Re: [PATCH v3 09/19] ovl: stop using lookup_one() in iterate_shared() handling.

On Mon, 27 Apr 2026, Amir Goldstein wrote:
> On Mon, Apr 27, 2026 at 6:07 AM NeilBrown <[email protected]> wrote:
> >
> > From: NeilBrown <[email protected]>
> >
> > lookup_one() is expected to be removed as it does not fit well with
> > proposed changes to directory locking.
> > Specifically d_alloc_parallel() will be ordered outside of i_rwsem
> > and as iterate_shared() is called with i_rwsem held it is not safe
> > to call d_alloc_parallel().
> >
> > We can instead call d_alloc_noblock() and then call the ->lookup, but
> > that can fail if there is a lookup attempt concurrent with the
> > readdir().
> >
> > ovl cannot afford for the lookup to fail as that could produce incorrect
> > results, and it cannot safely drop i_rwsem temporarily and that could
> > introduce races with handling of the directory cache.
> >
> > Instead we rely on the fact that ovl_iterate() has an exclusive lock on
> > the directory, so any concurrent lookup will wait for the ovl_iterate()
> > call to complete.  We allocate a separate dentry and if the lookup is
> > successful, it is hashed with the result.
> >
> > When the concurrent lookup gets i_rwsem it mustn't do its own lookup -
> > it must use the existing dentry.  This is found, if it exists, using
> > try_lookup_noperm().
> 
> Hi Niel,
> 
> One nit about the subject -
> Please change it to "ovl: stop using lookup_one() in ovl_iterate()"

That's reasonable.  And that for providing Claude's reflections.  All
fixed.

Thanks,
NeilBrown


> 
> Apart from that, I let Claude do the review, because it has surpassed
> my abilities in reviewing such subtle api changes.
> 
> I will copy its output verbatim, if for nothing else, to encourage you
> to start applying Claude review before sending out your patches...
> 
> >
> > Signed-off-by: NeilBrown <[email protected]>
> > ---
> >  fs/overlayfs/namei.c   | 12 ++++++++++++
> >  fs/overlayfs/readdir.c | 24 ++++++++++++++++++++++--
> >  2 files changed, 34 insertions(+), 2 deletions(-)
> >
> > diff --git a/fs/overlayfs/namei.c b/fs/overlayfs/namei.c
> > index ca899fdfaafd..69032eb2b1e2 100644
> > --- a/fs/overlayfs/namei.c
> > +++ b/fs/overlayfs/namei.c
> > @@ -1385,6 +1385,7 @@ struct dentry *ovl_lookup(struct inode *dir, struct dentry *dentry,
> >         struct ovl_fs *ofs = OVL_FS(dentry->d_sb);
> >         struct ovl_entry *poe = OVL_E(dentry->d_parent);
> >         bool check_redirect = (ovl_redirect_follow(ofs) || ofs->numdatalayer);
> > +       struct dentry *alias;
> >         int err;
> >         struct ovl_lookup_ctx ctx = {
> >                 .dentry = dentry,
> > @@ -1399,6 +1400,17 @@ struct dentry *ovl_lookup(struct inode *dir, struct dentry *dentry,
> >         if (dentry->d_name.len > ofs->namelen)
> >                 return ERR_PTR(-ENAMETOOLONG);
> >
> > +       /*
> > +        * The existance of this in-lookup dentry might have forced
> 
> "existance" should be "existence" in the namei.c comment
> 
> > +        * readdir to do the lookup with a new dentry.  If so we must
> > +        * return that one.
> > +        */
> > +       alias = try_lookup_noperm(&QSTR_LEN(dentry->d_name.name,
> > +                                           dentry->d_name.len),
> > +                                 dentry->d_parent);
> > +       if (alias && !IS_ERR(alias))
> > +               return alias;
> > +
> >         with_ovl_creds(dentry->d_sb)
> >                 err = ovl_lookup_layers(&ctx, &d);
> >
> > diff --git a/fs/overlayfs/readdir.c b/fs/overlayfs/readdir.c
> > index 1dcc75b3a90f..e03b32491941 100644
> > --- a/fs/overlayfs/readdir.c
> > +++ b/fs/overlayfs/readdir.c
> > @@ -574,8 +574,28 @@ static int ovl_cache_update(const struct path *path, struct ovl_cache_entry *p,
> >                 }
> >         }
> >         /* This checks also for xwhiteouts */
> > -       this = lookup_one(mnt_idmap(path->mnt), &QSTR_LEN(p->name, p->len), dir);
> > -       if (IS_ERR_OR_NULL(this) || !this->d_inode) {
> > +       this = d_alloc_noblock(dir, &QSTR_LEN(p->name, p->len));
> > +       if (this == ERR_PTR(-EWOULDBLOCK)) {
> > +               /*
> > +                * Some other thead is looking up this name and will
> 
> "thead" should be "thread" in the readdir.c comment
> 
> > +                * block on i_rwsem before it can complete the lookup.
> > +                * We will do the lookup in a new dentry and when that
> > +                * lookup gets a turn it will find and return this
> > +                * dentry.
> > +                */
> > +               this = d_alloc_name(dir, p->name);
> 
> Claude suggests this as the simplest fix to NULL deref bug below:
>                     if (!this)
>                             this = ERR_PTR(-ENOMEM);
> 
> > +       }
> > +       if (!IS_ERR(this) && !d_unhashed(this)) {
> > +               /* Either we got an in-lookup or we made our own unhashed */
> 
> The condition should be d_unhashed(this) (without !), matching the comment
> which says "Either we got an in-lookup or we made our own unhashed".
> 
> d_alloc_name() calls d_alloc() which returns NULL on allocation failure.
> The original code checked IS_ERR_OR_NULL(this), but the new code only
> checks IS_ERR(this). A NULL this would pass the !IS_ERR(this) check and
> then crash on !d_unhashed(this) (or d_unhashed(this) with the fix).
> 
> > +               struct dentry *alias = ovl_lookup(dir->d_inode, this, 0);
> > +
> > +               if (alias) {
> > +                       d_lookup_done(this);
> > +                       dput(this);
> > +                       this = alias;
> > +               }
> > +       }
> > +       if (IS_ERR(this) || !this->d_inode) {
> >                 /* Mark a stale entry */
> >                 p->is_whiteout = true;
> >                 if (IS_ERR(this)) {
> > --
> 
> Claude (Opus 4.6) review summary:
> 
> The patch has a critical condition inversion (!d_unhashed vs d_unhashed)
> that would break impure readdir entirely and cause use-after-free corruption
> in the dcache. It also has a missing NULL check for d_alloc_name() failure.
> 
> Thanks,
> Amir.
>