Annotation of src/sys/kern/uipc_socket.c, Revision 1.155
1.155 ! ad 1: /* $NetBSD: uipc_socket.c,v 1.154 2008/03/20 19:23:15 ad Exp $ */
1.64 thorpej 2:
3: /*-
1.154 ad 4: * Copyright (c) 2002, 2007, 2008 The NetBSD Foundation, Inc.
1.64 thorpej 5: * All rights reserved.
6: *
7: * This code is derived from software contributed to The NetBSD Foundation
8: * by Jason R. Thorpe of Wasabi Systems, Inc.
9: *
10: * Redistribution and use in source and binary forms, with or without
11: * modification, are permitted provided that the following conditions
12: * are met:
13: * 1. Redistributions of source code must retain the above copyright
14: * notice, this list of conditions and the following disclaimer.
15: * 2. Redistributions in binary form must reproduce the above copyright
16: * notice, this list of conditions and the following disclaimer in the
17: * documentation and/or other materials provided with the distribution.
18: * 3. All advertising materials mentioning features or use of this software
19: * must display the following acknowledgement:
20: * This product includes software developed by the NetBSD
21: * Foundation, Inc. and its contributors.
22: * 4. Neither the name of The NetBSD Foundation nor the names of its
23: * contributors may be used to endorse or promote products derived
24: * from this software without specific prior written permission.
25: *
26: * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
27: * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
28: * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
29: * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
30: * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
31: * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
32: * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
33: * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
34: * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
35: * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
36: * POSSIBILITY OF SUCH DAMAGE.
37: */
1.16 cgd 38:
1.1 cgd 39: /*
1.15 mycroft 40: * Copyright (c) 1982, 1986, 1988, 1990, 1993
41: * The Regents of the University of California. All rights reserved.
1.1 cgd 42: *
43: * Redistribution and use in source and binary forms, with or without
44: * modification, are permitted provided that the following conditions
45: * are met:
46: * 1. Redistributions of source code must retain the above copyright
47: * notice, this list of conditions and the following disclaimer.
48: * 2. Redistributions in binary form must reproduce the above copyright
49: * notice, this list of conditions and the following disclaimer in the
50: * documentation and/or other materials provided with the distribution.
1.85 agc 51: * 3. Neither the name of the University nor the names of its contributors
1.1 cgd 52: * may be used to endorse or promote products derived from this software
53: * without specific prior written permission.
54: *
55: * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
56: * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
57: * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
58: * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
59: * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
60: * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
61: * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
62: * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
63: * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
64: * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
65: * SUCH DAMAGE.
66: *
1.32 fvdl 67: * @(#)uipc_socket.c 8.6 (Berkeley) 5/2/95
1.1 cgd 68: */
1.59 lukem 69:
70: #include <sys/cdefs.h>
1.155 ! ad 71: __KERNEL_RCSID(0, "$NetBSD: uipc_socket.c,v 1.154 2008/03/20 19:23:15 ad Exp $");
1.64 thorpej 72:
73: #include "opt_sock_counters.h"
74: #include "opt_sosend_loan.h"
1.81 martin 75: #include "opt_mbuftrace.h"
1.84 ragge 76: #include "opt_somaxkva.h"
1.1 cgd 77:
1.9 mycroft 78: #include <sys/param.h>
79: #include <sys/systm.h>
80: #include <sys/proc.h>
81: #include <sys/file.h>
1.142 dyoung 82: #include <sys/filedesc.h>
1.9 mycroft 83: #include <sys/malloc.h>
84: #include <sys/mbuf.h>
85: #include <sys/domain.h>
86: #include <sys/kernel.h>
87: #include <sys/protosw.h>
88: #include <sys/socket.h>
89: #include <sys/socketvar.h>
1.21 christos 90: #include <sys/signalvar.h>
1.9 mycroft 91: #include <sys/resourcevar.h>
1.37 thorpej 92: #include <sys/pool.h>
1.72 jdolecek 93: #include <sys/event.h>
1.89 christos 94: #include <sys/poll.h>
1.118 elad 95: #include <sys/kauth.h>
1.136 ad 96: #include <sys/mutex.h>
97: #include <sys/condvar.h>
1.37 thorpej 98:
1.64 thorpej 99: #include <uvm/uvm.h>
100:
1.135 ad 101: POOL_INIT(socket_pool, sizeof(struct socket), 0, 0, 0, "sockpl", NULL,
102: IPL_SOFTNET);
1.77 thorpej 103:
104: MALLOC_DEFINE(M_SOOPTS, "soopts", "socket options");
105: MALLOC_DEFINE(M_SONAME, "soname", "socket name");
1.37 thorpej 106:
1.142 dyoung 107: extern const struct fileops socketops;
108:
1.54 lukem 109: extern int somaxconn; /* patchable (XXX sysctl) */
110: int somaxconn = SOMAXCONN;
1.49 jonathan 111:
1.64 thorpej 112: #ifdef SOSEND_COUNTERS
113: #include <sys/device.h>
114:
1.113 thorpej 115: static struct evcnt sosend_loan_big = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
1.64 thorpej 116: NULL, "sosend", "loan big");
1.113 thorpej 117: static struct evcnt sosend_copy_big = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
1.64 thorpej 118: NULL, "sosend", "copy big");
1.113 thorpej 119: static struct evcnt sosend_copy_small = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
1.64 thorpej 120: NULL, "sosend", "copy small");
1.113 thorpej 121: static struct evcnt sosend_kvalimit = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
1.64 thorpej 122: NULL, "sosend", "kva limit");
123:
124: #define SOSEND_COUNTER_INCR(ev) (ev)->ev_count++
125:
1.101 matt 126: EVCNT_ATTACH_STATIC(sosend_loan_big);
127: EVCNT_ATTACH_STATIC(sosend_copy_big);
128: EVCNT_ATTACH_STATIC(sosend_copy_small);
129: EVCNT_ATTACH_STATIC(sosend_kvalimit);
1.64 thorpej 130: #else
131:
132: #define SOSEND_COUNTER_INCR(ev) /* nothing */
133:
134: #endif /* SOSEND_COUNTERS */
135:
1.119 yamt 136: static struct callback_entry sokva_reclaimerentry;
1.1 cgd 137:
1.71 thorpej 138: #ifdef SOSEND_NO_LOAN
1.121 yamt 139: int sock_loan_thresh = -1;
1.71 thorpej 140: #else
1.121 yamt 141: int sock_loan_thresh = 4096;
1.65 thorpej 142: #endif
1.64 thorpej 143:
1.136 ad 144: static kmutex_t so_pendfree_lock;
1.113 thorpej 145: static struct mbuf *so_pendfree;
1.64 thorpej 146:
1.84 ragge 147: #ifndef SOMAXKVA
148: #define SOMAXKVA (16 * 1024 * 1024)
149: #endif
150: int somaxkva = SOMAXKVA;
1.113 thorpej 151: static int socurkva;
1.136 ad 152: static kcondvar_t socurkva_cv;
1.64 thorpej 153:
154: #define SOCK_LOAN_CHUNK 65536
155:
1.117 yamt 156: static size_t sodopendfree(void);
157: static size_t sodopendfreel(void);
1.93 yamt 158:
1.113 thorpej 159: static vsize_t
1.129 yamt 160: sokvareserve(struct socket *so, vsize_t len)
1.80 yamt 161: {
1.98 christos 162: int error;
1.80 yamt 163:
1.136 ad 164: mutex_enter(&so_pendfree_lock);
1.80 yamt 165: while (socurkva + len > somaxkva) {
1.93 yamt 166: size_t freed;
167:
168: /*
169: * try to do pendfree.
170: */
171:
1.117 yamt 172: freed = sodopendfreel();
1.93 yamt 173:
174: /*
175: * if some kva was freed, try again.
176: */
177:
178: if (freed)
1.80 yamt 179: continue;
1.93 yamt 180:
1.80 yamt 181: SOSEND_COUNTER_INCR(&sosend_kvalimit);
1.136 ad 182: error = cv_wait_sig(&socurkva_cv, &so_pendfree_lock);
1.98 christos 183: if (error) {
184: len = 0;
185: break;
186: }
1.80 yamt 187: }
1.93 yamt 188: socurkva += len;
1.136 ad 189: mutex_exit(&so_pendfree_lock);
1.98 christos 190: return len;
1.95 yamt 191: }
192:
1.113 thorpej 193: static void
1.95 yamt 194: sokvaunreserve(vsize_t len)
195: {
196:
1.136 ad 197: mutex_enter(&so_pendfree_lock);
1.95 yamt 198: socurkva -= len;
1.136 ad 199: cv_broadcast(&socurkva_cv);
200: mutex_exit(&so_pendfree_lock);
1.95 yamt 201: }
202:
203: /*
204: * sokvaalloc: allocate kva for loan.
205: */
206:
207: vaddr_t
208: sokvaalloc(vsize_t len, struct socket *so)
209: {
210: vaddr_t lva;
211:
212: /*
213: * reserve kva.
214: */
215:
1.98 christos 216: if (sokvareserve(so, len) == 0)
217: return 0;
1.93 yamt 218:
219: /*
220: * allocate kva.
221: */
1.80 yamt 222:
1.109 yamt 223: lva = uvm_km_alloc(kernel_map, len, 0, UVM_KMF_VAONLY | UVM_KMF_WAITVA);
1.95 yamt 224: if (lva == 0) {
225: sokvaunreserve(len);
1.80 yamt 226: return (0);
1.95 yamt 227: }
1.80 yamt 228:
229: return lva;
230: }
231:
1.93 yamt 232: /*
233: * sokvafree: free kva for loan.
234: */
235:
1.80 yamt 236: void
237: sokvafree(vaddr_t sva, vsize_t len)
238: {
1.93 yamt 239:
240: /*
241: * free kva.
242: */
1.80 yamt 243:
1.109 yamt 244: uvm_km_free(kernel_map, sva, len, UVM_KMF_VAONLY);
1.93 yamt 245:
246: /*
247: * unreserve kva.
248: */
249:
1.95 yamt 250: sokvaunreserve(len);
1.80 yamt 251: }
252:
1.64 thorpej 253: static void
1.134 christos 254: sodoloanfree(struct vm_page **pgs, void *buf, size_t size)
1.64 thorpej 255: {
256: vaddr_t va, sva, eva;
257: vsize_t len;
258: paddr_t pa;
259: int i, npgs;
260:
261: eva = round_page((vaddr_t) buf + size);
262: sva = trunc_page((vaddr_t) buf);
263: len = eva - sva;
264: npgs = len >> PAGE_SHIFT;
265:
1.79 thorpej 266: if (__predict_false(pgs == NULL)) {
267: pgs = alloca(npgs * sizeof(*pgs));
1.64 thorpej 268:
1.79 thorpej 269: for (i = 0, va = sva; va < eva; i++, va += PAGE_SIZE) {
1.133 thorpej 270: if (pmap_extract(pmap_kernel(), va, &pa) == false)
1.79 thorpej 271: panic("sodoloanfree: va 0x%lx not mapped", va);
272: pgs[i] = PHYS_TO_VM_PAGE(pa);
273: }
1.64 thorpej 274: }
275:
276: pmap_kremove(sva, len);
277: pmap_update(pmap_kernel());
278: uvm_unloan(pgs, npgs, UVM_LOAN_TOPAGE);
1.80 yamt 279: sokvafree(sva, len);
1.64 thorpej 280: }
281:
282: static size_t
1.152 matt 283: sodopendfree(void)
1.64 thorpej 284: {
1.93 yamt 285: size_t rv;
1.64 thorpej 286:
1.136 ad 287: mutex_enter(&so_pendfree_lock);
1.117 yamt 288: rv = sodopendfreel();
1.136 ad 289: mutex_exit(&so_pendfree_lock);
1.93 yamt 290:
291: return rv;
292: }
293:
294: /*
295: * sodopendfreel: free mbufs on "pendfree" list.
1.136 ad 296: * unlock and relock so_pendfree_lock when freeing mbufs.
1.93 yamt 297: *
1.136 ad 298: * => called with so_pendfree_lock held.
1.93 yamt 299: */
300:
301: static size_t
1.152 matt 302: sodopendfreel(void)
1.93 yamt 303: {
1.137 ad 304: struct mbuf *m, *next;
1.93 yamt 305: size_t rv = 0;
306:
1.136 ad 307: KASSERT(mutex_owned(&so_pendfree_lock));
1.64 thorpej 308:
1.137 ad 309: while (so_pendfree != NULL) {
1.64 thorpej 310: m = so_pendfree;
1.93 yamt 311: so_pendfree = NULL;
1.136 ad 312: mutex_exit(&so_pendfree_lock);
1.93 yamt 313:
314: for (; m != NULL; m = next) {
315: next = m->m_next;
316:
317: rv += m->m_ext.ext_size;
318: sodoloanfree((m->m_flags & M_EXT_PAGES) ?
319: m->m_ext.ext_pgs : NULL, m->m_ext.ext_buf,
320: m->m_ext.ext_size);
1.145 ad 321: pool_cache_put(mb_cache, m);
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:
1.142 dyoung 430: struct mbuf *
1.147 dyoung 431: getsombuf(struct socket *so, int type)
1.142 dyoung 432: {
433: struct mbuf *m;
434:
1.147 dyoung 435: m = m_get(M_WAIT, type);
1.142 dyoung 436: MCLAIM(m, so->so_mowner);
437: return m;
438: }
439:
440: struct mbuf *
441: m_intopt(struct socket *so, int val)
442: {
443: struct mbuf *m;
444:
1.147 dyoung 445: m = getsombuf(so, MT_SOOPTS);
1.142 dyoung 446: m->m_len = sizeof(int);
447: *mtod(m, int *) = val;
448: return m;
449: }
450:
1.119 yamt 451: void
452: soinit(void)
453: {
454:
1.148 ad 455: mutex_init(&so_pendfree_lock, MUTEX_DEFAULT, IPL_VM);
1.136 ad 456: cv_init(&socurkva_cv, "sokva");
457:
1.119 yamt 458: /* Set the initial adjusted socket buffer size. */
459: if (sb_max_set(sb_max))
460: panic("bad initial sb_max value: %lu", sb_max);
461:
462: callback_register(&vm_map_to_kernel(kernel_map)->vmk_reclaim_callback,
463: &sokva_reclaimerentry, NULL, sokva_reclaim_callback);
464: }
465:
1.1 cgd 466: /*
467: * Socket operation routines.
468: * These routines are called by the routines in
469: * sys_socket.c or from a system process, and
470: * implement the semantics of socket operations by
471: * switching out to the protocol specific routines.
472: */
473: /*ARGSUSED*/
1.3 andrew 474: int
1.114 christos 475: socreate(int dom, struct socket **aso, int type, int proto, struct lwp *l)
1.1 cgd 476: {
1.99 matt 477: const struct protosw *prp;
1.54 lukem 478: struct socket *so;
1.115 yamt 479: uid_t uid;
1.54 lukem 480: int error, s;
1.1 cgd 481:
1.132 elad 482: error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_SOCKET,
483: KAUTH_REQ_NETWORK_SOCKET_OPEN, KAUTH_ARG(dom), KAUTH_ARG(type),
484: KAUTH_ARG(proto));
1.140 dyoung 485: if (error != 0)
486: return error;
1.127 elad 487:
1.1 cgd 488: if (proto)
489: prp = pffindproto(dom, proto, type);
490: else
491: prp = pffindtype(dom, type);
1.140 dyoung 492: if (prp == NULL) {
1.120 ginsbach 493: /* no support for domain */
494: if (pffinddomain(dom) == 0)
1.140 dyoung 495: return EAFNOSUPPORT;
1.120 ginsbach 496: /* no support for socket type */
497: if (proto == 0 && type != 0)
1.140 dyoung 498: return EPROTOTYPE;
499: return EPROTONOSUPPORT;
1.120 ginsbach 500: }
1.140 dyoung 501: if (prp->pr_usrreq == NULL)
502: return EPROTONOSUPPORT;
1.1 cgd 503: if (prp->pr_type != type)
1.140 dyoung 504: return EPROTOTYPE;
1.39 matt 505: s = splsoftnet();
1.37 thorpej 506: so = pool_get(&socket_pool, PR_WAITOK);
1.140 dyoung 507: memset(so, 0, sizeof(*so));
1.31 thorpej 508: TAILQ_INIT(&so->so_q0);
509: TAILQ_INIT(&so->so_q);
1.1 cgd 510: so->so_type = type;
511: so->so_proto = prp;
1.33 matt 512: so->so_send = sosend;
513: so->so_receive = soreceive;
1.78 matt 514: #ifdef MBUFTRACE
515: so->so_rcv.sb_mowner = &prp->pr_domain->dom_mowner;
516: so->so_snd.sb_mowner = &prp->pr_domain->dom_mowner;
517: so->so_mowner = &prp->pr_domain->dom_mowner;
518: #endif
1.143 ad 519: selinit(&so->so_rcv.sb_sel);
520: selinit(&so->so_snd.sb_sel);
1.138 rmind 521: uid = kauth_cred_geteuid(l->l_cred);
1.115 yamt 522: so->so_uidinfo = uid_find(uid);
1.140 dyoung 523: error = (*prp->pr_usrreq)(so, PRU_ATTACH, NULL,
524: (struct mbuf *)(long)proto, NULL, l);
525: if (error != 0) {
1.1 cgd 526: so->so_state |= SS_NOFDREF;
527: sofree(so);
1.39 matt 528: splx(s);
1.140 dyoung 529: return error;
1.1 cgd 530: }
1.39 matt 531: splx(s);
1.1 cgd 532: *aso = so;
1.140 dyoung 533: return 0;
1.1 cgd 534: }
535:
1.142 dyoung 536: /* On success, write file descriptor to fdout and return zero. On
537: * failure, return non-zero; *fdout will be undefined.
538: */
539: int
540: fsocreate(int domain, struct socket **sop, int type, int protocol,
541: struct lwp *l, int *fdout)
542: {
543: struct socket *so;
544: struct file *fp;
545: int fd, error;
546:
1.155 ! ad 547: if ((error = fd_allocfile(&fp, &fd)) != 0)
1.142 dyoung 548: return (error);
549: fp->f_flag = FREAD|FWRITE;
550: fp->f_type = DTYPE_SOCKET;
551: fp->f_ops = &socketops;
552: error = socreate(domain, &so, type, protocol, l);
553: if (error != 0) {
1.155 ! ad 554: fd_abort(curproc, fp, fd);
1.142 dyoung 555: } else {
556: if (sop != NULL)
557: *sop = so;
558: fp->f_data = so;
1.155 ! ad 559: fd_affix(curproc, fp, fd);
1.142 dyoung 560: *fdout = fd;
561: }
562: return error;
563: }
564:
1.3 andrew 565: int
1.114 christos 566: sobind(struct socket *so, struct mbuf *nam, struct lwp *l)
1.1 cgd 567: {
1.54 lukem 568: int s, error;
1.1 cgd 569:
1.54 lukem 570: s = splsoftnet();
1.140 dyoung 571: error = (*so->so_proto->pr_usrreq)(so, PRU_BIND, NULL, nam, NULL, l);
1.1 cgd 572: splx(s);
1.140 dyoung 573: return error;
1.1 cgd 574: }
575:
1.3 andrew 576: int
1.150 elad 577: solisten(struct socket *so, int backlog, struct lwp *l)
1.1 cgd 578: {
1.54 lukem 579: int s, error;
1.1 cgd 580:
1.54 lukem 581: s = splsoftnet();
1.140 dyoung 582: error = (*so->so_proto->pr_usrreq)(so, PRU_LISTEN, NULL,
1.150 elad 583: NULL, NULL, l);
1.140 dyoung 584: if (error != 0) {
1.1 cgd 585: splx(s);
1.140 dyoung 586: return error;
1.1 cgd 587: }
1.63 matt 588: if (TAILQ_EMPTY(&so->so_q))
1.1 cgd 589: so->so_options |= SO_ACCEPTCONN;
590: if (backlog < 0)
591: backlog = 0;
1.49 jonathan 592: so->so_qlimit = min(backlog, somaxconn);
1.1 cgd 593: splx(s);
1.140 dyoung 594: return 0;
1.1 cgd 595: }
596:
1.21 christos 597: void
1.54 lukem 598: sofree(struct socket *so)
1.1 cgd 599: {
600:
1.43 mycroft 601: if (so->so_pcb || (so->so_state & SS_NOFDREF) == 0)
1.1 cgd 602: return;
1.43 mycroft 603: if (so->so_head) {
604: /*
605: * We must not decommission a socket that's on the accept(2)
606: * queue. If we do, then accept(2) may hang after select(2)
607: * indicated that the listening socket was ready.
608: */
609: if (!soqremque(so, 0))
610: return;
611: }
1.98 christos 612: if (so->so_rcv.sb_hiwat)
1.110 christos 613: (void)chgsbsize(so->so_uidinfo, &so->so_rcv.sb_hiwat, 0,
1.98 christos 614: RLIM_INFINITY);
615: if (so->so_snd.sb_hiwat)
1.110 christos 616: (void)chgsbsize(so->so_uidinfo, &so->so_snd.sb_hiwat, 0,
1.98 christos 617: RLIM_INFINITY);
618: sbrelease(&so->so_snd, so);
1.1 cgd 619: sorflush(so);
1.143 ad 620: seldestroy(&so->so_rcv.sb_sel);
621: seldestroy(&so->so_snd.sb_sel);
1.37 thorpej 622: pool_put(&socket_pool, so);
1.1 cgd 623: }
624:
625: /*
626: * Close a socket on last file table reference removal.
627: * Initiate disconnect if connected.
628: * Free socket when disconnect complete.
629: */
1.3 andrew 630: int
1.54 lukem 631: soclose(struct socket *so)
1.1 cgd 632: {
1.54 lukem 633: struct socket *so2;
634: int s, error;
1.1 cgd 635:
1.54 lukem 636: error = 0;
637: s = splsoftnet(); /* conservative */
1.1 cgd 638: if (so->so_options & SO_ACCEPTCONN) {
1.63 matt 639: while ((so2 = TAILQ_FIRST(&so->so_q0)) != 0) {
1.42 mycroft 640: (void) soqremque(so2, 0);
1.41 mycroft 641: (void) soabort(so2);
642: }
1.63 matt 643: while ((so2 = TAILQ_FIRST(&so->so_q)) != 0) {
1.42 mycroft 644: (void) soqremque(so2, 1);
1.41 mycroft 645: (void) soabort(so2);
646: }
1.1 cgd 647: }
648: if (so->so_pcb == 0)
649: goto discard;
650: if (so->so_state & SS_ISCONNECTED) {
651: if ((so->so_state & SS_ISDISCONNECTING) == 0) {
652: error = sodisconnect(so);
653: if (error)
654: goto drop;
655: }
656: if (so->so_options & SO_LINGER) {
1.151 ad 657: if ((so->so_state & SS_ISDISCONNECTING) && so->so_nbio)
1.1 cgd 658: goto drop;
1.21 christos 659: while (so->so_state & SS_ISCONNECTED) {
1.134 christos 660: error = tsleep((void *)&so->so_timeo,
1.21 christos 661: PSOCK | PCATCH, netcls,
1.30 thorpej 662: so->so_linger * hz);
1.21 christos 663: if (error)
1.1 cgd 664: break;
1.21 christos 665: }
1.1 cgd 666: }
667: }
1.54 lukem 668: drop:
1.1 cgd 669: if (so->so_pcb) {
1.22 mycroft 670: int error2 = (*so->so_proto->pr_usrreq)(so, PRU_DETACH,
1.140 dyoung 671: NULL, NULL, NULL, NULL);
1.1 cgd 672: if (error == 0)
673: error = error2;
674: }
1.54 lukem 675: discard:
1.1 cgd 676: if (so->so_state & SS_NOFDREF)
677: panic("soclose: NOFDREF");
678: so->so_state |= SS_NOFDREF;
679: sofree(so);
680: splx(s);
681: return (error);
682: }
683:
684: /*
1.20 mycroft 685: * Must be called at splsoftnet...
1.1 cgd 686: */
1.3 andrew 687: int
1.54 lukem 688: soabort(struct socket *so)
1.1 cgd 689: {
1.139 yamt 690: int error;
1.1 cgd 691:
1.139 yamt 692: KASSERT(so->so_head == NULL);
1.140 dyoung 693: error = (*so->so_proto->pr_usrreq)(so, PRU_ABORT, NULL,
694: NULL, NULL, NULL);
1.139 yamt 695: if (error) {
696: sofree(so);
697: }
698: return error;
1.1 cgd 699: }
700:
1.3 andrew 701: int
1.54 lukem 702: soaccept(struct socket *so, struct mbuf *nam)
1.1 cgd 703: {
1.54 lukem 704: int s, error;
1.1 cgd 705:
1.54 lukem 706: error = 0;
707: s = splsoftnet();
1.1 cgd 708: if ((so->so_state & SS_NOFDREF) == 0)
709: panic("soaccept: !NOFDREF");
710: so->so_state &= ~SS_NOFDREF;
1.55 thorpej 711: if ((so->so_state & SS_ISDISCONNECTED) == 0 ||
712: (so->so_proto->pr_flags & PR_ABRTACPTDIS) == 0)
1.41 mycroft 713: error = (*so->so_proto->pr_usrreq)(so, PRU_ACCEPT,
1.140 dyoung 714: NULL, nam, NULL, NULL);
1.41 mycroft 715: else
1.53 itojun 716: error = ECONNABORTED;
1.52 itojun 717:
1.1 cgd 718: splx(s);
719: return (error);
720: }
721:
1.3 andrew 722: int
1.114 christos 723: soconnect(struct socket *so, struct mbuf *nam, struct lwp *l)
1.1 cgd 724: {
1.54 lukem 725: int s, error;
1.1 cgd 726:
727: if (so->so_options & SO_ACCEPTCONN)
728: return (EOPNOTSUPP);
1.20 mycroft 729: s = splsoftnet();
1.1 cgd 730: /*
731: * If protocol is connection-based, can only connect once.
732: * Otherwise, if connected, try to disconnect first.
733: * This allows user to disconnect by connecting to, e.g.,
734: * a null address.
735: */
736: if (so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING) &&
737: ((so->so_proto->pr_flags & PR_CONNREQUIRED) ||
738: (error = sodisconnect(so))))
739: error = EISCONN;
740: else
741: error = (*so->so_proto->pr_usrreq)(so, PRU_CONNECT,
1.140 dyoung 742: NULL, nam, NULL, l);
1.1 cgd 743: splx(s);
744: return (error);
745: }
746:
1.3 andrew 747: int
1.54 lukem 748: soconnect2(struct socket *so1, struct socket *so2)
1.1 cgd 749: {
1.54 lukem 750: int s, error;
1.1 cgd 751:
1.54 lukem 752: s = splsoftnet();
1.22 mycroft 753: error = (*so1->so_proto->pr_usrreq)(so1, PRU_CONNECT2,
1.140 dyoung 754: NULL, (struct mbuf *)so2, NULL, NULL);
1.1 cgd 755: splx(s);
756: return (error);
757: }
758:
1.3 andrew 759: int
1.54 lukem 760: sodisconnect(struct socket *so)
1.1 cgd 761: {
1.54 lukem 762: int s, error;
1.1 cgd 763:
1.54 lukem 764: s = splsoftnet();
1.1 cgd 765: if ((so->so_state & SS_ISCONNECTED) == 0) {
766: error = ENOTCONN;
767: goto bad;
768: }
769: if (so->so_state & SS_ISDISCONNECTING) {
770: error = EALREADY;
771: goto bad;
772: }
1.22 mycroft 773: error = (*so->so_proto->pr_usrreq)(so, PRU_DISCONNECT,
1.140 dyoung 774: NULL, NULL, NULL, NULL);
1.54 lukem 775: bad:
1.1 cgd 776: splx(s);
1.117 yamt 777: sodopendfree();
1.1 cgd 778: return (error);
779: }
780:
1.15 mycroft 781: #define SBLOCKWAIT(f) (((f) & MSG_DONTWAIT) ? M_NOWAIT : M_WAITOK)
1.1 cgd 782: /*
783: * Send on a socket.
784: * If send must go all at once and message is larger than
785: * send buffering, then hard error.
786: * Lock against other senders.
787: * If must go all at once and not enough room now, then
788: * inform user that this would block and do nothing.
789: * Otherwise, if nonblocking, send as much as possible.
790: * The data to be sent is described by "uio" if nonzero,
791: * otherwise by the mbuf chain "top" (which must be null
792: * if uio is not). Data provided in mbuf chain must be small
793: * enough to send all at once.
794: *
795: * Returns nonzero on error, timeout or signal; callers
796: * must check for short counts if EINTR/ERESTART are returned.
797: * Data and control buffers are freed on return.
798: */
1.3 andrew 799: int
1.54 lukem 800: sosend(struct socket *so, struct mbuf *addr, struct uio *uio, struct mbuf *top,
1.114 christos 801: struct mbuf *control, int flags, struct lwp *l)
1.1 cgd 802: {
1.54 lukem 803: struct mbuf **mp, *m;
1.114 christos 804: struct proc *p;
1.58 jdolecek 805: long space, len, resid, clen, mlen;
806: int error, s, dontroute, atomic;
1.54 lukem 807:
1.114 christos 808: p = l->l_proc;
1.117 yamt 809: sodopendfree();
1.64 thorpej 810:
1.54 lukem 811: clen = 0;
812: atomic = sosendallatonce(so) || top;
1.1 cgd 813: if (uio)
814: resid = uio->uio_resid;
815: else
816: resid = top->m_pkthdr.len;
1.7 cgd 817: /*
818: * In theory resid should be unsigned.
819: * However, space must be signed, as it might be less than 0
820: * if we over-committed, and we must use a signed comparison
821: * of space and resid. On the other hand, a negative resid
822: * causes us to loop sending 0-length segments to the protocol.
823: */
1.29 mycroft 824: if (resid < 0) {
825: error = EINVAL;
826: goto out;
827: }
1.1 cgd 828: dontroute =
829: (flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 &&
830: (so->so_proto->pr_flags & PR_ATOMIC);
1.102 jonathan 831: if (p)
832: p->p_stats->p_ru.ru_msgsnd++;
1.1 cgd 833: if (control)
834: clen = control->m_len;
835: #define snderr(errno) { error = errno; splx(s); goto release; }
836:
1.54 lukem 837: restart:
1.21 christos 838: if ((error = sblock(&so->so_snd, SBLOCKWAIT(flags))) != 0)
1.1 cgd 839: goto out;
840: do {
1.20 mycroft 841: s = splsoftnet();
1.1 cgd 842: if (so->so_state & SS_CANTSENDMORE)
843: snderr(EPIPE);
1.48 thorpej 844: if (so->so_error) {
845: error = so->so_error;
846: so->so_error = 0;
847: splx(s);
848: goto release;
849: }
1.1 cgd 850: if ((so->so_state & SS_ISCONNECTED) == 0) {
851: if (so->so_proto->pr_flags & PR_CONNREQUIRED) {
852: if ((so->so_state & SS_ISCONFIRMING) == 0 &&
853: !(resid == 0 && clen != 0))
854: snderr(ENOTCONN);
855: } else if (addr == 0)
856: snderr(EDESTADDRREQ);
857: }
858: space = sbspace(&so->so_snd);
859: if (flags & MSG_OOB)
860: space += 1024;
1.21 christos 861: if ((atomic && resid > so->so_snd.sb_hiwat) ||
1.11 mycroft 862: clen > so->so_snd.sb_hiwat)
863: snderr(EMSGSIZE);
1.96 mycroft 864: if (space < resid + clen &&
1.1 cgd 865: (atomic || space < so->so_snd.sb_lowat || space < clen)) {
1.151 ad 866: if (so->so_nbio)
1.1 cgd 867: snderr(EWOULDBLOCK);
868: sbunlock(&so->so_snd);
869: error = sbwait(&so->so_snd);
870: splx(s);
871: if (error)
872: goto out;
873: goto restart;
874: }
875: splx(s);
876: mp = ⊤
877: space -= clen;
878: do {
1.45 tv 879: if (uio == NULL) {
880: /*
881: * Data is prepackaged in "top".
882: */
883: resid = 0;
884: if (flags & MSG_EOR)
885: top->m_flags |= M_EOR;
886: } else do {
1.144 dyoung 887: if (top == NULL) {
1.78 matt 888: m = m_gethdr(M_WAIT, MT_DATA);
1.45 tv 889: mlen = MHLEN;
890: m->m_pkthdr.len = 0;
1.140 dyoung 891: m->m_pkthdr.rcvif = NULL;
1.45 tv 892: } else {
1.78 matt 893: m = m_get(M_WAIT, MT_DATA);
1.45 tv 894: mlen = MLEN;
895: }
1.78 matt 896: MCLAIM(m, so->so_snd.sb_mowner);
1.121 yamt 897: if (sock_loan_thresh >= 0 &&
898: uio->uio_iov->iov_len >= sock_loan_thresh &&
899: space >= sock_loan_thresh &&
1.64 thorpej 900: (len = sosend_loan(so, uio, m,
901: space)) != 0) {
902: SOSEND_COUNTER_INCR(&sosend_loan_big);
903: space -= len;
904: goto have_data;
905: }
1.45 tv 906: if (resid >= MINCLSIZE && space >= MCLBYTES) {
1.64 thorpej 907: SOSEND_COUNTER_INCR(&sosend_copy_big);
1.78 matt 908: m_clget(m, M_WAIT);
1.45 tv 909: if ((m->m_flags & M_EXT) == 0)
910: goto nopages;
911: mlen = MCLBYTES;
912: if (atomic && top == 0) {
1.58 jdolecek 913: len = lmin(MCLBYTES - max_hdr,
1.54 lukem 914: resid);
1.45 tv 915: m->m_data += max_hdr;
916: } else
1.58 jdolecek 917: len = lmin(MCLBYTES, resid);
1.45 tv 918: space -= len;
919: } else {
1.64 thorpej 920: nopages:
921: SOSEND_COUNTER_INCR(&sosend_copy_small);
1.58 jdolecek 922: len = lmin(lmin(mlen, resid), space);
1.45 tv 923: space -= len;
924: /*
925: * For datagram protocols, leave room
926: * for protocol headers in first mbuf.
927: */
928: if (atomic && top == 0 && len < mlen)
929: MH_ALIGN(m, len);
930: }
1.144 dyoung 931: error = uiomove(mtod(m, void *), (int)len, uio);
1.64 thorpej 932: have_data:
1.45 tv 933: resid = uio->uio_resid;
934: m->m_len = len;
935: *mp = m;
936: top->m_pkthdr.len += len;
1.144 dyoung 937: if (error != 0)
1.45 tv 938: goto release;
939: mp = &m->m_next;
940: if (resid <= 0) {
941: if (flags & MSG_EOR)
942: top->m_flags |= M_EOR;
943: break;
944: }
945: } while (space > 0 && atomic);
1.108 perry 946:
1.46 sommerfe 947: s = splsoftnet();
948:
949: if (so->so_state & SS_CANTSENDMORE)
950: snderr(EPIPE);
1.45 tv 951:
952: if (dontroute)
953: so->so_options |= SO_DONTROUTE;
954: if (resid > 0)
955: so->so_state |= SS_MORETOCOME;
1.46 sommerfe 956: error = (*so->so_proto->pr_usrreq)(so,
957: (flags & MSG_OOB) ? PRU_SENDOOB : PRU_SEND,
1.114 christos 958: top, addr, control, curlwp); /* XXX */
1.45 tv 959: if (dontroute)
960: so->so_options &= ~SO_DONTROUTE;
961: if (resid > 0)
962: so->so_state &= ~SS_MORETOCOME;
1.46 sommerfe 963: splx(s);
964:
1.45 tv 965: clen = 0;
1.144 dyoung 966: control = NULL;
967: top = NULL;
1.45 tv 968: mp = ⊤
1.144 dyoung 969: if (error != 0)
1.1 cgd 970: goto release;
971: } while (resid && space > 0);
972: } while (resid);
973:
1.54 lukem 974: release:
1.1 cgd 975: sbunlock(&so->so_snd);
1.54 lukem 976: out:
1.1 cgd 977: if (top)
978: m_freem(top);
979: if (control)
980: m_freem(control);
981: return (error);
982: }
983:
984: /*
985: * Implement receive operations on a socket.
986: * We depend on the way that records are added to the sockbuf
987: * by sbappend*. In particular, each record (mbufs linked through m_next)
988: * must begin with an address if the protocol so specifies,
989: * followed by an optional mbuf or mbufs containing ancillary data,
990: * and then zero or more mbufs of data.
991: * In order to avoid blocking network interrupts for the entire time here,
992: * we splx() while doing the actual copy to user space.
993: * Although the sockbuf is locked, new data may still be appended,
994: * and thus we must maintain consistency of the sockbuf during that time.
995: *
996: * The caller may receive the data as a single mbuf chain by supplying
997: * an mbuf **mp0 for use in returning the chain. The uio is then used
998: * only for the count in uio_resid.
999: */
1.3 andrew 1000: int
1.54 lukem 1001: soreceive(struct socket *so, struct mbuf **paddr, struct uio *uio,
1002: struct mbuf **mp0, struct mbuf **controlp, int *flagsp)
1.1 cgd 1003: {
1.116 yamt 1004: struct lwp *l = curlwp;
1.54 lukem 1005: struct mbuf *m, **mp;
1.146 dyoung 1006: int atomic, flags, len, error, s, offset, moff, type, orig_resid;
1.99 matt 1007: const struct protosw *pr;
1.54 lukem 1008: struct mbuf *nextrecord;
1.67 he 1009: int mbuf_removed = 0;
1.146 dyoung 1010: const struct domain *dom;
1.64 thorpej 1011:
1.54 lukem 1012: pr = so->so_proto;
1.146 dyoung 1013: atomic = pr->pr_flags & PR_ATOMIC;
1014: dom = pr->pr_domain;
1.1 cgd 1015: mp = mp0;
1.54 lukem 1016: type = 0;
1017: orig_resid = uio->uio_resid;
1.102 jonathan 1018:
1.144 dyoung 1019: if (paddr != NULL)
1020: *paddr = NULL;
1021: if (controlp != NULL)
1022: *controlp = NULL;
1023: if (flagsp != NULL)
1.1 cgd 1024: flags = *flagsp &~ MSG_EOR;
1025: else
1026: flags = 0;
1.66 enami 1027:
1028: if ((flags & MSG_DONTWAIT) == 0)
1.117 yamt 1029: sodopendfree();
1.66 enami 1030:
1.1 cgd 1031: if (flags & MSG_OOB) {
1032: m = m_get(M_WAIT, MT_DATA);
1.17 cgd 1033: error = (*pr->pr_usrreq)(so, PRU_RCVOOB, m,
1.140 dyoung 1034: (struct mbuf *)(long)(flags & MSG_PEEK), NULL, l);
1.1 cgd 1035: if (error)
1036: goto bad;
1037: do {
1.134 christos 1038: error = uiomove(mtod(m, void *),
1.1 cgd 1039: (int) min(uio->uio_resid, m->m_len), uio);
1040: m = m_free(m);
1.144 dyoung 1041: } while (uio->uio_resid > 0 && error == 0 && m);
1.54 lukem 1042: bad:
1.144 dyoung 1043: if (m != NULL)
1.1 cgd 1044: m_freem(m);
1.144 dyoung 1045: return error;
1.1 cgd 1046: }
1.144 dyoung 1047: if (mp != NULL)
1.140 dyoung 1048: *mp = NULL;
1.1 cgd 1049: if (so->so_state & SS_ISCONFIRMING && uio->uio_resid)
1.140 dyoung 1050: (*pr->pr_usrreq)(so, PRU_RCVD, NULL, NULL, NULL, l);
1.1 cgd 1051:
1.54 lukem 1052: restart:
1.21 christos 1053: if ((error = sblock(&so->so_rcv, SBLOCKWAIT(flags))) != 0)
1.144 dyoung 1054: return error;
1.20 mycroft 1055: s = splsoftnet();
1.1 cgd 1056:
1057: m = so->so_rcv.sb_mb;
1058: /*
1059: * If we have less data than requested, block awaiting more
1060: * (subject to any timeout) if:
1.15 mycroft 1061: * 1. the current count is less than the low water mark,
1.1 cgd 1062: * 2. MSG_WAITALL is set, and it is possible to do the entire
1.15 mycroft 1063: * receive operation at once if we block (resid <= hiwat), or
1064: * 3. MSG_DONTWAIT is not set.
1.1 cgd 1065: * If MSG_WAITALL is set but resid is larger than the receive buffer,
1066: * we have to do the receive in sections, and thus risk returning
1067: * a short count if a timeout or signal occurs after we start.
1068: */
1.144 dyoung 1069: if (m == NULL ||
1070: ((flags & MSG_DONTWAIT) == 0 &&
1071: so->so_rcv.sb_cc < uio->uio_resid &&
1072: (so->so_rcv.sb_cc < so->so_rcv.sb_lowat ||
1073: ((flags & MSG_WAITALL) &&
1074: uio->uio_resid <= so->so_rcv.sb_hiwat)) &&
1.146 dyoung 1075: m->m_nextpkt == NULL && !atomic)) {
1.1 cgd 1076: #ifdef DIAGNOSTIC
1.144 dyoung 1077: if (m == NULL && so->so_rcv.sb_cc)
1.1 cgd 1078: panic("receive 1");
1079: #endif
1080: if (so->so_error) {
1.144 dyoung 1081: if (m != NULL)
1.15 mycroft 1082: goto dontblock;
1.1 cgd 1083: error = so->so_error;
1084: if ((flags & MSG_PEEK) == 0)
1085: so->so_error = 0;
1086: goto release;
1087: }
1088: if (so->so_state & SS_CANTRCVMORE) {
1.144 dyoung 1089: if (m != NULL)
1.15 mycroft 1090: goto dontblock;
1.1 cgd 1091: else
1092: goto release;
1093: }
1.144 dyoung 1094: for (; m != NULL; m = m->m_next)
1.1 cgd 1095: if (m->m_type == MT_OOBDATA || (m->m_flags & M_EOR)) {
1096: m = so->so_rcv.sb_mb;
1097: goto dontblock;
1098: }
1099: if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0 &&
1100: (so->so_proto->pr_flags & PR_CONNREQUIRED)) {
1101: error = ENOTCONN;
1102: goto release;
1103: }
1104: if (uio->uio_resid == 0)
1105: goto release;
1.151 ad 1106: if (so->so_nbio || (flags & MSG_DONTWAIT)) {
1.1 cgd 1107: error = EWOULDBLOCK;
1108: goto release;
1109: }
1.69 thorpej 1110: SBLASTRECORDCHK(&so->so_rcv, "soreceive sbwait 1");
1111: SBLASTMBUFCHK(&so->so_rcv, "soreceive sbwait 1");
1.1 cgd 1112: sbunlock(&so->so_rcv);
1113: error = sbwait(&so->so_rcv);
1114: splx(s);
1.144 dyoung 1115: if (error != 0)
1116: return error;
1.1 cgd 1117: goto restart;
1118: }
1.54 lukem 1119: dontblock:
1.69 thorpej 1120: /*
1121: * On entry here, m points to the first record of the socket buffer.
1122: * While we process the initial mbufs containing address and control
1123: * info, we save a copy of m->m_nextpkt into nextrecord.
1124: */
1.144 dyoung 1125: if (l != NULL)
1.114 christos 1126: l->l_proc->p_stats->p_ru.ru_msgrcv++;
1.69 thorpej 1127: KASSERT(m == so->so_rcv.sb_mb);
1128: SBLASTRECORDCHK(&so->so_rcv, "soreceive 1");
1129: SBLASTMBUFCHK(&so->so_rcv, "soreceive 1");
1.1 cgd 1130: nextrecord = m->m_nextpkt;
1131: if (pr->pr_flags & PR_ADDR) {
1132: #ifdef DIAGNOSTIC
1133: if (m->m_type != MT_SONAME)
1134: panic("receive 1a");
1135: #endif
1.3 andrew 1136: orig_resid = 0;
1.1 cgd 1137: if (flags & MSG_PEEK) {
1138: if (paddr)
1139: *paddr = m_copy(m, 0, m->m_len);
1140: m = m->m_next;
1141: } else {
1142: sbfree(&so->so_rcv, m);
1.67 he 1143: mbuf_removed = 1;
1.144 dyoung 1144: if (paddr != NULL) {
1.1 cgd 1145: *paddr = m;
1146: so->so_rcv.sb_mb = m->m_next;
1.144 dyoung 1147: m->m_next = NULL;
1.1 cgd 1148: m = so->so_rcv.sb_mb;
1149: } else {
1150: MFREE(m, so->so_rcv.sb_mb);
1151: m = so->so_rcv.sb_mb;
1152: }
1153: }
1154: }
1.144 dyoung 1155: while (m != NULL && m->m_type == MT_CONTROL && error == 0) {
1.1 cgd 1156: if (flags & MSG_PEEK) {
1.144 dyoung 1157: if (controlp != NULL)
1.1 cgd 1158: *controlp = m_copy(m, 0, m->m_len);
1159: m = m->m_next;
1160: } else {
1161: sbfree(&so->so_rcv, m);
1.67 he 1162: mbuf_removed = 1;
1.144 dyoung 1163: if (controlp != NULL) {
1.114 christos 1164: if (dom->dom_externalize && l &&
1.1 cgd 1165: mtod(m, struct cmsghdr *)->cmsg_type ==
1166: SCM_RIGHTS)
1.114 christos 1167: error = (*dom->dom_externalize)(m, l);
1.1 cgd 1168: *controlp = m;
1169: so->so_rcv.sb_mb = m->m_next;
1.144 dyoung 1170: m->m_next = NULL;
1.1 cgd 1171: m = so->so_rcv.sb_mb;
1172: } else {
1.106 itojun 1173: /*
1174: * Dispose of any SCM_RIGHTS message that went
1175: * through the read path rather than recv.
1176: */
1.146 dyoung 1177: if (dom->dom_dispose &&
1.106 itojun 1178: mtod(m, struct cmsghdr *)->cmsg_type == SCM_RIGHTS)
1.146 dyoung 1179: (*dom->dom_dispose)(m);
1.1 cgd 1180: MFREE(m, so->so_rcv.sb_mb);
1181: m = so->so_rcv.sb_mb;
1182: }
1183: }
1.144 dyoung 1184: if (controlp != NULL) {
1.3 andrew 1185: orig_resid = 0;
1.1 cgd 1186: controlp = &(*controlp)->m_next;
1.3 andrew 1187: }
1.1 cgd 1188: }
1.69 thorpej 1189:
1190: /*
1191: * If m is non-NULL, we have some data to read. From now on,
1192: * make sure to keep sb_lastrecord consistent when working on
1193: * the last packet on the chain (nextrecord == NULL) and we
1194: * change m->m_nextpkt.
1195: */
1.144 dyoung 1196: if (m != NULL) {
1.69 thorpej 1197: if ((flags & MSG_PEEK) == 0) {
1.1 cgd 1198: m->m_nextpkt = nextrecord;
1.69 thorpej 1199: /*
1200: * If nextrecord == NULL (this is a single chain),
1201: * then sb_lastrecord may not be valid here if m
1202: * was changed earlier.
1203: */
1204: if (nextrecord == NULL) {
1205: KASSERT(so->so_rcv.sb_mb == m);
1206: so->so_rcv.sb_lastrecord = m;
1207: }
1208: }
1.1 cgd 1209: type = m->m_type;
1210: if (type == MT_OOBDATA)
1211: flags |= MSG_OOB;
1.69 thorpej 1212: } else {
1213: if ((flags & MSG_PEEK) == 0) {
1214: KASSERT(so->so_rcv.sb_mb == m);
1215: so->so_rcv.sb_mb = nextrecord;
1.70 thorpej 1216: SB_EMPTY_FIXUP(&so->so_rcv);
1.69 thorpej 1217: }
1.1 cgd 1218: }
1.69 thorpej 1219: SBLASTRECORDCHK(&so->so_rcv, "soreceive 2");
1220: SBLASTMBUFCHK(&so->so_rcv, "soreceive 2");
1221:
1.1 cgd 1222: moff = 0;
1223: offset = 0;
1.144 dyoung 1224: while (m != NULL && uio->uio_resid > 0 && error == 0) {
1.1 cgd 1225: if (m->m_type == MT_OOBDATA) {
1226: if (type != MT_OOBDATA)
1227: break;
1228: } else if (type == MT_OOBDATA)
1229: break;
1230: #ifdef DIAGNOSTIC
1231: else if (m->m_type != MT_DATA && m->m_type != MT_HEADER)
1232: panic("receive 3");
1233: #endif
1234: so->so_state &= ~SS_RCVATMARK;
1235: len = uio->uio_resid;
1236: if (so->so_oobmark && len > so->so_oobmark - offset)
1237: len = so->so_oobmark - offset;
1238: if (len > m->m_len - moff)
1239: len = m->m_len - moff;
1240: /*
1241: * If mp is set, just pass back the mbufs.
1242: * Otherwise copy them out via the uio, then free.
1243: * Sockbuf must be consistent here (points to current mbuf,
1244: * it points to next record) when we drop priority;
1245: * we must note any additions to the sockbuf when we
1246: * block interrupts again.
1247: */
1.144 dyoung 1248: if (mp == NULL) {
1.69 thorpej 1249: SBLASTRECORDCHK(&so->so_rcv, "soreceive uiomove");
1250: SBLASTMBUFCHK(&so->so_rcv, "soreceive uiomove");
1.1 cgd 1251: splx(s);
1.134 christos 1252: error = uiomove(mtod(m, char *) + moff, (int)len, uio);
1.20 mycroft 1253: s = splsoftnet();
1.144 dyoung 1254: if (error != 0) {
1.67 he 1255: /*
1256: * If any part of the record has been removed
1257: * (such as the MT_SONAME mbuf, which will
1258: * happen when PR_ADDR, and thus also
1259: * PR_ATOMIC, is set), then drop the entire
1260: * record to maintain the atomicity of the
1261: * receive operation.
1262: *
1263: * This avoids a later panic("receive 1a")
1264: * when compiled with DIAGNOSTIC.
1265: */
1.146 dyoung 1266: if (m && mbuf_removed && atomic)
1.67 he 1267: (void) sbdroprecord(&so->so_rcv);
1268:
1.57 jdolecek 1269: goto release;
1.67 he 1270: }
1.1 cgd 1271: } else
1272: uio->uio_resid -= len;
1273: if (len == m->m_len - moff) {
1274: if (m->m_flags & M_EOR)
1275: flags |= MSG_EOR;
1276: if (flags & MSG_PEEK) {
1277: m = m->m_next;
1278: moff = 0;
1279: } else {
1280: nextrecord = m->m_nextpkt;
1281: sbfree(&so->so_rcv, m);
1282: if (mp) {
1283: *mp = m;
1284: mp = &m->m_next;
1285: so->so_rcv.sb_mb = m = m->m_next;
1.140 dyoung 1286: *mp = NULL;
1.1 cgd 1287: } else {
1288: MFREE(m, so->so_rcv.sb_mb);
1289: m = so->so_rcv.sb_mb;
1290: }
1.69 thorpej 1291: /*
1292: * If m != NULL, we also know that
1293: * so->so_rcv.sb_mb != NULL.
1294: */
1295: KASSERT(so->so_rcv.sb_mb == m);
1296: if (m) {
1.1 cgd 1297: m->m_nextpkt = nextrecord;
1.69 thorpej 1298: if (nextrecord == NULL)
1299: so->so_rcv.sb_lastrecord = m;
1300: } else {
1301: so->so_rcv.sb_mb = nextrecord;
1.70 thorpej 1302: SB_EMPTY_FIXUP(&so->so_rcv);
1.69 thorpej 1303: }
1304: SBLASTRECORDCHK(&so->so_rcv, "soreceive 3");
1305: SBLASTMBUFCHK(&so->so_rcv, "soreceive 3");
1.1 cgd 1306: }
1.144 dyoung 1307: } else if (flags & MSG_PEEK)
1308: moff += len;
1309: else {
1310: if (mp != NULL)
1311: *mp = m_copym(m, 0, len, M_WAIT);
1312: m->m_data += len;
1313: m->m_len -= len;
1314: so->so_rcv.sb_cc -= len;
1.1 cgd 1315: }
1316: if (so->so_oobmark) {
1317: if ((flags & MSG_PEEK) == 0) {
1318: so->so_oobmark -= len;
1319: if (so->so_oobmark == 0) {
1320: so->so_state |= SS_RCVATMARK;
1321: break;
1322: }
1.7 cgd 1323: } else {
1.1 cgd 1324: offset += len;
1.7 cgd 1325: if (offset == so->so_oobmark)
1326: break;
1327: }
1.1 cgd 1328: }
1329: if (flags & MSG_EOR)
1330: break;
1331: /*
1332: * If the MSG_WAITALL flag is set (for non-atomic socket),
1333: * we must not quit until "uio->uio_resid == 0" or an error
1334: * termination. If a signal/timeout occurs, return
1335: * with a short count but without error.
1336: * Keep sockbuf locked against other readers.
1337: */
1.144 dyoung 1338: while (flags & MSG_WAITALL && m == NULL && uio->uio_resid > 0 &&
1.3 andrew 1339: !sosendallatonce(so) && !nextrecord) {
1.1 cgd 1340: if (so->so_error || so->so_state & SS_CANTRCVMORE)
1341: break;
1.68 matt 1342: /*
1343: * If we are peeking and the socket receive buffer is
1344: * full, stop since we can't get more data to peek at.
1345: */
1346: if ((flags & MSG_PEEK) && sbspace(&so->so_rcv) <= 0)
1347: break;
1348: /*
1349: * If we've drained the socket buffer, tell the
1350: * protocol in case it needs to do something to
1351: * get it filled again.
1352: */
1353: if ((pr->pr_flags & PR_WANTRCVD) && so->so_pcb)
1354: (*pr->pr_usrreq)(so, PRU_RCVD,
1.140 dyoung 1355: NULL, (struct mbuf *)(long)flags, NULL, l);
1.69 thorpej 1356: SBLASTRECORDCHK(&so->so_rcv, "soreceive sbwait 2");
1357: SBLASTMBUFCHK(&so->so_rcv, "soreceive sbwait 2");
1.1 cgd 1358: error = sbwait(&so->so_rcv);
1.144 dyoung 1359: if (error != 0) {
1.1 cgd 1360: sbunlock(&so->so_rcv);
1361: splx(s);
1.144 dyoung 1362: return 0;
1.1 cgd 1363: }
1.21 christos 1364: if ((m = so->so_rcv.sb_mb) != NULL)
1.1 cgd 1365: nextrecord = m->m_nextpkt;
1366: }
1367: }
1.3 andrew 1368:
1.146 dyoung 1369: if (m && atomic) {
1.3 andrew 1370: flags |= MSG_TRUNC;
1371: if ((flags & MSG_PEEK) == 0)
1372: (void) sbdroprecord(&so->so_rcv);
1373: }
1.1 cgd 1374: if ((flags & MSG_PEEK) == 0) {
1.144 dyoung 1375: if (m == NULL) {
1.69 thorpej 1376: /*
1.70 thorpej 1377: * First part is an inline SB_EMPTY_FIXUP(). Second
1.69 thorpej 1378: * part makes sure sb_lastrecord is up-to-date if
1379: * there is still data in the socket buffer.
1380: */
1.1 cgd 1381: so->so_rcv.sb_mb = nextrecord;
1.69 thorpej 1382: if (so->so_rcv.sb_mb == NULL) {
1383: so->so_rcv.sb_mbtail = NULL;
1384: so->so_rcv.sb_lastrecord = NULL;
1385: } else if (nextrecord->m_nextpkt == NULL)
1386: so->so_rcv.sb_lastrecord = nextrecord;
1387: }
1388: SBLASTRECORDCHK(&so->so_rcv, "soreceive 4");
1389: SBLASTMBUFCHK(&so->so_rcv, "soreceive 4");
1.1 cgd 1390: if (pr->pr_flags & PR_WANTRCVD && so->so_pcb)
1.140 dyoung 1391: (*pr->pr_usrreq)(so, PRU_RCVD, NULL,
1392: (struct mbuf *)(long)flags, NULL, l);
1.1 cgd 1393: }
1.3 andrew 1394: if (orig_resid == uio->uio_resid && orig_resid &&
1395: (flags & MSG_EOR) == 0 && (so->so_state & SS_CANTRCVMORE) == 0) {
1396: sbunlock(&so->so_rcv);
1397: splx(s);
1398: goto restart;
1399: }
1.108 perry 1400:
1.144 dyoung 1401: if (flagsp != NULL)
1.1 cgd 1402: *flagsp |= flags;
1.54 lukem 1403: release:
1.1 cgd 1404: sbunlock(&so->so_rcv);
1405: splx(s);
1.144 dyoung 1406: return error;
1.1 cgd 1407: }
1408:
1.14 mycroft 1409: int
1.54 lukem 1410: soshutdown(struct socket *so, int how)
1.1 cgd 1411: {
1.99 matt 1412: const struct protosw *pr;
1.34 kleink 1413:
1.54 lukem 1414: pr = so->so_proto;
1.34 kleink 1415: if (!(how == SHUT_RD || how == SHUT_WR || how == SHUT_RDWR))
1416: return (EINVAL);
1.1 cgd 1417:
1.34 kleink 1418: if (how == SHUT_RD || how == SHUT_RDWR)
1.1 cgd 1419: sorflush(so);
1.34 kleink 1420: if (how == SHUT_WR || how == SHUT_RDWR)
1.140 dyoung 1421: return (*pr->pr_usrreq)(so, PRU_SHUTDOWN, NULL,
1422: NULL, NULL, NULL);
1.144 dyoung 1423: return 0;
1.1 cgd 1424: }
1425:
1.14 mycroft 1426: void
1.54 lukem 1427: sorflush(struct socket *so)
1.1 cgd 1428: {
1.54 lukem 1429: struct sockbuf *sb, asb;
1.99 matt 1430: const struct protosw *pr;
1.54 lukem 1431: int s;
1.1 cgd 1432:
1.54 lukem 1433: sb = &so->so_rcv;
1434: pr = so->so_proto;
1.1 cgd 1435: sb->sb_flags |= SB_NOINTR;
1.15 mycroft 1436: (void) sblock(sb, M_WAITOK);
1.56 thorpej 1437: s = splnet();
1.1 cgd 1438: socantrcvmore(so);
1439: sbunlock(sb);
1440: asb = *sb;
1.86 wrstuden 1441: /*
1442: * Clear most of the sockbuf structure, but leave some of the
1443: * fields valid.
1444: */
1445: memset(&sb->sb_startzero, 0,
1446: sizeof(*sb) - offsetof(struct sockbuf, sb_startzero));
1.1 cgd 1447: splx(s);
1448: if (pr->pr_flags & PR_RIGHTS && pr->pr_domain->dom_dispose)
1449: (*pr->pr_domain->dom_dispose)(asb.sb_mb);
1.98 christos 1450: sbrelease(&asb, so);
1.1 cgd 1451: }
1452:
1.142 dyoung 1453: static int
1454: sosetopt1(struct socket *so, int level, int optname, struct mbuf *m)
1.1 cgd 1455: {
1.142 dyoung 1456: int optval, val;
1.130 christos 1457: struct linger *l;
1.141 yamt 1458: struct sockbuf *sb;
1.142 dyoung 1459: struct timeval *tv;
1460:
1461: switch (optname) {
1462:
1463: case SO_LINGER:
1464: if (m == NULL || m->m_len != sizeof(struct linger))
1465: return EINVAL;
1466: l = mtod(m, struct linger *);
1467: if (l->l_linger < 0 || l->l_linger > USHRT_MAX ||
1468: l->l_linger > (INT_MAX / hz))
1469: return EDOM;
1470: so->so_linger = l->l_linger;
1471: if (l->l_onoff)
1472: so->so_options |= SO_LINGER;
1473: else
1474: so->so_options &= ~SO_LINGER;
1475: break;
1.1 cgd 1476:
1.142 dyoung 1477: case SO_DEBUG:
1478: case SO_KEEPALIVE:
1479: case SO_DONTROUTE:
1480: case SO_USELOOPBACK:
1481: case SO_BROADCAST:
1482: case SO_REUSEADDR:
1483: case SO_REUSEPORT:
1484: case SO_OOBINLINE:
1485: case SO_TIMESTAMP:
1486: if (m == NULL || m->m_len < sizeof(int))
1487: return EINVAL;
1488: if (*mtod(m, int *))
1489: so->so_options |= optname;
1490: else
1491: so->so_options &= ~optname;
1492: break;
1493:
1494: case SO_SNDBUF:
1495: case SO_RCVBUF:
1496: case SO_SNDLOWAT:
1497: case SO_RCVLOWAT:
1498: if (m == NULL || m->m_len < sizeof(int))
1499: return EINVAL;
1.1 cgd 1500:
1.142 dyoung 1501: /*
1502: * Values < 1 make no sense for any of these
1503: * options, so disallow them.
1504: */
1505: optval = *mtod(m, int *);
1506: if (optval < 1)
1507: return EINVAL;
1.1 cgd 1508:
1.142 dyoung 1509: switch (optname) {
1.1 cgd 1510:
1511: case SO_SNDBUF:
1512: case SO_RCVBUF:
1.142 dyoung 1513: sb = (optname == SO_SNDBUF) ?
1514: &so->so_snd : &so->so_rcv;
1515: if (sbreserve(sb, (u_long)optval, so) == 0)
1516: return ENOBUFS;
1517: sb->sb_flags &= ~SB_AUTOSIZE;
1518: break;
1519:
1520: /*
1521: * Make sure the low-water is never greater than
1522: * the high-water.
1523: */
1.1 cgd 1524: case SO_SNDLOWAT:
1.142 dyoung 1525: so->so_snd.sb_lowat =
1526: (optval > so->so_snd.sb_hiwat) ?
1527: so->so_snd.sb_hiwat : optval;
1528: break;
1.1 cgd 1529: case SO_RCVLOWAT:
1.142 dyoung 1530: so->so_rcv.sb_lowat =
1531: (optval > so->so_rcv.sb_hiwat) ?
1532: so->so_rcv.sb_hiwat : optval;
1533: break;
1534: }
1535: break;
1.28 thorpej 1536:
1.142 dyoung 1537: case SO_SNDTIMEO:
1538: case SO_RCVTIMEO:
1539: if (m == NULL || m->m_len < sizeof(*tv))
1540: return EINVAL;
1541: tv = mtod(m, struct timeval *);
1542: if (tv->tv_sec > (INT_MAX - tv->tv_usec / tick) / hz)
1543: return EDOM;
1544: val = tv->tv_sec * hz + tv->tv_usec / tick;
1545: if (val == 0 && tv->tv_usec != 0)
1546: val = 1;
1.28 thorpej 1547:
1.142 dyoung 1548: switch (optname) {
1.28 thorpej 1549:
1.142 dyoung 1550: case SO_SNDTIMEO:
1551: so->so_snd.sb_timeo = val;
1.1 cgd 1552: break;
1553: case SO_RCVTIMEO:
1.142 dyoung 1554: so->so_rcv.sb_timeo = val;
1555: break;
1556: }
1557: break;
1.1 cgd 1558:
1.142 dyoung 1559: default:
1560: return ENOPROTOOPT;
1561: }
1562: return 0;
1563: }
1.1 cgd 1564:
1.142 dyoung 1565: int
1566: sosetopt(struct socket *so, int level, int optname, struct mbuf *m)
1567: {
1568: int error, prerr;
1.1 cgd 1569:
1.142 dyoung 1570: if (level == SOL_SOCKET)
1571: error = sosetopt1(so, level, optname, m);
1572: else
1573: error = ENOPROTOOPT;
1.1 cgd 1574:
1.142 dyoung 1575: if ((error == 0 || error == ENOPROTOOPT) &&
1576: so->so_proto != NULL && so->so_proto->pr_ctloutput != NULL) {
1577: /* give the protocol stack a shot */
1578: prerr = (*so->so_proto->pr_ctloutput)(PRCO_SETOPT, so, level,
1579: optname, &m);
1580: if (prerr == 0)
1581: error = 0;
1582: else if (prerr != ENOPROTOOPT)
1583: error = prerr;
1584: } else if (m != NULL)
1585: (void)m_free(m);
1586: return error;
1.1 cgd 1587: }
1588:
1.14 mycroft 1589: int
1.54 lukem 1590: sogetopt(struct socket *so, int level, int optname, struct mbuf **mp)
1.1 cgd 1591: {
1.54 lukem 1592: struct mbuf *m;
1.1 cgd 1593:
1594: if (level != SOL_SOCKET) {
1595: if (so->so_proto && so->so_proto->pr_ctloutput) {
1596: return ((*so->so_proto->pr_ctloutput)
1597: (PRCO_GETOPT, so, level, optname, mp));
1598: } else
1599: return (ENOPROTOOPT);
1600: } else {
1601: m = m_get(M_WAIT, MT_SOOPTS);
1.36 perry 1602: m->m_len = sizeof(int);
1.1 cgd 1603:
1604: switch (optname) {
1605:
1606: case SO_LINGER:
1.36 perry 1607: m->m_len = sizeof(struct linger);
1.1 cgd 1608: mtod(m, struct linger *)->l_onoff =
1.131 christos 1609: (so->so_options & SO_LINGER) ? 1 : 0;
1.1 cgd 1610: mtod(m, struct linger *)->l_linger = so->so_linger;
1611: break;
1612:
1613: case SO_USELOOPBACK:
1614: case SO_DONTROUTE:
1615: case SO_DEBUG:
1616: case SO_KEEPALIVE:
1617: case SO_REUSEADDR:
1.15 mycroft 1618: case SO_REUSEPORT:
1.1 cgd 1619: case SO_BROADCAST:
1620: case SO_OOBINLINE:
1.26 thorpej 1621: case SO_TIMESTAMP:
1.131 christos 1622: *mtod(m, int *) = (so->so_options & optname) ? 1 : 0;
1.1 cgd 1623: break;
1624:
1625: case SO_TYPE:
1626: *mtod(m, int *) = so->so_type;
1627: break;
1628:
1629: case SO_ERROR:
1630: *mtod(m, int *) = so->so_error;
1631: so->so_error = 0;
1632: break;
1633:
1634: case SO_SNDBUF:
1635: *mtod(m, int *) = so->so_snd.sb_hiwat;
1636: break;
1637:
1638: case SO_RCVBUF:
1639: *mtod(m, int *) = so->so_rcv.sb_hiwat;
1640: break;
1641:
1642: case SO_SNDLOWAT:
1643: *mtod(m, int *) = so->so_snd.sb_lowat;
1644: break;
1645:
1646: case SO_RCVLOWAT:
1647: *mtod(m, int *) = so->so_rcv.sb_lowat;
1648: break;
1649:
1650: case SO_SNDTIMEO:
1651: case SO_RCVTIMEO:
1652: {
1653: int val = (optname == SO_SNDTIMEO ?
1654: so->so_snd.sb_timeo : so->so_rcv.sb_timeo);
1655:
1656: m->m_len = sizeof(struct timeval);
1657: mtod(m, struct timeval *)->tv_sec = val / hz;
1658: mtod(m, struct timeval *)->tv_usec =
1.27 kleink 1659: (val % hz) * tick;
1.1 cgd 1660: break;
1661: }
1662:
1.107 darrenr 1663: case SO_OVERFLOWED:
1664: *mtod(m, int *) = so->so_rcv.sb_overflowed;
1665: break;
1666:
1.1 cgd 1667: default:
1668: (void)m_free(m);
1669: return (ENOPROTOOPT);
1670: }
1671: *mp = m;
1672: return (0);
1673: }
1674: }
1675:
1.14 mycroft 1676: void
1.54 lukem 1677: sohasoutofband(struct socket *so)
1.1 cgd 1678: {
1.153 rmind 1679:
1.90 christos 1680: fownsignal(so->so_pgid, SIGURG, POLL_PRI, POLLPRI|POLLRDBAND, so);
1.153 rmind 1681: selnotify(&so->so_rcv.sb_sel, POLLPRI | POLLRDBAND, 0);
1.1 cgd 1682: }
1.72 jdolecek 1683:
1684: static void
1685: filt_sordetach(struct knote *kn)
1686: {
1687: struct socket *so;
1688:
1.155 ! ad 1689: so = ((file_t *)kn->kn_obj)->f_data;
1.73 christos 1690: SLIST_REMOVE(&so->so_rcv.sb_sel.sel_klist, kn, knote, kn_selnext);
1691: if (SLIST_EMPTY(&so->so_rcv.sb_sel.sel_klist))
1.72 jdolecek 1692: so->so_rcv.sb_flags &= ~SB_KNOTE;
1693: }
1694:
1695: /*ARGSUSED*/
1696: static int
1.129 yamt 1697: filt_soread(struct knote *kn, long hint)
1.72 jdolecek 1698: {
1699: struct socket *so;
1700:
1.155 ! ad 1701: so = ((file_t *)kn->kn_obj)->f_data;
1.72 jdolecek 1702: kn->kn_data = so->so_rcv.sb_cc;
1703: if (so->so_state & SS_CANTRCVMORE) {
1.108 perry 1704: kn->kn_flags |= EV_EOF;
1.72 jdolecek 1705: kn->kn_fflags = so->so_error;
1706: return (1);
1707: }
1708: if (so->so_error) /* temporary udp error */
1709: return (1);
1710: if (kn->kn_sfflags & NOTE_LOWAT)
1711: return (kn->kn_data >= kn->kn_sdata);
1712: return (kn->kn_data >= so->so_rcv.sb_lowat);
1713: }
1714:
1715: static void
1716: filt_sowdetach(struct knote *kn)
1717: {
1718: struct socket *so;
1719:
1.155 ! ad 1720: so = ((file_t *)kn->kn_obj)->f_data;
1.73 christos 1721: SLIST_REMOVE(&so->so_snd.sb_sel.sel_klist, kn, knote, kn_selnext);
1722: if (SLIST_EMPTY(&so->so_snd.sb_sel.sel_klist))
1.72 jdolecek 1723: so->so_snd.sb_flags &= ~SB_KNOTE;
1724: }
1725:
1726: /*ARGSUSED*/
1727: static int
1.129 yamt 1728: filt_sowrite(struct knote *kn, long hint)
1.72 jdolecek 1729: {
1730: struct socket *so;
1731:
1.155 ! ad 1732: so = ((file_t *)kn->kn_obj)->f_data;
1.72 jdolecek 1733: kn->kn_data = sbspace(&so->so_snd);
1734: if (so->so_state & SS_CANTSENDMORE) {
1.108 perry 1735: kn->kn_flags |= EV_EOF;
1.72 jdolecek 1736: kn->kn_fflags = so->so_error;
1737: return (1);
1738: }
1739: if (so->so_error) /* temporary udp error */
1740: return (1);
1741: if (((so->so_state & SS_ISCONNECTED) == 0) &&
1742: (so->so_proto->pr_flags & PR_CONNREQUIRED))
1743: return (0);
1744: if (kn->kn_sfflags & NOTE_LOWAT)
1745: return (kn->kn_data >= kn->kn_sdata);
1746: return (kn->kn_data >= so->so_snd.sb_lowat);
1747: }
1748:
1749: /*ARGSUSED*/
1750: static int
1.129 yamt 1751: filt_solisten(struct knote *kn, long hint)
1.72 jdolecek 1752: {
1753: struct socket *so;
1754:
1.155 ! ad 1755: so = ((file_t *)kn->kn_obj)->f_data;
1.72 jdolecek 1756:
1757: /*
1758: * Set kn_data to number of incoming connections, not
1759: * counting partial (incomplete) connections.
1.108 perry 1760: */
1.72 jdolecek 1761: kn->kn_data = so->so_qlen;
1762: return (kn->kn_data > 0);
1763: }
1764:
1765: static const struct filterops solisten_filtops =
1766: { 1, NULL, filt_sordetach, filt_solisten };
1767: static const struct filterops soread_filtops =
1768: { 1, NULL, filt_sordetach, filt_soread };
1769: static const struct filterops sowrite_filtops =
1770: { 1, NULL, filt_sowdetach, filt_sowrite };
1771:
1772: int
1.129 yamt 1773: soo_kqfilter(struct file *fp, struct knote *kn)
1.72 jdolecek 1774: {
1775: struct socket *so;
1776: struct sockbuf *sb;
1777:
1.155 ! ad 1778: so = ((file_t *)kn->kn_obj)->f_data;
1.72 jdolecek 1779: switch (kn->kn_filter) {
1780: case EVFILT_READ:
1781: if (so->so_options & SO_ACCEPTCONN)
1782: kn->kn_fop = &solisten_filtops;
1783: else
1784: kn->kn_fop = &soread_filtops;
1785: sb = &so->so_rcv;
1786: break;
1787: case EVFILT_WRITE:
1788: kn->kn_fop = &sowrite_filtops;
1789: sb = &so->so_snd;
1790: break;
1791: default:
1.149 pooka 1792: return (EINVAL);
1.72 jdolecek 1793: }
1.73 christos 1794: SLIST_INSERT_HEAD(&sb->sb_sel.sel_klist, kn, kn_selnext);
1.72 jdolecek 1795: sb->sb_flags |= SB_KNOTE;
1796: return (0);
1797: }
1798:
1.154 ad 1799: static int
1800: sodopoll(struct socket *so, int events)
1801: {
1802: int revents;
1803:
1804: revents = 0;
1805:
1806: if (events & (POLLIN | POLLRDNORM))
1807: if (soreadable(so))
1808: revents |= events & (POLLIN | POLLRDNORM);
1809:
1810: if (events & (POLLOUT | POLLWRNORM))
1811: if (sowritable(so))
1812: revents |= events & (POLLOUT | POLLWRNORM);
1813:
1814: if (events & (POLLPRI | POLLRDBAND))
1815: if (so->so_oobmark || (so->so_state & SS_RCVATMARK))
1816: revents |= events & (POLLPRI | POLLRDBAND);
1817:
1818: return revents;
1819: }
1820:
1821: int
1822: sopoll(struct socket *so, int events)
1823: {
1824: int revents = 0;
1825: int s;
1826:
1827: if ((revents = sodopoll(so, events)) != 0)
1828: return revents;
1829:
1830: KERNEL_LOCK(1, curlwp);
1831: s = splsoftnet();
1832:
1833: if ((revents = sodopoll(so, events)) == 0) {
1834: if (events & (POLLIN | POLLPRI | POLLRDNORM | POLLRDBAND)) {
1835: selrecord(curlwp, &so->so_rcv.sb_sel);
1836: so->so_rcv.sb_flags |= SB_SEL;
1837: }
1838:
1839: if (events & (POLLOUT | POLLWRNORM)) {
1840: selrecord(curlwp, &so->so_snd.sb_sel);
1841: so->so_snd.sb_flags |= SB_SEL;
1842: }
1843: }
1844:
1845: splx(s);
1846: KERNEL_UNLOCK_ONE(curlwp);
1847:
1848: return revents;
1849: }
1850:
1851:
1.94 yamt 1852: #include <sys/sysctl.h>
1853:
1854: static int sysctl_kern_somaxkva(SYSCTLFN_PROTO);
1855:
1856: /*
1857: * sysctl helper routine for kern.somaxkva. ensures that the given
1858: * value is not too small.
1859: * (XXX should we maybe make sure it's not too large as well?)
1860: */
1861: static int
1862: sysctl_kern_somaxkva(SYSCTLFN_ARGS)
1863: {
1864: int error, new_somaxkva;
1865: struct sysctlnode node;
1866:
1867: new_somaxkva = somaxkva;
1868: node = *rnode;
1869: node.sysctl_data = &new_somaxkva;
1870: error = sysctl_lookup(SYSCTLFN_CALL(&node));
1871: if (error || newp == NULL)
1872: return (error);
1873:
1874: if (new_somaxkva < (16 * 1024 * 1024)) /* sanity */
1875: return (EINVAL);
1876:
1.136 ad 1877: mutex_enter(&so_pendfree_lock);
1.94 yamt 1878: somaxkva = new_somaxkva;
1.136 ad 1879: cv_broadcast(&socurkva_cv);
1880: mutex_exit(&so_pendfree_lock);
1.94 yamt 1881:
1882: return (error);
1883: }
1884:
1885: SYSCTL_SETUP(sysctl_kern_somaxkva_setup, "sysctl kern.somaxkva setup")
1886: {
1887:
1.97 atatat 1888: sysctl_createv(clog, 0, NULL, NULL,
1889: CTLFLAG_PERMANENT,
1890: CTLTYPE_NODE, "kern", NULL,
1891: NULL, 0, NULL, 0,
1892: CTL_KERN, CTL_EOL);
1893:
1894: sysctl_createv(clog, 0, NULL, NULL,
1895: CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1.103 atatat 1896: CTLTYPE_INT, "somaxkva",
1897: SYSCTL_DESCR("Maximum amount of kernel memory to be "
1898: "used for socket buffers"),
1.94 yamt 1899: sysctl_kern_somaxkva, 0, NULL, 0,
1900: CTL_KERN, KERN_SOMAXKVA, CTL_EOL);
1901: }
CVSweb <webmaster@jp.NetBSD.org>