 0x06 -#define CTX_RING_BUFFER_START 0x08 -#define CTX_RING_BUFFER_CONTROL 0x0a -#define CTX_BB_HEAD_U 0x0c -#define CTX_BB_HEAD_L 0x0e -#define CTX_BB_STATE 0x10 -#define CTX_SECOND_BB_HEAD_U 0x12 -#define CTX_SECOND_BB_HEAD_L 0x14 -#define CTX_SECOND_BB_STATE 0x16 -#define CTX_BB_PER_CTX_PTR 0x18 -#define CTX_RCS_INDIRECT_CTX 0x1a -#define CTX_RCS_INDIRECT_CTX_OFFSET 0x1c -#define CTX_LRI_HEADER_1 0x21 -#define CTX_CTX_TIMESTAMP 0x22 -#define CTX_PDP3_UDW 0x24 -#define CTX_PDP3_LDW 0x26 -#define CTX_PDP2_UDW 0x28 -#define CTX_PDP2_LDW 0x2a -#define CTX_PDP1_UDW 0x2c -#define CTX_PDP1_LDW 0x2e -#define CTX_PDP0_UDW 0x30 -#define CTX_PDP0_LDW 0x32 -#define CTX_LRI_HEADER_2 0x41 -#define CTX_R_PWR_CLK_STATE 0x42 -#define CTX_END 0x44 - -#define CTX_REG(reg_state, pos, reg, val) do { \ - u32 *reg_state__ = (reg_state); \ - const u32 pos__ = (pos); \ - (reg_state__)[(pos__) + 0] = i915_mmio_reg_offset(reg); \ - (reg_state__)[(pos__) + 1] = (val); \ -} while (0) +/* GEN8 to GEN11 Reg State Context */ +#define CTX_CONTEXT_CONTROL (0x02 + 1) +#define CTX_RING_HEAD (0x04 + 1) +#define CTX_RING_TAIL (0x06 + 1) +#define CTX_RING_BUFFER_START (0x08 + 1) +#define CTX_RING_BUFFER_CONTROL (0x0a + 1) +#define CTX_BB_STATE (0x10 + 1) +#define CTX_BB_PER_CTX_PTR (0x18 + 1) +#define CTX_PDP3_UDW (0x24 + 1) +#define CTX_PDP3_LDW (0x26 + 1) +#define CTX_PDP2_UDW (0x28 + 1) +#define CTX_PDP2_LDW (0x2a + 1) +#define CTX_PDP1_UDW (0x2c + 1) +#define CTX_PDP1_LDW (0x2e + 1) +#define CTX_PDP0_UDW (0x30 + 1) +#define CTX_PDP0_LDW (0x32 + 1) +#define CTX_R_PWR_CLK_STATE (0x42 + 1) + +#define GEN9_CTX_RING_MI_MODE 0x54 + +/* GEN12+ Reg State Context */ +#define GEN12_CTX_BB_PER_CTX_PTR (0x12 + 1) #define ASSIGN_CTX_PDP(ppgtt, reg_state, n) do { \ u32 *reg_state__ = (reg_state); \ const u64 addr__ = i915_page_dir_dma_addr((ppgtt), (n)); \ - (reg_state__)[CTX_PDP ## n ## _UDW + 1] = upper_32_bits(addr__); \ - (reg_state__)[CTX_PDP ## n ## _LDW + 1] = lower_32_bits(addr__); \ + (reg_state__)[CTX_PDP ## n ## _UDW] = upper_32_bits(addr__); \ + (reg_state__)[CTX_PDP ## n ## _LDW] = lower_32_bits(addr__); \ } while (0) #define ASSIGN_CTX_PML4(ppgtt, reg_state) do { \ u32 *reg_state__ = (reg_state); \ const u64 addr__ = px_dma(ppgtt->pd); \ - (reg_state__)[CTX_PDP0_UDW + 1] = upper_32_bits(addr__); \ - (reg_state__)[CTX_PDP0_LDW + 1] = lower_32_bits(addr__); \ + (reg_state__)[CTX_PDP0_UDW] = upper_32_bits(addr__); \ + (reg_state__)[CTX_PDP0_LDW] = lower_32_bits(addr__); \ } while (0) #define GEN8_CTX_RCS_INDIRECT_CTX_OFFSET_DEFAULT 0x17 diff --git a/drivers/gpu/drm/i915/gt/intel_rc6.c b/drivers/gpu/drm/i915/gt/intel_rc6.c new file mode 100644 index 000000000000..71184aa72896 --- /dev/null +++ b/drivers/gpu/drm/i915/gt/intel_rc6.c @@ -0,0 +1,712 @@ +/* + * SPDX-License-Identifier: MIT + * + * Copyright © 2019 Intel Corporation + */ + +#include <linux/pm_runtime.h> + +#include "i915_drv.h" +#include "intel_gt.h" +#include "intel_gt_pm.h" +#include "intel_rc6.h" +#include "intel_sideband.h" + +/** + * DOC: RC6 + * + * RC6 is a special power stage which allows the GPU to enter an very + * low-voltage mode when idle, using down to 0V while at this stage. This + * stage is entered automatically when the GPU is idle when RC6 support is + * enabled, and as soon as new workload arises GPU wakes up automatically as + * well. + * + * There are different RC6 modes available in Intel GPU, which differentiate + * among each other with the latency required to enter and leave RC6 and + * voltage consumed by the GPU in different states. + * + * The combination of the following flags define which states GPU is allowed + * to enter, while RC6 is the normal RC6 state, RC6p is the deep RC6, and + * RC6pp is deepest RC6. Their support by hardware varies according to the + * GPU, BIOS, chipset and platform. RC6 is usually the safest one and the one + * which brings the most power savings; deeper states save more power, but + * require higher latency to switch to and wake up. + */ + +static struct intel_gt *rc6_to_gt(struct intel_rc6 *rc6) +{ + return container_of(rc6, struct intel_gt, rc6); +} + +static struct intel_uncore *rc6_to_uncore(struct intel_rc6 *rc) +{ + return rc6_to_gt(rc)->uncore; +} + +static struct drm_i915_private *rc6_to_i915(struct intel_rc6 *rc) +{ + return rc6_to_gt(rc)->i915; +} + +static inline void set(struct intel_uncore *uncore, i915_reg_t reg, u32 val) +{ + intel_uncore_write_fw(uncore, reg, val); +} + +static void gen11_rc6_enable(struct intel_rc6 *rc6) +{ + struct intel_uncore *uncore = rc6_to_uncore(rc6); + struct intel_engine_cs *engine; + enum intel_engine_id id; + + /* 2b: Program RC6 thresholds.*/ + set(uncore, GEN6_RC6_WAKE_RATE_LIMIT, 54 << 16 | 85); + set(uncore, GEN10_MEDIA_WAKE_RATE_LIMIT, 150); + + set(uncore, GEN6_RC_EVALUATION_INTERVAL, 125000); /* 12500 * 1280ns */ + set(uncore, GEN6_RC_IDLE_HYSTERSIS, 25); /* 25 * 1280ns */ + for_each_engine(engine, rc6_to_gt(rc6)->i915, id) + set(uncore, RING_MAX_IDLE(engine->mmio_base), 10); + + set(uncore, GUC_MAX_IDLE_COUNT, 0xA); + + set(uncore, GEN6_RC_SLEEP, 0); + + set(uncore, GEN6_RC6_THRESHOLD, 50000); /* 50/125ms per EI */ + + /* + * 2c: Program Coarse Power Gating Policies. + * + * Bspec's guidance is to use 25us (really 25 * 1280ns) here. What we + * use instead is a more conservative estimate for the maximum time + * it takes us to service a CS interrupt and submit a new ELSP - that + * is the time which the GPU is idle waiting for the CPU to select the + * next request to execute. If the idle hysteresis is less than that + * interrupt service latency, the hardware will automatically gate + * the power well and we will then incur the wake up cost on top of + * the service latency. A similar guide from plane_state is that we + * do not want the enable hysteresis to less than the wakeup latency. + * + * igt/gem_exec_nop/sequential provides a rough estimate for the + * service latency, and puts it around 10us for Broadwell (and other + * big core) and around 40us for Broxton (and other low power cores). + * [Note that for legacy ringbuffer submission, this is less than 1us!] + * However, the wakeup latency on Broxton is closer to 100us. To be + * conservative, we have to factor in a context switch on top (due + * to ksoftirqd). + */ + set(uncore, GEN9_MEDIA_PG_IDLE_HYSTERESIS, 250); + set(uncore, GEN9_RENDER_PG_IDLE_HYSTERESIS, 250); + + /* 3a: Enable RC6 */ + set(uncore, GEN6_RC_CONTROL, + GEN6_RC_CTL_HW_ENABLE | + GEN6_RC_CTL_RC6_ENABLE | + GEN6_RC_CTL_EI_MODE(1)); + + set(uncore, GEN9_PG_ENABLE, + GEN9_RENDER_PG_ENABLE | + GEN9_MEDIA_PG_ENABLE | + GEN11_MEDIA_SAMPLER_PG_ENABLE); +} + +static void gen9_rc6_enable(struct intel_rc6 *rc6) +{ + struct intel_uncore *uncore = rc6_to_uncore(rc6); + struct intel_engine_cs *engine; + enum intel_engine_id id; + u32 rc6_mode; + + /* 2b: Program RC6 thresholds.*/ + if (INTEL_GEN(rc6_to_i915(rc6)) >= 10) { + set(uncore, GEN6_RC6_WAKE_RATE_LIMIT, 54 << 16 | 85); + set(uncore, GEN10_MEDIA_WAKE_RATE_LIMIT, 150); + } else if (IS_SKYLAKE(rc6_to_i915(rc6))) { + /* + * WaRsDoubleRc6WrlWithCoarsePowerGating:skl Doubling WRL only + * when CPG is enabled + */ + set(uncore, GEN6_RC6_WAKE_RATE_LIMIT, 108 << 16); + } else { + set(uncore, GEN6_RC6_WAKE_RATE_LIMIT, 54 << 16); + } + + set(uncore, GEN6_RC_EVALUATION_INTERVAL, 125000); /* 12500 * 1280ns */ + set(uncore, GEN6_RC_IDLE_HYSTERSIS, 25); /* 25 * 1280ns */ + for_each_engine(engine, rc6_to_gt(rc6)->i915, id) + set(uncore, RING_MAX_IDLE(engine->mmio_base), 10); + + set(uncore, GUC_MAX_IDLE_COUNT, 0xA); + + set(uncore, GEN6_RC_SLEEP, 0); + + /* + * 2c: Program Coarse Power Gating Policies. + * + * Bspec's guidance is to use 25us (really 25 * 1280ns) here. What we + * use instead is a more conservative estimate for the maximum time + * it takes us to service a CS interrupt and submit a new ELSP - that + * is the time which the GPU is idle waiting for the CPU to select the + * next request to execute. If the idle hysteresis is less than that + * interrupt service latency, the hardware will automatically gate + * the power well and we will then incur the wake up cost on top of + * the service latency. A similar guide from plane_state is that we + * do not want the enable hysteresis to less than the wakeup latency. + * + * igt/gem_exec_nop/sequential provides a rough estimate for the + * service latency, and puts it around 10us for Broadwell (and other + * big core) and around 40us for Broxton (and other low power cores). + * [Note that for legacy ringbuffer submission, this is less than 1us!] + * However, the wakeup latency on Broxton is closer to 100us. To be + * conservative, we have to factor in a context switch on top (due + * to ksoftirqd). + */ + set(uncore, GEN9_MEDIA_PG_IDLE_HYSTERESIS, 250); + set(uncore, GEN9_RENDER_PG_IDLE_HYSTERESIS, 250); + + /* 3a: Enable RC6 */ + set(uncore, GEN6_RC6_THRESHOLD, 37500); /* 37.5/125ms per EI */ + + /* WaRsUseTimeoutMode:cnl (pre-prod) */ + if (IS_CNL_REVID(rc6_to_i915(rc6), CNL_REVID_A0, CNL_REVID_C0)) + rc6_mode = GEN7_RC_CTL_TO_MODE; + else + rc6_mode = GEN6_RC_CTL_EI_MODE(1); + + set(uncore, GEN6_RC_CONTROL, + GEN6_RC_CTL_HW_ENABLE | + GEN6_RC_CTL_RC6_ENABLE | + rc6_mode); + + set(uncore, GEN9_PG_ENABLE, + GEN9_RENDER_PG_ENABLE | GEN9_MEDIA_PG_ENABLE); +} + +static void gen8_rc6_enable(struct intel_rc6 *rc6) +{ + struct intel_uncore *uncore = rc6_to_uncore(rc6); + struct intel_engine_cs *engine; + enum intel_engine_id id; + + /* 2b: Program RC6 thresholds.*/ + set(uncore, GEN6_RC6_WAKE_RATE_LIMIT, 40 << 16); + set(uncore, GEN6_RC_EVALUATION_INTERVAL, 125000); /* 12500 * 1280ns */ + set(uncore, GEN6_RC_IDLE_HYSTERSIS, 25); /* 25 * 1280ns */ + for_each_engine(engine, rc6_to_gt(rc6)->i915, id) + set(uncore, RING_MAX_IDLE(engine->mmio_base), 10); + set(uncore, GEN6_RC_SLEEP, 0); + set(uncore, GEN6_RC6_THRESHOLD, 625); /* 800us/1.28 for TO */ + + /* 3: Enable RC6 */ + set(uncore, GEN6_RC_CONTROL, + GEN6_RC_CTL_HW_ENABLE | + GEN7_RC_CTL_TO_MODE | + GEN6_RC_CTL_RC6_ENABLE); +} + +static void gen6_rc6_enable(struct intel_rc6 *rc6) +{ + struct intel_uncore *uncore = rc6_to_uncore(rc6); + struct drm_i915_private *i915 = rc6_to_i915(rc6); + struct intel_engine_cs *engine; + enum intel_engine_id id; + u32 rc6vids, rc6_mask; + int ret; + + set(uncore, GEN6_RC1_WAKE_RATE_LIMIT, 1000 << 16); + set(uncore, GEN6_RC6_WAKE_RATE_LIMIT, 40 << 16 | 30); + set(uncore, GEN6_RC6pp_WAKE_RATE_LIMIT, 30); + set(uncore, GEN6_RC_EVALUATION_INTERVAL, 125000); + set(uncore, GEN6_RC_IDLE_HYSTERSIS, 25); + + for_each_engine(engine, i915, id) + set(uncore, RING_MAX_IDLE(engine->mmio_base), 10); + + set(uncore, GEN6_RC_SLEEP, 0); + set(uncore, GEN6_RC1e_THRESHOLD, 1000); + if (IS_IVYBRIDGE(i915)) + set(uncore, GEN6_RC6_THRESHOLD, 125000); + else + set(uncore, GEN6_RC6_THRESHOLD, 50000); + set(uncore, GEN6_RC6p_THRESHOLD, 150000); + set(uncore, GEN6_RC6pp_THRESHOLD, 64000); /* unused */ + + /* We don't use those on Haswell */ + rc6_mask = GEN6_RC_CTL_RC6_ENABLE; + if (HAS_RC6p(i915)) + rc6_mask |= GEN6_RC_CTL_RC6p_ENABLE; + if (HAS_RC6pp(i915)) + rc6_mask |= GEN6_RC_CTL_RC6pp_ENABLE; + set(uncore, GEN6_RC_CONTROL, + rc6_mask | + GEN6_RC_CTL_EI_MODE(1) | + GEN6_RC_CTL_HW_ENABLE); + + rc6vids = 0; + ret = sandybridge_pcode_read(i915, GEN6_PCODE_READ_RC6VIDS, + &rc6vids, NULL); + if (IS_GEN(i915, 6) && ret) { + DRM_DEBUG_DRIVER("Couldn't check for BIOS workaround\n"); + } else if (IS_GEN(i915, 6) && + (GEN6_DECODE_RC6_VID(rc6vids & 0xff) < 450)) { + DRM_DEBUG_DRIVER("You should update your BIOS. Correcting minimum rc6 voltage (%dmV->%dmV)\n", + GEN6_DECODE_RC6_VID(rc6vids & 0xff), 450); + rc6vids &= 0xffff00; + rc6vids |= GEN6_ENCODE_RC6_VID(450); + ret = sandybridge_pcode_write(i915, GEN6_PCODE_WRITE_RC6VIDS, rc6vids); + if (ret) + DRM_ERROR("Couldn't fix incorrect rc6 voltage\n"); + } +} + +/* Check that the pcbr address is not empty. */ +static int chv_rc6_init(struct intel_rc6 *rc6) +{ + struct intel_uncore *uncore = rc6_to_uncore(rc6); + resource_size_t pctx_paddr, paddr; + resource_size_t pctx_size = 32 * SZ_1K; + u32 pcbr; + + pcbr = intel_uncore_read(uncore, VLV_PCBR); + if ((pcbr >> VLV_PCBR_ADDR_SHIFT) == 0) { + DRM_DEBUG_DRIVER("BIOS didn't set up PCBR, fixing up\n"); + paddr = rc6_to_i915(rc6)->dsm.end + 1 - pctx_size; + GEM_BUG_ON(paddr > U32_MAX); + + pctx_paddr = (paddr & ~4095); + intel_uncore_write(uncore, VLV_PCBR, pctx_paddr); + } + + return 0; +} + +static int vlv_rc6_init(struct intel_rc6 *rc6) +{ + struct drm_i915_private *i915 = rc6_to_i915(rc6); + struct intel_uncore *uncore = rc6_to_uncore(rc6); + struct drm_i915_gem_object *pctx; + resource_size_t pctx_paddr; + resource_size_t pctx_size = 24 * SZ_1K; + u32 pcbr; + + pcbr = intel_uncore_read(uncore, VLV_PCBR); + if (pcbr) { + /* BIOS set it up already, grab the pre-alloc'd space */ + resource_size_t pcbr_offset; + + pcbr_offset = (pcbr & ~4095) - i915->dsm.start; + pctx = i915_gem_object_create_stolen_for_preallocated(i915, + pcbr_offset, + I915_GTT_OFFSET_NONE, + pctx_size); + if (IS_ERR(pctx)) + return PTR_ERR(pctx); + + goto out; + } + + DRM_DEBUG_DRIVER("BIOS didn't set up PCBR, fixing up\n"); + + /* + * From the Gunit register HAS: + * The Gfx driver is expected to program this register and ensure + * proper allocation within Gfx stolen memory. For example, this + * register should be programmed such than the PCBR range does not + * overlap with other ranges, such as the frame buffer, protected + * memory, or any other relevant ranges. + */ + pctx = i915_gem_object_create_stolen(i915, pctx_size); + if (IS_ERR(pctx)) { + DRM_DEBUG("not enough stolen space for PCTX, disabling\n"); + return PTR_ERR(pctx); + } + + GEM_BUG_ON(range_overflows_t(u64, + i915->dsm.start, + pctx->stolen->start, + U32_MAX)); + pctx_paddr = i915->dsm.start + pctx->stolen->start; + intel_uncore_write(uncore, VLV_PCBR, pctx_paddr); + +out: + rc6->pctx = pctx; + return 0; +} + +static void chv_rc6_enable(struct intel_rc6 *rc6) +{ + struct intel_uncore *uncore = rc6_to_uncore(rc6); + struct intel_engine_cs *engine; + enum intel_engine_id id; + + /* 2a: Program RC6 thresholds.*/ + set(uncore, GEN6_RC6_WAKE_RATE_LIMIT, 40 << 16); + set(uncore, GEN6_RC_EVALUATION_INTERVAL, 125000); /* 12500 * 1280ns */ + set(uncore, GEN6_RC_IDLE_HYSTERSIS, 25); /* 25 * 1280ns */ + + for_each_engine(engine, rc6_to_gt(rc6)->i915, id) + set(uncore, RING_MAX_IDLE(engine->mmio_base), 10); + set(uncore, GEN6_RC_SLEEP, 0); + + /* TO threshold set to 500 us (0x186 * 1.28 us) */ + set(uncore, GEN6_RC6_THRESHOLD, 0x186); + + /* Allows RC6 residency counter to work */ + set(uncore, VLV_COUNTER_CONTROL, + _MASKED_BIT_ENABLE(VLV_COUNT_RANGE_HIGH | + VLV_MEDIA_RC6_COUNT_EN | + VLV_RENDER_RC6_COUNT_EN)); + + /* 3: Enable RC6 */ + set(uncore, GEN6_RC_CONTROL, GEN7_RC_CTL_TO_MODE); +} + +static void vlv_rc6_enable(struct intel_rc6 *rc6) +{ + struct intel_uncore *uncore = rc6_to_uncore(rc6); + struct intel_engine_cs *engine; + enum intel_engine_id id; + + set(uncore, GEN6_RC6_WAKE_RATE_LIMIT, 0x00280000); + set(uncore, GEN6_RC_EVALUATION_INTERVAL, 125000); + set(uncore, GEN6_RC_IDLE_HYSTERSIS, 25); + + for_each_engine(engine, rc6_to_gt(rc6)->i915, id) + set(uncore, RING_MAX_IDLE(engine->mmio_base), 10); + + set(uncore, GEN6_RC6_THRESHOLD, 0x557); + + /* Allows RC6 residency counter to work */ + set(uncore, VLV_COUNTER_CONTROL, + _MASKED_BIT_ENABLE(VLV_COUNT_RANGE_HIGH | + VLV_MEDIA_RC0_COUNT_EN | + VLV_RENDER_RC0_COUNT_EN | + VLV_MEDIA_RC6_COUNT_EN | + VLV_RENDER_RC6_COUNT_EN)); + + set(uncore, GEN6_RC_CONTROL, + GEN7_RC_CTL_TO_MODE | VLV_RC_CTL_CTX_RST_PARALLEL); +} + +static bool bxt_check_bios_rc6_setup(struct intel_rc6 *rc6) +{ + struct intel_uncore *uncore = rc6_to_uncore(rc6); + struct drm_i915_private *i915 = rc6_to_i915(rc6); + u32 rc6_ctx_base, rc_ctl, rc_sw_target; + bool enable_rc6 = true; + + rc_ctl = intel_uncore_read(uncore, GEN6_RC_CONTROL); + rc_sw_target = intel_uncore_read(uncore, GEN6_RC_STATE); + rc_sw_target &= RC_SW_TARGET_STATE_MASK; + rc_sw_target >>= RC_SW_TARGET_STATE_SHIFT; + DRM_DEBUG_DRIVER("BIOS enabled RC states: " + "HW_CTRL %s HW_RC6 %s SW_TARGET_STATE %x\n", + onoff(rc_ctl & GEN6_RC_CTL_HW_ENABLE), + onoff(rc_ctl & GEN6_RC_CTL_RC6_ENABLE), + rc_sw_target); + + if (!(intel_uncore_read(uncore, RC6_LOCATION) & RC6_CTX_IN_DRAM)) { + DRM_DEBUG_DRIVER("RC6 Base location not set properly.\n"); + enable_rc6 = false; + } + + /* + * The exact context size is not known for BXT, so assume a page size + * for this check. + */ + rc6_ctx_base = + intel_uncore_read(uncore, RC6_CTX_BASE) & RC6_CTX_BASE_MASK; + if (!(rc6_ctx_base >= i915->dsm_reserved.start && + rc6_ctx_base + PAGE_SIZE < i915->dsm_reserved.end)) { + DRM_DEBUG_DRIVER("RC6 Base address not as expected.\n"); + enable_rc6 = false; + } + + if (!((intel_uncore_read(uncore, PWRCTX_MAXCNT_RCSUNIT) & IDLE_TIME_MASK) > 1 && + (intel_uncore_read(uncore, PWRCTX_MAXCNT_VCSUNIT0) & IDLE_TIME_MASK) > 1 && + (intel_uncore_read(uncore, PWRCTX_MAXCNT_BCSUNIT) & IDLE_TIME_MASK) > 1 && + (intel_uncore_read(uncore, PWRCTX_MAXCNT_VECSUNIT) & IDLE_TIME_MASK) > 1)) { + DRM_DEBUG_DRIVER("Engine Idle wait time not set properly.\n"); + enable_rc6 = false; + } + + if (!intel_uncore_read(uncore, GEN8_PUSHBUS_CONTROL) || + !intel_uncore_read(uncore, GEN8_PUSHBUS_ENABLE) || + !intel_uncore_read(uncore, GEN8_PUSHBUS_SHIFT)) { + DRM_DEBUG_DRIVER("Pushbus not setup properly.\n"); + enable_rc6 = false; + } + + if (!intel_uncore_read(uncore, GEN6_GFXPAUSE)) { + DRM_DEBUG_DRIVER("GFX pause not setup properly.\n"); + enable_rc6 = false; + } + + if (!intel_uncore_read(uncore, GEN8_MISC_CTRL0)) { + DRM_DEBUG_DRIVER("GPM control not setup properly.\n"); + enable_rc6 = false; + } + + return enable_rc6; +} + +static bool rc6_supported(struct intel_rc6 *rc6) +{ + struct drm_i915_private *i915 = rc6_to_i915(rc6); + + if (!HAS_RC6(i915)) + return false; + + if (intel_vgpu_active(i915)) + return false; + + if (is_mock_gt(rc6_to_gt(rc6))) + return false; + + if (IS_GEN9_LP(i915) && !bxt_check_bios_rc6_setup(rc6)) { + dev_notice(i915->drm.dev, + "RC6 and powersaving disabled by BIOS\n"); + return false; + } + + return true; +} + +static void rpm_get(struct intel_rc6 *rc6) +{ + GEM_BUG_ON(rc6->wakeref); + pm_runtime_get_sync(&rc6_to_i915(rc6)->drm.pdev->dev); + rc6->wakeref = true; +} + +static void rpm_put(struct intel_rc6 *rc6) +{ + GEM_BUG_ON(!rc6->wakeref); + pm_runtime_put(&rc6_to_i915(rc6)->drm.pdev->dev); + rc6->wakeref = false; +} + +static void __intel_rc6_disable(struct intel_rc6 *rc6) +{ + struct drm_i915_private *i915 = rc6_to_i915(rc6); + struct intel_uncore *uncore = rc6_to_uncore(rc6); + + intel_uncore_forcewake_get(uncore, FORCEWAKE_ALL); + if (INTEL_GEN(i915) >= 9) + set(uncore, GEN9_PG_ENABLE, 0); + set(uncore, GEN6_RC_CONTROL, 0); + set(uncore, GEN6_RC_STATE, 0); + intel_uncore_forcewake_put(uncore, FORCEWAKE_ALL); +} + +void intel_rc6_init(struct intel_rc6 *rc6) +{ + struct drm_i915_private *i915 = rc6_to_i915(rc6); + int err; + + /* Disable runtime-pm until we can save the GPU state with rc6 pctx */ + rpm_get(rc6); + + if (!rc6_supported(rc6)) + return; + + if (IS_CHERRYVIEW(i915)) + err = chv_rc6_init(rc6); + else if (IS_VALLEYVIEW(i915)) + err = vlv_rc6_init(rc6); + else + err = 0; + + /* Sanitize rc6, ensure it is disabled before we are ready. */ + __intel_rc6_disable(rc6); + + rc6->supported = err == 0; +} + +void intel_rc6_sanitize(struct intel_rc6 *rc6) +{ + if (rc6->supported) + __intel_rc6_disable(rc6); +} + +void intel_rc6_enable(struct intel_rc6 *rc6) +{ + struct drm_i915_private *i915 = rc6_to_i915(rc6); + struct intel_uncore *uncore = rc6_to_uncore(rc6); + + if (!rc6->supported) + return; + + GEM_BUG_ON(rc6->enabled); + + intel_uncore_forcewake_get(uncore, FORCEWAKE_ALL); + + if (IS_CHERRYVIEW(i915)) + chv_rc6_enable(rc6); + else if (IS_VALLEYVIEW(i915)) + vlv_rc6_enable(rc6); + else if (INTEL_GEN(i915) >= 11) + gen11_rc6_enable(rc6); + else if (INTEL_GEN(i915) >= 9) + gen9_rc6_enable(rc6); + else if (IS_BROADWELL(i915)) + gen8_rc6_enable(rc6); + else if (INTEL_GEN(i915) >= 6) + gen6_rc6_enable(rc6); + + intel_uncore_forcewake_put(uncore, FORCEWAKE_ALL); + + /* rc6 is ready, runtime-pm is go! */ + rpm_put(rc6); + rc6->enabled = true; +} + +void intel_rc6_disable(struct intel_rc6 *rc6) +{ + if (!rc6->enabled) + return; + + rpm_get(rc6); + rc6->enabled = false; + + __intel_rc6_disable(rc6); +} + +void intel_rc6_fini(struct intel_rc6 *rc6) +{ + struct drm_i915_gem_object *pctx; + + intel_rc6_disable(rc6); + + pctx = fetch_and_zero(&rc6->pctx); + if (pctx) + i915_gem_object_put(pctx); + + if (rc6->wakeref) + rpm_put(rc6); +} + +static u64 vlv_residency_raw(struct intel_uncore *uncore, const i915_reg_t reg) +{ + u32 lower, upper, tmp; + int loop = 2; + + /* + * The register accessed do not need forcewake. We borrow + * uncore lock to prevent concurrent access to range reg. + */ + lockdep_assert_held(&uncore->lock); + + /* + * vlv and chv residency counters are 40 bits in width. + * With a control bit, we can choose between upper or lower + * 32bit window into this counter. + * + * Although we always use the counter in high-range mode elsewhere, + * userspace may attempt to read the value before rc6 is initialised, + * before we have set the default VLV_COUNTER_CONTROL value. So always + * set the high bit to be safe. + */ + set(uncore, VLV_COUNTER_CONTROL, + _MASKED_BIT_ENABLE(VLV_COUNT_RANGE_HIGH)); + upper = intel_uncore_read_fw(uncore, reg); + do { + tmp = upper; + + set(uncore, VLV_COUNTER_CONTROL, + _MASKED_BIT_DISABLE(VLV_COUNT_RANGE_HIGH)); + lower = intel_uncore_read_fw(uncore, reg); + + set(uncore, VLV_COUNTER_CONTROL, + _MASKED_BIT_ENABLE(VLV_COUNT_RANGE_HIGH)); + upper = intel_uncore_read_fw(uncore, reg); + } while (upper != tmp && --loop); + + /* + * Everywhere else we always use VLV_COUNTER_CONTROL with the + * VLV_COUNT_RANGE_HIGH bit set - so it is safe to leave it set + * now. + */ + + return lower | (u64)upper << 8; +} + +u64 intel_rc6_residency_ns(struct intel_rc6 *rc6, const i915_reg_t reg) +{ + struct drm_i915_private *i915 = rc6_to_i915(rc6); + struct intel_uncore *uncore = rc6_to_uncore(rc6); + u64 time_hw, prev_hw, overflow_hw; + unsigned int fw_domains; + unsigned long flags; + unsigned int i; + u32 mul, div; + + if (!rc6->supported) + return 0; + + /* + * Store previous hw counter values for counter wrap-around handling. + * + * There are only four interesting registers and they live next to each + * other so we can use the relative address, compared to the smallest + * one as the index into driver storage. + */ + i = (i915_mmio_reg_offset(reg) - + i915_mmio_reg_offset(GEN6_GT_GFX_RC6_LOCKED)) / sizeof(u32); + if (WARN_ON_ONCE(i >= ARRAY_SIZE(rc6->cur_residency))) + return 0; + + fw_domains = intel_uncore_forcewake_for_reg(uncore, reg, FW_REG_READ); + + spin_lock_irqsave(&uncore->lock, flags); + intel_uncore_forcewake_get__locked(uncore, fw_domains); + + /* On VLV and CHV, residency time is in CZ units rather than 1.28us */ + if (IS_VALLEYVIEW(i915) || IS_CHERRYVIEW(i915)) { + mul = 1000000; + div = i915->czclk_freq; + overflow_hw = BIT_ULL(40); + time_hw = vlv_residency_raw(uncore, reg); + } else { + /* 833.33ns units on Gen9LP, 1.28us elsewhere. */ + if (IS_GEN9_LP(i915)) { + mul = 10000; + div = 12; + } else { + mul = 1280; + div = 1; + } + + overflow_hw = BIT_ULL(32); + time_hw = intel_uncore_read_fw(uncore, reg); + } + + /* + * Counter wrap handling. + * + * But relying on a sufficient frequency of queries otherwise counters + * can still wrap. + */ + prev_hw = rc6->prev_hw_residency[i]; + rc6->prev_hw_residency[i] = time_hw; + + /* RC6 delta from last sample. */ + if (time_hw >= prev_hw) + time_hw -= prev_hw; + else + time_hw += overflow_hw - prev_hw; + + /* Add delta to RC6 extended raw driver copy. */ + time_hw += rc6->cur_residency[i]; + rc6->cur_residency[i] = time_hw; + + intel_uncore_forcewake_put__locked(uncore, fw_domains); + spin_unlock_irqrestore(&uncore->lock, flags); + + return mul_u64_u32_div(time_hw, mul, div); +} + +u64 intel_rc6_residency_us(struct intel_rc6 *rc6, i915_reg_t reg) +{ + return DIV_ROUND_UP_ULL(intel_rc6_residency_ns(rc6, reg), 1000); +} diff --git a/drivers/gpu/drm/i915/gt/intel_rc6.h b/drivers/gpu/drm/i915/gt/intel_rc6.h new file mode 100644 index 000000000000..5e6711f36457 --- /dev/null +++ b/drivers/gpu/drm/i915/gt/intel_rc6.h @@ -0,0 +1,25 @@ +/* + * SPDX-License-Identifier: MIT + * + * Copyright © 2019 Intel Corporation + */ + +#ifndef INTEL_RC6_H +#define INTEL_RC6_H + +#include "i915_reg.h" + +struct intel_engine_cs; +struct intel_rc6; + +void intel_rc6_init(struct intel_rc6 *rc6); +void intel_rc6_fini(struct intel_rc6 *rc6); + +void intel_rc6_sanitize(struct intel_rc6 *rc6); +void intel_rc6_enable(struct intel_rc6 *rc6); +void intel_rc6_disable(struct intel_rc6 *rc6); + +u64 intel_rc6_residency_ns(struct intel_rc6 *rc6, i915_reg_t reg); +u64 intel_rc6_residency_us(struct intel_rc6 *rc6, i915_reg_t reg); + +#endif /* INTEL_RC6_H */ diff --git a/drivers/gpu/drm/i915/gt/intel_rc6_types.h b/drivers/gpu/drm/i915/gt/intel_rc6_types.h new file mode 100644 index 000000000000..214f354d6ae4 --- /dev/null +++ b/drivers/gpu/drm/i915/gt/intel_rc6_types.h @@ -0,0 +1,28 @@ +/* + * SPDX-License-Identifier: MIT + * + * Copyright © 2019 Intel Corporation + */ + +#ifndef INTEL_RC6_TYPES_H +#define INTEL_RC6_TYPES_H + +#include <linux/spinlock.h> +#include <linux/types.h> + +#include "intel_engine_types.h" + +struct drm_i915_gem_object; + +struct intel_rc6 { + u64 prev_hw_residency[4]; + u64 cur_residency[4]; + + struct drm_i915_gem_object *pctx; + + bool supported : 1; + bool enabled : 1; + bool wakeref : 1; +}; + +#endif /* INTEL_RC6_TYPES_H */ diff --git a/drivers/gpu/drm/i915/gt/intel_reset.c b/drivers/gpu/drm/i915/gt/intel_reset.c index b9d84d52e986..7b3d9d4517a0 100644 --- a/drivers/gpu/drm/i915/gt/intel_reset.c +++ b/drivers/gpu/drm/i915/gt/intel_reset.c @@ -42,11 +42,10 @@ static void engine_skip_context(struct i915_request *rq) struct intel_engine_cs *engine = rq->engine; struct i915_gem_context *hung_ctx = rq->gem_context; - lockdep_assert_held(&engine->active.lock); - if (!i915_request_is_active(rq)) return; + lockdep_assert_held(&engine->active.lock); list_for_each_entry_continue(rq, &engine->active.requests, sched.link) if (rq->gem_context == hung_ctx) i915_request_skip(rq, -EIO); @@ -123,7 +122,6 @@ void __i915_request_reset(struct i915_request *rq, bool guilty) rq->fence.seqno, yesno(guilty)); - lockdep_assert_held(&rq->engine->active.lock); GEM_BUG_ON(i915_request_completed(rq)); if (guilty) { @@ -309,7 +307,7 @@ static int gen6_reset_engines(struct intel_gt *gt, return gen6_hw_domain_reset(gt, hw_mask); } -static u32 gen11_lock_sfc(struct intel_engine_cs *engine) +static int gen11_lock_sfc(struct intel_engine_cs *engine, u32 *hw_mask) { struct intel_uncore *uncore = engine->uncore; u8 vdbox_sfc_access = RUNTIME_INFO(engine->i915)->vdbox_sfc_access; @@ -318,6 +316,7 @@ static u32 gen11_lock_sfc(struct intel_engine_cs *engine) i915_reg_t sfc_usage; u32 sfc_usage_bit; u32 sfc_reset_bit; + int ret; switch (engine->class) { case VIDEO_DECODE_CLASS: @@ -352,27 +351,33 @@ static u32 gen11_lock_sfc(struct intel_engine_cs *engine) } /* - * Tell the engine that a software reset is going to happen. The engine - * will then try to force lock the SFC (if currently locked, it will - * remain so until we tell the engine it is safe to unlock; if currently - * unlocked, it will ignore this and all new lock requests). If SFC - * ends up being locked to the engine we want to reset, we have to reset - * it as well (we will unlock it once the reset sequence is completed). + * If the engine is using a SFC, tell the engine that a software reset + * is going to happen. The engine will then try to force lock the SFC. + * If SFC ends up being locked to the engine we want to reset, we have + * to reset it as well (we will unlock it once the reset sequence is + * completed). */ + if (!(intel_uncore_read_fw(uncore, sfc_usage) & sfc_usage_bit)) + return 0; + rmw_set_fw(uncore, sfc_forced_lock, sfc_forced_lock_bit); - if (__intel_wait_for_register_fw(uncore, - sfc_forced_lock_ack, - sfc_forced_lock_ack_bit, - sfc_forced_lock_ack_bit, - 1000, 0, NULL)) { - DRM_DEBUG_DRIVER("Wait for SFC forced lock ack failed\n"); + ret = __intel_wait_for_register_fw(uncore, + sfc_forced_lock_ack, + sfc_forced_lock_ack_bit, + sfc_forced_lock_ack_bit, + 1000, 0, NULL); + + /* Was the SFC released while we were trying to lock it? */ + if (!(intel_uncore_read_fw(uncore, sfc_usage) & sfc_usage_bit)) return 0; - } - if (intel_uncore_read_fw(uncore, sfc_usage) & sfc_usage_bit) - return sfc_reset_bit; + if (ret) { + DRM_DEBUG_DRIVER("Wait for SFC forced lock ack failed\n"); + return ret; + } + *hw_mask |= sfc_reset_bit; return 0; } @@ -430,12 +435,21 @@ static int gen11_reset_engines(struct intel_gt *gt, 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 = gen11_lock_sfc(engine, &hw_mask); + if (ret) + goto sfc_unlock; } } ret = gen6_hw_domain_reset(gt, hw_mask); +sfc_unlock: + /* + * We unlock the SFC based on the lock status and not the result of + * gen11_lock_sfc to make sure that we clean properly if something + * wrong happened during the lock (e.g. lock acquired after timeout + * expiration). + */ if (engine_mask != ALL_ENGINES) for_each_engine_masked(engine, gt->i915, engine_mask, tmp) gen11_unlock_sfc(engine); @@ -528,13 +542,24 @@ skip_reset: return ret; } +static int mock_reset(struct intel_gt *gt, + intel_engine_mask_t mask, + unsigned int retry) +{ + return 0; +} + typedef int (*reset_func)(struct intel_gt *, intel_engine_mask_t engine_mask, unsigned int retry); -static reset_func intel_get_gpu_reset(struct drm_i915_private *i915) +static reset_func intel_get_gpu_reset(const struct intel_gt *gt) { - if (INTEL_GEN(i915) >= 8) + struct drm_i915_private *i915 = gt->i915; + + if (is_mock_gt(gt)) + return mock_reset; + else if (INTEL_GEN(i915) >= 8) return gen8_reset_engines; else if (INTEL_GEN(i915) >= 6) return gen6_reset_engines; @@ -557,7 +582,7 @@ int __intel_gt_reset(struct intel_gt *gt, intel_engine_mask_t engine_mask) int ret = -ETIMEDOUT; int retry; - reset = intel_get_gpu_reset(gt->i915); + reset = intel_get_gpu_reset(gt); if (!reset) return -ENODEV; @@ -577,17 +602,20 @@ int __intel_gt_reset(struct intel_gt *gt, intel_engine_mask_t engine_mask) return ret; } -bool intel_has_gpu_reset(struct drm_i915_private *i915) +bool intel_has_gpu_reset(const struct intel_gt *gt) { if (!i915_modparams.reset) return NULL; - return intel_get_gpu_reset(i915); + return intel_get_gpu_reset(gt); } -bool intel_has_reset_engine(struct drm_i915_private *i915) +bool intel_has_reset_engine(const struct intel_gt *gt) { - return INTEL_INFO(i915)->has_reset_engine && i915_modparams.reset >= 2; + if (i915_modparams.reset < 2) + return false; + + return INTEL_INFO(gt->i915)->has_reset_engine; } int intel_reset_guc(struct intel_gt *gt) @@ -697,7 +725,7 @@ static void reset_finish_engine(struct intel_engine_cs *engine) engine->reset.finish(engine); intel_uncore_forcewake_put(engine->uncore, FORCEWAKE_ALL); - intel_engine_signal_breadcrumbs(engine); + intel_engine_breadcrumbs_irq(engine); } static void reset_finish(struct intel_gt *gt, intel_engine_mask_t awake) @@ -793,11 +821,13 @@ static bool __intel_gt_unset_wedged(struct intel_gt *gt) struct intel_gt_timelines *timelines = >->timelines; struct intel_timeline *tl; unsigned long flags; + bool ok; if (!test_bit(I915_WEDGED, >->reset.flags)) return true; - if (!gt->scratch) /* Never full initialised, recovery impossible */ + /* Never fully initialised, recovery impossible */ + if (test_bit(I915_WEDGED_ON_INIT, >->reset.flags)) return false; GEM_TRACE("start\n"); @@ -814,10 +844,10 @@ static bool __intel_gt_unset_wedged(struct intel_gt *gt) */ spin_lock_irqsave(&timelines->lock, flags); list_for_each_entry(tl, &timelines->active_list, link) { - struct i915_request *rq; + struct dma_fence *fence; - rq = i915_active_request_get_unlocked(&tl->last_request); - if (!rq) + fence = i915_active_fence_get(&tl->last_request); + if (!fence) continue; spin_unlock_irqrestore(&timelines->lock, flags); @@ -829,8 +859,8 @@ static bool __intel_gt_unset_wedged(struct intel_gt *gt) * (I915_FENCE_TIMEOUT) so this wait should not be unbounded * in the worst case. */ - dma_fence_default_wait(&rq->fence, false, MAX_SCHEDULE_TIMEOUT); - i915_request_put(rq); + dma_fence_default_wait(fence, false, MAX_SCHEDULE_TIMEOUT); + dma_fence_put(fence); /* Restart iteration after droping lock */ spin_lock_irqsave(&timelines->lock, flags); @@ -838,7 +868,12 @@ static bool __intel_gt_unset_wedged(struct intel_gt *gt) } spin_unlock_irqrestore(&timelines->lock, flags); - intel_gt_sanitize(gt, false); + /* We must reset pending GPU events before restoring our submission */ + ok = !HAS_EXECLISTS(gt->i915); /* XXX better agnosticism desired */ + if (!INTEL_INFO(gt->i915)->gpu_reset_clobbers_display) + ok = __intel_gt_reset(gt, ALL_ENGINES) == 0; + if (!ok) + return false; /* * Undo nop_submit_request. We prevent all new i915 requests from @@ -943,7 +978,7 @@ void intel_gt_reset(struct intel_gt *gt, awake = reset_prepare(gt); - if (!intel_has_gpu_reset(gt->i915)) { + if (!intel_has_gpu_reset(gt)) { if (i915_modparams.reset) dev_err(gt->i915->drm.dev, "GPU reset not supported\n"); else @@ -972,7 +1007,7 @@ void intel_gt_reset(struct intel_gt *gt, * was running at the time of the reset (i.e. we weren't VT * switched away). */ - ret = i915_gem_init_hw(gt->i915); + ret = intel_gt_init_hw(gt); if (ret) { DRM_ERROR("Failed to initialise HW following reset (%d)\n", ret); @@ -1164,7 +1199,7 @@ void intel_gt_handle_error(struct intel_gt *gt, * Try engine reset when available. We fall back to full reset if * single reset fails. */ - if (intel_has_reset_engine(gt->i915) && !intel_gt_is_wedged(gt)) { + if (intel_has_reset_engine(gt) && !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, @@ -1214,10 +1249,8 @@ out: intel_runtime_pm_put(>->i915->runtime_pm, wakeref); } -int intel_gt_reset_trylock(struct intel_gt *gt) +int intel_gt_reset_trylock(struct intel_gt *gt, int *srcu) { - int srcu; - might_lock(>->reset.backoff_srcu); might_sleep(); @@ -1232,10 +1265,10 @@ int intel_gt_reset_trylock(struct intel_gt *gt) rcu_read_lock(); } - srcu = srcu_read_lock(>->reset.backoff_srcu); + *srcu = srcu_read_lock(>->reset.backoff_srcu); rcu_read_unlock(); - return srcu; + return 0; } void intel_gt_reset_unlock(struct intel_gt *gt, int tag) @@ -1255,10 +1288,6 @@ int intel_gt_terminally_wedged(struct intel_gt *gt) 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)) - return -EAGAIN; - if (wait_event_interruptible(gt->reset.queue, !test_bit(I915_RESET_BACKOFF, >->reset.flags))) @@ -1267,6 +1296,14 @@ int intel_gt_terminally_wedged(struct intel_gt *gt) return intel_gt_is_wedged(gt) ? -EIO : 0; } +void intel_gt_set_wedged_on_init(struct intel_gt *gt) +{ + BUILD_BUG_ON(I915_RESET_ENGINE + I915_NUM_ENGINES > + I915_WEDGED_ON_INIT); + intel_gt_set_wedged(gt); + set_bit(I915_WEDGED_ON_INIT, >->reset.flags); +} + void intel_gt_init_reset(struct intel_gt *gt) { init_waitqueue_head(>->reset.queue); diff --git a/drivers/gpu/drm/i915/gt/intel_reset.h b/drivers/gpu/drm/i915/gt/intel_reset.h index 37a987b17108..8e8d5f761166 100644 --- a/drivers/gpu/drm/i915/gt/intel_reset.h +++ b/drivers/gpu/drm/i915/gt/intel_reset.h @@ -14,7 +14,6 @@ #include "intel_engine_types.h" #include "intel_reset_types.h" -struct drm_i915_private; struct i915_request; struct intel_engine_cs; struct intel_gt; @@ -38,13 +37,19 @@ int intel_engine_reset(struct intel_engine_cs *engine, void __i915_request_reset(struct i915_request *rq, bool guilty); -int __must_check intel_gt_reset_trylock(struct intel_gt *gt); +int __must_check intel_gt_reset_trylock(struct intel_gt *gt, int *srcu); void intel_gt_reset_unlock(struct intel_gt *gt, int tag); 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); +/* + * There's no unset_wedged_on_init paired with this one. + * Once we're wedged on init, there's no going back. + */ +void intel_gt_set_wedged_on_init(struct intel_gt *gt); + int __intel_gt_reset(struct intel_gt *gt, intel_engine_mask_t engine_mask); int intel_reset_guc(struct intel_gt *gt); @@ -68,10 +73,13 @@ void __intel_fini_wedge(struct intel_wedge_me *w); static inline bool __intel_reset_failed(const struct intel_reset *reset) { + GEM_BUG_ON(test_bit(I915_WEDGED_ON_INIT, &reset->flags) ? + !test_bit(I915_WEDGED, &reset->flags) : false); + 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); +bool intel_has_gpu_reset(const struct intel_gt *gt); +bool intel_has_reset_engine(const struct intel_gt *gt); #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 index 31968356e0c0..f43bc3a0fe4f 100644 --- a/drivers/gpu/drm/i915/gt/intel_reset_types.h +++ b/drivers/gpu/drm/i915/gt/intel_reset_types.h @@ -29,11 +29,17 @@ struct intel_reset { * 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. + * + * #I915_WEDGED_ON_INIT - If we fail to initialize the GPU we can no + * longer use the GPU - similar to #I915_WEDGED bit. The difference in + * in the way we're handling "forced" unwedged (e.g. through debugfs), + * which is not allowed in case we failed to initialize. */ unsigned long flags; #define I915_RESET_BACKOFF 0 #define I915_RESET_MODESET 1 #define I915_RESET_ENGINE 2 +#define I915_WEDGED_ON_INIT (BITS_PER_LONG - 2) #define I915_WEDGED (BITS_PER_LONG - 1) struct mutex mutex; /* serialises wedging/unwedging */ diff --git a/drivers/gpu/drm/i915/gt/intel_ringbuffer.c b/drivers/gpu/drm/i915/gt/intel_ringbuffer.c index 601c16239fdf..311fdc0a21bc 100644 --- a/drivers/gpu/drm/i915/gt/intel_ringbuffer.c +++ b/drivers/gpu/drm/i915/gt/intel_ringbuffer.c @@ -322,7 +322,8 @@ static u32 *gen6_rcs_emit_breadcrumb(struct i915_request *rq, u32 *cs) PIPE_CONTROL_DC_FLUSH_ENABLE | PIPE_CONTROL_QW_WRITE | PIPE_CONTROL_CS_STALL); - *cs++ = rq->timeline->hwsp_offset | PIPE_CONTROL_GLOBAL_GTT; + *cs++ = i915_request_active_timeline(rq)->hwsp_offset | + PIPE_CONTROL_GLOBAL_GTT; *cs++ = rq->fence.seqno; *cs++ = MI_USER_INTERRUPT; @@ -425,7 +426,7 @@ static u32 *gen7_rcs_emit_breadcrumb(struct i915_request *rq, u32 *cs) PIPE_CONTROL_QW_WRITE | PIPE_CONTROL_GLOBAL_GTT_IVB | PIPE_CONTROL_CS_STALL); - *cs++ = rq->timeline->hwsp_offset; + *cs++ = i915_request_active_timeline(rq)->hwsp_offset; *cs++ = rq->fence.seqno; *cs++ = MI_USER_INTERRUPT; @@ -439,8 +440,8 @@ static u32 *gen7_rcs_emit_breadcrumb(struct i915_request *rq, u32 *cs) static u32 *gen6_xcs_emit_breadcrumb(struct i915_request *rq, u32 *cs) { - GEM_BUG_ON(rq->timeline->hwsp_ggtt != rq->engine->status_page.vma); - GEM_BUG_ON(offset_in_page(rq->timeline->hwsp_offset) != I915_GEM_HWS_SEQNO_ADDR); + GEM_BUG_ON(i915_request_active_timeline(rq)->hwsp_ggtt != rq->engine->status_page.vma); + GEM_BUG_ON(offset_in_page(i915_request_active_timeline(rq)->hwsp_offset) != I915_GEM_HWS_SEQNO_ADDR); *cs++ = MI_FLUSH_DW | MI_FLUSH_DW_OP_STOREDW | MI_FLUSH_DW_STORE_INDEX; *cs++ = I915_GEM_HWS_SEQNO_ADDR | MI_FLUSH_DW_USE_GTT; @@ -459,8 +460,8 @@ static u32 *gen7_xcs_emit_breadcrumb(struct i915_request *rq, u32 *cs) { int i; - GEM_BUG_ON(rq->timeline->hwsp_ggtt != rq->engine->status_page.vma); - GEM_BUG_ON(offset_in_page(rq->timeline->hwsp_offset) != I915_GEM_HWS_SEQNO_ADDR); + GEM_BUG_ON(i915_request_active_timeline(rq)->hwsp_ggtt != rq->engine->status_page.vma); + GEM_BUG_ON(offset_in_page(i915_request_active_timeline(rq)->hwsp_offset) != I915_GEM_HWS_SEQNO_ADDR); *cs++ = MI_FLUSH_DW | MI_FLUSH_DW_OP_STOREDW | MI_FLUSH_DW_STORE_INDEX; *cs++ = I915_GEM_HWS_SEQNO_ADDR | MI_FLUSH_DW_USE_GTT; @@ -930,6 +931,7 @@ static void cancel_requests(struct intel_engine_cs *engine) static void i9xx_submit_request(struct i915_request *request) { i915_request_submit(request); + wmb(); /* paranoid flush writes out of the WCB before mmio */ ENGINE_WRITE(request->engine, RING_TAIL, intel_ring_set_tail(request->ring, request->tail)); @@ -937,8 +939,8 @@ static void i9xx_submit_request(struct i915_request *request) static u32 *i9xx_emit_breadcrumb(struct i915_request *rq, u32 *cs) { - GEM_BUG_ON(rq->timeline->hwsp_ggtt != rq->engine->status_page.vma); - GEM_BUG_ON(offset_in_page(rq->timeline->hwsp_offset) != I915_GEM_HWS_SEQNO_ADDR); + GEM_BUG_ON(i915_request_active_timeline(rq)->hwsp_ggtt != rq->engine->status_page.vma); + GEM_BUG_ON(offset_in_page(i915_request_active_timeline(rq)->hwsp_offset) != I915_GEM_HWS_SEQNO_ADDR); *cs++ = MI_FLUSH; @@ -960,8 +962,8 @@ static u32 *gen5_emit_breadcrumb(struct i915_request *rq, u32 *cs) { int i; - GEM_BUG_ON(rq->timeline->hwsp_ggtt != rq->engine->status_page.vma); - GEM_BUG_ON(offset_in_page(rq->timeline->hwsp_offset) != I915_GEM_HWS_SEQNO_ADDR); + GEM_BUG_ON(i915_request_active_timeline(rq)->hwsp_ggtt != rq->engine->status_page.vma); + GEM_BUG_ON(offset_in_page(i915_request_active_timeline(rq)->hwsp_offset) != I915_GEM_HWS_SEQNO_ADDR); *cs++ = MI_FLUSH; @@ -1272,7 +1274,7 @@ static struct i915_vma *create_ring_vma(struct i915_ggtt *ggtt, int size) struct i915_vma *vma; obj = i915_gem_object_create_stolen(i915, size); - if (!obj) + if (IS_ERR(obj)) obj = i915_gem_object_create_internal(i915, size); if (IS_ERR(obj)) return ERR_CAST(obj); @@ -1336,15 +1338,13 @@ void intel_ring_free(struct kref *ref) { struct intel_ring *ring = container_of(ref, typeof(*ring), ref); - i915_vma_close(ring->vma); i915_vma_put(ring->vma); - kfree(ring); } static void __ring_context_fini(struct intel_context *ce) { - i915_gem_object_put(ce->state->obj); + i915_vma_put(ce->state); } static void ring_context_destroy(struct kref *ref) @@ -1573,7 +1573,7 @@ static inline int mi_set_context(struct i915_request *rq, u32 flags) struct intel_engine_cs *engine = rq->engine; enum intel_engine_id id; const int num_engines = - IS_HSW_GT1(i915) ? RUNTIME_INFO(i915)->num_engines - 1 : 0; + IS_HASWELL(i915) ? RUNTIME_INFO(i915)->num_engines - 1 : 0; bool force_restore = false; int len; u32 *cs; @@ -1741,46 +1741,22 @@ static int remap_l3(struct i915_request *rq) static int switch_context(struct i915_request *rq) { - struct intel_engine_cs *engine = rq->engine; - struct i915_address_space *vm = vm_alias(rq->hw_context); - unsigned int unwind_mm = 0; - u32 hw_flags = 0; + struct intel_context *ce = rq->hw_context; + struct i915_address_space *vm = vm_alias(ce); int ret; GEM_BUG_ON(HAS_EXECLISTS(rq->i915)); if (vm) { - struct i915_ppgtt *ppgtt = i915_vm_to_ppgtt(vm); - int loops; - - /* - * Baytail takes a little more convincing that it really needs - * to reload the PD between contexts. It is not just a little - * longer, as adding more stalls after the load_pd_dir (i.e. - * adding a long loop around flush_pd_dir) is not as effective - * as reloading the PD umpteen times. 32 is derived from - * experimentation (gem_exec_parallel/fds) and has no good - * explanation. - */ - loops = 1; - if (engine->id == BCS0 && IS_VALLEYVIEW(engine->i915)) - loops = 32; - - do { - ret = load_pd_dir(rq, ppgtt); - if (ret) - goto err; - } while (--loops); - - if (ppgtt->pd_dirty_engines & engine->mask) { - unwind_mm = engine->mask; - ppgtt->pd_dirty_engines &= ~unwind_mm; - hw_flags = MI_FORCE_RESTORE; - } + ret = load_pd_dir(rq, i915_vm_to_ppgtt(vm)); + if (ret) + return ret; } - if (rq->hw_context->state) { - GEM_BUG_ON(engine->id != RCS0); + if (ce->state) { + u32 hw_flags; + + GEM_BUG_ON(rq->engine->id != RCS0); /* * The kernel context(s) is treated as pure scratch and is not @@ -1789,22 +1765,25 @@ static int switch_context(struct i915_request *rq) * as nothing actually executes using the kernel context; it * is purely used for flushing user contexts. */ + hw_flags = 0; if (i915_gem_context_is_kernel(rq->gem_context)) hw_flags = MI_RESTORE_INHIBIT; ret = mi_set_context(rq, hw_flags); if (ret) - goto err_mm; + return ret; } if (vm) { + struct intel_engine_cs *engine = rq->engine; + ret = engine->emit_flush(rq, EMIT_INVALIDATE); if (ret) - goto err_mm; + return ret; ret = flush_pd_dir(rq); if (ret) - goto err_mm; + return ret; /* * Not only do we need a full barrier (post-sync write) after @@ -1816,24 +1795,18 @@ static int switch_context(struct i915_request *rq) */ ret = engine->emit_flush(rq, EMIT_INVALIDATE); if (ret) - goto err_mm; + return ret; ret = engine->emit_flush(rq, EMIT_FLUSH); if (ret) - goto err_mm; + return ret; } ret = remap_l3(rq); if (ret) - goto err_mm; + return ret; return 0; - -err_mm: - if (unwind_mm) - i915_vm_to_ppgtt(vm)->pd_dirty_engines |= unwind_mm; -err: - return ret; } static int ring_request_alloc(struct i915_request *request) @@ -1841,7 +1814,7 @@ static int ring_request_alloc(struct i915_request *request) int ret; GEM_BUG_ON(!intel_context_is_pinned(request->hw_context)); - GEM_BUG_ON(request->timeline->has_initial_breadcrumb); + GEM_BUG_ON(i915_request_timeline(request)->has_initial_breadcrumb); /* * Flush enough space to reduce the likelihood of waiting after @@ -1952,7 +1925,9 @@ u32 *intel_ring_begin(struct i915_request *rq, unsigned int num_dwords) */ GEM_BUG_ON(!rq->reserved_space); - ret = wait_for_space(ring, rq->timeline, total_bytes); + ret = wait_for_space(ring, + i915_request_timeline(rq), + total_bytes); if (unlikely(ret)) return ERR_PTR(ret); } diff --git a/drivers/gpu/drm/i915/gt/intel_sseu.c b/drivers/gpu/drm/i915/gt/intel_sseu.c index 6bf2d87da109..74f793423231 100644 --- a/drivers/gpu/drm/i915/gt/intel_sseu.c +++ b/drivers/gpu/drm/i915/gt/intel_sseu.c @@ -8,6 +8,19 @@ #include "intel_lrc_reg.h" #include "intel_sseu.h" +void intel_sseu_set_info(struct sseu_dev_info *sseu, u8 max_slices, + u8 max_subslices, u8 max_eus_per_subslice) +{ + sseu->max_slices = max_slices; + sseu->max_subslices = max_subslices; + sseu->max_eus_per_subslice = max_eus_per_subslice; + + sseu->ss_stride = GEN_SSEU_STRIDE(sseu->max_subslices); + GEM_BUG_ON(sseu->ss_stride > GEN_MAX_SUBSLICE_STRIDE); + sseu->eu_stride = GEN_SSEU_STRIDE(sseu->max_eus_per_subslice); + GEM_BUG_ON(sseu->eu_stride > GEN_MAX_EU_STRIDE); +} + unsigned int intel_sseu_subslice_total(const struct sseu_dev_info *sseu) { @@ -19,10 +32,32 @@ intel_sseu_subslice_total(const struct sseu_dev_info *sseu) return total; } +u32 intel_sseu_get_subslices(const struct sseu_dev_info *sseu, u8 slice) +{ + int i, offset = slice * sseu->ss_stride; + u32 mask = 0; + + GEM_BUG_ON(slice >= sseu->max_slices); + + for (i = 0; i < sseu->ss_stride; i++) + mask |= (u32)sseu->subslice_mask[offset + i] << + i * BITS_PER_BYTE; + + return mask; +} + +void intel_sseu_set_subslices(struct sseu_dev_info *sseu, int slice, + u32 ss_mask) +{ + int offset = slice * sseu->ss_stride; + + memcpy(&sseu->subslice_mask[offset], &ss_mask, sseu->ss_stride); +} + unsigned int intel_sseu_subslices_per_slice(const struct sseu_dev_info *sseu, u8 slice) { - return hweight8(sseu->subslice_mask[slice]); + return hweight32(intel_sseu_get_subslices(sseu, slice)); } u32 intel_sseu_make_rpcs(struct drm_i915_private *i915, diff --git a/drivers/gpu/drm/i915/gt/intel_sseu.h b/drivers/gpu/drm/i915/gt/intel_sseu.h index b50d0401a4e2..d1d225204f09 100644 --- a/drivers/gpu/drm/i915/gt/intel_sseu.h +++ b/drivers/gpu/drm/i915/gt/intel_sseu.h @@ -10,15 +10,21 @@ #include <linux/types.h> #include <linux/kernel.h> +#include "i915_gem.h" + struct drm_i915_private; #define GEN_MAX_SLICES (6) /* CNL upper bound */ #define GEN_MAX_SUBSLICES (8) /* ICL upper bound */ #define GEN_SSEU_STRIDE(max_entries) DIV_ROUND_UP(max_entries, BITS_PER_BYTE) +#define GEN_MAX_SUBSLICE_STRIDE GEN_SSEU_STRIDE(GEN_MAX_SUBSLICES) +#define GEN_MAX_EUS (16) /* TGL upper bound */ +#define GEN_MAX_EU_STRIDE GEN_SSEU_STRIDE(GEN_MAX_EUS) struct sseu_dev_info { u8 slice_mask; - u8 subslice_mask[GEN_MAX_SLICES]; + u8 subslice_mask[GEN_MAX_SLICES * GEN_MAX_SUBSLICE_STRIDE]; + u8 eu_mask[GEN_MAX_SLICES * GEN_MAX_SUBSLICES * GEN_MAX_EU_STRIDE]; u16 eu_total; u8 eu_per_subslice; u8 min_eu_in_pool; @@ -33,11 +39,8 @@ struct sseu_dev_info { u8 max_subslices; u8 max_eus_per_subslice; - /* We don't have more than 8 eus per subslice at the moment and as we - * store eus enabled using bits, no need to multiply by eus per - * subslice. - */ - u8 eu_mask[GEN_MAX_SLICES * GEN_MAX_SUBSLICES]; + u8 ss_stride; + u8 eu_stride; }; /* @@ -63,12 +66,34 @@ intel_sseu_from_device_info(const struct sseu_dev_info *sseu) return value; } +static inline bool +intel_sseu_has_subslice(const struct sseu_dev_info *sseu, int slice, + int subslice) +{ + u8 mask; + int ss_idx = subslice / BITS_PER_BYTE; + + GEM_BUG_ON(ss_idx >= sseu->ss_stride); + + mask = sseu->subslice_mask[slice * sseu->ss_stride + ss_idx]; + + return mask & BIT(subslice % BITS_PER_BYTE); +} + +void intel_sseu_set_info(struct sseu_dev_info *sseu, u8 max_slices, + u8 max_subslices, u8 max_eus_per_subslice); + unsigned int intel_sseu_subslice_total(const struct sseu_dev_info *sseu); unsigned int intel_sseu_subslices_per_slice(const struct sseu_dev_info *sseu, u8 slice); +u32 intel_sseu_get_subslices(const struct sseu_dev_info *sseu, u8 slice); + +void intel_sseu_set_subslices(struct sseu_dev_info *sseu, int slice, + u32 ss_mask); + u32 intel_sseu_make_rpcs(struct drm_i915_private *i915, const struct intel_sseu *req_sseu); diff --git a/drivers/gpu/drm/i915/gt/intel_timeline.c b/drivers/gpu/drm/i915/gt/intel_timeline.c index 9cb01d9828f1..0f959694303c 100644 --- a/drivers/gpu/drm/i915/gt/intel_timeline.c +++ b/drivers/gpu/drm/i915/gt/intel_timeline.c @@ -136,6 +136,7 @@ static void __idle_cacheline_free(struct intel_timeline_cacheline *cl) kfree(cl); } +__i915_active_call static void __cacheline_retire(struct i915_active *active) { struct intel_timeline_cacheline *cl = @@ -177,8 +178,7 @@ cacheline_alloc(struct intel_timeline_hwsp *hwsp, unsigned int cacheline) cl->hwsp = hwsp; cl->vaddr = page_pack_bits(vaddr, cacheline); - i915_active_init(hwsp->gt->i915, &cl->active, - __cacheline_active, __cacheline_retire); + i915_active_init(&cl->active, __cacheline_active, __cacheline_retire); return cl; } @@ -254,7 +254,7 @@ int intel_timeline_init(struct intel_timeline *timeline, mutex_init(&timeline->mutex); - INIT_ACTIVE_REQUEST(&timeline->last_request, &timeline->mutex); + INIT_ACTIVE_FENCE(&timeline->last_request, &timeline->mutex); INIT_LIST_HEAD(&timeline->requests); i915_syncmap_init(&timeline->sync); @@ -442,7 +442,7 @@ __intel_timeline_get_seqno(struct intel_timeline *tl, * 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, rq); + err = i915_active_ref(&tl->hwsp_cacheline->active, tl, &rq->fence); if (err) goto err_cacheline; @@ -493,24 +493,39 @@ int intel_timeline_get_seqno(struct intel_timeline *tl, static int cacheline_ref(struct intel_timeline_cacheline *cl, struct i915_request *rq) { - return i915_active_ref(&cl->active, rq->timeline, rq); + return i915_active_add_request(&cl->active, 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; + struct intel_timeline *tl; int err; - GEM_BUG_ON(to->timeline == tl); + rcu_read_lock(); + tl = rcu_dereference(from->timeline); + if (i915_request_completed(from) || !kref_get_unless_zero(&tl->kref)) + tl = NULL; + rcu_read_unlock(); + if (!tl) /* already completed */ + return 1; + + GEM_BUG_ON(rcu_access_pointer(to->timeline) == tl); + + err = -EBUSY; + if (mutex_trylock(&tl->mutex)) { + struct intel_timeline_cacheline *cl = from->hwsp_cacheline; + + if (i915_request_completed(from)) { + err = 1; + goto unlock; + } - mutex_lock_nested(&tl->mutex, SINGLE_DEPTH_NESTING); - err = i915_request_completed(from); - if (!err) err = cacheline_ref(cl, to); - if (!err) { + if (err) + goto unlock; + if (likely(cl == tl->hwsp_cacheline)) { *hwsp = tl->hwsp_offset; } else { /* across a seqno wrap, recover the original offset */ @@ -518,8 +533,11 @@ int intel_timeline_read_hwsp(struct i915_request *from, ptr_unmask_bits(cl->vaddr, CACHELINE_BITS) * CACHELINE_BYTES; } + +unlock: + mutex_unlock(&tl->mutex); } - mutex_unlock(&tl->mutex); + intel_timeline_put(tl); return err; } @@ -541,7 +559,7 @@ void __intel_timeline_free(struct kref *kref) container_of(kref, typeof(*timeline), kref); intel_timeline_fini(timeline); - kfree(timeline); + kfree_rcu(timeline, rcu); } static void timelines_fini(struct intel_gt *gt) diff --git a/drivers/gpu/drm/i915/gt/intel_timeline_types.h b/drivers/gpu/drm/i915/gt/intel_timeline_types.h index 2b1baf2fcc8e..98d9ee166379 100644 --- a/drivers/gpu/drm/i915/gt/intel_timeline_types.h +++ b/drivers/gpu/drm/i915/gt/intel_timeline_types.h @@ -58,12 +58,13 @@ struct intel_timeline { */ struct list_head requests; - /* Contains an RCU guarded pointer to the last request. No reference is + /* + * 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. + * the request using i915_active_fence_get(), or manage the RCU + * protection themselves (cf the i915_active_fence API). */ - struct i915_active_request last_request; + struct i915_active_fence last_request; /** * We track the most recent seqno that we wait on in every context so @@ -80,6 +81,7 @@ struct intel_timeline { struct intel_gt *gt; struct kref kref; + struct rcu_head rcu; }; #endif /* __I915_TIMELINE_TYPES_H__ */ diff --git a/drivers/gpu/drm/i915/gt/intel_workarounds.c b/drivers/gpu/drm/i915/gt/intel_workarounds.c index 45481eb1fa3c..ba65e5018978 100644 --- a/drivers/gpu/drm/i915/gt/intel_workarounds.c +++ b/drivers/gpu/drm/i915/gt/intel_workarounds.c @@ -567,6 +567,9 @@ static void icl_ctx_workarounds_init(struct intel_engine_cs *engine, static void tgl_ctx_workarounds_init(struct intel_engine_cs *engine, struct i915_wa_list *wal) { + /* Wa_1409142259 */ + WA_SET_BIT_MASKED(GEN11_COMMON_SLICE_CHICKEN3, + GEN12_DISABLE_CPS_AWARE_COLOR_PIPE); } static void @@ -796,11 +799,10 @@ wa_init_mcr(struct drm_i915_private *i915, struct i915_wa_list *wal) } slice = fls(sseu->slice_mask) - 1; - GEM_BUG_ON(slice >= ARRAY_SIZE(sseu->subslice_mask)); - subslice = fls(l3_en & sseu->subslice_mask[slice]); + subslice = fls(l3_en & intel_sseu_get_subslices(sseu, slice)); if (!subslice) { DRM_WARN("No common index found between subslice mask %x and L3 bank mask %x!\n", - sseu->subslice_mask[slice], l3_en); + intel_sseu_get_subslices(sseu, slice), l3_en); subslice = fls(l3_en); WARN_ON(!subslice); } @@ -1063,6 +1065,9 @@ static void gen9_whitelist_build(struct i915_wa_list *w) /* WaAllowUMDToModifyHDCChicken1:skl,bxt,kbl,glk,cfl */ whitelist_reg(w, GEN8_HDC_CHICKEN1); + + /* WaSendPushConstantsFromMMIO:skl,bxt */ + whitelist_reg(w, COMMON_SLICE_CHICKEN2); } static void skl_whitelist_build(struct intel_engine_cs *engine) @@ -1449,7 +1454,7 @@ static bool mcr_range(struct drm_i915_private *i915, u32 offset) * which only controls CPU initiated MMIO. Routing does not * work for CS access so we cannot verify them on this path. */ - if (INTEL_GEN(i915) >= 8 && (offset >= 0xb100 && offset <= 0xb3ff)) + if (INTEL_GEN(i915) >= 8 && (offset >= 0xb000 && offset <= 0xb4ff)) return true; return false; diff --git a/drivers/gpu/drm/i915/gt/selftest_context.c b/drivers/gpu/drm/i915/gt/selftest_context.c index 9d1ea26c7a2d..7c838a57e174 100644 --- a/drivers/gpu/drm/i915/gt/selftest_context.c +++ b/drivers/gpu/drm/i915/gt/selftest_context.c @@ -14,22 +14,28 @@ static int request_sync(struct i915_request *rq) { + struct intel_timeline *tl = i915_request_timeline(rq); long timeout; int err = 0; + intel_timeline_get(tl); i915_request_get(rq); - i915_request_add(rq); + /* Opencode i915_request_add() so we can keep the timeline locked. */ + __i915_request_commit(rq); + __i915_request_queue(rq, NULL); + timeout = i915_request_wait(rq, 0, HZ / 10); - if (timeout < 0) { + if (timeout < 0) err = timeout; - } else { - mutex_lock(&rq->timeline->mutex); + else i915_request_retire_upto(rq); - mutex_unlock(&rq->timeline->mutex); - } + + lockdep_unpin_lock(&tl->mutex, rq->cookie); + mutex_unlock(&tl->mutex); i915_request_put(rq); + intel_timeline_put(tl); return err; } @@ -41,24 +47,20 @@ static int context_sync(struct intel_context *ce) mutex_lock(&tl->mutex); do { - struct i915_request *rq; + struct dma_fence *fence; long timeout; - rcu_read_lock(); - rq = rcu_dereference(tl->last_request.request); - if (rq) - rq = i915_request_get_rcu(rq); - rcu_read_unlock(); - if (!rq) + fence = i915_active_fence_get(&tl->last_request); + if (!fence) break; - timeout = i915_request_wait(rq, 0, HZ / 10); + timeout = dma_fence_wait_timeout(fence, false, HZ / 10); if (timeout < 0) err = timeout; else - i915_request_retire_upto(rq); + i915_request_retire_upto(to_request(fence)); - i915_request_put(rq); + dma_fence_put(fence); } while (!err); mutex_unlock(&tl->mutex); @@ -153,13 +155,9 @@ static int live_context_size(void *arg) * HW tries to write past the end of one. */ - mutex_lock(>->i915->drm.struct_mutex); - fixme = kernel_context(gt->i915); - if (IS_ERR(fixme)) { - err = PTR_ERR(fixme); - goto unlock; - } + if (IS_ERR(fixme)) + return PTR_ERR(fixme); for_each_engine(engine, gt->i915, id) { struct { @@ -199,8 +197,6 @@ static int live_context_size(void *arg) } kernel_context_close(fixme); -unlock: - mutex_unlock(>->i915->drm.struct_mutex); return err; } @@ -303,12 +299,10 @@ static int live_active_context(void *arg) if (IS_ERR(file)) return PTR_ERR(file); - mutex_lock(>->i915->drm.struct_mutex); - fixme = live_context(gt->i915, file); if (IS_ERR(fixme)) { err = PTR_ERR(fixme); - goto unlock; + goto out_file; } for_each_engine(engine, gt->i915, id) { @@ -316,13 +310,12 @@ static int live_active_context(void *arg) if (err) break; - err = igt_flush_test(gt->i915, I915_WAIT_LOCKED); + err = igt_flush_test(gt->i915); if (err) break; } -unlock: - mutex_unlock(>->i915->drm.struct_mutex); +out_file: mock_file_free(gt->i915, file); return err; } @@ -416,12 +409,10 @@ static int live_remote_context(void *arg) if (IS_ERR(file)) return PTR_ERR(file); - mutex_lock(>->i915->drm.struct_mutex); - fixme = live_context(gt->i915, file); if (IS_ERR(fixme)) { err = PTR_ERR(fixme); - goto unlock; + goto out_file; } for_each_engine(engine, gt->i915, id) { @@ -429,13 +420,12 @@ static int live_remote_context(void *arg) if (err) break; - err = igt_flush_test(gt->i915, I915_WAIT_LOCKED); + err = igt_flush_test(gt->i915); if (err) break; } -unlock: - mutex_unlock(>->i915->drm.struct_mutex); +out_file: mock_file_free(gt->i915, file); return err; } diff --git a/drivers/gpu/drm/i915/gt/selftest_gt_pm.c b/drivers/gpu/drm/i915/gt/selftest_gt_pm.c new file mode 100644 index 000000000000..87985bd46423 --- /dev/null +++ b/drivers/gpu/drm/i915/gt/selftest_gt_pm.c @@ -0,0 +1,50 @@ + +/* + * SPDX-License-Identifier: MIT + * + * Copyright © 2019 Intel Corporation + */ + +static int live_gt_resume(void *arg) +{ + struct intel_gt *gt = arg; + IGT_TIMEOUT(end_time); + int err; + + /* Do several suspend/resume cycles to check we don't explode! */ + do { + intel_gt_suspend(gt); + + if (gt->rc6.enabled) { + pr_err("rc6 still enabled after suspend!\n"); + intel_gt_set_wedged_on_init(gt); + err = -EINVAL; + break; + } + + err = intel_gt_resume(gt); + if (err) + break; + + if (gt->rc6.supported && !gt->rc6.enabled) { + pr_err("rc6 not enabled upon resume!\n"); + intel_gt_set_wedged_on_init(gt); + err = -EINVAL; + break; + } + } while (!__igt_timeout(end_time, NULL)); + + return err; +} + +int intel_gt_pm_live_selftests(struct drm_i915_private *i915) +{ + static const struct i915_subtest tests[] = { + SUBTEST(live_gt_resume), + }; + + if (intel_gt_is_wedged(&i915->gt)) + return 0; + + return intel_gt_live_subtests(tests, &i915->gt); +} diff --git a/drivers/gpu/drm/i915/gt/selftest_hangcheck.c b/drivers/gpu/drm/i915/gt/selftest_hangcheck.c index a0098fc35921..e8a40df79bd0 100644 --- a/drivers/gpu/drm/i915/gt/selftest_hangcheck.c +++ b/drivers/gpu/drm/i915/gt/selftest_hangcheck.c @@ -131,7 +131,7 @@ static struct i915_request * hang_create_request(struct hang *h, struct intel_engine_cs *engine) { struct intel_gt *gt = h->gt; - struct i915_address_space *vm = h->ctx->vm ?: &engine->gt->ggtt->vm; + struct i915_address_space *vm = i915_gem_context_get_vm_rcu(h->ctx); struct drm_i915_gem_object *obj; struct i915_request *rq = NULL; struct i915_vma *hws, *vma; @@ -141,12 +141,15 @@ hang_create_request(struct hang *h, struct intel_engine_cs *engine) int err; obj = i915_gem_object_create_internal(gt->i915, PAGE_SIZE); - if (IS_ERR(obj)) + if (IS_ERR(obj)) { + i915_vm_put(vm); return ERR_CAST(obj); + } vaddr = i915_gem_object_pin_map(obj, i915_coherent_map_type(gt->i915)); if (IS_ERR(vaddr)) { i915_gem_object_put(obj); + i915_vm_put(vm); return ERR_CAST(vaddr); } @@ -157,16 +160,22 @@ hang_create_request(struct hang *h, struct intel_engine_cs *engine) h->batch = vaddr; vma = i915_vma_instance(h->obj, vm, NULL); - if (IS_ERR(vma)) + if (IS_ERR(vma)) { + i915_vm_put(vm); return ERR_CAST(vma); + } hws = i915_vma_instance(h->hws, vm, NULL); - if (IS_ERR(hws)) + if (IS_ERR(hws)) { + i915_vm_put(vm); return ERR_CAST(hws); + } err = i915_vma_pin(vma, 0, 0, PIN_USER); - if (err) + if (err) { + i915_vm_put(vm); return ERR_PTR(err); + } err = i915_vma_pin(hws, 0, 0, PIN_USER); if (err) @@ -264,6 +273,7 @@ unpin_hws: i915_vma_unpin(hws); unpin_vma: i915_vma_unpin(vma); + i915_vm_put(vm); return err ? ERR_PTR(err) : rq; } @@ -285,7 +295,7 @@ static void hang_fini(struct hang *h) kernel_context_close(h->ctx); - igt_flush_test(h->gt->i915, I915_WAIT_LOCKED); + igt_flush_test(h->gt->i915); } static bool wait_until_running(struct hang *h, struct i915_request *rq) @@ -309,10 +319,9 @@ 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, gt); if (err) - goto unlock; + return err; for_each_engine(engine, gt->i915, id) { struct intel_wedge_me w; @@ -355,8 +364,6 @@ static int igt_hang_sanitycheck(void *arg) fini: hang_fini(&h); -unlock: - mutex_unlock(>->i915->drm.struct_mutex); return err; } @@ -383,9 +390,7 @@ static int igt_reset_nop(void *arg) if (IS_ERR(file)) return PTR_ERR(file); - 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; @@ -395,8 +400,6 @@ static int igt_reset_nop(void *arg) reset_count = i915_reset_count(global); count = 0; do { - mutex_lock(>->i915->drm.struct_mutex); - for_each_engine(engine, gt->i915, id) { int i; @@ -417,7 +420,6 @@ static int igt_reset_nop(void *arg) intel_gt_reset(gt, ALL_ENGINES, NULL); igt_global_reset_unlock(gt); - mutex_unlock(>->i915->drm.struct_mutex); if (intel_gt_is_wedged(gt)) { err = -EIO; break; @@ -429,16 +431,13 @@ static int igt_reset_nop(void *arg) break; } - err = igt_flush_test(gt->i915, 0); + err = igt_flush_test(gt->i915); 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(gt->i915, I915_WAIT_LOCKED); - mutex_unlock(>->i915->drm.struct_mutex); - + err = igt_flush_test(gt->i915); out: mock_file_free(gt->i915, file); if (intel_gt_is_wedged(gt)) @@ -458,16 +457,14 @@ static int igt_reset_nop_engine(void *arg) /* Check that we can engine-reset during non-user portions */ - if (!intel_has_reset_engine(gt->i915)) + if (!intel_has_reset_engine(gt)) return 0; file = mock_file(gt->i915); if (IS_ERR(file)) return PTR_ERR(file); - 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; @@ -494,7 +491,6 @@ static int igt_reset_nop_engine(void *arg) break; } - mutex_lock(>->i915->drm.struct_mutex); for (i = 0; i < 16; i++) { struct i915_request *rq; @@ -507,7 +503,6 @@ static int igt_reset_nop_engine(void *arg) i915_request_add(rq); } err = intel_engine_reset(engine, NULL); - mutex_unlock(>->i915->drm.struct_mutex); if (err) { pr_err("i915_reset_engine failed\n"); break; @@ -533,15 +528,12 @@ static int igt_reset_nop_engine(void *arg) if (err) break; - err = igt_flush_test(gt->i915, 0); + err = igt_flush_test(gt->i915); if (err) break; } - mutex_lock(>->i915->drm.struct_mutex); - err = igt_flush_test(gt->i915, I915_WAIT_LOCKED); - mutex_unlock(>->i915->drm.struct_mutex); - + err = igt_flush_test(gt->i915); out: mock_file_free(gt->i915, file); if (intel_gt_is_wedged(gt)) @@ -559,13 +551,11 @@ static int __igt_reset_engine(struct intel_gt *gt, bool active) /* Check that we can issue an engine reset on an idle engine (no-op) */ - if (!intel_has_reset_engine(gt->i915)) + if (!intel_has_reset_engine(gt)) return 0; if (active) { - mutex_lock(>->i915->drm.struct_mutex); err = hang_init(&h, gt); - mutex_unlock(>->i915->drm.struct_mutex); if (err) return err; } @@ -593,17 +583,14 @@ static int __igt_reset_engine(struct intel_gt *gt, bool active) if (active) { struct i915_request *rq; - 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); break; } i915_request_get(rq); i915_request_add(rq); - mutex_unlock(>->i915->drm.struct_mutex); if (!wait_until_running(&h, rq)) { struct drm_printer p = drm_info_printer(gt->i915->drm.dev); @@ -647,7 +634,7 @@ static int __igt_reset_engine(struct intel_gt *gt, bool active) if (err) break; - err = igt_flush_test(gt->i915, 0); + err = igt_flush_test(gt->i915); if (err) break; } @@ -655,11 +642,8 @@ static int __igt_reset_engine(struct intel_gt *gt, bool active) if (intel_gt_is_wedged(gt)) err = -EIO; - if (active) { - mutex_lock(>->i915->drm.struct_mutex); + if (active) hang_fini(&h); - mutex_unlock(>->i915->drm.struct_mutex); - } return err; } @@ -725,9 +709,7 @@ static int active_engine(void *data) return PTR_ERR(file); for (count = 0; count < ARRAY_SIZE(ctx); count++) { - mutex_lock(&engine->i915->drm.struct_mutex); ctx[count] = live_context(engine->i915, file); - mutex_unlock(&engine->i915->drm.struct_mutex); if (IS_ERR(ctx[count])) { err = PTR_ERR(ctx[count]); while (--count) @@ -741,10 +723,8 @@ static int active_engine(void *data) struct i915_request *old = rq[idx]; struct i915_request *new; - mutex_lock(&engine->i915->drm.struct_mutex); new = igt_request_alloc(ctx[idx], engine); if (IS_ERR(new)) { - mutex_unlock(&engine->i915->drm.struct_mutex); err = PTR_ERR(new); break; } @@ -755,7 +735,6 @@ static int active_engine(void *data) rq[idx] = i915_request_get(new); i915_request_add(new); - mutex_unlock(&engine->i915->drm.struct_mutex); err = active_request_put(old); if (err) @@ -791,13 +770,11 @@ static int __igt_reset_engines(struct intel_gt *gt, * with any other engine. */ - if (!intel_has_reset_engine(gt->i915)) + if (!intel_has_reset_engine(gt)) return 0; if (flags & TEST_ACTIVE) { - mutex_lock(>->i915->drm.struct_mutex); err = hang_init(&h, gt); - mutex_unlock(>->i915->drm.struct_mutex); if (err) return err; @@ -855,17 +832,14 @@ static int __igt_reset_engines(struct intel_gt *gt, struct i915_request *rq = NULL; if (flags & TEST_ACTIVE) { - 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); break; } i915_request_get(rq); i915_request_add(rq); - mutex_unlock(>->i915->drm.struct_mutex); if (!wait_until_running(&h, rq)) { struct drm_printer p = drm_info_printer(gt->i915->drm.dev); @@ -977,9 +951,7 @@ unwind: if (err) break; - mutex_lock(>->i915->drm.struct_mutex); - err = igt_flush_test(gt->i915, I915_WAIT_LOCKED); - mutex_unlock(>->i915->drm.struct_mutex); + err = igt_flush_test(gt->i915); if (err) break; } @@ -987,11 +959,8 @@ unwind: if (intel_gt_is_wedged(gt)) err = -EIO; - if (flags & TEST_ACTIVE) { - mutex_lock(>->i915->drm.struct_mutex); + if (flags & TEST_ACTIVE) hang_fini(&h); - mutex_unlock(>->i915->drm.struct_mutex); - } return err; } @@ -1061,7 +1030,6 @@ static int igt_reset_wait(void *arg) igt_global_reset_lock(gt); - mutex_lock(>->i915->drm.struct_mutex); err = hang_init(&h, gt); if (err) goto unlock; @@ -1109,7 +1077,6 @@ out_rq: fini: hang_fini(&h); unlock: - mutex_unlock(>->i915->drm.struct_mutex); igt_global_reset_unlock(gt); if (intel_gt_is_wedged(gt)) @@ -1127,15 +1094,14 @@ static int evict_vma(void *data) { struct evict_vma *arg = data; struct i915_address_space *vm = arg->vma->vm; - struct drm_i915_private *i915 = vm->i915; struct drm_mm_node evict = arg->vma->node; int err; complete(&arg->completion); - mutex_lock(&i915->drm.struct_mutex); + mutex_lock(&vm->mutex); err = i915_gem_evict_for_node(vm, &evict, 0); - mutex_unlock(&i915->drm.struct_mutex); + mutex_unlock(&vm->mutex); return err; } @@ -1143,39 +1109,33 @@ static int evict_vma(void *data) static int evict_fence(void *data) { struct evict_vma *arg = data; - struct drm_i915_private *i915 = arg->vma->vm->i915; int err; complete(&arg->completion); - mutex_lock(&i915->drm.struct_mutex); - /* Mark the fence register as dirty to force the mmio update. */ err = i915_gem_object_set_tiling(arg->vma->obj, I915_TILING_Y, 512); if (err) { pr_err("Invalid Y-tiling settings; err:%d\n", err); - goto out_unlock; + return err; } err = i915_vma_pin(arg->vma, 0, 0, PIN_GLOBAL | PIN_MAPPABLE); if (err) { pr_err("Unable to pin vma for Y-tiled fence; err:%d\n", err); - goto out_unlock; + return err; } err = i915_vma_pin_fence(arg->vma); i915_vma_unpin(arg->vma); if (err) { pr_err("Unable to pin Y-tiled fence; err:%d\n", err); - goto out_unlock; + return err; } i915_vma_unpin_fence(arg->vma); -out_unlock: - mutex_unlock(&i915->drm.struct_mutex); - - return err; + return 0; } static int __igt_reset_evict_vma(struct intel_gt *gt, @@ -1196,10 +1156,9 @@ static int __igt_reset_evict_vma(struct intel_gt *gt, /* Check that we can recover an unbind stuck on a hanging request */ - mutex_lock(>->i915->drm.struct_mutex); err = hang_init(&h, gt); if (err) - goto unlock; + return err; obj = i915_gem_object_create_internal(gt->i915, SZ_1M); if (IS_ERR(obj)) { @@ -1262,8 +1221,6 @@ static int __igt_reset_evict_vma(struct intel_gt *gt, if (err) goto out_rq; - mutex_unlock(>->i915->drm.struct_mutex); - if (!wait_until_running(&h, rq)) { struct drm_printer p = drm_info_printer(gt->i915->drm.dev); @@ -1312,16 +1269,12 @@ out_reset: put_task_struct(tsk); } - mutex_lock(>->i915->drm.struct_mutex); out_rq: i915_request_put(rq); out_obj: i915_gem_object_put(obj); fini: hang_fini(&h); -unlock: - mutex_unlock(>->i915->drm.struct_mutex); - if (intel_gt_is_wedged(gt)) return -EIO; @@ -1340,6 +1293,7 @@ static int igt_reset_evict_ppgtt(void *arg) { struct intel_gt *gt = arg; struct i915_gem_context *ctx; + struct i915_address_space *vm; struct drm_file *file; int err; @@ -1347,18 +1301,20 @@ static int igt_reset_evict_ppgtt(void *arg) if (IS_ERR(file)) return PTR_ERR(file); - 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; } err = 0; - if (ctx->vm) /* aliasing == global gtt locking, covered above */ - err = __igt_reset_evict_vma(gt, ctx->vm, + vm = i915_gem_context_get_vm_rcu(ctx); + if (!i915_is_ggtt(vm)) { + /* aliasing == global gtt locking, covered above */ + err = __igt_reset_evict_vma(gt, vm, evict_vma, EXEC_OBJECT_WRITE); + } + i915_vm_put(vm); out: mock_file_free(gt->i915, file); @@ -1403,7 +1359,6 @@ static int igt_reset_queue(void *arg) igt_global_reset_lock(gt); - mutex_lock(>->i915->drm.struct_mutex); err = hang_init(&h, gt); if (err) goto unlock; @@ -1518,7 +1473,7 @@ static int igt_reset_queue(void *arg) i915_request_put(prev); - err = igt_flush_test(gt->i915, I915_WAIT_LOCKED); + err = igt_flush_test(gt->i915); if (err) break; } @@ -1526,7 +1481,6 @@ static int igt_reset_queue(void *arg) fini: hang_fini(&h); unlock: - mutex_unlock(>->i915->drm.struct_mutex); igt_global_reset_unlock(gt); if (intel_gt_is_wedged(gt)) @@ -1547,17 +1501,15 @@ static int igt_handle_error(void *arg) /* Check that we can issue a global GPU and engine reset */ - if (!intel_has_reset_engine(gt->i915)) + if (!intel_has_reset_engine(gt)) return 0; if (!engine || !intel_engine_can_store_dword(engine)) return 0; - mutex_lock(>->i915->drm.struct_mutex); - err = hang_init(&h, gt); if (err) - goto err_unlock; + return err; rq = hang_create_request(&h, engine); if (IS_ERR(rq)) { @@ -1581,8 +1533,6 @@ static int igt_handle_error(void *arg) goto err_request; } - mutex_unlock(>->i915->drm.struct_mutex); - /* Temporarily disable error capture */ error = xchg(&global->first_error, (void *)-1); @@ -1590,8 +1540,6 @@ static int igt_handle_error(void *arg) xchg(&global->first_error, error); - mutex_lock(>->i915->drm.struct_mutex); - if (rq->fence.error != -EIO) { pr_err("Guilty request not identified!\n"); err = -EINVAL; @@ -1602,8 +1550,6 @@ err_request: i915_request_put(rq); err_fini: hang_fini(&h); -err_unlock: - mutex_unlock(>->i915->drm.struct_mutex); return err; } @@ -1689,14 +1635,13 @@ static int igt_reset_engines_atomic(void *arg) /* Check that the engines resets are usable from atomic context */ - if (!intel_has_reset_engine(gt->i915)) + if (!intel_has_reset_engine(gt)) return 0; if (USES_GUC_SUBMISSION(gt->i915)) return 0; 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(gt)) @@ -1716,9 +1661,7 @@ static int igt_reset_engines_atomic(void *arg) out: /* As we poke around the guts, do a full reset before continuing. */ igt_force_reset(gt); - unlock: - mutex_unlock(>->i915->drm.struct_mutex); igt_global_reset_unlock(gt); return err; @@ -1746,7 +1689,7 @@ int intel_hangcheck_live_selftests(struct drm_i915_private *i915) bool saved_hangcheck; int err; - if (!intel_has_gpu_reset(gt->i915)) + if (!intel_has_gpu_reset(gt)) return 0; if (intel_gt_is_wedged(gt)) @@ -1758,10 +1701,6 @@ int intel_hangcheck_live_selftests(struct drm_i915_private *i915) err = intel_gt_live_subtests(tests, gt); - 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); diff --git a/drivers/gpu/drm/i915/gt/selftest_lrc.c b/drivers/gpu/drm/i915/gt/selftest_lrc.c index d791158988d6..393ae5321e1d 100644 --- a/drivers/gpu/drm/i915/gt/selftest_lrc.c +++ b/drivers/gpu/drm/i915/gt/selftest_lrc.c @@ -26,17 +26,13 @@ static int live_sanitycheck(void *arg) struct i915_gem_context *ctx; struct intel_context *ce; struct igt_spinner spin; - intel_wakeref_t wakeref; int err = -ENOMEM; if (!HAS_LOGICAL_RING_CONTEXTS(i915)) return 0; - mutex_lock(&i915->drm.struct_mutex); - wakeref = intel_runtime_pm_get(&i915->runtime_pm); - if (igt_spinner_init(&spin, &i915->gt)) - goto err_unlock; + return -ENOMEM; ctx = kernel_context(i915); if (!ctx) @@ -61,7 +57,7 @@ static int live_sanitycheck(void *arg) } igt_spinner_end(&spin); - if (igt_flush_test(i915, I915_WAIT_LOCKED)) { + if (igt_flush_test(i915)) { err = -EIO; goto err_ctx; } @@ -73,12 +69,187 @@ err_ctx: kernel_context_close(ctx); err_spin: igt_spinner_fini(&spin); + return err; +} + +static int live_unlite_restore(struct drm_i915_private *i915, int prio) +{ + struct intel_engine_cs *engine; + struct i915_gem_context *ctx; + enum intel_engine_id id; + intel_wakeref_t wakeref; + struct igt_spinner spin; + int err = -ENOMEM; + + /* + * Check that we can correctly context switch between 2 instances + * on the same engine from the same parent context. + */ + + mutex_lock(&i915->drm.struct_mutex); + wakeref = intel_runtime_pm_get(&i915->runtime_pm); + + if (igt_spinner_init(&spin, &i915->gt)) + goto err_unlock; + + ctx = kernel_context(i915); + if (!ctx) + goto err_spin; + + err = 0; + for_each_engine(engine, i915, id) { + struct intel_context *ce[2] = {}; + struct i915_request *rq[2]; + struct igt_live_test t; + int n; + + if (prio && !intel_engine_has_preemption(engine)) + continue; + + if (!intel_engine_can_store_dword(engine)) + continue; + + if (igt_live_test_begin(&t, i915, __func__, engine->name)) { + err = -EIO; + break; + } + + for (n = 0; n < ARRAY_SIZE(ce); n++) { + struct intel_context *tmp; + + tmp = intel_context_create(ctx, engine); + if (IS_ERR(tmp)) { + err = PTR_ERR(tmp); + goto err_ce; + } + + err = intel_context_pin(tmp); + if (err) { + intel_context_put(tmp); + goto err_ce; + } + + /* + * Setup the pair of contexts such that if we + * lite-restore using the RING_TAIL from ce[1] it + * will execute garbage from ce[0]->ring. + */ + memset(tmp->ring->vaddr, + POISON_INUSE, /* IPEHR: 0x5a5a5a5a [hung!] */ + tmp->ring->vma->size); + + ce[n] = tmp; + } + GEM_BUG_ON(!ce[1]->ring->size); + intel_ring_reset(ce[1]->ring, ce[1]->ring->size / 2); + + local_bh_disable(); /* appease lockdep */ + __context_pin_acquire(ce[1]); + __execlists_update_reg_state(ce[1], engine); + __context_pin_release(ce[1]); + local_bh_enable(); + + rq[0] = igt_spinner_create_request(&spin, ce[0], MI_ARB_CHECK); + if (IS_ERR(rq[0])) { + err = PTR_ERR(rq[0]); + goto err_ce; + } + + i915_request_get(rq[0]); + i915_request_add(rq[0]); + GEM_BUG_ON(rq[0]->postfix > ce[1]->ring->emit); + + if (!igt_wait_for_spinner(&spin, rq[0])) { + i915_request_put(rq[0]); + goto err_ce; + } + + rq[1] = i915_request_create(ce[1]); + if (IS_ERR(rq[1])) { + err = PTR_ERR(rq[1]); + i915_request_put(rq[0]); + goto err_ce; + } + + if (!prio) { + /* + * Ensure we do the switch to ce[1] on completion. + * + * rq[0] is already submitted, so this should reduce + * to a no-op (a wait on a request on the same engine + * uses the submit fence, not the completion fence), + * but it will install a dependency on rq[1] for rq[0] + * that will prevent the pair being reordered by + * timeslicing. + */ + i915_request_await_dma_fence(rq[1], &rq[0]->fence); + } + + i915_request_get(rq[1]); + i915_request_add(rq[1]); + GEM_BUG_ON(rq[1]->postfix <= rq[0]->postfix); + i915_request_put(rq[0]); + + if (prio) { + struct i915_sched_attr attr = { + .priority = prio, + }; + + /* Alternatively preempt the spinner with ce[1] */ + engine->schedule(rq[1], &attr); + } + + /* And switch back to ce[0] for good measure */ + rq[0] = i915_request_create(ce[0]); + if (IS_ERR(rq[0])) { + err = PTR_ERR(rq[0]); + i915_request_put(rq[1]); + goto err_ce; + } + + i915_request_await_dma_fence(rq[0], &rq[1]->fence); + i915_request_get(rq[0]); + i915_request_add(rq[0]); + GEM_BUG_ON(rq[0]->postfix > rq[1]->postfix); + i915_request_put(rq[1]); + i915_request_put(rq[0]); + +err_ce: + tasklet_kill(&engine->execlists.tasklet); /* flush submission */ + igt_spinner_end(&spin); + for (n = 0; n < ARRAY_SIZE(ce); n++) { + if (IS_ERR_OR_NULL(ce[n])) + break; + + intel_context_unpin(ce[n]); + intel_context_put(ce[n]); + } + + if (igt_live_test_end(&t)) + err = -EIO; + if (err) + break; + } + + kernel_context_close(ctx); +err_spin: + igt_spinner_fini(&spin); err_unlock: intel_runtime_pm_put(&i915->runtime_pm, wakeref); mutex_unlock(&i915->drm.struct_mutex); return err; } +static int live_unlite_switch(void *arg) +{ + return live_unlite_restore(arg, 0); +} + +static int live_unlite_preempt(void *arg) +{ + return live_unlite_restore(arg, I915_USER_PRIORITY(I915_PRIORITY_MAX)); +} + static int emit_semaphore_chain(struct i915_request *rq, struct i915_vma *vma, int idx) { @@ -206,8 +377,7 @@ slice_semaphore_queue(struct intel_engine_cs *outer, if (err) goto out; - if (i915_request_wait(head, - I915_WAIT_LOCKED, + if (i915_request_wait(head, 0, 2 * RUNTIME_INFO(outer->i915)->num_engines * (count + 2) * (count + 3)) < 0) { pr_err("Failed to slice along semaphore chain of length (%d, %d)!\n", count, n); @@ -225,7 +395,6 @@ static int live_timeslice_preempt(void *arg) { struct drm_i915_private *i915 = arg; struct drm_i915_gem_object *obj; - intel_wakeref_t wakeref; struct i915_vma *vma; void *vaddr; int err = 0; @@ -240,14 +409,9 @@ static int live_timeslice_preempt(void *arg) * ready task. */ - mutex_lock(&i915->drm.struct_mutex); - wakeref = intel_runtime_pm_get(&i915->runtime_pm); - obj = i915_gem_object_create_internal(i915, PAGE_SIZE); - if (IS_ERR(obj)) { - err = PTR_ERR(obj); - goto err_unlock; - } + if (IS_ERR(obj)) + return PTR_ERR(obj); vma = i915_vma_instance(obj, &i915->ggtt.vm, NULL); if (IS_ERR(vma)) { @@ -279,7 +443,7 @@ static int live_timeslice_preempt(void *arg) if (err) goto err_pin; - if (igt_flush_test(i915, I915_WAIT_LOCKED)) { + if (igt_flush_test(i915)) { err = -EIO; goto err_pin; } @@ -292,10 +456,6 @@ err_map: i915_gem_object_unpin_map(obj); err_obj: i915_gem_object_put(obj); -err_unlock: - intel_runtime_pm_put(&i915->runtime_pm, wakeref); - mutex_unlock(&i915->drm.struct_mutex); - return err; } @@ -307,7 +467,6 @@ static int live_busywait_preempt(void *arg) struct drm_i915_gem_object *obj; struct i915_vma *vma; enum intel_engine_id id; - intel_wakeref_t wakeref; int err = -ENOMEM; u32 *map; @@ -316,12 +475,9 @@ static int live_busywait_preempt(void *arg) * preempt the busywaits used to synchronise between rings. */ - mutex_lock(&i915->drm.struct_mutex); - wakeref = intel_runtime_pm_get(&i915->runtime_pm); - ctx_hi = kernel_context(i915); if (!ctx_hi) - goto err_unlock; + return -ENOMEM; ctx_hi->sched.priority = I915_USER_PRIORITY(I915_CONTEXT_MAX_USER_PRIORITY); @@ -475,9 +631,6 @@ err_ctx_lo: kernel_context_close(ctx_lo); err_ctx_hi: kernel_context_close(ctx_hi); -err_unlock: - intel_runtime_pm_put(&i915->runtime_pm, wakeref); - mutex_unlock(&i915->drm.struct_mutex); return err; } @@ -506,7 +659,6 @@ static int live_preempt(void *arg) struct igt_spinner spin_hi, spin_lo; struct intel_engine_cs *engine; enum intel_engine_id id; - intel_wakeref_t wakeref; int err = -ENOMEM; if (!HAS_LOGICAL_RING_PREEMPTION(i915)) @@ -515,11 +667,8 @@ static int live_preempt(void *arg) if (!(i915->caps.scheduler & I915_SCHEDULER_CAP_PREEMPTION)) pr_err("Logical preemption supported, but not exposed\n"); - mutex_lock(&i915->drm.struct_mutex); - wakeref = intel_runtime_pm_get(&i915->runtime_pm); - if (igt_spinner_init(&spin_hi, &i915->gt)) - goto err_unlock; + return -ENOMEM; if (igt_spinner_init(&spin_lo, &i915->gt)) goto err_spin_hi; @@ -599,9 +748,6 @@ err_spin_lo: igt_spinner_fini(&spin_lo); err_spin_hi: igt_spinner_fini(&spin_hi); -err_unlock: - intel_runtime_pm_put(&i915->runtime_pm, wakeref); - mutex_unlock(&i915->drm.struct_mutex); return err; } @@ -613,17 +759,13 @@ static int live_late_preempt(void *arg) struct intel_engine_cs *engine; struct i915_sched_attr attr = {}; enum intel_engine_id id; - intel_wakeref_t wakeref; int err = -ENOMEM; if (!HAS_LOGICAL_RING_PREEMPTION(i915)) return 0; - mutex_lock(&i915->drm.struct_mutex); - wakeref = intel_runtime_pm_get(&i915->runtime_pm); - if (igt_spinner_init(&spin_hi, &i915->gt)) - goto err_unlock; + return -ENOMEM; if (igt_spinner_init(&spin_lo, &i915->gt)) goto err_spin_hi; @@ -705,9 +847,6 @@ err_spin_lo: igt_spinner_fini(&spin_lo); err_spin_hi: igt_spinner_fini(&spin_hi); -err_unlock: - intel_runtime_pm_put(&i915->runtime_pm, wakeref); - mutex_unlock(&i915->drm.struct_mutex); return err; err_wedged: @@ -752,7 +891,6 @@ static int live_nopreempt(void *arg) struct intel_engine_cs *engine; struct preempt_client a, b; enum intel_engine_id id; - intel_wakeref_t wakeref; int err = -ENOMEM; /* @@ -763,11 +901,8 @@ static int live_nopreempt(void *arg) if (!HAS_LOGICAL_RING_PREEMPTION(i915)) return 0; - mutex_lock(&i915->drm.struct_mutex); - wakeref = intel_runtime_pm_get(&i915->runtime_pm); - if (preempt_client_init(i915, &a)) - goto err_unlock; + return -ENOMEM; if (preempt_client_init(i915, &b)) goto err_client_a; b.ctx->sched.priority = I915_USER_PRIORITY(I915_PRIORITY_MAX); @@ -832,7 +967,7 @@ static int live_nopreempt(void *arg) goto err_wedged; } - if (igt_flush_test(i915, I915_WAIT_LOCKED)) + if (igt_flush_test(i915)) goto err_wedged; } @@ -841,9 +976,6 @@ err_client_b: preempt_client_fini(&b); err_client_a: preempt_client_fini(&a); -err_unlock: - intel_runtime_pm_put(&i915->runtime_pm, wakeref); - mutex_unlock(&i915->drm.struct_mutex); return err; err_wedged: @@ -863,7 +995,6 @@ static int live_suppress_self_preempt(void *arg) }; struct preempt_client a, b; enum intel_engine_id id; - intel_wakeref_t wakeref; int err = -ENOMEM; /* @@ -882,11 +1013,8 @@ static int live_suppress_self_preempt(void *arg) if (intel_vgpu_active(i915)) return 0; /* GVT forces single port & request submission */ - mutex_lock(&i915->drm.struct_mutex); - wakeref = intel_runtime_pm_get(&i915->runtime_pm); - if (preempt_client_init(i915, &a)) - goto err_unlock; + return -ENOMEM; if (preempt_client_init(i915, &b)) goto err_client_a; @@ -897,6 +1025,10 @@ static int live_suppress_self_preempt(void *arg) if (!intel_engine_has_preemption(engine)) continue; + if (igt_flush_test(i915)) + goto err_wedged; + + intel_engine_pm_get(engine); engine->execlists.preempt_hang.count = 0; rq_a = spinner_create_request(&a.spin, @@ -904,12 +1036,14 @@ static int live_suppress_self_preempt(void *arg) MI_NOOP); if (IS_ERR(rq_a)) { err = PTR_ERR(rq_a); + intel_engine_pm_put(engine); goto err_client_b; } i915_request_add(rq_a); if (!igt_wait_for_spinner(&a.spin, rq_a)) { pr_err("First client failed to start\n"); + intel_engine_pm_put(engine); goto err_wedged; } @@ -921,6 +1055,7 @@ static int live_suppress_self_preempt(void *arg) MI_NOOP); if (IS_ERR(rq_b)) { err = PTR_ERR(rq_b); + intel_engine_pm_put(engine); goto err_client_b; } i915_request_add(rq_b); @@ -931,6 +1066,7 @@ static int live_suppress_self_preempt(void *arg) if (!igt_wait_for_spinner(&b.spin, rq_b)) { pr_err("Second client failed to start\n"); + intel_engine_pm_put(engine); goto err_wedged; } @@ -944,11 +1080,13 @@ static int live_suppress_self_preempt(void *arg) engine->name, engine->execlists.preempt_hang.count, depth); + intel_engine_pm_put(engine); err = -EINVAL; goto err_client_b; } - if (igt_flush_test(i915, I915_WAIT_LOCKED)) + intel_engine_pm_put(engine); + if (igt_flush_test(i915)) goto err_wedged; } @@ -957,9 +1095,6 @@ err_client_b: preempt_client_fini(&b); err_client_a: preempt_client_fini(&a); -err_unlock: - intel_runtime_pm_put(&i915->runtime_pm, wakeref); - mutex_unlock(&i915->drm.struct_mutex); return err; err_wedged: @@ -984,9 +1119,13 @@ static struct i915_request *dummy_request(struct intel_engine_cs *engine) if (!rq) return NULL; - INIT_LIST_HEAD(&rq->active_list); rq->engine = engine; + spin_lock_init(&rq->lock); + INIT_LIST_HEAD(&rq->fence.cb_list); + rq->fence.lock = &rq->lock; + rq->fence.ops = &i915_fence_ops; + i915_sched_node_init(&rq->sched); /* mark this request as permanently incomplete */ @@ -1025,7 +1164,6 @@ static int live_suppress_wait_preempt(void *arg) struct preempt_client client[4]; struct intel_engine_cs *engine; enum intel_engine_id id; - intel_wakeref_t wakeref; int err = -ENOMEM; int i; @@ -1038,11 +1176,8 @@ static int live_suppress_wait_preempt(void *arg) if (!HAS_LOGICAL_RING_PREEMPTION(i915)) return 0; - mutex_lock(&i915->drm.struct_mutex); - wakeref = intel_runtime_pm_get(&i915->runtime_pm); - if (preempt_client_init(i915, &client[0])) /* ELSP[0] */ - goto err_unlock; + return -ENOMEM; if (preempt_client_init(i915, &client[1])) /* ELSP[1] */ goto err_client_0; if (preempt_client_init(i915, &client[2])) /* head of queue */ @@ -1079,8 +1214,8 @@ static int live_suppress_wait_preempt(void *arg) } /* Disable NEWCLIENT promotion */ - __i915_active_request_set(&rq[i]->timeline->last_request, - dummy); + __i915_active_fence_set(&i915_request_timeline(rq[i])->last_request, + &dummy->fence); i915_request_add(rq[i]); } @@ -1105,7 +1240,7 @@ static int live_suppress_wait_preempt(void *arg) for (i = 0; i < ARRAY_SIZE(client); i++) igt_spinner_end(&client[i].spin); - if (igt_flush_test(i915, I915_WAIT_LOCKED)) + if (igt_flush_test(i915)) goto err_wedged; if (engine->execlists.preempt_hang.count) { @@ -1128,9 +1263,6 @@ err_client_1: preempt_client_fini(&client[1]); err_client_0: preempt_client_fini(&client[0]); -err_unlock: - intel_runtime_pm_put(&i915->runtime_pm, wakeref); - mutex_unlock(&i915->drm.struct_mutex); return err; err_wedged: @@ -1147,7 +1279,6 @@ static int live_chain_preempt(void *arg) struct intel_engine_cs *engine; struct preempt_client hi, lo; enum intel_engine_id id; - intel_wakeref_t wakeref; int err = -ENOMEM; /* @@ -1159,11 +1290,8 @@ static int live_chain_preempt(void *arg) if (!HAS_LOGICAL_RING_PREEMPTION(i915)) return 0; - mutex_lock(&i915->drm.struct_mutex); - wakeref = intel_runtime_pm_get(&i915->runtime_pm); - if (preempt_client_init(i915, &hi)) - goto err_unlock; + return -ENOMEM; if (preempt_client_init(i915, &lo)) goto err_client_hi; @@ -1274,9 +1402,6 @@ err_client_lo: preempt_client_fini(&lo); err_client_hi: preempt_client_fini(&hi); -err_unlock: - intel_runtime_pm_put(&i915->runtime_pm, wakeref); - mutex_unlock(&i915->drm.struct_mutex); return err; err_wedged: @@ -1294,20 +1419,16 @@ static int live_preempt_hang(void *arg) struct igt_spinner spin_hi, spin_lo; struct intel_engine_cs *engine; enum intel_engine_id id; - intel_wakeref_t wakeref; int err = -ENOMEM; if (!HAS_LOGICAL_RING_PREEMPTION(i915)) return 0; - if (!intel_has_reset_engine(i915)) + if (!intel_has_reset_engine(&i915->gt)) return 0; - mutex_lock(&i915->drm.struct_mutex); - wakeref = intel_runtime_pm_get(&i915->runtime_pm); - if (igt_spinner_init(&spin_hi, &i915->gt)) - goto err_unlock; + return -ENOMEM; if (igt_spinner_init(&spin_lo, &i915->gt)) goto err_spin_hi; @@ -1384,7 +1505,7 @@ static int live_preempt_hang(void *arg) igt_spinner_end(&spin_hi); igt_spinner_end(&spin_lo); - if (igt_flush_test(i915, I915_WAIT_LOCKED)) { + if (igt_flush_test(i915)) { err = -EIO; goto err_ctx_lo; } @@ -1399,9 +1520,6 @@ err_spin_lo: igt_spinner_fini(&spin_lo); err_spin_hi: igt_spinner_fini(&spin_hi); -err_unlock: - intel_runtime_pm_put(&i915->runtime_pm, wakeref); - mutex_unlock(&i915->drm.struct_mutex); return err; } @@ -1440,7 +1558,11 @@ static int smoke_submit(struct preempt_smoke *smoke, int err = 0; if (batch) { - vma = i915_vma_instance(batch, ctx->vm, NULL); + struct i915_address_space *vm; + + vm = i915_gem_context_get_vm_rcu(ctx); + vma = i915_vma_instance(batch, vm, NULL); + i915_vm_put(vm); if (IS_ERR(vma)) return PTR_ERR(vma); @@ -1489,11 +1611,9 @@ static int smoke_crescendo_thread(void *arg) struct i915_gem_context *ctx = smoke_context(smoke); int err; - mutex_lock(&smoke->i915->drm.struct_mutex); err = smoke_submit(smoke, ctx, count % I915_PRIORITY_MAX, smoke->batch); - mutex_unlock(&smoke->i915->drm.struct_mutex); if (err) return err; @@ -1514,8 +1634,6 @@ static int smoke_crescendo(struct preempt_smoke *smoke, unsigned int flags) unsigned long count; int err = 0; - mutex_unlock(&smoke->i915->drm.struct_mutex); - for_each_engine(engine, smoke->i915, id) { arg[id] = *smoke; arg[id].engine = engine; @@ -1548,8 +1666,6 @@ static int smoke_crescendo(struct preempt_smoke *smoke, unsigned int flags) put_task_struct(tsk[id]); } - mutex_lock(&smoke->i915->drm.struct_mutex); - pr_info("Submitted %lu crescendo:%x requests across %d engines and %d contexts\n", count, flags, RUNTIME_INFO(smoke->i915)->num_engines, smoke->ncontext); @@ -1592,7 +1708,6 @@ static int live_preempt_smoke(void *arg) .ncontext = 1024, }; const unsigned int phase[] = { 0, BATCH }; - intel_wakeref_t wakeref; struct igt_live_test t; int err = -ENOMEM; u32 *cs; @@ -1607,13 +1722,10 @@ static int live_preempt_smoke(void *arg) if (!smoke.contexts) return -ENOMEM; - mutex_lock(&smoke.i915->drm.struct_mutex); - wakeref = intel_runtime_pm_get(&smoke.i915->runtime_pm); - smoke.batch = i915_gem_object_create_internal(smoke.i915, PAGE_SIZE); if (IS_ERR(smoke.batch)) { err = PTR_ERR(smoke.batch); - goto err_unlock; + goto err_free; } cs = i915_gem_object_pin_map(smoke.batch, I915_MAP_WB); @@ -1660,9 +1772,7 @@ err_ctx: err_batch: i915_gem_object_put(smoke.batch); -err_unlock: - intel_runtime_pm_put(&smoke.i915->runtime_pm, wakeref); - mutex_unlock(&smoke.i915->drm.struct_mutex); +err_free: kfree(smoke.contexts); return err; @@ -1781,7 +1891,7 @@ static int nop_virtual_engine(struct drm_i915_private *i915, prime, div64_u64(ktime_to_ns(times[1]), prime)); out: - if (igt_flush_test(i915, I915_WAIT_LOCKED)) + if (igt_flush_test(i915)) err = -EIO; for (nc = 0; nc < nctx; nc++) { @@ -1800,19 +1910,17 @@ static int live_virtual_engine(void *arg) struct intel_gt *gt = &i915->gt; enum intel_engine_id id; unsigned int class, inst; - int err = -ENODEV; + int err; if (USES_GUC_SUBMISSION(i915)) return 0; - mutex_lock(&i915->drm.struct_mutex); - for_each_engine(engine, i915, id) { err = nop_virtual_engine(i915, &engine, 1, 1, 0); if (err) { pr_err("Failed to wrap engine %s: err=%d\n", engine->name, err); - goto out_unlock; + return err; } } @@ -1833,17 +1941,15 @@ static int live_virtual_engine(void *arg) err = nop_virtual_engine(i915, siblings, nsibling, n, 0); if (err) - goto out_unlock; + return err; } err = nop_virtual_engine(i915, siblings, nsibling, n, CHAIN); if (err) - goto out_unlock; + return err; } -out_unlock: - mutex_unlock(&i915->drm.struct_mutex); - return err; + return 0; } static int mask_virtual_engine(struct drm_i915_private *i915, @@ -1922,11 +2028,8 @@ static int mask_virtual_engine(struct drm_i915_private *i915, } err = igt_live_test_end(&t); - if (err) - goto out; - out: - if (igt_flush_test(i915, I915_WAIT_LOCKED)) + if (igt_flush_test(i915)) err = -EIO; for (n = 0; n < nsibling; n++) @@ -1947,13 +2050,11 @@ static int live_virtual_mask(void *arg) struct intel_engine_cs *siblings[MAX_ENGINE_INSTANCE + 1]; struct intel_gt *gt = &i915->gt; unsigned int class, inst; - int err = 0; + int err; if (USES_GUC_SUBMISSION(i915)) return 0; - mutex_lock(&i915->drm.struct_mutex); - for (class = 0; class <= MAX_ENGINE_CLASS; class++) { unsigned int nsibling; @@ -1969,12 +2070,10 @@ static int live_virtual_mask(void *arg) err = mask_virtual_engine(i915, siblings, nsibling); if (err) - goto out_unlock; + return err; } -out_unlock: - mutex_unlock(&i915->drm.struct_mutex); - return err; + return 0; } static int bond_virtual_engine(struct drm_i915_private *i915, @@ -2104,7 +2203,7 @@ static int bond_virtual_engine(struct drm_i915_private *i915, out: for (n = 0; !IS_ERR(rq[n]); n++) i915_request_put(rq[n]); - if (igt_flush_test(i915, I915_WAIT_LOCKED)) + if (igt_flush_test(i915)) err = -EIO; kernel_context_close(ctx); @@ -2125,13 +2224,11 @@ static int live_virtual_bond(void *arg) struct intel_engine_cs *siblings[MAX_ENGINE_INSTANCE + 1]; struct intel_gt *gt = &i915->gt; unsigned int class, inst; - int err = 0; + int err; if (USES_GUC_SUBMISSION(i915)) return 0; - mutex_lock(&i915->drm.struct_mutex); - for (class = 0; class <= MAX_ENGINE_CLASS; class++) { const struct phase *p; int nsibling; @@ -2154,20 +2251,20 @@ static int live_virtual_bond(void *arg) if (err) { pr_err("%s(%s): failed class=%d, nsibling=%d, err=%d\n", __func__, p->name, class, nsibling, err); - goto out_unlock; + return err; } } } -out_unlock: - mutex_unlock(&i915->drm.struct_mutex); - return err; + return 0; } int intel_execlists_live_selftests(struct drm_i915_private *i915) { static const struct i915_subtest tests[] = { SUBTEST(live_sanitycheck), + SUBTEST(live_unlite_switch), + SUBTEST(live_unlite_preempt), SUBTEST(live_timeslice_preempt), SUBTEST(live_busywait_preempt), SUBTEST(live_preempt), @@ -2191,3 +2288,145 @@ int intel_execlists_live_selftests(struct drm_i915_private *i915) return i915_live_subtests(tests, i915); } + +static void hexdump(const void *buf, size_t len) +{ + const size_t rowsize = 8 * sizeof(u32); + const void *prev = NULL; + bool skip = false; + size_t pos; + + for (pos = 0; pos < len; pos += rowsize) { + char line[128]; + + if (prev && !memcmp(prev, buf + pos, rowsize)) { + if (!skip) { + pr_info("*\n"); + skip = true; + } + continue; + } + + WARN_ON_ONCE(hex_dump_to_buffer(buf + pos, len - pos, + rowsize, sizeof(u32), + line, sizeof(line), + false) >= sizeof(line)); + pr_info("[%04zx] %s\n", pos, line); + + prev = buf + pos; + skip = false; + } +} + +static int live_lrc_layout(void *arg) +{ + struct intel_gt *gt = arg; + struct intel_engine_cs *engine; + enum intel_engine_id id; + u32 *mem; + int err; + + /* + * Check the registers offsets we use to create the initial reg state + * match the layout saved by HW. + */ + + mem = kmalloc(PAGE_SIZE, GFP_KERNEL); + if (!mem) + return -ENOMEM; + + err = 0; + for_each_engine(engine, gt->i915, id) { + u32 *hw, *lrc; + int dw; + + if (!engine->default_state) + continue; + + hw = i915_gem_object_pin_map(engine->default_state, + I915_MAP_WB); + if (IS_ERR(hw)) { + err = PTR_ERR(hw); + break; + } + hw += LRC_STATE_PN * PAGE_SIZE / sizeof(*hw); + + lrc = memset(mem, 0, PAGE_SIZE); + execlists_init_reg_state(lrc, + engine->kernel_context, + engine, + engine->kernel_context->ring, + true); + + dw = 0; + do { + u32 lri = hw[dw]; + + if (lri == 0) { + dw++; + continue; + } + + if ((lri & GENMASK(31, 23)) != MI_INSTR(0x22, 0)) { + pr_err("%s: Expected LRI command at dword %d, found %08x\n", + engine->name, dw, lri); + err = -EINVAL; + break; + } + + if (lrc[dw] != lri) { + pr_err("%s: LRI command mismatch at dword %d, expected %08x found %08x\n", + engine->name, dw, lri, lrc[dw]); + err = -EINVAL; + break; + } + + lri &= 0x7f; + lri++; + dw++; + + while (lri) { + if (hw[dw] != lrc[dw]) { + pr_err("%s: Different registers found at dword %d, expected %x, found %x\n", + engine->name, dw, hw[dw], lrc[dw]); + err = -EINVAL; + break; + } + + /* + * Skip over the actual register value as we + * expect that to differ. + */ + dw += 2; + lri -= 2; + } + } while ((lrc[dw] & ~BIT(0)) != MI_BATCH_BUFFER_END); + + if (err) { + pr_info("%s: HW register image:\n", engine->name); + hexdump(hw, PAGE_SIZE); + + pr_info("%s: SW register image:\n", engine->name); + hexdump(lrc, PAGE_SIZE); + } + + i915_gem_object_unpin_map(engine->default_state); + if (err) + break; + } + + kfree(mem); + return err; +} + +int intel_lrc_live_selftests(struct drm_i915_private *i915) +{ + static const struct i915_subtest tests[] = { + SUBTEST(live_lrc_layout), + }; + + if (!HAS_LOGICAL_RING_CONTEXTS(i915)) + return 0; + + return intel_gt_live_subtests(tests, &i915->gt); +} diff --git a/drivers/gpu/drm/i915/gt/selftest_reset.c b/drivers/gpu/drm/i915/gt/selftest_reset.c index 00a4f60cdfd5..d79482db7fe8 100644 --- a/drivers/gpu/drm/i915/gt/selftest_reset.c +++ b/drivers/gpu/drm/i915/gt/selftest_reset.c @@ -112,7 +112,7 @@ static int igt_atomic_engine_reset(void *arg) /* Check that the resets are usable from atomic context */ - if (!intel_has_reset_engine(gt->i915)) + if (!intel_has_reset_engine(gt)) return 0; if (USES_GUC_SUBMISSION(gt->i915)) @@ -170,7 +170,7 @@ int intel_reset_live_selftests(struct drm_i915_private *i915) }; struct intel_gt *gt = &i915->gt; - if (!intel_has_gpu_reset(gt->i915)) + if (!intel_has_gpu_reset(gt)) return 0; if (intel_gt_is_wedged(gt)) diff --git a/drivers/gpu/drm/i915/gt/selftest_timeline.c b/drivers/gpu/drm/i915/gt/selftest_timeline.c index 321481403165..d6df40cdc8a6 100644 --- a/drivers/gpu/drm/i915/gt/selftest_timeline.c +++ b/drivers/gpu/drm/i915/gt/selftest_timeline.c @@ -6,8 +6,9 @@ #include <linux/prime_numbers.h> -#include "gem/i915_gem_pm.h" +#include "intel_engine_pm.h" #include "intel_gt.h" +#include "intel_gt_requests.h" #include "../selftests/i915_random.h" #include "../i915_selftest.h" @@ -136,7 +137,6 @@ static int mock_hwsp_freelist(void *arg) 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) { @@ -149,7 +149,6 @@ static int mock_hwsp_freelist(void *arg) 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); @@ -449,8 +448,6 @@ 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); @@ -461,10 +458,14 @@ tl_write(struct intel_timeline *tl, struct intel_engine_cs *engine, u32 value) if (IS_ERR(rq)) goto out_unpin; + i915_request_get(rq); + err = emit_ggtt_store_dw(rq, tl->hwsp_offset, value); i915_request_add(rq); - if (err) + if (err) { + i915_request_put(rq); rq = ERR_PTR(err); + } out_unpin: intel_timeline_unpin(tl); @@ -500,7 +501,6 @@ static int live_hwsp_engine(void *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; @@ -515,14 +515,13 @@ static int live_hwsp_engine(void *arg) 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; + intel_engine_pm_get(engine); + for (n = 0; n < NUM_TIMELINES; n++) { struct intel_timeline *tl; struct i915_request *rq; @@ -530,22 +529,26 @@ static int live_hwsp_engine(void *arg) tl = checked_intel_timeline_create(i915); if (IS_ERR(tl)) { err = PTR_ERR(tl); - goto out; + break; } rq = tl_write(tl, engine, count); if (IS_ERR(rq)) { intel_timeline_put(tl); err = PTR_ERR(rq); - goto out; + break; } timelines[count++] = tl; + i915_request_put(rq); } + + intel_engine_pm_put(engine); + if (err) + break; } -out: - if (igt_flush_test(i915, I915_WAIT_LOCKED)) + if (igt_flush_test(i915)) err = -EIO; for (n = 0; n < count; n++) { @@ -559,11 +562,7 @@ out: 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 } @@ -575,7 +574,6 @@ static int live_hwsp_alternate(void *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; @@ -591,9 +589,6 @@ static int live_hwsp_alternate(void *arg) 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) { @@ -605,11 +600,14 @@ static int live_hwsp_alternate(void *arg) tl = checked_intel_timeline_create(i915); if (IS_ERR(tl)) { + intel_engine_pm_put(engine); err = PTR_ERR(tl); goto out; } + intel_engine_pm_get(engine); rq = tl_write(tl, engine, count); + intel_engine_pm_put(engine); if (IS_ERR(rq)) { intel_timeline_put(tl); err = PTR_ERR(rq); @@ -617,11 +615,12 @@ static int live_hwsp_alternate(void *arg) } timelines[count++] = tl; + i915_request_put(rq); } } out: - if (igt_flush_test(i915, I915_WAIT_LOCKED)) + if (igt_flush_test(i915)) err = -EIO; for (n = 0; n < count; n++) { @@ -635,11 +634,7 @@ out: 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 } @@ -647,10 +642,10 @@ out: static int live_hwsp_wrap(void *arg) { struct drm_i915_private *i915 = arg; + struct intel_gt *gt = &i915->gt; struct intel_engine_cs *engine; struct intel_timeline *tl; enum intel_engine_id id; - intel_wakeref_t wakeref; int err = 0; /* @@ -658,14 +653,10 @@ static int live_hwsp_wrap(void *arg) * foreign GPU references. */ - mutex_lock(&i915->drm.struct_mutex); - wakeref = intel_runtime_pm_get(&i915->runtime_pm); + tl = intel_timeline_create(gt, NULL); + if (IS_ERR(tl)) + return PTR_ERR(tl); - 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; @@ -673,7 +664,7 @@ static int live_hwsp_wrap(void *arg) if (err) goto out_free; - for_each_engine(engine, i915, id) { + for_each_engine(engine, gt->i915, id) { const u32 *hwsp_seqno[2]; struct i915_request *rq; u32 seqno[2]; @@ -681,7 +672,9 @@ static int live_hwsp_wrap(void *arg) if (!intel_engine_can_store_dword(engine)) continue; + intel_engine_pm_get(engine); rq = i915_request_create(engine->kernel_context); + intel_engine_pm_put(engine); if (IS_ERR(rq)) { err = PTR_ERR(rq); goto out; @@ -743,20 +736,16 @@ static int live_hwsp_wrap(void *arg) goto out; } - i915_retire_requests(i915); /* recycle HWSP */ + intel_gt_retire_requests(gt); /* recycle HWSP */ } out: - if (igt_flush_test(i915, I915_WAIT_LOCKED)) + if (igt_flush_test(i915)) 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; } @@ -765,7 +754,6 @@ 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; @@ -775,9 +763,6 @@ static int live_hwsp_recycle(void *arg) * 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); @@ -785,6 +770,8 @@ static int live_hwsp_recycle(void *arg) if (!intel_engine_can_store_dword(engine)) continue; + intel_engine_pm_get(engine); + do { struct intel_timeline *tl; struct i915_request *rq; @@ -792,21 +779,22 @@ static int live_hwsp_recycle(void *arg) tl = checked_intel_timeline_create(i915); if (IS_ERR(tl)) { err = PTR_ERR(tl); - goto out; + break; } rq = tl_write(tl, engine, count); if (IS_ERR(rq)) { intel_timeline_put(tl); err = PTR_ERR(rq); - goto out; + break; } if (i915_request_wait(rq, 0, HZ / 5) < 0) { pr_err("Wait for timeline writes timed out!\n"); + i915_request_put(rq); intel_timeline_put(tl); err = -EIO; - goto out; + break; } if (*tl->hwsp_seqno != count) { @@ -815,17 +803,18 @@ static int live_hwsp_recycle(void *arg) err = -EINVAL; } + i915_request_put(rq); intel_timeline_put(tl); count++; if (err) - goto out; + break; } while (!__igt_timeout(end_time, NULL)); - } -out: - intel_runtime_pm_put(&i915->runtime_pm, wakeref); - mutex_unlock(&i915->drm.struct_mutex); + intel_engine_pm_put(engine); + if (err) + break; + } return err; } diff --git a/drivers/gpu/drm/i915/gt/selftest_workarounds.c b/drivers/gpu/drm/i915/gt/selftest_workarounds.c index d06d68ac2a3b..95627e80f246 100644 --- a/drivers/gpu/drm/i915/gt/selftest_workarounds.c +++ b/drivers/gpu/drm/i915/gt/selftest_workarounds.c @@ -260,7 +260,6 @@ switch_to_scratch_context(struct intel_engine_cs *engine, rq = igt_spinner_create_request(spin, ce, MI_NOOP); intel_context_put(ce); - kernel_context_close(ctx); if (IS_ERR(rq)) { spin = NULL; @@ -279,6 +278,7 @@ err: if (err && spin) igt_spinner_end(spin); + kernel_context_close(ctx); return err; } @@ -355,6 +355,7 @@ out_ctx: static struct i915_vma *create_batch(struct i915_gem_context *ctx) { struct drm_i915_gem_object *obj; + struct i915_address_space *vm; struct i915_vma *vma; int err; @@ -362,7 +363,9 @@ static struct i915_vma *create_batch(struct i915_gem_context *ctx) if (IS_ERR(obj)) return ERR_CAST(obj); - vma = i915_vma_instance(obj, ctx->vm, NULL); + vm = i915_gem_context_get_vm_rcu(ctx); + vma = i915_vma_instance(obj, vm, NULL); + i915_vm_put(vm); if (IS_ERR(vma)) { err = PTR_ERR(vma); goto err_obj; @@ -463,12 +466,15 @@ static int check_dirty_whitelist(struct i915_gem_context *ctx, 0xffff00ff, 0xffffffff, }; + struct i915_address_space *vm; struct i915_vma *scratch; struct i915_vma *batch; int err = 0, i, v; u32 *cs, *results; - scratch = create_scratch(ctx->vm, 2 * ARRAY_SIZE(values) + 1); + vm = i915_gem_context_get_vm_rcu(ctx); + scratch = create_scratch(vm, 2 * ARRAY_SIZE(values) + 1); + i915_vm_put(vm); if (IS_ERR(scratch)) return PTR_ERR(scratch); @@ -565,6 +571,14 @@ static int check_dirty_whitelist(struct i915_gem_context *ctx, goto err_request; } + i915_vma_lock(batch); + err = i915_request_await_object(rq, batch->obj, false); + if (err == 0) + err = i915_vma_move_to_active(batch, rq, 0); + i915_vma_unlock(batch); + if (err) + goto err_request; + err = engine->emit_bb_start(rq, batch->node.start, PAGE_SIZE, 0); @@ -668,7 +682,7 @@ out_unpin: break; } - if (igt_flush_test(ctx->i915, I915_WAIT_LOCKED)) + if (igt_flush_test(ctx->i915)) err = -EIO; out_batch: i915_vma_unpin_and_release(&batch, 0); @@ -694,9 +708,7 @@ static int live_dirty_whitelist(void *arg) wakeref = intel_runtime_pm_get(&i915->runtime_pm); - mutex_unlock(&i915->drm.struct_mutex); file = mock_file(i915); - mutex_lock(&i915->drm.struct_mutex); if (IS_ERR(file)) { err = PTR_ERR(file); goto out_rpm; @@ -718,9 +730,7 @@ static int live_dirty_whitelist(void *arg) } out_file: - mutex_unlock(&i915->drm.struct_mutex); mock_file_free(i915, file); - mutex_lock(&i915->drm.struct_mutex); out_rpm: intel_runtime_pm_put(&i915->runtime_pm, wakeref); return err; @@ -739,7 +749,7 @@ static int live_reset_whitelist(void *arg) igt_global_reset_lock(&i915->gt); - if (intel_has_reset_engine(i915)) { + if (intel_has_reset_engine(&i915->gt)) { err = check_whitelist_across_reset(engine, do_engine_reset, "engine"); @@ -747,7 +757,7 @@ static int live_reset_whitelist(void *arg) goto out; } - if (intel_has_gpu_reset(i915)) { + if (intel_has_gpu_reset(&i915->gt)) { err = check_whitelist_across_reset(engine, do_device_reset, "device"); @@ -850,6 +860,14 @@ static int scrub_whitelisted_registers(struct i915_gem_context *ctx, goto err_request; } + i915_vma_lock(batch); + err = i915_request_await_object(rq, batch->obj, false); + if (err == 0) + err = i915_vma_move_to_active(batch, rq, 0); + i915_vma_unlock(batch); + if (err) + goto err_request; + /* Perform the writes from an unprivileged "user" batch */ err = engine->emit_bb_start(rq, batch->node.start, 0, 0); @@ -994,6 +1012,7 @@ static int live_isolated_whitelist(void *arg) return 0; for (i = 0; i < ARRAY_SIZE(client); i++) { + struct i915_address_space *vm; struct i915_gem_context *c; c = kernel_context(i915); @@ -1002,22 +1021,27 @@ static int live_isolated_whitelist(void *arg) goto err; } - client[i].scratch[0] = create_scratch(c->vm, 1024); + vm = i915_gem_context_get_vm_rcu(c); + + client[i].scratch[0] = create_scratch(vm, 1024); if (IS_ERR(client[i].scratch[0])) { err = PTR_ERR(client[i].scratch[0]); + i915_vm_put(vm); kernel_context_close(c); goto err; } - client[i].scratch[1] = create_scratch(c->vm, 1024); + client[i].scratch[1] = create_scratch(vm, 1024); if (IS_ERR(client[i].scratch[1])) { err = PTR_ERR(client[i].scratch[1]); i915_vma_unpin_and_release(&client[i].scratch[0], 0); + i915_vm_put(vm); kernel_context_close(c); goto err; } client[i].ctx = c; + i915_vm_put(vm); } for_each_engine(engine, i915, id) { @@ -1074,7 +1098,7 @@ err: kernel_context_close(client[i].ctx); } - if (igt_flush_test(i915, I915_WAIT_LOCKED)) + if (igt_flush_test(i915)) err = -EIO; return err; @@ -1115,7 +1139,7 @@ live_gpu_reset_workarounds(void *arg) struct wa_lists lists; bool ok; - if (!intel_has_gpu_reset(i915)) + if (!intel_has_gpu_reset(&i915->gt)) return 0; ctx = kernel_context(i915); @@ -1162,7 +1186,7 @@ live_engine_reset_workarounds(void *arg) struct wa_lists lists; int ret = 0; - if (!intel_has_reset_engine(i915)) + if (!intel_has_reset_engine(&i915->gt)) return 0; ctx = kernel_context(i915); @@ -1232,7 +1256,7 @@ err: igt_global_reset_unlock(&i915->gt); kernel_context_close(ctx); - igt_flush_test(i915, I915_WAIT_LOCKED); + igt_flush_test(i915); return ret; } @@ -1246,14 +1270,9 @@ int intel_workarounds_live_selftests(struct drm_i915_private *i915) SUBTEST(live_gpu_reset_workarounds), SUBTEST(live_engine_reset_workarounds), }; - int err; if (intel_gt_is_wedged(&i915->gt)) return 0; - mutex_lock(&i915->drm.struct_mutex); - err = i915_subtests(tests, i915); - mutex_unlock(&i915->drm.struct_mutex); - - return err; + 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 index 598170efcaf6..2a77c051f36a 100644 --- a/drivers/gpu/drm/i915/gt/selftests/mock_timeline.c +++ b/drivers/gpu/drm/i915/gt/selftests/mock_timeline.c @@ -15,7 +15,7 @@ void mock_timeline_init(struct intel_timeline *timeline, u64 context) mutex_init(&timeline->mutex); - INIT_ACTIVE_REQUEST(&timeline->last_request, &timeline->mutex); + INIT_ACTIVE_FENCE(&timeline->last_request, &timeline->mutex); INIT_LIST_HEAD(&timeline->requests); i915_syncmap_init(&timeline->sync); diff --git a/drivers/gpu/drm/i915/gt/uc/intel_guc_reg.h b/drivers/gpu/drm/i915/gt/uc/intel_guc_reg.h index edf194d23c6b..1949346e714e 100644 --- a/drivers/gpu/drm/i915/gt/uc/intel_guc_reg.h +++ b/drivers/gpu/drm/i915/gt/uc/intel_guc_reg.h @@ -83,6 +83,9 @@ #define GEN8_GTCR _MMIO(0x4274) #define GEN8_GTCR_INVALIDATE (1<<0) +#define GEN12_GUC_TLB_INV_CR _MMIO(0xcee8) +#define GEN12_GUC_TLB_INV_CR_INVALIDATE (1 << 0) + #define GUC_ARAT_C6DIS _MMIO(0xA178) #define GUC_SHIM_CONTROL _MMIO(0xc064) diff --git a/drivers/gpu/drm/i915/gt/uc/intel_uc.c b/drivers/gpu/drm/i915/gt/uc/intel_uc.c index 71ee7ab035cc..29a9eec60d2e 100644 --- a/drivers/gpu/drm/i915/gt/uc/intel_uc.c +++ b/drivers/gpu/drm/i915/gt/uc/intel_uc.c @@ -224,17 +224,7 @@ static int guc_enable_communication(struct intel_guc *guc) return 0; } -static void guc_stop_communication(struct intel_guc *guc) -{ - intel_guc_ct_stop(&guc->ct); - - guc->send = intel_guc_send_nop; - guc->handler = intel_guc_to_host_event_handler_nop; - - guc_clear_mmio_msg(guc); -} - -static void guc_disable_communication(struct intel_guc *guc) +static void __guc_stop_communication(struct intel_guc *guc) { /* * Events generated during or after CT disable are logged by guc in @@ -247,6 +237,20 @@ static void guc_disable_communication(struct intel_guc *guc) guc->send = intel_guc_send_nop; guc->handler = intel_guc_to_host_event_handler_nop; +} + +static void guc_stop_communication(struct intel_guc *guc) +{ + intel_guc_ct_stop(&guc->ct); + + __guc_stop_communication(guc); + + DRM_INFO("GuC communication stopped\n"); +} + +static void guc_disable_communication(struct intel_guc *guc) +{ + __guc_stop_communication(guc); intel_guc_ct_disable(&guc->ct); @@ -537,7 +541,9 @@ void intel_uc_fini_hw(struct intel_uc *uc) if (intel_uc_supports_guc_submission(uc)) intel_guc_submission_disable(guc); - guc_disable_communication(guc); + if (guc_communication_enabled(guc)) + guc_disable_communication(guc); + __uc_sanitize(uc); } diff --git a/drivers/gpu/drm/i915/gt/uc/intel_uc_fw.c b/drivers/gpu/drm/i915/gt/uc/intel_uc_fw.c index bd22bf11adad..bb4889d2346d 100644 --- a/drivers/gpu/drm/i915/gt/uc/intel_uc_fw.c +++ b/drivers/gpu/drm/i915/gt/uc/intel_uc_fw.c @@ -39,25 +39,27 @@ void intel_uc_fw_change_status(struct intel_uc_fw *uc_fw, * Must be ordered based on platform + revid, from newer to older. */ #define INTEL_UC_FIRMWARE_DEFS(fw_def, guc_def, huc_def) \ - fw_def(ICELAKE, 0, guc_def(icl, 33, 0, 0), huc_def(icl, 8, 4, 3238)) \ - fw_def(COFFEELAKE, 0, guc_def(kbl, 33, 0, 0), huc_def(kbl, 02, 00, 1810)) \ - fw_def(GEMINILAKE, 0, guc_def(glk, 33, 0, 0), huc_def(glk, 03, 01, 2893)) \ - fw_def(KABYLAKE, 0, guc_def(kbl, 33, 0, 0), huc_def(kbl, 02, 00, 1810)) \ - fw_def(BROXTON, 0, guc_def(bxt, 33, 0, 0), huc_def(bxt, 01, 8, 2893)) \ - fw_def(SKYLAKE, 0, guc_def(skl, 33, 0, 0), huc_def(skl, 01, 07, 1398)) - -#define __MAKE_UC_FW_PATH(prefix_, name_, separator_, major_, minor_, patch_) \ + fw_def(ELKHARTLAKE, 0, guc_def(ehl, 33, 0, 4), huc_def(ehl, 9, 0, 0)) \ + fw_def(ICELAKE, 0, guc_def(icl, 33, 0, 0), huc_def(icl, 9, 0, 0)) \ + fw_def(COFFEELAKE, 5, guc_def(cml, 33, 0, 0), huc_def(cml, 4, 0, 0)) \ + fw_def(COFFEELAKE, 0, guc_def(kbl, 33, 0, 0), huc_def(kbl, 4, 0, 0)) \ + fw_def(GEMINILAKE, 0, guc_def(glk, 33, 0, 0), huc_def(glk, 4, 0, 0)) \ + fw_def(KABYLAKE, 0, guc_def(kbl, 33, 0, 0), huc_def(kbl, 4, 0, 0)) \ + fw_def(BROXTON, 0, guc_def(bxt, 33, 0, 0), huc_def(bxt, 2, 0, 0)) \ + fw_def(SKYLAKE, 0, guc_def(skl, 33, 0, 0), huc_def(skl, 2, 0, 0)) + +#define __MAKE_UC_FW_PATH(prefix_, name_, major_, minor_, patch_) \ "i915/" \ __stringify(prefix_) name_ \ - __stringify(major_) separator_ \ - __stringify(minor_) separator_ \ + __stringify(major_) "." \ + __stringify(minor_) "." \ __stringify(patch_) ".bin" #define MAKE_GUC_FW_PATH(prefix_, major_, minor_, patch_) \ - __MAKE_UC_FW_PATH(prefix_, "_guc_", ".", major_, minor_, patch_) + __MAKE_UC_FW_PATH(prefix_, "_guc_", major_, minor_, patch_) #define MAKE_HUC_FW_PATH(prefix_, major_, minor_, bld_num_) \ - __MAKE_UC_FW_PATH(prefix_, "_huc_ver", "_", major_, minor_, bld_num_) + __MAKE_UC_FW_PATH(prefix_, "_huc_", major_, minor_, bld_num_) /* All blobs need to be declared via MODULE_FIRMWARE() */ #define INTEL_UC_MODULE_FW(platform_, revid_, guc_, huc_) \ @@ -337,25 +339,10 @@ int intel_uc_fw_fetch(struct intel_uc_fw *uc_fw, struct drm_i915_private *i915) } /* Get version numbers from the CSS header */ - switch (uc_fw->type) { - case INTEL_UC_FW_TYPE_GUC: - uc_fw->major_ver_found = FIELD_GET(CSS_SW_VERSION_GUC_MAJOR, - css->sw_version); - uc_fw->minor_ver_found = FIELD_GET(CSS_SW_VERSION_GUC_MINOR, - css->sw_version); - break; - - case INTEL_UC_FW_TYPE_HUC: - uc_fw->major_ver_found = FIELD_GET(CSS_SW_VERSION_HUC_MAJOR, - css->sw_version); - uc_fw->minor_ver_found = FIELD_GET(CSS_SW_VERSION_HUC_MINOR, - css->sw_version); - break; - - default: - MISSING_CASE(uc_fw->type); - break; - } + uc_fw->major_ver_found = FIELD_GET(CSS_SW_VERSION_UC_MAJOR, + css->sw_version); + uc_fw->minor_ver_found = FIELD_GET(CSS_SW_VERSION_UC_MINOR, + css->sw_version); if (uc_fw->major_ver_found != uc_fw->major_ver_wanted || uc_fw->minor_ver_found < uc_fw->minor_ver_wanted) { @@ -400,7 +387,7 @@ static u32 uc_fw_ggtt_offset(struct intel_uc_fw *uc_fw, struct i915_ggtt *ggtt) { struct drm_mm_node *node = &ggtt->uc_fw; - GEM_BUG_ON(!node->allocated); + GEM_BUG_ON(!drm_mm_node_allocated(node)); GEM_BUG_ON(upper_32_bits(node->start)); GEM_BUG_ON(upper_32_bits(node->start + node->size - 1)); diff --git a/drivers/gpu/drm/i915/gt/uc/intel_uc_fw_abi.h b/drivers/gpu/drm/i915/gt/uc/intel_uc_fw_abi.h index ae58e8a8c53b..f8f6c91a0df6 100644 --- a/drivers/gpu/drm/i915/gt/uc/intel_uc_fw_abi.h +++ b/drivers/gpu/drm/i915/gt/uc/intel_uc_fw_abi.h @@ -69,11 +69,9 @@ struct uc_css_header { char username[8]; char buildnumber[12]; u32 sw_version; -#define CSS_SW_VERSION_GUC_MAJOR (0xFF << 16) -#define CSS_SW_VERSION_GUC_MINOR (0xFF << 8) -#define CSS_SW_VERSION_GUC_PATCH (0xFF << 0) -#define CSS_SW_VERSION_HUC_MAJOR (0xFFFF << 16) -#define CSS_SW_VERSION_HUC_MINOR (0xFFFF << 0) +#define CSS_SW_VERSION_UC_MAJOR (0xFF << 16) +#define CSS_SW_VERSION_UC_MINOR (0xFF << 8) +#define CSS_SW_VERSION_UC_PATCH (0xFF << 0) u32 reserved[14]; u32 header_info; } __packed; diff --git a/drivers/gpu/drm/i915/gt/uc/selftest_guc.c b/drivers/gpu/drm/i915/gt/uc/selftest_guc.c index bba0eafe1cdb..f927f851aadf 100644 --- a/drivers/gpu/drm/i915/gt/uc/selftest_guc.c +++ b/drivers/gpu/drm/i915/gt/uc/selftest_guc.c @@ -116,7 +116,6 @@ static int igt_guc_clients(void *args) int err = 0; GEM_BUG_ON(!HAS_GT_UC(dev_priv)); - mutex_lock(&dev_priv->drm.struct_mutex); wakeref = intel_runtime_pm_get(&dev_priv->runtime_pm); guc = &dev_priv->gt.uc.guc; @@ -190,7 +189,6 @@ out: guc_clients_enable(guc); unlock: intel_runtime_pm_put(&dev_priv->runtime_pm, wakeref); - mutex_unlock(&dev_priv->drm.struct_mutex); return err; } @@ -208,7 +206,6 @@ static int igt_guc_doorbells(void *arg) u16 db_id; GEM_BUG_ON(!HAS_GT_UC(dev_priv)); - mutex_lock(&dev_priv->drm.struct_mutex); wakeref = intel_runtime_pm_get(&dev_priv->runtime_pm); guc = &dev_priv->gt.uc.guc; @@ -299,7 +296,6 @@ out: } unlock: intel_runtime_pm_put(&dev_priv->runtime_pm, wakeref); - mutex_unlock(&dev_priv->drm.struct_mutex); return err; } |
