1 /*- 2 * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1994, 1995 3 * The Regents of the University of California. All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 4. Neither the name of the University nor the names of its contributors 14 * may be used to endorse or promote products derived from this software 15 * without specific prior written permission. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27 * SUCH DAMAGE. 28 * 29 * @(#)tcp_input.c 8.12 (Berkeley) 5/24/95 30 */ 31 32 #include <sys/cdefs.h> 33 __FBSDID("$FreeBSD$"); 34 35 #include "opt_inet.h" 36 #include "opt_inet6.h" 37 #include "opt_tcpdebug.h" 38 39 #include <sys/param.h> 40 #include <sys/kernel.h> 41 #include <sys/malloc.h> 42 #include <sys/mbuf.h> 43 #include <sys/socket.h> 44 #include <sys/socketvar.h> 45 #include <sys/sysctl.h> 46 #include <sys/syslog.h> 47 #include <sys/systm.h> 48 #include <sys/vimage.h> 49 50 #include <vm/uma.h> 51 52 #include <net/if.h> 53 #include <net/route.h> 54 55 #include <netinet/in.h> 56 #include <netinet/in_pcb.h> 57 #include <netinet/in_systm.h> 58 #include <netinet/in_var.h> 59 #include <netinet/ip.h> 60 #include <netinet/ip_var.h> 61 #include <netinet/ip_options.h> 62 #include <netinet/ip6.h> 63 #include <netinet6/in6_pcb.h> 64 #include <netinet6/ip6_var.h> 65 #include <netinet6/nd6.h> 66 #include <netinet/tcp.h> 67 #include <netinet/tcp_fsm.h> 68 #include <netinet/tcp_seq.h> 69 #include <netinet/tcp_timer.h> 70 #include <netinet/tcp_var.h> 71 #include <netinet6/tcp6_var.h> 72 #include <netinet/tcpip.h> 73 #ifdef TCPDEBUG 74 #include <netinet/tcp_debug.h> 75 #endif /* TCPDEBUG */ 76 77 SYSCTL_NODE(_net_inet_tcp, OID_AUTO, reass, CTLFLAG_RW, 0, 78 "TCP Segment Reassembly Queue"); 79 80 static int tcp_reass_maxseg = 0; 81 SYSCTL_INT(_net_inet_tcp_reass, OID_AUTO, maxsegments, CTLFLAG_RDTUN, 82 &tcp_reass_maxseg, 0, 83 "Global maximum number of TCP Segments in Reassembly Queue"); 84 85 int tcp_reass_qsize = 0; 86 SYSCTL_INT(_net_inet_tcp_reass, OID_AUTO, cursegments, CTLFLAG_RD, 87 &tcp_reass_qsize, 0, 88 "Global number of TCP Segments currently in Reassembly Queue"); 89 90 static int tcp_reass_maxqlen = 48; 91 SYSCTL_INT(_net_inet_tcp_reass, OID_AUTO, maxqlen, CTLFLAG_RW, 92 &tcp_reass_maxqlen, 0, 93 "Maximum number of TCP Segments per individual Reassembly Queue"); 94 95 static int tcp_reass_overflows = 0; 96 SYSCTL_INT(_net_inet_tcp_reass, OID_AUTO, overflows, CTLFLAG_RD, 97 &tcp_reass_overflows, 0, 98 "Global number of TCP Segment Reassembly Queue Overflows"); 99 100 /* Initialize TCP reassembly queue */ 101 static void 102 tcp_reass_zone_change(void *tag) 103 { 104 105 V_tcp_reass_maxseg = nmbclusters / 16; 106 uma_zone_set_max(tcp_reass_zone, V_tcp_reass_maxseg); 107 } 108 109 uma_zone_t tcp_reass_zone; 110 111 void 112 tcp_reass_init(void) 113 { 114 115 V_tcp_reass_maxseg = nmbclusters / 16; 116 TUNABLE_INT_FETCH("net.inet.tcp.reass.maxsegments", 117 &V_tcp_reass_maxseg); 118 tcp_reass_zone = uma_zcreate("tcpreass", sizeof (struct tseg_qent), 119 NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE); 120 uma_zone_set_max(tcp_reass_zone, V_tcp_reass_maxseg); 121 EVENTHANDLER_REGISTER(nmbclusters_change, 122 tcp_reass_zone_change, NULL, EVENTHANDLER_PRI_ANY); 123 } 124 125 int 126 tcp_reass(struct tcpcb *tp, struct tcphdr *th, int *tlenp, struct mbuf *m) 127 { 128 struct tseg_qent *q; 129 struct tseg_qent *p = NULL; 130 struct tseg_qent *nq; 131 struct tseg_qent *te = NULL; 132 struct socket *so = tp->t_inpcb->inp_socket; 133 int flags; 134 135 INP_WLOCK_ASSERT(tp->t_inpcb); 136 137 /* 138 * XXX: tcp_reass() is rather inefficient with its data structures 139 * and should be rewritten (see NetBSD for optimizations). While 140 * doing that it should move to its own file tcp_reass.c. 141 */ 142 143 /* 144 * Call with th==NULL after become established to 145 * force pre-ESTABLISHED data up to user socket. 146 */ 147 if (th == NULL) 148 goto present; 149 150 /* 151 * Limit the number of segments in the reassembly queue to prevent 152 * holding on to too many segments (and thus running out of mbufs). 153 * Make sure to let the missing segment through which caused this 154 * queue. Always keep one global queue entry spare to be able to 155 * process the missing segment. 156 */ 157 if (th->th_seq != tp->rcv_nxt && 158 (V_tcp_reass_qsize + 1 >= V_tcp_reass_maxseg || 159 tp->t_segqlen >= V_tcp_reass_maxqlen)) { 160 V_tcp_reass_overflows++; 161 V_tcpstat.tcps_rcvmemdrop++; 162 m_freem(m); 163 *tlenp = 0; 164 return (0); 165 } 166 167 /* 168 * Allocate a new queue entry. If we can't, or hit the zone limit 169 * just drop the pkt. 170 */ 171 te = uma_zalloc(tcp_reass_zone, M_NOWAIT); 172 if (te == NULL) { 173 V_tcpstat.tcps_rcvmemdrop++; 174 m_freem(m); 175 *tlenp = 0; 176 return (0); 177 } 178 tp->t_segqlen++; 179 V_tcp_reass_qsize++; 180 181 /* 182 * Find a segment which begins after this one does. 183 */ 184 LIST_FOREACH(q, &tp->t_segq, tqe_q) { 185 if (SEQ_GT(q->tqe_th->th_seq, th->th_seq)) 186 break; 187 p = q; 188 } 189 190 /* 191 * If there is a preceding segment, it may provide some of 192 * our data already. If so, drop the data from the incoming 193 * segment. If it provides all of our data, drop us. 194 */ 195 if (p != NULL) { 196 int i; 197 /* conversion to int (in i) handles seq wraparound */ 198 i = p->tqe_th->th_seq + p->tqe_len - th->th_seq; 199 if (i > 0) { 200 if (i >= *tlenp) { 201 V_tcpstat.tcps_rcvduppack++; 202 V_tcpstat.tcps_rcvdupbyte += *tlenp; 203 m_freem(m); 204 uma_zfree(tcp_reass_zone, te); 205 tp->t_segqlen--; 206 V_tcp_reass_qsize--; 207 /* 208 * Try to present any queued data 209 * at the left window edge to the user. 210 * This is needed after the 3-WHS 211 * completes. 212 */ 213 goto present; /* ??? */ 214 } 215 m_adj(m, i); 216 *tlenp -= i; 217 th->th_seq += i; 218 } 219 } 220 V_tcpstat.tcps_rcvoopack++; 221 V_tcpstat.tcps_rcvoobyte += *tlenp; 222 223 /* 224 * While we overlap succeeding segments trim them or, 225 * if they are completely covered, dequeue them. 226 */ 227 while (q) { 228 int i = (th->th_seq + *tlenp) - q->tqe_th->th_seq; 229 if (i <= 0) 230 break; 231 if (i < q->tqe_len) { 232 q->tqe_th->th_seq += i; 233 q->tqe_len -= i; 234 m_adj(q->tqe_m, i); 235 break; 236 } 237 238 nq = LIST_NEXT(q, tqe_q); 239 LIST_REMOVE(q, tqe_q); 240 m_freem(q->tqe_m); 241 uma_zfree(tcp_reass_zone, q); 242 tp->t_segqlen--; 243 V_tcp_reass_qsize--; 244 q = nq; 245 } 246 247 /* Insert the new segment queue entry into place. */ 248 te->tqe_m = m; 249 te->tqe_th = th; 250 te->tqe_len = *tlenp; 251 252 if (p == NULL) { 253 LIST_INSERT_HEAD(&tp->t_segq, te, tqe_q); 254 } else { 255 LIST_INSERT_AFTER(p, te, tqe_q); 256 } 257 258 present: 259 /* 260 * Present data to user, advancing rcv_nxt through 261 * completed sequence space. 262 */ 263 if (!TCPS_HAVEESTABLISHED(tp->t_state)) 264 return (0); 265 q = LIST_FIRST(&tp->t_segq); 266 if (!q || q->tqe_th->th_seq != tp->rcv_nxt) 267 return (0); 268 SOCKBUF_LOCK(&so->so_rcv); 269 do { 270 tp->rcv_nxt += q->tqe_len; 271 flags = q->tqe_th->th_flags & TH_FIN; 272 nq = LIST_NEXT(q, tqe_q); 273 LIST_REMOVE(q, tqe_q); 274 if (so->so_rcv.sb_state & SBS_CANTRCVMORE) 275 m_freem(q->tqe_m); 276 else 277 sbappendstream_locked(&so->so_rcv, q->tqe_m); 278 uma_zfree(tcp_reass_zone, q); 279 tp->t_segqlen--; 280 V_tcp_reass_qsize--; 281 q = nq; 282 } while (q && q->tqe_th->th_seq == tp->rcv_nxt); 283 ND6_HINT(tp); 284 sorwakeup_locked(so); 285 return (flags); 286 } 287