| @@ -1,3232 +1,3232 @@ | | | @@ -1,3232 +1,3232 @@ |
1 | /* $NetBSD: tcp_input.c,v 1.295 2009/03/18 16:00:22 cegger Exp $ */ | | 1 | /* $NetBSD: tcp_input.c,v 1.296 2009/06/20 17:29:31 christos Exp $ */ |
2 | | | 2 | |
3 | /* | | 3 | /* |
4 | * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. | | 4 | * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. |
5 | * All rights reserved. | | 5 | * All rights reserved. |
6 | * | | 6 | * |
7 | * Redistribution and use in source and binary forms, with or without | | 7 | * Redistribution and use in source and binary forms, with or without |
8 | * modification, are permitted provided that the following conditions | | 8 | * modification, are permitted provided that the following conditions |
9 | * are met: | | 9 | * are met: |
10 | * 1. Redistributions of source code must retain the above copyright | | 10 | * 1. Redistributions of source code must retain the above copyright |
11 | * notice, this list of conditions and the following disclaimer. | | 11 | * notice, this list of conditions and the following disclaimer. |
12 | * 2. Redistributions in binary form must reproduce the above copyright | | 12 | * 2. Redistributions in binary form must reproduce the above copyright |
13 | * notice, this list of conditions and the following disclaimer in the | | 13 | * notice, this list of conditions and the following disclaimer in the |
14 | * documentation and/or other materials provided with the distribution. | | 14 | * documentation and/or other materials provided with the distribution. |
15 | * 3. Neither the name of the project nor the names of its contributors | | 15 | * 3. Neither the name of the project nor the names of its contributors |
16 | * may be used to endorse or promote products derived from this software | | 16 | * may be used to endorse or promote products derived from this software |
17 | * without specific prior written permission. | | 17 | * without specific prior written permission. |
18 | * | | 18 | * |
19 | * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND | | 19 | * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND |
20 | * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE | | 20 | * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE |
21 | * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE | | 21 | * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE |
22 | * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE | | 22 | * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE |
23 | * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL | | 23 | * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL |
24 | * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS | | 24 | * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS |
25 | * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) | | 25 | * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) |
26 | * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT | | 26 | * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT |
27 | * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY | | 27 | * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY |
28 | * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF | | 28 | * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF |
29 | * SUCH DAMAGE. | | 29 | * SUCH DAMAGE. |
30 | */ | | 30 | */ |
31 | | | 31 | |
32 | /* | | 32 | /* |
33 | * @(#)COPYRIGHT 1.1 (NRL) 17 January 1995 | | 33 | * @(#)COPYRIGHT 1.1 (NRL) 17 January 1995 |
34 | * | | 34 | * |
35 | * NRL grants permission for redistribution and use in source and binary | | 35 | * NRL grants permission for redistribution and use in source and binary |
36 | * forms, with or without modification, of the software and documentation | | 36 | * forms, with or without modification, of the software and documentation |
37 | * created at NRL provided that the following conditions are met: | | 37 | * created at NRL provided that the following conditions are met: |
38 | * | | 38 | * |
39 | * 1. Redistributions of source code must retain the above copyright | | 39 | * 1. Redistributions of source code must retain the above copyright |
40 | * notice, this list of conditions and the following disclaimer. | | 40 | * notice, this list of conditions and the following disclaimer. |
41 | * 2. Redistributions in binary form must reproduce the above copyright | | 41 | * 2. Redistributions in binary form must reproduce the above copyright |
42 | * notice, this list of conditions and the following disclaimer in the | | 42 | * notice, this list of conditions and the following disclaimer in the |
43 | * documentation and/or other materials provided with the distribution. | | 43 | * documentation and/or other materials provided with the distribution. |
44 | * 3. All advertising materials mentioning features or use of this software | | 44 | * 3. All advertising materials mentioning features or use of this software |
45 | * must display the following acknowledgements: | | 45 | * must display the following acknowledgements: |
46 | * This product includes software developed by the University of | | 46 | * This product includes software developed by the University of |
47 | * California, Berkeley and its contributors. | | 47 | * California, Berkeley and its contributors. |
48 | * This product includes software developed at the Information | | 48 | * This product includes software developed at the Information |
49 | * Technology Division, US Naval Research Laboratory. | | 49 | * Technology Division, US Naval Research Laboratory. |
50 | * 4. Neither the name of the NRL nor the names of its contributors | | 50 | * 4. Neither the name of the NRL nor the names of its contributors |
51 | * may be used to endorse or promote products derived from this software | | 51 | * may be used to endorse or promote products derived from this software |
52 | * without specific prior written permission. | | 52 | * without specific prior written permission. |
53 | * | | 53 | * |
54 | * THE SOFTWARE PROVIDED BY NRL IS PROVIDED BY NRL AND CONTRIBUTORS ``AS | | 54 | * THE SOFTWARE PROVIDED BY NRL IS PROVIDED BY NRL AND CONTRIBUTORS ``AS |
55 | * IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED | | 55 | * IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED |
56 | * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A | | 56 | * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A |
57 | * PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL NRL OR | | 57 | * PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL NRL OR |
58 | * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, | | 58 | * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, |
59 | * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, | | 59 | * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, |
60 | * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR | | 60 | * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR |
61 | * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF | | 61 | * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF |
62 | * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING | | 62 | * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING |
63 | * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS | | 63 | * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS |
64 | * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. | | 64 | * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. |
65 | * | | 65 | * |
66 | * The views and conclusions contained in the software and documentation | | 66 | * The views and conclusions contained in the software and documentation |
67 | * are those of the authors and should not be interpreted as representing | | 67 | * are those of the authors and should not be interpreted as representing |
68 | * official policies, either expressed or implied, of the US Naval | | 68 | * official policies, either expressed or implied, of the US Naval |
69 | * Research Laboratory (NRL). | | 69 | * Research Laboratory (NRL). |
70 | */ | | 70 | */ |
71 | | | 71 | |
72 | /*- | | 72 | /*- |
73 | * Copyright (c) 1997, 1998, 1999, 2001, 2005, 2006 The NetBSD Foundation, Inc. | | 73 | * Copyright (c) 1997, 1998, 1999, 2001, 2005, 2006 The NetBSD Foundation, Inc. |
74 | * All rights reserved. | | 74 | * All rights reserved. |
75 | * | | 75 | * |
76 | * This code is derived from software contributed to The NetBSD Foundation | | 76 | * This code is derived from software contributed to The NetBSD Foundation |
77 | * by Jason R. Thorpe and Kevin M. Lahey of the Numerical Aerospace Simulation | | 77 | * by Jason R. Thorpe and Kevin M. Lahey of the Numerical Aerospace Simulation |
78 | * Facility, NASA Ames Research Center. | | 78 | * Facility, NASA Ames Research Center. |
79 | * This code is derived from software contributed to The NetBSD Foundation | | 79 | * This code is derived from software contributed to The NetBSD Foundation |
80 | * by Charles M. Hannum. | | 80 | * by Charles M. Hannum. |
81 | * This code is derived from software contributed to The NetBSD Foundation | | 81 | * This code is derived from software contributed to The NetBSD Foundation |
82 | * by Rui Paulo. | | 82 | * by Rui Paulo. |
83 | * | | 83 | * |
84 | * Redistribution and use in source and binary forms, with or without | | 84 | * Redistribution and use in source and binary forms, with or without |
85 | * modification, are permitted provided that the following conditions | | 85 | * modification, are permitted provided that the following conditions |
86 | * are met: | | 86 | * are met: |
87 | * 1. Redistributions of source code must retain the above copyright | | 87 | * 1. Redistributions of source code must retain the above copyright |
88 | * notice, this list of conditions and the following disclaimer. | | 88 | * notice, this list of conditions and the following disclaimer. |
89 | * 2. Redistributions in binary form must reproduce the above copyright | | 89 | * 2. Redistributions in binary form must reproduce the above copyright |
90 | * notice, this list of conditions and the following disclaimer in the | | 90 | * notice, this list of conditions and the following disclaimer in the |
91 | * documentation and/or other materials provided with the distribution. | | 91 | * documentation and/or other materials provided with the distribution. |
92 | * | | 92 | * |
93 | * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS | | 93 | * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS |
94 | * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED | | 94 | * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED |
95 | * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR | | 95 | * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR |
96 | * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS | | 96 | * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS |
97 | * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR | | 97 | * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR |
98 | * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF | | 98 | * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF |
99 | * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS | | 99 | * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS |
100 | * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN | | 100 | * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN |
101 | * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) | | 101 | * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) |
102 | * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE | | 102 | * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE |
103 | * POSSIBILITY OF SUCH DAMAGE. | | 103 | * POSSIBILITY OF SUCH DAMAGE. |
104 | */ | | 104 | */ |
105 | | | 105 | |
106 | /* | | 106 | /* |
107 | * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1994, 1995 | | 107 | * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1994, 1995 |
108 | * The Regents of the University of California. All rights reserved. | | 108 | * The Regents of the University of California. All rights reserved. |
109 | * | | 109 | * |
110 | * Redistribution and use in source and binary forms, with or without | | 110 | * Redistribution and use in source and binary forms, with or without |
111 | * modification, are permitted provided that the following conditions | | 111 | * modification, are permitted provided that the following conditions |
112 | * are met: | | 112 | * are met: |
113 | * 1. Redistributions of source code must retain the above copyright | | 113 | * 1. Redistributions of source code must retain the above copyright |
114 | * notice, this list of conditions and the following disclaimer. | | 114 | * notice, this list of conditions and the following disclaimer. |
115 | * 2. Redistributions in binary form must reproduce the above copyright | | 115 | * 2. Redistributions in binary form must reproduce the above copyright |
116 | * notice, this list of conditions and the following disclaimer in the | | 116 | * notice, this list of conditions and the following disclaimer in the |
117 | * documentation and/or other materials provided with the distribution. | | 117 | * documentation and/or other materials provided with the distribution. |
118 | * 3. Neither the name of the University nor the names of its contributors | | 118 | * 3. Neither the name of the University nor the names of its contributors |
119 | * may be used to endorse or promote products derived from this software | | 119 | * may be used to endorse or promote products derived from this software |
120 | * without specific prior written permission. | | 120 | * without specific prior written permission. |
121 | * | | 121 | * |
122 | * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND | | 122 | * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND |
123 | * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE | | 123 | * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE |
124 | * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE | | 124 | * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE |
125 | * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE | | 125 | * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE |
126 | * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL | | 126 | * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL |
127 | * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS | | 127 | * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS |
128 | * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) | | 128 | * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) |
129 | * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT | | 129 | * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT |
130 | * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY | | 130 | * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY |
131 | * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF | | 131 | * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF |
132 | * SUCH DAMAGE. | | 132 | * SUCH DAMAGE. |
133 | * | | 133 | * |
134 | * @(#)tcp_input.c 8.12 (Berkeley) 5/24/95 | | 134 | * @(#)tcp_input.c 8.12 (Berkeley) 5/24/95 |
135 | */ | | 135 | */ |
136 | | | 136 | |
137 | /* | | 137 | /* |
138 | * TODO list for SYN cache stuff: | | 138 | * TODO list for SYN cache stuff: |
139 | * | | 139 | * |
140 | * Find room for a "state" field, which is needed to keep a | | 140 | * Find room for a "state" field, which is needed to keep a |
141 | * compressed state for TIME_WAIT TCBs. It's been noted already | | 141 | * compressed state for TIME_WAIT TCBs. It's been noted already |
142 | * that this is fairly important for very high-volume web and | | 142 | * that this is fairly important for very high-volume web and |
143 | * mail servers, which use a large number of short-lived | | 143 | * mail servers, which use a large number of short-lived |
144 | * connections. | | 144 | * connections. |
145 | */ | | 145 | */ |
146 | | | 146 | |
147 | #include <sys/cdefs.h> | | 147 | #include <sys/cdefs.h> |
148 | __KERNEL_RCSID(0, "$NetBSD: tcp_input.c,v 1.295 2009/03/18 16:00:22 cegger Exp $"); | | 148 | __KERNEL_RCSID(0, "$NetBSD: tcp_input.c,v 1.296 2009/06/20 17:29:31 christos Exp $"); |
149 | | | 149 | |
150 | #include "opt_inet.h" | | 150 | #include "opt_inet.h" |
151 | #include "opt_ipsec.h" | | 151 | #include "opt_ipsec.h" |
152 | #include "opt_inet_csum.h" | | 152 | #include "opt_inet_csum.h" |
153 | #include "opt_tcp_debug.h" | | 153 | #include "opt_tcp_debug.h" |
154 | | | 154 | |
155 | #include <sys/param.h> | | 155 | #include <sys/param.h> |
156 | #include <sys/systm.h> | | 156 | #include <sys/systm.h> |
157 | #include <sys/malloc.h> | | 157 | #include <sys/malloc.h> |
158 | #include <sys/mbuf.h> | | 158 | #include <sys/mbuf.h> |
159 | #include <sys/protosw.h> | | 159 | #include <sys/protosw.h> |
160 | #include <sys/socket.h> | | 160 | #include <sys/socket.h> |
161 | #include <sys/socketvar.h> | | 161 | #include <sys/socketvar.h> |
162 | #include <sys/errno.h> | | 162 | #include <sys/errno.h> |
163 | #include <sys/syslog.h> | | 163 | #include <sys/syslog.h> |
164 | #include <sys/pool.h> | | 164 | #include <sys/pool.h> |
165 | #include <sys/domain.h> | | 165 | #include <sys/domain.h> |
166 | #include <sys/kernel.h> | | 166 | #include <sys/kernel.h> |
167 | #ifdef TCP_SIGNATURE | | 167 | #ifdef TCP_SIGNATURE |
168 | #include <sys/md5.h> | | 168 | #include <sys/md5.h> |
169 | #endif | | 169 | #endif |
170 | #include <sys/lwp.h> /* for lwp0 */ | | 170 | #include <sys/lwp.h> /* for lwp0 */ |
171 | | | 171 | |
172 | #include <net/if.h> | | 172 | #include <net/if.h> |
173 | #include <net/route.h> | | 173 | #include <net/route.h> |
174 | #include <net/if_types.h> | | 174 | #include <net/if_types.h> |
175 | | | 175 | |
176 | #include <netinet/in.h> | | 176 | #include <netinet/in.h> |
177 | #include <netinet/in_systm.h> | | 177 | #include <netinet/in_systm.h> |
178 | #include <netinet/ip.h> | | 178 | #include <netinet/ip.h> |
179 | #include <netinet/in_pcb.h> | | 179 | #include <netinet/in_pcb.h> |
180 | #include <netinet/in_var.h> | | 180 | #include <netinet/in_var.h> |
181 | #include <netinet/ip_var.h> | | 181 | #include <netinet/ip_var.h> |
182 | #include <netinet/in_offload.h> | | 182 | #include <netinet/in_offload.h> |
183 | | | 183 | |
184 | #ifdef INET6 | | 184 | #ifdef INET6 |
185 | #ifndef INET | | 185 | #ifndef INET |
186 | #include <netinet/in.h> | | 186 | #include <netinet/in.h> |
187 | #endif | | 187 | #endif |
188 | #include <netinet/ip6.h> | | 188 | #include <netinet/ip6.h> |
189 | #include <netinet6/ip6_var.h> | | 189 | #include <netinet6/ip6_var.h> |
190 | #include <netinet6/in6_pcb.h> | | 190 | #include <netinet6/in6_pcb.h> |
191 | #include <netinet6/ip6_var.h> | | 191 | #include <netinet6/ip6_var.h> |
192 | #include <netinet6/in6_var.h> | | 192 | #include <netinet6/in6_var.h> |
193 | #include <netinet/icmp6.h> | | 193 | #include <netinet/icmp6.h> |
194 | #include <netinet6/nd6.h> | | 194 | #include <netinet6/nd6.h> |
195 | #ifdef TCP_SIGNATURE | | 195 | #ifdef TCP_SIGNATURE |
196 | #include <netinet6/scope6_var.h> | | 196 | #include <netinet6/scope6_var.h> |
197 | #endif | | 197 | #endif |
198 | #endif | | 198 | #endif |
199 | | | 199 | |
200 | #ifndef INET6 | | 200 | #ifndef INET6 |
201 | /* always need ip6.h for IP6_EXTHDR_GET */ | | 201 | /* always need ip6.h for IP6_EXTHDR_GET */ |
202 | #include <netinet/ip6.h> | | 202 | #include <netinet/ip6.h> |
203 | #endif | | 203 | #endif |
204 | | | 204 | |
205 | #include <netinet/tcp.h> | | 205 | #include <netinet/tcp.h> |
206 | #include <netinet/tcp_fsm.h> | | 206 | #include <netinet/tcp_fsm.h> |
207 | #include <netinet/tcp_seq.h> | | 207 | #include <netinet/tcp_seq.h> |
208 | #include <netinet/tcp_timer.h> | | 208 | #include <netinet/tcp_timer.h> |
209 | #include <netinet/tcp_var.h> | | 209 | #include <netinet/tcp_var.h> |
210 | #include <netinet/tcp_private.h> | | 210 | #include <netinet/tcp_private.h> |
211 | #include <netinet/tcpip.h> | | 211 | #include <netinet/tcpip.h> |
212 | #include <netinet/tcp_congctl.h> | | 212 | #include <netinet/tcp_congctl.h> |
213 | #include <netinet/tcp_debug.h> | | 213 | #include <netinet/tcp_debug.h> |
214 | | | 214 | |
215 | #include <machine/stdarg.h> | | 215 | #include <machine/stdarg.h> |
216 | | | 216 | |
217 | #ifdef IPSEC | | 217 | #ifdef IPSEC |
218 | #include <netinet6/ipsec.h> | | 218 | #include <netinet6/ipsec.h> |
219 | #include <netinet6/ipsec_private.h> | | 219 | #include <netinet6/ipsec_private.h> |
220 | #include <netkey/key.h> | | 220 | #include <netkey/key.h> |
221 | #endif /*IPSEC*/ | | 221 | #endif /*IPSEC*/ |
222 | #ifdef INET6 | | 222 | #ifdef INET6 |
223 | #include "faith.h" | | 223 | #include "faith.h" |
224 | #if defined(NFAITH) && NFAITH > 0 | | 224 | #if defined(NFAITH) && NFAITH > 0 |
225 | #include <net/if_faith.h> | | 225 | #include <net/if_faith.h> |
226 | #endif | | 226 | #endif |
227 | #endif /* IPSEC */ | | 227 | #endif /* IPSEC */ |
228 | | | 228 | |
229 | #ifdef FAST_IPSEC | | 229 | #ifdef FAST_IPSEC |
230 | #include <netipsec/ipsec.h> | | 230 | #include <netipsec/ipsec.h> |
231 | #include <netipsec/ipsec_var.h> | | 231 | #include <netipsec/ipsec_var.h> |
232 | #include <netipsec/ipsec_private.h> | | 232 | #include <netipsec/ipsec_private.h> |
233 | #include <netipsec/key.h> | | 233 | #include <netipsec/key.h> |
234 | #ifdef INET6 | | 234 | #ifdef INET6 |
235 | #include <netipsec/ipsec6.h> | | 235 | #include <netipsec/ipsec6.h> |
236 | #endif | | 236 | #endif |
237 | #endif /* FAST_IPSEC*/ | | 237 | #endif /* FAST_IPSEC*/ |
238 | | | 238 | |
239 | int tcprexmtthresh = 3; | | 239 | int tcprexmtthresh = 3; |
240 | int tcp_log_refused; | | 240 | int tcp_log_refused; |
241 | | | 241 | |
242 | int tcp_do_autorcvbuf = 0; | | 242 | int tcp_do_autorcvbuf = 0; |
243 | int tcp_autorcvbuf_inc = 16 * 1024; | | 243 | int tcp_autorcvbuf_inc = 16 * 1024; |
244 | int tcp_autorcvbuf_max = 256 * 1024; | | 244 | int tcp_autorcvbuf_max = 256 * 1024; |
245 | | | 245 | |
246 | static int tcp_rst_ppslim_count = 0; | | 246 | static int tcp_rst_ppslim_count = 0; |
247 | static struct timeval tcp_rst_ppslim_last; | | 247 | static struct timeval tcp_rst_ppslim_last; |
248 | static int tcp_ackdrop_ppslim_count = 0; | | 248 | static int tcp_ackdrop_ppslim_count = 0; |
249 | static struct timeval tcp_ackdrop_ppslim_last; | | 249 | static struct timeval tcp_ackdrop_ppslim_last; |
250 | | | 250 | |
251 | #define TCP_PAWS_IDLE (24U * 24 * 60 * 60 * PR_SLOWHZ) | | 251 | #define TCP_PAWS_IDLE (24U * 24 * 60 * 60 * PR_SLOWHZ) |
252 | | | 252 | |
253 | /* for modulo comparisons of timestamps */ | | 253 | /* for modulo comparisons of timestamps */ |
254 | #define TSTMP_LT(a,b) ((int)((a)-(b)) < 0) | | 254 | #define TSTMP_LT(a,b) ((int)((a)-(b)) < 0) |
255 | #define TSTMP_GEQ(a,b) ((int)((a)-(b)) >= 0) | | 255 | #define TSTMP_GEQ(a,b) ((int)((a)-(b)) >= 0) |
256 | | | 256 | |
257 | /* | | 257 | /* |
258 | * Neighbor Discovery, Neighbor Unreachability Detection Upper layer hint. | | 258 | * Neighbor Discovery, Neighbor Unreachability Detection Upper layer hint. |
259 | */ | | 259 | */ |
260 | #ifdef INET6 | | 260 | #ifdef INET6 |
261 | static inline void | | 261 | static inline void |
262 | nd6_hint(struct tcpcb *tp) | | 262 | nd6_hint(struct tcpcb *tp) |
263 | { | | 263 | { |
264 | struct rtentry *rt; | | 264 | struct rtentry *rt; |
265 | | | 265 | |
266 | if (tp != NULL && tp->t_in6pcb != NULL && tp->t_family == AF_INET6 && | | 266 | if (tp != NULL && tp->t_in6pcb != NULL && tp->t_family == AF_INET6 && |
267 | (rt = rtcache_validate(&tp->t_in6pcb->in6p_route)) != NULL) | | 267 | (rt = rtcache_validate(&tp->t_in6pcb->in6p_route)) != NULL) |
268 | nd6_nud_hint(rt, NULL, 0); | | 268 | nd6_nud_hint(rt, NULL, 0); |
269 | } | | 269 | } |
270 | #else | | 270 | #else |
271 | static inline void | | 271 | static inline void |
272 | nd6_hint(struct tcpcb *tp) | | 272 | nd6_hint(struct tcpcb *tp) |
273 | { | | 273 | { |
274 | } | | 274 | } |
275 | #endif | | 275 | #endif |
276 | | | 276 | |
277 | /* | | 277 | /* |
278 | * Compute ACK transmission behavior. Delay the ACK unless | | 278 | * Compute ACK transmission behavior. Delay the ACK unless |
279 | * we have already delayed an ACK (must send an ACK every two segments). | | 279 | * we have already delayed an ACK (must send an ACK every two segments). |
280 | * We also ACK immediately if we received a PUSH and the ACK-on-PUSH | | 280 | * We also ACK immediately if we received a PUSH and the ACK-on-PUSH |
281 | * option is enabled. | | 281 | * option is enabled. |
282 | */ | | 282 | */ |
283 | static void | | 283 | static void |
284 | tcp_setup_ack(struct tcpcb *tp, const struct tcphdr *th) | | 284 | tcp_setup_ack(struct tcpcb *tp, const struct tcphdr *th) |
285 | { | | 285 | { |
286 | | | 286 | |
287 | if (tp->t_flags & TF_DELACK || | | 287 | if (tp->t_flags & TF_DELACK || |
288 | (tcp_ack_on_push && th->th_flags & TH_PUSH)) | | 288 | (tcp_ack_on_push && th->th_flags & TH_PUSH)) |
289 | tp->t_flags |= TF_ACKNOW; | | 289 | tp->t_flags |= TF_ACKNOW; |
290 | else | | 290 | else |
291 | TCP_SET_DELACK(tp); | | 291 | TCP_SET_DELACK(tp); |
292 | } | | 292 | } |
293 | | | 293 | |
294 | static void | | 294 | static void |
295 | icmp_check(struct tcpcb *tp, const struct tcphdr *th, int acked) | | 295 | icmp_check(struct tcpcb *tp, const struct tcphdr *th, int acked) |
296 | { | | 296 | { |
297 | | | 297 | |
298 | /* | | 298 | /* |
299 | * If we had a pending ICMP message that refers to data that have | | 299 | * If we had a pending ICMP message that refers to data that have |
300 | * just been acknowledged, disregard the recorded ICMP message. | | 300 | * just been acknowledged, disregard the recorded ICMP message. |
301 | */ | | 301 | */ |
302 | if ((tp->t_flags & TF_PMTUD_PEND) && | | 302 | if ((tp->t_flags & TF_PMTUD_PEND) && |
303 | SEQ_GT(th->th_ack, tp->t_pmtud_th_seq)) | | 303 | SEQ_GT(th->th_ack, tp->t_pmtud_th_seq)) |
304 | tp->t_flags &= ~TF_PMTUD_PEND; | | 304 | tp->t_flags &= ~TF_PMTUD_PEND; |
305 | | | 305 | |
306 | /* | | 306 | /* |
307 | * Keep track of the largest chunk of data | | 307 | * Keep track of the largest chunk of data |
308 | * acknowledged since last PMTU update | | 308 | * acknowledged since last PMTU update |
309 | */ | | 309 | */ |
310 | if (tp->t_pmtud_mss_acked < acked) | | 310 | if (tp->t_pmtud_mss_acked < acked) |
311 | tp->t_pmtud_mss_acked = acked; | | 311 | tp->t_pmtud_mss_acked = acked; |
312 | } | | 312 | } |
313 | | | 313 | |
314 | /* | | 314 | /* |
315 | * Convert TCP protocol fields to host order for easier processing. | | 315 | * Convert TCP protocol fields to host order for easier processing. |
316 | */ | | 316 | */ |
317 | static void | | 317 | static void |
318 | tcp_fields_to_host(struct tcphdr *th) | | 318 | tcp_fields_to_host(struct tcphdr *th) |
319 | { | | 319 | { |
320 | | | 320 | |
321 | NTOHL(th->th_seq); | | 321 | NTOHL(th->th_seq); |
322 | NTOHL(th->th_ack); | | 322 | NTOHL(th->th_ack); |
323 | NTOHS(th->th_win); | | 323 | NTOHS(th->th_win); |
324 | NTOHS(th->th_urp); | | 324 | NTOHS(th->th_urp); |
325 | } | | 325 | } |
326 | | | 326 | |
327 | /* | | 327 | /* |
328 | * ... and reverse the above. | | 328 | * ... and reverse the above. |
329 | */ | | 329 | */ |
330 | static void | | 330 | static void |
331 | tcp_fields_to_net(struct tcphdr *th) | | 331 | tcp_fields_to_net(struct tcphdr *th) |
332 | { | | 332 | { |
333 | | | 333 | |
334 | HTONL(th->th_seq); | | 334 | HTONL(th->th_seq); |
335 | HTONL(th->th_ack); | | 335 | HTONL(th->th_ack); |
336 | HTONS(th->th_win); | | 336 | HTONS(th->th_win); |
337 | HTONS(th->th_urp); | | 337 | HTONS(th->th_urp); |
338 | } | | 338 | } |
339 | | | 339 | |
340 | #ifdef TCP_CSUM_COUNTERS | | 340 | #ifdef TCP_CSUM_COUNTERS |
341 | #include <sys/device.h> | | 341 | #include <sys/device.h> |
342 | | | 342 | |
343 | #if defined(INET) | | 343 | #if defined(INET) |
344 | extern struct evcnt tcp_hwcsum_ok; | | 344 | extern struct evcnt tcp_hwcsum_ok; |
345 | extern struct evcnt tcp_hwcsum_bad; | | 345 | extern struct evcnt tcp_hwcsum_bad; |
346 | extern struct evcnt tcp_hwcsum_data; | | 346 | extern struct evcnt tcp_hwcsum_data; |
347 | extern struct evcnt tcp_swcsum; | | 347 | extern struct evcnt tcp_swcsum; |
348 | #endif /* defined(INET) */ | | 348 | #endif /* defined(INET) */ |
349 | #if defined(INET6) | | 349 | #if defined(INET6) |
350 | extern struct evcnt tcp6_hwcsum_ok; | | 350 | extern struct evcnt tcp6_hwcsum_ok; |
351 | extern struct evcnt tcp6_hwcsum_bad; | | 351 | extern struct evcnt tcp6_hwcsum_bad; |
352 | extern struct evcnt tcp6_hwcsum_data; | | 352 | extern struct evcnt tcp6_hwcsum_data; |
353 | extern struct evcnt tcp6_swcsum; | | 353 | extern struct evcnt tcp6_swcsum; |
354 | #endif /* defined(INET6) */ | | 354 | #endif /* defined(INET6) */ |
355 | | | 355 | |
356 | #define TCP_CSUM_COUNTER_INCR(ev) (ev)->ev_count++ | | 356 | #define TCP_CSUM_COUNTER_INCR(ev) (ev)->ev_count++ |
357 | | | 357 | |
358 | #else | | 358 | #else |
359 | | | 359 | |
360 | #define TCP_CSUM_COUNTER_INCR(ev) /* nothing */ | | 360 | #define TCP_CSUM_COUNTER_INCR(ev) /* nothing */ |
361 | | | 361 | |
362 | #endif /* TCP_CSUM_COUNTERS */ | | 362 | #endif /* TCP_CSUM_COUNTERS */ |
363 | | | 363 | |
364 | #ifdef TCP_REASS_COUNTERS | | 364 | #ifdef TCP_REASS_COUNTERS |
365 | #include <sys/device.h> | | 365 | #include <sys/device.h> |
366 | | | 366 | |
367 | extern struct evcnt tcp_reass_; | | 367 | extern struct evcnt tcp_reass_; |
368 | extern struct evcnt tcp_reass_empty; | | 368 | extern struct evcnt tcp_reass_empty; |
369 | extern struct evcnt tcp_reass_iteration[8]; | | 369 | extern struct evcnt tcp_reass_iteration[8]; |
370 | extern struct evcnt tcp_reass_prependfirst; | | 370 | extern struct evcnt tcp_reass_prependfirst; |
371 | extern struct evcnt tcp_reass_prepend; | | 371 | extern struct evcnt tcp_reass_prepend; |
372 | extern struct evcnt tcp_reass_insert; | | 372 | extern struct evcnt tcp_reass_insert; |
373 | extern struct evcnt tcp_reass_inserttail; | | 373 | extern struct evcnt tcp_reass_inserttail; |
374 | extern struct evcnt tcp_reass_append; | | 374 | extern struct evcnt tcp_reass_append; |
375 | extern struct evcnt tcp_reass_appendtail; | | 375 | extern struct evcnt tcp_reass_appendtail; |
376 | extern struct evcnt tcp_reass_overlaptail; | | 376 | extern struct evcnt tcp_reass_overlaptail; |
377 | extern struct evcnt tcp_reass_overlapfront; | | 377 | extern struct evcnt tcp_reass_overlapfront; |
378 | extern struct evcnt tcp_reass_segdup; | | 378 | extern struct evcnt tcp_reass_segdup; |
379 | extern struct evcnt tcp_reass_fragdup; | | 379 | extern struct evcnt tcp_reass_fragdup; |
380 | | | 380 | |
381 | #define TCP_REASS_COUNTER_INCR(ev) (ev)->ev_count++ | | 381 | #define TCP_REASS_COUNTER_INCR(ev) (ev)->ev_count++ |
382 | | | 382 | |
383 | #else | | 383 | #else |
384 | | | 384 | |
385 | #define TCP_REASS_COUNTER_INCR(ev) /* nothing */ | | 385 | #define TCP_REASS_COUNTER_INCR(ev) /* nothing */ |
386 | | | 386 | |
387 | #endif /* TCP_REASS_COUNTERS */ | | 387 | #endif /* TCP_REASS_COUNTERS */ |
388 | | | 388 | |
389 | static int tcp_reass(struct tcpcb *, const struct tcphdr *, struct mbuf *, | | 389 | static int tcp_reass(struct tcpcb *, const struct tcphdr *, struct mbuf *, |
390 | int *); | | 390 | int *); |
391 | static int tcp_dooptions(struct tcpcb *, const u_char *, int, | | 391 | static int tcp_dooptions(struct tcpcb *, const u_char *, int, |
392 | struct tcphdr *, struct mbuf *, int, struct tcp_opt_info *); | | 392 | struct tcphdr *, struct mbuf *, int, struct tcp_opt_info *); |
393 | | | 393 | |
394 | #ifdef INET | | 394 | #ifdef INET |
395 | static void tcp4_log_refused(const struct ip *, const struct tcphdr *); | | 395 | static void tcp4_log_refused(const struct ip *, const struct tcphdr *); |
396 | #endif | | 396 | #endif |
397 | #ifdef INET6 | | 397 | #ifdef INET6 |
398 | static void tcp6_log_refused(const struct ip6_hdr *, const struct tcphdr *); | | 398 | static void tcp6_log_refused(const struct ip6_hdr *, const struct tcphdr *); |
399 | #endif | | 399 | #endif |
400 | | | 400 | |
401 | #define TRAVERSE(x) while ((x)->m_next) (x) = (x)->m_next | | 401 | #define TRAVERSE(x) while ((x)->m_next) (x) = (x)->m_next |
402 | | | 402 | |
403 | #if defined(MBUFTRACE) | | 403 | #if defined(MBUFTRACE) |
404 | struct mowner tcp_reass_mowner = MOWNER_INIT("tcp", "reass"); | | 404 | struct mowner tcp_reass_mowner = MOWNER_INIT("tcp", "reass"); |
405 | #endif /* defined(MBUFTRACE) */ | | 405 | #endif /* defined(MBUFTRACE) */ |
406 | | | 406 | |
407 | static struct pool tcpipqent_pool; | | 407 | static struct pool tcpipqent_pool; |
408 | | | 408 | |
409 | void | | 409 | void |
410 | tcpipqent_init(void) | | 410 | tcpipqent_init(void) |
411 | { | | 411 | { |
412 | | | 412 | |
413 | pool_init(&tcpipqent_pool, sizeof(struct ipqent), 0, 0, 0, "tcpipqepl", | | 413 | pool_init(&tcpipqent_pool, sizeof(struct ipqent), 0, 0, 0, "tcpipqepl", |
414 | NULL, IPL_VM); | | 414 | NULL, IPL_VM); |
415 | } | | 415 | } |
416 | | | 416 | |
417 | struct ipqent * | | 417 | struct ipqent * |
418 | tcpipqent_alloc(void) | | 418 | tcpipqent_alloc(void) |
419 | { | | 419 | { |
420 | struct ipqent *ipqe; | | 420 | struct ipqent *ipqe; |
421 | int s; | | 421 | int s; |
422 | | | 422 | |
423 | s = splvm(); | | 423 | s = splvm(); |
424 | ipqe = pool_get(&tcpipqent_pool, PR_NOWAIT); | | 424 | ipqe = pool_get(&tcpipqent_pool, PR_NOWAIT); |
425 | splx(s); | | 425 | splx(s); |
426 | | | 426 | |
427 | return ipqe; | | 427 | return ipqe; |
428 | } | | 428 | } |
429 | | | 429 | |
430 | void | | 430 | void |
431 | tcpipqent_free(struct ipqent *ipqe) | | 431 | tcpipqent_free(struct ipqent *ipqe) |
432 | { | | 432 | { |
433 | int s; | | 433 | int s; |
434 | | | 434 | |
435 | s = splvm(); | | 435 | s = splvm(); |
436 | pool_put(&tcpipqent_pool, ipqe); | | 436 | pool_put(&tcpipqent_pool, ipqe); |
437 | splx(s); | | 437 | splx(s); |
438 | } | | 438 | } |
439 | | | 439 | |
440 | static int | | 440 | static int |
441 | tcp_reass(struct tcpcb *tp, const struct tcphdr *th, struct mbuf *m, int *tlen) | | 441 | tcp_reass(struct tcpcb *tp, const struct tcphdr *th, struct mbuf *m, int *tlen) |
442 | { | | 442 | { |
443 | struct ipqent *p, *q, *nq, *tiqe = NULL; | | 443 | struct ipqent *p, *q, *nq, *tiqe = NULL; |
444 | struct socket *so = NULL; | | 444 | struct socket *so = NULL; |
445 | int pkt_flags; | | 445 | int pkt_flags; |
446 | tcp_seq pkt_seq; | | 446 | tcp_seq pkt_seq; |
447 | unsigned pkt_len; | | 447 | unsigned pkt_len; |
448 | u_long rcvpartdupbyte = 0; | | 448 | u_long rcvpartdupbyte = 0; |
449 | u_long rcvoobyte; | | 449 | u_long rcvoobyte; |
450 | #ifdef TCP_REASS_COUNTERS | | 450 | #ifdef TCP_REASS_COUNTERS |
451 | u_int count = 0; | | 451 | u_int count = 0; |
452 | #endif | | 452 | #endif |
453 | uint64_t *tcps; | | 453 | uint64_t *tcps; |
454 | | | 454 | |
455 | if (tp->t_inpcb) | | 455 | if (tp->t_inpcb) |
456 | so = tp->t_inpcb->inp_socket; | | 456 | so = tp->t_inpcb->inp_socket; |
457 | #ifdef INET6 | | 457 | #ifdef INET6 |
458 | else if (tp->t_in6pcb) | | 458 | else if (tp->t_in6pcb) |
459 | so = tp->t_in6pcb->in6p_socket; | | 459 | so = tp->t_in6pcb->in6p_socket; |
460 | #endif | | 460 | #endif |
461 | | | 461 | |
462 | TCP_REASS_LOCK_CHECK(tp); | | 462 | TCP_REASS_LOCK_CHECK(tp); |
463 | | | 463 | |
464 | /* | | 464 | /* |
465 | * Call with th==0 after become established to | | 465 | * Call with th==0 after become established to |
466 | * force pre-ESTABLISHED data up to user socket. | | 466 | * force pre-ESTABLISHED data up to user socket. |
467 | */ | | 467 | */ |
468 | if (th == 0) | | 468 | if (th == 0) |
469 | goto present; | | 469 | goto present; |
470 | | | 470 | |
471 | m_claimm(m, &tcp_reass_mowner); | | 471 | m_claimm(m, &tcp_reass_mowner); |
472 | | | 472 | |
473 | rcvoobyte = *tlen; | | 473 | rcvoobyte = *tlen; |
474 | /* | | 474 | /* |
475 | * Copy these to local variables because the tcpiphdr | | 475 | * Copy these to local variables because the tcpiphdr |
476 | * gets munged while we are collapsing mbufs. | | 476 | * gets munged while we are collapsing mbufs. |
477 | */ | | 477 | */ |
478 | pkt_seq = th->th_seq; | | 478 | pkt_seq = th->th_seq; |
479 | pkt_len = *tlen; | | 479 | pkt_len = *tlen; |
480 | pkt_flags = th->th_flags; | | 480 | pkt_flags = th->th_flags; |
481 | | | 481 | |
482 | TCP_REASS_COUNTER_INCR(&tcp_reass_); | | 482 | TCP_REASS_COUNTER_INCR(&tcp_reass_); |
483 | | | 483 | |
484 | if ((p = TAILQ_LAST(&tp->segq, ipqehead)) != NULL) { | | 484 | if ((p = TAILQ_LAST(&tp->segq, ipqehead)) != NULL) { |
485 | /* | | 485 | /* |
486 | * When we miss a packet, the vast majority of time we get | | 486 | * When we miss a packet, the vast majority of time we get |
487 | * packets that follow it in order. So optimize for that. | | 487 | * packets that follow it in order. So optimize for that. |
488 | */ | | 488 | */ |
489 | if (pkt_seq == p->ipqe_seq + p->ipqe_len) { | | 489 | if (pkt_seq == p->ipqe_seq + p->ipqe_len) { |
490 | p->ipqe_len += pkt_len; | | 490 | p->ipqe_len += pkt_len; |
491 | p->ipqe_flags |= pkt_flags; | | 491 | p->ipqe_flags |= pkt_flags; |
492 | m_cat(p->ipre_mlast, m); | | 492 | m_cat(p->ipre_mlast, m); |
493 | TRAVERSE(p->ipre_mlast); | | 493 | TRAVERSE(p->ipre_mlast); |
494 | m = NULL; | | 494 | m = NULL; |
495 | tiqe = p; | | 495 | tiqe = p; |
496 | TAILQ_REMOVE(&tp->timeq, p, ipqe_timeq); | | 496 | TAILQ_REMOVE(&tp->timeq, p, ipqe_timeq); |
497 | TCP_REASS_COUNTER_INCR(&tcp_reass_appendtail); | | 497 | TCP_REASS_COUNTER_INCR(&tcp_reass_appendtail); |
498 | goto skip_replacement; | | 498 | goto skip_replacement; |
499 | } | | 499 | } |
500 | /* | | 500 | /* |
501 | * While we're here, if the pkt is completely beyond | | 501 | * While we're here, if the pkt is completely beyond |
502 | * anything we have, just insert it at the tail. | | 502 | * anything we have, just insert it at the tail. |
503 | */ | | 503 | */ |
504 | if (SEQ_GT(pkt_seq, p->ipqe_seq + p->ipqe_len)) { | | 504 | if (SEQ_GT(pkt_seq, p->ipqe_seq + p->ipqe_len)) { |
505 | TCP_REASS_COUNTER_INCR(&tcp_reass_inserttail); | | 505 | TCP_REASS_COUNTER_INCR(&tcp_reass_inserttail); |
506 | goto insert_it; | | 506 | goto insert_it; |
507 | } | | 507 | } |
508 | } | | 508 | } |
509 | | | 509 | |
510 | q = TAILQ_FIRST(&tp->segq); | | 510 | q = TAILQ_FIRST(&tp->segq); |
511 | | | 511 | |
512 | if (q != NULL) { | | 512 | if (q != NULL) { |
513 | /* | | 513 | /* |
514 | * If this segment immediately precedes the first out-of-order | | 514 | * If this segment immediately precedes the first out-of-order |
515 | * block, simply slap the segment in front of it and (mostly) | | 515 | * block, simply slap the segment in front of it and (mostly) |
516 | * skip the complicated logic. | | 516 | * skip the complicated logic. |
517 | */ | | 517 | */ |
518 | if (pkt_seq + pkt_len == q->ipqe_seq) { | | 518 | if (pkt_seq + pkt_len == q->ipqe_seq) { |
519 | q->ipqe_seq = pkt_seq; | | 519 | q->ipqe_seq = pkt_seq; |
520 | q->ipqe_len += pkt_len; | | 520 | q->ipqe_len += pkt_len; |
521 | q->ipqe_flags |= pkt_flags; | | 521 | q->ipqe_flags |= pkt_flags; |
522 | m_cat(m, q->ipqe_m); | | 522 | m_cat(m, q->ipqe_m); |
523 | q->ipqe_m = m; | | 523 | q->ipqe_m = m; |
524 | q->ipre_mlast = m; /* last mbuf may have changed */ | | 524 | q->ipre_mlast = m; /* last mbuf may have changed */ |
525 | TRAVERSE(q->ipre_mlast); | | 525 | TRAVERSE(q->ipre_mlast); |
526 | tiqe = q; | | 526 | tiqe = q; |
527 | TAILQ_REMOVE(&tp->timeq, q, ipqe_timeq); | | 527 | TAILQ_REMOVE(&tp->timeq, q, ipqe_timeq); |
528 | TCP_REASS_COUNTER_INCR(&tcp_reass_prependfirst); | | 528 | TCP_REASS_COUNTER_INCR(&tcp_reass_prependfirst); |
529 | goto skip_replacement; | | 529 | goto skip_replacement; |
530 | } | | 530 | } |
531 | } else { | | 531 | } else { |
532 | TCP_REASS_COUNTER_INCR(&tcp_reass_empty); | | 532 | TCP_REASS_COUNTER_INCR(&tcp_reass_empty); |
533 | } | | 533 | } |
534 | | | 534 | |
535 | /* | | 535 | /* |
536 | * Find a segment which begins after this one does. | | 536 | * Find a segment which begins after this one does. |
537 | */ | | 537 | */ |
538 | for (p = NULL; q != NULL; q = nq) { | | 538 | for (p = NULL; q != NULL; q = nq) { |
539 | nq = TAILQ_NEXT(q, ipqe_q); | | 539 | nq = TAILQ_NEXT(q, ipqe_q); |
540 | #ifdef TCP_REASS_COUNTERS | | 540 | #ifdef TCP_REASS_COUNTERS |
541 | count++; | | 541 | count++; |
542 | #endif | | 542 | #endif |
543 | /* | | 543 | /* |
544 | * If the received segment is just right after this | | 544 | * If the received segment is just right after this |
545 | * fragment, merge the two together and then check | | 545 | * fragment, merge the two together and then check |
546 | * for further overlaps. | | 546 | * for further overlaps. |
547 | */ | | 547 | */ |
548 | if (q->ipqe_seq + q->ipqe_len == pkt_seq) { | | 548 | if (q->ipqe_seq + q->ipqe_len == pkt_seq) { |
549 | #ifdef TCPREASS_DEBUG | | 549 | #ifdef TCPREASS_DEBUG |
550 | printf("tcp_reass[%p]: concat %u:%u(%u) to %u:%u(%u)\n", | | 550 | printf("tcp_reass[%p]: concat %u:%u(%u) to %u:%u(%u)\n", |
551 | tp, pkt_seq, pkt_seq + pkt_len, pkt_len, | | 551 | tp, pkt_seq, pkt_seq + pkt_len, pkt_len, |
552 | q->ipqe_seq, q->ipqe_seq + q->ipqe_len, q->ipqe_len); | | 552 | q->ipqe_seq, q->ipqe_seq + q->ipqe_len, q->ipqe_len); |
553 | #endif | | 553 | #endif |
554 | pkt_len += q->ipqe_len; | | 554 | pkt_len += q->ipqe_len; |
555 | pkt_flags |= q->ipqe_flags; | | 555 | pkt_flags |= q->ipqe_flags; |
556 | pkt_seq = q->ipqe_seq; | | 556 | pkt_seq = q->ipqe_seq; |
557 | m_cat(q->ipre_mlast, m); | | 557 | m_cat(q->ipre_mlast, m); |
558 | TRAVERSE(q->ipre_mlast); | | 558 | TRAVERSE(q->ipre_mlast); |
559 | m = q->ipqe_m; | | 559 | m = q->ipqe_m; |
560 | TCP_REASS_COUNTER_INCR(&tcp_reass_append); | | 560 | TCP_REASS_COUNTER_INCR(&tcp_reass_append); |
561 | goto free_ipqe; | | 561 | goto free_ipqe; |
562 | } | | 562 | } |
563 | /* | | 563 | /* |
564 | * If the received segment is completely past this | | 564 | * If the received segment is completely past this |
565 | * fragment, we need to go the next fragment. | | 565 | * fragment, we need to go the next fragment. |
566 | */ | | 566 | */ |
567 | if (SEQ_LT(q->ipqe_seq + q->ipqe_len, pkt_seq)) { | | 567 | if (SEQ_LT(q->ipqe_seq + q->ipqe_len, pkt_seq)) { |
568 | p = q; | | 568 | p = q; |
569 | continue; | | 569 | continue; |
570 | } | | 570 | } |
571 | /* | | 571 | /* |
572 | * If the fragment is past the received segment, | | 572 | * If the fragment is past the received segment, |
573 | * it (or any following) can't be concatenated. | | 573 | * it (or any following) can't be concatenated. |
574 | */ | | 574 | */ |
575 | if (SEQ_GT(q->ipqe_seq, pkt_seq + pkt_len)) { | | 575 | if (SEQ_GT(q->ipqe_seq, pkt_seq + pkt_len)) { |
576 | TCP_REASS_COUNTER_INCR(&tcp_reass_insert); | | 576 | TCP_REASS_COUNTER_INCR(&tcp_reass_insert); |
577 | break; | | 577 | break; |
578 | } | | 578 | } |
579 | | | 579 | |
580 | /* | | 580 | /* |
581 | * We've received all the data in this segment before. | | 581 | * We've received all the data in this segment before. |
582 | * mark it as a duplicate and return. | | 582 | * mark it as a duplicate and return. |
583 | */ | | 583 | */ |
584 | if (SEQ_LEQ(q->ipqe_seq, pkt_seq) && | | 584 | if (SEQ_LEQ(q->ipqe_seq, pkt_seq) && |
585 | SEQ_GEQ(q->ipqe_seq + q->ipqe_len, pkt_seq + pkt_len)) { | | 585 | SEQ_GEQ(q->ipqe_seq + q->ipqe_len, pkt_seq + pkt_len)) { |
586 | tcps = TCP_STAT_GETREF(); | | 586 | tcps = TCP_STAT_GETREF(); |
587 | tcps[TCP_STAT_RCVDUPPACK]++; | | 587 | tcps[TCP_STAT_RCVDUPPACK]++; |
588 | tcps[TCP_STAT_RCVDUPBYTE] += pkt_len; | | 588 | tcps[TCP_STAT_RCVDUPBYTE] += pkt_len; |
589 | TCP_STAT_PUTREF(); | | 589 | TCP_STAT_PUTREF(); |
590 | tcp_new_dsack(tp, pkt_seq, pkt_len); | | 590 | tcp_new_dsack(tp, pkt_seq, pkt_len); |
591 | m_freem(m); | | 591 | m_freem(m); |
592 | if (tiqe != NULL) { | | 592 | if (tiqe != NULL) { |
593 | tcpipqent_free(tiqe); | | 593 | tcpipqent_free(tiqe); |
594 | } | | 594 | } |
595 | TCP_REASS_COUNTER_INCR(&tcp_reass_segdup); | | 595 | TCP_REASS_COUNTER_INCR(&tcp_reass_segdup); |
596 | return (0); | | 596 | return (0); |
597 | } | | 597 | } |
598 | /* | | 598 | /* |
599 | * Received segment completely overlaps this fragment | | 599 | * Received segment completely overlaps this fragment |
600 | * so we drop the fragment (this keeps the temporal | | 600 | * so we drop the fragment (this keeps the temporal |
601 | * ordering of segments correct). | | 601 | * ordering of segments correct). |
602 | */ | | 602 | */ |
603 | if (SEQ_GEQ(q->ipqe_seq, pkt_seq) && | | 603 | if (SEQ_GEQ(q->ipqe_seq, pkt_seq) && |
604 | SEQ_LEQ(q->ipqe_seq + q->ipqe_len, pkt_seq + pkt_len)) { | | 604 | SEQ_LEQ(q->ipqe_seq + q->ipqe_len, pkt_seq + pkt_len)) { |
605 | rcvpartdupbyte += q->ipqe_len; | | 605 | rcvpartdupbyte += q->ipqe_len; |
606 | m_freem(q->ipqe_m); | | 606 | m_freem(q->ipqe_m); |
607 | TCP_REASS_COUNTER_INCR(&tcp_reass_fragdup); | | 607 | TCP_REASS_COUNTER_INCR(&tcp_reass_fragdup); |
608 | goto free_ipqe; | | 608 | goto free_ipqe; |
609 | } | | 609 | } |
610 | /* | | 610 | /* |
611 | * RX'ed segment extends past the end of the | | 611 | * RX'ed segment extends past the end of the |
612 | * fragment. Drop the overlapping bytes. Then | | 612 | * fragment. Drop the overlapping bytes. Then |
613 | * merge the fragment and segment then treat as | | 613 | * merge the fragment and segment then treat as |
614 | * a longer received packet. | | 614 | * a longer received packet. |
615 | */ | | 615 | */ |
616 | if (SEQ_LT(q->ipqe_seq, pkt_seq) && | | 616 | if (SEQ_LT(q->ipqe_seq, pkt_seq) && |
617 | SEQ_GT(q->ipqe_seq + q->ipqe_len, pkt_seq)) { | | 617 | SEQ_GT(q->ipqe_seq + q->ipqe_len, pkt_seq)) { |
618 | int overlap = q->ipqe_seq + q->ipqe_len - pkt_seq; | | 618 | int overlap = q->ipqe_seq + q->ipqe_len - pkt_seq; |
619 | #ifdef TCPREASS_DEBUG | | 619 | #ifdef TCPREASS_DEBUG |
620 | printf("tcp_reass[%p]: trim starting %d bytes of %u:%u(%u)\n", | | 620 | printf("tcp_reass[%p]: trim starting %d bytes of %u:%u(%u)\n", |
621 | tp, overlap, | | 621 | tp, overlap, |
622 | pkt_seq, pkt_seq + pkt_len, pkt_len); | | 622 | pkt_seq, pkt_seq + pkt_len, pkt_len); |
623 | #endif | | 623 | #endif |
624 | m_adj(m, overlap); | | 624 | m_adj(m, overlap); |
625 | rcvpartdupbyte += overlap; | | 625 | rcvpartdupbyte += overlap; |
626 | m_cat(q->ipre_mlast, m); | | 626 | m_cat(q->ipre_mlast, m); |
627 | TRAVERSE(q->ipre_mlast); | | 627 | TRAVERSE(q->ipre_mlast); |
628 | m = q->ipqe_m; | | 628 | m = q->ipqe_m; |
629 | pkt_seq = q->ipqe_seq; | | 629 | pkt_seq = q->ipqe_seq; |
630 | pkt_len += q->ipqe_len - overlap; | | 630 | pkt_len += q->ipqe_len - overlap; |
631 | rcvoobyte -= overlap; | | 631 | rcvoobyte -= overlap; |
632 | TCP_REASS_COUNTER_INCR(&tcp_reass_overlaptail); | | 632 | TCP_REASS_COUNTER_INCR(&tcp_reass_overlaptail); |
633 | goto free_ipqe; | | 633 | goto free_ipqe; |
634 | } | | 634 | } |
635 | /* | | 635 | /* |
636 | * RX'ed segment extends past the front of the | | 636 | * RX'ed segment extends past the front of the |
637 | * fragment. Drop the overlapping bytes on the | | 637 | * fragment. Drop the overlapping bytes on the |
638 | * received packet. The packet will then be | | 638 | * received packet. The packet will then be |
639 | * contatentated with this fragment a bit later. | | 639 | * contatentated with this fragment a bit later. |
640 | */ | | 640 | */ |
641 | if (SEQ_GT(q->ipqe_seq, pkt_seq) && | | 641 | if (SEQ_GT(q->ipqe_seq, pkt_seq) && |
642 | SEQ_LT(q->ipqe_seq, pkt_seq + pkt_len)) { | | 642 | SEQ_LT(q->ipqe_seq, pkt_seq + pkt_len)) { |
643 | int overlap = pkt_seq + pkt_len - q->ipqe_seq; | | 643 | int overlap = pkt_seq + pkt_len - q->ipqe_seq; |
644 | #ifdef TCPREASS_DEBUG | | 644 | #ifdef TCPREASS_DEBUG |
645 | printf("tcp_reass[%p]: trim trailing %d bytes of %u:%u(%u)\n", | | 645 | printf("tcp_reass[%p]: trim trailing %d bytes of %u:%u(%u)\n", |
646 | tp, overlap, | | 646 | tp, overlap, |
647 | pkt_seq, pkt_seq + pkt_len, pkt_len); | | 647 | pkt_seq, pkt_seq + pkt_len, pkt_len); |
648 | #endif | | 648 | #endif |
649 | m_adj(m, -overlap); | | 649 | m_adj(m, -overlap); |
650 | pkt_len -= overlap; | | 650 | pkt_len -= overlap; |
651 | rcvpartdupbyte += overlap; | | 651 | rcvpartdupbyte += overlap; |
652 | TCP_REASS_COUNTER_INCR(&tcp_reass_overlapfront); | | 652 | TCP_REASS_COUNTER_INCR(&tcp_reass_overlapfront); |
653 | rcvoobyte -= overlap; | | 653 | rcvoobyte -= overlap; |
654 | } | | 654 | } |
655 | /* | | 655 | /* |
656 | * If the received segment immediates precedes this | | 656 | * If the received segment immediates precedes this |
657 | * fragment then tack the fragment onto this segment | | 657 | * fragment then tack the fragment onto this segment |
658 | * and reinsert the data. | | 658 | * and reinsert the data. |
659 | */ | | 659 | */ |
660 | if (q->ipqe_seq == pkt_seq + pkt_len) { | | 660 | if (q->ipqe_seq == pkt_seq + pkt_len) { |
661 | #ifdef TCPREASS_DEBUG | | 661 | #ifdef TCPREASS_DEBUG |
662 | printf("tcp_reass[%p]: append %u:%u(%u) to %u:%u(%u)\n", | | 662 | printf("tcp_reass[%p]: append %u:%u(%u) to %u:%u(%u)\n", |
663 | tp, q->ipqe_seq, q->ipqe_seq + q->ipqe_len, q->ipqe_len, | | 663 | tp, q->ipqe_seq, q->ipqe_seq + q->ipqe_len, q->ipqe_len, |
664 | pkt_seq, pkt_seq + pkt_len, pkt_len); | | 664 | pkt_seq, pkt_seq + pkt_len, pkt_len); |
665 | #endif | | 665 | #endif |
666 | pkt_len += q->ipqe_len; | | 666 | pkt_len += q->ipqe_len; |
667 | pkt_flags |= q->ipqe_flags; | | 667 | pkt_flags |= q->ipqe_flags; |
668 | m_cat(m, q->ipqe_m); | | 668 | m_cat(m, q->ipqe_m); |
669 | TAILQ_REMOVE(&tp->segq, q, ipqe_q); | | 669 | TAILQ_REMOVE(&tp->segq, q, ipqe_q); |
670 | TAILQ_REMOVE(&tp->timeq, q, ipqe_timeq); | | 670 | TAILQ_REMOVE(&tp->timeq, q, ipqe_timeq); |
671 | tp->t_segqlen--; | | 671 | tp->t_segqlen--; |
672 | KASSERT(tp->t_segqlen >= 0); | | 672 | KASSERT(tp->t_segqlen >= 0); |
673 | KASSERT(tp->t_segqlen != 0 || | | 673 | KASSERT(tp->t_segqlen != 0 || |
674 | (TAILQ_EMPTY(&tp->segq) && | | 674 | (TAILQ_EMPTY(&tp->segq) && |
675 | TAILQ_EMPTY(&tp->timeq))); | | 675 | TAILQ_EMPTY(&tp->timeq))); |
676 | if (tiqe == NULL) { | | 676 | if (tiqe == NULL) { |
677 | tiqe = q; | | 677 | tiqe = q; |
678 | } else { | | 678 | } else { |
679 | tcpipqent_free(q); | | 679 | tcpipqent_free(q); |
680 | } | | 680 | } |
681 | TCP_REASS_COUNTER_INCR(&tcp_reass_prepend); | | 681 | TCP_REASS_COUNTER_INCR(&tcp_reass_prepend); |
682 | break; | | 682 | break; |
683 | } | | 683 | } |
684 | /* | | 684 | /* |
685 | * If the fragment is before the segment, remember it. | | 685 | * If the fragment is before the segment, remember it. |
686 | * When this loop is terminated, p will contain the | | 686 | * When this loop is terminated, p will contain the |
687 | * pointer to fragment that is right before the received | | 687 | * pointer to fragment that is right before the received |
688 | * segment. | | 688 | * segment. |
689 | */ | | 689 | */ |
690 | if (SEQ_LEQ(q->ipqe_seq, pkt_seq)) | | 690 | if (SEQ_LEQ(q->ipqe_seq, pkt_seq)) |
691 | p = q; | | 691 | p = q; |
692 | | | 692 | |
693 | continue; | | 693 | continue; |
694 | | | 694 | |
695 | /* | | 695 | /* |
696 | * This is a common operation. It also will allow | | 696 | * This is a common operation. It also will allow |
697 | * to save doing a malloc/free in most instances. | | 697 | * to save doing a malloc/free in most instances. |
698 | */ | | 698 | */ |
699 | free_ipqe: | | 699 | free_ipqe: |
700 | TAILQ_REMOVE(&tp->segq, q, ipqe_q); | | 700 | TAILQ_REMOVE(&tp->segq, q, ipqe_q); |
701 | TAILQ_REMOVE(&tp->timeq, q, ipqe_timeq); | | 701 | TAILQ_REMOVE(&tp->timeq, q, ipqe_timeq); |
702 | tp->t_segqlen--; | | 702 | tp->t_segqlen--; |
703 | KASSERT(tp->t_segqlen >= 0); | | 703 | KASSERT(tp->t_segqlen >= 0); |
704 | KASSERT(tp->t_segqlen != 0 || | | 704 | KASSERT(tp->t_segqlen != 0 || |
705 | (TAILQ_EMPTY(&tp->segq) && TAILQ_EMPTY(&tp->timeq))); | | 705 | (TAILQ_EMPTY(&tp->segq) && TAILQ_EMPTY(&tp->timeq))); |
706 | if (tiqe == NULL) { | | 706 | if (tiqe == NULL) { |
707 | tiqe = q; | | 707 | tiqe = q; |
708 | } else { | | 708 | } else { |
709 | tcpipqent_free(q); | | 709 | tcpipqent_free(q); |
710 | } | | 710 | } |
711 | } | | 711 | } |
712 | | | 712 | |
713 | #ifdef TCP_REASS_COUNTERS | | 713 | #ifdef TCP_REASS_COUNTERS |
714 | if (count > 7) | | 714 | if (count > 7) |
715 | TCP_REASS_COUNTER_INCR(&tcp_reass_iteration[0]); | | 715 | TCP_REASS_COUNTER_INCR(&tcp_reass_iteration[0]); |
716 | else if (count > 0) | | 716 | else if (count > 0) |
717 | TCP_REASS_COUNTER_INCR(&tcp_reass_iteration[count]); | | 717 | TCP_REASS_COUNTER_INCR(&tcp_reass_iteration[count]); |
718 | #endif | | 718 | #endif |
719 | | | 719 | |
720 | insert_it: | | 720 | insert_it: |
721 | | | 721 | |
722 | /* | | 722 | /* |
723 | * Allocate a new queue entry since the received segment did not | | 723 | * Allocate a new queue entry since the received segment did not |
724 | * collapse onto any other out-of-order block; thus we are allocating | | 724 | * collapse onto any other out-of-order block; thus we are allocating |
725 | * a new block. If it had collapsed, tiqe would not be NULL and | | 725 | * a new block. If it had collapsed, tiqe would not be NULL and |
726 | * we would be reusing it. | | 726 | * we would be reusing it. |
727 | * XXX If we can't, just drop the packet. XXX | | 727 | * XXX If we can't, just drop the packet. XXX |
728 | */ | | 728 | */ |
729 | if (tiqe == NULL) { | | 729 | if (tiqe == NULL) { |
730 | tiqe = tcpipqent_alloc(); | | 730 | tiqe = tcpipqent_alloc(); |
731 | if (tiqe == NULL) { | | 731 | if (tiqe == NULL) { |
732 | TCP_STATINC(TCP_STAT_RCVMEMDROP); | | 732 | TCP_STATINC(TCP_STAT_RCVMEMDROP); |
733 | m_freem(m); | | 733 | m_freem(m); |
734 | return (0); | | 734 | return (0); |
735 | } | | 735 | } |
736 | } | | 736 | } |
737 | | | 737 | |
738 | /* | | 738 | /* |
739 | * Update the counters. | | 739 | * Update the counters. |
740 | */ | | 740 | */ |
741 | tcps = TCP_STAT_GETREF(); | | 741 | tcps = TCP_STAT_GETREF(); |
742 | tcps[TCP_STAT_RCVOOPACK]++; | | 742 | tcps[TCP_STAT_RCVOOPACK]++; |
743 | tcps[TCP_STAT_RCVOOBYTE] += rcvoobyte; | | 743 | tcps[TCP_STAT_RCVOOBYTE] += rcvoobyte; |
744 | if (rcvpartdupbyte) { | | 744 | if (rcvpartdupbyte) { |
745 | tcps[TCP_STAT_RCVPARTDUPPACK]++; | | 745 | tcps[TCP_STAT_RCVPARTDUPPACK]++; |
746 | tcps[TCP_STAT_RCVPARTDUPBYTE] += rcvpartdupbyte; | | 746 | tcps[TCP_STAT_RCVPARTDUPBYTE] += rcvpartdupbyte; |
747 | } | | 747 | } |
748 | TCP_STAT_PUTREF(); | | 748 | TCP_STAT_PUTREF(); |
749 | | | 749 | |
750 | /* | | 750 | /* |
751 | * Insert the new fragment queue entry into both queues. | | 751 | * Insert the new fragment queue entry into both queues. |
752 | */ | | 752 | */ |
753 | tiqe->ipqe_m = m; | | 753 | tiqe->ipqe_m = m; |
754 | tiqe->ipre_mlast = m; | | 754 | tiqe->ipre_mlast = m; |
755 | tiqe->ipqe_seq = pkt_seq; | | 755 | tiqe->ipqe_seq = pkt_seq; |
756 | tiqe->ipqe_len = pkt_len; | | 756 | tiqe->ipqe_len = pkt_len; |
757 | tiqe->ipqe_flags = pkt_flags; | | 757 | tiqe->ipqe_flags = pkt_flags; |
758 | if (p == NULL) { | | 758 | if (p == NULL) { |
759 | TAILQ_INSERT_HEAD(&tp->segq, tiqe, ipqe_q); | | 759 | TAILQ_INSERT_HEAD(&tp->segq, tiqe, ipqe_q); |
760 | #ifdef TCPREASS_DEBUG | | 760 | #ifdef TCPREASS_DEBUG |
761 | if (tiqe->ipqe_seq != tp->rcv_nxt) | | 761 | if (tiqe->ipqe_seq != tp->rcv_nxt) |
762 | printf("tcp_reass[%p]: insert %u:%u(%u) at front\n", | | 762 | printf("tcp_reass[%p]: insert %u:%u(%u) at front\n", |
763 | tp, pkt_seq, pkt_seq + pkt_len, pkt_len); | | 763 | tp, pkt_seq, pkt_seq + pkt_len, pkt_len); |
764 | #endif | | 764 | #endif |
765 | } else { | | 765 | } else { |
766 | TAILQ_INSERT_AFTER(&tp->segq, p, tiqe, ipqe_q); | | 766 | TAILQ_INSERT_AFTER(&tp->segq, p, tiqe, ipqe_q); |
767 | #ifdef TCPREASS_DEBUG | | 767 | #ifdef TCPREASS_DEBUG |
768 | printf("tcp_reass[%p]: insert %u:%u(%u) after %u:%u(%u)\n", | | 768 | printf("tcp_reass[%p]: insert %u:%u(%u) after %u:%u(%u)\n", |
769 | tp, pkt_seq, pkt_seq + pkt_len, pkt_len, | | 769 | tp, pkt_seq, pkt_seq + pkt_len, pkt_len, |
770 | p->ipqe_seq, p->ipqe_seq + p->ipqe_len, p->ipqe_len); | | 770 | p->ipqe_seq, p->ipqe_seq + p->ipqe_len, p->ipqe_len); |
771 | #endif | | 771 | #endif |
772 | } | | 772 | } |
773 | tp->t_segqlen++; | | 773 | tp->t_segqlen++; |
774 | | | 774 | |
775 | skip_replacement: | | 775 | skip_replacement: |
776 | | | 776 | |
777 | TAILQ_INSERT_HEAD(&tp->timeq, tiqe, ipqe_timeq); | | 777 | TAILQ_INSERT_HEAD(&tp->timeq, tiqe, ipqe_timeq); |
778 | | | 778 | |
779 | present: | | 779 | present: |
780 | /* | | 780 | /* |
781 | * Present data to user, advancing rcv_nxt through | | 781 | * Present data to user, advancing rcv_nxt through |
782 | * completed sequence space. | | 782 | * completed sequence space. |
783 | */ | | 783 | */ |
784 | if (TCPS_HAVEESTABLISHED(tp->t_state) == 0) | | 784 | if (TCPS_HAVEESTABLISHED(tp->t_state) == 0) |
785 | return (0); | | 785 | return (0); |
786 | q = TAILQ_FIRST(&tp->segq); | | 786 | q = TAILQ_FIRST(&tp->segq); |
787 | if (q == NULL || q->ipqe_seq != tp->rcv_nxt) | | 787 | if (q == NULL || q->ipqe_seq != tp->rcv_nxt) |
788 | return (0); | | 788 | return (0); |
789 | if (tp->t_state == TCPS_SYN_RECEIVED && q->ipqe_len) | | 789 | if (tp->t_state == TCPS_SYN_RECEIVED && q->ipqe_len) |
790 | return (0); | | 790 | return (0); |
791 | | | 791 | |
792 | tp->rcv_nxt += q->ipqe_len; | | 792 | tp->rcv_nxt += q->ipqe_len; |
793 | pkt_flags = q->ipqe_flags & TH_FIN; | | 793 | pkt_flags = q->ipqe_flags & TH_FIN; |
794 | nd6_hint(tp); | | 794 | nd6_hint(tp); |
795 | | | 795 | |
796 | TAILQ_REMOVE(&tp->segq, q, ipqe_q); | | 796 | TAILQ_REMOVE(&tp->segq, q, ipqe_q); |
797 | TAILQ_REMOVE(&tp->timeq, q, ipqe_timeq); | | 797 | TAILQ_REMOVE(&tp->timeq, q, ipqe_timeq); |
798 | tp->t_segqlen--; | | 798 | tp->t_segqlen--; |
799 | KASSERT(tp->t_segqlen >= 0); | | 799 | KASSERT(tp->t_segqlen >= 0); |
800 | KASSERT(tp->t_segqlen != 0 || | | 800 | KASSERT(tp->t_segqlen != 0 || |
801 | (TAILQ_EMPTY(&tp->segq) && TAILQ_EMPTY(&tp->timeq))); | | 801 | (TAILQ_EMPTY(&tp->segq) && TAILQ_EMPTY(&tp->timeq))); |
802 | if (so->so_state & SS_CANTRCVMORE) | | 802 | if (so->so_state & SS_CANTRCVMORE) |
803 | m_freem(q->ipqe_m); | | 803 | m_freem(q->ipqe_m); |
804 | else | | 804 | else |
805 | sbappendstream(&so->so_rcv, q->ipqe_m); | | 805 | sbappendstream(&so->so_rcv, q->ipqe_m); |
806 | tcpipqent_free(q); | | 806 | tcpipqent_free(q); |
807 | sorwakeup(so); | | 807 | sorwakeup(so); |
808 | return (pkt_flags); | | 808 | return (pkt_flags); |
809 | } | | 809 | } |
810 | | | 810 | |
811 | #ifdef INET6 | | 811 | #ifdef INET6 |
812 | int | | 812 | int |
813 | tcp6_input(struct mbuf **mp, int *offp, int proto) | | 813 | tcp6_input(struct mbuf **mp, int *offp, int proto) |
814 | { | | 814 | { |
815 | struct mbuf *m = *mp; | | 815 | struct mbuf *m = *mp; |
816 | | | 816 | |
817 | /* | | 817 | /* |
818 | * draft-itojun-ipv6-tcp-to-anycast | | 818 | * draft-itojun-ipv6-tcp-to-anycast |
819 | * better place to put this in? | | 819 | * better place to put this in? |
820 | */ | | 820 | */ |
821 | if (m->m_flags & M_ANYCAST6) { | | 821 | if (m->m_flags & M_ANYCAST6) { |
822 | struct ip6_hdr *ip6; | | 822 | struct ip6_hdr *ip6; |
823 | if (m->m_len < sizeof(struct ip6_hdr)) { | | 823 | if (m->m_len < sizeof(struct ip6_hdr)) { |
824 | if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) { | | 824 | if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) { |
825 | TCP_STATINC(TCP_STAT_RCVSHORT); | | 825 | TCP_STATINC(TCP_STAT_RCVSHORT); |
826 | return IPPROTO_DONE; | | 826 | return IPPROTO_DONE; |
827 | } | | 827 | } |
828 | } | | 828 | } |
829 | ip6 = mtod(m, struct ip6_hdr *); | | 829 | ip6 = mtod(m, struct ip6_hdr *); |
830 | icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_ADDR, | | 830 | icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_ADDR, |
831 | (char *)&ip6->ip6_dst - (char *)ip6); | | 831 | (char *)&ip6->ip6_dst - (char *)ip6); |
832 | return IPPROTO_DONE; | | 832 | return IPPROTO_DONE; |
833 | } | | 833 | } |
834 | | | 834 | |
835 | tcp_input(m, *offp, proto); | | 835 | tcp_input(m, *offp, proto); |
836 | return IPPROTO_DONE; | | 836 | return IPPROTO_DONE; |
837 | } | | 837 | } |
838 | #endif | | 838 | #endif |
839 | | | 839 | |
840 | #ifdef INET | | 840 | #ifdef INET |
841 | static void | | 841 | static void |
842 | tcp4_log_refused(const struct ip *ip, const struct tcphdr *th) | | 842 | tcp4_log_refused(const struct ip *ip, const struct tcphdr *th) |
843 | { | | 843 | { |
844 | char src[4*sizeof "123"]; | | 844 | char src[4*sizeof "123"]; |
845 | char dst[4*sizeof "123"]; | | 845 | char dst[4*sizeof "123"]; |
846 | | | 846 | |
847 | if (ip) { | | 847 | if (ip) { |
848 | strlcpy(src, inet_ntoa(ip->ip_src), sizeof(src)); | | 848 | strlcpy(src, inet_ntoa(ip->ip_src), sizeof(src)); |
849 | strlcpy(dst, inet_ntoa(ip->ip_dst), sizeof(dst)); | | 849 | strlcpy(dst, inet_ntoa(ip->ip_dst), sizeof(dst)); |
850 | } | | 850 | } |
851 | else { | | 851 | else { |
852 | strlcpy(src, "(unknown)", sizeof(src)); | | 852 | strlcpy(src, "(unknown)", sizeof(src)); |
853 | strlcpy(dst, "(unknown)", sizeof(dst)); | | 853 | strlcpy(dst, "(unknown)", sizeof(dst)); |
854 | } | | 854 | } |
855 | log(LOG_INFO, | | 855 | log(LOG_INFO, |
856 | "Connection attempt to TCP %s:%d from %s:%d\n", | | 856 | "Connection attempt to TCP %s:%d from %s:%d\n", |
857 | dst, ntohs(th->th_dport), | | 857 | dst, ntohs(th->th_dport), |
858 | src, ntohs(th->th_sport)); | | 858 | src, ntohs(th->th_sport)); |
859 | } | | 859 | } |
860 | #endif | | 860 | #endif |
861 | | | 861 | |
862 | #ifdef INET6 | | 862 | #ifdef INET6 |
863 | static void | | 863 | static void |
864 | tcp6_log_refused(const struct ip6_hdr *ip6, const struct tcphdr *th) | | 864 | tcp6_log_refused(const struct ip6_hdr *ip6, const struct tcphdr *th) |
865 | { | | 865 | { |
866 | char src[INET6_ADDRSTRLEN]; | | 866 | char src[INET6_ADDRSTRLEN]; |
867 | char dst[INET6_ADDRSTRLEN]; | | 867 | char dst[INET6_ADDRSTRLEN]; |
868 | | | 868 | |
869 | if (ip6) { | | 869 | if (ip6) { |
870 | strlcpy(src, ip6_sprintf(&ip6->ip6_src), sizeof(src)); | | 870 | strlcpy(src, ip6_sprintf(&ip6->ip6_src), sizeof(src)); |
871 | strlcpy(dst, ip6_sprintf(&ip6->ip6_dst), sizeof(dst)); | | 871 | strlcpy(dst, ip6_sprintf(&ip6->ip6_dst), sizeof(dst)); |
872 | } | | 872 | } |
873 | else { | | 873 | else { |
874 | strlcpy(src, "(unknown v6)", sizeof(src)); | | 874 | strlcpy(src, "(unknown v6)", sizeof(src)); |
875 | strlcpy(dst, "(unknown v6)", sizeof(dst)); | | 875 | strlcpy(dst, "(unknown v6)", sizeof(dst)); |
876 | } | | 876 | } |
877 | log(LOG_INFO, | | 877 | log(LOG_INFO, |
878 | "Connection attempt to TCP [%s]:%d from [%s]:%d\n", | | 878 | "Connection attempt to TCP [%s]:%d from [%s]:%d\n", |
879 | dst, ntohs(th->th_dport), | | 879 | dst, ntohs(th->th_dport), |
880 | src, ntohs(th->th_sport)); | | 880 | src, ntohs(th->th_sport)); |
881 | } | | 881 | } |
882 | #endif | | 882 | #endif |
883 | | | 883 | |
884 | /* | | 884 | /* |
885 | * Checksum extended TCP header and data. | | 885 | * Checksum extended TCP header and data. |
886 | */ | | 886 | */ |
887 | int | | 887 | int |
888 | tcp_input_checksum(int af, struct mbuf *m, const struct tcphdr *th, | | 888 | tcp_input_checksum(int af, struct mbuf *m, const struct tcphdr *th, |
889 | int toff, int off, int tlen) | | 889 | int toff, int off, int tlen) |
890 | { | | 890 | { |
891 | | | 891 | |
892 | /* | | 892 | /* |
893 | * XXX it's better to record and check if this mbuf is | | 893 | * XXX it's better to record and check if this mbuf is |
894 | * already checked. | | 894 | * already checked. |
895 | */ | | 895 | */ |
896 | | | 896 | |
897 | switch (af) { | | 897 | switch (af) { |
898 | #ifdef INET | | 898 | #ifdef INET |
899 | case AF_INET: | | 899 | case AF_INET: |
900 | switch (m->m_pkthdr.csum_flags & | | 900 | switch (m->m_pkthdr.csum_flags & |
901 | ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_TCPv4) | | | 901 | ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_TCPv4) | |
902 | M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) { | | 902 | M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) { |
903 | case M_CSUM_TCPv4|M_CSUM_TCP_UDP_BAD: | | 903 | case M_CSUM_TCPv4|M_CSUM_TCP_UDP_BAD: |
904 | TCP_CSUM_COUNTER_INCR(&tcp_hwcsum_bad); | | 904 | TCP_CSUM_COUNTER_INCR(&tcp_hwcsum_bad); |
905 | goto badcsum; | | 905 | goto badcsum; |
906 | | | 906 | |
907 | case M_CSUM_TCPv4|M_CSUM_DATA: { | | 907 | case M_CSUM_TCPv4|M_CSUM_DATA: { |
908 | u_int32_t hw_csum = m->m_pkthdr.csum_data; | | 908 | u_int32_t hw_csum = m->m_pkthdr.csum_data; |
909 | | | 909 | |
910 | TCP_CSUM_COUNTER_INCR(&tcp_hwcsum_data); | | 910 | TCP_CSUM_COUNTER_INCR(&tcp_hwcsum_data); |
911 | if (m->m_pkthdr.csum_flags & M_CSUM_NO_PSEUDOHDR) { | | 911 | if (m->m_pkthdr.csum_flags & M_CSUM_NO_PSEUDOHDR) { |
912 | const struct ip *ip = | | 912 | const struct ip *ip = |
913 | mtod(m, const struct ip *); | | 913 | mtod(m, const struct ip *); |
914 | | | 914 | |
915 | hw_csum = in_cksum_phdr(ip->ip_src.s_addr, | | 915 | hw_csum = in_cksum_phdr(ip->ip_src.s_addr, |
916 | ip->ip_dst.s_addr, | | 916 | ip->ip_dst.s_addr, |
917 | htons(hw_csum + tlen + off + IPPROTO_TCP)); | | 917 | htons(hw_csum + tlen + off + IPPROTO_TCP)); |
918 | } | | 918 | } |
919 | if ((hw_csum ^ 0xffff) != 0) | | 919 | if ((hw_csum ^ 0xffff) != 0) |
920 | goto badcsum; | | 920 | goto badcsum; |
921 | break; | | 921 | break; |
922 | } | | 922 | } |
923 | | | 923 | |
924 | case M_CSUM_TCPv4: | | 924 | case M_CSUM_TCPv4: |
925 | /* Checksum was okay. */ | | 925 | /* Checksum was okay. */ |
926 | TCP_CSUM_COUNTER_INCR(&tcp_hwcsum_ok); | | 926 | TCP_CSUM_COUNTER_INCR(&tcp_hwcsum_ok); |
927 | break; | | 927 | break; |
928 | | | 928 | |
929 | default: | | 929 | default: |
930 | /* | | 930 | /* |
931 | * Must compute it ourselves. Maybe skip checksum | | 931 | * Must compute it ourselves. Maybe skip checksum |
932 | * on loopback interfaces. | | 932 | * on loopback interfaces. |
933 | */ | | 933 | */ |
934 | if (__predict_true(!(m->m_pkthdr.rcvif->if_flags & | | 934 | if (__predict_true(!(m->m_pkthdr.rcvif->if_flags & |
935 | IFF_LOOPBACK) || | | 935 | IFF_LOOPBACK) || |
936 | tcp_do_loopback_cksum)) { | | 936 | tcp_do_loopback_cksum)) { |
937 | TCP_CSUM_COUNTER_INCR(&tcp_swcsum); | | 937 | TCP_CSUM_COUNTER_INCR(&tcp_swcsum); |
938 | if (in4_cksum(m, IPPROTO_TCP, toff, | | 938 | if (in4_cksum(m, IPPROTO_TCP, toff, |
939 | tlen + off) != 0) | | 939 | tlen + off) != 0) |
940 | goto badcsum; | | 940 | goto badcsum; |
941 | } | | 941 | } |
942 | break; | | 942 | break; |
943 | } | | 943 | } |
944 | break; | | 944 | break; |
945 | #endif /* INET4 */ | | 945 | #endif /* INET4 */ |
946 | | | 946 | |
947 | #ifdef INET6 | | 947 | #ifdef INET6 |
948 | case AF_INET6: | | 948 | case AF_INET6: |
949 | switch (m->m_pkthdr.csum_flags & | | 949 | switch (m->m_pkthdr.csum_flags & |
950 | ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_TCPv6) | | | 950 | ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_TCPv6) | |
951 | M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) { | | 951 | M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) { |
952 | case M_CSUM_TCPv6|M_CSUM_TCP_UDP_BAD: | | 952 | case M_CSUM_TCPv6|M_CSUM_TCP_UDP_BAD: |
953 | TCP_CSUM_COUNTER_INCR(&tcp6_hwcsum_bad); | | 953 | TCP_CSUM_COUNTER_INCR(&tcp6_hwcsum_bad); |
954 | goto badcsum; | | 954 | goto badcsum; |
955 | | | 955 | |
956 | #if 0 /* notyet */ | | 956 | #if 0 /* notyet */ |
957 | case M_CSUM_TCPv6|M_CSUM_DATA: | | 957 | case M_CSUM_TCPv6|M_CSUM_DATA: |
958 | #endif | | 958 | #endif |
959 | | | 959 | |
960 | case M_CSUM_TCPv6: | | 960 | case M_CSUM_TCPv6: |
961 | /* Checksum was okay. */ | | 961 | /* Checksum was okay. */ |
962 | TCP_CSUM_COUNTER_INCR(&tcp6_hwcsum_ok); | | 962 | TCP_CSUM_COUNTER_INCR(&tcp6_hwcsum_ok); |
963 | break; | | 963 | break; |
964 | | | 964 | |
965 | default: | | 965 | default: |
966 | /* | | 966 | /* |
967 | * Must compute it ourselves. Maybe skip checksum | | 967 | * Must compute it ourselves. Maybe skip checksum |
968 | * on loopback interfaces. | | 968 | * on loopback interfaces. |
969 | */ | | 969 | */ |
970 | if (__predict_true((m->m_flags & M_LOOP) == 0 || | | 970 | if (__predict_true((m->m_flags & M_LOOP) == 0 || |
971 | tcp_do_loopback_cksum)) { | | 971 | tcp_do_loopback_cksum)) { |
972 | TCP_CSUM_COUNTER_INCR(&tcp6_swcsum); | | 972 | TCP_CSUM_COUNTER_INCR(&tcp6_swcsum); |
973 | if (in6_cksum(m, IPPROTO_TCP, toff, | | 973 | if (in6_cksum(m, IPPROTO_TCP, toff, |
974 | tlen + off) != 0) | | 974 | tlen + off) != 0) |
975 | goto badcsum; | | 975 | goto badcsum; |
976 | } | | 976 | } |
977 | } | | 977 | } |
978 | break; | | 978 | break; |
979 | #endif /* INET6 */ | | 979 | #endif /* INET6 */ |
980 | } | | 980 | } |
981 | | | 981 | |
982 | return 0; | | 982 | return 0; |
983 | | | 983 | |
984 | badcsum: | | 984 | badcsum: |
985 | TCP_STATINC(TCP_STAT_RCVBADSUM); | | 985 | TCP_STATINC(TCP_STAT_RCVBADSUM); |
986 | return -1; | | 986 | return -1; |
987 | } | | 987 | } |
988 | | | 988 | |
989 | /* | | 989 | /* |
990 | * TCP input routine, follows pages 65-76 of RFC 793 very closely. | | 990 | * TCP input routine, follows pages 65-76 of RFC 793 very closely. |
991 | */ | | 991 | */ |
992 | void | | 992 | void |
993 | tcp_input(struct mbuf *m, ...) | | 993 | tcp_input(struct mbuf *m, ...) |
994 | { | | 994 | { |
995 | struct tcphdr *th; | | 995 | struct tcphdr *th; |
996 | struct ip *ip; | | 996 | struct ip *ip; |
997 | struct inpcb *inp; | | 997 | struct inpcb *inp; |
998 | #ifdef INET6 | | 998 | #ifdef INET6 |
999 | struct ip6_hdr *ip6; | | 999 | struct ip6_hdr *ip6; |
1000 | struct in6pcb *in6p; | | 1000 | struct in6pcb *in6p; |
1001 | #endif | | 1001 | #endif |
1002 | u_int8_t *optp = NULL; | | 1002 | u_int8_t *optp = NULL; |
1003 | int optlen = 0; | | 1003 | int optlen = 0; |
1004 | int len, tlen, toff, hdroptlen = 0; | | 1004 | int len, tlen, toff, hdroptlen = 0; |
1005 | struct tcpcb *tp = 0; | | 1005 | struct tcpcb *tp = 0; |
1006 | int tiflags; | | 1006 | int tiflags; |
1007 | struct socket *so = NULL; | | 1007 | struct socket *so = NULL; |
1008 | int todrop, dupseg, acked, ourfinisacked, needoutput = 0; | | 1008 | int todrop, dupseg, acked, ourfinisacked, needoutput = 0; |
1009 | #ifdef TCP_DEBUG | | 1009 | #ifdef TCP_DEBUG |
1010 | short ostate = 0; | | 1010 | short ostate = 0; |
1011 | #endif | | 1011 | #endif |
1012 | u_long tiwin; | | 1012 | u_long tiwin; |
1013 | struct tcp_opt_info opti; | | 1013 | struct tcp_opt_info opti; |
1014 | int off, iphlen; | | 1014 | int off, iphlen; |
1015 | va_list ap; | | 1015 | va_list ap; |
1016 | int af; /* af on the wire */ | | 1016 | int af; /* af on the wire */ |
1017 | struct mbuf *tcp_saveti = NULL; | | 1017 | struct mbuf *tcp_saveti = NULL; |
1018 | uint32_t ts_rtt; | | 1018 | uint32_t ts_rtt; |
1019 | uint8_t iptos; | | 1019 | uint8_t iptos; |
1020 | uint64_t *tcps; | | 1020 | uint64_t *tcps; |
1021 | | | 1021 | |
1022 | MCLAIM(m, &tcp_rx_mowner); | | 1022 | MCLAIM(m, &tcp_rx_mowner); |
1023 | va_start(ap, m); | | 1023 | va_start(ap, m); |
1024 | toff = va_arg(ap, int); | | 1024 | toff = va_arg(ap, int); |
1025 | (void)va_arg(ap, int); /* ignore value, advance ap */ | | 1025 | (void)va_arg(ap, int); /* ignore value, advance ap */ |
1026 | va_end(ap); | | 1026 | va_end(ap); |
1027 | | | 1027 | |
1028 | TCP_STATINC(TCP_STAT_RCVTOTAL); | | 1028 | TCP_STATINC(TCP_STAT_RCVTOTAL); |
1029 | | | 1029 | |
1030 | memset(&opti, 0, sizeof(opti)); | | 1030 | memset(&opti, 0, sizeof(opti)); |
1031 | opti.ts_present = 0; | | 1031 | opti.ts_present = 0; |
1032 | opti.maxseg = 0; | | 1032 | opti.maxseg = 0; |
1033 | | | 1033 | |
1034 | /* | | 1034 | /* |
1035 | * RFC1122 4.2.3.10, p. 104: discard bcast/mcast SYN. | | 1035 | * RFC1122 4.2.3.10, p. 104: discard bcast/mcast SYN. |
1036 | * | | 1036 | * |
1037 | * TCP is, by definition, unicast, so we reject all | | 1037 | * TCP is, by definition, unicast, so we reject all |
1038 | * multicast outright. | | 1038 | * multicast outright. |
1039 | * | | 1039 | * |
1040 | * Note, there are additional src/dst address checks in | | 1040 | * Note, there are additional src/dst address checks in |
1041 | * the AF-specific code below. | | 1041 | * the AF-specific code below. |
1042 | */ | | 1042 | */ |
1043 | if (m->m_flags & (M_BCAST|M_MCAST)) { | | 1043 | if (m->m_flags & (M_BCAST|M_MCAST)) { |
1044 | /* XXX stat */ | | 1044 | /* XXX stat */ |
1045 | goto drop; | | 1045 | goto drop; |
1046 | } | | 1046 | } |
1047 | #ifdef INET6 | | 1047 | #ifdef INET6 |
1048 | if (m->m_flags & M_ANYCAST6) { | | 1048 | if (m->m_flags & M_ANYCAST6) { |
1049 | /* XXX stat */ | | 1049 | /* XXX stat */ |
1050 | goto drop; | | 1050 | goto drop; |
1051 | } | | 1051 | } |
1052 | #endif | | 1052 | #endif |
1053 | | | 1053 | |
1054 | /* | | 1054 | /* |
1055 | * Get IP and TCP header. | | 1055 | * Get IP and TCP header. |
1056 | * Note: IP leaves IP header in first mbuf. | | 1056 | * Note: IP leaves IP header in first mbuf. |
1057 | */ | | 1057 | */ |
1058 | ip = mtod(m, struct ip *); | | 1058 | ip = mtod(m, struct ip *); |
1059 | #ifdef INET6 | | 1059 | #ifdef INET6 |
1060 | ip6 = NULL; | | 1060 | ip6 = NULL; |
1061 | #endif | | 1061 | #endif |
1062 | switch (ip->ip_v) { | | 1062 | switch (ip->ip_v) { |
1063 | #ifdef INET | | 1063 | #ifdef INET |
1064 | case 4: | | 1064 | case 4: |
1065 | af = AF_INET; | | 1065 | af = AF_INET; |
1066 | iphlen = sizeof(struct ip); | | 1066 | iphlen = sizeof(struct ip); |
1067 | ip = mtod(m, struct ip *); | | 1067 | ip = mtod(m, struct ip *); |
1068 | IP6_EXTHDR_GET(th, struct tcphdr *, m, toff, | | 1068 | IP6_EXTHDR_GET(th, struct tcphdr *, m, toff, |
1069 | sizeof(struct tcphdr)); | | 1069 | sizeof(struct tcphdr)); |
1070 | if (th == NULL) { | | 1070 | if (th == NULL) { |
1071 | TCP_STATINC(TCP_STAT_RCVSHORT); | | 1071 | TCP_STATINC(TCP_STAT_RCVSHORT); |
1072 | return; | | 1072 | return; |
1073 | } | | 1073 | } |
1074 | /* We do the checksum after PCB lookup... */ | | 1074 | /* We do the checksum after PCB lookup... */ |
1075 | len = ntohs(ip->ip_len); | | 1075 | len = ntohs(ip->ip_len); |
1076 | tlen = len - toff; | | 1076 | tlen = len - toff; |
1077 | iptos = ip->ip_tos; | | 1077 | iptos = ip->ip_tos; |
1078 | break; | | 1078 | break; |
1079 | #endif | | 1079 | #endif |
1080 | #ifdef INET6 | | 1080 | #ifdef INET6 |
1081 | case 6: | | 1081 | case 6: |
1082 | ip = NULL; | | 1082 | ip = NULL; |
1083 | iphlen = sizeof(struct ip6_hdr); | | 1083 | iphlen = sizeof(struct ip6_hdr); |
1084 | af = AF_INET6; | | 1084 | af = AF_INET6; |
1085 | ip6 = mtod(m, struct ip6_hdr *); | | 1085 | ip6 = mtod(m, struct ip6_hdr *); |
1086 | IP6_EXTHDR_GET(th, struct tcphdr *, m, toff, | | 1086 | IP6_EXTHDR_GET(th, struct tcphdr *, m, toff, |
1087 | sizeof(struct tcphdr)); | | 1087 | sizeof(struct tcphdr)); |
1088 | if (th == NULL) { | | 1088 | if (th == NULL) { |
1089 | TCP_STATINC(TCP_STAT_RCVSHORT); | | 1089 | TCP_STATINC(TCP_STAT_RCVSHORT); |
1090 | return; | | 1090 | return; |
1091 | } | | 1091 | } |
1092 | | | 1092 | |
1093 | /* Be proactive about malicious use of IPv4 mapped address */ | | 1093 | /* Be proactive about malicious use of IPv4 mapped address */ |
1094 | if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) || | | 1094 | if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) || |
1095 | IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) { | | 1095 | IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) { |
1096 | /* XXX stat */ | | 1096 | /* XXX stat */ |
1097 | goto drop; | | 1097 | goto drop; |
1098 | } | | 1098 | } |
1099 | | | 1099 | |
1100 | /* | | 1100 | /* |
1101 | * Be proactive about unspecified IPv6 address in source. | | 1101 | * Be proactive about unspecified IPv6 address in source. |
1102 | * As we use all-zero to indicate unbounded/unconnected pcb, | | 1102 | * As we use all-zero to indicate unbounded/unconnected pcb, |
1103 | * unspecified IPv6 address can be used to confuse us. | | 1103 | * unspecified IPv6 address can be used to confuse us. |
1104 | * | | 1104 | * |
1105 | * Note that packets with unspecified IPv6 destination is | | 1105 | * Note that packets with unspecified IPv6 destination is |
1106 | * already dropped in ip6_input. | | 1106 | * already dropped in ip6_input. |
1107 | */ | | 1107 | */ |
1108 | if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) { | | 1108 | if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) { |
1109 | /* XXX stat */ | | 1109 | /* XXX stat */ |
1110 | goto drop; | | 1110 | goto drop; |
1111 | } | | 1111 | } |
1112 | | | 1112 | |
1113 | /* | | 1113 | /* |
1114 | * Make sure destination address is not multicast. | | 1114 | * Make sure destination address is not multicast. |
1115 | * Source address checked in ip6_input(). | | 1115 | * Source address checked in ip6_input(). |
1116 | */ | | 1116 | */ |
1117 | if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { | | 1117 | if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { |
1118 | /* XXX stat */ | | 1118 | /* XXX stat */ |
1119 | goto drop; | | 1119 | goto drop; |
1120 | } | | 1120 | } |
1121 | | | 1121 | |
1122 | /* We do the checksum after PCB lookup... */ | | 1122 | /* We do the checksum after PCB lookup... */ |
1123 | len = m->m_pkthdr.len; | | 1123 | len = m->m_pkthdr.len; |
1124 | tlen = len - toff; | | 1124 | tlen = len - toff; |
1125 | iptos = (ntohl(ip6->ip6_flow) >> 20) & 0xff; | | 1125 | iptos = (ntohl(ip6->ip6_flow) >> 20) & 0xff; |
1126 | break; | | 1126 | break; |
1127 | #endif | | 1127 | #endif |
1128 | default: | | 1128 | default: |
1129 | m_freem(m); | | 1129 | m_freem(m); |
1130 | return; | | 1130 | return; |
1131 | } | | 1131 | } |
1132 | | | 1132 | |
1133 | KASSERT(TCP_HDR_ALIGNED_P(th)); | | 1133 | KASSERT(TCP_HDR_ALIGNED_P(th)); |
1134 | | | 1134 | |
1135 | /* | | 1135 | /* |
1136 | * Check that TCP offset makes sense, | | 1136 | * Check that TCP offset makes sense, |
1137 | * pull out TCP options and adjust length. XXX | | 1137 | * pull out TCP options and adjust length. XXX |
1138 | */ | | 1138 | */ |
1139 | off = th->th_off << 2; | | 1139 | off = th->th_off << 2; |
1140 | if (off < sizeof (struct tcphdr) || off > tlen) { | | 1140 | if (off < sizeof (struct tcphdr) || off > tlen) { |
1141 | TCP_STATINC(TCP_STAT_RCVBADOFF); | | 1141 | TCP_STATINC(TCP_STAT_RCVBADOFF); |
1142 | goto drop; | | 1142 | goto drop; |
1143 | } | | 1143 | } |
1144 | tlen -= off; | | 1144 | tlen -= off; |
1145 | | | 1145 | |
1146 | /* | | 1146 | /* |
1147 | * tcp_input() has been modified to use tlen to mean the TCP data | | 1147 | * tcp_input() has been modified to use tlen to mean the TCP data |
1148 | * length throughout the function. Other functions can use | | 1148 | * length throughout the function. Other functions can use |
1149 | * m->m_pkthdr.len as the basis for calculating the TCP data length. | | 1149 | * m->m_pkthdr.len as the basis for calculating the TCP data length. |
1150 | * rja | | 1150 | * rja |
1151 | */ | | 1151 | */ |
1152 | | | 1152 | |
1153 | if (off > sizeof (struct tcphdr)) { | | 1153 | if (off > sizeof (struct tcphdr)) { |
1154 | IP6_EXTHDR_GET(th, struct tcphdr *, m, toff, off); | | 1154 | IP6_EXTHDR_GET(th, struct tcphdr *, m, toff, off); |
1155 | if (th == NULL) { | | 1155 | if (th == NULL) { |
1156 | TCP_STATINC(TCP_STAT_RCVSHORT); | | 1156 | TCP_STATINC(TCP_STAT_RCVSHORT); |
1157 | return; | | 1157 | return; |
1158 | } | | 1158 | } |
1159 | /* | | 1159 | /* |
1160 | * NOTE: ip/ip6 will not be affected by m_pulldown() | | 1160 | * NOTE: ip/ip6 will not be affected by m_pulldown() |
1161 | * (as they're before toff) and we don't need to update those. | | 1161 | * (as they're before toff) and we don't need to update those. |
1162 | */ | | 1162 | */ |
1163 | KASSERT(TCP_HDR_ALIGNED_P(th)); | | 1163 | KASSERT(TCP_HDR_ALIGNED_P(th)); |
1164 | optlen = off - sizeof (struct tcphdr); | | 1164 | optlen = off - sizeof (struct tcphdr); |
1165 | optp = ((u_int8_t *)th) + sizeof(struct tcphdr); | | 1165 | optp = ((u_int8_t *)th) + sizeof(struct tcphdr); |
1166 | /* | | 1166 | /* |
1167 | * Do quick retrieval of timestamp options ("options | | 1167 | * Do quick retrieval of timestamp options ("options |
1168 | * prediction?"). If timestamp is the only option and it's | | 1168 | * prediction?"). If timestamp is the only option and it's |
1169 | * formatted as recommended in RFC 1323 appendix A, we | | 1169 | * formatted as recommended in RFC 1323 appendix A, we |
1170 | * quickly get the values now and not bother calling | | 1170 | * quickly get the values now and not bother calling |
1171 | * tcp_dooptions(), etc. | | 1171 | * tcp_dooptions(), etc. |
1172 | */ | | 1172 | */ |
1173 | if ((optlen == TCPOLEN_TSTAMP_APPA || | | 1173 | if ((optlen == TCPOLEN_TSTAMP_APPA || |
1174 | (optlen > TCPOLEN_TSTAMP_APPA && | | 1174 | (optlen > TCPOLEN_TSTAMP_APPA && |
1175 | optp[TCPOLEN_TSTAMP_APPA] == TCPOPT_EOL)) && | | 1175 | optp[TCPOLEN_TSTAMP_APPA] == TCPOPT_EOL)) && |
1176 | *(u_int32_t *)optp == htonl(TCPOPT_TSTAMP_HDR) && | | 1176 | *(u_int32_t *)optp == htonl(TCPOPT_TSTAMP_HDR) && |
1177 | (th->th_flags & TH_SYN) == 0) { | | 1177 | (th->th_flags & TH_SYN) == 0) { |
1178 | opti.ts_present = 1; | | 1178 | opti.ts_present = 1; |
1179 | opti.ts_val = ntohl(*(u_int32_t *)(optp + 4)); | | 1179 | opti.ts_val = ntohl(*(u_int32_t *)(optp + 4)); |
1180 | opti.ts_ecr = ntohl(*(u_int32_t *)(optp + 8)); | | 1180 | opti.ts_ecr = ntohl(*(u_int32_t *)(optp + 8)); |
1181 | optp = NULL; /* we've parsed the options */ | | 1181 | optp = NULL; /* we've parsed the options */ |
1182 | } | | 1182 | } |
1183 | } | | 1183 | } |
1184 | tiflags = th->th_flags; | | 1184 | tiflags = th->th_flags; |
1185 | | | 1185 | |
1186 | /* | | 1186 | /* |
1187 | * Locate pcb for segment. | | 1187 | * Locate pcb for segment. |
1188 | */ | | 1188 | */ |
1189 | findpcb: | | 1189 | findpcb: |
1190 | inp = NULL; | | 1190 | inp = NULL; |
1191 | #ifdef INET6 | | 1191 | #ifdef INET6 |
1192 | in6p = NULL; | | 1192 | in6p = NULL; |
1193 | #endif | | 1193 | #endif |
1194 | switch (af) { | | 1194 | switch (af) { |
1195 | #ifdef INET | | 1195 | #ifdef INET |
1196 | case AF_INET: | | 1196 | case AF_INET: |
1197 | inp = in_pcblookup_connect(&tcbtable, ip->ip_src, th->th_sport, | | 1197 | inp = in_pcblookup_connect(&tcbtable, ip->ip_src, th->th_sport, |
1198 | ip->ip_dst, th->th_dport); | | 1198 | ip->ip_dst, th->th_dport); |
1199 | if (inp == 0) { | | 1199 | if (inp == 0) { |
1200 | TCP_STATINC(TCP_STAT_PCBHASHMISS); | | 1200 | TCP_STATINC(TCP_STAT_PCBHASHMISS); |
1201 | inp = in_pcblookup_bind(&tcbtable, ip->ip_dst, th->th_dport); | | 1201 | inp = in_pcblookup_bind(&tcbtable, ip->ip_dst, th->th_dport); |
1202 | } | | 1202 | } |
1203 | #ifdef INET6 | | 1203 | #ifdef INET6 |
1204 | if (inp == 0) { | | 1204 | if (inp == 0) { |
1205 | struct in6_addr s, d; | | 1205 | struct in6_addr s, d; |
1206 | | | 1206 | |
1207 | /* mapped addr case */ | | 1207 | /* mapped addr case */ |
1208 | memset(&s, 0, sizeof(s)); | | 1208 | memset(&s, 0, sizeof(s)); |
1209 | s.s6_addr16[5] = htons(0xffff); | | 1209 | s.s6_addr16[5] = htons(0xffff); |
1210 | bcopy(&ip->ip_src, &s.s6_addr32[3], sizeof(ip->ip_src)); | | 1210 | bcopy(&ip->ip_src, &s.s6_addr32[3], sizeof(ip->ip_src)); |
1211 | memset(&d, 0, sizeof(d)); | | 1211 | memset(&d, 0, sizeof(d)); |
1212 | d.s6_addr16[5] = htons(0xffff); | | 1212 | d.s6_addr16[5] = htons(0xffff); |
1213 | bcopy(&ip->ip_dst, &d.s6_addr32[3], sizeof(ip->ip_dst)); | | 1213 | bcopy(&ip->ip_dst, &d.s6_addr32[3], sizeof(ip->ip_dst)); |
1214 | in6p = in6_pcblookup_connect(&tcbtable, &s, | | 1214 | in6p = in6_pcblookup_connect(&tcbtable, &s, |
1215 | th->th_sport, &d, th->th_dport, 0); | | 1215 | th->th_sport, &d, th->th_dport, 0); |
1216 | if (in6p == 0) { | | 1216 | if (in6p == 0) { |
1217 | TCP_STATINC(TCP_STAT_PCBHASHMISS); | | 1217 | TCP_STATINC(TCP_STAT_PCBHASHMISS); |
1218 | in6p = in6_pcblookup_bind(&tcbtable, &d, | | 1218 | in6p = in6_pcblookup_bind(&tcbtable, &d, |
1219 | th->th_dport, 0); | | 1219 | th->th_dport, 0); |
1220 | } | | 1220 | } |
1221 | } | | 1221 | } |
1222 | #endif | | 1222 | #endif |
1223 | #ifndef INET6 | | 1223 | #ifndef INET6 |
1224 | if (inp == 0) | | 1224 | if (inp == 0) |
1225 | #else | | 1225 | #else |
1226 | if (inp == 0 && in6p == 0) | | 1226 | if (inp == 0 && in6p == 0) |
1227 | #endif | | 1227 | #endif |
1228 | { | | 1228 | { |
1229 | TCP_STATINC(TCP_STAT_NOPORT); | | 1229 | TCP_STATINC(TCP_STAT_NOPORT); |
1230 | if (tcp_log_refused && | | 1230 | if (tcp_log_refused && |
1231 | (tiflags & (TH_RST|TH_ACK|TH_SYN)) == TH_SYN) { | | 1231 | (tiflags & (TH_RST|TH_ACK|TH_SYN)) == TH_SYN) { |
1232 | tcp4_log_refused(ip, th); | | 1232 | tcp4_log_refused(ip, th); |
1233 | } | | 1233 | } |
1234 | tcp_fields_to_host(th); | | 1234 | tcp_fields_to_host(th); |
1235 | goto dropwithreset_ratelim; | | 1235 | goto dropwithreset_ratelim; |
1236 | } | | 1236 | } |
1237 | #if defined(IPSEC) || defined(FAST_IPSEC) | | 1237 | #if defined(IPSEC) || defined(FAST_IPSEC) |
1238 | if (inp && (inp->inp_socket->so_options & SO_ACCEPTCONN) == 0 && | | 1238 | if (inp && (inp->inp_socket->so_options & SO_ACCEPTCONN) == 0 && |
1239 | ipsec4_in_reject(m, inp)) { | | 1239 | ipsec4_in_reject(m, inp)) { |
1240 | IPSEC_STATINC(IPSEC_STAT_IN_POLVIO); | | 1240 | IPSEC_STATINC(IPSEC_STAT_IN_POLVIO); |
1241 | goto drop; | | 1241 | goto drop; |
1242 | } | | 1242 | } |
1243 | #ifdef INET6 | | 1243 | #ifdef INET6 |
1244 | else if (in6p && | | 1244 | else if (in6p && |
1245 | (in6p->in6p_socket->so_options & SO_ACCEPTCONN) == 0 && | | 1245 | (in6p->in6p_socket->so_options & SO_ACCEPTCONN) == 0 && |
1246 | ipsec6_in_reject_so(m, in6p->in6p_socket)) { | | 1246 | ipsec6_in_reject_so(m, in6p->in6p_socket)) { |
1247 | IPSEC_STATINC(IPSEC_STAT_IN_POLVIO); | | 1247 | IPSEC_STATINC(IPSEC_STAT_IN_POLVIO); |
1248 | goto drop; | | 1248 | goto drop; |
1249 | } | | 1249 | } |
1250 | #endif | | 1250 | #endif |
1251 | #endif /*IPSEC*/ | | 1251 | #endif /*IPSEC*/ |
1252 | break; | | 1252 | break; |
1253 | #endif /*INET*/ | | 1253 | #endif /*INET*/ |
1254 | #ifdef INET6 | | 1254 | #ifdef INET6 |
1255 | case AF_INET6: | | 1255 | case AF_INET6: |
1256 | { | | 1256 | { |
1257 | int faith; | | 1257 | int faith; |
1258 | | | 1258 | |
1259 | #if defined(NFAITH) && NFAITH > 0 | | 1259 | #if defined(NFAITH) && NFAITH > 0 |
1260 | faith = faithprefix(&ip6->ip6_dst); | | 1260 | faith = faithprefix(&ip6->ip6_dst); |
1261 | #else | | 1261 | #else |
1262 | faith = 0; | | 1262 | faith = 0; |
1263 | #endif | | 1263 | #endif |
1264 | in6p = in6_pcblookup_connect(&tcbtable, &ip6->ip6_src, | | 1264 | in6p = in6_pcblookup_connect(&tcbtable, &ip6->ip6_src, |
1265 | th->th_sport, &ip6->ip6_dst, th->th_dport, faith); | | 1265 | th->th_sport, &ip6->ip6_dst, th->th_dport, faith); |
1266 | if (in6p == NULL) { | | 1266 | if (in6p == NULL) { |
1267 | TCP_STATINC(TCP_STAT_PCBHASHMISS); | | 1267 | TCP_STATINC(TCP_STAT_PCBHASHMISS); |
1268 | in6p = in6_pcblookup_bind(&tcbtable, &ip6->ip6_dst, | | 1268 | in6p = in6_pcblookup_bind(&tcbtable, &ip6->ip6_dst, |
1269 | th->th_dport, faith); | | 1269 | th->th_dport, faith); |
1270 | } | | 1270 | } |
1271 | if (in6p == NULL) { | | 1271 | if (in6p == NULL) { |
1272 | TCP_STATINC(TCP_STAT_NOPORT); | | 1272 | TCP_STATINC(TCP_STAT_NOPORT); |
1273 | if (tcp_log_refused && | | 1273 | if (tcp_log_refused && |
1274 | (tiflags & (TH_RST|TH_ACK|TH_SYN)) == TH_SYN) { | | 1274 | (tiflags & (TH_RST|TH_ACK|TH_SYN)) == TH_SYN) { |
1275 | tcp6_log_refused(ip6, th); | | 1275 | tcp6_log_refused(ip6, th); |
1276 | } | | 1276 | } |
1277 | tcp_fields_to_host(th); | | 1277 | tcp_fields_to_host(th); |
1278 | goto dropwithreset_ratelim; | | 1278 | goto dropwithreset_ratelim; |
1279 | } | | 1279 | } |
1280 | #if defined(IPSEC) || defined(FAST_IPSEC) | | 1280 | #if defined(IPSEC) || defined(FAST_IPSEC) |
1281 | if ((in6p->in6p_socket->so_options & SO_ACCEPTCONN) == 0 && | | 1281 | if ((in6p->in6p_socket->so_options & SO_ACCEPTCONN) == 0 && |
1282 | ipsec6_in_reject(m, in6p)) { | | 1282 | ipsec6_in_reject(m, in6p)) { |
1283 | IPSEC6_STATINC(IPSEC_STAT_IN_POLVIO); | | 1283 | IPSEC6_STATINC(IPSEC_STAT_IN_POLVIO); |
1284 | goto drop; | | 1284 | goto drop; |
1285 | } | | 1285 | } |
1286 | #endif /*IPSEC*/ | | 1286 | #endif /*IPSEC*/ |
1287 | break; | | 1287 | break; |
1288 | } | | 1288 | } |
1289 | #endif | | 1289 | #endif |
1290 | } | | 1290 | } |
1291 | | | 1291 | |
1292 | /* | | 1292 | /* |
1293 | * If the state is CLOSED (i.e., TCB does not exist) then | | 1293 | * If the state is CLOSED (i.e., TCB does not exist) then |
1294 | * all data in the incoming segment is discarded. | | 1294 | * all data in the incoming segment is discarded. |
1295 | * If the TCB exists but is in CLOSED state, it is embryonic, | | 1295 | * If the TCB exists but is in CLOSED state, it is embryonic, |
1296 | * but should either do a listen or a connect soon. | | 1296 | * but should either do a listen or a connect soon. |
1297 | */ | | 1297 | */ |
1298 | tp = NULL; | | 1298 | tp = NULL; |
1299 | so = NULL; | | 1299 | so = NULL; |
1300 | if (inp) { | | 1300 | if (inp) { |
1301 | tp = intotcpcb(inp); | | 1301 | tp = intotcpcb(inp); |
1302 | so = inp->inp_socket; | | 1302 | so = inp->inp_socket; |
1303 | } | | 1303 | } |
1304 | #ifdef INET6 | | 1304 | #ifdef INET6 |
1305 | else if (in6p) { | | 1305 | else if (in6p) { |
1306 | tp = in6totcpcb(in6p); | | 1306 | tp = in6totcpcb(in6p); |
1307 | so = in6p->in6p_socket; | | 1307 | so = in6p->in6p_socket; |
1308 | } | | 1308 | } |
1309 | #endif | | 1309 | #endif |
1310 | if (tp == 0) { | | 1310 | if (tp == 0) { |
1311 | tcp_fields_to_host(th); | | 1311 | tcp_fields_to_host(th); |
1312 | goto dropwithreset_ratelim; | | 1312 | goto dropwithreset_ratelim; |
1313 | } | | 1313 | } |
1314 | if (tp->t_state == TCPS_CLOSED) | | 1314 | if (tp->t_state == TCPS_CLOSED) |
1315 | goto drop; | | 1315 | goto drop; |
1316 | | | 1316 | |
1317 | KASSERT(so->so_lock == softnet_lock); | | 1317 | KASSERT(so->so_lock == softnet_lock); |
1318 | KASSERT(solocked(so)); | | 1318 | KASSERT(solocked(so)); |
1319 | | | 1319 | |
1320 | /* | | 1320 | /* |
1321 | * Checksum extended TCP header and data. | | 1321 | * Checksum extended TCP header and data. |
1322 | */ | | 1322 | */ |
1323 | if (tcp_input_checksum(af, m, th, toff, off, tlen)) | | 1323 | if (tcp_input_checksum(af, m, th, toff, off, tlen)) |
1324 | goto badcsum; | | 1324 | goto badcsum; |
1325 | | | 1325 | |
1326 | tcp_fields_to_host(th); | | 1326 | tcp_fields_to_host(th); |
1327 | | | 1327 | |
1328 | /* Unscale the window into a 32-bit value. */ | | 1328 | /* Unscale the window into a 32-bit value. */ |
1329 | if ((tiflags & TH_SYN) == 0) | | 1329 | if ((tiflags & TH_SYN) == 0) |
1330 | tiwin = th->th_win << tp->snd_scale; | | 1330 | tiwin = th->th_win << tp->snd_scale; |
1331 | else | | 1331 | else |
1332 | tiwin = th->th_win; | | 1332 | tiwin = th->th_win; |
1333 | | | 1333 | |
1334 | #ifdef INET6 | | 1334 | #ifdef INET6 |
1335 | /* save packet options if user wanted */ | | 1335 | /* save packet options if user wanted */ |
1336 | if (in6p && (in6p->in6p_flags & IN6P_CONTROLOPTS)) { | | 1336 | if (in6p && (in6p->in6p_flags & IN6P_CONTROLOPTS)) { |
1337 | if (in6p->in6p_options) { | | 1337 | if (in6p->in6p_options) { |
1338 | m_freem(in6p->in6p_options); | | 1338 | m_freem(in6p->in6p_options); |
1339 | in6p->in6p_options = 0; | | 1339 | in6p->in6p_options = 0; |
1340 | } | | 1340 | } |
1341 | KASSERT(ip6 != NULL); | | 1341 | KASSERT(ip6 != NULL); |
1342 | ip6_savecontrol(in6p, &in6p->in6p_options, ip6, m); | | 1342 | ip6_savecontrol(in6p, &in6p->in6p_options, ip6, m); |
1343 | } | | 1343 | } |
1344 | #endif | | 1344 | #endif |
1345 | | | 1345 | |
1346 | if (so->so_options & (SO_DEBUG|SO_ACCEPTCONN)) { | | 1346 | if (so->so_options & (SO_DEBUG|SO_ACCEPTCONN)) { |
1347 | union syn_cache_sa src; | | 1347 | union syn_cache_sa src; |
1348 | union syn_cache_sa dst; | | 1348 | union syn_cache_sa dst; |
1349 | | | 1349 | |
1350 | memset(&src, 0, sizeof(src)); | | 1350 | memset(&src, 0, sizeof(src)); |
1351 | memset(&dst, 0, sizeof(dst)); | | 1351 | memset(&dst, 0, sizeof(dst)); |
1352 | switch (af) { | | 1352 | switch (af) { |
1353 | #ifdef INET | | 1353 | #ifdef INET |
1354 | case AF_INET: | | 1354 | case AF_INET: |
1355 | src.sin.sin_len = sizeof(struct sockaddr_in); | | 1355 | src.sin.sin_len = sizeof(struct sockaddr_in); |
1356 | src.sin.sin_family = AF_INET; | | 1356 | src.sin.sin_family = AF_INET; |
1357 | src.sin.sin_addr = ip->ip_src; | | 1357 | src.sin.sin_addr = ip->ip_src; |
1358 | src.sin.sin_port = th->th_sport; | | 1358 | src.sin.sin_port = th->th_sport; |
1359 | | | 1359 | |
1360 | dst.sin.sin_len = sizeof(struct sockaddr_in); | | 1360 | dst.sin.sin_len = sizeof(struct sockaddr_in); |
1361 | dst.sin.sin_family = AF_INET; | | 1361 | dst.sin.sin_family = AF_INET; |
1362 | dst.sin.sin_addr = ip->ip_dst; | | 1362 | dst.sin.sin_addr = ip->ip_dst; |
1363 | dst.sin.sin_port = th->th_dport; | | 1363 | dst.sin.sin_port = th->th_dport; |
1364 | break; | | 1364 | break; |
1365 | #endif | | 1365 | #endif |
1366 | #ifdef INET6 | | 1366 | #ifdef INET6 |
1367 | case AF_INET6: | | 1367 | case AF_INET6: |
1368 | src.sin6.sin6_len = sizeof(struct sockaddr_in6); | | 1368 | src.sin6.sin6_len = sizeof(struct sockaddr_in6); |
1369 | src.sin6.sin6_family = AF_INET6; | | 1369 | src.sin6.sin6_family = AF_INET6; |
1370 | src.sin6.sin6_addr = ip6->ip6_src; | | 1370 | src.sin6.sin6_addr = ip6->ip6_src; |
1371 | src.sin6.sin6_port = th->th_sport; | | 1371 | src.sin6.sin6_port = th->th_sport; |
1372 | | | 1372 | |
1373 | dst.sin6.sin6_len = sizeof(struct sockaddr_in6); | | 1373 | dst.sin6.sin6_len = sizeof(struct sockaddr_in6); |
1374 | dst.sin6.sin6_family = AF_INET6; | | 1374 | dst.sin6.sin6_family = AF_INET6; |
1375 | dst.sin6.sin6_addr = ip6->ip6_dst; | | 1375 | dst.sin6.sin6_addr = ip6->ip6_dst; |
1376 | dst.sin6.sin6_port = th->th_dport; | | 1376 | dst.sin6.sin6_port = th->th_dport; |
1377 | break; | | 1377 | break; |
1378 | #endif /* INET6 */ | | 1378 | #endif /* INET6 */ |
1379 | default: | | 1379 | default: |
1380 | goto badsyn; /*sanity*/ | | 1380 | goto badsyn; /*sanity*/ |
1381 | } | | 1381 | } |
1382 | | | 1382 | |
1383 | if (so->so_options & SO_DEBUG) { | | 1383 | if (so->so_options & SO_DEBUG) { |
1384 | #ifdef TCP_DEBUG | | 1384 | #ifdef TCP_DEBUG |
1385 | ostate = tp->t_state; | | 1385 | ostate = tp->t_state; |
1386 | #endif | | 1386 | #endif |
1387 | | | 1387 | |
1388 | tcp_saveti = NULL; | | 1388 | tcp_saveti = NULL; |
1389 | if (iphlen + sizeof(struct tcphdr) > MHLEN) | | 1389 | if (iphlen + sizeof(struct tcphdr) > MHLEN) |
1390 | goto nosave; | | 1390 | goto nosave; |
1391 | | | 1391 | |
1392 | if (m->m_len > iphlen && (m->m_flags & M_EXT) == 0) { | | 1392 | if (m->m_len > iphlen && (m->m_flags & M_EXT) == 0) { |
1393 | tcp_saveti = m_copym(m, 0, iphlen, M_DONTWAIT); | | 1393 | tcp_saveti = m_copym(m, 0, iphlen, M_DONTWAIT); |
1394 | if (!tcp_saveti) | | 1394 | if (!tcp_saveti) |
1395 | goto nosave; | | 1395 | goto nosave; |
1396 | } else { | | 1396 | } else { |
1397 | MGETHDR(tcp_saveti, M_DONTWAIT, MT_HEADER); | | 1397 | MGETHDR(tcp_saveti, M_DONTWAIT, MT_HEADER); |
1398 | if (!tcp_saveti) | | 1398 | if (!tcp_saveti) |
1399 | goto nosave; | | 1399 | goto nosave; |
1400 | MCLAIM(m, &tcp_mowner); | | 1400 | MCLAIM(m, &tcp_mowner); |
1401 | tcp_saveti->m_len = iphlen; | | 1401 | tcp_saveti->m_len = iphlen; |
1402 | m_copydata(m, 0, iphlen, | | 1402 | m_copydata(m, 0, iphlen, |
1403 | mtod(tcp_saveti, void *)); | | 1403 | mtod(tcp_saveti, void *)); |
1404 | } | | 1404 | } |
1405 | | | 1405 | |
1406 | if (M_TRAILINGSPACE(tcp_saveti) < sizeof(struct tcphdr)) { | | 1406 | if (M_TRAILINGSPACE(tcp_saveti) < sizeof(struct tcphdr)) { |
1407 | m_freem(tcp_saveti); | | 1407 | m_freem(tcp_saveti); |
1408 | tcp_saveti = NULL; | | 1408 | tcp_saveti = NULL; |
1409 | } else { | | 1409 | } else { |
1410 | tcp_saveti->m_len += sizeof(struct tcphdr); | | 1410 | tcp_saveti->m_len += sizeof(struct tcphdr); |
1411 | memcpy(mtod(tcp_saveti, char *) + iphlen, th, | | 1411 | memcpy(mtod(tcp_saveti, char *) + iphlen, th, |
1412 | sizeof(struct tcphdr)); | | 1412 | sizeof(struct tcphdr)); |
1413 | } | | 1413 | } |
1414 | nosave:; | | 1414 | nosave:; |
1415 | } | | 1415 | } |
1416 | if (so->so_options & SO_ACCEPTCONN) { | | 1416 | if (so->so_options & SO_ACCEPTCONN) { |
1417 | if ((tiflags & (TH_RST|TH_ACK|TH_SYN)) != TH_SYN) { | | 1417 | if ((tiflags & (TH_RST|TH_ACK|TH_SYN)) != TH_SYN) { |
1418 | if (tiflags & TH_RST) { | | 1418 | if (tiflags & TH_RST) { |
1419 | syn_cache_reset(&src.sa, &dst.sa, th); | | 1419 | syn_cache_reset(&src.sa, &dst.sa, th); |
1420 | } else if ((tiflags & (TH_ACK|TH_SYN)) == | | 1420 | } else if ((tiflags & (TH_ACK|TH_SYN)) == |
1421 | (TH_ACK|TH_SYN)) { | | 1421 | (TH_ACK|TH_SYN)) { |
1422 | /* | | 1422 | /* |
1423 | * Received a SYN,ACK. This should | | 1423 | * Received a SYN,ACK. This should |
1424 | * never happen while we are in | | 1424 | * never happen while we are in |
1425 | * LISTEN. Send an RST. | | 1425 | * LISTEN. Send an RST. |
1426 | */ | | 1426 | */ |
1427 | goto badsyn; | | 1427 | goto badsyn; |
1428 | } else if (tiflags & TH_ACK) { | | 1428 | } else if (tiflags & TH_ACK) { |
1429 | so = syn_cache_get(&src.sa, &dst.sa, | | 1429 | so = syn_cache_get(&src.sa, &dst.sa, |
1430 | th, toff, tlen, so, m); | | 1430 | th, toff, tlen, so, m); |
1431 | if (so == NULL) { | | 1431 | if (so == NULL) { |
1432 | /* | | 1432 | /* |
1433 | * We don't have a SYN for | | 1433 | * We don't have a SYN for |
1434 | * this ACK; send an RST. | | 1434 | * this ACK; send an RST. |
1435 | */ | | 1435 | */ |
1436 | goto badsyn; | | 1436 | goto badsyn; |
1437 | } else if (so == | | 1437 | } else if (so == |
1438 | (struct socket *)(-1)) { | | 1438 | (struct socket *)(-1)) { |
1439 | /* | | 1439 | /* |
1440 | * We were unable to create | | 1440 | * We were unable to create |
1441 | * the connection. If the | | 1441 | * the connection. If the |
1442 | * 3-way handshake was | | 1442 | * 3-way handshake was |
1443 | * completed, and RST has | | 1443 | * completed, and RST has |
1444 | * been sent to the peer. | | 1444 | * been sent to the peer. |
1445 | * Since the mbuf might be | | 1445 | * Since the mbuf might be |
1446 | * in use for the reply, | | 1446 | * in use for the reply, |
1447 | * do not free it. | | 1447 | * do not free it. |
1448 | */ | | 1448 | */ |
1449 | m = NULL; | | 1449 | m = NULL; |
1450 | } else { | | 1450 | } else { |
1451 | /* | | 1451 | /* |
1452 | * We have created a | | 1452 | * We have created a |
1453 | * full-blown connection. | | 1453 | * full-blown connection. |
1454 | */ | | 1454 | */ |
1455 | tp = NULL; | | 1455 | tp = NULL; |
1456 | inp = NULL; | | 1456 | inp = NULL; |
1457 | #ifdef INET6 | | 1457 | #ifdef INET6 |
1458 | in6p = NULL; | | 1458 | in6p = NULL; |
1459 | #endif | | 1459 | #endif |
1460 | switch (so->so_proto->pr_domain->dom_family) { | | 1460 | switch (so->so_proto->pr_domain->dom_family) { |
1461 | #ifdef INET | | 1461 | #ifdef INET |
1462 | case AF_INET: | | 1462 | case AF_INET: |
1463 | inp = sotoinpcb(so); | | 1463 | inp = sotoinpcb(so); |
1464 | tp = intotcpcb(inp); | | 1464 | tp = intotcpcb(inp); |
1465 | break; | | 1465 | break; |
1466 | #endif | | 1466 | #endif |
1467 | #ifdef INET6 | | 1467 | #ifdef INET6 |
1468 | case AF_INET6: | | 1468 | case AF_INET6: |
1469 | in6p = sotoin6pcb(so); | | 1469 | in6p = sotoin6pcb(so); |
1470 | tp = in6totcpcb(in6p); | | 1470 | tp = in6totcpcb(in6p); |
1471 | break; | | 1471 | break; |
1472 | #endif | | 1472 | #endif |
1473 | } | | 1473 | } |
1474 | if (tp == NULL) | | 1474 | if (tp == NULL) |
1475 | goto badsyn; /*XXX*/ | | 1475 | goto badsyn; /*XXX*/ |
1476 | tiwin <<= tp->snd_scale; | | 1476 | tiwin <<= tp->snd_scale; |
1477 | goto after_listen; | | 1477 | goto after_listen; |
1478 | } | | 1478 | } |
1479 | } else { | | 1479 | } else { |
1480 | /* | | 1480 | /* |
1481 | * None of RST, SYN or ACK was set. | | 1481 | * None of RST, SYN or ACK was set. |
1482 | * This is an invalid packet for a | | 1482 | * This is an invalid packet for a |
1483 | * TCB in LISTEN state. Send a RST. | | 1483 | * TCB in LISTEN state. Send a RST. |
1484 | */ | | 1484 | */ |
1485 | goto badsyn; | | 1485 | goto badsyn; |
1486 | } | | 1486 | } |
1487 | } else { | | 1487 | } else { |
1488 | /* | | 1488 | /* |
1489 | * Received a SYN. | | 1489 | * Received a SYN. |
1490 | * | | 1490 | * |
1491 | * RFC1122 4.2.3.10, p. 104: discard bcast/mcast SYN | | 1491 | * RFC1122 4.2.3.10, p. 104: discard bcast/mcast SYN |
1492 | */ | | 1492 | */ |
1493 | if (m->m_flags & (M_BCAST|M_MCAST)) | | 1493 | if (m->m_flags & (M_BCAST|M_MCAST)) |
1494 | goto drop; | | 1494 | goto drop; |
1495 | | | 1495 | |
1496 | switch (af) { | | 1496 | switch (af) { |
1497 | #ifdef INET6 | | 1497 | #ifdef INET6 |
1498 | case AF_INET6: | | 1498 | case AF_INET6: |
1499 | if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) | | 1499 | if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) |
1500 | goto drop; | | 1500 | goto drop; |
1501 | break; | | 1501 | break; |
1502 | #endif /* INET6 */ | | 1502 | #endif /* INET6 */ |
1503 | case AF_INET: | | 1503 | case AF_INET: |
1504 | if (IN_MULTICAST(ip->ip_dst.s_addr) || | | 1504 | if (IN_MULTICAST(ip->ip_dst.s_addr) || |
1505 | in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif)) | | 1505 | in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif)) |
1506 | goto drop; | | 1506 | goto drop; |
1507 | break; | | 1507 | break; |
1508 | } | | 1508 | } |
1509 | | | 1509 | |
1510 | #ifdef INET6 | | 1510 | #ifdef INET6 |
1511 | /* | | 1511 | /* |
1512 | * If deprecated address is forbidden, we do | | 1512 | * If deprecated address is forbidden, we do |
1513 | * not accept SYN to deprecated interface | | 1513 | * not accept SYN to deprecated interface |
1514 | * address to prevent any new inbound | | 1514 | * address to prevent any new inbound |
1515 | * connection from getting established. | | 1515 | * connection from getting established. |
1516 | * When we do not accept SYN, we send a TCP | | 1516 | * When we do not accept SYN, we send a TCP |
1517 | * RST, with deprecated source address (instead | | 1517 | * RST, with deprecated source address (instead |
1518 | * of dropping it). We compromise it as it is | | 1518 | * of dropping it). We compromise it as it is |
1519 | * much better for peer to send a RST, and | | 1519 | * much better for peer to send a RST, and |
1520 | * RST will be the final packet for the | | 1520 | * RST will be the final packet for the |
1521 | * exchange. | | 1521 | * exchange. |
1522 | * | | 1522 | * |
1523 | * If we do not forbid deprecated addresses, we | | 1523 | * If we do not forbid deprecated addresses, we |
1524 | * accept the SYN packet. RFC2462 does not | | 1524 | * accept the SYN packet. RFC2462 does not |
1525 | * suggest dropping SYN in this case. | | 1525 | * suggest dropping SYN in this case. |
1526 | * If we decipher RFC2462 5.5.4, it says like | | 1526 | * If we decipher RFC2462 5.5.4, it says like |
1527 | * this: | | 1527 | * this: |
1528 | * 1. use of deprecated addr with existing | | 1528 | * 1. use of deprecated addr with existing |
1529 | * communication is okay - "SHOULD continue | | 1529 | * communication is okay - "SHOULD continue |
1530 | * to be used" | | 1530 | * to be used" |
1531 | * 2. use of it with new communication: | | 1531 | * 2. use of it with new communication: |
1532 | * (2a) "SHOULD NOT be used if alternate | | 1532 | * (2a) "SHOULD NOT be used if alternate |
1533 | * address with sufficient scope is | | 1533 | * address with sufficient scope is |
1534 | * available" | | 1534 | * available" |
1535 | * (2b) nothing mentioned otherwise. | | 1535 | * (2b) nothing mentioned otherwise. |
1536 | * Here we fall into (2b) case as we have no | | 1536 | * Here we fall into (2b) case as we have no |
1537 | * choice in our source address selection - we | | 1537 | * choice in our source address selection - we |
1538 | * must obey the peer. | | 1538 | * must obey the peer. |
1539 | * | | 1539 | * |
1540 | * The wording in RFC2462 is confusing, and | | 1540 | * The wording in RFC2462 is confusing, and |
1541 | * there are multiple description text for | | 1541 | * there are multiple description text for |
1542 | * deprecated address handling - worse, they | | 1542 | * deprecated address handling - worse, they |
1543 | * are not exactly the same. I believe 5.5.4 | | 1543 | * are not exactly the same. I believe 5.5.4 |
1544 | * is the best one, so we follow 5.5.4. | | 1544 | * is the best one, so we follow 5.5.4. |
1545 | */ | | 1545 | */ |
1546 | if (af == AF_INET6 && !ip6_use_deprecated) { | | 1546 | if (af == AF_INET6 && !ip6_use_deprecated) { |
1547 | struct in6_ifaddr *ia6; | | 1547 | struct in6_ifaddr *ia6; |
1548 | if ((ia6 = in6ifa_ifpwithaddr(m->m_pkthdr.rcvif, | | 1548 | if ((ia6 = in6ifa_ifpwithaddr(m->m_pkthdr.rcvif, |
1549 | &ip6->ip6_dst)) && | | 1549 | &ip6->ip6_dst)) && |
1550 | (ia6->ia6_flags & IN6_IFF_DEPRECATED)) { | | 1550 | (ia6->ia6_flags & IN6_IFF_DEPRECATED)) { |
1551 | tp = NULL; | | 1551 | tp = NULL; |
1552 | goto dropwithreset; | | 1552 | goto dropwithreset; |
1553 | } | | 1553 | } |
1554 | } | | 1554 | } |
1555 | #endif | | 1555 | #endif |
1556 | | | 1556 | |
1557 | #if defined(IPSEC) || defined(FAST_IPSEC) | | 1557 | #if defined(IPSEC) || defined(FAST_IPSEC) |
1558 | switch (af) { | | 1558 | switch (af) { |
1559 | #ifdef INET | | 1559 | #ifdef INET |
1560 | case AF_INET: | | 1560 | case AF_INET: |
1561 | if (ipsec4_in_reject_so(m, so)) { | | 1561 | if (ipsec4_in_reject_so(m, so)) { |
1562 | IPSEC_STATINC(IPSEC_STAT_IN_POLVIO); | | 1562 | IPSEC_STATINC(IPSEC_STAT_IN_POLVIO); |
1563 | tp = NULL; | | 1563 | tp = NULL; |
1564 | goto dropwithreset; | | 1564 | goto dropwithreset; |
1565 | } | | 1565 | } |
1566 | break; | | 1566 | break; |
1567 | #endif | | 1567 | #endif |
1568 | #ifdef INET6 | | 1568 | #ifdef INET6 |
1569 | case AF_INET6: | | 1569 | case AF_INET6: |
1570 | if (ipsec6_in_reject_so(m, so)) { | | 1570 | if (ipsec6_in_reject_so(m, so)) { |
1571 | IPSEC6_STATINC(IPSEC_STAT_IN_POLVIO); | | 1571 | IPSEC6_STATINC(IPSEC_STAT_IN_POLVIO); |
1572 | tp = NULL; | | 1572 | tp = NULL; |
1573 | goto dropwithreset; | | 1573 | goto dropwithreset; |
1574 | } | | 1574 | } |
1575 | break; | | 1575 | break; |
1576 | #endif /*INET6*/ | | 1576 | #endif /*INET6*/ |
1577 | } | | 1577 | } |
1578 | #endif /*IPSEC*/ | | 1578 | #endif /*IPSEC*/ |
1579 | | | 1579 | |
1580 | /* | | 1580 | /* |
1581 | * LISTEN socket received a SYN | | 1581 | * LISTEN socket received a SYN |
1582 | * from itself? This can't possibly | | 1582 | * from itself? This can't possibly |
1583 | * be valid; drop the packet. | | 1583 | * be valid; drop the packet. |
1584 | */ | | 1584 | */ |
1585 | if (th->th_sport == th->th_dport) { | | 1585 | if (th->th_sport == th->th_dport) { |
1586 | int i; | | 1586 | int i; |
1587 | | | 1587 | |
1588 | switch (af) { | | 1588 | switch (af) { |
1589 | #ifdef INET | | 1589 | #ifdef INET |
1590 | case AF_INET: | | 1590 | case AF_INET: |
1591 | i = in_hosteq(ip->ip_src, ip->ip_dst); | | 1591 | i = in_hosteq(ip->ip_src, ip->ip_dst); |
1592 | break; | | 1592 | break; |
1593 | #endif | | 1593 | #endif |
1594 | #ifdef INET6 | | 1594 | #ifdef INET6 |
1595 | case AF_INET6: | | 1595 | case AF_INET6: |
1596 | i = IN6_ARE_ADDR_EQUAL(&ip6->ip6_src, &ip6->ip6_dst); | | 1596 | i = IN6_ARE_ADDR_EQUAL(&ip6->ip6_src, &ip6->ip6_dst); |
1597 | break; | | 1597 | break; |
1598 | #endif | | 1598 | #endif |
1599 | default: | | 1599 | default: |
1600 | i = 1; | | 1600 | i = 1; |
1601 | } | | 1601 | } |
1602 | if (i) { | | 1602 | if (i) { |
1603 | TCP_STATINC(TCP_STAT_BADSYN); | | 1603 | TCP_STATINC(TCP_STAT_BADSYN); |
1604 | goto drop; | | 1604 | goto drop; |
1605 | } | | 1605 | } |
1606 | } | | 1606 | } |
1607 | | | 1607 | |
1608 | /* | | 1608 | /* |
1609 | * SYN looks ok; create compressed TCP | | 1609 | * SYN looks ok; create compressed TCP |
1610 | * state for it. | | 1610 | * state for it. |
1611 | */ | | 1611 | */ |
1612 | if (so->so_qlen <= so->so_qlimit && | | 1612 | if (so->so_qlen <= so->so_qlimit && |
1613 | syn_cache_add(&src.sa, &dst.sa, th, tlen, | | 1613 | syn_cache_add(&src.sa, &dst.sa, th, tlen, |
1614 | so, m, optp, optlen, &opti)) | | 1614 | so, m, optp, optlen, &opti)) |
1615 | m = NULL; | | 1615 | m = NULL; |
1616 | } | | 1616 | } |
1617 | goto drop; | | 1617 | goto drop; |
1618 | } | | 1618 | } |
1619 | } | | 1619 | } |
1620 | | | 1620 | |
1621 | after_listen: | | 1621 | after_listen: |
1622 | #ifdef DIAGNOSTIC | | 1622 | #ifdef DIAGNOSTIC |
1623 | /* | | 1623 | /* |
1624 | * Should not happen now that all embryonic connections | | 1624 | * Should not happen now that all embryonic connections |
1625 | * are handled with compressed state. | | 1625 | * are handled with compressed state. |
1626 | */ | | 1626 | */ |
1627 | if (tp->t_state == TCPS_LISTEN) | | 1627 | if (tp->t_state == TCPS_LISTEN) |
1628 | panic("tcp_input: TCPS_LISTEN"); | | 1628 | panic("tcp_input: TCPS_LISTEN"); |
1629 | #endif | | 1629 | #endif |
1630 | | | 1630 | |
1631 | /* | | 1631 | /* |
1632 | * Segment received on connection. | | 1632 | * Segment received on connection. |
1633 | * Reset idle time and keep-alive timer. | | 1633 | * Reset idle time and keep-alive timer. |
1634 | */ | | 1634 | */ |
1635 | tp->t_rcvtime = tcp_now; | | 1635 | tp->t_rcvtime = tcp_now; |
1636 | if (TCPS_HAVEESTABLISHED(tp->t_state)) | | 1636 | if (TCPS_HAVEESTABLISHED(tp->t_state)) |
1637 | TCP_TIMER_ARM(tp, TCPT_KEEP, tp->t_keepidle); | | 1637 | TCP_TIMER_ARM(tp, TCPT_KEEP, tp->t_keepidle); |
1638 | | | 1638 | |
1639 | /* | | 1639 | /* |
1640 | * Process options. | | 1640 | * Process options. |
1641 | */ | | 1641 | */ |
1642 | #ifdef TCP_SIGNATURE | | 1642 | #ifdef TCP_SIGNATURE |
1643 | if (optp || (tp->t_flags & TF_SIGNATURE)) | | 1643 | if (optp || (tp->t_flags & TF_SIGNATURE)) |
1644 | #else | | 1644 | #else |
1645 | if (optp) | | 1645 | if (optp) |
1646 | #endif | | 1646 | #endif |
1647 | if (tcp_dooptions(tp, optp, optlen, th, m, toff, &opti) < 0) | | 1647 | if (tcp_dooptions(tp, optp, optlen, th, m, toff, &opti) < 0) |
1648 | goto drop; | | 1648 | goto drop; |
1649 | | | 1649 | |
1650 | if (TCP_SACK_ENABLED(tp)) { | | 1650 | if (TCP_SACK_ENABLED(tp)) { |
1651 | tcp_del_sackholes(tp, th); | | 1651 | tcp_del_sackholes(tp, th); |
1652 | } | | 1652 | } |
1653 | | | 1653 | |
1654 | if (TCP_ECN_ALLOWED(tp)) { | | 1654 | if (TCP_ECN_ALLOWED(tp)) { |
1655 | switch (iptos & IPTOS_ECN_MASK) { | | 1655 | switch (iptos & IPTOS_ECN_MASK) { |
1656 | case IPTOS_ECN_CE: | | 1656 | case IPTOS_ECN_CE: |
1657 | tp->t_flags |= TF_ECN_SND_ECE; | | 1657 | tp->t_flags |= TF_ECN_SND_ECE; |
1658 | TCP_STATINC(TCP_STAT_ECN_CE); | | 1658 | TCP_STATINC(TCP_STAT_ECN_CE); |
1659 | break; | | 1659 | break; |
1660 | case IPTOS_ECN_ECT0: | | 1660 | case IPTOS_ECN_ECT0: |
1661 | TCP_STATINC(TCP_STAT_ECN_ECT); | | 1661 | TCP_STATINC(TCP_STAT_ECN_ECT); |
1662 | break; | | 1662 | break; |
1663 | case IPTOS_ECN_ECT1: | | 1663 | case IPTOS_ECN_ECT1: |
1664 | /* XXX: ignore for now -- rpaulo */ | | 1664 | /* XXX: ignore for now -- rpaulo */ |
1665 | break; | | 1665 | break; |
1666 | } | | 1666 | } |
1667 | | | 1667 | |
1668 | if (tiflags & TH_CWR) | | 1668 | if (tiflags & TH_CWR) |
1669 | tp->t_flags &= ~TF_ECN_SND_ECE; | | 1669 | tp->t_flags &= ~TF_ECN_SND_ECE; |
1670 | | | 1670 | |
1671 | /* | | 1671 | /* |
1672 | * Congestion experienced. | | 1672 | * Congestion experienced. |
1673 | * Ignore if we are already trying to recover. | | 1673 | * Ignore if we are already trying to recover. |
1674 | */ | | 1674 | */ |
1675 | if ((tiflags & TH_ECE) && SEQ_GEQ(tp->snd_una, tp->snd_recover)) | | 1675 | if ((tiflags & TH_ECE) && SEQ_GEQ(tp->snd_una, tp->snd_recover)) |
1676 | tp->t_congctl->cong_exp(tp); | | 1676 | tp->t_congctl->cong_exp(tp); |
1677 | } | | 1677 | } |
1678 | | | 1678 | |
1679 | if (opti.ts_present && opti.ts_ecr) { | | 1679 | if (opti.ts_present && opti.ts_ecr) { |
1680 | /* | | 1680 | /* |
1681 | * Calculate the RTT from the returned time stamp and the | | 1681 | * Calculate the RTT from the returned time stamp and the |
1682 | * connection's time base. If the time stamp is later than | | 1682 | * connection's time base. If the time stamp is later than |
1683 | * the current time, or is extremely old, fall back to non-1323 | | 1683 | * the current time, or is extremely old, fall back to non-1323 |
1684 | * RTT calculation. Since ts_ecr is unsigned, we can test both | | 1684 | * RTT calculation. Since ts_ecr is unsigned, we can test both |
1685 | * at the same time. | | 1685 | * at the same time. |
1686 | */ | | 1686 | */ |
1687 | ts_rtt = TCP_TIMESTAMP(tp) - opti.ts_ecr + 1; | | 1687 | ts_rtt = TCP_TIMESTAMP(tp) - opti.ts_ecr + 1; |
1688 | if (ts_rtt > TCP_PAWS_IDLE) | | 1688 | if (ts_rtt > TCP_PAWS_IDLE) |
1689 | ts_rtt = 0; | | 1689 | ts_rtt = 0; |
1690 | } else { | | 1690 | } else { |
1691 | ts_rtt = 0; | | 1691 | ts_rtt = 0; |
1692 | } | | 1692 | } |
1693 | | | 1693 | |
1694 | /* | | 1694 | /* |
1695 | * Header prediction: check for the two common cases | | 1695 | * Header prediction: check for the two common cases |
1696 | * of a uni-directional data xfer. If the packet has | | 1696 | * of a uni-directional data xfer. If the packet has |
1697 | * no control flags, is in-sequence, the window didn't | | 1697 | * no control flags, is in-sequence, the window didn't |
1698 | * change and we're not retransmitting, it's a | | 1698 | * change and we're not retransmitting, it's a |
1699 | * candidate. If the length is zero and the ack moved | | 1699 | * candidate. If the length is zero and the ack moved |
1700 | * forward, we're the sender side of the xfer. Just | | 1700 | * forward, we're the sender side of the xfer. Just |
1701 | * free the data acked & wake any higher level process | | 1701 | * free the data acked & wake any higher level process |
1702 | * that was blocked waiting for space. If the length | | 1702 | * that was blocked waiting for space. If the length |
1703 | * is non-zero and the ack didn't move, we're the | | 1703 | * is non-zero and the ack didn't move, we're the |
1704 | * receiver side. If we're getting packets in-order | | 1704 | * receiver side. If we're getting packets in-order |
1705 | * (the reassembly queue is empty), add the data to | | 1705 | * (the reassembly queue is empty), add the data to |
1706 | * the socket buffer and note that we need a delayed ack. | | 1706 | * the socket buffer and note that we need a delayed ack. |
1707 | */ | | 1707 | */ |
1708 | if (tp->t_state == TCPS_ESTABLISHED && | | 1708 | if (tp->t_state == TCPS_ESTABLISHED && |
1709 | (tiflags & (TH_SYN|TH_FIN|TH_RST|TH_URG|TH_ECE|TH_CWR|TH_ACK)) | | 1709 | (tiflags & (TH_SYN|TH_FIN|TH_RST|TH_URG|TH_ECE|TH_CWR|TH_ACK)) |
1710 | == TH_ACK && | | 1710 | == TH_ACK && |
1711 | (!opti.ts_present || TSTMP_GEQ(opti.ts_val, tp->ts_recent)) && | | 1711 | (!opti.ts_present || TSTMP_GEQ(opti.ts_val, tp->ts_recent)) && |
1712 | th->th_seq == tp->rcv_nxt && | | 1712 | th->th_seq == tp->rcv_nxt && |
1713 | tiwin && tiwin == tp->snd_wnd && | | 1713 | tiwin && tiwin == tp->snd_wnd && |
1714 | tp->snd_nxt == tp->snd_max) { | | 1714 | tp->snd_nxt == tp->snd_max) { |
1715 | | | 1715 | |
1716 | /* | | 1716 | /* |
1717 | * If last ACK falls within this segment's sequence numbers, | | 1717 | * If last ACK falls within this segment's sequence numbers, |
1718 | * record the timestamp. | | 1718 | * record the timestamp. |
1719 | * NOTE that the test is modified according to the latest | | 1719 | * NOTE that the test is modified according to the latest |
1720 | * proposal of the tcplw@cray.com list (Braden 1993/04/26). | | 1720 | * proposal of the tcplw@cray.com list (Braden 1993/04/26). |
1721 | * | | 1721 | * |
1722 | * note that we already know | | 1722 | * note that we already know |
1723 | * TSTMP_GEQ(opti.ts_val, tp->ts_recent) | | 1723 | * TSTMP_GEQ(opti.ts_val, tp->ts_recent) |
1724 | */ | | 1724 | */ |
1725 | if (opti.ts_present && | | 1725 | if (opti.ts_present && |
1726 | SEQ_LEQ(th->th_seq, tp->last_ack_sent)) { | | 1726 | SEQ_LEQ(th->th_seq, tp->last_ack_sent)) { |
1727 | tp->ts_recent_age = tcp_now; | | 1727 | tp->ts_recent_age = tcp_now; |
1728 | tp->ts_recent = opti.ts_val; | | 1728 | tp->ts_recent = opti.ts_val; |
1729 | } | | 1729 | } |
1730 | | | 1730 | |
1731 | if (tlen == 0) { | | 1731 | if (tlen == 0) { |
1732 | /* Ack prediction. */ | | 1732 | /* Ack prediction. */ |
1733 | if (SEQ_GT(th->th_ack, tp->snd_una) && | | 1733 | if (SEQ_GT(th->th_ack, tp->snd_una) && |
1734 | SEQ_LEQ(th->th_ack, tp->snd_max) && | | 1734 | SEQ_LEQ(th->th_ack, tp->snd_max) && |
1735 | tp->snd_cwnd >= tp->snd_wnd && | | 1735 | tp->snd_cwnd >= tp->snd_wnd && |
1736 | tp->t_partialacks < 0) { | | 1736 | tp->t_partialacks < 0) { |
1737 | /* | | 1737 | /* |
1738 | * this is a pure ack for outstanding data. | | 1738 | * this is a pure ack for outstanding data. |
1739 | */ | | 1739 | */ |
1740 | if (ts_rtt) | | 1740 | if (ts_rtt) |
1741 | tcp_xmit_timer(tp, ts_rtt); | | 1741 | tcp_xmit_timer(tp, ts_rtt); |
1742 | else if (tp->t_rtttime && | | 1742 | else if (tp->t_rtttime && |
1743 | SEQ_GT(th->th_ack, tp->t_rtseq)) | | 1743 | SEQ_GT(th->th_ack, tp->t_rtseq)) |
1744 | tcp_xmit_timer(tp, | | 1744 | tcp_xmit_timer(tp, |
1745 | tcp_now - tp->t_rtttime); | | 1745 | tcp_now - tp->t_rtttime); |
1746 | acked = th->th_ack - tp->snd_una; | | 1746 | acked = th->th_ack - tp->snd_una; |
1747 | tcps = TCP_STAT_GETREF(); | | 1747 | tcps = TCP_STAT_GETREF(); |
1748 | tcps[TCP_STAT_PREDACK]++; | | 1748 | tcps[TCP_STAT_PREDACK]++; |
1749 | tcps[TCP_STAT_RCVACKPACK]++; | | 1749 | tcps[TCP_STAT_RCVACKPACK]++; |
1750 | tcps[TCP_STAT_RCVACKBYTE] += acked; | | 1750 | tcps[TCP_STAT_RCVACKBYTE] += acked; |
1751 | TCP_STAT_PUTREF(); | | 1751 | TCP_STAT_PUTREF(); |
1752 | nd6_hint(tp); | | 1752 | nd6_hint(tp); |
1753 | | | 1753 | |
1754 | if (acked > (tp->t_lastoff - tp->t_inoff)) | | 1754 | if (acked > (tp->t_lastoff - tp->t_inoff)) |
1755 | tp->t_lastm = NULL; | | 1755 | tp->t_lastm = NULL; |
1756 | sbdrop(&so->so_snd, acked); | | 1756 | sbdrop(&so->so_snd, acked); |
1757 | tp->t_lastoff -= acked; | | 1757 | tp->t_lastoff -= acked; |
1758 | | | 1758 | |
1759 | icmp_check(tp, th, acked); | | 1759 | icmp_check(tp, th, acked); |
1760 | | | 1760 | |
1761 | tp->snd_una = th->th_ack; | | 1761 | tp->snd_una = th->th_ack; |
1762 | tp->snd_fack = tp->snd_una; | | 1762 | tp->snd_fack = tp->snd_una; |
1763 | if (SEQ_LT(tp->snd_high, tp->snd_una)) | | 1763 | if (SEQ_LT(tp->snd_high, tp->snd_una)) |
1764 | tp->snd_high = tp->snd_una; | | 1764 | tp->snd_high = tp->snd_una; |
1765 | m_freem(m); | | 1765 | m_freem(m); |
1766 | | | 1766 | |
1767 | /* | | 1767 | /* |
1768 | * If all outstanding data are acked, stop | | 1768 | * If all outstanding data are acked, stop |
1769 | * retransmit timer, otherwise restart timer | | 1769 | * retransmit timer, otherwise restart timer |
1770 | * using current (possibly backed-off) value. | | 1770 | * using current (possibly backed-off) value. |
1771 | * If process is waiting for space, | | 1771 | * If process is waiting for space, |
1772 | * wakeup/selnotify/signal. If data | | 1772 | * wakeup/selnotify/signal. If data |
1773 | * are ready to send, let tcp_output | | 1773 | * are ready to send, let tcp_output |
1774 | * decide between more output or persist. | | 1774 | * decide between more output or persist. |
1775 | */ | | 1775 | */ |
1776 | if (tp->snd_una == tp->snd_max) | | 1776 | if (tp->snd_una == tp->snd_max) |
1777 | TCP_TIMER_DISARM(tp, TCPT_REXMT); | | 1777 | TCP_TIMER_DISARM(tp, TCPT_REXMT); |
1778 | else if (TCP_TIMER_ISARMED(tp, | | 1778 | else if (TCP_TIMER_ISARMED(tp, |
1779 | TCPT_PERSIST) == 0) | | 1779 | TCPT_PERSIST) == 0) |
1780 | TCP_TIMER_ARM(tp, TCPT_REXMT, | | 1780 | TCP_TIMER_ARM(tp, TCPT_REXMT, |
1781 | tp->t_rxtcur); | | 1781 | tp->t_rxtcur); |
1782 | | | 1782 | |
1783 | sowwakeup(so); | | 1783 | sowwakeup(so); |
1784 | if (so->so_snd.sb_cc) | | 1784 | if (so->so_snd.sb_cc) |
1785 | (void) tcp_output(tp); | | 1785 | (void) tcp_output(tp); |
1786 | if (tcp_saveti) | | 1786 | if (tcp_saveti) |
1787 | m_freem(tcp_saveti); | | 1787 | m_freem(tcp_saveti); |
1788 | return; | | 1788 | return; |
1789 | } | | 1789 | } |
1790 | } else if (th->th_ack == tp->snd_una && | | 1790 | } else if (th->th_ack == tp->snd_una && |
1791 | TAILQ_FIRST(&tp->segq) == NULL && | | 1791 | TAILQ_FIRST(&tp->segq) == NULL && |
1792 | tlen <= sbspace(&so->so_rcv)) { | | 1792 | tlen <= sbspace(&so->so_rcv)) { |
1793 | int newsize = 0; /* automatic sockbuf scaling */ | | 1793 | int newsize = 0; /* automatic sockbuf scaling */ |
1794 | | | 1794 | |
1795 | /* | | 1795 | /* |
1796 | * this is a pure, in-sequence data packet | | 1796 | * this is a pure, in-sequence data packet |
1797 | * with nothing on the reassembly queue and | | 1797 | * with nothing on the reassembly queue and |
1798 | * we have enough buffer space to take it. | | 1798 | * we have enough buffer space to take it. |
1799 | */ | | 1799 | */ |
1800 | tp->rcv_nxt += tlen; | | 1800 | tp->rcv_nxt += tlen; |
1801 | tcps = TCP_STAT_GETREF(); | | 1801 | tcps = TCP_STAT_GETREF(); |
1802 | tcps[TCP_STAT_PREDDAT]++; | | 1802 | tcps[TCP_STAT_PREDDAT]++; |
1803 | tcps[TCP_STAT_RCVPACK]++; | | 1803 | tcps[TCP_STAT_RCVPACK]++; |
1804 | tcps[TCP_STAT_RCVBYTE] += tlen; | | 1804 | tcps[TCP_STAT_RCVBYTE] += tlen; |
1805 | TCP_STAT_PUTREF(); | | 1805 | TCP_STAT_PUTREF(); |
1806 | nd6_hint(tp); | | 1806 | nd6_hint(tp); |
1807 | | | 1807 | |
1808 | /* | | 1808 | /* |
1809 | * Automatic sizing enables the performance of large buffers | | 1809 | * Automatic sizing enables the performance of large buffers |
1810 | * and most of the efficiency of small ones by only allocating | | 1810 | * and most of the efficiency of small ones by only allocating |
1811 | * space when it is needed. | | 1811 | * space when it is needed. |
1812 | * | | 1812 | * |
1813 | * On the receive side the socket buffer memory is only rarely | | 1813 | * On the receive side the socket buffer memory is only rarely |
1814 | * used to any significant extent. This allows us to be much | | 1814 | * used to any significant extent. This allows us to be much |
1815 | * more aggressive in scaling the receive socket buffer. For | | 1815 | * more aggressive in scaling the receive socket buffer. For |
1816 | * the case that the buffer space is actually used to a large | | 1816 | * the case that the buffer space is actually used to a large |
1817 | * extent and we run out of kernel memory we can simply drop | | 1817 | * extent and we run out of kernel memory we can simply drop |
1818 | * the new segments; TCP on the sender will just retransmit it | | 1818 | * the new segments; TCP on the sender will just retransmit it |
1819 | * later. Setting the buffer size too big may only consume too | | 1819 | * later. Setting the buffer size too big may only consume too |
1820 | * much kernel memory if the application doesn't read() from | | 1820 | * much kernel memory if the application doesn't read() from |
1821 | * the socket or packet loss or reordering makes use of the | | 1821 | * the socket or packet loss or reordering makes use of the |
1822 | * reassembly queue. | | 1822 | * reassembly queue. |
1823 | * | | 1823 | * |
1824 | * The criteria to step up the receive buffer one notch are: | | 1824 | * The criteria to step up the receive buffer one notch are: |
1825 | * 1. the number of bytes received during the time it takes | | 1825 | * 1. the number of bytes received during the time it takes |
1826 | * one timestamp to be reflected back to us (the RTT); | | 1826 | * one timestamp to be reflected back to us (the RTT); |
1827 | * 2. received bytes per RTT is within seven eighth of the | | 1827 | * 2. received bytes per RTT is within seven eighth of the |
1828 | * current socket buffer size; | | 1828 | * current socket buffer size; |
1829 | * 3. receive buffer size has not hit maximal automatic size; | | 1829 | * 3. receive buffer size has not hit maximal automatic size; |
1830 | * | | 1830 | * |
1831 | * This algorithm does one step per RTT at most and only if | | 1831 | * This algorithm does one step per RTT at most and only if |
1832 | * we receive a bulk stream w/o packet losses or reorderings. | | 1832 | * we receive a bulk stream w/o packet losses or reorderings. |
1833 | * Shrinking the buffer during idle times is not necessary as | | 1833 | * Shrinking the buffer during idle times is not necessary as |
1834 | * it doesn't consume any memory when idle. | | 1834 | * it doesn't consume any memory when idle. |
1835 | * | | 1835 | * |
1836 | * TODO: Only step up if the application is actually serving | | 1836 | * TODO: Only step up if the application is actually serving |
1837 | * the buffer to better manage the socket buffer resources. | | 1837 | * the buffer to better manage the socket buffer resources. |
1838 | */ | | 1838 | */ |
1839 | if (tcp_do_autorcvbuf && | | 1839 | if (tcp_do_autorcvbuf && |
1840 | opti.ts_ecr && | | 1840 | opti.ts_ecr && |
1841 | (so->so_rcv.sb_flags & SB_AUTOSIZE)) { | | 1841 | (so->so_rcv.sb_flags & SB_AUTOSIZE)) { |
1842 | if (opti.ts_ecr > tp->rfbuf_ts && | | 1842 | if (opti.ts_ecr > tp->rfbuf_ts && |
1843 | opti.ts_ecr - tp->rfbuf_ts < PR_SLOWHZ) { | | 1843 | opti.ts_ecr - tp->rfbuf_ts < PR_SLOWHZ) { |
1844 | if (tp->rfbuf_cnt > | | 1844 | if (tp->rfbuf_cnt > |
1845 | (so->so_rcv.sb_hiwat / 8 * 7) && | | 1845 | (so->so_rcv.sb_hiwat / 8 * 7) && |
1846 | so->so_rcv.sb_hiwat < | | 1846 | so->so_rcv.sb_hiwat < |
1847 | tcp_autorcvbuf_max) { | | 1847 | tcp_autorcvbuf_max) { |
1848 | newsize = | | 1848 | newsize = |
1849 | min(so->so_rcv.sb_hiwat + | | 1849 | min(so->so_rcv.sb_hiwat + |
1850 | tcp_autorcvbuf_inc, | | 1850 | tcp_autorcvbuf_inc, |
1851 | tcp_autorcvbuf_max); | | 1851 | tcp_autorcvbuf_max); |
1852 | } | | 1852 | } |
1853 | /* Start over with next RTT. */ | | 1853 | /* Start over with next RTT. */ |
1854 | tp->rfbuf_ts = 0; | | 1854 | tp->rfbuf_ts = 0; |
1855 | tp->rfbuf_cnt = 0; | | 1855 | tp->rfbuf_cnt = 0; |
1856 | } else | | 1856 | } else |
1857 | tp->rfbuf_cnt += tlen; /* add up */ | | 1857 | tp->rfbuf_cnt += tlen; /* add up */ |
1858 | } | | 1858 | } |
1859 | | | 1859 | |
1860 | /* | | 1860 | /* |
1861 | * Drop TCP, IP headers and TCP options then add data | | 1861 | * Drop TCP, IP headers and TCP options then add data |
1862 | * to socket buffer. | | 1862 | * to socket buffer. |
1863 | */ | | 1863 | */ |
1864 | if (so->so_state & SS_CANTRCVMORE) | | 1864 | if (so->so_state & SS_CANTRCVMORE) |
1865 | m_freem(m); | | 1865 | m_freem(m); |
1866 | else { | | 1866 | else { |
1867 | /* | | 1867 | /* |
1868 | * Set new socket buffer size. | | 1868 | * Set new socket buffer size. |
1869 | * Give up when limit is reached. | | 1869 | * Give up when limit is reached. |
1870 | */ | | 1870 | */ |
1871 | if (newsize) | | 1871 | if (newsize) |
1872 | if (!sbreserve(&so->so_rcv, | | 1872 | if (!sbreserve(&so->so_rcv, |
1873 | newsize, so)) | | 1873 | newsize, so)) |
1874 | so->so_rcv.sb_flags &= ~SB_AUTOSIZE; | | 1874 | so->so_rcv.sb_flags &= ~SB_AUTOSIZE; |
1875 | m_adj(m, toff + off); | | 1875 | m_adj(m, toff + off); |
1876 | sbappendstream(&so->so_rcv, m); | | 1876 | sbappendstream(&so->so_rcv, m); |
1877 | } | | 1877 | } |
1878 | sorwakeup(so); | | 1878 | sorwakeup(so); |
1879 | tcp_setup_ack(tp, th); | | 1879 | tcp_setup_ack(tp, th); |
1880 | if (tp->t_flags & TF_ACKNOW) | | 1880 | if (tp->t_flags & TF_ACKNOW) |
1881 | (void) tcp_output(tp); | | 1881 | (void) tcp_output(tp); |
1882 | if (tcp_saveti) | | 1882 | if (tcp_saveti) |
1883 | m_freem(tcp_saveti); | | 1883 | m_freem(tcp_saveti); |
1884 | return; | | 1884 | return; |
1885 | } | | 1885 | } |
1886 | } | | 1886 | } |
1887 | | | 1887 | |
1888 | /* | | 1888 | /* |
1889 | * Compute mbuf offset to TCP data segment. | | 1889 | * Compute mbuf offset to TCP data segment. |
1890 | */ | | 1890 | */ |
1891 | hdroptlen = toff + off; | | 1891 | hdroptlen = toff + off; |
1892 | | | 1892 | |
1893 | /* | | 1893 | /* |
1894 | * Calculate amount of space in receive window, | | 1894 | * Calculate amount of space in receive window, |
1895 | * and then do TCP input processing. | | 1895 | * and then do TCP input processing. |
1896 | * Receive window is amount of space in rcv queue, | | 1896 | * Receive window is amount of space in rcv queue, |
1897 | * but not less than advertised window. | | 1897 | * but not less than advertised window. |
1898 | */ | | 1898 | */ |
1899 | { int win; | | 1899 | { int win; |
1900 | | | 1900 | |
1901 | win = sbspace(&so->so_rcv); | | 1901 | win = sbspace(&so->so_rcv); |
1902 | if (win < 0) | | 1902 | if (win < 0) |
1903 | win = 0; | | 1903 | win = 0; |
1904 | tp->rcv_wnd = imax(win, (int)(tp->rcv_adv - tp->rcv_nxt)); | | 1904 | tp->rcv_wnd = imax(win, (int)(tp->rcv_adv - tp->rcv_nxt)); |
1905 | } | | 1905 | } |
1906 | | | 1906 | |
1907 | /* Reset receive buffer auto scaling when not in bulk receive mode. */ | | 1907 | /* Reset receive buffer auto scaling when not in bulk receive mode. */ |
1908 | tp->rfbuf_ts = 0; | | 1908 | tp->rfbuf_ts = 0; |
1909 | tp->rfbuf_cnt = 0; | | 1909 | tp->rfbuf_cnt = 0; |
1910 | | | 1910 | |
1911 | switch (tp->t_state) { | | 1911 | switch (tp->t_state) { |
1912 | /* | | 1912 | /* |
1913 | * If the state is SYN_SENT: | | 1913 | * If the state is SYN_SENT: |
1914 | * if seg contains an ACK, but not for our SYN, drop the input. | | 1914 | * if seg contains an ACK, but not for our SYN, drop the input. |
1915 | * if seg contains a RST, then drop the connection. | | 1915 | * if seg contains a RST, then drop the connection. |
1916 | * if seg does not contain SYN, then drop it. | | 1916 | * if seg does not contain SYN, then drop it. |
1917 | * Otherwise this is an acceptable SYN segment | | 1917 | * Otherwise this is an acceptable SYN segment |
1918 | * initialize tp->rcv_nxt and tp->irs | | 1918 | * initialize tp->rcv_nxt and tp->irs |
1919 | * if seg contains ack then advance tp->snd_una | | 1919 | * if seg contains ack then advance tp->snd_una |
1920 | * if seg contains a ECE and ECN support is enabled, the stream | | 1920 | * if seg contains a ECE and ECN support is enabled, the stream |
1921 | * is ECN capable. | | 1921 | * is ECN capable. |
1922 | * if SYN has been acked change to ESTABLISHED else SYN_RCVD state | | 1922 | * if SYN has been acked change to ESTABLISHED else SYN_RCVD state |
1923 | * arrange for segment to be acked (eventually) | | 1923 | * arrange for segment to be acked (eventually) |
1924 | * continue processing rest of data/controls, beginning with URG | | 1924 | * continue processing rest of data/controls, beginning with URG |
1925 | */ | | 1925 | */ |
1926 | case TCPS_SYN_SENT: | | 1926 | case TCPS_SYN_SENT: |
1927 | if ((tiflags & TH_ACK) && | | 1927 | if ((tiflags & TH_ACK) && |
1928 | (SEQ_LEQ(th->th_ack, tp->iss) || | | 1928 | (SEQ_LEQ(th->th_ack, tp->iss) || |
1929 | SEQ_GT(th->th_ack, tp->snd_max))) | | 1929 | SEQ_GT(th->th_ack, tp->snd_max))) |
1930 | goto dropwithreset; | | 1930 | goto dropwithreset; |
1931 | if (tiflags & TH_RST) { | | 1931 | if (tiflags & TH_RST) { |
1932 | if (tiflags & TH_ACK) | | 1932 | if (tiflags & TH_ACK) |
1933 | tp = tcp_drop(tp, ECONNREFUSED); | | 1933 | tp = tcp_drop(tp, ECONNREFUSED); |
1934 | goto drop; | | 1934 | goto drop; |
1935 | } | | 1935 | } |
1936 | if ((tiflags & TH_SYN) == 0) | | 1936 | if ((tiflags & TH_SYN) == 0) |
1937 | goto drop; | | 1937 | goto drop; |
1938 | if (tiflags & TH_ACK) { | | 1938 | if (tiflags & TH_ACK) { |
1939 | tp->snd_una = th->th_ack; | | 1939 | tp->snd_una = th->th_ack; |
1940 | if (SEQ_LT(tp->snd_nxt, tp->snd_una)) | | 1940 | if (SEQ_LT(tp->snd_nxt, tp->snd_una)) |
1941 | tp->snd_nxt = tp->snd_una; | | 1941 | tp->snd_nxt = tp->snd_una; |
1942 | if (SEQ_LT(tp->snd_high, tp->snd_una)) | | 1942 | if (SEQ_LT(tp->snd_high, tp->snd_una)) |
1943 | tp->snd_high = tp->snd_una; | | 1943 | tp->snd_high = tp->snd_una; |
1944 | TCP_TIMER_DISARM(tp, TCPT_REXMT); | | 1944 | TCP_TIMER_DISARM(tp, TCPT_REXMT); |
1945 | | | 1945 | |
1946 | if ((tiflags & TH_ECE) && tcp_do_ecn) { | | 1946 | if ((tiflags & TH_ECE) && tcp_do_ecn) { |
1947 | tp->t_flags |= TF_ECN_PERMIT; | | 1947 | tp->t_flags |= TF_ECN_PERMIT; |
1948 | TCP_STATINC(TCP_STAT_ECN_SHS); | | 1948 | TCP_STATINC(TCP_STAT_ECN_SHS); |
1949 | } | | 1949 | } |
1950 | | | 1950 | |
1951 | } | | 1951 | } |
1952 | tp->irs = th->th_seq; | | 1952 | tp->irs = th->th_seq; |
1953 | tcp_rcvseqinit(tp); | | 1953 | tcp_rcvseqinit(tp); |
1954 | tp->t_flags |= TF_ACKNOW; | | 1954 | tp->t_flags |= TF_ACKNOW; |
1955 | tcp_mss_from_peer(tp, opti.maxseg); | | 1955 | tcp_mss_from_peer(tp, opti.maxseg); |
1956 | | | 1956 | |
1957 | /* | | 1957 | /* |
1958 | * Initialize the initial congestion window. If we | | 1958 | * Initialize the initial congestion window. If we |
1959 | * had to retransmit the SYN, we must initialize cwnd | | 1959 | * had to retransmit the SYN, we must initialize cwnd |
1960 | * to 1 segment (i.e. the Loss Window). | | 1960 | * to 1 segment (i.e. the Loss Window). |
1961 | */ | | 1961 | */ |
1962 | if (tp->t_flags & TF_SYN_REXMT) | | 1962 | if (tp->t_flags & TF_SYN_REXMT) |
1963 | tp->snd_cwnd = tp->t_peermss; | | 1963 | tp->snd_cwnd = tp->t_peermss; |
1964 | else { | | 1964 | else { |
1965 | int ss = tcp_init_win; | | 1965 | int ss = tcp_init_win; |
1966 | #ifdef INET | | 1966 | #ifdef INET |
1967 | if (inp != NULL && in_localaddr(inp->inp_faddr)) | | 1967 | if (inp != NULL && in_localaddr(inp->inp_faddr)) |
1968 | ss = tcp_init_win_local; | | 1968 | ss = tcp_init_win_local; |
1969 | #endif | | 1969 | #endif |
1970 | #ifdef INET6 | | 1970 | #ifdef INET6 |
1971 | if (in6p != NULL && in6_localaddr(&in6p->in6p_faddr)) | | 1971 | if (in6p != NULL && in6_localaddr(&in6p->in6p_faddr)) |
1972 | ss = tcp_init_win_local; | | 1972 | ss = tcp_init_win_local; |
1973 | #endif | | 1973 | #endif |
1974 | tp->snd_cwnd = TCP_INITIAL_WINDOW(ss, tp->t_peermss); | | 1974 | tp->snd_cwnd = TCP_INITIAL_WINDOW(ss, tp->t_peermss); |
1975 | } | | 1975 | } |
1976 | | | 1976 | |
1977 | tcp_rmx_rtt(tp); | | 1977 | tcp_rmx_rtt(tp); |
1978 | if (tiflags & TH_ACK) { | | 1978 | if (tiflags & TH_ACK) { |
1979 | TCP_STATINC(TCP_STAT_CONNECTS); | | 1979 | TCP_STATINC(TCP_STAT_CONNECTS); |
1980 | soisconnected(so); | | 1980 | soisconnected(so); |
1981 | tcp_established(tp); | | 1981 | tcp_established(tp); |
1982 | /* Do window scaling on this connection? */ | | 1982 | /* Do window scaling on this connection? */ |
1983 | if ((tp->t_flags & (TF_RCVD_SCALE|TF_REQ_SCALE)) == | | 1983 | if ((tp->t_flags & (TF_RCVD_SCALE|TF_REQ_SCALE)) == |
1984 | (TF_RCVD_SCALE|TF_REQ_SCALE)) { | | 1984 | (TF_RCVD_SCALE|TF_REQ_SCALE)) { |
1985 | tp->snd_scale = tp->requested_s_scale; | | 1985 | tp->snd_scale = tp->requested_s_scale; |
1986 | tp->rcv_scale = tp->request_r_scale; | | 1986 | tp->rcv_scale = tp->request_r_scale; |
1987 | } | | 1987 | } |
1988 | TCP_REASS_LOCK(tp); | | 1988 | TCP_REASS_LOCK(tp); |
1989 | (void) tcp_reass(tp, NULL, (struct mbuf *)0, &tlen); | | 1989 | (void) tcp_reass(tp, NULL, (struct mbuf *)0, &tlen); |
1990 | TCP_REASS_UNLOCK(tp); | | 1990 | TCP_REASS_UNLOCK(tp); |
1991 | /* | | 1991 | /* |
1992 | * if we didn't have to retransmit the SYN, | | 1992 | * if we didn't have to retransmit the SYN, |
1993 | * use its rtt as our initial srtt & rtt var. | | 1993 | * use its rtt as our initial srtt & rtt var. |
1994 | */ | | 1994 | */ |
1995 | if (tp->t_rtttime) | | 1995 | if (tp->t_rtttime) |
1996 | tcp_xmit_timer(tp, tcp_now - tp->t_rtttime); | | 1996 | tcp_xmit_timer(tp, tcp_now - tp->t_rtttime); |
1997 | } else | | 1997 | } else |
1998 | tp->t_state = TCPS_SYN_RECEIVED; | | 1998 | tp->t_state = TCPS_SYN_RECEIVED; |
1999 | | | 1999 | |
2000 | /* | | 2000 | /* |
2001 | * Advance th->th_seq to correspond to first data byte. | | 2001 | * Advance th->th_seq to correspond to first data byte. |
2002 | * If data, trim to stay within window, | | 2002 | * If data, trim to stay within window, |
2003 | * dropping FIN if necessary. | | 2003 | * dropping FIN if necessary. |
2004 | */ | | 2004 | */ |
2005 | th->th_seq++; | | 2005 | th->th_seq++; |
2006 | if (tlen > tp->rcv_wnd) { | | 2006 | if (tlen > tp->rcv_wnd) { |
2007 | todrop = tlen - tp->rcv_wnd; | | 2007 | todrop = tlen - tp->rcv_wnd; |
2008 | m_adj(m, -todrop); | | 2008 | m_adj(m, -todrop); |
2009 | tlen = tp->rcv_wnd; | | 2009 | tlen = tp->rcv_wnd; |
2010 | tiflags &= ~TH_FIN; | | 2010 | tiflags &= ~TH_FIN; |
2011 | tcps = TCP_STAT_GETREF(); | | 2011 | tcps = TCP_STAT_GETREF(); |
2012 | tcps[TCP_STAT_RCVPACKAFTERWIN]++; | | 2012 | tcps[TCP_STAT_RCVPACKAFTERWIN]++; |
2013 | tcps[TCP_STAT_RCVBYTEAFTERWIN] += todrop; | | 2013 | tcps[TCP_STAT_RCVBYTEAFTERWIN] += todrop; |
2014 | TCP_STAT_PUTREF(); | | 2014 | TCP_STAT_PUTREF(); |
2015 | } | | 2015 | } |
2016 | tp->snd_wl1 = th->th_seq - 1; | | 2016 | tp->snd_wl1 = th->th_seq - 1; |
2017 | tp->rcv_up = th->th_seq; | | 2017 | tp->rcv_up = th->th_seq; |
2018 | goto step6; | | 2018 | goto step6; |
2019 | | | 2019 | |
2020 | /* | | 2020 | /* |
2021 | * If the state is SYN_RECEIVED: | | 2021 | * If the state is SYN_RECEIVED: |
2022 | * If seg contains an ACK, but not for our SYN, drop the input | | 2022 | * If seg contains an ACK, but not for our SYN, drop the input |
2023 | * and generate an RST. See page 36, rfc793 | | 2023 | * and generate an RST. See page 36, rfc793 |
2024 | */ | | 2024 | */ |
2025 | case TCPS_SYN_RECEIVED: | | 2025 | case TCPS_SYN_RECEIVED: |
2026 | if ((tiflags & TH_ACK) && | | 2026 | if ((tiflags & TH_ACK) && |
2027 | (SEQ_LEQ(th->th_ack, tp->iss) || | | 2027 | (SEQ_LEQ(th->th_ack, tp->iss) || |
2028 | SEQ_GT(th->th_ack, tp->snd_max))) | | 2028 | SEQ_GT(th->th_ack, tp->snd_max))) |
2029 | goto dropwithreset; | | 2029 | goto dropwithreset; |
2030 | break; | | 2030 | break; |
2031 | } | | 2031 | } |
2032 | | | 2032 | |
2033 | /* | | 2033 | /* |
2034 | * States other than LISTEN or SYN_SENT. | | 2034 | * States other than LISTEN or SYN_SENT. |
2035 | * First check timestamp, if present. | | 2035 | * First check timestamp, if present. |
2036 | * Then check that at least some bytes of segment are within | | 2036 | * Then check that at least some bytes of segment are within |
2037 | * receive window. If segment begins before rcv_nxt, | | 2037 | * receive window. If segment begins before rcv_nxt, |
2038 | * drop leading data (and SYN); if nothing left, just ack. | | 2038 | * drop leading data (and SYN); if nothing left, just ack. |
2039 | * | | 2039 | * |
2040 | * RFC 1323 PAWS: If we have a timestamp reply on this segment | | 2040 | * RFC 1323 PAWS: If we have a timestamp reply on this segment |
2041 | * and it's less than ts_recent, drop it. | | 2041 | * and it's less than ts_recent, drop it. |
2042 | */ | | 2042 | */ |
2043 | if (opti.ts_present && (tiflags & TH_RST) == 0 && tp->ts_recent && | | 2043 | if (opti.ts_present && (tiflags & TH_RST) == 0 && tp->ts_recent && |
2044 | TSTMP_LT(opti.ts_val, tp->ts_recent)) { | | 2044 | TSTMP_LT(opti.ts_val, tp->ts_recent)) { |
2045 | | | 2045 | |
2046 | /* Check to see if ts_recent is over 24 days old. */ | | 2046 | /* Check to see if ts_recent is over 24 days old. */ |
2047 | if (tcp_now - tp->ts_recent_age > TCP_PAWS_IDLE) { | | 2047 | if (tcp_now - tp->ts_recent_age > TCP_PAWS_IDLE) { |
2048 | /* | | 2048 | /* |
2049 | * Invalidate ts_recent. If this segment updates | | 2049 | * Invalidate ts_recent. If this segment updates |
2050 | * ts_recent, the age will be reset later and ts_recent | | 2050 | * ts_recent, the age will be reset later and ts_recent |
2051 | * will get a valid value. If it does not, setting | | 2051 | * will get a valid value. If it does not, setting |
2052 | * ts_recent to zero will at least satisfy the | | 2052 | * ts_recent to zero will at least satisfy the |
2053 | * requirement that zero be placed in the timestamp | | 2053 | * requirement that zero be placed in the timestamp |
2054 | * echo reply when ts_recent isn't valid. The | | 2054 | * echo reply when ts_recent isn't valid. The |
2055 | * age isn't reset until we get a valid ts_recent | | 2055 | * age isn't reset until we get a valid ts_recent |
2056 | * because we don't want out-of-order segments to be | | 2056 | * because we don't want out-of-order segments to be |
2057 | * dropped when ts_recent is old. | | 2057 | * dropped when ts_recent is old. |
2058 | */ | | 2058 | */ |
2059 | tp->ts_recent = 0; | | 2059 | tp->ts_recent = 0; |
2060 | } else { | | 2060 | } else { |
2061 | tcps = TCP_STAT_GETREF(); | | 2061 | tcps = TCP_STAT_GETREF(); |
2062 | tcps[TCP_STAT_RCVDUPPACK]++; | | 2062 | tcps[TCP_STAT_RCVDUPPACK]++; |
2063 | tcps[TCP_STAT_RCVDUPBYTE] += tlen; | | 2063 | tcps[TCP_STAT_RCVDUPBYTE] += tlen; |
2064 | tcps[TCP_STAT_PAWSDROP]++; | | 2064 | tcps[TCP_STAT_PAWSDROP]++; |
2065 | TCP_STAT_PUTREF(); | | 2065 | TCP_STAT_PUTREF(); |
2066 | tcp_new_dsack(tp, th->th_seq, tlen); | | 2066 | tcp_new_dsack(tp, th->th_seq, tlen); |
2067 | goto dropafterack; | | 2067 | goto dropafterack; |
2068 | } | | 2068 | } |
2069 | } | | 2069 | } |
2070 | | | 2070 | |
2071 | todrop = tp->rcv_nxt - th->th_seq; | | 2071 | todrop = tp->rcv_nxt - th->th_seq; |
2072 | dupseg = false; | | 2072 | dupseg = false; |
2073 | if (todrop > 0) { | | 2073 | if (todrop > 0) { |
2074 | if (tiflags & TH_SYN) { | | 2074 | if (tiflags & TH_SYN) { |
2075 | tiflags &= ~TH_SYN; | | 2075 | tiflags &= ~TH_SYN; |
2076 | th->th_seq++; | | 2076 | th->th_seq++; |
2077 | if (th->th_urp > 1) | | 2077 | if (th->th_urp > 1) |
2078 | th->th_urp--; | | 2078 | th->th_urp--; |
2079 | else { | | 2079 | else { |
2080 | tiflags &= ~TH_URG; | | 2080 | tiflags &= ~TH_URG; |
2081 | th->th_urp = 0; | | 2081 | th->th_urp = 0; |
2082 | } | | 2082 | } |
2083 | todrop--; | | 2083 | todrop--; |
2084 | } | | 2084 | } |
2085 | if (todrop > tlen || | | 2085 | if (todrop > tlen || |
2086 | (todrop == tlen && (tiflags & TH_FIN) == 0)) { | | 2086 | (todrop == tlen && (tiflags & TH_FIN) == 0)) { |
2087 | /* | | 2087 | /* |
2088 | * Any valid FIN or RST must be to the left of the | | 2088 | * Any valid FIN or RST must be to the left of the |
2089 | * window. At this point the FIN or RST must be a | | 2089 | * window. At this point the FIN or RST must be a |
2090 | * duplicate or out of sequence; drop it. | | 2090 | * duplicate or out of sequence; drop it. |
2091 | */ | | 2091 | */ |
2092 | if (tiflags & TH_RST) | | 2092 | if (tiflags & TH_RST) |
2093 | goto drop; | | 2093 | goto drop; |
2094 | tiflags &= ~(TH_FIN|TH_RST); | | 2094 | tiflags &= ~(TH_FIN|TH_RST); |
2095 | /* | | 2095 | /* |
2096 | * Send an ACK to resynchronize and drop any data. | | 2096 | * Send an ACK to resynchronize and drop any data. |
2097 | * But keep on processing for RST or ACK. | | 2097 | * But keep on processing for RST or ACK. |
2098 | */ | | 2098 | */ |
2099 | tp->t_flags |= TF_ACKNOW; | | 2099 | tp->t_flags |= TF_ACKNOW; |
2100 | todrop = tlen; | | 2100 | todrop = tlen; |
2101 | dupseg = true; | | 2101 | dupseg = true; |
2102 | tcps = TCP_STAT_GETREF(); | | 2102 | tcps = TCP_STAT_GETREF(); |
2103 | tcps[TCP_STAT_RCVDUPPACK]++; | | 2103 | tcps[TCP_STAT_RCVDUPPACK]++; |
2104 | tcps[TCP_STAT_RCVDUPBYTE] += todrop; | | 2104 | tcps[TCP_STAT_RCVDUPBYTE] += todrop; |
2105 | TCP_STAT_PUTREF(); | | 2105 | TCP_STAT_PUTREF(); |
2106 | } else if ((tiflags & TH_RST) && | | 2106 | } else if ((tiflags & TH_RST) && |
2107 | th->th_seq != tp->last_ack_sent) { | | 2107 | th->th_seq != tp->rcv_nxt) { |
2108 | /* | | 2108 | /* |
2109 | * Test for reset before adjusting the sequence | | 2109 | * Test for reset before adjusting the sequence |
2110 | * number for overlapping data. | | 2110 | * number for overlapping data. |
2111 | */ | | 2111 | */ |
2112 | goto dropafterack_ratelim; | | 2112 | goto dropafterack_ratelim; |
2113 | } else { | | 2113 | } else { |
2114 | tcps = TCP_STAT_GETREF(); | | 2114 | tcps = TCP_STAT_GETREF(); |
2115 | tcps[TCP_STAT_RCVPARTDUPPACK]++; | | 2115 | tcps[TCP_STAT_RCVPARTDUPPACK]++; |
2116 | tcps[TCP_STAT_RCVPARTDUPBYTE] += todrop; | | 2116 | tcps[TCP_STAT_RCVPARTDUPBYTE] += todrop; |
2117 | TCP_STAT_PUTREF(); | | 2117 | TCP_STAT_PUTREF(); |
2118 | } | | 2118 | } |
2119 | tcp_new_dsack(tp, th->th_seq, todrop); | | 2119 | tcp_new_dsack(tp, th->th_seq, todrop); |
2120 | hdroptlen += todrop; /*drop from head afterwards*/ | | 2120 | hdroptlen += todrop; /*drop from head afterwards*/ |
2121 | th->th_seq += todrop; | | 2121 | th->th_seq += todrop; |
2122 | tlen -= todrop; | | 2122 | tlen -= todrop; |
2123 | if (th->th_urp > todrop) | | 2123 | if (th->th_urp > todrop) |
2124 | th->th_urp -= todrop; | | 2124 | th->th_urp -= todrop; |
2125 | else { | | 2125 | else { |
2126 | tiflags &= ~TH_URG; | | 2126 | tiflags &= ~TH_URG; |
2127 | th->th_urp = 0; | | 2127 | th->th_urp = 0; |
2128 | } | | 2128 | } |
2129 | } | | 2129 | } |
2130 | | | 2130 | |
2131 | /* | | 2131 | /* |
2132 | * If new data are received on a connection after the | | 2132 | * If new data are received on a connection after the |
2133 | * user processes are gone, then RST the other end. | | 2133 | * user processes are gone, then RST the other end. |
2134 | */ | | 2134 | */ |
2135 | if ((so->so_state & SS_NOFDREF) && | | 2135 | if ((so->so_state & SS_NOFDREF) && |
2136 | tp->t_state > TCPS_CLOSE_WAIT && tlen) { | | 2136 | tp->t_state > TCPS_CLOSE_WAIT && tlen) { |
2137 | tp = tcp_close(tp); | | 2137 | tp = tcp_close(tp); |
2138 | TCP_STATINC(TCP_STAT_RCVAFTERCLOSE); | | 2138 | TCP_STATINC(TCP_STAT_RCVAFTERCLOSE); |
2139 | goto dropwithreset; | | 2139 | goto dropwithreset; |
2140 | } | | 2140 | } |
2141 | | | 2141 | |
2142 | /* | | 2142 | /* |
2143 | * If segment ends after window, drop trailing data | | 2143 | * If segment ends after window, drop trailing data |
2144 | * (and PUSH and FIN); if nothing left, just ACK. | | 2144 | * (and PUSH and FIN); if nothing left, just ACK. |
2145 | */ | | 2145 | */ |
2146 | todrop = (th->th_seq + tlen) - (tp->rcv_nxt+tp->rcv_wnd); | | 2146 | todrop = (th->th_seq + tlen) - (tp->rcv_nxt+tp->rcv_wnd); |
2147 | if (todrop > 0) { | | 2147 | if (todrop > 0) { |
2148 | TCP_STATINC(TCP_STAT_RCVPACKAFTERWIN); | | 2148 | TCP_STATINC(TCP_STAT_RCVPACKAFTERWIN); |
2149 | if (todrop >= tlen) { | | 2149 | if (todrop >= tlen) { |
2150 | /* | | 2150 | /* |
2151 | * The segment actually starts after the window. | | 2151 | * The segment actually starts after the window. |
2152 | * th->th_seq + tlen - tp->rcv_nxt - tp->rcv_wnd >= tlen | | 2152 | * th->th_seq + tlen - tp->rcv_nxt - tp->rcv_wnd >= tlen |
2153 | * th->th_seq - tp->rcv_nxt - tp->rcv_wnd >= 0 | | 2153 | * th->th_seq - tp->rcv_nxt - tp->rcv_wnd >= 0 |
2154 | * th->th_seq >= tp->rcv_nxt + tp->rcv_wnd | | 2154 | * th->th_seq >= tp->rcv_nxt + tp->rcv_wnd |
2155 | */ | | 2155 | */ |
2156 | TCP_STATADD(TCP_STAT_RCVBYTEAFTERWIN, tlen); | | 2156 | TCP_STATADD(TCP_STAT_RCVBYTEAFTERWIN, tlen); |
2157 | /* | | 2157 | /* |
2158 | * If a new connection request is received | | 2158 | * If a new connection request is received |
2159 | * while in TIME_WAIT, drop the old connection | | 2159 | * while in TIME_WAIT, drop the old connection |
2160 | * and start over if the sequence numbers | | 2160 | * and start over if the sequence numbers |
2161 | * are above the previous ones. | | 2161 | * are above the previous ones. |
2162 | * | | 2162 | * |
2163 | * NOTE: We will checksum the packet again, and | | 2163 | * NOTE: We will checksum the packet again, and |
2164 | * so we need to put the header fields back into | | 2164 | * so we need to put the header fields back into |
2165 | * network order! | | 2165 | * network order! |
2166 | * XXX This kind of sucks, but we don't expect | | 2166 | * XXX This kind of sucks, but we don't expect |
2167 | * XXX this to happen very often, so maybe it | | 2167 | * XXX this to happen very often, so maybe it |
2168 | * XXX doesn't matter so much. | | 2168 | * XXX doesn't matter so much. |
2169 | */ | | 2169 | */ |
2170 | if (tiflags & TH_SYN && | | 2170 | if (tiflags & TH_SYN && |
2171 | tp->t_state == TCPS_TIME_WAIT && | | 2171 | tp->t_state == TCPS_TIME_WAIT && |
2172 | SEQ_GT(th->th_seq, tp->rcv_nxt)) { | | 2172 | SEQ_GT(th->th_seq, tp->rcv_nxt)) { |
2173 | tp = tcp_close(tp); | | 2173 | tp = tcp_close(tp); |
2174 | tcp_fields_to_net(th); | | 2174 | tcp_fields_to_net(th); |
2175 | goto findpcb; | | 2175 | goto findpcb; |
2176 | } | | 2176 | } |
2177 | /* | | 2177 | /* |
2178 | * If window is closed can only take segments at | | 2178 | * If window is closed can only take segments at |
2179 | * window edge, and have to drop data and PUSH from | | 2179 | * window edge, and have to drop data and PUSH from |
2180 | * incoming segments. Continue processing, but | | 2180 | * incoming segments. Continue processing, but |
2181 | * remember to ack. Otherwise, drop segment | | 2181 | * remember to ack. Otherwise, drop segment |
2182 | * and (if not RST) ack. | | 2182 | * and (if not RST) ack. |
2183 | */ | | 2183 | */ |
2184 | if (tp->rcv_wnd == 0 && th->th_seq == tp->rcv_nxt) { | | 2184 | if (tp->rcv_wnd == 0 && th->th_seq == tp->rcv_nxt) { |
2185 | tp->t_flags |= TF_ACKNOW; | | 2185 | tp->t_flags |= TF_ACKNOW; |
2186 | TCP_STATINC(TCP_STAT_RCVWINPROBE); | | 2186 | TCP_STATINC(TCP_STAT_RCVWINPROBE); |
2187 | } else | | 2187 | } else |
2188 | goto dropafterack; | | 2188 | goto dropafterack; |
2189 | } else | | 2189 | } else |
2190 | TCP_STATADD(TCP_STAT_RCVBYTEAFTERWIN, todrop); | | 2190 | TCP_STATADD(TCP_STAT_RCVBYTEAFTERWIN, todrop); |
2191 | m_adj(m, -todrop); | | 2191 | m_adj(m, -todrop); |
2192 | tlen -= todrop; | | 2192 | tlen -= todrop; |
2193 | tiflags &= ~(TH_PUSH|TH_FIN); | | 2193 | tiflags &= ~(TH_PUSH|TH_FIN); |
2194 | } | | 2194 | } |
2195 | | | 2195 | |
2196 | /* | | 2196 | /* |
2197 | * If last ACK falls within this segment's sequence numbers, | | 2197 | * If last ACK falls within this segment's sequence numbers, |
2198 | * record the timestamp. | | 2198 | * record the timestamp. |
2199 | * NOTE: | | 2199 | * NOTE: |
2200 | * 1) That the test incorporates suggestions from the latest | | 2200 | * 1) That the test incorporates suggestions from the latest |
2201 | * proposal of the tcplw@cray.com list (Braden 1993/04/26). | | 2201 | * proposal of the tcplw@cray.com list (Braden 1993/04/26). |
2202 | * 2) That updating only on newer timestamps interferes with | | 2202 | * 2) That updating only on newer timestamps interferes with |
2203 | * our earlier PAWS tests, so this check should be solely | | 2203 | * our earlier PAWS tests, so this check should be solely |
2204 | * predicated on the sequence space of this segment. | | 2204 | * predicated on the sequence space of this segment. |
2205 | * 3) That we modify the segment boundary check to be | | 2205 | * 3) That we modify the segment boundary check to be |
2206 | * Last.ACK.Sent <= SEG.SEQ + SEG.Len | | 2206 | * Last.ACK.Sent <= SEG.SEQ + SEG.Len |
2207 | * instead of RFC1323's | | 2207 | * instead of RFC1323's |
2208 | * Last.ACK.Sent < SEG.SEQ + SEG.Len, | | 2208 | * Last.ACK.Sent < SEG.SEQ + SEG.Len, |
2209 | * This modified check allows us to overcome RFC1323's | | 2209 | * This modified check allows us to overcome RFC1323's |
2210 | * limitations as described in Stevens TCP/IP Illustrated | | 2210 | * limitations as described in Stevens TCP/IP Illustrated |
2211 | * Vol. 2 p.869. In such cases, we can still calculate the | | 2211 | * Vol. 2 p.869. In such cases, we can still calculate the |
2212 | * RTT correctly when RCV.NXT == Last.ACK.Sent. | | 2212 | * RTT correctly when RCV.NXT == Last.ACK.Sent. |
2213 | */ | | 2213 | */ |
2214 | if (opti.ts_present && | | 2214 | if (opti.ts_present && |
2215 | SEQ_LEQ(th->th_seq, tp->last_ack_sent) && | | 2215 | SEQ_LEQ(th->th_seq, tp->last_ack_sent) && |
2216 | SEQ_LEQ(tp->last_ack_sent, th->th_seq + tlen + | | 2216 | SEQ_LEQ(tp->last_ack_sent, th->th_seq + tlen + |
2217 | ((tiflags & (TH_SYN|TH_FIN)) != 0))) { | | 2217 | ((tiflags & (TH_SYN|TH_FIN)) != 0))) { |
2218 | tp->ts_recent_age = tcp_now; | | 2218 | tp->ts_recent_age = tcp_now; |
2219 | tp->ts_recent = opti.ts_val; | | 2219 | tp->ts_recent = opti.ts_val; |
2220 | } | | 2220 | } |
2221 | | | 2221 | |
2222 | /* | | 2222 | /* |
2223 | * If the RST bit is set examine the state: | | 2223 | * If the RST bit is set examine the state: |
2224 | * SYN_RECEIVED STATE: | | 2224 | * SYN_RECEIVED STATE: |
2225 | * If passive open, return to LISTEN state. | | 2225 | * If passive open, return to LISTEN state. |
2226 | * If active open, inform user that connection was refused. | | 2226 | * If active open, inform user that connection was refused. |
2227 | * ESTABLISHED, FIN_WAIT_1, FIN_WAIT2, CLOSE_WAIT STATES: | | 2227 | * ESTABLISHED, FIN_WAIT_1, FIN_WAIT2, CLOSE_WAIT STATES: |
2228 | * Inform user that connection was reset, and close tcb. | | 2228 | * Inform user that connection was reset, and close tcb. |
2229 | * CLOSING, LAST_ACK, TIME_WAIT STATES | | 2229 | * CLOSING, LAST_ACK, TIME_WAIT STATES |
2230 | * Close the tcb. | | 2230 | * Close the tcb. |
2231 | */ | | 2231 | */ |
2232 | if (tiflags & TH_RST) { | | 2232 | if (tiflags & TH_RST) { |
2233 | if (th->th_seq != tp->last_ack_sent) | | 2233 | if (th->th_seq != tp->rcv_nxt) |
2234 | goto dropafterack_ratelim; | | 2234 | goto dropafterack_ratelim; |
2235 | | | 2235 | |
2236 | switch (tp->t_state) { | | 2236 | switch (tp->t_state) { |
2237 | case TCPS_SYN_RECEIVED: | | 2237 | case TCPS_SYN_RECEIVED: |
2238 | so->so_error = ECONNREFUSED; | | 2238 | so->so_error = ECONNREFUSED; |
2239 | goto close; | | 2239 | goto close; |
2240 | | | 2240 | |
2241 | case TCPS_ESTABLISHED: | | 2241 | case TCPS_ESTABLISHED: |
2242 | case TCPS_FIN_WAIT_1: | | 2242 | case TCPS_FIN_WAIT_1: |
2243 | case TCPS_FIN_WAIT_2: | | 2243 | case TCPS_FIN_WAIT_2: |
2244 | case TCPS_CLOSE_WAIT: | | 2244 | case TCPS_CLOSE_WAIT: |
2245 | so->so_error = ECONNRESET; | | 2245 | so->so_error = ECONNRESET; |
2246 | close: | | 2246 | close: |
2247 | tp->t_state = TCPS_CLOSED; | | 2247 | tp->t_state = TCPS_CLOSED; |
2248 | TCP_STATINC(TCP_STAT_DROPS); | | 2248 | TCP_STATINC(TCP_STAT_DROPS); |
2249 | tp = tcp_close(tp); | | 2249 | tp = tcp_close(tp); |
2250 | goto drop; | | 2250 | goto drop; |
2251 | | | 2251 | |
2252 | case TCPS_CLOSING: | | 2252 | case TCPS_CLOSING: |
2253 | case TCPS_LAST_ACK: | | 2253 | case TCPS_LAST_ACK: |
2254 | case TCPS_TIME_WAIT: | | 2254 | case TCPS_TIME_WAIT: |
2255 | tp = tcp_close(tp); | | 2255 | tp = tcp_close(tp); |
2256 | goto drop; | | 2256 | goto drop; |
2257 | } | | 2257 | } |
2258 | } | | 2258 | } |
2259 | | | 2259 | |
2260 | /* | | 2260 | /* |
2261 | * Since we've covered the SYN-SENT and SYN-RECEIVED states above | | 2261 | * Since we've covered the SYN-SENT and SYN-RECEIVED states above |
2262 | * we must be in a synchronized state. RFC791 states (under RST | | 2262 | * we must be in a synchronized state. RFC791 states (under RST |
2263 | * generation) that any unacceptable segment (an out-of-order SYN | | 2263 | * generation) that any unacceptable segment (an out-of-order SYN |
2264 | * qualifies) received in a synchronized state must elicit only an | | 2264 | * qualifies) received in a synchronized state must elicit only an |
2265 | * empty acknowledgment segment ... and the connection remains in | | 2265 | * empty acknowledgment segment ... and the connection remains in |
2266 | * the same state. | | 2266 | * the same state. |
2267 | */ | | 2267 | */ |
2268 | if (tiflags & TH_SYN) { | | 2268 | if (tiflags & TH_SYN) { |
2269 | if (tp->rcv_nxt == th->th_seq) { | | 2269 | if (tp->rcv_nxt == th->th_seq) { |
2270 | tcp_respond(tp, m, m, th, (tcp_seq)0, th->th_ack - 1, | | 2270 | tcp_respond(tp, m, m, th, (tcp_seq)0, th->th_ack - 1, |
2271 | TH_ACK); | | 2271 | TH_ACK); |
2272 | if (tcp_saveti) | | 2272 | if (tcp_saveti) |
2273 | m_freem(tcp_saveti); | | 2273 | m_freem(tcp_saveti); |
2274 | return; | | 2274 | return; |
2275 | } | | 2275 | } |
2276 | | | 2276 | |
2277 | goto dropafterack_ratelim; | | 2277 | goto dropafterack_ratelim; |
2278 | } | | 2278 | } |
2279 | | | 2279 | |
2280 | /* | | 2280 | /* |
2281 | * If the ACK bit is off we drop the segment and return. | | 2281 | * If the ACK bit is off we drop the segment and return. |
2282 | */ | | 2282 | */ |
2283 | if ((tiflags & TH_ACK) == 0) { | | 2283 | if ((tiflags & TH_ACK) == 0) { |
2284 | if (tp->t_flags & TF_ACKNOW) | | 2284 | if (tp->t_flags & TF_ACKNOW) |
2285 | goto dropafterack; | | 2285 | goto dropafterack; |
2286 | else | | 2286 | else |
2287 | goto drop; | | 2287 | goto drop; |
2288 | } | | 2288 | } |
2289 | | | 2289 | |
2290 | /* | | 2290 | /* |
2291 | * Ack processing. | | 2291 | * Ack processing. |
2292 | */ | | 2292 | */ |
2293 | switch (tp->t_state) { | | 2293 | switch (tp->t_state) { |
2294 | | | 2294 | |
2295 | /* | | 2295 | /* |
2296 | * In SYN_RECEIVED state if the ack ACKs our SYN then enter | | 2296 | * In SYN_RECEIVED state if the ack ACKs our SYN then enter |
2297 | * ESTABLISHED state and continue processing, otherwise | | 2297 | * ESTABLISHED state and continue processing, otherwise |
2298 | * send an RST. | | 2298 | * send an RST. |
2299 | */ | | 2299 | */ |
2300 | case TCPS_SYN_RECEIVED: | | 2300 | case TCPS_SYN_RECEIVED: |
2301 | if (SEQ_GT(tp->snd_una, th->th_ack) || | | 2301 | if (SEQ_GT(tp->snd_una, th->th_ack) || |
2302 | SEQ_GT(th->th_ack, tp->snd_max)) | | 2302 | SEQ_GT(th->th_ack, tp->snd_max)) |
2303 | goto dropwithreset; | | 2303 | goto dropwithreset; |
2304 | TCP_STATINC(TCP_STAT_CONNECTS); | | 2304 | TCP_STATINC(TCP_STAT_CONNECTS); |
2305 | soisconnected(so); | | 2305 | soisconnected(so); |
2306 | tcp_established(tp); | | 2306 | tcp_established(tp); |
2307 | /* Do window scaling? */ | | 2307 | /* Do window scaling? */ |
2308 | if ((tp->t_flags & (TF_RCVD_SCALE|TF_REQ_SCALE)) == | | 2308 | if ((tp->t_flags & (TF_RCVD_SCALE|TF_REQ_SCALE)) == |
2309 | (TF_RCVD_SCALE|TF_REQ_SCALE)) { | | 2309 | (TF_RCVD_SCALE|TF_REQ_SCALE)) { |
2310 | tp->snd_scale = tp->requested_s_scale; | | 2310 | tp->snd_scale = tp->requested_s_scale; |
2311 | tp->rcv_scale = tp->request_r_scale; | | 2311 | tp->rcv_scale = tp->request_r_scale; |
2312 | } | | 2312 | } |
2313 | TCP_REASS_LOCK(tp); | | 2313 | TCP_REASS_LOCK(tp); |
2314 | (void) tcp_reass(tp, NULL, (struct mbuf *)0, &tlen); | | 2314 | (void) tcp_reass(tp, NULL, (struct mbuf *)0, &tlen); |
2315 | TCP_REASS_UNLOCK(tp); | | 2315 | TCP_REASS_UNLOCK(tp); |
2316 | tp->snd_wl1 = th->th_seq - 1; | | 2316 | tp->snd_wl1 = th->th_seq - 1; |
2317 | /* fall into ... */ | | 2317 | /* fall into ... */ |
2318 | | | 2318 | |
2319 | /* | | 2319 | /* |
2320 | * In ESTABLISHED state: drop duplicate ACKs; ACK out of range | | 2320 | * In ESTABLISHED state: drop duplicate ACKs; ACK out of range |
2321 | * ACKs. If the ack is in the range | | 2321 | * ACKs. If the ack is in the range |
2322 | * tp->snd_una < th->th_ack <= tp->snd_max | | 2322 | * tp->snd_una < th->th_ack <= tp->snd_max |
2323 | * then advance tp->snd_una to th->th_ack and drop | | 2323 | * then advance tp->snd_una to th->th_ack and drop |
2324 | * data from the retransmission queue. If this ACK reflects | | 2324 | * data from the retransmission queue. If this ACK reflects |
2325 | * more up to date window information we update our window information. | | 2325 | * more up to date window information we update our window information. |
2326 | */ | | 2326 | */ |
2327 | case TCPS_ESTABLISHED: | | 2327 | case TCPS_ESTABLISHED: |
2328 | case TCPS_FIN_WAIT_1: | | 2328 | case TCPS_FIN_WAIT_1: |
2329 | case TCPS_FIN_WAIT_2: | | 2329 | case TCPS_FIN_WAIT_2: |
2330 | case TCPS_CLOSE_WAIT: | | 2330 | case TCPS_CLOSE_WAIT: |
2331 | case TCPS_CLOSING: | | 2331 | case TCPS_CLOSING: |
2332 | case TCPS_LAST_ACK: | | 2332 | case TCPS_LAST_ACK: |
2333 | case TCPS_TIME_WAIT: | | 2333 | case TCPS_TIME_WAIT: |
2334 | | | 2334 | |
2335 | if (SEQ_LEQ(th->th_ack, tp->snd_una)) { | | 2335 | if (SEQ_LEQ(th->th_ack, tp->snd_una)) { |
2336 | if (tlen == 0 && !dupseg && tiwin == tp->snd_wnd) { | | 2336 | if (tlen == 0 && !dupseg && tiwin == tp->snd_wnd) { |
2337 | TCP_STATINC(TCP_STAT_RCVDUPPACK); | | 2337 | TCP_STATINC(TCP_STAT_RCVDUPPACK); |
2338 | /* | | 2338 | /* |
2339 | * If we have outstanding data (other than | | 2339 | * If we have outstanding data (other than |
2340 | * a window probe), this is a completely | | 2340 | * a window probe), this is a completely |
2341 | * duplicate ack (ie, window info didn't | | 2341 | * duplicate ack (ie, window info didn't |
2342 | * change), the ack is the biggest we've | | 2342 | * change), the ack is the biggest we've |
2343 | * seen and we've seen exactly our rexmt | | 2343 | * seen and we've seen exactly our rexmt |
2344 | * threshhold of them, assume a packet | | 2344 | * threshhold of them, assume a packet |
2345 | * has been dropped and retransmit it. | | 2345 | * has been dropped and retransmit it. |
2346 | * Kludge snd_nxt & the congestion | | 2346 | * Kludge snd_nxt & the congestion |
2347 | * window so we send only this one | | 2347 | * window so we send only this one |
2348 | * packet. | | 2348 | * packet. |
2349 | */ | | 2349 | */ |
2350 | if (TCP_TIMER_ISARMED(tp, TCPT_REXMT) == 0 || | | 2350 | if (TCP_TIMER_ISARMED(tp, TCPT_REXMT) == 0 || |
2351 | th->th_ack != tp->snd_una) | | 2351 | th->th_ack != tp->snd_una) |
2352 | tp->t_dupacks = 0; | | 2352 | tp->t_dupacks = 0; |
2353 | else if (tp->t_partialacks < 0 && | | 2353 | else if (tp->t_partialacks < 0 && |
2354 | (++tp->t_dupacks == tcprexmtthresh || | | 2354 | (++tp->t_dupacks == tcprexmtthresh || |
2355 | TCP_FACK_FASTRECOV(tp))) { | | 2355 | TCP_FACK_FASTRECOV(tp))) { |
2356 | /* | | 2356 | /* |
2357 | * Do the fast retransmit, and adjust | | 2357 | * Do the fast retransmit, and adjust |
2358 | * congestion control paramenters. | | 2358 | * congestion control paramenters. |
2359 | */ | | 2359 | */ |
2360 | if (tp->t_congctl->fast_retransmit(tp, th)) { | | 2360 | if (tp->t_congctl->fast_retransmit(tp, th)) { |
2361 | /* False fast retransmit */ | | 2361 | /* False fast retransmit */ |
2362 | break; | | 2362 | break; |
2363 | } else | | 2363 | } else |
2364 | goto drop; | | 2364 | goto drop; |
2365 | } else if (tp->t_dupacks > tcprexmtthresh) { | | 2365 | } else if (tp->t_dupacks > tcprexmtthresh) { |
2366 | tp->snd_cwnd += tp->t_segsz; | | 2366 | tp->snd_cwnd += tp->t_segsz; |
2367 | (void) tcp_output(tp); | | 2367 | (void) tcp_output(tp); |
2368 | goto drop; | | 2368 | goto drop; |
2369 | } | | 2369 | } |
2370 | } else { | | 2370 | } else { |
2371 | /* | | 2371 | /* |
2372 | * If the ack appears to be very old, only | | 2372 | * If the ack appears to be very old, only |
2373 | * allow data that is in-sequence. This | | 2373 | * allow data that is in-sequence. This |
2374 | * makes it somewhat more difficult to insert | | 2374 | * makes it somewhat more difficult to insert |
2375 | * forged data by guessing sequence numbers. | | 2375 | * forged data by guessing sequence numbers. |
2376 | * Sent an ack to try to update the send | | 2376 | * Sent an ack to try to update the send |
2377 | * sequence number on the other side. | | 2377 | * sequence number on the other side. |
2378 | */ | | 2378 | */ |
2379 | if (tlen && th->th_seq != tp->rcv_nxt && | | 2379 | if (tlen && th->th_seq != tp->rcv_nxt && |
2380 | SEQ_LT(th->th_ack, | | 2380 | SEQ_LT(th->th_ack, |
2381 | tp->snd_una - tp->max_sndwnd)) | | 2381 | tp->snd_una - tp->max_sndwnd)) |
2382 | goto dropafterack; | | 2382 | goto dropafterack; |
2383 | } | | 2383 | } |
2384 | break; | | 2384 | break; |
2385 | } | | 2385 | } |
2386 | /* | | 2386 | /* |
2387 | * If the congestion window was inflated to account | | 2387 | * If the congestion window was inflated to account |
2388 | * for the other side's cached packets, retract it. | | 2388 | * for the other side's cached packets, retract it. |
2389 | */ | | 2389 | */ |
2390 | /* XXX: make SACK have his own congestion control | | 2390 | /* XXX: make SACK have his own congestion control |
2391 | * struct -- rpaulo */ | | 2391 | * struct -- rpaulo */ |
2392 | if (TCP_SACK_ENABLED(tp)) | | 2392 | if (TCP_SACK_ENABLED(tp)) |
2393 | tcp_sack_newack(tp, th); | | 2393 | tcp_sack_newack(tp, th); |
2394 | else | | 2394 | else |
2395 | tp->t_congctl->fast_retransmit_newack(tp, th); | | 2395 | tp->t_congctl->fast_retransmit_newack(tp, th); |
2396 | if (SEQ_GT(th->th_ack, tp->snd_max)) { | | 2396 | if (SEQ_GT(th->th_ack, tp->snd_max)) { |
2397 | TCP_STATINC(TCP_STAT_RCVACKTOOMUCH); | | 2397 | TCP_STATINC(TCP_STAT_RCVACKTOOMUCH); |
2398 | goto dropafterack; | | 2398 | goto dropafterack; |
2399 | } | | 2399 | } |
2400 | acked = th->th_ack - tp->snd_una; | | 2400 | acked = th->th_ack - tp->snd_una; |
2401 | tcps = TCP_STAT_GETREF(); | | 2401 | tcps = TCP_STAT_GETREF(); |
2402 | tcps[TCP_STAT_RCVACKPACK]++; | | 2402 | tcps[TCP_STAT_RCVACKPACK]++; |
2403 | tcps[TCP_STAT_RCVACKBYTE] += acked; | | 2403 | tcps[TCP_STAT_RCVACKBYTE] += acked; |
2404 | TCP_STAT_PUTREF(); | | 2404 | TCP_STAT_PUTREF(); |
2405 | | | 2405 | |
2406 | /* | | 2406 | /* |
2407 | * If we have a timestamp reply, update smoothed | | 2407 | * If we have a timestamp reply, update smoothed |
2408 | * round trip time. If no timestamp is present but | | 2408 | * round trip time. If no timestamp is present but |
2409 | * transmit timer is running and timed sequence | | 2409 | * transmit timer is running and timed sequence |
2410 | * number was acked, update smoothed round trip time. | | 2410 | * number was acked, update smoothed round trip time. |
2411 | * Since we now have an rtt measurement, cancel the | | 2411 | * Since we now have an rtt measurement, cancel the |
2412 | * timer backoff (cf., Phil Karn's retransmit alg.). | | 2412 | * timer backoff (cf., Phil Karn's retransmit alg.). |
2413 | * Recompute the initial retransmit timer. | | 2413 | * Recompute the initial retransmit timer. |
2414 | */ | | 2414 | */ |
2415 | if (ts_rtt) | | 2415 | if (ts_rtt) |
2416 | tcp_xmit_timer(tp, ts_rtt); | | 2416 | tcp_xmit_timer(tp, ts_rtt); |
2417 | else if (tp->t_rtttime && SEQ_GT(th->th_ack, tp->t_rtseq)) | | 2417 | else if (tp->t_rtttime && SEQ_GT(th->th_ack, tp->t_rtseq)) |
2418 | tcp_xmit_timer(tp, tcp_now - tp->t_rtttime); | | 2418 | tcp_xmit_timer(tp, tcp_now - tp->t_rtttime); |
2419 | | | 2419 | |
2420 | /* | | 2420 | /* |
2421 | * If all outstanding data is acked, stop retransmit | | 2421 | * If all outstanding data is acked, stop retransmit |
2422 | * timer and remember to restart (more output or persist). | | 2422 | * timer and remember to restart (more output or persist). |
2423 | * If there is more data to be acked, restart retransmit | | 2423 | * If there is more data to be acked, restart retransmit |
2424 | * timer, using current (possibly backed-off) value. | | 2424 | * timer, using current (possibly backed-off) value. |
2425 | */ | | 2425 | */ |
2426 | if (th->th_ack == tp->snd_max) { | | 2426 | if (th->th_ack == tp->snd_max) { |
2427 | TCP_TIMER_DISARM(tp, TCPT_REXMT); | | 2427 | TCP_TIMER_DISARM(tp, TCPT_REXMT); |
2428 | needoutput = 1; | | 2428 | needoutput = 1; |
2429 | } else if (TCP_TIMER_ISARMED(tp, TCPT_PERSIST) == 0) | | 2429 | } else if (TCP_TIMER_ISARMED(tp, TCPT_PERSIST) == 0) |
2430 | TCP_TIMER_ARM(tp, TCPT_REXMT, tp->t_rxtcur); | | 2430 | TCP_TIMER_ARM(tp, TCPT_REXMT, tp->t_rxtcur); |
2431 | | | 2431 | |
2432 | /* | | 2432 | /* |
2433 | * New data has been acked, adjust the congestion window. | | 2433 | * New data has been acked, adjust the congestion window. |
2434 | */ | | 2434 | */ |
2435 | tp->t_congctl->newack(tp, th); | | 2435 | tp->t_congctl->newack(tp, th); |
2436 | | | 2436 | |
2437 | nd6_hint(tp); | | 2437 | nd6_hint(tp); |
2438 | if (acked > so->so_snd.sb_cc) { | | 2438 | if (acked > so->so_snd.sb_cc) { |
2439 | tp->snd_wnd -= so->so_snd.sb_cc; | | 2439 | tp->snd_wnd -= so->so_snd.sb_cc; |
2440 | sbdrop(&so->so_snd, (int)so->so_snd.sb_cc); | | 2440 | sbdrop(&so->so_snd, (int)so->so_snd.sb_cc); |
2441 | ourfinisacked = 1; | | 2441 | ourfinisacked = 1; |
2442 | } else { | | 2442 | } else { |
2443 | if (acked > (tp->t_lastoff - tp->t_inoff)) | | 2443 | if (acked > (tp->t_lastoff - tp->t_inoff)) |
2444 | tp->t_lastm = NULL; | | 2444 | tp->t_lastm = NULL; |
2445 | sbdrop(&so->so_snd, acked); | | 2445 | sbdrop(&so->so_snd, acked); |
2446 | tp->t_lastoff -= acked; | | 2446 | tp->t_lastoff -= acked; |
2447 | tp->snd_wnd -= acked; | | 2447 | tp->snd_wnd -= acked; |
2448 | ourfinisacked = 0; | | 2448 | ourfinisacked = 0; |
2449 | } | | 2449 | } |
2450 | sowwakeup(so); | | 2450 | sowwakeup(so); |
2451 | | | 2451 | |
2452 | icmp_check(tp, th, acked); | | 2452 | icmp_check(tp, th, acked); |
2453 | | | 2453 | |
2454 | tp->snd_una = th->th_ack; | | 2454 | tp->snd_una = th->th_ack; |
2455 | if (SEQ_GT(tp->snd_una, tp->snd_fack)) | | 2455 | if (SEQ_GT(tp->snd_una, tp->snd_fack)) |
2456 | tp->snd_fack = tp->snd_una; | | 2456 | tp->snd_fack = tp->snd_una; |
2457 | if (SEQ_LT(tp->snd_nxt, tp->snd_una)) | | 2457 | if (SEQ_LT(tp->snd_nxt, tp->snd_una)) |
2458 | tp->snd_nxt = tp->snd_una; | | 2458 | tp->snd_nxt = tp->snd_una; |
2459 | if (SEQ_LT(tp->snd_high, tp->snd_una)) | | 2459 | if (SEQ_LT(tp->snd_high, tp->snd_una)) |
2460 | tp->snd_high = tp->snd_una; | | 2460 | tp->snd_high = tp->snd_una; |
2461 | | | 2461 | |
2462 | switch (tp->t_state) { | | 2462 | switch (tp->t_state) { |
2463 | | | 2463 | |
2464 | /* | | 2464 | /* |
2465 | * In FIN_WAIT_1 STATE in addition to the processing | | 2465 | * In FIN_WAIT_1 STATE in addition to the processing |
2466 | * for the ESTABLISHED state if our FIN is now acknowledged | | 2466 | * for the ESTABLISHED state if our FIN is now acknowledged |
2467 | * then enter FIN_WAIT_2. | | 2467 | * then enter FIN_WAIT_2. |
2468 | */ | | 2468 | */ |
2469 | case TCPS_FIN_WAIT_1: | | 2469 | case TCPS_FIN_WAIT_1: |
2470 | if (ourfinisacked) { | | 2470 | if (ourfinisacked) { |
2471 | /* | | 2471 | /* |
2472 | * If we can't receive any more | | 2472 | * If we can't receive any more |
2473 | * data, then closing user can proceed. | | 2473 | * data, then closing user can proceed. |
2474 | * Starting the timer is contrary to the | | 2474 | * Starting the timer is contrary to the |
2475 | * specification, but if we don't get a FIN | | 2475 | * specification, but if we don't get a FIN |
2476 | * we'll hang forever. | | 2476 | * we'll hang forever. |
2477 | */ | | 2477 | */ |
2478 | if (so->so_state & SS_CANTRCVMORE) { | | 2478 | if (so->so_state & SS_CANTRCVMORE) { |
2479 | soisdisconnected(so); | | 2479 | soisdisconnected(so); |
2480 | if (tp->t_maxidle > 0) | | 2480 | if (tp->t_maxidle > 0) |
2481 | TCP_TIMER_ARM(tp, TCPT_2MSL, | | 2481 | TCP_TIMER_ARM(tp, TCPT_2MSL, |
2482 | tp->t_maxidle); | | 2482 | tp->t_maxidle); |
2483 | } | | 2483 | } |
2484 | tp->t_state = TCPS_FIN_WAIT_2; | | 2484 | tp->t_state = TCPS_FIN_WAIT_2; |
2485 | } | | 2485 | } |
2486 | break; | | 2486 | break; |
2487 | | | 2487 | |
2488 | /* | | 2488 | /* |
2489 | * In CLOSING STATE in addition to the processing for | | 2489 | * In CLOSING STATE in addition to the processing for |
2490 | * the ESTABLISHED state if the ACK acknowledges our FIN | | 2490 | * the ESTABLISHED state if the ACK acknowledges our FIN |
2491 | * then enter the TIME-WAIT state, otherwise ignore | | 2491 | * then enter the TIME-WAIT state, otherwise ignore |
2492 | * the segment. | | 2492 | * the segment. |
2493 | */ | | 2493 | */ |
2494 | case TCPS_CLOSING: | | 2494 | case TCPS_CLOSING: |
2495 | if (ourfinisacked) { | | 2495 | if (ourfinisacked) { |
2496 | tp->t_state = TCPS_TIME_WAIT; | | 2496 | tp->t_state = TCPS_TIME_WAIT; |
2497 | tcp_canceltimers(tp); | | 2497 | tcp_canceltimers(tp); |
2498 | TCP_TIMER_ARM(tp, TCPT_2MSL, 2 * TCPTV_MSL); | | 2498 | TCP_TIMER_ARM(tp, TCPT_2MSL, 2 * TCPTV_MSL); |
2499 | soisdisconnected(so); | | 2499 | soisdisconnected(so); |
2500 | } | | 2500 | } |
2501 | break; | | 2501 | break; |
2502 | | | 2502 | |
2503 | /* | | 2503 | /* |
2504 | * In LAST_ACK, we may still be waiting for data to drain | | 2504 | * In LAST_ACK, we may still be waiting for data to drain |
2505 | * and/or to be acked, as well as for the ack of our FIN. | | 2505 | * and/or to be acked, as well as for the ack of our FIN. |
2506 | * If our FIN is now acknowledged, delete the TCB, | | 2506 | * If our FIN is now acknowledged, delete the TCB, |
2507 | * enter the closed state and return. | | 2507 | * enter the closed state and return. |
2508 | */ | | 2508 | */ |
2509 | case TCPS_LAST_ACK: | | 2509 | case TCPS_LAST_ACK: |
2510 | if (ourfinisacked) { | | 2510 | if (ourfinisacked) { |
2511 | tp = tcp_close(tp); | | 2511 | tp = tcp_close(tp); |
2512 | goto drop; | | 2512 | goto drop; |
2513 | } | | 2513 | } |
2514 | break; | | 2514 | break; |
2515 | | | 2515 | |
2516 | /* | | 2516 | /* |
2517 | * In TIME_WAIT state the only thing that should arrive | | 2517 | * In TIME_WAIT state the only thing that should arrive |
2518 | * is a retransmission of the remote FIN. Acknowledge | | 2518 | * is a retransmission of the remote FIN. Acknowledge |
2519 | * it and restart the finack timer. | | 2519 | * it and restart the finack timer. |
2520 | */ | | 2520 | */ |
2521 | case TCPS_TIME_WAIT: | | 2521 | case TCPS_TIME_WAIT: |
2522 | TCP_TIMER_ARM(tp, TCPT_2MSL, 2 * TCPTV_MSL); | | 2522 | TCP_TIMER_ARM(tp, TCPT_2MSL, 2 * TCPTV_MSL); |
2523 | goto dropafterack; | | 2523 | goto dropafterack; |
2524 | } | | 2524 | } |
2525 | } | | 2525 | } |
2526 | | | 2526 | |
2527 | step6: | | 2527 | step6: |
2528 | /* | | 2528 | /* |
2529 | * Update window information. | | 2529 | * Update window information. |
2530 | * Don't look at window if no ACK: TAC's send garbage on first SYN. | | 2530 | * Don't look at window if no ACK: TAC's send garbage on first SYN. |
2531 | */ | | 2531 | */ |
2532 | if ((tiflags & TH_ACK) && (SEQ_LT(tp->snd_wl1, th->th_seq) || | | 2532 | if ((tiflags & TH_ACK) && (SEQ_LT(tp->snd_wl1, th->th_seq) || |
2533 | (tp->snd_wl1 == th->th_seq && (SEQ_LT(tp->snd_wl2, th->th_ack) || | | 2533 | (tp->snd_wl1 == th->th_seq && (SEQ_LT(tp->snd_wl2, th->th_ack) || |
2534 | (tp->snd_wl2 == th->th_ack && tiwin > tp->snd_wnd))))) { | | 2534 | (tp->snd_wl2 == th->th_ack && tiwin > tp->snd_wnd))))) { |
2535 | /* keep track of pure window updates */ | | 2535 | /* keep track of pure window updates */ |
2536 | if (tlen == 0 && | | 2536 | if (tlen == 0 && |
2537 | tp->snd_wl2 == th->th_ack && tiwin > tp->snd_wnd) | | 2537 | tp->snd_wl2 == th->th_ack && tiwin > tp->snd_wnd) |
2538 | TCP_STATINC(TCP_STAT_RCVWINUPD); | | 2538 | TCP_STATINC(TCP_STAT_RCVWINUPD); |
2539 | tp->snd_wnd = tiwin; | | 2539 | tp->snd_wnd = tiwin; |
2540 | tp->snd_wl1 = th->th_seq; | | 2540 | tp->snd_wl1 = th->th_seq; |
2541 | tp->snd_wl2 = th->th_ack; | | 2541 | tp->snd_wl2 = th->th_ack; |
2542 | if (tp->snd_wnd > tp->max_sndwnd) | | 2542 | if (tp->snd_wnd > tp->max_sndwnd) |
2543 | tp->max_sndwnd = tp->snd_wnd; | | 2543 | tp->max_sndwnd = tp->snd_wnd; |
2544 | needoutput = 1; | | 2544 | needoutput = 1; |
2545 | } | | 2545 | } |
2546 | | | 2546 | |
2547 | /* | | 2547 | /* |
2548 | * Process segments with URG. | | 2548 | * Process segments with URG. |
2549 | */ | | 2549 | */ |
2550 | if ((tiflags & TH_URG) && th->th_urp && | | 2550 | if ((tiflags & TH_URG) && th->th_urp && |
2551 | TCPS_HAVERCVDFIN(tp->t_state) == 0) { | | 2551 | TCPS_HAVERCVDFIN(tp->t_state) == 0) { |
2552 | /* | | 2552 | /* |
2553 | * This is a kludge, but if we receive and accept | | 2553 | * This is a kludge, but if we receive and accept |
2554 | * random urgent pointers, we'll crash in | | 2554 | * random urgent pointers, we'll crash in |
2555 | * soreceive. It's hard to imagine someone | | 2555 | * soreceive. It's hard to imagine someone |
2556 | * actually wanting to send this much urgent data. | | 2556 | * actually wanting to send this much urgent data. |
2557 | */ | | 2557 | */ |
2558 | if (th->th_urp + so->so_rcv.sb_cc > sb_max) { | | 2558 | if (th->th_urp + so->so_rcv.sb_cc > sb_max) { |
2559 | th->th_urp = 0; /* XXX */ | | 2559 | th->th_urp = 0; /* XXX */ |
2560 | tiflags &= ~TH_URG; /* XXX */ | | 2560 | tiflags &= ~TH_URG; /* XXX */ |
2561 | goto dodata; /* XXX */ | | 2561 | goto dodata; /* XXX */ |
2562 | } | | 2562 | } |
2563 | /* | | 2563 | /* |
2564 | * If this segment advances the known urgent pointer, | | 2564 | * If this segment advances the known urgent pointer, |
2565 | * then mark the data stream. This should not happen | | 2565 | * then mark the data stream. This should not happen |
2566 | * in CLOSE_WAIT, CLOSING, LAST_ACK or TIME_WAIT STATES since | | 2566 | * in CLOSE_WAIT, CLOSING, LAST_ACK or TIME_WAIT STATES since |
2567 | * a FIN has been received from the remote side. | | 2567 | * a FIN has been received from the remote side. |
2568 | * In these states we ignore the URG. | | 2568 | * In these states we ignore the URG. |
2569 | * | | 2569 | * |
2570 | * According to RFC961 (Assigned Protocols), | | 2570 | * According to RFC961 (Assigned Protocols), |
2571 | * the urgent pointer points to the last octet | | 2571 | * the urgent pointer points to the last octet |
2572 | * of urgent data. We continue, however, | | 2572 | * of urgent data. We continue, however, |
2573 | * to consider it to indicate the first octet | | 2573 | * to consider it to indicate the first octet |
2574 | * of data past the urgent section as the original | | 2574 | * of data past the urgent section as the original |
2575 | * spec states (in one of two places). | | 2575 | * spec states (in one of two places). |
2576 | */ | | 2576 | */ |
2577 | if (SEQ_GT(th->th_seq+th->th_urp, tp->rcv_up)) { | | 2577 | if (SEQ_GT(th->th_seq+th->th_urp, tp->rcv_up)) { |
2578 | tp->rcv_up = th->th_seq + th->th_urp; | | 2578 | tp->rcv_up = th->th_seq + th->th_urp; |
2579 | so->so_oobmark = so->so_rcv.sb_cc + | | 2579 | so->so_oobmark = so->so_rcv.sb_cc + |
2580 | (tp->rcv_up - tp->rcv_nxt) - 1; | | 2580 | (tp->rcv_up - tp->rcv_nxt) - 1; |
2581 | if (so->so_oobmark == 0) | | 2581 | if (so->so_oobmark == 0) |
2582 | so->so_state |= SS_RCVATMARK; | | 2582 | so->so_state |= SS_RCVATMARK; |
2583 | sohasoutofband(so); | | 2583 | sohasoutofband(so); |
2584 | tp->t_oobflags &= ~(TCPOOB_HAVEDATA | TCPOOB_HADDATA); | | 2584 | tp->t_oobflags &= ~(TCPOOB_HAVEDATA | TCPOOB_HADDATA); |
2585 | } | | 2585 | } |
2586 | /* | | 2586 | /* |
2587 | * Remove out of band data so doesn't get presented to user. | | 2587 | * Remove out of band data so doesn't get presented to user. |
2588 | * This can happen independent of advancing the URG pointer, | | 2588 | * This can happen independent of advancing the URG pointer, |
2589 | * but if two URG's are pending at once, some out-of-band | | 2589 | * but if two URG's are pending at once, some out-of-band |
2590 | * data may creep in... ick. | | 2590 | * data may creep in... ick. |
2591 | */ | | 2591 | */ |
2592 | if (th->th_urp <= (u_int16_t) tlen | | 2592 | if (th->th_urp <= (u_int16_t) tlen |
2593 | #ifdef SO_OOBINLINE | | 2593 | #ifdef SO_OOBINLINE |
2594 | && (so->so_options & SO_OOBINLINE) == 0 | | 2594 | && (so->so_options & SO_OOBINLINE) == 0 |
2595 | #endif | | 2595 | #endif |
2596 | ) | | 2596 | ) |
2597 | tcp_pulloutofband(so, th, m, hdroptlen); | | 2597 | tcp_pulloutofband(so, th, m, hdroptlen); |
2598 | } else | | 2598 | } else |
2599 | /* | | 2599 | /* |
2600 | * If no out of band data is expected, | | 2600 | * If no out of band data is expected, |
2601 | * pull receive urgent pointer along | | 2601 | * pull receive urgent pointer along |
2602 | * with the receive window. | | 2602 | * with the receive window. |
2603 | */ | | 2603 | */ |
2604 | if (SEQ_GT(tp->rcv_nxt, tp->rcv_up)) | | 2604 | if (SEQ_GT(tp->rcv_nxt, tp->rcv_up)) |
2605 | tp->rcv_up = tp->rcv_nxt; | | 2605 | tp->rcv_up = tp->rcv_nxt; |
2606 | dodata: /* XXX */ | | 2606 | dodata: /* XXX */ |
2607 | | | 2607 | |
2608 | /* | | 2608 | /* |
2609 | * Process the segment text, merging it into the TCP sequencing queue, | | 2609 | * Process the segment text, merging it into the TCP sequencing queue, |
2610 | * and arranging for acknowledgement of receipt if necessary. | | 2610 | * and arranging for acknowledgement of receipt if necessary. |
2611 | * This process logically involves adjusting tp->rcv_wnd as data | | 2611 | * This process logically involves adjusting tp->rcv_wnd as data |
2612 | * is presented to the user (this happens in tcp_usrreq.c, | | 2612 | * is presented to the user (this happens in tcp_usrreq.c, |
2613 | * case PRU_RCVD). If a FIN has already been received on this | | 2613 | * case PRU_RCVD). If a FIN has already been received on this |
2614 | * connection then we just ignore the text. | | 2614 | * connection then we just ignore the text. |
2615 | */ | | 2615 | */ |
2616 | if ((tlen || (tiflags & TH_FIN)) && | | 2616 | if ((tlen || (tiflags & TH_FIN)) && |
2617 | TCPS_HAVERCVDFIN(tp->t_state) == 0) { | | 2617 | TCPS_HAVERCVDFIN(tp->t_state) == 0) { |
2618 | /* | | 2618 | /* |
2619 | * Insert segment ti into reassembly queue of tcp with | | 2619 | * Insert segment ti into reassembly queue of tcp with |
2620 | * control block tp. Return TH_FIN if reassembly now includes | | 2620 | * control block tp. Return TH_FIN if reassembly now includes |
2621 | * a segment with FIN. The macro form does the common case | | 2621 | * a segment with FIN. The macro form does the common case |
2622 | * inline (segment is the next to be received on an | | 2622 | * inline (segment is the next to be received on an |
2623 | * established connection, and the queue is empty), | | 2623 | * established connection, and the queue is empty), |
2624 | * avoiding linkage into and removal from the queue and | | 2624 | * avoiding linkage into and removal from the queue and |
2625 | * repetition of various conversions. | | 2625 | * repetition of various conversions. |
2626 | * Set DELACK for segments received in order, but ack | | 2626 | * Set DELACK for segments received in order, but ack |
2627 | * immediately when segments are out of order | | 2627 | * immediately when segments are out of order |
2628 | * (so fast retransmit can work). | | 2628 | * (so fast retransmit can work). |
2629 | */ | | 2629 | */ |
2630 | /* NOTE: this was TCP_REASS() macro, but used only once */ | | 2630 | /* NOTE: this was TCP_REASS() macro, but used only once */ |
2631 | TCP_REASS_LOCK(tp); | | 2631 | TCP_REASS_LOCK(tp); |
2632 | if (th->th_seq == tp->rcv_nxt && | | 2632 | if (th->th_seq == tp->rcv_nxt && |
2633 | TAILQ_FIRST(&tp->segq) == NULL && | | 2633 | TAILQ_FIRST(&tp->segq) == NULL && |
2634 | tp->t_state == TCPS_ESTABLISHED) { | | 2634 | tp->t_state == TCPS_ESTABLISHED) { |
2635 | tcp_setup_ack(tp, th); | | 2635 | tcp_setup_ack(tp, th); |
2636 | tp->rcv_nxt += tlen; | | 2636 | tp->rcv_nxt += tlen; |
2637 | tiflags = th->th_flags & TH_FIN; | | 2637 | tiflags = th->th_flags & TH_FIN; |
2638 | tcps = TCP_STAT_GETREF(); | | 2638 | tcps = TCP_STAT_GETREF(); |
2639 | tcps[TCP_STAT_RCVPACK]++; | | 2639 | tcps[TCP_STAT_RCVPACK]++; |
2640 | tcps[TCP_STAT_RCVBYTE] += tlen; | | 2640 | tcps[TCP_STAT_RCVBYTE] += tlen; |
2641 | TCP_STAT_PUTREF(); | | 2641 | TCP_STAT_PUTREF(); |
2642 | nd6_hint(tp); | | 2642 | nd6_hint(tp); |
2643 | if (so->so_state & SS_CANTRCVMORE) | | 2643 | if (so->so_state & SS_CANTRCVMORE) |
2644 | m_freem(m); | | 2644 | m_freem(m); |
2645 | else { | | 2645 | else { |
2646 | m_adj(m, hdroptlen); | | 2646 | m_adj(m, hdroptlen); |
2647 | sbappendstream(&(so)->so_rcv, m); | | 2647 | sbappendstream(&(so)->so_rcv, m); |
2648 | } | | 2648 | } |
2649 | TCP_REASS_UNLOCK(tp); | | 2649 | TCP_REASS_UNLOCK(tp); |
2650 | sorwakeup(so); | | 2650 | sorwakeup(so); |
2651 | } else { | | 2651 | } else { |
2652 | m_adj(m, hdroptlen); | | 2652 | m_adj(m, hdroptlen); |
2653 | tiflags = tcp_reass(tp, th, m, &tlen); | | 2653 | tiflags = tcp_reass(tp, th, m, &tlen); |
2654 | tp->t_flags |= TF_ACKNOW; | | 2654 | tp->t_flags |= TF_ACKNOW; |
2655 | TCP_REASS_UNLOCK(tp); | | 2655 | TCP_REASS_UNLOCK(tp); |
2656 | } | | 2656 | } |
2657 | | | 2657 | |
2658 | /* | | 2658 | /* |
2659 | * Note the amount of data that peer has sent into | | 2659 | * Note the amount of data that peer has sent into |
2660 | * our window, in order to estimate the sender's | | 2660 | * our window, in order to estimate the sender's |
2661 | * buffer size. | | 2661 | * buffer size. |
2662 | */ | | 2662 | */ |
2663 | len = so->so_rcv.sb_hiwat - (tp->rcv_adv - tp->rcv_nxt); | | 2663 | len = so->so_rcv.sb_hiwat - (tp->rcv_adv - tp->rcv_nxt); |
2664 | } else { | | 2664 | } else { |
2665 | m_freem(m); | | 2665 | m_freem(m); |
2666 | m = NULL; | | 2666 | m = NULL; |
2667 | tiflags &= ~TH_FIN; | | 2667 | tiflags &= ~TH_FIN; |
2668 | } | | 2668 | } |
2669 | | | 2669 | |
2670 | /* | | 2670 | /* |
2671 | * If FIN is received ACK the FIN and let the user know | | 2671 | * If FIN is received ACK the FIN and let the user know |
2672 | * that the connection is closing. Ignore a FIN received before | | 2672 | * that the connection is closing. Ignore a FIN received before |
2673 | * the connection is fully established. | | 2673 | * the connection is fully established. |
2674 | */ | | 2674 | */ |
2675 | if ((tiflags & TH_FIN) && TCPS_HAVEESTABLISHED(tp->t_state)) { | | 2675 | if ((tiflags & TH_FIN) && TCPS_HAVEESTABLISHED(tp->t_state)) { |
2676 | if (TCPS_HAVERCVDFIN(tp->t_state) == 0) { | | 2676 | if (TCPS_HAVERCVDFIN(tp->t_state) == 0) { |
2677 | socantrcvmore(so); | | 2677 | socantrcvmore(so); |
2678 | tp->t_flags |= TF_ACKNOW; | | 2678 | tp->t_flags |= TF_ACKNOW; |
2679 | tp->rcv_nxt++; | | 2679 | tp->rcv_nxt++; |
2680 | } | | 2680 | } |
2681 | switch (tp->t_state) { | | 2681 | switch (tp->t_state) { |
2682 | | | 2682 | |
2683 | /* | | 2683 | /* |
2684 | * In ESTABLISHED STATE enter the CLOSE_WAIT state. | | 2684 | * In ESTABLISHED STATE enter the CLOSE_WAIT state. |
2685 | */ | | 2685 | */ |
2686 | case TCPS_ESTABLISHED: | | 2686 | case TCPS_ESTABLISHED: |
2687 | tp->t_state = TCPS_CLOSE_WAIT; | | 2687 | tp->t_state = TCPS_CLOSE_WAIT; |
2688 | break; | | 2688 | break; |
2689 | | | 2689 | |
2690 | /* | | 2690 | /* |
2691 | * If still in FIN_WAIT_1 STATE FIN has not been acked so | | 2691 | * If still in FIN_WAIT_1 STATE FIN has not been acked so |
2692 | * enter the CLOSING state. | | 2692 | * enter the CLOSING state. |
2693 | */ | | 2693 | */ |
2694 | case TCPS_FIN_WAIT_1: | | 2694 | case TCPS_FIN_WAIT_1: |
2695 | tp->t_state = TCPS_CLOSING; | | 2695 | tp->t_state = TCPS_CLOSING; |
2696 | break; | | 2696 | break; |
2697 | | | 2697 | |
2698 | /* | | 2698 | /* |
2699 | * In FIN_WAIT_2 state enter the TIME_WAIT state, | | 2699 | * In FIN_WAIT_2 state enter the TIME_WAIT state, |
2700 | * starting the time-wait timer, turning off the other | | 2700 | * starting the time-wait timer, turning off the other |
2701 | * standard timers. | | 2701 | * standard timers. |
2702 | */ | | 2702 | */ |
2703 | case TCPS_FIN_WAIT_2: | | 2703 | case TCPS_FIN_WAIT_2: |
2704 | tp->t_state = TCPS_TIME_WAIT; | | 2704 | tp->t_state = TCPS_TIME_WAIT; |
2705 | tcp_canceltimers(tp); | | 2705 | tcp_canceltimers(tp); |
2706 | TCP_TIMER_ARM(tp, TCPT_2MSL, 2 * TCPTV_MSL); | | 2706 | TCP_TIMER_ARM(tp, TCPT_2MSL, 2 * TCPTV_MSL); |
2707 | soisdisconnected(so); | | 2707 | soisdisconnected(so); |
2708 | break; | | 2708 | break; |
2709 | | | 2709 | |
2710 | /* | | 2710 | /* |
2711 | * In TIME_WAIT state restart the 2 MSL time_wait timer. | | 2711 | * In TIME_WAIT state restart the 2 MSL time_wait timer. |
2712 | */ | | 2712 | */ |
2713 | case TCPS_TIME_WAIT: | | 2713 | case TCPS_TIME_WAIT: |
2714 | TCP_TIMER_ARM(tp, TCPT_2MSL, 2 * TCPTV_MSL); | | 2714 | TCP_TIMER_ARM(tp, TCPT_2MSL, 2 * TCPTV_MSL); |
2715 | break; | | 2715 | break; |
2716 | } | | 2716 | } |
2717 | } | | 2717 | } |
2718 | #ifdef TCP_DEBUG | | 2718 | #ifdef TCP_DEBUG |
2719 | if (so->so_options & SO_DEBUG) | | 2719 | if (so->so_options & SO_DEBUG) |
2720 | tcp_trace(TA_INPUT, ostate, tp, tcp_saveti, 0); | | 2720 | tcp_trace(TA_INPUT, ostate, tp, tcp_saveti, 0); |
2721 | #endif | | 2721 | #endif |
2722 | | | 2722 | |
2723 | /* | | 2723 | /* |
2724 | * Return any desired output. | | 2724 | * Return any desired output. |
2725 | */ | | 2725 | */ |
2726 | if (needoutput || (tp->t_flags & TF_ACKNOW)) { | | 2726 | if (needoutput || (tp->t_flags & TF_ACKNOW)) { |
2727 | (void) tcp_output(tp); | | 2727 | (void) tcp_output(tp); |
2728 | } | | 2728 | } |
2729 | if (tcp_saveti) | | 2729 | if (tcp_saveti) |
2730 | m_freem(tcp_saveti); | | 2730 | m_freem(tcp_saveti); |
2731 | return; | | 2731 | return; |
2732 | | | 2732 | |
2733 | badsyn: | | 2733 | badsyn: |
2734 | /* | | 2734 | /* |
2735 | * Received a bad SYN. Increment counters and dropwithreset. | | 2735 | * Received a bad SYN. Increment counters and dropwithreset. |
2736 | */ | | 2736 | */ |
2737 | TCP_STATINC(TCP_STAT_BADSYN); | | 2737 | TCP_STATINC(TCP_STAT_BADSYN); |
2738 | tp = NULL; | | 2738 | tp = NULL; |
2739 | goto dropwithreset; | | 2739 | goto dropwithreset; |
2740 | | | 2740 | |
2741 | dropafterack: | | 2741 | dropafterack: |
2742 | /* | | 2742 | /* |
2743 | * Generate an ACK dropping incoming segment if it occupies | | 2743 | * Generate an ACK dropping incoming segment if it occupies |
2744 | * sequence space, where the ACK reflects our state. | | 2744 | * sequence space, where the ACK reflects our state. |
2745 | */ | | 2745 | */ |
2746 | if (tiflags & TH_RST) | | 2746 | if (tiflags & TH_RST) |
2747 | goto drop; | | 2747 | goto drop; |
2748 | goto dropafterack2; | | 2748 | goto dropafterack2; |
2749 | | | 2749 | |
2750 | dropafterack_ratelim: | | 2750 | dropafterack_ratelim: |
2751 | /* | | 2751 | /* |
2752 | * We may want to rate-limit ACKs against SYN/RST attack. | | 2752 | * We may want to rate-limit ACKs against SYN/RST attack. |
2753 | */ | | 2753 | */ |
2754 | if (ppsratecheck(&tcp_ackdrop_ppslim_last, &tcp_ackdrop_ppslim_count, | | 2754 | if (ppsratecheck(&tcp_ackdrop_ppslim_last, &tcp_ackdrop_ppslim_count, |
2755 | tcp_ackdrop_ppslim) == 0) { | | 2755 | tcp_ackdrop_ppslim) == 0) { |
2756 | /* XXX stat */ | | 2756 | /* XXX stat */ |
2757 | goto drop; | | 2757 | goto drop; |
2758 | } | | 2758 | } |
2759 | /* ...fall into dropafterack2... */ | | 2759 | /* ...fall into dropafterack2... */ |
2760 | | | 2760 | |
2761 | dropafterack2: | | 2761 | dropafterack2: |
2762 | m_freem(m); | | 2762 | m_freem(m); |
2763 | tp->t_flags |= TF_ACKNOW; | | 2763 | tp->t_flags |= TF_ACKNOW; |
2764 | (void) tcp_output(tp); | | 2764 | (void) tcp_output(tp); |
2765 | if (tcp_saveti) | | 2765 | if (tcp_saveti) |
2766 | m_freem(tcp_saveti); | | 2766 | m_freem(tcp_saveti); |
2767 | return; | | 2767 | return; |
2768 | | | 2768 | |
2769 | dropwithreset_ratelim: | | 2769 | dropwithreset_ratelim: |
2770 | /* | | 2770 | /* |
2771 | * We may want to rate-limit RSTs in certain situations, | | 2771 | * We may want to rate-limit RSTs in certain situations, |
2772 | * particularly if we are sending an RST in response to | | 2772 | * particularly if we are sending an RST in response to |
2773 | * an attempt to connect to or otherwise communicate with | | 2773 | * an attempt to connect to or otherwise communicate with |
2774 | * a port for which we have no socket. | | 2774 | * a port for which we have no socket. |
2775 | */ | | 2775 | */ |
2776 | if (ppsratecheck(&tcp_rst_ppslim_last, &tcp_rst_ppslim_count, | | 2776 | if (ppsratecheck(&tcp_rst_ppslim_last, &tcp_rst_ppslim_count, |
2777 | tcp_rst_ppslim) == 0) { | | 2777 | tcp_rst_ppslim) == 0) { |
2778 | /* XXX stat */ | | 2778 | /* XXX stat */ |
2779 | goto drop; | | 2779 | goto drop; |
2780 | } | | 2780 | } |
2781 | /* ...fall into dropwithreset... */ | | 2781 | /* ...fall into dropwithreset... */ |
2782 | | | 2782 | |
2783 | dropwithreset: | | 2783 | dropwithreset: |
2784 | /* | | 2784 | /* |
2785 | * Generate a RST, dropping incoming segment. | | 2785 | * Generate a RST, dropping incoming segment. |
2786 | * Make ACK acceptable to originator of segment. | | 2786 | * Make ACK acceptable to originator of segment. |
2787 | */ | | 2787 | */ |
2788 | if (tiflags & TH_RST) | | 2788 | if (tiflags & TH_RST) |
2789 | goto drop; | | 2789 | goto drop; |
2790 | | | 2790 | |
2791 | switch (af) { | | 2791 | switch (af) { |
2792 | #ifdef INET6 | | 2792 | #ifdef INET6 |
2793 | case AF_INET6: | | 2793 | case AF_INET6: |
2794 | /* For following calls to tcp_respond */ | | 2794 | /* For following calls to tcp_respond */ |
2795 | if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) | | 2795 | if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) |
2796 | goto drop; | | 2796 | goto drop; |
2797 | break; | | 2797 | break; |
2798 | #endif /* INET6 */ | | 2798 | #endif /* INET6 */ |
2799 | case AF_INET: | | 2799 | case AF_INET: |
2800 | if (IN_MULTICAST(ip->ip_dst.s_addr) || | | 2800 | if (IN_MULTICAST(ip->ip_dst.s_addr) || |
2801 | in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif)) | | 2801 | in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif)) |
2802 | goto drop; | | 2802 | goto drop; |
2803 | } | | 2803 | } |
2804 | | | 2804 | |
2805 | if (tiflags & TH_ACK) | | 2805 | if (tiflags & TH_ACK) |
2806 | (void)tcp_respond(tp, m, m, th, (tcp_seq)0, th->th_ack, TH_RST); | | 2806 | (void)tcp_respond(tp, m, m, th, (tcp_seq)0, th->th_ack, TH_RST); |
2807 | else { | | 2807 | else { |
2808 | if (tiflags & TH_SYN) | | 2808 | if (tiflags & TH_SYN) |
2809 | tlen++; | | 2809 | tlen++; |
2810 | (void)tcp_respond(tp, m, m, th, th->th_seq + tlen, (tcp_seq)0, | | 2810 | (void)tcp_respond(tp, m, m, th, th->th_seq + tlen, (tcp_seq)0, |
2811 | TH_RST|TH_ACK); | | 2811 | TH_RST|TH_ACK); |
2812 | } | | 2812 | } |
2813 | if (tcp_saveti) | | 2813 | if (tcp_saveti) |
2814 | m_freem(tcp_saveti); | | 2814 | m_freem(tcp_saveti); |
2815 | return; | | 2815 | return; |
2816 | | | 2816 | |
2817 | badcsum: | | 2817 | badcsum: |
2818 | drop: | | 2818 | drop: |
2819 | /* | | 2819 | /* |
2820 | * Drop space held by incoming segment and return. | | 2820 | * Drop space held by incoming segment and return. |
2821 | */ | | 2821 | */ |
2822 | if (tp) { | | 2822 | if (tp) { |
2823 | if (tp->t_inpcb) | | 2823 | if (tp->t_inpcb) |
2824 | so = tp->t_inpcb->inp_socket; | | 2824 | so = tp->t_inpcb->inp_socket; |
2825 | #ifdef INET6 | | 2825 | #ifdef INET6 |
2826 | else if (tp->t_in6pcb) | | 2826 | else if (tp->t_in6pcb) |
2827 | so = tp->t_in6pcb->in6p_socket; | | 2827 | so = tp->t_in6pcb->in6p_socket; |
2828 | #endif | | 2828 | #endif |
2829 | else | | 2829 | else |
2830 | so = NULL; | | 2830 | so = NULL; |
2831 | #ifdef TCP_DEBUG | | 2831 | #ifdef TCP_DEBUG |
2832 | if (so && (so->so_options & SO_DEBUG) != 0) | | 2832 | if (so && (so->so_options & SO_DEBUG) != 0) |
2833 | tcp_trace(TA_DROP, ostate, tp, tcp_saveti, 0); | | 2833 | tcp_trace(TA_DROP, ostate, tp, tcp_saveti, 0); |
2834 | #endif | | 2834 | #endif |
2835 | } | | 2835 | } |
2836 | if (tcp_saveti) | | 2836 | if (tcp_saveti) |
2837 | m_freem(tcp_saveti); | | 2837 | m_freem(tcp_saveti); |
2838 | m_freem(m); | | 2838 | m_freem(m); |
2839 | return; | | 2839 | return; |
2840 | } | | 2840 | } |
2841 | | | 2841 | |
2842 | #ifdef TCP_SIGNATURE | | 2842 | #ifdef TCP_SIGNATURE |
2843 | int | | 2843 | int |
2844 | tcp_signature_apply(void *fstate, void *data, u_int len) | | 2844 | tcp_signature_apply(void *fstate, void *data, u_int len) |
2845 | { | | 2845 | { |
2846 | | | 2846 | |
2847 | MD5Update(fstate, (u_char *)data, len); | | 2847 | MD5Update(fstate, (u_char *)data, len); |
2848 | return (0); | | 2848 | return (0); |
2849 | } | | 2849 | } |
2850 | | | 2850 | |
2851 | struct secasvar * | | 2851 | struct secasvar * |
2852 | tcp_signature_getsav(struct mbuf *m, struct tcphdr *th) | | 2852 | tcp_signature_getsav(struct mbuf *m, struct tcphdr *th) |
2853 | { | | 2853 | { |
2854 | struct secasvar *sav; | | 2854 | struct secasvar *sav; |
2855 | #ifdef FAST_IPSEC | | 2855 | #ifdef FAST_IPSEC |
2856 | union sockaddr_union dst; | | 2856 | union sockaddr_union dst; |
2857 | #endif | | 2857 | #endif |
2858 | struct ip *ip; | | 2858 | struct ip *ip; |
2859 | struct ip6_hdr *ip6; | | 2859 | struct ip6_hdr *ip6; |
2860 | | | 2860 | |
2861 | ip = mtod(m, struct ip *); | | 2861 | ip = mtod(m, struct ip *); |
2862 | switch (ip->ip_v) { | | 2862 | switch (ip->ip_v) { |
2863 | case 4: | | 2863 | case 4: |
2864 | ip = mtod(m, struct ip *); | | 2864 | ip = mtod(m, struct ip *); |
2865 | ip6 = NULL; | | 2865 | ip6 = NULL; |
2866 | break; | | 2866 | break; |
2867 | case 6: | | 2867 | case 6: |
2868 | ip = NULL; | | 2868 | ip = NULL; |
2869 | ip6 = mtod(m, struct ip6_hdr *); | | 2869 | ip6 = mtod(m, struct ip6_hdr *); |
2870 | break; | | 2870 | break; |
2871 | default: | | 2871 | default: |
2872 | return (NULL); | | 2872 | return (NULL); |
2873 | } | | 2873 | } |
2874 | | | 2874 | |
2875 | #ifdef FAST_IPSEC | | 2875 | #ifdef FAST_IPSEC |
2876 | /* Extract the destination from the IP header in the mbuf. */ | | 2876 | /* Extract the destination from the IP header in the mbuf. */ |
2877 | memset(&dst, 0, sizeof(union sockaddr_union)); | | 2877 | memset(&dst, 0, sizeof(union sockaddr_union)); |
2878 | if (ip !=NULL) { | | 2878 | if (ip !=NULL) { |
2879 | dst.sa.sa_len = sizeof(struct sockaddr_in); | | 2879 | dst.sa.sa_len = sizeof(struct sockaddr_in); |
2880 | dst.sa.sa_family = AF_INET; | | 2880 | dst.sa.sa_family = AF_INET; |
2881 | dst.sin.sin_addr = ip->ip_dst; | | 2881 | dst.sin.sin_addr = ip->ip_dst; |
2882 | } else { | | 2882 | } else { |
2883 | dst.sa.sa_len = sizeof(struct sockaddr_in6); | | 2883 | dst.sa.sa_len = sizeof(struct sockaddr_in6); |
2884 | dst.sa.sa_family = AF_INET6; | | 2884 | dst.sa.sa_family = AF_INET6; |
2885 | dst.sin6.sin6_addr = ip6->ip6_dst; | | 2885 | dst.sin6.sin6_addr = ip6->ip6_dst; |
2886 | } | | 2886 | } |
2887 | | | 2887 | |
2888 | /* | | 2888 | /* |
2889 | * Look up an SADB entry which matches the address of the peer. | | 2889 | * Look up an SADB entry which matches the address of the peer. |
2890 | */ | | 2890 | */ |
2891 | sav = KEY_ALLOCSA(&dst, IPPROTO_TCP, htonl(TCP_SIG_SPI)); | | 2891 | sav = KEY_ALLOCSA(&dst, IPPROTO_TCP, htonl(TCP_SIG_SPI)); |
2892 | #else | | 2892 | #else |
2893 | if (ip) | | 2893 | if (ip) |
2894 | sav = key_allocsa(AF_INET, (void *)&ip->ip_src, | | 2894 | sav = key_allocsa(AF_INET, (void *)&ip->ip_src, |
2895 | (void *)&ip->ip_dst, IPPROTO_TCP, | | 2895 | (void *)&ip->ip_dst, IPPROTO_TCP, |
2896 | htonl(TCP_SIG_SPI), 0, 0); | | 2896 | htonl(TCP_SIG_SPI), 0, 0); |
2897 | else | | 2897 | else |
2898 | sav = key_allocsa(AF_INET6, (void *)&ip6->ip6_src, | | 2898 | sav = key_allocsa(AF_INET6, (void *)&ip6->ip6_src, |
2899 | (void *)&ip6->ip6_dst, IPPROTO_TCP, | | 2899 | (void *)&ip6->ip6_dst, IPPROTO_TCP, |
2900 | htonl(TCP_SIG_SPI), 0, 0); | | 2900 | htonl(TCP_SIG_SPI), 0, 0); |
2901 | #endif | | 2901 | #endif |
2902 | | | 2902 | |
2903 | return (sav); /* freesav must be performed by caller */ | | 2903 | return (sav); /* freesav must be performed by caller */ |
2904 | } | | 2904 | } |
2905 | | | 2905 | |
2906 | int | | 2906 | int |
2907 | tcp_signature(struct mbuf *m, struct tcphdr *th, int thoff, | | 2907 | tcp_signature(struct mbuf *m, struct tcphdr *th, int thoff, |
2908 | struct secasvar *sav, char *sig) | | 2908 | struct secasvar *sav, char *sig) |
2909 | { | | 2909 | { |
2910 | MD5_CTX ctx; | | 2910 | MD5_CTX ctx; |
2911 | struct ip *ip; | | 2911 | struct ip *ip; |
2912 | struct ipovly *ipovly; | | 2912 | struct ipovly *ipovly; |
2913 | struct ip6_hdr *ip6; | | 2913 | struct ip6_hdr *ip6; |
2914 | struct ippseudo ippseudo; | | 2914 | struct ippseudo ippseudo; |
2915 | struct ip6_hdr_pseudo ip6pseudo; | | 2915 | struct ip6_hdr_pseudo ip6pseudo; |
2916 | struct tcphdr th0; | | 2916 | struct tcphdr th0; |
2917 | int l, tcphdrlen; | | 2917 | int l, tcphdrlen; |
2918 | | | 2918 | |
2919 | if (sav == NULL) | | 2919 | if (sav == NULL) |
2920 | return (-1); | | 2920 | return (-1); |
2921 | | | 2921 | |
2922 | tcphdrlen = th->th_off * 4; | | 2922 | tcphdrlen = th->th_off * 4; |
2923 | | | 2923 | |
2924 | switch (mtod(m, struct ip *)->ip_v) { | | 2924 | switch (mtod(m, struct ip *)->ip_v) { |
2925 | case 4: | | 2925 | case 4: |
2926 | ip = mtod(m, struct ip *); | | 2926 | ip = mtod(m, struct ip *); |
2927 | ip6 = NULL; | | 2927 | ip6 = NULL; |
2928 | break; | | 2928 | break; |
2929 | case 6: | | 2929 | case 6: |
2930 | ip = NULL; | | 2930 | ip = NULL; |
2931 | ip6 = mtod(m, struct ip6_hdr *); | | 2931 | ip6 = mtod(m, struct ip6_hdr *); |
2932 | break; | | 2932 | break; |
2933 | default: | | 2933 | default: |
2934 | return (-1); | | 2934 | return (-1); |
2935 | } | | 2935 | } |
2936 | | | 2936 | |
2937 | MD5Init(&ctx); | | 2937 | MD5Init(&ctx); |
2938 | | | 2938 | |
2939 | if (ip) { | | 2939 | if (ip) { |
2940 | memset(&ippseudo, 0, sizeof(ippseudo)); | | 2940 | memset(&ippseudo, 0, sizeof(ippseudo)); |
2941 | ipovly = (struct ipovly *)ip; | | 2941 | ipovly = (struct ipovly *)ip; |
2942 | ippseudo.ippseudo_src = ipovly->ih_src; | | 2942 | ippseudo.ippseudo_src = ipovly->ih_src; |
2943 | ippseudo.ippseudo_dst = ipovly->ih_dst; | | 2943 | ippseudo.ippseudo_dst = ipovly->ih_dst; |
2944 | ippseudo.ippseudo_pad = 0; | | 2944 | ippseudo.ippseudo_pad = 0; |
2945 | ippseudo.ippseudo_p = IPPROTO_TCP; | | 2945 | ippseudo.ippseudo_p = IPPROTO_TCP; |
2946 | ippseudo.ippseudo_len = htons(m->m_pkthdr.len - thoff); | | 2946 | ippseudo.ippseudo_len = htons(m->m_pkthdr.len - thoff); |
2947 | MD5Update(&ctx, (char *)&ippseudo, sizeof(ippseudo)); | | 2947 | MD5Update(&ctx, (char *)&ippseudo, sizeof(ippseudo)); |
2948 | } else { | | 2948 | } else { |
2949 | memset(&ip6pseudo, 0, sizeof(ip6pseudo)); | | 2949 | memset(&ip6pseudo, 0, sizeof(ip6pseudo)); |
2950 | ip6pseudo.ip6ph_src = ip6->ip6_src; | | 2950 | ip6pseudo.ip6ph_src = ip6->ip6_src; |
2951 | in6_clearscope(&ip6pseudo.ip6ph_src); | | 2951 | in6_clearscope(&ip6pseudo.ip6ph_src); |
2952 | ip6pseudo.ip6ph_dst = ip6->ip6_dst; | | 2952 | ip6pseudo.ip6ph_dst = ip6->ip6_dst; |
2953 | in6_clearscope(&ip6pseudo.ip6ph_dst); | | 2953 | in6_clearscope(&ip6pseudo.ip6ph_dst); |
2954 | ip6pseudo.ip6ph_len = htons(m->m_pkthdr.len - thoff); | | 2954 | ip6pseudo.ip6ph_len = htons(m->m_pkthdr.len - thoff); |
2955 | ip6pseudo.ip6ph_nxt = IPPROTO_TCP; | | 2955 | ip6pseudo.ip6ph_nxt = IPPROTO_TCP; |
2956 | MD5Update(&ctx, (char *)&ip6pseudo, sizeof(ip6pseudo)); | | 2956 | MD5Update(&ctx, (char *)&ip6pseudo, sizeof(ip6pseudo)); |
2957 | } | | 2957 | } |
2958 | | | 2958 | |
2959 | th0 = *th; | | 2959 | th0 = *th; |
2960 | th0.th_sum = 0; | | 2960 | th0.th_sum = 0; |
2961 | MD5Update(&ctx, (char *)&th0, sizeof(th0)); | | 2961 | MD5Update(&ctx, (char *)&th0, sizeof(th0)); |
2962 | | | 2962 | |
2963 | l = m->m_pkthdr.len - thoff - tcphdrlen; | | 2963 | l = m->m_pkthdr.len - thoff - tcphdrlen; |
2964 | if (l > 0) | | 2964 | if (l > 0) |
2965 | m_apply(m, thoff + tcphdrlen, | | 2965 | m_apply(m, thoff + tcphdrlen, |
2966 | m->m_pkthdr.len - thoff - tcphdrlen, | | 2966 | m->m_pkthdr.len - thoff - tcphdrlen, |
2967 | tcp_signature_apply, &ctx); | | 2967 | tcp_signature_apply, &ctx); |
2968 | | | 2968 | |
2969 | MD5Update(&ctx, _KEYBUF(sav->key_auth), _KEYLEN(sav->key_auth)); | | 2969 | MD5Update(&ctx, _KEYBUF(sav->key_auth), _KEYLEN(sav->key_auth)); |
2970 | MD5Final(sig, &ctx); | | 2970 | MD5Final(sig, &ctx); |
2971 | | | 2971 | |
2972 | return (0); | | 2972 | return (0); |
2973 | } | | 2973 | } |
2974 | #endif | | 2974 | #endif |
2975 | | | 2975 | |
2976 | static int | | 2976 | static int |
2977 | tcp_dooptions(struct tcpcb *tp, const u_char *cp, int cnt, | | 2977 | tcp_dooptions(struct tcpcb *tp, const u_char *cp, int cnt, |
2978 | struct tcphdr *th, | | 2978 | struct tcphdr *th, |
2979 | struct mbuf *m, int toff, struct tcp_opt_info *oi) | | 2979 | struct mbuf *m, int toff, struct tcp_opt_info *oi) |
2980 | { | | 2980 | { |
2981 | u_int16_t mss; | | 2981 | u_int16_t mss; |
2982 | int opt, optlen = 0; | | 2982 | int opt, optlen = 0; |
2983 | #ifdef TCP_SIGNATURE | | 2983 | #ifdef TCP_SIGNATURE |
2984 | void *sigp = NULL; | | 2984 | void *sigp = NULL; |
2985 | char sigbuf[TCP_SIGLEN]; | | 2985 | char sigbuf[TCP_SIGLEN]; |
2986 | struct secasvar *sav = NULL; | | 2986 | struct secasvar *sav = NULL; |
2987 | #endif | | 2987 | #endif |
2988 | | | 2988 | |
2989 | for (; cp && cnt > 0; cnt -= optlen, cp += optlen) { | | 2989 | for (; cp && cnt > 0; cnt -= optlen, cp += optlen) { |
2990 | opt = cp[0]; | | 2990 | opt = cp[0]; |
2991 | if (opt == TCPOPT_EOL) | | 2991 | if (opt == TCPOPT_EOL) |
2992 | break; | | 2992 | break; |
2993 | if (opt == TCPOPT_NOP) | | 2993 | if (opt == TCPOPT_NOP) |
2994 | optlen = 1; | | 2994 | optlen = 1; |
2995 | else { | | 2995 | else { |
2996 | if (cnt < 2) | | 2996 | if (cnt < 2) |
2997 | break; | | 2997 | break; |
2998 | optlen = cp[1]; | | 2998 | optlen = cp[1]; |
2999 | if (optlen < 2 || optlen > cnt) | | 2999 | if (optlen < 2 || optlen > cnt) |
3000 | break; | | 3000 | break; |
3001 | } | | 3001 | } |
3002 | switch (opt) { | | 3002 | switch (opt) { |
3003 | | | 3003 | |
3004 | default: | | 3004 | default: |
3005 | continue; | | 3005 | continue; |
3006 | | | 3006 | |
3007 | case TCPOPT_MAXSEG: | | 3007 | case TCPOPT_MAXSEG: |
3008 | if (optlen != TCPOLEN_MAXSEG) | | 3008 | if (optlen != TCPOLEN_MAXSEG) |
3009 | continue; | | 3009 | continue; |
3010 | if (!(th->th_flags & TH_SYN)) | | 3010 | if (!(th->th_flags & TH_SYN)) |
3011 | continue; | | 3011 | continue; |
3012 | if (TCPS_HAVERCVDSYN(tp->t_state)) | | 3012 | if (TCPS_HAVERCVDSYN(tp->t_state)) |
3013 | continue; | | 3013 | continue; |
3014 | bcopy(cp + 2, &mss, sizeof(mss)); | | 3014 | bcopy(cp + 2, &mss, sizeof(mss)); |
3015 | oi->maxseg = ntohs(mss); | | 3015 | oi->maxseg = ntohs(mss); |
3016 | break; | | 3016 | break; |
3017 | | | 3017 | |
3018 | case TCPOPT_WINDOW: | | 3018 | case TCPOPT_WINDOW: |
3019 | if (optlen != TCPOLEN_WINDOW) | | 3019 | if (optlen != TCPOLEN_WINDOW) |
3020 | continue; | | 3020 | continue; |
3021 | if (!(th->th_flags & TH_SYN)) | | 3021 | if (!(th->th_flags & TH_SYN)) |
3022 | continue; | | 3022 | continue; |
3023 | if (TCPS_HAVERCVDSYN(tp->t_state)) | | 3023 | if (TCPS_HAVERCVDSYN(tp->t_state)) |
3024 | continue; | | 3024 | continue; |
3025 | tp->t_flags |= TF_RCVD_SCALE; | | 3025 | tp->t_flags |= TF_RCVD_SCALE; |
3026 | tp->requested_s_scale = cp[2]; | | 3026 | tp->requested_s_scale = cp[2]; |
3027 | if (tp->requested_s_scale > TCP_MAX_WINSHIFT) { | | 3027 | if (tp->requested_s_scale > TCP_MAX_WINSHIFT) { |
3028 | #if 0 /*XXX*/ | | 3028 | #if 0 /*XXX*/ |
3029 | char *p; | | 3029 | char *p; |
3030 | | | 3030 | |
3031 | if (ip) | | 3031 | if (ip) |
3032 | p = ntohl(ip->ip_src); | | 3032 | p = ntohl(ip->ip_src); |
3033 | #ifdef INET6 | | 3033 | #ifdef INET6 |
3034 | else if (ip6) | | 3034 | else if (ip6) |
3035 | p = ip6_sprintf(&ip6->ip6_src); | | 3035 | p = ip6_sprintf(&ip6->ip6_src); |
3036 | #endif | | 3036 | #endif |
3037 | else | | 3037 | else |
3038 | p = "(unknown)"; | | 3038 | p = "(unknown)"; |
3039 | log(LOG_ERR, "TCP: invalid wscale %d from %s, " | | 3039 | log(LOG_ERR, "TCP: invalid wscale %d from %s, " |
3040 | "assuming %d\n", | | 3040 | "assuming %d\n", |
3041 | tp->requested_s_scale, p, | | 3041 | tp->requested_s_scale, p, |
3042 | TCP_MAX_WINSHIFT); | | 3042 | TCP_MAX_WINSHIFT); |
3043 | #else | | 3043 | #else |
3044 | log(LOG_ERR, "TCP: invalid wscale %d, " | | 3044 | log(LOG_ERR, "TCP: invalid wscale %d, " |
3045 | "assuming %d\n", | | 3045 | "assuming %d\n", |
3046 | tp->requested_s_scale, | | 3046 | tp->requested_s_scale, |
3047 | TCP_MAX_WINSHIFT); | | 3047 | TCP_MAX_WINSHIFT); |
3048 | #endif | | 3048 | #endif |
3049 | tp->requested_s_scale = TCP_MAX_WINSHIFT; | | 3049 | tp->requested_s_scale = TCP_MAX_WINSHIFT; |
3050 | } | | 3050 | } |
3051 | break; | | 3051 | break; |
3052 | | | 3052 | |
3053 | case TCPOPT_TIMESTAMP: | | 3053 | case TCPOPT_TIMESTAMP: |
3054 | if (optlen != TCPOLEN_TIMESTAMP) | | 3054 | if (optlen != TCPOLEN_TIMESTAMP) |
3055 | continue; | | 3055 | continue; |
3056 | oi->ts_present = 1; | | 3056 | oi->ts_present = 1; |
3057 | bcopy(cp + 2, &oi->ts_val, sizeof(oi->ts_val)); | | 3057 | bcopy(cp + 2, &oi->ts_val, sizeof(oi->ts_val)); |
3058 | NTOHL(oi->ts_val); | | 3058 | NTOHL(oi->ts_val); |
3059 | bcopy(cp + 6, &oi->ts_ecr, sizeof(oi->ts_ecr)); | | 3059 | bcopy(cp + 6, &oi->ts_ecr, sizeof(oi->ts_ecr)); |
3060 | NTOHL(oi->ts_ecr); | | 3060 | NTOHL(oi->ts_ecr); |
3061 | | | 3061 | |
3062 | if (!(th->th_flags & TH_SYN)) | | 3062 | if (!(th->th_flags & TH_SYN)) |
3063 | continue; | | 3063 | continue; |
3064 | if (TCPS_HAVERCVDSYN(tp->t_state)) | | 3064 | if (TCPS_HAVERCVDSYN(tp->t_state)) |
3065 | continue; | | 3065 | continue; |
3066 | /* | | 3066 | /* |
3067 | * A timestamp received in a SYN makes | | 3067 | * A timestamp received in a SYN makes |
3068 | * it ok to send timestamp requests and replies. | | 3068 | * it ok to send timestamp requests and replies. |
3069 | */ | | 3069 | */ |
3070 | tp->t_flags |= TF_RCVD_TSTMP; | | 3070 | tp->t_flags |= TF_RCVD_TSTMP; |
3071 | tp->ts_recent = oi->ts_val; | | 3071 | tp->ts_recent = oi->ts_val; |
3072 | tp->ts_recent_age = tcp_now; | | 3072 | tp->ts_recent_age = tcp_now; |
3073 | break; | | 3073 | break; |
3074 | | | 3074 | |
3075 | case TCPOPT_SACK_PERMITTED: | | 3075 | case TCPOPT_SACK_PERMITTED: |
3076 | if (optlen != TCPOLEN_SACK_PERMITTED) | | 3076 | if (optlen != TCPOLEN_SACK_PERMITTED) |
3077 | continue; | | 3077 | continue; |
3078 | if (!(th->th_flags & TH_SYN)) | | 3078 | if (!(th->th_flags & TH_SYN)) |
3079 | continue; | | 3079 | continue; |
3080 | if (TCPS_HAVERCVDSYN(tp->t_state)) | | 3080 | if (TCPS_HAVERCVDSYN(tp->t_state)) |
3081 | continue; | | 3081 | continue; |
3082 | if (tcp_do_sack) { | | 3082 | if (tcp_do_sack) { |
3083 | tp->t_flags |= TF_SACK_PERMIT; | | 3083 | tp->t_flags |= TF_SACK_PERMIT; |
3084 | tp->t_flags |= TF_WILL_SACK; | | 3084 | tp->t_flags |= TF_WILL_SACK; |
3085 | } | | 3085 | } |
3086 | break; | | 3086 | break; |
3087 | | | 3087 | |
3088 | case TCPOPT_SACK: | | 3088 | case TCPOPT_SACK: |
3089 | tcp_sack_option(tp, th, cp, optlen); | | 3089 | tcp_sack_option(tp, th, cp, optlen); |
3090 | break; | | 3090 | break; |
3091 | #ifdef TCP_SIGNATURE | | 3091 | #ifdef TCP_SIGNATURE |
3092 | case TCPOPT_SIGNATURE: | | 3092 | case TCPOPT_SIGNATURE: |
3093 | if (optlen != TCPOLEN_SIGNATURE) | | 3093 | if (optlen != TCPOLEN_SIGNATURE) |
3094 | continue; | | 3094 | continue; |
3095 | if (sigp && memcmp(sigp, cp + 2, TCP_SIGLEN)) | | 3095 | if (sigp && memcmp(sigp, cp + 2, TCP_SIGLEN)) |
3096 | return (-1); | | 3096 | return (-1); |
3097 | | | 3097 | |
3098 | sigp = sigbuf; | | 3098 | sigp = sigbuf; |
3099 | memcpy(sigbuf, cp + 2, TCP_SIGLEN); | | 3099 | memcpy(sigbuf, cp + 2, TCP_SIGLEN); |
3100 | tp->t_flags |= TF_SIGNATURE; | | 3100 | tp->t_flags |= TF_SIGNATURE; |
3101 | break; | | 3101 | break; |
3102 | #endif | | 3102 | #endif |
3103 | } | | 3103 | } |
3104 | } | | 3104 | } |
3105 | | | 3105 | |
3106 | #ifdef TCP_SIGNATURE | | 3106 | #ifdef TCP_SIGNATURE |
3107 | if (tp->t_flags & TF_SIGNATURE) { | | 3107 | if (tp->t_flags & TF_SIGNATURE) { |
3108 | | | 3108 | |
3109 | sav = tcp_signature_getsav(m, th); | | 3109 | sav = tcp_signature_getsav(m, th); |
3110 | | | 3110 | |
3111 | if (sav == NULL && tp->t_state == TCPS_LISTEN) | | 3111 | if (sav == NULL && tp->t_state == TCPS_LISTEN) |
3112 | return (-1); | | 3112 | return (-1); |
3113 | } | | 3113 | } |
3114 | | | 3114 | |
3115 | if ((sigp ? TF_SIGNATURE : 0) ^ (tp->t_flags & TF_SIGNATURE)) { | | 3115 | if ((sigp ? TF_SIGNATURE : 0) ^ (tp->t_flags & TF_SIGNATURE)) { |
3116 | if (sav == NULL) | | 3116 | if (sav == NULL) |
3117 | return (-1); | | 3117 | return (-1); |
3118 | #ifdef FAST_IPSEC | | 3118 | #ifdef FAST_IPSEC |
3119 | KEY_FREESAV(&sav); | | 3119 | KEY_FREESAV(&sav); |
3120 | #else | | 3120 | #else |
3121 | key_freesav(sav); | | 3121 | key_freesav(sav); |
3122 | #endif | | 3122 | #endif |
3123 | return (-1); | | 3123 | return (-1); |
3124 | } | | 3124 | } |
3125 | | | 3125 | |
3126 | if (sigp) { | | 3126 | if (sigp) { |
3127 | char sig[TCP_SIGLEN]; | | 3127 | char sig[TCP_SIGLEN]; |
3128 | | | 3128 | |
3129 | tcp_fields_to_net(th); | | 3129 | tcp_fields_to_net(th); |
3130 | if (tcp_signature(m, th, toff, sav, sig) < 0) { | | 3130 | if (tcp_signature(m, th, toff, sav, sig) < 0) { |
3131 | tcp_fields_to_host(th); | | 3131 | tcp_fields_to_host(th); |
3132 | if (sav == NULL) | | 3132 | if (sav == NULL) |
3133 | return (-1); | | 3133 | return (-1); |
3134 | #ifdef FAST_IPSEC | | 3134 | #ifdef FAST_IPSEC |
3135 | KEY_FREESAV(&sav); | | 3135 | KEY_FREESAV(&sav); |
3136 | #else | | 3136 | #else |
3137 | key_freesav(sav); | | 3137 | key_freesav(sav); |
3138 | #endif | | 3138 | #endif |
3139 | return (-1); | | 3139 | return (-1); |
3140 | } | | 3140 | } |
3141 | tcp_fields_to_host(th); | | 3141 | tcp_fields_to_host(th); |
3142 | | | 3142 | |
3143 | if (memcmp(sig, sigp, TCP_SIGLEN)) { | | 3143 | if (memcmp(sig, sigp, TCP_SIGLEN)) { |
3144 | TCP_STATINC(TCP_STAT_BADSIG); | | 3144 | TCP_STATINC(TCP_STAT_BADSIG); |
3145 | if (sav == NULL) | | 3145 | if (sav == NULL) |
3146 | return (-1); | | 3146 | return (-1); |
3147 | #ifdef FAST_IPSEC | | 3147 | #ifdef FAST_IPSEC |
3148 | KEY_FREESAV(&sav); | | 3148 | KEY_FREESAV(&sav); |
3149 | #else | | 3149 | #else |
3150 | key_freesav(sav); | | 3150 | key_freesav(sav); |
3151 | #endif | | 3151 | #endif |
3152 | return (-1); | | 3152 | return (-1); |
3153 | } else | | 3153 | } else |
3154 | TCP_STATINC(TCP_STAT_GOODSIG); | | 3154 | TCP_STATINC(TCP_STAT_GOODSIG); |
3155 | | | 3155 | |
3156 | key_sa_recordxfer(sav, m); | | 3156 | key_sa_recordxfer(sav, m); |
3157 | #ifdef FAST_IPSEC | | 3157 | #ifdef FAST_IPSEC |
3158 | KEY_FREESAV(&sav); | | 3158 | KEY_FREESAV(&sav); |
3159 | #else | | 3159 | #else |
3160 | key_freesav(sav); | | 3160 | key_freesav(sav); |
3161 | #endif | | 3161 | #endif |
3162 | } | | 3162 | } |
3163 | #endif | | 3163 | #endif |
3164 | | | 3164 | |
3165 | return (0); | | 3165 | return (0); |
3166 | } | | 3166 | } |
3167 | | | 3167 | |
3168 | /* | | 3168 | /* |
3169 | * Pull out of band byte out of a segment so | | 3169 | * Pull out of band byte out of a segment so |
3170 | * it doesn't appear in the user's data queue. | | 3170 | * it doesn't appear in the user's data queue. |
3171 | * It is still reflected in the segment length for | | 3171 | * It is still reflected in the segment length for |
3172 | * sequencing purposes. | | 3172 | * sequencing purposes. |
3173 | */ | | 3173 | */ |
3174 | void | | 3174 | void |
3175 | tcp_pulloutofband(struct socket *so, struct tcphdr *th, | | 3175 | tcp_pulloutofband(struct socket *so, struct tcphdr *th, |
3176 | struct mbuf *m, int off) | | 3176 | struct mbuf *m, int off) |
3177 | { | | 3177 | { |
3178 | int cnt = off + th->th_urp - 1; | | 3178 | int cnt = off + th->th_urp - 1; |
3179 | | | 3179 | |
3180 | while (cnt >= 0) { | | 3180 | while (cnt >= 0) { |
3181 | if (m->m_len > cnt) { | | 3181 | if (m->m_len > cnt) { |
3182 | char *cp = mtod(m, char *) + cnt; | | 3182 | char *cp = mtod(m, char *) + cnt; |
3183 | struct tcpcb *tp = sototcpcb(so); | | 3183 | struct tcpcb *tp = sototcpcb(so); |
3184 | | | 3184 | |
3185 | tp->t_iobc = *cp; | | 3185 | tp->t_iobc = *cp; |
3186 | tp->t_oobflags |= TCPOOB_HAVEDATA; | | 3186 | tp->t_oobflags |= TCPOOB_HAVEDATA; |
3187 | bcopy(cp+1, cp, (unsigned)(m->m_len - cnt - 1)); | | 3187 | bcopy(cp+1, cp, (unsigned)(m->m_len - cnt - 1)); |
3188 | m->m_len--; | | 3188 | m->m_len--; |
3189 | return; | | 3189 | return; |
3190 | } | | 3190 | } |
3191 | cnt -= m->m_len; | | 3191 | cnt -= m->m_len; |
3192 | m = m->m_next; | | 3192 | m = m->m_next; |
3193 | if (m == 0) | | 3193 | if (m == 0) |
3194 | break; | | 3194 | break; |
3195 | } | | 3195 | } |
3196 | panic("tcp_pulloutofband"); | | 3196 | panic("tcp_pulloutofband"); |
3197 | } | | 3197 | } |
3198 | | | 3198 | |
3199 | /* | | 3199 | /* |
3200 | * Collect new round-trip time estimate | | 3200 | * Collect new round-trip time estimate |
3201 | * and update averages and current timeout. | | 3201 | * and update averages and current timeout. |
3202 | */ | | 3202 | */ |
3203 | void | | 3203 | void |
3204 | tcp_xmit_timer(struct tcpcb *tp, uint32_t rtt) | | 3204 | tcp_xmit_timer(struct tcpcb *tp, uint32_t rtt) |
3205 | { | | 3205 | { |
3206 | int32_t delta; | | 3206 | int32_t delta; |
3207 | | | 3207 | |
3208 | TCP_STATINC(TCP_STAT_RTTUPDATED); | | 3208 | TCP_STATINC(TCP_STAT_RTTUPDATED); |
3209 | if (tp->t_srtt != 0) { | | 3209 | if (tp->t_srtt != 0) { |
3210 | /* | | 3210 | /* |
3211 | * srtt is stored as fixed point with 3 bits after the | | 3211 | * srtt is stored as fixed point with 3 bits after the |
3212 | * binary point (i.e., scaled by 8). The following magic | | 3212 | * binary point (i.e., scaled by 8). The following magic |
3213 | * is equivalent to the smoothing algorithm in rfc793 with | | 3213 | * is equivalent to the smoothing algorithm in rfc793 with |
3214 | * an alpha of .875 (srtt = rtt/8 + srtt*7/8 in fixed | | 3214 | * an alpha of .875 (srtt = rtt/8 + srtt*7/8 in fixed |
3215 | * point). Adjust rtt to origin 0. | | 3215 | * point). Adjust rtt to origin 0. |
3216 | */ | | 3216 | */ |
3217 | delta = (rtt << 2) - (tp->t_srtt >> TCP_RTT_SHIFT); | | 3217 | delta = (rtt << 2) - (tp->t_srtt >> TCP_RTT_SHIFT); |
3218 | if ((tp->t_srtt += delta) <= 0) | | 3218 | if ((tp->t_srtt += delta) <= 0) |
3219 | tp->t_srtt = 1 << 2; | | 3219 | tp->t_srtt = 1 << 2; |
3220 | /* | | 3220 | /* |
3221 | * We accumulate a smoothed rtt variance (actually, a | | 3221 | * We accumulate a smoothed rtt variance (actually, a |
3222 | * smoothed mean difference), then set the retransmit | | 3222 | * smoothed mean difference), then set the retransmit |
3223 | * timer to smoothed rtt + 4 times the smoothed variance. | | 3223 | * timer to smoothed rtt + 4 times the smoothed variance. |
3224 | * rttvar is stored as fixed point with 2 bits after the | | 3224 | * rttvar is stored as fixed point with 2 bits after the |
3225 | * binary point (scaled by 4). The following is | | 3225 | * binary point (scaled by 4). The following is |
3226 | * equivalent to rfc793 smoothing with an alpha of .75 | | 3226 | * equivalent to rfc793 smoothing with an alpha of .75 |
3227 | * (rttvar = rttvar*3/4 + |delta| / 4). This replaces | | 3227 | * (rttvar = rttvar*3/4 + |delta| / 4). This replaces |
3228 | * rfc793's wired-in beta. | | 3228 | * rfc793's wired-in beta. |
3229 | */ | | 3229 | */ |
3230 | if (delta < 0) | | 3230 | if (delta < 0) |
3231 | delta = -delta; | | 3231 | delta = -delta; |
3232 | delta -= (tp->t_rttvar >> TCP_RTTVAR_SHIFT); | | 3232 | delta -= (tp->t_rttvar >> TCP_RTTVAR_SHIFT); |