#include <sys/param.h>
#include <sys/systm.h>
#include <sys/kernel.h>
#include <sys/sysctl.h>
#include <sys/malloc.h>
#include <sys/mbuf.h>
#include <sys/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>
#include <net/ntstat.h>
#include <net/multi_layer_pkt_log.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>
#include <netinet/ip6.h>
#include <netinet/icmp6.h>
#include <netinet/in_pcb.h>
#include <netinet6/in6_pcb.h>
#include <netinet/in_var.h>
#include <netinet/ip_var.h>
#include <netinet/icmp_var.h>
#include <netinet6/ip6_var.h>
#include <netinet/mptcp_var.h>
#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>
#include <netinet6/tcp6_var.h>
#include <netinet/tcpip.h>
#if TCPDEBUG
#include <netinet/tcp_debug.h>
#endif
#include <netinet/tcp_log.h>
#include <netinet6/ip6protosw.h>
#if IPSEC
#include <netinet6/ipsec.h>
#include <netinet6/ipsec6.h>
#endif
#if NECP
#include <net/necp.h>
#endif
#undef tcp_minmssoverload
#include <corecrypto/ccaes.h>
#include <libkern/crypto/aes.h>
#include <libkern/crypto/md5.h>
#include <sys/kdebug.h>
#include <mach/sdt.h>
#include <atm/atm_internal.h>
#include <pexpert/pexpert.h>
#define DBG_FNC_TCP_CLOSE NETDBG_CODE(DBG_NETTCP, ((5 << 8) | 2))
static tcp_cc tcp_ccgen;
extern struct tcptimerlist tcp_timer_list;
extern struct tcptailq tcp_tw_tailq;
SYSCTL_SKMEM_TCP_INT(TCPCTL_MSSDFLT, mssdflt, CTLFLAG_RW | CTLFLAG_LOCKED,
int, tcp_mssdflt, TCP_MSS, "Default TCP Maximum Segment Size");
SYSCTL_SKMEM_TCP_INT(TCPCTL_V6MSSDFLT, v6mssdflt,
CTLFLAG_RW | CTLFLAG_LOCKED, int, tcp_v6mssdflt, TCP6_MSS,
"Default TCP Maximum Segment Size for IPv6");
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;
SYSCTL_SKMEM_TCP_INT(OID_AUTO, fastopen_backlog,
CTLFLAG_RW | CTLFLAG_LOCKED, int, tcp_tfo_backlog, 10,
"Backlog queue for half-open TFO connections");
SYSCTL_SKMEM_TCP_INT(OID_AUTO, fastopen, CTLFLAG_RW | CTLFLAG_LOCKED,
int, tcp_fastopen, TCP_FASTOPEN_CLIENT | TCP_FASTOPEN_SERVER,
"Enable TCP Fastopen (RFC 7413)");
SYSCTL_SKMEM_TCP_INT(OID_AUTO, now_init, CTLFLAG_RD | CTLFLAG_LOCKED,
uint32_t, tcp_now_init, 0, "Initial tcp now value");
SYSCTL_SKMEM_TCP_INT(OID_AUTO, microuptime_init, CTLFLAG_RD | CTLFLAG_LOCKED,
uint32_t, tcp_microuptime_init, 0, "Initial tcp uptime value in micro seconds");
SYSCTL_SKMEM_TCP_INT(OID_AUTO, minmss, CTLFLAG_RW | CTLFLAG_LOCKED,
int, tcp_minmss, TCP_MINMSS, "Minmum TCP Maximum Segment Size");
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");
SYSCTL_SKMEM_TCP_INT(OID_AUTO, icmp_may_rst, CTLFLAG_RW | CTLFLAG_LOCKED,
static int, icmp_may_rst, 1,
"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
SYSCTL_SKMEM_TCP_INT(OID_AUTO, rtt_min, CTLFLAG_RW | CTLFLAG_LOCKED,
int, tcp_TCPTV_MIN, 100, "min rtt value allowed");
SYSCTL_SKMEM_TCP_INT(OID_AUTO, rexmt_slop, CTLFLAG_RW,
int, tcp_rexmt_slop, TCPTV_REXMTSLOP, "Slop added to retransmit timeout");
SYSCTL_SKMEM_TCP_INT(OID_AUTO, randomize_ports, CTLFLAG_RW | CTLFLAG_LOCKED,
__private_extern__ int, tcp_use_randomport, 0,
"Randomize TCP port numbers");
SYSCTL_SKMEM_TCP_INT(OID_AUTO, win_scale_factor, CTLFLAG_RW | CTLFLAG_LOCKED,
__private_extern__ int, tcp_win_scale, 3, "Window scaling factor");
#if (DEVELOPMENT || DEBUG)
SYSCTL_SKMEM_TCP_INT(OID_AUTO, init_rtt_from_cache,
CTLFLAG_RW | CTLFLAG_LOCKED, static int, tcp_init_rtt_from_cache, 1,
"Initalize RTT from route cache");
#else
SYSCTL_SKMEM_TCP_INT(OID_AUTO, init_rtt_from_cache,
CTLFLAG_RD | CTLFLAG_LOCKED, static int, tcp_init_rtt_from_cache, 1,
"Initalize RTT from route cache");
#endif
static int tso_debug = 0;
SYSCTL_INT(_net_inet_tcp, OID_AUTO, tso_debug, CTLFLAG_RW | CTLFLAG_LOCKED,
&tso_debug, 0, "TSO verbosity");
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;
static void tcp_sbrcv_grow_rwin(struct tcpcb *tp, struct sockbuf *sb);
#define TCP_BWMEAS_BURST_MINSIZE 6
#define TCP_BWMEAS_BURST_MAXSIZE 25
#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);
os_log_t tcp_mpkl_log_object = NULL;
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 void tcp_flow_lim_stats(struct ifnet_stats_per_flow *ifs,
struct if_lim_perf_stat *stat);
static void tcp_flow_ecn_perf_stats(struct ifnet_stats_per_flow *ifs,
struct if_tcp_ecn_perf_stat *stat);
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];
int isipv6 = inp->inp_vflag & INP_IPV6;
VERIFY(blk_size == CCAES_BLOCK_SIZE);
bzero(&in[0], CCAES_BLOCK_SIZE);
bzero(&out[0], CCAES_BLOCK_SIZE);
if (isipv6) {
memcpy(in, &inp->in6p_faddr, sizeof(struct in6_addr));
} else {
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);
}
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(void)
{
u_char key[TCP_FASTOPEN_KEYLEN];
read_frandom(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;
uint32_t logging_config;
VERIFY((pp->pr_flags & (PR_INITIALIZED | PR_ATTACHED)) == PR_ATTACHED);
if (tcp_initialized) {
return;
}
tcp_initialized = 1;
tcp_ccgen = 1;
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_frandom(&tcp_now, sizeof(tcp_now));
tcp_now = tcp_now & 0x3fffffff;
tcp_now_init = tcp_now;
tcp_microuptime_init =
(uint32_t)(tcp_uptime.tv_usec + (tcp_uptime.tv_sec * USEC_PER_SEC));
SYSCTL_SKMEM_UPDATE_FIELD(tcp.microuptime_init, tcp_microuptime_init);
SYSCTL_SKMEM_UPDATE_FIELD(tcp.now_init, tcp_now_init);
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 = (vm_size_t)P2ROUNDUP(sizeof(struct inp_tp), sizeof(u_int64_t));
tcbinfo.ipi_zone = zone_create("tcpcb", str_size, ZC_NONE);
tcbinfo.ipi_gc = tcp_gc;
tcbinfo.ipi_timer = tcp_itimer;
in_pcbinfo_attach(&tcbinfo);
str_size = (vm_size_t)P2ROUNDUP(sizeof(struct sackhole), sizeof(u_int64_t));
sack_hole_zone = zone_create("sack_hole zone", str_size, ZC_NONE);
str_size = (vm_size_t)P2ROUNDUP(sizeof(struct tseg_qent), sizeof(u_int64_t));
tcp_reass_zone = zone_create("tcp_reass_zone", str_size, ZC_NONE);
str_size = (vm_size_t)P2ROUNDUP(sizeof(struct bwmeas), sizeof(u_int64_t));
tcp_bwmeas_zone = zone_create("tcp_bwmeas_zone", str_size, ZC_ZFREE_CLEARMEM);
str_size = (vm_size_t)P2ROUNDUP(sizeof(struct tcp_ccstate), sizeof(u_int64_t));
tcp_cc_zone = zone_create("tcp_cc_zone", str_size, ZC_NONE);
str_size = (vm_size_t)P2ROUNDUP(sizeof(struct tcp_rxt_seg), sizeof(u_int64_t));
tcp_rxt_seg_zone = zone_create("tcp_rxt_seg_zone", str_size, ZC_NONE);
#define TCP_MINPROTOHDR (sizeof(struct ip6_hdr) + sizeof(struct tcphdr))
if (max_protohdr < TCP_MINPROTOHDR) {
_max_protohdr = TCP_MINPROTOHDR;
_max_protohdr = (int)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");
}
;
tcp_timer_list.call = thread_call_allocate(tcp_run_timerlist, NULL);
if (tcp_timer_list.call == 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_cache_init();
tcp_mpkl_log_object = MPKL_CREATE_LOGOBJECT("com.apple.xnu.tcp");
if (tcp_mpkl_log_object == NULL) {
panic("MPKL_CREATE_LOGOBJECT failed");
}
logging_config = atm_get_diagnostic_config();
if (logging_config & 0x80000000) {
tcp_log_privacy = 1;
}
PE_parse_boot_argn("tcp_log", &tcp_log_enable_flags, sizeof(tcp_log_enable_flags));
if (mem_actual >= (1ULL << (GBSHIFT + 2))) {
tcp_autorcvbuf_max = 4 * 1024 * 1024;
tcp_autosndbuf_max = 4 * 1024 * 1024;
SYSCTL_SKMEM_UPDATE_FIELD(tcp.autorcvbufmax, tcp_autorcvbuf_max);
SYSCTL_SKMEM_UPDATE_FIELD(tcp.autosndbufmax, tcp_autosndbuf_max);
}
}
void
tcp_fillheaders(struct tcpcb *tp, void *ip_ptr, void *tcp_ptr)
{
struct inpcb *inp = tp->t_inpcb;
struct tcphdr *tcp_hdr = (struct tcphdr *)tcp_ptr;
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_plen = htons(sizeof(struct tcphdr));
ip6->ip6_nxt = IPPROTO_TCP;
ip6->ip6_hlim = 0;
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 {
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(struct tcpcb *tp)
{
struct mbuf *m;
struct tcptemp *n;
m = m_get(M_DONTWAIT, MT_HEADER);
if (m == NULL) {
return NULL;
}
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, uint8_t flags, struct tcp_respond_args *tra)
{
uint16_t tlen;
int win = 0;
struct route *ro = 0;
struct route sro;
struct ip *ip;
struct tcphdr *nth;
struct route_in6 *ro6 = 0;
struct route_in6 sro6;
struct ip6_hdr *ip6;
int isipv6;
struct ifnet *outif;
int sotc = SO_TC_UNSPEC;
bool check_qos_marking_again = FALSE;
isipv6 = IP_VHL_V(((struct ip *)ipgen)->ip_vhl) == 6;
ip6 = ipgen;
ip = ipgen;
if (tp) {
check_qos_marking_again = tp->t_inpcb->inp_socket->so_flags1 & SOF1_QOSMARKING_POLICY_OVERRIDE ? FALSE : TRUE;
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 (isipv6) {
ro6 = &tp->t_inpcb->in6p_route;
} else {
ro = &tp->t_inpcb->inp_route;
}
} else {
if (isipv6) {
ro6 = &sro6;
bzero(ro6, sizeof(*ro6));
} else {
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 (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 {
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 (isipv6) {
IP6_HDR_STRICT_ALIGNMENT_CHECK(ip6);
xchg(ip6->ip6_dst, ip6->ip6_src, struct in6_addr);
nth = (struct tcphdr *)(void *)(ip6 + 1);
} else {
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 (isipv6) {
ip6->ip6_plen = htons((u_short)(sizeof(struct tcphdr) +
tlen));
tlen += sizeof(struct ip6_hdr) + sizeof(struct tcphdr);
} else {
tlen += sizeof(struct tcpiphdr);
ip->ip_len = tlen;
ip->ip_ttl = (uint8_t)ip_defttl;
}
m->m_len = tlen;
m->m_pkthdr.len = tlen;
m->m_pkthdr.rcvif = 0;
if (tra->keep_alive) {
m->m_pkthdr.pkt_flags |= PKTF_KEEPALIVE;
}
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 (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 {
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, 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;
}
sotc = tp->t_inpcb->inp_socket->so_traffic_class;
if ((flags & TH_RST) == 0) {
set_packet_service_class(m, tp->t_inpcb->inp_socket,
sotc, svc_flags);
} else {
m_set_service_class(m, MBUF_SC_BK_SYS);
}
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 | PKTF_FLOW_ADV);
m->m_pkthdr.pkt_proto = IPPROTO_TCP;
m->m_pkthdr.tx_tcp_pid = tp->t_inpcb->inp_socket->last_pid;
m->m_pkthdr.tx_tcp_e_pid = tp->t_inpcb->inp_socket->e_pid;
if (flags & TH_RST) {
m->m_pkthdr.comp_gencnt = tp->t_comp_gencnt;
}
} else {
if (flags & TH_RST) {
m->m_pkthdr.comp_gencnt = TCP_ACK_COMPRESSION_DUMMY;
m_set_service_class(m, MBUF_SC_BK_SYS);
}
}
if (isipv6) {
struct ip6_out_args ip6oa;
bzero(&ip6oa, sizeof(ip6oa));
ip6oa.ip6oa_boundif = tra->ifscope;
ip6oa.ip6oa_flags = IP6OAF_SELECT_SRCIF | IP6OAF_BOUND_SRCADDR;
ip6oa.ip6oa_sotc = SO_TC_UNSPEC;
ip6oa.ip6oa_netsvctype = _NET_SERVICE_TYPE_UNSPEC;
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->noconstrained) {
ip6oa.ip6oa_flags |= IP6OAF_NO_CONSTRAINED;
}
if (tra->awdl_unrestricted) {
ip6oa.ip6oa_flags |= IP6OAF_AWDL_UNRESTRICTED;
}
if (tra->intcoproc_allowed) {
ip6oa.ip6oa_flags |= IP6OAF_INTCOPROC_ALLOWED;
}
ip6oa.ip6oa_sotc = sotc;
if (tp != NULL) {
if ((tp->t_inpcb->inp_socket->so_flags1 & SOF1_QOSMARKING_ALLOWED)) {
ip6oa.ip6oa_flags |= IP6OAF_QOSMARKING_ALLOWED;
}
ip6oa.qos_marking_gencount = tp->t_inpcb->inp_policyresult.results.qos_marking_gencount;
if (check_qos_marking_again) {
ip6oa.ip6oa_flags |= IP6OAF_REDO_QOSMARKING_POLICY;
}
ip6oa.ip6oa_netsvctype = tp->t_inpcb->inp_socket->so_netsvctype;
}
(void) ip6_output(m, NULL, ro6, IPV6_OUTARGS, NULL,
NULL, &ip6oa);
if (check_qos_marking_again) {
struct inpcb *inp = tp->t_inpcb;
inp->inp_policyresult.results.qos_marking_gencount = ip6oa.qos_marking_gencount;
if (ip6oa.ip6oa_flags & IP6OAF_QOSMARKING_ALLOWED) {
inp->inp_socket->so_flags1 |= SOF1_QOSMARKING_ALLOWED;
} else {
inp->inp_socket->so_flags1 &= ~SOF1_QOSMARKING_ALLOWED;
}
}
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 {
struct ip_out_args ipoa;
bzero(&ipoa, sizeof(ipoa));
ipoa.ipoa_boundif = tra->ifscope;
ipoa.ipoa_flags = IPOAF_SELECT_SRCIF | IPOAF_BOUND_SRCADDR;
ipoa.ipoa_sotc = SO_TC_UNSPEC;
ipoa.ipoa_netsvctype = _NET_SERVICE_TYPE_UNSPEC;
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->noconstrained) {
ipoa.ipoa_flags |= IPOAF_NO_CONSTRAINED;
}
if (tra->awdl_unrestricted) {
ipoa.ipoa_flags |= IPOAF_AWDL_UNRESTRICTED;
}
ipoa.ipoa_sotc = sotc;
if (tp != NULL) {
if ((tp->t_inpcb->inp_socket->so_flags1 & SOF1_QOSMARKING_ALLOWED)) {
ipoa.ipoa_flags |= IPOAF_QOSMARKING_ALLOWED;
}
if (!(tp->t_inpcb->inp_socket->so_flags1 & SOF1_QOSMARKING_POLICY_OVERRIDE)) {
ipoa.ipoa_flags |= IPOAF_REDO_QOSMARKING_POLICY;
}
ipoa.qos_marking_gencount = tp->t_inpcb->inp_policyresult.results.qos_marking_gencount;
ipoa.ipoa_netsvctype = tp->t_inpcb->inp_socket->so_netsvctype;
}
if (ro != &sro) {
inp_route_copyout(tp->t_inpcb, &sro);
}
(void) ip_output(m, NULL, &sro, IP_OUTARGS, NULL, &ipoa);
if (check_qos_marking_again) {
struct inpcb *inp = tp->t_inpcb;
inp->inp_policyresult.results.qos_marking_gencount = ipoa.qos_marking_gencount;
if (ipoa.ipoa_flags & IPOAF_QOSMARKING_ALLOWED) {
inp->inp_socket->so_flags1 |= SOF1_QOSMARKING_ALLOWED;
} else {
inp->inp_socket->so_flags1 &= ~SOF1_QOSMARKING_ALLOWED;
}
}
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(struct inpcb *inp)
{
struct inp_tp *it;
struct tcpcb *tp;
struct socket *so = inp->inp_socket;
int isipv6 = (inp->inp_vflag & INP_IPV6) != 0;
uint32_t random_32;
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 = isipv6 ? tcp_v6mssdflt : tcp_mssdflt;
tp->t_flags = (TF_REQ_SCALE | TF_REQ_TSTMP);
tp->t_flagsext |= TF_SACK_ENABLE;
TAILQ_INIT(&tp->snd_holes);
SLIST_INIT(&tp->t_rxt_segments);
SLIST_INIT(&tp->t_notify_ack);
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_initial_cwnd(tp);
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->rcv_unackwin = tcp_now;
tp->t_persist_timeout = tcp_max_persist_timeout;
tp->t_persist_stop = 0;
tp->t_flagsext |= TF_RCVUNACK_WAITSS;
tp->t_rexmtthresh = (uint8_t)tcprexmtthresh;
tp->rfbuf_ts = tcp_now;
tp->rfbuf_space = tcp_initial_cwnd(tp);
tp->t_forced_acks = TCP_FORCED_ACKS_COUNT;
tp->t_flagsext |= TF_MEASURESNDBW;
if (tp->t_bwmeas == NULL) {
tp->t_bwmeas = tcp_bwmeas_alloc(tp);
if (tp->t_bwmeas == NULL) {
tp->t_flagsext &= ~TF_MEASURESNDBW;
}
}
tp->t_twentry.tqe_next = NULL;
tp->t_twentry.tqe_prev = NULL;
read_frandom(&random_32, sizeof(random_32));
if (__probable(tcp_do_ack_compression)) {
tp->t_comp_gencnt = random_32;
if (tp->t_comp_gencnt <= TCP_ACK_COMPRESSION_DUMMY) {
tp->t_comp_gencnt = TCP_ACK_COMPRESSION_DUMMY + 1;
}
tp->t_comp_lastinc = tcp_now;
}
if (__probable(tcp_randomize_timestamps)) {
tp->t_ts_offset = random_32;
}
inp->inp_ip_ttl = (uint8_t)ip_defttl;
inp->inp_ppcb = (caddr_t)tp;
return tp;
}
struct tcpcb *
tcp_drop(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 = (u_short)errno;
TCP_LOG_CONNECTION_SUMMARY(tp);
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);
TCP_LOG_RTM_RTT(tp, rt);
if (rtt != 0 && tcp_init_rtt_from_cache != 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));
}
TCP_LOG_RTT_INFO(tp);
}
static inline void
tcp_create_ifnet_stats_per_flow(struct tcpcb *tp,
struct ifnet_stats_per_flow *ifs)
{
struct inpcb *inp;
struct socket *so;
if (tp == NULL || ifs == NULL) {
return;
}
bzero(ifs, sizeof(*ifs));
inp = tp->t_inpcb;
so = inp->inp_socket;
ifs->ipv4 = (inp->inp_vflag & INP_IPV6) ? 0 : 1;
ifs->local = (tp->t_flags & TF_LOCAL) ? 1 : 0;
ifs->connreset = (so->so_error == ECONNRESET) ? 1 : 0;
ifs->conntimeout = (so->so_error == ETIMEDOUT) ? 1 : 0;
ifs->ecn_flags = tp->ecn_flags;
ifs->txretransmitbytes = tp->t_stat.txretransmitbytes;
ifs->rxoutoforderbytes = tp->t_stat.rxoutoforderbytes;
ifs->rxmitpkts = tp->t_stat.rxmitpkts;
ifs->rcvoopack = tp->t_rcvoopack;
ifs->pawsdrop = tp->t_pawsdrop;
ifs->sack_recovery_episodes = tp->t_sack_recovery_episode;
ifs->reordered_pkts = tp->t_reordered_pkts;
ifs->dsack_sent = tp->t_dsack_sent;
ifs->dsack_recvd = tp->t_dsack_recvd;
ifs->srtt = tp->t_srtt;
ifs->rttupdated = tp->t_rttupdated;
ifs->rttvar = tp->t_rttvar;
ifs->rttmin = get_base_rtt(tp);
if (tp->t_bwmeas != NULL && tp->t_bwmeas->bw_sndbw_max > 0) {
ifs->bw_sndbw_max = tp->t_bwmeas->bw_sndbw_max;
} else {
ifs->bw_sndbw_max = 0;
}
if (tp->t_bwmeas != NULL && tp->t_bwmeas->bw_rcvbw_max > 0) {
ifs->bw_rcvbw_max = tp->t_bwmeas->bw_rcvbw_max;
} else {
ifs->bw_rcvbw_max = 0;
}
ifs->bk_txpackets = so->so_tc_stats[MBUF_TC_BK].txpackets;
ifs->txpackets = inp->inp_stat->txpackets;
ifs->rxpackets = inp->inp_stat->rxpackets;
}
static inline void
tcp_flow_ecn_perf_stats(struct ifnet_stats_per_flow *ifs,
struct if_tcp_ecn_perf_stat *stat)
{
u_int64_t curval, oldval;
stat->total_txpkts += ifs->txpackets;
stat->total_rxpkts += ifs->rxpackets;
stat->total_rxmitpkts += ifs->rxmitpkts;
stat->total_oopkts += ifs->rcvoopack;
stat->total_reorderpkts += (ifs->reordered_pkts +
ifs->pawsdrop + ifs->dsack_sent + ifs->dsack_recvd);
curval = ifs->srtt >> TCP_RTT_SHIFT;
if (curval > 0 && ifs->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 = ifs->rttvar >> TCP_RTTVAR_SHIFT;
if (curval > 0 && ifs->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 += ifs->sack_recovery_episodes;
if (ifs->connreset) {
stat->rst_drop++;
}
}
static inline void
tcp_flow_lim_stats(struct ifnet_stats_per_flow *ifs,
struct if_lim_perf_stat *stat)
{
u_int64_t curval, oldval;
stat->lim_total_txpkts += ifs->txpackets;
stat->lim_total_rxpkts += ifs->rxpackets;
stat->lim_total_retxpkts += ifs->rxmitpkts;
stat->lim_total_oopkts += ifs->rcvoopack;
if (ifs->bw_sndbw_max > 0) {
ifs->bw_sndbw_max *= 8000;
stat->lim_ul_max_bandwidth = MAX(stat->lim_ul_max_bandwidth,
ifs->bw_sndbw_max);
}
if (ifs->bw_rcvbw_max > 0) {
ifs->bw_rcvbw_max *= 8000;
stat->lim_dl_max_bandwidth = MAX(stat->lim_dl_max_bandwidth,
ifs->bw_rcvbw_max);
}
curval = ifs->srtt >> TCP_RTT_SHIFT;
if (curval > 0 && ifs->rttupdated >= 16) {
if (stat->lim_rtt_average == 0) {
stat->lim_rtt_average = curval;
} else {
oldval = stat->lim_rtt_average;
stat->lim_rtt_average =
((oldval << 4) - oldval + curval) >> 4;
}
}
curval = ifs->rttvar >> TCP_RTTVAR_SHIFT;
if (curval > 0 && ifs->rttupdated >= 16) {
if (stat->lim_rtt_variance == 0) {
stat->lim_rtt_variance = curval;
} else {
oldval = stat->lim_rtt_variance;
stat->lim_rtt_variance =
((oldval << 4) - oldval + curval) >> 4;
}
}
if (stat->lim_rtt_min == 0) {
stat->lim_rtt_min = ifs->rttmin;
} else {
stat->lim_rtt_min = MIN(stat->lim_rtt_min, ifs->rttmin);
}
stat->lim_conn_attempts++;
if (ifs->conntimeout) {
stat->lim_conn_timeouts++;
}
stat->lim_bk_txpkts += ifs->bk_txpackets;
}
struct tcpcb *
tcp_close(struct tcpcb *tp)
{
struct inpcb *inp = tp->t_inpcb;
struct socket *so = inp->inp_socket;
int isipv6 = (inp->inp_vflag & INP_IPV6) != 0;
struct route *ro;
struct rtentry *rt;
int dosavessthresh;
struct ifnet_stats_per_flow ifs;
if (inp->inp_ppcb == NULL) {
return NULL;
}
tcp_del_fsw_flow(tp);
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;
}
TCP_LOG_CONNECTION_SUMMARY(tp);
DTRACE_TCP4(state__change, void, NULL, struct inpcb *, inp,
struct tcpcb *, tp, int32_t, TCPS_CLOSED);
ro = (isipv6 ? (struct route *)&inp->in6p_route : &inp->inp_route);
rt = ro->ro_rt;
if (rt != NULL) {
RT_LOCK_SPIN(rt);
}
if (tp->t_rttupdated >= 16) {
u_int32_t i = 0;
bool log_rtt = false;
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 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++;
log_rtt = true;
}
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++;
log_rtt = true;
}
if (log_rtt) {
TCP_LOG_RTM_RTT(tp, rt);
TCP_LOG_RTT_INFO(tp);
}
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 +
isipv6 ? sizeof(struct ip6_hdr) +
sizeof(struct tcphdr) :
sizeof(struct tcpiphdr));
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)) {
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);
tcp_create_ifnet_stats_per_flow(tp, &ifs);
tcp_update_stats_per_flow(&ifs, inp->inp_last_outifp);
tcp_free_sackholes(tp);
tcp_notify_ack_free(tp);
inp_decr_sndbytes_allunsent(so, tp->snd_una);
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 (so->so_flags1 & SOF1_CACHED_IN_SOCK_LAYER) {
inp->inp_saved_ppcb = (caddr_t) tp;
}
tp->t_state = TCPS_CLOSED;
sodisconnectwakeup(so);
tcp_clear_keep_alive_offload(so);
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 (SOCK_CHECK_DOM(so, PF_INET6)) {
in6_pcbdetach(inp);
} else {
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(struct tcpcb *tp)
{
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(void)
{
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) {
socket_lock(inp->inp_socket, 1);
if (in_pcb_checkstate(inp, WNT_RELEASE, 1)
== WNT_STOPUSING) {
socket_unlock(inp->inp_socket, 1);
continue;
}
tp = intotcpcb(inp);
so_drain_extended_bk_idle(inp->inp_socket);
socket_unlock(inp->inp_socket, 1);
}
}
lck_rw_done(tcbinfo.ipi_lock);
}
static void
tcp_notify(struct inpcb *inp, int error)
{
struct tcpcb *tp;
if (inp == NULL || (inp->inp_state == INPCB_STATE_DEAD)) {
return;
}
tp = (struct tcpcb *)inp->inp_ppcb;
VERIFY(tp != NULL);
if (tp->t_state == TCPS_ESTABLISHED &&
(error == EHOSTUNREACH || error == ENETUNREACH ||
error == EHOSTDOWN)) {
if (inp->inp_route.ro_rt) {
rtfree(inp->inp_route.ro_rt);
inp->inp_route.ro_rt = (struct rtentry *)NULL;
}
} else if (tp->t_state < TCPS_ESTABLISHED && tp->t_rxtshift > 3 &&
tp->t_softerror) {
tcp_drop(tp, error);
} else {
tp->t_softerror = error;
}
}
struct bwmeas *
tcp_bwmeas_alloc(struct tcpcb *tp)
{
struct bwmeas *elm;
elm = zalloc_flags(tcp_bwmeas_zone, Z_ZERO | Z_WAITOK);
elm->bw_minsizepkts = TCP_BWMEAS_BURST_MINSIZE;
elm->bw_minsize = elm->bw_minsizepkts * 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);
}
int
get_tcp_inp_list(struct inpcb **inp_list, int n, inp_gen_t gencnt)
{
struct tcpcb *tp;
struct inpcb *inp;
int i = 0;
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;
}
}
return i;
}
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_list;
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;
}
n = get_tcp_inp_list(inp_list, n, gencnt);
error = 0;
for (i = 0; i < n; i++) {
struct xtcpcb xt;
caddr_t inp_ppcb;
struct inpcb *inp;
inp = inp_list[i];
if (in_pcb_checkstate(inp, WNT_ACQUIRE, 0) == WNT_STOPUSING) {
continue;
}
socket_lock(inp->inp_socket, 1);
if (in_pcb_checkstate(inp, WNT_RELEASE, 1) == WNT_STOPUSING) {
socket_unlock(inp->inp_socket, 1);
continue;
}
if (inp->inp_gencnt > gencnt) {
socket_unlock(inp->inp_socket, 1);
continue;
}
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);
}
socket_unlock(inp->inp_socket, 1);
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");
#if XNU_TARGET_OS_OSX
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_list;
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;
}
n = get_tcp_inp_list(inp_list, n, gencnt);
error = 0;
for (i = 0; i < n; i++) {
struct xtcpcb64 xt;
struct inpcb *inp;
inp = inp_list[i];
if (in_pcb_checkstate(inp, WNT_ACQUIRE, 0) == WNT_STOPUSING) {
continue;
}
socket_lock(inp->inp_socket, 1);
if (in_pcb_checkstate(inp, WNT_RELEASE, 1) == WNT_STOPUSING) {
socket_unlock(inp->inp_socket, 1);
continue;
}
if (inp->inp_gencnt > gencnt) {
socket_unlock(inp->inp_socket, 1);
continue;
}
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);
}
socket_unlock(inp->inp_socket, 1);
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");
#endif
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");
static int
tcp_progress_indicators SYSCTL_HANDLER_ARGS
{
#pragma unused(oidp, arg1, arg2)
return ntstat_tcp_progress_indicators(req);
}
SYSCTL_PROC(_net_inet_tcp, OID_AUTO, progress,
CTLTYPE_STRUCT | CTLFLAG_RW | CTLFLAG_LOCKED | CTLFLAG_ANYBODY, 0, 0,
tcp_progress_indicators, "S", "Various items that indicate the current state of progress on the link");
__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);
}
static void
tcp_handle_msgsize(struct ip *ip, struct inpcb *inp)
{
struct rtentry *rt = NULL;
u_short ifscope = IFSCOPE_NONE;
int mtu;
struct sockaddr_in icmpsrc = {
.sin_len = sizeof(struct sockaddr_in),
.sin_family = AF_INET, .sin_port = 0, .sin_addr = { .s_addr = 0 },
.sin_zero = { 0, 0, 0, 0, 0, 0, 0, 0 }
};
struct icmp *icp = NULL;
icp = (struct icmp *)(void *)
((caddr_t)ip - offsetof(struct icmp, icmp_ip));
icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
if (ROUTE_UNUSABLE(&(inp->inp_route)) == false) {
rt = inp->inp_route.ro_rt;
}
if (rt != NULL) {
ifscope = (rt->rt_ifp != NULL) ?
rt->rt_ifp->if_index : IFSCOPE_NONE;
} else {
ifscope = (inp->inp_last_outifp != NULL) ?
inp->inp_last_outifp->if_index : IFSCOPE_NONE;
}
if ((rt == NULL) ||
!(rt->rt_flags & RTF_HOST) ||
(rt->rt_flags & (RTF_CLONING | RTF_PRCLONING))) {
rt = rtalloc1_scoped((struct sockaddr *)&icmpsrc, 0,
RTF_CLONING | RTF_PRCLONING, ifscope);
} else if (rt) {
RT_LOCK(rt);
rtref(rt);
RT_UNLOCK(rt);
}
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 == 0) {
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);
}
}
void
tcp_ctlinput(int cmd, struct sockaddr *sa, void *vip, __unused struct ifnet *ifp)
{
tcp_seq icmp_tcp_seq;
struct ip *ip = vip;
struct in_addr faddr;
struct inpcb *inp;
struct tcpcb *tp;
struct tcphdr *th;
struct icmp *icp;
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_QUENCH) {
return;
}
if (cmd == PRC_MSGSIZE) {
notify = tcp_mtudisc;
} else if (icmp_may_rst && (cmd == PRC_UNREACH_ADMIN_PROHIB ||
cmd == PRC_UNREACH_PORT || cmd == PRC_UNREACH_PROTOCOL ||
cmd == PRC_TIMXCEED_INTRANS) && ip) {
notify = tcp_drop_syn_sent;
}
else if (cmd == PRC_HOSTDEAD) {
ip = NULL;
} else if (inetctlerrmap[cmd] == 0 && !PRC_IS_REDIRECT(cmd)) {
return;
}
if (ip == NULL) {
in_pcbnotifyall(&tcbinfo, faddr, inetctlerrmap[cmd], notify);
return;
}
icp = (struct icmp *)(void *)
((caddr_t)ip - offsetof(struct icmp, icmp_ip));
th = (struct tcphdr *)(void *)((caddr_t)ip + (IP_VHL_HL(ip->ip_vhl) << 2));
icmp_tcp_seq = ntohl(th->th_seq);
inp = in_pcblookup_hash(&tcbinfo, faddr, th->th_dport,
ip->ip_src, th->th_sport, 0, NULL);
if (inp == NULL ||
inp->inp_socket == NULL) {
return;
}
socket_lock(inp->inp_socket, 1);
if (in_pcb_checkstate(inp, WNT_RELEASE, 1) ==
WNT_STOPUSING) {
socket_unlock(inp->inp_socket, 1);
return;
}
if (PRC_IS_REDIRECT(cmd)) {
(*notify)(inp, EHOSTDOWN);
} else {
tp = intotcpcb(inp);
if (SEQ_GEQ(icmp_tcp_seq, tp->snd_una) &&
SEQ_LT(icmp_tcp_seq, tp->snd_max)) {
if (cmd == PRC_MSGSIZE) {
tcp_handle_msgsize(ip, inp);
}
(*notify)(inp, inetctlerrmap[cmd]);
}
}
socket_unlock(inp->inp_socket, 1);
}
void
tcp6_ctlinput(int cmd, struct sockaddr *sa, void *d, __unused struct ifnet *ifp)
{
tcp_seq icmp_tcp_seq;
struct in6_addr *dst;
void (*notify)(struct inpcb *, int) = tcp_notify;
struct ip6_hdr *ip6;
struct mbuf *m;
struct inpcb *inp;
struct tcpcb *tp;
struct icmp6_hdr *icmp6;
struct ip6ctlparam *ip6cp = NULL;
const struct sockaddr_in6 *sa6_src = NULL;
unsigned int mtu;
unsigned int off;
struct tcp_ports {
uint16_t th_sport;
uint16_t th_dport;
} t_ports;
if (sa->sa_family != AF_INET6 ||
sa->sa_len != sizeof(struct sockaddr_in6)) {
return;
}
if (cmd == PRC_QUENCH) {
return;
}
if ((unsigned)cmd >= PRC_NCMDS) {
return;
}
if (d != NULL) {
ip6cp = (struct ip6ctlparam *)d;
icmp6 = ip6cp->ip6c_icmp6;
m = ip6cp->ip6c_m;
ip6 = ip6cp->ip6c_ip6;
off = ip6cp->ip6c_off;
sa6_src = ip6cp->ip6c_src;
dst = ip6cp->ip6c_finaldst;
} else {
m = NULL;
ip6 = NULL;
off = 0;
sa6_src = &sa6_any;
dst = NULL;
}
if (cmd == PRC_MSGSIZE) {
notify = tcp_mtudisc;
} else if (icmp_may_rst && (cmd == PRC_UNREACH_ADMIN_PROHIB ||
cmd == PRC_UNREACH_PORT || cmd == PRC_TIMXCEED_INTRANS) &&
ip6 != NULL) {
notify = tcp_drop_syn_sent;
}
else if (cmd == PRC_HOSTDEAD) {
ip6 = NULL;
} else if (inet6ctlerrmap[cmd] == 0 && !PRC_IS_REDIRECT(cmd)) {
return;
}
if (ip6 == NULL) {
in6_pcbnotify(&tcbinfo, sa, 0, (struct sockaddr *)(size_t)sa6_src,
0, cmd, NULL, notify);
return;
}
if (m == NULL ||
(m->m_pkthdr.len <
(int32_t) (off + sizeof(struct tcp_ports)))) {
return;
}
bzero(&t_ports, sizeof(struct tcp_ports));
m_copydata(m, off, sizeof(struct tcp_ports), (caddr_t)&t_ports);
off += sizeof(struct tcp_ports);
if (m->m_pkthdr.len < (int32_t) (off + sizeof(tcp_seq))) {
return;
}
m_copydata(m, off, sizeof(tcp_seq), (caddr_t)&icmp_tcp_seq);
icmp_tcp_seq = ntohl(icmp_tcp_seq);
if (cmd == PRC_MSGSIZE) {
mtu = ntohl(icmp6->icmp6_mtu);
if (mtu < IPV6_MMTU) {
mtu = IPV6_MMTU - 8;
}
}
inp = in6_pcblookup_hash(&tcbinfo, &ip6->ip6_dst, t_ports.th_dport,
&ip6->ip6_src, t_ports.th_sport, 0, NULL);
if (inp == NULL ||
inp->inp_socket == NULL) {
return;
}
socket_lock(inp->inp_socket, 1);
if (in_pcb_checkstate(inp, WNT_RELEASE, 1) ==
WNT_STOPUSING) {
socket_unlock(inp->inp_socket, 1);
return;
}
if (PRC_IS_REDIRECT(cmd)) {
(*notify)(inp, EHOSTDOWN);
} else {
tp = intotcpcb(inp);
if (SEQ_GEQ(icmp_tcp_seq, tp->snd_una) &&
SEQ_LT(icmp_tcp_seq, tp->snd_max)) {
if (cmd == PRC_MSGSIZE) {
if (mtu < tp->t_maxseg +
(sizeof(struct tcphdr) + sizeof(struct ip6_hdr))) {
(*notify)(inp, inetctlerrmap[cmd]);
}
} else {
(*notify)(inp, inetctlerrmap[cmd]);
}
}
}
socket_unlock(inp->inp_socket, 1);
}
#define ISN_BYTES_PER_SECOND 1048576
tcp_seq
tcp_new_isn(struct tcpcb *tp)
{
u_int32_t md5_buffer[4];
tcp_seq new_isn;
struct timeval timenow;
u_char isn_secret[32];
long 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_frandom(&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 ((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 {
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(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 socket *so = inp->inp_socket;
int mss;
u_int32_t mtu;
u_int32_t protoHdrOverhead = sizeof(struct tcpiphdr);
int isipv6 = (tp->t_inpcb->inp_vflag & INP_IPV6) != 0;
if ((so->so_state & SS_DEFUNCT) || (tp != NULL && tp->t_state == TCPS_CLOSED)) {
return;
}
if (isipv6) {
protoHdrOverhead = sizeof(struct ip6_hdr) +
sizeof(struct tcphdr);
}
if (tp != NULL) {
if (isipv6) {
rt = tcp_rtlookup6(inp, IFSCOPE_NONE);
} else {
rt = tcp_rtlookup(inp, IFSCOPE_NONE);
}
if (!rt || !rt->rt_rmx.rmx_mtu) {
tp->t_maxopd = tp->t_maxseg =
isipv6 ? tcp_v6mssdflt :
tcp_mssdflt;
if (rt != NULL) {
RT_UNLOCK(rt);
}
return;
}
mtu = rt->rt_rmx.rmx_mtu;
RT_UNLOCK(rt);
#if NECP
mtu = necp_socket_get_effective_mtu(inp, mtu);
#endif
mss = mtu - protoHdrOverhead;
if (tp->t_maxopd) {
mss = min(mss, tp->t_maxopd);
}
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;
ASSERT(tp->t_maxseg);
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(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 (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 (inp->inp_last_outifp == NULL) {
inp->inp_last_outifp = 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;
}
struct rtentry *
tcp_rtlookup6(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 (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 (inp->inp_last_outifp == NULL) {
inp->inp_last_outifp = rt->rt_ifp;
}
if (!(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;
}
#if IPSEC
size_t
ipsec_hdrsiz_tcp(struct tcpcb *tp)
{
struct inpcb *inp;
struct mbuf *m;
size_t hdrsiz;
struct ip *ip;
struct ip6_hdr *ip6 = NULL;
struct tcphdr *th;
if ((tp == NULL) || ((inp = tp->t_inpcb) == NULL)) {
return 0;
}
MGETHDR(m, M_DONTWAIT, MT_DATA);
if (!m) {
return 0;
}
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 {
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
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;
}
retry:
if (so->so_pcb != NULL) {
if (so->so_flags & SOF_MP_SUBFLOW) {
struct mptcb *mp_tp = tptomptp(sototcpcb(so));
struct socket *mp_so = mptetoso(mp_tp->mpt_mpte);
socket_lock(mp_so, refcount);
if (!(so->so_flags & SOF_MP_SUBFLOW)) {
socket_unlock(mp_so, refcount);
goto retry;
}
} else {
lck_mtx_lock(&((struct inpcb *)so->so_pcb)->inpcb_mtx);
if (so->so_flags & SOF_MP_SUBFLOW) {
lck_mtx_unlock(&((struct inpcb *)so->so_pcb)->inpcb_mtx);
goto retry;
}
}
} 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 {
so->unlock_lr[so->next_unlock_lr] = lr_saved;
so->next_unlock_lr = (so->next_unlock_lr + 1) % SO_LCKDBG_MAX;
if (so->so_flags & SOF_MP_SUBFLOW) {
struct mptcb *mp_tp = tptomptp(sototcpcb(so));
struct socket *mp_so = mptetoso(mp_tp->mpt_mpte);
socket_lock_assert_owned(mp_so);
socket_unlock(mp_so, refcount);
} else {
LCK_MTX_ASSERT(&((struct inpcb *)so->so_pcb)->inpcb_mtx,
LCK_MTX_ASSERT_OWNED);
lck_mtx_unlock(&((struct inpcb *)so->so_pcb)->inpcb_mtx);
}
}
return 0;
}
lck_mtx_t *
tcp_getlock(struct socket *so, int flags)
{
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));
}
if (so->so_flags & SOF_MP_SUBFLOW) {
struct mptcb *mp_tp = tptomptp(sototcpcb(so));
struct socket *mp_so = mptetoso(mp_tp->mpt_mpte);
return mp_so->so_proto->pr_getlock(mp_so, flags);
} else {
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 (tcp_recv_bg == 1 || IS_TCP_RECV_BG(so)) {
return;
}
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 socket *so = tp->t_inpcb->inp_socket;
struct sockbuf *sb = &so->so_rcv;
u_int32_t rcvbuf;
int32_t space;
int32_t pending = 0;
if (so->so_flags & SOF_MP_SUBFLOW) {
tcp_sbrcv_grow_rwin(tp, sb);
return mptcp_sbspace(tptomptp(tp));
}
tcp_sbrcv_grow_rwin(tp, sb);
rcvbuf = sb->sb_hiwat;
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)
{
struct inpcb *inp;
int isipv6;
struct ifnet *tunnel_ifp = NULL;
#define IFNET_TSO_MASK (IFNET_TSO_IPV6 | IFNET_TSO_IPV4)
tp->t_flags &= ~TF_TSO;
if (ifp == NULL) {
return;
}
#if MPTCP
if (tp->t_mpflags & TMPF_MPTCP_TRUE) {
return;
}
#endif
inp = tp->t_inpcb;
isipv6 = (inp->inp_vflag & INP_IPV6) != 0;
if (inp->inp_policyresult.results.result == NECP_KERNEL_POLICY_RESULT_IP_TUNNEL) {
u_int tunnel_if_index = inp->inp_policyresult.results.result_parameter.tunnel_interface_index;
if (tunnel_if_index != 0) {
ifnet_head_lock_shared();
tunnel_ifp = ifindex2ifnet[tunnel_if_index];
ifnet_head_done();
}
if (tunnel_ifp == NULL) {
return;
}
if ((ifp->if_hwassist & IFNET_TSO_MASK) != (tunnel_ifp->if_hwassist & IFNET_TSO_MASK)) {
if (tso_debug > 0) {
os_log(OS_LOG_DEFAULT,
"%s: %u > %u TSO 0 tunnel_ifp %s hwassist mismatch with ifp %s",
__func__,
ntohs(tp->t_inpcb->inp_lport), ntohs(tp->t_inpcb->inp_fport),
tunnel_ifp->if_xname, ifp->if_xname);
}
return;
}
if (inp->inp_last_outifp != NULL &&
(inp->inp_last_outifp->if_hwassist & IFNET_TSO_MASK) != (tunnel_ifp->if_hwassist & IFNET_TSO_MASK)) {
if (tso_debug > 0) {
os_log(OS_LOG_DEFAULT,
"%s: %u > %u TSO 0 tunnel_ifp %s hwassist mismatch with inp_last_outifp %s",
__func__,
ntohs(tp->t_inpcb->inp_lport), ntohs(tp->t_inpcb->inp_fport),
tunnel_ifp->if_xname, inp->inp_last_outifp->if_xname);
}
return;
}
if ((inp->inp_flags & INP_BOUND_IF) && inp->inp_boundifp != NULL &&
(inp->inp_boundifp->if_hwassist & IFNET_TSO_MASK) != (tunnel_ifp->if_hwassist & IFNET_TSO_MASK)) {
if (tso_debug > 0) {
os_log(OS_LOG_DEFAULT,
"%s: %u > %u TSO 0 tunnel_ifp %s hwassist mismatch with inp_boundifp %s",
__func__,
ntohs(tp->t_inpcb->inp_lport), ntohs(tp->t_inpcb->inp_fport),
tunnel_ifp->if_xname, inp->inp_boundifp->if_xname);
}
return;
}
}
if (isipv6) {
if (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 {
if (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;
}
if (INTF_ADJUST_MTU_FOR_CLAT46(ifp)) {
tp->tso_max_segment_size -=
CLAT46_HDR_EXPANSION_OVERHD;
}
}
}
if (tso_debug > 1) {
os_log(OS_LOG_DEFAULT, "%s: %u > %u TSO %d ifp %s",
__func__,
ntohs(tp->t_inpcb->inp_lport),
ntohs(tp->t_inpcb->inp_fport),
(tp->t_flags & TF_TSO) != 0,
ifp != NULL ? ifp->if_xname : "<NULL>");
}
}
#define TIMEVAL_TO_TCPHZ(_tv_) ((uint32_t)((_tv_).tv_sec * TCP_RETRANSHZ + \
(_tv_).tv_usec / TCP_RETRANSHZ_TO_USEC))
void
calculate_tcp_clock(void)
{
struct timeval tv = tcp_uptime;
struct timeval interval = {.tv_sec = 0, .tv_usec = TCP_RETRANSHZ_TO_USEC};
struct timeval now, hold_now;
uint32_t incr = 0;
microuptime(&now);
net_update_uptime_with_time(&now);
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);
}
}
void
tcp_set_max_rwinscale(struct tcpcb *tp, struct socket *so)
{
uint32_t maxsockbufsize;
tp->request_r_scale = MAX((uint8_t)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 + 1) << 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);
}
if (tp == NULL) {
return 0;
}
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;
uint16_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);
}
}
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_set_spurious(struct tcpcb *tp, tcp_seq start, tcp_seq end)
{
struct tcp_rxt_seg *rxseg;
if (SLIST_EMPTY(&tp->t_rxt_segments)) {
return;
}
SLIST_FOREACH(rxseg, &tp->t_rxt_segments, rx_link) {
if (SEQ_GEQ(rxseg->rx_start, start) &&
SEQ_LEQ(rxseg->rx_end, end)) {
if (rxseg->rx_count == 1) {
rxseg->rx_flags |= TCP_RXT_SPURIOUS;
}
}
if (SEQ_GEQ(rxseg->rx_start, end)) {
break;
}
}
return;
}
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_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;
}
}
boolean_t
tfo_enabled(const struct tcpcb *tp)
{
return (tp->t_flagsext & TF_FASTOPEN)? TRUE : FALSE;
}
void
tcp_disable_tfo(struct tcpcb *tp)
{
tp->t_flagsext &= ~TF_FASTOPEN;
}
static struct mbuf *
tcp_make_keepalive_frame(struct tcpcb *tp, struct ifnet *ifp,
boolean_t is_probe)
{
struct inpcb *inp = tp->t_inpcb;
struct tcphdr *th;
u_int8_t *data;
int win = 0;
struct mbuf *m;
_CASSERT(sizeof(struct ip) + sizeof(struct tcphdr) <= _MHLEN);
_CASSERT(sizeof(struct ip6_hdr) + sizeof(struct tcphdr) <= _MHLEN);
MGETHDR(m, M_WAIT, MT_HEADER);
if (m == NULL) {
return NULL;
}
m->m_pkthdr.pkt_proto = IPPROTO_TCP;
data = mbuf_datastart(m);
if (inp->inp_vflag & INP_IPV4) {
bzero(data, sizeof(struct ip) + sizeof(struct tcphdr));
th = (struct tcphdr *)(void *) (data + sizeof(struct ip));
m->m_len = sizeof(struct ip) + sizeof(struct tcphdr);
m->m_pkthdr.len = m->m_len;
} else {
VERIFY(inp->inp_vflag & INP_IPV6);
bzero(data, sizeof(struct ip6_hdr)
+ sizeof(struct tcphdr));
th = (struct tcphdr *)(void *)(data + sizeof(struct ip6_hdr));
m->m_len = sizeof(struct ip6_hdr) +
sizeof(struct tcphdr);
m->m_pkthdr.len = m->m_len;
}
tcp_fillheaders(tp, data, th);
if (inp->inp_vflag & INP_IPV4) {
struct ip *ip;
ip = (__typeof__(ip))(void *)data;
ip->ip_id = rfc6864 ? 0 : ip_randomid();
ip->ip_off = htons(IP_DF);
ip->ip_len = htons(sizeof(struct ip) + sizeof(struct tcphdr));
ip->ip_ttl = inp->inp_ip_ttl;
ip->ip_tos |= (inp->inp_ip_tos & ~IPTOS_ECN_MASK);
ip->ip_sum = in_cksum_hdr(ip);
} else {
struct ip6_hdr *ip6;
ip6 = (__typeof__(ip6))(void *)data;
ip6->ip6_plen = htons(sizeof(struct tcphdr));
ip6->ip6_hlim = in6_selecthlim(inp, ifp);
ip6->ip6_flow = ip6->ip6_flow & ~IPV6_FLOW_ECN_MASK;
if (IN6_IS_SCOPE_EMBED(&ip6->ip6_src)) {
ip6->ip6_src.s6_addr16[1] = 0;
}
if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst)) {
ip6->ip6_dst.s6_addr16[1] = 0;
}
}
th->th_flags = TH_ACK;
win = tcp_sbspace(tp);
if (win > ((int32_t)TCP_MAXWIN << tp->rcv_scale)) {
win = (int32_t)TCP_MAXWIN << tp->rcv_scale;
}
th->th_win = htons((u_short) (win >> tp->rcv_scale));
if (is_probe) {
th->th_seq = htonl(tp->snd_una - 1);
} else {
th->th_seq = htonl(tp->snd_una);
}
th->th_ack = htonl(tp->rcv_nxt);
th->th_sum = 0;
if (inp->inp_vflag & INP_IPV4) {
th->th_sum = inet_cksum(m, IPPROTO_TCP,
sizeof(struct ip), sizeof(struct tcphdr));
} else {
th->th_sum = inet6_cksum(m, IPPROTO_TCP,
sizeof(struct ip6_hdr), sizeof(struct tcphdr));
}
return m;
}
void
tcp_fill_keepalive_offload_frames(ifnet_t ifp,
struct ifnet_keepalive_offload_frame *frames_array,
u_int32_t frames_array_count, size_t frame_data_offset,
u_int32_t *used_frames_count)
{
struct inpcb *inp;
inp_gen_t gencnt;
u_int32_t frame_index = *used_frames_count;
if (ifp == NULL || frames_array == NULL ||
frames_array_count == 0 ||
frame_index >= frames_array_count ||
frame_data_offset >= IFNET_KEEPALIVE_OFFLOAD_FRAME_DATA_SIZE) {
return;
}
calculate_tcp_clock();
lck_rw_lock_shared(tcbinfo.ipi_lock);
gencnt = tcbinfo.ipi_gencnt;
LIST_FOREACH(inp, tcbinfo.ipi_listhead, inp_list) {
struct socket *so;
struct ifnet_keepalive_offload_frame *frame;
struct mbuf *m = NULL;
struct tcpcb *tp = intotcpcb(inp);
if (frame_index >= frames_array_count) {
break;
}
if (inp->inp_gencnt > gencnt ||
inp->inp_state == INPCB_STATE_DEAD) {
continue;
}
if ((so = inp->inp_socket) == NULL ||
(so->so_state & SS_DEFUNCT)) {
continue;
}
if (!(inp->inp_flags2 & INP2_KEEPALIVE_OFFLOAD)) {
continue;
}
if (!(inp->inp_vflag & (INP_IPV4 | INP_IPV6))) {
continue;
}
if (inp->inp_ppcb == NULL ||
in_pcb_checkstate(inp, WNT_ACQUIRE, 0) == WNT_STOPUSING) {
continue;
}
socket_lock(so, 1);
if (inp->inp_ppcb == NULL ||
(in_pcb_checkstate(inp, WNT_RELEASE, 1) == WNT_STOPUSING)) {
socket_unlock(so, 1);
continue;
}
if ((inp->inp_vflag & INP_IPV4) &&
(inp->inp_laddr.s_addr == INADDR_ANY ||
inp->inp_faddr.s_addr == INADDR_ANY)) {
socket_unlock(so, 1);
continue;
}
if ((inp->inp_vflag & INP_IPV6) &&
(IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) ||
IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr))) {
socket_unlock(so, 1);
continue;
}
if (inp->inp_lport == 0 || inp->inp_fport == 0) {
socket_unlock(so, 1);
continue;
}
if (inp->inp_last_outifp == NULL ||
inp->inp_last_outifp->if_index != ifp->if_index) {
socket_unlock(so, 1);
continue;
}
if ((inp->inp_vflag & INP_IPV4) && frame_data_offset +
sizeof(struct ip) + sizeof(struct tcphdr) >
IFNET_KEEPALIVE_OFFLOAD_FRAME_DATA_SIZE) {
socket_unlock(so, 1);
continue;
} else if (!(inp->inp_vflag & INP_IPV4) && frame_data_offset +
sizeof(struct ip6_hdr) + sizeof(struct tcphdr) >
IFNET_KEEPALIVE_OFFLOAD_FRAME_DATA_SIZE) {
socket_unlock(so, 1);
continue;
}
if (tp->t_state != TCPS_ESTABLISHED) {
socket_unlock(so, 1);
continue;
}
frame = &frames_array[frame_index];
frame->type = IFNET_KEEPALIVE_OFFLOAD_FRAME_TCP;
frame->ether_type = (inp->inp_vflag & INP_IPV4) ?
IFNET_KEEPALIVE_OFFLOAD_FRAME_ETHERTYPE_IPV4 :
IFNET_KEEPALIVE_OFFLOAD_FRAME_ETHERTYPE_IPV6;
frame->interval = (uint16_t)(tp->t_keepidle > 0 ? tp->t_keepidle :
tcp_keepidle);
frame->keep_cnt = (uint8_t)TCP_CONN_KEEPCNT(tp);
frame->keep_retry = (uint16_t)TCP_CONN_KEEPINTVL(tp);
if (so->so_options & SO_NOWAKEFROMSLEEP) {
frame->flags |=
IFNET_KEEPALIVE_OFFLOAD_FLAG_NOWAKEFROMSLEEP;
}
frame->local_port = ntohs(inp->inp_lport);
frame->remote_port = ntohs(inp->inp_fport);
frame->local_seq = tp->snd_nxt;
frame->remote_seq = tp->rcv_nxt;
if (inp->inp_vflag & INP_IPV4) {
ASSERT(frame_data_offset + sizeof(struct ip) + sizeof(struct tcphdr) <= UINT8_MAX);
frame->length = (uint8_t)(frame_data_offset +
sizeof(struct ip) + sizeof(struct tcphdr));
frame->reply_length = frame->length;
frame->addr_length = sizeof(struct in_addr);
bcopy(&inp->inp_laddr, frame->local_addr,
sizeof(struct in_addr));
bcopy(&inp->inp_faddr, frame->remote_addr,
sizeof(struct in_addr));
} else {
struct in6_addr *ip6;
ASSERT(frame_data_offset + sizeof(struct ip6_hdr) + sizeof(struct tcphdr) <= UINT8_MAX);
frame->length = (uint8_t)(frame_data_offset +
sizeof(struct ip6_hdr) + sizeof(struct tcphdr));
frame->reply_length = frame->length;
frame->addr_length = sizeof(struct in6_addr);
ip6 = (struct in6_addr *)(void *)frame->local_addr;
bcopy(&inp->in6p_laddr, ip6, sizeof(struct in6_addr));
if (IN6_IS_SCOPE_EMBED(ip6)) {
ip6->s6_addr16[1] = 0;
}
ip6 = (struct in6_addr *)(void *)frame->remote_addr;
bcopy(&inp->in6p_faddr, ip6, sizeof(struct in6_addr));
if (IN6_IS_SCOPE_EMBED(ip6)) {
ip6->s6_addr16[1] = 0;
}
}
m = tcp_make_keepalive_frame(tp, ifp, TRUE);
if (m == NULL) {
socket_unlock(so, 1);
continue;
}
bcopy(m->m_data, frame->data + frame_data_offset,
m->m_len);
m_freem(m);
m = tcp_make_keepalive_frame(tp, ifp, FALSE);
if (m == NULL) {
socket_unlock(so, 1);
continue;
}
bcopy(m->m_data, frame->reply_data + frame_data_offset,
m->m_len);
m_freem(m);
frame_index++;
socket_unlock(so, 1);
}
lck_rw_done(tcbinfo.ipi_lock);
*used_frames_count = frame_index;
}
static bool
inp_matches_kao_frame(ifnet_t ifp, struct ifnet_keepalive_offload_frame *frame,
struct inpcb *inp)
{
if (inp->inp_ppcb == NULL) {
return false;
}
if (in_pcb_checkstate(inp, WNT_RELEASE, 1) == WNT_STOPUSING) {
return false;
}
if (inp->inp_last_outifp == NULL ||
inp->inp_last_outifp->if_index != ifp->if_index) {
return false;
}
if (frame->local_port != ntohs(inp->inp_lport) ||
frame->remote_port != ntohs(inp->inp_fport)) {
return false;
}
if (inp->inp_vflag & INP_IPV4) {
if (memcmp(&inp->inp_laddr, frame->local_addr,
sizeof(struct in_addr)) != 0 ||
memcmp(&inp->inp_faddr, frame->remote_addr,
sizeof(struct in_addr)) != 0) {
return false;
}
} else if (inp->inp_vflag & INP_IPV6) {
if (memcmp(&inp->inp_laddr, frame->local_addr,
sizeof(struct in6_addr)) != 0 ||
memcmp(&inp->inp_faddr, frame->remote_addr,
sizeof(struct in6_addr)) != 0) {
return false;
}
} else {
return false;
}
return true;
}
int
tcp_notify_kao_timeout(ifnet_t ifp,
struct ifnet_keepalive_offload_frame *frame)
{
struct inpcb *inp = NULL;
struct socket *so = NULL;
bool found = false;
lck_rw_lock_shared(tcbinfo.ipi_lock);
LIST_FOREACH(inp, tcbinfo.ipi_listhead, inp_list) {
if ((so = inp->inp_socket) == NULL ||
(so->so_state & SS_DEFUNCT)) {
continue;
}
if (!(inp->inp_flags2 & INP2_KEEPALIVE_OFFLOAD)) {
continue;
}
if (!(inp->inp_vflag & (INP_IPV4 | INP_IPV6))) {
continue;
}
if (inp->inp_ppcb == NULL ||
in_pcb_checkstate(inp, WNT_ACQUIRE, 0) == WNT_STOPUSING) {
continue;
}
socket_lock(so, 1);
if (inp_matches_kao_frame(ifp, frame, inp)) {
found = true;
break;
}
socket_unlock(so, 1);
}
lck_rw_done(tcbinfo.ipi_lock);
if (found) {
ASSERT(inp != NULL);
ASSERT(so != NULL);
ASSERT(so == inp->inp_socket);
struct tcpcb *tp = inp->inp_ppcb;
tcpstat.tcps_keepdrops++;
soevent(so,
(SO_FILT_HINT_LOCKED | SO_FILT_HINT_TIMEOUT));
tp = tcp_drop(tp, ETIMEDOUT);
tcpstat.tcps_ka_offload_drops++;
os_log_info(OS_LOG_DEFAULT, "%s: dropped lport %u fport %u\n",
__func__, frame->local_port, frame->remote_port);
socket_unlock(so, 1);
}
return 0;
}
errno_t
tcp_notify_ack_id_valid(struct tcpcb *tp, struct socket *so,
u_int32_t notify_id)
{
struct tcp_notify_ack_marker *elm;
if (so->so_snd.sb_cc == 0) {
return ENOBUFS;
}
SLIST_FOREACH(elm, &tp->t_notify_ack, notify_next) {
if (elm->notify_id == notify_id) {
return EINVAL;
}
if (elm->notify_snd_una == tp->snd_una + so->so_snd.sb_cc) {
return EINVAL;
}
}
return 0;
}
errno_t
tcp_add_notify_ack_marker(struct tcpcb *tp, u_int32_t notify_id)
{
struct tcp_notify_ack_marker *nm, *elm = NULL;
struct socket *so = tp->t_inpcb->inp_socket;
MALLOC(nm, struct tcp_notify_ack_marker *, sizeof(*nm),
M_TEMP, M_WAIT | M_ZERO);
if (nm == NULL) {
return ENOMEM;
}
nm->notify_id = notify_id;
nm->notify_snd_una = tp->snd_una + so->so_snd.sb_cc;
SLIST_FOREACH(elm, &tp->t_notify_ack, notify_next) {
if (SEQ_GT(nm->notify_snd_una, elm->notify_snd_una)) {
break;
}
}
if (elm == NULL) {
VERIFY(SLIST_EMPTY(&tp->t_notify_ack));
SLIST_INSERT_HEAD(&tp->t_notify_ack, nm, notify_next);
} else {
SLIST_INSERT_AFTER(elm, nm, notify_next);
}
tp->t_notify_ack_count++;
return 0;
}
void
tcp_notify_ack_free(struct tcpcb *tp)
{
struct tcp_notify_ack_marker *elm, *next;
if (SLIST_EMPTY(&tp->t_notify_ack)) {
return;
}
SLIST_FOREACH_SAFE(elm, &tp->t_notify_ack, notify_next, next) {
SLIST_REMOVE(&tp->t_notify_ack, elm, tcp_notify_ack_marker,
notify_next);
FREE(elm, M_TEMP);
}
SLIST_INIT(&tp->t_notify_ack);
tp->t_notify_ack_count = 0;
}
inline void
tcp_notify_acknowledgement(struct tcpcb *tp, struct socket *so)
{
struct tcp_notify_ack_marker *elm;
elm = SLIST_FIRST(&tp->t_notify_ack);
if (SEQ_GEQ(tp->snd_una, elm->notify_snd_una)) {
soevent(so, SO_FILT_HINT_LOCKED | SO_FILT_HINT_NOTIFY_ACK);
}
}
void
tcp_get_notify_ack_count(struct tcpcb *tp,
struct tcp_notify_ack_complete *retid)
{
struct tcp_notify_ack_marker *elm;
uint32_t complete = 0;
SLIST_FOREACH(elm, &tp->t_notify_ack, notify_next) {
if (SEQ_GEQ(tp->snd_una, elm->notify_snd_una)) {
ASSERT(complete < UINT32_MAX);
complete++;
} else {
break;
}
}
retid->notify_pending = tp->t_notify_ack_count - complete;
retid->notify_complete_count = min(TCP_MAX_NOTIFY_ACK, complete);
}
void
tcp_get_notify_ack_ids(struct tcpcb *tp,
struct tcp_notify_ack_complete *retid)
{
size_t i = 0;
struct tcp_notify_ack_marker *elm, *next;
SLIST_FOREACH_SAFE(elm, &tp->t_notify_ack, notify_next, next) {
if (i >= retid->notify_complete_count) {
break;
}
if (SEQ_GEQ(tp->snd_una, elm->notify_snd_una)) {
retid->notify_complete_id[i++] = elm->notify_id;
SLIST_REMOVE(&tp->t_notify_ack, elm,
tcp_notify_ack_marker, notify_next);
FREE(elm, M_TEMP);
tp->t_notify_ack_count--;
} else {
break;
}
}
}
bool
tcp_notify_ack_active(struct socket *so)
{
if ((SOCK_DOM(so) == PF_INET || SOCK_DOM(so) == PF_INET6) &&
SOCK_TYPE(so) == SOCK_STREAM) {
struct tcpcb *tp = intotcpcb(sotoinpcb(so));
if (!SLIST_EMPTY(&tp->t_notify_ack)) {
struct tcp_notify_ack_marker *elm;
elm = SLIST_FIRST(&tp->t_notify_ack);
if (SEQ_GEQ(tp->snd_una, elm->notify_snd_una)) {
return true;
}
}
}
return false;
}
inline int32_t
inp_get_sndbytes_allunsent(struct socket *so, u_int32_t th_ack)
{
struct inpcb *inp = sotoinpcb(so);
struct tcpcb *tp = intotcpcb(inp);
if ((so->so_snd.sb_flags & SB_SNDBYTE_CNT) &&
so->so_snd.sb_cc > 0) {
int32_t unsent, sent;
sent = tp->snd_max - th_ack;
if (tp->t_flags & TF_SENTFIN) {
sent--;
}
unsent = so->so_snd.sb_cc - sent;
return unsent;
}
return 0;
}
#define IFP_PER_FLOW_STAT(_ipv4_, _stat_) { \
if (_ipv4_) { \
ifp->if_ipv4_stat->_stat_++; \
} else { \
ifp->if_ipv6_stat->_stat_++; \
} \
}
#define FLOW_ECN_ENABLED(_flags_) \
((_flags_ & (TE_ECN_ON)) == (TE_ECN_ON))
void
tcp_update_stats_per_flow(struct ifnet_stats_per_flow *ifs,
struct ifnet *ifp)
{
if (ifp == NULL || !IF_FULLY_ATTACHED(ifp)) {
return;
}
ifnet_lock_shared(ifp);
if (ifs->ecn_flags & TE_SETUPSENT) {
if (ifs->ecn_flags & TE_CLIENT_SETUP) {
IFP_PER_FLOW_STAT(ifs->ipv4, ecn_client_setup);
if (FLOW_ECN_ENABLED(ifs->ecn_flags)) {
IFP_PER_FLOW_STAT(ifs->ipv4,
ecn_client_success);
} else if (ifs->ecn_flags & TE_LOST_SYN) {
IFP_PER_FLOW_STAT(ifs->ipv4,
ecn_syn_lost);
} else {
IFP_PER_FLOW_STAT(ifs->ipv4,
ecn_peer_nosupport);
}
} else {
IFP_PER_FLOW_STAT(ifs->ipv4, ecn_server_setup);
if (FLOW_ECN_ENABLED(ifs->ecn_flags)) {
IFP_PER_FLOW_STAT(ifs->ipv4,
ecn_server_success);
} else if (ifs->ecn_flags & TE_LOST_SYN) {
IFP_PER_FLOW_STAT(ifs->ipv4,
ecn_synack_lost);
} else {
IFP_PER_FLOW_STAT(ifs->ipv4,
ecn_peer_nosupport);
}
}
} else {
IFP_PER_FLOW_STAT(ifs->ipv4, ecn_off_conn);
}
if (FLOW_ECN_ENABLED(ifs->ecn_flags)) {
if (ifs->ecn_flags & TE_RECV_ECN_CE) {
tcpstat.tcps_ecn_conn_recv_ce++;
IFP_PER_FLOW_STAT(ifs->ipv4, ecn_conn_recv_ce);
}
if (ifs->ecn_flags & TE_RECV_ECN_ECE) {
tcpstat.tcps_ecn_conn_recv_ece++;
IFP_PER_FLOW_STAT(ifs->ipv4, ecn_conn_recv_ece);
}
if (ifs->ecn_flags & (TE_RECV_ECN_CE | TE_RECV_ECN_ECE)) {
if (ifs->txretransmitbytes > 0 ||
ifs->rxoutoforderbytes > 0) {
tcpstat.tcps_ecn_conn_pl_ce++;
IFP_PER_FLOW_STAT(ifs->ipv4, ecn_conn_plce);
} else {
tcpstat.tcps_ecn_conn_nopl_ce++;
IFP_PER_FLOW_STAT(ifs->ipv4, ecn_conn_noplce);
}
} else {
if (ifs->txretransmitbytes > 0 ||
ifs->rxoutoforderbytes > 0) {
tcpstat.tcps_ecn_conn_plnoce++;
IFP_PER_FLOW_STAT(ifs->ipv4, ecn_conn_plnoce);
}
}
}
if (ifs->local) {
ifnet_lock_done(ifp);
return;
}
if (ifs->ipv4) {
ifp->if_ipv4_stat->timestamp = net_uptime();
if (FLOW_ECN_ENABLED(ifs->ecn_flags)) {
tcp_flow_ecn_perf_stats(ifs, &ifp->if_ipv4_stat->ecn_on);
} else {
tcp_flow_ecn_perf_stats(ifs, &ifp->if_ipv4_stat->ecn_off);
}
} else {
ifp->if_ipv6_stat->timestamp = net_uptime();
if (FLOW_ECN_ENABLED(ifs->ecn_flags)) {
tcp_flow_ecn_perf_stats(ifs, &ifp->if_ipv6_stat->ecn_on);
} else {
tcp_flow_ecn_perf_stats(ifs, &ifp->if_ipv6_stat->ecn_off);
}
}
if (ifs->rxmit_drop) {
if (FLOW_ECN_ENABLED(ifs->ecn_flags)) {
IFP_PER_FLOW_STAT(ifs->ipv4, ecn_on.rxmit_drop);
} else {
IFP_PER_FLOW_STAT(ifs->ipv4, ecn_off.rxmit_drop);
}
}
if (ifs->ecn_fallback_synloss) {
IFP_PER_FLOW_STAT(ifs->ipv4, ecn_fallback_synloss);
}
if (ifs->ecn_fallback_droprst) {
IFP_PER_FLOW_STAT(ifs->ipv4, ecn_fallback_droprst);
}
if (ifs->ecn_fallback_droprxmt) {
IFP_PER_FLOW_STAT(ifs->ipv4, ecn_fallback_droprxmt);
}
if (ifs->ecn_fallback_ce) {
IFP_PER_FLOW_STAT(ifs->ipv4, ecn_fallback_ce);
}
if (ifs->ecn_fallback_reorder) {
IFP_PER_FLOW_STAT(ifs->ipv4, ecn_fallback_reorder);
}
if (ifs->ecn_recv_ce > 0) {
IFP_PER_FLOW_STAT(ifs->ipv4, ecn_recv_ce);
}
if (ifs->ecn_recv_ece > 0) {
IFP_PER_FLOW_STAT(ifs->ipv4, ecn_recv_ece);
}
tcp_flow_lim_stats(ifs, &ifp->if_lim_stat);
ifnet_lock_done(ifp);
}