#include <sys/param.h>
#include <sys/systm.h>
#include <sys/callout.h>
#include <sys/kernel.h>
#include <sys/sysctl.h>
#include <sys/malloc.h>
#include <sys/mbuf.h>
#include <sys/domain.h>
#include <sys/proc.h>
#include <sys/kauth.h>
#include <sys/socket.h>
#include <sys/socketvar.h>
#include <sys/protosw.h>
#include <sys/random.h>
#include <sys/syslog.h>
#include <sys/mcache.h>
#include <kern/locks.h>
#include <kern/zalloc.h>
#include <dev/random/randomdev.h>
#include <net/route.h>
#include <net/if.h>
#include <net/content_filter.h>
#define tcp_minmssoverload fring
#define _IP_VHL
#include <netinet/in.h>
#include <netinet/in_systm.h>
#include <netinet/ip.h>
#include <netinet/ip_icmp.h>
#if INET6
#include <netinet/ip6.h>
#endif
#include <netinet/in_pcb.h>
#if INET6
#include <netinet6/in6_pcb.h>
#endif
#include <netinet/in_var.h>
#include <netinet/ip_var.h>
#include <netinet/icmp_var.h>
#if INET6
#include <netinet6/ip6_var.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>
#include <netinet/tcp_cc.h>
#include <netinet/tcp_cache.h>
#include <kern/thread_call.h>
#if INET6
#include <netinet6/tcp6_var.h>
#endif
#include <netinet/tcpip.h>
#if TCPDEBUG
#include <netinet/tcp_debug.h>
#endif
#include <netinet6/ip6protosw.h>
#if IPSEC
#include <netinet6/ipsec.h>
#if INET6
#include <netinet6/ipsec6.h>
#endif
#endif
#if NECP
#include <net/necp.h>
#endif
#undef tcp_minmssoverload
#if CONFIG_MACF_NET
#include <security/mac_framework.h>
#endif
#include <corecrypto/ccaes.h>
#include <libkern/crypto/aes.h>
#include <libkern/crypto/md5.h>
#include <sys/kdebug.h>
#include <mach/sdt.h>
#include <netinet/lro_ext.h>
#define DBG_FNC_TCP_CLOSE NETDBG_CODE(DBG_NETTCP, ((5 << 8) | 2))
extern int tcp_lq_overflow;
extern struct tcptimerlist tcp_timer_list;
extern struct tcptailq tcp_tw_tailq;
int tcp_mssdflt = TCP_MSS;
SYSCTL_INT(_net_inet_tcp, TCPCTL_MSSDFLT, mssdflt, CTLFLAG_RW | CTLFLAG_LOCKED,
&tcp_mssdflt , 0, "Default TCP Maximum Segment Size");
#if INET6
int tcp_v6mssdflt = TCP6_MSS;
SYSCTL_INT(_net_inet_tcp, TCPCTL_V6MSSDFLT, v6mssdflt,
CTLFLAG_RW | CTLFLAG_LOCKED, &tcp_v6mssdflt , 0,
"Default TCP Maximum Segment Size for IPv6");
#endif
extern int tcp_do_autorcvbuf;
int tcp_sysctl_fastopenkey(struct sysctl_oid *, void *, int ,
struct sysctl_req *);
SYSCTL_PROC(_net_inet_tcp, OID_AUTO, fastopen_key,
CTLTYPE_STRING | CTLFLAG_WR,
0 , 0, tcp_sysctl_fastopenkey, "S", "TCP Fastopen key");
int tcp_tfo_halfcnt = 0;
int tcp_tfo_backlog = 10;
SYSCTL_INT(_net_inet_tcp, OID_AUTO, fastopen_backlog, CTLFLAG_RW | CTLFLAG_LOCKED,
&tcp_tfo_backlog, 0, "Backlog queue for half-open TFO connections");
int tcp_fastopen = TCP_FASTOPEN_CLIENT | TCP_FASTOPEN_SERVER;
SYSCTL_INT(_net_inet_tcp, OID_AUTO, fastopen, CTLFLAG_RW | CTLFLAG_LOCKED,
&tcp_fastopen, 0, "Enable TCP Fastopen (RFC 7413)");
int tcp_tfo_fallback_min = 10;
SYSCTL_INT(_net_inet_tcp, OID_AUTO, fastopen_fallback_min, CTLFLAG_RW | CTLFLAG_LOCKED,
&tcp_tfo_fallback_min, 0, "Mininum number of trials without TFO when in fallback mode");
int tcp_minmss = TCP_MINMSS;
SYSCTL_INT(_net_inet_tcp, OID_AUTO, minmss, CTLFLAG_RW | CTLFLAG_LOCKED,
&tcp_minmss , 0, "Minmum TCP Maximum Segment Size");
int tcp_do_rfc1323 = 1;
#if (DEVELOPMENT || DEBUG)
SYSCTL_INT(_net_inet_tcp, TCPCTL_DO_RFC1323, rfc1323,
CTLFLAG_RW | CTLFLAG_LOCKED, &tcp_do_rfc1323 , 0,
"Enable rfc1323 (high performance TCP) extensions");
#endif
static int tcp_do_rfc1644 = 0;
SYSCTL_INT(_net_inet_tcp, TCPCTL_DO_RFC1644, rfc1644, CTLFLAG_RW | CTLFLAG_LOCKED,
&tcp_do_rfc1644 , 0, "Enable rfc1644 (TTCP) extensions");
static int do_tcpdrain = 0;
SYSCTL_INT(_net_inet_tcp, OID_AUTO, do_tcpdrain, CTLFLAG_RW | CTLFLAG_LOCKED, &do_tcpdrain, 0,
"Enable tcp_drain routine for extra help when low on mbufs");
SYSCTL_INT(_net_inet_tcp, OID_AUTO, pcbcount, CTLFLAG_RD | CTLFLAG_LOCKED,
&tcbinfo.ipi_count, 0, "Number of active PCBs");
SYSCTL_INT(_net_inet_tcp, OID_AUTO, tw_pcbcount,
CTLFLAG_RD | CTLFLAG_LOCKED,
&tcbinfo.ipi_twcount, 0, "Number of pcbs in time-wait state");
static int icmp_may_rst = 1;
SYSCTL_INT(_net_inet_tcp, OID_AUTO, icmp_may_rst, CTLFLAG_RW | CTLFLAG_LOCKED, &icmp_may_rst, 0,
"Certain ICMP unreachable messages may abort connections in SYN_SENT");
static int tcp_strict_rfc1948 = 0;
static int tcp_isn_reseed_interval = 0;
#if (DEVELOPMENT || DEBUG)
SYSCTL_INT(_net_inet_tcp, OID_AUTO, strict_rfc1948,
CTLFLAG_RW | CTLFLAG_LOCKED,
&tcp_strict_rfc1948, 0, "Determines if RFC1948 is followed exactly");
SYSCTL_INT(_net_inet_tcp, OID_AUTO, isn_reseed_interval,
CTLFLAG_RW | CTLFLAG_LOCKED,
&tcp_isn_reseed_interval, 0, "Seconds between reseeding of ISN secret");
#endif
int tcp_TCPTV_MIN = 100;
SYSCTL_INT(_net_inet_tcp, OID_AUTO, rtt_min, CTLFLAG_RW | CTLFLAG_LOCKED,
&tcp_TCPTV_MIN, 0, "min rtt value allowed");
int tcp_rexmt_slop = TCPTV_REXMTSLOP;
SYSCTL_INT(_net_inet_tcp, OID_AUTO, rexmt_slop, CTLFLAG_RW,
&tcp_rexmt_slop, 0, "Slop added to retransmit timeout");
__private_extern__ int tcp_use_randomport = 0;
SYSCTL_INT(_net_inet_tcp, OID_AUTO, randomize_ports, CTLFLAG_RW | CTLFLAG_LOCKED,
&tcp_use_randomport, 0, "Randomize TCP port numbers");
__private_extern__ int tcp_win_scale = 3;
SYSCTL_INT(_net_inet_tcp, OID_AUTO, win_scale_factor,
CTLFLAG_RW | CTLFLAG_LOCKED,
&tcp_win_scale, 0, "Window scaling factor");
static void tcp_cleartaocache(void);
static void tcp_notify(struct inpcb *, int);
struct zone *sack_hole_zone;
struct zone *tcp_reass_zone;
struct zone *tcp_bwmeas_zone;
struct zone *tcp_rxt_seg_zone;
extern int slowlink_wsize;
extern int path_mtu_discovery;
extern u_int32_t tcp_autorcvbuf_max;
extern u_int32_t tcp_autorcvbuf_inc_shift;
static void tcp_sbrcv_grow_rwin(struct tcpcb *tp, struct sockbuf *sb);
#define TCP_BWMEAS_BURST_MINSIZE 6
#define TCP_BWMEAS_BURST_MAXSIZE 25
static uint32_t bwmeas_elm_size;
#ifndef TCBHASHSIZE
#define TCBHASHSIZE CONFIG_TCBHASHSIZE
#endif
__private_extern__ int tcp_tcbhashsize = TCBHASHSIZE;
SYSCTL_INT(_net_inet_tcp, OID_AUTO, tcbhashsize, CTLFLAG_RD | CTLFLAG_LOCKED,
&tcp_tcbhashsize, 0, "Size of TCP control-block hashtable");
#define ALIGNMENT 32
struct inp_tp {
struct inpcb inp;
struct tcpcb tcb __attribute__((aligned(ALIGNMENT)));
};
#undef ALIGNMENT
int get_inpcb_str_size(void);
int get_tcp_str_size(void);
static void tcpcb_to_otcpcb(struct tcpcb *, struct otcpcb *);
static lck_attr_t *tcp_uptime_mtx_attr = NULL;
static lck_grp_t *tcp_uptime_mtx_grp = NULL;
static lck_grp_attr_t *tcp_uptime_mtx_grp_attr = NULL;
int tcp_notsent_lowat_check(struct socket *so);
static aes_encrypt_ctx tfo_ctx;
void
tcp_tfo_gen_cookie(struct inpcb *inp, u_char *out, size_t blk_size)
{
u_char in[CCAES_BLOCK_SIZE];
#if INET6
int isipv6 = inp->inp_vflag & INP_IPV6;
#endif
VERIFY(blk_size == CCAES_BLOCK_SIZE);
bzero(&in[0], CCAES_BLOCK_SIZE);
bzero(&out[0], CCAES_BLOCK_SIZE);
#if INET6
if (isipv6)
memcpy(in, &inp->in6p_faddr, sizeof(struct in6_addr));
else
#endif
memcpy(in, &inp->inp_faddr, sizeof(struct in_addr));
aes_encrypt_cbc(in, NULL, 1, out, &tfo_ctx);
}
__private_extern__ int
tcp_sysctl_fastopenkey(__unused struct sysctl_oid *oidp, __unused void *arg1,
__unused int arg2, struct sysctl_req *req)
{
int error = 0;
char keystring[TCP_FASTOPEN_KEYLEN * 2 + 1];
u_int32_t key[TCP_FASTOPEN_KEYLEN / sizeof(u_int32_t)];
int i;
if (req->newlen != (sizeof(keystring) - 1)) {
error = EINVAL;
goto exit;
}
bzero(keystring, sizeof(keystring));
error = sysctl_io_string(req, keystring, sizeof(keystring), 0, NULL);
if (error)
goto exit;
for (i = 0; i < (TCP_FASTOPEN_KEYLEN / sizeof(u_int32_t)); i++) {
if (sscanf(&keystring[i * 8], "%8x", &key[i]) != 1) {
error = EINVAL;
goto exit;
}
}
aes_encrypt_key128((u_char *)key, &tfo_ctx);
exit:
return (error);
}
int get_inpcb_str_size(void)
{
return sizeof(struct inpcb);
}
int get_tcp_str_size(void)
{
return sizeof(struct tcpcb);
}
int tcp_freeq(struct tcpcb *tp);
static int scale_to_powerof2(int size);
static int scale_to_powerof2(int size) {
int ret = size ? size : 1;
if (!powerof2(ret)) {
while(!powerof2(size)) {
size = size & (size -1);
}
if (0 == (size << 1)) {
ret = size;
} else {
ret = size << 1;
}
}
return ret;
}
static void
tcp_tfo_init()
{
u_char key[TCP_FASTOPEN_KEYLEN];
read_random(key, sizeof(key));
aes_encrypt_key128(key, &tfo_ctx);
}
void
tcp_init(struct protosw *pp, struct domain *dp)
{
#pragma unused(dp)
static int tcp_initialized = 0;
vm_size_t str_size;
struct inpcbinfo *pcbinfo;
VERIFY((pp->pr_flags & (PR_INITIALIZED|PR_ATTACHED)) == PR_ATTACHED);
if (tcp_initialized)
return;
tcp_initialized = 1;
tcp_ccgen = 1;
tcp_cleartaocache();
tcp_keepinit = TCPTV_KEEP_INIT;
tcp_keepidle = TCPTV_KEEP_IDLE;
tcp_keepintvl = TCPTV_KEEPINTVL;
tcp_keepcnt = TCPTV_KEEPCNT;
tcp_maxpersistidle = TCPTV_KEEP_IDLE;
tcp_msl = TCPTV_MSL;
microuptime(&tcp_uptime);
read_random(&tcp_now, sizeof(tcp_now));
tcp_now = tcp_now & 0x3fffffff;
tcp_tfo_init();
LIST_INIT(&tcb);
tcbinfo.ipi_listhead = &tcb;
pcbinfo = &tcbinfo;
pcbinfo->ipi_lock_grp_attr = lck_grp_attr_alloc_init();
pcbinfo->ipi_lock_grp = lck_grp_alloc_init("tcppcb", pcbinfo->ipi_lock_grp_attr);
pcbinfo->ipi_lock_attr = lck_attr_alloc_init();
if ((pcbinfo->ipi_lock = lck_rw_alloc_init(pcbinfo->ipi_lock_grp,
pcbinfo->ipi_lock_attr)) == NULL) {
panic("%s: unable to allocate PCB lock\n", __func__);
}
if (tcp_tcbhashsize == 0) {
tcp_tcbhashsize = 512;
}
if (!powerof2(tcp_tcbhashsize)) {
int old_hash_size = tcp_tcbhashsize;
tcp_tcbhashsize = scale_to_powerof2(tcp_tcbhashsize);
if (tcp_tcbhashsize < 16) {
tcp_tcbhashsize = 16;
}
printf("WARNING: TCB hash size not a power of 2, "
"scaled from %d to %d.\n",
old_hash_size,
tcp_tcbhashsize);
}
tcbinfo.ipi_hashbase = hashinit(tcp_tcbhashsize, M_PCB, &tcbinfo.ipi_hashmask);
tcbinfo.ipi_porthashbase = hashinit(tcp_tcbhashsize, M_PCB,
&tcbinfo.ipi_porthashmask);
str_size = P2ROUNDUP(sizeof(struct inp_tp), sizeof(u_int64_t));
tcbinfo.ipi_zone = zinit(str_size, 120000*str_size, 8192, "tcpcb");
zone_change(tcbinfo.ipi_zone, Z_CALLERACCT, FALSE);
zone_change(tcbinfo.ipi_zone, Z_EXPAND, TRUE);
tcbinfo.ipi_gc = tcp_gc;
tcbinfo.ipi_timer = tcp_itimer;
in_pcbinfo_attach(&tcbinfo);
str_size = P2ROUNDUP(sizeof(struct sackhole), sizeof(u_int64_t));
sack_hole_zone = zinit(str_size, 120000*str_size, 8192, "sack_hole zone");
zone_change(sack_hole_zone, Z_CALLERACCT, FALSE);
zone_change(sack_hole_zone, Z_EXPAND, TRUE);
str_size = P2ROUNDUP(sizeof(struct tseg_qent), sizeof(u_int64_t));
tcp_reass_zone = zinit(str_size, (nmbclusters >> 4) * str_size,
0, "tcp_reass_zone");
if (tcp_reass_zone == NULL) {
panic("%s: failed allocating tcp_reass_zone", __func__);
}
zone_change(tcp_reass_zone, Z_CALLERACCT, FALSE);
zone_change(tcp_reass_zone, Z_EXPAND, TRUE);
bwmeas_elm_size = P2ROUNDUP(sizeof(struct bwmeas), sizeof(u_int64_t));
tcp_bwmeas_zone = zinit(bwmeas_elm_size, (100 * bwmeas_elm_size), 0, "tcp_bwmeas_zone");
if (tcp_bwmeas_zone == NULL) {
panic("%s: failed allocating tcp_bwmeas_zone", __func__);
}
zone_change(tcp_bwmeas_zone, Z_CALLERACCT, FALSE);
zone_change(tcp_bwmeas_zone, Z_EXPAND, TRUE);
str_size = P2ROUNDUP(sizeof(struct tcp_ccstate), sizeof(u_int64_t));
tcp_cc_zone = zinit(str_size, 20000 * str_size, 0, "tcp_cc_zone");
zone_change(tcp_cc_zone, Z_CALLERACCT, FALSE);
zone_change(tcp_cc_zone, Z_EXPAND, TRUE);
str_size = P2ROUNDUP(sizeof(struct tcp_rxt_seg), sizeof(u_int64_t));
tcp_rxt_seg_zone = zinit(str_size, 10000 * str_size, 0,
"tcp_rxt_seg_zone");
zone_change(tcp_rxt_seg_zone, Z_CALLERACCT, FALSE);
zone_change(tcp_rxt_seg_zone, Z_EXPAND, TRUE);
#if INET6
#define TCP_MINPROTOHDR (sizeof(struct ip6_hdr) + sizeof(struct tcphdr))
#else
#define TCP_MINPROTOHDR (sizeof(struct tcpiphdr))
#endif
if (max_protohdr < TCP_MINPROTOHDR) {
_max_protohdr = TCP_MINPROTOHDR;
_max_protohdr = max_protohdr;
}
if (max_linkhdr + max_protohdr > MCLBYTES)
panic("tcp_init");
#undef TCP_MINPROTOHDR
TAILQ_INIT(&tcp_tw_tailq);
bzero(&tcp_timer_list, sizeof(tcp_timer_list));
LIST_INIT(&tcp_timer_list.lhead);
tcp_timer_list.mtx_grp_attr = lck_grp_attr_alloc_init();
tcp_timer_list.mtx_grp = lck_grp_alloc_init("tcptimerlist", tcp_timer_list.mtx_grp_attr);
tcp_timer_list.mtx_attr = lck_attr_alloc_init();
if ((tcp_timer_list.mtx = lck_mtx_alloc_init(tcp_timer_list.mtx_grp, tcp_timer_list.mtx_attr)) == NULL) {
panic("failed to allocate memory for tcp_timer_list.mtx\n");
};
if ((tcp_timer_list.call = thread_call_allocate(tcp_run_timerlist, NULL)) == NULL) {
panic("failed to allocate call entry 1 in tcp_init\n");
}
tcp_uptime_mtx_grp_attr = lck_grp_attr_alloc_init();
tcp_uptime_mtx_grp = lck_grp_alloc_init("tcpuptime", tcp_uptime_mtx_grp_attr);
tcp_uptime_mtx_attr = lck_attr_alloc_init();
tcp_uptime_lock = lck_spin_alloc_init(tcp_uptime_mtx_grp, tcp_uptime_mtx_attr);
tcp_lro_init();
tcp_cache_init();
if (nmbclusters > 30720) {
tcp_autorcvbuf_max = 1024 * 1024;
tcp_autosndbuf_max = 1024 * 1024;
}
}
void
tcp_fillheaders(tp, ip_ptr, tcp_ptr)
struct tcpcb *tp;
void *ip_ptr;
void *tcp_ptr;
{
struct inpcb *inp = tp->t_inpcb;
struct tcphdr *tcp_hdr = (struct tcphdr *)tcp_ptr;
#if 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->inp_flow & 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 = in6_pseudo(&inp->in6p_laddr, &inp->in6p_faddr,
htonl(sizeof (struct tcphdr) + IPPROTO_TCP));
} 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;
}
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);
return (n);
}
void
tcp_respond(struct tcpcb *tp, void *ipgen, struct tcphdr *th, struct mbuf *m,
tcp_seq ack, tcp_seq seq, int flags, struct tcp_respond_args *tra)
{
int tlen;
int win = 0;
struct route *ro = 0;
struct route sro;
struct ip *ip;
struct tcphdr *nth;
#if INET6
struct route_in6 *ro6 = 0;
struct route_in6 sro6;
struct ip6_hdr *ip6;
int isipv6;
#endif
struct ifnet *outif;
#if INET6
isipv6 = IP_VHL_V(((struct ip *)ipgen)->ip_vhl) == 6;
ip6 = ipgen;
#endif
ip = ipgen;
if (tp) {
if (!(flags & TH_RST)) {
win = tcp_sbspace(tp);
if (win > (int32_t)TCP_MAXWIN << tp->rcv_scale)
win = (int32_t)TCP_MAXWIN << tp->rcv_scale;
}
#if INET6
if (isipv6)
ro6 = &tp->t_inpcb->in6p_route;
else
#endif
ro = &tp->t_inpcb->inp_route;
} else {
#if INET6
if (isipv6) {
ro6 = &sro6;
bzero(ro6, sizeof *ro6);
} else
#endif
{
ro = &sro;
bzero(ro, sizeof *ro);
}
}
if (m == 0) {
m = m_gethdr(M_DONTWAIT, MT_HEADER);
if (m == NULL)
return;
tlen = 0;
m->m_data += max_linkhdr;
#if INET6
if (isipv6) {
VERIFY((MHLEN - max_linkhdr) >=
(sizeof (*ip6) + sizeof (*nth)));
bcopy((caddr_t)ip6, mtod(m, caddr_t),
sizeof(struct ip6_hdr));
ip6 = mtod(m, struct ip6_hdr *);
nth = (struct tcphdr *)(void *)(ip6 + 1);
} else
#endif
{
VERIFY((MHLEN - max_linkhdr) >=
(sizeof (*ip) + sizeof (*nth)));
bcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip));
ip = mtod(m, struct ip *);
nth = (struct tcphdr *)(void *)(ip + 1);
}
bcopy((caddr_t)th, (caddr_t)nth, sizeof(struct tcphdr));
#if MPTCP
if ((tp) && (tp->t_mpflags & TMPF_RESET))
flags = (TH_RST | TH_ACK);
else
#endif
flags = TH_ACK;
} else {
m_freem(m->m_next);
m->m_next = 0;
m->m_data = (caddr_t)ipgen;
tlen = 0;
#define xchg(a,b,type) { type t; t=a; a=b; b=t; }
#if INET6
if (isipv6) {
IP6_HDR_STRICT_ALIGNMENT_CHECK(ip6);
xchg(ip6->ip6_dst, ip6->ip6_src, struct in6_addr);
nth = (struct tcphdr *)(void *)(ip6 + 1);
} else
#endif
{
IP_HDR_STRICT_ALIGNMENT_CHECK(ip);
xchg(ip->ip_dst.s_addr, ip->ip_src.s_addr, n_long);
nth = (struct tcphdr *)(void *)(ip + 1);
}
if (th != nth) {
nth->th_sport = th->th_sport;
nth->th_dport = th->th_dport;
}
xchg(nth->th_dport, nth->th_sport, n_short);
#undef xchg
}
#if INET6
if (isipv6) {
ip6->ip6_plen = htons((u_short)(sizeof (struct tcphdr) +
tlen));
tlen += sizeof (struct ip6_hdr) + sizeof (struct tcphdr);
} else
#endif
{
tlen += sizeof (struct tcpiphdr);
ip->ip_len = tlen;
ip->ip_ttl = ip_defttl;
}
m->m_len = tlen;
m->m_pkthdr.len = tlen;
m->m_pkthdr.rcvif = 0;
#if CONFIG_MACF_NET
if (tp != NULL && tp->t_inpcb != NULL) {
mac_mbuf_label_associate_inpcb(tp->t_inpcb, m);
} else {
mac_netinet_tcp_reply(m);
}
#endif
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;
if (tp)
nth->th_win = htons((u_short) (win >> tp->rcv_scale));
else
nth->th_win = htons((u_short)win);
nth->th_urp = 0;
#if INET6
if (isipv6) {
nth->th_sum = 0;
nth->th_sum = in6_pseudo(&ip6->ip6_src, &ip6->ip6_dst,
htonl((tlen - sizeof (struct ip6_hdr)) + IPPROTO_TCP));
m->m_pkthdr.csum_flags = CSUM_TCPIPV6;
m->m_pkthdr.csum_data = offsetof(struct tcphdr, th_sum);
ip6->ip6_hlim = in6_selecthlim(tp ? tp->t_inpcb : NULL,
ro6 && ro6->ro_rt ?
ro6->ro_rt->rt_ifp :
NULL);
} else
#endif
{
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);
}
#if 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
#if NECP
necp_mark_packet_from_socket(m, tp ? tp->t_inpcb : NULL, 0, 0);
#endif
#if IPSEC
if (tp != NULL && tp->t_inpcb->inp_sp != NULL &&
ipsec_setsocket(m, tp ? tp->t_inpcb->inp_socket : NULL) != 0) {
m_freem(m);
return;
}
#endif
if (tp != NULL) {
u_int32_t svc_flags = 0;
if (isipv6) {
svc_flags |= PKT_SCF_IPV6;
}
set_packet_service_class(m, tp->t_inpcb->inp_socket,
MBUF_SC_UNSPEC, svc_flags);
m->m_pkthdr.pkt_flowsrc = FLOWSRC_INPCB;
m->m_pkthdr.pkt_flowid = tp->t_inpcb->inp_flowhash;
m->m_pkthdr.pkt_flags |= PKTF_FLOW_ID | PKTF_FLOW_LOCALSRC;
#if MPTCP
if (!(tp->t_mpflags & TMPF_MPTCP_TRUE))
#endif
m->m_pkthdr.pkt_flags |= PKTF_FLOW_ADV;
m->m_pkthdr.pkt_proto = IPPROTO_TCP;
}
#if INET6
if (isipv6) {
struct ip6_out_args ip6oa = { tra->ifscope, { 0 },
IP6OAF_SELECT_SRCIF | IP6OAF_BOUND_SRCADDR, 0 };
if (tra->ifscope != IFSCOPE_NONE)
ip6oa.ip6oa_flags |= IP6OAF_BOUND_IF;
if (tra->nocell)
ip6oa.ip6oa_flags |= IP6OAF_NO_CELLULAR;
if (tra->noexpensive)
ip6oa.ip6oa_flags |= IP6OAF_NO_EXPENSIVE;
if (tra->awdl_unrestricted)
ip6oa.ip6oa_flags |= IP6OAF_AWDL_UNRESTRICTED;
(void) ip6_output(m, NULL, ro6, IPV6_OUTARGS, NULL,
NULL, &ip6oa);
if (tp != NULL && ro6 != NULL && ro6->ro_rt != NULL &&
(outif = ro6->ro_rt->rt_ifp) !=
tp->t_inpcb->in6p_last_outifp)
tp->t_inpcb->in6p_last_outifp = outif;
if (ro6 == &sro6)
ROUTE_RELEASE(ro6);
} else
#endif
{
struct ip_out_args ipoa = { tra->ifscope, { 0 },
IPOAF_SELECT_SRCIF | IPOAF_BOUND_SRCADDR, 0 };
if (tra->ifscope != IFSCOPE_NONE)
ipoa.ipoa_flags |= IPOAF_BOUND_IF;
if (tra->nocell)
ipoa.ipoa_flags |= IPOAF_NO_CELLULAR;
if (tra->noexpensive)
ipoa.ipoa_flags |= IPOAF_NO_EXPENSIVE;
if (tra->awdl_unrestricted)
ipoa.ipoa_flags |= IPOAF_AWDL_UNRESTRICTED;
if (ro != &sro) {
inp_route_copyout(tp->t_inpcb, &sro);
}
(void) ip_output(m, NULL, &sro, IP_OUTARGS, NULL, &ipoa);
if (tp != NULL && sro.ro_rt != NULL &&
(outif = sro.ro_rt->rt_ifp) !=
tp->t_inpcb->inp_last_outifp)
tp->t_inpcb->inp_last_outifp = outif;
if (ro != &sro) {
inp_route_copyin(tp->t_inpcb, &sro);
} else {
ROUTE_RELEASE(&sro);
}
}
}
struct tcpcb *
tcp_newtcpcb(inp)
struct inpcb *inp;
{
struct inp_tp *it;
register struct tcpcb *tp;
register struct socket *so = inp->inp_socket;
#if INET6
int isipv6 = (inp->inp_vflag & INP_IPV6) != 0;
#endif
calculate_tcp_clock();
if ((so->so_flags1 & SOF1_CACHED_IN_SOCK_LAYER) == 0) {
it = (struct inp_tp *)(void *)inp;
tp = &it->tcb;
} else {
tp = (struct tcpcb *)(void *)inp->inp_saved_ppcb;
}
bzero((char *) tp, sizeof(struct tcpcb));
LIST_INIT(&tp->t_segq);
tp->t_maxseg = tp->t_maxopd =
#if INET6
isipv6 ? tcp_v6mssdflt :
#endif
tcp_mssdflt;
if (tcp_do_rfc1323)
tp->t_flags = (TF_REQ_SCALE|TF_REQ_TSTMP);
if (tcp_do_sack)
tp->t_flagsext |= TF_SACK_ENABLE;
TAILQ_INIT(&tp->snd_holes);
SLIST_INIT(&tp->t_rxt_segments);
tp->t_inpcb = inp;
tp->t_srtt = TCPTV_SRTTBASE;
tp->t_rttvar = ((TCPTV_RTOBASE - TCPTV_SRTTBASE) << TCP_RTTVAR_SHIFT) / 4;
tp->t_rttmin = tcp_TCPTV_MIN;
tp->t_rxtcur = TCPTV_RTOBASE;
if (tcp_use_newreno)
tp->tcp_cc_index = TCP_CC_ALGO_NEWRENO_INDEX;
else
tp->tcp_cc_index = TCP_CC_ALGO_CUBIC_INDEX;
tcp_cc_allocate_state(tp);
if (CC_ALGO(tp)->init != NULL)
CC_ALGO(tp)->init(tp);
tp->snd_cwnd = TCP_CC_CWND_INIT_BYTES;
tp->snd_ssthresh = TCP_MAXWIN << TCP_MAX_WINSHIFT;
tp->snd_ssthresh_prev = TCP_MAXWIN << TCP_MAX_WINSHIFT;
tp->t_rcvtime = tcp_now;
tp->tentry.timer_start = tcp_now;
tp->t_persist_timeout = tcp_max_persist_timeout;
tp->t_persist_stop = 0;
tp->t_flagsext |= TF_RCVUNACK_WAITSS;
tp->t_rexmtthresh = tcprexmtthresh;
tp->t_twentry.tqe_next = NULL;
tp->t_twentry.tqe_prev = NULL;
inp->inp_ip_ttl = ip_defttl;
inp->inp_ppcb = (caddr_t)tp;
return (tp);
}
struct tcpcb *
tcp_drop(tp, errno)
register struct tcpcb *tp;
int errno;
{
struct socket *so = tp->t_inpcb->inp_socket;
#if CONFIG_DTRACE
struct inpcb *inp = tp->t_inpcb;
#endif
if (TCPS_HAVERCVDSYN(tp->t_state)) {
DTRACE_TCP4(state__change, void, NULL, struct inpcb *, inp,
struct tcpcb *, tp, int32_t, TCPS_CLOSED);
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));
}
void
tcp_getrt_rtt(struct tcpcb *tp, struct rtentry *rt)
{
u_int32_t rtt = rt->rt_rmx.rmx_rtt;
int isnetlocal = (tp->t_flags & TF_LOCAL);
if (rtt != 0) {
if (rt->rt_rmx.rmx_locks & RTV_RTT)
tp->t_rttmin = rtt / (RTM_RTTUNIT / TCP_RETRANSHZ);
else
tp->t_rttmin = isnetlocal ? tcp_TCPTV_MIN : TCPTV_REXMTMIN;
tp->t_srtt = rtt / (RTM_RTTUNIT / (TCP_RETRANSHZ * TCP_RTT_SCALE));
tcpstat.tcps_usedrtt++;
if (rt->rt_rmx.rmx_rttvar) {
tp->t_rttvar = rt->rt_rmx.rmx_rttvar /
(RTM_RTTUNIT / (TCP_RETRANSHZ * TCP_RTTVAR_SCALE));
tcpstat.tcps_usedrttvar++;
} else {
tp->t_rttvar =
tp->t_srtt * TCP_RTTVAR_SCALE / TCP_RTT_SCALE;
}
TCPT_RANGESET(tp->t_rxtcur,
((tp->t_srtt >> 2) + tp->t_rttvar) >> 1,
tp->t_rttmin, TCPTV_REXMTMAX,
TCP_ADD_REXMTSLOP(tp));
}
}
static inline void
tcp_update_ecn_perf_stats(struct tcpcb *tp,
struct if_tcp_ecn_perf_stat *stat)
{
u_int64_t curval, oldval;
struct inpcb *inp = tp->t_inpcb;
stat->total_txpkts += inp->inp_stat->txpackets;
stat->total_rxpkts += inp->inp_stat->rxpackets;
stat->total_rxmitpkts += tp->t_stat.rxmitpkts;
stat->total_oopkts += tp->t_rcvoopack;
stat->total_reorderpkts += (tp->t_reordered_pkts + tp->t_pawsdrop +
tp->t_dsack_sent + tp->t_dsack_recvd);
curval = (tp->t_srtt >> TCP_RTT_SHIFT);
if (curval > 0 && tp->t_rttupdated >= 16) {
if (stat->rtt_avg == 0) {
stat->rtt_avg = curval;
} else {
oldval = stat->rtt_avg;
stat->rtt_avg =
((oldval << 4) - oldval + curval) >> 4;
}
}
curval = tp->t_rttvar >> TCP_RTTVAR_SHIFT;
if (curval > 0 && tp->t_rttupdated >= 16) {
if (stat->rtt_var == 0) {
stat->rtt_var = curval;
} else {
oldval = stat->rtt_var;
stat->rtt_var =
((oldval << 4) - oldval + curval) >> 4;
}
}
stat->sack_episodes += tp->t_sack_recovery_episode;
if (inp->inp_socket->so_error == ECONNRESET)
stat->rst_drop++;
return;
}
struct tcpcb *
tcp_close(tp)
register struct tcpcb *tp;
{
struct inpcb *inp = tp->t_inpcb;
struct socket *so = inp->inp_socket;
#if INET6
int isipv6 = (inp->inp_vflag & INP_IPV6) != 0;
#endif
struct route *ro;
struct rtentry *rt;
int dosavessthresh;
if (inp->inp_ppcb == NULL)
return(NULL);
tcp_canceltimers(tp);
KERNEL_DEBUG(DBG_FNC_TCP_CLOSE | DBG_FUNC_START, tp,0,0,0,0);
if ((tp->t_flags & TF_CLOSING) ||
inp->inp_sndinprog_cnt > 0) {
tp->t_flags |= TF_CLOSING;
return (NULL);
}
DTRACE_TCP4(state__change, void, NULL, struct inpcb *, inp,
struct tcpcb *, tp, int32_t, TCPS_CLOSED);
#if INET6
ro = (isipv6 ? (struct route *)&inp->in6p_route : &inp->inp_route);
#else
ro = &inp->inp_route;
#endif
rt = ro->ro_rt;
if (rt != NULL)
RT_LOCK_SPIN(rt);
if (tp->t_rttupdated >= 16) {
register u_int32_t i = 0;
#if INET6
if (isipv6) {
struct sockaddr_in6 *sin6;
if (rt == NULL)
goto no_valid_rt;
sin6 = (struct sockaddr_in6 *)(void *)rt_key(rt);
if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))
goto no_valid_rt;
}
else
#endif
if (ROUTE_UNUSABLE(ro) ||
SIN(rt_key(rt))->sin_addr.s_addr == INADDR_ANY) {
DTRACE_TCP4(state__change, void, NULL,
struct inpcb *, inp, struct tcpcb *, tp,
int32_t, TCPS_CLOSED);
tp->t_state = TCPS_CLOSED;
goto no_valid_rt;
}
RT_LOCK_ASSERT_HELD(rt);
if ((rt->rt_rmx.rmx_locks & RTV_RTT) == 0) {
i = tp->t_srtt *
(RTM_RTTUNIT / (TCP_RETRANSHZ * TCP_RTT_SCALE));
if (rt->rt_rmx.rmx_rtt && i)
rt->rt_rmx.rmx_rtt =
(rt->rt_rmx.rmx_rtt + i) / 2;
else
rt->rt_rmx.rmx_rtt = i;
tcpstat.tcps_cachedrtt++;
}
if ((rt->rt_rmx.rmx_locks & RTV_RTTVAR) == 0) {
i = tp->t_rttvar *
(RTM_RTTUNIT / (TCP_RETRANSHZ * TCP_RTTVAR_SCALE));
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++;
}
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);
if (((rt->rt_rmx.rmx_locks & RTV_SSTHRESH) == 0 &&
i != 0 && rt->rt_rmx.rmx_ssthresh != 0)
|| dosavessthresh) {
i = (i + tp->t_maxseg / 2) / tp->t_maxseg;
if (i < 2)
i = 2;
i *= (u_int32_t)(tp->t_maxseg +
#if INET6
(isipv6 ? sizeof (struct ip6_hdr) +
sizeof (struct tcphdr) :
#endif
sizeof (struct tcpiphdr)
#if INET6
)
#endif
);
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++;
}
}
if (rt != NULL && (so->so_flags & SOF_OVERFLOW) && tcp_lq_overflow) {
if (!(rt->rt_rmx.rmx_locks & RTV_RTT) &&
rt->rt_rmx.rmx_rtt == 0) {
rt->rt_flags |= RTF_DELCLONE;
}
}
no_valid_rt:
if (rt != NULL)
RT_UNLOCK(rt);
(void) tcp_freeq(tp);
if (tp->ecn_flags & TE_SETUPSENT) {
if (tp->ecn_flags & TE_CLIENT_SETUP) {
INP_INC_IFNET_STAT(inp, ecn_client_setup);
if (TCP_ECN_ENABLED(tp)) {
INP_INC_IFNET_STAT(inp,
ecn_client_success);
} else if (tp->ecn_flags & TE_LOST_SYN) {
INP_INC_IFNET_STAT(inp, ecn_syn_lost);
} else {
INP_INC_IFNET_STAT(inp,
ecn_peer_nosupport);
}
} else {
INP_INC_IFNET_STAT(inp, ecn_server_setup);
if (TCP_ECN_ENABLED(tp)) {
INP_INC_IFNET_STAT(inp,
ecn_server_success);
} else if (tp->ecn_flags & TE_LOST_SYNACK) {
INP_INC_IFNET_STAT(inp,
ecn_synack_lost);
} else {
INP_INC_IFNET_STAT(inp,
ecn_peer_nosupport);
}
}
} else {
INP_INC_IFNET_STAT(inp, ecn_off_conn);
}
if (TCP_ECN_ENABLED(tp)) {
if (tp->ecn_flags & TE_RECV_ECN_CE) {
tcpstat.tcps_ecn_conn_recv_ce++;
INP_INC_IFNET_STAT(inp, ecn_conn_recv_ce);
}
if (tp->ecn_flags & TE_RECV_ECN_ECE) {
tcpstat.tcps_ecn_conn_recv_ece++;
INP_INC_IFNET_STAT(inp, ecn_conn_recv_ece);
}
if (tp->ecn_flags & (TE_RECV_ECN_CE | TE_RECV_ECN_ECE)) {
if (tp->t_stat.txretransmitbytes > 0 ||
tp->t_stat.rxoutoforderbytes > 0) {
tcpstat.tcps_ecn_conn_pl_ce++;
INP_INC_IFNET_STAT(inp, ecn_conn_plce);
} else {
tcpstat.tcps_ecn_conn_nopl_ce++;
INP_INC_IFNET_STAT(inp, ecn_conn_noplce);
}
} else {
if (tp->t_stat.txretransmitbytes > 0 ||
tp->t_stat.rxoutoforderbytes > 0) {
tcpstat.tcps_ecn_conn_plnoce++;
INP_INC_IFNET_STAT(inp, ecn_conn_plnoce);
}
}
}
if (inp->inp_last_outifp != NULL && !(tp->t_flags & TF_LOCAL)) {
struct ifnet *ifp = inp->inp_last_outifp;
ifnet_lock_shared(ifp);
if ((ifp->if_refflags & (IFRF_ATTACHED | IFRF_DETACHING)) ==
IFRF_ATTACHED) {
if (inp->inp_vflag & INP_IPV6) {
ifp->if_ipv6_stat->timestamp = net_uptime();
if (TCP_ECN_ENABLED(tp)) {
tcp_update_ecn_perf_stats(tp,
&ifp->if_ipv6_stat->ecn_on);
} else {
tcp_update_ecn_perf_stats(tp,
&ifp->if_ipv6_stat->ecn_off);
}
} else {
ifp->if_ipv4_stat->timestamp = net_uptime();
if (TCP_ECN_ENABLED(tp)) {
tcp_update_ecn_perf_stats(tp,
&ifp->if_ipv4_stat->ecn_on);
} else {
tcp_update_ecn_perf_stats(tp,
&ifp->if_ipv4_stat->ecn_off);
}
}
}
ifnet_lock_done(ifp);
}
tcp_free_sackholes(tp);
if (tp->t_bwmeas != NULL) {
tcp_bwmeas_free(tp);
}
tcp_rxtseg_clean(tp);
if (tp->t_pktlist_head != NULL)
m_freem_list(tp->t_pktlist_head);
TCP_PKTLIST_CLEAR(tp);
#if MPTCP
if ((so->so_flags & SOF_MPTCP_TRUE) ||
(so->so_flags & SOF_MP_SUBFLOW)) {
soevent(so, (SO_FILT_HINT_LOCKED | SO_FILT_HINT_DELETEOK));
}
tp->t_mpflags = 0;
tp->t_mptcb = NULL;
#endif
if (so->so_flags1 & SOF1_CACHED_IN_SOCK_LAYER)
inp->inp_saved_ppcb = (caddr_t) tp;
tp->t_state = TCPS_CLOSED;
sodisconnectwakeup(so);
if (tp->t_flagsext & TF_LRO_OFFLOADED) {
tcp_lro_remove_state(inp->inp_laddr, inp->inp_faddr,
inp->inp_lport, inp->inp_fport);
tp->t_flagsext &= ~TF_LRO_OFFLOADED;
}
if ((so->so_options & SO_NOWAKEFROMSLEEP) &&
inp->inp_state != INPCB_STATE_DEAD &&
!(inp->inp_flags2 & INP2_TIMEWAIT))
socket_post_kev_msg_closed(so);
if (CC_ALGO(tp)->cleanup != NULL) {
CC_ALGO(tp)->cleanup(tp);
}
if (tp->t_ccstate != NULL) {
zfree(tcp_cc_zone, tp->t_ccstate);
tp->t_ccstate = NULL;
}
tp->tcp_cc_index = TCP_CC_ALGO_NONE;
if ((tp->t_tfo_flags & TFO_F_COOKIE_VALID)) {
OSDecrementAtomic(&tcp_tfo_halfcnt);
VERIFY(tcp_tfo_halfcnt >= 0);
tp->t_tfo_flags &= ~TFO_F_COOKIE_VALID;
}
#if INET6
if (SOCK_CHECK_DOM(so, PF_INET6))
in6_pcbdetach(inp);
else
#endif
in_pcbdetach(inp);
soisdisconnected(so);
tcpstat.tcps_closed++;
KERNEL_DEBUG(DBG_FNC_TCP_CLOSE | DBG_FUNC_END,
tcpstat.tcps_closed, 0, 0, 0, 0);
return(NULL);
}
int
tcp_freeq(tp)
struct tcpcb *tp;
{
register struct tseg_qent *q;
int rv = 0;
while((q = LIST_FIRST(&tp->t_segq)) != NULL) {
LIST_REMOVE(q, tqe_q);
m_freem(q->tqe_m);
zfree(tcp_reass_zone, q);
rv = 1;
}
tp->t_reassqlen = 0;
return (rv);
}
void
tcp_drain()
{
struct inpcb *inp;
struct tcpcb *tp;
if (!lck_rw_try_lock_exclusive(tcbinfo.ipi_lock))
return;
LIST_FOREACH(inp, tcbinfo.ipi_listhead, inp_list) {
if (in_pcb_checkstate(inp, WNT_ACQUIRE, 0) !=
WNT_STOPUSING) {
tcp_lock(inp->inp_socket, 1, 0);
if (in_pcb_checkstate(inp, WNT_RELEASE, 1)
== WNT_STOPUSING) {
tcp_unlock(inp->inp_socket, 1, 0);
continue;
}
tp = intotcpcb(inp);
if (do_tcpdrain)
tcp_freeq(tp);
so_drain_extended_bk_idle(inp->inp_socket);
tcp_unlock(inp->inp_socket, 1, 0);
}
}
lck_rw_done(tcbinfo.ipi_lock);
}
static void
tcp_notify(inp, error)
struct inpcb *inp;
int error;
{
struct tcpcb *tp;
if (inp == NULL || (inp->inp_state == INPCB_STATE_DEAD))
return;
tp = (struct tcpcb *)inp->inp_ppcb;
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);
else
tp->t_softerror = error;
#if 0
wakeup((caddr_t) &so->so_timeo);
sorwakeup(so);
sowwakeup(so);
#endif
}
struct bwmeas*
tcp_bwmeas_alloc(struct tcpcb *tp)
{
struct bwmeas *elm;
elm = zalloc(tcp_bwmeas_zone);
if (elm == NULL)
return(elm);
bzero(elm, bwmeas_elm_size);
elm->bw_minsizepkts = TCP_BWMEAS_BURST_MINSIZE;
elm->bw_maxsizepkts = TCP_BWMEAS_BURST_MAXSIZE;
elm->bw_minsize = elm->bw_minsizepkts * tp->t_maxseg;
elm->bw_maxsize = elm->bw_maxsizepkts * tp->t_maxseg;
return(elm);
}
void
tcp_bwmeas_free(struct tcpcb* tp)
{
zfree(tcp_bwmeas_zone, tp->t_bwmeas);
tp->t_bwmeas = NULL;
tp->t_flagsext &= ~(TF_MEASURESNDBW);
}
static void
tcpcb_to_otcpcb(struct tcpcb *tp, struct otcpcb *otp)
{
otp->t_segq = (uint32_t)VM_KERNEL_ADDRPERM(tp->t_segq.lh_first);
otp->t_dupacks = tp->t_dupacks;
otp->t_timer[TCPT_REXMT_EXT] = tp->t_timer[TCPT_REXMT];
otp->t_timer[TCPT_PERSIST_EXT] = tp->t_timer[TCPT_PERSIST];
otp->t_timer[TCPT_KEEP_EXT] = tp->t_timer[TCPT_KEEP];
otp->t_timer[TCPT_2MSL_EXT] = tp->t_timer[TCPT_2MSL];
otp->t_inpcb = (_TCPCB_PTR(struct inpcb *))VM_KERNEL_ADDRPERM(tp->t_inpcb);
otp->t_state = tp->t_state;
otp->t_flags = tp->t_flags;
otp->t_force = (tp->t_flagsext & TF_FORCE) ? 1 : 0;
otp->snd_una = tp->snd_una;
otp->snd_max = tp->snd_max;
otp->snd_nxt = tp->snd_nxt;
otp->snd_up = tp->snd_up;
otp->snd_wl1 = tp->snd_wl1;
otp->snd_wl2 = tp->snd_wl2;
otp->iss = tp->iss;
otp->irs = tp->irs;
otp->rcv_nxt = tp->rcv_nxt;
otp->rcv_adv = tp->rcv_adv;
otp->rcv_wnd = tp->rcv_wnd;
otp->rcv_up = tp->rcv_up;
otp->snd_wnd = tp->snd_wnd;
otp->snd_cwnd = tp->snd_cwnd;
otp->snd_ssthresh = tp->snd_ssthresh;
otp->t_maxopd = tp->t_maxopd;
otp->t_rcvtime = tp->t_rcvtime;
otp->t_starttime = tp->t_starttime;
otp->t_rtttime = tp->t_rtttime;
otp->t_rtseq = tp->t_rtseq;
otp->t_rxtcur = tp->t_rxtcur;
otp->t_maxseg = tp->t_maxseg;
otp->t_srtt = tp->t_srtt;
otp->t_rttvar = tp->t_rttvar;
otp->t_rxtshift = tp->t_rxtshift;
otp->t_rttmin = tp->t_rttmin;
otp->t_rttupdated = tp->t_rttupdated;
otp->max_sndwnd = tp->max_sndwnd;
otp->t_softerror = tp->t_softerror;
otp->t_oobflags = tp->t_oobflags;
otp->t_iobc = tp->t_iobc;
otp->snd_scale = tp->snd_scale;
otp->rcv_scale = tp->rcv_scale;
otp->request_r_scale = tp->request_r_scale;
otp->requested_s_scale = tp->requested_s_scale;
otp->ts_recent = tp->ts_recent;
otp->ts_recent_age = tp->ts_recent_age;
otp->last_ack_sent = tp->last_ack_sent;
otp->cc_send = 0;
otp->cc_recv = 0;
otp->snd_recover = tp->snd_recover;
otp->snd_cwnd_prev = tp->snd_cwnd_prev;
otp->snd_ssthresh_prev = tp->snd_ssthresh_prev;
otp->t_badrxtwin = 0;
}
static int
tcp_pcblist SYSCTL_HANDLER_ARGS
{
#pragma unused(oidp, arg1, arg2)
int error, i = 0, n;
struct inpcb *inp, **inp_list;
struct tcpcb *tp;
inp_gen_t gencnt;
struct xinpgen xig;
lck_rw_lock_shared(tcbinfo.ipi_lock);
if (req->oldptr == USER_ADDR_NULL) {
n = tcbinfo.ipi_count;
req->oldidx = 2 * (sizeof xig)
+ (n + n/8) * sizeof(struct xtcpcb);
lck_rw_done(tcbinfo.ipi_lock);
return 0;
}
if (req->newptr != USER_ADDR_NULL) {
lck_rw_done(tcbinfo.ipi_lock);
return EPERM;
}
gencnt = tcbinfo.ipi_gencnt;
n = tcbinfo.ipi_count;
bzero(&xig, sizeof(xig));
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) {
lck_rw_done(tcbinfo.ipi_lock);
return error;
}
if (n == 0) {
lck_rw_done(tcbinfo.ipi_lock);
return 0;
}
inp_list = _MALLOC(n * sizeof *inp_list, M_TEMP, M_WAITOK);
if (inp_list == 0) {
lck_rw_done(tcbinfo.ipi_lock);
return ENOMEM;
}
LIST_FOREACH(inp, tcbinfo.ipi_listhead, inp_list) {
if (inp->inp_gencnt <= gencnt &&
inp->inp_state != INPCB_STATE_DEAD)
inp_list[i++] = inp;
if (i >= n) break;
}
TAILQ_FOREACH(tp, &tcp_tw_tailq, t_twentry) {
inp = tp->t_inpcb;
if (inp->inp_gencnt <= gencnt &&
inp->inp_state != INPCB_STATE_DEAD)
inp_list[i++] = inp;
if (i >= n) break;
}
n = i;
error = 0;
for (i = 0; i < n; i++) {
inp = inp_list[i];
if (inp->inp_gencnt <= gencnt &&
inp->inp_state != INPCB_STATE_DEAD) {
struct xtcpcb xt;
caddr_t inp_ppcb;
bzero(&xt, sizeof(xt));
xt.xt_len = sizeof xt;
inpcb_to_compat(inp, &xt.xt_inp);
inp_ppcb = inp->inp_ppcb;
if (inp_ppcb != NULL) {
tcpcb_to_otcpcb(
(struct tcpcb *)(void *)inp_ppcb,
&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) {
bzero(&xig, sizeof(xig));
xig.xig_len = sizeof xig;
xig.xig_gen = tcbinfo.ipi_gencnt;
xig.xig_sogen = so_gencnt;
xig.xig_count = tcbinfo.ipi_count;
error = SYSCTL_OUT(req, &xig, sizeof xig);
}
FREE(inp_list, M_TEMP);
lck_rw_done(tcbinfo.ipi_lock);
return error;
}
SYSCTL_PROC(_net_inet_tcp, TCPCTL_PCBLIST, pcblist,
CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_LOCKED, 0, 0,
tcp_pcblist, "S,xtcpcb", "List of active TCP connections");
static void
tcpcb_to_xtcpcb64(struct tcpcb *tp, struct xtcpcb64 *otp)
{
otp->t_segq = (uint32_t)VM_KERNEL_ADDRPERM(tp->t_segq.lh_first);
otp->t_dupacks = tp->t_dupacks;
otp->t_timer[TCPT_REXMT_EXT] = tp->t_timer[TCPT_REXMT];
otp->t_timer[TCPT_PERSIST_EXT] = tp->t_timer[TCPT_PERSIST];
otp->t_timer[TCPT_KEEP_EXT] = tp->t_timer[TCPT_KEEP];
otp->t_timer[TCPT_2MSL_EXT] = tp->t_timer[TCPT_2MSL];
otp->t_state = tp->t_state;
otp->t_flags = tp->t_flags;
otp->t_force = (tp->t_flagsext & TF_FORCE) ? 1 : 0;
otp->snd_una = tp->snd_una;
otp->snd_max = tp->snd_max;
otp->snd_nxt = tp->snd_nxt;
otp->snd_up = tp->snd_up;
otp->snd_wl1 = tp->snd_wl1;
otp->snd_wl2 = tp->snd_wl2;
otp->iss = tp->iss;
otp->irs = tp->irs;
otp->rcv_nxt = tp->rcv_nxt;
otp->rcv_adv = tp->rcv_adv;
otp->rcv_wnd = tp->rcv_wnd;
otp->rcv_up = tp->rcv_up;
otp->snd_wnd = tp->snd_wnd;
otp->snd_cwnd = tp->snd_cwnd;
otp->snd_ssthresh = tp->snd_ssthresh;
otp->t_maxopd = tp->t_maxopd;
otp->t_rcvtime = tp->t_rcvtime;
otp->t_starttime = tp->t_starttime;
otp->t_rtttime = tp->t_rtttime;
otp->t_rtseq = tp->t_rtseq;
otp->t_rxtcur = tp->t_rxtcur;
otp->t_maxseg = tp->t_maxseg;
otp->t_srtt = tp->t_srtt;
otp->t_rttvar = tp->t_rttvar;
otp->t_rxtshift = tp->t_rxtshift;
otp->t_rttmin = tp->t_rttmin;
otp->t_rttupdated = tp->t_rttupdated;
otp->max_sndwnd = tp->max_sndwnd;
otp->t_softerror = tp->t_softerror;
otp->t_oobflags = tp->t_oobflags;
otp->t_iobc = tp->t_iobc;
otp->snd_scale = tp->snd_scale;
otp->rcv_scale = tp->rcv_scale;
otp->request_r_scale = tp->request_r_scale;
otp->requested_s_scale = tp->requested_s_scale;
otp->ts_recent = tp->ts_recent;
otp->ts_recent_age = tp->ts_recent_age;
otp->last_ack_sent = tp->last_ack_sent;
otp->cc_send = 0;
otp->cc_recv = 0;
otp->snd_recover = tp->snd_recover;
otp->snd_cwnd_prev = tp->snd_cwnd_prev;
otp->snd_ssthresh_prev = tp->snd_ssthresh_prev;
otp->t_badrxtwin = 0;
}
static int
tcp_pcblist64 SYSCTL_HANDLER_ARGS
{
#pragma unused(oidp, arg1, arg2)
int error, i = 0, n;
struct inpcb *inp, **inp_list;
struct tcpcb *tp;
inp_gen_t gencnt;
struct xinpgen xig;
lck_rw_lock_shared(tcbinfo.ipi_lock);
if (req->oldptr == USER_ADDR_NULL) {
n = tcbinfo.ipi_count;
req->oldidx = 2 * (sizeof xig)
+ (n + n/8) * sizeof(struct xtcpcb64);
lck_rw_done(tcbinfo.ipi_lock);
return 0;
}
if (req->newptr != USER_ADDR_NULL) {
lck_rw_done(tcbinfo.ipi_lock);
return EPERM;
}
gencnt = tcbinfo.ipi_gencnt;
n = tcbinfo.ipi_count;
bzero(&xig, sizeof(xig));
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) {
lck_rw_done(tcbinfo.ipi_lock);
return error;
}
if (n == 0) {
lck_rw_done(tcbinfo.ipi_lock);
return 0;
}
inp_list = _MALLOC(n * sizeof *inp_list, M_TEMP, M_WAITOK);
if (inp_list == 0) {
lck_rw_done(tcbinfo.ipi_lock);
return ENOMEM;
}
LIST_FOREACH(inp, tcbinfo.ipi_listhead, inp_list) {
if (inp->inp_gencnt <= gencnt &&
inp->inp_state != INPCB_STATE_DEAD)
inp_list[i++] = inp;
if (i >= n) break;
}
TAILQ_FOREACH(tp, &tcp_tw_tailq, t_twentry) {
inp = tp->t_inpcb;
if (inp->inp_gencnt <= gencnt &&
inp->inp_state != INPCB_STATE_DEAD)
inp_list[i++] = inp;
if (i >= n) break;
}
n = i;
error = 0;
for (i = 0; i < n; i++) {
inp = inp_list[i];
if (inp->inp_gencnt <= gencnt && inp->inp_state != INPCB_STATE_DEAD) {
struct xtcpcb64 xt;
bzero(&xt, sizeof(xt));
xt.xt_len = sizeof xt;
inpcb_to_xinpcb64(inp, &xt.xt_inpcb);
xt.xt_inpcb.inp_ppcb = (uint64_t)VM_KERNEL_ADDRPERM(inp->inp_ppcb);
if (inp->inp_ppcb != NULL)
tcpcb_to_xtcpcb64((struct tcpcb *)inp->inp_ppcb, &xt);
if (inp->inp_socket)
sotoxsocket64(inp->inp_socket, &xt.xt_inpcb.xi_socket);
error = SYSCTL_OUT(req, &xt, sizeof xt);
}
}
if (!error) {
bzero(&xig, sizeof(xig));
xig.xig_len = sizeof xig;
xig.xig_gen = tcbinfo.ipi_gencnt;
xig.xig_sogen = so_gencnt;
xig.xig_count = tcbinfo.ipi_count;
error = SYSCTL_OUT(req, &xig, sizeof xig);
}
FREE(inp_list, M_TEMP);
lck_rw_done(tcbinfo.ipi_lock);
return error;
}
SYSCTL_PROC(_net_inet_tcp, OID_AUTO, pcblist64,
CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_LOCKED, 0, 0,
tcp_pcblist64, "S,xtcpcb64", "List of active TCP connections");
static int
tcp_pcblist_n SYSCTL_HANDLER_ARGS
{
#pragma unused(oidp, arg1, arg2)
int error = 0;
error = get_pcblist_n(IPPROTO_TCP, req, &tcbinfo);
return error;
}
SYSCTL_PROC(_net_inet_tcp, OID_AUTO, pcblist_n,
CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_LOCKED, 0, 0,
tcp_pcblist_n, "S,xtcpcb_n", "List of active TCP connections");
__private_extern__ void
tcp_get_ports_used(uint32_t ifindex, int protocol, uint32_t flags,
bitstr_t *bitfield)
{
inpcb_get_ports_used(ifindex, protocol, flags,
bitfield, &tcbinfo);
}
__private_extern__ uint32_t
tcp_count_opportunistic(unsigned int ifindex, u_int32_t flags)
{
return inpcb_count_opportunistic(ifindex, &tcbinfo, flags);
}
__private_extern__ uint32_t
tcp_find_anypcb_byaddr(struct ifaddr *ifa)
{
return inpcb_find_anypcb_byaddr(ifa, &tcbinfo);
}
void
tcp_ctlinput(cmd, sa, vip)
int cmd;
struct sockaddr *sa;
void *vip;
{
tcp_seq icmp_tcp_seq;
struct ip *ip = vip;
struct in_addr faddr;
struct inpcb *inp;
struct tcpcb *tp;
void (*notify)(struct inpcb *, int) = tcp_notify;
faddr = ((struct sockaddr_in *)(void *)sa)->sin_addr;
if (sa->sa_family != AF_INET || faddr.s_addr == INADDR_ANY)
return;
if ((unsigned)cmd >= PRC_NCMDS)
return;
if (cmd == PRC_MSGSIZE)
notify = tcp_mtudisc;
else if (icmp_may_rst && (cmd == PRC_UNREACH_ADMIN_PROHIB ||
cmd == PRC_UNREACH_PORT) && ip)
notify = tcp_drop_syn_sent;
else if (PRC_IS_REDIRECT(cmd)) {
ip = 0;
notify = in_rtchange;
} else if (cmd == PRC_HOSTDEAD)
ip = 0;
else if (cmd == PRC_QUENCH)
return;
else if (inetctlerrmap[cmd] == 0)
return;
if (ip) {
struct tcphdr th;
struct icmp *icp;
icp = (struct icmp *)(void *)
((caddr_t)ip - offsetof(struct icmp, icmp_ip));
bcopy(((caddr_t)ip + (IP_VHL_HL(ip->ip_vhl) << 2)),
&th, sizeof (th));
inp = in_pcblookup_hash(&tcbinfo, faddr, th.th_dport,
ip->ip_src, th.th_sport, 0, NULL);
if (inp != NULL && inp->inp_socket != NULL) {
tcp_lock(inp->inp_socket, 1, 0);
if (in_pcb_checkstate(inp, WNT_RELEASE, 1) == WNT_STOPUSING) {
tcp_unlock(inp->inp_socket, 1, 0);
return;
}
icmp_tcp_seq = htonl(th.th_seq);
tp = intotcpcb(inp);
if (SEQ_GEQ(icmp_tcp_seq, tp->snd_una) &&
SEQ_LT(icmp_tcp_seq, tp->snd_max)) {
if (cmd == PRC_MSGSIZE) {
struct rtentry *rt;
int mtu;
struct sockaddr_in icmpsrc = { sizeof (struct sockaddr_in), AF_INET,
0 , { 0 }, { 0,0,0,0,0,0,0,0 } };
icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
rt = rtalloc1((struct sockaddr *)&icmpsrc, 0,
RTF_CLONING | RTF_PRCLONING);
if (rt != NULL) {
RT_LOCK(rt);
if ((rt->rt_flags & RTF_HOST) &&
!(rt->rt_rmx.rmx_locks & RTV_MTU)) {
mtu = ntohs(icp->icmp_nextmtu);
if (!mtu)
mtu = ip_next_mtu(rt->rt_rmx.
rmx_mtu, 1);
#if DEBUG_MTUDISC
printf("MTU for %s reduced to %d\n",
inet_ntop(AF_INET,
&icmpsrc.sin_addr, ipv4str,
sizeof (ipv4str)), mtu);
#endif
if (mtu < max(296, (tcp_minmss +
sizeof (struct tcpiphdr)))) {
rt->rt_rmx.rmx_locks |= RTV_MTU;
} else if (rt->rt_rmx.rmx_mtu > mtu) {
rt->rt_rmx.rmx_mtu = mtu;
}
}
RT_UNLOCK(rt);
rtfree(rt);
}
}
(*notify)(inp, inetctlerrmap[cmd]);
}
tcp_unlock(inp->inp_socket, 1, 0);
}
} else
in_pcbnotifyall(&tcbinfo, faddr, inetctlerrmap[cmd], notify);
}
#if INET6
void
tcp6_ctlinput(cmd, sa, d)
int cmd;
struct sockaddr *sa;
void *d;
{
struct tcphdr th;
void (*notify)(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 ((unsigned)cmd >= PRC_NCMDS)
return;
if (cmd == PRC_MSGSIZE)
notify = tcp_mtudisc;
else if (!PRC_IS_REDIRECT(cmd) && (inet6ctlerrmap[cmd] == 0))
return;
else if (cmd == PRC_QUENCH)
return;
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;
sa6_src = &sa6_any;
}
if (ip6) {
if (m->m_pkthdr.len < off + sizeof(*thp))
return;
bzero(&th, sizeof(th));
m_copydata(m, off, sizeof(*thp), (caddr_t)&th);
in6_pcbnotify(&tcbinfo, sa, th.th_dport,
(struct sockaddr *)ip6cp->ip6c_src,
th.th_sport, cmd, NULL, notify);
} else {
in6_pcbnotify(&tcbinfo, sa, 0,
(struct sockaddr *)(size_t)sa6_src, 0, cmd, NULL, notify);
}
}
#endif
#define ISN_BYTES_PER_SECOND 1048576
tcp_seq
tcp_new_isn(tp)
struct tcpcb *tp;
{
u_int32_t md5_buffer[4];
tcp_seq new_isn;
struct timeval timenow;
u_char isn_secret[32];
int isn_last_reseed = 0;
MD5_CTX isn_ctx;
if (((tp->t_state == TCPS_LISTEN) || (tp->t_state == TCPS_TIME_WAIT))
&& tcp_strict_rfc1948 == 0)
#ifdef __APPLE__
return RandomULong();
#else
return arc4random();
#endif
getmicrotime(&timenow);
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)timenow.tv_sec))) {
#ifdef __APPLE__
read_random(&isn_secret, sizeof(isn_secret));
#else
read_random_unlimited(&isn_secret, sizeof(isn_secret));
#endif
isn_last_reseed = timenow.tv_sec;
}
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));
#if 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 += timenow.tv_sec * (ISN_BYTES_PER_SECOND / hz);
return new_isn;
}
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);
}
void
tcp_mtudisc(
struct inpcb *inp,
__unused int errno
)
{
struct tcpcb *tp = intotcpcb(inp);
struct rtentry *rt;
struct rmxp_tao *taop;
struct socket *so = inp->inp_socket;
int offered;
int mss;
u_int32_t mtu;
#if INET6
int isipv6 = (tp->t_inpcb->inp_vflag & INP_IPV6) != 0;
#endif
if (tp) {
#if INET6
if (isipv6)
rt = tcp_rtlookup6(inp, IFSCOPE_NONE);
else
#endif
rt = tcp_rtlookup(inp, IFSCOPE_NONE);
if (!rt || !rt->rt_rmx.rmx_mtu) {
tp->t_maxopd = tp->t_maxseg =
#if INET6
isipv6 ? tcp_v6mssdflt :
#endif
tcp_mssdflt;
if (rt != NULL)
RT_UNLOCK(rt);
return;
}
taop = rmx_taop(rt->rt_rmx);
offered = taop->tao_mssopt;
mtu = rt->rt_rmx.rmx_mtu;
RT_UNLOCK(rt);
#if NECP
mtu = necp_socket_get_effective_mtu(inp, mtu);
#endif
mss = mtu -
#if INET6
(isipv6 ?
sizeof(struct ip6_hdr) + sizeof(struct tcphdr) :
#endif
sizeof(struct tcpiphdr)
#if INET6
)
#endif
;
if (offered)
mss = min(mss, offered);
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 MPTCP
mss -= mptcp_adj_mss(tp, TRUE);
#endif
if (so->so_snd.sb_hiwat < mss)
mss = so->so_snd.sb_hiwat;
tp->t_maxseg = mss;
if (CC_ALGO(tp)->cwnd_init != NULL)
CC_ALGO(tp)->cwnd_init(tp);
tcpstat.tcps_mturesent++;
tp->t_rtttime = 0;
tp->snd_nxt = tp->snd_una;
tcp_output(tp);
}
}
struct rtentry *
tcp_rtlookup(inp, input_ifscope)
struct inpcb *inp;
unsigned int input_ifscope;
{
struct route *ro;
struct rtentry *rt;
struct tcpcb *tp;
lck_mtx_assert(rnh_lock, LCK_MTX_ASSERT_NOTOWNED);
ro = &inp->inp_route;
if ((rt = ro->ro_rt) != NULL)
RT_LOCK(rt);
if (ROUTE_UNUSABLE(ro)) {
if (rt != NULL) {
RT_UNLOCK(rt);
rt = NULL;
}
ROUTE_RELEASE(ro);
if (inp->inp_faddr.s_addr != INADDR_ANY) {
unsigned int ifscope;
ro->ro_dst.sa_family = AF_INET;
ro->ro_dst.sa_len = sizeof(struct sockaddr_in);
((struct sockaddr_in *)(void *)&ro->ro_dst)->sin_addr =
inp->inp_faddr;
ifscope = (inp->inp_flags & INP_BOUND_IF) ?
inp->inp_boundifp->if_index : input_ifscope;
rtalloc_scoped(ro, ifscope);
if ((rt = ro->ro_rt) != NULL)
RT_LOCK(rt);
}
}
if (rt != NULL)
RT_LOCK_ASSERT_HELD(rt);
tp = intotcpcb(inp);
if (!path_mtu_discovery || ((rt != NULL) &&
(!(rt->rt_flags & RTF_UP) || (rt->rt_rmx.rmx_locks & RTV_MTU))))
tp->t_flags &= ~TF_PMTUD;
else
tp->t_flags |= TF_PMTUD;
#if CONFIG_IFEF_NOWINDOWSCALE
if (tcp_obey_ifef_nowindowscale &&
tp->t_state == TCPS_SYN_SENT && rt != NULL && rt->rt_ifp != NULL &&
(rt->rt_ifp->if_eflags & IFEF_NOWINDOWSCALE)) {
tp->t_flags &= ~TF_REQ_SCALE;
}
#endif
if (rt != NULL && rt->rt_ifp != NULL) {
somultipages(inp->inp_socket,
(rt->rt_ifp->if_hwassist & IFNET_MULTIPAGES));
tcp_set_tso(tp, rt->rt_ifp);
soif2kcl(inp->inp_socket,
(rt->rt_ifp->if_eflags & IFEF_2KCL));
tcp_set_ecn(tp, rt->rt_ifp);
}
if (rt != NULL && !(rt->rt_ifp->if_flags & IFF_POINTOPOINT) &&
(rt->rt_gateway->sa_family == AF_LINK ||
rt->rt_ifp->if_flags & IFF_LOOPBACK ||
in_localaddr(inp->inp_faddr))) {
tp->t_flags |= TF_LOCAL;
}
return rt;
}
#if INET6
struct rtentry *
tcp_rtlookup6(inp, input_ifscope)
struct inpcb *inp;
unsigned int input_ifscope;
{
struct route_in6 *ro6;
struct rtentry *rt;
struct tcpcb *tp;
lck_mtx_assert(rnh_lock, LCK_MTX_ASSERT_NOTOWNED);
ro6 = &inp->in6p_route;
if ((rt = ro6->ro_rt) != NULL)
RT_LOCK(rt);
if (ROUTE_UNUSABLE(ro6)) {
if (rt != NULL) {
RT_UNLOCK(rt);
rt = NULL;
}
ROUTE_RELEASE(ro6);
if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
struct sockaddr_in6 *dst6;
unsigned int ifscope;
dst6 = (struct sockaddr_in6 *)&ro6->ro_dst;
dst6->sin6_family = AF_INET6;
dst6->sin6_len = sizeof(*dst6);
dst6->sin6_addr = inp->in6p_faddr;
ifscope = (inp->inp_flags & INP_BOUND_IF) ?
inp->inp_boundifp->if_index : input_ifscope;
rtalloc_scoped((struct route *)ro6, ifscope);
if ((rt = ro6->ro_rt) != NULL)
RT_LOCK(rt);
}
}
if (rt != NULL)
RT_LOCK_ASSERT_HELD(rt);
tp = intotcpcb(inp);
if (!path_mtu_discovery || ((rt != NULL) &&
(!(rt->rt_flags & RTF_UP) || (rt->rt_rmx.rmx_locks & RTV_MTU))))
tp->t_flags &= ~TF_PMTUD;
else
tp->t_flags |= TF_PMTUD;
#if CONFIG_IFEF_NOWINDOWSCALE
if (tcp_obey_ifef_nowindowscale &&
tp->t_state == TCPS_SYN_SENT && rt != NULL && rt->rt_ifp != NULL &&
(rt->rt_ifp->if_eflags & IFEF_NOWINDOWSCALE)) {
tp->t_flags &= ~TF_REQ_SCALE;
}
#endif
if (rt != NULL && rt->rt_ifp != NULL) {
somultipages(inp->inp_socket,
(rt->rt_ifp->if_hwassist & IFNET_MULTIPAGES));
tcp_set_tso(tp, rt->rt_ifp);
soif2kcl(inp->inp_socket,
(rt->rt_ifp->if_eflags & IFEF_2KCL));
tcp_set_ecn(tp, rt->rt_ifp);
}
if (rt != NULL && !(rt->rt_ifp->if_flags & IFF_POINTOPOINT) &&
(IN6_IS_ADDR_LOOPBACK(&inp->in6p_faddr) ||
IN6_IS_ADDR_LINKLOCAL(&inp->in6p_faddr) ||
rt->rt_gateway->sa_family == AF_LINK ||
in6_localaddr(&inp->in6p_faddr))) {
tp->t_flags |= TF_LOCAL;
}
return rt;
}
#endif
#if IPSEC
size_t
ipsec_hdrsiz_tcp(tp)
struct tcpcb *tp;
{
struct inpcb *inp;
struct mbuf *m;
size_t hdrsiz;
struct ip *ip;
#if INET6
struct ip6_hdr *ip6 = NULL;
#endif
struct tcphdr *th;
if ((tp == NULL) || ((inp = tp->t_inpcb) == NULL))
return 0;
MGETHDR(m, M_DONTWAIT, MT_DATA);
if (!m)
return 0;
#if INET6
if ((inp->inp_vflag & INP_IPV6) != 0) {
ip6 = mtod(m, struct ip6_hdr *);
th = (struct tcphdr *)(void *)(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
{
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
struct rmxp_tao *
tcp_gettaocache(inp)
struct inpcb *inp;
{
struct rtentry *rt;
struct rmxp_tao *taop;
#if INET6
if ((inp->inp_vflag & INP_IPV6) != 0)
rt = tcp_rtlookup6(inp, IFSCOPE_NONE);
else
#endif
rt = tcp_rtlookup(inp, IFSCOPE_NONE);
if (rt == NULL ||
(rt->rt_flags & (RTF_UP|RTF_HOST)) != (RTF_UP|RTF_HOST)) {
if (rt != NULL)
RT_UNLOCK(rt);
return NULL;
}
taop = rmx_taop(rt->rt_rmx);
RT_UNLOCK(rt);
return (taop);
}
static void
tcp_cleartaocache()
{
}
int
tcp_lock(struct socket *so, int refcount, void *lr)
{
void *lr_saved;
if (lr == NULL)
lr_saved = __builtin_return_address(0);
else
lr_saved = lr;
if (so->so_pcb != NULL) {
lck_mtx_lock(&((struct inpcb *)so->so_pcb)->inpcb_mtx);
} else {
panic("tcp_lock: so=%p NO PCB! lr=%p lrh= %s\n",
so, lr_saved, solockhistory_nr(so));
}
if (so->so_usecount < 0) {
panic("tcp_lock: so=%p so_pcb=%p lr=%p ref=%x lrh= %s\n",
so, so->so_pcb, lr_saved, so->so_usecount, solockhistory_nr(so));
}
if (refcount)
so->so_usecount++;
so->lock_lr[so->next_lock_lr] = lr_saved;
so->next_lock_lr = (so->next_lock_lr+1) % SO_LCKDBG_MAX;
return (0);
}
int
tcp_unlock(struct socket *so, int refcount, void *lr)
{
void *lr_saved;
if (lr == NULL)
lr_saved = __builtin_return_address(0);
else
lr_saved = lr;
#ifdef MORE_TCPLOCK_DEBUG
printf("tcp_unlock: so=0x%llx sopcb=0x%llx lock=0x%llx ref=%x "
"lr=0x%llx\n", (uint64_t)VM_KERNEL_ADDRPERM(so),
(uint64_t)VM_KERNEL_ADDRPERM(so->so_pcb),
(uint64_t)VM_KERNEL_ADDRPERM(&(sotoinpcb(so)->inpcb_mtx)),
so->so_usecount, (uint64_t)VM_KERNEL_ADDRPERM(lr_saved));
#endif
if (refcount)
so->so_usecount--;
if (so->so_usecount < 0) {
panic("tcp_unlock: so=%p usecount=%x lrh= %s\n",
so, so->so_usecount, solockhistory_nr(so));
}
if (so->so_pcb == NULL) {
panic("tcp_unlock: so=%p NO PCB usecount=%x lr=%p lrh= %s\n",
so, so->so_usecount, lr_saved, solockhistory_nr(so));
} else {
lck_mtx_assert(&((struct inpcb *)so->so_pcb)->inpcb_mtx,
LCK_MTX_ASSERT_OWNED);
so->unlock_lr[so->next_unlock_lr] = lr_saved;
so->next_unlock_lr = (so->next_unlock_lr+1) % SO_LCKDBG_MAX;
lck_mtx_unlock(&((struct inpcb *)so->so_pcb)->inpcb_mtx);
}
return (0);
}
lck_mtx_t *
tcp_getlock(
struct socket *so,
__unused int locktype)
{
struct inpcb *inp = sotoinpcb(so);
if (so->so_pcb) {
if (so->so_usecount < 0)
panic("tcp_getlock: so=%p usecount=%x lrh= %s\n",
so, so->so_usecount, solockhistory_nr(so));
return(&inp->inpcb_mtx);
}
else {
panic("tcp_getlock: so=%p NULL so_pcb %s\n",
so, solockhistory_nr(so));
return (so->so_proto->pr_domain->dom_mtx);
}
}
static void
tcp_sbrcv_grow_rwin(struct tcpcb *tp, struct sockbuf *sb)
{
u_int32_t rcvbufinc = tp->t_maxseg << 4;
u_int32_t rcvbuf = sb->sb_hiwat;
struct socket *so = tp->t_inpcb->inp_socket;
if (so->so_flags & SOF_ENABLE_MSGS)
rcvbuf = rcvbuf - so->so_msg_state->msg_uno_bytes;
if (tcp_do_autorcvbuf == 1 &&
tcp_cansbgrow(sb) &&
(tp->t_flags & TF_SLOWLINK) == 0 &&
(so->so_flags1 & SOF1_EXTEND_BK_IDLE_WANTED) == 0 &&
(rcvbuf - sb->sb_cc) < rcvbufinc &&
rcvbuf < tcp_autorcvbuf_max &&
(sb->sb_idealsize > 0 &&
sb->sb_hiwat <= (sb->sb_idealsize + rcvbufinc))) {
sbreserve(sb,
min((sb->sb_hiwat + rcvbufinc), tcp_autorcvbuf_max));
}
}
int32_t
tcp_sbspace(struct tcpcb *tp)
{
struct sockbuf *sb = &tp->t_inpcb->inp_socket->so_rcv;
u_int32_t rcvbuf = sb->sb_hiwat;
int32_t space;
struct socket *so = tp->t_inpcb->inp_socket;
int32_t pending = 0;
if (so->so_flags & SOF_ENABLE_MSGS)
rcvbuf = rcvbuf - so->so_msg_state->msg_uno_bytes;
tcp_sbrcv_grow_rwin(tp, sb);
space = ((int32_t) imin((rcvbuf - sb->sb_cc),
(sb->sb_mbmax - sb->sb_mbcnt)));
if (space < 0)
space = 0;
#if CONTENT_FILTER
pending = cfil_sock_data_space(sb);
#endif
if (pending > space)
space = 0;
else
space -= pending;
if (space < tp->t_maxseg)
return space;
if (((tp->t_flags & TF_SLOWLINK) != 0) && slowlink_wsize > 0 )
return imin(space, slowlink_wsize);
return space;
}
void
tcp_set_tso(struct tcpcb *tp, struct ifnet *ifp)
{
#if INET6
struct inpcb *inp;
int isipv6;
#endif
#if MPTCP
if (tp->t_mpflags & TMPF_MPTCP_TRUE) {
tp->t_flags &= ~TF_TSO;
return;
}
#endif
#if INET6
inp = tp->t_inpcb;
isipv6 = (inp->inp_vflag & INP_IPV6) != 0;
if (isipv6) {
if (ifp && (ifp->if_hwassist & IFNET_TSO_IPV6)) {
tp->t_flags |= TF_TSO;
if (ifp->if_tso_v6_mtu != 0)
tp->tso_max_segment_size = ifp->if_tso_v6_mtu;
else
tp->tso_max_segment_size = TCP_MAXWIN;
} else
tp->t_flags &= ~TF_TSO;
} else
#endif
{
if (ifp && (ifp->if_hwassist & IFNET_TSO_IPV4)) {
tp->t_flags |= TF_TSO;
if (ifp->if_tso_v4_mtu != 0)
tp->tso_max_segment_size = ifp->if_tso_v4_mtu;
else
tp->tso_max_segment_size = TCP_MAXWIN;
} else
tp->t_flags &= ~TF_TSO;
}
}
#define TIMEVAL_TO_TCPHZ(_tv_) ((_tv_).tv_sec * TCP_RETRANSHZ + (_tv_).tv_usec / TCP_RETRANSHZ_TO_USEC)
void
calculate_tcp_clock()
{
struct timeval tv = tcp_uptime;
struct timeval interval = {0, TCP_RETRANSHZ_TO_USEC};
struct timeval now, hold_now;
uint32_t incr = 0;
microuptime(&now);
net_update_uptime_secs(now.tv_sec);
timevaladd(&tv, &interval);
if (timevalcmp(&now, &tv, >)) {
lck_spin_lock(tcp_uptime_lock);
if (timevalcmp(&tcp_uptime, &now, >=)) {
lck_spin_unlock(tcp_uptime_lock);
return;
}
microuptime(&now);
hold_now = now;
tv = tcp_uptime;
timevalsub(&now, &tv);
incr = TIMEVAL_TO_TCPHZ(now);
if (incr > 0) {
tcp_uptime = hold_now;
tcp_now += incr;
}
lck_spin_unlock(tcp_uptime_lock);
}
return;
}
void
tcp_set_max_rwinscale(struct tcpcb *tp, struct socket *so) {
u_int32_t maxsockbufsize;
if (!tcp_do_rfc1323) {
tp->request_r_scale = 0;
return;
}
tp->request_r_scale = max(tcp_win_scale, tp->request_r_scale);
maxsockbufsize = ((so->so_rcv.sb_flags & SB_USRSIZE) != 0) ?
so->so_rcv.sb_hiwat : tcp_autorcvbuf_max;
while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
(TCP_MAXWIN << tp->request_r_scale) < maxsockbufsize)
tp->request_r_scale++;
tp->request_r_scale = min(tp->request_r_scale, TCP_MAX_WINSHIFT);
}
int
tcp_notsent_lowat_check(struct socket *so) {
struct inpcb *inp = sotoinpcb(so);
struct tcpcb *tp = NULL;
int notsent = 0;
if (inp != NULL) {
tp = intotcpcb(inp);
}
notsent = so->so_snd.sb_cc -
(tp->snd_nxt - tp->snd_una);
if (notsent <= tp->t_notsent_lowat) {
return(1);
}
if ((tp->t_flags & TF_NODELAY) == 0 &&
notsent > 0 && notsent < tp->t_maxseg) {
return(1);
}
return(0);
}
void
tcp_rxtseg_insert(struct tcpcb *tp, tcp_seq start, tcp_seq end) {
struct tcp_rxt_seg *rxseg = NULL, *prev = NULL, *next = NULL;
u_int32_t rxcount = 0;
if (SLIST_EMPTY(&tp->t_rxt_segments))
tp->t_dsack_lastuna = tp->snd_una;
SLIST_FOREACH(rxseg, &tp->t_rxt_segments, rx_link) {
if (SEQ_GT(rxseg->rx_start, start))
break;
prev = rxseg;
}
next = rxseg;
if (prev != NULL && SEQ_LEQ(prev->rx_start, start) &&
SEQ_GEQ(prev->rx_end, end)) {
prev->rx_count++;
return;
}
if (prev != NULL && SEQ_GT(prev->rx_end, start)) {
if (prev->rx_start == start && SEQ_GT(end, prev->rx_end)) {
start = prev->rx_end + 1;
prev->rx_count++;
} else {
prev->rx_end = (start - 1);
rxcount = prev->rx_count;
}
}
if (next != NULL && SEQ_LT(next->rx_start, end)) {
if (SEQ_LEQ(next->rx_end, end)) {
end = next->rx_start - 1;
next->rx_count++;
} else {
next->rx_start = end + 1;
rxcount = next->rx_count;
}
}
if (!SEQ_LT(start, end))
return;
rxseg = (struct tcp_rxt_seg *) zalloc(tcp_rxt_seg_zone);
if (rxseg == NULL) {
return;
}
bzero(rxseg, sizeof(*rxseg));
rxseg->rx_start = start;
rxseg->rx_end = end;
rxseg->rx_count = rxcount + 1;
if (prev != NULL) {
SLIST_INSERT_AFTER(prev, rxseg, rx_link);
} else {
SLIST_INSERT_HEAD(&tp->t_rxt_segments, rxseg, rx_link);
}
return;
}
struct tcp_rxt_seg *
tcp_rxtseg_find(struct tcpcb *tp, tcp_seq start, tcp_seq end)
{
struct tcp_rxt_seg *rxseg;
if (SLIST_EMPTY(&tp->t_rxt_segments))
return (NULL);
SLIST_FOREACH(rxseg, &tp->t_rxt_segments, rx_link) {
if (SEQ_LEQ(rxseg->rx_start, start) &&
SEQ_GEQ(rxseg->rx_end, end))
return (rxseg);
if (SEQ_GT(rxseg->rx_start, start))
break;
}
return (NULL);
}
void
tcp_rxtseg_clean(struct tcpcb *tp)
{
struct tcp_rxt_seg *rxseg, *next;
SLIST_FOREACH_SAFE(rxseg, &tp->t_rxt_segments, rx_link, next) {
SLIST_REMOVE(&tp->t_rxt_segments, rxseg,
tcp_rxt_seg, rx_link);
zfree(tcp_rxt_seg_zone, rxseg);
}
tp->t_dsack_lastuna = tp->snd_max;
}
boolean_t
tcp_rxtseg_detect_bad_rexmt(struct tcpcb *tp, tcp_seq th_ack)
{
boolean_t bad_rexmt;
struct tcp_rxt_seg *rxseg;
if (SLIST_EMPTY(&tp->t_rxt_segments))
return (FALSE);
if (SEQ_LT(th_ack, tp->snd_recover))
return (FALSE);
bad_rexmt = TRUE;
SLIST_FOREACH(rxseg, &tp->t_rxt_segments, rx_link) {
if (rxseg->rx_count > 1 ||
!(rxseg->rx_flags & TCP_RXT_SPURIOUS)) {
bad_rexmt = FALSE;
break;
}
}
return (bad_rexmt);
}
boolean_t
tcp_rxtseg_dsack_for_tlp(struct tcpcb *tp)
{
boolean_t dsack_for_tlp = FALSE;
struct tcp_rxt_seg *rxseg;
if (SLIST_EMPTY(&tp->t_rxt_segments))
return (FALSE);
SLIST_FOREACH(rxseg, &tp->t_rxt_segments, rx_link) {
if (rxseg->rx_count == 1 &&
SLIST_NEXT(rxseg,rx_link) == NULL &&
(rxseg->rx_flags & TCP_RXT_DSACK_FOR_TLP)) {
dsack_for_tlp = TRUE;
break;
}
}
return (dsack_for_tlp);
}
u_int32_t
tcp_rxtseg_total_size(struct tcpcb *tp) {
struct tcp_rxt_seg *rxseg;
u_int32_t total_size = 0;
SLIST_FOREACH(rxseg, &tp->t_rxt_segments, rx_link) {
total_size += (rxseg->rx_end - rxseg->rx_start) + 1;
}
return (total_size);
}
void
tcp_get_connectivity_status(struct tcpcb *tp,
struct tcp_conn_status *connstatus)
{
if (tp == NULL || connstatus == NULL)
return;
bzero(connstatus, sizeof(*connstatus));
if (tp->t_rxtshift >= TCP_CONNECTIVITY_PROBES_MAX) {
if (TCPS_HAVEESTABLISHED(tp->t_state)) {
connstatus->write_probe_failed = 1;
} else {
connstatus->conn_probe_failed = 1;
}
}
if (tp->t_rtimo_probes >= TCP_CONNECTIVITY_PROBES_MAX)
connstatus->read_probe_failed = 1;
if (tp->t_inpcb != NULL && tp->t_inpcb->inp_last_outifp != NULL
&& (tp->t_inpcb->inp_last_outifp->if_eflags & IFEF_PROBE_CONNECTIVITY))
connstatus->probe_activated = 1;
return;
}
boolean_t
tfo_enabled(const struct tcpcb *tp)
{
return !!(tp->t_flagsext & TF_FASTOPEN);
}
void
tcp_disable_tfo(struct tcpcb *tp)
{
tp->t_flagsext &= ~TF_FASTOPEN;
}