1/*	$NetBSD: svc.c,v 1.21 2000/07/06 03:10:35 christos Exp $	*/
2
3/*
4 * Sun RPC is a product of Sun Microsystems, Inc. and is provided for
5 * unrestricted use provided that this legend is included on all tape
6 * media and as a part of the software program in whole or part.  Users
7 * may copy or modify Sun RPC without charge, but are not authorized
8 * to license or distribute it to anyone else except as part of a product or
9 * program developed by the user.
10 *
11 * SUN RPC IS PROVIDED AS IS WITH NO WARRANTIES OF ANY KIND INCLUDING THE
12 * WARRANTIES OF DESIGN, MERCHANTIBILITY AND FITNESS FOR A PARTICULAR
13 * PURPOSE, OR ARISING FROM A COURSE OF DEALING, USAGE OR TRADE PRACTICE.
14 *
15 * Sun RPC is provided with no support and without any obligation on the
16 * part of Sun Microsystems, Inc. to assist in its use, correction,
17 * modification or enhancement.
18 *
19 * SUN MICROSYSTEMS, INC. SHALL HAVE NO LIABILITY WITH RESPECT TO THE
20 * INFRINGEMENT OF COPYRIGHTS, TRADE SECRETS OR ANY PATENTS BY SUN RPC
21 * OR ANY PART THEREOF.
22 *
23 * In no event will Sun Microsystems, Inc. be liable for any lost revenue
24 * or profits or other special, indirect and consequential damages, even if
25 * Sun has been advised of the possibility of such damages.
26 *
27 * Sun Microsystems, Inc.
28 * 2550 Garcia Avenue
29 * Mountain View, California  94043
30 */
31
32#if defined(LIBC_SCCS) && !defined(lint)
33static char *sccsid2 = "@(#)svc.c 1.44 88/02/08 Copyr 1984 Sun Micro";
34static char *sccsid = "@(#)svc.c	2.4 88/08/11 4.0 RPCSRC";
35#endif
36#include <sys/cdefs.h>
37__FBSDID("$FreeBSD$");
38
39/*
40 * svc.c, Server-side remote procedure call interface.
41 *
42 * There are two sets of procedures here.  The xprt routines are
43 * for handling transport handles.  The svc routines handle the
44 * list of service routines.
45 *
46 * Copyright (C) 1984, Sun Microsystems, Inc.
47 */
48
49#include <sys/param.h>
50#include <sys/lock.h>
51#include <sys/kernel.h>
52#include <sys/kthread.h>
53#include <sys/malloc.h>
54#include <sys/mbuf.h>
55#include <sys/mutex.h>
56#include <sys/proc.h>
57#include <sys/queue.h>
58#include <sys/socketvar.h>
59#include <sys/systm.h>
60#include <sys/ucred.h>
61
62#include <rpc/rpc.h>
63#include <rpc/rpcb_clnt.h>
64#include <rpc/replay.h>
65
66#include <rpc/rpc_com.h>
67
68#define SVC_VERSQUIET 0x0001		/* keep quiet about vers mismatch */
69#define version_keepquiet(xp) (SVC_EXT(xp)->xp_flags & SVC_VERSQUIET)
70
71static struct svc_callout *svc_find(SVCPOOL *pool, rpcprog_t, rpcvers_t,
72    char *);
73static void svc_new_thread(SVCPOOL *pool);
74static void xprt_unregister_locked(SVCXPRT *xprt);
75
76/* ***************  SVCXPRT related stuff **************** */
77
78static int svcpool_minthread_sysctl(SYSCTL_HANDLER_ARGS);
79static int svcpool_maxthread_sysctl(SYSCTL_HANDLER_ARGS);
80
81SVCPOOL*
82svcpool_create(const char *name, struct sysctl_oid_list *sysctl_base)
83{
84	SVCPOOL *pool;
85
86	pool = malloc(sizeof(SVCPOOL), M_RPC, M_WAITOK|M_ZERO);
87
88	mtx_init(&pool->sp_lock, "sp_lock", NULL, MTX_DEF);
89	pool->sp_name = name;
90	pool->sp_state = SVCPOOL_INIT;
91	pool->sp_proc = NULL;
92	TAILQ_INIT(&pool->sp_xlist);
93	TAILQ_INIT(&pool->sp_active);
94	TAILQ_INIT(&pool->sp_callouts);
95	LIST_INIT(&pool->sp_threads);
96	LIST_INIT(&pool->sp_idlethreads);
97	pool->sp_minthreads = 1;
98	pool->sp_maxthreads = 1;
99	pool->sp_threadcount = 0;
100
101	/*
102	 * Don't use more than a quarter of mbuf clusters or more than
103	 * 45Mb buffering requests.
104	 */
105	pool->sp_space_high = nmbclusters * MCLBYTES / 4;
106	if (pool->sp_space_high > 45 << 20)
107		pool->sp_space_high = 45 << 20;
108	pool->sp_space_low = 2 * pool->sp_space_high / 3;
109
110	sysctl_ctx_init(&pool->sp_sysctl);
111	if (sysctl_base) {
112		SYSCTL_ADD_PROC(&pool->sp_sysctl, sysctl_base, OID_AUTO,
113		    "minthreads", CTLTYPE_INT | CTLFLAG_RW,
114		    pool, 0, svcpool_minthread_sysctl, "I", "");
115		SYSCTL_ADD_PROC(&pool->sp_sysctl, sysctl_base, OID_AUTO,
116		    "maxthreads", CTLTYPE_INT | CTLFLAG_RW,
117		    pool, 0, svcpool_maxthread_sysctl, "I", "");
118		SYSCTL_ADD_INT(&pool->sp_sysctl, sysctl_base, OID_AUTO,
119		    "threads", CTLFLAG_RD, &pool->sp_threadcount, 0, "");
120
121		SYSCTL_ADD_UINT(&pool->sp_sysctl, sysctl_base, OID_AUTO,
122		    "request_space_used", CTLFLAG_RD,
123		    &pool->sp_space_used, 0,
124		    "Space in parsed but not handled requests.");
125
126		SYSCTL_ADD_UINT(&pool->sp_sysctl, sysctl_base, OID_AUTO,
127		    "request_space_used_highest", CTLFLAG_RD,
128		    &pool->sp_space_used_highest, 0,
129		    "Highest space used since reboot.");
130
131		SYSCTL_ADD_UINT(&pool->sp_sysctl, sysctl_base, OID_AUTO,
132		    "request_space_high", CTLFLAG_RW,
133		    &pool->sp_space_high, 0,
134		    "Maximum space in parsed but not handled requests.");
135
136		SYSCTL_ADD_UINT(&pool->sp_sysctl, sysctl_base, OID_AUTO,
137		    "request_space_low", CTLFLAG_RW,
138		    &pool->sp_space_low, 0,
139		    "Low water mark for request space.");
140
141		SYSCTL_ADD_INT(&pool->sp_sysctl, sysctl_base, OID_AUTO,
142		    "request_space_throttled", CTLFLAG_RD,
143		    &pool->sp_space_throttled, 0,
144		    "Whether nfs requests are currently throttled");
145
146		SYSCTL_ADD_INT(&pool->sp_sysctl, sysctl_base, OID_AUTO,
147		    "request_space_throttle_count", CTLFLAG_RD,
148		    &pool->sp_space_throttle_count, 0,
149		    "Count of times throttling based on request space has occurred");
150	}
151
152	return pool;
153}
154
155void
156svcpool_destroy(SVCPOOL *pool)
157{
158	SVCXPRT *xprt, *nxprt;
159	struct svc_callout *s;
160	struct svcxprt_list cleanup;
161
162	TAILQ_INIT(&cleanup);
163	mtx_lock(&pool->sp_lock);
164
165	while (TAILQ_FIRST(&pool->sp_xlist)) {
166		xprt = TAILQ_FIRST(&pool->sp_xlist);
167		xprt_unregister_locked(xprt);
168		TAILQ_INSERT_TAIL(&cleanup, xprt, xp_link);
169	}
170
171	while (TAILQ_FIRST(&pool->sp_callouts)) {
172		s = TAILQ_FIRST(&pool->sp_callouts);
173		mtx_unlock(&pool->sp_lock);
174		svc_unreg(pool, s->sc_prog, s->sc_vers);
175		mtx_lock(&pool->sp_lock);
176	}
177	mtx_unlock(&pool->sp_lock);
178
179	TAILQ_FOREACH_SAFE(xprt, &cleanup, xp_link, nxprt) {
180		SVC_RELEASE(xprt);
181	}
182
183	mtx_destroy(&pool->sp_lock);
184
185	if (pool->sp_rcache)
186		replay_freecache(pool->sp_rcache);
187
188	sysctl_ctx_free(&pool->sp_sysctl);
189	free(pool, M_RPC);
190}
191
192static bool_t
193svcpool_active(SVCPOOL *pool)
194{
195	enum svcpool_state state = pool->sp_state;
196
197	if (state == SVCPOOL_INIT || state == SVCPOOL_CLOSING)
198		return (FALSE);
199	return (TRUE);
200}
201
202/*
203 * Sysctl handler to set the minimum thread count on a pool
204 */
205static int
206svcpool_minthread_sysctl(SYSCTL_HANDLER_ARGS)
207{
208	SVCPOOL *pool;
209	int newminthreads, error, n;
210
211	pool = oidp->oid_arg1;
212	newminthreads = pool->sp_minthreads;
213	error = sysctl_handle_int(oidp, &newminthreads, 0, req);
214	if (error == 0 && newminthreads != pool->sp_minthreads) {
215		if (newminthreads > pool->sp_maxthreads)
216			return (EINVAL);
217		mtx_lock(&pool->sp_lock);
218		if (newminthreads > pool->sp_minthreads
219		    && svcpool_active(pool)) {
220			/*
221			 * If the pool is running and we are
222			 * increasing, create some more threads now.
223			 */
224			n = newminthreads - pool->sp_threadcount;
225			if (n > 0) {
226				mtx_unlock(&pool->sp_lock);
227				while (n--)
228					svc_new_thread(pool);
229				mtx_lock(&pool->sp_lock);
230			}
231		}
232		pool->sp_minthreads = newminthreads;
233		mtx_unlock(&pool->sp_lock);
234	}
235	return (error);
236}
237
238/*
239 * Sysctl handler to set the maximum thread count on a pool
240 */
241static int
242svcpool_maxthread_sysctl(SYSCTL_HANDLER_ARGS)
243{
244	SVCPOOL *pool;
245	SVCTHREAD *st;
246	int newmaxthreads, error;
247
248	pool = oidp->oid_arg1;
249	newmaxthreads = pool->sp_maxthreads;
250	error = sysctl_handle_int(oidp, &newmaxthreads, 0, req);
251	if (error == 0 && newmaxthreads != pool->sp_maxthreads) {
252		if (newmaxthreads < pool->sp_minthreads)
253			return (EINVAL);
254		mtx_lock(&pool->sp_lock);
255		if (newmaxthreads < pool->sp_maxthreads
256		    && svcpool_active(pool)) {
257			/*
258			 * If the pool is running and we are
259			 * decreasing, wake up some idle threads to
260			 * encourage them to exit.
261			 */
262			LIST_FOREACH(st, &pool->sp_idlethreads, st_ilink)
263				cv_signal(&st->st_cond);
264		}
265		pool->sp_maxthreads = newmaxthreads;
266		mtx_unlock(&pool->sp_lock);
267	}
268	return (error);
269}
270
271/*
272 * Activate a transport handle.
273 */
274void
275xprt_register(SVCXPRT *xprt)
276{
277	SVCPOOL *pool = xprt->xp_pool;
278
279	SVC_ACQUIRE(xprt);
280	mtx_lock(&pool->sp_lock);
281	xprt->xp_registered = TRUE;
282	xprt->xp_active = FALSE;
283	TAILQ_INSERT_TAIL(&pool->sp_xlist, xprt, xp_link);
284	mtx_unlock(&pool->sp_lock);
285}
286
287/*
288 * De-activate a transport handle. Note: the locked version doesn't
289 * release the transport - caller must do that after dropping the pool
290 * lock.
291 */
292static void
293xprt_unregister_locked(SVCXPRT *xprt)
294{
295	SVCPOOL *pool = xprt->xp_pool;
296
297	KASSERT(xprt->xp_registered == TRUE,
298	    ("xprt_unregister_locked: not registered"));
299	if (xprt->xp_active) {
300		TAILQ_REMOVE(&pool->sp_active, xprt, xp_alink);
301		xprt->xp_active = FALSE;
302	}
303	TAILQ_REMOVE(&pool->sp_xlist, xprt, xp_link);
304	xprt->xp_registered = FALSE;
305}
306
307void
308xprt_unregister(SVCXPRT *xprt)
309{
310	SVCPOOL *pool = xprt->xp_pool;
311
312	mtx_lock(&pool->sp_lock);
313	if (xprt->xp_registered == FALSE) {
314		/* Already unregistered by another thread */
315		mtx_unlock(&pool->sp_lock);
316		return;
317	}
318	xprt_unregister_locked(xprt);
319	mtx_unlock(&pool->sp_lock);
320
321	SVC_RELEASE(xprt);
322}
323
324static void
325xprt_assignthread(SVCXPRT *xprt)
326{
327	SVCPOOL *pool = xprt->xp_pool;
328	SVCTHREAD *st;
329
330	/*
331	 * Attempt to assign a service thread to this
332	 * transport.
333	 */
334	LIST_FOREACH(st, &pool->sp_idlethreads, st_ilink) {
335		if (st->st_xprt == NULL && STAILQ_EMPTY(&st->st_reqs))
336			break;
337	}
338	if (st) {
339		SVC_ACQUIRE(xprt);
340		xprt->xp_thread = st;
341		st->st_xprt = xprt;
342		cv_signal(&st->st_cond);
343	} else {
344		/*
345		 * See if we can create a new thread. The
346		 * actual thread creation happens in
347		 * svc_run_internal because our locking state
348		 * is poorly defined (we are typically called
349		 * from a socket upcall). Don't create more
350		 * than one thread per second.
351		 */
352		if (pool->sp_state == SVCPOOL_ACTIVE
353		    && pool->sp_lastcreatetime < time_uptime
354		    && pool->sp_threadcount < pool->sp_maxthreads) {
355			pool->sp_state = SVCPOOL_THREADWANTED;
356		}
357	}
358}
359
360void
361xprt_active(SVCXPRT *xprt)
362{
363	SVCPOOL *pool = xprt->xp_pool;
364
365	mtx_lock(&pool->sp_lock);
366
367	if (!xprt->xp_registered) {
368		/*
369		 * Race with xprt_unregister - we lose.
370		 */
371		mtx_unlock(&pool->sp_lock);
372		return;
373	}
374
375	if (!xprt->xp_active) {
376		TAILQ_INSERT_TAIL(&pool->sp_active, xprt, xp_alink);
377		xprt->xp_active = TRUE;
378		xprt_assignthread(xprt);
379	}
380
381	mtx_unlock(&pool->sp_lock);
382}
383
384void
385xprt_inactive_locked(SVCXPRT *xprt)
386{
387	SVCPOOL *pool = xprt->xp_pool;
388
389	if (xprt->xp_active) {
390		TAILQ_REMOVE(&pool->sp_active, xprt, xp_alink);
391		xprt->xp_active = FALSE;
392	}
393}
394
395void
396xprt_inactive(SVCXPRT *xprt)
397{
398	SVCPOOL *pool = xprt->xp_pool;
399
400	mtx_lock(&pool->sp_lock);
401	xprt_inactive_locked(xprt);
402	mtx_unlock(&pool->sp_lock);
403}
404
405/*
406 * Add a service program to the callout list.
407 * The dispatch routine will be called when a rpc request for this
408 * program number comes in.
409 */
410bool_t
411svc_reg(SVCXPRT *xprt, const rpcprog_t prog, const rpcvers_t vers,
412    void (*dispatch)(struct svc_req *, SVCXPRT *),
413    const struct netconfig *nconf)
414{
415	SVCPOOL *pool = xprt->xp_pool;
416	struct svc_callout *s;
417	char *netid = NULL;
418	int flag = 0;
419
420/* VARIABLES PROTECTED BY svc_lock: s, svc_head */
421
422	if (xprt->xp_netid) {
423		netid = strdup(xprt->xp_netid, M_RPC);
424		flag = 1;
425	} else if (nconf && nconf->nc_netid) {
426		netid = strdup(nconf->nc_netid, M_RPC);
427		flag = 1;
428	} /* must have been created with svc_raw_create */
429	if ((netid == NULL) && (flag == 1)) {
430		return (FALSE);
431	}
432
433	mtx_lock(&pool->sp_lock);
434	if ((s = svc_find(pool, prog, vers, netid)) != NULL) {
435		if (netid)
436			free(netid, M_RPC);
437		if (s->sc_dispatch == dispatch)
438			goto rpcb_it; /* he is registering another xptr */
439		mtx_unlock(&pool->sp_lock);
440		return (FALSE);
441	}
442	s = malloc(sizeof (struct svc_callout), M_RPC, M_NOWAIT);
443	if (s == NULL) {
444		if (netid)
445			free(netid, M_RPC);
446		mtx_unlock(&pool->sp_lock);
447		return (FALSE);
448	}
449
450	s->sc_prog = prog;
451	s->sc_vers = vers;
452	s->sc_dispatch = dispatch;
453	s->sc_netid = netid;
454	TAILQ_INSERT_TAIL(&pool->sp_callouts, s, sc_link);
455
456	if ((xprt->xp_netid == NULL) && (flag == 1) && netid)
457		((SVCXPRT *) xprt)->xp_netid = strdup(netid, M_RPC);
458
459rpcb_it:
460	mtx_unlock(&pool->sp_lock);
461	/* now register the information with the local binder service */
462	if (nconf) {
463		bool_t dummy;
464		struct netconfig tnc;
465		struct netbuf nb;
466		tnc = *nconf;
467		nb.buf = &xprt->xp_ltaddr;
468		nb.len = xprt->xp_ltaddr.ss_len;
469		dummy = rpcb_set(prog, vers, &tnc, &nb);
470		return (dummy);
471	}
472	return (TRUE);
473}
474
475/*
476 * Remove a service program from the callout list.
477 */
478void
479svc_unreg(SVCPOOL *pool, const rpcprog_t prog, const rpcvers_t vers)
480{
481	struct svc_callout *s;
482
483	/* unregister the information anyway */
484	(void) rpcb_unset(prog, vers, NULL);
485	mtx_lock(&pool->sp_lock);
486	while ((s = svc_find(pool, prog, vers, NULL)) != NULL) {
487		TAILQ_REMOVE(&pool->sp_callouts, s, sc_link);
488		if (s->sc_netid)
489			mem_free(s->sc_netid, sizeof (s->sc_netid) + 1);
490		mem_free(s, sizeof (struct svc_callout));
491	}
492	mtx_unlock(&pool->sp_lock);
493}
494
495/* ********************** CALLOUT list related stuff ************* */
496
497/*
498 * Search the callout list for a program number, return the callout
499 * struct.
500 */
501static struct svc_callout *
502svc_find(SVCPOOL *pool, rpcprog_t prog, rpcvers_t vers, char *netid)
503{
504	struct svc_callout *s;
505
506	mtx_assert(&pool->sp_lock, MA_OWNED);
507	TAILQ_FOREACH(s, &pool->sp_callouts, sc_link) {
508		if (s->sc_prog == prog && s->sc_vers == vers
509		    && (netid == NULL || s->sc_netid == NULL ||
510			strcmp(netid, s->sc_netid) == 0))
511			break;
512	}
513
514	return (s);
515}
516
517/* ******************* REPLY GENERATION ROUTINES  ************ */
518
519static bool_t
520svc_sendreply_common(struct svc_req *rqstp, struct rpc_msg *rply,
521    struct mbuf *body)
522{
523	SVCXPRT *xprt = rqstp->rq_xprt;
524	bool_t ok;
525
526	if (rqstp->rq_args) {
527		m_freem(rqstp->rq_args);
528		rqstp->rq_args = NULL;
529	}
530
531	if (xprt->xp_pool->sp_rcache)
532		replay_setreply(xprt->xp_pool->sp_rcache,
533		    rply, svc_getrpccaller(rqstp), body);
534
535	if (!SVCAUTH_WRAP(&rqstp->rq_auth, &body))
536		return (FALSE);
537
538	ok = SVC_REPLY(xprt, rply, rqstp->rq_addr, body);
539	if (rqstp->rq_addr) {
540		free(rqstp->rq_addr, M_SONAME);
541		rqstp->rq_addr = NULL;
542	}
543
544	return (ok);
545}
546
547/*
548 * Send a reply to an rpc request
549 */
550bool_t
551svc_sendreply(struct svc_req *rqstp, xdrproc_t xdr_results, void * xdr_location)
552{
553	struct rpc_msg rply;
554	struct mbuf *m;
555	XDR xdrs;
556	bool_t ok;
557
558	rply.rm_xid = rqstp->rq_xid;
559	rply.rm_direction = REPLY;
560	rply.rm_reply.rp_stat = MSG_ACCEPTED;
561	rply.acpted_rply.ar_verf = rqstp->rq_verf;
562	rply.acpted_rply.ar_stat = SUCCESS;
563	rply.acpted_rply.ar_results.where = NULL;
564	rply.acpted_rply.ar_results.proc = (xdrproc_t) xdr_void;
565
566	m = m_getcl(M_WAITOK, MT_DATA, 0);
567	xdrmbuf_create(&xdrs, m, XDR_ENCODE);
568	ok = xdr_results(&xdrs, xdr_location);
569	XDR_DESTROY(&xdrs);
570
571	if (ok) {
572		return (svc_sendreply_common(rqstp, &rply, m));
573	} else {
574		m_freem(m);
575		return (FALSE);
576	}
577}
578
579bool_t
580svc_sendreply_mbuf(struct svc_req *rqstp, struct mbuf *m)
581{
582	struct rpc_msg rply;
583
584	rply.rm_xid = rqstp->rq_xid;
585	rply.rm_direction = REPLY;
586	rply.rm_reply.rp_stat = MSG_ACCEPTED;
587	rply.acpted_rply.ar_verf = rqstp->rq_verf;
588	rply.acpted_rply.ar_stat = SUCCESS;
589	rply.acpted_rply.ar_results.where = NULL;
590	rply.acpted_rply.ar_results.proc = (xdrproc_t) xdr_void;
591
592	return (svc_sendreply_common(rqstp, &rply, m));
593}
594
595/*
596 * No procedure error reply
597 */
598void
599svcerr_noproc(struct svc_req *rqstp)
600{
601	SVCXPRT *xprt = rqstp->rq_xprt;
602	struct rpc_msg rply;
603
604	rply.rm_xid = rqstp->rq_xid;
605	rply.rm_direction = REPLY;
606	rply.rm_reply.rp_stat = MSG_ACCEPTED;
607	rply.acpted_rply.ar_verf = rqstp->rq_verf;
608	rply.acpted_rply.ar_stat = PROC_UNAVAIL;
609
610	if (xprt->xp_pool->sp_rcache)
611		replay_setreply(xprt->xp_pool->sp_rcache,
612		    &rply, svc_getrpccaller(rqstp), NULL);
613
614	svc_sendreply_common(rqstp, &rply, NULL);
615}
616
617/*
618 * Can't decode args error reply
619 */
620void
621svcerr_decode(struct svc_req *rqstp)
622{
623	SVCXPRT *xprt = rqstp->rq_xprt;
624	struct rpc_msg rply;
625
626	rply.rm_xid = rqstp->rq_xid;
627	rply.rm_direction = REPLY;
628	rply.rm_reply.rp_stat = MSG_ACCEPTED;
629	rply.acpted_rply.ar_verf = rqstp->rq_verf;
630	rply.acpted_rply.ar_stat = GARBAGE_ARGS;
631
632	if (xprt->xp_pool->sp_rcache)
633		replay_setreply(xprt->xp_pool->sp_rcache,
634		    &rply, (struct sockaddr *) &xprt->xp_rtaddr, NULL);
635
636	svc_sendreply_common(rqstp, &rply, NULL);
637}
638
639/*
640 * Some system error
641 */
642void
643svcerr_systemerr(struct svc_req *rqstp)
644{
645	SVCXPRT *xprt = rqstp->rq_xprt;
646	struct rpc_msg rply;
647
648	rply.rm_xid = rqstp->rq_xid;
649	rply.rm_direction = REPLY;
650	rply.rm_reply.rp_stat = MSG_ACCEPTED;
651	rply.acpted_rply.ar_verf = rqstp->rq_verf;
652	rply.acpted_rply.ar_stat = SYSTEM_ERR;
653
654	if (xprt->xp_pool->sp_rcache)
655		replay_setreply(xprt->xp_pool->sp_rcache,
656		    &rply, svc_getrpccaller(rqstp), NULL);
657
658	svc_sendreply_common(rqstp, &rply, NULL);
659}
660
661/*
662 * Authentication error reply
663 */
664void
665svcerr_auth(struct svc_req *rqstp, enum auth_stat why)
666{
667	SVCXPRT *xprt = rqstp->rq_xprt;
668	struct rpc_msg rply;
669
670	rply.rm_xid = rqstp->rq_xid;
671	rply.rm_direction = REPLY;
672	rply.rm_reply.rp_stat = MSG_DENIED;
673	rply.rjcted_rply.rj_stat = AUTH_ERROR;
674	rply.rjcted_rply.rj_why = why;
675
676	if (xprt->xp_pool->sp_rcache)
677		replay_setreply(xprt->xp_pool->sp_rcache,
678		    &rply, svc_getrpccaller(rqstp), NULL);
679
680	svc_sendreply_common(rqstp, &rply, NULL);
681}
682
683/*
684 * Auth too weak error reply
685 */
686void
687svcerr_weakauth(struct svc_req *rqstp)
688{
689
690	svcerr_auth(rqstp, AUTH_TOOWEAK);
691}
692
693/*
694 * Program unavailable error reply
695 */
696void
697svcerr_noprog(struct svc_req *rqstp)
698{
699	SVCXPRT *xprt = rqstp->rq_xprt;
700	struct rpc_msg rply;
701
702	rply.rm_xid = rqstp->rq_xid;
703	rply.rm_direction = REPLY;
704	rply.rm_reply.rp_stat = MSG_ACCEPTED;
705	rply.acpted_rply.ar_verf = rqstp->rq_verf;
706	rply.acpted_rply.ar_stat = PROG_UNAVAIL;
707
708	if (xprt->xp_pool->sp_rcache)
709		replay_setreply(xprt->xp_pool->sp_rcache,
710		    &rply, svc_getrpccaller(rqstp), NULL);
711
712	svc_sendreply_common(rqstp, &rply, NULL);
713}
714
715/*
716 * Program version mismatch error reply
717 */
718void
719svcerr_progvers(struct svc_req *rqstp, rpcvers_t low_vers, rpcvers_t high_vers)
720{
721	SVCXPRT *xprt = rqstp->rq_xprt;
722	struct rpc_msg rply;
723
724	rply.rm_xid = rqstp->rq_xid;
725	rply.rm_direction = REPLY;
726	rply.rm_reply.rp_stat = MSG_ACCEPTED;
727	rply.acpted_rply.ar_verf = rqstp->rq_verf;
728	rply.acpted_rply.ar_stat = PROG_MISMATCH;
729	rply.acpted_rply.ar_vers.low = (uint32_t)low_vers;
730	rply.acpted_rply.ar_vers.high = (uint32_t)high_vers;
731
732	if (xprt->xp_pool->sp_rcache)
733		replay_setreply(xprt->xp_pool->sp_rcache,
734		    &rply, svc_getrpccaller(rqstp), NULL);
735
736	svc_sendreply_common(rqstp, &rply, NULL);
737}
738
739/*
740 * Allocate a new server transport structure. All fields are
741 * initialized to zero and xp_p3 is initialized to point at an
742 * extension structure to hold various flags and authentication
743 * parameters.
744 */
745SVCXPRT *
746svc_xprt_alloc()
747{
748	SVCXPRT *xprt;
749	SVCXPRT_EXT *ext;
750
751	xprt = mem_alloc(sizeof(SVCXPRT));
752	memset(xprt, 0, sizeof(SVCXPRT));
753	ext = mem_alloc(sizeof(SVCXPRT_EXT));
754	memset(ext, 0, sizeof(SVCXPRT_EXT));
755	xprt->xp_p3 = ext;
756	refcount_init(&xprt->xp_refs, 1);
757
758	return (xprt);
759}
760
761/*
762 * Free a server transport structure.
763 */
764void
765svc_xprt_free(xprt)
766	SVCXPRT *xprt;
767{
768
769	mem_free(xprt->xp_p3, sizeof(SVCXPRT_EXT));
770	mem_free(xprt, sizeof(SVCXPRT));
771}
772
773/* ******************* SERVER INPUT STUFF ******************* */
774
775/*
776 * Read RPC requests from a transport and queue them to be
777 * executed. We handle authentication and replay cache replies here.
778 * Actually dispatching the RPC is deferred till svc_executereq.
779 */
780static enum xprt_stat
781svc_getreq(SVCXPRT *xprt, struct svc_req **rqstp_ret)
782{
783	SVCPOOL *pool = xprt->xp_pool;
784	struct svc_req *r;
785	struct rpc_msg msg;
786	struct mbuf *args;
787	enum xprt_stat stat;
788
789	/* now receive msgs from xprtprt (support batch calls) */
790	r = malloc(sizeof(*r), M_RPC, M_WAITOK|M_ZERO);
791
792	msg.rm_call.cb_cred.oa_base = r->rq_credarea;
793	msg.rm_call.cb_verf.oa_base = &r->rq_credarea[MAX_AUTH_BYTES];
794	r->rq_clntcred = &r->rq_credarea[2*MAX_AUTH_BYTES];
795	if (SVC_RECV(xprt, &msg, &r->rq_addr, &args)) {
796		enum auth_stat why;
797
798		/*
799		 * Handle replays and authenticate before queuing the
800		 * request to be executed.
801		 */
802		SVC_ACQUIRE(xprt);
803		r->rq_xprt = xprt;
804		if (pool->sp_rcache) {
805			struct rpc_msg repmsg;
806			struct mbuf *repbody;
807			enum replay_state rs;
808			rs = replay_find(pool->sp_rcache, &msg,
809			    svc_getrpccaller(r), &repmsg, &repbody);
810			switch (rs) {
811			case RS_NEW:
812				break;
813			case RS_DONE:
814				SVC_REPLY(xprt, &repmsg, r->rq_addr,
815				    repbody);
816				if (r->rq_addr) {
817					free(r->rq_addr, M_SONAME);
818					r->rq_addr = NULL;
819				}
820				m_freem(args);
821				goto call_done;
822
823			default:
824				m_freem(args);
825				goto call_done;
826			}
827		}
828
829		r->rq_xid = msg.rm_xid;
830		r->rq_prog = msg.rm_call.cb_prog;
831		r->rq_vers = msg.rm_call.cb_vers;
832		r->rq_proc = msg.rm_call.cb_proc;
833		r->rq_size = sizeof(*r) + m_length(args, NULL);
834		r->rq_args = args;
835		if ((why = _authenticate(r, &msg)) != AUTH_OK) {
836			/*
837			 * RPCSEC_GSS uses this return code
838			 * for requests that form part of its
839			 * context establishment protocol and
840			 * should not be dispatched to the
841			 * application.
842			 */
843			if (why != RPCSEC_GSS_NODISPATCH)
844				svcerr_auth(r, why);
845			goto call_done;
846		}
847
848		if (!SVCAUTH_UNWRAP(&r->rq_auth, &r->rq_args)) {
849			svcerr_decode(r);
850			goto call_done;
851		}
852
853		/*
854		 * Everything checks out, return request to caller.
855		 */
856		*rqstp_ret = r;
857		r = NULL;
858	}
859call_done:
860	if (r) {
861		svc_freereq(r);
862		r = NULL;
863	}
864	if ((stat = SVC_STAT(xprt)) == XPRT_DIED) {
865		xprt_unregister(xprt);
866	}
867
868	return (stat);
869}
870
871static void
872svc_executereq(struct svc_req *rqstp)
873{
874	SVCXPRT *xprt = rqstp->rq_xprt;
875	SVCPOOL *pool = xprt->xp_pool;
876	int prog_found;
877	rpcvers_t low_vers;
878	rpcvers_t high_vers;
879	struct svc_callout *s;
880
881	/* now match message with a registered service*/
882	prog_found = FALSE;
883	low_vers = (rpcvers_t) -1L;
884	high_vers = (rpcvers_t) 0L;
885	TAILQ_FOREACH(s, &pool->sp_callouts, sc_link) {
886		if (s->sc_prog == rqstp->rq_prog) {
887			if (s->sc_vers == rqstp->rq_vers) {
888				/*
889				 * We hand ownership of r to the
890				 * dispatch method - they must call
891				 * svc_freereq.
892				 */
893				(*s->sc_dispatch)(rqstp, xprt);
894				return;
895			}  /* found correct version */
896			prog_found = TRUE;
897			if (s->sc_vers < low_vers)
898				low_vers = s->sc_vers;
899			if (s->sc_vers > high_vers)
900				high_vers = s->sc_vers;
901		}   /* found correct program */
902	}
903
904	/*
905	 * if we got here, the program or version
906	 * is not served ...
907	 */
908	if (prog_found)
909		svcerr_progvers(rqstp, low_vers, high_vers);
910	else
911		svcerr_noprog(rqstp);
912
913	svc_freereq(rqstp);
914}
915
916static void
917svc_checkidle(SVCPOOL *pool)
918{
919	SVCXPRT *xprt, *nxprt;
920	time_t timo;
921	struct svcxprt_list cleanup;
922
923	TAILQ_INIT(&cleanup);
924	TAILQ_FOREACH_SAFE(xprt, &pool->sp_xlist, xp_link, nxprt) {
925		/*
926		 * Only some transports have idle timers. Don't time
927		 * something out which is just waking up.
928		 */
929		if (!xprt->xp_idletimeout || xprt->xp_thread)
930			continue;
931
932		timo = xprt->xp_lastactive + xprt->xp_idletimeout;
933		if (time_uptime > timo) {
934			xprt_unregister_locked(xprt);
935			TAILQ_INSERT_TAIL(&cleanup, xprt, xp_link);
936		}
937	}
938
939	mtx_unlock(&pool->sp_lock);
940	TAILQ_FOREACH_SAFE(xprt, &cleanup, xp_link, nxprt) {
941		SVC_RELEASE(xprt);
942	}
943	mtx_lock(&pool->sp_lock);
944
945}
946
947static void
948svc_assign_waiting_sockets(SVCPOOL *pool)
949{
950	SVCXPRT *xprt;
951
952	TAILQ_FOREACH(xprt, &pool->sp_active, xp_alink) {
953		if (!xprt->xp_thread) {
954			xprt_assignthread(xprt);
955		}
956	}
957}
958
959static bool_t
960svc_request_space_available(SVCPOOL *pool)
961{
962
963	mtx_assert(&pool->sp_lock, MA_OWNED);
964
965	if (pool->sp_space_throttled) {
966		/*
967		 * Below the low-water yet? If so, assign any waiting sockets.
968		 */
969		if (pool->sp_space_used < pool->sp_space_low) {
970			pool->sp_space_throttled = FALSE;
971			svc_assign_waiting_sockets(pool);
972			return TRUE;
973		}
974
975		return FALSE;
976	} else {
977		if (pool->sp_space_used
978		    >= pool->sp_space_high) {
979			pool->sp_space_throttled = TRUE;
980			pool->sp_space_throttle_count++;
981			return FALSE;
982		}
983
984		return TRUE;
985	}
986}
987
988static void
989svc_run_internal(SVCPOOL *pool, bool_t ismaster)
990{
991	SVCTHREAD *st, *stpref;
992	SVCXPRT *xprt;
993	enum xprt_stat stat;
994	struct svc_req *rqstp;
995	int error;
996
997	st = mem_alloc(sizeof(*st));
998	st->st_xprt = NULL;
999	STAILQ_INIT(&st->st_reqs);
1000	cv_init(&st->st_cond, "rpcsvc");
1001
1002	mtx_lock(&pool->sp_lock);
1003	LIST_INSERT_HEAD(&pool->sp_threads, st, st_link);
1004
1005	/*
1006	 * If we are a new thread which was spawned to cope with
1007	 * increased load, set the state back to SVCPOOL_ACTIVE.
1008	 */
1009	if (pool->sp_state == SVCPOOL_THREADSTARTING)
1010		pool->sp_state = SVCPOOL_ACTIVE;
1011
1012	while (pool->sp_state != SVCPOOL_CLOSING) {
1013		/*
1014		 * Check for idle transports once per second.
1015		 */
1016		if (time_uptime > pool->sp_lastidlecheck) {
1017			pool->sp_lastidlecheck = time_uptime;
1018			svc_checkidle(pool);
1019		}
1020
1021		xprt = st->st_xprt;
1022		if (!xprt && STAILQ_EMPTY(&st->st_reqs)) {
1023			/*
1024			 * Enforce maxthreads count.
1025			 */
1026			if (pool->sp_threadcount > pool->sp_maxthreads)
1027				break;
1028
1029			/*
1030			 * Before sleeping, see if we can find an
1031			 * active transport which isn't being serviced
1032			 * by a thread.
1033			 */
1034			if (svc_request_space_available(pool)) {
1035				TAILQ_FOREACH(xprt, &pool->sp_active,
1036				    xp_alink) {
1037					if (!xprt->xp_thread) {
1038						SVC_ACQUIRE(xprt);
1039						xprt->xp_thread = st;
1040						st->st_xprt = xprt;
1041						break;
1042					}
1043				}
1044			}
1045			if (st->st_xprt)
1046				continue;
1047
1048			LIST_INSERT_HEAD(&pool->sp_idlethreads, st, st_ilink);
1049			error = cv_timedwait_sig(&st->st_cond, &pool->sp_lock,
1050				5 * hz);
1051			LIST_REMOVE(st, st_ilink);
1052
1053			/*
1054			 * Reduce worker thread count when idle.
1055			 */
1056			if (error == EWOULDBLOCK) {
1057				if (!ismaster
1058				    && (pool->sp_threadcount
1059					> pool->sp_minthreads)
1060					&& !st->st_xprt
1061					&& STAILQ_EMPTY(&st->st_reqs))
1062					break;
1063			}
1064			if (error == EWOULDBLOCK)
1065				continue;
1066			if (error) {
1067				if (pool->sp_state != SVCPOOL_CLOSING) {
1068					mtx_unlock(&pool->sp_lock);
1069					svc_exit(pool);
1070					mtx_lock(&pool->sp_lock);
1071				}
1072				break;
1073			}
1074
1075			if (pool->sp_state == SVCPOOL_THREADWANTED) {
1076				pool->sp_state = SVCPOOL_THREADSTARTING;
1077				pool->sp_lastcreatetime = time_uptime;
1078				mtx_unlock(&pool->sp_lock);
1079				svc_new_thread(pool);
1080				mtx_lock(&pool->sp_lock);
1081			}
1082			continue;
1083		}
1084
1085		if (xprt) {
1086			/*
1087			 * Drain the transport socket and queue up any
1088			 * RPCs.
1089			 */
1090			xprt->xp_lastactive = time_uptime;
1091			stat = XPRT_IDLE;
1092			do {
1093				if (!svc_request_space_available(pool))
1094					break;
1095				rqstp = NULL;
1096				mtx_unlock(&pool->sp_lock);
1097				stat = svc_getreq(xprt, &rqstp);
1098				mtx_lock(&pool->sp_lock);
1099				if (rqstp) {
1100					/*
1101					 * See if the application has
1102					 * a preference for some other
1103					 * thread.
1104					 */
1105					stpref = st;
1106					if (pool->sp_assign)
1107						stpref = pool->sp_assign(st,
1108						    rqstp);
1109
1110					pool->sp_space_used +=
1111						rqstp->rq_size;
1112					if (pool->sp_space_used
1113					    > pool->sp_space_used_highest)
1114						pool->sp_space_used_highest =
1115							pool->sp_space_used;
1116					rqstp->rq_thread = stpref;
1117					STAILQ_INSERT_TAIL(&stpref->st_reqs,
1118					    rqstp, rq_link);
1119					stpref->st_reqcount++;
1120
1121					/*
1122					 * If we assigned the request
1123					 * to another thread, make
1124					 * sure its awake and continue
1125					 * reading from the
1126					 * socket. Otherwise, try to
1127					 * find some other thread to
1128					 * read from the socket and
1129					 * execute the request
1130					 * immediately.
1131					 */
1132					if (stpref != st) {
1133						cv_signal(&stpref->st_cond);
1134						continue;
1135					} else {
1136						break;
1137					}
1138				}
1139			} while (stat == XPRT_MOREREQS
1140			    && pool->sp_state != SVCPOOL_CLOSING);
1141
1142			/*
1143			 * Move this transport to the end of the
1144			 * active list to ensure fairness when
1145			 * multiple transports are active. If this was
1146			 * the last queued request, svc_getreq will
1147			 * end up calling xprt_inactive to remove from
1148			 * the active list.
1149			 */
1150			xprt->xp_thread = NULL;
1151			st->st_xprt = NULL;
1152			if (xprt->xp_active) {
1153				xprt_assignthread(xprt);
1154				TAILQ_REMOVE(&pool->sp_active, xprt, xp_alink);
1155				TAILQ_INSERT_TAIL(&pool->sp_active, xprt,
1156				    xp_alink);
1157			}
1158			mtx_unlock(&pool->sp_lock);
1159			SVC_RELEASE(xprt);
1160			mtx_lock(&pool->sp_lock);
1161		}
1162
1163		/*
1164		 * Execute what we have queued.
1165		 */
1166		while ((rqstp = STAILQ_FIRST(&st->st_reqs)) != NULL) {
1167			size_t sz = rqstp->rq_size;
1168			mtx_unlock(&pool->sp_lock);
1169			svc_executereq(rqstp);
1170			mtx_lock(&pool->sp_lock);
1171			pool->sp_space_used -= sz;
1172		}
1173	}
1174
1175	if (st->st_xprt) {
1176		xprt = st->st_xprt;
1177		st->st_xprt = NULL;
1178		SVC_RELEASE(xprt);
1179	}
1180
1181	KASSERT(STAILQ_EMPTY(&st->st_reqs), ("stray reqs on exit"));
1182	LIST_REMOVE(st, st_link);
1183	pool->sp_threadcount--;
1184
1185	mtx_unlock(&pool->sp_lock);
1186
1187	cv_destroy(&st->st_cond);
1188	mem_free(st, sizeof(*st));
1189
1190	if (!ismaster)
1191		wakeup(pool);
1192}
1193
1194static void
1195svc_thread_start(void *arg)
1196{
1197
1198	svc_run_internal((SVCPOOL *) arg, FALSE);
1199	kthread_exit();
1200}
1201
1202static void
1203svc_new_thread(SVCPOOL *pool)
1204{
1205	struct thread *td;
1206
1207	pool->sp_threadcount++;
1208	kthread_add(svc_thread_start, pool,
1209	    pool->sp_proc, &td, 0, 0,
1210	    "%s: service", pool->sp_name);
1211}
1212
1213void
1214svc_run(SVCPOOL *pool)
1215{
1216	int i;
1217	struct proc *p;
1218	struct thread *td;
1219
1220	p = curproc;
1221	td = curthread;
1222	snprintf(td->td_name, sizeof(td->td_name),
1223	    "%s: master", pool->sp_name);
1224	pool->sp_state = SVCPOOL_ACTIVE;
1225	pool->sp_proc = p;
1226	pool->sp_lastcreatetime = time_uptime;
1227	pool->sp_threadcount = 1;
1228
1229	for (i = 1; i < pool->sp_minthreads; i++) {
1230		svc_new_thread(pool);
1231	}
1232
1233	svc_run_internal(pool, TRUE);
1234
1235	mtx_lock(&pool->sp_lock);
1236	while (pool->sp_threadcount > 0)
1237		msleep(pool, &pool->sp_lock, 0, "svcexit", 0);
1238	mtx_unlock(&pool->sp_lock);
1239}
1240
1241void
1242svc_exit(SVCPOOL *pool)
1243{
1244	SVCTHREAD *st;
1245
1246	mtx_lock(&pool->sp_lock);
1247
1248	pool->sp_state = SVCPOOL_CLOSING;
1249	LIST_FOREACH(st, &pool->sp_idlethreads, st_ilink)
1250		cv_signal(&st->st_cond);
1251
1252	mtx_unlock(&pool->sp_lock);
1253}
1254
1255bool_t
1256svc_getargs(struct svc_req *rqstp, xdrproc_t xargs, void *args)
1257{
1258	struct mbuf *m;
1259	XDR xdrs;
1260	bool_t stat;
1261
1262	m = rqstp->rq_args;
1263	rqstp->rq_args = NULL;
1264
1265	xdrmbuf_create(&xdrs, m, XDR_DECODE);
1266	stat = xargs(&xdrs, args);
1267	XDR_DESTROY(&xdrs);
1268
1269	return (stat);
1270}
1271
1272bool_t
1273svc_freeargs(struct svc_req *rqstp, xdrproc_t xargs, void *args)
1274{
1275	XDR xdrs;
1276
1277	if (rqstp->rq_addr) {
1278		free(rqstp->rq_addr, M_SONAME);
1279		rqstp->rq_addr = NULL;
1280	}
1281
1282	xdrs.x_op = XDR_FREE;
1283	return (xargs(&xdrs, args));
1284}
1285
1286void
1287svc_freereq(struct svc_req *rqstp)
1288{
1289	SVCTHREAD *st;
1290	SVCXPRT *xprt;
1291	SVCPOOL *pool;
1292
1293	st = rqstp->rq_thread;
1294	xprt = rqstp->rq_xprt;
1295	if (xprt)
1296		pool = xprt->xp_pool;
1297	else
1298		pool = NULL;
1299	if (st) {
1300		mtx_lock(&pool->sp_lock);
1301		KASSERT(rqstp == STAILQ_FIRST(&st->st_reqs),
1302		    ("Freeing request out of order"));
1303		STAILQ_REMOVE_HEAD(&st->st_reqs, rq_link);
1304		st->st_reqcount--;
1305		if (pool->sp_done)
1306			pool->sp_done(st, rqstp);
1307		mtx_unlock(&pool->sp_lock);
1308	}
1309
1310	if (rqstp->rq_auth.svc_ah_ops)
1311		SVCAUTH_RELEASE(&rqstp->rq_auth);
1312
1313	if (rqstp->rq_xprt) {
1314		SVC_RELEASE(rqstp->rq_xprt);
1315	}
1316
1317	if (rqstp->rq_addr)
1318		free(rqstp->rq_addr, M_SONAME);
1319
1320	if (rqstp->rq_args)
1321		m_freem(rqstp->rq_args);
1322
1323	free(rqstp, M_RPC);
1324}
1325