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(V_tcp_reass_zone, V_tcp_reass_maxseg); 112 } 113 114 #ifdef VIMAGE_GLOBALS 115 uma_zone_t tcp_reass_zone; 116 #endif 117 118 void 119 tcp_reass_init(void) 120 { 121 INIT_VNET_INET(curvnet); 122 123 V_tcp_reass_maxseg = 0; 124 V_tcp_reass_qsize = 0; 125 V_tcp_reass_maxqlen = 48; 126 V_tcp_reass_overflows = 0; 127 128 V_tcp_reass_maxseg = nmbclusters / 16; 129 TUNABLE_INT_FETCH("net.inet.tcp.reass.maxsegments", 130 &V_tcp_reass_maxseg); 131 V_tcp_reass_zone = uma_zcreate("tcpreass", sizeof (struct tseg_qent), 132 NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE); 133 uma_zone_set_max(V_tcp_reass_zone, V_tcp_reass_maxseg); 134 EVENTHANDLER_REGISTER(nmbclusters_change, 135 tcp_reass_zone_change, NULL, EVENTHANDLER_PRI_ANY); 136 } 137 138 int 139 tcp_reass(struct tcpcb *tp, struct tcphdr *th, int *tlenp, struct mbuf *m) 140 { 141 INIT_VNET_INET(curvnet); 142 struct tseg_qent *q; 143 struct tseg_qent *p = NULL; 144 struct tseg_qent *nq; 145 struct tseg_qent *te = NULL; 146 struct socket *so = tp->t_inpcb->inp_socket; 147 int flags; 148 149 INP_WLOCK_ASSERT(tp->t_inpcb); 150 151 /* 152 * XXX: tcp_reass() is rather inefficient with its data structures 153 * and should be rewritten (see NetBSD for optimizations). While 154 * doing that it should move to its own file tcp_reass.c. 155 */ 156 157 /* 158 * Call with th==NULL after become established to 159 * force pre-ESTABLISHED data up to user socket. 160 */ 161 if (th == NULL) 162 goto present; 163 164 /* 165 * Limit the number of segments in the reassembly queue to prevent 166 * holding on to too many segments (and thus running out of mbufs). 167 * Make sure to let the missing segment through which caused this 168 * queue. Always keep one global queue entry spare to be able to 169 * process the missing segment. 170 */ 171 if (th->th_seq != tp->rcv_nxt && 172 (V_tcp_reass_qsize + 1 >= V_tcp_reass_maxseg || 173 tp->t_segqlen >= V_tcp_reass_maxqlen)) { 174 V_tcp_reass_overflows++; 175 TCPSTAT_INC(tcps_rcvmemdrop); 176 m_freem(m); 177 *tlenp = 0; 178 return (0); 179 } 180 181 /* 182 * Allocate a new queue entry. If we can't, or hit the zone limit 183 * just drop the pkt. 184 */ 185 te = uma_zalloc(V_tcp_reass_zone, M_NOWAIT); 186 if (te == NULL) { 187 TCPSTAT_INC(tcps_rcvmemdrop); 188 m_freem(m); 189 *tlenp = 0; 190 return (0); 191 } 192 tp->t_segqlen++; 193 V_tcp_reass_qsize++; 194 195 /* 196 * Find a segment which begins after this one does. 197 */ 198 LIST_FOREACH(q, &tp->t_segq, tqe_q) { 199 if (SEQ_GT(q->tqe_th->th_seq, th->th_seq)) 200 break; 201 p = q; 202 } 203 204 /* 205 * If there is a preceding segment, it may provide some of 206 * our data already. If so, drop the data from the incoming 207 * segment. If it provides all of our data, drop us. 208 */ 209 if (p != NULL) { 210 int i; 211 /* conversion to int (in i) handles seq wraparound */ 212 i = p->tqe_th->th_seq + p->tqe_len - th->th_seq; 213 if (i > 0) { 214 if (i >= *tlenp) { 215 TCPSTAT_INC(tcps_rcvduppack); 216 TCPSTAT_ADD(tcps_rcvdupbyte, *tlenp); 217 m_freem(m); 218 uma_zfree(V_tcp_reass_zone, te); 219 tp->t_segqlen--; 220 V_tcp_reass_qsize--; 221 /* 222 * Try to present any queued data 223 * at the left window edge to the user. 224 * This is needed after the 3-WHS 225 * completes. 226 */ 227 goto present; /* ??? */ 228 } 229 m_adj(m, i); 230 *tlenp -= i; 231 th->th_seq += i; 232 } 233 } 234 TCPSTAT_INC(tcps_rcvoopack); 235 TCPSTAT_ADD(tcps_rcvoobyte, *tlenp); 236 237 /* 238 * While we overlap succeeding segments trim them or, 239 * if they are completely covered, dequeue them. 240 */ 241 while (q) { 242 int i = (th->th_seq + *tlenp) - q->tqe_th->th_seq; 243 if (i <= 0) 244 break; 245 if (i < q->tqe_len) { 246 q->tqe_th->th_seq += i; 247 q->tqe_len -= i; 248 m_adj(q->tqe_m, i); 249 break; 250 } 251 252 nq = LIST_NEXT(q, tqe_q); 253 LIST_REMOVE(q, tqe_q); 254 m_freem(q->tqe_m); 255 uma_zfree(V_tcp_reass_zone, q); 256 tp->t_segqlen--; 257 V_tcp_reass_qsize--; 258 q = nq; 259 } 260 261 /* Insert the new segment queue entry into place. */ 262 te->tqe_m = m; 263 te->tqe_th = th; 264 te->tqe_len = *tlenp; 265 266 if (p == NULL) { 267 LIST_INSERT_HEAD(&tp->t_segq, te, tqe_q); 268 } else { 269 LIST_INSERT_AFTER(p, te, tqe_q); 270 } 271 272 present: 273 /* 274 * Present data to user, advancing rcv_nxt through 275 * completed sequence space. 276 */ 277 if (!TCPS_HAVEESTABLISHED(tp->t_state)) 278 return (0); 279 q = LIST_FIRST(&tp->t_segq); 280 if (!q || q->tqe_th->th_seq != tp->rcv_nxt) 281 return (0); 282 SOCKBUF_LOCK(&so->so_rcv); 283 do { 284 tp->rcv_nxt += q->tqe_len; 285 flags = q->tqe_th->th_flags & TH_FIN; 286 nq = LIST_NEXT(q, tqe_q); 287 LIST_REMOVE(q, tqe_q); 288 if (so->so_rcv.sb_state & SBS_CANTRCVMORE) 289 m_freem(q->tqe_m); 290 else 291 sbappendstream_locked(&so->so_rcv, q->tqe_m); 292 uma_zfree(V_tcp_reass_zone, q); 293 tp->t_segqlen--; 294 V_tcp_reass_qsize--; 295 q = nq; 296 } while (q && q->tqe_th->th_seq == tp->rcv_nxt); 297 ND6_HINT(tp); 298 sorwakeup_locked(so); 299 return (flags); 300 } 301