On Thu, 2016-07-07 at 12:14 +0800, Ian Kent wrote:
> On Sun, 2016-07-03 at 02:24 +0900, Tomohiro Kusumi wrote:
> > if (!dentry->d_fsdata) should be testing d_fsdata or ino without !.
>
> I've added both of these to the list of kernel patches I have, hopefully I'll
> be
> able to get the needed testing done to send them to Andrew after the next
> merge
> window.
>
> But I need to take a closer look at this one though.
>
> There's another ENOMEM return just above this where ino is not freed.
>
> So I'm wondering if it should be freed here at all.
>
> It belongs to the dentry and would get freed using autofs4_dentry_release() on
> final dput() and that takes care of additional house keeping too (although a
> symlink dentry is a bit different and shouldn't need that extra cleanup).
>
> What really worries me is that autofs4_free_ino() doesn't set d_fsdata to NULL
> so there's a chance of a double free when autofs4_dentry_release() is finally
> called.
I've removed the free from autofs4_dir_symlink(), it really should be done by
d_release() on final dput.
>
> >
> > Signed-off-by: Tomohiro Kusumi <kusumi.tomohiro@xxxxxxxxx>
> > ---
> > fs/autofs4/root.c | 2 +-
> > 1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/fs/autofs4/root.c b/fs/autofs4/root.c
> > index 93d3232..5a877bf 100644
> > --- a/fs/autofs4/root.c
> > +++ b/fs/autofs4/root.c
> > @@ -577,7 +577,7 @@ static int autofs4_dir_symlink(struct inode *dir,
> > inode = autofs4_get_inode(dir->i_sb, S_IFLNK | 0555);
> > if (!inode) {
> > kfree(cp);
> > - if (!dentry->d_fsdata)
> > + if (ino)
> > autofs4_free_ino(ino);
> > return -ENOMEM;
> > }
> --
> To unsubscribe from this list: send the line "unsubscribe autofs" in
--
To unsubscribe from this list: send the line "unsubscribe autofs" in
