Annotation of src/sys/kern/uipc_socket.c, Revision 1.54.2.17
1.54.2.15 nathanw 1: /* $NetBSD$ */
1.54.2.11 nathanw 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: */
1.54.2.6 nathanw 73:
74: #include <sys/cdefs.h>
1.54.2.15 nathanw 75: __KERNEL_RCSID(0, "$NetBSD$");
1.54.2.11 nathanw 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>
1.54.2.17! nathanw 94: #include <sys/event.h>
1.37 thorpej 95:
1.54.2.11 nathanw 96: #include <uvm/uvm.h>
97:
1.54 lukem 98: struct pool socket_pool;
1.37 thorpej 99:
1.54 lukem 100: extern int somaxconn; /* patchable (XXX sysctl) */
101: int somaxconn = SOMAXCONN;
1.49 jonathan 102:
1.54.2.11 nathanw 103: #ifdef SOSEND_COUNTERS
104: #include <sys/device.h>
105:
106: struct evcnt sosend_loan_big = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
107: NULL, "sosend", "loan big");
108: struct evcnt sosend_copy_big = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
109: NULL, "sosend", "copy big");
110: struct evcnt sosend_copy_small = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
111: NULL, "sosend", "copy small");
112: struct evcnt sosend_kvalimit = EVCNT_INITIALIZER(EVCNT_TYPE_MISC,
113: NULL, "sosend", "kva limit");
114:
115: #define SOSEND_COUNTER_INCR(ev) (ev)->ev_count++
116:
117: #else
118:
119: #define SOSEND_COUNTER_INCR(ev) /* nothing */
120:
121: #endif /* SOSEND_COUNTERS */
122:
1.37 thorpej 123: void
1.54 lukem 124: soinit(void)
1.37 thorpej 125: {
126:
127: pool_init(&socket_pool, sizeof(struct socket), 0, 0, 0,
1.54.2.9 nathanw 128: "sockpl", NULL);
1.54.2.11 nathanw 129:
130: #ifdef SOSEND_COUNTERS
131: evcnt_attach_static(&sosend_loan_big);
132: evcnt_attach_static(&sosend_copy_big);
133: evcnt_attach_static(&sosend_copy_small);
134: evcnt_attach_static(&sosend_kvalimit);
135: #endif /* SOSEND_COUNTERS */
136: }
137:
1.54.2.16 nathanw 138: #ifdef SOSEND_NO_LOAN
1.54.2.11 nathanw 139: int use_sosend_loan = 0;
1.54.2.16 nathanw 140: #else
141: int use_sosend_loan = 1;
1.54.2.11 nathanw 142: #endif
143:
144: struct mbuf *so_pendfree;
145:
146: int somaxkva = 16 * 1024 * 1024;
147: int socurkva;
148: int sokvawaiters;
149:
150: #define SOCK_LOAN_THRESH 4096
151: #define SOCK_LOAN_CHUNK 65536
152:
153: static void
154: sodoloanfree(caddr_t buf, u_int size)
155: {
156: struct vm_page **pgs;
157: vaddr_t va, sva, eva;
158: vsize_t len;
159: paddr_t pa;
160: int i, npgs;
161:
162: eva = round_page((vaddr_t) buf + size);
163: sva = trunc_page((vaddr_t) buf);
164: len = eva - sva;
165: npgs = len >> PAGE_SHIFT;
166:
167: pgs = alloca(npgs * sizeof(*pgs));
168:
169: for (i = 0, va = sva; va < eva; i++, va += PAGE_SIZE) {
170: if (pmap_extract(pmap_kernel(), va, &pa) == FALSE)
171: panic("sodoloanfree: va 0x%lx not mapped", va);
172: pgs[i] = PHYS_TO_VM_PAGE(pa);
173: }
174:
175: pmap_kremove(sva, len);
176: pmap_update(pmap_kernel());
177: uvm_unloan(pgs, npgs, UVM_LOAN_TOPAGE);
178: uvm_km_free(kernel_map, sva, len);
179: socurkva -= len;
180: if (sokvawaiters)
181: wakeup(&socurkva);
182: }
183:
184: static size_t
185: sodopendfree(struct socket *so)
186: {
187: struct mbuf *m;
188: size_t rv = 0;
189: int s;
190:
191: s = splvm();
192:
193: for (;;) {
194: m = so_pendfree;
195: if (m == NULL)
196: break;
197: so_pendfree = m->m_next;
198: splx(s);
199:
200: rv += m->m_ext.ext_size;
201: sodoloanfree(m->m_ext.ext_buf, m->m_ext.ext_size);
202: s = splvm();
203: pool_cache_put(&mbpool_cache, m);
204: }
205:
206: for (;;) {
207: m = so->so_pendfree;
208: if (m == NULL)
209: break;
210: so->so_pendfree = m->m_next;
211: splx(s);
212:
213: rv += m->m_ext.ext_size;
214: sodoloanfree(m->m_ext.ext_buf, m->m_ext.ext_size);
215: s = splvm();
216: pool_cache_put(&mbpool_cache, m);
217: }
218:
219: splx(s);
220: return (rv);
221: }
222:
223: static void
224: soloanfree(struct mbuf *m, caddr_t buf, u_int size, void *arg)
225: {
226: struct socket *so = arg;
227: int s;
228:
229: if (m == NULL) {
230: sodoloanfree(buf, size);
231: return;
232: }
233:
234: s = splvm();
235: m->m_next = so->so_pendfree;
236: so->so_pendfree = m;
237: splx(s);
238: if (sokvawaiters)
239: wakeup(&socurkva);
240: }
241:
242: static long
243: sosend_loan(struct socket *so, struct uio *uio, struct mbuf *m, long space)
244: {
245: struct iovec *iov = uio->uio_iov;
246: vaddr_t sva, eva;
247: vsize_t len;
248: struct vm_page **pgs;
249: vaddr_t lva, va;
250: int npgs, s, i, error;
251:
252: if (uio->uio_segflg != UIO_USERSPACE)
253: return (0);
254:
255: if (iov->iov_len < (size_t) space)
256: space = iov->iov_len;
257: if (space > SOCK_LOAN_CHUNK)
258: space = SOCK_LOAN_CHUNK;
259:
260: eva = round_page((vaddr_t) iov->iov_base + space);
261: sva = trunc_page((vaddr_t) iov->iov_base);
262: len = eva - sva;
263: npgs = len >> PAGE_SHIFT;
264:
265: while (socurkva + len > somaxkva) {
266: if (sodopendfree(so))
267: continue;
268: SOSEND_COUNTER_INCR(&sosend_kvalimit);
269: s = splvm();
270: sokvawaiters++;
271: (void) tsleep(&socurkva, PVM, "sokva", 0);
272: sokvawaiters--;
273: splx(s);
274: }
275:
276: lva = uvm_km_valloc_wait(kernel_map, len);
277: if (lva == 0)
278: return (0);
279: socurkva += len;
280:
281: pgs = alloca(npgs * sizeof(*pgs));
282:
283: error = uvm_loan(&uio->uio_procp->p_vmspace->vm_map, sva, len,
284: pgs, UVM_LOAN_TOPAGE);
285: if (error) {
286: uvm_km_free(kernel_map, lva, len);
287: socurkva -= len;
288: return (0);
289: }
290:
291: for (i = 0, va = lva; i < npgs; i++, va += PAGE_SIZE)
292: pmap_kenter_pa(va, VM_PAGE_TO_PHYS(pgs[i]), VM_PROT_READ);
293: pmap_update(pmap_kernel());
294:
295: lva += (vaddr_t) iov->iov_base & PAGE_MASK;
296:
297: MEXTADD(m, (caddr_t) lva, space, M_MBUF, soloanfree, so);
298:
299: uio->uio_resid -= space;
300: /* uio_offset not updated, not set/used for write(2) */
301: uio->uio_iov->iov_base = (caddr_t) uio->uio_iov->iov_base + space;
302: uio->uio_iov->iov_len -= space;
303: if (uio->uio_iov->iov_len == 0) {
304: uio->uio_iov++;
305: uio->uio_iovcnt--;
306: }
307:
308: return (space);
1.37 thorpej 309: }
1.1 cgd 310:
311: /*
312: * Socket operation routines.
313: * These routines are called by the routines in
314: * sys_socket.c or from a system process, and
315: * implement the semantics of socket operations by
316: * switching out to the protocol specific routines.
317: */
318: /*ARGSUSED*/
1.3 andrew 319: int
1.54 lukem 320: socreate(int dom, struct socket **aso, int type, int proto)
1.1 cgd 321: {
1.54 lukem 322: struct proc *p;
323: struct protosw *prp;
324: struct socket *so;
325: int error, s;
1.1 cgd 326:
1.54.2.14 nathanw 327: p = curproc; /* XXX */
1.1 cgd 328: if (proto)
329: prp = pffindproto(dom, proto, type);
330: else
331: prp = pffindtype(dom, type);
1.15 mycroft 332: if (prp == 0 || prp->pr_usrreq == 0)
1.1 cgd 333: return (EPROTONOSUPPORT);
334: if (prp->pr_type != type)
335: return (EPROTOTYPE);
1.39 matt 336: s = splsoftnet();
1.37 thorpej 337: so = pool_get(&socket_pool, PR_WAITOK);
1.38 perry 338: memset((caddr_t)so, 0, sizeof(*so));
1.31 thorpej 339: TAILQ_INIT(&so->so_q0);
340: TAILQ_INIT(&so->so_q);
1.1 cgd 341: so->so_type = type;
342: so->so_proto = prp;
1.33 matt 343: so->so_send = sosend;
344: so->so_receive = soreceive;
1.44 lukem 345: if (p != 0)
346: so->so_uid = p->p_ucred->cr_uid;
1.22 mycroft 347: error = (*prp->pr_usrreq)(so, PRU_ATTACH, (struct mbuf *)0,
348: (struct mbuf *)(long)proto, (struct mbuf *)0, p);
1.1 cgd 349: if (error) {
350: so->so_state |= SS_NOFDREF;
351: sofree(so);
1.39 matt 352: splx(s);
1.1 cgd 353: return (error);
354: }
1.39 matt 355: splx(s);
1.1 cgd 356: *aso = so;
357: return (0);
358: }
359:
1.3 andrew 360: int
1.54 lukem 361: sobind(struct socket *so, struct mbuf *nam, struct proc *p)
1.1 cgd 362: {
1.54 lukem 363: int s, error;
1.1 cgd 364:
1.54 lukem 365: s = splsoftnet();
1.22 mycroft 366: error = (*so->so_proto->pr_usrreq)(so, PRU_BIND, (struct mbuf *)0,
367: nam, (struct mbuf *)0, p);
1.1 cgd 368: splx(s);
369: return (error);
370: }
371:
1.3 andrew 372: int
1.54 lukem 373: solisten(struct socket *so, int backlog)
1.1 cgd 374: {
1.54 lukem 375: int s, error;
1.1 cgd 376:
1.54 lukem 377: s = splsoftnet();
1.22 mycroft 378: error = (*so->so_proto->pr_usrreq)(so, PRU_LISTEN, (struct mbuf *)0,
379: (struct mbuf *)0, (struct mbuf *)0, (struct proc *)0);
1.1 cgd 380: if (error) {
381: splx(s);
382: return (error);
383: }
1.54.2.10 nathanw 384: if (TAILQ_EMPTY(&so->so_q))
1.1 cgd 385: so->so_options |= SO_ACCEPTCONN;
386: if (backlog < 0)
387: backlog = 0;
1.49 jonathan 388: so->so_qlimit = min(backlog, somaxconn);
1.1 cgd 389: splx(s);
390: return (0);
391: }
392:
1.21 christos 393: void
1.54 lukem 394: sofree(struct socket *so)
1.1 cgd 395: {
1.54.2.11 nathanw 396: struct mbuf *m;
1.1 cgd 397:
1.43 mycroft 398: if (so->so_pcb || (so->so_state & SS_NOFDREF) == 0)
1.1 cgd 399: return;
1.43 mycroft 400: if (so->so_head) {
401: /*
402: * We must not decommission a socket that's on the accept(2)
403: * queue. If we do, then accept(2) may hang after select(2)
404: * indicated that the listening socket was ready.
405: */
406: if (!soqremque(so, 0))
407: return;
408: }
1.1 cgd 409: sbrelease(&so->so_snd);
410: sorflush(so);
1.54.2.11 nathanw 411: while ((m = so->so_pendfree) != NULL) {
412: so->so_pendfree = m->m_next;
413: m->m_next = so_pendfree;
414: so_pendfree = m;
415: }
1.37 thorpej 416: pool_put(&socket_pool, so);
1.1 cgd 417: }
418:
419: /*
420: * Close a socket on last file table reference removal.
421: * Initiate disconnect if connected.
422: * Free socket when disconnect complete.
423: */
1.3 andrew 424: int
1.54 lukem 425: soclose(struct socket *so)
1.1 cgd 426: {
1.54 lukem 427: struct socket *so2;
428: int s, error;
1.1 cgd 429:
1.54 lukem 430: error = 0;
431: s = splsoftnet(); /* conservative */
1.1 cgd 432: if (so->so_options & SO_ACCEPTCONN) {
1.54.2.10 nathanw 433: while ((so2 = TAILQ_FIRST(&so->so_q0)) != 0) {
1.42 mycroft 434: (void) soqremque(so2, 0);
1.41 mycroft 435: (void) soabort(so2);
436: }
1.54.2.10 nathanw 437: while ((so2 = TAILQ_FIRST(&so->so_q)) != 0) {
1.42 mycroft 438: (void) soqremque(so2, 1);
1.41 mycroft 439: (void) soabort(so2);
440: }
1.1 cgd 441: }
442: if (so->so_pcb == 0)
443: goto discard;
444: if (so->so_state & SS_ISCONNECTED) {
445: if ((so->so_state & SS_ISDISCONNECTING) == 0) {
446: error = sodisconnect(so);
447: if (error)
448: goto drop;
449: }
450: if (so->so_options & SO_LINGER) {
451: if ((so->so_state & SS_ISDISCONNECTING) &&
452: (so->so_state & SS_NBIO))
453: goto drop;
1.21 christos 454: while (so->so_state & SS_ISCONNECTED) {
455: error = tsleep((caddr_t)&so->so_timeo,
456: PSOCK | PCATCH, netcls,
1.30 thorpej 457: so->so_linger * hz);
1.21 christos 458: if (error)
1.1 cgd 459: break;
1.21 christos 460: }
1.1 cgd 461: }
462: }
1.54 lukem 463: drop:
1.1 cgd 464: if (so->so_pcb) {
1.22 mycroft 465: int error2 = (*so->so_proto->pr_usrreq)(so, PRU_DETACH,
466: (struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0,
467: (struct proc *)0);
1.1 cgd 468: if (error == 0)
469: error = error2;
470: }
1.54 lukem 471: discard:
1.1 cgd 472: if (so->so_state & SS_NOFDREF)
473: panic("soclose: NOFDREF");
474: so->so_state |= SS_NOFDREF;
475: sofree(so);
476: splx(s);
477: return (error);
478: }
479:
480: /*
1.20 mycroft 481: * Must be called at splsoftnet...
1.1 cgd 482: */
1.3 andrew 483: int
1.54 lukem 484: soabort(struct socket *so)
1.1 cgd 485: {
486:
1.22 mycroft 487: return (*so->so_proto->pr_usrreq)(so, PRU_ABORT, (struct mbuf *)0,
488: (struct mbuf *)0, (struct mbuf *)0, (struct proc *)0);
1.1 cgd 489: }
490:
1.3 andrew 491: int
1.54 lukem 492: soaccept(struct socket *so, struct mbuf *nam)
1.1 cgd 493: {
1.54 lukem 494: int s, error;
1.1 cgd 495:
1.54 lukem 496: error = 0;
497: s = splsoftnet();
1.1 cgd 498: if ((so->so_state & SS_NOFDREF) == 0)
499: panic("soaccept: !NOFDREF");
500: so->so_state &= ~SS_NOFDREF;
1.54.2.2 nathanw 501: if ((so->so_state & SS_ISDISCONNECTED) == 0 ||
502: (so->so_proto->pr_flags & PR_ABRTACPTDIS) == 0)
1.41 mycroft 503: error = (*so->so_proto->pr_usrreq)(so, PRU_ACCEPT,
504: (struct mbuf *)0, nam, (struct mbuf *)0, (struct proc *)0);
505: else
1.53 itojun 506: error = ECONNABORTED;
1.52 itojun 507:
1.1 cgd 508: splx(s);
509: return (error);
510: }
511:
1.3 andrew 512: int
1.54 lukem 513: soconnect(struct socket *so, struct mbuf *nam)
1.1 cgd 514: {
1.54 lukem 515: struct proc *p;
516: int s, error;
1.1 cgd 517:
1.54.2.12 nathanw 518: p = curproc; /* XXX */
1.1 cgd 519: if (so->so_options & SO_ACCEPTCONN)
520: return (EOPNOTSUPP);
1.20 mycroft 521: s = splsoftnet();
1.1 cgd 522: /*
523: * If protocol is connection-based, can only connect once.
524: * Otherwise, if connected, try to disconnect first.
525: * This allows user to disconnect by connecting to, e.g.,
526: * a null address.
527: */
528: if (so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING) &&
529: ((so->so_proto->pr_flags & PR_CONNREQUIRED) ||
530: (error = sodisconnect(so))))
531: error = EISCONN;
532: else
533: error = (*so->so_proto->pr_usrreq)(so, PRU_CONNECT,
1.23 mycroft 534: (struct mbuf *)0, nam, (struct mbuf *)0, p);
1.1 cgd 535: splx(s);
536: return (error);
537: }
538:
1.3 andrew 539: int
1.54 lukem 540: soconnect2(struct socket *so1, struct socket *so2)
1.1 cgd 541: {
1.54 lukem 542: int s, error;
1.1 cgd 543:
1.54 lukem 544: s = splsoftnet();
1.22 mycroft 545: error = (*so1->so_proto->pr_usrreq)(so1, PRU_CONNECT2,
546: (struct mbuf *)0, (struct mbuf *)so2, (struct mbuf *)0,
547: (struct proc *)0);
1.1 cgd 548: splx(s);
549: return (error);
550: }
551:
1.3 andrew 552: int
1.54 lukem 553: sodisconnect(struct socket *so)
1.1 cgd 554: {
1.54 lukem 555: int s, error;
1.1 cgd 556:
1.54 lukem 557: s = splsoftnet();
1.1 cgd 558: if ((so->so_state & SS_ISCONNECTED) == 0) {
559: error = ENOTCONN;
560: goto bad;
561: }
562: if (so->so_state & SS_ISDISCONNECTING) {
563: error = EALREADY;
564: goto bad;
565: }
1.22 mycroft 566: error = (*so->so_proto->pr_usrreq)(so, PRU_DISCONNECT,
567: (struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0,
568: (struct proc *)0);
1.54 lukem 569: bad:
1.1 cgd 570: splx(s);
1.54.2.11 nathanw 571: sodopendfree(so);
1.1 cgd 572: return (error);
573: }
574:
1.15 mycroft 575: #define SBLOCKWAIT(f) (((f) & MSG_DONTWAIT) ? M_NOWAIT : M_WAITOK)
1.1 cgd 576: /*
577: * Send on a socket.
578: * If send must go all at once and message is larger than
579: * send buffering, then hard error.
580: * Lock against other senders.
581: * If must go all at once and not enough room now, then
582: * inform user that this would block and do nothing.
583: * Otherwise, if nonblocking, send as much as possible.
584: * The data to be sent is described by "uio" if nonzero,
585: * otherwise by the mbuf chain "top" (which must be null
586: * if uio is not). Data provided in mbuf chain must be small
587: * enough to send all at once.
588: *
589: * Returns nonzero on error, timeout or signal; callers
590: * must check for short counts if EINTR/ERESTART are returned.
591: * Data and control buffers are freed on return.
592: */
1.3 andrew 593: int
1.54 lukem 594: sosend(struct socket *so, struct mbuf *addr, struct uio *uio, struct mbuf *top,
595: struct mbuf *control, int flags)
1.1 cgd 596: {
1.54 lukem 597: struct proc *p;
598: struct mbuf **mp, *m;
1.54.2.5 nathanw 599: long space, len, resid, clen, mlen;
600: int error, s, dontroute, atomic;
1.54 lukem 601:
1.54.2.11 nathanw 602: sodopendfree(so);
603:
1.54.2.12 nathanw 604: p = curproc; /* XXX */
1.54 lukem 605: clen = 0;
606: atomic = sosendallatonce(so) || top;
1.1 cgd 607: if (uio)
608: resid = uio->uio_resid;
609: else
610: resid = top->m_pkthdr.len;
1.7 cgd 611: /*
612: * In theory resid should be unsigned.
613: * However, space must be signed, as it might be less than 0
614: * if we over-committed, and we must use a signed comparison
615: * of space and resid. On the other hand, a negative resid
616: * causes us to loop sending 0-length segments to the protocol.
617: */
1.29 mycroft 618: if (resid < 0) {
619: error = EINVAL;
620: goto out;
621: }
1.1 cgd 622: dontroute =
623: (flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 &&
624: (so->so_proto->pr_flags & PR_ATOMIC);
1.12 mycroft 625: p->p_stats->p_ru.ru_msgsnd++;
1.1 cgd 626: if (control)
627: clen = control->m_len;
628: #define snderr(errno) { error = errno; splx(s); goto release; }
629:
1.54 lukem 630: restart:
1.21 christos 631: if ((error = sblock(&so->so_snd, SBLOCKWAIT(flags))) != 0)
1.1 cgd 632: goto out;
633: do {
1.20 mycroft 634: s = splsoftnet();
1.1 cgd 635: if (so->so_state & SS_CANTSENDMORE)
636: snderr(EPIPE);
1.48 thorpej 637: if (so->so_error) {
638: error = so->so_error;
639: so->so_error = 0;
640: splx(s);
641: goto release;
642: }
1.1 cgd 643: if ((so->so_state & SS_ISCONNECTED) == 0) {
644: if (so->so_proto->pr_flags & PR_CONNREQUIRED) {
645: if ((so->so_state & SS_ISCONFIRMING) == 0 &&
646: !(resid == 0 && clen != 0))
647: snderr(ENOTCONN);
648: } else if (addr == 0)
649: snderr(EDESTADDRREQ);
650: }
651: space = sbspace(&so->so_snd);
652: if (flags & MSG_OOB)
653: space += 1024;
1.21 christos 654: if ((atomic && resid > so->so_snd.sb_hiwat) ||
1.11 mycroft 655: clen > so->so_snd.sb_hiwat)
656: snderr(EMSGSIZE);
657: if (space < resid + clen && uio &&
1.1 cgd 658: (atomic || space < so->so_snd.sb_lowat || space < clen)) {
659: if (so->so_state & SS_NBIO)
660: snderr(EWOULDBLOCK);
661: sbunlock(&so->so_snd);
662: error = sbwait(&so->so_snd);
663: splx(s);
664: if (error)
665: goto out;
666: goto restart;
667: }
668: splx(s);
669: mp = ⊤
670: space -= clen;
671: do {
1.45 tv 672: if (uio == NULL) {
673: /*
674: * Data is prepackaged in "top".
675: */
676: resid = 0;
677: if (flags & MSG_EOR)
678: top->m_flags |= M_EOR;
679: } else do {
680: if (top == 0) {
681: MGETHDR(m, M_WAIT, MT_DATA);
682: mlen = MHLEN;
683: m->m_pkthdr.len = 0;
684: m->m_pkthdr.rcvif = (struct ifnet *)0;
685: } else {
686: MGET(m, M_WAIT, MT_DATA);
687: mlen = MLEN;
688: }
1.54.2.11 nathanw 689: if (use_sosend_loan &&
690: uio->uio_iov->iov_len >= SOCK_LOAN_THRESH &&
691: space >= SOCK_LOAN_THRESH &&
692: (len = sosend_loan(so, uio, m,
693: space)) != 0) {
694: SOSEND_COUNTER_INCR(&sosend_loan_big);
695: space -= len;
696: goto have_data;
697: }
1.45 tv 698: if (resid >= MINCLSIZE && space >= MCLBYTES) {
1.54.2.11 nathanw 699: SOSEND_COUNTER_INCR(&sosend_copy_big);
1.45 tv 700: MCLGET(m, M_WAIT);
701: if ((m->m_flags & M_EXT) == 0)
702: goto nopages;
703: mlen = MCLBYTES;
704: if (atomic && top == 0) {
1.54.2.5 nathanw 705: len = lmin(MCLBYTES - max_hdr,
1.54 lukem 706: resid);
1.45 tv 707: m->m_data += max_hdr;
708: } else
1.54.2.5 nathanw 709: len = lmin(MCLBYTES, resid);
1.45 tv 710: space -= len;
711: } else {
1.54.2.11 nathanw 712: nopages:
713: SOSEND_COUNTER_INCR(&sosend_copy_small);
1.54.2.5 nathanw 714: len = lmin(lmin(mlen, resid), space);
1.45 tv 715: space -= len;
716: /*
717: * For datagram protocols, leave room
718: * for protocol headers in first mbuf.
719: */
720: if (atomic && top == 0 && len < mlen)
721: MH_ALIGN(m, len);
722: }
1.54 lukem 723: error = uiomove(mtod(m, caddr_t), (int)len,
724: uio);
1.54.2.11 nathanw 725: have_data:
1.45 tv 726: resid = uio->uio_resid;
727: m->m_len = len;
728: *mp = m;
729: top->m_pkthdr.len += len;
730: if (error)
731: goto release;
732: mp = &m->m_next;
733: if (resid <= 0) {
734: if (flags & MSG_EOR)
735: top->m_flags |= M_EOR;
736: break;
737: }
738: } while (space > 0 && atomic);
1.46 sommerfe 739:
740: s = splsoftnet();
741:
742: if (so->so_state & SS_CANTSENDMORE)
743: snderr(EPIPE);
1.45 tv 744:
745: if (dontroute)
746: so->so_options |= SO_DONTROUTE;
747: if (resid > 0)
748: so->so_state |= SS_MORETOCOME;
1.46 sommerfe 749: error = (*so->so_proto->pr_usrreq)(so,
750: (flags & MSG_OOB) ? PRU_SENDOOB : PRU_SEND,
751: top, addr, control, p);
1.45 tv 752: if (dontroute)
753: so->so_options &= ~SO_DONTROUTE;
754: if (resid > 0)
755: so->so_state &= ~SS_MORETOCOME;
1.46 sommerfe 756: splx(s);
757:
1.45 tv 758: clen = 0;
759: control = 0;
760: top = 0;
761: mp = ⊤
1.1 cgd 762: if (error)
763: goto release;
764: } while (resid && space > 0);
765: } while (resid);
766:
1.54 lukem 767: release:
1.1 cgd 768: sbunlock(&so->so_snd);
1.54 lukem 769: out:
1.1 cgd 770: if (top)
771: m_freem(top);
772: if (control)
773: m_freem(control);
774: return (error);
775: }
776:
777: /*
778: * Implement receive operations on a socket.
779: * We depend on the way that records are added to the sockbuf
780: * by sbappend*. In particular, each record (mbufs linked through m_next)
781: * must begin with an address if the protocol so specifies,
782: * followed by an optional mbuf or mbufs containing ancillary data,
783: * and then zero or more mbufs of data.
784: * In order to avoid blocking network interrupts for the entire time here,
785: * we splx() while doing the actual copy to user space.
786: * Although the sockbuf is locked, new data may still be appended,
787: * and thus we must maintain consistency of the sockbuf during that time.
788: *
789: * The caller may receive the data as a single mbuf chain by supplying
790: * an mbuf **mp0 for use in returning the chain. The uio is then used
791: * only for the count in uio_resid.
792: */
1.3 andrew 793: int
1.54 lukem 794: soreceive(struct socket *so, struct mbuf **paddr, struct uio *uio,
795: struct mbuf **mp0, struct mbuf **controlp, int *flagsp)
1.1 cgd 796: {
1.54 lukem 797: struct mbuf *m, **mp;
798: int flags, len, error, s, offset, moff, type, orig_resid;
799: struct protosw *pr;
800: struct mbuf *nextrecord;
1.54.2.11 nathanw 801: int mbuf_removed = 0;
1.1 cgd 802:
1.54 lukem 803: pr = so->so_proto;
1.1 cgd 804: mp = mp0;
1.54 lukem 805: type = 0;
806: orig_resid = uio->uio_resid;
1.1 cgd 807: if (paddr)
808: *paddr = 0;
809: if (controlp)
810: *controlp = 0;
811: if (flagsp)
812: flags = *flagsp &~ MSG_EOR;
813: else
814: flags = 0;
1.54.2.11 nathanw 815:
816: if ((flags & MSG_DONTWAIT) == 0)
817: sodopendfree(so);
818:
1.1 cgd 819: if (flags & MSG_OOB) {
820: m = m_get(M_WAIT, MT_DATA);
1.17 cgd 821: error = (*pr->pr_usrreq)(so, PRU_RCVOOB, m,
1.22 mycroft 822: (struct mbuf *)(long)(flags & MSG_PEEK), (struct mbuf *)0,
823: (struct proc *)0);
1.1 cgd 824: if (error)
825: goto bad;
826: do {
827: error = uiomove(mtod(m, caddr_t),
828: (int) min(uio->uio_resid, m->m_len), uio);
829: m = m_free(m);
830: } while (uio->uio_resid && error == 0 && m);
1.54 lukem 831: bad:
1.1 cgd 832: if (m)
833: m_freem(m);
834: return (error);
835: }
836: if (mp)
837: *mp = (struct mbuf *)0;
838: if (so->so_state & SS_ISCONFIRMING && uio->uio_resid)
1.22 mycroft 839: (*pr->pr_usrreq)(so, PRU_RCVD, (struct mbuf *)0,
840: (struct mbuf *)0, (struct mbuf *)0, (struct proc *)0);
1.1 cgd 841:
1.54 lukem 842: restart:
1.21 christos 843: if ((error = sblock(&so->so_rcv, SBLOCKWAIT(flags))) != 0)
1.1 cgd 844: return (error);
1.20 mycroft 845: s = splsoftnet();
1.1 cgd 846:
847: m = so->so_rcv.sb_mb;
848: /*
849: * If we have less data than requested, block awaiting more
850: * (subject to any timeout) if:
1.15 mycroft 851: * 1. the current count is less than the low water mark,
1.1 cgd 852: * 2. MSG_WAITALL is set, and it is possible to do the entire
1.15 mycroft 853: * receive operation at once if we block (resid <= hiwat), or
854: * 3. MSG_DONTWAIT is not set.
1.1 cgd 855: * If MSG_WAITALL is set but resid is larger than the receive buffer,
856: * we have to do the receive in sections, and thus risk returning
857: * a short count if a timeout or signal occurs after we start.
858: */
1.21 christos 859: if (m == 0 || (((flags & MSG_DONTWAIT) == 0 &&
1.15 mycroft 860: so->so_rcv.sb_cc < uio->uio_resid) &&
1.1 cgd 861: (so->so_rcv.sb_cc < so->so_rcv.sb_lowat ||
862: ((flags & MSG_WAITALL) && uio->uio_resid <= so->so_rcv.sb_hiwat)) &&
1.21 christos 863: m->m_nextpkt == 0 && (pr->pr_flags & PR_ATOMIC) == 0)) {
1.1 cgd 864: #ifdef DIAGNOSTIC
865: if (m == 0 && so->so_rcv.sb_cc)
866: panic("receive 1");
867: #endif
868: if (so->so_error) {
869: if (m)
1.15 mycroft 870: goto dontblock;
1.1 cgd 871: error = so->so_error;
872: if ((flags & MSG_PEEK) == 0)
873: so->so_error = 0;
874: goto release;
875: }
876: if (so->so_state & SS_CANTRCVMORE) {
877: if (m)
1.15 mycroft 878: goto dontblock;
1.1 cgd 879: else
880: goto release;
881: }
882: for (; m; m = m->m_next)
883: if (m->m_type == MT_OOBDATA || (m->m_flags & M_EOR)) {
884: m = so->so_rcv.sb_mb;
885: goto dontblock;
886: }
887: if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0 &&
888: (so->so_proto->pr_flags & PR_CONNREQUIRED)) {
889: error = ENOTCONN;
890: goto release;
891: }
892: if (uio->uio_resid == 0)
893: goto release;
1.15 mycroft 894: if ((so->so_state & SS_NBIO) || (flags & MSG_DONTWAIT)) {
1.1 cgd 895: error = EWOULDBLOCK;
896: goto release;
897: }
1.54.2.15 nathanw 898: SBLASTRECORDCHK(&so->so_rcv, "soreceive sbwait 1");
899: SBLASTMBUFCHK(&so->so_rcv, "soreceive sbwait 1");
1.1 cgd 900: sbunlock(&so->so_rcv);
901: error = sbwait(&so->so_rcv);
902: splx(s);
903: if (error)
904: return (error);
905: goto restart;
906: }
1.54 lukem 907: dontblock:
1.54.2.15 nathanw 908: /*
909: * On entry here, m points to the first record of the socket buffer.
910: * While we process the initial mbufs containing address and control
911: * info, we save a copy of m->m_nextpkt into nextrecord.
912: */
1.15 mycroft 913: #ifdef notyet /* XXXX */
914: if (uio->uio_procp)
915: uio->uio_procp->p_stats->p_ru.ru_msgrcv++;
916: #endif
1.54.2.15 nathanw 917: KASSERT(m == so->so_rcv.sb_mb);
918: SBLASTRECORDCHK(&so->so_rcv, "soreceive 1");
919: SBLASTMBUFCHK(&so->so_rcv, "soreceive 1");
1.1 cgd 920: nextrecord = m->m_nextpkt;
921: if (pr->pr_flags & PR_ADDR) {
922: #ifdef DIAGNOSTIC
923: if (m->m_type != MT_SONAME)
924: panic("receive 1a");
925: #endif
1.3 andrew 926: orig_resid = 0;
1.1 cgd 927: if (flags & MSG_PEEK) {
928: if (paddr)
929: *paddr = m_copy(m, 0, m->m_len);
930: m = m->m_next;
931: } else {
932: sbfree(&so->so_rcv, m);
1.54.2.11 nathanw 933: mbuf_removed = 1;
1.1 cgd 934: if (paddr) {
935: *paddr = m;
936: so->so_rcv.sb_mb = m->m_next;
937: m->m_next = 0;
938: m = so->so_rcv.sb_mb;
939: } else {
940: MFREE(m, so->so_rcv.sb_mb);
941: m = so->so_rcv.sb_mb;
942: }
943: }
944: }
945: while (m && m->m_type == MT_CONTROL && error == 0) {
946: if (flags & MSG_PEEK) {
947: if (controlp)
948: *controlp = m_copy(m, 0, m->m_len);
949: m = m->m_next;
950: } else {
951: sbfree(&so->so_rcv, m);
1.54.2.11 nathanw 952: mbuf_removed = 1;
1.1 cgd 953: if (controlp) {
954: if (pr->pr_domain->dom_externalize &&
955: mtod(m, struct cmsghdr *)->cmsg_type ==
956: SCM_RIGHTS)
1.45 tv 957: error = (*pr->pr_domain->dom_externalize)(m);
1.1 cgd 958: *controlp = m;
959: so->so_rcv.sb_mb = m->m_next;
960: m->m_next = 0;
961: m = so->so_rcv.sb_mb;
962: } else {
963: MFREE(m, so->so_rcv.sb_mb);
964: m = so->so_rcv.sb_mb;
965: }
966: }
1.3 andrew 967: if (controlp) {
968: orig_resid = 0;
1.1 cgd 969: controlp = &(*controlp)->m_next;
1.3 andrew 970: }
1.1 cgd 971: }
1.54.2.15 nathanw 972:
973: /*
974: * If m is non-NULL, we have some data to read. From now on,
975: * make sure to keep sb_lastrecord consistent when working on
976: * the last packet on the chain (nextrecord == NULL) and we
977: * change m->m_nextpkt.
978: */
1.1 cgd 979: if (m) {
1.54.2.15 nathanw 980: if ((flags & MSG_PEEK) == 0) {
1.1 cgd 981: m->m_nextpkt = nextrecord;
1.54.2.15 nathanw 982: /*
983: * If nextrecord == NULL (this is a single chain),
984: * then sb_lastrecord may not be valid here if m
985: * was changed earlier.
986: */
987: if (nextrecord == NULL) {
988: KASSERT(so->so_rcv.sb_mb == m);
989: so->so_rcv.sb_lastrecord = m;
990: }
991: }
1.1 cgd 992: type = m->m_type;
993: if (type == MT_OOBDATA)
994: flags |= MSG_OOB;
1.54.2.15 nathanw 995: } else {
996: if ((flags & MSG_PEEK) == 0) {
997: KASSERT(so->so_rcv.sb_mb == m);
998: so->so_rcv.sb_mb = nextrecord;
999: SB_EMPTY_FIXUP(&so->so_rcv);
1000: }
1.1 cgd 1001: }
1.54.2.15 nathanw 1002: SBLASTRECORDCHK(&so->so_rcv, "soreceive 2");
1003: SBLASTMBUFCHK(&so->so_rcv, "soreceive 2");
1004:
1.1 cgd 1005: moff = 0;
1006: offset = 0;
1007: while (m && uio->uio_resid > 0 && error == 0) {
1008: if (m->m_type == MT_OOBDATA) {
1009: if (type != MT_OOBDATA)
1010: break;
1011: } else if (type == MT_OOBDATA)
1012: break;
1013: #ifdef DIAGNOSTIC
1014: else if (m->m_type != MT_DATA && m->m_type != MT_HEADER)
1015: panic("receive 3");
1016: #endif
1017: so->so_state &= ~SS_RCVATMARK;
1018: len = uio->uio_resid;
1019: if (so->so_oobmark && len > so->so_oobmark - offset)
1020: len = so->so_oobmark - offset;
1021: if (len > m->m_len - moff)
1022: len = m->m_len - moff;
1023: /*
1024: * If mp is set, just pass back the mbufs.
1025: * Otherwise copy them out via the uio, then free.
1026: * Sockbuf must be consistent here (points to current mbuf,
1027: * it points to next record) when we drop priority;
1028: * we must note any additions to the sockbuf when we
1029: * block interrupts again.
1030: */
1031: if (mp == 0) {
1.54.2.15 nathanw 1032: SBLASTRECORDCHK(&so->so_rcv, "soreceive uiomove");
1033: SBLASTMBUFCHK(&so->so_rcv, "soreceive uiomove");
1.1 cgd 1034: splx(s);
1035: error = uiomove(mtod(m, caddr_t) + moff, (int)len, uio);
1.20 mycroft 1036: s = splsoftnet();
1.54.2.11 nathanw 1037: if (error) {
1038: /*
1039: * If any part of the record has been removed
1040: * (such as the MT_SONAME mbuf, which will
1041: * happen when PR_ADDR, and thus also
1042: * PR_ATOMIC, is set), then drop the entire
1043: * record to maintain the atomicity of the
1044: * receive operation.
1045: *
1046: * This avoids a later panic("receive 1a")
1047: * when compiled with DIAGNOSTIC.
1048: */
1049: if (m && mbuf_removed
1050: && (pr->pr_flags & PR_ATOMIC))
1051: (void) sbdroprecord(&so->so_rcv);
1052:
1.54.2.4 nathanw 1053: goto release;
1.54.2.11 nathanw 1054: }
1.1 cgd 1055: } else
1056: uio->uio_resid -= len;
1057: if (len == m->m_len - moff) {
1058: if (m->m_flags & M_EOR)
1059: flags |= MSG_EOR;
1060: if (flags & MSG_PEEK) {
1061: m = m->m_next;
1062: moff = 0;
1063: } else {
1064: nextrecord = m->m_nextpkt;
1065: sbfree(&so->so_rcv, m);
1066: if (mp) {
1067: *mp = m;
1068: mp = &m->m_next;
1069: so->so_rcv.sb_mb = m = m->m_next;
1070: *mp = (struct mbuf *)0;
1071: } else {
1072: MFREE(m, so->so_rcv.sb_mb);
1073: m = so->so_rcv.sb_mb;
1074: }
1.54.2.15 nathanw 1075: /*
1076: * If m != NULL, we also know that
1077: * so->so_rcv.sb_mb != NULL.
1078: */
1079: KASSERT(so->so_rcv.sb_mb == m);
1080: if (m) {
1.1 cgd 1081: m->m_nextpkt = nextrecord;
1.54.2.15 nathanw 1082: if (nextrecord == NULL)
1083: so->so_rcv.sb_lastrecord = m;
1084: } else {
1085: so->so_rcv.sb_mb = nextrecord;
1086: SB_EMPTY_FIXUP(&so->so_rcv);
1087: }
1088: SBLASTRECORDCHK(&so->so_rcv, "soreceive 3");
1089: SBLASTMBUFCHK(&so->so_rcv, "soreceive 3");
1.1 cgd 1090: }
1091: } else {
1092: if (flags & MSG_PEEK)
1093: moff += len;
1094: else {
1095: if (mp)
1096: *mp = m_copym(m, 0, len, M_WAIT);
1097: m->m_data += len;
1098: m->m_len -= len;
1099: so->so_rcv.sb_cc -= len;
1100: }
1101: }
1102: if (so->so_oobmark) {
1103: if ((flags & MSG_PEEK) == 0) {
1104: so->so_oobmark -= len;
1105: if (so->so_oobmark == 0) {
1106: so->so_state |= SS_RCVATMARK;
1107: break;
1108: }
1.7 cgd 1109: } else {
1.1 cgd 1110: offset += len;
1.7 cgd 1111: if (offset == so->so_oobmark)
1112: break;
1113: }
1.1 cgd 1114: }
1115: if (flags & MSG_EOR)
1116: break;
1117: /*
1118: * If the MSG_WAITALL flag is set (for non-atomic socket),
1119: * we must not quit until "uio->uio_resid == 0" or an error
1120: * termination. If a signal/timeout occurs, return
1121: * with a short count but without error.
1122: * Keep sockbuf locked against other readers.
1123: */
1124: while (flags & MSG_WAITALL && m == 0 && uio->uio_resid > 0 &&
1.3 andrew 1125: !sosendallatonce(so) && !nextrecord) {
1.1 cgd 1126: if (so->so_error || so->so_state & SS_CANTRCVMORE)
1127: break;
1.54.2.11 nathanw 1128: /*
1129: * If we are peeking and the socket receive buffer is
1130: * full, stop since we can't get more data to peek at.
1131: */
1132: if ((flags & MSG_PEEK) && sbspace(&so->so_rcv) <= 0)
1133: break;
1134: /*
1135: * If we've drained the socket buffer, tell the
1136: * protocol in case it needs to do something to
1137: * get it filled again.
1138: */
1139: if ((pr->pr_flags & PR_WANTRCVD) && so->so_pcb)
1140: (*pr->pr_usrreq)(so, PRU_RCVD,
1141: (struct mbuf *)0,
1142: (struct mbuf *)(long)flags,
1143: (struct mbuf *)0,
1144: (struct proc *)0);
1.54.2.15 nathanw 1145: SBLASTRECORDCHK(&so->so_rcv, "soreceive sbwait 2");
1146: SBLASTMBUFCHK(&so->so_rcv, "soreceive sbwait 2");
1.1 cgd 1147: error = sbwait(&so->so_rcv);
1148: if (error) {
1149: sbunlock(&so->so_rcv);
1150: splx(s);
1151: return (0);
1152: }
1.21 christos 1153: if ((m = so->so_rcv.sb_mb) != NULL)
1.1 cgd 1154: nextrecord = m->m_nextpkt;
1155: }
1156: }
1.3 andrew 1157:
1158: if (m && pr->pr_flags & PR_ATOMIC) {
1159: flags |= MSG_TRUNC;
1160: if ((flags & MSG_PEEK) == 0)
1161: (void) sbdroprecord(&so->so_rcv);
1162: }
1.1 cgd 1163: if ((flags & MSG_PEEK) == 0) {
1.54.2.15 nathanw 1164: if (m == 0) {
1165: /*
1166: * First part is an inline SB_EMPTY_FIXUP(). Second
1167: * part makes sure sb_lastrecord is up-to-date if
1168: * there is still data in the socket buffer.
1169: */
1.1 cgd 1170: so->so_rcv.sb_mb = nextrecord;
1.54.2.15 nathanw 1171: if (so->so_rcv.sb_mb == NULL) {
1172: so->so_rcv.sb_mbtail = NULL;
1173: so->so_rcv.sb_lastrecord = NULL;
1174: } else if (nextrecord->m_nextpkt == NULL)
1175: so->so_rcv.sb_lastrecord = nextrecord;
1176: }
1177: SBLASTRECORDCHK(&so->so_rcv, "soreceive 4");
1178: SBLASTMBUFCHK(&so->so_rcv, "soreceive 4");
1.1 cgd 1179: if (pr->pr_flags & PR_WANTRCVD && so->so_pcb)
1.22 mycroft 1180: (*pr->pr_usrreq)(so, PRU_RCVD, (struct mbuf *)0,
1181: (struct mbuf *)(long)flags, (struct mbuf *)0,
1182: (struct proc *)0);
1.1 cgd 1183: }
1.3 andrew 1184: if (orig_resid == uio->uio_resid && orig_resid &&
1185: (flags & MSG_EOR) == 0 && (so->so_state & SS_CANTRCVMORE) == 0) {
1186: sbunlock(&so->so_rcv);
1187: splx(s);
1188: goto restart;
1189: }
1190:
1.1 cgd 1191: if (flagsp)
1192: *flagsp |= flags;
1.54 lukem 1193: release:
1.1 cgd 1194: sbunlock(&so->so_rcv);
1195: splx(s);
1196: return (error);
1197: }
1198:
1.14 mycroft 1199: int
1.54 lukem 1200: soshutdown(struct socket *so, int how)
1.1 cgd 1201: {
1.54 lukem 1202: struct protosw *pr;
1.34 kleink 1203:
1.54 lukem 1204: pr = so->so_proto;
1.34 kleink 1205: if (!(how == SHUT_RD || how == SHUT_WR || how == SHUT_RDWR))
1206: return (EINVAL);
1.1 cgd 1207:
1.34 kleink 1208: if (how == SHUT_RD || how == SHUT_RDWR)
1.1 cgd 1209: sorflush(so);
1.34 kleink 1210: if (how == SHUT_WR || how == SHUT_RDWR)
1.22 mycroft 1211: return (*pr->pr_usrreq)(so, PRU_SHUTDOWN, (struct mbuf *)0,
1212: (struct mbuf *)0, (struct mbuf *)0, (struct proc *)0);
1.1 cgd 1213: return (0);
1214: }
1215:
1.14 mycroft 1216: void
1.54 lukem 1217: sorflush(struct socket *so)
1.1 cgd 1218: {
1.54 lukem 1219: struct sockbuf *sb, asb;
1220: struct protosw *pr;
1221: int s;
1.1 cgd 1222:
1.54 lukem 1223: sb = &so->so_rcv;
1224: pr = so->so_proto;
1.1 cgd 1225: sb->sb_flags |= SB_NOINTR;
1.15 mycroft 1226: (void) sblock(sb, M_WAITOK);
1.54.2.3 nathanw 1227: s = splnet();
1.1 cgd 1228: socantrcvmore(so);
1229: sbunlock(sb);
1230: asb = *sb;
1.38 perry 1231: memset((caddr_t)sb, 0, sizeof(*sb));
1.1 cgd 1232: splx(s);
1233: if (pr->pr_flags & PR_RIGHTS && pr->pr_domain->dom_dispose)
1234: (*pr->pr_domain->dom_dispose)(asb.sb_mb);
1235: sbrelease(&asb);
1236: }
1237:
1.14 mycroft 1238: int
1.54 lukem 1239: sosetopt(struct socket *so, int level, int optname, struct mbuf *m0)
1.1 cgd 1240: {
1.54 lukem 1241: int error;
1242: struct mbuf *m;
1.1 cgd 1243:
1.54 lukem 1244: error = 0;
1245: m = m0;
1.1 cgd 1246: if (level != SOL_SOCKET) {
1247: if (so->so_proto && so->so_proto->pr_ctloutput)
1248: return ((*so->so_proto->pr_ctloutput)
1249: (PRCO_SETOPT, so, level, optname, &m0));
1250: error = ENOPROTOOPT;
1251: } else {
1252: switch (optname) {
1253:
1254: case SO_LINGER:
1.36 perry 1255: if (m == NULL || m->m_len != sizeof(struct linger)) {
1.1 cgd 1256: error = EINVAL;
1257: goto bad;
1258: }
1259: so->so_linger = mtod(m, struct linger *)->l_linger;
1260: /* fall thru... */
1261:
1262: case SO_DEBUG:
1263: case SO_KEEPALIVE:
1264: case SO_DONTROUTE:
1265: case SO_USELOOPBACK:
1266: case SO_BROADCAST:
1267: case SO_REUSEADDR:
1.15 mycroft 1268: case SO_REUSEPORT:
1.1 cgd 1269: case SO_OOBINLINE:
1.26 thorpej 1270: case SO_TIMESTAMP:
1.36 perry 1271: if (m == NULL || m->m_len < sizeof(int)) {
1.1 cgd 1272: error = EINVAL;
1273: goto bad;
1274: }
1275: if (*mtod(m, int *))
1276: so->so_options |= optname;
1277: else
1278: so->so_options &= ~optname;
1279: break;
1280:
1281: case SO_SNDBUF:
1282: case SO_RCVBUF:
1283: case SO_SNDLOWAT:
1284: case SO_RCVLOWAT:
1.28 thorpej 1285: {
1286: int optval;
1287:
1.36 perry 1288: if (m == NULL || m->m_len < sizeof(int)) {
1.1 cgd 1289: error = EINVAL;
1290: goto bad;
1291: }
1.28 thorpej 1292:
1293: /*
1294: * Values < 1 make no sense for any of these
1295: * options, so disallow them.
1296: */
1297: optval = *mtod(m, int *);
1298: if (optval < 1) {
1299: error = EINVAL;
1300: goto bad;
1301: }
1302:
1.1 cgd 1303: switch (optname) {
1304:
1305: case SO_SNDBUF:
1306: case SO_RCVBUF:
1307: if (sbreserve(optname == SO_SNDBUF ?
1308: &so->so_snd : &so->so_rcv,
1.28 thorpej 1309: (u_long) optval) == 0) {
1.1 cgd 1310: error = ENOBUFS;
1311: goto bad;
1312: }
1313: break;
1314:
1.28 thorpej 1315: /*
1316: * Make sure the low-water is never greater than
1317: * the high-water.
1318: */
1.1 cgd 1319: case SO_SNDLOWAT:
1.28 thorpej 1320: so->so_snd.sb_lowat =
1321: (optval > so->so_snd.sb_hiwat) ?
1322: so->so_snd.sb_hiwat : optval;
1.1 cgd 1323: break;
1324: case SO_RCVLOWAT:
1.28 thorpej 1325: so->so_rcv.sb_lowat =
1326: (optval > so->so_rcv.sb_hiwat) ?
1327: so->so_rcv.sb_hiwat : optval;
1.1 cgd 1328: break;
1329: }
1330: break;
1.28 thorpej 1331: }
1.1 cgd 1332:
1333: case SO_SNDTIMEO:
1334: case SO_RCVTIMEO:
1335: {
1336: struct timeval *tv;
1337: short val;
1338:
1.36 perry 1339: if (m == NULL || m->m_len < sizeof(*tv)) {
1.1 cgd 1340: error = EINVAL;
1341: goto bad;
1342: }
1343: tv = mtod(m, struct timeval *);
1.19 cgd 1344: if (tv->tv_sec * hz + tv->tv_usec / tick > SHRT_MAX) {
1.1 cgd 1345: error = EDOM;
1346: goto bad;
1347: }
1348: val = tv->tv_sec * hz + tv->tv_usec / tick;
1349:
1350: switch (optname) {
1351:
1352: case SO_SNDTIMEO:
1353: so->so_snd.sb_timeo = val;
1354: break;
1355: case SO_RCVTIMEO:
1356: so->so_rcv.sb_timeo = val;
1357: break;
1358: }
1359: break;
1360: }
1361:
1362: default:
1363: error = ENOPROTOOPT;
1364: break;
1365: }
1.15 mycroft 1366: if (error == 0 && so->so_proto && so->so_proto->pr_ctloutput) {
1367: (void) ((*so->so_proto->pr_ctloutput)
1368: (PRCO_SETOPT, so, level, optname, &m0));
1369: m = NULL; /* freed by protocol */
1370: }
1.1 cgd 1371: }
1.54 lukem 1372: bad:
1.1 cgd 1373: if (m)
1374: (void) m_free(m);
1375: return (error);
1376: }
1377:
1.14 mycroft 1378: int
1.54 lukem 1379: sogetopt(struct socket *so, int level, int optname, struct mbuf **mp)
1.1 cgd 1380: {
1.54 lukem 1381: struct mbuf *m;
1.1 cgd 1382:
1383: if (level != SOL_SOCKET) {
1384: if (so->so_proto && so->so_proto->pr_ctloutput) {
1385: return ((*so->so_proto->pr_ctloutput)
1386: (PRCO_GETOPT, so, level, optname, mp));
1387: } else
1388: return (ENOPROTOOPT);
1389: } else {
1390: m = m_get(M_WAIT, MT_SOOPTS);
1.36 perry 1391: m->m_len = sizeof(int);
1.1 cgd 1392:
1393: switch (optname) {
1394:
1395: case SO_LINGER:
1.36 perry 1396: m->m_len = sizeof(struct linger);
1.1 cgd 1397: mtod(m, struct linger *)->l_onoff =
1398: so->so_options & SO_LINGER;
1399: mtod(m, struct linger *)->l_linger = so->so_linger;
1400: break;
1401:
1402: case SO_USELOOPBACK:
1403: case SO_DONTROUTE:
1404: case SO_DEBUG:
1405: case SO_KEEPALIVE:
1406: case SO_REUSEADDR:
1.15 mycroft 1407: case SO_REUSEPORT:
1.1 cgd 1408: case SO_BROADCAST:
1409: case SO_OOBINLINE:
1.26 thorpej 1410: case SO_TIMESTAMP:
1.1 cgd 1411: *mtod(m, int *) = so->so_options & optname;
1412: break;
1413:
1414: case SO_TYPE:
1415: *mtod(m, int *) = so->so_type;
1416: break;
1417:
1418: case SO_ERROR:
1419: *mtod(m, int *) = so->so_error;
1420: so->so_error = 0;
1421: break;
1422:
1423: case SO_SNDBUF:
1424: *mtod(m, int *) = so->so_snd.sb_hiwat;
1425: break;
1426:
1427: case SO_RCVBUF:
1428: *mtod(m, int *) = so->so_rcv.sb_hiwat;
1429: break;
1430:
1431: case SO_SNDLOWAT:
1432: *mtod(m, int *) = so->so_snd.sb_lowat;
1433: break;
1434:
1435: case SO_RCVLOWAT:
1436: *mtod(m, int *) = so->so_rcv.sb_lowat;
1437: break;
1438:
1439: case SO_SNDTIMEO:
1440: case SO_RCVTIMEO:
1441: {
1442: int val = (optname == SO_SNDTIMEO ?
1443: so->so_snd.sb_timeo : so->so_rcv.sb_timeo);
1444:
1445: m->m_len = sizeof(struct timeval);
1446: mtod(m, struct timeval *)->tv_sec = val / hz;
1447: mtod(m, struct timeval *)->tv_usec =
1.27 kleink 1448: (val % hz) * tick;
1.1 cgd 1449: break;
1450: }
1451:
1452: default:
1453: (void)m_free(m);
1454: return (ENOPROTOOPT);
1455: }
1456: *mp = m;
1457: return (0);
1458: }
1459: }
1460:
1.14 mycroft 1461: void
1.54 lukem 1462: sohasoutofband(struct socket *so)
1.1 cgd 1463: {
1464: struct proc *p;
1465:
1466: if (so->so_pgid < 0)
1467: gsignal(-so->so_pgid, SIGURG);
1468: else if (so->so_pgid > 0 && (p = pfind(so->so_pgid)) != 0)
1469: psignal(p, SIGURG);
1.2 cgd 1470: selwakeup(&so->so_rcv.sb_sel);
1.1 cgd 1471: }
1.54.2.17! nathanw 1472:
! 1473: static void
! 1474: filt_sordetach(struct knote *kn)
! 1475: {
! 1476: struct socket *so;
! 1477:
! 1478: so = (struct socket *)kn->kn_fp->f_data;
! 1479: SLIST_REMOVE(&so->so_rcv.sb_sel.si_klist, kn, knote, kn_selnext);
! 1480: if (SLIST_EMPTY(&so->so_rcv.sb_sel.si_klist))
! 1481: so->so_rcv.sb_flags &= ~SB_KNOTE;
! 1482: }
! 1483:
! 1484: /*ARGSUSED*/
! 1485: static int
! 1486: filt_soread(struct knote *kn, long hint)
! 1487: {
! 1488: struct socket *so;
! 1489:
! 1490: so = (struct socket *)kn->kn_fp->f_data;
! 1491: kn->kn_data = so->so_rcv.sb_cc;
! 1492: if (so->so_state & SS_CANTRCVMORE) {
! 1493: kn->kn_flags |= EV_EOF;
! 1494: kn->kn_fflags = so->so_error;
! 1495: return (1);
! 1496: }
! 1497: if (so->so_error) /* temporary udp error */
! 1498: return (1);
! 1499: if (kn->kn_sfflags & NOTE_LOWAT)
! 1500: return (kn->kn_data >= kn->kn_sdata);
! 1501: return (kn->kn_data >= so->so_rcv.sb_lowat);
! 1502: }
! 1503:
! 1504: static void
! 1505: filt_sowdetach(struct knote *kn)
! 1506: {
! 1507: struct socket *so;
! 1508:
! 1509: so = (struct socket *)kn->kn_fp->f_data;
! 1510: SLIST_REMOVE(&so->so_snd.sb_sel.si_klist, kn, knote, kn_selnext);
! 1511: if (SLIST_EMPTY(&so->so_snd.sb_sel.si_klist))
! 1512: so->so_snd.sb_flags &= ~SB_KNOTE;
! 1513: }
! 1514:
! 1515: /*ARGSUSED*/
! 1516: static int
! 1517: filt_sowrite(struct knote *kn, long hint)
! 1518: {
! 1519: struct socket *so;
! 1520:
! 1521: so = (struct socket *)kn->kn_fp->f_data;
! 1522: kn->kn_data = sbspace(&so->so_snd);
! 1523: if (so->so_state & SS_CANTSENDMORE) {
! 1524: kn->kn_flags |= EV_EOF;
! 1525: kn->kn_fflags = so->so_error;
! 1526: return (1);
! 1527: }
! 1528: if (so->so_error) /* temporary udp error */
! 1529: return (1);
! 1530: if (((so->so_state & SS_ISCONNECTED) == 0) &&
! 1531: (so->so_proto->pr_flags & PR_CONNREQUIRED))
! 1532: return (0);
! 1533: if (kn->kn_sfflags & NOTE_LOWAT)
! 1534: return (kn->kn_data >= kn->kn_sdata);
! 1535: return (kn->kn_data >= so->so_snd.sb_lowat);
! 1536: }
! 1537:
! 1538: /*ARGSUSED*/
! 1539: static int
! 1540: filt_solisten(struct knote *kn, long hint)
! 1541: {
! 1542: struct socket *so;
! 1543:
! 1544: so = (struct socket *)kn->kn_fp->f_data;
! 1545:
! 1546: /*
! 1547: * Set kn_data to number of incoming connections, not
! 1548: * counting partial (incomplete) connections.
! 1549: */
! 1550: kn->kn_data = so->so_qlen;
! 1551: return (kn->kn_data > 0);
! 1552: }
! 1553:
! 1554: static const struct filterops solisten_filtops =
! 1555: { 1, NULL, filt_sordetach, filt_solisten };
! 1556: static const struct filterops soread_filtops =
! 1557: { 1, NULL, filt_sordetach, filt_soread };
! 1558: static const struct filterops sowrite_filtops =
! 1559: { 1, NULL, filt_sowdetach, filt_sowrite };
! 1560:
! 1561: int
! 1562: soo_kqfilter(struct file *fp, struct knote *kn)
! 1563: {
! 1564: struct socket *so;
! 1565: struct sockbuf *sb;
! 1566:
! 1567: so = (struct socket *)kn->kn_fp->f_data;
! 1568: switch (kn->kn_filter) {
! 1569: case EVFILT_READ:
! 1570: if (so->so_options & SO_ACCEPTCONN)
! 1571: kn->kn_fop = &solisten_filtops;
! 1572: else
! 1573: kn->kn_fop = &soread_filtops;
! 1574: sb = &so->so_rcv;
! 1575: break;
! 1576: case EVFILT_WRITE:
! 1577: kn->kn_fop = &sowrite_filtops;
! 1578: sb = &so->so_snd;
! 1579: break;
! 1580: default:
! 1581: return (1);
! 1582: }
! 1583: SLIST_INSERT_HEAD(&sb->sb_sel.si_klist, kn, kn_selnext);
! 1584: sb->sb_flags |= SB_KNOTE;
! 1585: return (0);
! 1586: }
! 1587:
CVSweb <webmaster@jp.NetBSD.org>