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: releng/10.0/sys/kern/kern_sysctl.c 254115 2013-08-09 01:04:44Z scottl $");
   40 
   41 #include "opt_capsicum.h"
   42 #include "opt_compat.h"
   43 #include "opt_ktrace.h"
   44 
   45 #include <sys/param.h>
   46 #include <sys/fail.h>
   47 #include <sys/systm.h>
   48 #include <sys/capability.h>
   49 #include <sys/kernel.h>
   50 #include <sys/sysctl.h>
   51 #include <sys/malloc.h>
   52 #include <sys/priv.h>
   53 #include <sys/proc.h>
   54 #include <sys/jail.h>
   55 #include <sys/lock.h>
   56 #include <sys/mutex.h>
   57 #include <sys/sbuf.h>
   58 #include <sys/sx.h>
   59 #include <sys/sysproto.h>
   60 #include <sys/uio.h>
   61 #ifdef KTRACE
   62 #include <sys/ktrace.h>
   63 #endif
   64 
   65 #include <net/vnet.h>
   66 
   67 #include <security/mac/mac_framework.h>
   68 
   69 #include <vm/vm.h>
   70 #include <vm/vm_extern.h>
   71 
   72 static MALLOC_DEFINE(M_SYSCTL, "sysctl", "sysctl internal magic");
   73 static MALLOC_DEFINE(M_SYSCTLOID, "sysctloid", "sysctl dynamic oids");
   74 static MALLOC_DEFINE(M_SYSCTLTMP, "sysctltmp", "sysctl temp output buffer");
   75 
   76 /*
   77  * The sysctllock protects the MIB tree.  It also protects sysctl
   78  * contexts used with dynamic sysctls.  The sysctl_register_oid() and
   79  * sysctl_unregister_oid() routines require the sysctllock to already
   80  * be held, so the sysctl_lock() and sysctl_unlock() routines are
   81  * provided for the few places in the kernel which need to use that
   82  * API rather than using the dynamic API.  Use of the dynamic API is
   83  * strongly encouraged for most code.
   84  *
   85  * The sysctlmemlock is used to limit the amount of user memory wired for
   86  * sysctl requests.  This is implemented by serializing any userland
   87  * sysctl requests larger than a single page via an exclusive lock.
   88  */
   89 static struct sx sysctllock;
   90 static struct sx sysctlmemlock;
   91 
   92 #define SYSCTL_XLOCK()          sx_xlock(&sysctllock)
   93 #define SYSCTL_XUNLOCK()        sx_xunlock(&sysctllock)
   94 #define SYSCTL_ASSERT_XLOCKED() sx_assert(&sysctllock, SA_XLOCKED)
   95 #define SYSCTL_INIT()           sx_init(&sysctllock, "sysctl lock")
   96 #define SYSCTL_SLEEP(ch, wmesg, timo)                                   \
   97                                 sx_sleep(ch, &sysctllock, 0, wmesg, timo)
   98 
   99 static int sysctl_root(SYSCTL_HANDLER_ARGS);
  100 
  101 struct sysctl_oid_list sysctl__children; /* root list */
  102 
  103 static int      sysctl_remove_oid_locked(struct sysctl_oid *oidp, int del,
  104                     int recurse);
  105 
  106 static struct sysctl_oid *
  107 sysctl_find_oidname(const char *name, struct sysctl_oid_list *list)
  108 {
  109         struct sysctl_oid *oidp;
  110 
  111         SYSCTL_ASSERT_XLOCKED();
  112         SLIST_FOREACH(oidp, list, oid_link) {
  113                 if (strcmp(oidp->oid_name, name) == 0) {
  114                         return (oidp);
  115                 }
  116         }
  117         return (NULL);
  118 }
  119 
  120 /*
  121  * Initialization of the MIB tree.
  122  *
  123  * Order by number in each list.
  124  */
  125 void
  126 sysctl_lock(void)
  127 {
  128 
  129         SYSCTL_XLOCK();
  130 }
  131 
  132 void
  133 sysctl_unlock(void)
  134 {
  135 
  136         SYSCTL_XUNLOCK();
  137 }
  138 
  139 void
  140 sysctl_register_oid(struct sysctl_oid *oidp)
  141 {
  142         struct sysctl_oid_list *parent = oidp->oid_parent;
  143         struct sysctl_oid *p;
  144         struct sysctl_oid *q;
  145 
  146         /*
  147          * First check if another oid with the same name already
  148          * exists in the parent's list.
  149          */
  150         SYSCTL_ASSERT_XLOCKED();
  151         p = sysctl_find_oidname(oidp->oid_name, parent);
  152         if (p != NULL) {
  153                 if ((p->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
  154                         p->oid_refcnt++;
  155                         return;
  156                 } else {
  157                         printf("can't re-use a leaf (%s)!\n", p->oid_name);
  158                         return;
  159                 }
  160         }
  161         /*
  162          * If this oid has a number OID_AUTO, give it a number which
  163          * is greater than any current oid.
  164          * NOTE: DO NOT change the starting value here, change it in
  165          * <sys/sysctl.h>, and make sure it is at least 256 to
  166          * accomodate e.g. net.inet.raw as a static sysctl node.
  167          */
  168         if (oidp->oid_number == OID_AUTO) {
  169                 static int newoid = CTL_AUTO_START;
  170 
  171                 oidp->oid_number = newoid++;
  172                 if (newoid == 0x7fffffff)
  173                         panic("out of oids");
  174         }
  175 #if 0
  176         else if (oidp->oid_number >= CTL_AUTO_START) {
  177                 /* do not panic; this happens when unregistering sysctl sets */
  178                 printf("static sysctl oid too high: %d", oidp->oid_number);
  179         }
  180 #endif
  181 
  182         /*
  183          * Insert the oid into the parent's list in order.
  184          */
  185         q = NULL;
  186         SLIST_FOREACH(p, parent, oid_link) {
  187                 if (oidp->oid_number < p->oid_number)
  188                         break;
  189                 q = p;
  190         }
  191         if (q)
  192                 SLIST_INSERT_AFTER(q, oidp, oid_link);
  193         else
  194                 SLIST_INSERT_HEAD(parent, oidp, oid_link);
  195 }
  196 
  197 void
  198 sysctl_unregister_oid(struct sysctl_oid *oidp)
  199 {
  200         struct sysctl_oid *p;
  201         int error;
  202 
  203         SYSCTL_ASSERT_XLOCKED();
  204         error = ENOENT;
  205         if (oidp->oid_number == OID_AUTO) {
  206                 error = EINVAL;
  207         } else {
  208                 SLIST_FOREACH(p, oidp->oid_parent, oid_link) {
  209                         if (p == oidp) {
  210                                 SLIST_REMOVE(oidp->oid_parent, oidp,
  211                                     sysctl_oid, oid_link);
  212                                 error = 0;
  213                                 break;
  214                         }
  215                 }
  216         }
  217 
  218         /* 
  219          * This can happen when a module fails to register and is
  220          * being unloaded afterwards.  It should not be a panic()
  221          * for normal use.
  222          */
  223         if (error)
  224                 printf("%s: failed to unregister sysctl\n", __func__);
  225 }
  226 
  227 /* Initialize a new context to keep track of dynamically added sysctls. */
  228 int
  229 sysctl_ctx_init(struct sysctl_ctx_list *c)
  230 {
  231 
  232         if (c == NULL) {
  233                 return (EINVAL);
  234         }
  235 
  236         /*
  237          * No locking here, the caller is responsible for not adding
  238          * new nodes to a context until after this function has
  239          * returned.
  240          */
  241         TAILQ_INIT(c);
  242         return (0);
  243 }
  244 
  245 /* Free the context, and destroy all dynamic oids registered in this context */
  246 int
  247 sysctl_ctx_free(struct sysctl_ctx_list *clist)
  248 {
  249         struct sysctl_ctx_entry *e, *e1;
  250         int error;
  251 
  252         error = 0;
  253         /*
  254          * First perform a "dry run" to check if it's ok to remove oids.
  255          * XXX FIXME
  256          * XXX This algorithm is a hack. But I don't know any
  257          * XXX better solution for now...
  258          */
  259         SYSCTL_XLOCK();
  260         TAILQ_FOREACH(e, clist, link) {
  261                 error = sysctl_remove_oid_locked(e->entry, 0, 0);
  262                 if (error)
  263                         break;
  264         }
  265         /*
  266          * Restore deregistered entries, either from the end,
  267          * or from the place where error occured.
  268          * e contains the entry that was not unregistered
  269          */
  270         if (error)
  271                 e1 = TAILQ_PREV(e, sysctl_ctx_list, link);
  272         else
  273                 e1 = TAILQ_LAST(clist, sysctl_ctx_list);
  274         while (e1 != NULL) {
  275                 sysctl_register_oid(e1->entry);
  276                 e1 = TAILQ_PREV(e1, sysctl_ctx_list, link);
  277         }
  278         if (error) {
  279                 SYSCTL_XUNLOCK();
  280                 return(EBUSY);
  281         }
  282         /* Now really delete the entries */
  283         e = TAILQ_FIRST(clist);
  284         while (e != NULL) {
  285                 e1 = TAILQ_NEXT(e, link);
  286                 error = sysctl_remove_oid_locked(e->entry, 1, 0);
  287                 if (error)
  288                         panic("sysctl_remove_oid: corrupt tree, entry: %s",
  289                             e->entry->oid_name);
  290                 free(e, M_SYSCTLOID);
  291                 e = e1;
  292         }
  293         SYSCTL_XUNLOCK();
  294         return (error);
  295 }
  296 
  297 /* Add an entry to the context */
  298 struct sysctl_ctx_entry *
  299 sysctl_ctx_entry_add(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
  300 {
  301         struct sysctl_ctx_entry *e;
  302 
  303         SYSCTL_ASSERT_XLOCKED();
  304         if (clist == NULL || oidp == NULL)
  305                 return(NULL);
  306         e = malloc(sizeof(struct sysctl_ctx_entry), M_SYSCTLOID, M_WAITOK);
  307         e->entry = oidp;
  308         TAILQ_INSERT_HEAD(clist, e, link);
  309         return (e);
  310 }
  311 
  312 /* Find an entry in the context */
  313 struct sysctl_ctx_entry *
  314 sysctl_ctx_entry_find(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
  315 {
  316         struct sysctl_ctx_entry *e;
  317 
  318         SYSCTL_ASSERT_XLOCKED();
  319         if (clist == NULL || oidp == NULL)
  320                 return(NULL);
  321         TAILQ_FOREACH(e, clist, link) {
  322                 if(e->entry == oidp)
  323                         return(e);
  324         }
  325         return (e);
  326 }
  327 
  328 /*
  329  * Delete an entry from the context.
  330  * NOTE: this function doesn't free oidp! You have to remove it
  331  * with sysctl_remove_oid().
  332  */
  333 int
  334 sysctl_ctx_entry_del(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
  335 {
  336         struct sysctl_ctx_entry *e;
  337 
  338         if (clist == NULL || oidp == NULL)
  339                 return (EINVAL);
  340         SYSCTL_XLOCK();
  341         e = sysctl_ctx_entry_find(clist, oidp);
  342         if (e != NULL) {
  343                 TAILQ_REMOVE(clist, e, link);
  344                 SYSCTL_XUNLOCK();
  345                 free(e, M_SYSCTLOID);
  346                 return (0);
  347         } else {
  348                 SYSCTL_XUNLOCK();
  349                 return (ENOENT);
  350         }
  351 }
  352 
  353 /*
  354  * Remove dynamically created sysctl trees.
  355  * oidp - top of the tree to be removed
  356  * del - if 0 - just deregister, otherwise free up entries as well
  357  * recurse - if != 0 traverse the subtree to be deleted
  358  */
  359 int
  360 sysctl_remove_oid(struct sysctl_oid *oidp, int del, int recurse)
  361 {
  362         int error;
  363 
  364         SYSCTL_XLOCK();
  365         error = sysctl_remove_oid_locked(oidp, del, recurse);
  366         SYSCTL_XUNLOCK();
  367         return (error);
  368 }
  369 
  370 int
  371 sysctl_remove_name(struct sysctl_oid *parent, const char *name,
  372     int del, int recurse)
  373 {
  374         struct sysctl_oid *p, *tmp;
  375         int error;
  376 
  377         error = ENOENT;
  378         SYSCTL_XLOCK();
  379         SLIST_FOREACH_SAFE(p, SYSCTL_CHILDREN(parent), oid_link, tmp) {
  380                 if (strcmp(p->oid_name, name) == 0) {
  381                         error = sysctl_remove_oid_locked(p, del, recurse);
  382                         break;
  383                 }
  384         }
  385         SYSCTL_XUNLOCK();
  386 
  387         return (error);
  388 }
  389 
  390 
  391 static int
  392 sysctl_remove_oid_locked(struct sysctl_oid *oidp, int del, int recurse)
  393 {
  394         struct sysctl_oid *p, *tmp;
  395         int error;
  396 
  397         SYSCTL_ASSERT_XLOCKED();
  398         if (oidp == NULL)
  399                 return(EINVAL);
  400         if ((oidp->oid_kind & CTLFLAG_DYN) == 0) {
  401                 printf("can't remove non-dynamic nodes!\n");
  402                 return (EINVAL);
  403         }
  404         /*
  405          * WARNING: normal method to do this should be through
  406          * sysctl_ctx_free(). Use recursing as the last resort
  407          * method to purge your sysctl tree of leftovers...
  408          * However, if some other code still references these nodes,
  409          * it will panic.
  410          */
  411         if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
  412                 if (oidp->oid_refcnt == 1) {
  413                         SLIST_FOREACH_SAFE(p,
  414                             SYSCTL_CHILDREN(oidp), oid_link, tmp) {
  415                                 if (!recurse) {
  416                                         printf("Warning: failed attempt to "
  417                                             "remove oid %s with child %s\n",
  418                                             oidp->oid_name, p->oid_name);
  419                                         return (ENOTEMPTY);
  420                                 }
  421                                 error = sysctl_remove_oid_locked(p, del,
  422                                     recurse);
  423                                 if (error)
  424                                         return (error);
  425                         }
  426                         if (del)
  427                                 free(SYSCTL_CHILDREN(oidp), M_SYSCTLOID);
  428                 }
  429         }
  430         if (oidp->oid_refcnt > 1 ) {
  431                 oidp->oid_refcnt--;
  432         } else {
  433                 if (oidp->oid_refcnt == 0) {
  434                         printf("Warning: bad oid_refcnt=%u (%s)!\n",
  435                                 oidp->oid_refcnt, oidp->oid_name);
  436                         return (EINVAL);
  437                 }
  438                 sysctl_unregister_oid(oidp);
  439                 if (del) {
  440                         /*
  441                          * Wait for all threads running the handler to drain.
  442                          * This preserves the previous behavior when the
  443                          * sysctl lock was held across a handler invocation,
  444                          * and is necessary for module unload correctness.
  445                          */
  446                         while (oidp->oid_running > 0) {
  447                                 oidp->oid_kind |= CTLFLAG_DYING;
  448                                 SYSCTL_SLEEP(&oidp->oid_running, "oidrm", 0);
  449                         }
  450                         if (oidp->oid_descr)
  451                                 free(__DECONST(char *, oidp->oid_descr),
  452                                     M_SYSCTLOID);
  453                         free(__DECONST(char *, oidp->oid_name), M_SYSCTLOID);
  454                         free(oidp, M_SYSCTLOID);
  455                 }
  456         }
  457         return (0);
  458 }
  459 /*
  460  * Create new sysctls at run time.
  461  * clist may point to a valid context initialized with sysctl_ctx_init().
  462  */
  463 struct sysctl_oid *
  464 sysctl_add_oid(struct sysctl_ctx_list *clist, struct sysctl_oid_list *parent,
  465         int number, const char *name, int kind, void *arg1, intptr_t arg2,
  466         int (*handler)(SYSCTL_HANDLER_ARGS), const char *fmt, const char *descr)
  467 {
  468         struct sysctl_oid *oidp;
  469 
  470         /* You have to hook up somewhere.. */
  471         if (parent == NULL)
  472                 return(NULL);
  473         /* Check if the node already exists, otherwise create it */
  474         SYSCTL_XLOCK();
  475         oidp = sysctl_find_oidname(name, parent);
  476         if (oidp != NULL) {
  477                 if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
  478                         oidp->oid_refcnt++;
  479                         /* Update the context */
  480                         if (clist != NULL)
  481                                 sysctl_ctx_entry_add(clist, oidp);
  482                         SYSCTL_XUNLOCK();
  483                         return (oidp);
  484                 } else {
  485                         SYSCTL_XUNLOCK();
  486                         printf("can't re-use a leaf (%s)!\n", name);
  487                         return (NULL);
  488                 }
  489         }
  490         oidp = malloc(sizeof(struct sysctl_oid), M_SYSCTLOID, M_WAITOK|M_ZERO);
  491         oidp->oid_parent = parent;
  492         SLIST_NEXT(oidp, oid_link) = NULL;
  493         oidp->oid_number = number;
  494         oidp->oid_refcnt = 1;
  495         oidp->oid_name = strdup(name, M_SYSCTLOID);
  496         oidp->oid_handler = handler;
  497         oidp->oid_kind = CTLFLAG_DYN | kind;
  498         if ((kind & CTLTYPE) == CTLTYPE_NODE) {
  499                 /* Allocate space for children */
  500                 SYSCTL_CHILDREN_SET(oidp, malloc(sizeof(struct sysctl_oid_list),
  501                     M_SYSCTLOID, M_WAITOK));
  502                 SLIST_INIT(SYSCTL_CHILDREN(oidp));
  503                 oidp->oid_arg2 = arg2;
  504         } else {
  505                 oidp->oid_arg1 = arg1;
  506                 oidp->oid_arg2 = arg2;
  507         }
  508         oidp->oid_fmt = fmt;
  509         if (descr)
  510                 oidp->oid_descr = strdup(descr, M_SYSCTLOID);
  511         /* Update the context, if used */
  512         if (clist != NULL)
  513                 sysctl_ctx_entry_add(clist, oidp);
  514         /* Register this oid */
  515         sysctl_register_oid(oidp);
  516         SYSCTL_XUNLOCK();
  517         return (oidp);
  518 }
  519 
  520 /*
  521  * Rename an existing oid.
  522  */
  523 void
  524 sysctl_rename_oid(struct sysctl_oid *oidp, const char *name)
  525 {
  526         char *newname;
  527         char *oldname;
  528 
  529         newname = strdup(name, M_SYSCTLOID);
  530         SYSCTL_XLOCK();
  531         oldname = __DECONST(char *, oidp->oid_name);
  532         oidp->oid_name = newname;
  533         SYSCTL_XUNLOCK();
  534         free(oldname, M_SYSCTLOID);
  535 }
  536 
  537 /*
  538  * Reparent an existing oid.
  539  */
  540 int
  541 sysctl_move_oid(struct sysctl_oid *oid, struct sysctl_oid_list *parent)
  542 {
  543         struct sysctl_oid *oidp;
  544 
  545         SYSCTL_XLOCK();
  546         if (oid->oid_parent == parent) {
  547                 SYSCTL_XUNLOCK();
  548                 return (0);
  549         }
  550         oidp = sysctl_find_oidname(oid->oid_name, parent);
  551         if (oidp != NULL) {
  552                 SYSCTL_XUNLOCK();
  553                 return (EEXIST);
  554         }
  555         sysctl_unregister_oid(oid);
  556         oid->oid_parent = parent;
  557         oid->oid_number = OID_AUTO;
  558         sysctl_register_oid(oid);
  559         SYSCTL_XUNLOCK();
  560         return (0);
  561 }
  562 
  563 /*
  564  * Register the kernel's oids on startup.
  565  */
  566 SET_DECLARE(sysctl_set, struct sysctl_oid);
  567 
  568 static void
  569 sysctl_register_all(void *arg)
  570 {
  571         struct sysctl_oid **oidp;
  572 
  573         sx_init(&sysctlmemlock, "sysctl mem");
  574         SYSCTL_INIT();
  575         SYSCTL_XLOCK();
  576         SET_FOREACH(oidp, sysctl_set)
  577                 sysctl_register_oid(*oidp);
  578         SYSCTL_XUNLOCK();
  579 }
  580 SYSINIT(sysctl, SI_SUB_KMEM, SI_ORDER_ANY, sysctl_register_all, 0);
  581 
  582 /*
  583  * "Staff-functions"
  584  *
  585  * These functions implement a presently undocumented interface 
  586  * used by the sysctl program to walk the tree, and get the type
  587  * so it can print the value.
  588  * This interface is under work and consideration, and should probably
  589  * be killed with a big axe by the first person who can find the time.
  590  * (be aware though, that the proper interface isn't as obvious as it
  591  * may seem, there are various conflicting requirements.
  592  *
  593  * {0,0}        printf the entire MIB-tree.
  594  * {0,1,...}    return the name of the "..." OID.
  595  * {0,2,...}    return the next OID.
  596  * {0,3}        return the OID of the name in "new"
  597  * {0,4,...}    return the kind & format info for the "..." OID.
  598  * {0,5,...}    return the description the "..." OID.
  599  */
  600 
  601 #ifdef SYSCTL_DEBUG
  602 static void
  603 sysctl_sysctl_debug_dump_node(struct sysctl_oid_list *l, int i)
  604 {
  605         int k;
  606         struct sysctl_oid *oidp;
  607 
  608         SYSCTL_ASSERT_XLOCKED();
  609         SLIST_FOREACH(oidp, l, oid_link) {
  610 
  611                 for (k=0; k<i; k++)
  612                         printf(" ");
  613 
  614                 printf("%d %s ", oidp->oid_number, oidp->oid_name);
  615 
  616                 printf("%c%c",
  617                         oidp->oid_kind & CTLFLAG_RD ? 'R':' ',
  618                         oidp->oid_kind & CTLFLAG_WR ? 'W':' ');
  619 
  620                 if (oidp->oid_handler)
  621                         printf(" *Handler");
  622 
  623                 switch (oidp->oid_kind & CTLTYPE) {
  624                         case CTLTYPE_NODE:
  625                                 printf(" Node\n");
  626                                 if (!oidp->oid_handler) {
  627                                         sysctl_sysctl_debug_dump_node(
  628                                                 oidp->oid_arg1, i+2);
  629                                 }
  630                                 break;
  631                         case CTLTYPE_INT:    printf(" Int\n"); break;
  632                         case CTLTYPE_UINT:   printf(" u_int\n"); break;
  633                         case CTLTYPE_LONG:   printf(" Long\n"); break;
  634                         case CTLTYPE_ULONG:  printf(" u_long\n"); break;
  635                         case CTLTYPE_STRING: printf(" String\n"); break;
  636                         case CTLTYPE_U64:    printf(" uint64_t\n"); break;
  637                         case CTLTYPE_S64:    printf(" int64_t\n"); break;
  638                         case CTLTYPE_OPAQUE: printf(" Opaque/struct\n"); break;
  639                         default:             printf("\n");
  640                 }
  641 
  642         }
  643 }
  644 
  645 static int
  646 sysctl_sysctl_debug(SYSCTL_HANDLER_ARGS)
  647 {
  648         int error;
  649 
  650         error = priv_check(req->td, PRIV_SYSCTL_DEBUG);
  651         if (error)
  652                 return (error);
  653         SYSCTL_XLOCK();
  654         sysctl_sysctl_debug_dump_node(&sysctl__children, 0);
  655         SYSCTL_XUNLOCK();
  656         return (ENOENT);
  657 }
  658 
  659 SYSCTL_PROC(_sysctl, 0, debug, CTLTYPE_STRING|CTLFLAG_RD,
  660         0, 0, sysctl_sysctl_debug, "-", "");
  661 #endif
  662 
  663 static int
  664 sysctl_sysctl_name(SYSCTL_HANDLER_ARGS)
  665 {
  666         int *name = (int *) arg1;
  667         u_int namelen = arg2;
  668         int error = 0;
  669         struct sysctl_oid *oid;
  670         struct sysctl_oid_list *lsp = &sysctl__children, *lsp2;
  671         char buf[10];
  672 
  673         SYSCTL_XLOCK();
  674         while (namelen) {
  675                 if (!lsp) {
  676                         snprintf(buf,sizeof(buf),"%d",*name);
  677                         if (req->oldidx)
  678                                 error = SYSCTL_OUT(req, ".", 1);
  679                         if (!error)
  680                                 error = SYSCTL_OUT(req, buf, strlen(buf));
  681                         if (error)
  682                                 goto out;
  683                         namelen--;
  684                         name++;
  685                         continue;
  686                 }
  687                 lsp2 = 0;
  688                 SLIST_FOREACH(oid, lsp, oid_link) {
  689                         if (oid->oid_number != *name)
  690                                 continue;
  691 
  692                         if (req->oldidx)
  693                                 error = SYSCTL_OUT(req, ".", 1);
  694                         if (!error)
  695                                 error = SYSCTL_OUT(req, oid->oid_name,
  696                                         strlen(oid->oid_name));
  697                         if (error)
  698                                 goto out;
  699 
  700                         namelen--;
  701                         name++;
  702 
  703                         if ((oid->oid_kind & CTLTYPE) != CTLTYPE_NODE) 
  704                                 break;
  705 
  706                         if (oid->oid_handler)
  707                                 break;
  708 
  709                         lsp2 = SYSCTL_CHILDREN(oid);
  710                         break;
  711                 }
  712                 lsp = lsp2;
  713         }
  714         error = SYSCTL_OUT(req, "", 1);
  715  out:
  716         SYSCTL_XUNLOCK();
  717         return (error);
  718 }
  719 
  720 /*
  721  * XXXRW/JA: Shouldn't return name data for nodes that we don't permit in
  722  * capability mode.
  723  */
  724 static SYSCTL_NODE(_sysctl, 1, name, CTLFLAG_RD | CTLFLAG_CAPRD,
  725     sysctl_sysctl_name, "");
  726 
  727 static int
  728 sysctl_sysctl_next_ls(struct sysctl_oid_list *lsp, int *name, u_int namelen, 
  729         int *next, int *len, int level, struct sysctl_oid **oidpp)
  730 {
  731         struct sysctl_oid *oidp;
  732 
  733         SYSCTL_ASSERT_XLOCKED();
  734         *len = level;
  735         SLIST_FOREACH(oidp, lsp, oid_link) {
  736                 *next = oidp->oid_number;
  737                 *oidpp = oidp;
  738 
  739                 if (oidp->oid_kind & CTLFLAG_SKIP)
  740                         continue;
  741 
  742                 if (!namelen) {
  743                         if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE) 
  744                                 return (0);
  745                         if (oidp->oid_handler) 
  746                                 /* We really should call the handler here...*/
  747                                 return (0);
  748                         lsp = SYSCTL_CHILDREN(oidp);
  749                         if (!sysctl_sysctl_next_ls(lsp, 0, 0, next+1, 
  750                                 len, level+1, oidpp))
  751                                 return (0);
  752                         goto emptynode;
  753                 }
  754 
  755                 if (oidp->oid_number < *name)
  756                         continue;
  757 
  758                 if (oidp->oid_number > *name) {
  759                         if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
  760                                 return (0);
  761                         if (oidp->oid_handler)
  762                                 return (0);
  763                         lsp = SYSCTL_CHILDREN(oidp);
  764                         if (!sysctl_sysctl_next_ls(lsp, name+1, namelen-1, 
  765                                 next+1, len, level+1, oidpp))
  766                                 return (0);
  767                         goto next;
  768                 }
  769                 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
  770                         continue;
  771 
  772                 if (oidp->oid_handler)
  773                         continue;
  774 
  775                 lsp = SYSCTL_CHILDREN(oidp);
  776                 if (!sysctl_sysctl_next_ls(lsp, name+1, namelen-1, next+1, 
  777                         len, level+1, oidpp))
  778                         return (0);
  779         next:
  780                 namelen = 1;
  781         emptynode:
  782                 *len = level;
  783         }
  784         return (1);
  785 }
  786 
  787 static int
  788 sysctl_sysctl_next(SYSCTL_HANDLER_ARGS)
  789 {
  790         int *name = (int *) arg1;
  791         u_int namelen = arg2;
  792         int i, j, error;
  793         struct sysctl_oid *oid;
  794         struct sysctl_oid_list *lsp = &sysctl__children;
  795         int newoid[CTL_MAXNAME];
  796 
  797         SYSCTL_XLOCK();
  798         i = sysctl_sysctl_next_ls(lsp, name, namelen, newoid, &j, 1, &oid);
  799         SYSCTL_XUNLOCK();
  800         if (i)
  801                 return (ENOENT);
  802         error = SYSCTL_OUT(req, newoid, j * sizeof (int));
  803         return (error);
  804 }
  805 
  806 /*
  807  * XXXRW/JA: Shouldn't return next data for nodes that we don't permit in
  808  * capability mode.
  809  */
  810 static SYSCTL_NODE(_sysctl, 2, next, CTLFLAG_RD | CTLFLAG_CAPRD,
  811     sysctl_sysctl_next, "");
  812 
  813 static int
  814 name2oid(char *name, int *oid, int *len, struct sysctl_oid **oidpp)
  815 {
  816         struct sysctl_oid *oidp;
  817         struct sysctl_oid_list *lsp = &sysctl__children;
  818         char *p;
  819 
  820         SYSCTL_ASSERT_XLOCKED();
  821 
  822         for (*len = 0; *len < CTL_MAXNAME;) {
  823                 p = strsep(&name, ".");
  824 
  825                 oidp = SLIST_FIRST(lsp);
  826                 for (;; oidp = SLIST_NEXT(oidp, oid_link)) {
  827                         if (oidp == NULL)
  828                                 return (ENOENT);
  829                         if (strcmp(p, oidp->oid_name) == 0)
  830                                 break;
  831                 }
  832                 *oid++ = oidp->oid_number;
  833                 (*len)++;
  834 
  835                 if (name == NULL || *name == '\0') {
  836                         if (oidpp)
  837                                 *oidpp = oidp;
  838                         return (0);
  839                 }
  840 
  841                 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
  842                         break;
  843 
  844                 if (oidp->oid_handler)
  845                         break;
  846 
  847                 lsp = SYSCTL_CHILDREN(oidp);
  848         }
  849         return (ENOENT);
  850 }
  851 
  852 static int
  853 sysctl_sysctl_name2oid(SYSCTL_HANDLER_ARGS)
  854 {
  855         char *p;
  856         int error, oid[CTL_MAXNAME], len = 0;
  857         struct sysctl_oid *op = 0;
  858 
  859         if (!req->newlen) 
  860                 return (ENOENT);
  861         if (req->newlen >= MAXPATHLEN)  /* XXX arbitrary, undocumented */
  862                 return (ENAMETOOLONG);
  863 
  864         p = malloc(req->newlen+1, M_SYSCTL, M_WAITOK);
  865 
  866         error = SYSCTL_IN(req, p, req->newlen);
  867         if (error) {
  868                 free(p, M_SYSCTL);
  869                 return (error);
  870         }
  871 
  872         p [req->newlen] = '\0';
  873 
  874         SYSCTL_XLOCK();
  875         error = name2oid(p, oid, &len, &op);
  876         SYSCTL_XUNLOCK();
  877 
  878         free(p, M_SYSCTL);
  879 
  880         if (error)
  881                 return (error);
  882 
  883         error = SYSCTL_OUT(req, oid, len * sizeof *oid);
  884         return (error);
  885 }
  886 
  887 /*
  888  * XXXRW/JA: Shouldn't return name2oid data for nodes that we don't permit in
  889  * capability mode.
  890  */
  891 SYSCTL_PROC(_sysctl, 3, name2oid,
  892     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE
  893     | CTLFLAG_CAPRW, 0, 0, sysctl_sysctl_name2oid, "I", "");
  894 
  895 static int
  896 sysctl_sysctl_oidfmt(SYSCTL_HANDLER_ARGS)
  897 {
  898         struct sysctl_oid *oid;
  899         int error;
  900 
  901         SYSCTL_XLOCK();
  902         error = sysctl_find_oid(arg1, arg2, &oid, NULL, req);
  903         if (error)
  904                 goto out;
  905 
  906         if (oid->oid_fmt == NULL) {
  907                 error = ENOENT;
  908                 goto out;
  909         }
  910         error = SYSCTL_OUT(req, &oid->oid_kind, sizeof(oid->oid_kind));
  911         if (error)
  912                 goto out;
  913         error = SYSCTL_OUT(req, oid->oid_fmt, strlen(oid->oid_fmt) + 1);
  914  out:
  915         SYSCTL_XUNLOCK();
  916         return (error);
  917 }
  918 
  919 
  920 static SYSCTL_NODE(_sysctl, 4, oidfmt, CTLFLAG_RD|CTLFLAG_MPSAFE|CTLFLAG_CAPRD,
  921     sysctl_sysctl_oidfmt, "");
  922 
  923 static int
  924 sysctl_sysctl_oiddescr(SYSCTL_HANDLER_ARGS)
  925 {
  926         struct sysctl_oid *oid;
  927         int error;
  928 
  929         SYSCTL_XLOCK();
  930         error = sysctl_find_oid(arg1, arg2, &oid, NULL, req);
  931         if (error)
  932                 goto out;
  933 
  934         if (oid->oid_descr == NULL) {
  935                 error = ENOENT;
  936                 goto out;
  937         }
  938         error = SYSCTL_OUT(req, oid->oid_descr, strlen(oid->oid_descr) + 1);
  939  out:
  940         SYSCTL_XUNLOCK();
  941         return (error);
  942 }
  943 
  944 static SYSCTL_NODE(_sysctl, 5, oiddescr, CTLFLAG_RD|CTLFLAG_CAPRD,
  945     sysctl_sysctl_oiddescr, "");
  946 
  947 /*
  948  * Default "handler" functions.
  949  */
  950 
  951 /*
  952  * Handle an int, signed or unsigned.
  953  * Two cases:
  954  *     a variable:  point arg1 at it.
  955  *     a constant:  pass it in arg2.
  956  */
  957 
  958 int
  959 sysctl_handle_int(SYSCTL_HANDLER_ARGS)
  960 {
  961         int tmpout, error = 0;
  962 
  963         /*
  964          * Attempt to get a coherent snapshot by making a copy of the data.
  965          */
  966         if (arg1)
  967                 tmpout = *(int *)arg1;
  968         else
  969                 tmpout = arg2;
  970         error = SYSCTL_OUT(req, &tmpout, sizeof(int));
  971 
  972         if (error || !req->newptr)
  973                 return (error);
  974 
  975         if (!arg1)
  976                 error = EPERM;
  977         else
  978                 error = SYSCTL_IN(req, arg1, sizeof(int));
  979         return (error);
  980 }
  981 
  982 /*
  983  * Based on on sysctl_handle_int() convert milliseconds into ticks.
  984  * Note: this is used by TCP.
  985  */
  986 
  987 int
  988 sysctl_msec_to_ticks(SYSCTL_HANDLER_ARGS)
  989 {
  990         int error, s, tt;
  991 
  992         tt = *(int *)arg1;
  993         s = (int)((int64_t)tt * 1000 / hz);
  994 
  995         error = sysctl_handle_int(oidp, &s, 0, req);
  996         if (error || !req->newptr)
  997                 return (error);
  998 
  999         tt = (int)((int64_t)s * hz / 1000);
 1000         if (tt < 1)
 1001                 return (EINVAL);
 1002 
 1003         *(int *)arg1 = tt;
 1004         return (0);
 1005 }
 1006 
 1007 
 1008 /*
 1009  * Handle a long, signed or unsigned.
 1010  * Two cases:
 1011  *     a variable:  point arg1 at it.
 1012  *     a constant:  pass it in arg2.
 1013  */
 1014 
 1015 int
 1016 sysctl_handle_long(SYSCTL_HANDLER_ARGS)
 1017 {
 1018         int error = 0;
 1019         long tmplong;
 1020 #ifdef SCTL_MASK32
 1021         int tmpint;
 1022 #endif
 1023 
 1024         /*
 1025          * Attempt to get a coherent snapshot by making a copy of the data.
 1026          */
 1027         if (arg1)
 1028                 tmplong = *(long *)arg1;
 1029         else
 1030                 tmplong = arg2;
 1031 #ifdef SCTL_MASK32
 1032         if (req->flags & SCTL_MASK32) {
 1033                 tmpint = tmplong;
 1034                 error = SYSCTL_OUT(req, &tmpint, sizeof(int));
 1035         } else
 1036 #endif
 1037                 error = SYSCTL_OUT(req, &tmplong, sizeof(long));
 1038 
 1039         if (error || !req->newptr)
 1040                 return (error);
 1041 
 1042         if (!arg1)
 1043                 error = EPERM;
 1044 #ifdef SCTL_MASK32
 1045         else if (req->flags & SCTL_MASK32) {
 1046                 error = SYSCTL_IN(req, &tmpint, sizeof(int));
 1047                 *(long *)arg1 = (long)tmpint;
 1048         }
 1049 #endif
 1050         else
 1051                 error = SYSCTL_IN(req, arg1, sizeof(long));
 1052         return (error);
 1053 }
 1054 
 1055 /*
 1056  * Handle a 64 bit int, signed or unsigned.
 1057  * Two cases:
 1058  *     a variable:  point arg1 at it.
 1059  *     a constant:  pass it in arg2.
 1060  */
 1061 int
 1062 sysctl_handle_64(SYSCTL_HANDLER_ARGS)
 1063 {
 1064         int error = 0;
 1065         uint64_t tmpout;
 1066 
 1067         /*
 1068          * Attempt to get a coherent snapshot by making a copy of the data.
 1069          */
 1070         if (arg1)
 1071                 tmpout = *(uint64_t *)arg1;
 1072         else
 1073                 tmpout = arg2;
 1074         error = SYSCTL_OUT(req, &tmpout, sizeof(uint64_t));
 1075 
 1076         if (error || !req->newptr)
 1077                 return (error);
 1078 
 1079         if (!arg1)
 1080                 error = EPERM;
 1081         else
 1082                 error = SYSCTL_IN(req, arg1, sizeof(uint64_t));
 1083         return (error);
 1084 }
 1085 
 1086 /*
 1087  * Handle our generic '\0' terminated 'C' string.
 1088  * Two cases:
 1089  *      a variable string:  point arg1 at it, arg2 is max length.
 1090  *      a constant string:  point arg1 at it, arg2 is zero.
 1091  */
 1092 
 1093 int
 1094 sysctl_handle_string(SYSCTL_HANDLER_ARGS)
 1095 {
 1096         int error=0;
 1097         char *tmparg;
 1098         size_t outlen;
 1099 
 1100         /*
 1101          * Attempt to get a coherent snapshot by copying to a
 1102          * temporary kernel buffer.
 1103          */
 1104 retry:
 1105         outlen = strlen((char *)arg1)+1;
 1106         tmparg = malloc(outlen, M_SYSCTLTMP, M_WAITOK);
 1107 
 1108         if (strlcpy(tmparg, (char *)arg1, outlen) >= outlen) {
 1109                 free(tmparg, M_SYSCTLTMP);
 1110                 goto retry;
 1111         }
 1112 
 1113         error = SYSCTL_OUT(req, tmparg, outlen);
 1114         free(tmparg, M_SYSCTLTMP);
 1115 
 1116         if (error || !req->newptr)
 1117                 return (error);
 1118 
 1119         if ((req->newlen - req->newidx) >= arg2) {
 1120                 error = EINVAL;
 1121         } else {
 1122                 arg2 = (req->newlen - req->newidx);
 1123                 error = SYSCTL_IN(req, arg1, arg2);
 1124                 ((char *)arg1)[arg2] = '\0';
 1125         }
 1126 
 1127         return (error);
 1128 }
 1129 
 1130 /*
 1131  * Handle any kind of opaque data.
 1132  * arg1 points to it, arg2 is the size.
 1133  */
 1134 
 1135 int
 1136 sysctl_handle_opaque(SYSCTL_HANDLER_ARGS)
 1137 {
 1138         int error, tries;
 1139         u_int generation;
 1140         struct sysctl_req req2;
 1141 
 1142         /*
 1143          * Attempt to get a coherent snapshot, by using the thread
 1144          * pre-emption counter updated from within mi_switch() to
 1145          * determine if we were pre-empted during a bcopy() or
 1146          * copyout(). Make 3 attempts at doing this before giving up.
 1147          * If we encounter an error, stop immediately.
 1148          */
 1149         tries = 0;
 1150         req2 = *req;
 1151 retry:
 1152         generation = curthread->td_generation;
 1153         error = SYSCTL_OUT(req, arg1, arg2);
 1154         if (error)
 1155                 return (error);
 1156         tries++;
 1157         if (generation != curthread->td_generation && tries < 3) {
 1158                 *req = req2;
 1159                 goto retry;
 1160         }
 1161 
 1162         error = SYSCTL_IN(req, arg1, arg2);
 1163 
 1164         return (error);
 1165 }
 1166 
 1167 /*
 1168  * Transfer functions to/from kernel space.
 1169  * XXX: rather untested at this point
 1170  */
 1171 static int
 1172 sysctl_old_kernel(struct sysctl_req *req, const void *p, size_t l)
 1173 {
 1174         size_t i = 0;
 1175 
 1176         if (req->oldptr) {
 1177                 i = l;
 1178                 if (req->oldlen <= req->oldidx)
 1179                         i = 0;
 1180                 else
 1181                         if (i > req->oldlen - req->oldidx)
 1182                                 i = req->oldlen - req->oldidx;
 1183                 if (i > 0)
 1184                         bcopy(p, (char *)req->oldptr + req->oldidx, i);
 1185         }
 1186         req->oldidx += l;
 1187         if (req->oldptr && i != l)
 1188                 return (ENOMEM);
 1189         return (0);
 1190 }
 1191 
 1192 static int
 1193 sysctl_new_kernel(struct sysctl_req *req, void *p, size_t l)
 1194 {
 1195         if (!req->newptr)
 1196                 return (0);
 1197         if (req->newlen - req->newidx < l)
 1198                 return (EINVAL);
 1199         bcopy((char *)req->newptr + req->newidx, p, l);
 1200         req->newidx += l;
 1201         return (0);
 1202 }
 1203 
 1204 int
 1205 kernel_sysctl(struct thread *td, int *name, u_int namelen, void *old,
 1206     size_t *oldlenp, void *new, size_t newlen, size_t *retval, int flags)
 1207 {
 1208         int error = 0;
 1209         struct sysctl_req req;
 1210 
 1211         bzero(&req, sizeof req);
 1212 
 1213         req.td = td;
 1214         req.flags = flags;
 1215 
 1216         if (oldlenp) {
 1217                 req.oldlen = *oldlenp;
 1218         }
 1219         req.validlen = req.oldlen;
 1220 
 1221         if (old) {
 1222                 req.oldptr= old;
 1223         }
 1224 
 1225         if (new != NULL) {
 1226                 req.newlen = newlen;
 1227                 req.newptr = new;
 1228         }
 1229 
 1230         req.oldfunc = sysctl_old_kernel;
 1231         req.newfunc = sysctl_new_kernel;
 1232         req.lock = REQ_UNWIRED;
 1233 
 1234         SYSCTL_XLOCK();
 1235         error = sysctl_root(0, name, namelen, &req);
 1236         SYSCTL_XUNLOCK();
 1237 
 1238         if (req.lock == REQ_WIRED && req.validlen > 0)
 1239                 vsunlock(req.oldptr, req.validlen);
 1240 
 1241         if (error && error != ENOMEM)
 1242                 return (error);
 1243 
 1244         if (retval) {
 1245                 if (req.oldptr && req.oldidx > req.validlen)
 1246                         *retval = req.validlen;
 1247                 else
 1248                         *retval = req.oldidx;
 1249         }
 1250         return (error);
 1251 }
 1252 
 1253 int
 1254 kernel_sysctlbyname(struct thread *td, char *name, void *old, size_t *oldlenp,
 1255     void *new, size_t newlen, size_t *retval, int flags)
 1256 {
 1257         int oid[CTL_MAXNAME];
 1258         size_t oidlen, plen;
 1259         int error;
 1260 
 1261         oid[0] = 0;             /* sysctl internal magic */
 1262         oid[1] = 3;             /* name2oid */
 1263         oidlen = sizeof(oid);
 1264 
 1265         error = kernel_sysctl(td, oid, 2, oid, &oidlen,
 1266             (void *)name, strlen(name), &plen, flags);
 1267         if (error)
 1268                 return (error);
 1269 
 1270         error = kernel_sysctl(td, oid, plen / sizeof(int), old, oldlenp,
 1271             new, newlen, retval, flags);
 1272         return (error);
 1273 }
 1274 
 1275 /*
 1276  * Transfer function to/from user space.
 1277  */
 1278 static int
 1279 sysctl_old_user(struct sysctl_req *req, const void *p, size_t l)
 1280 {
 1281         size_t i, len, origidx;
 1282         int error;
 1283 
 1284         origidx = req->oldidx;
 1285         req->oldidx += l;
 1286         if (req->oldptr == NULL)
 1287                 return (0);
 1288         /*
 1289          * If we have not wired the user supplied buffer and we are currently
 1290          * holding locks, drop a witness warning, as it's possible that
 1291          * write operations to the user page can sleep.
 1292          */
 1293         if (req->lock != REQ_WIRED)
 1294                 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,
 1295                     "sysctl_old_user()");
 1296         i = l;
 1297         len = req->validlen;
 1298         if (len <= origidx)
 1299                 i = 0;
 1300         else {
 1301                 if (i > len - origidx)
 1302                         i = len - origidx;
 1303                 if (req->lock == REQ_WIRED) {
 1304                         error = copyout_nofault(p, (char *)req->oldptr +
 1305                             origidx, i);
 1306                 } else
 1307                         error = copyout(p, (char *)req->oldptr + origidx, i);
 1308                 if (error != 0)
 1309                         return (error);
 1310         }
 1311         if (i < l)
 1312                 return (ENOMEM);
 1313         return (0);
 1314 }
 1315 
 1316 static int
 1317 sysctl_new_user(struct sysctl_req *req, void *p, size_t l)
 1318 {
 1319         int error;
 1320 
 1321         if (!req->newptr)
 1322                 return (0);
 1323         if (req->newlen - req->newidx < l)
 1324                 return (EINVAL);
 1325         WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,
 1326             "sysctl_new_user()");
 1327         error = copyin((char *)req->newptr + req->newidx, p, l);
 1328         req->newidx += l;
 1329         return (error);
 1330 }
 1331 
 1332 /*
 1333  * Wire the user space destination buffer.  If set to a value greater than
 1334  * zero, the len parameter limits the maximum amount of wired memory.
 1335  */
 1336 int
 1337 sysctl_wire_old_buffer(struct sysctl_req *req, size_t len)
 1338 {
 1339         int ret;
 1340         size_t wiredlen;
 1341 
 1342         wiredlen = (len > 0 && len < req->oldlen) ? len : req->oldlen;
 1343         ret = 0;
 1344         if (req->lock != REQ_WIRED && req->oldptr &&
 1345             req->oldfunc == sysctl_old_user) {
 1346                 if (wiredlen != 0) {
 1347                         ret = vslock(req->oldptr, wiredlen);
 1348                         if (ret != 0) {
 1349                                 if (ret != ENOMEM)
 1350                                         return (ret);
 1351                                 wiredlen = 0;
 1352                         }
 1353                 }
 1354                 req->lock = REQ_WIRED;
 1355                 req->validlen = wiredlen;
 1356         }
 1357         return (0);
 1358 }
 1359 
 1360 int
 1361 sysctl_find_oid(int *name, u_int namelen, struct sysctl_oid **noid,
 1362     int *nindx, struct sysctl_req *req)
 1363 {
 1364         struct sysctl_oid_list *lsp;
 1365         struct sysctl_oid *oid;
 1366         int indx;
 1367 
 1368         SYSCTL_ASSERT_XLOCKED();
 1369         lsp = &sysctl__children;
 1370         indx = 0;
 1371         while (indx < CTL_MAXNAME) {
 1372                 SLIST_FOREACH(oid, lsp, oid_link) {
 1373                         if (oid->oid_number == name[indx])
 1374                                 break;
 1375                 }
 1376                 if (oid == NULL)
 1377                         return (ENOENT);
 1378 
 1379                 indx++;
 1380                 if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
 1381                         if (oid->oid_handler != NULL || indx == namelen) {
 1382                                 *noid = oid;
 1383                                 if (nindx != NULL)
 1384                                         *nindx = indx;
 1385                                 KASSERT((oid->oid_kind & CTLFLAG_DYING) == 0,
 1386                                     ("%s found DYING node %p", __func__, oid));
 1387                                 return (0);
 1388                         }
 1389                         lsp = SYSCTL_CHILDREN(oid);
 1390                 } else if (indx == namelen) {
 1391                         *noid = oid;
 1392                         if (nindx != NULL)
 1393                                 *nindx = indx;
 1394                         KASSERT((oid->oid_kind & CTLFLAG_DYING) == 0,
 1395                             ("%s found DYING node %p", __func__, oid));
 1396                         return (0);
 1397                 } else {
 1398                         return (ENOTDIR);
 1399                 }
 1400         }
 1401         return (ENOENT);
 1402 }
 1403 
 1404 /*
 1405  * Traverse our tree, and find the right node, execute whatever it points
 1406  * to, and return the resulting error code.
 1407  */
 1408 
 1409 static int
 1410 sysctl_root(SYSCTL_HANDLER_ARGS)
 1411 {
 1412         struct sysctl_oid *oid;
 1413         int error, indx, lvl;
 1414 
 1415         SYSCTL_ASSERT_XLOCKED();
 1416 
 1417         error = sysctl_find_oid(arg1, arg2, &oid, &indx, req);
 1418         if (error)
 1419                 return (error);
 1420 
 1421         if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
 1422                 /*
 1423                  * You can't call a sysctl when it's a node, but has
 1424                  * no handler.  Inform the user that it's a node.
 1425                  * The indx may or may not be the same as namelen.
 1426                  */
 1427                 if (oid->oid_handler == NULL)
 1428                         return (EISDIR);
 1429         }
 1430 
 1431         /* Is this sysctl writable? */
 1432         if (req->newptr && !(oid->oid_kind & CTLFLAG_WR))
 1433                 return (EPERM);
 1434 
 1435         KASSERT(req->td != NULL, ("sysctl_root(): req->td == NULL"));
 1436 
 1437 #ifdef CAPABILITY_MODE
 1438         /*
 1439          * If the process is in capability mode, then don't permit reading or
 1440          * writing unless specifically granted for the node.
 1441          */
 1442         if (IN_CAPABILITY_MODE(req->td)) {
 1443                 if (req->oldptr && !(oid->oid_kind & CTLFLAG_CAPRD))
 1444                         return (EPERM);
 1445                 if (req->newptr && !(oid->oid_kind & CTLFLAG_CAPWR))
 1446                         return (EPERM);
 1447         }
 1448 #endif
 1449 
 1450         /* Is this sysctl sensitive to securelevels? */
 1451         if (req->newptr && (oid->oid_kind & CTLFLAG_SECURE)) {
 1452                 lvl = (oid->oid_kind & CTLMASK_SECURE) >> CTLSHIFT_SECURE;
 1453                 error = securelevel_gt(req->td->td_ucred, lvl);
 1454                 if (error)
 1455                         return (error);
 1456         }
 1457 
 1458         /* Is this sysctl writable by only privileged users? */
 1459         if (req->newptr && !(oid->oid_kind & CTLFLAG_ANYBODY)) {
 1460                 int priv;
 1461 
 1462                 if (oid->oid_kind & CTLFLAG_PRISON)
 1463                         priv = PRIV_SYSCTL_WRITEJAIL;
 1464 #ifdef VIMAGE
 1465                 else if ((oid->oid_kind & CTLFLAG_VNET) &&
 1466                      prison_owns_vnet(req->td->td_ucred))
 1467                         priv = PRIV_SYSCTL_WRITEJAIL;
 1468 #endif
 1469                 else
 1470                         priv = PRIV_SYSCTL_WRITE;
 1471                 error = priv_check(req->td, priv);
 1472                 if (error)
 1473                         return (error);
 1474         }
 1475 
 1476         if (!oid->oid_handler)
 1477                 return (EINVAL);
 1478 
 1479         if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
 1480                 arg1 = (int *)arg1 + indx;
 1481                 arg2 -= indx;
 1482         } else {
 1483                 arg1 = oid->oid_arg1;
 1484                 arg2 = oid->oid_arg2;
 1485         }
 1486 #ifdef MAC
 1487         error = mac_system_check_sysctl(req->td->td_ucred, oid, arg1, arg2,
 1488             req);
 1489         if (error != 0)
 1490                 return (error);
 1491 #endif
 1492         oid->oid_running++;
 1493         SYSCTL_XUNLOCK();
 1494 
 1495         if (!(oid->oid_kind & CTLFLAG_MPSAFE))
 1496                 mtx_lock(&Giant);
 1497         error = oid->oid_handler(oid, arg1, arg2, req);
 1498         if (!(oid->oid_kind & CTLFLAG_MPSAFE))
 1499                 mtx_unlock(&Giant);
 1500 
 1501         KFAIL_POINT_ERROR(_debug_fail_point, sysctl_running, error);
 1502 
 1503         SYSCTL_XLOCK();
 1504         oid->oid_running--;
 1505         if (oid->oid_running == 0 && (oid->oid_kind & CTLFLAG_DYING) != 0)
 1506                 wakeup(&oid->oid_running);
 1507         return (error);
 1508 }
 1509 
 1510 #ifndef _SYS_SYSPROTO_H_
 1511 struct sysctl_args {
 1512         int     *name;
 1513         u_int   namelen;
 1514         void    *old;
 1515         size_t  *oldlenp;
 1516         void    *new;
 1517         size_t  newlen;
 1518 };
 1519 #endif
 1520 int
 1521 sys___sysctl(struct thread *td, struct sysctl_args *uap)
 1522 {
 1523         int error, i, name[CTL_MAXNAME];
 1524         size_t j;
 1525 
 1526         if (uap->namelen > CTL_MAXNAME || uap->namelen < 2)
 1527                 return (EINVAL);
 1528 
 1529         error = copyin(uap->name, &name, uap->namelen * sizeof(int));
 1530         if (error)
 1531                 return (error);
 1532 
 1533         error = userland_sysctl(td, name, uap->namelen,
 1534                 uap->old, uap->oldlenp, 0,
 1535                 uap->new, uap->newlen, &j, 0);
 1536         if (error && error != ENOMEM)
 1537                 return (error);
 1538         if (uap->oldlenp) {
 1539                 i = copyout(&j, uap->oldlenp, sizeof(j));
 1540                 if (i)
 1541                         return (i);
 1542         }
 1543         return (error);
 1544 }
 1545 
 1546 /*
 1547  * This is used from various compatibility syscalls too.  That's why name
 1548  * must be in kernel space.
 1549  */
 1550 int
 1551 userland_sysctl(struct thread *td, int *name, u_int namelen, void *old,
 1552     size_t *oldlenp, int inkernel, void *new, size_t newlen, size_t *retval,
 1553     int flags)
 1554 {
 1555         int error = 0, memlocked;
 1556         struct sysctl_req req;
 1557 
 1558         bzero(&req, sizeof req);
 1559 
 1560         req.td = td;
 1561         req.flags = flags;
 1562 
 1563         if (oldlenp) {
 1564                 if (inkernel) {
 1565                         req.oldlen = *oldlenp;
 1566                 } else {
 1567                         error = copyin(oldlenp, &req.oldlen, sizeof(*oldlenp));
 1568                         if (error)
 1569                                 return (error);
 1570                 }
 1571         }
 1572         req.validlen = req.oldlen;
 1573 
 1574         if (old) {
 1575                 if (!useracc(old, req.oldlen, VM_PROT_WRITE))
 1576                         return (EFAULT);
 1577                 req.oldptr= old;
 1578         }
 1579 
 1580         if (new != NULL) {
 1581                 if (!useracc(new, newlen, VM_PROT_READ))
 1582                         return (EFAULT);
 1583                 req.newlen = newlen;
 1584                 req.newptr = new;
 1585         }
 1586 
 1587         req.oldfunc = sysctl_old_user;
 1588         req.newfunc = sysctl_new_user;
 1589         req.lock = REQ_UNWIRED;
 1590 
 1591 #ifdef KTRACE
 1592         if (KTRPOINT(curthread, KTR_SYSCTL))
 1593                 ktrsysctl(name, namelen);
 1594 #endif
 1595 
 1596         if (req.oldlen > PAGE_SIZE) {
 1597                 memlocked = 1;
 1598                 sx_xlock(&sysctlmemlock);
 1599         } else
 1600                 memlocked = 0;
 1601         CURVNET_SET(TD_TO_VNET(td));
 1602 
 1603         for (;;) {
 1604                 req.oldidx = 0;
 1605                 req.newidx = 0;
 1606                 SYSCTL_XLOCK();
 1607                 error = sysctl_root(0, name, namelen, &req);
 1608                 SYSCTL_XUNLOCK();
 1609                 if (error != EAGAIN)
 1610                         break;
 1611                 kern_yield(PRI_USER);
 1612         }
 1613 
 1614         CURVNET_RESTORE();
 1615 
 1616         if (req.lock == REQ_WIRED && req.validlen > 0)
 1617                 vsunlock(req.oldptr, req.validlen);
 1618         if (memlocked)
 1619                 sx_xunlock(&sysctlmemlock);
 1620 
 1621         if (error && error != ENOMEM)
 1622                 return (error);
 1623 
 1624         if (retval) {
 1625                 if (req.oldptr && req.oldidx > req.validlen)
 1626                         *retval = req.validlen;
 1627                 else
 1628                         *retval = req.oldidx;
 1629         }
 1630         return (error);
 1631 }
 1632 
 1633 /*
 1634  * Drain into a sysctl struct.  The user buffer should be wired if a page
 1635  * fault would cause issue.
 1636  */
 1637 static int
 1638 sbuf_sysctl_drain(void *arg, const char *data, int len)
 1639 {
 1640         struct sysctl_req *req = arg;
 1641         int error;
 1642 
 1643         error = SYSCTL_OUT(req, data, len);
 1644         KASSERT(error >= 0, ("Got unexpected negative value %d", error));
 1645         return (error == 0 ? len : -error);
 1646 }
 1647 
 1648 struct sbuf *
 1649 sbuf_new_for_sysctl(struct sbuf *s, char *buf, int length,
 1650     struct sysctl_req *req)
 1651 {
 1652 
 1653         s = sbuf_new(s, buf, length, SBUF_FIXEDLEN);
 1654         sbuf_set_drain(s, sbuf_sysctl_drain, req);
 1655         return (s);
 1656 }
Cache object: 2563c77a40362e7e5ad68c91b03472aa
[ 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
