Date: Thu, 2 Jul 2009 11:18:46 GMT From: Andre Oppermann <andre@FreeBSD.org> To: Perforce Change Reviews <perforce@freebsd.org> Subject: PERFORCE change 165539 for review Message-ID: <200907021118.n62BIktH076734@repoman.freebsd.org>
index | next in thread | raw e-mail
http://perforce.freebsd.org/chv.cgi?CH=165539 Change 165539 by andre@andre_flirtbox on 2009/07/02 11:18:45 Fix up function prototypes, add variables, comment out gone or currently unresolved stuff, hackery... Makes tcp_new syntactically correct enough to compile but not functional or useable yet. Affected files ... .. //depot/projects/tcp_new/modules/Makefile#3 edit .. //depot/projects/tcp_new/netinet/tcp.h#4 edit .. //depot/projects/tcp_new/netinet/tcp_input.c#12 edit .. //depot/projects/tcp_new/netinet/tcp_output.c#10 edit .. //depot/projects/tcp_new/netinet/tcp_reass.c#2 edit .. //depot/projects/tcp_new/netinet/tcp_sack.c#2 edit .. //depot/projects/tcp_new/netinet/tcp_seq.h#2 edit .. //depot/projects/tcp_new/netinet/tcp_subr.c#5 edit .. //depot/projects/tcp_new/netinet/tcp_syncache.c#5 edit .. //depot/projects/tcp_new/netinet/tcp_syncache.h#2 edit .. //depot/projects/tcp_new/netinet/tcp_timer.c#3 edit .. //depot/projects/tcp_new/netinet/tcp_timewait.c#2 edit .. //depot/projects/tcp_new/netinet/tcp_usrreq.c#4 edit .. //depot/projects/tcp_new/netinet/tcp_var.h#10 edit Differences ... ==== //depot/projects/tcp_new/modules/Makefile#3 (text+ko) ==== @@ -64,7 +64,6 @@ ${_ctau} \ cue \ ${_cx} \ - cxgb \ dc \ dcons \ dcons_crom \ ==== //depot/projects/tcp_new/netinet/tcp.h#4 (text+ko) ==== ==== //depot/projects/tcp_new/netinet/tcp_input.c#12 (text+ko) ==== @@ -128,10 +128,10 @@ &tcp_do_rfc3390, 0, "Enable RFC 3390 (Increasing TCP's Initial Congestion Window)"); -static int tcp_insecure_rst = 0; -SYSCTL_INT(_net_inet_tcp, OID_AUTO, insecure_rst, CTLFLAG_RW, - &tcp_insecure_rst, 0, - "Follow the old (insecure) criteria for accepting RST packets"); +static int tcp_secure_rst = 0; +SYSCTL_INT(_net_inet_tcp, OID_AUTO, secure_rst, CTLFLAG_RW, + &tcp_secure_rst, 0, + "Follow the new (secure) criteria for accepting RST packets"); int tcp_do_autorcvbuf = 1; SYSCTL_INT(_net_inet_tcp, OID_AUTO, recvbuf_auto, CTLFLAG_RW, @@ -150,21 +150,25 @@ #define tcb6 tcb /* for KAME src sync over BSD*'s */ struct inpcbinfo tcbinfo; -static void tcp_dooptions(struct tcpopt *, u_char *, int, int); +static void tcp_do_options(struct tcpopt *, u_char *, int, int); +static void tcp_do_ack(struct tcpcb *tp, struct tcphdr *th, int tiwin, + int acked, int tlen, int sacked); +static int tcp_do_rto(struct tcpcb *tp, int rtt); static void tcp_do_segment(struct mbuf *, struct tcphdr *, struct socket *, struct tcpcb *, int, int); static void tcp_do_time(struct tcpcb *tp, struct tcphdr *th, - struct tcpopt *to, int acked, int tlen); -static void tcp_do_urg(struct tcpcb *tp, struct tcphdr *th, int tlen); -static void tcp_do_wu(struct tcpcb *tp, struct tcphdr *th, + struct tcpopt *to, int acked, int tlen, int sacked); +static void tcp_do_urg(struct tcpcb *tp, struct socket *so, + struct tcphdr *th, int *tlen); +static int tcp_do_wu(struct tcpcb *tp, struct tcphdr *th, struct tcpopt *to, int tiwin, int acked, int tlen, int sacked); static void tcp_dropwithreset(struct mbuf *, struct tcphdr *, struct tcpcb *, int, int); +int tcp_init_cwnd(struct tcpcb *tp); +static void tcp_init_values(struct tcpcb *tp); static void tcp_pulloutofband(struct socket *, struct tcphdr *, struct mbuf *, int); -static void tcp_xmit_timer(struct tcpcb *, int); -static void tcp_newreno_partial_ack(struct tcpcb *, struct tcphdr *); /* Neighbor Discovery, Neighbor Unreachability Detection Upper layer hint. */ #ifdef INET6 @@ -504,7 +508,7 @@ */ if (inp->inp_vflag & INP_TIMEWAIT) { if (thflags & TH_SYN) - tcp_dooptions(&to, optp, optlen, TO_SYN); + tcp_do_options(&to, optp, optlen, TO_SYN); /* * NB: tcp_twcheck unlocks the INP and frees the mbuf. */ @@ -580,7 +584,7 @@ * syncookies need access to the reflected * timestamp. */ - tcp_dooptions(&to, optp, optlen, 0); + tcp_do_options(&to, optp, optlen, 0); /* * NB: syncache_expand() doesn't unlock * inp and tcpinfo locks. @@ -819,7 +823,7 @@ tcp_trace(TA_INPUT, ostate, tp, (void *)tcp_saveipgen, &tcp_savetcp, 0); #endif - tcp_dooptions(&to, optp, optlen, TO_SYN); + tcp_do_options(&to, optp, optlen, TO_SYN); syncache_add(&inc, &to, th, inp, &so, m); /* * Entry added to syncache and mbuf consumed. @@ -856,7 +860,7 @@ return; } -#define tcplog (s = tcp_log_addrs(tcpcbtoinc(tp), th, NULL, NULL)) +#define tcplog(x) (s = tcp_log_addrs(&tp->t_inpcb->inp_inc, th, NULL, NULL)) static void tcp_do_segment(struct mbuf *m, struct tcphdr *th, struct socket *so, @@ -898,7 +902,7 @@ * RFC1122: section 4.2.2.5 */ if ((th->th_off << 2) != sizeof(struct tcphdr)) - tcp_dooptions(&to, (u_char *)(th + 1), + tcp_do_options(&to, (u_char *)(th + 1), (th->th_off << 2) - sizeof(struct tcphdr), (thflags & TH_SYN) ? TO_SYN : 0); else @@ -1021,9 +1025,9 @@ * SYN and once when get back the SYN-ACK. */ if (to.to_flags & TOF_MSS) - tp->snd_mss = tcp_mss(tptoinpinc(tp), to.to_mss, 0); + tp->snd_mss = tcp_mss(tcpcbtoinc(tp), to.to_mss, 0); else - tp->snd_mss = tcp_mss(tptoinpinc(tp), 0, 0); + tp->snd_mss = tcp_mss(tcpcbtoinc(tp), 0, 0); /* * Do window scaling on this connection? @@ -1048,7 +1052,7 @@ */ tcplog("Window Scaling Option unexpected, " "connection aborted"); - tp->t_error = ENETRESET; /* XXXAO: Correct error? */ + tp->t_softerror = ENETRESET; /* XXXAO: Correct error? */ tp = tcp_close(tp); rstreason = BANDLIM_UNLIMITED; goto dropwithreset; @@ -1070,7 +1074,7 @@ */ tcplog("Timestamp Option unexpected, " "connection aborted"); - tp->t_error = ENETRESET; + tp->t_softerror = ENETRESET; tp = tcp_close(tp); rstreason = BANDLIM_UNLIMITED; goto dropwithreset; @@ -1092,7 +1096,7 @@ */ tcplog("SACK Permitted unexpected, " "connection aborted"); - tp->t_error = ENETRESET; + tp->t_softerror = ENETRESET; tp = tcp_close(tp); rstreason = BANDLIM_UNLIMITED; goto dropwithreset; @@ -1102,13 +1106,12 @@ * Initialize receive structure. * XXXAO: TODO */ - tp->rcv_adv += rwin; /* XXX */ - tp->irs = th->th_seq; + tp->rcv_advwin += rwin; /* XXX */ tp->rcv_up = th->th_seq; - tcp_rcvseqinit(tp); - tcp_init_rcv(tp, seq); /* XXXAO */ - tcp_init_snd(tp, ack); /* XXXAO */ + //tcp_init_rcv(tp, seq); /* XXXAO */ + //tcp_init_snd(tp, ack); /* XXXAO */ + tcp_init_values(tp); /* * Process SYN and integrate sequence number. @@ -1132,8 +1135,8 @@ * Otherwise go into idle mode and wait for * either side to start transmitting data. */ - if (tp->snd_una != snd_nxt) { - tcp_timer_activate(tp, TT_RXMIT, ); /* XXXAO */ + if (tp->snd_una != tp->snd_nxt) { + tcp_timer_activate(tp, TT_REXMT, 1); /* XXXAO */ nudgeoutput = 1; } else { tcp_timer_activate(tp, TT_KEEP, tcp_keepidle); /* XXXAO */ @@ -1191,7 +1194,7 @@ * RFC2018: section 1, page 2, last paragraph */ if ((to.to_flags & TOF_SACK) && - !(tp->t_flags & TF_SACK_PERM)) { + !(tp->t_flags & TF_SACK_PERMIT)) { tcplog("SACK unexpected, segment ignored"); goto drop; } @@ -1214,21 +1217,21 @@ * Linux 2.6.25, net/ipv4/tcp_input.c, tcp_disordered_ack() */ if (to.to_flags & TOF_TS) { - time_t uptime = tcp_uptime(); + time_t uptime = (time_t)ticks; /* XXXAO */ if (uptime - tp->t_rcvtime < ((tcp_ts)0x0 - 1) / (hz * 2)) { - if (TSTMP_LT(to.to_tsval, tp->snd_tsecr) { + if (TSTMP_LT(to.to_tsval, tp->snd_tsecr)) { tcplog("Timestamp too old, " "sending challenge ack"); goto dropafterack; } - if (TSTMP_GT(to.to_tsecr, tp->snd_tsval) { + if (TSTMP_GT(to.to_tsecr, tp->snd_tslast)) { tcplog("Timestamp too new, " "sending challenge ack"); goto dropafterack; } } else - to->to_flags &= ~TOF_TS; + to.to_flags &= ~TOF_TS; } /* @@ -1276,7 +1279,7 @@ if (th->th_seq == tp->rcv_nxt - 1 && th->th_ack == tp->snd_nxt && tlen <= 1 && !(thflags & TH_URG) && - !TCPS_HAVERCVDFIN(tp)) { + !TCPS_HAVERCVDFIN(tp->t_state)) { /* * The connection is idle and this * is a keepalive. Force an ACK and @@ -1293,9 +1296,9 @@ thflags &= ~TH_FIN; } tp->t_flags |= TF_ACKNOW; - tcps.tcps_rcv_keepalive++; + //tcpstat.tcps_rcv_keepalive++; } else if (th->th_seq == tp->rcv_nxt - 1 && - TCPS_HAVERCVDFIN(tp) && (thflags & TH_FIN)) { + TCPS_HAVERCVDFIN(tp->t_state) && (thflags & TH_FIN)) { /* * Continue with duplicate FIN. * When we received the orginal FIN @@ -1309,7 +1312,7 @@ */ tp->t_flags |= TF_ACKNOW; } else if (th->th_seq == tp->rcv_nxt && tlen > 0 && - (thflags & TH_URG) && th->th_urg > 0) { + (thflags & TH_URG) && th->th_urp > 0) { /* * We must accept urgent data even when * the window is closed. Continue @@ -1355,7 +1358,7 @@ * XXXAO: Informal. Not part of any RFC. */ if (SEQ_GT(th->th_seq, tp->snd_lastack) && - SEQ_LT(th->th_seq, tp->rcv_nxt) { + SEQ_LT(th->th_seq, tp->rcv_nxt)) { tcplog("Received retransmit before we sent delayed ACK," " no action"); } @@ -1433,7 +1436,7 @@ * a SYN. */ if (th->th_ack != tp->snd_una || - th->th_ack != th->snd_nxt) { + th->th_ack != tp->snd_nxt) { tcplog("RST does not match, segment ignored"); tcpstat.tcps_badrst++; goto drop; @@ -1476,14 +1479,14 @@ */ if (tcp_secure_rst && (SEQ_DELTA(th->th_seq, tp->rcv_nxt) > 1 || - SEQ_DELTA(th->th_seq, tp->snd_last_ack) > 1) || - SEQ_DELTA(th->th_seq, tp->snd_last_ack + rwin) > 1) { + SEQ_DELTA(th->th_seq, tp->snd_lastack) > 1 || + SEQ_DELTA(th->th_seq, tp->snd_lastack + rwin) > 1)) { tcplog("RST does not match (secure), segment ignored"); tcpstat.tcps_badrst++; goto drop; } else if (!tcp_secure_rst && - (SEQ_LT(th->th_seq, tp->snd_last_ack - 1) || - SEQ_GT(th->th_seq, tp->snd_last_ack + rwin))) { + (SEQ_LT(th->th_seq, tp->snd_lastack - 1) || + SEQ_GT(th->th_seq, tp->snd_lastack + rwin))) { tcplog("RST does not match (insecure), segment ignored"); tcpstat.tcps_badrst++; goto drop; @@ -1646,11 +1649,11 @@ * drop trailing data (and PUSH and FIN); * if nothing left, just ACK. */ - todrop = th->th_seq + tlen, tp->rcv_nxt + rwin; + todrop = SEQ_DELTA(th->th_seq + tlen, tp->rcv_nxt + rwin); if (todrop > 0) { KASSERT(todrop <= tlen, ("%s: right todrop > tlen", __func__)); - KASSERT(!TCPS_HAVERCVDFIN(tp), + KASSERT(!TCPS_HAVERCVDFIN(tp->t_state), ("%s: FIN received, todrop > 0", __func__)); /* @@ -1760,7 +1763,7 @@ /* * Update congestion control information. */ - tcp_cc_ack(tp, th, tiwin, acked, tlen, sacked); + //tcp_cc_ack(tp, th, tiwin, acked, tlen, sacked); KASSERT(tp->snd_cwnd > tp->snd_mss, ("%s: cwnd < 1*mss after congestion control function", __func__)); @@ -1768,14 +1771,14 @@ * Drop acknowledged data from send socket buffer. * RFC793: section 3.9, page 72, fifth check */ - if (acked > 0) + if (acked > 0) { SOCKBUF_LOCK(&so->so_snd); KASSERT(SEQ_GT(th->th_ack, tp->snd_nxt), ("%s: more acked than sent", __func__)); KASSERT(acked <= so->so_snd.sb_cc + 1, ("%s: more acked than in send buffer", __func__)); - KASSERT(!TCPS_HAVERCVDFIN(tp), + KASSERT(!TCPS_HAVERCVDFIN(tp->t_state), ("%s: FIN already processed but acked > 0", __func__)); /* @@ -1810,9 +1813,9 @@ ("%s: got ack for FIN but haven't sent FIN yet", __func__)); - KASSERT(!tcp_timer_active(TT_RXMIT), + KASSERT(!tcp_timer_active(tp, TT_REXMT), ("%s: ourfinisacked but RXMIT still active", - __func__); + __func__)); /* * Handle ack'ed FIN according to previous state. @@ -1903,7 +1906,7 @@ */ if (tlen == 0 && (th->th_flags & TH_FIN) == 0 && !TCPS_HAVERCVDFIN(tp->t_state)) { - INP_INFO_WUNLOCK(&tcpinfo); + INP_INFO_WUNLOCK(&tcbinfo); m_freem(m); m = NULL; th = NULL; @@ -1917,7 +1920,7 @@ */ if ((thflags & TH_URG) && th->th_urp > 0 && tlen > 0 && !TCPS_HAVERCVDFIN(tp->t_state)) { - tcp_do_urg(tp, th, &tlen); + tcp_do_urg(tp, so, th, &tlen); } else if (SEQ_GT(tp->rcv_nxt, tp->rcv_up)) { tp->rcv_up = tp->rcv_nxt; } @@ -1937,7 +1940,7 @@ * segment with urgent that got pulled and now is zero */ if (!TCPS_HAVERCVDFIN(tp->t_state) && - (tlen > 0 || (tp->rcv_trq != NULL && (th->th_flags & TH_FIN)))) { + (tlen > 0 || (!TAILQ_EMPTY(&tp->rcv_trq) && (th->th_flags & TH_FIN)))) { int newsize = 0; /* Rcvbuf autoscaling. */ /* @@ -1955,14 +1958,14 @@ * XXXAO: This breaks with reassembly queue and * tlen == 0 and TH_FIN. */ - m_demote(m); + m_demote(m, 1); while (m->m_len == 0 && m->m_next != NULL) m = m_free(m); KASSERT(m->m_len > 0, ("%s: drop_hdrlen too big or mbuf chain broken", __func__)); } - KASSERT(tlen == m_length(m), + KASSERT(tlen == m_length(m, NULL), ("%s: tlen != mbuf chain length", __func__)); /* @@ -1980,7 +1983,8 @@ * RFC1122: section 4.2.2.21 */ if (th->th_seq != tp->rcv_nxt || !TAILQ_EMPTY(&tp->rcv_trq)) { - m = tcp_reass(tp, m, tlen, thflags); + //m = tcp_reass(tp, m, tlen, thflags); + m = NULL; if (m != NULL && (m->m_flags & M_PROTO1)) thflags |= TH_FIN; @@ -2029,7 +2033,7 @@ */ if (m != NULL && tcp_do_autorcvbuf && (to.to_flags & TOF_TS) && (so->so_rcv.sb_flags & SB_AUTOSIZE)) { - if (TS_GT(to.to_tsecr, tp->rfbuf_ts) && + if (TSTMP_GT(to.to_tsecr, tp->rfbuf_ts) && to.to_tsecr - tp->rfbuf_ts < hz) { if (tp->rfbuf_cnt > (so->so_rcv.sb_hiwat / 8 * 7) && so->so_rcv.sb_hiwat < tcp_autorcvbuf_max) { @@ -2091,7 +2095,7 @@ /* * Update size of receive window. */ - tp->rcv_wnd = sbspace(so->so_rcv); + tp->rcv_wnd = (tcp_seq)sbspace(&so->so_rcv); /* * NB: sorwakeup_locked implicitly unlocks. @@ -2125,10 +2129,10 @@ */ } } else { - KASSERT(!TCPS_HAVERCVDFIN(tp) || TAILQ_EMPTY(&tp->rcv_trq), + KASSERT(!TCPS_HAVERCVDFIN(tp->t_state) || TAILQ_EMPTY(&tp->rcv_trq), ("%s: FIN received but reassembly queue not empty", __func__)); KASSERT(tlen == 0, - ("%s: data unexpected", __func__)) + ("%s: data unexpected", __func__)); /* * Dispose of segment. The data was already dropped @@ -2255,7 +2259,7 @@ * XXXAO: Multi-delack? * XXXAO: Always call into tcp_output and have it decide what to do. */ - (void)tcp_output(tp, TPO_TINPUT); + (void)tcp_output(tp); #if 0 if ((tp->t_flags & TF_ACKNOW) || tp->snd_delack > 1 || nudgeoutput || (tp->t_flags & TF_RXWIN0SENT) || @@ -2271,7 +2275,7 @@ return; dropafterack: - INP_INFO_LOCK_ASSERT(&tcpinfo); + INP_INFO_WLOCK_ASSERT(&tcbinfo); INP_LOCK_ASSERT(tp->t_inpcb); KASSERT(!(thflags & TH_RST), ("%s: dropafterack with RST", __func__)); @@ -2290,12 +2294,12 @@ */ tp->t_flags |= TF_ACKNOW; m_freem(m); - (void) tcp_output(tp, TPO_TINPUT); + (void) tcp_output(tp); INP_UNLOCK(tp->t_inpcb); return; dropwithreset: - INP_INFO_LOCK_ASSERT(&tcpinfo); + INP_INFO_WLOCK_ASSERT(&tcbinfo); INP_INFO_WUNLOCK(&tcbinfo); KASSERT(th != NULL, ("%s: th for tcp_dropwithreset() is NULL", __func__)); @@ -2310,7 +2314,7 @@ return; drop: - INP_INFO_LOCK_ASSERT(&tcpinfo); + INP_INFO_WLOCK_ASSERT(&tcbinfo); /* * Drop space held by incoming segment and return. */ @@ -2489,7 +2493,7 @@ int acked, int tlen, int sacked) { int rtt; - tcp_ts tick = tcp_ticks; + tcp_ts tick = ticks; INP_LOCK_ASSERT(tp->t_inpcb); KASSERT(tp != NULL && th != NULL && to != NULL, @@ -2500,8 +2504,7 @@ * * Make note of most recent segment received time. */ - tp->t_rcvtime = tcp_uptime(); - tp->t_rcvticks = tick; + tp->t_rcvtime = time_uptime; /* * 2. If timestamps are used decide which to reflect. @@ -2527,27 +2530,29 @@ * Does this give PAWS problems? */ if (to->to_flags & TOF_TS) { +#define TCP_RFC1323_BRADEN #ifdef TCP_RFC1323_BRADEN - if (TS_GEQ(to->to_tsval, tp->snd_tsecr) && + if (TSTMP_GEQ(to->to_tsval, tp->snd_tsecr) && SEQ_LEQ(th->th_ack, tp->snd_lastack)) { #endif #ifdef TCP_RFC1323bis_plusSACK - if (TS_GT(to->to_tsval, tp->snd_tsecr) && + if (TSTMP_GT(to->to_tsval, tp->snd_tsecr) && ((th->th_seq == tp->rcv_nxt && tp->snd_delack == 0) || - sacked > 0) { + sacked > 0)) { #endif tp->snd_tsecr = to->to_tsval; - tp->snd_tsecrts = tcp_ticks; + //tp->snd_tsecrts = ticks; } - KASSERT(!TS_GT(to->to_secr, tick), + KASSERT(!TSTMP_GT(to->to_tsecr, tick), ("%s: timestamp newer than our time", __func__)); /* * Remember highest most recent reflected TS. */ if (SEQ_LEQ(th->th_seq, tp->snd_lastack) && - TS_GT(to->to_tsecr > tp->ts_recent)) - tp->ts_recent = to->to_tsecr; - tp->ts_recentts = tick; + TSTMP_GT(to->to_tsecr, tp->tsval_recent)) { + tp->tsval_recent = to->to_tsecr; + //tp->ts_recentts = tick; + } } /* @@ -2557,7 +2562,7 @@ rtt = tick - to->to_tsecr; } else if (acked > 0 && tp->snd_rtseq != 0 && SEQ_GT(th->th_ack, tp->snd_rtseq) && - TAILQ_EMPTY(tp->rcv_trq) && tp->snd_rtoshift == 0) { + TAILQ_EMPTY(&tp->rcv_trq) /*&& tp->snd_rtoshift == 0*/ ) { rtt = tick - tp->snd_rtts; tp->snd_rtseq = 0; } else @@ -2587,8 +2592,8 @@ * the time measurement or our estimate will be way off. */ if (tick - tp->snd_tslast > tp->snd_rto) { - if (tp->t_rtseq != 0) - tp->t_rtseq = 0; + if (tp->snd_rtseq != 0) + tp->snd_rtseq = 0; return; } @@ -2878,8 +2883,8 @@ #endif #ifdef TCP_WU_BSDNEW if (((to->to_flags & TOF_TS) && - TS_GT(to->to_tsecr, tp->ts_reflected_last) || - TS_GT(to->to_tsval, tp->ts_recent)) || + (TSTMP_GT(to->to_tsecr, tp->snd_tsecr) || + TSTMP_GT(to->to_tsval, tp->snd_tslast))) || acked > 0 || (SEQ_GT(th->th_seq, tp->snd_wu_seq) && acked >= 0) || (th->th_seq == tp->snd_wu_seq && th->th_ack == tp->snd_una && @@ -2898,8 +2903,9 @@ * RFC793: section 3.7, page 42-44, "Managing the Window" * RFC1122: section 4.2.2.16 */ - if (SEQ_DELTA(tp->snd_nxt, tp->snd_una + acked) + tiwin < tp->snd_wnd) - tcplog("peer shrank the window"); + if (SEQ_DELTA(tp->snd_nxt, tp->snd_una + acked) + tiwin < tp->snd_wnd) { + //tcplog("peer shrank the window"); + } /* * Update the window and keep track of this update. @@ -2920,7 +2926,8 @@ } static void -tcp_do_ack(tp, th, tiwin, acked, tlen, sacked) +tcp_do_ack(struct tcpcb *tp, struct tcphdr *th, int tiwin, int acked, + int tlen, int sacked) { /* * Without SACK detecting a duplicate ACK is based on an @@ -2959,8 +2966,8 @@ if (acked > 0) tp->snd_una += acked; - KASSERT(tp->snd_una == tp->snd_nxt || tcp_timer_active(TT_RXMIT), - ("%s: outstanding data but RXMIT timer not active", __func__)); + KASSERT(tp->snd_una == tp->snd_nxt || tcp_timer_active(tp, TT_REXMT), + ("%s: outstanding data but REXMT timer not active", __func__)); /* * Stop the retransmit timer if all data we sent has been @@ -2970,9 +2977,9 @@ * XXXAO: Handle backoff on multiple retransmits. */ if (acked > 0 && tp->snd_una == tp->snd_nxt) - tcp_timer_activate(TT_RXMIT, 0); + tcp_timer_activate(tp, TT_REXMT, 0); else if (acked > 0) - tcp_timer_activate(TT_RXMIT, tp->snd_rto); + tcp_timer_activate(tp, TT_REXMT, tp->snd_rto); } @@ -2983,14 +2990,16 @@ * Finish this function and validate against all relevant RFCs. */ void -tcp_do_urg(struct tcpcb *tp, struct tcphdr *th, int tlen) +tcp_do_urg(struct tcpcb *tp, struct socket *so, struct tcphdr *th, int *tlen) { + struct mbuf *m = NULL; + int drop_hdrlen = 0; KASSERT(tp != NULL && th != NULL, ("%s: ", __func__)); KASSERT(tlen > 0, ("%s: ", __func__)); - KASSERT(!HAVERCVDFIN(tp), + KASSERT(!TCPS_HAVERCVDFIN(tp->t_state), ("%s: ", __func__)); INP_LOCK_ASSERT(tp->t_inpcb); @@ -3003,7 +3012,7 @@ SOCKBUF_LOCK(&so->so_rcv); if (th->th_urp + so->so_rcv.sb_cc > sb_max) { th->th_urp = 0; /* XXX */ - thflags &= ~TH_URG; /* XXX */ + //thflags &= ~TH_URG; /* XXX */ SOCKBUF_UNLOCK(&so->so_rcv); /* XXX */ return; /* XXX */ } @@ -3110,13 +3119,13 @@ #ifdef INET6 if (isipv6) { mss = tcp_v6mssdflt; - maxmtu = tcp_maxmtu6(inc, mtuflags); + maxmtu = tcp_maxmtu6(inc, &mtuflags); min_protoh = sizeof(struct ip6_hdr) + sizeof(struct tcphdr); } else #endif { mss = tcp_mssdflt; - maxmtu = tcp_maxmtu(inc, mtuflags); + maxmtu = tcp_maxmtu(inc, &mtuflags); min_protoh = sizeof(struct tcpiphdr); } thcmtu = tcp_hc_getmtu(inc); /* IPv4 and IPv6 */ @@ -3173,14 +3182,8 @@ if (tcp_do_rfc3390) cwnd = min(4 * tp->snd_mss, max(2 * tp->snd_mss, 4380)); -#ifdef INET6 - else if (isipv6 && in6_localaddr(&inp->in6p_faddr)) - cwnd = tp->snd_mss * ss_fltsz_local; -#endif - else if (in_localaddr(inp->inp_faddr)) - cwnd = tp->snd_mss * ss_fltsz_local; else - cwnd = tp->snd_mss * ss_fltsz; + cwnd = tp->snd_mss; return (cwnd); } @@ -3194,13 +3197,8 @@ struct inpcb *inp = tp->t_inpcb; struct socket *so = inp->inp_socket; u_long bufsize; - u_long maxmtu; - int rtt, mss; - int origoffer = offer; + int rtt, mss = 0; int mtuflags = 0; -#ifdef INET6 - int isipv6 = ((inp->inp_vflag & INP_IPV6) != 0) ? 1 : 0; -#endif struct hc_metrics_lite metrics; /* @@ -3255,7 +3253,7 @@ (void)sbreserve_locked(&so->so_snd, bufsize, so, NULL); } SOCKBUF_UNLOCK(&so->so_snd); - tp->t_maxseg = mss; + //tp->t_maxseg = mss; SOCKBUF_LOCK(&so->so_rcv); if ((so->so_rcv.sb_hiwat == tcp_recvspace) && metrics.rmx_recvpipe) @@ -3276,7 +3274,7 @@ */ if (tp->t_srtt == 0 && (rtt = metrics.rmx_rtt)) { tp->t_srtt = rtt; - tp->t_rttbest = tp->t_srtt + TCP_RTT_SCALE; + //tp->t_rttbest = tp->t_srtt + TCP_RTT_SCALE; tcpstat.tcps_usedrtt++; if (metrics.rmx_rttvar) { tp->t_rttvar = metrics.rmx_rttvar; @@ -3286,9 +3284,9 @@ tp->t_rttvar = tp->t_srtt * TCP_RTTVAR_SCALE / TCP_RTT_SCALE; } - TCPT_RANGESET(tp->t_rxtcur, - ((tp->t_srtt >> 2) + tp->t_rttvar) >> 1, - tp->t_rttmin, TCPTV_REXMTMAX); + //TCPT_RANGESET(tp->t_rxtcur, + // ((tp->t_srtt >> 2) + tp->t_rttvar) >> 1, + // tp->t_rttmin, TCPTV_REXMTMAX); } if (metrics.rmx_ssthresh) { /* ==== //depot/projects/tcp_new/netinet/tcp_output.c#10 (text+ko) ==== @@ -80,6 +80,7 @@ #include <security/mac/mac_framework.h> +int path_mtu_discovery = 1; /* XXXAO: old style */ int tcp_do_pmtud = 1; SYSCTL_INT(_net_inet_tcp, OID_AUTO, path_mtu_discovery, CTLFLAG_RW, &tcp_do_pmtud, 1, "Enable Path MTU Discovery"); @@ -100,19 +101,21 @@ SYSCTL_INT(_net_inet_tcp, OID_AUTO, sendbuf_max, CTLFLAG_RW, &tcp_autosndbuf_max, 0, "Max size of automatic send buffer"); -static int -tcp_send(struct tcpcb *tp, struct tcpopt *to, int len, int rwin, int flags); -static int -tcp_retransmit(struct tcpcb *tp, int *len); -static int -tcp_send_segments(struct tcpcb *tp, struct tcphdr *ths, struct tcpopt *opt, - int off, int *olen, int optlen); -static u_int -tcp_rcv_wnd(struct tcpcb *tp, struct socket *so); -static void -tcp_snd_pace(struct tcpcp *tp); -static void -tcp_options(struct tcpcb *tp, struct tcpopt *to, int flags); +static int ip_optlen(struct inpcb *inp); +static int tcp_send(struct tcpcb *tp, struct socket *so, + struct tcpopt *to, u_char *opt, int len, + int optlen, int rwin, int flags); +static int tcp_retransmit(struct tcpcb *tp, struct socket *so, + struct tcpopt *to, u_char *opt, int *len, + int optlen, int rwin, int flags); +static int tcp_send_segments(struct tcpcb *tp, struct tcphdr *ths, + u_char *opt, int off, int *olen, int optlen); +static u_int tcp_rcv_wnd(struct tcpcb *tp, struct socket *so); +static int tcp_snd_pace(struct tcpcb *tp, int len); +static void tcp_snd_autoscale(struct tcpcb *tp, struct socket *so, + int swin); +static int tcp_options(struct tcpcb *tp, struct socket *so, + struct tcpopt *to, u_char *opt, int flags); /* * Tcp output routine: figure out what should be sent and send it. @@ -139,13 +142,14 @@ * 7. Send based on flags */ int -tcp_output(struct tcpcb *tp, int reason) +tcp_output(struct tcpcb *tp) { - int off, flags, error, optlen; - tcp_win len, recwin, swin; + int flags, error, optlen = 0; + tcp_win len; + int duna, swnd, cwnd, dlen, inflight, rwin; + int tcp_min_idle = 1; /* XXXAO */ struct inpcb *inp = tp->t_inpcb; struct socket *so = inp->inp_socket; - struct tcphdr ths; struct tcpopt to; u_char opt[TCP_MAXOLEN]; #ifdef TCP_SIGNATURE @@ -159,10 +163,10 @@ KASSERT(tp->t_state < TCPS_TIME_WAIT, ("%s: TCPS_TIME_WAIT invalid", __func__)); - KASSERT(SEQ_GEQ(tp->snd_rxmit, tp->snd_una), - ("%s: snd_rxmit < snd_una", __func__)) - KASSERT(SEQ_LEQ(tp->snd_rxmit, tp->snd_nxt), - ("%s: snd_rxmit > snd_nxt", __func__)) + //KASSERT(SEQ_GEQ(tp->snd_rxmit, tp->snd_una), + // ("%s: snd_rxmit < snd_una", __func__)); + //KASSERT(SEQ_LEQ(tp->snd_rxmit, tp->snd_nxt), + // ("%s: snd_rxmit > snd_nxt", __func__)); /* * Get standard flags. Removal of inappropriate flags for a @@ -251,15 +255,15 @@ */ if (tp->snd_nxt == tp->snd_una && (ticks - tp->t_rcvtime) >= max(tp->t_rxtcur, tcp_min_idle)) { - tp->snd_cwnd = tcp_init_cwnd(tp); + tp->snd_cwnd = tp->snd_mss /*tcp_init_cwnd(tp)*/; } break; case TP_SENDING: break; case TP_LOSSRECOV: case TP_REXMT: - tcp_retransmit(tp, &len); - if (len = 0) + error = tcp_retransmit(tp, so, &to, &opt[0], &len, optlen, rwin, flags); + if (len == 0) return (0); break; case TP_PERSIST: @@ -478,13 +482,21 @@ return (0); send: - tcp_options(tp, &to, flags); - return (tcp_send(tp, &to, flags)); + optlen = tcp_options(tp, so, &to, &opt[0], flags); + error = tcp_send(tp, so, &to, &opt[0], len, optlen, rwin, flags); + + if (!error) + tcp_snd_autoscale(tp, so, swnd); + + return (error); } static int -tcp_send(struct tcpcb *tp, struct tcpopt *to, int len, int rwin, int flags) +tcp_send(struct tcpcb *tp, struct socket *so, struct tcpopt *to, + u_char *opt, int len, int optlen, int rwin, int flags) { + int off, error; + struct tcphdr ths, *th = &ths; KASSERT((flags & TH_SYN) || (tp->t_flags & TF_SENTSYN), ("%s: retransmitting SYN", __func__)); @@ -552,7 +564,7 @@ /* * Integrate FIN into sequence space. */ - if ((flags & TH_FIN) && !(tp-t_flags & TF_SENTFIN)) { + if ((flags & TH_FIN) && !(tp->t_flags & TF_SENTFIN)) { tp->snd_nxt++; tp->t_flags |= TF_SENTFIN; } @@ -584,7 +596,7 @@ * Clear related flags and disarm the delayed ACK timer. */ tp->t_flags &= ~(TF_ACKNOW | TF_DELACK); - if (SEQ_LT(tp->snd_lastack, tp->rcv_nxt) + if (SEQ_LT(tp->snd_lastack, tp->rcv_nxt)) tp->snd_lastack = tp->rcv_nxt; if (tcp_timer_active(tp, TT_DELACK)) tcp_timer_activate(tp, TT_DELACK, 0); @@ -620,9 +632,9 @@ * Time this transmission if not a retransmission and * not currently timing anything. */ - if (tp->t_rtttime == 0) { - tp->t_rtttime = ticks; - tp->t_rtseq = startseq; + if (tp->snd_rtts == 0) { + tp->snd_rtts = ticks; + tp->snd_rtseq = tp->snd_nxt - len; /* XXXAO */ tcpstat.tcps_segstimed++; } @@ -635,7 +647,7 @@ * of retransmit time. */ if (!tcp_timer_active(tp, TT_REXMT) && - (tp->snd_nxt != tp->snd_una))) { + (tp->snd_nxt != tp->snd_una)) { if (tcp_timer_active(tp, TT_PERSIST)) { tcp_timer_activate(tp, TT_PERSIST, 0); tp->t_rxtshift = 0; @@ -667,7 +679,7 @@ if (!tcp_timer_active(tp, TT_REXMT) && !tcp_timer_active(tp, TT_PERSIST)) tcp_timer_activate(tp, TT_REXMT, tp->t_rxtcur); - tp->snd_cwnd = tp->t_maxseg; + tp->snd_cwnd = tp->snd_mss; error = 0; break; @@ -687,7 +699,7 @@ * immediatly retry with MSS sized segments generated * by this function. */ - if (tso) + if (tp->t_flags & TF_TSO) tp->t_flags &= ~TF_TSO; else tcp_mtudisc(tp->t_inpcb, 0); @@ -724,11 +736,11 @@ * * XXXAO: Account for TSO. */ - if ((tp->t_flags & TF_FORCEDATA) && len == 1) + if ((tp->t_flags & TF_FORCEDATA) && len == 1) { tcpstat.tcps_sndprobe++; - else if (SEQ_LT(tp->snd_nxt, tp->snd_max)) { - tcpstat.tcps_sndrexmitpack++; - tcpstat.tcps_sndrexmitbyte += len; + //} else if (SEQ_LT(tp->snd_nxt, tp->snd_max)) { + // tcpstat.tcps_sndrexmitpack++; + // tcpstat.tcps_sndrexmitbyte += len; } else { tcpstat.tcps_sndpack++; tcpstat.tcps_sndbyte += len; @@ -752,9 +764,11 @@ * from the normal transmit case as the logic is quite a bit different. */ static int -tcp_retransmit(struct tcpcb *tp, int *len) +tcp_retransmit(struct tcpcb *tp, struct socket *so, struct tcpopt *to, + u_char *opt, int *len, int optlen, int rwin, int flags) { - struct tcphdr ths, *th; + int error, off, rlen = 0; + struct tcphdr ths, *th = &ths; /* * Retransmit over the SACK holes. @@ -780,12 +794,12 @@ /* Limited transmit */ if (tp->snd_dupack < 3) - *len = min(len, tp->snd_mss); /* one mss */ + *len = min(*len, tp->snd_mss); /* one mss */ else *len = 0; if (tp->snd_dupack < 3) - return; + return (0); /* * XXXAO: Temporary. @@ -811,10 +825,10 @@ SOCKBUF_LOCK(&so->so_snd); off = min(tp->snd_rxmit - tp->snd_una, so->so_snd.sb_cc); - error = tcp_send_segments(tp, &ths, opt, off, olen, optlen); + error = tcp_send_segments(tp, &ths, opt, off, &rlen, optlen); SOCKBUF_UNLOCK(&so->so_snd); - return; + return (error); } /* @@ -824,17 +838,18 @@ * XXXAO: We may leave some left-overs if we can't drain the whole send buffer. */ static int -tcp_send_segments(struct tcpcb *tp, struct tcphdr *ths, struct tcpopt *opt, +tcp_send_segments(struct tcpcb *tp, struct tcphdr *ths, u_char *opt, int off, int *olen, int optlen) { - int len, slen, hdrs, hdrlen, linkhdr, optlen, ipoptlen; + int len, slen, hdrs, hdrlen, linkhdr, ipoptlen; int error = 0; + struct mbuf *m; struct tcphdr *th; - struct ip *ip; + struct ip *ip = NULL; struct inpcb *inp = tp->t_inpcb; struct socket *so = inp->inp_socket; #ifdef INET6 - struct ip6_hdr *ip6; + struct ip6_hdr *ip6 = NULL; int isipv6 = (inp->inp_vflag & INP_IPV6) != 0; #endif INP_LOCK_ASSERT(tp->t_inpcb); @@ -905,7 +920,7 @@ * is already reflected in the MSS. */ if (tp->t_flags & TF_TSO) { - if (len > TCP_MAXWIN - (hdrs + linkhdr) { + if (len > TCP_MAXWIN - (hdrs + linkhdr)) { slen = TCP_MAXWIN - (hdrs + linkhdr); slen -= slen % (tp->snd_mss - optlen); } else @@ -975,7 +990,7 @@ * trailing space is available, or we attach an mbuf chain * with references to the mbuf storage in the socket buffer. */ - if (slen <= M_TAILINGSPACE(m)) { >>> TRUNCATED FOR MAIL (1000 lines) <<<help
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200907021118.n62BIktH076734>