Re: [PATCH] ubifs: Correctly handle 0 value hashes in ubifs_lookup()

From: Richard Weinberger
Date: Mon Apr 24 2017 - 17:41:56 EST


Am 24.04.2017 um 23:31 schrieb Richard Weinberger:
> Use fname_name(&nm) == NULL to detect whether a double hash
> lookup should be performed instead of checking for nm.hash
> being non-zero. zero is a valid hash value.
>
> Fixes: f4f61d2cc6d8 ("ubifs: Implement encrypted filenames")
> Cc: <stable@xxxxxxxxxxxxxxx>
> Signed-off-by: Richard Weinberger <richard@xxxxxx>
> ---
> fs/ubifs/dir.c | 3 +--
> 1 file changed, 1 insertion(+), 2 deletions(-)
>
> diff --git a/fs/ubifs/dir.c b/fs/ubifs/dir.c
> index b777bddaa1dd..ff77a0aa2f2b 100644
> --- a/fs/ubifs/dir.c
> +++ b/fs/ubifs/dir.c
> @@ -249,9 +249,8 @@ static struct dentry *ubifs_lookup(struct inode *dir, struct dentry *dentry,
> goto out_fname;
> }
>
> - if (nm.hash) {
> + if (fname_name(&nm) == NULL) {
> ubifs_assert(fname_len(&nm) == 0);
> - ubifs_assert(fname_name(&nm) == NULL);
> dent_key_init_hash(c, &key, dir->i_ino, nm.hash);
> err = ubifs_tnc_lookup_dh(c, &key, dent, nm.minor_hash);
> } else {
>

Just realized that checking for nm.hash being non-zero is still legit on UBIFS
since UBIFS uses hash values 0, 1 and 2 in readdir() as special values.
key_mask_hash() makes sure that no hash value is smaller than 3 that goes do disk.

Thanks,
//richard