| @@ -1,1445 +1,1450 @@ | | | @@ -1,1445 +1,1450 @@ |
1 | /* $NetBSD: uipc_socket.c,v 1.192 2009/10/03 01:41:39 elad Exp $ */ | | 1 | /* $NetBSD: uipc_socket.c,v 1.193 2009/10/03 03:59:39 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 1.192 2009/10/03 01:41:39 elad Exp $"); | | 66 | __KERNEL_RCSID(0, "$NetBSD: uipc_socket.c,v 1.193 2009/10/03 03:59:39 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 | (action != KAUTH_NETWORK_BIND)) |
444 | return result; | | 445 | return result; |
445 | | | 446 | |
446 | switch (req) { | | 447 | switch (req) { |
| | | 448 | case KAUTH_REQ_NETWORK_BIND_PORT: |
| | | 449 | result = KAUTH_RESULT_ALLOW; |
| | | 450 | break; |
| | | 451 | |
447 | case KAUTH_REQ_NETWORK_SOCKET_DROP: { | | 452 | case KAUTH_REQ_NETWORK_SOCKET_DROP: { |
448 | /* Normal users can only drop their own connections. */ | | 453 | /* Normal users can only drop their own connections. */ |
449 | struct socket *so = (struct socket *)arg1; | | 454 | struct socket *so = (struct socket *)arg1; |
450 | uid_t sockuid = so->so_uidinfo->ui_uid; | | 455 | uid_t sockuid = so->so_uidinfo->ui_uid; |
451 | | | 456 | |
452 | if (sockuid == kauth_cred_getuid(cred) || | | 457 | if (sockuid == kauth_cred_getuid(cred) || |
453 | sockuid == kauth_cred_geteuid(cred)) | | 458 | sockuid == kauth_cred_geteuid(cred)) |
454 | result = KAUTH_RESULT_ALLOW; | | 459 | result = KAUTH_RESULT_ALLOW; |
455 | | | 460 | |
456 | break; | | 461 | break; |
457 | } | | 462 | } |
458 | | | 463 | |
459 | case KAUTH_REQ_NETWORK_SOCKET_OPEN: | | 464 | case KAUTH_REQ_NETWORK_SOCKET_OPEN: |
460 | /* We allow "raw" routing/bluetooth sockets to anyone. */ | | 465 | /* We allow "raw" routing/bluetooth sockets to anyone. */ |
461 | if ((u_long)arg1 == PF_ROUTE || (u_long)arg1 == PF_BLUETOOTH) | | 466 | if ((u_long)arg1 == PF_ROUTE || (u_long)arg1 == PF_BLUETOOTH) |
462 | result = KAUTH_RESULT_ALLOW; | | 467 | result = KAUTH_RESULT_ALLOW; |
463 | else { | | 468 | else { |
464 | /* Privileged, let secmodel handle this. */ | | 469 | /* Privileged, let secmodel handle this. */ |
465 | if ((u_long)arg2 == SOCK_RAW) | | 470 | if ((u_long)arg2 == SOCK_RAW) |
466 | break; | | 471 | break; |
467 | } | | 472 | } |
468 | | | 473 | |
469 | result = KAUTH_RESULT_ALLOW; | | 474 | result = KAUTH_RESULT_ALLOW; |
470 | | | 475 | |
471 | break; | | 476 | break; |
472 | | | 477 | |
473 | case KAUTH_REQ_NETWORK_SOCKET_CANSEE: | | 478 | case KAUTH_REQ_NETWORK_SOCKET_CANSEE: |
474 | result = KAUTH_RESULT_ALLOW; | | 479 | result = KAUTH_RESULT_ALLOW; |
475 | | | 480 | |
476 | break; | | 481 | break; |
477 | | | 482 | |
478 | default: | | 483 | default: |
479 | break; | | 484 | break; |
480 | } | | 485 | } |
481 | | | 486 | |
482 | return result; | | 487 | return result; |
483 | } | | 488 | } |
484 | | | 489 | |
485 | void | | 490 | void |
486 | soinit(void) | | 491 | soinit(void) |
487 | { | | 492 | { |
488 | | | 493 | |
489 | sysctl_kern_somaxkva_setup(); | | 494 | sysctl_kern_somaxkva_setup(); |
490 | | | 495 | |
491 | mutex_init(&so_pendfree_lock, MUTEX_DEFAULT, IPL_VM); | | 496 | mutex_init(&so_pendfree_lock, MUTEX_DEFAULT, IPL_VM); |
492 | softnet_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE); | | 497 | softnet_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE); |
493 | cv_init(&socurkva_cv, "sokva"); | | 498 | cv_init(&socurkva_cv, "sokva"); |
494 | soinit2(); | | 499 | soinit2(); |
495 | | | 500 | |
496 | /* Set the initial adjusted socket buffer size. */ | | 501 | /* Set the initial adjusted socket buffer size. */ |
497 | if (sb_max_set(sb_max)) | | 502 | if (sb_max_set(sb_max)) |
498 | panic("bad initial sb_max value: %lu", sb_max); | | 503 | panic("bad initial sb_max value: %lu", sb_max); |
499 | | | 504 | |
500 | callback_register(&vm_map_to_kernel(kernel_map)->vmk_reclaim_callback, | | 505 | callback_register(&vm_map_to_kernel(kernel_map)->vmk_reclaim_callback, |
501 | &sokva_reclaimerentry, NULL, sokva_reclaim_callback); | | 506 | &sokva_reclaimerentry, NULL, sokva_reclaim_callback); |
502 | | | 507 | |
503 | socket_listener = kauth_listen_scope(KAUTH_SCOPE_NETWORK, | | 508 | socket_listener = kauth_listen_scope(KAUTH_SCOPE_NETWORK, |
504 | socket_listener_cb, NULL); | | 509 | socket_listener_cb, NULL); |
505 | } | | 510 | } |
506 | | | 511 | |
507 | /* | | 512 | /* |
508 | * Socket operation routines. | | 513 | * Socket operation routines. |
509 | * These routines are called by the routines in | | 514 | * These routines are called by the routines in |
510 | * sys_socket.c or from a system process, and | | 515 | * sys_socket.c or from a system process, and |
511 | * implement the semantics of socket operations by | | 516 | * implement the semantics of socket operations by |
512 | * switching out to the protocol specific routines. | | 517 | * switching out to the protocol specific routines. |
513 | */ | | 518 | */ |
514 | /*ARGSUSED*/ | | 519 | /*ARGSUSED*/ |
515 | int | | 520 | int |
516 | socreate(int dom, struct socket **aso, int type, int proto, struct lwp *l, | | 521 | socreate(int dom, struct socket **aso, int type, int proto, struct lwp *l, |
517 | struct socket *lockso) | | 522 | struct socket *lockso) |
518 | { | | 523 | { |
519 | const struct protosw *prp; | | 524 | const struct protosw *prp; |
520 | struct socket *so; | | 525 | struct socket *so; |
521 | uid_t uid; | | 526 | uid_t uid; |
522 | int error; | | 527 | int error; |
523 | kmutex_t *lock; | | 528 | kmutex_t *lock; |
524 | | | 529 | |
525 | error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_SOCKET, | | 530 | error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_SOCKET, |
526 | KAUTH_REQ_NETWORK_SOCKET_OPEN, KAUTH_ARG(dom), KAUTH_ARG(type), | | 531 | KAUTH_REQ_NETWORK_SOCKET_OPEN, KAUTH_ARG(dom), KAUTH_ARG(type), |
527 | KAUTH_ARG(proto)); | | 532 | KAUTH_ARG(proto)); |
528 | if (error != 0) | | 533 | if (error != 0) |
529 | return error; | | 534 | return error; |
530 | | | 535 | |
531 | if (proto) | | 536 | if (proto) |
532 | prp = pffindproto(dom, proto, type); | | 537 | prp = pffindproto(dom, proto, type); |
533 | else | | 538 | else |
534 | prp = pffindtype(dom, type); | | 539 | prp = pffindtype(dom, type); |
535 | if (prp == NULL) { | | 540 | if (prp == NULL) { |
536 | /* no support for domain */ | | 541 | /* no support for domain */ |
537 | if (pffinddomain(dom) == 0) | | 542 | if (pffinddomain(dom) == 0) |
538 | return EAFNOSUPPORT; | | 543 | return EAFNOSUPPORT; |
539 | /* no support for socket type */ | | 544 | /* no support for socket type */ |
540 | if (proto == 0 && type != 0) | | 545 | if (proto == 0 && type != 0) |
541 | return EPROTOTYPE; | | 546 | return EPROTOTYPE; |
542 | return EPROTONOSUPPORT; | | 547 | return EPROTONOSUPPORT; |
543 | } | | 548 | } |
544 | if (prp->pr_usrreq == NULL) | | 549 | if (prp->pr_usrreq == NULL) |
545 | return EPROTONOSUPPORT; | | 550 | return EPROTONOSUPPORT; |
546 | if (prp->pr_type != type) | | 551 | if (prp->pr_type != type) |
547 | return EPROTOTYPE; | | 552 | return EPROTOTYPE; |
548 | | | 553 | |
549 | so = soget(true); | | 554 | so = soget(true); |
550 | so->so_type = type; | | 555 | so->so_type = type; |
551 | so->so_proto = prp; | | 556 | so->so_proto = prp; |
552 | so->so_send = sosend; | | 557 | so->so_send = sosend; |
553 | so->so_receive = soreceive; | | 558 | so->so_receive = soreceive; |
554 | #ifdef MBUFTRACE | | 559 | #ifdef MBUFTRACE |
555 | so->so_rcv.sb_mowner = &prp->pr_domain->dom_mowner; | | 560 | so->so_rcv.sb_mowner = &prp->pr_domain->dom_mowner; |
556 | so->so_snd.sb_mowner = &prp->pr_domain->dom_mowner; | | 561 | so->so_snd.sb_mowner = &prp->pr_domain->dom_mowner; |
557 | so->so_mowner = &prp->pr_domain->dom_mowner; | | 562 | so->so_mowner = &prp->pr_domain->dom_mowner; |
558 | #endif | | 563 | #endif |
559 | /* so->so_cred = kauth_cred_dup(l->l_cred); */ | | 564 | /* so->so_cred = kauth_cred_dup(l->l_cred); */ |
560 | uid = kauth_cred_geteuid(l->l_cred); | | 565 | uid = kauth_cred_geteuid(l->l_cred); |
561 | so->so_uidinfo = uid_find(uid); | | 566 | so->so_uidinfo = uid_find(uid); |
562 | so->so_egid = kauth_cred_getegid(l->l_cred); | | 567 | so->so_egid = kauth_cred_getegid(l->l_cred); |
563 | so->so_cpid = l->l_proc->p_pid; | | 568 | so->so_cpid = l->l_proc->p_pid; |
564 | if (lockso != NULL) { | | 569 | if (lockso != NULL) { |
565 | /* Caller wants us to share a lock. */ | | 570 | /* Caller wants us to share a lock. */ |
566 | lock = lockso->so_lock; | | 571 | lock = lockso->so_lock; |
567 | so->so_lock = lock; | | 572 | so->so_lock = lock; |
568 | mutex_obj_hold(lock); | | 573 | mutex_obj_hold(lock); |
569 | mutex_enter(lock); | | 574 | mutex_enter(lock); |
570 | } else { | | 575 | } else { |
571 | /* Lock assigned and taken during PRU_ATTACH. */ | | 576 | /* Lock assigned and taken during PRU_ATTACH. */ |
572 | } | | 577 | } |
573 | error = (*prp->pr_usrreq)(so, PRU_ATTACH, NULL, | | 578 | error = (*prp->pr_usrreq)(so, PRU_ATTACH, NULL, |
574 | (struct mbuf *)(long)proto, NULL, l); | | 579 | (struct mbuf *)(long)proto, NULL, l); |
575 | KASSERT(solocked(so)); | | 580 | KASSERT(solocked(so)); |
576 | if (error != 0) { | | 581 | if (error != 0) { |
577 | so->so_state |= SS_NOFDREF; | | 582 | so->so_state |= SS_NOFDREF; |
578 | sofree(so); | | 583 | sofree(so); |
579 | return error; | | 584 | return error; |
580 | } | | 585 | } |
581 | sounlock(so); | | 586 | sounlock(so); |
582 | *aso = so; | | 587 | *aso = so; |
583 | return 0; | | 588 | return 0; |
584 | } | | 589 | } |
585 | | | 590 | |
586 | /* On success, write file descriptor to fdout and return zero. On | | 591 | /* On success, write file descriptor to fdout and return zero. On |
587 | * failure, return non-zero; *fdout will be undefined. | | 592 | * failure, return non-zero; *fdout will be undefined. |
588 | */ | | 593 | */ |
589 | int | | 594 | int |
590 | fsocreate(int domain, struct socket **sop, int type, int protocol, | | 595 | fsocreate(int domain, struct socket **sop, int type, int protocol, |
591 | struct lwp *l, int *fdout) | | 596 | struct lwp *l, int *fdout) |
592 | { | | 597 | { |
593 | struct socket *so; | | 598 | struct socket *so; |
594 | struct file *fp; | | 599 | struct file *fp; |
595 | int fd, error; | | 600 | int fd, error; |
596 | | | 601 | |
597 | if ((error = fd_allocfile(&fp, &fd)) != 0) | | 602 | if ((error = fd_allocfile(&fp, &fd)) != 0) |
598 | return (error); | | 603 | return (error); |
599 | fp->f_flag = FREAD|FWRITE; | | 604 | fp->f_flag = FREAD|FWRITE; |
600 | fp->f_type = DTYPE_SOCKET; | | 605 | fp->f_type = DTYPE_SOCKET; |
601 | fp->f_ops = &socketops; | | 606 | fp->f_ops = &socketops; |
602 | error = socreate(domain, &so, type, protocol, l, NULL); | | 607 | error = socreate(domain, &so, type, protocol, l, NULL); |
603 | if (error != 0) { | | 608 | if (error != 0) { |
604 | fd_abort(curproc, fp, fd); | | 609 | fd_abort(curproc, fp, fd); |
605 | } else { | | 610 | } else { |
606 | if (sop != NULL) | | 611 | if (sop != NULL) |
607 | *sop = so; | | 612 | *sop = so; |
608 | fp->f_data = so; | | 613 | fp->f_data = so; |
609 | fd_affix(curproc, fp, fd); | | 614 | fd_affix(curproc, fp, fd); |
610 | *fdout = fd; | | 615 | *fdout = fd; |
611 | } | | 616 | } |
612 | return error; | | 617 | return error; |
613 | } | | 618 | } |
614 | | | 619 | |
615 | int | | 620 | int |
616 | sofamily(const struct socket *so) | | 621 | sofamily(const struct socket *so) |
617 | { | | 622 | { |
618 | const struct protosw *pr; | | 623 | const struct protosw *pr; |
619 | const struct domain *dom; | | 624 | const struct domain *dom; |
620 | | | 625 | |
621 | if ((pr = so->so_proto) == NULL) | | 626 | if ((pr = so->so_proto) == NULL) |
622 | return AF_UNSPEC; | | 627 | return AF_UNSPEC; |
623 | if ((dom = pr->pr_domain) == NULL) | | 628 | if ((dom = pr->pr_domain) == NULL) |
624 | return AF_UNSPEC; | | 629 | return AF_UNSPEC; |
625 | return dom->dom_family; | | 630 | return dom->dom_family; |
626 | } | | 631 | } |
627 | | | 632 | |
628 | int | | 633 | int |
629 | sobind(struct socket *so, struct mbuf *nam, struct lwp *l) | | 634 | sobind(struct socket *so, struct mbuf *nam, struct lwp *l) |
630 | { | | 635 | { |
631 | int error; | | 636 | int error; |
632 | | | 637 | |
633 | solock(so); | | 638 | solock(so); |
634 | error = (*so->so_proto->pr_usrreq)(so, PRU_BIND, NULL, nam, NULL, l); | | 639 | error = (*so->so_proto->pr_usrreq)(so, PRU_BIND, NULL, nam, NULL, l); |
635 | sounlock(so); | | 640 | sounlock(so); |
636 | return error; | | 641 | return error; |
637 | } | | 642 | } |
638 | | | 643 | |
639 | int | | 644 | int |
640 | solisten(struct socket *so, int backlog, struct lwp *l) | | 645 | solisten(struct socket *so, int backlog, struct lwp *l) |
641 | { | | 646 | { |
642 | int error; | | 647 | int error; |
643 | | | 648 | |
644 | solock(so); | | 649 | solock(so); |
645 | if ((so->so_state & (SS_ISCONNECTED | SS_ISCONNECTING | | | 650 | if ((so->so_state & (SS_ISCONNECTED | SS_ISCONNECTING | |
646 | SS_ISDISCONNECTING)) != 0) { | | 651 | SS_ISDISCONNECTING)) != 0) { |
647 | sounlock(so); | | 652 | sounlock(so); |
648 | return (EOPNOTSUPP); | | 653 | return (EOPNOTSUPP); |
649 | } | | 654 | } |
650 | error = (*so->so_proto->pr_usrreq)(so, PRU_LISTEN, NULL, | | 655 | error = (*so->so_proto->pr_usrreq)(so, PRU_LISTEN, NULL, |
651 | NULL, NULL, l); | | 656 | NULL, NULL, l); |
652 | if (error != 0) { | | 657 | if (error != 0) { |
653 | sounlock(so); | | 658 | sounlock(so); |
654 | return error; | | 659 | return error; |
655 | } | | 660 | } |
656 | if (TAILQ_EMPTY(&so->so_q)) | | 661 | if (TAILQ_EMPTY(&so->so_q)) |
657 | so->so_options |= SO_ACCEPTCONN; | | 662 | so->so_options |= SO_ACCEPTCONN; |
658 | if (backlog < 0) | | 663 | if (backlog < 0) |
659 | backlog = 0; | | 664 | backlog = 0; |
660 | so->so_qlimit = min(backlog, somaxconn); | | 665 | so->so_qlimit = min(backlog, somaxconn); |
661 | sounlock(so); | | 666 | sounlock(so); |
662 | return 0; | | 667 | return 0; |
663 | } | | 668 | } |
664 | | | 669 | |
665 | void | | 670 | void |
666 | sofree(struct socket *so) | | 671 | sofree(struct socket *so) |
667 | { | | 672 | { |
668 | u_int refs; | | 673 | u_int refs; |
669 | | | 674 | |
670 | KASSERT(solocked(so)); | | 675 | KASSERT(solocked(so)); |
671 | | | 676 | |
672 | if (so->so_pcb || (so->so_state & SS_NOFDREF) == 0) { | | 677 | if (so->so_pcb || (so->so_state & SS_NOFDREF) == 0) { |
673 | sounlock(so); | | 678 | sounlock(so); |
674 | return; | | 679 | return; |
675 | } | | 680 | } |
676 | if (so->so_head) { | | 681 | if (so->so_head) { |
677 | /* | | 682 | /* |
678 | * We must not decommission a socket that's on the accept(2) | | 683 | * We must not decommission a socket that's on the accept(2) |
679 | * queue. If we do, then accept(2) may hang after select(2) | | 684 | * queue. If we do, then accept(2) may hang after select(2) |
680 | * indicated that the listening socket was ready. | | 685 | * indicated that the listening socket was ready. |
681 | */ | | 686 | */ |
682 | if (!soqremque(so, 0)) { | | 687 | if (!soqremque(so, 0)) { |
683 | sounlock(so); | | 688 | sounlock(so); |
684 | return; | | 689 | return; |
685 | } | | 690 | } |
686 | } | | 691 | } |
687 | if (so->so_rcv.sb_hiwat) | | 692 | if (so->so_rcv.sb_hiwat) |
688 | (void)chgsbsize(so->so_uidinfo, &so->so_rcv.sb_hiwat, 0, | | 693 | (void)chgsbsize(so->so_uidinfo, &so->so_rcv.sb_hiwat, 0, |
689 | RLIM_INFINITY); | | 694 | RLIM_INFINITY); |
690 | if (so->so_snd.sb_hiwat) | | 695 | if (so->so_snd.sb_hiwat) |
691 | (void)chgsbsize(so->so_uidinfo, &so->so_snd.sb_hiwat, 0, | | 696 | (void)chgsbsize(so->so_uidinfo, &so->so_snd.sb_hiwat, 0, |
692 | RLIM_INFINITY); | | 697 | RLIM_INFINITY); |
693 | sbrelease(&so->so_snd, so); | | 698 | sbrelease(&so->so_snd, so); |
694 | KASSERT(!cv_has_waiters(&so->so_cv)); | | 699 | KASSERT(!cv_has_waiters(&so->so_cv)); |
695 | KASSERT(!cv_has_waiters(&so->so_rcv.sb_cv)); | | 700 | KASSERT(!cv_has_waiters(&so->so_rcv.sb_cv)); |
696 | KASSERT(!cv_has_waiters(&so->so_snd.sb_cv)); | | 701 | KASSERT(!cv_has_waiters(&so->so_snd.sb_cv)); |
697 | sorflush(so); | | 702 | sorflush(so); |
698 | refs = so->so_aborting; /* XXX */ | | 703 | refs = so->so_aborting; /* XXX */ |
699 | /* Remove acccept filter if one is present. */ | | 704 | /* Remove acccept filter if one is present. */ |
700 | if (so->so_accf != NULL) | | 705 | if (so->so_accf != NULL) |
701 | (void)accept_filt_clear(so); | | 706 | (void)accept_filt_clear(so); |
702 | /* kauth_cred_free(so->so_cred); */ | | 707 | /* kauth_cred_free(so->so_cred); */ |
703 | sounlock(so); | | 708 | sounlock(so); |
704 | if (refs == 0) /* XXX */ | | 709 | if (refs == 0) /* XXX */ |
705 | soput(so); | | 710 | soput(so); |
706 | } | | 711 | } |
707 | | | 712 | |
708 | /* | | 713 | /* |
709 | * Close a socket on last file table reference removal. | | 714 | * Close a socket on last file table reference removal. |
710 | * Initiate disconnect if connected. | | 715 | * Initiate disconnect if connected. |
711 | * Free socket when disconnect complete. | | 716 | * Free socket when disconnect complete. |
712 | */ | | 717 | */ |
713 | int | | 718 | int |
714 | soclose(struct socket *so) | | 719 | soclose(struct socket *so) |
715 | { | | 720 | { |
716 | struct socket *so2; | | 721 | struct socket *so2; |
717 | int error; | | 722 | int error; |
718 | int error2; | | 723 | int error2; |
719 | | | 724 | |
720 | error = 0; | | 725 | error = 0; |
721 | solock(so); | | 726 | solock(so); |
722 | if (so->so_options & SO_ACCEPTCONN) { | | 727 | if (so->so_options & SO_ACCEPTCONN) { |
723 | for (;;) { | | 728 | for (;;) { |
724 | if ((so2 = TAILQ_FIRST(&so->so_q0)) != 0) { | | 729 | if ((so2 = TAILQ_FIRST(&so->so_q0)) != 0) { |
725 | KASSERT(solocked2(so, so2)); | | 730 | KASSERT(solocked2(so, so2)); |
726 | (void) soqremque(so2, 0); | | 731 | (void) soqremque(so2, 0); |
727 | /* soabort drops the lock. */ | | 732 | /* soabort drops the lock. */ |
728 | (void) soabort(so2); | | 733 | (void) soabort(so2); |
729 | solock(so); | | 734 | solock(so); |
730 | continue; | | 735 | continue; |
731 | } | | 736 | } |
732 | if ((so2 = TAILQ_FIRST(&so->so_q)) != 0) { | | 737 | if ((so2 = TAILQ_FIRST(&so->so_q)) != 0) { |
733 | KASSERT(solocked2(so, so2)); | | 738 | KASSERT(solocked2(so, so2)); |
734 | (void) soqremque(so2, 1); | | 739 | (void) soqremque(so2, 1); |
735 | /* soabort drops the lock. */ | | 740 | /* soabort drops the lock. */ |
736 | (void) soabort(so2); | | 741 | (void) soabort(so2); |
737 | solock(so); | | 742 | solock(so); |
738 | continue; | | 743 | continue; |
739 | } | | 744 | } |
740 | break; | | 745 | break; |
741 | } | | 746 | } |
742 | } | | 747 | } |
743 | if (so->so_pcb == 0) | | 748 | if (so->so_pcb == 0) |
744 | goto discard; | | 749 | goto discard; |
745 | if (so->so_state & SS_ISCONNECTED) { | | 750 | if (so->so_state & SS_ISCONNECTED) { |
746 | if ((so->so_state & SS_ISDISCONNECTING) == 0) { | | 751 | if ((so->so_state & SS_ISDISCONNECTING) == 0) { |
747 | error = sodisconnect(so); | | 752 | error = sodisconnect(so); |
748 | if (error) | | 753 | if (error) |
749 | goto drop; | | 754 | goto drop; |
750 | } | | 755 | } |
751 | if (so->so_options & SO_LINGER) { | | 756 | if (so->so_options & SO_LINGER) { |
752 | if ((so->so_state & SS_ISDISCONNECTING) && so->so_nbio) | | 757 | if ((so->so_state & SS_ISDISCONNECTING) && so->so_nbio) |
753 | goto drop; | | 758 | goto drop; |
754 | while (so->so_state & SS_ISCONNECTED) { | | 759 | while (so->so_state & SS_ISCONNECTED) { |
755 | error = sowait(so, true, so->so_linger * hz); | | 760 | error = sowait(so, true, so->so_linger * hz); |
756 | if (error) | | 761 | if (error) |
757 | break; | | 762 | break; |
758 | } | | 763 | } |
759 | } | | 764 | } |
760 | } | | 765 | } |
761 | drop: | | 766 | drop: |
762 | if (so->so_pcb) { | | 767 | if (so->so_pcb) { |
763 | error2 = (*so->so_proto->pr_usrreq)(so, PRU_DETACH, | | 768 | error2 = (*so->so_proto->pr_usrreq)(so, PRU_DETACH, |
764 | NULL, NULL, NULL, NULL); | | 769 | NULL, NULL, NULL, NULL); |
765 | if (error == 0) | | 770 | if (error == 0) |
766 | error = error2; | | 771 | error = error2; |
767 | } | | 772 | } |
768 | discard: | | 773 | discard: |
769 | if (so->so_state & SS_NOFDREF) | | 774 | if (so->so_state & SS_NOFDREF) |
770 | panic("soclose: NOFDREF"); | | 775 | panic("soclose: NOFDREF"); |
771 | so->so_state |= SS_NOFDREF; | | 776 | so->so_state |= SS_NOFDREF; |
772 | sofree(so); | | 777 | sofree(so); |
773 | return (error); | | 778 | return (error); |
774 | } | | 779 | } |
775 | | | 780 | |
776 | /* | | 781 | /* |
777 | * Must be called with the socket locked.. Will return with it unlocked. | | 782 | * Must be called with the socket locked.. Will return with it unlocked. |
778 | */ | | 783 | */ |
779 | int | | 784 | int |
780 | soabort(struct socket *so) | | 785 | soabort(struct socket *so) |
781 | { | | 786 | { |
782 | u_int refs; | | 787 | u_int refs; |
783 | int error; | | 788 | int error; |
784 | | | 789 | |
785 | KASSERT(solocked(so)); | | 790 | KASSERT(solocked(so)); |
786 | KASSERT(so->so_head == NULL); | | 791 | KASSERT(so->so_head == NULL); |
787 | | | 792 | |
788 | so->so_aborting++; /* XXX */ | | 793 | so->so_aborting++; /* XXX */ |
789 | error = (*so->so_proto->pr_usrreq)(so, PRU_ABORT, NULL, | | 794 | error = (*so->so_proto->pr_usrreq)(so, PRU_ABORT, NULL, |
790 | NULL, NULL, NULL); | | 795 | NULL, NULL, NULL); |
791 | refs = --so->so_aborting; /* XXX */ | | 796 | refs = --so->so_aborting; /* XXX */ |
792 | if (error || (refs == 0)) { | | 797 | if (error || (refs == 0)) { |
793 | sofree(so); | | 798 | sofree(so); |
794 | } else { | | 799 | } else { |
795 | sounlock(so); | | 800 | sounlock(so); |
796 | } | | 801 | } |
797 | return error; | | 802 | return error; |
798 | } | | 803 | } |
799 | | | 804 | |
800 | int | | 805 | int |
801 | soaccept(struct socket *so, struct mbuf *nam) | | 806 | soaccept(struct socket *so, struct mbuf *nam) |
802 | { | | 807 | { |
803 | int error; | | 808 | int error; |
804 | | | 809 | |
805 | KASSERT(solocked(so)); | | 810 | KASSERT(solocked(so)); |
806 | | | 811 | |
807 | error = 0; | | 812 | error = 0; |
808 | if ((so->so_state & SS_NOFDREF) == 0) | | 813 | if ((so->so_state & SS_NOFDREF) == 0) |
809 | panic("soaccept: !NOFDREF"); | | 814 | panic("soaccept: !NOFDREF"); |
810 | so->so_state &= ~SS_NOFDREF; | | 815 | so->so_state &= ~SS_NOFDREF; |
811 | if ((so->so_state & SS_ISDISCONNECTED) == 0 || | | 816 | if ((so->so_state & SS_ISDISCONNECTED) == 0 || |
812 | (so->so_proto->pr_flags & PR_ABRTACPTDIS) == 0) | | 817 | (so->so_proto->pr_flags & PR_ABRTACPTDIS) == 0) |
813 | error = (*so->so_proto->pr_usrreq)(so, PRU_ACCEPT, | | 818 | error = (*so->so_proto->pr_usrreq)(so, PRU_ACCEPT, |
814 | NULL, nam, NULL, NULL); | | 819 | NULL, nam, NULL, NULL); |
815 | else | | 820 | else |
816 | error = ECONNABORTED; | | 821 | error = ECONNABORTED; |
817 | | | 822 | |
818 | return (error); | | 823 | return (error); |
819 | } | | 824 | } |
820 | | | 825 | |
821 | int | | 826 | int |
822 | soconnect(struct socket *so, struct mbuf *nam, struct lwp *l) | | 827 | soconnect(struct socket *so, struct mbuf *nam, struct lwp *l) |
823 | { | | 828 | { |
824 | int error; | | 829 | int error; |
825 | | | 830 | |
826 | KASSERT(solocked(so)); | | 831 | KASSERT(solocked(so)); |
827 | | | 832 | |
828 | if (so->so_options & SO_ACCEPTCONN) | | 833 | if (so->so_options & SO_ACCEPTCONN) |
829 | return (EOPNOTSUPP); | | 834 | return (EOPNOTSUPP); |
830 | /* | | 835 | /* |
831 | * If protocol is connection-based, can only connect once. | | 836 | * If protocol is connection-based, can only connect once. |
832 | * Otherwise, if connected, try to disconnect first. | | 837 | * Otherwise, if connected, try to disconnect first. |
833 | * This allows user to disconnect by connecting to, e.g., | | 838 | * This allows user to disconnect by connecting to, e.g., |
834 | * a null address. | | 839 | * a null address. |
835 | */ | | 840 | */ |
836 | if (so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING) && | | 841 | if (so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING) && |
837 | ((so->so_proto->pr_flags & PR_CONNREQUIRED) || | | 842 | ((so->so_proto->pr_flags & PR_CONNREQUIRED) || |
838 | (error = sodisconnect(so)))) | | 843 | (error = sodisconnect(so)))) |
839 | error = EISCONN; | | 844 | error = EISCONN; |
840 | else | | 845 | else |
841 | error = (*so->so_proto->pr_usrreq)(so, PRU_CONNECT, | | 846 | error = (*so->so_proto->pr_usrreq)(so, PRU_CONNECT, |
842 | NULL, nam, NULL, l); | | 847 | NULL, nam, NULL, l); |
843 | return (error); | | 848 | return (error); |
844 | } | | 849 | } |
845 | | | 850 | |
846 | int | | 851 | int |
847 | soconnect2(struct socket *so1, struct socket *so2) | | 852 | soconnect2(struct socket *so1, struct socket *so2) |
848 | { | | 853 | { |
849 | int error; | | 854 | int error; |
850 | | | 855 | |
851 | KASSERT(solocked2(so1, so2)); | | 856 | KASSERT(solocked2(so1, so2)); |
852 | | | 857 | |
853 | error = (*so1->so_proto->pr_usrreq)(so1, PRU_CONNECT2, | | 858 | error = (*so1->so_proto->pr_usrreq)(so1, PRU_CONNECT2, |
854 | NULL, (struct mbuf *)so2, NULL, NULL); | | 859 | NULL, (struct mbuf *)so2, NULL, NULL); |
855 | return (error); | | 860 | return (error); |
856 | } | | 861 | } |
857 | | | 862 | |
858 | int | | 863 | int |
859 | sodisconnect(struct socket *so) | | 864 | sodisconnect(struct socket *so) |
860 | { | | 865 | { |
861 | int error; | | 866 | int error; |
862 | | | 867 | |
863 | KASSERT(solocked(so)); | | 868 | KASSERT(solocked(so)); |
864 | | | 869 | |
865 | if ((so->so_state & SS_ISCONNECTED) == 0) { | | 870 | if ((so->so_state & SS_ISCONNECTED) == 0) { |
866 | error = ENOTCONN; | | 871 | error = ENOTCONN; |
867 | } else if (so->so_state & SS_ISDISCONNECTING) { | | 872 | } else if (so->so_state & SS_ISDISCONNECTING) { |
868 | error = EALREADY; | | 873 | error = EALREADY; |
869 | } else { | | 874 | } else { |
870 | error = (*so->so_proto->pr_usrreq)(so, PRU_DISCONNECT, | | 875 | error = (*so->so_proto->pr_usrreq)(so, PRU_DISCONNECT, |
871 | NULL, NULL, NULL, NULL); | | 876 | NULL, NULL, NULL, NULL); |
872 | } | | 877 | } |
873 | sodopendfree(); | | 878 | sodopendfree(); |
874 | return (error); | | 879 | return (error); |
875 | } | | 880 | } |
876 | | | 881 | |
877 | #define SBLOCKWAIT(f) (((f) & MSG_DONTWAIT) ? M_NOWAIT : M_WAITOK) | | 882 | #define SBLOCKWAIT(f) (((f) & MSG_DONTWAIT) ? M_NOWAIT : M_WAITOK) |
878 | /* | | 883 | /* |
879 | * Send on a socket. | | 884 | * Send on a socket. |
880 | * If send must go all at once and message is larger than | | 885 | * If send must go all at once and message is larger than |
881 | * send buffering, then hard error. | | 886 | * send buffering, then hard error. |
882 | * Lock against other senders. | | 887 | * Lock against other senders. |
883 | * If must go all at once and not enough room now, then | | 888 | * If must go all at once and not enough room now, then |
884 | * inform user that this would block and do nothing. | | 889 | * inform user that this would block and do nothing. |
885 | * Otherwise, if nonblocking, send as much as possible. | | 890 | * Otherwise, if nonblocking, send as much as possible. |
886 | * The data to be sent is described by "uio" if nonzero, | | 891 | * The data to be sent is described by "uio" if nonzero, |
887 | * otherwise by the mbuf chain "top" (which must be null | | 892 | * otherwise by the mbuf chain "top" (which must be null |
888 | * if uio is not). Data provided in mbuf chain must be small | | 893 | * if uio is not). Data provided in mbuf chain must be small |
889 | * enough to send all at once. | | 894 | * enough to send all at once. |
890 | * | | 895 | * |
891 | * Returns nonzero on error, timeout or signal; callers | | 896 | * Returns nonzero on error, timeout or signal; callers |
892 | * must check for short counts if EINTR/ERESTART are returned. | | 897 | * must check for short counts if EINTR/ERESTART are returned. |
893 | * Data and control buffers are freed on return. | | 898 | * Data and control buffers are freed on return. |
894 | */ | | 899 | */ |
895 | int | | 900 | int |
896 | sosend(struct socket *so, struct mbuf *addr, struct uio *uio, struct mbuf *top, | | 901 | sosend(struct socket *so, struct mbuf *addr, struct uio *uio, struct mbuf *top, |
897 | struct mbuf *control, int flags, struct lwp *l) | | 902 | struct mbuf *control, int flags, struct lwp *l) |
898 | { | | 903 | { |
899 | struct mbuf **mp, *m; | | 904 | struct mbuf **mp, *m; |
900 | struct proc *p; | | 905 | struct proc *p; |
901 | long space, len, resid, clen, mlen; | | 906 | long space, len, resid, clen, mlen; |
902 | int error, s, dontroute, atomic; | | 907 | int error, s, dontroute, atomic; |
903 | | | 908 | |
904 | p = l->l_proc; | | 909 | p = l->l_proc; |
905 | sodopendfree(); | | 910 | sodopendfree(); |
906 | clen = 0; | | 911 | clen = 0; |
907 | | | 912 | |
908 | /* | | 913 | /* |
909 | * solock() provides atomicity of access. splsoftnet() prevents | | 914 | * solock() provides atomicity of access. splsoftnet() prevents |
910 | * protocol processing soft interrupts from interrupting us and | | 915 | * protocol processing soft interrupts from interrupting us and |
911 | * blocking (expensive). | | 916 | * blocking (expensive). |
912 | */ | | 917 | */ |
913 | s = splsoftnet(); | | 918 | s = splsoftnet(); |
914 | solock(so); | | 919 | solock(so); |
915 | atomic = sosendallatonce(so) || top; | | 920 | atomic = sosendallatonce(so) || top; |
916 | if (uio) | | 921 | if (uio) |
917 | resid = uio->uio_resid; | | 922 | resid = uio->uio_resid; |
918 | else | | 923 | else |
919 | resid = top->m_pkthdr.len; | | 924 | resid = top->m_pkthdr.len; |
920 | /* | | 925 | /* |
921 | * In theory resid should be unsigned. | | 926 | * In theory resid should be unsigned. |
922 | * However, space must be signed, as it might be less than 0 | | 927 | * However, space must be signed, as it might be less than 0 |
923 | * if we over-committed, and we must use a signed comparison | | 928 | * if we over-committed, and we must use a signed comparison |
924 | * of space and resid. On the other hand, a negative resid | | 929 | * of space and resid. On the other hand, a negative resid |
925 | * causes us to loop sending 0-length segments to the protocol. | | 930 | * causes us to loop sending 0-length segments to the protocol. |
926 | */ | | 931 | */ |
927 | if (resid < 0) { | | 932 | if (resid < 0) { |
928 | error = EINVAL; | | 933 | error = EINVAL; |
929 | goto out; | | 934 | goto out; |
930 | } | | 935 | } |
931 | dontroute = | | 936 | dontroute = |
932 | (flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 && | | 937 | (flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 && |
933 | (so->so_proto->pr_flags & PR_ATOMIC); | | 938 | (so->so_proto->pr_flags & PR_ATOMIC); |
934 | l->l_ru.ru_msgsnd++; | | 939 | l->l_ru.ru_msgsnd++; |
935 | if (control) | | 940 | if (control) |
936 | clen = control->m_len; | | 941 | clen = control->m_len; |
937 | restart: | | 942 | restart: |
938 | if ((error = sblock(&so->so_snd, SBLOCKWAIT(flags))) != 0) | | 943 | if ((error = sblock(&so->so_snd, SBLOCKWAIT(flags))) != 0) |
939 | goto out; | | 944 | goto out; |
940 | do { | | 945 | do { |
941 | if (so->so_state & SS_CANTSENDMORE) { | | 946 | if (so->so_state & SS_CANTSENDMORE) { |
942 | error = EPIPE; | | 947 | error = EPIPE; |
943 | goto release; | | 948 | goto release; |
944 | } | | 949 | } |
945 | if (so->so_error) { | | 950 | if (so->so_error) { |
946 | error = so->so_error; | | 951 | error = so->so_error; |
947 | so->so_error = 0; | | 952 | so->so_error = 0; |
948 | goto release; | | 953 | goto release; |
949 | } | | 954 | } |
950 | if ((so->so_state & SS_ISCONNECTED) == 0) { | | 955 | if ((so->so_state & SS_ISCONNECTED) == 0) { |
951 | if (so->so_proto->pr_flags & PR_CONNREQUIRED) { | | 956 | if (so->so_proto->pr_flags & PR_CONNREQUIRED) { |
952 | if ((so->so_state & SS_ISCONFIRMING) == 0 && | | 957 | if ((so->so_state & SS_ISCONFIRMING) == 0 && |
953 | !(resid == 0 && clen != 0)) { | | 958 | !(resid == 0 && clen != 0)) { |
954 | error = ENOTCONN; | | 959 | error = ENOTCONN; |
955 | goto release; | | 960 | goto release; |
956 | } | | 961 | } |
957 | } else if (addr == 0) { | | 962 | } else if (addr == 0) { |
958 | error = EDESTADDRREQ; | | 963 | error = EDESTADDRREQ; |
959 | goto release; | | 964 | goto release; |
960 | } | | 965 | } |
961 | } | | 966 | } |
962 | space = sbspace(&so->so_snd); | | 967 | space = sbspace(&so->so_snd); |
963 | if (flags & MSG_OOB) | | 968 | if (flags & MSG_OOB) |
964 | space += 1024; | | 969 | space += 1024; |
965 | if ((atomic && resid > so->so_snd.sb_hiwat) || | | 970 | if ((atomic && resid > so->so_snd.sb_hiwat) || |
966 | clen > so->so_snd.sb_hiwat) { | | 971 | clen > so->so_snd.sb_hiwat) { |
967 | error = EMSGSIZE; | | 972 | error = EMSGSIZE; |
968 | goto release; | | 973 | goto release; |
969 | } | | 974 | } |
970 | if (space < resid + clen && | | 975 | if (space < resid + clen && |
971 | (atomic || space < so->so_snd.sb_lowat || space < clen)) { | | 976 | (atomic || space < so->so_snd.sb_lowat || space < clen)) { |
972 | if (so->so_nbio) { | | 977 | if (so->so_nbio) { |
973 | error = EWOULDBLOCK; | | 978 | error = EWOULDBLOCK; |
974 | goto release; | | 979 | goto release; |
975 | } | | 980 | } |
976 | sbunlock(&so->so_snd); | | 981 | sbunlock(&so->so_snd); |
977 | error = sbwait(&so->so_snd); | | 982 | error = sbwait(&so->so_snd); |
978 | if (error) | | 983 | if (error) |
979 | goto out; | | 984 | goto out; |
980 | goto restart; | | 985 | goto restart; |
981 | } | | 986 | } |
982 | mp = ⊤ | | 987 | mp = ⊤ |
983 | space -= clen; | | 988 | space -= clen; |
984 | do { | | 989 | do { |
985 | if (uio == NULL) { | | 990 | if (uio == NULL) { |
986 | /* | | 991 | /* |
987 | * Data is prepackaged in "top". | | 992 | * Data is prepackaged in "top". |
988 | */ | | 993 | */ |
989 | resid = 0; | | 994 | resid = 0; |
990 | if (flags & MSG_EOR) | | 995 | if (flags & MSG_EOR) |
991 | top->m_flags |= M_EOR; | | 996 | top->m_flags |= M_EOR; |
992 | } else do { | | 997 | } else do { |
993 | sounlock(so); | | 998 | sounlock(so); |
994 | splx(s); | | 999 | splx(s); |
995 | if (top == NULL) { | | 1000 | if (top == NULL) { |
996 | m = m_gethdr(M_WAIT, MT_DATA); | | 1001 | m = m_gethdr(M_WAIT, MT_DATA); |
997 | mlen = MHLEN; | | 1002 | mlen = MHLEN; |
998 | m->m_pkthdr.len = 0; | | 1003 | m->m_pkthdr.len = 0; |
999 | m->m_pkthdr.rcvif = NULL; | | 1004 | m->m_pkthdr.rcvif = NULL; |
1000 | } else { | | 1005 | } else { |
1001 | m = m_get(M_WAIT, MT_DATA); | | 1006 | m = m_get(M_WAIT, MT_DATA); |
1002 | mlen = MLEN; | | 1007 | mlen = MLEN; |
1003 | } | | 1008 | } |
1004 | MCLAIM(m, so->so_snd.sb_mowner); | | 1009 | MCLAIM(m, so->so_snd.sb_mowner); |
1005 | if (sock_loan_thresh >= 0 && | | 1010 | if (sock_loan_thresh >= 0 && |
1006 | uio->uio_iov->iov_len >= sock_loan_thresh && | | 1011 | uio->uio_iov->iov_len >= sock_loan_thresh && |
1007 | space >= sock_loan_thresh && | | 1012 | space >= sock_loan_thresh && |
1008 | (len = sosend_loan(so, uio, m, | | 1013 | (len = sosend_loan(so, uio, m, |
1009 | space)) != 0) { | | 1014 | space)) != 0) { |
1010 | SOSEND_COUNTER_INCR(&sosend_loan_big); | | 1015 | SOSEND_COUNTER_INCR(&sosend_loan_big); |
1011 | space -= len; | | 1016 | space -= len; |
1012 | goto have_data; | | 1017 | goto have_data; |
1013 | } | | 1018 | } |
1014 | if (resid >= MINCLSIZE && space >= MCLBYTES) { | | 1019 | if (resid >= MINCLSIZE && space >= MCLBYTES) { |
1015 | SOSEND_COUNTER_INCR(&sosend_copy_big); | | 1020 | SOSEND_COUNTER_INCR(&sosend_copy_big); |
1016 | m_clget(m, M_WAIT); | | 1021 | m_clget(m, M_WAIT); |
1017 | if ((m->m_flags & M_EXT) == 0) | | 1022 | if ((m->m_flags & M_EXT) == 0) |
1018 | goto nopages; | | 1023 | goto nopages; |
1019 | mlen = MCLBYTES; | | 1024 | mlen = MCLBYTES; |
1020 | if (atomic && top == 0) { | | 1025 | if (atomic && top == 0) { |
1021 | len = lmin(MCLBYTES - max_hdr, | | 1026 | len = lmin(MCLBYTES - max_hdr, |
1022 | resid); | | 1027 | resid); |
1023 | m->m_data += max_hdr; | | 1028 | m->m_data += max_hdr; |
1024 | } else | | 1029 | } else |
1025 | len = lmin(MCLBYTES, resid); | | 1030 | len = lmin(MCLBYTES, resid); |
1026 | space -= len; | | 1031 | space -= len; |
1027 | } else { | | 1032 | } else { |
1028 | nopages: | | 1033 | nopages: |
1029 | SOSEND_COUNTER_INCR(&sosend_copy_small); | | 1034 | SOSEND_COUNTER_INCR(&sosend_copy_small); |
1030 | len = lmin(lmin(mlen, resid), space); | | 1035 | len = lmin(lmin(mlen, resid), space); |
1031 | space -= len; | | 1036 | space -= len; |
1032 | /* | | 1037 | /* |
1033 | * For datagram protocols, leave room | | 1038 | * For datagram protocols, leave room |
1034 | * for protocol headers in first mbuf. | | 1039 | * for protocol headers in first mbuf. |
1035 | */ | | 1040 | */ |
1036 | if (atomic && top == 0 && len < mlen) | | 1041 | if (atomic && top == 0 && len < mlen) |
1037 | MH_ALIGN(m, len); | | 1042 | MH_ALIGN(m, len); |
1038 | } | | 1043 | } |
1039 | error = uiomove(mtod(m, void *), (int)len, uio); | | 1044 | error = uiomove(mtod(m, void *), (int)len, uio); |
1040 | have_data: | | 1045 | have_data: |
1041 | resid = uio->uio_resid; | | 1046 | resid = uio->uio_resid; |
1042 | m->m_len = len; | | 1047 | m->m_len = len; |
1043 | *mp = m; | | 1048 | *mp = m; |
1044 | top->m_pkthdr.len += len; | | 1049 | top->m_pkthdr.len += len; |
1045 | s = splsoftnet(); | | 1050 | s = splsoftnet(); |
1046 | solock(so); | | 1051 | solock(so); |
1047 | if (error != 0) | | 1052 | if (error != 0) |
1048 | goto release; | | 1053 | goto release; |
1049 | mp = &m->m_next; | | 1054 | mp = &m->m_next; |
1050 | if (resid <= 0) { | | 1055 | if (resid <= 0) { |
1051 | if (flags & MSG_EOR) | | 1056 | if (flags & MSG_EOR) |
1052 | top->m_flags |= M_EOR; | | 1057 | top->m_flags |= M_EOR; |
1053 | break; | | 1058 | break; |
1054 | } | | 1059 | } |
1055 | } while (space > 0 && atomic); | | 1060 | } while (space > 0 && atomic); |
1056 | | | 1061 | |
1057 | if (so->so_state & SS_CANTSENDMORE) { | | 1062 | if (so->so_state & SS_CANTSENDMORE) { |
1058 | error = EPIPE; | | 1063 | error = EPIPE; |
1059 | goto release; | | 1064 | goto release; |
1060 | } | | 1065 | } |
1061 | if (dontroute) | | 1066 | if (dontroute) |
1062 | so->so_options |= SO_DONTROUTE; | | 1067 | so->so_options |= SO_DONTROUTE; |
1063 | if (resid > 0) | | 1068 | if (resid > 0) |
1064 | so->so_state |= SS_MORETOCOME; | | 1069 | so->so_state |= SS_MORETOCOME; |
1065 | error = (*so->so_proto->pr_usrreq)(so, | | 1070 | error = (*so->so_proto->pr_usrreq)(so, |
1066 | (flags & MSG_OOB) ? PRU_SENDOOB : PRU_SEND, | | 1071 | (flags & MSG_OOB) ? PRU_SENDOOB : PRU_SEND, |
1067 | top, addr, control, curlwp); | | 1072 | top, addr, control, curlwp); |
1068 | if (dontroute) | | 1073 | if (dontroute) |
1069 | so->so_options &= ~SO_DONTROUTE; | | 1074 | so->so_options &= ~SO_DONTROUTE; |
1070 | if (resid > 0) | | 1075 | if (resid > 0) |
1071 | so->so_state &= ~SS_MORETOCOME; | | 1076 | so->so_state &= ~SS_MORETOCOME; |
1072 | clen = 0; | | 1077 | clen = 0; |
1073 | control = NULL; | | 1078 | control = NULL; |
1074 | top = NULL; | | 1079 | top = NULL; |
1075 | mp = ⊤ | | 1080 | mp = ⊤ |
1076 | if (error != 0) | | 1081 | if (error != 0) |
1077 | goto release; | | 1082 | goto release; |
1078 | } while (resid && space > 0); | | 1083 | } while (resid && space > 0); |
1079 | } while (resid); | | 1084 | } while (resid); |
1080 | | | 1085 | |
1081 | release: | | 1086 | release: |
1082 | sbunlock(&so->so_snd); | | 1087 | sbunlock(&so->so_snd); |
1083 | out: | | 1088 | out: |
1084 | sounlock(so); | | 1089 | sounlock(so); |
1085 | splx(s); | | 1090 | splx(s); |
1086 | if (top) | | 1091 | if (top) |
1087 | m_freem(top); | | 1092 | m_freem(top); |
1088 | if (control) | | 1093 | if (control) |
1089 | m_freem(control); | | 1094 | m_freem(control); |
1090 | return (error); | | 1095 | return (error); |
1091 | } | | 1096 | } |
1092 | | | 1097 | |
1093 | /* | | 1098 | /* |
1094 | * Following replacement or removal of the first mbuf on the first | | 1099 | * Following replacement or removal of the first mbuf on the first |
1095 | * mbuf chain of a socket buffer, push necessary state changes back | | 1100 | * mbuf chain of a socket buffer, push necessary state changes back |
1096 | * into the socket buffer so that other consumers see the values | | 1101 | * into the socket buffer so that other consumers see the values |
1097 | * consistently. 'nextrecord' is the callers locally stored value of | | 1102 | * consistently. 'nextrecord' is the callers locally stored value of |
1098 | * the original value of sb->sb_mb->m_nextpkt which must be restored | | 1103 | * the original value of sb->sb_mb->m_nextpkt which must be restored |
1099 | * when the lead mbuf changes. NOTE: 'nextrecord' may be NULL. | | 1104 | * when the lead mbuf changes. NOTE: 'nextrecord' may be NULL. |
1100 | */ | | 1105 | */ |
1101 | static void | | 1106 | static void |
1102 | sbsync(struct sockbuf *sb, struct mbuf *nextrecord) | | 1107 | sbsync(struct sockbuf *sb, struct mbuf *nextrecord) |
1103 | { | | 1108 | { |
1104 | | | 1109 | |
1105 | KASSERT(solocked(sb->sb_so)); | | 1110 | KASSERT(solocked(sb->sb_so)); |
1106 | | | 1111 | |
1107 | /* | | 1112 | /* |
1108 | * First, update for the new value of nextrecord. If necessary, | | 1113 | * First, update for the new value of nextrecord. If necessary, |
1109 | * make it the first record. | | 1114 | * make it the first record. |
1110 | */ | | 1115 | */ |
1111 | if (sb->sb_mb != NULL) | | 1116 | if (sb->sb_mb != NULL) |
1112 | sb->sb_mb->m_nextpkt = nextrecord; | | 1117 | sb->sb_mb->m_nextpkt = nextrecord; |
1113 | else | | 1118 | else |
1114 | sb->sb_mb = nextrecord; | | 1119 | sb->sb_mb = nextrecord; |
1115 | | | 1120 | |
1116 | /* | | 1121 | /* |
1117 | * Now update any dependent socket buffer fields to reflect | | 1122 | * Now update any dependent socket buffer fields to reflect |
1118 | * the new state. This is an inline of SB_EMPTY_FIXUP, with | | 1123 | * the new state. This is an inline of SB_EMPTY_FIXUP, with |
1119 | * the addition of a second clause that takes care of the | | 1124 | * the addition of a second clause that takes care of the |
1120 | * case where sb_mb has been updated, but remains the last | | 1125 | * case where sb_mb has been updated, but remains the last |
1121 | * record. | | 1126 | * record. |
1122 | */ | | 1127 | */ |
1123 | if (sb->sb_mb == NULL) { | | 1128 | if (sb->sb_mb == NULL) { |
1124 | sb->sb_mbtail = NULL; | | 1129 | sb->sb_mbtail = NULL; |
1125 | sb->sb_lastrecord = NULL; | | 1130 | sb->sb_lastrecord = NULL; |
1126 | } else if (sb->sb_mb->m_nextpkt == NULL) | | 1131 | } else if (sb->sb_mb->m_nextpkt == NULL) |
1127 | sb->sb_lastrecord = sb->sb_mb; | | 1132 | sb->sb_lastrecord = sb->sb_mb; |
1128 | } | | 1133 | } |
1129 | | | 1134 | |
1130 | /* | | 1135 | /* |
1131 | * Implement receive operations on a socket. | | 1136 | * Implement receive operations on a socket. |
1132 | * We depend on the way that records are added to the sockbuf | | 1137 | * We depend on the way that records are added to the sockbuf |
1133 | * by sbappend*. In particular, each record (mbufs linked through m_next) | | 1138 | * by sbappend*. In particular, each record (mbufs linked through m_next) |
1134 | * must begin with an address if the protocol so specifies, | | 1139 | * must begin with an address if the protocol so specifies, |
1135 | * followed by an optional mbuf or mbufs containing ancillary data, | | 1140 | * followed by an optional mbuf or mbufs containing ancillary data, |
1136 | * and then zero or more mbufs of data. | | 1141 | * and then zero or more mbufs of data. |
1137 | * In order to avoid blocking network interrupts for the entire time here, | | 1142 | * In order to avoid blocking network interrupts for the entire time here, |
1138 | * we splx() while doing the actual copy to user space. | | 1143 | * we splx() while doing the actual copy to user space. |
1139 | * Although the sockbuf is locked, new data may still be appended, | | 1144 | * Although the sockbuf is locked, new data may still be appended, |
1140 | * and thus we must maintain consistency of the sockbuf during that time. | | 1145 | * and thus we must maintain consistency of the sockbuf during that time. |
1141 | * | | 1146 | * |
1142 | * The caller may receive the data as a single mbuf chain by supplying | | 1147 | * The caller may receive the data as a single mbuf chain by supplying |
1143 | * an mbuf **mp0 for use in returning the chain. The uio is then used | | 1148 | * an mbuf **mp0 for use in returning the chain. The uio is then used |
1144 | * only for the count in uio_resid. | | 1149 | * only for the count in uio_resid. |
1145 | */ | | 1150 | */ |
1146 | int | | 1151 | int |
1147 | soreceive(struct socket *so, struct mbuf **paddr, struct uio *uio, | | 1152 | soreceive(struct socket *so, struct mbuf **paddr, struct uio *uio, |
1148 | struct mbuf **mp0, struct mbuf **controlp, int *flagsp) | | 1153 | struct mbuf **mp0, struct mbuf **controlp, int *flagsp) |
1149 | { | | 1154 | { |
1150 | struct lwp *l = curlwp; | | 1155 | struct lwp *l = curlwp; |
1151 | struct mbuf *m, **mp, *mt; | | 1156 | struct mbuf *m, **mp, *mt; |
1152 | int atomic, flags, len, error, s, offset, moff, type, orig_resid; | | 1157 | int atomic, flags, len, error, s, offset, moff, type, orig_resid; |
1153 | const struct protosw *pr; | | 1158 | const struct protosw *pr; |
1154 | struct mbuf *nextrecord; | | 1159 | struct mbuf *nextrecord; |
1155 | int mbuf_removed = 0; | | 1160 | int mbuf_removed = 0; |
1156 | const struct domain *dom; | | 1161 | const struct domain *dom; |
1157 | | | 1162 | |
1158 | pr = so->so_proto; | | 1163 | pr = so->so_proto; |
1159 | atomic = pr->pr_flags & PR_ATOMIC; | | 1164 | atomic = pr->pr_flags & PR_ATOMIC; |
1160 | dom = pr->pr_domain; | | 1165 | dom = pr->pr_domain; |
1161 | mp = mp0; | | 1166 | mp = mp0; |
1162 | type = 0; | | 1167 | type = 0; |
1163 | orig_resid = uio->uio_resid; | | 1168 | orig_resid = uio->uio_resid; |
1164 | | | 1169 | |
1165 | if (paddr != NULL) | | 1170 | if (paddr != NULL) |
1166 | *paddr = NULL; | | 1171 | *paddr = NULL; |
1167 | if (controlp != NULL) | | 1172 | if (controlp != NULL) |
1168 | *controlp = NULL; | | 1173 | *controlp = NULL; |
1169 | if (flagsp != NULL) | | 1174 | if (flagsp != NULL) |
1170 | flags = *flagsp &~ MSG_EOR; | | 1175 | flags = *flagsp &~ MSG_EOR; |
1171 | else | | 1176 | else |
1172 | flags = 0; | | 1177 | flags = 0; |
1173 | | | 1178 | |
1174 | if ((flags & MSG_DONTWAIT) == 0) | | 1179 | if ((flags & MSG_DONTWAIT) == 0) |
1175 | sodopendfree(); | | 1180 | sodopendfree(); |
1176 | | | 1181 | |
1177 | if (flags & MSG_OOB) { | | 1182 | if (flags & MSG_OOB) { |
1178 | m = m_get(M_WAIT, MT_DATA); | | 1183 | m = m_get(M_WAIT, MT_DATA); |
1179 | solock(so); | | 1184 | solock(so); |
1180 | error = (*pr->pr_usrreq)(so, PRU_RCVOOB, m, | | 1185 | error = (*pr->pr_usrreq)(so, PRU_RCVOOB, m, |
1181 | (struct mbuf *)(long)(flags & MSG_PEEK), NULL, l); | | 1186 | (struct mbuf *)(long)(flags & MSG_PEEK), NULL, l); |
1182 | sounlock(so); | | 1187 | sounlock(so); |
1183 | if (error) | | 1188 | if (error) |
1184 | goto bad; | | 1189 | goto bad; |
1185 | do { | | 1190 | do { |
1186 | error = uiomove(mtod(m, void *), | | 1191 | error = uiomove(mtod(m, void *), |
1187 | (int) min(uio->uio_resid, m->m_len), uio); | | 1192 | (int) min(uio->uio_resid, m->m_len), uio); |
1188 | m = m_free(m); | | 1193 | m = m_free(m); |
1189 | } while (uio->uio_resid > 0 && error == 0 && m); | | 1194 | } while (uio->uio_resid > 0 && error == 0 && m); |
1190 | bad: | | 1195 | bad: |
1191 | if (m != NULL) | | 1196 | if (m != NULL) |
1192 | m_freem(m); | | 1197 | m_freem(m); |
1193 | return error; | | 1198 | return error; |
1194 | } | | 1199 | } |
1195 | if (mp != NULL) | | 1200 | if (mp != NULL) |
1196 | *mp = NULL; | | 1201 | *mp = NULL; |
1197 | | | 1202 | |
1198 | /* | | 1203 | /* |
1199 | * solock() provides atomicity of access. splsoftnet() prevents | | 1204 | * solock() provides atomicity of access. splsoftnet() prevents |
1200 | * protocol processing soft interrupts from interrupting us and | | 1205 | * protocol processing soft interrupts from interrupting us and |
1201 | * blocking (expensive). | | 1206 | * blocking (expensive). |
1202 | */ | | 1207 | */ |
1203 | s = splsoftnet(); | | 1208 | s = splsoftnet(); |
1204 | solock(so); | | 1209 | solock(so); |
1205 | if (so->so_state & SS_ISCONFIRMING && uio->uio_resid) | | 1210 | if (so->so_state & SS_ISCONFIRMING && uio->uio_resid) |
1206 | (*pr->pr_usrreq)(so, PRU_RCVD, NULL, NULL, NULL, l); | | 1211 | (*pr->pr_usrreq)(so, PRU_RCVD, NULL, NULL, NULL, l); |
1207 | | | 1212 | |
1208 | restart: | | 1213 | restart: |
1209 | if ((error = sblock(&so->so_rcv, SBLOCKWAIT(flags))) != 0) { | | 1214 | if ((error = sblock(&so->so_rcv, SBLOCKWAIT(flags))) != 0) { |
1210 | sounlock(so); | | 1215 | sounlock(so); |
1211 | splx(s); | | 1216 | splx(s); |
1212 | return error; | | 1217 | return error; |
1213 | } | | 1218 | } |
1214 | | | 1219 | |
1215 | m = so->so_rcv.sb_mb; | | 1220 | m = so->so_rcv.sb_mb; |
1216 | /* | | 1221 | /* |
1217 | * If we have less data than requested, block awaiting more | | 1222 | * If we have less data than requested, block awaiting more |
1218 | * (subject to any timeout) if: | | 1223 | * (subject to any timeout) if: |
1219 | * 1. the current count is less than the low water mark, | | 1224 | * 1. the current count is less than the low water mark, |
1220 | * 2. MSG_WAITALL is set, and it is possible to do the entire | | 1225 | * 2. MSG_WAITALL is set, and it is possible to do the entire |
1221 | * receive operation at once if we block (resid <= hiwat), or | | 1226 | * receive operation at once if we block (resid <= hiwat), or |
1222 | * 3. MSG_DONTWAIT is not set. | | 1227 | * 3. MSG_DONTWAIT is not set. |
1223 | * If MSG_WAITALL is set but resid is larger than the receive buffer, | | 1228 | * If MSG_WAITALL is set but resid is larger than the receive buffer, |
1224 | * we have to do the receive in sections, and thus risk returning | | 1229 | * we have to do the receive in sections, and thus risk returning |
1225 | * a short count if a timeout or signal occurs after we start. | | 1230 | * a short count if a timeout or signal occurs after we start. |
1226 | */ | | 1231 | */ |
1227 | if (m == NULL || | | 1232 | if (m == NULL || |
1228 | ((flags & MSG_DONTWAIT) == 0 && | | 1233 | ((flags & MSG_DONTWAIT) == 0 && |
1229 | so->so_rcv.sb_cc < uio->uio_resid && | | 1234 | so->so_rcv.sb_cc < uio->uio_resid && |
1230 | (so->so_rcv.sb_cc < so->so_rcv.sb_lowat || | | 1235 | (so->so_rcv.sb_cc < so->so_rcv.sb_lowat || |
1231 | ((flags & MSG_WAITALL) && | | 1236 | ((flags & MSG_WAITALL) && |
1232 | uio->uio_resid <= so->so_rcv.sb_hiwat)) && | | 1237 | uio->uio_resid <= so->so_rcv.sb_hiwat)) && |
1233 | m->m_nextpkt == NULL && !atomic)) { | | 1238 | m->m_nextpkt == NULL && !atomic)) { |
1234 | #ifdef DIAGNOSTIC | | 1239 | #ifdef DIAGNOSTIC |
1235 | if (m == NULL && so->so_rcv.sb_cc) | | 1240 | if (m == NULL && so->so_rcv.sb_cc) |
1236 | panic("receive 1"); | | 1241 | panic("receive 1"); |
1237 | #endif | | 1242 | #endif |
1238 | if (so->so_error) { | | 1243 | if (so->so_error) { |
1239 | if (m != NULL) | | 1244 | if (m != NULL) |
1240 | goto dontblock; | | 1245 | goto dontblock; |
1241 | error = so->so_error; | | 1246 | error = so->so_error; |
1242 | if ((flags & MSG_PEEK) == 0) | | 1247 | if ((flags & MSG_PEEK) == 0) |
1243 | so->so_error = 0; | | 1248 | so->so_error = 0; |
1244 | goto release; | | 1249 | goto release; |
1245 | } | | 1250 | } |
1246 | if (so->so_state & SS_CANTRCVMORE) { | | 1251 | if (so->so_state & SS_CANTRCVMORE) { |
1247 | if (m != NULL) | | 1252 | if (m != NULL) |
1248 | goto dontblock; | | 1253 | goto dontblock; |
1249 | else | | 1254 | else |
1250 | goto release; | | 1255 | goto release; |
1251 | } | | 1256 | } |
1252 | for (; m != NULL; m = m->m_next) | | 1257 | for (; m != NULL; m = m->m_next) |
1253 | if (m->m_type == MT_OOBDATA || (m->m_flags & M_EOR)) { | | 1258 | if (m->m_type == MT_OOBDATA || (m->m_flags & M_EOR)) { |
1254 | m = so->so_rcv.sb_mb; | | 1259 | m = so->so_rcv.sb_mb; |
1255 | goto dontblock; | | 1260 | goto dontblock; |
1256 | } | | 1261 | } |
1257 | if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0 && | | 1262 | if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0 && |
1258 | (so->so_proto->pr_flags & PR_CONNREQUIRED)) { | | 1263 | (so->so_proto->pr_flags & PR_CONNREQUIRED)) { |
1259 | error = ENOTCONN; | | 1264 | error = ENOTCONN; |
1260 | goto release; | | 1265 | goto release; |
1261 | } | | 1266 | } |
1262 | if (uio->uio_resid == 0) | | 1267 | if (uio->uio_resid == 0) |
1263 | goto release; | | 1268 | goto release; |
1264 | if (so->so_nbio || (flags & MSG_DONTWAIT)) { | | 1269 | if (so->so_nbio || (flags & MSG_DONTWAIT)) { |
1265 | error = EWOULDBLOCK; | | 1270 | error = EWOULDBLOCK; |
1266 | goto release; | | 1271 | goto release; |
1267 | } | | 1272 | } |
1268 | SBLASTRECORDCHK(&so->so_rcv, "soreceive sbwait 1"); | | 1273 | SBLASTRECORDCHK(&so->so_rcv, "soreceive sbwait 1"); |
1269 | SBLASTMBUFCHK(&so->so_rcv, "soreceive sbwait 1"); | | 1274 | SBLASTMBUFCHK(&so->so_rcv, "soreceive sbwait 1"); |
1270 | sbunlock(&so->so_rcv); | | 1275 | sbunlock(&so->so_rcv); |
1271 | error = sbwait(&so->so_rcv); | | 1276 | error = sbwait(&so->so_rcv); |
1272 | if (error != 0) { | | 1277 | if (error != 0) { |
1273 | sounlock(so); | | 1278 | sounlock(so); |
1274 | splx(s); | | 1279 | splx(s); |
1275 | return error; | | 1280 | return error; |
1276 | } | | 1281 | } |
1277 | goto restart; | | 1282 | goto restart; |
1278 | } | | 1283 | } |
1279 | dontblock: | | 1284 | dontblock: |
1280 | /* | | 1285 | /* |
1281 | * On entry here, m points to the first record of the socket buffer. | | 1286 | * On entry here, m points to the first record of the socket buffer. |
1282 | * From this point onward, we maintain 'nextrecord' as a cache of the | | 1287 | * From this point onward, we maintain 'nextrecord' as a cache of the |
1283 | * pointer to the next record in the socket buffer. We must keep the | | 1288 | * pointer to the next record in the socket buffer. We must keep the |
1284 | * various socket buffer pointers and local stack versions of the | | 1289 | * various socket buffer pointers and local stack versions of the |
1285 | * pointers in sync, pushing out modifications before dropping the | | 1290 | * pointers in sync, pushing out modifications before dropping the |
1286 | * socket lock, and re-reading them when picking it up. | | 1291 | * socket lock, and re-reading them when picking it up. |
1287 | * | | 1292 | * |
1288 | * Otherwise, we will race with the network stack appending new data | | 1293 | * Otherwise, we will race with the network stack appending new data |
1289 | * or records onto the socket buffer by using inconsistent/stale | | 1294 | * or records onto the socket buffer by using inconsistent/stale |
1290 | * versions of the field, possibly resulting in socket buffer | | 1295 | * versions of the field, possibly resulting in socket buffer |
1291 | * corruption. | | 1296 | * corruption. |
1292 | * | | 1297 | * |
1293 | * By holding the high-level sblock(), we prevent simultaneous | | 1298 | * By holding the high-level sblock(), we prevent simultaneous |
1294 | * readers from pulling off the front of the socket buffer. | | 1299 | * readers from pulling off the front of the socket buffer. |
1295 | */ | | 1300 | */ |
1296 | if (l != NULL) | | 1301 | if (l != NULL) |
1297 | l->l_ru.ru_msgrcv++; | | 1302 | l->l_ru.ru_msgrcv++; |
1298 | KASSERT(m == so->so_rcv.sb_mb); | | 1303 | KASSERT(m == so->so_rcv.sb_mb); |
1299 | SBLASTRECORDCHK(&so->so_rcv, "soreceive 1"); | | 1304 | SBLASTRECORDCHK(&so->so_rcv, "soreceive 1"); |
1300 | SBLASTMBUFCHK(&so->so_rcv, "soreceive 1"); | | 1305 | SBLASTMBUFCHK(&so->so_rcv, "soreceive 1"); |
1301 | nextrecord = m->m_nextpkt; | | 1306 | nextrecord = m->m_nextpkt; |
1302 | if (pr->pr_flags & PR_ADDR) { | | 1307 | if (pr->pr_flags & PR_ADDR) { |
1303 | #ifdef DIAGNOSTIC | | 1308 | #ifdef DIAGNOSTIC |
1304 | if (m->m_type != MT_SONAME) | | 1309 | if (m->m_type != MT_SONAME) |
1305 | panic("receive 1a"); | | 1310 | panic("receive 1a"); |
1306 | #endif | | 1311 | #endif |
1307 | orig_resid = 0; | | 1312 | orig_resid = 0; |
1308 | if (flags & MSG_PEEK) { | | 1313 | if (flags & MSG_PEEK) { |
1309 | if (paddr) | | 1314 | if (paddr) |
1310 | *paddr = m_copy(m, 0, m->m_len); | | 1315 | *paddr = m_copy(m, 0, m->m_len); |
1311 | m = m->m_next; | | 1316 | m = m->m_next; |
1312 | } else { | | 1317 | } else { |
1313 | sbfree(&so->so_rcv, m); | | 1318 | sbfree(&so->so_rcv, m); |
1314 | mbuf_removed = 1; | | 1319 | mbuf_removed = 1; |
1315 | if (paddr != NULL) { | | 1320 | if (paddr != NULL) { |
1316 | *paddr = m; | | 1321 | *paddr = m; |
1317 | so->so_rcv.sb_mb = m->m_next; | | 1322 | so->so_rcv.sb_mb = m->m_next; |
1318 | m->m_next = NULL; | | 1323 | m->m_next = NULL; |
1319 | m = so->so_rcv.sb_mb; | | 1324 | m = so->so_rcv.sb_mb; |
1320 | } else { | | 1325 | } else { |
1321 | MFREE(m, so->so_rcv.sb_mb); | | 1326 | MFREE(m, so->so_rcv.sb_mb); |
1322 | m = so->so_rcv.sb_mb; | | 1327 | m = so->so_rcv.sb_mb; |
1323 | } | | 1328 | } |
1324 | sbsync(&so->so_rcv, nextrecord); | | 1329 | sbsync(&so->so_rcv, nextrecord); |
1325 | } | | 1330 | } |
1326 | } | | 1331 | } |
1327 | | | 1332 | |
1328 | /* | | 1333 | /* |
1329 | * Process one or more MT_CONTROL mbufs present before any data mbufs | | 1334 | * Process one or more MT_CONTROL mbufs present before any data mbufs |
1330 | * in the first mbuf chain on the socket buffer. If MSG_PEEK, we | | 1335 | * in the first mbuf chain on the socket buffer. If MSG_PEEK, we |
1331 | * just copy the data; if !MSG_PEEK, we call into the protocol to | | 1336 | * just copy the data; if !MSG_PEEK, we call into the protocol to |
1332 | * perform externalization (or freeing if controlp == NULL). | | 1337 | * perform externalization (or freeing if controlp == NULL). |
1333 | */ | | 1338 | */ |
1334 | if (__predict_false(m != NULL && m->m_type == MT_CONTROL)) { | | 1339 | if (__predict_false(m != NULL && m->m_type == MT_CONTROL)) { |
1335 | struct mbuf *cm = NULL, *cmn; | | 1340 | struct mbuf *cm = NULL, *cmn; |
1336 | struct mbuf **cme = &cm; | | 1341 | struct mbuf **cme = &cm; |
1337 | | | 1342 | |
1338 | do { | | 1343 | do { |
1339 | if (flags & MSG_PEEK) { | | 1344 | if (flags & MSG_PEEK) { |
1340 | if (controlp != NULL) { | | 1345 | if (controlp != NULL) { |
1341 | *controlp = m_copy(m, 0, m->m_len); | | 1346 | *controlp = m_copy(m, 0, m->m_len); |
1342 | controlp = &(*controlp)->m_next; | | 1347 | controlp = &(*controlp)->m_next; |
1343 | } | | 1348 | } |
1344 | m = m->m_next; | | 1349 | m = m->m_next; |
1345 | } else { | | 1350 | } else { |
1346 | sbfree(&so->so_rcv, m); | | 1351 | sbfree(&so->so_rcv, m); |
1347 | so->so_rcv.sb_mb = m->m_next; | | 1352 | so->so_rcv.sb_mb = m->m_next; |
1348 | m->m_next = NULL; | | 1353 | m->m_next = NULL; |
1349 | *cme = m; | | 1354 | *cme = m; |
1350 | cme = &(*cme)->m_next; | | 1355 | cme = &(*cme)->m_next; |
1351 | m = so->so_rcv.sb_mb; | | 1356 | m = so->so_rcv.sb_mb; |
1352 | } | | 1357 | } |
1353 | } while (m != NULL && m->m_type == MT_CONTROL); | | 1358 | } while (m != NULL && m->m_type == MT_CONTROL); |
1354 | if ((flags & MSG_PEEK) == 0) | | 1359 | if ((flags & MSG_PEEK) == 0) |
1355 | sbsync(&so->so_rcv, nextrecord); | | 1360 | sbsync(&so->so_rcv, nextrecord); |
1356 | for (; cm != NULL; cm = cmn) { | | 1361 | for (; cm != NULL; cm = cmn) { |
1357 | cmn = cm->m_next; | | 1362 | cmn = cm->m_next; |
1358 | cm->m_next = NULL; | | 1363 | cm->m_next = NULL; |
1359 | type = mtod(cm, struct cmsghdr *)->cmsg_type; | | 1364 | type = mtod(cm, struct cmsghdr *)->cmsg_type; |
1360 | if (controlp != NULL) { | | 1365 | if (controlp != NULL) { |
1361 | if (dom->dom_externalize != NULL && | | 1366 | if (dom->dom_externalize != NULL && |
1362 | type == SCM_RIGHTS) { | | 1367 | type == SCM_RIGHTS) { |
1363 | sounlock(so); | | 1368 | sounlock(so); |
1364 | splx(s); | | 1369 | splx(s); |
1365 | error = (*dom->dom_externalize)(cm, l); | | 1370 | error = (*dom->dom_externalize)(cm, l); |
1366 | s = splsoftnet(); | | 1371 | s = splsoftnet(); |
1367 | solock(so); | | 1372 | solock(so); |
1368 | } | | 1373 | } |
1369 | *controlp = cm; | | 1374 | *controlp = cm; |
1370 | while (*controlp != NULL) | | 1375 | while (*controlp != NULL) |
1371 | controlp = &(*controlp)->m_next; | | 1376 | controlp = &(*controlp)->m_next; |
1372 | } else { | | 1377 | } else { |
1373 | /* | | 1378 | /* |
1374 | * Dispose of any SCM_RIGHTS message that went | | 1379 | * Dispose of any SCM_RIGHTS message that went |
1375 | * through the read path rather than recv. | | 1380 | * through the read path rather than recv. |
1376 | */ | | 1381 | */ |
1377 | if (dom->dom_dispose != NULL && | | 1382 | if (dom->dom_dispose != NULL && |
1378 | type == SCM_RIGHTS) { | | 1383 | type == SCM_RIGHTS) { |
1379 | sounlock(so); | | 1384 | sounlock(so); |
1380 | (*dom->dom_dispose)(cm); | | 1385 | (*dom->dom_dispose)(cm); |
1381 | solock(so); | | 1386 | solock(so); |
1382 | } | | 1387 | } |
1383 | m_freem(cm); | | 1388 | m_freem(cm); |
1384 | } | | 1389 | } |
1385 | } | | 1390 | } |
1386 | if (m != NULL) | | 1391 | if (m != NULL) |
1387 | nextrecord = so->so_rcv.sb_mb->m_nextpkt; | | 1392 | nextrecord = so->so_rcv.sb_mb->m_nextpkt; |
1388 | else | | 1393 | else |
1389 | nextrecord = so->so_rcv.sb_mb; | | 1394 | nextrecord = so->so_rcv.sb_mb; |
1390 | orig_resid = 0; | | 1395 | orig_resid = 0; |
1391 | } | | 1396 | } |
1392 | | | 1397 | |
1393 | /* If m is non-NULL, we have some data to read. */ | | 1398 | /* If m is non-NULL, we have some data to read. */ |
1394 | if (__predict_true(m != NULL)) { | | 1399 | if (__predict_true(m != NULL)) { |
1395 | type = m->m_type; | | 1400 | type = m->m_type; |
1396 | if (type == MT_OOBDATA) | | 1401 | if (type == MT_OOBDATA) |
1397 | flags |= MSG_OOB; | | 1402 | flags |= MSG_OOB; |
1398 | } | | 1403 | } |
1399 | SBLASTRECORDCHK(&so->so_rcv, "soreceive 2"); | | 1404 | SBLASTRECORDCHK(&so->so_rcv, "soreceive 2"); |
1400 | SBLASTMBUFCHK(&so->so_rcv, "soreceive 2"); | | 1405 | SBLASTMBUFCHK(&so->so_rcv, "soreceive 2"); |
1401 | | | 1406 | |
1402 | moff = 0; | | 1407 | moff = 0; |
1403 | offset = 0; | | 1408 | offset = 0; |
1404 | while (m != NULL && uio->uio_resid > 0 && error == 0) { | | 1409 | while (m != NULL && uio->uio_resid > 0 && error == 0) { |
1405 | if (m->m_type == MT_OOBDATA) { | | 1410 | if (m->m_type == MT_OOBDATA) { |
1406 | if (type != MT_OOBDATA) | | 1411 | if (type != MT_OOBDATA) |
1407 | break; | | 1412 | break; |
1408 | } else if (type == MT_OOBDATA) | | 1413 | } else if (type == MT_OOBDATA) |
1409 | break; | | 1414 | break; |
1410 | #ifdef DIAGNOSTIC | | 1415 | #ifdef DIAGNOSTIC |
1411 | else if (m->m_type != MT_DATA && m->m_type != MT_HEADER) | | 1416 | else if (m->m_type != MT_DATA && m->m_type != MT_HEADER) |
1412 | panic("receive 3"); | | 1417 | panic("receive 3"); |
1413 | #endif | | 1418 | #endif |
1414 | so->so_state &= ~SS_RCVATMARK; | | 1419 | so->so_state &= ~SS_RCVATMARK; |
1415 | len = uio->uio_resid; | | 1420 | len = uio->uio_resid; |
1416 | if (so->so_oobmark && len > so->so_oobmark - offset) | | 1421 | if (so->so_oobmark && len > so->so_oobmark - offset) |
1417 | len = so->so_oobmark - offset; | | 1422 | len = so->so_oobmark - offset; |
1418 | if (len > m->m_len - moff) | | 1423 | if (len > m->m_len - moff) |
1419 | len = m->m_len - moff; | | 1424 | len = m->m_len - moff; |
1420 | /* | | 1425 | /* |
1421 | * If mp is set, just pass back the mbufs. | | 1426 | * If mp is set, just pass back the mbufs. |
1422 | * Otherwise copy them out via the uio, then free. | | 1427 | * Otherwise copy them out via the uio, then free. |
1423 | * Sockbuf must be consistent here (points to current mbuf, | | 1428 | * Sockbuf must be consistent here (points to current mbuf, |
1424 | * it points to next record) when we drop priority; | | 1429 | * it points to next record) when we drop priority; |
1425 | * we must note any additions to the sockbuf when we | | 1430 | * we must note any additions to the sockbuf when we |
1426 | * block interrupts again. | | 1431 | * block interrupts again. |
1427 | */ | | 1432 | */ |
1428 | if (mp == NULL) { | | 1433 | if (mp == NULL) { |
1429 | SBLASTRECORDCHK(&so->so_rcv, "soreceive uiomove"); | | 1434 | SBLASTRECORDCHK(&so->so_rcv, "soreceive uiomove"); |
1430 | SBLASTMBUFCHK(&so->so_rcv, "soreceive uiomove"); | | 1435 | SBLASTMBUFCHK(&so->so_rcv, "soreceive uiomove"); |
1431 | sounlock(so); | | 1436 | sounlock(so); |
1432 | splx(s); | | 1437 | splx(s); |
1433 | error = uiomove(mtod(m, char *) + moff, (int)len, uio); | | 1438 | error = uiomove(mtod(m, char *) + moff, (int)len, uio); |
1434 | s = splsoftnet(); | | 1439 | s = splsoftnet(); |
1435 | solock(so); | | 1440 | solock(so); |
1436 | if (error != 0) { | | 1441 | if (error != 0) { |
1437 | /* | | 1442 | /* |
1438 | * If any part of the record has been removed | | 1443 | * If any part of the record has been removed |
1439 | * (such as the MT_SONAME mbuf, which will | | 1444 | * (such as the MT_SONAME mbuf, which will |
1440 | * happen when PR_ADDR, and thus also | | 1445 | * happen when PR_ADDR, and thus also |
1441 | * PR_ATOMIC, is set), then drop the entire | | 1446 | * PR_ATOMIC, is set), then drop the entire |
1442 | * record to maintain the atomicity of the | | 1447 | * record to maintain the atomicity of the |
1443 | * receive operation. | | 1448 | * receive operation. |
1444 | * | | 1449 | * |
1445 | * This avoids a later panic("receive 1a") | | 1450 | * This avoids a later panic("receive 1a") |