freebsd-skq/sys/netinet/tcp_timewait.c

1494 lines
39 KiB
C
Raw Normal View History

1994-05-24 10:09:53 +00:00
/*
* Copyright (c) 1982, 1986, 1988, 1990, 1993, 1995
1994-05-24 10:09:53 +00:00
* The Regents of the University of California. All rights reserved.
*
* 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. All advertising materials mentioning features or use of this software
* must display the following acknowledgement:
* This product includes software developed by the University of
* California, Berkeley and its contributors.
* 4. Neither the name of the University nor the names of its contributors
* may be used to endorse or promote products derived from this software
* without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE REGENTS 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 REGENTS 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.
*
* @(#)tcp_subr.c 8.2 (Berkeley) 5/24/95
1999-08-28 01:08:13 +00:00
* $FreeBSD$
1994-05-24 10:09:53 +00:00
*/
1998-01-25 04:23:33 +00:00
#include "opt_compat.h"
#include "opt_inet6.h"
#include "opt_ipsec.h"
1997-09-16 18:36:06 +00:00
#include "opt_tcpdebug.h"
1994-05-24 10:09:53 +00:00
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/callout.h>
#include <sys/kernel.h>
#include <sys/sysctl.h>
1994-05-24 10:09:53 +00:00
#include <sys/malloc.h>
#include <sys/mbuf.h>
#ifdef INET6
#include <sys/domain.h>
#endif
#include <sys/proc.h>
1994-05-24 10:09:53 +00:00
#include <sys/socket.h>
#include <sys/socketvar.h>
#include <sys/protosw.h>
#include <sys/random.h>
#include <vm/vm_zone.h>
1994-05-24 10:09:53 +00:00
#include <net/route.h>
#include <net/if.h>
#define _IP_VHL
1994-05-24 10:09:53 +00:00
#include <netinet/in.h>
#include <netinet/in_systm.h>
#include <netinet/ip.h>
#ifdef INET6
#include <netinet/ip6.h>
#endif
1994-05-24 10:09:53 +00:00
#include <netinet/in_pcb.h>
#ifdef INET6
#include <netinet6/in6_pcb.h>
#endif
#include <netinet/in_var.h>
1994-05-24 10:09:53 +00:00
#include <netinet/ip_var.h>
#ifdef INET6
#include <netinet6/ip6_var.h>
#endif
1994-05-24 10:09:53 +00:00
#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
1994-05-24 10:09:53 +00:00
#include <netinet/tcpip.h>
#ifdef TCPDEBUG
#include <netinet/tcp_debug.h>
#endif
#include <netinet6/ip6protosw.h>
#ifdef IPSEC
#include <netinet6/ipsec.h>
#ifdef INET6
#include <netinet6/ipsec6.h>
#endif
#endif /*IPSEC*/
1994-05-24 10:09:53 +00:00
#include <machine/in_cksum.h>
#include <sys/md5.h>
1994-05-24 10:09:53 +00:00
int tcp_mssdflt = TCP_MSS;
SYSCTL_INT(_net_inet_tcp, TCPCTL_MSSDFLT, mssdflt, CTLFLAG_RW,
&tcp_mssdflt , 0, "Default TCP Maximum Segment Size");
#ifdef INET6
int tcp_v6mssdflt = TCP6_MSS;
SYSCTL_INT(_net_inet_tcp, TCPCTL_V6MSSDFLT, v6mssdflt,
CTLFLAG_RW, &tcp_v6mssdflt , 0,
"Default TCP Maximum Segment Size for IPv6");
#endif
#if 0
static int tcp_rttdflt = TCPTV_SRTTDFLT / PR_SLOWHZ;
SYSCTL_INT(_net_inet_tcp, TCPCTL_RTTDFLT, rttdflt, CTLFLAG_RW,
&tcp_rttdflt , 0, "Default maximum TCP Round Trip Time");
#endif
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");
int tcp_do_rfc1644 = 0;
SYSCTL_INT(_net_inet_tcp, TCPCTL_DO_RFC1644, rfc1644, CTLFLAG_RW,
&tcp_do_rfc1644 , 0, "Enable rfc1644 (TTCP) extensions");
static int tcp_tcbhashsize = 0;
SYSCTL_INT(_net_inet_tcp, OID_AUTO, tcbhashsize, CTLFLAG_RD,
&tcp_tcbhashsize, 0, "Size of TCP control-block hashtable");
static int do_tcpdrain = 1;
SYSCTL_INT(_net_inet_tcp, OID_AUTO, do_tcpdrain, CTLFLAG_RW, &do_tcpdrain, 0,
"Enable tcp_drain routine for extra help when low on mbufs");
SYSCTL_INT(_net_inet_tcp, OID_AUTO, pcbcount, CTLFLAG_RD,
&tcbinfo.ipi_count, 0, "Number of active PCBs");
static int icmp_may_rst = 1;
SYSCTL_INT(_net_inet_tcp, OID_AUTO, icmp_may_rst, CTLFLAG_RW, &icmp_may_rst, 0,
"Certain ICMP unreachable messages may abort connections in SYN_SENT");
We currently does not react to ICMP administratively prohibited messages send by routers when they deny our traffic, this causes a timeout when trying to connect to TCP ports/services on a remote host, which is blocked by routers or firewalls. rfc1122 (Requirements for Internet Hosts) section 3.2.2.1 actually requi re that we treat such a message for a TCP session, that we treat it like if we had recieved a RST. quote begin. A Destination Unreachable message that is received MUST be reported to the transport layer. The transport layer SHOULD use the information appropriately; for example, see Sections 4.1.3.3, 4.2.3.9, and 4.2.4 below. A transport protocol that has its own mechanism for notifying the sender that a port is unreachable (e.g., TCP, which sends RST segments) MUST nevertheless accept an ICMP Port Unreachable for the same purpose. quote end. I've written a small extension that implement this, it also create a sysctl "net.inet.tcp.icmp_admin_prohib_like_rst" to control if this new behaviour is activated. When it's activated (set to 1) we'll treat a ICMP administratively prohibited message (icmp type 3 code 9, 10 and 13) for a TCP sessions, as if we recived a TCP RST, but only if the TCP session is in SYN_SENT state. The reason for only reacting when in SYN_SENT state, is that this will solve the problem, and at the same time minimize the risk of this being abused. I suggest that we enable this new behaviour by default, but it would be a change of current behaviour, so if people prefer to leave it disabled by default, at least for now, this would be ok for me, the attached diff actually have the sysctl set to 0 by default. PR: 23086 Submitted by: Jesper Skriver <jesper@skriver.dk>
2000-12-16 19:42:06 +00:00
static int tcp_strict_rfc1948 = 0;
SYSCTL_INT(_net_inet_tcp, OID_AUTO, strict_rfc1948, CTLFLAG_RW,
&tcp_strict_rfc1948, 0, "Determines if RFC1948 is followed exactly");
static int tcp_isn_reseed_interval = 0;
SYSCTL_INT(_net_inet_tcp, OID_AUTO, isn_reseed_interval, CTLFLAG_RW,
&tcp_isn_reseed_interval, 0, "Seconds between reseeding of ISN secret");
1997-09-16 11:44:05 +00:00
static void tcp_cleartaocache __P((void));
static void tcp_notify __P((struct inpcb *, int));
1994-05-24 10:09:53 +00:00
/*
Improved connection establishment performance by doing local port lookups via a hashed port list. In the new scheme, in_pcblookup() goes away and is replaced by a new routine, in_pcblookup_local() for doing the local port check. Note that this implementation is space inefficient in that the PCB struct is now too large to fit into 128 bytes. I might deal with this in the future by using the new zone allocator, but I wanted these changes to be extensively tested in their current form first. Also: 1) Fixed off-by-one errors in the port lookup loops in in_pcbbind(). 2) Got rid of some unneeded rehashing. Adding a new routine, in_pcbinshash() to do the initialial hash insertion. 3) Renamed in_pcblookuphash() to in_pcblookup_hash() for easier readability. 4) Added a new routine, in_pcbremlists() to remove the PCB from the various hash lists. 5) Added/deleted comments where appropriate. 6) Removed unnecessary splnet() locking. In general, the PCB functions should be called at splnet()...there are unfortunately a few exceptions, however. 7) Reorganized a few structs for better cache line behavior. 8) Killed my TCP_ACK_HACK kludge. It may come back in a different form in the future, however. These changes have been tested on wcarchive for more than a month. In tests done here, connection establishment overhead is reduced by more than 50 times, thus getting rid of one of the major networking scalability problems. Still to do: make tcp_fastimo/tcp_slowtimo scale well for systems with a large number of connections. tcp_fastimo is easy; tcp_slowtimo is difficult. WARNING: Anything that knows about inpcb and tcpcb structs will have to be recompiled; at the very least, this includes netstat(1).
1998-01-27 09:15:13 +00:00
* Target size of TCP PCB hash tables. Must be a power of two.
*
* Note that this can be overridden by the kernel environment
* variable net.inet.tcp.tcbhashsize
*/
#ifndef TCBHASHSIZE
Improved connection establishment performance by doing local port lookups via a hashed port list. In the new scheme, in_pcblookup() goes away and is replaced by a new routine, in_pcblookup_local() for doing the local port check. Note that this implementation is space inefficient in that the PCB struct is now too large to fit into 128 bytes. I might deal with this in the future by using the new zone allocator, but I wanted these changes to be extensively tested in their current form first. Also: 1) Fixed off-by-one errors in the port lookup loops in in_pcbbind(). 2) Got rid of some unneeded rehashing. Adding a new routine, in_pcbinshash() to do the initialial hash insertion. 3) Renamed in_pcblookuphash() to in_pcblookup_hash() for easier readability. 4) Added a new routine, in_pcbremlists() to remove the PCB from the various hash lists. 5) Added/deleted comments where appropriate. 6) Removed unnecessary splnet() locking. In general, the PCB functions should be called at splnet()...there are unfortunately a few exceptions, however. 7) Reorganized a few structs for better cache line behavior. 8) Killed my TCP_ACK_HACK kludge. It may come back in a different form in the future, however. These changes have been tested on wcarchive for more than a month. In tests done here, connection establishment overhead is reduced by more than 50 times, thus getting rid of one of the major networking scalability problems. Still to do: make tcp_fastimo/tcp_slowtimo scale well for systems with a large number of connections. tcp_fastimo is easy; tcp_slowtimo is difficult. WARNING: Anything that knows about inpcb and tcpcb structs will have to be recompiled; at the very least, this includes netstat(1).
1998-01-27 09:15:13 +00:00
#define TCBHASHSIZE 512
#endif
1994-05-24 10:09:53 +00:00
/*
* This is the actual shape of what we allocate using the zone
* allocator. Doing it this way allows us to protect both structures
* using the same generation count, and also eliminates the overhead
* of allocating tcpcbs separately. By hiding the structure here,
* we avoid changing most of the rest of the code (although it needs
* to be changed, eventually, for greater efficiency).
*/
#define ALIGNMENT 32
#define ALIGNM1 (ALIGNMENT - 1)
struct inp_tp {
union {
struct inpcb inp;
char align[(sizeof(struct inpcb) + ALIGNM1) & ~ALIGNM1];
} inp_tp_u;
struct tcpcb tcb;
struct callout inp_tp_rexmt, inp_tp_persist, inp_tp_keep, inp_tp_2msl;
struct callout inp_tp_delack;
};
#undef ALIGNMENT
#undef ALIGNM1
1994-05-24 10:09:53 +00:00
/*
* Tcp initialization
*/
void
tcp_init()
{
int hashsize = TCBHASHSIZE;
tcp_ccgen = 1;
tcp_cleartaocache();
tcp_delacktime = TCPTV_DELACK;
tcp_keepinit = TCPTV_KEEP_INIT;
tcp_keepidle = TCPTV_KEEP_IDLE;
tcp_keepintvl = TCPTV_KEEPINTVL;
tcp_maxpersistidle = TCPTV_KEEP_IDLE;
tcp_msl = TCPTV_MSL;
LIST_INIT(&tcb);
tcbinfo.listhead = &tcb;
TUNABLE_INT_FETCH("net.inet.tcp.tcbhashsize", &hashsize);
if (!powerof2(hashsize)) {
printf("WARNING: TCB hash size not a power of 2\n");
hashsize = 512; /* safe default */
}
tcp_tcbhashsize = hashsize;
tcbinfo.hashbase = hashinit(hashsize, M_PCB, &tcbinfo.hashmask);
tcbinfo.porthashbase = hashinit(hashsize, M_PCB,
&tcbinfo.porthashmask);
tcbinfo.ipi_zone = zinit("tcpcb", sizeof(struct inp_tp), maxsockets,
ZONE_INTERRUPT, 0);
#ifdef INET6
#define TCP_MINPROTOHDR (sizeof(struct ip6_hdr) + sizeof(struct tcphdr))
#else /* INET6 */
#define TCP_MINPROTOHDR (sizeof(struct tcpiphdr))
#endif /* INET6 */
if (max_protohdr < TCP_MINPROTOHDR)
max_protohdr = TCP_MINPROTOHDR;
if (max_linkhdr + TCP_MINPROTOHDR > MHLEN)
1994-05-24 10:09:53 +00:00
panic("tcp_init");
#undef TCP_MINPROTOHDR
syncache_init();
1994-05-24 10:09:53 +00:00
}
/*
* Fill in the IP and TCP headers for an outgoing packet, given the tcpcb.
* tcp_template used to store this data in mbufs, but we now recopy it out
* of the tcpcb each time to conserve mbufs.
1994-05-24 10:09:53 +00:00
*/
void
tcp_fillheaders(tp, ip_ptr, tcp_ptr)
1994-05-24 10:09:53 +00:00
struct tcpcb *tp;
void *ip_ptr;
void *tcp_ptr;
1994-05-24 10:09:53 +00:00
{
struct inpcb *inp = tp->t_inpcb;
struct tcphdr *tcp_hdr = (struct tcphdr *)tcp_ptr;
1994-05-24 10:09:53 +00:00
#ifdef INET6
if ((inp->inp_vflag & INP_IPV6) != 0) {
struct ip6_hdr *ip6;
ip6 = (struct ip6_hdr *)ip_ptr;
ip6->ip6_flow = (ip6->ip6_flow & ~IPV6_FLOWINFO_MASK) |
(inp->in6p_flowinfo & IPV6_FLOWINFO_MASK);
ip6->ip6_vfc = (ip6->ip6_vfc & ~IPV6_VERSION_MASK) |
(IPV6_VERSION & IPV6_VERSION_MASK);
ip6->ip6_nxt = IPPROTO_TCP;
ip6->ip6_plen = sizeof(struct tcphdr);
ip6->ip6_src = inp->in6p_laddr;
ip6->ip6_dst = inp->in6p_faddr;
tcp_hdr->th_sum = 0;
} else
#endif
{
struct ip *ip = (struct ip *) ip_ptr;
ip->ip_vhl = IP_VHL_BORING;
ip->ip_tos = 0;
ip->ip_len = 0;
ip->ip_id = 0;
ip->ip_off = 0;
ip->ip_ttl = 0;
ip->ip_sum = 0;
ip->ip_p = IPPROTO_TCP;
ip->ip_src = inp->inp_laddr;
ip->ip_dst = inp->inp_faddr;
tcp_hdr->th_sum = in_pseudo(ip->ip_src.s_addr, ip->ip_dst.s_addr,
htons(sizeof(struct tcphdr) + IPPROTO_TCP));
}
tcp_hdr->th_sport = inp->inp_lport;
tcp_hdr->th_dport = inp->inp_fport;
tcp_hdr->th_seq = 0;
tcp_hdr->th_ack = 0;
tcp_hdr->th_x2 = 0;
tcp_hdr->th_off = 5;
tcp_hdr->th_flags = 0;
tcp_hdr->th_win = 0;
tcp_hdr->th_urp = 0;
}
/*
* Create template to be used to send tcp packets on a connection.
* Allocates an mbuf and fills in a skeletal tcp/ip header. The only
* use for this function is in keepalives, which use tcp_respond.
*/
struct tcptemp *
tcp_maketemplate(tp)
struct tcpcb *tp;
{
struct mbuf *m;
struct tcptemp *n;
m = m_get(M_DONTWAIT, MT_HEADER);
if (m == NULL)
return (0);
m->m_len = sizeof(struct tcptemp);
n = mtod(m, struct tcptemp *);
tcp_fillheaders(tp, (void *)&n->tt_ipgen, (void *)&n->tt_t);
1994-05-24 10:09:53 +00:00
return (n);
}
/*
* Send a single message to the TCP at address specified by
* the given TCP/IP header. If m == 0, then we make a copy
* of the tcpiphdr at ti and send directly to the addressed host.
* This is used to force keep alive messages out using the TCP
* template for a connection. If flags are given then we send
* a message back to the TCP which originated the * segment ti,
* and discard the mbuf containing it and any other attached mbufs.
1994-05-24 10:09:53 +00:00
*
* In any case the ack and sequence number of the transmitted
* segment are as specified by the parameters.
*
* NOTE: If m != NULL, then ti must point to *inside* the mbuf.
1994-05-24 10:09:53 +00:00
*/
void
tcp_respond(tp, ipgen, th, m, ack, seq, flags)
1994-05-24 10:09:53 +00:00
struct tcpcb *tp;
void *ipgen;
register struct tcphdr *th;
1994-05-24 10:09:53 +00:00
register struct mbuf *m;
tcp_seq ack, seq;
int flags;
{
register int tlen;
int win = 0;
struct route *ro = 0;
struct route sro;
struct ip *ip;
struct tcphdr *nth;
#ifdef INET6
struct route_in6 *ro6 = 0;
struct route_in6 sro6;
struct ip6_hdr *ip6;
int isipv6;
#endif /* INET6 */
int ipflags = 0;
#ifdef INET6
isipv6 = IP_VHL_V(((struct ip *)ipgen)->ip_vhl) == 6;
ip6 = ipgen;
#endif /* INET6 */
ip = ipgen;
1994-05-24 10:09:53 +00:00
if (tp) {
if (!(flags & TH_RST)) {
win = sbspace(&tp->t_inpcb->inp_socket->so_rcv);
if (win > (long)TCP_MAXWIN << tp->rcv_scale)
win = (long)TCP_MAXWIN << tp->rcv_scale;
}
#ifdef INET6
if (isipv6)
ro6 = &tp->t_inpcb->in6p_route;
else
#endif /* INET6 */
1994-05-24 10:09:53 +00:00
ro = &tp->t_inpcb->inp_route;
} else {
#ifdef INET6
if (isipv6) {
ro6 = &sro6;
bzero(ro6, sizeof *ro6);
} else
#endif /* INET6 */
{
ro = &sro;
bzero(ro, sizeof *ro);
}
1994-05-24 10:09:53 +00:00
}
if (m == 0) {
m = m_gethdr(M_DONTWAIT, MT_HEADER);
if (m == NULL)
return;
tlen = 0;
m->m_data += max_linkhdr;
#ifdef INET6
if (isipv6) {
bcopy((caddr_t)ip6, mtod(m, caddr_t),
sizeof(struct ip6_hdr));
ip6 = mtod(m, struct ip6_hdr *);
nth = (struct tcphdr *)(ip6 + 1);
} else
#endif /* INET6 */
{
bcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip));
ip = mtod(m, struct ip *);
nth = (struct tcphdr *)(ip + 1);
}
bcopy((caddr_t)th, (caddr_t)nth, sizeof(struct tcphdr));
1994-05-24 10:09:53 +00:00
flags = TH_ACK;
} else {
m_freem(m->m_next);
m->m_next = 0;
m->m_data = (caddr_t)ipgen;
/* m_len is set later */
1994-05-24 10:09:53 +00:00
tlen = 0;
#define xchg(a,b,type) { type t; t=a; a=b; b=t; }
#ifdef INET6
if (isipv6) {
xchg(ip6->ip6_dst, ip6->ip6_src, struct in6_addr);
nth = (struct tcphdr *)(ip6 + 1);
} else
#endif /* INET6 */
{
xchg(ip->ip_dst.s_addr, ip->ip_src.s_addr, n_long);
nth = (struct tcphdr *)(ip + 1);
}
if (th != nth) {
/*
* this is usually a case when an extension header
* exists between the IPv6 header and the
* TCP header.
*/
nth->th_sport = th->th_sport;
nth->th_dport = th->th_dport;
}
xchg(nth->th_dport, nth->th_sport, n_short);
1994-05-24 10:09:53 +00:00
#undef xchg
}
#ifdef INET6
if (isipv6) {
ip6->ip6_flow = 0;
ip6->ip6_vfc = IPV6_VERSION;
ip6->ip6_nxt = IPPROTO_TCP;
ip6->ip6_plen = htons((u_short)(sizeof (struct tcphdr) +
tlen));
tlen += sizeof (struct ip6_hdr) + sizeof (struct tcphdr);
} else
#endif
{
1994-05-24 10:09:53 +00:00
tlen += sizeof (struct tcpiphdr);
ip->ip_len = tlen;
ip->ip_ttl = ip_defttl;
}
1994-05-24 10:09:53 +00:00
m->m_len = tlen;
m->m_pkthdr.len = tlen;
m->m_pkthdr.rcvif = (struct ifnet *) 0;
nth->th_seq = htonl(seq);
nth->th_ack = htonl(ack);
nth->th_x2 = 0;
nth->th_off = sizeof (struct tcphdr) >> 2;
nth->th_flags = flags;
1994-05-24 10:09:53 +00:00
if (tp)
nth->th_win = htons((u_short) (win >> tp->rcv_scale));
1994-05-24 10:09:53 +00:00
else
nth->th_win = htons((u_short)win);
nth->th_urp = 0;
#ifdef INET6
if (isipv6) {
nth->th_sum = 0;
nth->th_sum = in6_cksum(m, IPPROTO_TCP,
sizeof(struct ip6_hdr),
tlen - sizeof(struct ip6_hdr));
ip6->ip6_hlim = in6_selecthlim(tp ? tp->t_inpcb : NULL,
ro6 && ro6->ro_rt ?
ro6->ro_rt->rt_ifp :
NULL);
} else
#endif /* INET6 */
{
nth->th_sum = in_pseudo(ip->ip_src.s_addr, ip->ip_dst.s_addr,
htons((u_short)(tlen - sizeof(struct ip) + ip->ip_p)));
m->m_pkthdr.csum_flags = CSUM_TCP;
m->m_pkthdr.csum_data = offsetof(struct tcphdr, th_sum);
}
#ifdef TCPDEBUG
if (tp == NULL || (tp->t_inpcb->inp_socket->so_options & SO_DEBUG))
tcp_trace(TA_OUTPUT, 0, tp, mtod(m, void *), th, 0);
#endif
#ifdef IPSEC
if (ipsec_setsocket(m, tp ? tp->t_inpcb->inp_socket : NULL) != 0) {
m_freem(m);
return;
}
#endif
#ifdef INET6
if (isipv6) {
(void)ip6_output(m, NULL, ro6, ipflags, NULL, NULL);
if (ro6 == &sro6 && ro6->ro_rt) {
RTFREE(ro6->ro_rt);
ro6->ro_rt = NULL;
}
} else
#endif /* INET6 */
{
(void) ip_output(m, NULL, ro, ipflags, NULL);
if (ro == &sro && ro->ro_rt) {
RTFREE(ro->ro_rt);
ro->ro_rt = NULL;
}
}
1994-05-24 10:09:53 +00:00
}
/*
* Create a new TCP control block, making an
* empty reassembly queue and hooking it to the argument
* protocol control block. The `inp' parameter must have
* come from the zone allocator set up in tcp_init().
1994-05-24 10:09:53 +00:00
*/
struct tcpcb *
tcp_newtcpcb(inp)
struct inpcb *inp;
{
struct inp_tp *it;
1994-05-24 10:09:53 +00:00
register struct tcpcb *tp;
#ifdef INET6
int isipv6 = (inp->inp_vflag & INP_IPV6) != 0;
#endif /* INET6 */
1994-05-24 10:09:53 +00:00
it = (struct inp_tp *)inp;
tp = &it->tcb;
1994-05-24 10:09:53 +00:00
bzero((char *) tp, sizeof(struct tcpcb));
LIST_INIT(&tp->t_segq);
tp->t_maxseg = tp->t_maxopd =
#ifdef INET6
isipv6 ? tcp_v6mssdflt :
#endif /* INET6 */
tcp_mssdflt;
1994-05-24 10:09:53 +00:00
/* Set up our timeouts. */
callout_init(tp->tt_rexmt = &it->inp_tp_rexmt, 0);
callout_init(tp->tt_persist = &it->inp_tp_persist, 0);
callout_init(tp->tt_keep = &it->inp_tp_keep, 0);
callout_init(tp->tt_2msl = &it->inp_tp_2msl, 0);
callout_init(tp->tt_delack = &it->inp_tp_delack, 0);
if (tcp_do_rfc1323)
tp->t_flags = (TF_REQ_SCALE|TF_REQ_TSTMP);
if (tcp_do_rfc1644)
tp->t_flags |= TF_REQ_CC;
tp->t_inpcb = inp; /* XXX */
1994-05-24 10:09:53 +00:00
/*
* Init srtt to TCPTV_SRTTBASE (0), so we can tell that we have no
* rtt estimate. Set rttvar so that srtt + 4 * rttvar gives
1994-05-24 10:09:53 +00:00
* reasonable initial retransmit time.
*/
tp->t_srtt = TCPTV_SRTTBASE;
tp->t_rttvar = ((TCPTV_RTOBASE - TCPTV_SRTTBASE) << TCP_RTTVAR_SHIFT) / 4;
1994-05-24 10:09:53 +00:00
tp->t_rttmin = TCPTV_MIN;
tp->t_rxtcur = TCPTV_RTOBASE;
1994-05-24 10:09:53 +00:00
tp->snd_cwnd = TCP_MAXWIN << TCP_MAX_WINSHIFT;
tp->snd_ssthresh = TCP_MAXWIN << TCP_MAX_WINSHIFT;
tp->t_rcvtime = ticks;
/*
* IPv4 TTL initialization is necessary for an IPv6 socket as well,
* because the socket may be bound to an IPv6 wildcard address,
* which may match an IPv4-mapped IPv6 address.
*/
inp->inp_ip_ttl = ip_defttl;
1994-05-24 10:09:53 +00:00
inp->inp_ppcb = (caddr_t)tp;
return (tp); /* XXX */
1994-05-24 10:09:53 +00:00
}
/*
* Drop a TCP connection, reporting
* the specified error. If connection is synchronized,
* then send a RST to peer.
*/
struct tcpcb *
tcp_drop(tp, errno)
register struct tcpcb *tp;
int errno;
{
struct socket *so = tp->t_inpcb->inp_socket;
if (TCPS_HAVERCVDSYN(tp->t_state)) {
tp->t_state = TCPS_CLOSED;
(void) tcp_output(tp);
tcpstat.tcps_drops++;
} else
tcpstat.tcps_conndrops++;
if (errno == ETIMEDOUT && tp->t_softerror)
errno = tp->t_softerror;
so->so_error = errno;
return (tcp_close(tp));
}
/*
* Close a TCP control block:
* discard all space held by the tcp
* discard internet protocol block
* wake up any sleepers
*/
struct tcpcb *
tcp_close(tp)
register struct tcpcb *tp;
{
register struct tseg_qent *q;
1994-05-24 10:09:53 +00:00
struct inpcb *inp = tp->t_inpcb;
struct socket *so = inp->inp_socket;
#ifdef INET6
int isipv6 = (inp->inp_vflag & INP_IPV6) != 0;
#endif /* INET6 */
1994-05-24 10:09:53 +00:00
register struct rtentry *rt;
int dosavessthresh;
1994-05-24 10:09:53 +00:00
/*
* Make sure that all of our timers are stopped before we
* delete the PCB.
*/
callout_stop(tp->tt_rexmt);
callout_stop(tp->tt_persist);
callout_stop(tp->tt_keep);
callout_stop(tp->tt_2msl);
callout_stop(tp->tt_delack);
1994-05-24 10:09:53 +00:00
/*
* If we got enough samples through the srtt filter,
* save the rtt and rttvar in the routing entry.
* 'Enough' is arbitrarily defined as the 16 samples.
* 16 samples is enough for the srtt filter to converge
* to within 5% of the correct value; fewer samples and
* we could save a very bogus rtt.
1994-05-24 10:09:53 +00:00
*
* Don't update the default route's characteristics and don't
* update anything that the user "locked".
*/
if (tp->t_rttupdated >= 16) {
register u_long i = 0;
#ifdef INET6
if (isipv6) {
struct sockaddr_in6 *sin6;
if ((rt = inp->in6p_route.ro_rt) == NULL)
goto no_valid_rt;
sin6 = (struct sockaddr_in6 *)rt_key(rt);
if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))
goto no_valid_rt;
}
else
#endif /* INET6 */
if ((rt = inp->inp_route.ro_rt) == NULL ||
((struct sockaddr_in *)rt_key(rt))->sin_addr.s_addr
== INADDR_ANY)
goto no_valid_rt;
1994-05-24 10:09:53 +00:00
if ((rt->rt_rmx.rmx_locks & RTV_RTT) == 0) {
i = tp->t_srtt *
(RTM_RTTUNIT / (hz * TCP_RTT_SCALE));
1994-05-24 10:09:53 +00:00
if (rt->rt_rmx.rmx_rtt && i)
/*
* filter this update to half the old & half
* the new values, converting scale.
* See route.h and tcp_var.h for a
* description of the scaling constants.
*/
rt->rt_rmx.rmx_rtt =
(rt->rt_rmx.rmx_rtt + i) / 2;
else
rt->rt_rmx.rmx_rtt = i;
tcpstat.tcps_cachedrtt++;
1994-05-24 10:09:53 +00:00
}
if ((rt->rt_rmx.rmx_locks & RTV_RTTVAR) == 0) {
i = tp->t_rttvar *
(RTM_RTTUNIT / (hz * TCP_RTTVAR_SCALE));
1994-05-24 10:09:53 +00:00
if (rt->rt_rmx.rmx_rttvar && i)
rt->rt_rmx.rmx_rttvar =
(rt->rt_rmx.rmx_rttvar + i) / 2;
else
rt->rt_rmx.rmx_rttvar = i;
tcpstat.tcps_cachedrttvar++;
1994-05-24 10:09:53 +00:00
}
/*
* The old comment here said:
1994-05-24 10:09:53 +00:00
* update the pipelimit (ssthresh) if it has been updated
* already or if a pipesize was specified & the threshhold
* got below half the pipesize. I.e., wait for bad news
* before we start updating, then update on both good
* and bad news.
*
* But we want to save the ssthresh even if no pipesize is
* specified explicitly in the route, because such
* connections still have an implicit pipesize specified
* by the global tcp_sendspace. In the absence of a reliable
* way to calculate the pipesize, it will have to do.
1994-05-24 10:09:53 +00:00
*/
i = tp->snd_ssthresh;
if (rt->rt_rmx.rmx_sendpipe != 0)
dosavessthresh = (i < rt->rt_rmx.rmx_sendpipe / 2);
else
dosavessthresh = (i < so->so_snd.sb_hiwat / 2);
1994-10-08 22:39:58 +00:00
if (((rt->rt_rmx.rmx_locks & RTV_SSTHRESH) == 0 &&
i != 0 && rt->rt_rmx.rmx_ssthresh != 0)
|| dosavessthresh) {
1994-05-24 10:09:53 +00:00
/*
* convert the limit from user data bytes to
* packets then to packet data bytes.
*/
i = (i + tp->t_maxseg / 2) / tp->t_maxseg;
if (i < 2)
i = 2;
i *= (u_long)(tp->t_maxseg +
#ifdef INET6
(isipv6 ? sizeof (struct ip6_hdr) +
sizeof (struct tcphdr) :
#endif
sizeof (struct tcpiphdr)
#ifdef INET6
)
#endif
);
1994-05-24 10:09:53 +00:00
if (rt->rt_rmx.rmx_ssthresh)
rt->rt_rmx.rmx_ssthresh =
(rt->rt_rmx.rmx_ssthresh + i) / 2;
else
rt->rt_rmx.rmx_ssthresh = i;
tcpstat.tcps_cachedssthresh++;
1994-05-24 10:09:53 +00:00
}
}
no_valid_rt:
1994-05-24 10:09:53 +00:00
/* free the reassembly queue, if any */
while((q = LIST_FIRST(&tp->t_segq)) != NULL) {
LIST_REMOVE(q, tqe_q);
m_freem(q->tqe_m);
FREE(q, M_TSEGQ);
1994-05-24 10:09:53 +00:00
}
Improved connection establishment performance by doing local port lookups via a hashed port list. In the new scheme, in_pcblookup() goes away and is replaced by a new routine, in_pcblookup_local() for doing the local port check. Note that this implementation is space inefficient in that the PCB struct is now too large to fit into 128 bytes. I might deal with this in the future by using the new zone allocator, but I wanted these changes to be extensively tested in their current form first. Also: 1) Fixed off-by-one errors in the port lookup loops in in_pcbbind(). 2) Got rid of some unneeded rehashing. Adding a new routine, in_pcbinshash() to do the initialial hash insertion. 3) Renamed in_pcblookuphash() to in_pcblookup_hash() for easier readability. 4) Added a new routine, in_pcbremlists() to remove the PCB from the various hash lists. 5) Added/deleted comments where appropriate. 6) Removed unnecessary splnet() locking. In general, the PCB functions should be called at splnet()...there are unfortunately a few exceptions, however. 7) Reorganized a few structs for better cache line behavior. 8) Killed my TCP_ACK_HACK kludge. It may come back in a different form in the future, however. These changes have been tested on wcarchive for more than a month. In tests done here, connection establishment overhead is reduced by more than 50 times, thus getting rid of one of the major networking scalability problems. Still to do: make tcp_fastimo/tcp_slowtimo scale well for systems with a large number of connections. tcp_fastimo is easy; tcp_slowtimo is difficult. WARNING: Anything that knows about inpcb and tcpcb structs will have to be recompiled; at the very least, this includes netstat(1).
1998-01-27 09:15:13 +00:00
inp->inp_ppcb = NULL;
1994-05-24 10:09:53 +00:00
soisdisconnected(so);
#ifdef INET6
if (INP_CHECK_SOCKAF(so, AF_INET6))
in6_pcbdetach(inp);
else
#endif /* INET6 */
1994-05-24 10:09:53 +00:00
in_pcbdetach(inp);
tcpstat.tcps_closed++;
return ((struct tcpcb *)0);
}
void
tcp_drain()
{
if (do_tcpdrain)
{
struct inpcb *inpb;
struct tcpcb *tcpb;
struct tseg_qent *te;
/*
* Walk the tcpbs, if existing, and flush the reassembly queue,
* if there is one...
* XXX: The "Net/3" implementation doesn't imply that the TCP
* reassembly queue should be flushed, but in a situation
* where we're really low on mbufs, this is potentially
* usefull.
*/
2001-03-16 20:00:53 +00:00
LIST_FOREACH(inpb, tcbinfo.listhead, inp_list) {
if ((tcpb = intotcpcb(inpb))) {
while ((te = LIST_FIRST(&tcpb->t_segq))
!= NULL) {
LIST_REMOVE(te, tqe_q);
m_freem(te->tqe_m);
FREE(te, M_TSEGQ);
}
}
}
}
1994-05-24 10:09:53 +00:00
}
/*
* Notify a tcp user of an asynchronous error;
* store error as soft error, but wake up user
* (for now, won't do anything until can select for soft error).
*
* Do not wake up user since there currently is no mechanism for
* reporting soft errors (yet - a kqueue filter may be added).
1994-05-24 10:09:53 +00:00
*/
static void
1994-05-24 10:09:53 +00:00
tcp_notify(inp, error)
struct inpcb *inp;
int error;
{
struct tcpcb *tp = (struct tcpcb *)inp->inp_ppcb;
1994-05-24 10:09:53 +00:00
/*
* Ignore some errors if we are hooked up.
* If connection hasn't completed, has retransmitted several times,
* and receives a second error, give up now. This is better
* than waiting a long time to establish a connection that
* can never complete.
*/
if (tp->t_state == TCPS_ESTABLISHED &&
(error == EHOSTUNREACH || error == ENETUNREACH ||
error == EHOSTDOWN)) {
return;
} else if (tp->t_state < TCPS_ESTABLISHED && tp->t_rxtshift > 3 &&
tp->t_softerror)
tcp_drop(tp, error);
1995-05-30 08:16:23 +00:00
else
1994-05-24 10:09:53 +00:00
tp->t_softerror = error;
#if 0
1994-05-24 10:09:53 +00:00
wakeup((caddr_t) &so->so_timeo);
sorwakeup(so);
sowwakeup(so);
#endif
1994-05-24 10:09:53 +00:00
}
static int
tcp_pcblist(SYSCTL_HANDLER_ARGS)
{
int error, i, n, s;
struct inpcb *inp, **inp_list;
inp_gen_t gencnt;
struct xinpgen xig;
/*
* The process of preparing the TCB list is too time-consuming and
* resource-intensive to repeat twice on every request.
*/
if (req->oldptr == 0) {
n = tcbinfo.ipi_count;
req->oldidx = 2 * (sizeof xig)
+ (n + n/8) * sizeof(struct xtcpcb);
return 0;
}
if (req->newptr != 0)
return EPERM;
/*
* OK, now we're committed to doing something.
*/
s = splnet();
gencnt = tcbinfo.ipi_gencnt;
n = tcbinfo.ipi_count;
splx(s);
xig.xig_len = sizeof xig;
xig.xig_count = n;
xig.xig_gen = gencnt;
xig.xig_sogen = so_gencnt;
error = SYSCTL_OUT(req, &xig, sizeof xig);
if (error)
return error;
inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK);
if (inp_list == 0)
return ENOMEM;
s = splnet();
for (inp = LIST_FIRST(tcbinfo.listhead), i = 0; inp && i < n;
inp = LIST_NEXT(inp, inp_list)) {
if (inp->inp_gencnt <= gencnt) {
if (cr_cansee(req->td->td_proc->p_ucred,
inp->inp_socket->so_cred))
continue;
inp_list[i++] = inp;
}
}
splx(s);
n = i;
error = 0;
for (i = 0; i < n; i++) {
inp = inp_list[i];
if (inp->inp_gencnt <= gencnt) {
struct xtcpcb xt;
caddr_t inp_ppcb;
xt.xt_len = sizeof xt;
/* XXX should avoid extra copy */
bcopy(inp, &xt.xt_inp, sizeof *inp);
inp_ppcb = inp->inp_ppcb;
if (inp_ppcb != NULL)
bcopy(inp_ppcb, &xt.xt_tp, sizeof xt.xt_tp);
else
bzero((char *) &xt.xt_tp, sizeof xt.xt_tp);
if (inp->inp_socket)
sotoxsocket(inp->inp_socket, &xt.xt_socket);
error = SYSCTL_OUT(req, &xt, sizeof xt);
}
}
if (!error) {
/*
* Give the user an updated idea of our state.
* If the generation differs from what we told
* her before, she knows that something happened
* while we were processing this request, and it
* might be necessary to retry.
*/
s = splnet();
xig.xig_gen = tcbinfo.ipi_gencnt;
xig.xig_sogen = so_gencnt;
xig.xig_count = tcbinfo.ipi_count;
splx(s);
error = SYSCTL_OUT(req, &xig, sizeof xig);
}
free(inp_list, M_TEMP);
return error;
}
SYSCTL_PROC(_net_inet_tcp, TCPCTL_PCBLIST, pcblist, CTLFLAG_RD, 0, 0,
tcp_pcblist, "S,xtcpcb", "List of active TCP connections");
static int
tcp_getcred(SYSCTL_HANDLER_ARGS)
{
struct xucred xuc;
struct sockaddr_in addrs[2];
struct inpcb *inp;
int error, s;
error = suser_xxx(0, req->td->td_proc, PRISON_ROOT);
if (error)
return (error);
error = SYSCTL_IN(req, addrs, sizeof(addrs));
if (error)
return (error);
s = splnet();
inp = in_pcblookup_hash(&tcbinfo, addrs[1].sin_addr, addrs[1].sin_port,
addrs[0].sin_addr, addrs[0].sin_port, 0, NULL);
if (inp == NULL || inp->inp_socket == NULL) {
error = ENOENT;
goto out;
}
error = cr_cansee(req->td->td_proc->p_ucred, inp->inp_socket->so_cred);
if (error)
goto out;
bzero(&xuc, sizeof(xuc));
xuc.cr_uid = inp->inp_socket->so_cred->cr_uid;
xuc.cr_ngroups = inp->inp_socket->so_cred->cr_ngroups;
bcopy(inp->inp_socket->so_cred->cr_groups, xuc.cr_groups,
sizeof(xuc.cr_groups));
error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred));
out:
splx(s);
return (error);
}
SYSCTL_PROC(_net_inet_tcp, OID_AUTO, getcred,
CTLTYPE_OPAQUE|CTLFLAG_RW|CTLFLAG_PRISON, 0, 0,
tcp_getcred, "S,xucred", "Get the xucred of a TCP connection");
#ifdef INET6
static int
tcp6_getcred(SYSCTL_HANDLER_ARGS)
{
struct xucred xuc;
struct sockaddr_in6 addrs[2];
struct inpcb *inp;
int error, s, mapped = 0;
error = suser_xxx(0, req->td->td_proc, PRISON_ROOT);
if (error)
return (error);
error = SYSCTL_IN(req, addrs, sizeof(addrs));
if (error)
return (error);
if (IN6_IS_ADDR_V4MAPPED(&addrs[0].sin6_addr)) {
if (IN6_IS_ADDR_V4MAPPED(&addrs[1].sin6_addr))
mapped = 1;
else
return (EINVAL);
}
s = splnet();
if (mapped == 1)
inp = in_pcblookup_hash(&tcbinfo,
*(struct in_addr *)&addrs[1].sin6_addr.s6_addr[12],
addrs[1].sin6_port,
*(struct in_addr *)&addrs[0].sin6_addr.s6_addr[12],
addrs[0].sin6_port,
0, NULL);
else
inp = in6_pcblookup_hash(&tcbinfo, &addrs[1].sin6_addr,
addrs[1].sin6_port,
&addrs[0].sin6_addr, addrs[0].sin6_port,
0, NULL);
if (inp == NULL || inp->inp_socket == NULL) {
error = ENOENT;
goto out;
}
error = cr_cansee(req->td->td_proc->p_ucred, inp->inp_socket->so_cred);
if (error)
goto out;
bzero(&xuc, sizeof(xuc));
xuc.cr_uid = inp->inp_socket->so_cred->cr_uid;
xuc.cr_ngroups = inp->inp_socket->so_cred->cr_ngroups;
bcopy(inp->inp_socket->so_cred->cr_groups, xuc.cr_groups,
sizeof(xuc.cr_groups));
error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred));
out:
splx(s);
return (error);
}
SYSCTL_PROC(_net_inet6_tcp6, OID_AUTO, getcred,
CTLTYPE_OPAQUE|CTLFLAG_RW|CTLFLAG_PRISON, 0, 0,
tcp6_getcred, "S,xucred", "Get the xucred of a TCP6 connection");
#endif
1994-05-24 10:09:53 +00:00
void
tcp_ctlinput(cmd, sa, vip)
1994-05-24 10:09:53 +00:00
int cmd;
struct sockaddr *sa;
void *vip;
1994-05-24 10:09:53 +00:00
{
struct ip *ip = vip;
struct tcphdr *th;
struct in_addr faddr;
struct inpcb *inp;
struct tcpcb *tp;
1994-05-24 10:09:53 +00:00
void (*notify) __P((struct inpcb *, int)) = tcp_notify;
tcp_seq icmp_seq;
int s;
faddr = ((struct sockaddr_in *)sa)->sin_addr;
if (sa->sa_family != AF_INET || faddr.s_addr == INADDR_ANY)
return;
1994-05-24 10:09:53 +00:00
if (cmd == PRC_QUENCH)
notify = tcp_quench;
else if (icmp_may_rst && (cmd == PRC_UNREACH_ADMIN_PROHIB ||
cmd == PRC_UNREACH_PORT) && ip)
notify = tcp_drop_syn_sent;
else if (cmd == PRC_MSGSIZE)
notify = tcp_mtudisc;
else if (PRC_IS_REDIRECT(cmd)) {
ip = 0;
notify = in_rtchange;
} else if (cmd == PRC_HOSTDEAD)
ip = 0;
else if ((unsigned)cmd > PRC_NCMDS || inetctlerrmap[cmd] == 0)
1994-05-24 10:09:53 +00:00
return;
if (ip) {
s = splnet();
th = (struct tcphdr *)((caddr_t)ip
+ (IP_VHL_HL(ip->ip_vhl) << 2));
inp = in_pcblookup_hash(&tcbinfo, faddr, th->th_dport,
ip->ip_src, th->th_sport, 0, NULL);
if (inp != NULL && inp->inp_socket != NULL) {
icmp_seq = htonl(th->th_seq);
tp = intotcpcb(inp);
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);
1994-05-24 10:09:53 +00:00
} else
in_pcbnotifyall(&tcb, faddr, inetctlerrmap[cmd], notify);
1994-05-24 10:09:53 +00:00
}
#ifdef INET6
void
tcp6_ctlinput(cmd, sa, d)
int cmd;
struct sockaddr *sa;
void *d;
{
struct tcphdr th;
void (*notify) __P((struct inpcb *, int)) = tcp_notify;
struct ip6_hdr *ip6;
struct mbuf *m;
struct ip6ctlparam *ip6cp = NULL;
const struct sockaddr_in6 *sa6_src = NULL;
int off;
struct tcp_portonly {
u_int16_t th_sport;
u_int16_t th_dport;
} *thp;
if (sa->sa_family != AF_INET6 ||
sa->sa_len != sizeof(struct sockaddr_in6))
return;
if (cmd == PRC_QUENCH)
notify = tcp_quench;
else if (cmd == PRC_MSGSIZE)
notify = tcp_mtudisc;
else if (!PRC_IS_REDIRECT(cmd) &&
((unsigned)cmd > PRC_NCMDS || inet6ctlerrmap[cmd] == 0))
return;
/* if the parameter is from icmp6, decode it. */
if (d != NULL) {
ip6cp = (struct ip6ctlparam *)d;
m = ip6cp->ip6c_m;
ip6 = ip6cp->ip6c_ip6;
off = ip6cp->ip6c_off;
sa6_src = ip6cp->ip6c_src;
} else {
m = NULL;
ip6 = NULL;
off = 0; /* fool gcc */
sa6_src = &sa6_any;
}
if (ip6) {
struct in_conninfo inc;
/*
* XXX: We assume that when IPV6 is non NULL,
* M and OFF are valid.
*/
/* check if we can safely examine src and dst ports */
if (m->m_pkthdr.len < off + sizeof(*thp))
return;
bzero(&th, sizeof(th));
m_copydata(m, off, sizeof(*thp), (caddr_t)&th);
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);
}
#endif /* INET6 */
/*
* Following is where TCP initial sequence number generation occurs.
*
* There are two places where we must use initial sequence numbers:
* 1. In SYN-ACK packets.
* 2. In SYN packets.
*
* The ISNs in SYN-ACK packets have no monotonicity requirement,
* and should be as unpredictable as possible to avoid the possibility
* of spoofing and/or connection hijacking. To satisfy this
* requirement, SYN-ACK ISNs are generated via the arc4random()
* function. If exact RFC 1948 compliance is requested via sysctl,
* these ISNs will be generated just like those in SYN packets.
*
* The ISNs in SYN packets must be monotonic; TIME_WAIT recycling
* depends on this property. In addition, these ISNs should be
* unguessable so as to prevent connection hijacking. To satisfy
* the requirements of this situation, the algorithm outlined in
* RFC 1948 is used to generate sequence numbers.
*
* For more information on the theory of operation, please see
* RFC 1948.
*
* Implementation details:
*
* Time is based off the system timer, and is corrected so that it
* increases by one megabyte per second. This allows for proper
* recycling on high speed LANs while still leaving over an hour
* before rollover.
*
* Two sysctls control the generation of ISNs:
*
* net.inet.tcp.isn_reseed_interval controls the number of seconds
* between seeding of isn_secret. This is normally set to zero,
* as reseeding should not be necessary.
*
* net.inet.tcp.strict_rfc1948 controls whether RFC 1948 is followed
* strictly. When strict compliance is requested, reseeding is
* disabled and SYN-ACKs will be generated in the same manner as
* SYNs. Strict mode is disabled by default.
*
*/
#define ISN_BYTES_PER_SECOND 1048576
u_char isn_secret[32];
int isn_last_reseed;
MD5_CTX isn_ctx;
tcp_seq
tcp_new_isn(tp)
struct tcpcb *tp;
{
u_int32_t md5_buffer[4];
tcp_seq new_isn;
/* Use arc4random for SYN-ACKs when not in exact RFC1948 mode. */
if (((tp->t_state == TCPS_LISTEN) || (tp->t_state == TCPS_TIME_WAIT))
&& tcp_strict_rfc1948 == 0)
return arc4random();
/* Seed if this is the first use, reseed if requested. */
if ((isn_last_reseed == 0) ||
((tcp_strict_rfc1948 == 0) && (tcp_isn_reseed_interval > 0) &&
(((u_int)isn_last_reseed + (u_int)tcp_isn_reseed_interval*hz)
< (u_int)ticks))) {
read_random(&isn_secret, sizeof(isn_secret));
isn_last_reseed = ticks;
}
/* Compute the md5 hash and return the ISN. */
MD5Init(&isn_ctx);
MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->inp_fport, sizeof(u_short));
MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->inp_lport, sizeof(u_short));
#ifdef INET6
if ((tp->t_inpcb->inp_vflag & INP_IPV6) != 0) {
MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->in6p_faddr,
sizeof(struct in6_addr));
MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->in6p_laddr,
sizeof(struct in6_addr));
} else
#endif
{
MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->inp_faddr,
sizeof(struct in_addr));
MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->inp_laddr,
sizeof(struct in_addr));
}
MD5Update(&isn_ctx, (u_char *) &isn_secret, sizeof(isn_secret));
MD5Final((u_char *) &md5_buffer, &isn_ctx);
new_isn = (tcp_seq) md5_buffer[0];
new_isn += ticks * (ISN_BYTES_PER_SECOND / hz);
return new_isn;
}
1994-05-24 10:09:53 +00:00
/*
* When a source quench is received, close congestion window
* to one segment. We will gradually open it again as we proceed.
*/
void
tcp_quench(inp, errno)
struct inpcb *inp;
int errno;
{
struct tcpcb *tp = intotcpcb(inp);
if (tp)
tp->snd_cwnd = tp->t_maxseg;
}
We currently does not react to ICMP administratively prohibited messages send by routers when they deny our traffic, this causes a timeout when trying to connect to TCP ports/services on a remote host, which is blocked by routers or firewalls. rfc1122 (Requirements for Internet Hosts) section 3.2.2.1 actually requi re that we treat such a message for a TCP session, that we treat it like if we had recieved a RST. quote begin. A Destination Unreachable message that is received MUST be reported to the transport layer. The transport layer SHOULD use the information appropriately; for example, see Sections 4.1.3.3, 4.2.3.9, and 4.2.4 below. A transport protocol that has its own mechanism for notifying the sender that a port is unreachable (e.g., TCP, which sends RST segments) MUST nevertheless accept an ICMP Port Unreachable for the same purpose. quote end. I've written a small extension that implement this, it also create a sysctl "net.inet.tcp.icmp_admin_prohib_like_rst" to control if this new behaviour is activated. When it's activated (set to 1) we'll treat a ICMP administratively prohibited message (icmp type 3 code 9, 10 and 13) for a TCP sessions, as if we recived a TCP RST, but only if the TCP session is in SYN_SENT state. The reason for only reacting when in SYN_SENT state, is that this will solve the problem, and at the same time minimize the risk of this being abused. I suggest that we enable this new behaviour by default, but it would be a change of current behaviour, so if people prefer to leave it disabled by default, at least for now, this would be ok for me, the attached diff actually have the sysctl set to 0 by default. PR: 23086 Submitted by: Jesper Skriver <jesper@skriver.dk>
2000-12-16 19:42:06 +00:00
/*
* When a specific ICMP unreachable message is received and the
* connection state is SYN-SENT, drop the connection. This behavior
* is controlled by the icmp_may_rst sysctl.
We currently does not react to ICMP administratively prohibited messages send by routers when they deny our traffic, this causes a timeout when trying to connect to TCP ports/services on a remote host, which is blocked by routers or firewalls. rfc1122 (Requirements for Internet Hosts) section 3.2.2.1 actually requi re that we treat such a message for a TCP session, that we treat it like if we had recieved a RST. quote begin. A Destination Unreachable message that is received MUST be reported to the transport layer. The transport layer SHOULD use the information appropriately; for example, see Sections 4.1.3.3, 4.2.3.9, and 4.2.4 below. A transport protocol that has its own mechanism for notifying the sender that a port is unreachable (e.g., TCP, which sends RST segments) MUST nevertheless accept an ICMP Port Unreachable for the same purpose. quote end. I've written a small extension that implement this, it also create a sysctl "net.inet.tcp.icmp_admin_prohib_like_rst" to control if this new behaviour is activated. When it's activated (set to 1) we'll treat a ICMP administratively prohibited message (icmp type 3 code 9, 10 and 13) for a TCP sessions, as if we recived a TCP RST, but only if the TCP session is in SYN_SENT state. The reason for only reacting when in SYN_SENT state, is that this will solve the problem, and at the same time minimize the risk of this being abused. I suggest that we enable this new behaviour by default, but it would be a change of current behaviour, so if people prefer to leave it disabled by default, at least for now, this would be ok for me, the attached diff actually have the sysctl set to 0 by default. PR: 23086 Submitted by: Jesper Skriver <jesper@skriver.dk>
2000-12-16 19:42:06 +00:00
*/
void
tcp_drop_syn_sent(inp, errno)
struct inpcb *inp;
int errno;
{
struct tcpcb *tp = intotcpcb(inp);
if (tp && tp->t_state == TCPS_SYN_SENT)
tcp_drop(tp, errno);
}
/*
* When `need fragmentation' ICMP is received, update our idea of the MSS
* based on the new value in the route. Also nudge TCP to send something,
* since we know the packet we just sent was dropped.
* This duplicates some code in the tcp_mss() function in tcp_input.c.
*/
void
tcp_mtudisc(inp, errno)
struct inpcb *inp;
int errno;
{
struct tcpcb *tp = intotcpcb(inp);
struct rtentry *rt;
struct rmxp_tao *taop;
struct socket *so = inp->inp_socket;
int offered;
int mss;
#ifdef INET6
int isipv6 = (tp->t_inpcb->inp_vflag & INP_IPV6) != 0;
#endif /* INET6 */
if (tp) {
#ifdef INET6
if (isipv6)
rt = tcp_rtlookup6(&inp->inp_inc);
else
#endif /* INET6 */
rt = tcp_rtlookup(&inp->inp_inc);
if (!rt || !rt->rt_rmx.rmx_mtu) {
tp->t_maxopd = tp->t_maxseg =
#ifdef INET6
isipv6 ? tcp_v6mssdflt :
#endif /* INET6 */
tcp_mssdflt;
return;
}
taop = rmx_taop(rt->rt_rmx);
offered = taop->tao_mssopt;
mss = rt->rt_rmx.rmx_mtu -
#ifdef INET6
(isipv6 ?
sizeof(struct ip6_hdr) + sizeof(struct tcphdr) :
#endif /* INET6 */
sizeof(struct tcpiphdr)
#ifdef INET6
)
#endif /* INET6 */
;
if (offered)
mss = min(mss, offered);
/*
* XXX - The above conditional probably violates the TCP
* spec. The problem is that, since we don't know the
* other end's MSS, we are supposed to use a conservative
* default. But, if we do that, then MTU discovery will
* never actually take place, because the conservative
* default is much less than the MTUs typically seen
* on the Internet today. For the moment, we'll sweep
* this under the carpet.
*
* The conservative default might not actually be a problem
* if the only case this occurs is when sending an initial
* SYN with options and data to a host we've never talked
* to before. Then, they will reply with an MSS value which
* will get recorded and the new parameters should get
* recomputed. For Further Study.
*/
if (tp->t_maxopd <= mss)
return;
tp->t_maxopd = mss;
if ((tp->t_flags & (TF_REQ_TSTMP|TF_NOOPT)) == TF_REQ_TSTMP &&
(tp->t_flags & TF_RCVD_TSTMP) == TF_RCVD_TSTMP)
mss -= TCPOLEN_TSTAMP_APPA;
if ((tp->t_flags & (TF_REQ_CC|TF_NOOPT)) == TF_REQ_CC &&
(tp->t_flags & TF_RCVD_CC) == TF_RCVD_CC)
mss -= TCPOLEN_CC_APPA;
#if (MCLBYTES & (MCLBYTES - 1)) == 0
if (mss > MCLBYTES)
mss &= ~(MCLBYTES-1);
#else
if (mss > MCLBYTES)
mss = mss / MCLBYTES * MCLBYTES;
#endif
if (so->so_snd.sb_hiwat < mss)
mss = so->so_snd.sb_hiwat;
tp->t_maxseg = mss;
tcpstat.tcps_mturesent++;
tp->t_rtttime = 0;
tp->snd_nxt = tp->snd_una;
tcp_output(tp);
}
}
/*
* Look-up the routing entry to the peer of this inpcb. If no route
* is found and it cannot be allocated the return NULL. This routine
* is called by TCP routines that access the rmx structure and by tcp_mss
* to get the interface MTU.
*/
struct rtentry *
tcp_rtlookup(inc)
struct in_conninfo *inc;
{
struct route *ro;
struct rtentry *rt;
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 (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 =
inc->inc_faddr;
rtalloc(ro);
rt = ro->ro_rt;
}
}
return rt;
}
#ifdef INET6
struct rtentry *
tcp_rtlookup6(inc)
struct in_conninfo *inc;
{
struct route_in6 *ro6;
struct rtentry *rt;
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(&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;
}
}
return rt;
}
#endif /* INET6 */
#ifdef IPSEC
/* compute ESP/AH header size for TCP, including outer IP header. */
size_t
ipsec_hdrsiz_tcp(tp)
struct tcpcb *tp;
{
struct inpcb *inp;
struct mbuf *m;
size_t hdrsiz;
struct ip *ip;
#ifdef INET6
struct ip6_hdr *ip6;
#endif /* INET6 */
struct tcphdr *th;
if ((tp == NULL) || ((inp = tp->t_inpcb) == NULL))
return 0;
MGETHDR(m, M_DONTWAIT, MT_DATA);
if (!m)
return 0;
#ifdef INET6
if ((inp->inp_vflag & INP_IPV6) != 0) {
ip6 = mtod(m, struct ip6_hdr *);
th = (struct tcphdr *)(ip6 + 1);
m->m_pkthdr.len = m->m_len =
sizeof(struct ip6_hdr) + sizeof(struct tcphdr);
tcp_fillheaders(tp, ip6, th);
hdrsiz = ipsec6_hdrsiz(m, IPSEC_DIR_OUTBOUND, inp);
} else
#endif /* INET6 */
{
ip = mtod(m, struct ip *);
th = (struct tcphdr *)(ip + 1);
m->m_pkthdr.len = m->m_len = sizeof(struct tcpiphdr);
tcp_fillheaders(tp, ip, th);
hdrsiz = ipsec4_hdrsiz(m, IPSEC_DIR_OUTBOUND, inp);
}
m_free(m);
return hdrsiz;
}
#endif /*IPSEC*/
/*
* Return a pointer to the cached information about the remote host.
* The cached information is stored in the protocol specific part of
* the route metrics.
*/
struct rmxp_tao *
tcp_gettaocache(inc)
struct in_conninfo *inc;
{
struct rtentry *rt;
#ifdef INET6
if (inc->inc_isipv6)
rt = tcp_rtlookup6(inc);
else
#endif /* INET6 */
rt = tcp_rtlookup(inc);
/* Make sure this is a host route and is up. */
if (rt == NULL ||
(rt->rt_flags & (RTF_UP|RTF_HOST)) != (RTF_UP|RTF_HOST))
return NULL;
return rmx_taop(rt->rt_rmx);
}
/*
* Clear all the TAO cache entries, called from tcp_init.
*
* XXX
* This routine is just an empty one, because we assume that the routing
* routing tables are initialized at the same time when TCP, so there is
* nothing in the cache left over.
*/
static void
1997-09-16 11:44:05 +00:00
tcp_cleartaocache()
{
}