#define _IP_VHL
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/kernel.h>
#include <sys/sysctl.h>
#include <sys/mbuf.h>
#include <sys/domain.h>
#include <sys/protosw.h>
#include <sys/socket.h>
#include <sys/socketvar.h>
#include <net/route.h>
#include <net/ntstat.h>
#include <net/if_var.h>
#include <net/if.h>
#include <net/if_types.h>
#include <net/dlil.h>
#include <netinet/in.h>
#include <netinet/in_systm.h>
#include <netinet/in_var.h>
#include <netinet/in_tclass.h>
#include <netinet/ip.h>
#include <netinet/in_pcb.h>
#include <netinet/ip_var.h>
#include <mach/sdt.h>
#if INET6
#include <netinet6/in6_pcb.h>
#include <netinet/ip6.h>
#include <netinet6/ip6_var.h>
#endif
#include <netinet/tcp.h>
#define TCPOUTFLAGS
#include <netinet/tcp_cache.h>
#include <netinet/tcp_fsm.h>
#include <netinet/tcp_seq.h>
#include <netinet/tcp_timer.h>
#include <netinet/tcp_var.h>
#include <netinet/tcpip.h>
#include <netinet/tcp_cc.h>
#if TCPDEBUG
#include <netinet/tcp_debug.h>
#endif
#include <sys/kdebug.h>
#include <mach/sdt.h>
#if IPSEC
#include <netinet6/ipsec.h>
#endif
#if CONFIG_MACF_NET
#include <security/mac_framework.h>
#endif
#include <netinet/lro_ext.h>
#if MPTCP
#include <netinet/mptcp_var.h>
#include <netinet/mptcp.h>
#include <netinet/mptcp_opt.h>
#endif
#include <corecrypto/ccaes.h>
#define DBG_LAYER_BEG NETDBG_CODE(DBG_NETTCP, 1)
#define DBG_LAYER_END NETDBG_CODE(DBG_NETTCP, 3)
#define DBG_FNC_TCP_OUTPUT NETDBG_CODE(DBG_NETTCP, (4 << 8) | 1)
SYSCTL_SKMEM_TCP_INT(OID_AUTO, path_mtu_discovery,
CTLFLAG_RW | CTLFLAG_LOCKED, int, path_mtu_discovery, 1,
"Enable Path MTU Discovery");
SYSCTL_SKMEM_TCP_INT(OID_AUTO, slowstart_flightsize,
CTLFLAG_RW | CTLFLAG_LOCKED, int, ss_fltsz, 1,
"Slow start flight size");
SYSCTL_SKMEM_TCP_INT(OID_AUTO, local_slowstart_flightsize,
CTLFLAG_RW | CTLFLAG_LOCKED, int, ss_fltsz_local, 8,
"Slow start flight size for local networks");
int tcp_do_tso = 1;
SYSCTL_INT(_net_inet_tcp, OID_AUTO, tso, CTLFLAG_RW | CTLFLAG_LOCKED,
&tcp_do_tso, 0, "Enable TCP Segmentation Offload");
SYSCTL_SKMEM_TCP_INT(OID_AUTO, ecn_setup_percentage,
CTLFLAG_RW | CTLFLAG_LOCKED, int, tcp_ecn_setup_percentage, 100,
"Max ECN setup percentage");
static int
sysctl_change_ecn_setting SYSCTL_HANDLER_ARGS
{
#pragma unused(oidp, arg1, arg2)
int i, err = 0, changed = 0;
struct ifnet *ifp;
err = sysctl_io_number(req, tcp_ecn_outbound, sizeof(int32_t),
&i, &changed);
if (err != 0 || req->newptr == USER_ADDR_NULL)
return(err);
if (changed) {
if ((tcp_ecn_outbound == 0 || tcp_ecn_outbound == 1) &&
(i == 0 || i == 1)) {
tcp_ecn_outbound = i;
SYSCTL_SKMEM_UPDATE_FIELD(tcp.ecn_initiate_out, tcp_ecn_outbound);
return(err);
}
if (tcp_ecn_outbound == 2 && (i == 0 || i == 1)) {
ifnet_head_lock_shared();
TAILQ_FOREACH(ifp, &ifnet_head, if_link) {
if (!IFNET_IS_CELLULAR(ifp)) {
ifnet_lock_exclusive(ifp);
ifp->if_eflags &= ~IFEF_ECN_DISABLE;
ifp->if_eflags &= ~IFEF_ECN_ENABLE;
ifnet_lock_done(ifp);
}
}
ifnet_head_done();
} else {
ifnet_head_lock_shared();
TAILQ_FOREACH(ifp, &ifnet_head, if_link) {
if (!IFNET_IS_CELLULAR(ifp)) {
ifnet_lock_exclusive(ifp);
ifp->if_eflags |= IFEF_ECN_ENABLE;
ifp->if_eflags &= ~IFEF_ECN_DISABLE;
ifnet_lock_done(ifp);
}
}
ifnet_head_done();
}
tcp_ecn_outbound = i;
SYSCTL_SKMEM_UPDATE_FIELD(tcp.ecn_initiate_out, tcp_ecn_outbound);
}
if (i == 2 || tcp_ecn_inbound == 2) {
tcp_ecn_inbound = i;
SYSCTL_SKMEM_UPDATE_FIELD(tcp.ecn_negotiate_in, tcp_ecn_inbound);
}
return (err);
}
int tcp_ecn_outbound = 2;
SYSCTL_PROC(_net_inet_tcp, OID_AUTO, ecn_initiate_out,
CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_LOCKED, &tcp_ecn_outbound, 0,
sysctl_change_ecn_setting, "IU",
"Initiate ECN for outbound connections");
int tcp_ecn_inbound = 2;
SYSCTL_PROC(_net_inet_tcp, OID_AUTO, ecn_negotiate_in,
CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_LOCKED, &tcp_ecn_inbound, 0,
sysctl_change_ecn_setting, "IU",
"Initiate ECN for inbound connections");
SYSCTL_SKMEM_TCP_INT(OID_AUTO, packetchain,
CTLFLAG_RW | CTLFLAG_LOCKED, int, tcp_packet_chaining, 50,
"Enable TCP output packet chaining");
SYSCTL_SKMEM_TCP_INT(OID_AUTO, socket_unlocked_on_output,
CTLFLAG_RW | CTLFLAG_LOCKED, int, tcp_output_unlocked, 1,
"Unlock TCP when sending packets down to IP");
SYSCTL_SKMEM_TCP_INT(OID_AUTO, rfc3390,
CTLFLAG_RW | CTLFLAG_LOCKED, int, tcp_do_rfc3390, 1,
"Calculate intial slowstart cwnd depending on MSS");
SYSCTL_SKMEM_TCP_INT(OID_AUTO, min_iaj_win,
CTLFLAG_RW | CTLFLAG_LOCKED, int, tcp_min_iaj_win, MIN_IAJ_WIN,
"Minimum recv win based on inter-packet arrival jitter");
SYSCTL_SKMEM_TCP_INT(OID_AUTO, acc_iaj_react_limit,
CTLFLAG_RW | CTLFLAG_LOCKED, int, tcp_acc_iaj_react_limit,
ACC_IAJ_REACT_LIMIT, "Accumulated IAJ when receiver starts to react");
SYSCTL_SKMEM_TCP_INT(OID_AUTO, doautosndbuf,
CTLFLAG_RW | CTLFLAG_LOCKED, uint32_t, tcp_do_autosendbuf, 1,
"Enable send socket buffer auto-tuning");
SYSCTL_SKMEM_TCP_INT(OID_AUTO, autosndbufinc,
CTLFLAG_RW | CTLFLAG_LOCKED, uint32_t, tcp_autosndbuf_inc,
8 * 1024, "Increment in send socket bufffer size");
SYSCTL_SKMEM_TCP_INT(OID_AUTO, autosndbufmax,
CTLFLAG_RW | CTLFLAG_LOCKED, uint32_t, tcp_autosndbuf_max, 512 * 1024,
"Maximum send socket buffer size");
SYSCTL_SKMEM_TCP_INT(OID_AUTO, ack_prioritize,
CTLFLAG_RW | CTLFLAG_LOCKED, uint32_t, tcp_prioritize_acks, 1,
"Prioritize pure acks");
SYSCTL_SKMEM_TCP_INT(OID_AUTO, rtt_recvbg,
CTLFLAG_RW | CTLFLAG_LOCKED, uint32_t, tcp_use_rtt_recvbg, 1,
"Use RTT for bg recv algorithm");
SYSCTL_SKMEM_TCP_INT(OID_AUTO, recv_throttle_minwin,
CTLFLAG_RW | CTLFLAG_LOCKED, uint32_t, tcp_recv_throttle_minwin, 16 * 1024,
"Minimum recv win for throttling");
SYSCTL_SKMEM_TCP_INT(OID_AUTO, enable_tlp,
CTLFLAG_RW | CTLFLAG_LOCKED,
int32_t, tcp_enable_tlp, 1, "Enable Tail loss probe");
static int32_t packchain_newlist = 0;
static int32_t packchain_looped = 0;
static int32_t packchain_sent = 0;
#if IPSEC
extern int ipsec_bypass;
#endif
extern int slowlink_wsize;
#if IPFIREWALL
extern int fw_enable;
extern int fw_bypass;
#endif
extern u_int32_t dlil_filter_disable_tso_count;
extern u_int32_t kipf_count;
static int tcp_ip_output(struct socket *, struct tcpcb *, struct mbuf *,
int, struct mbuf *, int, int, boolean_t);
static struct mbuf* tcp_send_lroacks(struct tcpcb *tp, struct mbuf *m, struct tcphdr *th);
static int tcp_recv_throttle(struct tcpcb *tp);
static int32_t tcp_tfo_check(struct tcpcb *tp, int32_t len)
{
struct socket *so = tp->t_inpcb->inp_socket;
unsigned int optlen = 0;
unsigned int cookie_len;
if (tp->t_flags & TF_NOOPT)
goto fallback;
if ((so->so_flags1 & SOF1_DATA_AUTHENTICATED) &&
!(tp->t_flagsext & TF_FASTOPEN_HEUR))
return (len);
if (!tcp_heuristic_do_tfo(tp)) {
tp->t_tfo_stats |= TFO_S_HEURISTICS_DISABLE;
tcpstat.tcps_tfo_heuristics_disable++;
goto fallback;
}
if (so->so_flags1 & SOF1_DATA_AUTHENTICATED)
return (len);
optlen += TCPOLEN_MAXSEG;
if (tp->t_flags & TF_REQ_SCALE)
optlen += 4;
#if MPTCP
if ((so->so_flags & SOF_MP_SUBFLOW) && mptcp_enable &&
tp->t_rxtshift <= mptcp_mpcap_retries)
optlen += sizeof(struct mptcp_mpcapable_opt_common) + sizeof(mptcp_key_t);
#endif
if (tp->t_flags & TF_REQ_TSTMP)
optlen += TCPOLEN_TSTAMP_APPA;
if (SACK_ENABLED(tp))
optlen += TCPOLEN_SACK_PERMITTED;
if (MAX_TCPOPTLEN - optlen < TCPOLEN_FASTOPEN_REQ)
goto fallback;
cookie_len = tcp_cache_get_cookie_len(tp);
if (cookie_len == 0)
return (0);
if (MAX_TCPOPTLEN - optlen < cookie_len)
goto fallback;
if (so->so_snd.sb_cc > (tp->t_maxopd - MAX_TCPOPTLEN))
goto fallback;
return (len);
fallback:
tp->t_flagsext &= ~TF_FASTOPEN;
return (0);
}
static unsigned
tcp_tfo_write_cookie_rep(struct tcpcb *tp, unsigned optlen, u_char *opt)
{
u_char out[CCAES_BLOCK_SIZE];
unsigned ret = 0;
u_char *bp;
if ((MAX_TCPOPTLEN - optlen) <
(TCPOLEN_FASTOPEN_REQ + TFO_COOKIE_LEN_DEFAULT))
return (ret);
tcp_tfo_gen_cookie(tp->t_inpcb, out, sizeof(out));
bp = opt + optlen;
*bp++ = TCPOPT_FASTOPEN;
*bp++ = 2 + TFO_COOKIE_LEN_DEFAULT;
memcpy(bp, out, TFO_COOKIE_LEN_DEFAULT);
ret += 2 + TFO_COOKIE_LEN_DEFAULT;
tp->t_tfo_stats |= TFO_S_COOKIE_SENT;
tcpstat.tcps_tfo_cookie_sent++;
return (ret);
}
static unsigned
tcp_tfo_write_cookie(struct tcpcb *tp, unsigned optlen, int32_t len,
u_char *opt)
{
u_int8_t tfo_len = MAX_TCPOPTLEN - optlen - TCPOLEN_FASTOPEN_REQ;
struct socket *so = tp->t_inpcb->inp_socket;
unsigned ret = 0;
int res;
u_char *bp;
if (so->so_flags1 & SOF1_DATA_AUTHENTICATED) {
if (len > 0) {
tp->t_tfo_stats |= TFO_S_SYN_DATA_SENT;
tcpstat.tcps_tfo_syn_data_sent++;
}
return (0);
}
bp = opt + optlen;
res = tcp_cache_get_cookie(tp, bp + TCPOLEN_FASTOPEN_REQ, &tfo_len);
if (res == 0) {
*bp++ = TCPOPT_FASTOPEN;
*bp++ = TCPOLEN_FASTOPEN_REQ;
ret += TCPOLEN_FASTOPEN_REQ;
tp->t_tfo_flags |= TFO_F_COOKIE_REQ;
tp->t_tfo_stats |= TFO_S_COOKIE_REQ;
tcpstat.tcps_tfo_cookie_req++;
} else {
*bp++ = TCPOPT_FASTOPEN;
*bp++ = TCPOLEN_FASTOPEN_REQ + tfo_len;
ret += TCPOLEN_FASTOPEN_REQ + tfo_len;
tp->t_tfo_flags |= TFO_F_COOKIE_SENT;
if (len > 0) {
tp->t_tfo_stats |= TFO_S_SYN_DATA_SENT;
tcpstat.tcps_tfo_syn_data_sent++;
}
}
return (ret);
}
static inline bool
tcp_send_ecn_flags_on_syn(struct tcpcb *tp, struct socket *so)
{
return(!((tp->ecn_flags & TE_SETUPSENT) ||
(so->so_flags & SOF_MP_SUBFLOW) ||
(tp->t_flagsext & TF_FASTOPEN)));
}
void
tcp_set_ecn(struct tcpcb *tp, struct ifnet *ifp)
{
boolean_t inbound;
if (tp->ecn_flags & TE_ECN_MODE_ENABLE) {
tp->ecn_flags |= TE_ENABLE_ECN;
goto check_heuristic;
}
if (tp->ecn_flags & TE_ECN_MODE_DISABLE) {
tp->ecn_flags &= ~TE_ENABLE_ECN;
return;
}
if (ifp != NULL) {
if (ifp->if_eflags & IFEF_ECN_ENABLE) {
tp->ecn_flags |= TE_ENABLE_ECN;
goto check_heuristic;
}
if (ifp->if_eflags & IFEF_ECN_DISABLE) {
tp->ecn_flags &= ~TE_ENABLE_ECN;
return;
}
}
inbound = (tp->t_inpcb->inp_socket->so_head != NULL);
if ((inbound && tcp_ecn_inbound == 1) ||
(!inbound && tcp_ecn_outbound == 1)) {
tp->ecn_flags |= TE_ENABLE_ECN;
goto check_heuristic;
} else {
tp->ecn_flags &= ~TE_ENABLE_ECN;
}
return;
check_heuristic:
if (!tcp_heuristic_do_ecn(tp))
tp->ecn_flags &= ~TE_ENABLE_ECN;
if ((tp->ecn_flags & (TE_ECN_MODE_ENABLE | TE_ECN_MODE_DISABLE
| TE_ENABLE_ECN)) == TE_ENABLE_ECN) {
if ((tp->iss % 100) >= tcp_ecn_setup_percentage)
tp->ecn_flags &= ~TE_ENABLE_ECN;
}
}
int
tcp_output(struct tcpcb *tp)
{
struct inpcb *inp = tp->t_inpcb;
struct socket *so = inp->inp_socket;
int32_t len, recwin, sendwin, off;
int flags, error;
struct mbuf *m;
struct ip *ip = NULL;
struct ipovly *ipov = NULL;
#if INET6
struct ip6_hdr *ip6 = NULL;
#endif
struct tcphdr *th;
u_char opt[TCP_MAXOLEN];
unsigned ipoptlen, optlen, hdrlen;
int idle, sendalot, lost = 0;
int i, sack_rxmit;
int tso = 0;
int sack_bytes_rxmt;
tcp_seq old_snd_nxt = 0;
struct sackhole *p;
#if IPSEC
unsigned ipsec_optlen = 0;
#endif
int idle_time = 0;
struct mbuf *packetlist = NULL;
struct mbuf *tp_inp_options = inp->inp_depend4.inp4_options;
#if INET6
int isipv6 = inp->inp_vflag & INP_IPV6 ;
#else
int isipv6 = 0;
#endif
short packchain_listadd = 0;
int so_options = so->so_options;
struct rtentry *rt;
u_int32_t svc_flags = 0, allocated_len;
u_int32_t lro_ackmore = (tp->t_lropktlen != 0) ? 1 : 0;
struct mbuf *mnext = NULL;
int sackoptlen = 0;
#if MPTCP
boolean_t mptcp_acknow;
#endif
boolean_t cell = FALSE;
boolean_t wifi = FALSE;
boolean_t wired = FALSE;
boolean_t sack_rescue_rxt = FALSE;
int sotc = so->so_traffic_class;
idle = (tp->t_flags & TF_LASTIDLE) || (tp->snd_max == tp->snd_una);
idle_time = tcp_now - tp->t_rcvtime;
if (idle && idle_time >= TCP_IDLETIMEOUT(tp)) {
if (CC_ALGO(tp)->after_idle != NULL &&
(tp->tcp_cc_index != TCP_CC_ALGO_CUBIC_INDEX ||
idle_time >= TCP_CC_CWND_NONVALIDATED_PERIOD)) {
CC_ALGO(tp)->after_idle(tp);
tcp_ccdbg_trace(tp, NULL, TCP_CC_IDLE_TIMEOUT);
}
if (!SLIST_EMPTY(&tp->t_rxt_segments))
tcp_rxtseg_clean(tp);
tcp_cc_after_idle_stretchack(tp);
}
tp->t_flags &= ~TF_LASTIDLE;
if (idle) {
if (tp->t_flags & TF_MORETOCOME) {
tp->t_flags |= TF_LASTIDLE;
idle = 0;
}
}
#if MPTCP
if (tp->t_mpflags & TMPF_RESET) {
tcp_check_timer_state(tp);
return (0);
}
#endif
again:
#if MPTCP
mptcp_acknow = FALSE;
#endif
KERNEL_DEBUG(DBG_FNC_TCP_OUTPUT | DBG_FUNC_START, 0,0,0,0,0);
#if INET6
if (isipv6) {
KERNEL_DEBUG(DBG_LAYER_BEG,
((inp->inp_fport << 16) | inp->inp_lport),
(((inp->in6p_laddr.s6_addr16[0] & 0xffff) << 16) |
(inp->in6p_faddr.s6_addr16[0] & 0xffff)),
sendalot,0,0);
} else
#endif
{
KERNEL_DEBUG(DBG_LAYER_BEG,
((inp->inp_fport << 16) | inp->inp_lport),
(((inp->inp_laddr.s_addr & 0xffff) << 16) |
(inp->inp_faddr.s_addr & 0xffff)),
sendalot,0,0);
}
rt = inp->inp_route.ro_rt;
if (rt != NULL && ROUTE_UNUSABLE(&tp->t_inpcb->inp_route)) {
struct ifnet *ifp;
struct in_ifaddr *ia = NULL;
struct in6_ifaddr *ia6 = NULL;
int found_srcaddr = 0;
somultipages(so, FALSE);
tp->t_flags &= ~TF_TSO;
soif2kcl(so, FALSE);
if (isipv6) {
ia6 = ifa_foraddr6(&inp->in6p_laddr);
if (ia6 != NULL)
found_srcaddr = 1;
} else {
ia = ifa_foraddr(inp->inp_laddr.s_addr);
if (ia != NULL)
found_srcaddr = 1;
}
if (found_srcaddr == 0) {
soevent(so,
(SO_FILT_HINT_LOCKED | SO_FILT_HINT_NOSRCADDR));
if (tp->t_state >= TCPS_CLOSE_WAIT) {
tcp_drop(tp, EADDRNOTAVAIL);
return(EADDRNOTAVAIL);
}
if (!tp->t_timer[TCPT_REXMT]) {
tp->t_timer[TCPT_REXMT] =
OFFSET_FROM_START(tp, tp->t_rxtcur);
if (tp->t_timer[TCPT_PERSIST]) {
tp->t_timer[TCPT_PERSIST] = 0;
tp->t_persist_stop = 0;
TCP_RESET_REXMT_STATE(tp);
}
}
if (tp->t_pktlist_head != NULL)
m_freem_list(tp->t_pktlist_head);
TCP_PKTLIST_CLEAR(tp);
if (so->so_flags & SOF_NOADDRAVAIL) {
tcp_drop(tp, EADDRNOTAVAIL);
return(EADDRNOTAVAIL);
} else {
tcp_check_timer_state(tp);
return(0);
}
}
if (ia != NULL)
IFA_REMREF(&ia->ia_ifa);
if (ia6 != NULL)
IFA_REMREF(&ia6->ia_ifa);
RT_LOCK(rt);
if ((ifp = rt->rt_ifp) != NULL) {
somultipages(so, (ifp->if_hwassist & IFNET_MULTIPAGES));
tcp_set_tso(tp, ifp);
soif2kcl(so, (ifp->if_eflags & IFEF_2KCL));
tcp_set_ecn(tp, ifp);
}
if (rt->rt_flags & RTF_UP)
RT_GENID_SYNC(rt);
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;
RT_UNLOCK(rt);
}
if (rt != NULL) {
cell = IFNET_IS_CELLULAR(rt->rt_ifp);
wifi = (!cell && IFNET_IS_WIFI(rt->rt_ifp));
wired = (!wifi && IFNET_IS_WIRED(rt->rt_ifp));
}
if (SACK_ENABLED(tp) && SEQ_LT(tp->snd_nxt, tp->snd_max))
tcp_sack_adjust(tp);
sendalot = 0;
off = tp->snd_nxt - tp->snd_una;
sendwin = min(tp->snd_wnd, tp->snd_cwnd);
if (tp->t_flags & TF_SLOWLINK && slowlink_wsize > 0)
sendwin = min(sendwin, slowlink_wsize);
flags = tcp_outflags[tp->t_state];
sack_rxmit = 0;
sack_bytes_rxmt = 0;
len = 0;
p = NULL;
if (SACK_ENABLED(tp) && IN_FASTRECOVERY(tp) &&
(p = tcp_sack_output(tp, &sack_bytes_rxmt))) {
int32_t cwin;
cwin = min(tp->snd_wnd, tp->snd_cwnd) - sack_bytes_rxmt;
if (cwin < 0)
cwin = 0;
if (SEQ_GT(p->end, tp->snd_recover)) {
if (SEQ_GEQ(p->rxmit, tp->snd_recover)) {
p = NULL;
goto after_sack_rexmit;
} else
len = ((int32_t)min(cwin,
tp->snd_recover - p->rxmit));
} else {
len = ((int32_t)min(cwin, p->end - p->rxmit));
}
if (len > 0) {
off = p->rxmit - tp->snd_una;
sack_rxmit = 1;
sendalot = 1;
tcpstat.tcps_sack_rexmits++;
tcpstat.tcps_sack_rexmit_bytes +=
min(len, tp->t_maxseg);
} else {
len = 0;
}
}
after_sack_rexmit:
if (tp->t_flags & TF_NEEDFIN)
flags |= TH_FIN;
if (tp->t_flags & TF_NEEDSYN)
flags |= TH_SYN;
if (tp->t_flagsext & TF_FORCE) {
if (sendwin == 0) {
if (off < so->so_snd.sb_cc)
flags &= ~TH_FIN;
sendwin = 1;
} else {
tp->t_timer[TCPT_PERSIST] = 0;
tp->t_persist_stop = 0;
TCP_RESET_REXMT_STATE(tp);
}
}
if (sack_rxmit == 0) {
if (sack_bytes_rxmt == 0) {
len = min(so->so_snd.sb_cc, sendwin) - off;
} else {
int32_t cwin;
cwin = tp->snd_cwnd -
(tp->snd_nxt - tp->sack_newdata) -
sack_bytes_rxmt;
if (cwin < 0)
cwin = 0;
len = min(so->so_snd.sb_cc, tp->snd_wnd)
- off;
if (len > 0) {
len = imin(len, cwin);
} else {
len = 0;
}
if (len == 0 && cwin > 0 &&
SEQ_LT(tp->snd_fack, tp->snd_recover) &&
!(tp->t_flagsext & TF_RESCUE_RXT)) {
len = min((tp->snd_recover - tp->snd_fack),
tp->t_maxseg);
len = imin(len, cwin);
old_snd_nxt = tp->snd_nxt;
sack_rescue_rxt = TRUE;
tp->snd_nxt = tp->snd_recover - len;
if ((tp->t_flags & TF_SENTFIN) &&
tp->snd_max == tp->snd_recover)
tp->snd_nxt--;
off = tp->snd_nxt - tp->snd_una;
sendalot = 0;
tp->t_flagsext |= TF_RESCUE_RXT;
}
}
}
if ((flags & TH_SYN) && SEQ_GT(tp->snd_nxt, tp->snd_una)) {
if (tp->t_state != TCPS_SYN_RECEIVED || tfo_enabled(tp))
flags &= ~TH_SYN;
off--;
len++;
if (len > 0 && tp->t_state == TCPS_SYN_SENT) {
while (inp->inp_sndinprog_cnt == 0 &&
tp->t_pktlist_head != NULL) {
packetlist = tp->t_pktlist_head;
packchain_listadd = tp->t_lastchain;
packchain_sent++;
TCP_PKTLIST_CLEAR(tp);
error = tcp_ip_output(so, tp, packetlist,
packchain_listadd, tp_inp_options,
(so_options & SO_DONTROUTE),
(sack_rxmit || (sack_bytes_rxmt != 0)),
isipv6);
}
if (inp->inp_sndinprog_cnt == 0 &&
(tp->t_flags & TF_CLOSING)) {
tp->t_flags &= ~TF_CLOSING;
(void) tcp_close(tp);
} else {
tcp_check_timer_state(tp);
}
KERNEL_DEBUG(DBG_FNC_TCP_OUTPUT | DBG_FUNC_END,
0,0,0,0,0);
return(0);
}
}
if ((flags & TH_SYN) && !tfo_enabled(tp)) {
len = 0;
flags &= ~TH_FIN;
}
if ((flags & TH_SYN) && tp->t_state <= TCPS_SYN_SENT && tfo_enabled(tp))
len = tcp_tfo_check(tp, len);
if (len <= 0 && !(flags & TH_SYN)) {
len = 0;
if (sendwin == 0) {
tp->t_timer[TCPT_REXMT] = 0;
tp->t_timer[TCPT_PTO] = 0;
TCP_RESET_REXMT_STATE(tp);
tp->snd_nxt = tp->snd_una;
off = 0;
if (tp->t_timer[TCPT_PERSIST] == 0)
tcp_setpersist(tp);
}
}
if (tcp_do_autosendbuf == 1 &&
!INP_WAIT_FOR_IF_FEEDBACK(inp) && !IN_FASTRECOVERY(tp) &&
(so->so_snd.sb_flags & (SB_AUTOSIZE | SB_TRIM)) == SB_AUTOSIZE &&
tcp_cansbgrow(&so->so_snd)) {
if ((tp->snd_wnd / 4 * 5) >= so->so_snd.sb_hiwat &&
so->so_snd.sb_cc >= (so->so_snd.sb_hiwat / 8 * 7) &&
sendwin >= (so->so_snd.sb_cc - (tp->snd_nxt - tp->snd_una))) {
if (sbreserve(&so->so_snd,
min(so->so_snd.sb_hiwat + tcp_autosndbuf_inc,
tcp_autosndbuf_max)) == 1) {
so->so_snd.sb_idealsize = so->so_snd.sb_hiwat;
}
}
}
#if IPSEC
if (ipsec_bypass == 0)
ipsec_optlen = ipsec_hdrsiz_tcp(tp);
#endif
if (len > tp->t_maxseg) {
if ((tp->t_flags & TF_TSO) && tcp_do_tso && hwcksum_tx &&
ip_use_randomid && kipf_count == 0 &&
dlil_filter_disable_tso_count == 0 &&
tp->rcv_numsacks == 0 && sack_rxmit == 0 &&
sack_bytes_rxmt == 0 &&
inp->inp_options == NULL &&
inp->in6p_options == NULL
#if IPSEC
&& ipsec_optlen == 0
#endif
#if IPFIREWALL
&& (fw_enable == 0 || fw_bypass)
#endif
) {
tso = 1;
sendalot = 0;
} else {
len = tp->t_maxseg;
sendalot = 1;
tso = 0;
}
}
if (tp->t_flagsext & TF_SENT_TLPROBE) {
len = min(len, tp->t_maxseg);
sendalot = 0;
tso = 0;
}
#if MPTCP
if ((so->so_flags & SOF_MP_SUBFLOW) &&
!(tp->t_mpflags & TMPF_TCP_FALLBACK)) {
int newlen = len;
if (tp->t_state >= TCPS_ESTABLISHED &&
(tp->t_mpflags & TMPF_SND_MPPRIO ||
tp->t_mpflags & TMPF_SND_REM_ADDR ||
tp->t_mpflags & TMPF_SND_MPFAIL ||
tp->t_mpflags & TMPF_SND_KEYS ||
tp->t_mpflags & TMPF_SND_JACK)) {
if (len > 0) {
len = 0;
}
if (!(tp->t_mpflags & TMPF_PREESTABLISHED))
sendalot = 1;
mptcp_acknow = TRUE;
} else {
mptcp_acknow = FALSE;
}
if (len > 0)
newlen = mptcp_adj_sendlen(so, off);
if (newlen < len) {
len = newlen;
sendalot = 1;
}
}
#endif
if (so->so_flags & SOF_ENABLE_MSGS) {
if ((off + len) >
so->so_msg_state->msg_serial_bytes) {
sbpull_unordered_data(so, off, len);
if ((off + len) >
so->so_msg_state->msg_serial_bytes) {
len = so->so_msg_state->msg_serial_bytes - off;
if (len <= 0) {
len = 0;
tcpstat.tcps_msg_sndwaithipri++;
}
}
}
}
if (sack_rxmit) {
if (SEQ_LT(p->rxmit + len, tp->snd_una + so->so_snd.sb_cc))
flags &= ~TH_FIN;
} else {
if (SEQ_LT(tp->snd_nxt + len, tp->snd_una + so->so_snd.sb_cc))
flags &= ~TH_FIN;
}
recwin = tcp_sbspace(tp);
#if MPTCP
if (so->so_flags & SOF_MP_SUBFLOW) {
struct mptcb *mp_tp = tptomptp(tp);
if (mp_tp != NULL) {
mpte_lock_assert_held(mp_tp->mpt_mpte);
recwin = imin(recwin, mptcp_sbspace(mp_tp));
}
}
#endif
if (recwin < (int32_t)(so->so_rcv.sb_hiwat / 4) &&
recwin < (int)tp->t_maxseg)
recwin = 0;
#if TRAFFIC_MGT
if (tcp_recv_bg == 1 || IS_TCP_RECV_BG(so)) {
if (recwin > 0 && tcp_recv_throttle(tp)) {
uint32_t min_iaj_win = tcp_min_iaj_win * tp->t_maxseg;
uint32_t bg_rwintop = tp->rcv_adv;
if (SEQ_LT(bg_rwintop, tp->rcv_nxt + min_iaj_win))
bg_rwintop = tp->rcv_nxt + min_iaj_win;
recwin = imin((int32_t)(bg_rwintop - tp->rcv_nxt),
recwin);
if (recwin < 0)
recwin = 0;
}
}
#endif
if (recwin > (int32_t)(TCP_MAXWIN << tp->rcv_scale))
recwin = (int32_t)(TCP_MAXWIN << tp->rcv_scale);
if (!(so->so_flags & SOF_MPTCP_TRUE) &&
recwin < (int32_t)(tp->rcv_adv - tp->rcv_nxt))
recwin = (int32_t)(tp->rcv_adv - tp->rcv_nxt);
if (len) {
if (tp->t_flagsext & TF_FORCE)
goto send;
if (SEQ_LT(tp->snd_nxt, tp->snd_max))
goto send;
if (sack_rxmit)
goto send;
if (!INP_WAIT_FOR_IF_FEEDBACK(inp) ||
tp->t_state != TCPS_ESTABLISHED) {
if (len >= tp->t_maxseg)
goto send;
if (!(tp->t_flags & TF_MORETOCOME) &&
(idle || tp->t_flags & TF_NODELAY ||
(tp->t_flags & TF_MAXSEGSNT) ||
ALLOW_LIMITED_TRANSMIT(tp)) &&
(tp->t_flags & TF_NOPUSH) == 0 &&
(len + off >= so->so_snd.sb_cc ||
so->so_flags & SOF_MP_SUBFLOW))
goto send;
if (len >= tp->max_sndwnd / 2 && tp->max_sndwnd > 0)
goto send;
} else {
tcpstat.tcps_fcholdpacket++;
}
}
if (recwin > 0 && !(tp->t_flags & TF_NEEDSYN)) {
int32_t adv, oldwin = 0;
adv = imin(recwin, (int)TCP_MAXWIN << tp->rcv_scale) -
(tp->rcv_adv - tp->rcv_nxt);
if (SEQ_GT(tp->rcv_adv, tp->rcv_nxt))
oldwin = tp->rcv_adv - tp->rcv_nxt;
if (adv >= (int32_t) (2 * tp->t_maxseg)) {
if (!(tp->t_flags & TF_STRETCHACK) &&
(tp->last_ack_sent != tp->rcv_nxt ||
((oldwin + adv) >> tp->rcv_scale) >
(oldwin >> tp->rcv_scale))) {
goto send;
}
}
if (4 * adv >= (int32_t) so->so_rcv.sb_hiwat)
goto send;
if ((tp->t_flags & TF_STRETCHACK) &&
!(tp->t_flags & TF_DELACK)) {
tp->t_flags |= TF_DELACK;
tp->t_timer[TCPT_DELACK] =
OFFSET_FROM_START(tp, tcp_delack);
}
}
if (tp->t_flags & TF_ACKNOW)
goto send;
if ((flags & TH_RST) ||
((flags & TH_SYN) && (tp->t_flags & TF_NEEDSYN) == 0))
goto send;
if (SEQ_GT(tp->snd_up, tp->snd_una))
goto send;
#if MPTCP
if (mptcp_acknow)
goto send;
#endif
if ((flags & TH_FIN) &&
(!(tp->t_flags & TF_SENTFIN) || tp->snd_nxt == tp->snd_una))
goto send;
if (SACK_ENABLED(tp) && (tp->t_state >= TCPS_ESTABLISHED) &&
SEQ_GT(tp->snd_max, tp->snd_una) &&
tp->t_timer[TCPT_REXMT] == 0 &&
tp->t_timer[TCPT_PERSIST] == 0) {
tp->t_timer[TCPT_REXMT] = OFFSET_FROM_START(tp,
tp->t_rxtcur);
goto just_return;
}
if (so->so_snd.sb_cc && tp->t_timer[TCPT_REXMT] == 0 &&
tp->t_timer[TCPT_PERSIST] == 0) {
TCP_RESET_REXMT_STATE(tp);
tcp_setpersist(tp);
}
just_return:
while (inp->inp_sndinprog_cnt == 0 &&
tp->t_pktlist_head != NULL) {
packetlist = tp->t_pktlist_head;
packchain_listadd = tp->t_lastchain;
packchain_sent++;
TCP_PKTLIST_CLEAR(tp);
error = tcp_ip_output(so, tp, packetlist,
packchain_listadd,
tp_inp_options, (so_options & SO_DONTROUTE),
(sack_rxmit || (sack_bytes_rxmt != 0)), isipv6);
}
if (inp->inp_sndinprog_cnt == 0 &&
(tp->t_flags & TF_CLOSING)) {
tp->t_flags &= ~TF_CLOSING;
(void) tcp_close(tp);
} else {
tcp_check_timer_state(tp);
}
KERNEL_DEBUG(DBG_FNC_TCP_OUTPUT | DBG_FUNC_END, 0,0,0,0,0);
return (0);
send:
if (len > 0) {
if (len >= tp->t_maxseg)
tp->t_flags |= TF_MAXSEGSNT;
else
tp->t_flags &= ~TF_MAXSEGSNT;
}
optlen = 0;
#if INET6
if (isipv6)
hdrlen = sizeof (struct ip6_hdr) + sizeof (struct tcphdr);
else
#endif
hdrlen = sizeof (struct tcpiphdr);
if (flags & TH_SYN) {
tp->snd_nxt = tp->iss;
if ((tp->t_flags & TF_NOOPT) == 0) {
u_short mss;
opt[0] = TCPOPT_MAXSEG;
opt[1] = TCPOLEN_MAXSEG;
mss = htons((u_short) tcp_mssopt(tp));
(void)memcpy(opt + 2, &mss, sizeof(mss));
optlen = TCPOLEN_MAXSEG;
if ((tp->t_flags & TF_REQ_SCALE) &&
((flags & TH_ACK) == 0 ||
(tp->t_flags & TF_RCVD_SCALE))) {
*((u_int32_t *)(void *)(opt + optlen)) = htonl(
TCPOPT_NOP << 24 |
TCPOPT_WINDOW << 16 |
TCPOLEN_WINDOW << 8 |
tp->request_r_scale);
optlen += 4;
}
#if MPTCP
if (mptcp_enable && (so->so_flags & SOF_MP_SUBFLOW)) {
optlen = mptcp_setup_syn_opts(so, opt, optlen);
}
#endif
}
}
if ((tp->t_flags & (TF_REQ_TSTMP|TF_NOOPT)) == TF_REQ_TSTMP &&
(flags & TH_RST) == 0 &&
((flags & TH_ACK) == 0 ||
(tp->t_flags & TF_RCVD_TSTMP))) {
u_int32_t *lp = (u_int32_t *)(void *)(opt + optlen);
*lp++ = htonl(TCPOPT_TSTAMP_HDR);
*lp++ = htonl(tcp_now);
*lp = htonl(tp->ts_recent);
optlen += TCPOLEN_TSTAMP_APPA;
}
if (tp->rfbuf_ts == 0 && (so->so_rcv.sb_flags & SB_AUTOSIZE))
tp->rfbuf_ts = tcp_now;
if (SACK_ENABLED(tp) && ((tp->t_flags & TF_NOOPT) == 0)) {
if ((flags & TH_SYN) &&
(!(flags & TH_ACK) || (tp->t_flags & TF_SACK_PERMIT))) {
u_char *bp;
bp = (u_char *)opt + optlen;
*bp++ = TCPOPT_SACK_PERMITTED;
*bp++ = TCPOLEN_SACK_PERMITTED;
optlen += TCPOLEN_SACK_PERMITTED;
}
}
#if MPTCP
if (so->so_flags & SOF_MP_SUBFLOW) {
if (len != 0) {
tp->t_mpflags |= (TMPF_SEND_DSN |
TMPF_MPTCP_ACKNOW);
} else {
tp->t_mpflags |= TMPF_MPTCP_ACKNOW;
}
optlen = mptcp_setup_opts(tp, off, &opt[0], optlen, flags,
len, &mptcp_acknow);
tp->t_mpflags &= ~TMPF_SEND_DSN;
}
#endif
if (tfo_enabled(tp) && !(tp->t_flags & TF_NOOPT) &&
(flags & (TH_SYN | TH_ACK)) == TH_SYN)
optlen += tcp_tfo_write_cookie(tp, optlen, len, opt);
if (tfo_enabled(tp) &&
(flags & (TH_SYN | TH_ACK)) == (TH_SYN | TH_ACK) &&
(tp->t_tfo_flags & TFO_F_OFFER_COOKIE))
optlen += tcp_tfo_write_cookie_rep(tp, optlen, opt);
if (SACK_ENABLED(tp) && ((tp->t_flags & TF_NOOPT) == 0)) {
if (TCPS_HAVEESTABLISHED(tp->t_state) &&
(tp->t_flags & TF_SACK_PERMIT) &&
(tp->rcv_numsacks > 0 || TCP_SEND_DSACK_OPT(tp)) &&
MAX_TCPOPTLEN - optlen - 2 >= TCPOLEN_SACK) {
int nsack, padlen;
u_char *bp = (u_char *)opt + optlen;
u_int32_t *lp;
nsack = (MAX_TCPOPTLEN - optlen - 2) / TCPOLEN_SACK;
nsack = min(nsack, (tp->rcv_numsacks +
(TCP_SEND_DSACK_OPT(tp) ? 1 : 0)));
sackoptlen = (2 + nsack * TCPOLEN_SACK);
padlen = (MAX_TCPOPTLEN - optlen - sackoptlen) % 4;
optlen += padlen;
while (padlen-- > 0)
*bp++ = TCPOPT_NOP;
tcpstat.tcps_sack_send_blocks++;
*bp++ = TCPOPT_SACK;
*bp++ = sackoptlen;
lp = (u_int32_t *)(void *)bp;
if (TCP_SEND_DSACK_OPT(tp) && nsack > 0 &&
(tp->rcv_numsacks == 0 || nsack > 1)) {
*lp++ = htonl(tp->t_dsack_lseq);
*lp++ = htonl(tp->t_dsack_rseq);
tcpstat.tcps_dsack_sent++;
tp->t_dsack_sent++;
nsack--;
}
VERIFY(nsack == 0 || tp->rcv_numsacks >= nsack);
for (i = 0; i < nsack; i++) {
struct sackblk sack = tp->sackblks[i];
*lp++ = htonl(sack.start);
*lp++ = htonl(sack.end);
}
optlen += sackoptlen;
}
}
if (optlen < MAX_TCPOPTLEN && (optlen % sizeof(u_int32_t))) {
int pad = sizeof(u_int32_t) - (optlen % sizeof(u_int32_t));
u_char *bp = (u_char *)opt + optlen;
optlen += pad;
while (pad) {
*bp++ = TCPOPT_EOL;
pad--;
}
}
if ((flags & (TH_SYN | TH_ACK)) == (TH_SYN | TH_ACK) &&
(tp->ecn_flags & TE_ENABLE_ECN)) {
if (tp->ecn_flags & TE_SETUPRECEIVED) {
if (tcp_send_ecn_flags_on_syn(tp, so)) {
flags |= TH_ECE;
tp->ecn_flags |= (TE_SETUPSENT|TE_SENDIPECT);
tcpstat.tcps_ecn_server_setup++;
tcpstat.tcps_ecn_server_success++;
} else {
if (tp->ecn_flags & TE_SETUPSENT) {
tcpstat.tcps_ecn_lost_synack++;
tcpstat.tcps_ecn_server_success--;
tp->ecn_flags |= TE_LOST_SYNACK;
}
tp->ecn_flags &=
~(TE_SETUPRECEIVED | TE_SENDIPECT |
TE_SENDCWR);
}
}
} else if ((flags & (TH_SYN | TH_ACK)) == TH_SYN &&
(tp->ecn_flags & TE_ENABLE_ECN)) {
if (tcp_send_ecn_flags_on_syn(tp, so)) {
flags |= (TH_ECE | TH_CWR);
tcpstat.tcps_ecn_client_setup++;
tp->ecn_flags |= TE_CLIENT_SETUP;
tp->ecn_flags |= (TE_SETUPSENT | TE_SENDIPECT);
} else {
if (tp->ecn_flags & TE_SETUPSENT) {
tcpstat.tcps_ecn_lost_syn++;
tp->ecn_flags |= TE_LOST_SYN;
}
tp->ecn_flags &= ~TE_SENDIPECT;
}
}
if ((tp->ecn_flags & TE_SENDCWR) != 0 && len != 0 &&
!SEQ_LT(tp->snd_nxt, tp->snd_max) && !sack_rxmit) {
flags |= TH_CWR;
tp->ecn_flags &= ~TE_SENDCWR;
}
if ((tp->ecn_flags & TE_SENDECE) != 0 && len == 0) {
flags |= TH_ECE;
tcpstat.tcps_ecn_sent_ece++;
}
hdrlen += optlen;
tp->t_dsack_lseq = 0;
tp->t_dsack_rseq = 0;
#if INET6
if (isipv6)
ipoptlen = ip6_optlen(inp);
else
#endif
{
if (tp_inp_options) {
ipoptlen = tp_inp_options->m_len -
offsetof(struct ipoption, ipopt_list);
} else {
ipoptlen = 0;
}
}
#if IPSEC
ipoptlen += ipsec_optlen;
#endif
if (len + optlen + ipoptlen > tp->t_maxopd) {
flags &= ~TH_FIN;
if (tso) {
int32_t tso_maxlen;
tso_maxlen = tp->tso_max_segment_size ?
tp->tso_max_segment_size : TCP_MAXWIN;
if (len > tso_maxlen - hdrlen - optlen) {
len = tso_maxlen - hdrlen - optlen;
len = len - (len % (tp->t_maxopd - optlen));
sendalot = 1;
} else if (tp->t_flags & TF_NEEDFIN) {
sendalot = 1;
}
} else {
len = tp->t_maxopd - optlen - ipoptlen;
sendalot = 1;
}
}
if (max_linkhdr + hdrlen > MCLBYTES)
panic("tcphdr too big");
if ((tp->t_flagsext & TF_MEASURESNDBW) != 0 &&
(tp->t_bwmeas != NULL) &&
(tp->t_flagsext & TF_BWMEAS_INPROGRESS) == 0) {
tp->t_bwmeas->bw_size = min(min(
(so->so_snd.sb_cc - (tp->snd_max - tp->snd_una)),
tp->snd_cwnd), tp->snd_wnd);
if (tp->t_bwmeas->bw_minsize > 0 &&
tp->t_bwmeas->bw_size < tp->t_bwmeas->bw_minsize)
tp->t_bwmeas->bw_size = 0;
if (tp->t_bwmeas->bw_maxsize > 0)
tp->t_bwmeas->bw_size = min(tp->t_bwmeas->bw_size,
tp->t_bwmeas->bw_maxsize);
if (tp->t_bwmeas->bw_size > 0) {
tp->t_flagsext |= TF_BWMEAS_INPROGRESS;
tp->t_bwmeas->bw_start = tp->snd_max;
tp->t_bwmeas->bw_ts = tcp_now;
}
}
VERIFY(inp->inp_flowhash != 0);
if (len) {
tp->t_pmtud_lastseg_size = len + optlen + ipoptlen;
if ((tp->t_flagsext & TF_FORCE) && len == 1)
tcpstat.tcps_sndprobe++;
else if (SEQ_LT(tp->snd_nxt, tp->snd_max) || sack_rxmit) {
tcpstat.tcps_sndrexmitpack++;
tcpstat.tcps_sndrexmitbyte += len;
if (nstat_collect) {
nstat_route_tx(inp->inp_route.ro_rt, 1,
len, NSTAT_TX_FLAG_RETRANSMIT);
INP_ADD_STAT(inp, cell, wifi, wired,
txpackets, 1);
INP_ADD_STAT(inp, cell, wifi, wired,
txbytes, len);
tp->t_stat.txretransmitbytes += len;
tp->t_stat.rxmitpkts++;
}
} else {
tcpstat.tcps_sndpack++;
tcpstat.tcps_sndbyte += len;
if (nstat_collect) {
INP_ADD_STAT(inp, cell, wifi, wired,
txpackets, 1);
INP_ADD_STAT(inp, cell, wifi, wired,
txbytes, len);
}
inp_decr_sndbytes_unsent(so, len);
}
inp_set_activity_bitmap(inp);
#if MPTCP
if (tp->t_mpflags & TMPF_MPTCP_TRUE) {
tcpstat.tcps_mp_sndpacks++;
tcpstat.tcps_mp_sndbytes += len;
}
#endif
m = NULL;
allocated_len = MHLEN;
if (MHLEN < hdrlen + max_linkhdr) {
MGETHDR(m, M_DONTWAIT, MT_HEADER);
if (m == NULL) {
error = ENOBUFS;
goto out;
}
MCLGET(m, M_DONTWAIT);
if ((m->m_flags & M_EXT) == 0) {
m_freem(m);
error = ENOBUFS;
goto out;
}
m->m_data += max_linkhdr;
m->m_len = hdrlen;
allocated_len = MCLBYTES;
}
if (len <= allocated_len - hdrlen - max_linkhdr) {
if (m == NULL) {
VERIFY(allocated_len <= MHLEN);
MGETHDR(m, M_DONTWAIT, MT_HEADER);
if (m == NULL) {
error = ENOBUFS;
goto out;
}
m->m_data += max_linkhdr;
m->m_len = hdrlen;
}
if (so->so_snd.sb_mb == NULL || off < 0) {
if (m != NULL) m_freem(m);
error = 0;
goto out;
}
m_copydata(so->so_snd.sb_mb, off, (int) len,
mtod(m, caddr_t) + hdrlen);
m->m_len += len;
} else {
uint32_t copymode;
copymode = M_COPYM_MOVE_HDR;
#if MPTCP
if (so->so_flags & SOF_MP_SUBFLOW) {
copymode = M_COPYM_NOOP_HDR;
}
#endif
if (m != NULL) {
m->m_next = m_copym_mode(so->so_snd.sb_mb,
off, (int)len, M_DONTWAIT, copymode);
if (m->m_next == NULL) {
(void) m_free(m);
error = ENOBUFS;
goto out;
}
} else {
if (so->so_snd.sb_mb == NULL) {
error = 0;
goto out;
}
if ((m = m_copym_with_hdrs(so->so_snd.sb_mb,
off, len, M_DONTWAIT, NULL, NULL,
copymode)) == NULL) {
error = ENOBUFS;
goto out;
}
m->m_data += max_linkhdr;
m->m_len = hdrlen;
}
}
if (off + len == so->so_snd.sb_cc && !(flags & TH_SYN))
flags |= TH_PUSH;
} else {
if (tp->t_flags & TF_ACKNOW)
tcpstat.tcps_sndacks++;
else if (flags & (TH_SYN|TH_FIN|TH_RST))
tcpstat.tcps_sndctrl++;
else if (SEQ_GT(tp->snd_up, tp->snd_una))
tcpstat.tcps_sndurg++;
else
tcpstat.tcps_sndwinup++;
MGETHDR(m, M_DONTWAIT, MT_HEADER);
if (m == NULL) {
error = ENOBUFS;
goto out;
}
if (MHLEN < (hdrlen + max_linkhdr)) {
MCLGET(m, M_DONTWAIT);
if ((m->m_flags & M_EXT) == 0) {
m_freem(m);
error = ENOBUFS;
goto out;
}
}
m->m_data += max_linkhdr;
m->m_len = hdrlen;
}
m->m_pkthdr.rcvif = 0;
#if CONFIG_MACF_NET
mac_mbuf_label_associate_inpcb(inp, m);
#endif
#if INET6
if (isipv6) {
ip6 = mtod(m, struct ip6_hdr *);
th = (struct tcphdr *)(void *)(ip6 + 1);
tcp_fillheaders(tp, ip6, th);
if ((tp->ecn_flags & TE_SENDIPECT) != 0 && len &&
!SEQ_LT(tp->snd_nxt, tp->snd_max) && !sack_rxmit) {
ip6->ip6_flow |= htonl(IPTOS_ECN_ECT0 << 20);
}
svc_flags |= PKT_SCF_IPV6;
#if PF_ECN
m_pftag(m)->pftag_hdr = (void *)ip6;
m_pftag(m)->pftag_flags |= PF_TAG_HDR_INET6;
#endif
} else
#endif
{
ip = mtod(m, struct ip *);
ipov = (struct ipovly *)ip;
th = (struct tcphdr *)(void *)(ip + 1);
tcp_fillheaders(tp, ip, th);
if ((tp->ecn_flags & TE_SENDIPECT) != 0 && len &&
!SEQ_LT(tp->snd_nxt, tp->snd_max) &&
!sack_rxmit && !(flags & TH_SYN)) {
ip->ip_tos |= IPTOS_ECN_ECT0;
}
#if PF_ECN
m_pftag(m)->pftag_hdr = (void *)ip;
m_pftag(m)->pftag_flags |= PF_TAG_HDR_INET;
#endif
}
if ((flags & TH_FIN) && (tp->t_flags & TF_SENTFIN) &&
tp->snd_nxt == tp->snd_max)
tp->snd_nxt--;
if (sack_rxmit == 0) {
if (len || (flags & (TH_SYN|TH_FIN)) ||
tp->t_timer[TCPT_PERSIST]) {
th->th_seq = htonl(tp->snd_nxt);
if (len > 0) {
m->m_pkthdr.tx_start_seq = tp->snd_nxt;
m->m_pkthdr.pkt_flags |= PKTF_START_SEQ;
}
if (SEQ_LT(tp->snd_nxt, tp->snd_max)) {
if (SACK_ENABLED(tp) && len > 1) {
tcp_rxtseg_insert(tp, tp->snd_nxt,
(tp->snd_nxt + len - 1));
}
if (len > 0)
m->m_pkthdr.pkt_flags |=
PKTF_TCP_REXMT;
}
} else {
th->th_seq = htonl(tp->snd_max);
}
} else {
th->th_seq = htonl(p->rxmit);
if (len > 0) {
m->m_pkthdr.pkt_flags |=
(PKTF_TCP_REXMT | PKTF_START_SEQ);
m->m_pkthdr.tx_start_seq = p->rxmit;
}
tcp_rxtseg_insert(tp, p->rxmit, (p->rxmit + len - 1));
p->rxmit += len;
tp->sackhint.sack_bytes_rexmit += len;
}
th->th_ack = htonl(tp->rcv_nxt);
tp->last_ack_sent = tp->rcv_nxt;
if (optlen) {
bcopy(opt, th + 1, optlen);
th->th_off = (sizeof (struct tcphdr) + optlen) >> 2;
}
th->th_flags = flags;
th->th_win = htons((u_short) (recwin>>tp->rcv_scale));
if (recwin > 0 && SEQ_LT(tp->rcv_adv, tp->rcv_nxt + recwin))
tp->rcv_adv = tp->rcv_nxt + recwin;
if (th->th_win == 0)
tp->t_flags |= TF_RXWIN0SENT;
else
tp->t_flags &= ~TF_RXWIN0SENT;
if (SEQ_GT(tp->snd_up, tp->snd_nxt)) {
th->th_urp = htons((u_short)(tp->snd_up - tp->snd_nxt));
th->th_flags |= TH_URG;
} else {
tp->snd_up = tp->snd_una;
}
m->m_pkthdr.len = hdrlen + len;
if (tp->t_state != TCPS_ESTABLISHED &&
(tp->t_state == TCPS_CLOSING || tp->t_state == TCPS_TIME_WAIT
|| tp->t_state == TCPS_LAST_ACK || (th->th_flags & TH_RST)))
m->m_pkthdr.pkt_flags |= PKTF_LAST_PKT;
#if INET6
if (isipv6) {
m->m_pkthdr.csum_flags = CSUM_TCPIPV6;
m->m_pkthdr.csum_data = offsetof(struct tcphdr, th_sum);
if (len + optlen)
th->th_sum = in_addword(th->th_sum,
htons((u_short)(optlen + len)));
}
else
#endif
{
m->m_pkthdr.csum_flags = CSUM_TCP;
m->m_pkthdr.csum_data = offsetof(struct tcphdr, th_sum);
if (len + optlen)
th->th_sum = in_addword(th->th_sum,
htons((u_short)(optlen + len)));
}
if (tso) {
#if INET6
if (isipv6)
m->m_pkthdr.csum_flags |= CSUM_TSO_IPV6;
else
#endif
m->m_pkthdr.csum_flags |= CSUM_TSO_IPV4;
m->m_pkthdr.tso_segsz = tp->t_maxopd - optlen;
} else {
m->m_pkthdr.tso_segsz = 0;
}
if (!(tp->t_flagsext & TF_FORCE)
|| tp->t_timer[TCPT_PERSIST] == 0) {
tcp_seq startseq = tp->snd_nxt;
if (flags & (TH_SYN|TH_FIN)) {
if (flags & TH_SYN)
tp->snd_nxt++;
if ((flags & TH_FIN) &&
!(tp->t_flags & TF_SENTFIN)) {
tp->snd_nxt++;
tp->t_flags |= TF_SENTFIN;
}
}
if (sack_rxmit)
goto timer;
if (sack_rescue_rxt == TRUE) {
tp->snd_nxt = old_snd_nxt;
sack_rescue_rxt = FALSE;
tcpstat.tcps_pto_in_recovery++;
} else {
tp->snd_nxt += len;
}
if (SEQ_GT(tp->snd_nxt, tp->snd_max)) {
tp->snd_max = tp->snd_nxt;
tp->t_sndtime = tcp_now;
if (tp->t_rtttime == 0) {
tp->t_rtttime = tcp_now;
tp->t_rtseq = startseq;
tcpstat.tcps_segstimed++;
tp->t_pipeack_lastuna = tp->snd_una;
}
}
timer:
if (tp->t_timer[TCPT_REXMT] == 0 &&
((sack_rxmit && tp->snd_nxt != tp->snd_max) ||
tp->snd_nxt != tp->snd_una || (flags & TH_FIN))) {
if (tp->t_timer[TCPT_PERSIST]) {
tp->t_timer[TCPT_PERSIST] = 0;
tp->t_persist_stop = 0;
TCP_RESET_REXMT_STATE(tp);
}
tp->t_timer[TCPT_REXMT] =
OFFSET_FROM_START(tp, tp->t_rxtcur);
}
if (tcp_enable_tlp && tp->t_state == TCPS_ESTABLISHED &&
SACK_ENABLED(tp) && !IN_FASTRECOVERY(tp)
&& tp->snd_nxt == tp->snd_max
&& SEQ_GT(tp->snd_nxt, tp->snd_una)
&& tp->t_rxtshift == 0
&& (tp->t_flagsext & (TF_SENT_TLPROBE|TF_PKTS_REORDERED)) == 0) {
u_int32_t pto, srtt, new_rto = 0;
srtt = tp->t_srtt >> TCP_RTT_SHIFT;
pto = ((TCP_REXMTVAL(tp)) * 3) >> 1;
pto = max (2 * srtt, pto);
if ((tp->snd_max - tp->snd_una) == tp->t_maxseg)
pto = max(pto,
(((3 * pto) >> 2) + tcp_delack * 2));
else
pto = max(10, pto);
if (tp->t_rxtcur < pto) {
pto = tp->t_rxtcur;
TCPT_RANGESET(new_rto,
(pto + TCP_REXMTVAL(tp)),
max(tp->t_rttmin, tp->t_rttcur + 2),
TCPTV_REXMTMAX, 0);
tp->t_timer[TCPT_REXMT] =
OFFSET_FROM_START(tp, new_rto);
}
tp->t_timer[TCPT_PTO] = OFFSET_FROM_START(tp, pto);
}
} else {
int xlen = len;
if (flags & TH_SYN)
++xlen;
if ((flags & TH_FIN) &&
!(tp->t_flags & TF_SENTFIN)) {
++xlen;
tp->t_flags |= TF_SENTFIN;
}
if (SEQ_GT(tp->snd_nxt + xlen, tp->snd_max)) {
tp->snd_max = tp->snd_nxt + len;
tp->t_sndtime = tcp_now;
}
}
#if TCPDEBUG
if (so_options & SO_DEBUG)
tcp_trace(TA_OUTPUT, tp->t_state, tp, mtod(m, void *), th, 0);
#endif
#if INET6
if (isipv6) {
ip6->ip6_hlim = in6_selecthlim(inp, inp->in6p_route.ro_rt ?
inp->in6p_route.ro_rt->rt_ifp : NULL);
KERNEL_DEBUG(DBG_LAYER_BEG,
((inp->inp_fport << 16) | inp->inp_lport),
(((inp->in6p_laddr.s6_addr16[0] & 0xffff) << 16) |
(inp->in6p_faddr.s6_addr16[0] & 0xffff)),
sendalot,0,0);
} else
#endif
{
ip->ip_len = m->m_pkthdr.len;
ip->ip_ttl = inp->inp_ip_ttl;
ip->ip_tos |= (inp->inp_ip_tos & ~IPTOS_ECN_MASK);
KERNEL_DEBUG(DBG_LAYER_BEG,
((inp->inp_fport << 16) | inp->inp_lport),
(((inp->inp_laddr.s_addr & 0xffff) << 16) |
(inp->inp_faddr.s_addr & 0xffff)), 0,0,0);
}
#if INET6
if (!isipv6)
#endif
if (path_mtu_discovery && (tp->t_flags & TF_PMTUD))
ip->ip_off |= IP_DF;
#if NECP
{
necp_kernel_policy_id policy_id;
u_int32_t route_rule_id;
if (!necp_socket_is_allowed_to_send_recv(inp, &policy_id, &route_rule_id)) {
m_freem(m);
error = EHOSTUNREACH;
goto out;
}
necp_mark_packet_from_socket(m, inp, policy_id, route_rule_id);
if (net_qos_policy_restricted != 0) {
necp_socket_update_qos_marking(inp, inp->inp_route.ro_rt,
NULL, route_rule_id);
}
}
#endif
#if IPSEC
if (inp->inp_sp != NULL)
ipsec_setsocket(m, so);
#endif
lost = 0;
m->m_pkthdr.pkt_flowsrc = FLOWSRC_INPCB;
m->m_pkthdr.pkt_flowid = inp->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_nextpkt = NULL;
if (inp->inp_last_outifp != NULL &&
!(inp->inp_last_outifp->if_flags & IFF_LOOPBACK)) {
if (tcp_prioritize_acks != 0 && len == 0 &&
(inp->inp_last_outifp->if_eflags &
(IFEF_TXSTART | IFEF_NOACKPRI)) == IFEF_TXSTART) {
if (th->th_flags == TH_ACK &&
tp->snd_una == tp->snd_max &&
tp->t_timer[TCPT_REXMT] == 0)
svc_flags |= PKT_SCF_TCP_ACK;
if (th->th_flags & TH_SYN)
svc_flags |= PKT_SCF_TCP_SYN;
}
set_packet_service_class(m, so, sotc, svc_flags);
} else {
(void) m_set_service_class(m, so_tc2msc(sotc));
}
tp->t_pktlist_sentlen += len;
tp->t_lastchain++;
#if INET6
if (isipv6) {
DTRACE_TCP5(send, struct mbuf *, m, struct inpcb *, inp,
struct ip6 *, ip6, struct tcpcb *, tp, struct tcphdr *,
th);
} else
#endif
{
DTRACE_TCP5(send, struct mbuf *, m, struct inpcb *, inp,
struct ip *, ip, struct tcpcb *, tp, struct tcphdr *, th);
}
if (tp->t_pktlist_head != NULL) {
tp->t_pktlist_tail->m_nextpkt = m;
tp->t_pktlist_tail = m;
} else {
packchain_newlist++;
tp->t_pktlist_head = tp->t_pktlist_tail = m;
}
if ((lro_ackmore) && (!sackoptlen) && (!tp->t_timer[TCPT_PERSIST]) &&
((th->th_flags & TH_ACK) == TH_ACK) && (!len) &&
(tp->t_state == TCPS_ESTABLISHED)) {
mnext = tcp_send_lroacks(tp, m, th);
if (mnext) {
tp->t_pktlist_tail->m_nextpkt = mnext;
if (mnext->m_nextpkt == NULL) {
tp->t_pktlist_tail = mnext;
tp->t_lastchain++;
} else {
struct mbuf *tail, *next;
next = mnext->m_nextpkt;
tail = next->m_nextpkt;
while (tail) {
next = tail;
tail = tail->m_nextpkt;
tp->t_lastchain++;
}
tp->t_pktlist_tail = next;
}
}
}
if (sendalot == 0 || (tp->t_state != TCPS_ESTABLISHED) ||
(tp->snd_cwnd <= (tp->snd_wnd / 8)) ||
(tp->t_flags & TF_ACKNOW) ||
(tp->t_flagsext & TF_FORCE) ||
tp->t_lastchain >= tcp_packet_chaining) {
error = 0;
while (inp->inp_sndinprog_cnt == 0 &&
tp->t_pktlist_head != NULL) {
packetlist = tp->t_pktlist_head;
packchain_listadd = tp->t_lastchain;
packchain_sent++;
lost = tp->t_pktlist_sentlen;
TCP_PKTLIST_CLEAR(tp);
error = tcp_ip_output(so, tp, packetlist,
packchain_listadd, tp_inp_options,
(so_options & SO_DONTROUTE),
(sack_rxmit || (sack_bytes_rxmt != 0)), isipv6);
if (error) {
lost += tp->t_pktlist_sentlen;
break;
} else {
lost = 0;
}
}
if (inp->inp_sndinprog_cnt == 0 &&
(tp->t_flags & TF_CLOSING)) {
tp->t_flags &= ~TF_CLOSING;
(void) tcp_close(tp);
return (0);
}
} else {
error = 0;
packchain_looped++;
tcpstat.tcps_sndtotal++;
goto again;
}
if (error) {
if (!(tp->t_flagsext & TF_FORCE)
|| tp->t_timer[TCPT_PERSIST] == 0) {
if ((flags & TH_SYN) == 0) {
if (sack_rxmit) {
if (SEQ_GT((p->rxmit - lost),
tp->snd_una)) {
p->rxmit -= lost;
} else {
lost = p->rxmit - tp->snd_una;
p->rxmit = tp->snd_una;
}
tp->sackhint.sack_bytes_rexmit -= lost;
} else {
if (SEQ_GT((tp->snd_nxt - lost),
tp->snd_una))
tp->snd_nxt -= lost;
else
tp->snd_nxt = tp->snd_una;
}
}
}
out:
if (tp->t_pktlist_head != NULL)
m_freem_list(tp->t_pktlist_head);
TCP_PKTLIST_CLEAR(tp);
if (error == ENOBUFS) {
if (!tp->t_timer[TCPT_REXMT] &&
!tp->t_timer[TCPT_PERSIST] &&
(SEQ_GT(tp->snd_max, tp->snd_una) ||
so->so_snd.sb_cc > 0))
tp->t_timer[TCPT_REXMT] =
OFFSET_FROM_START(tp, tp->t_rxtcur);
tp->snd_cwnd = tp->t_maxseg;
tp->t_bytes_acked = 0;
tcp_check_timer_state(tp);
KERNEL_DEBUG(DBG_FNC_TCP_OUTPUT | DBG_FUNC_END, 0,0,0,0,0);
tcp_ccdbg_trace(tp, NULL, TCP_CC_OUTPUT_ERROR);
return (0);
}
if (error == EMSGSIZE) {
if (tso)
tp->t_flags &= ~TF_TSO;
tcp_mtudisc(inp, 0);
tcp_check_timer_state(tp);
KERNEL_DEBUG(DBG_FNC_TCP_OUTPUT | DBG_FUNC_END, 0,0,0,0,0);
return 0;
}
if ((error == EHOSTUNREACH || error == ENETDOWN) &&
TCPS_HAVERCVDSYN(tp->t_state) &&
!inp_restricted_send(inp, inp->inp_last_outifp)) {
tp->t_softerror = error;
error = 0;
}
tcp_check_timer_state(tp);
KERNEL_DEBUG(DBG_FNC_TCP_OUTPUT | DBG_FUNC_END, 0,0,0,0,0);
return (error);
}
tcpstat.tcps_sndtotal++;
KERNEL_DEBUG(DBG_FNC_TCP_OUTPUT | DBG_FUNC_END,0,0,0,0,0);
if (sendalot)
goto again;
tcp_check_timer_state(tp);
return (0);
}
static int
tcp_ip_output(struct socket *so, struct tcpcb *tp, struct mbuf *pkt,
int cnt, struct mbuf *opt, int flags, int sack_in_progress, boolean_t isipv6)
{
int error = 0;
boolean_t chain;
boolean_t unlocked = FALSE;
boolean_t ifdenied = FALSE;
struct inpcb *inp = tp->t_inpcb;
struct ip_out_args ipoa =
{ IFSCOPE_NONE, { 0 }, IPOAF_SELECT_SRCIF|IPOAF_BOUND_SRCADDR, 0,
SO_TC_UNSPEC, _NET_SERVICE_TYPE_UNSPEC };
struct route ro;
struct ifnet *outif = NULL;
#if INET6
struct ip6_out_args ip6oa =
{ IFSCOPE_NONE, { 0 }, IP6OAF_SELECT_SRCIF|IP6OAF_BOUND_SRCADDR, 0,
SO_TC_UNSPEC, _NET_SERVICE_TYPE_UNSPEC };
struct route_in6 ro6;
struct flowadv *adv =
(isipv6 ? &ip6oa.ip6oa_flowadv : &ipoa.ipoa_flowadv);
#else
struct flowadv *adv = &ipoa.ipoa_flowadv;
#endif
if (inp->inp_flags & INP_BOUND_IF) {
#if INET6
if (isipv6) {
ip6oa.ip6oa_boundif = inp->inp_boundifp->if_index;
ip6oa.ip6oa_flags |= IP6OAF_BOUND_IF;
} else
#endif
{
ipoa.ipoa_boundif = inp->inp_boundifp->if_index;
ipoa.ipoa_flags |= IPOAF_BOUND_IF;
}
}
if (INP_NO_CELLULAR(inp)) {
#if INET6
if (isipv6)
ip6oa.ip6oa_flags |= IP6OAF_NO_CELLULAR;
else
#endif
ipoa.ipoa_flags |= IPOAF_NO_CELLULAR;
}
if (INP_NO_EXPENSIVE(inp)) {
#if INET6
if (isipv6)
ip6oa.ip6oa_flags |= IP6OAF_NO_EXPENSIVE;
else
#endif
ipoa.ipoa_flags |= IPOAF_NO_EXPENSIVE;
}
if (INP_AWDL_UNRESTRICTED(inp)) {
#if INET6
if (isipv6)
ip6oa.ip6oa_flags |= IP6OAF_AWDL_UNRESTRICTED;
else
#endif
ipoa.ipoa_flags |= IPOAF_AWDL_UNRESTRICTED;
}
#if INET6
if (INP_INTCOPROC_ALLOWED(inp) && isipv6) {
ip6oa.ip6oa_flags |= IP6OAF_INTCOPROC_ALLOWED;
}
if (isipv6) {
ip6oa.ip6oa_sotc = so->so_traffic_class;
ip6oa.ip6oa_netsvctype = so->so_netsvctype;
} else
#endif
{
ipoa.ipoa_sotc = so->so_traffic_class;
ipoa.ipoa_netsvctype = so->so_netsvctype;
}
if ((so->so_flags1 & SOF1_QOSMARKING_ALLOWED)) {
#if INET6
if (isipv6)
ip6oa.ip6oa_flags |= IP6OAF_QOSMARKING_ALLOWED;
else
#endif
ipoa.ipoa_flags |= IPOAF_QOSMARKING_ALLOWED;
}
#if INET6
if (isipv6)
flags |= IPV6_OUTARGS;
else
#endif
flags |= IP_OUTARGS;
#if INET6
if (isipv6)
in6p_route_copyout(inp, &ro6);
else
#endif
inp_route_copyout(inp, &ro);
tp->last_ack_sent = tp->rcv_nxt;
tp->t_flags &= ~(TF_ACKNOW | TF_DELACK);
tp->t_timer[TCPT_DELACK] = 0;
tp->t_unacksegs = 0;
inp->inp_sndinprog_cnt++;
if (tcp_output_unlocked && !so->so_upcallusecount &&
(tp->t_state == TCPS_ESTABLISHED) && (sack_in_progress == 0) &&
!IN_FASTRECOVERY(tp) && !(so->so_flags & SOF_MP_SUBFLOW)) {
unlocked = TRUE;
socket_unlock(so, 0);
}
chain = tcp_packet_chaining > 1
#if IPSEC
&& ipsec_bypass
#endif
#if IPFIREWALL
&& (fw_enable == 0 || fw_bypass)
#endif
;
while (pkt != NULL) {
struct mbuf *npkt = pkt->m_nextpkt;
if (!chain) {
pkt->m_nextpkt = NULL;
cnt = 0;
}
#if INET6
if (isipv6) {
error = ip6_output_list(pkt, cnt,
inp->in6p_outputopts, &ro6, flags, NULL, NULL,
&ip6oa);
ifdenied = (ip6oa.ip6oa_retflags & IP6OARF_IFDENIED);
} else {
#endif
error = ip_output_list(pkt, cnt, opt, &ro, flags, NULL,
&ipoa);
ifdenied = (ipoa.ipoa_retflags & IPOARF_IFDENIED);
}
if (chain || error) {
if (!chain)
m_freem_list(npkt);
break;
}
pkt = npkt;
}
if (unlocked)
socket_lock(so, 0);
if (((adv->code == FADV_FLOW_CONTROLLED && !IN_FASTRECOVERY(tp)) ||
adv->code == FADV_SUSPENDED) &&
!(tp->t_flags & TF_CLOSING) &&
tp->t_state == TCPS_ESTABLISHED &&
SEQ_GT(tp->snd_max, tp->snd_una)) {
int rc;
rc = inp_set_fc_state(inp, adv->code);
if (rc == 1)
tcp_ccdbg_trace(tp, NULL,
((adv->code == FADV_FLOW_CONTROLLED) ?
TCP_CC_FLOW_CONTROL : TCP_CC_SUSPEND));
}
if (adv->code == FADV_SUSPENDED)
error = ENOBUFS;
VERIFY(inp->inp_sndinprog_cnt > 0);
if ( --inp->inp_sndinprog_cnt == 0)
inp->inp_flags &= ~(INP_FC_FEEDBACK);
#if INET6
if (isipv6) {
if (ro6.ro_rt != NULL)
outif = ro6.ro_rt->rt_ifp;
} else
#endif
if (ro.ro_rt != NULL)
outif = ro.ro_rt->rt_ifp;
if (outif != NULL && outif != inp->inp_last_outifp &&
so->so_snd.sb_cc > 0) {
if (so->so_snd.sb_flags & SB_SNDBYTE_CNT) {
inp_decr_sndbytes_total(so, so->so_snd.sb_cc);
inp_decr_sndbytes_allunsent(so, tp->snd_una);
so->so_snd.sb_flags &= ~SB_SNDBYTE_CNT;
}
inp->inp_last_outifp = outif;
}
if (error != 0 && ifdenied &&
(INP_NO_CELLULAR(inp) || INP_NO_EXPENSIVE(inp)))
soevent(so,
(SO_FILT_HINT_LOCKED|SO_FILT_HINT_IFDENIED));
#if INET6
if (isipv6)
in6p_route_copyin(inp, &ro6);
else
#endif
inp_route_copyin(inp, &ro);
if (tp->t_state < TCPS_ESTABLISHED && tp->t_rxtshift == 0 &&
tp->t_inpcb->inp_route.ro_rt != NULL) {
tcp_getrt_rtt(tp, tp->t_inpcb->in6p_route.ro_rt);
tp->t_timer[TCPT_REXMT] = OFFSET_FROM_START(tp, tp->t_rxtcur);
}
return (error);
}
int tcptv_persmin_val = TCPTV_PERSMIN;
void
tcp_setpersist(struct tcpcb *tp)
{
int t = ((tp->t_srtt >> 2) + tp->t_rttvar) >> 1;
if ((tp->t_persist_timeout != 0) &&
(tp->t_timer[TCPT_PERSIST] == 0) &&
(tp->t_persist_stop == 0)) {
tp->t_persist_stop = tcp_now + tp->t_persist_timeout;
}
TCPT_RANGESET(tp->t_timer[TCPT_PERSIST],
t * tcp_backoff[tp->t_rxtshift],
tcptv_persmin_val, TCPTV_PERSMAX, 0);
tp->t_timer[TCPT_PERSIST] = OFFSET_FROM_START(tp, tp->t_timer[TCPT_PERSIST]);
if (tp->t_rxtshift < TCP_MAXRXTSHIFT)
tp->t_rxtshift++;
}
static struct mbuf*
tcp_send_lroacks(struct tcpcb *tp, struct mbuf *m, struct tcphdr *th)
{
struct mbuf *mnext = NULL, *ack_chain = NULL, *tail = NULL;
int count = 0;
tcp_seq org_ack = ntohl(th->th_ack);
tcp_seq prev_ack = 0;
int tack_offset = 28;
int twin_offset = 34;
int ack_size = (tp->t_flags & TF_STRETCHACK) ?
(maxseg_unacked * tp->t_maxseg) : (tp->t_maxseg << 1);
int segs_acked = (tp->t_flags & TF_STRETCHACK) ? maxseg_unacked : 2;
struct mbuf *prev_ack_pkt = NULL;
struct socket *so = tp->t_inpcb->inp_socket;
unsigned short winsz = ntohs(th->th_win);
unsigned int scaled_win = winsz<<tp->rcv_scale;
tcp_seq win_rtedge = org_ack + scaled_win;
count = tp->t_lropktlen/tp->t_maxseg;
prev_ack = (org_ack - tp->t_lropktlen) + ack_size;
if (prev_ack < org_ack) {
ack_chain = m_dup(m, M_DONTWAIT);
if (ack_chain) {
th->th_ack = htonl(prev_ack);
scaled_win = win_rtedge - prev_ack;
if (scaled_win > (int32_t)(TCP_MAXWIN << tp->rcv_scale))
scaled_win = (int32_t)(TCP_MAXWIN << tp->rcv_scale);
th->th_win = htons(scaled_win>>tp->rcv_scale);
if (lrodebug == 5) {
printf("%s: win = %d winsz = %d sc = %d"
" lro_len %d %d\n",
__func__, scaled_win>>tp->rcv_scale, winsz,
tp->rcv_scale, tp->t_lropktlen, count);
}
tail = ack_chain;
count -= segs_acked;
count = (count <= segs_acked) ? 0 : count - segs_acked;
tcpstat.tcps_sndacks++;
so_tc_update_stats(m, so, m_get_service_class(m));
} else {
return NULL;
}
}
else {
tp->t_lropktlen = 0;
return NULL;
}
prev_ack_pkt = ack_chain;
while (count > 0) {
if ((prev_ack + ack_size) < org_ack) {
prev_ack += ack_size;
} else {
prev_ack = org_ack;
}
mnext = m_dup(prev_ack_pkt, M_DONTWAIT);
if (mnext) {
scaled_win = win_rtedge - prev_ack;
if (scaled_win > (int32_t)(TCP_MAXWIN << tp->rcv_scale))
scaled_win = (int32_t)(TCP_MAXWIN << tp->rcv_scale);
winsz = htons(scaled_win>>tp->rcv_scale);
if (lrodebug == 5) {
printf("%s: winsz = %d ack %x count %d\n",
__func__, scaled_win>>tp->rcv_scale,
prev_ack, count);
}
bcopy(&winsz, mtod(prev_ack_pkt, caddr_t) + twin_offset, 2);
HTONL(prev_ack);
bcopy(&prev_ack, mtod(prev_ack_pkt, caddr_t) + tack_offset, 4);
NTOHL(prev_ack);
tail->m_nextpkt = mnext;
tail = mnext;
count -= segs_acked;
tcpstat.tcps_sndacks++;
so_tc_update_stats(m, so, m_get_service_class(m));
} else {
if (lrodebug == 5) {
printf("%s: failed to alloc mbuf.\n", __func__);
}
break;
}
prev_ack_pkt = mnext;
}
tp->t_lropktlen = 0;
return ack_chain;
}
static int
tcp_recv_throttle (struct tcpcb *tp)
{
uint32_t base_rtt, newsize;
struct sockbuf *sbrcv = &tp->t_inpcb->inp_socket->so_rcv;
if (tcp_use_rtt_recvbg == 1 &&
TSTMP_SUPPORTED(tp)) {
if (tp->t_flagsext & TF_RECV_THROTTLE)
return (1);
base_rtt = get_base_rtt(tp);
if (base_rtt != 0 && tp->t_rttcur != 0) {
if (tp->t_rttcur > (base_rtt + target_qdelay)) {
tp->t_flagsext |= TF_RECV_THROTTLE;
if (tp->t_recv_throttle_ts == 0)
tp->t_recv_throttle_ts = tcp_now;
if (sbrcv->sb_idealsize >
tcp_recv_throttle_minwin) {
newsize = sbrcv->sb_idealsize >> 1;
newsize =
max(newsize,
tcp_recv_throttle_minwin);
sbrcv->sb_idealsize = newsize;
}
return (1);
} else {
return (0);
}
}
}
if (tp->acc_iaj > tcp_acc_iaj_react_limit)
return (1);
return (0);
}