Annotation of src/sys/kern/uipc_socket.c, Revision 1.79
1.79 ! thorpej 1: /* $NetBSD: uipc_socket.c,v 1.78 2003/02/26 06:31:11 matt 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>
1.79 ! thorpej 75: __KERNEL_RCSID(0, "$NetBSD: uipc_socket.c,v 1.78 2003/02/26 06:31:11 matt 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
1.79 ! thorpej 157: sodoloanfree(struct vm_page **pgs, caddr_t buf, size_t size)
1.64 thorpej 158: {
159: vaddr_t va, sva, eva;
160: vsize_t len;
161: paddr_t pa;
162: int i, npgs;
163:
164: eva = round_page((vaddr_t) buf + size);
165: sva = trunc_page((vaddr_t) buf);
166: len = eva - sva;
167: npgs = len >> PAGE_SHIFT;
168:
1.79 ! thorpej 169: if (__predict_false(pgs == NULL)) {
! 170: pgs = alloca(npgs * sizeof(*pgs));
1.64 thorpej 171:
1.79 ! thorpej 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: }
1.64 thorpej 177: }
178:
179: pmap_kremove(sva, len);
180: pmap_update(pmap_kernel());
181: uvm_unloan(pgs, npgs, UVM_LOAN_TOPAGE);
182: uvm_km_free(kernel_map, sva, len);
183: socurkva -= len;
184: if (sokvawaiters)
185: wakeup(&socurkva);
186: }
187:
188: static size_t
189: sodopendfree(struct socket *so)
190: {
191: struct mbuf *m;
192: size_t rv = 0;
193: int s;
194:
195: s = splvm();
196:
197: for (;;) {
198: m = so_pendfree;
199: if (m == NULL)
200: break;
201: so_pendfree = m->m_next;
202: splx(s);
203:
204: rv += m->m_ext.ext_size;
1.79 ! thorpej 205: sodoloanfree((m->m_flags & M_EXT_PAGES) ?
! 206: m->m_ext.ext_pgs : NULL, m->m_ext.ext_buf,
! 207: m->m_ext.ext_size);
1.64 thorpej 208: s = splvm();
209: pool_cache_put(&mbpool_cache, m);
210: }
211:
212: for (;;) {
213: m = so->so_pendfree;
214: if (m == NULL)
215: break;
216: so->so_pendfree = m->m_next;
217: splx(s);
218:
219: rv += m->m_ext.ext_size;
1.79 ! thorpej 220: sodoloanfree((m->m_flags & M_EXT_PAGES) ?
! 221: m->m_ext.ext_pgs : NULL, m->m_ext.ext_buf,
! 222: m->m_ext.ext_size);
1.64 thorpej 223: s = splvm();
224: pool_cache_put(&mbpool_cache, m);
225: }
226:
227: splx(s);
228: return (rv);
229: }
230:
231: static void
1.76 thorpej 232: soloanfree(struct mbuf *m, caddr_t buf, size_t size, void *arg)
1.64 thorpej 233: {
234: struct socket *so = arg;
235: int s;
236:
237: if (m == NULL) {
1.79 ! thorpej 238: sodoloanfree(NULL, buf, size);
1.64 thorpej 239: return;
240: }
241:
242: s = splvm();
243: m->m_next = so->so_pendfree;
244: so->so_pendfree = m;
245: splx(s);
246: if (sokvawaiters)
247: wakeup(&socurkva);
248: }
249:
250: static long
251: sosend_loan(struct socket *so, struct uio *uio, struct mbuf *m, long space)
252: {
253: struct iovec *iov = uio->uio_iov;
254: vaddr_t sva, eva;
255: vsize_t len;
256: vaddr_t lva, va;
257: int npgs, s, i, error;
258:
259: if (uio->uio_segflg != UIO_USERSPACE)
260: return (0);
261:
262: if (iov->iov_len < (size_t) space)
263: space = iov->iov_len;
264: if (space > SOCK_LOAN_CHUNK)
265: space = SOCK_LOAN_CHUNK;
266:
267: eva = round_page((vaddr_t) iov->iov_base + space);
268: sva = trunc_page((vaddr_t) iov->iov_base);
269: len = eva - sva;
270: npgs = len >> PAGE_SHIFT;
271:
1.79 ! thorpej 272: /* XXX KDASSERT */
! 273: KASSERT(npgs <= M_EXT_MAXPAGES);
! 274:
1.64 thorpej 275: while (socurkva + len > somaxkva) {
276: if (sodopendfree(so))
277: continue;
278: SOSEND_COUNTER_INCR(&sosend_kvalimit);
279: s = splvm();
280: sokvawaiters++;
281: (void) tsleep(&socurkva, PVM, "sokva", 0);
282: sokvawaiters--;
283: splx(s);
284: }
285:
286: lva = uvm_km_valloc_wait(kernel_map, len);
287: if (lva == 0)
288: return (0);
289: socurkva += len;
290:
291: error = uvm_loan(&uio->uio_procp->p_vmspace->vm_map, sva, len,
1.79 ! thorpej 292: m->m_ext.ext_pgs, UVM_LOAN_TOPAGE);
1.64 thorpej 293: if (error) {
294: uvm_km_free(kernel_map, lva, len);
295: socurkva -= len;
296: return (0);
297: }
298:
299: for (i = 0, va = lva; i < npgs; i++, va += PAGE_SIZE)
1.79 ! thorpej 300: pmap_kenter_pa(va, VM_PAGE_TO_PHYS(m->m_ext.ext_pgs[i]),
! 301: VM_PROT_READ);
1.64 thorpej 302: pmap_update(pmap_kernel());
303:
304: lva += (vaddr_t) iov->iov_base & PAGE_MASK;
305:
306: MEXTADD(m, (caddr_t) lva, space, M_MBUF, soloanfree, so);
1.79 ! thorpej 307: m->m_flags |= M_EXT_PAGES | M_EXT_ROMAP;
1.64 thorpej 308:
309: uio->uio_resid -= space;
310: /* uio_offset not updated, not set/used for write(2) */
311: uio->uio_iov->iov_base = (caddr_t) uio->uio_iov->iov_base + space;
312: uio->uio_iov->iov_len -= space;
313: if (uio->uio_iov->iov_len == 0) {
314: uio->uio_iov++;
315: uio->uio_iovcnt--;
316: }
317:
318: return (space);
319: }
320:
1.1 cgd 321: /*
322: * Socket operation routines.
323: * These routines are called by the routines in
324: * sys_socket.c or from a system process, and
325: * implement the semantics of socket operations by
326: * switching out to the protocol specific routines.
327: */
328: /*ARGSUSED*/
1.3 andrew 329: int
1.54 lukem 330: socreate(int dom, struct socket **aso, int type, int proto)
1.1 cgd 331: {
1.54 lukem 332: struct proc *p;
333: struct protosw *prp;
334: struct socket *so;
335: int error, s;
1.1 cgd 336:
1.54 lukem 337: p = curproc; /* XXX */
1.1 cgd 338: if (proto)
339: prp = pffindproto(dom, proto, type);
340: else
341: prp = pffindtype(dom, type);
1.15 mycroft 342: if (prp == 0 || prp->pr_usrreq == 0)
1.1 cgd 343: return (EPROTONOSUPPORT);
344: if (prp->pr_type != type)
345: return (EPROTOTYPE);
1.39 matt 346: s = splsoftnet();
1.37 thorpej 347: so = pool_get(&socket_pool, PR_WAITOK);
1.38 perry 348: memset((caddr_t)so, 0, sizeof(*so));
1.31 thorpej 349: TAILQ_INIT(&so->so_q0);
350: TAILQ_INIT(&so->so_q);
1.1 cgd 351: so->so_type = type;
352: so->so_proto = prp;
1.33 matt 353: so->so_send = sosend;
354: so->so_receive = soreceive;
1.78 matt 355: #ifdef MBUFTRACE
356: so->so_rcv.sb_mowner = &prp->pr_domain->dom_mowner;
357: so->so_snd.sb_mowner = &prp->pr_domain->dom_mowner;
358: so->so_mowner = &prp->pr_domain->dom_mowner;
359: #endif
1.44 lukem 360: if (p != 0)
361: so->so_uid = p->p_ucred->cr_uid;
1.22 mycroft 362: error = (*prp->pr_usrreq)(so, PRU_ATTACH, (struct mbuf *)0,
363: (struct mbuf *)(long)proto, (struct mbuf *)0, p);
1.1 cgd 364: if (error) {
365: so->so_state |= SS_NOFDREF;
366: sofree(so);
1.39 matt 367: splx(s);
1.1 cgd 368: return (error);
369: }
1.39 matt 370: splx(s);
1.1 cgd 371: *aso = so;
372: return (0);
373: }
374:
1.3 andrew 375: int
1.54 lukem 376: sobind(struct socket *so, struct mbuf *nam, struct proc *p)
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_BIND, (struct mbuf *)0,
382: nam, (struct mbuf *)0, p);
1.1 cgd 383: splx(s);
384: return (error);
385: }
386:
1.3 andrew 387: int
1.54 lukem 388: solisten(struct socket *so, int backlog)
1.1 cgd 389: {
1.54 lukem 390: int s, error;
1.1 cgd 391:
1.54 lukem 392: s = splsoftnet();
1.22 mycroft 393: error = (*so->so_proto->pr_usrreq)(so, PRU_LISTEN, (struct mbuf *)0,
394: (struct mbuf *)0, (struct mbuf *)0, (struct proc *)0);
1.1 cgd 395: if (error) {
396: splx(s);
397: return (error);
398: }
1.63 matt 399: if (TAILQ_EMPTY(&so->so_q))
1.1 cgd 400: so->so_options |= SO_ACCEPTCONN;
401: if (backlog < 0)
402: backlog = 0;
1.49 jonathan 403: so->so_qlimit = min(backlog, somaxconn);
1.1 cgd 404: splx(s);
405: return (0);
406: }
407:
1.21 christos 408: void
1.54 lukem 409: sofree(struct socket *so)
1.1 cgd 410: {
1.64 thorpej 411: struct mbuf *m;
1.1 cgd 412:
1.43 mycroft 413: if (so->so_pcb || (so->so_state & SS_NOFDREF) == 0)
1.1 cgd 414: return;
1.43 mycroft 415: if (so->so_head) {
416: /*
417: * We must not decommission a socket that's on the accept(2)
418: * queue. If we do, then accept(2) may hang after select(2)
419: * indicated that the listening socket was ready.
420: */
421: if (!soqremque(so, 0))
422: return;
423: }
1.1 cgd 424: sbrelease(&so->so_snd);
425: sorflush(so);
1.64 thorpej 426: while ((m = so->so_pendfree) != NULL) {
427: so->so_pendfree = m->m_next;
428: m->m_next = so_pendfree;
429: so_pendfree = m;
430: }
1.37 thorpej 431: pool_put(&socket_pool, so);
1.1 cgd 432: }
433:
434: /*
435: * Close a socket on last file table reference removal.
436: * Initiate disconnect if connected.
437: * Free socket when disconnect complete.
438: */
1.3 andrew 439: int
1.54 lukem 440: soclose(struct socket *so)
1.1 cgd 441: {
1.54 lukem 442: struct socket *so2;
443: int s, error;
1.1 cgd 444:
1.54 lukem 445: error = 0;
446: s = splsoftnet(); /* conservative */
1.1 cgd 447: if (so->so_options & SO_ACCEPTCONN) {
1.63 matt 448: while ((so2 = TAILQ_FIRST(&so->so_q0)) != 0) {
1.42 mycroft 449: (void) soqremque(so2, 0);
1.41 mycroft 450: (void) soabort(so2);
451: }
1.63 matt 452: while ((so2 = TAILQ_FIRST(&so->so_q)) != 0) {
1.42 mycroft 453: (void) soqremque(so2, 1);
1.41 mycroft 454: (void) soabort(so2);
455: }
1.1 cgd 456: }
457: if (so->so_pcb == 0)
458: goto discard;
459: if (so->so_state & SS_ISCONNECTED) {
460: if ((so->so_state & SS_ISDISCONNECTING) == 0) {
461: error = sodisconnect(so);
462: if (error)
463: goto drop;
464: }
465: if (so->so_options & SO_LINGER) {
466: if ((so->so_state & SS_ISDISCONNECTING) &&
467: (so->so_state & SS_NBIO))
468: goto drop;
1.21 christos 469: while (so->so_state & SS_ISCONNECTED) {
470: error = tsleep((caddr_t)&so->so_timeo,
471: PSOCK | PCATCH, netcls,
1.30 thorpej 472: so->so_linger * hz);
1.21 christos 473: if (error)
1.1 cgd 474: break;
1.21 christos 475: }
1.1 cgd 476: }
477: }
1.54 lukem 478: drop:
1.1 cgd 479: if (so->so_pcb) {
1.22 mycroft 480: int error2 = (*so->so_proto->pr_usrreq)(so, PRU_DETACH,
481: (struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0,
482: (struct proc *)0);
1.1 cgd 483: if (error == 0)
484: error = error2;
485: }
1.54 lukem 486: discard:
1.1 cgd 487: if (so->so_state & SS_NOFDREF)
488: panic("soclose: NOFDREF");
489: so->so_state |= SS_NOFDREF;
490: sofree(so);
491: splx(s);
492: return (error);
493: }
494:
495: /*
1.20 mycroft 496: * Must be called at splsoftnet...
1.1 cgd 497: */
1.3 andrew 498: int
1.54 lukem 499: soabort(struct socket *so)
1.1 cgd 500: {
501:
1.22 mycroft 502: return (*so->so_proto->pr_usrreq)(so, PRU_ABORT, (struct mbuf *)0,
503: (struct mbuf *)0, (struct mbuf *)0, (struct proc *)0);
1.1 cgd 504: }
505:
1.3 andrew 506: int
1.54 lukem 507: soaccept(struct socket *so, struct mbuf *nam)
1.1 cgd 508: {
1.54 lukem 509: int s, error;
1.1 cgd 510:
1.54 lukem 511: error = 0;
512: s = splsoftnet();
1.1 cgd 513: if ((so->so_state & SS_NOFDREF) == 0)
514: panic("soaccept: !NOFDREF");
515: so->so_state &= ~SS_NOFDREF;
1.55 thorpej 516: if ((so->so_state & SS_ISDISCONNECTED) == 0 ||
517: (so->so_proto->pr_flags & PR_ABRTACPTDIS) == 0)
1.41 mycroft 518: error = (*so->so_proto->pr_usrreq)(so, PRU_ACCEPT,
519: (struct mbuf *)0, nam, (struct mbuf *)0, (struct proc *)0);
520: else
1.53 itojun 521: error = ECONNABORTED;
1.52 itojun 522:
1.1 cgd 523: splx(s);
524: return (error);
525: }
526:
1.3 andrew 527: int
1.54 lukem 528: soconnect(struct socket *so, struct mbuf *nam)
1.1 cgd 529: {
1.54 lukem 530: struct proc *p;
531: int s, error;
1.1 cgd 532:
1.54 lukem 533: p = curproc; /* XXX */
1.1 cgd 534: if (so->so_options & SO_ACCEPTCONN)
535: return (EOPNOTSUPP);
1.20 mycroft 536: s = splsoftnet();
1.1 cgd 537: /*
538: * If protocol is connection-based, can only connect once.
539: * Otherwise, if connected, try to disconnect first.
540: * This allows user to disconnect by connecting to, e.g.,
541: * a null address.
542: */
543: if (so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING) &&
544: ((so->so_proto->pr_flags & PR_CONNREQUIRED) ||
545: (error = sodisconnect(so))))
546: error = EISCONN;
547: else
548: error = (*so->so_proto->pr_usrreq)(so, PRU_CONNECT,
1.23 mycroft 549: (struct mbuf *)0, nam, (struct mbuf *)0, p);
1.1 cgd 550: splx(s);
551: return (error);
552: }
553:
1.3 andrew 554: int
1.54 lukem 555: soconnect2(struct socket *so1, struct socket *so2)
1.1 cgd 556: {
1.54 lukem 557: int s, error;
1.1 cgd 558:
1.54 lukem 559: s = splsoftnet();
1.22 mycroft 560: error = (*so1->so_proto->pr_usrreq)(so1, PRU_CONNECT2,
561: (struct mbuf *)0, (struct mbuf *)so2, (struct mbuf *)0,
562: (struct proc *)0);
1.1 cgd 563: splx(s);
564: return (error);
565: }
566:
1.3 andrew 567: int
1.54 lukem 568: sodisconnect(struct socket *so)
1.1 cgd 569: {
1.54 lukem 570: int s, error;
1.1 cgd 571:
1.54 lukem 572: s = splsoftnet();
1.1 cgd 573: if ((so->so_state & SS_ISCONNECTED) == 0) {
574: error = ENOTCONN;
575: goto bad;
576: }
577: if (so->so_state & SS_ISDISCONNECTING) {
578: error = EALREADY;
579: goto bad;
580: }
1.22 mycroft 581: error = (*so->so_proto->pr_usrreq)(so, PRU_DISCONNECT,
582: (struct mbuf *)0, (struct mbuf *)0, (struct mbuf *)0,
583: (struct proc *)0);
1.54 lukem 584: bad:
1.1 cgd 585: splx(s);
1.64 thorpej 586: sodopendfree(so);
1.1 cgd 587: return (error);
588: }
589:
1.15 mycroft 590: #define SBLOCKWAIT(f) (((f) & MSG_DONTWAIT) ? M_NOWAIT : M_WAITOK)
1.1 cgd 591: /*
592: * Send on a socket.
593: * If send must go all at once and message is larger than
594: * send buffering, then hard error.
595: * Lock against other senders.
596: * If must go all at once and not enough room now, then
597: * inform user that this would block and do nothing.
598: * Otherwise, if nonblocking, send as much as possible.
599: * The data to be sent is described by "uio" if nonzero,
600: * otherwise by the mbuf chain "top" (which must be null
601: * if uio is not). Data provided in mbuf chain must be small
602: * enough to send all at once.
603: *
604: * Returns nonzero on error, timeout or signal; callers
605: * must check for short counts if EINTR/ERESTART are returned.
606: * Data and control buffers are freed on return.
607: */
1.3 andrew 608: int
1.54 lukem 609: sosend(struct socket *so, struct mbuf *addr, struct uio *uio, struct mbuf *top,
610: struct mbuf *control, int flags)
1.1 cgd 611: {
1.54 lukem 612: struct proc *p;
613: struct mbuf **mp, *m;
1.58 jdolecek 614: long space, len, resid, clen, mlen;
615: int error, s, dontroute, atomic;
1.54 lukem 616:
1.64 thorpej 617: sodopendfree(so);
618:
1.54 lukem 619: p = curproc; /* XXX */
620: clen = 0;
621: atomic = sosendallatonce(so) || top;
1.1 cgd 622: if (uio)
623: resid = uio->uio_resid;
624: else
625: resid = top->m_pkthdr.len;
1.7 cgd 626: /*
627: * In theory resid should be unsigned.
628: * However, space must be signed, as it might be less than 0
629: * if we over-committed, and we must use a signed comparison
630: * of space and resid. On the other hand, a negative resid
631: * causes us to loop sending 0-length segments to the protocol.
632: */
1.29 mycroft 633: if (resid < 0) {
634: error = EINVAL;
635: goto out;
636: }
1.1 cgd 637: dontroute =
638: (flags & MSG_DONTROUTE) && (so->so_options & SO_DONTROUTE) == 0 &&
639: (so->so_proto->pr_flags & PR_ATOMIC);
1.12 mycroft 640: p->p_stats->p_ru.ru_msgsnd++;
1.1 cgd 641: if (control)
642: clen = control->m_len;
643: #define snderr(errno) { error = errno; splx(s); goto release; }
644:
1.54 lukem 645: restart:
1.21 christos 646: if ((error = sblock(&so->so_snd, SBLOCKWAIT(flags))) != 0)
1.1 cgd 647: goto out;
648: do {
1.20 mycroft 649: s = splsoftnet();
1.1 cgd 650: if (so->so_state & SS_CANTSENDMORE)
651: snderr(EPIPE);
1.48 thorpej 652: if (so->so_error) {
653: error = so->so_error;
654: so->so_error = 0;
655: splx(s);
656: goto release;
657: }
1.1 cgd 658: if ((so->so_state & SS_ISCONNECTED) == 0) {
659: if (so->so_proto->pr_flags & PR_CONNREQUIRED) {
660: if ((so->so_state & SS_ISCONFIRMING) == 0 &&
661: !(resid == 0 && clen != 0))
662: snderr(ENOTCONN);
663: } else if (addr == 0)
664: snderr(EDESTADDRREQ);
665: }
666: space = sbspace(&so->so_snd);
667: if (flags & MSG_OOB)
668: space += 1024;
1.21 christos 669: if ((atomic && resid > so->so_snd.sb_hiwat) ||
1.11 mycroft 670: clen > so->so_snd.sb_hiwat)
671: snderr(EMSGSIZE);
672: if (space < resid + clen && uio &&
1.1 cgd 673: (atomic || space < so->so_snd.sb_lowat || space < clen)) {
674: if (so->so_state & SS_NBIO)
675: snderr(EWOULDBLOCK);
676: sbunlock(&so->so_snd);
677: error = sbwait(&so->so_snd);
678: splx(s);
679: if (error)
680: goto out;
681: goto restart;
682: }
683: splx(s);
684: mp = ⊤
685: space -= clen;
686: do {
1.45 tv 687: if (uio == NULL) {
688: /*
689: * Data is prepackaged in "top".
690: */
691: resid = 0;
692: if (flags & MSG_EOR)
693: top->m_flags |= M_EOR;
694: } else do {
695: if (top == 0) {
1.78 matt 696: m = m_gethdr(M_WAIT, MT_DATA);
1.45 tv 697: mlen = MHLEN;
698: m->m_pkthdr.len = 0;
699: m->m_pkthdr.rcvif = (struct ifnet *)0;
700: } else {
1.78 matt 701: m = m_get(M_WAIT, MT_DATA);
1.45 tv 702: mlen = MLEN;
703: }
1.78 matt 704: MCLAIM(m, so->so_snd.sb_mowner);
1.65 thorpej 705: if (use_sosend_loan &&
706: uio->uio_iov->iov_len >= SOCK_LOAN_THRESH &&
1.64 thorpej 707: space >= SOCK_LOAN_THRESH &&
708: (len = sosend_loan(so, uio, m,
709: space)) != 0) {
710: SOSEND_COUNTER_INCR(&sosend_loan_big);
711: space -= len;
712: goto have_data;
713: }
1.45 tv 714: if (resid >= MINCLSIZE && space >= MCLBYTES) {
1.64 thorpej 715: SOSEND_COUNTER_INCR(&sosend_copy_big);
1.78 matt 716: m_clget(m, M_WAIT);
1.45 tv 717: if ((m->m_flags & M_EXT) == 0)
718: goto nopages;
719: mlen = MCLBYTES;
720: if (atomic && top == 0) {
1.58 jdolecek 721: len = lmin(MCLBYTES - max_hdr,
1.54 lukem 722: resid);
1.45 tv 723: m->m_data += max_hdr;
724: } else
1.58 jdolecek 725: len = lmin(MCLBYTES, resid);
1.45 tv 726: space -= len;
727: } else {
1.64 thorpej 728: nopages:
729: SOSEND_COUNTER_INCR(&sosend_copy_small);
1.58 jdolecek 730: len = lmin(lmin(mlen, resid), space);
1.45 tv 731: space -= len;
732: /*
733: * For datagram protocols, leave room
734: * for protocol headers in first mbuf.
735: */
736: if (atomic && top == 0 && len < mlen)
737: MH_ALIGN(m, len);
738: }
1.54 lukem 739: error = uiomove(mtod(m, caddr_t), (int)len,
740: uio);
1.64 thorpej 741: have_data:
1.45 tv 742: resid = uio->uio_resid;
743: m->m_len = len;
744: *mp = m;
745: top->m_pkthdr.len += len;
746: if (error)
747: goto release;
748: mp = &m->m_next;
749: if (resid <= 0) {
750: if (flags & MSG_EOR)
751: top->m_flags |= M_EOR;
752: break;
753: }
754: } while (space > 0 && atomic);
1.46 sommerfe 755:
756: s = splsoftnet();
757:
758: if (so->so_state & SS_CANTSENDMORE)
759: snderr(EPIPE);
1.45 tv 760:
761: if (dontroute)
762: so->so_options |= SO_DONTROUTE;
763: if (resid > 0)
764: so->so_state |= SS_MORETOCOME;
1.46 sommerfe 765: error = (*so->so_proto->pr_usrreq)(so,
766: (flags & MSG_OOB) ? PRU_SENDOOB : PRU_SEND,
767: top, addr, control, p);
1.45 tv 768: if (dontroute)
769: so->so_options &= ~SO_DONTROUTE;
770: if (resid > 0)
771: so->so_state &= ~SS_MORETOCOME;
1.46 sommerfe 772: splx(s);
773:
1.45 tv 774: clen = 0;
775: control = 0;
776: top = 0;
777: mp = ⊤
1.1 cgd 778: if (error)
779: goto release;
780: } while (resid && space > 0);
781: } while (resid);
782:
1.54 lukem 783: release:
1.1 cgd 784: sbunlock(&so->so_snd);
1.54 lukem 785: out:
1.1 cgd 786: if (top)
787: m_freem(top);
788: if (control)
789: m_freem(control);
790: return (error);
791: }
792:
793: /*
794: * Implement receive operations on a socket.
795: * We depend on the way that records are added to the sockbuf
796: * by sbappend*. In particular, each record (mbufs linked through m_next)
797: * must begin with an address if the protocol so specifies,
798: * followed by an optional mbuf or mbufs containing ancillary data,
799: * and then zero or more mbufs of data.
800: * In order to avoid blocking network interrupts for the entire time here,
801: * we splx() while doing the actual copy to user space.
802: * Although the sockbuf is locked, new data may still be appended,
803: * and thus we must maintain consistency of the sockbuf during that time.
804: *
805: * The caller may receive the data as a single mbuf chain by supplying
806: * an mbuf **mp0 for use in returning the chain. The uio is then used
807: * only for the count in uio_resid.
808: */
1.3 andrew 809: int
1.54 lukem 810: soreceive(struct socket *so, struct mbuf **paddr, struct uio *uio,
811: struct mbuf **mp0, struct mbuf **controlp, int *flagsp)
1.1 cgd 812: {
1.54 lukem 813: struct mbuf *m, **mp;
814: int flags, len, error, s, offset, moff, type, orig_resid;
815: struct protosw *pr;
816: struct mbuf *nextrecord;
1.67 he 817: int mbuf_removed = 0;
1.64 thorpej 818:
1.54 lukem 819: pr = so->so_proto;
1.1 cgd 820: mp = mp0;
1.54 lukem 821: type = 0;
822: orig_resid = uio->uio_resid;
1.1 cgd 823: if (paddr)
824: *paddr = 0;
825: if (controlp)
826: *controlp = 0;
827: if (flagsp)
828: flags = *flagsp &~ MSG_EOR;
829: else
830: flags = 0;
1.66 enami 831:
832: if ((flags & MSG_DONTWAIT) == 0)
833: sodopendfree(so);
834:
1.1 cgd 835: if (flags & MSG_OOB) {
836: m = m_get(M_WAIT, MT_DATA);
1.17 cgd 837: error = (*pr->pr_usrreq)(so, PRU_RCVOOB, m,
1.22 mycroft 838: (struct mbuf *)(long)(flags & MSG_PEEK), (struct mbuf *)0,
839: (struct proc *)0);
1.1 cgd 840: if (error)
841: goto bad;
842: do {
843: error = uiomove(mtod(m, caddr_t),
844: (int) min(uio->uio_resid, m->m_len), uio);
845: m = m_free(m);
846: } while (uio->uio_resid && error == 0 && m);
1.54 lukem 847: bad:
1.1 cgd 848: if (m)
849: m_freem(m);
850: return (error);
851: }
852: if (mp)
853: *mp = (struct mbuf *)0;
854: if (so->so_state & SS_ISCONFIRMING && uio->uio_resid)
1.22 mycroft 855: (*pr->pr_usrreq)(so, PRU_RCVD, (struct mbuf *)0,
856: (struct mbuf *)0, (struct mbuf *)0, (struct proc *)0);
1.1 cgd 857:
1.54 lukem 858: restart:
1.21 christos 859: if ((error = sblock(&so->so_rcv, SBLOCKWAIT(flags))) != 0)
1.1 cgd 860: return (error);
1.20 mycroft 861: s = splsoftnet();
1.1 cgd 862:
863: m = so->so_rcv.sb_mb;
864: /*
865: * If we have less data than requested, block awaiting more
866: * (subject to any timeout) if:
1.15 mycroft 867: * 1. the current count is less than the low water mark,
1.1 cgd 868: * 2. MSG_WAITALL is set, and it is possible to do the entire
1.15 mycroft 869: * receive operation at once if we block (resid <= hiwat), or
870: * 3. MSG_DONTWAIT is not set.
1.1 cgd 871: * If MSG_WAITALL is set but resid is larger than the receive buffer,
872: * we have to do the receive in sections, and thus risk returning
873: * a short count if a timeout or signal occurs after we start.
874: */
1.21 christos 875: if (m == 0 || (((flags & MSG_DONTWAIT) == 0 &&
1.15 mycroft 876: so->so_rcv.sb_cc < uio->uio_resid) &&
1.1 cgd 877: (so->so_rcv.sb_cc < so->so_rcv.sb_lowat ||
878: ((flags & MSG_WAITALL) && uio->uio_resid <= so->so_rcv.sb_hiwat)) &&
1.21 christos 879: m->m_nextpkt == 0 && (pr->pr_flags & PR_ATOMIC) == 0)) {
1.1 cgd 880: #ifdef DIAGNOSTIC
881: if (m == 0 && so->so_rcv.sb_cc)
882: panic("receive 1");
883: #endif
884: if (so->so_error) {
885: if (m)
1.15 mycroft 886: goto dontblock;
1.1 cgd 887: error = so->so_error;
888: if ((flags & MSG_PEEK) == 0)
889: so->so_error = 0;
890: goto release;
891: }
892: if (so->so_state & SS_CANTRCVMORE) {
893: if (m)
1.15 mycroft 894: goto dontblock;
1.1 cgd 895: else
896: goto release;
897: }
898: for (; m; m = m->m_next)
899: if (m->m_type == MT_OOBDATA || (m->m_flags & M_EOR)) {
900: m = so->so_rcv.sb_mb;
901: goto dontblock;
902: }
903: if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0 &&
904: (so->so_proto->pr_flags & PR_CONNREQUIRED)) {
905: error = ENOTCONN;
906: goto release;
907: }
908: if (uio->uio_resid == 0)
909: goto release;
1.15 mycroft 910: if ((so->so_state & SS_NBIO) || (flags & MSG_DONTWAIT)) {
1.1 cgd 911: error = EWOULDBLOCK;
912: goto release;
913: }
1.69 thorpej 914: SBLASTRECORDCHK(&so->so_rcv, "soreceive sbwait 1");
915: SBLASTMBUFCHK(&so->so_rcv, "soreceive sbwait 1");
1.1 cgd 916: sbunlock(&so->so_rcv);
917: error = sbwait(&so->so_rcv);
918: splx(s);
919: if (error)
920: return (error);
921: goto restart;
922: }
1.54 lukem 923: dontblock:
1.69 thorpej 924: /*
925: * On entry here, m points to the first record of the socket buffer.
926: * While we process the initial mbufs containing address and control
927: * info, we save a copy of m->m_nextpkt into nextrecord.
928: */
1.15 mycroft 929: #ifdef notyet /* XXXX */
930: if (uio->uio_procp)
931: uio->uio_procp->p_stats->p_ru.ru_msgrcv++;
932: #endif
1.69 thorpej 933: KASSERT(m == so->so_rcv.sb_mb);
934: SBLASTRECORDCHK(&so->so_rcv, "soreceive 1");
935: SBLASTMBUFCHK(&so->so_rcv, "soreceive 1");
1.1 cgd 936: nextrecord = m->m_nextpkt;
937: if (pr->pr_flags & PR_ADDR) {
938: #ifdef DIAGNOSTIC
939: if (m->m_type != MT_SONAME)
940: panic("receive 1a");
941: #endif
1.3 andrew 942: orig_resid = 0;
1.1 cgd 943: if (flags & MSG_PEEK) {
944: if (paddr)
945: *paddr = m_copy(m, 0, m->m_len);
946: m = m->m_next;
947: } else {
948: sbfree(&so->so_rcv, m);
1.67 he 949: mbuf_removed = 1;
1.1 cgd 950: if (paddr) {
951: *paddr = m;
952: so->so_rcv.sb_mb = m->m_next;
953: m->m_next = 0;
954: m = so->so_rcv.sb_mb;
955: } else {
956: MFREE(m, so->so_rcv.sb_mb);
957: m = so->so_rcv.sb_mb;
958: }
959: }
960: }
961: while (m && m->m_type == MT_CONTROL && error == 0) {
962: if (flags & MSG_PEEK) {
963: if (controlp)
964: *controlp = m_copy(m, 0, m->m_len);
965: m = m->m_next;
966: } else {
967: sbfree(&so->so_rcv, m);
1.67 he 968: mbuf_removed = 1;
1.1 cgd 969: if (controlp) {
970: if (pr->pr_domain->dom_externalize &&
971: mtod(m, struct cmsghdr *)->cmsg_type ==
972: SCM_RIGHTS)
1.45 tv 973: error = (*pr->pr_domain->dom_externalize)(m);
1.1 cgd 974: *controlp = m;
975: so->so_rcv.sb_mb = m->m_next;
976: m->m_next = 0;
977: m = so->so_rcv.sb_mb;
978: } else {
979: MFREE(m, so->so_rcv.sb_mb);
980: m = so->so_rcv.sb_mb;
981: }
982: }
1.3 andrew 983: if (controlp) {
984: orig_resid = 0;
1.1 cgd 985: controlp = &(*controlp)->m_next;
1.3 andrew 986: }
1.1 cgd 987: }
1.69 thorpej 988:
989: /*
990: * If m is non-NULL, we have some data to read. From now on,
991: * make sure to keep sb_lastrecord consistent when working on
992: * the last packet on the chain (nextrecord == NULL) and we
993: * change m->m_nextpkt.
994: */
1.1 cgd 995: if (m) {
1.69 thorpej 996: if ((flags & MSG_PEEK) == 0) {
1.1 cgd 997: m->m_nextpkt = nextrecord;
1.69 thorpej 998: /*
999: * If nextrecord == NULL (this is a single chain),
1000: * then sb_lastrecord may not be valid here if m
1001: * was changed earlier.
1002: */
1003: if (nextrecord == NULL) {
1004: KASSERT(so->so_rcv.sb_mb == m);
1005: so->so_rcv.sb_lastrecord = m;
1006: }
1007: }
1.1 cgd 1008: type = m->m_type;
1009: if (type == MT_OOBDATA)
1010: flags |= MSG_OOB;
1.69 thorpej 1011: } else {
1012: if ((flags & MSG_PEEK) == 0) {
1013: KASSERT(so->so_rcv.sb_mb == m);
1014: so->so_rcv.sb_mb = nextrecord;
1.70 thorpej 1015: SB_EMPTY_FIXUP(&so->so_rcv);
1.69 thorpej 1016: }
1.1 cgd 1017: }
1.69 thorpej 1018: SBLASTRECORDCHK(&so->so_rcv, "soreceive 2");
1019: SBLASTMBUFCHK(&so->so_rcv, "soreceive 2");
1020:
1.1 cgd 1021: moff = 0;
1022: offset = 0;
1023: while (m && uio->uio_resid > 0 && error == 0) {
1024: if (m->m_type == MT_OOBDATA) {
1025: if (type != MT_OOBDATA)
1026: break;
1027: } else if (type == MT_OOBDATA)
1028: break;
1029: #ifdef DIAGNOSTIC
1030: else if (m->m_type != MT_DATA && m->m_type != MT_HEADER)
1031: panic("receive 3");
1032: #endif
1033: so->so_state &= ~SS_RCVATMARK;
1034: len = uio->uio_resid;
1035: if (so->so_oobmark && len > so->so_oobmark - offset)
1036: len = so->so_oobmark - offset;
1037: if (len > m->m_len - moff)
1038: len = m->m_len - moff;
1039: /*
1040: * If mp is set, just pass back the mbufs.
1041: * Otherwise copy them out via the uio, then free.
1042: * Sockbuf must be consistent here (points to current mbuf,
1043: * it points to next record) when we drop priority;
1044: * we must note any additions to the sockbuf when we
1045: * block interrupts again.
1046: */
1047: if (mp == 0) {
1.69 thorpej 1048: SBLASTRECORDCHK(&so->so_rcv, "soreceive uiomove");
1049: SBLASTMBUFCHK(&so->so_rcv, "soreceive uiomove");
1.1 cgd 1050: splx(s);
1051: error = uiomove(mtod(m, caddr_t) + moff, (int)len, uio);
1.20 mycroft 1052: s = splsoftnet();
1.67 he 1053: if (error) {
1054: /*
1055: * If any part of the record has been removed
1056: * (such as the MT_SONAME mbuf, which will
1057: * happen when PR_ADDR, and thus also
1058: * PR_ATOMIC, is set), then drop the entire
1059: * record to maintain the atomicity of the
1060: * receive operation.
1061: *
1062: * This avoids a later panic("receive 1a")
1063: * when compiled with DIAGNOSTIC.
1064: */
1065: if (m && mbuf_removed
1066: && (pr->pr_flags & PR_ATOMIC))
1067: (void) sbdroprecord(&so->so_rcv);
1068:
1.57 jdolecek 1069: goto release;
1.67 he 1070: }
1.1 cgd 1071: } else
1072: uio->uio_resid -= len;
1073: if (len == m->m_len - moff) {
1074: if (m->m_flags & M_EOR)
1075: flags |= MSG_EOR;
1076: if (flags & MSG_PEEK) {
1077: m = m->m_next;
1078: moff = 0;
1079: } else {
1080: nextrecord = m->m_nextpkt;
1081: sbfree(&so->so_rcv, m);
1082: if (mp) {
1083: *mp = m;
1084: mp = &m->m_next;
1085: so->so_rcv.sb_mb = m = m->m_next;
1086: *mp = (struct mbuf *)0;
1087: } else {
1088: MFREE(m, so->so_rcv.sb_mb);
1089: m = so->so_rcv.sb_mb;
1090: }
1.69 thorpej 1091: /*
1092: * If m != NULL, we also know that
1093: * so->so_rcv.sb_mb != NULL.
1094: */
1095: KASSERT(so->so_rcv.sb_mb == m);
1096: if (m) {
1.1 cgd 1097: m->m_nextpkt = nextrecord;
1.69 thorpej 1098: if (nextrecord == NULL)
1099: so->so_rcv.sb_lastrecord = m;
1100: } else {
1101: so->so_rcv.sb_mb = nextrecord;
1.70 thorpej 1102: SB_EMPTY_FIXUP(&so->so_rcv);
1.69 thorpej 1103: }
1104: SBLASTRECORDCHK(&so->so_rcv, "soreceive 3");
1105: SBLASTMBUFCHK(&so->so_rcv, "soreceive 3");
1.1 cgd 1106: }
1107: } else {
1108: if (flags & MSG_PEEK)
1109: moff += len;
1110: else {
1111: if (mp)
1112: *mp = m_copym(m, 0, len, M_WAIT);
1113: m->m_data += len;
1114: m->m_len -= len;
1115: so->so_rcv.sb_cc -= len;
1116: }
1117: }
1118: if (so->so_oobmark) {
1119: if ((flags & MSG_PEEK) == 0) {
1120: so->so_oobmark -= len;
1121: if (so->so_oobmark == 0) {
1122: so->so_state |= SS_RCVATMARK;
1123: break;
1124: }
1.7 cgd 1125: } else {
1.1 cgd 1126: offset += len;
1.7 cgd 1127: if (offset == so->so_oobmark)
1128: break;
1129: }
1.1 cgd 1130: }
1131: if (flags & MSG_EOR)
1132: break;
1133: /*
1134: * If the MSG_WAITALL flag is set (for non-atomic socket),
1135: * we must not quit until "uio->uio_resid == 0" or an error
1136: * termination. If a signal/timeout occurs, return
1137: * with a short count but without error.
1138: * Keep sockbuf locked against other readers.
1139: */
1140: while (flags & MSG_WAITALL && m == 0 && uio->uio_resid > 0 &&
1.3 andrew 1141: !sosendallatonce(so) && !nextrecord) {
1.1 cgd 1142: if (so->so_error || so->so_state & SS_CANTRCVMORE)
1143: break;
1.68 matt 1144: /*
1145: * If we are peeking and the socket receive buffer is
1146: * full, stop since we can't get more data to peek at.
1147: */
1148: if ((flags & MSG_PEEK) && sbspace(&so->so_rcv) <= 0)
1149: break;
1150: /*
1151: * If we've drained the socket buffer, tell the
1152: * protocol in case it needs to do something to
1153: * get it filled again.
1154: */
1155: if ((pr->pr_flags & PR_WANTRCVD) && so->so_pcb)
1156: (*pr->pr_usrreq)(so, PRU_RCVD,
1157: (struct mbuf *)0,
1158: (struct mbuf *)(long)flags,
1159: (struct mbuf *)0,
1160: (struct proc *)0);
1.69 thorpej 1161: SBLASTRECORDCHK(&so->so_rcv, "soreceive sbwait 2");
1162: SBLASTMBUFCHK(&so->so_rcv, "soreceive sbwait 2");
1.1 cgd 1163: error = sbwait(&so->so_rcv);
1164: if (error) {
1165: sbunlock(&so->so_rcv);
1166: splx(s);
1167: return (0);
1168: }
1.21 christos 1169: if ((m = so->so_rcv.sb_mb) != NULL)
1.1 cgd 1170: nextrecord = m->m_nextpkt;
1171: }
1172: }
1.3 andrew 1173:
1174: if (m && pr->pr_flags & PR_ATOMIC) {
1175: flags |= MSG_TRUNC;
1176: if ((flags & MSG_PEEK) == 0)
1177: (void) sbdroprecord(&so->so_rcv);
1178: }
1.1 cgd 1179: if ((flags & MSG_PEEK) == 0) {
1.69 thorpej 1180: if (m == 0) {
1181: /*
1.70 thorpej 1182: * First part is an inline SB_EMPTY_FIXUP(). Second
1.69 thorpej 1183: * part makes sure sb_lastrecord is up-to-date if
1184: * there is still data in the socket buffer.
1185: */
1.1 cgd 1186: so->so_rcv.sb_mb = nextrecord;
1.69 thorpej 1187: if (so->so_rcv.sb_mb == NULL) {
1188: so->so_rcv.sb_mbtail = NULL;
1189: so->so_rcv.sb_lastrecord = NULL;
1190: } else if (nextrecord->m_nextpkt == NULL)
1191: so->so_rcv.sb_lastrecord = nextrecord;
1192: }
1193: SBLASTRECORDCHK(&so->so_rcv, "soreceive 4");
1194: SBLASTMBUFCHK(&so->so_rcv, "soreceive 4");
1.1 cgd 1195: if (pr->pr_flags & PR_WANTRCVD && so->so_pcb)
1.22 mycroft 1196: (*pr->pr_usrreq)(so, PRU_RCVD, (struct mbuf *)0,
1197: (struct mbuf *)(long)flags, (struct mbuf *)0,
1198: (struct proc *)0);
1.1 cgd 1199: }
1.3 andrew 1200: if (orig_resid == uio->uio_resid && orig_resid &&
1201: (flags & MSG_EOR) == 0 && (so->so_state & SS_CANTRCVMORE) == 0) {
1202: sbunlock(&so->so_rcv);
1203: splx(s);
1204: goto restart;
1205: }
1206:
1.1 cgd 1207: if (flagsp)
1208: *flagsp |= flags;
1.54 lukem 1209: release:
1.1 cgd 1210: sbunlock(&so->so_rcv);
1211: splx(s);
1212: return (error);
1213: }
1214:
1.14 mycroft 1215: int
1.54 lukem 1216: soshutdown(struct socket *so, int how)
1.1 cgd 1217: {
1.54 lukem 1218: struct protosw *pr;
1.34 kleink 1219:
1.54 lukem 1220: pr = so->so_proto;
1.34 kleink 1221: if (!(how == SHUT_RD || how == SHUT_WR || how == SHUT_RDWR))
1222: return (EINVAL);
1.1 cgd 1223:
1.34 kleink 1224: if (how == SHUT_RD || how == SHUT_RDWR)
1.1 cgd 1225: sorflush(so);
1.34 kleink 1226: if (how == SHUT_WR || how == SHUT_RDWR)
1.22 mycroft 1227: return (*pr->pr_usrreq)(so, PRU_SHUTDOWN, (struct mbuf *)0,
1228: (struct mbuf *)0, (struct mbuf *)0, (struct proc *)0);
1.1 cgd 1229: return (0);
1230: }
1231:
1.14 mycroft 1232: void
1.54 lukem 1233: sorflush(struct socket *so)
1.1 cgd 1234: {
1.54 lukem 1235: struct sockbuf *sb, asb;
1236: struct protosw *pr;
1237: int s;
1.1 cgd 1238:
1.54 lukem 1239: sb = &so->so_rcv;
1240: pr = so->so_proto;
1.1 cgd 1241: sb->sb_flags |= SB_NOINTR;
1.15 mycroft 1242: (void) sblock(sb, M_WAITOK);
1.56 thorpej 1243: s = splnet();
1.1 cgd 1244: socantrcvmore(so);
1245: sbunlock(sb);
1246: asb = *sb;
1.38 perry 1247: memset((caddr_t)sb, 0, sizeof(*sb));
1.1 cgd 1248: splx(s);
1249: if (pr->pr_flags & PR_RIGHTS && pr->pr_domain->dom_dispose)
1250: (*pr->pr_domain->dom_dispose)(asb.sb_mb);
1251: sbrelease(&asb);
1252: }
1253:
1.14 mycroft 1254: int
1.54 lukem 1255: sosetopt(struct socket *so, int level, int optname, struct mbuf *m0)
1.1 cgd 1256: {
1.54 lukem 1257: int error;
1258: struct mbuf *m;
1.1 cgd 1259:
1.54 lukem 1260: error = 0;
1261: m = m0;
1.1 cgd 1262: if (level != SOL_SOCKET) {
1263: if (so->so_proto && so->so_proto->pr_ctloutput)
1264: return ((*so->so_proto->pr_ctloutput)
1265: (PRCO_SETOPT, so, level, optname, &m0));
1266: error = ENOPROTOOPT;
1267: } else {
1268: switch (optname) {
1269:
1270: case SO_LINGER:
1.36 perry 1271: if (m == NULL || m->m_len != sizeof(struct linger)) {
1.1 cgd 1272: error = EINVAL;
1273: goto bad;
1274: }
1275: so->so_linger = mtod(m, struct linger *)->l_linger;
1276: /* fall thru... */
1277:
1278: case SO_DEBUG:
1279: case SO_KEEPALIVE:
1280: case SO_DONTROUTE:
1281: case SO_USELOOPBACK:
1282: case SO_BROADCAST:
1283: case SO_REUSEADDR:
1.15 mycroft 1284: case SO_REUSEPORT:
1.1 cgd 1285: case SO_OOBINLINE:
1.26 thorpej 1286: case SO_TIMESTAMP:
1.36 perry 1287: if (m == NULL || m->m_len < sizeof(int)) {
1.1 cgd 1288: error = EINVAL;
1289: goto bad;
1290: }
1291: if (*mtod(m, int *))
1292: so->so_options |= optname;
1293: else
1294: so->so_options &= ~optname;
1295: break;
1296:
1297: case SO_SNDBUF:
1298: case SO_RCVBUF:
1299: case SO_SNDLOWAT:
1300: case SO_RCVLOWAT:
1.28 thorpej 1301: {
1302: int optval;
1303:
1.36 perry 1304: if (m == NULL || m->m_len < sizeof(int)) {
1.1 cgd 1305: error = EINVAL;
1306: goto bad;
1307: }
1.28 thorpej 1308:
1309: /*
1310: * Values < 1 make no sense for any of these
1311: * options, so disallow them.
1312: */
1313: optval = *mtod(m, int *);
1314: if (optval < 1) {
1315: error = EINVAL;
1316: goto bad;
1317: }
1318:
1.1 cgd 1319: switch (optname) {
1320:
1321: case SO_SNDBUF:
1322: case SO_RCVBUF:
1323: if (sbreserve(optname == SO_SNDBUF ?
1324: &so->so_snd : &so->so_rcv,
1.28 thorpej 1325: (u_long) optval) == 0) {
1.1 cgd 1326: error = ENOBUFS;
1327: goto bad;
1328: }
1329: break;
1330:
1.28 thorpej 1331: /*
1332: * Make sure the low-water is never greater than
1333: * the high-water.
1334: */
1.1 cgd 1335: case SO_SNDLOWAT:
1.28 thorpej 1336: so->so_snd.sb_lowat =
1337: (optval > so->so_snd.sb_hiwat) ?
1338: so->so_snd.sb_hiwat : optval;
1.1 cgd 1339: break;
1340: case SO_RCVLOWAT:
1.28 thorpej 1341: so->so_rcv.sb_lowat =
1342: (optval > so->so_rcv.sb_hiwat) ?
1343: so->so_rcv.sb_hiwat : optval;
1.1 cgd 1344: break;
1345: }
1346: break;
1.28 thorpej 1347: }
1.1 cgd 1348:
1349: case SO_SNDTIMEO:
1350: case SO_RCVTIMEO:
1351: {
1352: struct timeval *tv;
1353: short val;
1354:
1.36 perry 1355: if (m == NULL || m->m_len < sizeof(*tv)) {
1.1 cgd 1356: error = EINVAL;
1357: goto bad;
1358: }
1359: tv = mtod(m, struct timeval *);
1.75 itojun 1360: if (tv->tv_sec > (SHRT_MAX - tv->tv_usec / tick) / hz) {
1.1 cgd 1361: error = EDOM;
1362: goto bad;
1363: }
1364: val = tv->tv_sec * hz + tv->tv_usec / tick;
1.74 itojun 1365: if (val == 0 && tv->tv_usec != 0)
1366: val = 1;
1.1 cgd 1367:
1368: switch (optname) {
1369:
1370: case SO_SNDTIMEO:
1371: so->so_snd.sb_timeo = val;
1372: break;
1373: case SO_RCVTIMEO:
1374: so->so_rcv.sb_timeo = val;
1375: break;
1376: }
1377: break;
1378: }
1379:
1380: default:
1381: error = ENOPROTOOPT;
1382: break;
1383: }
1.15 mycroft 1384: if (error == 0 && so->so_proto && so->so_proto->pr_ctloutput) {
1385: (void) ((*so->so_proto->pr_ctloutput)
1386: (PRCO_SETOPT, so, level, optname, &m0));
1387: m = NULL; /* freed by protocol */
1388: }
1.1 cgd 1389: }
1.54 lukem 1390: bad:
1.1 cgd 1391: if (m)
1392: (void) m_free(m);
1393: return (error);
1394: }
1395:
1.14 mycroft 1396: int
1.54 lukem 1397: sogetopt(struct socket *so, int level, int optname, struct mbuf **mp)
1.1 cgd 1398: {
1.54 lukem 1399: struct mbuf *m;
1.1 cgd 1400:
1401: if (level != SOL_SOCKET) {
1402: if (so->so_proto && so->so_proto->pr_ctloutput) {
1403: return ((*so->so_proto->pr_ctloutput)
1404: (PRCO_GETOPT, so, level, optname, mp));
1405: } else
1406: return (ENOPROTOOPT);
1407: } else {
1408: m = m_get(M_WAIT, MT_SOOPTS);
1.36 perry 1409: m->m_len = sizeof(int);
1.1 cgd 1410:
1411: switch (optname) {
1412:
1413: case SO_LINGER:
1.36 perry 1414: m->m_len = sizeof(struct linger);
1.1 cgd 1415: mtod(m, struct linger *)->l_onoff =
1416: so->so_options & SO_LINGER;
1417: mtod(m, struct linger *)->l_linger = so->so_linger;
1418: break;
1419:
1420: case SO_USELOOPBACK:
1421: case SO_DONTROUTE:
1422: case SO_DEBUG:
1423: case SO_KEEPALIVE:
1424: case SO_REUSEADDR:
1.15 mycroft 1425: case SO_REUSEPORT:
1.1 cgd 1426: case SO_BROADCAST:
1427: case SO_OOBINLINE:
1.26 thorpej 1428: case SO_TIMESTAMP:
1.1 cgd 1429: *mtod(m, int *) = so->so_options & optname;
1430: break;
1431:
1432: case SO_TYPE:
1433: *mtod(m, int *) = so->so_type;
1434: break;
1435:
1436: case SO_ERROR:
1437: *mtod(m, int *) = so->so_error;
1438: so->so_error = 0;
1439: break;
1440:
1441: case SO_SNDBUF:
1442: *mtod(m, int *) = so->so_snd.sb_hiwat;
1443: break;
1444:
1445: case SO_RCVBUF:
1446: *mtod(m, int *) = so->so_rcv.sb_hiwat;
1447: break;
1448:
1449: case SO_SNDLOWAT:
1450: *mtod(m, int *) = so->so_snd.sb_lowat;
1451: break;
1452:
1453: case SO_RCVLOWAT:
1454: *mtod(m, int *) = so->so_rcv.sb_lowat;
1455: break;
1456:
1457: case SO_SNDTIMEO:
1458: case SO_RCVTIMEO:
1459: {
1460: int val = (optname == SO_SNDTIMEO ?
1461: so->so_snd.sb_timeo : so->so_rcv.sb_timeo);
1462:
1463: m->m_len = sizeof(struct timeval);
1464: mtod(m, struct timeval *)->tv_sec = val / hz;
1465: mtod(m, struct timeval *)->tv_usec =
1.27 kleink 1466: (val % hz) * tick;
1.1 cgd 1467: break;
1468: }
1469:
1470: default:
1471: (void)m_free(m);
1472: return (ENOPROTOOPT);
1473: }
1474: *mp = m;
1475: return (0);
1476: }
1477: }
1478:
1.14 mycroft 1479: void
1.54 lukem 1480: sohasoutofband(struct socket *so)
1.1 cgd 1481: {
1482: struct proc *p;
1483:
1484: if (so->so_pgid < 0)
1485: gsignal(-so->so_pgid, SIGURG);
1486: else if (so->so_pgid > 0 && (p = pfind(so->so_pgid)) != 0)
1487: psignal(p, SIGURG);
1.2 cgd 1488: selwakeup(&so->so_rcv.sb_sel);
1.1 cgd 1489: }
1.72 jdolecek 1490:
1491: static void
1492: filt_sordetach(struct knote *kn)
1493: {
1494: struct socket *so;
1495:
1496: so = (struct socket *)kn->kn_fp->f_data;
1.73 christos 1497: SLIST_REMOVE(&so->so_rcv.sb_sel.sel_klist, kn, knote, kn_selnext);
1498: if (SLIST_EMPTY(&so->so_rcv.sb_sel.sel_klist))
1.72 jdolecek 1499: so->so_rcv.sb_flags &= ~SB_KNOTE;
1500: }
1501:
1502: /*ARGSUSED*/
1503: static int
1504: filt_soread(struct knote *kn, long hint)
1505: {
1506: struct socket *so;
1507:
1508: so = (struct socket *)kn->kn_fp->f_data;
1509: kn->kn_data = so->so_rcv.sb_cc;
1510: if (so->so_state & SS_CANTRCVMORE) {
1511: kn->kn_flags |= EV_EOF;
1512: kn->kn_fflags = so->so_error;
1513: return (1);
1514: }
1515: if (so->so_error) /* temporary udp error */
1516: return (1);
1517: if (kn->kn_sfflags & NOTE_LOWAT)
1518: return (kn->kn_data >= kn->kn_sdata);
1519: return (kn->kn_data >= so->so_rcv.sb_lowat);
1520: }
1521:
1522: static void
1523: filt_sowdetach(struct knote *kn)
1524: {
1525: struct socket *so;
1526:
1527: so = (struct socket *)kn->kn_fp->f_data;
1.73 christos 1528: SLIST_REMOVE(&so->so_snd.sb_sel.sel_klist, kn, knote, kn_selnext);
1529: if (SLIST_EMPTY(&so->so_snd.sb_sel.sel_klist))
1.72 jdolecek 1530: so->so_snd.sb_flags &= ~SB_KNOTE;
1531: }
1532:
1533: /*ARGSUSED*/
1534: static int
1535: filt_sowrite(struct knote *kn, long hint)
1536: {
1537: struct socket *so;
1538:
1539: so = (struct socket *)kn->kn_fp->f_data;
1540: kn->kn_data = sbspace(&so->so_snd);
1541: if (so->so_state & SS_CANTSENDMORE) {
1542: kn->kn_flags |= EV_EOF;
1543: kn->kn_fflags = so->so_error;
1544: return (1);
1545: }
1546: if (so->so_error) /* temporary udp error */
1547: return (1);
1548: if (((so->so_state & SS_ISCONNECTED) == 0) &&
1549: (so->so_proto->pr_flags & PR_CONNREQUIRED))
1550: return (0);
1551: if (kn->kn_sfflags & NOTE_LOWAT)
1552: return (kn->kn_data >= kn->kn_sdata);
1553: return (kn->kn_data >= so->so_snd.sb_lowat);
1554: }
1555:
1556: /*ARGSUSED*/
1557: static int
1558: filt_solisten(struct knote *kn, long hint)
1559: {
1560: struct socket *so;
1561:
1562: so = (struct socket *)kn->kn_fp->f_data;
1563:
1564: /*
1565: * Set kn_data to number of incoming connections, not
1566: * counting partial (incomplete) connections.
1567: */
1568: kn->kn_data = so->so_qlen;
1569: return (kn->kn_data > 0);
1570: }
1571:
1572: static const struct filterops solisten_filtops =
1573: { 1, NULL, filt_sordetach, filt_solisten };
1574: static const struct filterops soread_filtops =
1575: { 1, NULL, filt_sordetach, filt_soread };
1576: static const struct filterops sowrite_filtops =
1577: { 1, NULL, filt_sowdetach, filt_sowrite };
1578:
1579: int
1580: soo_kqfilter(struct file *fp, struct knote *kn)
1581: {
1582: struct socket *so;
1583: struct sockbuf *sb;
1584:
1585: so = (struct socket *)kn->kn_fp->f_data;
1586: switch (kn->kn_filter) {
1587: case EVFILT_READ:
1588: if (so->so_options & SO_ACCEPTCONN)
1589: kn->kn_fop = &solisten_filtops;
1590: else
1591: kn->kn_fop = &soread_filtops;
1592: sb = &so->so_rcv;
1593: break;
1594: case EVFILT_WRITE:
1595: kn->kn_fop = &sowrite_filtops;
1596: sb = &so->so_snd;
1597: break;
1598: default:
1599: return (1);
1600: }
1.73 christos 1601: SLIST_INSERT_HEAD(&sb->sb_sel.sel_klist, kn, kn_selnext);
1.72 jdolecek 1602: sb->sb_flags |= SB_KNOTE;
1603: return (0);
1604: }
1605:
CVSweb <webmaster@jp.NetBSD.org>