xref: /freebsd/sys/rpc/clnt_dg.c (revision d0b2dbfa)
1 /*	$NetBSD: clnt_dg.c,v 1.4 2000/07/14 08:40:41 fvdl Exp $	*/
2 
3 /*-
4  * SPDX-License-Identifier: BSD-3-Clause
5  *
6  * Copyright (c) 2009, Sun Microsystems, Inc.
7  * All rights reserved.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions are met:
11  * - Redistributions of source code must retain the above copyright notice,
12  *   this list of conditions and the following disclaimer.
13  * - Redistributions in binary form must reproduce the above copyright notice,
14  *   this list of conditions and the following disclaimer in the documentation
15  *   and/or other materials provided with the distribution.
16  * - Neither the name of Sun Microsystems, Inc. nor the names of its
17  *   contributors may be used to endorse or promote products derived
18  *   from this software without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
21  * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
24  * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
25  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
26  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
27  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
28  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
29  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
30  * POSSIBILITY OF SUCH DAMAGE.
31  */
32 /*
33  * Copyright (c) 1986-1991 by Sun Microsystems Inc.
34  */
35 
36 #if defined(LIBC_SCCS) && !defined(lint)
37 #ident	"@(#)clnt_dg.c	1.23	94/04/22 SMI"
38 static char sccsid[] = "@(#)clnt_dg.c 1.19 89/03/16 Copyr 1988 Sun Micro";
39 #endif
40 #include <sys/cdefs.h>
41 /*
42  * Implements a connectionless client side RPC.
43  */
44 
45 #include <sys/param.h>
46 #include <sys/systm.h>
47 #include <sys/kernel.h>
48 #include <sys/lock.h>
49 #include <sys/malloc.h>
50 #include <sys/mbuf.h>
51 #include <sys/mutex.h>
52 #include <sys/pcpu.h>
53 #include <sys/proc.h>
54 #include <sys/socket.h>
55 #include <sys/socketvar.h>
56 #include <sys/time.h>
57 #include <sys/uio.h>
58 
59 #include <net/vnet.h>
60 
61 #include <rpc/rpc.h>
62 #include <rpc/rpc_com.h>
63 
64 
65 #ifdef _FREEFALL_CONFIG
66 /*
67  * Disable RPC exponential back-off for FreeBSD.org systems.
68  */
69 #define	RPC_MAX_BACKOFF		1 /* second */
70 #else
71 #define	RPC_MAX_BACKOFF		30 /* seconds */
72 #endif
73 
74 static bool_t time_not_ok(struct timeval *);
75 static enum clnt_stat clnt_dg_call(CLIENT *, struct rpc_callextra *,
76     rpcproc_t, struct mbuf *, struct mbuf **, struct timeval);
77 static void clnt_dg_geterr(CLIENT *, struct rpc_err *);
78 static bool_t clnt_dg_freeres(CLIENT *, xdrproc_t, void *);
79 static void clnt_dg_abort(CLIENT *);
80 static bool_t clnt_dg_control(CLIENT *, u_int, void *);
81 static void clnt_dg_close(CLIENT *);
82 static void clnt_dg_destroy(CLIENT *);
83 static int clnt_dg_soupcall(struct socket *so, void *arg, int waitflag);
84 
85 static const struct clnt_ops clnt_dg_ops = {
86 	.cl_call =	clnt_dg_call,
87 	.cl_abort =	clnt_dg_abort,
88 	.cl_geterr =	clnt_dg_geterr,
89 	.cl_freeres =	clnt_dg_freeres,
90 	.cl_close =	clnt_dg_close,
91 	.cl_destroy =	clnt_dg_destroy,
92 	.cl_control =	clnt_dg_control
93 };
94 
95 static volatile uint32_t rpc_xid = 0;
96 
97 /*
98  * A pending RPC request which awaits a reply. Requests which have
99  * received their reply will have cr_xid set to zero and cr_mrep to
100  * the mbuf chain of the reply.
101  */
102 struct cu_request {
103 	TAILQ_ENTRY(cu_request) cr_link;
104 	CLIENT			*cr_client;	/* owner */
105 	uint32_t		cr_xid;		/* XID of request */
106 	struct mbuf		*cr_mrep;	/* reply received by upcall */
107 	int			cr_error;	/* any error from upcall */
108 	char			cr_verf[MAX_AUTH_BYTES]; /* reply verf */
109 };
110 
111 TAILQ_HEAD(cu_request_list, cu_request);
112 
113 #define MCALL_MSG_SIZE 24
114 
115 /*
116  * This structure is pointed to by the socket buffer's sb_upcallarg
117  * member. It is separate from the client private data to facilitate
118  * multiple clients sharing the same socket. The cs_lock mutex is used
119  * to protect all fields of this structure, the socket's receive
120  * buffer SOCKBUF_LOCK is used to ensure that exactly one of these
121  * structures is installed on the socket.
122  */
123 struct cu_socket {
124 	struct mtx		cs_lock;
125 	int			cs_refs;	/* Count of clients */
126 	struct cu_request_list	cs_pending;	/* Requests awaiting replies */
127 	int			cs_upcallrefs;	/* Refcnt of upcalls in prog.*/
128 };
129 
130 static void clnt_dg_upcallsdone(struct socket *, struct cu_socket *);
131 
132 /*
133  * Private data kept per client handle
134  */
135 struct cu_data {
136 	int			cu_threads;	/* # threads in clnt_vc_call */
137 	bool_t			cu_closing;	/* TRUE if we are closing */
138 	bool_t			cu_closed;	/* TRUE if we are closed */
139 	struct socket		*cu_socket;	/* connection socket */
140 	bool_t			cu_closeit;	/* opened by library */
141 	struct sockaddr_storage	cu_raddr;	/* remote address */
142 	int			cu_rlen;
143 	struct timeval		cu_wait;	/* retransmit interval */
144 	struct timeval		cu_total;	/* total time for the call */
145 	struct rpc_err		cu_error;
146 	uint32_t		cu_xid;
147 	char			cu_mcallc[MCALL_MSG_SIZE]; /* marshalled callmsg */
148 	size_t			cu_mcalllen;
149 	size_t			cu_sendsz;	/* send size */
150 	size_t			cu_recvsz;	/* recv size */
151 	int			cu_async;
152 	int			cu_connect;	/* Use connect(). */
153 	int			cu_connected;	/* Have done connect(). */
154 	const char		*cu_waitchan;
155 	int			cu_waitflag;
156 	int			cu_cwnd;	/* congestion window */
157 	int			cu_sent;	/* number of in-flight RPCs */
158 	bool_t			cu_cwnd_wait;
159 };
160 
161 #define CWNDSCALE	256
162 #define MAXCWND		(32 * CWNDSCALE)
163 
164 /*
165  * Connection less client creation returns with client handle parameters.
166  * Default options are set, which the user can change using clnt_control().
167  * fd should be open and bound.
168  * NB: The rpch->cl_auth is initialized to null authentication.
169  * 	Caller may wish to set this something more useful.
170  *
171  * sendsz and recvsz are the maximum allowable packet sizes that can be
172  * sent and received. Normally they are the same, but they can be
173  * changed to improve the program efficiency and buffer allocation.
174  * If they are 0, use the transport default.
175  *
176  * If svcaddr is NULL, returns NULL.
177  */
178 CLIENT *
179 clnt_dg_create(
180 	struct socket *so,
181 	struct sockaddr *svcaddr,	/* servers address */
182 	rpcprog_t program,		/* program number */
183 	rpcvers_t version,		/* version number */
184 	size_t sendsz,			/* buffer recv size */
185 	size_t recvsz)			/* buffer send size */
186 {
187 	CLIENT *cl = NULL;		/* client handle */
188 	struct cu_data *cu = NULL;	/* private data */
189 	struct cu_socket *cs = NULL;
190 	struct sockbuf *sb;
191 	struct timeval now;
192 	struct rpc_msg call_msg;
193 	struct __rpc_sockinfo si;
194 	XDR xdrs;
195 	int error;
196 	uint32_t newxid;
197 
198 	if (svcaddr == NULL) {
199 		rpc_createerr.cf_stat = RPC_UNKNOWNADDR;
200 		return (NULL);
201 	}
202 
203 	if (!__rpc_socket2sockinfo(so, &si)) {
204 		rpc_createerr.cf_stat = RPC_TLIERROR;
205 		rpc_createerr.cf_error.re_errno = 0;
206 		return (NULL);
207 	}
208 
209 	/*
210 	 * Find the receive and the send size
211 	 */
212 	sendsz = __rpc_get_t_size(si.si_af, si.si_proto, (int)sendsz);
213 	recvsz = __rpc_get_t_size(si.si_af, si.si_proto, (int)recvsz);
214 	if ((sendsz == 0) || (recvsz == 0)) {
215 		rpc_createerr.cf_stat = RPC_TLIERROR; /* XXX */
216 		rpc_createerr.cf_error.re_errno = 0;
217 		return (NULL);
218 	}
219 
220 	cl = mem_alloc(sizeof (CLIENT));
221 
222 	/*
223 	 * Should be multiple of 4 for XDR.
224 	 */
225 	sendsz = rounddown(sendsz + 3, 4);
226 	recvsz = rounddown(recvsz + 3, 4);
227 	cu = mem_alloc(sizeof (*cu));
228 	cu->cu_threads = 0;
229 	cu->cu_closing = FALSE;
230 	cu->cu_closed = FALSE;
231 	(void) memcpy(&cu->cu_raddr, svcaddr, (size_t)svcaddr->sa_len);
232 	cu->cu_rlen = svcaddr->sa_len;
233 	/* Other values can also be set through clnt_control() */
234 	cu->cu_wait.tv_sec = 3;	/* heuristically chosen */
235 	cu->cu_wait.tv_usec = 0;
236 	cu->cu_total.tv_sec = -1;
237 	cu->cu_total.tv_usec = -1;
238 	cu->cu_sendsz = sendsz;
239 	cu->cu_recvsz = recvsz;
240 	cu->cu_async = FALSE;
241 	cu->cu_connect = FALSE;
242 	cu->cu_connected = FALSE;
243 	cu->cu_waitchan = "rpcrecv";
244 	cu->cu_waitflag = 0;
245 	cu->cu_cwnd = MAXCWND / 2;
246 	cu->cu_sent = 0;
247 	cu->cu_cwnd_wait = FALSE;
248 	(void) getmicrotime(&now);
249 	/* Clip at 28bits so that it will not wrap around. */
250 	newxid = __RPC_GETXID(&now) & 0xfffffff;
251 	atomic_cmpset_32(&rpc_xid, 0, newxid);
252 	call_msg.rm_xid = atomic_fetchadd_32(&rpc_xid, 1);
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 	error = soreserve(so, (u_long)sendsz, (u_long)recvsz);
271 	if (error != 0) {
272 		rpc_createerr.cf_stat = RPC_FAILED;
273 		rpc_createerr.cf_error.re_errno = error;
274 		goto err2;
275 	}
276 
277 	sb = &so->so_rcv;
278 	SOCKBUF_LOCK(&so->so_rcv);
279 recheck_socket:
280 	if (sb->sb_upcall) {
281 		if (sb->sb_upcall != clnt_dg_soupcall) {
282 			SOCKBUF_UNLOCK(&so->so_rcv);
283 			printf("clnt_dg_create(): socket already has an incompatible upcall\n");
284 			goto err2;
285 		}
286 		cs = (struct cu_socket *) sb->sb_upcallarg;
287 		mtx_lock(&cs->cs_lock);
288 		cs->cs_refs++;
289 		mtx_unlock(&cs->cs_lock);
290 	} else {
291 		/*
292 		 * We are the first on this socket - allocate the
293 		 * structure and install it in the socket.
294 		 */
295 		SOCKBUF_UNLOCK(&so->so_rcv);
296 		cs = mem_alloc(sizeof(*cs));
297 		SOCKBUF_LOCK(&so->so_rcv);
298 		if (sb->sb_upcall) {
299 			/*
300 			 * We have lost a race with some other client.
301 			 */
302 			mem_free(cs, sizeof(*cs));
303 			goto recheck_socket;
304 		}
305 		mtx_init(&cs->cs_lock, "cs->cs_lock", NULL, MTX_DEF);
306 		cs->cs_refs = 1;
307 		cs->cs_upcallrefs = 0;
308 		TAILQ_INIT(&cs->cs_pending);
309 		soupcall_set(so, SO_RCV, clnt_dg_soupcall, cs);
310 	}
311 	SOCKBUF_UNLOCK(&so->so_rcv);
312 
313 	cl->cl_refs = 1;
314 	cl->cl_ops = &clnt_dg_ops;
315 	cl->cl_private = (caddr_t)(void *)cu;
316 	cl->cl_auth = authnone_create();
317 	cl->cl_tp = NULL;
318 	cl->cl_netid = NULL;
319 	return (cl);
320 err2:
321 	mem_free(cl, sizeof (CLIENT));
322 	mem_free(cu, sizeof (*cu));
323 
324 	return (NULL);
325 }
326 
327 static enum clnt_stat
328 clnt_dg_call(
329 	CLIENT		*cl,		/* client handle */
330 	struct rpc_callextra *ext,	/* call metadata */
331 	rpcproc_t	proc,		/* procedure number */
332 	struct mbuf	*args,		/* pointer to args */
333 	struct mbuf	**resultsp,	/* pointer to results */
334 	struct timeval	utimeout)	/* seconds to wait before giving up */
335 {
336 	struct cu_data *cu = (struct cu_data *)cl->cl_private;
337 	struct cu_socket *cs;
338 	struct rpc_timers *rt;
339 	AUTH *auth;
340 	struct rpc_err *errp;
341 	enum clnt_stat stat;
342 	XDR xdrs;
343 	struct rpc_msg reply_msg;
344 	bool_t ok;
345 	int retrans;			/* number of re-transmits so far */
346 	int nrefreshes = 2;		/* number of times to refresh cred */
347 	struct timeval *tvp;
348 	int timeout;
349 	int retransmit_time;
350 	int next_sendtime, starttime, rtt, time_waited, tv = 0;
351 	struct sockaddr *sa;
352 	uint32_t xid = 0;
353 	struct mbuf *mreq = NULL, *results;
354 	struct cu_request *cr;
355 	int error;
356 
357 	cs = cu->cu_socket->so_rcv.sb_upcallarg;
358 	cr = malloc(sizeof(struct cu_request), M_RPC, M_WAITOK);
359 
360 	mtx_lock(&cs->cs_lock);
361 
362 	if (cu->cu_closing || cu->cu_closed) {
363 		mtx_unlock(&cs->cs_lock);
364 		free(cr, M_RPC);
365 		return (RPC_CANTSEND);
366 	}
367 	cu->cu_threads++;
368 
369 	if (ext) {
370 		auth = ext->rc_auth;
371 		errp = &ext->rc_err;
372 	} else {
373 		auth = cl->cl_auth;
374 		errp = &cu->cu_error;
375 	}
376 
377 	cr->cr_client = cl;
378 	cr->cr_mrep = NULL;
379 	cr->cr_error = 0;
380 
381 	if (cu->cu_total.tv_usec == -1) {
382 		tvp = &utimeout; /* use supplied timeout */
383 	} else {
384 		tvp = &cu->cu_total; /* use default timeout */
385 	}
386 	if (tvp->tv_sec || tvp->tv_usec)
387 		timeout = tvtohz(tvp);
388 	else
389 		timeout = 0;
390 
391 	if (cu->cu_connect && !cu->cu_connected) {
392 		mtx_unlock(&cs->cs_lock);
393 		error = soconnect(cu->cu_socket,
394 		    (struct sockaddr *)&cu->cu_raddr, curthread);
395 		mtx_lock(&cs->cs_lock);
396 		if (error) {
397 			errp->re_errno = error;
398 			errp->re_status = stat = RPC_CANTSEND;
399 			goto out;
400 		}
401 		cu->cu_connected = 1;
402 	}
403 	if (cu->cu_connected)
404 		sa = NULL;
405 	else
406 		sa = (struct sockaddr *)&cu->cu_raddr;
407 	time_waited = 0;
408 	retrans = 0;
409 	if (ext && ext->rc_timers) {
410 		rt = ext->rc_timers;
411 		if (!rt->rt_rtxcur)
412 			rt->rt_rtxcur = tvtohz(&cu->cu_wait);
413 		retransmit_time = next_sendtime = rt->rt_rtxcur;
414 	} else {
415 		rt = NULL;
416 		retransmit_time = next_sendtime = tvtohz(&cu->cu_wait);
417 	}
418 
419 	starttime = ticks;
420 
421 call_again:
422 	mtx_assert(&cs->cs_lock, MA_OWNED);
423 
424 	xid = atomic_fetchadd_32(&rpc_xid, 1);
425 
426 send_again:
427 	mtx_unlock(&cs->cs_lock);
428 
429 	mreq = m_gethdr(M_WAITOK, MT_DATA);
430 	KASSERT(cu->cu_mcalllen <= MHLEN, ("RPC header too big"));
431 	bcopy(cu->cu_mcallc, mreq->m_data, cu->cu_mcalllen);
432 	mreq->m_len = cu->cu_mcalllen;
433 
434 	/*
435 	 * The XID is the first thing in the request.
436 	 */
437 	*mtod(mreq, uint32_t *) = htonl(xid);
438 
439 	xdrmbuf_create(&xdrs, mreq, XDR_ENCODE);
440 
441 	if (cu->cu_async == TRUE && args == NULL)
442 		goto get_reply;
443 
444 	if ((! XDR_PUTINT32(&xdrs, &proc)) ||
445 	    (! AUTH_MARSHALL(auth, xid, &xdrs,
446 		m_copym(args, 0, M_COPYALL, M_WAITOK)))) {
447 		errp->re_status = stat = RPC_CANTENCODEARGS;
448 		mtx_lock(&cs->cs_lock);
449 		goto out;
450 	}
451 	mreq->m_pkthdr.len = m_length(mreq, NULL);
452 
453 	cr->cr_xid = xid;
454 	mtx_lock(&cs->cs_lock);
455 
456 	/*
457 	 * Try to get a place in the congestion window.
458 	 */
459 	while (cu->cu_sent >= cu->cu_cwnd) {
460 		cu->cu_cwnd_wait = TRUE;
461 		error = msleep(&cu->cu_cwnd_wait, &cs->cs_lock,
462 		    cu->cu_waitflag, "rpccwnd", 0);
463 		if (error) {
464 			errp->re_errno = error;
465 			if (error == EINTR || error == ERESTART)
466 				errp->re_status = stat = RPC_INTR;
467 			else
468 				errp->re_status = stat = RPC_CANTSEND;
469 			goto out;
470 		}
471 	}
472 	cu->cu_sent += CWNDSCALE;
473 
474 	TAILQ_INSERT_TAIL(&cs->cs_pending, cr, cr_link);
475 	mtx_unlock(&cs->cs_lock);
476 
477 	/*
478 	 * sosend consumes mreq.
479 	 */
480 	error = sosend(cu->cu_socket, sa, NULL, mreq, NULL, 0, curthread);
481 	mreq = NULL;
482 
483 	/*
484 	 * sub-optimal code appears here because we have
485 	 * some clock time to spare while the packets are in flight.
486 	 * (We assume that this is actually only executed once.)
487 	 */
488 	reply_msg.acpted_rply.ar_verf.oa_flavor = AUTH_NULL;
489 	reply_msg.acpted_rply.ar_verf.oa_base = cr->cr_verf;
490 	reply_msg.acpted_rply.ar_verf.oa_length = 0;
491 	reply_msg.acpted_rply.ar_results.where = NULL;
492 	reply_msg.acpted_rply.ar_results.proc = (xdrproc_t)xdr_void;
493 
494 	mtx_lock(&cs->cs_lock);
495 	if (error) {
496 		TAILQ_REMOVE(&cs->cs_pending, cr, cr_link);
497 		errp->re_errno = error;
498 		errp->re_status = stat = RPC_CANTSEND;
499 		cu->cu_sent -= CWNDSCALE;
500 		if (cu->cu_cwnd_wait) {
501 			cu->cu_cwnd_wait = FALSE;
502 			wakeup(&cu->cu_cwnd_wait);
503 		}
504 		goto out;
505 	}
506 
507 	/*
508 	 * Check to see if we got an upcall while waiting for the
509 	 * lock.
510 	 */
511 	if (cr->cr_error) {
512 		TAILQ_REMOVE(&cs->cs_pending, cr, cr_link);
513 		errp->re_errno = cr->cr_error;
514 		errp->re_status = stat = RPC_CANTRECV;
515 		cu->cu_sent -= CWNDSCALE;
516 		if (cu->cu_cwnd_wait) {
517 			cu->cu_cwnd_wait = FALSE;
518 			wakeup(&cu->cu_cwnd_wait);
519 		}
520 		goto out;
521 	}
522 	if (cr->cr_mrep) {
523 		TAILQ_REMOVE(&cs->cs_pending, cr, cr_link);
524 		cu->cu_sent -= CWNDSCALE;
525 		if (cu->cu_cwnd_wait) {
526 			cu->cu_cwnd_wait = FALSE;
527 			wakeup(&cu->cu_cwnd_wait);
528 		}
529 		goto got_reply;
530 	}
531 
532 	/*
533 	 * Hack to provide rpc-based message passing
534 	 */
535 	if (timeout == 0) {
536 		TAILQ_REMOVE(&cs->cs_pending, cr, cr_link);
537 		errp->re_status = stat = RPC_TIMEDOUT;
538 		cu->cu_sent -= CWNDSCALE;
539 		if (cu->cu_cwnd_wait) {
540 			cu->cu_cwnd_wait = FALSE;
541 			wakeup(&cu->cu_cwnd_wait);
542 		}
543 		goto out;
544 	}
545 
546 get_reply:
547 	for (;;) {
548 		/* Decide how long to wait. */
549 		if (next_sendtime < timeout)
550 			tv = next_sendtime;
551 		else
552 			tv = timeout;
553 		tv -= time_waited;
554 
555 		if (tv > 0) {
556 			if (cu->cu_closing || cu->cu_closed) {
557 				error = 0;
558 				cr->cr_error = ESHUTDOWN;
559 			} else {
560 				error = msleep(cr, &cs->cs_lock,
561 				    cu->cu_waitflag, cu->cu_waitchan, tv);
562 			}
563 		} else {
564 			error = EWOULDBLOCK;
565 		}
566 
567 		TAILQ_REMOVE(&cs->cs_pending, cr, cr_link);
568 		cu->cu_sent -= CWNDSCALE;
569 		if (cu->cu_cwnd_wait) {
570 			cu->cu_cwnd_wait = FALSE;
571 			wakeup(&cu->cu_cwnd_wait);
572 		}
573 
574 		if (!error) {
575 			/*
576 			 * We were woken up by the upcall.  If the
577 			 * upcall had a receive error, report that,
578 			 * otherwise we have a reply.
579 			 */
580 			if (cr->cr_error) {
581 				errp->re_errno = cr->cr_error;
582 				errp->re_status = stat = RPC_CANTRECV;
583 				goto out;
584 			}
585 
586 			cu->cu_cwnd += (CWNDSCALE * CWNDSCALE
587 			    + cu->cu_cwnd / 2) / cu->cu_cwnd;
588 			if (cu->cu_cwnd > MAXCWND)
589 				cu->cu_cwnd = MAXCWND;
590 
591 			if (rt) {
592 				/*
593 				 * Add one to the time since a tick
594 				 * count of N means that the actual
595 				 * time taken was somewhere between N
596 				 * and N+1.
597 				 */
598 				rtt = ticks - starttime + 1;
599 
600 				/*
601 				 * Update our estimate of the round
602 				 * trip time using roughly the
603 				 * algorithm described in RFC
604 				 * 2988. Given an RTT sample R:
605 				 *
606 				 * RTTVAR = (1-beta) * RTTVAR + beta * |SRTT-R|
607 				 * SRTT = (1-alpha) * SRTT + alpha * R
608 				 *
609 				 * where alpha = 0.125 and beta = 0.25.
610 				 *
611 				 * The initial retransmit timeout is
612 				 * SRTT + 4*RTTVAR and doubles on each
613 				 * retransmision.
614 				 */
615 				if (rt->rt_srtt == 0) {
616 					rt->rt_srtt = rtt;
617 					rt->rt_deviate = rtt / 2;
618 				} else {
619 					int32_t error = rtt - rt->rt_srtt;
620 					rt->rt_srtt += error / 8;
621 					error = abs(error) - rt->rt_deviate;
622 					rt->rt_deviate += error / 4;
623 				}
624 				rt->rt_rtxcur = rt->rt_srtt + 4*rt->rt_deviate;
625 			}
626 
627 			break;
628 		}
629 
630 		/*
631 		 * The sleep returned an error so our request is still
632 		 * on the list. If we got EWOULDBLOCK, we may want to
633 		 * re-send the request.
634 		 */
635 		if (error != EWOULDBLOCK) {
636 			errp->re_errno = error;
637 			if (error == EINTR || error == ERESTART)
638 				errp->re_status = stat = RPC_INTR;
639 			else
640 				errp->re_status = stat = RPC_CANTRECV;
641 			goto out;
642 		}
643 
644 		time_waited = ticks - starttime;
645 
646 		/* Check for timeout. */
647 		if (time_waited > timeout) {
648 			errp->re_errno = EWOULDBLOCK;
649 			errp->re_status = stat = RPC_TIMEDOUT;
650 			goto out;
651 		}
652 
653 		/* Retransmit if necessary. */
654 		if (time_waited >= next_sendtime) {
655 			cu->cu_cwnd /= 2;
656 			if (cu->cu_cwnd < CWNDSCALE)
657 				cu->cu_cwnd = CWNDSCALE;
658 			if (ext && ext->rc_feedback) {
659 				mtx_unlock(&cs->cs_lock);
660 				if (retrans == 0)
661 					ext->rc_feedback(FEEDBACK_REXMIT1,
662 					    proc, ext->rc_feedback_arg);
663 				else
664 					ext->rc_feedback(FEEDBACK_REXMIT2,
665 					    proc, ext->rc_feedback_arg);
666 				mtx_lock(&cs->cs_lock);
667 			}
668 			if (cu->cu_closing || cu->cu_closed) {
669 				errp->re_errno = ESHUTDOWN;
670 				errp->re_status = stat = RPC_CANTRECV;
671 				goto out;
672 			}
673 			retrans++;
674 			/* update retransmit_time */
675 			if (retransmit_time < RPC_MAX_BACKOFF * hz)
676 				retransmit_time = 2 * retransmit_time;
677 			next_sendtime += retransmit_time;
678 			goto send_again;
679 		}
680 		cu->cu_sent += CWNDSCALE;
681 		TAILQ_INSERT_TAIL(&cs->cs_pending, cr, cr_link);
682 	}
683 
684 got_reply:
685 	/*
686 	 * Now decode and validate the response. We need to drop the
687 	 * lock since xdr_replymsg may end up sleeping in malloc.
688 	 */
689 	mtx_unlock(&cs->cs_lock);
690 
691 	if (ext && ext->rc_feedback)
692 		ext->rc_feedback(FEEDBACK_OK, proc, ext->rc_feedback_arg);
693 
694 	xdrmbuf_create(&xdrs, cr->cr_mrep, XDR_DECODE);
695 	ok = xdr_replymsg(&xdrs, &reply_msg);
696 	cr->cr_mrep = NULL;
697 
698 	if (ok) {
699 		if ((reply_msg.rm_reply.rp_stat == MSG_ACCEPTED) &&
700 		    (reply_msg.acpted_rply.ar_stat == SUCCESS))
701 			errp->re_status = stat = RPC_SUCCESS;
702 		else
703 			stat = _seterr_reply(&reply_msg, &(cu->cu_error));
704 
705 		if (errp->re_status == RPC_SUCCESS) {
706 			results = xdrmbuf_getall(&xdrs);
707 			if (! AUTH_VALIDATE(auth, xid,
708 				&reply_msg.acpted_rply.ar_verf,
709 				&results)) {
710 				errp->re_status = stat = RPC_AUTHERROR;
711 				errp->re_why = AUTH_INVALIDRESP;
712 				if (retrans &&
713 				    auth->ah_cred.oa_flavor == RPCSEC_GSS) {
714 					/*
715 					 * If we retransmitted, its
716 					 * possible that we will
717 					 * receive a reply for one of
718 					 * the earlier transmissions
719 					 * (which will use an older
720 					 * RPCSEC_GSS sequence
721 					 * number). In this case, just
722 					 * go back and listen for a
723 					 * new reply. We could keep a
724 					 * record of all the seq
725 					 * numbers we have transmitted
726 					 * so far so that we could
727 					 * accept a reply for any of
728 					 * them here.
729 					 */
730 					XDR_DESTROY(&xdrs);
731 					mtx_lock(&cs->cs_lock);
732 					cu->cu_sent += CWNDSCALE;
733 					TAILQ_INSERT_TAIL(&cs->cs_pending,
734 					    cr, cr_link);
735 					cr->cr_mrep = NULL;
736 					goto get_reply;
737 				}
738 			} else {
739 				*resultsp = results;
740 			}
741 		}		/* end successful completion */
742 		/*
743 		 * If unsuccessful AND error is an authentication error
744 		 * then refresh credentials and try again, else break
745 		 */
746 		else if (stat == RPC_AUTHERROR)
747 			/* maybe our credentials need to be refreshed ... */
748 			if (nrefreshes > 0 &&
749 			    AUTH_REFRESH(auth, &reply_msg)) {
750 				nrefreshes--;
751 				XDR_DESTROY(&xdrs);
752 				mtx_lock(&cs->cs_lock);
753 				goto call_again;
754 			}
755 		/* end of unsuccessful completion */
756 	}	/* end of valid reply message */
757 	else {
758 		errp->re_status = stat = RPC_CANTDECODERES;
759 
760 	}
761 	XDR_DESTROY(&xdrs);
762 	mtx_lock(&cs->cs_lock);
763 out:
764 	mtx_assert(&cs->cs_lock, MA_OWNED);
765 
766 	if (mreq)
767 		m_freem(mreq);
768 	if (cr->cr_mrep)
769 		m_freem(cr->cr_mrep);
770 
771 	cu->cu_threads--;
772 	if (cu->cu_closing)
773 		wakeup(cu);
774 
775 	mtx_unlock(&cs->cs_lock);
776 
777 	if (auth && stat != RPC_SUCCESS)
778 		AUTH_VALIDATE(auth, xid, NULL, NULL);
779 
780 	free(cr, M_RPC);
781 
782 	return (stat);
783 }
784 
785 static void
786 clnt_dg_geterr(CLIENT *cl, struct rpc_err *errp)
787 {
788 	struct cu_data *cu = (struct cu_data *)cl->cl_private;
789 
790 	*errp = cu->cu_error;
791 }
792 
793 static bool_t
794 clnt_dg_freeres(CLIENT *cl, xdrproc_t xdr_res, void *res_ptr)
795 {
796 	XDR xdrs;
797 	bool_t dummy;
798 
799 	xdrs.x_op = XDR_FREE;
800 	dummy = (*xdr_res)(&xdrs, res_ptr);
801 
802 	return (dummy);
803 }
804 
805 /*ARGSUSED*/
806 static void
807 clnt_dg_abort(CLIENT *h)
808 {
809 }
810 
811 static bool_t
812 clnt_dg_control(CLIENT *cl, u_int request, void *info)
813 {
814 	struct cu_data *cu = (struct cu_data *)cl->cl_private;
815 	struct cu_socket *cs;
816 	struct sockaddr *addr;
817 
818 	cs = cu->cu_socket->so_rcv.sb_upcallarg;
819 	mtx_lock(&cs->cs_lock);
820 
821 	switch (request) {
822 	case CLSET_FD_CLOSE:
823 		cu->cu_closeit = TRUE;
824 		mtx_unlock(&cs->cs_lock);
825 		return (TRUE);
826 	case CLSET_FD_NCLOSE:
827 		cu->cu_closeit = FALSE;
828 		mtx_unlock(&cs->cs_lock);
829 		return (TRUE);
830 	}
831 
832 	/* for other requests which use info */
833 	if (info == NULL) {
834 		mtx_unlock(&cs->cs_lock);
835 		return (FALSE);
836 	}
837 	switch (request) {
838 	case CLSET_TIMEOUT:
839 		if (time_not_ok((struct timeval *)info)) {
840 			mtx_unlock(&cs->cs_lock);
841 			return (FALSE);
842 		}
843 		cu->cu_total = *(struct timeval *)info;
844 		break;
845 	case CLGET_TIMEOUT:
846 		*(struct timeval *)info = cu->cu_total;
847 		break;
848 	case CLSET_RETRY_TIMEOUT:
849 		if (time_not_ok((struct timeval *)info)) {
850 			mtx_unlock(&cs->cs_lock);
851 			return (FALSE);
852 		}
853 		cu->cu_wait = *(struct timeval *)info;
854 		break;
855 	case CLGET_RETRY_TIMEOUT:
856 		*(struct timeval *)info = cu->cu_wait;
857 		break;
858 	case CLGET_SVC_ADDR:
859 		/*
860 		 * Slightly different semantics to userland - we use
861 		 * sockaddr instead of netbuf.
862 		 */
863 		memcpy(info, &cu->cu_raddr, cu->cu_raddr.ss_len);
864 		break;
865 	case CLSET_SVC_ADDR:		/* set to new address */
866 		addr = (struct sockaddr *)info;
867 		(void) memcpy(&cu->cu_raddr, addr, addr->sa_len);
868 		break;
869 	case CLGET_XID:
870 		*(uint32_t *)info = atomic_load_32(&rpc_xid);
871 		break;
872 
873 	case CLSET_XID:
874 		/* This will set the xid of the NEXT call */
875 		/* decrement by 1 as clnt_dg_call() increments once */
876 		atomic_store_32(&rpc_xid, *(uint32_t *)info - 1);
877 		break;
878 
879 	case CLGET_VERS:
880 		/*
881 		 * This RELIES on the information that, in the call body,
882 		 * the version number field is the fifth field from the
883 		 * beginning of the RPC header. MUST be changed if the
884 		 * call_struct is changed
885 		 */
886 		*(uint32_t *)info =
887 		    ntohl(*(uint32_t *)(void *)(cu->cu_mcallc +
888 		    4 * BYTES_PER_XDR_UNIT));
889 		break;
890 
891 	case CLSET_VERS:
892 		*(uint32_t *)(void *)(cu->cu_mcallc + 4 * BYTES_PER_XDR_UNIT)
893 			= htonl(*(uint32_t *)info);
894 		break;
895 
896 	case CLGET_PROG:
897 		/*
898 		 * This RELIES on the information that, in the call body,
899 		 * the program number field is the fourth field from the
900 		 * beginning of the RPC header. MUST be changed if the
901 		 * call_struct is changed
902 		 */
903 		*(uint32_t *)info =
904 		    ntohl(*(uint32_t *)(void *)(cu->cu_mcallc +
905 		    3 * BYTES_PER_XDR_UNIT));
906 		break;
907 
908 	case CLSET_PROG:
909 		*(uint32_t *)(void *)(cu->cu_mcallc + 3 * BYTES_PER_XDR_UNIT)
910 			= htonl(*(uint32_t *)info);
911 		break;
912 	case CLSET_ASYNC:
913 		cu->cu_async = *(int *)info;
914 		break;
915 	case CLSET_CONNECT:
916 		cu->cu_connect = *(int *)info;
917 		break;
918 	case CLSET_WAITCHAN:
919 		cu->cu_waitchan = (const char *)info;
920 		break;
921 	case CLGET_WAITCHAN:
922 		*(const char **) info = cu->cu_waitchan;
923 		break;
924 	case CLSET_INTERRUPTIBLE:
925 		if (*(int *) info)
926 			cu->cu_waitflag = PCATCH;
927 		else
928 			cu->cu_waitflag = 0;
929 		break;
930 	case CLGET_INTERRUPTIBLE:
931 		if (cu->cu_waitflag)
932 			*(int *) info = TRUE;
933 		else
934 			*(int *) info = FALSE;
935 		break;
936 	default:
937 		mtx_unlock(&cs->cs_lock);
938 		return (FALSE);
939 	}
940 	mtx_unlock(&cs->cs_lock);
941 	return (TRUE);
942 }
943 
944 static void
945 clnt_dg_close(CLIENT *cl)
946 {
947 	struct cu_data *cu = (struct cu_data *)cl->cl_private;
948 	struct cu_socket *cs;
949 	struct cu_request *cr;
950 
951 	cs = cu->cu_socket->so_rcv.sb_upcallarg;
952 	mtx_lock(&cs->cs_lock);
953 
954 	if (cu->cu_closed) {
955 		mtx_unlock(&cs->cs_lock);
956 		return;
957 	}
958 
959 	if (cu->cu_closing) {
960 		while (cu->cu_closing)
961 			msleep(cu, &cs->cs_lock, 0, "rpcclose", 0);
962 		KASSERT(cu->cu_closed, ("client should be closed"));
963 		mtx_unlock(&cs->cs_lock);
964 		return;
965 	}
966 
967 	/*
968 	 * Abort any pending requests and wait until everyone
969 	 * has finished with clnt_vc_call.
970 	 */
971 	cu->cu_closing = TRUE;
972 	TAILQ_FOREACH(cr, &cs->cs_pending, cr_link) {
973 		if (cr->cr_client == cl) {
974 			cr->cr_xid = 0;
975 			cr->cr_error = ESHUTDOWN;
976 			wakeup(cr);
977 		}
978 	}
979 
980 	while (cu->cu_threads)
981 		msleep(cu, &cs->cs_lock, 0, "rpcclose", 0);
982 
983 	cu->cu_closing = FALSE;
984 	cu->cu_closed = TRUE;
985 
986 	mtx_unlock(&cs->cs_lock);
987 	wakeup(cu);
988 }
989 
990 static void
991 clnt_dg_destroy(CLIENT *cl)
992 {
993 	struct cu_data *cu = (struct cu_data *)cl->cl_private;
994 	struct cu_socket *cs;
995 	struct socket *so = NULL;
996 	bool_t lastsocketref;
997 
998 	cs = cu->cu_socket->so_rcv.sb_upcallarg;
999 	clnt_dg_close(cl);
1000 
1001 	SOCKBUF_LOCK(&cu->cu_socket->so_rcv);
1002 	mtx_lock(&cs->cs_lock);
1003 
1004 	cs->cs_refs--;
1005 	if (cs->cs_refs == 0) {
1006 		mtx_unlock(&cs->cs_lock);
1007 		soupcall_clear(cu->cu_socket, SO_RCV);
1008 		clnt_dg_upcallsdone(cu->cu_socket, cs);
1009 		SOCKBUF_UNLOCK(&cu->cu_socket->so_rcv);
1010 		mtx_destroy(&cs->cs_lock);
1011 		mem_free(cs, sizeof(*cs));
1012 		lastsocketref = TRUE;
1013 	} else {
1014 		mtx_unlock(&cs->cs_lock);
1015 		SOCKBUF_UNLOCK(&cu->cu_socket->so_rcv);
1016 		lastsocketref = FALSE;
1017 	}
1018 
1019 	if (cu->cu_closeit && lastsocketref) {
1020 		so = cu->cu_socket;
1021 		cu->cu_socket = NULL;
1022 	}
1023 
1024 	if (so)
1025 		soclose(so);
1026 
1027 	if (cl->cl_netid && cl->cl_netid[0])
1028 		mem_free(cl->cl_netid, strlen(cl->cl_netid) +1);
1029 	if (cl->cl_tp && cl->cl_tp[0])
1030 		mem_free(cl->cl_tp, strlen(cl->cl_tp) +1);
1031 	mem_free(cu, sizeof (*cu));
1032 	mem_free(cl, sizeof (CLIENT));
1033 }
1034 
1035 /*
1036  * Make sure that the time is not garbage.  -1 value is allowed.
1037  */
1038 static bool_t
1039 time_not_ok(struct timeval *t)
1040 {
1041 	return (t->tv_sec < -1 || t->tv_sec > 100000000 ||
1042 		t->tv_usec < -1 || t->tv_usec > 1000000);
1043 }
1044 
1045 int
1046 clnt_dg_soupcall(struct socket *so, void *arg, int waitflag)
1047 {
1048 	struct cu_socket *cs = (struct cu_socket *) arg;
1049 	struct uio uio;
1050 	struct mbuf *m;
1051 	struct mbuf *control;
1052 	struct cu_request *cr;
1053 	int error, rcvflag, foundreq;
1054 	uint32_t xid;
1055 
1056 	cs->cs_upcallrefs++;
1057 	uio.uio_resid = 1000000000;
1058 	uio.uio_td = curthread;
1059 	do {
1060 		SOCKBUF_UNLOCK(&so->so_rcv);
1061 		m = NULL;
1062 		control = NULL;
1063 		rcvflag = MSG_DONTWAIT;
1064 		error = soreceive(so, NULL, &uio, &m, &control, &rcvflag);
1065 		if (control)
1066 			m_freem(control);
1067 		SOCKBUF_LOCK(&so->so_rcv);
1068 
1069 		if (error == EWOULDBLOCK)
1070 			break;
1071 
1072 		/*
1073 		 * If there was an error, wake up all pending
1074 		 * requests.
1075 		 */
1076 		if (error) {
1077 			mtx_lock(&cs->cs_lock);
1078 			TAILQ_FOREACH(cr, &cs->cs_pending, cr_link) {
1079 				cr->cr_xid = 0;
1080 				cr->cr_error = error;
1081 				wakeup(cr);
1082 			}
1083 			mtx_unlock(&cs->cs_lock);
1084 			break;
1085 		}
1086 
1087 		/*
1088 		 * The XID is in the first uint32_t of the reply.
1089 		 */
1090 		if (m->m_len < sizeof(xid) && m_length(m, NULL) < sizeof(xid)) {
1091 			/*
1092 			 * Should never happen.
1093 			 */
1094 			m_freem(m);
1095 			continue;
1096 		}
1097 
1098 		m_copydata(m, 0, sizeof(xid), (char *)&xid);
1099 		xid = ntohl(xid);
1100 
1101 		/*
1102 		 * Attempt to match this reply with a pending request.
1103 		 */
1104 		mtx_lock(&cs->cs_lock);
1105 		foundreq = 0;
1106 		TAILQ_FOREACH(cr, &cs->cs_pending, cr_link) {
1107 			if (cr->cr_xid == xid) {
1108 				/*
1109 				 * This one matches. We leave the
1110 				 * reply mbuf in cr->cr_mrep. Set the
1111 				 * XID to zero so that we will ignore
1112 				 * any duplicated replies that arrive
1113 				 * before clnt_dg_call removes it from
1114 				 * the queue.
1115 				 */
1116 				cr->cr_xid = 0;
1117 				cr->cr_mrep = m;
1118 				cr->cr_error = 0;
1119 				foundreq = 1;
1120 				wakeup(cr);
1121 				break;
1122 			}
1123 		}
1124 		mtx_unlock(&cs->cs_lock);
1125 
1126 		/*
1127 		 * If we didn't find the matching request, just drop
1128 		 * it - its probably a repeated reply.
1129 		 */
1130 		if (!foundreq)
1131 			m_freem(m);
1132 	} while (m);
1133 	cs->cs_upcallrefs--;
1134 	if (cs->cs_upcallrefs < 0)
1135 		panic("rpcdg upcall refcnt");
1136 	if (cs->cs_upcallrefs == 0)
1137 		wakeup(&cs->cs_upcallrefs);
1138 	return (SU_OK);
1139 }
1140 
1141 /*
1142  * Wait for all upcalls in progress to complete.
1143  */
1144 static void
1145 clnt_dg_upcallsdone(struct socket *so, struct cu_socket *cs)
1146 {
1147 
1148 	SOCKBUF_LOCK_ASSERT(&so->so_rcv);
1149 
1150 	while (cs->cs_upcallrefs > 0)
1151 		(void) msleep(&cs->cs_upcallrefs, SOCKBUF_MTX(&so->so_rcv), 0,
1152 		    "rpcdgup", 0);
1153 }
1154