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 #include <netinet/vinet.h> 77 78 #ifdef VIMAGE_GLOBALS 79 static int tcp_reass_maxseg; 80 int tcp_reass_qsize; 81 static int tcp_reass_maxqlen; 82 static int tcp_reass_overflows; 83 #endif 84 85 SYSCTL_NODE(_net_inet_tcp, OID_AUTO, reass, CTLFLAG_RW, 0, 86 "TCP Segment Reassembly Queue"); 87 88 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_tcp_reass, OID_AUTO, maxsegments, 89 CTLFLAG_RDTUN, tcp_reass_maxseg, 0, 90 "Global maximum number of TCP Segments in Reassembly Queue"); 91 92 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_tcp_reass, OID_AUTO, cursegments, 93 CTLFLAG_RD, tcp_reass_qsize, 0, 94 "Global number of TCP Segments currently in Reassembly Queue"); 95 96 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_tcp_reass, OID_AUTO, maxqlen, 97 CTLFLAG_RW, tcp_reass_maxqlen, 0, 98 "Maximum number of TCP Segments per individual Reassembly Queue"); 99 100 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_tcp_reass, OID_AUTO, overflows, 101 CTLFLAG_RD, tcp_reass_overflows, 0, 102 "Global number of TCP Segment Reassembly Queue Overflows"); 103 104 /* Initialize TCP reassembly queue */ 105 static void 106 tcp_reass_zone_change(void *tag) 107 { 108 INIT_VNET_INET(curvnet); 109 110 V_tcp_reass_maxseg = nmbclusters / 16; 111 uma_zone_set_max(tcp_reass_zone, V_tcp_reass_maxseg); 112 } 113 114 uma_zone_t tcp_reass_zone; 115 116 void 117 tcp_reass_init(void) 118 { 119 INIT_VNET_INET(curvnet); 120 121 V_tcp_reass_maxseg = 0; 122 V_tcp_reass_qsize = 0; 123 V_tcp_reass_maxqlen = 48; 124 V_tcp_reass_overflows = 0; 125 126 V_tcp_reass_maxseg = nmbclusters / 16; 127 TUNABLE_INT_FETCH("net.inet.tcp.reass.maxsegments", 128 &V_tcp_reass_maxseg); 129 tcp_reass_zone = uma_zcreate("tcpreass", sizeof (struct tseg_qent), 130 NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE); 131 uma_zone_set_max(tcp_reass_zone, V_tcp_reass_maxseg); 132 EVENTHANDLER_REGISTER(nmbclusters_change, 133 tcp_reass_zone_change, NULL, EVENTHANDLER_PRI_ANY); 134 } 135 136 int 137 tcp_reass(struct tcpcb *tp, struct tcphdr *th, int *tlenp, struct mbuf *m) 138 { 139 INIT_VNET_INET(curvnet); 140 struct tseg_qent *q; 141 struct tseg_qent *p = NULL; 142 struct tseg_qent *nq; 143 struct tseg_qent *te = NULL; 144 struct socket *so = tp->t_inpcb->inp_socket; 145 int flags; 146 147 INP_WLOCK_ASSERT(tp->t_inpcb); 148 149 /* 150 * XXX: tcp_reass() is rather inefficient with its data structures 151 * and should be rewritten (see NetBSD for optimizations). While 152 * doing that it should move to its own file tcp_reass.c. 153 */ 154 155 /* 156 * Call with th==NULL after become established to 157 * force pre-ESTABLISHED data up to user socket. 158 */ 159 if (th == NULL) 160 goto present; 161 162 /* 163 * Limit the number of segments in the reassembly queue to prevent 164 * holding on to too many segments (and thus running out of mbufs). 165 * Make sure to let the missing segment through which caused this 166 * queue. Always keep one global queue entry spare to be able to 167 * process the missing segment. 168 */ 169 if (th->th_seq != tp->rcv_nxt && 170 (V_tcp_reass_qsize + 1 >= V_tcp_reass_maxseg || 171 tp->t_segqlen >= V_tcp_reass_maxqlen)) { 172 V_tcp_reass_overflows++; 173 V_tcpstat.tcps_rcvmemdrop++; 174 m_freem(m); 175 *tlenp = 0; 176 return (0); 177 } 178 179 /* 180 * Allocate a new queue entry. If we can't, or hit the zone limit 181 * just drop the pkt. 182 */ 183 te = uma_zalloc(tcp_reass_zone, M_NOWAIT); 184 if (te == NULL) { 185 V_tcpstat.tcps_rcvmemdrop++; 186 m_freem(m); 187 *tlenp = 0; 188 return (0); 189 } 190 tp->t_segqlen++; 191 V_tcp_reass_qsize++; 192 193 /* 194 * Find a segment which begins after this one does. 195 */ 196 LIST_FOREACH(q, &tp->t_segq, tqe_q) { 197 if (SEQ_GT(q->tqe_th->th_seq, th->th_seq)) 198 break; 199 p = q; 200 } 201 202 /* 203 * If there is a preceding segment, it may provide some of 204 * our data already. If so, drop the data from the incoming 205 * segment. If it provides all of our data, drop us. 206 */ 207 if (p != NULL) { 208 int i; 209 /* conversion to int (in i) handles seq wraparound */ 210 i = p->tqe_th->th_seq + p->tqe_len - th->th_seq; 211 if (i > 0) { 212 if (i >= *tlenp) { 213 V_tcpstat.tcps_rcvduppack++; 214 V_tcpstat.tcps_rcvdupbyte += *tlenp; 215 m_freem(m); 216 uma_zfree(tcp_reass_zone, te); 217 tp->t_segqlen--; 218 V_tcp_reass_qsize--; 219 /* 220 * Try to present any queued data 221 * at the left window edge to the user. 222 * This is needed after the 3-WHS 223 * completes. 224 */ 225 goto present; /* ??? */ 226 } 227 m_adj(m, i); 228 *tlenp -= i; 229 th->th_seq += i; 230 } 231 } 232 V_tcpstat.tcps_rcvoopack++; 233 V_tcpstat.tcps_rcvoobyte += *tlenp; 234 235 /* 236 * While we overlap succeeding segments trim them or, 237 * if they are completely covered, dequeue them. 238 */ 239 while (q) { 240 int i = (th->th_seq + *tlenp) - q->tqe_th->th_seq; 241 if (i <= 0) 242 break; 243 if (i < q->tqe_len) { 244 q->tqe_th->th_seq += i; 245 q->tqe_len -= i; 246 m_adj(q->tqe_m, i); 247 break; 248 } 249 250 nq = LIST_NEXT(q, tqe_q); 251 LIST_REMOVE(q, tqe_q); 252 m_freem(q->tqe_m); 253 uma_zfree(tcp_reass_zone, q); 254 tp->t_segqlen--; 255 V_tcp_reass_qsize--; 256 q = nq; 257 } 258 259 /* Insert the new segment queue entry into place. */ 260 te->tqe_m = m; 261 te->tqe_th = th; 262 te->tqe_len = *tlenp; 263 264 if (p == NULL) { 265 LIST_INSERT_HEAD(&tp->t_segq, te, tqe_q); 266 } else { 267 LIST_INSERT_AFTER(p, te, tqe_q); 268 } 269 270 present: 271 /* 272 * Present data to user, advancing rcv_nxt through 273 * completed sequence space. 274 */ 275 if (!TCPS_HAVEESTABLISHED(tp->t_state)) 276 return (0); 277 q = LIST_FIRST(&tp->t_segq); 278 if (!q || q->tqe_th->th_seq != tp->rcv_nxt) 279 return (0); 280 SOCKBUF_LOCK(&so->so_rcv); 281 do { 282 tp->rcv_nxt += q->tqe_len; 283 flags = q->tqe_th->th_flags & TH_FIN; 284 nq = LIST_NEXT(q, tqe_q); 285 LIST_REMOVE(q, tqe_q); 286 if (so->so_rcv.sb_state & SBS_CANTRCVMORE) 287 m_freem(q->tqe_m); 288 else 289 sbappendstream_locked(&so->so_rcv, q->tqe_m); 290 uma_zfree(tcp_reass_zone, q); 291 tp->t_segqlen--; 292 V_tcp_reass_qsize--; 293 q = nq; 294 } while (q && q->tqe_th->th_seq == tp->rcv_nxt); 295 ND6_HINT(tp); 296 sorwakeup_locked(so); 297 return (flags); 298 } 299