xref: /freebsd/sys/rpc/clnt_dg.c (revision 1670a1c2a47d10ecccd001970b859caf93cd3b6e)
1 /*	$NetBSD: clnt_dg.c,v 1.4 2000/07/14 08:40:41 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  * Copyright (c) 1986-1991 by Sun Microsystems Inc.
33  */
34 
35 #if defined(LIBC_SCCS) && !defined(lint)
36 #ident	"@(#)clnt_dg.c	1.23	94/04/22 SMI"
37 static char sccsid[] = "@(#)clnt_dg.c 1.19 89/03/16 Copyr 1988 Sun Micro";
38 #endif
39 #include <sys/cdefs.h>
40 __FBSDID("$FreeBSD$");
41 
42 /*
43  * Implements a connectionless client side RPC.
44  */
45 
46 #include <sys/param.h>
47 #include <sys/systm.h>
48 #include <sys/kernel.h>
49 #include <sys/lock.h>
50 #include <sys/malloc.h>
51 #include <sys/mbuf.h>
52 #include <sys/mutex.h>
53 #include <sys/pcpu.h>
54 #include <sys/proc.h>
55 #include <sys/socket.h>
56 #include <sys/socketvar.h>
57 #include <sys/time.h>
58 #include <sys/uio.h>
59 
60 #include <net/vnet.h>
61 
62 #include <rpc/rpc.h>
63 #include <rpc/rpc_com.h>
64 
65 
66 #ifdef _FREEFALL_CONFIG
67 /*
68  * Disable RPC exponential back-off for FreeBSD.org systems.
69  */
70 #define	RPC_MAX_BACKOFF		1 /* second */
71 #else
72 #define	RPC_MAX_BACKOFF		30 /* seconds */
73 #endif
74 
75 static bool_t time_not_ok(struct timeval *);
76 static enum clnt_stat clnt_dg_call(CLIENT *, struct rpc_callextra *,
77     rpcproc_t, struct mbuf *, struct mbuf **, struct timeval);
78 static void clnt_dg_geterr(CLIENT *, struct rpc_err *);
79 static bool_t clnt_dg_freeres(CLIENT *, xdrproc_t, void *);
80 static void clnt_dg_abort(CLIENT *);
81 static bool_t clnt_dg_control(CLIENT *, u_int, void *);
82 static void clnt_dg_close(CLIENT *);
83 static void clnt_dg_destroy(CLIENT *);
84 static int clnt_dg_soupcall(struct socket *so, void *arg, int waitflag);
85 
86 static struct clnt_ops clnt_dg_ops = {
87 	.cl_call =	clnt_dg_call,
88 	.cl_abort =	clnt_dg_abort,
89 	.cl_geterr =	clnt_dg_geterr,
90 	.cl_freeres =	clnt_dg_freeres,
91 	.cl_close =	clnt_dg_close,
92 	.cl_destroy =	clnt_dg_destroy,
93 	.cl_control =	clnt_dg_control
94 };
95 
96 static const char mem_err_clnt_dg[] = "clnt_dg_create: out of memory";
97 
98 /*
99  * A pending RPC request which awaits a reply. Requests which have
100  * received their reply will have cr_xid set to zero and cr_mrep to
101  * the mbuf chain of the reply.
102  */
103 struct cu_request {
104 	TAILQ_ENTRY(cu_request) cr_link;
105 	CLIENT			*cr_client;	/* owner */
106 	uint32_t		cr_xid;		/* XID of request */
107 	struct mbuf		*cr_mrep;	/* reply received by upcall */
108 	int			cr_error;	/* any error from upcall */
109 	char			cr_verf[MAX_AUTH_BYTES]; /* reply verf */
110 };
111 
112 TAILQ_HEAD(cu_request_list, cu_request);
113 
114 #define MCALL_MSG_SIZE 24
115 
116 /*
117  * This structure is pointed to by the socket buffer's sb_upcallarg
118  * member. It is separate from the client private data to facilitate
119  * multiple clients sharing the same socket. The cs_lock mutex is used
120  * to protect all fields of this structure, the socket's receive
121  * buffer SOCKBUF_LOCK is used to ensure that exactly one of these
122  * structures is installed on the socket.
123  */
124 struct cu_socket {
125 	struct mtx		cs_lock;
126 	int			cs_refs;	/* Count of clients */
127 	struct cu_request_list	cs_pending;	/* Requests awaiting replies */
128 	int			cs_upcallrefs;	/* Refcnt of upcalls in prog.*/
129 };
130 
131 static void clnt_dg_upcallsdone(struct socket *, struct cu_socket *);
132 
133 /*
134  * Private data kept per client handle
135  */
136 struct cu_data {
137 	int			cu_threads;	/* # threads in clnt_vc_call */
138 	bool_t			cu_closing;	/* TRUE if we are closing */
139 	bool_t			cu_closed;	/* TRUE if we are closed */
140 	struct socket		*cu_socket;	/* connection socket */
141 	bool_t			cu_closeit;	/* opened by library */
142 	struct sockaddr_storage	cu_raddr;	/* remote address */
143 	int			cu_rlen;
144 	struct timeval		cu_wait;	/* retransmit interval */
145 	struct timeval		cu_total;	/* total time for the call */
146 	struct rpc_err		cu_error;
147 	uint32_t		cu_xid;
148 	char			cu_mcallc[MCALL_MSG_SIZE]; /* marshalled callmsg */
149 	size_t			cu_mcalllen;
150 	size_t			cu_sendsz;	/* send size */
151 	size_t			cu_recvsz;	/* recv size */
152 	int			cu_async;
153 	int			cu_connect;	/* Use connect(). */
154 	int			cu_connected;	/* Have done connect(). */
155 	const char		*cu_waitchan;
156 	int			cu_waitflag;
157 	int			cu_cwnd;	/* congestion window */
158 	int			cu_sent;	/* number of in-flight RPCs */
159 	bool_t			cu_cwnd_wait;
160 };
161 
162 #define CWNDSCALE	256
163 #define MAXCWND		(32 * CWNDSCALE)
164 
165 /*
166  * Connection less client creation returns with client handle parameters.
167  * Default options are set, which the user can change using clnt_control().
168  * fd should be open and bound.
169  * NB: The rpch->cl_auth is initialized to null authentication.
170  * 	Caller may wish to set this something more useful.
171  *
172  * sendsz and recvsz are the maximum allowable packet sizes that can be
173  * sent and received. Normally they are the same, but they can be
174  * changed to improve the program efficiency and buffer allocation.
175  * If they are 0, use the transport default.
176  *
177  * If svcaddr is NULL, returns NULL.
178  */
179 CLIENT *
180 clnt_dg_create(
181 	struct socket *so,
182 	struct sockaddr *svcaddr,	/* servers address */
183 	rpcprog_t program,		/* program number */
184 	rpcvers_t version,		/* version number */
185 	size_t sendsz,			/* buffer recv size */
186 	size_t recvsz)			/* buffer send size */
187 {
188 	CLIENT *cl = NULL;		/* client handle */
189 	struct cu_data *cu = NULL;	/* private data */
190 	struct cu_socket *cs = NULL;
191 	struct sockbuf *sb;
192 	struct timeval now;
193 	struct rpc_msg call_msg;
194 	struct __rpc_sockinfo si;
195 	XDR xdrs;
196 
197 	if (svcaddr == NULL) {
198 		rpc_createerr.cf_stat = RPC_UNKNOWNADDR;
199 		return (NULL);
200 	}
201 
202 	CURVNET_SET(so->so_vnet);
203 	if (!__rpc_socket2sockinfo(so, &si)) {
204 		rpc_createerr.cf_stat = RPC_TLIERROR;
205 		rpc_createerr.cf_error.re_errno = 0;
206 		CURVNET_RESTORE();
207 		return (NULL);
208 	}
209 	CURVNET_RESTORE();
210 
211 	/*
212 	 * Find the receive and the send size
213 	 */
214 	sendsz = __rpc_get_t_size(si.si_af, si.si_proto, (int)sendsz);
215 	recvsz = __rpc_get_t_size(si.si_af, si.si_proto, (int)recvsz);
216 	if ((sendsz == 0) || (recvsz == 0)) {
217 		rpc_createerr.cf_stat = RPC_TLIERROR; /* XXX */
218 		rpc_createerr.cf_error.re_errno = 0;
219 		return (NULL);
220 	}
221 
222 	cl = mem_alloc(sizeof (CLIENT));
223 
224 	/*
225 	 * Should be multiple of 4 for XDR.
226 	 */
227 	sendsz = ((sendsz + 3) / 4) * 4;
228 	recvsz = ((recvsz + 3) / 4) * 4;
229 	cu = mem_alloc(sizeof (*cu));
230 	cu->cu_threads = 0;
231 	cu->cu_closing = FALSE;
232 	cu->cu_closed = FALSE;
233 	(void) memcpy(&cu->cu_raddr, svcaddr, (size_t)svcaddr->sa_len);
234 	cu->cu_rlen = svcaddr->sa_len;
235 	/* Other values can also be set through clnt_control() */
236 	cu->cu_wait.tv_sec = 3;	/* heuristically chosen */
237 	cu->cu_wait.tv_usec = 0;
238 	cu->cu_total.tv_sec = -1;
239 	cu->cu_total.tv_usec = -1;
240 	cu->cu_sendsz = sendsz;
241 	cu->cu_recvsz = recvsz;
242 	cu->cu_async = FALSE;
243 	cu->cu_connect = FALSE;
244 	cu->cu_connected = FALSE;
245 	cu->cu_waitchan = "rpcrecv";
246 	cu->cu_waitflag = 0;
247 	cu->cu_cwnd = MAXCWND / 2;
248 	cu->cu_sent = 0;
249 	cu->cu_cwnd_wait = FALSE;
250 	(void) getmicrotime(&now);
251 	cu->cu_xid = __RPC_GETXID(&now);
252 	call_msg.rm_xid = cu->cu_xid;
253 	call_msg.rm_call.cb_prog = program;
254 	call_msg.rm_call.cb_vers = version;
255 	xdrmem_create(&xdrs, cu->cu_mcallc, MCALL_MSG_SIZE, XDR_ENCODE);
256 	if (! xdr_callhdr(&xdrs, &call_msg)) {
257 		rpc_createerr.cf_stat = RPC_CANTENCODEARGS;  /* XXX */
258 		rpc_createerr.cf_error.re_errno = 0;
259 		goto err2;
260 	}
261 	cu->cu_mcalllen = XDR_GETPOS(&xdrs);
262 
263 	/*
264 	 * By default, closeit is always FALSE. It is users responsibility
265 	 * to do a close on it, else the user may use clnt_control
266 	 * to let clnt_destroy do it for him/her.
267 	 */
268 	cu->cu_closeit = FALSE;
269 	cu->cu_socket = so;
270 	soreserve(so, 256*1024, 256*1024);
271 
272 	sb = &so->so_rcv;
273 	SOCKBUF_LOCK(&so->so_rcv);
274 recheck_socket:
275 	if (sb->sb_upcall) {
276 		if (sb->sb_upcall != clnt_dg_soupcall) {
277 			SOCKBUF_UNLOCK(&so->so_rcv);
278 			printf("clnt_dg_create(): socket already has an incompatible upcall\n");
279 			goto err2;
280 		}
281 		cs = (struct cu_socket *) sb->sb_upcallarg;
282 		mtx_lock(&cs->cs_lock);
283 		cs->cs_refs++;
284 		mtx_unlock(&cs->cs_lock);
285 	} else {
286 		/*
287 		 * We are the first on this socket - allocate the
288 		 * structure and install it in the socket.
289 		 */
290 		SOCKBUF_UNLOCK(&so->so_rcv);
291 		cs = mem_alloc(sizeof(*cs));
292 		SOCKBUF_LOCK(&so->so_rcv);
293 		if (sb->sb_upcall) {
294 			/*
295 			 * We have lost a race with some other client.
296 			 */
297 			mem_free(cs, sizeof(*cs));
298 			goto recheck_socket;
299 		}
300 		mtx_init(&cs->cs_lock, "cs->cs_lock", NULL, MTX_DEF);
301 		cs->cs_refs = 1;
302 		cs->cs_upcallrefs = 0;
303 		TAILQ_INIT(&cs->cs_pending);
304 		soupcall_set(so, SO_RCV, clnt_dg_soupcall, cs);
305 	}
306 	SOCKBUF_UNLOCK(&so->so_rcv);
307 
308 	cl->cl_refs = 1;
309 	cl->cl_ops = &clnt_dg_ops;
310 	cl->cl_private = (caddr_t)(void *)cu;
311 	cl->cl_auth = authnone_create();
312 	cl->cl_tp = NULL;
313 	cl->cl_netid = NULL;
314 	return (cl);
315 err2:
316 	if (cl) {
317 		mem_free(cl, sizeof (CLIENT));
318 		if (cu)
319 			mem_free(cu, sizeof (*cu));
320 	}
321 	return (NULL);
322 }
323 
324 static enum clnt_stat
325 clnt_dg_call(
326 	CLIENT		*cl,		/* client handle */
327 	struct rpc_callextra *ext,	/* call metadata */
328 	rpcproc_t	proc,		/* procedure number */
329 	struct mbuf	*args,		/* pointer to args */
330 	struct mbuf	**resultsp,	/* pointer to results */
331 	struct timeval	utimeout)	/* seconds to wait before giving up */
332 {
333 	struct cu_data *cu = (struct cu_data *)cl->cl_private;
334 	struct cu_socket *cs;
335 	struct rpc_timers *rt;
336 	AUTH *auth;
337 	struct rpc_err *errp;
338 	enum clnt_stat stat;
339 	XDR xdrs;
340 	struct rpc_msg reply_msg;
341 	bool_t ok;
342 	int retrans;			/* number of re-transmits so far */
343 	int nrefreshes = 2;		/* number of times to refresh cred */
344 	struct timeval *tvp;
345 	int timeout;
346 	int retransmit_time;
347 	int next_sendtime, starttime, rtt, time_waited, tv = 0;
348 	struct sockaddr *sa;
349 	socklen_t salen;
350 	uint32_t xid = 0;
351 	struct mbuf *mreq = NULL, *results;
352 	struct cu_request *cr;
353 	int error;
354 
355 	cs = cu->cu_socket->so_rcv.sb_upcallarg;
356 	cr = malloc(sizeof(struct cu_request), M_RPC, M_WAITOK);
357 
358 	mtx_lock(&cs->cs_lock);
359 
360 	if (cu->cu_closing || cu->cu_closed) {
361 		mtx_unlock(&cs->cs_lock);
362 		free(cr, M_RPC);
363 		return (RPC_CANTSEND);
364 	}
365 	cu->cu_threads++;
366 
367 	if (ext) {
368 		auth = ext->rc_auth;
369 		errp = &ext->rc_err;
370 	} else {
371 		auth = cl->cl_auth;
372 		errp = &cu->cu_error;
373 	}
374 
375 	cr->cr_client = cl;
376 	cr->cr_mrep = NULL;
377 	cr->cr_error = 0;
378 
379 	if (cu->cu_total.tv_usec == -1) {
380 		tvp = &utimeout; /* use supplied timeout */
381 	} else {
382 		tvp = &cu->cu_total; /* use default timeout */
383 	}
384 	if (tvp->tv_sec || tvp->tv_usec)
385 		timeout = tvtohz(tvp);
386 	else
387 		timeout = 0;
388 
389 	if (cu->cu_connect && !cu->cu_connected) {
390 		mtx_unlock(&cs->cs_lock);
391 		error = soconnect(cu->cu_socket,
392 		    (struct sockaddr *)&cu->cu_raddr, curthread);
393 		mtx_lock(&cs->cs_lock);
394 		if (error) {
395 			errp->re_errno = error;
396 			errp->re_status = stat = RPC_CANTSEND;
397 			goto out;
398 		}
399 		cu->cu_connected = 1;
400 	}
401 	if (cu->cu_connected) {
402 		sa = NULL;
403 		salen = 0;
404 	} else {
405 		sa = (struct sockaddr *)&cu->cu_raddr;
406 		salen = cu->cu_rlen;
407 	}
408 	time_waited = 0;
409 	retrans = 0;
410 	if (ext && ext->rc_timers) {
411 		rt = ext->rc_timers;
412 		if (!rt->rt_rtxcur)
413 			rt->rt_rtxcur = tvtohz(&cu->cu_wait);
414 		retransmit_time = next_sendtime = rt->rt_rtxcur;
415 	} else {
416 		rt = NULL;
417 		retransmit_time = next_sendtime = tvtohz(&cu->cu_wait);
418 	}
419 
420 	starttime = ticks;
421 
422 call_again:
423 	mtx_assert(&cs->cs_lock, MA_OWNED);
424 
425 	cu->cu_xid++;
426 	xid = cu->cu_xid;
427 
428 send_again:
429 	mtx_unlock(&cs->cs_lock);
430 
431 	MGETHDR(mreq, M_WAIT, MT_DATA);
432 	KASSERT(cu->cu_mcalllen <= MHLEN, ("RPC header too big"));
433 	bcopy(cu->cu_mcallc, mreq->m_data, cu->cu_mcalllen);
434 	mreq->m_len = cu->cu_mcalllen;
435 
436 	/*
437 	 * The XID is the first thing in the request.
438 	 */
439 	*mtod(mreq, uint32_t *) = htonl(xid);
440 
441 	xdrmbuf_create(&xdrs, mreq, XDR_ENCODE);
442 
443 	if (cu->cu_async == TRUE && args == NULL)
444 		goto get_reply;
445 
446 	if ((! XDR_PUTINT32(&xdrs, &proc)) ||
447 	    (! AUTH_MARSHALL(auth, xid, &xdrs,
448 		m_copym(args, 0, M_COPYALL, M_WAITOK)))) {
449 		errp->re_status = stat = RPC_CANTENCODEARGS;
450 		mtx_lock(&cs->cs_lock);
451 		goto out;
452 	}
453 	mreq->m_pkthdr.len = m_length(mreq, NULL);
454 
455 	cr->cr_xid = xid;
456 	mtx_lock(&cs->cs_lock);
457 
458 	/*
459 	 * Try to get a place in the congestion window.
460 	 */
461 	while (cu->cu_sent >= cu->cu_cwnd) {
462 		cu->cu_cwnd_wait = TRUE;
463 		error = msleep(&cu->cu_cwnd_wait, &cs->cs_lock,
464 		    cu->cu_waitflag, "rpccwnd", 0);
465 		if (error) {
466 			errp->re_errno = error;
467 			errp->re_status = stat = RPC_CANTSEND;
468 			goto out;
469 		}
470 	}
471 	cu->cu_sent += CWNDSCALE;
472 
473 	TAILQ_INSERT_TAIL(&cs->cs_pending, cr, cr_link);
474 	mtx_unlock(&cs->cs_lock);
475 
476 	/*
477 	 * sosend consumes mreq.
478 	 */
479 	error = sosend(cu->cu_socket, sa, NULL, mreq, NULL, 0, curthread);
480 	mreq = NULL;
481 
482 	/*
483 	 * sub-optimal code appears here because we have
484 	 * some clock time to spare while the packets are in flight.
485 	 * (We assume that this is actually only executed once.)
486 	 */
487 	reply_msg.acpted_rply.ar_verf.oa_flavor = AUTH_NULL;
488 	reply_msg.acpted_rply.ar_verf.oa_base = cr->cr_verf;
489 	reply_msg.acpted_rply.ar_verf.oa_length = 0;
490 	reply_msg.acpted_rply.ar_results.where = NULL;
491 	reply_msg.acpted_rply.ar_results.proc = (xdrproc_t)xdr_void;
492 
493 	mtx_lock(&cs->cs_lock);
494 	if (error) {
495 		TAILQ_REMOVE(&cs->cs_pending, cr, cr_link);
496 		errp->re_errno = error;
497 		errp->re_status = stat = RPC_CANTSEND;
498 		cu->cu_sent -= CWNDSCALE;
499 		if (cu->cu_cwnd_wait) {
500 			cu->cu_cwnd_wait = FALSE;
501 			wakeup(&cu->cu_cwnd_wait);
502 		}
503 		goto out;
504 	}
505 
506 	/*
507 	 * Check to see if we got an upcall while waiting for the
508 	 * lock.
509 	 */
510 	if (cr->cr_error) {
511 		TAILQ_REMOVE(&cs->cs_pending, cr, cr_link);
512 		errp->re_errno = cr->cr_error;
513 		errp->re_status = stat = RPC_CANTRECV;
514 		cu->cu_sent -= CWNDSCALE;
515 		if (cu->cu_cwnd_wait) {
516 			cu->cu_cwnd_wait = FALSE;
517 			wakeup(&cu->cu_cwnd_wait);
518 		}
519 		goto out;
520 	}
521 	if (cr->cr_mrep) {
522 		TAILQ_REMOVE(&cs->cs_pending, cr, cr_link);
523 		cu->cu_sent -= CWNDSCALE;
524 		if (cu->cu_cwnd_wait) {
525 			cu->cu_cwnd_wait = FALSE;
526 			wakeup(&cu->cu_cwnd_wait);
527 		}
528 		goto got_reply;
529 	}
530 
531 	/*
532 	 * Hack to provide rpc-based message passing
533 	 */
534 	if (timeout == 0) {
535 		TAILQ_REMOVE(&cs->cs_pending, cr, cr_link);
536 		errp->re_status = stat = RPC_TIMEDOUT;
537 		cu->cu_sent -= CWNDSCALE;
538 		if (cu->cu_cwnd_wait) {
539 			cu->cu_cwnd_wait = FALSE;
540 			wakeup(&cu->cu_cwnd_wait);
541 		}
542 		goto out;
543 	}
544 
545 get_reply:
546 	for (;;) {
547 		/* Decide how long to wait. */
548 		if (next_sendtime < timeout)
549 			tv = next_sendtime;
550 		else
551 			tv = timeout;
552 		tv -= time_waited;
553 
554 		if (tv > 0) {
555 			if (cu->cu_closing || cu->cu_closed) {
556 				error = 0;
557 				cr->cr_error = ESHUTDOWN;
558 			} else {
559 				error = msleep(cr, &cs->cs_lock,
560 				    cu->cu_waitflag, cu->cu_waitchan, tv);
561 			}
562 		} else {
563 			error = EWOULDBLOCK;
564 		}
565 
566 		TAILQ_REMOVE(&cs->cs_pending, cr, cr_link);
567 		cu->cu_sent -= CWNDSCALE;
568 		if (cu->cu_cwnd_wait) {
569 			cu->cu_cwnd_wait = FALSE;
570 			wakeup(&cu->cu_cwnd_wait);
571 		}
572 
573 		if (!error) {
574 			/*
575 			 * We were woken up by the upcall.  If the
576 			 * upcall had a receive error, report that,
577 			 * otherwise we have a reply.
578 			 */
579 			if (cr->cr_error) {
580 				errp->re_errno = cr->cr_error;
581 				errp->re_status = stat = RPC_CANTRECV;
582 				goto out;
583 			}
584 
585 			cu->cu_cwnd += (CWNDSCALE * CWNDSCALE
586 			    + cu->cu_cwnd / 2) / cu->cu_cwnd;
587 			if (cu->cu_cwnd > MAXCWND)
588 				cu->cu_cwnd = MAXCWND;
589 
590 			if (rt) {
591 				/*
592 				 * Add one to the time since a tick
593 				 * count of N means that the actual
594 				 * time taken was somewhere between N
595 				 * and N+1.
596 				 */
597 				rtt = ticks - starttime + 1;
598 
599 				/*
600 				 * Update our estimate of the round
601 				 * trip time using roughly the
602 				 * algorithm described in RFC
603 				 * 2988. Given an RTT sample R:
604 				 *
605 				 * RTTVAR = (1-beta) * RTTVAR + beta * |SRTT-R|
606 				 * SRTT = (1-alpha) * SRTT + alpha * R
607 				 *
608 				 * where alpha = 0.125 and beta = 0.25.
609 				 *
610 				 * The initial retransmit timeout is
611 				 * SRTT + 4*RTTVAR and doubles on each
612 				 * retransmision.
613 				 */
614 				if (rt->rt_srtt == 0) {
615 					rt->rt_srtt = rtt;
616 					rt->rt_deviate = rtt / 2;
617 				} else {
618 					int32_t error = rtt - rt->rt_srtt;
619 					rt->rt_srtt += error / 8;
620 					error = abs(error) - rt->rt_deviate;
621 					rt->rt_deviate += error / 4;
622 				}
623 				rt->rt_rtxcur = rt->rt_srtt + 4*rt->rt_deviate;
624 			}
625 
626 			break;
627 		}
628 
629 		/*
630 		 * The sleep returned an error so our request is still
631 		 * on the list. If we got EWOULDBLOCK, we may want to
632 		 * re-send the request.
633 		 */
634 		if (error != EWOULDBLOCK) {
635 			errp->re_errno = error;
636 			if (error == EINTR)
637 				errp->re_status = stat = RPC_INTR;
638 			else
639 				errp->re_status = stat = RPC_CANTRECV;
640 			goto out;
641 		}
642 
643 		time_waited = ticks - starttime;
644 
645 		/* Check for timeout. */
646 		if (time_waited > timeout) {
647 			errp->re_errno = EWOULDBLOCK;
648 			errp->re_status = stat = RPC_TIMEDOUT;
649 			goto out;
650 		}
651 
652 		/* Retransmit if necessary. */
653 		if (time_waited >= next_sendtime) {
654 			cu->cu_cwnd /= 2;
655 			if (cu->cu_cwnd < CWNDSCALE)
656 				cu->cu_cwnd = CWNDSCALE;
657 			if (ext && ext->rc_feedback) {
658 				mtx_unlock(&cs->cs_lock);
659 				if (retrans == 0)
660 					ext->rc_feedback(FEEDBACK_REXMIT1,
661 					    proc, ext->rc_feedback_arg);
662 				else
663 					ext->rc_feedback(FEEDBACK_REXMIT2,
664 					    proc, ext->rc_feedback_arg);
665 				mtx_lock(&cs->cs_lock);
666 			}
667 			if (cu->cu_closing || cu->cu_closed) {
668 				errp->re_errno = ESHUTDOWN;
669 				errp->re_status = stat = RPC_CANTRECV;
670 				goto out;
671 			}
672 			retrans++;
673 			/* update retransmit_time */
674 			if (retransmit_time < RPC_MAX_BACKOFF * hz)
675 				retransmit_time = 2 * retransmit_time;
676 			next_sendtime += retransmit_time;
677 			goto send_again;
678 		}
679 		TAILQ_INSERT_TAIL(&cs->cs_pending, cr, cr_link);
680 	}
681 
682 got_reply:
683 	/*
684 	 * Now decode and validate the response. We need to drop the
685 	 * lock since xdr_replymsg may end up sleeping in malloc.
686 	 */
687 	mtx_unlock(&cs->cs_lock);
688 
689 	if (ext && ext->rc_feedback)
690 		ext->rc_feedback(FEEDBACK_OK, proc, ext->rc_feedback_arg);
691 
692 	xdrmbuf_create(&xdrs, cr->cr_mrep, XDR_DECODE);
693 	ok = xdr_replymsg(&xdrs, &reply_msg);
694 	cr->cr_mrep = NULL;
695 
696 	if (ok) {
697 		if ((reply_msg.rm_reply.rp_stat == MSG_ACCEPTED) &&
698 		    (reply_msg.acpted_rply.ar_stat == SUCCESS))
699 			errp->re_status = stat = RPC_SUCCESS;
700 		else
701 			stat = _seterr_reply(&reply_msg, &(cu->cu_error));
702 
703 		if (errp->re_status == RPC_SUCCESS) {
704 			results = xdrmbuf_getall(&xdrs);
705 			if (! AUTH_VALIDATE(auth, xid,
706 				&reply_msg.acpted_rply.ar_verf,
707 				&results)) {
708 				errp->re_status = stat = RPC_AUTHERROR;
709 				errp->re_why = AUTH_INVALIDRESP;
710 				if (retrans &&
711 				    auth->ah_cred.oa_flavor == RPCSEC_GSS) {
712 					/*
713 					 * If we retransmitted, its
714 					 * possible that we will
715 					 * receive a reply for one of
716 					 * the earlier transmissions
717 					 * (which will use an older
718 					 * RPCSEC_GSS sequence
719 					 * number). In this case, just
720 					 * go back and listen for a
721 					 * new reply. We could keep a
722 					 * record of all the seq
723 					 * numbers we have transmitted
724 					 * so far so that we could
725 					 * accept a reply for any of
726 					 * them here.
727 					 */
728 					XDR_DESTROY(&xdrs);
729 					mtx_lock(&cs->cs_lock);
730 					TAILQ_INSERT_TAIL(&cs->cs_pending,
731 					    cr, cr_link);
732 					cr->cr_mrep = NULL;
733 					goto get_reply;
734 				}
735 			} else {
736 				*resultsp = results;
737 			}
738 		}		/* end successful completion */
739 		/*
740 		 * If unsuccesful AND error is an authentication error
741 		 * then refresh credentials and try again, else break
742 		 */
743 		else if (stat == RPC_AUTHERROR)
744 			/* maybe our credentials need to be refreshed ... */
745 			if (nrefreshes > 0 &&
746 			    AUTH_REFRESH(auth, &reply_msg)) {
747 				nrefreshes--;
748 				XDR_DESTROY(&xdrs);
749 				mtx_lock(&cs->cs_lock);
750 				goto call_again;
751 			}
752 		/* end of unsuccessful completion */
753 	}	/* end of valid reply message */
754 	else {
755 		errp->re_status = stat = RPC_CANTDECODERES;
756 
757 	}
758 	XDR_DESTROY(&xdrs);
759 	mtx_lock(&cs->cs_lock);
760 out:
761 	mtx_assert(&cs->cs_lock, MA_OWNED);
762 
763 	if (mreq)
764 		m_freem(mreq);
765 	if (cr->cr_mrep)
766 		m_freem(cr->cr_mrep);
767 
768 	cu->cu_threads--;
769 	if (cu->cu_closing)
770 		wakeup(cu);
771 
772 	mtx_unlock(&cs->cs_lock);
773 
774 	if (auth && stat != RPC_SUCCESS)
775 		AUTH_VALIDATE(auth, xid, NULL, NULL);
776 
777 	free(cr, M_RPC);
778 
779 	return (stat);
780 }
781 
782 static void
783 clnt_dg_geterr(CLIENT *cl, struct rpc_err *errp)
784 {
785 	struct cu_data *cu = (struct cu_data *)cl->cl_private;
786 
787 	*errp = cu->cu_error;
788 }
789 
790 static bool_t
791 clnt_dg_freeres(CLIENT *cl, xdrproc_t xdr_res, void *res_ptr)
792 {
793 	XDR xdrs;
794 	bool_t dummy;
795 
796 	xdrs.x_op = XDR_FREE;
797 	dummy = (*xdr_res)(&xdrs, res_ptr);
798 
799 	return (dummy);
800 }
801 
802 /*ARGSUSED*/
803 static void
804 clnt_dg_abort(CLIENT *h)
805 {
806 }
807 
808 static bool_t
809 clnt_dg_control(CLIENT *cl, u_int request, void *info)
810 {
811 	struct cu_data *cu = (struct cu_data *)cl->cl_private;
812 	struct cu_socket *cs;
813 	struct sockaddr *addr;
814 
815 	cs = cu->cu_socket->so_rcv.sb_upcallarg;
816 	mtx_lock(&cs->cs_lock);
817 
818 	switch (request) {
819 	case CLSET_FD_CLOSE:
820 		cu->cu_closeit = TRUE;
821 		mtx_unlock(&cs->cs_lock);
822 		return (TRUE);
823 	case CLSET_FD_NCLOSE:
824 		cu->cu_closeit = FALSE;
825 		mtx_unlock(&cs->cs_lock);
826 		return (TRUE);
827 	}
828 
829 	/* for other requests which use info */
830 	if (info == NULL) {
831 		mtx_unlock(&cs->cs_lock);
832 		return (FALSE);
833 	}
834 	switch (request) {
835 	case CLSET_TIMEOUT:
836 		if (time_not_ok((struct timeval *)info)) {
837 			mtx_unlock(&cs->cs_lock);
838 			return (FALSE);
839 		}
840 		cu->cu_total = *(struct timeval *)info;
841 		break;
842 	case CLGET_TIMEOUT:
843 		*(struct timeval *)info = cu->cu_total;
844 		break;
845 	case CLSET_RETRY_TIMEOUT:
846 		if (time_not_ok((struct timeval *)info)) {
847 			mtx_unlock(&cs->cs_lock);
848 			return (FALSE);
849 		}
850 		cu->cu_wait = *(struct timeval *)info;
851 		break;
852 	case CLGET_RETRY_TIMEOUT:
853 		*(struct timeval *)info = cu->cu_wait;
854 		break;
855 	case CLGET_SVC_ADDR:
856 		/*
857 		 * Slightly different semantics to userland - we use
858 		 * sockaddr instead of netbuf.
859 		 */
860 		memcpy(info, &cu->cu_raddr, cu->cu_raddr.ss_len);
861 		break;
862 	case CLSET_SVC_ADDR:		/* set to new address */
863 		addr = (struct sockaddr *)info;
864 		(void) memcpy(&cu->cu_raddr, addr, addr->sa_len);
865 		break;
866 	case CLGET_XID:
867 		*(uint32_t *)info = cu->cu_xid;
868 		break;
869 
870 	case CLSET_XID:
871 		/* This will set the xid of the NEXT call */
872 		/* decrement by 1 as clnt_dg_call() increments once */
873 		cu->cu_xid = *(uint32_t *)info - 1;
874 		break;
875 
876 	case CLGET_VERS:
877 		/*
878 		 * This RELIES on the information that, in the call body,
879 		 * the version number field is the fifth field from the
880 		 * begining of the RPC header. MUST be changed if the
881 		 * call_struct is changed
882 		 */
883 		*(uint32_t *)info =
884 		    ntohl(*(uint32_t *)(void *)(cu->cu_mcallc +
885 		    4 * BYTES_PER_XDR_UNIT));
886 		break;
887 
888 	case CLSET_VERS:
889 		*(uint32_t *)(void *)(cu->cu_mcallc + 4 * BYTES_PER_XDR_UNIT)
890 			= htonl(*(uint32_t *)info);
891 		break;
892 
893 	case CLGET_PROG:
894 		/*
895 		 * This RELIES on the information that, in the call body,
896 		 * the program number field is the fourth field from the
897 		 * begining of the RPC header. MUST be changed if the
898 		 * call_struct is changed
899 		 */
900 		*(uint32_t *)info =
901 		    ntohl(*(uint32_t *)(void *)(cu->cu_mcallc +
902 		    3 * BYTES_PER_XDR_UNIT));
903 		break;
904 
905 	case CLSET_PROG:
906 		*(uint32_t *)(void *)(cu->cu_mcallc + 3 * BYTES_PER_XDR_UNIT)
907 			= htonl(*(uint32_t *)info);
908 		break;
909 	case CLSET_ASYNC:
910 		cu->cu_async = *(int *)info;
911 		break;
912 	case CLSET_CONNECT:
913 		cu->cu_connect = *(int *)info;
914 		break;
915 	case CLSET_WAITCHAN:
916 		cu->cu_waitchan = (const char *)info;
917 		break;
918 	case CLGET_WAITCHAN:
919 		*(const char **) info = cu->cu_waitchan;
920 		break;
921 	case CLSET_INTERRUPTIBLE:
922 		if (*(int *) info)
923 			cu->cu_waitflag = PCATCH;
924 		else
925 			cu->cu_waitflag = 0;
926 		break;
927 	case CLGET_INTERRUPTIBLE:
928 		if (cu->cu_waitflag)
929 			*(int *) info = TRUE;
930 		else
931 			*(int *) info = FALSE;
932 		break;
933 	default:
934 		mtx_unlock(&cs->cs_lock);
935 		return (FALSE);
936 	}
937 	mtx_unlock(&cs->cs_lock);
938 	return (TRUE);
939 }
940 
941 static void
942 clnt_dg_close(CLIENT *cl)
943 {
944 	struct cu_data *cu = (struct cu_data *)cl->cl_private;
945 	struct cu_socket *cs;
946 	struct cu_request *cr;
947 
948 	cs = cu->cu_socket->so_rcv.sb_upcallarg;
949 	mtx_lock(&cs->cs_lock);
950 
951 	if (cu->cu_closed) {
952 		mtx_unlock(&cs->cs_lock);
953 		return;
954 	}
955 
956 	if (cu->cu_closing) {
957 		while (cu->cu_closing)
958 			msleep(cu, &cs->cs_lock, 0, "rpcclose", 0);
959 		KASSERT(cu->cu_closed, ("client should be closed"));
960 		mtx_unlock(&cs->cs_lock);
961 		return;
962 	}
963 
964 	/*
965 	 * Abort any pending requests and wait until everyone
966 	 * has finished with clnt_vc_call.
967 	 */
968 	cu->cu_closing = TRUE;
969 	TAILQ_FOREACH(cr, &cs->cs_pending, cr_link) {
970 		if (cr->cr_client == cl) {
971 			cr->cr_xid = 0;
972 			cr->cr_error = ESHUTDOWN;
973 			wakeup(cr);
974 		}
975 	}
976 
977 	while (cu->cu_threads)
978 		msleep(cu, &cs->cs_lock, 0, "rpcclose", 0);
979 
980 	cu->cu_closing = FALSE;
981 	cu->cu_closed = TRUE;
982 
983 	mtx_unlock(&cs->cs_lock);
984 	wakeup(cu);
985 }
986 
987 static void
988 clnt_dg_destroy(CLIENT *cl)
989 {
990 	struct cu_data *cu = (struct cu_data *)cl->cl_private;
991 	struct cu_socket *cs;
992 	struct socket *so = NULL;
993 	bool_t lastsocketref;
994 
995 	cs = cu->cu_socket->so_rcv.sb_upcallarg;
996 	clnt_dg_close(cl);
997 
998 	mtx_lock(&cs->cs_lock);
999 
1000 	cs->cs_refs--;
1001 	if (cs->cs_refs == 0) {
1002 		mtx_unlock(&cs->cs_lock);
1003 		SOCKBUF_LOCK(&cu->cu_socket->so_rcv);
1004 		soupcall_clear(cu->cu_socket, SO_RCV);
1005 		clnt_dg_upcallsdone(cu->cu_socket, cs);
1006 		SOCKBUF_UNLOCK(&cu->cu_socket->so_rcv);
1007 		mtx_destroy(&cs->cs_lock);
1008 		mem_free(cs, sizeof(*cs));
1009 		lastsocketref = TRUE;
1010 	} else {
1011 		mtx_unlock(&cs->cs_lock);
1012 		lastsocketref = FALSE;
1013 	}
1014 
1015 	if (cu->cu_closeit && lastsocketref) {
1016 		so = cu->cu_socket;
1017 		cu->cu_socket = NULL;
1018 	}
1019 
1020 	if (so)
1021 		soclose(so);
1022 
1023 	if (cl->cl_netid && cl->cl_netid[0])
1024 		mem_free(cl->cl_netid, strlen(cl->cl_netid) +1);
1025 	if (cl->cl_tp && cl->cl_tp[0])
1026 		mem_free(cl->cl_tp, strlen(cl->cl_tp) +1);
1027 	mem_free(cu, sizeof (*cu));
1028 	mem_free(cl, sizeof (CLIENT));
1029 }
1030 
1031 /*
1032  * Make sure that the time is not garbage.  -1 value is allowed.
1033  */
1034 static bool_t
1035 time_not_ok(struct timeval *t)
1036 {
1037 	return (t->tv_sec < -1 || t->tv_sec > 100000000 ||
1038 		t->tv_usec < -1 || t->tv_usec > 1000000);
1039 }
1040 
1041 int
1042 clnt_dg_soupcall(struct socket *so, void *arg, int waitflag)
1043 {
1044 	struct cu_socket *cs = (struct cu_socket *) arg;
1045 	struct uio uio;
1046 	struct mbuf *m;
1047 	struct mbuf *control;
1048 	struct cu_request *cr;
1049 	int error, rcvflag, foundreq;
1050 	uint32_t xid;
1051 
1052 	cs->cs_upcallrefs++;
1053 	uio.uio_resid = 1000000000;
1054 	uio.uio_td = curthread;
1055 	do {
1056 		SOCKBUF_UNLOCK(&so->so_rcv);
1057 		m = NULL;
1058 		control = NULL;
1059 		rcvflag = MSG_DONTWAIT;
1060 		error = soreceive(so, NULL, &uio, &m, &control, &rcvflag);
1061 		if (control)
1062 			m_freem(control);
1063 		SOCKBUF_LOCK(&so->so_rcv);
1064 
1065 		if (error == EWOULDBLOCK)
1066 			break;
1067 
1068 		/*
1069 		 * If there was an error, wake up all pending
1070 		 * requests.
1071 		 */
1072 		if (error) {
1073 			mtx_lock(&cs->cs_lock);
1074 			TAILQ_FOREACH(cr, &cs->cs_pending, cr_link) {
1075 				cr->cr_xid = 0;
1076 				cr->cr_error = error;
1077 				wakeup(cr);
1078 			}
1079 			mtx_unlock(&cs->cs_lock);
1080 			break;
1081 		}
1082 
1083 		/*
1084 		 * The XID is in the first uint32_t of the reply.
1085 		 */
1086 		if (m->m_len < sizeof(xid))
1087 			m = m_pullup(m, sizeof(xid));
1088 		if (!m)
1089 			/*
1090 			 * Should never happen.
1091 			 */
1092 			continue;
1093 
1094 		xid = ntohl(*mtod(m, uint32_t *));
1095 
1096 		/*
1097 		 * Attempt to match this reply with a pending request.
1098 		 */
1099 		mtx_lock(&cs->cs_lock);
1100 		foundreq = 0;
1101 		TAILQ_FOREACH(cr, &cs->cs_pending, cr_link) {
1102 			if (cr->cr_xid == xid) {
1103 				/*
1104 				 * This one matches. We leave the
1105 				 * reply mbuf in cr->cr_mrep. Set the
1106 				 * XID to zero so that we will ignore
1107 				 * any duplicated replies that arrive
1108 				 * before clnt_dg_call removes it from
1109 				 * the queue.
1110 				 */
1111 				cr->cr_xid = 0;
1112 				cr->cr_mrep = m;
1113 				cr->cr_error = 0;
1114 				foundreq = 1;
1115 				wakeup(cr);
1116 				break;
1117 			}
1118 		}
1119 		mtx_unlock(&cs->cs_lock);
1120 
1121 		/*
1122 		 * If we didn't find the matching request, just drop
1123 		 * it - its probably a repeated reply.
1124 		 */
1125 		if (!foundreq)
1126 			m_freem(m);
1127 	} while (m);
1128 	cs->cs_upcallrefs--;
1129 	if (cs->cs_upcallrefs < 0)
1130 		panic("rpcdg upcall refcnt");
1131 	if (cs->cs_upcallrefs == 0)
1132 		wakeup(&cs->cs_upcallrefs);
1133 	return (SU_OK);
1134 }
1135 
1136 /*
1137  * Wait for all upcalls in progress to complete.
1138  */
1139 static void
1140 clnt_dg_upcallsdone(struct socket *so, struct cu_socket *cs)
1141 {
1142 
1143 	SOCKBUF_LOCK_ASSERT(&so->so_rcv);
1144 
1145 	while (cs->cs_upcallrefs > 0)
1146 		(void) msleep(&cs->cs_upcallrefs, SOCKBUF_MTX(&so->so_rcv), 0,
1147 		    "rpcdgup", 0);
1148 }
1149