RE: [PATCH net-next] sctp: Pull the user copies out of the individual sockopt functions.

From: David Laight <David.Laight@ACULAB.COM>
To: 'Marcelo Ricardo Leitner' <marcelo.leitner@gmail.com>
Cc: "netdev@vger.kernel.org" <netdev@vger.kernel.org>,
	"linux-sctp@vger.kernel.org" <linux-sctp@vger.kernel.org>,
	Neil Horman <nhorman@tuxdriver.com>
Subject: RE: [PATCH net-next] sctp: Pull the user copies out of the individual sockopt functions.
Date: Thu, 21 May 2020 07:32:14 +0000	[thread overview]
Message-ID: <e777874fbd0e4ccb813e08145f3c3359@AcuMS.aculab.com> (raw)
In-Reply-To: <20200521001725.GW2491@localhost.localdomain>

From: 'Marcelo Ricardo Leitner'
> Sent: 21 May 2020 01:17
> On Wed, May 20, 2020 at 03:08:13PM +0000, David Laight wrote:
> ...
> > Only SCTP_SOCKOPT_CONNECTX3 contains an indirect pointer.
> > It is also the only getsockopt() that wants to return a buffer
> > and an error code. It is also definitely abusing getsockopt().
> ...
> > @@ -1375,11 +1350,11 @@ struct compat_sctp_getaddrs_old {
> >  #endif
> >
> >  static int sctp_getsockopt_connectx3(struct sock *sk, int len,
> > -				     char __user *optval,
> > -				     int __user *optlen)
> > +				     struct sctp_getaddrs_old *param,
> > +				     int *optlen)
> >  {
> > -	struct sctp_getaddrs_old param;
> >  	sctp_assoc_t assoc_id = 0;
> > +	struct sockaddr *addrs;
> >  	int err = 0;
> >
> >  #ifdef CONFIG_COMPAT
..
> >  	} else
> >  #endif
> >  	{
> > -		if (len < sizeof(param))
> > +		if (len < sizeof(*param))
> >  			return -EINVAL;
> > -		if (copy_from_user(&param, optval, sizeof(param)))
> > -			return -EFAULT;
> >  	}
> >
> > -	err = __sctp_setsockopt_connectx(sk, (struct sockaddr __user *)
> > -					 param.addrs, param.addr_num,
> > +	addrs = memdup_user(param->addrs, param->addr_num);
> 
> I'm staring at this for a while now but I don't get this memdup_user.
> AFAICT, params->addrs is not __user anymore here, because
> sctp_getsockopt() copied the whole thing already, no?
> Also weird because it is being called from kernel_sctp_getsockopt(),
> which now has no knowledge of __user buffers.
> Maybe I didn't get something from the patch description.

The connectx3 sockopt buffer contains a pointer to the user buffer
that contains the actual addresses.
So a second copy_from_user() is needed.

This does mean that this option can only be actioned from userspace.

Kernel code can get the same functionality using one of the
other interfaces to connectx().

	David

-
Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK
Registration No: 1397386 (Wales)