Skip to contentSkip to conversation

fs/cramfs: Convert kmap() to kmap_local_data()

PATCH RESEND In review Fabio M. De Francesco 2 participants 2 messages Jan 9, 2023 – Jan 11, 2023 Cramfs filesystem

lore
Share & tools
lore
#0 lore raw

PATCH RESEND 1/1 fs/cramfs: Convert kmap() to kmap_local_data()

The use of kmap() is being deprecated in favor of kmap_local_page().

There are two main problems with kmap(): (1) It comes with an overhead as the mapping space is restricted and protected by a global lock for synchronization and (2) it also requires global TLB invalidation when the kmap’s pool wraps and it might block when the mapping space is fully utilized until a slot becomes available.

With kmap_local_page() the mappings are per thread, CPU local, can take page faults, and can be called from any context (including interrupts). It is faster than kmap() in kernels with HIGHMEM enabled. Furthermore, the tasks can be preempted and, when they are scheduled to run again, the kernel virtual addresses are restored and still valid.

Since its use in fs/cramfs is safe everywhere, it should be preferred.

Therefore, replace kmap() with kmap_local_page() in fs/cramfs. Instead of open-coding kmap_local_page() + memcpy(), use memcpy_from_page().

---

I'm resending this patch to Al Viro as it was asked by Nicolas[1]. In the meantime I'm also adding three "Reviewed-by" tags from Christian, Ira and Nicolas (thanks to you all!).

[1] https://lore.kernel.org/lkml/[email protected]/

 fs/cramfs/inode.c | 9 ++++-----
 1 file changed, 4 insertions(+), 5 deletions(-)
fs/cramfs/inode.c +4 -5
index 61ccf7722fc3..c17cbba5d45e 100644
@@ -238,8 +238,7 @@ static void *cramfs_blkdev_read(struct super_block *sb, unsigned int offset,
 		struct page *page = pages[i];
 
 		if (page) {
-			memcpy(data, kmap(page), PAGE_SIZE);
-			kunmap(page);
+			memcpy_from_page(data, page, 0, PAGE_SIZE);
 			put_page(page);
 		} else
 			memset(data, 0, PAGE_SIZE);
@@ -815,7 +814,7 @@ static int cramfs_read_folio(struct file *file, struct folio *folio)
 
 	maxblock = (inode->i_size + PAGE_SIZE - 1) >> PAGE_SHIFT;
 	bytes_filled = 0;
-	pgdata = kmap(page);
+	pgdata = kmap_local_page(page);
 
 	if (page->index < maxblock) {
 		struct super_block *sb = inode->i_sb;
@@ -903,13 +902,13 @@ static int cramfs_read_folio(struct file *file, struct folio *folio)
 
 	memset(pgdata + bytes_filled, 0, PAGE_SIZE - bytes_filled);
 	flush_dcache_page(page);
-	kunmap(page);
+	kunmap_local(pgdata);
 	SetPageUptodate(page);
 	unlock_page(page);
 	return 0;
 
 err:
-	kunmap(page);
+	kunmap_local(pgdata);
 	ClearPageUptodate(page);
 	SetPageError(page);
 	unlock_page(page);
Signature · 1 line
2.39.0
#1 lore raw
Quoted 14 lines · Fabio M. De Francesco

The use of kmap() is being deprecated in favor of kmap_local_page().

There are two main problems with kmap(): (1) It comes with an overhead as the mapping space is restricted and protected by a global lock for synchronization and (2) it also requires global TLB invalidation when the kmap’s pool wraps and it might block when the mapping space is fully utilized until a slot becomes available.

With kmap_local_page() the mappings are per thread, CPU local, can take page faults, and can be called from any context (including interrupts). It is faster than kmap() in kernels with HIGHMEM enabled. Furthermore, the tasks can be preempted and, when they are scheduled to run again, the kernel virtual addresses are restored and still valid.

Quoted 4 lines

Since its use in fs/cramfs is safe everywhere, it should be preferred.

Therefore, replace kmap() with kmap_local_page() in fs/cramfs. Instead of open-coding kmap_local_page() + memcpy(), use memcpy_from_page().

Applied (#work.misc).