BackMerge v4.18-rc7 into drm-next

rmk requested this for armada and I think we've had a few
conflicts build up.

Signed-off-by: Dave Airlie <airlied@redhat.com>

8 files changed, 144 insertions, 4 deletions

diff --git a/drivers/gpu/drm/i915/gvt/cmd_parser.c b/drivers/gpu/drm/i915/gvt/cmd_parser.c
index 0651e63b25fb..45e89b1e0481 100644
--- a/drivers/gpu/drm/i915/gvt/cmd_parser.c
+++ b/drivers/gpu/drm/i915/gvt/cmd_parser.c
@@ -863,6 +863,7 @@ static int cmd_reg_handler(struct parser_exec_state *s,
 {
 	struct intel_vgpu *vgpu = s->vgpu;
 	struct intel_gvt *gvt = vgpu->gvt;
+	u32 ctx_sr_ctl;
 
 	if (offset + 4 > gvt->device_info.mmio_size) {
 		gvt_vgpu_err("%s access to (%x) outside of MMIO range\n",
@@ -895,6 +896,28 @@ static int cmd_reg_handler(struct parser_exec_state *s,
 		patch_value(s, cmd_ptr(s, index), VGT_PVINFO_PAGE);
 	}
 
+	/* TODO
+	 * Right now only scan LRI command on KBL and in inhibit context.
+	 * It's good enough to support initializing mmio by lri command in
+	 * vgpu inhibit context on KBL.
+	 */
+	if (IS_KABYLAKE(s->vgpu->gvt->dev_priv) &&
+			intel_gvt_mmio_is_in_ctx(gvt, offset) &&
+			!strncmp(cmd, "lri", 3)) {
+		intel_gvt_hypervisor_read_gpa(s->vgpu,
+			s->workload->ring_context_gpa + 12, &ctx_sr_ctl, 4);
+		/* check inhibit context */
+		if (ctx_sr_ctl & 1) {
+			u32 data = cmd_val(s, index + 1);
+
+			if (intel_gvt_mmio_has_mode_mask(s->vgpu->gvt, offset))
+				intel_vgpu_mask_mmio_write(vgpu,
+							offset, &data, 4);
+			else
+				vgpu_vreg(vgpu, offset) = data;
+		}
+	}
+
 	/* TODO: Update the global mask if this MMIO is a masked-MMIO */
 	intel_gvt_mmio_set_cmd_accessed(gvt, offset);
 	return 0;
diff --git a/drivers/gpu/drm/i915/gvt/display.c b/drivers/gpu/drm/i915/gvt/display.c
index 6ee50cb328f8..3019dbc39aef 100644
--- a/drivers/gpu/drm/i915/gvt/display.c
+++ b/drivers/gpu/drm/i915/gvt/display.c
@@ -219,7 +219,7 @@ static void emulate_monitor_status_change(struct intel_vgpu *vgpu)
 			~(TRANS_DDI_BPC_MASK | TRANS_DDI_MODE_SELECT_MASK |
 			TRANS_DDI_PORT_MASK);
 		vgpu_vreg_t(vgpu, TRANS_DDI_FUNC_CTL(TRANSCODER_A)) |=
-			(TRANS_DDI_BPC_8 | TRANS_DDI_MODE_SELECT_DP_SST |
+			(TRANS_DDI_BPC_8 | TRANS_DDI_MODE_SELECT_DVI |
 			(PORT_B << TRANS_DDI_PORT_SHIFT) |
 			TRANS_DDI_FUNC_ENABLE);
 		if (IS_BROADWELL(dev_priv)) {
@@ -239,7 +239,7 @@ static void emulate_monitor_status_change(struct intel_vgpu *vgpu)
 			~(TRANS_DDI_BPC_MASK | TRANS_DDI_MODE_SELECT_MASK |
 			TRANS_DDI_PORT_MASK);
 		vgpu_vreg_t(vgpu, TRANS_DDI_FUNC_CTL(TRANSCODER_A)) |=
-			(TRANS_DDI_BPC_8 | TRANS_DDI_MODE_SELECT_DP_SST |
+			(TRANS_DDI_BPC_8 | TRANS_DDI_MODE_SELECT_DVI |
 			(PORT_C << TRANS_DDI_PORT_SHIFT) |
 			TRANS_DDI_FUNC_ENABLE);
 		if (IS_BROADWELL(dev_priv)) {
@@ -259,7 +259,7 @@ static void emulate_monitor_status_change(struct intel_vgpu *vgpu)
 			~(TRANS_DDI_BPC_MASK | TRANS_DDI_MODE_SELECT_MASK |
 			TRANS_DDI_PORT_MASK);
 		vgpu_vreg_t(vgpu, TRANS_DDI_FUNC_CTL(TRANSCODER_A)) |=
-			(TRANS_DDI_BPC_8 | TRANS_DDI_MODE_SELECT_DP_SST |
+			(TRANS_DDI_BPC_8 | TRANS_DDI_MODE_SELECT_DVI |
 			(PORT_D << TRANS_DDI_PORT_SHIFT) |
 			TRANS_DDI_FUNC_ENABLE);
 		if (IS_BROADWELL(dev_priv)) {
diff --git a/drivers/gpu/drm/i915/gvt/gtt.c b/drivers/gpu/drm/i915/gvt/gtt.c
index 39980dfbbebd..00aad8164dec 100644
--- a/drivers/gpu/drm/i915/gvt/gtt.c
+++ b/drivers/gpu/drm/i915/gvt/gtt.c
@@ -1901,6 +1901,7 @@ static struct intel_vgpu_mm *intel_vgpu_create_ggtt_mm(struct intel_vgpu *vgpu)
 		vgpu_free_mm(mm);
 		return ERR_PTR(-ENOMEM);
 	}
+	mm->ggtt_mm.last_partial_off = -1UL;
 
 	return mm;
 }
@@ -1925,6 +1926,7 @@ void _intel_vgpu_mm_release(struct kref *mm_ref)
 		invalidate_ppgtt_mm(mm);
 	} else {
 		vfree(mm->ggtt_mm.virtual_ggtt);
+		mm->ggtt_mm.last_partial_off = -1UL;
 	}
 
 	vgpu_free_mm(mm);
@@ -2177,6 +2179,62 @@ static int emulate_ggtt_mmio_write(struct intel_vgpu *vgpu, unsigned int off,
 	memcpy((void *)&e.val64 + (off & (info->gtt_entry_size - 1)), p_data,
 			bytes);
 
+	/* If ggtt entry size is 8 bytes, and it's split into two 4 bytes
+	 * write, we assume the two 4 bytes writes are consecutive.
+	 * Otherwise, we abort and report error
+	 */
+	if (bytes < info->gtt_entry_size) {
+		if (ggtt_mm->ggtt_mm.last_partial_off == -1UL) {
+			/* the first partial part*/
+			ggtt_mm->ggtt_mm.last_partial_off = off;
+			ggtt_mm->ggtt_mm.last_partial_data = e.val64;
+			return 0;
+		} else if ((g_gtt_index ==
+				(ggtt_mm->ggtt_mm.last_partial_off >>
+				info->gtt_entry_size_shift)) &&
+			(off !=	ggtt_mm->ggtt_mm.last_partial_off)) {
+			/* the second partial part */
+
+			int last_off = ggtt_mm->ggtt_mm.last_partial_off &
+				(info->gtt_entry_size - 1);
+
+			memcpy((void *)&e.val64 + last_off,
+				(void *)&ggtt_mm->ggtt_mm.last_partial_data +
+				last_off, bytes);
+
+			ggtt_mm->ggtt_mm.last_partial_off = -1UL;
+		} else {
+			int last_offset;
+
+			gvt_vgpu_err("failed to populate guest ggtt entry: abnormal ggtt entry write sequence, last_partial_off=%lx, offset=%x, bytes=%d, ggtt entry size=%d\n",
+					ggtt_mm->ggtt_mm.last_partial_off, off,
+					bytes, info->gtt_entry_size);
+
+			/* set host ggtt entry to scratch page and clear
+			 * virtual ggtt entry as not present for last
+			 * partially write offset
+			 */
+			last_offset = ggtt_mm->ggtt_mm.last_partial_off &
+					(~(info->gtt_entry_size - 1));
+
+			ggtt_get_host_entry(ggtt_mm, &m, last_offset);
+			ggtt_invalidate_pte(vgpu, &m);
+			ops->set_pfn(&m, gvt->gtt.scratch_mfn);
+			ops->clear_present(&m);
+			ggtt_set_host_entry(ggtt_mm, &m, last_offset);
+			ggtt_invalidate(gvt->dev_priv);
+
+			ggtt_get_guest_entry(ggtt_mm, &e, last_offset);
+			ops->clear_present(&e);
+			ggtt_set_guest_entry(ggtt_mm, &e, last_offset);
+
+			ggtt_mm->ggtt_mm.last_partial_off = off;
+			ggtt_mm->ggtt_mm.last_partial_data = e.val64;
+
+			return 0;
+		}
+	}
+
 	if (ops->test_present(&e)) {
 		gfn = ops->get_pfn(&e);
 		m = e;
diff --git a/drivers/gpu/drm/i915/gvt/gtt.h b/drivers/gpu/drm/i915/gvt/gtt.h
index b7bf68cc8418..7a9b36176efb 100644
--- a/drivers/gpu/drm/i915/gvt/gtt.h
+++ b/drivers/gpu/drm/i915/gvt/gtt.h
@@ -157,6 +157,8 @@ struct intel_vgpu_mm {
 		} ppgtt_mm;
 		struct {
 			void *virtual_ggtt;
+			unsigned long last_partial_off;
+			u64 last_partial_data;
 		} ggtt_mm;
 	};
 };
diff --git a/drivers/gpu/drm/i915/gvt/gvt.h b/drivers/gpu/drm/i915/gvt/gvt.h
index de2a3a2580be..9a9671522774 100644
--- a/drivers/gpu/drm/i915/gvt/gvt.h
+++ b/drivers/gpu/drm/i915/gvt/gvt.h
@@ -274,6 +274,8 @@ struct intel_gvt_mmio {
 #define F_CMD_ACCESSED	(1 << 5)
 /* This reg could be accessed by unaligned address */
 #define F_UNALIGN	(1 << 6)
+/* This reg is saved/restored in context */
+#define F_IN_CTX	(1 << 7)
 
 	struct gvt_mmio_block *mmio_block;
 	unsigned int num_mmio_block;
@@ -655,6 +657,33 @@ static inline bool intel_gvt_mmio_has_mode_mask(
 	return gvt->mmio.mmio_attribute[offset >> 2] & F_MODE_MASK;
 }
 
+/**
+ * intel_gvt_mmio_is_in_ctx - check if a MMIO has in-ctx mask
+ * @gvt: a GVT device
+ * @offset: register offset
+ *
+ * Returns:
+ * True if a MMIO has a in-context mask, false if it isn't.
+ *
+ */
+static inline bool intel_gvt_mmio_is_in_ctx(
+			struct intel_gvt *gvt, unsigned int offset)
+{
+	return gvt->mmio.mmio_attribute[offset >> 2] & F_IN_CTX;
+}
+
+/**
+ * intel_gvt_mmio_set_in_ctx - mask a MMIO in logical context
+ * @gvt: a GVT device
+ * @offset: register offset
+ *
+ */
+static inline void intel_gvt_mmio_set_in_ctx(
+			struct intel_gvt *gvt, unsigned int offset)
+{
+	gvt->mmio.mmio_attribute[offset >> 2] |= F_IN_CTX;
+}
+
 int intel_gvt_debugfs_add_vgpu(struct intel_vgpu *vgpu);
 void intel_gvt_debugfs_remove_vgpu(struct intel_vgpu *vgpu);
 int intel_gvt_debugfs_init(struct intel_gvt *gvt);
diff --git a/drivers/gpu/drm/i915/gvt/handlers.c b/drivers/gpu/drm/i915/gvt/handlers.c
index 6b50f850dc28..7a58ca555197 100644
--- a/drivers/gpu/drm/i915/gvt/handlers.c
+++ b/drivers/gpu/drm/i915/gvt/handlers.c
@@ -3388,6 +3388,30 @@ int intel_vgpu_default_mmio_write(struct intel_vgpu *vgpu, unsigned int offset,
 }
 
 /**
+ * intel_vgpu_mask_mmio_write - write mask register
+ * @vgpu: a vGPU
+ * @offset: access offset
+ * @p_data: write data buffer
+ * @bytes: access data length
+ *
+ * Returns:
+ * Zero on success, negative error code if failed.
+ */
+int intel_vgpu_mask_mmio_write(struct intel_vgpu *vgpu, unsigned int offset,
+		void *p_data, unsigned int bytes)
+{
+	u32 mask, old_vreg;
+
+	old_vreg = vgpu_vreg(vgpu, offset);
+	write_vreg(vgpu, offset, p_data, bytes);
+	mask = vgpu_vreg(vgpu, offset) >> 16;
+	vgpu_vreg(vgpu, offset) = (old_vreg & ~mask) |
+				(vgpu_vreg(vgpu, offset) & mask);
+
+	return 0;
+}
+
+/**
  * intel_gvt_in_force_nonpriv_whitelist - if a mmio is in whitelist to be
  * force-nopriv register
  *
diff --git a/drivers/gpu/drm/i915/gvt/mmio.h b/drivers/gpu/drm/i915/gvt/mmio.h
index e474188b46d2..1ffc69eba30e 100644
--- a/drivers/gpu/drm/i915/gvt/mmio.h
+++ b/drivers/gpu/drm/i915/gvt/mmio.h
@@ -99,4 +99,6 @@ bool intel_gvt_in_force_nonpriv_whitelist(struct intel_gvt *gvt,
 int intel_vgpu_mmio_reg_rw(struct intel_vgpu *vgpu, unsigned int offset,
 			   void *pdata, unsigned int bytes, bool is_read);
 
+int intel_vgpu_mask_mmio_write(struct intel_vgpu *vgpu, unsigned int offset,
+				  void *p_data, unsigned int bytes);
 #endif
diff --git a/drivers/gpu/drm/i915/gvt/mmio_context.c b/drivers/gpu/drm/i915/gvt/mmio_context.c
index 20be9a92600f..42e1e6bdcc2c 100644
--- a/drivers/gpu/drm/i915/gvt/mmio_context.c
+++ b/drivers/gpu/drm/i915/gvt/mmio_context.c
@@ -587,7 +587,9 @@ 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)
+		if (mmio->in_context) {
 			gvt->engine_mmio_list.ctx_mmio_count[mmio->ring_id]++;
+			intel_gvt_mmio_set_in_ctx(gvt, mmio->reg.reg);
+		}
 	}
 }