| @@ -1,2603 +1,2599 @@ | | | @@ -1,2603 +1,2599 @@ |
1 | /* $NetBSD: tcp_input.c,v 1.357.4.3 2018/03/30 11:17:19 martin Exp $ */ | | 1 | /* $NetBSD: tcp_input.c,v 1.357.4.4 2020/07/07 11:56:57 martin 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, | | 73 | * Copyright (c) 1997, 1998, 1999, 2001, 2005, 2006, |
74 | * 2011 The NetBSD Foundation, Inc. | | 74 | * 2011 The NetBSD Foundation, Inc. |
75 | * All rights reserved. | | 75 | * All rights reserved. |
76 | * | | 76 | * |
77 | * This code is derived from software contributed to The NetBSD Foundation | | 77 | * This code is derived from software contributed to The NetBSD Foundation |
78 | * by Coyote Point Systems, Inc. | | 78 | * by Coyote Point Systems, Inc. |
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 Jason R. Thorpe and Kevin M. Lahey of the Numerical Aerospace Simulation | | 80 | * by Jason R. Thorpe and Kevin M. Lahey of the Numerical Aerospace Simulation |
81 | * Facility, NASA Ames Research Center. | | 81 | * Facility, NASA Ames Research Center. |
82 | * This code is derived from software contributed to The NetBSD Foundation | | 82 | * This code is derived from software contributed to The NetBSD Foundation |
83 | * by Charles M. Hannum. | | 83 | * by Charles M. Hannum. |
84 | * This code is derived from software contributed to The NetBSD Foundation | | 84 | * This code is derived from software contributed to The NetBSD Foundation |
85 | * by Rui Paulo. | | 85 | * by Rui Paulo. |
86 | * | | 86 | * |
87 | * Redistribution and use in source and binary forms, with or without | | 87 | * Redistribution and use in source and binary forms, with or without |
88 | * modification, are permitted provided that the following conditions | | 88 | * modification, are permitted provided that the following conditions |
89 | * are met: | | 89 | * are met: |
90 | * 1. Redistributions of source code must retain the above copyright | | 90 | * 1. Redistributions of source code must retain the above copyright |
91 | * notice, this list of conditions and the following disclaimer. | | 91 | * notice, this list of conditions and the following disclaimer. |
92 | * 2. Redistributions in binary form must reproduce the above copyright | | 92 | * 2. Redistributions in binary form must reproduce the above copyright |
93 | * notice, this list of conditions and the following disclaimer in the | | 93 | * notice, this list of conditions and the following disclaimer in the |
94 | * documentation and/or other materials provided with the distribution. | | 94 | * documentation and/or other materials provided with the distribution. |
95 | * | | 95 | * |
96 | * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS | | 96 | * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS |
97 | * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED | | 97 | * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED |
98 | * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR | | 98 | * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR |
99 | * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS | | 99 | * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS |
100 | * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR | | 100 | * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR |
101 | * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF | | 101 | * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF |
102 | * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS | | 102 | * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS |
103 | * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN | | 103 | * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN |
104 | * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) | | 104 | * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) |
105 | * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE | | 105 | * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE |
106 | * POSSIBILITY OF SUCH DAMAGE. | | 106 | * POSSIBILITY OF SUCH DAMAGE. |
107 | */ | | 107 | */ |
108 | | | 108 | |
109 | /* | | 109 | /* |
110 | * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1994, 1995 | | 110 | * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1994, 1995 |
111 | * The Regents of the University of California. All rights reserved. | | 111 | * The Regents of the University of California. All rights reserved. |
112 | * | | 112 | * |
113 | * Redistribution and use in source and binary forms, with or without | | 113 | * Redistribution and use in source and binary forms, with or without |
114 | * modification, are permitted provided that the following conditions | | 114 | * modification, are permitted provided that the following conditions |
115 | * are met: | | 115 | * are met: |
116 | * 1. Redistributions of source code must retain the above copyright | | 116 | * 1. Redistributions of source code must retain the above copyright |
117 | * notice, this list of conditions and the following disclaimer. | | 117 | * notice, this list of conditions and the following disclaimer. |
118 | * 2. Redistributions in binary form must reproduce the above copyright | | 118 | * 2. Redistributions in binary form must reproduce the above copyright |
119 | * notice, this list of conditions and the following disclaimer in the | | 119 | * notice, this list of conditions and the following disclaimer in the |
120 | * documentation and/or other materials provided with the distribution. | | 120 | * documentation and/or other materials provided with the distribution. |
121 | * 3. Neither the name of the University nor the names of its contributors | | 121 | * 3. Neither the name of the University nor the names of its contributors |
122 | * may be used to endorse or promote products derived from this software | | 122 | * may be used to endorse or promote products derived from this software |
123 | * without specific prior written permission. | | 123 | * without specific prior written permission. |
124 | * | | 124 | * |
125 | * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND | | 125 | * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND |
126 | * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE | | 126 | * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE |
127 | * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE | | 127 | * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE |
128 | * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE | | 128 | * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE |
129 | * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL | | 129 | * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL |
130 | * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS | | 130 | * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS |
131 | * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) | | 131 | * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) |
132 | * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT | | 132 | * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT |
133 | * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY | | 133 | * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY |
134 | * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF | | 134 | * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF |
135 | * SUCH DAMAGE. | | 135 | * SUCH DAMAGE. |
136 | * | | 136 | * |
137 | * @(#)tcp_input.c 8.12 (Berkeley) 5/24/95 | | 137 | * @(#)tcp_input.c 8.12 (Berkeley) 5/24/95 |
138 | */ | | 138 | */ |
139 | | | 139 | |
140 | /* | | 140 | /* |
141 | * TODO list for SYN cache stuff: | | 141 | * TODO list for SYN cache stuff: |
142 | * | | 142 | * |
143 | * Find room for a "state" field, which is needed to keep a | | 143 | * Find room for a "state" field, which is needed to keep a |
144 | * compressed state for TIME_WAIT TCBs. It's been noted already | | 144 | * compressed state for TIME_WAIT TCBs. It's been noted already |
145 | * that this is fairly important for very high-volume web and | | 145 | * that this is fairly important for very high-volume web and |
146 | * mail servers, which use a large number of short-lived | | 146 | * mail servers, which use a large number of short-lived |
147 | * connections. | | 147 | * connections. |
148 | */ | | 148 | */ |
149 | | | 149 | |
150 | #include <sys/cdefs.h> | | 150 | #include <sys/cdefs.h> |
151 | __KERNEL_RCSID(0, "$NetBSD: tcp_input.c,v 1.357.4.3 2018/03/30 11:17:19 martin Exp $"); | | 151 | __KERNEL_RCSID(0, "$NetBSD: tcp_input.c,v 1.357.4.4 2020/07/07 11:56:57 martin Exp $"); |
152 | | | 152 | |
153 | #ifdef _KERNEL_OPT | | 153 | #ifdef _KERNEL_OPT |
154 | #include "opt_inet.h" | | 154 | #include "opt_inet.h" |
155 | #include "opt_ipsec.h" | | 155 | #include "opt_ipsec.h" |
156 | #include "opt_inet_csum.h" | | 156 | #include "opt_inet_csum.h" |
157 | #include "opt_tcp_debug.h" | | 157 | #include "opt_tcp_debug.h" |
158 | #endif | | 158 | #endif |
159 | | | 159 | |
160 | #include <sys/param.h> | | 160 | #include <sys/param.h> |
161 | #include <sys/systm.h> | | 161 | #include <sys/systm.h> |
162 | #include <sys/malloc.h> | | 162 | #include <sys/malloc.h> |
163 | #include <sys/mbuf.h> | | 163 | #include <sys/mbuf.h> |
164 | #include <sys/protosw.h> | | 164 | #include <sys/protosw.h> |
165 | #include <sys/socket.h> | | 165 | #include <sys/socket.h> |
166 | #include <sys/socketvar.h> | | 166 | #include <sys/socketvar.h> |
167 | #include <sys/errno.h> | | 167 | #include <sys/errno.h> |
168 | #include <sys/syslog.h> | | 168 | #include <sys/syslog.h> |
169 | #include <sys/pool.h> | | 169 | #include <sys/pool.h> |
170 | #include <sys/domain.h> | | 170 | #include <sys/domain.h> |
171 | #include <sys/kernel.h> | | 171 | #include <sys/kernel.h> |
172 | #ifdef TCP_SIGNATURE | | 172 | #ifdef TCP_SIGNATURE |
173 | #include <sys/md5.h> | | 173 | #include <sys/md5.h> |
174 | #endif | | 174 | #endif |
175 | #include <sys/lwp.h> /* for lwp0 */ | | 175 | #include <sys/lwp.h> /* for lwp0 */ |
176 | #include <sys/cprng.h> | | 176 | #include <sys/cprng.h> |
177 | | | 177 | |
178 | #include <net/if.h> | | 178 | #include <net/if.h> |
179 | #include <net/if_types.h> | | 179 | #include <net/if_types.h> |
180 | | | 180 | |
181 | #include <netinet/in.h> | | 181 | #include <netinet/in.h> |
182 | #include <netinet/in_systm.h> | | 182 | #include <netinet/in_systm.h> |
183 | #include <netinet/ip.h> | | 183 | #include <netinet/ip.h> |
184 | #include <netinet/in_pcb.h> | | 184 | #include <netinet/in_pcb.h> |
185 | #include <netinet/in_var.h> | | 185 | #include <netinet/in_var.h> |
186 | #include <netinet/ip_var.h> | | 186 | #include <netinet/ip_var.h> |
187 | #include <netinet/in_offload.h> | | 187 | #include <netinet/in_offload.h> |
188 | | | 188 | |
189 | #ifdef INET6 | | 189 | #ifdef INET6 |
190 | #ifndef INET | | 190 | #ifndef INET |
191 | #include <netinet/in.h> | | 191 | #include <netinet/in.h> |
192 | #endif | | 192 | #endif |
193 | #include <netinet/ip6.h> | | 193 | #include <netinet/ip6.h> |
194 | #include <netinet6/ip6_var.h> | | 194 | #include <netinet6/ip6_var.h> |
195 | #include <netinet6/in6_pcb.h> | | 195 | #include <netinet6/in6_pcb.h> |
196 | #include <netinet6/ip6_var.h> | | 196 | #include <netinet6/ip6_var.h> |
197 | #include <netinet6/in6_var.h> | | 197 | #include <netinet6/in6_var.h> |
198 | #include <netinet/icmp6.h> | | 198 | #include <netinet/icmp6.h> |
199 | #include <netinet6/nd6.h> | | 199 | #include <netinet6/nd6.h> |
200 | #ifdef TCP_SIGNATURE | | 200 | #ifdef TCP_SIGNATURE |
201 | #include <netinet6/scope6_var.h> | | 201 | #include <netinet6/scope6_var.h> |
202 | #endif | | 202 | #endif |
203 | #endif | | 203 | #endif |
204 | | | 204 | |
205 | #ifndef INET6 | | 205 | #ifndef INET6 |
206 | /* always need ip6.h for IP6_EXTHDR_GET */ | | 206 | /* always need ip6.h for IP6_EXTHDR_GET */ |
207 | #include <netinet/ip6.h> | | 207 | #include <netinet/ip6.h> |
208 | #endif | | 208 | #endif |
209 | | | 209 | |
210 | #include <netinet/tcp.h> | | 210 | #include <netinet/tcp.h> |
211 | #include <netinet/tcp_fsm.h> | | 211 | #include <netinet/tcp_fsm.h> |
212 | #include <netinet/tcp_seq.h> | | 212 | #include <netinet/tcp_seq.h> |
213 | #include <netinet/tcp_timer.h> | | 213 | #include <netinet/tcp_timer.h> |
214 | #include <netinet/tcp_var.h> | | 214 | #include <netinet/tcp_var.h> |
215 | #include <netinet/tcp_private.h> | | 215 | #include <netinet/tcp_private.h> |
216 | #include <netinet/tcpip.h> | | 216 | #include <netinet/tcpip.h> |
217 | #include <netinet/tcp_congctl.h> | | 217 | #include <netinet/tcp_congctl.h> |
218 | #include <netinet/tcp_debug.h> | | 218 | #include <netinet/tcp_debug.h> |
219 | | | 219 | |
220 | #ifdef INET6 | | 220 | #ifdef INET6 |
221 | #include "faith.h" | | 221 | #include "faith.h" |
222 | #if defined(NFAITH) && NFAITH > 0 | | 222 | #if defined(NFAITH) && NFAITH > 0 |
223 | #include <net/if_faith.h> | | 223 | #include <net/if_faith.h> |
224 | #endif | | 224 | #endif |
225 | #endif /* INET6 */ | | 225 | #endif /* INET6 */ |
226 | | | 226 | |
227 | #ifdef IPSEC | | 227 | #ifdef IPSEC |
228 | #include <netipsec/ipsec.h> | | 228 | #include <netipsec/ipsec.h> |
229 | #include <netipsec/ipsec_var.h> | | 229 | #include <netipsec/ipsec_var.h> |
230 | #include <netipsec/ipsec_private.h> | | 230 | #include <netipsec/ipsec_private.h> |
231 | #include <netipsec/key.h> | | 231 | #include <netipsec/key.h> |
232 | #ifdef INET6 | | 232 | #ifdef INET6 |
233 | #include <netipsec/ipsec6.h> | | 233 | #include <netipsec/ipsec6.h> |
234 | #endif | | 234 | #endif |
235 | #endif /* IPSEC*/ | | 235 | #endif /* IPSEC*/ |
236 | | | 236 | |
237 | #include <netinet/tcp_vtw.h> | | 237 | #include <netinet/tcp_vtw.h> |
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 = 1; | | 242 | int tcp_do_autorcvbuf = 1; |
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 | int tcp_msl = (TCPTV_MSL / PR_SLOWHZ); | | 245 | int tcp_msl = (TCPTV_MSL / PR_SLOWHZ); |
246 | | | 246 | |
247 | static int tcp_rst_ppslim_count = 0; | | 247 | static int tcp_rst_ppslim_count = 0; |
248 | static struct timeval tcp_rst_ppslim_last; | | 248 | static struct timeval tcp_rst_ppslim_last; |
249 | static int tcp_ackdrop_ppslim_count = 0; | | 249 | static int tcp_ackdrop_ppslim_count = 0; |
250 | static struct timeval tcp_ackdrop_ppslim_last; | | 250 | static struct timeval tcp_ackdrop_ppslim_last; |
251 | | | 251 | |
252 | #define TCP_PAWS_IDLE (24U * 24 * 60 * 60 * PR_SLOWHZ) | | 252 | #define TCP_PAWS_IDLE (24U * 24 * 60 * 60 * PR_SLOWHZ) |
253 | | | 253 | |
254 | /* for modulo comparisons of timestamps */ | | 254 | /* for modulo comparisons of timestamps */ |
255 | #define TSTMP_LT(a,b) ((int)((a)-(b)) < 0) | | 255 | #define TSTMP_LT(a,b) ((int)((a)-(b)) < 0) |
256 | #define TSTMP_GEQ(a,b) ((int)((a)-(b)) >= 0) | | 256 | #define TSTMP_GEQ(a,b) ((int)((a)-(b)) >= 0) |
257 | | | 257 | |
258 | /* | | 258 | /* |
259 | * Neighbor Discovery, Neighbor Unreachability Detection Upper layer hint. | | 259 | * Neighbor Discovery, Neighbor Unreachability Detection Upper layer hint. |
260 | */ | | 260 | */ |
261 | #ifdef INET6 | | 261 | #ifdef INET6 |
262 | static inline void | | 262 | static inline void |
263 | nd6_hint(struct tcpcb *tp) | | 263 | nd6_hint(struct tcpcb *tp) |
264 | { | | 264 | { |
265 | struct rtentry *rt = NULL; | | 265 | struct rtentry *rt = NULL; |
266 | | | 266 | |
267 | if (tp != NULL && tp->t_in6pcb != NULL && tp->t_family == AF_INET6 && | | 267 | if (tp != NULL && tp->t_in6pcb != NULL && tp->t_family == AF_INET6 && |
268 | (rt = rtcache_validate(&tp->t_in6pcb->in6p_route)) != NULL) | | 268 | (rt = rtcache_validate(&tp->t_in6pcb->in6p_route)) != NULL) |
269 | nd6_nud_hint(rt); | | 269 | nd6_nud_hint(rt); |
270 | rtcache_unref(rt, &tp->t_in6pcb->in6p_route); | | 270 | rtcache_unref(rt, &tp->t_in6pcb->in6p_route); |
271 | } | | 271 | } |
272 | #else | | 272 | #else |
273 | static inline void | | 273 | static inline void |
274 | nd6_hint(struct tcpcb *tp) | | 274 | nd6_hint(struct tcpcb *tp) |
275 | { | | 275 | { |
276 | } | | 276 | } |
277 | #endif | | 277 | #endif |
278 | | | 278 | |
279 | /* | | 279 | /* |
280 | * Compute ACK transmission behavior. Delay the ACK unless | | 280 | * Compute ACK transmission behavior. Delay the ACK unless |
281 | * we have already delayed an ACK (must send an ACK every two segments). | | 281 | * we have already delayed an ACK (must send an ACK every two segments). |
282 | * We also ACK immediately if we received a PUSH and the ACK-on-PUSH | | 282 | * We also ACK immediately if we received a PUSH and the ACK-on-PUSH |
283 | * option is enabled. | | 283 | * option is enabled. |
284 | */ | | 284 | */ |
285 | static void | | 285 | static void |
286 | tcp_setup_ack(struct tcpcb *tp, const struct tcphdr *th) | | 286 | tcp_setup_ack(struct tcpcb *tp, const struct tcphdr *th) |
287 | { | | 287 | { |
288 | | | 288 | |
289 | if (tp->t_flags & TF_DELACK || | | 289 | if (tp->t_flags & TF_DELACK || |
290 | (tcp_ack_on_push && th->th_flags & TH_PUSH)) | | 290 | (tcp_ack_on_push && th->th_flags & TH_PUSH)) |
291 | tp->t_flags |= TF_ACKNOW; | | 291 | tp->t_flags |= TF_ACKNOW; |
292 | else | | 292 | else |
293 | TCP_SET_DELACK(tp); | | 293 | TCP_SET_DELACK(tp); |
294 | } | | 294 | } |
295 | | | 295 | |
296 | static void | | 296 | static void |
297 | icmp_check(struct tcpcb *tp, const struct tcphdr *th, int acked) | | 297 | icmp_check(struct tcpcb *tp, const struct tcphdr *th, int acked) |
298 | { | | 298 | { |
299 | | | 299 | |
300 | /* | | 300 | /* |
301 | * If we had a pending ICMP message that refers to data that have | | 301 | * If we had a pending ICMP message that refers to data that have |
302 | * just been acknowledged, disregard the recorded ICMP message. | | 302 | * just been acknowledged, disregard the recorded ICMP message. |
303 | */ | | 303 | */ |
304 | if ((tp->t_flags & TF_PMTUD_PEND) && | | 304 | if ((tp->t_flags & TF_PMTUD_PEND) && |
305 | SEQ_GT(th->th_ack, tp->t_pmtud_th_seq)) | | 305 | SEQ_GT(th->th_ack, tp->t_pmtud_th_seq)) |
306 | tp->t_flags &= ~TF_PMTUD_PEND; | | 306 | tp->t_flags &= ~TF_PMTUD_PEND; |
307 | | | 307 | |
308 | /* | | 308 | /* |
309 | * Keep track of the largest chunk of data | | 309 | * Keep track of the largest chunk of data |
310 | * acknowledged since last PMTU update | | 310 | * acknowledged since last PMTU update |
311 | */ | | 311 | */ |
312 | if (tp->t_pmtud_mss_acked < acked) | | 312 | if (tp->t_pmtud_mss_acked < acked) |
313 | tp->t_pmtud_mss_acked = acked; | | 313 | tp->t_pmtud_mss_acked = acked; |
314 | } | | 314 | } |
315 | | | 315 | |
316 | /* | | 316 | /* |
317 | * Convert TCP protocol fields to host order for easier processing. | | 317 | * Convert TCP protocol fields to host order for easier processing. |
318 | */ | | 318 | */ |
319 | static void | | 319 | static void |
320 | tcp_fields_to_host(struct tcphdr *th) | | 320 | tcp_fields_to_host(struct tcphdr *th) |
321 | { | | 321 | { |
322 | | | 322 | |
323 | NTOHL(th->th_seq); | | 323 | NTOHL(th->th_seq); |
324 | NTOHL(th->th_ack); | | 324 | NTOHL(th->th_ack); |
325 | NTOHS(th->th_win); | | 325 | NTOHS(th->th_win); |
326 | NTOHS(th->th_urp); | | 326 | NTOHS(th->th_urp); |
327 | } | | 327 | } |
328 | | | 328 | |
329 | /* | | 329 | /* |
330 | * ... and reverse the above. | | 330 | * ... and reverse the above. |
331 | */ | | 331 | */ |
332 | static void | | 332 | static void |
333 | tcp_fields_to_net(struct tcphdr *th) | | 333 | tcp_fields_to_net(struct tcphdr *th) |
334 | { | | 334 | { |
335 | | | 335 | |
336 | HTONL(th->th_seq); | | 336 | HTONL(th->th_seq); |
337 | HTONL(th->th_ack); | | 337 | HTONL(th->th_ack); |
338 | HTONS(th->th_win); | | 338 | HTONS(th->th_win); |
339 | HTONS(th->th_urp); | | 339 | HTONS(th->th_urp); |
340 | } | | 340 | } |
341 | | | 341 | |
342 | #ifdef TCP_CSUM_COUNTERS | | 342 | #ifdef TCP_CSUM_COUNTERS |
343 | #include <sys/device.h> | | 343 | #include <sys/device.h> |
344 | | | 344 | |
345 | #if defined(INET) | | 345 | #if defined(INET) |
346 | extern struct evcnt tcp_hwcsum_ok; | | 346 | extern struct evcnt tcp_hwcsum_ok; |
347 | extern struct evcnt tcp_hwcsum_bad; | | 347 | extern struct evcnt tcp_hwcsum_bad; |
348 | extern struct evcnt tcp_hwcsum_data; | | 348 | extern struct evcnt tcp_hwcsum_data; |
349 | extern struct evcnt tcp_swcsum; | | 349 | extern struct evcnt tcp_swcsum; |
350 | #endif /* defined(INET) */ | | 350 | #endif /* defined(INET) */ |
351 | #if defined(INET6) | | 351 | #if defined(INET6) |
352 | extern struct evcnt tcp6_hwcsum_ok; | | 352 | extern struct evcnt tcp6_hwcsum_ok; |
353 | extern struct evcnt tcp6_hwcsum_bad; | | 353 | extern struct evcnt tcp6_hwcsum_bad; |
354 | extern struct evcnt tcp6_hwcsum_data; | | 354 | extern struct evcnt tcp6_hwcsum_data; |
355 | extern struct evcnt tcp6_swcsum; | | 355 | extern struct evcnt tcp6_swcsum; |
356 | #endif /* defined(INET6) */ | | 356 | #endif /* defined(INET6) */ |
357 | | | 357 | |
358 | #define TCP_CSUM_COUNTER_INCR(ev) (ev)->ev_count++ | | 358 | #define TCP_CSUM_COUNTER_INCR(ev) (ev)->ev_count++ |
359 | | | 359 | |
360 | #else | | 360 | #else |
361 | | | 361 | |
362 | #define TCP_CSUM_COUNTER_INCR(ev) /* nothing */ | | 362 | #define TCP_CSUM_COUNTER_INCR(ev) /* nothing */ |
363 | | | 363 | |
364 | #endif /* TCP_CSUM_COUNTERS */ | | 364 | #endif /* TCP_CSUM_COUNTERS */ |
365 | | | 365 | |
366 | #ifdef TCP_REASS_COUNTERS | | 366 | #ifdef TCP_REASS_COUNTERS |
367 | #include <sys/device.h> | | 367 | #include <sys/device.h> |
368 | | | 368 | |
369 | extern struct evcnt tcp_reass_; | | 369 | extern struct evcnt tcp_reass_; |
370 | extern struct evcnt tcp_reass_empty; | | 370 | extern struct evcnt tcp_reass_empty; |
371 | extern struct evcnt tcp_reass_iteration[8]; | | 371 | extern struct evcnt tcp_reass_iteration[8]; |
372 | extern struct evcnt tcp_reass_prependfirst; | | 372 | extern struct evcnt tcp_reass_prependfirst; |
373 | extern struct evcnt tcp_reass_prepend; | | 373 | extern struct evcnt tcp_reass_prepend; |
374 | extern struct evcnt tcp_reass_insert; | | 374 | extern struct evcnt tcp_reass_insert; |
375 | extern struct evcnt tcp_reass_inserttail; | | 375 | extern struct evcnt tcp_reass_inserttail; |
376 | extern struct evcnt tcp_reass_append; | | 376 | extern struct evcnt tcp_reass_append; |
377 | extern struct evcnt tcp_reass_appendtail; | | 377 | extern struct evcnt tcp_reass_appendtail; |
378 | extern struct evcnt tcp_reass_overlaptail; | | 378 | extern struct evcnt tcp_reass_overlaptail; |
379 | extern struct evcnt tcp_reass_overlapfront; | | 379 | extern struct evcnt tcp_reass_overlapfront; |
380 | extern struct evcnt tcp_reass_segdup; | | 380 | extern struct evcnt tcp_reass_segdup; |
381 | extern struct evcnt tcp_reass_fragdup; | | 381 | extern struct evcnt tcp_reass_fragdup; |
382 | | | 382 | |
383 | #define TCP_REASS_COUNTER_INCR(ev) (ev)->ev_count++ | | 383 | #define TCP_REASS_COUNTER_INCR(ev) (ev)->ev_count++ |
384 | | | 384 | |
385 | #else | | 385 | #else |
386 | | | 386 | |
387 | #define TCP_REASS_COUNTER_INCR(ev) /* nothing */ | | 387 | #define TCP_REASS_COUNTER_INCR(ev) /* nothing */ |
388 | | | 388 | |
389 | #endif /* TCP_REASS_COUNTERS */ | | 389 | #endif /* TCP_REASS_COUNTERS */ |
390 | | | 390 | |
391 | static int tcp_reass(struct tcpcb *, const struct tcphdr *, struct mbuf *, | | 391 | static int tcp_reass(struct tcpcb *, const struct tcphdr *, struct mbuf *, |
392 | int *); | | 392 | int *); |
393 | static int tcp_dooptions(struct tcpcb *, const u_char *, int, | | 393 | static int tcp_dooptions(struct tcpcb *, const u_char *, int, |
394 | struct tcphdr *, struct mbuf *, int, struct tcp_opt_info *); | | 394 | struct tcphdr *, struct mbuf *, int, struct tcp_opt_info *); |
395 | | | 395 | |
396 | #ifdef INET | | 396 | #ifdef INET |
397 | static void tcp4_log_refused(const struct ip *, const struct tcphdr *); | | 397 | static void tcp4_log_refused(const struct ip *, const struct tcphdr *); |
398 | #endif | | 398 | #endif |
399 | #ifdef INET6 | | 399 | #ifdef INET6 |
400 | static void tcp6_log_refused(const struct ip6_hdr *, const struct tcphdr *); | | 400 | static void tcp6_log_refused(const struct ip6_hdr *, const struct tcphdr *); |
401 | #endif | | 401 | #endif |
402 | | | 402 | |
403 | #define TRAVERSE(x) while ((x)->m_next) (x) = (x)->m_next | | 403 | #define TRAVERSE(x) while ((x)->m_next) (x) = (x)->m_next |
404 | | | 404 | |
405 | #if defined(MBUFTRACE) | | 405 | #if defined(MBUFTRACE) |
406 | struct mowner tcp_reass_mowner = MOWNER_INIT("tcp", "reass"); | | 406 | struct mowner tcp_reass_mowner = MOWNER_INIT("tcp", "reass"); |
407 | #endif /* defined(MBUFTRACE) */ | | 407 | #endif /* defined(MBUFTRACE) */ |
408 | | | 408 | |
409 | static struct pool tcpipqent_pool; | | 409 | static struct pool tcpipqent_pool; |
410 | | | 410 | |
411 | void | | 411 | void |
412 | tcpipqent_init(void) | | 412 | tcpipqent_init(void) |
413 | { | | 413 | { |
414 | | | 414 | |
415 | pool_init(&tcpipqent_pool, sizeof(struct ipqent), 0, 0, 0, "tcpipqepl", | | 415 | pool_init(&tcpipqent_pool, sizeof(struct ipqent), 0, 0, 0, "tcpipqepl", |
416 | NULL, IPL_VM); | | 416 | NULL, IPL_VM); |
417 | } | | 417 | } |
418 | | | 418 | |
419 | struct ipqent * | | 419 | struct ipqent * |
420 | tcpipqent_alloc(void) | | 420 | tcpipqent_alloc(void) |
421 | { | | 421 | { |
422 | struct ipqent *ipqe; | | 422 | struct ipqent *ipqe; |
423 | int s; | | 423 | int s; |
424 | | | 424 | |
425 | s = splvm(); | | 425 | s = splvm(); |
426 | ipqe = pool_get(&tcpipqent_pool, PR_NOWAIT); | | 426 | ipqe = pool_get(&tcpipqent_pool, PR_NOWAIT); |
427 | splx(s); | | 427 | splx(s); |
428 | | | 428 | |
429 | return ipqe; | | 429 | return ipqe; |
430 | } | | 430 | } |
431 | | | 431 | |
432 | void | | 432 | void |
433 | tcpipqent_free(struct ipqent *ipqe) | | 433 | tcpipqent_free(struct ipqent *ipqe) |
434 | { | | 434 | { |
435 | int s; | | 435 | int s; |
436 | | | 436 | |
437 | s = splvm(); | | 437 | s = splvm(); |
438 | pool_put(&tcpipqent_pool, ipqe); | | 438 | pool_put(&tcpipqent_pool, ipqe); |
439 | splx(s); | | 439 | splx(s); |
440 | } | | 440 | } |
441 | | | 441 | |
442 | static int | | 442 | static int |
443 | tcp_reass(struct tcpcb *tp, const struct tcphdr *th, struct mbuf *m, int *tlen) | | 443 | tcp_reass(struct tcpcb *tp, const struct tcphdr *th, struct mbuf *m, int *tlen) |
444 | { | | 444 | { |
445 | struct ipqent *p, *q, *nq, *tiqe = NULL; | | 445 | struct ipqent *p, *q, *nq, *tiqe = NULL; |
446 | struct socket *so = NULL; | | 446 | struct socket *so = NULL; |
447 | int pkt_flags; | | 447 | int pkt_flags; |
448 | tcp_seq pkt_seq; | | 448 | tcp_seq pkt_seq; |
449 | unsigned pkt_len; | | 449 | unsigned pkt_len; |
450 | u_long rcvpartdupbyte = 0; | | 450 | u_long rcvpartdupbyte = 0; |
451 | u_long rcvoobyte; | | 451 | u_long rcvoobyte; |
452 | #ifdef TCP_REASS_COUNTERS | | 452 | #ifdef TCP_REASS_COUNTERS |
453 | u_int count = 0; | | 453 | u_int count = 0; |
454 | #endif | | 454 | #endif |
455 | uint64_t *tcps; | | 455 | uint64_t *tcps; |
456 | | | 456 | |
457 | if (tp->t_inpcb) | | 457 | if (tp->t_inpcb) |
458 | so = tp->t_inpcb->inp_socket; | | 458 | so = tp->t_inpcb->inp_socket; |
459 | #ifdef INET6 | | 459 | #ifdef INET6 |
460 | else if (tp->t_in6pcb) | | 460 | else if (tp->t_in6pcb) |
461 | so = tp->t_in6pcb->in6p_socket; | | 461 | so = tp->t_in6pcb->in6p_socket; |
462 | #endif | | 462 | #endif |
463 | | | 463 | |
464 | TCP_REASS_LOCK_CHECK(tp); | | 464 | TCP_REASS_LOCK_CHECK(tp); |
465 | | | 465 | |
466 | /* | | 466 | /* |
467 | * Call with th==0 after become established to | | 467 | * Call with th==0 after become established to |
468 | * force pre-ESTABLISHED data up to user socket. | | 468 | * force pre-ESTABLISHED data up to user socket. |
469 | */ | | 469 | */ |
470 | if (th == 0) | | 470 | if (th == 0) |
471 | goto present; | | 471 | goto present; |
472 | | | 472 | |
473 | m_claimm(m, &tcp_reass_mowner); | | 473 | m_claimm(m, &tcp_reass_mowner); |
474 | | | 474 | |
475 | rcvoobyte = *tlen; | | 475 | rcvoobyte = *tlen; |
476 | /* | | 476 | /* |
477 | * Copy these to local variables because the tcpiphdr | | 477 | * Copy these to local variables because the tcpiphdr |
478 | * gets munged while we are collapsing mbufs. | | 478 | * gets munged while we are collapsing mbufs. |
479 | */ | | 479 | */ |
480 | pkt_seq = th->th_seq; | | 480 | pkt_seq = th->th_seq; |
481 | pkt_len = *tlen; | | 481 | pkt_len = *tlen; |
482 | pkt_flags = th->th_flags; | | 482 | pkt_flags = th->th_flags; |
483 | | | 483 | |
484 | TCP_REASS_COUNTER_INCR(&tcp_reass_); | | 484 | TCP_REASS_COUNTER_INCR(&tcp_reass_); |
485 | | | 485 | |
486 | if ((p = TAILQ_LAST(&tp->segq, ipqehead)) != NULL) { | | 486 | if ((p = TAILQ_LAST(&tp->segq, ipqehead)) != NULL) { |
487 | /* | | 487 | /* |
488 | * When we miss a packet, the vast majority of time we get | | 488 | * When we miss a packet, the vast majority of time we get |
489 | * packets that follow it in order. So optimize for that. | | 489 | * packets that follow it in order. So optimize for that. |
490 | */ | | 490 | */ |
491 | if (pkt_seq == p->ipqe_seq + p->ipqe_len) { | | 491 | if (pkt_seq == p->ipqe_seq + p->ipqe_len) { |
492 | p->ipqe_len += pkt_len; | | 492 | p->ipqe_len += pkt_len; |
493 | p->ipqe_flags |= pkt_flags; | | 493 | p->ipqe_flags |= pkt_flags; |
494 | m_cat(p->ipre_mlast, m); | | 494 | m_cat(p->ipre_mlast, m); |
495 | TRAVERSE(p->ipre_mlast); | | 495 | TRAVERSE(p->ipre_mlast); |
496 | m = NULL; | | 496 | m = NULL; |
497 | tiqe = p; | | 497 | tiqe = p; |
498 | TAILQ_REMOVE(&tp->timeq, p, ipqe_timeq); | | 498 | TAILQ_REMOVE(&tp->timeq, p, ipqe_timeq); |
499 | TCP_REASS_COUNTER_INCR(&tcp_reass_appendtail); | | 499 | TCP_REASS_COUNTER_INCR(&tcp_reass_appendtail); |
500 | goto skip_replacement; | | 500 | goto skip_replacement; |
501 | } | | 501 | } |
502 | /* | | 502 | /* |
503 | * While we're here, if the pkt is completely beyond | | 503 | * While we're here, if the pkt is completely beyond |
504 | * anything we have, just insert it at the tail. | | 504 | * anything we have, just insert it at the tail. |
505 | */ | | 505 | */ |
506 | if (SEQ_GT(pkt_seq, p->ipqe_seq + p->ipqe_len)) { | | 506 | if (SEQ_GT(pkt_seq, p->ipqe_seq + p->ipqe_len)) { |
507 | TCP_REASS_COUNTER_INCR(&tcp_reass_inserttail); | | 507 | TCP_REASS_COUNTER_INCR(&tcp_reass_inserttail); |
508 | goto insert_it; | | 508 | goto insert_it; |
509 | } | | 509 | } |
510 | } | | 510 | } |
511 | | | 511 | |
512 | q = TAILQ_FIRST(&tp->segq); | | 512 | q = TAILQ_FIRST(&tp->segq); |
513 | | | 513 | |
514 | if (q != NULL) { | | 514 | if (q != NULL) { |
515 | /* | | 515 | /* |
516 | * If this segment immediately precedes the first out-of-order | | 516 | * If this segment immediately precedes the first out-of-order |
517 | * block, simply slap the segment in front of it and (mostly) | | 517 | * block, simply slap the segment in front of it and (mostly) |
518 | * skip the complicated logic. | | 518 | * skip the complicated logic. |
519 | */ | | 519 | */ |
520 | if (pkt_seq + pkt_len == q->ipqe_seq) { | | 520 | if (pkt_seq + pkt_len == q->ipqe_seq) { |
521 | q->ipqe_seq = pkt_seq; | | 521 | q->ipqe_seq = pkt_seq; |
522 | q->ipqe_len += pkt_len; | | 522 | q->ipqe_len += pkt_len; |
523 | q->ipqe_flags |= pkt_flags; | | 523 | q->ipqe_flags |= pkt_flags; |
524 | m_cat(m, q->ipqe_m); | | 524 | m_cat(m, q->ipqe_m); |
525 | q->ipqe_m = m; | | 525 | q->ipqe_m = m; |
526 | q->ipre_mlast = m; /* last mbuf may have changed */ | | 526 | q->ipre_mlast = m; /* last mbuf may have changed */ |
527 | TRAVERSE(q->ipre_mlast); | | 527 | TRAVERSE(q->ipre_mlast); |
528 | tiqe = q; | | 528 | tiqe = q; |
529 | TAILQ_REMOVE(&tp->timeq, q, ipqe_timeq); | | 529 | TAILQ_REMOVE(&tp->timeq, q, ipqe_timeq); |
530 | TCP_REASS_COUNTER_INCR(&tcp_reass_prependfirst); | | 530 | TCP_REASS_COUNTER_INCR(&tcp_reass_prependfirst); |
531 | goto skip_replacement; | | 531 | goto skip_replacement; |
532 | } | | 532 | } |
533 | } else { | | 533 | } else { |
534 | TCP_REASS_COUNTER_INCR(&tcp_reass_empty); | | 534 | TCP_REASS_COUNTER_INCR(&tcp_reass_empty); |
535 | } | | 535 | } |
536 | | | 536 | |
537 | /* | | 537 | /* |
538 | * Find a segment which begins after this one does. | | 538 | * Find a segment which begins after this one does. |
539 | */ | | 539 | */ |
540 | for (p = NULL; q != NULL; q = nq) { | | 540 | for (p = NULL; q != NULL; q = nq) { |
541 | nq = TAILQ_NEXT(q, ipqe_q); | | 541 | nq = TAILQ_NEXT(q, ipqe_q); |
542 | #ifdef TCP_REASS_COUNTERS | | 542 | #ifdef TCP_REASS_COUNTERS |
543 | count++; | | 543 | count++; |
544 | #endif | | 544 | #endif |
545 | /* | | 545 | /* |
546 | * If the received segment is just right after this | | 546 | * If the received segment is just right after this |
547 | * fragment, merge the two together and then check | | 547 | * fragment, merge the two together and then check |
548 | * for further overlaps. | | 548 | * for further overlaps. |
549 | */ | | 549 | */ |
550 | if (q->ipqe_seq + q->ipqe_len == pkt_seq) { | | 550 | if (q->ipqe_seq + q->ipqe_len == pkt_seq) { |
551 | #ifdef TCPREASS_DEBUG | | 551 | #ifdef TCPREASS_DEBUG |
552 | printf("tcp_reass[%p]: concat %u:%u(%u) to %u:%u(%u)\n", | | 552 | printf("tcp_reass[%p]: concat %u:%u(%u) to %u:%u(%u)\n", |
553 | tp, pkt_seq, pkt_seq + pkt_len, pkt_len, | | 553 | tp, pkt_seq, pkt_seq + pkt_len, pkt_len, |
554 | q->ipqe_seq, q->ipqe_seq + q->ipqe_len, q->ipqe_len); | | 554 | q->ipqe_seq, q->ipqe_seq + q->ipqe_len, q->ipqe_len); |
555 | #endif | | 555 | #endif |
556 | pkt_len += q->ipqe_len; | | 556 | pkt_len += q->ipqe_len; |
557 | pkt_flags |= q->ipqe_flags; | | 557 | pkt_flags |= q->ipqe_flags; |
558 | pkt_seq = q->ipqe_seq; | | 558 | pkt_seq = q->ipqe_seq; |
559 | m_cat(q->ipre_mlast, m); | | 559 | m_cat(q->ipre_mlast, m); |
560 | TRAVERSE(q->ipre_mlast); | | 560 | TRAVERSE(q->ipre_mlast); |
561 | m = q->ipqe_m; | | 561 | m = q->ipqe_m; |
562 | TCP_REASS_COUNTER_INCR(&tcp_reass_append); | | 562 | TCP_REASS_COUNTER_INCR(&tcp_reass_append); |
563 | goto free_ipqe; | | 563 | goto free_ipqe; |
564 | } | | 564 | } |
565 | /* | | 565 | /* |
566 | * If the received segment is completely past this | | 566 | * If the received segment is completely past this |
567 | * fragment, we need to go the next fragment. | | 567 | * fragment, we need to go the next fragment. |
568 | */ | | 568 | */ |
569 | if (SEQ_LT(q->ipqe_seq + q->ipqe_len, pkt_seq)) { | | 569 | if (SEQ_LT(q->ipqe_seq + q->ipqe_len, pkt_seq)) { |
570 | p = q; | | 570 | p = q; |
571 | continue; | | 571 | continue; |
572 | } | | 572 | } |
573 | /* | | 573 | /* |
574 | * If the fragment is past the received segment, | | 574 | * If the fragment is past the received segment, |
575 | * it (or any following) can't be concatenated. | | 575 | * it (or any following) can't be concatenated. |
576 | */ | | 576 | */ |
577 | if (SEQ_GT(q->ipqe_seq, pkt_seq + pkt_len)) { | | 577 | if (SEQ_GT(q->ipqe_seq, pkt_seq + pkt_len)) { |
578 | TCP_REASS_COUNTER_INCR(&tcp_reass_insert); | | 578 | TCP_REASS_COUNTER_INCR(&tcp_reass_insert); |
579 | break; | | 579 | break; |
580 | } | | 580 | } |
581 | | | 581 | |
582 | /* | | 582 | /* |
583 | * We've received all the data in this segment before. | | 583 | * We've received all the data in this segment before. |
584 | * mark it as a duplicate and return. | | 584 | * mark it as a duplicate and return. |
585 | */ | | 585 | */ |
586 | if (SEQ_LEQ(q->ipqe_seq, pkt_seq) && | | 586 | if (SEQ_LEQ(q->ipqe_seq, pkt_seq) && |
587 | SEQ_GEQ(q->ipqe_seq + q->ipqe_len, pkt_seq + pkt_len)) { | | 587 | SEQ_GEQ(q->ipqe_seq + q->ipqe_len, pkt_seq + pkt_len)) { |
588 | tcps = TCP_STAT_GETREF(); | | 588 | tcps = TCP_STAT_GETREF(); |
589 | tcps[TCP_STAT_RCVDUPPACK]++; | | 589 | tcps[TCP_STAT_RCVDUPPACK]++; |
590 | tcps[TCP_STAT_RCVDUPBYTE] += pkt_len; | | 590 | tcps[TCP_STAT_RCVDUPBYTE] += pkt_len; |
591 | TCP_STAT_PUTREF(); | | 591 | TCP_STAT_PUTREF(); |
592 | tcp_new_dsack(tp, pkt_seq, pkt_len); | | 592 | tcp_new_dsack(tp, pkt_seq, pkt_len); |
593 | m_freem(m); | | 593 | m_freem(m); |
594 | if (tiqe != NULL) { | | 594 | if (tiqe != NULL) { |
595 | tcpipqent_free(tiqe); | | 595 | tcpipqent_free(tiqe); |
596 | } | | 596 | } |
597 | TCP_REASS_COUNTER_INCR(&tcp_reass_segdup); | | 597 | TCP_REASS_COUNTER_INCR(&tcp_reass_segdup); |
598 | goto out; | | 598 | goto out; |
599 | } | | 599 | } |
600 | /* | | 600 | /* |
601 | * Received segment completely overlaps this fragment | | 601 | * Received segment completely overlaps this fragment |
602 | * so we drop the fragment (this keeps the temporal | | 602 | * so we drop the fragment (this keeps the temporal |
603 | * ordering of segments correct). | | 603 | * ordering of segments correct). |
604 | */ | | 604 | */ |
605 | if (SEQ_GEQ(q->ipqe_seq, pkt_seq) && | | 605 | if (SEQ_GEQ(q->ipqe_seq, pkt_seq) && |
606 | SEQ_LEQ(q->ipqe_seq + q->ipqe_len, pkt_seq + pkt_len)) { | | 606 | SEQ_LEQ(q->ipqe_seq + q->ipqe_len, pkt_seq + pkt_len)) { |
607 | rcvpartdupbyte += q->ipqe_len; | | 607 | rcvpartdupbyte += q->ipqe_len; |
608 | m_freem(q->ipqe_m); | | 608 | m_freem(q->ipqe_m); |
609 | TCP_REASS_COUNTER_INCR(&tcp_reass_fragdup); | | 609 | TCP_REASS_COUNTER_INCR(&tcp_reass_fragdup); |
610 | goto free_ipqe; | | 610 | goto free_ipqe; |
611 | } | | 611 | } |
612 | /* | | 612 | /* |
613 | * RX'ed segment extends past the end of the | | 613 | * RX'ed segment extends past the end of the |
614 | * fragment. Drop the overlapping bytes. Then | | 614 | * fragment. Drop the overlapping bytes. Then |
615 | * merge the fragment and segment then treat as | | 615 | * merge the fragment and segment then treat as |
616 | * a longer received packet. | | 616 | * a longer received packet. |
617 | */ | | 617 | */ |
618 | if (SEQ_LT(q->ipqe_seq, pkt_seq) && | | 618 | if (SEQ_LT(q->ipqe_seq, pkt_seq) && |
619 | SEQ_GT(q->ipqe_seq + q->ipqe_len, pkt_seq)) { | | 619 | SEQ_GT(q->ipqe_seq + q->ipqe_len, pkt_seq)) { |
620 | int overlap = q->ipqe_seq + q->ipqe_len - pkt_seq; | | 620 | int overlap = q->ipqe_seq + q->ipqe_len - pkt_seq; |
621 | #ifdef TCPREASS_DEBUG | | 621 | #ifdef TCPREASS_DEBUG |
622 | printf("tcp_reass[%p]: trim starting %d bytes of %u:%u(%u)\n", | | 622 | printf("tcp_reass[%p]: trim starting %d bytes of %u:%u(%u)\n", |
623 | tp, overlap, | | 623 | tp, overlap, |
624 | pkt_seq, pkt_seq + pkt_len, pkt_len); | | 624 | pkt_seq, pkt_seq + pkt_len, pkt_len); |
625 | #endif | | 625 | #endif |
626 | m_adj(m, overlap); | | 626 | m_adj(m, overlap); |
627 | rcvpartdupbyte += overlap; | | 627 | rcvpartdupbyte += overlap; |
628 | m_cat(q->ipre_mlast, m); | | 628 | m_cat(q->ipre_mlast, m); |
629 | TRAVERSE(q->ipre_mlast); | | 629 | TRAVERSE(q->ipre_mlast); |
630 | m = q->ipqe_m; | | 630 | m = q->ipqe_m; |
631 | pkt_seq = q->ipqe_seq; | | 631 | pkt_seq = q->ipqe_seq; |
632 | pkt_len += q->ipqe_len - overlap; | | 632 | pkt_len += q->ipqe_len - overlap; |
633 | rcvoobyte -= overlap; | | 633 | rcvoobyte -= overlap; |
634 | TCP_REASS_COUNTER_INCR(&tcp_reass_overlaptail); | | 634 | TCP_REASS_COUNTER_INCR(&tcp_reass_overlaptail); |
635 | goto free_ipqe; | | 635 | goto free_ipqe; |
636 | } | | 636 | } |
637 | /* | | 637 | /* |
638 | * RX'ed segment extends past the front of the | | 638 | * RX'ed segment extends past the front of the |
639 | * fragment. Drop the overlapping bytes on the | | 639 | * fragment. Drop the overlapping bytes on the |
640 | * received packet. The packet will then be | | 640 | * received packet. The packet will then be |
641 | * contatentated with this fragment a bit later. | | 641 | * contatentated with this fragment a bit later. |
642 | */ | | 642 | */ |
643 | if (SEQ_GT(q->ipqe_seq, pkt_seq) && | | 643 | if (SEQ_GT(q->ipqe_seq, pkt_seq) && |
644 | SEQ_LT(q->ipqe_seq, pkt_seq + pkt_len)) { | | 644 | SEQ_LT(q->ipqe_seq, pkt_seq + pkt_len)) { |
645 | int overlap = pkt_seq + pkt_len - q->ipqe_seq; | | 645 | int overlap = pkt_seq + pkt_len - q->ipqe_seq; |
646 | #ifdef TCPREASS_DEBUG | | 646 | #ifdef TCPREASS_DEBUG |
647 | printf("tcp_reass[%p]: trim trailing %d bytes of %u:%u(%u)\n", | | 647 | printf("tcp_reass[%p]: trim trailing %d bytes of %u:%u(%u)\n", |
648 | tp, overlap, | | 648 | tp, overlap, |
649 | pkt_seq, pkt_seq + pkt_len, pkt_len); | | 649 | pkt_seq, pkt_seq + pkt_len, pkt_len); |
650 | #endif | | 650 | #endif |
651 | m_adj(m, -overlap); | | 651 | m_adj(m, -overlap); |
652 | pkt_len -= overlap; | | 652 | pkt_len -= overlap; |
653 | rcvpartdupbyte += overlap; | | 653 | rcvpartdupbyte += overlap; |
654 | TCP_REASS_COUNTER_INCR(&tcp_reass_overlapfront); | | 654 | TCP_REASS_COUNTER_INCR(&tcp_reass_overlapfront); |
655 | rcvoobyte -= overlap; | | 655 | rcvoobyte -= overlap; |
656 | } | | 656 | } |
657 | /* | | 657 | /* |
658 | * If the received segment immediates precedes this | | 658 | * If the received segment immediates precedes this |
659 | * fragment then tack the fragment onto this segment | | 659 | * fragment then tack the fragment onto this segment |
660 | * and reinsert the data. | | 660 | * and reinsert the data. |
661 | */ | | 661 | */ |
662 | if (q->ipqe_seq == pkt_seq + pkt_len) { | | 662 | if (q->ipqe_seq == pkt_seq + pkt_len) { |
663 | #ifdef TCPREASS_DEBUG | | 663 | #ifdef TCPREASS_DEBUG |
664 | printf("tcp_reass[%p]: append %u:%u(%u) to %u:%u(%u)\n", | | 664 | printf("tcp_reass[%p]: append %u:%u(%u) to %u:%u(%u)\n", |
665 | tp, q->ipqe_seq, q->ipqe_seq + q->ipqe_len, q->ipqe_len, | | 665 | tp, q->ipqe_seq, q->ipqe_seq + q->ipqe_len, q->ipqe_len, |
666 | pkt_seq, pkt_seq + pkt_len, pkt_len); | | 666 | pkt_seq, pkt_seq + pkt_len, pkt_len); |
667 | #endif | | 667 | #endif |
668 | pkt_len += q->ipqe_len; | | 668 | pkt_len += q->ipqe_len; |
669 | pkt_flags |= q->ipqe_flags; | | 669 | pkt_flags |= q->ipqe_flags; |
670 | m_cat(m, q->ipqe_m); | | 670 | m_cat(m, q->ipqe_m); |
671 | TAILQ_REMOVE(&tp->segq, q, ipqe_q); | | 671 | TAILQ_REMOVE(&tp->segq, q, ipqe_q); |
672 | TAILQ_REMOVE(&tp->timeq, q, ipqe_timeq); | | 672 | TAILQ_REMOVE(&tp->timeq, q, ipqe_timeq); |
673 | tp->t_segqlen--; | | 673 | tp->t_segqlen--; |
674 | KASSERT(tp->t_segqlen >= 0); | | 674 | KASSERT(tp->t_segqlen >= 0); |
675 | KASSERT(tp->t_segqlen != 0 || | | 675 | KASSERT(tp->t_segqlen != 0 || |
676 | (TAILQ_EMPTY(&tp->segq) && | | 676 | (TAILQ_EMPTY(&tp->segq) && |
677 | TAILQ_EMPTY(&tp->timeq))); | | 677 | TAILQ_EMPTY(&tp->timeq))); |
678 | if (tiqe == NULL) { | | 678 | if (tiqe == NULL) { |
679 | tiqe = q; | | 679 | tiqe = q; |
680 | } else { | | 680 | } else { |
681 | tcpipqent_free(q); | | 681 | tcpipqent_free(q); |
682 | } | | 682 | } |
683 | TCP_REASS_COUNTER_INCR(&tcp_reass_prepend); | | 683 | TCP_REASS_COUNTER_INCR(&tcp_reass_prepend); |
684 | break; | | 684 | break; |
685 | } | | 685 | } |
686 | /* | | 686 | /* |
687 | * If the fragment is before the segment, remember it. | | 687 | * If the fragment is before the segment, remember it. |
688 | * When this loop is terminated, p will contain the | | 688 | * When this loop is terminated, p will contain the |
689 | * pointer to fragment that is right before the received | | 689 | * pointer to fragment that is right before the received |
690 | * segment. | | 690 | * segment. |
691 | */ | | 691 | */ |
692 | if (SEQ_LEQ(q->ipqe_seq, pkt_seq)) | | 692 | if (SEQ_LEQ(q->ipqe_seq, pkt_seq)) |
693 | p = q; | | 693 | p = q; |
694 | | | 694 | |
695 | continue; | | 695 | continue; |
696 | | | 696 | |
697 | /* | | 697 | /* |
698 | * This is a common operation. It also will allow | | 698 | * This is a common operation. It also will allow |
699 | * to save doing a malloc/free in most instances. | | 699 | * to save doing a malloc/free in most instances. |
700 | */ | | 700 | */ |
701 | free_ipqe: | | 701 | free_ipqe: |
702 | TAILQ_REMOVE(&tp->segq, q, ipqe_q); | | 702 | TAILQ_REMOVE(&tp->segq, q, ipqe_q); |
703 | TAILQ_REMOVE(&tp->timeq, q, ipqe_timeq); | | 703 | TAILQ_REMOVE(&tp->timeq, q, ipqe_timeq); |
704 | tp->t_segqlen--; | | 704 | tp->t_segqlen--; |
705 | KASSERT(tp->t_segqlen >= 0); | | 705 | KASSERT(tp->t_segqlen >= 0); |
706 | KASSERT(tp->t_segqlen != 0 || | | 706 | KASSERT(tp->t_segqlen != 0 || |
707 | (TAILQ_EMPTY(&tp->segq) && TAILQ_EMPTY(&tp->timeq))); | | 707 | (TAILQ_EMPTY(&tp->segq) && TAILQ_EMPTY(&tp->timeq))); |
708 | if (tiqe == NULL) { | | 708 | if (tiqe == NULL) { |
709 | tiqe = q; | | 709 | tiqe = q; |
710 | } else { | | 710 | } else { |
711 | tcpipqent_free(q); | | 711 | tcpipqent_free(q); |
712 | } | | 712 | } |
713 | } | | 713 | } |
714 | | | 714 | |
715 | #ifdef TCP_REASS_COUNTERS | | 715 | #ifdef TCP_REASS_COUNTERS |
716 | if (count > 7) | | 716 | if (count > 7) |
717 | TCP_REASS_COUNTER_INCR(&tcp_reass_iteration[0]); | | 717 | TCP_REASS_COUNTER_INCR(&tcp_reass_iteration[0]); |
718 | else if (count > 0) | | 718 | else if (count > 0) |
719 | TCP_REASS_COUNTER_INCR(&tcp_reass_iteration[count]); | | 719 | TCP_REASS_COUNTER_INCR(&tcp_reass_iteration[count]); |
720 | #endif | | 720 | #endif |
721 | | | 721 | |
722 | insert_it: | | 722 | insert_it: |
723 | | | 723 | |
724 | /* | | 724 | /* |
725 | * Allocate a new queue entry since the received segment did not | | 725 | * Allocate a new queue entry since the received segment did not |
726 | * collapse onto any other out-of-order block; thus we are allocating | | 726 | * collapse onto any other out-of-order block; thus we are allocating |
727 | * a new block. If it had collapsed, tiqe would not be NULL and | | 727 | * a new block. If it had collapsed, tiqe would not be NULL and |
728 | * we would be reusing it. | | 728 | * we would be reusing it. |
729 | * XXX If we can't, just drop the packet. XXX | | 729 | * XXX If we can't, just drop the packet. XXX |
730 | */ | | 730 | */ |
731 | if (tiqe == NULL) { | | 731 | if (tiqe == NULL) { |
732 | tiqe = tcpipqent_alloc(); | | 732 | tiqe = tcpipqent_alloc(); |
733 | if (tiqe == NULL) { | | 733 | if (tiqe == NULL) { |
734 | TCP_STATINC(TCP_STAT_RCVMEMDROP); | | 734 | TCP_STATINC(TCP_STAT_RCVMEMDROP); |
735 | m_freem(m); | | 735 | m_freem(m); |
736 | goto out; | | 736 | goto out; |
737 | } | | 737 | } |
738 | } | | 738 | } |
739 | | | 739 | |
740 | /* | | 740 | /* |
741 | * Update the counters. | | 741 | * Update the counters. |
742 | */ | | 742 | */ |
743 | tp->t_rcvoopack++; | | 743 | tp->t_rcvoopack++; |
744 | tcps = TCP_STAT_GETREF(); | | 744 | tcps = TCP_STAT_GETREF(); |
745 | tcps[TCP_STAT_RCVOOPACK]++; | | 745 | tcps[TCP_STAT_RCVOOPACK]++; |
746 | tcps[TCP_STAT_RCVOOBYTE] += rcvoobyte; | | 746 | tcps[TCP_STAT_RCVOOBYTE] += rcvoobyte; |
747 | if (rcvpartdupbyte) { | | 747 | if (rcvpartdupbyte) { |
748 | tcps[TCP_STAT_RCVPARTDUPPACK]++; | | 748 | tcps[TCP_STAT_RCVPARTDUPPACK]++; |
749 | tcps[TCP_STAT_RCVPARTDUPBYTE] += rcvpartdupbyte; | | 749 | tcps[TCP_STAT_RCVPARTDUPBYTE] += rcvpartdupbyte; |
750 | } | | 750 | } |
751 | TCP_STAT_PUTREF(); | | 751 | TCP_STAT_PUTREF(); |
752 | | | 752 | |
753 | /* | | 753 | /* |
754 | * Insert the new fragment queue entry into both queues. | | 754 | * Insert the new fragment queue entry into both queues. |
755 | */ | | 755 | */ |
756 | tiqe->ipqe_m = m; | | 756 | tiqe->ipqe_m = m; |
757 | tiqe->ipre_mlast = m; | | 757 | tiqe->ipre_mlast = m; |
758 | tiqe->ipqe_seq = pkt_seq; | | 758 | tiqe->ipqe_seq = pkt_seq; |
759 | tiqe->ipqe_len = pkt_len; | | 759 | tiqe->ipqe_len = pkt_len; |
760 | tiqe->ipqe_flags = pkt_flags; | | 760 | tiqe->ipqe_flags = pkt_flags; |
761 | if (p == NULL) { | | 761 | if (p == NULL) { |
762 | TAILQ_INSERT_HEAD(&tp->segq, tiqe, ipqe_q); | | 762 | TAILQ_INSERT_HEAD(&tp->segq, tiqe, ipqe_q); |
763 | #ifdef TCPREASS_DEBUG | | 763 | #ifdef TCPREASS_DEBUG |
764 | if (tiqe->ipqe_seq != tp->rcv_nxt) | | 764 | if (tiqe->ipqe_seq != tp->rcv_nxt) |
765 | printf("tcp_reass[%p]: insert %u:%u(%u) at front\n", | | 765 | printf("tcp_reass[%p]: insert %u:%u(%u) at front\n", |
766 | tp, pkt_seq, pkt_seq + pkt_len, pkt_len); | | 766 | tp, pkt_seq, pkt_seq + pkt_len, pkt_len); |
767 | #endif | | 767 | #endif |
768 | } else { | | 768 | } else { |
769 | TAILQ_INSERT_AFTER(&tp->segq, p, tiqe, ipqe_q); | | 769 | TAILQ_INSERT_AFTER(&tp->segq, p, tiqe, ipqe_q); |
770 | #ifdef TCPREASS_DEBUG | | 770 | #ifdef TCPREASS_DEBUG |
771 | printf("tcp_reass[%p]: insert %u:%u(%u) after %u:%u(%u)\n", | | 771 | printf("tcp_reass[%p]: insert %u:%u(%u) after %u:%u(%u)\n", |
772 | tp, pkt_seq, pkt_seq + pkt_len, pkt_len, | | 772 | tp, pkt_seq, pkt_seq + pkt_len, pkt_len, |
773 | p->ipqe_seq, p->ipqe_seq + p->ipqe_len, p->ipqe_len); | | 773 | p->ipqe_seq, p->ipqe_seq + p->ipqe_len, p->ipqe_len); |
774 | #endif | | 774 | #endif |
775 | } | | 775 | } |
776 | tp->t_segqlen++; | | 776 | tp->t_segqlen++; |
777 | | | 777 | |
778 | skip_replacement: | | 778 | skip_replacement: |
779 | | | 779 | |
780 | TAILQ_INSERT_HEAD(&tp->timeq, tiqe, ipqe_timeq); | | 780 | TAILQ_INSERT_HEAD(&tp->timeq, tiqe, ipqe_timeq); |
781 | | | 781 | |
782 | present: | | 782 | present: |
783 | /* | | 783 | /* |
784 | * Present data to user, advancing rcv_nxt through | | 784 | * Present data to user, advancing rcv_nxt through |
785 | * completed sequence space. | | 785 | * completed sequence space. |
786 | */ | | 786 | */ |
787 | if (TCPS_HAVEESTABLISHED(tp->t_state) == 0) | | 787 | if (TCPS_HAVEESTABLISHED(tp->t_state) == 0) |
788 | goto out; | | 788 | goto out; |
789 | q = TAILQ_FIRST(&tp->segq); | | 789 | q = TAILQ_FIRST(&tp->segq); |
790 | if (q == NULL || q->ipqe_seq != tp->rcv_nxt) | | 790 | if (q == NULL || q->ipqe_seq != tp->rcv_nxt) |
791 | goto out; | | 791 | goto out; |
792 | if (tp->t_state == TCPS_SYN_RECEIVED && q->ipqe_len) | | 792 | if (tp->t_state == TCPS_SYN_RECEIVED && q->ipqe_len) |
793 | goto out; | | 793 | goto out; |
794 | | | 794 | |
795 | tp->rcv_nxt += q->ipqe_len; | | 795 | tp->rcv_nxt += q->ipqe_len; |
796 | pkt_flags = q->ipqe_flags & TH_FIN; | | 796 | pkt_flags = q->ipqe_flags & TH_FIN; |
797 | nd6_hint(tp); | | 797 | nd6_hint(tp); |
798 | | | 798 | |
799 | TAILQ_REMOVE(&tp->segq, q, ipqe_q); | | 799 | TAILQ_REMOVE(&tp->segq, q, ipqe_q); |
800 | TAILQ_REMOVE(&tp->timeq, q, ipqe_timeq); | | 800 | TAILQ_REMOVE(&tp->timeq, q, ipqe_timeq); |
801 | tp->t_segqlen--; | | 801 | tp->t_segqlen--; |
802 | KASSERT(tp->t_segqlen >= 0); | | 802 | KASSERT(tp->t_segqlen >= 0); |
803 | KASSERT(tp->t_segqlen != 0 || | | 803 | KASSERT(tp->t_segqlen != 0 || |
804 | (TAILQ_EMPTY(&tp->segq) && TAILQ_EMPTY(&tp->timeq))); | | 804 | (TAILQ_EMPTY(&tp->segq) && TAILQ_EMPTY(&tp->timeq))); |
805 | if (so->so_state & SS_CANTRCVMORE) | | 805 | if (so->so_state & SS_CANTRCVMORE) |
806 | m_freem(q->ipqe_m); | | 806 | m_freem(q->ipqe_m); |
807 | else | | 807 | else |
808 | sbappendstream(&so->so_rcv, q->ipqe_m); | | 808 | sbappendstream(&so->so_rcv, q->ipqe_m); |
809 | tcpipqent_free(q); | | 809 | tcpipqent_free(q); |
810 | TCP_REASS_UNLOCK(tp); | | 810 | TCP_REASS_UNLOCK(tp); |
811 | sorwakeup(so); | | 811 | sorwakeup(so); |
812 | return (pkt_flags); | | 812 | return (pkt_flags); |
813 | out: | | 813 | out: |
814 | TCP_REASS_UNLOCK(tp); | | 814 | TCP_REASS_UNLOCK(tp); |
815 | return (0); | | 815 | return (0); |
816 | } | | 816 | } |
817 | | | 817 | |
818 | #ifdef INET6 | | 818 | #ifdef INET6 |
819 | int | | 819 | int |
820 | tcp6_input(struct mbuf **mp, int *offp, int proto) | | 820 | tcp6_input(struct mbuf **mp, int *offp, int proto) |
821 | { | | 821 | { |
822 | struct mbuf *m = *mp; | | 822 | struct mbuf *m = *mp; |
823 | | | 823 | |
824 | /* | | 824 | /* |
825 | * draft-itojun-ipv6-tcp-to-anycast | | 825 | * draft-itojun-ipv6-tcp-to-anycast |
826 | * better place to put this in? | | 826 | * better place to put this in? |
827 | */ | | 827 | */ |
828 | if (m->m_flags & M_ANYCAST6) { | | 828 | if (m->m_flags & M_ANYCAST6) { |
829 | struct ip6_hdr *ip6; | | 829 | struct ip6_hdr *ip6; |
830 | if (m->m_len < sizeof(struct ip6_hdr)) { | | 830 | if (m->m_len < sizeof(struct ip6_hdr)) { |
831 | if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) { | | 831 | if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) { |
832 | TCP_STATINC(TCP_STAT_RCVSHORT); | | 832 | TCP_STATINC(TCP_STAT_RCVSHORT); |
833 | return IPPROTO_DONE; | | 833 | return IPPROTO_DONE; |
834 | } | | 834 | } |
835 | } | | 835 | } |
836 | ip6 = mtod(m, struct ip6_hdr *); | | 836 | ip6 = mtod(m, struct ip6_hdr *); |
837 | icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_ADDR, | | 837 | icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_ADDR, |
838 | (char *)&ip6->ip6_dst - (char *)ip6); | | 838 | (char *)&ip6->ip6_dst - (char *)ip6); |
839 | return IPPROTO_DONE; | | 839 | return IPPROTO_DONE; |
840 | } | | 840 | } |
841 | | | 841 | |
842 | tcp_input(m, *offp, proto); | | 842 | tcp_input(m, *offp, proto); |
843 | return IPPROTO_DONE; | | 843 | return IPPROTO_DONE; |
844 | } | | 844 | } |
845 | #endif | | 845 | #endif |
846 | | | 846 | |
847 | #ifdef INET | | 847 | #ifdef INET |
848 | static void | | 848 | static void |
849 | tcp4_log_refused(const struct ip *ip, const struct tcphdr *th) | | 849 | tcp4_log_refused(const struct ip *ip, const struct tcphdr *th) |
850 | { | | 850 | { |
851 | char src[INET_ADDRSTRLEN]; | | 851 | char src[INET_ADDRSTRLEN]; |
852 | char dst[INET_ADDRSTRLEN]; | | 852 | char dst[INET_ADDRSTRLEN]; |
853 | | | 853 | |
854 | if (ip) { | | 854 | if (ip) { |
855 | in_print(src, sizeof(src), &ip->ip_src); | | 855 | in_print(src, sizeof(src), &ip->ip_src); |
856 | in_print(dst, sizeof(dst), &ip->ip_dst); | | 856 | in_print(dst, sizeof(dst), &ip->ip_dst); |
857 | } | | 857 | } |
858 | else { | | 858 | else { |
859 | strlcpy(src, "(unknown)", sizeof(src)); | | 859 | strlcpy(src, "(unknown)", sizeof(src)); |
860 | strlcpy(dst, "(unknown)", sizeof(dst)); | | 860 | strlcpy(dst, "(unknown)", sizeof(dst)); |
861 | } | | 861 | } |
862 | log(LOG_INFO, | | 862 | log(LOG_INFO, |
863 | "Connection attempt to TCP %s:%d from %s:%d\n", | | 863 | "Connection attempt to TCP %s:%d from %s:%d\n", |
864 | dst, ntohs(th->th_dport), | | 864 | dst, ntohs(th->th_dport), |
865 | src, ntohs(th->th_sport)); | | 865 | src, ntohs(th->th_sport)); |
866 | } | | 866 | } |
867 | #endif | | 867 | #endif |
868 | | | 868 | |
869 | #ifdef INET6 | | 869 | #ifdef INET6 |
870 | static void | | 870 | static void |
871 | tcp6_log_refused(const struct ip6_hdr *ip6, const struct tcphdr *th) | | 871 | tcp6_log_refused(const struct ip6_hdr *ip6, const struct tcphdr *th) |
872 | { | | 872 | { |
873 | char src[INET6_ADDRSTRLEN]; | | 873 | char src[INET6_ADDRSTRLEN]; |
874 | char dst[INET6_ADDRSTRLEN]; | | 874 | char dst[INET6_ADDRSTRLEN]; |
875 | | | 875 | |
876 | if (ip6) { | | 876 | if (ip6) { |
877 | in6_print(src, sizeof(src), &ip6->ip6_src); | | 877 | in6_print(src, sizeof(src), &ip6->ip6_src); |
878 | in6_print(dst, sizeof(dst), &ip6->ip6_dst); | | 878 | in6_print(dst, sizeof(dst), &ip6->ip6_dst); |
879 | } | | 879 | } |
880 | else { | | 880 | else { |
881 | strlcpy(src, "(unknown v6)", sizeof(src)); | | 881 | strlcpy(src, "(unknown v6)", sizeof(src)); |
882 | strlcpy(dst, "(unknown v6)", sizeof(dst)); | | 882 | strlcpy(dst, "(unknown v6)", sizeof(dst)); |
883 | } | | 883 | } |
884 | log(LOG_INFO, | | 884 | log(LOG_INFO, |
885 | "Connection attempt to TCP [%s]:%d from [%s]:%d\n", | | 885 | "Connection attempt to TCP [%s]:%d from [%s]:%d\n", |
886 | dst, ntohs(th->th_dport), | | 886 | dst, ntohs(th->th_dport), |
887 | src, ntohs(th->th_sport)); | | 887 | src, ntohs(th->th_sport)); |
888 | } | | 888 | } |
889 | #endif | | 889 | #endif |
890 | | | 890 | |
891 | /* | | 891 | /* |
892 | * Checksum extended TCP header and data. | | 892 | * Checksum extended TCP header and data. |
893 | */ | | 893 | */ |
894 | int | | 894 | int |
895 | tcp_input_checksum(int af, struct mbuf *m, const struct tcphdr *th, | | 895 | tcp_input_checksum(int af, struct mbuf *m, const struct tcphdr *th, |
896 | int toff, int off, int tlen) | | 896 | int toff, int off, int tlen) |
897 | { | | 897 | { |
898 | struct ifnet *rcvif; | | 898 | struct ifnet *rcvif; |
899 | int s; | | 899 | int s; |
900 | | | 900 | |
901 | /* | | 901 | /* |
902 | * XXX it's better to record and check if this mbuf is | | 902 | * XXX it's better to record and check if this mbuf is |
903 | * already checked. | | 903 | * already checked. |
904 | */ | | 904 | */ |
905 | | | 905 | |
906 | rcvif = m_get_rcvif(m, &s); | | 906 | rcvif = m_get_rcvif(m, &s); |
907 | if (__predict_false(rcvif == NULL)) | | 907 | if (__predict_false(rcvif == NULL)) |
908 | goto badcsum; /* XXX */ | | 908 | goto badcsum; /* XXX */ |
909 | | | 909 | |
910 | switch (af) { | | 910 | switch (af) { |
911 | #ifdef INET | | 911 | #ifdef INET |
912 | case AF_INET: | | 912 | case AF_INET: |
913 | switch (m->m_pkthdr.csum_flags & | | 913 | switch (m->m_pkthdr.csum_flags & |
914 | ((rcvif->if_csum_flags_rx & M_CSUM_TCPv4) | | | 914 | ((rcvif->if_csum_flags_rx & M_CSUM_TCPv4) | |
915 | M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) { | | 915 | M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) { |
916 | case M_CSUM_TCPv4|M_CSUM_TCP_UDP_BAD: | | 916 | case M_CSUM_TCPv4|M_CSUM_TCP_UDP_BAD: |
917 | TCP_CSUM_COUNTER_INCR(&tcp_hwcsum_bad); | | 917 | TCP_CSUM_COUNTER_INCR(&tcp_hwcsum_bad); |
918 | goto badcsum; | | 918 | goto badcsum; |
919 | | | 919 | |
920 | case M_CSUM_TCPv4|M_CSUM_DATA: { | | 920 | case M_CSUM_TCPv4|M_CSUM_DATA: { |
921 | u_int32_t hw_csum = m->m_pkthdr.csum_data; | | 921 | u_int32_t hw_csum = m->m_pkthdr.csum_data; |
922 | | | 922 | |
923 | TCP_CSUM_COUNTER_INCR(&tcp_hwcsum_data); | | 923 | TCP_CSUM_COUNTER_INCR(&tcp_hwcsum_data); |
924 | if (m->m_pkthdr.csum_flags & M_CSUM_NO_PSEUDOHDR) { | | 924 | if (m->m_pkthdr.csum_flags & M_CSUM_NO_PSEUDOHDR) { |
925 | const struct ip *ip = | | 925 | const struct ip *ip = |
926 | mtod(m, const struct ip *); | | 926 | mtod(m, const struct ip *); |
927 | | | 927 | |
928 | hw_csum = in_cksum_phdr(ip->ip_src.s_addr, | | 928 | hw_csum = in_cksum_phdr(ip->ip_src.s_addr, |
929 | ip->ip_dst.s_addr, | | 929 | ip->ip_dst.s_addr, |
930 | htons(hw_csum + tlen + off + IPPROTO_TCP)); | | 930 | htons(hw_csum + tlen + off + IPPROTO_TCP)); |
931 | } | | 931 | } |
932 | if ((hw_csum ^ 0xffff) != 0) | | 932 | if ((hw_csum ^ 0xffff) != 0) |
933 | goto badcsum; | | 933 | goto badcsum; |
934 | break; | | 934 | break; |
935 | } | | 935 | } |
936 | | | 936 | |
937 | case M_CSUM_TCPv4: | | 937 | case M_CSUM_TCPv4: |
938 | /* Checksum was okay. */ | | 938 | /* Checksum was okay. */ |
939 | TCP_CSUM_COUNTER_INCR(&tcp_hwcsum_ok); | | 939 | TCP_CSUM_COUNTER_INCR(&tcp_hwcsum_ok); |
940 | break; | | 940 | break; |
941 | | | 941 | |
942 | default: | | 942 | default: |
943 | /* | | 943 | /* |
944 | * Must compute it ourselves. Maybe skip checksum | | 944 | * Must compute it ourselves. Maybe skip checksum |
945 | * on loopback interfaces. | | 945 | * on loopback interfaces. |
946 | */ | | 946 | */ |
947 | if (__predict_true(!(rcvif->if_flags & IFF_LOOPBACK) || | | 947 | if (__predict_true(!(rcvif->if_flags & IFF_LOOPBACK) || |
948 | tcp_do_loopback_cksum)) { | | 948 | tcp_do_loopback_cksum)) { |
949 | TCP_CSUM_COUNTER_INCR(&tcp_swcsum); | | 949 | TCP_CSUM_COUNTER_INCR(&tcp_swcsum); |
950 | if (in4_cksum(m, IPPROTO_TCP, toff, | | 950 | if (in4_cksum(m, IPPROTO_TCP, toff, |
951 | tlen + off) != 0) | | 951 | tlen + off) != 0) |
952 | goto badcsum; | | 952 | goto badcsum; |
953 | } | | 953 | } |
954 | break; | | 954 | break; |
955 | } | | 955 | } |
956 | break; | | 956 | break; |
957 | #endif /* INET4 */ | | 957 | #endif /* INET4 */ |
958 | | | 958 | |
959 | #ifdef INET6 | | 959 | #ifdef INET6 |
960 | case AF_INET6: | | 960 | case AF_INET6: |
961 | switch (m->m_pkthdr.csum_flags & | | 961 | switch (m->m_pkthdr.csum_flags & |
962 | ((rcvif->if_csum_flags_rx & M_CSUM_TCPv6) | | | 962 | ((rcvif->if_csum_flags_rx & M_CSUM_TCPv6) | |
963 | M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) { | | 963 | M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) { |
964 | case M_CSUM_TCPv6|M_CSUM_TCP_UDP_BAD: | | 964 | case M_CSUM_TCPv6|M_CSUM_TCP_UDP_BAD: |
965 | TCP_CSUM_COUNTER_INCR(&tcp6_hwcsum_bad); | | 965 | TCP_CSUM_COUNTER_INCR(&tcp6_hwcsum_bad); |
966 | goto badcsum; | | 966 | goto badcsum; |
967 | | | 967 | |
968 | #if 0 /* notyet */ | | 968 | #if 0 /* notyet */ |
969 | case M_CSUM_TCPv6|M_CSUM_DATA: | | 969 | case M_CSUM_TCPv6|M_CSUM_DATA: |
970 | #endif | | 970 | #endif |
971 | | | 971 | |
972 | case M_CSUM_TCPv6: | | 972 | case M_CSUM_TCPv6: |
973 | /* Checksum was okay. */ | | 973 | /* Checksum was okay. */ |
974 | TCP_CSUM_COUNTER_INCR(&tcp6_hwcsum_ok); | | 974 | TCP_CSUM_COUNTER_INCR(&tcp6_hwcsum_ok); |
975 | break; | | 975 | break; |
976 | | | 976 | |
977 | default: | | 977 | default: |
978 | /* | | 978 | /* |
979 | * Must compute it ourselves. Maybe skip checksum | | 979 | * Must compute it ourselves. Maybe skip checksum |
980 | * on loopback interfaces. | | 980 | * on loopback interfaces. |
981 | */ | | 981 | */ |
982 | if (__predict_true((m->m_flags & M_LOOP) == 0 || | | 982 | if (__predict_true((m->m_flags & M_LOOP) == 0 || |
983 | tcp_do_loopback_cksum)) { | | 983 | tcp_do_loopback_cksum)) { |
984 | TCP_CSUM_COUNTER_INCR(&tcp6_swcsum); | | 984 | TCP_CSUM_COUNTER_INCR(&tcp6_swcsum); |
985 | if (in6_cksum(m, IPPROTO_TCP, toff, | | 985 | if (in6_cksum(m, IPPROTO_TCP, toff, |
986 | tlen + off) != 0) | | 986 | tlen + off) != 0) |
987 | goto badcsum; | | 987 | goto badcsum; |
988 | } | | 988 | } |
989 | } | | 989 | } |
990 | break; | | 990 | break; |
991 | #endif /* INET6 */ | | 991 | #endif /* INET6 */ |
992 | } | | 992 | } |
993 | m_put_rcvif(rcvif, &s); | | 993 | m_put_rcvif(rcvif, &s); |
994 | | | 994 | |
995 | return 0; | | 995 | return 0; |
996 | | | 996 | |
997 | badcsum: | | 997 | badcsum: |
998 | m_put_rcvif(rcvif, &s); | | 998 | m_put_rcvif(rcvif, &s); |
999 | TCP_STATINC(TCP_STAT_RCVBADSUM); | | 999 | TCP_STATINC(TCP_STAT_RCVBADSUM); |
1000 | return -1; | | 1000 | return -1; |
1001 | } | | 1001 | } |
1002 | | | 1002 | |
1003 | /* When a packet arrives addressed to a vestigial tcpbp, we | | 1003 | /* When a packet arrives addressed to a vestigial tcpbp, we |
1004 | * nevertheless have to respond to it per the spec. | | 1004 | * nevertheless have to respond to it per the spec. |
1005 | */ | | 1005 | */ |
1006 | static void tcp_vtw_input(struct tcphdr *th, vestigial_inpcb_t *vp, | | 1006 | static void tcp_vtw_input(struct tcphdr *th, vestigial_inpcb_t *vp, |
1007 | struct mbuf *m, int tlen, int multicast) | | 1007 | struct mbuf *m, int tlen, int multicast) |
1008 | { | | 1008 | { |
1009 | int tiflags; | | 1009 | int tiflags; |
1010 | int todrop; | | 1010 | int todrop; |
1011 | uint32_t t_flags = 0; | | 1011 | uint32_t t_flags = 0; |
1012 | uint64_t *tcps; | | 1012 | uint64_t *tcps; |
1013 | | | 1013 | |
1014 | tiflags = th->th_flags; | | 1014 | tiflags = th->th_flags; |
1015 | todrop = vp->rcv_nxt - th->th_seq; | | 1015 | todrop = vp->rcv_nxt - th->th_seq; |
1016 | | | 1016 | |
1017 | if (todrop > 0) { | | 1017 | if (todrop > 0) { |
1018 | if (tiflags & TH_SYN) { | | 1018 | if (tiflags & TH_SYN) { |
1019 | tiflags &= ~TH_SYN; | | 1019 | tiflags &= ~TH_SYN; |
1020 | ++th->th_seq; | | 1020 | ++th->th_seq; |
1021 | if (th->th_urp > 1) | | 1021 | if (th->th_urp > 1) |
1022 | --th->th_urp; | | 1022 | --th->th_urp; |
1023 | else { | | 1023 | else { |
1024 | tiflags &= ~TH_URG; | | 1024 | tiflags &= ~TH_URG; |
1025 | th->th_urp = 0; | | 1025 | th->th_urp = 0; |
1026 | } | | 1026 | } |
1027 | --todrop; | | 1027 | --todrop; |
1028 | } | | 1028 | } |
1029 | if (todrop > tlen || | | 1029 | if (todrop > tlen || |
1030 | (todrop == tlen && (tiflags & TH_FIN) == 0)) { | | 1030 | (todrop == tlen && (tiflags & TH_FIN) == 0)) { |
1031 | /* | | 1031 | /* |
1032 | * Any valid FIN or RST must be to the left of the | | 1032 | * Any valid FIN or RST must be to the left of the |
1033 | * window. At this point the FIN or RST must be a | | 1033 | * window. At this point the FIN or RST must be a |
1034 | * duplicate or out of sequence; drop it. | | 1034 | * duplicate or out of sequence; drop it. |
1035 | */ | | 1035 | */ |
1036 | if (tiflags & TH_RST) | | 1036 | if (tiflags & TH_RST) |
1037 | goto drop; | | 1037 | goto drop; |
1038 | tiflags &= ~(TH_FIN|TH_RST); | | 1038 | tiflags &= ~(TH_FIN|TH_RST); |
1039 | /* | | 1039 | /* |
1040 | * Send an ACK to resynchronize and drop any data. | | 1040 | * Send an ACK to resynchronize and drop any data. |
1041 | * But keep on processing for RST or ACK. | | 1041 | * But keep on processing for RST or ACK. |
1042 | */ | | 1042 | */ |
1043 | t_flags |= TF_ACKNOW; | | 1043 | t_flags |= TF_ACKNOW; |
1044 | todrop = tlen; | | 1044 | todrop = tlen; |
1045 | tcps = TCP_STAT_GETREF(); | | 1045 | tcps = TCP_STAT_GETREF(); |
1046 | tcps[TCP_STAT_RCVDUPPACK] += 1; | | 1046 | tcps[TCP_STAT_RCVDUPPACK] += 1; |
1047 | tcps[TCP_STAT_RCVDUPBYTE] += todrop; | | 1047 | tcps[TCP_STAT_RCVDUPBYTE] += todrop; |
1048 | TCP_STAT_PUTREF(); | | 1048 | TCP_STAT_PUTREF(); |
1049 | } else if ((tiflags & TH_RST) | | 1049 | } else if ((tiflags & TH_RST) |
1050 | && th->th_seq != vp->rcv_nxt) { | | 1050 | && th->th_seq != vp->rcv_nxt) { |
1051 | /* | | 1051 | /* |
1052 | * Test for reset before adjusting the sequence | | 1052 | * Test for reset before adjusting the sequence |
1053 | * number for overlapping data. | | 1053 | * number for overlapping data. |
1054 | */ | | 1054 | */ |
1055 | goto dropafterack_ratelim; | | 1055 | goto dropafterack_ratelim; |
1056 | } else { | | 1056 | } else { |
1057 | tcps = TCP_STAT_GETREF(); | | 1057 | tcps = TCP_STAT_GETREF(); |
1058 | tcps[TCP_STAT_RCVPARTDUPPACK] += 1; | | 1058 | tcps[TCP_STAT_RCVPARTDUPPACK] += 1; |
1059 | tcps[TCP_STAT_RCVPARTDUPBYTE] += todrop; | | 1059 | tcps[TCP_STAT_RCVPARTDUPBYTE] += todrop; |
1060 | TCP_STAT_PUTREF(); | | 1060 | TCP_STAT_PUTREF(); |
1061 | } | | 1061 | } |
1062 | | | 1062 | |
1063 | // tcp_new_dsack(tp, th->th_seq, todrop); | | 1063 | // tcp_new_dsack(tp, th->th_seq, todrop); |
1064 | // hdroptlen += todrop; /*drop from head afterwards*/ | | 1064 | // hdroptlen += todrop; /*drop from head afterwards*/ |
1065 | | | 1065 | |
1066 | th->th_seq += todrop; | | 1066 | th->th_seq += todrop; |
1067 | tlen -= todrop; | | 1067 | tlen -= todrop; |
1068 | | | 1068 | |
1069 | if (th->th_urp > todrop) | | 1069 | if (th->th_urp > todrop) |
1070 | th->th_urp -= todrop; | | 1070 | th->th_urp -= todrop; |
1071 | else { | | 1071 | else { |
1072 | tiflags &= ~TH_URG; | | 1072 | tiflags &= ~TH_URG; |
1073 | th->th_urp = 0; | | 1073 | th->th_urp = 0; |
1074 | } | | 1074 | } |
1075 | } | | 1075 | } |
1076 | | | 1076 | |
1077 | /* | | 1077 | /* |
1078 | * If new data are received on a connection after the | | 1078 | * If new data are received on a connection after the |
1079 | * user processes are gone, then RST the other end. | | 1079 | * user processes are gone, then RST the other end. |
1080 | */ | | 1080 | */ |
1081 | if (tlen) { | | 1081 | if (tlen) { |
1082 | TCP_STATINC(TCP_STAT_RCVAFTERCLOSE); | | 1082 | TCP_STATINC(TCP_STAT_RCVAFTERCLOSE); |
1083 | goto dropwithreset; | | 1083 | goto dropwithreset; |
1084 | } | | 1084 | } |
1085 | | | 1085 | |
1086 | /* | | 1086 | /* |
1087 | * If segment ends after window, drop trailing data | | 1087 | * If segment ends after window, drop trailing data |
1088 | * (and PUSH and FIN); if nothing left, just ACK. | | 1088 | * (and PUSH and FIN); if nothing left, just ACK. |
1089 | */ | | 1089 | */ |
1090 | todrop = (th->th_seq + tlen) - (vp->rcv_nxt+vp->rcv_wnd); | | 1090 | todrop = (th->th_seq + tlen) - (vp->rcv_nxt+vp->rcv_wnd); |
1091 | | | 1091 | |
1092 | if (todrop > 0) { | | 1092 | if (todrop > 0) { |
1093 | TCP_STATINC(TCP_STAT_RCVPACKAFTERWIN); | | 1093 | TCP_STATINC(TCP_STAT_RCVPACKAFTERWIN); |
1094 | if (todrop >= tlen) { | | 1094 | if (todrop >= tlen) { |
1095 | /* | | 1095 | /* |
1096 | * The segment actually starts after the window. | | 1096 | * The segment actually starts after the window. |
1097 | * th->th_seq + tlen - vp->rcv_nxt - vp->rcv_wnd >= tlen | | 1097 | * th->th_seq + tlen - vp->rcv_nxt - vp->rcv_wnd >= tlen |
1098 | * th->th_seq - vp->rcv_nxt - vp->rcv_wnd >= 0 | | 1098 | * th->th_seq - vp->rcv_nxt - vp->rcv_wnd >= 0 |
1099 | * th->th_seq >= vp->rcv_nxt + vp->rcv_wnd | | 1099 | * th->th_seq >= vp->rcv_nxt + vp->rcv_wnd |
1100 | */ | | 1100 | */ |
1101 | TCP_STATADD(TCP_STAT_RCVBYTEAFTERWIN, tlen); | | 1101 | TCP_STATADD(TCP_STAT_RCVBYTEAFTERWIN, tlen); |
1102 | /* | | 1102 | /* |
1103 | * If a new connection request is received | | 1103 | * If a new connection request is received |
1104 | * while in TIME_WAIT, drop the old connection | | 1104 | * while in TIME_WAIT, drop the old connection |
1105 | * and start over if the sequence numbers | | 1105 | * and start over if the sequence numbers |
1106 | * are above the previous ones. | | 1106 | * are above the previous ones. |
1107 | */ | | 1107 | */ |
1108 | if ((tiflags & TH_SYN) | | 1108 | if ((tiflags & TH_SYN) |
1109 | && SEQ_GT(th->th_seq, vp->rcv_nxt)) { | | 1109 | && SEQ_GT(th->th_seq, vp->rcv_nxt)) { |
1110 | /* We only support this in the !NOFDREF case, which | | 1110 | /* We only support this in the !NOFDREF case, which |
1111 | * is to say: not here. | | 1111 | * is to say: not here. |
1112 | */ | | 1112 | */ |
1113 | goto dropwithreset; | | 1113 | goto dropwithreset; |
1114 | } | | 1114 | } |
1115 | /* | | 1115 | /* |
1116 | * If window is closed can only take segments at | | 1116 | * If window is closed can only take segments at |
1117 | * window edge, and have to drop data and PUSH from | | 1117 | * window edge, and have to drop data and PUSH from |
1118 | * incoming segments. Continue processing, but | | 1118 | * incoming segments. Continue processing, but |
1119 | * remember to ack. Otherwise, drop segment | | 1119 | * remember to ack. Otherwise, drop segment |
1120 | * and (if not RST) ack. | | 1120 | * and (if not RST) ack. |
1121 | */ | | 1121 | */ |
1122 | if (vp->rcv_wnd == 0 && th->th_seq == vp->rcv_nxt) { | | 1122 | if (vp->rcv_wnd == 0 && th->th_seq == vp->rcv_nxt) { |
1123 | t_flags |= TF_ACKNOW; | | 1123 | t_flags |= TF_ACKNOW; |
1124 | TCP_STATINC(TCP_STAT_RCVWINPROBE); | | 1124 | TCP_STATINC(TCP_STAT_RCVWINPROBE); |
1125 | } else | | 1125 | } else |
1126 | goto dropafterack; | | 1126 | goto dropafterack; |
1127 | } else | | 1127 | } else |
1128 | TCP_STATADD(TCP_STAT_RCVBYTEAFTERWIN, todrop); | | 1128 | TCP_STATADD(TCP_STAT_RCVBYTEAFTERWIN, todrop); |
1129 | m_adj(m, -todrop); | | 1129 | m_adj(m, -todrop); |
1130 | tlen -= todrop; | | 1130 | tlen -= todrop; |
1131 | tiflags &= ~(TH_PUSH|TH_FIN); | | 1131 | tiflags &= ~(TH_PUSH|TH_FIN); |
1132 | } | | 1132 | } |
1133 | | | 1133 | |
1134 | if (tiflags & TH_RST) { | | 1134 | if (tiflags & TH_RST) { |
1135 | if (th->th_seq != vp->rcv_nxt) | | 1135 | if (th->th_seq != vp->rcv_nxt) |
1136 | goto dropafterack_ratelim; | | 1136 | goto dropafterack_ratelim; |
1137 | | | 1137 | |
1138 | vtw_del(vp->ctl, vp->vtw); | | 1138 | vtw_del(vp->ctl, vp->vtw); |
1139 | goto drop; | | 1139 | goto drop; |
1140 | } | | 1140 | } |
1141 | | | 1141 | |
1142 | /* | | 1142 | /* |
1143 | * If the ACK bit is off we drop the segment and return. | | 1143 | * If the ACK bit is off we drop the segment and return. |
1144 | */ | | 1144 | */ |
1145 | if ((tiflags & TH_ACK) == 0) { | | 1145 | if ((tiflags & TH_ACK) == 0) { |
1146 | if (t_flags & TF_ACKNOW) | | 1146 | if (t_flags & TF_ACKNOW) |
1147 | goto dropafterack; | | 1147 | goto dropafterack; |
1148 | else | | 1148 | else |
1149 | goto drop; | | 1149 | goto drop; |
1150 | } | | 1150 | } |
1151 | | | 1151 | |
1152 | /* | | 1152 | /* |
1153 | * In TIME_WAIT state the only thing that should arrive | | 1153 | * In TIME_WAIT state the only thing that should arrive |
1154 | * is a retransmission of the remote FIN. Acknowledge | | 1154 | * is a retransmission of the remote FIN. Acknowledge |
1155 | * it and restart the finack timer. | | 1155 | * it and restart the finack timer. |
1156 | */ | | 1156 | */ |
1157 | vtw_restart(vp); | | 1157 | vtw_restart(vp); |
1158 | goto dropafterack; | | 1158 | goto dropafterack; |
1159 | | | 1159 | |
1160 | dropafterack: | | 1160 | dropafterack: |
1161 | /* | | 1161 | /* |
1162 | * Generate an ACK dropping incoming segment if it occupies | | 1162 | * Generate an ACK dropping incoming segment if it occupies |
1163 | * sequence space, where the ACK reflects our state. | | 1163 | * sequence space, where the ACK reflects our state. |
1164 | */ | | 1164 | */ |
1165 | if (tiflags & TH_RST) | | 1165 | if (tiflags & TH_RST) |
1166 | goto drop; | | 1166 | goto drop; |
1167 | goto dropafterack2; | | 1167 | goto dropafterack2; |
1168 | | | 1168 | |
1169 | dropafterack_ratelim: | | 1169 | dropafterack_ratelim: |
1170 | /* | | 1170 | /* |
1171 | * We may want to rate-limit ACKs against SYN/RST attack. | | 1171 | * We may want to rate-limit ACKs against SYN/RST attack. |
1172 | */ | | 1172 | */ |
1173 | if (ppsratecheck(&tcp_ackdrop_ppslim_last, &tcp_ackdrop_ppslim_count, | | 1173 | if (ppsratecheck(&tcp_ackdrop_ppslim_last, &tcp_ackdrop_ppslim_count, |
1174 | tcp_ackdrop_ppslim) == 0) { | | 1174 | tcp_ackdrop_ppslim) == 0) { |
1175 | /* XXX stat */ | | 1175 | /* XXX stat */ |
1176 | goto drop; | | 1176 | goto drop; |
1177 | } | | 1177 | } |
1178 | /* ...fall into dropafterack2... */ | | 1178 | /* ...fall into dropafterack2... */ |
1179 | | | 1179 | |
1180 | dropafterack2: | | 1180 | dropafterack2: |
1181 | (void)tcp_respond(0, m, m, th, th->th_seq + tlen, th->th_ack, | | 1181 | (void)tcp_respond(0, m, m, th, th->th_seq + tlen, th->th_ack, |
1182 | TH_ACK); | | 1182 | TH_ACK); |
1183 | return; | | 1183 | return; |
1184 | | | 1184 | |
1185 | dropwithreset: | | 1185 | dropwithreset: |
1186 | /* | | 1186 | /* |
1187 | * Generate a RST, dropping incoming segment. | | 1187 | * Generate a RST, dropping incoming segment. |
1188 | * Make ACK acceptable to originator of segment. | | 1188 | * Make ACK acceptable to originator of segment. |
1189 | */ | | 1189 | */ |
1190 | if (tiflags & TH_RST) | | 1190 | if (tiflags & TH_RST) |
1191 | goto drop; | | 1191 | goto drop; |
1192 | | | 1192 | |
1193 | if (tiflags & TH_ACK) | | 1193 | if (tiflags & TH_ACK) |
1194 | tcp_respond(0, m, m, th, (tcp_seq)0, th->th_ack, TH_RST); | | 1194 | tcp_respond(0, m, m, th, (tcp_seq)0, th->th_ack, TH_RST); |
1195 | else { | | 1195 | else { |
1196 | if (tiflags & TH_SYN) | | 1196 | if (tiflags & TH_SYN) |
1197 | ++tlen; | | 1197 | ++tlen; |
1198 | (void)tcp_respond(0, m, m, th, th->th_seq + tlen, (tcp_seq)0, | | 1198 | (void)tcp_respond(0, m, m, th, th->th_seq + tlen, (tcp_seq)0, |
1199 | TH_RST|TH_ACK); | | 1199 | TH_RST|TH_ACK); |
1200 | } | | 1200 | } |
1201 | return; | | 1201 | return; |
1202 | drop: | | 1202 | drop: |
1203 | m_freem(m); | | 1203 | m_freem(m); |
1204 | } | | 1204 | } |
1205 | | | 1205 | |
1206 | /* | | 1206 | /* |
1207 | * TCP input routine, follows pages 65-76 of RFC 793 very closely. | | 1207 | * TCP input routine, follows pages 65-76 of RFC 793 very closely. |
1208 | */ | | 1208 | */ |
1209 | void | | 1209 | void |
1210 | tcp_input(struct mbuf *m, ...) | | 1210 | tcp_input(struct mbuf *m, ...) |
1211 | { | | 1211 | { |
1212 | struct tcphdr *th; | | 1212 | struct tcphdr *th; |
1213 | struct ip *ip; | | 1213 | struct ip *ip; |
1214 | struct inpcb *inp; | | 1214 | struct inpcb *inp; |
1215 | #ifdef INET6 | | 1215 | #ifdef INET6 |
1216 | struct ip6_hdr *ip6; | | 1216 | struct ip6_hdr *ip6; |
1217 | struct in6pcb *in6p; | | 1217 | struct in6pcb *in6p; |
1218 | #endif | | 1218 | #endif |
1219 | u_int8_t *optp = NULL; | | 1219 | u_int8_t *optp = NULL; |
1220 | int optlen = 0; | | 1220 | int optlen = 0; |
1221 | int len, tlen, toff, hdroptlen = 0; | | 1221 | int len, tlen, toff, hdroptlen = 0; |
1222 | struct tcpcb *tp = 0; | | 1222 | struct tcpcb *tp = 0; |
1223 | int tiflags; | | 1223 | int tiflags; |
1224 | struct socket *so = NULL; | | 1224 | struct socket *so = NULL; |
1225 | int todrop, acked, ourfinisacked, needoutput = 0; | | 1225 | int todrop, acked, ourfinisacked, needoutput = 0; |
1226 | bool dupseg; | | 1226 | bool dupseg; |
1227 | #ifdef TCP_DEBUG | | 1227 | #ifdef TCP_DEBUG |
1228 | short ostate = 0; | | 1228 | short ostate = 0; |
1229 | #endif | | 1229 | #endif |
1230 | u_long tiwin; | | 1230 | u_long tiwin; |
1231 | struct tcp_opt_info opti; | | 1231 | struct tcp_opt_info opti; |
1232 | int off, iphlen; | | 1232 | int off, iphlen; |
1233 | va_list ap; | | 1233 | va_list ap; |
1234 | int af; /* af on the wire */ | | 1234 | int af; /* af on the wire */ |
1235 | struct mbuf *tcp_saveti = NULL; | | 1235 | struct mbuf *tcp_saveti = NULL; |
1236 | uint32_t ts_rtt; | | 1236 | uint32_t ts_rtt; |
1237 | uint8_t iptos; | | 1237 | uint8_t iptos; |
1238 | uint64_t *tcps; | | 1238 | uint64_t *tcps; |
1239 | vestigial_inpcb_t vestige; | | 1239 | vestigial_inpcb_t vestige; |
1240 | | | 1240 | |
1241 | vestige.valid = 0; | | 1241 | vestige.valid = 0; |
1242 | | | 1242 | |
1243 | MCLAIM(m, &tcp_rx_mowner); | | 1243 | MCLAIM(m, &tcp_rx_mowner); |
1244 | va_start(ap, m); | | 1244 | va_start(ap, m); |
1245 | toff = va_arg(ap, int); | | 1245 | toff = va_arg(ap, int); |
1246 | (void)va_arg(ap, int); /* ignore value, advance ap */ | | 1246 | (void)va_arg(ap, int); /* ignore value, advance ap */ |
1247 | va_end(ap); | | 1247 | va_end(ap); |
1248 | | | 1248 | |
1249 | TCP_STATINC(TCP_STAT_RCVTOTAL); | | 1249 | TCP_STATINC(TCP_STAT_RCVTOTAL); |
1250 | | | 1250 | |
1251 | memset(&opti, 0, sizeof(opti)); | | 1251 | memset(&opti, 0, sizeof(opti)); |
1252 | opti.ts_present = 0; | | 1252 | opti.ts_present = 0; |
1253 | opti.maxseg = 0; | | 1253 | opti.maxseg = 0; |
1254 | | | 1254 | |
1255 | /* | | 1255 | /* |
1256 | * RFC1122 4.2.3.10, p. 104: discard bcast/mcast SYN. | | 1256 | * RFC1122 4.2.3.10, p. 104: discard bcast/mcast SYN. |
1257 | * | | 1257 | * |
1258 | * TCP is, by definition, unicast, so we reject all | | 1258 | * TCP is, by definition, unicast, so we reject all |
1259 | * multicast outright. | | 1259 | * multicast outright. |
1260 | * | | 1260 | * |
1261 | * Note, there are additional src/dst address checks in | | 1261 | * Note, there are additional src/dst address checks in |
1262 | * the AF-specific code below. | | 1262 | * the AF-specific code below. |
1263 | */ | | 1263 | */ |
1264 | if (m->m_flags & (M_BCAST|M_MCAST)) { | | 1264 | if (m->m_flags & (M_BCAST|M_MCAST)) { |
1265 | /* XXX stat */ | | 1265 | /* XXX stat */ |
1266 | goto drop; | | 1266 | goto drop; |
1267 | } | | 1267 | } |
1268 | #ifdef INET6 | | 1268 | #ifdef INET6 |
1269 | if (m->m_flags & M_ANYCAST6) { | | 1269 | if (m->m_flags & M_ANYCAST6) { |
1270 | /* XXX stat */ | | 1270 | /* XXX stat */ |
1271 | goto drop; | | 1271 | goto drop; |
1272 | } | | 1272 | } |
1273 | #endif | | 1273 | #endif |
1274 | | | 1274 | |
1275 | /* | | 1275 | /* |
| | | 1276 | * Enforce alignment requirements that are violated in |
| | | 1277 | * some cases, see kern/50766 for details. |
| | | 1278 | */ |
| | | 1279 | if (TCP_HDR_ALIGNED_P(th) == 0) { |
| | | 1280 | m = m_copyup(m, toff + sizeof(struct tcphdr), 0); |
| | | 1281 | if (m == NULL) { |
| | | 1282 | TCP_STATINC(TCP_STAT_RCVSHORT); |
| | | 1283 | return; |
| | | 1284 | } |
| | | 1285 | th = (struct tcphdr *)(mtod(m, char *) + toff); |
| | | 1286 | } |
| | | 1287 | KASSERT(TCP_HDR_ALIGNED_P(th)); |
| | | 1288 | |
| | | 1289 | /* |
1276 | * Get IP and TCP header. | | 1290 | * Get IP and TCP header. |
1277 | * Note: IP leaves IP header in first mbuf. | | 1291 | * Note: IP leaves IP header in first mbuf. |
1278 | */ | | 1292 | */ |
| | | 1293 | #ifdef INET6 |
| | | 1294 | ip6 = mtod(m, struct ip6_hdr *); |
| | | 1295 | #endif |
| | | 1296 | #ifdef INET |
1279 | ip = mtod(m, struct ip *); | | 1297 | ip = mtod(m, struct ip *); |
| | | 1298 | #endif |
1280 | switch (ip->ip_v) { | | 1299 | switch (ip->ip_v) { |
1281 | #ifdef INET | | 1300 | #ifdef INET |
1282 | case 4: | | 1301 | case 4: |
1283 | #ifdef INET6 | | | |
1284 | ip6 = NULL; | | | |
1285 | #endif | | | |
1286 | af = AF_INET; | | 1302 | af = AF_INET; |
1287 | iphlen = sizeof(struct ip); | | 1303 | iphlen = sizeof(struct ip); |
1288 | IP6_EXTHDR_GET(th, struct tcphdr *, m, toff, | | 1304 | IP6_EXTHDR_GET(th, struct tcphdr *, m, toff, |
1289 | sizeof(struct tcphdr)); | | 1305 | sizeof(struct tcphdr)); |
1290 | if (th == NULL) { | | 1306 | if (th == NULL) { |
1291 | TCP_STATINC(TCP_STAT_RCVSHORT); | | 1307 | TCP_STATINC(TCP_STAT_RCVSHORT); |
1292 | return; | | 1308 | return; |
1293 | } | | 1309 | } |
1294 | /* We do the checksum after PCB lookup... */ | | 1310 | /* We do the checksum after PCB lookup... */ |
1295 | len = ntohs(ip->ip_len); | | 1311 | len = ntohs(ip->ip_len); |
1296 | tlen = len - toff; | | 1312 | tlen = len - toff; |
1297 | iptos = ip->ip_tos; | | 1313 | iptos = ip->ip_tos; |
1298 | break; | | 1314 | break; |
1299 | #endif | | 1315 | #endif |
1300 | #ifdef INET6 | | 1316 | #ifdef INET6 |
1301 | case 6: | | 1317 | case 6: |
1302 | ip = NULL; | | | |
1303 | iphlen = sizeof(struct ip6_hdr); | | 1318 | iphlen = sizeof(struct ip6_hdr); |
1304 | af = AF_INET6; | | 1319 | af = AF_INET6; |
1305 | ip6 = mtod(m, struct ip6_hdr *); | | | |
1306 | IP6_EXTHDR_GET(th, struct tcphdr *, m, toff, | | 1320 | IP6_EXTHDR_GET(th, struct tcphdr *, m, toff, |
1307 | sizeof(struct tcphdr)); | | 1321 | sizeof(struct tcphdr)); |
1308 | if (th == NULL) { | | 1322 | if (th == NULL) { |
1309 | TCP_STATINC(TCP_STAT_RCVSHORT); | | 1323 | TCP_STATINC(TCP_STAT_RCVSHORT); |
1310 | return; | | 1324 | return; |
1311 | } | | 1325 | } |
1312 | | | 1326 | |
1313 | /* Be proactive about malicious use of IPv4 mapped address */ | | 1327 | /* Be proactive about malicious use of IPv4 mapped address */ |
1314 | if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) || | | 1328 | if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) || |
1315 | IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) { | | 1329 | IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) { |
1316 | /* XXX stat */ | | 1330 | /* XXX stat */ |
1317 | goto drop; | | 1331 | goto drop; |
1318 | } | | 1332 | } |
1319 | | | 1333 | |
1320 | /* | | 1334 | /* |
1321 | * Be proactive about unspecified IPv6 address in source. | | 1335 | * Be proactive about unspecified IPv6 address in source. |
1322 | * As we use all-zero to indicate unbounded/unconnected pcb, | | 1336 | * As we use all-zero to indicate unbounded/unconnected pcb, |
1323 | * unspecified IPv6 address can be used to confuse us. | | 1337 | * unspecified IPv6 address can be used to confuse us. |
1324 | * | | 1338 | * |
1325 | * Note that packets with unspecified IPv6 destination is | | 1339 | * Note that packets with unspecified IPv6 destination is |
1326 | * already dropped in ip6_input. | | 1340 | * already dropped in ip6_input. |
1327 | */ | | 1341 | */ |
1328 | if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) { | | 1342 | if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) { |
1329 | /* XXX stat */ | | 1343 | /* XXX stat */ |
1330 | goto drop; | | 1344 | goto drop; |
1331 | } | | 1345 | } |
1332 | | | 1346 | |
1333 | /* | | 1347 | /* |
1334 | * Make sure destination address is not multicast. | | 1348 | * Make sure destination address is not multicast. |
1335 | * Source address checked in ip6_input(). | | 1349 | * Source address checked in ip6_input(). |
1336 | */ | | 1350 | */ |
1337 | if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { | | 1351 | if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) { |
1338 | /* XXX stat */ | | 1352 | /* XXX stat */ |
1339 | goto drop; | | 1353 | goto drop; |
1340 | } | | 1354 | } |
1341 | | | 1355 | |
1342 | /* We do the checksum after PCB lookup... */ | | 1356 | /* We do the checksum after PCB lookup... */ |
1343 | len = m->m_pkthdr.len; | | 1357 | len = m->m_pkthdr.len; |
1344 | tlen = len - toff; | | 1358 | tlen = len - toff; |
1345 | iptos = (ntohl(ip6->ip6_flow) >> 20) & 0xff; | | 1359 | iptos = (ntohl(ip6->ip6_flow) >> 20) & 0xff; |
1346 | break; | | 1360 | break; |
1347 | #endif | | 1361 | #endif |
1348 | default: | | 1362 | default: |
1349 | m_freem(m); | | 1363 | m_freem(m); |
1350 | return; | | 1364 | return; |
1351 | } | | 1365 | } |
1352 | /* | | | |
1353 | * Enforce alignment requirements that are violated in | | | |
1354 | * some cases, see kern/50766 for details. | | | |
1355 | */ | | | |
1356 | if (TCP_HDR_ALIGNED_P(th) == 0) { | | | |
1357 | m = m_copyup(m, toff + sizeof(struct tcphdr), 0); | | | |
1358 | if (m == NULL) { | | | |
1359 | TCP_STATINC(TCP_STAT_RCVSHORT); | | | |
1360 | return; | | | |
1361 | } | | | |
1362 | ip = mtod(m, struct ip *); | | | |
1363 | #ifdef INET6 | | | |
1364 | ip6 = mtod(m, struct ip6_hdr *); | | | |
1365 | #endif | | | |
1366 | th = (struct tcphdr *)(mtod(m, char *) + toff); | | | |
1367 | } | | | |
1368 | KASSERT(TCP_HDR_ALIGNED_P(th)); | | | |
1369 | | | 1366 | |
1370 | /* | | 1367 | /* |
1371 | * Check that TCP offset makes sense, | | 1368 | * Check that TCP offset makes sense, |
1372 | * pull out TCP options and adjust length. XXX | | 1369 | * pull out TCP options and adjust length. XXX |
1373 | */ | | 1370 | */ |
1374 | off = th->th_off << 2; | | 1371 | off = th->th_off << 2; |
1375 | if (off < sizeof (struct tcphdr) || off > tlen) { | | 1372 | if (off < sizeof (struct tcphdr) || off > tlen) { |
1376 | TCP_STATINC(TCP_STAT_RCVBADOFF); | | 1373 | TCP_STATINC(TCP_STAT_RCVBADOFF); |
1377 | goto drop; | | 1374 | goto drop; |
1378 | } | | 1375 | } |
1379 | tlen -= off; | | 1376 | tlen -= off; |
1380 | | | 1377 | |
1381 | /* | | 1378 | /* |
1382 | * tcp_input() has been modified to use tlen to mean the TCP data | | 1379 | * tcp_input() has been modified to use tlen to mean the TCP data |
1383 | * length throughout the function. Other functions can use | | 1380 | * length throughout the function. Other functions can use |
1384 | * m->m_pkthdr.len as the basis for calculating the TCP data length. | | 1381 | * m->m_pkthdr.len as the basis for calculating the TCP data length. |
1385 | * rja | | 1382 | * rja |
1386 | */ | | 1383 | */ |
1387 | | | 1384 | |
1388 | if (off > sizeof (struct tcphdr)) { | | 1385 | if (off > sizeof (struct tcphdr)) { |
1389 | IP6_EXTHDR_GET(th, struct tcphdr *, m, toff, off); | | 1386 | IP6_EXTHDR_GET(th, struct tcphdr *, m, toff, off); |
1390 | if (th == NULL) { | | 1387 | if (th == NULL) { |
1391 | TCP_STATINC(TCP_STAT_RCVSHORT); | | 1388 | TCP_STATINC(TCP_STAT_RCVSHORT); |
1392 | return; | | 1389 | return; |
1393 | } | | 1390 | } |
1394 | /* | | 1391 | /* |
1395 | * NOTE: ip/ip6 will not be affected by m_pulldown() | | 1392 | * NOTE: ip/ip6 will not be affected by m_pulldown() |
1396 | * (as they're before toff) and we don't need to update those. | | 1393 | * (as they're before toff) and we don't need to update those. |
1397 | */ | | 1394 | */ |
1398 | KASSERT(TCP_HDR_ALIGNED_P(th)); | | 1395 | KASSERT(TCP_HDR_ALIGNED_P(th)); |
1399 | optlen = off - sizeof (struct tcphdr); | | 1396 | optlen = off - sizeof (struct tcphdr); |
1400 | optp = ((u_int8_t *)th) + sizeof(struct tcphdr); | | 1397 | optp = ((u_int8_t *)th) + sizeof(struct tcphdr); |
1401 | /* | | 1398 | /* |
1402 | * Do quick retrieval of timestamp options ("options | | 1399 | * Do quick retrieval of timestamp options ("options |
1403 | * prediction?"). If timestamp is the only option and it's | | 1400 | * prediction?"). If timestamp is the only option and it's |
1404 | * formatted as recommended in RFC 1323 appendix A, we | | 1401 | * formatted as recommended in RFC 1323 appendix A, we |
1405 | * quickly get the values now and not bother calling | | 1402 | * quickly get the values now and not bother calling |
1406 | * tcp_dooptions(), etc. | | 1403 | * tcp_dooptions(), etc. |
1407 | */ | | 1404 | */ |
1408 | if ((optlen == TCPOLEN_TSTAMP_APPA || | | 1405 | if ((optlen == TCPOLEN_TSTAMP_APPA || |
1409 | (optlen > TCPOLEN_TSTAMP_APPA && | | 1406 | (optlen > TCPOLEN_TSTAMP_APPA && |
1410 | optp[TCPOLEN_TSTAMP_APPA] == TCPOPT_EOL)) && | | 1407 | optp[TCPOLEN_TSTAMP_APPA] == TCPOPT_EOL)) && |
1411 | *(u_int32_t *)optp == htonl(TCPOPT_TSTAMP_HDR) && | | 1408 | *(u_int32_t *)optp == htonl(TCPOPT_TSTAMP_HDR) && |
1412 | (th->th_flags & TH_SYN) == 0) { | | 1409 | (th->th_flags & TH_SYN) == 0) { |
1413 | opti.ts_present = 1; | | 1410 | opti.ts_present = 1; |
1414 | opti.ts_val = ntohl(*(u_int32_t *)(optp + 4)); | | 1411 | opti.ts_val = ntohl(*(u_int32_t *)(optp + 4)); |
1415 | opti.ts_ecr = ntohl(*(u_int32_t *)(optp + 8)); | | 1412 | opti.ts_ecr = ntohl(*(u_int32_t *)(optp + 8)); |
1416 | optp = NULL; /* we've parsed the options */ | | 1413 | optp = NULL; /* we've parsed the options */ |
1417 | } | | 1414 | } |
1418 | } | | 1415 | } |
1419 | tiflags = th->th_flags; | | 1416 | tiflags = th->th_flags; |
1420 | | | 1417 | |
1421 | /* | | 1418 | /* |
1422 | * Checksum extended TCP header and data | | 1419 | * Checksum extended TCP header and data |
1423 | */ | | 1420 | */ |
1424 | if (tcp_input_checksum(af, m, th, toff, off, tlen)) | | 1421 | if (tcp_input_checksum(af, m, th, toff, off, tlen)) |
1425 | goto badcsum; | | 1422 | goto badcsum; |
1426 | | | 1423 | |
1427 | /* | | 1424 | /* |
1428 | * Locate pcb for segment. | | 1425 | * Locate pcb for segment. |
1429 | */ | | 1426 | */ |
1430 | findpcb: | | 1427 | findpcb: |
1431 | inp = NULL; | | 1428 | inp = NULL; |
1432 | #ifdef INET6 | | 1429 | #ifdef INET6 |
1433 | in6p = NULL; | | 1430 | in6p = NULL; |
1434 | #endif | | 1431 | #endif |
1435 | switch (af) { | | 1432 | switch (af) { |
1436 | #ifdef INET | | 1433 | #ifdef INET |
1437 | case AF_INET: | | 1434 | case AF_INET: |
1438 | inp = in_pcblookup_connect(&tcbtable, ip->ip_src, th->th_sport, | | 1435 | inp = in_pcblookup_connect(&tcbtable, ip->ip_src, th->th_sport, |
1439 | ip->ip_dst, th->th_dport, | | 1436 | ip->ip_dst, th->th_dport, |
1440 | &vestige); | | 1437 | &vestige); |
1441 | if (inp == 0 && !vestige.valid) { | | 1438 | if (inp == 0 && !vestige.valid) { |
1442 | TCP_STATINC(TCP_STAT_PCBHASHMISS); | | 1439 | TCP_STATINC(TCP_STAT_PCBHASHMISS); |
1443 | inp = in_pcblookup_bind(&tcbtable, ip->ip_dst, th->th_dport); | | 1440 | inp = in_pcblookup_bind(&tcbtable, ip->ip_dst, th->th_dport); |
1444 | } | | 1441 | } |
1445 | #ifdef INET6 | | 1442 | #ifdef INET6 |
1446 | if (inp == 0 && !vestige.valid) { | | 1443 | if (inp == 0 && !vestige.valid) { |
1447 | struct in6_addr s, d; | | 1444 | struct in6_addr s, d; |
1448 | | | 1445 | |
1449 | /* mapped addr case */ | | 1446 | /* mapped addr case */ |
1450 | in6_in_2_v4mapin6(&ip->ip_src, &s); | | 1447 | in6_in_2_v4mapin6(&ip->ip_src, &s); |
1451 | in6_in_2_v4mapin6(&ip->ip_dst, &d); | | 1448 | in6_in_2_v4mapin6(&ip->ip_dst, &d); |
1452 | in6p = in6_pcblookup_connect(&tcbtable, &s, | | 1449 | in6p = in6_pcblookup_connect(&tcbtable, &s, |
1453 | th->th_sport, &d, th->th_dport, | | 1450 | th->th_sport, &d, th->th_dport, |
1454 | 0, &vestige); | | 1451 | 0, &vestige); |
1455 | if (in6p == 0 && !vestige.valid) { | | 1452 | if (in6p == 0 && !vestige.valid) { |
1456 | TCP_STATINC(TCP_STAT_PCBHASHMISS); | | 1453 | TCP_STATINC(TCP_STAT_PCBHASHMISS); |
1457 | in6p = in6_pcblookup_bind(&tcbtable, &d, | | 1454 | in6p = in6_pcblookup_bind(&tcbtable, &d, |
1458 | th->th_dport, 0); | | 1455 | th->th_dport, 0); |
1459 | } | | 1456 | } |
1460 | } | | 1457 | } |
1461 | #endif | | 1458 | #endif |
1462 | #ifndef INET6 | | 1459 | #ifndef INET6 |
1463 | if (inp == 0 && !vestige.valid) | | 1460 | if (inp == 0 && !vestige.valid) |
1464 | #else | | 1461 | #else |
1465 | if (inp == 0 && in6p == 0 && !vestige.valid) | | 1462 | if (inp == 0 && in6p == 0 && !vestige.valid) |
1466 | #endif | | 1463 | #endif |
1467 | { | | 1464 | { |
1468 | TCP_STATINC(TCP_STAT_NOPORT); | | 1465 | TCP_STATINC(TCP_STAT_NOPORT); |
1469 | if (tcp_log_refused && | | 1466 | if (tcp_log_refused && |
1470 | (tiflags & (TH_RST|TH_ACK|TH_SYN)) == TH_SYN) { | | 1467 | (tiflags & (TH_RST|TH_ACK|TH_SYN)) == TH_SYN) { |
1471 | tcp4_log_refused(ip, th); | | 1468 | tcp4_log_refused(ip, th); |
1472 | } | | 1469 | } |
1473 | tcp_fields_to_host(th); | | 1470 | tcp_fields_to_host(th); |
1474 | goto dropwithreset_ratelim; | | 1471 | goto dropwithreset_ratelim; |
1475 | } | | 1472 | } |
1476 | #if defined(IPSEC) | | 1473 | #if defined(IPSEC) |
1477 | if (ipsec_used) { | | 1474 | if (ipsec_used) { |
1478 | if (inp && ipsec4_in_reject(m, inp)) { | | 1475 | if (inp && ipsec4_in_reject(m, inp)) { |
1479 | IPSEC_STATINC(IPSEC_STAT_IN_POLVIO); | | 1476 | IPSEC_STATINC(IPSEC_STAT_IN_POLVIO); |
1480 | goto drop; | | 1477 | goto drop; |
1481 | } | | 1478 | } |
1482 | #ifdef INET6 | | 1479 | #ifdef INET6 |
1483 | else if (in6p && ipsec6_in_reject(m, in6p)) { | | 1480 | else if (in6p && ipsec6_in_reject(m, in6p)) { |
1484 | IPSEC_STATINC(IPSEC_STAT_IN_POLVIO); | | 1481 | IPSEC_STATINC(IPSEC_STAT_IN_POLVIO); |
1485 | goto drop; | | 1482 | goto drop; |
1486 | } | | 1483 | } |
1487 | #endif | | 1484 | #endif |
1488 | } | | 1485 | } |
1489 | #endif /*IPSEC*/ | | 1486 | #endif /*IPSEC*/ |
1490 | break; | | 1487 | break; |
1491 | #endif /*INET*/ | | 1488 | #endif /*INET*/ |
1492 | #ifdef INET6 | | 1489 | #ifdef INET6 |
1493 | case AF_INET6: | | 1490 | case AF_INET6: |
1494 | { | | 1491 | { |
1495 | int faith; | | 1492 | int faith; |
1496 | | | 1493 | |
1497 | #if defined(NFAITH) && NFAITH > 0 | | 1494 | #if defined(NFAITH) && NFAITH > 0 |
1498 | faith = faithprefix(&ip6->ip6_dst); | | 1495 | faith = faithprefix(&ip6->ip6_dst); |
1499 | #else | | 1496 | #else |
1500 | faith = 0; | | 1497 | faith = 0; |
1501 | #endif | | 1498 | #endif |
1502 | in6p = in6_pcblookup_connect(&tcbtable, &ip6->ip6_src, | | 1499 | in6p = in6_pcblookup_connect(&tcbtable, &ip6->ip6_src, |
1503 | th->th_sport, &ip6->ip6_dst, th->th_dport, faith, &vestige); | | 1500 | th->th_sport, &ip6->ip6_dst, th->th_dport, faith, &vestige); |
1504 | if (!in6p && !vestige.valid) { | | 1501 | if (!in6p && !vestige.valid) { |
1505 | TCP_STATINC(TCP_STAT_PCBHASHMISS); | | 1502 | TCP_STATINC(TCP_STAT_PCBHASHMISS); |
1506 | in6p = in6_pcblookup_bind(&tcbtable, &ip6->ip6_dst, | | 1503 | in6p = in6_pcblookup_bind(&tcbtable, &ip6->ip6_dst, |
1507 | th->th_dport, faith); | | 1504 | th->th_dport, faith); |
1508 | } | | 1505 | } |
1509 | if (!in6p && !vestige.valid) { | | 1506 | if (!in6p && !vestige.valid) { |
1510 | TCP_STATINC(TCP_STAT_NOPORT); | | 1507 | TCP_STATINC(TCP_STAT_NOPORT); |
1511 | if (tcp_log_refused && | | 1508 | if (tcp_log_refused && |
1512 | (tiflags & (TH_RST|TH_ACK|TH_SYN)) == TH_SYN) { | | 1509 | (tiflags & (TH_RST|TH_ACK|TH_SYN)) == TH_SYN) { |
1513 | tcp6_log_refused(ip6, th); | | 1510 | tcp6_log_refused(ip6, th); |
1514 | } | | 1511 | } |
1515 | tcp_fields_to_host(th); | | 1512 | tcp_fields_to_host(th); |
1516 | goto dropwithreset_ratelim; | | 1513 | goto dropwithreset_ratelim; |
1517 | } | | 1514 | } |
1518 | #if defined(IPSEC) | | 1515 | #if defined(IPSEC) |
1519 | if (ipsec_used && in6p && ipsec6_in_reject(m, in6p)) { | | 1516 | if (ipsec_used && in6p && ipsec6_in_reject(m, in6p)) { |
1520 | IPSEC6_STATINC(IPSEC_STAT_IN_POLVIO); | | 1517 | IPSEC6_STATINC(IPSEC_STAT_IN_POLVIO); |
1521 | goto drop; | | 1518 | goto drop; |
1522 | } | | 1519 | } |
1523 | #endif /*IPSEC*/ | | 1520 | #endif /*IPSEC*/ |
1524 | break; | | 1521 | break; |
1525 | } | | 1522 | } |
1526 | #endif | | 1523 | #endif |
1527 | } | | 1524 | } |
1528 | | | 1525 | |
1529 | /* | | 1526 | /* |
1530 | * If the state is CLOSED (i.e., TCB does not exist) then | | 1527 | * If the state is CLOSED (i.e., TCB does not exist) then |
1531 | * all data in the incoming segment is discarded. | | 1528 | * all data in the incoming segment is discarded. |
1532 | * If the TCB exists but is in CLOSED state, it is embryonic, | | 1529 | * If the TCB exists but is in CLOSED state, it is embryonic, |
1533 | * but should either do a listen or a connect soon. | | 1530 | * but should either do a listen or a connect soon. |
1534 | */ | | 1531 | */ |
1535 | tp = NULL; | | 1532 | tp = NULL; |
1536 | so = NULL; | | 1533 | so = NULL; |
1537 | if (inp) { | | 1534 | if (inp) { |
1538 | /* Check the minimum TTL for socket. */ | | 1535 | /* Check the minimum TTL for socket. */ |
1539 | if (ip->ip_ttl < inp->inp_ip_minttl) | | 1536 | if (ip->ip_ttl < inp->inp_ip_minttl) |
1540 | goto drop; | | 1537 | goto drop; |
1541 | | | 1538 | |
1542 | tp = intotcpcb(inp); | | 1539 | tp = intotcpcb(inp); |
1543 | so = inp->inp_socket; | | 1540 | so = inp->inp_socket; |
1544 | } | | 1541 | } |
1545 | #ifdef INET6 | | 1542 | #ifdef INET6 |
1546 | else if (in6p) { | | 1543 | else if (in6p) { |
1547 | tp = in6totcpcb(in6p); | | 1544 | tp = in6totcpcb(in6p); |
1548 | so = in6p->in6p_socket; | | 1545 | so = in6p->in6p_socket; |
1549 | } | | 1546 | } |
1550 | #endif | | 1547 | #endif |
1551 | else if (vestige.valid) { | | 1548 | else if (vestige.valid) { |
1552 | int mc = 0; | | 1549 | int mc = 0; |
1553 | | | 1550 | |
1554 | /* We do not support the resurrection of vtw tcpcps. | | 1551 | /* We do not support the resurrection of vtw tcpcps. |
1555 | */ | | 1552 | */ |
1556 | if (tcp_input_checksum(af, m, th, toff, off, tlen)) | | 1553 | if (tcp_input_checksum(af, m, th, toff, off, tlen)) |
1557 | goto badcsum; | | 1554 | goto badcsum; |
1558 | | | 1555 | |
1559 | switch (af) { | | 1556 | switch (af) { |
1560 | #ifdef INET6 | | 1557 | #ifdef INET6 |
1561 | case AF_INET6: | | 1558 | case AF_INET6: |
1562 | mc = IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst); | | 1559 | mc = IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst); |
1563 | break; | | 1560 | break; |
1564 | #endif | | 1561 | #endif |
1565 | | | 1562 | |
1566 | case AF_INET: | | 1563 | case AF_INET: |
1567 | mc = (IN_MULTICAST(ip->ip_dst.s_addr) | | 1564 | mc = (IN_MULTICAST(ip->ip_dst.s_addr) |
1568 | || in_broadcast(ip->ip_dst, | | 1565 | || in_broadcast(ip->ip_dst, |
1569 | m_get_rcvif_NOMPSAFE(m))); | | 1566 | m_get_rcvif_NOMPSAFE(m))); |
1570 | break; | | 1567 | break; |
1571 | } | | 1568 | } |
1572 | | | 1569 | |
1573 | tcp_fields_to_host(th); | | 1570 | tcp_fields_to_host(th); |
1574 | tcp_vtw_input(th, &vestige, m, tlen, mc); | | 1571 | tcp_vtw_input(th, &vestige, m, tlen, mc); |
1575 | m = 0; | | 1572 | m = 0; |
1576 | goto drop; | | 1573 | goto drop; |
1577 | } | | 1574 | } |
1578 | | | 1575 | |
1579 | if (tp == 0) { | | 1576 | if (tp == 0) { |
1580 | tcp_fields_to_host(th); | | 1577 | tcp_fields_to_host(th); |
1581 | goto dropwithreset_ratelim; | | 1578 | goto dropwithreset_ratelim; |
1582 | } | | 1579 | } |
1583 | if (tp->t_state == TCPS_CLOSED) | | 1580 | if (tp->t_state == TCPS_CLOSED) |
1584 | goto drop; | | 1581 | goto drop; |
1585 | | | 1582 | |
1586 | KASSERT(so->so_lock == softnet_lock); | | 1583 | KASSERT(so->so_lock == softnet_lock); |
1587 | KASSERT(solocked(so)); | | 1584 | KASSERT(solocked(so)); |
1588 | | | 1585 | |
1589 | tcp_fields_to_host(th); | | 1586 | tcp_fields_to_host(th); |
1590 | | | 1587 | |
1591 | /* Unscale the window into a 32-bit value. */ | | 1588 | /* Unscale the window into a 32-bit value. */ |
1592 | if ((tiflags & TH_SYN) == 0) | | 1589 | if ((tiflags & TH_SYN) == 0) |
1593 | tiwin = th->th_win << tp->snd_scale; | | 1590 | tiwin = th->th_win << tp->snd_scale; |
1594 | else | | 1591 | else |
1595 | tiwin = th->th_win; | | 1592 | tiwin = th->th_win; |
1596 | | | 1593 | |
1597 | #ifdef INET6 | | 1594 | #ifdef INET6 |
1598 | /* save packet options if user wanted */ | | 1595 | /* save packet options if user wanted */ |
1599 | if (in6p && (in6p->in6p_flags & IN6P_CONTROLOPTS)) { | | 1596 | if (in6p && (in6p->in6p_flags & IN6P_CONTROLOPTS)) { |
1600 | if (in6p->in6p_options) { | | 1597 | if (in6p->in6p_options) { |
1601 | m_freem(in6p->in6p_options); | | 1598 | m_freem(in6p->in6p_options); |
1602 | in6p->in6p_options = 0; | | 1599 | in6p->in6p_options = 0; |
1603 | } | | 1600 | } |
1604 | KASSERT(ip6 != NULL); | | | |
1605 | ip6_savecontrol(in6p, &in6p->in6p_options, ip6, m); | | 1601 | ip6_savecontrol(in6p, &in6p->in6p_options, ip6, m); |
1606 | } | | 1602 | } |
1607 | #endif | | 1603 | #endif |
1608 | | | 1604 | |
1609 | if (so->so_options & (SO_DEBUG|SO_ACCEPTCONN)) { | | 1605 | if (so->so_options & (SO_DEBUG|SO_ACCEPTCONN)) { |
1610 | union syn_cache_sa src; | | 1606 | union syn_cache_sa src; |
1611 | union syn_cache_sa dst; | | 1607 | union syn_cache_sa dst; |
1612 | | | 1608 | |
1613 | memset(&src, 0, sizeof(src)); | | 1609 | memset(&src, 0, sizeof(src)); |
1614 | memset(&dst, 0, sizeof(dst)); | | 1610 | memset(&dst, 0, sizeof(dst)); |
1615 | switch (af) { | | 1611 | switch (af) { |
1616 | #ifdef INET | | 1612 | #ifdef INET |
1617 | case AF_INET: | | 1613 | case AF_INET: |
1618 | src.sin.sin_len = sizeof(struct sockaddr_in); | | 1614 | src.sin.sin_len = sizeof(struct sockaddr_in); |
1619 | src.sin.sin_family = AF_INET; | | 1615 | src.sin.sin_family = AF_INET; |
1620 | src.sin.sin_addr = ip->ip_src; | | 1616 | src.sin.sin_addr = ip->ip_src; |
1621 | src.sin.sin_port = th->th_sport; | | 1617 | src.sin.sin_port = th->th_sport; |
1622 | | | 1618 | |
1623 | dst.sin.sin_len = sizeof(struct sockaddr_in); | | 1619 | dst.sin.sin_len = sizeof(struct sockaddr_in); |
1624 | dst.sin.sin_family = AF_INET; | | 1620 | dst.sin.sin_family = AF_INET; |
1625 | dst.sin.sin_addr = ip->ip_dst; | | 1621 | dst.sin.sin_addr = ip->ip_dst; |
1626 | dst.sin.sin_port = th->th_dport; | | 1622 | dst.sin.sin_port = th->th_dport; |
1627 | break; | | 1623 | break; |
1628 | #endif | | 1624 | #endif |
1629 | #ifdef INET6 | | 1625 | #ifdef INET6 |
1630 | case AF_INET6: | | 1626 | case AF_INET6: |
1631 | src.sin6.sin6_len = sizeof(struct sockaddr_in6); | | 1627 | src.sin6.sin6_len = sizeof(struct sockaddr_in6); |
1632 | src.sin6.sin6_family = AF_INET6; | | 1628 | src.sin6.sin6_family = AF_INET6; |
1633 | src.sin6.sin6_addr = ip6->ip6_src; | | 1629 | src.sin6.sin6_addr = ip6->ip6_src; |
1634 | src.sin6.sin6_port = th->th_sport; | | 1630 | src.sin6.sin6_port = th->th_sport; |
1635 | | | 1631 | |
1636 | dst.sin6.sin6_len = sizeof(struct sockaddr_in6); | | 1632 | dst.sin6.sin6_len = sizeof(struct sockaddr_in6); |
1637 | dst.sin6.sin6_family = AF_INET6; | | 1633 | dst.sin6.sin6_family = AF_INET6; |
1638 | dst.sin6.sin6_addr = ip6->ip6_dst; | | 1634 | dst.sin6.sin6_addr = ip6->ip6_dst; |
1639 | dst.sin6.sin6_port = th->th_dport; | | 1635 | dst.sin6.sin6_port = th->th_dport; |
1640 | break; | | 1636 | break; |
1641 | #endif /* INET6 */ | | 1637 | #endif /* INET6 */ |
1642 | default: | | 1638 | default: |
1643 | goto badsyn; /*sanity*/ | | 1639 | goto badsyn; /*sanity*/ |
1644 | } | | 1640 | } |
1645 | | | 1641 | |
1646 | if (so->so_options & SO_DEBUG) { | | 1642 | if (so->so_options & SO_DEBUG) { |
1647 | #ifdef TCP_DEBUG | | 1643 | #ifdef TCP_DEBUG |
1648 | ostate = tp->t_state; | | 1644 | ostate = tp->t_state; |
1649 | #endif | | 1645 | #endif |
1650 | | | 1646 | |
1651 | tcp_saveti = NULL; | | 1647 | tcp_saveti = NULL; |
1652 | if (iphlen + sizeof(struct tcphdr) > MHLEN) | | 1648 | if (iphlen + sizeof(struct tcphdr) > MHLEN) |
1653 | goto nosave; | | 1649 | goto nosave; |
1654 | | | 1650 | |
1655 | if (m->m_len > iphlen && (m->m_flags & M_EXT) == 0) { | | 1651 | if (m->m_len > iphlen && (m->m_flags & M_EXT) == 0) { |
1656 | tcp_saveti = m_copym(m, 0, iphlen, M_DONTWAIT); | | 1652 | tcp_saveti = m_copym(m, 0, iphlen, M_DONTWAIT); |
1657 | if (!tcp_saveti) | | 1653 | if (!tcp_saveti) |
1658 | goto nosave; | | 1654 | goto nosave; |
1659 | } else { | | 1655 | } else { |
1660 | MGETHDR(tcp_saveti, M_DONTWAIT, MT_HEADER); | | 1656 | MGETHDR(tcp_saveti, M_DONTWAIT, MT_HEADER); |
1661 | if (!tcp_saveti) | | 1657 | if (!tcp_saveti) |
1662 | goto nosave; | | 1658 | goto nosave; |
1663 | MCLAIM(m, &tcp_mowner); | | 1659 | MCLAIM(m, &tcp_mowner); |
1664 | tcp_saveti->m_len = iphlen; | | 1660 | tcp_saveti->m_len = iphlen; |
1665 | m_copydata(m, 0, iphlen, | | 1661 | m_copydata(m, 0, iphlen, |
1666 | mtod(tcp_saveti, void *)); | | 1662 | mtod(tcp_saveti, void *)); |
1667 | } | | 1663 | } |
1668 | | | 1664 | |
1669 | if (M_TRAILINGSPACE(tcp_saveti) < sizeof(struct tcphdr)) { | | 1665 | if (M_TRAILINGSPACE(tcp_saveti) < sizeof(struct tcphdr)) { |
1670 | m_freem(tcp_saveti); | | 1666 | m_freem(tcp_saveti); |
1671 | tcp_saveti = NULL; | | 1667 | tcp_saveti = NULL; |
1672 | } else { | | 1668 | } else { |
1673 | tcp_saveti->m_len += sizeof(struct tcphdr); | | 1669 | tcp_saveti->m_len += sizeof(struct tcphdr); |
1674 | memcpy(mtod(tcp_saveti, char *) + iphlen, th, | | 1670 | memcpy(mtod(tcp_saveti, char *) + iphlen, th, |
1675 | sizeof(struct tcphdr)); | | 1671 | sizeof(struct tcphdr)); |
1676 | } | | 1672 | } |
1677 | nosave:; | | 1673 | nosave:; |
1678 | } | | 1674 | } |
1679 | if (so->so_options & SO_ACCEPTCONN) { | | 1675 | if (so->so_options & SO_ACCEPTCONN) { |
1680 | if ((tiflags & (TH_RST|TH_ACK|TH_SYN)) != TH_SYN) { | | 1676 | if ((tiflags & (TH_RST|TH_ACK|TH_SYN)) != TH_SYN) { |
1681 | if (tiflags & TH_RST) { | | 1677 | if (tiflags & TH_RST) { |
1682 | syn_cache_reset(&src.sa, &dst.sa, th); | | 1678 | syn_cache_reset(&src.sa, &dst.sa, th); |
1683 | } else if ((tiflags & (TH_ACK|TH_SYN)) == | | 1679 | } else if ((tiflags & (TH_ACK|TH_SYN)) == |
1684 | (TH_ACK|TH_SYN)) { | | 1680 | (TH_ACK|TH_SYN)) { |
1685 | /* | | 1681 | /* |
1686 | * Received a SYN,ACK. This should | | 1682 | * Received a SYN,ACK. This should |
1687 | * never happen while we are in | | 1683 | * never happen while we are in |
1688 | * LISTEN. Send an RST. | | 1684 | * LISTEN. Send an RST. |
1689 | */ | | 1685 | */ |
1690 | goto badsyn; | | 1686 | goto badsyn; |
1691 | } else if (tiflags & TH_ACK) { | | 1687 | } else if (tiflags & TH_ACK) { |
1692 | so = syn_cache_get(&src.sa, &dst.sa, | | 1688 | so = syn_cache_get(&src.sa, &dst.sa, |
1693 | th, toff, tlen, so, m); | | 1689 | th, toff, tlen, so, m); |
1694 | if (so == NULL) { | | 1690 | if (so == NULL) { |
1695 | /* | | 1691 | /* |
1696 | * We don't have a SYN for | | 1692 | * We don't have a SYN for |
1697 | * this ACK; send an RST. | | 1693 | * this ACK; send an RST. |
1698 | */ | | 1694 | */ |
1699 | goto badsyn; | | 1695 | goto badsyn; |
1700 | } else if (so == | | 1696 | } else if (so == |
1701 | (struct socket *)(-1)) { | | 1697 | (struct socket *)(-1)) { |
1702 | /* | | 1698 | /* |
1703 | * We were unable to create | | 1699 | * We were unable to create |
1704 | * the connection. If the | | 1700 | * the connection. If the |
1705 | * 3-way handshake was | | 1701 | * 3-way handshake was |
1706 | * completed, and RST has | | 1702 | * completed, and RST has |
1707 | * been sent to the peer. | | 1703 | * been sent to the peer. |
1708 | * Since the mbuf might be | | 1704 | * Since the mbuf might be |
1709 | * in use for the reply, | | 1705 | * in use for the reply, |
1710 | * do not free it. | | 1706 | * do not free it. |
1711 | */ | | 1707 | */ |
1712 | m = NULL; | | 1708 | m = NULL; |
1713 | } else { | | 1709 | } else { |
1714 | /* | | 1710 | /* |
1715 | * We have created a | | 1711 | * We have created a |
1716 | * full-blown connection. | | 1712 | * full-blown connection. |
1717 | */ | | 1713 | */ |
1718 | tp = NULL; | | 1714 | tp = NULL; |
1719 | inp = NULL; | | 1715 | inp = NULL; |
1720 | #ifdef INET6 | | 1716 | #ifdef INET6 |
1721 | in6p = NULL; | | 1717 | in6p = NULL; |
1722 | #endif | | 1718 | #endif |
1723 | switch (so->so_proto->pr_domain->dom_family) { | | 1719 | switch (so->so_proto->pr_domain->dom_family) { |
1724 | #ifdef INET | | 1720 | #ifdef INET |
1725 | case AF_INET: | | 1721 | case AF_INET: |
1726 | inp = sotoinpcb(so); | | 1722 | inp = sotoinpcb(so); |
1727 | tp = intotcpcb(inp); | | 1723 | tp = intotcpcb(inp); |
1728 | break; | | 1724 | break; |
1729 | #endif | | 1725 | #endif |
1730 | #ifdef INET6 | | 1726 | #ifdef INET6 |
1731 | case AF_INET6: | | 1727 | case AF_INET6: |
1732 | in6p = sotoin6pcb(so); | | 1728 | in6p = sotoin6pcb(so); |
1733 | tp = in6totcpcb(in6p); | | 1729 | tp = in6totcpcb(in6p); |
1734 | break; | | 1730 | break; |
1735 | #endif | | 1731 | #endif |
1736 | } | | 1732 | } |
1737 | if (tp == NULL) | | 1733 | if (tp == NULL) |
1738 | goto badsyn; /*XXX*/ | | 1734 | goto badsyn; /*XXX*/ |
1739 | tiwin <<= tp->snd_scale; | | 1735 | tiwin <<= tp->snd_scale; |
1740 | goto after_listen; | | 1736 | goto after_listen; |
1741 | } | | 1737 | } |
1742 | } else { | | 1738 | } else { |
1743 | /* | | 1739 | /* |
1744 | * None of RST, SYN or ACK was set. | | 1740 | * None of RST, SYN or ACK was set. |
1745 | * This is an invalid packet for a | | 1741 | * This is an invalid packet for a |
1746 | * TCB in LISTEN state. Send a RST. | | 1742 | * TCB in LISTEN state. Send a RST. |
1747 | */ | | 1743 | */ |
1748 | goto badsyn; | | 1744 | goto badsyn; |
1749 | } | | 1745 | } |
1750 | } else { | | 1746 | } else { |
1751 | /* | | 1747 | /* |
1752 | * Received a SYN. | | 1748 | * Received a SYN. |
1753 | * | | 1749 | * |
1754 | * RFC1122 4.2.3.10, p. 104: discard bcast/mcast SYN | | 1750 | * RFC1122 4.2.3.10, p. 104: discard bcast/mcast SYN |
1755 | */ | | 1751 | */ |
1756 | if (m->m_flags & (M_BCAST|M_MCAST)) | | 1752 | if (m->m_flags & (M_BCAST|M_MCAST)) |
1757 | goto drop; | | 1753 | goto drop; |
1758 | | | 1754 | |
1759 | switch (af) { | | 1755 | switch (af) { |
1760 | #ifdef INET6 | | 1756 | #ifdef INET6 |
1761 | case AF_INET6: | | 1757 | case AF_INET6: |
1762 | if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) | | 1758 | if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) |
1763 | goto drop; | | 1759 | goto drop; |
1764 | break; | | 1760 | break; |
1765 | #endif /* INET6 */ | | 1761 | #endif /* INET6 */ |
1766 | case AF_INET: | | 1762 | case AF_INET: |
1767 | if (IN_MULTICAST(ip->ip_dst.s_addr) || | | 1763 | if (IN_MULTICAST(ip->ip_dst.s_addr) || |
1768 | in_broadcast(ip->ip_dst, | | 1764 | in_broadcast(ip->ip_dst, |
1769 | m_get_rcvif_NOMPSAFE(m))) | | 1765 | m_get_rcvif_NOMPSAFE(m))) |
1770 | goto drop; | | 1766 | goto drop; |
1771 | break; | | 1767 | break; |
1772 | } | | 1768 | } |
1773 | | | 1769 | |
1774 | #ifdef INET6 | | 1770 | #ifdef INET6 |
1775 | /* | | 1771 | /* |
1776 | * If deprecated address is forbidden, we do | | 1772 | * If deprecated address is forbidden, we do |
1777 | * not accept SYN to deprecated interface | | 1773 | * not accept SYN to deprecated interface |
1778 | * address to prevent any new inbound | | 1774 | * address to prevent any new inbound |
1779 | * connection from getting established. | | 1775 | * connection from getting established. |
1780 | * When we do not accept SYN, we send a TCP | | 1776 | * When we do not accept SYN, we send a TCP |
1781 | * RST, with deprecated source address (instead | | 1777 | * RST, with deprecated source address (instead |
1782 | * of dropping it). We compromise it as it is | | 1778 | * of dropping it). We compromise it as it is |
1783 | * much better for peer to send a RST, and | | 1779 | * much better for peer to send a RST, and |
1784 | * RST will be the final packet for the | | 1780 | * RST will be the final packet for the |
1785 | * exchange. | | 1781 | * exchange. |
1786 | * | | 1782 | * |
1787 | * If we do not forbid deprecated addresses, we | | 1783 | * If we do not forbid deprecated addresses, we |
1788 | * accept the SYN packet. RFC2462 does not | | 1784 | * accept the SYN packet. RFC2462 does not |
1789 | * suggest dropping SYN in this case. | | 1785 | * suggest dropping SYN in this case. |
1790 | * If we decipher RFC2462 5.5.4, it says like | | 1786 | * If we decipher RFC2462 5.5.4, it says like |
1791 | * this: | | 1787 | * this: |
1792 | * 1. use of deprecated addr with existing | | 1788 | * 1. use of deprecated addr with existing |
1793 | * communication is okay - "SHOULD continue | | 1789 | * communication is okay - "SHOULD continue |
1794 | * to be used" | | 1790 | * to be used" |
1795 | * 2. use of it with new communication: | | 1791 | * 2. use of it with new communication: |
1796 | * (2a) "SHOULD NOT be used if alternate | | 1792 | * (2a) "SHOULD NOT be used if alternate |
1797 | * address with sufficient scope is | | 1793 | * address with sufficient scope is |
1798 | * available" | | 1794 | * available" |
1799 | * (2b) nothing mentioned otherwise. | | 1795 | * (2b) nothing mentioned otherwise. |
1800 | * Here we fall into (2b) case as we have no | | 1796 | * Here we fall into (2b) case as we have no |
1801 | * choice in our source address selection - we | | 1797 | * choice in our source address selection - we |
1802 | * must obey the peer. | | 1798 | * must obey the peer. |
1803 | * | | 1799 | * |
1804 | * The wording in RFC2462 is confusing, and | | 1800 | * The wording in RFC2462 is confusing, and |
1805 | * there are multiple description text for | | 1801 | * there are multiple description text for |
1806 | * deprecated address handling - worse, they | | 1802 | * deprecated address handling - worse, they |
1807 | * are not exactly the same. I believe 5.5.4 | | 1803 | * are not exactly the same. I believe 5.5.4 |
1808 | * is the best one, so we follow 5.5.4. | | 1804 | * is the best one, so we follow 5.5.4. |
1809 | */ | | 1805 | */ |
1810 | if (af == AF_INET6 && !ip6_use_deprecated) { | | 1806 | if (af == AF_INET6 && !ip6_use_deprecated) { |
1811 | struct in6_ifaddr *ia6; | | 1807 | struct in6_ifaddr *ia6; |
1812 | int s; | | 1808 | int s; |
1813 | struct ifnet *rcvif = m_get_rcvif(m, &s); | | 1809 | struct ifnet *rcvif = m_get_rcvif(m, &s); |
1814 | if (rcvif == NULL) | | 1810 | if (rcvif == NULL) |
1815 | goto dropwithreset; /* XXX */ | | 1811 | goto dropwithreset; /* XXX */ |
1816 | if ((ia6 = in6ifa_ifpwithaddr(rcvif, | | 1812 | if ((ia6 = in6ifa_ifpwithaddr(rcvif, |
1817 | &ip6->ip6_dst)) && | | 1813 | &ip6->ip6_dst)) && |
1818 | (ia6->ia6_flags & IN6_IFF_DEPRECATED)) { | | 1814 | (ia6->ia6_flags & IN6_IFF_DEPRECATED)) { |
1819 | tp = NULL; | | 1815 | tp = NULL; |
1820 | m_put_rcvif(rcvif, &s); | | 1816 | m_put_rcvif(rcvif, &s); |
1821 | goto dropwithreset; | | 1817 | goto dropwithreset; |
1822 | } | | 1818 | } |
1823 | m_put_rcvif(rcvif, &s); | | 1819 | m_put_rcvif(rcvif, &s); |
1824 | } | | 1820 | } |
1825 | #endif | | 1821 | #endif |
1826 | | | 1822 | |
1827 | /* | | 1823 | /* |
1828 | * LISTEN socket received a SYN | | 1824 | * LISTEN socket received a SYN |
1829 | * from itself? This can't possibly | | 1825 | * from itself? This can't possibly |
1830 | * be valid; drop the packet. | | 1826 | * be valid; drop the packet. |
1831 | */ | | 1827 | */ |
1832 | if (th->th_sport == th->th_dport) { | | 1828 | if (th->th_sport == th->th_dport) { |
1833 | int i; | | 1829 | int i; |
1834 | | | 1830 | |
1835 | switch (af) { | | 1831 | switch (af) { |
1836 | #ifdef INET | | 1832 | #ifdef INET |
1837 | case AF_INET: | | 1833 | case AF_INET: |
1838 | i = in_hosteq(ip->ip_src, ip->ip_dst); | | 1834 | i = in_hosteq(ip->ip_src, ip->ip_dst); |
1839 | break; | | 1835 | break; |
1840 | #endif | | 1836 | #endif |
1841 | #ifdef INET6 | | 1837 | #ifdef INET6 |
1842 | case AF_INET6: | | 1838 | case AF_INET6: |
1843 | i = IN6_ARE_ADDR_EQUAL(&ip6->ip6_src, &ip6->ip6_dst); | | 1839 | i = IN6_ARE_ADDR_EQUAL(&ip6->ip6_src, &ip6->ip6_dst); |
1844 | break; | | 1840 | break; |
1845 | #endif | | 1841 | #endif |
1846 | default: | | 1842 | default: |
1847 | i = 1; | | 1843 | i = 1; |
1848 | } | | 1844 | } |
1849 | if (i) { | | 1845 | if (i) { |
1850 | TCP_STATINC(TCP_STAT_BADSYN); | | 1846 | TCP_STATINC(TCP_STAT_BADSYN); |
1851 | goto drop; | | 1847 | goto drop; |
1852 | } | | 1848 | } |
1853 | } | | 1849 | } |
1854 | | | 1850 | |
1855 | /* | | 1851 | /* |
1856 | * SYN looks ok; create compressed TCP | | 1852 | * SYN looks ok; create compressed TCP |
1857 | * state for it. | | 1853 | * state for it. |
1858 | */ | | 1854 | */ |
1859 | if (so->so_qlen <= so->so_qlimit && | | 1855 | if (so->so_qlen <= so->so_qlimit && |
1860 | syn_cache_add(&src.sa, &dst.sa, th, tlen, | | 1856 | syn_cache_add(&src.sa, &dst.sa, th, tlen, |
1861 | so, m, optp, optlen, &opti)) | | 1857 | so, m, optp, optlen, &opti)) |
1862 | m = NULL; | | 1858 | m = NULL; |
1863 | } | | 1859 | } |
1864 | goto drop; | | 1860 | goto drop; |
1865 | } | | 1861 | } |
1866 | } | | 1862 | } |
1867 | | | 1863 | |
1868 | after_listen: | | 1864 | after_listen: |
1869 | #ifdef DIAGNOSTIC | | 1865 | #ifdef DIAGNOSTIC |
1870 | /* | | 1866 | /* |
1871 | * Should not happen now that all embryonic connections | | 1867 | * Should not happen now that all embryonic connections |
1872 | * are handled with compressed state. | | 1868 | * are handled with compressed state. |
1873 | */ | | 1869 | */ |
1874 | if (tp->t_state == TCPS_LISTEN) | | 1870 | if (tp->t_state == TCPS_LISTEN) |
1875 | panic("tcp_input: TCPS_LISTEN"); | | 1871 | panic("tcp_input: TCPS_LISTEN"); |
1876 | #endif | | 1872 | #endif |
1877 | | | 1873 | |
1878 | /* | | 1874 | /* |
1879 | * Segment received on connection. | | 1875 | * Segment received on connection. |
1880 | * Reset idle time and keep-alive timer. | | 1876 | * Reset idle time and keep-alive timer. |
1881 | */ | | 1877 | */ |
1882 | tp->t_rcvtime = tcp_now; | | 1878 | tp->t_rcvtime = tcp_now; |
1883 | if (TCPS_HAVEESTABLISHED(tp->t_state)) | | 1879 | if (TCPS_HAVEESTABLISHED(tp->t_state)) |
1884 | TCP_TIMER_ARM(tp, TCPT_KEEP, tp->t_keepidle); | | 1880 | TCP_TIMER_ARM(tp, TCPT_KEEP, tp->t_keepidle); |
1885 | | | 1881 | |
1886 | /* | | 1882 | /* |
1887 | * Process options. | | 1883 | * Process options. |
1888 | */ | | 1884 | */ |
1889 | #ifdef TCP_SIGNATURE | | 1885 | #ifdef TCP_SIGNATURE |
1890 | if (optp || (tp->t_flags & TF_SIGNATURE)) | | 1886 | if (optp || (tp->t_flags & TF_SIGNATURE)) |
1891 | #else | | 1887 | #else |
1892 | if (optp) | | 1888 | if (optp) |
1893 | #endif | | 1889 | #endif |
1894 | if (tcp_dooptions(tp, optp, optlen, th, m, toff, &opti) < 0) | | 1890 | if (tcp_dooptions(tp, optp, optlen, th, m, toff, &opti) < 0) |
1895 | goto drop; | | 1891 | goto drop; |
1896 | | | 1892 | |
1897 | if (TCP_SACK_ENABLED(tp)) { | | 1893 | if (TCP_SACK_ENABLED(tp)) { |
1898 | tcp_del_sackholes(tp, th); | | 1894 | tcp_del_sackholes(tp, th); |
1899 | } | | 1895 | } |
1900 | | | 1896 | |
1901 | if (TCP_ECN_ALLOWED(tp)) { | | 1897 | if (TCP_ECN_ALLOWED(tp)) { |
1902 | if (tiflags & TH_CWR) { | | 1898 | if (tiflags & TH_CWR) { |
1903 | tp->t_flags &= ~TF_ECN_SND_ECE; | | 1899 | tp->t_flags &= ~TF_ECN_SND_ECE; |
1904 | } | | 1900 | } |
1905 | switch (iptos & IPTOS_ECN_MASK) { | | 1901 | switch (iptos & IPTOS_ECN_MASK) { |
1906 | case IPTOS_ECN_CE: | | 1902 | case IPTOS_ECN_CE: |
1907 | tp->t_flags |= TF_ECN_SND_ECE; | | 1903 | tp->t_flags |= TF_ECN_SND_ECE; |
1908 | TCP_STATINC(TCP_STAT_ECN_CE); | | 1904 | TCP_STATINC(TCP_STAT_ECN_CE); |
1909 | break; | | 1905 | break; |
1910 | case IPTOS_ECN_ECT0: | | 1906 | case IPTOS_ECN_ECT0: |
1911 | TCP_STATINC(TCP_STAT_ECN_ECT); | | 1907 | TCP_STATINC(TCP_STAT_ECN_ECT); |
1912 | break; | | 1908 | break; |
1913 | case IPTOS_ECN_ECT1: | | 1909 | case IPTOS_ECN_ECT1: |
1914 | /* XXX: ignore for now -- rpaulo */ | | 1910 | /* XXX: ignore for now -- rpaulo */ |
1915 | break; | | 1911 | break; |
1916 | } | | 1912 | } |
1917 | /* | | 1913 | /* |
1918 | * Congestion experienced. | | 1914 | * Congestion experienced. |
1919 | * Ignore if we are already trying to recover. | | 1915 | * Ignore if we are already trying to recover. |
1920 | */ | | 1916 | */ |
1921 | if ((tiflags & TH_ECE) && SEQ_GEQ(tp->snd_una, tp->snd_recover)) | | 1917 | if ((tiflags & TH_ECE) && SEQ_GEQ(tp->snd_una, tp->snd_recover)) |
1922 | tp->t_congctl->cong_exp(tp); | | 1918 | tp->t_congctl->cong_exp(tp); |
1923 | } | | 1919 | } |
1924 | | | 1920 | |
1925 | if (opti.ts_present && opti.ts_ecr) { | | 1921 | if (opti.ts_present && opti.ts_ecr) { |
1926 | /* | | 1922 | /* |
1927 | * Calculate the RTT from the returned time stamp and the | | 1923 | * Calculate the RTT from the returned time stamp and the |
1928 | * connection's time base. If the time stamp is later than | | 1924 | * connection's time base. If the time stamp is later than |
1929 | * the current time, or is extremely old, fall back to non-1323 | | 1925 | * the current time, or is extremely old, fall back to non-1323 |
1930 | * RTT calculation. Since ts_rtt is unsigned, we can test both | | 1926 | * RTT calculation. Since ts_rtt is unsigned, we can test both |
1931 | * at the same time. | | 1927 | * at the same time. |
1932 | * | | 1928 | * |
1933 | * Note that ts_rtt is in units of slow ticks (500 | | 1929 | * Note that ts_rtt is in units of slow ticks (500 |
1934 | * ms). Since most earthbound RTTs are < 500 ms, | | 1930 | * ms). Since most earthbound RTTs are < 500 ms, |
1935 | * observed values will have large quantization noise. | | 1931 | * observed values will have large quantization noise. |
1936 | * Our smoothed RTT is then the fraction of observed | | 1932 | * Our smoothed RTT is then the fraction of observed |
1937 | * samples that are 1 tick instead of 0 (times 500 | | 1933 | * samples that are 1 tick instead of 0 (times 500 |
1938 | * ms). | | 1934 | * ms). |
1939 | * | | 1935 | * |
1940 | * ts_rtt is increased by 1 to denote a valid sample, | | 1936 | * ts_rtt is increased by 1 to denote a valid sample, |
1941 | * with 0 indicating an invalid measurement. This | | 1937 | * with 0 indicating an invalid measurement. This |
1942 | * extra 1 must be removed when ts_rtt is used, or | | 1938 | * extra 1 must be removed when ts_rtt is used, or |
1943 | * else an an erroneous extra 500 ms will result. | | 1939 | * else an an erroneous extra 500 ms will result. |
1944 | */ | | 1940 | */ |
1945 | ts_rtt = TCP_TIMESTAMP(tp) - opti.ts_ecr + 1; | | 1941 | ts_rtt = TCP_TIMESTAMP(tp) - opti.ts_ecr + 1; |
1946 | if (ts_rtt > TCP_PAWS_IDLE) | | 1942 | if (ts_rtt > TCP_PAWS_IDLE) |
1947 | ts_rtt = 0; | | 1943 | ts_rtt = 0; |
1948 | } else { | | 1944 | } else { |
1949 | ts_rtt = 0; | | 1945 | ts_rtt = 0; |
1950 | } | | 1946 | } |
1951 | | | 1947 | |
1952 | /* | | 1948 | /* |
1953 | * Header prediction: check for the two common cases | | 1949 | * Header prediction: check for the two common cases |
1954 | * of a uni-directional data xfer. If the packet has | | 1950 | * of a uni-directional data xfer. If the packet has |
1955 | * no control flags, is in-sequence, the window didn't | | 1951 | * no control flags, is in-sequence, the window didn't |
1956 | * change and we're not retransmitting, it's a | | 1952 | * change and we're not retransmitting, it's a |
1957 | * candidate. If the length is zero and the ack moved | | 1953 | * candidate. If the length is zero and the ack moved |
1958 | * forward, we're the sender side of the xfer. Just | | 1954 | * forward, we're the sender side of the xfer. Just |
1959 | * free the data acked & wake any higher level process | | 1955 | * free the data acked & wake any higher level process |
1960 | * that was blocked waiting for space. If the length | | 1956 | * that was blocked waiting for space. If the length |
1961 | * is non-zero and the ack didn't move, we're the | | 1957 | * is non-zero and the ack didn't move, we're the |
1962 | * receiver side. If we're getting packets in-order | | 1958 | * receiver side. If we're getting packets in-order |
1963 | * (the reassembly queue is empty), add the data to | | 1959 | * (the reassembly queue is empty), add the data to |
1964 | * the socket buffer and note that we need a delayed ack. | | 1960 | * the socket buffer and note that we need a delayed ack. |
1965 | */ | | 1961 | */ |
1966 | if (tp->t_state == TCPS_ESTABLISHED && | | 1962 | if (tp->t_state == TCPS_ESTABLISHED && |
1967 | (tiflags & (TH_SYN|TH_FIN|TH_RST|TH_URG|TH_ECE|TH_CWR|TH_ACK)) | | 1963 | (tiflags & (TH_SYN|TH_FIN|TH_RST|TH_URG|TH_ECE|TH_CWR|TH_ACK)) |
1968 | == TH_ACK && | | 1964 | == TH_ACK && |
1969 | (!opti.ts_present || TSTMP_GEQ(opti.ts_val, tp->ts_recent)) && | | 1965 | (!opti.ts_present || TSTMP_GEQ(opti.ts_val, tp->ts_recent)) && |
1970 | th->th_seq == tp->rcv_nxt && | | 1966 | th->th_seq == tp->rcv_nxt && |
1971 | tiwin && tiwin == tp->snd_wnd && | | 1967 | tiwin && tiwin == tp->snd_wnd && |
1972 | tp->snd_nxt == tp->snd_max) { | | 1968 | tp->snd_nxt == tp->snd_max) { |
1973 | | | 1969 | |
1974 | /* | | 1970 | /* |
1975 | * If last ACK falls within this segment's sequence numbers, | | 1971 | * If last ACK falls within this segment's sequence numbers, |
1976 | * record the timestamp. | | 1972 | * record the timestamp. |
1977 | * NOTE that the test is modified according to the latest | | 1973 | * NOTE that the test is modified according to the latest |
1978 | * proposal of the tcplw@cray.com list (Braden 1993/04/26). | | 1974 | * proposal of the tcplw@cray.com list (Braden 1993/04/26). |
1979 | * | | 1975 | * |
1980 | * note that we already know | | 1976 | * note that we already know |
1981 | * TSTMP_GEQ(opti.ts_val, tp->ts_recent) | | 1977 | * TSTMP_GEQ(opti.ts_val, tp->ts_recent) |
1982 | */ | | 1978 | */ |
1983 | if (opti.ts_present && | | 1979 | if (opti.ts_present && |
1984 | SEQ_LEQ(th->th_seq, tp->last_ack_sent)) { | | 1980 | SEQ_LEQ(th->th_seq, tp->last_ack_sent)) { |
1985 | tp->ts_recent_age = tcp_now; | | 1981 | tp->ts_recent_age = tcp_now; |
1986 | tp->ts_recent = opti.ts_val; | | 1982 | tp->ts_recent = opti.ts_val; |
1987 | } | | 1983 | } |
1988 | | | 1984 | |
1989 | if (tlen == 0) { | | 1985 | if (tlen == 0) { |
1990 | /* Ack prediction. */ | | 1986 | /* Ack prediction. */ |
1991 | if (SEQ_GT(th->th_ack, tp->snd_una) && | | 1987 | if (SEQ_GT(th->th_ack, tp->snd_una) && |
1992 | SEQ_LEQ(th->th_ack, tp->snd_max) && | | 1988 | SEQ_LEQ(th->th_ack, tp->snd_max) && |
1993 | tp->snd_cwnd >= tp->snd_wnd && | | 1989 | tp->snd_cwnd >= tp->snd_wnd && |
1994 | tp->t_partialacks < 0) { | | 1990 | tp->t_partialacks < 0) { |
1995 | /* | | 1991 | /* |
1996 | * this is a pure ack for outstanding data. | | 1992 | * this is a pure ack for outstanding data. |
1997 | */ | | 1993 | */ |
1998 | if (ts_rtt) | | 1994 | if (ts_rtt) |
1999 | tcp_xmit_timer(tp, ts_rtt - 1); | | 1995 | tcp_xmit_timer(tp, ts_rtt - 1); |
2000 | else if (tp->t_rtttime && | | 1996 | else if (tp->t_rtttime && |
2001 | SEQ_GT(th->th_ack, tp->t_rtseq)) | | 1997 | SEQ_GT(th->th_ack, tp->t_rtseq)) |
2002 | tcp_xmit_timer(tp, | | 1998 | tcp_xmit_timer(tp, |
2003 | tcp_now - tp->t_rtttime); | | 1999 | tcp_now - tp->t_rtttime); |
2004 | acked = th->th_ack - tp->snd_una; | | 2000 | acked = th->th_ack - tp->snd_una; |
2005 | tcps = TCP_STAT_GETREF(); | | 2001 | tcps = TCP_STAT_GETREF(); |
2006 | tcps[TCP_STAT_PREDACK]++; | | 2002 | tcps[TCP_STAT_PREDACK]++; |
2007 | tcps[TCP_STAT_RCVACKPACK]++; | | 2003 | tcps[TCP_STAT_RCVACKPACK]++; |
2008 | tcps[TCP_STAT_RCVACKBYTE] += acked; | | 2004 | tcps[TCP_STAT_RCVACKBYTE] += acked; |
2009 | TCP_STAT_PUTREF(); | | 2005 | TCP_STAT_PUTREF(); |
2010 | nd6_hint(tp); | | 2006 | nd6_hint(tp); |
2011 | | | 2007 | |
2012 | if (acked > (tp->t_lastoff - tp->t_inoff)) | | 2008 | if (acked > (tp->t_lastoff - tp->t_inoff)) |
2013 | tp->t_lastm = NULL; | | 2009 | tp->t_lastm = NULL; |
2014 | sbdrop(&so->so_snd, acked); | | 2010 | sbdrop(&so->so_snd, acked); |
2015 | tp->t_lastoff -= acked; | | 2011 | tp->t_lastoff -= acked; |
2016 | | | 2012 | |
2017 | icmp_check(tp, th, acked); | | 2013 | icmp_check(tp, th, acked); |
2018 | | | 2014 | |
2019 | tp->snd_una = th->th_ack; | | 2015 | tp->snd_una = th->th_ack; |
2020 | tp->snd_fack = tp->snd_una; | | 2016 | tp->snd_fack = tp->snd_una; |
2021 | if (SEQ_LT(tp->snd_high, tp->snd_una)) | | 2017 | if (SEQ_LT(tp->snd_high, tp->snd_una)) |
2022 | tp->snd_high = tp->snd_una; | | 2018 | tp->snd_high = tp->snd_una; |
2023 | m_freem(m); | | 2019 | m_freem(m); |
2024 | | | 2020 | |
2025 | /* | | 2021 | /* |
2026 | * If all outstanding data are acked, stop | | 2022 | * If all outstanding data are acked, stop |
2027 | * retransmit timer, otherwise restart timer | | 2023 | * retransmit timer, otherwise restart timer |
2028 | * using current (possibly backed-off) value. | | 2024 | * using current (possibly backed-off) value. |
2029 | * If process is waiting for space, | | 2025 | * If process is waiting for space, |
2030 | * wakeup/selnotify/signal. If data | | 2026 | * wakeup/selnotify/signal. If data |
2031 | * are ready to send, let tcp_output | | 2027 | * are ready to send, let tcp_output |
2032 | * decide between more output or persist. | | 2028 | * decide between more output or persist. |
2033 | */ | | 2029 | */ |
2034 | if (tp->snd_una == tp->snd_max) | | 2030 | if (tp->snd_una == tp->snd_max) |
2035 | TCP_TIMER_DISARM(tp, TCPT_REXMT); | | 2031 | TCP_TIMER_DISARM(tp, TCPT_REXMT); |
2036 | else if (TCP_TIMER_ISARMED(tp, | | 2032 | else if (TCP_TIMER_ISARMED(tp, |
2037 | TCPT_PERSIST) == 0) | | 2033 | TCPT_PERSIST) == 0) |
2038 | TCP_TIMER_ARM(tp, TCPT_REXMT, | | 2034 | TCP_TIMER_ARM(tp, TCPT_REXMT, |
2039 | tp->t_rxtcur); | | 2035 | tp->t_rxtcur); |
2040 | | | 2036 | |
2041 | sowwakeup(so); | | 2037 | sowwakeup(so); |
2042 | if (so->so_snd.sb_cc) { | | 2038 | if (so->so_snd.sb_cc) { |
2043 | KERNEL_LOCK(1, NULL); | | 2039 | KERNEL_LOCK(1, NULL); |
2044 | (void) tcp_output(tp); | | 2040 | (void) tcp_output(tp); |
2045 | KERNEL_UNLOCK_ONE(NULL); | | 2041 | KERNEL_UNLOCK_ONE(NULL); |
2046 | } | | 2042 | } |
2047 | if (tcp_saveti) | | 2043 | if (tcp_saveti) |
2048 | m_freem(tcp_saveti); | | 2044 | m_freem(tcp_saveti); |
2049 | return; | | 2045 | return; |
2050 | } | | 2046 | } |
2051 | } else if (th->th_ack == tp->snd_una && | | 2047 | } else if (th->th_ack == tp->snd_una && |
2052 | TAILQ_FIRST(&tp->segq) == NULL && | | 2048 | TAILQ_FIRST(&tp->segq) == NULL && |
2053 | tlen <= sbspace(&so->so_rcv)) { | | 2049 | tlen <= sbspace(&so->so_rcv)) { |
2054 | int newsize = 0; /* automatic sockbuf scaling */ | | 2050 | int newsize = 0; /* automatic sockbuf scaling */ |
2055 | | | 2051 | |
2056 | /* | | 2052 | /* |
2057 | * this is a pure, in-sequence data packet | | 2053 | * this is a pure, in-sequence data packet |
2058 | * with nothing on the reassembly queue and | | 2054 | * with nothing on the reassembly queue and |
2059 | * we have enough buffer space to take it. | | 2055 | * we have enough buffer space to take it. |
2060 | */ | | 2056 | */ |
2061 | tp->rcv_nxt += tlen; | | 2057 | tp->rcv_nxt += tlen; |
2062 | tcps = TCP_STAT_GETREF(); | | 2058 | tcps = TCP_STAT_GETREF(); |
2063 | tcps[TCP_STAT_PREDDAT]++; | | 2059 | tcps[TCP_STAT_PREDDAT]++; |
2064 | tcps[TCP_STAT_RCVPACK]++; | | 2060 | tcps[TCP_STAT_RCVPACK]++; |
2065 | tcps[TCP_STAT_RCVBYTE] += tlen; | | 2061 | tcps[TCP_STAT_RCVBYTE] += tlen; |
2066 | TCP_STAT_PUTREF(); | | 2062 | TCP_STAT_PUTREF(); |
2067 | nd6_hint(tp); | | 2063 | nd6_hint(tp); |
2068 | | | 2064 | |
2069 | /* | | 2065 | /* |
2070 | * Automatic sizing enables the performance of large buffers | | 2066 | * Automatic sizing enables the performance of large buffers |
2071 | * and most of the efficiency of small ones by only allocating | | 2067 | * and most of the efficiency of small ones by only allocating |
2072 | * space when it is needed. | | 2068 | * space when it is needed. |
2073 | * | | 2069 | * |
2074 | * On the receive side the socket buffer memory is only rarely | | 2070 | * On the receive side the socket buffer memory is only rarely |
2075 | * used to any significant extent. This allows us to be much | | 2071 | * used to any significant extent. This allows us to be much |
2076 | * more aggressive in scaling the receive socket buffer. For | | 2072 | * more aggressive in scaling the receive socket buffer. For |
2077 | * the case that the buffer space is actually used to a large | | 2073 | * the case that the buffer space is actually used to a large |
2078 | * extent and we run out of kernel memory we can simply drop | | 2074 | * extent and we run out of kernel memory we can simply drop |
2079 | * the new segments; TCP on the sender will just retransmit it | | 2075 | * the new segments; TCP on the sender will just retransmit it |
2080 | * later. Setting the buffer size too big may only consume too | | 2076 | * later. Setting the buffer size too big may only consume too |
2081 | * much kernel memory if the application doesn't read() from | | 2077 | * much kernel memory if the application doesn't read() from |
2082 | * the socket or packet loss or reordering makes use of the | | 2078 | * the socket or packet loss or reordering makes use of the |
2083 | * reassembly queue. | | 2079 | * reassembly queue. |
2084 | * | | 2080 | * |
2085 | * The criteria to step up the receive buffer one notch are: | | 2081 | * The criteria to step up the receive buffer one notch are: |
2086 | * 1. the number of bytes received during the time it takes | | 2082 | * 1. the number of bytes received during the time it takes |
2087 | * one timestamp to be reflected back to us (the RTT); | | 2083 | * one timestamp to be reflected back to us (the RTT); |
2088 | * 2. received bytes per RTT is within seven eighth of the | | 2084 | * 2. received bytes per RTT is within seven eighth of the |
2089 | * current socket buffer size; | | 2085 | * current socket buffer size; |
2090 | * 3. receive buffer size has not hit maximal automatic size; | | 2086 | * 3. receive buffer size has not hit maximal automatic size; |
2091 | * | | 2087 | * |
2092 | * This algorithm does one step per RTT at most and only if | | 2088 | * This algorithm does one step per RTT at most and only if |
2093 | * we receive a bulk stream w/o packet losses or reorderings. | | 2089 | * we receive a bulk stream w/o packet losses or reorderings. |
2094 | * Shrinking the buffer during idle times is not necessary as | | 2090 | * Shrinking the buffer during idle times is not necessary as |
2095 | * it doesn't consume any memory when idle. | | 2091 | * it doesn't consume any memory when idle. |
2096 | * | | 2092 | * |
2097 | * TODO: Only step up if the application is actually serving | | 2093 | * TODO: Only step up if the application is actually serving |
2098 | * the buffer to better manage the socket buffer resources. | | 2094 | * the buffer to better manage the socket buffer resources. |
2099 | */ | | 2095 | */ |
2100 | if (tcp_do_autorcvbuf && | | 2096 | if (tcp_do_autorcvbuf && |
2101 | opti.ts_ecr && | | 2097 | opti.ts_ecr && |
2102 | (so->so_rcv.sb_flags & SB_AUTOSIZE)) { | | 2098 | (so->so_rcv.sb_flags & SB_AUTOSIZE)) { |
2103 | if (opti.ts_ecr > tp->rfbuf_ts && | | 2099 | if (opti.ts_ecr > tp->rfbuf_ts && |
2104 | opti.ts_ecr - tp->rfbuf_ts < PR_SLOWHZ) { | | 2100 | opti.ts_ecr - tp->rfbuf_ts < PR_SLOWHZ) { |
2105 | if (tp->rfbuf_cnt > | | 2101 | if (tp->rfbuf_cnt > |
2106 | (so->so_rcv.sb_hiwat / 8 * 7) && | | 2102 | (so->so_rcv.sb_hiwat / 8 * 7) && |
2107 | so->so_rcv.sb_hiwat < | | 2103 | so->so_rcv.sb_hiwat < |
2108 | tcp_autorcvbuf_max) { | | 2104 | tcp_autorcvbuf_max) { |
2109 | newsize = | | 2105 | newsize = |
2110 | min(so->so_rcv.sb_hiwat + | | 2106 | min(so->so_rcv.sb_hiwat + |
2111 | tcp_autorcvbuf_inc, | | 2107 | tcp_autorcvbuf_inc, |
2112 | tcp_autorcvbuf_max); | | 2108 | tcp_autorcvbuf_max); |
2113 | } | | 2109 | } |
2114 | /* Start over with next RTT. */ | | 2110 | /* Start over with next RTT. */ |
2115 | tp->rfbuf_ts = 0; | | 2111 | tp->rfbuf_ts = 0; |
2116 | tp->rfbuf_cnt = 0; | | 2112 | tp->rfbuf_cnt = 0; |
2117 | } else | | 2113 | } else |
2118 | tp->rfbuf_cnt += tlen; /* add up */ | | 2114 | tp->rfbuf_cnt += tlen; /* add up */ |
2119 | } | | 2115 | } |
2120 | | | 2116 | |
2121 | /* | | 2117 | /* |
2122 | * Drop TCP, IP headers and TCP options then add data | | 2118 | * Drop TCP, IP headers and TCP options then add data |
2123 | * to socket buffer. | | 2119 | * to socket buffer. |
2124 | */ | | 2120 | */ |
2125 | if (so->so_state & SS_CANTRCVMORE) | | 2121 | if (so->so_state & SS_CANTRCVMORE) |
2126 | m_freem(m); | | 2122 | m_freem(m); |
2127 | else { | | 2123 | else { |
2128 | /* | | 2124 | /* |
2129 | * Set new socket buffer size. | | 2125 | * Set new socket buffer size. |
2130 | * Give up when limit is reached. | | 2126 | * Give up when limit is reached. |
2131 | */ | | 2127 | */ |
2132 | if (newsize) | | 2128 | if (newsize) |
2133 | if (!sbreserve(&so->so_rcv, | | 2129 | if (!sbreserve(&so->so_rcv, |
2134 | newsize, so)) | | 2130 | newsize, so)) |
2135 | so->so_rcv.sb_flags &= ~SB_AUTOSIZE; | | 2131 | so->so_rcv.sb_flags &= ~SB_AUTOSIZE; |
2136 | m_adj(m, toff + off); | | 2132 | m_adj(m, toff + off); |
2137 | sbappendstream(&so->so_rcv, m); | | 2133 | sbappendstream(&so->so_rcv, m); |
2138 | } | | 2134 | } |
2139 | sorwakeup(so); | | 2135 | sorwakeup(so); |
2140 | tcp_setup_ack(tp, th); | | 2136 | tcp_setup_ack(tp, th); |
2141 | if (tp->t_flags & TF_ACKNOW) { | | 2137 | if (tp->t_flags & TF_ACKNOW) { |
2142 | KERNEL_LOCK(1, NULL); | | 2138 | KERNEL_LOCK(1, NULL); |
2143 | (void) tcp_output(tp); | | 2139 | (void) tcp_output(tp); |
2144 | KERNEL_UNLOCK_ONE(NULL); | | 2140 | KERNEL_UNLOCK_ONE(NULL); |
2145 | } | | 2141 | } |
2146 | if (tcp_saveti) | | 2142 | if (tcp_saveti) |
2147 | m_freem(tcp_saveti); | | 2143 | m_freem(tcp_saveti); |
2148 | return; | | 2144 | return; |
2149 | } | | 2145 | } |
2150 | } | | 2146 | } |
2151 | | | 2147 | |
2152 | /* | | 2148 | /* |
2153 | * Compute mbuf offset to TCP data segment. | | 2149 | * Compute mbuf offset to TCP data segment. |
2154 | */ | | 2150 | */ |
2155 | hdroptlen = toff + off; | | 2151 | hdroptlen = toff + off; |
2156 | | | 2152 | |
2157 | /* | | 2153 | /* |
2158 | * Calculate amount of space in receive window, | | 2154 | * Calculate amount of space in receive window, |
2159 | * and then do TCP input processing. | | 2155 | * and then do TCP input processing. |
2160 | * Receive window is amount of space in rcv queue, | | 2156 | * Receive window is amount of space in rcv queue, |
2161 | * but not less than advertised window. | | 2157 | * but not less than advertised window. |
2162 | */ | | 2158 | */ |
2163 | { int win; | | 2159 | { int win; |
2164 | | | 2160 | |
2165 | win = sbspace(&so->so_rcv); | | 2161 | win = sbspace(&so->so_rcv); |
2166 | if (win < 0) | | 2162 | if (win < 0) |
2167 | win = 0; | | 2163 | win = 0; |
2168 | tp->rcv_wnd = imax(win, (int)(tp->rcv_adv - tp->rcv_nxt)); | | 2164 | tp->rcv_wnd = imax(win, (int)(tp->rcv_adv - tp->rcv_nxt)); |
2169 | } | | 2165 | } |
2170 | | | 2166 | |
2171 | /* Reset receive buffer auto scaling when not in bulk receive mode. */ | | 2167 | /* Reset receive buffer auto scaling when not in bulk receive mode. */ |
2172 | tp->rfbuf_ts = 0; | | 2168 | tp->rfbuf_ts = 0; |
2173 | tp->rfbuf_cnt = 0; | | 2169 | tp->rfbuf_cnt = 0; |
2174 | | | 2170 | |
2175 | switch (tp->t_state) { | | 2171 | switch (tp->t_state) { |
2176 | /* | | 2172 | /* |
2177 | * If the state is SYN_SENT: | | 2173 | * If the state is SYN_SENT: |
2178 | * if seg contains an ACK, but not for our SYN, drop the input. | | 2174 | * if seg contains an ACK, but not for our SYN, drop the input. |
2179 | * if seg contains a RST, then drop the connection. | | 2175 | * if seg contains a RST, then drop the connection. |
2180 | * if seg does not contain SYN, then drop it. | | 2176 | * if seg does not contain SYN, then drop it. |
2181 | * Otherwise this is an acceptable SYN segment | | 2177 | * Otherwise this is an acceptable SYN segment |
2182 | * initialize tp->rcv_nxt and tp->irs | | 2178 | * initialize tp->rcv_nxt and tp->irs |
2183 | * if seg contains ack then advance tp->snd_una | | 2179 | * if seg contains ack then advance tp->snd_una |
2184 | * if seg contains a ECE and ECN support is enabled, the stream | | 2180 | * if seg contains a ECE and ECN support is enabled, the stream |
2185 | * is ECN capable. | | 2181 | * is ECN capable. |
2186 | * if SYN has been acked change to ESTABLISHED else SYN_RCVD state | | 2182 | * if SYN has been acked change to ESTABLISHED else SYN_RCVD state |
2187 | * arrange for segment to be acked (eventually) | | 2183 | * arrange for segment to be acked (eventually) |
2188 | * continue processing rest of data/controls, beginning with URG | | 2184 | * continue processing rest of data/controls, beginning with URG |
2189 | */ | | 2185 | */ |
2190 | case TCPS_SYN_SENT: | | 2186 | case TCPS_SYN_SENT: |
2191 | if ((tiflags & TH_ACK) && | | 2187 | if ((tiflags & TH_ACK) && |
2192 | (SEQ_LEQ(th->th_ack, tp->iss) || | | 2188 | (SEQ_LEQ(th->th_ack, tp->iss) || |
2193 | SEQ_GT(th->th_ack, tp->snd_max))) | | 2189 | SEQ_GT(th->th_ack, tp->snd_max))) |
2194 | goto dropwithreset; | | 2190 | goto dropwithreset; |
2195 | if (tiflags & TH_RST) { | | 2191 | if (tiflags & TH_RST) { |
2196 | if (tiflags & TH_ACK) | | 2192 | if (tiflags & TH_ACK) |
2197 | tp = tcp_drop(tp, ECONNREFUSED); | | 2193 | tp = tcp_drop(tp, ECONNREFUSED); |
2198 | goto drop; | | 2194 | goto drop; |
2199 | } | | 2195 | } |
2200 | if ((tiflags & TH_SYN) == 0) | | 2196 | if ((tiflags & TH_SYN) == 0) |
2201 | goto drop; | | 2197 | goto drop; |
2202 | if (tiflags & TH_ACK) { | | 2198 | if (tiflags & TH_ACK) { |
2203 | tp->snd_una = th->th_ack; | | 2199 | tp->snd_una = th->th_ack; |
2204 | if (SEQ_LT(tp->snd_nxt, tp->snd_una)) | | 2200 | if (SEQ_LT(tp->snd_nxt, tp->snd_una)) |
2205 | tp->snd_nxt = tp->snd_una; | | 2201 | tp->snd_nxt = tp->snd_una; |
2206 | if (SEQ_LT(tp->snd_high, tp->snd_una)) | | 2202 | if (SEQ_LT(tp->snd_high, tp->snd_una)) |
2207 | tp->snd_high = tp->snd_una; | | 2203 | tp->snd_high = tp->snd_una; |
2208 | TCP_TIMER_DISARM(tp, TCPT_REXMT); | | 2204 | TCP_TIMER_DISARM(tp, TCPT_REXMT); |
2209 | | | 2205 | |
2210 | if ((tiflags & TH_ECE) && tcp_do_ecn) { | | 2206 | if ((tiflags & TH_ECE) && tcp_do_ecn) { |
2211 | tp->t_flags |= TF_ECN_PERMIT; | | 2207 | tp->t_flags |= TF_ECN_PERMIT; |
2212 | TCP_STATINC(TCP_STAT_ECN_SHS); | | 2208 | TCP_STATINC(TCP_STAT_ECN_SHS); |
2213 | } | | 2209 | } |
2214 | | | 2210 | |
2215 | } | | 2211 | } |
2216 | tp->irs = th->th_seq; | | 2212 | tp->irs = th->th_seq; |
2217 | tcp_rcvseqinit(tp); | | 2213 | tcp_rcvseqinit(tp); |
2218 | tp->t_flags |= TF_ACKNOW; | | 2214 | tp->t_flags |= TF_ACKNOW; |
2219 | tcp_mss_from_peer(tp, opti.maxseg); | | 2215 | tcp_mss_from_peer(tp, opti.maxseg); |
2220 | | | 2216 | |
2221 | /* | | 2217 | /* |
2222 | * Initialize the initial congestion window. If we | | 2218 | * Initialize the initial congestion window. If we |
2223 | * had to retransmit the SYN, we must initialize cwnd | | 2219 | * had to retransmit the SYN, we must initialize cwnd |
2224 | * to 1 segment (i.e. the Loss Window). | | 2220 | * to 1 segment (i.e. the Loss Window). |
2225 | */ | | 2221 | */ |
2226 | if (tp->t_flags & TF_SYN_REXMT) | | 2222 | if (tp->t_flags & TF_SYN_REXMT) |
2227 | tp->snd_cwnd = tp->t_peermss; | | 2223 | tp->snd_cwnd = tp->t_peermss; |
2228 | else { | | 2224 | else { |
2229 | int ss = tcp_init_win; | | 2225 | int ss = tcp_init_win; |
2230 | #ifdef INET | | 2226 | #ifdef INET |
2231 | if (inp != NULL && in_localaddr(inp->inp_faddr)) | | 2227 | if (inp != NULL && in_localaddr(inp->inp_faddr)) |
2232 | ss = tcp_init_win_local; | | 2228 | ss = tcp_init_win_local; |
2233 | #endif | | 2229 | #endif |
2234 | #ifdef INET6 | | 2230 | #ifdef INET6 |
2235 | if (in6p != NULL && in6_localaddr(&in6p->in6p_faddr)) | | 2231 | if (in6p != NULL && in6_localaddr(&in6p->in6p_faddr)) |
2236 | ss = tcp_init_win_local; | | 2232 | ss = tcp_init_win_local; |
2237 | #endif | | 2233 | #endif |
2238 | tp->snd_cwnd = TCP_INITIAL_WINDOW(ss, tp->t_peermss); | | 2234 | tp->snd_cwnd = TCP_INITIAL_WINDOW(ss, tp->t_peermss); |
2239 | } | | 2235 | } |
2240 | | | 2236 | |
2241 | tcp_rmx_rtt(tp); | | 2237 | tcp_rmx_rtt(tp); |
2242 | if (tiflags & TH_ACK) { | | 2238 | if (tiflags & TH_ACK) { |
2243 | TCP_STATINC(TCP_STAT_CONNECTS); | | 2239 | TCP_STATINC(TCP_STAT_CONNECTS); |
2244 | /* | | 2240 | /* |
2245 | * move tcp_established before soisconnected | | 2241 | * move tcp_established before soisconnected |
2246 | * because upcall handler can drive tcp_output | | 2242 | * because upcall handler can drive tcp_output |
2247 | * functionality. | | 2243 | * functionality. |
2248 | * XXX we might call soisconnected at the end of | | 2244 | * XXX we might call soisconnected at the end of |
2249 | * all processing | | 2245 | * all processing |
2250 | */ | | 2246 | */ |
2251 | tcp_established(tp); | | 2247 | tcp_established(tp); |
2252 | soisconnected(so); | | 2248 | soisconnected(so); |
2253 | /* Do window scaling on this connection? */ | | 2249 | /* Do window scaling on this connection? */ |
2254 | if ((tp->t_flags & (TF_RCVD_SCALE|TF_REQ_SCALE)) == | | 2250 | if ((tp->t_flags & (TF_RCVD_SCALE|TF_REQ_SCALE)) == |
2255 | (TF_RCVD_SCALE|TF_REQ_SCALE)) { | | 2251 | (TF_RCVD_SCALE|TF_REQ_SCALE)) { |
2256 | tp->snd_scale = tp->requested_s_scale; | | 2252 | tp->snd_scale = tp->requested_s_scale; |
2257 | tp->rcv_scale = tp->request_r_scale; | | 2253 | tp->rcv_scale = tp->request_r_scale; |
2258 | } | | 2254 | } |
2259 | TCP_REASS_LOCK(tp); | | 2255 | TCP_REASS_LOCK(tp); |
2260 | (void) tcp_reass(tp, NULL, NULL, &tlen); | | 2256 | (void) tcp_reass(tp, NULL, NULL, &tlen); |
2261 | /* | | 2257 | /* |
2262 | * if we didn't have to retransmit the SYN, | | 2258 | * if we didn't have to retransmit the SYN, |
2263 | * use its rtt as our initial srtt & rtt var. | | 2259 | * use its rtt as our initial srtt & rtt var. |
2264 | */ | | 2260 | */ |
2265 | if (tp->t_rtttime) | | 2261 | if (tp->t_rtttime) |
2266 | tcp_xmit_timer(tp, tcp_now - tp->t_rtttime); | | 2262 | tcp_xmit_timer(tp, tcp_now - tp->t_rtttime); |
2267 | } else | | 2263 | } else |
2268 | tp->t_state = TCPS_SYN_RECEIVED; | | 2264 | tp->t_state = TCPS_SYN_RECEIVED; |
2269 | | | 2265 | |
2270 | /* | | 2266 | /* |
2271 | * Advance th->th_seq to correspond to first data byte. | | 2267 | * Advance th->th_seq to correspond to first data byte. |
2272 | * If data, trim to stay within window, | | 2268 | * If data, trim to stay within window, |
2273 | * dropping FIN if necessary. | | 2269 | * dropping FIN if necessary. |
2274 | */ | | 2270 | */ |
2275 | th->th_seq++; | | 2271 | th->th_seq++; |
2276 | if (tlen > tp->rcv_wnd) { | | 2272 | if (tlen > tp->rcv_wnd) { |
2277 | todrop = tlen - tp->rcv_wnd; | | 2273 | todrop = tlen - tp->rcv_wnd; |
2278 | m_adj(m, -todrop); | | 2274 | m_adj(m, -todrop); |
2279 | tlen = tp->rcv_wnd; | | 2275 | tlen = tp->rcv_wnd; |
2280 | tiflags &= ~TH_FIN; | | 2276 | tiflags &= ~TH_FIN; |
2281 | tcps = TCP_STAT_GETREF(); | | 2277 | tcps = TCP_STAT_GETREF(); |
2282 | tcps[TCP_STAT_RCVPACKAFTERWIN]++; | | 2278 | tcps[TCP_STAT_RCVPACKAFTERWIN]++; |
2283 | tcps[TCP_STAT_RCVBYTEAFTERWIN] += todrop; | | 2279 | tcps[TCP_STAT_RCVBYTEAFTERWIN] += todrop; |
2284 | TCP_STAT_PUTREF(); | | 2280 | TCP_STAT_PUTREF(); |
2285 | } | | 2281 | } |
2286 | tp->snd_wl1 = th->th_seq - 1; | | 2282 | tp->snd_wl1 = th->th_seq - 1; |
2287 | tp->rcv_up = th->th_seq; | | 2283 | tp->rcv_up = th->th_seq; |
2288 | goto step6; | | 2284 | goto step6; |
2289 | | | 2285 | |
2290 | /* | | 2286 | /* |
2291 | * If the state is SYN_RECEIVED: | | 2287 | * If the state is SYN_RECEIVED: |
2292 | * If seg contains an ACK, but not for our SYN, drop the input | | 2288 | * If seg contains an ACK, but not for our SYN, drop the input |
2293 | * and generate an RST. See page 36, rfc793 | | 2289 | * and generate an RST. See page 36, rfc793 |
2294 | */ | | 2290 | */ |
2295 | case TCPS_SYN_RECEIVED: | | 2291 | case TCPS_SYN_RECEIVED: |
2296 | if ((tiflags & TH_ACK) && | | 2292 | if ((tiflags & TH_ACK) && |
2297 | (SEQ_LEQ(th->th_ack, tp->iss) || | | 2293 | (SEQ_LEQ(th->th_ack, tp->iss) || |
2298 | SEQ_GT(th->th_ack, tp->snd_max))) | | 2294 | SEQ_GT(th->th_ack, tp->snd_max))) |
2299 | goto dropwithreset; | | 2295 | goto dropwithreset; |
2300 | break; | | 2296 | break; |
2301 | } | | 2297 | } |
2302 | | | 2298 | |
2303 | /* | | 2299 | /* |
2304 | * States other than LISTEN or SYN_SENT. | | 2300 | * States other than LISTEN or SYN_SENT. |
2305 | * First check timestamp, if present. | | 2301 | * First check timestamp, if present. |
2306 | * Then check that at least some bytes of segment are within | | 2302 | * Then check that at least some bytes of segment are within |
2307 | * receive window. If segment begins before rcv_nxt, | | 2303 | * receive window. If segment begins before rcv_nxt, |
2308 | * drop leading data (and SYN); if nothing left, just ack. | | 2304 | * drop leading data (and SYN); if nothing left, just ack. |
2309 | * | | 2305 | * |
2310 | * RFC 1323 PAWS: If we have a timestamp reply on this segment | | 2306 | * RFC 1323 PAWS: If we have a timestamp reply on this segment |
2311 | * and it's less than ts_recent, drop it. | | 2307 | * and it's less than ts_recent, drop it. |
2312 | */ | | 2308 | */ |
2313 | if (opti.ts_present && (tiflags & TH_RST) == 0 && tp->ts_recent && | | 2309 | if (opti.ts_present && (tiflags & TH_RST) == 0 && tp->ts_recent && |
2314 | TSTMP_LT(opti.ts_val, tp->ts_recent)) { | | 2310 | TSTMP_LT(opti.ts_val, tp->ts_recent)) { |
2315 | | | 2311 | |
2316 | /* Check to see if ts_recent is over 24 days old. */ | | 2312 | /* Check to see if ts_recent is over 24 days old. */ |
2317 | if (tcp_now - tp->ts_recent_age > TCP_PAWS_IDLE) { | | 2313 | if (tcp_now - tp->ts_recent_age > TCP_PAWS_IDLE) { |
2318 | /* | | 2314 | /* |
2319 | * Invalidate ts_recent. If this segment updates | | 2315 | * Invalidate ts_recent. If this segment updates |
2320 | * ts_recent, the age will be reset later and ts_recent | | 2316 | * ts_recent, the age will be reset later and ts_recent |
2321 | * will get a valid value. If it does not, setting | | 2317 | * will get a valid value. If it does not, setting |
2322 | * ts_recent to zero will at least satisfy the | | 2318 | * ts_recent to zero will at least satisfy the |
2323 | * requirement that zero be placed in the timestamp | | 2319 | * requirement that zero be placed in the timestamp |
2324 | * echo reply when ts_recent isn't valid. The | | 2320 | * echo reply when ts_recent isn't valid. The |
2325 | * age isn't reset until we get a valid ts_recent | | 2321 | * age isn't reset until we get a valid ts_recent |
2326 | * because we don't want out-of-order segments to be | | 2322 | * because we don't want out-of-order segments to be |
2327 | * dropped when ts_recent is old. | | 2323 | * dropped when ts_recent is old. |
2328 | */ | | 2324 | */ |
2329 | tp->ts_recent = 0; | | 2325 | tp->ts_recent = 0; |
2330 | } else { | | 2326 | } else { |
2331 | tcps = TCP_STAT_GETREF(); | | 2327 | tcps = TCP_STAT_GETREF(); |
2332 | tcps[TCP_STAT_RCVDUPPACK]++; | | 2328 | tcps[TCP_STAT_RCVDUPPACK]++; |
2333 | tcps[TCP_STAT_RCVDUPBYTE] += tlen; | | 2329 | tcps[TCP_STAT_RCVDUPBYTE] += tlen; |
2334 | tcps[TCP_STAT_PAWSDROP]++; | | 2330 | tcps[TCP_STAT_PAWSDROP]++; |
2335 | TCP_STAT_PUTREF(); | | 2331 | TCP_STAT_PUTREF(); |
2336 | tcp_new_dsack(tp, th->th_seq, tlen); | | 2332 | tcp_new_dsack(tp, th->th_seq, tlen); |
2337 | goto dropafterack; | | 2333 | goto dropafterack; |
2338 | } | | 2334 | } |
2339 | } | | 2335 | } |
2340 | | | 2336 | |
2341 | todrop = tp->rcv_nxt - th->th_seq; | | 2337 | todrop = tp->rcv_nxt - th->th_seq; |
2342 | dupseg = false; | | 2338 | dupseg = false; |
2343 | if (todrop > 0) { | | 2339 | if (todrop > 0) { |
2344 | if (tiflags & TH_SYN) { | | 2340 | if (tiflags & TH_SYN) { |
2345 | tiflags &= ~TH_SYN; | | 2341 | tiflags &= ~TH_SYN; |
2346 | th->th_seq++; | | 2342 | th->th_seq++; |
2347 | if (th->th_urp > 1) | | 2343 | if (th->th_urp > 1) |
2348 | th->th_urp--; | | 2344 | th->th_urp--; |
2349 | else { | | 2345 | else { |
2350 | tiflags &= ~TH_URG; | | 2346 | tiflags &= ~TH_URG; |
2351 | th->th_urp = 0; | | 2347 | th->th_urp = 0; |
2352 | } | | 2348 | } |
2353 | todrop--; | | 2349 | todrop--; |
2354 | } | | 2350 | } |
2355 | if (todrop > tlen || | | 2351 | if (todrop > tlen || |
2356 | (todrop == tlen && (tiflags & TH_FIN) == 0)) { | | 2352 | (todrop == tlen && (tiflags & TH_FIN) == 0)) { |
2357 | /* | | 2353 | /* |
2358 | * Any valid FIN or RST must be to the left of the | | 2354 | * Any valid FIN or RST must be to the left of the |
2359 | * window. At this point the FIN or RST must be a | | 2355 | * window. At this point the FIN or RST must be a |
2360 | * duplicate or out of sequence; drop it. | | 2356 | * duplicate or out of sequence; drop it. |
2361 | */ | | 2357 | */ |
2362 | if (tiflags & TH_RST) | | 2358 | if (tiflags & TH_RST) |
2363 | goto drop; | | 2359 | goto drop; |
2364 | tiflags &= ~(TH_FIN|TH_RST); | | 2360 | tiflags &= ~(TH_FIN|TH_RST); |
2365 | /* | | 2361 | /* |
2366 | * Send an ACK to resynchronize and drop any data. | | 2362 | * Send an ACK to resynchronize and drop any data. |
2367 | * But keep on processing for RST or ACK. | | 2363 | * But keep on processing for RST or ACK. |
2368 | */ | | 2364 | */ |
2369 | tp->t_flags |= TF_ACKNOW; | | 2365 | tp->t_flags |= TF_ACKNOW; |
2370 | todrop = tlen; | | 2366 | todrop = tlen; |
2371 | dupseg = true; | | 2367 | dupseg = true; |
2372 | tcps = TCP_STAT_GETREF(); | | 2368 | tcps = TCP_STAT_GETREF(); |
2373 | tcps[TCP_STAT_RCVDUPPACK]++; | | 2369 | tcps[TCP_STAT_RCVDUPPACK]++; |
2374 | tcps[TCP_STAT_RCVDUPBYTE] += todrop; | | 2370 | tcps[TCP_STAT_RCVDUPBYTE] += todrop; |
2375 | TCP_STAT_PUTREF(); | | 2371 | TCP_STAT_PUTREF(); |
2376 | } else if ((tiflags & TH_RST) && | | 2372 | } else if ((tiflags & TH_RST) && |
2377 | th->th_seq != tp->rcv_nxt) { | | 2373 | th->th_seq != tp->rcv_nxt) { |
2378 | /* | | 2374 | /* |
2379 | * Test for reset before adjusting the sequence | | 2375 | * Test for reset before adjusting the sequence |
2380 | * number for overlapping data. | | 2376 | * number for overlapping data. |
2381 | */ | | 2377 | */ |
2382 | goto dropafterack_ratelim; | | 2378 | goto dropafterack_ratelim; |
2383 | } else { | | 2379 | } else { |
2384 | tcps = TCP_STAT_GETREF(); | | 2380 | tcps = TCP_STAT_GETREF(); |
2385 | tcps[TCP_STAT_RCVPARTDUPPACK]++; | | 2381 | tcps[TCP_STAT_RCVPARTDUPPACK]++; |
2386 | tcps[TCP_STAT_RCVPARTDUPBYTE] += todrop; | | 2382 | tcps[TCP_STAT_RCVPARTDUPBYTE] += todrop; |
2387 | TCP_STAT_PUTREF(); | | 2383 | TCP_STAT_PUTREF(); |
2388 | } | | 2384 | } |
2389 | tcp_new_dsack(tp, th->th_seq, todrop); | | 2385 | tcp_new_dsack(tp, th->th_seq, todrop); |
2390 | hdroptlen += todrop; /*drop from head afterwards*/ | | 2386 | hdroptlen += todrop; /*drop from head afterwards*/ |
2391 | th->th_seq += todrop; | | 2387 | th->th_seq += todrop; |
2392 | tlen -= todrop; | | 2388 | tlen -= todrop; |
2393 | if (th->th_urp > todrop) | | 2389 | if (th->th_urp > todrop) |
2394 | th->th_urp -= todrop; | | 2390 | th->th_urp -= todrop; |
2395 | else { | | 2391 | else { |
2396 | tiflags &= ~TH_URG; | | 2392 | tiflags &= ~TH_URG; |
2397 | th->th_urp = 0; | | 2393 | th->th_urp = 0; |
2398 | } | | 2394 | } |
2399 | } | | 2395 | } |
2400 | | | 2396 | |
2401 | /* | | 2397 | /* |
2402 | * If new data are received on a connection after the | | 2398 | * If new data are received on a connection after the |
2403 | * user processes are gone, then RST the other end. | | 2399 | * user processes are gone, then RST the other end. |
2404 | */ | | 2400 | */ |
2405 | if ((so->so_state & SS_NOFDREF) && | | 2401 | if ((so->so_state & SS_NOFDREF) && |
2406 | tp->t_state > TCPS_CLOSE_WAIT && tlen) { | | 2402 | tp->t_state > TCPS_CLOSE_WAIT && tlen) { |
2407 | tp = tcp_close(tp); | | 2403 | tp = tcp_close(tp); |
2408 | TCP_STATINC(TCP_STAT_RCVAFTERCLOSE); | | 2404 | TCP_STATINC(TCP_STAT_RCVAFTERCLOSE); |
2409 | goto dropwithreset; | | 2405 | goto dropwithreset; |
2410 | } | | 2406 | } |
2411 | | | 2407 | |
2412 | /* | | 2408 | /* |
2413 | * If segment ends after window, drop trailing data | | 2409 | * If segment ends after window, drop trailing data |
2414 | * (and PUSH and FIN); if nothing left, just ACK. | | 2410 | * (and PUSH and FIN); if nothing left, just ACK. |
2415 | */ | | 2411 | */ |
2416 | todrop = (th->th_seq + tlen) - (tp->rcv_nxt+tp->rcv_wnd); | | 2412 | todrop = (th->th_seq + tlen) - (tp->rcv_nxt+tp->rcv_wnd); |
2417 | if (todrop > 0) { | | 2413 | if (todrop > 0) { |
2418 | TCP_STATINC(TCP_STAT_RCVPACKAFTERWIN); | | 2414 | TCP_STATINC(TCP_STAT_RCVPACKAFTERWIN); |
2419 | if (todrop >= tlen) { | | 2415 | if (todrop >= tlen) { |
2420 | /* | | 2416 | /* |
2421 | * The segment actually starts after the window. | | 2417 | * The segment actually starts after the window. |
2422 | * th->th_seq + tlen - tp->rcv_nxt - tp->rcv_wnd >= tlen | | 2418 | * th->th_seq + tlen - tp->rcv_nxt - tp->rcv_wnd >= tlen |
2423 | * th->th_seq - tp->rcv_nxt - tp->rcv_wnd >= 0 | | 2419 | * th->th_seq - tp->rcv_nxt - tp->rcv_wnd >= 0 |
2424 | * th->th_seq >= tp->rcv_nxt + tp->rcv_wnd | | 2420 | * th->th_seq >= tp->rcv_nxt + tp->rcv_wnd |
2425 | */ | | 2421 | */ |
2426 | TCP_STATADD(TCP_STAT_RCVBYTEAFTERWIN, tlen); | | 2422 | TCP_STATADD(TCP_STAT_RCVBYTEAFTERWIN, tlen); |
2427 | /* | | 2423 | /* |
2428 | * If a new connection request is received | | 2424 | * If a new connection request is received |
2429 | * while in TIME_WAIT, drop the old connection | | 2425 | * while in TIME_WAIT, drop the old connection |
2430 | * and start over if the sequence numbers | | 2426 | * and start over if the sequence numbers |
2431 | * are above the previous ones. | | 2427 | * are above the previous ones. |
2432 | * | | 2428 | * |
2433 | * NOTE: We will checksum the packet again, and | | 2429 | * NOTE: We will checksum the packet again, and |
2434 | * so we need to put the header fields back into | | 2430 | * so we need to put the header fields back into |
2435 | * network order! | | 2431 | * network order! |
2436 | * XXX This kind of sucks, but we don't expect | | 2432 | * XXX This kind of sucks, but we don't expect |
2437 | * XXX this to happen very often, so maybe it | | 2433 | * XXX this to happen very often, so maybe it |
2438 | * XXX doesn't matter so much. | | 2434 | * XXX doesn't matter so much. |
2439 | */ | | 2435 | */ |
2440 | if (tiflags & TH_SYN && | | 2436 | if (tiflags & TH_SYN && |
2441 | tp->t_state == TCPS_TIME_WAIT && | | 2437 | tp->t_state == TCPS_TIME_WAIT && |
2442 | SEQ_GT(th->th_seq, tp->rcv_nxt)) { | | 2438 | SEQ_GT(th->th_seq, tp->rcv_nxt)) { |
2443 | tp = tcp_close(tp); | | 2439 | tp = tcp_close(tp); |
2444 | tcp_fields_to_net(th); | | 2440 | tcp_fields_to_net(th); |
2445 | goto findpcb; | | 2441 | goto findpcb; |
2446 | } | | 2442 | } |
2447 | /* | | 2443 | /* |
2448 | * If window is closed can only take segments at | | 2444 | * If window is closed can only take segments at |
2449 | * window edge, and have to drop data and PUSH from | | 2445 | * window edge, and have to drop data and PUSH from |
2450 | * incoming segments. Continue processing, but | | 2446 | * incoming segments. Continue processing, but |
2451 | * remember to ack. Otherwise, drop segment | | 2447 | * remember to ack. Otherwise, drop segment |
2452 | * and (if not RST) ack. | | 2448 | * and (if not RST) ack. |
2453 | */ | | 2449 | */ |
2454 | if (tp->rcv_wnd == 0 && th->th_seq == tp->rcv_nxt) { | | 2450 | if (tp->rcv_wnd == 0 && th->th_seq == tp->rcv_nxt) { |
2455 | tp->t_flags |= TF_ACKNOW; | | 2451 | tp->t_flags |= TF_ACKNOW; |
2456 | TCP_STATINC(TCP_STAT_RCVWINPROBE); | | 2452 | TCP_STATINC(TCP_STAT_RCVWINPROBE); |
2457 | } else | | 2453 | } else |
2458 | goto dropafterack; | | 2454 | goto dropafterack; |
2459 | } else | | 2455 | } else |
2460 | TCP_STATADD(TCP_STAT_RCVBYTEAFTERWIN, todrop); | | 2456 | TCP_STATADD(TCP_STAT_RCVBYTEAFTERWIN, todrop); |
2461 | m_adj(m, -todrop); | | 2457 | m_adj(m, -todrop); |
2462 | tlen -= todrop; | | 2458 | tlen -= todrop; |
2463 | tiflags &= ~(TH_PUSH|TH_FIN); | | 2459 | tiflags &= ~(TH_PUSH|TH_FIN); |
2464 | } | | 2460 | } |
2465 | | | 2461 | |
2466 | /* | | 2462 | /* |
2467 | * If last ACK falls within this segment's sequence numbers, | | 2463 | * If last ACK falls within this segment's sequence numbers, |
2468 | * record the timestamp. | | 2464 | * record the timestamp. |
2469 | * NOTE: | | 2465 | * NOTE: |
2470 | * 1) That the test incorporates suggestions from the latest | | 2466 | * 1) That the test incorporates suggestions from the latest |
2471 | * proposal of the tcplw@cray.com list (Braden 1993/04/26). | | 2467 | * proposal of the tcplw@cray.com list (Braden 1993/04/26). |
2472 | * 2) That updating only on newer timestamps interferes with | | 2468 | * 2) That updating only on newer timestamps interferes with |
2473 | * our earlier PAWS tests, so this check should be solely | | 2469 | * our earlier PAWS tests, so this check should be solely |
2474 | * predicated on the sequence space of this segment. | | 2470 | * predicated on the sequence space of this segment. |
2475 | * 3) That we modify the segment boundary check to be | | 2471 | * 3) That we modify the segment boundary check to be |
2476 | * Last.ACK.Sent <= SEG.SEQ + SEG.Len | | 2472 | * Last.ACK.Sent <= SEG.SEQ + SEG.Len |
2477 | * instead of RFC1323's | | 2473 | * instead of RFC1323's |
2478 | * Last.ACK.Sent < SEG.SEQ + SEG.Len, | | 2474 | * Last.ACK.Sent < SEG.SEQ + SEG.Len, |
2479 | * This modified check allows us to overcome RFC1323's | | 2475 | * This modified check allows us to overcome RFC1323's |
2480 | * limitations as described in Stevens TCP/IP Illustrated | | 2476 | * limitations as described in Stevens TCP/IP Illustrated |
2481 | * Vol. 2 p.869. In such cases, we can still calculate the | | 2477 | * Vol. 2 p.869. In such cases, we can still calculate the |
2482 | * RTT correctly when RCV.NXT == Last.ACK.Sent. | | 2478 | * RTT correctly when RCV.NXT == Last.ACK.Sent. |
2483 | */ | | 2479 | */ |
2484 | if (opti.ts_present && | | 2480 | if (opti.ts_present && |
2485 | SEQ_LEQ(th->th_seq, tp->last_ack_sent) && | | 2481 | SEQ_LEQ(th->th_seq, tp->last_ack_sent) && |
2486 | SEQ_LEQ(tp->last_ack_sent, th->th_seq + tlen + | | 2482 | SEQ_LEQ(tp->last_ack_sent, th->th_seq + tlen + |
2487 | ((tiflags & (TH_SYN|TH_FIN)) != 0))) { | | 2483 | ((tiflags & (TH_SYN|TH_FIN)) != 0))) { |
2488 | tp->ts_recent_age = tcp_now; | | 2484 | tp->ts_recent_age = tcp_now; |
2489 | tp->ts_recent = opti.ts_val; | | 2485 | tp->ts_recent = opti.ts_val; |
2490 | } | | 2486 | } |
2491 | | | 2487 | |
2492 | /* | | 2488 | /* |
2493 | * If the RST bit is set examine the state: | | 2489 | * If the RST bit is set examine the state: |
2494 | * SYN_RECEIVED STATE: | | 2490 | * SYN_RECEIVED STATE: |
2495 | * If passive open, return to LISTEN state. | | 2491 | * If passive open, return to LISTEN state. |
2496 | * If active open, inform user that connection was refused. | | 2492 | * If active open, inform user that connection was refused. |
2497 | * ESTABLISHED, FIN_WAIT_1, FIN_WAIT2, CLOSE_WAIT STATES: | | 2493 | * ESTABLISHED, FIN_WAIT_1, FIN_WAIT2, CLOSE_WAIT STATES: |
2498 | * Inform user that connection was reset, and close tcb. | | 2494 | * Inform user that connection was reset, and close tcb. |
2499 | * CLOSING, LAST_ACK, TIME_WAIT STATES | | 2495 | * CLOSING, LAST_ACK, TIME_WAIT STATES |
2500 | * Close the tcb. | | 2496 | * Close the tcb. |
2501 | */ | | 2497 | */ |
2502 | if (tiflags & TH_RST) { | | 2498 | if (tiflags & TH_RST) { |
2503 | if (th->th_seq != tp->rcv_nxt) | | 2499 | if (th->th_seq != tp->rcv_nxt) |
2504 | goto dropafterack_ratelim; | | 2500 | goto dropafterack_ratelim; |
2505 | | | 2501 | |
2506 | switch (tp->t_state) { | | 2502 | switch (tp->t_state) { |
2507 | case TCPS_SYN_RECEIVED: | | 2503 | case TCPS_SYN_RECEIVED: |
2508 | so->so_error = ECONNREFUSED; | | 2504 | so->so_error = ECONNREFUSED; |
2509 | goto close; | | 2505 | goto close; |
2510 | | | 2506 | |
2511 | case TCPS_ESTABLISHED: | | 2507 | case TCPS_ESTABLISHED: |
2512 | case TCPS_FIN_WAIT_1: | | 2508 | case TCPS_FIN_WAIT_1: |
2513 | case TCPS_FIN_WAIT_2: | | 2509 | case TCPS_FIN_WAIT_2: |
2514 | case TCPS_CLOSE_WAIT: | | 2510 | case TCPS_CLOSE_WAIT: |
2515 | so->so_error = ECONNRESET; | | 2511 | so->so_error = ECONNRESET; |
2516 | close: | | 2512 | close: |
2517 | tp->t_state = TCPS_CLOSED; | | 2513 | tp->t_state = TCPS_CLOSED; |
2518 | TCP_STATINC(TCP_STAT_DROPS); | | 2514 | TCP_STATINC(TCP_STAT_DROPS); |
2519 | tp = tcp_close(tp); | | 2515 | tp = tcp_close(tp); |
2520 | goto drop; | | 2516 | goto drop; |
2521 | | | 2517 | |
2522 | case TCPS_CLOSING: | | 2518 | case TCPS_CLOSING: |
2523 | case TCPS_LAST_ACK: | | 2519 | case TCPS_LAST_ACK: |
2524 | case TCPS_TIME_WAIT: | | 2520 | case TCPS_TIME_WAIT: |
2525 | tp = tcp_close(tp); | | 2521 | tp = tcp_close(tp); |
2526 | goto drop; | | 2522 | goto drop; |
2527 | } | | 2523 | } |
2528 | } | | 2524 | } |
2529 | | | 2525 | |
2530 | /* | | 2526 | /* |
2531 | * Since we've covered the SYN-SENT and SYN-RECEIVED states above | | 2527 | * Since we've covered the SYN-SENT and SYN-RECEIVED states above |
2532 | * we must be in a synchronized state. RFC791 states (under RST | | 2528 | * we must be in a synchronized state. RFC791 states (under RST |
2533 | * generation) that any unacceptable segment (an out-of-order SYN | | 2529 | * generation) that any unacceptable segment (an out-of-order SYN |
2534 | * qualifies) received in a synchronized state must elicit only an | | 2530 | * qualifies) received in a synchronized state must elicit only an |
2535 | * empty acknowledgment segment ... and the connection remains in | | 2531 | * empty acknowledgment segment ... and the connection remains in |
2536 | * the same state. | | 2532 | * the same state. |
2537 | */ | | 2533 | */ |
2538 | if (tiflags & TH_SYN) { | | 2534 | if (tiflags & TH_SYN) { |
2539 | if (tp->rcv_nxt == th->th_seq) { | | 2535 | if (tp->rcv_nxt == th->th_seq) { |
2540 | tcp_respond(tp, m, m, th, (tcp_seq)0, th->th_ack - 1, | | 2536 | tcp_respond(tp, m, m, th, (tcp_seq)0, th->th_ack - 1, |
2541 | TH_ACK); | | 2537 | TH_ACK); |
2542 | if (tcp_saveti) | | 2538 | if (tcp_saveti) |
2543 | m_freem(tcp_saveti); | | 2539 | m_freem(tcp_saveti); |
2544 | return; | | 2540 | return; |
2545 | } | | 2541 | } |
2546 | | | 2542 | |
2547 | goto dropafterack_ratelim; | | 2543 | goto dropafterack_ratelim; |
2548 | } | | 2544 | } |
2549 | | | 2545 | |
2550 | /* | | 2546 | /* |
2551 | * If the ACK bit is off we drop the segment and return. | | 2547 | * If the ACK bit is off we drop the segment and return. |
2552 | */ | | 2548 | */ |
2553 | if ((tiflags & TH_ACK) == 0) { | | 2549 | if ((tiflags & TH_ACK) == 0) { |
2554 | if (tp->t_flags & TF_ACKNOW) | | 2550 | if (tp->t_flags & TF_ACKNOW) |
2555 | goto dropafterack; | | 2551 | goto dropafterack; |
2556 | else | | 2552 | else |
2557 | goto drop; | | 2553 | goto drop; |
2558 | } | | 2554 | } |
2559 | | | 2555 | |
2560 | /* | | 2556 | /* |
2561 | * Ack processing. | | 2557 | * Ack processing. |
2562 | */ | | 2558 | */ |
2563 | switch (tp->t_state) { | | 2559 | switch (tp->t_state) { |
2564 | | | 2560 | |
2565 | /* | | 2561 | /* |
2566 | * In SYN_RECEIVED state if the ack ACKs our SYN then enter | | 2562 | * In SYN_RECEIVED state if the ack ACKs our SYN then enter |
2567 | * ESTABLISHED state and continue processing, otherwise | | 2563 | * ESTABLISHED state and continue processing, otherwise |
2568 | * send an RST. | | 2564 | * send an RST. |
2569 | */ | | 2565 | */ |
2570 | case TCPS_SYN_RECEIVED: | | 2566 | case TCPS_SYN_RECEIVED: |
2571 | if (SEQ_GT(tp->snd_una, th->th_ack) || | | 2567 | if (SEQ_GT(tp->snd_una, th->th_ack) || |
2572 | SEQ_GT(th->th_ack, tp->snd_max)) | | 2568 | SEQ_GT(th->th_ack, tp->snd_max)) |
2573 | goto dropwithreset; | | 2569 | goto dropwithreset; |
2574 | TCP_STATINC(TCP_STAT_CONNECTS); | | 2570 | TCP_STATINC(TCP_STAT_CONNECTS); |
2575 | soisconnected(so); | | 2571 | soisconnected(so); |
2576 | tcp_established(tp); | | 2572 | tcp_established(tp); |
2577 | /* Do window scaling? */ | | 2573 | /* Do window scaling? */ |
2578 | if ((tp->t_flags & (TF_RCVD_SCALE|TF_REQ_SCALE)) == | | 2574 | if ((tp->t_flags & (TF_RCVD_SCALE|TF_REQ_SCALE)) == |
2579 | (TF_RCVD_SCALE|TF_REQ_SCALE)) { | | 2575 | (TF_RCVD_SCALE|TF_REQ_SCALE)) { |
2580 | tp->snd_scale = tp->requested_s_scale; | | 2576 | tp->snd_scale = tp->requested_s_scale; |
2581 | tp->rcv_scale = tp->request_r_scale; | | 2577 | tp->rcv_scale = tp->request_r_scale; |
2582 | } | | 2578 | } |
2583 | TCP_REASS_LOCK(tp); | | 2579 | TCP_REASS_LOCK(tp); |
2584 | (void) tcp_reass(tp, NULL, NULL, &tlen); | | 2580 | (void) tcp_reass(tp, NULL, NULL, &tlen); |
2585 | tp->snd_wl1 = th->th_seq - 1; | | 2581 | tp->snd_wl1 = th->th_seq - 1; |
2586 | /* fall into ... */ | | 2582 | /* fall into ... */ |
2587 | | | 2583 | |
2588 | /* | | 2584 | /* |
2589 | * In ESTABLISHED state: drop duplicate ACKs; ACK out of range | | 2585 | * In ESTABLISHED state: drop duplicate ACKs; ACK out of range |
2590 | * ACKs. If the ack is in the range | | 2586 | * ACKs. If the ack is in the range |
2591 | * tp->snd_una < th->th_ack <= tp->snd_max | | 2587 | * tp->snd_una < th->th_ack <= tp->snd_max |
2592 | * then advance tp->snd_una to th->th_ack and drop | | 2588 | * then advance tp->snd_una to th->th_ack and drop |
2593 | * data from the retransmission queue. If this ACK reflects | | 2589 | * data from the retransmission queue. If this ACK reflects |
2594 | * more up to date window information we update our window information. | | 2590 | * more up to date window information we update our window information. |
2595 | */ | | 2591 | */ |
2596 | case TCPS_ESTABLISHED: | | 2592 | case TCPS_ESTABLISHED: |
2597 | case TCPS_FIN_WAIT_1: | | 2593 | case TCPS_FIN_WAIT_1: |
2598 | case TCPS_FIN_WAIT_2: | | 2594 | case TCPS_FIN_WAIT_2: |
2599 | case TCPS_CLOSE_WAIT: | | 2595 | case TCPS_CLOSE_WAIT: |
2600 | case TCPS_CLOSING: | | 2596 | case TCPS_CLOSING: |
2601 | case TCPS_LAST_ACK: | | 2597 | case TCPS_LAST_ACK: |
2602 | case TCPS_TIME_WAIT: | | 2598 | case TCPS_TIME_WAIT: |
2603 | | | 2599 | |