1 /* $NetBSD: svc.c,v 1.21 2000/07/06 03:10:35 christos Exp $ */
4 * Copyright (c) 2009, Sun Microsystems, Inc.
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.
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.
31 #if defined(LIBC_SCCS) && !defined(lint)
32 static char *sccsid2 = "@(#)svc.c 1.44 88/02/08 Copyr 1984 Sun Micro";
33 static char *sccsid = "@(#)svc.c 2.4 88/08/11 4.0 RPCSRC";
35 #include <sys/cdefs.h>
36 __FBSDID("$FreeBSD$");
39 * svc.c, Server-side remote procedure call interface.
41 * There are two sets of procedures here. The xprt routines are
42 * for handling transport handles. The svc routines handle the
43 * list of service routines.
45 * Copyright (C) 1984, Sun Microsystems, Inc.
48 #include <sys/param.h>
50 #include <sys/kernel.h>
51 #include <sys/kthread.h>
52 #include <sys/malloc.h>
54 #include <sys/mutex.h>
56 #include <sys/queue.h>
57 #include <sys/socketvar.h>
58 #include <sys/systm.h>
61 #include <sys/ucred.h>
64 #include <rpc/rpcb_clnt.h>
65 #include <rpc/replay.h>
67 #include <rpc/rpc_com.h>
69 #define SVC_VERSQUIET 0x0001 /* keep quiet about vers mismatch */
70 #define version_keepquiet(xp) (SVC_EXT(xp)->xp_flags & SVC_VERSQUIET)
72 static struct svc_callout *svc_find(SVCPOOL *pool, rpcprog_t, rpcvers_t,
74 static void svc_new_thread(SVCGROUP *grp);
75 static void xprt_unregister_locked(SVCXPRT *xprt);
76 static void svc_change_space_used(SVCPOOL *pool, long delta);
77 static bool_t svc_request_space_available(SVCPOOL *pool);
79 /* *************** SVCXPRT related stuff **************** */
81 static int svcpool_minthread_sysctl(SYSCTL_HANDLER_ARGS);
82 static int svcpool_maxthread_sysctl(SYSCTL_HANDLER_ARGS);
83 static int svcpool_threads_sysctl(SYSCTL_HANDLER_ARGS);
86 svcpool_create(const char *name, struct sysctl_oid_list *sysctl_base)
92 pool = malloc(sizeof(SVCPOOL), M_RPC, M_WAITOK|M_ZERO);
94 mtx_init(&pool->sp_lock, "sp_lock", NULL, MTX_DEF);
96 pool->sp_state = SVCPOOL_INIT;
98 TAILQ_INIT(&pool->sp_callouts);
99 TAILQ_INIT(&pool->sp_lcallouts);
100 pool->sp_minthreads = 1;
101 pool->sp_maxthreads = 1;
102 pool->sp_groupcount = 1;
103 for (g = 0; g < SVC_MAXGROUPS; g++) {
104 grp = &pool->sp_groups[g];
105 mtx_init(&grp->sg_lock, "sg_lock", NULL, MTX_DEF);
107 grp->sg_state = SVCPOOL_ACTIVE;
108 TAILQ_INIT(&grp->sg_xlist);
109 TAILQ_INIT(&grp->sg_active);
110 LIST_INIT(&grp->sg_idlethreads);
111 grp->sg_minthreads = 1;
112 grp->sg_maxthreads = 1;
116 * Don't use more than a quarter of mbuf clusters. Nota bene:
117 * nmbclusters is an int, but nmbclusters*MCLBYTES may overflow
118 * on LP64 architectures, so cast to u_long to avoid undefined
119 * behavior. (ILP32 architectures cannot have nmbclusters
120 * large enough to overflow for other reasons.)
122 pool->sp_space_high = (u_long)nmbclusters * MCLBYTES / 4;
123 pool->sp_space_low = (pool->sp_space_high / 3) * 2;
125 sysctl_ctx_init(&pool->sp_sysctl);
127 SYSCTL_ADD_PROC(&pool->sp_sysctl, sysctl_base, OID_AUTO,
128 "minthreads", CTLTYPE_INT | CTLFLAG_RW,
129 pool, 0, svcpool_minthread_sysctl, "I",
130 "Minimal number of threads");
131 SYSCTL_ADD_PROC(&pool->sp_sysctl, sysctl_base, OID_AUTO,
132 "maxthreads", CTLTYPE_INT | CTLFLAG_RW,
133 pool, 0, svcpool_maxthread_sysctl, "I",
134 "Maximal number of threads");
135 SYSCTL_ADD_PROC(&pool->sp_sysctl, sysctl_base, OID_AUTO,
136 "threads", CTLTYPE_INT | CTLFLAG_RD,
137 pool, 0, svcpool_threads_sysctl, "I",
138 "Current number of threads");
139 SYSCTL_ADD_INT(&pool->sp_sysctl, sysctl_base, OID_AUTO,
140 "groups", CTLFLAG_RD, &pool->sp_groupcount, 0,
141 "Number of thread groups");
143 SYSCTL_ADD_ULONG(&pool->sp_sysctl, sysctl_base, OID_AUTO,
144 "request_space_used", CTLFLAG_RD,
145 &pool->sp_space_used,
146 "Space in parsed but not handled requests.");
148 SYSCTL_ADD_ULONG(&pool->sp_sysctl, sysctl_base, OID_AUTO,
149 "request_space_used_highest", CTLFLAG_RD,
150 &pool->sp_space_used_highest,
151 "Highest space used since reboot.");
153 SYSCTL_ADD_ULONG(&pool->sp_sysctl, sysctl_base, OID_AUTO,
154 "request_space_high", CTLFLAG_RW,
155 &pool->sp_space_high,
156 "Maximum space in parsed but not handled requests.");
158 SYSCTL_ADD_ULONG(&pool->sp_sysctl, sysctl_base, OID_AUTO,
159 "request_space_low", CTLFLAG_RW,
161 "Low water mark for request space.");
163 SYSCTL_ADD_INT(&pool->sp_sysctl, sysctl_base, OID_AUTO,
164 "request_space_throttled", CTLFLAG_RD,
165 &pool->sp_space_throttled, 0,
166 "Whether nfs requests are currently throttled");
168 SYSCTL_ADD_INT(&pool->sp_sysctl, sysctl_base, OID_AUTO,
169 "request_space_throttle_count", CTLFLAG_RD,
170 &pool->sp_space_throttle_count, 0,
171 "Count of times throttling based on request space has occurred");
178 svcpool_destroy(SVCPOOL *pool)
181 SVCXPRT *xprt, *nxprt;
182 struct svc_callout *s;
183 struct svc_loss_callout *sl;
184 struct svcxprt_list cleanup;
187 TAILQ_INIT(&cleanup);
189 for (g = 0; g < SVC_MAXGROUPS; g++) {
190 grp = &pool->sp_groups[g];
191 mtx_lock(&grp->sg_lock);
192 while ((xprt = TAILQ_FIRST(&grp->sg_xlist)) != NULL) {
193 xprt_unregister_locked(xprt);
194 TAILQ_INSERT_TAIL(&cleanup, xprt, xp_link);
196 mtx_unlock(&grp->sg_lock);
198 TAILQ_FOREACH_SAFE(xprt, &cleanup, xp_link, nxprt) {
202 mtx_lock(&pool->sp_lock);
203 while ((s = TAILQ_FIRST(&pool->sp_callouts)) != NULL) {
204 mtx_unlock(&pool->sp_lock);
205 svc_unreg(pool, s->sc_prog, s->sc_vers);
206 mtx_lock(&pool->sp_lock);
208 while ((sl = TAILQ_FIRST(&pool->sp_lcallouts)) != NULL) {
209 mtx_unlock(&pool->sp_lock);
210 svc_loss_unreg(pool, sl->slc_dispatch);
211 mtx_lock(&pool->sp_lock);
213 mtx_unlock(&pool->sp_lock);
215 for (g = 0; g < SVC_MAXGROUPS; g++) {
216 grp = &pool->sp_groups[g];
217 mtx_destroy(&grp->sg_lock);
219 mtx_destroy(&pool->sp_lock);
222 replay_freecache(pool->sp_rcache);
224 sysctl_ctx_free(&pool->sp_sysctl);
229 * Sysctl handler to get the present thread count on a pool
232 svcpool_threads_sysctl(SYSCTL_HANDLER_ARGS)
235 int threads, error, g;
237 pool = oidp->oid_arg1;
239 mtx_lock(&pool->sp_lock);
240 for (g = 0; g < pool->sp_groupcount; g++)
241 threads += pool->sp_groups[g].sg_threadcount;
242 mtx_unlock(&pool->sp_lock);
243 error = sysctl_handle_int(oidp, &threads, 0, req);
248 * Sysctl handler to set the minimum thread count on a pool
251 svcpool_minthread_sysctl(SYSCTL_HANDLER_ARGS)
254 int newminthreads, error, g;
256 pool = oidp->oid_arg1;
257 newminthreads = pool->sp_minthreads;
258 error = sysctl_handle_int(oidp, &newminthreads, 0, req);
259 if (error == 0 && newminthreads != pool->sp_minthreads) {
260 if (newminthreads > pool->sp_maxthreads)
262 mtx_lock(&pool->sp_lock);
263 pool->sp_minthreads = newminthreads;
264 for (g = 0; g < pool->sp_groupcount; g++) {
265 pool->sp_groups[g].sg_minthreads = max(1,
266 pool->sp_minthreads / pool->sp_groupcount);
268 mtx_unlock(&pool->sp_lock);
274 * Sysctl handler to set the maximum thread count on a pool
277 svcpool_maxthread_sysctl(SYSCTL_HANDLER_ARGS)
280 int newmaxthreads, error, g;
282 pool = oidp->oid_arg1;
283 newmaxthreads = pool->sp_maxthreads;
284 error = sysctl_handle_int(oidp, &newmaxthreads, 0, req);
285 if (error == 0 && newmaxthreads != pool->sp_maxthreads) {
286 if (newmaxthreads < pool->sp_minthreads)
288 mtx_lock(&pool->sp_lock);
289 pool->sp_maxthreads = newmaxthreads;
290 for (g = 0; g < pool->sp_groupcount; g++) {
291 pool->sp_groups[g].sg_maxthreads = max(1,
292 pool->sp_maxthreads / pool->sp_groupcount);
294 mtx_unlock(&pool->sp_lock);
300 * Activate a transport handle.
303 xprt_register(SVCXPRT *xprt)
305 SVCPOOL *pool = xprt->xp_pool;
310 g = atomic_fetchadd_int(&pool->sp_nextgroup, 1) % pool->sp_groupcount;
311 xprt->xp_group = grp = &pool->sp_groups[g];
312 mtx_lock(&grp->sg_lock);
313 xprt->xp_registered = TRUE;
314 xprt->xp_active = FALSE;
315 TAILQ_INSERT_TAIL(&grp->sg_xlist, xprt, xp_link);
316 mtx_unlock(&grp->sg_lock);
320 * De-activate a transport handle. Note: the locked version doesn't
321 * release the transport - caller must do that after dropping the pool
325 xprt_unregister_locked(SVCXPRT *xprt)
327 SVCGROUP *grp = xprt->xp_group;
329 mtx_assert(&grp->sg_lock, MA_OWNED);
330 KASSERT(xprt->xp_registered == TRUE,
331 ("xprt_unregister_locked: not registered"));
332 xprt_inactive_locked(xprt);
333 TAILQ_REMOVE(&grp->sg_xlist, xprt, xp_link);
334 xprt->xp_registered = FALSE;
338 xprt_unregister(SVCXPRT *xprt)
340 SVCGROUP *grp = xprt->xp_group;
342 mtx_lock(&grp->sg_lock);
343 if (xprt->xp_registered == FALSE) {
344 /* Already unregistered by another thread */
345 mtx_unlock(&grp->sg_lock);
348 xprt_unregister_locked(xprt);
349 mtx_unlock(&grp->sg_lock);
355 * Attempt to assign a service thread to this transport.
358 xprt_assignthread(SVCXPRT *xprt)
360 SVCGROUP *grp = xprt->xp_group;
363 mtx_assert(&grp->sg_lock, MA_OWNED);
364 st = LIST_FIRST(&grp->sg_idlethreads);
366 LIST_REMOVE(st, st_ilink);
368 xprt->xp_thread = st;
370 cv_signal(&st->st_cond);
374 * See if we can create a new thread. The
375 * actual thread creation happens in
376 * svc_run_internal because our locking state
377 * is poorly defined (we are typically called
378 * from a socket upcall). Don't create more
379 * than one thread per second.
381 if (grp->sg_state == SVCPOOL_ACTIVE
382 && grp->sg_lastcreatetime < time_uptime
383 && grp->sg_threadcount < grp->sg_maxthreads) {
384 grp->sg_state = SVCPOOL_THREADWANTED;
391 xprt_active(SVCXPRT *xprt)
393 SVCGROUP *grp = xprt->xp_group;
395 mtx_lock(&grp->sg_lock);
397 if (!xprt->xp_registered) {
399 * Race with xprt_unregister - we lose.
401 mtx_unlock(&grp->sg_lock);
405 if (!xprt->xp_active) {
406 xprt->xp_active = TRUE;
407 if (xprt->xp_thread == NULL) {
408 if (!svc_request_space_available(xprt->xp_pool) ||
409 !xprt_assignthread(xprt))
410 TAILQ_INSERT_TAIL(&grp->sg_active, xprt,
415 mtx_unlock(&grp->sg_lock);
419 xprt_inactive_locked(SVCXPRT *xprt)
421 SVCGROUP *grp = xprt->xp_group;
423 mtx_assert(&grp->sg_lock, MA_OWNED);
424 if (xprt->xp_active) {
425 if (xprt->xp_thread == NULL)
426 TAILQ_REMOVE(&grp->sg_active, xprt, xp_alink);
427 xprt->xp_active = FALSE;
432 xprt_inactive(SVCXPRT *xprt)
434 SVCGROUP *grp = xprt->xp_group;
436 mtx_lock(&grp->sg_lock);
437 xprt_inactive_locked(xprt);
438 mtx_unlock(&grp->sg_lock);
442 * Variant of xprt_inactive() for use only when sure that port is
443 * assigned to thread. For example, withing receive handlers.
446 xprt_inactive_self(SVCXPRT *xprt)
449 KASSERT(xprt->xp_thread != NULL,
450 ("xprt_inactive_self(%p) with NULL xp_thread", xprt));
451 xprt->xp_active = FALSE;
455 * Add a service program to the callout list.
456 * The dispatch routine will be called when a rpc request for this
457 * program number comes in.
460 svc_reg(SVCXPRT *xprt, const rpcprog_t prog, const rpcvers_t vers,
461 void (*dispatch)(struct svc_req *, SVCXPRT *),
462 const struct netconfig *nconf)
464 SVCPOOL *pool = xprt->xp_pool;
465 struct svc_callout *s;
469 /* VARIABLES PROTECTED BY svc_lock: s, svc_head */
471 if (xprt->xp_netid) {
472 netid = strdup(xprt->xp_netid, M_RPC);
474 } else if (nconf && nconf->nc_netid) {
475 netid = strdup(nconf->nc_netid, M_RPC);
477 } /* must have been created with svc_raw_create */
478 if ((netid == NULL) && (flag == 1)) {
482 mtx_lock(&pool->sp_lock);
483 if ((s = svc_find(pool, prog, vers, netid)) != NULL) {
486 if (s->sc_dispatch == dispatch)
487 goto rpcb_it; /* he is registering another xptr */
488 mtx_unlock(&pool->sp_lock);
491 s = malloc(sizeof (struct svc_callout), M_RPC, M_NOWAIT);
495 mtx_unlock(&pool->sp_lock);
501 s->sc_dispatch = dispatch;
503 TAILQ_INSERT_TAIL(&pool->sp_callouts, s, sc_link);
505 if ((xprt->xp_netid == NULL) && (flag == 1) && netid)
506 ((SVCXPRT *) xprt)->xp_netid = strdup(netid, M_RPC);
509 mtx_unlock(&pool->sp_lock);
510 /* now register the information with the local binder service */
513 struct netconfig tnc;
516 nb.buf = &xprt->xp_ltaddr;
517 nb.len = xprt->xp_ltaddr.ss_len;
518 dummy = rpcb_set(prog, vers, &tnc, &nb);
525 * Remove a service program from the callout list.
528 svc_unreg(SVCPOOL *pool, const rpcprog_t prog, const rpcvers_t vers)
530 struct svc_callout *s;
532 /* unregister the information anyway */
533 (void) rpcb_unset(prog, vers, NULL);
534 mtx_lock(&pool->sp_lock);
535 while ((s = svc_find(pool, prog, vers, NULL)) != NULL) {
536 TAILQ_REMOVE(&pool->sp_callouts, s, sc_link);
538 mem_free(s->sc_netid, sizeof (s->sc_netid) + 1);
539 mem_free(s, sizeof (struct svc_callout));
541 mtx_unlock(&pool->sp_lock);
545 * Add a service connection loss program to the callout list.
546 * The dispatch routine will be called when some port in ths pool die.
549 svc_loss_reg(SVCXPRT *xprt, void (*dispatch)(SVCXPRT *))
551 SVCPOOL *pool = xprt->xp_pool;
552 struct svc_loss_callout *s;
554 mtx_lock(&pool->sp_lock);
555 TAILQ_FOREACH(s, &pool->sp_lcallouts, slc_link) {
556 if (s->slc_dispatch == dispatch)
560 mtx_unlock(&pool->sp_lock);
563 s = malloc(sizeof (struct svc_callout), M_RPC, M_NOWAIT);
565 mtx_unlock(&pool->sp_lock);
568 s->slc_dispatch = dispatch;
569 TAILQ_INSERT_TAIL(&pool->sp_lcallouts, s, slc_link);
570 mtx_unlock(&pool->sp_lock);
575 * Remove a service connection loss program from the callout list.
578 svc_loss_unreg(SVCPOOL *pool, void (*dispatch)(SVCXPRT *))
580 struct svc_loss_callout *s;
582 mtx_lock(&pool->sp_lock);
583 TAILQ_FOREACH(s, &pool->sp_lcallouts, slc_link) {
584 if (s->slc_dispatch == dispatch) {
585 TAILQ_REMOVE(&pool->sp_lcallouts, s, slc_link);
590 mtx_unlock(&pool->sp_lock);
593 /* ********************** CALLOUT list related stuff ************* */
596 * Search the callout list for a program number, return the callout
599 static struct svc_callout *
600 svc_find(SVCPOOL *pool, rpcprog_t prog, rpcvers_t vers, char *netid)
602 struct svc_callout *s;
604 mtx_assert(&pool->sp_lock, MA_OWNED);
605 TAILQ_FOREACH(s, &pool->sp_callouts, sc_link) {
606 if (s->sc_prog == prog && s->sc_vers == vers
607 && (netid == NULL || s->sc_netid == NULL ||
608 strcmp(netid, s->sc_netid) == 0))
615 /* ******************* REPLY GENERATION ROUTINES ************ */
618 svc_sendreply_common(struct svc_req *rqstp, struct rpc_msg *rply,
621 SVCXPRT *xprt = rqstp->rq_xprt;
624 if (rqstp->rq_args) {
625 m_freem(rqstp->rq_args);
626 rqstp->rq_args = NULL;
629 if (xprt->xp_pool->sp_rcache)
630 replay_setreply(xprt->xp_pool->sp_rcache,
631 rply, svc_getrpccaller(rqstp), body);
633 if (!SVCAUTH_WRAP(&rqstp->rq_auth, &body))
636 ok = SVC_REPLY(xprt, rply, rqstp->rq_addr, body, &rqstp->rq_reply_seq);
637 if (rqstp->rq_addr) {
638 free(rqstp->rq_addr, M_SONAME);
639 rqstp->rq_addr = NULL;
646 * Send a reply to an rpc request
649 svc_sendreply(struct svc_req *rqstp, xdrproc_t xdr_results, void * xdr_location)
656 rply.rm_xid = rqstp->rq_xid;
657 rply.rm_direction = REPLY;
658 rply.rm_reply.rp_stat = MSG_ACCEPTED;
659 rply.acpted_rply.ar_verf = rqstp->rq_verf;
660 rply.acpted_rply.ar_stat = SUCCESS;
661 rply.acpted_rply.ar_results.where = NULL;
662 rply.acpted_rply.ar_results.proc = (xdrproc_t) xdr_void;
664 m = m_getcl(M_WAITOK, MT_DATA, 0);
665 xdrmbuf_create(&xdrs, m, XDR_ENCODE);
666 ok = xdr_results(&xdrs, xdr_location);
670 return (svc_sendreply_common(rqstp, &rply, m));
678 svc_sendreply_mbuf(struct svc_req *rqstp, struct mbuf *m)
682 rply.rm_xid = rqstp->rq_xid;
683 rply.rm_direction = REPLY;
684 rply.rm_reply.rp_stat = MSG_ACCEPTED;
685 rply.acpted_rply.ar_verf = rqstp->rq_verf;
686 rply.acpted_rply.ar_stat = SUCCESS;
687 rply.acpted_rply.ar_results.where = NULL;
688 rply.acpted_rply.ar_results.proc = (xdrproc_t) xdr_void;
690 return (svc_sendreply_common(rqstp, &rply, m));
694 * No procedure error reply
697 svcerr_noproc(struct svc_req *rqstp)
699 SVCXPRT *xprt = rqstp->rq_xprt;
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 = PROC_UNAVAIL;
708 if (xprt->xp_pool->sp_rcache)
709 replay_setreply(xprt->xp_pool->sp_rcache,
710 &rply, svc_getrpccaller(rqstp), NULL);
712 svc_sendreply_common(rqstp, &rply, NULL);
716 * Can't decode args error reply
719 svcerr_decode(struct svc_req *rqstp)
721 SVCXPRT *xprt = rqstp->rq_xprt;
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 = GARBAGE_ARGS;
730 if (xprt->xp_pool->sp_rcache)
731 replay_setreply(xprt->xp_pool->sp_rcache,
732 &rply, (struct sockaddr *) &xprt->xp_rtaddr, NULL);
734 svc_sendreply_common(rqstp, &rply, NULL);
741 svcerr_systemerr(struct svc_req *rqstp)
743 SVCXPRT *xprt = rqstp->rq_xprt;
746 rply.rm_xid = rqstp->rq_xid;
747 rply.rm_direction = REPLY;
748 rply.rm_reply.rp_stat = MSG_ACCEPTED;
749 rply.acpted_rply.ar_verf = rqstp->rq_verf;
750 rply.acpted_rply.ar_stat = SYSTEM_ERR;
752 if (xprt->xp_pool->sp_rcache)
753 replay_setreply(xprt->xp_pool->sp_rcache,
754 &rply, svc_getrpccaller(rqstp), NULL);
756 svc_sendreply_common(rqstp, &rply, NULL);
760 * Authentication error reply
763 svcerr_auth(struct svc_req *rqstp, enum auth_stat why)
765 SVCXPRT *xprt = rqstp->rq_xprt;
768 rply.rm_xid = rqstp->rq_xid;
769 rply.rm_direction = REPLY;
770 rply.rm_reply.rp_stat = MSG_DENIED;
771 rply.rjcted_rply.rj_stat = AUTH_ERROR;
772 rply.rjcted_rply.rj_why = why;
774 if (xprt->xp_pool->sp_rcache)
775 replay_setreply(xprt->xp_pool->sp_rcache,
776 &rply, svc_getrpccaller(rqstp), NULL);
778 svc_sendreply_common(rqstp, &rply, NULL);
782 * Auth too weak error reply
785 svcerr_weakauth(struct svc_req *rqstp)
788 svcerr_auth(rqstp, AUTH_TOOWEAK);
792 * Program unavailable error reply
795 svcerr_noprog(struct svc_req *rqstp)
797 SVCXPRT *xprt = rqstp->rq_xprt;
800 rply.rm_xid = rqstp->rq_xid;
801 rply.rm_direction = REPLY;
802 rply.rm_reply.rp_stat = MSG_ACCEPTED;
803 rply.acpted_rply.ar_verf = rqstp->rq_verf;
804 rply.acpted_rply.ar_stat = PROG_UNAVAIL;
806 if (xprt->xp_pool->sp_rcache)
807 replay_setreply(xprt->xp_pool->sp_rcache,
808 &rply, svc_getrpccaller(rqstp), NULL);
810 svc_sendreply_common(rqstp, &rply, NULL);
814 * Program version mismatch error reply
817 svcerr_progvers(struct svc_req *rqstp, rpcvers_t low_vers, rpcvers_t high_vers)
819 SVCXPRT *xprt = rqstp->rq_xprt;
822 rply.rm_xid = rqstp->rq_xid;
823 rply.rm_direction = REPLY;
824 rply.rm_reply.rp_stat = MSG_ACCEPTED;
825 rply.acpted_rply.ar_verf = rqstp->rq_verf;
826 rply.acpted_rply.ar_stat = PROG_MISMATCH;
827 rply.acpted_rply.ar_vers.low = (uint32_t)low_vers;
828 rply.acpted_rply.ar_vers.high = (uint32_t)high_vers;
830 if (xprt->xp_pool->sp_rcache)
831 replay_setreply(xprt->xp_pool->sp_rcache,
832 &rply, svc_getrpccaller(rqstp), NULL);
834 svc_sendreply_common(rqstp, &rply, NULL);
838 * Allocate a new server transport structure. All fields are
839 * initialized to zero and xp_p3 is initialized to point at an
840 * extension structure to hold various flags and authentication
849 xprt = mem_alloc(sizeof(SVCXPRT));
850 memset(xprt, 0, sizeof(SVCXPRT));
851 ext = mem_alloc(sizeof(SVCXPRT_EXT));
852 memset(ext, 0, sizeof(SVCXPRT_EXT));
854 refcount_init(&xprt->xp_refs, 1);
860 * Free a server transport structure.
867 mem_free(xprt->xp_p3, sizeof(SVCXPRT_EXT));
868 mem_free(xprt, sizeof(SVCXPRT));
871 /* ******************* SERVER INPUT STUFF ******************* */
874 * Read RPC requests from a transport and queue them to be
875 * executed. We handle authentication and replay cache replies here.
876 * Actually dispatching the RPC is deferred till svc_executereq.
878 static enum xprt_stat
879 svc_getreq(SVCXPRT *xprt, struct svc_req **rqstp_ret)
881 SVCPOOL *pool = xprt->xp_pool;
885 struct svc_loss_callout *s;
888 /* now receive msgs from xprtprt (support batch calls) */
889 r = malloc(sizeof(*r), M_RPC, M_WAITOK|M_ZERO);
891 msg.rm_call.cb_cred.oa_base = r->rq_credarea;
892 msg.rm_call.cb_verf.oa_base = &r->rq_credarea[MAX_AUTH_BYTES];
893 r->rq_clntcred = &r->rq_credarea[2*MAX_AUTH_BYTES];
894 if (SVC_RECV(xprt, &msg, &r->rq_addr, &args)) {
898 * Handle replays and authenticate before queuing the
899 * request to be executed.
903 if (pool->sp_rcache) {
904 struct rpc_msg repmsg;
905 struct mbuf *repbody;
906 enum replay_state rs;
907 rs = replay_find(pool->sp_rcache, &msg,
908 svc_getrpccaller(r), &repmsg, &repbody);
913 SVC_REPLY(xprt, &repmsg, r->rq_addr,
914 repbody, &r->rq_reply_seq);
916 free(r->rq_addr, M_SONAME);
928 r->rq_xid = msg.rm_xid;
929 r->rq_prog = msg.rm_call.cb_prog;
930 r->rq_vers = msg.rm_call.cb_vers;
931 r->rq_proc = msg.rm_call.cb_proc;
932 r->rq_size = sizeof(*r) + m_length(args, NULL);
934 if ((why = _authenticate(r, &msg)) != AUTH_OK) {
936 * RPCSEC_GSS uses this return code
937 * for requests that form part of its
938 * context establishment protocol and
939 * should not be dispatched to the
942 if (why != RPCSEC_GSS_NODISPATCH)
947 if (!SVCAUTH_UNWRAP(&r->rq_auth, &r->rq_args)) {
953 * Everything checks out, return request to caller.
963 if ((stat = SVC_STAT(xprt)) == XPRT_DIED) {
964 TAILQ_FOREACH(s, &pool->sp_lcallouts, slc_link)
965 (*s->slc_dispatch)(xprt);
966 xprt_unregister(xprt);
973 svc_executereq(struct svc_req *rqstp)
975 SVCXPRT *xprt = rqstp->rq_xprt;
976 SVCPOOL *pool = xprt->xp_pool;
980 struct svc_callout *s;
982 /* now match message with a registered service*/
984 low_vers = (rpcvers_t) -1L;
985 high_vers = (rpcvers_t) 0L;
986 TAILQ_FOREACH(s, &pool->sp_callouts, sc_link) {
987 if (s->sc_prog == rqstp->rq_prog) {
988 if (s->sc_vers == rqstp->rq_vers) {
990 * We hand ownership of r to the
991 * dispatch method - they must call
994 (*s->sc_dispatch)(rqstp, xprt);
996 } /* found correct version */
998 if (s->sc_vers < low_vers)
999 low_vers = s->sc_vers;
1000 if (s->sc_vers > high_vers)
1001 high_vers = s->sc_vers;
1002 } /* found correct program */
1006 * if we got here, the program or version
1010 svcerr_progvers(rqstp, low_vers, high_vers);
1012 svcerr_noprog(rqstp);
1018 svc_checkidle(SVCGROUP *grp)
1020 SVCXPRT *xprt, *nxprt;
1022 struct svcxprt_list cleanup;
1024 TAILQ_INIT(&cleanup);
1025 TAILQ_FOREACH_SAFE(xprt, &grp->sg_xlist, xp_link, nxprt) {
1027 * Only some transports have idle timers. Don't time
1028 * something out which is just waking up.
1030 if (!xprt->xp_idletimeout || xprt->xp_thread)
1033 timo = xprt->xp_lastactive + xprt->xp_idletimeout;
1034 if (time_uptime > timo) {
1035 xprt_unregister_locked(xprt);
1036 TAILQ_INSERT_TAIL(&cleanup, xprt, xp_link);
1040 mtx_unlock(&grp->sg_lock);
1041 TAILQ_FOREACH_SAFE(xprt, &cleanup, xp_link, nxprt) {
1044 mtx_lock(&grp->sg_lock);
1048 svc_assign_waiting_sockets(SVCPOOL *pool)
1054 for (g = 0; g < pool->sp_groupcount; g++) {
1055 grp = &pool->sp_groups[g];
1056 mtx_lock(&grp->sg_lock);
1057 while ((xprt = TAILQ_FIRST(&grp->sg_active)) != NULL) {
1058 if (xprt_assignthread(xprt))
1059 TAILQ_REMOVE(&grp->sg_active, xprt, xp_alink);
1063 mtx_unlock(&grp->sg_lock);
1068 svc_change_space_used(SVCPOOL *pool, long delta)
1070 unsigned long value;
1072 value = atomic_fetchadd_long(&pool->sp_space_used, delta) + delta;
1074 if (value >= pool->sp_space_high && !pool->sp_space_throttled) {
1075 pool->sp_space_throttled = TRUE;
1076 pool->sp_space_throttle_count++;
1078 if (value > pool->sp_space_used_highest)
1079 pool->sp_space_used_highest = value;
1081 if (value < pool->sp_space_low && pool->sp_space_throttled) {
1082 pool->sp_space_throttled = FALSE;
1083 svc_assign_waiting_sockets(pool);
1089 svc_request_space_available(SVCPOOL *pool)
1092 if (pool->sp_space_throttled)
1098 svc_run_internal(SVCGROUP *grp, bool_t ismaster)
1100 SVCPOOL *pool = grp->sg_pool;
1101 SVCTHREAD *st, *stpref;
1103 enum xprt_stat stat;
1104 struct svc_req *rqstp;
1109 st = mem_alloc(sizeof(*st));
1110 mtx_init(&st->st_lock, "st_lock", NULL, MTX_DEF);
1113 STAILQ_INIT(&st->st_reqs);
1114 cv_init(&st->st_cond, "rpcsvc");
1116 mtx_lock(&grp->sg_lock);
1119 * If we are a new thread which was spawned to cope with
1120 * increased load, set the state back to SVCPOOL_ACTIVE.
1122 if (grp->sg_state == SVCPOOL_THREADSTARTING)
1123 grp->sg_state = SVCPOOL_ACTIVE;
1125 while (grp->sg_state != SVCPOOL_CLOSING) {
1127 * Create new thread if requested.
1129 if (grp->sg_state == SVCPOOL_THREADWANTED) {
1130 grp->sg_state = SVCPOOL_THREADSTARTING;
1131 grp->sg_lastcreatetime = time_uptime;
1132 mtx_unlock(&grp->sg_lock);
1133 svc_new_thread(grp);
1134 mtx_lock(&grp->sg_lock);
1139 * Check for idle transports once per second.
1141 if (time_uptime > grp->sg_lastidlecheck) {
1142 grp->sg_lastidlecheck = time_uptime;
1149 * Enforce maxthreads count.
1151 if (grp->sg_threadcount > grp->sg_maxthreads)
1155 * Before sleeping, see if we can find an
1156 * active transport which isn't being serviced
1159 if (svc_request_space_available(pool) &&
1160 (xprt = TAILQ_FIRST(&grp->sg_active)) != NULL) {
1161 TAILQ_REMOVE(&grp->sg_active, xprt, xp_alink);
1163 xprt->xp_thread = st;
1168 LIST_INSERT_HEAD(&grp->sg_idlethreads, st, st_ilink);
1169 if (ismaster || (!ismaster &&
1170 grp->sg_threadcount > grp->sg_minthreads))
1171 error = cv_timedwait_sig(&st->st_cond,
1172 &grp->sg_lock, 5 * hz);
1174 error = cv_wait_sig(&st->st_cond,
1176 if (st->st_xprt == NULL)
1177 LIST_REMOVE(st, st_ilink);
1180 * Reduce worker thread count when idle.
1182 if (error == EWOULDBLOCK) {
1184 && (grp->sg_threadcount
1185 > grp->sg_minthreads)
1188 } else if (error != 0) {
1189 KASSERT(error == EINTR || error == ERESTART,
1190 ("non-signal error %d", error));
1191 mtx_unlock(&grp->sg_lock);
1194 if (P_SHOULDSTOP(p) ||
1195 (p->p_flag & P_TOTAL_STOP) != 0) {
1196 thread_suspend_check(0);
1198 mtx_lock(&grp->sg_lock);
1202 mtx_lock(&grp->sg_lock);
1208 mtx_unlock(&grp->sg_lock);
1211 * Drain the transport socket and queue up any RPCs.
1213 xprt->xp_lastactive = time_uptime;
1215 if (!svc_request_space_available(pool))
1218 stat = svc_getreq(xprt, &rqstp);
1220 svc_change_space_used(pool, rqstp->rq_size);
1222 * See if the application has a preference
1223 * for some other thread.
1225 if (pool->sp_assign) {
1226 stpref = pool->sp_assign(st, rqstp);
1227 rqstp->rq_thread = stpref;
1228 STAILQ_INSERT_TAIL(&stpref->st_reqs,
1230 mtx_unlock(&stpref->st_lock);
1234 rqstp->rq_thread = st;
1235 STAILQ_INSERT_TAIL(&st->st_reqs,
1239 } while (rqstp == NULL && stat == XPRT_MOREREQS
1240 && grp->sg_state != SVCPOOL_CLOSING);
1243 * Move this transport to the end of the active list to
1244 * ensure fairness when multiple transports are active.
1245 * If this was the last queued request, svc_getreq will end
1246 * up calling xprt_inactive to remove from the active list.
1248 mtx_lock(&grp->sg_lock);
1249 xprt->xp_thread = NULL;
1251 if (xprt->xp_active) {
1252 if (!svc_request_space_available(pool) ||
1253 !xprt_assignthread(xprt))
1254 TAILQ_INSERT_TAIL(&grp->sg_active,
1257 mtx_unlock(&grp->sg_lock);
1261 * Execute what we have queued.
1263 mtx_lock(&st->st_lock);
1264 while ((rqstp = STAILQ_FIRST(&st->st_reqs)) != NULL) {
1265 STAILQ_REMOVE_HEAD(&st->st_reqs, rq_link);
1266 mtx_unlock(&st->st_lock);
1267 sz = (long)rqstp->rq_size;
1268 svc_executereq(rqstp);
1269 svc_change_space_used(pool, -sz);
1270 mtx_lock(&st->st_lock);
1272 mtx_unlock(&st->st_lock);
1273 mtx_lock(&grp->sg_lock);
1281 KASSERT(STAILQ_EMPTY(&st->st_reqs), ("stray reqs on exit"));
1282 mtx_destroy(&st->st_lock);
1283 cv_destroy(&st->st_cond);
1284 mem_free(st, sizeof(*st));
1286 grp->sg_threadcount--;
1289 mtx_unlock(&grp->sg_lock);
1293 svc_thread_start(void *arg)
1296 svc_run_internal((SVCGROUP *) arg, FALSE);
1301 svc_new_thread(SVCGROUP *grp)
1303 SVCPOOL *pool = grp->sg_pool;
1306 grp->sg_threadcount++;
1307 kthread_add(svc_thread_start, grp, pool->sp_proc, &td, 0, 0,
1308 "%s: service", pool->sp_name);
1312 svc_run(SVCPOOL *pool)
1321 snprintf(td->td_name, sizeof(td->td_name),
1322 "%s: master", pool->sp_name);
1323 pool->sp_state = SVCPOOL_ACTIVE;
1326 /* Choose group count based on number of threads and CPUs. */
1327 pool->sp_groupcount = max(1, min(SVC_MAXGROUPS,
1328 min(pool->sp_maxthreads / 2, mp_ncpus) / 6));
1329 for (g = 0; g < pool->sp_groupcount; g++) {
1330 grp = &pool->sp_groups[g];
1331 grp->sg_minthreads = max(1,
1332 pool->sp_minthreads / pool->sp_groupcount);
1333 grp->sg_maxthreads = max(1,
1334 pool->sp_maxthreads / pool->sp_groupcount);
1335 grp->sg_lastcreatetime = time_uptime;
1338 /* Starting threads */
1339 for (g = 0; g < pool->sp_groupcount; g++) {
1340 grp = &pool->sp_groups[g];
1341 for (i = ((g == 0) ? 1 : 0); i < grp->sg_minthreads; i++)
1342 svc_new_thread(grp);
1344 pool->sp_groups[0].sg_threadcount++;
1345 svc_run_internal(&pool->sp_groups[0], TRUE);
1347 /* Waiting for threads to stop. */
1348 for (g = 0; g < pool->sp_groupcount; g++) {
1349 grp = &pool->sp_groups[g];
1350 mtx_lock(&grp->sg_lock);
1351 while (grp->sg_threadcount > 0)
1352 msleep(grp, &grp->sg_lock, 0, "svcexit", 0);
1353 mtx_unlock(&grp->sg_lock);
1358 svc_exit(SVCPOOL *pool)
1364 pool->sp_state = SVCPOOL_CLOSING;
1365 for (g = 0; g < pool->sp_groupcount; g++) {
1366 grp = &pool->sp_groups[g];
1367 mtx_lock(&grp->sg_lock);
1368 if (grp->sg_state != SVCPOOL_CLOSING) {
1369 grp->sg_state = SVCPOOL_CLOSING;
1370 LIST_FOREACH(st, &grp->sg_idlethreads, st_ilink)
1371 cv_signal(&st->st_cond);
1373 mtx_unlock(&grp->sg_lock);
1378 svc_getargs(struct svc_req *rqstp, xdrproc_t xargs, void *args)
1385 rqstp->rq_args = NULL;
1387 xdrmbuf_create(&xdrs, m, XDR_DECODE);
1388 stat = xargs(&xdrs, args);
1395 svc_freeargs(struct svc_req *rqstp, xdrproc_t xargs, void *args)
1399 if (rqstp->rq_addr) {
1400 free(rqstp->rq_addr, M_SONAME);
1401 rqstp->rq_addr = NULL;
1404 xdrs.x_op = XDR_FREE;
1405 return (xargs(&xdrs, args));
1409 svc_freereq(struct svc_req *rqstp)
1414 st = rqstp->rq_thread;
1418 pool->sp_done(st, rqstp);
1421 if (rqstp->rq_auth.svc_ah_ops)
1422 SVCAUTH_RELEASE(&rqstp->rq_auth);
1424 if (rqstp->rq_xprt) {
1425 SVC_RELEASE(rqstp->rq_xprt);
1429 free(rqstp->rq_addr, M_SONAME);
1432 m_freem(rqstp->rq_args);