1 files changed, 183 insertions, 108 deletions

diff --git a/drivers/gpu/drm/i915/i915_gem_context.c b/drivers/gpu/drm/i915/i915_gem_context.c
index e5acc3916f75..2aedd188473d 100644
--- a/drivers/gpu/drm/i915/i915_gem_context.c
+++ b/drivers/gpu/drm/i915/i915_gem_context.c
@@ -90,6 +90,8 @@
 #include "i915_drv.h"
 #include "i915_trace.h"
 
+#define ALL_L3_SLICES(dev) (1 << NUM_L3_SLICES(dev)) - 1
+
 /* This is a HW constraint. The value below is the largest known requirement
  * I've seen in a spec to date, and that was a workaround for a non-shipping
  * part. It should be safe to decrease this, but it's more future proof as is.
@@ -97,28 +99,27 @@
 #define GEN6_CONTEXT_ALIGN (64<<10)
 #define GEN7_CONTEXT_ALIGN 4096
 
-static size_t get_context_alignment(struct drm_device *dev)
+static size_t get_context_alignment(struct drm_i915_private *dev_priv)
 {
-	if (IS_GEN6(dev))
+	if (IS_GEN6(dev_priv))
 		return GEN6_CONTEXT_ALIGN;
 
 	return GEN7_CONTEXT_ALIGN;
 }
 
-static int get_context_size(struct drm_device *dev)
+static int get_context_size(struct drm_i915_private *dev_priv)
 {
-	struct drm_i915_private *dev_priv = dev->dev_private;
 	int ret;
 	u32 reg;
 
-	switch (INTEL_INFO(dev)->gen) {
+	switch (INTEL_GEN(dev_priv)) {
 	case 6:
 		reg = I915_READ(CXT_SIZE);
 		ret = GEN6_CXT_TOTAL_SIZE(reg) * 64;
 		break;
 	case 7:
 		reg = I915_READ(GEN7_CXT_SIZE);
-		if (IS_HASWELL(dev))
+		if (IS_HASWELL(dev_priv))
 			ret = HSW_CXT_TOTAL_SIZE;
 		else
 			ret = GEN7_CXT_TOTAL_SIZE(reg) * 64;
@@ -169,6 +170,8 @@ void i915_gem_context_free(struct kref *ctx_ref)
 	if (ctx->legacy_hw_ctx.rcs_state)
 		drm_gem_object_unreference(&ctx->legacy_hw_ctx.rcs_state->base);
 	list_del(&ctx->link);
+
+	ida_simple_remove(&ctx->i915->context_hw_ida, ctx->hw_id);
 	kfree(ctx);
 }
 
@@ -178,9 +181,9 @@ i915_gem_alloc_context_obj(struct drm_device *dev, size_t size)
 	struct drm_i915_gem_object *obj;
 	int ret;
 
-	obj = i915_gem_alloc_object(dev, size);
-	if (obj == NULL)
-		return ERR_PTR(-ENOMEM);
+	obj = i915_gem_object_create(dev, size);
+	if (IS_ERR(obj))
+		return obj;
 
 	/*
 	 * Try to make the context utilize L3 as well as LLC.
@@ -209,6 +212,28 @@ i915_gem_alloc_context_obj(struct drm_device *dev, size_t size)
 	return obj;
 }
 
+static int assign_hw_id(struct drm_i915_private *dev_priv, unsigned *out)
+{
+	int ret;
+
+	ret = ida_simple_get(&dev_priv->context_hw_ida,
+			     0, MAX_CONTEXT_HW_ID, GFP_KERNEL);
+	if (ret < 0) {
+		/* Contexts are only released when no longer active.
+		 * Flush any pending retires to hopefully release some
+		 * stale contexts and try again.
+		 */
+		i915_gem_retire_requests(dev_priv);
+		ret = ida_simple_get(&dev_priv->context_hw_ida,
+				     0, MAX_CONTEXT_HW_ID, GFP_KERNEL);
+		if (ret < 0)
+			return ret;
+	}
+
+	*out = ret;
+	return 0;
+}
+
 static struct intel_context *
 __create_hw_context(struct drm_device *dev,
 		    struct drm_i915_file_private *file_priv)
@@ -221,6 +246,12 @@ __create_hw_context(struct drm_device *dev,
 	if (ctx == NULL)
 		return ERR_PTR(-ENOMEM);
 
+	ret = assign_hw_id(dev_priv, &ctx->hw_id);
+	if (ret) {
+		kfree(ctx);
+		return ERR_PTR(ret);
+	}
+
 	kref_init(&ctx->ref);
 	list_add_tail(&ctx->link, &dev_priv->context_list);
 	ctx->i915 = dev_priv;
@@ -249,7 +280,7 @@ __create_hw_context(struct drm_device *dev,
 	/* NB: Mark all slices as needing a remap so that when the context first
 	 * loads it will restore whatever remap state already exists. If there
 	 * is no remap info, it will be a NOP. */
-	ctx->remap_slice = (1 << NUM_L3_SLICES(dev)) - 1;
+	ctx->remap_slice = ALL_L3_SLICES(dev_priv);
 
 	ctx->hang_stats.ban_period_seconds = DRM_I915_CTX_BAN_PERIOD;
 
@@ -288,7 +319,7 @@ i915_gem_create_context(struct drm_device *dev,
 		 * context.
 		 */
 		ret = i915_gem_obj_ggtt_pin(ctx->legacy_hw_ctx.rcs_state,
-					    get_context_alignment(dev), 0);
+					    get_context_alignment(to_i915(dev)), 0);
 		if (ret) {
 			DRM_DEBUG_DRIVER("Couldn't pin %d\n", ret);
 			goto err_destroy;
@@ -336,7 +367,6 @@ static void i915_gem_context_unpin(struct intel_context *ctx,
 void i915_gem_context_reset(struct drm_device *dev)
 {
 	struct drm_i915_private *dev_priv = dev->dev_private;
-	int i;
 
 	if (i915.enable_execlists) {
 		struct intel_context *ctx;
@@ -345,17 +375,7 @@ void i915_gem_context_reset(struct drm_device *dev)
 			intel_lr_context_reset(dev_priv, ctx);
 	}
 
-	for (i = 0; i < I915_NUM_ENGINES; i++) {
-		struct intel_engine_cs *engine = &dev_priv->engine[i];
-
-		if (engine->last_context) {
-			i915_gem_context_unpin(engine->last_context, engine);
-			engine->last_context = NULL;
-		}
-	}
-
-	/* Force the GPU state to be reinitialised on enabling */
-	dev_priv->kernel_context->legacy_hw_ctx.initialized = false;
+	i915_gem_context_lost(dev_priv);
 }
 
 int i915_gem_context_init(struct drm_device *dev)
@@ -368,19 +388,25 @@ int i915_gem_context_init(struct drm_device *dev)
 	if (WARN_ON(dev_priv->kernel_context))
 		return 0;
 
-	if (intel_vgpu_active(dev) && HAS_LOGICAL_RING_CONTEXTS(dev)) {
+	if (intel_vgpu_active(dev_priv) &&
+	    HAS_LOGICAL_RING_CONTEXTS(dev_priv)) {
 		if (!i915.enable_execlists) {
 			DRM_INFO("Only EXECLIST mode is supported in vgpu.\n");
 			return -EINVAL;
 		}
 	}
 
+	/* Using the simple ida interface, the max is limited by sizeof(int) */
+	BUILD_BUG_ON(MAX_CONTEXT_HW_ID > INT_MAX);
+	ida_init(&dev_priv->context_hw_ida);
+
 	if (i915.enable_execlists) {
 		/* NB: intentionally left blank. We will allocate our own
 		 * backing objects as we need them, thank you very much */
 		dev_priv->hw_context_size = 0;
-	} else if (HAS_HW_CONTEXTS(dev)) {
-		dev_priv->hw_context_size = round_up(get_context_size(dev), 4096);
+	} else if (HAS_HW_CONTEXTS(dev_priv)) {
+		dev_priv->hw_context_size =
+			round_up(get_context_size(dev_priv), 4096);
 		if (dev_priv->hw_context_size > (1<<20)) {
 			DRM_DEBUG_DRIVER("Disabling HW Contexts; invalid size %d\n",
 					 dev_priv->hw_context_size);
@@ -403,61 +429,35 @@ int i915_gem_context_init(struct drm_device *dev)
 	return 0;
 }
 
-void i915_gem_context_fini(struct drm_device *dev)
+void i915_gem_context_lost(struct drm_i915_private *dev_priv)
 {
-	struct drm_i915_private *dev_priv = dev->dev_private;
-	struct intel_context *dctx = dev_priv->kernel_context;
-	int i;
-
-	if (dctx->legacy_hw_ctx.rcs_state) {
-		/* The only known way to stop the gpu from accessing the hw context is
-		 * to reset it. Do this as the very last operation to avoid confusing
-		 * other code, leading to spurious errors. */
-		intel_gpu_reset(dev, ALL_ENGINES);
-
-		/* When default context is created and switched to, base object refcount
-		 * will be 2 (+1 from object creation and +1 from do_switch()).
-		 * i915_gem_context_fini() will be called after gpu_idle() has switched
-		 * to default context. So we need to unreference the base object once
-		 * to offset the do_switch part, so that i915_gem_context_unreference()
-		 * can then free the base object correctly. */
-		WARN_ON(!dev_priv->engine[RCS].last_context);
+	struct intel_engine_cs *engine;
 
-		i915_gem_object_ggtt_unpin(dctx->legacy_hw_ctx.rcs_state);
-	}
-
-	for (i = I915_NUM_ENGINES; --i >= 0;) {
-		struct intel_engine_cs *engine = &dev_priv->engine[i];
+	for_each_engine(engine, dev_priv) {
+		if (engine->last_context == NULL)
+			continue;
 
-		if (engine->last_context) {
-			i915_gem_context_unpin(engine->last_context, engine);
-			engine->last_context = NULL;
-		}
+		i915_gem_context_unpin(engine->last_context, engine);
+		engine->last_context = NULL;
 	}
 
-	i915_gem_context_unreference(dctx);
-	dev_priv->kernel_context = NULL;
+	/* Force the GPU state to be reinitialised on enabling */
+	dev_priv->kernel_context->legacy_hw_ctx.initialized = false;
+	dev_priv->kernel_context->remap_slice = ALL_L3_SLICES(dev_priv);
 }
 
-int i915_gem_context_enable(struct drm_i915_gem_request *req)
+void i915_gem_context_fini(struct drm_device *dev)
 {
-	struct intel_engine_cs *engine = req->engine;
-	int ret;
-
-	if (i915.enable_execlists) {
-		if (engine->init_context == NULL)
-			return 0;
+	struct drm_i915_private *dev_priv = dev->dev_private;
+	struct intel_context *dctx = dev_priv->kernel_context;
 
-		ret = engine->init_context(req);
-	} else
-		ret = i915_switch_context(req);
+	if (dctx->legacy_hw_ctx.rcs_state)
+		i915_gem_object_ggtt_unpin(dctx->legacy_hw_ctx.rcs_state);
 
-	if (ret) {
-		DRM_ERROR("ring init context: %d\n", ret);
-		return ret;
-	}
+	i915_gem_context_unreference(dctx);
+	dev_priv->kernel_context = NULL;
 
-	return 0;
+	ida_destroy(&dev_priv->context_hw_ida);
 }
 
 static int context_idr_cleanup(int id, void *p, void *data)
@@ -510,12 +510,13 @@ i915_gem_context_get(struct drm_i915_file_private *file_priv, u32 id)
 static inline int
 mi_set_context(struct drm_i915_gem_request *req, u32 hw_flags)
 {
+	struct drm_i915_private *dev_priv = req->i915;
 	struct intel_engine_cs *engine = req->engine;
 	u32 flags = hw_flags | MI_MM_SPACE_GTT;
 	const int num_rings =
 		/* Use an extended w/a on ivb+ if signalling from other rings */
-		i915_semaphore_is_enabled(engine->dev) ?
-		hweight32(INTEL_INFO(engine->dev)->ring_mask) - 1 :
+		i915_semaphore_is_enabled(dev_priv) ?
+		hweight32(INTEL_INFO(dev_priv)->ring_mask) - 1 :
 		0;
 	int len, ret;
 
@@ -524,21 +525,21 @@ mi_set_context(struct drm_i915_gem_request *req, u32 hw_flags)
 	 * explicitly, so we rely on the value at ring init, stored in
 	 * itlb_before_ctx_switch.
 	 */
-	if (IS_GEN6(engine->dev)) {
+	if (IS_GEN6(dev_priv)) {
 		ret = engine->flush(req, I915_GEM_GPU_DOMAINS, 0);
 		if (ret)
 			return ret;
 	}
 
 	/* These flags are for resource streamer on HSW+ */
-	if (IS_HASWELL(engine->dev) || INTEL_INFO(engine->dev)->gen >= 8)
+	if (IS_HASWELL(dev_priv) || INTEL_GEN(dev_priv) >= 8)
 		flags |= (HSW_MI_RS_SAVE_STATE_EN | HSW_MI_RS_RESTORE_STATE_EN);
-	else if (INTEL_INFO(engine->dev)->gen < 8)
+	else if (INTEL_GEN(dev_priv) < 8)
 		flags |= (MI_SAVE_EXT_STATE_EN | MI_RESTORE_EXT_STATE_EN);
 
 
 	len = 4;
-	if (INTEL_INFO(engine->dev)->gen >= 7)
+	if (INTEL_GEN(dev_priv) >= 7)
 		len += 2 + (num_rings ? 4*num_rings + 6 : 0);
 
 	ret = intel_ring_begin(req, len);
@@ -546,14 +547,14 @@ mi_set_context(struct drm_i915_gem_request *req, u32 hw_flags)
 		return ret;
 
 	/* WaProgramMiArbOnOffAroundMiSetContext:ivb,vlv,hsw,bdw,chv */
-	if (INTEL_INFO(engine->dev)->gen >= 7) {
+	if (INTEL_GEN(dev_priv) >= 7) {
 		intel_ring_emit(engine, MI_ARB_ON_OFF | MI_ARB_DISABLE);
 		if (num_rings) {
 			struct intel_engine_cs *signaller;
 
 			intel_ring_emit(engine,
 					MI_LOAD_REGISTER_IMM(num_rings));
-			for_each_engine(signaller, to_i915(engine->dev)) {
+			for_each_engine(signaller, dev_priv) {
 				if (signaller == engine)
 					continue;
 
@@ -576,14 +577,14 @@ mi_set_context(struct drm_i915_gem_request *req, u32 hw_flags)
 	 */
 	intel_ring_emit(engine, MI_NOOP);
 
-	if (INTEL_INFO(engine->dev)->gen >= 7) {
+	if (INTEL_GEN(dev_priv) >= 7) {
 		if (num_rings) {
 			struct intel_engine_cs *signaller;
 			i915_reg_t last_reg = {}; /* keep gcc quiet */
 
 			intel_ring_emit(engine,
 					MI_LOAD_REGISTER_IMM(num_rings));
-			for_each_engine(signaller, to_i915(engine->dev)) {
+			for_each_engine(signaller, dev_priv) {
 				if (signaller == engine)
 					continue;
 
@@ -609,7 +610,37 @@ mi_set_context(struct drm_i915_gem_request *req, u32 hw_flags)
 	return ret;
 }
 
-static inline bool skip_rcs_switch(struct intel_engine_cs *engine,
+static int remap_l3(struct drm_i915_gem_request *req, int slice)
+{
+	u32 *remap_info = req->i915->l3_parity.remap_info[slice];
+	struct intel_engine_cs *engine = req->engine;
+	int i, ret;
+
+	if (!remap_info)
+		return 0;
+
+	ret = intel_ring_begin(req, GEN7_L3LOG_SIZE/4 * 2 + 2);
+	if (ret)
+		return ret;
+
+	/*
+	 * Note: We do not worry about the concurrent register cacheline hang
+	 * here because no other code should access these registers other than
+	 * at initialization time.
+	 */
+	intel_ring_emit(engine, MI_LOAD_REGISTER_IMM(GEN7_L3LOG_SIZE/4));
+	for (i = 0; i < GEN7_L3LOG_SIZE/4; i++) {
+		intel_ring_emit_reg(engine, GEN7_L3LOG(slice, i));
+		intel_ring_emit(engine, remap_info[i]);
+	}
+	intel_ring_emit(engine, MI_NOOP);
+	intel_ring_advance(engine);
+
+	return 0;
+}
+
+static inline bool skip_rcs_switch(struct i915_hw_ppgtt *ppgtt,
+				   struct intel_engine_cs *engine,
 				   struct intel_context *to)
 {
 	if (to->remap_slice)
@@ -618,36 +649,44 @@ static inline bool skip_rcs_switch(struct intel_engine_cs *engine,
 	if (!to->legacy_hw_ctx.initialized)
 		return false;
 
-	if (to->ppgtt &&
-	    !(intel_engine_flag(engine) & to->ppgtt->pd_dirty_rings))
+	if (ppgtt && (intel_engine_flag(engine) & ppgtt->pd_dirty_rings))
 		return false;
 
 	return to == engine->last_context;
 }
 
 static bool
-needs_pd_load_pre(struct intel_engine_cs *engine, struct intel_context *to)
+needs_pd_load_pre(struct i915_hw_ppgtt *ppgtt,
+		  struct intel_engine_cs *engine,
+		  struct intel_context *to)
 {
-	if (!to->ppgtt)
+	if (!ppgtt)
 		return false;
 
+	/* Always load the ppgtt on first use */
+	if (!engine->last_context)
+		return true;
+
+	/* Same context without new entries, skip */
 	if (engine->last_context == to &&
-	    !(intel_engine_flag(engine) & to->ppgtt->pd_dirty_rings))
+	    !(intel_engine_flag(engine) & ppgtt->pd_dirty_rings))
 		return false;
 
 	if (engine->id != RCS)
 		return true;
 
-	if (INTEL_INFO(engine->dev)->gen < 8)
+	if (INTEL_GEN(engine->i915) < 8)
 		return true;
 
 	return false;
 }
 
 static bool
-needs_pd_load_post(struct intel_context *to, u32 hw_flags)
+needs_pd_load_post(struct i915_hw_ppgtt *ppgtt,
+		   struct intel_context *to,
+		   u32 hw_flags)
 {
-	if (!to->ppgtt)
+	if (!ppgtt)
 		return false;
 
 	if (!IS_GEN8(to->i915))
@@ -663,16 +702,17 @@ static int do_rcs_switch(struct drm_i915_gem_request *req)
 {
 	struct intel_context *to = req->ctx;
 	struct intel_engine_cs *engine = req->engine;
+	struct i915_hw_ppgtt *ppgtt = to->ppgtt ?: req->i915->mm.aliasing_ppgtt;
 	struct intel_context *from;
 	u32 hw_flags;
 	int ret, i;
 
-	if (skip_rcs_switch(engine, to))
+	if (skip_rcs_switch(ppgtt, engine, to))
 		return 0;
 
 	/* Trying to pin first makes error handling easier. */
 	ret = i915_gem_obj_ggtt_pin(to->legacy_hw_ctx.rcs_state,
-				    get_context_alignment(engine->dev),
+				    get_context_alignment(engine->i915),
 				    0);
 	if (ret)
 		return ret;
@@ -698,13 +738,13 @@ static int do_rcs_switch(struct drm_i915_gem_request *req)
 	if (ret)
 		goto unpin_out;
 
-	if (needs_pd_load_pre(engine, to)) {
+	if (needs_pd_load_pre(ppgtt, engine, to)) {
 		/* Older GENs and non render rings still want the load first,
 		 * "PP_DCLV followed by PP_DIR_BASE register through Load
 		 * Register Immediate commands in Ring Buffer before submitting
 		 * a context."*/
 		trace_switch_mm(engine, to);
-		ret = to->ppgtt->switch_mm(to->ppgtt, req);
+		ret = ppgtt->switch_mm(ppgtt, req);
 		if (ret)
 			goto unpin_out;
 	}
@@ -715,16 +755,11 @@ static int do_rcs_switch(struct drm_i915_gem_request *req)
 		 * space. This means we must enforce that a page table load
 		 * occur when this occurs. */
 		hw_flags = MI_RESTORE_INHIBIT;
-	else if (to->ppgtt &&
-		 intel_engine_flag(engine) & to->ppgtt->pd_dirty_rings)
+	else if (ppgtt && intel_engine_flag(engine) & ppgtt->pd_dirty_rings)
 		hw_flags = MI_FORCE_RESTORE;
 	else
 		hw_flags = 0;
 
-	/* We should never emit switch_mm more than once */
-	WARN_ON(needs_pd_load_pre(engine, to) &&
-		needs_pd_load_post(to, hw_flags));
-
 	if (to != from || (hw_flags & MI_FORCE_RESTORE)) {
 		ret = mi_set_context(req, hw_flags);
 		if (ret)
@@ -759,9 +794,9 @@ static int do_rcs_switch(struct drm_i915_gem_request *req)
 	/* GEN8 does *not* require an explicit reload if the PDPs have been
 	 * setup, and we do not wish to move them.
 	 */
-	if (needs_pd_load_post(to, hw_flags)) {
+	if (needs_pd_load_post(ppgtt, to, hw_flags)) {
 		trace_switch_mm(engine, to);
-		ret = to->ppgtt->switch_mm(to->ppgtt, req);
+		ret = ppgtt->switch_mm(ppgtt, req);
 		/* The hardware context switch is emitted, but we haven't
 		 * actually changed the state - so it's probably safe to bail
 		 * here. Still, let the user know something dangerous has
@@ -771,14 +806,14 @@ static int do_rcs_switch(struct drm_i915_gem_request *req)
 			return ret;
 	}
 
-	if (to->ppgtt)
-		to->ppgtt->pd_dirty_rings &= ~intel_engine_flag(engine);
+	if (ppgtt)
+		ppgtt->pd_dirty_rings &= ~intel_engine_flag(engine);
 
 	for (i = 0; i < MAX_L3_SLICES; i++) {
 		if (!(to->remap_slice & (1<<i)))
 			continue;
 
-		ret = i915_gem_l3_remap(req, i);
+		ret = remap_l3(req, i);
 		if (ret)
 			return ret;
 
@@ -825,17 +860,18 @@ int i915_switch_context(struct drm_i915_gem_request *req)
 	if (engine->id != RCS ||
 	    req->ctx->legacy_hw_ctx.rcs_state == NULL) {
 		struct intel_context *to = req->ctx;
+		struct i915_hw_ppgtt *ppgtt =
+			to->ppgtt ?: req->i915->mm.aliasing_ppgtt;
 
-		if (needs_pd_load_pre(engine, to)) {
+		if (needs_pd_load_pre(ppgtt, engine, to)) {
 			int ret;
 
 			trace_switch_mm(engine, to);
-			ret = to->ppgtt->switch_mm(to->ppgtt, req);
+			ret = ppgtt->switch_mm(ppgtt, req);
 			if (ret)
 				return ret;
 
-			/* Doing a PD load always reloads the page dirs */
-			to->ppgtt->pd_dirty_rings &= ~intel_engine_flag(engine);
+			ppgtt->pd_dirty_rings &= ~intel_engine_flag(engine);
 		}
 
 		if (to != engine->last_context) {
@@ -1004,3 +1040,42 @@ int i915_gem_context_setparam_ioctl(struct drm_device *dev, void *data,
 
 	return ret;
 }
+
+int i915_gem_context_reset_stats_ioctl(struct drm_device *dev,
+				       void *data, struct drm_file *file)
+{
+	struct drm_i915_private *dev_priv = dev->dev_private;
+	struct drm_i915_reset_stats *args = data;
+	struct i915_ctx_hang_stats *hs;
+	struct intel_context *ctx;
+	int ret;
+
+	if (args->flags || args->pad)
+		return -EINVAL;
+
+	if (args->ctx_id == DEFAULT_CONTEXT_HANDLE && !capable(CAP_SYS_ADMIN))
+		return -EPERM;
+
+	ret = i915_mutex_lock_interruptible(dev);
+	if (ret)
+		return ret;
+
+	ctx = i915_gem_context_get(file->driver_priv, args->ctx_id);
+	if (IS_ERR(ctx)) {
+		mutex_unlock(&dev->struct_mutex);
+		return PTR_ERR(ctx);
+	}
+	hs = &ctx->hang_stats;
+
+	if (capable(CAP_SYS_ADMIN))
+		args->reset_count = i915_reset_count(&dev_priv->gpu_error);
+	else
+		args->reset_count = 0;
+
+	args->batch_active = hs->batch_active;
+	args->batch_pending = hs->batch_pending;
+
+	mutex_unlock(&dev->struct_mutex);
+
+	return 0;
+}