From 2d14045224f2233aed386eddf155d10a81892c3f Mon Sep 17 00:00:00 2001 From: Martin Mares Date: Wed, 3 Mar 1999 19:49:56 +0000 Subject: Rewrote the kernel syncer. The old layering was horrible. The new kernel syncer is cleanly split between generic UNIX module and OS dependent submodules: - krt.c (the generic part) - krt-iface (low-level functions for interface handling) - krt-scan (low-level functions for routing table scanning) - krt-set (low-level functions for setting of kernel routes) krt-set and krt-iface are common for all BSD-like Unices, krt-scan is heavily system dependent (most Unices require /dev/kmem parsing, Linux uses /proc), Netlink substitues all three modules. We expect each UNIX port supports kernel routing table scanning, kernel interface table scanning, kernel route manipulation and possibly also asynchronous event notifications (new route, interface state change; not implemented yet) and build the KRT protocol on the top of these primitive operations. --- sysdep/linux/netlink/Modules | 8 +- sysdep/linux/netlink/krt-iface.h | 26 +++ sysdep/linux/netlink/krt-scan.h | 26 +++ sysdep/linux/netlink/krt-set.h | 26 +++ sysdep/linux/netlink/krt.Y | 46 ---- sysdep/linux/netlink/krt.c | 447 --------------------------------------- sysdep/linux/netlink/krt.h | 30 --- sysdep/linux/netlink/netlink.Y | 25 +++ sysdep/linux/netlink/netlink.c | 440 ++++++++++++++++++++++++++++++++++++++ 9 files changed, 548 insertions(+), 526 deletions(-) create mode 100644 sysdep/linux/netlink/krt-iface.h create mode 100644 sysdep/linux/netlink/krt-scan.h create mode 100644 sysdep/linux/netlink/krt-set.h delete mode 100644 sysdep/linux/netlink/krt.Y delete mode 100644 sysdep/linux/netlink/krt.c delete mode 100644 sysdep/linux/netlink/krt.h create mode 100644 sysdep/linux/netlink/netlink.Y create mode 100644 sysdep/linux/netlink/netlink.c (limited to 'sysdep/linux/netlink') diff --git a/sysdep/linux/netlink/Modules b/sysdep/linux/netlink/Modules index 8033810..c26f7f7 100644 --- a/sysdep/linux/netlink/Modules +++ b/sysdep/linux/netlink/Modules @@ -1,3 +1,5 @@ -krt.c -krt.h -krt.Y +krt-iface.h +krt-set.h +krt-scan.h +netlink.c +netlink.Y diff --git a/sysdep/linux/netlink/krt-iface.h b/sysdep/linux/netlink/krt-iface.h new file mode 100644 index 0000000..5dfb934 --- /dev/null +++ b/sysdep/linux/netlink/krt-iface.h @@ -0,0 +1,26 @@ +/* + * BIRD -- Unix Kernel Netlink Interface Syncer -- Dummy Include File + * + * (c) 1998--1999 Martin Mares + * + * Can be freely distributed and used under the terms of the GNU GPL. + */ + +#ifndef _BIRD_KRT_IFACE_H_ +#define _BIRD_KRT_IFACE_H_ + +/* + * We don't have split iface/scan/set parts. See krt-scan.h. + */ + +struct krt_if_params { +}; + +struct krt_if_status { +}; + +static inline void krt_if_preconfig(struct krt_config *c) { }; +static inline void krt_if_start(struct krt_proto *p) { }; +static inline void krt_if_shutdown(struct krt_proto *p) { }; + +#endif diff --git a/sysdep/linux/netlink/krt-scan.h b/sysdep/linux/netlink/krt-scan.h new file mode 100644 index 0000000..cd118de --- /dev/null +++ b/sysdep/linux/netlink/krt-scan.h @@ -0,0 +1,26 @@ +/* + * BIRD -- Linux Kernel Netlink Route Syncer -- Scanning + * + * (c) 1998--1999 Martin Mares + * + * Can be freely distributed and used under the terms of the GNU GPL. + */ + +#ifndef _BIRD_KRT_SCAN_H_ +#define _BIRD_KRT_SCAN_H_ + +/* + * We don't have split iface/scan/set for Netlink. All options + * and run-time parameters are declared here instead of splitting + * to krt-set.h, krt-iface.h and this file. + */ + +struct krt_scan_params { + int async; /* Allow asynchronous events */ +}; + +struct krt_scan_status { + list temp_ifs; /* Temporary interfaces */ +}; + +#endif diff --git a/sysdep/linux/netlink/krt-set.h b/sysdep/linux/netlink/krt-set.h new file mode 100644 index 0000000..32fdad0 --- /dev/null +++ b/sysdep/linux/netlink/krt-set.h @@ -0,0 +1,26 @@ +/* + * BIRD -- Unix Kernel Netlink Route Syncer -- Dummy Include File + * + * (c) 1998--1999 Martin Mares + * + * Can be freely distributed and used under the terms of the GNU GPL. + */ + +#ifndef _BIRD_KRT_SET_H_ +#define _BIRD_KRT_SET_H_ + +/* + * We don't have split iface/scan/set parts. See krt-scan.h. + */ + +struct krt_set_params { +}; + +struct krt_set_status { +}; + +static inline void krt_set_preconfig(struct krt_config *c) { }; +static inline void krt_set_start(struct krt_proto *p) { }; +static inline void krt_set_shutdown(struct krt_proto *p) { }; + +#endif diff --git a/sysdep/linux/netlink/krt.Y b/sysdep/linux/netlink/krt.Y deleted file mode 100644 index c572640..0000000 --- a/sysdep/linux/netlink/krt.Y +++ /dev/null @@ -1,46 +0,0 @@ -/* - * BIRD -- Netlink Interface Configuration - * - * (c) 1999 Martin Mares - * - * Can be freely distributed and used under the terms of the GNU GPL. - */ - -CF_HDR - -#include "lib/krt.h" - -#define KRT_PROTO ((struct krt_config *) this_proto) - -CF_DECLS - -CF_KEYWORDS(KERNEL, PERSIST, ROUTE, SCAN, TIME, LEARN) - -CF_GRAMMAR - -/* Kernel protocol */ - -CF_ADDTO(proto, kern_proto '}') - -kern_proto_start: proto_start KERNEL { - if (!(this_proto = cf_krt)) cf_error("Kernel protocol already defined"); - cf_krt = NULL; - } - ; - -kern_proto: - kern_proto_start '{' - | kern_proto proto_item ';' - | kern_proto kern_item ';' - ; - -kern_item: - PERSIST bool { KRT_PROTO->persist = $2; } - | SCAN TIME expr { KRT_PROTO->scan_time = $3; } - | LEARN bool { KRT_PROTO->learn = $2; } - | ROUTE SCAN TIME expr { KRT_PROTO->route_scan_time = $4; } - ; - -CF_CODE - -CF_END diff --git a/sysdep/linux/netlink/krt.c b/sysdep/linux/netlink/krt.c deleted file mode 100644 index c399e20..0000000 --- a/sysdep/linux/netlink/krt.c +++ /dev/null @@ -1,447 +0,0 @@ -/* - * BIRD -- Linux Netlink Interface - * - * (c) 1999 Martin Mares - * - * Can be freely distributed and used under the terms of the GNU GPL. - */ - -#include -#include -#include -#include -#include -#include - -#define LOCAL_DEBUG - -#undef ASYNC_NETLINK /* Define if async notifications should be used (debug) */ - -#include "nest/bird.h" -#include "nest/route.h" -#include "nest/protocol.h" -#include "nest/iface.h" -#include "lib/timer.h" -#include "lib/unix.h" -#include "lib/krt.h" -#include "lib/socket.h" - -/* - * We need to work around namespace conflicts between us and the kernel, - * but I prefer this way to being forced to rename our configuration symbols. - * This will disappear as soon as netlink headers become part of the libc. - */ - -#undef CONFIG_NETLINK -#include -#ifndef CONFIG_NETLINK -#error "Kernel not configured to support netlink" -#endif - -#include -#include -#include - -#ifndef MSG_TRUNC /* FIXME: Hack to circumvent omissions in glibc includes */ -#define MSG_TRUNC 0x20 -#endif - -struct proto_config *cf_krt; - -/* - * Synchronous Netlink interface - */ - -static int nl_sync_fd; /* Unix socket for synchronous netlink actions */ -static u32 nl_sync_seq; /* Sequence number of last request sent */ - -static byte *nl_rx_buffer; /* Receive buffer */ -static int nl_rx_size = 8192; - -static struct nlmsghdr *nl_last_hdr; /* Recently received packet */ -static unsigned int nl_last_size; - -static void -nl_send(void *rq, int size) -{ - struct nlmsghdr *nh = rq; - struct sockaddr_nl sa; - - memset(&sa, 0, sizeof(sa)); - sa.nl_family = AF_NETLINK; - nh->nlmsg_len = size; - nh->nlmsg_pid = 0; - nh->nlmsg_seq = ++nl_sync_seq; - if (sendto(nl_sync_fd, rq, size, 0, (struct sockaddr *)&sa, sizeof(sa)) < 0) - die("rtnetlink sendto: %m"); - nl_last_hdr = NULL; -} - -static void -nl_request_dump(int cmd) -{ - struct { - struct nlmsghdr nh; - struct rtgenmsg g; - } req; - req.nh.nlmsg_type = cmd; - req.nh.nlmsg_flags = NLM_F_REQUEST | NLM_F_DUMP; - req.g.rtgen_family = PF_INET; - nl_send(&req, sizeof(req)); -} - -static struct nlmsghdr * -nl_get_reply(void) -{ - for(;;) - { - if (!nl_last_hdr) - { - struct iovec iov = { nl_rx_buffer, nl_rx_size }; - struct sockaddr_nl sa; - struct msghdr m = { (struct sockaddr *) &sa, sizeof(sa), &iov, 1, NULL, 0, 0 }; - int x = recvmsg(nl_sync_fd, &m, 0); - if (x < 0) - die("nl_get_reply: %m"); - if (sa.nl_pid) /* It isn't from the kernel */ - { - DBG("Non-kernel packet\n"); - continue; - } - nl_last_size = x; - nl_last_hdr = (void *) nl_rx_buffer; - if (m.msg_flags & MSG_TRUNC) - bug("nl_get_reply: got truncated reply which should be impossible"); - } - if (NLMSG_OK(nl_last_hdr, nl_last_size)) - { - struct nlmsghdr *h = nl_last_hdr; - if (h->nlmsg_seq != nl_sync_seq) - { - log(L_WARN "nl_get_reply: Ignoring out of sequence netlink packet (%x != %x)", - h->nlmsg_seq, nl_sync_seq); - continue; - } - nl_last_hdr = NLMSG_NEXT(h, nl_last_size); - return h; - } - if (nl_last_size) - log(L_WARN "nl_get_reply: Found packet remnant of size %d", nl_last_size); - nl_last_hdr = NULL; - } -} - -static char * -nl_error(struct nlmsghdr *h) -{ - struct nlmsgerr *e = NLMSG_DATA(h); - if (h->nlmsg_len < NLMSG_LENGTH(sizeof(struct nlmsgerr))) - return "Error message truncated"; - else - return strerror(-e->error); -} - -static struct nlmsghdr * -nl_get_scan(void) -{ - struct nlmsghdr *h = nl_get_reply(); - - if (h->nlmsg_type == NLMSG_DONE) - return NULL; - if (h->nlmsg_type == NLMSG_ERROR) - { - log(L_ERR "Netlink error: %s", nl_error(h)); - return NULL; - } - return h; -} - -/* - * Parsing of Netlink attributes - */ - -static int nl_attr_len; - -static void * -nl_checkin(struct nlmsghdr *h, int lsize) -{ - nl_attr_len = h->nlmsg_len - NLMSG_LENGTH(lsize); - if (nl_attr_len < 0) - { - log(L_ERR "nl_checkin: underrun by %d bytes", -nl_attr_len); - return NULL; - } - return NLMSG_DATA(h); -} - -static int -nl_parse_attrs(struct rtattr *a, struct rtattr **k, int ksize) -{ - int max = ksize / sizeof(struct rtattr *); - bzero(k, ksize); - while (RTA_OK(a, nl_attr_len)) - { - if (a->rta_type < max) - k[a->rta_type] = a; - a = RTA_NEXT(a, nl_attr_len); - } - if (nl_attr_len) - { - log(L_ERR "nl_parse_attrs: remnant of size %d", nl_attr_len); - return 0; - } - else - return 1; -} - -/* - * Scanning of interfaces - */ - -static void -nl_parse_link(struct nlmsghdr *h, int scan) -{ - struct ifinfomsg *i; - struct rtattr *a[IFLA_STATS+1]; - int new = h->nlmsg_type == RTM_NEWLINK; - struct iface f; - struct iface *ifi; - char *name; - u32 mtu; - unsigned int fl; - - if (!(i = nl_checkin(h, sizeof(*i))) || !nl_parse_attrs(IFLA_RTA(i), a, sizeof(a))) - return; - if (!a[IFLA_IFNAME] || RTA_PAYLOAD(a[IFLA_IFNAME]) < 2 || - !a[IFLA_MTU] || RTA_PAYLOAD(a[IFLA_MTU]) != 4) - { - log(L_ERR "nl_parse_link: Malformed message received"); - return; - } - name = RTA_DATA(a[IFLA_IFNAME]); - memcpy(&mtu, RTA_DATA(a[IFLA_MTU]), sizeof(u32)); - - ifi = if_find_by_index(i->ifi_index); - if (!new) - { - DBG("KRT: IF%d(%s) goes down\n", i->ifi_index, name); - if (ifi && !scan) - { - memcpy(&f, ifi, sizeof(struct iface)); - f.flags |= IF_ADMIN_DOWN; - if_update(&f); - } - } - else - { - DBG("KRT: IF%d(%s) goes up (mtu=%d,flg=%x)\n", i->ifi_index, name, mtu, i->ifi_flags); - if (ifi) - memcpy(&f, ifi, sizeof(f)); - else - { - bzero(&f, sizeof(f)); - f.index = i->ifi_index; - } - strncpy(f.name, RTA_DATA(a[IFLA_IFNAME]), sizeof(f.name)-1); - f.mtu = mtu; - f.flags = 0; - fl = i->ifi_flags; - if (fl & IFF_UP) - f.flags |= IF_LINK_UP; - if (fl & IFF_POINTOPOINT) - f.flags |= IF_UNNUMBERED | IF_MULTICAST; - if (fl & IFF_LOOPBACK) - f.flags |= IF_LOOPBACK | IF_IGNORE; - if (fl & IFF_BROADCAST) - f.flags |= IF_BROADCAST | IF_MULTICAST; - if_update(&f); - } -} - -static void -nl_parse_addr(struct nlmsghdr *h) -{ - struct ifaddrmsg *i; - struct rtattr *a[IFA_ANYCAST+1]; - int new = h->nlmsg_type == RTM_NEWADDR; - struct iface f; - struct iface *ifi; - - if (!(i = nl_checkin(h, sizeof(*i))) || !nl_parse_attrs(IFA_RTA(i), a, sizeof(a))) - return; - if (i->ifa_family != AF_INET) - return; - if (!a[IFA_ADDRESS] || RTA_PAYLOAD(a[IFA_ADDRESS]) != 4 || - !a[IFA_LOCAL] || RTA_PAYLOAD(a[IFA_LOCAL]) != 4 || - (a[IFA_BROADCAST] && RTA_PAYLOAD(a[IFA_BROADCAST]) != 4)) - { - log(L_ERR "nl_parse_addr: Malformed message received"); - return; - } - if (i->ifa_flags & IFA_F_SECONDARY) - { - DBG("KRT: Received address message for secondary address which is not supported.\n"); /* FIXME */ - return; - } - - ifi = if_find_by_index(i->ifa_index); - if (!ifi) - { - log(L_ERR "KRT: Received address message for unknown interface %d\n", i->ifa_index); - return; - } - memcpy(&f, ifi, sizeof(f)); - - if (i->ifa_prefixlen > 32 || i->ifa_prefixlen == 31 || - (f.flags & IF_UNNUMBERED) && i->ifa_prefixlen != 32) - { - log(L_ERR "KRT: Invalid prefix length for interface %s: %d\n", f.name, i->ifa_prefixlen); - new = 0; - } - - f.ip = f.brd = f.opposite = IPA_NONE; - if (!new) - { - DBG("KRT: IF%d IP address deleted\n"); - f.pxlen = 0; - } - else - { - memcpy(&f.ip, RTA_DATA(a[IFA_LOCAL]), sizeof(f.ip)); - f.ip = ipa_ntoh(f.ip); - f.pxlen = i->ifa_prefixlen; - if (f.flags & IF_UNNUMBERED) - { - memcpy(&f.opposite, RTA_DATA(a[IFA_ADDRESS]), sizeof(f.opposite)); - f.opposite = f.brd = ipa_ntoh(f.opposite); - } - else if ((f.flags & IF_BROADCAST) && a[IFA_BROADCAST]) - { - memcpy(&f.brd, RTA_DATA(a[IFA_BROADCAST]), sizeof(f.brd)); - f.brd = ipa_ntoh(f.brd); - } - /* else a NBMA link */ - f.prefix = ipa_and(f.ip, ipa_mkmask(f.pxlen)); - DBG("KRT: IF%d IP address set to %I, net %I/%d, brd %I, opp %I\n", f.index, f.ip, f.prefix, f.pxlen, f.brd, f.opposite); - } - if_update(&f); -} - -static void -nl_scan_ifaces(void) -{ - struct nlmsghdr *h; - - if_start_update(); - - nl_request_dump(RTM_GETLINK); - while (h = nl_get_scan()) - if (h->nlmsg_type == RTM_NEWLINK || h->nlmsg_type == RTM_DELLINK) - nl_parse_link(h, 1); - else - log(L_DEBUG "nl_scan_ifaces: Unknown packet received (type=%d)", h->nlmsg_type); - - nl_request_dump(RTM_GETADDR); - while (h = nl_get_scan()) - if (h->nlmsg_type == RTM_NEWADDR || h->nlmsg_type == RTM_DELADDR) - nl_parse_addr(h); - else - log(L_DEBUG "nl_scan_ifaces: Unknown packet received (type=%d)", h->nlmsg_type); - - if_end_update(); -} - -/* - * Asynchronous Netlink interface - */ - -static sock *nl_async_sk; /* BIRD socket for asynchronous notifications */ - -static int -nl_async_hook(sock *sk, int size) -{ - DBG("nl_async_hook\n"); - return 0; -} - -/* - * Protocol core - */ - -static void -krt_preconfig(struct protocol *x, struct config *c) -{ - struct krt_config *z = proto_config_new(&proto_unix_kernel, sizeof(struct krt_config)); - - cf_krt = &z->c; - z->c.preference = DEF_PREF_UKR; -} - -static struct proto * -krt_init(struct proto_config *c) -{ - struct krt_proto *p = proto_new(c, sizeof(struct krt_proto)); - - return &p->p; -} - -static void -nl_open_async(struct proto *p) -{ - sock *sk; - struct sockaddr_nl sa; - - sk = nl_async_sk = sk_new(p->pool); - sk->type = SK_MAGIC; - sk->rx_hook = nl_async_hook; - sk->fd = socket(PF_NETLINK, SOCK_RAW, NETLINK_ROUTE); - if (sk->fd < 0 || sk_open(sk)) - die("Unable to open secondary rtnetlink socket: %m"); - - bzero(&sa, sizeof(sa)); - sa.nl_family = AF_NETLINK; - sa.nl_groups = RTMGRP_LINK | RTMGRP_IPV4_IFADDR | RTMGRP_IPV4_ROUTE; - if (bind(sk->fd, (struct sockaddr *) &sa, sizeof(sa)) < 0) - die("Unable to bind secondary rtnetlink socket: %m"); -} - -static int -krt_start(struct proto *p) -{ -#ifdef ASYNC_NETLINK - nl_open_async(p); -#endif - - /* FIXME: Filter kernel routing table etc. */ - - return PS_UP; -} - -static int -krt_shutdown(struct proto *p) -{ - /* FIXME: Remove all our routes from the kernel */ - - return PS_DOWN; -} - -void -scan_if_init(void) -{ - nl_sync_fd = socket(PF_NETLINK, SOCK_RAW, NETLINK_ROUTE); - if (nl_sync_fd < 0) - die("Unable to open rtnetlink socket: %m"); - nl_sync_seq = now; - nl_rx_buffer = xmalloc(nl_rx_size); - /* FIXME: Should we fetch our local address and compare it with addresses of all incoming messages? */ - - nl_scan_ifaces(); -} - -struct protocol proto_unix_kernel = { - name: "Kernel", - preconfig: krt_preconfig, - init: krt_init, - start: krt_start, - shutdown: krt_shutdown -}; diff --git a/sysdep/linux/netlink/krt.h b/sysdep/linux/netlink/krt.h deleted file mode 100644 index 8dd55cc..0000000 --- a/sysdep/linux/netlink/krt.h +++ /dev/null @@ -1,30 +0,0 @@ -/* - * BIRD -- Linux Netlink Interface - * - * (c) 1999 Martin Mares - * - * Can be freely distributed and used under the terms of the GNU GPL. - */ - -#ifndef _BIRD_NETLINK_KRT_H_ -#define _BIRD_NETLINK_KRT_H_ - -extern struct protocol proto_unix_kernel; - -struct krt_config { - struct proto_config c; - int persist; /* Keep routes when we exit */ - int scan_time; /* How often we re-scan interfaces */ - int route_scan_time; /* How often we re-scan routes */ - int learn; /* Learn routes from other sources */ -}; - -extern struct proto_config *cf_krt; - -struct krt_proto { - struct proto p; -}; - -void scan_if_init(void); - -#endif diff --git a/sysdep/linux/netlink/netlink.Y b/sysdep/linux/netlink/netlink.Y new file mode 100644 index 0000000..b5c45f2 --- /dev/null +++ b/sysdep/linux/netlink/netlink.Y @@ -0,0 +1,25 @@ +/* + * BIRD -- Linux Netlink Configuration + * + * (c) 1999 Martin Mares + * + * Can be freely distributed and used under the terms of the GNU GPL. + */ + +CF_HDR + +CF_DECLS + +CF_KEYWORDS(ASYNC) + +CF_GRAMMAR + +CF_ADDTO(kern_proto, kern_proto nl_item ';') + +nl_item: + ASYNC bool { THIS_KRT->scan.async = $2; } + ; + +CF_CODE + +CF_END diff --git a/sysdep/linux/netlink/netlink.c b/sysdep/linux/netlink/netlink.c new file mode 100644 index 0000000..6e6fafd --- /dev/null +++ b/sysdep/linux/netlink/netlink.c @@ -0,0 +1,440 @@ +/* + * BIRD -- Linux Netlink Interface + * + * (c) 1999 Martin Mares + * + * Can be freely distributed and used under the terms of the GNU GPL. + */ + +#include +#include +#include +#include +#include +#include + +#define LOCAL_DEBUG + +#include "nest/bird.h" +#include "nest/route.h" +#include "nest/protocol.h" +#include "nest/iface.h" +#include "lib/timer.h" +#include "lib/unix.h" +#include "lib/krt.h" +#include "lib/socket.h" + +/* + * We need to work around namespace conflicts between us and the kernel, + * but I prefer this way to being forced to rename our configuration symbols. + * This will disappear as soon as netlink headers become part of the libc. + */ + +#undef CONFIG_NETLINK +#include +#ifndef CONFIG_NETLINK +#error "Kernel not configured to support netlink" +#endif + +#include +#include +#include + +#ifndef MSG_TRUNC /* FIXME: Hack to circumvent omissions in glibc includes */ +#define MSG_TRUNC 0x20 +#endif + +/* + * Synchronous Netlink interface + */ + +static int nl_sync_fd = -1; /* Unix socket for synchronous netlink actions */ +static u32 nl_sync_seq; /* Sequence number of last request sent */ + +static byte *nl_rx_buffer; /* Receive buffer */ +static int nl_rx_size = 8192; + +static struct nlmsghdr *nl_last_hdr; /* Recently received packet */ +static unsigned int nl_last_size; + +static void +nl_open(void) +{ + if (nl_sync_fd < 0) + { + nl_sync_fd = socket(PF_NETLINK, SOCK_RAW, NETLINK_ROUTE); + if (nl_sync_fd < 0) + die("Unable to open rtnetlink socket: %m"); + nl_sync_seq = now; + nl_rx_buffer = xmalloc(nl_rx_size); + } +} + +static void +nl_send(void *rq, int size) +{ + struct nlmsghdr *nh = rq; + struct sockaddr_nl sa; + + memset(&sa, 0, sizeof(sa)); + sa.nl_family = AF_NETLINK; + nh->nlmsg_len = size; + nh->nlmsg_pid = 0; + nh->nlmsg_seq = ++nl_sync_seq; + if (sendto(nl_sync_fd, rq, size, 0, (struct sockaddr *)&sa, sizeof(sa)) < 0) + die("rtnetlink sendto: %m"); + nl_last_hdr = NULL; +} + +static void +nl_request_dump(int cmd) +{ + struct { + struct nlmsghdr nh; + struct rtgenmsg g; + } req; + req.nh.nlmsg_type = cmd; + req.nh.nlmsg_flags = NLM_F_REQUEST | NLM_F_DUMP; + req.g.rtgen_family = PF_INET; + nl_send(&req, sizeof(req)); +} + +static struct nlmsghdr * +nl_get_reply(void) +{ + for(;;) + { + if (!nl_last_hdr) + { + struct iovec iov = { nl_rx_buffer, nl_rx_size }; + struct sockaddr_nl sa; + struct msghdr m = { (struct sockaddr *) &sa, sizeof(sa), &iov, 1, NULL, 0, 0 }; + int x = recvmsg(nl_sync_fd, &m, 0); + if (x < 0) + die("nl_get_reply: %m"); + if (sa.nl_pid) /* It isn't from the kernel */ + { + DBG("Non-kernel packet\n"); + continue; + } + nl_last_size = x; + nl_last_hdr = (void *) nl_rx_buffer; + if (m.msg_flags & MSG_TRUNC) + bug("nl_get_reply: got truncated reply which should be impossible"); + } + if (NLMSG_OK(nl_last_hdr, nl_last_size)) + { + struct nlmsghdr *h = nl_last_hdr; + if (h->nlmsg_seq != nl_sync_seq) + { + log(L_WARN "nl_get_reply: Ignoring out of sequence netlink packet (%x != %x)", + h->nlmsg_seq, nl_sync_seq); + continue; + } + nl_last_hdr = NLMSG_NEXT(h, nl_last_size); + return h; + } + if (nl_last_size) + log(L_WARN "nl_get_reply: Found packet remnant of size %d", nl_last_size); + nl_last_hdr = NULL; + } +} + +static char * +nl_error(struct nlmsghdr *h) +{ + struct nlmsgerr *e = NLMSG_DATA(h); + if (h->nlmsg_len < NLMSG_LENGTH(sizeof(struct nlmsgerr))) + return "Error message truncated"; + else + return strerror(-e->error); +} + +static struct nlmsghdr * +nl_get_scan(void) +{ + struct nlmsghdr *h = nl_get_reply(); + + if (h->nlmsg_type == NLMSG_DONE) + return NULL; + if (h->nlmsg_type == NLMSG_ERROR) + { + log(L_ERR "Netlink error: %s", nl_error(h)); + return NULL; + } + return h; +} + +/* + * Parsing of Netlink attributes + */ + +static int nl_attr_len; + +static void * +nl_checkin(struct nlmsghdr *h, int lsize) +{ + nl_attr_len = h->nlmsg_len - NLMSG_LENGTH(lsize); + if (nl_attr_len < 0) + { + log(L_ERR "nl_checkin: underrun by %d bytes", -nl_attr_len); + return NULL; + } + return NLMSG_DATA(h); +} + +static int +nl_parse_attrs(struct rtattr *a, struct rtattr **k, int ksize) +{ + int max = ksize / sizeof(struct rtattr *); + bzero(k, ksize); + while (RTA_OK(a, nl_attr_len)) + { + if (a->rta_type < max) + k[a->rta_type] = a; + a = RTA_NEXT(a, nl_attr_len); + } + if (nl_attr_len) + { + log(L_ERR "nl_parse_attrs: remnant of size %d", nl_attr_len); + return 0; + } + else + return 1; +} + +/* + * Scanning of interfaces + */ + +static void +nl_parse_link(struct nlmsghdr *h, int scan) +{ + struct ifinfomsg *i; + struct rtattr *a[IFLA_STATS+1]; + int new = h->nlmsg_type == RTM_NEWLINK; + struct iface f; + struct iface *ifi; + char *name; + u32 mtu; + unsigned int fl; + + if (!(i = nl_checkin(h, sizeof(*i))) || !nl_parse_attrs(IFLA_RTA(i), a, sizeof(a))) + return; + if (!a[IFLA_IFNAME] || RTA_PAYLOAD(a[IFLA_IFNAME]) < 2 || + !a[IFLA_MTU] || RTA_PAYLOAD(a[IFLA_MTU]) != 4) + { + log(L_ERR "nl_parse_link: Malformed message received"); + return; + } + name = RTA_DATA(a[IFLA_IFNAME]); + memcpy(&mtu, RTA_DATA(a[IFLA_MTU]), sizeof(u32)); + + ifi = if_find_by_index(i->ifi_index); + if (!new) + { + DBG("KRT: IF%d(%s) goes down\n", i->ifi_index, name); + if (ifi && !scan) + { + memcpy(&f, ifi, sizeof(struct iface)); + f.flags |= IF_ADMIN_DOWN; + if_update(&f); + } + } + else + { + DBG("KRT: IF%d(%s) goes up (mtu=%d,flg=%x)\n", i->ifi_index, name, mtu, i->ifi_flags); + if (ifi) + memcpy(&f, ifi, sizeof(f)); + else + { + bzero(&f, sizeof(f)); + f.index = i->ifi_index; + } + strncpy(f.name, RTA_DATA(a[IFLA_IFNAME]), sizeof(f.name)-1); + f.mtu = mtu; + f.flags = 0; + fl = i->ifi_flags; + if (fl & IFF_UP) + f.flags |= IF_LINK_UP; + if (fl & IFF_POINTOPOINT) + f.flags |= IF_UNNUMBERED | IF_MULTICAST; + if (fl & IFF_LOOPBACK) + f.flags |= IF_LOOPBACK | IF_IGNORE; + if (fl & IFF_BROADCAST) + f.flags |= IF_BROADCAST | IF_MULTICAST; + if_update(&f); + } +} + +static void +nl_parse_addr(struct nlmsghdr *h) +{ + struct ifaddrmsg *i; + struct rtattr *a[IFA_ANYCAST+1]; + int new = h->nlmsg_type == RTM_NEWADDR; + struct iface f; + struct iface *ifi; + + if (!(i = nl_checkin(h, sizeof(*i))) || !nl_parse_attrs(IFA_RTA(i), a, sizeof(a))) + return; + if (i->ifa_family != AF_INET) + return; + if (!a[IFA_ADDRESS] || RTA_PAYLOAD(a[IFA_ADDRESS]) != 4 || + !a[IFA_LOCAL] || RTA_PAYLOAD(a[IFA_LOCAL]) != 4 || + (a[IFA_BROADCAST] && RTA_PAYLOAD(a[IFA_BROADCAST]) != 4)) + { + log(L_ERR "nl_parse_addr: Malformed message received"); + return; + } + if (i->ifa_flags & IFA_F_SECONDARY) + { + DBG("KRT: Received address message for secondary address which is not supported.\n"); /* FIXME */ + return; + } + + ifi = if_find_by_index(i->ifa_index); + if (!ifi) + { + log(L_ERR "KRT: Received address message for unknown interface %d\n", i->ifa_index); + return; + } + memcpy(&f, ifi, sizeof(f)); + + if (i->ifa_prefixlen > 32 || i->ifa_prefixlen == 31 || + (f.flags & IF_UNNUMBERED) && i->ifa_prefixlen != 32) + { + log(L_ERR "KRT: Invalid prefix length for interface %s: %d\n", f.name, i->ifa_prefixlen); + new = 0; + } + + f.ip = f.brd = f.opposite = IPA_NONE; + if (!new) + { + DBG("KRT: IF%d IP address deleted\n"); + f.pxlen = 0; + } + else + { + memcpy(&f.ip, RTA_DATA(a[IFA_LOCAL]), sizeof(f.ip)); + f.ip = ipa_ntoh(f.ip); + f.pxlen = i->ifa_prefixlen; + if (f.flags & IF_UNNUMBERED) + { + memcpy(&f.opposite, RTA_DATA(a[IFA_ADDRESS]), sizeof(f.opposite)); + f.opposite = f.brd = ipa_ntoh(f.opposite); + } + else if ((f.flags & IF_BROADCAST) && a[IFA_BROADCAST]) + { + memcpy(&f.brd, RTA_DATA(a[IFA_BROADCAST]), sizeof(f.brd)); + f.brd = ipa_ntoh(f.brd); + } + /* else a NBMA link */ + f.prefix = ipa_and(f.ip, ipa_mkmask(f.pxlen)); + DBG("KRT: IF%d IP address set to %I, net %I/%d, brd %I, opp %I\n", f.index, f.ip, f.prefix, f.pxlen, f.brd, f.opposite); + } + if_update(&f); +} + +void +krt_if_scan(struct krt_proto *p) +{ + struct nlmsghdr *h; + + if_start_update(); + + nl_request_dump(RTM_GETLINK); + while (h = nl_get_scan()) + if (h->nlmsg_type == RTM_NEWLINK || h->nlmsg_type == RTM_DELLINK) + nl_parse_link(h, 1); + else + log(L_DEBUG "nl_scan_ifaces: Unknown packet received (type=%d)", h->nlmsg_type); + + nl_request_dump(RTM_GETADDR); + while (h = nl_get_scan()) + if (h->nlmsg_type == RTM_NEWADDR || h->nlmsg_type == RTM_DELADDR) + nl_parse_addr(h); + else + log(L_DEBUG "nl_scan_ifaces: Unknown packet received (type=%d)", h->nlmsg_type); + + if_end_update(); +} + +/* + * Routes + */ + +int +krt_capable(rte *e) +{ + return 1; /* FIXME */ +} + +void +krt_set_notify(struct proto *p, net *n, rte *new, rte *old) +{ + /* FIXME */ +} + +void +krt_scan_fire(struct krt_proto *p) +{ +} + +/* + * Asynchronous Netlink interface + */ + +static sock *nl_async_sk; /* BIRD socket for asynchronous notifications */ + +static int +nl_async_hook(sock *sk, int size) +{ + DBG("nl_async_hook\n"); + return 0; +} + +static void +nl_open_async(struct krt_proto *p) +{ + sock *sk; + struct sockaddr_nl sa; + + DBG("KRT: Opening async netlink socket\n"); + + sk = nl_async_sk = sk_new(p->p.pool); + sk->type = SK_MAGIC; + sk->rx_hook = nl_async_hook; + sk->fd = socket(PF_NETLINK, SOCK_RAW, NETLINK_ROUTE); + if (sk->fd < 0 || sk_open(sk)) + die("Unable to open secondary rtnetlink socket: %m"); + + bzero(&sa, sizeof(sa)); + sa.nl_family = AF_NETLINK; + sa.nl_groups = RTMGRP_LINK | RTMGRP_IPV4_IFADDR | RTMGRP_IPV4_ROUTE; + if (bind(sk->fd, (struct sockaddr *) &sa, sizeof(sa)) < 0) + die("Unable to bind secondary rtnetlink socket: %m"); +} + +/* + * Interface to the UNIX krt module + */ + +void +krt_scan_preconfig(struct krt_config *x) +{ + x->scan.async = 1; +} + +void +krt_scan_start(struct krt_proto *p) +{ + nl_open(); + if (KRT_CF->scan.async) + nl_open_async(p); +} + +void +krt_scan_shutdown(struct krt_proto *p) +{ +} -- cgit v1.2.3