Skip to contentSkip to conversation

iommu: Improve the performance for direct_mapping

PATCH Applied Yong Wu 2 participants 2 messages Dec 7, 2020

Subsystems (2)
lore
Share & tools
lore

Yong Wu

author
#0 lore raw

PATCH v2 1/1 iommu: Improve the performance for direct_mapping

Currently direct_mapping always use the smallest pgsize which is SZ_4K normally to mapping. This is unnecessary. we could gather the size, and call iommu_map then, iommu_map could decide how to map better with the just right pgsize.

From the original comment, we should take care overlap, otherwise, iommu_map may return -EEXIST. In this overlap case, we should map the previous region before overlap firstly. then map the left part.

Each a iommu device will call this direct_mapping when its iommu initialize, This patch is effective to improve the boot/initialization time especially while it only needs level 1 mapping.

---
change notes:
v2: Refine the code flow.
v1: https://lore.kernel.org/linux-iommu/[email protected]/
base on v5.10-rc1.

---
 drivers/iommu/iommu.c | 22 +++++++++++++++++-----
 1 file changed, 17 insertions(+), 5 deletions(-)
drivers/iommu/iommu.c +17 -5
index 8c470f451a32..1a91decb95fa 100644
@@ -737,6 +737,7 @@ static int iommu_create_device_direct_mappings(struct iommu_group *group,
 	/* We need to consider overlapping regions for different devices */
 	list_for_each_entry(entry, &mappings, list) {
 		dma_addr_t start, end, addr;
+		size_t map_size = 0;
 
 		if (domain->ops->apply_resv_region)
 			domain->ops->apply_resv_region(dev, domain, entry);
@@ -748,16 +749,27 @@ static int iommu_create_device_direct_mappings(struct iommu_group *group,
 		    entry->type != IOMMU_RESV_DIRECT_RELAXABLE)
 			continue;
 
-		for (addr = start; addr < end; addr += pg_size) {
+		for (addr = start; addr <= end; addr += pg_size) {
 			phys_addr_t phys_addr;
 
+			if (addr == end)
+				goto map_end;
+
 			phys_addr = iommu_iova_to_phys(domain, addr);
-			if (phys_addr)
+			if (!phys_addr) {
+				map_size += pg_size;
 				continue;
+			}
 
-			ret = iommu_map(domain, addr, addr, pg_size, entry->prot);
-			if (ret)
-				goto out;
+map_end:
+			if (map_size) {
+				ret = iommu_map(domain, addr - map_size,
+						addr - map_size, map_size,
+						entry->prot);
+				if (ret)
+					goto out;
+				map_size = 0;
+			}
 		}
 
 	}
Signature · 6 lines
2.18.0

_______________________________________________
iommu mailing list
[email protected]
https://lists.linuxfoundation.org/mailman/listinfo/iommu
maintainer Applied
#1 lore raw
Quoted 10 lines · Yong Wu

Currently direct_mapping always use the smallest pgsize which is SZ_4K normally to mapping. This is unnecessary. we could gather the size, and call iommu_map then, iommu_map could decide how to map better with the just right pgsize.

Quoted 1 line

From the original comment, we should take care overlap, otherwise,

iommu_map may return -EEXIST. In this overlap case, we should map the previous region before overlap firstly. then map the left part.

[...]

Applied to arm64 (for-next/iommu/misc), thanks!

[1/1] iommu: Improve the performance for direct_mapping

      https://git.kernel.org/arm64/c/093b32a849b3

Cheers,

Signature · 9 lines
Will

https://fixes.arm64.dev
https://next.arm64.dev
https://will.arm64.dev
_______________________________________________
iommu mailing list
[email protected]
https://lists.linuxfoundation.org/mailman/listinfo/iommu