Re: [PATCH v4 3/7] libfs: Validate negative dentries in case-insensitive directories

[Gabriel Krisman Bertazi <krisman@suse.de>]

Eric Biggers <ebiggers@kernel.org> writes:

> On Thu, Jul 27, 2023 at 01:28:39PM -0400, Gabriel Krisman Bertazi wrote:
>> dentry->d_name is only checked by the case-insensitive d_revalidate hook
>> in the LOOKUP_CREATE/LOOKUP_RENAME_TARGET case since, for these cases,
>> d_revalidate is always called with the parent inode read-locked, and
>> therefore the name cannot change from under us.
>
> "at least read-locked"?  Or do you actually mean write-locked?

No. I mean read-locked, as in holding the read-part of the inode lock.
This is the case for lookup_slow, which is safe, despite the d_add_ci
case we discussed in the previous iteration.  I'll reword to say "at
least read-locked and mention it is the case in lookup_slow".

>> +static inline int generic_ci_d_revalidate(struct dentry *dentry,
>> +					  const struct qstr *name,
>> +					  unsigned int flags)
>
> No need for inline here.

sorry, I missed the inline from your previuos review.  Will fix it up
for this one.


>
> - Eric

-- 
Gabriel Krisman Bertazi