d/freebsd-nq
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
1de8bd1c41
freebsd-nq/sys/netinet/ip_mroute.c
Jeffrey Hsu 9ba208b413 * Bug fix in bw_meter_process(): the periodically processed bins 
of bw_meter entries were processed up to one second ahead.
  After an unappropriate rescheduling of some of the bw_meter
  entries, the upcalls weren't delivered.

* pim_register_prepare() uses the appropriate sw_csum flag to
  call ip_fragment() so the IP checksum is computed properly.

* Modify pim_register_prepare() to take care of IP packets that
  don't need fragmentation.

* Add-back in_delayed_cksum() to encap_send(), because it seems it
  should be there.

Submitted by:	Pavlin Radoslavov <pavlin@icir.org>
2003-08-19 17:22:51 +00:00

3296 lines
84 KiB
C
Raw Blame History

/*
* IP multicast forwarding procedures
*
* Written by David Waitzman, BBN Labs, August 1988.
* Modified by Steve Deering, Stanford, February 1989.
* Modified by Mark J. Steiglitz, Stanford, May, 1991
* Modified by Van Jacobson, LBL, January 1993
* Modified by Ajit Thyagarajan, PARC, August 1993
* Modified by Bill Fenner, PARC, April 1995
* Modified by Ahmed Helmy, SGI, June 1996
* Modified by George Edmond Eddy (Rusty), ISI, February 1998
* Modified by Pavlin Radoslavov, USC/ISI, May 1998, August 1999, October 2000
* Modified by Hitoshi Asaeda, WIDE, August 2000
* Modified by Pavlin Radoslavov, ICSI, October 2002
*
* MROUTING Revision: 3.5
* and PIM-SMv2 and PIM-DM support, advanced API support,
* bandwidth metering and signaling
*
* $FreeBSD$
*/
#include "opt_mac.h"
#include "opt_mrouting.h"
#include "opt_random_ip_id.h"
#ifdef PIM
#define _PIM_VT 1
#endif
#include <sys/param.h>
#include <sys/kernel.h>
#include <sys/lock.h>
#include <sys/mac.h>
#include <sys/malloc.h>
#include <sys/mbuf.h>
#include <sys/protosw.h>
#include <sys/signalvar.h>
#include <sys/socket.h>
#include <sys/socketvar.h>
#include <sys/sockio.h>
#include <sys/sx.h>
#include <sys/sysctl.h>
#include <sys/syslog.h>
#include <sys/systm.h>
#include <sys/time.h>
#include <net/if.h>
#include <net/netisr.h>
#include <net/route.h>
#include <netinet/in.h>
#include <netinet/igmp.h>
#include <netinet/in_systm.h>
#include <netinet/in_var.h>
#include <netinet/ip.h>
#include <netinet/ip_encap.h>
#include <netinet/ip_mroute.h>
#include <netinet/ip_var.h>
#ifdef PIM
#include <netinet/pim.h>
#include <netinet/pim_var.h>
#endif
#include <netinet/udp.h>
#include <machine/in_cksum.h>
/*
* Control debugging code for rsvp and multicast routing code.
* Can only set them with the debugger.
*/
static u_int rsvpdebug; /* non-zero enables debugging */
static u_int mrtdebug; /* any set of the flags below */
#define DEBUG_MFC 0x02
#define DEBUG_FORWARD 0x04
#define DEBUG_EXPIRE 0x08
#define DEBUG_XMIT 0x10
#define DEBUG_PIM 0x20
#define VIFI_INVALID ((vifi_t) -1)
#define M_HASCL(m) ((m)->m_flags & M_EXT)
static MALLOC_DEFINE(M_MRTABLE, "mroutetbl", "multicast routing tables");
static struct mrtstat mrtstat;
SYSCTL_STRUCT(_net_inet_ip, OID_AUTO, mrtstat, CTLFLAG_RW,
&mrtstat, mrtstat,
"Multicast Routing Statistics (struct mrtstat, netinet/ip_mroute.h)");
static struct mfc *mfctable[MFCTBLSIZ];
SYSCTL_OPAQUE(_net_inet_ip, OID_AUTO, mfctable, CTLFLAG_RD,
&mfctable, sizeof(mfctable), "S,*mfc[MFCTBLSIZ]",
"Multicast Forwarding Table (struct *mfc[MFCTBLSIZ], netinet/ip_mroute.h)");
static struct vif viftable[MAXVIFS];
SYSCTL_OPAQUE(_net_inet_ip, OID_AUTO, viftable, CTLFLAG_RD,
&viftable, sizeof(viftable), "S,vif[MAXVIFS]",
"Multicast Virtual Interfaces (struct vif[MAXVIFS], netinet/ip_mroute.h)");
static u_char nexpire[MFCTBLSIZ];
static struct callout_handle expire_upcalls_ch;
#define EXPIRE_TIMEOUT (hz / 4) /* 4x / second */
#define UPCALL_EXPIRE 6 /* number of timeouts */
/*
* Define the token bucket filter structures
* tbftable -> each vif has one of these for storing info
*/
static struct tbf tbftable[MAXVIFS];
#define TBF_REPROCESS (hz / 100) /* 100x / second */
/*
* 'Interfaces' associated with decapsulator (so we can tell
* packets that went through it from ones that get reflected
* by a broken gateway). These interfaces are never linked into
* the system ifnet list & no routes point to them. I.e., packets
* can't be sent this way. They only exist as a placeholder for
* multicast source verification.
*/
static struct ifnet multicast_decap_if[MAXVIFS];
#define ENCAP_TTL 64
#define ENCAP_PROTO IPPROTO_IPIP /* 4 */
/* prototype IP hdr for encapsulated packets */
static struct ip multicast_encap_iphdr = {
#if BYTE_ORDER == LITTLE_ENDIAN
sizeof(struct ip) >> 2, IPVERSION,
#else
IPVERSION, sizeof(struct ip) >> 2,
#endif
0, /* tos */
sizeof(struct ip), /* total length */
0, /* id */
0, /* frag offset */
ENCAP_TTL, ENCAP_PROTO,
0, /* checksum */
};
/*
* Bandwidth meter variables and constants
*/
static MALLOC_DEFINE(M_BWMETER, "bwmeter", "multicast upcall bw meters");
/*
* Pending timeouts are stored in a hash table, the key being the
* expiration time. Periodically, the entries are analysed and processed.
*/
#define BW_METER_BUCKETS 1024
static struct bw_meter *bw_meter_timers[BW_METER_BUCKETS];
static struct callout_handle bw_meter_ch;
#define BW_METER_PERIOD (hz) /* periodical handling of bw meters */
/*
* Pending upcalls are stored in a vector which is flushed when
* full, or periodically
*/
static struct bw_upcall bw_upcalls[BW_UPCALLS_MAX];
static u_int bw_upcalls_n; /* # of pending upcalls */
static struct callout_handle bw_upcalls_ch;
#define BW_UPCALLS_PERIOD (hz) /* periodical flush of bw upcalls */
#ifdef PIM
static struct pimstat pimstat;
SYSCTL_STRUCT(_net_inet_pim, PIMCTL_STATS, stats, CTLFLAG_RD,
&pimstat, pimstat,
"PIM Statistics (struct pimstat, netinet/pim_var.h)");
/*
* Note: the PIM Register encapsulation adds the following in front of a
* data packet:
*
* struct pim_encap_hdr {
* struct ip ip;
* struct pim_encap_pimhdr pim;
* }
*
*/
struct pim_encap_pimhdr {
struct pim pim;
uint32_t flags;
};
static struct ip pim_encap_iphdr = {
#if BYTE_ORDER == LITTLE_ENDIAN
sizeof(struct ip) >> 2,
IPVERSION,
#else
IPVERSION,
sizeof(struct ip) >> 2,
#endif
0, /* tos */
sizeof(struct ip), /* total length */
0, /* id */
0, /* frag offset */
ENCAP_TTL,
IPPROTO_PIM,
0, /* checksum */
};
static struct pim_encap_pimhdr pim_encap_pimhdr = {
{
PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */
0, /* reserved */
0, /* checksum */
},
0 /* flags */
};
static struct ifnet multicast_register_if;
static vifi_t reg_vif_num = VIFI_INVALID;
#endif /* PIM */
/*
* Private variables.
*/
static vifi_t numvifs;
static const struct encaptab *encap_cookie;
/*
* one-back cache used by mroute_encapcheck to locate a tunnel's vif
* given a datagram's src ip address.
*/
static u_long last_encap_src;
static struct vif *last_encap_vif;
static u_long X_ip_mcast_src(int vifi);
static int X_ip_mforward(struct ip *ip, struct ifnet *ifp,
struct mbuf *m, struct ip_moptions *imo);
static int X_ip_mrouter_done(void);
static int X_ip_mrouter_get(struct socket *so, struct sockopt *m);
static int X_ip_mrouter_set(struct socket *so, struct sockopt *m);
static int X_legal_vif_num(int vif);
static int X_mrt_ioctl(int cmd, caddr_t data);
static int get_sg_cnt(struct sioc_sg_req *);
static int get_vif_cnt(struct sioc_vif_req *);
static int ip_mrouter_init(struct socket *, int);
static int add_vif(struct vifctl *);
static int del_vif(vifi_t);
static int add_mfc(struct mfcctl2 *);
static int del_mfc(struct mfcctl2 *);
static int set_api_config(uint32_t *); /* chose API capabilities */
static int socket_send(struct socket *, struct mbuf *, struct sockaddr_in *);
static int set_assert(int);
static void expire_upcalls(void *);
static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, vifi_t);
static void phyint_send(struct ip *, struct vif *, struct mbuf *);
static void encap_send(struct ip *, struct vif *, struct mbuf *);
static void tbf_control(struct vif *, struct mbuf *, struct ip *, u_long);
static void tbf_queue(struct vif *, struct mbuf *);
static void tbf_process_q(struct vif *);
static void tbf_reprocess_q(void *);
static int tbf_dq_sel(struct vif *, struct ip *);
static void tbf_send_packet(struct vif *, struct mbuf *);
static void tbf_update_tokens(struct vif *);
static int priority(struct vif *, struct ip *);
/*
* Bandwidth monitoring
*/
static void free_bw_list(struct bw_meter *list);
static int add_bw_upcall(struct bw_upcall *);
static int del_bw_upcall(struct bw_upcall *);
static void bw_meter_receive_packet(struct bw_meter *x, int plen,
struct timeval *nowp);
static void bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp);
static void bw_upcalls_send(void);
static void schedule_bw_meter(struct bw_meter *x, struct timeval *nowp);
static void unschedule_bw_meter(struct bw_meter *x);
static void bw_meter_process(void);
static void expire_bw_upcalls_send(void *);
static void expire_bw_meter_process(void *);
#ifdef PIM
static int pim_register_send(struct ip *, struct vif *,
struct mbuf *, struct mfc *);
static int pim_register_send_rp(struct ip *, struct vif *,
struct mbuf *, struct mfc *);
static int pim_register_send_upcall(struct ip *, struct vif *,
struct mbuf *, struct mfc *);
static struct mbuf *pim_register_prepare(struct ip *, struct mbuf *);
#endif
/*
* whether or not special PIM assert processing is enabled.
*/
static int pim_assert;
/*
* Rate limit for assert notification messages, in usec
*/
#define ASSERT_MSG_TIME 3000000
/*
* Kernel multicast routing API capabilities and setup.
* If more API capabilities are added to the kernel, they should be
* recorded in `mrt_api_support'.
*/
static const uint32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF |
MRT_MFC_FLAGS_BORDER_VIF |
MRT_MFC_RP |
MRT_MFC_BW_UPCALL);
static uint32_t mrt_api_config = 0;
/*
* Hash function for a source, group entry
*/
#define MFCHASH(a, g) MFCHASHMOD(((a) >> 20) ^ ((a) >> 10) ^ (a) ^ \
((g) >> 20) ^ ((g) >> 10) ^ (g))
/*
* Find a route for a given origin IP address and Multicast group address
* Type of service parameter to be added in the future!!!
* Statistics are updated by the caller if needed
* (mrtstat.mrts_mfc_lookups and mrtstat.mrts_mfc_misses)
*/
static struct mfc *
mfc_find(in_addr_t o, in_addr_t g)
{
struct mfc *rt;
for (rt = mfctable[MFCHASH(o,g)]; rt; rt = rt->mfc_next)
if ((rt->mfc_origin.s_addr == o) &&
(rt->mfc_mcastgrp.s_addr == g) && (rt->mfc_stall == NULL))
break;
return rt;
}
/*
* Macros to compute elapsed time efficiently
* Borrowed from Van Jacobson's scheduling code
*/
#define TV_DELTA(a, b, delta) { \
int xxs; \
delta = (a).tv_usec - (b).tv_usec; \
if ((xxs = (a).tv_sec - (b).tv_sec)) { \
switch (xxs) { \
case 2: \
delta += 1000000; \
/* FALLTHROUGH */ \
case 1: \
delta += 1000000; \
break; \
default: \
delta += (1000000 * xxs); \
} \
} \
}
#define TV_LT(a, b) (((a).tv_usec < (b).tv_usec && \
(a).tv_sec <= (b).tv_sec) || (a).tv_sec < (b).tv_sec)
/*
* Handle MRT setsockopt commands to modify the multicast routing tables.
*/
static int
X_ip_mrouter_set(struct socket *so, struct sockopt *sopt)
{
int error, optval;
vifi_t vifi;
struct vifctl vifc;
struct mfcctl2 mfc;
struct bw_upcall bw_upcall;
uint32_t i;
if (so != ip_mrouter && sopt->sopt_name != MRT_INIT)
return EPERM;
error = 0;
switch (sopt->sopt_name) {
case MRT_INIT:
error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
if (error)
break;
error = ip_mrouter_init(so, optval);
break;
case MRT_DONE:
error = ip_mrouter_done();
break;
case MRT_ADD_VIF:
error = sooptcopyin(sopt, &vifc, sizeof vifc, sizeof vifc);
if (error)
break;
error = add_vif(&vifc);
break;
case MRT_DEL_VIF:
error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi);
if (error)
break;
error = del_vif(vifi);
break;
case MRT_ADD_MFC:
case MRT_DEL_MFC:
bzero((caddr_t)&mfc, sizeof(mfc)); /* zero out all fields */
/*
* select data size depending on API version.
*/
if (sopt->sopt_name == MRT_ADD_MFC &&
mrt_api_config & MRT_API_FLAGS_ALL) {
error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl2),
sizeof(struct mfcctl2));
} else {
error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl),
sizeof(struct mfcctl));
bzero((caddr_t)&mfc + sizeof(struct mfcctl),
sizeof(mfc) - sizeof(struct mfcctl));
}
if (error)
break;
if (sopt->sopt_name == MRT_ADD_MFC)
error = add_mfc(&mfc);
else
error = del_mfc(&mfc);
break;
case MRT_ASSERT:
error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
if (error)
break;
set_assert(optval);
break;
case MRT_API_CONFIG:
error = sooptcopyin(sopt, &i, sizeof i, sizeof i);
if (!error)
error = set_api_config(&i);
if (!error)
error = sooptcopyout(sopt, &i, sizeof i);
break;
case MRT_ADD_BW_UPCALL:
case MRT_DEL_BW_UPCALL:
error = sooptcopyin(sopt, &bw_upcall, sizeof bw_upcall,
sizeof bw_upcall);
if (error)
break;
if (sopt->sopt_name == MRT_ADD_BW_UPCALL)
error = add_bw_upcall(&bw_upcall);
else
error = del_bw_upcall(&bw_upcall);
break;
default:
error = EOPNOTSUPP;
break;
}
return error;
}
/*
* Handle MRT getsockopt commands
*/
static int
X_ip_mrouter_get(struct socket *so, struct sockopt *sopt)
{
int error;
static int version = 0x0305; /* !!! why is this here? XXX */
switch (sopt->sopt_name) {
case MRT_VERSION:
error = sooptcopyout(sopt, &version, sizeof version);
break;
case MRT_ASSERT:
error = sooptcopyout(sopt, &pim_assert, sizeof pim_assert);
break;
case MRT_API_SUPPORT:
error = sooptcopyout(sopt, &mrt_api_support, sizeof mrt_api_support);
break;
case MRT_API_CONFIG:
error = sooptcopyout(sopt, &mrt_api_config, sizeof mrt_api_config);
break;
default:
error = EOPNOTSUPP;
break;
}
return error;
}
/*
* Handle ioctl commands to obtain information from the cache
*/
static int
X_mrt_ioctl(int cmd, caddr_t data)
{
int error = 0;
switch (cmd) {
case (SIOCGETVIFCNT):
error = get_vif_cnt((struct sioc_vif_req *)data);
break;
case (SIOCGETSGCNT):
error = get_sg_cnt((struct sioc_sg_req *)data);
break;
default:
error = EINVAL;
break;
}
return error;
}
/*
* returns the packet, byte, rpf-failure count for the source group provided
*/
static int
get_sg_cnt(struct sioc_sg_req *req)
{
int s;
struct mfc *rt;
s = splnet();
rt = mfc_find(req->src.s_addr, req->grp.s_addr);
splx(s);
if (rt == NULL) {
req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
return EADDRNOTAVAIL;
}
req->pktcnt = rt->mfc_pkt_cnt;
req->bytecnt = rt->mfc_byte_cnt;
req->wrong_if = rt->mfc_wrong_if;
return 0;
}
/*
* returns the input and output packet and byte counts on the vif provided
*/
static int
get_vif_cnt(struct sioc_vif_req *req)
{
vifi_t vifi = req->vifi;
if (vifi >= numvifs)
return EINVAL;
req->icount = viftable[vifi].v_pkt_in;
req->ocount = viftable[vifi].v_pkt_out;
req->ibytes = viftable[vifi].v_bytes_in;
req->obytes = viftable[vifi].v_bytes_out;
return 0;
}
/*
* Enable multicast routing
*/
static int
ip_mrouter_init(struct socket *so, int version)
{
if (mrtdebug)
log(LOG_DEBUG, "ip_mrouter_init: so_type = %d, pr_protocol = %d\n",
so->so_type, so->so_proto->pr_protocol);
if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_IGMP)
return EOPNOTSUPP;
if (version != 1)
return ENOPROTOOPT;
if (ip_mrouter != NULL)
return EADDRINUSE;
ip_mrouter = so;
bzero((caddr_t)mfctable, sizeof(mfctable));
bzero((caddr_t)nexpire, sizeof(nexpire));
pim_assert = 0;
expire_upcalls_ch = timeout(expire_upcalls, NULL, EXPIRE_TIMEOUT);
bw_upcalls_n = 0;
bzero((caddr_t)bw_meter_timers, sizeof(bw_meter_timers));
bw_upcalls_ch = timeout(expire_bw_upcalls_send, NULL, BW_UPCALLS_PERIOD);
bw_meter_ch = timeout(expire_bw_meter_process, NULL, BW_METER_PERIOD);
mrt_api_config = 0;
if (mrtdebug)
log(LOG_DEBUG, "ip_mrouter_init\n");
return 0;
}
/*
* Disable multicast routing
*/
static int
X_ip_mrouter_done(void)
{
vifi_t vifi;
int i;
struct ifnet *ifp;
struct ifreq ifr;
struct mfc *rt;
struct rtdetq *rte;
int s;
s = splnet();
/*
* For each phyint in use, disable promiscuous reception of all IP
* multicasts.
*/
for (vifi = 0; vifi < numvifs; vifi++) {
if (viftable[vifi].v_lcl_addr.s_addr != 0 &&
!(viftable[vifi].v_flags & (VIFF_TUNNEL | VIFF_REGISTER))) {
struct sockaddr_in *so = (struct sockaddr_in *)&(ifr.ifr_addr);
so->sin_len = sizeof(struct sockaddr_in);
so->sin_family = AF_INET;
so->sin_addr.s_addr = INADDR_ANY;
ifp = viftable[vifi].v_ifp;
if_allmulti(ifp, 0);
}
}
bzero((caddr_t)tbftable, sizeof(tbftable));
bzero((caddr_t)viftable, sizeof(viftable));
numvifs = 0;
pim_assert = 0;
untimeout(expire_upcalls, NULL, expire_upcalls_ch);
mrt_api_config = 0;
bw_upcalls_n = 0;
untimeout(expire_bw_upcalls_send, NULL, bw_upcalls_ch);
untimeout(expire_bw_meter_process, NULL, bw_meter_ch);
/*
* Free all multicast forwarding cache entries.
*/
for (i = 0; i < MFCTBLSIZ; i++) {
for (rt = mfctable[i]; rt != NULL; ) {
struct mfc *nr = rt->mfc_next;
for (rte = rt->mfc_stall; rte != NULL; ) {
struct rtdetq *n = rte->next;
m_freem(rte->m);
free(rte, M_MRTABLE);
rte = n;
}
free_bw_list(rt->mfc_bw_meter);
free(rt, M_MRTABLE);
rt = nr;
}
}
bzero((caddr_t)mfctable, sizeof(mfctable));
bzero(bw_meter_timers, sizeof(bw_meter_timers));
/*
* Reset de-encapsulation cache
*/
last_encap_src = INADDR_ANY;
last_encap_vif = NULL;
#ifdef PIM
reg_vif_num = VIFI_INVALID;
#endif
if (encap_cookie) {
encap_detach(encap_cookie);
encap_cookie = NULL;
}
ip_mrouter = NULL;
splx(s);
if (mrtdebug)
log(LOG_DEBUG, "ip_mrouter_done\n");
return 0;
}
/*
* Set PIM assert processing global
*/
static int
set_assert(int i)
{
if ((i != 1) && (i != 0))
return EINVAL;
pim_assert = i;
return 0;
}
/*
* Configure API capabilities
*/
int
set_api_config(uint32_t *apival)
{
int i;
/*
* We can set the API capabilities only if it is the first operation
* after MRT_INIT. I.e.:
* - there are no vifs installed
* - pim_assert is not enabled
* - the MFC table is empty
*/
if (numvifs > 0) {
*apival = 0;
return EPERM;
}
if (pim_assert) {
*apival = 0;
return EPERM;
}
for (i = 0; i < MFCTBLSIZ; i++) {
if (mfctable[i] != NULL) {
*apival = 0;
return EPERM;
}
}
mrt_api_config = *apival & mrt_api_support;
*apival = mrt_api_config;
return 0;
}
/*
* Decide if a packet is from a tunnelled peer.
* Return 0 if not, 64 if so. XXX yuck.. 64 ???
*/
static int
mroute_encapcheck(const struct mbuf *m, int off, int proto, void *arg)
{
struct ip *ip = mtod(m, struct ip *);
int hlen = ip->ip_hl << 2;
/*
* don't claim the packet if it's not to a multicast destination or if
* we don't have an encapsulating tunnel with the source.
* Note: This code assumes that the remote site IP address
* uniquely identifies the tunnel (i.e., that this site has
* at most one tunnel with the remote site).
*/
if (!IN_MULTICAST(ntohl(((struct ip *)((char *)ip+hlen))->ip_dst.s_addr)))
return 0;
if (ip->ip_src.s_addr != last_encap_src) {
struct vif *vifp = viftable;
struct vif *vife = vifp + numvifs;
last_encap_src = ip->ip_src.s_addr;
last_encap_vif = NULL;
for ( ; vifp < vife; ++vifp)
if (vifp->v_rmt_addr.s_addr == ip->ip_src.s_addr) {
if ((vifp->v_flags & (VIFF_TUNNEL|VIFF_SRCRT)) == VIFF_TUNNEL)
last_encap_vif = vifp;
break;
}
}
if (last_encap_vif == NULL) {
last_encap_src = INADDR_ANY;
return 0;
}
return 64;
}
/*
* De-encapsulate a packet and feed it back through ip input (this
* routine is called whenever IP gets a packet that mroute_encap_func()
* claimed).
*/
static void
mroute_encap_input(struct mbuf *m, int off)
{
struct ip *ip = mtod(m, struct ip *);
int hlen = ip->ip_hl << 2;
if (hlen > sizeof(struct ip))
ip_stripoptions(m, (struct mbuf *) 0);
m->m_data += sizeof(struct ip);
m->m_len -= sizeof(struct ip);
m->m_pkthdr.len -= sizeof(struct ip);
m->m_pkthdr.rcvif = last_encap_vif->v_ifp;
netisr_queue(NETISR_IP, m);
/*
* normally we would need a "schednetisr(NETISR_IP)"
* here but we were called by ip_input and it is going
* to loop back & try to dequeue the packet we just
* queued as soon as we return so we avoid the
* unnecessary software interrrupt.
*
* XXX
* This no longer holds - we may have direct-dispatched the packet,
* or there may be a queue processing limit.
*/
}
extern struct domain inetdomain;
static struct protosw mroute_encap_protosw =
{ SOCK_RAW, &inetdomain, IPPROTO_IPV4, PR_ATOMIC|PR_ADDR,
mroute_encap_input, 0, 0, rip_ctloutput,
0,
0, 0, 0, 0,
&rip_usrreqs
};
/*
* Add a vif to the vif table
*/
static int
add_vif(struct vifctl *vifcp)
{
struct vif *vifp = viftable + vifcp->vifc_vifi;
struct sockaddr_in sin = {sizeof sin, AF_INET};
struct ifaddr *ifa;
struct ifnet *ifp;
int error, s;
struct tbf *v_tbf = tbftable + vifcp->vifc_vifi;
if (vifcp->vifc_vifi >= MAXVIFS)
return EINVAL;
if (vifp->v_lcl_addr.s_addr != INADDR_ANY)
return EADDRINUSE;
if (vifcp->vifc_lcl_addr.s_addr == INADDR_ANY)
return EADDRNOTAVAIL;
/* Find the interface with an address in AF_INET family */
#ifdef PIM
if (vifcp->vifc_flags & VIFF_REGISTER) {
/*
* XXX: Because VIFF_REGISTER does not really need a valid
* local interface (e.g. it could be 127.0.0.2), we don't
* check its address.
*/
ifp = NULL;
} else
#endif
{
sin.sin_addr = vifcp->vifc_lcl_addr;
ifa = ifa_ifwithaddr((struct sockaddr *)&sin);
if (ifa == NULL)
return EADDRNOTAVAIL;
ifp = ifa->ifa_ifp;
}
if (vifcp->vifc_flags & VIFF_TUNNEL) {
if ((vifcp->vifc_flags & VIFF_SRCRT) == 0) {
/*
* An encapsulating tunnel is wanted. Tell
* mroute_encap_input() to start paying attention
* to encapsulated packets.
*/
if (encap_cookie == NULL) {
encap_cookie = encap_attach_func(AF_INET, IPPROTO_IPV4,
mroute_encapcheck,
(struct protosw *)&mroute_encap_protosw, NULL);
if (encap_cookie == NULL) {
printf("ip_mroute: unable to attach encap\n");
return EIO; /* XXX */
}
for (s = 0; s < MAXVIFS; ++s) {
multicast_decap_if[s].if_name = "mdecap";
multicast_decap_if[s].if_unit = s;
}
}
/*
* Set interface to fake encapsulator interface
*/
ifp = &multicast_decap_if[vifcp->vifc_vifi];
/*
* Prepare cached route entry
*/
bzero(&vifp->v_route, sizeof(vifp->v_route));
} else {
log(LOG_ERR, "source routed tunnels not supported\n");
return EOPNOTSUPP;
}
#ifdef PIM
} else if (vifcp->vifc_flags & VIFF_REGISTER) {
ifp = &multicast_register_if;
if (mrtdebug)
log(LOG_DEBUG, "Adding a register vif, ifp: %p\n",
(void *)&multicast_register_if);
if (reg_vif_num == VIFI_INVALID) {
multicast_register_if.if_name = "register_vif";
multicast_register_if.if_unit = 0;
multicast_register_if.if_flags = IFF_LOOPBACK;
bzero(&vifp->v_route, sizeof(vifp->v_route));
reg_vif_num = vifcp->vifc_vifi;
}
#endif
} else { /* Make sure the interface supports multicast */
if ((ifp->if_flags & IFF_MULTICAST) == 0)
return EOPNOTSUPP;
/* Enable promiscuous reception of all IP multicasts from the if */
s = splnet();
error = if_allmulti(ifp, 1);
splx(s);
if (error)
return error;
}
s = splnet();
/* define parameters for the tbf structure */
vifp->v_tbf = v_tbf;
GET_TIME(vifp->v_tbf->tbf_last_pkt_t);
vifp->v_tbf->tbf_n_tok = 0;
vifp->v_tbf->tbf_q_len = 0;
vifp->v_tbf->tbf_max_q_len = MAXQSIZE;
vifp->v_tbf->tbf_q = vifp->v_tbf->tbf_t = NULL;
vifp->v_flags = vifcp->vifc_flags;
vifp->v_threshold = vifcp->vifc_threshold;
vifp->v_lcl_addr = vifcp->vifc_lcl_addr;
vifp->v_rmt_addr = vifcp->vifc_rmt_addr;
vifp->v_ifp = ifp;
/* scaling up here allows division by 1024 in critical code */
vifp->v_rate_limit= vifcp->vifc_rate_limit * 1024 / 1000;
vifp->v_rsvp_on = 0;
vifp->v_rsvpd = NULL;
/* initialize per vif pkt counters */
vifp->v_pkt_in = 0;
vifp->v_pkt_out = 0;
vifp->v_bytes_in = 0;
vifp->v_bytes_out = 0;
splx(s);
/* Adjust numvifs up if the vifi is higher than numvifs */
if (numvifs <= vifcp->vifc_vifi) numvifs = vifcp->vifc_vifi + 1;
if (mrtdebug)
log(LOG_DEBUG, "add_vif #%d, lcladdr %lx, %s %lx, thresh %x, rate %d\n",
vifcp->vifc_vifi,
(u_long)ntohl(vifcp->vifc_lcl_addr.s_addr),
(vifcp->vifc_flags & VIFF_TUNNEL) ? "rmtaddr" : "mask",
(u_long)ntohl(vifcp->vifc_rmt_addr.s_addr),
vifcp->vifc_threshold,
vifcp->vifc_rate_limit);
return 0;
}
/*
* Delete a vif from the vif table
*/
static int
del_vif(vifi_t vifi)
{
struct vif *vifp;
int s;
if (vifi >= numvifs)
return EINVAL;
vifp = &viftable[vifi];
if (vifp->v_lcl_addr.s_addr == INADDR_ANY)
return EADDRNOTAVAIL;
s = splnet();
if (!(vifp->v_flags & (VIFF_TUNNEL | VIFF_REGISTER)))
if_allmulti(vifp->v_ifp, 0);
if (vifp == last_encap_vif) {
last_encap_vif = NULL;
last_encap_src = INADDR_ANY;
}
/*
* Free packets queued at the interface
*/
while (vifp->v_tbf->tbf_q) {
struct mbuf *m = vifp->v_tbf->tbf_q;
vifp->v_tbf->tbf_q = m->m_act;
m_freem(m);
}
#ifdef PIM
if (vifp->v_flags & VIFF_REGISTER)
reg_vif_num = VIFI_INVALID;
#endif
bzero((caddr_t)vifp->v_tbf, sizeof(*(vifp->v_tbf)));
bzero((caddr_t)vifp, sizeof (*vifp));
if (mrtdebug)
log(LOG_DEBUG, "del_vif %d, numvifs %d\n", vifi, numvifs);
/* Adjust numvifs down */
for (vifi = numvifs; vifi > 0; vifi--)
if (viftable[vifi-1].v_lcl_addr.s_addr != INADDR_ANY)
break;
numvifs = vifi;
splx(s);
return 0;
}
/*
* update an mfc entry without resetting counters and S,G addresses.
*/
static void
update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
{
int i;
rt->mfc_parent = mfccp->mfcc_parent;
for (i = 0; i < numvifs; i++) {
rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
rt->mfc_flags[i] = mfccp->mfcc_flags[i] & mrt_api_config &
MRT_MFC_FLAGS_ALL;
}
/* set the RP address */
if (mrt_api_config & MRT_MFC_RP)
rt->mfc_rp = mfccp->mfcc_rp;
else
rt->mfc_rp.s_addr = INADDR_ANY;
}
/*
* fully initialize an mfc entry from the parameter.
*/
static void
init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
{
rt->mfc_origin = mfccp->mfcc_origin;
rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp;
update_mfc_params(rt, mfccp);
/* initialize pkt counters per src-grp */
rt->mfc_pkt_cnt = 0;
rt->mfc_byte_cnt = 0;
rt->mfc_wrong_if = 0;
rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0;
}
/*
* Add an mfc entry
*/
static int
add_mfc(struct mfcctl2 *mfccp)
{
struct mfc *rt;
u_long hash;
struct rtdetq *rte;
u_short nstl;
int s;
rt = mfc_find(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr);
/* If an entry already exists, just update the fields */
if (rt) {
if (mrtdebug & DEBUG_MFC)
log(LOG_DEBUG,"add_mfc update o %lx g %lx p %x\n",
(u_long)ntohl(mfccp->mfcc_origin.s_addr),
(u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
mfccp->mfcc_parent);
s = splnet();
update_mfc_params(rt, mfccp);
splx(s);
return 0;
}
/*
* Find the entry for which the upcall was made and update
*/
s = splnet();
hash = MFCHASH(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr);
for (rt = mfctable[hash], nstl = 0; rt; rt = rt->mfc_next) {
if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) &&
(rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr) &&
(rt->mfc_stall != NULL)) {
if (nstl++)
log(LOG_ERR, "add_mfc %s o %lx g %lx p %x dbx %p\n",
"multiple kernel entries",
(u_long)ntohl(mfccp->mfcc_origin.s_addr),
(u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
mfccp->mfcc_parent, (void *)rt->mfc_stall);
if (mrtdebug & DEBUG_MFC)
log(LOG_DEBUG,"add_mfc o %lx g %lx p %x dbg %p\n",
(u_long)ntohl(mfccp->mfcc_origin.s_addr),
(u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
mfccp->mfcc_parent, (void *)rt->mfc_stall);
init_mfc_params(rt, mfccp);
rt->mfc_expire = 0; /* Don't clean this guy up */
nexpire[hash]--;
/* free packets Qed at the end of this entry */
for (rte = rt->mfc_stall; rte != NULL; ) {
struct rtdetq *n = rte->next;
ip_mdq(rte->m, rte->ifp, rt, -1);
m_freem(rte->m);
free(rte, M_MRTABLE);
rte = n;
}
rt->mfc_stall = NULL;
}
}
/*
* It is possible that an entry is being inserted without an upcall
*/
if (nstl == 0) {
if (mrtdebug & DEBUG_MFC)
log(LOG_DEBUG,"add_mfc no upcall h %lu o %lx g %lx p %x\n",
hash, (u_long)ntohl(mfccp->mfcc_origin.s_addr),
(u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
mfccp->mfcc_parent);
for (rt = mfctable[hash]; rt != NULL; rt = rt->mfc_next) {
if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) &&
(rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr)) {
init_mfc_params(rt, mfccp);
if (rt->mfc_expire)
nexpire[hash]--;
rt->mfc_expire = 0;
break; /* XXX */
}
}
if (rt == NULL) { /* no upcall, so make a new entry */
rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
if (rt == NULL) {
splx(s);
return ENOBUFS;
}
init_mfc_params(rt, mfccp);
rt->mfc_expire = 0;
rt->mfc_stall = NULL;
rt->mfc_bw_meter = NULL;
/* insert new entry at head of hash chain */
rt->mfc_next = mfctable[hash];
mfctable[hash] = rt;
}
}
splx(s);
return 0;
}
/*
* Delete an mfc entry
*/
static int
del_mfc(struct mfcctl2 *mfccp)
{
struct in_addr origin;
struct in_addr mcastgrp;
struct mfc *rt;
struct mfc **nptr;
u_long hash;
int s;
struct bw_meter *list;
origin = mfccp->mfcc_origin;
mcastgrp = mfccp->mfcc_mcastgrp;
if (mrtdebug & DEBUG_MFC)
log(LOG_DEBUG,"del_mfc orig %lx mcastgrp %lx\n",
(u_long)ntohl(origin.s_addr), (u_long)ntohl(mcastgrp.s_addr));
s = splnet();
hash = MFCHASH(origin.s_addr, mcastgrp.s_addr);
for (nptr = &mfctable[hash]; (rt = *nptr) != NULL; nptr = &rt->mfc_next)
if (origin.s_addr == rt->mfc_origin.s_addr &&
mcastgrp.s_addr == rt->mfc_mcastgrp.s_addr &&
rt->mfc_stall == NULL)
break;
if (rt == NULL) {
splx(s);
return EADDRNOTAVAIL;
}
*nptr = rt->mfc_next;
/*
* free the bw_meter entries
*/
list = rt->mfc_bw_meter;
rt->mfc_bw_meter = NULL;
free(rt, M_MRTABLE);
splx(s);
free_bw_list(list);
return 0;
}
/*
* Send a message to mrouted on the multicast routing socket
*/
static int
socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in *src)
{
if (s) {
if (sbappendaddr(&s->so_rcv, (struct sockaddr *)src, mm, NULL) != 0) {
sorwakeup(s);
return 0;
}
}
m_freem(mm);
return -1;
}
/*
* IP multicast forwarding function. This function assumes that the packet
* pointed to by "ip" has arrived on (or is about to be sent to) the interface
* pointed to by "ifp", and the packet is to be relayed to other networks
* that have members of the packet's destination IP multicast group.
*
* The packet is returned unscathed to the caller, unless it is
* erroneous, in which case a non-zero return value tells the caller to
* discard it.
*/
#define TUNNEL_LEN 12 /* # bytes of IP option for tunnel encapsulation */
static int
X_ip_mforward(struct ip *ip, struct ifnet *ifp, struct mbuf *m,
struct ip_moptions *imo)
{
struct mfc *rt;
int s;
vifi_t vifi;
if (mrtdebug & DEBUG_FORWARD)
log(LOG_DEBUG, "ip_mforward: src %lx, dst %lx, ifp %p\n",
(u_long)ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr),
(void *)ifp);
if (ip->ip_hl < (sizeof(struct ip) + TUNNEL_LEN) >> 2 ||
((u_char *)(ip + 1))[1] != IPOPT_LSRR ) {
/*
* Packet arrived via a physical interface or
* an encapsulated tunnel or a register_vif.
*/
} else {
/*
* Packet arrived through a source-route tunnel.
* Source-route tunnels are no longer supported.
*/
static int last_log;
if (last_log != time_second) {
last_log = time_second;
log(LOG_ERR,
"ip_mforward: received source-routed packet from %lx\n",
(u_long)ntohl(ip->ip_src.s_addr));
}
return 1;
}
if (imo && ((vifi = imo->imo_multicast_vif) < numvifs)) {
if (ip->ip_ttl < 255)
ip->ip_ttl++; /* compensate for -1 in *_send routines */
if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) {
struct vif *vifp = viftable + vifi;
printf("Sending IPPROTO_RSVP from %lx to %lx on vif %d (%s%s%d)\n",
(long)ntohl(ip->ip_src.s_addr), (long)ntohl(ip->ip_dst.s_addr),
vifi,
(vifp->v_flags & VIFF_TUNNEL) ? "tunnel on " : "",
vifp->v_ifp->if_name, vifp->v_ifp->if_unit);
}
return ip_mdq(m, ifp, NULL, vifi);
}
if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) {
printf("Warning: IPPROTO_RSVP from %lx to %lx without vif option\n",
(long)ntohl(ip->ip_src.s_addr), (long)ntohl(ip->ip_dst.s_addr));
if (!imo)
printf("In fact, no options were specified at all\n");
}
/*
* Don't forward a packet with time-to-live of zero or one,
* or a packet destined to a local-only group.
*/
if (ip->ip_ttl <= 1 || ntohl(ip->ip_dst.s_addr) <= INADDR_MAX_LOCAL_GROUP)
return 0;
/*
* Determine forwarding vifs from the forwarding cache table
*/
s = splnet();
++mrtstat.mrts_mfc_lookups;
rt = mfc_find(ip->ip_src.s_addr, ip->ip_dst.s_addr);
/* Entry exists, so forward if necessary */
if (rt != NULL) {
splx(s);
return ip_mdq(m, ifp, rt, -1);
} else {
/*
* If we don't have a route for packet's origin,
* Make a copy of the packet & send message to routing daemon
*/
struct mbuf *mb0;
struct rtdetq *rte;
u_long hash;
int hlen = ip->ip_hl << 2;
++mrtstat.mrts_mfc_misses;
mrtstat.mrts_no_route++;
if (mrtdebug & (DEBUG_FORWARD | DEBUG_MFC))
log(LOG_DEBUG, "ip_mforward: no rte s %lx g %lx\n",
(u_long)ntohl(ip->ip_src.s_addr),
(u_long)ntohl(ip->ip_dst.s_addr));
/*
* Allocate mbufs early so that we don't do extra work if we are
* just going to fail anyway. Make sure to pullup the header so
* that other people can't step on it.
*/
rte = (struct rtdetq *)malloc((sizeof *rte), M_MRTABLE, M_NOWAIT);
if (rte == NULL) {
splx(s);
return ENOBUFS;
}
mb0 = m_copypacket(m, M_DONTWAIT);
if (mb0 && (M_HASCL(mb0) || mb0->m_len < hlen))
mb0 = m_pullup(mb0, hlen);
if (mb0 == NULL) {
free(rte, M_MRTABLE);
splx(s);
return ENOBUFS;
}
/* is there an upcall waiting for this flow ? */
hash = MFCHASH(ip->ip_src.s_addr, ip->ip_dst.s_addr);
for (rt = mfctable[hash]; rt; rt = rt->mfc_next) {
if ((ip->ip_src.s_addr == rt->mfc_origin.s_addr) &&
(ip->ip_dst.s_addr == rt->mfc_mcastgrp.s_addr) &&
(rt->mfc_stall != NULL))
break;
}
if (rt == NULL) {
int i;
struct igmpmsg *im;
struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
struct mbuf *mm;
/*
* Locate the vifi for the incoming interface for this packet.
* If none found, drop packet.
*/
for (vifi=0; vifi < numvifs && viftable[vifi].v_ifp != ifp; vifi++)
;
if (vifi >= numvifs) /* vif not found, drop packet */
goto non_fatal;
/* no upcall, so make a new entry */
rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
if (rt == NULL)
goto fail;
/* Make a copy of the header to send to the user level process */
mm = m_copy(mb0, 0, hlen);
if (mm == NULL)
goto fail1;
/*
* Send message to routing daemon to install
* a route into the kernel table
*/
im = mtod(mm, struct igmpmsg *);
im->im_msgtype = IGMPMSG_NOCACHE;
im->im_mbz = 0;
im->im_vif = vifi;
mrtstat.mrts_upcalls++;
k_igmpsrc.sin_addr = ip->ip_src;
if (socket_send(ip_mrouter, mm, &k_igmpsrc) < 0) {
log(LOG_WARNING, "ip_mforward: ip_mrouter socket queue full\n");
++mrtstat.mrts_upq_sockfull;
fail1:
free(rt, M_MRTABLE);
fail:
free(rte, M_MRTABLE);
m_freem(mb0);
splx(s);
return ENOBUFS;
}
/* insert new entry at head of hash chain */
rt->mfc_origin.s_addr = ip->ip_src.s_addr;
rt->mfc_mcastgrp.s_addr = ip->ip_dst.s_addr;
rt->mfc_expire = UPCALL_EXPIRE;
nexpire[hash]++;
for (i = 0; i < numvifs; i++) {
rt->mfc_ttls[i] = 0;
rt->mfc_flags[i] = 0;
}
rt->mfc_parent = -1;
rt->mfc_rp.s_addr = INADDR_ANY; /* clear the RP address */
rt->mfc_bw_meter = NULL;
/* link into table */
rt->mfc_next = mfctable[hash];
mfctable[hash] = rt;
rt->mfc_stall = rte;
} else {
/* determine if q has overflowed */
int npkts = 0;
struct rtdetq **p;
/*
* XXX ouch! we need to append to the list, but we
* only have a pointer to the front, so we have to
* scan the entire list every time.
*/
for (p = &rt->mfc_stall; *p != NULL; p = &(*p)->next)
npkts++;
if (npkts > MAX_UPQ) {
mrtstat.mrts_upq_ovflw++;
non_fatal:
free(rte, M_MRTABLE);
m_freem(mb0);
splx(s);
return 0;
}
/* Add this entry to the end of the queue */
*p = rte;
}
rte->m = mb0;
rte->ifp = ifp;
rte->next = NULL;
splx(s);
return 0;
}
}
/*
* Clean up the cache entry if upcall is not serviced
*/
static void
expire_upcalls(void *unused)
{
struct rtdetq *rte;
struct mfc *mfc, **nptr;
int i;
int s;
s = splnet();
for (i = 0; i < MFCTBLSIZ; i++) {
if (nexpire[i] == 0)
continue;
nptr = &mfctable[i];
for (mfc = *nptr; mfc != NULL; mfc = *nptr) {
/*
* Skip real cache entries
* Make sure it wasn't marked to not expire (shouldn't happen)
* If it expires now
*/
if (mfc->mfc_stall != NULL && mfc->mfc_expire != 0 &&
--mfc->mfc_expire == 0) {
if (mrtdebug & DEBUG_EXPIRE)
log(LOG_DEBUG, "expire_upcalls: expiring (%lx %lx)\n",
(u_long)ntohl(mfc->mfc_origin.s_addr),
(u_long)ntohl(mfc->mfc_mcastgrp.s_addr));
/*
* drop all the packets
* free the mbuf with the pkt, if, timing info
*/
for (rte = mfc->mfc_stall; rte; ) {
struct rtdetq *n = rte->next;
m_freem(rte->m);
free(rte, M_MRTABLE);
rte = n;
}
++mrtstat.mrts_cache_cleanups;
nexpire[i]--;
/*
* free the bw_meter entries
*/
while (mfc->mfc_bw_meter != NULL) {
struct bw_meter *x = mfc->mfc_bw_meter;
mfc->mfc_bw_meter = x->bm_mfc_next;
free(x, M_BWMETER);
}
*nptr = mfc->mfc_next;
free(mfc, M_MRTABLE);
} else {
nptr = &mfc->mfc_next;
}
}
}
splx(s);
expire_upcalls_ch = timeout(expire_upcalls, NULL, EXPIRE_TIMEOUT);
}
/*
* Packet forwarding routine once entry in the cache is made
*/
static int
ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif)
{
struct ip *ip = mtod(m, struct ip *);
vifi_t vifi;
int plen = ip->ip_len;
/*
* Macro to send packet on vif. Since RSVP packets don't get counted on
* input, they shouldn't get counted on output, so statistics keeping is
* separate.
*/
#define MC_SEND(ip,vifp,m) { \
if ((vifp)->v_flags & VIFF_TUNNEL) \
encap_send((ip), (vifp), (m)); \
else \
phyint_send((ip), (vifp), (m)); \
}
/*
* If xmt_vif is not -1, send on only the requested vif.
*
* (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.)
*/
if (xmt_vif < numvifs) {
#ifdef PIM
if (viftable[xmt_vif].v_flags & VIFF_REGISTER)
pim_register_send(ip, viftable + xmt_vif, m, rt);
else
#endif
MC_SEND(ip, viftable + xmt_vif, m);
return 1;
}
/*
* Don't forward if it didn't arrive from the parent vif for its origin.
*/
vifi = rt->mfc_parent;
if ((vifi >= numvifs) || (viftable[vifi].v_ifp != ifp)) {
/* came in the wrong interface */
if (mrtdebug & DEBUG_FORWARD)
log(LOG_DEBUG, "wrong if: ifp %p vifi %d vififp %p\n",
(void *)ifp, vifi, (void *)viftable[vifi].v_ifp);
++mrtstat.mrts_wrong_if;
++rt->mfc_wrong_if;
/*
* If we are doing PIM assert processing, send a message
* to the routing daemon.
*
* XXX: A PIM-SM router needs the WRONGVIF detection so it
* can complete the SPT switch, regardless of the type
* of the iif (broadcast media, GRE tunnel, etc).
*/
if (pim_assert && (vifi < numvifs) && viftable[vifi].v_ifp) {
struct timeval now;
u_long delta;
#ifdef PIM
if (ifp == &multicast_register_if)
pimstat.pims_rcv_registers_wrongiif++;
#endif
/* Get vifi for the incoming packet */
for (vifi=0; vifi < numvifs && viftable[vifi].v_ifp != ifp; vifi++)
;
if (vifi >= numvifs)
return 0; /* The iif is not found: ignore the packet. */
if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_DISABLE_WRONGVIF)
return 0; /* WRONGVIF disabled: ignore the packet */
GET_TIME(now);
TV_DELTA(rt->mfc_last_assert, now, delta);
if (delta > ASSERT_MSG_TIME) {
struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
struct igmpmsg *im;
int hlen = ip->ip_hl << 2;
struct mbuf *mm = m_copy(m, 0, hlen);
if (mm && (M_HASCL(mm) || mm->m_len < hlen))
mm = m_pullup(mm, hlen);
if (mm == NULL)
return ENOBUFS;
rt->mfc_last_assert = now;
im = mtod(mm, struct igmpmsg *);
im->im_msgtype = IGMPMSG_WRONGVIF;
im->im_mbz = 0;
im->im_vif = vifi;
mrtstat.mrts_upcalls++;
k_igmpsrc.sin_addr = im->im_src;
if (socket_send(ip_mrouter, mm, &k_igmpsrc) < 0) {
log(LOG_WARNING,
"ip_mforward: ip_mrouter socket queue full\n");
++mrtstat.mrts_upq_sockfull;
return ENOBUFS;
}
}
}
return 0;
}
/* If I sourced this packet, it counts as output, else it was input. */
if (ip->ip_src.s_addr == viftable[vifi].v_lcl_addr.s_addr) {
viftable[vifi].v_pkt_out++;
viftable[vifi].v_bytes_out += plen;
} else {
viftable[vifi].v_pkt_in++;
viftable[vifi].v_bytes_in += plen;
}
rt->mfc_pkt_cnt++;
rt->mfc_byte_cnt += plen;
/*
* For each vif, decide if a copy of the packet should be forwarded.
* Forward if:
* - the ttl exceeds the vif's threshold
* - there are group members downstream on interface
*/
for (vifi = 0; vifi < numvifs; vifi++)
if ((rt->mfc_ttls[vifi] > 0) && (ip->ip_ttl > rt->mfc_ttls[vifi])) {
viftable[vifi].v_pkt_out++;
viftable[vifi].v_bytes_out += plen;
#ifdef PIM
if (viftable[vifi].v_flags & VIFF_REGISTER)
pim_register_send(ip, viftable + vifi, m, rt);
else
#endif
MC_SEND(ip, viftable+vifi, m);
}
/*
* Perform upcall-related bw measuring.
*/
if (rt->mfc_bw_meter != NULL) {
struct bw_meter *x;
struct timeval now;
GET_TIME(now);
for (x = rt->mfc_bw_meter; x != NULL; x = x->bm_mfc_next)
bw_meter_receive_packet(x, plen, &now);
}
return 0;
}
/*
* check if a vif number is legal/ok. This is used by ip_output.
*/
static int
X_legal_vif_num(int vif)
{
return (vif >= 0 && vif < numvifs);
}
/*
* Return the local address used by this vif
*/
static u_long
X_ip_mcast_src(int vifi)
{
if (vifi >= 0 && vifi < numvifs)
return viftable[vifi].v_lcl_addr.s_addr;
else
return INADDR_ANY;
}
static void
phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
{
struct mbuf *mb_copy;
int hlen = ip->ip_hl << 2;
/*
* Make a new reference to the packet; make sure that
* the IP header is actually copied, not just referenced,
* so that ip_output() only scribbles on the copy.
*/
mb_copy = m_copypacket(m, M_DONTWAIT);
if (mb_copy && (M_HASCL(mb_copy) || mb_copy->m_len < hlen))
mb_copy = m_pullup(mb_copy, hlen);
if (mb_copy == NULL)
return;
if (vifp->v_rate_limit == 0)
tbf_send_packet(vifp, mb_copy);
else
tbf_control(vifp, mb_copy, mtod(mb_copy, struct ip *), ip->ip_len);
}
static void
encap_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
{
struct mbuf *mb_copy;
struct ip *ip_copy;
int i, len = ip->ip_len;
/* Take care of delayed checksums */
if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
in_delayed_cksum(m);
m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
}
/*
* copy the old packet & pullup its IP header into the
* new mbuf so we can modify it. Try to fill the new
* mbuf since if we don't the ethernet driver will.
*/
MGETHDR(mb_copy, M_DONTWAIT, MT_HEADER);
if (mb_copy == NULL)
return;
#ifdef MAC
mac_create_mbuf_multicast_encap(m, vifp->v_ifp, mb_copy);
#endif
mb_copy->m_data += max_linkhdr;
mb_copy->m_len = sizeof(multicast_encap_iphdr);
if ((mb_copy->m_next = m_copypacket(m, M_DONTWAIT)) == NULL) {
m_freem(mb_copy);
return;
}
i = MHLEN - M_LEADINGSPACE(mb_copy);
if (i > len)
i = len;
mb_copy = m_pullup(mb_copy, i);
if (mb_copy == NULL)
return;
mb_copy->m_pkthdr.len = len + sizeof(multicast_encap_iphdr);
/*
* fill in the encapsulating IP header.
*/
ip_copy = mtod(mb_copy, struct ip *);
*ip_copy = multicast_encap_iphdr;
#ifdef RANDOM_IP_ID
ip_copy->ip_id = ip_randomid();
#else
ip_copy->ip_id = htons(ip_id++);
#endif
ip_copy->ip_len += len;
ip_copy->ip_src = vifp->v_lcl_addr;
ip_copy->ip_dst = vifp->v_rmt_addr;
/*
* turn the encapsulated IP header back into a valid one.
*/
ip = (struct ip *)((caddr_t)ip_copy + sizeof(multicast_encap_iphdr));
--ip->ip_ttl;
ip->ip_len = htons(ip->ip_len);
ip->ip_off = htons(ip->ip_off);
ip->ip_sum = 0;
mb_copy->m_data += sizeof(multicast_encap_iphdr);
ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
mb_copy->m_data -= sizeof(multicast_encap_iphdr);
if (vifp->v_rate_limit == 0)
tbf_send_packet(vifp, mb_copy);
else
tbf_control(vifp, mb_copy, ip, ip_copy->ip_len);
}
/*
* Token bucket filter module
*/
static void
tbf_control(struct vif *vifp, struct mbuf *m, struct ip *ip, u_long p_len)
{
struct tbf *t = vifp->v_tbf;
if (p_len > MAX_BKT_SIZE) { /* drop if packet is too large */
mrtstat.mrts_pkt2large++;
m_freem(m);
return;
}
tbf_update_tokens(vifp);
if (t->tbf_q_len == 0) { /* queue empty... */
if (p_len <= t->tbf_n_tok) { /* send packet if enough tokens */
t->tbf_n_tok -= p_len;
tbf_send_packet(vifp, m);
} else { /* no, queue packet and try later */
tbf_queue(vifp, m);
timeout(tbf_reprocess_q, (caddr_t)vifp, TBF_REPROCESS);
}
} else if (t->tbf_q_len < t->tbf_max_q_len) {
/* finite queue length, so queue pkts and process queue */
tbf_queue(vifp, m);
tbf_process_q(vifp);
} else {
/* queue full, try to dq and queue and process */
if (!tbf_dq_sel(vifp, ip)) {
mrtstat.mrts_q_overflow++;
m_freem(m);
} else {
tbf_queue(vifp, m);
tbf_process_q(vifp);
}
}
}
/*
* adds a packet to the queue at the interface
*/
static void
tbf_queue(struct vif *vifp, struct mbuf *m)
{
int s = splnet();
struct tbf *t = vifp->v_tbf;
if (t->tbf_t == NULL) /* Queue was empty */
t->tbf_q = m;
else /* Insert at tail */
t->tbf_t->m_act = m;
t->tbf_t = m; /* Set new tail pointer */
#ifdef DIAGNOSTIC
/* Make sure we didn't get fed a bogus mbuf */
if (m->m_act)
panic("tbf_queue: m_act");
#endif
m->m_act = NULL;
t->tbf_q_len++;
splx(s);
}
/*
* processes the queue at the interface
*/
static void
tbf_process_q(struct vif *vifp)
{
int s = splnet();
struct tbf *t = vifp->v_tbf;
/* loop through the queue at the interface and send as many packets
* as possible
*/
while (t->tbf_q_len > 0) {
struct mbuf *m = t->tbf_q;
int len = mtod(m, struct ip *)->ip_len;
/* determine if the packet can be sent */
if (len > t->tbf_n_tok) /* not enough tokens, we are done */
break;
/* ok, reduce no of tokens, dequeue and send the packet. */
t->tbf_n_tok -= len;
t->tbf_q = m->m_act;
if (--t->tbf_q_len == 0)
t->tbf_t = NULL;
m->m_act = NULL;
tbf_send_packet(vifp, m);
}
splx(s);
}
static void
tbf_reprocess_q(void *xvifp)
{
struct vif *vifp = xvifp;
if (ip_mrouter == NULL)
return;
tbf_update_tokens(vifp);
tbf_process_q(vifp);
if (vifp->v_tbf->tbf_q_len)
timeout(tbf_reprocess_q, (caddr_t)vifp, TBF_REPROCESS);
}
/* function that will selectively discard a member of the queue
* based on the precedence value and the priority
*/
static int
tbf_dq_sel(struct vif *vifp, struct ip *ip)
{
int s = splnet();
u_int p;
struct mbuf *m, *last;
struct mbuf **np;
struct tbf *t = vifp->v_tbf;
p = priority(vifp, ip);
np = &t->tbf_q;
last = NULL;
while ((m = *np) != NULL) {
if (p > priority(vifp, mtod(m, struct ip *))) {
*np = m->m_act;
/* If we're removing the last packet, fix the tail pointer */
if (m == t->tbf_t)
t->tbf_t = last;
m_freem(m);
/* It's impossible for the queue to be empty, but check anyways. */
if (--t->tbf_q_len == 0)
t->tbf_t = NULL;
splx(s);
mrtstat.mrts_drop_sel++;
return 1;
}
np = &m->m_act;
last = m;
}
splx(s);
return 0;
}
static void
tbf_send_packet(struct vif *vifp, struct mbuf *m)
{
int s = splnet();
if (vifp->v_flags & VIFF_TUNNEL) /* If tunnel options */
ip_output(m, NULL, &vifp->v_route, IP_FORWARDING, NULL, NULL);
else {
struct ip_moptions imo;
int error;
static struct route ro; /* XXX check this */
imo.imo_multicast_ifp = vifp->v_ifp;
imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1;
imo.imo_multicast_loop = 1;
imo.imo_multicast_vif = -1;
/*
* Re-entrancy should not be a problem here, because
* the packets that we send out and are looped back at us
* should get rejected because they appear to come from
* the loopback interface, thus preventing looping.
*/
error = ip_output(m, NULL, &ro, IP_FORWARDING, &imo, NULL);
if (mrtdebug & DEBUG_XMIT)
log(LOG_DEBUG, "phyint_send on vif %d err %d\n",
(int)(vifp - viftable), error);
}
splx(s);
}
/* determine the current time and then
* the elapsed time (between the last time and time now)
* in milliseconds & update the no. of tokens in the bucket
*/
static void
tbf_update_tokens(struct vif *vifp)
{
struct timeval tp;
u_long tm;
int s = splnet();
struct tbf *t = vifp->v_tbf;
GET_TIME(tp);
TV_DELTA(tp, t->tbf_last_pkt_t, tm);
/*
* This formula is actually
* "time in seconds" * "bytes/second".
*
* (tm / 1000000) * (v_rate_limit * 1000 * (1000/1024) / 8)
*
* The (1000/1024) was introduced in add_vif to optimize
* this divide into a shift.
*/
t->tbf_n_tok += tm * vifp->v_rate_limit / 1024 / 8;
t->tbf_last_pkt_t = tp;
if (t->tbf_n_tok > MAX_BKT_SIZE)
t->tbf_n_tok = MAX_BKT_SIZE;
splx(s);
}
static int
priority(struct vif *vifp, struct ip *ip)
{
int prio = 50; /* the lowest priority -- default case */
/* temporary hack; may add general packet classifier some day */
/*
* The UDP port space is divided up into four priority ranges:
* [0, 16384) : unclassified - lowest priority
* [16384, 32768) : audio - highest priority
* [32768, 49152) : whiteboard - medium priority
* [49152, 65536) : video - low priority
*
* Everything else gets lowest priority.
*/
if (ip->ip_p == IPPROTO_UDP) {
struct udphdr *udp = (struct udphdr *)(((char *)ip) + (ip->ip_hl << 2));
switch (ntohs(udp->uh_dport) & 0xc000) {
case 0x4000:
prio = 70;
break;
case 0x8000:
prio = 60;
break;
case 0xc000:
prio = 55;
break;
}
}
return prio;
}
/*
* End of token bucket filter modifications
*/
static int
X_ip_rsvp_vif(struct socket *so, struct sockopt *sopt)
{
int error, vifi, s;
if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP)
return EOPNOTSUPP;
error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi);
if (error)
return error;
s = splnet();
if (vifi < 0 || vifi >= numvifs) { /* Error if vif is invalid */
splx(s);
return EADDRNOTAVAIL;
}
if (sopt->sopt_name == IP_RSVP_VIF_ON) {
/* Check if socket is available. */
if (viftable[vifi].v_rsvpd != NULL) {
splx(s);
return EADDRINUSE;
}
viftable[vifi].v_rsvpd = so;
/* This may seem silly, but we need to be sure we don't over-increment
* the RSVP counter, in case something slips up.
*/
if (!viftable[vifi].v_rsvp_on) {
viftable[vifi].v_rsvp_on = 1;
rsvp_on++;
}
} else { /* must be VIF_OFF */
/*
* XXX as an additional consistency check, one could make sure
* that viftable[vifi].v_rsvpd == so, otherwise passing so as
* first parameter is pretty useless.
*/
viftable[vifi].v_rsvpd = NULL;
/*
* This may seem silly, but we need to be sure we don't over-decrement
* the RSVP counter, in case something slips up.
*/
if (viftable[vifi].v_rsvp_on) {
viftable[vifi].v_rsvp_on = 0;
rsvp_on--;
}
}
splx(s);
return 0;
}
static void
X_ip_rsvp_force_done(struct socket *so)
{
int vifi;
int s;
/* Don't bother if it is not the right type of socket. */
if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP)
return;
s = splnet();
/* The socket may be attached to more than one vif...this
* is perfectly legal.
*/
for (vifi = 0; vifi < numvifs; vifi++) {
if (viftable[vifi].v_rsvpd == so) {
viftable[vifi].v_rsvpd = NULL;
/* This may seem silly, but we need to be sure we don't
* over-decrement the RSVP counter, in case something slips up.
*/
if (viftable[vifi].v_rsvp_on) {
viftable[vifi].v_rsvp_on = 0;
rsvp_on--;
}
}
}
splx(s);
}
static void
X_rsvp_input(struct mbuf *m, int off)
{
int vifi;
struct ip *ip = mtod(m, struct ip *);
struct sockaddr_in rsvp_src = { sizeof rsvp_src, AF_INET };
int s;
struct ifnet *ifp;
if (rsvpdebug)
printf("rsvp_input: rsvp_on %d\n",rsvp_on);
/* Can still get packets with rsvp_on = 0 if there is a local member
* of the group to which the RSVP packet is addressed. But in this
* case we want to throw the packet away.
*/
if (!rsvp_on) {
m_freem(m);
return;
}
s = splnet();
if (rsvpdebug)
printf("rsvp_input: check vifs\n");
#ifdef DIAGNOSTIC
M_ASSERTPKTHDR(m);
#endif
ifp = m->m_pkthdr.rcvif;
/* Find which vif the packet arrived on. */
for (vifi = 0; vifi < numvifs; vifi++)
if (viftable[vifi].v_ifp == ifp)
break;
if (vifi == numvifs || viftable[vifi].v_rsvpd == NULL) {
/*
* If the old-style non-vif-associated socket is set,
* then use it. Otherwise, drop packet since there
* is no specific socket for this vif.
*/
if (ip_rsvpd != NULL) {
if (rsvpdebug)
printf("rsvp_input: Sending packet up old-style socket\n");
rip_input(m, off); /* xxx */
} else {
if (rsvpdebug && vifi == numvifs)
printf("rsvp_input: Can't find vif for packet.\n");
else if (rsvpdebug && viftable[vifi].v_rsvpd == NULL)
printf("rsvp_input: No socket defined for vif %d\n",vifi);
m_freem(m);
}
splx(s);
return;
}
rsvp_src.sin_addr = ip->ip_src;
if (rsvpdebug && m)
printf("rsvp_input: m->m_len = %d, sbspace() = %ld\n",
m->m_len,sbspace(&(viftable[vifi].v_rsvpd->so_rcv)));
if (socket_send(viftable[vifi].v_rsvpd, m, &rsvp_src) < 0) {
if (rsvpdebug)
printf("rsvp_input: Failed to append to socket\n");
} else {
if (rsvpdebug)
printf("rsvp_input: send packet up\n");
}
splx(s);
}
/*
* Code for bandwidth monitors
*/
/*
* Define common interface for timeval-related methods
*/
#define BW_TIMEVALCMP(tvp, uvp, cmp) timevalcmp((tvp), (uvp), cmp)
#define BW_TIMEVALDECR(vvp, uvp) timevalsub((vvp), (uvp))
#define BW_TIMEVALADD(vvp, uvp) timevaladd((vvp), (uvp))
static uint32_t
compute_bw_meter_flags(struct bw_upcall *req)
{
uint32_t flags = 0;
if (req->bu_flags & BW_UPCALL_UNIT_PACKETS)
flags |= BW_METER_UNIT_PACKETS;
if (req->bu_flags & BW_UPCALL_UNIT_BYTES)
flags |= BW_METER_UNIT_BYTES;
if (req->bu_flags & BW_UPCALL_GEQ)
flags |= BW_METER_GEQ;
if (req->bu_flags & BW_UPCALL_LEQ)
flags |= BW_METER_LEQ;
return flags;
}
/*
* Add a bw_meter entry
*/
static int
add_bw_upcall(struct bw_upcall *req)
{
struct mfc *mfc;
struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC,
BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC };
struct timeval now;
struct bw_meter *x;
uint32_t flags;
int s;
if (!(mrt_api_config & MRT_MFC_BW_UPCALL))
return EOPNOTSUPP;
/* Test if the flags are valid */
if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES)))
return EINVAL;
if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)))
return EINVAL;
if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
== (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
return EINVAL;
/* Test if the threshold time interval is valid */
if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <))
return EINVAL;
flags = compute_bw_meter_flags(req);
/*
* Find if we have already same bw_meter entry
*/
s = splnet();
mfc = mfc_find(req->bu_src.s_addr, req->bu_dst.s_addr);
if (mfc == NULL) {
splx(s);
return EADDRNOTAVAIL;
}
for (x = mfc->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) {
if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
&req->bu_threshold.b_time, ==)) &&
(x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
(x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
(x->bm_flags & BW_METER_USER_FLAGS) == flags) {
splx(s);
return 0; /* XXX Already installed */
}
}
splx(s);
/* Allocate the new bw_meter entry */
x = (struct bw_meter *)malloc(sizeof(*x), M_BWMETER, M_NOWAIT);
if (x == NULL)
return ENOBUFS;
/* Set the new bw_meter entry */
x->bm_threshold.b_time = req->bu_threshold.b_time;
GET_TIME(now);
x->bm_start_time = now;
x->bm_threshold.b_packets = req->bu_threshold.b_packets;
x->bm_threshold.b_bytes = req->bu_threshold.b_bytes;
x->bm_measured.b_packets = 0;
x->bm_measured.b_bytes = 0;
x->bm_flags = flags;
x->bm_time_next = NULL;
x->bm_time_hash = BW_METER_BUCKETS;
/* Add the new bw_meter entry to the front of entries for this MFC */
s = splnet();
x->bm_mfc = mfc;
x->bm_mfc_next = mfc->mfc_bw_meter;
mfc->mfc_bw_meter = x;
schedule_bw_meter(x, &now);
splx(s);
return 0;
}
static void
free_bw_list(struct bw_meter *list)
{
while (list != NULL) {
struct bw_meter *x = list;
list = list->bm_mfc_next;
unschedule_bw_meter(x);
free(x, M_BWMETER);
}
}
/*
* Delete one or multiple bw_meter entries
*/
static int
del_bw_upcall(struct bw_upcall *req)
{
struct mfc *mfc;
struct bw_meter *x;
int s;
if (!(mrt_api_config & MRT_MFC_BW_UPCALL))
return EOPNOTSUPP;
s = splnet();
/* Find the corresponding MFC entry */
mfc = mfc_find(req->bu_src.s_addr, req->bu_dst.s_addr);
if (mfc == NULL) {
splx(s);
return EADDRNOTAVAIL;
} else if (req->bu_flags & BW_UPCALL_DELETE_ALL) {
/*
* Delete all bw_meter entries for this mfc
*/
struct bw_meter *list;
list = mfc->mfc_bw_meter;
mfc->mfc_bw_meter = NULL;
splx(s);
free_bw_list(list);
return 0;
} else { /* Delete a single bw_meter entry */
struct bw_meter *prev;
uint32_t flags = 0;
flags = compute_bw_meter_flags(req);
/* Find the bw_meter entry to delete */
for (prev = NULL, x = mfc->mfc_bw_meter; x != NULL;
x = x->bm_mfc_next) {
if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
&req->bu_threshold.b_time, ==)) &&
(x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
(x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
(x->bm_flags & BW_METER_USER_FLAGS) == flags)
break;
}
if (x != NULL) { /* Delete entry from the list for this MFC */
if (prev != NULL)
prev->bm_mfc_next = x->bm_mfc_next; /* remove from middle*/
else
x->bm_mfc->mfc_bw_meter = x->bm_mfc_next;/* new head of list */
splx(s);
unschedule_bw_meter(x);
/* Free the bw_meter entry */
free(x, M_BWMETER);
return 0;
} else {
splx(s);
return EINVAL;
}
}
/* NOTREACHED */
}
/*
* Perform bandwidth measurement processing that may result in an upcall
*/
static void
bw_meter_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp)
{
struct timeval delta;
int s;
s = splnet();
delta = *nowp;
BW_TIMEVALDECR(&delta, &x->bm_start_time);
if (x->bm_flags & BW_METER_GEQ) {
/*
* Processing for ">=" type of bw_meter entry
*/
if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
/* Reset the bw_meter entry */
x->bm_start_time = *nowp;
x->bm_measured.b_packets = 0;
x->bm_measured.b_bytes = 0;
x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
}
/* Record that a packet is received */
x->bm_measured.b_packets++;
x->bm_measured.b_bytes += plen;
/*
* Test if we should deliver an upcall
*/
if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) {
if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
(x->bm_measured.b_packets >= x->bm_threshold.b_packets)) ||
((x->bm_flags & BW_METER_UNIT_BYTES) &&
(x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) {
/* Prepare an upcall for delivery */
bw_meter_prepare_upcall(x, nowp);
x->bm_flags |= BW_METER_UPCALL_DELIVERED;
}
}
} else if (x->bm_flags & BW_METER_LEQ) {
/*
* Processing for "<=" type of bw_meter entry
*/
if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
/*
* We are behind time with the multicast forwarding table
* scanning for "<=" type of bw_meter entries, so test now
* if we should deliver an upcall.
*/
if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
(x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
((x->bm_flags & BW_METER_UNIT_BYTES) &&
(x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
/* Prepare an upcall for delivery */
bw_meter_prepare_upcall(x, nowp);
}
/* Reschedule the bw_meter entry */
unschedule_bw_meter(x);
schedule_bw_meter(x, nowp);
}
/* Record that a packet is received */
x->bm_measured.b_packets++;
x->bm_measured.b_bytes += plen;
/*
* Test if we should restart the measuring interval
*/
if ((x->bm_flags & BW_METER_UNIT_PACKETS &&
x->bm_measured.b_packets <= x->bm_threshold.b_packets) ||
(x->bm_flags & BW_METER_UNIT_BYTES &&
x->bm_measured.b_bytes <= x->bm_threshold.b_bytes)) {
/* Don't restart the measuring interval */
} else {
/* Do restart the measuring interval */
/*
* XXX: note that we don't unschedule and schedule, because this
* might be too much overhead per packet. Instead, when we process
* all entries for a given timer hash bin, we check whether it is
* really a timeout. If not, we reschedule at that time.
*/
x->bm_start_time = *nowp;
x->bm_measured.b_packets = 0;
x->bm_measured.b_bytes = 0;
x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
}
}
splx(s);
}
/*
* Prepare a bandwidth-related upcall
*/
static void
bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp)
{
struct timeval delta;
struct bw_upcall *u;
int s;
s = splnet();
/*
* Compute the measured time interval
*/
delta = *nowp;
BW_TIMEVALDECR(&delta, &x->bm_start_time);
/*
* If there are too many pending upcalls, deliver them now
*/
if (bw_upcalls_n >= BW_UPCALLS_MAX)
bw_upcalls_send();
/*
* Set the bw_upcall entry
*/
u = &bw_upcalls[bw_upcalls_n++];
u->bu_src = x->bm_mfc->mfc_origin;
u->bu_dst = x->bm_mfc->mfc_mcastgrp;
u->bu_threshold.b_time = x->bm_threshold.b_time;
u->bu_threshold.b_packets = x->bm_threshold.b_packets;
u->bu_threshold.b_bytes = x->bm_threshold.b_bytes;
u->bu_measured.b_time = delta;
u->bu_measured.b_packets = x->bm_measured.b_packets;
u->bu_measured.b_bytes = x->bm_measured.b_bytes;
u->bu_flags = 0;
if (x->bm_flags & BW_METER_UNIT_PACKETS)
u->bu_flags |= BW_UPCALL_UNIT_PACKETS;
if (x->bm_flags & BW_METER_UNIT_BYTES)
u->bu_flags |= BW_UPCALL_UNIT_BYTES;
if (x->bm_flags & BW_METER_GEQ)
u->bu_flags |= BW_UPCALL_GEQ;
if (x->bm_flags & BW_METER_LEQ)
u->bu_flags |= BW_UPCALL_LEQ;
splx(s);
}
/*
* Send the pending bandwidth-related upcalls
*/
static void
bw_upcalls_send(void)
{
struct mbuf *m;
int len = bw_upcalls_n * sizeof(bw_upcalls[0]);
struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
static struct igmpmsg igmpmsg = { 0, /* unused1 */
0, /* unused2 */
IGMPMSG_BW_UPCALL,/* im_msgtype */
0, /* im_mbz */
0, /* im_vif */
0, /* unused3 */
{ 0 }, /* im_src */
{ 0 } }; /* im_dst */
if (bw_upcalls_n == 0)
return; /* No pending upcalls */
bw_upcalls_n = 0;
/*
* Allocate a new mbuf, initialize it with the header and
* the payload for the pending calls.
*/
MGETHDR(m, M_DONTWAIT, MT_HEADER);
if (m == NULL) {
log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n");
return;
}
m->m_len = m->m_pkthdr.len = 0;
m_copyback(m, 0, sizeof(struct igmpmsg), (caddr_t)&igmpmsg);
m_copyback(m, sizeof(struct igmpmsg), len, (caddr_t)&bw_upcalls[0]);
/*
* Send the upcalls
* XXX do we need to set the address in k_igmpsrc ?
*/
mrtstat.mrts_upcalls++;
if (socket_send(ip_mrouter, m, &k_igmpsrc) < 0) {
log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n");
++mrtstat.mrts_upq_sockfull;
}
}
/*
* Compute the timeout hash value for the bw_meter entries
*/
#define BW_METER_TIMEHASH(bw_meter, hash) \
do { \
struct timeval next_timeval = (bw_meter)->bm_start_time; \
\
BW_TIMEVALADD(&next_timeval, &(bw_meter)->bm_threshold.b_time); \
(hash) = next_timeval.tv_sec; \
if (next_timeval.tv_usec) \
(hash)++; /* XXX: make sure we don't timeout early */ \
(hash) %= BW_METER_BUCKETS; \
} while (0)
/*
* Schedule a timer to process periodically bw_meter entry of type "<="
* by linking the entry in the proper hash bucket.
*/
static void
schedule_bw_meter(struct bw_meter *x, struct timeval *nowp)
{
int time_hash, s;
if (!(x->bm_flags & BW_METER_LEQ))
return; /* XXX: we schedule timers only for "<=" entries */
/*
* Reset the bw_meter entry
*/
s = splnet();
x->bm_start_time = *nowp;
x->bm_measured.b_packets = 0;
x->bm_measured.b_bytes = 0;
x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
splx(s);
/*
* Compute the timeout hash value and insert the entry
*/
BW_METER_TIMEHASH(x, time_hash);
x->bm_time_next = bw_meter_timers[time_hash];
bw_meter_timers[time_hash] = x;
x->bm_time_hash = time_hash;
}
/*
* Unschedule the periodic timer that processes bw_meter entry of type "<="
* by removing the entry from the proper hash bucket.
*/
static void
unschedule_bw_meter(struct bw_meter *x)
{
int time_hash;
struct bw_meter *prev, *tmp;
if (!(x->bm_flags & BW_METER_LEQ))
return; /* XXX: we schedule timers only for "<=" entries */
/*
* Compute the timeout hash value and delete the entry
*/
time_hash = x->bm_time_hash;
if (time_hash >= BW_METER_BUCKETS)
return; /* Entry was not scheduled */
for (prev = NULL, tmp = bw_meter_timers[time_hash];
tmp != NULL; prev = tmp, tmp = tmp->bm_time_next)
if (tmp == x)
break;
if (tmp == NULL)
panic("unschedule_bw_meter: bw_meter entry not found");
if (prev != NULL)
prev->bm_time_next = x->bm_time_next;
else
bw_meter_timers[time_hash] = x->bm_time_next;
x->bm_time_next = NULL;
x->bm_time_hash = BW_METER_BUCKETS;
}
/*
* Process all "<=" type of bw_meter that should be processed now,
* and for each entry prepare an upcall if necessary. Each processed
* entry is rescheduled again for the (periodic) processing.
*
* This is run periodically (once per second normally). On each round,
* all the potentially matching entries are in the hash slot that we are
* looking at.
*/
static void
bw_meter_process()
{
static uint32_t last_tv_sec; /* last time we processed this */
uint32_t loops;
int i, s;
struct timeval now, process_endtime;
GET_TIME(now);
if (last_tv_sec == now.tv_sec)
return; /* nothing to do */
s = splnet();
loops = now.tv_sec - last_tv_sec;
last_tv_sec = now.tv_sec;
if (loops > BW_METER_BUCKETS)
loops = BW_METER_BUCKETS;
/*
* Process all bins of bw_meter entries from the one after the last
* processed to the current one. On entry, i points to the last bucket
* visited, so we need to increment i at the beginning of the loop.
*/
for (i = (now.tv_sec - loops) % BW_METER_BUCKETS; loops > 0; loops--) {
struct bw_meter *x, *tmp_list;
if (++i >= BW_METER_BUCKETS)
i = 0;
/* Disconnect the list of bw_meter entries from the bin */
tmp_list = bw_meter_timers[i];
bw_meter_timers[i] = NULL;
/* Process the list of bw_meter entries */
while (tmp_list != NULL) {
x = tmp_list;
tmp_list = tmp_list->bm_time_next;
/* Test if the time interval is over */
process_endtime = x->bm_start_time;
BW_TIMEVALADD(&process_endtime, &x->bm_threshold.b_time);
if (BW_TIMEVALCMP(&process_endtime, &now, >)) {
/* Not yet: reschedule, but don't reset */
int time_hash;
BW_METER_TIMEHASH(x, time_hash);
if (time_hash == i && process_endtime.tv_sec == now.tv_sec) {
/*
* XXX: somehow the bin processing is a bit ahead of time.
* Put the entry in the next bin.
*/
if (++time_hash >= BW_METER_BUCKETS)
time_hash = 0;
}
x->bm_time_next = bw_meter_timers[time_hash];
bw_meter_timers[time_hash] = x;
x->bm_time_hash = time_hash;
continue;
}
/*
* Test if we should deliver an upcall
*/
if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
(x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
((x->bm_flags & BW_METER_UNIT_BYTES) &&
(x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
/* Prepare an upcall for delivery */
bw_meter_prepare_upcall(x, &now);
}
/*
* Reschedule for next processing
*/
schedule_bw_meter(x, &now);
}
}
splx(s);
/* Send all upcalls that are pending delivery */
bw_upcalls_send();
}
/*
* A periodic function for sending all upcalls that are pending delivery
*/
static void
expire_bw_upcalls_send(void *unused)
{
bw_upcalls_send();
bw_upcalls_ch = timeout(expire_bw_upcalls_send, NULL, BW_UPCALLS_PERIOD);
}
/*
* A periodic function for periodic scanning of the multicast forwarding
* table for processing all "<=" bw_meter entries.
*/
static void
expire_bw_meter_process(void *unused)
{
if (mrt_api_config & MRT_MFC_BW_UPCALL)
bw_meter_process();
bw_meter_ch = timeout(expire_bw_meter_process, NULL, BW_METER_PERIOD);
}
/*
* End of bandwidth monitoring code
*/
#ifdef PIM
/*
* Send the packet up to the user daemon, or eventually do kernel encapsulation
*
*/
static int
pim_register_send(struct ip *ip, struct vif *vifp,
struct mbuf *m, struct mfc *rt)
{
struct mbuf *mb_copy, *mm;
if (mrtdebug & DEBUG_PIM)
log(LOG_DEBUG, "pim_register_send: ");
mb_copy = pim_register_prepare(ip, m);
if (mb_copy == NULL)
return ENOBUFS;
/*
* Send all the fragments. Note that the mbuf for each fragment
* is freed by the sending machinery.
*/
for (mm = mb_copy; mm; mm = mb_copy) {
mb_copy = mm->m_nextpkt;
mm->m_nextpkt = 0;
mm = m_pullup(mm, sizeof(struct ip));
if (mm != NULL) {
ip = mtod(mm, struct ip *);
if ((mrt_api_config & MRT_MFC_RP) &&
(rt->mfc_rp.s_addr != INADDR_ANY)) {
pim_register_send_rp(ip, vifp, mm, rt);
} else {
pim_register_send_upcall(ip, vifp, mm, rt);
}
}
}
return 0;
}
/*
* Return a copy of the data packet that is ready for PIM Register
* encapsulation.
* XXX: Note that in the returned copy the IP header is a valid one.
*/
static struct mbuf *
pim_register_prepare(struct ip *ip, struct mbuf *m)
{
struct mbuf *mb_copy = NULL;
int mtu;
/* Take care of delayed checksums */
if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
in_delayed_cksum(m);
m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
}
/*
* Copy the old packet & pullup its IP header into the
* new mbuf so we can modify it.
*/
mb_copy = m_copypacket(m, M_DONTWAIT);
if (mb_copy == NULL)
return NULL;
mb_copy = m_pullup(mb_copy, ip->ip_hl << 2);
if (mb_copy == NULL)
return NULL;
/* take care of the TTL */
ip = mtod(mb_copy, struct ip *);
--ip->ip_ttl;
/* Compute the MTU after the PIM Register encapsulation */
mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr);
if (ip->ip_len <= mtu) {
/* Turn the IP header into a valid one */
ip->ip_len = htons(ip->ip_len);
ip->ip_off = htons(ip->ip_off);
ip->ip_sum = 0;
ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
} else {
/* Fragment the packet */
if (ip_fragment(ip, &mb_copy, mtu, 0, CSUM_DELAY_IP) != 0) {
m_freem(mb_copy);
return NULL;
}
}
return mb_copy;
}
/*
* Send an upcall with the data packet to the user-level process.
*/
static int
pim_register_send_upcall(struct ip *ip, struct vif *vifp,
struct mbuf *mb_copy, struct mfc *rt)
{
struct mbuf *mb_first;
int len = ntohs(ip->ip_len);
struct igmpmsg *im;
struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
/*
* Add a new mbuf with an upcall header
*/
MGETHDR(mb_first, M_DONTWAIT, MT_HEADER);
if (mb_first == NULL) {
m_freem(mb_copy);
return ENOBUFS;
}
mb_first->m_data += max_linkhdr;
mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg);
mb_first->m_len = sizeof(struct igmpmsg);
mb_first->m_next = mb_copy;
/* Send message to routing daemon */
im = mtod(mb_first, struct igmpmsg *);
im->im_msgtype = IGMPMSG_WHOLEPKT;
im->im_mbz = 0;
im->im_vif = vifp - viftable;
im->im_src = ip->ip_src;
im->im_dst = ip->ip_dst;
k_igmpsrc.sin_addr = ip->ip_src;
mrtstat.mrts_upcalls++;
if (socket_send(ip_mrouter, mb_first, &k_igmpsrc) < 0) {
if (mrtdebug & DEBUG_PIM)
log(LOG_WARNING,
"mcast: pim_register_send_upcall: ip_mrouter socket queue full");
++mrtstat.mrts_upq_sockfull;
return ENOBUFS;
}
/* Keep statistics */
pimstat.pims_snd_registers_msgs++;
pimstat.pims_snd_registers_bytes += len;
return 0;
}
/*
* Encapsulate the data packet in PIM Register message and send it to the RP.
*/
static int
pim_register_send_rp(struct ip *ip, struct vif *vifp,
struct mbuf *mb_copy, struct mfc *rt)
{
struct mbuf *mb_first;
struct ip *ip_outer;
struct pim_encap_pimhdr *pimhdr;
int len = ntohs(ip->ip_len);
vifi_t vifi = rt->mfc_parent;
if ((vifi >= numvifs) || (viftable[vifi].v_lcl_addr.s_addr == 0)) {
m_freem(mb_copy);
return EADDRNOTAVAIL; /* The iif vif is invalid */
}
/*
* Add a new mbuf with the encapsulating header
*/
MGETHDR(mb_first, M_DONTWAIT, MT_HEADER);
if (mb_first == NULL) {
m_freem(mb_copy);
return ENOBUFS;
}
mb_first->m_data += max_linkhdr;
mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
mb_first->m_next = mb_copy;
mb_first->m_pkthdr.len = len + mb_first->m_len;
/*
* Fill in the encapsulating IP and PIM header
*/
ip_outer = mtod(mb_first, struct ip *);
*ip_outer = pim_encap_iphdr;
#ifdef RANDOM_IP_ID
ip_outer->ip_id = ip_randomid();
#else
ip_outer->ip_id = htons(ip_id++);
#endif
ip_outer->ip_len = len + sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
ip_outer->ip_src = viftable[vifi].v_lcl_addr;
ip_outer->ip_dst = rt->mfc_rp;
/*
* Copy the inner header TOS to the outer header, and take care of the
* IP_DF bit.
*/
ip_outer->ip_tos = ip->ip_tos;
if (ntohs(ip->ip_off) & IP_DF)
ip_outer->ip_off |= IP_DF;
pimhdr = (struct pim_encap_pimhdr *)((caddr_t)ip_outer
+ sizeof(pim_encap_iphdr));
*pimhdr = pim_encap_pimhdr;
/* If the iif crosses a border, set the Border-bit */
if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & mrt_api_config)
pimhdr->flags |= htonl(PIM_BORDER_REGISTER);
mb_first->m_data += sizeof(pim_encap_iphdr);
pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr));
mb_first->m_data -= sizeof(pim_encap_iphdr);
if (vifp->v_rate_limit == 0)
tbf_send_packet(vifp, mb_first);
else
tbf_control(vifp, mb_first, ip, ip_outer->ip_len);
/* Keep statistics */
pimstat.pims_snd_registers_msgs++;
pimstat.pims_snd_registers_bytes += len;
return 0;
}
/*
* PIM-SMv2 and PIM-DM messages processing.
* Receives and verifies the PIM control messages, and passes them
* up to the listening socket, using rip_input().
* The only message with special processing is the PIM_REGISTER message
* (used by PIM-SM): the PIM header is stripped off, and the inner packet
* is passed to if_simloop().
*/
void
pim_input(struct mbuf *m, int off)
{
struct ip *ip = mtod(m, struct ip *);
struct pim *pim;
int minlen;
int datalen = ip->ip_len;
int ip_tos;
int iphlen = off;
/* Keep statistics */
pimstat.pims_rcv_total_msgs++;
pimstat.pims_rcv_total_bytes += datalen;
/*
* Validate lengths
*/
if (datalen < PIM_MINLEN) {
pimstat.pims_rcv_tooshort++;
log(LOG_ERR, "pim_input: packet size too small %d from %lx\n",
datalen, (u_long)ip->ip_src.s_addr);
m_freem(m);
return;
}
/*
* If the packet is at least as big as a REGISTER, go agead
* and grab the PIM REGISTER header size, to avoid another
* possible m_pullup() later.
*
* PIM_MINLEN == pimhdr + u_int32_t == 4 + 4 = 8
* PIM_REG_MINLEN == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28
*/
minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN);
/*
* Get the IP and PIM headers in contiguous memory, and
* possibly the PIM REGISTER header.
*/
if ((m->m_flags & M_EXT || m->m_len < minlen) &&
(m = m_pullup(m, minlen)) == 0) {
log(LOG_ERR, "pim_input: m_pullup failure\n");
return;
}
/* m_pullup() may have given us a new mbuf so reset ip. */
ip = mtod(m, struct ip *);
ip_tos = ip->ip_tos;
/* adjust mbuf to point to the PIM header */
m->m_data += iphlen;
m->m_len -= iphlen;
pim = mtod(m, struct pim *);
/*
* Validate checksum. If PIM REGISTER, exclude the data packet.
*
* XXX: some older PIMv2 implementations don't make this distinction,
* so for compatibility reason perform the checksum over part of the
* message, and if error, then over the whole message.
*/
if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) {
/* do nothing, checksum okay */
} else if (in_cksum(m, datalen)) {
pimstat.pims_rcv_badsum++;
if (mrtdebug & DEBUG_PIM)
log(LOG_DEBUG, "pim_input: invalid checksum");
m_freem(m);
return;
}
/* PIM version check */
if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) {
pimstat.pims_rcv_badversion++;
log(LOG_ERR, "pim_input: incorrect version %d, expecting %d\n",
PIM_VT_V(pim->pim_vt), PIM_VERSION);
m_freem(m);
return;
}
/* restore mbuf back to the outer IP */
m->m_data -= iphlen;
m->m_len += iphlen;
if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) {
/*
* Since this is a REGISTER, we'll make a copy of the register
* headers ip + pim + u_int32 + encap_ip, to be passed up to the
* routing daemon.
*/
struct sockaddr_in dst = { sizeof(dst), AF_INET };
struct mbuf *mcp;
struct ip *encap_ip;
u_int32_t *reghdr;
if ((reg_vif_num >= numvifs) || (reg_vif_num == VIFI_INVALID)) {
if (mrtdebug & DEBUG_PIM)
log(LOG_DEBUG,
"pim_input: register vif not set: %d\n", reg_vif_num);
m_freem(m);
return;
}
/*
* Validate length
*/
if (datalen < PIM_REG_MINLEN) {
pimstat.pims_rcv_tooshort++;
pimstat.pims_rcv_badregisters++;
log(LOG_ERR,
"pim_input: register packet size too small %d from %lx\n",
datalen, (u_long)ip->ip_src.s_addr);
m_freem(m);
return;
}
reghdr = (u_int32_t *)(pim + 1);
encap_ip = (struct ip *)(reghdr + 1);
if (mrtdebug & DEBUG_PIM) {
log(LOG_DEBUG,
"pim_input[register], encap_ip: %lx -> %lx, encap_ip len %d\n",
(u_long)ntohl(encap_ip->ip_src.s_addr),
(u_long)ntohl(encap_ip->ip_dst.s_addr),
ntohs(encap_ip->ip_len));
}
/* verify the version number of the inner packet */
if (encap_ip->ip_v != IPVERSION) {
pimstat.pims_rcv_badregisters++;
if (mrtdebug & DEBUG_PIM) {
log(LOG_DEBUG, "pim_input: invalid IP version (%d) "
"of the inner packet\n", encap_ip->ip_v);
}
m_freem(m);
return;
}
/* verify the inner packet is destined to a mcast group */
if (!IN_MULTICAST(ntohl(encap_ip->ip_dst.s_addr))) {
pimstat.pims_rcv_badregisters++;
if (mrtdebug & DEBUG_PIM)
log(LOG_DEBUG,
"pim_input: inner packet of register is not "
"multicast %lx\n",
(u_long)ntohl(encap_ip->ip_dst.s_addr));
m_freem(m);
return;
}
/*
* Copy the TOS from the outer IP header to the inner IP header.
*/
if (encap_ip->ip_tos != ip_tos) {
/* Outer TOS -> inner TOS */
encap_ip->ip_tos = ip_tos;
/* Recompute the inner header checksum. Sigh... */
/* adjust mbuf to point to the inner IP header */
m->m_data += (iphlen + PIM_MINLEN);
m->m_len -= (iphlen + PIM_MINLEN);
encap_ip->ip_sum = 0;
encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2);
/* restore mbuf to point back to the outer IP header */
m->m_data -= (iphlen + PIM_MINLEN);
m->m_len += (iphlen + PIM_MINLEN);
}
/* If a NULL_REGISTER, pass it to the daemon */
if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
goto pim_input_to_daemon;
/*
* Decapsulate the inner IP packet and loopback to forward it
* as a normal multicast packet. Also, make a copy of the
* outer_iphdr + pimhdr + reghdr + encap_iphdr
* to pass to the daemon later, so it can take the appropriate
* actions (e.g., send back PIM_REGISTER_STOP).
* XXX: here m->m_data points to the outer IP header.
*/
mcp = m_copy(m, 0, iphlen + PIM_REG_MINLEN);
if (mcp == NULL) {
log(LOG_ERR,
"pim_input: pim register: could not copy register head\n");
m_freem(m);
return;
}
/* Keep statistics */
/* XXX: registers_bytes include only the encap. mcast pkt */
pimstat.pims_rcv_registers_msgs++;
pimstat.pims_rcv_registers_bytes += ntohs(encap_ip->ip_len);
/*
* forward the inner ip packet; point m_data at the inner ip.
*/
m_adj(m, iphlen + PIM_MINLEN);
if (mrtdebug & DEBUG_PIM) {
log(LOG_DEBUG,
"pim_input: forwarding decapsulated register: "
"src %lx, dst %lx, vif %d\n",
(u_long)ntohl(encap_ip->ip_src.s_addr),
(u_long)ntohl(encap_ip->ip_dst.s_addr),
reg_vif_num);
}
if_simloop(viftable[reg_vif_num].v_ifp, m, dst.sin_family, 0);
/* prepare the register head to send to the mrouting daemon */
m = mcp;
}
pim_input_to_daemon:
/*
* Pass the PIM message up to the daemon; if it is a Register message,
* pass the 'head' only up to the daemon. This includes the
* outer IP header, PIM header, PIM-Register header and the
* inner IP header.
* XXX: the outer IP header pkt size of a Register is not adjust to
* reflect the fact that the inner multicast data is truncated.
*/
rip_input(m, iphlen);
return;
}
#endif /* PIM */
static int
ip_mroute_modevent(module_t mod, int type, void *unused)
{
int s;
switch (type) {
case MOD_LOAD:
s = splnet();
/* XXX Protect against multiple loading */
ip_mcast_src = X_ip_mcast_src;
ip_mforward = X_ip_mforward;
ip_mrouter_done = X_ip_mrouter_done;
ip_mrouter_get = X_ip_mrouter_get;
ip_mrouter_set = X_ip_mrouter_set;
ip_rsvp_force_done = X_ip_rsvp_force_done;
ip_rsvp_vif = X_ip_rsvp_vif;
legal_vif_num = X_legal_vif_num;
mrt_ioctl = X_mrt_ioctl;
rsvp_input_p = X_rsvp_input;
splx(s);
break;
case MOD_UNLOAD:
if (ip_mrouter)
return EINVAL;
s = splnet();
ip_mcast_src = NULL;
ip_mforward = NULL;
ip_mrouter_done = NULL;
ip_mrouter_get = NULL;
ip_mrouter_set = NULL;
ip_rsvp_force_done = NULL;
ip_rsvp_vif = NULL;
legal_vif_num = NULL;
mrt_ioctl = NULL;
rsvp_input_p = NULL;
splx(s);
break;
}
return 0;
}
static moduledata_t ip_mroutemod = {
"ip_mroute",
ip_mroute_modevent,
0
};
DECLARE_MODULE(ip_mroute, ip_mroutemod, SI_SUB_PSEUDO, SI_ORDER_ANY);
Reference in New Issue View Git Blame Copy Permalink