/* User-kernel network link */
-/* We support a variety of methods for extracting packets from the
- kernel: userv-ipif, ipif on its own (when we run as root), the
- kernel TUN driver. Possible future methods: SLIP to a pty, an
- external netlink daemon. There is a performance/security
- tradeoff. */
+/* See RFCs 791, 792, 1123 and 1812 */
-/* When dealing with SLIP (to a pty, or ipif) we have separate rx, tx
- and client buffers. When receiving we may read() any amount, not
- just whole packets. When transmitting we need to bytestuff anyway,
- and may be part-way through receiving. */
+/* The netlink device is actually a router. Tunnels are unnumbered
+ point-to-point lines (RFC1812 section 2.2.7); the router has a
+ single address (the 'router-id'). */
-/* Each netlink device is actually a router, with its own IP address.
- We do things like decreasing the TTL and recalculating the header
- checksum, generating ICMP, responding to pings, etc. */
+/* This is where we currently have the anti-spoofing paranoia - before
+ sending a packet to the kernel we check that the tunnel it came
+ over could reasonably have produced it. */
-/* This is where we have the anti-spoofing paranoia - before sending a
- packet to the kernel we check that the tunnel it came over could
- reasonably have produced it. */
-#include "secnet.h"
-#include <stdio.h>
-#include <string.h>
-#include <unistd.h>
-#include <fcntl.h>
-#include <sys/ioctl.h>
-#include "util.h"
+/* Points to note from RFC1812 (which may require changes in this
+ file):
-#ifdef HAVE_LINUX_IF_H
-#include <linux/if.h>
-#include <linux/if_tun.h>
-#endif
+3.3.4 Maximum Transmission Unit - MTU
-/* XXX where do we find if_tun on other platforms? */
+ The MTU of each logical interface MUST be configurable within the
+ range of legal MTUs for the interface.
-#define DEFAULT_BUFSIZE 2048
-#define DEFAULT_MTU 1000
-#define ICMP_BUFSIZE 1024
+ Many Link Layer protocols define a maximum frame size that may be
+ sent. In such cases, a router MUST NOT allow an MTU to be set which
+ would allow sending of frames larger than those allowed by the Link
+ Layer protocol. However, a router SHOULD be willing to receive a
+ packet as large as the maximum frame size even if that is larger than
+ the MTU.
-#define SLIP_END 192
-#define SLIP_ESC 219
-#define SLIP_ESCEND 220
-#define SLIP_ESCESC 221
+4.2.1 A router SHOULD count datagrams discarded.
-struct netlink_client {
- struct subnet_list *networks;
- netlink_deliver_fn *deliver;
- void *dst;
- string_t name;
- uint32_t link_quality;
- struct netlink_client *next;
-};
+4.2.2.1 Source route options - we probably should implement processing
+of source routes, even though mostly the security policy will prevent
+their use.
-struct netlink_route {
- struct subnet net;
- struct netlink_client *c;
-};
+5.3.13.4 Source Route Options
-/* Netlink provides one function to the device driver, to call to deliver
- a packet from the device. The device driver provides one function to
- netlink, for it to call to deliver a packet to the device. */
+ A router MUST implement support for source route options in forwarded
+ packets. A router MAY implement a configuration option that, when
+ enabled, causes all source-routed packets to be discarded. However,
+ such an option MUST NOT be enabled by default.
-struct netlink {
- closure_t cl;
- struct netlink_if ops;
- void *dst; /* Pointer to host interface state */
- string_t name;
- uint32_t max_start_pad;
- uint32_t max_end_pad;
- struct subnet_list networks;
- struct subnet_list exclude_remote_networks;
- uint32_t local_address; /* host interface address */
- uint32_t secnet_address; /* our own address */
- uint32_t mtu;
- struct netlink_client *clients;
- netlink_deliver_fn *deliver_to_host; /* Provided by driver */
- struct buffer_if icmp; /* Buffer for assembly of outgoing ICMP */
- uint32_t n_routes; /* How many routes do we know about? */
- struct netlink_route *routes;
-};
+5.3.13.5 Record Route Option
+
+ Routers MUST support the Record Route option in forwarded packets.
+
+ A router MAY provide a configuration option that, if enabled, will
+ cause the router to ignore (i.e., pass through unchanged) Record
+ Route options in forwarded packets. If provided, such an option MUST
+ default to enabling the record-route. This option should not affect
+ the processing of Record Route options in datagrams received by the
+ router itself (in particular, Record Route options in ICMP echo
+ requests will still be processed according to Section [4.3.3.6]).
+
+5.3.13.6 Timestamp Option
+
+ Routers MUST support the timestamp option in forwarded packets. A
+ timestamp value MUST follow the rules given [INTRO:2].
+
+ If the flags field = 3 (timestamp and prespecified address), the
+ router MUST add its timestamp if the next prespecified address
+ matches any of the router's IP addresses. It is not necessary that
+ the prespecified address be either the address of the interface on
+ which the packet arrived or the address of the interface over which
+ it will be sent.
+
+
+4.2.2.7 Fragmentation: RFC 791 Section 3.2
+
+ Fragmentation, as described in [INTERNET:1], MUST be supported by a
+ router.
+
+4.2.2.8 Reassembly: RFC 791 Section 3.2
+
+ As specified in the corresponding section of [INTRO:2], a router MUST
+ support reassembly of datagrams that it delivers to itself.
+
+4.2.2.9 Time to Live: RFC 791 Section 3.2
+
+ Note in particular that a router MUST NOT check the TTL of a packet
+ except when forwarding it.
+
+ A router MUST NOT discard a datagram just because it was received
+ with TTL equal to zero or one; if it is to the router and otherwise
+ valid, the router MUST attempt to receive it.
+
+ On messages the router originates, the IP layer MUST provide a means
+ for the transport layer to set the TTL field of every datagram that
+ is sent. When a fixed TTL value is used, it MUST be configurable.
+
+
+8.1 The Simple Network Management Protocol - SNMP
+8.1.1 SNMP Protocol Elements
+
+ Routers MUST be manageable by SNMP [MGT:3]. The SNMP MUST operate
+ using UDP/IP as its transport and network protocols.
+
+
+*/
+
+#include <string.h>
+#include <assert.h>
+#include <limits.h>
+#include "secnet.h"
+#include "util.h"
+#include "ipaddr.h"
+#include "netlink.h"
+#include "process.h"
+
+#define ICMP_TYPE_ECHO_REPLY 0
+
+#define ICMP_TYPE_UNREACHABLE 3
+#define ICMP_CODE_NET_UNREACHABLE 0
+#define ICMP_CODE_PROTOCOL_UNREACHABLE 2
+#define ICMP_CODE_FRAGMENTATION_REQUIRED 4
+#define ICMP_CODE_NET_PROHIBITED 13
+
+#define ICMP_TYPE_ECHO_REQUEST 8
+
+#define ICMP_TYPE_TIME_EXCEEDED 11
+#define ICMP_CODE_TTL_EXCEEDED 0
/* Generic IP checksum routine */
-static inline uint16_t ip_csum(uint8_t *iph,uint32_t count)
+static inline uint16_t ip_csum(uint8_t *iph,int32_t count)
{
register uint32_t sum=0;
* By Jorge Cwik <jorge@laser.satlink.net>, adapted for linux by
* Arnt Gulbrandsen.
*/
-static inline uint16_t ip_fast_csum(uint8_t *iph, uint32_t ihl) {
+static inline uint16_t ip_fast_csum(uint8_t *iph, int32_t ihl) {
uint32_t sum;
- __asm__ __volatile__("
- movl (%1), %0
- subl $4, %2
- jbe 2f
- addl 4(%1), %0
- adcl 8(%1), %0
- adcl 12(%1), %0
-1: adcl 16(%1), %0
- lea 4(%1), %1
- decl %2
- jne 1b
- adcl $0, %0
- movl %0, %2
- shrl $16, %0
- addw %w2, %w0
- adcl $0, %0
- notl %0
-2:
- "
+ __asm__ __volatile__(
+ "movl (%1), %0 ;\n"
+ "subl $4, %2 ;\n"
+ "jbe 2f ;\n"
+ "addl 4(%1), %0 ;\n"
+ "adcl 8(%1), %0 ;\n"
+ "adcl 12(%1), %0 ;\n"
+"1: adcl 16(%1), %0 ;\n"
+ "lea 4(%1), %1 ;\n"
+ "decl %2 ;\n"
+ "jne 1b ;\n"
+ "adcl $0, %0 ;\n"
+ "movl %0, %2 ;\n"
+ "shrl $16, %0 ;\n"
+ "addw %w2, %w0 ;\n"
+ "adcl $0, %0 ;\n"
+ "notl %0 ;\n"
+"2: ;\n"
/* Since the input registers which are loaded with iph and ipl
are modified, we must also specify them as outputs, or gcc
will assume they contain their original values. */
: "=r" (sum), "=r" (iph), "=r" (ihl)
- : "1" (iph), "2" (ihl));
+ : "1" (iph), "2" (ihl)
+ : "memory");
return sum;
}
#else
-static inline uint16_t ip_fast_csum(uint8_t *iph, uint32_t ihl)
+static inline uint16_t ip_fast_csum(uint8_t *iph, int32_t ihl)
{
+ assert(ihl < INT_MAX/4);
return ip_csum(iph,ihl*4);
}
#endif
struct netlink_client *client,
struct buffer_if *buf);
+/* XXX RFC1812 4.3.2.5:
+ All other ICMP error messages (Destination Unreachable,
+ Redirect, Time Exceeded, and Parameter Problem) SHOULD have their
+ precedence value set to 6 (INTERNETWORK CONTROL) or 7 (NETWORK
+ CONTROL). The IP Precedence value for these error messages MAY be
+ settable.
+ */
static struct icmphdr *netlink_icmp_tmpl(struct netlink *st,
uint32_t dest,uint16_t len)
{
h->iph.tot_len=htons(len+(h->iph.ihl*4)+8);
h->iph.id=0;
h->iph.frag_off=0;
- h->iph.ttl=255;
+ h->iph.ttl=255; /* XXX should be configurable */
h->iph.protocol=1;
h->iph.saddr=htonl(st->secnet_address);
h->iph.daddr=htonl(dest);
/* Fill in the ICMP checksum field correctly */
static void netlink_icmp_csum(struct icmphdr *h)
{
- uint32_t len;
+ int32_t len;
len=ntohs(h->iph.tot_len)-(4*h->iph.ihl);
h->check=0;
static bool_t netlink_icmp_may_reply(struct buffer_if *buf)
{
struct iphdr *iph;
+ struct icmphdr *icmph;
uint32_t source;
iph=(struct iphdr *)buf->start;
- if (iph->protocol==1) return False; /* Overly-broad; we may reply to
- eg. icmp echo-request */
+ icmph=(struct icmphdr *)buf->start;
+ if (iph->protocol==1) {
+ switch(icmph->type) {
+ case 3: /* Destination unreachable */
+ case 11: /* Time Exceeded */
+ case 12: /* Parameter Problem */
+ return False;
+ }
+ }
/* How do we spot broadcast destination addresses? */
if (ntohs(iph->frag_off)&0x1fff) return False; /* Non-initial fragment */
source=ntohl(iph->saddr);
/* How much of the original IP packet do we include in its ICMP
response? The header plus up to 64 bits. */
+
+/* XXX TODO RFC1812:
+4.3.2.3 Original Message Header
+
+ Historically, every ICMP error message has included the Internet
+ header and at least the first 8 data bytes of the datagram that
+ triggered the error. This is no longer adequate, due to the use of
+ IP-in-IP tunneling and other technologies. Therefore, the ICMP
+ datagram SHOULD contain as much of the original datagram as possible
+ without the length of the ICMP datagram exceeding 576 bytes. The
+ returned IP header (and user data) MUST be identical to that which
+ was received, except that the router is not required to undo any
+ modifications to the IP header that are normally performed in
+ forwarding that were performed before the error was detected (e.g.,
+ decrementing the TTL, or updating options). Note that the
+ requirements of Section [4.3.3.5] supersede this requirement in some
+ cases (i.e., for a Parameter Problem message, if the problem is in a
+ modified field, the router must undo the modification). See Section
+ [4.3.3.5]).
+ */
static uint16_t netlink_icmp_reply_len(struct buffer_if *buf)
{
struct iphdr *iph=(struct iphdr *)buf->start;
/*
* RFC1122: 3.1.2.2 MUST silently discard any IP frame that fails the
* checksum.
+ * RFC1812: 4.2.2.5 MUST discard messages containing invalid checksums.
*
* Is the datagram acceptable?
*
* 3. Checksums correctly.
* 4. Doesn't have a bogus length
*/
-static bool_t netlink_check(struct netlink *st, struct buffer_if *buf)
+static bool_t netlink_check(struct netlink *st, struct buffer_if *buf,
+ char *errmsgbuf, int errmsgbuflen)
{
+#define BAD(...) do{ \
+ snprintf(errmsgbuf,errmsgbuflen,__VA_ARGS__); \
+ return False; \
+ }while(0)
+
struct iphdr *iph=(struct iphdr *)buf->start;
- uint32_t len;
+ int32_t len;
- if (iph->ihl < 5 || iph->version != 4) {
- printf("ihl/version check failed\n");
- return False;
- }
- if (buf->size < iph->ihl*4) {
- printf("buffer size check failed\n");
- return False;
- }
- if (ip_fast_csum((uint8_t *)iph, iph->ihl)!=0) {
- printf("checksum failed\n");
- return False;
- }
+ if (iph->ihl < 5) BAD("ihl %u",iph->ihl);
+ if (iph->version != 4) BAD("version %u",iph->version);
+ if (buf->size < iph->ihl*4) BAD("size %"PRId32"<%u*4",buf->size,iph->ihl);
+ if (ip_fast_csum((uint8_t *)iph, iph->ihl)!=0) BAD("csum");
len=ntohs(iph->tot_len);
/* There should be no padding */
- if (buf->size!=len || len<(iph->ihl<<2)) {
- printf("length check failed buf->size=%d len=%d\n",buf->size,len);
- return False;
- }
-
+ if (buf->size!=len) BAD("len %"PRId32"!=%"PRId32,buf->size,len);
+ if (len<(iph->ihl<<2)) BAD("len %"PRId32"<(%u<<2)",len,iph->ihl);
/* XXX check that there's no source route specified */
return True;
+
+#undef BAD
}
-/* Deliver a packet. "client" points to the _origin_ of the packet, not
- its destination. (May be used when sending ICMP response - avoid
- asymmetric routing.) */
+/* Deliver a packet. "client" is the _origin_ of the packet, not its
+ destination, and is NULL for packets from the host and packets
+ generated internally in secnet. */
static void netlink_packet_deliver(struct netlink *st,
struct netlink_client *client,
struct buffer_if *buf)
uint32_t dest=ntohl(iph->daddr);
uint32_t source=ntohl(iph->saddr);
uint32_t best_quality;
+ bool_t allow_route=False;
+ bool_t found_allowed=False;
int best_match;
int i;
BUF_ASSERT_USED(buf);
if (dest==st->secnet_address) {
- Message(M_ERROR,"%s: trying to deliver a packet to myself!\n");
+ Message(M_ERR,"%s: trying to deliver a packet to myself!\n",st->name);
BUF_FREE(buf);
return;
}
- if (!client) {
- /* Origin of packet is host or secnet. Might be for a tunnel. */
- best_quality=0;
- best_match=-1;
- for (i=0; i<st->n_routes; i++) {
- if (subnet_match(&st->routes[i].net,dest)) {
- if (st->routes[i].c->link_quality>best_quality
- || best_quality==0) {
- best_quality=st->routes[i].c->link_quality;
- best_match=i;
- /* If quality isn't perfect we may wish to
- consider kicking the tunnel with a 0-length
- packet to prompt it to perform a key setup.
- Then it'll eventually decide it's up or
- down. */
- /* If quality is perfect we don't need to search
- any more. */
- if (best_quality>=MAXIMUM_LINK_QUALITY) break;
- }
+ /* Packets from the host (client==NULL) may always be routed. Packets
+ from clients with the allow_route option will also be routed. */
+ if (!client || (client && (client->options & OPT_ALLOWROUTE)))
+ allow_route=True;
+
+ /* If !allow_route, we check the routing table anyway, and if
+ there's a suitable route with OPT_ALLOWROUTE set we use it. If
+ there's a suitable route, but none with OPT_ALLOWROUTE set then
+ we generate ICMP 'communication with destination network
+ administratively prohibited'. */
+
+ best_quality=0;
+ best_match=-1;
+ for (i=0; i<st->n_clients; i++) {
+ if (st->routes[i]->up &&
+ ipset_contains_addr(st->routes[i]->networks,dest)) {
+ /* It's an available route to the correct destination. But is
+ it better than the one we already have? */
+
+ /* If we have already found an allowed route then we don't
+ bother looking at routes we're not allowed to use. If
+ we don't yet have an allowed route we'll consider any. */
+ if (!allow_route && found_allowed) {
+ if (!(st->routes[i]->options&OPT_ALLOWROUTE)) continue;
}
- }
- if (best_match==-1) {
- /* Not going down a tunnel. Might be for the host.
- XXX think about this - only situation should be if we're
- sending ICMP. */
- if (source!=st->secnet_address) {
- Message(M_ERROR,"netlink_packet_deliver: outgoing packet "
- "from host that won't fit down any of our tunnels!\n");
- BUF_FREE(buf);
- } else {
- st->deliver_to_host(st->dst,NULL,buf);
- BUF_ASSERT_FREE(buf);
+
+ if (st->routes[i]->link_quality>best_quality
+ || best_quality==0) {
+ best_quality=st->routes[i]->link_quality;
+ best_match=i;
+ if (st->routes[i]->options&OPT_ALLOWROUTE)
+ found_allowed=True;
+ /* If quality isn't perfect we may wish to
+ consider kicking the tunnel with a 0-length
+ packet to prompt it to perform a key setup.
+ Then it'll eventually decide it's up or
+ down. */
+ /* If quality is perfect and we're allowed to use the
+ route we don't need to search any more. */
+ if (best_quality>=MAXIMUM_LINK_QUALITY &&
+ (allow_route || found_allowed)) break;
}
+ }
+ }
+ if (best_match==-1) {
+ /* The packet's not going down a tunnel. It might (ought to)
+ be for the host. */
+ if (ipset_contains_addr(st->networks,dest)) {
+ st->deliver_to_host(st->dst,buf);
+ st->outcount++;
+ BUF_ASSERT_FREE(buf);
+ } else {
+ string_t s,d;
+ s=ipaddr_to_string(source);
+ d=ipaddr_to_string(dest);
+ Message(M_DEBUG,"%s: don't know where to deliver packet "
+ "(s=%s, d=%s)\n", st->name, s, d);
+ free(s); free(d);
+ netlink_icmp_simple(st,buf,client,ICMP_TYPE_UNREACHABLE,
+ ICMP_CODE_NET_UNREACHABLE);
+ BUF_FREE(buf);
+ }
+ } else {
+ if (!allow_route &&
+ !(st->routes[best_match]->options&OPT_ALLOWROUTE)) {
+ string_t s,d;
+ s=ipaddr_to_string(source);
+ d=ipaddr_to_string(dest);
+ /* We have a usable route but aren't allowed to use it.
+ Generate ICMP destination unreachable: communication
+ with destination network administratively prohibited */
+ Message(M_NOTICE,"%s: denied forwarding for packet (s=%s, d=%s)\n",
+ st->name,s,d);
+ free(s); free(d);
+
+ netlink_icmp_simple(st,buf,client,ICMP_TYPE_UNREACHABLE,
+ ICMP_CODE_NET_PROHIBITED);
+ BUF_FREE(buf);
} else {
if (best_quality>0) {
- st->routes[best_match].c->deliver(
- st->routes[best_match].c->dst,
- st->routes[best_match].c, buf);
+ /* XXX Fragment if required */
+ st->routes[best_match]->deliver(
+ st->routes[best_match]->dst, buf);
+ st->routes[best_match]->outcount++;
BUF_ASSERT_FREE(buf);
} else {
/* Generate ICMP destination unreachable */
- netlink_icmp_simple(st,buf,client,3,0); /* client==NULL */
+ netlink_icmp_simple(st,buf,client,ICMP_TYPE_UNREACHABLE,
+ ICMP_CODE_NET_UNREACHABLE); /* client==NULL */
BUF_FREE(buf);
}
}
- } else { /* client is set */
- /* We know the origin is a tunnel - packet must be for the host */
- if (subnet_matches_list(&st->networks,dest)) {
- st->deliver_to_host(st->dst,NULL,buf);
- BUF_ASSERT_FREE(buf);
- } else {
- Message(M_ERROR,"%s: packet from tunnel %s can't be delivered "
- "to the host\n",st->name,client->name);
- netlink_icmp_simple(st,buf,client,3,0);
- BUF_FREE(buf);
- }
}
BUF_ASSERT_FREE(buf);
}
/* Packet has already been checked */
if (iph->ttl<=1) {
/* Generate ICMP time exceeded */
- netlink_icmp_simple(st,buf,client,11,0);
+ netlink_icmp_simple(st,buf,client,ICMP_TYPE_TIME_EXCEEDED,
+ ICMP_CODE_TTL_EXCEEDED);
BUF_FREE(buf);
return;
}
BUF_ASSERT_FREE(buf);
}
-/* Someone has been foolish enough to address a packet to us. I
- suppose we should reply to it, just to be polite. */
+/* Deal with packets addressed explicitly to us */
static void netlink_packet_local(struct netlink *st,
struct netlink_client *client,
struct buffer_if *buf)
{
struct icmphdr *h;
+ st->localcount++;
+
h=(struct icmphdr *)buf->start;
if ((ntohs(h->iph.frag_off)&0xbfff)!=0) {
- Message(M_WARNING,"%s: fragmented packet addressed to us\n",st->name);
+ Message(M_WARNING,"%s: fragmented packet addressed to secnet; "
+ "ignoring it\n",st->name);
BUF_FREE(buf);
return;
}
if (h->iph.protocol==1) {
/* It's ICMP */
- if (h->type==8 && h->code==0) {
+ if (h->type==ICMP_TYPE_ECHO_REQUEST && h->code==0) {
/* ICMP echo-request. Special case: we re-use the buffer
to construct the reply. */
- h->type=0;
+ h->type=ICMP_TYPE_ECHO_REPLY;
h->iph.daddr=h->iph.saddr;
h->iph.saddr=htonl(st->secnet_address);
- h->iph.ttl=255; /* Be nice and bump it up again... */
+ h->iph.ttl=255;
h->iph.check=0;
h->iph.check=ip_fast_csum((uint8_t *)h,h->iph.ihl);
netlink_icmp_csum(h);
Message(M_WARNING,"%s: unknown incoming ICMP\n",st->name);
} else {
/* Send ICMP protocol unreachable */
- netlink_icmp_simple(st,buf,client,3,2);
+ netlink_icmp_simple(st,buf,client,ICMP_TYPE_UNREACHABLE,
+ ICMP_CODE_PROTOCOL_UNREACHABLE);
BUF_FREE(buf);
return;
}
BUF_FREE(buf);
}
-/* Called by site code when remote packet is available */
-/* buf is allocated on entry and free on return */
-static void netlink_from_tunnel(void *sst, void *cst, struct buffer_if *buf)
+/* If cid==NULL packet is from host, otherwise cid specifies which tunnel
+ it came from. */
+static void netlink_incoming(struct netlink *st, struct netlink_client *client,
+ struct buffer_if *buf)
{
- struct netlink *st=sst;
- struct netlink_client *client=cst;
uint32_t source,dest;
struct iphdr *iph;
+ char errmsgbuf[50];
BUF_ASSERT_USED(buf);
- if (!netlink_check(st,buf)) {
- Message(M_WARNING,"%s: bad IP packet from tunnel %s\n",
- st->name,client->name);
+ if (!netlink_check(st,buf,errmsgbuf,sizeof(errmsgbuf))) {
+ Message(M_WARNING,"%s: bad IP packet from %s: %s\n",
+ st->name,client?client->name:"host",
+ errmsgbuf);
BUF_FREE(buf);
return;
}
source=ntohl(iph->saddr);
dest=ntohl(iph->daddr);
- /* Check that the packet source is in 'nets' and its destination is
- in st->networks */
- if (!subnet_matches_list(client->networks,source)) {
- string_t s,d;
- s=ipaddr_to_string(source);
- d=ipaddr_to_string(dest);
- Message(M_WARNING,"%s: packet from tunnel %s with bad source address "
- "(s=%s,d=%s)\n",st->name,client->name,s,d);
- free(s); free(d);
- BUF_FREE(buf);
- return;
+ /* Check source. If we don't like the source, there's no point
+ generating ICMP because we won't know how to get it to the
+ source of the packet. */
+ if (client) {
+ /* Check that the packet source is appropriate for the tunnel
+ it came down */
+ if (!ipset_contains_addr(client->networks,source)) {
+ string_t s,d;
+ s=ipaddr_to_string(source);
+ d=ipaddr_to_string(dest);
+ Message(M_WARNING,"%s: packet from tunnel %s with bad "
+ "source address (s=%s,d=%s)\n",st->name,client->name,s,d);
+ free(s); free(d);
+ BUF_FREE(buf);
+ return;
+ }
+ } else {
+ /* Check that the packet originates in our configured local
+ network, and hasn't been forwarded from elsewhere or
+ generated with the wrong source address */
+ if (!ipset_contains_addr(st->networks,source)) {
+ string_t s,d;
+ s=ipaddr_to_string(source);
+ d=ipaddr_to_string(dest);
+ Message(M_WARNING,"%s: outgoing packet with bad source address "
+ "(s=%s,d=%s)\n",st->name,s,d);
+ free(s); free(d);
+ BUF_FREE(buf);
+ return;
+ }
}
- /* (st->secnet_address needs checking before matching against
- st->networks because secnet's IP address may not be in the
- range the host is willing to deal with) */
- if (dest==st->secnet_address) {
- netlink_packet_local(st,client,buf);
+
+ /* If this is a point-to-point device we don't examine the
+ destination address at all; we blindly send it down our
+ one-and-only registered tunnel, or to the host, depending on
+ where it came from. It's up to external software to check
+ address validity and generate ICMP, etc. */
+ if (st->ptp) {
+ if (client) {
+ st->deliver_to_host(st->dst,buf);
+ } else {
+ st->clients->deliver(st->clients->dst,buf);
+ }
BUF_ASSERT_FREE(buf);
return;
}
- if (!subnet_matches_list(&st->networks,dest)) {
- string_t s,d;
- s=ipaddr_to_string(source);
- d=ipaddr_to_string(dest);
- Message(M_WARNING,"%s: incoming packet from tunnel %s "
- "with bad destination address "
- "(s=%s,d=%s)\n",st->name,client->name,s,d);
- free(s); free(d);
- BUF_FREE(buf);
+
+ /* st->secnet_address needs checking before matching destination
+ addresses */
+ if (dest==st->secnet_address) {
+ netlink_packet_local(st,client,buf);
+ BUF_ASSERT_FREE(buf);
return;
}
-
netlink_packet_forward(st,client,buf);
-
BUF_ASSERT_FREE(buf);
}
-/* Called by driver code when packet is received from kernel */
-/* cid should be NULL */
-/* buf should be allocated on entry, and is free on return */
-static void netlink_from_host(void *sst, void *cid, struct buffer_if *buf)
+static void netlink_inst_incoming(void *sst, struct buffer_if *buf)
{
- struct netlink *st=sst;
- uint32_t source,dest;
- struct iphdr *iph;
+ struct netlink_client *c=sst;
+ struct netlink *st=c->nst;
- BUF_ASSERT_USED(buf);
- if (!netlink_check(st,buf)) {
- Message(M_WARNING,"%s: bad IP packet from host\n",
- st->name);
- BUF_FREE(buf);
- return;
- }
- iph=(struct iphdr *)buf->start;
+ netlink_incoming(st,c,buf);
+}
- source=ntohl(iph->saddr);
- dest=ntohl(iph->daddr);
+static void netlink_dev_incoming(void *sst, struct buffer_if *buf)
+{
+ struct netlink *st=sst;
- if (!subnet_matches_list(&st->networks,source)) {
- string_t s,d;
- s=ipaddr_to_string(source);
- d=ipaddr_to_string(dest);
- Message(M_WARNING,"%s: outgoing packet with bad source address "
- "(s=%s,d=%s)\n",st->name,s,d);
- free(s); free(d);
- BUF_FREE(buf);
- return;
- }
- if (dest==st->secnet_address) {
- netlink_packet_local(st,NULL,buf);
- BUF_ASSERT_FREE(buf);
- return;
- }
- netlink_packet_forward(st,NULL,buf);
- BUF_ASSERT_FREE(buf);
+ netlink_incoming(st,NULL,buf);
}
-static void netlink_set_quality(void *sst, void *cid, uint32_t quality)
+static void netlink_set_quality(void *sst, uint32_t quality)
{
- struct netlink_client *c=cid;
+ struct netlink_client *c=sst;
+ struct netlink *st=c->nst;
c->link_quality=quality;
+ c->up=(c->link_quality==LINK_QUALITY_DOWN)?False:True;
+ if (c->options&OPT_SOFTROUTE) {
+ st->set_routes(st->dst,c);
+ }
}
-static void *netlink_regnets(void *sst, struct subnet_list *nets,
- netlink_deliver_fn *deliver, void *dst,
- uint32_t max_start_pad, uint32_t max_end_pad,
- string_t client_name)
+static void netlink_output_subnets(struct netlink *st, uint32_t loglevel,
+ struct subnet_list *snets)
{
- struct netlink *st=sst;
- struct netlink_client *c;
+ int32_t i;
+ string_t net;
- Message(M_DEBUG_CONFIG,"netlink_regnets: request for %d networks, "
- "max_start_pad=%d, max_end_pad=%d\n",
- nets->entries,max_start_pad,max_end_pad);
-
- /* Check that nets does not intersect with st->networks or
- st->exclude_remote_networks; refuse to register if it does. */
- if (subnet_lists_intersect(&st->networks,nets)) {
- Message(M_ERROR,"%s: site %s specifies networks that "
- "intersect with our local networks\n",st->name,client_name);
- return False;
- }
- if (subnet_lists_intersect(&st->exclude_remote_networks,nets)) {
- Message(M_ERROR,"%s: site %s specifies networks that "
- "intersect with the explicitly excluded remote networks\n",
- st->name,client_name);
- return False;
+ for (i=0; i<snets->entries; i++) {
+ net=subnet_to_string(snets->list[i]);
+ Message(loglevel,"%s ",net);
+ free(net);
}
+}
- c=safe_malloc(sizeof(*c),"netlink_regnets");
- c->networks=nets;
- c->deliver=deliver;
- c->dst=dst;
- c->name=client_name; /* XXX copy it? */
- c->link_quality=LINK_QUALITY_DOWN;
- c->next=st->clients;
- st->clients=c;
- if (max_start_pad > st->max_start_pad) st->max_start_pad=max_start_pad;
- if (max_end_pad > st->max_end_pad) st->max_end_pad=max_end_pad;
- st->n_routes+=nets->entries;
-
- return c;
+static void netlink_dump_routes(struct netlink *st, bool_t requested)
+{
+ int i;
+ string_t net;
+ uint32_t c=M_INFO;
+
+ if (requested) c=M_WARNING;
+ if (st->ptp) {
+ net=ipaddr_to_string(st->secnet_address);
+ Message(c,"%s: point-to-point (remote end is %s); routes:\n",
+ st->name, net);
+ free(net);
+ netlink_output_subnets(st,c,st->clients->subnets);
+ Message(c,"\n");
+ } else {
+ Message(c,"%s: routing table:\n",st->name);
+ for (i=0; i<st->n_clients; i++) {
+ netlink_output_subnets(st,c,st->routes[i]->subnets);
+ Message(c,"-> tunnel %s (%s,mtu %d,%s routes,%s,"
+ "quality %d,use %d,pri %lu)\n",
+ st->routes[i]->name,
+ st->routes[i]->up?"up":"down",
+ st->routes[i]->mtu,
+ st->routes[i]->options&OPT_SOFTROUTE?"soft":"hard",
+ st->routes[i]->options&OPT_ALLOWROUTE?"free":"restricted",
+ st->routes[i]->link_quality,
+ st->routes[i]->outcount,
+ (unsigned long)st->routes[i]->priority);
+ }
+ net=ipaddr_to_string(st->secnet_address);
+ Message(c,"%s/32 -> netlink \"%s\" (use %d)\n",
+ net,st->name,st->localcount);
+ free(net);
+ for (i=0; i<st->subnets->entries; i++) {
+ net=subnet_to_string(st->subnets->list[i]);
+ Message(c,"%s ",net);
+ free(net);
+ }
+ if (i>0)
+ Message(c,"-> host (use %d)\n",st->outcount);
+ }
}
-static int netlink_compare_route_specificity(const void *ap, const void *bp)
+/* ap is a pointer to a member of the routes array */
+static int netlink_compare_client_priority(const void *ap, const void *bp)
{
- const struct netlink_route *a=ap;
- const struct netlink_route *b=bp;
+ const struct netlink_client *const*a=ap;
+ const struct netlink_client *const*b=bp;
- if (a->net.len==b->net.len) return 0;
- if (a->net.len<b->net.len) return 1;
+ if ((*a)->priority==(*b)->priority) return 0;
+ if ((*a)->priority<(*b)->priority) return 1;
return -1;
}
{
struct netlink *st=sst;
struct netlink_client *c;
- uint32_t i,j;
+ int32_t i;
/* All the networks serviced by the various tunnels should now
* have been registered. We build a routing table by sorting the
- * routes into most-specific-first order. */
- st->routes=safe_malloc(st->n_routes*sizeof(*st->routes),
- "netlink_phase_hook");
+ * clients by priority. */
+ st->routes=safe_malloc_ary(sizeof(*st->routes),st->n_clients,
+ "netlink_phase_hook");
/* Fill the table */
i=0;
for (c=st->clients; c; c=c->next) {
- for (j=0; j<c->networks->entries; j++) {
- st->routes[i].net=c->networks->list[j];
- st->routes[i].c=c;
- i++;
- }
+ assert(i<INT_MAX);
+ st->routes[i++]=c;
}
- /* ASSERT i==st->n_routes */
- if (i!=st->n_routes) {
- fatal("netlink: route count error: expected %d got %d\n",
- st->n_routes,i);
- }
- /* Sort the table in descending order of specificity */
- qsort(st->routes,st->n_routes,sizeof(*st->routes),
- netlink_compare_route_specificity);
- Message(M_INFO,"%s: routing table:\n",st->name);
- for (i=0; i<st->n_routes; i++) {
- string_t net;
- net=subnet_to_string(&st->routes[i].net);
- Message(M_INFO,"%s -> tunnel %s\n",net,st->routes[i].c->name);
- free(net);
- }
- Message(M_INFO,"%s/32 -> netlink \"%s\"\n",
- ipaddr_to_string(st->secnet_address),st->name);
- for (i=0; i<st->networks.entries; i++) {
- string_t net;
- net=subnet_to_string(&st->networks.list[i]);
- Message(M_INFO,"%s -> host\n",net);
- free(net);
- }
-}
+ /* Sort the table in descending order of priority */
+ qsort(st->routes,st->n_clients,sizeof(*st->routes),
+ netlink_compare_client_priority);
-static netlink_deliver_fn *netlink_init(struct netlink *st,
- void *dst, struct cloc loc,
- dict_t *dict, string_t description,
- netlink_deliver_fn *to_host)
-{
- st->dst=dst;
- st->cl.description=description;
- st->cl.type=CL_NETLINK;
- st->cl.apply=NULL;
- st->cl.interface=&st->ops;
- st->ops.st=st;
- st->ops.regnets=netlink_regnets;
- st->ops.deliver=netlink_from_tunnel;
- st->ops.set_quality=netlink_set_quality;
- st->max_start_pad=0;
- st->max_end_pad=0;
- st->clients=NULL;
- st->deliver_to_host=to_host;
-
- st->name=dict_read_string(dict,"name",False,"netlink",loc);
- if (!st->name) st->name=description;
- dict_read_subnet_list(dict, "networks", True, "netlink", loc,
- &st->networks);
- dict_read_subnet_list(dict, "exclude-remote-networks", False, "netlink",
- loc, &st->exclude_remote_networks);
- /* local-address and secnet-address do not have to be in local-networks;
- however, they should be advertised in the 'sites' file for the
- local site. */
- st->local_address=string_to_ipaddr(
- dict_find_item(dict,"local-address", True, "netlink", loc),"netlink");
- st->secnet_address=string_to_ipaddr(
- dict_find_item(dict,"secnet-address", True, "netlink", loc),"netlink");
- st->mtu=dict_read_number(dict, "mtu", False, "netlink", loc, DEFAULT_MTU);
- buffer_new(&st->icmp,ICMP_BUFSIZE);
- st->n_routes=0;
- st->routes=NULL;
-
- add_hook(PHASE_SETUP,netlink_phase_hook,st);
-
- return netlink_from_host;
+ netlink_dump_routes(st,False);
}
-/* Connection to the kernel through userv-ipif */
-
-struct userv {
- struct netlink nl;
- int txfd; /* We transmit to userv */
- int rxfd; /* We receive from userv */
- string_t userv_path;
- string_t service_user;
- string_t service_name;
- uint32_t txbuflen;
- struct buffer_if *buff; /* We unstuff received packets into here
- and send them to the site code. */
- bool_t pending_esc;
- netlink_deliver_fn *netlink_to_tunnel;
-};
-
-static int userv_beforepoll(void *sst, struct pollfd *fds, int *nfds_io,
- int *timeout_io, const struct timeval *tv_now,
- uint64_t *now)
+static void netlink_signal_handler(void *sst, int signum)
{
- struct userv *st=sst;
- *nfds_io=2;
- fds[0].fd=st->txfd;
- fds[0].events=POLLERR; /* Might want to pick up POLLOUT sometime */
- fds[1].fd=st->rxfd;
- fds[1].events=POLLIN|POLLERR|POLLHUP;
- return 0;
+ struct netlink *st=sst;
+ Message(M_INFO,"%s: route dump requested by SIGUSR1\n",st->name);
+ netlink_dump_routes(st,True);
}
-static void userv_afterpoll(void *sst, struct pollfd *fds, int nfds,
- const struct timeval *tv_now, uint64_t *now)
+static void netlink_inst_set_mtu(void *sst, int32_t new_mtu)
{
- struct userv *st=sst;
- uint8_t rxbuf[DEFAULT_BUFSIZE];
- int l,i;
+ struct netlink_client *c=sst;
- if (fds[1].revents&POLLERR) {
- printf("userv_afterpoll: hup!\n");
- }
- if (fds[1].revents&POLLIN) {
- l=read(st->rxfd,rxbuf,DEFAULT_BUFSIZE);
- if (l<0) {
- fatal_perror("userv_afterpoll: read(rxfd)");
- }
- if (l==0) {
- fatal("userv_afterpoll: read(rxfd)=0; userv gone away?\n");
- }
- /* XXX really crude unstuff code */
- /* XXX check for buffer overflow */
- BUF_ASSERT_USED(st->buff);
- for (i=0; i<l; i++) {
- if (st->pending_esc) {
- st->pending_esc=False;
- switch(rxbuf[i]) {
- case SLIP_ESCEND:
- *(uint8_t *)buf_append(st->buff,1)=SLIP_END;
- break;
- case SLIP_ESCESC:
- *(uint8_t *)buf_append(st->buff,1)=SLIP_ESC;
- break;
- default:
- fatal("userv_afterpoll: bad SLIP escape character\n");
- }
- } else {
- switch (rxbuf[i]) {
- case SLIP_END:
- if (st->buff->size>0) {
- st->netlink_to_tunnel(&st->nl,NULL,
- st->buff);
- BUF_ALLOC(st->buff,"userv_afterpoll");
- }
- buffer_init(st->buff,st->nl.max_start_pad);
- break;
- case SLIP_ESC:
- st->pending_esc=True;
- break;
- default:
- *(uint8_t *)buf_append(st->buff,1)=rxbuf[i];
- break;
- }
- }
- }
- }
+ c->mtu=new_mtu;
}
-/* Send buf to the kernel. Free buf before returning. */
-static void userv_deliver_to_kernel(void *sst, void *cid,
- struct buffer_if *buf)
+static void netlink_inst_reg(void *sst, netlink_deliver_fn *deliver,
+ void *dst, int32_t max_start_pad)
{
- struct userv *st=sst;
- uint8_t txbuf[DEFAULT_BUFSIZE];
- uint8_t *i;
- uint32_t j;
+ struct netlink_client *c=sst;
+ struct netlink *st=c->nst;
- BUF_ASSERT_USED(buf);
-
- /* Spit the packet at userv-ipif: SLIP start marker, then
- bytestuff the packet, then SLIP end marker */
- /* XXX crunchy bytestuff code */
- j=0;
- txbuf[j++]=SLIP_END;
- for (i=buf->start; i<(buf->start+buf->size); i++) {
- switch (*i) {
- case SLIP_END:
- txbuf[j++]=SLIP_ESC;
- txbuf[j++]=SLIP_ESCEND;
- break;
- case SLIP_ESC:
- txbuf[j++]=SLIP_ESC;
- txbuf[j++]=SLIP_ESCESC;
- break;
- default:
- txbuf[j++]=*i;
- break;
- }
- }
- txbuf[j++]=SLIP_END;
- if (write(st->txfd,txbuf,j)<0) {
- fatal_perror("userv_deliver_to_kernel: write()");
- }
- BUF_FREE(buf);
+ if (max_start_pad > st->max_start_pad) st->max_start_pad=max_start_pad;
+ c->deliver=deliver;
+ c->dst=dst;
}
-static void userv_phase_hook(void *sst, uint32_t newphase)
+static struct flagstr netlink_option_table[]={
+ { "soft", OPT_SOFTROUTE },
+ { "allow-route", OPT_ALLOWROUTE },
+ { NULL, 0}
+};
+/* This is the routine that gets called when the closure that's
+ returned by an invocation of a netlink device closure (eg. tun,
+ userv-ipif) is invoked. It's used to create routes and pass in
+ information about them; the closure it returns is used by site
+ code. */
+static closure_t *netlink_inst_create(struct netlink *st,
+ struct cloc loc, dict_t *dict)
{
- struct userv *st=sst;
- pid_t child;
- int c_stdin[2];
- int c_stdout[2];
- string_t addrs;
- string_t nets;
- string_t s;
struct netlink_client *c;
- int i;
+ string_t name;
+ struct ipset *networks;
+ uint32_t options,priority;
+ int32_t mtu;
+ list_t *l;
+
+ name=dict_read_string(dict, "name", True, st->name, loc);
+
+ l=dict_lookup(dict,"routes");
+ if (!l)
+ cfgfatal(loc,st->name,"required parameter \"routes\" not found\n");
+ networks=string_list_to_ipset(l,loc,st->name,"routes");
+ options=string_list_to_word(dict_lookup(dict,"options"),
+ netlink_option_table,st->name);
+
+ priority=dict_read_number(dict,"priority",False,st->name,loc,0);
+ mtu=dict_read_number(dict,"mtu",False,st->name,loc,0);
+
+ if ((options&OPT_SOFTROUTE) && !st->set_routes) {
+ cfgfatal(loc,st->name,"this netlink device does not support "
+ "soft routes.\n");
+ return NULL;
+ }
- /* This is where we actually invoke userv - all the networks we'll
- be using should already have been registered. */
-
- addrs=safe_malloc(512,"userv_phase_hook:addrs");
- snprintf(addrs,512,"%s,%s,%d,slip",ipaddr_to_string(st->nl.local_address),
- ipaddr_to_string(st->nl.secnet_address),st->nl.mtu);
-
- nets=safe_malloc(1024,"userv_phase_hook:nets");
- *nets=0;
- for (c=st->nl.clients; c; c=c->next) {
- for (i=0; i<c->networks->entries; i++) {
- s=subnet_to_string(&c->networks->list[i]);
- strcat(nets,s);
- strcat(nets,",");
- free(s);
+ if (options&OPT_SOFTROUTE) {
+ /* XXX for now we assume that soft routes require root privilege;
+ this may not always be true. The device driver can tell us. */
+ require_root_privileges=True;
+ require_root_privileges_explanation="netlink: soft routes";
+ if (st->ptp) {
+ cfgfatal(loc,st->name,"point-to-point netlinks do not support "
+ "soft routes.\n");
+ return NULL;
}
}
- nets[strlen(nets)-1]=0;
-
- Message(M_INFO,"\nuserv_phase_hook: %s %s %s %s %s\n",st->userv_path,
- st->service_user,st->service_name,addrs,nets);
-
- /* Allocate buffer, plus space for padding. Make sure we end up
- with the start of the packet well-aligned. */
- /* ALIGN(st->max_start_pad,16); */
- /* ALIGN(st->max_end_pad,16); */
- st->pending_esc=False;
-
- /* Invoke userv */
- if (pipe(c_stdin)!=0) {
- fatal_perror("userv_phase_hook: pipe(c_stdin)");
- }
- if (pipe(c_stdout)!=0) {
- fatal_perror("userv_phase_hook: pipe(c_stdout)");
+ /* Check that nets are a subset of st->remote_networks;
+ refuse to register if they are not. */
+ if (!ipset_is_subset(st->remote_networks,networks)) {
+ cfgfatal(loc,st->name,"routes are not allowed\n");
+ return NULL;
}
- st->txfd=c_stdin[1];
- st->rxfd=c_stdout[0];
- child=fork();
- if (child==-1) {
- fatal_perror("userv_phase_hook: fork()");
- }
- if (child==0) {
- char **argv;
-
- /* We are the child. Modify our stdin and stdout, then exec userv */
- dup2(c_stdin[0],0);
- dup2(c_stdout[1],1);
- close(c_stdin[1]);
- close(c_stdout[0]);
-
- /* The arguments are:
- userv
- service-user
- service-name
- local-addr,secnet-addr,mtu,protocol
- route1,route2,... */
- argv=malloc(sizeof(*argv)*6);
- argv[0]=st->userv_path;
- argv[1]=st->service_user;
- argv[2]=st->service_name;
- argv[3]=addrs;
- argv[4]=nets;
- argv[5]=NULL;
- execvp(st->userv_path,argv);
- perror("netlink-userv-ipif: execvp");
-
- exit(1);
- }
- /* We are the parent... */
-
- /* Register for poll() */
- register_for_poll(st, userv_beforepoll, userv_afterpoll, 2, st->nl.name);
+ c=safe_malloc(sizeof(*c),"netlink_inst_create");
+ c->cl.description=name;
+ c->cl.type=CL_NETLINK;
+ c->cl.apply=NULL;
+ c->cl.interface=&c->ops;
+ c->ops.st=c;
+ c->ops.reg=netlink_inst_reg;
+ c->ops.deliver=netlink_inst_incoming;
+ c->ops.set_quality=netlink_set_quality;
+ c->ops.set_mtu=netlink_inst_set_mtu;
+ c->nst=st;
+
+ c->networks=networks;
+ c->subnets=ipset_to_subnet_list(networks);
+ c->priority=priority;
+ c->deliver=NULL;
+ c->dst=NULL;
+ c->name=name;
+ c->link_quality=LINK_QUALITY_UNUSED;
+ c->mtu=mtu?mtu:st->mtu;
+ c->options=options;
+ c->outcount=0;
+ c->up=False;
+ c->kup=False;
+ c->next=st->clients;
+ st->clients=c;
+ assert(st->n_clients < INT_MAX);
+ st->n_clients++;
+
+ return &c->cl;
}
-static list_t *userv_apply(closure_t *self, struct cloc loc, dict_t *context,
- list_t *args)
+static list_t *netlink_inst_apply(closure_t *self, struct cloc loc,
+ dict_t *context, list_t *args)
{
- struct userv *st;
- item_t *item;
- dict_t *dict;
+ struct netlink *st=self->interface;
- st=safe_malloc(sizeof(*st),"userv_apply");
+ dict_t *dict;
+ item_t *item;
+ closure_t *cl;
- /* First parameter must be a dict */
item=list_elem(args,0);
- if (!item || item->type!=t_dict)
- cfgfatal(loc,"userv-ipif","parameter must be a dictionary\n");
-
+ if (!item || item->type!=t_dict) {
+ cfgfatal(loc,st->name,"must have a dictionary argument\n");
+ }
dict=item->data.dict;
- st->netlink_to_tunnel=
- netlink_init(&st->nl,st,loc,dict,
- "netlink-userv-ipif",userv_deliver_to_kernel);
-
- st->userv_path=dict_read_string(dict,"userv-path",False,"userv-netlink",
- loc);
- st->service_user=dict_read_string(dict,"service-user",False,
- "userv-netlink",loc);
- st->service_name=dict_read_string(dict,"service-name",False,
- "userv-netlink",loc);
- if (!st->userv_path) st->userv_path="userv";
- if (!st->service_user) st->service_user="root";
- if (!st->service_name) st->service_name="ipif";
- st->buff=find_cl_if(dict,"buffer",CL_BUFFER,True,"userv-netlink",loc);
- BUF_ALLOC(st->buff,"netlink:userv_apply");
-
- st->rxfd=-1; st->txfd=-1;
- add_hook(PHASE_DROPPRIV,userv_phase_hook,st);
+ cl=netlink_inst_create(st,loc,dict);
- return new_closure(&st->nl.cl);
+ return new_closure(cl);
}
-/* Connection to the kernel through the universal TUN/TAP driver */
-
-struct tun {
- struct netlink nl;
- int fd;
- string_t device_path;
- string_t interface_name;
- string_t ifconfig_path;
- string_t route_path;
- bool_t tun_old;
- bool_t search_for_if; /* Applies to tun-old only */
- struct buffer_if *buff; /* We receive packets into here
- and send them to the netlink code. */
- netlink_deliver_fn *netlink_to_tunnel;
-};
-
-static int tun_beforepoll(void *sst, struct pollfd *fds, int *nfds_io,
- int *timeout_io, const struct timeval *tv_now,
- uint64_t *now)
+netlink_deliver_fn *netlink_init(struct netlink *st,
+ void *dst, struct cloc loc,
+ dict_t *dict, cstring_t description,
+ netlink_route_fn *set_routes,
+ netlink_deliver_fn *to_host)
{
- struct tun *st=sst;
- *nfds_io=1;
- fds[0].fd=st->fd;
- fds[0].events=POLLIN|POLLERR|POLLHUP;
- return 0;
-}
+ item_t *sa, *ptpa;
+ list_t *l;
-static void tun_afterpoll(void *sst, struct pollfd *fds, int nfds,
- const struct timeval *tv_now, uint64_t *now)
-{
- struct tun *st=sst;
- int l;
+ st->dst=dst;
+ st->cl.description=description;
+ st->cl.type=CL_PURE;
+ st->cl.apply=netlink_inst_apply;
+ st->cl.interface=st;
+ st->max_start_pad=0;
+ st->clients=NULL;
+ st->routes=NULL;
+ st->n_clients=0;
+ st->set_routes=set_routes;
+ st->deliver_to_host=to_host;
- if (fds[0].revents&POLLERR) {
- printf("tun_afterpoll: hup!\n");
- }
- if (fds[0].revents&POLLIN) {
- BUF_ALLOC(st->buff,"tun_afterpoll");
- buffer_init(st->buff,st->nl.max_start_pad);
- l=read(st->fd,st->buff->start,st->buff->len-st->nl.max_start_pad);
- if (l<0) {
- fatal_perror("tun_afterpoll: read()");
- }
- if (l==0) {
- fatal("tun_afterpoll: read()=0; device gone away?\n");
- }
- if (l>0) {
- st->buff->size=l;
- st->netlink_to_tunnel(&st->nl,NULL,st->buff);
- BUF_ASSERT_FREE(st->buff);
- }
+ st->name=dict_read_string(dict,"name",False,description,loc);
+ if (!st->name) st->name=description;
+ l=dict_lookup(dict,"networks");
+ if (l)
+ st->networks=string_list_to_ipset(l,loc,st->name,"networks");
+ else {
+ struct ipset *empty;
+ empty=ipset_new();
+ st->networks=ipset_complement(empty);
+ ipset_free(empty);
}
-}
-
-static void tun_deliver_to_kernel(void *sst, void *cid,
- struct buffer_if *buf)
-{
- struct tun *st=sst;
-
- BUF_ASSERT_USED(buf);
-
- /* No error checking, because we'd just throw the packet away anyway */
- write(st->fd,buf->start,buf->size);
- BUF_FREE(buf);
-}
-
-static void tun_phase_hook(void *sst, uint32_t newphase)
-{
- struct tun *st=sst;
- string_t hostaddr,secnetaddr;
- uint8_t mtu[6];
- string_t network,mask;
- struct netlink_client *c;
- int i;
-
- if (st->tun_old) {
- if (st->search_for_if) {
- string_t dname;
- int i;
-
- /* ASSERT st->interface_name */
- dname=safe_malloc(strlen(st->device_path)+4,"tun_old_apply");
- st->interface_name=safe_malloc(8,"tun_phase_hook");
-
- for (i=0; i<255; i++) {
- sprintf(dname,"%s%d",st->device_path,i);
- if ((st->fd=open(dname,O_RDWR))>0) {
- sprintf(st->interface_name,"tun%d",i);
- Message(M_INFO,"%s: allocated network interface %s "
- "through %s\n",st->nl.name,st->interface_name,
- dname);
- break;
- }
- }
- if (st->fd==-1) {
- fatal("%s: unable to open any TUN device (%s...)\n",
- st->nl.name,st->device_path);
- }
- } else {
- st->fd=open(st->device_path,O_RDWR);
- if (st->fd==-1) {
- fatal_perror("%s: unable to open TUN device file %s",
- st->nl.name,st->device_path);
- }
- }
+ l=dict_lookup(dict,"remote-networks");
+ if (l) {
+ st->remote_networks=string_list_to_ipset(l,loc,st->name,
+ "remote-networks");
} else {
-#ifdef HAVE_LINUX_IF_H
- struct ifreq ifr;
-
- /* New TUN interface: open the device, then do ioctl TUNSETIFF
- to set or find out the network interface name. */
- st->fd=open(st->device_path,O_RDWR);
- if (st->fd==-1) {
- fatal_perror("%s: can't open device file %s",st->nl.name,
- st->device_path);
- }
- memset(&ifr,0,sizeof(ifr));
- ifr.ifr_flags = IFF_TUN | IFF_NO_PI; /* Just send/receive IP packets,
- no extra headers */
- if (st->interface_name)
- strncpy(ifr.ifr_name,st->interface_name,IFNAMSIZ);
- Message(M_INFO,"%s: about to ioctl(TUNSETIFF)...\n",st->nl.name);
- if (ioctl(st->fd,TUNSETIFF,&ifr)<0) {
- fatal_perror("%s: ioctl(TUNSETIFF)",st->nl.name);
- }
- if (!st->interface_name) {
- st->interface_name=safe_malloc(strlen(ifr.ifr_name)+1,"tun_apply");
- strcpy(st->interface_name,ifr.ifr_name);
- Message(M_INFO,"%s: allocated network interface %s\n",st->nl.name,
- st->interface_name);
- }
-#else
- fatal("netlink.c:tun_phase_hook:!tun_old unexpected\n");
-#endif /* HAVE_LINUX_IF_H */
- }
- /* All the networks we'll be using have been registered. Invoke ifconfig
- to set the TUN device's address, and route to add routes to all
- our networks. */
-
- hostaddr=ipaddr_to_string(st->nl.local_address);
- secnetaddr=ipaddr_to_string(st->nl.secnet_address);
- snprintf(mtu,6,"%d",st->nl.mtu);
- mtu[5]=0;
-
- sys_cmd(st->ifconfig_path,"ifconfig",st->interface_name,
- hostaddr,"netmask","255.255.255.255","-broadcast",
- "pointopoint",secnetaddr,"mtu",mtu,"up",(char *)0);
-
- for (c=st->nl.clients; c; c=c->next) {
- for (i=0; i<c->networks->entries; i++) {
- network=ipaddr_to_string(c->networks->list[i].prefix);
- mask=ipaddr_to_string(c->networks->list[i].mask);
- sys_cmd(st->route_path,"route","add","-net",network,
- "netmask",mask,"gw",secnetaddr,(char *)0);
- }
+ struct ipset *empty;
+ empty=ipset_new();
+ st->remote_networks=ipset_complement(empty);
+ ipset_free(empty);
}
- /* Register for poll() */
- register_for_poll(st, tun_beforepoll, tun_afterpoll, 1, st->nl.name);
-}
-
-#ifdef HAVE_LINUX_IF_H
-static list_t *tun_apply(closure_t *self, struct cloc loc, dict_t *context,
- list_t *args)
-{
- struct tun *st;
- item_t *item;
- dict_t *dict;
-
- st=safe_malloc(sizeof(*st),"tun_apply");
-
- /* First parameter must be a dict */
- item=list_elem(args,0);
- if (!item || item->type!=t_dict)
- cfgfatal(loc,"tun","parameter must be a dictionary\n");
-
- dict=item->data.dict;
-
- st->netlink_to_tunnel=
- netlink_init(&st->nl,st,loc,dict,
- "netlink-tun",tun_deliver_to_kernel);
-
- st->tun_old=False;
- st->device_path=dict_read_string(dict,"device",False,"tun-netlink",loc);
- st->interface_name=dict_read_string(dict,"interface",False,
- "tun-netlink",loc);
- st->ifconfig_path=dict_read_string(dict,"ifconfig-path",
- False,"tun-netlink",loc);
- st->route_path=dict_read_string(dict,"route-path",
- False,"tun-netlink",loc);
-
- if (!st->device_path) st->device_path="/dev/net/tun";
- if (!st->ifconfig_path) st->ifconfig_path="ifconfig";
- if (!st->route_path) st->route_path="route";
- st->buff=find_cl_if(dict,"buffer",CL_BUFFER,True,"tun-netlink",loc);
-
- add_hook(PHASE_GETRESOURCES,tun_phase_hook,st);
-
- return new_closure(&st->nl.cl);
-}
-#endif /* HAVE_LINUX_IF_H */
-
-static list_t *tun_old_apply(closure_t *self, struct cloc loc, dict_t *context,
- list_t *args)
-{
- struct tun *st;
- item_t *item;
- dict_t *dict;
-
- st=safe_malloc(sizeof(*st),"tun_old_apply");
-
- Message(M_WARNING,"the tun-old code has never been tested. Please report "
- "success or failure to steve@greenend.org.uk\n");
-
- /* First parameter must be a dict */
- item=list_elem(args,0);
- if (!item || item->type!=t_dict)
- cfgfatal(loc,"tun","parameter must be a dictionary\n");
-
- dict=item->data.dict;
-
- st->netlink_to_tunnel=
- netlink_init(&st->nl,st,loc,dict,
- "netlink-tun",tun_deliver_to_kernel);
-
- st->tun_old=True;
- st->device_path=dict_read_string(dict,"device",False,"tun-netlink",loc);
- st->interface_name=dict_read_string(dict,"interface",False,
- "tun-netlink",loc);
- st->search_for_if=dict_read_bool(dict,"interface-search",False,
- "tun-netlink",loc,st->device_path==NULL);
- st->ifconfig_path=dict_read_string(dict,"ifconfig-path",False,
- "tun-netlink",loc);
- st->route_path=dict_read_string(dict,"route-path",False,"tun-netlink",loc);
-
- if (!st->device_path) st->device_path="/dev/tun";
- if (!st->ifconfig_path) st->ifconfig_path="ifconfig";
- if (!st->route_path) st->route_path="route";
- st->buff=find_cl_if(dict,"buffer",CL_BUFFER,True,"tun-netlink",loc);
-
- /* Old TUN interface: the network interface name depends on which
- /dev/tunX file we open. If 'interface-search' is set to true, treat
- 'device' as the prefix and try numbers from 0--255. If it's set
- to false, treat 'device' as the whole name, and require than an
- appropriate interface name be specified. */
- if (st->search_for_if && st->interface_name) {
- cfgfatal(loc,"tun-old","you may not specify an interface name "
- "in interface-search mode\n");
+ sa=dict_find_item(dict,"secnet-address",False,"netlink",loc);
+ ptpa=dict_find_item(dict,"ptp-address",False,"netlink",loc);
+ if (sa && ptpa) {
+ cfgfatal(loc,st->name,"you may not specify secnet-address and "
+ "ptp-address in the same netlink device\n");
}
- if (!st->search_for_if && !st->interface_name) {
- cfgfatal(loc,"tun-old","you must specify an interface name "
- "when you explicitly specify a TUN device file\n");
+ if (!(sa || ptpa)) {
+ cfgfatal(loc,st->name,"you must specify secnet-address or "
+ "ptp-address for this netlink device\n");
}
+ if (sa) {
+ st->secnet_address=string_item_to_ipaddr(sa,"netlink");
+ st->ptp=False;
+ } else {
+ st->secnet_address=string_item_to_ipaddr(ptpa,"netlink");
+ st->ptp=True;
+ }
+ /* To be strictly correct we could subtract secnet_address from
+ networks here. It shouldn't make any practical difference,
+ though, and will make the route dump look complicated... */
+ st->subnets=ipset_to_subnet_list(st->networks);
+ st->mtu=dict_read_number(dict, "mtu", False, "netlink", loc, DEFAULT_MTU);
+ buffer_new(&st->icmp,ICMP_BUFSIZE);
+ st->outcount=0;
+ st->localcount=0;
-
- add_hook(PHASE_GETRESOURCES,tun_phase_hook,st);
-
- return new_closure(&st->nl.cl);
+ add_hook(PHASE_SETUP,netlink_phase_hook,st);
+ request_signal_notification(SIGUSR1, netlink_signal_handler, st);
+
+ /* If we're point-to-point then we return a CL_NETLINK directly,
+ rather than a CL_NETLINK_OLD or pure closure (depending on
+ compatibility). This CL_NETLINK is for our one and only
+ client. Our cl.apply function is NULL. */
+ if (st->ptp) {
+ closure_t *cl;
+ cl=netlink_inst_create(st,loc,dict);
+ st->cl=*cl;
+ }
+ return netlink_dev_incoming;
}
/* No connection to the kernel at all... */
struct netlink nl;
};
-static void null_deliver(void *sst, void *cid, struct buffer_if *buf)
+static bool_t null_set_route(void *sst, struct netlink_client *routes)
+{
+ struct null *st=sst;
+
+ if (routes->up!=routes->kup) {
+ Message(M_INFO,"%s: setting routes for tunnel %s to state %s\n",
+ st->nl.name,routes->name,
+ routes->up?"up":"down");
+ routes->kup=routes->up;
+ return True;
+ }
+ return False;
+}
+
+static void null_deliver(void *sst, struct buffer_if *buf)
{
return;
}
dict=item->data.dict;
- netlink_init(&st->nl,st,loc,dict,"null-netlink",null_deliver);
+ netlink_init(&st->nl,st,loc,dict,"null-netlink",null_set_route,
+ null_deliver);
return new_closure(&st->nl.cl);
}
-init_module netlink_module;
void netlink_module(dict_t *dict)
{
- add_closure(dict,"userv-ipif",userv_apply);
-#ifdef HAVE_LINUX_IF_H
- add_closure(dict,"tun",tun_apply);
-#endif
- add_closure(dict,"tun-old",tun_old_apply);
add_closure(dict,"null-netlink",null_apply);
-#if 0
- /* TODO */
- add_closure(dict,"pty-slip",ptyslip_apply);
- add_closure(dict,"slipd",slipd_apply);
-#endif /* 0 */
}
~ian / secnet.git / blobdiff