On Tue, Feb 21, 2017 at 05:23:26PM -0500, James Bottomley wrote:
> On Tue, 2017-02-21 at 11:06 -0500, Vivek Goyal wrote:
> > On Mon, Feb 20, 2017 at 04:47:05PM -0800, James Bottomley wrote:
> >
> > [..]
> > > +static struct dentry *shiftfs_lookup(struct inode *dir, struct
> > > dentry *dentry,
> > > + unsigned int flags)
> > > +{
> > > + struct dentry *real = dir->i_private, *new;
> > > + struct inode *reali = real->d_inode, *newi;
> >
> > newi needs to be initialized to NULL. Otherwise if a file does not
> > exist, we try to call d_splice_alias() with uninitialized value of
> > newi.
> >
> > This crashes for the new file creation case.
>
> I think you've still got an older version. The code around that region
> should read:
>
>
>
> oldcred = shiftfs_new_creds(&newcred, dentry->d_sb);
> new = lookup_one_len(dentry->d_name.name, real, dentry->d_name.len);
> shiftfs_old_creds(oldcred, &newcred);
> inode_unlock(reali);
>
> if (IS_ERR(new))
> return new;
>
> dentry->d_fsdata = new;
>
> newi = NULL;
> if (!new->d_inode)
> goto out;
>
> newi = shiftfs_new_inode(dentry->d_sb, new->d_inode->i_mode, new);
> if (!newi) {
> dput(new);
> return ERR_PTR(-ENOMEM);
> }
>
> out:
> return d_splice_alias(newi, dentry);
>
> So newi is set to NULL before out is taken. I checked: this hunk did make it to the v2 patch.
Aha.., you are right. I did not pay attention of newi=NULL in V2 and went
on to provide feeback based on my V1 testing.
Vivek
