vfio/spapr_tce: Invalidate multiple TCEs at once

Invalidating a TCE cache entry for each updated TCE is quite expensive.
This makes use of the new iommu_table_ops::xchg_no_kill()/tce_kill()
callbacks to bring down the time spent in mapping a huge guest DMA window.
Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20190829085252.72370-4-aik@ozlabs.ru

drivers/vfio/vfio_iommu_spapr_tce.c

@@ -435,7 +435,7 @@ static int tce_iommu_clear(struct tce_container *container,
 	unsigned long oldhpa;
 	long ret;
 	enum dma_data_direction direction;
-	unsigned long lastentry = entry + pages;
+	unsigned long lastentry = entry + pages, firstentry = entry;
 
 	for ( ; entry < lastentry; ++entry) {
 		if (tbl->it_indirect_levels && tbl->it_userspace) {
@@ -460,7 +460,7 @@ static int tce_iommu_clear(struct tce_container *container,
 
 		direction = DMA_NONE;
 		oldhpa = 0;
-		ret = iommu_tce_xchg(container->mm, tbl, entry, &oldhpa,
+		ret = iommu_tce_xchg_no_kill(container->mm, tbl, entry, &oldhpa,
 				&direction);
 		if (ret)
 			continue;
@@ -476,6 +476,8 @@ static int tce_iommu_clear(struct tce_container *container,
 		tce_iommu_unuse_page(container, oldhpa);
 	}
 
+	iommu_tce_kill(tbl, firstentry, pages);
+
 	return 0;
 }
 
@@ -518,8 +520,8 @@ static long tce_iommu_build(struct tce_container *container,
 
 		hpa |= offset;
 		dirtmp = direction;
-		ret = iommu_tce_xchg(container->mm, tbl, entry + i, &hpa,
-				&dirtmp);
+		ret = iommu_tce_xchg_no_kill(container->mm, tbl, entry + i,
+				&hpa, &dirtmp);
 		if (ret) {
 			tce_iommu_unuse_page(container, hpa);
 			pr_err("iommu_tce: %s failed ioba=%lx, tce=%lx, ret=%ld\n",
@@ -536,6 +538,8 @@ static long tce_iommu_build(struct tce_container *container,
 
 	if (ret)
 		tce_iommu_clear(container, tbl, entry, i);
+	else
+		iommu_tce_kill(tbl, entry, pages);
 
 	return ret;
 }
@@ -572,8 +576,8 @@ static long tce_iommu_build_v2(struct tce_container *container,
 		if (mm_iommu_mapped_inc(mem))
 			break;
 
-		ret = iommu_tce_xchg(container->mm, tbl, entry + i, &hpa,
-				&dirtmp);
+		ret = iommu_tce_xchg_no_kill(container->mm, tbl, entry + i,
+				&hpa, &dirtmp);
 		if (ret) {
 			/* dirtmp cannot be DMA_NONE here */
 			tce_iommu_unuse_page_v2(container, tbl, entry + i);
@@ -593,6 +597,8 @@ static long tce_iommu_build_v2(struct tce_container *container,
 
 	if (ret)
 		tce_iommu_clear(container, tbl, entry, i);
+	else
+		iommu_tce_kill(tbl, entry, pages);
 
 	return ret;
 }