Re: [PATCH 2/2] autofs4 - fix device ioctl mount lookup

[Ian Kent <raven@themaw.net>]

On Sun, 2013-09-08 at 07:33 -0400, Jeff Layton wrote:
> On Sun, 08 Sep 2013 16:47:23 +0800
> Ian Kent <raven@themaw.net> wrote:
> 
> > When reconnecting to automounts at startup an autofs ioctl is used
> > to find the device and inode of existing mounts so they can be used
> > to open a file descriptor of possibly covered mounts.
> > 
> > At this time the the caller might not yet "own" the mount so it can
> > trigger calling ->d_automount(). This causes automount to hang when
> > trying to reconnect to direct or offset mount types.
> > 
> > Consequently kern_path() can't be used but path_mntpointat() can be.
> > 
> > Signed-off-by: Ian Kent <raven@themaw.net>
> > Cc: Jeff Layton <jlayton@redhat.com>
> > Cc: Al Viro <viro@ZenIV.linux.org.uk>
> > ---
> >  fs/autofs4/dev-ioctl.c |   23 ++++++++++++-----------
> >  1 file changed, 12 insertions(+), 11 deletions(-)
> > 
> > diff --git a/fs/autofs4/dev-ioctl.c b/fs/autofs4/dev-ioctl.c
> > index 9183821..228866f 100644
> > --- a/fs/autofs4/dev-ioctl.c
> > +++ b/fs/autofs4/dev-ioctl.c
> > @@ -183,13 +183,14 @@ static int autofs_dev_ioctl_protosubver(struct file *fp,
> >  	return 0;
> >  }
> >  
> > +/* Find the topmost mount satisfying test() */
> >  static int find_autofs_mount(const char *pathname,
> >  			     struct path *res,
> >  			     int test(struct path *path, void *data),
> >  			     void *data)
> >  {
> >  	struct path path;
> > -	int err = kern_path(pathname, 0, &path);
> > +	int err = user_path_mntpointat(AT_FDCWD, pathname, 0, &path);
> 
> This looks wrong. "pathname" is a kernel string, not a __user one. I
> think what you need to do here is to turn user_path_mntpointat into a
> wrapper around a kern_path_mntpointat equivalent and then call that
> here.

In both cases the path comes from a structure passed from user space.
So I started thinking it wasn't correct previously.

> 
> >  	if (err)
> >  		return err;
> >  	err = -ENOENT;
> > @@ -197,10 +198,9 @@ static int find_autofs_mount(const char *pathname,
> >  		if (path.dentry->d_sb->s_magic == AUTOFS_SUPER_MAGIC) {
> >  			if (test(&path, data)) {
> >  				path_get(&path);
> > -				if (!err) /* already found some */
> > -					path_put(res);
> >  				*res = path;
> >  				err = 0;
> > +				break;
> >  			}
> >  		}
> >  		if (!follow_up(&path))
> > @@ -498,12 +498,11 @@ static int autofs_dev_ioctl_askumount(struct file *fp,
> >   * mount if there is one or 0 if it isn't a mountpoint.
> >   *
> >   * If we aren't supplied with a file descriptor then we
> > - * lookup the nameidata of the path and check if it is the
> > - * root of a mount. If a type is given we are looking for
> > - * a particular autofs mount and if we don't find a match
> > - * we return fail. If the located nameidata path is the
> > - * root of a mount we return 1 along with the super magic
> > - * of the mount or 0 otherwise.
> > + * lookup the path and check if it is the root of a mount.
> > + * If a type is given we are looking for a particular autofs
> > + * mount and if we don't find a match we return fail. If the
> > + * located path is the root of a mount we return 1 along with
> > + * the super magic of the mount or 0 otherwise.
> >   *
> >   * In both cases the the device number (as returned by
> >   * new_encode_dev()) is also returned.
> > @@ -531,9 +530,11 @@ static int autofs_dev_ioctl_ismountpoint(struct file *fp,
> >  
> >  	if (!fp || param->ioctlfd == -1) {
> >  		if (autofs_type_any(type))
> > -			err = kern_path(name, LOOKUP_FOLLOW, &path);
> > +			err = user_path_mntpointat(AT_FDCWD,
> > +						   name, LOOKUP_FOLLOW, &path);
> >  		else
> > -			err = find_autofs_mount(name, &path, test_by_type, &type);
> > +			err = find_autofs_mount(name, &path,
> > +						test_by_type, &type);
> 
> 
> ...ditto in these spots of course...
> 
> >  		if (err)
> >  			goto out;
> >  		devid = new_encode_dev(path.dentry->d_sb->s_dev);
> > 
> 
>