Annotation of src/sys/kern/uipc_socket.c, Revision 1.158
1.158 ! ad 1: /* $NetBSD: uipc_socket.c,v 1.157 2008/03/27 19:06:52 ad Exp $ */
1.64 thorpej 2:
3: /*-
1.154 ad 4: * Copyright (c) 2002, 2007, 2008 The NetBSD Foundation, Inc.
1.64 thorpej 5: * All rights reserved.
6: *
7: * This code is derived from software contributed to The NetBSD Foundation
8: * by Jason R. Thorpe of Wasabi Systems, Inc.
9: *
10: * Redistribution and use in source and binary forms, with or without
11: * modification, are permitted provided that the following conditions
12: * are met:
13: * 1. Redistributions of source code must retain the above copyright
14: * notice, this list of conditions and the following disclaimer.
15: * 2. Redistributions in binary form must reproduce the above copyright
16: * notice, this list of conditions and the following disclaimer in the
17: * documentation and/or other materials provided with the distribution.
18: * 3. All advertising materials mentioning features or use of this software
19: * must display the following acknowledgement:
20: * This product includes software developed by the NetBSD
21: * Foundation, Inc. and its contributors.
22: * 4. Neither the name of The NetBSD Foundation nor the names of its
23: * contributors may be used to endorse or promote products derived
24: * from this software without specific prior written permission.
25: *
26: * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
27: * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
28: * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
29: * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
30: * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
31: * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
32: * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
33: * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
34: * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
35: * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
36: * POSSIBILITY OF SUCH DAMAGE.
37: */
1.16 cgd 38:
1.1 cgd 39: /*
1.15 mycroft 40: * Copyright (c) 1982, 1986, 1988, 1990, 1993
41: * The Regents of the University of California. All rights reserved.
1.1 cgd 42: *
43: * Redistribution and use in source and binary forms, with or without
44: * modification, are permitted provided that the following conditions
45: * are met:
46: * 1. Redistributions of source code must retain the above copyright
47: * notice, this list of conditions and the following disclaimer.
48: * 2. Redistributions in binary form must reproduce the above copyright
49: * notice, this list of conditions and the following disclaimer in the
50: * documentation and/or other materials provided with the distribution.
1.85 agc 51: * 3. Neither the name of the University nor the names of its contributors
1.1 cgd 52: * may be used to endorse or promote products derived from this software
53: * without specific prior written permission.
54: *
55: * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
56: * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
57: * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
58: * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
59: * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
60: * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
61: * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
62: * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
63: * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
64: * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
65: * SUCH DAMAGE.
66: *
1.32 fvdl 67: * @(#)uipc_socket.c 8.6 (Berkeley) 5/2/95
1.1 cgd 68: */
1.59 lukem 69:
70: #include <sys/cdefs.h>
1.158 ! ad 71: __KERNEL_RCSID(0, "$NetBSD: uipc_socket.c,v 1.157 2008/03/27 19:06:52 ad Exp $");
1.64 thorpej 72:
73: #include "opt_sock_counters.h"
74: #include "opt_sosend_loan.h"
1.81 martin 75: #include "opt_mbuftrace.h"
1.84 ragge 76: #include "opt_somaxkva.h"
1.1 cgd 77:
1.9 mycroft 78: #include <sys/param.h>
79: #include <sys/systm.h>
80: #include <sys/proc.h>
81: #include <sys/file.h>
1.142 dyoung 82: #include <sys/filedesc.h>
1.9 mycroft 83: #include <sys/malloc.h>
84: #include <sys/mbuf.h>
85: #include <sys/domain.h>
86: #include <sys/kernel.h>
87: #include <sys/protosw.h>
88: #include <sys/socket.h>
89: #include <sys/socketvar.h>
1.21 christos 90: #include <sys/signalvar.h>
1.9 mycroft 91: #include <sys/resourcevar.h>
1.37 thorpej 92: #include <sys/pool.h>
1.72 jdolecek 93: #include <sys/event.h>
1.89 christos 94: #include <sys/poll.h>
1.118 elad 95: #include <sys/kauth.h>
1.136 ad 96: #include <sys/mutex.h>
97: #include <sys/condvar.h>
1.37 thorpej 98:
1.64 thorpej 99: #include <uvm/uvm.h>
100:
1.135 ad 101: POOL_INIT(socket_pool, sizeof(struct socket), 0, 0, 0, "sockpl", NULL,
102: IPL_SOFTNET);
1.77 thorpej 103:
104: MALLOC_DEFINE(M_SOOPTS, "soopts", "socket options");
105: MALLOC_DEFINE(M_SONAME, "soname", "socket name");
1.37 thorpej 106:
1.142 dyoung 107: extern const struct fileops socketops;
108:
1.54 lukem 109: extern int somaxconn; /* patchable (XXX sysctl) */
110: int somaxconn = SOMAXCONN;
1.49 jonathan 111:
1.64 thorpej 112: #ifdef SOSEND_COUNTERS
113: #include <sys/device.h>
114:
1.113 thorpej 115: static struct evcnt sosend_loan_big = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
1.64 thorpej 116: NULL, "sosend", "loan big");
1.113 thorpej 117: static struct evcnt sosend_copy_big = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
1.64 thorpej 118: NULL, "sosend", "copy big");
1.113 thorpej 119: static struct evcnt sosend_copy_small = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
1.64 thorpej 120: NULL, "sosend", "copy small");
1.113 thorpej 121: static struct evcnt sosend_kvalimit = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
1.64 thorpej 122: NULL, "sosend", "kva limit");
123:
124: #define SOSEND_COUNTER_INCR(ev) (ev)->ev_count++
125:
1.101 matt 126: EVCNT_ATTACH_STATIC(sosend_loan_big);
127: EVCNT_ATTACH_STATIC(sosend_copy_big);
128: EVCNT_ATTACH_STATIC(sosend_copy_small);
129: EVCNT_ATTACH_STATIC(sosend_kvalimit);
1.64 thorpej 130: #else
131:
132: #define SOSEND_COUNTER_INCR(ev) /* nothing */
133:
134: #endif /* SOSEND_COUNTERS */
135:
1.119 yamt 136: static struct callback_entry sokva_reclaimerentry;
1.1 cgd 137:
1.71 thorpej 138: #ifdef SOSEND_NO_LOAN
1.121 yamt 139: int sock_loan_thresh = -1;
1.71 thorpej 140: #else
1.121 yamt 141: int sock_loan_thresh = 4096;
1.65 thorpej 142: #endif
1.64 thorpej 143:
1.136 ad 144: static kmutex_t so_pendfree_lock;
1.113 thorpej 145: static struct mbuf *so_pendfree;
1.64 thorpej 146:
1.84 ragge 147: #ifndef SOMAXKVA
148: #define SOMAXKVA (16 * 1024 * 1024)
149: #endif
150: int somaxkva = SOMAXKVA;
1.113 thorpej 151: static int socurkva;
1.136 ad 152: static kcondvar_t socurkva_cv;
1.64 thorpej 153:
154: #define SOCK_LOAN_CHUNK 65536
155:
1.117 yamt 156: static size_t sodopendfree(void);
157: static size_t sodopendfreel(void);
1.93 yamt 158:
1.113 thorpej 159: static vsize_t
1.129 yamt 160: sokvareserve(struct socket *so, vsize_t len)
1.80 yamt 161: {
1.98 christos 162: int error;
1.80 yamt 163:
1.136 ad 164: mutex_enter(&so_pendfree_lock);
1.80 yamt 165: while (socurkva + len > somaxkva) {
1.93 yamt 166: size_t freed;
167:
168: /*
169: * try to do pendfree.
170: */
171:
1.117 yamt 172: freed = sodopendfreel();
1.93 yamt 173:
174: /*
175: * if some kva was freed, try again.
176: */
177:
178: if (freed)
1.80 yamt 179: continue;
1.93 yamt 180:
1.80 yamt 181: SOSEND_COUNTER_INCR(&sosend_kvalimit);
1.136 ad 182: error = cv_wait_sig(&socurkva_cv, &so_pendfree_lock);
1.98 christos 183: if (error) {
184: len = 0;
185: break;
186: }
1.80 yamt 187: }
1.93 yamt 188: socurkva += len;
1.136 ad 189: mutex_exit(&so_pendfree_lock);
1.98 christos 190: return len;
1.95 yamt 191: }
192:
1.113 thorpej 193: static void
1.95 yamt 194: sokvaunreserve(vsize_t len)
195: {
196:
1.136 ad 197: mutex_enter(&so_pendfree_lock);
1.95 yamt 198: socurkva -= len;
1.136 ad 199: cv_broadcast(&socurkva_cv);
200: mutex_exit(&so_pendfree_lock);
1.95 yamt 201: }
202:
203: /*
204: * sokvaalloc: allocate kva for loan.
205: */
206:
207: vaddr_t
208: sokvaalloc(vsize_t len, struct socket *so)
209: {
210: vaddr_t lva;
211:
212: /*
213: * reserve kva.
214: */
215:
1.98 christos 216: if (sokvareserve(so, len) == 0)
217: return 0;
1.93 yamt 218:
219: /*
220: * allocate kva.
221: */
1.80 yamt 222:
1.109 yamt 223: lva = uvm_km_alloc(kernel_map, len, 0, UVM_KMF_VAONLY | UVM_KMF_WAITVA);
1.95 yamt 224: if (lva == 0) {
225: sokvaunreserve(len);
1.80 yamt 226: return (0);
1.95 yamt 227: }
1.80 yamt 228:
229: return lva;
230: }
231:
1.93 yamt 232: /*
233: * sokvafree: free kva for loan.
234: */
235:
1.80 yamt 236: void
237: sokvafree(vaddr_t sva, vsize_t len)
238: {
1.93 yamt 239:
240: /*
241: * free kva.
242: */
1.80 yamt 243:
1.109 yamt 244: uvm_km_free(kernel_map, sva, len, UVM_KMF_VAONLY);
1.93 yamt 245:
246: /*
247: * unreserve kva.
248: */
249:
1.95 yamt 250: sokvaunreserve(len);
1.80 yamt 251: }
252:
1.64 thorpej 253: static void
1.134 christos 254: sodoloanfree(struct vm_page **pgs, void *buf, size_t size)
1.64 thorpej 255: {
1.156 yamt 256: vaddr_t sva, eva;
1.64 thorpej 257: vsize_t len;
1.156 yamt 258: int npgs;
259:
260: KASSERT(pgs != NULL);
1.64 thorpej 261:
262: eva = round_page((vaddr_t) buf + size);
263: sva = trunc_page((vaddr_t) buf);
264: len = eva - sva;
265: npgs = len >> PAGE_SHIFT;
266:
267: pmap_kremove(sva, len);
268: pmap_update(pmap_kernel());
269: uvm_unloan(pgs, npgs, UVM_LOAN_TOPAGE);
1.80 yamt 270: sokvafree(sva, len);
1.64 thorpej 271: }
272:
273: static size_t
1.152 matt 274: sodopendfree(void)
1.64 thorpej 275: {
1.93 yamt 276: size_t rv;
1.64 thorpej 277:
1.136 ad 278: mutex_enter(&so_pendfree_lock);
1.117 yamt 279: rv = sodopendfreel();
1.136 ad 280: mutex_exit(&so_pendfree_lock);
1.93 yamt 281:
282: return rv;
283: }
284:
285: /*
286: * sodopendfreel: free mbufs on "pendfree" list.
1.136 ad 287: * unlock and relock so_pendfree_lock when freeing mbufs.
1.93 yamt 288: *
1.136 ad 289: * => called with so_pendfree_lock held.
1.93 yamt 290: */
291:
292: static size_t
1.152 matt 293: sodopendfreel(void)
1.93 yamt 294: {
1.137 ad 295: struct mbuf *m, *next;
1.93 yamt 296: size_t rv = 0;
297:
1.136 ad 298: KASSERT(mutex_owned(&so_pendfree_lock));
1.64 thorpej 299:
1.137 ad 300: while (so_pendfree != NULL) {
1.64 thorpej 301: m = so_pendfree;
1.93 yamt 302: so_pendfree = NULL;
1.136 ad 303: mutex_exit(&so_pendfree_lock);
1.93 yamt 304:
305: for (; m != NULL; m = next) {
306: next = m->m_next;
1.156 yamt 307: KASSERT((~m->m_flags & (M_EXT|M_EXT_PAGES)) == 0);
308: KASSERT(m->m_ext.ext_refcnt == 0);
1.93 yamt 309:
310: rv += m->m_ext.ext_size;
1.156 yamt 311: sodoloanfree(m->m_ext.ext_pgs, m->m_ext.ext_buf,
1.93 yamt 312: m->m_ext.ext_size);
1.145 ad 313: pool_cache_put(mb_cache, m);
1.93 yamt 314: }
1.64 thorpej 315:
1.136 ad 316: mutex_enter(&so_pendfree_lock);
1.64 thorpej 317: }
318:
319: return (rv);
320: }
321:
1.80 yamt 322: void
1.134 christos 323: soloanfree(struct mbuf *m, void *buf, size_t size, void *arg)
1.64 thorpej 324: {
325:
1.156 yamt 326: KASSERT(m != NULL);
1.64 thorpej 327:
1.93 yamt 328: /*
329: * postpone freeing mbuf.
330: *
331: * we can't do it in interrupt context
332: * because we need to put kva back to kernel_map.
333: */
334:
1.136 ad 335: mutex_enter(&so_pendfree_lock);
1.92 yamt 336: m->m_next = so_pendfree;
337: so_pendfree = m;
1.136 ad 338: cv_broadcast(&socurkva_cv);
339: mutex_exit(&so_pendfree_lock);
1.64 thorpej 340: }
341:
342: static long
343: sosend_loan(struct socket *so, struct uio *uio, struct mbuf *m, long space)
344: {
345: struct iovec *iov = uio->uio_iov;
346: vaddr_t sva, eva;
347: vsize_t len;
1.156 yamt 348: vaddr_t lva;
349: int npgs, error;
350: vaddr_t va;
351: int i;
1.64 thorpej 352:
1.116 yamt 353: if (VMSPACE_IS_KERNEL_P(uio->uio_vmspace))
1.64 thorpej 354: return (0);
355:
356: if (iov->iov_len < (size_t) space)
357: space = iov->iov_len;
358: if (space > SOCK_LOAN_CHUNK)
359: space = SOCK_LOAN_CHUNK;
360:
361: eva = round_page((vaddr_t) iov->iov_base + space);
362: sva = trunc_page((vaddr_t) iov->iov_base);
363: len = eva - sva;
364: npgs = len >> PAGE_SHIFT;
365:
1.79 thorpej 366: /* XXX KDASSERT */
367: KASSERT(npgs <= M_EXT_MAXPAGES);
368:
1.80 yamt 369: lva = sokvaalloc(len, so);
1.64 thorpej 370: if (lva == 0)
1.80 yamt 371: return 0;
1.64 thorpej 372:
1.116 yamt 373: error = uvm_loan(&uio->uio_vmspace->vm_map, sva, len,
1.79 thorpej 374: m->m_ext.ext_pgs, UVM_LOAN_TOPAGE);
1.64 thorpej 375: if (error) {
1.80 yamt 376: sokvafree(lva, len);
1.64 thorpej 377: return (0);
378: }
379:
380: for (i = 0, va = lva; i < npgs; i++, va += PAGE_SIZE)
1.79 thorpej 381: pmap_kenter_pa(va, VM_PAGE_TO_PHYS(m->m_ext.ext_pgs[i]),
382: VM_PROT_READ);
1.64 thorpej 383: pmap_update(pmap_kernel());
384:
385: lva += (vaddr_t) iov->iov_base & PAGE_MASK;
386:
1.134 christos 387: MEXTADD(m, (void *) lva, space, M_MBUF, soloanfree, so);
1.79 thorpej 388: m->m_flags |= M_EXT_PAGES | M_EXT_ROMAP;
1.64 thorpej 389:
390: uio->uio_resid -= space;
391: /* uio_offset not updated, not set/used for write(2) */
1.134 christos 392: uio->uio_iov->iov_base = (char *)uio->uio_iov->iov_base + space;
1.64 thorpej 393: uio->uio_iov->iov_len -= space;
394: if (uio->uio_iov->iov_len == 0) {
395: uio->uio_iov++;
396: uio->uio_iovcnt--;
397: }
398:
399: return (space);
400: }
401:
1.119 yamt 402: static int
1.129 yamt 403: sokva_reclaim_callback(struct callback_entry *ce, void *obj, void *arg)
1.119 yamt 404: {
405:
406: KASSERT(ce == &sokva_reclaimerentry);
407: KASSERT(obj == NULL);
408:
409: sodopendfree();
410: if (!vm_map_starved_p(kernel_map)) {
411: return CALLBACK_CHAIN_ABORT;
412: }
413: return CALLBACK_CHAIN_CONTINUE;
414: }
415:
1.142 dyoung 416: struct mbuf *
1.147 dyoung 417: getsombuf(struct socket *so, int type)
1.142 dyoung 418: {
419: struct mbuf *m;
420:
1.147 dyoung 421: m = m_get(M_WAIT, type);
1.142 dyoung 422: MCLAIM(m, so->so_mowner);
423: return m;
424: }
425:
426: struct mbuf *
427: m_intopt(struct socket *so, int val)
428: {
429: struct mbuf *m;
430:
1.147 dyoung 431: m = getsombuf(so, MT_SOOPTS);
1.142 dyoung 432: m->m_len = sizeof(int);
433: *mtod(m, int *) = val;
434: return m;
435: }
436:
1.119 yamt 437: void
438: soinit(void)
439: {
440:
1.148 ad 441: mutex_init(&so_pendfree_lock, MUTEX_DEFAULT, IPL_VM);
1.136 ad 442: cv_init(&socurkva_cv, "sokva");
443:
1.119 yamt 444: /* Set the initial adjusted socket buffer size. */
445: if (sb_max_set(sb_max))
446: panic("bad initial sb_max value: %lu", sb_max);
447:
448: callback_register(&vm_map_to_kernel(kernel_map)->vmk_reclaim_callback,
449: &sokva_reclaimerentry, NULL, sokva_reclaim_callback);
450: }
451:
1.1 cgd 452: /*
453: * Socket operation routines.
454: * These routines are called by the routines in
455: * sys_socket.c or from a system process, and
456: * implement the semantics of socket operations by
457: * switching out to the protocol specific routines.
458: */
459: /*ARGSUSED*/
1.3 andrew 460: int
1.114 christos 461: socreate(int dom, struct socket **aso, int type, int proto, struct lwp *l)
1.1 cgd 462: {
1.99 matt 463: const struct protosw *prp;
1.54 lukem 464: struct socket *so;
1.115 yamt 465: uid_t uid;
1.54 lukem 466: int error, s;
1.1 cgd 467:
1.132 elad 468: error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_SOCKET,
469: KAUTH_REQ_NETWORK_SOCKET_OPEN, KAUTH_ARG(dom), KAUTH_ARG(type),
470: KAUTH_ARG(proto));
1.140 dyoung 471: if (error != 0)
472: return error;
1.127 elad 473:
1.1 cgd 474: if (proto)
475: prp = pffindproto(dom, proto, type);
476: else
477: prp = pffindtype(dom, type);
1.140 dyoung 478: if (prp == NULL) {
1.120 ginsbach 479: /* no support for domain */
480: if (pffinddomain(dom) == 0)
1.140 dyoung 481: return EAFNOSUPPORT;
1.120 ginsbach 482: /* no support for socket type */
483: if (proto == 0 && type != 0)
1.140 dyoung 484: return EPROTOTYPE;
485: return EPROTONOSUPPORT;
1.120 ginsbach 486: }
1.140 dyoung 487: if (prp->pr_usrreq == NULL)
488: return EPROTONOSUPPORT;
1.1 cgd 489: if (prp->pr_type != type)
1.140 dyoung 490: return EPROTOTYPE;
1.39 matt 491: s = splsoftnet();
1.37 thorpej 492: so = pool_get(&socket_pool, PR_WAITOK);
1.140 dyoung 493: memset(so, 0, sizeof(*so));
1.31 thorpej 494: TAILQ_INIT(&so->so_q0);
495: TAILQ_INIT(&so->so_q);
1.1 cgd 496: so->so_type = type;
497: so->so_proto = prp;
1.33 matt 498: so->so_send = sosend;
499: so->so_receive = soreceive;
1.78 matt 500: #ifdef MBUFTRACE
501: so->so_rcv.sb_mowner = &prp->pr_domain->dom_mowner;
502: so->so_snd.sb_mowner = &prp->pr_domain->dom_mowner;
503: so->so_mowner = &prp->pr_domain->dom_mowner;
504: #endif
1.143 ad 505: selinit(&so->so_rcv.sb_sel);
506: selinit(&so->so_snd.sb_sel);
1.138 rmind 507: uid = kauth_cred_geteuid(l->l_cred);
1.115 yamt 508: so->so_uidinfo = uid_find(uid);
1.140 dyoung 509: error = (*prp->pr_usrreq)(so, PRU_ATTACH, NULL,
510: (struct mbuf *)(long)proto, NULL, l);
511: if (error != 0) {
1.1 cgd 512: so->so_state |= SS_NOFDREF;
513: sofree(so);
1.39 matt 514: splx(s);
1.140 dyoung 515: return error;
1.1 cgd 516: }
1.39 matt 517: splx(s);
1.1 cgd 518: *aso = so;
1.140 dyoung 519: return 0;
1.1 cgd 520: }
521:
1.142 dyoung 522: /* On success, write file descriptor to fdout and return zero. On
523: * failure, return non-zero; *fdout will be undefined.
524: */
525: int
526: fsocreate(int domain, struct socket **sop, int type, int protocol,
527: struct lwp *l, int *fdout)
528: {
529: struct socket *so;
530: struct file *fp;
531: int fd, error;
532:
1.155 ad 533: if ((error = fd_allocfile(&fp, &fd)) != 0)
1.142 dyoung 534: return (error);
535: fp->f_flag = FREAD|FWRITE;
536: fp->f_type = DTYPE_SOCKET;
537: fp->f_ops = &socketops;
538: error = socreate(domain, &so, type, protocol, l);
539: if (error != 0) {
1.155 ad 540: fd_abort(curproc, fp, fd);
1.142 dyoung 541: } else {
542: if (sop != NULL)
543: *sop = so;
544: fp->f_data = so;
1.155 ad 545: fd_affix(curproc, fp, fd);
1.142 dyoung 546: *fdout = fd;
547: }
548: return error;
549: }
550:
1.3 andrew 551: int
1.114 christos 552: sobind(struct socket *so, struct mbuf *nam, struct lwp *l)
1.1 cgd 553: {
1.54 lukem 554: int s, error;
1.1 cgd 555:
1.54 lukem 556: s = splsoftnet();
1.140 dyoung 557: error = (*so->so_proto->pr_usrreq)(so, PRU_BIND, NULL, nam, NULL, l);
1.1 cgd 558: splx(s);
1.140 dyoung 559: return error;
1.1 cgd 560: }
561:
1.3 andrew 562: int
1.150 elad 563: solisten(struct socket *so, int backlog, struct lwp *l)
1.1 cgd 564: {
1.54 lukem 565: int s, error;
1.1 cgd 566:
1.54 lukem 567: s = splsoftnet();
1.158 ! ad 568: if ((so->so_state & (SS_ISCONNECTED | SS_ISCONNECTING |
! 569: SS_ISDISCONNECTING)) != 0)
! 570: return (EOPNOTSUPP);
1.140 dyoung 571: error = (*so->so_proto->pr_usrreq)(so, PRU_LISTEN, NULL,
1.150 elad 572: NULL, NULL, l);
1.140 dyoung 573: if (error != 0) {
1.1 cgd 574: splx(s);
1.140 dyoung 575: return error;
1.1 cgd 576: }
1.63 matt 577: if (TAILQ_EMPTY(&so->so_q))
1.1 cgd 578: so->so_options |= SO_ACCEPTCONN;
579: if (backlog < 0)
580: backlog = 0;
1.49 jonathan 581: so->so_qlimit = min(backlog, somaxconn);
1.1 cgd 582: splx(s);
1.140 dyoung 583: return 0;
1.1 cgd 584: }
585:
1.21 christos 586: void
1.54 lukem 587: sofree(struct socket *so)
1.1 cgd 588: {
589:
1.43 mycroft 590: if (so->so_pcb || (so->so_state & SS_NOFDREF) == 0)
1.1 cgd 591: return;
1.43 mycroft 592: if (so->so_head) {
593: /*
594: * We must not decommission a socket that's on the accept(2)
595: * queue. If we do, then accept(2) may hang after select(2)
596: * indicated that the listening socket was ready.
597: */
598: if (!soqremque(so, 0))
599: return;
600: }
1.98 christos 601: if (so->so_rcv.sb_hiwat)
1.110 christos 602: (void)chgsbsize(so->so_uidinfo, &so->so_rcv.sb_hiwat, 0,
1.98 christos 603: RLIM_INFINITY);
604: if (so->so_snd.sb_hiwat)
1.110 christos 605: (void)chgsbsize(so->so_uidinfo, &so->so_snd.sb_hiwat, 0,
1.98 christos 606: RLIM_INFINITY);
607: sbrelease(&so->so_snd, so);
1.1 cgd 608: sorflush(so);
1.143 ad 609: seldestroy(&so->so_rcv.sb_sel);
610: seldestroy(&so->so_snd.sb_sel);
1.37 thorpej 611: pool_put(&socket_pool, so);
1.1 cgd 612: }
613:
614: /*
615: * Close a socket on last file table reference removal.
616: * Initiate disconnect if connected.
617: * Free socket when disconnect complete.
618: */
1.3 andrew 619: int
1.54 lukem 620: soclose(struct socket *so)
1.1 cgd 621: {
1.54 lukem 622: struct socket *so2;
623: int s, error;
1.1 cgd 624:
1.54 lukem 625: error = 0;
626: s = splsoftnet(); /* conservative */
1.1 cgd 627: if (so->so_options & SO_ACCEPTCONN) {
1.63 matt 628: while ((so2 = TAILQ_FIRST(&so->so_q0)) != 0) {
1.42 mycroft 629: (void) soqremque(so2, 0);
1.41 mycroft 630: (void) soabort(so2);
631: }
1.63 matt 632: while ((so2 = TAILQ_FIRST(&so->so_q)) != 0) {
1.42 mycroft 633: (void) soqremque(so2, 1);
1.41 mycroft 634: (void) soabort(so2);
635: }
1.1 cgd 636: }
637: if (so->so_pcb == 0)
638: goto discard;
639: if (so->so_state & SS_ISCONNECTED) {
640: if ((so->so_state & SS_ISDISCONNECTING) == 0) {
641: error = sodisconnect(so);
642: if (error)
643: goto drop;
644: }
645: if (so->so_options & SO_LINGER) {
1.151 ad 646: if ((so->so_state & SS_ISDISCONNECTING) && so->so_nbio)
1.1 cgd 647: goto drop;
1.21 christos 648: while (so->so_state & SS_ISCONNECTED) {
1.134 christos 649: error = tsleep((void *)&so->so_timeo,
1.21 christos 650: PSOCK | PCATCH, netcls,
1.30 thorpej 651: so->so_linger * hz);
1.21 christos 652: if (error)
1.1 cgd 653: break;
1.21 christos 654: }
1.1 cgd 655: }
656: }
1.54 lukem 657: drop:
1.1 cgd 658: if (so->so_pcb) {
1.22 mycroft 659: int error2 = (*so->so_proto->pr_usrreq)(so, PRU_DETACH,
1.140 dyoung 660: NULL, NULL, NULL, NULL);
1.1 cgd 661: if (error == 0)
662: error = error2;
663: }
1.54 lukem 664: discard:
1.1 cgd 665: if (so->so_state & SS_NOFDREF)
666: panic("soclose: NOFDREF");
667: so->so_state |= SS_NOFDREF;
668: sofree(so);
669: splx(s);
670: return (error);
671: }
672:
673: /*
1.20 mycroft 674: * Must be called at splsoftnet...
1.1 cgd 675: */
1.3 andrew 676: int
1.54 lukem 677: soabort(struct socket *so)
1.1 cgd 678: {
1.139 yamt 679: int error;
1.1 cgd 680:
1.139 yamt 681: KASSERT(so->so_head == NULL);
1.140 dyoung 682: error = (*so->so_proto->pr_usrreq)(so, PRU_ABORT, NULL,
683: NULL, NULL, NULL);
1.139 yamt 684: if (error) {
685: sofree(so);
686: }
687: return error;
1.1 cgd 688: }
689:
1.3 andrew 690: int
1.54 lukem 691: soaccept(struct socket *so, struct mbuf *nam)
1.1 cgd 692: {
1.54 lukem 693: int s, error;
1.1 cgd 694:
1.54 lukem 695: error = 0;
696: s = splsoftnet();
1.1 cgd 697: if ((so->so_state & SS_NOFDREF) == 0)
698: panic("soaccept: !NOFDREF");
699: so->so_state &= ~SS_NOFDREF;
1.55 thorpej 700: if ((so->so_state & SS_ISDISCONNECTED) == 0 ||
701: (so->so_proto->pr_flags & PR_ABRTACPTDIS) == 0)
1.41 mycroft 702: error = (*so->so_proto->pr_usrreq)(so, PRU_ACCEPT,
1.140 dyoung 703: NULL, nam, NULL, NULL);
1.41 mycroft 704: else
1.53 itojun 705: error = ECONNABORTED;
1.52 itojun 706:
1.1 cgd 707: splx(s);
708: return (error);
709: }
710:
1.3 andrew 711: int
1.114 christos 712: soconnect(struct socket *so, struct mbuf *nam, struct lwp *l)
1.1 cgd 713: {
1.54 lukem 714: int s, error;
1.1 cgd 715:
716: if (so->so_options & SO_ACCEPTCONN)
717: return (EOPNOTSUPP);
1.20 mycroft 718: s = splsoftnet();
1.1 cgd 719: /*
720: * If protocol is connection-based, can only connect once.
721: * Otherwise, if connected, try to disconnect first.
722: * This allows user to disconnect by connecting to, e.g.,
723: * a null address.
724: */
725: if (so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING) &&
726: ((so->so_proto->pr_flags & PR_CONNREQUIRED) ||
727: (error = sodisconnect(so))))
728: error = EISCONN;
729: else
730: error = (*so->so_proto->pr_usrreq)(so, PRU_CONNECT,
1.140 dyoung 731: NULL, nam, NULL, l);
1.1 cgd 732: splx(s);
733: return (error);
734: }
735:
1.3 andrew 736: int
1.54 lukem 737: soconnect2(struct socket *so1, struct socket *so2)
1.1 cgd 738: {
1.54 lukem 739: int s, error;
1.1 cgd 740:
1.54 lukem 741: s = splsoftnet();
1.22 mycroft 742: error = (*so1->so_proto->pr_usrreq)(so1, PRU_CONNECT2,
1.140 dyoung 743: NULL, (struct mbuf *)so2, NULL, NULL);
1.1 cgd 744: splx(s);
745: return (error);
746: }
747:
1.3 andrew 748: int
1.54 lukem 749: sodisconnect(struct socket *so)
1.1 cgd 750: {
1.54 lukem 751: int s, error;
1.1 cgd 752:
1.54 lukem 753: s = splsoftnet();
1.1 cgd 754: if ((so->so_state & SS_ISCONNECTED) == 0) {
755: error = ENOTCONN;
756: goto bad;
757: }
758: if (so->so_state & SS_ISDISCONNECTING) {
759: error = EALREADY;
760: goto bad;
761: }
1.22 mycroft 762: error = (*so->so_proto->pr_usrreq)(so, PRU_DISCONNECT,
1.140 dyoung 763: NULL, NULL, NULL, NULL);
1.54 lukem 764: bad:
1.1 cgd 765: splx(s);
1.117 yamt 766: sodopendfree();
1.1 cgd 767: return (error);
768: }
769:
1.15 mycroft 770: #define SBLOCKWAIT(f) (((f) & MSG_DONTWAIT) ? M_NOWAIT : M_WAITOK)
1.1 cgd 771: /*
772: * Send on a socket.
773: * If send must go all at once and message is larger than
774: * send buffering, then hard error.
775: * Lock against other senders.
776: * If must go all at once and not enough room now, then
777: * inform user that this would block and do nothing.
778: * Otherwise, if nonblocking, send as much as possible.
779: * The data to be sent is described by "uio" if nonzero,
780: * otherwise by the mbuf chain "top" (which must be null
781: * if uio is not). Data provided in mbuf chain must be small
782: * enough to send all at once.
783: *
784: * Returns nonzero on error, timeout or signal; callers
785: * must check for short counts if EINTR/ERESTART are returned.
786: * Data and control buffers are freed on return.
787: */
1.3 andrew 788: int
1.54 lukem 789: sosend(struct socket *so, struct mbuf *addr, struct uio *uio, struct mbuf *top,
1.114 christos 790: struct mbuf *control, int flags, struct lwp *l)
1.1 cgd 791: {
1.54 lukem 792: struct mbuf **mp, *m;
1.114 christos 793: struct proc *p;
1.58 jdolecek 794: long space, len, resid, clen, mlen;
795: int error, s, dontroute, atomic;
1.54 lukem 796:
1.114 christos 797: p = l->l_proc;
1.117 yamt 798: sodopendfree();
1.64 thorpej 799:
1.54 lukem 800: clen = 0;
801: atomic = sosendallatonce(so) || top;
1.1 cgd 802: if (uio)
803: resid = uio->uio_resid;
804: else
805: resid = top->m_pkthdr.len;
1.7 cgd 806: /*
807: * In theory resid should be unsigned.
808: * However, space must be signed, as it might be less than 0
809: * if we over-committed, and we must use a signed comparison
810: * of space and resid. On the other hand, a negative resid
811: * causes us to loop sending 0-length segments to the protocol.
812: */
1.29 mycroft 813: if (resid < 0) {
814: error = EINVAL;
815: goto out;
816: }
1.1 cgd 817: dontroute =
818: (flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 &&
819: (so->so_proto->pr_flags & PR_ATOMIC);
1.157 ad 820: if (l)
821: l->l_ru.ru_msgsnd++;
1.1 cgd 822: if (control)
823: clen = control->m_len;
824: #define snderr(errno) { error = errno; splx(s); goto release; }
825:
1.54 lukem 826: restart:
1.21 christos 827: if ((error = sblock(&so->so_snd, SBLOCKWAIT(flags))) != 0)
1.1 cgd 828: goto out;
829: do {
1.20 mycroft 830: s = splsoftnet();
1.1 cgd 831: if (so->so_state & SS_CANTSENDMORE)
832: snderr(EPIPE);
1.48 thorpej 833: if (so->so_error) {
834: error = so->so_error;
835: so->so_error = 0;
836: splx(s);
837: goto release;
838: }
1.1 cgd 839: if ((so->so_state & SS_ISCONNECTED) == 0) {
840: if (so->so_proto->pr_flags & PR_CONNREQUIRED) {
841: if ((so->so_state & SS_ISCONFIRMING) == 0 &&
842: !(resid == 0 && clen != 0))
843: snderr(ENOTCONN);
844: } else if (addr == 0)
845: snderr(EDESTADDRREQ);
846: }
847: space = sbspace(&so->so_snd);
848: if (flags & MSG_OOB)
849: space += 1024;
1.21 christos 850: if ((atomic && resid > so->so_snd.sb_hiwat) ||
1.11 mycroft 851: clen > so->so_snd.sb_hiwat)
852: snderr(EMSGSIZE);
1.96 mycroft 853: if (space < resid + clen &&
1.1 cgd 854: (atomic || space < so->so_snd.sb_lowat || space < clen)) {
1.151 ad 855: if (so->so_nbio)
1.1 cgd 856: snderr(EWOULDBLOCK);
857: sbunlock(&so->so_snd);
858: error = sbwait(&so->so_snd);
859: splx(s);
860: if (error)
861: goto out;
862: goto restart;
863: }
864: splx(s);
865: mp = ⊤
866: space -= clen;
867: do {
1.45 tv 868: if (uio == NULL) {
869: /*
870: * Data is prepackaged in "top".
871: */
872: resid = 0;
873: if (flags & MSG_EOR)
874: top->m_flags |= M_EOR;
875: } else do {
1.144 dyoung 876: if (top == NULL) {
1.78 matt 877: m = m_gethdr(M_WAIT, MT_DATA);
1.45 tv 878: mlen = MHLEN;
879: m->m_pkthdr.len = 0;
1.140 dyoung 880: m->m_pkthdr.rcvif = NULL;
1.45 tv 881: } else {
1.78 matt 882: m = m_get(M_WAIT, MT_DATA);
1.45 tv 883: mlen = MLEN;
884: }
1.78 matt 885: MCLAIM(m, so->so_snd.sb_mowner);
1.121 yamt 886: if (sock_loan_thresh >= 0 &&
887: uio->uio_iov->iov_len >= sock_loan_thresh &&
888: space >= sock_loan_thresh &&
1.64 thorpej 889: (len = sosend_loan(so, uio, m,
890: space)) != 0) {
891: SOSEND_COUNTER_INCR(&sosend_loan_big);
892: space -= len;
893: goto have_data;
894: }
1.45 tv 895: if (resid >= MINCLSIZE && space >= MCLBYTES) {
1.64 thorpej 896: SOSEND_COUNTER_INCR(&sosend_copy_big);
1.78 matt 897: m_clget(m, M_WAIT);
1.45 tv 898: if ((m->m_flags & M_EXT) == 0)
899: goto nopages;
900: mlen = MCLBYTES;
901: if (atomic && top == 0) {
1.58 jdolecek 902: len = lmin(MCLBYTES - max_hdr,
1.54 lukem 903: resid);
1.45 tv 904: m->m_data += max_hdr;
905: } else
1.58 jdolecek 906: len = lmin(MCLBYTES, resid);
1.45 tv 907: space -= len;
908: } else {
1.64 thorpej 909: nopages:
910: SOSEND_COUNTER_INCR(&sosend_copy_small);
1.58 jdolecek 911: len = lmin(lmin(mlen, resid), space);
1.45 tv 912: space -= len;
913: /*
914: * For datagram protocols, leave room
915: * for protocol headers in first mbuf.
916: */
917: if (atomic && top == 0 && len < mlen)
918: MH_ALIGN(m, len);
919: }
1.144 dyoung 920: error = uiomove(mtod(m, void *), (int)len, uio);
1.64 thorpej 921: have_data:
1.45 tv 922: resid = uio->uio_resid;
923: m->m_len = len;
924: *mp = m;
925: top->m_pkthdr.len += len;
1.144 dyoung 926: if (error != 0)
1.45 tv 927: goto release;
928: mp = &m->m_next;
929: if (resid <= 0) {
930: if (flags & MSG_EOR)
931: top->m_flags |= M_EOR;
932: break;
933: }
934: } while (space > 0 && atomic);
1.108 perry 935:
1.46 sommerfe 936: s = splsoftnet();
937:
938: if (so->so_state & SS_CANTSENDMORE)
939: snderr(EPIPE);
1.45 tv 940:
941: if (dontroute)
942: so->so_options |= SO_DONTROUTE;
943: if (resid > 0)
944: so->so_state |= SS_MORETOCOME;
1.46 sommerfe 945: error = (*so->so_proto->pr_usrreq)(so,
946: (flags & MSG_OOB) ? PRU_SENDOOB : PRU_SEND,
1.114 christos 947: top, addr, control, curlwp); /* XXX */
1.45 tv 948: if (dontroute)
949: so->so_options &= ~SO_DONTROUTE;
950: if (resid > 0)
951: so->so_state &= ~SS_MORETOCOME;
1.46 sommerfe 952: splx(s);
953:
1.45 tv 954: clen = 0;
1.144 dyoung 955: control = NULL;
956: top = NULL;
1.45 tv 957: mp = ⊤
1.144 dyoung 958: if (error != 0)
1.1 cgd 959: goto release;
960: } while (resid && space > 0);
961: } while (resid);
962:
1.54 lukem 963: release:
1.1 cgd 964: sbunlock(&so->so_snd);
1.54 lukem 965: out:
1.1 cgd 966: if (top)
967: m_freem(top);
968: if (control)
969: m_freem(control);
970: return (error);
971: }
972:
973: /*
974: * Implement receive operations on a socket.
975: * We depend on the way that records are added to the sockbuf
976: * by sbappend*. In particular, each record (mbufs linked through m_next)
977: * must begin with an address if the protocol so specifies,
978: * followed by an optional mbuf or mbufs containing ancillary data,
979: * and then zero or more mbufs of data.
980: * In order to avoid blocking network interrupts for the entire time here,
981: * we splx() while doing the actual copy to user space.
982: * Although the sockbuf is locked, new data may still be appended,
983: * and thus we must maintain consistency of the sockbuf during that time.
984: *
985: * The caller may receive the data as a single mbuf chain by supplying
986: * an mbuf **mp0 for use in returning the chain. The uio is then used
987: * only for the count in uio_resid.
988: */
1.3 andrew 989: int
1.54 lukem 990: soreceive(struct socket *so, struct mbuf **paddr, struct uio *uio,
991: struct mbuf **mp0, struct mbuf **controlp, int *flagsp)
1.1 cgd 992: {
1.116 yamt 993: struct lwp *l = curlwp;
1.54 lukem 994: struct mbuf *m, **mp;
1.146 dyoung 995: int atomic, flags, len, error, s, offset, moff, type, orig_resid;
1.99 matt 996: const struct protosw *pr;
1.54 lukem 997: struct mbuf *nextrecord;
1.67 he 998: int mbuf_removed = 0;
1.146 dyoung 999: const struct domain *dom;
1.64 thorpej 1000:
1.54 lukem 1001: pr = so->so_proto;
1.146 dyoung 1002: atomic = pr->pr_flags & PR_ATOMIC;
1003: dom = pr->pr_domain;
1.1 cgd 1004: mp = mp0;
1.54 lukem 1005: type = 0;
1006: orig_resid = uio->uio_resid;
1.102 jonathan 1007:
1.144 dyoung 1008: if (paddr != NULL)
1009: *paddr = NULL;
1010: if (controlp != NULL)
1011: *controlp = NULL;
1012: if (flagsp != NULL)
1.1 cgd 1013: flags = *flagsp &~ MSG_EOR;
1014: else
1015: flags = 0;
1.66 enami 1016:
1017: if ((flags & MSG_DONTWAIT) == 0)
1.117 yamt 1018: sodopendfree();
1.66 enami 1019:
1.1 cgd 1020: if (flags & MSG_OOB) {
1021: m = m_get(M_WAIT, MT_DATA);
1.17 cgd 1022: error = (*pr->pr_usrreq)(so, PRU_RCVOOB, m,
1.140 dyoung 1023: (struct mbuf *)(long)(flags & MSG_PEEK), NULL, l);
1.1 cgd 1024: if (error)
1025: goto bad;
1026: do {
1.134 christos 1027: error = uiomove(mtod(m, void *),
1.1 cgd 1028: (int) min(uio->uio_resid, m->m_len), uio);
1029: m = m_free(m);
1.144 dyoung 1030: } while (uio->uio_resid > 0 && error == 0 && m);
1.54 lukem 1031: bad:
1.144 dyoung 1032: if (m != NULL)
1.1 cgd 1033: m_freem(m);
1.144 dyoung 1034: return error;
1.1 cgd 1035: }
1.144 dyoung 1036: if (mp != NULL)
1.140 dyoung 1037: *mp = NULL;
1.1 cgd 1038: if (so->so_state & SS_ISCONFIRMING && uio->uio_resid)
1.140 dyoung 1039: (*pr->pr_usrreq)(so, PRU_RCVD, NULL, NULL, NULL, l);
1.1 cgd 1040:
1.54 lukem 1041: restart:
1.21 christos 1042: if ((error = sblock(&so->so_rcv, SBLOCKWAIT(flags))) != 0)
1.144 dyoung 1043: return error;
1.20 mycroft 1044: s = splsoftnet();
1.1 cgd 1045:
1046: m = so->so_rcv.sb_mb;
1047: /*
1048: * If we have less data than requested, block awaiting more
1049: * (subject to any timeout) if:
1.15 mycroft 1050: * 1. the current count is less than the low water mark,
1.1 cgd 1051: * 2. MSG_WAITALL is set, and it is possible to do the entire
1.15 mycroft 1052: * receive operation at once if we block (resid <= hiwat), or
1053: * 3. MSG_DONTWAIT is not set.
1.1 cgd 1054: * If MSG_WAITALL is set but resid is larger than the receive buffer,
1055: * we have to do the receive in sections, and thus risk returning
1056: * a short count if a timeout or signal occurs after we start.
1057: */
1.144 dyoung 1058: if (m == NULL ||
1059: ((flags & MSG_DONTWAIT) == 0 &&
1060: so->so_rcv.sb_cc < uio->uio_resid &&
1061: (so->so_rcv.sb_cc < so->so_rcv.sb_lowat ||
1062: ((flags & MSG_WAITALL) &&
1063: uio->uio_resid <= so->so_rcv.sb_hiwat)) &&
1.146 dyoung 1064: m->m_nextpkt == NULL && !atomic)) {
1.1 cgd 1065: #ifdef DIAGNOSTIC
1.144 dyoung 1066: if (m == NULL && so->so_rcv.sb_cc)
1.1 cgd 1067: panic("receive 1");
1068: #endif
1069: if (so->so_error) {
1.144 dyoung 1070: if (m != NULL)
1.15 mycroft 1071: goto dontblock;
1.1 cgd 1072: error = so->so_error;
1073: if ((flags & MSG_PEEK) == 0)
1074: so->so_error = 0;
1075: goto release;
1076: }
1077: if (so->so_state & SS_CANTRCVMORE) {
1.144 dyoung 1078: if (m != NULL)
1.15 mycroft 1079: goto dontblock;
1.1 cgd 1080: else
1081: goto release;
1082: }
1.144 dyoung 1083: for (; m != NULL; m = m->m_next)
1.1 cgd 1084: if (m->m_type == MT_OOBDATA || (m->m_flags & M_EOR)) {
1085: m = so->so_rcv.sb_mb;
1086: goto dontblock;
1087: }
1088: if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0 &&
1089: (so->so_proto->pr_flags & PR_CONNREQUIRED)) {
1090: error = ENOTCONN;
1091: goto release;
1092: }
1093: if (uio->uio_resid == 0)
1094: goto release;
1.151 ad 1095: if (so->so_nbio || (flags & MSG_DONTWAIT)) {
1.1 cgd 1096: error = EWOULDBLOCK;
1097: goto release;
1098: }
1.69 thorpej 1099: SBLASTRECORDCHK(&so->so_rcv, "soreceive sbwait 1");
1100: SBLASTMBUFCHK(&so->so_rcv, "soreceive sbwait 1");
1.1 cgd 1101: sbunlock(&so->so_rcv);
1102: error = sbwait(&so->so_rcv);
1103: splx(s);
1.144 dyoung 1104: if (error != 0)
1105: return error;
1.1 cgd 1106: goto restart;
1107: }
1.54 lukem 1108: dontblock:
1.69 thorpej 1109: /*
1110: * On entry here, m points to the first record of the socket buffer.
1111: * While we process the initial mbufs containing address and control
1112: * info, we save a copy of m->m_nextpkt into nextrecord.
1113: */
1.144 dyoung 1114: if (l != NULL)
1.157 ad 1115: l->l_ru.ru_msgrcv++;
1.69 thorpej 1116: KASSERT(m == so->so_rcv.sb_mb);
1117: SBLASTRECORDCHK(&so->so_rcv, "soreceive 1");
1118: SBLASTMBUFCHK(&so->so_rcv, "soreceive 1");
1.1 cgd 1119: nextrecord = m->m_nextpkt;
1120: if (pr->pr_flags & PR_ADDR) {
1121: #ifdef DIAGNOSTIC
1122: if (m->m_type != MT_SONAME)
1123: panic("receive 1a");
1124: #endif
1.3 andrew 1125: orig_resid = 0;
1.1 cgd 1126: if (flags & MSG_PEEK) {
1127: if (paddr)
1128: *paddr = m_copy(m, 0, m->m_len);
1129: m = m->m_next;
1130: } else {
1131: sbfree(&so->so_rcv, m);
1.67 he 1132: mbuf_removed = 1;
1.144 dyoung 1133: if (paddr != NULL) {
1.1 cgd 1134: *paddr = m;
1135: so->so_rcv.sb_mb = m->m_next;
1.144 dyoung 1136: m->m_next = NULL;
1.1 cgd 1137: m = so->so_rcv.sb_mb;
1138: } else {
1139: MFREE(m, so->so_rcv.sb_mb);
1140: m = so->so_rcv.sb_mb;
1141: }
1142: }
1143: }
1.144 dyoung 1144: while (m != NULL && m->m_type == MT_CONTROL && error == 0) {
1.1 cgd 1145: if (flags & MSG_PEEK) {
1.144 dyoung 1146: if (controlp != NULL)
1.1 cgd 1147: *controlp = m_copy(m, 0, m->m_len);
1148: m = m->m_next;
1149: } else {
1150: sbfree(&so->so_rcv, m);
1.67 he 1151: mbuf_removed = 1;
1.144 dyoung 1152: if (controlp != NULL) {
1.114 christos 1153: if (dom->dom_externalize && l &&
1.1 cgd 1154: mtod(m, struct cmsghdr *)->cmsg_type ==
1155: SCM_RIGHTS)
1.114 christos 1156: error = (*dom->dom_externalize)(m, l);
1.1 cgd 1157: *controlp = m;
1158: so->so_rcv.sb_mb = m->m_next;
1.144 dyoung 1159: m->m_next = NULL;
1.1 cgd 1160: m = so->so_rcv.sb_mb;
1161: } else {
1.106 itojun 1162: /*
1163: * Dispose of any SCM_RIGHTS message that went
1164: * through the read path rather than recv.
1165: */
1.146 dyoung 1166: if (dom->dom_dispose &&
1.106 itojun 1167: mtod(m, struct cmsghdr *)->cmsg_type == SCM_RIGHTS)
1.146 dyoung 1168: (*dom->dom_dispose)(m);
1.1 cgd 1169: MFREE(m, so->so_rcv.sb_mb);
1170: m = so->so_rcv.sb_mb;
1171: }
1172: }
1.144 dyoung 1173: if (controlp != NULL) {
1.3 andrew 1174: orig_resid = 0;
1.1 cgd 1175: controlp = &(*controlp)->m_next;
1.3 andrew 1176: }
1.1 cgd 1177: }
1.69 thorpej 1178:
1179: /*
1180: * If m is non-NULL, we have some data to read. From now on,
1181: * make sure to keep sb_lastrecord consistent when working on
1182: * the last packet on the chain (nextrecord == NULL) and we
1183: * change m->m_nextpkt.
1184: */
1.144 dyoung 1185: if (m != NULL) {
1.69 thorpej 1186: if ((flags & MSG_PEEK) == 0) {
1.1 cgd 1187: m->m_nextpkt = nextrecord;
1.69 thorpej 1188: /*
1189: * If nextrecord == NULL (this is a single chain),
1190: * then sb_lastrecord may not be valid here if m
1191: * was changed earlier.
1192: */
1193: if (nextrecord == NULL) {
1194: KASSERT(so->so_rcv.sb_mb == m);
1195: so->so_rcv.sb_lastrecord = m;
1196: }
1197: }
1.1 cgd 1198: type = m->m_type;
1199: if (type == MT_OOBDATA)
1200: flags |= MSG_OOB;
1.69 thorpej 1201: } else {
1202: if ((flags & MSG_PEEK) == 0) {
1203: KASSERT(so->so_rcv.sb_mb == m);
1204: so->so_rcv.sb_mb = nextrecord;
1.70 thorpej 1205: SB_EMPTY_FIXUP(&so->so_rcv);
1.69 thorpej 1206: }
1.1 cgd 1207: }
1.69 thorpej 1208: SBLASTRECORDCHK(&so->so_rcv, "soreceive 2");
1209: SBLASTMBUFCHK(&so->so_rcv, "soreceive 2");
1210:
1.1 cgd 1211: moff = 0;
1212: offset = 0;
1.144 dyoung 1213: while (m != NULL && uio->uio_resid > 0 && error == 0) {
1.1 cgd 1214: if (m->m_type == MT_OOBDATA) {
1215: if (type != MT_OOBDATA)
1216: break;
1217: } else if (type == MT_OOBDATA)
1218: break;
1219: #ifdef DIAGNOSTIC
1220: else if (m->m_type != MT_DATA && m->m_type != MT_HEADER)
1221: panic("receive 3");
1222: #endif
1223: so->so_state &= ~SS_RCVATMARK;
1224: len = uio->uio_resid;
1225: if (so->so_oobmark && len > so->so_oobmark - offset)
1226: len = so->so_oobmark - offset;
1227: if (len > m->m_len - moff)
1228: len = m->m_len - moff;
1229: /*
1230: * If mp is set, just pass back the mbufs.
1231: * Otherwise copy them out via the uio, then free.
1232: * Sockbuf must be consistent here (points to current mbuf,
1233: * it points to next record) when we drop priority;
1234: * we must note any additions to the sockbuf when we
1235: * block interrupts again.
1236: */
1.144 dyoung 1237: if (mp == NULL) {
1.69 thorpej 1238: SBLASTRECORDCHK(&so->so_rcv, "soreceive uiomove");
1239: SBLASTMBUFCHK(&so->so_rcv, "soreceive uiomove");
1.1 cgd 1240: splx(s);
1.134 christos 1241: error = uiomove(mtod(m, char *) + moff, (int)len, uio);
1.20 mycroft 1242: s = splsoftnet();
1.144 dyoung 1243: if (error != 0) {
1.67 he 1244: /*
1245: * If any part of the record has been removed
1246: * (such as the MT_SONAME mbuf, which will
1247: * happen when PR_ADDR, and thus also
1248: * PR_ATOMIC, is set), then drop the entire
1249: * record to maintain the atomicity of the
1250: * receive operation.
1251: *
1252: * This avoids a later panic("receive 1a")
1253: * when compiled with DIAGNOSTIC.
1254: */
1.146 dyoung 1255: if (m && mbuf_removed && atomic)
1.67 he 1256: (void) sbdroprecord(&so->so_rcv);
1257:
1.57 jdolecek 1258: goto release;
1.67 he 1259: }
1.1 cgd 1260: } else
1261: uio->uio_resid -= len;
1262: if (len == m->m_len - moff) {
1263: if (m->m_flags & M_EOR)
1264: flags |= MSG_EOR;
1265: if (flags & MSG_PEEK) {
1266: m = m->m_next;
1267: moff = 0;
1268: } else {
1269: nextrecord = m->m_nextpkt;
1270: sbfree(&so->so_rcv, m);
1271: if (mp) {
1272: *mp = m;
1273: mp = &m->m_next;
1274: so->so_rcv.sb_mb = m = m->m_next;
1.140 dyoung 1275: *mp = NULL;
1.1 cgd 1276: } else {
1277: MFREE(m, so->so_rcv.sb_mb);
1278: m = so->so_rcv.sb_mb;
1279: }
1.69 thorpej 1280: /*
1281: * If m != NULL, we also know that
1282: * so->so_rcv.sb_mb != NULL.
1283: */
1284: KASSERT(so->so_rcv.sb_mb == m);
1285: if (m) {
1.1 cgd 1286: m->m_nextpkt = nextrecord;
1.69 thorpej 1287: if (nextrecord == NULL)
1288: so->so_rcv.sb_lastrecord = m;
1289: } else {
1290: so->so_rcv.sb_mb = nextrecord;
1.70 thorpej 1291: SB_EMPTY_FIXUP(&so->so_rcv);
1.69 thorpej 1292: }
1293: SBLASTRECORDCHK(&so->so_rcv, "soreceive 3");
1294: SBLASTMBUFCHK(&so->so_rcv, "soreceive 3");
1.1 cgd 1295: }
1.144 dyoung 1296: } else if (flags & MSG_PEEK)
1297: moff += len;
1298: else {
1299: if (mp != NULL)
1300: *mp = m_copym(m, 0, len, M_WAIT);
1301: m->m_data += len;
1302: m->m_len -= len;
1303: so->so_rcv.sb_cc -= len;
1.1 cgd 1304: }
1305: if (so->so_oobmark) {
1306: if ((flags & MSG_PEEK) == 0) {
1307: so->so_oobmark -= len;
1308: if (so->so_oobmark == 0) {
1309: so->so_state |= SS_RCVATMARK;
1310: break;
1311: }
1.7 cgd 1312: } else {
1.1 cgd 1313: offset += len;
1.7 cgd 1314: if (offset == so->so_oobmark)
1315: break;
1316: }
1.1 cgd 1317: }
1318: if (flags & MSG_EOR)
1319: break;
1320: /*
1321: * If the MSG_WAITALL flag is set (for non-atomic socket),
1322: * we must not quit until "uio->uio_resid == 0" or an error
1323: * termination. If a signal/timeout occurs, return
1324: * with a short count but without error.
1325: * Keep sockbuf locked against other readers.
1326: */
1.144 dyoung 1327: while (flags & MSG_WAITALL && m == NULL && uio->uio_resid > 0 &&
1.3 andrew 1328: !sosendallatonce(so) && !nextrecord) {
1.1 cgd 1329: if (so->so_error || so->so_state & SS_CANTRCVMORE)
1330: break;
1.68 matt 1331: /*
1332: * If we are peeking and the socket receive buffer is
1333: * full, stop since we can't get more data to peek at.
1334: */
1335: if ((flags & MSG_PEEK) && sbspace(&so->so_rcv) <= 0)
1336: break;
1337: /*
1338: * If we've drained the socket buffer, tell the
1339: * protocol in case it needs to do something to
1340: * get it filled again.
1341: */
1342: if ((pr->pr_flags & PR_WANTRCVD) && so->so_pcb)
1343: (*pr->pr_usrreq)(so, PRU_RCVD,
1.140 dyoung 1344: NULL, (struct mbuf *)(long)flags, NULL, l);
1.69 thorpej 1345: SBLASTRECORDCHK(&so->so_rcv, "soreceive sbwait 2");
1346: SBLASTMBUFCHK(&so->so_rcv, "soreceive sbwait 2");
1.1 cgd 1347: error = sbwait(&so->so_rcv);
1.144 dyoung 1348: if (error != 0) {
1.1 cgd 1349: sbunlock(&so->so_rcv);
1350: splx(s);
1.144 dyoung 1351: return 0;
1.1 cgd 1352: }
1.21 christos 1353: if ((m = so->so_rcv.sb_mb) != NULL)
1.1 cgd 1354: nextrecord = m->m_nextpkt;
1355: }
1356: }
1.3 andrew 1357:
1.146 dyoung 1358: if (m && atomic) {
1.3 andrew 1359: flags |= MSG_TRUNC;
1360: if ((flags & MSG_PEEK) == 0)
1361: (void) sbdroprecord(&so->so_rcv);
1362: }
1.1 cgd 1363: if ((flags & MSG_PEEK) == 0) {
1.144 dyoung 1364: if (m == NULL) {
1.69 thorpej 1365: /*
1.70 thorpej 1366: * First part is an inline SB_EMPTY_FIXUP(). Second
1.69 thorpej 1367: * part makes sure sb_lastrecord is up-to-date if
1368: * there is still data in the socket buffer.
1369: */
1.1 cgd 1370: so->so_rcv.sb_mb = nextrecord;
1.69 thorpej 1371: if (so->so_rcv.sb_mb == NULL) {
1372: so->so_rcv.sb_mbtail = NULL;
1373: so->so_rcv.sb_lastrecord = NULL;
1374: } else if (nextrecord->m_nextpkt == NULL)
1375: so->so_rcv.sb_lastrecord = nextrecord;
1376: }
1377: SBLASTRECORDCHK(&so->so_rcv, "soreceive 4");
1378: SBLASTMBUFCHK(&so->so_rcv, "soreceive 4");
1.1 cgd 1379: if (pr->pr_flags & PR_WANTRCVD && so->so_pcb)
1.140 dyoung 1380: (*pr->pr_usrreq)(so, PRU_RCVD, NULL,
1381: (struct mbuf *)(long)flags, NULL, l);
1.1 cgd 1382: }
1.3 andrew 1383: if (orig_resid == uio->uio_resid && orig_resid &&
1384: (flags & MSG_EOR) == 0 && (so->so_state & SS_CANTRCVMORE) == 0) {
1385: sbunlock(&so->so_rcv);
1386: splx(s);
1387: goto restart;
1388: }
1.108 perry 1389:
1.144 dyoung 1390: if (flagsp != NULL)
1.1 cgd 1391: *flagsp |= flags;
1.54 lukem 1392: release:
1.1 cgd 1393: sbunlock(&so->so_rcv);
1394: splx(s);
1.144 dyoung 1395: return error;
1.1 cgd 1396: }
1397:
1.14 mycroft 1398: int
1.54 lukem 1399: soshutdown(struct socket *so, int how)
1.1 cgd 1400: {
1.99 matt 1401: const struct protosw *pr;
1.34 kleink 1402:
1.54 lukem 1403: pr = so->so_proto;
1.34 kleink 1404: if (!(how == SHUT_RD || how == SHUT_WR || how == SHUT_RDWR))
1405: return (EINVAL);
1.1 cgd 1406:
1.34 kleink 1407: if (how == SHUT_RD || how == SHUT_RDWR)
1.1 cgd 1408: sorflush(so);
1.34 kleink 1409: if (how == SHUT_WR || how == SHUT_RDWR)
1.140 dyoung 1410: return (*pr->pr_usrreq)(so, PRU_SHUTDOWN, NULL,
1411: NULL, NULL, NULL);
1.144 dyoung 1412: return 0;
1.1 cgd 1413: }
1414:
1.14 mycroft 1415: void
1.54 lukem 1416: sorflush(struct socket *so)
1.1 cgd 1417: {
1.54 lukem 1418: struct sockbuf *sb, asb;
1.99 matt 1419: const struct protosw *pr;
1.54 lukem 1420: int s;
1.1 cgd 1421:
1.54 lukem 1422: sb = &so->so_rcv;
1423: pr = so->so_proto;
1.1 cgd 1424: sb->sb_flags |= SB_NOINTR;
1.15 mycroft 1425: (void) sblock(sb, M_WAITOK);
1.56 thorpej 1426: s = splnet();
1.1 cgd 1427: socantrcvmore(so);
1428: sbunlock(sb);
1429: asb = *sb;
1.86 wrstuden 1430: /*
1431: * Clear most of the sockbuf structure, but leave some of the
1432: * fields valid.
1433: */
1434: memset(&sb->sb_startzero, 0,
1435: sizeof(*sb) - offsetof(struct sockbuf, sb_startzero));
1.1 cgd 1436: splx(s);
1437: if (pr->pr_flags & PR_RIGHTS && pr->pr_domain->dom_dispose)
1438: (*pr->pr_domain->dom_dispose)(asb.sb_mb);
1.98 christos 1439: sbrelease(&asb, so);
1.1 cgd 1440: }
1441:
1.142 dyoung 1442: static int
1443: sosetopt1(struct socket *so, int level, int optname, struct mbuf *m)
1.1 cgd 1444: {
1.142 dyoung 1445: int optval, val;
1.130 christos 1446: struct linger *l;
1.141 yamt 1447: struct sockbuf *sb;
1.142 dyoung 1448: struct timeval *tv;
1449:
1450: switch (optname) {
1451:
1452: case SO_LINGER:
1453: if (m == NULL || m->m_len != sizeof(struct linger))
1454: return EINVAL;
1455: l = mtod(m, struct linger *);
1456: if (l->l_linger < 0 || l->l_linger > USHRT_MAX ||
1457: l->l_linger > (INT_MAX / hz))
1458: return EDOM;
1459: so->so_linger = l->l_linger;
1460: if (l->l_onoff)
1461: so->so_options |= SO_LINGER;
1462: else
1463: so->so_options &= ~SO_LINGER;
1464: break;
1.1 cgd 1465:
1.142 dyoung 1466: case SO_DEBUG:
1467: case SO_KEEPALIVE:
1468: case SO_DONTROUTE:
1469: case SO_USELOOPBACK:
1470: case SO_BROADCAST:
1471: case SO_REUSEADDR:
1472: case SO_REUSEPORT:
1473: case SO_OOBINLINE:
1474: case SO_TIMESTAMP:
1475: if (m == NULL || m->m_len < sizeof(int))
1476: return EINVAL;
1477: if (*mtod(m, int *))
1478: so->so_options |= optname;
1479: else
1480: so->so_options &= ~optname;
1481: break;
1482:
1483: case SO_SNDBUF:
1484: case SO_RCVBUF:
1485: case SO_SNDLOWAT:
1486: case SO_RCVLOWAT:
1487: if (m == NULL || m->m_len < sizeof(int))
1488: return EINVAL;
1.1 cgd 1489:
1.142 dyoung 1490: /*
1491: * Values < 1 make no sense for any of these
1492: * options, so disallow them.
1493: */
1494: optval = *mtod(m, int *);
1495: if (optval < 1)
1496: return EINVAL;
1.1 cgd 1497:
1.142 dyoung 1498: switch (optname) {
1.1 cgd 1499:
1500: case SO_SNDBUF:
1501: case SO_RCVBUF:
1.142 dyoung 1502: sb = (optname == SO_SNDBUF) ?
1503: &so->so_snd : &so->so_rcv;
1504: if (sbreserve(sb, (u_long)optval, so) == 0)
1505: return ENOBUFS;
1506: sb->sb_flags &= ~SB_AUTOSIZE;
1507: break;
1508:
1509: /*
1510: * Make sure the low-water is never greater than
1511: * the high-water.
1512: */
1.1 cgd 1513: case SO_SNDLOWAT:
1.142 dyoung 1514: so->so_snd.sb_lowat =
1515: (optval > so->so_snd.sb_hiwat) ?
1516: so->so_snd.sb_hiwat : optval;
1517: break;
1.1 cgd 1518: case SO_RCVLOWAT:
1.142 dyoung 1519: so->so_rcv.sb_lowat =
1520: (optval > so->so_rcv.sb_hiwat) ?
1521: so->so_rcv.sb_hiwat : optval;
1522: break;
1523: }
1524: break;
1.28 thorpej 1525:
1.142 dyoung 1526: case SO_SNDTIMEO:
1527: case SO_RCVTIMEO:
1528: if (m == NULL || m->m_len < sizeof(*tv))
1529: return EINVAL;
1530: tv = mtod(m, struct timeval *);
1531: if (tv->tv_sec > (INT_MAX - tv->tv_usec / tick) / hz)
1532: return EDOM;
1533: val = tv->tv_sec * hz + tv->tv_usec / tick;
1534: if (val == 0 && tv->tv_usec != 0)
1535: val = 1;
1.28 thorpej 1536:
1.142 dyoung 1537: switch (optname) {
1.28 thorpej 1538:
1.142 dyoung 1539: case SO_SNDTIMEO:
1540: so->so_snd.sb_timeo = val;
1.1 cgd 1541: break;
1542: case SO_RCVTIMEO:
1.142 dyoung 1543: so->so_rcv.sb_timeo = val;
1544: break;
1545: }
1546: break;
1.1 cgd 1547:
1.142 dyoung 1548: default:
1549: return ENOPROTOOPT;
1550: }
1551: return 0;
1552: }
1.1 cgd 1553:
1.142 dyoung 1554: int
1555: sosetopt(struct socket *so, int level, int optname, struct mbuf *m)
1556: {
1557: int error, prerr;
1.1 cgd 1558:
1.142 dyoung 1559: if (level == SOL_SOCKET)
1560: error = sosetopt1(so, level, optname, m);
1561: else
1562: error = ENOPROTOOPT;
1.1 cgd 1563:
1.142 dyoung 1564: if ((error == 0 || error == ENOPROTOOPT) &&
1565: so->so_proto != NULL && so->so_proto->pr_ctloutput != NULL) {
1566: /* give the protocol stack a shot */
1567: prerr = (*so->so_proto->pr_ctloutput)(PRCO_SETOPT, so, level,
1568: optname, &m);
1569: if (prerr == 0)
1570: error = 0;
1571: else if (prerr != ENOPROTOOPT)
1572: error = prerr;
1573: } else if (m != NULL)
1574: (void)m_free(m);
1575: return error;
1.1 cgd 1576: }
1577:
1.14 mycroft 1578: int
1.54 lukem 1579: sogetopt(struct socket *so, int level, int optname, struct mbuf **mp)
1.1 cgd 1580: {
1.54 lukem 1581: struct mbuf *m;
1.1 cgd 1582:
1583: if (level != SOL_SOCKET) {
1584: if (so->so_proto && so->so_proto->pr_ctloutput) {
1585: return ((*so->so_proto->pr_ctloutput)
1586: (PRCO_GETOPT, so, level, optname, mp));
1587: } else
1588: return (ENOPROTOOPT);
1589: } else {
1590: m = m_get(M_WAIT, MT_SOOPTS);
1.36 perry 1591: m->m_len = sizeof(int);
1.1 cgd 1592:
1593: switch (optname) {
1594:
1595: case SO_LINGER:
1.36 perry 1596: m->m_len = sizeof(struct linger);
1.1 cgd 1597: mtod(m, struct linger *)->l_onoff =
1.131 christos 1598: (so->so_options & SO_LINGER) ? 1 : 0;
1.1 cgd 1599: mtod(m, struct linger *)->l_linger = so->so_linger;
1600: break;
1601:
1602: case SO_USELOOPBACK:
1603: case SO_DONTROUTE:
1604: case SO_DEBUG:
1605: case SO_KEEPALIVE:
1606: case SO_REUSEADDR:
1.15 mycroft 1607: case SO_REUSEPORT:
1.1 cgd 1608: case SO_BROADCAST:
1609: case SO_OOBINLINE:
1.26 thorpej 1610: case SO_TIMESTAMP:
1.131 christos 1611: *mtod(m, int *) = (so->so_options & optname) ? 1 : 0;
1.1 cgd 1612: break;
1613:
1614: case SO_TYPE:
1615: *mtod(m, int *) = so->so_type;
1616: break;
1617:
1618: case SO_ERROR:
1619: *mtod(m, int *) = so->so_error;
1620: so->so_error = 0;
1621: break;
1622:
1623: case SO_SNDBUF:
1624: *mtod(m, int *) = so->so_snd.sb_hiwat;
1625: break;
1626:
1627: case SO_RCVBUF:
1628: *mtod(m, int *) = so->so_rcv.sb_hiwat;
1629: break;
1630:
1631: case SO_SNDLOWAT:
1632: *mtod(m, int *) = so->so_snd.sb_lowat;
1633: break;
1634:
1635: case SO_RCVLOWAT:
1636: *mtod(m, int *) = so->so_rcv.sb_lowat;
1637: break;
1638:
1639: case SO_SNDTIMEO:
1640: case SO_RCVTIMEO:
1641: {
1642: int val = (optname == SO_SNDTIMEO ?
1643: so->so_snd.sb_timeo : so->so_rcv.sb_timeo);
1644:
1645: m->m_len = sizeof(struct timeval);
1646: mtod(m, struct timeval *)->tv_sec = val / hz;
1647: mtod(m, struct timeval *)->tv_usec =
1.27 kleink 1648: (val % hz) * tick;
1.1 cgd 1649: break;
1650: }
1651:
1.107 darrenr 1652: case SO_OVERFLOWED:
1653: *mtod(m, int *) = so->so_rcv.sb_overflowed;
1654: break;
1655:
1.1 cgd 1656: default:
1657: (void)m_free(m);
1658: return (ENOPROTOOPT);
1659: }
1660: *mp = m;
1661: return (0);
1662: }
1663: }
1664:
1.14 mycroft 1665: void
1.54 lukem 1666: sohasoutofband(struct socket *so)
1.1 cgd 1667: {
1.153 rmind 1668:
1.90 christos 1669: fownsignal(so->so_pgid, SIGURG, POLL_PRI, POLLPRI|POLLRDBAND, so);
1.153 rmind 1670: selnotify(&so->so_rcv.sb_sel, POLLPRI | POLLRDBAND, 0);
1.1 cgd 1671: }
1.72 jdolecek 1672:
1673: static void
1674: filt_sordetach(struct knote *kn)
1675: {
1676: struct socket *so;
1677:
1.155 ad 1678: so = ((file_t *)kn->kn_obj)->f_data;
1.73 christos 1679: SLIST_REMOVE(&so->so_rcv.sb_sel.sel_klist, kn, knote, kn_selnext);
1680: if (SLIST_EMPTY(&so->so_rcv.sb_sel.sel_klist))
1.72 jdolecek 1681: so->so_rcv.sb_flags &= ~SB_KNOTE;
1682: }
1683:
1684: /*ARGSUSED*/
1685: static int
1.129 yamt 1686: filt_soread(struct knote *kn, long hint)
1.72 jdolecek 1687: {
1688: struct socket *so;
1689:
1.155 ad 1690: so = ((file_t *)kn->kn_obj)->f_data;
1.72 jdolecek 1691: kn->kn_data = so->so_rcv.sb_cc;
1692: if (so->so_state & SS_CANTRCVMORE) {
1.108 perry 1693: kn->kn_flags |= EV_EOF;
1.72 jdolecek 1694: kn->kn_fflags = so->so_error;
1695: return (1);
1696: }
1697: if (so->so_error) /* temporary udp error */
1698: return (1);
1699: if (kn->kn_sfflags & NOTE_LOWAT)
1700: return (kn->kn_data >= kn->kn_sdata);
1701: return (kn->kn_data >= so->so_rcv.sb_lowat);
1702: }
1703:
1704: static void
1705: filt_sowdetach(struct knote *kn)
1706: {
1707: struct socket *so;
1708:
1.155 ad 1709: so = ((file_t *)kn->kn_obj)->f_data;
1.73 christos 1710: SLIST_REMOVE(&so->so_snd.sb_sel.sel_klist, kn, knote, kn_selnext);
1711: if (SLIST_EMPTY(&so->so_snd.sb_sel.sel_klist))
1.72 jdolecek 1712: so->so_snd.sb_flags &= ~SB_KNOTE;
1713: }
1714:
1715: /*ARGSUSED*/
1716: static int
1.129 yamt 1717: filt_sowrite(struct knote *kn, long hint)
1.72 jdolecek 1718: {
1719: struct socket *so;
1720:
1.155 ad 1721: so = ((file_t *)kn->kn_obj)->f_data;
1.72 jdolecek 1722: kn->kn_data = sbspace(&so->so_snd);
1723: if (so->so_state & SS_CANTSENDMORE) {
1.108 perry 1724: kn->kn_flags |= EV_EOF;
1.72 jdolecek 1725: kn->kn_fflags = so->so_error;
1726: return (1);
1727: }
1728: if (so->so_error) /* temporary udp error */
1729: return (1);
1730: if (((so->so_state & SS_ISCONNECTED) == 0) &&
1731: (so->so_proto->pr_flags & PR_CONNREQUIRED))
1732: return (0);
1733: if (kn->kn_sfflags & NOTE_LOWAT)
1734: return (kn->kn_data >= kn->kn_sdata);
1735: return (kn->kn_data >= so->so_snd.sb_lowat);
1736: }
1737:
1738: /*ARGSUSED*/
1739: static int
1.129 yamt 1740: filt_solisten(struct knote *kn, long hint)
1.72 jdolecek 1741: {
1742: struct socket *so;
1743:
1.155 ad 1744: so = ((file_t *)kn->kn_obj)->f_data;
1.72 jdolecek 1745:
1746: /*
1747: * Set kn_data to number of incoming connections, not
1748: * counting partial (incomplete) connections.
1.108 perry 1749: */
1.72 jdolecek 1750: kn->kn_data = so->so_qlen;
1751: return (kn->kn_data > 0);
1752: }
1753:
1754: static const struct filterops solisten_filtops =
1755: { 1, NULL, filt_sordetach, filt_solisten };
1756: static const struct filterops soread_filtops =
1757: { 1, NULL, filt_sordetach, filt_soread };
1758: static const struct filterops sowrite_filtops =
1759: { 1, NULL, filt_sowdetach, filt_sowrite };
1760:
1761: int
1.129 yamt 1762: soo_kqfilter(struct file *fp, struct knote *kn)
1.72 jdolecek 1763: {
1764: struct socket *so;
1765: struct sockbuf *sb;
1766:
1.155 ad 1767: so = ((file_t *)kn->kn_obj)->f_data;
1.72 jdolecek 1768: switch (kn->kn_filter) {
1769: case EVFILT_READ:
1770: if (so->so_options & SO_ACCEPTCONN)
1771: kn->kn_fop = &solisten_filtops;
1772: else
1773: kn->kn_fop = &soread_filtops;
1774: sb = &so->so_rcv;
1775: break;
1776: case EVFILT_WRITE:
1777: kn->kn_fop = &sowrite_filtops;
1778: sb = &so->so_snd;
1779: break;
1780: default:
1.149 pooka 1781: return (EINVAL);
1.72 jdolecek 1782: }
1.73 christos 1783: SLIST_INSERT_HEAD(&sb->sb_sel.sel_klist, kn, kn_selnext);
1.72 jdolecek 1784: sb->sb_flags |= SB_KNOTE;
1785: return (0);
1786: }
1787:
1.154 ad 1788: static int
1789: sodopoll(struct socket *so, int events)
1790: {
1791: int revents;
1792:
1793: revents = 0;
1794:
1795: if (events & (POLLIN | POLLRDNORM))
1796: if (soreadable(so))
1797: revents |= events & (POLLIN | POLLRDNORM);
1798:
1799: if (events & (POLLOUT | POLLWRNORM))
1800: if (sowritable(so))
1801: revents |= events & (POLLOUT | POLLWRNORM);
1802:
1803: if (events & (POLLPRI | POLLRDBAND))
1804: if (so->so_oobmark || (so->so_state & SS_RCVATMARK))
1805: revents |= events & (POLLPRI | POLLRDBAND);
1806:
1807: return revents;
1808: }
1809:
1810: int
1811: sopoll(struct socket *so, int events)
1812: {
1813: int revents = 0;
1814: int s;
1815:
1816: if ((revents = sodopoll(so, events)) != 0)
1817: return revents;
1818:
1819: KERNEL_LOCK(1, curlwp);
1820: s = splsoftnet();
1821:
1822: if ((revents = sodopoll(so, events)) == 0) {
1823: if (events & (POLLIN | POLLPRI | POLLRDNORM | POLLRDBAND)) {
1824: selrecord(curlwp, &so->so_rcv.sb_sel);
1825: so->so_rcv.sb_flags |= SB_SEL;
1826: }
1827:
1828: if (events & (POLLOUT | POLLWRNORM)) {
1829: selrecord(curlwp, &so->so_snd.sb_sel);
1830: so->so_snd.sb_flags |= SB_SEL;
1831: }
1832: }
1833:
1834: splx(s);
1835: KERNEL_UNLOCK_ONE(curlwp);
1836:
1837: return revents;
1838: }
1839:
1840:
1.94 yamt 1841: #include <sys/sysctl.h>
1842:
1843: static int sysctl_kern_somaxkva(SYSCTLFN_PROTO);
1844:
1845: /*
1846: * sysctl helper routine for kern.somaxkva. ensures that the given
1847: * value is not too small.
1848: * (XXX should we maybe make sure it's not too large as well?)
1849: */
1850: static int
1851: sysctl_kern_somaxkva(SYSCTLFN_ARGS)
1852: {
1853: int error, new_somaxkva;
1854: struct sysctlnode node;
1855:
1856: new_somaxkva = somaxkva;
1857: node = *rnode;
1858: node.sysctl_data = &new_somaxkva;
1859: error = sysctl_lookup(SYSCTLFN_CALL(&node));
1860: if (error || newp == NULL)
1861: return (error);
1862:
1863: if (new_somaxkva < (16 * 1024 * 1024)) /* sanity */
1864: return (EINVAL);
1865:
1.136 ad 1866: mutex_enter(&so_pendfree_lock);
1.94 yamt 1867: somaxkva = new_somaxkva;
1.136 ad 1868: cv_broadcast(&socurkva_cv);
1869: mutex_exit(&so_pendfree_lock);
1.94 yamt 1870:
1871: return (error);
1872: }
1873:
1874: SYSCTL_SETUP(sysctl_kern_somaxkva_setup, "sysctl kern.somaxkva setup")
1875: {
1876:
1.97 atatat 1877: sysctl_createv(clog, 0, NULL, NULL,
1878: CTLFLAG_PERMANENT,
1879: CTLTYPE_NODE, "kern", NULL,
1880: NULL, 0, NULL, 0,
1881: CTL_KERN, CTL_EOL);
1882:
1883: sysctl_createv(clog, 0, NULL, NULL,
1884: CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1.103 atatat 1885: CTLTYPE_INT, "somaxkva",
1886: SYSCTL_DESCR("Maximum amount of kernel memory to be "
1887: "used for socket buffers"),
1.94 yamt 1888: sysctl_kern_somaxkva, 0, NULL, 0,
1889: CTL_KERN, KERN_SOMAXKVA, CTL_EOL);
1890: }
CVSweb <webmaster@jp.NetBSD.org>