#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/proc.h>
#include <sys/protosw.h>
#include <sys/socket.h>
#include <sys/socketvar.h>
#include <sys/syslog.h>
#include <sys/mcache.h>
#if !CONFIG_EMBEDDED
#include <sys/kasl.h>
#endif
#include <sys/kauth.h>
#include <kern/cpu_number.h>
#include <machine/endian.h>
#include <net/if.h>
#include <net/if_types.h>
#include <net/route.h>
#include <net/ntstat.h>
#include <net/content_filter.h>
#include <net/dlil.h>
#include <net/multi_layer_pkt_log.h>
#include <netinet/in.h>
#include <netinet/in_systm.h>
#include <netinet/ip.h>
#include <netinet/ip_icmp.h>
#include <netinet/in_var.h>
#include <netinet/icmp_var.h>
#include <netinet/in_pcb.h>
#include <netinet/ip_var.h>
#include <mach/sdt.h>
#if INET6
#include <netinet/ip6.h>
#include <netinet/icmp6.h>
#include <netinet6/nd6.h>
#include <netinet6/ip6_var.h>
#include <netinet6/in6_pcb.h>
#endif
#include <netinet/tcp.h>
#include <netinet/tcp_cache.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 <dev/random/randomdev.h>
#include <kern/zalloc.h>
#if INET6
#include <netinet6/tcp6_var.h>
#endif
#include <netinet/tcpip.h>
#if TCPDEBUG
#include <netinet/tcp_debug.h>
u_char tcp_saveipgen[40];
struct tcphdr tcp_savetcp;
#endif
#include <netinet/tcp_log.h>
#if IPSEC
#include <netinet6/ipsec.h>
#if INET6
#include <netinet6/ipsec6.h>
#endif
#include <netkey/key.h>
#endif
#if CONFIG_MACF_NET || CONFIG_MACF_SOCKET
#include <security/mac_framework.h>
#endif
#include <sys/kdebug.h>
#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, 0)
#define DBG_LAYER_END NETDBG_CODE(DBG_NETTCP, 2)
#define DBG_FNC_TCP_INPUT NETDBG_CODE(DBG_NETTCP, (3 << 8))
#define DBG_FNC_TCP_NEWCONN NETDBG_CODE(DBG_NETTCP, (7 << 8))
#define TCP_RTT_HISTORY_EXPIRE_TIME (60 * TCP_RETRANSHZ)
#define TCP_RECV_THROTTLE_WIN (5 * TCP_RETRANSHZ)
#define TCP_STRETCHACK_ENABLE_PKTCNT 2000
struct tcpstat tcpstat;
static int log_in_vain = 0;
SYSCTL_INT(_net_inet_tcp, OID_AUTO, log_in_vain,
CTLFLAG_RW | CTLFLAG_LOCKED, &log_in_vain, 0,
"Log all incoming TCP connections");
static int blackhole = 0;
SYSCTL_INT(_net_inet_tcp, OID_AUTO, blackhole,
CTLFLAG_RW | CTLFLAG_LOCKED, &blackhole, 0,
"Do not send RST when dropping refused connections");
SYSCTL_SKMEM_TCP_INT(OID_AUTO, delayed_ack,
CTLFLAG_RW | CTLFLAG_LOCKED, int, tcp_delack_enabled, 3,
"Delay ACK to try and piggyback it onto a data packet");
SYSCTL_SKMEM_TCP_INT(OID_AUTO, tcp_lq_overflow,
CTLFLAG_RW | CTLFLAG_LOCKED, int, tcp_lq_overflow, 1,
"Listen Queue Overflow");
SYSCTL_SKMEM_TCP_INT(OID_AUTO, recvbg, CTLFLAG_RW | CTLFLAG_LOCKED,
int, tcp_recv_bg, 0, "Receive background");
#if TCP_DROP_SYNFIN
SYSCTL_SKMEM_TCP_INT(OID_AUTO, drop_synfin,
CTLFLAG_RW | CTLFLAG_LOCKED, static int, drop_synfin, 1,
"Drop TCP packets with SYN+FIN set");
#endif
SYSCTL_NODE(_net_inet_tcp, OID_AUTO, reass, CTLFLAG_RW | CTLFLAG_LOCKED, 0,
"TCP Segment Reassembly Queue");
static int tcp_reass_overflows = 0;
SYSCTL_INT(_net_inet_tcp_reass, OID_AUTO, overflows,
CTLFLAG_RD | CTLFLAG_LOCKED, &tcp_reass_overflows, 0,
"Global number of TCP Segment Reassembly Queue Overflows");
SYSCTL_SKMEM_TCP_INT(OID_AUTO, slowlink_wsize, CTLFLAG_RW | CTLFLAG_LOCKED,
__private_extern__ int, slowlink_wsize, 8192,
"Maximum advertised window size for slowlink");
SYSCTL_SKMEM_TCP_INT(OID_AUTO, maxseg_unacked,
CTLFLAG_RW | CTLFLAG_LOCKED, int, maxseg_unacked, 8,
"Maximum number of outstanding segments left unacked");
SYSCTL_SKMEM_TCP_INT(OID_AUTO, rfc3465, CTLFLAG_RW | CTLFLAG_LOCKED,
int, tcp_do_rfc3465, 1, "");
SYSCTL_SKMEM_TCP_INT(OID_AUTO, rfc3465_lim2,
CTLFLAG_RW | CTLFLAG_LOCKED, int, tcp_do_rfc3465_lim2, 1,
"Appropriate bytes counting w/ L=2*SMSS");
int rtt_samples_per_slot = 20;
int tcp_acc_iaj_high_thresh = ACC_IAJ_HIGH_THRESH;
u_int32_t tcp_autorcvbuf_inc_shift = 3;
SYSCTL_SKMEM_TCP_INT(OID_AUTO, recv_allowed_iaj,
CTLFLAG_RW | CTLFLAG_LOCKED, int, tcp_allowed_iaj, ALLOWED_IAJ,
"Allowed inter-packet arrival jiter");
#if (DEVELOPMENT || DEBUG)
SYSCTL_INT(_net_inet_tcp, OID_AUTO, acc_iaj_high_thresh,
CTLFLAG_RW | CTLFLAG_LOCKED, &tcp_acc_iaj_high_thresh, 0,
"Used in calculating maximum accumulated IAJ");
SYSCTL_INT(_net_inet_tcp, OID_AUTO, autorcvbufincshift,
CTLFLAG_RW | CTLFLAG_LOCKED, &tcp_autorcvbuf_inc_shift, 0,
"Shift for increment in receive socket buffer size");
#endif
SYSCTL_SKMEM_TCP_INT(OID_AUTO, doautorcvbuf,
CTLFLAG_RW | CTLFLAG_LOCKED, u_int32_t, tcp_do_autorcvbuf, 1,
"Enable automatic socket buffer tuning");
SYSCTL_SKMEM_TCP_INT(OID_AUTO, autotunereorder,
CTLFLAG_RW | CTLFLAG_LOCKED, u_int32_t, tcp_autotune_reorder, 1,
"Enable automatic socket buffer tuning even when reordering is present");
SYSCTL_SKMEM_TCP_INT(OID_AUTO, autorcvbufmax,
CTLFLAG_RW | CTLFLAG_LOCKED, u_int32_t, tcp_autorcvbuf_max, 512 * 1024,
"Maximum receive socket buffer size");
#if CONFIG_EMBEDDED
int sw_lro = 1;
#else
int sw_lro = 0;
#endif
SYSCTL_INT(_net_inet_tcp, OID_AUTO, lro, CTLFLAG_RW | CTLFLAG_LOCKED,
&sw_lro, 0, "Used to coalesce TCP packets");
int lrodebug = 0;
SYSCTL_INT(_net_inet_tcp, OID_AUTO, lrodbg,
CTLFLAG_RW | CTLFLAG_LOCKED, &lrodebug, 0,
"Used to debug SW LRO");
int lro_start = 4;
SYSCTL_INT(_net_inet_tcp, OID_AUTO, lro_startcnt,
CTLFLAG_RW | CTLFLAG_LOCKED, &lro_start, 0,
"Segments for starting LRO computed as power of 2");
int limited_txmt = 1;
int early_rexmt = 1;
int sack_ackadv = 1;
int tcp_dsack_enable = 1;
#if (DEVELOPMENT || DEBUG)
SYSCTL_INT(_net_inet_tcp, OID_AUTO, limited_transmit,
CTLFLAG_RW | CTLFLAG_LOCKED, &limited_txmt, 0,
"Enable limited transmit");
SYSCTL_INT(_net_inet_tcp, OID_AUTO, early_rexmt,
CTLFLAG_RW | CTLFLAG_LOCKED, &early_rexmt, 0,
"Enable Early Retransmit");
SYSCTL_INT(_net_inet_tcp, OID_AUTO, sack_ackadv,
CTLFLAG_RW | CTLFLAG_LOCKED, &sack_ackadv, 0,
"Use SACK with cumulative ack advancement as a dupack");
SYSCTL_INT(_net_inet_tcp, OID_AUTO, dsack_enable,
CTLFLAG_RW | CTLFLAG_LOCKED, &tcp_dsack_enable, 0,
"use DSACK TCP option to report duplicate segments");
#endif
int tcp_disable_access_to_stats = 1;
SYSCTL_INT(_net_inet_tcp, OID_AUTO, disable_access_to_stats,
CTLFLAG_RW | CTLFLAG_LOCKED, &tcp_disable_access_to_stats, 0,
"Disable access to tcpstat");
SYSCTL_SKMEM_TCP_INT(OID_AUTO, challengeack_limit,
CTLFLAG_RW | CTLFLAG_LOCKED, uint32_t, tcp_challengeack_limit, 10,
"Maximum number of challenge ACKs per connection per second");
SYSCTL_SKMEM_TCP_INT(OID_AUTO, do_rfc5961,
CTLFLAG_RW | CTLFLAG_LOCKED, static int, tcp_do_rfc5961, 1,
"Enable/Disable full RFC 5961 compliance");
extern int tcp_acc_iaj_high;
extern int tcp_acc_iaj_react_limit;
int tcprexmtthresh = 3;
u_int32_t tcp_now;
struct timeval tcp_uptime;
lck_spin_t *tcp_uptime_lock;
struct inpcbhead tcb;
#define tcb6 tcb
struct inpcbinfo tcbinfo;
static void tcp_dooptions(struct tcpcb *, u_char *, int, struct tcphdr *,
struct tcpopt *);
static void tcp_finalize_options(struct tcpcb *, struct tcpopt *, unsigned int);
static void tcp_pulloutofband(struct socket *,
struct tcphdr *, struct mbuf *, int);
static void tcp_xmit_timer(struct tcpcb *, int, u_int32_t, tcp_seq);
static inline unsigned int tcp_maxmtu(struct rtentry *);
static inline int tcp_stretch_ack_enable(struct tcpcb *tp, int thflags);
static inline void tcp_adaptive_rwtimo_check(struct tcpcb *, int);
#if TRAFFIC_MGT
static inline void update_iaj_state(struct tcpcb *tp, uint32_t tlen,
int reset_size);
void compute_iaj(struct tcpcb *tp, int nlropkts, int lro_delay_factor);
static void compute_iaj_meat(struct tcpcb *tp, uint32_t cur_iaj);
#endif
#if INET6
static inline unsigned int tcp_maxmtu6(struct rtentry *);
#endif
unsigned int get_maxmtu(struct rtentry *);
static void tcp_sbrcv_grow(struct tcpcb *tp, struct sockbuf *sb,
struct tcpopt *to, u_int32_t tlen, u_int32_t rcvbuf_max);
void tcp_sbrcv_trim(struct tcpcb *tp, struct sockbuf *sb);
static void tcp_sbsnd_trim(struct sockbuf *sbsnd);
static inline void tcp_sbrcv_tstmp_check(struct tcpcb *tp);
static inline void tcp_sbrcv_reserve(struct tcpcb *tp, struct sockbuf *sb,
u_int32_t newsize, u_int32_t idealsize, u_int32_t rcvbuf_max);
static void tcp_bad_rexmt_restore_state(struct tcpcb *tp, struct tcphdr *th);
static void tcp_compute_rtt(struct tcpcb *tp, struct tcpopt *to,
struct tcphdr *th);
static void tcp_early_rexmt_check(struct tcpcb *tp, struct tcphdr *th);
static void tcp_bad_rexmt_check(struct tcpcb *tp, struct tcphdr *th,
struct tcpopt *to);
#define TCPTV_RCVNOTS_QUANTUM 100
#define TCP_RCVNOTS_BYTELEVEL 204800
#define TCP_EARLY_REXMT_WIN (60 * TCP_RETRANSHZ)
#define TCP_EARLY_REXMT_LIMIT 10
extern void ipfwsyslog( int level, const char *format, ...);
extern int fw_verbose;
#if IPFIREWALL
extern void ipfw_stealth_stats_incr_tcp(void);
#define log_in_vain_log( a ) { \
if ( (log_in_vain == 3 ) && (fw_verbose == 2)) { \
ipfwsyslog a ; \
} else if ( (log_in_vain == 4 ) && (fw_verbose == 2)) { \
ipfw_stealth_stats_incr_tcp(); \
} \
else log a ; \
}
#else
#define log_in_vain_log( a ) { log a; }
#endif
int tcp_rcvunackwin = TCPTV_UNACKWIN;
int tcp_maxrcvidle = TCPTV_MAXRCVIDLE;
SYSCTL_SKMEM_TCP_INT(OID_AUTO, rcvsspktcnt, CTLFLAG_RW | CTLFLAG_LOCKED,
int, tcp_rcvsspktcnt, TCP_RCV_SS_PKTCOUNT, "packets to be seen before receiver stretches acks");
#define DELAY_ACK(tp, th) \
(CC_ALGO(tp)->delay_ack != NULL && CC_ALGO(tp)->delay_ack(tp, th))
static int tcp_dropdropablreq(struct socket *head);
static void tcp_newreno_partial_ack(struct tcpcb *tp, struct tcphdr *th);
static void update_base_rtt(struct tcpcb *tp, uint32_t rtt);
void tcp_set_background_cc(struct socket *so);
void tcp_set_foreground_cc(struct socket *so);
static void tcp_set_new_cc(struct socket *so, uint16_t cc_index);
static void tcp_bwmeas_check(struct tcpcb *tp);
#if TRAFFIC_MGT
void
reset_acc_iaj(struct tcpcb *tp)
{
tp->acc_iaj = 0;
CLEAR_IAJ_STATE(tp);
}
static inline void
update_iaj_state(struct tcpcb *tp, uint32_t size, int rst_size)
{
if (rst_size > 0) {
tp->iaj_size = 0;
}
if (tp->iaj_size == 0 || size >= tp->iaj_size) {
tp->iaj_size = size;
tp->iaj_rcv_ts = tcp_now;
tp->iaj_small_pkt = 0;
}
}
static inline int
isqrt(unsigned int val)
{
unsigned int sqrt_cache[11] = {0, 1, 4, 9, 16, 25, 36, 49, 64, 81, 100};
unsigned int temp, g = 0, b = 0x8000, bshft = 15;
if (val <= 100) {
for (g = 0; g <= 10; ++g) {
if (sqrt_cache[g] > val) {
g--;
break;
} else if (sqrt_cache[g] == val) {
break;
}
}
} else {
do {
temp = (((g << 1) + b) << (bshft--));
if (val >= temp) {
g += b;
val -= temp;
}
b >>= 1;
} while (b > 0 && val > 0);
}
return g;
}
void
compute_iaj(struct tcpcb *tp, int nlropkts, int lro_delay_factor)
{
uint32_t cur_iaj = tcp_now - tp->iaj_rcv_ts;
uint32_t timediff = 0;
if (cur_iaj >= lro_delay_factor) {
cur_iaj = cur_iaj - lro_delay_factor;
}
compute_iaj_meat(tp, cur_iaj);
if (nlropkts <= 1) {
return;
}
nlropkts--;
timediff = lro_delay_factor / nlropkts;
while (nlropkts > 0) {
compute_iaj_meat(tp, timediff);
nlropkts--;
}
}
static
void
compute_iaj_meat(struct tcpcb *tp, uint32_t cur_iaj)
{
#define MAX_ACC_IAJ (tcp_acc_iaj_high_thresh + tcp_acc_iaj_react_limit)
#define IAJ_DIV_SHIFT 4
#define IAJ_ROUNDUP_CONST (1 << (IAJ_DIV_SHIFT - 1))
uint32_t allowed_iaj, acc_iaj = 0;
uint32_t mean, temp;
int32_t cur_iaj_dev;
cur_iaj_dev = (cur_iaj - tp->avg_iaj);
allowed_iaj = tp->avg_iaj + tp->std_dev_iaj;
if (allowed_iaj < tcp_allowed_iaj) {
allowed_iaj = tcp_allowed_iaj;
}
if (tp->iaj_pktcnt > IAJ_IGNORE_PKTCNT) {
if (cur_iaj <= allowed_iaj) {
if (tp->acc_iaj >= 2) {
acc_iaj = tp->acc_iaj - 2;
} else {
acc_iaj = 0;
}
} else {
acc_iaj = tp->acc_iaj + (cur_iaj - allowed_iaj);
}
if (acc_iaj > MAX_ACC_IAJ) {
acc_iaj = MAX_ACC_IAJ;
}
tp->acc_iaj = acc_iaj;
}
tp->avg_iaj = (((tp->avg_iaj << IAJ_DIV_SHIFT) - tp->avg_iaj)
+ cur_iaj + IAJ_ROUNDUP_CONST) >> IAJ_DIV_SHIFT;
temp = tp->std_dev_iaj * tp->std_dev_iaj;
mean = (((temp << IAJ_DIV_SHIFT) - temp)
+ (cur_iaj_dev * cur_iaj_dev)
+ IAJ_ROUNDUP_CONST) >> IAJ_DIV_SHIFT;
tp->std_dev_iaj = isqrt(mean);
DTRACE_TCP3(iaj, struct tcpcb *, tp, uint32_t, cur_iaj,
uint32_t, allowed_iaj);
return;
}
#endif
static boolean_t
tcp_is_ack_ratelimited(struct tcpcb *tp)
{
boolean_t ret = TRUE;
uint32_t now = tcp_now;
int32_t diff = 0;
diff = timer_diff(now, 0, tp->t_challengeack_last, 0);
if (tp->t_challengeack_last == 0 || diff >= 1000) {
tp->t_challengeack_last = now;
tp->t_challengeack_count = 0;
ret = FALSE;
} else if (tp->t_challengeack_count < tcp_challengeack_limit) {
ret = FALSE;
}
if (ret == FALSE && (tp->t_challengeack_count + 1 > 0)) {
tp->t_challengeack_count++;
}
return ret;
}
static void
tcp_bwmeas_check(struct tcpcb *tp)
{
int32_t bw_meas_bytes;
uint32_t bw, bytes, elapsed_time;
if (SEQ_LEQ(tp->snd_una, tp->t_bwmeas->bw_start)) {
return;
}
bw_meas_bytes = tp->snd_una - tp->t_bwmeas->bw_start;
if ((tp->t_flagsext & TF_BWMEAS_INPROGRESS) &&
bw_meas_bytes >= (int32_t)(tp->t_bwmeas->bw_size)) {
bytes = bw_meas_bytes;
elapsed_time = tcp_now - tp->t_bwmeas->bw_ts;
if (elapsed_time > 0) {
bw = bytes / elapsed_time;
if (bw > 0) {
if (tp->t_bwmeas->bw_sndbw > 0) {
tp->t_bwmeas->bw_sndbw =
(((tp->t_bwmeas->bw_sndbw << 3)
- tp->t_bwmeas->bw_sndbw)
+ bw) >> 3;
} else {
tp->t_bwmeas->bw_sndbw = bw;
}
if (tp->t_bwmeas->bw_sndbw_max == 0) {
tp->t_bwmeas->bw_sndbw_max =
tp->t_bwmeas->bw_sndbw;
} else {
tp->t_bwmeas->bw_sndbw_max =
max(tp->t_bwmeas->bw_sndbw,
tp->t_bwmeas->bw_sndbw_max);
}
}
}
tp->t_flagsext &= ~(TF_BWMEAS_INPROGRESS);
}
}
static int
tcp_reass(struct tcpcb *tp, struct tcphdr *th, int *tlenp, struct mbuf *m,
struct ifnet *ifp, int *dowakeup)
{
struct tseg_qent *q;
struct tseg_qent *p = NULL;
struct tseg_qent *nq;
struct tseg_qent *te = NULL;
struct inpcb *inp = tp->t_inpcb;
struct socket *so = inp->inp_socket;
int flags = 0;
struct mbuf *oodata = NULL;
int copy_oodata = 0;
u_int16_t qlimit;
boolean_t cell = IFNET_IS_CELLULAR(ifp);
boolean_t wifi = (!cell && IFNET_IS_WIFI(ifp));
boolean_t wired = (!wifi && IFNET_IS_WIRED(ifp));
boolean_t dsack_set = FALSE;
if (th == NULL) {
goto present;
}
tcp_reset_stretch_ack(tp);
#if TRAFFIC_MGT
if (tp->acc_iaj > 0) {
reset_acc_iaj(tp);
}
#endif
qlimit = min(max(100, so->so_rcv.sb_hiwat >> 10),
(TCP_AUTORCVBUF_MAX(ifp) >> 10));
if (th->th_seq != tp->rcv_nxt &&
(tp->t_reassqlen + 1) >= qlimit) {
tcp_reass_overflows++;
tcpstat.tcps_rcvmemdrop++;
m_freem(m);
*tlenp = 0;
return 0;
}
te = (struct tseg_qent *) zalloc(tcp_reass_zone);
if (te == NULL) {
tcpstat.tcps_rcvmemdrop++;
m_freem(m);
return 0;
}
tp->t_reassqlen++;
LIST_FOREACH(q, &tp->t_segq, tqe_q) {
if (SEQ_GT(q->tqe_th->th_seq, th->th_seq)) {
break;
}
p = q;
}
if (p != NULL) {
int i;
i = p->tqe_th->th_seq + p->tqe_len - th->th_seq;
if (i > 0) {
if (TCP_DSACK_ENABLED(tp) && i > 1) {
tp->t_dsack_lseq = th->th_seq;
tp->t_dsack_rseq = th->th_seq +
min(i, *tlenp);
dsack_set = TRUE;
}
if (i >= *tlenp) {
tcpstat.tcps_rcvduppack++;
tcpstat.tcps_rcvdupbyte += *tlenp;
if (nstat_collect) {
nstat_route_rx(inp->inp_route.ro_rt,
1, *tlenp,
NSTAT_RX_FLAG_DUPLICATE);
INP_ADD_STAT(inp, cell, wifi, wired,
rxpackets, 1);
INP_ADD_STAT(inp, cell, wifi, wired,
rxbytes, *tlenp);
tp->t_stat.rxduplicatebytes += *tlenp;
inp_set_activity_bitmap(inp);
}
m_freem(m);
zfree(tcp_reass_zone, te);
te = NULL;
tp->t_reassqlen--;
goto present;
}
m_adj(m, i);
*tlenp -= i;
th->th_seq += i;
}
}
tp->t_rcvoopack++;
tcpstat.tcps_rcvoopack++;
tcpstat.tcps_rcvoobyte += *tlenp;
if (nstat_collect) {
nstat_route_rx(inp->inp_route.ro_rt, 1, *tlenp,
NSTAT_RX_FLAG_OUT_OF_ORDER);
INP_ADD_STAT(inp, cell, wifi, wired, rxpackets, 1);
INP_ADD_STAT(inp, cell, wifi, wired, rxbytes, *tlenp);
tp->t_stat.rxoutoforderbytes += *tlenp;
inp_set_activity_bitmap(inp);
}
while (q) {
int i = (th->th_seq + *tlenp) - q->tqe_th->th_seq;
if (i <= 0) {
break;
}
if (TCP_DSACK_ENABLED(tp) && i > 1 && !dsack_set) {
if (tp->t_dsack_lseq == 0) {
tp->t_dsack_lseq = q->tqe_th->th_seq;
tp->t_dsack_rseq =
tp->t_dsack_lseq + min(i, q->tqe_len);
} else {
tp->t_dsack_rseq =
tp->t_dsack_rseq + min(i, q->tqe_len);
}
}
if (i < q->tqe_len) {
q->tqe_th->th_seq += i;
q->tqe_len -= i;
m_adj(q->tqe_m, i);
break;
}
nq = LIST_NEXT(q, tqe_q);
LIST_REMOVE(q, tqe_q);
m_freem(q->tqe_m);
zfree(tcp_reass_zone, q);
tp->t_reassqlen--;
q = nq;
}
te->tqe_m = m;
te->tqe_th = th;
te->tqe_len = *tlenp;
if (p == NULL) {
LIST_INSERT_HEAD(&tp->t_segq, te, tqe_q);
} else {
LIST_INSERT_AFTER(p, te, tqe_q);
}
if (!(so->so_state & SS_CANTRCVMORE)) {
copy_oodata = 1;
}
present:
if (!TCPS_HAVEESTABLISHED(tp->t_state)) {
return 0;
}
q = LIST_FIRST(&tp->t_segq);
if (!q || q->tqe_th->th_seq != tp->rcv_nxt) {
if (tp->t_flagsext & TF_LRO_OFFLOADED) {
tcp_lro_remove_state(inp->inp_laddr, inp->inp_faddr,
th->th_dport, th->th_sport);
tp->t_flagsext &= ~TF_LRO_OFFLOADED;
}
if (q && (so->so_flags & SOF_ENABLE_MSGS)) {
goto msg_unordered_delivery;
}
return 0;
}
if (tp->t_flagsext & TF_REASS_INPROG) {
return 0;
}
tp->t_flagsext |= TF_REASS_INPROG;
tcpstat.tcps_recovered_pkts++;
do {
tp->rcv_nxt += q->tqe_len;
flags = q->tqe_th->th_flags & TH_FIN;
LIST_REMOVE(q, tqe_q);
if (so->so_state & SS_CANTRCVMORE) {
m_freem(q->tqe_m);
} else {
struct tcphdr saved_tcphdr = {};
so_recv_data_stat(so, q->tqe_m, 0);
if (so->so_flags & SOF_ENABLE_MSGS) {
if (q == te) {
copy_oodata = 0;
}
}
memcpy(&saved_tcphdr, th, sizeof(struct tcphdr));
if (sbappendstream_rcvdemux(so, q->tqe_m,
q->tqe_th->th_seq - (tp->irs + 1), 0)) {
*dowakeup = 1;
}
th = &saved_tcphdr;
if (tp->t_flagsext & TF_LRO_OFFLOADED) {
tcp_update_lro_seq(tp->rcv_nxt,
inp->inp_laddr, inp->inp_faddr,
th->th_dport, th->th_sport);
}
}
zfree(tcp_reass_zone, q);
tp->t_reassqlen--;
q = LIST_FIRST(&tp->t_segq);
} while (q && q->tqe_th->th_seq == tp->rcv_nxt);
tp->t_flagsext &= ~TF_REASS_INPROG;
#if INET6
if ((inp->inp_vflag & INP_IPV6) != 0) {
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)),
0, 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)),
0, 0, 0);
}
msg_unordered_delivery:
if (te && (so->so_flags & SOF_ENABLE_MSGS) && copy_oodata && te->tqe_len) {
oodata = m_copym(te->tqe_m, 0, M_COPYALL, M_DONTWAIT);
if (oodata != NULL) {
if (sbappendmsgstream_rcv(&so->so_rcv, oodata,
te->tqe_th->th_seq - (tp->irs + 1), 1)) {
*dowakeup = 1;
tcpstat.tcps_msg_unopkts++;
} else {
tcpstat.tcps_msg_unoappendfail++;
}
}
}
return flags;
}
static void
tcp_reduce_congestion_window(
struct tcpcb *tp)
{
if (CC_ALGO(tp)->pre_fr != NULL) {
CC_ALGO(tp)->pre_fr(tp);
}
ENTER_FASTRECOVERY(tp);
if (tp->t_flags & TF_SENTFIN) {
tp->snd_recover = tp->snd_max - 1;
} else {
tp->snd_recover = tp->snd_max;
}
tp->t_timer[TCPT_REXMT] = 0;
tp->t_timer[TCPT_PTO] = 0;
tp->t_rtttime = 0;
if (tp->t_flagsext & TF_CWND_NONVALIDATED) {
tcp_cc_adjust_nonvalidated_cwnd(tp);
} else {
tp->snd_cwnd = tp->snd_ssthresh +
tp->t_maxseg * tcprexmtthresh;
}
}
static inline void
tcp_adaptive_rwtimo_check(struct tcpcb *tp, int tlen)
{
struct ifnet *outifp = tp->t_inpcb->inp_last_outifp;
if ((tp->t_adaptive_rtimo > 0 ||
(outifp != NULL &&
(outifp->if_eflags & IFEF_PROBE_CONNECTIVITY)))
&& tlen > 0 &&
tp->t_state == TCPS_ESTABLISHED) {
tp->t_timer[TCPT_KEEP] = OFFSET_FROM_START(tp,
(TCP_REXMTVAL(tp) << 1));
tp->t_flagsext |= TF_DETECT_READSTALL;
tp->t_rtimo_probes = 0;
}
}
inline void
tcp_keepalive_reset(struct tcpcb *tp)
{
tp->t_timer[TCPT_KEEP] = OFFSET_FROM_START(tp,
TCP_CONN_KEEPIDLE(tp));
tp->t_flagsext &= ~(TF_DETECT_READSTALL);
tp->t_rtimo_probes = 0;
}
#if INET6
int
tcp6_input(struct mbuf **mp, int *offp, int proto)
{
#pragma unused(proto)
struct mbuf *m = *mp;
uint32_t ia6_flags;
struct ifnet *ifp = m->m_pkthdr.rcvif;
IP6_EXTHDR_CHECK(m, *offp, sizeof(struct tcphdr), return IPPROTO_DONE);
MBUF_STRICT_DATA_ALIGNMENT_CHECK_32(m);
if (ip6_getdstifaddr_info(m, NULL, &ia6_flags) == 0) {
if (ia6_flags & IN6_IFF_ANYCAST) {
struct ip6_hdr *ip6;
ip6 = mtod(m, struct ip6_hdr *);
icmp6_error(m, ICMP6_DST_UNREACH,
ICMP6_DST_UNREACH_ADDR,
(caddr_t)&ip6->ip6_dst - (caddr_t)ip6);
IF_TCP_STATINC(ifp, icmp6unreach);
return IPPROTO_DONE;
}
}
tcp_input(m, *offp);
return IPPROTO_DONE;
}
#endif
u_int8_t
tcp_cansbgrow(struct sockbuf *sb)
{
u_int32_t mblim = ((nmbclusters >> 1) << (MCLSHIFT - MSIZESHIFT));
u_int32_t sbspacelim = ((nmbclusters >> 4) << MCLSHIFT);
if ((total_sbmb_cnt < mblim) &&
(sb->sb_hiwat < sbspacelim)) {
return 1;
} else {
OSIncrementAtomic64(&sbmb_limreached);
}
return 0;
}
static void
tcp_sbrcv_reserve(struct tcpcb *tp, struct sockbuf *sbrcv,
u_int32_t newsize, u_int32_t idealsize, u_int32_t rcvbuf_max)
{
newsize = min(newsize, rcvbuf_max);
newsize = min(newsize, TCP_MAXWIN << tp->rcv_scale);
if (newsize > sbrcv->sb_hiwat &&
(sbreserve(sbrcv, newsize) == 1)) {
sbrcv->sb_idealsize = min(max(sbrcv->sb_idealsize,
(idealsize != 0) ? idealsize : newsize), rcvbuf_max);
sbrcv->sb_idealsize = min(sbrcv->sb_idealsize,
TCP_MAXWIN << tp->rcv_scale);
}
}
static void
tcp_sbrcv_grow(struct tcpcb *tp, struct sockbuf *sbrcv,
struct tcpopt *to, u_int32_t pktlen, u_int32_t rcvbuf_max)
{
struct socket *so = sbrcv->sb_so;
if (tcp_do_autorcvbuf == 0 ||
(sbrcv->sb_flags & SB_AUTOSIZE) == 0 ||
tcp_cansbgrow(sbrcv) == 0 ||
sbrcv->sb_hiwat >= rcvbuf_max ||
(tp->t_flagsext & TF_RECV_THROTTLE) ||
(so->so_flags1 & SOF1_EXTEND_BK_IDLE_WANTED) ||
(!tcp_autotune_reorder && !LIST_EMPTY(&tp->t_segq))) {
goto out;
}
if (TSTMP_GT(tcp_now,
tp->rfbuf_ts + TCPTV_RCVBUFIDLE)) {
goto out;
}
if (!TSTMP_SUPPORTED(tp)) {
if (TSTMP_GEQ(tcp_now,
tp->rfbuf_ts + TCPTV_RCVNOTS_QUANTUM)) {
if (tp->rfbuf_cnt + pktlen >= TCP_RCVNOTS_BYTELEVEL) {
tcp_sbrcv_reserve(tp, sbrcv,
tcp_autorcvbuf_max, 0,
tcp_autorcvbuf_max);
}
goto out;
} else {
tp->rfbuf_cnt += pktlen;
return;
}
} else if (to->to_tsecr != 0) {
if (TSTMP_GEQ(to->to_tsecr, tp->rfbuf_ts)) {
if (tp->rfbuf_cnt + pktlen > (sbrcv->sb_hiwat -
(sbrcv->sb_hiwat >> 1))) {
int32_t rcvbuf_inc, min_incr;
tp->rfbuf_cnt += pktlen;
rcvbuf_inc = (tp->rfbuf_cnt << 1)
- sbrcv->sb_hiwat;
min_incr = tp->t_maxseg << tcp_autorcvbuf_inc_shift;
if (rcvbuf_inc < min_incr) {
rcvbuf_inc = min_incr;
}
rcvbuf_inc =
(rcvbuf_inc / tp->t_maxseg) * tp->t_maxseg;
tcp_sbrcv_reserve(tp, sbrcv,
sbrcv->sb_hiwat + rcvbuf_inc,
(tp->rfbuf_cnt << 1), rcvbuf_max);
}
if (tp->t_bwmeas != NULL && tp->rfbuf_cnt > 0 &&
TSTMP_GT(tcp_now, tp->rfbuf_ts)) {
u_int32_t rcv_bw;
rcv_bw = tp->rfbuf_cnt /
(int)(tcp_now - tp->rfbuf_ts);
if (tp->t_bwmeas->bw_rcvbw_max == 0) {
tp->t_bwmeas->bw_rcvbw_max = rcv_bw;
} else {
tp->t_bwmeas->bw_rcvbw_max = max(
tp->t_bwmeas->bw_rcvbw_max, rcv_bw);
}
}
goto out;
} else {
tp->rfbuf_cnt += pktlen;
return;
}
}
out:
tp->rfbuf_ts = 0;
tp->rfbuf_cnt = 0;
return;
}
void
tcp_sbrcv_trim(struct tcpcb *tp, struct sockbuf *sbrcv)
{
if (tcp_do_autorcvbuf == 1 && sbrcv->sb_idealsize > 0 &&
sbrcv->sb_hiwat > sbrcv->sb_idealsize) {
int32_t trim;
u_int32_t diff = sbrcv->sb_hiwat - sbrcv->sb_idealsize;
u_int32_t advwin = tp->rcv_adv - tp->rcv_nxt;
u_int32_t leave = max(advwin, (sbrcv->sb_idealsize >> 4));
if (leave == 0) {
leave = tp->t_maxseg << tcp_autorcvbuf_inc_shift;
}
trim = sbrcv->sb_hiwat - (sbrcv->sb_cc + leave);
trim = imin(trim, (int32_t)diff);
if (trim > 0) {
sbreserve(sbrcv, (sbrcv->sb_hiwat - trim));
}
}
}
void
tcp_sbsnd_trim(struct sockbuf *sbsnd)
{
if (tcp_do_autosendbuf == 1 &&
((sbsnd->sb_flags & (SB_AUTOSIZE | SB_TRIM)) ==
(SB_AUTOSIZE | SB_TRIM)) &&
(sbsnd->sb_idealsize > 0) &&
(sbsnd->sb_hiwat > sbsnd->sb_idealsize)) {
u_int32_t trim = 0;
if (sbsnd->sb_cc <= sbsnd->sb_idealsize) {
trim = sbsnd->sb_hiwat - sbsnd->sb_idealsize;
} else {
trim = sbsnd->sb_hiwat - sbsnd->sb_cc;
}
sbreserve(sbsnd, (sbsnd->sb_hiwat - trim));
}
if (sbsnd->sb_hiwat <= sbsnd->sb_idealsize) {
sbsnd->sb_flags &= ~(SB_TRIM);
}
}
static inline void
tcp_sbrcv_tstmp_check(struct tcpcb *tp)
{
struct socket *so = tp->t_inpcb->inp_socket;
u_int32_t newsize = 2 * tcp_recvspace;
struct sockbuf *sbrcv = &so->so_rcv;
if ((tp->t_flags & (TF_REQ_TSTMP | TF_RCVD_TSTMP)) !=
(TF_REQ_TSTMP | TF_RCVD_TSTMP) &&
(sbrcv->sb_flags & SB_AUTOSIZE) != 0) {
tcp_sbrcv_reserve(tp, sbrcv, newsize, 0, newsize);
}
}
static inline int
tcp_stretch_ack_enable(struct tcpcb *tp, int thflags)
{
if (tp->rcv_by_unackwin >= (maxseg_unacked * tp->t_maxseg) &&
TSTMP_GEQ(tp->rcv_unackwin, tcp_now)) {
tp->t_flags |= TF_STREAMING_ON;
} else {
tp->t_flags &= ~TF_STREAMING_ON;
}
if (TSTMP_GT(tcp_now, tp->rcv_unackwin + tcp_maxrcvidle)) {
tp->t_flags &= ~TF_STREAMING_ON;
}
if (thflags & ~TH_ACK) {
tp->t_flags &= ~TF_STREAMING_ON;
}
if (tp->t_flagsext & TF_DISABLE_STRETCHACK) {
if (tp->rcv_nostrack_pkts >= TCP_STRETCHACK_ENABLE_PKTCNT) {
tp->t_flagsext &= ~TF_DISABLE_STRETCHACK;
tp->rcv_nostrack_pkts = 0;
tp->rcv_nostrack_ts = 0;
} else {
tp->rcv_nostrack_pkts++;
}
}
if (!(tp->t_flagsext & (TF_NOSTRETCHACK | TF_DISABLE_STRETCHACK)) &&
(tp->t_flags & TF_STREAMING_ON) &&
(!(tp->t_flagsext & TF_RCVUNACK_WAITSS) ||
(tp->rcv_waitforss >= tcp_rcvsspktcnt))) {
return 1;
}
return 0;
}
void
tcp_reset_stretch_ack(struct tcpcb *tp)
{
tp->t_flags &= ~(TF_STRETCHACK | TF_STREAMING_ON);
tp->rcv_by_unackwin = 0;
tp->rcv_by_unackhalfwin = 0;
tp->rcv_unackwin = tcp_now + tcp_rcvunackwin;
tp->t_flagsext |= TF_RCVUNACK_WAITSS;
tp->rcv_waitforss = 0;
}
int
tcp_detect_bad_rexmt(struct tcpcb *tp, struct tcphdr *th,
struct tcpopt *to, u_int32_t rxtime)
{
int32_t tdiff, bad_rexmt_win;
bad_rexmt_win = (tp->t_srtt >> (TCP_RTT_SHIFT + 1));
if (TCP_ECN_ENABLED(tp) && (th->th_flags & TH_ECE)) {
return 0;
}
if (TSTMP_SUPPORTED(tp)) {
if (rxtime > 0 && (to->to_flags & TOF_TS)
&& to->to_tsecr != 0
&& TSTMP_LT(to->to_tsecr, rxtime)) {
return 1;
}
} else {
if ((tp->t_rxtshift == 1
|| (tp->t_flagsext & TF_SENT_TLPROBE))
&& rxtime > 0) {
tdiff = (int32_t)(tcp_now - rxtime);
if (tdiff < bad_rexmt_win) {
return 1;
}
}
}
return 0;
}
static void
tcp_bad_rexmt_restore_state(struct tcpcb *tp, struct tcphdr *th)
{
if (TSTMP_SUPPORTED(tp)) {
u_int32_t fsize, acked;
fsize = tp->snd_max - th->th_ack;
acked = BYTES_ACKED(th, tp);
tp->snd_ssthresh = tp->snd_ssthresh_prev;
if (CC_ALGO(tp)->cwnd_init != NULL) {
CC_ALGO(tp)->cwnd_init(tp);
}
tp->snd_cwnd = fsize + min(acked, tp->snd_cwnd);
} else {
tp->snd_cwnd = tp->snd_cwnd_prev;
tp->snd_ssthresh = tp->snd_ssthresh_prev;
if (tp->t_flags & TF_WASFRECOVERY) {
ENTER_FASTRECOVERY(tp);
}
tp->t_lossflightsize = 0;
}
tp->snd_cwnd = max(tp->snd_cwnd, TCP_CC_CWND_INIT_BYTES);
tp->snd_recover = tp->snd_recover_prev;
tp->snd_nxt = tp->snd_max;
tcp_bad_rexmt_fix_sndbuf(tp);
tp->t_flagsext |= TF_RECOMPUTE_RTT;
tp->t_badrexmt_time = tcp_now;
tp->t_rtttime = 0;
}
static void
tcp_bad_rexmt_check(struct tcpcb *tp, struct tcphdr *th, struct tcpopt *to)
{
if (tp->t_rxtshift > 0 &&
tcp_detect_bad_rexmt(tp, th, to, tp->t_rxtstart)) {
++tcpstat.tcps_sndrexmitbad;
tcp_bad_rexmt_restore_state(tp, th);
tcp_ccdbg_trace(tp, th, TCP_CC_BAD_REXMT_RECOVERY);
} else if ((tp->t_flagsext & TF_SENT_TLPROBE)
&& tp->t_tlphighrxt > 0
&& SEQ_GEQ(th->th_ack, tp->t_tlphighrxt)
&& !tcp_detect_bad_rexmt(tp, th, to, tp->t_tlpstart)) {
if (tcp_rxtseg_dsack_for_tlp(tp)) {
tcp_rxtseg_clean(tp);
goto out;
}
++tcpstat.tcps_tlp_recoverlastpkt;
if (!IN_FASTRECOVERY(tp)) {
tcp_reduce_congestion_window(tp);
EXIT_FASTRECOVERY(tp);
}
tcp_ccdbg_trace(tp, th, TCP_CC_TLP_RECOVER_LASTPACKET);
} else if (tcp_rxtseg_detect_bad_rexmt(tp, th->th_ack)) {
tcpstat.tcps_dsack_badrexmt++;
tcp_bad_rexmt_restore_state(tp, th);
tcp_ccdbg_trace(tp, th, TCP_CC_DSACK_BAD_REXMT);
tcp_rxtseg_clean(tp);
}
out:
tp->t_flagsext &= ~(TF_SENT_TLPROBE);
tp->t_tlphighrxt = 0;
tp->t_tlpstart = 0;
if (tp->t_rxtshift > 0 && (tp->t_flags & TF_BLACKHOLE) &&
tp->t_pmtud_start_ts > 0 && TSTMP_SUPPORTED(tp)) {
if ((to->to_flags & TOF_TS) && to->to_tsecr != 0
&& TSTMP_LT(to->to_tsecr, tp->t_pmtud_start_ts)) {
tcp_pmtud_revert_segment_size(tp);
}
}
if (tp->t_pmtud_start_ts > 0) {
tp->t_pmtud_start_ts = 0;
}
}
static void
tcp_early_rexmt_check(struct tcpcb *tp, struct tcphdr *th)
{
u_int32_t obytes, snd_off;
int32_t snd_len;
struct socket *so = tp->t_inpcb->inp_socket;
if (early_rexmt && (SACK_ENABLED(tp) ||
tp->t_early_rexmt_count < TCP_EARLY_REXMT_LIMIT) &&
SEQ_GT(tp->snd_max, tp->snd_una) &&
(tp->t_dupacks == 1 ||
(SACK_ENABLED(tp) &&
!TAILQ_EMPTY(&tp->snd_holes)))) {
if (SACK_ENABLED(tp) &&
!TAILQ_EMPTY(&tp->snd_holes)) {
obytes = (tp->snd_max - tp->snd_fack) +
tp->sackhint.sack_bytes_rexmit;
} else {
obytes = (tp->snd_max - tp->snd_una);
}
snd_off = tp->snd_max - tp->snd_una;
snd_len = min(so->so_snd.sb_cc, tp->snd_wnd) - snd_off;
if (obytes < (tp->t_maxseg << 2) &&
snd_len <= 0) {
u_int32_t osegs;
osegs = obytes / tp->t_maxseg;
if ((osegs * tp->t_maxseg) < obytes) {
osegs++;
}
osegs += tp->t_dupacks;
if (osegs < 4) {
tp->t_rexmtthresh =
((osegs - 1) > 1) ? (osegs - 1) : 1;
tp->t_rexmtthresh =
min(tp->t_rexmtthresh, tcprexmtthresh);
tp->t_rexmtthresh =
max(tp->t_rexmtthresh, tp->t_dupacks);
if (tp->t_early_rexmt_count == 0) {
tp->t_early_rexmt_win = tcp_now;
}
if (tp->t_flagsext & TF_SENT_TLPROBE) {
tcpstat.tcps_tlp_recovery++;
tcp_ccdbg_trace(tp, th,
TCP_CC_TLP_RECOVERY);
} else {
tcpstat.tcps_early_rexmt++;
tp->t_early_rexmt_count++;
tcp_ccdbg_trace(tp, th,
TCP_CC_EARLY_RETRANSMIT);
}
}
}
}
tp->t_flagsext &= ~(TF_SENT_TLPROBE);
tp->t_tlphighrxt = 0;
tp->t_tlpstart = 0;
}
static boolean_t
tcp_tfo_syn(struct tcpcb *tp, struct tcpopt *to)
{
u_char out[CCAES_BLOCK_SIZE];
unsigned char len;
if (!(to->to_flags & (TOF_TFO | TOF_TFOREQ)) ||
!(tcp_fastopen & TCP_FASTOPEN_SERVER)) {
return FALSE;
}
if ((to->to_flags & TOF_TFOREQ)) {
tp->t_tfo_flags |= TFO_F_OFFER_COOKIE;
tp->t_tfo_stats |= TFO_S_COOKIEREQ_RECV;
tcpstat.tcps_tfo_cookie_req_rcv++;
return FALSE;
}
tcp_tfo_gen_cookie(tp->t_inpcb, out, sizeof(out));
len = *to->to_tfo - TCPOLEN_FASTOPEN_REQ;
to->to_tfo++;
if (memcmp(out, to->to_tfo, len)) {
tp->t_tfo_flags |= TFO_F_OFFER_COOKIE;
tp->t_tfo_stats |= TFO_S_COOKIE_INVALID;
tcpstat.tcps_tfo_cookie_invalid++;
return FALSE;
}
if (OSIncrementAtomic(&tcp_tfo_halfcnt) >= tcp_tfo_backlog) {
OSDecrementAtomic(&tcp_tfo_halfcnt);
return FALSE;
}
tp->t_tfo_flags |= TFO_F_COOKIE_VALID;
tp->t_tfo_stats |= TFO_S_SYNDATA_RCV;
tcpstat.tcps_tfo_syn_data_rcv++;
return TRUE;
}
static void
tcp_tfo_synack(struct tcpcb *tp, struct tcpopt *to)
{
if (to->to_flags & TOF_TFO) {
unsigned char len = *to->to_tfo - TCPOLEN_FASTOPEN_REQ;
VERIFY(len <= TFO_COOKIE_LEN_MAX);
to->to_tfo++;
tcp_cache_set_cookie(tp, to->to_tfo, len);
tcp_heuristic_tfo_success(tp);
tp->t_tfo_stats |= TFO_S_COOKIE_RCV;
tcpstat.tcps_tfo_cookie_rcv++;
if (tp->t_tfo_flags & TFO_F_COOKIE_SENT) {
tcpstat.tcps_tfo_cookie_wrong++;
tp->t_tfo_stats |= TFO_S_COOKIE_WRONG;
}
} else {
if (!(tp->t_flagsext & TF_FASTOPEN_FORCE_ENABLE) &&
tp->t_tfo_flags & TFO_F_SYN_LOSS) {
tp->t_tfo_stats |= TFO_S_SYN_LOSS;
tcpstat.tcps_tfo_syn_loss++;
tcp_heuristic_tfo_loss(tp);
} else {
if (tp->t_tfo_flags & TFO_F_COOKIE_REQ) {
tp->t_tfo_stats |= TFO_S_NO_COOKIE_RCV;
tcpstat.tcps_tfo_no_cookie_rcv++;
}
tcp_heuristic_tfo_success(tp);
}
}
}
static void
tcp_tfo_rcv_probe(struct tcpcb *tp, int tlen)
{
if (tlen != 0) {
return;
}
tp->t_tfo_probe_state = TFO_PROBE_PROBING;
tp->t_timer[TCPT_KEEP] = OFFSET_FROM_START(tp, (TCP_REXMTVAL(tp)));
}
static void
tcp_tfo_rcv_data(struct tcpcb *tp)
{
if (tp->t_tfo_probe_state >= TFO_PROBE_PROBING) {
tp->t_tfo_probe_state = TFO_PROBE_NONE;
}
}
static void
tcp_tfo_rcv_ack(struct tcpcb *tp, struct tcphdr *th)
{
if (tp->t_tfo_probe_state == TFO_PROBE_PROBING &&
tp->t_tfo_probes > 0) {
if (th->th_seq == tp->rcv_nxt) {
tp->t_tfo_probe_state = TFO_PROBE_NONE;
} else if (SEQ_GT(th->th_seq, tp->rcv_nxt)) {
tp->t_tfo_probe_state = TFO_PROBE_WAIT_DATA;
tp->t_timer[TCPT_KEEP] = OFFSET_FROM_START(tp,
TCP_REXMTVAL(tp));
}
}
}
static inline bool
tcp_update_window(struct tcpcb *tp, int thflags, struct tcphdr * th,
u_int32_t tiwin, int tlen)
{
if ((thflags & TH_ACK) &&
(SEQ_LT(tp->snd_wl1, th->th_seq) ||
(tp->snd_wl1 == th->th_seq && (SEQ_LT(tp->snd_wl2, th->th_ack) ||
(tp->snd_wl2 == th->th_ack && tiwin > tp->snd_wnd))))) {
if (tlen == 0 &&
tp->snd_wl2 == th->th_ack && tiwin > tp->snd_wnd) {
tcpstat.tcps_rcvwinupd++;
}
tp->snd_wnd = tiwin;
tp->snd_wl1 = th->th_seq;
tp->snd_wl2 = th->th_ack;
if (tp->snd_wnd > tp->max_sndwnd) {
tp->max_sndwnd = tp->snd_wnd;
}
if (tp->t_inpcb->inp_socket->so_flags & SOF_MP_SUBFLOW) {
mptcp_update_window_wakeup(tp);
}
return true;
}
return false;
}
static void
tcp_handle_wakeup(struct socket *so, int read_wakeup, int write_wakeup)
{
if (read_wakeup != 0) {
sorwakeup(so);
}
if (write_wakeup != 0) {
sowwakeup(so);
}
}
void
tcp_input(struct mbuf *m, int off0)
{
struct tcphdr *th;
struct ip *ip = NULL;
struct inpcb *inp;
u_char *optp = NULL;
int optlen = 0;
int tlen, off;
int drop_hdrlen;
struct tcpcb *tp = 0;
int thflags;
struct socket *so = 0;
int todrop, acked, ourfinisacked, needoutput = 0;
int read_wakeup = 0;
int write_wakeup = 0;
struct in_addr laddr;
#if INET6
struct in6_addr laddr6;
#endif
int dropsocket = 0;
int iss = 0, nosock = 0;
u_int32_t tiwin, sack_bytes_acked = 0;
struct tcpopt to;
#if TCPDEBUG
short ostate = 0;
#endif
#if IPFIREWALL
struct sockaddr_in *next_hop = NULL;
struct m_tag *fwd_tag;
#endif
u_char ip_ecn = IPTOS_ECN_NOTECT;
unsigned int ifscope;
uint8_t isconnected, isdisconnected;
struct ifnet *ifp = m->m_pkthdr.rcvif;
int pktf_sw_lro_pkt = (m->m_pkthdr.pkt_flags & PKTF_SW_LRO_PKT) ? 1 : 0;
int nlropkts = (pktf_sw_lro_pkt == 1) ? m->m_pkthdr.lro_npkts : 1;
int turnoff_lro = 0, win;
#if MPTCP
struct mptcb *mp_tp = NULL;
#endif
boolean_t cell = IFNET_IS_CELLULAR(ifp);
boolean_t wifi = (!cell && IFNET_IS_WIFI(ifp));
boolean_t wired = (!wifi && IFNET_IS_WIRED(ifp));
boolean_t recvd_dsack = FALSE;
struct tcp_respond_args tra;
int prev_t_state;
boolean_t check_cfil = cfil_filter_present();
bool findpcb_iterated = false;
struct tcphdr saved_tcphdr = {};
#define MAX_IPWORDS ((sizeof(struct ip) + MAX_IPOPTLEN) / sizeof(uint32_t))
uint32_t saved_hdr[MAX_IPWORDS];
#define TCP_INC_VAR(stat, npkts) do { \
stat += npkts; \
} while (0)
TCP_INC_VAR(tcpstat.tcps_rcvtotal, nlropkts);
#if IPFIREWALL
if (!SLIST_EMPTY(&m->m_pkthdr.tags)) {
fwd_tag = m_tag_locate(m, KERNEL_MODULE_TAG_ID,
KERNEL_TAG_TYPE_IPFORWARD, NULL);
} else {
fwd_tag = NULL;
}
if (fwd_tag != NULL) {
struct ip_fwd_tag *ipfwd_tag =
(struct ip_fwd_tag *)(fwd_tag + 1);
next_hop = ipfwd_tag->next_hop;
m_tag_delete(m, fwd_tag);
}
#endif
#if INET6
struct ip6_hdr *ip6 = NULL;
int isipv6;
#endif
int rstreason;
struct proc *proc0 = current_proc();
KERNEL_DEBUG(DBG_FNC_TCP_INPUT | DBG_FUNC_START, 0, 0, 0, 0, 0);
#if INET6
isipv6 = (mtod(m, struct ip *)->ip_v == 6) ? 1 : 0;
#endif
bzero((char *)&to, sizeof(to));
#if INET6
if (isipv6) {
MBUF_STRICT_DATA_ALIGNMENT_CHECK_32(m);
ip6 = mtod(m, struct ip6_hdr *);
tlen = sizeof(*ip6) + ntohs(ip6->ip6_plen) - off0;
th = (struct tcphdr *)(void *)((caddr_t)ip6 + off0);
if (tcp_input_checksum(AF_INET6, m, th, off0, tlen)) {
TCP_LOG_DROP_PKT(ip6, th, ifp, "IPv6 bad tcp checksum");
goto dropnosock;
}
KERNEL_DEBUG(DBG_LAYER_BEG, ((th->th_dport << 16) | th->th_sport),
(((ip6->ip6_src.s6_addr16[0]) << 16) | (ip6->ip6_dst.s6_addr16[0])),
th->th_seq, th->th_ack, th->th_win);
if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) {
IF_TCP_STATINC(ifp, unspecv6);
TCP_LOG_DROP_PKT(ip6, th, ifp, "src IPv6 address unspecified");
goto dropnosock;
}
DTRACE_TCP5(receive, struct mbuf *, m, struct inpcb *, NULL,
struct ip6_hdr *, ip6, struct tcpcb *, NULL,
struct tcphdr *, th);
ip_ecn = (ntohl(ip6->ip6_flow) >> 20) & IPTOS_ECN_MASK;
} else
#endif
{
if (off0 > sizeof(struct ip)) {
ip_stripoptions(m);
off0 = sizeof(struct ip);
}
if (m->m_len < sizeof(struct tcpiphdr)) {
if ((m = m_pullup(m, sizeof(struct tcpiphdr))) == 0) {
tcpstat.tcps_rcvshort++;
return;
}
}
MBUF_STRICT_DATA_ALIGNMENT_CHECK_32(m);
ip = mtod(m, struct ip *);
th = (struct tcphdr *)(void *)((caddr_t)ip + off0);
tlen = ip->ip_len;
if (tcp_input_checksum(AF_INET, m, th, off0, tlen)) {
TCP_LOG_DROP_PKT(ip, th, ifp, "IPv4 bad tcp checksum");
goto dropnosock;
}
#if INET6
ip->ip_v = IPVERSION;
#endif
ip_ecn = (ip->ip_tos & IPTOS_ECN_MASK);
DTRACE_TCP5(receive, struct mbuf *, m, struct inpcb *, NULL,
struct ip *, ip, struct tcpcb *, NULL, struct tcphdr *, th);
KERNEL_DEBUG(DBG_LAYER_BEG, ((th->th_dport << 16) | th->th_sport),
(((ip->ip_src.s_addr & 0xffff) << 16) | (ip->ip_dst.s_addr & 0xffff)),
th->th_seq, th->th_ack, th->th_win);
}
#define TCP_LOG_HDR (isipv6 ? (void *)ip6 : (void *)ip)
off = th->th_off << 2;
if (off < sizeof(struct tcphdr) || off > tlen) {
tcpstat.tcps_rcvbadoff++;
IF_TCP_STATINC(ifp, badformat);
TCP_LOG_DROP_PKT(TCP_LOG_HDR, th, ifp, "bad tcp offset");
goto dropnosock;
}
tlen -= off;
if (off > sizeof(struct tcphdr)) {
#if INET6
if (isipv6) {
IP6_EXTHDR_CHECK(m, off0, off, return );
ip6 = mtod(m, struct ip6_hdr *);
th = (struct tcphdr *)(void *)((caddr_t)ip6 + off0);
} else
#endif
{
if (m->m_len < sizeof(struct ip) + off) {
if ((m = m_pullup(m, sizeof(struct ip) + off)) == 0) {
tcpstat.tcps_rcvshort++;
return;
}
ip = mtod(m, struct ip *);
th = (struct tcphdr *)(void *)((caddr_t)ip + off0);
}
}
optlen = off - sizeof(struct tcphdr);
optp = (u_char *)(th + 1);
if ((optlen == TCPOLEN_TSTAMP_APPA ||
(optlen > TCPOLEN_TSTAMP_APPA &&
optp[TCPOLEN_TSTAMP_APPA] == TCPOPT_EOL)) &&
*(u_int32_t *)(void *)optp == htonl(TCPOPT_TSTAMP_HDR) &&
(th->th_flags & TH_SYN) == 0) {
to.to_flags |= TOF_TS;
to.to_tsval = ntohl(*(u_int32_t *)(void *)(optp + 4));
to.to_tsecr = ntohl(*(u_int32_t *)(void *)(optp + 8));
optp = NULL;
}
}
thflags = th->th_flags;
#if TCP_DROP_SYNFIN
if (drop_synfin && (thflags & (TH_SYN | TH_FIN)) == (TH_SYN | TH_FIN)) {
IF_TCP_STATINC(ifp, synfin);
TCP_LOG_DROP_PKT(TCP_LOG_HDR, th, ifp, "drop SYN FIN");
goto dropnosock;
}
#endif
drop_hdrlen = off0 + off;
calculate_tcp_clock();
if ((m->m_pkthdr.pkt_flags & PKTF_LOOP) || m->m_pkthdr.rcvif == NULL) {
ifscope = IFSCOPE_NONE;
} else {
ifscope = m->m_pkthdr.rcvif->if_index;
}
#if BYTE_ORDER != BIG_ENDIAN
NTOHL(th->th_seq);
NTOHL(th->th_ack);
NTOHS(th->th_win);
NTOHS(th->th_urp);
#endif
findpcb:
isconnected = FALSE;
isdisconnected = FALSE;
#if IPFIREWALL_FORWARD
if (next_hop != NULL
#if INET6
&& isipv6 == 0
#endif
) {
inp = in_pcblookup_hash(&tcbinfo, ip->ip_src, th->th_sport,
ip->ip_dst, th->th_dport, 0, m->m_pkthdr.rcvif);
if (!inp) {
if (!next_hop->sin_port) {
inp = in_pcblookup_hash(&tcbinfo, ip->ip_src,
th->th_sport, next_hop->sin_addr,
th->th_dport, 1, m->m_pkthdr.rcvif);
} else {
inp = in_pcblookup_hash(&tcbinfo,
ip->ip_src, th->th_sport,
next_hop->sin_addr,
ntohs(next_hop->sin_port), 1,
m->m_pkthdr.rcvif);
}
}
} else
#endif
{
#if INET6
if (isipv6) {
inp = in6_pcblookup_hash(&tcbinfo, &ip6->ip6_src, th->th_sport,
&ip6->ip6_dst, th->th_dport, 1,
m->m_pkthdr.rcvif);
} else
#endif
inp = in_pcblookup_hash(&tcbinfo, ip->ip_src, th->th_sport,
ip->ip_dst, th->th_dport, 1, m->m_pkthdr.rcvif);
}
if (inp != NULL && (inp->inp_flags & INP_BOUND_IF)) {
ifscope = inp->inp_boundifp->if_index;
}
if (inp == NULL) {
if (log_in_vain) {
#if INET6
char dbuf[MAX_IPv6_STR_LEN], sbuf[MAX_IPv6_STR_LEN];
#else
char dbuf[MAX_IPv4_STR_LEN], sbuf[MAX_IPv4_STR_LEN];
#endif
#if INET6
if (isipv6) {
inet_ntop(AF_INET6, &ip6->ip6_dst, dbuf, sizeof(dbuf));
inet_ntop(AF_INET6, &ip6->ip6_src, sbuf, sizeof(sbuf));
} else
#endif
{
inet_ntop(AF_INET, &ip->ip_dst, dbuf, sizeof(dbuf));
inet_ntop(AF_INET, &ip->ip_src, sbuf, sizeof(sbuf));
}
switch (log_in_vain) {
case 1:
if (thflags & TH_SYN) {
log(LOG_INFO,
"Connection attempt to TCP %s:%d from %s:%d\n",
dbuf, ntohs(th->th_dport),
sbuf,
ntohs(th->th_sport));
}
break;
case 2:
log(LOG_INFO,
"Connection attempt to TCP %s:%d from %s:%d flags:0x%x\n",
dbuf, ntohs(th->th_dport), sbuf,
ntohs(th->th_sport), thflags);
break;
case 3:
case 4:
if ((thflags & TH_SYN) && !(thflags & TH_ACK) &&
!(m->m_flags & (M_BCAST | M_MCAST)) &&
#if INET6
((isipv6 && !IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &ip6->ip6_src)) ||
(!isipv6 && ip->ip_dst.s_addr != ip->ip_src.s_addr))
#else
ip->ip_dst.s_addr != ip->ip_src.s_addr
#endif
) {
log_in_vain_log((LOG_INFO,
"Stealth Mode connection attempt to TCP %s:%d from %s:%d\n",
dbuf, ntohs(th->th_dport),
sbuf,
ntohs(th->th_sport)));
}
break;
default:
break;
}
}
if (blackhole) {
if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type != IFT_LOOP) {
switch (blackhole) {
case 1:
if (thflags & TH_SYN) {
TCP_LOG_DROP_PKT(TCP_LOG_HDR, th, ifp, "blackhole 1 syn for closed port");
goto dropnosock;
}
break;
case 2:
TCP_LOG_DROP_PKT(TCP_LOG_HDR, th, ifp, "blackhole 2 closed port");
goto dropnosock;
default:
TCP_LOG_DROP_PKT(TCP_LOG_HDR, th, ifp, "blackhole closed port");
goto dropnosock;
}
}
}
rstreason = BANDLIM_RST_CLOSEDPORT;
IF_TCP_STATINC(ifp, noconnnolist);
TCP_LOG_DROP_PKT(TCP_LOG_HDR, th, ifp, "closed port");
goto dropwithresetnosock;
}
so = inp->inp_socket;
if (so == NULL) {
inp = NULL;
#if TEMPDEBUG
printf("tcp_input: no more socket for inp=%x. This shouldn't happen\n", inp);
#endif
TCP_LOG_DROP_PKT(TCP_LOG_HDR, th, ifp, "inp_socket NULL");
goto dropnosock;
}
socket_lock(so, 1);
if (in_pcb_checkstate(inp, WNT_RELEASE, 1) == WNT_STOPUSING) {
socket_unlock(so, 1);
inp = NULL; TCP_LOG_DROP_PKT(TCP_LOG_HDR, th, ifp, "inp state WNT_STOPUSING");
goto dropnosock;
}
if (!isipv6 && inp->inp_faddr.s_addr != INADDR_ANY) {
if (inp->inp_faddr.s_addr != ip->ip_src.s_addr ||
inp->inp_laddr.s_addr != ip->ip_dst.s_addr ||
inp->inp_fport != th->th_sport ||
inp->inp_lport != th->th_dport) {
os_log_error(OS_LOG_DEFAULT, "%s 5-tuple does not match: %u:%u %u:%u\n",
__func__,
ntohs(inp->inp_fport), ntohs(th->th_sport),
ntohs(inp->inp_lport), ntohs(th->th_dport));
if (findpcb_iterated) {
goto drop;
}
findpcb_iterated = true;
socket_unlock(so, 1);
inp = NULL;
goto findpcb;
}
} else if (isipv6 && !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, &ip6->ip6_src) ||
!IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, &ip6->ip6_dst) ||
inp->inp_fport != th->th_sport ||
inp->inp_lport != th->th_dport) {
os_log_error(OS_LOG_DEFAULT, "%s 5-tuple does not match: %u:%u %u:%u\n",
__func__,
ntohs(inp->inp_fport), ntohs(th->th_sport),
ntohs(inp->inp_lport), ntohs(th->th_dport));
if (findpcb_iterated) {
goto drop;
}
findpcb_iterated = true;
socket_unlock(so, 1);
inp = NULL;
goto findpcb;
}
}
tp = intotcpcb(inp);
if (tp == NULL) {
rstreason = BANDLIM_RST_CLOSEDPORT;
IF_TCP_STATINC(ifp, noconnlist);
TCP_LOG_DROP_PKT(TCP_LOG_HDR, th, ifp, "tp is NULL");
goto dropwithreset;
}
TCP_LOG_TH_FLAGS(TCP_LOG_HDR, th, tp, false, ifp);
if (tp->t_state == TCPS_CLOSED) {
TCP_LOG_DROP_PCB(TCP_LOG_HDR, th, tp, false, "tp state TCPS_CLOSED");
goto drop;
}
#if NECP
if (so->so_state & SS_ISCONNECTED) {
if (!necp_socket_is_allowed_to_send_recv(inp, ifp, NULL, NULL, NULL)) {
TCP_LOG_DROP_NECP(TCP_LOG_HDR, th, intotcpcb(inp), false);
IF_TCP_STATINC(ifp, badformat);
goto drop;
}
} else {
(void) inp_update_policy(inp);
#if INET6
if (isipv6) {
if (!necp_socket_is_allowed_to_send_recv_v6(inp,
th->th_dport, th->th_sport, &ip6->ip6_dst,
&ip6->ip6_src, ifp, NULL, NULL, NULL)) {
TCP_LOG_DROP_NECP(TCP_LOG_HDR, th, intotcpcb(inp), false);
IF_TCP_STATINC(ifp, badformat);
goto drop;
}
} else
#endif
{
if (!necp_socket_is_allowed_to_send_recv_v4(inp,
th->th_dport, th->th_sport, &ip->ip_dst, &ip->ip_src,
ifp, NULL, NULL, NULL)) {
TCP_LOG_DROP_NECP(TCP_LOG_HDR, th, intotcpcb(inp), false);
IF_TCP_STATINC(ifp, badformat);
goto drop;
}
}
}
#endif
prev_t_state = tp->t_state;
if (tcp_do_rfc5961 && (thflags & TH_ACCEPT) == 0) {
TCP_LOG_DROP_PCB(TCP_LOG_HDR, th, tp, false, "rfc5961 TH_ACCEPT == 0");
goto drop;
}
if ((thflags & TH_SYN) == 0) {
tiwin = th->th_win << tp->snd_scale;
} else {
tiwin = th->th_win;
}
#if CONFIG_MACF_NET
if (mac_inpcb_check_deliver(inp, m, AF_INET, SOCK_STREAM)) {
TCP_LOG_DROP_PCB(TCP_LOG_HDR, th, tp, false, "mac_inpcb_check_deliver failed");
goto drop;
}
#endif
if (tp->t_state == TCPS_LISTEN &&
(so->so_options & SO_ACCEPTCONN) == 0) {
TCP_LOG_DROP_PCB(TCP_LOG_HDR, th, tp, false, "closing a listening socket");
goto drop;
}
if (so->so_options & (SO_DEBUG | SO_ACCEPTCONN)) {
#if TCPDEBUG
if (so->so_options & SO_DEBUG) {
ostate = tp->t_state;
#if INET6
if (isipv6) {
bcopy((char *)ip6, (char *)tcp_saveipgen,
sizeof(*ip6));
} else
#endif
bcopy((char *)ip, (char *)tcp_saveipgen, sizeof(*ip));
tcp_savetcp = *th;
}
#endif
if (so->so_options & SO_ACCEPTCONN) {
struct tcpcb *tp0 = tp;
struct socket *so2;
struct socket *oso;
struct sockaddr_storage from;
struct sockaddr_storage to2;
#if INET6
struct inpcb *oinp = sotoinpcb(so);
#endif
struct ifnet *head_ifscope;
unsigned int head_nocell, head_recvanyif,
head_noexpensive, head_awdl_unrestricted,
head_intcoproc_allowed, head_external_port,
head_noconstrained;
head_ifscope = (inp->inp_flags & INP_BOUND_IF) ?
inp->inp_boundifp : NULL;
head_nocell = INP_NO_CELLULAR(inp);
head_recvanyif = (inp->inp_flags & INP_RECV_ANYIF);
head_noexpensive = INP_NO_EXPENSIVE(inp);
head_noconstrained = INP_NO_CONSTRAINED(inp);
head_awdl_unrestricted = INP_AWDL_UNRESTRICTED(inp);
head_intcoproc_allowed = INP_INTCOPROC_ALLOWED(inp);
head_external_port = (inp->inp_flags2 & INP2_EXTERNAL_PORT);
if ((thflags & (TH_RST | TH_ACK | TH_SYN)) != TH_SYN) {
IF_TCP_STATINC(ifp, listbadsyn);
if (thflags & TH_RST) {
TCP_LOG_DROP_PCB(TCP_LOG_HDR, th, tp, false, "SYN with RST");
goto drop;
}
if (thflags & TH_ACK) {
TCP_LOG_DROP_PCB(TCP_LOG_HDR, th, tp, false, "SYN with ACK");
tp = NULL;
tcpstat.tcps_badsyn++;
rstreason = BANDLIM_RST_OPENPORT;
goto dropwithreset;
}
tcpstat.tcps_badsyn++;
TCP_LOG_DROP_PCB(TCP_LOG_HDR, th, tp, false, "bad SYN");
goto drop;
}
KERNEL_DEBUG(DBG_FNC_TCP_NEWCONN | DBG_FUNC_START, 0, 0, 0, 0, 0);
if (th->th_dport == th->th_sport) {
#if INET6
if (isipv6) {
if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
&ip6->ip6_src)) {
TCP_LOG_DROP_PCB(TCP_LOG_HDR, th, tp, false, "bad tuple same port");
goto drop;
}
} else
#endif
if (ip->ip_dst.s_addr == ip->ip_src.s_addr) {
TCP_LOG_DROP_PCB(TCP_LOG_HDR, th, tp, false, "bad tuple same IPv4 address");
goto drop;
}
}
if (m->m_flags & (M_BCAST | M_MCAST)) {
TCP_LOG_DROP_PCB(TCP_LOG_HDR, th, tp, false, "mbuf M_BCAST | M_MCAST");
goto drop;
}
#if INET6
if (isipv6) {
if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
TCP_LOG_DROP_PCB(TCP_LOG_HDR, th, tp, false, "IN6_IS_ADDR_MULTICAST");
goto drop;
}
} else
#endif
if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) ||
IN_MULTICAST(ntohl(ip->ip_src.s_addr)) ||
ip->ip_src.s_addr == htonl(INADDR_BROADCAST) ||
in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif)) {
TCP_LOG_DROP_PCB(TCP_LOG_HDR, th, tp, false, "multicast or broadcast address");
goto drop;
}
#if INET6
if (isipv6 && !ip6_use_deprecated) {
uint32_t ia6_flags;
if (ip6_getdstifaddr_info(m, NULL,
&ia6_flags) == 0) {
if (ia6_flags & IN6_IFF_DEPRECATED) {
tp = NULL;
rstreason = BANDLIM_RST_OPENPORT;
IF_TCP_STATINC(ifp, deprecate6);
TCP_LOG_DROP_PCB(TCP_LOG_HDR, th, tp, false, "deprecated IPv6 address");
goto dropwithreset;
}
}
}
#endif
if (so->so_filt || check_cfil) {
#if INET6
if (isipv6) {
struct sockaddr_in6 *sin6 = (struct sockaddr_in6*)&from;
sin6->sin6_len = sizeof(*sin6);
sin6->sin6_family = AF_INET6;
sin6->sin6_port = th->th_sport;
sin6->sin6_flowinfo = 0;
sin6->sin6_addr = ip6->ip6_src;
sin6->sin6_scope_id = 0;
sin6 = (struct sockaddr_in6*)&to2;
sin6->sin6_len = sizeof(struct sockaddr_in6);
sin6->sin6_family = AF_INET6;
sin6->sin6_port = th->th_dport;
sin6->sin6_flowinfo = 0;
sin6->sin6_addr = ip6->ip6_dst;
sin6->sin6_scope_id = 0;
} else
#endif
{
struct sockaddr_in *sin = (struct sockaddr_in*)&from;
sin->sin_len = sizeof(*sin);
sin->sin_family = AF_INET;
sin->sin_port = th->th_sport;
sin->sin_addr = ip->ip_src;
sin = (struct sockaddr_in*)&to2;
sin->sin_len = sizeof(struct sockaddr_in);
sin->sin_family = AF_INET;
sin->sin_port = th->th_dport;
sin->sin_addr = ip->ip_dst;
}
}
if (so->so_filt) {
so2 = sonewconn(so, 0, (struct sockaddr*)&from);
} else {
so2 = sonewconn(so, 0, NULL);
}
if (so2 == 0) {
tcpstat.tcps_listendrop++;
if (tcp_dropdropablreq(so)) {
if (so->so_filt) {
so2 = sonewconn(so, 0, (struct sockaddr*)&from);
} else {
so2 = sonewconn(so, 0, NULL);
}
}
if (!so2) {
TCP_LOG_DROP_PCB(TCP_LOG_HDR, th, tp, false, " listen drop");
goto drop;
}
}
inp = (struct inpcb *)so2->so_pcb;
tp = intotcpcb(inp);
oso = so;
socket_unlock(so, 0);
so = so2;
socket_lock(so, 1);
dropsocket++;
if (head_ifscope != NULL) {
inp->inp_flags |= INP_BOUND_IF;
inp->inp_boundifp = head_ifscope;
} else {
inp->inp_flags &= ~INP_BOUND_IF;
}
if (head_nocell) {
inp_set_nocellular(inp);
}
if (head_noexpensive) {
inp_set_noexpensive(inp);
}
if (head_noconstrained) {
inp_set_noconstrained(inp);
}
if (head_awdl_unrestricted) {
inp_set_awdl_unrestricted(inp);
}
if (head_intcoproc_allowed) {
inp_set_intcoproc_allowed(inp);
}
if (head_recvanyif) {
inp->inp_flags |= INP_RECV_ANYIF;
} else {
inp->inp_flags &= ~INP_RECV_ANYIF;
}
if (head_external_port) {
inp->inp_flags2 |= INP2_EXTERNAL_PORT;
}
#if INET6
if (isipv6) {
inp->in6p_laddr = ip6->ip6_dst;
} else {
inp->inp_vflag &= ~INP_IPV6;
inp->inp_vflag |= INP_IPV4;
#endif
inp->inp_laddr = ip->ip_dst;
#if INET6
}
#endif
inp->inp_lport = th->th_dport;
if (in_pcbinshash(inp, 0) != 0) {
#if INET6
if (isipv6) {
inp->in6p_laddr = in6addr_any;
} else
#endif
inp->inp_laddr.s_addr = INADDR_ANY;
inp->inp_lport = 0;
socket_lock(oso, 0);
socket_unlock(oso, 1);
TCP_LOG_DROP_PCB(TCP_LOG_HDR, th, tp, false, " in_pcbinshash failed");
goto drop;
}
socket_lock(oso, 0);
#if INET6
if (isipv6) {
inp->inp_flags |=
oinp->inp_flags & INP_CONTROLOPTS;
if (oinp->in6p_outputopts) {
inp->in6p_outputopts =
ip6_copypktopts(oinp->in6p_outputopts,
M_NOWAIT);
}
} else
#endif
{
inp->inp_options = ip_srcroute();
inp->inp_ip_tos = oinp->inp_ip_tos;
}
#if IPSEC
if (sotoinpcb(oso)->inp_sp) {
int error = 0;
if (inp->inp_sp != NULL) {
error = ipsec_init_policy(so, &inp->inp_sp);
}
if (error != 0 || ipsec_copy_policy(sotoinpcb(oso)->inp_sp, inp->inp_sp)) {
printf("tcp_input: could not copy policy\n");
}
}
#endif
DTRACE_TCP4(state__change, void, NULL, struct inpcb *, inp,
struct tcpcb *, tp, int32_t, TCPS_LISTEN);
tp->t_state = TCPS_LISTEN;
tp->t_flags |= tp0->t_flags & (TF_NOPUSH | TF_NOOPT | TF_NODELAY);
tp->t_flagsext |= (tp0->t_flagsext & (TF_RXTFINDROP | TF_NOTIMEWAIT | TF_FASTOPEN));
tp->t_keepinit = tp0->t_keepinit;
tp->t_keepcnt = tp0->t_keepcnt;
tp->t_keepintvl = tp0->t_keepintvl;
tp->t_adaptive_wtimo = tp0->t_adaptive_wtimo;
tp->t_adaptive_rtimo = tp0->t_adaptive_rtimo;
tp->t_inpcb->inp_ip_ttl = tp0->t_inpcb->inp_ip_ttl;
if ((so->so_flags & SOF_NOTSENT_LOWAT) != 0) {
tp->t_notsent_lowat = tp0->t_notsent_lowat;
}
tp->t_inpcb->inp_flags2 |=
tp0->t_inpcb->inp_flags2 & INP2_KEEPALIVE_OFFLOAD;
socket_unlock(oso, 1);
tcp_set_max_rwinscale(tp, so, ifp);
#if CONTENT_FILTER
if (check_cfil) {
int error = cfil_sock_attach(so2, (struct sockaddr*)&to2, (struct sockaddr*)&from,
CFS_CONNECTION_DIR_IN);
if (error != 0) {
TCP_LOG_DROP_PCB(TCP_LOG_HDR, th, tp, false, " cfil_sock_attach failed");
goto drop;
}
}
#endif
KERNEL_DEBUG(DBG_FNC_TCP_NEWCONN | DBG_FUNC_END, 0, 0, 0, 0, 0);
}
}
socket_lock_assert_owned(so);
if (net_mpklog_enabled && (m->m_pkthdr.rcvif->if_xflags & IFXF_MPK_LOG)) {
MPKL_TCP_INPUT(tcp_mpkl_log_object,
ntohs(tp->t_inpcb->inp_lport), ntohs(tp->t_inpcb->inp_fport),
th->th_seq, th->th_ack, tlen, thflags,
so->last_pid, so->so_log_seqn++);
}
if (tp->t_state == TCPS_ESTABLISHED && tlen > 0) {
if ((tp->t_flagsext & TF_RCVUNACK_WAITSS) != 0) {
TCP_INC_VAR(tp->rcv_waitforss, nlropkts);
}
if (tcp_stretch_ack_enable(tp, thflags)) {
tp->t_flags |= TF_STRETCHACK;
tp->t_flagsext &= ~(TF_RCVUNACK_WAITSS);
tp->rcv_waitforss = 0;
} else {
tp->t_flags &= ~(TF_STRETCHACK);
}
if (TSTMP_GT(tp->rcv_unackwin - (tcp_rcvunackwin >> 1), tcp_now)) {
tp->rcv_by_unackhalfwin += (tlen + off);
tp->rcv_by_unackwin += (tlen + off);
} else {
tp->rcv_unackwin = tcp_now + tcp_rcvunackwin;
tp->rcv_by_unackwin = tp->rcv_by_unackhalfwin + tlen + off;
tp->rcv_by_unackhalfwin = tlen + off;
}
}
if ((pktf_sw_lro_pkt) && (nlropkts > 2)) {
tp->t_lropktlen += tlen;
}
if (ip_ecn == IPTOS_ECN_CE && tp->t_state == TCPS_ESTABLISHED &&
TCP_ECN_ENABLED(tp) && tlen > 0 &&
SEQ_GEQ(th->th_seq, tp->last_ack_sent) &&
SEQ_LT(th->th_seq, tp->last_ack_sent + tp->rcv_wnd)) {
tp->t_ecn_recv_ce++;
tcpstat.tcps_ecn_recv_ce++;
INP_INC_IFNET_STAT(inp, ecn_recv_ce);
tp->ecn_flags |= TE_RECV_ECN_CE;
tp->ecn_flags |= TE_SENDECE;
}
if (thflags & TH_CWR) {
tp->ecn_flags &= ~TE_SENDECE;
tp->t_ecn_recv_cwr++;
}
if (tp->t_state == TCPS_ESTABLISHED
&& (tp->ecn_flags & TE_SETUPSENT)
&& (ip_ecn == IPTOS_ECN_CE || (thflags & TH_CWR))) {
tcp_reset_stretch_ack(tp);
CLEAR_IAJ_STATE(tp);
}
if (ip_ecn == IPTOS_ECN_CE && tp->t_state == TCPS_ESTABLISHED &&
!TCP_ECN_ENABLED(tp) && !(tp->ecn_flags & TE_CEHEURI_SET)) {
tcpstat.tcps_ecn_fallback_ce++;
tcp_heuristic_ecn_aggressive(tp);
tp->ecn_flags |= TE_CEHEURI_SET;
}
if (tp->t_state == TCPS_ESTABLISHED && TCP_ECN_ENABLED(tp) &&
ip_ecn == IPTOS_ECN_CE && !(tp->ecn_flags & TE_CEHEURI_SET)) {
if (inp->inp_stat->rxpackets < ECN_MIN_CE_PROBES) {
tp->t_ecn_recv_ce_pkt++;
} else if (tp->t_ecn_recv_ce_pkt > ECN_MAX_CE_RATIO) {
tcpstat.tcps_ecn_fallback_ce++;
tcp_heuristic_ecn_aggressive(tp);
tp->ecn_flags |= TE_CEHEURI_SET;
INP_INC_IFNET_STAT(inp, ecn_fallback_ce);
} else {
tp->ecn_flags |= TE_CEHEURI_SET;
}
}
if (sw_lro && (tp->t_flagsext & TF_LRO_OFFLOADED) &&
((tcp_now - tp->t_rcvtime) >= (TCP_IDLETIMEOUT(tp)))) {
turnoff_lro = 1;
}
tp->t_rcvtime = tcp_now;
if (TCPS_HAVEESTABLISHED(tp->t_state)) {
tcp_keepalive_reset(tp);
if (tp->t_mpsub) {
mptcp_reset_keepalive(tp);
}
}
if (tp->t_state != TCPS_LISTEN && optp) {
tcp_dooptions(tp, optp, optlen, th, &to);
}
#if MPTCP
if (tp->t_state != TCPS_LISTEN && (so->so_flags & SOF_MP_SUBFLOW) &&
mptcp_input_preproc(tp, m, th, drop_hdrlen) != 0) {
tp->t_flags |= TF_ACKNOW;
(void) tcp_output(tp);
tcp_check_timer_state(tp);
socket_unlock(so, 1);
KERNEL_DEBUG(DBG_FNC_TCP_INPUT |
DBG_FUNC_END, 0, 0, 0, 0, 0);
return;
}
#endif
if (tp->t_state == TCPS_SYN_SENT && (thflags & TH_SYN)) {
if (!(thflags & TH_ACK) ||
(SEQ_GT(th->th_ack, tp->iss) &&
SEQ_LEQ(th->th_ack, tp->snd_max))) {
tcp_finalize_options(tp, &to, ifscope);
}
}
#if TRAFFIC_MGT
if (tp->t_state == TCPS_ESTABLISHED &&
(thflags & (TH_SYN | TH_FIN | TH_RST | TH_URG | TH_ACK | TH_ECE | TH_PUSH)) == TH_ACK &&
((tp->t_flags & (TF_NEEDSYN | TF_NEEDFIN)) == 0) &&
((to.to_flags & TOF_TS) == 0 ||
TSTMP_GEQ(to.to_tsval, tp->ts_recent)) &&
th->th_seq == tp->rcv_nxt && LIST_EMPTY(&tp->t_segq)) {
int seg_size = tlen;
if (tp->iaj_pktcnt <= IAJ_IGNORE_PKTCNT) {
TCP_INC_VAR(tp->iaj_pktcnt, nlropkts);
}
if (m->m_pkthdr.pkt_flags & PKTF_SW_LRO_PKT) {
seg_size = m->m_pkthdr.lro_pktlen;
}
if (tp->iaj_size == 0 || seg_size > tp->iaj_size ||
(seg_size == tp->iaj_size && tp->iaj_rcv_ts == 0)) {
update_iaj_state(tp, seg_size, 0);
} else {
if (seg_size == tp->iaj_size) {
if (pktf_sw_lro_pkt) {
compute_iaj(tp, nlropkts,
m->m_pkthdr.lro_elapsed);
} else {
compute_iaj(tp, 1, 0);
}
}
if (seg_size < tp->iaj_size) {
TCP_INC_VAR(tp->iaj_small_pkt, nlropkts);
if (tp->iaj_small_pkt > RESET_IAJ_SIZE_THRESH) {
update_iaj_state(tp, seg_size, 1);
} else {
CLEAR_IAJ_STATE(tp);
}
} else {
update_iaj_state(tp, seg_size, 0);
}
}
} else {
CLEAR_IAJ_STATE(tp);
}
#endif
if (tp->t_state == TCPS_ESTABLISHED &&
(thflags & (TH_SYN | TH_FIN | TH_RST | TH_URG | TH_ACK | TH_ECE | TH_CWR)) == TH_ACK &&
((tp->t_flags & (TF_NEEDSYN | TF_NEEDFIN)) == 0) &&
((to.to_flags & TOF_TS) == 0 ||
TSTMP_GEQ(to.to_tsval, tp->ts_recent)) &&
th->th_seq == tp->rcv_nxt &&
tiwin && tiwin == tp->snd_wnd &&
tp->snd_nxt == tp->snd_max) {
if ((to.to_flags & TOF_TS) != 0 &&
SEQ_LEQ(th->th_seq, tp->last_ack_sent)) {
tp->ts_recent_age = tcp_now;
tp->ts_recent = to.to_tsval;
}
if (tlen == 0) {
if (SEQ_GT(th->th_ack, tp->snd_una) &&
SEQ_LEQ(th->th_ack, tp->snd_max) &&
tp->snd_cwnd >= tp->snd_ssthresh &&
(!IN_FASTRECOVERY(tp) &&
((!(SACK_ENABLED(tp)) &&
tp->t_dupacks < tp->t_rexmtthresh) ||
(SACK_ENABLED(tp) && to.to_nsacks == 0 &&
TAILQ_EMPTY(&tp->snd_holes))))) {
++tcpstat.tcps_predack;
tcp_bad_rexmt_check(tp, th, &to);
tcp_compute_rtt(tp, &to, th);
VERIFY(SEQ_GEQ(th->th_ack, tp->snd_una));
acked = BYTES_ACKED(th, tp);
tcpstat.tcps_rcvackpack++;
tcpstat.tcps_rcvackbyte += acked;
if (CC_ALGO(tp)->congestion_avd != NULL) {
CC_ALGO(tp)->congestion_avd(tp, th);
}
tcp_ccdbg_trace(tp, th, TCP_CC_INSEQ_ACK_RCVD);
sbdrop(&so->so_snd, acked);
if (so->so_flags & SOF_ENABLE_MSGS) {
VERIFY(acked <= so->so_msg_state->msg_serial_bytes);
so->so_msg_state->msg_serial_bytes -= acked;
}
tcp_sbsnd_trim(&so->so_snd);
if (SEQ_GT(tp->snd_una, tp->snd_recover) &&
SEQ_LEQ(th->th_ack, tp->snd_recover)) {
tp->snd_recover = th->th_ack - 1;
}
tp->snd_una = th->th_ack;
TCP_RESET_REXMT_STATE(tp);
tp->snd_wl2 = th->th_ack;
if (tp->t_dupacks > 0) {
tp->t_dupacks = 0;
tp->t_rexmtthresh = tcprexmtthresh;
}
if (tp->snd_una == tp->snd_max) {
tp->t_timer[TCPT_REXMT] = 0;
tp->t_timer[TCPT_PTO] = 0;
} else if (tp->t_timer[TCPT_PERSIST] == 0) {
tp->t_timer[TCPT_REXMT] =
OFFSET_FROM_START(tp,
tp->t_rxtcur);
}
if (!SLIST_EMPTY(&tp->t_rxt_segments) &&
!TCP_DSACK_SEQ_IN_WINDOW(tp,
tp->t_dsack_lastuna, tp->snd_una)) {
tcp_rxtseg_clean(tp);
}
if ((tp->t_flagsext & TF_MEASURESNDBW) != 0 &&
tp->t_bwmeas != NULL) {
tcp_bwmeas_check(tp);
}
write_wakeup = 1;
if (!SLIST_EMPTY(&tp->t_notify_ack)) {
tcp_notify_acknowledgement(tp, so);
}
if ((so->so_snd.sb_cc) || (tp->t_flags & TF_ACKNOW)) {
(void) tcp_output(tp);
}
tcp_tfo_rcv_ack(tp, th);
m_freem(m);
tcp_check_timer_state(tp);
tcp_handle_wakeup(so, read_wakeup, write_wakeup);
socket_unlock(so, 1);
KERNEL_DEBUG(DBG_FNC_TCP_INPUT | DBG_FUNC_END, 0, 0, 0, 0, 0);
return;
}
} else if (th->th_ack == tp->snd_una &&
LIST_EMPTY(&tp->t_segq) &&
tlen <= tcp_sbspace(tp)) {
if (turnoff_lro) {
tcp_lro_remove_state(tp->t_inpcb->inp_laddr,
tp->t_inpcb->inp_faddr,
tp->t_inpcb->inp_lport,
tp->t_inpcb->inp_fport);
tp->t_flagsext &= ~TF_LRO_OFFLOADED;
tp->t_idleat = tp->rcv_nxt;
} else if (sw_lro && !pktf_sw_lro_pkt && !isipv6 &&
(so->so_flags & SOF_USELRO) &&
!IFNET_IS_CELLULAR(m->m_pkthdr.rcvif) &&
(m->m_pkthdr.rcvif->if_type != IFT_LOOP) &&
((th->th_seq - tp->irs) >
(tp->t_maxseg << lro_start)) &&
((tp->t_idleat == 0) || ((th->th_seq -
tp->t_idleat) > (tp->t_maxseg << lro_start)))) {
tp->t_flagsext |= TF_LRO_OFFLOADED;
tcp_start_coalescing(ip, th, tlen);
tp->t_idleat = 0;
}
if (SACK_ENABLED(tp) && tp->rcv_numsacks) {
tcp_clean_sackreport(tp);
}
++tcpstat.tcps_preddat;
tp->rcv_nxt += tlen;
tp->snd_wl1 = th->th_seq;
tp->rcv_up = tp->rcv_nxt;
TCP_INC_VAR(tcpstat.tcps_rcvpack, nlropkts);
tcpstat.tcps_rcvbyte += tlen;
if (nstat_collect) {
if (m->m_pkthdr.pkt_flags & PKTF_SW_LRO_PKT) {
INP_ADD_STAT(inp, cell, wifi, wired,
rxpackets, m->m_pkthdr.lro_npkts);
} else {
INP_ADD_STAT(inp, cell, wifi, wired,
rxpackets, 1);
}
INP_ADD_STAT(inp, cell, wifi, wired, rxbytes,
tlen);
inp_set_activity_bitmap(inp);
}
if (tp->t_flags & TF_STREAMING_ON) {
tcp_compute_rtt(tp, &to, th);
}
tcp_sbrcv_grow(tp, &so->so_rcv, &to, tlen,
TCP_AUTORCVBUF_MAX(ifp));
so_recv_data_stat(so, m, 0);
m_adj(m, drop_hdrlen);
if (isipv6) {
memcpy(&saved_hdr, ip6, sizeof(struct ip6_hdr));
ip6 = (struct ip6_hdr *)&saved_hdr[0];
} else {
memcpy(&saved_hdr, ip, ip->ip_hl << 2);
ip = (struct ip *)&saved_hdr[0];
}
memcpy(&saved_tcphdr, th, sizeof(struct tcphdr));
if (sbappendstream_rcvdemux(so, m,
th->th_seq - (tp->irs + 1), 0)) {
mptcp_handle_input(so);
read_wakeup = 1;
}
th = &saved_tcphdr;
#if INET6
if (isipv6) {
KERNEL_DEBUG(DBG_LAYER_END, ((th->th_dport << 16) | th->th_sport),
(((ip6->ip6_src.s6_addr16[0]) << 16) | (ip6->ip6_dst.s6_addr16[0])),
th->th_seq, th->th_ack, th->th_win);
} else
#endif
{
KERNEL_DEBUG(DBG_LAYER_END, ((th->th_dport << 16) | th->th_sport),
(((ip->ip_src.s_addr & 0xffff) << 16) | (ip->ip_dst.s_addr & 0xffff)),
th->th_seq, th->th_ack, th->th_win);
}
TCP_INC_VAR(tp->t_unacksegs, nlropkts);
if (DELAY_ACK(tp, th)) {
if ((tp->t_flags & TF_DELACK) == 0) {
tp->t_flags |= TF_DELACK;
tp->t_timer[TCPT_DELACK] = OFFSET_FROM_START(tp, tcp_delack);
}
} else {
tp->t_flags |= TF_ACKNOW;
tcp_output(tp);
}
tcp_adaptive_rwtimo_check(tp, tlen);
if (tlen > 0) {
tcp_tfo_rcv_data(tp);
}
tcp_check_timer_state(tp);
tcp_handle_wakeup(so, read_wakeup, write_wakeup);
socket_unlock(so, 1);
KERNEL_DEBUG(DBG_FNC_TCP_INPUT | DBG_FUNC_END, 0, 0, 0, 0, 0);
return;
}
}
socket_lock_assert_owned(so);
win = tcp_sbspace(tp);
if (win < 0) {
win = 0;
} else {
if (tp->t_flags & TF_SLOWLINK && slowlink_wsize > 0) {
win = min(win, slowlink_wsize);
}
}
tp->rcv_wnd = imax(win, (int)(tp->rcv_adv - tp->rcv_nxt));
#if MPTCP
if ((tp->t_mpflags & TMPF_MPTCP_TRUE) &&
(mp_tp = tptomptp(tp))) {
socket_lock_assert_owned(mptetoso(mp_tp->mpt_mpte));
if (tp->rcv_wnd > (int)(mp_tp->mpt_rcvadv - (uint32_t)mp_tp->mpt_rcvnxt)) {
tp->rcv_wnd = mp_tp->mpt_rcvadv - (uint32_t)mp_tp->mpt_rcvnxt;
tcpstat.tcps_mp_reducedwin++;
}
}
#endif
switch (tp->t_state) {
case TCPS_LISTEN: {
struct sockaddr_in *sin;
#if INET6
struct sockaddr_in6 *sin6;
#endif
socket_lock_assert_owned(so);
tp->t_log_flags = 0;
tp->t_flagsext &= ~TF_LOGGED_CONN_SUMMARY;
#if INET6
if (isipv6) {
MALLOC(sin6, struct sockaddr_in6 *, sizeof *sin6,
M_SONAME, M_NOWAIT);
if (sin6 == NULL) {
TCP_LOG_DROP_PCB(TCP_LOG_HDR, th, tp, false, "LISTEN malloc M_SONAME failed");
goto drop;
}
bzero(sin6, sizeof(*sin6));
sin6->sin6_family = AF_INET6;
sin6->sin6_len = sizeof(*sin6);
sin6->sin6_addr = ip6->ip6_src;
sin6->sin6_port = th->th_sport;
laddr6 = inp->in6p_laddr;
if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
inp->in6p_laddr = ip6->ip6_dst;
}
if (in6_pcbconnect(inp, (struct sockaddr *)sin6,
proc0)) {
inp->in6p_laddr = laddr6;
FREE(sin6, M_SONAME);
TCP_LOG_DROP_PCB(TCP_LOG_HDR, th, tp, false, " LISTEN in6_pcbconnect failed");
goto drop;
}
FREE(sin6, M_SONAME);
} else
#endif
{
socket_lock_assert_owned(so);
MALLOC(sin, struct sockaddr_in *, sizeof *sin, M_SONAME,
M_NOWAIT);
if (sin == NULL) {
TCP_LOG_DROP_PCB(TCP_LOG_HDR, th, tp, false, "LISTEN malloc M_SONAME failed");
goto drop;
}
sin->sin_family = AF_INET;
sin->sin_len = sizeof(*sin);
sin->sin_addr = ip->ip_src;
sin->sin_port = th->th_sport;
bzero((caddr_t)sin->sin_zero, sizeof(sin->sin_zero));
laddr = inp->inp_laddr;
if (inp->inp_laddr.s_addr == INADDR_ANY) {
inp->inp_laddr = ip->ip_dst;
}
if (in_pcbconnect(inp, (struct sockaddr *)sin, proc0,
IFSCOPE_NONE, NULL)) {
inp->inp_laddr = laddr;
FREE(sin, M_SONAME);
TCP_LOG_DROP_PCB(TCP_LOG_HDR, th, tp, false, " LISTEN in_pcbconnect failed");
goto drop;
}
FREE(sin, M_SONAME);
}
tcp_dooptions(tp, optp, optlen, th, &to);
tcp_finalize_options(tp, &to, ifscope);
if (tfo_enabled(tp) && tcp_tfo_syn(tp, &to)) {
isconnected = TRUE;
}
if (iss) {
tp->iss = iss;
} else {
tp->iss = tcp_new_isn(tp);
}
tp->irs = th->th_seq;
tcp_sendseqinit(tp);
tcp_rcvseqinit(tp);
tp->snd_recover = tp->snd_una;
tp->snd_wnd = tiwin;
tp->max_sndwnd = tp->snd_wnd;
tp->t_flags |= TF_ACKNOW;
tp->t_unacksegs = 0;
DTRACE_TCP4(state__change, void, NULL, struct inpcb *, inp,
struct tcpcb *, tp, int32_t, TCPS_SYN_RECEIVED);
tp->t_state = TCPS_SYN_RECEIVED;
tp->t_timer[TCPT_KEEP] = OFFSET_FROM_START(tp,
TCP_CONN_KEEPINIT(tp));
tp->t_connect_time = tcp_now;
dropsocket = 0;
if (inp->inp_flowhash == 0) {
inp->inp_flowhash = inp_calc_flowhash(inp);
}
#if INET6
if (inp->inp_flow == 0 &&
inp->in6p_flags & IN6P_AUTOFLOWLABEL) {
inp->inp_flow &= ~IPV6_FLOWLABEL_MASK;
inp->inp_flow |=
(htonl(inp->inp_flowhash) & IPV6_FLOWLABEL_MASK);
}
#endif
so->so_flags &= ~(SOF_INCOMP_INPROGRESS);
tcpstat.tcps_accepts++;
if ((thflags & (TH_ECE | TH_CWR)) == (TH_ECE | TH_CWR)) {
tp->ecn_flags |= (TE_SETUPRECEIVED | TE_SENDIPECT);
}
TCP_LOG_CONNECT(tp, false, 0);
goto trimthenstep6;
}
case TCPS_SYN_RECEIVED:
if ((thflags & TH_ACK) &&
(SEQ_LEQ(th->th_ack, tp->snd_una) ||
SEQ_GT(th->th_ack, tp->snd_max))) {
rstreason = BANDLIM_RST_OPENPORT;
IF_TCP_STATINC(ifp, ooopacket);
TCP_LOG_DROP_PCB(TCP_LOG_HDR, th, tp, false, "SYN_RECEIVED bad ACK");
goto dropwithreset;
}
if ((thflags & TH_SYN) &&
(tp->irs == th->th_seq) &&
!(to.to_flags & TOF_SCALE)) {
tp->t_flags &= ~TF_RCVD_SCALE;
}
break;
case TCPS_SYN_SENT:
if ((thflags & TH_ACK) &&
(SEQ_LEQ(th->th_ack, tp->iss) ||
SEQ_GT(th->th_ack, tp->snd_max))) {
rstreason = BANDLIM_UNLIMITED;
IF_TCP_STATINC(ifp, ooopacket);
TCP_LOG_DROP_PCB(TCP_LOG_HDR, th, tp, false, "SYN_SENT bad ACK");
goto dropwithreset;
}
if (thflags & TH_RST) {
if ((thflags & TH_ACK) != 0) {
if (tfo_enabled(tp) &&
!(tp->t_flagsext & TF_FASTOPEN_FORCE_ENABLE)) {
tcp_heuristic_tfo_rst(tp);
}
if ((tp->ecn_flags & (TE_SETUPSENT | TE_RCVD_SYN_RST)) == TE_SETUPSENT) {
if (tp->t_flags & TF_LOCAL) {
tp->ecn_flags |= TE_RCVD_SYN_RST;
goto drop;
} else {
tcp_heuristic_ecn_synrst(tp);
}
}
soevent(so,
(SO_FILT_HINT_LOCKED |
SO_FILT_HINT_CONNRESET));
tp = tcp_drop(tp, ECONNREFUSED);
postevent(so, 0, EV_RESET);
}
TCP_LOG_DROP_PCB(TCP_LOG_HDR, th, tp, false, "SYN_SENT got RST");
goto drop;
}
if ((thflags & TH_SYN) == 0) {
TCP_LOG_DROP_PCB(TCP_LOG_HDR, th, tp, false, "SYN_SENT no SYN");
goto drop;
}
tp->snd_wnd = th->th_win;
tp->max_sndwnd = tp->snd_wnd;
tp->irs = th->th_seq;
tcp_rcvseqinit(tp);
if (thflags & TH_ACK) {
tcpstat.tcps_connects++;
if ((thflags & (TH_ECE | TH_CWR)) == (TH_ECE)) {
tp->ecn_flags |= TE_SETUPRECEIVED;
if (TCP_ECN_ENABLED(tp)) {
tcp_heuristic_ecn_success(tp);
tcpstat.tcps_ecn_client_success++;
}
} else {
if (tp->ecn_flags & TE_SETUPSENT &&
tp->t_rxtshift == 0) {
tcp_heuristic_ecn_success(tp);
tcpstat.tcps_ecn_not_supported++;
}
if (tp->ecn_flags & TE_SETUPSENT &&
tp->t_rxtshift > 0) {
tcp_heuristic_ecn_loss(tp);
}
tp->ecn_flags &= ~TE_SENDIPECT;
}
#if CONFIG_MACF_NET && CONFIG_MACF_SOCKET
mac_socketpeer_label_associate_mbuf(m, so);
#endif
if (TCP_WINDOW_SCALE_ENABLED(tp)) {
tp->snd_scale = tp->requested_s_scale;
tp->rcv_scale = tp->request_r_scale;
}
tp->rcv_adv += min(tp->rcv_wnd, TCP_MAXWIN << tp->rcv_scale);
tp->snd_una++;
if (SEQ_LT(tp->snd_nxt, tp->snd_una)) {
tp->snd_nxt = tp->snd_una;
}
if (SEQ_GT(tp->snd_nxt, th->th_ack)) {
if (!(tp->t_flagsext & TF_FASTOPEN_FORCE_ENABLE) &&
(tp->t_tfo_stats & TFO_S_SYN_DATA_SENT) &&
tp->snd_max == th->th_ack + 1 &&
tp->snd_max > tp->snd_una + 1) {
tcp_heuristic_tfo_middlebox(tp);
so->so_error = ENODATA;
soevent(so,
(SO_FILT_HINT_LOCKED | SO_FILT_HINT_MP_SUB_ERROR));
tp->t_tfo_stats |= TFO_S_ONE_BYTE_PROXY;
}
tp->snd_max = tp->snd_nxt = th->th_ack;
}
TCP_INC_VAR(tp->t_unacksegs, nlropkts);
if (DELAY_ACK(tp, th) && tlen != 0) {
if ((tp->t_flags & TF_DELACK) == 0) {
tp->t_flags |= TF_DELACK;
tp->t_timer[TCPT_DELACK] = OFFSET_FROM_START(tp, tcp_delack);
}
} else {
tp->t_flags |= TF_ACKNOW;
}
tp->t_starttime = tcp_now;
tcp_sbrcv_tstmp_check(tp);
if (tp->t_flags & TF_NEEDFIN) {
DTRACE_TCP4(state__change, void, NULL,
struct inpcb *, inp,
struct tcpcb *, tp, int32_t,
TCPS_FIN_WAIT_1);
tp->t_state = TCPS_FIN_WAIT_1;
tp->t_flags &= ~TF_NEEDFIN;
thflags &= ~TH_SYN;
TCP_LOG_CONNECTION_SUMMARY(tp);
} else {
DTRACE_TCP4(state__change, void, NULL,
struct inpcb *, inp, struct tcpcb *,
tp, int32_t, TCPS_ESTABLISHED);
tp->t_state = TCPS_ESTABLISHED;
tp->t_timer[TCPT_KEEP] =
OFFSET_FROM_START(tp,
TCP_CONN_KEEPIDLE(tp));
if (nstat_collect) {
nstat_route_connect_success(
inp->inp_route.ro_rt);
}
inp_count_sndbytes(inp, th->th_ack);
}
#if MPTCP
if ((!(tp->t_mpflags & TMPF_MPTCP_TRUE)) &&
(tp->t_mpflags & TMPF_SENT_JOIN)) {
isconnected = FALSE;
} else
#endif
isconnected = TRUE;
if ((tp->t_tfo_flags & (TFO_F_COOKIE_REQ | TFO_F_COOKIE_SENT)) ||
(tp->t_tfo_stats & TFO_S_SYN_DATA_SENT)) {
tcp_tfo_synack(tp, &to);
if ((tp->t_tfo_stats & TFO_S_SYN_DATA_SENT) &&
SEQ_LT(tp->snd_una, th->th_ack)) {
tp->t_tfo_stats |= TFO_S_SYN_DATA_ACKED;
tcpstat.tcps_tfo_syn_data_acked++;
#if MPTCP
if (so->so_flags & SOF_MP_SUBFLOW) {
so->so_flags1 |= SOF1_TFO_REWIND;
}
#endif
tcp_tfo_rcv_probe(tp, tlen);
}
}
} else {
tp->t_flags |= TF_ACKNOW;
tp->t_timer[TCPT_REXMT] = 0;
DTRACE_TCP4(state__change, void, NULL, struct inpcb *, inp,
struct tcpcb *, tp, int32_t, TCPS_SYN_RECEIVED);
tp->t_state = TCPS_SYN_RECEIVED;
tcp_disable_tfo(tp);
}
trimthenstep6:
th->th_seq++;
if (tlen > tp->rcv_wnd) {
todrop = tlen - tp->rcv_wnd;
m_adj(m, -todrop);
tlen = tp->rcv_wnd;
thflags &= ~TH_FIN;
tcpstat.tcps_rcvpackafterwin++;
tcpstat.tcps_rcvbyteafterwin += todrop;
}
tp->snd_wl1 = th->th_seq - 1;
tp->rcv_up = th->th_seq;
if (thflags & TH_ACK) {
goto process_ACK;
}
goto step6;
case TCPS_LAST_ACK:
case TCPS_CLOSING:
case TCPS_TIME_WAIT:
break;
case TCPS_ESTABLISHED:
if (thflags & TH_SYN) {
if (tcp_do_rfc5961 && tcp_is_ack_ratelimited(tp)) {
TCP_LOG_DROP_PCB(TCP_LOG_HDR, th, tp, false, "ESTABLISHED rfc5961 rate limited");
goto drop;
} else {
tcpstat.tcps_synchallenge++;
TCP_LOG_DROP_PCB(TCP_LOG_HDR, th, tp, false, "ESTABLISHED rfc5961 challenge ACK");
goto dropafterack;
}
}
break;
}
if (thflags & TH_RST) {
if ((SEQ_GEQ(th->th_seq, tp->last_ack_sent) &&
SEQ_LT(th->th_seq, tp->last_ack_sent + tp->rcv_wnd)) ||
(tp->rcv_wnd == 0 &&
((tp->last_ack_sent == th->th_seq) ||
((tp->last_ack_sent - 1) == th->th_seq)))) {
if (tcp_do_rfc5961 == 0 || tp->last_ack_sent == th->th_seq) {
switch (tp->t_state) {
case TCPS_SYN_RECEIVED:
IF_TCP_STATINC(ifp, rstinsynrcv);
so->so_error = ECONNREFUSED;
goto close;
case TCPS_ESTABLISHED:
if (tcp_do_rfc5961 == 0 && tp->last_ack_sent != th->th_seq) {
tcpstat.tcps_badrst++;
TCP_LOG_DROP_PCB(TCP_LOG_HDR, th, tp, false, "ESTABLISHED rfc5961 bad RST");
goto drop;
}
if (TCP_ECN_ENABLED(tp) &&
tp->snd_una == tp->iss + 1 &&
SEQ_GT(tp->snd_max, tp->snd_una)) {
tcp_heuristic_ecn_droprst(tp);
}
case TCPS_FIN_WAIT_1:
case TCPS_CLOSE_WAIT:
case TCPS_FIN_WAIT_2:
so->so_error = ECONNRESET;
close:
postevent(so, 0, EV_RESET);
soevent(so,
(SO_FILT_HINT_LOCKED |
SO_FILT_HINT_CONNRESET));
tcpstat.tcps_drops++;
tp = tcp_close(tp);
break;
case TCPS_CLOSING:
case TCPS_LAST_ACK:
tp = tcp_close(tp);
break;
case TCPS_TIME_WAIT:
break;
}
} else if (tcp_do_rfc5961) {
tcpstat.tcps_badrst++;
if (tcp_is_ack_ratelimited(tp)) {
TCP_LOG_DROP_PCB(TCP_LOG_HDR, th, tp, false, "ESTABLISHED rfc5961 rate limited");
goto drop;
} else {
tcpstat.tcps_rstchallenge++;
TCP_LOG_DROP_PCB(TCP_LOG_HDR, th, tp, false, "ESTABLISHED rfc5961 challenge ACK");
goto dropafterack;
}
}
}
goto drop;
}
if ((to.to_flags & TOF_TS) != 0 && tp->ts_recent &&
TSTMP_LT(to.to_tsval, tp->ts_recent)) {
if ((int)(tcp_now - tp->ts_recent_age) > TCP_PAWS_IDLE) {
tp->ts_recent = 0;
} else {
tcpstat.tcps_rcvduppack++;
tcpstat.tcps_rcvdupbyte += tlen;
tp->t_pawsdrop++;
tcpstat.tcps_pawsdrop++;
if (TCP_ECN_ENABLED(tp) &&
inp->inp_stat->rxbytes < 2147483648) {
INP_INC_IFNET_STAT(inp, ecn_fallback_reorder);
tcpstat.tcps_ecn_fallback_reorder++;
tcp_heuristic_ecn_aggressive(tp);
}
if (nstat_collect) {
nstat_route_rx(tp->t_inpcb->inp_route.ro_rt,
1, tlen, NSTAT_RX_FLAG_DUPLICATE);
INP_ADD_STAT(inp, cell, wifi, wired,
rxpackets, 1);
INP_ADD_STAT(inp, cell, wifi, wired,
rxbytes, tlen);
tp->t_stat.rxduplicatebytes += tlen;
inp_set_activity_bitmap(inp);
}
if (tlen > 0) {
goto dropafterack;
}
goto drop;
}
}
if (tp->t_state == TCPS_SYN_RECEIVED && SEQ_LT(th->th_seq, tp->irs)) {
rstreason = BANDLIM_RST_OPENPORT;
IF_TCP_STATINC(ifp, dospacket);
TCP_LOG_DROP_PCB(TCP_LOG_HDR, th, tp, false, "SYN_RECEIVED bad SEQ");
goto dropwithreset;
}
todrop = tp->rcv_nxt - th->th_seq;
if (todrop > 0) {
boolean_t is_syn_set = FALSE;
if (thflags & TH_SYN) {
is_syn_set = TRUE;
thflags &= ~TH_SYN;
th->th_seq++;
if (th->th_urp > 1) {
th->th_urp--;
} else {
thflags &= ~TH_URG;
}
todrop--;
}
if (todrop > tlen
|| (todrop == tlen && (thflags & TH_FIN) == 0)) {
thflags &= ~TH_FIN;
if (tcp_do_rfc5961 && is_syn_set) {
if (!tcp_is_ack_ratelimited(tp)) {
tcpstat.tcps_synchallenge++;
tp->t_flags |= TF_ACKNOW;
}
} else {
tp->t_flags |= TF_ACKNOW;
}
if (todrop == 1) {
soevent(so, SO_FILT_HINT_LOCKED |
SO_FILT_HINT_KEEPALIVE);
}
todrop = tlen;
tcpstat.tcps_rcvduppack++;
tcpstat.tcps_rcvdupbyte += todrop;
} else {
tcpstat.tcps_rcvpartduppack++;
tcpstat.tcps_rcvpartdupbyte += todrop;
}
if (TCP_DSACK_ENABLED(tp) && todrop > 1) {
tp->t_dsack_lseq = th->th_seq;
tp->t_dsack_rseq = th->th_seq + todrop;
tp->t_flags |= TF_ACKNOW;
}
if (nstat_collect) {
nstat_route_rx(tp->t_inpcb->inp_route.ro_rt, 1,
todrop, NSTAT_RX_FLAG_DUPLICATE);
INP_ADD_STAT(inp, cell, wifi, wired, rxpackets, 1);
INP_ADD_STAT(inp, cell, wifi, wired, rxbytes, todrop);
tp->t_stat.rxduplicatebytes += todrop;
inp_set_activity_bitmap(inp);
}
drop_hdrlen += todrop;
th->th_seq += todrop;
tlen -= todrop;
if (th->th_urp > todrop) {
th->th_urp -= todrop;
} else {
thflags &= ~TH_URG;
th->th_urp = 0;
}
}
if (tlen) {
boolean_t close_it = FALSE;
if (!(so->so_flags & SOF_MP_SUBFLOW) && (so->so_state & SS_NOFDREF) &&
tp->t_state > TCPS_CLOSE_WAIT) {
TCP_LOG_DROP_PCB(TCP_LOG_HDR, th, tp, false, "SS_NOFDREF");
close_it = TRUE;
}
if ((so->so_flags & SOF_MP_SUBFLOW) && (mptetoso(tptomptp(tp)->mpt_mpte)->so_state & SS_NOFDREF) &&
tp->t_state > TCPS_CLOSE_WAIT) {
TCP_LOG_DROP_PCB(TCP_LOG_HDR, th, tp, false, "SOF_MP_SUBFLOW SS_NOFDREF");
close_it = TRUE;
}
if ((so->so_flags & SOF_DEFUNCT) && tp->t_state > TCPS_FIN_WAIT_1) {
TCP_LOG_DROP_PCB(TCP_LOG_HDR, th, tp, false, "SOF_DEFUNCT");
close_it = TRUE;
}
if (close_it) {
tp = tcp_close(tp);
tcpstat.tcps_rcvafterclose++;
rstreason = BANDLIM_UNLIMITED;
IF_TCP_STATINC(ifp, cleanup);
goto dropwithreset;
}
}
todrop = (th->th_seq + tlen) - (tp->rcv_nxt + tp->rcv_wnd);
if (todrop > 0) {
tcpstat.tcps_rcvpackafterwin++;
if (todrop >= tlen) {
tcpstat.tcps_rcvbyteafterwin += tlen;
if (thflags & TH_SYN &&
tp->t_state == TCPS_TIME_WAIT &&
SEQ_GT(th->th_seq, tp->rcv_nxt)) {
iss = tcp_new_isn(tp);
tp = tcp_close(tp);
socket_unlock(so, 1);
goto findpcb;
}
if (tp->rcv_wnd == 0 && th->th_seq == tp->rcv_nxt) {
tp->t_flags |= TF_ACKNOW;
tcpstat.tcps_rcvwinprobe++;
} else {
goto dropafterack;
}
} else {
tcpstat.tcps_rcvbyteafterwin += todrop;
}
m_adj(m, -todrop);
tlen -= todrop;
thflags &= ~(TH_PUSH | TH_FIN);
}
if ((to.to_flags & TOF_TS) != 0 &&
SEQ_LEQ(th->th_seq, tp->last_ack_sent) &&
SEQ_LEQ(tp->last_ack_sent, th->th_seq + tlen +
((thflags & (TH_SYN | TH_FIN)) != 0))) {
tp->ts_recent_age = tcp_now;
tp->ts_recent = to.to_tsval;
}
if (thflags & TH_SYN) {
if (tcp_do_rfc5961) {
tcpstat.tcps_badsyn++;
if (tcp_is_ack_ratelimited(tp)) {
TCP_LOG_DROP_PCB(TCP_LOG_HDR, th, tp, false, "rfc5961 bad SYN rate limited");
goto drop;
} else {
tcpstat.tcps_synchallenge++;
TCP_LOG_DROP_PCB(TCP_LOG_HDR, th, tp, false, "rfc5961 bad SYN challenge ack");
goto dropafterack;
}
} else {
tp = tcp_drop(tp, ECONNRESET);
rstreason = BANDLIM_UNLIMITED;
postevent(so, 0, EV_RESET);
IF_TCP_STATINC(ifp, synwindow);
TCP_LOG_DROP_PCB(TCP_LOG_HDR, th, tp, false, "bad SYN");
goto dropwithreset;
}
}
if ((thflags & TH_ACK) == 0) {
if (tp->t_state == TCPS_SYN_RECEIVED ||
(tp->t_flags & TF_NEEDSYN)) {
if ((tfo_enabled(tp))) {
tp->snd_nxt = tp->snd_una;
if (th->th_flags & TH_SYN) {
needoutput = 1;
}
}
goto step6;
} else if (tp->t_flags & TF_ACKNOW) {
TCP_LOG_DROP_PCB(TCP_LOG_HDR, th, tp, false, "bad ACK");
goto dropafterack;
} else {
TCP_LOG_DROP_PCB(TCP_LOG_HDR, th, tp, false, "bad ACK");
goto drop;
}
}
switch (tp->t_state) {
case TCPS_SYN_RECEIVED:
tcpstat.tcps_connects++;
if (TCP_WINDOW_SCALE_ENABLED(tp)) {
tp->snd_scale = tp->requested_s_scale;
tp->rcv_scale = tp->request_r_scale;
tp->snd_wnd = th->th_win << tp->snd_scale;
tp->max_sndwnd = tp->snd_wnd;
tiwin = tp->snd_wnd;
}
tp->t_starttime = tcp_now;
tcp_sbrcv_tstmp_check(tp);
if (tp->t_flags & TF_NEEDFIN) {
DTRACE_TCP4(state__change, void, NULL,
struct inpcb *, inp,
struct tcpcb *, tp, int32_t, TCPS_FIN_WAIT_1);
tp->t_state = TCPS_FIN_WAIT_1;
tp->t_flags &= ~TF_NEEDFIN;
TCP_LOG_CONNECTION_SUMMARY(tp);
} else {
DTRACE_TCP4(state__change, void, NULL,
struct inpcb *, inp,
struct tcpcb *, tp, int32_t, TCPS_ESTABLISHED);
tp->t_state = TCPS_ESTABLISHED;
tp->t_timer[TCPT_KEEP] = OFFSET_FROM_START(tp,
TCP_CONN_KEEPIDLE(tp));
if (nstat_collect) {
nstat_route_connect_success(
tp->t_inpcb->inp_route.ro_rt);
}
inp_count_sndbytes(inp, th->th_ack);
}
if (tlen == 0 && (thflags & TH_FIN) == 0) {
if (isipv6) {
memcpy(&saved_hdr, ip6, sizeof(struct ip6_hdr));
ip6 = (struct ip6_hdr *)&saved_hdr[0];
} else {
memcpy(&saved_hdr, ip, ip->ip_hl << 2);
ip = (struct ip *)&saved_hdr[0];
}
memcpy(&saved_tcphdr, th, sizeof(struct tcphdr));
(void) tcp_reass(tp, (struct tcphdr *)0, &tlen,
NULL, ifp, &read_wakeup);
th = &saved_tcphdr;
}
tp->snd_wl1 = th->th_seq - 1;
#if MPTCP
if ((!(tp->t_mpflags & TMPF_MPTCP_TRUE)) &&
(tp->t_mpflags & TMPF_SENT_JOIN)) {
isconnected = FALSE;
} else
#endif
isconnected = TRUE;
if ((tp->t_tfo_flags & TFO_F_COOKIE_VALID)) {
isconnected = FALSE;
OSDecrementAtomic(&tcp_tfo_halfcnt);
VERIFY(tcp_tfo_halfcnt >= 0);
tp->t_tfo_flags &= ~TFO_F_COOKIE_VALID;
}
if (so->so_snd.sb_cc) {
tp->snd_una++;
if (SEQ_LT(tp->snd_nxt, tp->snd_una)) {
tp->snd_nxt = tp->snd_una;
}
if (SACK_ENABLED(tp) &&
(to.to_nsacks > 0 ||
!TAILQ_EMPTY(&tp->snd_holes))) {
tcp_sack_doack(tp, &to, th,
&sack_bytes_acked);
}
goto process_ACK;
}
case TCPS_ESTABLISHED:
case TCPS_FIN_WAIT_1:
case TCPS_FIN_WAIT_2:
case TCPS_CLOSE_WAIT:
case TCPS_CLOSING:
case TCPS_LAST_ACK:
case TCPS_TIME_WAIT:
if (SEQ_GT(th->th_ack, tp->snd_max)) {
tcpstat.tcps_rcvacktoomuch++;
if (tcp_do_rfc5961 && tcp_is_ack_ratelimited(tp)) {
TCP_LOG_DROP_PCB(TCP_LOG_HDR, th, tp, false, "rfc5961 rcvacktoomuch");
goto drop;
} else {
goto dropafterack;
}
}
if (tcp_do_rfc5961 && SEQ_LT(th->th_ack, tp->snd_una - tp->max_sndwnd)) {
if (tcp_is_ack_ratelimited(tp)) {
TCP_LOG_DROP_PCB(TCP_LOG_HDR, th, tp, false, "rfc5961 bad ACK");
goto drop;
} else {
goto dropafterack;
}
}
if (SACK_ENABLED(tp) && to.to_nsacks > 0) {
recvd_dsack = tcp_sack_process_dsack(tp, &to, th);
if (recvd_dsack &&
SEQ_LEQ(th->th_ack, tp->snd_una) &&
to.to_nsacks == 0) {
tcp_bad_rexmt_check(tp, th, &to);
goto drop;
}
}
if (SACK_ENABLED(tp) &&
(to.to_nsacks > 0 || !TAILQ_EMPTY(&tp->snd_holes))) {
tcp_sack_doack(tp, &to, th, &sack_bytes_acked);
}
#if MPTCP
if (tp->t_mpuna && SEQ_GEQ(th->th_ack, tp->t_mpuna)) {
if (tp->t_mpflags & TMPF_PREESTABLISHED) {
tp->t_mpuna = 0;
if (tp->t_mpflags & TMPF_JOINED_FLOW) {
if (tp->t_mpflags & TMPF_SENT_JOIN) {
tp->t_mpflags &=
~TMPF_PREESTABLISHED;
tp->t_mpflags |=
TMPF_MPTCP_TRUE;
tp->t_timer[TCPT_JACK_RXMT] = 0;
tp->t_mprxtshift = 0;
isconnected = TRUE;
} else {
isconnected = FALSE;
}
} else {
isconnected = TRUE;
}
}
}
#endif
tcp_tfo_rcv_ack(tp, th);
if (SEQ_LEQ(th->th_ack, tp->snd_una)) {
if (tlen == 0 && (tiwin == tp->snd_wnd ||
(to.to_nsacks > 0 && sack_bytes_acked > 0))) {
if ((thflags & TH_FIN) &&
!TCPS_HAVERCVDFIN(tp->t_state)) {
break;
}
process_dupack:
#if MPTCP
if (to.to_flags & TOF_MPTCP) {
goto drop;
}
if ((isconnected) && (tp->t_mpflags & TMPF_JOINED_FLOW)) {
mptcplog((LOG_DEBUG, "MPTCP "
"Sockets: bypass ack recovery\n"),
MPTCP_SOCKET_DBG,
MPTCP_LOGLVL_VERBOSE);
break;
}
#endif
if (tp->ecn_flags & TE_INRECOVERY) {
tp->ecn_flags &= ~TE_INRECOVERY;
}
tcpstat.tcps_rcvdupack++;
++tp->t_dupacks;
if (tp->t_early_rexmt_count > 0 &&
TSTMP_GEQ(tcp_now,
(tp->t_early_rexmt_win +
TCP_EARLY_REXMT_WIN))) {
tp->t_early_rexmt_count = 0;
}
if (!IN_FASTRECOVERY(tp)) {
tcp_early_rexmt_check(tp, th);
}
if (tp->t_timer[TCPT_REXMT] == 0 ||
(th->th_ack != tp->snd_una
&& sack_bytes_acked == 0)) {
tp->t_dupacks = 0;
tp->t_rexmtthresh = tcprexmtthresh;
} else if (tp->t_dupacks > tp->t_rexmtthresh ||
IN_FASTRECOVERY(tp)) {
if (SACK_ENABLED(tp) && !IN_FASTRECOVERY(tp) &&
(tp->t_flagsext &
(TF_PKTS_REORDERED | TF_DELAY_RECOVERY)) ==
(TF_PKTS_REORDERED | TF_DELAY_RECOVERY)) {
break;
}
if (SACK_ENABLED(tp)
&& IN_FASTRECOVERY(tp)) {
int awnd;
awnd = (tp->snd_nxt - tp->snd_fack) +
tp->sackhint.sack_bytes_rexmit;
if (awnd < tp->snd_ssthresh) {
tp->snd_cwnd += tp->t_maxseg;
if (tp->snd_cwnd > tp->snd_ssthresh) {
tp->snd_cwnd = tp->snd_ssthresh;
}
}
} else {
tp->snd_cwnd += tp->t_maxseg;
}
if (tiwin > tp->snd_wnd) {
tcp_update_window(tp, thflags,
th, tiwin, tlen);
}
tcp_ccdbg_trace(tp, th,
TCP_CC_IN_FASTRECOVERY);
(void) tcp_output(tp);
goto drop;
} else if (tp->t_dupacks == tp->t_rexmtthresh) {
tcp_seq onxt = tp->snd_nxt;
if (SACK_ENABLED(tp)) {
if (IN_FASTRECOVERY(tp)) {
tp->t_dupacks = 0;
break;
} else if (tp->t_flagsext & TF_DELAY_RECOVERY) {
break;
}
} else {
if (SEQ_LEQ(th->th_ack,
tp->snd_recover)) {
tp->t_dupacks = 0;
break;
}
}
if (tp->t_flags & TF_SENTFIN) {
tp->snd_recover = tp->snd_max - 1;
} else {
tp->snd_recover = tp->snd_max;
}
tp->t_timer[TCPT_PTO] = 0;
tp->t_rtttime = 0;
if (SACK_ENABLED(tp) &&
(tp->t_flagsext &
(TF_PKTS_REORDERED | TF_DELAY_RECOVERY))
== TF_PKTS_REORDERED &&
!IN_FASTRECOVERY(tp) &&
tp->t_reorderwin > 0 &&
(tp->t_state == TCPS_ESTABLISHED ||
tp->t_state == TCPS_FIN_WAIT_1)) {
tp->t_timer[TCPT_DELAYFR] =
OFFSET_FROM_START(tp,
tp->t_reorderwin);
tp->t_flagsext |= TF_DELAY_RECOVERY;
tcpstat.tcps_delay_recovery++;
tcp_ccdbg_trace(tp, th,
TCP_CC_DELAY_FASTRECOVERY);
break;
}
tcp_rexmt_save_state(tp);
if (CC_ALGO(tp)->pre_fr != NULL) {
CC_ALGO(tp)->pre_fr(tp);
}
ENTER_FASTRECOVERY(tp);
tp->t_timer[TCPT_REXMT] = 0;
if (TCP_ECN_ENABLED(tp)) {
tp->ecn_flags |= TE_SENDCWR;
}
if (SACK_ENABLED(tp)) {
tcpstat.tcps_sack_recovery_episode++;
tp->t_sack_recovery_episode++;
tp->sack_newdata = tp->snd_nxt;
tp->snd_cwnd = tp->t_maxseg;
tp->t_flagsext &=
~TF_CWND_NONVALIDATED;
if (tiwin > tp->snd_wnd) {
tcp_update_window(
tp, thflags,
th, tiwin, tlen);
}
tcp_ccdbg_trace(tp, th,
TCP_CC_ENTER_FASTRECOVERY);
(void) tcp_output(tp);
goto drop;
}
tp->snd_nxt = th->th_ack;
tp->snd_cwnd = tp->t_maxseg;
if (tiwin > tp->snd_wnd) {
tcp_update_window(tp,
thflags,
th, tiwin, tlen);
}
(void) tcp_output(tp);
if (tp->t_flagsext & TF_CWND_NONVALIDATED) {
tcp_cc_adjust_nonvalidated_cwnd(tp);
} else {
tp->snd_cwnd = tp->snd_ssthresh +
tp->t_maxseg * tp->t_dupacks;
}
if (SEQ_GT(onxt, tp->snd_nxt)) {
tp->snd_nxt = onxt;
}
tcp_ccdbg_trace(tp, th,
TCP_CC_ENTER_FASTRECOVERY);
goto drop;
} else if (limited_txmt &&
ALLOW_LIMITED_TRANSMIT(tp) &&
(!(SACK_ENABLED(tp)) || sack_bytes_acked > 0) &&
(so->so_snd.sb_cc - (tp->snd_max - tp->snd_una)) > 0) {
u_int32_t incr = (tp->t_maxseg * tp->t_dupacks);
tp->snd_cwnd += incr;
tcpstat.tcps_limited_txt++;
(void) tcp_output(tp);
tcp_ccdbg_trace(tp, th, TCP_CC_LIMITED_TRANSMIT);
tp->snd_cwnd -= incr;
}
}
break;
}
if (IN_FASTRECOVERY(tp)) {
if (SEQ_LT(th->th_ack, tp->snd_recover)) {
if (tp->ecn_flags & TE_INRECOVERY) {
goto process_ACK;
}
if (SACK_ENABLED(tp)) {
tcp_sack_partialack(tp, th);
} else {
tcp_newreno_partial_ack(tp, th);
}
tcp_ccdbg_trace(tp, th, TCP_CC_PARTIAL_ACK);
} else {
EXIT_FASTRECOVERY(tp);
if (CC_ALGO(tp)->post_fr != NULL) {
CC_ALGO(tp)->post_fr(tp, th);
}
tp->t_pipeack = 0;
tcp_clear_pipeack_state(tp);
tcp_ccdbg_trace(tp, th,
TCP_CC_EXIT_FASTRECOVERY);
}
} else if ((tp->t_flagsext &
(TF_PKTS_REORDERED | TF_DELAY_RECOVERY))
== (TF_PKTS_REORDERED | TF_DELAY_RECOVERY)) {
if (SEQ_GEQ(th->th_ack, tp->snd_recover) ||
TAILQ_EMPTY(&tp->snd_holes)) {
tp->t_timer[TCPT_DELAYFR] = 0;
tp->t_flagsext &= ~TF_DELAY_RECOVERY;
EXIT_FASTRECOVERY(tp);
tcp_ccdbg_trace(tp, th,
TCP_CC_EXIT_FASTRECOVERY);
}
} else {
tp->t_dupacks = 0;
tp->t_rexmtthresh = tcprexmtthresh;
}
if (tp->t_flags & TF_NEEDSYN) {
tp->t_flags &= ~TF_NEEDSYN;
tp->snd_una++;
if (TCP_WINDOW_SCALE_ENABLED(tp)) {
tp->snd_scale = tp->requested_s_scale;
tp->rcv_scale = tp->request_r_scale;
}
}
process_ACK:
VERIFY(SEQ_GEQ(th->th_ack, tp->snd_una));
acked = BYTES_ACKED(th, tp);
tcpstat.tcps_rcvackpack++;
tcpstat.tcps_rcvackbyte += acked;
tcp_bad_rexmt_check(tp, th, &to);
tcp_compute_rtt(tp, &to, th);
TCP_RESET_REXMT_STATE(tp);
TCPT_RANGESET(tp->t_rxtcur, TCP_REXMTVAL(tp),
tp->t_rttmin, TCPTV_REXMTMAX,
TCP_ADD_REXMTSLOP(tp));
if (th->th_ack == tp->snd_max) {
tp->t_timer[TCPT_REXMT] = 0;
tp->t_timer[TCPT_PTO] = 0;
needoutput = 1;
} else if (tp->t_timer[TCPT_PERSIST] == 0) {
tp->t_timer[TCPT_REXMT] = OFFSET_FROM_START(tp,
tp->t_rxtcur);
}
if ((prev_t_state == TCPS_SYN_SENT ||
prev_t_state == TCPS_SYN_RECEIVED) &&
tp->t_state == TCPS_ESTABLISHED) {
TCP_LOG_RTT_INFO(tp);
}
if (acked == 0) {
goto step6;
}
if ((tp->t_tfo_stats & TFO_S_SYN_DATA_SENT) &&
!(tp->t_tfo_flags & TFO_F_NO_SNDPROBING) &&
!(th->th_flags & TH_SYN)) {
tp->t_tfo_flags |= TFO_F_NO_SNDPROBING;
}
if ((thflags & TH_ECE) != 0 && TCP_ECN_ENABLED(tp) &&
(tp->ecn_flags & TE_SENDIPECT)) {
if (!IN_FASTRECOVERY(tp)) {
tcp_reduce_congestion_window(tp);
tp->ecn_flags |= (TE_INRECOVERY | TE_SENDCWR);
tp->ecn_flags |= TE_RECV_ECN_ECE;
INP_INC_IFNET_STAT(inp, ecn_recv_ece);
tcpstat.tcps_ecn_recv_ece++;
tcp_ccdbg_trace(tp, th, TCP_CC_ECN_RCVD);
}
}
if (!IN_FASTRECOVERY(tp)) {
if (CC_ALGO(tp)->ack_rcvd != NULL) {
CC_ALGO(tp)->ack_rcvd(tp, th);
}
tcp_ccdbg_trace(tp, th, TCP_CC_ACK_RCVD);
}
if (acked > so->so_snd.sb_cc) {
tp->snd_wnd -= so->so_snd.sb_cc;
sbdrop(&so->so_snd, (int)so->so_snd.sb_cc);
if (so->so_flags & SOF_ENABLE_MSGS) {
so->so_msg_state->msg_serial_bytes -=
(int)so->so_snd.sb_cc;
}
ourfinisacked = 1;
} else {
sbdrop(&so->so_snd, acked);
if (so->so_flags & SOF_ENABLE_MSGS) {
so->so_msg_state->msg_serial_bytes -=
acked;
}
tcp_sbsnd_trim(&so->so_snd);
tp->snd_wnd -= acked;
ourfinisacked = 0;
}
if (!IN_FASTRECOVERY(tp) &&
SEQ_GT(tp->snd_una, tp->snd_recover) &&
SEQ_LEQ(th->th_ack, tp->snd_recover)) {
tp->snd_recover = th->th_ack - 1;
}
if (IN_FASTRECOVERY(tp) &&
SEQ_GEQ(th->th_ack, tp->snd_recover)) {
EXIT_FASTRECOVERY(tp);
}
tp->snd_una = th->th_ack;
if (SACK_ENABLED(tp)) {
if (SEQ_GT(tp->snd_una, tp->snd_recover)) {
tp->snd_recover = tp->snd_una;
}
}
if (SEQ_LT(tp->snd_nxt, tp->snd_una)) {
tp->snd_nxt = tp->snd_una;
}
if (!SLIST_EMPTY(&tp->t_rxt_segments) &&
!TCP_DSACK_SEQ_IN_WINDOW(tp, tp->t_dsack_lastuna,
tp->snd_una)) {
tcp_rxtseg_clean(tp);
}
if ((tp->t_flagsext & TF_MEASURESNDBW) != 0 &&
tp->t_bwmeas != NULL) {
tcp_bwmeas_check(tp);
}
write_wakeup = 1;
if (!SLIST_EMPTY(&tp->t_notify_ack)) {
tcp_notify_acknowledgement(tp, so);
}
switch (tp->t_state) {
case TCPS_FIN_WAIT_1:
if (ourfinisacked) {
if (so->so_state & SS_CANTRCVMORE) {
tp->t_timer[TCPT_2MSL] = OFFSET_FROM_START(tp,
TCP_CONN_MAXIDLE(tp));
isconnected = FALSE;
isdisconnected = TRUE;
}
DTRACE_TCP4(state__change, void, NULL,
struct inpcb *, inp,
struct tcpcb *, tp,
int32_t, TCPS_FIN_WAIT_2);
tp->t_state = TCPS_FIN_WAIT_2;
}
break;
case TCPS_CLOSING:
if (ourfinisacked) {
DTRACE_TCP4(state__change, void, NULL,
struct inpcb *, inp,
struct tcpcb *, tp,
int32_t, TCPS_TIME_WAIT);
tp->t_state = TCPS_TIME_WAIT;
tcp_canceltimers(tp);
if (tp->t_flagsext & TF_NOTIMEWAIT) {
tp->t_flags |= TF_CLOSING;
} else {
add_to_time_wait(tp, 2 * tcp_msl);
}
isconnected = FALSE;
isdisconnected = TRUE;
}
break;
case TCPS_LAST_ACK:
if (ourfinisacked) {
tp = tcp_close(tp);
goto drop;
}
break;
case TCPS_TIME_WAIT:
add_to_time_wait(tp, 2 * tcp_msl);
goto dropafterack;
}
if (sack_ackadv == 1 &&
tp->t_state == TCPS_ESTABLISHED &&
SACK_ENABLED(tp) && sack_bytes_acked > 0 &&
to.to_nsacks > 0 && tp->t_dupacks == 0 &&
SEQ_LEQ(th->th_ack, tp->snd_una) && tlen == 0 &&
!(tp->t_flagsext & TF_PKTS_REORDERED)) {
tcpstat.tcps_sack_ackadv++;
goto process_dupack;
}
}
step6:
if (tcp_update_window(tp, thflags, th, tiwin, tlen)) {
needoutput = 1;
}
if ((thflags & TH_URG) && th->th_urp &&
TCPS_HAVERCVDFIN(tp->t_state) == 0) {
if (th->th_urp + so->so_rcv.sb_cc > sb_max) {
th->th_urp = 0;
thflags &= ~TH_URG;
goto dodata;
}
if (SEQ_GT(th->th_seq + th->th_urp, tp->rcv_up)) {
tp->rcv_up = th->th_seq + th->th_urp;
so->so_oobmark = so->so_rcv.sb_cc +
(tp->rcv_up - tp->rcv_nxt) - 1;
if (so->so_oobmark == 0) {
so->so_state |= SS_RCVATMARK;
postevent(so, 0, EV_OOB);
}
sohasoutofband(so);
tp->t_oobflags &= ~(TCPOOB_HAVEDATA | TCPOOB_HADDATA);
}
if (th->th_urp <= (u_int32_t)tlen
#if SO_OOBINLINE
&& (so->so_options & SO_OOBINLINE) == 0
#endif
) {
tcp_pulloutofband(so, th, m,
drop_hdrlen);
}
} else {
if (SEQ_GT(tp->rcv_nxt, tp->rcv_up)) {
tp->rcv_up = tp->rcv_nxt;
}
}
dodata:
if (isconnected) {
soisconnected(so);
} else if (isdisconnected) {
soisdisconnected(so);
}
if (inp->inp_state == INPCB_STATE_DEAD) {
TCP_LOG_DROP_PCB(TCP_LOG_HDR, th, tp, false, "INPCB_STATE_DEAD");
goto drop;
}
if ((tlen || (thflags & TH_FIN)) &&
TCPS_HAVERCVDFIN(tp->t_state) == 0 &&
(TCPS_HAVEESTABLISHED(tp->t_state) ||
(tp->t_state == TCPS_SYN_RECEIVED &&
(tp->t_tfo_flags & TFO_F_COOKIE_VALID)))) {
tcp_seq save_start = th->th_seq;
tcp_seq save_end = th->th_seq + tlen;
m_adj(m, drop_hdrlen);
if (th->th_seq == tp->rcv_nxt && LIST_EMPTY(&tp->t_segq)) {
TCP_INC_VAR(tp->t_unacksegs, nlropkts);
if (tp->t_flags & TF_STREAMING_ON) {
tcp_compute_rtt(tp, &to, th);
}
if (DELAY_ACK(tp, th) &&
((tp->t_flags & TF_ACKNOW) == 0)) {
if ((tp->t_flags & TF_DELACK) == 0) {
tp->t_flags |= TF_DELACK;
tp->t_timer[TCPT_DELACK] =
OFFSET_FROM_START(tp, tcp_delack);
}
} else {
tp->t_flags |= TF_ACKNOW;
}
tp->rcv_nxt += tlen;
thflags = th->th_flags & TH_FIN;
TCP_INC_VAR(tcpstat.tcps_rcvpack, nlropkts);
tcpstat.tcps_rcvbyte += tlen;
if (nstat_collect) {
if (m->m_pkthdr.pkt_flags & PKTF_SW_LRO_PKT) {
INP_ADD_STAT(inp, cell, wifi, wired,
rxpackets, m->m_pkthdr.lro_npkts);
} else {
INP_ADD_STAT(inp, cell, wifi, wired,
rxpackets, 1);
}
INP_ADD_STAT(inp, cell, wifi, wired,
rxbytes, tlen);
inp_set_activity_bitmap(inp);
}
tcp_sbrcv_grow(tp, &so->so_rcv, &to, tlen,
TCP_AUTORCVBUF_MAX(ifp));
so_recv_data_stat(so, m, drop_hdrlen);
if (isipv6) {
memcpy(&saved_hdr, ip6, sizeof(struct ip6_hdr));
ip6 = (struct ip6_hdr *)&saved_hdr[0];
} else {
memcpy(&saved_hdr, ip, ip->ip_hl << 2);
ip = (struct ip *)&saved_hdr[0];
}
memcpy(&saved_tcphdr, th, sizeof(struct tcphdr));
if (sbappendstream_rcvdemux(so, m,
th->th_seq - (tp->irs + 1), 0)) {
read_wakeup = 1;
}
th = &saved_tcphdr;
} else {
if (isipv6) {
memcpy(&saved_hdr, ip6, sizeof(struct ip6_hdr));
ip6 = (struct ip6_hdr *)&saved_hdr[0];
} else {
memcpy(&saved_hdr, ip, ip->ip_hl << 2);
ip = (struct ip *)&saved_hdr[0];
}
if (tcp_autotune_reorder) {
tcp_sbrcv_grow(tp, &so->so_rcv, &to, tlen, TCP_AUTORCVBUF_MAX(ifp));
}
memcpy(&saved_tcphdr, th, sizeof(struct tcphdr));
thflags = tcp_reass(tp, th, &tlen, m, ifp, &read_wakeup);
th = &saved_tcphdr;
tp->t_flags |= TF_ACKNOW;
}
if ((tlen > 0 || (th->th_flags & TH_FIN)) && SACK_ENABLED(tp)) {
if (th->th_flags & TH_FIN) {
save_end++;
}
tcp_update_sack_list(tp, save_start, save_end);
}
tcp_adaptive_rwtimo_check(tp, tlen);
if (tlen > 0) {
tcp_tfo_rcv_data(tp);
}
if (tp->t_flags & TF_DELACK) {
#if INET6
if (isipv6) {
KERNEL_DEBUG(DBG_LAYER_END, ((th->th_dport << 16) | th->th_sport),
(((ip6->ip6_src.s6_addr16[0]) << 16) | (ip6->ip6_dst.s6_addr16[0])),
th->th_seq, th->th_ack, th->th_win);
} else
#endif
{
KERNEL_DEBUG(DBG_LAYER_END, ((th->th_dport << 16) | th->th_sport),
(((ip->ip_src.s_addr & 0xffff) << 16) | (ip->ip_dst.s_addr & 0xffff)),
th->th_seq, th->th_ack, th->th_win);
}
}
} else {
if ((so->so_flags & SOF_MP_SUBFLOW) && tlen == 0 &&
(m->m_pkthdr.pkt_flags & PKTF_MPTCP_DFIN) &&
(m->m_pkthdr.pkt_flags & PKTF_MPTCP)) {
m_adj(m, drop_hdrlen);
mptcp_input(tptomptp(tp)->mpt_mpte, m);
tp->t_flags |= TF_ACKNOW;
} else {
m_freem(m);
}
thflags &= ~TH_FIN;
}
if (thflags & TH_FIN) {
if (TCPS_HAVERCVDFIN(tp->t_state) == 0) {
socantrcvmore(so);
postevent(so, 0, EV_FIN);
TCP_INC_VAR(tp->t_unacksegs, nlropkts);
if (DELAY_ACK(tp, th) && (tp->t_flags & TF_NEEDSYN)) {
if ((tp->t_flags & TF_DELACK) == 0) {
tp->t_flags |= TF_DELACK;
tp->t_timer[TCPT_DELACK] = OFFSET_FROM_START(tp, tcp_delack);
}
} else {
tp->t_flags |= TF_ACKNOW;
}
tp->rcv_nxt++;
}
switch (tp->t_state) {
case TCPS_SYN_RECEIVED:
tp->t_starttime = tcp_now;
case TCPS_ESTABLISHED:
DTRACE_TCP4(state__change, void, NULL, struct inpcb *, inp,
struct tcpcb *, tp, int32_t, TCPS_CLOSE_WAIT);
tp->t_state = TCPS_CLOSE_WAIT;
break;
case TCPS_FIN_WAIT_1:
DTRACE_TCP4(state__change, void, NULL, struct inpcb *, inp,
struct tcpcb *, tp, int32_t, TCPS_CLOSING);
tp->t_state = TCPS_CLOSING;
break;
case TCPS_FIN_WAIT_2:
DTRACE_TCP4(state__change, void, NULL,
struct inpcb *, inp,
struct tcpcb *, tp,
int32_t, TCPS_TIME_WAIT);
tp->t_state = TCPS_TIME_WAIT;
tcp_canceltimers(tp);
tp->t_flags |= TF_ACKNOW;
if (tp->t_flagsext & TF_NOTIMEWAIT) {
tp->t_flags |= TF_CLOSING;
} else {
add_to_time_wait(tp, 2 * tcp_msl);
}
soisdisconnected(so);
break;
case TCPS_TIME_WAIT:
add_to_time_wait(tp, 2 * tcp_msl);
break;
}
}
#if TCPDEBUG
if (so->so_options & SO_DEBUG) {
tcp_trace(TA_INPUT, ostate, tp, (void *)tcp_saveipgen,
&tcp_savetcp, 0);
}
#endif
if (read_wakeup) {
mptcp_handle_input(so);
}
if (needoutput || (tp->t_flags & TF_ACKNOW)) {
(void) tcp_output(tp);
}
tcp_check_timer_state(tp);
tcp_handle_wakeup(so, read_wakeup, write_wakeup);
socket_unlock(so, 1);
KERNEL_DEBUG(DBG_FNC_TCP_INPUT | DBG_FUNC_END, 0, 0, 0, 0, 0);
return;
dropafterack:
if (tp->t_state == TCPS_SYN_RECEIVED && (thflags & TH_ACK) &&
(SEQ_GT(tp->snd_una, th->th_ack) ||
SEQ_GT(th->th_ack, tp->snd_max))) {
rstreason = BANDLIM_RST_OPENPORT;
IF_TCP_STATINC(ifp, dospacket);
goto dropwithreset;
}
#if TCPDEBUG
if (so->so_options & SO_DEBUG) {
tcp_trace(TA_DROP, ostate, tp, (void *)tcp_saveipgen,
&tcp_savetcp, 0);
}
#endif
m_freem(m);
tp->t_flags |= TF_ACKNOW;
(void) tcp_output(tp);
tcp_handle_wakeup(so, read_wakeup, write_wakeup);
socket_unlock(so, 1);
KERNEL_DEBUG(DBG_FNC_TCP_INPUT | DBG_FUNC_END, 0, 0, 0, 0, 0);
return;
dropwithresetnosock:
nosock = 1;
dropwithreset:
if ((thflags & TH_RST) || m->m_flags & (M_BCAST | M_MCAST)) {
goto drop;
}
#if INET6
if (isipv6) {
if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
goto drop;
}
} else
#endif
if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) ||
IN_MULTICAST(ntohl(ip->ip_src.s_addr)) ||
ip->ip_src.s_addr == htonl(INADDR_BROADCAST) ||
in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif)) {
goto drop;
}
#if ICMP_BANDLIM
if (badport_bandlim(rstreason) < 0) {
goto drop;
}
#endif
#if TCPDEBUG
if (tp == 0 || (tp->t_inpcb->inp_socket->so_options & SO_DEBUG)) {
tcp_trace(TA_DROP, ostate, tp, (void *)tcp_saveipgen,
&tcp_savetcp, 0);
}
#endif
bzero(&tra, sizeof(tra));
tra.ifscope = ifscope;
tra.awdl_unrestricted = 1;
tra.intcoproc_allowed = 1;
if (thflags & TH_ACK) {
tcp_respond(tp, mtod(m, void *), th, m, (tcp_seq)0, th->th_ack,
TH_RST, &tra);
} else {
if (thflags & TH_SYN) {
tlen++;
}
tcp_respond(tp, mtod(m, void *), th, m, th->th_seq + tlen,
(tcp_seq)0, TH_RST | TH_ACK, &tra);
}
if (dropsocket) {
(void) soabort(so);
socket_unlock(so, 1);
} else if ((inp != NULL) && (nosock == 0)) {
tcp_handle_wakeup(so, read_wakeup, write_wakeup);
socket_unlock(so, 1);
}
KERNEL_DEBUG(DBG_FNC_TCP_INPUT | DBG_FUNC_END, 0, 0, 0, 0, 0);
return;
dropnosock:
nosock = 1;
drop:
#if TCPDEBUG
if (tp == 0 || (tp->t_inpcb->inp_socket->so_options & SO_DEBUG)) {
tcp_trace(TA_DROP, ostate, tp, (void *)tcp_saveipgen,
&tcp_savetcp, 0);
}
#endif
m_freem(m);
if (dropsocket) {
(void) soabort(so);
socket_unlock(so, 1);
} else if (nosock == 0) {
tcp_handle_wakeup(so, read_wakeup, write_wakeup);
socket_unlock(so, 1);
}
KERNEL_DEBUG(DBG_FNC_TCP_INPUT | DBG_FUNC_END, 0, 0, 0, 0, 0);
return;
}
static void
tcp_dooptions(struct tcpcb *tp, u_char *cp, int cnt, struct tcphdr *th,
struct tcpopt *to)
{
u_short mss = 0;
int opt, optlen;
for (; cnt > 0; cnt -= optlen, cp += optlen) {
opt = cp[0];
if (opt == TCPOPT_EOL) {
break;
}
if (opt == TCPOPT_NOP) {
optlen = 1;
} else {
if (cnt < 2) {
break;
}
optlen = cp[1];
if (optlen < 2 || optlen > cnt) {
break;
}
}
switch (opt) {
default:
continue;
case TCPOPT_MAXSEG:
if (optlen != TCPOLEN_MAXSEG) {
continue;
}
if (!(th->th_flags & TH_SYN)) {
continue;
}
bcopy((char *) cp + 2, (char *) &mss, sizeof(mss));
NTOHS(mss);
to->to_mss = mss;
to->to_flags |= TOF_MSS;
break;
case TCPOPT_WINDOW:
if (optlen != TCPOLEN_WINDOW) {
continue;
}
if (!(th->th_flags & TH_SYN)) {
continue;
}
to->to_flags |= TOF_SCALE;
to->to_requested_s_scale = min(cp[2], TCP_MAX_WINSHIFT);
break;
case TCPOPT_TIMESTAMP:
if (optlen != TCPOLEN_TIMESTAMP) {
continue;
}
to->to_flags |= TOF_TS;
bcopy((char *)cp + 2,
(char *)&to->to_tsval, sizeof(to->to_tsval));
NTOHL(to->to_tsval);
bcopy((char *)cp + 6,
(char *)&to->to_tsecr, sizeof(to->to_tsecr));
NTOHL(to->to_tsecr);
if (!(tp->t_flags & TF_REQ_TSTMP) &&
tcp_do_rfc1323 == 1) {
tp->t_flags |= TF_REQ_TSTMP;
}
break;
case TCPOPT_SACK_PERMITTED:
if (!tcp_do_sack ||
optlen != TCPOLEN_SACK_PERMITTED) {
continue;
}
if (th->th_flags & TH_SYN) {
to->to_flags |= TOF_SACK;
}
break;
case TCPOPT_SACK:
if (optlen <= 2 || (optlen - 2) % TCPOLEN_SACK != 0) {
continue;
}
to->to_nsacks = (optlen - 2) / TCPOLEN_SACK;
to->to_sacks = cp + 2;
tcpstat.tcps_sack_rcv_blocks++;
break;
case TCPOPT_FASTOPEN:
if (optlen == TCPOLEN_FASTOPEN_REQ) {
if (tp->t_state != TCPS_LISTEN) {
continue;
}
to->to_flags |= TOF_TFOREQ;
} else {
if (optlen < TCPOLEN_FASTOPEN_REQ ||
(optlen - TCPOLEN_FASTOPEN_REQ) > TFO_COOKIE_LEN_MAX ||
(optlen - TCPOLEN_FASTOPEN_REQ) < TFO_COOKIE_LEN_MIN) {
continue;
}
if (tp->t_state != TCPS_LISTEN &&
tp->t_state != TCPS_SYN_SENT) {
continue;
}
to->to_flags |= TOF_TFO;
to->to_tfo = cp + 1;
}
break;
#if MPTCP
case TCPOPT_MULTIPATH:
tcp_do_mptcp_options(tp, cp, th, to, optlen);
break;
#endif
}
}
}
static void
tcp_finalize_options(struct tcpcb *tp, struct tcpopt *to, unsigned int ifscope)
{
if (to->to_flags & TOF_TS) {
tp->t_flags |= TF_RCVD_TSTMP;
tp->ts_recent = to->to_tsval;
tp->ts_recent_age = tcp_now;
}
if (to->to_flags & TOF_MSS) {
tcp_mss(tp, to->to_mss, ifscope);
}
if (SACK_ENABLED(tp)) {
if (!(to->to_flags & TOF_SACK)) {
tp->t_flagsext &= ~(TF_SACK_ENABLE);
} else {
tp->t_flags |= TF_SACK_PERMIT;
}
}
if (to->to_flags & TOF_SCALE) {
tp->t_flags |= TF_RCVD_SCALE;
tp->requested_s_scale = to->to_requested_s_scale;
if (tp->request_r_scale > 0) {
tp->t_flags |= TF_REQ_SCALE;
}
}
}
static void
tcp_pulloutofband(struct socket *so, struct tcphdr *th, struct mbuf *m, int off)
{
int cnt = off + th->th_urp - 1;
while (cnt >= 0) {
if (m->m_len > cnt) {
char *cp = mtod(m, caddr_t) + cnt;
struct tcpcb *tp = sototcpcb(so);
tp->t_iobc = *cp;
tp->t_oobflags |= TCPOOB_HAVEDATA;
bcopy(cp + 1, cp, (unsigned)(m->m_len - cnt - 1));
m->m_len--;
if (m->m_flags & M_PKTHDR) {
m->m_pkthdr.len--;
}
return;
}
cnt -= m->m_len;
m = m->m_next;
if (m == 0) {
break;
}
}
panic("tcp_pulloutofband");
}
uint32_t
get_base_rtt(struct tcpcb *tp)
{
struct rtentry *rt = tp->t_inpcb->inp_route.ro_rt;
return (rt == NULL) ? 0 : rt->rtt_min;
}
void
update_base_rtt(struct tcpcb *tp, uint32_t rtt)
{
u_int32_t base_rtt, i;
struct rtentry *rt;
if ((rt = tp->t_inpcb->inp_route.ro_rt) == NULL) {
return;
}
if (rt->rtt_expire_ts == 0) {
RT_LOCK_SPIN(rt);
if (rt->rtt_expire_ts != 0) {
RT_UNLOCK(rt);
goto update;
}
rt->rtt_expire_ts = tcp_now;
rt->rtt_index = 0;
rt->rtt_hist[0] = rtt;
rt->rtt_min = rtt;
RT_UNLOCK(rt);
return;
}
update:
#if TRAFFIC_MGT
if ((tp->t_flagsext & TF_RECV_THROTTLE) &&
(int)(tcp_now - tp->t_recv_throttle_ts) >= TCP_RECV_THROTTLE_WIN) {
base_rtt = rt->rtt_min;
if (tp->t_rttcur <= (base_rtt + target_qdelay)) {
tp->t_flagsext &= ~TF_RECV_THROTTLE;
tp->t_recv_throttle_ts = 0;
}
}
#endif
if ((int)(tcp_now - rt->rtt_expire_ts) >=
TCP_RTT_HISTORY_EXPIRE_TIME) {
RT_LOCK_SPIN(rt);
if ((int)(tcp_now - rt->rtt_expire_ts) >=
TCP_RTT_HISTORY_EXPIRE_TIME) {
rt->rtt_index++;
if (rt->rtt_index >= NRTT_HIST) {
rt->rtt_index = 0;
}
rt->rtt_hist[rt->rtt_index] = rtt;
rt->rtt_expire_ts = tcp_now;
} else {
rt->rtt_hist[rt->rtt_index] =
min(rt->rtt_hist[rt->rtt_index], rtt);
}
rt->rtt_min = 0;
for (i = 0; i < NRTT_HIST; ++i) {
if (rt->rtt_hist[i] != 0 &&
(rt->rtt_min == 0 ||
rt->rtt_hist[i] < rt->rtt_min)) {
rt->rtt_min = rt->rtt_hist[i];
}
}
RT_UNLOCK(rt);
} else {
rt->rtt_hist[rt->rtt_index] =
min(rt->rtt_hist[rt->rtt_index], rtt);
if (rt->rtt_min == 0) {
rt->rtt_min = rtt;
} else {
rt->rtt_min = min(rt->rtt_min, rtt);
}
}
}
static void
tcp_compute_rtt(struct tcpcb *tp, struct tcpopt *to, struct tcphdr *th)
{
int rtt = 0;
VERIFY(to != NULL && th != NULL);
if (tp->t_rtttime != 0 && SEQ_GT(th->th_ack, tp->t_rtseq)) {
u_int32_t pipe_ack_val;
rtt = tcp_now - tp->t_rtttime;
if (SEQ_GT(th->th_ack, tp->t_pipeack_lastuna)) {
pipe_ack_val = th->th_ack - tp->t_pipeack_lastuna;
tp->t_pipeack_sample[tp->t_pipeack_ind++] =
pipe_ack_val;
tp->t_pipeack_ind %= TCP_PIPEACK_SAMPLE_COUNT;
pipe_ack_val = tcp_get_max_pipeack(tp);
tp->t_pipeack = (pipe_ack_val >
TCP_CC_CWND_INIT_BYTES) ?
pipe_ack_val : 0;
}
tp->t_rtttime = 0;
}
if (((to->to_flags & TOF_TS) != 0) &&
(to->to_tsecr != 0) &&
TSTMP_GEQ(tcp_now, to->to_tsecr)) {
tcp_xmit_timer(tp, (tcp_now - to->to_tsecr),
to->to_tsecr, th->th_ack);
} else if (rtt > 0) {
tcp_xmit_timer(tp, rtt, 0, th->th_ack);
}
}
static void
tcp_xmit_timer(struct tcpcb *tp, int rtt,
u_int32_t tsecr, tcp_seq th_ack)
{
int delta;
int old_srtt = tp->t_srtt;
int old_rttvar = tp->t_rttvar;
bool log_rtt = false;
if (tp->t_inpcb->inp_last_outifp != NULL &&
(tp->t_inpcb->inp_last_outifp->if_eflags & IFEF_AWDL) &&
th_ack == tp->iss + 1) {
return;
}
if (tp->t_flagsext & TF_RECOMPUTE_RTT) {
if (SEQ_GT(th_ack, tp->snd_una) &&
SEQ_LEQ(th_ack, tp->snd_max) &&
(tsecr == 0 ||
TSTMP_GEQ(tsecr, tp->t_badrexmt_time))) {
tp->t_flagsext &= ~(TF_RECOMPUTE_RTT);
tp->t_badrexmt_time = 0;
tp->t_srtt = max(tp->t_srtt_prev, rtt);
tp->t_srtt = tp->t_srtt << TCP_RTT_SHIFT;
tp->t_rttvar = max(tp->t_rttvar_prev, (rtt >> 1));
tp->t_rttvar = tp->t_rttvar << TCP_RTTVAR_SHIFT;
if (tp->t_rttbest > (tp->t_srtt + tp->t_rttvar)) {
tp->t_rttbest = tp->t_srtt + tp->t_rttvar;
}
goto compute_rto;
} else {
return;
}
}
tcpstat.tcps_rttupdated++;
tp->t_rttupdated++;
if (rtt > 0) {
tp->t_rttcur = rtt;
update_base_rtt(tp, rtt);
}
if (tp->t_srtt != 0) {
delta = (rtt << TCP_DELTA_SHIFT)
- (tp->t_srtt >> (TCP_RTT_SHIFT - TCP_DELTA_SHIFT));
if ((tp->t_srtt += delta) <= 0) {
tp->t_srtt = 1;
}
if (delta < 0) {
delta = -delta;
}
delta -= tp->t_rttvar >> (TCP_RTTVAR_SHIFT - TCP_DELTA_SHIFT);
if ((tp->t_rttvar += delta) <= 0) {
tp->t_rttvar = 1;
}
if (tp->t_rttbest == 0 ||
tp->t_rttbest > (tp->t_srtt + tp->t_rttvar)) {
tp->t_rttbest = tp->t_srtt + tp->t_rttvar;
}
} else {
tp->t_srtt = rtt << TCP_RTT_SHIFT;
tp->t_rttvar = rtt << (TCP_RTTVAR_SHIFT - 1);
}
compute_rto:
nstat_route_rtt(tp->t_inpcb->inp_route.ro_rt, tp->t_srtt,
tp->t_rttvar);
TCPT_RANGESET(tp->t_rxtcur, TCP_REXMTVAL(tp),
max(tp->t_rttmin, rtt + 2), TCPTV_REXMTMAX,
TCP_ADD_REXMTSLOP(tp));
tp->t_softerror = 0;
if (log_rtt) {
TCP_LOG_RTT_INFO(tp);
}
TCP_LOG_RTT_CHANGE(tp, old_srtt, old_rttvar);
}
static inline unsigned int
tcp_maxmtu(struct rtentry *rt)
{
unsigned int maxmtu;
int interface_mtu = 0;
RT_LOCK_ASSERT_HELD(rt);
interface_mtu = rt->rt_ifp->if_mtu;
if (rt_key(rt)->sa_family == AF_INET &&
INTF_ADJUST_MTU_FOR_CLAT46(rt->rt_ifp)) {
interface_mtu = IN6_LINKMTU(rt->rt_ifp);
interface_mtu -= CLAT46_HDR_EXPANSION_OVERHD;
}
if (rt->rt_rmx.rmx_mtu == 0) {
maxmtu = interface_mtu;
} else {
maxmtu = MIN(rt->rt_rmx.rmx_mtu, interface_mtu);
}
return maxmtu;
}
#if INET6
static inline unsigned int
tcp_maxmtu6(struct rtentry *rt)
{
unsigned int maxmtu;
struct nd_ifinfo *ndi = NULL;
RT_LOCK_ASSERT_HELD(rt);
if ((ndi = ND_IFINFO(rt->rt_ifp)) != NULL && !ndi->initialized) {
ndi = NULL;
}
if (ndi != NULL) {
lck_mtx_lock(&ndi->lock);
}
if (rt->rt_rmx.rmx_mtu == 0) {
maxmtu = IN6_LINKMTU(rt->rt_ifp);
} else {
maxmtu = MIN(rt->rt_rmx.rmx_mtu, IN6_LINKMTU(rt->rt_ifp));
}
if (ndi != NULL) {
lck_mtx_unlock(&ndi->lock);
}
return maxmtu;
}
#endif
unsigned int
get_maxmtu(struct rtentry *rt)
{
unsigned int maxmtu = 0;
RT_LOCK_ASSERT_NOTHELD(rt);
RT_LOCK(rt);
if (rt_key(rt)->sa_family == AF_INET6) {
maxmtu = tcp_maxmtu6(rt);
} else {
maxmtu = tcp_maxmtu(rt);
}
RT_UNLOCK(rt);
return maxmtu;
}
void
tcp_mss(struct tcpcb *tp, int offer, unsigned int input_ifscope)
{
struct rtentry *rt;
struct ifnet *ifp;
int rtt, mss;
u_int32_t bufsize;
struct inpcb *inp;
struct socket *so;
struct rmxp_tao *taop;
int origoffer = offer;
u_int32_t sb_max_corrected;
int isnetlocal = 0;
#if INET6
int isipv6;
int min_protoh;
#endif
inp = tp->t_inpcb;
so = inp->inp_socket;
if ((so->so_state & SS_DEFUNCT) || tp->t_state == TCPS_CLOSED) {
return;
}
#if INET6
isipv6 = ((inp->inp_vflag & INP_IPV6) != 0) ? 1 : 0;
min_protoh = isipv6 ? sizeof(struct ip6_hdr) + sizeof(struct tcphdr)
: sizeof(struct tcpiphdr);
#else
#define min_protoh (sizeof (struct tcpiphdr))
#endif
#if INET6
if (isipv6) {
rt = tcp_rtlookup6(inp, input_ifscope);
} else
#endif
{
rt = tcp_rtlookup(inp, input_ifscope);
}
isnetlocal = (tp->t_flags & TF_LOCAL);
if (rt == NULL) {
tp->t_maxopd = tp->t_maxseg =
#if INET6
isipv6 ? tcp_v6mssdflt :
#endif
tcp_mssdflt;
return;
}
ifp = rt->rt_ifp;
if (ifp->if_type == IFT_PPP && slowlink_wsize > 0 &&
ifp->if_baudrate > 9600 && ifp->if_baudrate <= 128000) {
tp->t_flags |= TF_SLOWLINK;
}
taop = rmx_taop(rt->rt_rmx);
if (offer == -1) {
offer = taop->tao_mssopt;
}
if (offer == 0) {
offer =
#if INET6
isipv6 ? tcp_v6mssdflt :
#endif
tcp_mssdflt;
} else {
offer = max(offer, tcp_minmss);
offer = max(offer, 64);
}
taop->tao_mssopt = offer;
if (tp->t_srtt == 0 && (rtt = rt->rt_rmx.rmx_rtt) != 0) {
tcp_getrt_rtt(tp, rt);
} else {
tp->t_rttmin = isnetlocal ? tcp_TCPTV_MIN : TCPTV_REXMTMIN;
}
#if INET6
mss = (isipv6 ? tcp_maxmtu6(rt) : tcp_maxmtu(rt));
#else
mss = tcp_maxmtu(rt);
#endif
#if NECP
mss = necp_socket_get_effective_mtu(inp, mss);
#endif
mss -= min_protoh;
if (rt->rt_rmx.rmx_mtu == 0) {
#if INET6
if (isipv6) {
if (!isnetlocal) {
mss = min(mss, tcp_v6mssdflt);
}
} else
#endif
if (!isnetlocal) {
mss = min(mss, tcp_mssdflt);
}
}
mss = min(mss, offer);
tp->t_maxopd = mss;
if ((tp->t_flags & (TF_REQ_TSTMP | TF_NOOPT)) == TF_REQ_TSTMP &&
(origoffer == -1 ||
(tp->t_flags & TF_RCVD_TSTMP) == TF_RCVD_TSTMP)) {
mss -= TCPOLEN_TSTAMP_APPA;
}
#if MPTCP
mss -= mptcp_adj_mss(tp, FALSE);
#endif
tp->t_maxseg = mss;
sb_max_corrected = (sb_max * (u_int64_t)MCLBYTES) / (MSIZE + MCLBYTES);
#if RTV_SPIPE
bufsize = rt->rt_rmx.rmx_sendpipe;
if (bufsize < so->so_snd.sb_hiwat)
#endif
bufsize = so->so_snd.sb_hiwat;
if (bufsize < mss) {
mss = bufsize;
} else {
bufsize = (((bufsize + (u_int64_t)mss - 1) / (u_int64_t)mss) * (u_int64_t)mss);
if (bufsize > sb_max_corrected) {
bufsize = sb_max_corrected;
}
(void)sbreserve(&so->so_snd, bufsize);
}
tp->t_maxseg = mss;
ASSERT(tp->t_maxseg);
tcp_update_mss_locked(so, ifp);
#if RTV_RPIPE
bufsize = rt->rt_rmx.rmx_recvpipe;
if (bufsize < so->so_rcv.sb_hiwat)
#endif
bufsize = so->so_rcv.sb_hiwat;
if (bufsize > mss) {
bufsize = (((bufsize + (u_int64_t)mss - 1) / (u_int64_t)mss) * (u_int64_t)mss);
if (bufsize > sb_max_corrected) {
bufsize = sb_max_corrected;
}
(void)sbreserve(&so->so_rcv, bufsize);
}
set_tcp_stream_priority(so);
if (rt->rt_rmx.rmx_ssthresh) {
tp->snd_ssthresh = max(2 * mss, rt->rt_rmx.rmx_ssthresh);
tcpstat.tcps_usedssthresh++;
} else {
tp->snd_ssthresh = TCP_MAXWIN << TCP_MAX_WINSHIFT;
}
if (CC_ALGO(tp)->cwnd_init != NULL) {
CC_ALGO(tp)->cwnd_init(tp);
}
tcp_ccdbg_trace(tp, NULL, TCP_CC_CWND_INIT);
RT_UNLOCK(rt);
}
int
tcp_mssopt(struct tcpcb *tp)
{
struct rtentry *rt;
int mss;
#if INET6
int isipv6;
int min_protoh;
#endif
#if INET6
isipv6 = ((tp->t_inpcb->inp_vflag & INP_IPV6) != 0) ? 1 : 0;
min_protoh = isipv6 ? sizeof(struct ip6_hdr) + sizeof(struct tcphdr)
: sizeof(struct tcpiphdr);
#else
#define min_protoh (sizeof (struct tcpiphdr))
#endif
#if INET6
if (isipv6) {
rt = tcp_rtlookup6(tp->t_inpcb, IFSCOPE_NONE);
} else
#endif
rt = tcp_rtlookup(tp->t_inpcb, IFSCOPE_NONE);
if (rt == NULL) {
return
#if INET6
isipv6 ? tcp_v6mssdflt :
#endif
tcp_mssdflt;
}
if (rt->rt_ifp->if_type == IFT_PPP && slowlink_wsize > 0 &&
rt->rt_ifp->if_baudrate > 9600 && rt->rt_ifp->if_baudrate <= 128000) {
tp->t_flags |= TF_SLOWLINK;
}
#if INET6
mss = (isipv6 ? tcp_maxmtu6(rt) : tcp_maxmtu(rt));
#else
mss = tcp_maxmtu(rt);
#endif
RT_UNLOCK(rt);
#if NECP
mss = necp_socket_get_effective_mtu(tp->t_inpcb, mss);
#endif
return mss - min_protoh;
}
static void
tcp_newreno_partial_ack(struct tcpcb *tp, struct tcphdr *th)
{
tcp_seq onxt = tp->snd_nxt;
u_int32_t ocwnd = tp->snd_cwnd;
tp->t_timer[TCPT_REXMT] = 0;
tp->t_timer[TCPT_PTO] = 0;
tp->t_rtttime = 0;
tp->snd_nxt = th->th_ack;
tp->snd_cwnd = tp->t_maxseg + BYTES_ACKED(th, tp);
tp->t_flags |= TF_ACKNOW;
(void) tcp_output(tp);
tp->snd_cwnd = ocwnd;
if (SEQ_GT(onxt, tp->snd_nxt)) {
tp->snd_nxt = onxt;
}
if (tp->snd_cwnd > BYTES_ACKED(th, tp)) {
tp->snd_cwnd -= BYTES_ACKED(th, tp);
} else {
tp->snd_cwnd = 0;
}
tp->snd_cwnd += tp->t_maxseg;
}
static int
tcp_dropdropablreq(struct socket *head)
{
struct socket *so, *sonext;
unsigned int i, j, qlen;
static u_int32_t rnd = 0;
static u_int64_t old_runtime;
static unsigned int cur_cnt, old_cnt;
u_int64_t now_sec;
struct inpcb *inp = NULL;
struct tcpcb *tp;
if ((head->so_options & SO_ACCEPTCONN) == 0) {
return 0;
}
if (TAILQ_EMPTY(&head->so_incomp)) {
return 0;
}
so_acquire_accept_list(head, NULL);
socket_unlock(head, 0);
TAILQ_FOREACH_SAFE(so, &head->so_incomp, so_list, sonext) {
inp = sotoinpcb(so);
tp = intotcpcb(inp);
if (tp != NULL && tp->t_state == TCPS_CLOSED &&
so->so_head != NULL &&
(so->so_state & (SS_INCOMP | SS_CANTSENDMORE | SS_CANTRCVMORE)) ==
(SS_INCOMP | SS_CANTSENDMORE | SS_CANTRCVMORE)) {
if (socket_try_lock(so)) {
so->so_usecount++;
goto found_victim;
} else {
continue;
}
}
}
so = TAILQ_FIRST(&head->so_incomp);
now_sec = net_uptime();
if ((i = (now_sec - old_runtime)) != 0) {
old_runtime = now_sec;
old_cnt = cur_cnt / i;
cur_cnt = 0;
}
qlen = head->so_incqlen;
if (rnd == 0) {
rnd = RandomULong();
}
if (++cur_cnt > qlen || old_cnt > qlen) {
rnd = (314159 * rnd + 66329) & 0xffff;
j = ((qlen + 1) * rnd) >> 16;
while (j-- && so) {
so = TAILQ_NEXT(so, so_list);
}
}
while (so) {
inp = (struct inpcb *)so->so_pcb;
sonext = TAILQ_NEXT(so, so_list);
if (in_pcb_checkstate(inp, WNT_ACQUIRE, 0) != WNT_STOPUSING) {
if (socket_try_lock(so)) {
so->so_usecount++;
if ((so->so_usecount == 2) &&
(so->so_state & SS_INCOMP) &&
!(so->so_flags & SOF_INCOMP_INPROGRESS)) {
break;
} else {
in_pcb_checkstate(inp, WNT_RELEASE, 1);
socket_unlock(so, 1);
}
} else {
in_pcb_checkstate(inp, WNT_RELEASE, 1);
}
}
so = sonext;
}
if (so == NULL) {
socket_lock(head, 0);
so_release_accept_list(head);
return 0;
}
if (in_pcb_checkstate(inp, WNT_RELEASE, 1) == WNT_STOPUSING) {
socket_unlock(so, 1);
socket_lock(head, 0);
so_release_accept_list(head);
return 0;
}
found_victim:
if (so->so_usecount != 2 || !(so->so_state & SS_INCOMP)) {
socket_unlock(so, 1);
socket_lock(head, 0);
so_release_accept_list(head);
return 0;
}
socket_lock(head, 0);
TAILQ_REMOVE(&head->so_incomp, so, so_list);
head->so_incqlen--;
head->so_qlen--;
so->so_state &= ~SS_INCOMP;
so->so_flags |= SOF_OVERFLOW;
so->so_head = NULL;
so_release_accept_list(head);
socket_unlock(head, 0);
socket_lock_assert_owned(so);
tp = sototcpcb(so);
tcp_close(tp);
if (inp->inp_wantcnt > 0 && inp->inp_wantcnt != WNT_STOPUSING) {
VERIFY(so->so_usecount > 0);
so->so_usecount--;
socket_unlock(so, 1);
} else {
socket_unlock(so, 0);
lck_rw_lock_exclusive(tcbinfo.ipi_lock);
socket_lock(so, 0);
VERIFY(so->so_usecount > 0);
so->so_usecount--;
if (so->so_usecount != 1 ||
(inp->inp_wantcnt > 0 &&
inp->inp_wantcnt != WNT_STOPUSING)) {
socket_unlock(so, 1);
} else {
VERIFY(so->so_usecount > 0);
so->so_usecount--;
in_pcbdispose(inp);
}
lck_rw_done(tcbinfo.ipi_lock);
}
tcpstat.tcps_drops++;
socket_lock(head, 0);
return 1;
}
void
tcp_set_background_cc(struct socket *so)
{
tcp_set_new_cc(so, TCP_CC_ALGO_BACKGROUND_INDEX);
}
void
tcp_set_foreground_cc(struct socket *so)
{
if (tcp_use_newreno) {
tcp_set_new_cc(so, TCP_CC_ALGO_NEWRENO_INDEX);
} else {
tcp_set_new_cc(so, TCP_CC_ALGO_CUBIC_INDEX);
}
}
static void
tcp_set_new_cc(struct socket *so, uint16_t cc_index)
{
struct inpcb *inp = sotoinpcb(so);
struct tcpcb *tp = intotcpcb(inp);
u_char old_cc_index = 0;
if (tp->tcp_cc_index != cc_index) {
old_cc_index = tp->tcp_cc_index;
if (CC_ALGO(tp)->cleanup != NULL) {
CC_ALGO(tp)->cleanup(tp);
}
tp->tcp_cc_index = cc_index;
tcp_cc_allocate_state(tp);
if (CC_ALGO(tp)->switch_to != NULL) {
CC_ALGO(tp)->switch_to(tp, old_cc_index);
}
tcp_ccdbg_trace(tp, NULL, TCP_CC_CHANGE_ALGO);
}
}
void
tcp_set_recv_bg(struct socket *so)
{
if (!IS_TCP_RECV_BG(so)) {
so->so_flags1 |= SOF1_TRAFFIC_MGT_TCP_RECVBG;
}
so_set_lro(so, SO_TC_BK);
}
void
tcp_clear_recv_bg(struct socket *so)
{
if (IS_TCP_RECV_BG(so)) {
so->so_flags1 &= ~(SOF1_TRAFFIC_MGT_TCP_RECVBG);
}
so_set_lro(so, so->so_traffic_class);
}
void
inp_fc_unthrottle_tcp(struct inpcb *inp)
{
struct tcpcb *tp = inp->inp_ppcb;
if (CC_ALGO(tp)->pre_fr != NULL) {
CC_ALGO(tp)->pre_fr(tp);
}
tp->snd_cwnd = tp->snd_ssthresh;
tp->t_flagsext &= ~TF_CWND_NONVALIDATED;
tp->t_bytes_acked = 0;
TCP_RESET_REXMT_STATE(tp);
tcp_output(tp);
}
static int
tcp_getstat SYSCTL_HANDLER_ARGS
{
#pragma unused(oidp, arg1, arg2)
int error;
struct tcpstat *stat;
stat = &tcpstat;
#if !CONFIG_EMBEDDED
struct tcpstat zero_stat;
if (tcp_disable_access_to_stats &&
!kauth_cred_issuser(kauth_cred_get())) {
bzero(&zero_stat, sizeof(zero_stat));
stat = &zero_stat;
}
#endif
if (req->oldptr == 0) {
req->oldlen = (size_t)sizeof(struct tcpstat);
}
error = SYSCTL_OUT(req, stat, MIN(sizeof(tcpstat), req->oldlen));
return error;
}
int
tcp_input_checksum(int af, struct mbuf *m, struct tcphdr *th, int off, int tlen)
{
struct ifnet *ifp = m->m_pkthdr.rcvif;
switch (af) {
case AF_INET: {
struct ip *ip = mtod(m, struct ip *);
struct ipovly *ipov = (struct ipovly *)ip;
if (m->m_pkthdr.pkt_flags & PKTF_SW_LRO_DID_CSUM) {
return 0;
}
ASSERT((ip->ip_hl << 2) == sizeof(*ip));
if ((hwcksum_rx || (ifp->if_flags & IFF_LOOPBACK) ||
(m->m_pkthdr.pkt_flags & PKTF_LOOP)) &&
(m->m_pkthdr.csum_flags & CSUM_DATA_VALID)) {
if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR) {
th->th_sum = m->m_pkthdr.csum_rx_val;
} else {
uint32_t sum = m->m_pkthdr.csum_rx_val;
uint32_t start = m->m_pkthdr.csum_rx_start;
int32_t trailer = (m_pktlen(m) - (off + tlen));
ASSERT(trailer >= 0);
if ((m->m_pkthdr.csum_flags & CSUM_PARTIAL) &&
((start != 0 && start != off) || trailer)) {
uint32_t swbytes = (uint32_t)trailer;
if (start < off) {
ip->ip_len += sizeof(*ip);
#if BYTE_ORDER != BIG_ENDIAN
HTONS(ip->ip_len);
HTONS(ip->ip_off);
#endif
}
sum = m_adj_sum16(m, start, off,
tlen, sum);
if (off > start) {
swbytes += (off - start);
} else {
swbytes += (start - off);
}
if (start < off) {
#if BYTE_ORDER != BIG_ENDIAN
NTOHS(ip->ip_off);
NTOHS(ip->ip_len);
#endif
ip->ip_len -= sizeof(*ip);
}
if (swbytes != 0) {
tcp_in_cksum_stats(swbytes);
}
if (trailer != 0) {
m_adj(m, -trailer);
}
}
th->th_sum = in_pseudo(ip->ip_src.s_addr,
ip->ip_dst.s_addr,
sum + htonl(tlen + IPPROTO_TCP));
}
th->th_sum ^= 0xffff;
} else {
uint16_t ip_sum;
int len;
char b[9];
bcopy(ipov->ih_x1, b, sizeof(ipov->ih_x1));
bzero(ipov->ih_x1, sizeof(ipov->ih_x1));
ip_sum = ipov->ih_len;
ipov->ih_len = (u_short)tlen;
#if BYTE_ORDER != BIG_ENDIAN
HTONS(ipov->ih_len);
#endif
len = sizeof(struct ip) + tlen;
th->th_sum = in_cksum(m, len);
bcopy(b, ipov->ih_x1, sizeof(ipov->ih_x1));
ipov->ih_len = ip_sum;
tcp_in_cksum_stats(len);
}
break;
}
#if INET6
case AF_INET6: {
struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
if (m->m_pkthdr.pkt_flags & PKTF_SW_LRO_DID_CSUM) {
return 0;
}
if ((hwcksum_rx || (ifp->if_flags & IFF_LOOPBACK) ||
(m->m_pkthdr.pkt_flags & PKTF_LOOP)) &&
(m->m_pkthdr.csum_flags & CSUM_DATA_VALID)) {
if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR) {
th->th_sum = m->m_pkthdr.csum_rx_val;
} else {
uint32_t sum = m->m_pkthdr.csum_rx_val;
uint32_t start = m->m_pkthdr.csum_rx_start;
int32_t trailer = (m_pktlen(m) - (off + tlen));
ASSERT(trailer >= 0);
if ((m->m_pkthdr.csum_flags & CSUM_PARTIAL) &&
(start != off || trailer != 0)) {
uint16_t s = 0, d = 0;
uint32_t swbytes = (uint32_t)trailer;
if (IN6_IS_SCOPE_EMBED(&ip6->ip6_src)) {
s = ip6->ip6_src.s6_addr16[1];
ip6->ip6_src.s6_addr16[1] = 0;
}
if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst)) {
d = ip6->ip6_dst.s6_addr16[1];
ip6->ip6_dst.s6_addr16[1] = 0;
}
sum = m_adj_sum16(m, start, off,
tlen, sum);
if (off > start) {
swbytes += (off - start);
} else {
swbytes += (start - off);
}
if (IN6_IS_SCOPE_EMBED(&ip6->ip6_src)) {
ip6->ip6_src.s6_addr16[1] = s;
}
if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst)) {
ip6->ip6_dst.s6_addr16[1] = d;
}
if (swbytes != 0) {
tcp_in6_cksum_stats(swbytes);
}
if (trailer != 0) {
m_adj(m, -trailer);
}
}
th->th_sum = in6_pseudo(
&ip6->ip6_src, &ip6->ip6_dst,
sum + htonl(tlen + IPPROTO_TCP));
}
th->th_sum ^= 0xffff;
} else {
tcp_in6_cksum_stats(tlen);
th->th_sum = in6_cksum(m, IPPROTO_TCP, off, tlen);
}
break;
}
#endif
default:
VERIFY(0);
}
if (th->th_sum != 0) {
tcpstat.tcps_rcvbadsum++;
IF_TCP_STATINC(ifp, badformat);
return -1;
}
return 0;
}
SYSCTL_PROC(_net_inet_tcp, TCPCTL_STATS, stats,
CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_LOCKED, 0, 0, tcp_getstat,
"S,tcpstat", "TCP statistics (struct tcpstat, netinet/tcp_var.h)");
static int
sysctl_rexmtthresh SYSCTL_HANDLER_ARGS
{
#pragma unused(arg1, arg2)
int error, val = tcprexmtthresh;
error = sysctl_handle_int(oidp, &val, 0, req);
if (error || !req->newptr) {
return error;
}
if (val < 2 || val > 3) {
return EINVAL;
}
tcprexmtthresh = val;
return 0;
}
SYSCTL_PROC(_net_inet_tcp, OID_AUTO, rexmt_thresh, CTLTYPE_INT | CTLFLAG_RW |
CTLFLAG_LOCKED, &tcprexmtthresh, 0, &sysctl_rexmtthresh, "I",
"Duplicate ACK Threshold for Fast Retransmit");