Date:   Fri, 16 Feb 2018 23:28:09 -0500
From:   Neil Horman <nhorman@...driver.com>
To:     Marcelo Ricardo Leitner <marcelo.leitner@...il.com>
Cc:     Richard Haines <richard_c_haines@...nternet.com>,
        selinux@...ho.nsa.gov, netdev@...r.kernel.org,
        linux-sctp@...r.kernel.org, linux-security-module@...r.kernel.org,
        paul@...l-moore.com, vyasevich@...il.com, sds@...ho.nsa.gov,
        eparis@...isplace.org, casey@...aufler-ca.com,
        james.l.morris@...cle.com
Subject: Re: [PATCH V6 2/4] sctp: Add ip option support

On Fri, Feb 16, 2018 at 07:51:02PM -0200, Marcelo Ricardo Leitner wrote:
> On Fri, Feb 16, 2018 at 03:14:35PM -0500, Neil Horman wrote:
> > On Fri, Feb 16, 2018 at 10:56:07AM -0200, Marcelo Ricardo Leitner wrote:
> > > On Thu, Feb 15, 2018 at 09:15:40AM -0500, Neil Horman wrote:
> > > > On Tue, Feb 13, 2018 at 08:54:44PM +0000, Richard Haines wrote:
> > > > > Add ip option support to allow LSM security modules to utilise CIPSO/IPv4
> > > > > and CALIPSO/IPv6 services.
> > > > > 
> > > > > Signed-off-by: Richard Haines <richard_c_haines@...nternet.com>
> > > > > ---
> > > > >  include/net/sctp/sctp.h    |  4 +++-
> > > > >  include/net/sctp/structs.h |  2 ++
> > > > >  net/sctp/chunk.c           | 12 +++++++-----
> > > > >  net/sctp/ipv6.c            | 42 +++++++++++++++++++++++++++++++++++-------
> > > > >  net/sctp/output.c          |  5 ++++-
> > > > >  net/sctp/protocol.c        | 36 ++++++++++++++++++++++++++++++++++++
> > > > >  net/sctp/socket.c          | 14 ++++++++++----
> > > > >  7 files changed, 97 insertions(+), 18 deletions(-)
> > > > > 
> > > > > diff --git a/include/net/sctp/sctp.h b/include/net/sctp/sctp.h
> > > > > index f7ae6b0..25c5c87 100644
> > > > > --- a/include/net/sctp/sctp.h
> > > > > +++ b/include/net/sctp/sctp.h
> > > > > @@ -441,9 +441,11 @@ static inline int sctp_list_single_entry(struct list_head *head)
> > > > >  static inline int sctp_frag_point(const struct sctp_association *asoc, int pmtu)
> > > > >  {
> > > > >  	struct sctp_sock *sp = sctp_sk(asoc->base.sk);
> > > > > +	struct sctp_af *af = sp->pf->af;
> > > > >  	int frag = pmtu;
> > > > >  
> > > > > -	frag -= sp->pf->af->net_header_len;
> > > > > +	frag -= af->ip_options_len(asoc->base.sk);
> > > > > +	frag -= af->net_header_len;
> > > > >  	frag -= sizeof(struct sctphdr) + sctp_datachk_len(&asoc->stream);
> > > > >  
> > > > >  	if (asoc->user_frag)
> > > > > diff --git a/include/net/sctp/structs.h b/include/net/sctp/structs.h
> > > > > index 03e92dd..ead5fce 100644
> > > > > --- a/include/net/sctp/structs.h
> > > > > +++ b/include/net/sctp/structs.h
> > > > > @@ -491,6 +491,7 @@ struct sctp_af {
> > > > >  	void		(*ecn_capable)(struct sock *sk);
> > > > >  	__u16		net_header_len;
> > > > >  	int		sockaddr_len;
> > > > > +	int		(*ip_options_len)(struct sock *sk);
> > > > >  	sa_family_t	sa_family;
> > > > >  	struct list_head list;
> > > > >  };
> > > > > @@ -515,6 +516,7 @@ struct sctp_pf {
> > > > >  	int (*addr_to_user)(struct sctp_sock *sk, union sctp_addr *addr);
> > > > >  	void (*to_sk_saddr)(union sctp_addr *, struct sock *sk);
> > > > >  	void (*to_sk_daddr)(union sctp_addr *, struct sock *sk);
> > > > > +	void (*copy_ip_options)(struct sock *sk, struct sock *newsk);
> > > > >  	struct sctp_af *af;
> > > > >  };
> > > > >  
> > > > > diff --git a/net/sctp/chunk.c b/net/sctp/chunk.c
> > > > > index 991a530..d5c0ef7 100644
> > > > > --- a/net/sctp/chunk.c
> > > > > +++ b/net/sctp/chunk.c
> > > > > @@ -154,7 +154,6 @@ static void sctp_datamsg_assign(struct sctp_datamsg *msg, struct sctp_chunk *chu
> > > > >  	chunk->msg = msg;
> > > > >  }
> > > > >  
> > > > > -
> > > > >  /* A data chunk can have a maximum payload of (2^16 - 20).  Break
> > > > >   * down any such message into smaller chunks.  Opportunistically, fragment
> > > > >   * the chunks down to the current MTU constraints.  We may get refragmented
> > > > > @@ -171,6 +170,8 @@ struct sctp_datamsg *sctp_datamsg_from_user(struct sctp_association *asoc,
> > > > >  	struct list_head *pos, *temp;
> > > > >  	struct sctp_chunk *chunk;
> > > > >  	struct sctp_datamsg *msg;
> > > > > +	struct sctp_sock *sp;
> > > > > +	struct sctp_af *af;
> > > > >  	int err;
> > > > >  
> > > > >  	msg = sctp_datamsg_new(GFP_KERNEL);
> > > > > @@ -189,9 +190,11 @@ struct sctp_datamsg *sctp_datamsg_from_user(struct sctp_association *asoc,
> > > > >  	/* This is the biggest possible DATA chunk that can fit into
> > > > >  	 * the packet
> > > > >  	 */
> > > > > -	max_data = asoc->pathmtu -
> > > > > -		   sctp_sk(asoc->base.sk)->pf->af->net_header_len -
> > > > > -		   sizeof(struct sctphdr) - sctp_datachk_len(&asoc->stream);
> > > > > +	sp = sctp_sk(asoc->base.sk);
> > > > > +	af = sp->pf->af;
> > > > > +	max_data = asoc->pathmtu - af->net_header_len -
> > > > > +		   sizeof(struct sctphdr) - sctp_datachk_len(&asoc->stream) -
> > > > > +		   af->ip_options_len(asoc->base.sk);
> > > > >  	max_data = SCTP_TRUNC4(max_data);
> > > > >  
> > > > >  	/* If the the peer requested that we authenticate DATA chunks
> > > > > @@ -211,7 +214,6 @@ struct sctp_datamsg *sctp_datamsg_from_user(struct sctp_association *asoc,
> > > > >  
> > > > >  	/* Set first_len and then account for possible bundles on first frag */
> > > > >  	first_len = max_data;
> > > > > -
> > > > >  	/* Check to see if we have a pending SACK and try to let it be bundled
> > > > >  	 * with this message.  Do this if we don't have any data queued already.
> > > > >  	 * To check that, look at out_qlen and retransmit list.
> > > > > diff --git a/net/sctp/ipv6.c b/net/sctp/ipv6.c
> > > > > index e35d4f7..0b0f895 100644
> > > > > --- a/net/sctp/ipv6.c
> > > > > +++ b/net/sctp/ipv6.c
> > > > > @@ -427,6 +427,38 @@ static void sctp_v6_copy_addrlist(struct list_head *addrlist,
> > > > >  	rcu_read_unlock();
> > > > >  }
> > > > >  
> > > > > +/* Copy over any ip options */
> > > > > +static void sctp_v6_copy_ip_options(struct sock *sk, struct sock *newsk)
> > > > > +{
> > > > > +	struct ipv6_pinfo *newnp, *np = inet6_sk(sk);
> > > > > +	struct ipv6_txoptions *opt;
> > > > > +
> > > > > +	newnp = inet6_sk(newsk);
> > > > > +
> > > > > +	rcu_read_lock();
> > > > > +	opt = rcu_dereference(np->opt);
> > > > > +	if (opt)
> > > > > +		opt = ipv6_dup_options(newsk, opt);
> > > > 		do you want to print a warning here in the event the allocation
> > > > for the dup operation fails?
> > > > 
> > > > > +	RCU_INIT_POINTER(newnp->opt, opt);
> > > > > +	rcu_read_unlock();
> > > > > +}
> > > > > +
> > > > > +/* Account for the IP options */
> > > > > +static int sctp_v6_ip_options_len(struct sock *sk)
> > > > > +{
> > > > > +	struct ipv6_pinfo *np = inet6_sk(sk);
> > > > > +	struct ipv6_txoptions *opt;
> > > > > +	int len = 0;
> > > > > +
> > > > > +	rcu_read_lock();
> > > > > +	opt = rcu_dereference(np->opt);
> > > > > +	if (opt)
> > > > > +		len = opt->opt_flen + opt->opt_nflen;
> > > > > +
> > > > > +	rcu_read_unlock();
> > > > > +	return len;
> > > > > +}
> > > > > +
> > > > >  /* Initialize a sockaddr_storage from in incoming skb. */
> > > > >  static void sctp_v6_from_skb(union sctp_addr *addr, struct sk_buff *skb,
> > > > >  			     int is_saddr)
> > > > > @@ -666,7 +698,6 @@ static struct sock *sctp_v6_create_accept_sk(struct sock *sk,
> > > > >  	struct sock *newsk;
> > > > >  	struct ipv6_pinfo *newnp, *np = inet6_sk(sk);
> > > > >  	struct sctp6_sock *newsctp6sk;
> > > > > -	struct ipv6_txoptions *opt;
> > > > >  
> > > > >  	newsk = sk_alloc(sock_net(sk), PF_INET6, GFP_KERNEL, sk->sk_prot, kern);
> > > > >  	if (!newsk)
> > > > > @@ -689,12 +720,7 @@ static struct sock *sctp_v6_create_accept_sk(struct sock *sk,
> > > > >  	newnp->ipv6_ac_list = NULL;
> > > > >  	newnp->ipv6_fl_list = NULL;
> > > > >  
> > > > > -	rcu_read_lock();
> > > > > -	opt = rcu_dereference(np->opt);
> > > > > -	if (opt)
> > > > > -		opt = ipv6_dup_options(newsk, opt);
> > > > > -	RCU_INIT_POINTER(newnp->opt, opt);
> > > > > -	rcu_read_unlock();
> > > > > +	sctp_v6_copy_ip_options(sk, newsk);
> > > > >  
> > > > >  	/* Initialize sk's sport, dport, rcv_saddr and daddr for getsockname()
> > > > >  	 * and getpeername().
> > > > > @@ -1041,6 +1067,7 @@ static struct sctp_af sctp_af_inet6 = {
> > > > >  	.ecn_capable	   = sctp_v6_ecn_capable,
> > > > >  	.net_header_len	   = sizeof(struct ipv6hdr),
> > > > >  	.sockaddr_len	   = sizeof(struct sockaddr_in6),
> > > > > +	.ip_options_len	   = sctp_v6_ip_options_len,
> > > > >  #ifdef CONFIG_COMPAT
> > > > >  	.compat_setsockopt = compat_ipv6_setsockopt,
> > > > >  	.compat_getsockopt = compat_ipv6_getsockopt,
> > > > > @@ -1059,6 +1086,7 @@ static struct sctp_pf sctp_pf_inet6 = {
> > > > >  	.addr_to_user  = sctp_v6_addr_to_user,
> > > > >  	.to_sk_saddr   = sctp_v6_to_sk_saddr,
> > > > >  	.to_sk_daddr   = sctp_v6_to_sk_daddr,
> > > > > +	.copy_ip_options = sctp_v6_copy_ip_options,
> > > > >  	.af            = &sctp_af_inet6,
> > > > >  };
> > > > >  
> > > > > diff --git a/net/sctp/output.c b/net/sctp/output.c
> > > > > index 01a26ee..668e2fa 100644
> > > > > --- a/net/sctp/output.c
> > > > > +++ b/net/sctp/output.c
> > > > > @@ -151,7 +151,10 @@ void sctp_packet_init(struct sctp_packet *packet,
> > > > >  	INIT_LIST_HEAD(&packet->chunk_list);
> > > > >  	if (asoc) {
> > > > >  		struct sctp_sock *sp = sctp_sk(asoc->base.sk);
> > > > > -		overhead = sp->pf->af->net_header_len;
> > > > > +		struct sctp_af *af = sp->pf->af;
> > > > > +
> > > > > +		overhead = af->net_header_len +
> > > > > +			   af->ip_options_len(asoc->base.sk);
> > > > >  	} else {
> > > > >  		overhead = sizeof(struct ipv6hdr);
> > > > >  	}
> > > > I'm a bit worried about this mechanism.  Unlike tcp or udp, where a packet is
> > > > allocated its options field is pushed within the same call stack (or more
> > > > notably, during a single cycle in which the sock lock is held), sctp allocates a
> > > > packet here, and holds it for potentially multiple calls from userspace while
> > > > chunks are collected and added to it.  During those multiple calls the socket
> > > 
> > > Not sure if you simplified it here but that's not exactly how it
> > > works. The packet is not built chunk by chunk per sendmsg() call as
> > > you described, but instead it will collect the chunks in a list (outq)
> > > up to the point that it notices that it's time to send the packet.
> > > Then, it will call sctp_outq_flush(), which will assemble the packet
> > > and send to IP layer as needed. Chunks that won't fit on the packet
> > > and that will only be sent later, they aren't added to any packet but
> > > remains on outq list.
> > > 
> > Yes, I simplified it, and yes, given that I've maintained this code
> > since 2012, I know how it works.
> 
> That's really not how I meant it. I had to read the paragraph 3 times
> before seeing the simplification.  But Richard is not that acquainted
> with the code and the simplification was to say the least risky for
> his understanding and the implementation he is doing.
> 
> > 
> > > The packet is never freed, it's embedded into the transport. It's just
> > > reconfigured.
> > > 
> > > Nevertheless, I agree the issue is there.
> > > 
> > Which is really the salient point.
> > 
> > > > lock is released and reaquired, during which time the set of configured ip
> > > > options might change.  Then when the packet is passed to the ip layer and the
> > > > options copied into the packet, we might have a different option length leading
> > > > to an skb_over panic.
> > > > 
> > > > Suggest that it might be better to buffer any changes in options and only have
> > > > them take effect any time a new packet is allocated.
> > > 
> > > s/allocated/configured/ :)
> > Seriously? You clearly knew what I was saying. I understand I misued the term,
>                               yes, ^ I knew, but I doubt Richard also knew
> 
> Sorry if somehow I had this connotation. That wasn't the idea.
> 
Its ok I'm sorry as well.  You were responding to me, and I perfectly well know
how this code works, So I assumed you were trying to explain it to me.

> > but do you really want to harp on it? 
> 
> No. Just want Richard to understand what is meant here.
> sctp_packet_init and sctp_packet_config are two distinct moments in
> there.
> 
> > 
> > > I think we can fix it by moving this code to sctp_packet_config()
> > > instead. On a quick check here, seems all packet->overhead references
> > > are after it gets called for a packet that is about to be built.
> > > 
> > Yeah, I agree we could move it there, though I think I would prefer to see it in
> > sctp_packet_transmit.  If we do it there, then we only have to compute the
> > overhead size once before building the skb (that is to say, if we do it in
> > packet_config, we potentially compute it multiple times if we change
> > transports). In fact, if we do it in sct_packet_transmit, then we potentially
> > can eliminate the overhead member from the sctp_packet struct entirely, as we
> > can just store the computed overhead in a stack variable and use it in the
> > skb_reserve call.
> 
> in sctp_packet_transmit I think it would be too late because we have
> to know the entire overhead in upfront in order to know if the chunks
> that are getting enlisted on the packet actually fit in there.
> 
>   Marcelo
Thats a fair point, but we have a backoff path (the err label) in
sctp_packet_trasmit.  If the ip options means a chunk doesn't fit, we  follow
the error path, reset the packet, and try again.  Its slow, to be sure, but I
wonder what the trade off is a net gain(i.e. is it better to hit the rare case where ip
options change and cause a packet to get discarded and retransmitted, or better
to somewhat more frequently recompute the overhead lengtha net gain)

I suppose its a bit academic.  We're talking about a few memory dereferences and
an add or two.  Lets just go with sctp_packet_config for the overhead
computation location.

Neil
> 

Powered by blists - more mailing lists