On Wed, Feb 12, 2020 at 05:55:11AM +0000, Al Viro wrote:
> On Tue, Feb 11, 2020 at 09:10:13PM -0800, Eric Biggers wrote:
>
> > How about:
> >
> > int f2fs_ci_compare(const struct inode *parent, const struct qstr *name,
> > u8 *de_name, size_t de_name_len, bool quick)
> > {
> > const struct super_block *sb = parent->i_sb;
> > const struct unicode_map *um = sb->s_encoding;
> > struct fscrypt_str decrypted_name = FSTR_INIT(NULL, de_name_len);
> > struct qstr entry = QSTR_INIT(de_name, de_name_len);
> > int ret;
> >
> > if (IS_ENCRYPTED(parent)) {
>
> oops. parent->d_inode is unstable here; could have become NULL by that
> point.
>
> > if (quick)
> > ret = utf8_strncasecmp_folded(um, name, &entry);
> > else
> > ret = utf8_strncasecmp(um, name, &entry);
> > if (ret < 0) {
> > /* Handle invalid character sequence as either an error
> > * or as an opaque byte sequence.
> > */
>
> Really? How would the callers possibly tell mismatch from an
> error? And if they could, would would they *do* with that
> error, seeing that it might be an effect of a race with
> rename()?
>
> Again, ->d_compare() is NOT given a stable name. Or *parent. Or
> (parent->d_inode).
After the patch earlier in the series that created generic_ci_d_compare() and
switched f2fs to use it, f2fs_ci_compare() is only called when the filesystem is
actually searching a directory, not from ->d_compare(). So the names and
parent->d_inode are stable in it.
But, that also means the GFP_ATOMIC isn't needed, and f2fs_ci_compare() should
be made 'static'.
- Eric
