diff mbox series

[v2,4/9] ext4: don't call fscrypt_get_encryption_info() from dx_show_leaf()

Message ID 20201203022041.230976-5-ebiggers@kernel.org (mailing list archive)
State Accepted
Headers show
Series Allow deleting files with unsupported encryption policy | expand

Commit Message

Eric Biggers Dec. 3, 2020, 2:20 a.m. UTC
From: Eric Biggers <ebiggers@google.com>

The call to fscrypt_get_encryption_info() in dx_show_leaf() is too low
in the call tree; fscrypt_get_encryption_info() should have already been
called when starting the directory operation.  And indeed, it already
is.  Moreover, the encryption key is guaranteed to already be available
because dx_show_leaf() is only called when adding a new directory entry.

And even if the key wasn't available, dx_show_leaf() uses
fscrypt_fname_disk_to_usr() which knows how to create a no-key name.

So for the above reasons, and because it would be desirable to stop
exporting fscrypt_get_encryption_info() directly to filesystems, remove
the call to fscrypt_get_encryption_info() from dx_show_leaf().

Reviewed-by: Andreas Dilger <adilger@dilger.ca>
Signed-off-by: Eric Biggers <ebiggers@google.com>
---
 fs/ext4/namei.c | 8 +-------
 1 file changed, 1 insertion(+), 7 deletions(-)
diff mbox series

Patch

diff --git a/fs/ext4/namei.c b/fs/ext4/namei.c
index 793fc7db9d28f..7b31aea3e0253 100644
--- a/fs/ext4/namei.c
+++ b/fs/ext4/namei.c
@@ -643,13 +643,7 @@  static struct stats dx_show_leaf(struct inode *dir,
 
 				name  = de->name;
 				len = de->name_len;
-				if (IS_ENCRYPTED(dir))
-					res = fscrypt_get_encryption_info(dir);
-				if (res) {
-					printk(KERN_WARNING "Error setting up"
-					       " fname crypto: %d\n", res);
-				}
-				if (!fscrypt_has_encryption_key(dir)) {
+				if (!IS_ENCRYPTED(dir)) {
 					/* Directory is not encrypted */
 					ext4fs_dirhash(dir, de->name,
 						de->name_len, &h);