Annotation of src/sys/kern/uipc_socket.c, Revision 1.139
1.139 ! yamt 1: /* $NetBSD$ */
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>
1.139 ! yamt 71: __KERNEL_RCSID(0, "$NetBSD$");
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));
464: if (error)
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);
1.120 ginsbach 471: if (prp == 0) {
472: /* no support for domain */
473: if (pffinddomain(dom) == 0)
474: return (EAFNOSUPPORT);
475: /* no support for socket type */
476: if (proto == 0 && type != 0)
477: return (EPROTOTYPE);
478: return (EPROTONOSUPPORT);
479: }
480: if (prp->pr_usrreq == 0)
1.1 cgd 481: return (EPROTONOSUPPORT);
482: if (prp->pr_type != type)
483: return (EPROTOTYPE);
1.39 matt 484: s = splsoftnet();
1.37 thorpej 485: so = pool_get(&socket_pool, PR_WAITOK);
1.134 christos 486: memset((void *)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);
1.22 mycroft 500: error = (*prp->pr_usrreq)(so, PRU_ATTACH, (struct mbuf *)0,
1.114 christos 501: (struct mbuf *)(long)proto, (struct mbuf *)0, l);
1.1 cgd 502: if (error) {
503: so->so_state |= SS_NOFDREF;
504: sofree(so);
1.39 matt 505: splx(s);
1.1 cgd 506: return (error);
507: }
1.39 matt 508: splx(s);
1.1 cgd 509: *aso = so;
510: return (0);
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();
1.22 mycroft 519: error = (*so->so_proto->pr_usrreq)(so, PRU_BIND, (struct mbuf *)0,
1.114 christos 520: nam, (struct mbuf *)0, l);
1.1 cgd 521: splx(s);
522: return (error);
523: }
524:
1.3 andrew 525: int
1.54 lukem 526: solisten(struct socket *so, int backlog)
1.1 cgd 527: {
1.54 lukem 528: int s, error;
1.1 cgd 529:
1.54 lukem 530: s = splsoftnet();
1.22 mycroft 531: error = (*so->so_proto->pr_usrreq)(so, PRU_LISTEN, (struct mbuf *)0,
1.114 christos 532: (struct mbuf *)0, (struct mbuf *)0, (struct lwp *)0);
1.1 cgd 533: if (error) {
534: splx(s);
535: return (error);
536: }
1.63 matt 537: if (TAILQ_EMPTY(&so->so_q))
1.1 cgd 538: so->so_options |= SO_ACCEPTCONN;
539: if (backlog < 0)
540: backlog = 0;
1.49 jonathan 541: so->so_qlimit = min(backlog, somaxconn);
1.1 cgd 542: splx(s);
543: return (0);
544: }
545:
1.21 christos 546: void
1.54 lukem 547: sofree(struct socket *so)
1.1 cgd 548: {
549:
1.43 mycroft 550: if (so->so_pcb || (so->so_state & SS_NOFDREF) == 0)
1.1 cgd 551: return;
1.43 mycroft 552: if (so->so_head) {
553: /*
554: * We must not decommission a socket that's on the accept(2)
555: * queue. If we do, then accept(2) may hang after select(2)
556: * indicated that the listening socket was ready.
557: */
558: if (!soqremque(so, 0))
559: return;
560: }
1.98 christos 561: if (so->so_rcv.sb_hiwat)
1.110 christos 562: (void)chgsbsize(so->so_uidinfo, &so->so_rcv.sb_hiwat, 0,
1.98 christos 563: RLIM_INFINITY);
564: if (so->so_snd.sb_hiwat)
1.110 christos 565: (void)chgsbsize(so->so_uidinfo, &so->so_snd.sb_hiwat, 0,
1.98 christos 566: RLIM_INFINITY);
567: sbrelease(&so->so_snd, so);
1.1 cgd 568: sorflush(so);
1.37 thorpej 569: pool_put(&socket_pool, so);
1.1 cgd 570: }
571:
572: /*
573: * Close a socket on last file table reference removal.
574: * Initiate disconnect if connected.
575: * Free socket when disconnect complete.
576: */
1.3 andrew 577: int
1.54 lukem 578: soclose(struct socket *so)
1.1 cgd 579: {
1.54 lukem 580: struct socket *so2;
581: int s, error;
1.1 cgd 582:
1.54 lukem 583: error = 0;
584: s = splsoftnet(); /* conservative */
1.1 cgd 585: if (so->so_options & SO_ACCEPTCONN) {
1.63 matt 586: while ((so2 = TAILQ_FIRST(&so->so_q0)) != 0) {
1.42 mycroft 587: (void) soqremque(so2, 0);
1.41 mycroft 588: (void) soabort(so2);
589: }
1.63 matt 590: while ((so2 = TAILQ_FIRST(&so->so_q)) != 0) {
1.42 mycroft 591: (void) soqremque(so2, 1);
1.41 mycroft 592: (void) soabort(so2);
593: }
1.1 cgd 594: }
595: if (so->so_pcb == 0)
596: goto discard;
597: if (so->so_state & SS_ISCONNECTED) {
598: if ((so->so_state & SS_ISDISCONNECTING) == 0) {
599: error = sodisconnect(so);
600: if (error)
601: goto drop;
602: }
603: if (so->so_options & SO_LINGER) {
604: if ((so->so_state & SS_ISDISCONNECTING) &&
605: (so->so_state & SS_NBIO))
606: goto drop;
1.21 christos 607: while (so->so_state & SS_ISCONNECTED) {
1.134 christos 608: error = tsleep((void *)&so->so_timeo,
1.21 christos 609: PSOCK | PCATCH, netcls,
1.30 thorpej 610: so->so_linger * hz);
1.21 christos 611: if (error)
1.1 cgd 612: break;
1.21 christos 613: }
1.1 cgd 614: }
615: }
1.54 lukem 616: drop:
1.1 cgd 617: if (so->so_pcb) {
1.22 mycroft 618: int error2 = (*so->so_proto->pr_usrreq)(so, PRU_DETACH,
619: (struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0,
1.114 christos 620: (struct lwp *)0);
1.1 cgd 621: if (error == 0)
622: error = error2;
623: }
1.54 lukem 624: discard:
1.1 cgd 625: if (so->so_state & SS_NOFDREF)
626: panic("soclose: NOFDREF");
627: so->so_state |= SS_NOFDREF;
628: sofree(so);
629: splx(s);
630: return (error);
631: }
632:
633: /*
1.20 mycroft 634: * Must be called at splsoftnet...
1.1 cgd 635: */
1.3 andrew 636: int
1.54 lukem 637: soabort(struct socket *so)
1.1 cgd 638: {
1.139 ! yamt 639: int error;
1.1 cgd 640:
1.139 ! yamt 641: KASSERT(so->so_head == NULL);
! 642: error = (*so->so_proto->pr_usrreq)(so, PRU_ABORT, (struct mbuf *)0,
1.114 christos 643: (struct mbuf *)0, (struct mbuf *)0, (struct lwp *)0);
1.139 ! yamt 644: if (error) {
! 645: sofree(so);
! 646: }
! 647: return error;
1.1 cgd 648: }
649:
1.3 andrew 650: int
1.54 lukem 651: soaccept(struct socket *so, struct mbuf *nam)
1.1 cgd 652: {
1.54 lukem 653: int s, error;
1.1 cgd 654:
1.54 lukem 655: error = 0;
656: s = splsoftnet();
1.1 cgd 657: if ((so->so_state & SS_NOFDREF) == 0)
658: panic("soaccept: !NOFDREF");
659: so->so_state &= ~SS_NOFDREF;
1.55 thorpej 660: if ((so->so_state & SS_ISDISCONNECTED) == 0 ||
661: (so->so_proto->pr_flags & PR_ABRTACPTDIS) == 0)
1.41 mycroft 662: error = (*so->so_proto->pr_usrreq)(so, PRU_ACCEPT,
1.114 christos 663: (struct mbuf *)0, nam, (struct mbuf *)0, (struct lwp *)0);
1.41 mycroft 664: else
1.53 itojun 665: error = ECONNABORTED;
1.52 itojun 666:
1.1 cgd 667: splx(s);
668: return (error);
669: }
670:
1.3 andrew 671: int
1.114 christos 672: soconnect(struct socket *so, struct mbuf *nam, struct lwp *l)
1.1 cgd 673: {
1.54 lukem 674: int s, error;
1.1 cgd 675:
676: if (so->so_options & SO_ACCEPTCONN)
677: return (EOPNOTSUPP);
1.20 mycroft 678: s = splsoftnet();
1.1 cgd 679: /*
680: * If protocol is connection-based, can only connect once.
681: * Otherwise, if connected, try to disconnect first.
682: * This allows user to disconnect by connecting to, e.g.,
683: * a null address.
684: */
685: if (so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING) &&
686: ((so->so_proto->pr_flags & PR_CONNREQUIRED) ||
687: (error = sodisconnect(so))))
688: error = EISCONN;
689: else
690: error = (*so->so_proto->pr_usrreq)(so, PRU_CONNECT,
1.114 christos 691: (struct mbuf *)0, nam, (struct mbuf *)0, l);
1.1 cgd 692: splx(s);
693: return (error);
694: }
695:
1.3 andrew 696: int
1.54 lukem 697: soconnect2(struct socket *so1, struct socket *so2)
1.1 cgd 698: {
1.54 lukem 699: int s, error;
1.1 cgd 700:
1.54 lukem 701: s = splsoftnet();
1.22 mycroft 702: error = (*so1->so_proto->pr_usrreq)(so1, PRU_CONNECT2,
703: (struct mbuf *)0, (struct mbuf *)so2, (struct mbuf *)0,
1.114 christos 704: (struct lwp *)0);
1.1 cgd 705: splx(s);
706: return (error);
707: }
708:
1.3 andrew 709: int
1.54 lukem 710: sodisconnect(struct socket *so)
1.1 cgd 711: {
1.54 lukem 712: int s, error;
1.1 cgd 713:
1.54 lukem 714: s = splsoftnet();
1.1 cgd 715: if ((so->so_state & SS_ISCONNECTED) == 0) {
716: error = ENOTCONN;
717: goto bad;
718: }
719: if (so->so_state & SS_ISDISCONNECTING) {
720: error = EALREADY;
721: goto bad;
722: }
1.22 mycroft 723: error = (*so->so_proto->pr_usrreq)(so, PRU_DISCONNECT,
724: (struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0,
1.114 christos 725: (struct lwp *)0);
1.54 lukem 726: bad:
1.1 cgd 727: splx(s);
1.117 yamt 728: sodopendfree();
1.1 cgd 729: return (error);
730: }
731:
1.15 mycroft 732: #define SBLOCKWAIT(f) (((f) & MSG_DONTWAIT) ? M_NOWAIT : M_WAITOK)
1.1 cgd 733: /*
734: * Send on a socket.
735: * If send must go all at once and message is larger than
736: * send buffering, then hard error.
737: * Lock against other senders.
738: * If must go all at once and not enough room now, then
739: * inform user that this would block and do nothing.
740: * Otherwise, if nonblocking, send as much as possible.
741: * The data to be sent is described by "uio" if nonzero,
742: * otherwise by the mbuf chain "top" (which must be null
743: * if uio is not). Data provided in mbuf chain must be small
744: * enough to send all at once.
745: *
746: * Returns nonzero on error, timeout or signal; callers
747: * must check for short counts if EINTR/ERESTART are returned.
748: * Data and control buffers are freed on return.
749: */
1.3 andrew 750: int
1.54 lukem 751: sosend(struct socket *so, struct mbuf *addr, struct uio *uio, struct mbuf *top,
1.114 christos 752: struct mbuf *control, int flags, struct lwp *l)
1.1 cgd 753: {
1.54 lukem 754: struct mbuf **mp, *m;
1.114 christos 755: struct proc *p;
1.58 jdolecek 756: long space, len, resid, clen, mlen;
757: int error, s, dontroute, atomic;
1.54 lukem 758:
1.114 christos 759: p = l->l_proc;
1.117 yamt 760: sodopendfree();
1.64 thorpej 761:
1.54 lukem 762: clen = 0;
763: atomic = sosendallatonce(so) || top;
1.1 cgd 764: if (uio)
765: resid = uio->uio_resid;
766: else
767: resid = top->m_pkthdr.len;
1.7 cgd 768: /*
769: * In theory resid should be unsigned.
770: * However, space must be signed, as it might be less than 0
771: * if we over-committed, and we must use a signed comparison
772: * of space and resid. On the other hand, a negative resid
773: * causes us to loop sending 0-length segments to the protocol.
774: */
1.29 mycroft 775: if (resid < 0) {
776: error = EINVAL;
777: goto out;
778: }
1.1 cgd 779: dontroute =
780: (flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 &&
781: (so->so_proto->pr_flags & PR_ATOMIC);
1.102 jonathan 782: if (p)
783: p->p_stats->p_ru.ru_msgsnd++;
1.1 cgd 784: if (control)
785: clen = control->m_len;
786: #define snderr(errno) { error = errno; splx(s); goto release; }
787:
1.54 lukem 788: restart:
1.21 christos 789: if ((error = sblock(&so->so_snd, SBLOCKWAIT(flags))) != 0)
1.1 cgd 790: goto out;
791: do {
1.20 mycroft 792: s = splsoftnet();
1.1 cgd 793: if (so->so_state & SS_CANTSENDMORE)
794: snderr(EPIPE);
1.48 thorpej 795: if (so->so_error) {
796: error = so->so_error;
797: so->so_error = 0;
798: splx(s);
799: goto release;
800: }
1.1 cgd 801: if ((so->so_state & SS_ISCONNECTED) == 0) {
802: if (so->so_proto->pr_flags & PR_CONNREQUIRED) {
803: if ((so->so_state & SS_ISCONFIRMING) == 0 &&
804: !(resid == 0 && clen != 0))
805: snderr(ENOTCONN);
806: } else if (addr == 0)
807: snderr(EDESTADDRREQ);
808: }
809: space = sbspace(&so->so_snd);
810: if (flags & MSG_OOB)
811: space += 1024;
1.21 christos 812: if ((atomic && resid > so->so_snd.sb_hiwat) ||
1.11 mycroft 813: clen > so->so_snd.sb_hiwat)
814: snderr(EMSGSIZE);
1.96 mycroft 815: if (space < resid + clen &&
1.1 cgd 816: (atomic || space < so->so_snd.sb_lowat || space < clen)) {
817: if (so->so_state & SS_NBIO)
818: snderr(EWOULDBLOCK);
819: sbunlock(&so->so_snd);
820: error = sbwait(&so->so_snd);
821: splx(s);
822: if (error)
823: goto out;
824: goto restart;
825: }
826: splx(s);
827: mp = ⊤
828: space -= clen;
829: do {
1.45 tv 830: if (uio == NULL) {
831: /*
832: * Data is prepackaged in "top".
833: */
834: resid = 0;
835: if (flags & MSG_EOR)
836: top->m_flags |= M_EOR;
837: } else do {
838: if (top == 0) {
1.78 matt 839: m = m_gethdr(M_WAIT, MT_DATA);
1.45 tv 840: mlen = MHLEN;
841: m->m_pkthdr.len = 0;
842: m->m_pkthdr.rcvif = (struct ifnet *)0;
843: } else {
1.78 matt 844: m = m_get(M_WAIT, MT_DATA);
1.45 tv 845: mlen = MLEN;
846: }
1.78 matt 847: MCLAIM(m, so->so_snd.sb_mowner);
1.121 yamt 848: if (sock_loan_thresh >= 0 &&
849: uio->uio_iov->iov_len >= sock_loan_thresh &&
850: space >= sock_loan_thresh &&
1.64 thorpej 851: (len = sosend_loan(so, uio, m,
852: space)) != 0) {
853: SOSEND_COUNTER_INCR(&sosend_loan_big);
854: space -= len;
855: goto have_data;
856: }
1.45 tv 857: if (resid >= MINCLSIZE && space >= MCLBYTES) {
1.64 thorpej 858: SOSEND_COUNTER_INCR(&sosend_copy_big);
1.78 matt 859: m_clget(m, M_WAIT);
1.45 tv 860: if ((m->m_flags & M_EXT) == 0)
861: goto nopages;
862: mlen = MCLBYTES;
863: if (atomic && top == 0) {
1.58 jdolecek 864: len = lmin(MCLBYTES - max_hdr,
1.54 lukem 865: resid);
1.45 tv 866: m->m_data += max_hdr;
867: } else
1.58 jdolecek 868: len = lmin(MCLBYTES, resid);
1.45 tv 869: space -= len;
870: } else {
1.64 thorpej 871: nopages:
872: SOSEND_COUNTER_INCR(&sosend_copy_small);
1.58 jdolecek 873: len = lmin(lmin(mlen, resid), space);
1.45 tv 874: space -= len;
875: /*
876: * For datagram protocols, leave room
877: * for protocol headers in first mbuf.
878: */
879: if (atomic && top == 0 && len < mlen)
880: MH_ALIGN(m, len);
881: }
1.134 christos 882: error = uiomove(mtod(m, void *), (int)len,
1.54 lukem 883: uio);
1.64 thorpej 884: have_data:
1.45 tv 885: resid = uio->uio_resid;
886: m->m_len = len;
887: *mp = m;
888: top->m_pkthdr.len += len;
889: if (error)
890: goto release;
891: mp = &m->m_next;
892: if (resid <= 0) {
893: if (flags & MSG_EOR)
894: top->m_flags |= M_EOR;
895: break;
896: }
897: } while (space > 0 && atomic);
1.108 perry 898:
1.46 sommerfe 899: s = splsoftnet();
900:
901: if (so->so_state & SS_CANTSENDMORE)
902: snderr(EPIPE);
1.45 tv 903:
904: if (dontroute)
905: so->so_options |= SO_DONTROUTE;
906: if (resid > 0)
907: so->so_state |= SS_MORETOCOME;
1.46 sommerfe 908: error = (*so->so_proto->pr_usrreq)(so,
909: (flags & MSG_OOB) ? PRU_SENDOOB : PRU_SEND,
1.114 christos 910: top, addr, control, curlwp); /* XXX */
1.45 tv 911: if (dontroute)
912: so->so_options &= ~SO_DONTROUTE;
913: if (resid > 0)
914: so->so_state &= ~SS_MORETOCOME;
1.46 sommerfe 915: splx(s);
916:
1.45 tv 917: clen = 0;
918: control = 0;
919: top = 0;
920: mp = ⊤
1.1 cgd 921: if (error)
922: goto release;
923: } while (resid && space > 0);
924: } while (resid);
925:
1.54 lukem 926: release:
1.1 cgd 927: sbunlock(&so->so_snd);
1.54 lukem 928: out:
1.1 cgd 929: if (top)
930: m_freem(top);
931: if (control)
932: m_freem(control);
933: return (error);
934: }
935:
936: /*
937: * Implement receive operations on a socket.
938: * We depend on the way that records are added to the sockbuf
939: * by sbappend*. In particular, each record (mbufs linked through m_next)
940: * must begin with an address if the protocol so specifies,
941: * followed by an optional mbuf or mbufs containing ancillary data,
942: * and then zero or more mbufs of data.
943: * In order to avoid blocking network interrupts for the entire time here,
944: * we splx() while doing the actual copy to user space.
945: * Although the sockbuf is locked, new data may still be appended,
946: * and thus we must maintain consistency of the sockbuf during that time.
947: *
948: * The caller may receive the data as a single mbuf chain by supplying
949: * an mbuf **mp0 for use in returning the chain. The uio is then used
950: * only for the count in uio_resid.
951: */
1.3 andrew 952: int
1.54 lukem 953: soreceive(struct socket *so, struct mbuf **paddr, struct uio *uio,
954: struct mbuf **mp0, struct mbuf **controlp, int *flagsp)
1.1 cgd 955: {
1.116 yamt 956: struct lwp *l = curlwp;
1.54 lukem 957: struct mbuf *m, **mp;
958: int flags, len, error, s, offset, moff, type, orig_resid;
1.99 matt 959: const struct protosw *pr;
1.54 lukem 960: struct mbuf *nextrecord;
1.67 he 961: int mbuf_removed = 0;
1.64 thorpej 962:
1.54 lukem 963: pr = so->so_proto;
1.1 cgd 964: mp = mp0;
1.54 lukem 965: type = 0;
966: orig_resid = uio->uio_resid;
1.102 jonathan 967:
1.1 cgd 968: if (paddr)
969: *paddr = 0;
970: if (controlp)
971: *controlp = 0;
972: if (flagsp)
973: flags = *flagsp &~ MSG_EOR;
974: else
975: flags = 0;
1.66 enami 976:
977: if ((flags & MSG_DONTWAIT) == 0)
1.117 yamt 978: sodopendfree();
1.66 enami 979:
1.1 cgd 980: if (flags & MSG_OOB) {
981: m = m_get(M_WAIT, MT_DATA);
1.17 cgd 982: error = (*pr->pr_usrreq)(so, PRU_RCVOOB, m,
1.102 jonathan 983: (struct mbuf *)(long)(flags & MSG_PEEK),
1.114 christos 984: (struct mbuf *)0, l);
1.1 cgd 985: if (error)
986: goto bad;
987: do {
1.134 christos 988: error = uiomove(mtod(m, void *),
1.1 cgd 989: (int) min(uio->uio_resid, m->m_len), uio);
990: m = m_free(m);
991: } while (uio->uio_resid && error == 0 && m);
1.54 lukem 992: bad:
1.1 cgd 993: if (m)
994: m_freem(m);
995: return (error);
996: }
997: if (mp)
998: *mp = (struct mbuf *)0;
999: if (so->so_state & SS_ISCONFIRMING && uio->uio_resid)
1.22 mycroft 1000: (*pr->pr_usrreq)(so, PRU_RCVD, (struct mbuf *)0,
1.114 christos 1001: (struct mbuf *)0, (struct mbuf *)0, l);
1.1 cgd 1002:
1.54 lukem 1003: restart:
1.21 christos 1004: if ((error = sblock(&so->so_rcv, SBLOCKWAIT(flags))) != 0)
1.1 cgd 1005: return (error);
1.20 mycroft 1006: s = splsoftnet();
1.1 cgd 1007:
1008: m = so->so_rcv.sb_mb;
1009: /*
1010: * If we have less data than requested, block awaiting more
1011: * (subject to any timeout) if:
1.15 mycroft 1012: * 1. the current count is less than the low water mark,
1.1 cgd 1013: * 2. MSG_WAITALL is set, and it is possible to do the entire
1.15 mycroft 1014: * receive operation at once if we block (resid <= hiwat), or
1015: * 3. MSG_DONTWAIT is not set.
1.1 cgd 1016: * If MSG_WAITALL is set but resid is larger than the receive buffer,
1017: * we have to do the receive in sections, and thus risk returning
1018: * a short count if a timeout or signal occurs after we start.
1019: */
1.21 christos 1020: if (m == 0 || (((flags & MSG_DONTWAIT) == 0 &&
1.15 mycroft 1021: so->so_rcv.sb_cc < uio->uio_resid) &&
1.1 cgd 1022: (so->so_rcv.sb_cc < so->so_rcv.sb_lowat ||
1023: ((flags & MSG_WAITALL) && uio->uio_resid <= so->so_rcv.sb_hiwat)) &&
1.21 christos 1024: m->m_nextpkt == 0 && (pr->pr_flags & PR_ATOMIC) == 0)) {
1.1 cgd 1025: #ifdef DIAGNOSTIC
1026: if (m == 0 && so->so_rcv.sb_cc)
1027: panic("receive 1");
1028: #endif
1029: if (so->so_error) {
1030: if (m)
1.15 mycroft 1031: goto dontblock;
1.1 cgd 1032: error = so->so_error;
1033: if ((flags & MSG_PEEK) == 0)
1034: so->so_error = 0;
1035: goto release;
1036: }
1037: if (so->so_state & SS_CANTRCVMORE) {
1038: if (m)
1.15 mycroft 1039: goto dontblock;
1.1 cgd 1040: else
1041: goto release;
1042: }
1043: for (; m; m = m->m_next)
1044: if (m->m_type == MT_OOBDATA || (m->m_flags & M_EOR)) {
1045: m = so->so_rcv.sb_mb;
1046: goto dontblock;
1047: }
1048: if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0 &&
1049: (so->so_proto->pr_flags & PR_CONNREQUIRED)) {
1050: error = ENOTCONN;
1051: goto release;
1052: }
1053: if (uio->uio_resid == 0)
1054: goto release;
1.15 mycroft 1055: if ((so->so_state & SS_NBIO) || (flags & MSG_DONTWAIT)) {
1.1 cgd 1056: error = EWOULDBLOCK;
1057: goto release;
1058: }
1.69 thorpej 1059: SBLASTRECORDCHK(&so->so_rcv, "soreceive sbwait 1");
1060: SBLASTMBUFCHK(&so->so_rcv, "soreceive sbwait 1");
1.1 cgd 1061: sbunlock(&so->so_rcv);
1062: error = sbwait(&so->so_rcv);
1063: splx(s);
1064: if (error)
1065: return (error);
1066: goto restart;
1067: }
1.54 lukem 1068: dontblock:
1.69 thorpej 1069: /*
1070: * On entry here, m points to the first record of the socket buffer.
1071: * While we process the initial mbufs containing address and control
1072: * info, we save a copy of m->m_nextpkt into nextrecord.
1073: */
1.114 christos 1074: if (l)
1075: l->l_proc->p_stats->p_ru.ru_msgrcv++;
1.69 thorpej 1076: KASSERT(m == so->so_rcv.sb_mb);
1077: SBLASTRECORDCHK(&so->so_rcv, "soreceive 1");
1078: SBLASTMBUFCHK(&so->so_rcv, "soreceive 1");
1.1 cgd 1079: nextrecord = m->m_nextpkt;
1080: if (pr->pr_flags & PR_ADDR) {
1081: #ifdef DIAGNOSTIC
1082: if (m->m_type != MT_SONAME)
1083: panic("receive 1a");
1084: #endif
1.3 andrew 1085: orig_resid = 0;
1.1 cgd 1086: if (flags & MSG_PEEK) {
1087: if (paddr)
1088: *paddr = m_copy(m, 0, m->m_len);
1089: m = m->m_next;
1090: } else {
1091: sbfree(&so->so_rcv, m);
1.67 he 1092: mbuf_removed = 1;
1.1 cgd 1093: if (paddr) {
1094: *paddr = m;
1095: so->so_rcv.sb_mb = m->m_next;
1096: m->m_next = 0;
1097: m = so->so_rcv.sb_mb;
1098: } else {
1099: MFREE(m, so->so_rcv.sb_mb);
1100: m = so->so_rcv.sb_mb;
1101: }
1102: }
1103: }
1104: while (m && m->m_type == MT_CONTROL && error == 0) {
1105: if (flags & MSG_PEEK) {
1106: if (controlp)
1107: *controlp = m_copy(m, 0, m->m_len);
1108: m = m->m_next;
1109: } else {
1110: sbfree(&so->so_rcv, m);
1.67 he 1111: mbuf_removed = 1;
1.1 cgd 1112: if (controlp) {
1.102 jonathan 1113: struct domain *dom = pr->pr_domain;
1.114 christos 1114: if (dom->dom_externalize && l &&
1.1 cgd 1115: mtod(m, struct cmsghdr *)->cmsg_type ==
1116: SCM_RIGHTS)
1.114 christos 1117: error = (*dom->dom_externalize)(m, l);
1.1 cgd 1118: *controlp = m;
1119: so->so_rcv.sb_mb = m->m_next;
1120: m->m_next = 0;
1121: m = so->so_rcv.sb_mb;
1122: } else {
1.106 itojun 1123: /*
1124: * Dispose of any SCM_RIGHTS message that went
1125: * through the read path rather than recv.
1126: */
1127: if (pr->pr_domain->dom_dispose &&
1128: mtod(m, struct cmsghdr *)->cmsg_type == SCM_RIGHTS)
1129: (*pr->pr_domain->dom_dispose)(m);
1.1 cgd 1130: MFREE(m, so->so_rcv.sb_mb);
1131: m = so->so_rcv.sb_mb;
1132: }
1133: }
1.3 andrew 1134: if (controlp) {
1135: orig_resid = 0;
1.1 cgd 1136: controlp = &(*controlp)->m_next;
1.3 andrew 1137: }
1.1 cgd 1138: }
1.69 thorpej 1139:
1140: /*
1141: * If m is non-NULL, we have some data to read. From now on,
1142: * make sure to keep sb_lastrecord consistent when working on
1143: * the last packet on the chain (nextrecord == NULL) and we
1144: * change m->m_nextpkt.
1145: */
1.1 cgd 1146: if (m) {
1.69 thorpej 1147: if ((flags & MSG_PEEK) == 0) {
1.1 cgd 1148: m->m_nextpkt = nextrecord;
1.69 thorpej 1149: /*
1150: * If nextrecord == NULL (this is a single chain),
1151: * then sb_lastrecord may not be valid here if m
1152: * was changed earlier.
1153: */
1154: if (nextrecord == NULL) {
1155: KASSERT(so->so_rcv.sb_mb == m);
1156: so->so_rcv.sb_lastrecord = m;
1157: }
1158: }
1.1 cgd 1159: type = m->m_type;
1160: if (type == MT_OOBDATA)
1161: flags |= MSG_OOB;
1.69 thorpej 1162: } else {
1163: if ((flags & MSG_PEEK) == 0) {
1164: KASSERT(so->so_rcv.sb_mb == m);
1165: so->so_rcv.sb_mb = nextrecord;
1.70 thorpej 1166: SB_EMPTY_FIXUP(&so->so_rcv);
1.69 thorpej 1167: }
1.1 cgd 1168: }
1.69 thorpej 1169: SBLASTRECORDCHK(&so->so_rcv, "soreceive 2");
1170: SBLASTMBUFCHK(&so->so_rcv, "soreceive 2");
1171:
1.1 cgd 1172: moff = 0;
1173: offset = 0;
1174: while (m && uio->uio_resid > 0 && error == 0) {
1175: if (m->m_type == MT_OOBDATA) {
1176: if (type != MT_OOBDATA)
1177: break;
1178: } else if (type == MT_OOBDATA)
1179: break;
1180: #ifdef DIAGNOSTIC
1181: else if (m->m_type != MT_DATA && m->m_type != MT_HEADER)
1182: panic("receive 3");
1183: #endif
1184: so->so_state &= ~SS_RCVATMARK;
1185: len = uio->uio_resid;
1186: if (so->so_oobmark && len > so->so_oobmark - offset)
1187: len = so->so_oobmark - offset;
1188: if (len > m->m_len - moff)
1189: len = m->m_len - moff;
1190: /*
1191: * If mp is set, just pass back the mbufs.
1192: * Otherwise copy them out via the uio, then free.
1193: * Sockbuf must be consistent here (points to current mbuf,
1194: * it points to next record) when we drop priority;
1195: * we must note any additions to the sockbuf when we
1196: * block interrupts again.
1197: */
1198: if (mp == 0) {
1.69 thorpej 1199: SBLASTRECORDCHK(&so->so_rcv, "soreceive uiomove");
1200: SBLASTMBUFCHK(&so->so_rcv, "soreceive uiomove");
1.1 cgd 1201: splx(s);
1.134 christos 1202: error = uiomove(mtod(m, char *) + moff, (int)len, uio);
1.20 mycroft 1203: s = splsoftnet();
1.67 he 1204: if (error) {
1205: /*
1206: * If any part of the record has been removed
1207: * (such as the MT_SONAME mbuf, which will
1208: * happen when PR_ADDR, and thus also
1209: * PR_ATOMIC, is set), then drop the entire
1210: * record to maintain the atomicity of the
1211: * receive operation.
1212: *
1213: * This avoids a later panic("receive 1a")
1214: * when compiled with DIAGNOSTIC.
1215: */
1216: if (m && mbuf_removed
1217: && (pr->pr_flags & PR_ATOMIC))
1218: (void) sbdroprecord(&so->so_rcv);
1219:
1.57 jdolecek 1220: goto release;
1.67 he 1221: }
1.1 cgd 1222: } else
1223: uio->uio_resid -= len;
1224: if (len == m->m_len - moff) {
1225: if (m->m_flags & M_EOR)
1226: flags |= MSG_EOR;
1227: if (flags & MSG_PEEK) {
1228: m = m->m_next;
1229: moff = 0;
1230: } else {
1231: nextrecord = m->m_nextpkt;
1232: sbfree(&so->so_rcv, m);
1233: if (mp) {
1234: *mp = m;
1235: mp = &m->m_next;
1236: so->so_rcv.sb_mb = m = m->m_next;
1237: *mp = (struct mbuf *)0;
1238: } else {
1239: MFREE(m, so->so_rcv.sb_mb);
1240: m = so->so_rcv.sb_mb;
1241: }
1.69 thorpej 1242: /*
1243: * If m != NULL, we also know that
1244: * so->so_rcv.sb_mb != NULL.
1245: */
1246: KASSERT(so->so_rcv.sb_mb == m);
1247: if (m) {
1.1 cgd 1248: m->m_nextpkt = nextrecord;
1.69 thorpej 1249: if (nextrecord == NULL)
1250: so->so_rcv.sb_lastrecord = m;
1251: } else {
1252: so->so_rcv.sb_mb = nextrecord;
1.70 thorpej 1253: SB_EMPTY_FIXUP(&so->so_rcv);
1.69 thorpej 1254: }
1255: SBLASTRECORDCHK(&so->so_rcv, "soreceive 3");
1256: SBLASTMBUFCHK(&so->so_rcv, "soreceive 3");
1.1 cgd 1257: }
1258: } else {
1259: if (flags & MSG_PEEK)
1260: moff += len;
1261: else {
1262: if (mp)
1263: *mp = m_copym(m, 0, len, M_WAIT);
1264: m->m_data += len;
1265: m->m_len -= len;
1266: so->so_rcv.sb_cc -= len;
1267: }
1268: }
1269: if (so->so_oobmark) {
1270: if ((flags & MSG_PEEK) == 0) {
1271: so->so_oobmark -= len;
1272: if (so->so_oobmark == 0) {
1273: so->so_state |= SS_RCVATMARK;
1274: break;
1275: }
1.7 cgd 1276: } else {
1.1 cgd 1277: offset += len;
1.7 cgd 1278: if (offset == so->so_oobmark)
1279: break;
1280: }
1.1 cgd 1281: }
1282: if (flags & MSG_EOR)
1283: break;
1284: /*
1285: * If the MSG_WAITALL flag is set (for non-atomic socket),
1286: * we must not quit until "uio->uio_resid == 0" or an error
1287: * termination. If a signal/timeout occurs, return
1288: * with a short count but without error.
1289: * Keep sockbuf locked against other readers.
1290: */
1291: while (flags & MSG_WAITALL && m == 0 && uio->uio_resid > 0 &&
1.3 andrew 1292: !sosendallatonce(so) && !nextrecord) {
1.1 cgd 1293: if (so->so_error || so->so_state & SS_CANTRCVMORE)
1294: break;
1.68 matt 1295: /*
1296: * If we are peeking and the socket receive buffer is
1297: * full, stop since we can't get more data to peek at.
1298: */
1299: if ((flags & MSG_PEEK) && sbspace(&so->so_rcv) <= 0)
1300: break;
1301: /*
1302: * If we've drained the socket buffer, tell the
1303: * protocol in case it needs to do something to
1304: * get it filled again.
1305: */
1306: if ((pr->pr_flags & PR_WANTRCVD) && so->so_pcb)
1307: (*pr->pr_usrreq)(so, PRU_RCVD,
1308: (struct mbuf *)0,
1309: (struct mbuf *)(long)flags,
1.114 christos 1310: (struct mbuf *)0, l);
1.69 thorpej 1311: SBLASTRECORDCHK(&so->so_rcv, "soreceive sbwait 2");
1312: SBLASTMBUFCHK(&so->so_rcv, "soreceive sbwait 2");
1.1 cgd 1313: error = sbwait(&so->so_rcv);
1314: if (error) {
1315: sbunlock(&so->so_rcv);
1316: splx(s);
1317: return (0);
1318: }
1.21 christos 1319: if ((m = so->so_rcv.sb_mb) != NULL)
1.1 cgd 1320: nextrecord = m->m_nextpkt;
1321: }
1322: }
1.3 andrew 1323:
1324: if (m && pr->pr_flags & PR_ATOMIC) {
1325: flags |= MSG_TRUNC;
1326: if ((flags & MSG_PEEK) == 0)
1327: (void) sbdroprecord(&so->so_rcv);
1328: }
1.1 cgd 1329: if ((flags & MSG_PEEK) == 0) {
1.69 thorpej 1330: if (m == 0) {
1331: /*
1.70 thorpej 1332: * First part is an inline SB_EMPTY_FIXUP(). Second
1.69 thorpej 1333: * part makes sure sb_lastrecord is up-to-date if
1334: * there is still data in the socket buffer.
1335: */
1.1 cgd 1336: so->so_rcv.sb_mb = nextrecord;
1.69 thorpej 1337: if (so->so_rcv.sb_mb == NULL) {
1338: so->so_rcv.sb_mbtail = NULL;
1339: so->so_rcv.sb_lastrecord = NULL;
1340: } else if (nextrecord->m_nextpkt == NULL)
1341: so->so_rcv.sb_lastrecord = nextrecord;
1342: }
1343: SBLASTRECORDCHK(&so->so_rcv, "soreceive 4");
1344: SBLASTMBUFCHK(&so->so_rcv, "soreceive 4");
1.1 cgd 1345: if (pr->pr_flags & PR_WANTRCVD && so->so_pcb)
1.22 mycroft 1346: (*pr->pr_usrreq)(so, PRU_RCVD, (struct mbuf *)0,
1.114 christos 1347: (struct mbuf *)(long)flags, (struct mbuf *)0, l);
1.1 cgd 1348: }
1.3 andrew 1349: if (orig_resid == uio->uio_resid && orig_resid &&
1350: (flags & MSG_EOR) == 0 && (so->so_state & SS_CANTRCVMORE) == 0) {
1351: sbunlock(&so->so_rcv);
1352: splx(s);
1353: goto restart;
1354: }
1.108 perry 1355:
1.1 cgd 1356: if (flagsp)
1357: *flagsp |= flags;
1.54 lukem 1358: release:
1.1 cgd 1359: sbunlock(&so->so_rcv);
1360: splx(s);
1361: return (error);
1362: }
1363:
1.14 mycroft 1364: int
1.54 lukem 1365: soshutdown(struct socket *so, int how)
1.1 cgd 1366: {
1.99 matt 1367: const struct protosw *pr;
1.34 kleink 1368:
1.54 lukem 1369: pr = so->so_proto;
1.34 kleink 1370: if (!(how == SHUT_RD || how == SHUT_WR || how == SHUT_RDWR))
1371: return (EINVAL);
1.1 cgd 1372:
1.34 kleink 1373: if (how == SHUT_RD || how == SHUT_RDWR)
1.1 cgd 1374: sorflush(so);
1.34 kleink 1375: if (how == SHUT_WR || how == SHUT_RDWR)
1.22 mycroft 1376: return (*pr->pr_usrreq)(so, PRU_SHUTDOWN, (struct mbuf *)0,
1.114 christos 1377: (struct mbuf *)0, (struct mbuf *)0, (struct lwp *)0);
1.1 cgd 1378: return (0);
1379: }
1380:
1.14 mycroft 1381: void
1.54 lukem 1382: sorflush(struct socket *so)
1.1 cgd 1383: {
1.54 lukem 1384: struct sockbuf *sb, asb;
1.99 matt 1385: const struct protosw *pr;
1.54 lukem 1386: int s;
1.1 cgd 1387:
1.54 lukem 1388: sb = &so->so_rcv;
1389: pr = so->so_proto;
1.1 cgd 1390: sb->sb_flags |= SB_NOINTR;
1.15 mycroft 1391: (void) sblock(sb, M_WAITOK);
1.56 thorpej 1392: s = splnet();
1.1 cgd 1393: socantrcvmore(so);
1394: sbunlock(sb);
1395: asb = *sb;
1.86 wrstuden 1396: /*
1397: * Clear most of the sockbuf structure, but leave some of the
1398: * fields valid.
1399: */
1400: memset(&sb->sb_startzero, 0,
1401: sizeof(*sb) - offsetof(struct sockbuf, sb_startzero));
1.1 cgd 1402: splx(s);
1403: if (pr->pr_flags & PR_RIGHTS && pr->pr_domain->dom_dispose)
1404: (*pr->pr_domain->dom_dispose)(asb.sb_mb);
1.98 christos 1405: sbrelease(&asb, so);
1.1 cgd 1406: }
1407:
1.14 mycroft 1408: int
1.54 lukem 1409: sosetopt(struct socket *so, int level, int optname, struct mbuf *m0)
1.1 cgd 1410: {
1.54 lukem 1411: int error;
1412: struct mbuf *m;
1.130 christos 1413: struct linger *l;
1.1 cgd 1414:
1.54 lukem 1415: error = 0;
1416: m = m0;
1.1 cgd 1417: if (level != SOL_SOCKET) {
1418: if (so->so_proto && so->so_proto->pr_ctloutput)
1419: return ((*so->so_proto->pr_ctloutput)
1420: (PRCO_SETOPT, so, level, optname, &m0));
1421: error = ENOPROTOOPT;
1422: } else {
1423: switch (optname) {
1424:
1425: case SO_LINGER:
1.36 perry 1426: if (m == NULL || m->m_len != sizeof(struct linger)) {
1.1 cgd 1427: error = EINVAL;
1428: goto bad;
1429: }
1.130 christos 1430: l = mtod(m, struct linger *);
1.131 christos 1431: if (l->l_linger < 0 || l->l_linger > USHRT_MAX ||
1432: l->l_linger > (INT_MAX / hz)) {
1.112 nathanw 1433: error = EDOM;
1434: goto bad;
1435: }
1.130 christos 1436: so->so_linger = l->l_linger;
1437: if (l->l_onoff)
1438: so->so_options |= SO_LINGER;
1439: else
1440: so->so_options &= ~SO_LINGER;
1441: break;
1.1 cgd 1442:
1443: case SO_DEBUG:
1444: case SO_KEEPALIVE:
1445: case SO_DONTROUTE:
1446: case SO_USELOOPBACK:
1447: case SO_BROADCAST:
1448: case SO_REUSEADDR:
1.15 mycroft 1449: case SO_REUSEPORT:
1.1 cgd 1450: case SO_OOBINLINE:
1.26 thorpej 1451: case SO_TIMESTAMP:
1.36 perry 1452: if (m == NULL || m->m_len < sizeof(int)) {
1.1 cgd 1453: error = EINVAL;
1454: goto bad;
1455: }
1456: if (*mtod(m, int *))
1457: so->so_options |= optname;
1458: else
1459: so->so_options &= ~optname;
1460: break;
1461:
1462: case SO_SNDBUF:
1463: case SO_RCVBUF:
1464: case SO_SNDLOWAT:
1465: case SO_RCVLOWAT:
1.28 thorpej 1466: {
1467: int optval;
1468:
1.36 perry 1469: if (m == NULL || m->m_len < sizeof(int)) {
1.1 cgd 1470: error = EINVAL;
1471: goto bad;
1472: }
1.28 thorpej 1473:
1474: /*
1475: * Values < 1 make no sense for any of these
1476: * options, so disallow them.
1477: */
1478: optval = *mtod(m, int *);
1479: if (optval < 1) {
1480: error = EINVAL;
1481: goto bad;
1482: }
1483:
1.1 cgd 1484: switch (optname) {
1485:
1486: case SO_SNDBUF:
1487: case SO_RCVBUF:
1488: if (sbreserve(optname == SO_SNDBUF ?
1489: &so->so_snd : &so->so_rcv,
1.98 christos 1490: (u_long) optval, so) == 0) {
1.1 cgd 1491: error = ENOBUFS;
1492: goto bad;
1493: }
1494: break;
1495:
1.28 thorpej 1496: /*
1497: * Make sure the low-water is never greater than
1498: * the high-water.
1499: */
1.1 cgd 1500: case SO_SNDLOWAT:
1.28 thorpej 1501: so->so_snd.sb_lowat =
1502: (optval > so->so_snd.sb_hiwat) ?
1503: so->so_snd.sb_hiwat : optval;
1.1 cgd 1504: break;
1505: case SO_RCVLOWAT:
1.28 thorpej 1506: so->so_rcv.sb_lowat =
1507: (optval > so->so_rcv.sb_hiwat) ?
1508: so->so_rcv.sb_hiwat : optval;
1.1 cgd 1509: break;
1510: }
1511: break;
1.28 thorpej 1512: }
1.1 cgd 1513:
1514: case SO_SNDTIMEO:
1515: case SO_RCVTIMEO:
1516: {
1517: struct timeval *tv;
1.104 yamt 1518: int val;
1.1 cgd 1519:
1.36 perry 1520: if (m == NULL || m->m_len < sizeof(*tv)) {
1.1 cgd 1521: error = EINVAL;
1522: goto bad;
1523: }
1524: tv = mtod(m, struct timeval *);
1.104 yamt 1525: if (tv->tv_sec > (INT_MAX - tv->tv_usec / tick) / hz) {
1.1 cgd 1526: error = EDOM;
1527: goto bad;
1528: }
1529: val = tv->tv_sec * hz + tv->tv_usec / tick;
1.74 itojun 1530: if (val == 0 && tv->tv_usec != 0)
1531: val = 1;
1.1 cgd 1532:
1533: switch (optname) {
1534:
1535: case SO_SNDTIMEO:
1536: so->so_snd.sb_timeo = val;
1537: break;
1538: case SO_RCVTIMEO:
1539: so->so_rcv.sb_timeo = val;
1540: break;
1541: }
1542: break;
1543: }
1544:
1545: default:
1546: error = ENOPROTOOPT;
1547: break;
1548: }
1.15 mycroft 1549: if (error == 0 && so->so_proto && so->so_proto->pr_ctloutput) {
1550: (void) ((*so->so_proto->pr_ctloutput)
1551: (PRCO_SETOPT, so, level, optname, &m0));
1552: m = NULL; /* freed by protocol */
1553: }
1.1 cgd 1554: }
1.54 lukem 1555: bad:
1.1 cgd 1556: if (m)
1557: (void) m_free(m);
1558: return (error);
1559: }
1560:
1.14 mycroft 1561: int
1.54 lukem 1562: sogetopt(struct socket *so, int level, int optname, struct mbuf **mp)
1.1 cgd 1563: {
1.54 lukem 1564: struct mbuf *m;
1.1 cgd 1565:
1566: if (level != SOL_SOCKET) {
1567: if (so->so_proto && so->so_proto->pr_ctloutput) {
1568: return ((*so->so_proto->pr_ctloutput)
1569: (PRCO_GETOPT, so, level, optname, mp));
1570: } else
1571: return (ENOPROTOOPT);
1572: } else {
1573: m = m_get(M_WAIT, MT_SOOPTS);
1.36 perry 1574: m->m_len = sizeof(int);
1.1 cgd 1575:
1576: switch (optname) {
1577:
1578: case SO_LINGER:
1.36 perry 1579: m->m_len = sizeof(struct linger);
1.1 cgd 1580: mtod(m, struct linger *)->l_onoff =
1.131 christos 1581: (so->so_options & SO_LINGER) ? 1 : 0;
1.1 cgd 1582: mtod(m, struct linger *)->l_linger = so->so_linger;
1583: break;
1584:
1585: case SO_USELOOPBACK:
1586: case SO_DONTROUTE:
1587: case SO_DEBUG:
1588: case SO_KEEPALIVE:
1589: case SO_REUSEADDR:
1.15 mycroft 1590: case SO_REUSEPORT:
1.1 cgd 1591: case SO_BROADCAST:
1592: case SO_OOBINLINE:
1.26 thorpej 1593: case SO_TIMESTAMP:
1.131 christos 1594: *mtod(m, int *) = (so->so_options & optname) ? 1 : 0;
1.1 cgd 1595: break;
1596:
1597: case SO_TYPE:
1598: *mtod(m, int *) = so->so_type;
1599: break;
1600:
1601: case SO_ERROR:
1602: *mtod(m, int *) = so->so_error;
1603: so->so_error = 0;
1604: break;
1605:
1606: case SO_SNDBUF:
1607: *mtod(m, int *) = so->so_snd.sb_hiwat;
1608: break;
1609:
1610: case SO_RCVBUF:
1611: *mtod(m, int *) = so->so_rcv.sb_hiwat;
1612: break;
1613:
1614: case SO_SNDLOWAT:
1615: *mtod(m, int *) = so->so_snd.sb_lowat;
1616: break;
1617:
1618: case SO_RCVLOWAT:
1619: *mtod(m, int *) = so->so_rcv.sb_lowat;
1620: break;
1621:
1622: case SO_SNDTIMEO:
1623: case SO_RCVTIMEO:
1624: {
1625: int val = (optname == SO_SNDTIMEO ?
1626: so->so_snd.sb_timeo : so->so_rcv.sb_timeo);
1627:
1628: m->m_len = sizeof(struct timeval);
1629: mtod(m, struct timeval *)->tv_sec = val / hz;
1630: mtod(m, struct timeval *)->tv_usec =
1.27 kleink 1631: (val % hz) * tick;
1.1 cgd 1632: break;
1633: }
1634:
1.107 darrenr 1635: case SO_OVERFLOWED:
1636: *mtod(m, int *) = so->so_rcv.sb_overflowed;
1637: break;
1638:
1.1 cgd 1639: default:
1640: (void)m_free(m);
1641: return (ENOPROTOOPT);
1642: }
1643: *mp = m;
1644: return (0);
1645: }
1646: }
1647:
1.14 mycroft 1648: void
1.54 lukem 1649: sohasoutofband(struct socket *so)
1.1 cgd 1650: {
1.90 christos 1651: fownsignal(so->so_pgid, SIGURG, POLL_PRI, POLLPRI|POLLRDBAND, so);
1.2 cgd 1652: selwakeup(&so->so_rcv.sb_sel);
1.1 cgd 1653: }
1.72 jdolecek 1654:
1655: static void
1656: filt_sordetach(struct knote *kn)
1657: {
1658: struct socket *so;
1659:
1660: so = (struct socket *)kn->kn_fp->f_data;
1.73 christos 1661: SLIST_REMOVE(&so->so_rcv.sb_sel.sel_klist, kn, knote, kn_selnext);
1662: if (SLIST_EMPTY(&so->so_rcv.sb_sel.sel_klist))
1.72 jdolecek 1663: so->so_rcv.sb_flags &= ~SB_KNOTE;
1664: }
1665:
1666: /*ARGSUSED*/
1667: static int
1.129 yamt 1668: filt_soread(struct knote *kn, long hint)
1.72 jdolecek 1669: {
1670: struct socket *so;
1671:
1672: so = (struct socket *)kn->kn_fp->f_data;
1673: kn->kn_data = so->so_rcv.sb_cc;
1674: if (so->so_state & SS_CANTRCVMORE) {
1.108 perry 1675: kn->kn_flags |= EV_EOF;
1.72 jdolecek 1676: kn->kn_fflags = so->so_error;
1677: return (1);
1678: }
1679: if (so->so_error) /* temporary udp error */
1680: return (1);
1681: if (kn->kn_sfflags & NOTE_LOWAT)
1682: return (kn->kn_data >= kn->kn_sdata);
1683: return (kn->kn_data >= so->so_rcv.sb_lowat);
1684: }
1685:
1686: static void
1687: filt_sowdetach(struct knote *kn)
1688: {
1689: struct socket *so;
1690:
1691: so = (struct socket *)kn->kn_fp->f_data;
1.73 christos 1692: SLIST_REMOVE(&so->so_snd.sb_sel.sel_klist, kn, knote, kn_selnext);
1693: if (SLIST_EMPTY(&so->so_snd.sb_sel.sel_klist))
1.72 jdolecek 1694: so->so_snd.sb_flags &= ~SB_KNOTE;
1695: }
1696:
1697: /*ARGSUSED*/
1698: static int
1.129 yamt 1699: filt_sowrite(struct knote *kn, long hint)
1.72 jdolecek 1700: {
1701: struct socket *so;
1702:
1703: so = (struct socket *)kn->kn_fp->f_data;
1704: kn->kn_data = sbspace(&so->so_snd);
1705: if (so->so_state & SS_CANTSENDMORE) {
1.108 perry 1706: kn->kn_flags |= EV_EOF;
1.72 jdolecek 1707: kn->kn_fflags = so->so_error;
1708: return (1);
1709: }
1710: if (so->so_error) /* temporary udp error */
1711: return (1);
1712: if (((so->so_state & SS_ISCONNECTED) == 0) &&
1713: (so->so_proto->pr_flags & PR_CONNREQUIRED))
1714: return (0);
1715: if (kn->kn_sfflags & NOTE_LOWAT)
1716: return (kn->kn_data >= kn->kn_sdata);
1717: return (kn->kn_data >= so->so_snd.sb_lowat);
1718: }
1719:
1720: /*ARGSUSED*/
1721: static int
1.129 yamt 1722: filt_solisten(struct knote *kn, long hint)
1.72 jdolecek 1723: {
1724: struct socket *so;
1725:
1726: so = (struct socket *)kn->kn_fp->f_data;
1727:
1728: /*
1729: * Set kn_data to number of incoming connections, not
1730: * counting partial (incomplete) connections.
1.108 perry 1731: */
1.72 jdolecek 1732: kn->kn_data = so->so_qlen;
1733: return (kn->kn_data > 0);
1734: }
1735:
1736: static const struct filterops solisten_filtops =
1737: { 1, NULL, filt_sordetach, filt_solisten };
1738: static const struct filterops soread_filtops =
1739: { 1, NULL, filt_sordetach, filt_soread };
1740: static const struct filterops sowrite_filtops =
1741: { 1, NULL, filt_sowdetach, filt_sowrite };
1742:
1743: int
1.129 yamt 1744: soo_kqfilter(struct file *fp, struct knote *kn)
1.72 jdolecek 1745: {
1746: struct socket *so;
1747: struct sockbuf *sb;
1748:
1749: so = (struct socket *)kn->kn_fp->f_data;
1750: switch (kn->kn_filter) {
1751: case EVFILT_READ:
1752: if (so->so_options & SO_ACCEPTCONN)
1753: kn->kn_fop = &solisten_filtops;
1754: else
1755: kn->kn_fop = &soread_filtops;
1756: sb = &so->so_rcv;
1757: break;
1758: case EVFILT_WRITE:
1759: kn->kn_fop = &sowrite_filtops;
1760: sb = &so->so_snd;
1761: break;
1762: default:
1763: return (1);
1764: }
1.73 christos 1765: SLIST_INSERT_HEAD(&sb->sb_sel.sel_klist, kn, kn_selnext);
1.72 jdolecek 1766: sb->sb_flags |= SB_KNOTE;
1767: return (0);
1768: }
1769:
1.94 yamt 1770: #include <sys/sysctl.h>
1771:
1772: static int sysctl_kern_somaxkva(SYSCTLFN_PROTO);
1773:
1774: /*
1775: * sysctl helper routine for kern.somaxkva. ensures that the given
1776: * value is not too small.
1777: * (XXX should we maybe make sure it's not too large as well?)
1778: */
1779: static int
1780: sysctl_kern_somaxkva(SYSCTLFN_ARGS)
1781: {
1782: int error, new_somaxkva;
1783: struct sysctlnode node;
1784:
1785: new_somaxkva = somaxkva;
1786: node = *rnode;
1787: node.sysctl_data = &new_somaxkva;
1788: error = sysctl_lookup(SYSCTLFN_CALL(&node));
1789: if (error || newp == NULL)
1790: return (error);
1791:
1792: if (new_somaxkva < (16 * 1024 * 1024)) /* sanity */
1793: return (EINVAL);
1794:
1.136 ad 1795: mutex_enter(&so_pendfree_lock);
1.94 yamt 1796: somaxkva = new_somaxkva;
1.136 ad 1797: cv_broadcast(&socurkva_cv);
1798: mutex_exit(&so_pendfree_lock);
1.94 yamt 1799:
1800: return (error);
1801: }
1802:
1803: SYSCTL_SETUP(sysctl_kern_somaxkva_setup, "sysctl kern.somaxkva setup")
1804: {
1805:
1.97 atatat 1806: sysctl_createv(clog, 0, NULL, NULL,
1807: CTLFLAG_PERMANENT,
1808: CTLTYPE_NODE, "kern", NULL,
1809: NULL, 0, NULL, 0,
1810: CTL_KERN, CTL_EOL);
1811:
1812: sysctl_createv(clog, 0, NULL, NULL,
1813: CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1.103 atatat 1814: CTLTYPE_INT, "somaxkva",
1815: SYSCTL_DESCR("Maximum amount of kernel memory to be "
1816: "used for socket buffers"),
1.94 yamt 1817: sysctl_kern_somaxkva, 0, NULL, 0,
1818: CTL_KERN, KERN_SOMAXKVA, CTL_EOL);
1819: }
CVSweb <webmaster@jp.NetBSD.org>
![[BACK]](/icons/back.gif){kind=icon}
Return to uipc_socket.c CVS log ![[TXT]](/icons/text.gif){kind=icon}
![[DIR]](/icons/dir.gif){kind=icon}
Up to [cvs.NetBSD.org] / src / sys / kern