Skip to contentSkip to conversation

ecryptfs: use filemap_dirty_folio for address space operations

PATCH In review Aditya Prakash Srivastava 3 participants 4 messages Jul 3 – Jul 21 Ecrypt file system

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 ecryptfs: use filemap_dirty_folio for address space operations

ecryptfs does not use buffer_heads. The legacy block_dirty_folio and block_invalidate_folio mapping operations were only added as a temporary compatibility fallback under CONFIG_BLOCK.

Since ecryptfs does not attach private metadata (such as buffer_heads) to its folios, block_dirty_folio is unnecessary.

Modernize ecryptfs to use filemap_dirty_folio for its dirty_folio address space operation. This allows removing the block_dirty_folio and block_invalidate_folio fallbacks, removing the buffer_head header include, and removing the CONFIG_BLOCK dependency inside ecryptfs_aops.

---
 fs/ecryptfs/mmap.c | 15 +--------------
 1 file changed, 1 insertion(+), 14 deletions(-)
fs/ecryptfs/mmap.c +1 -14
index 2c2b12fedeae..a057472b409c 100644
@@ -510,21 +510,8 @@ static sector_t ecryptfs_bmap(struct address_space *mapping, sector_t block)
 	return block;
 }
 
-#include <linux/buffer_head.h>
-
 const struct address_space_operations ecryptfs_aops = {
-	/*
-	 * XXX: This is pretty broken for multiple reasons: ecryptfs does not
-	 * actually use buffer_heads, and ecryptfs will crash without
-	 * CONFIG_BLOCK.  But it matches the behavior before the default for
-	 * address_space_operations without the ->dirty_folio method was
-	 * cleaned up, so this is the best we can do without maintainer
-	 * feedback.
-	 */
-#ifdef CONFIG_BLOCK
-	.dirty_folio	= block_dirty_folio,
-	.invalidate_folio = block_invalidate_folio,
-#endif
+	.dirty_folio	= filemap_dirty_folio,
 	.writepages = ecryptfs_writepages,
 	.read_folio = ecryptfs_read_folio,
 	.write_begin = ecryptfs_write_begin,
Signature · 1 line
2.47.3
maintainer
#1 lore raw
Quoted 7 lines · Aditya Prakash Srivastava

ecryptfs does not use buffer_heads. The legacy block_dirty_folio and block_invalidate_folio mapping operations were only added as a temporary compatibility fallback under CONFIG_BLOCK.

Since ecryptfs does not attach private metadata (such as buffer_heads) to its folios, block_dirty_folio is unnecessary.

Quoted 6 lines

Modernize ecryptfs to use filemap_dirty_folio for its dirty_folio address space operation. This allows removing the block_dirty_folio and block_invalidate_folio fallbacks, removing the buffer_head header include, and removing the CONFIG_BLOCK dependency inside ecryptfs_aops.

This looks correct to me and I've successfully ran the kernel tests in the ecryptfs-utils test suite with the patch applied.

Adding Christoph for comment since this addresses the concern he documented in the code comment below.

Tyler

Quoted 34 lines
---
 fs/ecryptfs/mmap.c | 15 +--------------
 1 file changed, 1 insertion(+), 14 deletions(-)
fs/ecryptfs/mmap.c +1 -14
index 2c2b12fedeae..a057472b409c 100644
@@ -510,21 +510,8 @@ static sector_t ecryptfs_bmap(struct address_space *mapping, sector_t block)
 	return block;
 }
 
-#include <linux/buffer_head.h>
-
 const struct address_space_operations ecryptfs_aops = {
-	/*
-	 * XXX: This is pretty broken for multiple reasons: ecryptfs does not
-	 * actually use buffer_heads, and ecryptfs will crash without
-	 * CONFIG_BLOCK.  But it matches the behavior before the default for
-	 * address_space_operations without the ->dirty_folio method was
-	 * cleaned up, so this is the best we can do without maintainer
-	 * feedback.
-	 */
-#ifdef CONFIG_BLOCK
-	.dirty_folio	= block_dirty_folio,
-	.invalidate_folio = block_invalidate_folio,
-#endif
+	.dirty_folio	= filemap_dirty_folio,
 	.writepages = ecryptfs_writepages,
 	.read_folio = ecryptfs_read_folio,
 	.write_begin = ecryptfs_write_begin,
Signature · 1 line
2.47.3
Reviewed-by
#2 lore raw
Quoted 5 lines · Tyler Hicks

This looks correct to me and I've successfully ran the kernel tests in the ecryptfs-utils test suite with the patch applied.

Adding Christoph for comment since this addresses the concern he documented in the code comment below.

Looks good to me:

maintainer
#3 lore raw
In reply to Aditya Prakash Srivastava
Quoted 8 lines · Aditya Prakash Srivastava

ecryptfs does not use buffer_heads. The legacy block_dirty_folio and block_invalidate_folio mapping operations were only added as a temporary compatibility fallback under CONFIG_BLOCK.

Since ecryptfs does not attach private metadata (such as buffer_heads) to its folios, block_dirty_folio is unnecessary.

[...]

Thank you! This has been applied to the next branch of the tyhicks/ecryptfs.git tree.

You can find a direct link below but please be aware that the commit hash is unstable and, therefore, the URL may not be valid in the future.

[1/1] ecryptfs: use filemap_dirty_folio for address space operations

      https://git.kernel.org/tyhicks/ecryptfs/c/21a6287ae2489426d23979e2213033dd204aa9fa

Tyler