fscrypt: explicitly require that inode->i_blkbits be set

Document that fscrypt_prepare_new_inode() requires inode->i_blkbits to
be set, and make it WARN if it's not.  This would have made the CephFS
bug https://tracker.ceph.com/issues/64035 a bit easier to debug.

Signed-off-by: Xiubo Li <xiubli@redhat.com>
Link: https://lore.kernel.org/r/20240201003525.1788594-1-xiubli@redhat.com
Signed-off-by: Eric Biggers <ebiggers@google.com>

diff --git a/fs/crypto/keysetup.c b/fs/crypto/keysetup.c
index d71f7c799e79ef5f652d02ce9fa683dab51b123c..9a0a40c81bf2962b54cfdd85ecc63983f74dd14c 100644 (file)
--- a/fs/crypto/keysetup.c
+++ b/fs/crypto/keysetup.c
@@ -687,7 +687,7 @@ int fscrypt_get_encryption_info(struct inode *inode, bool allow_unsupported)
 /**
  * fscrypt_prepare_new_inode() - prepare to create a new inode in a directory
  * @dir: a possibly-encrypted directory
- * @inode: the new inode.  ->i_mode must be set already.
+ * @inode: the new inode.  ->i_mode and ->i_blkbits must be set already.
  *        ->i_ino doesn't need to be set yet.
  * @encrypt_ret: (output) set to %true if the new inode will be encrypted
  *
@@ -717,6 +717,9 @@ int fscrypt_prepare_new_inode(struct inode *dir, struct inode *inode,
        if (IS_ERR(policy))
                return PTR_ERR(policy);
 
+       if (WARN_ON_ONCE(inode->i_blkbits == 0))
+               return -EINVAL;
+
        if (WARN_ON_ONCE(inode->i_mode == 0))
                return -EINVAL;