| @@ -1,1471 +1,1476 @@ | | | @@ -1,1471 +1,1476 @@ |
1 | /* $NetBSD: uipc_socket.c,v 1.191 2009/10/02 23:50:16 elad Exp $ */ | | 1 | /* $NetBSD: uipc_socket.c,v 1.192 2009/10/03 01:41:39 elad Exp $ */ |
2 | | | 2 | |
3 | /*- | | 3 | /*- |
4 | * Copyright (c) 2002, 2007, 2008, 2009 The NetBSD Foundation, Inc. | | 4 | * Copyright (c) 2002, 2007, 2008, 2009 The NetBSD Foundation, Inc. |
5 | * All rights reserved. | | 5 | * All rights reserved. |
6 | * | | 6 | * |
7 | * This code is derived from software contributed to The NetBSD Foundation | | 7 | * This code is derived from software contributed to The NetBSD Foundation |
8 | * by Jason R. Thorpe of Wasabi Systems, Inc, and by Andrew Doran. | | 8 | * by Jason R. Thorpe of Wasabi Systems, Inc, and by Andrew Doran. |
9 | * | | 9 | * |
10 | * Redistribution and use in source and binary forms, with or without | | 10 | * Redistribution and use in source and binary forms, with or without |
11 | * modification, are permitted provided that the following conditions | | 11 | * modification, are permitted provided that the following conditions |
12 | * are met: | | 12 | * are met: |
13 | * 1. Redistributions of source code must retain the above copyright | | 13 | * 1. Redistributions of source code must retain the above copyright |
14 | * notice, this list of conditions and the following disclaimer. | | 14 | * notice, this list of conditions and the following disclaimer. |
15 | * 2. Redistributions in binary form must reproduce the above copyright | | 15 | * 2. Redistributions in binary form must reproduce the above copyright |
16 | * notice, this list of conditions and the following disclaimer in the | | 16 | * notice, this list of conditions and the following disclaimer in the |
17 | * documentation and/or other materials provided with the distribution. | | 17 | * documentation and/or other materials provided with the distribution. |
18 | * | | 18 | * |
19 | * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS | | 19 | * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS |
20 | * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED | | 20 | * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED |
21 | * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR | | 21 | * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR |
22 | * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS | | 22 | * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS |
23 | * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR | | 23 | * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR |
24 | * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF | | 24 | * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF |
25 | * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS | | 25 | * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS |
26 | * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN | | 26 | * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN |
27 | * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) | | 27 | * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) |
28 | * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE | | 28 | * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE |
29 | * POSSIBILITY OF SUCH DAMAGE. | | 29 | * POSSIBILITY OF SUCH DAMAGE. |
30 | */ | | 30 | */ |
31 | | | 31 | |
32 | /* | | 32 | /* |
33 | * Copyright (c) 2004 The FreeBSD Foundation | | 33 | * Copyright (c) 2004 The FreeBSD Foundation |
34 | * Copyright (c) 2004 Robert Watson | | 34 | * Copyright (c) 2004 Robert Watson |
35 | * Copyright (c) 1982, 1986, 1988, 1990, 1993 | | 35 | * Copyright (c) 1982, 1986, 1988, 1990, 1993 |
36 | * The Regents of the University of California. All rights reserved. | | 36 | * The Regents of the University of California. All rights reserved. |
37 | * | | 37 | * |
38 | * Redistribution and use in source and binary forms, with or without | | 38 | * Redistribution and use in source and binary forms, with or without |
39 | * modification, are permitted provided that the following conditions | | 39 | * modification, are permitted provided that the following conditions |
40 | * are met: | | 40 | * are met: |
41 | * 1. Redistributions of source code must retain the above copyright | | 41 | * 1. Redistributions of source code must retain the above copyright |
42 | * notice, this list of conditions and the following disclaimer. | | 42 | * notice, this list of conditions and the following disclaimer. |
43 | * 2. Redistributions in binary form must reproduce the above copyright | | 43 | * 2. Redistributions in binary form must reproduce the above copyright |
44 | * notice, this list of conditions and the following disclaimer in the | | 44 | * notice, this list of conditions and the following disclaimer in the |
45 | * documentation and/or other materials provided with the distribution. | | 45 | * documentation and/or other materials provided with the distribution. |
46 | * 3. Neither the name of the University nor the names of its contributors | | 46 | * 3. Neither the name of the University nor the names of its contributors |
47 | * may be used to endorse or promote products derived from this software | | 47 | * may be used to endorse or promote products derived from this software |
48 | * without specific prior written permission. | | 48 | * without specific prior written permission. |
49 | * | | 49 | * |
50 | * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND | | 50 | * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND |
51 | * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE | | 51 | * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE |
52 | * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE | | 52 | * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE |
53 | * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE | | 53 | * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE |
54 | * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL | | 54 | * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL |
55 | * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS | | 55 | * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS |
56 | * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) | | 56 | * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) |
57 | * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT | | 57 | * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT |
58 | * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY | | 58 | * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY |
59 | * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF | | 59 | * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF |
60 | * SUCH DAMAGE. | | 60 | * SUCH DAMAGE. |
61 | * | | 61 | * |
62 | * @(#)uipc_socket.c 8.6 (Berkeley) 5/2/95 | | 62 | * @(#)uipc_socket.c 8.6 (Berkeley) 5/2/95 |
63 | */ | | 63 | */ |
64 | | | 64 | |
65 | #include <sys/cdefs.h> | | 65 | #include <sys/cdefs.h> |
66 | __KERNEL_RCSID(0, "$NetBSD: uipc_socket.c,v 1.191 2009/10/02 23:50:16 elad Exp $"); | | 66 | __KERNEL_RCSID(0, "$NetBSD: uipc_socket.c,v 1.192 2009/10/03 01:41:39 elad Exp $"); |
67 | | | 67 | |
68 | #include "opt_compat_netbsd.h" | | 68 | #include "opt_compat_netbsd.h" |
69 | #include "opt_sock_counters.h" | | 69 | #include "opt_sock_counters.h" |
70 | #include "opt_sosend_loan.h" | | 70 | #include "opt_sosend_loan.h" |
71 | #include "opt_mbuftrace.h" | | 71 | #include "opt_mbuftrace.h" |
72 | #include "opt_somaxkva.h" | | 72 | #include "opt_somaxkva.h" |
73 | #include "opt_multiprocessor.h" /* XXX */ | | 73 | #include "opt_multiprocessor.h" /* XXX */ |
74 | | | 74 | |
75 | #include <sys/param.h> | | 75 | #include <sys/param.h> |
76 | #include <sys/systm.h> | | 76 | #include <sys/systm.h> |
77 | #include <sys/proc.h> | | 77 | #include <sys/proc.h> |
78 | #include <sys/file.h> | | 78 | #include <sys/file.h> |
79 | #include <sys/filedesc.h> | | 79 | #include <sys/filedesc.h> |
80 | #include <sys/kmem.h> | | 80 | #include <sys/kmem.h> |
81 | #include <sys/mbuf.h> | | 81 | #include <sys/mbuf.h> |
82 | #include <sys/domain.h> | | 82 | #include <sys/domain.h> |
83 | #include <sys/kernel.h> | | 83 | #include <sys/kernel.h> |
84 | #include <sys/protosw.h> | | 84 | #include <sys/protosw.h> |
85 | #include <sys/socket.h> | | 85 | #include <sys/socket.h> |
86 | #include <sys/socketvar.h> | | 86 | #include <sys/socketvar.h> |
87 | #include <sys/signalvar.h> | | 87 | #include <sys/signalvar.h> |
88 | #include <sys/resourcevar.h> | | 88 | #include <sys/resourcevar.h> |
89 | #include <sys/uidinfo.h> | | 89 | #include <sys/uidinfo.h> |
90 | #include <sys/event.h> | | 90 | #include <sys/event.h> |
91 | #include <sys/poll.h> | | 91 | #include <sys/poll.h> |
92 | #include <sys/kauth.h> | | 92 | #include <sys/kauth.h> |
93 | #include <sys/mutex.h> | | 93 | #include <sys/mutex.h> |
94 | #include <sys/condvar.h> | | 94 | #include <sys/condvar.h> |
95 | | | 95 | |
96 | #ifdef COMPAT_50 | | 96 | #ifdef COMPAT_50 |
97 | #include <compat/sys/time.h> | | 97 | #include <compat/sys/time.h> |
98 | #include <compat/sys/socket.h> | | 98 | #include <compat/sys/socket.h> |
99 | #endif | | 99 | #endif |
100 | | | 100 | |
101 | #include <uvm/uvm.h> | | 101 | #include <uvm/uvm.h> |
102 | | | 102 | |
103 | MALLOC_DEFINE(M_SOOPTS, "soopts", "socket options"); | | 103 | MALLOC_DEFINE(M_SOOPTS, "soopts", "socket options"); |
104 | MALLOC_DEFINE(M_SONAME, "soname", "socket name"); | | 104 | MALLOC_DEFINE(M_SONAME, "soname", "socket name"); |
105 | | | 105 | |
106 | extern const struct fileops socketops; | | 106 | extern const struct fileops socketops; |
107 | | | 107 | |
108 | extern int somaxconn; /* patchable (XXX sysctl) */ | | 108 | extern int somaxconn; /* patchable (XXX sysctl) */ |
109 | int somaxconn = SOMAXCONN; | | 109 | int somaxconn = SOMAXCONN; |
110 | kmutex_t *softnet_lock; | | 110 | kmutex_t *softnet_lock; |
111 | | | 111 | |
112 | #ifdef SOSEND_COUNTERS | | 112 | #ifdef SOSEND_COUNTERS |
113 | #include <sys/device.h> | | 113 | #include <sys/device.h> |
114 | | | 114 | |
115 | static struct evcnt sosend_loan_big = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, | | 115 | static struct evcnt sosend_loan_big = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, |
116 | NULL, "sosend", "loan big"); | | 116 | NULL, "sosend", "loan big"); |
117 | static struct evcnt sosend_copy_big = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, | | 117 | static struct evcnt sosend_copy_big = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, |
118 | NULL, "sosend", "copy big"); | | 118 | NULL, "sosend", "copy big"); |
119 | static struct evcnt sosend_copy_small = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, | | 119 | static struct evcnt sosend_copy_small = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, |
120 | NULL, "sosend", "copy small"); | | 120 | NULL, "sosend", "copy small"); |
121 | static struct evcnt sosend_kvalimit = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, | | 121 | static struct evcnt sosend_kvalimit = EVCNT_INITIALIZER(EVCNT_TYPE_MISC, |
122 | NULL, "sosend", "kva limit"); | | 122 | NULL, "sosend", "kva limit"); |
123 | | | 123 | |
124 | #define SOSEND_COUNTER_INCR(ev) (ev)->ev_count++ | | 124 | #define SOSEND_COUNTER_INCR(ev) (ev)->ev_count++ |
125 | | | 125 | |
126 | EVCNT_ATTACH_STATIC(sosend_loan_big); | | 126 | EVCNT_ATTACH_STATIC(sosend_loan_big); |
127 | EVCNT_ATTACH_STATIC(sosend_copy_big); | | 127 | EVCNT_ATTACH_STATIC(sosend_copy_big); |
128 | EVCNT_ATTACH_STATIC(sosend_copy_small); | | 128 | EVCNT_ATTACH_STATIC(sosend_copy_small); |
129 | EVCNT_ATTACH_STATIC(sosend_kvalimit); | | 129 | EVCNT_ATTACH_STATIC(sosend_kvalimit); |
130 | #else | | 130 | #else |
131 | | | 131 | |
132 | #define SOSEND_COUNTER_INCR(ev) /* nothing */ | | 132 | #define SOSEND_COUNTER_INCR(ev) /* nothing */ |
133 | | | 133 | |
134 | #endif /* SOSEND_COUNTERS */ | | 134 | #endif /* SOSEND_COUNTERS */ |
135 | | | 135 | |
136 | static struct callback_entry sokva_reclaimerentry; | | 136 | static struct callback_entry sokva_reclaimerentry; |
137 | | | 137 | |
138 | #if defined(SOSEND_NO_LOAN) || defined(MULTIPROCESSOR) | | 138 | #if defined(SOSEND_NO_LOAN) || defined(MULTIPROCESSOR) |
139 | int sock_loan_thresh = -1; | | 139 | int sock_loan_thresh = -1; |
140 | #else | | 140 | #else |
141 | int sock_loan_thresh = 4096; | | 141 | int sock_loan_thresh = 4096; |
142 | #endif | | 142 | #endif |
143 | | | 143 | |
144 | static kmutex_t so_pendfree_lock; | | 144 | static kmutex_t so_pendfree_lock; |
145 | static struct mbuf *so_pendfree; | | 145 | static struct mbuf *so_pendfree; |
146 | | | 146 | |
147 | #ifndef SOMAXKVA | | 147 | #ifndef SOMAXKVA |
148 | #define SOMAXKVA (16 * 1024 * 1024) | | 148 | #define SOMAXKVA (16 * 1024 * 1024) |
149 | #endif | | 149 | #endif |
150 | int somaxkva = SOMAXKVA; | | 150 | int somaxkva = SOMAXKVA; |
151 | static int socurkva; | | 151 | static int socurkva; |
152 | static kcondvar_t socurkva_cv; | | 152 | static kcondvar_t socurkva_cv; |
153 | | | 153 | |
154 | static kauth_listener_t socket_listener; | | 154 | static kauth_listener_t socket_listener; |
155 | | | 155 | |
156 | #define SOCK_LOAN_CHUNK 65536 | | 156 | #define SOCK_LOAN_CHUNK 65536 |
157 | | | 157 | |
158 | static size_t sodopendfree(void); | | 158 | static size_t sodopendfree(void); |
159 | static size_t sodopendfreel(void); | | 159 | static size_t sodopendfreel(void); |
160 | | | 160 | |
161 | static void sysctl_kern_somaxkva_setup(void); | | 161 | static void sysctl_kern_somaxkva_setup(void); |
162 | static struct sysctllog *socket_sysctllog; | | 162 | static struct sysctllog *socket_sysctllog; |
163 | | | 163 | |
164 | static vsize_t | | 164 | static vsize_t |
165 | sokvareserve(struct socket *so, vsize_t len) | | 165 | sokvareserve(struct socket *so, vsize_t len) |
166 | { | | 166 | { |
167 | int error; | | 167 | int error; |
168 | | | 168 | |
169 | mutex_enter(&so_pendfree_lock); | | 169 | mutex_enter(&so_pendfree_lock); |
170 | while (socurkva + len > somaxkva) { | | 170 | while (socurkva + len > somaxkva) { |
171 | size_t freed; | | 171 | size_t freed; |
172 | | | 172 | |
173 | /* | | 173 | /* |
174 | * try to do pendfree. | | 174 | * try to do pendfree. |
175 | */ | | 175 | */ |
176 | | | 176 | |
177 | freed = sodopendfreel(); | | 177 | freed = sodopendfreel(); |
178 | | | 178 | |
179 | /* | | 179 | /* |
180 | * if some kva was freed, try again. | | 180 | * if some kva was freed, try again. |
181 | */ | | 181 | */ |
182 | | | 182 | |
183 | if (freed) | | 183 | if (freed) |
184 | continue; | | 184 | continue; |
185 | | | 185 | |
186 | SOSEND_COUNTER_INCR(&sosend_kvalimit); | | 186 | SOSEND_COUNTER_INCR(&sosend_kvalimit); |
187 | error = cv_wait_sig(&socurkva_cv, &so_pendfree_lock); | | 187 | error = cv_wait_sig(&socurkva_cv, &so_pendfree_lock); |
188 | if (error) { | | 188 | if (error) { |
189 | len = 0; | | 189 | len = 0; |
190 | break; | | 190 | break; |
191 | } | | 191 | } |
192 | } | | 192 | } |
193 | socurkva += len; | | 193 | socurkva += len; |
194 | mutex_exit(&so_pendfree_lock); | | 194 | mutex_exit(&so_pendfree_lock); |
195 | return len; | | 195 | return len; |
196 | } | | 196 | } |
197 | | | 197 | |
198 | static void | | 198 | static void |
199 | sokvaunreserve(vsize_t len) | | 199 | sokvaunreserve(vsize_t len) |
200 | { | | 200 | { |
201 | | | 201 | |
202 | mutex_enter(&so_pendfree_lock); | | 202 | mutex_enter(&so_pendfree_lock); |
203 | socurkva -= len; | | 203 | socurkva -= len; |
204 | cv_broadcast(&socurkva_cv); | | 204 | cv_broadcast(&socurkva_cv); |
205 | mutex_exit(&so_pendfree_lock); | | 205 | mutex_exit(&so_pendfree_lock); |
206 | } | | 206 | } |
207 | | | 207 | |
208 | /* | | 208 | /* |
209 | * sokvaalloc: allocate kva for loan. | | 209 | * sokvaalloc: allocate kva for loan. |
210 | */ | | 210 | */ |
211 | | | 211 | |
212 | vaddr_t | | 212 | vaddr_t |
213 | sokvaalloc(vsize_t len, struct socket *so) | | 213 | sokvaalloc(vsize_t len, struct socket *so) |
214 | { | | 214 | { |
215 | vaddr_t lva; | | 215 | vaddr_t lva; |
216 | | | 216 | |
217 | /* | | 217 | /* |
218 | * reserve kva. | | 218 | * reserve kva. |
219 | */ | | 219 | */ |
220 | | | 220 | |
221 | if (sokvareserve(so, len) == 0) | | 221 | if (sokvareserve(so, len) == 0) |
222 | return 0; | | 222 | return 0; |
223 | | | 223 | |
224 | /* | | 224 | /* |
225 | * allocate kva. | | 225 | * allocate kva. |
226 | */ | | 226 | */ |
227 | | | 227 | |
228 | lva = uvm_km_alloc(kernel_map, len, 0, UVM_KMF_VAONLY | UVM_KMF_WAITVA); | | 228 | lva = uvm_km_alloc(kernel_map, len, 0, UVM_KMF_VAONLY | UVM_KMF_WAITVA); |
229 | if (lva == 0) { | | 229 | if (lva == 0) { |
230 | sokvaunreserve(len); | | 230 | sokvaunreserve(len); |
231 | return (0); | | 231 | return (0); |
232 | } | | 232 | } |
233 | | | 233 | |
234 | return lva; | | 234 | return lva; |
235 | } | | 235 | } |
236 | | | 236 | |
237 | /* | | 237 | /* |
238 | * sokvafree: free kva for loan. | | 238 | * sokvafree: free kva for loan. |
239 | */ | | 239 | */ |
240 | | | 240 | |
241 | void | | 241 | void |
242 | sokvafree(vaddr_t sva, vsize_t len) | | 242 | sokvafree(vaddr_t sva, vsize_t len) |
243 | { | | 243 | { |
244 | | | 244 | |
245 | /* | | 245 | /* |
246 | * free kva. | | 246 | * free kva. |
247 | */ | | 247 | */ |
248 | | | 248 | |
249 | uvm_km_free(kernel_map, sva, len, UVM_KMF_VAONLY); | | 249 | uvm_km_free(kernel_map, sva, len, UVM_KMF_VAONLY); |
250 | | | 250 | |
251 | /* | | 251 | /* |
252 | * unreserve kva. | | 252 | * unreserve kva. |
253 | */ | | 253 | */ |
254 | | | 254 | |
255 | sokvaunreserve(len); | | 255 | sokvaunreserve(len); |
256 | } | | 256 | } |
257 | | | 257 | |
258 | static void | | 258 | static void |
259 | sodoloanfree(struct vm_page **pgs, void *buf, size_t size) | | 259 | sodoloanfree(struct vm_page **pgs, void *buf, size_t size) |
260 | { | | 260 | { |
261 | vaddr_t sva, eva; | | 261 | vaddr_t sva, eva; |
262 | vsize_t len; | | 262 | vsize_t len; |
263 | int npgs; | | 263 | int npgs; |
264 | | | 264 | |
265 | KASSERT(pgs != NULL); | | 265 | KASSERT(pgs != NULL); |
266 | | | 266 | |
267 | eva = round_page((vaddr_t) buf + size); | | 267 | eva = round_page((vaddr_t) buf + size); |
268 | sva = trunc_page((vaddr_t) buf); | | 268 | sva = trunc_page((vaddr_t) buf); |
269 | len = eva - sva; | | 269 | len = eva - sva; |
270 | npgs = len >> PAGE_SHIFT; | | 270 | npgs = len >> PAGE_SHIFT; |
271 | | | 271 | |
272 | pmap_kremove(sva, len); | | 272 | pmap_kremove(sva, len); |
273 | pmap_update(pmap_kernel()); | | 273 | pmap_update(pmap_kernel()); |
274 | uvm_unloan(pgs, npgs, UVM_LOAN_TOPAGE); | | 274 | uvm_unloan(pgs, npgs, UVM_LOAN_TOPAGE); |
275 | sokvafree(sva, len); | | 275 | sokvafree(sva, len); |
276 | } | | 276 | } |
277 | | | 277 | |
278 | static size_t | | 278 | static size_t |
279 | sodopendfree(void) | | 279 | sodopendfree(void) |
280 | { | | 280 | { |
281 | size_t rv; | | 281 | size_t rv; |
282 | | | 282 | |
283 | if (__predict_true(so_pendfree == NULL)) | | 283 | if (__predict_true(so_pendfree == NULL)) |
284 | return 0; | | 284 | return 0; |
285 | | | 285 | |
286 | mutex_enter(&so_pendfree_lock); | | 286 | mutex_enter(&so_pendfree_lock); |
287 | rv = sodopendfreel(); | | 287 | rv = sodopendfreel(); |
288 | mutex_exit(&so_pendfree_lock); | | 288 | mutex_exit(&so_pendfree_lock); |
289 | | | 289 | |
290 | return rv; | | 290 | return rv; |
291 | } | | 291 | } |
292 | | | 292 | |
293 | /* | | 293 | /* |
294 | * sodopendfreel: free mbufs on "pendfree" list. | | 294 | * sodopendfreel: free mbufs on "pendfree" list. |
295 | * unlock and relock so_pendfree_lock when freeing mbufs. | | 295 | * unlock and relock so_pendfree_lock when freeing mbufs. |
296 | * | | 296 | * |
297 | * => called with so_pendfree_lock held. | | 297 | * => called with so_pendfree_lock held. |
298 | */ | | 298 | */ |
299 | | | 299 | |
300 | static size_t | | 300 | static size_t |
301 | sodopendfreel(void) | | 301 | sodopendfreel(void) |
302 | { | | 302 | { |
303 | struct mbuf *m, *next; | | 303 | struct mbuf *m, *next; |
304 | size_t rv = 0; | | 304 | size_t rv = 0; |
305 | | | 305 | |
306 | KASSERT(mutex_owned(&so_pendfree_lock)); | | 306 | KASSERT(mutex_owned(&so_pendfree_lock)); |
307 | | | 307 | |
308 | while (so_pendfree != NULL) { | | 308 | while (so_pendfree != NULL) { |
309 | m = so_pendfree; | | 309 | m = so_pendfree; |
310 | so_pendfree = NULL; | | 310 | so_pendfree = NULL; |
311 | mutex_exit(&so_pendfree_lock); | | 311 | mutex_exit(&so_pendfree_lock); |
312 | | | 312 | |
313 | for (; m != NULL; m = next) { | | 313 | for (; m != NULL; m = next) { |
314 | next = m->m_next; | | 314 | next = m->m_next; |
315 | KASSERT((~m->m_flags & (M_EXT|M_EXT_PAGES)) == 0); | | 315 | KASSERT((~m->m_flags & (M_EXT|M_EXT_PAGES)) == 0); |
316 | KASSERT(m->m_ext.ext_refcnt == 0); | | 316 | KASSERT(m->m_ext.ext_refcnt == 0); |
317 | | | 317 | |
318 | rv += m->m_ext.ext_size; | | 318 | rv += m->m_ext.ext_size; |
319 | sodoloanfree(m->m_ext.ext_pgs, m->m_ext.ext_buf, | | 319 | sodoloanfree(m->m_ext.ext_pgs, m->m_ext.ext_buf, |
320 | m->m_ext.ext_size); | | 320 | m->m_ext.ext_size); |
321 | pool_cache_put(mb_cache, m); | | 321 | pool_cache_put(mb_cache, m); |
322 | } | | 322 | } |
323 | | | 323 | |
324 | mutex_enter(&so_pendfree_lock); | | 324 | mutex_enter(&so_pendfree_lock); |
325 | } | | 325 | } |
326 | | | 326 | |
327 | return (rv); | | 327 | return (rv); |
328 | } | | 328 | } |
329 | | | 329 | |
330 | void | | 330 | void |
331 | soloanfree(struct mbuf *m, void *buf, size_t size, void *arg) | | 331 | soloanfree(struct mbuf *m, void *buf, size_t size, void *arg) |
332 | { | | 332 | { |
333 | | | 333 | |
334 | KASSERT(m != NULL); | | 334 | KASSERT(m != NULL); |
335 | | | 335 | |
336 | /* | | 336 | /* |
337 | * postpone freeing mbuf. | | 337 | * postpone freeing mbuf. |
338 | * | | 338 | * |
339 | * we can't do it in interrupt context | | 339 | * we can't do it in interrupt context |
340 | * because we need to put kva back to kernel_map. | | 340 | * because we need to put kva back to kernel_map. |
341 | */ | | 341 | */ |
342 | | | 342 | |
343 | mutex_enter(&so_pendfree_lock); | | 343 | mutex_enter(&so_pendfree_lock); |
344 | m->m_next = so_pendfree; | | 344 | m->m_next = so_pendfree; |
345 | so_pendfree = m; | | 345 | so_pendfree = m; |
346 | cv_broadcast(&socurkva_cv); | | 346 | cv_broadcast(&socurkva_cv); |
347 | mutex_exit(&so_pendfree_lock); | | 347 | mutex_exit(&so_pendfree_lock); |
348 | } | | 348 | } |
349 | | | 349 | |
350 | static long | | 350 | static long |
351 | sosend_loan(struct socket *so, struct uio *uio, struct mbuf *m, long space) | | 351 | sosend_loan(struct socket *so, struct uio *uio, struct mbuf *m, long space) |
352 | { | | 352 | { |
353 | struct iovec *iov = uio->uio_iov; | | 353 | struct iovec *iov = uio->uio_iov; |
354 | vaddr_t sva, eva; | | 354 | vaddr_t sva, eva; |
355 | vsize_t len; | | 355 | vsize_t len; |
356 | vaddr_t lva; | | 356 | vaddr_t lva; |
357 | int npgs, error; | | 357 | int npgs, error; |
358 | vaddr_t va; | | 358 | vaddr_t va; |
359 | int i; | | 359 | int i; |
360 | | | 360 | |
361 | if (VMSPACE_IS_KERNEL_P(uio->uio_vmspace)) | | 361 | if (VMSPACE_IS_KERNEL_P(uio->uio_vmspace)) |
362 | return (0); | | 362 | return (0); |
363 | | | 363 | |
364 | if (iov->iov_len < (size_t) space) | | 364 | if (iov->iov_len < (size_t) space) |
365 | space = iov->iov_len; | | 365 | space = iov->iov_len; |
366 | if (space > SOCK_LOAN_CHUNK) | | 366 | if (space > SOCK_LOAN_CHUNK) |
367 | space = SOCK_LOAN_CHUNK; | | 367 | space = SOCK_LOAN_CHUNK; |
368 | | | 368 | |
369 | eva = round_page((vaddr_t) iov->iov_base + space); | | 369 | eva = round_page((vaddr_t) iov->iov_base + space); |
370 | sva = trunc_page((vaddr_t) iov->iov_base); | | 370 | sva = trunc_page((vaddr_t) iov->iov_base); |
371 | len = eva - sva; | | 371 | len = eva - sva; |
372 | npgs = len >> PAGE_SHIFT; | | 372 | npgs = len >> PAGE_SHIFT; |
373 | | | 373 | |
374 | KASSERT(npgs <= M_EXT_MAXPAGES); | | 374 | KASSERT(npgs <= M_EXT_MAXPAGES); |
375 | | | 375 | |
376 | lva = sokvaalloc(len, so); | | 376 | lva = sokvaalloc(len, so); |
377 | if (lva == 0) | | 377 | if (lva == 0) |
378 | return 0; | | 378 | return 0; |
379 | | | 379 | |
380 | error = uvm_loan(&uio->uio_vmspace->vm_map, sva, len, | | 380 | error = uvm_loan(&uio->uio_vmspace->vm_map, sva, len, |
381 | m->m_ext.ext_pgs, UVM_LOAN_TOPAGE); | | 381 | m->m_ext.ext_pgs, UVM_LOAN_TOPAGE); |
382 | if (error) { | | 382 | if (error) { |
383 | sokvafree(lva, len); | | 383 | sokvafree(lva, len); |
384 | return (0); | | 384 | return (0); |
385 | } | | 385 | } |
386 | | | 386 | |
387 | for (i = 0, va = lva; i < npgs; i++, va += PAGE_SIZE) | | 387 | for (i = 0, va = lva; i < npgs; i++, va += PAGE_SIZE) |
388 | pmap_kenter_pa(va, VM_PAGE_TO_PHYS(m->m_ext.ext_pgs[i]), | | 388 | pmap_kenter_pa(va, VM_PAGE_TO_PHYS(m->m_ext.ext_pgs[i]), |
389 | VM_PROT_READ); | | 389 | VM_PROT_READ); |
390 | pmap_update(pmap_kernel()); | | 390 | pmap_update(pmap_kernel()); |
391 | | | 391 | |
392 | lva += (vaddr_t) iov->iov_base & PAGE_MASK; | | 392 | lva += (vaddr_t) iov->iov_base & PAGE_MASK; |
393 | | | 393 | |
394 | MEXTADD(m, (void *) lva, space, M_MBUF, soloanfree, so); | | 394 | MEXTADD(m, (void *) lva, space, M_MBUF, soloanfree, so); |
395 | m->m_flags |= M_EXT_PAGES | M_EXT_ROMAP; | | 395 | m->m_flags |= M_EXT_PAGES | M_EXT_ROMAP; |
396 | | | 396 | |
397 | uio->uio_resid -= space; | | 397 | uio->uio_resid -= space; |
398 | /* uio_offset not updated, not set/used for write(2) */ | | 398 | /* uio_offset not updated, not set/used for write(2) */ |
399 | uio->uio_iov->iov_base = (char *)uio->uio_iov->iov_base + space; | | 399 | uio->uio_iov->iov_base = (char *)uio->uio_iov->iov_base + space; |
400 | uio->uio_iov->iov_len -= space; | | 400 | uio->uio_iov->iov_len -= space; |
401 | if (uio->uio_iov->iov_len == 0) { | | 401 | if (uio->uio_iov->iov_len == 0) { |
402 | uio->uio_iov++; | | 402 | uio->uio_iov++; |
403 | uio->uio_iovcnt--; | | 403 | uio->uio_iovcnt--; |
404 | } | | 404 | } |
405 | | | 405 | |
406 | return (space); | | 406 | return (space); |
407 | } | | 407 | } |
408 | | | 408 | |
409 | static int | | 409 | static int |
410 | sokva_reclaim_callback(struct callback_entry *ce, void *obj, void *arg) | | 410 | sokva_reclaim_callback(struct callback_entry *ce, void *obj, void *arg) |
411 | { | | 411 | { |
412 | | | 412 | |
413 | KASSERT(ce == &sokva_reclaimerentry); | | 413 | KASSERT(ce == &sokva_reclaimerentry); |
414 | KASSERT(obj == NULL); | | 414 | KASSERT(obj == NULL); |
415 | | | 415 | |
416 | sodopendfree(); | | 416 | sodopendfree(); |
417 | if (!vm_map_starved_p(kernel_map)) { | | 417 | if (!vm_map_starved_p(kernel_map)) { |
418 | return CALLBACK_CHAIN_ABORT; | | 418 | return CALLBACK_CHAIN_ABORT; |
419 | } | | 419 | } |
420 | return CALLBACK_CHAIN_CONTINUE; | | 420 | return CALLBACK_CHAIN_CONTINUE; |
421 | } | | 421 | } |
422 | | | 422 | |
423 | struct mbuf * | | 423 | struct mbuf * |
424 | getsombuf(struct socket *so, int type) | | 424 | getsombuf(struct socket *so, int type) |
425 | { | | 425 | { |
426 | struct mbuf *m; | | 426 | struct mbuf *m; |
427 | | | 427 | |
428 | m = m_get(M_WAIT, type); | | 428 | m = m_get(M_WAIT, type); |
429 | MCLAIM(m, so->so_mowner); | | 429 | MCLAIM(m, so->so_mowner); |
430 | return m; | | 430 | return m; |
431 | } | | 431 | } |
432 | | | 432 | |
433 | static int | | 433 | static int |
434 | socket_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie, | | 434 | socket_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie, |
435 | void *arg0, void *arg1, void *arg2, void *arg3) | | 435 | void *arg0, void *arg1, void *arg2, void *arg3) |
436 | { | | 436 | { |
437 | int result; | | 437 | int result; |
438 | enum kauth_network_req req; | | 438 | enum kauth_network_req req; |
439 | | | 439 | |
440 | result = KAUTH_RESULT_DEFER; | | 440 | result = KAUTH_RESULT_DEFER; |
441 | req = (enum kauth_network_req)arg0; | | 441 | req = (enum kauth_network_req)arg0; |
442 | | | 442 | |
443 | if (action != KAUTH_NETWORK_SOCKET) | | 443 | if (action != KAUTH_NETWORK_SOCKET) |
444 | return result; | | 444 | return result; |
445 | | | 445 | |
446 | switch (req) { | | 446 | switch (req) { |
447 | case KAUTH_REQ_NETWORK_SOCKET_DROP: { | | 447 | case KAUTH_REQ_NETWORK_SOCKET_DROP: { |
448 | /* Normal users can only drop their own connections. */ | | 448 | /* Normal users can only drop their own connections. */ |
449 | struct socket *so = (struct socket *)arg1; | | 449 | struct socket *so = (struct socket *)arg1; |
450 | uid_t sockuid = so->so_uidinfo->ui_uid; | | 450 | uid_t sockuid = so->so_uidinfo->ui_uid; |
451 | | | 451 | |
452 | if (sockuid == kauth_cred_getuid(cred) || | | 452 | if (sockuid == kauth_cred_getuid(cred) || |
453 | sockuid == kauth_cred_geteuid(cred)) | | 453 | sockuid == kauth_cred_geteuid(cred)) |
454 | result = KAUTH_RESULT_ALLOW; | | 454 | result = KAUTH_RESULT_ALLOW; |
455 | | | 455 | |
456 | break; | | 456 | break; |
457 | } | | 457 | } |
458 | | | 458 | |
459 | case KAUTH_REQ_NETWORK_SOCKET_OPEN: | | 459 | case KAUTH_REQ_NETWORK_SOCKET_OPEN: |
460 | /* We allow "raw" routing/bluetooth sockets to anyone. */ | | 460 | /* We allow "raw" routing/bluetooth sockets to anyone. */ |
461 | if ((u_long)arg1 == PF_ROUTE || (u_long)arg1 == PF_BLUETOOTH) | | 461 | if ((u_long)arg1 == PF_ROUTE || (u_long)arg1 == PF_BLUETOOTH) |
462 | result = KAUTH_RESULT_ALLOW; | | 462 | result = KAUTH_RESULT_ALLOW; |
463 | else { | | 463 | else { |
464 | /* Privileged, let secmodel handle this. */ | | 464 | /* Privileged, let secmodel handle this. */ |
465 | if ((u_long)arg2 == SOCK_RAW) | | 465 | if ((u_long)arg2 == SOCK_RAW) |
466 | break; | | 466 | break; |
467 | } | | 467 | } |
468 | | | 468 | |
469 | result = KAUTH_RESULT_ALLOW; | | 469 | result = KAUTH_RESULT_ALLOW; |
470 | | | 470 | |
471 | break; | | 471 | break; |
472 | | | 472 | |
| | | 473 | case KAUTH_REQ_NETWORK_SOCKET_CANSEE: |
| | | 474 | result = KAUTH_RESULT_ALLOW; |
| | | 475 | |
| | | 476 | break; |
| | | 477 | |
473 | default: | | 478 | default: |
474 | break; | | 479 | break; |
475 | } | | 480 | } |
476 | | | 481 | |
477 | return result; | | 482 | return result; |
478 | } | | 483 | } |
479 | | | 484 | |
480 | void | | 485 | void |
481 | soinit(void) | | 486 | soinit(void) |
482 | { | | 487 | { |
483 | | | 488 | |
484 | sysctl_kern_somaxkva_setup(); | | 489 | sysctl_kern_somaxkva_setup(); |
485 | | | 490 | |
486 | mutex_init(&so_pendfree_lock, MUTEX_DEFAULT, IPL_VM); | | 491 | mutex_init(&so_pendfree_lock, MUTEX_DEFAULT, IPL_VM); |
487 | softnet_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE); | | 492 | softnet_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE); |
488 | cv_init(&socurkva_cv, "sokva"); | | 493 | cv_init(&socurkva_cv, "sokva"); |
489 | soinit2(); | | 494 | soinit2(); |
490 | | | 495 | |
491 | /* Set the initial adjusted socket buffer size. */ | | 496 | /* Set the initial adjusted socket buffer size. */ |
492 | if (sb_max_set(sb_max)) | | 497 | if (sb_max_set(sb_max)) |
493 | panic("bad initial sb_max value: %lu", sb_max); | | 498 | panic("bad initial sb_max value: %lu", sb_max); |
494 | | | 499 | |
495 | callback_register(&vm_map_to_kernel(kernel_map)->vmk_reclaim_callback, | | 500 | callback_register(&vm_map_to_kernel(kernel_map)->vmk_reclaim_callback, |
496 | &sokva_reclaimerentry, NULL, sokva_reclaim_callback); | | 501 | &sokva_reclaimerentry, NULL, sokva_reclaim_callback); |
497 | | | 502 | |
498 | socket_listener = kauth_listen_scope(KAUTH_SCOPE_NETWORK, | | 503 | socket_listener = kauth_listen_scope(KAUTH_SCOPE_NETWORK, |
499 | socket_listener_cb, NULL); | | 504 | socket_listener_cb, NULL); |
500 | } | | 505 | } |
501 | | | 506 | |
502 | /* | | 507 | /* |
503 | * Socket operation routines. | | 508 | * Socket operation routines. |
504 | * These routines are called by the routines in | | 509 | * These routines are called by the routines in |
505 | * sys_socket.c or from a system process, and | | 510 | * sys_socket.c or from a system process, and |
506 | * implement the semantics of socket operations by | | 511 | * implement the semantics of socket operations by |
507 | * switching out to the protocol specific routines. | | 512 | * switching out to the protocol specific routines. |
508 | */ | | 513 | */ |
509 | /*ARGSUSED*/ | | 514 | /*ARGSUSED*/ |
510 | int | | 515 | int |
511 | socreate(int dom, struct socket **aso, int type, int proto, struct lwp *l, | | 516 | socreate(int dom, struct socket **aso, int type, int proto, struct lwp *l, |
512 | struct socket *lockso) | | 517 | struct socket *lockso) |
513 | { | | 518 | { |
514 | const struct protosw *prp; | | 519 | const struct protosw *prp; |
515 | struct socket *so; | | 520 | struct socket *so; |
516 | uid_t uid; | | 521 | uid_t uid; |
517 | int error; | | 522 | int error; |
518 | kmutex_t *lock; | | 523 | kmutex_t *lock; |
519 | | | 524 | |
520 | error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_SOCKET, | | 525 | error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_SOCKET, |
521 | KAUTH_REQ_NETWORK_SOCKET_OPEN, KAUTH_ARG(dom), KAUTH_ARG(type), | | 526 | KAUTH_REQ_NETWORK_SOCKET_OPEN, KAUTH_ARG(dom), KAUTH_ARG(type), |
522 | KAUTH_ARG(proto)); | | 527 | KAUTH_ARG(proto)); |
523 | if (error != 0) | | 528 | if (error != 0) |
524 | return error; | | 529 | return error; |
525 | | | 530 | |
526 | if (proto) | | 531 | if (proto) |
527 | prp = pffindproto(dom, proto, type); | | 532 | prp = pffindproto(dom, proto, type); |
528 | else | | 533 | else |
529 | prp = pffindtype(dom, type); | | 534 | prp = pffindtype(dom, type); |
530 | if (prp == NULL) { | | 535 | if (prp == NULL) { |
531 | /* no support for domain */ | | 536 | /* no support for domain */ |
532 | if (pffinddomain(dom) == 0) | | 537 | if (pffinddomain(dom) == 0) |
533 | return EAFNOSUPPORT; | | 538 | return EAFNOSUPPORT; |
534 | /* no support for socket type */ | | 539 | /* no support for socket type */ |
535 | if (proto == 0 && type != 0) | | 540 | if (proto == 0 && type != 0) |
536 | return EPROTOTYPE; | | 541 | return EPROTOTYPE; |
537 | return EPROTONOSUPPORT; | | 542 | return EPROTONOSUPPORT; |
538 | } | | 543 | } |
539 | if (prp->pr_usrreq == NULL) | | 544 | if (prp->pr_usrreq == NULL) |
540 | return EPROTONOSUPPORT; | | 545 | return EPROTONOSUPPORT; |
541 | if (prp->pr_type != type) | | 546 | if (prp->pr_type != type) |
542 | return EPROTOTYPE; | | 547 | return EPROTOTYPE; |
543 | | | 548 | |
544 | so = soget(true); | | 549 | so = soget(true); |
545 | so->so_type = type; | | 550 | so->so_type = type; |
546 | so->so_proto = prp; | | 551 | so->so_proto = prp; |
547 | so->so_send = sosend; | | 552 | so->so_send = sosend; |
548 | so->so_receive = soreceive; | | 553 | so->so_receive = soreceive; |
549 | #ifdef MBUFTRACE | | 554 | #ifdef MBUFTRACE |
550 | so->so_rcv.sb_mowner = &prp->pr_domain->dom_mowner; | | 555 | so->so_rcv.sb_mowner = &prp->pr_domain->dom_mowner; |
551 | so->so_snd.sb_mowner = &prp->pr_domain->dom_mowner; | | 556 | so->so_snd.sb_mowner = &prp->pr_domain->dom_mowner; |
552 | so->so_mowner = &prp->pr_domain->dom_mowner; | | 557 | so->so_mowner = &prp->pr_domain->dom_mowner; |
553 | #endif | | 558 | #endif |
554 | /* so->so_cred = kauth_cred_dup(l->l_cred); */ | | 559 | /* so->so_cred = kauth_cred_dup(l->l_cred); */ |
555 | uid = kauth_cred_geteuid(l->l_cred); | | 560 | uid = kauth_cred_geteuid(l->l_cred); |
556 | so->so_uidinfo = uid_find(uid); | | 561 | so->so_uidinfo = uid_find(uid); |
557 | so->so_egid = kauth_cred_getegid(l->l_cred); | | 562 | so->so_egid = kauth_cred_getegid(l->l_cred); |
558 | so->so_cpid = l->l_proc->p_pid; | | 563 | so->so_cpid = l->l_proc->p_pid; |
559 | if (lockso != NULL) { | | 564 | if (lockso != NULL) { |
560 | /* Caller wants us to share a lock. */ | | 565 | /* Caller wants us to share a lock. */ |
561 | lock = lockso->so_lock; | | 566 | lock = lockso->so_lock; |
562 | so->so_lock = lock; | | 567 | so->so_lock = lock; |
563 | mutex_obj_hold(lock); | | 568 | mutex_obj_hold(lock); |
564 | mutex_enter(lock); | | 569 | mutex_enter(lock); |
565 | } else { | | 570 | } else { |
566 | /* Lock assigned and taken during PRU_ATTACH. */ | | 571 | /* Lock assigned and taken during PRU_ATTACH. */ |
567 | } | | 572 | } |
568 | error = (*prp->pr_usrreq)(so, PRU_ATTACH, NULL, | | 573 | error = (*prp->pr_usrreq)(so, PRU_ATTACH, NULL, |
569 | (struct mbuf *)(long)proto, NULL, l); | | 574 | (struct mbuf *)(long)proto, NULL, l); |
570 | KASSERT(solocked(so)); | | 575 | KASSERT(solocked(so)); |
571 | if (error != 0) { | | 576 | if (error != 0) { |
572 | so->so_state |= SS_NOFDREF; | | 577 | so->so_state |= SS_NOFDREF; |
573 | sofree(so); | | 578 | sofree(so); |
574 | return error; | | 579 | return error; |
575 | } | | 580 | } |
576 | sounlock(so); | | 581 | sounlock(so); |
577 | *aso = so; | | 582 | *aso = so; |
578 | return 0; | | 583 | return 0; |
579 | } | | 584 | } |
580 | | | 585 | |
581 | /* On success, write file descriptor to fdout and return zero. On | | 586 | /* On success, write file descriptor to fdout and return zero. On |
582 | * failure, return non-zero; *fdout will be undefined. | | 587 | * failure, return non-zero; *fdout will be undefined. |
583 | */ | | 588 | */ |
584 | int | | 589 | int |
585 | fsocreate(int domain, struct socket **sop, int type, int protocol, | | 590 | fsocreate(int domain, struct socket **sop, int type, int protocol, |
586 | struct lwp *l, int *fdout) | | 591 | struct lwp *l, int *fdout) |
587 | { | | 592 | { |
588 | struct socket *so; | | 593 | struct socket *so; |
589 | struct file *fp; | | 594 | struct file *fp; |
590 | int fd, error; | | 595 | int fd, error; |
591 | | | 596 | |
592 | if ((error = fd_allocfile(&fp, &fd)) != 0) | | 597 | if ((error = fd_allocfile(&fp, &fd)) != 0) |
593 | return (error); | | 598 | return (error); |
594 | fp->f_flag = FREAD|FWRITE; | | 599 | fp->f_flag = FREAD|FWRITE; |
595 | fp->f_type = DTYPE_SOCKET; | | 600 | fp->f_type = DTYPE_SOCKET; |
596 | fp->f_ops = &socketops; | | 601 | fp->f_ops = &socketops; |
597 | error = socreate(domain, &so, type, protocol, l, NULL); | | 602 | error = socreate(domain, &so, type, protocol, l, NULL); |
598 | if (error != 0) { | | 603 | if (error != 0) { |
599 | fd_abort(curproc, fp, fd); | | 604 | fd_abort(curproc, fp, fd); |
600 | } else { | | 605 | } else { |
601 | if (sop != NULL) | | 606 | if (sop != NULL) |
602 | *sop = so; | | 607 | *sop = so; |
603 | fp->f_data = so; | | 608 | fp->f_data = so; |
604 | fd_affix(curproc, fp, fd); | | 609 | fd_affix(curproc, fp, fd); |
605 | *fdout = fd; | | 610 | *fdout = fd; |
606 | } | | 611 | } |
607 | return error; | | 612 | return error; |
608 | } | | 613 | } |
609 | | | 614 | |
610 | int | | 615 | int |
611 | sofamily(const struct socket *so) | | 616 | sofamily(const struct socket *so) |
612 | { | | 617 | { |
613 | const struct protosw *pr; | | 618 | const struct protosw *pr; |
614 | const struct domain *dom; | | 619 | const struct domain *dom; |
615 | | | 620 | |
616 | if ((pr = so->so_proto) == NULL) | | 621 | if ((pr = so->so_proto) == NULL) |
617 | return AF_UNSPEC; | | 622 | return AF_UNSPEC; |
618 | if ((dom = pr->pr_domain) == NULL) | | 623 | if ((dom = pr->pr_domain) == NULL) |
619 | return AF_UNSPEC; | | 624 | return AF_UNSPEC; |
620 | return dom->dom_family; | | 625 | return dom->dom_family; |
621 | } | | 626 | } |
622 | | | 627 | |
623 | int | | 628 | int |
624 | sobind(struct socket *so, struct mbuf *nam, struct lwp *l) | | 629 | sobind(struct socket *so, struct mbuf *nam, struct lwp *l) |
625 | { | | 630 | { |
626 | int error; | | 631 | int error; |
627 | | | 632 | |
628 | solock(so); | | 633 | solock(so); |
629 | error = (*so->so_proto->pr_usrreq)(so, PRU_BIND, NULL, nam, NULL, l); | | 634 | error = (*so->so_proto->pr_usrreq)(so, PRU_BIND, NULL, nam, NULL, l); |
630 | sounlock(so); | | 635 | sounlock(so); |
631 | return error; | | 636 | return error; |
632 | } | | 637 | } |
633 | | | 638 | |
634 | int | | 639 | int |
635 | solisten(struct socket *so, int backlog, struct lwp *l) | | 640 | solisten(struct socket *so, int backlog, struct lwp *l) |
636 | { | | 641 | { |
637 | int error; | | 642 | int error; |
638 | | | 643 | |
639 | solock(so); | | 644 | solock(so); |
640 | if ((so->so_state & (SS_ISCONNECTED | SS_ISCONNECTING | | | 645 | if ((so->so_state & (SS_ISCONNECTED | SS_ISCONNECTING | |
641 | SS_ISDISCONNECTING)) != 0) { | | 646 | SS_ISDISCONNECTING)) != 0) { |
642 | sounlock(so); | | 647 | sounlock(so); |
643 | return (EOPNOTSUPP); | | 648 | return (EOPNOTSUPP); |
644 | } | | 649 | } |
645 | error = (*so->so_proto->pr_usrreq)(so, PRU_LISTEN, NULL, | | 650 | error = (*so->so_proto->pr_usrreq)(so, PRU_LISTEN, NULL, |
646 | NULL, NULL, l); | | 651 | NULL, NULL, l); |
647 | if (error != 0) { | | 652 | if (error != 0) { |
648 | sounlock(so); | | 653 | sounlock(so); |
649 | return error; | | 654 | return error; |
650 | } | | 655 | } |
651 | if (TAILQ_EMPTY(&so->so_q)) | | 656 | if (TAILQ_EMPTY(&so->so_q)) |
652 | so->so_options |= SO_ACCEPTCONN; | | 657 | so->so_options |= SO_ACCEPTCONN; |
653 | if (backlog < 0) | | 658 | if (backlog < 0) |
654 | backlog = 0; | | 659 | backlog = 0; |
655 | so->so_qlimit = min(backlog, somaxconn); | | 660 | so->so_qlimit = min(backlog, somaxconn); |
656 | sounlock(so); | | 661 | sounlock(so); |
657 | return 0; | | 662 | return 0; |
658 | } | | 663 | } |
659 | | | 664 | |
660 | void | | 665 | void |
661 | sofree(struct socket *so) | | 666 | sofree(struct socket *so) |
662 | { | | 667 | { |
663 | u_int refs; | | 668 | u_int refs; |
664 | | | 669 | |
665 | KASSERT(solocked(so)); | | 670 | KASSERT(solocked(so)); |
666 | | | 671 | |
667 | if (so->so_pcb || (so->so_state & SS_NOFDREF) == 0) { | | 672 | if (so->so_pcb || (so->so_state & SS_NOFDREF) == 0) { |
668 | sounlock(so); | | 673 | sounlock(so); |
669 | return; | | 674 | return; |
670 | } | | 675 | } |
671 | if (so->so_head) { | | 676 | if (so->so_head) { |
672 | /* | | 677 | /* |
673 | * We must not decommission a socket that's on the accept(2) | | 678 | * We must not decommission a socket that's on the accept(2) |
674 | * queue. If we do, then accept(2) may hang after select(2) | | 679 | * queue. If we do, then accept(2) may hang after select(2) |
675 | * indicated that the listening socket was ready. | | 680 | * indicated that the listening socket was ready. |
676 | */ | | 681 | */ |
677 | if (!soqremque(so, 0)) { | | 682 | if (!soqremque(so, 0)) { |
678 | sounlock(so); | | 683 | sounlock(so); |
679 | return; | | 684 | return; |
680 | } | | 685 | } |
681 | } | | 686 | } |
682 | if (so->so_rcv.sb_hiwat) | | 687 | if (so->so_rcv.sb_hiwat) |
683 | (void)chgsbsize(so->so_uidinfo, &so->so_rcv.sb_hiwat, 0, | | 688 | (void)chgsbsize(so->so_uidinfo, &so->so_rcv.sb_hiwat, 0, |
684 | RLIM_INFINITY); | | 689 | RLIM_INFINITY); |
685 | if (so->so_snd.sb_hiwat) | | 690 | if (so->so_snd.sb_hiwat) |
686 | (void)chgsbsize(so->so_uidinfo, &so->so_snd.sb_hiwat, 0, | | 691 | (void)chgsbsize(so->so_uidinfo, &so->so_snd.sb_hiwat, 0, |
687 | RLIM_INFINITY); | | 692 | RLIM_INFINITY); |
688 | sbrelease(&so->so_snd, so); | | 693 | sbrelease(&so->so_snd, so); |
689 | KASSERT(!cv_has_waiters(&so->so_cv)); | | 694 | KASSERT(!cv_has_waiters(&so->so_cv)); |
690 | KASSERT(!cv_has_waiters(&so->so_rcv.sb_cv)); | | 695 | KASSERT(!cv_has_waiters(&so->so_rcv.sb_cv)); |
691 | KASSERT(!cv_has_waiters(&so->so_snd.sb_cv)); | | 696 | KASSERT(!cv_has_waiters(&so->so_snd.sb_cv)); |
692 | sorflush(so); | | 697 | sorflush(so); |
693 | refs = so->so_aborting; /* XXX */ | | 698 | refs = so->so_aborting; /* XXX */ |
694 | /* Remove acccept filter if one is present. */ | | 699 | /* Remove acccept filter if one is present. */ |
695 | if (so->so_accf != NULL) | | 700 | if (so->so_accf != NULL) |
696 | (void)accept_filt_clear(so); | | 701 | (void)accept_filt_clear(so); |
697 | /* kauth_cred_free(so->so_cred); */ | | 702 | /* kauth_cred_free(so->so_cred); */ |
698 | sounlock(so); | | 703 | sounlock(so); |
699 | if (refs == 0) /* XXX */ | | 704 | if (refs == 0) /* XXX */ |
700 | soput(so); | | 705 | soput(so); |
701 | } | | 706 | } |
702 | | | 707 | |
703 | /* | | 708 | /* |
704 | * Close a socket on last file table reference removal. | | 709 | * Close a socket on last file table reference removal. |
705 | * Initiate disconnect if connected. | | 710 | * Initiate disconnect if connected. |
706 | * Free socket when disconnect complete. | | 711 | * Free socket when disconnect complete. |
707 | */ | | 712 | */ |
708 | int | | 713 | int |
709 | soclose(struct socket *so) | | 714 | soclose(struct socket *so) |
710 | { | | 715 | { |
711 | struct socket *so2; | | 716 | struct socket *so2; |
712 | int error; | | 717 | int error; |
713 | int error2; | | 718 | int error2; |
714 | | | 719 | |
715 | error = 0; | | 720 | error = 0; |
716 | solock(so); | | 721 | solock(so); |
717 | if (so->so_options & SO_ACCEPTCONN) { | | 722 | if (so->so_options & SO_ACCEPTCONN) { |
718 | for (;;) { | | 723 | for (;;) { |
719 | if ((so2 = TAILQ_FIRST(&so->so_q0)) != 0) { | | 724 | if ((so2 = TAILQ_FIRST(&so->so_q0)) != 0) { |
720 | KASSERT(solocked2(so, so2)); | | 725 | KASSERT(solocked2(so, so2)); |
721 | (void) soqremque(so2, 0); | | 726 | (void) soqremque(so2, 0); |
722 | /* soabort drops the lock. */ | | 727 | /* soabort drops the lock. */ |
723 | (void) soabort(so2); | | 728 | (void) soabort(so2); |
724 | solock(so); | | 729 | solock(so); |
725 | continue; | | 730 | continue; |
726 | } | | 731 | } |
727 | if ((so2 = TAILQ_FIRST(&so->so_q)) != 0) { | | 732 | if ((so2 = TAILQ_FIRST(&so->so_q)) != 0) { |
728 | KASSERT(solocked2(so, so2)); | | 733 | KASSERT(solocked2(so, so2)); |
729 | (void) soqremque(so2, 1); | | 734 | (void) soqremque(so2, 1); |
730 | /* soabort drops the lock. */ | | 735 | /* soabort drops the lock. */ |
731 | (void) soabort(so2); | | 736 | (void) soabort(so2); |
732 | solock(so); | | 737 | solock(so); |
733 | continue; | | 738 | continue; |
734 | } | | 739 | } |
735 | break; | | 740 | break; |
736 | } | | 741 | } |
737 | } | | 742 | } |
738 | if (so->so_pcb == 0) | | 743 | if (so->so_pcb == 0) |
739 | goto discard; | | 744 | goto discard; |
740 | if (so->so_state & SS_ISCONNECTED) { | | 745 | if (so->so_state & SS_ISCONNECTED) { |
741 | if ((so->so_state & SS_ISDISCONNECTING) == 0) { | | 746 | if ((so->so_state & SS_ISDISCONNECTING) == 0) { |
742 | error = sodisconnect(so); | | 747 | error = sodisconnect(so); |
743 | if (error) | | 748 | if (error) |
744 | goto drop; | | 749 | goto drop; |
745 | } | | 750 | } |
746 | if (so->so_options & SO_LINGER) { | | 751 | if (so->so_options & SO_LINGER) { |
747 | if ((so->so_state & SS_ISDISCONNECTING) && so->so_nbio) | | 752 | if ((so->so_state & SS_ISDISCONNECTING) && so->so_nbio) |
748 | goto drop; | | 753 | goto drop; |
749 | while (so->so_state & SS_ISCONNECTED) { | | 754 | while (so->so_state & SS_ISCONNECTED) { |
750 | error = sowait(so, true, so->so_linger * hz); | | 755 | error = sowait(so, true, so->so_linger * hz); |
751 | if (error) | | 756 | if (error) |
752 | break; | | 757 | break; |
753 | } | | 758 | } |
754 | } | | 759 | } |
755 | } | | 760 | } |
756 | drop: | | 761 | drop: |
757 | if (so->so_pcb) { | | 762 | if (so->so_pcb) { |
758 | error2 = (*so->so_proto->pr_usrreq)(so, PRU_DETACH, | | 763 | error2 = (*so->so_proto->pr_usrreq)(so, PRU_DETACH, |
759 | NULL, NULL, NULL, NULL); | | 764 | NULL, NULL, NULL, NULL); |
760 | if (error == 0) | | 765 | if (error == 0) |
761 | error = error2; | | 766 | error = error2; |
762 | } | | 767 | } |
763 | discard: | | 768 | discard: |
764 | if (so->so_state & SS_NOFDREF) | | 769 | if (so->so_state & SS_NOFDREF) |
765 | panic("soclose: NOFDREF"); | | 770 | panic("soclose: NOFDREF"); |
766 | so->so_state |= SS_NOFDREF; | | 771 | so->so_state |= SS_NOFDREF; |
767 | sofree(so); | | 772 | sofree(so); |
768 | return (error); | | 773 | return (error); |
769 | } | | 774 | } |
770 | | | 775 | |
771 | /* | | 776 | /* |
772 | * Must be called with the socket locked.. Will return with it unlocked. | | 777 | * Must be called with the socket locked.. Will return with it unlocked. |
773 | */ | | 778 | */ |
774 | int | | 779 | int |
775 | soabort(struct socket *so) | | 780 | soabort(struct socket *so) |
776 | { | | 781 | { |
777 | u_int refs; | | 782 | u_int refs; |
778 | int error; | | 783 | int error; |
779 | | | 784 | |
780 | KASSERT(solocked(so)); | | 785 | KASSERT(solocked(so)); |
781 | KASSERT(so->so_head == NULL); | | 786 | KASSERT(so->so_head == NULL); |
782 | | | 787 | |
783 | so->so_aborting++; /* XXX */ | | 788 | so->so_aborting++; /* XXX */ |
784 | error = (*so->so_proto->pr_usrreq)(so, PRU_ABORT, NULL, | | 789 | error = (*so->so_proto->pr_usrreq)(so, PRU_ABORT, NULL, |
785 | NULL, NULL, NULL); | | 790 | NULL, NULL, NULL); |
786 | refs = --so->so_aborting; /* XXX */ | | 791 | refs = --so->so_aborting; /* XXX */ |
787 | if (error || (refs == 0)) { | | 792 | if (error || (refs == 0)) { |
788 | sofree(so); | | 793 | sofree(so); |
789 | } else { | | 794 | } else { |
790 | sounlock(so); | | 795 | sounlock(so); |
791 | } | | 796 | } |
792 | return error; | | 797 | return error; |
793 | } | | 798 | } |
794 | | | 799 | |
795 | int | | 800 | int |
796 | soaccept(struct socket *so, struct mbuf *nam) | | 801 | soaccept(struct socket *so, struct mbuf *nam) |
797 | { | | 802 | { |
798 | int error; | | 803 | int error; |
799 | | | 804 | |
800 | KASSERT(solocked(so)); | | 805 | KASSERT(solocked(so)); |
801 | | | 806 | |
802 | error = 0; | | 807 | error = 0; |
803 | if ((so->so_state & SS_NOFDREF) == 0) | | 808 | if ((so->so_state & SS_NOFDREF) == 0) |
804 | panic("soaccept: !NOFDREF"); | | 809 | panic("soaccept: !NOFDREF"); |
805 | so->so_state &= ~SS_NOFDREF; | | 810 | so->so_state &= ~SS_NOFDREF; |
806 | if ((so->so_state & SS_ISDISCONNECTED) == 0 || | | 811 | if ((so->so_state & SS_ISDISCONNECTED) == 0 || |
807 | (so->so_proto->pr_flags & PR_ABRTACPTDIS) == 0) | | 812 | (so->so_proto->pr_flags & PR_ABRTACPTDIS) == 0) |
808 | error = (*so->so_proto->pr_usrreq)(so, PRU_ACCEPT, | | 813 | error = (*so->so_proto->pr_usrreq)(so, PRU_ACCEPT, |
809 | NULL, nam, NULL, NULL); | | 814 | NULL, nam, NULL, NULL); |
810 | else | | 815 | else |
811 | error = ECONNABORTED; | | 816 | error = ECONNABORTED; |
812 | | | 817 | |
813 | return (error); | | 818 | return (error); |
814 | } | | 819 | } |
815 | | | 820 | |
816 | int | | 821 | int |
817 | soconnect(struct socket *so, struct mbuf *nam, struct lwp *l) | | 822 | soconnect(struct socket *so, struct mbuf *nam, struct lwp *l) |
818 | { | | 823 | { |
819 | int error; | | 824 | int error; |
820 | | | 825 | |
821 | KASSERT(solocked(so)); | | 826 | KASSERT(solocked(so)); |
822 | | | 827 | |
823 | if (so->so_options & SO_ACCEPTCONN) | | 828 | if (so->so_options & SO_ACCEPTCONN) |
824 | return (EOPNOTSUPP); | | 829 | return (EOPNOTSUPP); |
825 | /* | | 830 | /* |
826 | * If protocol is connection-based, can only connect once. | | 831 | * If protocol is connection-based, can only connect once. |
827 | * Otherwise, if connected, try to disconnect first. | | 832 | * Otherwise, if connected, try to disconnect first. |
828 | * This allows user to disconnect by connecting to, e.g., | | 833 | * This allows user to disconnect by connecting to, e.g., |
829 | * a null address. | | 834 | * a null address. |
830 | */ | | 835 | */ |
831 | if (so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING) && | | 836 | if (so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING) && |
832 | ((so->so_proto->pr_flags & PR_CONNREQUIRED) || | | 837 | ((so->so_proto->pr_flags & PR_CONNREQUIRED) || |
833 | (error = sodisconnect(so)))) | | 838 | (error = sodisconnect(so)))) |
834 | error = EISCONN; | | 839 | error = EISCONN; |
835 | else | | 840 | else |
836 | error = (*so->so_proto->pr_usrreq)(so, PRU_CONNECT, | | 841 | error = (*so->so_proto->pr_usrreq)(so, PRU_CONNECT, |
837 | NULL, nam, NULL, l); | | 842 | NULL, nam, NULL, l); |
838 | return (error); | | 843 | return (error); |
839 | } | | 844 | } |
840 | | | 845 | |
841 | int | | 846 | int |
842 | soconnect2(struct socket *so1, struct socket *so2) | | 847 | soconnect2(struct socket *so1, struct socket *so2) |
843 | { | | 848 | { |
844 | int error; | | 849 | int error; |
845 | | | 850 | |
846 | KASSERT(solocked2(so1, so2)); | | 851 | KASSERT(solocked2(so1, so2)); |
847 | | | 852 | |
848 | error = (*so1->so_proto->pr_usrreq)(so1, PRU_CONNECT2, | | 853 | error = (*so1->so_proto->pr_usrreq)(so1, PRU_CONNECT2, |
849 | NULL, (struct mbuf *)so2, NULL, NULL); | | 854 | NULL, (struct mbuf *)so2, NULL, NULL); |
850 | return (error); | | 855 | return (error); |
851 | } | | 856 | } |
852 | | | 857 | |
853 | int | | 858 | int |
854 | sodisconnect(struct socket *so) | | 859 | sodisconnect(struct socket *so) |
855 | { | | 860 | { |
856 | int error; | | 861 | int error; |
857 | | | 862 | |
858 | KASSERT(solocked(so)); | | 863 | KASSERT(solocked(so)); |
859 | | | 864 | |
860 | if ((so->so_state & SS_ISCONNECTED) == 0) { | | 865 | if ((so->so_state & SS_ISCONNECTED) == 0) { |
861 | error = ENOTCONN; | | 866 | error = ENOTCONN; |
862 | } else if (so->so_state & SS_ISDISCONNECTING) { | | 867 | } else if (so->so_state & SS_ISDISCONNECTING) { |
863 | error = EALREADY; | | 868 | error = EALREADY; |
864 | } else { | | 869 | } else { |
865 | error = (*so->so_proto->pr_usrreq)(so, PRU_DISCONNECT, | | 870 | error = (*so->so_proto->pr_usrreq)(so, PRU_DISCONNECT, |
866 | NULL, NULL, NULL, NULL); | | 871 | NULL, NULL, NULL, NULL); |
867 | } | | 872 | } |
868 | sodopendfree(); | | 873 | sodopendfree(); |
869 | return (error); | | 874 | return (error); |
870 | } | | 875 | } |
871 | | | 876 | |
872 | #define SBLOCKWAIT(f) (((f) & MSG_DONTWAIT) ? M_NOWAIT : M_WAITOK) | | 877 | #define SBLOCKWAIT(f) (((f) & MSG_DONTWAIT) ? M_NOWAIT : M_WAITOK) |
873 | /* | | 878 | /* |
874 | * Send on a socket. | | 879 | * Send on a socket. |
875 | * If send must go all at once and message is larger than | | 880 | * If send must go all at once and message is larger than |
876 | * send buffering, then hard error. | | 881 | * send buffering, then hard error. |
877 | * Lock against other senders. | | 882 | * Lock against other senders. |
878 | * If must go all at once and not enough room now, then | | 883 | * If must go all at once and not enough room now, then |
879 | * inform user that this would block and do nothing. | | 884 | * inform user that this would block and do nothing. |
880 | * Otherwise, if nonblocking, send as much as possible. | | 885 | * Otherwise, if nonblocking, send as much as possible. |
881 | * The data to be sent is described by "uio" if nonzero, | | 886 | * The data to be sent is described by "uio" if nonzero, |
882 | * otherwise by the mbuf chain "top" (which must be null | | 887 | * otherwise by the mbuf chain "top" (which must be null |
883 | * if uio is not). Data provided in mbuf chain must be small | | 888 | * if uio is not). Data provided in mbuf chain must be small |
884 | * enough to send all at once. | | 889 | * enough to send all at once. |
885 | * | | 890 | * |
886 | * Returns nonzero on error, timeout or signal; callers | | 891 | * Returns nonzero on error, timeout or signal; callers |
887 | * must check for short counts if EINTR/ERESTART are returned. | | 892 | * must check for short counts if EINTR/ERESTART are returned. |
888 | * Data and control buffers are freed on return. | | 893 | * Data and control buffers are freed on return. |
889 | */ | | 894 | */ |
890 | int | | 895 | int |
891 | sosend(struct socket *so, struct mbuf *addr, struct uio *uio, struct mbuf *top, | | 896 | sosend(struct socket *so, struct mbuf *addr, struct uio *uio, struct mbuf *top, |
892 | struct mbuf *control, int flags, struct lwp *l) | | 897 | struct mbuf *control, int flags, struct lwp *l) |
893 | { | | 898 | { |
894 | struct mbuf **mp, *m; | | 899 | struct mbuf **mp, *m; |
895 | struct proc *p; | | 900 | struct proc *p; |
896 | long space, len, resid, clen, mlen; | | 901 | long space, len, resid, clen, mlen; |
897 | int error, s, dontroute, atomic; | | 902 | int error, s, dontroute, atomic; |
898 | | | 903 | |
899 | p = l->l_proc; | | 904 | p = l->l_proc; |
900 | sodopendfree(); | | 905 | sodopendfree(); |
901 | clen = 0; | | 906 | clen = 0; |
902 | | | 907 | |
903 | /* | | 908 | /* |
904 | * solock() provides atomicity of access. splsoftnet() prevents | | 909 | * solock() provides atomicity of access. splsoftnet() prevents |
905 | * protocol processing soft interrupts from interrupting us and | | 910 | * protocol processing soft interrupts from interrupting us and |
906 | * blocking (expensive). | | 911 | * blocking (expensive). |
907 | */ | | 912 | */ |
908 | s = splsoftnet(); | | 913 | s = splsoftnet(); |
909 | solock(so); | | 914 | solock(so); |
910 | atomic = sosendallatonce(so) || top; | | 915 | atomic = sosendallatonce(so) || top; |
911 | if (uio) | | 916 | if (uio) |
912 | resid = uio->uio_resid; | | 917 | resid = uio->uio_resid; |
913 | else | | 918 | else |
914 | resid = top->m_pkthdr.len; | | 919 | resid = top->m_pkthdr.len; |
915 | /* | | 920 | /* |
916 | * In theory resid should be unsigned. | | 921 | * In theory resid should be unsigned. |
917 | * However, space must be signed, as it might be less than 0 | | 922 | * However, space must be signed, as it might be less than 0 |
918 | * if we over-committed, and we must use a signed comparison | | 923 | * if we over-committed, and we must use a signed comparison |
919 | * of space and resid. On the other hand, a negative resid | | 924 | * of space and resid. On the other hand, a negative resid |
920 | * causes us to loop sending 0-length segments to the protocol. | | 925 | * causes us to loop sending 0-length segments to the protocol. |
921 | */ | | 926 | */ |
922 | if (resid < 0) { | | 927 | if (resid < 0) { |
923 | error = EINVAL; | | 928 | error = EINVAL; |
924 | goto out; | | 929 | goto out; |
925 | } | | 930 | } |
926 | dontroute = | | 931 | dontroute = |
927 | (flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 && | | 932 | (flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 && |
928 | (so->so_proto->pr_flags & PR_ATOMIC); | | 933 | (so->so_proto->pr_flags & PR_ATOMIC); |
929 | l->l_ru.ru_msgsnd++; | | 934 | l->l_ru.ru_msgsnd++; |
930 | if (control) | | 935 | if (control) |
931 | clen = control->m_len; | | 936 | clen = control->m_len; |
932 | restart: | | 937 | restart: |
933 | if ((error = sblock(&so->so_snd, SBLOCKWAIT(flags))) != 0) | | 938 | if ((error = sblock(&so->so_snd, SBLOCKWAIT(flags))) != 0) |
934 | goto out; | | 939 | goto out; |
935 | do { | | 940 | do { |
936 | if (so->so_state & SS_CANTSENDMORE) { | | 941 | if (so->so_state & SS_CANTSENDMORE) { |
937 | error = EPIPE; | | 942 | error = EPIPE; |
938 | goto release; | | 943 | goto release; |
939 | } | | 944 | } |
940 | if (so->so_error) { | | 945 | if (so->so_error) { |
941 | error = so->so_error; | | 946 | error = so->so_error; |
942 | so->so_error = 0; | | 947 | so->so_error = 0; |
943 | goto release; | | 948 | goto release; |
944 | } | | 949 | } |
945 | if ((so->so_state & SS_ISCONNECTED) == 0) { | | 950 | if ((so->so_state & SS_ISCONNECTED) == 0) { |
946 | if (so->so_proto->pr_flags & PR_CONNREQUIRED) { | | 951 | if (so->so_proto->pr_flags & PR_CONNREQUIRED) { |
947 | if ((so->so_state & SS_ISCONFIRMING) == 0 && | | 952 | if ((so->so_state & SS_ISCONFIRMING) == 0 && |
948 | !(resid == 0 && clen != 0)) { | | 953 | !(resid == 0 && clen != 0)) { |
949 | error = ENOTCONN; | | 954 | error = ENOTCONN; |
950 | goto release; | | 955 | goto release; |
951 | } | | 956 | } |
952 | } else if (addr == 0) { | | 957 | } else if (addr == 0) { |
953 | error = EDESTADDRREQ; | | 958 | error = EDESTADDRREQ; |
954 | goto release; | | 959 | goto release; |
955 | } | | 960 | } |
956 | } | | 961 | } |
957 | space = sbspace(&so->so_snd); | | 962 | space = sbspace(&so->so_snd); |
958 | if (flags & MSG_OOB) | | 963 | if (flags & MSG_OOB) |
959 | space += 1024; | | 964 | space += 1024; |
960 | if ((atomic && resid > so->so_snd.sb_hiwat) || | | 965 | if ((atomic && resid > so->so_snd.sb_hiwat) || |
961 | clen > so->so_snd.sb_hiwat) { | | 966 | clen > so->so_snd.sb_hiwat) { |
962 | error = EMSGSIZE; | | 967 | error = EMSGSIZE; |
963 | goto release; | | 968 | goto release; |
964 | } | | 969 | } |
965 | if (space < resid + clen && | | 970 | if (space < resid + clen && |
966 | (atomic || space < so->so_snd.sb_lowat || space < clen)) { | | 971 | (atomic || space < so->so_snd.sb_lowat || space < clen)) { |
967 | if (so->so_nbio) { | | 972 | if (so->so_nbio) { |
968 | error = EWOULDBLOCK; | | 973 | error = EWOULDBLOCK; |
969 | goto release; | | 974 | goto release; |
970 | } | | 975 | } |
971 | sbunlock(&so->so_snd); | | 976 | sbunlock(&so->so_snd); |
972 | error = sbwait(&so->so_snd); | | 977 | error = sbwait(&so->so_snd); |
973 | if (error) | | 978 | if (error) |
974 | goto out; | | 979 | goto out; |
975 | goto restart; | | 980 | goto restart; |
976 | } | | 981 | } |
977 | mp = ⊤ | | 982 | mp = ⊤ |
978 | space -= clen; | | 983 | space -= clen; |
979 | do { | | 984 | do { |
980 | if (uio == NULL) { | | 985 | if (uio == NULL) { |
981 | /* | | 986 | /* |
982 | * Data is prepackaged in "top". | | 987 | * Data is prepackaged in "top". |
983 | */ | | 988 | */ |
984 | resid = 0; | | 989 | resid = 0; |
985 | if (flags & MSG_EOR) | | 990 | if (flags & MSG_EOR) |
986 | top->m_flags |= M_EOR; | | 991 | top->m_flags |= M_EOR; |
987 | } else do { | | 992 | } else do { |
988 | sounlock(so); | | 993 | sounlock(so); |
989 | splx(s); | | 994 | splx(s); |
990 | if (top == NULL) { | | 995 | if (top == NULL) { |
991 | m = m_gethdr(M_WAIT, MT_DATA); | | 996 | m = m_gethdr(M_WAIT, MT_DATA); |
992 | mlen = MHLEN; | | 997 | mlen = MHLEN; |
993 | m->m_pkthdr.len = 0; | | 998 | m->m_pkthdr.len = 0; |
994 | m->m_pkthdr.rcvif = NULL; | | 999 | m->m_pkthdr.rcvif = NULL; |
995 | } else { | | 1000 | } else { |
996 | m = m_get(M_WAIT, MT_DATA); | | 1001 | m = m_get(M_WAIT, MT_DATA); |
997 | mlen = MLEN; | | 1002 | mlen = MLEN; |
998 | } | | 1003 | } |
999 | MCLAIM(m, so->so_snd.sb_mowner); | | 1004 | MCLAIM(m, so->so_snd.sb_mowner); |
1000 | if (sock_loan_thresh >= 0 && | | 1005 | if (sock_loan_thresh >= 0 && |
1001 | uio->uio_iov->iov_len >= sock_loan_thresh && | | 1006 | uio->uio_iov->iov_len >= sock_loan_thresh && |
1002 | space >= sock_loan_thresh && | | 1007 | space >= sock_loan_thresh && |
1003 | (len = sosend_loan(so, uio, m, | | 1008 | (len = sosend_loan(so, uio, m, |
1004 | space)) != 0) { | | 1009 | space)) != 0) { |
1005 | SOSEND_COUNTER_INCR(&sosend_loan_big); | | 1010 | SOSEND_COUNTER_INCR(&sosend_loan_big); |
1006 | space -= len; | | 1011 | space -= len; |
1007 | goto have_data; | | 1012 | goto have_data; |
1008 | } | | 1013 | } |
1009 | if (resid >= MINCLSIZE && space >= MCLBYTES) { | | 1014 | if (resid >= MINCLSIZE && space >= MCLBYTES) { |
1010 | SOSEND_COUNTER_INCR(&sosend_copy_big); | | 1015 | SOSEND_COUNTER_INCR(&sosend_copy_big); |
1011 | m_clget(m, M_WAIT); | | 1016 | m_clget(m, M_WAIT); |
1012 | if ((m->m_flags & M_EXT) == 0) | | 1017 | if ((m->m_flags & M_EXT) == 0) |
1013 | goto nopages; | | 1018 | goto nopages; |
1014 | mlen = MCLBYTES; | | 1019 | mlen = MCLBYTES; |
1015 | if (atomic && top == 0) { | | 1020 | if (atomic && top == 0) { |
1016 | len = lmin(MCLBYTES - max_hdr, | | 1021 | len = lmin(MCLBYTES - max_hdr, |
1017 | resid); | | 1022 | resid); |
1018 | m->m_data += max_hdr; | | 1023 | m->m_data += max_hdr; |
1019 | } else | | 1024 | } else |
1020 | len = lmin(MCLBYTES, resid); | | 1025 | len = lmin(MCLBYTES, resid); |
1021 | space -= len; | | 1026 | space -= len; |
1022 | } else { | | 1027 | } else { |
1023 | nopages: | | 1028 | nopages: |
1024 | SOSEND_COUNTER_INCR(&sosend_copy_small); | | 1029 | SOSEND_COUNTER_INCR(&sosend_copy_small); |
1025 | len = lmin(lmin(mlen, resid), space); | | 1030 | len = lmin(lmin(mlen, resid), space); |
1026 | space -= len; | | 1031 | space -= len; |
1027 | /* | | 1032 | /* |
1028 | * For datagram protocols, leave room | | 1033 | * For datagram protocols, leave room |
1029 | * for protocol headers in first mbuf. | | 1034 | * for protocol headers in first mbuf. |
1030 | */ | | 1035 | */ |
1031 | if (atomic && top == 0 && len < mlen) | | 1036 | if (atomic && top == 0 && len < mlen) |
1032 | MH_ALIGN(m, len); | | 1037 | MH_ALIGN(m, len); |
1033 | } | | 1038 | } |
1034 | error = uiomove(mtod(m, void *), (int)len, uio); | | 1039 | error = uiomove(mtod(m, void *), (int)len, uio); |
1035 | have_data: | | 1040 | have_data: |
1036 | resid = uio->uio_resid; | | 1041 | resid = uio->uio_resid; |
1037 | m->m_len = len; | | 1042 | m->m_len = len; |
1038 | *mp = m; | | 1043 | *mp = m; |
1039 | top->m_pkthdr.len += len; | | 1044 | top->m_pkthdr.len += len; |
1040 | s = splsoftnet(); | | 1045 | s = splsoftnet(); |
1041 | solock(so); | | 1046 | solock(so); |
1042 | if (error != 0) | | 1047 | if (error != 0) |
1043 | goto release; | | 1048 | goto release; |
1044 | mp = &m->m_next; | | 1049 | mp = &m->m_next; |
1045 | if (resid <= 0) { | | 1050 | if (resid <= 0) { |
1046 | if (flags & MSG_EOR) | | 1051 | if (flags & MSG_EOR) |
1047 | top->m_flags |= M_EOR; | | 1052 | top->m_flags |= M_EOR; |
1048 | break; | | 1053 | break; |
1049 | } | | 1054 | } |
1050 | } while (space > 0 && atomic); | | 1055 | } while (space > 0 && atomic); |
1051 | | | 1056 | |
1052 | if (so->so_state & SS_CANTSENDMORE) { | | 1057 | if (so->so_state & SS_CANTSENDMORE) { |
1053 | error = EPIPE; | | 1058 | error = EPIPE; |
1054 | goto release; | | 1059 | goto release; |
1055 | } | | 1060 | } |
1056 | if (dontroute) | | 1061 | if (dontroute) |
1057 | so->so_options |= SO_DONTROUTE; | | 1062 | so->so_options |= SO_DONTROUTE; |
1058 | if (resid > 0) | | 1063 | if (resid > 0) |
1059 | so->so_state |= SS_MORETOCOME; | | 1064 | so->so_state |= SS_MORETOCOME; |
1060 | error = (*so->so_proto->pr_usrreq)(so, | | 1065 | error = (*so->so_proto->pr_usrreq)(so, |
1061 | (flags & MSG_OOB) ? PRU_SENDOOB : PRU_SEND, | | 1066 | (flags & MSG_OOB) ? PRU_SENDOOB : PRU_SEND, |
1062 | top, addr, control, curlwp); | | 1067 | top, addr, control, curlwp); |
1063 | if (dontroute) | | 1068 | if (dontroute) |
1064 | so->so_options &= ~SO_DONTROUTE; | | 1069 | so->so_options &= ~SO_DONTROUTE; |
1065 | if (resid > 0) | | 1070 | if (resid > 0) |
1066 | so->so_state &= ~SS_MORETOCOME; | | 1071 | so->so_state &= ~SS_MORETOCOME; |
1067 | clen = 0; | | 1072 | clen = 0; |
1068 | control = NULL; | | 1073 | control = NULL; |
1069 | top = NULL; | | 1074 | top = NULL; |
1070 | mp = ⊤ | | 1075 | mp = ⊤ |
1071 | if (error != 0) | | 1076 | if (error != 0) |
1072 | goto release; | | 1077 | goto release; |
1073 | } while (resid && space > 0); | | 1078 | } while (resid && space > 0); |
1074 | } while (resid); | | 1079 | } while (resid); |
1075 | | | 1080 | |
1076 | release: | | 1081 | release: |
1077 | sbunlock(&so->so_snd); | | 1082 | sbunlock(&so->so_snd); |
1078 | out: | | 1083 | out: |
1079 | sounlock(so); | | 1084 | sounlock(so); |
1080 | splx(s); | | 1085 | splx(s); |
1081 | if (top) | | 1086 | if (top) |
1082 | m_freem(top); | | 1087 | m_freem(top); |
1083 | if (control) | | 1088 | if (control) |
1084 | m_freem(control); | | 1089 | m_freem(control); |
1085 | return (error); | | 1090 | return (error); |
1086 | } | | 1091 | } |
1087 | | | 1092 | |
1088 | /* | | 1093 | /* |
1089 | * Following replacement or removal of the first mbuf on the first | | 1094 | * Following replacement or removal of the first mbuf on the first |
1090 | * mbuf chain of a socket buffer, push necessary state changes back | | 1095 | * mbuf chain of a socket buffer, push necessary state changes back |
1091 | * into the socket buffer so that other consumers see the values | | 1096 | * into the socket buffer so that other consumers see the values |
1092 | * consistently. 'nextrecord' is the callers locally stored value of | | 1097 | * consistently. 'nextrecord' is the callers locally stored value of |
1093 | * the original value of sb->sb_mb->m_nextpkt which must be restored | | 1098 | * the original value of sb->sb_mb->m_nextpkt which must be restored |
1094 | * when the lead mbuf changes. NOTE: 'nextrecord' may be NULL. | | 1099 | * when the lead mbuf changes. NOTE: 'nextrecord' may be NULL. |
1095 | */ | | 1100 | */ |
1096 | static void | | 1101 | static void |
1097 | sbsync(struct sockbuf *sb, struct mbuf *nextrecord) | | 1102 | sbsync(struct sockbuf *sb, struct mbuf *nextrecord) |
1098 | { | | 1103 | { |
1099 | | | 1104 | |
1100 | KASSERT(solocked(sb->sb_so)); | | 1105 | KASSERT(solocked(sb->sb_so)); |
1101 | | | 1106 | |
1102 | /* | | 1107 | /* |
1103 | * First, update for the new value of nextrecord. If necessary, | | 1108 | * First, update for the new value of nextrecord. If necessary, |
1104 | * make it the first record. | | 1109 | * make it the first record. |
1105 | */ | | 1110 | */ |
1106 | if (sb->sb_mb != NULL) | | 1111 | if (sb->sb_mb != NULL) |
1107 | sb->sb_mb->m_nextpkt = nextrecord; | | 1112 | sb->sb_mb->m_nextpkt = nextrecord; |
1108 | else | | 1113 | else |
1109 | sb->sb_mb = nextrecord; | | 1114 | sb->sb_mb = nextrecord; |
1110 | | | 1115 | |
1111 | /* | | 1116 | /* |
1112 | * Now update any dependent socket buffer fields to reflect | | 1117 | * Now update any dependent socket buffer fields to reflect |
1113 | * the new state. This is an inline of SB_EMPTY_FIXUP, with | | 1118 | * the new state. This is an inline of SB_EMPTY_FIXUP, with |
1114 | * the addition of a second clause that takes care of the | | 1119 | * the addition of a second clause that takes care of the |
1115 | * case where sb_mb has been updated, but remains the last | | 1120 | * case where sb_mb has been updated, but remains the last |
1116 | * record. | | 1121 | * record. |
1117 | */ | | 1122 | */ |
1118 | if (sb->sb_mb == NULL) { | | 1123 | if (sb->sb_mb == NULL) { |
1119 | sb->sb_mbtail = NULL; | | 1124 | sb->sb_mbtail = NULL; |
1120 | sb->sb_lastrecord = NULL; | | 1125 | sb->sb_lastrecord = NULL; |
1121 | } else if (sb->sb_mb->m_nextpkt == NULL) | | 1126 | } else if (sb->sb_mb->m_nextpkt == NULL) |
1122 | sb->sb_lastrecord = sb->sb_mb; | | 1127 | sb->sb_lastrecord = sb->sb_mb; |
1123 | } | | 1128 | } |
1124 | | | 1129 | |
1125 | /* | | 1130 | /* |
1126 | * Implement receive operations on a socket. | | 1131 | * Implement receive operations on a socket. |
1127 | * We depend on the way that records are added to the sockbuf | | 1132 | * We depend on the way that records are added to the sockbuf |
1128 | * by sbappend*. In particular, each record (mbufs linked through m_next) | | 1133 | * by sbappend*. In particular, each record (mbufs linked through m_next) |
1129 | * must begin with an address if the protocol so specifies, | | 1134 | * must begin with an address if the protocol so specifies, |
1130 | * followed by an optional mbuf or mbufs containing ancillary data, | | 1135 | * followed by an optional mbuf or mbufs containing ancillary data, |
1131 | * and then zero or more mbufs of data. | | 1136 | * and then zero or more mbufs of data. |
1132 | * In order to avoid blocking network interrupts for the entire time here, | | 1137 | * In order to avoid blocking network interrupts for the entire time here, |
1133 | * we splx() while doing the actual copy to user space. | | 1138 | * we splx() while doing the actual copy to user space. |
1134 | * Although the sockbuf is locked, new data may still be appended, | | 1139 | * Although the sockbuf is locked, new data may still be appended, |
1135 | * and thus we must maintain consistency of the sockbuf during that time. | | 1140 | * and thus we must maintain consistency of the sockbuf during that time. |
1136 | * | | 1141 | * |
1137 | * The caller may receive the data as a single mbuf chain by supplying | | 1142 | * The caller may receive the data as a single mbuf chain by supplying |
1138 | * an mbuf **mp0 for use in returning the chain. The uio is then used | | 1143 | * an mbuf **mp0 for use in returning the chain. The uio is then used |
1139 | * only for the count in uio_resid. | | 1144 | * only for the count in uio_resid. |
1140 | */ | | 1145 | */ |
1141 | int | | 1146 | int |
1142 | soreceive(struct socket *so, struct mbuf **paddr, struct uio *uio, | | 1147 | soreceive(struct socket *so, struct mbuf **paddr, struct uio *uio, |
1143 | struct mbuf **mp0, struct mbuf **controlp, int *flagsp) | | 1148 | struct mbuf **mp0, struct mbuf **controlp, int *flagsp) |
1144 | { | | 1149 | { |
1145 | struct lwp *l = curlwp; | | 1150 | struct lwp *l = curlwp; |
1146 | struct mbuf *m, **mp, *mt; | | 1151 | struct mbuf *m, **mp, *mt; |
1147 | int atomic, flags, len, error, s, offset, moff, type, orig_resid; | | 1152 | int atomic, flags, len, error, s, offset, moff, type, orig_resid; |
1148 | const struct protosw *pr; | | 1153 | const struct protosw *pr; |
1149 | struct mbuf *nextrecord; | | 1154 | struct mbuf *nextrecord; |
1150 | int mbuf_removed = 0; | | 1155 | int mbuf_removed = 0; |
1151 | const struct domain *dom; | | 1156 | const struct domain *dom; |
1152 | | | 1157 | |
1153 | pr = so->so_proto; | | 1158 | pr = so->so_proto; |
1154 | atomic = pr->pr_flags & PR_ATOMIC; | | 1159 | atomic = pr->pr_flags & PR_ATOMIC; |
1155 | dom = pr->pr_domain; | | 1160 | dom = pr->pr_domain; |
1156 | mp = mp0; | | 1161 | mp = mp0; |
1157 | type = 0; | | 1162 | type = 0; |
1158 | orig_resid = uio->uio_resid; | | 1163 | orig_resid = uio->uio_resid; |
1159 | | | 1164 | |
1160 | if (paddr != NULL) | | 1165 | if (paddr != NULL) |
1161 | *paddr = NULL; | | 1166 | *paddr = NULL; |
1162 | if (controlp != NULL) | | 1167 | if (controlp != NULL) |
1163 | *controlp = NULL; | | 1168 | *controlp = NULL; |
1164 | if (flagsp != NULL) | | 1169 | if (flagsp != NULL) |
1165 | flags = *flagsp &~ MSG_EOR; | | 1170 | flags = *flagsp &~ MSG_EOR; |
1166 | else | | 1171 | else |
1167 | flags = 0; | | 1172 | flags = 0; |
1168 | | | 1173 | |
1169 | if ((flags & MSG_DONTWAIT) == 0) | | 1174 | if ((flags & MSG_DONTWAIT) == 0) |
1170 | sodopendfree(); | | 1175 | sodopendfree(); |
1171 | | | 1176 | |
1172 | if (flags & MSG_OOB) { | | 1177 | if (flags & MSG_OOB) { |
1173 | m = m_get(M_WAIT, MT_DATA); | | 1178 | m = m_get(M_WAIT, MT_DATA); |
1174 | solock(so); | | 1179 | solock(so); |
1175 | error = (*pr->pr_usrreq)(so, PRU_RCVOOB, m, | | 1180 | error = (*pr->pr_usrreq)(so, PRU_RCVOOB, m, |
1176 | (struct mbuf *)(long)(flags & MSG_PEEK), NULL, l); | | 1181 | (struct mbuf *)(long)(flags & MSG_PEEK), NULL, l); |
1177 | sounlock(so); | | 1182 | sounlock(so); |
1178 | if (error) | | 1183 | if (error) |
1179 | goto bad; | | 1184 | goto bad; |
1180 | do { | | 1185 | do { |
1181 | error = uiomove(mtod(m, void *), | | 1186 | error = uiomove(mtod(m, void *), |
1182 | (int) min(uio->uio_resid, m->m_len), uio); | | 1187 | (int) min(uio->uio_resid, m->m_len), uio); |
1183 | m = m_free(m); | | 1188 | m = m_free(m); |
1184 | } while (uio->uio_resid > 0 && error == 0 && m); | | 1189 | } while (uio->uio_resid > 0 && error == 0 && m); |
1185 | bad: | | 1190 | bad: |
1186 | if (m != NULL) | | 1191 | if (m != NULL) |
1187 | m_freem(m); | | 1192 | m_freem(m); |
1188 | return error; | | 1193 | return error; |
1189 | } | | 1194 | } |
1190 | if (mp != NULL) | | 1195 | if (mp != NULL) |
1191 | *mp = NULL; | | 1196 | *mp = NULL; |
1192 | | | 1197 | |
1193 | /* | | 1198 | /* |
1194 | * solock() provides atomicity of access. splsoftnet() prevents | | 1199 | * solock() provides atomicity of access. splsoftnet() prevents |
1195 | * protocol processing soft interrupts from interrupting us and | | 1200 | * protocol processing soft interrupts from interrupting us and |
1196 | * blocking (expensive). | | 1201 | * blocking (expensive). |
1197 | */ | | 1202 | */ |
1198 | s = splsoftnet(); | | 1203 | s = splsoftnet(); |
1199 | solock(so); | | 1204 | solock(so); |
1200 | if (so->so_state & SS_ISCONFIRMING && uio->uio_resid) | | 1205 | if (so->so_state & SS_ISCONFIRMING && uio->uio_resid) |
1201 | (*pr->pr_usrreq)(so, PRU_RCVD, NULL, NULL, NULL, l); | | 1206 | (*pr->pr_usrreq)(so, PRU_RCVD, NULL, NULL, NULL, l); |
1202 | | | 1207 | |
1203 | restart: | | 1208 | restart: |
1204 | if ((error = sblock(&so->so_rcv, SBLOCKWAIT(flags))) != 0) { | | 1209 | if ((error = sblock(&so->so_rcv, SBLOCKWAIT(flags))) != 0) { |
1205 | sounlock(so); | | 1210 | sounlock(so); |
1206 | splx(s); | | 1211 | splx(s); |
1207 | return error; | | 1212 | return error; |
1208 | } | | 1213 | } |
1209 | | | 1214 | |
1210 | m = so->so_rcv.sb_mb; | | 1215 | m = so->so_rcv.sb_mb; |
1211 | /* | | 1216 | /* |
1212 | * If we have less data than requested, block awaiting more | | 1217 | * If we have less data than requested, block awaiting more |
1213 | * (subject to any timeout) if: | | 1218 | * (subject to any timeout) if: |
1214 | * 1. the current count is less than the low water mark, | | 1219 | * 1. the current count is less than the low water mark, |
1215 | * 2. MSG_WAITALL is set, and it is possible to do the entire | | 1220 | * 2. MSG_WAITALL is set, and it is possible to do the entire |
1216 | * receive operation at once if we block (resid <= hiwat), or | | 1221 | * receive operation at once if we block (resid <= hiwat), or |
1217 | * 3. MSG_DONTWAIT is not set. | | 1222 | * 3. MSG_DONTWAIT is not set. |
1218 | * If MSG_WAITALL is set but resid is larger than the receive buffer, | | 1223 | * If MSG_WAITALL is set but resid is larger than the receive buffer, |
1219 | * we have to do the receive in sections, and thus risk returning | | 1224 | * we have to do the receive in sections, and thus risk returning |
1220 | * a short count if a timeout or signal occurs after we start. | | 1225 | * a short count if a timeout or signal occurs after we start. |
1221 | */ | | 1226 | */ |
1222 | if (m == NULL || | | 1227 | if (m == NULL || |
1223 | ((flags & MSG_DONTWAIT) == 0 && | | 1228 | ((flags & MSG_DONTWAIT) == 0 && |
1224 | so->so_rcv.sb_cc < uio->uio_resid && | | 1229 | so->so_rcv.sb_cc < uio->uio_resid && |
1225 | (so->so_rcv.sb_cc < so->so_rcv.sb_lowat || | | 1230 | (so->so_rcv.sb_cc < so->so_rcv.sb_lowat || |
1226 | ((flags & MSG_WAITALL) && | | 1231 | ((flags & MSG_WAITALL) && |
1227 | uio->uio_resid <= so->so_rcv.sb_hiwat)) && | | 1232 | uio->uio_resid <= so->so_rcv.sb_hiwat)) && |
1228 | m->m_nextpkt == NULL && !atomic)) { | | 1233 | m->m_nextpkt == NULL && !atomic)) { |
1229 | #ifdef DIAGNOSTIC | | 1234 | #ifdef DIAGNOSTIC |
1230 | if (m == NULL && so->so_rcv.sb_cc) | | 1235 | if (m == NULL && so->so_rcv.sb_cc) |
1231 | panic("receive 1"); | | 1236 | panic("receive 1"); |
1232 | #endif | | 1237 | #endif |
1233 | if (so->so_error) { | | 1238 | if (so->so_error) { |
1234 | if (m != NULL) | | 1239 | if (m != NULL) |
1235 | goto dontblock; | | 1240 | goto dontblock; |
1236 | error = so->so_error; | | 1241 | error = so->so_error; |
1237 | if ((flags & MSG_PEEK) == 0) | | 1242 | if ((flags & MSG_PEEK) == 0) |
1238 | so->so_error = 0; | | 1243 | so->so_error = 0; |
1239 | goto release; | | 1244 | goto release; |
1240 | } | | 1245 | } |
1241 | if (so->so_state & SS_CANTRCVMORE) { | | 1246 | if (so->so_state & SS_CANTRCVMORE) { |
1242 | if (m != NULL) | | 1247 | if (m != NULL) |
1243 | goto dontblock; | | 1248 | goto dontblock; |
1244 | else | | 1249 | else |
1245 | goto release; | | 1250 | goto release; |
1246 | } | | 1251 | } |
1247 | for (; m != NULL; m = m->m_next) | | 1252 | for (; m != NULL; m = m->m_next) |
1248 | if (m->m_type == MT_OOBDATA || (m->m_flags & M_EOR)) { | | 1253 | if (m->m_type == MT_OOBDATA || (m->m_flags & M_EOR)) { |
1249 | m = so->so_rcv.sb_mb; | | 1254 | m = so->so_rcv.sb_mb; |
1250 | goto dontblock; | | 1255 | goto dontblock; |
1251 | } | | 1256 | } |
1252 | if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0 && | | 1257 | if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0 && |
1253 | (so->so_proto->pr_flags & PR_CONNREQUIRED)) { | | 1258 | (so->so_proto->pr_flags & PR_CONNREQUIRED)) { |
1254 | error = ENOTCONN; | | 1259 | error = ENOTCONN; |
1255 | goto release; | | 1260 | goto release; |
1256 | } | | 1261 | } |
1257 | if (uio->uio_resid == 0) | | 1262 | if (uio->uio_resid == 0) |
1258 | goto release; | | 1263 | goto release; |
1259 | if (so->so_nbio || (flags & MSG_DONTWAIT)) { | | 1264 | if (so->so_nbio || (flags & MSG_DONTWAIT)) { |
1260 | error = EWOULDBLOCK; | | 1265 | error = EWOULDBLOCK; |
1261 | goto release; | | 1266 | goto release; |
1262 | } | | 1267 | } |
1263 | SBLASTRECORDCHK(&so->so_rcv, "soreceive sbwait 1"); | | 1268 | SBLASTRECORDCHK(&so->so_rcv, "soreceive sbwait 1"); |
1264 | SBLASTMBUFCHK(&so->so_rcv, "soreceive sbwait 1"); | | 1269 | SBLASTMBUFCHK(&so->so_rcv, "soreceive sbwait 1"); |
1265 | sbunlock(&so->so_rcv); | | 1270 | sbunlock(&so->so_rcv); |
1266 | error = sbwait(&so->so_rcv); | | 1271 | error = sbwait(&so->so_rcv); |
1267 | if (error != 0) { | | 1272 | if (error != 0) { |
1268 | sounlock(so); | | 1273 | sounlock(so); |
1269 | splx(s); | | 1274 | splx(s); |
1270 | return error; | | 1275 | return error; |
1271 | } | | 1276 | } |
1272 | goto restart; | | 1277 | goto restart; |
1273 | } | | 1278 | } |
1274 | dontblock: | | 1279 | dontblock: |
1275 | /* | | 1280 | /* |
1276 | * On entry here, m points to the first record of the socket buffer. | | 1281 | * On entry here, m points to the first record of the socket buffer. |
1277 | * From this point onward, we maintain 'nextrecord' as a cache of the | | 1282 | * From this point onward, we maintain 'nextrecord' as a cache of the |
1278 | * pointer to the next record in the socket buffer. We must keep the | | 1283 | * pointer to the next record in the socket buffer. We must keep the |
1279 | * various socket buffer pointers and local stack versions of the | | 1284 | * various socket buffer pointers and local stack versions of the |
1280 | * pointers in sync, pushing out modifications before dropping the | | 1285 | * pointers in sync, pushing out modifications before dropping the |
1281 | * socket lock, and re-reading them when picking it up. | | 1286 | * socket lock, and re-reading them when picking it up. |
1282 | * | | 1287 | * |
1283 | * Otherwise, we will race with the network stack appending new data | | 1288 | * Otherwise, we will race with the network stack appending new data |
1284 | * or records onto the socket buffer by using inconsistent/stale | | 1289 | * or records onto the socket buffer by using inconsistent/stale |
1285 | * versions of the field, possibly resulting in socket buffer | | 1290 | * versions of the field, possibly resulting in socket buffer |
1286 | * corruption. | | 1291 | * corruption. |
1287 | * | | 1292 | * |
1288 | * By holding the high-level sblock(), we prevent simultaneous | | 1293 | * By holding the high-level sblock(), we prevent simultaneous |
1289 | * readers from pulling off the front of the socket buffer. | | 1294 | * readers from pulling off the front of the socket buffer. |
1290 | */ | | 1295 | */ |
1291 | if (l != NULL) | | 1296 | if (l != NULL) |
1292 | l->l_ru.ru_msgrcv++; | | 1297 | l->l_ru.ru_msgrcv++; |
1293 | KASSERT(m == so->so_rcv.sb_mb); | | 1298 | KASSERT(m == so->so_rcv.sb_mb); |
1294 | SBLASTRECORDCHK(&so->so_rcv, "soreceive 1"); | | 1299 | SBLASTRECORDCHK(&so->so_rcv, "soreceive 1"); |
1295 | SBLASTMBUFCHK(&so->so_rcv, "soreceive 1"); | | 1300 | SBLASTMBUFCHK(&so->so_rcv, "soreceive 1"); |
1296 | nextrecord = m->m_nextpkt; | | 1301 | nextrecord = m->m_nextpkt; |
1297 | if (pr->pr_flags & PR_ADDR) { | | 1302 | if (pr->pr_flags & PR_ADDR) { |
1298 | #ifdef DIAGNOSTIC | | 1303 | #ifdef DIAGNOSTIC |
1299 | if (m->m_type != MT_SONAME) | | 1304 | if (m->m_type != MT_SONAME) |
1300 | panic("receive 1a"); | | 1305 | panic("receive 1a"); |
1301 | #endif | | 1306 | #endif |
1302 | orig_resid = 0; | | 1307 | orig_resid = 0; |
1303 | if (flags & MSG_PEEK) { | | 1308 | if (flags & MSG_PEEK) { |
1304 | if (paddr) | | 1309 | if (paddr) |
1305 | *paddr = m_copy(m, 0, m->m_len); | | 1310 | *paddr = m_copy(m, 0, m->m_len); |
1306 | m = m->m_next; | | 1311 | m = m->m_next; |
1307 | } else { | | 1312 | } else { |
1308 | sbfree(&so->so_rcv, m); | | 1313 | sbfree(&so->so_rcv, m); |
1309 | mbuf_removed = 1; | | 1314 | mbuf_removed = 1; |
1310 | if (paddr != NULL) { | | 1315 | if (paddr != NULL) { |
1311 | *paddr = m; | | 1316 | *paddr = m; |
1312 | so->so_rcv.sb_mb = m->m_next; | | 1317 | so->so_rcv.sb_mb = m->m_next; |
1313 | m->m_next = NULL; | | 1318 | m->m_next = NULL; |
1314 | m = so->so_rcv.sb_mb; | | 1319 | m = so->so_rcv.sb_mb; |
1315 | } else { | | 1320 | } else { |
1316 | MFREE(m, so->so_rcv.sb_mb); | | 1321 | MFREE(m, so->so_rcv.sb_mb); |
1317 | m = so->so_rcv.sb_mb; | | 1322 | m = so->so_rcv.sb_mb; |
1318 | } | | 1323 | } |
1319 | sbsync(&so->so_rcv, nextrecord); | | 1324 | sbsync(&so->so_rcv, nextrecord); |
1320 | } | | 1325 | } |
1321 | } | | 1326 | } |
1322 | | | 1327 | |
1323 | /* | | 1328 | /* |
1324 | * Process one or more MT_CONTROL mbufs present before any data mbufs | | 1329 | * Process one or more MT_CONTROL mbufs present before any data mbufs |
1325 | * in the first mbuf chain on the socket buffer. If MSG_PEEK, we | | 1330 | * in the first mbuf chain on the socket buffer. If MSG_PEEK, we |
1326 | * just copy the data; if !MSG_PEEK, we call into the protocol to | | 1331 | * just copy the data; if !MSG_PEEK, we call into the protocol to |
1327 | * perform externalization (or freeing if controlp == NULL). | | 1332 | * perform externalization (or freeing if controlp == NULL). |
1328 | */ | | 1333 | */ |
1329 | if (__predict_false(m != NULL && m->m_type == MT_CONTROL)) { | | 1334 | if (__predict_false(m != NULL && m->m_type == MT_CONTROL)) { |
1330 | struct mbuf *cm = NULL, *cmn; | | 1335 | struct mbuf *cm = NULL, *cmn; |
1331 | struct mbuf **cme = &cm; | | 1336 | struct mbuf **cme = &cm; |
1332 | | | 1337 | |
1333 | do { | | 1338 | do { |
1334 | if (flags & MSG_PEEK) { | | 1339 | if (flags & MSG_PEEK) { |
1335 | if (controlp != NULL) { | | 1340 | if (controlp != NULL) { |
1336 | *controlp = m_copy(m, 0, m->m_len); | | 1341 | *controlp = m_copy(m, 0, m->m_len); |
1337 | controlp = &(*controlp)->m_next; | | 1342 | controlp = &(*controlp)->m_next; |
1338 | } | | 1343 | } |
1339 | m = m->m_next; | | 1344 | m = m->m_next; |
1340 | } else { | | 1345 | } else { |
1341 | sbfree(&so->so_rcv, m); | | 1346 | sbfree(&so->so_rcv, m); |
1342 | so->so_rcv.sb_mb = m->m_next; | | 1347 | so->so_rcv.sb_mb = m->m_next; |
1343 | m->m_next = NULL; | | 1348 | m->m_next = NULL; |
1344 | *cme = m; | | 1349 | *cme = m; |
1345 | cme = &(*cme)->m_next; | | 1350 | cme = &(*cme)->m_next; |
1346 | m = so->so_rcv.sb_mb; | | 1351 | m = so->so_rcv.sb_mb; |
1347 | } | | 1352 | } |
1348 | } while (m != NULL && m->m_type == MT_CONTROL); | | 1353 | } while (m != NULL && m->m_type == MT_CONTROL); |
1349 | if ((flags & MSG_PEEK) == 0) | | 1354 | if ((flags & MSG_PEEK) == 0) |
1350 | sbsync(&so->so_rcv, nextrecord); | | 1355 | sbsync(&so->so_rcv, nextrecord); |
1351 | for (; cm != NULL; cm = cmn) { | | 1356 | for (; cm != NULL; cm = cmn) { |
1352 | cmn = cm->m_next; | | 1357 | cmn = cm->m_next; |
1353 | cm->m_next = NULL; | | 1358 | cm->m_next = NULL; |
1354 | type = mtod(cm, struct cmsghdr *)->cmsg_type; | | 1359 | type = mtod(cm, struct cmsghdr *)->cmsg_type; |
1355 | if (controlp != NULL) { | | 1360 | if (controlp != NULL) { |
1356 | if (dom->dom_externalize != NULL && | | 1361 | if (dom->dom_externalize != NULL && |
1357 | type == SCM_RIGHTS) { | | 1362 | type == SCM_RIGHTS) { |
1358 | sounlock(so); | | 1363 | sounlock(so); |
1359 | splx(s); | | 1364 | splx(s); |
1360 | error = (*dom->dom_externalize)(cm, l); | | 1365 | error = (*dom->dom_externalize)(cm, l); |
1361 | s = splsoftnet(); | | 1366 | s = splsoftnet(); |
1362 | solock(so); | | 1367 | solock(so); |
1363 | } | | 1368 | } |
1364 | *controlp = cm; | | 1369 | *controlp = cm; |
1365 | while (*controlp != NULL) | | 1370 | while (*controlp != NULL) |
1366 | controlp = &(*controlp)->m_next; | | 1371 | controlp = &(*controlp)->m_next; |
1367 | } else { | | 1372 | } else { |
1368 | /* | | 1373 | /* |
1369 | * Dispose of any SCM_RIGHTS message that went | | 1374 | * Dispose of any SCM_RIGHTS message that went |
1370 | * through the read path rather than recv. | | 1375 | * through the read path rather than recv. |
1371 | */ | | 1376 | */ |
1372 | if (dom->dom_dispose != NULL && | | 1377 | if (dom->dom_dispose != NULL && |
1373 | type == SCM_RIGHTS) { | | 1378 | type == SCM_RIGHTS) { |
1374 | sounlock(so); | | 1379 | sounlock(so); |
1375 | (*dom->dom_dispose)(cm); | | 1380 | (*dom->dom_dispose)(cm); |
1376 | solock(so); | | 1381 | solock(so); |
1377 | } | | 1382 | } |
1378 | m_freem(cm); | | 1383 | m_freem(cm); |
1379 | } | | 1384 | } |
1380 | } | | 1385 | } |
1381 | if (m != NULL) | | 1386 | if (m != NULL) |
1382 | nextrecord = so->so_rcv.sb_mb->m_nextpkt; | | 1387 | nextrecord = so->so_rcv.sb_mb->m_nextpkt; |
1383 | else | | 1388 | else |
1384 | nextrecord = so->so_rcv.sb_mb; | | 1389 | nextrecord = so->so_rcv.sb_mb; |
1385 | orig_resid = 0; | | 1390 | orig_resid = 0; |
1386 | } | | 1391 | } |
1387 | | | 1392 | |
1388 | /* If m is non-NULL, we have some data to read. */ | | 1393 | /* If m is non-NULL, we have some data to read. */ |
1389 | if (__predict_true(m != NULL)) { | | 1394 | if (__predict_true(m != NULL)) { |
1390 | type = m->m_type; | | 1395 | type = m->m_type; |
1391 | if (type == MT_OOBDATA) | | 1396 | if (type == MT_OOBDATA) |
1392 | flags |= MSG_OOB; | | 1397 | flags |= MSG_OOB; |
1393 | } | | 1398 | } |
1394 | SBLASTRECORDCHK(&so->so_rcv, "soreceive 2"); | | 1399 | SBLASTRECORDCHK(&so->so_rcv, "soreceive 2"); |
1395 | SBLASTMBUFCHK(&so->so_rcv, "soreceive 2"); | | 1400 | SBLASTMBUFCHK(&so->so_rcv, "soreceive 2"); |
1396 | | | 1401 | |
1397 | moff = 0; | | 1402 | moff = 0; |
1398 | offset = 0; | | 1403 | offset = 0; |
1399 | while (m != NULL && uio->uio_resid > 0 && error == 0) { | | 1404 | while (m != NULL && uio->uio_resid > 0 && error == 0) { |
1400 | if (m->m_type == MT_OOBDATA) { | | 1405 | if (m->m_type == MT_OOBDATA) { |
1401 | if (type != MT_OOBDATA) | | 1406 | if (type != MT_OOBDATA) |
1402 | break; | | 1407 | break; |
1403 | } else if (type == MT_OOBDATA) | | 1408 | } else if (type == MT_OOBDATA) |
1404 | break; | | 1409 | break; |
1405 | #ifdef DIAGNOSTIC | | 1410 | #ifdef DIAGNOSTIC |
1406 | else if (m->m_type != MT_DATA && m->m_type != MT_HEADER) | | 1411 | else if (m->m_type != MT_DATA && m->m_type != MT_HEADER) |
1407 | panic("receive 3"); | | 1412 | panic("receive 3"); |
1408 | #endif | | 1413 | #endif |
1409 | so->so_state &= ~SS_RCVATMARK; | | 1414 | so->so_state &= ~SS_RCVATMARK; |
1410 | len = uio->uio_resid; | | 1415 | len = uio->uio_resid; |
1411 | if (so->so_oobmark && len > so->so_oobmark - offset) | | 1416 | if (so->so_oobmark && len > so->so_oobmark - offset) |
1412 | len = so->so_oobmark - offset; | | 1417 | len = so->so_oobmark - offset; |
1413 | if (len > m->m_len - moff) | | 1418 | if (len > m->m_len - moff) |
1414 | len = m->m_len - moff; | | 1419 | len = m->m_len - moff; |
1415 | /* | | 1420 | /* |
1416 | * If mp is set, just pass back the mbufs. | | 1421 | * If mp is set, just pass back the mbufs. |
1417 | * Otherwise copy them out via the uio, then free. | | 1422 | * Otherwise copy them out via the uio, then free. |
1418 | * Sockbuf must be consistent here (points to current mbuf, | | 1423 | * Sockbuf must be consistent here (points to current mbuf, |
1419 | * it points to next record) when we drop priority; | | 1424 | * it points to next record) when we drop priority; |
1420 | * we must note any additions to the sockbuf when we | | 1425 | * we must note any additions to the sockbuf when we |
1421 | * block interrupts again. | | 1426 | * block interrupts again. |
1422 | */ | | 1427 | */ |
1423 | if (mp == NULL) { | | 1428 | if (mp == NULL) { |
1424 | SBLASTRECORDCHK(&so->so_rcv, "soreceive uiomove"); | | 1429 | SBLASTRECORDCHK(&so->so_rcv, "soreceive uiomove"); |
1425 | SBLASTMBUFCHK(&so->so_rcv, "soreceive uiomove"); | | 1430 | SBLASTMBUFCHK(&so->so_rcv, "soreceive uiomove"); |
1426 | sounlock(so); | | 1431 | sounlock(so); |
1427 | splx(s); | | 1432 | splx(s); |
1428 | error = uiomove(mtod(m, char *) + moff, (int)len, uio); | | 1433 | error = uiomove(mtod(m, char *) + moff, (int)len, uio); |
1429 | s = splsoftnet(); | | 1434 | s = splsoftnet(); |
1430 | solock(so); | | 1435 | solock(so); |
1431 | if (error != 0) { | | 1436 | if (error != 0) { |
1432 | /* | | 1437 | /* |
1433 | * If any part of the record has been removed | | 1438 | * If any part of the record has been removed |
1434 | * (such as the MT_SONAME mbuf, which will | | 1439 | * (such as the MT_SONAME mbuf, which will |
1435 | * happen when PR_ADDR, and thus also | | 1440 | * happen when PR_ADDR, and thus also |
1436 | * PR_ATOMIC, is set), then drop the entire | | 1441 | * PR_ATOMIC, is set), then drop the entire |
1437 | * record to maintain the atomicity of the | | 1442 | * record to maintain the atomicity of the |
1438 | * receive operation. | | 1443 | * receive operation. |
1439 | * | | 1444 | * |
1440 | * This avoids a later panic("receive 1a") | | 1445 | * This avoids a later panic("receive 1a") |
1441 | * when compiled with DIAGNOSTIC. | | 1446 | * when compiled with DIAGNOSTIC. |
1442 | */ | | 1447 | */ |
1443 | if (m && mbuf_removed && atomic) | | 1448 | if (m && mbuf_removed && atomic) |
1444 | (void) sbdroprecord(&so->so_rcv); | | 1449 | (void) sbdroprecord(&so->so_rcv); |
1445 | | | 1450 | |
1446 | goto release; | | 1451 | goto release; |
1447 | } | | 1452 | } |
1448 | } else | | 1453 | } else |
1449 | uio->uio_resid -= len; | | 1454 | uio->uio_resid -= len; |
1450 | if (len == m->m_len - moff) { | | 1455 | if (len == m->m_len - moff) { |
1451 | if (m->m_flags & M_EOR) | | 1456 | if (m->m_flags & M_EOR) |
1452 | flags |= MSG_EOR; | | 1457 | flags |= MSG_EOR; |
1453 | if (flags & MSG_PEEK) { | | 1458 | if (flags & MSG_PEEK) { |
1454 | m = m->m_next; | | 1459 | m = m->m_next; |
1455 | moff = 0; | | 1460 | moff = 0; |
1456 | } else { | | 1461 | } else { |
1457 | nextrecord = m->m_nextpkt; | | 1462 | nextrecord = m->m_nextpkt; |
1458 | sbfree(&so->so_rcv, m); | | 1463 | sbfree(&so->so_rcv, m); |
1459 | if (mp) { | | 1464 | if (mp) { |
1460 | *mp = m; | | 1465 | *mp = m; |
1461 | mp = &m->m_next; | | 1466 | mp = &m->m_next; |
1462 | so->so_rcv.sb_mb = m = m->m_next; | | 1467 | so->so_rcv.sb_mb = m = m->m_next; |
1463 | *mp = NULL; | | 1468 | *mp = NULL; |
1464 | } else { | | 1469 | } else { |
1465 | MFREE(m, so->so_rcv.sb_mb); | | 1470 | MFREE(m, so->so_rcv.sb_mb); |
1466 | m = so->so_rcv.sb_mb; | | 1471 | m = so->so_rcv.sb_mb; |
1467 | } | | 1472 | } |
1468 | /* | | 1473 | /* |
1469 | * If m != NULL, we also know that | | 1474 | * If m != NULL, we also know that |
1470 | * so->so_rcv.sb_mb != NULL. | | 1475 | * so->so_rcv.sb_mb != NULL. |
1471 | */ | | 1476 | */ |