1/*	$NetBSD: clnt_dg.c,v 1.4 2000/07/14 08:40:41 fvdl Exp $	*/
2
3/*-
4 * Copyright (c) 2009, Sun Microsystems, Inc.
5 * All rights reserved.
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions are met:
9 * - Redistributions of source code must retain the above copyright notice,
10 *   this list of conditions and the following disclaimer.
11 * - Redistributions in binary form must reproduce the above copyright notice,
12 *   this list of conditions and the following disclaimer in the documentation
13 *   and/or other materials provided with the distribution.
14 * - Neither the name of Sun Microsystems, Inc. nor the names of its
15 *   contributors may be used to endorse or promote products derived
16 *   from this software without specific prior written permission.
17 *
18 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
19 * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
20 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
21 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
22 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
23 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
24 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
25 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
26 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
27 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
28 * POSSIBILITY OF SUCH DAMAGE.
29 */
30/*
31 * Copyright (c) 1986-1991 by Sun Microsystems Inc.
32 */
33
34#if defined(LIBC_SCCS) && !defined(lint)
35#ident	"@(#)clnt_dg.c	1.23	94/04/22 SMI"
36static char sccsid[] = "@(#)clnt_dg.c 1.19 89/03/16 Copyr 1988 Sun Micro";
37#endif
38#include <sys/cdefs.h>
39__FBSDID("$FreeBSD$");
40
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
74static bool_t time_not_ok(struct timeval *);
75static enum clnt_stat clnt_dg_call(CLIENT *, struct rpc_callextra *,
76    rpcproc_t, struct mbuf *, struct mbuf **, struct timeval);
77static void clnt_dg_geterr(CLIENT *, struct rpc_err *);
78static bool_t clnt_dg_freeres(CLIENT *, xdrproc_t, void *);
79static void clnt_dg_abort(CLIENT *);
80static bool_t clnt_dg_control(CLIENT *, u_int, void *);
81static void clnt_dg_close(CLIENT *);
82static void clnt_dg_destroy(CLIENT *);
83static int clnt_dg_soupcall(struct socket *so, void *arg, int waitflag);
84
85static 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/*
96 * A pending RPC request which awaits a reply. Requests which have
97 * received their reply will have cr_xid set to zero and cr_mrep to
98 * the mbuf chain of the reply.
99 */
100struct cu_request {
101	TAILQ_ENTRY(cu_request) cr_link;
102	CLIENT			*cr_client;	/* owner */
103	uint32_t		cr_xid;		/* XID of request */
104	struct mbuf		*cr_mrep;	/* reply received by upcall */
105	int			cr_error;	/* any error from upcall */
106	char			cr_verf[MAX_AUTH_BYTES]; /* reply verf */
107};
108
109TAILQ_HEAD(cu_request_list, cu_request);
110
111#define MCALL_MSG_SIZE 24
112
113/*
114 * This structure is pointed to by the socket buffer's sb_upcallarg
115 * member. It is separate from the client private data to facilitate
116 * multiple clients sharing the same socket. The cs_lock mutex is used
117 * to protect all fields of this structure, the socket's receive
118 * buffer SOCKBUF_LOCK is used to ensure that exactly one of these
119 * structures is installed on the socket.
120 */
121struct cu_socket {
122	struct mtx		cs_lock;
123	int			cs_refs;	/* Count of clients */
124	struct cu_request_list	cs_pending;	/* Requests awaiting replies */
125	int			cs_upcallrefs;	/* Refcnt of upcalls in prog.*/
126};
127
128static void clnt_dg_upcallsdone(struct socket *, struct cu_socket *);
129
130/*
131 * Private data kept per client handle
132 */
133struct cu_data {
134	int			cu_threads;	/* # threads in clnt_vc_call */
135	bool_t			cu_closing;	/* TRUE if we are closing */
136	bool_t			cu_closed;	/* TRUE if we are closed */
137	struct socket		*cu_socket;	/* connection socket */
138	bool_t			cu_closeit;	/* opened by library */
139	struct sockaddr_storage	cu_raddr;	/* remote address */
140	int			cu_rlen;
141	struct timeval		cu_wait;	/* retransmit interval */
142	struct timeval		cu_total;	/* total time for the call */
143	struct rpc_err		cu_error;
144	uint32_t		cu_xid;
145	char			cu_mcallc[MCALL_MSG_SIZE]; /* marshalled callmsg */
146	size_t			cu_mcalllen;
147	size_t			cu_sendsz;	/* send size */
148	size_t			cu_recvsz;	/* recv size */
149	int			cu_async;
150	int			cu_connect;	/* Use connect(). */
151	int			cu_connected;	/* Have done connect(). */
152	const char		*cu_waitchan;
153	int			cu_waitflag;
154	int			cu_cwnd;	/* congestion window */
155	int			cu_sent;	/* number of in-flight RPCs */
156	bool_t			cu_cwnd_wait;
157};
158
159#define CWNDSCALE	256
160#define MAXCWND		(32 * CWNDSCALE)
161
162/*
163 * Connection less client creation returns with client handle parameters.
164 * Default options are set, which the user can change using clnt_control().
165 * fd should be open and bound.
166 * NB: The rpch->cl_auth is initialized to null authentication.
167 * 	Caller may wish to set this something more useful.
168 *
169 * sendsz and recvsz are the maximum allowable packet sizes that can be
170 * sent and received. Normally they are the same, but they can be
171 * changed to improve the program efficiency and buffer allocation.
172 * If they are 0, use the transport default.
173 *
174 * If svcaddr is NULL, returns NULL.
175 */
176CLIENT *
177clnt_dg_create(
178	struct socket *so,
179	struct sockaddr *svcaddr,	/* servers address */
180	rpcprog_t program,		/* program number */
181	rpcvers_t version,		/* version number */
182	size_t sendsz,			/* buffer recv size */
183	size_t recvsz)			/* buffer send size */
184{
185	CLIENT *cl = NULL;		/* client handle */
186	struct cu_data *cu = NULL;	/* private data */
187	struct cu_socket *cs = NULL;
188	struct sockbuf *sb;
189	struct timeval now;
190	struct rpc_msg call_msg;
191	struct __rpc_sockinfo si;
192	XDR xdrs;
193	int error;
194
195	if (svcaddr == NULL) {
196		rpc_createerr.cf_stat = RPC_UNKNOWNADDR;
197		return (NULL);
198	}
199
200	if (!__rpc_socket2sockinfo(so, &si)) {
201		rpc_createerr.cf_stat = RPC_TLIERROR;
202		rpc_createerr.cf_error.re_errno = 0;
203		return (NULL);
204	}
205
206	/*
207	 * Find the receive and the send size
208	 */
209	sendsz = __rpc_get_t_size(si.si_af, si.si_proto, (int)sendsz);
210	recvsz = __rpc_get_t_size(si.si_af, si.si_proto, (int)recvsz);
211	if ((sendsz == 0) || (recvsz == 0)) {
212		rpc_createerr.cf_stat = RPC_TLIERROR; /* XXX */
213		rpc_createerr.cf_error.re_errno = 0;
214		return (NULL);
215	}
216
217	cl = mem_alloc(sizeof (CLIENT));
218
219	/*
220	 * Should be multiple of 4 for XDR.
221	 */
222	sendsz = ((sendsz + 3) / 4) * 4;
223	recvsz = ((recvsz + 3) / 4) * 4;
224	cu = mem_alloc(sizeof (*cu));
225	cu->cu_threads = 0;
226	cu->cu_closing = FALSE;
227	cu->cu_closed = FALSE;
228	(void) memcpy(&cu->cu_raddr, svcaddr, (size_t)svcaddr->sa_len);
229	cu->cu_rlen = svcaddr->sa_len;
230	/* Other values can also be set through clnt_control() */
231	cu->cu_wait.tv_sec = 3;	/* heuristically chosen */
232	cu->cu_wait.tv_usec = 0;
233	cu->cu_total.tv_sec = -1;
234	cu->cu_total.tv_usec = -1;
235	cu->cu_sendsz = sendsz;
236	cu->cu_recvsz = recvsz;
237	cu->cu_async = FALSE;
238	cu->cu_connect = FALSE;
239	cu->cu_connected = FALSE;
240	cu->cu_waitchan = "rpcrecv";
241	cu->cu_waitflag = 0;
242	cu->cu_cwnd = MAXCWND / 2;
243	cu->cu_sent = 0;
244	cu->cu_cwnd_wait = FALSE;
245	(void) getmicrotime(&now);
246	cu->cu_xid = __RPC_GETXID(&now);
247	call_msg.rm_xid = cu->cu_xid;
248	call_msg.rm_call.cb_prog = program;
249	call_msg.rm_call.cb_vers = version;
250	xdrmem_create(&xdrs, cu->cu_mcallc, MCALL_MSG_SIZE, XDR_ENCODE);
251	if (! xdr_callhdr(&xdrs, &call_msg)) {
252		rpc_createerr.cf_stat = RPC_CANTENCODEARGS;  /* XXX */
253		rpc_createerr.cf_error.re_errno = 0;
254		goto err2;
255	}
256	cu->cu_mcalllen = XDR_GETPOS(&xdrs);
257
258	/*
259	 * By default, closeit is always FALSE. It is users responsibility
260	 * to do a close on it, else the user may use clnt_control
261	 * to let clnt_destroy do it for him/her.
262	 */
263	cu->cu_closeit = FALSE;
264	cu->cu_socket = so;
265	error = soreserve(so, (u_long)sendsz, (u_long)recvsz);
266	if (error != 0) {
267		rpc_createerr.cf_stat = RPC_FAILED;
268		rpc_createerr.cf_error.re_errno = error;
269		goto err2;
270	}
271
272	sb = &so->so_rcv;
273	SOCKBUF_LOCK(&so->so_rcv);
274recheck_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);
315err2:
316	mem_free(cl, sizeof (CLIENT));
317	mem_free(cu, sizeof (*cu));
318
319	return (NULL);
320}
321
322static enum clnt_stat
323clnt_dg_call(
324	CLIENT		*cl,		/* client handle */
325	struct rpc_callextra *ext,	/* call metadata */
326	rpcproc_t	proc,		/* procedure number */
327	struct mbuf	*args,		/* pointer to args */
328	struct mbuf	**resultsp,	/* pointer to results */
329	struct timeval	utimeout)	/* seconds to wait before giving up */
330{
331	struct cu_data *cu = (struct cu_data *)cl->cl_private;
332	struct cu_socket *cs;
333	struct rpc_timers *rt;
334	AUTH *auth;
335	struct rpc_err *errp;
336	enum clnt_stat stat;
337	XDR xdrs;
338	struct rpc_msg reply_msg;
339	bool_t ok;
340	int retrans;			/* number of re-transmits so far */
341	int nrefreshes = 2;		/* number of times to refresh cred */
342	struct timeval *tvp;
343	int timeout;
344	int retransmit_time;
345	int next_sendtime, starttime, rtt, time_waited, tv = 0;
346	struct sockaddr *sa;
347	socklen_t salen;
348	uint32_t xid = 0;
349	struct mbuf *mreq = NULL, *results;
350	struct cu_request *cr;
351	int error;
352
353	cs = cu->cu_socket->so_rcv.sb_upcallarg;
354	cr = malloc(sizeof(struct cu_request), M_RPC, M_WAITOK);
355
356	mtx_lock(&cs->cs_lock);
357
358	if (cu->cu_closing || cu->cu_closed) {
359		mtx_unlock(&cs->cs_lock);
360		free(cr, M_RPC);
361		return (RPC_CANTSEND);
362	}
363	cu->cu_threads++;
364
365	if (ext) {
366		auth = ext->rc_auth;
367		errp = &ext->rc_err;
368	} else {
369		auth = cl->cl_auth;
370		errp = &cu->cu_error;
371	}
372
373	cr->cr_client = cl;
374	cr->cr_mrep = NULL;
375	cr->cr_error = 0;
376
377	if (cu->cu_total.tv_usec == -1) {
378		tvp = &utimeout; /* use supplied timeout */
379	} else {
380		tvp = &cu->cu_total; /* use default timeout */
381	}
382	if (tvp->tv_sec || tvp->tv_usec)
383		timeout = tvtohz(tvp);
384	else
385		timeout = 0;
386
387	if (cu->cu_connect && !cu->cu_connected) {
388		mtx_unlock(&cs->cs_lock);
389		error = soconnect(cu->cu_socket,
390		    (struct sockaddr *)&cu->cu_raddr, curthread);
391		mtx_lock(&cs->cs_lock);
392		if (error) {
393			errp->re_errno = error;
394			errp->re_status = stat = RPC_CANTSEND;
395			goto out;
396		}
397		cu->cu_connected = 1;
398	}
399	if (cu->cu_connected) {
400		sa = NULL;
401		salen = 0;
402	} else {
403		sa = (struct sockaddr *)&cu->cu_raddr;
404		salen = cu->cu_rlen;
405	}
406	time_waited = 0;
407	retrans = 0;
408	if (ext && ext->rc_timers) {
409		rt = ext->rc_timers;
410		if (!rt->rt_rtxcur)
411			rt->rt_rtxcur = tvtohz(&cu->cu_wait);
412		retransmit_time = next_sendtime = rt->rt_rtxcur;
413	} else {
414		rt = NULL;
415		retransmit_time = next_sendtime = tvtohz(&cu->cu_wait);
416	}
417
418	starttime = ticks;
419
420call_again:
421	mtx_assert(&cs->cs_lock, MA_OWNED);
422
423	cu->cu_xid++;
424	xid = cu->cu_xid;
425
426send_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
546get_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
684got_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 unsuccesful 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);
763out:
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
785static void
786clnt_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
793static bool_t
794clnt_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*/
806static void
807clnt_dg_abort(CLIENT *h)
808{
809}
810
811static bool_t
812clnt_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 = cu->cu_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		cu->cu_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		 * begining 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		 * begining 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
944static void
945clnt_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
990static void
991clnt_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 */
1038static bool_t
1039time_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
1045int
1046clnt_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 */
1144static void
1145clnt_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