sys/kern/kern_sysctl.c

   1 /*-
   2  * Copyright (c) 1982, 1986, 1989, 1993
   3  *      The Regents of the University of California.  All rights reserved.
   4  *
   5  * This code is derived from software contributed to Berkeley by
   6  * Mike Karels at Berkeley Software Design, Inc.
   7  *
   8  * Quite extensively rewritten by Poul-Henning Kamp of the FreeBSD
   9  * project, to make these variables more userfriendly.
  10  *
  11  * Redistribution and use in source and binary forms, with or without
  12  * modification, are permitted provided that the following conditions
  13  * are met:
  14  * 1. Redistributions of source code must retain the above copyright
  15  *    notice, this list of conditions and the following disclaimer.
  16  * 2. Redistributions in binary form must reproduce the above copyright
  17  *    notice, this list of conditions and the following disclaimer in the
  18  *    documentation and/or other materials provided with the distribution.
  19  * 3. All advertising materials mentioning features or use of this software
  20  *    must display the following acknowledgement:
  21  *      This product includes software developed by the University of
  22  *      California, Berkeley and its contributors.
  23  * 4. Neither the name of the University nor the names of its contributors
  24  *    may be used to endorse or promote products derived from this software
  25  *    without specific prior written permission.
  26  *
  27  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  28  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  29  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  30  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  31  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  32  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  33  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  34  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  35  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  36  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  37  * SUCH DAMAGE.
  38  *
  39  *      @(#)kern_sysctl.c       8.4 (Berkeley) 4/14/94
  40  * $FreeBSD: src/sys/kern/kern_sysctl.c,v 1.92.2.9 2003/05/01 22:48:09 trhodes Exp $
  41  * $DragonFly: src/sys/kern/kern_sysctl.c,v 1.27 2007/04/30 07:18:54 dillon Exp $
  42  */
  43
  44 #include <sys/param.h>
  45 #include <sys/systm.h>
  46 #include <sys/kernel.h>
  47 #include <sys/buf.h>
  48 #include <sys/sysctl.h>
  49 #include <sys/malloc.h>
  50 #include <sys/proc.h>
  51 #include <sys/sysproto.h>
  52 #include <sys/lock.h>
  53 #include <vm/vm.h>
  54 #include <vm/vm_extern.h>
  55
  56 static MALLOC_DEFINE(M_SYSCTL, "sysctl", "sysctl internal magic");
  57 static MALLOC_DEFINE(M_SYSCTLOID, "sysctloid", "sysctl dynamic oids");
  58
  59 static struct lock sysctl_lkp;
  60 static struct lock sysctl_ctx_lkp;
  61
  62 static void     sysctl_lock(int type);
  63 static void     sysctl_unlock(void);
  64 static void     sysctl_ctx_lock(int type);
  65 static void     sysctl_ctx_unlock(void);
  66
  67 static int      sysctl_root(SYSCTL_HANDLER_ARGS);
  68 static void     sysctl_register_oid_int(struct sysctl_oid *oipd);
  69 static void     sysctl_unregister_oid_int(struct sysctl_oid *oipd);
  70 static struct sysctl_ctx_entry* sysctl_ctx_entry_find_int
  71         (struct sysctl_ctx_list *, struct sysctl_oid *oidp);
  72
  73 struct sysctl_oid_list sysctl__children; /* root list */
  74
  75 static struct sysctl_oid *
  76 sysctl_find_oidname(const char *name, struct sysctl_oid_list *list, int lock)
  77 {
  78         struct sysctl_oid *oidp;
  79
  80         SLIST_FOREACH(oidp, list, oid_link) {
  81                 if (strcmp(oidp->oid_name, name) == 0) {
  82                         break;
  83                 }
  84         }
  85         return (oidp);
  86 }
  87
  88 /*
  89  * Initialization of the MIB tree.
  90  *
  91  * Order by number in each list.
  92  */
  93
  94 void
  95 sysctl_register_oid(struct sysctl_oid *oidp)
  96 {
  97         sysctl_lock(LK_EXCLUSIVE);
  98         sysctl_register_oid_int(oidp);
  99         sysctl_unlock();
 100 }
 101
 102 static void
 103 sysctl_register_oid_int(struct sysctl_oid *oidp)
 104 {
 105         struct sysctl_oid_list *parent = oidp->oid_parent;
 106         struct sysctl_oid *p;
 107         struct sysctl_oid *q;
 108
 109         /*
 110          * First check if another oid with the same name already
 111          * exists in the parent's list.
 112          */
 113         p = sysctl_find_oidname(oidp->oid_name, parent, 0);
 114         if (p != NULL) {
 115                 if ((p->oid_kind & CTLTYPE) == CTLTYPE_NODE)
 116                         p->oid_refcnt++;
 117                 else
 118                         kprintf("can't re-use a leaf (%s)!\n", p->oid_name);
 119                 return;
 120         }
 121
 122         /*
 123          * If this oid has a number OID_AUTO, give it a number which
 124          * is greater than any current oid.  Make sure it is at least
 125          * 256 to leave space for pre-assigned oid numbers.
 126          */
 127         if (oidp->oid_number == OID_AUTO) {
 128                 int newoid = 0x100;     /* minimum AUTO oid */
 129
 130                 /*
 131                  * Adjust based on highest oid in parent list
 132                  */
 133                 SLIST_FOREACH(p, parent, oid_link) {
 134                         if (newoid <= p->oid_number)
 135                                 newoid = p->oid_number + 1;
 136                 }
 137                 oidp->oid_number = newoid;
 138         }
 139
 140         /*
 141          * Insert the oid into the parent's list in order.
 142          */
 143         q = NULL;
 144         SLIST_FOREACH(p, parent, oid_link) {
 145                 if (oidp->oid_number < p->oid_number)
 146                         break;
 147                 q = p;
 148         }
 149         if (q)
 150                 SLIST_INSERT_AFTER(q, oidp, oid_link);
 151         else
 152                 SLIST_INSERT_HEAD(parent, oidp, oid_link);
 153 }
 154
 155 void
 156 sysctl_unregister_oid(struct sysctl_oid *oidp)
 157 {
 158         sysctl_lock(LK_EXCLUSIVE);
 159         sysctl_unregister_oid_int(oidp);
 160         sysctl_unlock();
 161 }
 162
 163 static void
 164 sysctl_unregister_oid_int(struct sysctl_oid *oidp)
 165 {
 166         struct sysctl_oid *p;
 167
 168         if (oidp->oid_number == OID_AUTO)
 169                 panic("Trying to unregister OID_AUTO entry: %p", oidp);
 170
 171         SLIST_FOREACH(p, oidp->oid_parent, oid_link) {
 172                 if (p != oidp)
 173                         continue;
 174                 SLIST_REMOVE(oidp->oid_parent, oidp, sysctl_oid, oid_link);
 175                 return;
 176         }
 177
 178         /*
 179          * This can happen when a module fails to register and is
 180          * being unloaded afterwards.  It should not be a panic()
 181          * for normal use.
 182          */
 183         kprintf("%s: failed to unregister sysctl\n", __func__);
 184 }
 185
 186 /* Initialize a new context to keep track of dynamically added sysctls. */
 187 int
 188 sysctl_ctx_init(struct sysctl_ctx_list *c)
 189 {
 190         if (c == NULL)
 191                 return(EINVAL);
 192         TAILQ_INIT(c);
 193         return(0);
 194 }
 195
 196 /* Free the context, and destroy all dynamic oids registered in this context */
 197 int
 198 sysctl_ctx_free(struct sysctl_ctx_list *clist)
 199 {
 200         struct sysctl_ctx_entry *e, *e1;
 201         int error;
 202
 203         error = 0;
 204         sysctl_ctx_lock(LK_EXCLUSIVE);
 205         /*
 206          * First perform a "dry run" to check if it's ok to remove oids.
 207          * XXX FIXME
 208          * XXX This algorithm is a hack. But I don't know any
 209          * XXX better solution for now...
 210          */
 211         TAILQ_FOREACH(e, clist, link) {
 212                 error = sysctl_remove_oid(e->entry, 0, 0);
 213                 if (error)
 214                         break;
 215         }
 216         /*
 217          * Restore deregistered entries, either from the end,
 218          * or from the place where error occured.
 219          * e contains the entry that was not unregistered
 220          */
 221         if (error)
 222                 e1 = TAILQ_PREV(e, sysctl_ctx_list, link);
 223         else
 224                 e1 = TAILQ_LAST(clist, sysctl_ctx_list);
 225         while (e1 != NULL) {
 226                 sysctl_register_oid(e1->entry);
 227                 e1 = TAILQ_PREV(e1, sysctl_ctx_list, link);
 228         }
 229         if (error) {
 230                 sysctl_ctx_unlock();
 231                 return(EBUSY);
 232         }
 233         /* Now really delete the entries */
 234         e = TAILQ_FIRST(clist);
 235         while (e != NULL) {
 236                 e1 = TAILQ_NEXT(e, link);
 237                 error = sysctl_remove_oid(e->entry, 1, 0);
 238                 if (error)
 239                         panic("sysctl_remove_oid: corrupt tree, entry: %s",
 240                             e->entry->oid_name);
 241                 kfree(e, M_SYSCTLOID);
 242                 e = e1;
 243         }
 244         sysctl_ctx_unlock();
 245         return (error);
 246 }
 247
 248 /* Add an entry to the context */
 249 struct sysctl_ctx_entry *
 250 sysctl_ctx_entry_add(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
 251 {
 252         struct sysctl_ctx_entry *e;
 253
 254         if (clist == NULL || oidp == NULL)
 255                 return(NULL);
 256         e = kmalloc(sizeof(struct sysctl_ctx_entry), M_SYSCTLOID, M_WAITOK);
 257         e->entry = oidp;
 258         sysctl_ctx_lock(LK_EXCLUSIVE);
 259         TAILQ_INSERT_HEAD(clist, e, link);
 260         sysctl_ctx_unlock();
 261         return (e);
 262 }
 263
 264 /* Find an entry in the context */
 265 struct sysctl_ctx_entry *
 266 sysctl_ctx_entry_find(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
 267 {
 268         struct sysctl_ctx_entry *e;
 269
 270         if (clist == NULL || oidp == NULL)
 271                 return(NULL);
 272
 273         sysctl_ctx_lock(LK_SHARED);
 274         e = sysctl_ctx_entry_find_int(clist, oidp);
 275         sysctl_ctx_unlock();
 276
 277         return(e);
 278 }
 279
 280 struct sysctl_ctx_entry *
 281 sysctl_ctx_entry_find_int(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
 282 {
 283         struct sysctl_ctx_entry *e;
 284
 285         KKASSERT(clist != NULL && oidp != NULL);
 286
 287         for (e = TAILQ_FIRST(clist); e != NULL; e = TAILQ_NEXT(e, link)) {
 288                 if(e->entry == oidp)
 289                         break;
 290         }
 291
 292         return (e);
 293 }
 294
 295 /*
 296  * Delete an entry from the context.
 297  * NOTE: this function doesn't free oidp! You have to remove it
 298  * with sysctl_remove_oid().
 299  */
 300 int
 301 sysctl_ctx_entry_del(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
 302 {
 303         struct sysctl_ctx_entry *e;
 304
 305         if (clist == NULL || oidp == NULL)
 306                 return (EINVAL);
 307
 308         sysctl_ctx_lock(LK_EXCLUSIVE);
 309         e = sysctl_ctx_entry_find_int(clist, oidp);
 310         if (e == NULL) {
 311                 sysctl_ctx_unlock();
 312                 return (ENOENT);
 313         }
 314         TAILQ_REMOVE(clist, e, link);
 315         kfree(e, M_SYSCTLOID);
 316         sysctl_ctx_unlock();
 317
 318         return(0);
 319 }
 320
 321 /*
 322  * Remove dynamically created sysctl trees.
 323  * oidp - top of the tree to be removed
 324  * del - if 0 - just deregister, otherwise free up entries as well
 325  * recurse - if != 0 traverse the subtree to be deleted
 326  */
 327 int
 328 sysctl_remove_oid(struct sysctl_oid *oidp, int del, int recurse)
 329 {
 330         struct sysctl_oid *p;
 331         int error;
 332
 333         if (oidp == NULL)
 334                 return(EINVAL);
 335         if ((oidp->oid_kind & CTLFLAG_DYN) == 0) {
 336                 kprintf("can't remove non-dynamic nodes!\n");
 337                 return (EINVAL);
 338         }
 339         sysctl_lock(LK_EXCLUSIVE | LK_CANRECURSE);
 340         /*
 341          * WARNING: normal method to do this should be through
 342          * sysctl_ctx_free(). Use recursing as the last resort
 343          * method to purge your sysctl tree of leftovers...
 344          * However, if some other code still references these nodes,
 345          * it will panic.
 346          */
 347         if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
 348                 if (oidp->oid_refcnt == 1) {
 349                         SLIST_FOREACH(p, SYSCTL_CHILDREN(oidp), oid_link) {
 350                                 if (!recurse) {
 351                                         sysctl_unlock();
 352                                         return(ENOTEMPTY);
 353                                 }
 354                                 error = sysctl_remove_oid(p, del, recurse);
 355                                 if (error) {
 356                                         sysctl_unlock();
 357                                         return(error);
 358                                 }
 359                         }
 360                         if (del)
 361                                 kfree(SYSCTL_CHILDREN(oidp), M_SYSCTLOID);
 362                 }
 363         }
 364         if (oidp->oid_refcnt > 1 ) {
 365                 oidp->oid_refcnt--;
 366         } else {
 367                 if (oidp->oid_refcnt == 0) {
 368                         kprintf("Warning: bad oid_refcnt=%u (%s)!\n",
 369                                oidp->oid_refcnt, oidp->oid_name);
 370                         sysctl_unlock();
 371                         return(EINVAL);
 372                 }
 373                 sysctl_unregister_oid_int(oidp);
 374                 if (del) {
 375                         if (oidp->oid_descr)
 376                                 kfree(__DECONST(char *,oidp->oid_descr),
 377                                      M_SYSCTLOID);
 378                         kfree(__DECONST(char *, oidp->oid_name), M_SYSCTLOID);
 379                         kfree(oidp, M_SYSCTLOID);
 380                 }
 381         }
 382         sysctl_unlock();
 383         return(0);
 384 }
 385
 386 /*
 387  * Create new sysctls at run time.
 388  * clist may point to a valid context initialized with sysctl_ctx_init().
 389  */
 390 struct sysctl_oid *
 391 sysctl_add_oid(struct sysctl_ctx_list *clist, struct sysctl_oid_list *parent,
 392         int number, const char *name, int kind, void *arg1, int arg2,
 393         int (*handler)(SYSCTL_HANDLER_ARGS), const char *fmt, const char *descr)
 394 {
 395         struct sysctl_oid *oidp;
 396         ssize_t len;
 397         char *newname;
 398
 399         /* You have to hook up somewhere.. */
 400         if (parent == NULL)
 401                 return(NULL);
 402         sysctl_lock(LK_EXCLUSIVE);
 403         /* Check if the node already exists, otherwise create it */
 404         oidp = sysctl_find_oidname(name, parent, 0);
 405         if (oidp != NULL) {
 406                 if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
 407                         oidp->oid_refcnt++;
 408                         /* Update the context */
 409                         if (clist != NULL)
 410                                 sysctl_ctx_entry_add(clist, oidp);
 411                         sysctl_unlock();
 412                         return (oidp);
 413                 } else {
 414                         kprintf("can't re-use a leaf (%s)!\n", name);
 415                         sysctl_unlock();
 416                         return (NULL);
 417                 }
 418         }
 419         oidp = kmalloc(sizeof(struct sysctl_oid), M_SYSCTLOID, M_WAITOK | M_ZERO);
 420         oidp->oid_parent = parent;
 421         SLIST_NEXT(oidp, oid_link) = NULL;
 422         oidp->oid_number = number;
 423         oidp->oid_refcnt = 1;
 424         len = strlen(name);
 425         newname = kmalloc(len + 1, M_SYSCTLOID, M_WAITOK);
 426         bcopy(name, newname, len + 1);
 427         newname[len] = '\0';
 428         oidp->oid_name = newname;
 429         oidp->oid_handler = handler;
 430         oidp->oid_kind = CTLFLAG_DYN | kind;
 431         if ((kind & CTLTYPE) == CTLTYPE_NODE) {
 432                 struct sysctl_oid_list *children;
 433
 434                 /* Allocate space for children */
 435                 children = kmalloc(sizeof(*children), M_SYSCTLOID, M_WAITOK);
 436                 SYSCTL_SET_CHILDREN(oidp, children);
 437                 SLIST_INIT(children);
 438         } else {
 439                 oidp->oid_arg1 = arg1;
 440                 oidp->oid_arg2 = arg2;
 441         }
 442         oidp->oid_fmt = fmt;
 443         if (descr) {
 444                 int len = strlen(descr) + 1;
 445                 oidp->oid_descr = kmalloc(len, M_SYSCTLOID, M_WAITOK);
 446                 if (oidp->oid_descr)
 447                         strcpy((char *)(uintptr_t)(const void *)oidp->oid_descr, descr);
 448         };
 449         /* Update the context, if used */
 450         if (clist != NULL)
 451                 sysctl_ctx_entry_add(clist, oidp);
 452         /* Register this oid */
 453         sysctl_register_oid_int(oidp);
 454         sysctl_unlock();
 455         return (oidp);
 456 }
 457
 458 /*
 459  * Register the kernel's oids on startup.
 460  */
 461 SET_DECLARE(sysctl_set, struct sysctl_oid);
 462
 463 static void
 464 sysctl_register_all(void *arg)
 465 {
 466         struct sysctl_oid **oidp;
 467
 468         lockinit(&sysctl_lkp, "sysctl", 0, 0);
 469         lockinit(&sysctl_ctx_lkp, "sysctl ctx", 0, 0);
 470         SET_FOREACH(oidp, sysctl_set)
 471                 sysctl_register_oid_int(*oidp);
 472 }
 473
 474 SYSINIT(sysctl, SI_BOOT1_POST, SI_ORDER_ANY, sysctl_register_all, 0);
 475
 476 /*
 477  * "Staff-functions"
 478  *
 479  * These functions implement a presently undocumented interface
 480  * used by the sysctl program to walk the tree, and get the type
 481  * so it can print the value.
 482  * This interface is under work and consideration, and should probably
 483  * be killed with a big axe by the first person who can find the time.
 484  * (be aware though, that the proper interface isn't as obvious as it
 485  * may seem, there are various conflicting requirements.
 486  *
 487  * {0,0}        kprintf the entire MIB-tree.
 488  * {0,1,...}    return the name of the "..." OID.
 489  * {0,2,...}    return the next OID.
 490  * {0,3}        return the OID of the name in "new"
 491  * {0,4,...}    return the kind & format info for the "..." OID.
 492  */
 493
 494 static void
 495 sysctl_sysctl_debug_dump_node(struct sysctl_oid_list *l, int i)
 496 {
 497         int k;
 498         struct sysctl_oid *oidp;
 499
 500         sysctl_lock(LK_SHARED);
 501         SLIST_FOREACH(oidp, l, oid_link) {
 502
 503                 for (k=0; k<i; k++)
 504                         kprintf(" ");
 505
 506                 kprintf("%d %s ", oidp->oid_number, oidp->oid_name);
 507
 508                 kprintf("%c%c",
 509                         oidp->oid_kind & CTLFLAG_RD ? 'R':' ',
 510                         oidp->oid_kind & CTLFLAG_WR ? 'W':' ');
 511
 512                 if (oidp->oid_handler)
 513                         kprintf(" *Handler");
 514
 515                 switch (oidp->oid_kind & CTLTYPE) {
 516                         case CTLTYPE_NODE:
 517                                 kprintf(" Node\n");
 518                                 if (!oidp->oid_handler) {
 519                                         sysctl_sysctl_debug_dump_node(
 520                                                 oidp->oid_arg1, i+2);
 521                                 }
 522                                 break;
 523                         case CTLTYPE_INT:    kprintf(" Int\n"); break;
 524                         case CTLTYPE_STRING: kprintf(" String\n"); break;
 525                         case CTLTYPE_QUAD:   kprintf(" Quad\n"); break;
 526                         case CTLTYPE_OPAQUE: kprintf(" Opaque/struct\n"); break;
 527                         default:             kprintf("\n");
 528                 }
 529
 530         }
 531         sysctl_unlock();
 532 }
 533
 534 static int
 535 sysctl_sysctl_debug(SYSCTL_HANDLER_ARGS)
 536 {
 537         int error;
 538
 539         error = suser(req->td);
 540         if (error)
 541                 return error;
 542         sysctl_sysctl_debug_dump_node(&sysctl__children, 0);
 543         return ENOENT;
 544 }
 545
 546 SYSCTL_PROC(_sysctl, 0, debug, CTLTYPE_STRING|CTLFLAG_RD,
 547         0, 0, sysctl_sysctl_debug, "-", "");
 548
 549 static int
 550 sysctl_sysctl_name(SYSCTL_HANDLER_ARGS)
 551 {
 552         int *name = (int *) arg1;
 553         u_int namelen = arg2;
 554         int error = 0;
 555         struct sysctl_oid *oid;
 556         struct sysctl_oid_list *lsp = &sysctl__children, *lsp2;
 557         char buf[10];
 558
 559         sysctl_lock(LK_SHARED);
 560         while (namelen) {
 561                 if (!lsp) {
 562                         ksnprintf(buf,sizeof(buf),"%d",*name);
 563                         if (req->oldidx)
 564                                 error = SYSCTL_OUT(req, ".", 1);
 565                         if (!error)
 566                                 error = SYSCTL_OUT(req, buf, strlen(buf));
 567                         if (error) {
 568                                 sysctl_unlock();
 569                                 return (error);
 570                         }
 571                         namelen--;
 572                         name++;
 573                         continue;
 574                 }
 575                 lsp2 = 0;
 576                 SLIST_FOREACH(oid, lsp, oid_link) {
 577                         if (oid->oid_number != *name)
 578                                 continue;
 579
 580                         if (req->oldidx)
 581                                 error = SYSCTL_OUT(req, ".", 1);
 582                         if (!error)
 583                                 error = SYSCTL_OUT(req, oid->oid_name,
 584                                         strlen(oid->oid_name));
 585                         if (error) {
 586                                 sysctl_unlock();
 587                                 return (error);
 588                         }
 589
 590                         namelen--;
 591                         name++;
 592
 593                         if ((oid->oid_kind & CTLTYPE) != CTLTYPE_NODE)
 594                                 break;
 595
 596                         if (oid->oid_handler)
 597                                 break;
 598
 599                         lsp2 = (struct sysctl_oid_list *)oid->oid_arg1;
 600                         break;
 601                 }
 602                 lsp = lsp2;
 603         }
 604         sysctl_unlock();
 605         return (SYSCTL_OUT(req, "", 1));
 606 }
 607
 608 SYSCTL_NODE(_sysctl, 1, name, CTLFLAG_RD, sysctl_sysctl_name, "");
 609
 610 static int
 611 sysctl_sysctl_next_ls(struct sysctl_oid_list *lsp, int *name, u_int namelen,
 612         int *next, int *len, int level, struct sysctl_oid **oidpp)
 613 {
 614         struct sysctl_oid *oidp;
 615
 616         *len = level;
 617         sysctl_lock(LK_SHARED);
 618         SLIST_FOREACH(oidp, lsp, oid_link) {
 619                 *next = oidp->oid_number;
 620                 *oidpp = oidp;
 621
 622                 if (!namelen) {
 623                         if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE) {
 624                                 sysctl_unlock();
 625                                 return 0;
 626                         }
 627                         if (oidp->oid_handler) {
 628                                 /* We really should call the handler here...*/
 629                                 sysctl_unlock();
 630                                 return 0;
 631                         }
 632                         lsp = (struct sysctl_oid_list *)oidp->oid_arg1;
 633                         if (!sysctl_sysctl_next_ls(lsp, 0, 0, next+1,
 634                                 len, level+1, oidpp)) {
 635                                 sysctl_unlock();
 636                                 return 0;
 637                         }
 638                         goto emptynode;
 639                 }
 640
 641                 if (oidp->oid_number < *name)
 642                         continue;
 643
 644                 if (oidp->oid_number > *name) {
 645                         if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE) {
 646                                 sysctl_unlock();
 647                                 return 0;
 648                         }
 649                         if (oidp->oid_handler) {
 650                                 sysctl_unlock();
 651                                 return 0;
 652                         }
 653                         lsp = (struct sysctl_oid_list *)oidp->oid_arg1;
 654                         if (!sysctl_sysctl_next_ls(lsp, name+1, namelen-1,
 655                                 next+1, len, level+1, oidpp)) {
 656                                 sysctl_unlock();
 657                                 return (0);
 658                         }
 659                         goto next;
 660                 }
 661                 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
 662                         continue;
 663
 664                 if (oidp->oid_handler)
 665                         continue;
 666
 667                 lsp = (struct sysctl_oid_list *)oidp->oid_arg1;
 668                 if (!sysctl_sysctl_next_ls(lsp, name+1, namelen-1, next+1,
 669                         len, level+1, oidpp)) {
 670                         sysctl_unlock();
 671                         return (0);
 672                 }
 673         next:
 674                 namelen = 1;
 675                 *len = level;
 676         emptynode:
 677                 *len = level;
 678         }
 679         sysctl_unlock();
 680         return 1;
 681 }
 682
 683 static int
 684 sysctl_sysctl_next(SYSCTL_HANDLER_ARGS)
 685 {
 686         int *name = (int *) arg1;
 687         u_int namelen = arg2;
 688         int i, j, error;
 689         struct sysctl_oid *oid;
 690         struct sysctl_oid_list *lsp = &sysctl__children;
 691         int newoid[CTL_MAXNAME];
 692
 693         i = sysctl_sysctl_next_ls(lsp, name, namelen, newoid, &j, 1, &oid);
 694         if (i)
 695                 return ENOENT;
 696         error = SYSCTL_OUT(req, newoid, j * sizeof (int));
 697         return (error);
 698 }
 699
 700 SYSCTL_NODE(_sysctl, 2, next, CTLFLAG_RD, sysctl_sysctl_next, "");
 701
 702 static int
 703 name2oid (char *name, int *oid, int *len, struct sysctl_oid **oidpp)
 704 {
 705         int i;
 706         struct sysctl_oid *oidp;
 707         struct sysctl_oid_list *lsp = &sysctl__children;
 708         char *p;
 709
 710         if (!*name)
 711                 return ENOENT;
 712
 713         p = name + strlen(name) - 1 ;
 714         if (*p == '.')
 715                 *p = '\0';
 716
 717         *len = 0;
 718
 719         for (p = name; *p && *p != '.'; p++)
 720                 ;
 721         i = *p;
 722         if (i == '.')
 723                 *p = '\0';
 724
 725         sysctl_lock(LK_SHARED);
 726         oidp = SLIST_FIRST(lsp);
 727
 728         while (oidp && *len < CTL_MAXNAME) {
 729                 if (strcmp(name, oidp->oid_name)) {
 730                         oidp = SLIST_NEXT(oidp, oid_link);
 731                         continue;
 732                 }
 733                 *oid++ = oidp->oid_number;
 734                 (*len)++;
 735
 736                 if (!i) {
 737                         if (oidpp)
 738                                 *oidpp = oidp;
 739                         sysctl_unlock();
 740                         return (0);
 741                 }
 742
 743                 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
 744                         break;
 745
 746                 if (oidp->oid_handler)
 747                         break;
 748
 749                 lsp = (struct sysctl_oid_list *)oidp->oid_arg1;
 750                 oidp = SLIST_FIRST(lsp);
 751                 name = p+1;
 752                 for (p = name; *p && *p != '.'; p++)
 753                                 ;
 754                 i = *p;
 755                 if (i == '.')
 756                         *p = '\0';
 757         }
 758         sysctl_unlock();
 759         return ENOENT;
 760 }
 761
 762 static int
 763 sysctl_sysctl_name2oid(SYSCTL_HANDLER_ARGS)
 764 {
 765         char *p;
 766         int error, oid[CTL_MAXNAME], len;
 767         struct sysctl_oid *op = 0;
 768
 769         if (!req->newlen)
 770                 return ENOENT;
 771         if (req->newlen >= MAXPATHLEN)  /* XXX arbitrary, undocumented */
 772                 return (ENAMETOOLONG);
 773
 774         p = kmalloc(req->newlen+1, M_SYSCTL, M_WAITOK);
 775
 776         error = SYSCTL_IN(req, p, req->newlen);
 777         if (error) {
 778                 kfree(p, M_SYSCTL);
 779                 return (error);
 780         }
 781
 782         p [req->newlen] = '\0';
 783
 784         error = name2oid(p, oid, &len, &op);
 785
 786         kfree(p, M_SYSCTL);
 787
 788         if (error)
 789                 return (error);
 790
 791         error = SYSCTL_OUT(req, oid, len * sizeof *oid);
 792         return (error);
 793 }
 794
 795 SYSCTL_PROC(_sysctl, 3, name2oid, CTLFLAG_RW|CTLFLAG_ANYBODY, 0, 0,
 796         sysctl_sysctl_name2oid, "I", "");
 797
 798 static int
 799 sysctl_sysctl_oidfmt(SYSCTL_HANDLER_ARGS)
 800 {
 801         struct sysctl_oid *oid;
 802         int error;
 803
 804         error = sysctl_find_oid(arg1, arg2, &oid, NULL, req);
 805         if (error)
 806                 return (error);
 807
 808         if (!oid->oid_fmt)
 809                 return (ENOENT);
 810         error = SYSCTL_OUT(req, &oid->oid_kind, sizeof(oid->oid_kind));
 811         if (error)
 812                 return (error);
 813         error = SYSCTL_OUT(req, oid->oid_fmt, strlen(oid->oid_fmt) + 1);
 814         return (error);
 815 }
 816
 817
 818 SYSCTL_NODE(_sysctl, 4, oidfmt, CTLFLAG_RD, sysctl_sysctl_oidfmt, "");
 819
 820 static int
 821 sysctl_sysctl_oiddescr(SYSCTL_HANDLER_ARGS)
 822 {
 823         struct sysctl_oid *oid;
 824         int error;
 825
 826         error = sysctl_find_oid(arg1, arg2, &oid, NULL, req);
 827         if (error)
 828                 return (error);
 829
 830         if (!oid->oid_descr)
 831                 return (ENOENT);
 832         error = SYSCTL_OUT(req, oid->oid_descr, strlen(oid->oid_descr) + 1);
 833         return (error);
 834 }
 835
 836 SYSCTL_NODE(_sysctl, 5, oiddescr, CTLFLAG_RD, sysctl_sysctl_oiddescr, "");
 837
 838 /*
 839  * Default "handler" functions.
 840  */
 841
 842 /*
 843  * Handle an int, signed or unsigned.
 844  * Two cases:
 845  *     a variable:  point arg1 at it.
 846  *     a constant:  pass it in arg2.
 847  */
 848
 849 int
 850 sysctl_handle_int(SYSCTL_HANDLER_ARGS)
 851 {
 852         int error = 0;
 853
 854         if (arg1)
 855                 error = SYSCTL_OUT(req, arg1, sizeof(int));
 856         else
 857                 error = SYSCTL_OUT(req, &arg2, sizeof(int));
 858
 859         if (error || !req->newptr)
 860                 return (error);
 861
 862         if (!arg1)
 863                 error = EPERM;
 864         else
 865                 error = SYSCTL_IN(req, arg1, sizeof(int));
 866         return (error);
 867 }
 868
 869 /*
 870  * Handle a long, signed or unsigned.  arg1 points to it.
 871  */
 872
 873 int
 874 sysctl_handle_long(SYSCTL_HANDLER_ARGS)
 875 {
 876         int error = 0;
 877
 878         if (!arg1)
 879                 return (EINVAL);
 880         error = SYSCTL_OUT(req, arg1, sizeof(long));
 881
 882         if (error || !req->newptr)
 883                 return (error);
 884
 885         error = SYSCTL_IN(req, arg1, sizeof(long));
 886         return (error);
 887 }
 888
 889 /*
 890  * Handle a quad, signed or unsigned.  arg1 points to it.
 891  */
 892
 893 int
 894 sysctl_handle_quad(SYSCTL_HANDLER_ARGS)
 895 {
 896         int error = 0;
 897
 898         if (!arg1)
 899                 return (EINVAL);
 900         error = SYSCTL_OUT(req, arg1, sizeof(quad_t));
 901
 902         if (error || !req->newptr)
 903                 return (error);
 904
 905         error = SYSCTL_IN(req, arg1, sizeof(quad_t));
 906         return (error);
 907 }
 908
 909 /*
 910  * Handle our generic '\0' terminated 'C' string.
 911  * Two cases:
 912  *      a variable string:  point arg1 at it, arg2 is max length.
 913  *      a constant string:  point arg1 at it, arg2 is zero.
 914  */
 915
 916 int
 917 sysctl_handle_string(SYSCTL_HANDLER_ARGS)
 918 {
 919         int error=0;
 920
 921         error = SYSCTL_OUT(req, arg1, strlen((char *)arg1)+1);
 922
 923         if (error || !req->newptr)
 924                 return (error);
 925
 926         if ((req->newlen - req->newidx) >= arg2) {
 927                 error = EINVAL;
 928         } else {
 929                 arg2 = (req->newlen - req->newidx);
 930                 error = SYSCTL_IN(req, arg1, arg2);
 931                 ((char *)arg1)[arg2] = '\0';
 932         }
 933
 934         return (error);
 935 }
 936
 937 /*
 938  * Handle any kind of opaque data.
 939  * arg1 points to it, arg2 is the size.
 940  */
 941
 942 int
 943 sysctl_handle_opaque(SYSCTL_HANDLER_ARGS)
 944 {
 945         int error;
 946
 947         error = SYSCTL_OUT(req, arg1, arg2);
 948
 949         if (error || !req->newptr)
 950                 return (error);
 951
 952         error = SYSCTL_IN(req, arg1, arg2);
 953
 954         return (error);
 955 }
 956
 957 /*
 958  * Transfer functions to/from kernel space.
 959  * XXX: rather untested at this point
 960  */
 961 static int
 962 sysctl_old_kernel(struct sysctl_req *req, const void *p, size_t l)
 963 {
 964         size_t i = 0;
 965
 966         if (req->oldptr) {
 967                 i = l;
 968                 if (i > req->oldlen - req->oldidx)
 969                         i = req->oldlen - req->oldidx;
 970                 if (i > 0)
 971                         bcopy(p, (char *)req->oldptr + req->oldidx, i);
 972         }
 973         req->oldidx += l;
 974         if (req->oldptr && i != l)
 975                 return (ENOMEM);
 976         return (0);
 977 }
 978
 979 static int
 980 sysctl_new_kernel(struct sysctl_req *req, void *p, size_t l)
 981 {
 982
 983         if (!req->newptr)
 984                 return 0;
 985         if (req->newlen - req->newidx < l)
 986                 return (EINVAL);
 987         bcopy((char *)req->newptr + req->newidx, p, l);
 988         req->newidx += l;
 989         return (0);
 990 }
 991
 992 int
 993 kernel_sysctl(int *name, u_int namelen, void *old, size_t *oldlenp, void *new, size_t newlen, size_t *retval)
 994 {
 995         int error = 0;
 996         struct sysctl_req req;
 997
 998         bzero(&req, sizeof req);
 999
1000         req.td = curthread;
1001
1002         if (oldlenp) {
1003                 req.oldlen = *oldlenp;
1004         }
1005
1006         if (old) {
1007                 req.oldptr = old;
1008         }
1009
1010         if (new != NULL) {
1011                 req.newlen = newlen;
1012                 req.newptr = new;
1013         }
1014
1015         req.oldfunc = sysctl_old_kernel;
1016         req.newfunc = sysctl_new_kernel;
1017         req.lock = 1;
1018
1019         sysctl_lock(LK_SHARED);
1020
1021         error = sysctl_root(0, name, namelen, &req);
1022
1023         if (req.lock == 2)
1024                 vsunlock(req.oldptr, req.oldlen);
1025
1026         sysctl_unlock();
1027
1028         if (error && error != ENOMEM)
1029                 return (error);
1030
1031         if (retval) {
1032                 if (req.oldptr && req.oldidx > req.oldlen)
1033                         *retval = req.oldlen;
1034                 else
1035                         *retval = req.oldidx;
1036         }
1037         return (error);
1038 }
1039
1040 int
1041 kernel_sysctlbyname(char *name, void *old, size_t *oldlenp,
1042     void *new, size_t newlen, size_t *retval)
1043 {
1044         int oid[CTL_MAXNAME];
1045         size_t oidlen, plen;
1046         int error;
1047
1048         oid[0] = 0;             /* sysctl internal magic */
1049         oid[1] = 3;             /* name2oid */
1050         oidlen = sizeof(oid);
1051
1052         error = kernel_sysctl(oid, 2, oid, &oidlen, (void *)name,
1053                     strlen(name), &plen);
1054         if (error)
1055                 return (error);
1056
1057         error = kernel_sysctl(oid, plen / sizeof(int), old, oldlenp,
1058             new, newlen, retval);
1059         return (error);
1060 }
1061
1062 /*
1063  * Transfer function to/from user space.
1064  */
1065 static int
1066 sysctl_old_user(struct sysctl_req *req, const void *p, size_t l)
1067 {
1068         int error = 0;
1069         size_t i = 0;
1070
1071         if (req->lock == 1 && req->oldptr) {
1072                 vslock(req->oldptr, req->oldlen);
1073                 req->lock = 2;
1074         }
1075         if (req->oldptr) {
1076                 i = l;
1077                 if (i > req->oldlen - req->oldidx)
1078                         i = req->oldlen - req->oldidx;
1079                 if (i > 0)
1080                         error = copyout(p, (char *)req->oldptr + req->oldidx,
1081                                         i);
1082         }
1083         req->oldidx += l;
1084         if (error)
1085                 return (error);
1086         if (req->oldptr && i < l)
1087                 return (ENOMEM);
1088         return (0);
1089 }
1090
1091 static int
1092 sysctl_new_user(struct sysctl_req *req, void *p, size_t l)
1093 {
1094         int error;
1095
1096         if (!req->newptr)
1097                 return 0;
1098         if (req->newlen - req->newidx < l)
1099                 return (EINVAL);
1100         error = copyin((char *)req->newptr + req->newidx, p, l);
1101         req->newidx += l;
1102         return (error);
1103 }
1104
1105 int
1106 sysctl_find_oid(int *name, u_int namelen, struct sysctl_oid **noid,
1107     int *nindx, struct sysctl_req *req)
1108 {
1109         struct sysctl_oid *oid;
1110         int indx;
1111
1112         sysctl_lock(LK_SHARED);
1113         oid = SLIST_FIRST(&sysctl__children);
1114         indx = 0;
1115         while (oid && indx < CTL_MAXNAME) {
1116                 if (oid->oid_number == name[indx]) {
1117                         indx++;
1118                         if (oid->oid_kind & CTLFLAG_NOLOCK)
1119                                 req->lock = 0;
1120                         if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
1121                                 if (oid->oid_handler != NULL ||
1122                                     indx == namelen) {
1123                                         *noid = oid;
1124                                         if (nindx != NULL)
1125                                                 *nindx = indx;
1126                                         sysctl_unlock();
1127                                         return (0);
1128                                 }
1129                                 oid = SLIST_FIRST(
1130                                     (struct sysctl_oid_list *)oid->oid_arg1);
1131                         } else if (indx == namelen) {
1132                                 *noid = oid;
1133                                 if (nindx != NULL)
1134                                         *nindx = indx;
1135                                 sysctl_unlock();
1136                                 return (0);
1137                         } else {
1138                                 sysctl_unlock();
1139                                 return (ENOTDIR);
1140                         }
1141                 } else {
1142                         oid = SLIST_NEXT(oid, oid_link);
1143                 }
1144         }
1145         sysctl_unlock();
1146         return (ENOENT);
1147 }
1148
1149 /*
1150  * Traverse our tree, and find the right node, execute whatever it points
1151  * to, and return the resulting error code.
1152  */
1153
1154 int
1155 sysctl_root(SYSCTL_HANDLER_ARGS)
1156 {
1157         struct thread *td = req->td;
1158         struct proc *p = td ? td->td_proc : NULL;
1159         struct sysctl_oid *oid;
1160         int error, indx;
1161
1162         error = sysctl_find_oid(arg1, arg2, &oid, &indx, req);
1163         if (error)
1164                 return (error);
1165
1166         if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
1167                 /*
1168                  * You can't call a sysctl when it's a node, but has
1169                  * no handler.  Inform the user that it's a node.
1170                  * The indx may or may not be the same as namelen.
1171                  */
1172                 if (oid->oid_handler == NULL)
1173                         return (EISDIR);
1174         }
1175
1176         /* If writing isn't allowed */
1177         if (req->newptr && (!(oid->oid_kind & CTLFLAG_WR) ||
1178             ((oid->oid_kind & CTLFLAG_SECURE) && securelevel > 0)))
1179                 return (EPERM);
1180
1181         /* Most likely only root can write */
1182         if (!(oid->oid_kind & CTLFLAG_ANYBODY) && req->newptr && p &&
1183             (error = suser_cred(p->p_ucred,
1184              (oid->oid_kind & CTLFLAG_PRISON) ? PRISON_ROOT : 0)))
1185                 return (error);
1186
1187         if (!oid->oid_handler)
1188                 return EINVAL;
1189
1190         if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE)
1191                 error = oid->oid_handler(oid, (int *)arg1 + indx, arg2 - indx,
1192                     req);
1193         else
1194                 error = oid->oid_handler(oid, oid->oid_arg1, oid->oid_arg2,
1195                     req);
1196         return (error);
1197 }
1198
1199 int
1200 sys___sysctl(struct sysctl_args *uap)
1201 {
1202         int error, i, name[CTL_MAXNAME];
1203         size_t j;
1204
1205         if (uap->namelen > CTL_MAXNAME || uap->namelen < 2)
1206                 return (EINVAL);
1207
1208         error = copyin(uap->name, &name, uap->namelen * sizeof(int));
1209         if (error)
1210                 return (error);
1211
1212         error = userland_sysctl(name, uap->namelen,
1213                 uap->old, uap->oldlenp, 0,
1214                 uap->new, uap->newlen, &j);
1215         if (error && error != ENOMEM)
1216                 return (error);
1217         if (uap->oldlenp) {
1218                 i = copyout(&j, uap->oldlenp, sizeof(j));
1219                 if (i)
1220                         return (i);
1221         }
1222         return (error);
1223 }
1224
1225 /*
1226  * This is used from various compatibility syscalls too.  That's why name
1227  * must be in kernel space.
1228  */
1229 int
1230 userland_sysctl(int *name, u_int namelen, void *old, size_t *oldlenp, int inkernel, void *new, size_t newlen, size_t *retval)
1231 {
1232         int error = 0;
1233         struct sysctl_req req, req2;
1234
1235         bzero(&req, sizeof req);
1236
1237         if (oldlenp) {
1238                 if (inkernel) {
1239                         req.oldlen = *oldlenp;
1240                 } else {
1241                         error = copyin(oldlenp, &req.oldlen, sizeof(*oldlenp));
1242                         if (error)
1243                                 return (error);
1244                 }
1245         }
1246
1247         if (old) {
1248                 if (!useracc(old, req.oldlen, VM_PROT_WRITE))
1249                         return (EFAULT);
1250                 req.oldptr= old;
1251         }
1252
1253         if (new != NULL) {
1254                 if (!useracc(new, req.newlen, VM_PROT_READ))
1255                         return (EFAULT);
1256                 req.newlen = newlen;
1257                 req.newptr = new;
1258         }
1259
1260         req.oldfunc = sysctl_old_user;
1261         req.newfunc = sysctl_new_user;
1262         req.lock = 1;
1263         req.td = curthread;
1264
1265         sysctl_lock(LK_SHARED);
1266
1267         do {
1268             req2 = req;
1269             error = sysctl_root(0, name, namelen, &req2);
1270         } while (error == EAGAIN);
1271
1272         req = req2;
1273         if (req.lock == 2)
1274                 vsunlock(req.oldptr, req.oldlen);
1275
1276         sysctl_unlock();
1277
1278         if (error && error != ENOMEM)
1279                 return (error);
1280
1281         if (retval) {
1282                 if (req.oldptr && req.oldidx > req.oldlen)
1283                         *retval = req.oldlen;
1284                 else
1285                         *retval = req.oldidx;
1286         }
1287         return (error);
1288 }
1289
1290 static void
1291 sysctl_lock(int flag)
1292 {
1293         lockmgr(&sysctl_lkp, flag);
1294 }
1295
1296 static void
1297 sysctl_unlock(void)
1298 {
1299         lockmgr(&sysctl_lkp, LK_RELEASE);
1300 }
1301
1302 static void
1303 sysctl_ctx_lock(int flag)
1304 {
1305         lockmgr(&sysctl_ctx_lkp, flag);
1306 }
1307
1308 static void
1309 sysctl_ctx_unlock(void)
1310 {
1311         lockmgr(&sysctl_ctx_lkp, LK_RELEASE);
1312 }
1313