![[BACK]](/cvsweb/icons/back.gif){kind=icon}
Return to uipc_usrreq.c CVS log ![[TXT]](/cvsweb/icons/text.gif){kind=icon}
![[DIR]](/cvsweb/icons/dir.gif){kind=icon}
Up to [Apple XNU] / XNU / bsd / kern|
Return to uipc_usrreq.c CVS log | Up to [Apple XNU] / XNU / bsd / kern |
1.1 root 1: /*
2: * Copyright (c) 2000 Apple Computer, Inc. All rights reserved.
3: *
4: * @APPLE_LICENSE_HEADER_START@
5: *
6: * The contents of this file constitute Original Code as defined in and
7: * are subject to the Apple Public Source License Version 1.1 (the
8: * "License"). You may not use this file except in compliance with the
9: * License. Please obtain a copy of the License at
10: * http://www.apple.com/publicsource and read it before using this file.
11: *
12: * This Original Code and all software distributed under the License are
13: * distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY KIND, EITHER
14: * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
15: * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
16: * FITNESS FOR A PARTICULAR PURPOSE OR NON-INFRINGEMENT. Please see the
17: * License for the specific language governing rights and limitations
18: * under the License.
19: *
20: * @APPLE_LICENSE_HEADER_END@
21: */
22: /*
23: * Copyright (c) 1982, 1986, 1989, 1991, 1993
24: * The Regents of the University of California. All rights reserved.
25: *
26: * Redistribution and use in source and binary forms, with or without
27: * modification, are permitted provided that the following conditions
28: * are met:
29: * 1. Redistributions of source code must retain the above copyright
30: * notice, this list of conditions and the following disclaimer.
31: * 2. Redistributions in binary form must reproduce the above copyright
32: * notice, this list of conditions and the following disclaimer in the
33: * documentation and/or other materials provided with the distribution.
34: * 3. All advertising materials mentioning features or use of this software
35: * must display the following acknowledgement:
36: * This product includes software developed by the University of
37: * California, Berkeley and its contributors.
38: * 4. Neither the name of the University nor the names of its contributors
39: * may be used to endorse or promote products derived from this software
40: * without specific prior written permission.
41: *
42: * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
43: * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
44: * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
45: * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
46: * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
47: * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
48: * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
49: * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
50: * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
51: * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
52: * SUCH DAMAGE.
53: *
54: * From: @(#)uipc_usrreq.c 8.3 (Berkeley) 1/4/94
55: */
56:
57: #include <sys/param.h>
58: #include <sys/systm.h>
59: #include <sys/kernel.h>
60: #include <sys/domain.h>
61: #include <sys/fcntl.h>
62: #include <sys/malloc.h> /* XXX must be before <sys/file.h> */
63: #include <sys/file.h>
64: #include <sys/filedesc.h>
65: #include <sys/lock.h>
66: #include <sys/mbuf.h>
67: #include <sys/namei.h>
68: #include <sys/proc.h>
69: #include <sys/protosw.h>
70: #include <sys/socket.h>
71: #include <sys/socketvar.h>
72: #include <sys/stat.h>
73: #include <sys/sysctl.h>
74: #include <sys/un.h>
75: #include <sys/unpcb.h>
76: #include <sys/vnode.h>
77:
78: #include <kern/zalloc.h>
79:
80: struct zone *unp_zone;
81: static unp_gen_t unp_gencnt;
82: static u_int unp_count;
83:
84: static struct unp_head unp_shead, unp_dhead;
85:
86: /*
87: * Unix communications domain.
88: *
89: * TODO:
90: * SEQPACKET, RDM
91: * rethink name space problems
92: * need a proper out-of-band
93: * lock pushdown
94: */
95: static struct sockaddr sun_noname = { sizeof(sun_noname), AF_LOCAL };
96: static ino_t unp_ino; /* prototype for fake inode numbers */
97:
98: static int unp_attach __P((struct socket *));
99: static void unp_detach __P((struct unpcb *));
100: static int unp_bind __P((struct unpcb *,struct sockaddr *, struct proc *));
101: static int unp_connect __P((struct socket *,struct sockaddr *,
102: struct proc *));
103: static void unp_disconnect __P((struct unpcb *));
104: static void unp_shutdown __P((struct unpcb *));
105: static void unp_drop __P((struct unpcb *, int));
106: static void unp_gc __P((void));
107: static void unp_scan __P((struct mbuf *, void (*)(struct file *)));
108: static void unp_mark __P((struct file *));
109: static void unp_discard __P((struct file *));
110: static int unp_internalize __P((struct mbuf *, struct proc *));
111:
112: static int
113: uipc_abort(struct socket *so)
114: {
115: struct unpcb *unp = sotounpcb(so);
116:
117: if (unp == 0)
118: return EINVAL;
119: unp_drop(unp, ECONNABORTED);
120: return 0;
121: }
122:
123: static int
124: uipc_accept(struct socket *so, struct sockaddr **nam)
125: {
126: struct unpcb *unp = sotounpcb(so);
127:
128: if (unp == 0)
129: return EINVAL;
130:
131: /*
132: * Pass back name of connected socket,
133: * if it was bound and we are still connected
134: * (our peer may have closed already!).
135: */
136: if (unp->unp_conn && unp->unp_conn->unp_addr) {
137: *nam = dup_sockaddr((struct sockaddr *)unp->unp_conn->unp_addr,
138: 1);
139: } else {
140: *nam = dup_sockaddr((struct sockaddr *)&sun_noname, 1);
141: }
142: return 0;
143: }
144:
145: static int
146: uipc_attach(struct socket *so, int proto, struct proc *p)
147: {
148: struct unpcb *unp = sotounpcb(so);
149:
150: if (unp != 0)
151: return EISCONN;
152: return unp_attach(so);
153: }
154:
155: static int
156: uipc_bind(struct socket *so, struct sockaddr *nam, struct proc *p)
157: {
158: struct unpcb *unp = sotounpcb(so);
159:
160: if (unp == 0)
161: return EINVAL;
162:
163: return unp_bind(unp, nam, p);
164: }
165:
166: static int
167: uipc_connect(struct socket *so, struct sockaddr *nam, struct proc *p)
168: {
169: struct unpcb *unp = sotounpcb(so);
170:
171: if (unp == 0)
172: return EINVAL;
173: return unp_connect(so, nam, p);
174: }
175:
176: static int
177: uipc_connect2(struct socket *so1, struct socket *so2)
178: {
179: struct unpcb *unp = sotounpcb(so1);
180:
181: if (unp == 0)
182: return EINVAL;
183:
184: return unp_connect2(so1, so2);
185: }
186:
187: /* control is EOPNOTSUPP */
188:
189: static int
190: uipc_detach(struct socket *so)
191: {
192: struct unpcb *unp = sotounpcb(so);
193:
194: if (unp == 0)
195: return EINVAL;
196:
197: unp_detach(unp);
198: return 0;
199: }
200:
201: static int
202: uipc_disconnect(struct socket *so)
203: {
204: struct unpcb *unp = sotounpcb(so);
205:
206: if (unp == 0)
207: return EINVAL;
208: unp_disconnect(unp);
209: return 0;
210: }
211:
212: static int
213: uipc_listen(struct socket *so, struct proc *p)
214: {
215: struct unpcb *unp = sotounpcb(so);
216:
217: if (unp == 0 || unp->unp_vnode == 0)
218: return EINVAL;
219: return 0;
220: }
221:
222: static int
223: uipc_peeraddr(struct socket *so, struct sockaddr **nam)
224: {
225: struct unpcb *unp = sotounpcb(so);
226:
227: if (unp == 0)
228: return EINVAL;
229: if (unp->unp_conn && unp->unp_conn->unp_addr)
230: *nam = dup_sockaddr((struct sockaddr *)unp->unp_conn->unp_addr,
231: 1);
232: return 0;
233: }
234:
235: static int
236: uipc_rcvd(struct socket *so, int flags)
237: {
238: struct unpcb *unp = sotounpcb(so);
239: struct socket *so2;
240:
241: if (unp == 0)
242: return EINVAL;
243: switch (so->so_type) {
244: case SOCK_DGRAM:
245: panic("uipc_rcvd DGRAM?");
246: /*NOTREACHED*/
247:
248: case SOCK_STREAM:
249: #define rcv (&so->so_rcv)
250: #define snd (&so2->so_snd)
251: if (unp->unp_conn == 0)
252: break;
253: so2 = unp->unp_conn->unp_socket;
254: /*
255: * Adjust backpressure on sender
256: * and wakeup any waiting to write.
257: */
258: snd->sb_mbmax += unp->unp_mbcnt - rcv->sb_mbcnt;
259: unp->unp_mbcnt = rcv->sb_mbcnt;
260: snd->sb_hiwat += unp->unp_cc - rcv->sb_cc;
261: unp->unp_cc = rcv->sb_cc;
262: sowwakeup(so2);
263: #undef snd
264: #undef rcv
265: break;
266:
267: default:
268: panic("uipc_rcvd unknown socktype");
269: }
270: return 0;
271: }
272:
273: /* pru_rcvoob is EOPNOTSUPP */
274:
275: static int
276: uipc_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
277: struct mbuf *control, struct proc *p)
278: {
279: int error = 0;
280: struct unpcb *unp = sotounpcb(so);
281: struct socket *so2;
282:
283: if (unp == 0) {
284: error = EINVAL;
285: goto release;
286: }
287: if (flags & PRUS_OOB) {
288: error = EOPNOTSUPP;
289: goto release;
290: }
291:
292: if (control && (error = unp_internalize(control, p)))
293: goto release;
294:
295: switch (so->so_type) {
296: case SOCK_DGRAM:
297: {
298: struct sockaddr *from;
299:
300: if (nam) {
301: if (unp->unp_conn) {
302: error = EISCONN;
303: break;
304: }
305: error = unp_connect(so, nam, p);
306: if (error)
307: break;
308: } else {
309: if (unp->unp_conn == 0) {
310: error = ENOTCONN;
311: break;
312: }
313: }
314: so2 = unp->unp_conn->unp_socket;
315: if (unp->unp_addr)
316: from = (struct sockaddr *)unp->unp_addr;
317: else
318: from = &sun_noname;
319: if (sbappendaddr(&so2->so_rcv, from, m, control)) {
320: sorwakeup(so2);
321: m = 0;
322: control = 0;
323: } else
324: error = ENOBUFS;
325: if (nam)
326: unp_disconnect(unp);
327: break;
328: }
329:
330: case SOCK_STREAM:
331: #define rcv (&so2->so_rcv)
332: #define snd (&so->so_snd)
333: /* Connect if not connected yet. */
334: /*
335: * Note: A better implementation would complain
336: * if not equal to the peer's address.
337: */
338: if ((so->so_state & SS_ISCONNECTED) == 0) {
339: if (nam) {
340: error = unp_connect(so, nam, p);
341: if (error)
342: break; /* XXX */
343: } else {
344: error = ENOTCONN;
345: break;
346: }
347: }
348:
349: if (so->so_state & SS_CANTSENDMORE) {
350: error = EPIPE;
351: break;
352: }
353: if (unp->unp_conn == 0)
354: panic("uipc_send connected but no connection?");
355: so2 = unp->unp_conn->unp_socket;
356: /*
357: * Send to paired receive port, and then reduce
358: * send buffer hiwater marks to maintain backpressure.
359: * Wake up readers.
360: */
361: if (control) {
362: if (sbappendcontrol(rcv, m, control))
363: control = 0;
364: } else
365: sbappend(rcv, m);
366: snd->sb_mbmax -=
367: rcv->sb_mbcnt - unp->unp_conn->unp_mbcnt;
368: unp->unp_conn->unp_mbcnt = rcv->sb_mbcnt;
369: snd->sb_hiwat -= rcv->sb_cc - unp->unp_conn->unp_cc;
370: unp->unp_conn->unp_cc = rcv->sb_cc;
371: sorwakeup(so2);
372: m = 0;
373: #undef snd
374: #undef rcv
375: break;
376:
377: default:
378: panic("uipc_send unknown socktype");
379: }
380:
381: /*
382: * SEND_EOF is equivalent to a SEND followed by
383: * a SHUTDOWN.
384: */
385: if (flags & PRUS_EOF) {
386: socantsendmore(so);
387: unp_shutdown(unp);
388: }
389:
390: release:
391: if (control)
392: m_freem(control);
393: if (m)
394: m_freem(m);
395: return error;
396: }
397:
398: static int
399: uipc_sense(struct socket *so, struct stat *sb)
400: {
401: struct unpcb *unp = sotounpcb(so);
402: struct socket *so2;
403:
404: if (unp == 0)
405: return EINVAL;
406: sb->st_blksize = so->so_snd.sb_hiwat;
407: if (so->so_type == SOCK_STREAM && unp->unp_conn != 0) {
408: so2 = unp->unp_conn->unp_socket;
409: sb->st_blksize += so2->so_rcv.sb_cc;
410: }
411: sb->st_dev = NODEV;
412: if (unp->unp_ino == 0)
413: unp->unp_ino = unp_ino++;
414: sb->st_ino = unp->unp_ino;
415: return (0);
416: }
417:
418: static int
419: uipc_shutdown(struct socket *so)
420: {
421: struct unpcb *unp = sotounpcb(so);
422:
423: if (unp == 0)
424: return EINVAL;
425: socantsendmore(so);
426: unp_shutdown(unp);
427: return 0;
428: }
429:
430: static int
431: uipc_sockaddr(struct socket *so, struct sockaddr **nam)
432: {
433: struct unpcb *unp = sotounpcb(so);
434:
435: if (unp == 0)
436: return EINVAL;
437: if (unp->unp_addr)
438: *nam = dup_sockaddr((struct sockaddr *)unp->unp_addr, 1);
439: return 0;
440: }
441:
442: struct pr_usrreqs uipc_usrreqs = {
443: uipc_abort, uipc_accept, uipc_attach, uipc_bind, uipc_connect,
444: uipc_connect2, pru_control_notsupp, uipc_detach, uipc_disconnect,
445: uipc_listen, uipc_peeraddr, uipc_rcvd, pru_rcvoob_notsupp,
446: uipc_send, uipc_sense, uipc_shutdown, uipc_sockaddr,
447: sosend, soreceive, sopoll
448: };
449:
450: /*
451: * Both send and receive buffers are allocated PIPSIZ bytes of buffering
452: * for stream sockets, although the total for sender and receiver is
453: * actually only PIPSIZ.
454: * Datagram sockets really use the sendspace as the maximum datagram size,
455: * and don't really want to reserve the sendspace. Their recvspace should
456: * be large enough for at least one max-size datagram plus address.
457: */
458: #ifndef PIPSIZ
459: #define PIPSIZ 8192
460: #endif
461: static u_long unpst_sendspace = PIPSIZ;
462: static u_long unpst_recvspace = PIPSIZ;
463: static u_long unpdg_sendspace = 2*1024; /* really max datagram size */
464: static u_long unpdg_recvspace = 4*1024;
465:
466: static int unp_rights; /* file descriptors in flight */
467:
468: SYSCTL_DECL(_net_local_stream);
469: SYSCTL_INT(_net_local_stream, OID_AUTO, sendspace, CTLFLAG_RW,
470: &unpst_sendspace, 0, "");
471: SYSCTL_INT(_net_local_stream, OID_AUTO, recvspace, CTLFLAG_RW,
472: &unpst_recvspace, 0, "");
473: SYSCTL_DECL(_net_local_dgram);
474: SYSCTL_INT(_net_local_dgram, OID_AUTO, maxdgram, CTLFLAG_RW,
475: &unpdg_sendspace, 0, "");
476: SYSCTL_INT(_net_local_dgram, OID_AUTO, recvspace, CTLFLAG_RW,
477: &unpdg_recvspace, 0, "");
478: SYSCTL_DECL(_net_local);
479: SYSCTL_INT(_net_local, OID_AUTO, inflight, CTLFLAG_RD, &unp_rights, 0, "");
480:
481: static int
482: unp_attach(so)
483: struct socket *so;
484: {
485: register struct unpcb *unp;
486: int error;
487:
488: if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
489: switch (so->so_type) {
490:
491: case SOCK_STREAM:
492: error = soreserve(so, unpst_sendspace, unpst_recvspace);
493: break;
494:
495: case SOCK_DGRAM:
496: error = soreserve(so, unpdg_sendspace, unpdg_recvspace);
497: break;
498:
499: default:
500: panic("unp_attach");
501: }
502: if (error)
503: return (error);
504: }
505: unp = zalloc(unp_zone);
506: if (unp == NULL)
507: return (ENOBUFS);
508: bzero(unp, sizeof *unp);
509: unp->unp_gencnt = ++unp_gencnt;
510: unp_count++;
511: LIST_INIT(&unp->unp_refs);
512: unp->unp_socket = so;
513: LIST_INSERT_HEAD(so->so_type == SOCK_DGRAM ? &unp_dhead
514: : &unp_shead, unp, unp_link);
515: so->so_pcb = (caddr_t)unp;
516: return (0);
517: }
518:
519: static void
520: unp_detach(unp)
521: register struct unpcb *unp;
522: {
523: LIST_REMOVE(unp, unp_link);
524: unp->unp_gencnt = ++unp_gencnt;
525: --unp_count;
526: if (unp->unp_vnode) {
527: unp->unp_vnode->v_socket = 0;
528: vrele(unp->unp_vnode);
529: unp->unp_vnode = 0;
530: }
531: if (unp->unp_conn)
532: unp_disconnect(unp);
533: while (unp->unp_refs.lh_first)
534: unp_drop(unp->unp_refs.lh_first, ECONNRESET);
535: soisdisconnected(unp->unp_socket);
536: unp->unp_socket->so_pcb = 0;
537: if (unp_rights) {
538: /*
539: * Normally the receive buffer is flushed later,
540: * in sofree, but if our receive buffer holds references
541: * to descriptors that are now garbage, we will dispose
542: * of those descriptor references after the garbage collector
543: * gets them (resulting in a "panic: closef: count < 0").
544: */
545: sorflush(unp->unp_socket);
546: unp_gc();
547: }
548: if (unp->unp_addr)
549: FREE(unp->unp_addr, M_SONAME);
550: zfree(unp_zone, unp);
551: }
552:
553: static int
554: unp_bind(unp, nam, p)
555: struct unpcb *unp;
556: struct sockaddr *nam;
557: struct proc *p;
558: {
559: struct sockaddr_un *soun = (struct sockaddr_un *)nam;
560: register struct vnode *vp;
561: struct vattr vattr;
562: int error, namelen;
563: struct nameidata nd;
564: char buf[SOCK_MAXADDRLEN];
565:
566: if (unp->unp_vnode != NULL)
567: return (EINVAL);
568: #define offsetof(s, e) ((char *)&((s *)0)->e - (char *)((s *)0))
569: namelen = soun->sun_len - offsetof(struct sockaddr_un, sun_path);
570: if (namelen <= 0)
571: return EINVAL;
572: strncpy(buf, soun->sun_path, namelen);
573: buf[namelen] = 0; /* null-terminate the string */
574: NDINIT(&nd, CREATE, FOLLOW | LOCKPARENT, UIO_SYSSPACE,
575: buf, p);
576: /* SHOULD BE ABLE TO ADOPT EXISTING AND wakeup() ALA FIFO's */
577: error = namei(&nd);
578: if (error)
579: return (error);
580: vp = nd.ni_vp;
581: if (vp != NULL) {
582: VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
583: if (nd.ni_dvp == vp)
584: vrele(nd.ni_dvp);
585: else
586: vput(nd.ni_dvp);
587: vrele(vp);
588: return (EADDRINUSE);
589: }
590: VATTR_NULL(&vattr);
591: vattr.va_type = VSOCK;
592: vattr.va_mode = (ACCESSPERMS & ~p->p_fd->fd_cmask);
593: VOP_LEASE(nd.ni_dvp, p, p->p_ucred, LEASE_WRITE);
594: error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &vattr);
595: #if 0
596: /* In FreeBSD create leave s parent held ; not here */
597: vput(nd.ni_dvp);
598: #endif
599: if (error)
600: return (error);
601: vp = nd.ni_vp;
602: vp->v_socket = unp->unp_socket;
603: unp->unp_vnode = vp;
604: unp->unp_addr = (struct sockaddr_un *)dup_sockaddr(nam, 1);
605: VOP_UNLOCK(vp, 0, p);
606: return (0);
607: }
608:
609: static int
610: unp_connect(so, nam, p)
611: struct socket *so;
612: struct sockaddr *nam;
613: struct proc *p;
614: {
615: register struct sockaddr_un *soun = (struct sockaddr_un *)nam;
616: register struct vnode *vp;
617: register struct socket *so2, *so3;
618: struct unpcb *unp2, *unp3;
619: int error, len;
620: struct nameidata nd;
621: char buf[SOCK_MAXADDRLEN];
622:
623: len = nam->sa_len - offsetof(struct sockaddr_un, sun_path);
624: if (len <= 0)
625: return EINVAL;
626: strncpy(buf, soun->sun_path, len);
627: buf[len] = 0;
628:
629: NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, buf, p);
630: error = namei(&nd);
631: if (error)
632: return (error);
633: vp = nd.ni_vp;
634: if (vp->v_type != VSOCK) {
635: error = ENOTSOCK;
636: goto bad;
637: }
638: error = VOP_ACCESS(vp, VWRITE, p->p_ucred, p);
639: if (error)
640: goto bad;
641: so2 = vp->v_socket;
642: if (so2 == 0) {
643: error = ECONNREFUSED;
644: goto bad;
645: }
646: if (so->so_type != so2->so_type) {
647: error = EPROTOTYPE;
648: goto bad;
649: }
650: if (so->so_proto->pr_flags & PR_CONNREQUIRED) {
651: if ((so2->so_options & SO_ACCEPTCONN) == 0 ||
652: (so3 = sonewconn(so2, 0)) == 0) {
653: error = ECONNREFUSED;
654: goto bad;
655: }
656: unp2 = sotounpcb(so2);
657: unp3 = sotounpcb(so3);
658: if (unp2->unp_addr)
659: unp3->unp_addr = (struct sockaddr_un *)
660: dup_sockaddr((struct sockaddr *)
661: unp2->unp_addr, 1);
662: so2 = so3;
663: }
664: error = unp_connect2(so, so2);
665: bad:
666: vput(vp);
667: return (error);
668: }
669:
670: int
671: unp_connect2(so, so2)
672: register struct socket *so;
673: register struct socket *so2;
674: {
675: register struct unpcb *unp = sotounpcb(so);
676: register struct unpcb *unp2;
677:
678: if (so2->so_type != so->so_type)
679: return (EPROTOTYPE);
680: unp2 = sotounpcb(so2);
681: unp->unp_conn = unp2;
682: switch (so->so_type) {
683:
684: case SOCK_DGRAM:
685: LIST_INSERT_HEAD(&unp2->unp_refs, unp, unp_reflink);
686: soisconnected(so);
687: break;
688:
689: case SOCK_STREAM:
690: unp2->unp_conn = unp;
691: soisconnected(so);
692: soisconnected(so2);
693: break;
694:
695: default:
696: panic("unp_connect2");
697: }
698: return (0);
699: }
700:
701: static void
702: unp_disconnect(unp)
703: struct unpcb *unp;
704: {
705: register struct unpcb *unp2 = unp->unp_conn;
706:
707: if (unp2 == 0)
708: return;
709: unp->unp_conn = 0;
710: switch (unp->unp_socket->so_type) {
711:
712: case SOCK_DGRAM:
713: LIST_REMOVE(unp, unp_reflink);
714: unp->unp_socket->so_state &= ~SS_ISCONNECTED;
715: break;
716:
717: case SOCK_STREAM:
718: soisdisconnected(unp->unp_socket);
719: unp2->unp_conn = 0;
720: soisdisconnected(unp2->unp_socket);
721: break;
722: }
723: }
724:
725: #ifdef notdef
726: void
727: unp_abort(unp)
728: struct unpcb *unp;
729: {
730:
731: unp_detach(unp);
732: }
733: #endif
734:
735: static int
736: unp_pcblist SYSCTL_HANDLER_ARGS
737: {
738: int error, i, n;
739: struct unpcb *unp, **unp_list;
740: unp_gen_t gencnt;
741: struct xunpgen xug;
742: struct unp_head *head;
743:
744: head = ((intptr_t)arg1 == SOCK_DGRAM ? &unp_dhead : &unp_shead);
745:
746: /*
747: * The process of preparing the PCB list is too time-consuming and
748: * resource-intensive to repeat twice on every request.
749: */
750: if (req->oldptr == 0) {
751: n = unp_count;
752: req->oldidx = 2 * (sizeof xug)
753: + (n + n/8) * sizeof(struct xunpcb);
754: return 0;
755: }
756:
757: if (req->newptr != 0)
758: return EPERM;
759:
760: /*
761: * OK, now we're committed to doing something.
762: */
763: gencnt = unp_gencnt;
764: n = unp_count;
765:
766: xug.xug_len = sizeof xug;
767: xug.xug_count = n;
768: xug.xug_gen = gencnt;
769: xug.xug_sogen = so_gencnt;
770: error = SYSCTL_OUT(req, &xug, sizeof xug);
771: if (error)
772: return error;
773:
774: unp_list = _MALLOC(n * sizeof *unp_list, M_TEMP, M_WAITOK);
775: if (unp_list == 0)
776: return ENOMEM;
777:
778: for (unp = head->lh_first, i = 0; unp && i < n;
779: unp = unp->unp_link.le_next) {
780: if (unp->unp_gencnt <= gencnt)
781: unp_list[i++] = unp;
782: }
783: n = i; /* in case we lost some during malloc */
784:
785: error = 0;
786: for (i = 0; i < n; i++) {
787: unp = unp_list[i];
788: if (unp->unp_gencnt <= gencnt) {
789: struct xunpcb xu;
790: xu.xu_len = sizeof xu;
791: xu.xu_unpp = unp;
792: /*
793: * XXX - need more locking here to protect against
794: * connect/disconnect races for SMP.
795: */
796: if (unp->unp_addr)
797: bcopy(unp->unp_addr, &xu.xu_addr,
798: unp->unp_addr->sun_len);
799: if (unp->unp_conn && unp->unp_conn->unp_addr)
800: bcopy(unp->unp_conn->unp_addr,
801: &xu.xu_caddr,
802: unp->unp_conn->unp_addr->sun_len);
803: bcopy(unp, &xu.xu_unp, sizeof *unp);
804: sotoxsocket(unp->unp_socket, &xu.xu_socket);
805: error = SYSCTL_OUT(req, &xu, sizeof xu);
806: }
807: }
808: if (!error) {
809: /*
810: * Give the user an updated idea of our state.
811: * If the generation differs from what we told
812: * her before, she knows that something happened
813: * while we were processing this request, and it
814: * might be necessary to retry.
815: */
816: xug.xug_gen = unp_gencnt;
817: xug.xug_sogen = so_gencnt;
818: xug.xug_count = unp_count;
819: error = SYSCTL_OUT(req, &xug, sizeof xug);
820: }
821: FREE(unp_list, M_TEMP);
822: return error;
823: }
824:
825: SYSCTL_PROC(_net_local_dgram, OID_AUTO, pcblist, CTLFLAG_RD,
826: (caddr_t)(long)SOCK_DGRAM, 0, unp_pcblist, "S,xunpcb",
827: "List of active local datagram sockets");
828: SYSCTL_PROC(_net_local_stream, OID_AUTO, pcblist, CTLFLAG_RD,
829: (caddr_t)(long)SOCK_STREAM, 0, unp_pcblist, "S,xunpcb",
830: "List of active local stream sockets");
831:
832: static void
833: unp_shutdown(unp)
834: struct unpcb *unp;
835: {
836: struct socket *so;
837:
838: if (unp->unp_socket->so_type == SOCK_STREAM && unp->unp_conn &&
839: (so = unp->unp_conn->unp_socket))
840: socantrcvmore(so);
841: }
842:
843: static void
844: unp_drop(unp, errno)
845: struct unpcb *unp;
846: int errno;
847: {
848: struct socket *so = unp->unp_socket;
849:
850: so->so_error = errno;
851: unp_disconnect(unp);
852: if (so->so_head) {
853: LIST_REMOVE(unp, unp_link);
854: unp->unp_gencnt = ++unp_gencnt;
855: unp_count--;
856: so->so_pcb = (caddr_t) 0;
857: if (unp->unp_addr)
858: FREE(unp->unp_addr, M_SONAME);
859: zfree(unp_zone, unp);
860: sofree(so);
861: }
862: }
863:
864: #ifdef notdef
865: void
866: unp_drain()
867: {
868:
869: }
870: #endif
871:
872: int
873: unp_externalize(rights)
874: struct mbuf *rights;
875: {
876: struct proc *p = current_proc(); /* XXX */
877: register int i;
878: register struct cmsghdr *cm = mtod(rights, struct cmsghdr *);
879: register struct file **rp = (struct file **)(cm + 1);
880: register struct file *fp;
881: int newfds = (cm->cmsg_len - sizeof(*cm)) / sizeof (int);
882: int f;
883:
884: /*
885: * if the new FD's will not fit, then we free them all
886: */
887: if (!fdavail(p, newfds)) {
888: for (i = 0; i < newfds; i++) {
889: fp = *rp;
890: unp_discard(fp);
891: *rp++ = 0;
892: }
893: return (EMSGSIZE);
894: }
895: /*
896: * now change each pointer to an fd in the global table to
897: * an integer that is the index to the local fd table entry
898: * that we set up to point to the global one we are transferring.
899: * XXX this assumes a pointer and int are the same size...!
900: */
901: for (i = 0; i < newfds; i++) {
902: if (fdalloc(p, 0, &f))
903: panic("unp_externalize");
904: fp = *rp;
905: p->p_fd->fd_ofiles[f] = fp;
906: fp->f_msgcount--;
907: unp_rights--;
908: *(int *)rp++ = f;
909: }
910: return (0);
911: }
912:
913: void
914: unp_init(void)
915: {
916: unp_zone = zinit(sizeof(struct unpcb),
917: (nmbclusters * sizeof(struct unpcb)),
918: 4096, "unpzone");
919: if (unp_zone == 0)
920: panic("unp_init");
921: LIST_INIT(&unp_dhead);
922: LIST_INIT(&unp_shead);
923: }
924:
925: #ifndef MIN
926: #define MIN(a,b) (((a)<(b))?(a):(b))
927: #endif
928:
929: static int
930: unp_internalize(control, p)
931: struct mbuf *control;
932: struct proc *p;
933: {
934: register struct cmsghdr *cm = mtod(control, struct cmsghdr *);
935: register struct file **rp;
936: struct file *fp;
937: register int i, error;
938: int oldfds;
939:
940: if (cm->cmsg_type != SCM_RIGHTS || cm->cmsg_level != SOL_SOCKET ||
941: cm->cmsg_len != control->m_len)
942: return (EINVAL);
943: oldfds = (cm->cmsg_len - sizeof (*cm)) / sizeof (int);
944: rp = (struct file **)(cm + 1);
945: for (i = 0; i < oldfds; i++)
946: if (error = fdgetf(p, *(int *)rp++, 0))
947: return (error);
948: rp = (struct file **)(cm + 1);
949: for (i = 0; i < oldfds; i++) {
950: (void) fdgetf(p, *(int *)rp, &fp);
951: *rp++ = fp;
952: fp->f_count++;
953: if (fp->f_count <= 0)
954: panic("unp_internalize f_count");
955: fp->f_msgcount++;
956: unp_rights++;
957: }
958: return (0);
959: }
960:
961: static int unp_defer, unp_gcing;
962:
963: static void
964: unp_gc()
965: {
966: register struct file *fp, *nextfp;
967: register struct socket *so;
968: struct file **extra_ref, **fpp;
969: int nunref, i;
970:
971: if (unp_gcing)
972: return;
973: unp_gcing = 1;
974: unp_defer = 0;
975: /*
976: * before going through all this, set all FDs to
977: * be NOT defered and NOT externally accessible
978: */
979: for (fp = filehead.lh_first; fp != 0; fp = fp->f_list.le_next)
980: fp->f_flag &= ~(FMARK|FDEFER);
981: do {
982: for (fp = filehead.lh_first; fp != 0; fp = fp->f_list.le_next) {
983: /*
984: * If the file is not open, skip it
985: */
986: if (fp->f_count == 0)
987: continue;
988: /*
989: * If we already marked it as 'defer' in a
990: * previous pass, then try process it this time
991: * and un-mark it
992: */
993: if (fp->f_flag & FDEFER) {
994: fp->f_flag &= ~FDEFER;
995: unp_defer--;
996: } else {
997: /*
998: * if it's not defered, then check if it's
999: * already marked.. if so skip it
1000: */
1001: if (fp->f_flag & FMARK)
1002: continue;
1003: /*
1004: * If all references are from messages
1005: * in transit, then skip it. it's not
1006: * externally accessible.
1007: */
1008: if (fp->f_count == fp->f_msgcount)
1009: continue;
1010: /*
1011: * If it got this far then it must be
1012: * externally accessible.
1013: */
1014: fp->f_flag |= FMARK;
1015: }
1016: /*
1017: * either it was defered, or it is externally
1018: * accessible and not already marked so.
1019: * Now check if it is possibly one of OUR sockets.
1020: */
1021: if (fp->f_type != DTYPE_SOCKET ||
1022: (so = (struct socket *)fp->f_data) == 0)
1023: continue;
1024: if (so->so_proto->pr_domain != &localdomain ||
1025: (so->so_proto->pr_flags&PR_RIGHTS) == 0)
1026: continue;
1027: #ifdef notdef
1028: if (so->so_rcv.sb_flags & SB_LOCK) {
1029: /*
1030: * This is problematical; it's not clear
1031: * we need to wait for the sockbuf to be
1032: * unlocked (on a uniprocessor, at least),
1033: * and it's also not clear what to do
1034: * if sbwait returns an error due to receipt
1035: * of a signal. If sbwait does return
1036: * an error, we'll go into an infinite
1037: * loop. Delete all of this for now.
1038: */
1039: (void) sbwait(&so->so_rcv);
1040: goto restart;
1041: }
1042: #endif
1043: /*
1044: * So, Ok, it's one of our sockets and it IS externally
1045: * accessible (or was defered). Now we look
1046: * to see if we hold any file descriptors in its
1047: * message buffers. Follow those links and mark them
1048: * as accessible too.
1049: */
1050: unp_scan(so->so_rcv.sb_mb, unp_mark);
1051: }
1052: } while (unp_defer);
1053: /*
1054: * We grab an extra reference to each of the file table entries
1055: * that are not otherwise accessible and then free the rights
1056: * that are stored in messages on them.
1057: *
1058: * The bug in the orginal code is a little tricky, so I'll describe
1059: * what's wrong with it here.
1060: *
1061: * It is incorrect to simply unp_discard each entry for f_msgcount
1062: * times -- consider the case of sockets A and B that contain
1063: * references to each other. On a last close of some other socket,
1064: * we trigger a gc since the number of outstanding rights (unp_rights)
1065: * is non-zero. If during the sweep phase the gc code un_discards,
1066: * we end up doing a (full) closef on the descriptor. A closef on A
1067: * results in the following chain. Closef calls soo_close, which
1068: * calls soclose. Soclose calls first (through the switch
1069: * uipc_usrreq) unp_detach, which re-invokes unp_gc. Unp_gc simply
1070: * returns because the previous instance had set unp_gcing, and
1071: * we return all the way back to soclose, which marks the socket
1072: * with SS_NOFDREF, and then calls sofree. Sofree calls sorflush
1073: * to free up the rights that are queued in messages on the socket A,
1074: * i.e., the reference on B. The sorflush calls via the dom_dispose
1075: * switch unp_dispose, which unp_scans with unp_discard. This second
1076: * instance of unp_discard just calls closef on B.
1077: *
1078: * Well, a similar chain occurs on B, resulting in a sorflush on B,
1079: * which results in another closef on A. Unfortunately, A is already
1080: * being closed, and the descriptor has already been marked with
1081: * SS_NOFDREF, and soclose panics at this point.
1082: *
1083: * Here, we first take an extra reference to each inaccessible
1084: * descriptor. Then, we call sorflush ourself, since we know
1085: * it is a Unix domain socket anyhow. After we destroy all the
1086: * rights carried in messages, we do a last closef to get rid
1087: * of our extra reference. This is the last close, and the
1088: * unp_detach etc will shut down the socket.
1089: *
1090: * 91/09/19, [email protected]
1091: */
1092: extra_ref = _MALLOC(nfiles * sizeof(struct file *), M_FILE, M_WAITOK);
1093: for (nunref = 0, fp = filehead.lh_first, fpp = extra_ref; fp != 0;
1094: fp = nextfp) {
1095: nextfp = fp->f_list.le_next;
1096: /*
1097: * If it's not open, skip it
1098: */
1099: if (fp->f_count == 0)
1100: continue;
1101: /*
1102: * If all refs are from msgs, and it's not marked accessible
1103: * then it must be referenced from some unreachable cycle
1104: * of (shut-down) FDs, so include it in our
1105: * list of FDs to remove
1106: */
1107: if (fp->f_count == fp->f_msgcount && !(fp->f_flag & FMARK)) {
1108: *fpp++ = fp;
1109: nunref++;
1110: fp->f_count++;
1111: if (fp->f_count <= 0)
1112: panic("unp_gc f_count");
1113: }
1114: }
1115: /*
1116: * for each FD on our hit list, do the following two things
1117: */
1118: for (i = nunref, fpp = extra_ref; --i >= 0; ++fpp) {
1119: struct file *tfp = *fpp;
1120: if (tfp->f_type == DTYPE_SOCKET && tfp->f_data != NULL)
1121: sorflush((struct socket *)(tfp->f_data));
1122: }
1123:
1124:
1125: for (i = nunref, fpp = extra_ref; --i >= 0; ++fpp)
1126: closef(*fpp, (struct proc *) NULL);
1127: FREE((caddr_t)extra_ref, M_FILE);
1128: unp_gcing = 0;
1129: }
1130:
1131: void
1132: unp_dispose(m)
1133: struct mbuf *m;
1134: {
1135:
1136: if (m)
1137: unp_scan(m, unp_discard);
1138: }
1139:
1140: static void
1141: unp_scan(m0, op)
1142: register struct mbuf *m0;
1143: void (*op) __P((struct file *));
1144: {
1145: register struct mbuf *m;
1146: register struct file **rp;
1147: register struct cmsghdr *cm;
1148: register int i;
1149: int qfds;
1150:
1151: while (m0) {
1152: for (m = m0; m; m = m->m_next)
1153: if (m->m_type == MT_CONTROL &&
1154: m->m_len >= sizeof(*cm)) {
1155: cm = mtod(m, struct cmsghdr *);
1156: if (cm->cmsg_level != SOL_SOCKET ||
1157: cm->cmsg_type != SCM_RIGHTS)
1158: continue;
1159: qfds = (cm->cmsg_len - sizeof *cm)
1160: / sizeof (struct file *);
1161: rp = (struct file **)(cm + 1);
1162: for (i = 0; i < qfds; i++)
1163: (*op)(*rp++);
1164: break; /* XXX, but saves time */
1165: }
1166: m0 = m0->m_act;
1167: }
1168: }
1169:
1170: static void
1171: unp_mark(fp)
1172: struct file *fp;
1173: {
1174:
1175: if (fp->f_flag & FMARK)
1176: return;
1177: unp_defer++;
1178: fp->f_flag |= (FMARK|FDEFER);
1179: }
1180:
1181: static void
1182: unp_discard(fp)
1183: struct file *fp;
1184: {
1185:
1186: fp->f_msgcount--;
1187: unp_rights--;
1188: (void) closef(fp, (struct proc *)NULL);
1189: }
This archive runs on limited infrastructure. Preserving old code on modern bandwidth. Automated agents are requested to crawl responsibly.