xref: /freebsd/sys/rpc/clnt_vc.c (revision 9517e866259191fcd39434a97ad849a9b59b9b9f)
1 /*	$NetBSD: clnt_vc.c,v 1.4 2000/07/14 08:40:42 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)
33 static char *sccsid2 = "@(#)clnt_tcp.c 1.37 87/10/05 Copyr 1984 Sun Micro";
34 static char *sccsid = "@(#)clnt_tcp.c	2.2 88/08/01 4.0 RPCSRC";
35 static char sccsid3[] = "@(#)clnt_vc.c 1.19 89/03/16 Copyr 1988 Sun Micro";
36 #endif
37 #include <sys/cdefs.h>
38 __FBSDID("$FreeBSD$");
39 
40 /*
41  * clnt_tcp.c, Implements a TCP/IP based, client side RPC.
42  *
43  * Copyright (C) 1984, Sun Microsystems, Inc.
44  *
45  * TCP based RPC supports 'batched calls'.
46  * A sequence of calls may be batched-up in a send buffer.  The rpc call
47  * return immediately to the client even though the call was not necessarily
48  * sent.  The batching occurs if the results' xdr routine is NULL (0) AND
49  * the rpc timeout value is zero (see clnt.h, rpc).
50  *
51  * Clients should NOT casually batch calls that in fact return results; that is,
52  * the server side should be aware that a call is batched and not produce any
53  * return message.  Batched calls that produce many result messages can
54  * deadlock (netlock) the client and the server....
55  *
56  * Now go hang yourself.
57  */
58 
59 #include <sys/param.h>
60 #include <sys/systm.h>
61 #include <sys/lock.h>
62 #include <sys/malloc.h>
63 #include <sys/mbuf.h>
64 #include <sys/mutex.h>
65 #include <sys/pcpu.h>
66 #include <sys/proc.h>
67 #include <sys/protosw.h>
68 #include <sys/socket.h>
69 #include <sys/socketvar.h>
70 #include <sys/syslog.h>
71 #include <sys/time.h>
72 #include <sys/uio.h>
73 #include <netinet/tcp.h>
74 
75 #include <rpc/rpc.h>
76 #include <rpc/rpc_com.h>
77 
78 #define MCALL_MSG_SIZE 24
79 
80 struct cmessage {
81         struct cmsghdr cmsg;
82         struct cmsgcred cmcred;
83 };
84 
85 static enum clnt_stat clnt_vc_call(CLIENT *, struct rpc_callextra *,
86     rpcproc_t, struct mbuf *, struct mbuf **, struct timeval);
87 static void clnt_vc_geterr(CLIENT *, struct rpc_err *);
88 static bool_t clnt_vc_freeres(CLIENT *, xdrproc_t, void *);
89 static void clnt_vc_abort(CLIENT *);
90 static bool_t clnt_vc_control(CLIENT *, u_int, void *);
91 static void clnt_vc_close(CLIENT *);
92 static void clnt_vc_destroy(CLIENT *);
93 static bool_t time_not_ok(struct timeval *);
94 static int clnt_vc_soupcall(struct socket *so, void *arg, int waitflag);
95 
96 static struct clnt_ops clnt_vc_ops = {
97 	.cl_call =	clnt_vc_call,
98 	.cl_abort =	clnt_vc_abort,
99 	.cl_geterr =	clnt_vc_geterr,
100 	.cl_freeres =	clnt_vc_freeres,
101 	.cl_close =	clnt_vc_close,
102 	.cl_destroy =	clnt_vc_destroy,
103 	.cl_control =	clnt_vc_control
104 };
105 
106 /*
107  * A pending RPC request which awaits a reply. Requests which have
108  * received their reply will have cr_xid set to zero and cr_mrep to
109  * the mbuf chain of the reply.
110  */
111 struct ct_request {
112 	TAILQ_ENTRY(ct_request) cr_link;
113 	uint32_t		cr_xid;		/* XID of request */
114 	struct mbuf		*cr_mrep;	/* reply received by upcall */
115 	int			cr_error;	/* any error from upcall */
116 	char			cr_verf[MAX_AUTH_BYTES]; /* reply verf */
117 };
118 
119 TAILQ_HEAD(ct_request_list, ct_request);
120 
121 struct ct_data {
122 	struct mtx	ct_lock;
123 	int		ct_threads;	/* number of threads in clnt_vc_call */
124 	bool_t		ct_closing;	/* TRUE if we are closing */
125 	bool_t		ct_closed;	/* TRUE if we are closed */
126 	struct socket	*ct_socket;	/* connection socket */
127 	bool_t		ct_closeit;	/* close it on destroy */
128 	struct timeval	ct_wait;	/* wait interval in milliseconds */
129 	struct sockaddr_storage	ct_addr; /* remote addr */
130 	struct rpc_err	ct_error;
131 	uint32_t	ct_xid;
132 	char		ct_mcallc[MCALL_MSG_SIZE]; /* marshalled callmsg */
133 	size_t		ct_mpos;	/* pos after marshal */
134 	const char	*ct_waitchan;
135 	int		ct_waitflag;
136 	struct mbuf	*ct_record;	/* current reply record */
137 	size_t		ct_record_resid; /* how much left of reply to read */
138 	bool_t		ct_record_eor;	 /* true if reading last fragment */
139 	struct ct_request_list ct_pending;
140 	int		ct_upcallrefs;	/* Ref cnt of upcalls in prog. */
141 };
142 
143 static void clnt_vc_upcallsdone(struct ct_data *);
144 
145 static const char clnt_vc_errstr[] = "%s : %s";
146 static const char clnt_vc_str[] = "clnt_vc_create";
147 static const char clnt_read_vc_str[] = "read_vc";
148 static const char __no_mem_str[] = "out of memory";
149 
150 /*
151  * Create a client handle for a connection.
152  * Default options are set, which the user can change using clnt_control()'s.
153  * The rpc/vc package does buffering similar to stdio, so the client
154  * must pick send and receive buffer sizes, 0 => use the default.
155  * NB: fd is copied into a private area.
156  * NB: The rpch->cl_auth is set null authentication. Caller may wish to
157  * set this something more useful.
158  *
159  * fd should be an open socket
160  */
161 CLIENT *
162 clnt_vc_create(
163 	struct socket *so,		/* open file descriptor */
164 	struct sockaddr *raddr,		/* servers address */
165 	const rpcprog_t prog,		/* program number */
166 	const rpcvers_t vers,		/* version number */
167 	size_t sendsz,			/* buffer recv size */
168 	size_t recvsz)			/* buffer send size */
169 {
170 	CLIENT *cl;			/* client handle */
171 	struct ct_data *ct = NULL;	/* client handle */
172 	struct timeval now;
173 	struct rpc_msg call_msg;
174 	static uint32_t disrupt;
175 	struct __rpc_sockinfo si;
176 	XDR xdrs;
177 	int error, interrupted, one = 1;
178 	struct sockopt sopt;
179 
180 	if (disrupt == 0)
181 		disrupt = (uint32_t)(long)raddr;
182 
183 	cl = (CLIENT *)mem_alloc(sizeof (*cl));
184 	ct = (struct ct_data *)mem_alloc(sizeof (*ct));
185 
186 	mtx_init(&ct->ct_lock, "ct->ct_lock", NULL, MTX_DEF);
187 	ct->ct_threads = 0;
188 	ct->ct_closing = FALSE;
189 	ct->ct_closed = FALSE;
190 	ct->ct_upcallrefs = 0;
191 
192 	if ((so->so_state & (SS_ISCONNECTED|SS_ISCONFIRMING)) == 0) {
193 		error = soconnect(so, raddr, curthread);
194 		SOCK_LOCK(so);
195 		interrupted = 0;
196 		while ((so->so_state & SS_ISCONNECTING)
197 		    && so->so_error == 0) {
198 			error = msleep(&so->so_timeo, SOCK_MTX(so),
199 			    PSOCK | PCATCH | PBDRY, "connec", 0);
200 			if (error) {
201 				if (error == EINTR || error == ERESTART)
202 					interrupted = 1;
203 				break;
204 			}
205 		}
206 		if (error == 0) {
207 			error = so->so_error;
208 			so->so_error = 0;
209 		}
210 		SOCK_UNLOCK(so);
211 		if (error) {
212 			if (!interrupted)
213 				so->so_state &= ~SS_ISCONNECTING;
214 			rpc_createerr.cf_stat = RPC_SYSTEMERROR;
215 			rpc_createerr.cf_error.re_errno = error;
216 			goto err;
217 		}
218 	}
219 
220 	if (!__rpc_socket2sockinfo(so, &si))
221 		goto err;
222 
223 	if (so->so_proto->pr_flags & PR_CONNREQUIRED) {
224 		bzero(&sopt, sizeof(sopt));
225 		sopt.sopt_dir = SOPT_SET;
226 		sopt.sopt_level = SOL_SOCKET;
227 		sopt.sopt_name = SO_KEEPALIVE;
228 		sopt.sopt_val = &one;
229 		sopt.sopt_valsize = sizeof(one);
230 		sosetopt(so, &sopt);
231 	}
232 
233 	if (so->so_proto->pr_protocol == IPPROTO_TCP) {
234 		bzero(&sopt, sizeof(sopt));
235 		sopt.sopt_dir = SOPT_SET;
236 		sopt.sopt_level = IPPROTO_TCP;
237 		sopt.sopt_name = TCP_NODELAY;
238 		sopt.sopt_val = &one;
239 		sopt.sopt_valsize = sizeof(one);
240 		sosetopt(so, &sopt);
241 	}
242 
243 	ct->ct_closeit = FALSE;
244 
245 	/*
246 	 * Set up private data struct
247 	 */
248 	ct->ct_socket = so;
249 	ct->ct_wait.tv_sec = -1;
250 	ct->ct_wait.tv_usec = -1;
251 	memcpy(&ct->ct_addr, raddr, raddr->sa_len);
252 
253 	/*
254 	 * Initialize call message
255 	 */
256 	getmicrotime(&now);
257 	ct->ct_xid = ((uint32_t)++disrupt) ^ __RPC_GETXID(&now);
258 	call_msg.rm_xid = ct->ct_xid;
259 	call_msg.rm_direction = CALL;
260 	call_msg.rm_call.cb_rpcvers = RPC_MSG_VERSION;
261 	call_msg.rm_call.cb_prog = (uint32_t)prog;
262 	call_msg.rm_call.cb_vers = (uint32_t)vers;
263 
264 	/*
265 	 * pre-serialize the static part of the call msg and stash it away
266 	 */
267 	xdrmem_create(&xdrs, ct->ct_mcallc, MCALL_MSG_SIZE,
268 	    XDR_ENCODE);
269 	if (! xdr_callhdr(&xdrs, &call_msg)) {
270 		if (ct->ct_closeit) {
271 			soclose(ct->ct_socket);
272 		}
273 		goto err;
274 	}
275 	ct->ct_mpos = XDR_GETPOS(&xdrs);
276 	XDR_DESTROY(&xdrs);
277 	ct->ct_waitchan = "rpcrecv";
278 	ct->ct_waitflag = 0;
279 
280 	/*
281 	 * Create a client handle which uses xdrrec for serialization
282 	 * and authnone for authentication.
283 	 */
284 	cl->cl_refs = 1;
285 	cl->cl_ops = &clnt_vc_ops;
286 	cl->cl_private = ct;
287 	cl->cl_auth = authnone_create();
288 	sendsz = __rpc_get_t_size(si.si_af, si.si_proto, (int)sendsz);
289 	recvsz = __rpc_get_t_size(si.si_af, si.si_proto, (int)recvsz);
290 	soreserve(ct->ct_socket, sendsz, recvsz);
291 
292 	SOCKBUF_LOCK(&ct->ct_socket->so_rcv);
293 	soupcall_set(ct->ct_socket, SO_RCV, clnt_vc_soupcall, ct);
294 	SOCKBUF_UNLOCK(&ct->ct_socket->so_rcv);
295 
296 	ct->ct_record = NULL;
297 	ct->ct_record_resid = 0;
298 	TAILQ_INIT(&ct->ct_pending);
299 	return (cl);
300 
301 err:
302 	if (cl) {
303 		if (ct) {
304 			mtx_destroy(&ct->ct_lock);
305 			mem_free(ct, sizeof (struct ct_data));
306 		}
307 		if (cl)
308 			mem_free(cl, sizeof (CLIENT));
309 	}
310 	return ((CLIENT *)NULL);
311 }
312 
313 static enum clnt_stat
314 clnt_vc_call(
315 	CLIENT		*cl,		/* client handle */
316 	struct rpc_callextra *ext,	/* call metadata */
317 	rpcproc_t	proc,		/* procedure number */
318 	struct mbuf	*args,		/* pointer to args */
319 	struct mbuf	**resultsp,	/* pointer to results */
320 	struct timeval	utimeout)
321 {
322 	struct ct_data *ct = (struct ct_data *) cl->cl_private;
323 	AUTH *auth;
324 	struct rpc_err *errp;
325 	enum clnt_stat stat;
326 	XDR xdrs;
327 	struct rpc_msg reply_msg;
328 	bool_t ok;
329 	int nrefreshes = 2;		/* number of times to refresh cred */
330 	struct timeval timeout;
331 	uint32_t xid;
332 	struct mbuf *mreq = NULL, *results;
333 	struct ct_request *cr;
334 	int error;
335 
336 	cr = malloc(sizeof(struct ct_request), M_RPC, M_WAITOK);
337 
338 	mtx_lock(&ct->ct_lock);
339 
340 	if (ct->ct_closing || ct->ct_closed) {
341 		mtx_unlock(&ct->ct_lock);
342 		free(cr, M_RPC);
343 		return (RPC_CANTSEND);
344 	}
345 	ct->ct_threads++;
346 
347 	if (ext) {
348 		auth = ext->rc_auth;
349 		errp = &ext->rc_err;
350 	} else {
351 		auth = cl->cl_auth;
352 		errp = &ct->ct_error;
353 	}
354 
355 	cr->cr_mrep = NULL;
356 	cr->cr_error = 0;
357 
358 	if (ct->ct_wait.tv_usec == -1) {
359 		timeout = utimeout;	/* use supplied timeout */
360 	} else {
361 		timeout = ct->ct_wait;	/* use default timeout */
362 	}
363 
364 call_again:
365 	mtx_assert(&ct->ct_lock, MA_OWNED);
366 
367 	ct->ct_xid++;
368 	xid = ct->ct_xid;
369 
370 	mtx_unlock(&ct->ct_lock);
371 
372 	/*
373 	 * Leave space to pre-pend the record mark.
374 	 */
375 	MGETHDR(mreq, M_WAIT, MT_DATA);
376 	mreq->m_data += sizeof(uint32_t);
377 	KASSERT(ct->ct_mpos + sizeof(uint32_t) <= MHLEN,
378 	    ("RPC header too big"));
379 	bcopy(ct->ct_mcallc, mreq->m_data, ct->ct_mpos);
380 	mreq->m_len = ct->ct_mpos;
381 
382 	/*
383 	 * The XID is the first thing in the request.
384 	 */
385 	*mtod(mreq, uint32_t *) = htonl(xid);
386 
387 	xdrmbuf_create(&xdrs, mreq, XDR_ENCODE);
388 
389 	errp->re_status = stat = RPC_SUCCESS;
390 
391 	if ((! XDR_PUTINT32(&xdrs, &proc)) ||
392 	    (! AUTH_MARSHALL(auth, xid, &xdrs,
393 		m_copym(args, 0, M_COPYALL, M_WAITOK)))) {
394 		errp->re_status = stat = RPC_CANTENCODEARGS;
395 		mtx_lock(&ct->ct_lock);
396 		goto out;
397 	}
398 	mreq->m_pkthdr.len = m_length(mreq, NULL);
399 
400 	/*
401 	 * Prepend a record marker containing the packet length.
402 	 */
403 	M_PREPEND(mreq, sizeof(uint32_t), M_WAIT);
404 	*mtod(mreq, uint32_t *) =
405 		htonl(0x80000000 | (mreq->m_pkthdr.len - sizeof(uint32_t)));
406 
407 	cr->cr_xid = xid;
408 	mtx_lock(&ct->ct_lock);
409 	TAILQ_INSERT_TAIL(&ct->ct_pending, cr, cr_link);
410 	mtx_unlock(&ct->ct_lock);
411 
412 	/*
413 	 * sosend consumes mreq.
414 	 */
415 	error = sosend(ct->ct_socket, NULL, NULL, mreq, NULL, 0, curthread);
416 	mreq = NULL;
417 	if (error == EMSGSIZE) {
418 		SOCKBUF_LOCK(&ct->ct_socket->so_snd);
419 		sbwait(&ct->ct_socket->so_snd);
420 		SOCKBUF_UNLOCK(&ct->ct_socket->so_snd);
421 		AUTH_VALIDATE(auth, xid, NULL, NULL);
422 		mtx_lock(&ct->ct_lock);
423 		TAILQ_REMOVE(&ct->ct_pending, cr, cr_link);
424 		goto call_again;
425 	}
426 
427 	reply_msg.acpted_rply.ar_verf.oa_flavor = AUTH_NULL;
428 	reply_msg.acpted_rply.ar_verf.oa_base = cr->cr_verf;
429 	reply_msg.acpted_rply.ar_verf.oa_length = 0;
430 	reply_msg.acpted_rply.ar_results.where = NULL;
431 	reply_msg.acpted_rply.ar_results.proc = (xdrproc_t)xdr_void;
432 
433 	mtx_lock(&ct->ct_lock);
434 	if (error) {
435 		TAILQ_REMOVE(&ct->ct_pending, cr, cr_link);
436 		errp->re_errno = error;
437 		errp->re_status = stat = RPC_CANTSEND;
438 		goto out;
439 	}
440 
441 	/*
442 	 * Check to see if we got an upcall while waiting for the
443 	 * lock. In both these cases, the request has been removed
444 	 * from ct->ct_pending.
445 	 */
446 	if (cr->cr_error) {
447 		TAILQ_REMOVE(&ct->ct_pending, cr, cr_link);
448 		errp->re_errno = cr->cr_error;
449 		errp->re_status = stat = RPC_CANTRECV;
450 		goto out;
451 	}
452 	if (cr->cr_mrep) {
453 		TAILQ_REMOVE(&ct->ct_pending, cr, cr_link);
454 		goto got_reply;
455 	}
456 
457 	/*
458 	 * Hack to provide rpc-based message passing
459 	 */
460 	if (timeout.tv_sec == 0 && timeout.tv_usec == 0) {
461 		TAILQ_REMOVE(&ct->ct_pending, cr, cr_link);
462 		errp->re_status = stat = RPC_TIMEDOUT;
463 		goto out;
464 	}
465 
466 	error = msleep(cr, &ct->ct_lock, ct->ct_waitflag, ct->ct_waitchan,
467 	    tvtohz(&timeout));
468 
469 	TAILQ_REMOVE(&ct->ct_pending, cr, cr_link);
470 
471 	if (error) {
472 		/*
473 		 * The sleep returned an error so our request is still
474 		 * on the list. Turn the error code into an
475 		 * appropriate client status.
476 		 */
477 		errp->re_errno = error;
478 		switch (error) {
479 		case EINTR:
480 		case ERESTART:
481 			stat = RPC_INTR;
482 			break;
483 		case EWOULDBLOCK:
484 			stat = RPC_TIMEDOUT;
485 			break;
486 		default:
487 			stat = RPC_CANTRECV;
488 		}
489 		errp->re_status = stat;
490 		goto out;
491 	} else {
492 		/*
493 		 * We were woken up by the upcall.  If the
494 		 * upcall had a receive error, report that,
495 		 * otherwise we have a reply.
496 		 */
497 		if (cr->cr_error) {
498 			errp->re_errno = cr->cr_error;
499 			errp->re_status = stat = RPC_CANTRECV;
500 			goto out;
501 		}
502 	}
503 
504 got_reply:
505 	/*
506 	 * Now decode and validate the response. We need to drop the
507 	 * lock since xdr_replymsg may end up sleeping in malloc.
508 	 */
509 	mtx_unlock(&ct->ct_lock);
510 
511 	if (ext && ext->rc_feedback)
512 		ext->rc_feedback(FEEDBACK_OK, proc, ext->rc_feedback_arg);
513 
514 	xdrmbuf_create(&xdrs, cr->cr_mrep, XDR_DECODE);
515 	ok = xdr_replymsg(&xdrs, &reply_msg);
516 	cr->cr_mrep = NULL;
517 
518 	if (ok) {
519 		if ((reply_msg.rm_reply.rp_stat == MSG_ACCEPTED) &&
520 		    (reply_msg.acpted_rply.ar_stat == SUCCESS))
521 			errp->re_status = stat = RPC_SUCCESS;
522 		else
523 			stat = _seterr_reply(&reply_msg, errp);
524 
525 		if (stat == RPC_SUCCESS) {
526 			results = xdrmbuf_getall(&xdrs);
527 			if (!AUTH_VALIDATE(auth, xid,
528 				&reply_msg.acpted_rply.ar_verf,
529 				&results)) {
530 				errp->re_status = stat = RPC_AUTHERROR;
531 				errp->re_why = AUTH_INVALIDRESP;
532 			} else {
533 				KASSERT(results,
534 				    ("auth validated but no result"));
535 				*resultsp = results;
536 			}
537 		}		/* end successful completion */
538 		/*
539 		 * If unsuccesful AND error is an authentication error
540 		 * then refresh credentials and try again, else break
541 		 */
542 		else if (stat == RPC_AUTHERROR)
543 			/* maybe our credentials need to be refreshed ... */
544 			if (nrefreshes > 0 &&
545 			    AUTH_REFRESH(auth, &reply_msg)) {
546 				nrefreshes--;
547 				XDR_DESTROY(&xdrs);
548 				mtx_lock(&ct->ct_lock);
549 				goto call_again;
550 			}
551 		/* end of unsuccessful completion */
552 	}	/* end of valid reply message */
553 	else {
554 		errp->re_status = stat = RPC_CANTDECODERES;
555 	}
556 	XDR_DESTROY(&xdrs);
557 	mtx_lock(&ct->ct_lock);
558 out:
559 	mtx_assert(&ct->ct_lock, MA_OWNED);
560 
561 	KASSERT(stat != RPC_SUCCESS || *resultsp,
562 	    ("RPC_SUCCESS without reply"));
563 
564 	if (mreq)
565 		m_freem(mreq);
566 	if (cr->cr_mrep)
567 		m_freem(cr->cr_mrep);
568 
569 	ct->ct_threads--;
570 	if (ct->ct_closing)
571 		wakeup(ct);
572 
573 	mtx_unlock(&ct->ct_lock);
574 
575 	if (auth && stat != RPC_SUCCESS)
576 		AUTH_VALIDATE(auth, xid, NULL, NULL);
577 
578 	free(cr, M_RPC);
579 
580 	return (stat);
581 }
582 
583 static void
584 clnt_vc_geterr(CLIENT *cl, struct rpc_err *errp)
585 {
586 	struct ct_data *ct = (struct ct_data *) cl->cl_private;
587 
588 	*errp = ct->ct_error;
589 }
590 
591 static bool_t
592 clnt_vc_freeres(CLIENT *cl, xdrproc_t xdr_res, void *res_ptr)
593 {
594 	XDR xdrs;
595 	bool_t dummy;
596 
597 	xdrs.x_op = XDR_FREE;
598 	dummy = (*xdr_res)(&xdrs, res_ptr);
599 
600 	return (dummy);
601 }
602 
603 /*ARGSUSED*/
604 static void
605 clnt_vc_abort(CLIENT *cl)
606 {
607 }
608 
609 static bool_t
610 clnt_vc_control(CLIENT *cl, u_int request, void *info)
611 {
612 	struct ct_data *ct = (struct ct_data *)cl->cl_private;
613 	void *infop = info;
614 
615 	mtx_lock(&ct->ct_lock);
616 
617 	switch (request) {
618 	case CLSET_FD_CLOSE:
619 		ct->ct_closeit = TRUE;
620 		mtx_unlock(&ct->ct_lock);
621 		return (TRUE);
622 	case CLSET_FD_NCLOSE:
623 		ct->ct_closeit = FALSE;
624 		mtx_unlock(&ct->ct_lock);
625 		return (TRUE);
626 	default:
627 		break;
628 	}
629 
630 	/* for other requests which use info */
631 	if (info == NULL) {
632 		mtx_unlock(&ct->ct_lock);
633 		return (FALSE);
634 	}
635 	switch (request) {
636 	case CLSET_TIMEOUT:
637 		if (time_not_ok((struct timeval *)info)) {
638 			mtx_unlock(&ct->ct_lock);
639 			return (FALSE);
640 		}
641 		ct->ct_wait = *(struct timeval *)infop;
642 		break;
643 	case CLGET_TIMEOUT:
644 		*(struct timeval *)infop = ct->ct_wait;
645 		break;
646 	case CLGET_SERVER_ADDR:
647 		(void) memcpy(info, &ct->ct_addr, (size_t)ct->ct_addr.ss_len);
648 		break;
649 	case CLGET_SVC_ADDR:
650 		/*
651 		 * Slightly different semantics to userland - we use
652 		 * sockaddr instead of netbuf.
653 		 */
654 		memcpy(info, &ct->ct_addr, ct->ct_addr.ss_len);
655 		break;
656 	case CLSET_SVC_ADDR:		/* set to new address */
657 		mtx_unlock(&ct->ct_lock);
658 		return (FALSE);
659 	case CLGET_XID:
660 		*(uint32_t *)info = ct->ct_xid;
661 		break;
662 	case CLSET_XID:
663 		/* This will set the xid of the NEXT call */
664 		/* decrement by 1 as clnt_vc_call() increments once */
665 		ct->ct_xid = *(uint32_t *)info - 1;
666 		break;
667 	case CLGET_VERS:
668 		/*
669 		 * This RELIES on the information that, in the call body,
670 		 * the version number field is the fifth field from the
671 		 * begining of the RPC header. MUST be changed if the
672 		 * call_struct is changed
673 		 */
674 		*(uint32_t *)info =
675 		    ntohl(*(uint32_t *)(void *)(ct->ct_mcallc +
676 		    4 * BYTES_PER_XDR_UNIT));
677 		break;
678 
679 	case CLSET_VERS:
680 		*(uint32_t *)(void *)(ct->ct_mcallc +
681 		    4 * BYTES_PER_XDR_UNIT) =
682 		    htonl(*(uint32_t *)info);
683 		break;
684 
685 	case CLGET_PROG:
686 		/*
687 		 * This RELIES on the information that, in the call body,
688 		 * the program number field is the fourth field from the
689 		 * begining of the RPC header. MUST be changed if the
690 		 * call_struct is changed
691 		 */
692 		*(uint32_t *)info =
693 		    ntohl(*(uint32_t *)(void *)(ct->ct_mcallc +
694 		    3 * BYTES_PER_XDR_UNIT));
695 		break;
696 
697 	case CLSET_PROG:
698 		*(uint32_t *)(void *)(ct->ct_mcallc +
699 		    3 * BYTES_PER_XDR_UNIT) =
700 		    htonl(*(uint32_t *)info);
701 		break;
702 
703 	case CLSET_WAITCHAN:
704 		ct->ct_waitchan = (const char *)info;
705 		break;
706 
707 	case CLGET_WAITCHAN:
708 		*(const char **) info = ct->ct_waitchan;
709 		break;
710 
711 	case CLSET_INTERRUPTIBLE:
712 		if (*(int *) info)
713 			ct->ct_waitflag = PCATCH | PBDRY;
714 		else
715 			ct->ct_waitflag = 0;
716 		break;
717 
718 	case CLGET_INTERRUPTIBLE:
719 		if (ct->ct_waitflag)
720 			*(int *) info = TRUE;
721 		else
722 			*(int *) info = FALSE;
723 		break;
724 
725 	default:
726 		mtx_unlock(&ct->ct_lock);
727 		return (FALSE);
728 	}
729 
730 	mtx_unlock(&ct->ct_lock);
731 	return (TRUE);
732 }
733 
734 static void
735 clnt_vc_close(CLIENT *cl)
736 {
737 	struct ct_data *ct = (struct ct_data *) cl->cl_private;
738 	struct ct_request *cr;
739 
740 	mtx_lock(&ct->ct_lock);
741 
742 	if (ct->ct_closed) {
743 		mtx_unlock(&ct->ct_lock);
744 		return;
745 	}
746 
747 	if (ct->ct_closing) {
748 		while (ct->ct_closing)
749 			msleep(ct, &ct->ct_lock, 0, "rpcclose", 0);
750 		KASSERT(ct->ct_closed, ("client should be closed"));
751 		mtx_unlock(&ct->ct_lock);
752 		return;
753 	}
754 
755 	if (ct->ct_socket) {
756 		ct->ct_closing = TRUE;
757 		mtx_unlock(&ct->ct_lock);
758 
759 		SOCKBUF_LOCK(&ct->ct_socket->so_rcv);
760 		soupcall_clear(ct->ct_socket, SO_RCV);
761 		clnt_vc_upcallsdone(ct);
762 		SOCKBUF_UNLOCK(&ct->ct_socket->so_rcv);
763 
764 		/*
765 		 * Abort any pending requests and wait until everyone
766 		 * has finished with clnt_vc_call.
767 		 */
768 		mtx_lock(&ct->ct_lock);
769 		TAILQ_FOREACH(cr, &ct->ct_pending, cr_link) {
770 			cr->cr_xid = 0;
771 			cr->cr_error = ESHUTDOWN;
772 			wakeup(cr);
773 		}
774 
775 		while (ct->ct_threads)
776 			msleep(ct, &ct->ct_lock, 0, "rpcclose", 0);
777 	}
778 
779 	ct->ct_closing = FALSE;
780 	ct->ct_closed = TRUE;
781 	mtx_unlock(&ct->ct_lock);
782 	wakeup(ct);
783 }
784 
785 static void
786 clnt_vc_destroy(CLIENT *cl)
787 {
788 	struct ct_data *ct = (struct ct_data *) cl->cl_private;
789 	struct socket *so = NULL;
790 
791 	clnt_vc_close(cl);
792 
793 	mtx_lock(&ct->ct_lock);
794 
795 	if (ct->ct_socket) {
796 		if (ct->ct_closeit) {
797 			so = ct->ct_socket;
798 		}
799 	}
800 
801 	mtx_unlock(&ct->ct_lock);
802 
803 	mtx_destroy(&ct->ct_lock);
804 	if (so) {
805 		soshutdown(so, SHUT_WR);
806 		soclose(so);
807 	}
808 	mem_free(ct, sizeof(struct ct_data));
809 	mem_free(cl, sizeof(CLIENT));
810 }
811 
812 /*
813  * Make sure that the time is not garbage.   -1 value is disallowed.
814  * Note this is different from time_not_ok in clnt_dg.c
815  */
816 static bool_t
817 time_not_ok(struct timeval *t)
818 {
819 	return (t->tv_sec <= -1 || t->tv_sec > 100000000 ||
820 		t->tv_usec <= -1 || t->tv_usec > 1000000);
821 }
822 
823 int
824 clnt_vc_soupcall(struct socket *so, void *arg, int waitflag)
825 {
826 	struct ct_data *ct = (struct ct_data *) arg;
827 	struct uio uio;
828 	struct mbuf *m;
829 	struct ct_request *cr;
830 	int error, rcvflag, foundreq;
831 	uint32_t xid, header;
832 	bool_t do_read;
833 
834 	ct->ct_upcallrefs++;
835 	uio.uio_td = curthread;
836 	do {
837 		/*
838 		 * If ct_record_resid is zero, we are waiting for a
839 		 * record mark.
840 		 */
841 		if (ct->ct_record_resid == 0) {
842 
843 			/*
844 			 * Make sure there is either a whole record
845 			 * mark in the buffer or there is some other
846 			 * error condition
847 			 */
848 			do_read = FALSE;
849 			if (so->so_rcv.sb_cc >= sizeof(uint32_t)
850 			    || (so->so_rcv.sb_state & SBS_CANTRCVMORE)
851 			    || so->so_error)
852 				do_read = TRUE;
853 
854 			if (!do_read)
855 				break;
856 
857 			SOCKBUF_UNLOCK(&so->so_rcv);
858 			uio.uio_resid = sizeof(uint32_t);
859 			m = NULL;
860 			rcvflag = MSG_DONTWAIT | MSG_SOCALLBCK;
861 			error = soreceive(so, NULL, &uio, &m, NULL, &rcvflag);
862 			SOCKBUF_LOCK(&so->so_rcv);
863 
864 			if (error == EWOULDBLOCK)
865 				break;
866 
867 			/*
868 			 * If there was an error, wake up all pending
869 			 * requests.
870 			 */
871 			if (error || uio.uio_resid > 0) {
872 			wakeup_all:
873 				mtx_lock(&ct->ct_lock);
874 				if (!error) {
875 					/*
876 					 * We must have got EOF trying
877 					 * to read from the stream.
878 					 */
879 					error = ECONNRESET;
880 				}
881 				ct->ct_error.re_status = RPC_CANTRECV;
882 				ct->ct_error.re_errno = error;
883 				TAILQ_FOREACH(cr, &ct->ct_pending, cr_link) {
884 					cr->cr_error = error;
885 					wakeup(cr);
886 				}
887 				mtx_unlock(&ct->ct_lock);
888 				break;
889 			}
890 			bcopy(mtod(m, uint32_t *), &header, sizeof(uint32_t));
891 			header = ntohl(header);
892 			ct->ct_record = NULL;
893 			ct->ct_record_resid = header & 0x7fffffff;
894 			ct->ct_record_eor = ((header & 0x80000000) != 0);
895 			m_freem(m);
896 		} else {
897 			/*
898 			 * Wait until the socket has the whole record
899 			 * buffered.
900 			 */
901 			do_read = FALSE;
902 			if (so->so_rcv.sb_cc >= ct->ct_record_resid
903 			    || (so->so_rcv.sb_state & SBS_CANTRCVMORE)
904 			    || so->so_error)
905 				do_read = TRUE;
906 
907 			if (!do_read)
908 				break;
909 
910 			/*
911 			 * We have the record mark. Read as much as
912 			 * the socket has buffered up to the end of
913 			 * this record.
914 			 */
915 			SOCKBUF_UNLOCK(&so->so_rcv);
916 			uio.uio_resid = ct->ct_record_resid;
917 			m = NULL;
918 			rcvflag = MSG_DONTWAIT | MSG_SOCALLBCK;
919 			error = soreceive(so, NULL, &uio, &m, NULL, &rcvflag);
920 			SOCKBUF_LOCK(&so->so_rcv);
921 
922 			if (error == EWOULDBLOCK)
923 				break;
924 
925 			if (error || uio.uio_resid == ct->ct_record_resid)
926 				goto wakeup_all;
927 
928 			/*
929 			 * If we have part of the record already,
930 			 * chain this bit onto the end.
931 			 */
932 			if (ct->ct_record)
933 				m_last(ct->ct_record)->m_next = m;
934 			else
935 				ct->ct_record = m;
936 
937 			ct->ct_record_resid = uio.uio_resid;
938 
939 			/*
940 			 * If we have the entire record, see if we can
941 			 * match it to a request.
942 			 */
943 			if (ct->ct_record_resid == 0
944 			    && ct->ct_record_eor) {
945 				/*
946 				 * The XID is in the first uint32_t of
947 				 * the reply.
948 				 */
949 				if (ct->ct_record->m_len < sizeof(xid))
950 					ct->ct_record =
951 						m_pullup(ct->ct_record,
952 						    sizeof(xid));
953 				if (!ct->ct_record)
954 					break;
955 				bcopy(mtod(ct->ct_record, uint32_t *),
956 				    &xid, sizeof(uint32_t));
957 				xid = ntohl(xid);
958 
959 				mtx_lock(&ct->ct_lock);
960 				foundreq = 0;
961 				TAILQ_FOREACH(cr, &ct->ct_pending, cr_link) {
962 					if (cr->cr_xid == xid) {
963 						/*
964 						 * This one
965 						 * matches. We leave
966 						 * the reply mbuf in
967 						 * cr->cr_mrep. Set
968 						 * the XID to zero so
969 						 * that we will ignore
970 						 * any duplicaed
971 						 * replies.
972 						 */
973 						cr->cr_xid = 0;
974 						cr->cr_mrep = ct->ct_record;
975 						cr->cr_error = 0;
976 						foundreq = 1;
977 						wakeup(cr);
978 						break;
979 					}
980 				}
981 				mtx_unlock(&ct->ct_lock);
982 
983 				if (!foundreq)
984 					m_freem(ct->ct_record);
985 				ct->ct_record = NULL;
986 			}
987 		}
988 	} while (m);
989 	ct->ct_upcallrefs--;
990 	if (ct->ct_upcallrefs < 0)
991 		panic("rpcvc upcall refcnt");
992 	if (ct->ct_upcallrefs == 0)
993 		wakeup(&ct->ct_upcallrefs);
994 	return (SU_OK);
995 }
996 
997 /*
998  * Wait for all upcalls in progress to complete.
999  */
1000 static void
1001 clnt_vc_upcallsdone(struct ct_data *ct)
1002 {
1003 
1004 	SOCKBUF_LOCK_ASSERT(&ct->ct_socket->so_rcv);
1005 
1006 	while (ct->ct_upcallrefs > 0)
1007 		(void) msleep(&ct->ct_upcallrefs,
1008 		    SOCKBUF_MTX(&ct->ct_socket->so_rcv), 0, "rpcvcup", 0);
1009 }
1010