summaryrefslogtreecommitdiffstats
path: root/sys
diff options
context:
space:
mode:
authorjlemon <jlemon@FreeBSD.org>2001-11-22 04:50:44 +0000
committerjlemon <jlemon@FreeBSD.org>2001-11-22 04:50:44 +0000
commita3c1c9fdb4ec0da65e5e02c396bbd5bb22a16c8b (patch)
tree7e69286bc240807dcaf2ce9e68a4bb59848d76df /sys
parent7e4737a17a89fe0f71b6b62c33f5cb8b254ddbd8 (diff)
downloadFreeBSD-src-a3c1c9fdb4ec0da65e5e02c396bbd5bb22a16c8b.zip
FreeBSD-src-a3c1c9fdb4ec0da65e5e02c396bbd5bb22a16c8b.tar.gz
Introduce a syncache, which enables FreeBSD to withstand a SYN flood
DoS in an improved fashion over the existing code. Reviewed by: silby (in a previous iteration) Sponsored by: DARPA, NAI Labs
Diffstat (limited to 'sys')
-rw-r--r--sys/conf/files1
-rw-r--r--sys/net/route.h2
-rw-r--r--sys/netinet/in_pcb.c19
-rw-r--r--sys/netinet/in_pcb.h89
-rw-r--r--sys/netinet/tcp_input.c731
-rw-r--r--sys/netinet/tcp_output.c2
-rw-r--r--sys/netinet/tcp_reass.c731
-rw-r--r--sys/netinet/tcp_subr.c78
-rw-r--r--sys/netinet/tcp_syncache.c1161
-rw-r--r--sys/netinet/tcp_timewait.c78
-rw-r--r--sys/netinet/tcp_usrreq.c4
-rw-r--r--sys/netinet/tcp_var.h82
-rw-r--r--sys/netinet6/tcp6_var.h2
13 files changed, 1912 insertions, 1068 deletions
diff --git a/sys/conf/files b/sys/conf/files
index e63074c..abe4bdb 100644
--- a/sys/conf/files
+++ b/sys/conf/files
@@ -1091,6 +1091,7 @@ netinet/tcp_debug.c optional tcpdebug
netinet/tcp_input.c optional inet
netinet/tcp_output.c optional inet
netinet/tcp_subr.c optional inet
+netinet/tcp_syncache.c optional inet
netinet/tcp_timer.c optional inet
netinet/tcp_usrreq.c optional inet
netinet/udp_usrreq.c optional inet
diff --git a/sys/net/route.h b/sys/net/route.h
index 1a38ceb..d4de208 100644
--- a/sys/net/route.h
+++ b/sys/net/route.h
@@ -139,7 +139,7 @@ struct ortentry {
#define RTF_DYNAMIC 0x10 /* created dynamically (by redirect) */
#define RTF_MODIFIED 0x20 /* modified dynamically (by redirect) */
#define RTF_DONE 0x40 /* message confirmed */
-#define RTF_DELCLONE 0x80 /* delete cloned route */
+/* 0x80 unused, was RTF_DELCLONE */
#define RTF_CLONING 0x100 /* generate new routes on use */
#define RTF_XRESOLVE 0x200 /* external daemon resolves name */
#define RTF_LLINFO 0x400 /* generated by link layer (e.g. ARP) */
diff --git a/sys/netinet/in_pcb.c b/sys/netinet/in_pcb.c
index 4fecd6a..41987af 100644
--- a/sys/netinet/in_pcb.c
+++ b/sys/netinet/in_pcb.c
@@ -555,7 +555,6 @@ in_pcbdetach(inp)
{
struct socket *so = inp->inp_socket;
struct inpcbinfo *ipi = inp->inp_pcbinfo;
- struct rtentry *rt = inp->inp_route.ro_rt;
#ifdef IPSEC
ipsec4_delete_pcbpolicy(inp);
@@ -566,22 +565,8 @@ in_pcbdetach(inp)
sotryfree(so);
if (inp->inp_options)
(void)m_free(inp->inp_options);
- if (rt) {
- /*
- * route deletion requires reference count to be <= zero
- */
- if ((rt->rt_flags & RTF_DELCLONE) &&
- (rt->rt_flags & RTF_WASCLONED) &&
- (rt->rt_refcnt <= 1)) {
- rt->rt_refcnt--;
- rt->rt_flags &= ~RTF_UP;
- rtrequest(RTM_DELETE, rt_key(rt),
- rt->rt_gateway, rt_mask(rt),
- rt->rt_flags, (struct rtentry **)0);
- }
- else
- rtfree(rt);
- }
+ if (inp->inp_route.ro_rt)
+ rtfree(inp->inp_route.ro_rt);
ip_freemoptions(inp->inp_moptions);
inp->inp_vflag = 0;
zfree(ipi->ipi_zone, inp);
diff --git a/sys/netinet/in_pcb.h b/sys/netinet/in_pcb.h
index 8423462..eb1ac3d 100644
--- a/sys/netinet/in_pcb.h
+++ b/sys/netinet/in_pcb.h
@@ -39,7 +39,6 @@
#include <sys/queue.h>
-
#include <netinet6/ipsec.h> /* for IPSEC */
#define in6pcb inpcb /* for KAME src sync over BSD*'s */
@@ -67,6 +66,58 @@ struct in_addr_4in6 {
};
/*
+ * NOTE: ipv6 addrs should be 64-bit aligned, per RFC 2553.
+ * in_conninfo has some extra padding to accomplish this.
+ */
+struct in_endpoints {
+ u_int16_t ie_fport; /* foreign port */
+ u_int16_t ie_lport; /* local port */
+ /* protocol dependent part, local and foreign addr */
+ union {
+ /* foreign host table entry */
+ struct in_addr_4in6 ie46_foreign;
+ struct in6_addr ie6_foreign;
+ } ie_dependfaddr;
+ union {
+ /* local host table entry */
+ struct in_addr_4in6 ie46_local;
+ struct in6_addr ie6_local;
+ } ie_dependladdr;
+#define ie_faddr ie_dependfaddr.ie46_foreign.ia46_addr4
+#define ie_laddr ie_dependladdr.ie46_local.ia46_addr4
+#define ie6_faddr ie_dependfaddr.ie6_foreign
+#define ie6_laddr ie_dependladdr.ie6_local
+};
+
+/*
+ * XXX
+ * At some point struct route should possibly change to:
+ * struct rtentry *rt
+ * struct in_endpoints *ie;
+ */
+struct in_conninfo {
+ u_int8_t inc_flags;
+ u_int8_t inc_len;
+ u_int16_t inc_pad; /* XXX alignment for in_endpoints */
+ /* protocol dependent part; cached route */
+ struct in_endpoints inc_ie;
+ union {
+ /* placeholder for routing entry */
+ struct route inc4_route;
+ struct route_in6 inc6_route;
+ } inc_dependroute;
+};
+#define inc_isipv6 inc_flags /* temp compatability */
+#define inc_fport inc_ie.ie_fport
+#define inc_lport inc_ie.ie_lport
+#define inc_faddr inc_ie.ie_faddr
+#define inc_laddr inc_ie.ie_laddr
+#define inc_route inc_dependroute.inc4_route
+#define inc6_faddr inc_ie.ie6_faddr
+#define inc6_laddr inc_ie.ie6_laddr
+#define inc6_route inc_dependroute.inc6_route
+
+/*
* NB: the zone allocator is type-stable EXCEPT FOR THE FIRST TWO LONGS
* of the structure. Therefore, it is important that the members in
* that position not contain any information which is required to be
@@ -76,22 +127,11 @@ struct icmp6_filter;
struct inpcb {
LIST_ENTRY(inpcb) inp_hash; /* hash list */
- u_short inp_fport; /* foreign port */
- u_short inp_lport; /* local port */
LIST_ENTRY(inpcb) inp_list; /* list for all PCBs of this proto */
u_int32_t inp_flow;
- /* protocol dependent part, local and foreign addr */
- union {
- /* foreign host table entry */
- struct in_addr_4in6 inp46_foreign;
- struct in6_addr inp6_foreign;
- } inp_dependfaddr;
- union {
- /* local host table entry */
- struct in_addr_4in6 inp46_local;
- struct in6_addr inp6_local;
- } inp_dependladdr;
+ /* local and foreign ports, local and foreign addr */
+ struct in_conninfo inp_inc;
caddr_t inp_ppcb; /* pointer to per-protocol pcb */
struct inpcbinfo *inp_pcbinfo; /* PCB list info */
@@ -99,13 +139,6 @@ struct inpcb {
/* list for this PCB's local port */
int inp_flags; /* generic IP/datagram flags */
- /* protocol dependent part; cached route */
- union {
- /* placeholder for routing entry */
- struct route inp4_route;
- struct route_in6 inp6_route;
- } inp_dependroute;
-
struct inpcbpolicy *inp_sp; /* for IPSEC */
u_char inp_vflag;
#define INP_IPV4 0x1
@@ -119,9 +152,11 @@ struct inpcb {
struct mbuf *inp4_options; /* IP options */
struct ip_moptions *inp4_moptions; /* IP multicast options */
} inp_depend4;
-#define inp_faddr inp_dependfaddr.inp46_foreign.ia46_addr4
-#define inp_laddr inp_dependladdr.inp46_local.ia46_addr4
-#define inp_route inp_dependroute.inp4_route
+#define inp_fport inp_inc.inc_fport
+#define inp_lport inp_inc.inc_lport
+#define inp_faddr inp_inc.inc_faddr
+#define inp_laddr inp_inc.inc_laddr
+#define inp_route inp_inc.inc_route
#define inp_ip_tos inp_depend4.inp4_ip_tos
#define inp_options inp_depend4.inp4_options
#define inp_moptions inp_depend4.inp4_moptions
@@ -143,9 +178,9 @@ struct inpcb {
LIST_ENTRY(inpcb) inp_portlist;
struct inpcbport *inp_phd; /* head of this list */
inp_gen_t inp_gencnt; /* generation count of this instance */
-#define in6p_faddr inp_dependfaddr.inp6_foreign
-#define in6p_laddr inp_dependladdr.inp6_local
-#define in6p_route inp_dependroute.inp6_route
+#define in6p_faddr inp_inc.inc6_faddr
+#define in6p_laddr inp_inc.inc6_laddr
+#define in6p_route inp_inc.inc6_route
#define in6p_ip6_hlim inp_depend6.inp6_hlim
#define in6p_hops inp_depend6.inp6_hops /* default hop limit */
#define in6p_ip6_nxt inp_ip_p
diff --git a/sys/netinet/tcp_input.c b/sys/netinet/tcp_input.c
index 7ab594e..6740ebe 100644
--- a/sys/netinet/tcp_input.c
+++ b/sys/netinet/tcp_input.c
@@ -120,11 +120,6 @@ SYSCTL_INT(_net_inet_tcp, OID_AUTO, delayed_ack, CTLFLAG_RW,
&tcp_delack_enabled, 0,
"Delay ACK to try and piggyback it onto a data packet");
-int tcp_lq_overflow = 1;
-SYSCTL_INT(_net_inet_tcp, OID_AUTO, tcp_lq_overflow, CTLFLAG_RW,
- &tcp_lq_overflow, 0,
- "Listen Queue Overflow");
-
#ifdef TCP_DROP_SYNFIN
static int drop_synfin = 0;
SYSCTL_INT(_net_inet_tcp, OID_AUTO, drop_synfin, CTLFLAG_RW,
@@ -135,10 +130,9 @@ struct inpcbhead tcb;
#define tcb6 tcb /* for KAME src sync over BSD*'s */
struct inpcbinfo tcbinfo;
-static void tcp_dooptions __P((struct tcpcb *,
- u_char *, int, struct tcphdr *, struct tcpopt *));
+static void tcp_dooptions __P((struct tcpopt *, u_char *, int, int));
static void tcp_pulloutofband __P((struct socket *,
- struct tcphdr *, struct mbuf *, int));
+ struct tcphdr *, struct mbuf *, int));
static int tcp_reass __P((struct tcpcb *, struct tcphdr *, int *,
struct mbuf *));
static void tcp_xmit_timer __P((struct tcpcb *, int));
@@ -344,11 +338,6 @@ tcp_input(m, off0)
register int thflags;
struct socket *so = 0;
int todrop, acked, ourfinisacked, needoutput = 0;
- struct in_addr laddr;
-#ifdef INET6
- struct in6_addr laddr6;
-#endif
- int dropsocket = 0;
int iss = 0;
u_long tiwin;
struct tcpopt to; /* options in this segment */
@@ -643,6 +632,7 @@ findpcb:
so = inp->inp_socket;
if (so->so_options & (SO_DEBUG|SO_ACCEPTCONN)) {
+ struct in_conninfo inc;
#ifdef TCPDEBUG
if (so->so_options & SO_DEBUG) {
ostate = tp->t_state;
@@ -656,212 +646,232 @@ findpcb:
tcp_savetcp = *th;
}
#endif
- if (so->so_options & SO_ACCEPTCONN) {
- register struct tcpcb *tp0 = tp;
- struct socket *so2;
-#ifdef IPSEC
- struct socket *oso;
-#endif
+ /* skip if this isn't a listen socket */
+ if ((so->so_options & SO_ACCEPTCONN) == 0)
+ goto after_listen;
#ifdef INET6
- struct inpcb *oinp = sotoinpcb(so);
-#endif /* INET6 */
+ inc.inc_isipv6 = isipv6;
+ if (isipv6) {
+ inc.inc6_faddr = ip6->ip6_src;
+ inc.inc6_laddr = ip6->ip6_dst;
+ inc.inc6_route.ro_rt = NULL; /* XXX */
-#ifndef IPSEC
- /*
- * Current IPsec implementation makes incorrect IPsec
- * cache if this check is done here.
- * So delay this until duplicated socket is created.
- */
- if ((thflags & (TH_RST|TH_ACK|TH_SYN)) != TH_SYN) {
- /*
- * Note: dropwithreset makes sure we don't
- * send a RST in response to a RST.
- */
- if (thflags & TH_ACK) {
+ } else
+#endif /* INET6 */
+ {
+ inc.inc_faddr = ip->ip_src;
+ inc.inc_laddr = ip->ip_dst;
+ inc.inc_route.ro_rt = NULL; /* XXX */
+ }
+ inc.inc_fport = th->th_sport;
+ inc.inc_lport = th->th_dport;
+
+ /*
+ * If the state is LISTEN then ignore segment if it contains
+ * a RST. If the segment contains an ACK then it is bad and
+ * send a RST. If it does not contain a SYN then it is not
+ * interesting; drop it.
+ *
+ * If the state is SYN_RECEIVED (syncache) and seg contains
+ * an ACK, but not for our SYN/ACK, send a RST. If the seg
+ * contains a RST, check the sequence number to see if it
+ * is a valid reset segment.
+ */
+ if ((thflags & (TH_RST|TH_ACK|TH_SYN)) != TH_SYN) {
+ if ((thflags & (TH_RST|TH_ACK|TH_SYN)) == TH_ACK) {
+ if (!syncache_expand(&inc, th, &so, m)) {
+ /*
+ * No syncache entry, or ACK was not
+ * for our SYN/ACK. Send a RST.
+ */
tcpstat.tcps_badsyn++;
rstreason = BANDLIM_RST_OPENPORT;
goto dropwithreset;
}
+ if (so == NULL)
+ /*
+ * Could not complete 3-way handshake,
+ * connection is being closed down, and
+ * syncache will free mbuf.
+ */
+ return;
+ /*
+ * Socket is created in state SYN_RECEIVED.
+ * Continue processing segment.
+ */
+ inp = sotoinpcb(so);
+ tp = intotcpcb(inp);
+ /*
+ * This is what would have happened in
+ * tcp_ouput() when the SYN,ACK was sent.
+ */
+ tp->snd_up = tp->snd_una;
+ tp->snd_max = tp->snd_nxt = tp->iss + 1;
+ tp->last_ack_sent = tp->rcv_nxt;
+/*
+ * XXX possible bug - it doesn't appear that tp->snd_wnd is unscaled
+ * until the _second_ ACK is received:
+ * rcv SYN (set wscale opts) --> send SYN/ACK, set snd_wnd = window.
+ * rcv ACK, calculate tiwin --> process SYN_RECEIVED, determine wscale,
+ * move to ESTAB, set snd_wnd to tiwin.
+ */
+ tp->snd_wnd = tiwin; /* unscaled */
+ goto after_listen;
+ }
+ if (thflags & TH_RST) {
+ syncache_chkrst(&inc, th);
goto drop;
}
-#endif
+ if (thflags & TH_ACK) {
+ syncache_badack(&inc);
+ tcpstat.tcps_badsyn++;
+ rstreason = BANDLIM_RST_OPENPORT;
+ goto dropwithreset;
+ }
+ goto drop;
+ }
+ /*
+ * Segment's flags are (SYN) or (SYN|FIN).
+ */
#ifdef INET6
- /*
- * If deprecated address is forbidden,
- * we do not accept SYN to deprecated interface
- * address to prevent any new inbound connection from
- * getting established.
- * When we do not accept SYN, we send a TCP RST,
- * with deprecated source address (instead of dropping
- * it). We compromise it as it is much better for peer
- * to send a RST, and RST will be the final packet
- * for the exchange.
- *
- * If we do not forbid deprecated addresses, we accept
- * the SYN packet. RFC2462 does not suggest dropping
- * SYN in this case.
- * If we decipher RFC2462 5.5.4, it says like this:
- * 1. use of deprecated addr with existing
- * communication is okay - "SHOULD continue to be
- * used"
- * 2. use of it with new communication:
- * (2a) "SHOULD NOT be used if alternate address
- * with sufficient scope is available"
- * (2b) nothing mentioned otherwise.
- * Here we fall into (2b) case as we have no choice in
- * our source address selection - we must obey the peer.
- *
- * The wording in RFC2462 is confusing, and there are
- * multiple description text for deprecated address
- * handling - worse, they are not exactly the same.
- * I believe 5.5.4 is the best one, so we follow 5.5.4.
- */
- if (isipv6 && !ip6_use_deprecated) {
- struct in6_ifaddr *ia6;
-
- if ((ia6 = ip6_getdstifaddr(m)) &&
- (ia6->ia6_flags & IN6_IFF_DEPRECATED)) {
- tp = NULL;
- rstreason = BANDLIM_RST_OPENPORT;
- goto dropwithreset;
- }
- }
-#endif
+ /*
+ * If deprecated address is forbidden,
+ * we do not accept SYN to deprecated interface
+ * address to prevent any new inbound connection from
+ * getting established.
+ * When we do not accept SYN, we send a TCP RST,
+ * with deprecated source address (instead of dropping
+ * it). We compromise it as it is much better for peer
+ * to send a RST, and RST will be the final packet
+ * for the exchange.
+ *
+ * If we do not forbid deprecated addresses, we accept
+ * the SYN packet. RFC2462 does not suggest dropping
+ * SYN in this case.
+ * If we decipher RFC2462 5.5.4, it says like this:
+ * 1. use of deprecated addr with existing
+ * communication is okay - "SHOULD continue to be
+ * used"
+ * 2. use of it with new communication:
+ * (2a) "SHOULD NOT be used if alternate address
+ * with sufficient scope is available"
+ * (2b) nothing mentioned otherwise.
+ * Here we fall into (2b) case as we have no choice in
+ * our source address selection - we must obey the peer.
+ *
+ * The wording in RFC2462 is confusing, and there are
+ * multiple description text for deprecated address
+ * handling - worse, they are not exactly the same.
+ * I believe 5.5.4 is the best one, so we follow 5.5.4.
+ */
+ if (isipv6 && !ip6_use_deprecated) {
+ struct in6_ifaddr *ia6;
- so2 = sonewconn(so, 0);
- if (so2 == 0) {
- /*
- * If we were unable to create a new socket
- * for this SYN, we call sodropablereq to
- * see if there are any other sockets we
- * can kick out of the listen queue. If
- * so, we'll silently drop the socket
- * sodropablereq told us to drop and
- * create a new one.
- *
- * If sodropablereq returns 0, we'll
- * simply drop the incoming SYN, as we
- * can not allocate a socket for it.
- */
- tcpstat.tcps_listendrop++;
- so2 = sodropablereq(so);
- if (so2) {
- if (tcp_lq_overflow)
- sototcpcb(so2)->t_flags |=
- TF_LQ_OVERFLOW;
- tcp_close(sototcpcb(so2));
- so2 = sonewconn(so, 0);
- }
- if (!so2)
- goto drop;
+ if ((ia6 = ip6_getdstifaddr(m)) &&
+ (ia6->ia6_flags & IN6_IFF_DEPRECATED)) {
+ tp = NULL;
+ rstreason = BANDLIM_RST_OPENPORT;
+ goto dropwithreset;
}
-#ifdef IPSEC
- oso = so;
+ }
#endif
- so = so2;
- /*
- * This is ugly, but ....
- *
- * Mark socket as temporary until we're
- * committed to keeping it. The code at
- * ``drop'' and ``dropwithreset'' check the
- * flag dropsocket to see if the temporary
- * socket created here should be discarded.
- * We mark the socket as discardable until
- * we're committed to it below in TCPS_LISTEN.
- */
- dropsocket++;
- inp = (struct inpcb *)so->so_pcb;
+ /*
+ * If it is from this socket, drop it, it must be forged.
+ * Don't bother responding if the destination was a broadcast.
+ */
+ if (th->th_dport == th->th_sport) {
#ifdef INET6
- if (isipv6)
- inp->in6p_laddr = ip6->ip6_dst;
- else {
- inp->inp_vflag &= ~INP_IPV6;
- inp->inp_vflag |= INP_IPV4;
+ if (isipv6) {
+ if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
+ &ip6->ip6_src))
+ goto drop;
+ } else
#endif /* INET6 */
- inp->inp_laddr = ip->ip_dst;
+ if (ip->ip_dst.s_addr == ip->ip_src.s_addr)
+ goto drop;
+ }
+ /*
+ * RFC1122 4.2.3.10, p. 104: discard bcast/mcast SYN
+ * in_broadcast() should never return true on a received
+ * packet with M_BCAST not set.
+ *
+ * Packets with a multicast source address should also
+ * be discarded.
+ */
+ if (m->m_flags & (M_BCAST|M_MCAST))
+ goto drop;
#ifdef INET6
- }
-#endif /* INET6 */
- inp->inp_lport = th->th_dport;
- if (in_pcbinshash(inp) != 0) {
+ if (isipv6) {
+ if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
+ IN6_IS_ADDR_MULTICAST(&ip6->ip6_src))
+ goto drop;
+ } else
+#endif
+ if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) ||
+ IN_MULTICAST(ntohl(ip->ip_src.s_addr)) ||
+ ip->ip_src.s_addr == htonl(INADDR_BROADCAST))
+ goto drop;
+ /*
+ * SYN appears to be valid; create compressed TCP state
+ * for syncache, or perform t/tcp connection.
+ */
+ if (so->so_qlen <= so->so_qlimit) {
+ tcp_dooptions(&to, optp, optlen, 1);
+ if (!syncache_add(&inc, &to, th, &so, m))
+ goto drop;
+ if (so == NULL)
/*
- * Undo the assignments above if we failed to
- * put the PCB on the hash lists.
+ * Entry added to syncache, mbuf used to
+ * send SYN,ACK packet.
*/
-#ifdef INET6
- if (isipv6)
- inp->in6p_laddr = in6addr_any;
- else
-#endif /* INET6 */
- inp->inp_laddr.s_addr = INADDR_ANY;
- inp->inp_lport = 0;
- goto drop;
- }
-#ifdef IPSEC
+ return;
/*
- * To avoid creating incorrectly cached IPsec
- * association, this is need to be done here.
- *
- * Subject: (KAME-snap 748)
- * From: Wayne Knowles <w.knowles@niwa.cri.nz>
- * ftp://ftp.kame.net/pub/mail-list/snap-users/748
+ * Segment passed TAO tests.
*/
- if ((thflags & (TH_RST|TH_ACK|TH_SYN)) != TH_SYN) {
- /*
- * Note: dropwithreset makes sure we don't
- * send a RST in response to a RST.
- */
- if (thflags & TH_ACK) {
- tcpstat.tcps_badsyn++;
- rstreason = BANDLIM_RST_OPENPORT;
- goto dropwithreset;
- }
- goto drop;
- }
+ inp = sotoinpcb(so);
+ tp = intotcpcb(inp);
+ tp->snd_wnd = tiwin;
+ tp->t_starttime = ticks;
+ tp->t_state = TCPS_ESTABLISHED;
+
+ /*
+ * If there is a FIN, or if there is data and the
+ * connection is local, then delay SYN,ACK(SYN) in
+ * the hope of piggy-backing it on a response
+ * segment. Otherwise must send ACK now in case
+ * the other side is slow starting.
+ */
+ if (DELAY_ACK(tp) && ((thflags & TH_FIN) ||
+ (tlen != 0 &&
+#ifdef INET6
+ ((isipv6 && in6_localaddr(&inp->in6p_faddr))
+ ||
+ (!isipv6 &&
#endif
+ in_localaddr(inp->inp_faddr)
#ifdef INET6
- if (isipv6) {
- /*
- * Inherit socket options from the listening
- * socket.
- * Note that in6p_inputopts are not (even
- * should not be) copied, since it stores
- * previously received options and is used to
- * detect if each new option is different than
- * the previous one and hence should be passed
- * to a user.
- * If we copied in6p_inputopts, a user would
- * not be able to receive options just after
- * calling the accept system call.
- */
- inp->inp_flags |=
- oinp->inp_flags & INP_CONTROLOPTS;
- if (oinp->in6p_outputopts)
- inp->in6p_outputopts =
- ip6_copypktopts(oinp->in6p_outputopts,
- M_NOWAIT);
- } else
-#endif /* INET6 */
- inp->inp_options = ip_srcroute();
-#ifdef IPSEC
- /* copy old policy into new socket's */
- if (ipsec_copy_policy(sotoinpcb(oso)->inp_sp,
- inp->inp_sp))
- printf("tcp_input: could not copy policy\n");
+ ))
#endif
- tp = intotcpcb(inp);
- tp->t_state = TCPS_LISTEN;
- tp->t_flags |= tp0->t_flags & (TF_NOPUSH|TF_NOOPT);
-
- /* Compute proper scaling value from buffer space */
- while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
- TCP_MAXWIN << tp->request_r_scale <
- so->so_rcv.sb_hiwat)
- tp->request_r_scale++;
+ ))) {
+ callout_reset(tp->tt_delack, tcp_delacktime,
+ tcp_timer_delack, tp);
+ tp->t_flags |= TF_NEEDSYN;
+ } else
+ tp->t_flags |= (TF_ACKNOW | TF_NEEDSYN);
+
+ tcpstat.tcps_connects++;
+ soisconnected(so);
+ goto trimthenstep6;
}
+ goto drop;
}
+after_listen:
+
+/* XXX temp debugging */
+ /* should not happen - syncache should pick up these connections */
+ if (tp->t_state == TCPS_LISTEN)
+ panic("tcp_input: TCPS_LISTEN");
/*
* Segment received on connection.
@@ -872,11 +882,25 @@ findpcb:
callout_reset(tp->tt_keep, tcp_keepidle, tcp_timer_keep, tp);
/*
- * Process options if not in LISTEN state,
- * else do it below (after getting remote address).
+ * Process options.
+ * XXX this is tradtitional behavior, may need to be cleaned up.
*/
- if (tp->t_state != TCPS_LISTEN)
- tcp_dooptions(tp, optp, optlen, th, &to);
+ tcp_dooptions(&to, optp, optlen, thflags & TH_SYN);
+ if (thflags & TH_SYN) {
+ if (to.to_flags & TOF_SCALE) {
+ tp->t_flags |= TF_RCVD_SCALE;
+ tp->requested_s_scale = to.to_requested_s_scale;
+ }
+ if (to.to_flags & TOF_TS) {
+ tp->t_flags |= TF_RCVD_TSTMP;
+ tp->ts_recent = to.to_tsval;
+ tp->ts_recent_age = ticks;
+ }
+ if (to.to_flags & (TOF_CC|TOF_CCNEW))
+ tp->t_flags |= TF_RCVD_CC;
+ if (to.to_flags & TOF_MSS)
+ tcp_mss(tp, to.to_mss);
+ }
/*
* Header prediction: check for the two common cases
@@ -898,7 +922,7 @@ findpcb:
if (tp->t_state == TCPS_ESTABLISHED &&
(thflags & (TH_SYN|TH_FIN|TH_RST|TH_URG|TH_ACK)) == TH_ACK &&
((tp->t_flags & (TF_NEEDSYN|TF_NEEDFIN)) == 0) &&
- ((to.to_flag & TOF_TS) == 0 ||
+ ((to.to_flags & TOF_TS) == 0 ||
TSTMP_GEQ(to.to_tsval, tp->ts_recent)) &&
/*
* Using the CC option is compulsory if once started:
@@ -906,7 +930,7 @@ findpcb:
* if the segment has a CC option equal to CCrecv
*/
((tp->t_flags & (TF_REQ_CC|TF_RCVD_CC)) != (TF_REQ_CC|TF_RCVD_CC) ||
- ((to.to_flag & TOF_CC) != 0 && to.to_cc == tp->cc_recv)) &&
+ ((to.to_flags & TOF_CC) != 0 && to.to_cc == tp->cc_recv)) &&
th->th_seq == tp->rcv_nxt &&
tiwin && tiwin == tp->snd_wnd &&
tp->snd_nxt == tp->snd_max) {
@@ -917,7 +941,7 @@ findpcb:
* NOTE that the test is modified according to the latest
* proposal of the tcplw@cray.com list (Braden 1993/04/26).
*/
- if ((to.to_flag & TOF_TS) != 0 &&
+ if ((to.to_flags & TOF_TS) != 0 &&
SEQ_LEQ(th->th_seq, tp->last_ack_sent)) {
tp->ts_recent_age = ticks;
tp->ts_recent = to.to_tsval;
@@ -943,7 +967,7 @@ findpcb:
tp->snd_nxt = tp->snd_max;
tp->t_badrxtwin = 0;
}
- if ((to.to_flag & TOF_TS) != 0)
+ if ((to.to_flags & TOF_TS) != 0)
tcp_xmit_timer(tp,
ticks - to.to_tsecr + 1);
else if (tp->t_rtttime &&
@@ -1025,209 +1049,6 @@ findpcb:
switch (tp->t_state) {
/*
- * If the state is LISTEN then ignore segment if it contains an RST.
- * If the segment contains an ACK then it is bad and send a RST.
- * If it does not contain a SYN then it is not interesting; drop it.
- * If it is from this socket, drop it, it must be forged.
- * Don't bother responding if the destination was a broadcast.
- * Otherwise initialize tp->rcv_nxt, and tp->irs, select an initial
- * tp->iss, and send a segment:
- * <SEQ=ISS><ACK=RCV_NXT><CTL=SYN,ACK>
- * Also initialize tp->snd_nxt to tp->iss+1 and tp->snd_una to tp->iss.
- * Fill in remote peer address fields if not previously specified.
- * Enter SYN_RECEIVED state, and process any other fields of this
- * segment in this state.
- */
- case TCPS_LISTEN: {
- register struct sockaddr_in *sin;
-#ifdef INET6
- register struct sockaddr_in6 *sin6;
-#endif
-
- if (thflags & TH_RST)
- goto drop;
- if (thflags & TH_ACK) {
- rstreason = BANDLIM_RST_OPENPORT;
- goto dropwithreset;
- }
- if ((thflags & TH_SYN) == 0)
- goto drop;
- if (th->th_dport == th->th_sport) {
-#ifdef INET6
- if (isipv6) {
- if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
- &ip6->ip6_src))
- goto drop;
- } else
-#endif /* INET6 */
- if (ip->ip_dst.s_addr == ip->ip_src.s_addr)
- goto drop;
- }
- /*
- * RFC1122 4.2.3.10, p. 104: discard bcast/mcast SYN
- * in_broadcast() should never return true on a received
- * packet with M_BCAST not set.
- *
- * Packets with a multicast source address should also
- * be discarded.
- */
- if (m->m_flags & (M_BCAST|M_MCAST))
- goto drop;
-#ifdef INET6
- if (isipv6) {
- if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
- IN6_IS_ADDR_MULTICAST(&ip6->ip6_src))
- goto drop;
- } else
-#endif
- if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) ||
- IN_MULTICAST(ntohl(ip->ip_src.s_addr)) ||
- ip->ip_src.s_addr == htonl(INADDR_BROADCAST))
- goto drop;
-#ifdef INET6
- if (isipv6) {
- MALLOC(sin6, struct sockaddr_in6 *, sizeof *sin6,
- M_SONAME, M_NOWAIT | M_ZERO);
- if (sin6 == NULL)
- goto drop;
- sin6->sin6_family = AF_INET6;
- sin6->sin6_len = sizeof(*sin6);
- sin6->sin6_addr = ip6->ip6_src;
- sin6->sin6_port = th->th_sport;
- laddr6 = inp->in6p_laddr;
- if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr))
- inp->in6p_laddr = ip6->ip6_dst;
- if (in6_pcbconnect(inp, (struct sockaddr *)sin6,
- thread0)) {
- inp->in6p_laddr = laddr6;
- FREE(sin6, M_SONAME);
- goto drop;
- }
- FREE(sin6, M_SONAME);
- } else
-#endif
- {
- MALLOC(sin, struct sockaddr_in *, sizeof *sin, M_SONAME,
- M_NOWAIT);
- if (sin == NULL)
- goto drop;
- sin->sin_family = AF_INET;
- sin->sin_len = sizeof(*sin);
- sin->sin_addr = ip->ip_src;
- sin->sin_port = th->th_sport;
- bzero((caddr_t)sin->sin_zero, sizeof(sin->sin_zero));
- laddr = inp->inp_laddr;
- if (inp->inp_laddr.s_addr == INADDR_ANY)
- inp->inp_laddr = ip->ip_dst;
- if (in_pcbconnect(inp, (struct sockaddr *)sin, thread0)) {
- inp->inp_laddr = laddr;
- FREE(sin, M_SONAME);
- goto drop;
- }
- FREE(sin, M_SONAME);
- }
- if ((taop = tcp_gettaocache(inp)) == NULL) {
- taop = &tao_noncached;
- bzero(taop, sizeof(*taop));
- }
- tcp_dooptions(tp, optp, optlen, th, &to);
- if (iss)
- tp->iss = iss;
- else {
- tp->iss = tcp_new_isn(tp);
- }
- tp->irs = th->th_seq;
- tcp_sendseqinit(tp);
- tcp_rcvseqinit(tp);
- /*
- * Initialization of the tcpcb for transaction;
- * set SND.WND = SEG.WND,
- * initialize CCsend and CCrecv.
- */
- tp->snd_wnd = tiwin; /* initial send-window */
- tp->cc_send = CC_INC(tcp_ccgen);
- tp->cc_recv = to.to_cc;
- /*
- * Perform TAO test on incoming CC (SEG.CC) option, if any.
- * - compare SEG.CC against cached CC from the same host,
- * if any.
- * - if SEG.CC > chached value, SYN must be new and is accepted
- * immediately: save new CC in the cache, mark the socket
- * connected, enter ESTABLISHED state, turn on flag to
- * send a SYN in the next segment.
- * A virtual advertised window is set in rcv_adv to
- * initialize SWS prevention. Then enter normal segment
- * processing: drop SYN, process data and FIN.
- * - otherwise do a normal 3-way handshake.
- */
- if ((to.to_flag & TOF_CC) != 0) {
- if (((tp->t_flags & TF_NOPUSH) != 0) &&
- taop->tao_cc != 0 && CC_GT(to.to_cc, taop->tao_cc)) {
-
- taop->tao_cc = to.to_cc;
- tp->t_starttime = ticks;
- tp->t_state = TCPS_ESTABLISHED;
-
- /*
- * If there is a FIN, or if there is data and the
- * connection is local, then delay SYN,ACK(SYN) in
- * the hope of piggy-backing it on a response
- * segment. Otherwise must send ACK now in case
- * the other side is slow starting.
- */
- if (DELAY_ACK(tp) && ((thflags & TH_FIN) ||
- (tlen != 0 &&
-#ifdef INET6
- ((isipv6 && in6_localaddr(&inp->in6p_faddr))
- ||
- (!isipv6 &&
-#endif
- in_localaddr(inp->inp_faddr)
-#ifdef INET6
- ))
-#endif
- ))) {
- callout_reset(tp->tt_delack, tcp_delacktime,
- tcp_timer_delack, tp);
- tp->t_flags |= TF_NEEDSYN;
- } else
- tp->t_flags |= (TF_ACKNOW | TF_NEEDSYN);
-
- /*
- * Limit the `virtual advertised window' to TCP_MAXWIN
- * here. Even if we requested window scaling, it will
- * become effective only later when our SYN is acked.
- */
- tp->rcv_adv += min(tp->rcv_wnd, TCP_MAXWIN);
- tcpstat.tcps_connects++;
- soisconnected(so);
- callout_reset(tp->tt_keep, tcp_keepinit,
- tcp_timer_keep, tp);
- dropsocket = 0; /* committed to socket */
- tcpstat.tcps_accepts++;
- goto trimthenstep6;
- }
- /* else do standard 3-way handshake */
- } else {
- /*
- * No CC option, but maybe CC.NEW:
- * invalidate cached value.
- */
- taop->tao_cc = 0;
- }
- /*
- * TAO test failed or there was no CC option,
- * do a standard 3-way handshake.
- */
- tp->t_flags |= TF_ACKNOW;
- tp->t_state = TCPS_SYN_RECEIVED;
- callout_reset(tp->tt_keep, tcp_keepinit, tcp_timer_keep, tp);
- dropsocket = 0; /* committed to socket */
- tcpstat.tcps_accepts++;
- goto trimthenstep6;
- }
-
- /*
* If the state is SYN_RECEIVED:
* if seg contains an ACK, but not for our SYN/ACK, send a RST.
*/
@@ -1253,7 +1074,7 @@ findpcb:
* continue processing rest of data/controls, beginning with URG
*/
case TCPS_SYN_SENT:
- if ((taop = tcp_gettaocache(inp)) == NULL) {
+ if ((taop = tcp_gettaocache(&inp->inp_inc)) == NULL) {
taop = &tao_noncached;
bzero(taop, sizeof(*taop));
}
@@ -1297,7 +1118,7 @@ findpcb:
* by the old rules. If no CC.ECHO option, make sure
* we don't get fooled into using T/TCP.
*/
- if (to.to_flag & TOF_CCECHO) {
+ if (to.to_flags & TOF_CCECHO) {
if (tp->cc_send != to.to_ccecho) {
if (taop->tao_ccsent != 0)
goto drop;
@@ -1359,7 +1180,7 @@ findpcb:
*/
tp->t_flags |= TF_ACKNOW;
callout_stop(tp->tt_rexmt);
- if (to.to_flag & TOF_CC) {
+ if (to.to_flags & TOF_CC) {
if (taop->tao_cc != 0 &&
CC_GT(to.to_cc, taop->tao_cc)) {
/*
@@ -1434,7 +1255,7 @@ trimthenstep6:
case TCPS_CLOSING:
case TCPS_TIME_WAIT:
if ((thflags & TH_SYN) &&
- (to.to_flag & TOF_CC) && tp->cc_recv != 0) {
+ (to.to_flags & TOF_CC) && tp->cc_recv != 0) {
if (tp->t_state == TCPS_TIME_WAIT &&
(ticks - tp->t_starttime) > tcp_msl) {
rstreason = BANDLIM_UNLIMITED;
@@ -1542,7 +1363,7 @@ trimthenstep6:
* RFC 1323 PAWS: If we have a timestamp reply on this segment
* and it's less than ts_recent, drop it.
*/
- if ((to.to_flag & TOF_TS) != 0 && tp->ts_recent &&
+ if ((to.to_flags & TOF_TS) != 0 && tp->ts_recent &&
TSTMP_LT(to.to_tsval, tp->ts_recent)) {
/* Check to see if ts_recent is over 24 days old. */
@@ -1574,7 +1395,7 @@ trimthenstep6:
* RST segments do not have to comply with this.
*/
if ((tp->t_flags & (TF_REQ_CC|TF_RCVD_CC)) == (TF_REQ_CC|TF_RCVD_CC) &&
- ((to.to_flag & TOF_CC) == 0 || tp->cc_recv != to.to_cc))
+ ((to.to_flags & TOF_CC) == 0 || tp->cc_recv != to.to_cc))
goto dropafterack;
/*
@@ -1694,7 +1515,7 @@ trimthenstep6:
* NOTE that the test is modified according to the latest
* proposal of the tcplw@cray.com list (Braden 1993/04/26).
*/
- if ((to.to_flag & TOF_TS) != 0 &&
+ if ((to.to_flags & TOF_TS) != 0 &&
SEQ_LEQ(th->th_seq, tp->last_ack_sent)) {
tp->ts_recent_age = ticks;
tp->ts_recent = to.to_tsval;
@@ -1748,7 +1569,7 @@ trimthenstep6:
* update cache.CC if it was undefined, pass any queued
* data to the user, and advance state appropriately.
*/
- if ((taop = tcp_gettaocache(inp)) != NULL &&
+ if ((taop = tcp_gettaocache(&inp->inp_inc)) != NULL &&
taop->tao_cc == 0)
taop->tao_cc = tp->cc_recv;
@@ -1940,7 +1761,7 @@ process_ACK:
* timer backoff (cf., Phil Karn's retransmit alg.).
* Recompute the initial retransmit timer.
*/
- if (to.to_flag & TOF_TS)
+ if (to.to_flags & TOF_TS)
tcp_xmit_timer(tp, ticks - to.to_tsecr + 1);
else if (tp->t_rtttime && SEQ_GT(th->th_ack, tp->t_rtseq))
tcp_xmit_timer(tp, ticks - tp->t_rtttime);
@@ -2371,9 +2192,6 @@ dropwithreset:
tcp_respond(tp, mtod(m, void *), th, m, th->th_seq+tlen,
(tcp_seq)0, TH_RST|TH_ACK);
}
- /* destroy temporarily created socket */
- if (dropsocket)
- (void) soabort(so);
return;
drop:
@@ -2386,23 +2204,21 @@ drop:
&tcp_savetcp, 0);
#endif
m_freem(m);
- /* destroy temporarily created socket */
- if (dropsocket)
- (void) soabort(so);
return;
}
+/*
+ * Parse TCP options and place in tcpopt.
+ */
static void
-tcp_dooptions(tp, cp, cnt, th, to)
- struct tcpcb *tp;
+tcp_dooptions(to, cp, cnt, is_syn)
+ struct tcpopt *to;
u_char *cp;
int cnt;
- struct tcphdr *th;
- struct tcpopt *to;
{
- u_short mss = 0;
int opt, optlen;
+ to->to_flags = 0;
for (; cnt > 0; cnt -= optlen, cp += optlen) {
opt = cp[0];
if (opt == TCPOPT_EOL)
@@ -2417,92 +2233,67 @@ tcp_dooptions(tp, cp, cnt, th, to)
break;
}
switch (opt) {
-
- default:
- continue;
-
case TCPOPT_MAXSEG:
if (optlen != TCPOLEN_MAXSEG)
continue;
- if (!(th->th_flags & TH_SYN))
+ if (!is_syn)
continue;
- bcopy((char *) cp + 2, (char *) &mss, sizeof(mss));
- NTOHS(mss);
+ to->to_flags |= TOF_MSS;
+ bcopy((char *)cp + 2,
+ (char *)&to->to_mss, sizeof(to->to_mss));
+ NTOHS(to->to_mss);
break;
-
case TCPOPT_WINDOW:
if (optlen != TCPOLEN_WINDOW)
continue;
- if (!(th->th_flags & TH_SYN))
+ if (! is_syn)
continue;
- tp->t_flags |= TF_RCVD_SCALE;
- tp->requested_s_scale = min(cp[2], TCP_MAX_WINSHIFT);
+ to->to_flags |= TOF_SCALE;
+ to->to_requested_s_scale = min(cp[2], TCP_MAX_WINSHIFT);
break;
-
case TCPOPT_TIMESTAMP:
if (optlen != TCPOLEN_TIMESTAMP)
continue;
- to->to_flag |= TOF_TS;
+ to->to_flags |= TOF_TS;
bcopy((char *)cp + 2,
(char *)&to->to_tsval, sizeof(to->to_tsval));
NTOHL(to->to_tsval);
bcopy((char *)cp + 6,
(char *)&to->to_tsecr, sizeof(to->to_tsecr));
NTOHL(to->to_tsecr);
-
- /*
- * A timestamp received in a SYN makes
- * it ok to send timestamp requests and replies.
- */
- if (th->th_flags & TH_SYN) {
- tp->t_flags |= TF_RCVD_TSTMP;
- tp->ts_recent = to->to_tsval;
- tp->ts_recent_age = ticks;
- }
break;
case TCPOPT_CC:
if (optlen != TCPOLEN_CC)
continue;
- to->to_flag |= TOF_CC;
+ to->to_flags |= TOF_CC;
bcopy((char *)cp + 2,
(char *)&to->to_cc, sizeof(to->to_cc));
NTOHL(to->to_cc);
- /*
- * A CC or CC.new option received in a SYN makes
- * it ok to send CC in subsequent segments.
- */
- if (th->th_flags & TH_SYN)
- tp->t_flags |= TF_RCVD_CC;
break;
case TCPOPT_CCNEW:
if (optlen != TCPOLEN_CC)
continue;
- if (!(th->th_flags & TH_SYN))
+ if (!is_syn)
continue;
- to->to_flag |= TOF_CCNEW;
+ to->to_flags |= TOF_CCNEW;
bcopy((char *)cp + 2,
(char *)&to->to_cc, sizeof(to->to_cc));
NTOHL(to->to_cc);
- /*
- * A CC or CC.new option received in a SYN makes
- * it ok to send CC in subsequent segments.
- */
- tp->t_flags |= TF_RCVD_CC;
break;
case TCPOPT_CCECHO:
if (optlen != TCPOLEN_CC)
continue;
- if (!(th->th_flags & TH_SYN))
+ if (!is_syn)
continue;
- to->to_flag |= TOF_CCECHO;
+ to->to_flags |= TOF_CCECHO;
bcopy((char *)cp + 2,
(char *)&to->to_ccecho, sizeof(to->to_ccecho));
NTOHL(to->to_ccecho);
break;
+ default:
+ continue;
}
}
- if (th->th_flags & TH_SYN)
- tcp_mss(tp, mss); /* sets t_maxseg */
}
/*
@@ -2675,10 +2466,10 @@ tcp_mss(tp, offer)
#endif
#ifdef INET6
if (isipv6)
- rt = tcp_rtlookup6(inp);
+ rt = tcp_rtlookup6(&inp->inp_inc);
else
#endif
- rt = tcp_rtlookup(inp);
+ rt = tcp_rtlookup(&inp->inp_inc);
if (rt == NULL) {
tp->t_maxopd = tp->t_maxseg =
#ifdef INET6
@@ -2884,10 +2675,10 @@ tcp_mssopt(tp)
#endif
#ifdef INET6
if (isipv6)
- rt = tcp_rtlookup6(tp->t_inpcb);
+ rt = tcp_rtlookup6(&tp->t_inpcb->inp_inc);
else
#endif /* INET6 */
- rt = tcp_rtlookup(tp->t_inpcb);
+ rt = tcp_rtlookup(&tp->t_inpcb->inp_inc);
if (rt == NULL)
return
#ifdef INET6
diff --git a/sys/netinet/tcp_output.c b/sys/netinet/tcp_output.c
index e473b8e..683eeef 100644
--- a/sys/netinet/tcp_output.c
+++ b/sys/netinet/tcp_output.c
@@ -216,7 +216,7 @@ again:
len = (long)ulmin(so->so_snd.sb_cc, win) - off;
- if ((taop = tcp_gettaocache(tp->t_inpcb)) == NULL) {
+ if ((taop = tcp_gettaocache(&tp->t_inpcb->inp_inc)) == NULL) {
taop = &tao_noncached;
bzero(taop, sizeof(*taop));
}
diff --git a/sys/netinet/tcp_reass.c b/sys/netinet/tcp_reass.c
index 7ab594e..6740ebe 100644
--- a/sys/netinet/tcp_reass.c
+++ b/sys/netinet/tcp_reass.c
@@ -120,11 +120,6 @@ SYSCTL_INT(_net_inet_tcp, OID_AUTO, delayed_ack, CTLFLAG_RW,
&tcp_delack_enabled, 0,
"Delay ACK to try and piggyback it onto a data packet");
-int tcp_lq_overflow = 1;
-SYSCTL_INT(_net_inet_tcp, OID_AUTO, tcp_lq_overflow, CTLFLAG_RW,
- &tcp_lq_overflow, 0,
- "Listen Queue Overflow");
-
#ifdef TCP_DROP_SYNFIN
static int drop_synfin = 0;
SYSCTL_INT(_net_inet_tcp, OID_AUTO, drop_synfin, CTLFLAG_RW,
@@ -135,10 +130,9 @@ struct inpcbhead tcb;
#define tcb6 tcb /* for KAME src sync over BSD*'s */
struct inpcbinfo tcbinfo;
-static void tcp_dooptions __P((struct tcpcb *,
- u_char *, int, struct tcphdr *, struct tcpopt *));
+static void tcp_dooptions __P((struct tcpopt *, u_char *, int, int));
static void tcp_pulloutofband __P((struct socket *,
- struct tcphdr *, struct mbuf *, int));
+ struct tcphdr *, struct mbuf *, int));
static int tcp_reass __P((struct tcpcb *, struct tcphdr *, int *,
struct mbuf *));
static void tcp_xmit_timer __P((struct tcpcb *, int));
@@ -344,11 +338,6 @@ tcp_input(m, off0)
register int thflags;
struct socket *so = 0;
int todrop, acked, ourfinisacked, needoutput = 0;
- struct in_addr laddr;
-#ifdef INET6
- struct in6_addr laddr6;
-#endif
- int dropsocket = 0;
int iss = 0;
u_long tiwin;
struct tcpopt to; /* options in this segment */
@@ -643,6 +632,7 @@ findpcb:
so = inp->inp_socket;
if (so->so_options & (SO_DEBUG|SO_ACCEPTCONN)) {
+ struct in_conninfo inc;
#ifdef TCPDEBUG
if (so->so_options & SO_DEBUG) {
ostate = tp->t_state;
@@ -656,212 +646,232 @@ findpcb:
tcp_savetcp = *th;
}
#endif
- if (so->so_options & SO_ACCEPTCONN) {
- register struct tcpcb *tp0 = tp;
- struct socket *so2;
-#ifdef IPSEC
- struct socket *oso;
-#endif
+ /* skip if this isn't a listen socket */
+ if ((so->so_options & SO_ACCEPTCONN) == 0)
+ goto after_listen;
#ifdef INET6
- struct inpcb *oinp = sotoinpcb(so);
-#endif /* INET6 */
+ inc.inc_isipv6 = isipv6;
+ if (isipv6) {
+ inc.inc6_faddr = ip6->ip6_src;
+ inc.inc6_laddr = ip6->ip6_dst;
+ inc.inc6_route.ro_rt = NULL; /* XXX */
-#ifndef IPSEC
- /*
- * Current IPsec implementation makes incorrect IPsec
- * cache if this check is done here.
- * So delay this until duplicated socket is created.
- */
- if ((thflags & (TH_RST|TH_ACK|TH_SYN)) != TH_SYN) {
- /*
- * Note: dropwithreset makes sure we don't
- * send a RST in response to a RST.
- */
- if (thflags & TH_ACK) {
+ } else
+#endif /* INET6 */
+ {
+ inc.inc_faddr = ip->ip_src;
+ inc.inc_laddr = ip->ip_dst;
+ inc.inc_route.ro_rt = NULL; /* XXX */
+ }
+ inc.inc_fport = th->th_sport;
+ inc.inc_lport = th->th_dport;
+
+ /*
+ * If the state is LISTEN then ignore segment if it contains
+ * a RST. If the segment contains an ACK then it is bad and
+ * send a RST. If it does not contain a SYN then it is not
+ * interesting; drop it.
+ *
+ * If the state is SYN_RECEIVED (syncache) and seg contains
+ * an ACK, but not for our SYN/ACK, send a RST. If the seg
+ * contains a RST, check the sequence number to see if it
+ * is a valid reset segment.
+ */
+ if ((thflags & (TH_RST|TH_ACK|TH_SYN)) != TH_SYN) {
+ if ((thflags & (TH_RST|TH_ACK|TH_SYN)) == TH_ACK) {
+ if (!syncache_expand(&inc, th, &so, m)) {
+ /*
+ * No syncache entry, or ACK was not
+ * for our SYN/ACK. Send a RST.
+ */
tcpstat.tcps_badsyn++;
rstreason = BANDLIM_RST_OPENPORT;
goto dropwithreset;
}
+ if (so == NULL)
+ /*
+ * Could not complete 3-way handshake,
+ * connection is being closed down, and
+ * syncache will free mbuf.
+ */
+ return;
+ /*
+ * Socket is created in state SYN_RECEIVED.
+ * Continue processing segment.
+ */
+ inp = sotoinpcb(so);
+ tp = intotcpcb(inp);
+ /*
+ * This is what would have happened in
+ * tcp_ouput() when the SYN,ACK was sent.
+ */
+ tp->snd_up = tp->snd_una;
+ tp->snd_max = tp->snd_nxt = tp->iss + 1;
+ tp->last_ack_sent = tp->rcv_nxt;
+/*
+ * XXX possible bug - it doesn't appear that tp->snd_wnd is unscaled
+ * until the _second_ ACK is received:
+ * rcv SYN (set wscale opts) --> send SYN/ACK, set snd_wnd = window.
+ * rcv ACK, calculate tiwin --> process SYN_RECEIVED, determine wscale,
+ * move to ESTAB, set snd_wnd to tiwin.
+ */
+ tp->snd_wnd = tiwin; /* unscaled */
+ goto after_listen;
+ }
+ if (thflags & TH_RST) {
+ syncache_chkrst(&inc, th);
goto drop;
}
-#endif
+ if (thflags & TH_ACK) {
+ syncache_badack(&inc);
+ tcpstat.tcps_badsyn++;
+ rstreason = BANDLIM_RST_OPENPORT;
+ goto dropwithreset;
+ }
+ goto drop;
+ }
+ /*
+ * Segment's flags are (SYN) or (SYN|FIN).
+ */
#ifdef INET6
- /*
- * If deprecated address is forbidden,
- * we do not accept SYN to deprecated interface
- * address to prevent any new inbound connection from
- * getting established.
- * When we do not accept SYN, we send a TCP RST,
- * with deprecated source address (instead of dropping
- * it). We compromise it as it is much better for peer
- * to send a RST, and RST will be the final packet
- * for the exchange.
- *
- * If we do not forbid deprecated addresses, we accept
- * the SYN packet. RFC2462 does not suggest dropping
- * SYN in this case.
- * If we decipher RFC2462 5.5.4, it says like this:
- * 1. use of deprecated addr with existing
- * communication is okay - "SHOULD continue to be
- * used"
- * 2. use of it with new communication:
- * (2a) "SHOULD NOT be used if alternate address
- * with sufficient scope is available"
- * (2b) nothing mentioned otherwise.
- * Here we fall into (2b) case as we have no choice in
- * our source address selection - we must obey the peer.
- *
- * The wording in RFC2462 is confusing, and there are
- * multiple description text for deprecated address
- * handling - worse, they are not exactly the same.
- * I believe 5.5.4 is the best one, so we follow 5.5.4.
- */
- if (isipv6 && !ip6_use_deprecated) {
- struct in6_ifaddr *ia6;
-
- if ((ia6 = ip6_getdstifaddr(m)) &&
- (ia6->ia6_flags & IN6_IFF_DEPRECATED)) {
- tp = NULL;
- rstreason = BANDLIM_RST_OPENPORT;
- goto dropwithreset;
- }
- }
-#endif
+ /*
+ * If deprecated address is forbidden,
+ * we do not accept SYN to deprecated interface
+ * address to prevent any new inbound connection from
+ * getting established.
+ * When we do not accept SYN, we send a TCP RST,
+ * with deprecated source address (instead of dropping
+ * it). We compromise it as it is much better for peer
+ * to send a RST, and RST will be the final packet
+ * for the exchange.
+ *
+ * If we do not forbid deprecated addresses, we accept
+ * the SYN packet. RFC2462 does not suggest dropping
+ * SYN in this case.
+ * If we decipher RFC2462 5.5.4, it says like this:
+ * 1. use of deprecated addr with existing
+ * communication is okay - "SHOULD continue to be
+ * used"
+ * 2. use of it with new communication:
+ * (2a) "SHOULD NOT be used if alternate address
+ * with sufficient scope is available"
+ * (2b) nothing mentioned otherwise.
+ * Here we fall into (2b) case as we have no choice in
+ * our source address selection - we must obey the peer.
+ *
+ * The wording in RFC2462 is confusing, and there are
+ * multiple description text for deprecated address
+ * handling - worse, they are not exactly the same.
+ * I believe 5.5.4 is the best one, so we follow 5.5.4.
+ */
+ if (isipv6 && !ip6_use_deprecated) {
+ struct in6_ifaddr *ia6;
- so2 = sonewconn(so, 0);
- if (so2 == 0) {
- /*
- * If we were unable to create a new socket
- * for this SYN, we call sodropablereq to
- * see if there are any other sockets we
- * can kick out of the listen queue. If
- * so, we'll silently drop the socket
- * sodropablereq told us to drop and
- * create a new one.
- *
- * If sodropablereq returns 0, we'll
- * simply drop the incoming SYN, as we
- * can not allocate a socket for it.
- */
- tcpstat.tcps_listendrop++;
- so2 = sodropablereq(so);
- if (so2) {
- if (tcp_lq_overflow)
- sototcpcb(so2)->t_flags |=
- TF_LQ_OVERFLOW;
- tcp_close(sototcpcb(so2));
- so2 = sonewconn(so, 0);
- }
- if (!so2)
- goto drop;
+ if ((ia6 = ip6_getdstifaddr(m)) &&
+ (ia6->ia6_flags & IN6_IFF_DEPRECATED)) {
+ tp = NULL;
+ rstreason = BANDLIM_RST_OPENPORT;
+ goto dropwithreset;
}
-#ifdef IPSEC
- oso = so;
+ }
#endif
- so = so2;
- /*
- * This is ugly, but ....
- *
- * Mark socket as temporary until we're
- * committed to keeping it. The code at
- * ``drop'' and ``dropwithreset'' check the
- * flag dropsocket to see if the temporary
- * socket created here should be discarded.
- * We mark the socket as discardable until
- * we're committed to it below in TCPS_LISTEN.
- */
- dropsocket++;
- inp = (struct inpcb *)so->so_pcb;
+ /*
+ * If it is from this socket, drop it, it must be forged.
+ * Don't bother responding if the destination was a broadcast.
+ */
+ if (th->th_dport == th->th_sport) {
#ifdef INET6
- if (isipv6)
- inp->in6p_laddr = ip6->ip6_dst;
- else {
- inp->inp_vflag &= ~INP_IPV6;
- inp->inp_vflag |= INP_IPV4;
+ if (isipv6) {
+ if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
+ &ip6->ip6_src))
+ goto drop;
+ } else
#endif /* INET6 */
- inp->inp_laddr = ip->ip_dst;
+ if (ip->ip_dst.s_addr == ip->ip_src.s_addr)
+ goto drop;
+ }
+ /*
+ * RFC1122 4.2.3.10, p. 104: discard bcast/mcast SYN
+ * in_broadcast() should never return true on a received
+ * packet with M_BCAST not set.
+ *
+ * Packets with a multicast source address should also
+ * be discarded.
+ */
+ if (m->m_flags & (M_BCAST|M_MCAST))
+ goto drop;
#ifdef INET6
- }
-#endif /* INET6 */
- inp->inp_lport = th->th_dport;
- if (in_pcbinshash(inp) != 0) {
+ if (isipv6) {
+ if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
+ IN6_IS_ADDR_MULTICAST(&ip6->ip6_src))
+ goto drop;
+ } else
+#endif
+ if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) ||
+ IN_MULTICAST(ntohl(ip->ip_src.s_addr)) ||
+ ip->ip_src.s_addr == htonl(INADDR_BROADCAST))
+ goto drop;
+ /*
+ * SYN appears to be valid; create compressed TCP state
+ * for syncache, or perform t/tcp connection.
+ */
+ if (so->so_qlen <= so->so_qlimit) {
+ tcp_dooptions(&to, optp, optlen, 1);
+ if (!syncache_add(&inc, &to, th, &so, m))
+ goto drop;
+ if (so == NULL)
/*
- * Undo the assignments above if we failed to
- * put the PCB on the hash lists.
+ * Entry added to syncache, mbuf used to
+ * send SYN,ACK packet.
*/
-#ifdef INET6
- if (isipv6)
- inp->in6p_laddr = in6addr_any;
- else
-#endif /* INET6 */
- inp->inp_laddr.s_addr = INADDR_ANY;
- inp->inp_lport = 0;
- goto drop;
- }
-#ifdef IPSEC
+ return;
/*
- * To avoid creating incorrectly cached IPsec
- * association, this is need to be done here.
- *
- * Subject: (KAME-snap 748)
- * From: Wayne Knowles <w.knowles@niwa.cri.nz>
- * ftp://ftp.kame.net/pub/mail-list/snap-users/748
+ * Segment passed TAO tests.
*/
- if ((thflags & (TH_RST|TH_ACK|TH_SYN)) != TH_SYN) {
- /*
- * Note: dropwithreset makes sure we don't
- * send a RST in response to a RST.
- */
- if (thflags & TH_ACK) {
- tcpstat.tcps_badsyn++;
- rstreason = BANDLIM_RST_OPENPORT;
- goto dropwithreset;
- }
- goto drop;
- }
+ inp = sotoinpcb(so);
+ tp = intotcpcb(inp);
+ tp->snd_wnd = tiwin;
+ tp->t_starttime = ticks;
+ tp->t_state = TCPS_ESTABLISHED;
+
+ /*
+ * If there is a FIN, or if there is data and the
+ * connection is local, then delay SYN,ACK(SYN) in
+ * the hope of piggy-backing it on a response
+ * segment. Otherwise must send ACK now in case
+ * the other side is slow starting.
+ */
+ if (DELAY_ACK(tp) && ((thflags & TH_FIN) ||
+ (tlen != 0 &&
+#ifdef INET6
+ ((isipv6 && in6_localaddr(&inp->in6p_faddr))
+ ||
+ (!isipv6 &&
#endif
+ in_localaddr(inp->inp_faddr)
#ifdef INET6
- if (isipv6) {
- /*
- * Inherit socket options from the listening
- * socket.
- * Note that in6p_inputopts are not (even
- * should not be) copied, since it stores
- * previously received options and is used to
- * detect if each new option is different than
- * the previous one and hence should be passed
- * to a user.
- * If we copied in6p_inputopts, a user would
- * not be able to receive options just after
- * calling the accept system call.
- */
- inp->inp_flags |=
- oinp->inp_flags & INP_CONTROLOPTS;
- if (oinp->in6p_outputopts)
- inp->in6p_outputopts =
- ip6_copypktopts(oinp->in6p_outputopts,
- M_NOWAIT);
- } else
-#endif /* INET6 */
- inp->inp_options = ip_srcroute();
-#ifdef IPSEC
- /* copy old policy into new socket's */
- if (ipsec_copy_policy(sotoinpcb(oso)->inp_sp,
- inp->inp_sp))
- printf("tcp_input: could not copy policy\n");
+ ))
#endif
- tp = intotcpcb(inp);
- tp->t_state = TCPS_LISTEN;
- tp->t_flags |= tp0->t_flags & (TF_NOPUSH|TF_NOOPT);
-
- /* Compute proper scaling value from buffer space */
- while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
- TCP_MAXWIN << tp->request_r_scale <
- so->so_rcv.sb_hiwat)
- tp->request_r_scale++;
+ ))) {
+ callout_reset(tp->tt_delack, tcp_delacktime,
+ tcp_timer_delack, tp);
+ tp->t_flags |= TF_NEEDSYN;
+ } else
+ tp->t_flags |= (TF_ACKNOW | TF_NEEDSYN);
+
+ tcpstat.tcps_connects++;
+ soisconnected(so);
+ goto trimthenstep6;
}
+ goto drop;
}
+after_listen:
+
+/* XXX temp debugging */
+ /* should not happen - syncache should pick up these connections */
+ if (tp->t_state == TCPS_LISTEN)
+ panic("tcp_input: TCPS_LISTEN");
/*
* Segment received on connection.
@@ -872,11 +882,25 @@ findpcb:
callout_reset(tp->tt_keep, tcp_keepidle, tcp_timer_keep, tp);
/*
- * Process options if not in LISTEN state,
- * else do it below (after getting remote address).
+ * Process options.
+ * XXX this is tradtitional behavior, may need to be cleaned up.
*/
- if (tp->t_state != TCPS_LISTEN)
- tcp_dooptions(tp, optp, optlen, th, &to);
+ tcp_dooptions(&to, optp, optlen, thflags & TH_SYN);
+ if (thflags & TH_SYN) {
+ if (to.to_flags & TOF_SCALE) {
+ tp->t_flags |= TF_RCVD_SCALE;
+ tp->requested_s_scale = to.to_requested_s_scale;
+ }
+ if (to.to_flags & TOF_TS) {
+ tp->t_flags |= TF_RCVD_TSTMP;
+ tp->ts_recent = to.to_tsval;
+ tp->ts_recent_age = ticks;
+ }
+ if (to.to_flags & (TOF_CC|TOF_CCNEW))
+ tp->t_flags |= TF_RCVD_CC;
+ if (to.to_flags & TOF_MSS)
+ tcp_mss(tp, to.to_mss);
+ }
/*
* Header prediction: check for the two common cases
@@ -898,7 +922,7 @@ findpcb:
if (tp->t_state == TCPS_ESTABLISHED &&
(thflags & (TH_SYN|TH_FIN|TH_RST|TH_URG|TH_ACK)) == TH_ACK &&
((tp->t_flags & (TF_NEEDSYN|TF_NEEDFIN)) == 0) &&
- ((to.to_flag & TOF_TS) == 0 ||
+ ((to.to_flags & TOF_TS) == 0 ||
TSTMP_GEQ(to.to_tsval, tp->ts_recent)) &&
/*
* Using the CC option is compulsory if once started:
@@ -906,7 +930,7 @@ findpcb:
* if the segment has a CC option equal to CCrecv
*/
((tp->t_flags & (TF_REQ_CC|TF_RCVD_CC)) != (TF_REQ_CC|TF_RCVD_CC) ||
- ((to.to_flag & TOF_CC) != 0 && to.to_cc == tp->cc_recv)) &&
+ ((to.to_flags & TOF_CC) != 0 && to.to_cc == tp->cc_recv)) &&
th->th_seq == tp->rcv_nxt &&
tiwin && tiwin == tp->snd_wnd &&
tp->snd_nxt == tp->snd_max) {
@@ -917,7 +941,7 @@ findpcb:
* NOTE that the test is modified according to the latest
* proposal of the tcplw@cray.com list (Braden 1993/04/26).
*/
- if ((to.to_flag & TOF_TS) != 0 &&
+ if ((to.to_flags & TOF_TS) != 0 &&
SEQ_LEQ(th->th_seq, tp->last_ack_sent)) {
tp->ts_recent_age = ticks;
tp->ts_recent = to.to_tsval;
@@ -943,7 +967,7 @@ findpcb:
tp->snd_nxt = tp->snd_max;
tp->t_badrxtwin = 0;
}
- if ((to.to_flag & TOF_TS) != 0)
+ if ((to.to_flags & TOF_TS) != 0)
tcp_xmit_timer(tp,
ticks - to.to_tsecr + 1);
else if (tp->t_rtttime &&
@@ -1025,209 +1049,6 @@ findpcb:
switch (tp->t_state) {
/*
- * If the state is LISTEN then ignore segment if it contains an RST.
- * If the segment contains an ACK then it is bad and send a RST.
- * If it does not contain a SYN then it is not interesting; drop it.
- * If it is from this socket, drop it, it must be forged.
- * Don't bother responding if the destination was a broadcast.
- * Otherwise initialize tp->rcv_nxt, and tp->irs, select an initial
- * tp->iss, and send a segment:
- * <SEQ=ISS><ACK=RCV_NXT><CTL=SYN,ACK>
- * Also initialize tp->snd_nxt to tp->iss+1 and tp->snd_una to tp->iss.
- * Fill in remote peer address fields if not previously specified.
- * Enter SYN_RECEIVED state, and process any other fields of this
- * segment in this state.
- */
- case TCPS_LISTEN: {
- register struct sockaddr_in *sin;
-#ifdef INET6
- register struct sockaddr_in6 *sin6;
-#endif
-
- if (thflags & TH_RST)
- goto drop;
- if (thflags & TH_ACK) {
- rstreason = BANDLIM_RST_OPENPORT;
- goto dropwithreset;
- }
- if ((thflags & TH_SYN) == 0)
- goto drop;
- if (th->th_dport == th->th_sport) {
-#ifdef INET6
- if (isipv6) {
- if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
- &ip6->ip6_src))
- goto drop;
- } else
-#endif /* INET6 */
- if (ip->ip_dst.s_addr == ip->ip_src.s_addr)
- goto drop;
- }
- /*
- * RFC1122 4.2.3.10, p. 104: discard bcast/mcast SYN
- * in_broadcast() should never return true on a received
- * packet with M_BCAST not set.
- *
- * Packets with a multicast source address should also
- * be discarded.
- */
- if (m->m_flags & (M_BCAST|M_MCAST))
- goto drop;
-#ifdef INET6
- if (isipv6) {
- if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
- IN6_IS_ADDR_MULTICAST(&ip6->ip6_src))
- goto drop;
- } else
-#endif
- if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) ||
- IN_MULTICAST(ntohl(ip->ip_src.s_addr)) ||
- ip->ip_src.s_addr == htonl(INADDR_BROADCAST))
- goto drop;
-#ifdef INET6
- if (isipv6) {
- MALLOC(sin6, struct sockaddr_in6 *, sizeof *sin6,
- M_SONAME, M_NOWAIT | M_ZERO);
- if (sin6 == NULL)
- goto drop;
- sin6->sin6_family = AF_INET6;
- sin6->sin6_len = sizeof(*sin6);
- sin6->sin6_addr = ip6->ip6_src;
- sin6->sin6_port = th->th_sport;
- laddr6 = inp->in6p_laddr;
- if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr))
- inp->in6p_laddr = ip6->ip6_dst;
- if (in6_pcbconnect(inp, (struct sockaddr *)sin6,
- thread0)) {
- inp->in6p_laddr = laddr6;
- FREE(sin6, M_SONAME);
- goto drop;
- }
- FREE(sin6, M_SONAME);
- } else
-#endif
- {
- MALLOC(sin, struct sockaddr_in *, sizeof *sin, M_SONAME,
- M_NOWAIT);
- if (sin == NULL)
- goto drop;
- sin->sin_family = AF_INET;
- sin->sin_len = sizeof(*sin);
- sin->sin_addr = ip->ip_src;
- sin->sin_port = th->th_sport;
- bzero((caddr_t)sin->sin_zero, sizeof(sin->sin_zero));
- laddr = inp->inp_laddr;
- if (inp->inp_laddr.s_addr == INADDR_ANY)
- inp->inp_laddr = ip->ip_dst;
- if (in_pcbconnect(inp, (struct sockaddr *)sin, thread0)) {
- inp->inp_laddr = laddr;
- FREE(sin, M_SONAME);
- goto drop;
- }
- FREE(sin, M_SONAME);
- }
- if ((taop = tcp_gettaocache(inp)) == NULL) {
- taop = &tao_noncached;
- bzero(taop, sizeof(*taop));
- }
- tcp_dooptions(tp, optp, optlen, th, &to);
- if (iss)
- tp->iss = iss;
- else {
- tp->iss = tcp_new_isn(tp);
- }
- tp->irs = th->th_seq;
- tcp_sendseqinit(tp);
- tcp_rcvseqinit(tp);
- /*
- * Initialization of the tcpcb for transaction;
- * set SND.WND = SEG.WND,
- * initialize CCsend and CCrecv.
- */
- tp->snd_wnd = tiwin; /* initial send-window */
- tp->cc_send = CC_INC(tcp_ccgen);
- tp->cc_recv = to.to_cc;
- /*
- * Perform TAO test on incoming CC (SEG.CC) option, if any.
- * - compare SEG.CC against cached CC from the same host,
- * if any.
- * - if SEG.CC > chached value, SYN must be new and is accepted
- * immediately: save new CC in the cache, mark the socket
- * connected, enter ESTABLISHED state, turn on flag to
- * send a SYN in the next segment.
- * A virtual advertised window is set in rcv_adv to
- * initialize SWS prevention. Then enter normal segment
- * processing: drop SYN, process data and FIN.
- * - otherwise do a normal 3-way handshake.
- */
- if ((to.to_flag & TOF_CC) != 0) {
- if (((tp->t_flags & TF_NOPUSH) != 0) &&
- taop->tao_cc != 0 && CC_GT(to.to_cc, taop->tao_cc)) {
-
- taop->tao_cc = to.to_cc;
- tp->t_starttime = ticks;
- tp->t_state = TCPS_ESTABLISHED;
-
- /*
- * If there is a FIN, or if there is data and the
- * connection is local, then delay SYN,ACK(SYN) in
- * the hope of piggy-backing it on a response
- * segment. Otherwise must send ACK now in case
- * the other side is slow starting.
- */
- if (DELAY_ACK(tp) && ((thflags & TH_FIN) ||
- (tlen != 0 &&
-#ifdef INET6
- ((isipv6 && in6_localaddr(&inp->in6p_faddr))
- ||
- (!isipv6 &&
-#endif
- in_localaddr(inp->inp_faddr)
-#ifdef INET6
- ))
-#endif
- ))) {
- callout_reset(tp->tt_delack, tcp_delacktime,
- tcp_timer_delack, tp);
- tp->t_flags |= TF_NEEDSYN;
- } else
- tp->t_flags |= (TF_ACKNOW | TF_NEEDSYN);
-
- /*
- * Limit the `virtual advertised window' to TCP_MAXWIN
- * here. Even if we requested window scaling, it will
- * become effective only later when our SYN is acked.
- */
- tp->rcv_adv += min(tp->rcv_wnd, TCP_MAXWIN);
- tcpstat.tcps_connects++;
- soisconnected(so);
- callout_reset(tp->tt_keep, tcp_keepinit,
- tcp_timer_keep, tp);
- dropsocket = 0; /* committed to socket */
- tcpstat.tcps_accepts++;
- goto trimthenstep6;
- }
- /* else do standard 3-way handshake */
- } else {
- /*
- * No CC option, but maybe CC.NEW:
- * invalidate cached value.
- */
- taop->tao_cc = 0;
- }
- /*
- * TAO test failed or there was no CC option,
- * do a standard 3-way handshake.
- */
- tp->t_flags |= TF_ACKNOW;
- tp->t_state = TCPS_SYN_RECEIVED;
- callout_reset(tp->tt_keep, tcp_keepinit, tcp_timer_keep, tp);
- dropsocket = 0; /* committed to socket */
- tcpstat.tcps_accepts++;
- goto trimthenstep6;
- }
-
- /*
* If the state is SYN_RECEIVED:
* if seg contains an ACK, but not for our SYN/ACK, send a RST.
*/
@@ -1253,7 +1074,7 @@ findpcb:
* continue processing rest of data/controls, beginning with URG
*/
case TCPS_SYN_SENT:
- if ((taop = tcp_gettaocache(inp)) == NULL) {
+ if ((taop = tcp_gettaocache(&inp->inp_inc)) == NULL) {
taop = &tao_noncached;
bzero(taop, sizeof(*taop));
}
@@ -1297,7 +1118,7 @@ findpcb:
* by the old rules. If no CC.ECHO option, make sure
* we don't get fooled into using T/TCP.
*/
- if (to.to_flag & TOF_CCECHO) {
+ if (to.to_flags & TOF_CCECHO) {
if (tp->cc_send != to.to_ccecho) {
if (taop->tao_ccsent != 0)
goto drop;
@@ -1359,7 +1180,7 @@ findpcb:
*/
tp->t_flags |= TF_ACKNOW;
callout_stop(tp->tt_rexmt);
- if (to.to_flag & TOF_CC) {
+ if (to.to_flags & TOF_CC) {
if (taop->tao_cc != 0 &&
CC_GT(to.to_cc, taop->tao_cc)) {
/*
@@ -1434,7 +1255,7 @@ trimthenstep6:
case TCPS_CLOSING:
case TCPS_TIME_WAIT:
if ((thflags & TH_SYN) &&
- (to.to_flag & TOF_CC) && tp->cc_recv != 0) {
+ (to.to_flags & TOF_CC) && tp->cc_recv != 0) {
if (tp->t_state == TCPS_TIME_WAIT &&
(ticks - tp->t_starttime) > tcp_msl) {
rstreason = BANDLIM_UNLIMITED;
@@ -1542,7 +1363,7 @@ trimthenstep6:
* RFC 1323 PAWS: If we have a timestamp reply on this segment
* and it's less than ts_recent, drop it.
*/
- if ((to.to_flag & TOF_TS) != 0 && tp->ts_recent &&
+ if ((to.to_flags & TOF_TS) != 0 && tp->ts_recent &&
TSTMP_LT(to.to_tsval, tp->ts_recent)) {
/* Check to see if ts_recent is over 24 days old. */
@@ -1574,7 +1395,7 @@ trimthenstep6:
* RST segments do not have to comply with this.
*/
if ((tp->t_flags & (TF_REQ_CC|TF_RCVD_CC)) == (TF_REQ_CC|TF_RCVD_CC) &&
- ((to.to_flag & TOF_CC) == 0 || tp->cc_recv != to.to_cc))
+ ((to.to_flags & TOF_CC) == 0 || tp->cc_recv != to.to_cc))
goto dropafterack;
/*
@@ -1694,7 +1515,7 @@ trimthenstep6:
* NOTE that the test is modified according to the latest
* proposal of the tcplw@cray.com list (Braden 1993/04/26).
*/
- if ((to.to_flag & TOF_TS) != 0 &&
+ if ((to.to_flags & TOF_TS) != 0 &&
SEQ_LEQ(th->th_seq, tp->last_ack_sent)) {
tp->ts_recent_age = ticks;
tp->ts_recent = to.to_tsval;
@@ -1748,7 +1569,7 @@ trimthenstep6:
* update cache.CC if it was undefined, pass any queued
* data to the user, and advance state appropriately.
*/
- if ((taop = tcp_gettaocache(inp)) != NULL &&
+ if ((taop = tcp_gettaocache(&inp->inp_inc)) != NULL &&
taop->tao_cc == 0)
taop->tao_cc = tp->cc_recv;
@@ -1940,7 +1761,7 @@ process_ACK:
* timer backoff (cf., Phil Karn's retransmit alg.).
* Recompute the initial retransmit timer.
*/
- if (to.to_flag & TOF_TS)
+ if (to.to_flags & TOF_TS)
tcp_xmit_timer(tp, ticks - to.to_tsecr + 1);
else if (tp->t_rtttime && SEQ_GT(th->th_ack, tp->t_rtseq))
tcp_xmit_timer(tp, ticks - tp->t_rtttime);
@@ -2371,9 +2192,6 @@ dropwithreset:
tcp_respond(tp, mtod(m, void *), th, m, th->th_seq+tlen,
(tcp_seq)0, TH_RST|TH_ACK);
}
- /* destroy temporarily created socket */
- if (dropsocket)
- (void) soabort(so);
return;
drop:
@@ -2386,23 +2204,21 @@ drop:
&tcp_savetcp, 0);
#endif
m_freem(m);
- /* destroy temporarily created socket */
- if (dropsocket)
- (void) soabort(so);
return;
}
+/*
+ * Parse TCP options and place in tcpopt.
+ */
static void
-tcp_dooptions(tp, cp, cnt, th, to)
- struct tcpcb *tp;
+tcp_dooptions(to, cp, cnt, is_syn)
+ struct tcpopt *to;
u_char *cp;
int cnt;
- struct tcphdr *th;
- struct tcpopt *to;
{
- u_short mss = 0;
int opt, optlen;
+ to->to_flags = 0;
for (; cnt > 0; cnt -= optlen, cp += optlen) {
opt = cp[0];
if (opt == TCPOPT_EOL)
@@ -2417,92 +2233,67 @@ tcp_dooptions(tp, cp, cnt, th, to)
break;
}
switch (opt) {
-
- default:
- continue;
-
case TCPOPT_MAXSEG:
if (optlen != TCPOLEN_MAXSEG)
continue;
- if (!(th->th_flags & TH_SYN))
+ if (!is_syn)
continue;
- bcopy((char *) cp + 2, (char *) &mss, sizeof(mss));
- NTOHS(mss);
+ to->to_flags |= TOF_MSS;
+ bcopy((char *)cp + 2,
+ (char *)&to->to_mss, sizeof(to->to_mss));
+ NTOHS(to->to_mss);
break;
-
case TCPOPT_WINDOW:
if (optlen != TCPOLEN_WINDOW)
continue;
- if (!(th->th_flags & TH_SYN))
+ if (! is_syn)
continue;
- tp->t_flags |= TF_RCVD_SCALE;
- tp->requested_s_scale = min(cp[2], TCP_MAX_WINSHIFT);
+ to->to_flags |= TOF_SCALE;
+ to->to_requested_s_scale = min(cp[2], TCP_MAX_WINSHIFT);
break;
-
case TCPOPT_TIMESTAMP:
if (optlen != TCPOLEN_TIMESTAMP)
continue;
- to->to_flag |= TOF_TS;
+ to->to_flags |= TOF_TS;
bcopy((char *)cp + 2,
(char *)&to->to_tsval, sizeof(to->to_tsval));
NTOHL(to->to_tsval);
bcopy((char *)cp + 6,
(char *)&to->to_tsecr, sizeof(to->to_tsecr));
NTOHL(to->to_tsecr);
-
- /*
- * A timestamp received in a SYN makes
- * it ok to send timestamp requests and replies.
- */
- if (th->th_flags & TH_SYN) {
- tp->t_flags |= TF_RCVD_TSTMP;
- tp->ts_recent = to->to_tsval;
- tp->ts_recent_age = ticks;
- }
break;
case TCPOPT_CC:
if (optlen != TCPOLEN_CC)
continue;
- to->to_flag |= TOF_CC;
+ to->to_flags |= TOF_CC;
bcopy((char *)cp + 2,
(char *)&to->to_cc, sizeof(to->to_cc));
NTOHL(to->to_cc);
- /*
- * A CC or CC.new option received in a SYN makes
- * it ok to send CC in subsequent segments.
- */
- if (th->th_flags & TH_SYN)
- tp->t_flags |= TF_RCVD_CC;
break;
case TCPOPT_CCNEW:
if (optlen != TCPOLEN_CC)
continue;
- if (!(th->th_flags & TH_SYN))
+ if (!is_syn)
continue;
- to->to_flag |= TOF_CCNEW;
+ to->to_flags |= TOF_CCNEW;
bcopy((char *)cp + 2,
(char *)&to->to_cc, sizeof(to->to_cc));
NTOHL(to->to_cc);
- /*
- * A CC or CC.new option received in a SYN makes
- * it ok to send CC in subsequent segments.
- */
- tp->t_flags |= TF_RCVD_CC;
break;
case TCPOPT_CCECHO:
if (optlen != TCPOLEN_CC)
continue;
- if (!(th->th_flags & TH_SYN))
+ if (!is_syn)
continue;
- to->to_flag |= TOF_CCECHO;
+ to->to_flags |= TOF_CCECHO;
bcopy((char *)cp + 2,
(char *)&to->to_ccecho, sizeof(to->to_ccecho));
NTOHL(to->to_ccecho);
break;
+ default:
+ continue;
}
}
- if (th->th_flags & TH_SYN)
- tcp_mss(tp, mss); /* sets t_maxseg */
}
/*
@@ -2675,10 +2466,10 @@ tcp_mss(tp, offer)
#endif
#ifdef INET6
if (isipv6)
- rt = tcp_rtlookup6(inp);
+ rt = tcp_rtlookup6(&inp->inp_inc);
else
#endif
- rt = tcp_rtlookup(inp);
+ rt = tcp_rtlookup(&inp->inp_inc);
if (rt == NULL) {
tp->t_maxopd = tp->t_maxseg =
#ifdef INET6
@@ -2884,10 +2675,10 @@ tcp_mssopt(tp)
#endif
#ifdef INET6
if (isipv6)
- rt = tcp_rtlookup6(tp->t_inpcb);
+ rt = tcp_rtlookup6(&tp->t_inpcb->inp_inc);
else
#endif /* INET6 */
- rt = tcp_rtlookup(tp->t_inpcb);
+ rt = tcp_rtlookup(&tp->t_inpcb->inp_inc);
if (rt == NULL)
return
#ifdef INET6
diff --git a/sys/netinet/tcp_subr.c b/sys/netinet/tcp_subr.c
index 6a6c9f8..27d54eb 100644
--- a/sys/netinet/tcp_subr.c
+++ b/sys/netinet/tcp_subr.c
@@ -117,11 +117,11 @@ SYSCTL_INT(_net_inet_tcp, TCPCTL_RTTDFLT, rttdflt, CTLFLAG_RW,
&tcp_rttdflt , 0, "Default maximum TCP Round Trip Time");
#endif
-static int tcp_do_rfc1323 = 1;
+int tcp_do_rfc1323 = 1;
SYSCTL_INT(_net_inet_tcp, TCPCTL_DO_RFC1323, rfc1323, CTLFLAG_RW,
&tcp_do_rfc1323 , 0, "Enable rfc1323 (high performance TCP) extensions");
-static int tcp_do_rfc1644 = 0;
+int tcp_do_rfc1644 = 0;
SYSCTL_INT(_net_inet_tcp, TCPCTL_DO_RFC1644, rfc1644, CTLFLAG_RW,
&tcp_do_rfc1644 , 0, "Enable rfc1644 (TTCP) extensions");
@@ -224,6 +224,8 @@ tcp_init()
if (max_linkhdr + TCP_MINPROTOHDR > MHLEN)
panic("tcp_init");
#undef TCP_MINPROTOHDR
+
+ syncache_init();
}
/*
@@ -710,18 +712,6 @@ tcp_close(tp)
tcpstat.tcps_cachedssthresh++;
}
}
- rt = inp->inp_route.ro_rt;
- if (rt) {
- /*
- * mark route for deletion if no information is
- * cached.
- */
- if ((tp->t_flags & TF_LQ_OVERFLOW) &&
- ((rt->rt_rmx.rmx_locks & RTV_RTT) == 0)){
- if (rt->rt_rmx.rmx_rtt == 0)
- rt->rt_flags |= RTF_DELCLONE;
- }
- }
no_valid_rt:
/* free the reassembly queue, if any */
while((q = LIST_FIRST(&tp->t_segq)) != NULL) {
@@ -1047,6 +1037,17 @@ tcp_ctlinput(cmd, sa, vip)
if (SEQ_GEQ(icmp_seq, tp->snd_una) &&
SEQ_LT(icmp_seq, tp->snd_max))
(*notify)(inp, inetctlerrmap[cmd]);
+ } else {
+ struct in_conninfo inc;
+
+ inc.inc_fport = th->th_dport;
+ inc.inc_lport = th->th_sport;
+ inc.inc_faddr = faddr;
+ inc.inc_laddr = ip->ip_src;
+#ifdef INET6
+ inc.inc_isipv6 = 0;
+#endif
+ syncache_unreach(&inc, th);
}
splx(s);
} else
@@ -1099,6 +1100,7 @@ tcp6_ctlinput(cmd, sa, d)
}
if (ip6) {
+ struct in_conninfo inc;
/*
* XXX: We assume that when IPV6 is non NULL,
* M and OFF are valid.
@@ -1114,6 +1116,13 @@ tcp6_ctlinput(cmd, sa, d)
in6_pcbnotify(&tcb, sa, th.th_dport,
(struct sockaddr *)ip6cp->ip6c_src,
th.th_sport, cmd, notify);
+
+ inc.inc_fport = th.th_dport;
+ inc.inc_lport = th.th_sport;
+ inc.inc6_faddr = ((struct sockaddr_in6 *)sa)->sin6_addr;
+ inc.inc6_laddr = ip6cp->ip6c_src->sin6_addr;
+ inc.inc_isipv6 = 1;
+ syncache_unreach(&inc, &th);
} else
in6_pcbnotify(&tcb, sa, 0, (struct sockaddr *)sa6_src,
0, cmd, notify);
@@ -1271,10 +1280,10 @@ tcp_mtudisc(inp, errno)
if (tp) {
#ifdef INET6
if (isipv6)
- rt = tcp_rtlookup6(inp);
+ rt = tcp_rtlookup6(&inp->inp_inc);
else
#endif /* INET6 */
- rt = tcp_rtlookup(inp);
+ rt = tcp_rtlookup(&inp->inp_inc);
if (!rt || !rt->rt_rmx.rmx_mtu) {
tp->t_maxopd = tp->t_maxseg =
#ifdef INET6
@@ -1351,21 +1360,21 @@ tcp_mtudisc(inp, errno)
* to get the interface MTU.
*/
struct rtentry *
-tcp_rtlookup(inp)
- struct inpcb *inp;
+tcp_rtlookup(inc)
+ struct in_conninfo *inc;
{
struct route *ro;
struct rtentry *rt;
- ro = &inp->inp_route;
+ ro = &inc->inc_route;
rt = ro->ro_rt;
if (rt == NULL || !(rt->rt_flags & RTF_UP)) {
/* No route yet, so try to acquire one */
- if (inp->inp_faddr.s_addr != INADDR_ANY) {
+ if (inc->inc_faddr.s_addr != INADDR_ANY) {
ro->ro_dst.sa_family = AF_INET;
ro->ro_dst.sa_len = sizeof(struct sockaddr_in);
((struct sockaddr_in *) &ro->ro_dst)->sin_addr =
- inp->inp_faddr;
+ inc->inc_faddr;
rtalloc(ro);
rt = ro->ro_rt;
}
@@ -1375,23 +1384,20 @@ tcp_rtlookup(inp)
#ifdef INET6
struct rtentry *
-tcp_rtlookup6(inp)
- struct inpcb *inp;
+tcp_rtlookup6(inc)
+ struct in_conninfo *inc;
{
struct route_in6 *ro6;
struct rtentry *rt;
- ro6 = &inp->in6p_route;
+ ro6 = &inc->inc6_route;
rt = ro6->ro_rt;
if (rt == NULL || !(rt->rt_flags & RTF_UP)) {
/* No route yet, so try to acquire one */
- if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
- struct sockaddr_in6 *dst6;
-
- dst6 = (struct sockaddr_in6 *)&ro6->ro_dst;
- dst6->sin6_family = AF_INET6;
- dst6->sin6_len = sizeof(*dst6);
- dst6->sin6_addr = inp->in6p_faddr;
+ if (!IN6_IS_ADDR_UNSPECIFIED(&inc->inc6_faddr)) {
+ ro6->ro_dst.sin6_family = AF_INET6;
+ ro6->ro_dst.sin6_len = sizeof(struct sockaddr_in6);
+ ro6->ro_dst.sin6_addr = inc->inc6_faddr;
rtalloc((struct route *)ro6);
rt = ro6->ro_rt;
}
@@ -1450,17 +1456,17 @@ ipsec_hdrsiz_tcp(tp)
* the route metrics.
*/
struct rmxp_tao *
-tcp_gettaocache(inp)
- struct inpcb *inp;
+tcp_gettaocache(inc)
+ struct in_conninfo *inc;
{
struct rtentry *rt;
#ifdef INET6
- if ((inp->inp_vflag & INP_IPV6) != 0)
- rt = tcp_rtlookup6(inp);
+ if (inc->inc_isipv6)
+ rt = tcp_rtlookup6(inc);
else
#endif /* INET6 */
- rt = tcp_rtlookup(inp);
+ rt = tcp_rtlookup(inc);
/* Make sure this is a host route and is up. */
if (rt == NULL ||
diff --git a/sys/netinet/tcp_syncache.c b/sys/netinet/tcp_syncache.c
new file mode 100644
index 0000000..73860e0
--- /dev/null
+++ b/sys/netinet/tcp_syncache.c
@@ -0,0 +1,1161 @@
+/*-
+ * Copyright (c) 2001 Networks Associates Technologies, Inc.
+ * All rights reserved.
+ *
+ * This software was developed for the FreeBSD Project by Jonathan Lemon
+ * and NAI Labs, the Security Research Division of Network Associates, Inc.
+ * under DARPA/SPAWAR contract N66001-01-C-8035 ("CBOSS"), as part of the
+ * DARPA CHATS research program.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ * notice, this list of conditions and the following disclaimer in the
+ * documentation and/or other materials provided with the distribution.
+ * 3. The name of the author may not be used to endorse or promote
+ * products derived from this software without specific prior written
+ * permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
+ * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
+ * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+ * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
+ * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
+ * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
+ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
+ * SUCH DAMAGE.
+ *
+ * $FreeBSD$
+ */
+
+#include "opt_inet6.h"
+#include "opt_ipsec.h"
+
+#include <sys/param.h>
+#include <sys/systm.h>
+#include <sys/kernel.h>
+#include <sys/sysctl.h>
+#include <sys/malloc.h>
+#include <sys/mbuf.h>
+#include <sys/md5.h>
+#include <sys/proc.h> /* for proc0 declaration */
+#include <sys/random.h>
+#include <sys/socket.h>
+#include <sys/socketvar.h>
+
+#include <net/if.h>
+#include <net/route.h>
+
+#include <netinet/in.h>
+#include <netinet/in_systm.h>
+#include <netinet/ip.h>
+#include <netinet/in_var.h>
+#include <netinet/in_pcb.h>
+#include <netinet/ip_var.h>
+#ifdef INET6
+#include <netinet/ip6.h>
+#include <netinet/icmp6.h>
+#include <netinet6/nd6.h>
+#include <netinet6/ip6_var.h>
+#include <netinet6/in6_pcb.h>
+#endif
+#include <netinet/tcp.h>
+#include <netinet/tcp_fsm.h>
+#include <netinet/tcp_seq.h>
+#include <netinet/tcp_timer.h>
+#include <netinet/tcp_var.h>
+#ifdef INET6
+#include <netinet6/tcp6_var.h>
+#endif
+
+#ifdef IPSEC
+#include <netinet6/ipsec.h>
+#ifdef INET6
+#include <netinet6/ipsec6.h>
+#endif
+#include <netkey/key.h>
+#endif /*IPSEC*/
+
+#include <machine/in_cksum.h>
+#include <vm/vm_zone.h>
+
+static void syncache_drop(struct syncache *, struct syncache_head *);
+static void syncache_free(struct syncache *);
+static int syncache_insert(struct syncache *, struct syncache_head *);
+struct syncache *syncache_lookup(struct in_conninfo *, struct syncache_head **);
+static int syncache_respond(struct syncache *, struct mbuf *);
+static struct socket *syncache_socket(struct syncache *, struct socket *);
+static void syncache_timer(void *);
+
+/*
+ * Transmit the SYN,ACK fewer times than TCP_MAXRXTSHIFT specifies.
+ * 3 retransmits corresponds to a timeout of (1 + 2 + 4 + 8 == 15) seconds,
+ * the odds are that the user has given up attempting to connect by then.
+ */
+#define SYNCACHE_MAXREXMTS 3
+
+/* Arbitrary values */
+#define TCP_SYNCACHE_HASHSIZE 512
+#define TCP_SYNCACHE_BUCKETLIMIT 30
+
+struct tcp_syncache {
+ struct syncache_head *hashbase;
+ struct vm_zone *zone;
+ u_int hashsize;
+ u_int hashmask;
+ u_int bucket_limit;
+ u_int cache_count;
+ u_int cache_limit;
+ u_int rexmt_limit;
+ u_int hash_secret;
+ u_int next_reseed;
+ TAILQ_HEAD(, syncache) timerq[SYNCACHE_MAXREXMTS + 1];
+ struct callout tt_timerq[SYNCACHE_MAXREXMTS + 1];
+};
+static struct tcp_syncache tcp_syncache;
+
+SYSCTL_NODE(_net_inet_tcp, OID_AUTO, syncache, CTLFLAG_RW, 0, "TCP SYN cache");
+
+SYSCTL_INT(_net_inet_tcp_syncache, OID_AUTO, bucketlimit, CTLFLAG_RD,
+ &tcp_syncache.bucket_limit, 0, "Per-bucket hash limit for syncache");
+
+SYSCTL_INT(_net_inet_tcp_syncache, OID_AUTO, cachelimit, CTLFLAG_RD,
+ &tcp_syncache.cache_limit, 0, "Overall entry limit for syncache");
+
+SYSCTL_INT(_net_inet_tcp_syncache, OID_AUTO, count, CTLFLAG_RD,
+ &tcp_syncache.cache_count, 0, "Current number of entries in syncache");
+
+SYSCTL_INT(_net_inet_tcp_syncache, OID_AUTO, hashsize, CTLFLAG_RD,
+ &tcp_syncache.hashsize, 0, "Size of TCP syncache hashtable");
+
+SYSCTL_INT(_net_inet_tcp_syncache, OID_AUTO, rexmtlimit, CTLFLAG_RW,
+ &tcp_syncache.rexmt_limit, 0, "Limit on SYN/ACK retransmissions");
+
+static MALLOC_DEFINE(M_SYNCACHE, "syncache", "TCP syncache");
+
+#define SYNCACHE_HASH(inc, mask) \
+ ((tcp_syncache.hash_secret ^ \
+ (inc)->inc_faddr.s_addr ^ \
+ ((inc)->inc_faddr.s_addr >> 16) ^ \
+ (inc)->inc_fport ^ (inc)->inc_lport) & mask)
+
+#define SYNCACHE_HASH6(inc, mask) \
+ ((tcp_syncache.hash_secret ^ \
+ (inc)->inc6_faddr.s6_addr32[0] ^ \
+ (inc)->inc6_faddr.s6_addr32[3] ^ \
+ (inc)->inc_fport ^ (inc)->inc_lport) & mask)
+
+#define ENDPTS_EQ(a, b) ( \
+ (a)->ie_fport == (a)->ie_fport && \
+ (a)->ie_lport == (b)->ie_lport && \
+ (a)->ie_faddr.s_addr == (b)->ie_faddr.s_addr && \
+ (a)->ie_laddr.s_addr == (b)->ie_laddr.s_addr \
+)
+
+#define ENDPTS6_EQ(a, b) (memcmp(a, b, sizeof(*a)) == 0)
+
+#define SYNCACHE_TIMEOUT(sc, slot) do { \
+ sc->sc_rxtslot = slot; \
+ sc->sc_rxttime = ticks + TCPTV_RTOBASE * tcp_backoff[slot]; \
+ TAILQ_INSERT_TAIL(&tcp_syncache.timerq[slot], sc, sc_timerq); \
+ if (!callout_active(&tcp_syncache.tt_timerq[slot])) \
+ callout_reset(&tcp_syncache.tt_timerq[slot], \
+ TCPTV_RTOBASE * tcp_backoff[slot], \
+ syncache_timer, (void *)((int)slot)); \
+} while (0)
+
+static void
+syncache_free(struct syncache *sc)
+{
+ struct rtentry *rt;
+
+ if (sc->sc_ipopts)
+ (void) m_free(sc->sc_ipopts);
+#ifdef INET6
+ if (sc->sc_inc.inc_isipv6)
+ rt = sc->sc_route6.ro_rt;
+ else
+#endif
+ rt = sc->sc_route.ro_rt;
+ if (rt != NULL) {
+ /*
+ * If this is the only reference to a protocol cloned
+ * route, remove it immediately.
+ */
+ if (rt->rt_flags & RTF_WASCLONED &&
+ (sc->sc_flags & SCF_KEEPROUTE) == 0 &&
+ rt->rt_refcnt == 1)
+ rtrequest(RTM_DELETE, rt_key(rt),
+ rt->rt_gateway, rt_mask(rt),
+ rt->rt_flags, NULL);
+ RTFREE(rt);
+ }
+ zfree(tcp_syncache.zone, sc);
+}
+
+void
+syncache_init(void)
+{
+ int i;
+
+ tcp_syncache.cache_count = 0;
+ tcp_syncache.hashsize = TCP_SYNCACHE_HASHSIZE;
+ tcp_syncache.bucket_limit = TCP_SYNCACHE_BUCKETLIMIT;
+ tcp_syncache.cache_limit =
+ tcp_syncache.hashsize * tcp_syncache.bucket_limit;
+ tcp_syncache.rexmt_limit = SYNCACHE_MAXREXMTS;
+ tcp_syncache.next_reseed = 0;
+ tcp_syncache.hash_secret = arc4random();
+
+ TUNABLE_INT_FETCH("net.inet.tcp.syncache.hashsize",
+ &tcp_syncache.hashsize);
+ TUNABLE_INT_FETCH("net.inet.tcp.syncache.cachelimit",
+ &tcp_syncache.cache_limit);
+ TUNABLE_INT_FETCH("net.inet.tcp.syncache.bucketlimit",
+ &tcp_syncache.bucket_limit);
+ if (!powerof2(tcp_syncache.hashsize)) {
+ printf("WARNING: syncache hash size is not a power of 2.\n");
+ tcp_syncache.hashsize = 512; /* safe default */
+ }
+ tcp_syncache.hashmask = tcp_syncache.hashsize - 1;
+
+ /* Allocate the hash table. */
+ MALLOC(tcp_syncache.hashbase, struct syncache_head *,
+ tcp_syncache.hashsize * sizeof(struct syncache_head),
+ M_SYNCACHE, M_WAITOK);
+
+ /* Initialize the hash buckets. */
+ for (i = 0; i < tcp_syncache.hashsize; i++) {
+ TAILQ_INIT(&tcp_syncache.hashbase[i].sch_bucket);
+ tcp_syncache.hashbase[i].sch_length = 0;
+ }
+
+ /* Initialize the timer queues. */
+ for (i = 0; i <= TCP_MAXRXTSHIFT; i++) {
+ TAILQ_INIT(&tcp_syncache.timerq[i]);
+ callout_init(&tcp_syncache.tt_timerq[i], 0);
+ }
+
+ /*
+ * Allocate the syncache entries. Allow the zone to allocate one
+ * more entry than cache limit, so a new entry can bump out an
+ * older one.
+ */
+ tcp_syncache.cache_limit -= 1;
+ tcp_syncache.zone = zinit("syncache", sizeof(struct syncache),
+ tcp_syncache.cache_limit, ZONE_INTERRUPT, 0);
+}
+
+static int
+syncache_insert(sc, sch)
+ struct syncache *sc;
+ struct syncache_head *sch;
+{
+ struct syncache *sc2;
+ int s, i;
+
+ /*
+ * Make sure that we don't overflow the per-bucket
+ * limit or the total cache size limit.
+ */
+ s = splnet();
+ if (sch->sch_length >= tcp_syncache.bucket_limit) {
+ /*
+ * The bucket is full, toss the oldest element.
+ */
+ sc2 = TAILQ_FIRST(&sch->sch_bucket);
+ syncache_drop(sc2, sch);
+ tcpstat.tcps_sc_bucketoverflow++;
+ } else if (tcp_syncache.cache_count >= tcp_syncache.cache_limit) {
+ /*
+ * The cache is full. Toss the oldest entry in the
+ * entire cache. This is the front entry in the
+ * first non-empty timer queue with the largest
+ * timeout value.
+ */
+ for (i = SYNCACHE_MAXREXMTS; i >= 0; i--) {
+ sc2 = TAILQ_FIRST(&tcp_syncache.timerq[i]);
+ if (sc2 != NULL)
+ break;
+ }
+ syncache_drop(sc2, NULL);
+ tcpstat.tcps_sc_cacheoverflow++;
+ }
+
+ /* Initialize the entry's timer. */
+ SYNCACHE_TIMEOUT(sc, 0);
+
+ /* Put it into the bucket. */
+ TAILQ_INSERT_TAIL(&sch->sch_bucket, sc, sc_hash);
+ sch->sch_length++;
+ tcp_syncache.cache_count++;
+ tcpstat.tcps_sc_added++;
+ splx(s);
+ return (1);
+}
+
+static void
+syncache_drop(sc, sch)
+ struct syncache *sc;
+ struct syncache_head *sch;
+{
+ int s;
+
+ if (sch == NULL) {
+#ifdef INET6
+ if (sc->sc_inc.inc_isipv6) {
+ sch = &tcp_syncache.hashbase[
+ SYNCACHE_HASH6(&sc->sc_inc, tcp_syncache.hashmask)];
+ } else
+#endif
+ {
+ sch = &tcp_syncache.hashbase[
+ SYNCACHE_HASH(&sc->sc_inc, tcp_syncache.hashmask)];
+ }
+ }
+
+ s = splnet();
+
+ TAILQ_REMOVE(&sch->sch_bucket, sc, sc_hash);
+ sch->sch_length--;
+ tcp_syncache.cache_count--;
+
+ TAILQ_REMOVE(&tcp_syncache.timerq[sc->sc_rxtslot], sc, sc_timerq);
+ if (TAILQ_EMPTY(&tcp_syncache.timerq[sc->sc_rxtslot]))
+ callout_stop(&tcp_syncache.tt_timerq[sc->sc_rxtslot]);
+ splx(s);
+
+ syncache_free(sc);
+}
+
+/*
+ * Walk the timer queues, looking for SYN,ACKs that need to be retransmitted.
+ * If we have retransmitted an entry the maximum number of times, expire it.
+ */
+static void
+syncache_timer(xslot)
+ void *xslot;
+{
+ int slot = (int)xslot;
+ struct syncache *sc, *nsc;
+ struct inpcb *inp;
+ int s;
+
+ s = splnet();
+ if (callout_pending(&tcp_syncache.tt_timerq[slot]) ||
+ !callout_active(&tcp_syncache.tt_timerq[slot])) {
+ splx(s);
+ return;
+ }
+ callout_deactivate(&tcp_syncache.tt_timerq[slot]);
+
+ nsc = TAILQ_FIRST(&tcp_syncache.timerq[slot]);
+ while (nsc != NULL) {
+ if (ticks < nsc->sc_rxttime)
+ break;
+ sc = nsc;
+ nsc = TAILQ_NEXT(sc, sc_timerq);
+ inp = sc->sc_tp->t_inpcb;
+ if (slot == SYNCACHE_MAXREXMTS ||
+ slot >= tcp_syncache.rexmt_limit ||
+ inp->inp_gencnt != sc->sc_inp_gencnt) {
+ syncache_drop(sc, NULL);
+ tcpstat.tcps_sc_stale++;
+ continue;
+ }
+ (void) syncache_respond(sc, NULL);
+ tcpstat.tcps_sc_retransmitted++;
+ TAILQ_REMOVE(&tcp_syncache.timerq[slot], sc, sc_timerq);
+ SYNCACHE_TIMEOUT(sc, slot + 1);
+ }
+ if (nsc != NULL)
+ callout_reset(&tcp_syncache.tt_timerq[slot],
+ nsc->sc_rxttime - ticks, syncache_timer, (void *)(slot));
+ splx(s);
+}
+
+/*
+ * Find an entry in the syncache.
+ */
+struct syncache *
+syncache_lookup(inc, schp)
+ struct in_conninfo *inc;
+ struct syncache_head **schp;
+{
+ struct syncache *sc;
+ struct syncache_head *sch;
+ int s;
+
+#ifdef INET6
+ if (inc->inc_isipv6) {
+ sch = &tcp_syncache.hashbase[
+ SYNCACHE_HASH6(inc, tcp_syncache.hashmask)];
+ *schp = sch;
+ s = splnet();
+ TAILQ_FOREACH(sc, &sch->sch_bucket, sc_hash) {
+ if (ENDPTS6_EQ(&inc->inc_ie, &sc->sc_inc.inc_ie)) {
+ splx(s);
+ return (sc);
+ }
+ }
+ splx(s);
+ } else
+#endif
+ {
+ sch = &tcp_syncache.hashbase[
+ SYNCACHE_HASH(inc, tcp_syncache.hashmask)];
+ *schp = sch;
+ s = splnet();
+ TAILQ_FOREACH(sc, &sch->sch_bucket, sc_hash) {
+#ifdef INET6
+ if (sc->sc_inc.inc_isipv6)
+ continue;
+#endif
+ if (ENDPTS_EQ(&inc->inc_ie, &sc->sc_inc.inc_ie)) {
+ splx(s);
+ return (sc);
+ }
+ }
+ splx(s);
+ }
+ return (NULL);
+}
+
+/*
+ * This function is called when we get a RST for a
+ * non-existent connection, so that we can see if the
+ * connection is in the syn cache. If it is, zap it.
+ */
+void
+syncache_chkrst(inc, th)
+ struct in_conninfo *inc;
+ struct tcphdr *th;
+{
+ struct syncache *sc;
+ struct syncache_head *sch;
+
+ sc = syncache_lookup(inc, &sch);
+ if (sc == NULL)
+ return;
+ /*
+ * If the RST bit is set, check the sequence number to see
+ * if this is a valid reset segment.
+ * RFC 793 page 37:
+ * In all states except SYN-SENT, all reset (RST) segments
+ * are validated by checking their SEQ-fields. A reset is
+ * valid if its sequence number is in the window.
+ *
+ * The sequence number in the reset segment is normally an
+ * echo of our outgoing acknowlegement numbers, but some hosts
+ * send a reset with the sequence number at the rightmost edge
+ * of our receive window, and we have to handle this case.
+ */
+ if (SEQ_GEQ(th->th_seq, sc->sc_irs) &&
+ SEQ_LEQ(th->th_seq, sc->sc_irs + sc->sc_wnd)) {
+ syncache_drop(sc, sch);
+ tcpstat.tcps_sc_reset++;
+ }
+}
+
+void
+syncache_badack(inc)
+ struct in_conninfo *inc;
+{
+ struct syncache *sc;
+ struct syncache_head *sch;
+
+ sc = syncache_lookup(inc, &sch);
+ if (sc != NULL) {
+ syncache_drop(sc, sch);
+ tcpstat.tcps_sc_badack++;
+ }
+}
+
+void
+syncache_unreach(inc, th)
+ struct in_conninfo *inc;
+ struct tcphdr *th;
+{
+ struct syncache *sc;
+ struct syncache_head *sch;
+
+ /* we are called at splnet() here */
+ sc = syncache_lookup(inc, &sch);
+ if (sc == NULL)
+ return;
+
+ /* If the sequence number != sc_iss, then it's a bogus ICMP msg */
+ if (ntohl(th->th_seq) != sc->sc_iss)
+ return;
+
+ /*
+ * If we've rertransmitted 3 times and this is our second error,
+ * we remove the entry. Otherwise, we allow it to continue on.
+ * This prevents us from incorrectly nuking an entry during a
+ * spurious network outage.
+ *
+ * See tcp_notify().
+ */
+ if ((sc->sc_flags & SCF_UNREACH) == 0 || sc->sc_rxtslot < 3) {
+ sc->sc_flags |= SCF_UNREACH;
+ return;
+ }
+ syncache_drop(sc, sch);
+ tcpstat.tcps_sc_unreach++;
+}
+
+/*
+ * Build a new TCP socket structure from a syncache entry.
+ */
+static struct socket *
+syncache_socket(sc, lso)
+ struct syncache *sc;
+ struct socket *lso;
+{
+ struct inpcb *inp = NULL;
+ struct socket *so;
+ struct tcpcb *tp;
+
+ /*
+ * Ok, create the full blown connection, and set things up
+ * as they would have been set up if we had created the
+ * connection when the SYN arrived. If we can't create
+ * the connection, abort it.
+ */
+ so = sonewconn(lso, SS_ISCONNECTED);
+ if (so == NULL) {
+ /*
+ * Drop the connection; we will send a RST if the peer
+ * retransmits the ACK,
+ */
+ tcpstat.tcps_listendrop++;
+ goto abort;
+ }
+
+ inp = sotoinpcb(so);
+
+ /*
+ * Insert new socket into hash list.
+ */
+#ifdef INET6
+ if (sc->sc_inc.inc_isipv6) {
+ inp->in6p_laddr = sc->sc_inc.inc6_laddr;
+ } else {
+ inp->inp_vflag &= ~INP_IPV6;
+ inp->inp_vflag |= INP_IPV4;
+#endif
+ inp->inp_laddr = sc->sc_inc.inc_laddr;
+#ifdef INET6
+ }
+#endif
+ inp->inp_lport = sc->sc_inc.inc_lport;
+ if (in_pcbinshash(inp) != 0) {
+ /*
+ * Undo the assignments above if we failed to
+ * put the PCB on the hash lists.
+ */
+#ifdef INET6
+ if (sc->sc_inc.inc_isipv6)
+ inp->in6p_laddr = in6addr_any;
+ else
+#endif
+ inp->inp_laddr.s_addr = INADDR_ANY;
+ inp->inp_lport = 0;
+ goto abort;
+ }
+#ifdef IPSEC
+ /* copy old policy into new socket's */
+ if (ipsec_copy_policy(sotoinpcb(lso)->inp_sp, inp->inp_sp))
+ printf("syncache_expand: could not copy policy\n");
+#endif
+#ifdef INET6
+ if (sc->sc_inc.inc_isipv6) {
+ struct inpcb *oinp = sotoinpcb(lso);
+ struct in6_addr laddr6;
+ struct sockaddr_in6 *sin6;
+ /*
+ * Inherit socket options from the listening socket.
+ * Note that in6p_inputopts are not (and should not be)
+ * copied, since it stores previously received options and is
+ * used to detect if each new option is different than the
+ * previous one and hence should be passed to a user.
+ * If we copied in6p_inputopts, a user would not be able to
+ * receive options just after calling the accept system call.
+ */
+ inp->inp_flags |= oinp->inp_flags & INP_CONTROLOPTS;
+ if (oinp->in6p_outputopts)
+ inp->in6p_outputopts =
+ ip6_copypktopts(oinp->in6p_outputopts, M_NOWAIT);
+ inp->in6p_route = sc->sc_route6;
+ sc->sc_route6.ro_rt = NULL;
+
+ MALLOC(sin6, struct sockaddr_in6 *, sizeof *sin6,
+ M_SONAME, M_NOWAIT);
+ if (sin6 == NULL)
+ goto abort;
+ bzero(sin6, sizeof(*sin6));
+ sin6->sin6_family = AF_INET6;
+ sin6->sin6_len = sizeof(*sin6);
+ sin6->sin6_addr = sc->sc_inc.inc6_faddr;
+ sin6->sin6_port = sc->sc_inc.inc_fport;
+ laddr6 = inp->in6p_laddr;
+ if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr))
+ inp->in6p_laddr = sc->sc_inc.inc6_laddr;
+ if (in6_pcbconnect(inp, (struct sockaddr *)sin6, thread0)) {
+ inp->in6p_laddr = laddr6;
+ FREE(sin6, M_SONAME);
+ goto abort;
+ }
+ FREE(sin6, M_SONAME);
+ } else
+#endif
+ {
+ struct in_addr laddr;
+ struct sockaddr_in *sin;
+
+ inp->inp_options = ip_srcroute();
+ if (inp->inp_options == NULL) {
+ inp->inp_options = sc->sc_ipopts;
+ sc->sc_ipopts = NULL;
+ }
+ inp->inp_route = sc->sc_route;
+ sc->sc_route.ro_rt = NULL;
+
+ MALLOC(sin, struct sockaddr_in *, sizeof *sin,
+ M_SONAME, M_NOWAIT);
+ if (sin == NULL)
+ goto abort;
+ sin->sin_family = AF_INET;
+ sin->sin_len = sizeof(*sin);
+ sin->sin_addr = sc->sc_inc.inc_faddr;
+ sin->sin_port = sc->sc_inc.inc_fport;
+ bzero((caddr_t)sin->sin_zero, sizeof(sin->sin_zero));
+ laddr = inp->inp_laddr;
+ if (inp->inp_laddr.s_addr == INADDR_ANY)
+ inp->inp_laddr = sc->sc_inc.inc_laddr;
+ if (in_pcbconnect(inp, (struct sockaddr *)sin, thread0)) {
+ inp->inp_laddr = laddr;
+ FREE(sin, M_SONAME);
+ goto abort;
+ }
+ FREE(sin, M_SONAME);
+ }
+
+ tp = intotcpcb(inp);
+ tp->t_state = TCPS_SYN_RECEIVED;
+ tp->iss = sc->sc_iss;
+ tp->irs = sc->sc_irs;
+ tcp_rcvseqinit(tp);
+ tcp_sendseqinit(tp);
+ tp->snd_wl1 = sc->sc_irs;
+ tp->rcv_up = sc->sc_irs + 1;
+ tp->rcv_wnd = sc->sc_wnd;
+ tp->rcv_adv += tp->rcv_wnd;
+
+ tp->t_flags = sc->sc_tp->t_flags & TF_NOPUSH;
+ if (sc->sc_flags & SCF_NOOPT)
+ tp->t_flags |= TF_NOOPT;
+ if (sc->sc_flags & SCF_WINSCALE) {
+ tp->t_flags |= TF_REQ_SCALE|TF_RCVD_SCALE;
+ tp->requested_s_scale = sc->sc_requested_s_scale;
+ tp->request_r_scale = sc->sc_request_r_scale;
+ }
+ if (sc->sc_flags & SCF_TIMESTAMP) {
+ tp->t_flags |= TF_REQ_TSTMP|TF_RCVD_TSTMP;
+ tp->ts_recent = sc->sc_tsrecent;
+ tp->ts_recent_age = ticks;
+ }
+ if (sc->sc_flags & SCF_CC) {
+ /*
+ * Initialization of the tcpcb for transaction;
+ * set SND.WND = SEG.WND,
+ * initialize CCsend and CCrecv.
+ */
+ tp->t_flags |= TF_REQ_CC|TF_RCVD_CC;
+ tp->cc_send = sc->sc_cc_send;
+ tp->cc_recv = sc->sc_cc_recv;
+ }
+
+ tcp_mss(tp, sc->sc_peer_mss);
+
+ /*
+ * If the SYN,ACK was retransmitted, reset cwnd to 1 segment.
+ */
+ if (sc->sc_rxtslot != 0)
+ tp->snd_cwnd = tp->t_maxseg;
+ callout_reset(tp->tt_keep, tcp_keepinit, tcp_timer_keep, tp);
+
+ tcpstat.tcps_accepts++;
+ return (so);
+
+abort:
+ if (so != NULL)
+ (void) soabort(so);
+ return (NULL);
+}
+
+/*
+ * This function gets called when we receive an ACK for a
+ * socket in the LISTEN state. We look up the connection
+ * in the syncache, and if its there, we pull it out of
+ * the cache and turn it into a full-blown connection in
+ * the SYN-RECEIVED state.
+ */
+int
+syncache_expand(inc, th, sop, m)
+ struct in_conninfo *inc;
+ struct tcphdr *th;
+ struct socket **sop;
+ struct mbuf *m;
+{
+ struct syncache *sc;
+ struct syncache_head *sch;
+ struct socket *so;
+
+ sc = syncache_lookup(inc, &sch);
+ if (sc == NULL)
+ return (0);
+
+ /*
+ * If seg contains an ACK, but not for our SYN/ACK, send a RST.
+ */
+ if (th->th_ack != sc->sc_iss + 1)
+ return (0);
+
+ so = syncache_socket(sc, *sop);
+ if (so == NULL) {
+#if 0
+resetandabort:
+ /* XXXjlemon check this - is this correct? */
+ (void) tcp_respond(NULL, m, m, th,
+ th->th_seq + tlen, (tcp_seq)0, TH_RST|TH_ACK);
+#endif
+ m_freem(m); /* XXX only needed for above */
+ tcpstat.tcps_sc_aborted++;
+ } else {
+ sc->sc_flags |= SCF_KEEPROUTE;
+ tcpstat.tcps_sc_completed++;
+ }
+ if (sch == NULL)
+ syncache_free(sc);
+ else
+ syncache_drop(sc, sch);
+ *sop = so;
+ return (1);
+}
+
+/*
+ * Given a LISTEN socket and an inbound SYN request, add
+ * this to the syn cache, and send back a segment:
+ * <SEQ=ISS><ACK=RCV_NXT><CTL=SYN,ACK>
+ * to the source.
+ *
+ * IMPORTANT NOTE: We do _NOT_ ACK data that might accompany the SYN.
+ * Doing so would require that we hold onto the data and deliver it
+ * to the application. However, if we are the target of a SYN-flood
+ * DoS attack, an attacker could send data which would eventually
+ * consume all available buffer space if it were ACKed. By not ACKing
+ * the data, we avoid this DoS scenario.
+ */
+int
+syncache_add(inc, to, th, sop, m)
+ struct in_conninfo *inc;
+ struct tcpopt *to;
+ struct tcphdr *th;
+ struct socket **sop;
+ struct mbuf *m;
+{
+ struct tcpcb *tp;
+ struct socket *so;
+ struct syncache *sc = NULL;
+ struct syncache_head *sch;
+ struct mbuf *ipopts = NULL;
+ struct rmxp_tao *taop;
+ int i, s, win;
+
+ so = *sop;
+ tp = sototcpcb(so);
+
+ /*
+ * Remember the IP options, if any.
+ */
+#ifdef INET6
+ if (!inc->inc_isipv6)
+#endif
+ ipopts = ip_srcroute();
+
+ /*
+ * See if we already have an entry for this connection.
+ * If we do, resend the SYN,ACK, and reset the retransmit timer.
+ *
+ * XXX
+ * should the syncache be re-initialized with the contents
+ * of the new SYN here (which may have different options?)
+ */
+ sc = syncache_lookup(inc, &sch);
+ if (sc != NULL) {
+ tcpstat.tcps_sc_dupsyn++;
+ if (ipopts) {
+ /*
+ * If we were remembering a previous source route,
+ * forget it and use the new one we've been given.
+ */
+ if (sc->sc_ipopts)
+ (void) m_free(sc->sc_ipopts);
+ sc->sc_ipopts = ipopts;
+ }
+ /*
+ * Update timestamp if present.
+ */
+ if (sc->sc_flags & SCF_TIMESTAMP)
+ sc->sc_tsrecent = to->to_tsval;
+ if (syncache_respond(sc, m) == 0) {
+ s = splnet();
+ TAILQ_REMOVE(&tcp_syncache.timerq[sc->sc_rxtslot],
+ sc, sc_timerq);
+ SYNCACHE_TIMEOUT(sc, sc->sc_rxtslot);
+ splx(s);
+ tcpstat.tcps_sndacks++;
+ tcpstat.tcps_sndtotal++;
+ }
+ *sop = NULL;
+ return (1);
+ }
+
+ sc = zalloc(tcp_syncache.zone);
+ if (sc == NULL) {
+ /*
+ * The zone allocator couldn't provide more entries.
+ * Treat this as if the cache was full; drop the oldest
+ * entry and insert the new one.
+ */
+ s = splnet();
+ for (i = SYNCACHE_MAXREXMTS; i >= 0; i--) {
+ sc = TAILQ_FIRST(&tcp_syncache.timerq[i]);
+ if (sc != NULL)
+ break;
+ }
+ syncache_drop(sc, NULL);
+ splx(s);
+ tcpstat.tcps_sc_zonefail++;
+ sc = zalloc(tcp_syncache.zone);
+ if (sc == NULL) {
+ if (ipopts)
+ (void) m_free(ipopts);
+ return (0);
+ }
+ }
+
+ /*
+ * Fill in the syncache values.
+ */
+ sc->sc_tp = tp;
+ sc->sc_inp_gencnt = tp->t_inpcb->inp_gencnt;
+ sc->sc_ipopts = ipopts;
+ sc->sc_inc.inc_fport = inc->inc_fport;
+ sc->sc_inc.inc_lport = inc->inc_lport;
+#ifdef INET6
+ sc->sc_inc.inc_isipv6 = inc->inc_isipv6;
+ if (inc->inc_isipv6) {
+ sc->sc_inc.inc6_faddr = inc->inc6_faddr;
+ sc->sc_inc.inc6_laddr = inc->inc6_laddr;
+ sc->sc_route6.ro_rt = NULL;
+ } else
+#endif
+ {
+ sc->sc_inc.inc_faddr = inc->inc_faddr;
+ sc->sc_inc.inc_laddr = inc->inc_laddr;
+ sc->sc_route.ro_rt = NULL;
+ }
+ sc->sc_irs = th->th_seq;
+
+ /* Initial receive window: clip sbspace to [0 .. TCP_MAXWIN] */
+ win = sbspace(&so->so_rcv);
+ win = imax(win, 0);
+ win = imin(win, TCP_MAXWIN);
+ sc->sc_wnd = win;
+
+ sc->sc_flags = 0;
+ sc->sc_peer_mss = to->to_flags & TOF_MSS ? to->to_mss : 0;
+ if (tcp_do_rfc1323) {
+ /*
+ * A timestamp received in a SYN makes
+ * it ok to send timestamp requests and replies.
+ */
+ if (to->to_flags & TOF_TS) {
+ sc->sc_tsrecent = to->to_tsval;
+ sc->sc_flags |= SCF_TIMESTAMP;
+ }
+ if (to->to_flags & TOF_SCALE) {
+ int wscale = 0;
+
+ /* Compute proper scaling value from buffer space */
+ while (wscale < TCP_MAX_WINSHIFT &&
+ (TCP_MAXWIN << wscale) < so->so_rcv.sb_hiwat)
+ wscale++;
+ sc->sc_request_r_scale = wscale;
+ sc->sc_requested_s_scale = to->to_requested_s_scale;
+ sc->sc_flags |= SCF_WINSCALE;
+ }
+ }
+ if (tcp_do_rfc1644) {
+ /*
+ * A CC or CC.new option received in a SYN makes
+ * it ok to send CC in subsequent segments.
+ */
+ if (to->to_flags & (TOF_CC|TOF_CCNEW)) {
+ sc->sc_cc_recv = to->to_cc;
+ sc->sc_cc_send = CC_INC(tcp_ccgen);
+ sc->sc_flags |= SCF_CC;
+ }
+ }
+ if (tp->t_flags & TF_NOOPT)
+ sc->sc_flags = SCF_NOOPT;
+
+ /*
+ * XXX
+ * We have the option here of not doing TAO (even if the segment
+ * qualifies) and instead fall back to a normal 3WHS via the syncache.
+ * This allows us to apply synflood protection to TAO-qualifying SYNs
+ * also. However, there should be a hueristic to determine when to
+ * do this, and is not present at the moment.
+ */
+
+ /*
+ * Perform TAO test on incoming CC (SEG.CC) option, if any.
+ * - compare SEG.CC against cached CC from the same host, if any.
+ * - if SEG.CC > chached value, SYN must be new and is accepted
+ * immediately: save new CC in the cache, mark the socket
+ * connected, enter ESTABLISHED state, turn on flag to
+ * send a SYN in the next segment.
+ * A virtual advertised window is set in rcv_adv to
+ * initialize SWS prevention. Then enter normal segment
+ * processing: drop SYN, process data and FIN.
+ * - otherwise do a normal 3-way handshake.
+ */
+ taop = tcp_gettaocache(&sc->sc_inc);
+ if ((to->to_flags & TOF_CC) != 0) {
+ if (((tp->t_flags & TF_NOPUSH) != 0) &&
+ sc->sc_flags & SCF_CC &&
+ taop != NULL && taop->tao_cc != 0 &&
+ CC_GT(to->to_cc, taop->tao_cc)) {
+ sc->sc_rxtslot = 0;
+ so = syncache_socket(sc, *sop);
+ if (so != NULL) {
+ sc->sc_flags |= SCF_KEEPROUTE;
+ taop->tao_cc = to->to_cc;
+ *sop = so;
+ }
+ syncache_free(sc);
+ return (so != NULL);
+ }
+ } else {
+ /*
+ * No CC option, but maybe CC.NEW: invalidate cached value.
+ */
+ if (taop != NULL)
+ taop->tao_cc = 0;
+ }
+ /*
+ * TAO test failed or there was no CC option,
+ * do a standard 3-way handshake.
+ */
+ sc->sc_iss = arc4random();
+ if (syncache_insert(sc, sch)) {
+ if (syncache_respond(sc, m) == 0) {
+ tcpstat.tcps_sndacks++;
+ tcpstat.tcps_sndtotal++;
+ } else {
+ syncache_drop(sc, sch);
+ tcpstat.tcps_sc_dropped++;
+ }
+ } else {
+ syncache_free(sc);
+ }
+ *sop = NULL;
+ return (1);
+}
+
+static int
+syncache_respond(sc, m)
+ struct syncache *sc;
+ struct mbuf *m;
+{
+ u_int8_t *optp;
+ int optlen, error;
+ u_int16_t tlen, hlen, mssopt;
+ struct ip *ip = NULL;
+ struct rtentry *rt;
+ struct tcphdr *th;
+#ifdef INET6
+ struct ip6_hdr *ip6 = NULL;
+#endif
+
+#ifdef INET6
+ if (sc->sc_inc.inc_isipv6) {
+ rt = tcp_rtlookup6(&sc->sc_inc);
+ if (rt != NULL)
+ mssopt = rt->rt_ifp->if_mtu -
+ (sizeof(struct ip6_hdr) + sizeof(struct tcphdr));
+ else
+ mssopt = tcp_v6mssdflt;
+ hlen = sizeof(struct ip6_hdr);
+ } else
+#endif
+ {
+ rt = tcp_rtlookup(&sc->sc_inc);
+ if (rt != NULL)
+ mssopt = rt->rt_ifp->if_mtu -
+ (sizeof(struct ip) + sizeof(struct tcphdr));
+ else
+ mssopt = tcp_mssdflt;
+ hlen = sizeof(struct ip);
+ }
+
+ /* Compute the size of the TCP options. */
+ if (sc->sc_flags & SCF_NOOPT) {
+ optlen = 0;
+ } else {
+ optlen = TCPOLEN_MAXSEG +
+ ((sc->sc_flags & SCF_WINSCALE) ? 4 : 0) +
+ ((sc->sc_flags & SCF_TIMESTAMP) ? TCPOLEN_TSTAMP_APPA : 0) +
+ ((sc->sc_flags & SCF_CC) ? TCPOLEN_CC_APPA * 2 : 0);
+ }
+ tlen = hlen + sizeof(struct tcphdr) + optlen;
+
+ /*
+ * XXX
+ * assume that the entire packet will fit in a header mbuf
+ */
+ KASSERT(max_linkhdr + tlen <= MHLEN, ("syncache: mbuf too small"));
+
+ /*
+ * XXX shouldn't this reuse the mbuf if possible ?
+ * Create the IP+TCP header from scratch.
+ */
+ if (m)
+ m_freem(m);
+
+ m = m_gethdr(M_DONTWAIT, MT_HEADER);
+ if (m == NULL)
+ return (ENOBUFS);
+ m->m_data += max_linkhdr;
+ m->m_len = tlen;
+ m->m_pkthdr.len = tlen;
+ m->m_pkthdr.rcvif = NULL;
+
+#ifdef IPSEC
+ /* use IPsec policy on listening socket to send SYN,ACK */
+ if (ipsec_setsocket(m, sc->sc_tp->t_inpcb->inp_socket) != 0) {
+ m_freem(m);
+ return (ENOBUFS);
+ }
+#endif
+
+#ifdef INET6
+ if (sc->sc_inc.inc_isipv6) {
+ ip6 = mtod(m, struct ip6_hdr *);
+ ip6->ip6_vfc = IPV6_VERSION;
+ ip6->ip6_nxt = IPPROTO_TCP;
+ ip6->ip6_src = sc->sc_inc.inc6_laddr;
+ ip6->ip6_dst = sc->sc_inc.inc6_faddr;
+ ip6->ip6_plen = htons(tlen - hlen);
+ /* ip6_hlim is set after checksum */
+ /* ip6_flow = ??? */
+
+ th = (struct tcphdr *)(ip6 + 1);
+ } else
+#endif
+ {
+ ip = mtod(m, struct ip *);
+ ip->ip_v = IPVERSION;
+ ip->ip_hl = sizeof(struct ip) >> 2;
+ ip->ip_tos = 0;
+ ip->ip_len = tlen;
+ ip->ip_id = 0;
+ ip->ip_off = 0;
+ ip->ip_ttl = ip_defttl;
+ ip->ip_sum = 0;
+ ip->ip_p = IPPROTO_TCP;
+ ip->ip_src = sc->sc_inc.inc_laddr;
+ ip->ip_dst = sc->sc_inc.inc_faddr;
+
+ th = (struct tcphdr *)(ip + 1);
+ }
+ th->th_sport = sc->sc_inc.inc_lport;
+ th->th_dport = sc->sc_inc.inc_fport;
+
+ th->th_seq = htonl(sc->sc_iss);
+ th->th_ack = htonl(sc->sc_irs + 1);
+ th->th_off = (sizeof(struct tcphdr) + optlen) >> 2;
+ th->th_x2 = 0;
+ th->th_flags = TH_SYN|TH_ACK;
+ th->th_win = htons(sc->sc_wnd);
+ th->th_urp = 0;
+
+ /* Tack on the TCP options. */
+ if (optlen == 0)
+ goto no_options;
+ optp = (u_int8_t *)(th + 1);
+ *optp++ = TCPOPT_MAXSEG;
+ *optp++ = TCPOLEN_MAXSEG;
+ *optp++ = (mssopt >> 8) & 0xff;
+ *optp++ = mssopt & 0xff;
+
+ if (sc->sc_flags & SCF_WINSCALE) {
+ *((u_int32_t *)optp) = htonl(TCPOPT_NOP << 24 |
+ TCPOPT_WINDOW << 16 | TCPOLEN_WINDOW << 8 |
+ sc->sc_request_r_scale);
+ optp += 4;
+ }
+
+ if (sc->sc_flags & SCF_TIMESTAMP) {
+ u_int32_t *lp = (u_int32_t *)(optp);
+
+ /* Form timestamp option as shown in appendix A of RFC 1323. */
+ *lp++ = htonl(TCPOPT_TSTAMP_HDR);
+ *lp++ = htonl(ticks);
+ *lp = htonl(sc->sc_tsrecent);
+ optp += TCPOLEN_TSTAMP_APPA;
+ }
+
+ /*
+ * Send CC and CC.echo if we received CC from our peer.
+ */
+ if (sc->sc_flags & SCF_CC) {
+ u_int32_t *lp = (u_int32_t *)(optp);
+
+ *lp++ = htonl(TCPOPT_CC_HDR(TCPOPT_CC));
+ *lp++ = htonl(sc->sc_cc_send);
+ *lp++ = htonl(TCPOPT_CC_HDR(TCPOPT_CCECHO));
+ *lp = htonl(sc->sc_cc_recv);
+ optp += TCPOLEN_CC_APPA * 2;
+ }
+no_options:
+
+#ifdef INET6
+ if (sc->sc_inc.inc_isipv6) {
+ struct route_in6 *ro6 = &sc->sc_route6;
+
+ th->th_sum = 0;
+ th->th_sum = in6_cksum(m, IPPROTO_TCP, hlen, tlen - hlen);
+ ip6->ip6_hlim = in6_selecthlim(NULL,
+ ro6->ro_rt ? ro6->ro_rt->rt_ifp : NULL);
+ error = ip6_output(m, NULL, ro6, 0, NULL, NULL);
+ } else
+#endif
+ {
+ th->th_sum = in_pseudo(ip->ip_src.s_addr, ip->ip_dst.s_addr,
+ htons(tlen - hlen + IPPROTO_TCP));
+ m->m_pkthdr.csum_flags = CSUM_TCP;
+ m->m_pkthdr.csum_data = offsetof(struct tcphdr, th_sum);
+ error = ip_output(m, sc->sc_ipopts, &sc->sc_route, 0, NULL);
+ }
+ return (error);
+}
diff --git a/sys/netinet/tcp_timewait.c b/sys/netinet/tcp_timewait.c
index 6a6c9f8..27d54eb 100644
--- a/sys/netinet/tcp_timewait.c
+++ b/sys/netinet/tcp_timewait.c
@@ -117,11 +117,11 @@ SYSCTL_INT(_net_inet_tcp, TCPCTL_RTTDFLT, rttdflt, CTLFLAG_RW,
&tcp_rttdflt , 0, "Default maximum TCP Round Trip Time");
#endif
-static int tcp_do_rfc1323 = 1;
+int tcp_do_rfc1323 = 1;
SYSCTL_INT(_net_inet_tcp, TCPCTL_DO_RFC1323, rfc1323, CTLFLAG_RW,
&tcp_do_rfc1323 , 0, "Enable rfc1323 (high performance TCP) extensions");
-static int tcp_do_rfc1644 = 0;
+int tcp_do_rfc1644 = 0;
SYSCTL_INT(_net_inet_tcp, TCPCTL_DO_RFC1644, rfc1644, CTLFLAG_RW,
&tcp_do_rfc1644 , 0, "Enable rfc1644 (TTCP) extensions");
@@ -224,6 +224,8 @@ tcp_init()
if (max_linkhdr + TCP_MINPROTOHDR > MHLEN)
panic("tcp_init");
#undef TCP_MINPROTOHDR
+
+ syncache_init();
}
/*
@@ -710,18 +712,6 @@ tcp_close(tp)
tcpstat.tcps_cachedssthresh++;
}
}
- rt = inp->inp_route.ro_rt;
- if (rt) {
- /*
- * mark route for deletion if no information is
- * cached.
- */
- if ((tp->t_flags & TF_LQ_OVERFLOW) &&
- ((rt->rt_rmx.rmx_locks & RTV_RTT) == 0)){
- if (rt->rt_rmx.rmx_rtt == 0)
- rt->rt_flags |= RTF_DELCLONE;
- }
- }
no_valid_rt:
/* free the reassembly queue, if any */
while((q = LIST_FIRST(&tp->t_segq)) != NULL) {
@@ -1047,6 +1037,17 @@ tcp_ctlinput(cmd, sa, vip)
if (SEQ_GEQ(icmp_seq, tp->snd_una) &&
SEQ_LT(icmp_seq, tp->snd_max))
(*notify)(inp, inetctlerrmap[cmd]);
+ } else {
+ struct in_conninfo inc;
+
+ inc.inc_fport = th->th_dport;
+ inc.inc_lport = th->th_sport;
+ inc.inc_faddr = faddr;
+ inc.inc_laddr = ip->ip_src;
+#ifdef INET6
+ inc.inc_isipv6 = 0;
+#endif
+ syncache_unreach(&inc, th);
}
splx(s);
} else
@@ -1099,6 +1100,7 @@ tcp6_ctlinput(cmd, sa, d)
}
if (ip6) {
+ struct in_conninfo inc;
/*
* XXX: We assume that when IPV6 is non NULL,
* M and OFF are valid.
@@ -1114,6 +1116,13 @@ tcp6_ctlinput(cmd, sa, d)
in6_pcbnotify(&tcb, sa, th.th_dport,
(struct sockaddr *)ip6cp->ip6c_src,
th.th_sport, cmd, notify);
+
+ inc.inc_fport = th.th_dport;
+ inc.inc_lport = th.th_sport;
+ inc.inc6_faddr = ((struct sockaddr_in6 *)sa)->sin6_addr;
+ inc.inc6_laddr = ip6cp->ip6c_src->sin6_addr;
+ inc.inc_isipv6 = 1;
+ syncache_unreach(&inc, &th);
} else
in6_pcbnotify(&tcb, sa, 0, (struct sockaddr *)sa6_src,
0, cmd, notify);
@@ -1271,10 +1280,10 @@ tcp_mtudisc(inp, errno)
if (tp) {
#ifdef INET6
if (isipv6)
- rt = tcp_rtlookup6(inp);
+ rt = tcp_rtlookup6(&inp->inp_inc);
else
#endif /* INET6 */
- rt = tcp_rtlookup(inp);
+ rt = tcp_rtlookup(&inp->inp_inc);
if (!rt || !rt->rt_rmx.rmx_mtu) {
tp->t_maxopd = tp->t_maxseg =
#ifdef INET6
@@ -1351,21 +1360,21 @@ tcp_mtudisc(inp, errno)
* to get the interface MTU.
*/
struct rtentry *
-tcp_rtlookup(inp)
- struct inpcb *inp;
+tcp_rtlookup(inc)
+ struct in_conninfo *inc;
{
struct route *ro;
struct rtentry *rt;
- ro = &inp->inp_route;
+ ro = &inc->inc_route;
rt = ro->ro_rt;
if (rt == NULL || !(rt->rt_flags & RTF_UP)) {
/* No route yet, so try to acquire one */
- if (inp->inp_faddr.s_addr != INADDR_ANY) {
+ if (inc->inc_faddr.s_addr != INADDR_ANY) {
ro->ro_dst.sa_family = AF_INET;
ro->ro_dst.sa_len = sizeof(struct sockaddr_in);
((struct sockaddr_in *) &ro->ro_dst)->sin_addr =
- inp->inp_faddr;
+ inc->inc_faddr;
rtalloc(ro);
rt = ro->ro_rt;
}
@@ -1375,23 +1384,20 @@ tcp_rtlookup(inp)
#ifdef INET6
struct rtentry *
-tcp_rtlookup6(inp)
- struct inpcb *inp;
+tcp_rtlookup6(inc)
+ struct in_conninfo *inc;
{
struct route_in6 *ro6;
struct rtentry *rt;
- ro6 = &inp->in6p_route;
+ ro6 = &inc->inc6_route;
rt = ro6->ro_rt;
if (rt == NULL || !(rt->rt_flags & RTF_UP)) {
/* No route yet, so try to acquire one */
- if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
- struct sockaddr_in6 *dst6;
-
- dst6 = (struct sockaddr_in6 *)&ro6->ro_dst;
- dst6->sin6_family = AF_INET6;
- dst6->sin6_len = sizeof(*dst6);
- dst6->sin6_addr = inp->in6p_faddr;
+ if (!IN6_IS_ADDR_UNSPECIFIED(&inc->inc6_faddr)) {
+ ro6->ro_dst.sin6_family = AF_INET6;
+ ro6->ro_dst.sin6_len = sizeof(struct sockaddr_in6);
+ ro6->ro_dst.sin6_addr = inc->inc6_faddr;
rtalloc((struct route *)ro6);
rt = ro6->ro_rt;
}
@@ -1450,17 +1456,17 @@ ipsec_hdrsiz_tcp(tp)
* the route metrics.
*/
struct rmxp_tao *
-tcp_gettaocache(inp)
- struct inpcb *inp;
+tcp_gettaocache(inc)
+ struct in_conninfo *inc;
{
struct rtentry *rt;
#ifdef INET6
- if ((inp->inp_vflag & INP_IPV6) != 0)
- rt = tcp_rtlookup6(inp);
+ if (inc->inc_isipv6)
+ rt = tcp_rtlookup6(inc);
else
#endif /* INET6 */
- rt = tcp_rtlookup(inp);
+ rt = tcp_rtlookup(inc);
/* Make sure this is a host route and is up. */
if (rt == NULL ||
diff --git a/sys/netinet/tcp_usrreq.c b/sys/netinet/tcp_usrreq.c
index 2485a91..e3a35d7 100644
--- a/sys/netinet/tcp_usrreq.c
+++ b/sys/netinet/tcp_usrreq.c
@@ -765,7 +765,7 @@ tcp_connect(tp, nam, td)
* Generate a CC value for this connection and
* check whether CC or CCnew should be used.
*/
- if ((taop = tcp_gettaocache(tp->t_inpcb)) == NULL) {
+ if ((taop = tcp_gettaocache(&tp->t_inpcb->inp_inc)) == NULL) {
taop = &tao_noncached;
bzero(taop, sizeof(*taop));
}
@@ -851,7 +851,7 @@ tcp6_connect(tp, nam, td)
* Generate a CC value for this connection and
* check whether CC or CCnew should be used.
*/
- if ((taop = tcp_gettaocache(tp->t_inpcb)) == NULL) {
+ if ((taop = tcp_gettaocache(&tp->t_inpcb->inp_inc)) == NULL) {
taop = &tao_noncached;
bzero(taop, sizeof(*taop));
}
diff --git a/sys/netinet/tcp_var.h b/sys/netinet/tcp_var.h
index 7dc1bac..0f71278 100644
--- a/sys/netinet/tcp_var.h
+++ b/sys/netinet/tcp_var.h
@@ -36,9 +36,14 @@
#ifndef _NETINET_TCP_VAR_H_
#define _NETINET_TCP_VAR_H_
+
+#include <netinet/in_pcb.h> /* needed for in_conninfo, inp_gen_t */
+
/*
* Kernel variables for tcp.
*/
+extern int tcp_do_rfc1323;
+extern int tcp_do_rfc1644;
/* TCP segment queue entry */
struct tseg_qent {
@@ -171,17 +176,56 @@ struct tcpcb {
* to tcp_dooptions.
*/
struct tcpopt {
- u_long to_flag; /* which options are present */
+ u_long to_flags; /* which options are present */
#define TOF_TS 0x0001 /* timestamp */
#define TOF_CC 0x0002 /* CC and CCnew are exclusive */
#define TOF_CCNEW 0x0004
#define TOF_CCECHO 0x0008
- u_long to_tsval;
- u_long to_tsecr;
- tcp_cc to_cc; /* holds CC or CCnew */
- tcp_cc to_ccecho;
+#define TOF_MSS 0x0010
+#define TOF_SCALE 0x0020
+ u_int32_t to_tsval;
+ u_int32_t to_tsecr;
+ tcp_cc to_cc; /* holds CC or CCnew */
+ tcp_cc to_ccecho;
+ u_int16_t to_mss;
+ u_int8_t to_requested_s_scale;
+ u_int8_t to_pad;
+};
+
+struct syncache {
+ inp_gen_t sc_inp_gencnt; /* pointer check */
+ struct tcpcb *sc_tp; /* tcb for listening socket */
+ struct mbuf *sc_ipopts; /* source route */
+ struct in_conninfo sc_inc; /* addresses */
+#define sc_route sc_inc.inc_route
+#define sc_route6 sc_inc.inc6_route
+ u_int32_t sc_tsrecent;
+ tcp_cc sc_cc_send; /* holds CC or CCnew */
+ tcp_cc sc_cc_recv;
+ tcp_seq sc_irs; /* seq from peer */
+ tcp_seq sc_iss; /* our ISS */
+ u_long sc_rxttime; /* retransmit time */
+ u_int16_t sc_rxtslot; /* retransmit counter */
+ u_int16_t sc_peer_mss; /* peer's MSS */
+ u_int16_t sc_wnd; /* advertised window */
+ u_int8_t sc_requested_s_scale:4,
+ sc_request_r_scale:4;
+ u_int8_t sc_flags;
+#define SCF_NOOPT 0x01 /* no TCP options */
+#define SCF_WINSCALE 0x02 /* negotiated window scaling */
+#define SCF_TIMESTAMP 0x04 /* negotiated timestamps */
+#define SCF_CC 0x08 /* negotiated CC */
+#define SCF_UNREACH 0x10 /* icmp unreachable received */
+#define SCF_KEEPROUTE 0x20 /* keep cloned route */
+ TAILQ_ENTRY(syncache) sc_hash;
+ TAILQ_ENTRY(syncache) sc_timerq;
};
+struct syncache_head {
+ TAILQ_HEAD(, syncache) sch_bucket;
+ u_int sch_length;
+};
+
/*
* The TAO cache entry which is stored in the protocol family specific
* portion of the route metrics.
@@ -306,6 +350,22 @@ struct tcpstat {
u_long tcps_badsyn; /* bogus SYN, e.g. premature ACK */
u_long tcps_mturesent; /* resends due to MTU discovery */
u_long tcps_listendrop; /* listen queue overflows */
+
+ u_long tcps_sc_added; /* entry added to syncache */
+ u_long tcps_sc_retransmitted; /* syncache entry was retransmitted */
+ u_long tcps_sc_dupsyn; /* duplicate SYN packet */
+ u_long tcps_sc_dropped; /* could not reply to packet */
+ u_long tcps_sc_completed; /* successful extraction of entry */
+ u_long tcps_sc_bucketoverflow; /* syncache per-bucket limit hit */
+ u_long tcps_sc_cacheoverflow; /* syncache cache limit hit */
+ u_long tcps_sc_reset; /* RST removed entry from syncache */
+ u_long tcps_sc_stale; /* timed out or listen socket gone */
+ u_long tcps_sc_aborted; /* syncache entry aborted */
+ u_long tcps_sc_badack; /* removed due to bad ACK */
+ u_long tcps_sc_unreach; /* ICMP unreachable received */
+ u_long tcps_sc_zonefail; /* zalloc() failed */
+ u_long tcps_sc_sendcookie; /* SYN cookie sent */
+ u_long tcps_sc_recvcookie; /* SYN cookie received */
};
/*
@@ -383,7 +443,7 @@ struct tcpcb *
void tcp_drain __P((void));
void tcp_fasttimo __P((void));
struct rmxp_tao *
- tcp_gettaocache __P((struct inpcb *));
+ tcp_gettaocache __P((struct in_conninfo *));
void tcp_init __P((void));
void tcp_input __P((struct mbuf *, int));
void tcp_mss __P((struct tcpcb *, int));
@@ -397,7 +457,7 @@ void tcp_quench __P((struct inpcb *, int));
void tcp_respond __P((struct tcpcb *, void *,
struct tcphdr *, struct mbuf *, tcp_seq, tcp_seq, int));
struct rtentry *
- tcp_rtlookup __P((struct inpcb *));
+ tcp_rtlookup __P((struct in_conninfo *));
void tcp_setpersist __P((struct tcpcb *));
void tcp_slowtimo __P((void));
struct tcptemp *
@@ -407,6 +467,14 @@ struct tcpcb *
tcp_timers __P((struct tcpcb *, int));
void tcp_trace __P((int, int, struct tcpcb *, void *, struct tcphdr *,
int));
+void syncache_init(void);
+void syncache_unreach(struct in_conninfo *, struct tcphdr *);
+int syncache_expand(struct in_conninfo *, struct tcphdr *,
+ struct socket **, struct mbuf *);
+int syncache_add(struct in_conninfo *, struct tcpopt *,
+ struct tcphdr *, struct socket **, struct mbuf *);
+void syncache_chkrst(struct in_conninfo *, struct tcphdr *);
+void syncache_badack(struct in_conninfo *);
extern struct pr_usrreqs tcp_usrreqs;
extern u_long tcp_sendspace;
diff --git a/sys/netinet6/tcp6_var.h b/sys/netinet6/tcp6_var.h
index e283212..d911167 100644
--- a/sys/netinet6/tcp6_var.h
+++ b/sys/netinet6/tcp6_var.h
@@ -79,7 +79,7 @@ struct ip6_hdr;
void tcp6_ctlinput __P((int, struct sockaddr *, void *));
void tcp6_init __P((void));
int tcp6_input __P((struct mbuf **, int *, int));
-struct rtentry *tcp_rtlookup6 __P((struct inpcb *));
+struct rtentry *tcp_rtlookup6(struct in_conninfo *);
extern struct pr_usrreqs tcp6_usrreqs;
OpenPOWER on IntegriCloud