drm/rockchip: gem: add IOMMU_TLB_SHOT_ENTIRE when call iommu_map_sg

Add IOMMU_TLB_SHOT_ENTIRE prot for iommu_map_sg, this shoot down entire
iommu tlb one time after iommu_map_sg.This may save about 5ms when DRM
mapping 12MB buffer.

Change-Id: I618aff1b3928bd1ec1dd3d896db746e09e96acdc
Signed-off-by: Simon Xue <xxm@rock-chips.com>

drivers/gpu/drm/rockchip/rockchip_drm_gem.c

@@ -30,6 +30,10 @@ static int rockchip_gem_iommu_map(struct rockchip_gem_object *rk_obj)
 	int prot = IOMMU_READ | IOMMU_WRITE;
 	ssize_t ret;
 
+#ifdef IOMMU_TLB_SHOT_ENTIRE
+	prot |= IOMMU_TLB_SHOT_ENTIRE;
+#endif
+
 	mutex_lock(&private->mm_lock);
 	ret = drm_mm_insert_node_generic(&private->mm, &rk_obj->mm,
 					 rk_obj->base.size, PAGE_SIZE,

drivers/iommu/iommu.c

@@ -2636,6 +2636,12 @@ static size_t __iommu_map_sg(struct iommu_domain *domain, unsigned long iova,
 
 	if (ops->iotlb_sync_map)
 		ops->iotlb_sync_map(domain, iova, mapped);
+
+#ifdef IOMMU_TLB_SHOT_ENTIRE
+	if (domain->ops->flush_iotlb_all && (prot & IOMMU_TLB_SHOT_ENTIRE))
+		domain->ops->flush_iotlb_all(domain);
+#endif
+
 	return mapped;
 
 out_err:

drivers/iommu/rockchip-iommu.c

@@ -993,7 +993,13 @@ static int rk_iommu_map_iova(struct rk_iommu_domain *rk_domain, u32 *pte_addr,
 	 * We only zap the first and last iova, since only they could have
 	 * dte or pte shared with an existing mapping.
 	 */
+#ifdef IOMMU_TLB_SHOT_ENTIRE
+	/* Do not zap tlb cache line if IOMMU_TLB_SHOT_ENTIRE set */
+	if (!(prot & IOMMU_TLB_SHOT_ENTIRE))
+		rk_iommu_zap_iova_first_last(rk_domain, iova, size);
+#else
 	rk_iommu_zap_iova_first_last(rk_domain, iova, size);
+#endif
 
 	return 0;
 unwind:
@@ -1038,7 +1044,13 @@ static int rk_iommu_map_iova_v2(struct rk_iommu_domain *rk_domain, u32 *pte_addr
 	 * We only zap the first and last iova, since only they could have
 	 * dte or pte shared with an existing mapping.
 	 */
+#ifdef IOMMU_TLB_SHOT_ENTIRE
+	/* Do not zap tlb cache line if IOMMU_TLB_SHOT_ENTIRE set */
+	if (!(prot & IOMMU_TLB_SHOT_ENTIRE))
+		rk_iommu_zap_iova_first_last(rk_domain, iova, size);
+#else
 	rk_iommu_zap_iova_first_last(rk_domain, iova, size);
+#endif
 
 	return 0;
 unwind:

include/linux/iommu.h

@@ -45,6 +45,13 @@
  */
 #define IOMMU_SYS_CACHE_ONLY_NWA (1 << 7)
 
+#ifdef CONFIG_NO_GKI
+
+/* For shoting entire IOMMU tlb once */
+#define IOMMU_TLB_SHOT_ENTIRE	(1 << 8)
+
+#endif
+
 struct iommu_ops;
 struct iommu_group;
 struct bus_type;