Annotation of src/sys/kern/uipc_socket.c, Revision 1.77
1.77 ! thorpej 1: /* $NetBSD: uipc_socket.c,v 1.76 2003/01/31 05:00:24 thorpej Exp $ */
1.64 thorpej 2:
3: /*-
4: * Copyright (c) 2002 The NetBSD Foundation, Inc.
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.
51: * 3. All advertising materials mentioning features or use of this software
52: * must display the following acknowledgement:
53: * This product includes software developed by the University of
54: * California, Berkeley and its contributors.
55: * 4. Neither the name of the University nor the names of its contributors
56: * may be used to endorse or promote products derived from this software
57: * without specific prior written permission.
58: *
59: * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
60: * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
61: * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
62: * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
63: * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
64: * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
65: * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
66: * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
67: * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
68: * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
69: * SUCH DAMAGE.
70: *
1.32 fvdl 71: * @(#)uipc_socket.c 8.6 (Berkeley) 5/2/95
1.1 cgd 72: */
1.59 lukem 73:
74: #include <sys/cdefs.h>
1.77 ! thorpej 75: __KERNEL_RCSID(0, "$NetBSD: uipc_socket.c,v 1.76 2003/01/31 05:00:24 thorpej Exp $");
1.64 thorpej 76:
77: #include "opt_sock_counters.h"
78: #include "opt_sosend_loan.h"
1.1 cgd 79:
1.9 mycroft 80: #include <sys/param.h>
81: #include <sys/systm.h>
82: #include <sys/proc.h>
83: #include <sys/file.h>
84: #include <sys/malloc.h>
85: #include <sys/mbuf.h>
86: #include <sys/domain.h>
87: #include <sys/kernel.h>
88: #include <sys/protosw.h>
89: #include <sys/socket.h>
90: #include <sys/socketvar.h>
1.21 christos 91: #include <sys/signalvar.h>
1.9 mycroft 92: #include <sys/resourcevar.h>
1.37 thorpej 93: #include <sys/pool.h>
1.72 jdolecek 94: #include <sys/event.h>
1.37 thorpej 95:
1.64 thorpej 96: #include <uvm/uvm.h>
97:
1.54 lukem 98: struct pool socket_pool;
1.77 ! thorpej 99:
! 100: MALLOC_DEFINE(M_SOOPTS, "soopts", "socket options");
! 101: MALLOC_DEFINE(M_SONAME, "soname", "socket name");
1.37 thorpej 102:
1.54 lukem 103: extern int somaxconn; /* patchable (XXX sysctl) */
104: int somaxconn = SOMAXCONN;
1.49 jonathan 105:
1.64 thorpej 106: #ifdef SOSEND_COUNTERS
107: #include <sys/device.h>
108:
109: struct evcnt sosend_loan_big = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
110: NULL, "sosend", "loan big");
111: struct evcnt sosend_copy_big = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
112: NULL, "sosend", "copy big");
113: struct evcnt sosend_copy_small = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
114: NULL, "sosend", "copy small");
115: struct evcnt sosend_kvalimit = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
116: NULL, "sosend", "kva limit");
117:
118: #define SOSEND_COUNTER_INCR(ev) (ev)->ev_count++
119:
120: #else
121:
122: #define SOSEND_COUNTER_INCR(ev) /* nothing */
123:
124: #endif /* SOSEND_COUNTERS */
125:
1.37 thorpej 126: void
1.54 lukem 127: soinit(void)
1.37 thorpej 128: {
129:
130: pool_init(&socket_pool, sizeof(struct socket), 0, 0, 0,
1.62 thorpej 131: "sockpl", NULL);
1.64 thorpej 132:
133: #ifdef SOSEND_COUNTERS
134: evcnt_attach_static(&sosend_loan_big);
135: evcnt_attach_static(&sosend_copy_big);
136: evcnt_attach_static(&sosend_copy_small);
137: evcnt_attach_static(&sosend_kvalimit);
138: #endif /* SOSEND_COUNTERS */
1.37 thorpej 139: }
1.1 cgd 140:
1.71 thorpej 141: #ifdef SOSEND_NO_LOAN
142: int use_sosend_loan = 0;
143: #else
1.65 thorpej 144: int use_sosend_loan = 1;
145: #endif
1.64 thorpej 146:
147: struct mbuf *so_pendfree;
148:
149: int somaxkva = 16 * 1024 * 1024;
150: int socurkva;
151: int sokvawaiters;
152:
153: #define SOCK_LOAN_THRESH 4096
154: #define SOCK_LOAN_CHUNK 65536
155:
156: static void
1.76 thorpej 157: sodoloanfree(caddr_t buf, size_t size)
1.64 thorpej 158: {
159: struct vm_page **pgs;
160: vaddr_t va, sva, eva;
161: vsize_t len;
162: paddr_t pa;
163: int i, npgs;
164:
165: eva = round_page((vaddr_t) buf + size);
166: sva = trunc_page((vaddr_t) buf);
167: len = eva - sva;
168: npgs = len >> PAGE_SHIFT;
169:
170: pgs = alloca(npgs * sizeof(*pgs));
171:
172: for (i = 0, va = sva; va < eva; i++, va += PAGE_SIZE) {
173: if (pmap_extract(pmap_kernel(), va, &pa) == FALSE)
174: panic("sodoloanfree: va 0x%lx not mapped", va);
175: pgs[i] = PHYS_TO_VM_PAGE(pa);
176: }
177:
178: pmap_kremove(sva, len);
179: pmap_update(pmap_kernel());
180: uvm_unloan(pgs, npgs, UVM_LOAN_TOPAGE);
181: uvm_km_free(kernel_map, sva, len);
182: socurkva -= len;
183: if (sokvawaiters)
184: wakeup(&socurkva);
185: }
186:
187: static size_t
188: sodopendfree(struct socket *so)
189: {
190: struct mbuf *m;
191: size_t rv = 0;
192: int s;
193:
194: s = splvm();
195:
196: for (;;) {
197: m = so_pendfree;
198: if (m == NULL)
199: break;
200: so_pendfree = m->m_next;
201: splx(s);
202:
203: rv += m->m_ext.ext_size;
204: sodoloanfree(m->m_ext.ext_buf, m->m_ext.ext_size);
205: s = splvm();
206: pool_cache_put(&mbpool_cache, m);
207: }
208:
209: for (;;) {
210: m = so->so_pendfree;
211: if (m == NULL)
212: break;
213: so->so_pendfree = m->m_next;
214: splx(s);
215:
216: rv += m->m_ext.ext_size;
217: sodoloanfree(m->m_ext.ext_buf, m->m_ext.ext_size);
218: s = splvm();
219: pool_cache_put(&mbpool_cache, m);
220: }
221:
222: splx(s);
223: return (rv);
224: }
225:
226: static void
1.76 thorpej 227: soloanfree(struct mbuf *m, caddr_t buf, size_t size, void *arg)
1.64 thorpej 228: {
229: struct socket *so = arg;
230: int s;
231:
232: if (m == NULL) {
233: sodoloanfree(buf, size);
234: return;
235: }
236:
237: s = splvm();
238: m->m_next = so->so_pendfree;
239: so->so_pendfree = m;
240: splx(s);
241: if (sokvawaiters)
242: wakeup(&socurkva);
243: }
244:
245: static long
246: sosend_loan(struct socket *so, struct uio *uio, struct mbuf *m, long space)
247: {
248: struct iovec *iov = uio->uio_iov;
249: vaddr_t sva, eva;
250: vsize_t len;
251: struct vm_page **pgs;
252: vaddr_t lva, va;
253: int npgs, s, i, error;
254:
255: if (uio->uio_segflg != UIO_USERSPACE)
256: return (0);
257:
258: if (iov->iov_len < (size_t) space)
259: space = iov->iov_len;
260: if (space > SOCK_LOAN_CHUNK)
261: space = SOCK_LOAN_CHUNK;
262:
263: eva = round_page((vaddr_t) iov->iov_base + space);
264: sva = trunc_page((vaddr_t) iov->iov_base);
265: len = eva - sva;
266: npgs = len >> PAGE_SHIFT;
267:
268: while (socurkva + len > somaxkva) {
269: if (sodopendfree(so))
270: continue;
271: SOSEND_COUNTER_INCR(&sosend_kvalimit);
272: s = splvm();
273: sokvawaiters++;
274: (void) tsleep(&socurkva, PVM, "sokva", 0);
275: sokvawaiters--;
276: splx(s);
277: }
278:
279: lva = uvm_km_valloc_wait(kernel_map, len);
280: if (lva == 0)
281: return (0);
282: socurkva += len;
283:
284: pgs = alloca(npgs * sizeof(*pgs));
285:
286: error = uvm_loan(&uio->uio_procp->p_vmspace->vm_map, sva, len,
287: pgs, UVM_LOAN_TOPAGE);
288: if (error) {
289: uvm_km_free(kernel_map, lva, len);
290: socurkva -= len;
291: return (0);
292: }
293:
294: for (i = 0, va = lva; i < npgs; i++, va += PAGE_SIZE)
295: pmap_kenter_pa(va, VM_PAGE_TO_PHYS(pgs[i]), VM_PROT_READ);
296: pmap_update(pmap_kernel());
297:
298: lva += (vaddr_t) iov->iov_base & PAGE_MASK;
299:
300: MEXTADD(m, (caddr_t) lva, space, M_MBUF, soloanfree, so);
301:
302: uio->uio_resid -= space;
303: /* uio_offset not updated, not set/used for write(2) */
304: uio->uio_iov->iov_base = (caddr_t) uio->uio_iov->iov_base + space;
305: uio->uio_iov->iov_len -= space;
306: if (uio->uio_iov->iov_len == 0) {
307: uio->uio_iov++;
308: uio->uio_iovcnt--;
309: }
310:
311: return (space);
312: }
313:
1.1 cgd 314: /*
315: * Socket operation routines.
316: * These routines are called by the routines in
317: * sys_socket.c or from a system process, and
318: * implement the semantics of socket operations by
319: * switching out to the protocol specific routines.
320: */
321: /*ARGSUSED*/
1.3 andrew 322: int
1.54 lukem 323: socreate(int dom, struct socket **aso, int type, int proto)
1.1 cgd 324: {
1.54 lukem 325: struct proc *p;
326: struct protosw *prp;
327: struct socket *so;
328: int error, s;
1.1 cgd 329:
1.54 lukem 330: p = curproc; /* XXX */
1.1 cgd 331: if (proto)
332: prp = pffindproto(dom, proto, type);
333: else
334: prp = pffindtype(dom, type);
1.15 mycroft 335: if (prp == 0 || prp->pr_usrreq == 0)
1.1 cgd 336: return (EPROTONOSUPPORT);
337: if (prp->pr_type != type)
338: return (EPROTOTYPE);
1.39 matt 339: s = splsoftnet();
1.37 thorpej 340: so = pool_get(&socket_pool, PR_WAITOK);
1.38 perry 341: memset((caddr_t)so, 0, sizeof(*so));
1.31 thorpej 342: TAILQ_INIT(&so->so_q0);
343: TAILQ_INIT(&so->so_q);
1.1 cgd 344: so->so_type = type;
345: so->so_proto = prp;
1.33 matt 346: so->so_send = sosend;
347: so->so_receive = soreceive;
1.44 lukem 348: if (p != 0)
349: so->so_uid = p->p_ucred->cr_uid;
1.22 mycroft 350: error = (*prp->pr_usrreq)(so, PRU_ATTACH, (struct mbuf *)0,
351: (struct mbuf *)(long)proto, (struct mbuf *)0, p);
1.1 cgd 352: if (error) {
353: so->so_state |= SS_NOFDREF;
354: sofree(so);
1.39 matt 355: splx(s);
1.1 cgd 356: return (error);
357: }
1.39 matt 358: splx(s);
1.1 cgd 359: *aso = so;
360: return (0);
361: }
362:
1.3 andrew 363: int
1.54 lukem 364: sobind(struct socket *so, struct mbuf *nam, struct proc *p)
1.1 cgd 365: {
1.54 lukem 366: int s, error;
1.1 cgd 367:
1.54 lukem 368: s = splsoftnet();
1.22 mycroft 369: error = (*so->so_proto->pr_usrreq)(so, PRU_BIND, (struct mbuf *)0,
370: nam, (struct mbuf *)0, p);
1.1 cgd 371: splx(s);
372: return (error);
373: }
374:
1.3 andrew 375: int
1.54 lukem 376: solisten(struct socket *so, int backlog)
1.1 cgd 377: {
1.54 lukem 378: int s, error;
1.1 cgd 379:
1.54 lukem 380: s = splsoftnet();
1.22 mycroft 381: error = (*so->so_proto->pr_usrreq)(so, PRU_LISTEN, (struct mbuf *)0,
382: (struct mbuf *)0, (struct mbuf *)0, (struct proc *)0);
1.1 cgd 383: if (error) {
384: splx(s);
385: return (error);
386: }
1.63 matt 387: if (TAILQ_EMPTY(&so->so_q))
1.1 cgd 388: so->so_options |= SO_ACCEPTCONN;
389: if (backlog < 0)
390: backlog = 0;
1.49 jonathan 391: so->so_qlimit = min(backlog, somaxconn);
1.1 cgd 392: splx(s);
393: return (0);
394: }
395:
1.21 christos 396: void
1.54 lukem 397: sofree(struct socket *so)
1.1 cgd 398: {
1.64 thorpej 399: struct mbuf *m;
1.1 cgd 400:
1.43 mycroft 401: if (so->so_pcb || (so->so_state & SS_NOFDREF) == 0)
1.1 cgd 402: return;
1.43 mycroft 403: if (so->so_head) {
404: /*
405: * We must not decommission a socket that's on the accept(2)
406: * queue. If we do, then accept(2) may hang after select(2)
407: * indicated that the listening socket was ready.
408: */
409: if (!soqremque(so, 0))
410: return;
411: }
1.1 cgd 412: sbrelease(&so->so_snd);
413: sorflush(so);
1.64 thorpej 414: while ((m = so->so_pendfree) != NULL) {
415: so->so_pendfree = m->m_next;
416: m->m_next = so_pendfree;
417: so_pendfree = m;
418: }
1.37 thorpej 419: pool_put(&socket_pool, so);
1.1 cgd 420: }
421:
422: /*
423: * Close a socket on last file table reference removal.
424: * Initiate disconnect if connected.
425: * Free socket when disconnect complete.
426: */
1.3 andrew 427: int
1.54 lukem 428: soclose(struct socket *so)
1.1 cgd 429: {
1.54 lukem 430: struct socket *so2;
431: int s, error;
1.1 cgd 432:
1.54 lukem 433: error = 0;
434: s = splsoftnet(); /* conservative */
1.1 cgd 435: if (so->so_options & SO_ACCEPTCONN) {
1.63 matt 436: while ((so2 = TAILQ_FIRST(&so->so_q0)) != 0) {
1.42 mycroft 437: (void) soqremque(so2, 0);
1.41 mycroft 438: (void) soabort(so2);
439: }
1.63 matt 440: while ((so2 = TAILQ_FIRST(&so->so_q)) != 0) {
1.42 mycroft 441: (void) soqremque(so2, 1);
1.41 mycroft 442: (void) soabort(so2);
443: }
1.1 cgd 444: }
445: if (so->so_pcb == 0)
446: goto discard;
447: if (so->so_state & SS_ISCONNECTED) {
448: if ((so->so_state & SS_ISDISCONNECTING) == 0) {
449: error = sodisconnect(so);
450: if (error)
451: goto drop;
452: }
453: if (so->so_options & SO_LINGER) {
454: if ((so->so_state & SS_ISDISCONNECTING) &&
455: (so->so_state & SS_NBIO))
456: goto drop;
1.21 christos 457: while (so->so_state & SS_ISCONNECTED) {
458: error = tsleep((caddr_t)&so->so_timeo,
459: PSOCK | PCATCH, netcls,
1.30 thorpej 460: so->so_linger * hz);
1.21 christos 461: if (error)
1.1 cgd 462: break;
1.21 christos 463: }
1.1 cgd 464: }
465: }
1.54 lukem 466: drop:
1.1 cgd 467: if (so->so_pcb) {
1.22 mycroft 468: int error2 = (*so->so_proto->pr_usrreq)(so, PRU_DETACH,
469: (struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0,
470: (struct proc *)0);
1.1 cgd 471: if (error == 0)
472: error = error2;
473: }
1.54 lukem 474: discard:
1.1 cgd 475: if (so->so_state & SS_NOFDREF)
476: panic("soclose: NOFDREF");
477: so->so_state |= SS_NOFDREF;
478: sofree(so);
479: splx(s);
480: return (error);
481: }
482:
483: /*
1.20 mycroft 484: * Must be called at splsoftnet...
1.1 cgd 485: */
1.3 andrew 486: int
1.54 lukem 487: soabort(struct socket *so)
1.1 cgd 488: {
489:
1.22 mycroft 490: return (*so->so_proto->pr_usrreq)(so, PRU_ABORT, (struct mbuf *)0,
491: (struct mbuf *)0, (struct mbuf *)0, (struct proc *)0);
1.1 cgd 492: }
493:
1.3 andrew 494: int
1.54 lukem 495: soaccept(struct socket *so, struct mbuf *nam)
1.1 cgd 496: {
1.54 lukem 497: int s, error;
1.1 cgd 498:
1.54 lukem 499: error = 0;
500: s = splsoftnet();
1.1 cgd 501: if ((so->so_state & SS_NOFDREF) == 0)
502: panic("soaccept: !NOFDREF");
503: so->so_state &= ~SS_NOFDREF;
1.55 thorpej 504: if ((so->so_state & SS_ISDISCONNECTED) == 0 ||
505: (so->so_proto->pr_flags & PR_ABRTACPTDIS) == 0)
1.41 mycroft 506: error = (*so->so_proto->pr_usrreq)(so, PRU_ACCEPT,
507: (struct mbuf *)0, nam, (struct mbuf *)0, (struct proc *)0);
508: else
1.53 itojun 509: error = ECONNABORTED;
1.52 itojun 510:
1.1 cgd 511: splx(s);
512: return (error);
513: }
514:
1.3 andrew 515: int
1.54 lukem 516: soconnect(struct socket *so, struct mbuf *nam)
1.1 cgd 517: {
1.54 lukem 518: struct proc *p;
519: int s, error;
1.1 cgd 520:
1.54 lukem 521: p = curproc; /* XXX */
1.1 cgd 522: if (so->so_options & SO_ACCEPTCONN)
523: return (EOPNOTSUPP);
1.20 mycroft 524: s = splsoftnet();
1.1 cgd 525: /*
526: * If protocol is connection-based, can only connect once.
527: * Otherwise, if connected, try to disconnect first.
528: * This allows user to disconnect by connecting to, e.g.,
529: * a null address.
530: */
531: if (so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING) &&
532: ((so->so_proto->pr_flags & PR_CONNREQUIRED) ||
533: (error = sodisconnect(so))))
534: error = EISCONN;
535: else
536: error = (*so->so_proto->pr_usrreq)(so, PRU_CONNECT,
1.23 mycroft 537: (struct mbuf *)0, nam, (struct mbuf *)0, p);
1.1 cgd 538: splx(s);
539: return (error);
540: }
541:
1.3 andrew 542: int
1.54 lukem 543: soconnect2(struct socket *so1, struct socket *so2)
1.1 cgd 544: {
1.54 lukem 545: int s, error;
1.1 cgd 546:
1.54 lukem 547: s = splsoftnet();
1.22 mycroft 548: error = (*so1->so_proto->pr_usrreq)(so1, PRU_CONNECT2,
549: (struct mbuf *)0, (struct mbuf *)so2, (struct mbuf *)0,
550: (struct proc *)0);
1.1 cgd 551: splx(s);
552: return (error);
553: }
554:
1.3 andrew 555: int
1.54 lukem 556: sodisconnect(struct socket *so)
1.1 cgd 557: {
1.54 lukem 558: int s, error;
1.1 cgd 559:
1.54 lukem 560: s = splsoftnet();
1.1 cgd 561: if ((so->so_state & SS_ISCONNECTED) == 0) {
562: error = ENOTCONN;
563: goto bad;
564: }
565: if (so->so_state & SS_ISDISCONNECTING) {
566: error = EALREADY;
567: goto bad;
568: }
1.22 mycroft 569: error = (*so->so_proto->pr_usrreq)(so, PRU_DISCONNECT,
570: (struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0,
571: (struct proc *)0);
1.54 lukem 572: bad:
1.1 cgd 573: splx(s);
1.64 thorpej 574: sodopendfree(so);
1.1 cgd 575: return (error);
576: }
577:
1.15 mycroft 578: #define SBLOCKWAIT(f) (((f) & MSG_DONTWAIT) ? M_NOWAIT : M_WAITOK)
1.1 cgd 579: /*
580: * Send on a socket.
581: * If send must go all at once and message is larger than
582: * send buffering, then hard error.
583: * Lock against other senders.
584: * If must go all at once and not enough room now, then
585: * inform user that this would block and do nothing.
586: * Otherwise, if nonblocking, send as much as possible.
587: * The data to be sent is described by "uio" if nonzero,
588: * otherwise by the mbuf chain "top" (which must be null
589: * if uio is not). Data provided in mbuf chain must be small
590: * enough to send all at once.
591: *
592: * Returns nonzero on error, timeout or signal; callers
593: * must check for short counts if EINTR/ERESTART are returned.
594: * Data and control buffers are freed on return.
595: */
1.3 andrew 596: int
1.54 lukem 597: sosend(struct socket *so, struct mbuf *addr, struct uio *uio, struct mbuf *top,
598: struct mbuf *control, int flags)
1.1 cgd 599: {
1.54 lukem 600: struct proc *p;
601: struct mbuf **mp, *m;
1.58 jdolecek 602: long space, len, resid, clen, mlen;
603: int error, s, dontroute, atomic;
1.54 lukem 604:
1.64 thorpej 605: sodopendfree(so);
606:
1.54 lukem 607: p = curproc; /* XXX */
608: clen = 0;
609: atomic = sosendallatonce(so) || top;
1.1 cgd 610: if (uio)
611: resid = uio->uio_resid;
612: else
613: resid = top->m_pkthdr.len;
1.7 cgd 614: /*
615: * In theory resid should be unsigned.
616: * However, space must be signed, as it might be less than 0
617: * if we over-committed, and we must use a signed comparison
618: * of space and resid. On the other hand, a negative resid
619: * causes us to loop sending 0-length segments to the protocol.
620: */
1.29 mycroft 621: if (resid < 0) {
622: error = EINVAL;
623: goto out;
624: }
1.1 cgd 625: dontroute =
626: (flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 &&
627: (so->so_proto->pr_flags & PR_ATOMIC);
1.12 mycroft 628: p->p_stats->p_ru.ru_msgsnd++;
1.1 cgd 629: if (control)
630: clen = control->m_len;
631: #define snderr(errno) { error = errno; splx(s); goto release; }
632:
1.54 lukem 633: restart:
1.21 christos 634: if ((error = sblock(&so->so_snd, SBLOCKWAIT(flags))) != 0)
1.1 cgd 635: goto out;
636: do {
1.20 mycroft 637: s = splsoftnet();
1.1 cgd 638: if (so->so_state & SS_CANTSENDMORE)
639: snderr(EPIPE);
1.48 thorpej 640: if (so->so_error) {
641: error = so->so_error;
642: so->so_error = 0;
643: splx(s);
644: goto release;
645: }
1.1 cgd 646: if ((so->so_state & SS_ISCONNECTED) == 0) {
647: if (so->so_proto->pr_flags & PR_CONNREQUIRED) {
648: if ((so->so_state & SS_ISCONFIRMING) == 0 &&
649: !(resid == 0 && clen != 0))
650: snderr(ENOTCONN);
651: } else if (addr == 0)
652: snderr(EDESTADDRREQ);
653: }
654: space = sbspace(&so->so_snd);
655: if (flags & MSG_OOB)
656: space += 1024;
1.21 christos 657: if ((atomic && resid > so->so_snd.sb_hiwat) ||
1.11 mycroft 658: clen > so->so_snd.sb_hiwat)
659: snderr(EMSGSIZE);
660: if (space < resid + clen && uio &&
1.1 cgd 661: (atomic || space < so->so_snd.sb_lowat || space < clen)) {
662: if (so->so_state & SS_NBIO)
663: snderr(EWOULDBLOCK);
664: sbunlock(&so->so_snd);
665: error = sbwait(&so->so_snd);
666: splx(s);
667: if (error)
668: goto out;
669: goto restart;
670: }
671: splx(s);
672: mp = ⊤
673: space -= clen;
674: do {
1.45 tv 675: if (uio == NULL) {
676: /*
677: * Data is prepackaged in "top".
678: */
679: resid = 0;
680: if (flags & MSG_EOR)
681: top->m_flags |= M_EOR;
682: } else do {
683: if (top == 0) {
684: MGETHDR(m, M_WAIT, MT_DATA);
685: mlen = MHLEN;
686: m->m_pkthdr.len = 0;
687: m->m_pkthdr.rcvif = (struct ifnet *)0;
688: } else {
689: MGET(m, M_WAIT, MT_DATA);
690: mlen = MLEN;
691: }
1.65 thorpej 692: if (use_sosend_loan &&
693: uio->uio_iov->iov_len >= SOCK_LOAN_THRESH &&
1.64 thorpej 694: space >= SOCK_LOAN_THRESH &&
695: (len = sosend_loan(so, uio, m,
696: space)) != 0) {
697: SOSEND_COUNTER_INCR(&sosend_loan_big);
698: space -= len;
699: goto have_data;
700: }
1.45 tv 701: if (resid >= MINCLSIZE && space >= MCLBYTES) {
1.64 thorpej 702: SOSEND_COUNTER_INCR(&sosend_copy_big);
1.45 tv 703: MCLGET(m, M_WAIT);
704: if ((m->m_flags & M_EXT) == 0)
705: goto nopages;
706: mlen = MCLBYTES;
707: if (atomic && top == 0) {
1.58 jdolecek 708: len = lmin(MCLBYTES - max_hdr,
1.54 lukem 709: resid);
1.45 tv 710: m->m_data += max_hdr;
711: } else
1.58 jdolecek 712: len = lmin(MCLBYTES, resid);
1.45 tv 713: space -= len;
714: } else {
1.64 thorpej 715: nopages:
716: SOSEND_COUNTER_INCR(&sosend_copy_small);
1.58 jdolecek 717: len = lmin(lmin(mlen, resid), space);
1.45 tv 718: space -= len;
719: /*
720: * For datagram protocols, leave room
721: * for protocol headers in first mbuf.
722: */
723: if (atomic && top == 0 && len < mlen)
724: MH_ALIGN(m, len);
725: }
1.54 lukem 726: error = uiomove(mtod(m, caddr_t), (int)len,
727: uio);
1.64 thorpej 728: have_data:
1.45 tv 729: resid = uio->uio_resid;
730: m->m_len = len;
731: *mp = m;
732: top->m_pkthdr.len += len;
733: if (error)
734: goto release;
735: mp = &m->m_next;
736: if (resid <= 0) {
737: if (flags & MSG_EOR)
738: top->m_flags |= M_EOR;
739: break;
740: }
741: } while (space > 0 && atomic);
1.46 sommerfe 742:
743: s = splsoftnet();
744:
745: if (so->so_state & SS_CANTSENDMORE)
746: snderr(EPIPE);
1.45 tv 747:
748: if (dontroute)
749: so->so_options |= SO_DONTROUTE;
750: if (resid > 0)
751: so->so_state |= SS_MORETOCOME;
1.46 sommerfe 752: error = (*so->so_proto->pr_usrreq)(so,
753: (flags & MSG_OOB) ? PRU_SENDOOB : PRU_SEND,
754: top, addr, control, p);
1.45 tv 755: if (dontroute)
756: so->so_options &= ~SO_DONTROUTE;
757: if (resid > 0)
758: so->so_state &= ~SS_MORETOCOME;
1.46 sommerfe 759: splx(s);
760:
1.45 tv 761: clen = 0;
762: control = 0;
763: top = 0;
764: mp = ⊤
1.1 cgd 765: if (error)
766: goto release;
767: } while (resid && space > 0);
768: } while (resid);
769:
1.54 lukem 770: release:
1.1 cgd 771: sbunlock(&so->so_snd);
1.54 lukem 772: out:
1.1 cgd 773: if (top)
774: m_freem(top);
775: if (control)
776: m_freem(control);
777: return (error);
778: }
779:
780: /*
781: * Implement receive operations on a socket.
782: * We depend on the way that records are added to the sockbuf
783: * by sbappend*. In particular, each record (mbufs linked through m_next)
784: * must begin with an address if the protocol so specifies,
785: * followed by an optional mbuf or mbufs containing ancillary data,
786: * and then zero or more mbufs of data.
787: * In order to avoid blocking network interrupts for the entire time here,
788: * we splx() while doing the actual copy to user space.
789: * Although the sockbuf is locked, new data may still be appended,
790: * and thus we must maintain consistency of the sockbuf during that time.
791: *
792: * The caller may receive the data as a single mbuf chain by supplying
793: * an mbuf **mp0 for use in returning the chain. The uio is then used
794: * only for the count in uio_resid.
795: */
1.3 andrew 796: int
1.54 lukem 797: soreceive(struct socket *so, struct mbuf **paddr, struct uio *uio,
798: struct mbuf **mp0, struct mbuf **controlp, int *flagsp)
1.1 cgd 799: {
1.54 lukem 800: struct mbuf *m, **mp;
801: int flags, len, error, s, offset, moff, type, orig_resid;
802: struct protosw *pr;
803: struct mbuf *nextrecord;
1.67 he 804: int mbuf_removed = 0;
1.64 thorpej 805:
1.54 lukem 806: pr = so->so_proto;
1.1 cgd 807: mp = mp0;
1.54 lukem 808: type = 0;
809: orig_resid = uio->uio_resid;
1.1 cgd 810: if (paddr)
811: *paddr = 0;
812: if (controlp)
813: *controlp = 0;
814: if (flagsp)
815: flags = *flagsp &~ MSG_EOR;
816: else
817: flags = 0;
1.66 enami 818:
819: if ((flags & MSG_DONTWAIT) == 0)
820: sodopendfree(so);
821:
1.1 cgd 822: if (flags & MSG_OOB) {
823: m = m_get(M_WAIT, MT_DATA);
1.17 cgd 824: error = (*pr->pr_usrreq)(so, PRU_RCVOOB, m,
1.22 mycroft 825: (struct mbuf *)(long)(flags & MSG_PEEK), (struct mbuf *)0,
826: (struct proc *)0);
1.1 cgd 827: if (error)
828: goto bad;
829: do {
830: error = uiomove(mtod(m, caddr_t),
831: (int) min(uio->uio_resid, m->m_len), uio);
832: m = m_free(m);
833: } while (uio->uio_resid && error == 0 && m);
1.54 lukem 834: bad:
1.1 cgd 835: if (m)
836: m_freem(m);
837: return (error);
838: }
839: if (mp)
840: *mp = (struct mbuf *)0;
841: if (so->so_state & SS_ISCONFIRMING && uio->uio_resid)
1.22 mycroft 842: (*pr->pr_usrreq)(so, PRU_RCVD, (struct mbuf *)0,
843: (struct mbuf *)0, (struct mbuf *)0, (struct proc *)0);
1.1 cgd 844:
1.54 lukem 845: restart:
1.21 christos 846: if ((error = sblock(&so->so_rcv, SBLOCKWAIT(flags))) != 0)
1.1 cgd 847: return (error);
1.20 mycroft 848: s = splsoftnet();
1.1 cgd 849:
850: m = so->so_rcv.sb_mb;
851: /*
852: * If we have less data than requested, block awaiting more
853: * (subject to any timeout) if:
1.15 mycroft 854: * 1. the current count is less than the low water mark,
1.1 cgd 855: * 2. MSG_WAITALL is set, and it is possible to do the entire
1.15 mycroft 856: * receive operation at once if we block (resid <= hiwat), or
857: * 3. MSG_DONTWAIT is not set.
1.1 cgd 858: * If MSG_WAITALL is set but resid is larger than the receive buffer,
859: * we have to do the receive in sections, and thus risk returning
860: * a short count if a timeout or signal occurs after we start.
861: */
1.21 christos 862: if (m == 0 || (((flags & MSG_DONTWAIT) == 0 &&
1.15 mycroft 863: so->so_rcv.sb_cc < uio->uio_resid) &&
1.1 cgd 864: (so->so_rcv.sb_cc < so->so_rcv.sb_lowat ||
865: ((flags & MSG_WAITALL) && uio->uio_resid <= so->so_rcv.sb_hiwat)) &&
1.21 christos 866: m->m_nextpkt == 0 && (pr->pr_flags & PR_ATOMIC) == 0)) {
1.1 cgd 867: #ifdef DIAGNOSTIC
868: if (m == 0 && so->so_rcv.sb_cc)
869: panic("receive 1");
870: #endif
871: if (so->so_error) {
872: if (m)
1.15 mycroft 873: goto dontblock;
1.1 cgd 874: error = so->so_error;
875: if ((flags & MSG_PEEK) == 0)
876: so->so_error = 0;
877: goto release;
878: }
879: if (so->so_state & SS_CANTRCVMORE) {
880: if (m)
1.15 mycroft 881: goto dontblock;
1.1 cgd 882: else
883: goto release;
884: }
885: for (; m; m = m->m_next)
886: if (m->m_type == MT_OOBDATA || (m->m_flags & M_EOR)) {
887: m = so->so_rcv.sb_mb;
888: goto dontblock;
889: }
890: if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0 &&
891: (so->so_proto->pr_flags & PR_CONNREQUIRED)) {
892: error = ENOTCONN;
893: goto release;
894: }
895: if (uio->uio_resid == 0)
896: goto release;
1.15 mycroft 897: if ((so->so_state & SS_NBIO) || (flags & MSG_DONTWAIT)) {
1.1 cgd 898: error = EWOULDBLOCK;
899: goto release;
900: }
1.69 thorpej 901: SBLASTRECORDCHK(&so->so_rcv, "soreceive sbwait 1");
902: SBLASTMBUFCHK(&so->so_rcv, "soreceive sbwait 1");
1.1 cgd 903: sbunlock(&so->so_rcv);
904: error = sbwait(&so->so_rcv);
905: splx(s);
906: if (error)
907: return (error);
908: goto restart;
909: }
1.54 lukem 910: dontblock:
1.69 thorpej 911: /*
912: * On entry here, m points to the first record of the socket buffer.
913: * While we process the initial mbufs containing address and control
914: * info, we save a copy of m->m_nextpkt into nextrecord.
915: */
1.15 mycroft 916: #ifdef notyet /* XXXX */
917: if (uio->uio_procp)
918: uio->uio_procp->p_stats->p_ru.ru_msgrcv++;
919: #endif
1.69 thorpej 920: KASSERT(m == so->so_rcv.sb_mb);
921: SBLASTRECORDCHK(&so->so_rcv, "soreceive 1");
922: SBLASTMBUFCHK(&so->so_rcv, "soreceive 1");
1.1 cgd 923: nextrecord = m->m_nextpkt;
924: if (pr->pr_flags & PR_ADDR) {
925: #ifdef DIAGNOSTIC
926: if (m->m_type != MT_SONAME)
927: panic("receive 1a");
928: #endif
1.3 andrew 929: orig_resid = 0;
1.1 cgd 930: if (flags & MSG_PEEK) {
931: if (paddr)
932: *paddr = m_copy(m, 0, m->m_len);
933: m = m->m_next;
934: } else {
935: sbfree(&so->so_rcv, m);
1.67 he 936: mbuf_removed = 1;
1.1 cgd 937: if (paddr) {
938: *paddr = m;
939: so->so_rcv.sb_mb = m->m_next;
940: m->m_next = 0;
941: m = so->so_rcv.sb_mb;
942: } else {
943: MFREE(m, so->so_rcv.sb_mb);
944: m = so->so_rcv.sb_mb;
945: }
946: }
947: }
948: while (m && m->m_type == MT_CONTROL && error == 0) {
949: if (flags & MSG_PEEK) {
950: if (controlp)
951: *controlp = m_copy(m, 0, m->m_len);
952: m = m->m_next;
953: } else {
954: sbfree(&so->so_rcv, m);
1.67 he 955: mbuf_removed = 1;
1.1 cgd 956: if (controlp) {
957: if (pr->pr_domain->dom_externalize &&
958: mtod(m, struct cmsghdr *)->cmsg_type ==
959: SCM_RIGHTS)
1.45 tv 960: error = (*pr->pr_domain->dom_externalize)(m);
1.1 cgd 961: *controlp = m;
962: so->so_rcv.sb_mb = m->m_next;
963: m->m_next = 0;
964: m = so->so_rcv.sb_mb;
965: } else {
966: MFREE(m, so->so_rcv.sb_mb);
967: m = so->so_rcv.sb_mb;
968: }
969: }
1.3 andrew 970: if (controlp) {
971: orig_resid = 0;
1.1 cgd 972: controlp = &(*controlp)->m_next;
1.3 andrew 973: }
1.1 cgd 974: }
1.69 thorpej 975:
976: /*
977: * If m is non-NULL, we have some data to read. From now on,
978: * make sure to keep sb_lastrecord consistent when working on
979: * the last packet on the chain (nextrecord == NULL) and we
980: * change m->m_nextpkt.
981: */
1.1 cgd 982: if (m) {
1.69 thorpej 983: if ((flags & MSG_PEEK) == 0) {
1.1 cgd 984: m->m_nextpkt = nextrecord;
1.69 thorpej 985: /*
986: * If nextrecord == NULL (this is a single chain),
987: * then sb_lastrecord may not be valid here if m
988: * was changed earlier.
989: */
990: if (nextrecord == NULL) {
991: KASSERT(so->so_rcv.sb_mb == m);
992: so->so_rcv.sb_lastrecord = m;
993: }
994: }
1.1 cgd 995: type = m->m_type;
996: if (type == MT_OOBDATA)
997: flags |= MSG_OOB;
1.69 thorpej 998: } else {
999: if ((flags & MSG_PEEK) == 0) {
1000: KASSERT(so->so_rcv.sb_mb == m);
1001: so->so_rcv.sb_mb = nextrecord;
1.70 thorpej 1002: SB_EMPTY_FIXUP(&so->so_rcv);
1.69 thorpej 1003: }
1.1 cgd 1004: }
1.69 thorpej 1005: SBLASTRECORDCHK(&so->so_rcv, "soreceive 2");
1006: SBLASTMBUFCHK(&so->so_rcv, "soreceive 2");
1007:
1.1 cgd 1008: moff = 0;
1009: offset = 0;
1010: while (m && uio->uio_resid > 0 && error == 0) {
1011: if (m->m_type == MT_OOBDATA) {
1012: if (type != MT_OOBDATA)
1013: break;
1014: } else if (type == MT_OOBDATA)
1015: break;
1016: #ifdef DIAGNOSTIC
1017: else if (m->m_type != MT_DATA && m->m_type != MT_HEADER)
1018: panic("receive 3");
1019: #endif
1020: so->so_state &= ~SS_RCVATMARK;
1021: len = uio->uio_resid;
1022: if (so->so_oobmark && len > so->so_oobmark - offset)
1023: len = so->so_oobmark - offset;
1024: if (len > m->m_len - moff)
1025: len = m->m_len - moff;
1026: /*
1027: * If mp is set, just pass back the mbufs.
1028: * Otherwise copy them out via the uio, then free.
1029: * Sockbuf must be consistent here (points to current mbuf,
1030: * it points to next record) when we drop priority;
1031: * we must note any additions to the sockbuf when we
1032: * block interrupts again.
1033: */
1034: if (mp == 0) {
1.69 thorpej 1035: SBLASTRECORDCHK(&so->so_rcv, "soreceive uiomove");
1036: SBLASTMBUFCHK(&so->so_rcv, "soreceive uiomove");
1.1 cgd 1037: splx(s);
1038: error = uiomove(mtod(m, caddr_t) + moff, (int)len, uio);
1.20 mycroft 1039: s = splsoftnet();
1.67 he 1040: if (error) {
1041: /*
1042: * If any part of the record has been removed
1043: * (such as the MT_SONAME mbuf, which will
1044: * happen when PR_ADDR, and thus also
1045: * PR_ATOMIC, is set), then drop the entire
1046: * record to maintain the atomicity of the
1047: * receive operation.
1048: *
1049: * This avoids a later panic("receive 1a")
1050: * when compiled with DIAGNOSTIC.
1051: */
1052: if (m && mbuf_removed
1053: && (pr->pr_flags & PR_ATOMIC))
1054: (void) sbdroprecord(&so->so_rcv);
1055:
1.57 jdolecek 1056: goto release;
1.67 he 1057: }
1.1 cgd 1058: } else
1059: uio->uio_resid -= len;
1060: if (len == m->m_len - moff) {
1061: if (m->m_flags & M_EOR)
1062: flags |= MSG_EOR;
1063: if (flags & MSG_PEEK) {
1064: m = m->m_next;
1065: moff = 0;
1066: } else {
1067: nextrecord = m->m_nextpkt;
1068: sbfree(&so->so_rcv, m);
1069: if (mp) {
1070: *mp = m;
1071: mp = &m->m_next;
1072: so->so_rcv.sb_mb = m = m->m_next;
1073: *mp = (struct mbuf *)0;
1074: } else {
1075: MFREE(m, so->so_rcv.sb_mb);
1076: m = so->so_rcv.sb_mb;
1077: }
1.69 thorpej 1078: /*
1079: * If m != NULL, we also know that
1080: * so->so_rcv.sb_mb != NULL.
1081: */
1082: KASSERT(so->so_rcv.sb_mb == m);
1083: if (m) {
1.1 cgd 1084: m->m_nextpkt = nextrecord;
1.69 thorpej 1085: if (nextrecord == NULL)
1086: so->so_rcv.sb_lastrecord = m;
1087: } else {
1088: so->so_rcv.sb_mb = nextrecord;
1.70 thorpej 1089: SB_EMPTY_FIXUP(&so->so_rcv);
1.69 thorpej 1090: }
1091: SBLASTRECORDCHK(&so->so_rcv, "soreceive 3");
1092: SBLASTMBUFCHK(&so->so_rcv, "soreceive 3");
1.1 cgd 1093: }
1094: } else {
1095: if (flags & MSG_PEEK)
1096: moff += len;
1097: else {
1098: if (mp)
1099: *mp = m_copym(m, 0, len, M_WAIT);
1100: m->m_data += len;
1101: m->m_len -= len;
1102: so->so_rcv.sb_cc -= len;
1103: }
1104: }
1105: if (so->so_oobmark) {
1106: if ((flags & MSG_PEEK) == 0) {
1107: so->so_oobmark -= len;
1108: if (so->so_oobmark == 0) {
1109: so->so_state |= SS_RCVATMARK;
1110: break;
1111: }
1.7 cgd 1112: } else {
1.1 cgd 1113: offset += len;
1.7 cgd 1114: if (offset == so->so_oobmark)
1115: break;
1116: }
1.1 cgd 1117: }
1118: if (flags & MSG_EOR)
1119: break;
1120: /*
1121: * If the MSG_WAITALL flag is set (for non-atomic socket),
1122: * we must not quit until "uio->uio_resid == 0" or an error
1123: * termination. If a signal/timeout occurs, return
1124: * with a short count but without error.
1125: * Keep sockbuf locked against other readers.
1126: */
1127: while (flags & MSG_WAITALL && m == 0 && uio->uio_resid > 0 &&
1.3 andrew 1128: !sosendallatonce(so) && !nextrecord) {
1.1 cgd 1129: if (so->so_error || so->so_state & SS_CANTRCVMORE)
1130: break;
1.68 matt 1131: /*
1132: * If we are peeking and the socket receive buffer is
1133: * full, stop since we can't get more data to peek at.
1134: */
1135: if ((flags & MSG_PEEK) && sbspace(&so->so_rcv) <= 0)
1136: break;
1137: /*
1138: * If we've drained the socket buffer, tell the
1139: * protocol in case it needs to do something to
1140: * get it filled again.
1141: */
1142: if ((pr->pr_flags & PR_WANTRCVD) && so->so_pcb)
1143: (*pr->pr_usrreq)(so, PRU_RCVD,
1144: (struct mbuf *)0,
1145: (struct mbuf *)(long)flags,
1146: (struct mbuf *)0,
1147: (struct proc *)0);
1.69 thorpej 1148: SBLASTRECORDCHK(&so->so_rcv, "soreceive sbwait 2");
1149: SBLASTMBUFCHK(&so->so_rcv, "soreceive sbwait 2");
1.1 cgd 1150: error = sbwait(&so->so_rcv);
1151: if (error) {
1152: sbunlock(&so->so_rcv);
1153: splx(s);
1154: return (0);
1155: }
1.21 christos 1156: if ((m = so->so_rcv.sb_mb) != NULL)
1.1 cgd 1157: nextrecord = m->m_nextpkt;
1158: }
1159: }
1.3 andrew 1160:
1161: if (m && pr->pr_flags & PR_ATOMIC) {
1162: flags |= MSG_TRUNC;
1163: if ((flags & MSG_PEEK) == 0)
1164: (void) sbdroprecord(&so->so_rcv);
1165: }
1.1 cgd 1166: if ((flags & MSG_PEEK) == 0) {
1.69 thorpej 1167: if (m == 0) {
1168: /*
1.70 thorpej 1169: * First part is an inline SB_EMPTY_FIXUP(). Second
1.69 thorpej 1170: * part makes sure sb_lastrecord is up-to-date if
1171: * there is still data in the socket buffer.
1172: */
1.1 cgd 1173: so->so_rcv.sb_mb = nextrecord;
1.69 thorpej 1174: if (so->so_rcv.sb_mb == NULL) {
1175: so->so_rcv.sb_mbtail = NULL;
1176: so->so_rcv.sb_lastrecord = NULL;
1177: } else if (nextrecord->m_nextpkt == NULL)
1178: so->so_rcv.sb_lastrecord = nextrecord;
1179: }
1180: SBLASTRECORDCHK(&so->so_rcv, "soreceive 4");
1181: SBLASTMBUFCHK(&so->so_rcv, "soreceive 4");
1.1 cgd 1182: if (pr->pr_flags & PR_WANTRCVD && so->so_pcb)
1.22 mycroft 1183: (*pr->pr_usrreq)(so, PRU_RCVD, (struct mbuf *)0,
1184: (struct mbuf *)(long)flags, (struct mbuf *)0,
1185: (struct proc *)0);
1.1 cgd 1186: }
1.3 andrew 1187: if (orig_resid == uio->uio_resid && orig_resid &&
1188: (flags & MSG_EOR) == 0 && (so->so_state & SS_CANTRCVMORE) == 0) {
1189: sbunlock(&so->so_rcv);
1190: splx(s);
1191: goto restart;
1192: }
1193:
1.1 cgd 1194: if (flagsp)
1195: *flagsp |= flags;
1.54 lukem 1196: release:
1.1 cgd 1197: sbunlock(&so->so_rcv);
1198: splx(s);
1199: return (error);
1200: }
1201:
1.14 mycroft 1202: int
1.54 lukem 1203: soshutdown(struct socket *so, int how)
1.1 cgd 1204: {
1.54 lukem 1205: struct protosw *pr;
1.34 kleink 1206:
1.54 lukem 1207: pr = so->so_proto;
1.34 kleink 1208: if (!(how == SHUT_RD || how == SHUT_WR || how == SHUT_RDWR))
1209: return (EINVAL);
1.1 cgd 1210:
1.34 kleink 1211: if (how == SHUT_RD || how == SHUT_RDWR)
1.1 cgd 1212: sorflush(so);
1.34 kleink 1213: if (how == SHUT_WR || how == SHUT_RDWR)
1.22 mycroft 1214: return (*pr->pr_usrreq)(so, PRU_SHUTDOWN, (struct mbuf *)0,
1215: (struct mbuf *)0, (struct mbuf *)0, (struct proc *)0);
1.1 cgd 1216: return (0);
1217: }
1218:
1.14 mycroft 1219: void
1.54 lukem 1220: sorflush(struct socket *so)
1.1 cgd 1221: {
1.54 lukem 1222: struct sockbuf *sb, asb;
1223: struct protosw *pr;
1224: int s;
1.1 cgd 1225:
1.54 lukem 1226: sb = &so->so_rcv;
1227: pr = so->so_proto;
1.1 cgd 1228: sb->sb_flags |= SB_NOINTR;
1.15 mycroft 1229: (void) sblock(sb, M_WAITOK);
1.56 thorpej 1230: s = splnet();
1.1 cgd 1231: socantrcvmore(so);
1232: sbunlock(sb);
1233: asb = *sb;
1.38 perry 1234: memset((caddr_t)sb, 0, sizeof(*sb));
1.1 cgd 1235: splx(s);
1236: if (pr->pr_flags & PR_RIGHTS && pr->pr_domain->dom_dispose)
1237: (*pr->pr_domain->dom_dispose)(asb.sb_mb);
1238: sbrelease(&asb);
1239: }
1240:
1.14 mycroft 1241: int
1.54 lukem 1242: sosetopt(struct socket *so, int level, int optname, struct mbuf *m0)
1.1 cgd 1243: {
1.54 lukem 1244: int error;
1245: struct mbuf *m;
1.1 cgd 1246:
1.54 lukem 1247: error = 0;
1248: m = m0;
1.1 cgd 1249: if (level != SOL_SOCKET) {
1250: if (so->so_proto && so->so_proto->pr_ctloutput)
1251: return ((*so->so_proto->pr_ctloutput)
1252: (PRCO_SETOPT, so, level, optname, &m0));
1253: error = ENOPROTOOPT;
1254: } else {
1255: switch (optname) {
1256:
1257: case SO_LINGER:
1.36 perry 1258: if (m == NULL || m->m_len != sizeof(struct linger)) {
1.1 cgd 1259: error = EINVAL;
1260: goto bad;
1261: }
1262: so->so_linger = mtod(m, struct linger *)->l_linger;
1263: /* fall thru... */
1264:
1265: case SO_DEBUG:
1266: case SO_KEEPALIVE:
1267: case SO_DONTROUTE:
1268: case SO_USELOOPBACK:
1269: case SO_BROADCAST:
1270: case SO_REUSEADDR:
1.15 mycroft 1271: case SO_REUSEPORT:
1.1 cgd 1272: case SO_OOBINLINE:
1.26 thorpej 1273: case SO_TIMESTAMP:
1.36 perry 1274: if (m == NULL || m->m_len < sizeof(int)) {
1.1 cgd 1275: error = EINVAL;
1276: goto bad;
1277: }
1278: if (*mtod(m, int *))
1279: so->so_options |= optname;
1280: else
1281: so->so_options &= ~optname;
1282: break;
1283:
1284: case SO_SNDBUF:
1285: case SO_RCVBUF:
1286: case SO_SNDLOWAT:
1287: case SO_RCVLOWAT:
1.28 thorpej 1288: {
1289: int optval;
1290:
1.36 perry 1291: if (m == NULL || m->m_len < sizeof(int)) {
1.1 cgd 1292: error = EINVAL;
1293: goto bad;
1294: }
1.28 thorpej 1295:
1296: /*
1297: * Values < 1 make no sense for any of these
1298: * options, so disallow them.
1299: */
1300: optval = *mtod(m, int *);
1301: if (optval < 1) {
1302: error = EINVAL;
1303: goto bad;
1304: }
1305:
1.1 cgd 1306: switch (optname) {
1307:
1308: case SO_SNDBUF:
1309: case SO_RCVBUF:
1310: if (sbreserve(optname == SO_SNDBUF ?
1311: &so->so_snd : &so->so_rcv,
1.28 thorpej 1312: (u_long) optval) == 0) {
1.1 cgd 1313: error = ENOBUFS;
1314: goto bad;
1315: }
1316: break;
1317:
1.28 thorpej 1318: /*
1319: * Make sure the low-water is never greater than
1320: * the high-water.
1321: */
1.1 cgd 1322: case SO_SNDLOWAT:
1.28 thorpej 1323: so->so_snd.sb_lowat =
1324: (optval > so->so_snd.sb_hiwat) ?
1325: so->so_snd.sb_hiwat : optval;
1.1 cgd 1326: break;
1327: case SO_RCVLOWAT:
1.28 thorpej 1328: so->so_rcv.sb_lowat =
1329: (optval > so->so_rcv.sb_hiwat) ?
1330: so->so_rcv.sb_hiwat : optval;
1.1 cgd 1331: break;
1332: }
1333: break;
1.28 thorpej 1334: }
1.1 cgd 1335:
1336: case SO_SNDTIMEO:
1337: case SO_RCVTIMEO:
1338: {
1339: struct timeval *tv;
1340: short val;
1341:
1.36 perry 1342: if (m == NULL || m->m_len < sizeof(*tv)) {
1.1 cgd 1343: error = EINVAL;
1344: goto bad;
1345: }
1346: tv = mtod(m, struct timeval *);
1.75 itojun 1347: if (tv->tv_sec > (SHRT_MAX - tv->tv_usec / tick) / hz) {
1.1 cgd 1348: error = EDOM;
1349: goto bad;
1350: }
1351: val = tv->tv_sec * hz + tv->tv_usec / tick;
1.74 itojun 1352: if (val == 0 && tv->tv_usec != 0)
1353: val = 1;
1.1 cgd 1354:
1355: switch (optname) {
1356:
1357: case SO_SNDTIMEO:
1358: so->so_snd.sb_timeo = val;
1359: break;
1360: case SO_RCVTIMEO:
1361: so->so_rcv.sb_timeo = val;
1362: break;
1363: }
1364: break;
1365: }
1366:
1367: default:
1368: error = ENOPROTOOPT;
1369: break;
1370: }
1.15 mycroft 1371: if (error == 0 && so->so_proto && so->so_proto->pr_ctloutput) {
1372: (void) ((*so->so_proto->pr_ctloutput)
1373: (PRCO_SETOPT, so, level, optname, &m0));
1374: m = NULL; /* freed by protocol */
1375: }
1.1 cgd 1376: }
1.54 lukem 1377: bad:
1.1 cgd 1378: if (m)
1379: (void) m_free(m);
1380: return (error);
1381: }
1382:
1.14 mycroft 1383: int
1.54 lukem 1384: sogetopt(struct socket *so, int level, int optname, struct mbuf **mp)
1.1 cgd 1385: {
1.54 lukem 1386: struct mbuf *m;
1.1 cgd 1387:
1388: if (level != SOL_SOCKET) {
1389: if (so->so_proto && so->so_proto->pr_ctloutput) {
1390: return ((*so->so_proto->pr_ctloutput)
1391: (PRCO_GETOPT, so, level, optname, mp));
1392: } else
1393: return (ENOPROTOOPT);
1394: } else {
1395: m = m_get(M_WAIT, MT_SOOPTS);
1.36 perry 1396: m->m_len = sizeof(int);
1.1 cgd 1397:
1398: switch (optname) {
1399:
1400: case SO_LINGER:
1.36 perry 1401: m->m_len = sizeof(struct linger);
1.1 cgd 1402: mtod(m, struct linger *)->l_onoff =
1403: so->so_options & SO_LINGER;
1404: mtod(m, struct linger *)->l_linger = so->so_linger;
1405: break;
1406:
1407: case SO_USELOOPBACK:
1408: case SO_DONTROUTE:
1409: case SO_DEBUG:
1410: case SO_KEEPALIVE:
1411: case SO_REUSEADDR:
1.15 mycroft 1412: case SO_REUSEPORT:
1.1 cgd 1413: case SO_BROADCAST:
1414: case SO_OOBINLINE:
1.26 thorpej 1415: case SO_TIMESTAMP:
1.1 cgd 1416: *mtod(m, int *) = so->so_options & optname;
1417: break;
1418:
1419: case SO_TYPE:
1420: *mtod(m, int *) = so->so_type;
1421: break;
1422:
1423: case SO_ERROR:
1424: *mtod(m, int *) = so->so_error;
1425: so->so_error = 0;
1426: break;
1427:
1428: case SO_SNDBUF:
1429: *mtod(m, int *) = so->so_snd.sb_hiwat;
1430: break;
1431:
1432: case SO_RCVBUF:
1433: *mtod(m, int *) = so->so_rcv.sb_hiwat;
1434: break;
1435:
1436: case SO_SNDLOWAT:
1437: *mtod(m, int *) = so->so_snd.sb_lowat;
1438: break;
1439:
1440: case SO_RCVLOWAT:
1441: *mtod(m, int *) = so->so_rcv.sb_lowat;
1442: break;
1443:
1444: case SO_SNDTIMEO:
1445: case SO_RCVTIMEO:
1446: {
1447: int val = (optname == SO_SNDTIMEO ?
1448: so->so_snd.sb_timeo : so->so_rcv.sb_timeo);
1449:
1450: m->m_len = sizeof(struct timeval);
1451: mtod(m, struct timeval *)->tv_sec = val / hz;
1452: mtod(m, struct timeval *)->tv_usec =
1.27 kleink 1453: (val % hz) * tick;
1.1 cgd 1454: break;
1455: }
1456:
1457: default:
1458: (void)m_free(m);
1459: return (ENOPROTOOPT);
1460: }
1461: *mp = m;
1462: return (0);
1463: }
1464: }
1465:
1.14 mycroft 1466: void
1.54 lukem 1467: sohasoutofband(struct socket *so)
1.1 cgd 1468: {
1469: struct proc *p;
1470:
1471: if (so->so_pgid < 0)
1472: gsignal(-so->so_pgid, SIGURG);
1473: else if (so->so_pgid > 0 && (p = pfind(so->so_pgid)) != 0)
1474: psignal(p, SIGURG);
1.2 cgd 1475: selwakeup(&so->so_rcv.sb_sel);
1.1 cgd 1476: }
1.72 jdolecek 1477:
1478: static void
1479: filt_sordetach(struct knote *kn)
1480: {
1481: struct socket *so;
1482:
1483: so = (struct socket *)kn->kn_fp->f_data;
1.73 christos 1484: SLIST_REMOVE(&so->so_rcv.sb_sel.sel_klist, kn, knote, kn_selnext);
1485: if (SLIST_EMPTY(&so->so_rcv.sb_sel.sel_klist))
1.72 jdolecek 1486: so->so_rcv.sb_flags &= ~SB_KNOTE;
1487: }
1488:
1489: /*ARGSUSED*/
1490: static int
1491: filt_soread(struct knote *kn, long hint)
1492: {
1493: struct socket *so;
1494:
1495: so = (struct socket *)kn->kn_fp->f_data;
1496: kn->kn_data = so->so_rcv.sb_cc;
1497: if (so->so_state & SS_CANTRCVMORE) {
1498: kn->kn_flags |= EV_EOF;
1499: kn->kn_fflags = so->so_error;
1500: return (1);
1501: }
1502: if (so->so_error) /* temporary udp error */
1503: return (1);
1504: if (kn->kn_sfflags & NOTE_LOWAT)
1505: return (kn->kn_data >= kn->kn_sdata);
1506: return (kn->kn_data >= so->so_rcv.sb_lowat);
1507: }
1508:
1509: static void
1510: filt_sowdetach(struct knote *kn)
1511: {
1512: struct socket *so;
1513:
1514: so = (struct socket *)kn->kn_fp->f_data;
1.73 christos 1515: SLIST_REMOVE(&so->so_snd.sb_sel.sel_klist, kn, knote, kn_selnext);
1516: if (SLIST_EMPTY(&so->so_snd.sb_sel.sel_klist))
1.72 jdolecek 1517: so->so_snd.sb_flags &= ~SB_KNOTE;
1518: }
1519:
1520: /*ARGSUSED*/
1521: static int
1522: filt_sowrite(struct knote *kn, long hint)
1523: {
1524: struct socket *so;
1525:
1526: so = (struct socket *)kn->kn_fp->f_data;
1527: kn->kn_data = sbspace(&so->so_snd);
1528: if (so->so_state & SS_CANTSENDMORE) {
1529: kn->kn_flags |= EV_EOF;
1530: kn->kn_fflags = so->so_error;
1531: return (1);
1532: }
1533: if (so->so_error) /* temporary udp error */
1534: return (1);
1535: if (((so->so_state & SS_ISCONNECTED) == 0) &&
1536: (so->so_proto->pr_flags & PR_CONNREQUIRED))
1537: return (0);
1538: if (kn->kn_sfflags & NOTE_LOWAT)
1539: return (kn->kn_data >= kn->kn_sdata);
1540: return (kn->kn_data >= so->so_snd.sb_lowat);
1541: }
1542:
1543: /*ARGSUSED*/
1544: static int
1545: filt_solisten(struct knote *kn, long hint)
1546: {
1547: struct socket *so;
1548:
1549: so = (struct socket *)kn->kn_fp->f_data;
1550:
1551: /*
1552: * Set kn_data to number of incoming connections, not
1553: * counting partial (incomplete) connections.
1554: */
1555: kn->kn_data = so->so_qlen;
1556: return (kn->kn_data > 0);
1557: }
1558:
1559: static const struct filterops solisten_filtops =
1560: { 1, NULL, filt_sordetach, filt_solisten };
1561: static const struct filterops soread_filtops =
1562: { 1, NULL, filt_sordetach, filt_soread };
1563: static const struct filterops sowrite_filtops =
1564: { 1, NULL, filt_sowdetach, filt_sowrite };
1565:
1566: int
1567: soo_kqfilter(struct file *fp, struct knote *kn)
1568: {
1569: struct socket *so;
1570: struct sockbuf *sb;
1571:
1572: so = (struct socket *)kn->kn_fp->f_data;
1573: switch (kn->kn_filter) {
1574: case EVFILT_READ:
1575: if (so->so_options & SO_ACCEPTCONN)
1576: kn->kn_fop = &solisten_filtops;
1577: else
1578: kn->kn_fop = &soread_filtops;
1579: sb = &so->so_rcv;
1580: break;
1581: case EVFILT_WRITE:
1582: kn->kn_fop = &sowrite_filtops;
1583: sb = &so->so_snd;
1584: break;
1585: default:
1586: return (1);
1587: }
1.73 christos 1588: SLIST_INSERT_HEAD(&sb->sb_sel.sel_klist, kn, kn_selnext);
1.72 jdolecek 1589: sb->sb_flags |= SB_KNOTE;
1590: return (0);
1591: }
1592:
CVSweb <webmaster@jp.NetBSD.org>