kern_sysctl.c

Version: - FREEBSD - FREEBSD-13-STABLE - FREEBSD-13-0 - FREEBSD-12-STABLE - FREEBSD-12-0 - FREEBSD-11-STABLE - FREEBSD-11-0 - FREEBSD-10-STABLE - FREEBSD-10-0 - FREEBSD-9-STABLE - FREEBSD-9-0 - FREEBSD-8-STABLE - FREEBSD-8-0 - FREEBSD-7-STABLE - FREEBSD-7-0 - FREEBSD-6-STABLE - FREEBSD-6-0 - FREEBSD-5-STABLE - FREEBSD-5-0 - FREEBSD-4-STABLE - FREEBSD-3-STABLE - FREEBSD22 - l41 - OPENBSD - linux-2.6 - MK84 - PLAN9 - xnu-8792
SearchContext: - none - 3 - 10
    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  * 4. Neither the name of the University nor the names of its contributors
   20  *    may be used to endorse or promote products derived from this software
   21  *    without specific prior written permission.
   22  *
   23  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
   24  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
   25  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
   26  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
   27  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
   28  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
   29  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
   30  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
   31  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
   32  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
   33  * SUCH DAMAGE.
   34  *
   35  *      @(#)kern_sysctl.c       8.4 (Berkeley) 4/14/94
   36  */
   37 
   38 #include <sys/cdefs.h>
   39 __FBSDID("$FreeBSD$");
   40 
   41 #include "opt_compat.h"
   42 #include "opt_ktrace.h"
   43 
   44 #include <sys/param.h>
   45 #include <sys/fail.h>
   46 #include <sys/systm.h>
   47 #include <sys/kernel.h>
   48 #include <sys/sysctl.h>
   49 #include <sys/malloc.h>
   50 #include <sys/priv.h>
   51 #include <sys/proc.h>
   52 #include <sys/jail.h>
   53 #include <sys/lock.h>
   54 #include <sys/mutex.h>
   55 #include <sys/sbuf.h>
   56 #include <sys/sx.h>
   57 #include <sys/sysproto.h>
   58 #include <sys/uio.h>
   59 #ifdef KTRACE
   60 #include <sys/ktrace.h>
   61 #endif
   62 
   63 #include <net/vnet.h>
   64 
   65 #include <security/mac/mac_framework.h>
   66 
   67 #include <vm/vm.h>
   68 #include <vm/vm_extern.h>
   69 
   70 static MALLOC_DEFINE(M_SYSCTL, "sysctl", "sysctl internal magic");
   71 static MALLOC_DEFINE(M_SYSCTLOID, "sysctloid", "sysctl dynamic oids");
   72 static MALLOC_DEFINE(M_SYSCTLTMP, "sysctltmp", "sysctl temp output buffer");
   73 
   74 /*
   75  * The sysctllock protects the MIB tree.  It also protects sysctl
   76  * contexts used with dynamic sysctls.  The sysctl_register_oid() and
   77  * sysctl_unregister_oid() routines require the sysctllock to already
   78  * be held, so the sysctl_lock() and sysctl_unlock() routines are
   79  * provided for the few places in the kernel which need to use that
   80  * API rather than using the dynamic API.  Use of the dynamic API is
   81  * strongly encouraged for most code.
   82  *
   83  * The sysctlmemlock is used to limit the amount of user memory wired for
   84  * sysctl requests.  This is implemented by serializing any userland
   85  * sysctl requests larger than a single page via an exclusive lock.
   86  */
   87 static struct sx sysctllock;
   88 static struct sx sysctlmemlock;
   89 
   90 #define SYSCTL_XLOCK()          sx_xlock(&sysctllock)
   91 #define SYSCTL_XUNLOCK()        sx_xunlock(&sysctllock)
   92 #define SYSCTL_ASSERT_XLOCKED() sx_assert(&sysctllock, SA_XLOCKED)
   93 #define SYSCTL_INIT()           sx_init(&sysctllock, "sysctl lock")
   94 #define SYSCTL_SLEEP(ch, wmesg, timo)                                   \
   95                                 sx_sleep(ch, &sysctllock, 0, wmesg, timo)
   96 
   97 static int sysctl_root(SYSCTL_HANDLER_ARGS);
   98 
   99 struct sysctl_oid_list sysctl__children; /* root list */
  100 
  101 static int      sysctl_remove_oid_locked(struct sysctl_oid *oidp, int del,
  102                     int recurse);
  103 
  104 static struct sysctl_oid *
  105 sysctl_find_oidname(const char *name, struct sysctl_oid_list *list)
  106 {
  107         struct sysctl_oid *oidp;
  108 
  109         SYSCTL_ASSERT_XLOCKED();
  110         SLIST_FOREACH(oidp, list, oid_link) {
  111                 if (strcmp(oidp->oid_name, name) == 0) {
  112                         return (oidp);
  113                 }
  114         }
  115         return (NULL);
  116 }
  117 
  118 /*
  119  * Initialization of the MIB tree.
  120  *
  121  * Order by number in each list.
  122  */
  123 void
  124 sysctl_lock(void)
  125 {
  126 
  127         SYSCTL_XLOCK();
  128 }
  129 
  130 void
  131 sysctl_unlock(void)
  132 {
  133 
  134         SYSCTL_XUNLOCK();
  135 }
  136 
  137 void
  138 sysctl_register_oid(struct sysctl_oid *oidp)
  139 {
  140         struct sysctl_oid_list *parent = oidp->oid_parent;
  141         struct sysctl_oid *p;
  142         struct sysctl_oid *q;
  143 
  144         /*
  145          * First check if another oid with the same name already
  146          * exists in the parent's list.
  147          */
  148         SYSCTL_ASSERT_XLOCKED();
  149         p = sysctl_find_oidname(oidp->oid_name, parent);
  150         if (p != NULL) {
  151                 if ((p->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
  152                         p->oid_refcnt++;
  153                         return;
  154                 } else {
  155                         printf("can't re-use a leaf (%s)!\n", p->oid_name);
  156                         return;
  157                 }
  158         }
  159         /*
  160          * If this oid has a number OID_AUTO, give it a number which
  161          * is greater than any current oid.
  162          * NOTE: DO NOT change the starting value here, change it in
  163          * <sys/sysctl.h>, and make sure it is at least 256 to
  164          * accomodate e.g. net.inet.raw as a static sysctl node.
  165          */
  166         if (oidp->oid_number == OID_AUTO) {
  167                 static int newoid = CTL_AUTO_START;
  168 
  169                 oidp->oid_number = newoid++;
  170                 if (newoid == 0x7fffffff)
  171                         panic("out of oids");
  172         }
  173 #if 0
  174         else if (oidp->oid_number >= CTL_AUTO_START) {
  175                 /* do not panic; this happens when unregistering sysctl sets */
  176                 printf("static sysctl oid too high: %d", oidp->oid_number);
  177         }
  178 #endif
  179 
  180         /*
  181          * Insert the oid into the parent's list in order.
  182          */
  183         q = NULL;
  184         SLIST_FOREACH(p, parent, oid_link) {
  185                 if (oidp->oid_number < p->oid_number)
  186                         break;
  187                 q = p;
  188         }
  189         if (q)
  190                 SLIST_INSERT_AFTER(q, oidp, oid_link);
  191         else
  192                 SLIST_INSERT_HEAD(parent, oidp, oid_link);
  193 }
  194 
  195 void
  196 sysctl_unregister_oid(struct sysctl_oid *oidp)
  197 {
  198         struct sysctl_oid *p;
  199         int error;
  200 
  201         SYSCTL_ASSERT_XLOCKED();
  202         error = ENOENT;
  203         if (oidp->oid_number == OID_AUTO) {
  204                 error = EINVAL;
  205         } else {
  206                 SLIST_FOREACH(p, oidp->oid_parent, oid_link) {
  207                         if (p == oidp) {
  208                                 SLIST_REMOVE(oidp->oid_parent, oidp,
  209                                     sysctl_oid, oid_link);
  210                                 error = 0;
  211                                 break;
  212                         }
  213                 }
  214         }
  215 
  216         /* 
  217          * This can happen when a module fails to register and is
  218          * being unloaded afterwards.  It should not be a panic()
  219          * for normal use.
  220          */
  221         if (error)
  222                 printf("%s: failed to unregister sysctl\n", __func__);
  223 }
  224 
  225 /* Initialize a new context to keep track of dynamically added sysctls. */
  226 int
  227 sysctl_ctx_init(struct sysctl_ctx_list *c)
  228 {
  229 
  230         if (c == NULL) {
  231                 return (EINVAL);
  232         }
  233 
  234         /*
  235          * No locking here, the caller is responsible for not adding
  236          * new nodes to a context until after this function has
  237          * returned.
  238          */
  239         TAILQ_INIT(c);
  240         return (0);
  241 }
  242 
  243 /* Free the context, and destroy all dynamic oids registered in this context */
  244 int
  245 sysctl_ctx_free(struct sysctl_ctx_list *clist)
  246 {
  247         struct sysctl_ctx_entry *e, *e1;
  248         int error;
  249 
  250         error = 0;
  251         /*
  252          * First perform a "dry run" to check if it's ok to remove oids.
  253          * XXX FIXME
  254          * XXX This algorithm is a hack. But I don't know any
  255          * XXX better solution for now...
  256          */
  257         SYSCTL_XLOCK();
  258         TAILQ_FOREACH(e, clist, link) {
  259                 error = sysctl_remove_oid_locked(e->entry, 0, 0);
  260                 if (error)
  261                         break;
  262         }
  263         /*
  264          * Restore deregistered entries, either from the end,
  265          * or from the place where error occured.
  266          * e contains the entry that was not unregistered
  267          */
  268         if (error)
  269                 e1 = TAILQ_PREV(e, sysctl_ctx_list, link);
  270         else
  271                 e1 = TAILQ_LAST(clist, sysctl_ctx_list);
  272         while (e1 != NULL) {
  273                 sysctl_register_oid(e1->entry);
  274                 e1 = TAILQ_PREV(e1, sysctl_ctx_list, link);
  275         }
  276         if (error) {
  277                 SYSCTL_XUNLOCK();
  278                 return(EBUSY);
  279         }
  280         /* Now really delete the entries */
  281         e = TAILQ_FIRST(clist);
  282         while (e != NULL) {
  283                 e1 = TAILQ_NEXT(e, link);
  284                 error = sysctl_remove_oid_locked(e->entry, 1, 0);
  285                 if (error)
  286                         panic("sysctl_remove_oid: corrupt tree, entry: %s",
  287                             e->entry->oid_name);
  288                 free(e, M_SYSCTLOID);
  289                 e = e1;
  290         }
  291         SYSCTL_XUNLOCK();
  292         return (error);
  293 }
  294 
  295 /* Add an entry to the context */
  296 struct sysctl_ctx_entry *
  297 sysctl_ctx_entry_add(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
  298 {
  299         struct sysctl_ctx_entry *e;
  300 
  301         SYSCTL_ASSERT_XLOCKED();
  302         if (clist == NULL || oidp == NULL)
  303                 return(NULL);
  304         e = malloc(sizeof(struct sysctl_ctx_entry), M_SYSCTLOID, M_WAITOK);
  305         e->entry = oidp;
  306         TAILQ_INSERT_HEAD(clist, e, link);
  307         return (e);
  308 }
  309 
  310 /* Find an entry in the context */
  311 struct sysctl_ctx_entry *
  312 sysctl_ctx_entry_find(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
  313 {
  314         struct sysctl_ctx_entry *e;
  315 
  316         SYSCTL_ASSERT_XLOCKED();
  317         if (clist == NULL || oidp == NULL)
  318                 return(NULL);
  319         TAILQ_FOREACH(e, clist, link) {
  320                 if(e->entry == oidp)
  321                         return(e);
  322         }
  323         return (e);
  324 }
  325 
  326 /*
  327  * Delete an entry from the context.
  328  * NOTE: this function doesn't free oidp! You have to remove it
  329  * with sysctl_remove_oid().
  330  */
  331 int
  332 sysctl_ctx_entry_del(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
  333 {
  334         struct sysctl_ctx_entry *e;
  335 
  336         if (clist == NULL || oidp == NULL)
  337                 return (EINVAL);
  338         SYSCTL_XLOCK();
  339         e = sysctl_ctx_entry_find(clist, oidp);
  340         if (e != NULL) {
  341                 TAILQ_REMOVE(clist, e, link);
  342                 SYSCTL_XUNLOCK();
  343                 free(e, M_SYSCTLOID);
  344                 return (0);
  345         } else {
  346                 SYSCTL_XUNLOCK();
  347                 return (ENOENT);
  348         }
  349 }
  350 
  351 /*
  352  * Remove dynamically created sysctl trees.
  353  * oidp - top of the tree to be removed
  354  * del - if 0 - just deregister, otherwise free up entries as well
  355  * recurse - if != 0 traverse the subtree to be deleted
  356  */
  357 int
  358 sysctl_remove_oid(struct sysctl_oid *oidp, int del, int recurse)
  359 {
  360         int error;
  361 
  362         SYSCTL_XLOCK();
  363         error = sysctl_remove_oid_locked(oidp, del, recurse);
  364         SYSCTL_XUNLOCK();
  365         return (error);
  366 }
  367 
  368 static int
  369 sysctl_remove_oid_locked(struct sysctl_oid *oidp, int del, int recurse)
  370 {
  371         struct sysctl_oid *p;
  372         int error;
  373 
  374         SYSCTL_ASSERT_XLOCKED();
  375         if (oidp == NULL)
  376                 return(EINVAL);
  377         if ((oidp->oid_kind & CTLFLAG_DYN) == 0) {
  378                 printf("can't remove non-dynamic nodes!\n");
  379                 return (EINVAL);
  380         }
  381         /*
  382          * WARNING: normal method to do this should be through
  383          * sysctl_ctx_free(). Use recursing as the last resort
  384          * method to purge your sysctl tree of leftovers...
  385          * However, if some other code still references these nodes,
  386          * it will panic.
  387          */
  388         if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
  389                 if (oidp->oid_refcnt == 1) {
  390                         SLIST_FOREACH(p, SYSCTL_CHILDREN(oidp), oid_link) {
  391                                 if (!recurse)
  392                                         return (ENOTEMPTY);
  393                                 error = sysctl_remove_oid_locked(p, del,
  394                                     recurse);
  395                                 if (error)
  396                                         return (error);
  397                         }
  398                         if (del)
  399                                 free(SYSCTL_CHILDREN(oidp), M_SYSCTLOID);
  400                 }
  401         }
  402         if (oidp->oid_refcnt > 1 ) {
  403                 oidp->oid_refcnt--;
  404         } else {
  405                 if (oidp->oid_refcnt == 0) {
  406                         printf("Warning: bad oid_refcnt=%u (%s)!\n",
  407                                 oidp->oid_refcnt, oidp->oid_name);
  408                         return (EINVAL);
  409                 }
  410                 sysctl_unregister_oid(oidp);
  411                 if (del) {
  412                         /*
  413                          * Wait for all threads running the handler to drain.
  414                          * This preserves the previous behavior when the
  415                          * sysctl lock was held across a handler invocation,
  416                          * and is necessary for module unload correctness.
  417                          */
  418                         while (oidp->oid_running > 0) {
  419                                 oidp->oid_kind |= CTLFLAG_DYING;
  420                                 SYSCTL_SLEEP(&oidp->oid_running, "oidrm", 0);
  421                         }
  422                         if (oidp->oid_descr)
  423                                 free(__DECONST(char *, oidp->oid_descr),
  424                                     M_SYSCTLOID);
  425                         free(__DECONST(char *, oidp->oid_name), M_SYSCTLOID);
  426                         free(oidp, M_SYSCTLOID);
  427                 }
  428         }
  429         return (0);
  430 }
  431 
  432 /*
  433  * Create new sysctls at run time.
  434  * clist may point to a valid context initialized with sysctl_ctx_init().
  435  */
  436 struct sysctl_oid *
  437 sysctl_add_oid(struct sysctl_ctx_list *clist, struct sysctl_oid_list *parent,
  438         int number, const char *name, int kind, void *arg1, int arg2,
  439         int (*handler)(SYSCTL_HANDLER_ARGS), const char *fmt, const char *descr)
  440 {
  441         struct sysctl_oid *oidp;
  442 
  443         /* You have to hook up somewhere.. */
  444         if (parent == NULL)
  445                 return(NULL);
  446         /* Check if the node already exists, otherwise create it */
  447         SYSCTL_XLOCK();
  448         oidp = sysctl_find_oidname(name, parent);
  449         if (oidp != NULL) {
  450                 if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
  451                         oidp->oid_refcnt++;
  452                         /* Update the context */
  453                         if (clist != NULL)
  454                                 sysctl_ctx_entry_add(clist, oidp);
  455                         SYSCTL_XUNLOCK();
  456                         return (oidp);
  457                 } else {
  458                         SYSCTL_XUNLOCK();
  459                         printf("can't re-use a leaf (%s)!\n", name);
  460                         return (NULL);
  461                 }
  462         }
  463         oidp = malloc(sizeof(struct sysctl_oid), M_SYSCTLOID, M_WAITOK|M_ZERO);
  464         oidp->oid_parent = parent;
  465         SLIST_NEXT(oidp, oid_link) = NULL;
  466         oidp->oid_number = number;
  467         oidp->oid_refcnt = 1;
  468         oidp->oid_name = strdup(name, M_SYSCTLOID);
  469         oidp->oid_handler = handler;
  470         oidp->oid_kind = CTLFLAG_DYN | kind;
  471         if ((kind & CTLTYPE) == CTLTYPE_NODE) {
  472                 /* Allocate space for children */
  473                 SYSCTL_CHILDREN_SET(oidp, malloc(sizeof(struct sysctl_oid_list),
  474                     M_SYSCTLOID, M_WAITOK));
  475                 SLIST_INIT(SYSCTL_CHILDREN(oidp));
  476         } else {
  477                 oidp->oid_arg1 = arg1;
  478                 oidp->oid_arg2 = arg2;
  479         }
  480         oidp->oid_fmt = fmt;
  481         if (descr)
  482                 oidp->oid_descr = strdup(descr, M_SYSCTLOID);
  483         /* Update the context, if used */
  484         if (clist != NULL)
  485                 sysctl_ctx_entry_add(clist, oidp);
  486         /* Register this oid */
  487         sysctl_register_oid(oidp);
  488         SYSCTL_XUNLOCK();
  489         return (oidp);
  490 }
  491 
  492 /*
  493  * Rename an existing oid.
  494  */
  495 void
  496 sysctl_rename_oid(struct sysctl_oid *oidp, const char *name)
  497 {
  498         char *newname;
  499         char *oldname;
  500 
  501         newname = strdup(name, M_SYSCTLOID);
  502         SYSCTL_XLOCK();
  503         oldname = __DECONST(char *, oidp->oid_name);
  504         oidp->oid_name = newname;
  505         SYSCTL_XUNLOCK();
  506         free(oldname, M_SYSCTLOID);
  507 }
  508 
  509 /*
  510  * Reparent an existing oid.
  511  */
  512 int
  513 sysctl_move_oid(struct sysctl_oid *oid, struct sysctl_oid_list *parent)
  514 {
  515         struct sysctl_oid *oidp;
  516 
  517         SYSCTL_XLOCK();
  518         if (oid->oid_parent == parent) {
  519                 SYSCTL_XUNLOCK();
  520                 return (0);
  521         }
  522         oidp = sysctl_find_oidname(oid->oid_name, parent);
  523         if (oidp != NULL) {
  524                 SYSCTL_XUNLOCK();
  525                 return (EEXIST);
  526         }
  527         sysctl_unregister_oid(oid);
  528         oid->oid_parent = parent;
  529         oid->oid_number = OID_AUTO;
  530         sysctl_register_oid(oid);
  531         SYSCTL_XUNLOCK();
  532         return (0);
  533 }
  534 
  535 /*
  536  * Register the kernel's oids on startup.
  537  */
  538 SET_DECLARE(sysctl_set, struct sysctl_oid);
  539 
  540 static void
  541 sysctl_register_all(void *arg)
  542 {
  543         struct sysctl_oid **oidp;
  544 
  545         sx_init(&sysctlmemlock, "sysctl mem");
  546         SYSCTL_INIT();
  547         SYSCTL_XLOCK();
  548         SET_FOREACH(oidp, sysctl_set)
  549                 sysctl_register_oid(*oidp);
  550         SYSCTL_XUNLOCK();
  551 }
  552 SYSINIT(sysctl, SI_SUB_KMEM, SI_ORDER_ANY, sysctl_register_all, 0);
  553 
  554 /*
  555  * "Staff-functions"
  556  *
  557  * These functions implement a presently undocumented interface 
  558  * used by the sysctl program to walk the tree, and get the type
  559  * so it can print the value.
  560  * This interface is under work and consideration, and should probably
  561  * be killed with a big axe by the first person who can find the time.
  562  * (be aware though, that the proper interface isn't as obvious as it
  563  * may seem, there are various conflicting requirements.
  564  *
  565  * {0,0}        printf the entire MIB-tree.
  566  * {0,1,...}    return the name of the "..." OID.
  567  * {0,2,...}    return the next OID.
  568  * {0,3}        return the OID of the name in "new"
  569  * {0,4,...}    return the kind & format info for the "..." OID.
  570  * {0,5,...}    return the description the "..." OID.
  571  */
  572 
  573 #ifdef SYSCTL_DEBUG
  574 static void
  575 sysctl_sysctl_debug_dump_node(struct sysctl_oid_list *l, int i)
  576 {
  577         int k;
  578         struct sysctl_oid *oidp;
  579 
  580         SYSCTL_ASSERT_XLOCKED();
  581         SLIST_FOREACH(oidp, l, oid_link) {
  582 
  583                 for (k=0; k<i; k++)
  584                         printf(" ");
  585 
  586                 printf("%d %s ", oidp->oid_number, oidp->oid_name);
  587 
  588                 printf("%c%c",
  589                         oidp->oid_kind & CTLFLAG_RD ? 'R':' ',
  590                         oidp->oid_kind & CTLFLAG_WR ? 'W':' ');
  591 
  592                 if (oidp->oid_handler)
  593                         printf(" *Handler");
  594 
  595                 switch (oidp->oid_kind & CTLTYPE) {
  596                         case CTLTYPE_NODE:
  597                                 printf(" Node\n");
  598                                 if (!oidp->oid_handler) {
  599                                         sysctl_sysctl_debug_dump_node(
  600                                                 oidp->oid_arg1, i+2);
  601                                 }
  602                                 break;
  603                         case CTLTYPE_INT:    printf(" Int\n"); break;
  604                         case CTLTYPE_STRING: printf(" String\n"); break;
  605                         case CTLTYPE_QUAD:   printf(" Quad\n"); break;
  606                         case CTLTYPE_OPAQUE: printf(" Opaque/struct\n"); break;
  607                         default:             printf("\n");
  608                 }
  609 
  610         }
  611 }
  612 
  613 static int
  614 sysctl_sysctl_debug(SYSCTL_HANDLER_ARGS)
  615 {
  616         int error;
  617 
  618         error = priv_check(req->td, PRIV_SYSCTL_DEBUG);
  619         if (error)
  620                 return (error);
  621         SYSCTL_XLOCK();
  622         sysctl_sysctl_debug_dump_node(&sysctl__children, 0);
  623         SYSCTL_XUNLOCK();
  624         return (ENOENT);
  625 }
  626 
  627 SYSCTL_PROC(_sysctl, 0, debug, CTLTYPE_STRING|CTLFLAG_RD,
  628         0, 0, sysctl_sysctl_debug, "-", "");
  629 #endif
  630 
  631 static int
  632 sysctl_sysctl_name(SYSCTL_HANDLER_ARGS)
  633 {
  634         int *name = (int *) arg1;
  635         u_int namelen = arg2;
  636         int error = 0;
  637         struct sysctl_oid *oid;
  638         struct sysctl_oid_list *lsp = &sysctl__children, *lsp2;
  639         char buf[10];
  640 
  641         SYSCTL_XLOCK();
  642         while (namelen) {
  643                 if (!lsp) {
  644                         snprintf(buf,sizeof(buf),"%d",*name);
  645                         if (req->oldidx)
  646                                 error = SYSCTL_OUT(req, ".", 1);
  647                         if (!error)
  648                                 error = SYSCTL_OUT(req, buf, strlen(buf));
  649                         if (error)
  650                                 goto out;
  651                         namelen--;
  652                         name++;
  653                         continue;
  654                 }
  655                 lsp2 = 0;
  656                 SLIST_FOREACH(oid, lsp, oid_link) {
  657                         if (oid->oid_number != *name)
  658                                 continue;
  659 
  660                         if (req->oldidx)
  661                                 error = SYSCTL_OUT(req, ".", 1);
  662                         if (!error)
  663                                 error = SYSCTL_OUT(req, oid->oid_name,
  664                                         strlen(oid->oid_name));
  665                         if (error)
  666                                 goto out;
  667 
  668                         namelen--;
  669                         name++;
  670 
  671                         if ((oid->oid_kind & CTLTYPE) != CTLTYPE_NODE) 
  672                                 break;
  673 
  674                         if (oid->oid_handler)
  675                                 break;
  676 
  677                         lsp2 = SYSCTL_CHILDREN(oid);
  678                         break;
  679                 }
  680                 lsp = lsp2;
  681         }
  682         error = SYSCTL_OUT(req, "", 1);
  683  out:
  684         SYSCTL_XUNLOCK();
  685         return (error);
  686 }
  687 
  688 static SYSCTL_NODE(_sysctl, 1, name, CTLFLAG_RD, sysctl_sysctl_name, "");
  689 
  690 static int
  691 sysctl_sysctl_next_ls(struct sysctl_oid_list *lsp, int *name, u_int namelen, 
  692         int *next, int *len, int level, struct sysctl_oid **oidpp)
  693 {
  694         struct sysctl_oid *oidp;
  695 
  696         SYSCTL_ASSERT_XLOCKED();
  697         *len = level;
  698         SLIST_FOREACH(oidp, lsp, oid_link) {
  699                 *next = oidp->oid_number;
  700                 *oidpp = oidp;
  701 
  702                 if (oidp->oid_kind & CTLFLAG_SKIP)
  703                         continue;
  704 
  705                 if (!namelen) {
  706                         if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE) 
  707                                 return (0);
  708                         if (oidp->oid_handler) 
  709                                 /* We really should call the handler here...*/
  710                                 return (0);
  711                         lsp = SYSCTL_CHILDREN(oidp);
  712                         if (!sysctl_sysctl_next_ls(lsp, 0, 0, next+1, 
  713                                 len, level+1, oidpp))
  714                                 return (0);
  715                         goto emptynode;
  716                 }
  717 
  718                 if (oidp->oid_number < *name)
  719                         continue;
  720 
  721                 if (oidp->oid_number > *name) {
  722                         if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
  723                                 return (0);
  724                         if (oidp->oid_handler)
  725                                 return (0);
  726                         lsp = SYSCTL_CHILDREN(oidp);
  727                         if (!sysctl_sysctl_next_ls(lsp, name+1, namelen-1, 
  728                                 next+1, len, level+1, oidpp))
  729                                 return (0);
  730                         goto next;
  731                 }
  732                 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
  733                         continue;
  734 
  735                 if (oidp->oid_handler)
  736                         continue;
  737 
  738                 lsp = SYSCTL_CHILDREN(oidp);
  739                 if (!sysctl_sysctl_next_ls(lsp, name+1, namelen-1, next+1, 
  740                         len, level+1, oidpp))
  741                         return (0);
  742         next:
  743                 namelen = 1;
  744         emptynode:
  745                 *len = level;
  746         }
  747         return (1);
  748 }
  749 
  750 static int
  751 sysctl_sysctl_next(SYSCTL_HANDLER_ARGS)
  752 {
  753         int *name = (int *) arg1;
  754         u_int namelen = arg2;
  755         int i, j, error;
  756         struct sysctl_oid *oid;
  757         struct sysctl_oid_list *lsp = &sysctl__children;
  758         int newoid[CTL_MAXNAME];
  759 
  760         SYSCTL_XLOCK();
  761         i = sysctl_sysctl_next_ls(lsp, name, namelen, newoid, &j, 1, &oid);
  762         SYSCTL_XUNLOCK();
  763         if (i)
  764                 return (ENOENT);
  765         error = SYSCTL_OUT(req, newoid, j * sizeof (int));
  766         return (error);
  767 }
  768 
  769 static SYSCTL_NODE(_sysctl, 2, next, CTLFLAG_RD, sysctl_sysctl_next, "");
  770 
  771 static int
  772 name2oid(char *name, int *oid, int *len, struct sysctl_oid **oidpp)
  773 {
  774         struct sysctl_oid *oidp;
  775         struct sysctl_oid_list *lsp = &sysctl__children;
  776         char *p;
  777 
  778         SYSCTL_ASSERT_XLOCKED();
  779 
  780         for (*len = 0; *len < CTL_MAXNAME;) {
  781                 p = strsep(&name, ".");
  782 
  783                 oidp = SLIST_FIRST(lsp);
  784                 for (;; oidp = SLIST_NEXT(oidp, oid_link)) {
  785                         if (oidp == NULL)
  786                                 return (ENOENT);
  787                         if (strcmp(p, oidp->oid_name) == 0)
  788                                 break;
  789                 }
  790                 *oid++ = oidp->oid_number;
  791                 (*len)++;
  792 
  793                 if (name == NULL || *name == '\0') {
  794                         if (oidpp)
  795                                 *oidpp = oidp;
  796                         return (0);
  797                 }
  798 
  799                 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
  800                         break;
  801 
  802                 if (oidp->oid_handler)
  803                         break;
  804 
  805                 lsp = SYSCTL_CHILDREN(oidp);
  806         }
  807         return (ENOENT);
  808 }
  809 
  810 static int
  811 sysctl_sysctl_name2oid(SYSCTL_HANDLER_ARGS)
  812 {
  813         char *p;
  814         int error, oid[CTL_MAXNAME], len;
  815         struct sysctl_oid *op = 0;
  816 
  817         if (!req->newlen) 
  818                 return (ENOENT);
  819         if (req->newlen >= MAXPATHLEN)  /* XXX arbitrary, undocumented */
  820                 return (ENAMETOOLONG);
  821 
  822         p = malloc(req->newlen+1, M_SYSCTL, M_WAITOK);
  823 
  824         error = SYSCTL_IN(req, p, req->newlen);
  825         if (error) {
  826                 free(p, M_SYSCTL);
  827                 return (error);
  828         }
  829 
  830         p [req->newlen] = '\0';
  831         len = 0;
  832         SYSCTL_XLOCK();
  833         error = name2oid(p, oid, &len, &op);
  834         SYSCTL_XUNLOCK();
  835 
  836         free(p, M_SYSCTL);
  837 
  838         if (error)
  839                 return (error);
  840 
  841         error = SYSCTL_OUT(req, oid, len * sizeof *oid);
  842         return (error);
  843 }
  844 
  845 SYSCTL_PROC(_sysctl, 3, name2oid, CTLFLAG_RW|CTLFLAG_ANYBODY|CTLFLAG_MPSAFE,
  846     0, 0, sysctl_sysctl_name2oid, "I", "");
  847 
  848 static int
  849 sysctl_sysctl_oidfmt(SYSCTL_HANDLER_ARGS)
  850 {
  851         struct sysctl_oid *oid;
  852         int error;
  853 
  854         SYSCTL_XLOCK();
  855         error = sysctl_find_oid(arg1, arg2, &oid, NULL, req);
  856         if (error)
  857                 goto out;
  858 
  859         if (oid->oid_fmt == NULL) {
  860                 error = ENOENT;
  861                 goto out;
  862         }
  863         error = SYSCTL_OUT(req, &oid->oid_kind, sizeof(oid->oid_kind));
  864         if (error)
  865                 goto out;
  866         error = SYSCTL_OUT(req, oid->oid_fmt, strlen(oid->oid_fmt) + 1);
  867  out:
  868         SYSCTL_XUNLOCK();
  869         return (error);
  870 }
  871 
  872 
  873 static SYSCTL_NODE(_sysctl, 4, oidfmt, CTLFLAG_RD|CTLFLAG_MPSAFE,
  874     sysctl_sysctl_oidfmt, "");
  875 
  876 static int
  877 sysctl_sysctl_oiddescr(SYSCTL_HANDLER_ARGS)
  878 {
  879         struct sysctl_oid *oid;
  880         int error;
  881 
  882         SYSCTL_XLOCK();
  883         error = sysctl_find_oid(arg1, arg2, &oid, NULL, req);
  884         if (error)
  885                 goto out;
  886 
  887         if (oid->oid_descr == NULL) {
  888                 error = ENOENT;
  889                 goto out;
  890         }
  891         error = SYSCTL_OUT(req, oid->oid_descr, strlen(oid->oid_descr) + 1);
  892  out:
  893         SYSCTL_XUNLOCK();
  894         return (error);
  895 }
  896 
  897 static SYSCTL_NODE(_sysctl, 5, oiddescr, CTLFLAG_RD, sysctl_sysctl_oiddescr, "");
  898 
  899 /*
  900  * Default "handler" functions.
  901  */
  902 
  903 /*
  904  * Handle an int, signed or unsigned.
  905  * Two cases:
  906  *     a variable:  point arg1 at it.
  907  *     a constant:  pass it in arg2.
  908  */
  909 
  910 int
  911 sysctl_handle_int(SYSCTL_HANDLER_ARGS)
  912 {
  913         int tmpout, error = 0;
  914 
  915         /*
  916          * Attempt to get a coherent snapshot by making a copy of the data.
  917          */
  918         if (arg1)
  919                 tmpout = *(int *)arg1;
  920         else
  921                 tmpout = arg2;
  922         error = SYSCTL_OUT(req, &tmpout, sizeof(int));
  923 
  924         if (error || !req->newptr)
  925                 return (error);
  926 
  927         if (!arg1)
  928                 error = EPERM;
  929         else
  930                 error = SYSCTL_IN(req, arg1, sizeof(int));
  931         return (error);
  932 }
  933 
  934 /*
  935  * Based on on sysctl_handle_int() convert milliseconds into ticks.
  936  * Note: this is used by TCP.
  937  */
  938 
  939 int
  940 sysctl_msec_to_ticks(SYSCTL_HANDLER_ARGS)
  941 {
  942         int error, s, tt;
  943 
  944         tt = *(int *)arg1;
  945         s = (int)((int64_t)tt * 1000 / hz);
  946 
  947         error = sysctl_handle_int(oidp, &s, 0, req);
  948         if (error || !req->newptr)
  949                 return (error);
  950 
  951         tt = (int)((int64_t)s * hz / 1000);
  952         if (tt < 1)
  953                 return (EINVAL);
  954 
  955         *(int *)arg1 = tt;
  956         return (0);
  957 }
  958 
  959 
  960 /*
  961  * Handle a long, signed or unsigned.
  962  * Two cases:
  963  *     a variable:  point arg1 at it.
  964  *     a constant:  pass it in arg2.
  965  */
  966 
  967 int
  968 sysctl_handle_long(SYSCTL_HANDLER_ARGS)
  969 {
  970         int error = 0;
  971         long tmplong;
  972 #ifdef SCTL_MASK32
  973         int tmpint;
  974 #endif
  975 
  976         /*
  977          * Attempt to get a coherent snapshot by making a copy of the data.
  978          */
  979         if (arg1)
  980                 tmplong = *(long *)arg1;
  981         else
  982                 tmplong = arg2;
  983 #ifdef SCTL_MASK32
  984         if (req->flags & SCTL_MASK32) {
  985                 tmpint = tmplong;
  986                 error = SYSCTL_OUT(req, &tmpint, sizeof(int));
  987         } else
  988 #endif
  989                 error = SYSCTL_OUT(req, &tmplong, sizeof(long));
  990 
  991         if (error || !req->newptr)
  992                 return (error);
  993 
  994         if (!arg1)
  995                 error = EPERM;
  996 #ifdef SCTL_MASK32
  997         else if (req->flags & SCTL_MASK32) {
  998                 error = SYSCTL_IN(req, &tmpint, sizeof(int));
  999                 *(long *)arg1 = (long)tmpint;
 1000         }
 1001 #endif
 1002         else
 1003                 error = SYSCTL_IN(req, arg1, sizeof(long));
 1004         return (error);
 1005 }
 1006 
 1007 /*
 1008  * Handle a 64 bit int, signed or unsigned.
 1009  * Two cases:
 1010  *     a variable:  point arg1 at it.
 1011  *     a constant:  pass it in arg2.
 1012  */
 1013 
 1014 int
 1015 sysctl_handle_quad(SYSCTL_HANDLER_ARGS)
 1016 {
 1017         int error = 0;
 1018         uint64_t tmpout;
 1019 
 1020         /*
 1021          * Attempt to get a coherent snapshot by making a copy of the data.
 1022          */
 1023         if (arg1)
 1024                 tmpout = *(uint64_t *)arg1;
 1025         else
 1026                 tmpout = arg2;
 1027         error = SYSCTL_OUT(req, &tmpout, sizeof(uint64_t));
 1028 
 1029         if (error || !req->newptr)
 1030                 return (error);
 1031 
 1032         if (!arg1)
 1033                 error = EPERM;
 1034         else
 1035                 error = SYSCTL_IN(req, arg1, sizeof(uint64_t));
 1036         return (error);
 1037 }
 1038 
 1039 /*
 1040  * Handle our generic '\0' terminated 'C' string.
 1041  * Two cases:
 1042  *      a variable string:  point arg1 at it, arg2 is max length.
 1043  *      a constant string:  point arg1 at it, arg2 is zero.
 1044  */
 1045 
 1046 int
 1047 sysctl_handle_string(SYSCTL_HANDLER_ARGS)
 1048 {
 1049         int error=0;
 1050         char *tmparg;
 1051         size_t outlen;
 1052 
 1053         /*
 1054          * Attempt to get a coherent snapshot by copying to a
 1055          * temporary kernel buffer.
 1056          */
 1057 retry:
 1058         outlen = strlen((char *)arg1)+1;
 1059         tmparg = malloc(outlen, M_SYSCTLTMP, M_WAITOK);
 1060 
 1061         if (strlcpy(tmparg, (char *)arg1, outlen) >= outlen) {
 1062                 free(tmparg, M_SYSCTLTMP);
 1063                 goto retry;
 1064         }
 1065 
 1066         error = SYSCTL_OUT(req, tmparg, outlen);
 1067         free(tmparg, M_SYSCTLTMP);
 1068 
 1069         if (error || !req->newptr)
 1070                 return (error);
 1071 
 1072         if ((req->newlen - req->newidx) >= arg2) {
 1073                 error = EINVAL;
 1074         } else {
 1075                 arg2 = (req->newlen - req->newidx);
 1076                 error = SYSCTL_IN(req, arg1, arg2);
 1077                 ((char *)arg1)[arg2] = '\0';
 1078         }
 1079 
 1080         return (error);
 1081 }
 1082 
 1083 /*
 1084  * Handle any kind of opaque data.
 1085  * arg1 points to it, arg2 is the size.
 1086  */
 1087 
 1088 int
 1089 sysctl_handle_opaque(SYSCTL_HANDLER_ARGS)
 1090 {
 1091         int error, tries;
 1092         u_int generation;
 1093         struct sysctl_req req2;
 1094 
 1095         /*
 1096          * Attempt to get a coherent snapshot, by using the thread
 1097          * pre-emption counter updated from within mi_switch() to
 1098          * determine if we were pre-empted during a bcopy() or
 1099          * copyout(). Make 3 attempts at doing this before giving up.
 1100          * If we encounter an error, stop immediately.
 1101          */
 1102         tries = 0;
 1103         req2 = *req;
 1104 retry:
 1105         generation = curthread->td_generation;
 1106         error = SYSCTL_OUT(req, arg1, arg2);
 1107         if (error)
 1108                 return (error);
 1109         tries++;
 1110         if (generation != curthread->td_generation && tries < 3) {
 1111                 *req = req2;
 1112                 goto retry;
 1113         }
 1114 
 1115         error = SYSCTL_IN(req, arg1, arg2);
 1116 
 1117         return (error);
 1118 }
 1119 
 1120 /*
 1121  * Transfer functions to/from kernel space.
 1122  * XXX: rather untested at this point
 1123  */
 1124 static int
 1125 sysctl_old_kernel(struct sysctl_req *req, const void *p, size_t l)
 1126 {
 1127         size_t i = 0;
 1128 
 1129         if (req->oldptr) {
 1130                 i = l;
 1131                 if (req->oldlen <= req->oldidx)
 1132                         i = 0;
 1133                 else
 1134                         if (i > req->oldlen - req->oldidx)
 1135                                 i = req->oldlen - req->oldidx;
 1136                 if (i > 0)
 1137                         bcopy(p, (char *)req->oldptr + req->oldidx, i);
 1138         }
 1139         req->oldidx += l;
 1140         if (req->oldptr && i != l)
 1141                 return (ENOMEM);
 1142         return (0);
 1143 }
 1144 
 1145 static int
 1146 sysctl_new_kernel(struct sysctl_req *req, void *p, size_t l)
 1147 {
 1148         if (!req->newptr)
 1149                 return (0);
 1150         if (req->newlen - req->newidx < l)
 1151                 return (EINVAL);
 1152         bcopy((char *)req->newptr + req->newidx, p, l);
 1153         req->newidx += l;
 1154         return (0);
 1155 }
 1156 
 1157 int
 1158 kernel_sysctl(struct thread *td, int *name, u_int namelen, void *old,
 1159     size_t *oldlenp, void *new, size_t newlen, size_t *retval, int flags)
 1160 {
 1161         int error = 0;
 1162         struct sysctl_req req;
 1163 
 1164         bzero(&req, sizeof req);
 1165 
 1166         req.td = td;
 1167         req.flags = flags;
 1168 
 1169         if (oldlenp) {
 1170                 req.oldlen = *oldlenp;
 1171         }
 1172         req.validlen = req.oldlen;
 1173 
 1174         if (old) {
 1175                 req.oldptr= old;
 1176         }
 1177 
 1178         if (new != NULL) {
 1179                 req.newlen = newlen;
 1180                 req.newptr = new;
 1181         }
 1182 
 1183         req.oldfunc = sysctl_old_kernel;
 1184         req.newfunc = sysctl_new_kernel;
 1185         req.lock = REQ_LOCKED;
 1186 
 1187         SYSCTL_XLOCK();
 1188         error = sysctl_root(0, name, namelen, &req);
 1189         SYSCTL_XUNLOCK();
 1190 
 1191         if (req.lock == REQ_WIRED && req.validlen > 0)
 1192                 vsunlock(req.oldptr, req.validlen);
 1193 
 1194         if (error && error != ENOMEM)
 1195                 return (error);
 1196 
 1197         if (retval) {
 1198                 if (req.oldptr && req.oldidx > req.validlen)
 1199                         *retval = req.validlen;
 1200                 else
 1201                         *retval = req.oldidx;
 1202         }
 1203         return (error);
 1204 }
 1205 
 1206 int
 1207 kernel_sysctlbyname(struct thread *td, char *name, void *old, size_t *oldlenp,
 1208     void *new, size_t newlen, size_t *retval, int flags)
 1209 {
 1210         int oid[CTL_MAXNAME];
 1211         size_t oidlen, plen;
 1212         int error;
 1213 
 1214         oid[0] = 0;             /* sysctl internal magic */
 1215         oid[1] = 3;             /* name2oid */
 1216         oidlen = sizeof(oid);
 1217 
 1218         error = kernel_sysctl(td, oid, 2, oid, &oidlen,
 1219             (void *)name, strlen(name), &plen, flags);
 1220         if (error)
 1221                 return (error);
 1222 
 1223         error = kernel_sysctl(td, oid, plen / sizeof(int), old, oldlenp,
 1224             new, newlen, retval, flags);
 1225         return (error);
 1226 }
 1227 
 1228 /*
 1229  * Transfer function to/from user space.
 1230  */
 1231 static int
 1232 sysctl_old_user(struct sysctl_req *req, const void *p, size_t l)
 1233 {
 1234         size_t i, len, origidx;
 1235         int error;
 1236 
 1237         origidx = req->oldidx;
 1238         req->oldidx += l;
 1239         if (req->oldptr == NULL)
 1240                 return (0);
 1241         /*
 1242          * If we have not wired the user supplied buffer and we are currently
 1243          * holding locks, drop a witness warning, as it's possible that
 1244          * write operations to the user page can sleep.
 1245          */
 1246         if (req->lock != REQ_WIRED)
 1247                 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,
 1248                     "sysctl_old_user()");
 1249         i = l;
 1250         len = req->validlen;
 1251         if (len <= origidx)
 1252                 i = 0;
 1253         else {
 1254                 if (i > len - origidx)
 1255                         i = len - origidx;
 1256                 if (req->lock == REQ_WIRED) {
 1257                         error = copyout_nofault(p, (char *)req->oldptr +
 1258                             origidx, i);
 1259                 } else
 1260                         error = copyout(p, (char *)req->oldptr + origidx, i);
 1261                 if (error != 0)
 1262                         return (error);
 1263         }
 1264         if (i < l)
 1265                 return (ENOMEM);
 1266         return (0);
 1267 }
 1268 
 1269 static int
 1270 sysctl_new_user(struct sysctl_req *req, void *p, size_t l)
 1271 {
 1272         int error;
 1273 
 1274         if (!req->newptr)
 1275                 return (0);
 1276         if (req->newlen - req->newidx < l)
 1277                 return (EINVAL);
 1278         WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,
 1279             "sysctl_new_user()");
 1280         error = copyin((char *)req->newptr + req->newidx, p, l);
 1281         req->newidx += l;
 1282         return (error);
 1283 }
 1284 
 1285 /*
 1286  * Wire the user space destination buffer.  If set to a value greater than
 1287  * zero, the len parameter limits the maximum amount of wired memory.
 1288  */
 1289 int
 1290 sysctl_wire_old_buffer(struct sysctl_req *req, size_t len)
 1291 {
 1292         int ret;
 1293         size_t wiredlen;
 1294 
 1295         wiredlen = (len > 0 && len < req->oldlen) ? len : req->oldlen;
 1296         ret = 0;
 1297         if (req->lock == REQ_LOCKED && req->oldptr &&
 1298             req->oldfunc == sysctl_old_user) {
 1299                 if (wiredlen != 0) {
 1300                         ret = vslock(req->oldptr, wiredlen);
 1301                         if (ret != 0) {
 1302                                 if (ret != ENOMEM)
 1303                                         return (ret);
 1304                                 wiredlen = 0;
 1305                         }
 1306                 }
 1307                 req->lock = REQ_WIRED;
 1308                 req->validlen = wiredlen;
 1309         }
 1310         return (0);
 1311 }
 1312 
 1313 int
 1314 sysctl_find_oid(int *name, u_int namelen, struct sysctl_oid **noid,
 1315     int *nindx, struct sysctl_req *req)
 1316 {
 1317         struct sysctl_oid_list *lsp;
 1318         struct sysctl_oid *oid;
 1319         int indx;
 1320 
 1321         SYSCTL_ASSERT_XLOCKED();
 1322         lsp = &sysctl__children;
 1323         indx = 0;
 1324         while (indx < CTL_MAXNAME) {
 1325                 SLIST_FOREACH(oid, lsp, oid_link) {
 1326                         if (oid->oid_number == name[indx])
 1327                                 break;
 1328                 }
 1329                 if (oid == NULL)
 1330                         return (ENOENT);
 1331 
 1332                 indx++;
 1333                 if (oid->oid_kind & CTLFLAG_NOLOCK)
 1334                         req->lock = REQ_UNLOCKED;
 1335                 if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
 1336                         if (oid->oid_handler != NULL || indx == namelen) {
 1337                                 *noid = oid;
 1338                                 if (nindx != NULL)
 1339                                         *nindx = indx;
 1340                                 KASSERT((oid->oid_kind & CTLFLAG_DYING) == 0,
 1341                                     ("%s found DYING node %p", __func__, oid));
 1342                                 return (0);
 1343                         }
 1344                         lsp = SYSCTL_CHILDREN(oid);
 1345                 } else if (indx == namelen) {
 1346                         *noid = oid;
 1347                         if (nindx != NULL)
 1348                                 *nindx = indx;
 1349                         KASSERT((oid->oid_kind & CTLFLAG_DYING) == 0,
 1350                             ("%s found DYING node %p", __func__, oid));
 1351                         return (0);
 1352                 } else {
 1353                         return (ENOTDIR);
 1354                 }
 1355         }
 1356         return (ENOENT);
 1357 }
 1358 
 1359 /*
 1360  * Traverse our tree, and find the right node, execute whatever it points
 1361  * to, and return the resulting error code.
 1362  */
 1363 
 1364 static int
 1365 sysctl_root(SYSCTL_HANDLER_ARGS)
 1366 {
 1367         struct sysctl_oid *oid;
 1368         int error, indx, lvl;
 1369 
 1370         SYSCTL_ASSERT_XLOCKED();
 1371 
 1372         error = sysctl_find_oid(arg1, arg2, &oid, &indx, req);
 1373         if (error)
 1374                 return (error);
 1375 
 1376         if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
 1377                 /*
 1378                  * You can't call a sysctl when it's a node, but has
 1379                  * no handler.  Inform the user that it's a node.
 1380                  * The indx may or may not be the same as namelen.
 1381                  */
 1382                 if (oid->oid_handler == NULL)
 1383                         return (EISDIR);
 1384         }
 1385 
 1386         /* Is this sysctl writable? */
 1387         if (req->newptr && !(oid->oid_kind & CTLFLAG_WR))
 1388                 return (EPERM);
 1389 
 1390         KASSERT(req->td != NULL, ("sysctl_root(): req->td == NULL"));
 1391 
 1392         /* Is this sysctl sensitive to securelevels? */
 1393         if (req->newptr && (oid->oid_kind & CTLFLAG_SECURE)) {
 1394                 lvl = (oid->oid_kind & CTLMASK_SECURE) >> CTLSHIFT_SECURE;
 1395                 error = securelevel_gt(req->td->td_ucred, lvl);
 1396                 if (error)
 1397                         return (error);
 1398         }
 1399 
 1400         /* Is this sysctl writable by only privileged users? */
 1401         if (req->newptr && !(oid->oid_kind & CTLFLAG_ANYBODY)) {
 1402                 int priv;
 1403 
 1404                 if (oid->oid_kind & CTLFLAG_PRISON)
 1405                         priv = PRIV_SYSCTL_WRITEJAIL;
 1406 #ifdef VIMAGE
 1407                 else if ((oid->oid_kind & CTLFLAG_VNET) &&
 1408                      prison_owns_vnet(req->td->td_ucred))
 1409                         priv = PRIV_SYSCTL_WRITEJAIL;
 1410 #endif
 1411                 else
 1412                         priv = PRIV_SYSCTL_WRITE;
 1413                 error = priv_check(req->td, priv);
 1414                 if (error)
 1415                         return (error);
 1416         }
 1417 
 1418         if (!oid->oid_handler)
 1419                 return (EINVAL);
 1420 
 1421         if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
 1422                 arg1 = (int *)arg1 + indx;
 1423                 arg2 -= indx;
 1424         } else {
 1425                 arg1 = oid->oid_arg1;
 1426                 arg2 = oid->oid_arg2;
 1427         }
 1428 #ifdef MAC
 1429         error = mac_system_check_sysctl(req->td->td_ucred, oid, arg1, arg2,
 1430             req);
 1431         if (error != 0)
 1432                 return (error);
 1433 #endif
 1434         oid->oid_running++;
 1435         SYSCTL_XUNLOCK();
 1436 
 1437         if (!(oid->oid_kind & CTLFLAG_MPSAFE))
 1438                 mtx_lock(&Giant);
 1439         error = oid->oid_handler(oid, arg1, arg2, req);
 1440         if (!(oid->oid_kind & CTLFLAG_MPSAFE))
 1441                 mtx_unlock(&Giant);
 1442 
 1443         KFAIL_POINT_ERROR(_debug_fail_point, sysctl_running, error);
 1444 
 1445         SYSCTL_XLOCK();
 1446         oid->oid_running--;
 1447         if (oid->oid_running == 0 && (oid->oid_kind & CTLFLAG_DYING) != 0)
 1448                 wakeup(&oid->oid_running);
 1449         return (error);
 1450 }
 1451 
 1452 #ifndef _SYS_SYSPROTO_H_
 1453 struct sysctl_args {
 1454         int     *name;
 1455         u_int   namelen;
 1456         void    *old;
 1457         size_t  *oldlenp;
 1458         void    *new;
 1459         size_t  newlen;
 1460 };
 1461 #endif
 1462 int
 1463 __sysctl(struct thread *td, struct sysctl_args *uap)
 1464 {
 1465         int error, i, name[CTL_MAXNAME];
 1466         size_t j;
 1467 
 1468         if (uap->namelen > CTL_MAXNAME || uap->namelen < 2)
 1469                 return (EINVAL);
 1470 
 1471         error = copyin(uap->name, &name, uap->namelen * sizeof(int));
 1472         if (error)
 1473                 return (error);
 1474 
 1475         error = userland_sysctl(td, name, uap->namelen,
 1476                 uap->old, uap->oldlenp, 0,
 1477                 uap->new, uap->newlen, &j, 0);
 1478         if (error && error != ENOMEM)
 1479                 return (error);
 1480         if (uap->oldlenp) {
 1481                 i = copyout(&j, uap->oldlenp, sizeof(j));
 1482                 if (i)
 1483                         return (i);
 1484         }
 1485         return (error);
 1486 }
 1487 
 1488 /*
 1489  * This is used from various compatibility syscalls too.  That's why name
 1490  * must be in kernel space.
 1491  */
 1492 int
 1493 userland_sysctl(struct thread *td, int *name, u_int namelen, void *old,
 1494     size_t *oldlenp, int inkernel, void *new, size_t newlen, size_t *retval,
 1495     int flags)
 1496 {
 1497         int error = 0, memlocked;
 1498         struct sysctl_req req;
 1499 
 1500         bzero(&req, sizeof req);
 1501 
 1502         req.td = td;
 1503         req.flags = flags;
 1504 
 1505         if (oldlenp) {
 1506                 if (inkernel) {
 1507                         req.oldlen = *oldlenp;
 1508                 } else {
 1509                         error = copyin(oldlenp, &req.oldlen, sizeof(*oldlenp));
 1510                         if (error)
 1511                                 return (error);
 1512                 }
 1513         }
 1514         req.validlen = req.oldlen;
 1515 
 1516         if (old) {
 1517                 if (!useracc(old, req.oldlen, VM_PROT_WRITE))
 1518                         return (EFAULT);
 1519                 req.oldptr= old;
 1520         }
 1521 
 1522         if (new != NULL) {
 1523                 if (!useracc(new, newlen, VM_PROT_READ))
 1524                         return (EFAULT);
 1525                 req.newlen = newlen;
 1526                 req.newptr = new;
 1527         }
 1528 
 1529         req.oldfunc = sysctl_old_user;
 1530         req.newfunc = sysctl_new_user;
 1531         req.lock = REQ_LOCKED;
 1532 
 1533 #ifdef KTRACE
 1534         if (KTRPOINT(curthread, KTR_SYSCTL))
 1535                 ktrsysctl(name, namelen);
 1536 #endif
 1537 
 1538         if (req.oldlen > PAGE_SIZE) {
 1539                 memlocked = 1;
 1540                 sx_xlock(&sysctlmemlock);
 1541         } else
 1542                 memlocked = 0;
 1543         CURVNET_SET(TD_TO_VNET(td));
 1544 
 1545         for (;;) {
 1546                 req.oldidx = 0;
 1547                 req.newidx = 0;
 1548                 SYSCTL_XLOCK();
 1549                 error = sysctl_root(0, name, namelen, &req);
 1550                 SYSCTL_XUNLOCK();
 1551                 if (error != EAGAIN)
 1552                         break;
 1553                 uio_yield();
 1554         }
 1555 
 1556         CURVNET_RESTORE();
 1557 
 1558         if (req.lock == REQ_WIRED && req.validlen > 0)
 1559                 vsunlock(req.oldptr, req.validlen);
 1560         if (memlocked)
 1561                 sx_xunlock(&sysctlmemlock);
 1562 
 1563         if (error && error != ENOMEM)
 1564                 return (error);
 1565 
 1566         if (retval) {
 1567                 if (req.oldptr && req.oldidx > req.validlen)
 1568                         *retval = req.validlen;
 1569                 else
 1570                         *retval = req.oldidx;
 1571         }
 1572         return (error);
 1573 }
 1574 
 1575 /*
 1576  * Drain into a sysctl struct.  The user buffer should be wired if a page
 1577  * fault would cause issue.
 1578  */
 1579 static int
 1580 sbuf_sysctl_drain(void *arg, const char *data, int len)
 1581 {
 1582         struct sysctl_req *req = arg;
 1583         int error;
 1584 
 1585         error = SYSCTL_OUT(req, data, len);
 1586         KASSERT(error >= 0, ("Got unexpected negative value %d", error));
 1587         return (error == 0 ? len : -error);
 1588 }
 1589 
 1590 struct sbuf *
 1591 sbuf_new_for_sysctl(struct sbuf *s, char *buf, int length,
 1592     struct sysctl_req *req)
 1593 {
 1594 
 1595         s = sbuf_new(s, buf, length, SBUF_FIXEDLEN);
 1596         sbuf_set_drain(s, sbuf_sysctl_drain, req);
 1597         return (s);
 1598 }
Cache object: 2dc48a5a7ad23effc8121cbfcfbfe253
[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]
This page is part of the FreeBSD/Linux Linux Kernel Cross-Reference, and was automatically generated using a modified version of the LXR engine.
FreeBSD/Linux Kernel Cross Reference sys/kern/kern_sysctl.c

FreeBSD/Linux Kernel Cross Reference
sys/kern/kern_sysctl.c
