clnt_dg.c revision 193272
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"
37static char sccsid[] = "@(#)clnt_dg.c 1.19 89/03/16 Copyr 1988 Sun Micro";
38#endif
39#include <sys/cdefs.h>
40__FBSDID("$FreeBSD: head/sys/rpc/clnt_dg.c 193272 2009-06-01 21:17:03Z jhb $");
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 <rpc/rpc.h>
61#include <rpc/rpc_com.h>
62
63
64#ifdef _FREEFALL_CONFIG
65/*
66 * Disable RPC exponential back-off for FreeBSD.org systems.
67 */
68#define	RPC_MAX_BACKOFF		1 /* second */
69#else
70#define	RPC_MAX_BACKOFF		30 /* seconds */
71#endif
72
73static bool_t time_not_ok(struct timeval *);
74static enum clnt_stat clnt_dg_call(CLIENT *, struct rpc_callextra *,
75    rpcproc_t, struct mbuf *, struct mbuf **, struct timeval);
76static void clnt_dg_geterr(CLIENT *, struct rpc_err *);
77static bool_t clnt_dg_freeres(CLIENT *, xdrproc_t, void *);
78static void clnt_dg_abort(CLIENT *);
79static bool_t clnt_dg_control(CLIENT *, u_int, void *);
80static void clnt_dg_close(CLIENT *);
81static void clnt_dg_destroy(CLIENT *);
82static int clnt_dg_soupcall(struct socket *so, void *arg, int waitflag);
83
84static struct clnt_ops clnt_dg_ops = {
85	.cl_call =	clnt_dg_call,
86	.cl_abort =	clnt_dg_abort,
87	.cl_geterr =	clnt_dg_geterr,
88	.cl_freeres =	clnt_dg_freeres,
89	.cl_close =	clnt_dg_close,
90	.cl_destroy =	clnt_dg_destroy,
91	.cl_control =	clnt_dg_control
92};
93
94static const char mem_err_clnt_dg[] = "clnt_dg_create: out of memory";
95
96/*
97 * A pending RPC request which awaits a reply. Requests which have
98 * received their reply will have cr_xid set to zero and cr_mrep to
99 * the mbuf chain of the reply.
100 */
101struct cu_request {
102	TAILQ_ENTRY(cu_request) cr_link;
103	CLIENT			*cr_client;	/* owner */
104	uint32_t		cr_xid;		/* XID of request */
105	struct mbuf		*cr_mrep;	/* reply received by upcall */
106	int			cr_error;	/* any error from upcall */
107	char			cr_verf[MAX_AUTH_BYTES]; /* reply verf */
108};
109
110TAILQ_HEAD(cu_request_list, cu_request);
111
112#define MCALL_MSG_SIZE 24
113
114/*
115 * This structure is pointed to by the socket buffer's sb_upcallarg
116 * member. It is separate from the client private data to facilitate
117 * multiple clients sharing the same socket. The cs_lock mutex is used
118 * to protect all fields of this structure, the socket's receive
119 * buffer SOCKBUF_LOCK is used to ensure that exactly one of these
120 * structures is installed on the socket.
121 */
122struct cu_socket {
123	struct mtx		cs_lock;
124	int			cs_refs;	/* Count of clients */
125	struct cu_request_list	cs_pending;	/* Requests awaiting replies */
126};
127
128/*
129 * Private data kept per client handle
130 */
131struct cu_data {
132	int			cu_threads;	/* # threads in clnt_vc_call */
133	bool_t			cu_closing;	/* TRUE if we are closing */
134	bool_t			cu_closed;	/* TRUE if we are closed */
135	struct socket		*cu_socket;	/* connection socket */
136	bool_t			cu_closeit;	/* opened by library */
137	struct sockaddr_storage	cu_raddr;	/* remote address */
138	int			cu_rlen;
139	struct timeval		cu_wait;	/* retransmit interval */
140	struct timeval		cu_total;	/* total time for the call */
141	struct rpc_err		cu_error;
142	uint32_t		cu_xid;
143	char			cu_mcallc[MCALL_MSG_SIZE]; /* marshalled callmsg */
144	size_t			cu_mcalllen;
145	size_t			cu_sendsz;	/* send size */
146	size_t			cu_recvsz;	/* recv size */
147	int			cu_async;
148	int			cu_connect;	/* Use connect(). */
149	int			cu_connected;	/* Have done connect(). */
150	const char		*cu_waitchan;
151	int			cu_waitflag;
152	int			cu_cwnd;	/* congestion window */
153	int			cu_sent;	/* number of in-flight RPCs */
154	bool_t			cu_cwnd_wait;
155};
156
157#define CWNDSCALE	256
158#define MAXCWND		(32 * CWNDSCALE)
159
160/*
161 * Connection less client creation returns with client handle parameters.
162 * Default options are set, which the user can change using clnt_control().
163 * fd should be open and bound.
164 * NB: The rpch->cl_auth is initialized to null authentication.
165 * 	Caller may wish to set this something more useful.
166 *
167 * sendsz and recvsz are the maximum allowable packet sizes that can be
168 * sent and received. Normally they are the same, but they can be
169 * changed to improve the program efficiency and buffer allocation.
170 * If they are 0, use the transport default.
171 *
172 * If svcaddr is NULL, returns NULL.
173 */
174CLIENT *
175clnt_dg_create(
176	struct socket *so,
177	struct sockaddr *svcaddr,	/* servers address */
178	rpcprog_t program,		/* program number */
179	rpcvers_t version,		/* version number */
180	size_t sendsz,			/* buffer recv size */
181	size_t recvsz)			/* buffer send size */
182{
183	CLIENT *cl = NULL;		/* client handle */
184	struct cu_data *cu = NULL;	/* private data */
185	struct cu_socket *cs = NULL;
186	struct sockbuf *sb;
187	struct timeval now;
188	struct rpc_msg call_msg;
189	struct __rpc_sockinfo si;
190	XDR xdrs;
191
192	if (svcaddr == NULL) {
193		rpc_createerr.cf_stat = RPC_UNKNOWNADDR;
194		return (NULL);
195	}
196
197	if (!__rpc_socket2sockinfo(so, &si)) {
198		rpc_createerr.cf_stat = RPC_TLIERROR;
199		rpc_createerr.cf_error.re_errno = 0;
200		return (NULL);
201	}
202
203	/*
204	 * Find the receive and the send size
205	 */
206	sendsz = __rpc_get_t_size(si.si_af, si.si_proto, (int)sendsz);
207	recvsz = __rpc_get_t_size(si.si_af, si.si_proto, (int)recvsz);
208	if ((sendsz == 0) || (recvsz == 0)) {
209		rpc_createerr.cf_stat = RPC_TLIERROR; /* XXX */
210		rpc_createerr.cf_error.re_errno = 0;
211		return (NULL);
212	}
213
214	cl = mem_alloc(sizeof (CLIENT));
215
216	/*
217	 * Should be multiple of 4 for XDR.
218	 */
219	sendsz = ((sendsz + 3) / 4) * 4;
220	recvsz = ((recvsz + 3) / 4) * 4;
221	cu = mem_alloc(sizeof (*cu));
222	cu->cu_threads = 0;
223	cu->cu_closing = FALSE;
224	cu->cu_closed = FALSE;
225	(void) memcpy(&cu->cu_raddr, svcaddr, (size_t)svcaddr->sa_len);
226	cu->cu_rlen = svcaddr->sa_len;
227	/* Other values can also be set through clnt_control() */
228	cu->cu_wait.tv_sec = 3;	/* heuristically chosen */
229	cu->cu_wait.tv_usec = 0;
230	cu->cu_total.tv_sec = -1;
231	cu->cu_total.tv_usec = -1;
232	cu->cu_sendsz = sendsz;
233	cu->cu_recvsz = recvsz;
234	cu->cu_async = FALSE;
235	cu->cu_connect = FALSE;
236	cu->cu_connected = FALSE;
237	cu->cu_waitchan = "rpcrecv";
238	cu->cu_waitflag = 0;
239	cu->cu_cwnd = MAXCWND / 2;
240	cu->cu_sent = 0;
241	cu->cu_cwnd_wait = FALSE;
242	(void) getmicrotime(&now);
243	cu->cu_xid = __RPC_GETXID(&now);
244	call_msg.rm_xid = cu->cu_xid;
245	call_msg.rm_call.cb_prog = program;
246	call_msg.rm_call.cb_vers = version;
247	xdrmem_create(&xdrs, cu->cu_mcallc, MCALL_MSG_SIZE, XDR_ENCODE);
248	if (! xdr_callhdr(&xdrs, &call_msg)) {
249		rpc_createerr.cf_stat = RPC_CANTENCODEARGS;  /* XXX */
250		rpc_createerr.cf_error.re_errno = 0;
251		goto err2;
252	}
253	cu->cu_mcalllen = XDR_GETPOS(&xdrs);;
254
255	/*
256	 * By default, closeit is always FALSE. It is users responsibility
257	 * to do a close on it, else the user may use clnt_control
258	 * to let clnt_destroy do it for him/her.
259	 */
260	cu->cu_closeit = FALSE;
261	cu->cu_socket = so;
262	soreserve(so, 256*1024, 256*1024);
263
264	sb = &so->so_rcv;
265	SOCKBUF_LOCK(&so->so_rcv);
266recheck_socket:
267	if (sb->sb_upcall) {
268		if (sb->sb_upcall != clnt_dg_soupcall) {
269			SOCKBUF_UNLOCK(&so->so_rcv);
270			printf("clnt_dg_create(): socket already has an incompatible upcall\n");
271			goto err2;
272		}
273		cs = (struct cu_socket *) sb->sb_upcallarg;
274		mtx_lock(&cs->cs_lock);
275		cs->cs_refs++;
276		mtx_unlock(&cs->cs_lock);
277	} else {
278		/*
279		 * We are the first on this socket - allocate the
280		 * structure and install it in the socket.
281		 */
282		SOCKBUF_UNLOCK(&so->so_rcv);
283		cs = mem_alloc(sizeof(*cs));
284		SOCKBUF_LOCK(&so->so_rcv);
285		if (sb->sb_upcall) {
286			/*
287			 * We have lost a race with some other client.
288			 */
289			mem_free(cs, sizeof(*cs));
290			goto recheck_socket;
291		}
292		mtx_init(&cs->cs_lock, "cs->cs_lock", NULL, MTX_DEF);
293		cs->cs_refs = 1;
294		TAILQ_INIT(&cs->cs_pending);
295		soupcall_set(so, SO_RCV, clnt_dg_soupcall, cs);
296	}
297	SOCKBUF_UNLOCK(&so->so_rcv);
298
299	cl->cl_refs = 1;
300	cl->cl_ops = &clnt_dg_ops;
301	cl->cl_private = (caddr_t)(void *)cu;
302	cl->cl_auth = authnone_create();
303	cl->cl_tp = NULL;
304	cl->cl_netid = NULL;
305	return (cl);
306err2:
307	if (cl) {
308		mem_free(cl, sizeof (CLIENT));
309		if (cu)
310			mem_free(cu, sizeof (*cu));
311	}
312	return (NULL);
313}
314
315static enum clnt_stat
316clnt_dg_call(
317	CLIENT		*cl,		/* client handle */
318	struct rpc_callextra *ext,	/* call metadata */
319	rpcproc_t	proc,		/* procedure number */
320	struct mbuf	*args,		/* pointer to args */
321	struct mbuf	**resultsp,	/* pointer to results */
322	struct timeval	utimeout)	/* seconds to wait before giving up */
323{
324	struct cu_data *cu = (struct cu_data *)cl->cl_private;
325	struct cu_socket *cs;
326	struct rpc_timers *rt;
327	AUTH *auth;
328	struct rpc_err *errp;
329	enum clnt_stat stat;
330	XDR xdrs;
331	struct rpc_msg reply_msg;
332	bool_t ok;
333	int retrans;			/* number of re-transmits so far */
334	int nrefreshes = 2;		/* number of times to refresh cred */
335	struct timeval *tvp;
336	int timeout;
337	int retransmit_time;
338	int next_sendtime, starttime, rtt, time_waited, tv = 0;
339	struct sockaddr *sa;
340	socklen_t salen;
341	uint32_t xid = 0;
342	struct mbuf *mreq = NULL, *results;
343	struct cu_request *cr;
344	int error;
345
346	cs = cu->cu_socket->so_rcv.sb_upcallarg;
347	cr = malloc(sizeof(struct cu_request), M_RPC, M_WAITOK);
348
349	mtx_lock(&cs->cs_lock);
350
351	if (cu->cu_closing || cu->cu_closed) {
352		mtx_unlock(&cs->cs_lock);
353		free(cr, M_RPC);
354		return (RPC_CANTSEND);
355	}
356	cu->cu_threads++;
357
358	if (ext) {
359		auth = ext->rc_auth;
360		errp = &ext->rc_err;
361	} else {
362		auth = cl->cl_auth;
363		errp = &cu->cu_error;
364	}
365
366	cr->cr_client = cl;
367	cr->cr_mrep = NULL;
368	cr->cr_error = 0;
369
370	if (cu->cu_total.tv_usec == -1) {
371		tvp = &utimeout; /* use supplied timeout */
372	} else {
373		tvp = &cu->cu_total; /* use default timeout */
374	}
375	if (tvp->tv_sec || tvp->tv_usec)
376		timeout = tvtohz(tvp);
377	else
378		timeout = 0;
379
380	if (cu->cu_connect && !cu->cu_connected) {
381		mtx_unlock(&cs->cs_lock);
382		error = soconnect(cu->cu_socket,
383		    (struct sockaddr *)&cu->cu_raddr, curthread);
384		mtx_lock(&cs->cs_lock);
385		if (error) {
386			errp->re_errno = error;
387			errp->re_status = stat = RPC_CANTSEND;
388			goto out;
389		}
390		cu->cu_connected = 1;
391	}
392	if (cu->cu_connected) {
393		sa = NULL;
394		salen = 0;
395	} else {
396		sa = (struct sockaddr *)&cu->cu_raddr;
397		salen = cu->cu_rlen;
398	}
399	time_waited = 0;
400	retrans = 0;
401	if (ext && ext->rc_timers) {
402		rt = ext->rc_timers;
403		if (!rt->rt_rtxcur)
404			rt->rt_rtxcur = tvtohz(&cu->cu_wait);
405		retransmit_time = next_sendtime = rt->rt_rtxcur;
406	} else {
407		rt = NULL;
408		retransmit_time = next_sendtime = tvtohz(&cu->cu_wait);
409	}
410
411	starttime = ticks;
412
413call_again:
414	mtx_assert(&cs->cs_lock, MA_OWNED);
415
416	cu->cu_xid++;
417	xid = cu->cu_xid;
418
419send_again:
420	mtx_unlock(&cs->cs_lock);
421
422	MGETHDR(mreq, M_WAIT, MT_DATA);
423	KASSERT(cu->cu_mcalllen <= MHLEN, ("RPC header too big"));
424	bcopy(cu->cu_mcallc, mreq->m_data, cu->cu_mcalllen);
425	mreq->m_len = cu->cu_mcalllen;
426
427	/*
428	 * The XID is the first thing in the request.
429	 */
430	*mtod(mreq, uint32_t *) = htonl(xid);
431
432	xdrmbuf_create(&xdrs, mreq, XDR_ENCODE);
433
434	if (cu->cu_async == TRUE && args == NULL)
435		goto get_reply;
436
437	if ((! XDR_PUTINT32(&xdrs, &proc)) ||
438	    (! AUTH_MARSHALL(auth, xid, &xdrs,
439		m_copym(args, 0, M_COPYALL, M_WAITOK)))) {
440		errp->re_status = stat = RPC_CANTENCODEARGS;
441		mtx_lock(&cs->cs_lock);
442		goto out;
443	}
444	mreq->m_pkthdr.len = m_length(mreq, NULL);
445
446	cr->cr_xid = xid;
447	mtx_lock(&cs->cs_lock);
448
449	/*
450	 * Try to get a place in the congestion window.
451	 */
452	while (cu->cu_sent >= cu->cu_cwnd) {
453		cu->cu_cwnd_wait = TRUE;
454		error = msleep(&cu->cu_cwnd_wait, &cs->cs_lock,
455		    cu->cu_waitflag, "rpccwnd", 0);
456		if (error) {
457			errp->re_errno = error;
458			errp->re_status = stat = RPC_CANTSEND;
459			goto out;
460		}
461	}
462	cu->cu_sent += CWNDSCALE;
463
464	TAILQ_INSERT_TAIL(&cs->cs_pending, cr, cr_link);
465	mtx_unlock(&cs->cs_lock);
466
467	/*
468	 * sosend consumes mreq.
469	 */
470	error = sosend(cu->cu_socket, sa, NULL, mreq, NULL, 0, curthread);
471	mreq = NULL;
472
473	/*
474	 * sub-optimal code appears here because we have
475	 * some clock time to spare while the packets are in flight.
476	 * (We assume that this is actually only executed once.)
477	 */
478	reply_msg.acpted_rply.ar_verf.oa_flavor = AUTH_NULL;
479	reply_msg.acpted_rply.ar_verf.oa_base = cr->cr_verf;
480	reply_msg.acpted_rply.ar_verf.oa_length = 0;
481	reply_msg.acpted_rply.ar_results.where = NULL;
482	reply_msg.acpted_rply.ar_results.proc = (xdrproc_t)xdr_void;
483
484	mtx_lock(&cs->cs_lock);
485	if (error) {
486		TAILQ_REMOVE(&cs->cs_pending, cr, cr_link);
487		errp->re_errno = error;
488		errp->re_status = stat = RPC_CANTSEND;
489		cu->cu_sent -= CWNDSCALE;
490		if (cu->cu_cwnd_wait) {
491			cu->cu_cwnd_wait = FALSE;
492			wakeup(&cu->cu_cwnd_wait);
493		}
494		goto out;
495	}
496
497	/*
498	 * Check to see if we got an upcall while waiting for the
499	 * lock.
500	 */
501	if (cr->cr_error) {
502		TAILQ_REMOVE(&cs->cs_pending, cr, cr_link);
503		errp->re_errno = cr->cr_error;
504		errp->re_status = stat = RPC_CANTRECV;
505		cu->cu_sent -= CWNDSCALE;
506		if (cu->cu_cwnd_wait) {
507			cu->cu_cwnd_wait = FALSE;
508			wakeup(&cu->cu_cwnd_wait);
509		}
510		goto out;
511	}
512	if (cr->cr_mrep) {
513		TAILQ_REMOVE(&cs->cs_pending, cr, cr_link);
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 got_reply;
520	}
521
522	/*
523	 * Hack to provide rpc-based message passing
524	 */
525	if (timeout == 0) {
526		TAILQ_REMOVE(&cs->cs_pending, cr, cr_link);
527		errp->re_status = stat = RPC_TIMEDOUT;
528		cu->cu_sent -= CWNDSCALE;
529		if (cu->cu_cwnd_wait) {
530			cu->cu_cwnd_wait = FALSE;
531			wakeup(&cu->cu_cwnd_wait);
532		}
533		goto out;
534	}
535
536get_reply:
537	for (;;) {
538		/* Decide how long to wait. */
539		if (next_sendtime < timeout)
540			tv = next_sendtime;
541		else
542			tv = timeout;
543		tv -= time_waited;
544
545		if (tv > 0) {
546			if (cu->cu_closing || cu->cu_closed)
547				error = 0;
548			else
549				error = msleep(cr, &cs->cs_lock,
550				    cu->cu_waitflag, cu->cu_waitchan, tv);
551		} else {
552			error = EWOULDBLOCK;
553		}
554
555		TAILQ_REMOVE(&cs->cs_pending, cr, cr_link);
556		cu->cu_sent -= CWNDSCALE;
557		if (cu->cu_cwnd_wait) {
558			cu->cu_cwnd_wait = FALSE;
559			wakeup(&cu->cu_cwnd_wait);
560		}
561
562		if (!error) {
563			/*
564			 * We were woken up by the upcall.  If the
565			 * upcall had a receive error, report that,
566			 * otherwise we have a reply.
567			 */
568			if (cr->cr_error) {
569				errp->re_errno = cr->cr_error;
570				errp->re_status = stat = RPC_CANTRECV;
571				goto out;
572			}
573
574			cu->cu_cwnd += (CWNDSCALE * CWNDSCALE
575			    + cu->cu_cwnd / 2) / cu->cu_cwnd;
576			if (cu->cu_cwnd > MAXCWND)
577				cu->cu_cwnd = MAXCWND;
578
579			if (rt) {
580				/*
581				 * Add one to the time since a tick
582				 * count of N means that the actual
583				 * time taken was somewhere between N
584				 * and N+1.
585				 */
586				rtt = ticks - starttime + 1;
587
588				/*
589				 * Update our estimate of the round
590				 * trip time using roughly the
591				 * algorithm described in RFC
592				 * 2988. Given an RTT sample R:
593				 *
594				 * RTTVAR = (1-beta) * RTTVAR + beta * |SRTT-R|
595				 * SRTT = (1-alpha) * SRTT + alpha * R
596				 *
597				 * where alpha = 0.125 and beta = 0.25.
598				 *
599				 * The initial retransmit timeout is
600				 * SRTT + 4*RTTVAR and doubles on each
601				 * retransmision.
602				 */
603				if (rt->rt_srtt == 0) {
604					rt->rt_srtt = rtt;
605					rt->rt_deviate = rtt / 2;
606				} else {
607					int32_t error = rtt - rt->rt_srtt;
608					rt->rt_srtt += error / 8;
609					error = abs(error) - rt->rt_deviate;
610					rt->rt_deviate += error / 4;
611				}
612				rt->rt_rtxcur = rt->rt_srtt + 4*rt->rt_deviate;
613			}
614
615			break;
616		}
617
618		/*
619		 * The sleep returned an error so our request is still
620		 * on the list. If we got EWOULDBLOCK, we may want to
621		 * re-send the request.
622		 */
623		if (error != EWOULDBLOCK) {
624			errp->re_errno = error;
625			if (error == EINTR)
626				errp->re_status = stat = RPC_INTR;
627			else
628				errp->re_status = stat = RPC_CANTRECV;
629			goto out;
630		}
631
632		time_waited = ticks - starttime;
633
634		/* Check for timeout. */
635		if (time_waited > timeout) {
636			errp->re_errno = EWOULDBLOCK;
637			errp->re_status = stat = RPC_TIMEDOUT;
638			goto out;
639		}
640
641		/* Retransmit if necessary. */
642		if (time_waited >= next_sendtime) {
643			cu->cu_cwnd /= 2;
644			if (cu->cu_cwnd < CWNDSCALE)
645				cu->cu_cwnd = CWNDSCALE;
646			if (ext && ext->rc_feedback) {
647				mtx_unlock(&cs->cs_lock);
648				if (retrans == 0)
649					ext->rc_feedback(FEEDBACK_REXMIT1,
650					    proc, ext->rc_feedback_arg);
651				else
652					ext->rc_feedback(FEEDBACK_REXMIT2,
653					    proc, ext->rc_feedback_arg);
654				mtx_lock(&cs->cs_lock);
655			}
656			if (cu->cu_closing || cu->cu_closed) {
657				errp->re_errno = ESHUTDOWN;
658				errp->re_status = stat = RPC_CANTRECV;
659				goto out;
660			}
661			retrans++;
662			/* update retransmit_time */
663			if (retransmit_time < RPC_MAX_BACKOFF * hz)
664				retransmit_time = 2 * retransmit_time;
665			next_sendtime += retransmit_time;
666			goto send_again;
667		}
668		TAILQ_INSERT_TAIL(&cs->cs_pending, cr, cr_link);
669	}
670
671got_reply:
672	/*
673	 * Now decode and validate the response. We need to drop the
674	 * lock since xdr_replymsg may end up sleeping in malloc.
675	 */
676	mtx_unlock(&cs->cs_lock);
677
678	if (ext && ext->rc_feedback)
679		ext->rc_feedback(FEEDBACK_OK, proc, ext->rc_feedback_arg);
680
681	xdrmbuf_create(&xdrs, cr->cr_mrep, XDR_DECODE);
682	ok = xdr_replymsg(&xdrs, &reply_msg);
683	cr->cr_mrep = NULL;
684
685	if (ok) {
686		if ((reply_msg.rm_reply.rp_stat == MSG_ACCEPTED) &&
687		    (reply_msg.acpted_rply.ar_stat == SUCCESS))
688			errp->re_status = stat = RPC_SUCCESS;
689		else
690			stat = _seterr_reply(&reply_msg, &(cu->cu_error));
691
692		if (errp->re_status == RPC_SUCCESS) {
693			results = xdrmbuf_getall(&xdrs);
694			if (! AUTH_VALIDATE(auth, xid,
695				&reply_msg.acpted_rply.ar_verf,
696				&results)) {
697				errp->re_status = stat = RPC_AUTHERROR;
698				errp->re_why = AUTH_INVALIDRESP;
699				if (retrans &&
700				    auth->ah_cred.oa_flavor == RPCSEC_GSS) {
701					/*
702					 * If we retransmitted, its
703					 * possible that we will
704					 * receive a reply for one of
705					 * the earlier transmissions
706					 * (which will use an older
707					 * RPCSEC_GSS sequence
708					 * number). In this case, just
709					 * go back and listen for a
710					 * new reply. We could keep a
711					 * record of all the seq
712					 * numbers we have transmitted
713					 * so far so that we could
714					 * accept a reply for any of
715					 * them here.
716					 */
717					XDR_DESTROY(&xdrs);
718					mtx_lock(&cs->cs_lock);
719					TAILQ_INSERT_TAIL(&cs->cs_pending,
720					    cr, cr_link);
721					cr->cr_mrep = NULL;
722					goto get_reply;
723				}
724			} else {
725				*resultsp = results;
726			}
727		}		/* end successful completion */
728		/*
729		 * If unsuccesful AND error is an authentication error
730		 * then refresh credentials and try again, else break
731		 */
732		else if (stat == RPC_AUTHERROR)
733			/* maybe our credentials need to be refreshed ... */
734			if (nrefreshes > 0 &&
735			    AUTH_REFRESH(auth, &reply_msg)) {
736				nrefreshes--;
737				XDR_DESTROY(&xdrs);
738				mtx_lock(&cs->cs_lock);
739				goto call_again;
740			}
741		/* end of unsuccessful completion */
742	}	/* end of valid reply message */
743	else {
744		errp->re_status = stat = RPC_CANTDECODERES;
745
746	}
747	XDR_DESTROY(&xdrs);
748	mtx_lock(&cs->cs_lock);
749out:
750	mtx_assert(&cs->cs_lock, MA_OWNED);
751
752	if (mreq)
753		m_freem(mreq);
754	if (cr->cr_mrep)
755		m_freem(cr->cr_mrep);
756
757	cu->cu_threads--;
758	if (cu->cu_closing)
759		wakeup(cu);
760
761	mtx_unlock(&cs->cs_lock);
762
763	if (auth && stat != RPC_SUCCESS)
764		AUTH_VALIDATE(auth, xid, NULL, NULL);
765
766	free(cr, M_RPC);
767
768	return (stat);
769}
770
771static void
772clnt_dg_geterr(CLIENT *cl, struct rpc_err *errp)
773{
774	struct cu_data *cu = (struct cu_data *)cl->cl_private;
775
776	*errp = cu->cu_error;
777}
778
779static bool_t
780clnt_dg_freeres(CLIENT *cl, xdrproc_t xdr_res, void *res_ptr)
781{
782	XDR xdrs;
783	bool_t dummy;
784
785	xdrs.x_op = XDR_FREE;
786	dummy = (*xdr_res)(&xdrs, res_ptr);
787
788	return (dummy);
789}
790
791/*ARGSUSED*/
792static void
793clnt_dg_abort(CLIENT *h)
794{
795}
796
797static bool_t
798clnt_dg_control(CLIENT *cl, u_int request, void *info)
799{
800	struct cu_data *cu = (struct cu_data *)cl->cl_private;
801	struct cu_socket *cs;
802	struct sockaddr *addr;
803
804	cs = cu->cu_socket->so_rcv.sb_upcallarg;
805	mtx_lock(&cs->cs_lock);
806
807	switch (request) {
808	case CLSET_FD_CLOSE:
809		cu->cu_closeit = TRUE;
810		mtx_unlock(&cs->cs_lock);
811		return (TRUE);
812	case CLSET_FD_NCLOSE:
813		cu->cu_closeit = FALSE;
814		mtx_unlock(&cs->cs_lock);
815		return (TRUE);
816	}
817
818	/* for other requests which use info */
819	if (info == NULL) {
820		mtx_unlock(&cs->cs_lock);
821		return (FALSE);
822	}
823	switch (request) {
824	case CLSET_TIMEOUT:
825		if (time_not_ok((struct timeval *)info)) {
826			mtx_unlock(&cs->cs_lock);
827			return (FALSE);
828		}
829		cu->cu_total = *(struct timeval *)info;
830		break;
831	case CLGET_TIMEOUT:
832		*(struct timeval *)info = cu->cu_total;
833		break;
834	case CLSET_RETRY_TIMEOUT:
835		if (time_not_ok((struct timeval *)info)) {
836			mtx_unlock(&cs->cs_lock);
837			return (FALSE);
838		}
839		cu->cu_wait = *(struct timeval *)info;
840		break;
841	case CLGET_RETRY_TIMEOUT:
842		*(struct timeval *)info = cu->cu_wait;
843		break;
844	case CLGET_SVC_ADDR:
845		/*
846		 * Slightly different semantics to userland - we use
847		 * sockaddr instead of netbuf.
848		 */
849		memcpy(info, &cu->cu_raddr, cu->cu_raddr.ss_len);
850		break;
851	case CLSET_SVC_ADDR:		/* set to new address */
852		addr = (struct sockaddr *)info;
853		(void) memcpy(&cu->cu_raddr, addr, addr->sa_len);
854		break;
855	case CLGET_XID:
856		*(uint32_t *)info = cu->cu_xid;
857		break;
858
859	case CLSET_XID:
860		/* This will set the xid of the NEXT call */
861		/* decrement by 1 as clnt_dg_call() increments once */
862		cu->cu_xid = *(uint32_t *)info - 1;
863		break;
864
865	case CLGET_VERS:
866		/*
867		 * This RELIES on the information that, in the call body,
868		 * the version number field is the fifth field from the
869		 * begining of the RPC header. MUST be changed if the
870		 * call_struct is changed
871		 */
872		*(uint32_t *)info =
873		    ntohl(*(uint32_t *)(void *)(cu->cu_mcallc +
874		    4 * BYTES_PER_XDR_UNIT));
875		break;
876
877	case CLSET_VERS:
878		*(uint32_t *)(void *)(cu->cu_mcallc + 4 * BYTES_PER_XDR_UNIT)
879			= htonl(*(uint32_t *)info);
880		break;
881
882	case CLGET_PROG:
883		/*
884		 * This RELIES on the information that, in the call body,
885		 * the program number field is the fourth field from the
886		 * begining of the RPC header. MUST be changed if the
887		 * call_struct is changed
888		 */
889		*(uint32_t *)info =
890		    ntohl(*(uint32_t *)(void *)(cu->cu_mcallc +
891		    3 * BYTES_PER_XDR_UNIT));
892		break;
893
894	case CLSET_PROG:
895		*(uint32_t *)(void *)(cu->cu_mcallc + 3 * BYTES_PER_XDR_UNIT)
896			= htonl(*(uint32_t *)info);
897		break;
898	case CLSET_ASYNC:
899		cu->cu_async = *(int *)info;
900		break;
901	case CLSET_CONNECT:
902		cu->cu_connect = *(int *)info;
903		break;
904	case CLSET_WAITCHAN:
905		cu->cu_waitchan = (const char *)info;
906		break;
907	case CLGET_WAITCHAN:
908		*(const char **) info = cu->cu_waitchan;
909		break;
910	case CLSET_INTERRUPTIBLE:
911		if (*(int *) info)
912			cu->cu_waitflag = PCATCH;
913		else
914			cu->cu_waitflag = 0;
915		break;
916	case CLGET_INTERRUPTIBLE:
917		if (cu->cu_waitflag)
918			*(int *) info = TRUE;
919		else
920			*(int *) info = FALSE;
921		break;
922	default:
923		mtx_unlock(&cs->cs_lock);
924		return (FALSE);
925	}
926	mtx_unlock(&cs->cs_lock);
927	return (TRUE);
928}
929
930static void
931clnt_dg_close(CLIENT *cl)
932{
933	struct cu_data *cu = (struct cu_data *)cl->cl_private;
934	struct cu_socket *cs;
935	struct cu_request *cr;
936
937	cs = cu->cu_socket->so_rcv.sb_upcallarg;
938	mtx_lock(&cs->cs_lock);
939
940	if (cu->cu_closed) {
941		mtx_unlock(&cs->cs_lock);
942		return;
943	}
944
945	if (cu->cu_closing) {
946		while (cu->cu_closing)
947			msleep(cu, &cs->cs_lock, 0, "rpcclose", 0);
948		KASSERT(cu->cu_closed, ("client should be closed"));
949		mtx_unlock(&cs->cs_lock);
950		return;
951	}
952
953	/*
954	 * Abort any pending requests and wait until everyone
955	 * has finished with clnt_vc_call.
956	 */
957	cu->cu_closing = TRUE;
958	TAILQ_FOREACH(cr, &cs->cs_pending, cr_link) {
959		if (cr->cr_client == cl) {
960			cr->cr_xid = 0;
961			cr->cr_error = ESHUTDOWN;
962			wakeup(cr);
963		}
964	}
965
966	while (cu->cu_threads)
967		msleep(cu, &cs->cs_lock, 0, "rpcclose", 0);
968
969	cu->cu_closing = FALSE;
970	cu->cu_closed = TRUE;
971
972	mtx_unlock(&cs->cs_lock);
973	wakeup(cu);
974}
975
976static void
977clnt_dg_destroy(CLIENT *cl)
978{
979	struct cu_data *cu = (struct cu_data *)cl->cl_private;
980	struct cu_socket *cs;
981	struct socket *so = NULL;
982	bool_t lastsocketref;
983
984	cs = cu->cu_socket->so_rcv.sb_upcallarg;
985	clnt_dg_close(cl);
986
987	mtx_lock(&cs->cs_lock);
988
989	cs->cs_refs--;
990	if (cs->cs_refs == 0) {
991		mtx_destroy(&cs->cs_lock);
992		SOCKBUF_LOCK(&cu->cu_socket->so_rcv);
993		soupcall_clear(cu->cu_socket, SO_RCV);
994		SOCKBUF_UNLOCK(&cu->cu_socket->so_rcv);
995		mem_free(cs, sizeof(*cs));
996		lastsocketref = TRUE;
997	} else {
998		mtx_unlock(&cs->cs_lock);
999		lastsocketref = FALSE;
1000	}
1001
1002	if (cu->cu_closeit && lastsocketref) {
1003		so = cu->cu_socket;
1004		cu->cu_socket = NULL;
1005	}
1006
1007	if (so)
1008		soclose(so);
1009
1010	if (cl->cl_netid && cl->cl_netid[0])
1011		mem_free(cl->cl_netid, strlen(cl->cl_netid) +1);
1012	if (cl->cl_tp && cl->cl_tp[0])
1013		mem_free(cl->cl_tp, strlen(cl->cl_tp) +1);
1014	mem_free(cu, sizeof (*cu));
1015	mem_free(cl, sizeof (CLIENT));
1016}
1017
1018/*
1019 * Make sure that the time is not garbage.  -1 value is allowed.
1020 */
1021static bool_t
1022time_not_ok(struct timeval *t)
1023{
1024	return (t->tv_sec < -1 || t->tv_sec > 100000000 ||
1025		t->tv_usec < -1 || t->tv_usec > 1000000);
1026}
1027
1028int
1029clnt_dg_soupcall(struct socket *so, void *arg, int waitflag)
1030{
1031	struct cu_socket *cs = (struct cu_socket *) arg;
1032	struct uio uio;
1033	struct mbuf *m;
1034	struct mbuf *control;
1035	struct cu_request *cr;
1036	int error, rcvflag, foundreq;
1037	uint32_t xid;
1038
1039	uio.uio_resid = 1000000000;
1040	uio.uio_td = curthread;
1041	do {
1042		SOCKBUF_UNLOCK(&so->so_rcv);
1043		m = NULL;
1044		control = NULL;
1045		rcvflag = MSG_DONTWAIT;
1046		error = soreceive(so, NULL, &uio, &m, &control, &rcvflag);
1047		if (control)
1048			m_freem(control);
1049		SOCKBUF_LOCK(&so->so_rcv);
1050
1051		if (error == EWOULDBLOCK)
1052			break;
1053
1054		/*
1055		 * If there was an error, wake up all pending
1056		 * requests.
1057		 */
1058		if (error) {
1059			mtx_lock(&cs->cs_lock);
1060			TAILQ_FOREACH(cr, &cs->cs_pending, cr_link) {
1061				cr->cr_xid = 0;
1062				cr->cr_error = error;
1063				wakeup(cr);
1064			}
1065			mtx_unlock(&cs->cs_lock);
1066			break;
1067		}
1068
1069		/*
1070		 * The XID is in the first uint32_t of the reply.
1071		 */
1072		if (m->m_len < sizeof(xid))
1073			m = m_pullup(m, sizeof(xid));
1074		if (!m)
1075			/*
1076			 * Should never happen.
1077			 */
1078			continue;
1079
1080		xid = ntohl(*mtod(m, uint32_t *));
1081
1082		/*
1083		 * Attempt to match this reply with a pending request.
1084		 */
1085		mtx_lock(&cs->cs_lock);
1086		foundreq = 0;
1087		TAILQ_FOREACH(cr, &cs->cs_pending, cr_link) {
1088			if (cr->cr_xid == xid) {
1089				/*
1090				 * This one matches. We leave the
1091				 * reply mbuf in cr->cr_mrep. Set the
1092				 * XID to zero so that we will ignore
1093				 * any duplicated replies that arrive
1094				 * before clnt_dg_call removes it from
1095				 * the queue.
1096				 */
1097				cr->cr_xid = 0;
1098				cr->cr_mrep = m;
1099				cr->cr_error = 0;
1100				foundreq = 1;
1101				wakeup(cr);
1102				break;
1103			}
1104		}
1105		mtx_unlock(&cs->cs_lock);
1106
1107		/*
1108		 * If we didn't find the matching request, just drop
1109		 * it - its probably a repeated reply.
1110		 */
1111		if (!foundreq)
1112			m_freem(m);
1113	} while (m);
1114	return (SU_OK);
1115}
1116
1117