sys/rpc/svc.c

   1 /*      $NetBSD: svc.c,v 1.21 2000/07/06 03:10:35 christos Exp $        */
   2
   3 /*-
   4  * Copyright (c) 2009, Sun Microsystems, Inc.
   5  * All rights reserved.
   6  *
   7  * Redistribution and use in source and binary forms, with or without
   8  * modification, are permitted provided that the following conditions are met:
   9  * - Redistributions of source code must retain the above copyright notice,
  10  *   this list of conditions and the following disclaimer.
  11  * - Redistributions in binary form must reproduce the above copyright notice,
  12  *   this list of conditions and the following disclaimer in the documentation
  13  *   and/or other materials provided with the distribution.
  14  * - Neither the name of Sun Microsystems, Inc. nor the names of its
  15  *   contributors may be used to endorse or promote products derived
  16  *   from this software without specific prior written permission.
  17  *
  18  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
  19  * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  20  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  21  * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
  22  * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  23  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  24  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
  25  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
  26  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
  27  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
  28  * POSSIBILITY OF SUCH DAMAGE.
  29  */
  30
  31 #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";
  34 #endif
  35 #include <sys/cdefs.h>
  36 __FBSDID("$FreeBSD$");
  37
  38 /*
  39  * svc.c, Server-side remote procedure call interface.
  40  *
  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.
  44  *
  45  * Copyright (C) 1984, Sun Microsystems, Inc.
  46  */
  47
  48 #include <sys/param.h>
  49 #include <sys/lock.h>
  50 #include <sys/kernel.h>
  51 #include <sys/kthread.h>
  52 #include <sys/malloc.h>
  53 #include <sys/mbuf.h>
  54 #include <sys/mutex.h>
  55 #include <sys/proc.h>
  56 #include <sys/queue.h>
  57 #include <sys/socketvar.h>
  58 #include <sys/systm.h>
  59 #include <sys/smp.h>
  60 #include <sys/sx.h>
  61 #include <sys/ucred.h>
  62
  63 #include <rpc/rpc.h>
  64 #include <rpc/rpcb_clnt.h>
  65 #include <rpc/replay.h>
  66
  67 #include <rpc/rpc_com.h>
  68
  69 #define SVC_VERSQUIET 0x0001            /* keep quiet about vers mismatch */
  70 #define version_keepquiet(xp) (SVC_EXT(xp)->xp_flags & SVC_VERSQUIET)
  71
  72 static struct svc_callout *svc_find(SVCPOOL *pool, rpcprog_t, rpcvers_t,
  73     char *);
  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);
  78
  79 /* ***************  SVCXPRT related stuff **************** */
  80
  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);
  84
  85 SVCPOOL*
  86 svcpool_create(const char *name, struct sysctl_oid_list *sysctl_base)
  87 {
  88         SVCPOOL *pool;
  89         SVCGROUP *grp;
  90         int g;
  91
  92         pool = malloc(sizeof(SVCPOOL), M_RPC, M_WAITOK|M_ZERO);
  93
  94         mtx_init(&pool->sp_lock, "sp_lock", NULL, MTX_DEF);
  95         pool->sp_name = name;
  96         pool->sp_state = SVCPOOL_INIT;
  97         pool->sp_proc = NULL;
  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);
 106                 grp->sg_pool = pool;
 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;
 113         }
 114
 115         /*
 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.)
 121          */
 122         pool->sp_space_high = (u_long)nmbclusters * MCLBYTES / 4;
 123         pool->sp_space_low = (pool->sp_space_high / 3) * 2;
 124
 125         sysctl_ctx_init(&pool->sp_sysctl);
 126         if (sysctl_base) {
 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");
 142
 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.");
 147
 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.");
 152
 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.");
 157
 158                 SYSCTL_ADD_ULONG(&pool->sp_sysctl, sysctl_base, OID_AUTO,
 159                     "request_space_low", CTLFLAG_RW,
 160                     &pool->sp_space_low,
 161                     "Low water mark for request space.");
 162
 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");
 167
 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");
 172         }
 173
 174         return pool;
 175 }
 176
 177 void
 178 svcpool_destroy(SVCPOOL *pool)
 179 {
 180         SVCGROUP *grp;
 181         SVCXPRT *xprt, *nxprt;
 182         struct svc_callout *s;
 183         struct svc_loss_callout *sl;
 184         struct svcxprt_list cleanup;
 185         int g;
 186
 187         TAILQ_INIT(&cleanup);
 188
 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);
 195                 }
 196                 mtx_unlock(&grp->sg_lock);
 197         }
 198         TAILQ_FOREACH_SAFE(xprt, &cleanup, xp_link, nxprt) {
 199                 SVC_RELEASE(xprt);
 200         }
 201
 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);
 207         }
 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);
 212         }
 213         mtx_unlock(&pool->sp_lock);
 214
 215         for (g = 0; g < SVC_MAXGROUPS; g++) {
 216                 grp = &pool->sp_groups[g];
 217                 mtx_destroy(&grp->sg_lock);
 218         }
 219         mtx_destroy(&pool->sp_lock);
 220
 221         if (pool->sp_rcache)
 222                 replay_freecache(pool->sp_rcache);
 223
 224         sysctl_ctx_free(&pool->sp_sysctl);
 225         free(pool, M_RPC);
 226 }
 227
 228 /*
 229  * Sysctl handler to get the present thread count on a pool
 230  */
 231 static int
 232 svcpool_threads_sysctl(SYSCTL_HANDLER_ARGS)
 233 {
 234         SVCPOOL *pool;
 235         int threads, error, g;
 236
 237         pool = oidp->oid_arg1;
 238         threads = 0;
 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);
 244         return (error);
 245 }
 246
 247 /*
 248  * Sysctl handler to set the minimum thread count on a pool
 249  */
 250 static int
 251 svcpool_minthread_sysctl(SYSCTL_HANDLER_ARGS)
 252 {
 253         SVCPOOL *pool;
 254         int newminthreads, error, g;
 255
 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)
 261                         return (EINVAL);
 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);
 267                 }
 268                 mtx_unlock(&pool->sp_lock);
 269         }
 270         return (error);
 271 }
 272
 273 /*
 274  * Sysctl handler to set the maximum thread count on a pool
 275  */
 276 static int
 277 svcpool_maxthread_sysctl(SYSCTL_HANDLER_ARGS)
 278 {
 279         SVCPOOL *pool;
 280         int newmaxthreads, error, g;
 281
 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)
 287                         return (EINVAL);
 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);
 293                 }
 294                 mtx_unlock(&pool->sp_lock);
 295         }
 296         return (error);
 297 }
 298
 299 /*
 300  * Activate a transport handle.
 301  */
 302 void
 303 xprt_register(SVCXPRT *xprt)
 304 {
 305         SVCPOOL *pool = xprt->xp_pool;
 306         SVCGROUP *grp;
 307         int g;
 308
 309         SVC_ACQUIRE(xprt);
 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);
 317 }
 318
 319 /*
 320  * De-activate a transport handle. Note: the locked version doesn't
 321  * release the transport - caller must do that after dropping the pool
 322  * lock.
 323  */
 324 static void
 325 xprt_unregister_locked(SVCXPRT *xprt)
 326 {
 327         SVCGROUP *grp = xprt->xp_group;
 328
 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;
 335 }
 336
 337 void
 338 xprt_unregister(SVCXPRT *xprt)
 339 {
 340         SVCGROUP *grp = xprt->xp_group;
 341
 342         mtx_lock(&grp->sg_lock);
 343         if (xprt->xp_registered == FALSE) {
 344                 /* Already unregistered by another thread */
 345                 mtx_unlock(&grp->sg_lock);
 346                 return;
 347         }
 348         xprt_unregister_locked(xprt);
 349         mtx_unlock(&grp->sg_lock);
 350
 351         SVC_RELEASE(xprt);
 352 }
 353
 354 /*
 355  * Attempt to assign a service thread to this transport.
 356  */
 357 static int
 358 xprt_assignthread(SVCXPRT *xprt)
 359 {
 360         SVCGROUP *grp = xprt->xp_group;
 361         SVCTHREAD *st;
 362
 363         mtx_assert(&grp->sg_lock, MA_OWNED);
 364         st = LIST_FIRST(&grp->sg_idlethreads);
 365         if (st) {
 366                 LIST_REMOVE(st, st_ilink);
 367                 SVC_ACQUIRE(xprt);
 368                 xprt->xp_thread = st;
 369                 st->st_xprt = xprt;
 370                 cv_signal(&st->st_cond);
 371                 return (TRUE);
 372         } else {
 373                 /*
 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.
 380                  */
 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;
 385                 }
 386         }
 387         return (FALSE);
 388 }
 389
 390 void
 391 xprt_active(SVCXPRT *xprt)
 392 {
 393         SVCGROUP *grp = xprt->xp_group;
 394
 395         mtx_lock(&grp->sg_lock);
 396
 397         if (!xprt->xp_registered) {
 398                 /*
 399                  * Race with xprt_unregister - we lose.
 400                  */
 401                 mtx_unlock(&grp->sg_lock);
 402                 return;
 403         }
 404
 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,
 411                                     xp_alink);
 412                 }
 413         }
 414
 415         mtx_unlock(&grp->sg_lock);
 416 }
 417
 418 void
 419 xprt_inactive_locked(SVCXPRT *xprt)
 420 {
 421         SVCGROUP *grp = xprt->xp_group;
 422
 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;
 428         }
 429 }
 430
 431 void
 432 xprt_inactive(SVCXPRT *xprt)
 433 {
 434         SVCGROUP *grp = xprt->xp_group;
 435
 436         mtx_lock(&grp->sg_lock);
 437         xprt_inactive_locked(xprt);
 438         mtx_unlock(&grp->sg_lock);
 439 }
 440
 441 /*
 442  * Variant of xprt_inactive() for use only when sure that port is
 443  * assigned to thread. For example, withing receive handlers.
 444  */
 445 void
 446 xprt_inactive_self(SVCXPRT *xprt)
 447 {
 448
 449         KASSERT(xprt->xp_thread != NULL,
 450             ("xprt_inactive_self(%p) with NULL xp_thread", xprt));
 451         xprt->xp_active = FALSE;
 452 }
 453
 454 /*
 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.
 458  */
 459 bool_t
 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)
 463 {
 464         SVCPOOL *pool = xprt->xp_pool;
 465         struct svc_callout *s;
 466         char *netid = NULL;
 467         int flag = 0;
 468
 469 /* VARIABLES PROTECTED BY svc_lock: s, svc_head */
 470
 471         if (xprt->xp_netid) {
 472                 netid = strdup(xprt->xp_netid, M_RPC);
 473                 flag = 1;
 474         } else if (nconf && nconf->nc_netid) {
 475                 netid = strdup(nconf->nc_netid, M_RPC);
 476                 flag = 1;
 477         } /* must have been created with svc_raw_create */
 478         if ((netid == NULL) && (flag == 1)) {
 479                 return (FALSE);
 480         }
 481
 482         mtx_lock(&pool->sp_lock);
 483         if ((s = svc_find(pool, prog, vers, netid)) != NULL) {
 484                 if (netid)
 485                         free(netid, M_RPC);
 486                 if (s->sc_dispatch == dispatch)
 487                         goto rpcb_it; /* he is registering another xptr */
 488                 mtx_unlock(&pool->sp_lock);
 489                 return (FALSE);
 490         }
 491         s = malloc(sizeof (struct svc_callout), M_RPC, M_NOWAIT);
 492         if (s == NULL) {
 493                 if (netid)
 494                         free(netid, M_RPC);
 495                 mtx_unlock(&pool->sp_lock);
 496                 return (FALSE);
 497         }
 498
 499         s->sc_prog = prog;
 500         s->sc_vers = vers;
 501         s->sc_dispatch = dispatch;
 502         s->sc_netid = netid;
 503         TAILQ_INSERT_TAIL(&pool->sp_callouts, s, sc_link);
 504
 505         if ((xprt->xp_netid == NULL) && (flag == 1) && netid)
 506                 ((SVCXPRT *) xprt)->xp_netid = strdup(netid, M_RPC);
 507
 508 rpcb_it:
 509         mtx_unlock(&pool->sp_lock);
 510         /* now register the information with the local binder service */
 511         if (nconf) {
 512                 bool_t dummy;
 513                 struct netconfig tnc;
 514                 struct netbuf nb;
 515                 tnc = *nconf;
 516                 nb.buf = &xprt->xp_ltaddr;
 517                 nb.len = xprt->xp_ltaddr.ss_len;
 518                 dummy = rpcb_set(prog, vers, &tnc, &nb);
 519                 return (dummy);
 520         }
 521         return (TRUE);
 522 }
 523
 524 /*
 525  * Remove a service program from the callout list.
 526  */
 527 void
 528 svc_unreg(SVCPOOL *pool, const rpcprog_t prog, const rpcvers_t vers)
 529 {
 530         struct svc_callout *s;
 531
 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);
 537                 if (s->sc_netid)
 538                         mem_free(s->sc_netid, sizeof (s->sc_netid) + 1);
 539                 mem_free(s, sizeof (struct svc_callout));
 540         }
 541         mtx_unlock(&pool->sp_lock);
 542 }
 543
 544 /*
 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.
 547  */
 548 bool_t
 549 svc_loss_reg(SVCXPRT *xprt, void (*dispatch)(SVCXPRT *))
 550 {
 551         SVCPOOL *pool = xprt->xp_pool;
 552         struct svc_loss_callout *s;
 553
 554         mtx_lock(&pool->sp_lock);
 555         TAILQ_FOREACH(s, &pool->sp_lcallouts, slc_link) {
 556                 if (s->slc_dispatch == dispatch)
 557                         break;
 558         }
 559         if (s != NULL) {
 560                 mtx_unlock(&pool->sp_lock);
 561                 return (TRUE);
 562         }
 563         s = malloc(sizeof(struct svc_loss_callout), M_RPC, M_NOWAIT);
 564         if (s == NULL) {
 565                 mtx_unlock(&pool->sp_lock);
 566                 return (FALSE);
 567         }
 568         s->slc_dispatch = dispatch;
 569         TAILQ_INSERT_TAIL(&pool->sp_lcallouts, s, slc_link);
 570         mtx_unlock(&pool->sp_lock);
 571         return (TRUE);
 572 }
 573
 574 /*
 575  * Remove a service connection loss program from the callout list.
 576  */
 577 void
 578 svc_loss_unreg(SVCPOOL *pool, void (*dispatch)(SVCXPRT *))
 579 {
 580         struct svc_loss_callout *s;
 581
 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);
 586                         free(s, M_RPC);
 587                         break;
 588                 }
 589         }
 590         mtx_unlock(&pool->sp_lock);
 591 }
 592
 593 /* ********************** CALLOUT list related stuff ************* */
 594
 595 /*
 596  * Search the callout list for a program number, return the callout
 597  * struct.
 598  */
 599 static struct svc_callout *
 600 svc_find(SVCPOOL *pool, rpcprog_t prog, rpcvers_t vers, char *netid)
 601 {
 602         struct svc_callout *s;
 603
 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))
 609                         break;
 610         }
 611
 612         return (s);
 613 }
 614
 615 /* ******************* REPLY GENERATION ROUTINES  ************ */
 616
 617 static bool_t
 618 svc_sendreply_common(struct svc_req *rqstp, struct rpc_msg *rply,
 619     struct mbuf *body)
 620 {
 621         SVCXPRT *xprt = rqstp->rq_xprt;
 622         bool_t ok;
 623
 624         if (rqstp->rq_args) {
 625                 m_freem(rqstp->rq_args);
 626                 rqstp->rq_args = NULL;
 627         }
 628
 629         if (xprt->xp_pool->sp_rcache)
 630                 replay_setreply(xprt->xp_pool->sp_rcache,
 631                     rply, svc_getrpccaller(rqstp), body);
 632
 633         if (!SVCAUTH_WRAP(&rqstp->rq_auth, &body))
 634                 return (FALSE);
 635
 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;
 640         }
 641
 642         return (ok);
 643 }
 644
 645 /*
 646  * Send a reply to an rpc request
 647  */
 648 bool_t
 649 svc_sendreply(struct svc_req *rqstp, xdrproc_t xdr_results, void * xdr_location)
 650 {
 651         struct rpc_msg rply;
 652         struct mbuf *m;
 653         XDR xdrs;
 654         bool_t ok;
 655
 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;
 663
 664         m = m_getcl(M_WAITOK, MT_DATA, 0);
 665         xdrmbuf_create(&xdrs, m, XDR_ENCODE);
 666         ok = xdr_results(&xdrs, xdr_location);
 667         XDR_DESTROY(&xdrs);
 668
 669         if (ok) {
 670                 return (svc_sendreply_common(rqstp, &rply, m));
 671         } else {
 672                 m_freem(m);
 673                 return (FALSE);
 674         }
 675 }
 676
 677 bool_t
 678 svc_sendreply_mbuf(struct svc_req *rqstp, struct mbuf *m)
 679 {
 680         struct rpc_msg rply;
 681
 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;
 689
 690         return (svc_sendreply_common(rqstp, &rply, m));
 691 }
 692
 693 /*
 694  * No procedure error reply
 695  */
 696 void
 697 svcerr_noproc(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 = PROC_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  * Can't decode args error reply
 717  */
 718 void
 719 svcerr_decode(struct svc_req *rqstp)
 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 = GARBAGE_ARGS;
 729
 730         if (xprt->xp_pool->sp_rcache)
 731                 replay_setreply(xprt->xp_pool->sp_rcache,
 732                     &rply, (struct sockaddr *) &xprt->xp_rtaddr, NULL);
 733
 734         svc_sendreply_common(rqstp, &rply, NULL);
 735 }
 736
 737 /*
 738  * Some system error
 739  */
 740 void
 741 svcerr_systemerr(struct svc_req *rqstp)
 742 {
 743         SVCXPRT *xprt = rqstp->rq_xprt;
 744         struct rpc_msg rply;
 745
 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;
 751
 752         if (xprt->xp_pool->sp_rcache)
 753                 replay_setreply(xprt->xp_pool->sp_rcache,
 754                     &rply, svc_getrpccaller(rqstp), NULL);
 755
 756         svc_sendreply_common(rqstp, &rply, NULL);
 757 }
 758
 759 /*
 760  * Authentication error reply
 761  */
 762 void
 763 svcerr_auth(struct svc_req *rqstp, enum auth_stat why)
 764 {
 765         SVCXPRT *xprt = rqstp->rq_xprt;
 766         struct rpc_msg rply;
 767
 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;
 773
 774         if (xprt->xp_pool->sp_rcache)
 775                 replay_setreply(xprt->xp_pool->sp_rcache,
 776                     &rply, svc_getrpccaller(rqstp), NULL);
 777
 778         svc_sendreply_common(rqstp, &rply, NULL);
 779 }
 780
 781 /*
 782  * Auth too weak error reply
 783  */
 784 void
 785 svcerr_weakauth(struct svc_req *rqstp)
 786 {
 787
 788         svcerr_auth(rqstp, AUTH_TOOWEAK);
 789 }
 790
 791 /*
 792  * Program unavailable error reply
 793  */
 794 void
 795 svcerr_noprog(struct svc_req *rqstp)
 796 {
 797         SVCXPRT *xprt = rqstp->rq_xprt;
 798         struct rpc_msg rply;
 799
 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;
 805
 806         if (xprt->xp_pool->sp_rcache)
 807                 replay_setreply(xprt->xp_pool->sp_rcache,
 808                     &rply, svc_getrpccaller(rqstp), NULL);
 809
 810         svc_sendreply_common(rqstp, &rply, NULL);
 811 }
 812
 813 /*
 814  * Program version mismatch error reply
 815  */
 816 void
 817 svcerr_progvers(struct svc_req *rqstp, rpcvers_t low_vers, rpcvers_t high_vers)
 818 {
 819         SVCXPRT *xprt = rqstp->rq_xprt;
 820         struct rpc_msg rply;
 821
 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;
 829
 830         if (xprt->xp_pool->sp_rcache)
 831                 replay_setreply(xprt->xp_pool->sp_rcache,
 832                     &rply, svc_getrpccaller(rqstp), NULL);
 833
 834         svc_sendreply_common(rqstp, &rply, NULL);
 835 }
 836
 837 /*
 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
 841  * parameters.
 842  */
 843 SVCXPRT *
 844 svc_xprt_alloc(void)
 845 {
 846         SVCXPRT *xprt;
 847         SVCXPRT_EXT *ext;
 848
 849         xprt = mem_alloc(sizeof(SVCXPRT));
 850         ext = mem_alloc(sizeof(SVCXPRT_EXT));
 851         xprt->xp_p3 = ext;
 852         refcount_init(&xprt->xp_refs, 1);
 853
 854         return (xprt);
 855 }
 856
 857 /*
 858  * Free a server transport structure.
 859  */
 860 void
 861 svc_xprt_free(SVCXPRT *xprt)
 862 {
 863
 864         mem_free(xprt->xp_p3, sizeof(SVCXPRT_EXT));
 865         mem_free(xprt, sizeof(SVCXPRT));
 866 }
 867
 868 /* ******************* SERVER INPUT STUFF ******************* */
 869
 870 /*
 871  * Read RPC requests from a transport and queue them to be
 872  * executed. We handle authentication and replay cache replies here.
 873  * Actually dispatching the RPC is deferred till svc_executereq.
 874  */
 875 static enum xprt_stat
 876 svc_getreq(SVCXPRT *xprt, struct svc_req **rqstp_ret)
 877 {
 878         SVCPOOL *pool = xprt->xp_pool;
 879         struct svc_req *r;
 880         struct rpc_msg msg;
 881         struct mbuf *args;
 882         struct svc_loss_callout *s;
 883         enum xprt_stat stat;
 884
 885         /* now receive msgs from xprtprt (support batch calls) */
 886         r = malloc(sizeof(*r), M_RPC, M_WAITOK|M_ZERO);
 887
 888         msg.rm_call.cb_cred.oa_base = r->rq_credarea;
 889         msg.rm_call.cb_verf.oa_base = &r->rq_credarea[MAX_AUTH_BYTES];
 890         r->rq_clntcred = &r->rq_credarea[2*MAX_AUTH_BYTES];
 891         if (SVC_RECV(xprt, &msg, &r->rq_addr, &args)) {
 892                 enum auth_stat why;
 893
 894                 /*
 895                  * Handle replays and authenticate before queuing the
 896                  * request to be executed.
 897                  */
 898                 SVC_ACQUIRE(xprt);
 899                 r->rq_xprt = xprt;
 900                 if (pool->sp_rcache) {
 901                         struct rpc_msg repmsg;
 902                         struct mbuf *repbody;
 903                         enum replay_state rs;
 904                         rs = replay_find(pool->sp_rcache, &msg,
 905                             svc_getrpccaller(r), &repmsg, &repbody);
 906                         switch (rs) {
 907                         case RS_NEW:
 908                                 break;
 909                         case RS_DONE:
 910                                 SVC_REPLY(xprt, &repmsg, r->rq_addr,
 911                                     repbody, &r->rq_reply_seq);
 912                                 if (r->rq_addr) {
 913                                         free(r->rq_addr, M_SONAME);
 914                                         r->rq_addr = NULL;
 915                                 }
 916                                 m_freem(args);
 917                                 goto call_done;
 918
 919                         default:
 920                                 m_freem(args);
 921                                 goto call_done;
 922                         }
 923                 }
 924
 925                 r->rq_xid = msg.rm_xid;
 926                 r->rq_prog = msg.rm_call.cb_prog;
 927                 r->rq_vers = msg.rm_call.cb_vers;
 928                 r->rq_proc = msg.rm_call.cb_proc;
 929                 r->rq_size = sizeof(*r) + m_length(args, NULL);
 930                 r->rq_args = args;
 931                 if ((why = _authenticate(r, &msg)) != AUTH_OK) {
 932                         /*
 933                          * RPCSEC_GSS uses this return code
 934                          * for requests that form part of its
 935                          * context establishment protocol and
 936                          * should not be dispatched to the
 937                          * application.
 938                          */
 939                         if (why != RPCSEC_GSS_NODISPATCH)
 940                                 svcerr_auth(r, why);
 941                         goto call_done;
 942                 }
 943
 944                 if (!SVCAUTH_UNWRAP(&r->rq_auth, &r->rq_args)) {
 945                         svcerr_decode(r);
 946                         goto call_done;
 947                 }
 948
 949                 /*
 950                  * Everything checks out, return request to caller.
 951                  */
 952                 *rqstp_ret = r;
 953                 r = NULL;
 954         }
 955 call_done:
 956         if (r) {
 957                 svc_freereq(r);
 958                 r = NULL;
 959         }
 960         if ((stat = SVC_STAT(xprt)) == XPRT_DIED) {
 961                 TAILQ_FOREACH(s, &pool->sp_lcallouts, slc_link)
 962                         (*s->slc_dispatch)(xprt);
 963                 xprt_unregister(xprt);
 964         }
 965
 966         return (stat);
 967 }
 968
 969 static void
 970 svc_executereq(struct svc_req *rqstp)
 971 {
 972         SVCXPRT *xprt = rqstp->rq_xprt;
 973         SVCPOOL *pool = xprt->xp_pool;
 974         int prog_found;
 975         rpcvers_t low_vers;
 976         rpcvers_t high_vers;
 977         struct svc_callout *s;
 978
 979         /* now match message with a registered service*/
 980         prog_found = FALSE;
 981         low_vers = (rpcvers_t) -1L;
 982         high_vers = (rpcvers_t) 0L;
 983         TAILQ_FOREACH(s, &pool->sp_callouts, sc_link) {
 984                 if (s->sc_prog == rqstp->rq_prog) {
 985                         if (s->sc_vers == rqstp->rq_vers) {
 986                                 /*
 987                                  * We hand ownership of r to the
 988                                  * dispatch method - they must call
 989                                  * svc_freereq.
 990                                  */
 991                                 (*s->sc_dispatch)(rqstp, xprt);
 992                                 return;
 993                         }  /* found correct version */
 994                         prog_found = TRUE;
 995                         if (s->sc_vers < low_vers)
 996                                 low_vers = s->sc_vers;
 997                         if (s->sc_vers > high_vers)
 998                                 high_vers = s->sc_vers;
 999                 }   /* found correct program */
1000         }
1001
1002         /*
1003          * if we got here, the program or version
1004          * is not served ...
1005          */
1006         if (prog_found)
1007                 svcerr_progvers(rqstp, low_vers, high_vers);
1008         else
1009                 svcerr_noprog(rqstp);
1010
1011         svc_freereq(rqstp);
1012 }
1013
1014 static void
1015 svc_checkidle(SVCGROUP *grp)
1016 {
1017         SVCXPRT *xprt, *nxprt;
1018         time_t timo;
1019         struct svcxprt_list cleanup;
1020
1021         TAILQ_INIT(&cleanup);
1022         TAILQ_FOREACH_SAFE(xprt, &grp->sg_xlist, xp_link, nxprt) {
1023                 /*
1024                  * Only some transports have idle timers. Don't time
1025                  * something out which is just waking up.
1026                  */
1027                 if (!xprt->xp_idletimeout || xprt->xp_thread)
1028                         continue;
1029
1030                 timo = xprt->xp_lastactive + xprt->xp_idletimeout;
1031                 if (time_uptime > timo) {
1032                         xprt_unregister_locked(xprt);
1033                         TAILQ_INSERT_TAIL(&cleanup, xprt, xp_link);
1034                 }
1035         }
1036
1037         mtx_unlock(&grp->sg_lock);
1038         TAILQ_FOREACH_SAFE(xprt, &cleanup, xp_link, nxprt) {
1039                 SVC_RELEASE(xprt);
1040         }
1041         mtx_lock(&grp->sg_lock);
1042 }
1043
1044 static void
1045 svc_assign_waiting_sockets(SVCPOOL *pool)
1046 {
1047         SVCGROUP *grp;
1048         SVCXPRT *xprt;
1049         int g;
1050
1051         for (g = 0; g < pool->sp_groupcount; g++) {
1052                 grp = &pool->sp_groups[g];
1053                 mtx_lock(&grp->sg_lock);
1054                 while ((xprt = TAILQ_FIRST(&grp->sg_active)) != NULL) {
1055                         if (xprt_assignthread(xprt))
1056                                 TAILQ_REMOVE(&grp->sg_active, xprt, xp_alink);
1057                         else
1058                                 break;
1059                 }
1060                 mtx_unlock(&grp->sg_lock);
1061         }
1062 }
1063
1064 static void
1065 svc_change_space_used(SVCPOOL *pool, long delta)
1066 {
1067         unsigned long value;
1068
1069         value = atomic_fetchadd_long(&pool->sp_space_used, delta) + delta;
1070         if (delta > 0) {
1071                 if (value >= pool->sp_space_high && !pool->sp_space_throttled) {
1072                         pool->sp_space_throttled = TRUE;
1073                         pool->sp_space_throttle_count++;
1074                 }
1075                 if (value > pool->sp_space_used_highest)
1076                         pool->sp_space_used_highest = value;
1077         } else {
1078                 if (value < pool->sp_space_low && pool->sp_space_throttled) {
1079                         pool->sp_space_throttled = FALSE;
1080                         svc_assign_waiting_sockets(pool);
1081                 }
1082         }
1083 }
1084
1085 static bool_t
1086 svc_request_space_available(SVCPOOL *pool)
1087 {
1088
1089         if (pool->sp_space_throttled)
1090                 return (FALSE);
1091         return (TRUE);
1092 }
1093
1094 static void
1095 svc_run_internal(SVCGROUP *grp, bool_t ismaster)
1096 {
1097         SVCPOOL *pool = grp->sg_pool;
1098         SVCTHREAD *st, *stpref;
1099         SVCXPRT *xprt;
1100         enum xprt_stat stat;
1101         struct svc_req *rqstp;
1102         struct proc *p;
1103         long sz;
1104         int error;
1105
1106         st = mem_alloc(sizeof(*st));
1107         mtx_init(&st->st_lock, "st_lock", NULL, MTX_DEF);
1108         st->st_pool = pool;
1109         st->st_xprt = NULL;
1110         STAILQ_INIT(&st->st_reqs);
1111         cv_init(&st->st_cond, "rpcsvc");
1112
1113         mtx_lock(&grp->sg_lock);
1114
1115         /*
1116          * If we are a new thread which was spawned to cope with
1117          * increased load, set the state back to SVCPOOL_ACTIVE.
1118          */
1119         if (grp->sg_state == SVCPOOL_THREADSTARTING)
1120                 grp->sg_state = SVCPOOL_ACTIVE;
1121
1122         while (grp->sg_state != SVCPOOL_CLOSING) {
1123                 /*
1124                  * Create new thread if requested.
1125                  */
1126                 if (grp->sg_state == SVCPOOL_THREADWANTED) {
1127                         grp->sg_state = SVCPOOL_THREADSTARTING;
1128                         grp->sg_lastcreatetime = time_uptime;
1129                         mtx_unlock(&grp->sg_lock);
1130                         svc_new_thread(grp);
1131                         mtx_lock(&grp->sg_lock);
1132                         continue;
1133                 }
1134
1135                 /*
1136                  * Check for idle transports once per second.
1137                  */
1138                 if (time_uptime > grp->sg_lastidlecheck) {
1139                         grp->sg_lastidlecheck = time_uptime;
1140                         svc_checkidle(grp);
1141                 }
1142
1143                 xprt = st->st_xprt;
1144                 if (!xprt) {
1145                         /*
1146                          * Enforce maxthreads count.
1147                          */
1148                         if (grp->sg_threadcount > grp->sg_maxthreads)
1149                                 break;
1150
1151                         /*
1152                          * Before sleeping, see if we can find an
1153                          * active transport which isn't being serviced
1154                          * by a thread.
1155                          */
1156                         if (svc_request_space_available(pool) &&
1157                             (xprt = TAILQ_FIRST(&grp->sg_active)) != NULL) {
1158                                 TAILQ_REMOVE(&grp->sg_active, xprt, xp_alink);
1159                                 SVC_ACQUIRE(xprt);
1160                                 xprt->xp_thread = st;
1161                                 st->st_xprt = xprt;
1162                                 continue;
1163                         }
1164
1165                         LIST_INSERT_HEAD(&grp->sg_idlethreads, st, st_ilink);
1166                         if (ismaster || (!ismaster &&
1167                             grp->sg_threadcount > grp->sg_minthreads))
1168                                 error = cv_timedwait_sig(&st->st_cond,
1169                                     &grp->sg_lock, 5 * hz);
1170                         else
1171                                 error = cv_wait_sig(&st->st_cond,
1172                                     &grp->sg_lock);
1173                         if (st->st_xprt == NULL)
1174                                 LIST_REMOVE(st, st_ilink);
1175
1176                         /*
1177                          * Reduce worker thread count when idle.
1178                          */
1179                         if (error == EWOULDBLOCK) {
1180                                 if (!ismaster
1181                                     && (grp->sg_threadcount
1182                                         > grp->sg_minthreads)
1183                                         && !st->st_xprt)
1184                                         break;
1185                         } else if (error != 0) {
1186                                 KASSERT(error == EINTR || error == ERESTART,
1187                                     ("non-signal error %d", error));
1188                                 mtx_unlock(&grp->sg_lock);
1189                                 p = curproc;
1190                                 PROC_LOCK(p);
1191                                 if (P_SHOULDSTOP(p) ||
1192                                     (p->p_flag & P_TOTAL_STOP) != 0) {
1193                                         thread_suspend_check(0);
1194                                         PROC_UNLOCK(p);
1195                                         mtx_lock(&grp->sg_lock);
1196                                 } else {
1197                                         PROC_UNLOCK(p);
1198                                         svc_exit(pool);
1199                                         mtx_lock(&grp->sg_lock);
1200                                         break;
1201                                 }
1202                         }
1203                         continue;
1204                 }
1205                 mtx_unlock(&grp->sg_lock);
1206
1207                 /*
1208                  * Drain the transport socket and queue up any RPCs.
1209                  */
1210                 xprt->xp_lastactive = time_uptime;
1211                 do {
1212                         if (!svc_request_space_available(pool))
1213                                 break;
1214                         rqstp = NULL;
1215                         stat = svc_getreq(xprt, &rqstp);
1216                         if (rqstp) {
1217                                 svc_change_space_used(pool, rqstp->rq_size);
1218                                 /*
1219                                  * See if the application has a preference
1220                                  * for some other thread.
1221                                  */
1222                                 if (pool->sp_assign) {
1223                                         stpref = pool->sp_assign(st, rqstp);
1224                                         rqstp->rq_thread = stpref;
1225                                         STAILQ_INSERT_TAIL(&stpref->st_reqs,
1226                                             rqstp, rq_link);
1227                                         mtx_unlock(&stpref->st_lock);
1228                                         if (stpref != st)
1229                                                 rqstp = NULL;
1230                                 } else {
1231                                         rqstp->rq_thread = st;
1232                                         STAILQ_INSERT_TAIL(&st->st_reqs,
1233                                             rqstp, rq_link);
1234                                 }
1235                         }
1236                 } while (rqstp == NULL && stat == XPRT_MOREREQS
1237                     && grp->sg_state != SVCPOOL_CLOSING);
1238
1239                 /*
1240                  * Move this transport to the end of the active list to
1241                  * ensure fairness when multiple transports are active.
1242                  * If this was the last queued request, svc_getreq will end
1243                  * up calling xprt_inactive to remove from the active list.
1244                  */
1245                 mtx_lock(&grp->sg_lock);
1246                 xprt->xp_thread = NULL;
1247                 st->st_xprt = NULL;
1248                 if (xprt->xp_active) {
1249                         if (!svc_request_space_available(pool) ||
1250                             !xprt_assignthread(xprt))
1251                                 TAILQ_INSERT_TAIL(&grp->sg_active,
1252                                     xprt, xp_alink);
1253                 }
1254                 mtx_unlock(&grp->sg_lock);
1255                 SVC_RELEASE(xprt);
1256
1257                 /*
1258                  * Execute what we have queued.
1259                  */
1260                 mtx_lock(&st->st_lock);
1261                 while ((rqstp = STAILQ_FIRST(&st->st_reqs)) != NULL) {
1262                         STAILQ_REMOVE_HEAD(&st->st_reqs, rq_link);
1263                         mtx_unlock(&st->st_lock);
1264                         sz = (long)rqstp->rq_size;
1265                         svc_executereq(rqstp);
1266                         svc_change_space_used(pool, -sz);
1267                         mtx_lock(&st->st_lock);
1268                 }
1269                 mtx_unlock(&st->st_lock);
1270                 mtx_lock(&grp->sg_lock);
1271         }
1272
1273         if (st->st_xprt) {
1274                 xprt = st->st_xprt;
1275                 st->st_xprt = NULL;
1276                 SVC_RELEASE(xprt);
1277         }
1278         KASSERT(STAILQ_EMPTY(&st->st_reqs), ("stray reqs on exit"));
1279         mtx_destroy(&st->st_lock);
1280         cv_destroy(&st->st_cond);
1281         mem_free(st, sizeof(*st));
1282
1283         grp->sg_threadcount--;
1284         if (!ismaster)
1285                 wakeup(grp);
1286         mtx_unlock(&grp->sg_lock);
1287 }
1288
1289 static void
1290 svc_thread_start(void *arg)
1291 {
1292
1293         svc_run_internal((SVCGROUP *) arg, FALSE);
1294         kthread_exit();
1295 }
1296
1297 static void
1298 svc_new_thread(SVCGROUP *grp)
1299 {
1300         SVCPOOL *pool = grp->sg_pool;
1301         struct thread *td;
1302
1303         mtx_lock(&grp->sg_lock);
1304         grp->sg_threadcount++;
1305         mtx_unlock(&grp->sg_lock);
1306         kthread_add(svc_thread_start, grp, pool->sp_proc, &td, 0, 0,
1307             "%s: service", pool->sp_name);
1308 }
1309
1310 void
1311 svc_run(SVCPOOL *pool)
1312 {
1313         int g, i;
1314         struct proc *p;
1315         struct thread *td;
1316         SVCGROUP *grp;
1317
1318         p = curproc;
1319         td = curthread;
1320         snprintf(td->td_name, sizeof(td->td_name),
1321             "%s: master", pool->sp_name);
1322         pool->sp_state = SVCPOOL_ACTIVE;
1323         pool->sp_proc = p;
1324
1325         /* Choose group count based on number of threads and CPUs. */
1326         pool->sp_groupcount = max(1, min(SVC_MAXGROUPS,
1327             min(pool->sp_maxthreads / 2, mp_ncpus) / 6));
1328         for (g = 0; g < pool->sp_groupcount; g++) {
1329                 grp = &pool->sp_groups[g];
1330                 grp->sg_minthreads = max(1,
1331                     pool->sp_minthreads / pool->sp_groupcount);
1332                 grp->sg_maxthreads = max(1,
1333                     pool->sp_maxthreads / pool->sp_groupcount);
1334                 grp->sg_lastcreatetime = time_uptime;
1335         }
1336
1337         /* Starting threads */
1338         pool->sp_groups[0].sg_threadcount++;
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);
1343         }
1344         svc_run_internal(&pool->sp_groups[0], TRUE);
1345
1346         /* Waiting for threads to stop. */
1347         for (g = 0; g < pool->sp_groupcount; g++) {
1348                 grp = &pool->sp_groups[g];
1349                 mtx_lock(&grp->sg_lock);
1350                 while (grp->sg_threadcount > 0)
1351                         msleep(grp, &grp->sg_lock, 0, "svcexit", 0);
1352                 mtx_unlock(&grp->sg_lock);
1353         }
1354 }
1355
1356 void
1357 svc_exit(SVCPOOL *pool)
1358 {
1359         SVCGROUP *grp;
1360         SVCTHREAD *st;
1361         int g;
1362
1363         pool->sp_state = SVCPOOL_CLOSING;
1364         for (g = 0; g < pool->sp_groupcount; g++) {
1365                 grp = &pool->sp_groups[g];
1366                 mtx_lock(&grp->sg_lock);
1367                 if (grp->sg_state != SVCPOOL_CLOSING) {
1368                         grp->sg_state = SVCPOOL_CLOSING;
1369                         LIST_FOREACH(st, &grp->sg_idlethreads, st_ilink)
1370                                 cv_signal(&st->st_cond);
1371                 }
1372                 mtx_unlock(&grp->sg_lock);
1373         }
1374 }
1375
1376 bool_t
1377 svc_getargs(struct svc_req *rqstp, xdrproc_t xargs, void *args)
1378 {
1379         struct mbuf *m;
1380         XDR xdrs;
1381         bool_t stat;
1382
1383         m = rqstp->rq_args;
1384         rqstp->rq_args = NULL;
1385
1386         xdrmbuf_create(&xdrs, m, XDR_DECODE);
1387         stat = xargs(&xdrs, args);
1388         XDR_DESTROY(&xdrs);
1389
1390         return (stat);
1391 }
1392
1393 bool_t
1394 svc_freeargs(struct svc_req *rqstp, xdrproc_t xargs, void *args)
1395 {
1396         XDR xdrs;
1397
1398         if (rqstp->rq_addr) {
1399                 free(rqstp->rq_addr, M_SONAME);
1400                 rqstp->rq_addr = NULL;
1401         }
1402
1403         xdrs.x_op = XDR_FREE;
1404         return (xargs(&xdrs, args));
1405 }
1406
1407 void
1408 svc_freereq(struct svc_req *rqstp)
1409 {
1410         SVCTHREAD *st;
1411         SVCPOOL *pool;
1412
1413         st = rqstp->rq_thread;
1414         if (st) {
1415                 pool = st->st_pool;
1416                 if (pool->sp_done)
1417                         pool->sp_done(st, rqstp);
1418         }
1419
1420         if (rqstp->rq_auth.svc_ah_ops)
1421                 SVCAUTH_RELEASE(&rqstp->rq_auth);
1422
1423         if (rqstp->rq_xprt) {
1424                 SVC_RELEASE(rqstp->rq_xprt);
1425         }
1426
1427         if (rqstp->rq_addr)
1428                 free(rqstp->rq_addr, M_SONAME);
1429
1430         if (rqstp->rq_args)
1431                 m_freem(rqstp->rq_args);
1432
1433         free(rqstp, M_RPC);
1434 }