drm/i915/gvt: support inconsecutive partial gtt entry write

Previously we assumed two 4-byte writes to the same PTE coming in sequence.
But recently we observed inconsecutive partial write happening as well. So
this patch enhances the previous solution. It now uses a list to save more
partial writes. If one partial write can be combined with another one in
the list to construct a full PTE, update its shadow entry. Otherwise, save
the partial write in the list.

v2: invalidate old entry and flush ggtt (Zhenyu)
v3: split old ggtt page unmap to another patch (Zhenyu)
v4: refine codes (Zhenyu)

Signed-off-by: Hang Yuan <hang.yuan@linux.intel.com>
Cc: Yan Zhao <yan.y.zhao@intel.com>
Cc: Xiaolin Zhang <xiaolin.zhang@intel.com>
Cc: Zhenyu Wang <zhenyu.z.wang@intel.com>
Reviewed-by: Xiaolin Zhang <xiaolin.zhang@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>

1 files changed, 7 insertions, 2 deletions

diff --git a/drivers/gpu/drm/i915/gvt/gtt.h b/drivers/gpu/drm/i915/gvt/gtt.h
index 7a9b36176efb..a11bfee1e0c8 100644
--- a/drivers/gpu/drm/i915/gvt/gtt.h
+++ b/drivers/gpu/drm/i915/gvt/gtt.h
@@ -133,6 +133,12 @@ enum intel_gvt_mm_type {
 
 #define GVT_RING_CTX_NR_PDPS	GEN8_3LVL_PDPES
 
+struct intel_gvt_partial_pte {
+	unsigned long offset;
+	u64 data;
+	struct list_head list;
+};
+
 struct intel_vgpu_mm {
 	enum intel_gvt_mm_type type;
 	struct intel_vgpu *vgpu;
@@ -157,8 +163,7 @@ struct intel_vgpu_mm {
 		} ppgtt_mm;
 		struct {
 			void *virtual_ggtt;
-			unsigned long last_partial_off;
-			u64 last_partial_data;
+			struct list_head partial_pte_list;
 		} ggtt_mm;
 	};
 };