On Tue, Mar 06, 2018 at 03:03:13PM +0800, Xin Long wrote:
> On Tue, Mar 6, 2018 at 7:39 AM, Marcelo Ricardo Leitner
> <marcelo.leitner@xxxxxxxxx> wrote:
> > On Mon, Mar 05, 2018 at 08:44:19PM +0800, Xin Long wrote:
> >> This patch is to add support for Destination IPv4/6 Address options
> >> for sendmsg, as described in section 5.3.9/10 of RFC6458.
> >>
> >> With this option, you can provide more than one destination addrs
> >> to sendmsg when creating asoc, like sctp_connectx.
> >>
> >> It's also a necessary send info for sctp_sendv.
> >>
> >> Signed-off-by: Xin Long <lucien.xin@xxxxxxxxx>
> >> ---
> >> include/net/sctp/structs.h | 1 +
> >> include/uapi/linux/sctp.h | 6 ++++
> >> net/sctp/socket.c | 77 ++++++++++++++++++++++++++++++++++++++++++++++
> >> 3 files changed, 84 insertions(+)
> >>
> >> diff --git a/include/net/sctp/structs.h b/include/net/sctp/structs.h
> >> index d40a2a3..ec6e46b 100644
> >> --- a/include/net/sctp/structs.h
> >> +++ b/include/net/sctp/structs.h
> >> @@ -2113,6 +2113,7 @@ struct sctp_cmsgs {
> >> struct sctp_sndrcvinfo *srinfo;
> >> struct sctp_sndinfo *sinfo;
> >> struct sctp_prinfo *prinfo;
> >> + struct msghdr *addrs_msg;
> >> };
> >>
> >> /* Structure for tracking memory objects */
> >> diff --git a/include/uapi/linux/sctp.h b/include/uapi/linux/sctp.h
> >> index 0dd1f82..a1bc350 100644
> >> --- a/include/uapi/linux/sctp.h
> >> +++ b/include/uapi/linux/sctp.h
> >> @@ -308,6 +308,12 @@ typedef enum sctp_cmsg_type {
> >> #define SCTP_NXTINFO SCTP_NXTINFO
> >> SCTP_PRINFO, /* 5.3.7 SCTP PR-SCTP Information Structure */
> >> #define SCTP_PRINFO SCTP_PRINFO
> >> + SCTP_AUTHINFO, /* 5.3.8 SCTP AUTH Information Structure (RESERVED) */
> >> +#define SCTP_AUTHINFO SCTP_AUTHINFO
> >> + SCTP_DSTADDRV4, /* 5.3.9 SCTP Destination IPv4 Address Structure */
> >> +#define SCTP_DSTADDRV4 SCTP_DSTADDRV4
> >> + SCTP_DSTADDRV6, /* 5.3.10 SCTP Destination IPv6 Address Structure */
> >> +#define SCTP_DSTADDRV6 SCTP_DSTADDRV6
> >> } sctp_cmsg_t;
> >>
> >> /*
> >> diff --git a/net/sctp/socket.c b/net/sctp/socket.c
> >> index fdde697..067b57a 100644
> >> --- a/net/sctp/socket.c
> >> +++ b/net/sctp/socket.c
> >> @@ -1676,6 +1676,7 @@ static int sctp_sendmsg_new_asoc(struct sock *sk, __u16 sflags,
> >> struct net *net = sock_net(sk);
> >> struct sctp_association *asoc;
> >> enum sctp_scope scope;
> >> + struct cmsghdr *cmsg;
> >> int err = -EINVAL;
> >>
> >> *tp = NULL;
> >> @@ -1741,6 +1742,67 @@ static int sctp_sendmsg_new_asoc(struct sock *sk, __u16 sflags,
> >> goto free;
> >> }
> >>
> >> + if (!cmsgs->addrs_msg)
> >> + return 0;
> >> +
> >> + /* sendv addr list parse */
> >> + for_each_cmsghdr(cmsg, cmsgs->addrs_msg) {
> >> + struct sctp_transport *transport;
> >> + struct sctp_association *old;
> >> + union sctp_addr _daddr;
> >> + int dlen;
> >> +
> >> + if (cmsg->cmsg_level != IPPROTO_SCTP ||
> >> + (cmsg->cmsg_type != SCTP_DSTADDRV4 &&
> >> + cmsg->cmsg_type != SCTP_DSTADDRV6))
> >> + continue;
> >> +
> >> + daddr = &_daddr;
> >> + memset(daddr, 0, sizeof(*daddr));
> >> + dlen = cmsg->cmsg_len - sizeof(struct cmsghdr);
> >> + if (cmsg->cmsg_type == SCTP_DSTADDRV4) {
> >> + if (dlen < sizeof(struct in_addr))
> >> + goto free;
> >> +
> >> + dlen = sizeof(struct in_addr);
> >> + daddr->v4.sin_family = AF_INET;
> >> + daddr->v4.sin_port = htons(asoc->peer.port);
> >> + memcpy(&daddr->v4.sin_addr, CMSG_DATA(cmsg), dlen);
> >> + } else {
> >> + if (dlen < sizeof(struct in6_addr))
> >> + goto free;
> >> +
> >> + dlen = sizeof(struct in6_addr);
> >> + daddr->v6.sin6_family = AF_INET6;
> >> + daddr->v6.sin6_port = htons(asoc->peer.port);
> >> + memcpy(&daddr->v6.sin6_addr, CMSG_DATA(cmsg), dlen);
> >> + }
> >> + err = sctp_verify_addr(sk, daddr, sizeof(*daddr));
> >> + if (err)
> >> + goto free;
> >> +
> >> + old = sctp_endpoint_lookup_assoc(ep, daddr, &transport);
> >> + if (old && old != asoc) {
> >> + if (old->state >= SCTP_STATE_ESTABLISHED)
> >> + err = -EISCONN;
> >> + else
> >> + err = -EALREADY;
> >> + goto free;
> >> + }
> >> +
> >> + if (sctp_endpoint_is_peeled_off(ep, daddr)) {
> >> + err = -EADDRNOTAVAIL;
> >> + goto free;
> >> + }
> >> +
> >> + transport = sctp_assoc_add_peer(asoc, daddr, GFP_KERNEL,
> >> + SCTP_UNKNOWN);
> >> + if (!transport) {
> >> + err = -ENOMEM;
> >> + goto free;
> >> + }
> >> + }
> >> +
> >> return 0;
> >>
> >> free:
> >> @@ -7778,6 +7840,21 @@ static int sctp_msghdr_parse(const struct msghdr *msg, struct sctp_cmsgs *cmsgs)
> >> if (cmsgs->prinfo->pr_policy == SCTP_PR_SCTP_NONE)
> >> cmsgs->prinfo->pr_value = 0;
> >> break;
> >> + case SCTP_DSTADDRV4:
> >> + case SCTP_DSTADDRV6:
> >> + /* SCTP Socket API Extension
> >> + * 5.3.9/10 SCTP Destination IPv4/6 Address Structure (SCTP_DSTADDRV4/6)
> >> + *
> >> + * This cmsghdr structure specifies SCTP options for sendmsg().
> >> + *
> >> + * cmsg_level cmsg_type cmsg_data[]
> >> + * ------------ ------------ ---------------------
> >> + * IPPROTO_SCTP SCTP_DSTADDRV4 struct in_addr
> >> + * ------------ ------------ ---------------------
> >> + * IPPROTO_SCTP SCTP_DSTADDRV6 struct in6_addr
> >> + */
> >> + cmsgs->addrs_msg = my_msg;
> >
> > I think this is being handled differently from what the RFC specifies.
> > AFAIU the rfc defines these two options being each one an array of a
> > specific element type, in_addr and in6_addr.
> Hi, Marcelo.
>
> From section 5.3.9 of RFC6458:
>
> This ancillary data can be used to provide more than one destination
> address to sendmsg(). It can be used to implement sctp_sendv() using
> sendmsg().
>
> It doesn't say to put "more than one destination address" into an array of
> one option, or put "more than one destination address" into many options.
Spot on. And I went with the first interpretation, thus why the
difference.
>
> I was also confused, then I checked sctp_process_cmsgs_for_init() in
> freebsd, I think it's in MANY options.
Nice, ok
>
>
> >
> > The patch is saving both options into a single pointer, which later is
> > processed as a list of (in_addr or in6_addr) elements.
> It's not, 'my_msg' is the whole msghdr, not only in_addr or in6_addr.
> later in sctp_sendmsg_new_asoc, it will parse this msghdr again but
> this time only to pick up SCTP_DSTADDRV4/6 options.
> (by this way I don't need to alloc list or array to save them)
Which is nice, btw.
>
> Note that: one option is just one address, NOT a list of addresses.
> like, it can even be:
> SCTP_DSTADDRV4 (one option)
> SCTP_DSTADDRV6 (one option)
> SCTP_DSTADDRV4 (one option)
> SCTP_DSTADDRV4 (one option)
> SCTP_DSTADDRV6 (one option)
>
> Maybe the name of cmsgs->addrs_msg is confusing here, but I
> actually think it as a flag to see if there is addrs info in the msghdr,
> and also as a parameter to pass the msghdr into sctp_sendmsg_new_asoc.
What was more confusing to me is that it seems to overwrite the
pointer at every other addr option, but I should have noticed that it
is saving the entire cmsg.
>
> >
> > If an application specifies both options, each with a list of its own
> > addresses, it will ignore one of the options.
> If an application specifies many options, some are SCTP_DSTADDRV4
> and some are SCTP_DSTADDRV6, none of them will be ignored in
> sctp_sendmsg_new_asoc().
Yep. Thanks.
>
> >
> >> + break;
> >> default:
> >> return -EINVAL;
> >> }
> >> --
> >> 2.1.0
> >>
--
To unsubscribe from this list: send the line "unsubscribe linux-sctp" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
