1/* $NetBSD: svc_vc.c,v 1.7 2000/08/03 00:01:53 fvdl Exp $ */
2
3/*
4 * Sun RPC is a product of Sun Microsystems, Inc. and is provided for
5 * unrestricted use provided that this legend is included on all tape
6 * media and as a part of the software program in whole or part. Users
7 * may copy or modify Sun RPC without charge, but are not authorized
8 * to license or distribute it to anyone else except as part of a product or
9 * program developed by the user.
10 *
11 * SUN RPC IS PROVIDED AS IS WITH NO WARRANTIES OF ANY KIND INCLUDING THE
12 * WARRANTIES OF DESIGN, MERCHANTIBILITY AND FITNESS FOR A PARTICULAR
13 * PURPOSE, OR ARISING FROM A COURSE OF DEALING, USAGE OR TRADE PRACTICE.
14 *
15 * Sun RPC is provided with no support and without any obligation on the
16 * part of Sun Microsystems, Inc. to assist in its use, correction,
17 * modification or enhancement.
18 *
19 * SUN MICROSYSTEMS, INC. SHALL HAVE NO LIABILITY WITH RESPECT TO THE
20 * INFRINGEMENT OF COPYRIGHTS, TRADE SECRETS OR ANY PATENTS BY SUN RPC
21 * OR ANY PART THEREOF.
22 *
23 * In no event will Sun Microsystems, Inc. be liable for any lost revenue
24 * or profits or other special, indirect and consequential damages, even if
25 * Sun has been advised of the possibility of such damages.
26 *
27 * Sun Microsystems, Inc.
28 * 2550 Garcia Avenue
29 * Mountain View, California 94043
30 */
31
32#if defined(LIBC_SCCS) && !defined(lint)
33static char *sccsid2 = "@(#)svc_tcp.c 1.21 87/08/11 Copyr 1984 Sun Micro";
34static char *sccsid = "@(#)svc_tcp.c 2.2 88/08/01 4.0 RPCSRC";
35#endif
36#include <sys/cdefs.h>
37__FBSDID("$FreeBSD: head/sys/rpc/svc_vc.c 217242 2011-01-10 21:35:10Z rmacklem $");
38
39/*
40 * svc_vc.c, Server side for Connection Oriented based RPC.
41 *
42 * Actually implements two flavors of transporter -
43 * a tcp rendezvouser (a listner and connection establisher)
44 * and a record/tcp stream.
45 */
46
47#include <sys/param.h>
48#include <sys/lock.h>
49#include <sys/kernel.h>
50#include <sys/malloc.h>
51#include <sys/mbuf.h>
52#include <sys/mutex.h>
53#include <sys/proc.h>
54#include <sys/protosw.h>
55#include <sys/queue.h>
56#include <sys/socket.h>
57#include <sys/socketvar.h>
58#include <sys/sx.h>
59#include <sys/systm.h>
60#include <sys/uio.h>
61
62#include <net/vnet.h>
63
64#include <netinet/tcp.h>
65
66#include <rpc/rpc.h>
67
68#include <rpc/rpc_com.h>
69
70#include <security/mac/mac_framework.h>
71
72static bool_t svc_vc_rendezvous_recv(SVCXPRT *, struct rpc_msg *,
73 struct sockaddr **, struct mbuf **);
74static enum xprt_stat svc_vc_rendezvous_stat(SVCXPRT *);
75static void svc_vc_rendezvous_destroy(SVCXPRT *);
76static bool_t svc_vc_null(void);
77static void svc_vc_destroy(SVCXPRT *);
78static enum xprt_stat svc_vc_stat(SVCXPRT *);
79static bool_t svc_vc_recv(SVCXPRT *, struct rpc_msg *,
80 struct sockaddr **, struct mbuf **);
81static bool_t svc_vc_reply(SVCXPRT *, struct rpc_msg *,
82 struct sockaddr *, struct mbuf *);
83static bool_t svc_vc_control(SVCXPRT *xprt, const u_int rq, void *in);
84static bool_t svc_vc_rendezvous_control (SVCXPRT *xprt, const u_int rq,
85 void *in);
86static SVCXPRT *svc_vc_create_conn(SVCPOOL *pool, struct socket *so,
87 struct sockaddr *raddr);
88static int svc_vc_accept(struct socket *head, struct socket **sop);
89static int svc_vc_soupcall(struct socket *so, void *arg, int waitflag);
90
91static struct xp_ops svc_vc_rendezvous_ops = {
92 .xp_recv = svc_vc_rendezvous_recv,
93 .xp_stat = svc_vc_rendezvous_stat,
94 .xp_reply = (bool_t (*)(SVCXPRT *, struct rpc_msg *,
95 struct sockaddr *, struct mbuf *))svc_vc_null,
96 .xp_destroy = svc_vc_rendezvous_destroy,
97 .xp_control = svc_vc_rendezvous_control
98};
99
100static struct xp_ops svc_vc_ops = {
101 .xp_recv = svc_vc_recv,
102 .xp_stat = svc_vc_stat,
103 .xp_reply = svc_vc_reply,
104 .xp_destroy = svc_vc_destroy,
105 .xp_control = svc_vc_control
106};
107
108struct cf_conn { /* kept in xprt->xp_p1 for actual connection */
109 enum xprt_stat strm_stat;
110 struct mbuf *mpending; /* unparsed data read from the socket */
111 struct mbuf *mreq; /* current record being built from mpending */
112 uint32_t resid; /* number of bytes needed for fragment */
113 bool_t eor; /* reading last fragment of current record */
114};
115
116/*
117 * Usage:
118 * xprt = svc_vc_create(sock, send_buf_size, recv_buf_size);
119 *
120 * Creates, registers, and returns a (rpc) tcp based transporter.
121 * Once *xprt is initialized, it is registered as a transporter
122 * see (svc.h, xprt_register). This routine returns
123 * a NULL if a problem occurred.
124 *
125 * The filedescriptor passed in is expected to refer to a bound, but
126 * not yet connected socket.
127 *
128 * Since streams do buffered io similar to stdio, the caller can specify
129 * how big the send and receive buffers are via the second and third parms;
130 * 0 => use the system default.
131 */
132SVCXPRT *
133svc_vc_create(SVCPOOL *pool, struct socket *so, size_t sendsize,
134 size_t recvsize)
135{
136 SVCXPRT *xprt;
137 struct sockaddr* sa;
138 int error;
139
140 if (so->so_state & SS_ISCONNECTED) {
141 error = so->so_proto->pr_usrreqs->pru_peeraddr(so, &sa);
142 if (error)
143 return (NULL);
144 xprt = svc_vc_create_conn(pool, so, sa);
145 free(sa, M_SONAME);
146 return (xprt);
147 }
148
149 xprt = svc_xprt_alloc();
150 sx_init(&xprt->xp_lock, "xprt->xp_lock");
151 xprt->xp_pool = pool;
152 xprt->xp_socket = so;
153 xprt->xp_p1 = NULL;
154 xprt->xp_p2 = NULL;
155 xprt->xp_ops = &svc_vc_rendezvous_ops;
156
157 CURVNET_SET(so->so_vnet);
158 error = so->so_proto->pr_usrreqs->pru_sockaddr(so, &sa);
159 if (error) {
160 CURVNET_RESTORE();
161 goto cleanup_svc_vc_create;
162 }
163
164 memcpy(&xprt->xp_ltaddr, sa, sa->sa_len);
165 free(sa, M_SONAME);
166
167 xprt_register(xprt);
168
169 solisten(so, SOMAXCONN, curthread);
170 CURVNET_RESTORE();
171
172 SOCKBUF_LOCK(&so->so_rcv);
173 xprt->xp_upcallset = 1;
174 soupcall_set(so, SO_RCV, svc_vc_soupcall, xprt);
175 SOCKBUF_UNLOCK(&so->so_rcv);
176
177 return (xprt);
178cleanup_svc_vc_create:
179 if (xprt)
180 svc_xprt_free(xprt);
181 return (NULL);
182}
183
184/*
185 * Create a new transport for a socket optained via soaccept().
186 */
187SVCXPRT *
188svc_vc_create_conn(SVCPOOL *pool, struct socket *so, struct sockaddr *raddr)
189{
190 SVCXPRT *xprt = NULL;
191 struct cf_conn *cd = NULL;
192 struct sockaddr* sa = NULL;
193 struct sockopt opt;
194 int one = 1;
195 int error;
196
197 bzero(&opt, sizeof(struct sockopt));
198 opt.sopt_dir = SOPT_SET;
199 opt.sopt_level = SOL_SOCKET;
200 opt.sopt_name = SO_KEEPALIVE;
201 opt.sopt_val = &one;
202 opt.sopt_valsize = sizeof(one);
203 CURVNET_SET(so->so_vnet);
204 error = sosetopt(so, &opt);
205 if (error) {
206 CURVNET_RESTORE();
207 return (NULL);
208 }
209
210 if (so->so_proto->pr_protocol == IPPROTO_TCP) {
211 bzero(&opt, sizeof(struct sockopt));
212 opt.sopt_dir = SOPT_SET;
213 opt.sopt_level = IPPROTO_TCP;
214 opt.sopt_name = TCP_NODELAY;
215 opt.sopt_val = &one;
216 opt.sopt_valsize = sizeof(one);
217 error = sosetopt(so, &opt);
218 if (error) {
219 CURVNET_RESTORE();
220 return (NULL);
221 }
222 }
223 CURVNET_RESTORE();
224
225 cd = mem_alloc(sizeof(*cd));
226 cd->strm_stat = XPRT_IDLE;
227
228 xprt = svc_xprt_alloc();
229 sx_init(&xprt->xp_lock, "xprt->xp_lock");
230 xprt->xp_pool = pool;
231 xprt->xp_socket = so;
232 xprt->xp_p1 = cd;
233 xprt->xp_p2 = NULL;
234 xprt->xp_ops = &svc_vc_ops;
235
236 /*
237 * See http://www.connectathon.org/talks96/nfstcp.pdf - client
238 * has a 5 minute timer, server has a 6 minute timer.
239 */
240 xprt->xp_idletimeout = 6 * 60;
241
242 memcpy(&xprt->xp_rtaddr, raddr, raddr->sa_len);
243
244 error = so->so_proto->pr_usrreqs->pru_sockaddr(so, &sa);
245 if (error)
246 goto cleanup_svc_vc_create;
247
248 memcpy(&xprt->xp_ltaddr, sa, sa->sa_len);
249 free(sa, M_SONAME);
250
251 xprt_register(xprt);
252
253 SOCKBUF_LOCK(&so->so_rcv);
254 xprt->xp_upcallset = 1;
255 soupcall_set(so, SO_RCV, svc_vc_soupcall, xprt);
256 SOCKBUF_UNLOCK(&so->so_rcv);
257
258 /*
259 * Throw the transport into the active list in case it already
260 * has some data buffered.
261 */
262 sx_xlock(&xprt->xp_lock);
263 xprt_active(xprt);
264 sx_xunlock(&xprt->xp_lock);
265
266 return (xprt);
267cleanup_svc_vc_create:
268 if (xprt) {
269 mem_free(xprt, sizeof(*xprt));
270 }
271 if (cd)
272 mem_free(cd, sizeof(*cd));
273 return (NULL);
274}
275
276/*
277 * This does all of the accept except the final call to soaccept. The
278 * caller will call soaccept after dropping its locks (soaccept may
279 * call malloc).
280 */
281int
282svc_vc_accept(struct socket *head, struct socket **sop)
283{
284 int error = 0;
285 struct socket *so;
286
287 if ((head->so_options & SO_ACCEPTCONN) == 0) {
288 error = EINVAL;
289 goto done;
290 }
291#ifdef MAC
292 error = mac_socket_check_accept(curthread->td_ucred, head);
293 if (error != 0)
294 goto done;
295#endif
296 ACCEPT_LOCK();
297 if (TAILQ_EMPTY(&head->so_comp)) {
298 ACCEPT_UNLOCK();
299 error = EWOULDBLOCK;
300 goto done;
301 }
302 so = TAILQ_FIRST(&head->so_comp);
303 KASSERT(!(so->so_qstate & SQ_INCOMP), ("svc_vc_accept: so SQ_INCOMP"));
304 KASSERT(so->so_qstate & SQ_COMP, ("svc_vc_accept: so not SQ_COMP"));
305
306 /*
307 * Before changing the flags on the socket, we have to bump the
308 * reference count. Otherwise, if the protocol calls sofree(),
309 * the socket will be released due to a zero refcount.
310 * XXX might not need soref() since this is simpler than kern_accept.
311 */
312 SOCK_LOCK(so); /* soref() and so_state update */
313 soref(so); /* file descriptor reference */
314
315 TAILQ_REMOVE(&head->so_comp, so, so_list);
316 head->so_qlen--;
317 so->so_state |= (head->so_state & SS_NBIO);
318 so->so_qstate &= ~SQ_COMP;
319 so->so_head = NULL;
320
321 SOCK_UNLOCK(so);
322 ACCEPT_UNLOCK();
323
324 *sop = so;
325
326 /* connection has been removed from the listen queue */
327 KNOTE_UNLOCKED(&head->so_rcv.sb_sel.si_note, 0);
328done:
329 return (error);
330}
331
332/*ARGSUSED*/
333static bool_t
334svc_vc_rendezvous_recv(SVCXPRT *xprt, struct rpc_msg *msg,
335 struct sockaddr **addrp, struct mbuf **mp)
336{
337 struct socket *so = NULL;
338 struct sockaddr *sa = NULL;
339 int error;
340 SVCXPRT *new_xprt;
341
342 /*
343 * The socket upcall calls xprt_active() which will eventually
344 * cause the server to call us here. We attempt to accept a
345 * connection from the socket and turn it into a new
346 * transport. If the accept fails, we have drained all pending
347 * connections so we call xprt_inactive().
348 */
349 sx_xlock(&xprt->xp_lock);
350
351 error = svc_vc_accept(xprt->xp_socket, &so);
352
353 if (error == EWOULDBLOCK) {
354 /*
355 * We must re-test for new connections after taking
356 * the lock to protect us in the case where a new
357 * connection arrives after our call to accept fails
358 * with EWOULDBLOCK. The pool lock protects us from
359 * racing the upcall after our TAILQ_EMPTY() call
360 * returns false.
361 */
362 ACCEPT_LOCK();
363 mtx_lock(&xprt->xp_pool->sp_lock);
364 if (TAILQ_EMPTY(&xprt->xp_socket->so_comp))
365 xprt_inactive_locked(xprt);
366 mtx_unlock(&xprt->xp_pool->sp_lock);
367 ACCEPT_UNLOCK();
368 sx_xunlock(&xprt->xp_lock);
369 return (FALSE);
370 }
371
372 if (error) {
373 SOCKBUF_LOCK(&xprt->xp_socket->so_rcv);
374 if (xprt->xp_upcallset) {
375 xprt->xp_upcallset = 0;
376 soupcall_clear(xprt->xp_socket, SO_RCV);
377 }
378 SOCKBUF_UNLOCK(&xprt->xp_socket->so_rcv);
379 xprt_inactive(xprt);
380 sx_xunlock(&xprt->xp_lock);
381 return (FALSE);
382 }
383
384 sx_xunlock(&xprt->xp_lock);
385
386 sa = 0;
387 error = soaccept(so, &sa);
388
389 if (error) {
390 /*
391 * XXX not sure if I need to call sofree or soclose here.
392 */
393 if (sa)
394 free(sa, M_SONAME);
395 return (FALSE);
396 }
397
398 /*
399 * svc_vc_create_conn will call xprt_register - we don't need
400 * to do anything with the new connection except derefence it.
401 */
402 new_xprt = svc_vc_create_conn(xprt->xp_pool, so, sa);
403 if (!new_xprt) {
404 soclose(so);
405 } else {
406 SVC_RELEASE(new_xprt);
407 }
408
409 free(sa, M_SONAME);
410
411 return (FALSE); /* there is never an rpc msg to be processed */
412}
413
414/*ARGSUSED*/
415static enum xprt_stat
416svc_vc_rendezvous_stat(SVCXPRT *xprt)
417{
418
419 return (XPRT_IDLE);
420}
421
422static void
423svc_vc_destroy_common(SVCXPRT *xprt)
424{
425 SOCKBUF_LOCK(&xprt->xp_socket->so_rcv);
426 if (xprt->xp_upcallset) {
427 xprt->xp_upcallset = 0;
428 soupcall_clear(xprt->xp_socket, SO_RCV);
429 }
430 SOCKBUF_UNLOCK(&xprt->xp_socket->so_rcv);
431
432 sx_destroy(&xprt->xp_lock);
433 if (xprt->xp_socket)
434 (void)soclose(xprt->xp_socket);
435
436 if (xprt->xp_netid)
437 (void) mem_free(xprt->xp_netid, strlen(xprt->xp_netid) + 1);
438 svc_xprt_free(xprt);
439}
440
441static void
442svc_vc_rendezvous_destroy(SVCXPRT *xprt)
443{
444
445 svc_vc_destroy_common(xprt);
446}
447
448static void
449svc_vc_destroy(SVCXPRT *xprt)
450{
451 struct cf_conn *cd = (struct cf_conn *)xprt->xp_p1;
452
453 svc_vc_destroy_common(xprt);
454
455 if (cd->mreq)
456 m_freem(cd->mreq);
457 if (cd->mpending)
458 m_freem(cd->mpending);
459 mem_free(cd, sizeof(*cd));
460}
461
462/*ARGSUSED*/
463static bool_t
464svc_vc_control(SVCXPRT *xprt, const u_int rq, void *in)
465{
466 return (FALSE);
467}
468
469static bool_t
470svc_vc_rendezvous_control(SVCXPRT *xprt, const u_int rq, void *in)
471{
472
473 return (FALSE);
474}
475
476static enum xprt_stat
477svc_vc_stat(SVCXPRT *xprt)
478{
479 struct cf_conn *cd;
480 struct mbuf *m;
481 size_t n;
482
483 cd = (struct cf_conn *)(xprt->xp_p1);
484
485 if (cd->strm_stat == XPRT_DIED)
486 return (XPRT_DIED);
487
488 /*
489 * Return XPRT_MOREREQS if we have buffered data and we are
490 * mid-record or if we have enough data for a record
491 * marker. Since this is only a hint, we read mpending and
492 * resid outside the lock. We do need to take the lock if we
493 * have to traverse the mbuf chain.
494 */
495 if (cd->mpending) {
496 if (cd->resid)
497 return (XPRT_MOREREQS);
498 n = 0;
499 sx_xlock(&xprt->xp_lock);
500 m = cd->mpending;
501 while (m && n < sizeof(uint32_t)) {
502 n += m->m_len;
503 m = m->m_next;
504 }
505 sx_xunlock(&xprt->xp_lock);
506 if (n >= sizeof(uint32_t))
507 return (XPRT_MOREREQS);
508 }
509
510 if (soreadable(xprt->xp_socket))
511 return (XPRT_MOREREQS);
512
513 return (XPRT_IDLE);
514}
515
516static bool_t
517svc_vc_recv(SVCXPRT *xprt, struct rpc_msg *msg,
518 struct sockaddr **addrp, struct mbuf **mp)
519{
520 struct cf_conn *cd = (struct cf_conn *) xprt->xp_p1;
521 struct uio uio;
522 struct mbuf *m;
523 XDR xdrs;
524 int error, rcvflag;
525
526 /*
527 * Serialise access to the socket and our own record parsing
528 * state.
529 */
530 sx_xlock(&xprt->xp_lock);
531
532 for (;;) {
533 /*
534 * If we have an mbuf chain in cd->mpending, try to parse a
535 * record from it, leaving the result in cd->mreq. If we don't
536 * have a complete record, leave the partial result in
537 * cd->mreq and try to read more from the socket.
538 */
539 if (cd->mpending) {
540 /*
541 * If cd->resid is non-zero, we have part of the
542 * record already, otherwise we are expecting a record
543 * marker.
544 */
545 if (!cd->resid) {
546 /*
547 * See if there is enough data buffered to
548 * make up a record marker. Make sure we can
549 * handle the case where the record marker is
550 * split across more than one mbuf.
551 */
552 size_t n = 0;
553 uint32_t header;
554
555 m = cd->mpending;
556 while (n < sizeof(uint32_t) && m) {
557 n += m->m_len;
558 m = m->m_next;
559 }
560 if (n < sizeof(uint32_t))
561 goto readmore;
562 m_copydata(cd->mpending, 0, sizeof(header),
563 (char *)&header);
564 header = ntohl(header);
565 cd->eor = (header & 0x80000000) != 0;
566 cd->resid = header & 0x7fffffff;
567 m_adj(cd->mpending, sizeof(uint32_t));
568 }
569
570 /*
571 * Start pulling off mbufs from cd->mpending
572 * until we either have a complete record or
573 * we run out of data. We use m_split to pull
574 * data - it will pull as much as possible and
575 * split the last mbuf if necessary.
576 */
577 while (cd->mpending && cd->resid) {
578 m = cd->mpending;
579 if (cd->mpending->m_next
580 || cd->mpending->m_len > cd->resid)
581 cd->mpending = m_split(cd->mpending,
582 cd->resid, M_WAIT);
583 else
584 cd->mpending = NULL;
585 if (cd->mreq)
586 m_last(cd->mreq)->m_next = m;
587 else
588 cd->mreq = m;
589 while (m) {
590 cd->resid -= m->m_len;
591 m = m->m_next;
592 }
593 }
594
595 /*
596 * If cd->resid is zero now, we have managed to
597 * receive a record fragment from the stream. Check
598 * for the end-of-record mark to see if we need more.
599 */
600 if (cd->resid == 0) {
601 if (!cd->eor)
602 continue;
603
604 /*
605 * Success - we have a complete record in
606 * cd->mreq.
607 */
608 xdrmbuf_create(&xdrs, cd->mreq, XDR_DECODE);
609 cd->mreq = NULL;
610 sx_xunlock(&xprt->xp_lock);
611
612 if (! xdr_callmsg(&xdrs, msg)) {
613 XDR_DESTROY(&xdrs);
614 return (FALSE);
615 }
616
617 *addrp = NULL;
618 *mp = xdrmbuf_getall(&xdrs);
619 XDR_DESTROY(&xdrs);
620
621 return (TRUE);
622 }
623 }
624
625 readmore:
626 /*
627 * The socket upcall calls xprt_active() which will eventually
628 * cause the server to call us here. We attempt to
629 * read as much as possible from the socket and put
630 * the result in cd->mpending. If the read fails,
631 * we have drained both cd->mpending and the socket so
632 * we can call xprt_inactive().
633 */
634 uio.uio_resid = 1000000000;
635 uio.uio_td = curthread;
636 m = NULL;
637 rcvflag = MSG_DONTWAIT;
638 CURVNET_SET(xprt->xp_socket->so_vnet);
639 error = soreceive(xprt->xp_socket, NULL, &uio, &m, NULL,
640 &rcvflag);
641 CURVNET_RESTORE();
642
643 if (error == EWOULDBLOCK) {
644 /*
645 * We must re-test for readability after
646 * taking the lock to protect us in the case
647 * where a new packet arrives on the socket
648 * after our call to soreceive fails with
649 * EWOULDBLOCK. The pool lock protects us from
650 * racing the upcall after our soreadable()
651 * call returns false.
652 */
653 mtx_lock(&xprt->xp_pool->sp_lock);
654 if (!soreadable(xprt->xp_socket))
655 xprt_inactive_locked(xprt);
656 mtx_unlock(&xprt->xp_pool->sp_lock);
657 sx_xunlock(&xprt->xp_lock);
658 return (FALSE);
659 }
660
661 if (error) {
662 SOCKBUF_LOCK(&xprt->xp_socket->so_rcv);
663 if (xprt->xp_upcallset) {
664 xprt->xp_upcallset = 0;
665 soupcall_clear(xprt->xp_socket, SO_RCV);
666 }
667 SOCKBUF_UNLOCK(&xprt->xp_socket->so_rcv);
668 xprt_inactive(xprt);
669 cd->strm_stat = XPRT_DIED;
670 sx_xunlock(&xprt->xp_lock);
671 return (FALSE);
672 }
673
674 if (!m) {
675 /*
676 * EOF - the other end has closed the socket.
677 */
678 xprt_inactive(xprt);
679 cd->strm_stat = XPRT_DIED;
680 sx_xunlock(&xprt->xp_lock);
681 return (FALSE);
682 }
683
684 if (cd->mpending)
685 m_last(cd->mpending)->m_next = m;
686 else
687 cd->mpending = m;
688 }
689}
690
691static bool_t
692svc_vc_reply(SVCXPRT *xprt, struct rpc_msg *msg,
693 struct sockaddr *addr, struct mbuf *m)
694{
695 XDR xdrs;
696 struct mbuf *mrep;
697 bool_t stat = TRUE;
698 int error;
699
700 /*
701 * Leave space for record mark.
702 */
703 MGETHDR(mrep, M_WAIT, MT_DATA);
704 mrep->m_len = 0;
705 mrep->m_data += sizeof(uint32_t);
706
707 xdrmbuf_create(&xdrs, mrep, XDR_ENCODE);
708
709 if (msg->rm_reply.rp_stat == MSG_ACCEPTED &&
710 msg->rm_reply.rp_acpt.ar_stat == SUCCESS) {
711 if (!xdr_replymsg(&xdrs, msg))
712 stat = FALSE;
713 else
714 xdrmbuf_append(&xdrs, m);
715 } else {
716 stat = xdr_replymsg(&xdrs, msg);
717 }
718
719 if (stat) {
720 m_fixhdr(mrep);
721
722 /*
723 * Prepend a record marker containing the reply length.
724 */
725 M_PREPEND(mrep, sizeof(uint32_t), M_WAIT);
726 *mtod(mrep, uint32_t *) =
727 htonl(0x80000000 | (mrep->m_pkthdr.len
728 - sizeof(uint32_t)));
729 error = sosend(xprt->xp_socket, NULL, NULL, mrep, NULL,
730 0, curthread);
731 if (!error) {
732 stat = TRUE;
733 }
734 } else {
735 m_freem(mrep);
736 }
737
738 XDR_DESTROY(&xdrs);
739 xprt->xp_p2 = NULL;
740
741 return (stat);
742}
743
744static bool_t
745svc_vc_null()
746{
747
748 return (FALSE);
749}
750
751static int
752svc_vc_soupcall(struct socket *so, void *arg, int waitflag)
753{
754 SVCXPRT *xprt = (SVCXPRT *) arg;
755
756 xprt_active(xprt);
757 return (SU_OK);
758}
759
760#if 0
761/*
762 * Get the effective UID of the sending process. Used by rpcbind, keyserv
763 * and rpc.yppasswdd on AF_LOCAL.
764 */
765int
766__rpc_get_local_uid(SVCXPRT *transp, uid_t *uid) {
767 int sock, ret;
768 gid_t egid;
769 uid_t euid;
770 struct sockaddr *sa;
771
772 sock = transp->xp_fd;
773 sa = (struct sockaddr *)transp->xp_rtaddr;
774 if (sa->sa_family == AF_LOCAL) {
775 ret = getpeereid(sock, &euid, &egid);
776 if (ret == 0)
777 *uid = euid;
778 return (ret);
779 } else
780 return (-1);
781}
782#endif
2
3/*
4 * Sun RPC is a product of Sun Microsystems, Inc. and is provided for
5 * unrestricted use provided that this legend is included on all tape
6 * media and as a part of the software program in whole or part. Users
7 * may copy or modify Sun RPC without charge, but are not authorized
8 * to license or distribute it to anyone else except as part of a product or
9 * program developed by the user.
10 *
11 * SUN RPC IS PROVIDED AS IS WITH NO WARRANTIES OF ANY KIND INCLUDING THE
12 * WARRANTIES OF DESIGN, MERCHANTIBILITY AND FITNESS FOR A PARTICULAR
13 * PURPOSE, OR ARISING FROM A COURSE OF DEALING, USAGE OR TRADE PRACTICE.
14 *
15 * Sun RPC is provided with no support and without any obligation on the
16 * part of Sun Microsystems, Inc. to assist in its use, correction,
17 * modification or enhancement.
18 *
19 * SUN MICROSYSTEMS, INC. SHALL HAVE NO LIABILITY WITH RESPECT TO THE
20 * INFRINGEMENT OF COPYRIGHTS, TRADE SECRETS OR ANY PATENTS BY SUN RPC
21 * OR ANY PART THEREOF.
22 *
23 * In no event will Sun Microsystems, Inc. be liable for any lost revenue
24 * or profits or other special, indirect and consequential damages, even if
25 * Sun has been advised of the possibility of such damages.
26 *
27 * Sun Microsystems, Inc.
28 * 2550 Garcia Avenue
29 * Mountain View, California 94043
30 */
31
32#if defined(LIBC_SCCS) && !defined(lint)
33static char *sccsid2 = "@(#)svc_tcp.c 1.21 87/08/11 Copyr 1984 Sun Micro";
34static char *sccsid = "@(#)svc_tcp.c 2.2 88/08/01 4.0 RPCSRC";
35#endif
36#include <sys/cdefs.h>
37__FBSDID("$FreeBSD: head/sys/rpc/svc_vc.c 217242 2011-01-10 21:35:10Z rmacklem $");
38
39/*
40 * svc_vc.c, Server side for Connection Oriented based RPC.
41 *
42 * Actually implements two flavors of transporter -
43 * a tcp rendezvouser (a listner and connection establisher)
44 * and a record/tcp stream.
45 */
46
47#include <sys/param.h>
48#include <sys/lock.h>
49#include <sys/kernel.h>
50#include <sys/malloc.h>
51#include <sys/mbuf.h>
52#include <sys/mutex.h>
53#include <sys/proc.h>
54#include <sys/protosw.h>
55#include <sys/queue.h>
56#include <sys/socket.h>
57#include <sys/socketvar.h>
58#include <sys/sx.h>
59#include <sys/systm.h>
60#include <sys/uio.h>
61
62#include <net/vnet.h>
63
64#include <netinet/tcp.h>
65
66#include <rpc/rpc.h>
67
68#include <rpc/rpc_com.h>
69
70#include <security/mac/mac_framework.h>
71
72static bool_t svc_vc_rendezvous_recv(SVCXPRT *, struct rpc_msg *,
73 struct sockaddr **, struct mbuf **);
74static enum xprt_stat svc_vc_rendezvous_stat(SVCXPRT *);
75static void svc_vc_rendezvous_destroy(SVCXPRT *);
76static bool_t svc_vc_null(void);
77static void svc_vc_destroy(SVCXPRT *);
78static enum xprt_stat svc_vc_stat(SVCXPRT *);
79static bool_t svc_vc_recv(SVCXPRT *, struct rpc_msg *,
80 struct sockaddr **, struct mbuf **);
81static bool_t svc_vc_reply(SVCXPRT *, struct rpc_msg *,
82 struct sockaddr *, struct mbuf *);
83static bool_t svc_vc_control(SVCXPRT *xprt, const u_int rq, void *in);
84static bool_t svc_vc_rendezvous_control (SVCXPRT *xprt, const u_int rq,
85 void *in);
86static SVCXPRT *svc_vc_create_conn(SVCPOOL *pool, struct socket *so,
87 struct sockaddr *raddr);
88static int svc_vc_accept(struct socket *head, struct socket **sop);
89static int svc_vc_soupcall(struct socket *so, void *arg, int waitflag);
90
91static struct xp_ops svc_vc_rendezvous_ops = {
92 .xp_recv = svc_vc_rendezvous_recv,
93 .xp_stat = svc_vc_rendezvous_stat,
94 .xp_reply = (bool_t (*)(SVCXPRT *, struct rpc_msg *,
95 struct sockaddr *, struct mbuf *))svc_vc_null,
96 .xp_destroy = svc_vc_rendezvous_destroy,
97 .xp_control = svc_vc_rendezvous_control
98};
99
100static struct xp_ops svc_vc_ops = {
101 .xp_recv = svc_vc_recv,
102 .xp_stat = svc_vc_stat,
103 .xp_reply = svc_vc_reply,
104 .xp_destroy = svc_vc_destroy,
105 .xp_control = svc_vc_control
106};
107
108struct cf_conn { /* kept in xprt->xp_p1 for actual connection */
109 enum xprt_stat strm_stat;
110 struct mbuf *mpending; /* unparsed data read from the socket */
111 struct mbuf *mreq; /* current record being built from mpending */
112 uint32_t resid; /* number of bytes needed for fragment */
113 bool_t eor; /* reading last fragment of current record */
114};
115
116/*
117 * Usage:
118 * xprt = svc_vc_create(sock, send_buf_size, recv_buf_size);
119 *
120 * Creates, registers, and returns a (rpc) tcp based transporter.
121 * Once *xprt is initialized, it is registered as a transporter
122 * see (svc.h, xprt_register). This routine returns
123 * a NULL if a problem occurred.
124 *
125 * The filedescriptor passed in is expected to refer to a bound, but
126 * not yet connected socket.
127 *
128 * Since streams do buffered io similar to stdio, the caller can specify
129 * how big the send and receive buffers are via the second and third parms;
130 * 0 => use the system default.
131 */
132SVCXPRT *
133svc_vc_create(SVCPOOL *pool, struct socket *so, size_t sendsize,
134 size_t recvsize)
135{
136 SVCXPRT *xprt;
137 struct sockaddr* sa;
138 int error;
139
140 if (so->so_state & SS_ISCONNECTED) {
141 error = so->so_proto->pr_usrreqs->pru_peeraddr(so, &sa);
142 if (error)
143 return (NULL);
144 xprt = svc_vc_create_conn(pool, so, sa);
145 free(sa, M_SONAME);
146 return (xprt);
147 }
148
149 xprt = svc_xprt_alloc();
150 sx_init(&xprt->xp_lock, "xprt->xp_lock");
151 xprt->xp_pool = pool;
152 xprt->xp_socket = so;
153 xprt->xp_p1 = NULL;
154 xprt->xp_p2 = NULL;
155 xprt->xp_ops = &svc_vc_rendezvous_ops;
156
157 CURVNET_SET(so->so_vnet);
158 error = so->so_proto->pr_usrreqs->pru_sockaddr(so, &sa);
159 if (error) {
160 CURVNET_RESTORE();
161 goto cleanup_svc_vc_create;
162 }
163
164 memcpy(&xprt->xp_ltaddr, sa, sa->sa_len);
165 free(sa, M_SONAME);
166
167 xprt_register(xprt);
168
169 solisten(so, SOMAXCONN, curthread);
170 CURVNET_RESTORE();
171
172 SOCKBUF_LOCK(&so->so_rcv);
173 xprt->xp_upcallset = 1;
174 soupcall_set(so, SO_RCV, svc_vc_soupcall, xprt);
175 SOCKBUF_UNLOCK(&so->so_rcv);
176
177 return (xprt);
178cleanup_svc_vc_create:
179 if (xprt)
180 svc_xprt_free(xprt);
181 return (NULL);
182}
183
184/*
185 * Create a new transport for a socket optained via soaccept().
186 */
187SVCXPRT *
188svc_vc_create_conn(SVCPOOL *pool, struct socket *so, struct sockaddr *raddr)
189{
190 SVCXPRT *xprt = NULL;
191 struct cf_conn *cd = NULL;
192 struct sockaddr* sa = NULL;
193 struct sockopt opt;
194 int one = 1;
195 int error;
196
197 bzero(&opt, sizeof(struct sockopt));
198 opt.sopt_dir = SOPT_SET;
199 opt.sopt_level = SOL_SOCKET;
200 opt.sopt_name = SO_KEEPALIVE;
201 opt.sopt_val = &one;
202 opt.sopt_valsize = sizeof(one);
203 CURVNET_SET(so->so_vnet);
204 error = sosetopt(so, &opt);
205 if (error) {
206 CURVNET_RESTORE();
207 return (NULL);
208 }
209
210 if (so->so_proto->pr_protocol == IPPROTO_TCP) {
211 bzero(&opt, sizeof(struct sockopt));
212 opt.sopt_dir = SOPT_SET;
213 opt.sopt_level = IPPROTO_TCP;
214 opt.sopt_name = TCP_NODELAY;
215 opt.sopt_val = &one;
216 opt.sopt_valsize = sizeof(one);
217 error = sosetopt(so, &opt);
218 if (error) {
219 CURVNET_RESTORE();
220 return (NULL);
221 }
222 }
223 CURVNET_RESTORE();
224
225 cd = mem_alloc(sizeof(*cd));
226 cd->strm_stat = XPRT_IDLE;
227
228 xprt = svc_xprt_alloc();
229 sx_init(&xprt->xp_lock, "xprt->xp_lock");
230 xprt->xp_pool = pool;
231 xprt->xp_socket = so;
232 xprt->xp_p1 = cd;
233 xprt->xp_p2 = NULL;
234 xprt->xp_ops = &svc_vc_ops;
235
236 /*
237 * See http://www.connectathon.org/talks96/nfstcp.pdf - client
238 * has a 5 minute timer, server has a 6 minute timer.
239 */
240 xprt->xp_idletimeout = 6 * 60;
241
242 memcpy(&xprt->xp_rtaddr, raddr, raddr->sa_len);
243
244 error = so->so_proto->pr_usrreqs->pru_sockaddr(so, &sa);
245 if (error)
246 goto cleanup_svc_vc_create;
247
248 memcpy(&xprt->xp_ltaddr, sa, sa->sa_len);
249 free(sa, M_SONAME);
250
251 xprt_register(xprt);
252
253 SOCKBUF_LOCK(&so->so_rcv);
254 xprt->xp_upcallset = 1;
255 soupcall_set(so, SO_RCV, svc_vc_soupcall, xprt);
256 SOCKBUF_UNLOCK(&so->so_rcv);
257
258 /*
259 * Throw the transport into the active list in case it already
260 * has some data buffered.
261 */
262 sx_xlock(&xprt->xp_lock);
263 xprt_active(xprt);
264 sx_xunlock(&xprt->xp_lock);
265
266 return (xprt);
267cleanup_svc_vc_create:
268 if (xprt) {
269 mem_free(xprt, sizeof(*xprt));
270 }
271 if (cd)
272 mem_free(cd, sizeof(*cd));
273 return (NULL);
274}
275
276/*
277 * This does all of the accept except the final call to soaccept. The
278 * caller will call soaccept after dropping its locks (soaccept may
279 * call malloc).
280 */
281int
282svc_vc_accept(struct socket *head, struct socket **sop)
283{
284 int error = 0;
285 struct socket *so;
286
287 if ((head->so_options & SO_ACCEPTCONN) == 0) {
288 error = EINVAL;
289 goto done;
290 }
291#ifdef MAC
292 error = mac_socket_check_accept(curthread->td_ucred, head);
293 if (error != 0)
294 goto done;
295#endif
296 ACCEPT_LOCK();
297 if (TAILQ_EMPTY(&head->so_comp)) {
298 ACCEPT_UNLOCK();
299 error = EWOULDBLOCK;
300 goto done;
301 }
302 so = TAILQ_FIRST(&head->so_comp);
303 KASSERT(!(so->so_qstate & SQ_INCOMP), ("svc_vc_accept: so SQ_INCOMP"));
304 KASSERT(so->so_qstate & SQ_COMP, ("svc_vc_accept: so not SQ_COMP"));
305
306 /*
307 * Before changing the flags on the socket, we have to bump the
308 * reference count. Otherwise, if the protocol calls sofree(),
309 * the socket will be released due to a zero refcount.
310 * XXX might not need soref() since this is simpler than kern_accept.
311 */
312 SOCK_LOCK(so); /* soref() and so_state update */
313 soref(so); /* file descriptor reference */
314
315 TAILQ_REMOVE(&head->so_comp, so, so_list);
316 head->so_qlen--;
317 so->so_state |= (head->so_state & SS_NBIO);
318 so->so_qstate &= ~SQ_COMP;
319 so->so_head = NULL;
320
321 SOCK_UNLOCK(so);
322 ACCEPT_UNLOCK();
323
324 *sop = so;
325
326 /* connection has been removed from the listen queue */
327 KNOTE_UNLOCKED(&head->so_rcv.sb_sel.si_note, 0);
328done:
329 return (error);
330}
331
332/*ARGSUSED*/
333static bool_t
334svc_vc_rendezvous_recv(SVCXPRT *xprt, struct rpc_msg *msg,
335 struct sockaddr **addrp, struct mbuf **mp)
336{
337 struct socket *so = NULL;
338 struct sockaddr *sa = NULL;
339 int error;
340 SVCXPRT *new_xprt;
341
342 /*
343 * The socket upcall calls xprt_active() which will eventually
344 * cause the server to call us here. We attempt to accept a
345 * connection from the socket and turn it into a new
346 * transport. If the accept fails, we have drained all pending
347 * connections so we call xprt_inactive().
348 */
349 sx_xlock(&xprt->xp_lock);
350
351 error = svc_vc_accept(xprt->xp_socket, &so);
352
353 if (error == EWOULDBLOCK) {
354 /*
355 * We must re-test for new connections after taking
356 * the lock to protect us in the case where a new
357 * connection arrives after our call to accept fails
358 * with EWOULDBLOCK. The pool lock protects us from
359 * racing the upcall after our TAILQ_EMPTY() call
360 * returns false.
361 */
362 ACCEPT_LOCK();
363 mtx_lock(&xprt->xp_pool->sp_lock);
364 if (TAILQ_EMPTY(&xprt->xp_socket->so_comp))
365 xprt_inactive_locked(xprt);
366 mtx_unlock(&xprt->xp_pool->sp_lock);
367 ACCEPT_UNLOCK();
368 sx_xunlock(&xprt->xp_lock);
369 return (FALSE);
370 }
371
372 if (error) {
373 SOCKBUF_LOCK(&xprt->xp_socket->so_rcv);
374 if (xprt->xp_upcallset) {
375 xprt->xp_upcallset = 0;
376 soupcall_clear(xprt->xp_socket, SO_RCV);
377 }
378 SOCKBUF_UNLOCK(&xprt->xp_socket->so_rcv);
379 xprt_inactive(xprt);
380 sx_xunlock(&xprt->xp_lock);
381 return (FALSE);
382 }
383
384 sx_xunlock(&xprt->xp_lock);
385
386 sa = 0;
387 error = soaccept(so, &sa);
388
389 if (error) {
390 /*
391 * XXX not sure if I need to call sofree or soclose here.
392 */
393 if (sa)
394 free(sa, M_SONAME);
395 return (FALSE);
396 }
397
398 /*
399 * svc_vc_create_conn will call xprt_register - we don't need
400 * to do anything with the new connection except derefence it.
401 */
402 new_xprt = svc_vc_create_conn(xprt->xp_pool, so, sa);
403 if (!new_xprt) {
404 soclose(so);
405 } else {
406 SVC_RELEASE(new_xprt);
407 }
408
409 free(sa, M_SONAME);
410
411 return (FALSE); /* there is never an rpc msg to be processed */
412}
413
414/*ARGSUSED*/
415static enum xprt_stat
416svc_vc_rendezvous_stat(SVCXPRT *xprt)
417{
418
419 return (XPRT_IDLE);
420}
421
422static void
423svc_vc_destroy_common(SVCXPRT *xprt)
424{
425 SOCKBUF_LOCK(&xprt->xp_socket->so_rcv);
426 if (xprt->xp_upcallset) {
427 xprt->xp_upcallset = 0;
428 soupcall_clear(xprt->xp_socket, SO_RCV);
429 }
430 SOCKBUF_UNLOCK(&xprt->xp_socket->so_rcv);
431
432 sx_destroy(&xprt->xp_lock);
433 if (xprt->xp_socket)
434 (void)soclose(xprt->xp_socket);
435
436 if (xprt->xp_netid)
437 (void) mem_free(xprt->xp_netid, strlen(xprt->xp_netid) + 1);
438 svc_xprt_free(xprt);
439}
440
441static void
442svc_vc_rendezvous_destroy(SVCXPRT *xprt)
443{
444
445 svc_vc_destroy_common(xprt);
446}
447
448static void
449svc_vc_destroy(SVCXPRT *xprt)
450{
451 struct cf_conn *cd = (struct cf_conn *)xprt->xp_p1;
452
453 svc_vc_destroy_common(xprt);
454
455 if (cd->mreq)
456 m_freem(cd->mreq);
457 if (cd->mpending)
458 m_freem(cd->mpending);
459 mem_free(cd, sizeof(*cd));
460}
461
462/*ARGSUSED*/
463static bool_t
464svc_vc_control(SVCXPRT *xprt, const u_int rq, void *in)
465{
466 return (FALSE);
467}
468
469static bool_t
470svc_vc_rendezvous_control(SVCXPRT *xprt, const u_int rq, void *in)
471{
472
473 return (FALSE);
474}
475
476static enum xprt_stat
477svc_vc_stat(SVCXPRT *xprt)
478{
479 struct cf_conn *cd;
480 struct mbuf *m;
481 size_t n;
482
483 cd = (struct cf_conn *)(xprt->xp_p1);
484
485 if (cd->strm_stat == XPRT_DIED)
486 return (XPRT_DIED);
487
488 /*
489 * Return XPRT_MOREREQS if we have buffered data and we are
490 * mid-record or if we have enough data for a record
491 * marker. Since this is only a hint, we read mpending and
492 * resid outside the lock. We do need to take the lock if we
493 * have to traverse the mbuf chain.
494 */
495 if (cd->mpending) {
496 if (cd->resid)
497 return (XPRT_MOREREQS);
498 n = 0;
499 sx_xlock(&xprt->xp_lock);
500 m = cd->mpending;
501 while (m && n < sizeof(uint32_t)) {
502 n += m->m_len;
503 m = m->m_next;
504 }
505 sx_xunlock(&xprt->xp_lock);
506 if (n >= sizeof(uint32_t))
507 return (XPRT_MOREREQS);
508 }
509
510 if (soreadable(xprt->xp_socket))
511 return (XPRT_MOREREQS);
512
513 return (XPRT_IDLE);
514}
515
516static bool_t
517svc_vc_recv(SVCXPRT *xprt, struct rpc_msg *msg,
518 struct sockaddr **addrp, struct mbuf **mp)
519{
520 struct cf_conn *cd = (struct cf_conn *) xprt->xp_p1;
521 struct uio uio;
522 struct mbuf *m;
523 XDR xdrs;
524 int error, rcvflag;
525
526 /*
527 * Serialise access to the socket and our own record parsing
528 * state.
529 */
530 sx_xlock(&xprt->xp_lock);
531
532 for (;;) {
533 /*
534 * If we have an mbuf chain in cd->mpending, try to parse a
535 * record from it, leaving the result in cd->mreq. If we don't
536 * have a complete record, leave the partial result in
537 * cd->mreq and try to read more from the socket.
538 */
539 if (cd->mpending) {
540 /*
541 * If cd->resid is non-zero, we have part of the
542 * record already, otherwise we are expecting a record
543 * marker.
544 */
545 if (!cd->resid) {
546 /*
547 * See if there is enough data buffered to
548 * make up a record marker. Make sure we can
549 * handle the case where the record marker is
550 * split across more than one mbuf.
551 */
552 size_t n = 0;
553 uint32_t header;
554
555 m = cd->mpending;
556 while (n < sizeof(uint32_t) && m) {
557 n += m->m_len;
558 m = m->m_next;
559 }
560 if (n < sizeof(uint32_t))
561 goto readmore;
562 m_copydata(cd->mpending, 0, sizeof(header),
563 (char *)&header);
564 header = ntohl(header);
565 cd->eor = (header & 0x80000000) != 0;
566 cd->resid = header & 0x7fffffff;
567 m_adj(cd->mpending, sizeof(uint32_t));
568 }
569
570 /*
571 * Start pulling off mbufs from cd->mpending
572 * until we either have a complete record or
573 * we run out of data. We use m_split to pull
574 * data - it will pull as much as possible and
575 * split the last mbuf if necessary.
576 */
577 while (cd->mpending && cd->resid) {
578 m = cd->mpending;
579 if (cd->mpending->m_next
580 || cd->mpending->m_len > cd->resid)
581 cd->mpending = m_split(cd->mpending,
582 cd->resid, M_WAIT);
583 else
584 cd->mpending = NULL;
585 if (cd->mreq)
586 m_last(cd->mreq)->m_next = m;
587 else
588 cd->mreq = m;
589 while (m) {
590 cd->resid -= m->m_len;
591 m = m->m_next;
592 }
593 }
594
595 /*
596 * If cd->resid is zero now, we have managed to
597 * receive a record fragment from the stream. Check
598 * for the end-of-record mark to see if we need more.
599 */
600 if (cd->resid == 0) {
601 if (!cd->eor)
602 continue;
603
604 /*
605 * Success - we have a complete record in
606 * cd->mreq.
607 */
608 xdrmbuf_create(&xdrs, cd->mreq, XDR_DECODE);
609 cd->mreq = NULL;
610 sx_xunlock(&xprt->xp_lock);
611
612 if (! xdr_callmsg(&xdrs, msg)) {
613 XDR_DESTROY(&xdrs);
614 return (FALSE);
615 }
616
617 *addrp = NULL;
618 *mp = xdrmbuf_getall(&xdrs);
619 XDR_DESTROY(&xdrs);
620
621 return (TRUE);
622 }
623 }
624
625 readmore:
626 /*
627 * The socket upcall calls xprt_active() which will eventually
628 * cause the server to call us here. We attempt to
629 * read as much as possible from the socket and put
630 * the result in cd->mpending. If the read fails,
631 * we have drained both cd->mpending and the socket so
632 * we can call xprt_inactive().
633 */
634 uio.uio_resid = 1000000000;
635 uio.uio_td = curthread;
636 m = NULL;
637 rcvflag = MSG_DONTWAIT;
638 CURVNET_SET(xprt->xp_socket->so_vnet);
639 error = soreceive(xprt->xp_socket, NULL, &uio, &m, NULL,
640 &rcvflag);
641 CURVNET_RESTORE();
642
643 if (error == EWOULDBLOCK) {
644 /*
645 * We must re-test for readability after
646 * taking the lock to protect us in the case
647 * where a new packet arrives on the socket
648 * after our call to soreceive fails with
649 * EWOULDBLOCK. The pool lock protects us from
650 * racing the upcall after our soreadable()
651 * call returns false.
652 */
653 mtx_lock(&xprt->xp_pool->sp_lock);
654 if (!soreadable(xprt->xp_socket))
655 xprt_inactive_locked(xprt);
656 mtx_unlock(&xprt->xp_pool->sp_lock);
657 sx_xunlock(&xprt->xp_lock);
658 return (FALSE);
659 }
660
661 if (error) {
662 SOCKBUF_LOCK(&xprt->xp_socket->so_rcv);
663 if (xprt->xp_upcallset) {
664 xprt->xp_upcallset = 0;
665 soupcall_clear(xprt->xp_socket, SO_RCV);
666 }
667 SOCKBUF_UNLOCK(&xprt->xp_socket->so_rcv);
668 xprt_inactive(xprt);
669 cd->strm_stat = XPRT_DIED;
670 sx_xunlock(&xprt->xp_lock);
671 return (FALSE);
672 }
673
674 if (!m) {
675 /*
676 * EOF - the other end has closed the socket.
677 */
678 xprt_inactive(xprt);
679 cd->strm_stat = XPRT_DIED;
680 sx_xunlock(&xprt->xp_lock);
681 return (FALSE);
682 }
683
684 if (cd->mpending)
685 m_last(cd->mpending)->m_next = m;
686 else
687 cd->mpending = m;
688 }
689}
690
691static bool_t
692svc_vc_reply(SVCXPRT *xprt, struct rpc_msg *msg,
693 struct sockaddr *addr, struct mbuf *m)
694{
695 XDR xdrs;
696 struct mbuf *mrep;
697 bool_t stat = TRUE;
698 int error;
699
700 /*
701 * Leave space for record mark.
702 */
703 MGETHDR(mrep, M_WAIT, MT_DATA);
704 mrep->m_len = 0;
705 mrep->m_data += sizeof(uint32_t);
706
707 xdrmbuf_create(&xdrs, mrep, XDR_ENCODE);
708
709 if (msg->rm_reply.rp_stat == MSG_ACCEPTED &&
710 msg->rm_reply.rp_acpt.ar_stat == SUCCESS) {
711 if (!xdr_replymsg(&xdrs, msg))
712 stat = FALSE;
713 else
714 xdrmbuf_append(&xdrs, m);
715 } else {
716 stat = xdr_replymsg(&xdrs, msg);
717 }
718
719 if (stat) {
720 m_fixhdr(mrep);
721
722 /*
723 * Prepend a record marker containing the reply length.
724 */
725 M_PREPEND(mrep, sizeof(uint32_t), M_WAIT);
726 *mtod(mrep, uint32_t *) =
727 htonl(0x80000000 | (mrep->m_pkthdr.len
728 - sizeof(uint32_t)));
729 error = sosend(xprt->xp_socket, NULL, NULL, mrep, NULL,
730 0, curthread);
731 if (!error) {
732 stat = TRUE;
733 }
734 } else {
735 m_freem(mrep);
736 }
737
738 XDR_DESTROY(&xdrs);
739 xprt->xp_p2 = NULL;
740
741 return (stat);
742}
743
744static bool_t
745svc_vc_null()
746{
747
748 return (FALSE);
749}
750
751static int
752svc_vc_soupcall(struct socket *so, void *arg, int waitflag)
753{
754 SVCXPRT *xprt = (SVCXPRT *) arg;
755
756 xprt_active(xprt);
757 return (SU_OK);
758}
759
760#if 0
761/*
762 * Get the effective UID of the sending process. Used by rpcbind, keyserv
763 * and rpc.yppasswdd on AF_LOCAL.
764 */
765int
766__rpc_get_local_uid(SVCXPRT *transp, uid_t *uid) {
767 int sock, ret;
768 gid_t egid;
769 uid_t euid;
770 struct sockaddr *sa;
771
772 sock = transp->xp_fd;
773 sa = (struct sockaddr *)transp->xp_rtaddr;
774 if (sa->sa_family == AF_LOCAL) {
775 ret = getpeereid(sock, &euid, &egid);
776 if (ret == 0)
777 *uid = euid;
778 return (ret);
779 } else
780 return (-1);
781}
782#endif