On Wed, Mar 25, 2020 at 05:58:30AM +0000, Al Viro wrote:
> On Wed, Mar 25, 2020 at 04:03:59AM +0000, Al Viro wrote:
>
> > Lovely. So
> > * we really do get NULL nd->path.dentry there; I've not misread the
> > trace.
> > * on the entry into link_path_walk() nd->path.dentry is non-NULL.
> > * *ALL* components should've been LAST_NORM ones
> > * not a single symlink in sight, unless the setup is rather unusual
> > * possibly not even a single mountpoint along the way (depending
> > upon the userland used)
>
> OK, I see one place where that could occur, but I really don't see how that
> could be triggered on this pathname, short of very odd symlink layout in
> the filesystem on the testbox.
... which, apparently, is what you've got there (/var/run -> ../run), so
stepping into that braino is not implausible. Could you check if the
fix below fixes what you've observed? I am folding it in anyway (into
"lift all calls of step_into() out of follow_dotdot/follow_dotdot_rcu") -
it's an obvious braino introduced in the commit in question, but I'd
like a confirmation that this _is_ what you've caught.
> Does the following fix your reproducer?
>
> diff --git a/fs/namei.c b/fs/namei.c
> index 311e33dbac63..4082b70f32ff 100644
> --- a/fs/namei.c
> +++ b/fs/namei.c
> @@ -1805,6 +1805,8 @@ static const char *handle_dots(struct nameidata *nd, int type)
> error = step_into(nd, WALK_NOFOLLOW,
> parent, inode, seq);
> }
> + if (unlikely(error))
> + return ERR_PTR(error);
>
> if (unlikely(nd->flags & LOOKUP_IS_SCOPED)) {
> /*
