Re: [PATCH] fuse: don't keep inode-less dentry at fuse_ctl_add_dentry().

[Al Viro <viro@ZenIV.linux.org.uk>]

On Fri, May 11, 2018 at 09:55:00AM +0200, Miklos Szeredi wrote:
> On Thu, May 10, 2018 at 10:07 PM, Al Viro <viro@zeniv.linux.org.uk> wrote:
> > On Wed, May 09, 2018 at 07:58:58PM +0900, Tetsuo Handa wrote:
> >> >From 9f41081f8bd6762a6f629e5e23e6d07a62bba69c Mon Sep 17 00:00:00 2001
> >> From: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
> >> Date: Sat, 28 Apr 2018 11:24:09 +0900
> >> Subject: [PATCH] fuse: don't keep inode-less dentry at fuse_ctl_add_dentry().
> >>
> >> syzbot is reporting NULL pointer dereference at fuse_ctl_remove_conn() [1].
> >> Since fc->ctl_ndents is incremented by fuse_ctl_add_conn() when new_inode()
> >> failed, fuse_ctl_remove_conn() reaches an inode-less dentry and tries to
> >> clear d_inode(dentry)->i_private field. Fix this by calling dput() rather
> >> than incrementing fc->ctl_ndents when new_inode() failed.
> >
> > That looks bloody awful.  Sure, everything that accesses fc->ctl_dentry is
> > synchronous with this, but it would've been much easier to follow if
> > shoving dentry into that array happened only after it's been fully set
> > up.
> >
> > Incidentally, there's a nasty headache waiting to happen in that code -
> > consider a twit mounting something on that.  And think what happens when
> > connection gets shut down...
> 
> Need to call d_invalidate() instead of d_drop() in there.  Is that
> what you are referring to?

Yes, and do that once on the entire subdirectory...