drm/i915/gvt: Wean gvt off dev_priv->engine[]

Stop trying to escape out of the gvt layer to find the engine that we
initially setup for use with gvt. Record the engines during initialisation
and use them henceforth.

add/remove: 1/4 grow/shrink: 22/28 up/down: 341/-1410 (-1069)

[Zhenyu: rebase, fix nonpriv register check fault, fix gvt engine
thread run failure.]

Cc: Ding Zhuocheng <zhuocheng.ding@intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Acked-by: Zhenyu Wang <zhenyuw@linux.intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20200304032307.2983-2-zhenyuw@linux.intel.com

1 files changed, 54 insertions, 57 deletions

diff --git a/drivers/gpu/drm/i915/gvt/mmio_context.c b/drivers/gpu/drm/i915/gvt/mmio_context.c
index 1213c8e23317..c65042c6d50d 100644
--- a/drivers/gpu/drm/i915/gvt/mmio_context.c
+++ b/drivers/gpu/drm/i915/gvt/mmio_context.c
@@ -214,13 +214,11 @@ restore_context_mmio_for_inhibit(struct intel_vgpu *vgpu,
 	*cs++ = MI_LOAD_REGISTER_IMM(count);
 	for (mmio = gvt->engine_mmio_list.mmio;
 	     i915_mmio_reg_valid(mmio->reg); mmio++) {
-		if (mmio->ring_id != ring_id ||
-		    !mmio->in_context)
+		if (mmio->id != ring_id || !mmio->in_context)
 			continue;
 
 		*cs++ = i915_mmio_reg_offset(mmio->reg);
-		*cs++ = vgpu_vreg_t(vgpu, mmio->reg) |
-				(mmio->mask << 16);
+		*cs++ = vgpu_vreg_t(vgpu, mmio->reg) | (mmio->mask << 16);
 		gvt_dbg_core("add lri reg pair 0x%x:0x%x in inhibit ctx, vgpu:%d, rind_id:%d\n",
 			      *(cs-2), *(cs-1), vgpu->id, ring_id);
 	}
@@ -344,10 +342,10 @@ static u32 gen8_tlb_mmio_offset_list[] = {
 	[VECS0] = 0x4270,
 };
 
-static void handle_tlb_pending_event(struct intel_vgpu *vgpu, int ring_id)
+static void handle_tlb_pending_event(struct intel_vgpu *vgpu,
+				     const struct intel_engine_cs *engine)
 {
-	struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv;
-	struct intel_uncore *uncore = &dev_priv->uncore;
+	struct intel_uncore *uncore = engine->uncore;
 	struct intel_vgpu_submission *s = &vgpu->submission;
 	u32 *regs = vgpu->gvt->engine_mmio_list.tlb_mmio_offset_list;
 	u32 cnt = vgpu->gvt->engine_mmio_list.tlb_mmio_offset_list_cnt;
@@ -357,13 +355,13 @@ static void handle_tlb_pending_event(struct intel_vgpu *vgpu, int ring_id)
 	if (!regs)
 		return;
 
-	if (drm_WARN_ON(&dev_priv->drm, ring_id >= cnt))
+	if (drm_WARN_ON(&engine->i915->drm, engine->id >= cnt))
 		return;
 
-	if (!test_and_clear_bit(ring_id, (void *)s->tlb_handle_pending))
+	if (!test_and_clear_bit(engine->id, (void *)s->tlb_handle_pending))
 		return;
 
-	reg = _MMIO(regs[ring_id]);
+	reg = _MMIO(regs[engine->id]);
 
 	/* WaForceWakeRenderDuringMmioTLBInvalidate:skl
 	 * we need to put a forcewake when invalidating RCS TLB caches,
@@ -372,30 +370,27 @@ static void handle_tlb_pending_event(struct intel_vgpu *vgpu, int ring_id)
 	 */
 	fw = intel_uncore_forcewake_for_reg(uncore, reg,
 					    FW_REG_READ | FW_REG_WRITE);
-	if (ring_id == RCS0 && INTEL_GEN(dev_priv) >= 9)
+	if (engine->id == RCS0 && INTEL_GEN(engine->i915) >= 9)
 		fw |= FORCEWAKE_RENDER;
 
 	intel_uncore_forcewake_get(uncore, fw);
 
 	intel_uncore_write_fw(uncore, reg, 0x1);
 
-	if (wait_for_atomic((intel_uncore_read_fw(uncore, reg) == 0), 50))
-		gvt_vgpu_err("timeout in invalidate ring (%d) tlb\n", ring_id);
+	if (wait_for_atomic(intel_uncore_read_fw(uncore, reg) == 0, 50))
+		gvt_vgpu_err("timeout in invalidate ring %s tlb\n",
+			     engine->name);
 	else
 		vgpu_vreg_t(vgpu, reg) = 0;
 
 	intel_uncore_forcewake_put(uncore, fw);
 
-	gvt_dbg_core("invalidate TLB for ring %d\n", ring_id);
+	gvt_dbg_core("invalidate TLB for ring %s\n", engine->name);
 }
 
 static void switch_mocs(struct intel_vgpu *pre, struct intel_vgpu *next,
-			int ring_id)
+			const struct intel_engine_cs *engine)
 {
-	struct drm_i915_private *dev_priv;
-	i915_reg_t offset, l3_offset;
-	u32 old_v, new_v;
-
 	u32 regs[] = {
 		[RCS0]  = 0xc800,
 		[VCS0]  = 0xc900,
@@ -403,36 +398,38 @@ static void switch_mocs(struct intel_vgpu *pre, struct intel_vgpu *next,
 		[BCS0]  = 0xcc00,
 		[VECS0] = 0xcb00,
 	};
+	struct intel_uncore *uncore = engine->uncore;
+	i915_reg_t offset, l3_offset;
+	u32 old_v, new_v;
 	int i;
 
-	dev_priv = pre ? pre->gvt->dev_priv : next->gvt->dev_priv;
-	if (drm_WARN_ON(&dev_priv->drm, ring_id >= ARRAY_SIZE(regs)))
+	if (drm_WARN_ON(&engine->i915->drm, engine->id >= ARRAY_SIZE(regs)))
 		return;
 
-	if (ring_id == RCS0 && IS_GEN(dev_priv, 9))
+	if (engine->id == RCS0 && IS_GEN(engine->i915, 9))
 		return;
 
 	if (!pre && !gen9_render_mocs.initialized)
-		load_render_mocs(dev_priv);
+		load_render_mocs(engine->i915);
 
-	offset.reg = regs[ring_id];
+	offset.reg = regs[engine->id];
 	for (i = 0; i < GEN9_MOCS_SIZE; i++) {
 		if (pre)
 			old_v = vgpu_vreg_t(pre, offset);
 		else
-			old_v = gen9_render_mocs.control_table[ring_id][i];
+			old_v = gen9_render_mocs.control_table[engine->id][i];
 		if (next)
 			new_v = vgpu_vreg_t(next, offset);
 		else
-			new_v = gen9_render_mocs.control_table[ring_id][i];
+			new_v = gen9_render_mocs.control_table[engine->id][i];
 
 		if (old_v != new_v)
-			I915_WRITE_FW(offset, new_v);
+			intel_uncore_write_fw(uncore, offset, new_v);
 
 		offset.reg += 4;
 	}
 
-	if (ring_id == RCS0) {
+	if (engine->id == RCS0) {
 		l3_offset.reg = 0xb020;
 		for (i = 0; i < GEN9_MOCS_SIZE / 2; i++) {
 			if (pre)
@@ -445,7 +442,7 @@ static void switch_mocs(struct intel_vgpu *pre, struct intel_vgpu *next,
 				new_v = gen9_render_mocs.l3cc_table[i];
 
 			if (old_v != new_v)
-				I915_WRITE_FW(l3_offset, new_v);
+				intel_uncore_write_fw(uncore, l3_offset, new_v);
 
 			l3_offset.reg += 4;
 		}
@@ -467,38 +464,40 @@ bool is_inhibit_context(struct intel_context *ce)
 /* Switch ring mmio values (context). */
 static void switch_mmio(struct intel_vgpu *pre,
 			struct intel_vgpu *next,
-			int ring_id)
+			const struct intel_engine_cs *engine)
 {
-	struct drm_i915_private *dev_priv;
+	struct intel_uncore *uncore = engine->uncore;
 	struct intel_vgpu_submission *s;
 	struct engine_mmio *mmio;
 	u32 old_v, new_v;
 
-	dev_priv = pre ? pre->gvt->dev_priv : next->gvt->dev_priv;
-	if (INTEL_GEN(dev_priv) >= 9)
-		switch_mocs(pre, next, ring_id);
+	if (INTEL_GEN(engine->i915) >= 9)
+		switch_mocs(pre, next, engine);
 
-	for (mmio = dev_priv->gvt->engine_mmio_list.mmio;
+	for (mmio = engine->i915->gvt->engine_mmio_list.mmio;
 	     i915_mmio_reg_valid(mmio->reg); mmio++) {
-		if (mmio->ring_id != ring_id)
+		if (mmio->id != engine->id)
 			continue;
 		/*
 		 * No need to do save or restore of the mmio which is in context
 		 * state image on gen9, it's initialized by lri command and
 		 * save or restore with context together.
 		 */
-		if (IS_GEN(dev_priv, 9) && mmio->in_context)
+		if (IS_GEN(engine->i915, 9) && mmio->in_context)
 			continue;
 
 		// save
 		if (pre) {
-			vgpu_vreg_t(pre, mmio->reg) = I915_READ_FW(mmio->reg);
+			vgpu_vreg_t(pre, mmio->reg) =
+				intel_uncore_read_fw(uncore, mmio->reg);
 			if (mmio->mask)
 				vgpu_vreg_t(pre, mmio->reg) &=
-						~(mmio->mask << 16);
+					~(mmio->mask << 16);
 			old_v = vgpu_vreg_t(pre, mmio->reg);
-		} else
-			old_v = mmio->value = I915_READ_FW(mmio->reg);
+		} else {
+			old_v = mmio->value =
+				intel_uncore_read_fw(uncore, mmio->reg);
+		}
 
 		// restore
 		if (next) {
@@ -509,12 +508,12 @@ static void switch_mmio(struct intel_vgpu *pre,
 			 * itself.
 			 */
 			if (mmio->in_context &&
-			    !is_inhibit_context(s->shadow[ring_id]))
+			    !is_inhibit_context(s->shadow[engine->id]))
 				continue;
 
 			if (mmio->mask)
 				new_v = vgpu_vreg_t(next, mmio->reg) |
-							(mmio->mask << 16);
+					(mmio->mask << 16);
 			else
 				new_v = vgpu_vreg_t(next, mmio->reg);
 		} else {
@@ -526,7 +525,7 @@ static void switch_mmio(struct intel_vgpu *pre,
 				new_v = mmio->value;
 		}
 
-		I915_WRITE_FW(mmio->reg, new_v);
+		intel_uncore_write_fw(uncore, mmio->reg, new_v);
 
 		trace_render_mmio(pre ? pre->id : 0,
 				  next ? next->id : 0,
@@ -536,39 +535,37 @@ static void switch_mmio(struct intel_vgpu *pre,
 	}
 
 	if (next)
-		handle_tlb_pending_event(next, ring_id);
+		handle_tlb_pending_event(next, engine);
 }
 
 /**
  * intel_gvt_switch_render_mmio - switch mmio context of specific engine
  * @pre: the last vGPU that own the engine
  * @next: the vGPU to switch to
- * @ring_id: specify the engine
+ * @engine: the engine
  *
  * If pre is null indicates that host own the engine. If next is null
  * indicates that we are switching to host workload.
  */
 void intel_gvt_switch_mmio(struct intel_vgpu *pre,
-			   struct intel_vgpu *next, int ring_id)
+			   struct intel_vgpu *next,
+			   const struct intel_engine_cs *engine)
 {
-	struct drm_i915_private *dev_priv;
-
-	if (WARN(!pre && !next, "switch ring %d from host to HOST\n", ring_id))
+	if (WARN(!pre && !next, "switch ring %s from host to HOST\n",
+		 engine->name))
 		return;
 
-	gvt_dbg_render("switch ring %d from %s to %s\n", ring_id,
+	gvt_dbg_render("switch ring %s from %s to %s\n", engine->name,
 		       pre ? "vGPU" : "host", next ? "vGPU" : "HOST");
 
-	dev_priv = pre ? pre->gvt->dev_priv : next->gvt->dev_priv;
-
 	/**
 	 * We are using raw mmio access wrapper to improve the
 	 * performace for batch mmio read/write, so we need
 	 * handle forcewake mannually.
 	 */
-	intel_uncore_forcewake_get(&dev_priv->uncore, FORCEWAKE_ALL);
-	switch_mmio(pre, next, ring_id);
-	intel_uncore_forcewake_put(&dev_priv->uncore, FORCEWAKE_ALL);
+	intel_uncore_forcewake_get(engine->uncore, FORCEWAKE_ALL);
+	switch_mmio(pre, next, engine);
+	intel_uncore_forcewake_put(engine->uncore, FORCEWAKE_ALL);
 }
 
 /**
@@ -595,7 +592,7 @@ void intel_gvt_init_engine_mmio_context(struct intel_gvt *gvt)
 	for (mmio = gvt->engine_mmio_list.mmio;
 	     i915_mmio_reg_valid(mmio->reg); mmio++) {
 		if (mmio->in_context) {
-			gvt->engine_mmio_list.ctx_mmio_count[mmio->ring_id]++;
+			gvt->engine_mmio_list.ctx_mmio_count[mmio->id]++;
 			intel_gvt_mmio_set_in_ctx(gvt, mmio->reg.reg);
 		}
 	}