Merge tag 'drm-intel-next-fixes-2015-06-18' of...

Merge tag 'drm-intel-next-fixes-2015-06-18' of git://anongit.freedesktop.org/drm-intel into drm-next

i915 fixes for stuff in next

* tag 'drm-intel-next-fixes-2015-06-18' of git://anongit.freedesktop.org/drm-intel:
  drm/i915: Don't set enabled value of all CRTCs when restoring the mode
  drm/i915: Don't update staged config during force restore modesets
  drm/i915: Don't check modeset state in the hw state force restore path
  drm/i915: Add SCRATCH1 and ROW_CHICKEN3 to the register whitelist.
  drm/i915: Extend the parser to check register writes against a mask/value pair.
  drm/i915: Fix command parser to validate multiple register access with the same command.
  drm/i915: Don't skip request retirement if the active list is empty

drivers/gpu/drm/i915/i915_cmd_parser.c

@@ -123,7 +123,7 @@ static const struct drm_i915_cmd_descriptor common_cmds[] = {
 	CMD(  MI_SEMAPHORE_MBOX,                SMI,   !F,  0xFF,   R  ),
 	CMD(  MI_STORE_DWORD_INDEX,             SMI,   !F,  0xFF,   R  ),
 	CMD(  MI_LOAD_REGISTER_IMM(1),          SMI,   !F,  0xFF,   W,
-	      .reg = { .offset = 1, .mask = 0x007FFFFC }               ),
+	      .reg = { .offset = 1, .mask = 0x007FFFFC, .step = 2 }    ),
 	CMD(  MI_STORE_REGISTER_MEM(1),         SMI,   !F,  0xFF,   W | B,
 	      .reg = { .offset = 1, .mask = 0x007FFFFC },
 	      .bits = {{
@@ -395,16 +395,38 @@ static const struct drm_i915_cmd_table hsw_blt_ring_cmds[] = {
 
 /*
  * Register whitelists, sorted by increasing register offset.
+ */
+
+/*
+ * An individual whitelist entry granting access to register addr.  If
+ * mask is non-zero the argument of immediate register writes will be
+ * AND-ed with mask, and the command will be rejected if the result
+ * doesn't match value.
+ *
+ * Registers with non-zero mask are only allowed to be written using
+ * LRI.
+ */
+struct drm_i915_reg_descriptor {
+	u32 addr;
+	u32 mask;
+	u32 value;
+};
+
+/* Convenience macro for adding 32-bit registers. */
+#define REG32(address, ...)                             \
+	{ .addr = address, __VA_ARGS__ }
+
+/*
+ * Convenience macro for adding 64-bit registers.
  *
  * Some registers that userspace accesses are 64 bits. The register
  * access commands only allow 32-bit accesses. Hence, we have to include
  * entries for both halves of the 64-bit registers.
  */
+#define REG64(addr)                                     \
+	REG32(addr), REG32(addr + sizeof(u32))
 
-/* Convenience macro for adding 64-bit registers */
-#define REG64(addr) (addr), (addr + sizeof(u32))
-
-static const u32 gen7_render_regs[] = {
+static const struct drm_i915_reg_descriptor gen7_render_regs[] = {
 	REG64(GPGPU_THREADS_DISPATCHED),
 	REG64(HS_INVOCATION_COUNT),
 	REG64(DS_INVOCATION_COUNT),
@@ -417,15 +439,15 @@ static const u32 gen7_render_regs[] = {
 	REG64(CL_PRIMITIVES_COUNT),
 	REG64(PS_INVOCATION_COUNT),
 	REG64(PS_DEPTH_COUNT),
-	OACONTROL, /* Only allowed for LRI and SRM. See below. */
+	REG32(OACONTROL), /* Only allowed for LRI and SRM. See below. */
 	REG64(MI_PREDICATE_SRC0),
 	REG64(MI_PREDICATE_SRC1),
-	GEN7_3DPRIM_END_OFFSET,
-	GEN7_3DPRIM_START_VERTEX,
-	GEN7_3DPRIM_VERTEX_COUNT,
-	GEN7_3DPRIM_INSTANCE_COUNT,
-	GEN7_3DPRIM_START_INSTANCE,
-	GEN7_3DPRIM_BASE_VERTEX,
+	REG32(GEN7_3DPRIM_END_OFFSET),
+	REG32(GEN7_3DPRIM_START_VERTEX),
+	REG32(GEN7_3DPRIM_VERTEX_COUNT),
+	REG32(GEN7_3DPRIM_INSTANCE_COUNT),
+	REG32(GEN7_3DPRIM_START_INSTANCE),
+	REG32(GEN7_3DPRIM_BASE_VERTEX),
 	REG64(GEN7_SO_NUM_PRIMS_WRITTEN(0)),
 	REG64(GEN7_SO_NUM_PRIMS_WRITTEN(1)),
 	REG64(GEN7_SO_NUM_PRIMS_WRITTEN(2)),
@@ -434,33 +456,41 @@ static const u32 gen7_render_regs[] = {
 	REG64(GEN7_SO_PRIM_STORAGE_NEEDED(1)),
 	REG64(GEN7_SO_PRIM_STORAGE_NEEDED(2)),
 	REG64(GEN7_SO_PRIM_STORAGE_NEEDED(3)),
-	GEN7_SO_WRITE_OFFSET(0),
-	GEN7_SO_WRITE_OFFSET(1),
-	GEN7_SO_WRITE_OFFSET(2),
-	GEN7_SO_WRITE_OFFSET(3),
-	GEN7_L3SQCREG1,
-	GEN7_L3CNTLREG2,
-	GEN7_L3CNTLREG3,
+	REG32(GEN7_SO_WRITE_OFFSET(0)),
+	REG32(GEN7_SO_WRITE_OFFSET(1)),
+	REG32(GEN7_SO_WRITE_OFFSET(2)),
+	REG32(GEN7_SO_WRITE_OFFSET(3)),
+	REG32(GEN7_L3SQCREG1),
+	REG32(GEN7_L3CNTLREG2),
+	REG32(GEN7_L3CNTLREG3),
+	REG32(HSW_SCRATCH1,
+	      .mask = ~HSW_SCRATCH1_L3_DATA_ATOMICS_DISABLE,
+	      .value = 0),
+	REG32(HSW_ROW_CHICKEN3,
+	      .mask = ~(HSW_ROW_CHICKEN3_L3_GLOBAL_ATOMICS_DISABLE << 16 |
+                        HSW_ROW_CHICKEN3_L3_GLOBAL_ATOMICS_DISABLE),
+	      .value = 0),
 };
 
-static const u32 gen7_blt_regs[] = {
-	BCS_SWCTRL,
+static const struct drm_i915_reg_descriptor gen7_blt_regs[] = {
+	REG32(BCS_SWCTRL),
 };
 
-static const u32 ivb_master_regs[] = {
-	FORCEWAKE_MT,
-	DERRMR,
-	GEN7_PIPE_DE_LOAD_SL(PIPE_A),
-	GEN7_PIPE_DE_LOAD_SL(PIPE_B),
-	GEN7_PIPE_DE_LOAD_SL(PIPE_C),
+static const struct drm_i915_reg_descriptor ivb_master_regs[] = {
+	REG32(FORCEWAKE_MT),
+	REG32(DERRMR),
+	REG32(GEN7_PIPE_DE_LOAD_SL(PIPE_A)),
+	REG32(GEN7_PIPE_DE_LOAD_SL(PIPE_B)),
+	REG32(GEN7_PIPE_DE_LOAD_SL(PIPE_C)),
 };
 
-static const u32 hsw_master_regs[] = {
-	FORCEWAKE_MT,
-	DERRMR,
+static const struct drm_i915_reg_descriptor hsw_master_regs[] = {
+	REG32(FORCEWAKE_MT),
+	REG32(DERRMR),
 };
 
 #undef REG64
+#undef REG32
 
 static u32 gen7_render_get_cmd_length_mask(u32 cmd_header)
 {
@@ -550,14 +580,16 @@ static bool validate_cmds_sorted(struct intel_engine_cs *ring,
 	return ret;
 }
 
-static bool check_sorted(int ring_id, const u32 *reg_table, int reg_count)
+static bool check_sorted(int ring_id,
+			 const struct drm_i915_reg_descriptor *reg_table,
+			 int reg_count)
 {
 	int i;
 	u32 previous = 0;
 	bool ret = true;
 
 	for (i = 0; i < reg_count; i++) {
-		u32 curr = reg_table[i];
+		u32 curr = reg_table[i].addr;
 
 		if (curr < previous) {
 			DRM_ERROR("CMD: table not sorted ring=%d entry=%d reg=0x%08X prev=0x%08X\n",
@@ -804,18 +836,20 @@ find_cmd(struct intel_engine_cs *ring,
 	return default_desc;
 }
 
-static bool valid_reg(const u32 *table, int count, u32 addr)
+static const struct drm_i915_reg_descriptor *
+find_reg(const struct drm_i915_reg_descriptor *table,
+	 int count, u32 addr)
 {
-	if (table && count != 0) {
+	if (table) {
 		int i;
 
 		for (i = 0; i < count; i++) {
-			if (table[i] == addr)
-				return true;
+			if (table[i].addr == addr)
+				return &table[i];
 		}
 	}
 
-	return false;
+	return NULL;
 }
 
 static u32 *vmap_batch(struct drm_i915_gem_object *obj,
@@ -934,7 +968,7 @@ bool i915_needs_cmd_parser(struct intel_engine_cs *ring)
 
 static bool check_cmd(const struct intel_engine_cs *ring,
 		      const struct drm_i915_cmd_descriptor *desc,
-		      const u32 *cmd,
+		      const u32 *cmd, u32 length,
 		      const bool is_master,
 		      bool *oacontrol_set)
 {
@@ -950,16 +984,41 @@ static bool check_cmd(const struct intel_engine_cs *ring,
 	}
 
 	if (desc->flags & CMD_DESC_REGISTER) {
-		u32 reg_addr = cmd[desc->reg.offset] & desc->reg.mask;
+		/*
+		 * Get the distance between individual register offset
+		 * fields if the command can perform more than one
+		 * access at a time.
+		 */
+		const u32 step = desc->reg.step ? desc->reg.step : length;
+		u32 offset;
+
+		for (offset = desc->reg.offset; offset < length;
+		     offset += step) {
+			const u32 reg_addr = cmd[offset] & desc->reg.mask;
+			const struct drm_i915_reg_descriptor *reg =
+				find_reg(ring->reg_table, ring->reg_count,
+					 reg_addr);
+
+			if (!reg && is_master)
+				reg = find_reg(ring->master_reg_table,
+					       ring->master_reg_count,
+					       reg_addr);
+
+			if (!reg) {
+				DRM_DEBUG_DRIVER("CMD: Rejected register 0x%08X in command: 0x%08X (ring=%d)\n",
+						 reg_addr, *cmd, ring->id);
+				return false;
+			}
 
 			/*
-		 * OACONTROL requires some special handling for writes. We
-		 * want to make sure that any batch which enables OA also
-		 * disables it before the end of the batch. The goal is to
-		 * prevent one process from snooping on the perf data from
-		 * another process. To do that, we need to check the value
-		 * that will be written to the register. Hence, limit
-		 * OACONTROL writes to only MI_LOAD_REGISTER_IMM commands.
+			 * OACONTROL requires some special handling for
+			 * writes. We want to make sure that any batch which
+			 * enables OA also disables it before the end of the
+			 * batch. The goal is to prevent one process from
+			 * snooping on the perf data from another process. To do
+			 * that, we need to check the value that will be written
+			 * to the register. Hence, limit OACONTROL writes to
+			 * only MI_LOAD_REGISTER_IMM commands.
 			 */
 			if (reg_addr == OACONTROL) {
 				if (desc->cmd.value == MI_LOAD_REGISTER_MEM) {
@@ -968,21 +1027,28 @@ static bool check_cmd(const struct intel_engine_cs *ring,
 				}
 
 				if (desc->cmd.value == MI_LOAD_REGISTER_IMM(1))
-				*oacontrol_set = (cmd[2] != 0);
+					*oacontrol_set = (cmd[offset + 1] != 0);
 			}
 
-		if (!valid_reg(ring->reg_table,
-			       ring->reg_count, reg_addr)) {
-			if (!is_master ||
-			    !valid_reg(ring->master_reg_table,
-				       ring->master_reg_count,
-				       reg_addr)) {
-				DRM_DEBUG_DRIVER("CMD: Rejected register 0x%08X in command: 0x%08X (ring=%d)\n",
-						 reg_addr,
-						 *cmd,
-						 ring->id);
+			/*
+			 * Check the value written to the register against the
+			 * allowed mask/value pair given in the whitelist entry.
+			 */
+			if (reg->mask) {
+				if (desc->cmd.value == MI_LOAD_REGISTER_MEM) {
+					DRM_DEBUG_DRIVER("CMD: Rejected LRM to masked register 0x%08X\n",
+							 reg_addr);
 					return false;
 				}
+
+				if (desc->cmd.value == MI_LOAD_REGISTER_IMM(1) &&
+				    (offset + 2 > length ||
+				     (cmd[offset + 1] & reg->mask) != reg->value)) {
+					DRM_DEBUG_DRIVER("CMD: Rejected LRI to masked register 0x%08X\n",
+							 reg_addr);
+					return false;
+				}
+			}
 		}
 	}
 
@@ -1105,7 +1171,8 @@ int i915_parse_cmds(struct intel_engine_cs *ring,
 			break;
 		}
 
-		if (!check_cmd(ring, desc, cmd, is_master, &oacontrol_set)) {
+		if (!check_cmd(ring, desc, cmd, length, is_master,
+			       &oacontrol_set)) {
 			ret = -EINVAL;
 			break;
 		}

drivers/gpu/drm/i915/i915_drv.h

@@ -2300,10 +2300,15 @@ struct drm_i915_cmd_descriptor {
 	 * Describes where to find a register address in the command to check
 	 * against the ring's register whitelist. Only valid if flags has the
 	 * CMD_DESC_REGISTER bit set.
+	 *
+	 * A non-zero step value implies that the command may access multiple
+	 * registers in sequence (e.g. LRI), in that case step gives the
+	 * distance in dwords between individual offset fields.
 	 */
 	struct {
 		u32 offset;
 		u32 mask;
+		u32 step;
 	} reg;
 
 #define MAX_CMD_DESC_BITMASKS 3

drivers/gpu/drm/i915/i915_gem.c

@@ -2813,9 +2813,6 @@ i915_gem_retire_requests_ring(struct intel_engine_cs *ring)
 {
 	WARN_ON(i915_verify_lists(ring->dev));
 
-	if (list_empty(&ring->active_list))
-		return;
-
 	/* Retire requests first as we use it above for the early return.
 	 * If we retire requests last, we may use a later seqno and so clear
 	 * the requests lists without clearing the active list, leading to

drivers/gpu/drm/i915/intel_display.c

@@ -87,7 +87,8 @@ static void ironlake_pch_clock_get(struct intel_crtc *crtc,
 				   struct intel_crtc_state *pipe_config);
 
 static int intel_set_mode(struct drm_crtc *crtc,
-			  struct drm_atomic_state *state);
+			  struct drm_atomic_state *state,
+			  bool force_restore);
 static int intel_framebuffer_init(struct drm_device *dev,
 				  struct intel_framebuffer *ifb,
 				  struct drm_mode_fb_cmd2 *mode_cmd,
@@ -10096,7 +10097,7 @@ bool intel_get_load_detect_pipe(struct drm_connector *connector,
 
 	drm_mode_copy(&crtc_state->base.mode, mode);
 
-	if (intel_set_mode(crtc, state)) {
+	if (intel_set_mode(crtc, state, true)) {
 		DRM_DEBUG_KMS("failed to set mode on load-detect pipe\n");
 		if (old->release_fb)
 			old->release_fb->funcs->destroy(old->release_fb);
@@ -10170,7 +10171,7 @@ void intel_release_load_detect_pipe(struct drm_connector *connector,
 		if (ret)
 			goto fail;
 
-		ret = intel_set_mode(crtc, state);
+		ret = intel_set_mode(crtc, state, true);
 		if (ret)
 			goto fail;
 
@@ -11385,10 +11386,6 @@ static void intel_modeset_fixup_state(struct drm_atomic_state *state)
 		crtc->base.enabled = crtc->base.state->enable;
 		crtc->config = to_intel_crtc_state(crtc->base.state);
 	}
-
-	/* Copy the new configuration to the staged state, to keep the few
-	 * pieces of code that haven't been converted yet happy */
-	intel_modeset_update_staged_output_state(state->dev);
 }
 
 static void
@@ -12646,20 +12643,24 @@ static int __intel_set_mode(struct drm_crtc *modeset_crtc,
 }
 
 static int intel_set_mode_with_config(struct drm_crtc *crtc,
-				      struct intel_crtc_state *pipe_config)
+				      struct intel_crtc_state *pipe_config,
+				      bool force_restore)
 {
 	int ret;
 
 	ret = __intel_set_mode(crtc, pipe_config);
 
-	if (ret == 0)
+	if (ret == 0 && force_restore) {
+		intel_modeset_update_staged_output_state(crtc->dev);
 		intel_modeset_check_state(crtc->dev);
+	}
 
 	return ret;
 }
 
 static int intel_set_mode(struct drm_crtc *crtc,
-			  struct drm_atomic_state *state)
+			  struct drm_atomic_state *state,
+			  bool force_restore)
 {
 	struct intel_crtc_state *pipe_config;
 	int ret = 0;
@@ -12670,7 +12671,7 @@ static int intel_set_mode(struct drm_crtc *crtc,
 		goto out;
 	}
 
-	ret = intel_set_mode_with_config(crtc, pipe_config);
+	ret = intel_set_mode_with_config(crtc, pipe_config, force_restore);
 	if (ret)
 		goto out;
 
@@ -12682,7 +12683,6 @@ void intel_crtc_restore_mode(struct drm_crtc *crtc)
 {
 	struct drm_device *dev = crtc->dev;
 	struct drm_atomic_state *state;
-	struct intel_crtc *intel_crtc;
 	struct intel_encoder *encoder;
 	struct intel_connector *connector;
 	struct drm_connector_state *connector_state;
@@ -12725,29 +12725,23 @@ void intel_crtc_restore_mode(struct drm_crtc *crtc)
 		}
 	}
 
-	for_each_intel_crtc(dev, intel_crtc) {
-		if (intel_crtc->new_enabled == intel_crtc->base.enabled)
-			continue;
-
-		crtc_state = intel_atomic_get_crtc_state(state, intel_crtc);
+	crtc_state = intel_atomic_get_crtc_state(state, to_intel_crtc(crtc));
 	if (IS_ERR(crtc_state)) {
 		DRM_DEBUG_KMS("Failed to add [CRTC:%d] to state: %ld\n",
-				      intel_crtc->base.base.id,
-				      PTR_ERR(crtc_state));
-			continue;
+			      crtc->base.id, PTR_ERR(crtc_state));
+		drm_atomic_state_free(state);
+		return;
 	}
 
 	crtc_state->base.active = crtc_state->base.enable =
-			intel_crtc->new_enabled;
+		to_intel_crtc(crtc)->new_enabled;
 
-		if (&intel_crtc->base == crtc)
 	drm_mode_copy(&crtc_state->base.mode, &crtc->mode);
-	}
 
 	intel_modeset_setup_plane_state(state, crtc, &crtc->mode,
 					crtc->primary->fb, crtc->x, crtc->y);
 
-	ret = intel_set_mode(crtc, state);
+	ret = intel_set_mode(crtc, state, false);
 	if (ret)
 		drm_atomic_state_free(state);
 }
@@ -12947,7 +12941,7 @@ static int intel_crtc_set_config(struct drm_mode_set *set)
 
 	primary_plane_was_visible = primary_plane_visible(set->crtc);
 
-	ret = intel_set_mode_with_config(set->crtc, pipe_config);
+	ret = intel_set_mode_with_config(set->crtc, pipe_config, true);
 
 	if (ret == 0 &&
 	    pipe_config->base.enable &&

drivers/gpu/drm/i915/intel_ringbuffer.h

@@ -118,6 +118,7 @@ struct intel_ringbuffer {
 };
 
 struct	intel_context;
+struct drm_i915_reg_descriptor;
 
 struct  intel_engine_cs {
 	const char	*name;
@@ -300,14 +301,14 @@ struct  intel_engine_cs {
 	/*
 	 * Table of registers allowed in commands that read/write registers.
 	 */
-	const u32 *reg_table;
+	const struct drm_i915_reg_descriptor *reg_table;
 	int reg_count;
 
 	/*
 	 * Table of registers allowed in commands that read/write registers, but
 	 * only from the DRM master.
 	 */
-	const u32 *master_reg_table;
+	const struct drm_i915_reg_descriptor *master_reg_table;
 	int master_reg_count;
 
 	/*