Annotation of src/sys/kern/uipc_socket.c, Revision 1.140.6.1
1.140.6.1! jmcneill 1: /* $NetBSD: uipc_socket.c,v 1.141 2007/08/06 11:41:52 yamt Exp $ */
1.64 thorpej 2:
3: /*-
1.137 ad 4: * Copyright (c) 2002, 2007 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>
1.140.6.1! jmcneill 71: __KERNEL_RCSID(0, "$NetBSD: uipc_socket.c,v 1.141 2007/08/06 11:41:52 yamt 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>
82: #include <sys/malloc.h>
83: #include <sys/mbuf.h>
84: #include <sys/domain.h>
85: #include <sys/kernel.h>
86: #include <sys/protosw.h>
87: #include <sys/socket.h>
88: #include <sys/socketvar.h>
1.21 christos 89: #include <sys/signalvar.h>
1.9 mycroft 90: #include <sys/resourcevar.h>
1.37 thorpej 91: #include <sys/pool.h>
1.72 jdolecek 92: #include <sys/event.h>
1.89 christos 93: #include <sys/poll.h>
1.118 elad 94: #include <sys/kauth.h>
1.136 ad 95: #include <sys/mutex.h>
96: #include <sys/condvar.h>
1.37 thorpej 97:
1.64 thorpej 98: #include <uvm/uvm.h>
99:
1.135 ad 100: POOL_INIT(socket_pool, sizeof(struct socket), 0, 0, 0, "sockpl", NULL,
101: IPL_SOFTNET);
1.77 thorpej 102:
103: MALLOC_DEFINE(M_SOOPTS, "soopts", "socket options");
104: MALLOC_DEFINE(M_SONAME, "soname", "socket name");
1.37 thorpej 105:
1.54 lukem 106: extern int somaxconn; /* patchable (XXX sysctl) */
107: int somaxconn = SOMAXCONN;
1.49 jonathan 108:
1.64 thorpej 109: #ifdef SOSEND_COUNTERS
110: #include <sys/device.h>
111:
1.113 thorpej 112: static struct evcnt sosend_loan_big = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
1.64 thorpej 113: NULL, "sosend", "loan big");
1.113 thorpej 114: static struct evcnt sosend_copy_big = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
1.64 thorpej 115: NULL, "sosend", "copy big");
1.113 thorpej 116: static struct evcnt sosend_copy_small = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
1.64 thorpej 117: NULL, "sosend", "copy small");
1.113 thorpej 118: static struct evcnt sosend_kvalimit = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
1.64 thorpej 119: NULL, "sosend", "kva limit");
120:
121: #define SOSEND_COUNTER_INCR(ev) (ev)->ev_count++
122:
1.101 matt 123: EVCNT_ATTACH_STATIC(sosend_loan_big);
124: EVCNT_ATTACH_STATIC(sosend_copy_big);
125: EVCNT_ATTACH_STATIC(sosend_copy_small);
126: EVCNT_ATTACH_STATIC(sosend_kvalimit);
1.64 thorpej 127: #else
128:
129: #define SOSEND_COUNTER_INCR(ev) /* nothing */
130:
131: #endif /* SOSEND_COUNTERS */
132:
1.119 yamt 133: static struct callback_entry sokva_reclaimerentry;
1.1 cgd 134:
1.71 thorpej 135: #ifdef SOSEND_NO_LOAN
1.121 yamt 136: int sock_loan_thresh = -1;
1.71 thorpej 137: #else
1.121 yamt 138: int sock_loan_thresh = 4096;
1.65 thorpej 139: #endif
1.64 thorpej 140:
1.136 ad 141: static kmutex_t so_pendfree_lock;
1.113 thorpej 142: static struct mbuf *so_pendfree;
1.64 thorpej 143:
1.84 ragge 144: #ifndef SOMAXKVA
145: #define SOMAXKVA (16 * 1024 * 1024)
146: #endif
147: int somaxkva = SOMAXKVA;
1.113 thorpej 148: static int socurkva;
1.136 ad 149: static kcondvar_t socurkva_cv;
1.64 thorpej 150:
151: #define SOCK_LOAN_CHUNK 65536
152:
1.117 yamt 153: static size_t sodopendfree(void);
154: static size_t sodopendfreel(void);
1.93 yamt 155:
1.113 thorpej 156: static vsize_t
1.129 yamt 157: sokvareserve(struct socket *so, vsize_t len)
1.80 yamt 158: {
1.98 christos 159: int error;
1.80 yamt 160:
1.136 ad 161: mutex_enter(&so_pendfree_lock);
1.80 yamt 162: while (socurkva + len > somaxkva) {
1.93 yamt 163: size_t freed;
164:
165: /*
166: * try to do pendfree.
167: */
168:
1.117 yamt 169: freed = sodopendfreel();
1.93 yamt 170:
171: /*
172: * if some kva was freed, try again.
173: */
174:
175: if (freed)
1.80 yamt 176: continue;
1.93 yamt 177:
1.80 yamt 178: SOSEND_COUNTER_INCR(&sosend_kvalimit);
1.136 ad 179: error = cv_wait_sig(&socurkva_cv, &so_pendfree_lock);
1.98 christos 180: if (error) {
181: len = 0;
182: break;
183: }
1.80 yamt 184: }
1.93 yamt 185: socurkva += len;
1.136 ad 186: mutex_exit(&so_pendfree_lock);
1.98 christos 187: return len;
1.95 yamt 188: }
189:
1.113 thorpej 190: static void
1.95 yamt 191: sokvaunreserve(vsize_t len)
192: {
193:
1.136 ad 194: mutex_enter(&so_pendfree_lock);
1.95 yamt 195: socurkva -= len;
1.136 ad 196: cv_broadcast(&socurkva_cv);
197: mutex_exit(&so_pendfree_lock);
1.95 yamt 198: }
199:
200: /*
201: * sokvaalloc: allocate kva for loan.
202: */
203:
204: vaddr_t
205: sokvaalloc(vsize_t len, struct socket *so)
206: {
207: vaddr_t lva;
208:
209: /*
210: * reserve kva.
211: */
212:
1.98 christos 213: if (sokvareserve(so, len) == 0)
214: return 0;
1.93 yamt 215:
216: /*
217: * allocate kva.
218: */
1.80 yamt 219:
1.109 yamt 220: lva = uvm_km_alloc(kernel_map, len, 0, UVM_KMF_VAONLY | UVM_KMF_WAITVA);
1.95 yamt 221: if (lva == 0) {
222: sokvaunreserve(len);
1.80 yamt 223: return (0);
1.95 yamt 224: }
1.80 yamt 225:
226: return lva;
227: }
228:
1.93 yamt 229: /*
230: * sokvafree: free kva for loan.
231: */
232:
1.80 yamt 233: void
234: sokvafree(vaddr_t sva, vsize_t len)
235: {
1.93 yamt 236:
237: /*
238: * free kva.
239: */
1.80 yamt 240:
1.109 yamt 241: uvm_km_free(kernel_map, sva, len, UVM_KMF_VAONLY);
1.93 yamt 242:
243: /*
244: * unreserve kva.
245: */
246:
1.95 yamt 247: sokvaunreserve(len);
1.80 yamt 248: }
249:
1.64 thorpej 250: static void
1.134 christos 251: sodoloanfree(struct vm_page **pgs, void *buf, size_t size)
1.64 thorpej 252: {
253: vaddr_t va, sva, eva;
254: vsize_t len;
255: paddr_t pa;
256: int i, npgs;
257:
258: eva = round_page((vaddr_t) buf + size);
259: sva = trunc_page((vaddr_t) buf);
260: len = eva - sva;
261: npgs = len >> PAGE_SHIFT;
262:
1.79 thorpej 263: if (__predict_false(pgs == NULL)) {
264: pgs = alloca(npgs * sizeof(*pgs));
1.64 thorpej 265:
1.79 thorpej 266: for (i = 0, va = sva; va < eva; i++, va += PAGE_SIZE) {
1.133 thorpej 267: if (pmap_extract(pmap_kernel(), va, &pa) == false)
1.79 thorpej 268: panic("sodoloanfree: va 0x%lx not mapped", va);
269: pgs[i] = PHYS_TO_VM_PAGE(pa);
270: }
1.64 thorpej 271: }
272:
273: pmap_kremove(sva, len);
274: pmap_update(pmap_kernel());
275: uvm_unloan(pgs, npgs, UVM_LOAN_TOPAGE);
1.80 yamt 276: sokvafree(sva, len);
1.64 thorpej 277: }
278:
279: static size_t
1.117 yamt 280: sodopendfree()
1.64 thorpej 281: {
1.93 yamt 282: size_t rv;
1.64 thorpej 283:
1.136 ad 284: mutex_enter(&so_pendfree_lock);
1.117 yamt 285: rv = sodopendfreel();
1.136 ad 286: mutex_exit(&so_pendfree_lock);
1.93 yamt 287:
288: return rv;
289: }
290:
291: /*
292: * sodopendfreel: free mbufs on "pendfree" list.
1.136 ad 293: * unlock and relock so_pendfree_lock when freeing mbufs.
1.93 yamt 294: *
1.136 ad 295: * => called with so_pendfree_lock held.
1.93 yamt 296: */
297:
298: static size_t
1.117 yamt 299: sodopendfreel()
1.93 yamt 300: {
1.137 ad 301: struct mbuf *m, *next;
1.93 yamt 302: size_t rv = 0;
1.137 ad 303: int s;
1.93 yamt 304:
1.136 ad 305: KASSERT(mutex_owned(&so_pendfree_lock));
1.64 thorpej 306:
1.137 ad 307: while (so_pendfree != NULL) {
1.64 thorpej 308: m = so_pendfree;
1.93 yamt 309: so_pendfree = NULL;
1.136 ad 310: mutex_exit(&so_pendfree_lock);
1.93 yamt 311:
312: for (; m != NULL; m = next) {
313: next = m->m_next;
314:
315: rv += m->m_ext.ext_size;
316: sodoloanfree((m->m_flags & M_EXT_PAGES) ?
317: m->m_ext.ext_pgs : NULL, m->m_ext.ext_buf,
318: m->m_ext.ext_size);
1.137 ad 319: s = splvm();
1.93 yamt 320: pool_cache_put(&mbpool_cache, m);
1.137 ad 321: splx(s);
1.93 yamt 322: }
1.64 thorpej 323:
1.136 ad 324: mutex_enter(&so_pendfree_lock);
1.64 thorpej 325: }
326:
327: return (rv);
328: }
329:
1.80 yamt 330: void
1.134 christos 331: soloanfree(struct mbuf *m, void *buf, size_t size, void *arg)
1.64 thorpej 332: {
333:
334: if (m == NULL) {
1.93 yamt 335:
336: /*
337: * called from MEXTREMOVE.
338: */
339:
1.79 thorpej 340: sodoloanfree(NULL, buf, size);
1.64 thorpej 341: return;
342: }
343:
1.93 yamt 344: /*
345: * postpone freeing mbuf.
346: *
347: * we can't do it in interrupt context
348: * because we need to put kva back to kernel_map.
349: */
350:
1.136 ad 351: mutex_enter(&so_pendfree_lock);
1.92 yamt 352: m->m_next = so_pendfree;
353: so_pendfree = m;
1.136 ad 354: cv_broadcast(&socurkva_cv);
355: mutex_exit(&so_pendfree_lock);
1.64 thorpej 356: }
357:
358: static long
359: sosend_loan(struct socket *so, struct uio *uio, struct mbuf *m, long space)
360: {
361: struct iovec *iov = uio->uio_iov;
362: vaddr_t sva, eva;
363: vsize_t len;
364: vaddr_t lva, va;
1.80 yamt 365: int npgs, i, error;
1.64 thorpej 366:
1.116 yamt 367: if (VMSPACE_IS_KERNEL_P(uio->uio_vmspace))
1.64 thorpej 368: return (0);
369:
370: if (iov->iov_len < (size_t) space)
371: space = iov->iov_len;
372: if (space > SOCK_LOAN_CHUNK)
373: space = SOCK_LOAN_CHUNK;
374:
375: eva = round_page((vaddr_t) iov->iov_base + space);
376: sva = trunc_page((vaddr_t) iov->iov_base);
377: len = eva - sva;
378: npgs = len >> PAGE_SHIFT;
379:
1.79 thorpej 380: /* XXX KDASSERT */
381: KASSERT(npgs <= M_EXT_MAXPAGES);
382:
1.80 yamt 383: lva = sokvaalloc(len, so);
1.64 thorpej 384: if (lva == 0)
1.80 yamt 385: return 0;
1.64 thorpej 386:
1.116 yamt 387: error = uvm_loan(&uio->uio_vmspace->vm_map, sva, len,
1.79 thorpej 388: m->m_ext.ext_pgs, UVM_LOAN_TOPAGE);
1.64 thorpej 389: if (error) {
1.80 yamt 390: sokvafree(lva, len);
1.64 thorpej 391: return (0);
392: }
393:
394: for (i = 0, va = lva; i < npgs; i++, va += PAGE_SIZE)
1.79 thorpej 395: pmap_kenter_pa(va, VM_PAGE_TO_PHYS(m->m_ext.ext_pgs[i]),
396: VM_PROT_READ);
1.64 thorpej 397: pmap_update(pmap_kernel());
398:
399: lva += (vaddr_t) iov->iov_base & PAGE_MASK;
400:
1.134 christos 401: MEXTADD(m, (void *) lva, space, M_MBUF, soloanfree, so);
1.79 thorpej 402: m->m_flags |= M_EXT_PAGES | M_EXT_ROMAP;
1.64 thorpej 403:
404: uio->uio_resid -= space;
405: /* uio_offset not updated, not set/used for write(2) */
1.134 christos 406: uio->uio_iov->iov_base = (char *)uio->uio_iov->iov_base + space;
1.64 thorpej 407: uio->uio_iov->iov_len -= space;
408: if (uio->uio_iov->iov_len == 0) {
409: uio->uio_iov++;
410: uio->uio_iovcnt--;
411: }
412:
413: return (space);
414: }
415:
1.119 yamt 416: static int
1.129 yamt 417: sokva_reclaim_callback(struct callback_entry *ce, void *obj, void *arg)
1.119 yamt 418: {
419:
420: KASSERT(ce == &sokva_reclaimerentry);
421: KASSERT(obj == NULL);
422:
423: sodopendfree();
424: if (!vm_map_starved_p(kernel_map)) {
425: return CALLBACK_CHAIN_ABORT;
426: }
427: return CALLBACK_CHAIN_CONTINUE;
428: }
429:
430: void
431: soinit(void)
432: {
433:
1.136 ad 434: mutex_init(&so_pendfree_lock, MUTEX_DRIVER, IPL_VM);
435: cv_init(&socurkva_cv, "sokva");
436:
1.119 yamt 437: /* Set the initial adjusted socket buffer size. */
438: if (sb_max_set(sb_max))
439: panic("bad initial sb_max value: %lu", sb_max);
440:
441: callback_register(&vm_map_to_kernel(kernel_map)->vmk_reclaim_callback,
442: &sokva_reclaimerentry, NULL, sokva_reclaim_callback);
443: }
444:
1.1 cgd 445: /*
446: * Socket operation routines.
447: * These routines are called by the routines in
448: * sys_socket.c or from a system process, and
449: * implement the semantics of socket operations by
450: * switching out to the protocol specific routines.
451: */
452: /*ARGSUSED*/
1.3 andrew 453: int
1.114 christos 454: socreate(int dom, struct socket **aso, int type, int proto, struct lwp *l)
1.1 cgd 455: {
1.99 matt 456: const struct protosw *prp;
1.54 lukem 457: struct socket *so;
1.115 yamt 458: uid_t uid;
1.54 lukem 459: int error, s;
1.1 cgd 460:
1.132 elad 461: error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_SOCKET,
462: KAUTH_REQ_NETWORK_SOCKET_OPEN, KAUTH_ARG(dom), KAUTH_ARG(type),
463: KAUTH_ARG(proto));
1.140 dyoung 464: if (error != 0)
465: return error;
1.127 elad 466:
1.1 cgd 467: if (proto)
468: prp = pffindproto(dom, proto, type);
469: else
470: prp = pffindtype(dom, type);
1.140 dyoung 471: if (prp == NULL) {
1.120 ginsbach 472: /* no support for domain */
473: if (pffinddomain(dom) == 0)
1.140 dyoung 474: return EAFNOSUPPORT;
1.120 ginsbach 475: /* no support for socket type */
476: if (proto == 0 && type != 0)
1.140 dyoung 477: return EPROTOTYPE;
478: return EPROTONOSUPPORT;
1.120 ginsbach 479: }
1.140 dyoung 480: if (prp->pr_usrreq == NULL)
481: return EPROTONOSUPPORT;
1.1 cgd 482: if (prp->pr_type != type)
1.140 dyoung 483: return EPROTOTYPE;
1.39 matt 484: s = splsoftnet();
1.37 thorpej 485: so = pool_get(&socket_pool, PR_WAITOK);
1.140 dyoung 486: memset(so, 0, sizeof(*so));
1.31 thorpej 487: TAILQ_INIT(&so->so_q0);
488: TAILQ_INIT(&so->so_q);
1.1 cgd 489: so->so_type = type;
490: so->so_proto = prp;
1.33 matt 491: so->so_send = sosend;
492: so->so_receive = soreceive;
1.78 matt 493: #ifdef MBUFTRACE
494: so->so_rcv.sb_mowner = &prp->pr_domain->dom_mowner;
495: so->so_snd.sb_mowner = &prp->pr_domain->dom_mowner;
496: so->so_mowner = &prp->pr_domain->dom_mowner;
497: #endif
1.138 rmind 498: uid = kauth_cred_geteuid(l->l_cred);
1.115 yamt 499: so->so_uidinfo = uid_find(uid);
1.140 dyoung 500: error = (*prp->pr_usrreq)(so, PRU_ATTACH, NULL,
501: (struct mbuf *)(long)proto, NULL, l);
502: if (error != 0) {
1.1 cgd 503: so->so_state |= SS_NOFDREF;
504: sofree(so);
1.39 matt 505: splx(s);
1.140 dyoung 506: return error;
1.1 cgd 507: }
1.39 matt 508: splx(s);
1.1 cgd 509: *aso = so;
1.140 dyoung 510: return 0;
1.1 cgd 511: }
512:
1.3 andrew 513: int
1.114 christos 514: sobind(struct socket *so, struct mbuf *nam, struct lwp *l)
1.1 cgd 515: {
1.54 lukem 516: int s, error;
1.1 cgd 517:
1.54 lukem 518: s = splsoftnet();
1.140 dyoung 519: error = (*so->so_proto->pr_usrreq)(so, PRU_BIND, NULL, nam, NULL, l);
1.1 cgd 520: splx(s);
1.140 dyoung 521: return error;
1.1 cgd 522: }
523:
1.3 andrew 524: int
1.54 lukem 525: solisten(struct socket *so, int backlog)
1.1 cgd 526: {
1.54 lukem 527: int s, error;
1.1 cgd 528:
1.54 lukem 529: s = splsoftnet();
1.140 dyoung 530: error = (*so->so_proto->pr_usrreq)(so, PRU_LISTEN, NULL,
531: NULL, NULL, NULL);
532: if (error != 0) {
1.1 cgd 533: splx(s);
1.140 dyoung 534: return error;
1.1 cgd 535: }
1.63 matt 536: if (TAILQ_EMPTY(&so->so_q))
1.1 cgd 537: so->so_options |= SO_ACCEPTCONN;
538: if (backlog < 0)
539: backlog = 0;
1.49 jonathan 540: so->so_qlimit = min(backlog, somaxconn);
1.1 cgd 541: splx(s);
1.140 dyoung 542: return 0;
1.1 cgd 543: }
544:
1.21 christos 545: void
1.54 lukem 546: sofree(struct socket *so)
1.1 cgd 547: {
548:
1.43 mycroft 549: if (so->so_pcb || (so->so_state & SS_NOFDREF) == 0)
1.1 cgd 550: return;
1.43 mycroft 551: if (so->so_head) {
552: /*
553: * We must not decommission a socket that's on the accept(2)
554: * queue. If we do, then accept(2) may hang after select(2)
555: * indicated that the listening socket was ready.
556: */
557: if (!soqremque(so, 0))
558: return;
559: }
1.98 christos 560: if (so->so_rcv.sb_hiwat)
1.110 christos 561: (void)chgsbsize(so->so_uidinfo, &so->so_rcv.sb_hiwat, 0,
1.98 christos 562: RLIM_INFINITY);
563: if (so->so_snd.sb_hiwat)
1.110 christos 564: (void)chgsbsize(so->so_uidinfo, &so->so_snd.sb_hiwat, 0,
1.98 christos 565: RLIM_INFINITY);
566: sbrelease(&so->so_snd, so);
1.1 cgd 567: sorflush(so);
1.37 thorpej 568: pool_put(&socket_pool, so);
1.1 cgd 569: }
570:
571: /*
572: * Close a socket on last file table reference removal.
573: * Initiate disconnect if connected.
574: * Free socket when disconnect complete.
575: */
1.3 andrew 576: int
1.54 lukem 577: soclose(struct socket *so)
1.1 cgd 578: {
1.54 lukem 579: struct socket *so2;
580: int s, error;
1.1 cgd 581:
1.54 lukem 582: error = 0;
583: s = splsoftnet(); /* conservative */
1.1 cgd 584: if (so->so_options & SO_ACCEPTCONN) {
1.63 matt 585: while ((so2 = TAILQ_FIRST(&so->so_q0)) != 0) {
1.42 mycroft 586: (void) soqremque(so2, 0);
1.41 mycroft 587: (void) soabort(so2);
588: }
1.63 matt 589: while ((so2 = TAILQ_FIRST(&so->so_q)) != 0) {
1.42 mycroft 590: (void) soqremque(so2, 1);
1.41 mycroft 591: (void) soabort(so2);
592: }
1.1 cgd 593: }
594: if (so->so_pcb == 0)
595: goto discard;
596: if (so->so_state & SS_ISCONNECTED) {
597: if ((so->so_state & SS_ISDISCONNECTING) == 0) {
598: error = sodisconnect(so);
599: if (error)
600: goto drop;
601: }
602: if (so->so_options & SO_LINGER) {
603: if ((so->so_state & SS_ISDISCONNECTING) &&
604: (so->so_state & SS_NBIO))
605: goto drop;
1.21 christos 606: while (so->so_state & SS_ISCONNECTED) {
1.134 christos 607: error = tsleep((void *)&so->so_timeo,
1.21 christos 608: PSOCK | PCATCH, netcls,
1.30 thorpej 609: so->so_linger * hz);
1.21 christos 610: if (error)
1.1 cgd 611: break;
1.21 christos 612: }
1.1 cgd 613: }
614: }
1.54 lukem 615: drop:
1.1 cgd 616: if (so->so_pcb) {
1.22 mycroft 617: int error2 = (*so->so_proto->pr_usrreq)(so, PRU_DETACH,
1.140 dyoung 618: NULL, NULL, NULL, NULL);
1.1 cgd 619: if (error == 0)
620: error = error2;
621: }
1.54 lukem 622: discard:
1.1 cgd 623: if (so->so_state & SS_NOFDREF)
624: panic("soclose: NOFDREF");
625: so->so_state |= SS_NOFDREF;
626: sofree(so);
627: splx(s);
628: return (error);
629: }
630:
631: /*
1.20 mycroft 632: * Must be called at splsoftnet...
1.1 cgd 633: */
1.3 andrew 634: int
1.54 lukem 635: soabort(struct socket *so)
1.1 cgd 636: {
1.139 yamt 637: int error;
1.1 cgd 638:
1.139 yamt 639: KASSERT(so->so_head == NULL);
1.140 dyoung 640: error = (*so->so_proto->pr_usrreq)(so, PRU_ABORT, NULL,
641: NULL, NULL, NULL);
1.139 yamt 642: if (error) {
643: sofree(so);
644: }
645: return error;
1.1 cgd 646: }
647:
1.3 andrew 648: int
1.54 lukem 649: soaccept(struct socket *so, struct mbuf *nam)
1.1 cgd 650: {
1.54 lukem 651: int s, error;
1.1 cgd 652:
1.54 lukem 653: error = 0;
654: s = splsoftnet();
1.1 cgd 655: if ((so->so_state & SS_NOFDREF) == 0)
656: panic("soaccept: !NOFDREF");
657: so->so_state &= ~SS_NOFDREF;
1.55 thorpej 658: if ((so->so_state & SS_ISDISCONNECTED) == 0 ||
659: (so->so_proto->pr_flags & PR_ABRTACPTDIS) == 0)
1.41 mycroft 660: error = (*so->so_proto->pr_usrreq)(so, PRU_ACCEPT,
1.140 dyoung 661: NULL, nam, NULL, NULL);
1.41 mycroft 662: else
1.53 itojun 663: error = ECONNABORTED;
1.52 itojun 664:
1.1 cgd 665: splx(s);
666: return (error);
667: }
668:
1.3 andrew 669: int
1.114 christos 670: soconnect(struct socket *so, struct mbuf *nam, struct lwp *l)
1.1 cgd 671: {
1.54 lukem 672: int s, error;
1.1 cgd 673:
674: if (so->so_options & SO_ACCEPTCONN)
675: return (EOPNOTSUPP);
1.20 mycroft 676: s = splsoftnet();
1.1 cgd 677: /*
678: * If protocol is connection-based, can only connect once.
679: * Otherwise, if connected, try to disconnect first.
680: * This allows user to disconnect by connecting to, e.g.,
681: * a null address.
682: */
683: if (so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING) &&
684: ((so->so_proto->pr_flags & PR_CONNREQUIRED) ||
685: (error = sodisconnect(so))))
686: error = EISCONN;
687: else
688: error = (*so->so_proto->pr_usrreq)(so, PRU_CONNECT,
1.140 dyoung 689: NULL, nam, NULL, l);
1.1 cgd 690: splx(s);
691: return (error);
692: }
693:
1.3 andrew 694: int
1.54 lukem 695: soconnect2(struct socket *so1, struct socket *so2)
1.1 cgd 696: {
1.54 lukem 697: int s, error;
1.1 cgd 698:
1.54 lukem 699: s = splsoftnet();
1.22 mycroft 700: error = (*so1->so_proto->pr_usrreq)(so1, PRU_CONNECT2,
1.140 dyoung 701: NULL, (struct mbuf *)so2, NULL, NULL);
1.1 cgd 702: splx(s);
703: return (error);
704: }
705:
1.3 andrew 706: int
1.54 lukem 707: sodisconnect(struct socket *so)
1.1 cgd 708: {
1.54 lukem 709: int s, error;
1.1 cgd 710:
1.54 lukem 711: s = splsoftnet();
1.1 cgd 712: if ((so->so_state & SS_ISCONNECTED) == 0) {
713: error = ENOTCONN;
714: goto bad;
715: }
716: if (so->so_state & SS_ISDISCONNECTING) {
717: error = EALREADY;
718: goto bad;
719: }
1.22 mycroft 720: error = (*so->so_proto->pr_usrreq)(so, PRU_DISCONNECT,
1.140 dyoung 721: NULL, NULL, NULL, NULL);
1.54 lukem 722: bad:
1.1 cgd 723: splx(s);
1.117 yamt 724: sodopendfree();
1.1 cgd 725: return (error);
726: }
727:
1.15 mycroft 728: #define SBLOCKWAIT(f) (((f) & MSG_DONTWAIT) ? M_NOWAIT : M_WAITOK)
1.1 cgd 729: /*
730: * Send on a socket.
731: * If send must go all at once and message is larger than
732: * send buffering, then hard error.
733: * Lock against other senders.
734: * If must go all at once and not enough room now, then
735: * inform user that this would block and do nothing.
736: * Otherwise, if nonblocking, send as much as possible.
737: * The data to be sent is described by "uio" if nonzero,
738: * otherwise by the mbuf chain "top" (which must be null
739: * if uio is not). Data provided in mbuf chain must be small
740: * enough to send all at once.
741: *
742: * Returns nonzero on error, timeout or signal; callers
743: * must check for short counts if EINTR/ERESTART are returned.
744: * Data and control buffers are freed on return.
745: */
1.3 andrew 746: int
1.54 lukem 747: sosend(struct socket *so, struct mbuf *addr, struct uio *uio, struct mbuf *top,
1.114 christos 748: struct mbuf *control, int flags, struct lwp *l)
1.1 cgd 749: {
1.54 lukem 750: struct mbuf **mp, *m;
1.114 christos 751: struct proc *p;
1.58 jdolecek 752: long space, len, resid, clen, mlen;
753: int error, s, dontroute, atomic;
1.54 lukem 754:
1.114 christos 755: p = l->l_proc;
1.117 yamt 756: sodopendfree();
1.64 thorpej 757:
1.54 lukem 758: clen = 0;
759: atomic = sosendallatonce(so) || top;
1.1 cgd 760: if (uio)
761: resid = uio->uio_resid;
762: else
763: resid = top->m_pkthdr.len;
1.7 cgd 764: /*
765: * In theory resid should be unsigned.
766: * However, space must be signed, as it might be less than 0
767: * if we over-committed, and we must use a signed comparison
768: * of space and resid. On the other hand, a negative resid
769: * causes us to loop sending 0-length segments to the protocol.
770: */
1.29 mycroft 771: if (resid < 0) {
772: error = EINVAL;
773: goto out;
774: }
1.1 cgd 775: dontroute =
776: (flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 &&
777: (so->so_proto->pr_flags & PR_ATOMIC);
1.102 jonathan 778: if (p)
779: p->p_stats->p_ru.ru_msgsnd++;
1.1 cgd 780: if (control)
781: clen = control->m_len;
782: #define snderr(errno) { error = errno; splx(s); goto release; }
783:
1.54 lukem 784: restart:
1.21 christos 785: if ((error = sblock(&so->so_snd, SBLOCKWAIT(flags))) != 0)
1.1 cgd 786: goto out;
787: do {
1.20 mycroft 788: s = splsoftnet();
1.1 cgd 789: if (so->so_state & SS_CANTSENDMORE)
790: snderr(EPIPE);
1.48 thorpej 791: if (so->so_error) {
792: error = so->so_error;
793: so->so_error = 0;
794: splx(s);
795: goto release;
796: }
1.1 cgd 797: if ((so->so_state & SS_ISCONNECTED) == 0) {
798: if (so->so_proto->pr_flags & PR_CONNREQUIRED) {
799: if ((so->so_state & SS_ISCONFIRMING) == 0 &&
800: !(resid == 0 && clen != 0))
801: snderr(ENOTCONN);
802: } else if (addr == 0)
803: snderr(EDESTADDRREQ);
804: }
805: space = sbspace(&so->so_snd);
806: if (flags & MSG_OOB)
807: space += 1024;
1.21 christos 808: if ((atomic && resid > so->so_snd.sb_hiwat) ||
1.11 mycroft 809: clen > so->so_snd.sb_hiwat)
810: snderr(EMSGSIZE);
1.96 mycroft 811: if (space < resid + clen &&
1.1 cgd 812: (atomic || space < so->so_snd.sb_lowat || space < clen)) {
813: if (so->so_state & SS_NBIO)
814: snderr(EWOULDBLOCK);
815: sbunlock(&so->so_snd);
816: error = sbwait(&so->so_snd);
817: splx(s);
818: if (error)
819: goto out;
820: goto restart;
821: }
822: splx(s);
823: mp = ⊤
824: space -= clen;
825: do {
1.45 tv 826: if (uio == NULL) {
827: /*
828: * Data is prepackaged in "top".
829: */
830: resid = 0;
831: if (flags & MSG_EOR)
832: top->m_flags |= M_EOR;
833: } else do {
834: if (top == 0) {
1.78 matt 835: m = m_gethdr(M_WAIT, MT_DATA);
1.45 tv 836: mlen = MHLEN;
837: m->m_pkthdr.len = 0;
1.140 dyoung 838: m->m_pkthdr.rcvif = NULL;
1.45 tv 839: } else {
1.78 matt 840: m = m_get(M_WAIT, MT_DATA);
1.45 tv 841: mlen = MLEN;
842: }
1.78 matt 843: MCLAIM(m, so->so_snd.sb_mowner);
1.121 yamt 844: if (sock_loan_thresh >= 0 &&
845: uio->uio_iov->iov_len >= sock_loan_thresh &&
846: space >= sock_loan_thresh &&
1.64 thorpej 847: (len = sosend_loan(so, uio, m,
848: space)) != 0) {
849: SOSEND_COUNTER_INCR(&sosend_loan_big);
850: space -= len;
851: goto have_data;
852: }
1.45 tv 853: if (resid >= MINCLSIZE && space >= MCLBYTES) {
1.64 thorpej 854: SOSEND_COUNTER_INCR(&sosend_copy_big);
1.78 matt 855: m_clget(m, M_WAIT);
1.45 tv 856: if ((m->m_flags & M_EXT) == 0)
857: goto nopages;
858: mlen = MCLBYTES;
859: if (atomic && top == 0) {
1.58 jdolecek 860: len = lmin(MCLBYTES - max_hdr,
1.54 lukem 861: resid);
1.45 tv 862: m->m_data += max_hdr;
863: } else
1.58 jdolecek 864: len = lmin(MCLBYTES, resid);
1.45 tv 865: space -= len;
866: } else {
1.64 thorpej 867: nopages:
868: SOSEND_COUNTER_INCR(&sosend_copy_small);
1.58 jdolecek 869: len = lmin(lmin(mlen, resid), space);
1.45 tv 870: space -= len;
871: /*
872: * For datagram protocols, leave room
873: * for protocol headers in first mbuf.
874: */
875: if (atomic && top == 0 && len < mlen)
876: MH_ALIGN(m, len);
877: }
1.134 christos 878: error = uiomove(mtod(m, void *), (int)len,
1.54 lukem 879: uio);
1.64 thorpej 880: have_data:
1.45 tv 881: resid = uio->uio_resid;
882: m->m_len = len;
883: *mp = m;
884: top->m_pkthdr.len += len;
885: if (error)
886: goto release;
887: mp = &m->m_next;
888: if (resid <= 0) {
889: if (flags & MSG_EOR)
890: top->m_flags |= M_EOR;
891: break;
892: }
893: } while (space > 0 && atomic);
1.108 perry 894:
1.46 sommerfe 895: s = splsoftnet();
896:
897: if (so->so_state & SS_CANTSENDMORE)
898: snderr(EPIPE);
1.45 tv 899:
900: if (dontroute)
901: so->so_options |= SO_DONTROUTE;
902: if (resid > 0)
903: so->so_state |= SS_MORETOCOME;
1.46 sommerfe 904: error = (*so->so_proto->pr_usrreq)(so,
905: (flags & MSG_OOB) ? PRU_SENDOOB : PRU_SEND,
1.114 christos 906: top, addr, control, curlwp); /* XXX */
1.45 tv 907: if (dontroute)
908: so->so_options &= ~SO_DONTROUTE;
909: if (resid > 0)
910: so->so_state &= ~SS_MORETOCOME;
1.46 sommerfe 911: splx(s);
912:
1.45 tv 913: clen = 0;
914: control = 0;
915: top = 0;
916: mp = ⊤
1.1 cgd 917: if (error)
918: goto release;
919: } while (resid && space > 0);
920: } while (resid);
921:
1.54 lukem 922: release:
1.1 cgd 923: sbunlock(&so->so_snd);
1.54 lukem 924: out:
1.1 cgd 925: if (top)
926: m_freem(top);
927: if (control)
928: m_freem(control);
929: return (error);
930: }
931:
932: /*
933: * Implement receive operations on a socket.
934: * We depend on the way that records are added to the sockbuf
935: * by sbappend*. In particular, each record (mbufs linked through m_next)
936: * must begin with an address if the protocol so specifies,
937: * followed by an optional mbuf or mbufs containing ancillary data,
938: * and then zero or more mbufs of data.
939: * In order to avoid blocking network interrupts for the entire time here,
940: * we splx() while doing the actual copy to user space.
941: * Although the sockbuf is locked, new data may still be appended,
942: * and thus we must maintain consistency of the sockbuf during that time.
943: *
944: * The caller may receive the data as a single mbuf chain by supplying
945: * an mbuf **mp0 for use in returning the chain. The uio is then used
946: * only for the count in uio_resid.
947: */
1.3 andrew 948: int
1.54 lukem 949: soreceive(struct socket *so, struct mbuf **paddr, struct uio *uio,
950: struct mbuf **mp0, struct mbuf **controlp, int *flagsp)
1.1 cgd 951: {
1.116 yamt 952: struct lwp *l = curlwp;
1.54 lukem 953: struct mbuf *m, **mp;
954: int flags, len, error, s, offset, moff, type, orig_resid;
1.99 matt 955: const struct protosw *pr;
1.54 lukem 956: struct mbuf *nextrecord;
1.67 he 957: int mbuf_removed = 0;
1.64 thorpej 958:
1.54 lukem 959: pr = so->so_proto;
1.1 cgd 960: mp = mp0;
1.54 lukem 961: type = 0;
962: orig_resid = uio->uio_resid;
1.102 jonathan 963:
1.1 cgd 964: if (paddr)
965: *paddr = 0;
966: if (controlp)
967: *controlp = 0;
968: if (flagsp)
969: flags = *flagsp &~ MSG_EOR;
970: else
971: flags = 0;
1.66 enami 972:
973: if ((flags & MSG_DONTWAIT) == 0)
1.117 yamt 974: sodopendfree();
1.66 enami 975:
1.1 cgd 976: if (flags & MSG_OOB) {
977: m = m_get(M_WAIT, MT_DATA);
1.17 cgd 978: error = (*pr->pr_usrreq)(so, PRU_RCVOOB, m,
1.140 dyoung 979: (struct mbuf *)(long)(flags & MSG_PEEK), NULL, l);
1.1 cgd 980: if (error)
981: goto bad;
982: do {
1.134 christos 983: error = uiomove(mtod(m, void *),
1.1 cgd 984: (int) min(uio->uio_resid, m->m_len), uio);
985: m = m_free(m);
986: } while (uio->uio_resid && error == 0 && m);
1.54 lukem 987: bad:
1.1 cgd 988: if (m)
989: m_freem(m);
990: return (error);
991: }
992: if (mp)
1.140 dyoung 993: *mp = NULL;
1.1 cgd 994: if (so->so_state & SS_ISCONFIRMING && uio->uio_resid)
1.140 dyoung 995: (*pr->pr_usrreq)(so, PRU_RCVD, NULL, NULL, NULL, l);
1.1 cgd 996:
1.54 lukem 997: restart:
1.21 christos 998: if ((error = sblock(&so->so_rcv, SBLOCKWAIT(flags))) != 0)
1.1 cgd 999: return (error);
1.20 mycroft 1000: s = splsoftnet();
1.1 cgd 1001:
1002: m = so->so_rcv.sb_mb;
1003: /*
1004: * If we have less data than requested, block awaiting more
1005: * (subject to any timeout) if:
1.15 mycroft 1006: * 1. the current count is less than the low water mark,
1.1 cgd 1007: * 2. MSG_WAITALL is set, and it is possible to do the entire
1.15 mycroft 1008: * receive operation at once if we block (resid <= hiwat), or
1009: * 3. MSG_DONTWAIT is not set.
1.1 cgd 1010: * If MSG_WAITALL is set but resid is larger than the receive buffer,
1011: * we have to do the receive in sections, and thus risk returning
1012: * a short count if a timeout or signal occurs after we start.
1013: */
1.21 christos 1014: if (m == 0 || (((flags & MSG_DONTWAIT) == 0 &&
1.15 mycroft 1015: so->so_rcv.sb_cc < uio->uio_resid) &&
1.1 cgd 1016: (so->so_rcv.sb_cc < so->so_rcv.sb_lowat ||
1017: ((flags & MSG_WAITALL) && uio->uio_resid <= so->so_rcv.sb_hiwat)) &&
1.21 christos 1018: m->m_nextpkt == 0 && (pr->pr_flags & PR_ATOMIC) == 0)) {
1.1 cgd 1019: #ifdef DIAGNOSTIC
1020: if (m == 0 && so->so_rcv.sb_cc)
1021: panic("receive 1");
1022: #endif
1023: if (so->so_error) {
1024: if (m)
1.15 mycroft 1025: goto dontblock;
1.1 cgd 1026: error = so->so_error;
1027: if ((flags & MSG_PEEK) == 0)
1028: so->so_error = 0;
1029: goto release;
1030: }
1031: if (so->so_state & SS_CANTRCVMORE) {
1032: if (m)
1.15 mycroft 1033: goto dontblock;
1.1 cgd 1034: else
1035: goto release;
1036: }
1037: for (; m; m = m->m_next)
1038: if (m->m_type == MT_OOBDATA || (m->m_flags & M_EOR)) {
1039: m = so->so_rcv.sb_mb;
1040: goto dontblock;
1041: }
1042: if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0 &&
1043: (so->so_proto->pr_flags & PR_CONNREQUIRED)) {
1044: error = ENOTCONN;
1045: goto release;
1046: }
1047: if (uio->uio_resid == 0)
1048: goto release;
1.15 mycroft 1049: if ((so->so_state & SS_NBIO) || (flags & MSG_DONTWAIT)) {
1.1 cgd 1050: error = EWOULDBLOCK;
1051: goto release;
1052: }
1.69 thorpej 1053: SBLASTRECORDCHK(&so->so_rcv, "soreceive sbwait 1");
1054: SBLASTMBUFCHK(&so->so_rcv, "soreceive sbwait 1");
1.1 cgd 1055: sbunlock(&so->so_rcv);
1056: error = sbwait(&so->so_rcv);
1057: splx(s);
1058: if (error)
1059: return (error);
1060: goto restart;
1061: }
1.54 lukem 1062: dontblock:
1.69 thorpej 1063: /*
1064: * On entry here, m points to the first record of the socket buffer.
1065: * While we process the initial mbufs containing address and control
1066: * info, we save a copy of m->m_nextpkt into nextrecord.
1067: */
1.114 christos 1068: if (l)
1069: l->l_proc->p_stats->p_ru.ru_msgrcv++;
1.69 thorpej 1070: KASSERT(m == so->so_rcv.sb_mb);
1071: SBLASTRECORDCHK(&so->so_rcv, "soreceive 1");
1072: SBLASTMBUFCHK(&so->so_rcv, "soreceive 1");
1.1 cgd 1073: nextrecord = m->m_nextpkt;
1074: if (pr->pr_flags & PR_ADDR) {
1075: #ifdef DIAGNOSTIC
1076: if (m->m_type != MT_SONAME)
1077: panic("receive 1a");
1078: #endif
1.3 andrew 1079: orig_resid = 0;
1.1 cgd 1080: if (flags & MSG_PEEK) {
1081: if (paddr)
1082: *paddr = m_copy(m, 0, m->m_len);
1083: m = m->m_next;
1084: } else {
1085: sbfree(&so->so_rcv, m);
1.67 he 1086: mbuf_removed = 1;
1.1 cgd 1087: if (paddr) {
1088: *paddr = m;
1089: so->so_rcv.sb_mb = m->m_next;
1090: m->m_next = 0;
1091: m = so->so_rcv.sb_mb;
1092: } else {
1093: MFREE(m, so->so_rcv.sb_mb);
1094: m = so->so_rcv.sb_mb;
1095: }
1096: }
1097: }
1098: while (m && m->m_type == MT_CONTROL && error == 0) {
1099: if (flags & MSG_PEEK) {
1100: if (controlp)
1101: *controlp = m_copy(m, 0, m->m_len);
1102: m = m->m_next;
1103: } else {
1104: sbfree(&so->so_rcv, m);
1.67 he 1105: mbuf_removed = 1;
1.1 cgd 1106: if (controlp) {
1.102 jonathan 1107: struct domain *dom = pr->pr_domain;
1.114 christos 1108: if (dom->dom_externalize && l &&
1.1 cgd 1109: mtod(m, struct cmsghdr *)->cmsg_type ==
1110: SCM_RIGHTS)
1.114 christos 1111: error = (*dom->dom_externalize)(m, l);
1.1 cgd 1112: *controlp = m;
1113: so->so_rcv.sb_mb = m->m_next;
1114: m->m_next = 0;
1115: m = so->so_rcv.sb_mb;
1116: } else {
1.106 itojun 1117: /*
1118: * Dispose of any SCM_RIGHTS message that went
1119: * through the read path rather than recv.
1120: */
1121: if (pr->pr_domain->dom_dispose &&
1122: mtod(m, struct cmsghdr *)->cmsg_type == SCM_RIGHTS)
1123: (*pr->pr_domain->dom_dispose)(m);
1.1 cgd 1124: MFREE(m, so->so_rcv.sb_mb);
1125: m = so->so_rcv.sb_mb;
1126: }
1127: }
1.3 andrew 1128: if (controlp) {
1129: orig_resid = 0;
1.1 cgd 1130: controlp = &(*controlp)->m_next;
1.3 andrew 1131: }
1.1 cgd 1132: }
1.69 thorpej 1133:
1134: /*
1135: * If m is non-NULL, we have some data to read. From now on,
1136: * make sure to keep sb_lastrecord consistent when working on
1137: * the last packet on the chain (nextrecord == NULL) and we
1138: * change m->m_nextpkt.
1139: */
1.1 cgd 1140: if (m) {
1.69 thorpej 1141: if ((flags & MSG_PEEK) == 0) {
1.1 cgd 1142: m->m_nextpkt = nextrecord;
1.69 thorpej 1143: /*
1144: * If nextrecord == NULL (this is a single chain),
1145: * then sb_lastrecord may not be valid here if m
1146: * was changed earlier.
1147: */
1148: if (nextrecord == NULL) {
1149: KASSERT(so->so_rcv.sb_mb == m);
1150: so->so_rcv.sb_lastrecord = m;
1151: }
1152: }
1.1 cgd 1153: type = m->m_type;
1154: if (type == MT_OOBDATA)
1155: flags |= MSG_OOB;
1.69 thorpej 1156: } else {
1157: if ((flags & MSG_PEEK) == 0) {
1158: KASSERT(so->so_rcv.sb_mb == m);
1159: so->so_rcv.sb_mb = nextrecord;
1.70 thorpej 1160: SB_EMPTY_FIXUP(&so->so_rcv);
1.69 thorpej 1161: }
1.1 cgd 1162: }
1.69 thorpej 1163: SBLASTRECORDCHK(&so->so_rcv, "soreceive 2");
1164: SBLASTMBUFCHK(&so->so_rcv, "soreceive 2");
1165:
1.1 cgd 1166: moff = 0;
1167: offset = 0;
1168: while (m && uio->uio_resid > 0 && error == 0) {
1169: if (m->m_type == MT_OOBDATA) {
1170: if (type != MT_OOBDATA)
1171: break;
1172: } else if (type == MT_OOBDATA)
1173: break;
1174: #ifdef DIAGNOSTIC
1175: else if (m->m_type != MT_DATA && m->m_type != MT_HEADER)
1176: panic("receive 3");
1177: #endif
1178: so->so_state &= ~SS_RCVATMARK;
1179: len = uio->uio_resid;
1180: if (so->so_oobmark && len > so->so_oobmark - offset)
1181: len = so->so_oobmark - offset;
1182: if (len > m->m_len - moff)
1183: len = m->m_len - moff;
1184: /*
1185: * If mp is set, just pass back the mbufs.
1186: * Otherwise copy them out via the uio, then free.
1187: * Sockbuf must be consistent here (points to current mbuf,
1188: * it points to next record) when we drop priority;
1189: * we must note any additions to the sockbuf when we
1190: * block interrupts again.
1191: */
1192: if (mp == 0) {
1.69 thorpej 1193: SBLASTRECORDCHK(&so->so_rcv, "soreceive uiomove");
1194: SBLASTMBUFCHK(&so->so_rcv, "soreceive uiomove");
1.1 cgd 1195: splx(s);
1.134 christos 1196: error = uiomove(mtod(m, char *) + moff, (int)len, uio);
1.20 mycroft 1197: s = splsoftnet();
1.67 he 1198: if (error) {
1199: /*
1200: * If any part of the record has been removed
1201: * (such as the MT_SONAME mbuf, which will
1202: * happen when PR_ADDR, and thus also
1203: * PR_ATOMIC, is set), then drop the entire
1204: * record to maintain the atomicity of the
1205: * receive operation.
1206: *
1207: * This avoids a later panic("receive 1a")
1208: * when compiled with DIAGNOSTIC.
1209: */
1210: if (m && mbuf_removed
1211: && (pr->pr_flags & PR_ATOMIC))
1212: (void) sbdroprecord(&so->so_rcv);
1213:
1.57 jdolecek 1214: goto release;
1.67 he 1215: }
1.1 cgd 1216: } else
1217: uio->uio_resid -= len;
1218: if (len == m->m_len - moff) {
1219: if (m->m_flags & M_EOR)
1220: flags |= MSG_EOR;
1221: if (flags & MSG_PEEK) {
1222: m = m->m_next;
1223: moff = 0;
1224: } else {
1225: nextrecord = m->m_nextpkt;
1226: sbfree(&so->so_rcv, m);
1227: if (mp) {
1228: *mp = m;
1229: mp = &m->m_next;
1230: so->so_rcv.sb_mb = m = m->m_next;
1.140 dyoung 1231: *mp = NULL;
1.1 cgd 1232: } else {
1233: MFREE(m, so->so_rcv.sb_mb);
1234: m = so->so_rcv.sb_mb;
1235: }
1.69 thorpej 1236: /*
1237: * If m != NULL, we also know that
1238: * so->so_rcv.sb_mb != NULL.
1239: */
1240: KASSERT(so->so_rcv.sb_mb == m);
1241: if (m) {
1.1 cgd 1242: m->m_nextpkt = nextrecord;
1.69 thorpej 1243: if (nextrecord == NULL)
1244: so->so_rcv.sb_lastrecord = m;
1245: } else {
1246: so->so_rcv.sb_mb = nextrecord;
1.70 thorpej 1247: SB_EMPTY_FIXUP(&so->so_rcv);
1.69 thorpej 1248: }
1249: SBLASTRECORDCHK(&so->so_rcv, "soreceive 3");
1250: SBLASTMBUFCHK(&so->so_rcv, "soreceive 3");
1.1 cgd 1251: }
1252: } else {
1253: if (flags & MSG_PEEK)
1254: moff += len;
1255: else {
1256: if (mp)
1257: *mp = m_copym(m, 0, len, M_WAIT);
1258: m->m_data += len;
1259: m->m_len -= len;
1260: so->so_rcv.sb_cc -= len;
1261: }
1262: }
1263: if (so->so_oobmark) {
1264: if ((flags & MSG_PEEK) == 0) {
1265: so->so_oobmark -= len;
1266: if (so->so_oobmark == 0) {
1267: so->so_state |= SS_RCVATMARK;
1268: break;
1269: }
1.7 cgd 1270: } else {
1.1 cgd 1271: offset += len;
1.7 cgd 1272: if (offset == so->so_oobmark)
1273: break;
1274: }
1.1 cgd 1275: }
1276: if (flags & MSG_EOR)
1277: break;
1278: /*
1279: * If the MSG_WAITALL flag is set (for non-atomic socket),
1280: * we must not quit until "uio->uio_resid == 0" or an error
1281: * termination. If a signal/timeout occurs, return
1282: * with a short count but without error.
1283: * Keep sockbuf locked against other readers.
1284: */
1285: while (flags & MSG_WAITALL && m == 0 && uio->uio_resid > 0 &&
1.3 andrew 1286: !sosendallatonce(so) && !nextrecord) {
1.1 cgd 1287: if (so->so_error || so->so_state & SS_CANTRCVMORE)
1288: break;
1.68 matt 1289: /*
1290: * If we are peeking and the socket receive buffer is
1291: * full, stop since we can't get more data to peek at.
1292: */
1293: if ((flags & MSG_PEEK) && sbspace(&so->so_rcv) <= 0)
1294: break;
1295: /*
1296: * If we've drained the socket buffer, tell the
1297: * protocol in case it needs to do something to
1298: * get it filled again.
1299: */
1300: if ((pr->pr_flags & PR_WANTRCVD) && so->so_pcb)
1301: (*pr->pr_usrreq)(so, PRU_RCVD,
1.140 dyoung 1302: NULL, (struct mbuf *)(long)flags, NULL, l);
1.69 thorpej 1303: SBLASTRECORDCHK(&so->so_rcv, "soreceive sbwait 2");
1304: SBLASTMBUFCHK(&so->so_rcv, "soreceive sbwait 2");
1.1 cgd 1305: error = sbwait(&so->so_rcv);
1306: if (error) {
1307: sbunlock(&so->so_rcv);
1308: splx(s);
1309: return (0);
1310: }
1.21 christos 1311: if ((m = so->so_rcv.sb_mb) != NULL)
1.1 cgd 1312: nextrecord = m->m_nextpkt;
1313: }
1314: }
1.3 andrew 1315:
1316: if (m && pr->pr_flags & PR_ATOMIC) {
1317: flags |= MSG_TRUNC;
1318: if ((flags & MSG_PEEK) == 0)
1319: (void) sbdroprecord(&so->so_rcv);
1320: }
1.1 cgd 1321: if ((flags & MSG_PEEK) == 0) {
1.69 thorpej 1322: if (m == 0) {
1323: /*
1.70 thorpej 1324: * First part is an inline SB_EMPTY_FIXUP(). Second
1.69 thorpej 1325: * part makes sure sb_lastrecord is up-to-date if
1326: * there is still data in the socket buffer.
1327: */
1.1 cgd 1328: so->so_rcv.sb_mb = nextrecord;
1.69 thorpej 1329: if (so->so_rcv.sb_mb == NULL) {
1330: so->so_rcv.sb_mbtail = NULL;
1331: so->so_rcv.sb_lastrecord = NULL;
1332: } else if (nextrecord->m_nextpkt == NULL)
1333: so->so_rcv.sb_lastrecord = nextrecord;
1334: }
1335: SBLASTRECORDCHK(&so->so_rcv, "soreceive 4");
1336: SBLASTMBUFCHK(&so->so_rcv, "soreceive 4");
1.1 cgd 1337: if (pr->pr_flags & PR_WANTRCVD && so->so_pcb)
1.140 dyoung 1338: (*pr->pr_usrreq)(so, PRU_RCVD, NULL,
1339: (struct mbuf *)(long)flags, NULL, l);
1.1 cgd 1340: }
1.3 andrew 1341: if (orig_resid == uio->uio_resid && orig_resid &&
1342: (flags & MSG_EOR) == 0 && (so->so_state & SS_CANTRCVMORE) == 0) {
1343: sbunlock(&so->so_rcv);
1344: splx(s);
1345: goto restart;
1346: }
1.108 perry 1347:
1.1 cgd 1348: if (flagsp)
1349: *flagsp |= flags;
1.54 lukem 1350: release:
1.1 cgd 1351: sbunlock(&so->so_rcv);
1352: splx(s);
1353: return (error);
1354: }
1355:
1.14 mycroft 1356: int
1.54 lukem 1357: soshutdown(struct socket *so, int how)
1.1 cgd 1358: {
1.99 matt 1359: const struct protosw *pr;
1.34 kleink 1360:
1.54 lukem 1361: pr = so->so_proto;
1.34 kleink 1362: if (!(how == SHUT_RD || how == SHUT_WR || how == SHUT_RDWR))
1363: return (EINVAL);
1.1 cgd 1364:
1.34 kleink 1365: if (how == SHUT_RD || how == SHUT_RDWR)
1.1 cgd 1366: sorflush(so);
1.34 kleink 1367: if (how == SHUT_WR || how == SHUT_RDWR)
1.140 dyoung 1368: return (*pr->pr_usrreq)(so, PRU_SHUTDOWN, NULL,
1369: NULL, NULL, NULL);
1.1 cgd 1370: return (0);
1371: }
1372:
1.14 mycroft 1373: void
1.54 lukem 1374: sorflush(struct socket *so)
1.1 cgd 1375: {
1.54 lukem 1376: struct sockbuf *sb, asb;
1.99 matt 1377: const struct protosw *pr;
1.54 lukem 1378: int s;
1.1 cgd 1379:
1.54 lukem 1380: sb = &so->so_rcv;
1381: pr = so->so_proto;
1.1 cgd 1382: sb->sb_flags |= SB_NOINTR;
1.15 mycroft 1383: (void) sblock(sb, M_WAITOK);
1.56 thorpej 1384: s = splnet();
1.1 cgd 1385: socantrcvmore(so);
1386: sbunlock(sb);
1387: asb = *sb;
1.86 wrstuden 1388: /*
1389: * Clear most of the sockbuf structure, but leave some of the
1390: * fields valid.
1391: */
1392: memset(&sb->sb_startzero, 0,
1393: sizeof(*sb) - offsetof(struct sockbuf, sb_startzero));
1.1 cgd 1394: splx(s);
1395: if (pr->pr_flags & PR_RIGHTS && pr->pr_domain->dom_dispose)
1396: (*pr->pr_domain->dom_dispose)(asb.sb_mb);
1.98 christos 1397: sbrelease(&asb, so);
1.1 cgd 1398: }
1399:
1.14 mycroft 1400: int
1.54 lukem 1401: sosetopt(struct socket *so, int level, int optname, struct mbuf *m0)
1.1 cgd 1402: {
1.54 lukem 1403: int error;
1404: struct mbuf *m;
1.130 christos 1405: struct linger *l;
1.140.6.1! jmcneill 1406: struct sockbuf *sb;
1.1 cgd 1407:
1.54 lukem 1408: error = 0;
1409: m = m0;
1.1 cgd 1410: if (level != SOL_SOCKET) {
1411: if (so->so_proto && so->so_proto->pr_ctloutput)
1412: return ((*so->so_proto->pr_ctloutput)
1413: (PRCO_SETOPT, so, level, optname, &m0));
1414: error = ENOPROTOOPT;
1415: } else {
1416: switch (optname) {
1417:
1418: case SO_LINGER:
1.36 perry 1419: if (m == NULL || m->m_len != sizeof(struct linger)) {
1.1 cgd 1420: error = EINVAL;
1421: goto bad;
1422: }
1.130 christos 1423: l = mtod(m, struct linger *);
1.131 christos 1424: if (l->l_linger < 0 || l->l_linger > USHRT_MAX ||
1425: l->l_linger > (INT_MAX / hz)) {
1.112 nathanw 1426: error = EDOM;
1427: goto bad;
1428: }
1.130 christos 1429: so->so_linger = l->l_linger;
1430: if (l->l_onoff)
1431: so->so_options |= SO_LINGER;
1432: else
1433: so->so_options &= ~SO_LINGER;
1434: break;
1.1 cgd 1435:
1436: case SO_DEBUG:
1437: case SO_KEEPALIVE:
1438: case SO_DONTROUTE:
1439: case SO_USELOOPBACK:
1440: case SO_BROADCAST:
1441: case SO_REUSEADDR:
1.15 mycroft 1442: case SO_REUSEPORT:
1.1 cgd 1443: case SO_OOBINLINE:
1.26 thorpej 1444: case SO_TIMESTAMP:
1.36 perry 1445: if (m == NULL || m->m_len < sizeof(int)) {
1.1 cgd 1446: error = EINVAL;
1447: goto bad;
1448: }
1449: if (*mtod(m, int *))
1450: so->so_options |= optname;
1451: else
1452: so->so_options &= ~optname;
1453: break;
1454:
1455: case SO_SNDBUF:
1456: case SO_RCVBUF:
1457: case SO_SNDLOWAT:
1458: case SO_RCVLOWAT:
1.28 thorpej 1459: {
1460: int optval;
1461:
1.36 perry 1462: if (m == NULL || m->m_len < sizeof(int)) {
1.1 cgd 1463: error = EINVAL;
1464: goto bad;
1465: }
1.28 thorpej 1466:
1467: /*
1468: * Values < 1 make no sense for any of these
1469: * options, so disallow them.
1470: */
1471: optval = *mtod(m, int *);
1472: if (optval < 1) {
1473: error = EINVAL;
1474: goto bad;
1475: }
1476:
1.1 cgd 1477: switch (optname) {
1478:
1479: case SO_SNDBUF:
1480: case SO_RCVBUF:
1.140.6.1! jmcneill 1481: sb = (optname == SO_SNDBUF) ?
! 1482: &so->so_snd : &so->so_rcv;
! 1483: if (sbreserve(sb, (u_long)optval, so) == 0) {
1.1 cgd 1484: error = ENOBUFS;
1485: goto bad;
1486: }
1.140.6.1! jmcneill 1487: sb->sb_flags &= ~SB_AUTOSIZE;
1.1 cgd 1488: break;
1489:
1.28 thorpej 1490: /*
1491: * Make sure the low-water is never greater than
1492: * the high-water.
1493: */
1.1 cgd 1494: case SO_SNDLOWAT:
1.28 thorpej 1495: so->so_snd.sb_lowat =
1496: (optval > so->so_snd.sb_hiwat) ?
1497: so->so_snd.sb_hiwat : optval;
1.1 cgd 1498: break;
1499: case SO_RCVLOWAT:
1.28 thorpej 1500: so->so_rcv.sb_lowat =
1501: (optval > so->so_rcv.sb_hiwat) ?
1502: so->so_rcv.sb_hiwat : optval;
1.1 cgd 1503: break;
1504: }
1505: break;
1.28 thorpej 1506: }
1.1 cgd 1507:
1508: case SO_SNDTIMEO:
1509: case SO_RCVTIMEO:
1510: {
1511: struct timeval *tv;
1.104 yamt 1512: int val;
1.1 cgd 1513:
1.36 perry 1514: if (m == NULL || m->m_len < sizeof(*tv)) {
1.1 cgd 1515: error = EINVAL;
1516: goto bad;
1517: }
1518: tv = mtod(m, struct timeval *);
1.104 yamt 1519: if (tv->tv_sec > (INT_MAX - tv->tv_usec / tick) / hz) {
1.1 cgd 1520: error = EDOM;
1521: goto bad;
1522: }
1523: val = tv->tv_sec * hz + tv->tv_usec / tick;
1.74 itojun 1524: if (val == 0 && tv->tv_usec != 0)
1525: val = 1;
1.1 cgd 1526:
1527: switch (optname) {
1528:
1529: case SO_SNDTIMEO:
1530: so->so_snd.sb_timeo = val;
1531: break;
1532: case SO_RCVTIMEO:
1533: so->so_rcv.sb_timeo = val;
1534: break;
1535: }
1536: break;
1537: }
1538:
1539: default:
1540: error = ENOPROTOOPT;
1541: break;
1542: }
1.15 mycroft 1543: if (error == 0 && so->so_proto && so->so_proto->pr_ctloutput) {
1544: (void) ((*so->so_proto->pr_ctloutput)
1545: (PRCO_SETOPT, so, level, optname, &m0));
1546: m = NULL; /* freed by protocol */
1547: }
1.1 cgd 1548: }
1.54 lukem 1549: bad:
1.1 cgd 1550: if (m)
1551: (void) m_free(m);
1552: return (error);
1553: }
1554:
1.14 mycroft 1555: int
1.54 lukem 1556: sogetopt(struct socket *so, int level, int optname, struct mbuf **mp)
1.1 cgd 1557: {
1.54 lukem 1558: struct mbuf *m;
1.1 cgd 1559:
1560: if (level != SOL_SOCKET) {
1561: if (so->so_proto && so->so_proto->pr_ctloutput) {
1562: return ((*so->so_proto->pr_ctloutput)
1563: (PRCO_GETOPT, so, level, optname, mp));
1564: } else
1565: return (ENOPROTOOPT);
1566: } else {
1567: m = m_get(M_WAIT, MT_SOOPTS);
1.36 perry 1568: m->m_len = sizeof(int);
1.1 cgd 1569:
1570: switch (optname) {
1571:
1572: case SO_LINGER:
1.36 perry 1573: m->m_len = sizeof(struct linger);
1.1 cgd 1574: mtod(m, struct linger *)->l_onoff =
1.131 christos 1575: (so->so_options & SO_LINGER) ? 1 : 0;
1.1 cgd 1576: mtod(m, struct linger *)->l_linger = so->so_linger;
1577: break;
1578:
1579: case SO_USELOOPBACK:
1580: case SO_DONTROUTE:
1581: case SO_DEBUG:
1582: case SO_KEEPALIVE:
1583: case SO_REUSEADDR:
1.15 mycroft 1584: case SO_REUSEPORT:
1.1 cgd 1585: case SO_BROADCAST:
1586: case SO_OOBINLINE:
1.26 thorpej 1587: case SO_TIMESTAMP:
1.131 christos 1588: *mtod(m, int *) = (so->so_options & optname) ? 1 : 0;
1.1 cgd 1589: break;
1590:
1591: case SO_TYPE:
1592: *mtod(m, int *) = so->so_type;
1593: break;
1594:
1595: case SO_ERROR:
1596: *mtod(m, int *) = so->so_error;
1597: so->so_error = 0;
1598: break;
1599:
1600: case SO_SNDBUF:
1601: *mtod(m, int *) = so->so_snd.sb_hiwat;
1602: break;
1603:
1604: case SO_RCVBUF:
1605: *mtod(m, int *) = so->so_rcv.sb_hiwat;
1606: break;
1607:
1608: case SO_SNDLOWAT:
1609: *mtod(m, int *) = so->so_snd.sb_lowat;
1610: break;
1611:
1612: case SO_RCVLOWAT:
1613: *mtod(m, int *) = so->so_rcv.sb_lowat;
1614: break;
1615:
1616: case SO_SNDTIMEO:
1617: case SO_RCVTIMEO:
1618: {
1619: int val = (optname == SO_SNDTIMEO ?
1620: so->so_snd.sb_timeo : so->so_rcv.sb_timeo);
1621:
1622: m->m_len = sizeof(struct timeval);
1623: mtod(m, struct timeval *)->tv_sec = val / hz;
1624: mtod(m, struct timeval *)->tv_usec =
1.27 kleink 1625: (val % hz) * tick;
1.1 cgd 1626: break;
1627: }
1628:
1.107 darrenr 1629: case SO_OVERFLOWED:
1630: *mtod(m, int *) = so->so_rcv.sb_overflowed;
1631: break;
1632:
1.1 cgd 1633: default:
1634: (void)m_free(m);
1635: return (ENOPROTOOPT);
1636: }
1637: *mp = m;
1638: return (0);
1639: }
1640: }
1641:
1.14 mycroft 1642: void
1.54 lukem 1643: sohasoutofband(struct socket *so)
1.1 cgd 1644: {
1.90 christos 1645: fownsignal(so->so_pgid, SIGURG, POLL_PRI, POLLPRI|POLLRDBAND, so);
1.2 cgd 1646: selwakeup(&so->so_rcv.sb_sel);
1.1 cgd 1647: }
1.72 jdolecek 1648:
1649: static void
1650: filt_sordetach(struct knote *kn)
1651: {
1652: struct socket *so;
1653:
1654: so = (struct socket *)kn->kn_fp->f_data;
1.73 christos 1655: SLIST_REMOVE(&so->so_rcv.sb_sel.sel_klist, kn, knote, kn_selnext);
1656: if (SLIST_EMPTY(&so->so_rcv.sb_sel.sel_klist))
1.72 jdolecek 1657: so->so_rcv.sb_flags &= ~SB_KNOTE;
1658: }
1659:
1660: /*ARGSUSED*/
1661: static int
1.129 yamt 1662: filt_soread(struct knote *kn, long hint)
1.72 jdolecek 1663: {
1664: struct socket *so;
1665:
1666: so = (struct socket *)kn->kn_fp->f_data;
1667: kn->kn_data = so->so_rcv.sb_cc;
1668: if (so->so_state & SS_CANTRCVMORE) {
1.108 perry 1669: kn->kn_flags |= EV_EOF;
1.72 jdolecek 1670: kn->kn_fflags = so->so_error;
1671: return (1);
1672: }
1673: if (so->so_error) /* temporary udp error */
1674: return (1);
1675: if (kn->kn_sfflags & NOTE_LOWAT)
1676: return (kn->kn_data >= kn->kn_sdata);
1677: return (kn->kn_data >= so->so_rcv.sb_lowat);
1678: }
1679:
1680: static void
1681: filt_sowdetach(struct knote *kn)
1682: {
1683: struct socket *so;
1684:
1685: so = (struct socket *)kn->kn_fp->f_data;
1.73 christos 1686: SLIST_REMOVE(&so->so_snd.sb_sel.sel_klist, kn, knote, kn_selnext);
1687: if (SLIST_EMPTY(&so->so_snd.sb_sel.sel_klist))
1.72 jdolecek 1688: so->so_snd.sb_flags &= ~SB_KNOTE;
1689: }
1690:
1691: /*ARGSUSED*/
1692: static int
1.129 yamt 1693: filt_sowrite(struct knote *kn, long hint)
1.72 jdolecek 1694: {
1695: struct socket *so;
1696:
1697: so = (struct socket *)kn->kn_fp->f_data;
1698: kn->kn_data = sbspace(&so->so_snd);
1699: if (so->so_state & SS_CANTSENDMORE) {
1.108 perry 1700: kn->kn_flags |= EV_EOF;
1.72 jdolecek 1701: kn->kn_fflags = so->so_error;
1702: return (1);
1703: }
1704: if (so->so_error) /* temporary udp error */
1705: return (1);
1706: if (((so->so_state & SS_ISCONNECTED) == 0) &&
1707: (so->so_proto->pr_flags & PR_CONNREQUIRED))
1708: return (0);
1709: if (kn->kn_sfflags & NOTE_LOWAT)
1710: return (kn->kn_data >= kn->kn_sdata);
1711: return (kn->kn_data >= so->so_snd.sb_lowat);
1712: }
1713:
1714: /*ARGSUSED*/
1715: static int
1.129 yamt 1716: filt_solisten(struct knote *kn, long hint)
1.72 jdolecek 1717: {
1718: struct socket *so;
1719:
1720: so = (struct socket *)kn->kn_fp->f_data;
1721:
1722: /*
1723: * Set kn_data to number of incoming connections, not
1724: * counting partial (incomplete) connections.
1.108 perry 1725: */
1.72 jdolecek 1726: kn->kn_data = so->so_qlen;
1727: return (kn->kn_data > 0);
1728: }
1729:
1730: static const struct filterops solisten_filtops =
1731: { 1, NULL, filt_sordetach, filt_solisten };
1732: static const struct filterops soread_filtops =
1733: { 1, NULL, filt_sordetach, filt_soread };
1734: static const struct filterops sowrite_filtops =
1735: { 1, NULL, filt_sowdetach, filt_sowrite };
1736:
1737: int
1.129 yamt 1738: soo_kqfilter(struct file *fp, struct knote *kn)
1.72 jdolecek 1739: {
1740: struct socket *so;
1741: struct sockbuf *sb;
1742:
1743: so = (struct socket *)kn->kn_fp->f_data;
1744: switch (kn->kn_filter) {
1745: case EVFILT_READ:
1746: if (so->so_options & SO_ACCEPTCONN)
1747: kn->kn_fop = &solisten_filtops;
1748: else
1749: kn->kn_fop = &soread_filtops;
1750: sb = &so->so_rcv;
1751: break;
1752: case EVFILT_WRITE:
1753: kn->kn_fop = &sowrite_filtops;
1754: sb = &so->so_snd;
1755: break;
1756: default:
1757: return (1);
1758: }
1.73 christos 1759: SLIST_INSERT_HEAD(&sb->sb_sel.sel_klist, kn, kn_selnext);
1.72 jdolecek 1760: sb->sb_flags |= SB_KNOTE;
1761: return (0);
1762: }
1763:
1.94 yamt 1764: #include <sys/sysctl.h>
1765:
1766: static int sysctl_kern_somaxkva(SYSCTLFN_PROTO);
1767:
1768: /*
1769: * sysctl helper routine for kern.somaxkva. ensures that the given
1770: * value is not too small.
1771: * (XXX should we maybe make sure it's not too large as well?)
1772: */
1773: static int
1774: sysctl_kern_somaxkva(SYSCTLFN_ARGS)
1775: {
1776: int error, new_somaxkva;
1777: struct sysctlnode node;
1778:
1779: new_somaxkva = somaxkva;
1780: node = *rnode;
1781: node.sysctl_data = &new_somaxkva;
1782: error = sysctl_lookup(SYSCTLFN_CALL(&node));
1783: if (error || newp == NULL)
1784: return (error);
1785:
1786: if (new_somaxkva < (16 * 1024 * 1024)) /* sanity */
1787: return (EINVAL);
1788:
1.136 ad 1789: mutex_enter(&so_pendfree_lock);
1.94 yamt 1790: somaxkva = new_somaxkva;
1.136 ad 1791: cv_broadcast(&socurkva_cv);
1792: mutex_exit(&so_pendfree_lock);
1.94 yamt 1793:
1794: return (error);
1795: }
1796:
1797: SYSCTL_SETUP(sysctl_kern_somaxkva_setup, "sysctl kern.somaxkva setup")
1798: {
1799:
1.97 atatat 1800: sysctl_createv(clog, 0, NULL, NULL,
1801: CTLFLAG_PERMANENT,
1802: CTLTYPE_NODE, "kern", NULL,
1803: NULL, 0, NULL, 0,
1804: CTL_KERN, CTL_EOL);
1805:
1806: sysctl_createv(clog, 0, NULL, NULL,
1807: CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1.103 atatat 1808: CTLTYPE_INT, "somaxkva",
1809: SYSCTL_DESCR("Maximum amount of kernel memory to be "
1810: "used for socket buffers"),
1.94 yamt 1811: sysctl_kern_somaxkva, 0, NULL, 0,
1812: CTL_KERN, KERN_SOMAXKVA, CTL_EOL);
1813: }
CVSweb <webmaster@jp.NetBSD.org>