Re: [RESEND PATCH] autofs4 - remove string terminator check

[Ian Kent <raven@themaw.net>]

On Tue, 2008-10-28 at 11:11 +0900, Ian Kent wrote:
> On Tue, 2008-10-28 at 11:07 +0900, Ian Kent wrote:
> > On Mon, 2008-10-27 at 18:54 -0700, Andrew Morton wrote:
> > > On Tue, 28 Oct 2008 10:14:30 +0900 Ian Kent <raven@themaw.net> wrote:
> > > 
> > > > In a previous patch a comment was made that checking for the existence of
> > > > a NULL terminator in strings copied from userspace wasn't needed as this
> > > > is done in many places in the kernel without problem. This patch removes
> > > > this string terminator check.
> > > > 
> > > 
> > > ah, OK.  Now I'm worried.
> > > 
> > > > 
> > > >  fs/autofs4/dev-ioctl.c |   20 --------------------
> > > >  1 files changed, 0 insertions(+), 20 deletions(-)
> > > > 
> > > > diff --git a/fs/autofs4/dev-ioctl.c b/fs/autofs4/dev-ioctl.c
> > > > index 625abf5..304c1ff 100644
> > > > --- a/fs/autofs4/dev-ioctl.c
> > > > +++ b/fs/autofs4/dev-ioctl.c
> > > > @@ -51,18 +51,6 @@ static int check_name(const char *name)
> > > >  }
> > > >  
> > > >  /*
> > > > - * Check a string doesn't overrun the chunk of
> > > > - * memory we copied from user land.
> > > > - */
> > > > -static int invalid_str(char *str, void *end)
> > > > -{
> > > > -	while ((void *) str <= end)
> > > > -		if (!*str++)
> > > > -			return 0;
> > > > -	return -EINVAL;
> > > > -}
> > > > -
> > > > -/*
> > > >   * Check that the user compiled against correct version of autofs
> > > >   * misc device code.
> > > >   *
> > > > @@ -143,14 +131,6 @@ static int validate_dev_ioctl(int cmd, struct autofs_dev_ioctl *param)
> > > >  				    cmd);
> > > >  			goto out;
> > > >  		}
> > > > -
> > > > -		err = invalid_str(param->path,
> > > > -				 (void *) ((size_t) param + param->size));
> > > > -		if (err) {
> > > > -			AUTOFS_WARN("invalid path supplied for cmd(0x%08x)",
> > > > -				    cmd);
> > > > -			goto out;
> > > > -		}
> > > >  	}
> > > >  
> > > >  	err = 0;
> > > 
> > > What will now happen if userspace passes in a non-null-terminated
> > > string (if that's possible)?
> > > 
> > > Presumably that isn't possible, or it's never been tested, because
> > > before we check for null-termination we run check_name(), which
> > > _assumes_ null-termination!
> 
> Oh .. hang on, point taken.
> 
> > > 
> > > The comment over validate_dev_ioctl() will need the "and is terminated"
> > > removed after this change, yes?
> > 
> > Yes, but now I think I shouldn't have removed it.
> > For my part I would have been happy to keep this and now I think I
> > should so let's drop this patch.

Just to confirm.
Please drop this patch and I'll send a patch to fix the check order.

> > 
> > Ian
> >