[IPv6]: Export userland ND options through netlink (RDNSS support)
Pierre Ynard [Thu, 11 Oct 2007 04:22:05 +0000 (21:22 -0700)]
As discussed before, this patch provides userland with a way to access
relevant options in Router Advertisements, after they are processed
and validated by the kernel. Extra options are processed in a generic
way; this patch only exports RDNSS options described in RFC5006, but
support to control which options are exported could be easily added.

A new rtnetlink message type is defined, to transport Neighbor
Discovery options, along with optional context information. At the
moment only the address of the router sending an RDNSS option is
included, but additional attributes may be later defined, if needed by
new use cases.

Signed-off-by: Pierre Ynard <linkfanel@yahoo.fr>
Signed-off-by: David S. Miller <davem@davemloft.net>

include/linux/rtnetlink.h
include/net/ndisc.h
net/ipv6/ndisc.c

index dff3192..5bf6182 100644 (file)
@@ -97,6 +97,9 @@ enum {
        RTM_SETNEIGHTBL,
 #define RTM_SETNEIGHTBL        RTM_SETNEIGHTBL
 
+       RTM_NEWNDUSEROPT = 68,
+#define RTM_NEWNDUSEROPT RTM_NEWNDUSEROPT
+
        __RTM_MAX,
 #define RTM_MAX                (((__RTM_MAX + 3) & ~3) - 1)
 };
@@ -479,6 +482,30 @@ enum
 #define TCA_RTA(r)  ((struct rtattr*)(((char*)(r)) + NLMSG_ALIGN(sizeof(struct tcmsg))))
 #define TCA_PAYLOAD(n) NLMSG_PAYLOAD(n,sizeof(struct tcmsg))
 
+/********************************************************************
+ *             Neighbor Discovery userland options
+ ****/
+
+struct nduseroptmsg
+{
+       unsigned char   nduseropt_family;
+       unsigned char   nduseropt_pad1;
+       unsigned short  nduseropt_opts_len;     /* Total length of options */
+       __u8            nduseropt_icmp_type;
+       __u8            nduseropt_icmp_code;
+       unsigned short  nduseropt_pad2;
+       /* Followed by one or more ND options */
+};
+
+enum
+{
+       NDUSEROPT_UNSPEC,
+       NDUSEROPT_SRCADDR,
+       __NDUSEROPT_MAX
+};
+
+#define NDUSEROPT_MAX  (__NDUSEROPT_MAX - 1)
+
 #ifndef __KERNEL__
 /* RTnetlink multicast groups - backwards compatibility for userspace */
 #define RTMGRP_LINK            1
@@ -542,6 +569,8 @@ enum rtnetlink_groups {
 #define RTNLGRP_IPV6_PREFIX    RTNLGRP_IPV6_PREFIX
        RTNLGRP_IPV6_RULE,
 #define RTNLGRP_IPV6_RULE      RTNLGRP_IPV6_RULE
+       RTNLGRP_ND_USEROPT,
+#define RTNLGRP_ND_USEROPT     RTNLGRP_ND_USEROPT
        __RTNLGRP_MAX
 };
 #define RTNLGRP_MAX    (__RTNLGRP_MAX - 1)
index 475b10c..6684f7e 100644 (file)
@@ -24,6 +24,7 @@ enum {
        ND_OPT_MTU = 5,                 /* RFC2461 */
        __ND_OPT_ARRAY_MAX,
        ND_OPT_ROUTE_INFO = 24,         /* RFC4191 */
+       ND_OPT_RDNSS = 25,              /* RFC5006 */
        __ND_OPT_MAX
 };
 
index d4acd28..6cc33dc 100644 (file)
 /*
  *     Changes:
  *
+ *     Pierre Ynard                    :       export userland ND options
+ *                                             through netlink (RDNSS support)
  *     Lars Fenneberg                  :       fixed MTU setting on receipt
  *                                             of an RA.
- *
  *     Janos Farkas                    :       kmalloc failure checks
  *     Alexey Kuznetsov                :       state machine reworked
  *                                             and moved to net/core.
@@ -78,6 +79,9 @@
 #include <net/addrconf.h>
 #include <net/icmp.h>
 
+#include <net/netlink.h>
+#include <linux/rtnetlink.h>
+
 #include <net/flow.h>
 #include <net/ip6_checksum.h>
 #include <linux/proc_fs.h>
@@ -161,6 +165,8 @@ struct ndisc_options {
        struct nd_opt_hdr *nd_opts_ri;
        struct nd_opt_hdr *nd_opts_ri_end;
 #endif
+       struct nd_opt_hdr *nd_useropts;
+       struct nd_opt_hdr *nd_useropts_end;
 };
 
 #define nd_opts_src_lladdr     nd_opt_array[ND_OPT_SOURCE_LL_ADDR]
@@ -225,6 +231,22 @@ static struct nd_opt_hdr *ndisc_next_option(struct nd_opt_hdr *cur,
        return (cur <= end && cur->nd_opt_type == type ? cur : NULL);
 }
 
+static inline int ndisc_is_useropt(struct nd_opt_hdr *opt)
+{
+       return (opt->nd_opt_type == ND_OPT_RDNSS);
+}
+
+static struct nd_opt_hdr *ndisc_next_useropt(struct nd_opt_hdr *cur,
+                                            struct nd_opt_hdr *end)
+{
+       if (!cur || !end || cur >= end)
+               return NULL;
+       do {
+               cur = ((void *)cur) + (cur->nd_opt_len << 3);
+       } while(cur < end && !ndisc_is_useropt(cur));
+       return (cur <= end && ndisc_is_useropt(cur) ? cur : NULL);
+}
+
 static struct ndisc_options *ndisc_parse_options(u8 *opt, int opt_len,
                                                 struct ndisc_options *ndopts)
 {
@@ -267,14 +289,21 @@ static struct ndisc_options *ndisc_parse_options(u8 *opt, int opt_len,
                        break;
 #endif
                default:
-                       /*
-                        * Unknown options must be silently ignored,
-                        * to accommodate future extension to the protocol.
-                        */
-                       ND_PRINTK2(KERN_NOTICE
-                                  "%s(): ignored unsupported option; type=%d, len=%d\n",
-                                  __FUNCTION__,
-                                  nd_opt->nd_opt_type, nd_opt->nd_opt_len);
+                       if (ndisc_is_useropt(nd_opt)) {
+                               ndopts->nd_useropts_end = nd_opt;
+                               if (!ndopts->nd_useropts)
+                                       ndopts->nd_useropts = nd_opt;
+                       } else {
+                               /*
+                                * Unknown options must be silently ignored,
+                                * to accommodate future extension to the
+                                * protocol.
+                                */
+                               ND_PRINTK2(KERN_NOTICE
+                                          "%s(): ignored unsupported option; type=%d, len=%d\n",
+                                          __FUNCTION__,
+                                          nd_opt->nd_opt_type, nd_opt->nd_opt_len);
+                       }
                }
                opt_len -= l;
                nd_opt = ((void *)nd_opt) + l;
@@ -984,6 +1013,53 @@ out:
        in6_dev_put(idev);
 }
 
+static void ndisc_ra_useropt(struct sk_buff *ra, struct nd_opt_hdr *opt)
+{
+       struct icmp6hdr *icmp6h = (struct icmp6hdr *)skb_transport_header(ra);
+       struct sk_buff *skb;
+       struct nlmsghdr *nlh;
+       struct nduseroptmsg *ndmsg;
+       int err;
+       int base_size = NLMSG_ALIGN(sizeof(struct nduseroptmsg)
+                                   + (opt->nd_opt_len << 3));
+       size_t msg_size = base_size + nla_total_size(sizeof(struct in6_addr));
+
+       skb = nlmsg_new(msg_size, GFP_ATOMIC);
+       if (skb == NULL) {
+               err = -ENOBUFS;
+               goto errout;
+       }
+
+       nlh = nlmsg_put(skb, 0, 0, RTM_NEWNDUSEROPT, base_size, 0);
+       if (nlh == NULL) {
+               goto nla_put_failure;
+       }
+
+       ndmsg = nlmsg_data(nlh);
+       ndmsg->nduseropt_family = AF_INET6;
+       ndmsg->nduseropt_icmp_type = icmp6h->icmp6_type;
+       ndmsg->nduseropt_icmp_code = icmp6h->icmp6_code;
+       ndmsg->nduseropt_opts_len = opt->nd_opt_len << 3;
+
+       memcpy(ndmsg + 1, opt, opt->nd_opt_len << 3);
+
+       NLA_PUT(skb, NDUSEROPT_SRCADDR, sizeof(struct in6_addr),
+               &ipv6_hdr(ra)->saddr);
+       nlmsg_end(skb, nlh);
+
+       err = rtnl_notify(skb, 0, RTNLGRP_ND_USEROPT, NULL, GFP_ATOMIC);
+       if (err < 0)
+               goto errout;
+
+       return;
+
+nla_put_failure:
+       nlmsg_free(skb);
+       err = -EMSGSIZE;
+errout:
+       rtnl_set_sk_err(RTNLGRP_ND_USEROPT, err);
+}
+
 static void ndisc_router_discovery(struct sk_buff *skb)
 {
        struct ra_msg *ra_msg = (struct ra_msg *)skb_transport_header(skb);
@@ -1216,6 +1292,15 @@ skip_defrtr:
                }
        }
 
+       if (ndopts.nd_useropts) {
+               struct nd_opt_hdr *opt;
+               for (opt = ndopts.nd_useropts;
+                    opt;
+                    opt = ndisc_next_useropt(opt, ndopts.nd_useropts_end)) {
+                               ndisc_ra_useropt(skb, opt);
+               }
+       }
+
        if (ndopts.nd_opts_tgt_lladdr || ndopts.nd_opts_rh) {
                ND_PRINTK2(KERN_WARNING
                           "ICMPv6 RA: invalid RA options");