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  * SPDX-License-Identifier: BSD-3-Clause
    3  *
    4  * Copyright (c) 1982, 1986, 1989, 1993
    5  *      The Regents of the University of California.  All rights reserved.
    6  *
    7  * This code is derived from software contributed to Berkeley by
    8  * Mike Karels at Berkeley Software Design, Inc.
    9  *
   10  * Quite extensively rewritten by Poul-Henning Kamp of the FreeBSD
   11  * project, to make these variables more userfriendly.
   12  *
   13  * Redistribution and use in source and binary forms, with or without
   14  * modification, are permitted provided that the following conditions
   15  * are met:
   16  * 1. Redistributions of source code must retain the above copyright
   17  *    notice, this list of conditions and the following disclaimer.
   18  * 2. Redistributions in binary form must reproduce the above copyright
   19  *    notice, this list of conditions and the following disclaimer in the
   20  *    documentation and/or other materials provided with the distribution.
   21  * 3. Neither the name of the University nor the names of its contributors
   22  *    may be used to endorse or promote products derived from this software
   23  *    without specific prior written permission.
   24  *
   25  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
   26  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
   27  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
   28  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
   29  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
   30  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
   31  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
   32  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
   33  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
   34  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
   35  * SUCH DAMAGE.
   36  *
   37  *      @(#)kern_sysctl.c       8.4 (Berkeley) 4/14/94
   38  */
   39 
   40 #include <sys/cdefs.h>
   41 __FBSDID("$FreeBSD: releng/12.0/sys/kern/kern_sysctl.c 335461 2018-06-20 20:04:20Z hselasky $");
   42 
   43 #include "opt_capsicum.h"
   44 #include "opt_ktrace.h"
   45 
   46 #include <sys/param.h>
   47 #include <sys/fail.h>
   48 #include <sys/systm.h>
   49 #include <sys/capsicum.h>
   50 #include <sys/kernel.h>
   51 #include <sys/sysctl.h>
   52 #include <sys/malloc.h>
   53 #include <sys/priv.h>
   54 #include <sys/proc.h>
   55 #include <sys/jail.h>
   56 #include <sys/lock.h>
   57 #include <sys/mutex.h>
   58 #include <sys/rmlock.h>
   59 #include <sys/sbuf.h>
   60 #include <sys/sx.h>
   61 #include <sys/sysproto.h>
   62 #include <sys/uio.h>
   63 #ifdef KTRACE
   64 #include <sys/ktrace.h>
   65 #endif
   66 
   67 #include <net/vnet.h>
   68 
   69 #include <security/mac/mac_framework.h>
   70 
   71 #include <vm/vm.h>
   72 #include <vm/vm_extern.h>
   73 
   74 static MALLOC_DEFINE(M_SYSCTL, "sysctl", "sysctl internal magic");
   75 static MALLOC_DEFINE(M_SYSCTLOID, "sysctloid", "sysctl dynamic oids");
   76 static MALLOC_DEFINE(M_SYSCTLTMP, "sysctltmp", "sysctl temp output buffer");
   77 
   78 /*
   79  * The sysctllock protects the MIB tree.  It also protects sysctl
   80  * contexts used with dynamic sysctls.  The sysctl_register_oid() and
   81  * sysctl_unregister_oid() routines require the sysctllock to already
   82  * be held, so the sysctl_wlock() and sysctl_wunlock() routines are
   83  * provided for the few places in the kernel which need to use that
   84  * API rather than using the dynamic API.  Use of the dynamic API is
   85  * strongly encouraged for most code.
   86  *
   87  * The sysctlmemlock is used to limit the amount of user memory wired for
   88  * sysctl requests.  This is implemented by serializing any userland
   89  * sysctl requests larger than a single page via an exclusive lock.
   90  */
   91 static struct rmlock sysctllock;
   92 static struct sx __exclusive_cache_line sysctlmemlock;
   93 
   94 #define SYSCTL_WLOCK()          rm_wlock(&sysctllock)
   95 #define SYSCTL_WUNLOCK()        rm_wunlock(&sysctllock)
   96 #define SYSCTL_RLOCK(tracker)   rm_rlock(&sysctllock, (tracker))
   97 #define SYSCTL_RUNLOCK(tracker) rm_runlock(&sysctllock, (tracker))
   98 #define SYSCTL_WLOCKED()        rm_wowned(&sysctllock)
   99 #define SYSCTL_ASSERT_LOCKED()  rm_assert(&sysctllock, RA_LOCKED)
  100 #define SYSCTL_ASSERT_WLOCKED() rm_assert(&sysctllock, RA_WLOCKED)
  101 #define SYSCTL_ASSERT_RLOCKED() rm_assert(&sysctllock, RA_RLOCKED)
  102 #define SYSCTL_INIT()           rm_init_flags(&sysctllock, "sysctl lock", \
  103                                     RM_SLEEPABLE)
  104 #define SYSCTL_SLEEP(ch, wmesg, timo)                                   \
  105                                 rm_sleep(ch, &sysctllock, 0, wmesg, timo)
  106 
  107 static int sysctl_root(SYSCTL_HANDLER_ARGS);
  108 
  109 /* Root list */
  110 struct sysctl_oid_list sysctl__children = SLIST_HEAD_INITIALIZER(&sysctl__children);
  111 
  112 static int      sysctl_remove_oid_locked(struct sysctl_oid *oidp, int del,
  113                     int recurse);
  114 static int      sysctl_old_kernel(struct sysctl_req *, const void *, size_t);
  115 static int      sysctl_new_kernel(struct sysctl_req *, void *, size_t);
  116 
  117 static struct sysctl_oid *
  118 sysctl_find_oidname(const char *name, struct sysctl_oid_list *list)
  119 {
  120         struct sysctl_oid *oidp;
  121 
  122         SYSCTL_ASSERT_LOCKED();
  123         SLIST_FOREACH(oidp, list, oid_link) {
  124                 if (strcmp(oidp->oid_name, name) == 0) {
  125                         return (oidp);
  126                 }
  127         }
  128         return (NULL);
  129 }
  130 
  131 /*
  132  * Initialization of the MIB tree.
  133  *
  134  * Order by number in each list.
  135  */
  136 void
  137 sysctl_wlock(void)
  138 {
  139 
  140         SYSCTL_WLOCK();
  141 }
  142 
  143 void
  144 sysctl_wunlock(void)
  145 {
  146 
  147         SYSCTL_WUNLOCK();
  148 }
  149 
  150 static int
  151 sysctl_root_handler_locked(struct sysctl_oid *oid, void *arg1, intmax_t arg2,
  152     struct sysctl_req *req, struct rm_priotracker *tracker)
  153 {
  154         int error;
  155 
  156         if (oid->oid_kind & CTLFLAG_DYN)
  157                 atomic_add_int(&oid->oid_running, 1);
  158 
  159         if (tracker != NULL)
  160                 SYSCTL_RUNLOCK(tracker);
  161         else
  162                 SYSCTL_WUNLOCK();
  163 
  164         if (!(oid->oid_kind & CTLFLAG_MPSAFE))
  165                 mtx_lock(&Giant);
  166         error = oid->oid_handler(oid, arg1, arg2, req);
  167         if (!(oid->oid_kind & CTLFLAG_MPSAFE))
  168                 mtx_unlock(&Giant);
  169 
  170         KFAIL_POINT_ERROR(_debug_fail_point, sysctl_running, error);
  171 
  172         if (tracker != NULL)
  173                 SYSCTL_RLOCK(tracker);
  174         else
  175                 SYSCTL_WLOCK();
  176 
  177         if (oid->oid_kind & CTLFLAG_DYN) {
  178                 if (atomic_fetchadd_int(&oid->oid_running, -1) == 1 &&
  179                     (oid->oid_kind & CTLFLAG_DYING) != 0)
  180                         wakeup(&oid->oid_running);
  181         }
  182 
  183         return (error);
  184 }
  185 
  186 static void
  187 sysctl_load_tunable_by_oid_locked(struct sysctl_oid *oidp)
  188 {
  189         struct sysctl_req req;
  190         struct sysctl_oid *curr;
  191         char *penv = NULL;
  192         char path[96];
  193         ssize_t rem = sizeof(path);
  194         ssize_t len;
  195         uint8_t data[512] __aligned(sizeof(uint64_t));
  196         int size;
  197         int error;
  198 
  199         path[--rem] = 0;
  200 
  201         for (curr = oidp; curr != NULL; curr = SYSCTL_PARENT(curr)) {
  202                 len = strlen(curr->oid_name);
  203                 rem -= len;
  204                 if (curr != oidp)
  205                         rem -= 1;
  206                 if (rem < 0) {
  207                         printf("OID path exceeds %d bytes\n", (int)sizeof(path));
  208                         return;
  209                 }
  210                 memcpy(path + rem, curr->oid_name, len);
  211                 if (curr != oidp)
  212                         path[rem + len] = '.';
  213         }
  214 
  215         memset(&req, 0, sizeof(req));
  216 
  217         req.td = curthread;
  218         req.oldfunc = sysctl_old_kernel;
  219         req.newfunc = sysctl_new_kernel;
  220         req.lock = REQ_UNWIRED;
  221 
  222         switch (oidp->oid_kind & CTLTYPE) {
  223         case CTLTYPE_INT:
  224                 if (getenv_array(path + rem, data, sizeof(data), &size,
  225                     sizeof(int), GETENV_SIGNED) == 0)
  226                         return;
  227                 req.newlen = size;
  228                 req.newptr = data;
  229                 break;
  230         case CTLTYPE_UINT:
  231                 if (getenv_array(path + rem, data, sizeof(data), &size,
  232                     sizeof(int), GETENV_UNSIGNED) == 0)
  233                         return;
  234                 req.newlen = size;
  235                 req.newptr = data;
  236                 break;
  237         case CTLTYPE_LONG:
  238                 if (getenv_array(path + rem, data, sizeof(data), &size,
  239                     sizeof(long), GETENV_SIGNED) == 0)
  240                         return;
  241                 req.newlen = size;
  242                 req.newptr = data;
  243                 break;
  244         case CTLTYPE_ULONG:
  245                 if (getenv_array(path + rem, data, sizeof(data), &size,
  246                     sizeof(long), GETENV_UNSIGNED) == 0)
  247                         return;
  248                 req.newlen = size;
  249                 req.newptr = data;
  250                 break;
  251         case CTLTYPE_S8:
  252                 if (getenv_array(path + rem, data, sizeof(data), &size,
  253                     sizeof(int8_t), GETENV_SIGNED) == 0)
  254                         return;
  255                 req.newlen = size;
  256                 req.newptr = data;
  257                 break;
  258         case CTLTYPE_S16:
  259                 if (getenv_array(path + rem, data, sizeof(data), &size,
  260                     sizeof(int16_t), GETENV_SIGNED) == 0)
  261                         return;
  262                 req.newlen = size;
  263                 req.newptr = data;
  264                 break;
  265         case CTLTYPE_S32:
  266                 if (getenv_array(path + rem, data, sizeof(data), &size,
  267                     sizeof(int32_t), GETENV_SIGNED) == 0)
  268                         return;
  269                 req.newlen = size;
  270                 req.newptr = data;
  271                 break;
  272         case CTLTYPE_S64:
  273                 if (getenv_array(path + rem, data, sizeof(data), &size,
  274                     sizeof(int64_t), GETENV_SIGNED) == 0)
  275                         return;
  276                 req.newlen = size;
  277                 req.newptr = data;
  278                 break;
  279         case CTLTYPE_U8:
  280                 if (getenv_array(path + rem, data, sizeof(data), &size,
  281                     sizeof(uint8_t), GETENV_UNSIGNED) == 0)
  282                         return;
  283                 req.newlen = size;
  284                 req.newptr = data;
  285                 break;
  286         case CTLTYPE_U16:
  287                 if (getenv_array(path + rem, data, sizeof(data), &size,
  288                     sizeof(uint16_t), GETENV_UNSIGNED) == 0)
  289                         return;
  290                 req.newlen = size;
  291                 req.newptr = data;
  292                 break;
  293         case CTLTYPE_U32:
  294                 if (getenv_array(path + rem, data, sizeof(data), &size,
  295                     sizeof(uint32_t), GETENV_UNSIGNED) == 0)
  296                         return;
  297                 req.newlen = size;
  298                 req.newptr = data;
  299                 break;
  300         case CTLTYPE_U64:
  301                 if (getenv_array(path + rem, data, sizeof(data), &size,
  302                     sizeof(uint64_t), GETENV_UNSIGNED) == 0)
  303                         return;
  304                 req.newlen = size;
  305                 req.newptr = data;
  306                 break;
  307         case CTLTYPE_STRING:
  308                 penv = kern_getenv(path + rem);
  309                 if (penv == NULL)
  310                         return;
  311                 req.newlen = strlen(penv);
  312                 req.newptr = penv;
  313                 break;
  314         default:
  315                 return;
  316         }
  317         error = sysctl_root_handler_locked(oidp, oidp->oid_arg1,
  318             oidp->oid_arg2, &req, NULL);
  319         if (error != 0)
  320                 printf("Setting sysctl %s failed: %d\n", path + rem, error);
  321         if (penv != NULL)
  322                 freeenv(penv);
  323 }
  324 
  325 static int
  326 sbuf_printf_drain(void *arg __unused, const char *data, int len)
  327 {
  328 
  329         return (printf("%.*s", len, data));
  330 }
  331 
  332 /*
  333  * Locate the path to a given oid.  Returns the length of the resulting path,
  334  * or -1 if the oid was not found.  nodes must have room for CTL_MAXNAME
  335  * elements and be NULL initialized.
  336  */
  337 static int
  338 sysctl_search_oid(struct sysctl_oid **nodes, struct sysctl_oid *needle)
  339 {
  340         int indx;
  341 
  342         SYSCTL_ASSERT_LOCKED();
  343         indx = 0;
  344         while (indx < CTL_MAXNAME && indx >= 0) {
  345                 if (nodes[indx] == NULL && indx == 0)
  346                         nodes[indx] = SLIST_FIRST(&sysctl__children);
  347                 else if (nodes[indx] == NULL)
  348                         nodes[indx] = SLIST_FIRST(&nodes[indx - 1]->oid_children);
  349                 else
  350                         nodes[indx] = SLIST_NEXT(nodes[indx], oid_link);
  351 
  352                 if (nodes[indx] == needle)
  353                         return (indx + 1);
  354 
  355                 if (nodes[indx] == NULL) {
  356                         indx--;
  357                         continue;
  358                 }
  359 
  360                 if ((nodes[indx]->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
  361                         indx++;
  362                         continue;
  363                 }
  364         }
  365         return (-1);
  366 }
  367 
  368 static void
  369 sysctl_warn_reuse(const char *func, struct sysctl_oid *leaf)
  370 {
  371         struct sysctl_oid *nodes[CTL_MAXNAME];
  372         char buf[128];
  373         struct sbuf sb;
  374         int rc, i;
  375 
  376         (void)sbuf_new(&sb, buf, sizeof(buf), SBUF_FIXEDLEN | SBUF_INCLUDENUL);
  377         sbuf_set_drain(&sb, sbuf_printf_drain, NULL);
  378 
  379         sbuf_printf(&sb, "%s: can't re-use a leaf (", __func__);
  380 
  381         memset(nodes, 0, sizeof(nodes));
  382         rc = sysctl_search_oid(nodes, leaf);
  383         if (rc > 0) {
  384                 for (i = 0; i < rc; i++)
  385                         sbuf_printf(&sb, "%s%.*s", nodes[i]->oid_name,
  386                             i != (rc - 1), ".");
  387         } else {
  388                 sbuf_printf(&sb, "%s", leaf->oid_name);
  389         }
  390         sbuf_printf(&sb, ")!\n");
  391 
  392         (void)sbuf_finish(&sb);
  393 }
  394 
  395 #ifdef SYSCTL_DEBUG
  396 static int
  397 sysctl_reuse_test(SYSCTL_HANDLER_ARGS)
  398 {
  399         struct rm_priotracker tracker;
  400 
  401         SYSCTL_RLOCK(&tracker);
  402         sysctl_warn_reuse(__func__, oidp);
  403         SYSCTL_RUNLOCK(&tracker);
  404         return (0);
  405 }
  406 SYSCTL_PROC(_sysctl, 0, reuse_test, CTLTYPE_STRING|CTLFLAG_RD|CTLFLAG_MPSAFE,
  407         0, 0, sysctl_reuse_test, "-", "");
  408 #endif
  409 
  410 void
  411 sysctl_register_oid(struct sysctl_oid *oidp)
  412 {
  413         struct sysctl_oid_list *parent = oidp->oid_parent;
  414         struct sysctl_oid *p;
  415         struct sysctl_oid *q;
  416         int oid_number;
  417         int timeout = 2;
  418 
  419         /*
  420          * First check if another oid with the same name already
  421          * exists in the parent's list.
  422          */
  423         SYSCTL_ASSERT_WLOCKED();
  424         p = sysctl_find_oidname(oidp->oid_name, parent);
  425         if (p != NULL) {
  426                 if ((p->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
  427                         p->oid_refcnt++;
  428                         return;
  429                 } else {
  430                         sysctl_warn_reuse(__func__, p);
  431                         return;
  432                 }
  433         }
  434         /* get current OID number */
  435         oid_number = oidp->oid_number;
  436 
  437 #if (OID_AUTO >= 0)
  438 #error "OID_AUTO is expected to be a negative value"
  439 #endif  
  440         /*
  441          * Any negative OID number qualifies as OID_AUTO. Valid OID
  442          * numbers should always be positive.
  443          *
  444          * NOTE: DO NOT change the starting value here, change it in
  445          * <sys/sysctl.h>, and make sure it is at least 256 to
  446          * accommodate e.g. net.inet.raw as a static sysctl node.
  447          */
  448         if (oid_number < 0) {
  449                 static int newoid;
  450 
  451                 /*
  452                  * By decrementing the next OID number we spend less
  453                  * time inserting the OIDs into a sorted list.
  454                  */
  455                 if (--newoid < CTL_AUTO_START)
  456                         newoid = 0x7fffffff;
  457 
  458                 oid_number = newoid;
  459         }
  460 
  461         /*
  462          * Insert the OID into the parent's list sorted by OID number.
  463          */
  464 retry:
  465         q = NULL;
  466         SLIST_FOREACH(p, parent, oid_link) {
  467                 /* check if the current OID number is in use */
  468                 if (oid_number == p->oid_number) {
  469                         /* get the next valid OID number */
  470                         if (oid_number < CTL_AUTO_START ||
  471                             oid_number == 0x7fffffff) {
  472                                 /* wraparound - restart */
  473                                 oid_number = CTL_AUTO_START;
  474                                 /* don't loop forever */
  475                                 if (!timeout--)
  476                                         panic("sysctl: Out of OID numbers\n");
  477                                 goto retry;
  478                         } else {
  479                                 oid_number++;
  480                         }
  481                 } else if (oid_number < p->oid_number)
  482                         break;
  483                 q = p;
  484         }
  485         /* check for non-auto OID number collision */
  486         if (oidp->oid_number >= 0 && oidp->oid_number < CTL_AUTO_START &&
  487             oid_number >= CTL_AUTO_START) {
  488                 printf("sysctl: OID number(%d) is already in use for '%s'\n",
  489                     oidp->oid_number, oidp->oid_name);
  490         }
  491         /* update the OID number, if any */
  492         oidp->oid_number = oid_number;
  493         if (q != NULL)
  494                 SLIST_INSERT_AFTER(q, oidp, oid_link);
  495         else
  496                 SLIST_INSERT_HEAD(parent, oidp, oid_link);
  497 
  498         if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE &&
  499 #ifdef VIMAGE
  500             (oidp->oid_kind & CTLFLAG_VNET) == 0 &&
  501 #endif
  502             (oidp->oid_kind & CTLFLAG_TUN) != 0 &&
  503             (oidp->oid_kind & CTLFLAG_NOFETCH) == 0) {
  504                 /* only fetch value once */
  505                 oidp->oid_kind |= CTLFLAG_NOFETCH;
  506                 /* try to fetch value from kernel environment */
  507                 sysctl_load_tunable_by_oid_locked(oidp);
  508         }
  509 }
  510 
  511 void
  512 sysctl_register_disabled_oid(struct sysctl_oid *oidp)
  513 {
  514 
  515         /*
  516          * Mark the leaf as dormant if it's not to be immediately enabled.
  517          * We do not disable nodes as they can be shared between modules
  518          * and it is always safe to access a node.
  519          */
  520         KASSERT((oidp->oid_kind & CTLFLAG_DORMANT) == 0,
  521             ("internal flag is set in oid_kind"));
  522         if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
  523                 oidp->oid_kind |= CTLFLAG_DORMANT;
  524         sysctl_register_oid(oidp);
  525 }
  526 
  527 void
  528 sysctl_enable_oid(struct sysctl_oid *oidp)
  529 {
  530 
  531         SYSCTL_ASSERT_WLOCKED();
  532         if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
  533                 KASSERT((oidp->oid_kind & CTLFLAG_DORMANT) == 0,
  534                     ("sysctl node is marked as dormant"));
  535                 return;
  536         }
  537         KASSERT((oidp->oid_kind & CTLFLAG_DORMANT) != 0,
  538             ("enabling already enabled sysctl oid"));
  539         oidp->oid_kind &= ~CTLFLAG_DORMANT;
  540 }
  541 
  542 void
  543 sysctl_unregister_oid(struct sysctl_oid *oidp)
  544 {
  545         struct sysctl_oid *p;
  546         int error;
  547 
  548         SYSCTL_ASSERT_WLOCKED();
  549         error = ENOENT;
  550         if (oidp->oid_number == OID_AUTO) {
  551                 error = EINVAL;
  552         } else {
  553                 SLIST_FOREACH(p, oidp->oid_parent, oid_link) {
  554                         if (p == oidp) {
  555                                 SLIST_REMOVE(oidp->oid_parent, oidp,
  556                                     sysctl_oid, oid_link);
  557                                 error = 0;
  558                                 break;
  559                         }
  560                 }
  561         }
  562 
  563         /* 
  564          * This can happen when a module fails to register and is
  565          * being unloaded afterwards.  It should not be a panic()
  566          * for normal use.
  567          */
  568         if (error)
  569                 printf("%s: failed to unregister sysctl\n", __func__);
  570 }
  571 
  572 /* Initialize a new context to keep track of dynamically added sysctls. */
  573 int
  574 sysctl_ctx_init(struct sysctl_ctx_list *c)
  575 {
  576 
  577         if (c == NULL) {
  578                 return (EINVAL);
  579         }
  580 
  581         /*
  582          * No locking here, the caller is responsible for not adding
  583          * new nodes to a context until after this function has
  584          * returned.
  585          */
  586         TAILQ_INIT(c);
  587         return (0);
  588 }
  589 
  590 /* Free the context, and destroy all dynamic oids registered in this context */
  591 int
  592 sysctl_ctx_free(struct sysctl_ctx_list *clist)
  593 {
  594         struct sysctl_ctx_entry *e, *e1;
  595         int error;
  596 
  597         error = 0;
  598         /*
  599          * First perform a "dry run" to check if it's ok to remove oids.
  600          * XXX FIXME
  601          * XXX This algorithm is a hack. But I don't know any
  602          * XXX better solution for now...
  603          */
  604         SYSCTL_WLOCK();
  605         TAILQ_FOREACH(e, clist, link) {
  606                 error = sysctl_remove_oid_locked(e->entry, 0, 0);
  607                 if (error)
  608                         break;
  609         }
  610         /*
  611          * Restore deregistered entries, either from the end,
  612          * or from the place where error occurred.
  613          * e contains the entry that was not unregistered
  614          */
  615         if (error)
  616                 e1 = TAILQ_PREV(e, sysctl_ctx_list, link);
  617         else
  618                 e1 = TAILQ_LAST(clist, sysctl_ctx_list);
  619         while (e1 != NULL) {
  620                 sysctl_register_oid(e1->entry);
  621                 e1 = TAILQ_PREV(e1, sysctl_ctx_list, link);
  622         }
  623         if (error) {
  624                 SYSCTL_WUNLOCK();
  625                 return(EBUSY);
  626         }
  627         /* Now really delete the entries */
  628         e = TAILQ_FIRST(clist);
  629         while (e != NULL) {
  630                 e1 = TAILQ_NEXT(e, link);
  631                 error = sysctl_remove_oid_locked(e->entry, 1, 0);
  632                 if (error)
  633                         panic("sysctl_remove_oid: corrupt tree, entry: %s",
  634                             e->entry->oid_name);
  635                 free(e, M_SYSCTLOID);
  636                 e = e1;
  637         }
  638         SYSCTL_WUNLOCK();
  639         return (error);
  640 }
  641 
  642 /* Add an entry to the context */
  643 struct sysctl_ctx_entry *
  644 sysctl_ctx_entry_add(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
  645 {
  646         struct sysctl_ctx_entry *e;
  647 
  648         SYSCTL_ASSERT_WLOCKED();
  649         if (clist == NULL || oidp == NULL)
  650                 return(NULL);
  651         e = malloc(sizeof(struct sysctl_ctx_entry), M_SYSCTLOID, M_WAITOK);
  652         e->entry = oidp;
  653         TAILQ_INSERT_HEAD(clist, e, link);
  654         return (e);
  655 }
  656 
  657 /* Find an entry in the context */
  658 struct sysctl_ctx_entry *
  659 sysctl_ctx_entry_find(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
  660 {
  661         struct sysctl_ctx_entry *e;
  662 
  663         SYSCTL_ASSERT_WLOCKED();
  664         if (clist == NULL || oidp == NULL)
  665                 return(NULL);
  666         TAILQ_FOREACH(e, clist, link) {
  667                 if(e->entry == oidp)
  668                         return(e);
  669         }
  670         return (e);
  671 }
  672 
  673 /*
  674  * Delete an entry from the context.
  675  * NOTE: this function doesn't free oidp! You have to remove it
  676  * with sysctl_remove_oid().
  677  */
  678 int
  679 sysctl_ctx_entry_del(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp)
  680 {
  681         struct sysctl_ctx_entry *e;
  682 
  683         if (clist == NULL || oidp == NULL)
  684                 return (EINVAL);
  685         SYSCTL_WLOCK();
  686         e = sysctl_ctx_entry_find(clist, oidp);
  687         if (e != NULL) {
  688                 TAILQ_REMOVE(clist, e, link);
  689                 SYSCTL_WUNLOCK();
  690                 free(e, M_SYSCTLOID);
  691                 return (0);
  692         } else {
  693                 SYSCTL_WUNLOCK();
  694                 return (ENOENT);
  695         }
  696 }
  697 
  698 /*
  699  * Remove dynamically created sysctl trees.
  700  * oidp - top of the tree to be removed
  701  * del - if 0 - just deregister, otherwise free up entries as well
  702  * recurse - if != 0 traverse the subtree to be deleted
  703  */
  704 int
  705 sysctl_remove_oid(struct sysctl_oid *oidp, int del, int recurse)
  706 {
  707         int error;
  708 
  709         SYSCTL_WLOCK();
  710         error = sysctl_remove_oid_locked(oidp, del, recurse);
  711         SYSCTL_WUNLOCK();
  712         return (error);
  713 }
  714 
  715 int
  716 sysctl_remove_name(struct sysctl_oid *parent, const char *name,
  717     int del, int recurse)
  718 {
  719         struct sysctl_oid *p, *tmp;
  720         int error;
  721 
  722         error = ENOENT;
  723         SYSCTL_WLOCK();
  724         SLIST_FOREACH_SAFE(p, SYSCTL_CHILDREN(parent), oid_link, tmp) {
  725                 if (strcmp(p->oid_name, name) == 0) {
  726                         error = sysctl_remove_oid_locked(p, del, recurse);
  727                         break;
  728                 }
  729         }
  730         SYSCTL_WUNLOCK();
  731 
  732         return (error);
  733 }
  734 
  735 
  736 static int
  737 sysctl_remove_oid_locked(struct sysctl_oid *oidp, int del, int recurse)
  738 {
  739         struct sysctl_oid *p, *tmp;
  740         int error;
  741 
  742         SYSCTL_ASSERT_WLOCKED();
  743         if (oidp == NULL)
  744                 return(EINVAL);
  745         if ((oidp->oid_kind & CTLFLAG_DYN) == 0) {
  746                 printf("Warning: can't remove non-dynamic nodes (%s)!\n",
  747                     oidp->oid_name);
  748                 return (EINVAL);
  749         }
  750         /*
  751          * WARNING: normal method to do this should be through
  752          * sysctl_ctx_free(). Use recursing as the last resort
  753          * method to purge your sysctl tree of leftovers...
  754          * However, if some other code still references these nodes,
  755          * it will panic.
  756          */
  757         if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
  758                 if (oidp->oid_refcnt == 1) {
  759                         SLIST_FOREACH_SAFE(p,
  760                             SYSCTL_CHILDREN(oidp), oid_link, tmp) {
  761                                 if (!recurse) {
  762                                         printf("Warning: failed attempt to "
  763                                             "remove oid %s with child %s\n",
  764                                             oidp->oid_name, p->oid_name);
  765                                         return (ENOTEMPTY);
  766                                 }
  767                                 error = sysctl_remove_oid_locked(p, del,
  768                                     recurse);
  769                                 if (error)
  770                                         return (error);
  771                         }
  772                 }
  773         }
  774         if (oidp->oid_refcnt > 1 ) {
  775                 oidp->oid_refcnt--;
  776         } else {
  777                 if (oidp->oid_refcnt == 0) {
  778                         printf("Warning: bad oid_refcnt=%u (%s)!\n",
  779                                 oidp->oid_refcnt, oidp->oid_name);
  780                         return (EINVAL);
  781                 }
  782                 sysctl_unregister_oid(oidp);
  783                 if (del) {
  784                         /*
  785                          * Wait for all threads running the handler to drain.
  786                          * This preserves the previous behavior when the
  787                          * sysctl lock was held across a handler invocation,
  788                          * and is necessary for module unload correctness.
  789                          */
  790                         while (oidp->oid_running > 0) {
  791                                 oidp->oid_kind |= CTLFLAG_DYING;
  792                                 SYSCTL_SLEEP(&oidp->oid_running, "oidrm", 0);
  793                         }
  794                         if (oidp->oid_descr)
  795                                 free(__DECONST(char *, oidp->oid_descr),
  796                                     M_SYSCTLOID);
  797                         if (oidp->oid_label)
  798                                 free(__DECONST(char *, oidp->oid_label),
  799                                     M_SYSCTLOID);
  800                         free(__DECONST(char *, oidp->oid_name), M_SYSCTLOID);
  801                         free(oidp, M_SYSCTLOID);
  802                 }
  803         }
  804         return (0);
  805 }
  806 /*
  807  * Create new sysctls at run time.
  808  * clist may point to a valid context initialized with sysctl_ctx_init().
  809  */
  810 struct sysctl_oid *
  811 sysctl_add_oid(struct sysctl_ctx_list *clist, struct sysctl_oid_list *parent,
  812         int number, const char *name, int kind, void *arg1, intmax_t arg2,
  813         int (*handler)(SYSCTL_HANDLER_ARGS), const char *fmt, const char *descr,
  814         const char *label)
  815 {
  816         struct sysctl_oid *oidp;
  817 
  818         /* You have to hook up somewhere.. */
  819         if (parent == NULL)
  820                 return(NULL);
  821         /* Check if the node already exists, otherwise create it */
  822         SYSCTL_WLOCK();
  823         oidp = sysctl_find_oidname(name, parent);
  824         if (oidp != NULL) {
  825                 if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
  826                         oidp->oid_refcnt++;
  827                         /* Update the context */
  828                         if (clist != NULL)
  829                                 sysctl_ctx_entry_add(clist, oidp);
  830                         SYSCTL_WUNLOCK();
  831                         return (oidp);
  832                 } else {
  833                         sysctl_warn_reuse(__func__, oidp);
  834                         SYSCTL_WUNLOCK();
  835                         return (NULL);
  836                 }
  837         }
  838         oidp = malloc(sizeof(struct sysctl_oid), M_SYSCTLOID, M_WAITOK|M_ZERO);
  839         oidp->oid_parent = parent;
  840         SLIST_INIT(&oidp->oid_children);
  841         oidp->oid_number = number;
  842         oidp->oid_refcnt = 1;
  843         oidp->oid_name = strdup(name, M_SYSCTLOID);
  844         oidp->oid_handler = handler;
  845         oidp->oid_kind = CTLFLAG_DYN | kind;
  846         oidp->oid_arg1 = arg1;
  847         oidp->oid_arg2 = arg2;
  848         oidp->oid_fmt = fmt;
  849         if (descr != NULL)
  850                 oidp->oid_descr = strdup(descr, M_SYSCTLOID);
  851         if (label != NULL)
  852                 oidp->oid_label = strdup(label, M_SYSCTLOID);
  853         /* Update the context, if used */
  854         if (clist != NULL)
  855                 sysctl_ctx_entry_add(clist, oidp);
  856         /* Register this oid */
  857         sysctl_register_oid(oidp);
  858         SYSCTL_WUNLOCK();
  859         return (oidp);
  860 }
  861 
  862 /*
  863  * Rename an existing oid.
  864  */
  865 void
  866 sysctl_rename_oid(struct sysctl_oid *oidp, const char *name)
  867 {
  868         char *newname;
  869         char *oldname;
  870 
  871         newname = strdup(name, M_SYSCTLOID);
  872         SYSCTL_WLOCK();
  873         oldname = __DECONST(char *, oidp->oid_name);
  874         oidp->oid_name = newname;
  875         SYSCTL_WUNLOCK();
  876         free(oldname, M_SYSCTLOID);
  877 }
  878 
  879 /*
  880  * Reparent an existing oid.
  881  */
  882 int
  883 sysctl_move_oid(struct sysctl_oid *oid, struct sysctl_oid_list *parent)
  884 {
  885         struct sysctl_oid *oidp;
  886 
  887         SYSCTL_WLOCK();
  888         if (oid->oid_parent == parent) {
  889                 SYSCTL_WUNLOCK();
  890                 return (0);
  891         }
  892         oidp = sysctl_find_oidname(oid->oid_name, parent);
  893         if (oidp != NULL) {
  894                 SYSCTL_WUNLOCK();
  895                 return (EEXIST);
  896         }
  897         sysctl_unregister_oid(oid);
  898         oid->oid_parent = parent;
  899         oid->oid_number = OID_AUTO;
  900         sysctl_register_oid(oid);
  901         SYSCTL_WUNLOCK();
  902         return (0);
  903 }
  904 
  905 /*
  906  * Register the kernel's oids on startup.
  907  */
  908 SET_DECLARE(sysctl_set, struct sysctl_oid);
  909 
  910 static void
  911 sysctl_register_all(void *arg)
  912 {
  913         struct sysctl_oid **oidp;
  914 
  915         sx_init(&sysctlmemlock, "sysctl mem");
  916         SYSCTL_INIT();
  917         SYSCTL_WLOCK();
  918         SET_FOREACH(oidp, sysctl_set)
  919                 sysctl_register_oid(*oidp);
  920         SYSCTL_WUNLOCK();
  921 }
  922 SYSINIT(sysctl, SI_SUB_KMEM, SI_ORDER_FIRST, sysctl_register_all, NULL);
  923 
  924 /*
  925  * "Staff-functions"
  926  *
  927  * These functions implement a presently undocumented interface 
  928  * used by the sysctl program to walk the tree, and get the type
  929  * so it can print the value.
  930  * This interface is under work and consideration, and should probably
  931  * be killed with a big axe by the first person who can find the time.
  932  * (be aware though, that the proper interface isn't as obvious as it
  933  * may seem, there are various conflicting requirements.
  934  *
  935  * {0,0}        printf the entire MIB-tree.
  936  * {0,1,...}    return the name of the "..." OID.
  937  * {0,2,...}    return the next OID.
  938  * {0,3}        return the OID of the name in "new"
  939  * {0,4,...}    return the kind & format info for the "..." OID.
  940  * {0,5,...}    return the description of the "..." OID.
  941  * {0,6,...}    return the aggregation label of the "..." OID.
  942  */
  943 
  944 #ifdef SYSCTL_DEBUG
  945 static void
  946 sysctl_sysctl_debug_dump_node(struct sysctl_oid_list *l, int i)
  947 {
  948         int k;
  949         struct sysctl_oid *oidp;
  950 
  951         SYSCTL_ASSERT_LOCKED();
  952         SLIST_FOREACH(oidp, l, oid_link) {
  953 
  954                 for (k=0; k<i; k++)
  955                         printf(" ");
  956 
  957                 printf("%d %s ", oidp->oid_number, oidp->oid_name);
  958 
  959                 printf("%c%c",
  960                         oidp->oid_kind & CTLFLAG_RD ? 'R':' ',
  961                         oidp->oid_kind & CTLFLAG_WR ? 'W':' ');
  962 
  963                 if (oidp->oid_handler)
  964                         printf(" *Handler");
  965 
  966                 switch (oidp->oid_kind & CTLTYPE) {
  967                         case CTLTYPE_NODE:
  968                                 printf(" Node\n");
  969                                 if (!oidp->oid_handler) {
  970                                         sysctl_sysctl_debug_dump_node(
  971                                             SYSCTL_CHILDREN(oidp), i + 2);
  972                                 }
  973                                 break;
  974                         case CTLTYPE_INT:    printf(" Int\n"); break;
  975                         case CTLTYPE_UINT:   printf(" u_int\n"); break;
  976                         case CTLTYPE_LONG:   printf(" Long\n"); break;
  977                         case CTLTYPE_ULONG:  printf(" u_long\n"); break;
  978                         case CTLTYPE_STRING: printf(" String\n"); break;
  979                         case CTLTYPE_S8:     printf(" int8_t\n"); break;
  980                         case CTLTYPE_S16:    printf(" int16_t\n"); break;
  981                         case CTLTYPE_S32:    printf(" int32_t\n"); break;
  982                         case CTLTYPE_S64:    printf(" int64_t\n"); break;
  983                         case CTLTYPE_U8:     printf(" uint8_t\n"); break;
  984                         case CTLTYPE_U16:    printf(" uint16_t\n"); break;
  985                         case CTLTYPE_U32:    printf(" uint32_t\n"); break;
  986                         case CTLTYPE_U64:    printf(" uint64_t\n"); break;
  987                         case CTLTYPE_OPAQUE: printf(" Opaque/struct\n"); break;
  988                         default:             printf("\n");
  989                 }
  990 
  991         }
  992 }
  993 
  994 static int
  995 sysctl_sysctl_debug(SYSCTL_HANDLER_ARGS)
  996 {
  997         struct rm_priotracker tracker;
  998         int error;
  999 
 1000         error = priv_check(req->td, PRIV_SYSCTL_DEBUG);
 1001         if (error)
 1002                 return (error);
 1003         SYSCTL_RLOCK(&tracker);
 1004         sysctl_sysctl_debug_dump_node(&sysctl__children, 0);
 1005         SYSCTL_RUNLOCK(&tracker);
 1006         return (ENOENT);
 1007 }
 1008 
 1009 SYSCTL_PROC(_sysctl, 0, debug, CTLTYPE_STRING|CTLFLAG_RD|CTLFLAG_MPSAFE,
 1010         0, 0, sysctl_sysctl_debug, "-", "");
 1011 #endif
 1012 
 1013 static int
 1014 sysctl_sysctl_name(SYSCTL_HANDLER_ARGS)
 1015 {
 1016         int *name = (int *) arg1;
 1017         u_int namelen = arg2;
 1018         int error = 0;
 1019         struct sysctl_oid *oid;
 1020         struct sysctl_oid_list *lsp = &sysctl__children, *lsp2;
 1021         struct rm_priotracker tracker;
 1022         char buf[10];
 1023 
 1024         SYSCTL_RLOCK(&tracker);
 1025         while (namelen) {
 1026                 if (!lsp) {
 1027                         snprintf(buf,sizeof(buf),"%d",*name);
 1028                         if (req->oldidx)
 1029                                 error = SYSCTL_OUT(req, ".", 1);
 1030                         if (!error)
 1031                                 error = SYSCTL_OUT(req, buf, strlen(buf));
 1032                         if (error)
 1033                                 goto out;
 1034                         namelen--;
 1035                         name++;
 1036                         continue;
 1037                 }
 1038                 lsp2 = NULL;
 1039                 SLIST_FOREACH(oid, lsp, oid_link) {
 1040                         if (oid->oid_number != *name)
 1041                                 continue;
 1042 
 1043                         if (req->oldidx)
 1044                                 error = SYSCTL_OUT(req, ".", 1);
 1045                         if (!error)
 1046                                 error = SYSCTL_OUT(req, oid->oid_name,
 1047                                         strlen(oid->oid_name));
 1048                         if (error)
 1049                                 goto out;
 1050 
 1051                         namelen--;
 1052                         name++;
 1053 
 1054                         if ((oid->oid_kind & CTLTYPE) != CTLTYPE_NODE) 
 1055                                 break;
 1056 
 1057                         if (oid->oid_handler)
 1058                                 break;
 1059 
 1060                         lsp2 = SYSCTL_CHILDREN(oid);
 1061                         break;
 1062                 }
 1063                 lsp = lsp2;
 1064         }
 1065         error = SYSCTL_OUT(req, "", 1);
 1066  out:
 1067         SYSCTL_RUNLOCK(&tracker);
 1068         return (error);
 1069 }
 1070 
 1071 /*
 1072  * XXXRW/JA: Shouldn't return name data for nodes that we don't permit in
 1073  * capability mode.
 1074  */
 1075 static SYSCTL_NODE(_sysctl, 1, name, CTLFLAG_RD | CTLFLAG_MPSAFE | CTLFLAG_CAPRD,
 1076     sysctl_sysctl_name, "");
 1077 
 1078 static int
 1079 sysctl_sysctl_next_ls(struct sysctl_oid_list *lsp, int *name, u_int namelen, 
 1080         int *next, int *len, int level, struct sysctl_oid **oidpp)
 1081 {
 1082         struct sysctl_oid *oidp;
 1083 
 1084         SYSCTL_ASSERT_LOCKED();
 1085         *len = level;
 1086         SLIST_FOREACH(oidp, lsp, oid_link) {
 1087                 *next = oidp->oid_number;
 1088                 *oidpp = oidp;
 1089 
 1090                 if ((oidp->oid_kind & (CTLFLAG_SKIP | CTLFLAG_DORMANT)) != 0)
 1091                         continue;
 1092 
 1093                 if (!namelen) {
 1094                         if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE) 
 1095                                 return (0);
 1096                         if (oidp->oid_handler) 
 1097                                 /* We really should call the handler here...*/
 1098                                 return (0);
 1099                         lsp = SYSCTL_CHILDREN(oidp);
 1100                         if (!sysctl_sysctl_next_ls(lsp, 0, 0, next+1, 
 1101                                 len, level+1, oidpp))
 1102                                 return (0);
 1103                         goto emptynode;
 1104                 }
 1105 
 1106                 if (oidp->oid_number < *name)
 1107                         continue;
 1108 
 1109                 if (oidp->oid_number > *name) {
 1110                         if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
 1111                                 return (0);
 1112                         if (oidp->oid_handler)
 1113                                 return (0);
 1114                         lsp = SYSCTL_CHILDREN(oidp);
 1115                         if (!sysctl_sysctl_next_ls(lsp, name+1, namelen-1, 
 1116                                 next+1, len, level+1, oidpp))
 1117                                 return (0);
 1118                         goto next;
 1119                 }
 1120                 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
 1121                         continue;
 1122 
 1123                 if (oidp->oid_handler)
 1124                         continue;
 1125 
 1126                 lsp = SYSCTL_CHILDREN(oidp);
 1127                 if (!sysctl_sysctl_next_ls(lsp, name+1, namelen-1, next+1, 
 1128                         len, level+1, oidpp))
 1129                         return (0);
 1130         next:
 1131                 namelen = 1;
 1132         emptynode:
 1133                 *len = level;
 1134         }
 1135         return (1);
 1136 }
 1137 
 1138 static int
 1139 sysctl_sysctl_next(SYSCTL_HANDLER_ARGS)
 1140 {
 1141         int *name = (int *) arg1;
 1142         u_int namelen = arg2;
 1143         int i, j, error;
 1144         struct sysctl_oid *oid;
 1145         struct sysctl_oid_list *lsp = &sysctl__children;
 1146         struct rm_priotracker tracker;
 1147         int newoid[CTL_MAXNAME];
 1148 
 1149         SYSCTL_RLOCK(&tracker);
 1150         i = sysctl_sysctl_next_ls(lsp, name, namelen, newoid, &j, 1, &oid);
 1151         SYSCTL_RUNLOCK(&tracker);
 1152         if (i)
 1153                 return (ENOENT);
 1154         error = SYSCTL_OUT(req, newoid, j * sizeof (int));
 1155         return (error);
 1156 }
 1157 
 1158 /*
 1159  * XXXRW/JA: Shouldn't return next data for nodes that we don't permit in
 1160  * capability mode.
 1161  */
 1162 static SYSCTL_NODE(_sysctl, 2, next, CTLFLAG_RD | CTLFLAG_MPSAFE | CTLFLAG_CAPRD,
 1163     sysctl_sysctl_next, "");
 1164 
 1165 static int
 1166 name2oid(char *name, int *oid, int *len, struct sysctl_oid **oidpp)
 1167 {
 1168         struct sysctl_oid *oidp;
 1169         struct sysctl_oid_list *lsp = &sysctl__children;
 1170         char *p;
 1171 
 1172         SYSCTL_ASSERT_LOCKED();
 1173 
 1174         for (*len = 0; *len < CTL_MAXNAME;) {
 1175                 p = strsep(&name, ".");
 1176 
 1177                 oidp = SLIST_FIRST(lsp);
 1178                 for (;; oidp = SLIST_NEXT(oidp, oid_link)) {
 1179                         if (oidp == NULL)
 1180                                 return (ENOENT);
 1181                         if (strcmp(p, oidp->oid_name) == 0)
 1182                                 break;
 1183                 }
 1184                 *oid++ = oidp->oid_number;
 1185                 (*len)++;
 1186 
 1187                 if (name == NULL || *name == '\0') {
 1188                         if (oidpp)
 1189                                 *oidpp = oidp;
 1190                         return (0);
 1191                 }
 1192 
 1193                 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE)
 1194                         break;
 1195 
 1196                 if (oidp->oid_handler)
 1197                         break;
 1198 
 1199                 lsp = SYSCTL_CHILDREN(oidp);
 1200         }
 1201         return (ENOENT);
 1202 }
 1203 
 1204 static int
 1205 sysctl_sysctl_name2oid(SYSCTL_HANDLER_ARGS)
 1206 {
 1207         char *p;
 1208         int error, oid[CTL_MAXNAME], len = 0;
 1209         struct sysctl_oid *op = NULL;
 1210         struct rm_priotracker tracker;
 1211         char buf[32];
 1212 
 1213         if (!req->newlen) 
 1214                 return (ENOENT);
 1215         if (req->newlen >= MAXPATHLEN)  /* XXX arbitrary, undocumented */
 1216                 return (ENAMETOOLONG);
 1217 
 1218         p = buf;
 1219         if (req->newlen >= sizeof(buf))
 1220                 p = malloc(req->newlen+1, M_SYSCTL, M_WAITOK);
 1221 
 1222         error = SYSCTL_IN(req, p, req->newlen);
 1223         if (error) {
 1224                 if (p != buf)
 1225                         free(p, M_SYSCTL);
 1226                 return (error);
 1227         }
 1228 
 1229         p [req->newlen] = '\0';
 1230 
 1231         SYSCTL_RLOCK(&tracker);
 1232         error = name2oid(p, oid, &len, &op);
 1233         SYSCTL_RUNLOCK(&tracker);
 1234 
 1235         if (p != buf)
 1236                 free(p, M_SYSCTL);
 1237 
 1238         if (error)
 1239                 return (error);
 1240 
 1241         error = SYSCTL_OUT(req, oid, len * sizeof *oid);
 1242         return (error);
 1243 }
 1244 
 1245 /*
 1246  * XXXRW/JA: Shouldn't return name2oid data for nodes that we don't permit in
 1247  * capability mode.
 1248  */
 1249 SYSCTL_PROC(_sysctl, 3, name2oid,
 1250     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE
 1251     | CTLFLAG_CAPRW, 0, 0, sysctl_sysctl_name2oid, "I", "");
 1252 
 1253 static int
 1254 sysctl_sysctl_oidfmt(SYSCTL_HANDLER_ARGS)
 1255 {
 1256         struct sysctl_oid *oid;
 1257         struct rm_priotracker tracker;
 1258         int error;
 1259 
 1260         SYSCTL_RLOCK(&tracker);
 1261         error = sysctl_find_oid(arg1, arg2, &oid, NULL, req);
 1262         if (error)
 1263                 goto out;
 1264 
 1265         if (oid->oid_fmt == NULL) {
 1266                 error = ENOENT;
 1267                 goto out;
 1268         }
 1269         error = SYSCTL_OUT(req, &oid->oid_kind, sizeof(oid->oid_kind));
 1270         if (error)
 1271                 goto out;
 1272         error = SYSCTL_OUT(req, oid->oid_fmt, strlen(oid->oid_fmt) + 1);
 1273  out:
 1274         SYSCTL_RUNLOCK(&tracker);
 1275         return (error);
 1276 }
 1277 
 1278 
 1279 static SYSCTL_NODE(_sysctl, 4, oidfmt, CTLFLAG_RD|CTLFLAG_MPSAFE|CTLFLAG_CAPRD,
 1280     sysctl_sysctl_oidfmt, "");
 1281 
 1282 static int
 1283 sysctl_sysctl_oiddescr(SYSCTL_HANDLER_ARGS)
 1284 {
 1285         struct sysctl_oid *oid;
 1286         struct rm_priotracker tracker;
 1287         int error;
 1288 
 1289         SYSCTL_RLOCK(&tracker);
 1290         error = sysctl_find_oid(arg1, arg2, &oid, NULL, req);
 1291         if (error)
 1292                 goto out;
 1293 
 1294         if (oid->oid_descr == NULL) {
 1295                 error = ENOENT;
 1296                 goto out;
 1297         }
 1298         error = SYSCTL_OUT(req, oid->oid_descr, strlen(oid->oid_descr) + 1);
 1299  out:
 1300         SYSCTL_RUNLOCK(&tracker);
 1301         return (error);
 1302 }
 1303 
 1304 static SYSCTL_NODE(_sysctl, 5, oiddescr, CTLFLAG_RD|CTLFLAG_MPSAFE|CTLFLAG_CAPRD,
 1305     sysctl_sysctl_oiddescr, "");
 1306 
 1307 static int
 1308 sysctl_sysctl_oidlabel(SYSCTL_HANDLER_ARGS)
 1309 {
 1310         struct sysctl_oid *oid;
 1311         struct rm_priotracker tracker;
 1312         int error;
 1313 
 1314         SYSCTL_RLOCK(&tracker);
 1315         error = sysctl_find_oid(arg1, arg2, &oid, NULL, req);
 1316         if (error)
 1317                 goto out;
 1318 
 1319         if (oid->oid_label == NULL) {
 1320                 error = ENOENT;
 1321                 goto out;
 1322         }
 1323         error = SYSCTL_OUT(req, oid->oid_label, strlen(oid->oid_label) + 1);
 1324  out:
 1325         SYSCTL_RUNLOCK(&tracker);
 1326         return (error);
 1327 }
 1328 
 1329 static SYSCTL_NODE(_sysctl, 6, oidlabel,
 1330     CTLFLAG_RD | CTLFLAG_MPSAFE | CTLFLAG_CAPRD, sysctl_sysctl_oidlabel, "");
 1331 
 1332 /*
 1333  * Default "handler" functions.
 1334  */
 1335 
 1336 /*
 1337  * Handle a bool.
 1338  * Two cases:
 1339  *     a variable:  point arg1 at it.
 1340  *     a constant:  pass it in arg2.
 1341  */
 1342 
 1343 int
 1344 sysctl_handle_bool(SYSCTL_HANDLER_ARGS)
 1345 {
 1346         uint8_t temp;
 1347         int error;
 1348 
 1349         /*
 1350          * Attempt to get a coherent snapshot by making a copy of the data.
 1351          */
 1352         if (arg1)
 1353                 temp = *(bool *)arg1 ? 1 : 0;
 1354         else
 1355                 temp = arg2 ? 1 : 0;
 1356 
 1357         error = SYSCTL_OUT(req, &temp, sizeof(temp));
 1358         if (error || !req->newptr)
 1359                 return (error);
 1360 
 1361         if (!arg1)
 1362                 error = EPERM;
 1363         else {
 1364                 error = SYSCTL_IN(req, &temp, sizeof(temp));
 1365                 if (!error)
 1366                         *(bool *)arg1 = temp ? 1 : 0;
 1367         }
 1368         return (error);
 1369 }
 1370 
 1371 /*
 1372  * Handle an int8_t, signed or unsigned.
 1373  * Two cases:
 1374  *     a variable:  point arg1 at it.
 1375  *     a constant:  pass it in arg2.
 1376  */
 1377 
 1378 int
 1379 sysctl_handle_8(SYSCTL_HANDLER_ARGS)
 1380 {
 1381         int8_t tmpout;
 1382         int error = 0;
 1383 
 1384         /*
 1385          * Attempt to get a coherent snapshot by making a copy of the data.
 1386          */
 1387         if (arg1)
 1388                 tmpout = *(int8_t *)arg1;
 1389         else
 1390                 tmpout = arg2;
 1391         error = SYSCTL_OUT(req, &tmpout, sizeof(tmpout));
 1392 
 1393         if (error || !req->newptr)
 1394                 return (error);
 1395 
 1396         if (!arg1)
 1397                 error = EPERM;
 1398         else
 1399                 error = SYSCTL_IN(req, arg1, sizeof(tmpout));
 1400         return (error);
 1401 }
 1402 
 1403 /*
 1404  * Handle an int16_t, signed or unsigned.
 1405  * Two cases:
 1406  *     a variable:  point arg1 at it.
 1407  *     a constant:  pass it in arg2.
 1408  */
 1409 
 1410 int
 1411 sysctl_handle_16(SYSCTL_HANDLER_ARGS)
 1412 {
 1413         int16_t tmpout;
 1414         int error = 0;
 1415 
 1416         /*
 1417          * Attempt to get a coherent snapshot by making a copy of the data.
 1418          */
 1419         if (arg1)
 1420                 tmpout = *(int16_t *)arg1;
 1421         else
 1422                 tmpout = arg2;
 1423         error = SYSCTL_OUT(req, &tmpout, sizeof(tmpout));
 1424 
 1425         if (error || !req->newptr)
 1426                 return (error);
 1427 
 1428         if (!arg1)
 1429                 error = EPERM;
 1430         else
 1431                 error = SYSCTL_IN(req, arg1, sizeof(tmpout));
 1432         return (error);
 1433 }
 1434 
 1435 /*
 1436  * Handle an int32_t, signed or unsigned.
 1437  * Two cases:
 1438  *     a variable:  point arg1 at it.
 1439  *     a constant:  pass it in arg2.
 1440  */
 1441 
 1442 int
 1443 sysctl_handle_32(SYSCTL_HANDLER_ARGS)
 1444 {
 1445         int32_t tmpout;
 1446         int error = 0;
 1447 
 1448         /*
 1449          * Attempt to get a coherent snapshot by making a copy of the data.
 1450          */
 1451         if (arg1)
 1452                 tmpout = *(int32_t *)arg1;
 1453         else
 1454                 tmpout = arg2;
 1455         error = SYSCTL_OUT(req, &tmpout, sizeof(tmpout));
 1456 
 1457         if (error || !req->newptr)
 1458                 return (error);
 1459 
 1460         if (!arg1)
 1461                 error = EPERM;
 1462         else
 1463                 error = SYSCTL_IN(req, arg1, sizeof(tmpout));
 1464         return (error);
 1465 }
 1466 
 1467 /*
 1468  * Handle an int, signed or unsigned.
 1469  * Two cases:
 1470  *     a variable:  point arg1 at it.
 1471  *     a constant:  pass it in arg2.
 1472  */
 1473 
 1474 int
 1475 sysctl_handle_int(SYSCTL_HANDLER_ARGS)
 1476 {
 1477         int tmpout, error = 0;
 1478 
 1479         /*
 1480          * Attempt to get a coherent snapshot by making a copy of the data.
 1481          */
 1482         if (arg1)
 1483                 tmpout = *(int *)arg1;
 1484         else
 1485                 tmpout = arg2;
 1486         error = SYSCTL_OUT(req, &tmpout, sizeof(int));
 1487 
 1488         if (error || !req->newptr)
 1489                 return (error);
 1490 
 1491         if (!arg1)
 1492                 error = EPERM;
 1493         else
 1494                 error = SYSCTL_IN(req, arg1, sizeof(int));
 1495         return (error);
 1496 }
 1497 
 1498 /*
 1499  * Based on on sysctl_handle_int() convert milliseconds into ticks.
 1500  * Note: this is used by TCP.
 1501  */
 1502 
 1503 int
 1504 sysctl_msec_to_ticks(SYSCTL_HANDLER_ARGS)
 1505 {
 1506         int error, s, tt;
 1507 
 1508         tt = *(int *)arg1;
 1509         s = (int)((int64_t)tt * 1000 / hz);
 1510 
 1511         error = sysctl_handle_int(oidp, &s, 0, req);
 1512         if (error || !req->newptr)
 1513                 return (error);
 1514 
 1515         tt = (int)((int64_t)s * hz / 1000);
 1516         if (tt < 1)
 1517                 return (EINVAL);
 1518 
 1519         *(int *)arg1 = tt;
 1520         return (0);
 1521 }
 1522 
 1523 
 1524 /*
 1525  * Handle a long, signed or unsigned.
 1526  * Two cases:
 1527  *     a variable:  point arg1 at it.
 1528  *     a constant:  pass it in arg2.
 1529  */
 1530 
 1531 int
 1532 sysctl_handle_long(SYSCTL_HANDLER_ARGS)
 1533 {
 1534         int error = 0;
 1535         long tmplong;
 1536 #ifdef SCTL_MASK32
 1537         int tmpint;
 1538 #endif
 1539 
 1540         /*
 1541          * Attempt to get a coherent snapshot by making a copy of the data.
 1542          */
 1543         if (arg1)
 1544                 tmplong = *(long *)arg1;
 1545         else
 1546                 tmplong = arg2;
 1547 #ifdef SCTL_MASK32
 1548         if (req->flags & SCTL_MASK32) {
 1549                 tmpint = tmplong;
 1550                 error = SYSCTL_OUT(req, &tmpint, sizeof(int));
 1551         } else
 1552 #endif
 1553                 error = SYSCTL_OUT(req, &tmplong, sizeof(long));
 1554 
 1555         if (error || !req->newptr)
 1556                 return (error);
 1557 
 1558         if (!arg1)
 1559                 error = EPERM;
 1560 #ifdef SCTL_MASK32
 1561         else if (req->flags & SCTL_MASK32) {
 1562                 error = SYSCTL_IN(req, &tmpint, sizeof(int));
 1563                 *(long *)arg1 = (long)tmpint;
 1564         }
 1565 #endif
 1566         else
 1567                 error = SYSCTL_IN(req, arg1, sizeof(long));
 1568         return (error);
 1569 }
 1570 
 1571 /*
 1572  * Handle a 64 bit int, signed or unsigned.
 1573  * Two cases:
 1574  *     a variable:  point arg1 at it.
 1575  *     a constant:  pass it in arg2.
 1576  */
 1577 int
 1578 sysctl_handle_64(SYSCTL_HANDLER_ARGS)
 1579 {
 1580         int error = 0;
 1581         uint64_t tmpout;
 1582 
 1583         /*
 1584          * Attempt to get a coherent snapshot by making a copy of the data.
 1585          */
 1586         if (arg1)
 1587                 tmpout = *(uint64_t *)arg1;
 1588         else
 1589                 tmpout = arg2;
 1590         error = SYSCTL_OUT(req, &tmpout, sizeof(uint64_t));
 1591 
 1592         if (error || !req->newptr)
 1593                 return (error);
 1594 
 1595         if (!arg1)
 1596                 error = EPERM;
 1597         else
 1598                 error = SYSCTL_IN(req, arg1, sizeof(uint64_t));
 1599         return (error);
 1600 }
 1601 
 1602 /*
 1603  * Handle our generic '\0' terminated 'C' string.
 1604  * Two cases:
 1605  *      a variable string:  point arg1 at it, arg2 is max length.
 1606  *      a constant string:  point arg1 at it, arg2 is zero.
 1607  */
 1608 
 1609 int
 1610 sysctl_handle_string(SYSCTL_HANDLER_ARGS)
 1611 {
 1612         size_t outlen;
 1613         int error = 0, ro_string = 0;
 1614 
 1615         /*
 1616          * A zero-length buffer indicates a fixed size read-only
 1617          * string:
 1618          */
 1619         if (arg2 == 0) {
 1620                 arg2 = strlen((char *)arg1) + 1;
 1621                 ro_string = 1;
 1622         }
 1623 
 1624         if (req->oldptr != NULL) {
 1625                 char *tmparg;
 1626 
 1627                 if (ro_string) {
 1628                         tmparg = arg1;
 1629                 } else {
 1630                         /* try to make a coherent snapshot of the string */
 1631                         tmparg = malloc(arg2, M_SYSCTLTMP, M_WAITOK);
 1632                         memcpy(tmparg, arg1, arg2);
 1633                 }
 1634 
 1635                 outlen = strnlen(tmparg, arg2 - 1) + 1;
 1636                 error = SYSCTL_OUT(req, tmparg, outlen);
 1637 
 1638                 if (!ro_string)
 1639                         free(tmparg, M_SYSCTLTMP);
 1640         } else {
 1641                 outlen = strnlen((char *)arg1, arg2 - 1) + 1;
 1642                 error = SYSCTL_OUT(req, NULL, outlen);
 1643         }
 1644         if (error || !req->newptr)
 1645                 return (error);
 1646 
 1647         if ((req->newlen - req->newidx) >= arg2) {
 1648                 error = EINVAL;
 1649         } else {
 1650                 arg2 = (req->newlen - req->newidx);
 1651                 error = SYSCTL_IN(req, arg1, arg2);
 1652                 ((char *)arg1)[arg2] = '\0';
 1653         }
 1654         return (error);
 1655 }
 1656 
 1657 /*
 1658  * Handle any kind of opaque data.
 1659  * arg1 points to it, arg2 is the size.
 1660  */
 1661 
 1662 int
 1663 sysctl_handle_opaque(SYSCTL_HANDLER_ARGS)
 1664 {
 1665         int error, tries;
 1666         u_int generation;
 1667         struct sysctl_req req2;
 1668 
 1669         /*
 1670          * Attempt to get a coherent snapshot, by using the thread
 1671          * pre-emption counter updated from within mi_switch() to
 1672          * determine if we were pre-empted during a bcopy() or
 1673          * copyout(). Make 3 attempts at doing this before giving up.
 1674          * If we encounter an error, stop immediately.
 1675          */
 1676         tries = 0;
 1677         req2 = *req;
 1678 retry:
 1679         generation = curthread->td_generation;
 1680         error = SYSCTL_OUT(req, arg1, arg2);
 1681         if (error)
 1682                 return (error);
 1683         tries++;
 1684         if (generation != curthread->td_generation && tries < 3) {
 1685                 *req = req2;
 1686                 goto retry;
 1687         }
 1688 
 1689         error = SYSCTL_IN(req, arg1, arg2);
 1690 
 1691         return (error);
 1692 }
 1693 
 1694 /*
 1695  * Transfer functions to/from kernel space.
 1696  * XXX: rather untested at this point
 1697  */
 1698 static int
 1699 sysctl_old_kernel(struct sysctl_req *req, const void *p, size_t l)
 1700 {
 1701         size_t i = 0;
 1702 
 1703         if (req->oldptr) {
 1704                 i = l;
 1705                 if (req->oldlen <= req->oldidx)
 1706                         i = 0;
 1707                 else
 1708                         if (i > req->oldlen - req->oldidx)
 1709                                 i = req->oldlen - req->oldidx;
 1710                 if (i > 0)
 1711                         bcopy(p, (char *)req->oldptr + req->oldidx, i);
 1712         }
 1713         req->oldidx += l;
 1714         if (req->oldptr && i != l)
 1715                 return (ENOMEM);
 1716         return (0);
 1717 }
 1718 
 1719 static int
 1720 sysctl_new_kernel(struct sysctl_req *req, void *p, size_t l)
 1721 {
 1722         if (!req->newptr)
 1723                 return (0);
 1724         if (req->newlen - req->newidx < l)
 1725                 return (EINVAL);
 1726         bcopy((char *)req->newptr + req->newidx, p, l);
 1727         req->newidx += l;
 1728         return (0);
 1729 }
 1730 
 1731 int
 1732 kernel_sysctl(struct thread *td, int *name, u_int namelen, void *old,
 1733     size_t *oldlenp, void *new, size_t newlen, size_t *retval, int flags)
 1734 {
 1735         int error = 0;
 1736         struct sysctl_req req;
 1737 
 1738         bzero(&req, sizeof req);
 1739 
 1740         req.td = td;
 1741         req.flags = flags;
 1742 
 1743         if (oldlenp) {
 1744                 req.oldlen = *oldlenp;
 1745         }
 1746         req.validlen = req.oldlen;
 1747 
 1748         if (old) {
 1749                 req.oldptr= old;
 1750         }
 1751 
 1752         if (new != NULL) {
 1753                 req.newlen = newlen;
 1754                 req.newptr = new;
 1755         }
 1756 
 1757         req.oldfunc = sysctl_old_kernel;
 1758         req.newfunc = sysctl_new_kernel;
 1759         req.lock = REQ_UNWIRED;
 1760 
 1761         error = sysctl_root(0, name, namelen, &req);
 1762 
 1763         if (req.lock == REQ_WIRED && req.validlen > 0)
 1764                 vsunlock(req.oldptr, req.validlen);
 1765 
 1766         if (error && error != ENOMEM)
 1767                 return (error);
 1768 
 1769         if (retval) {
 1770                 if (req.oldptr && req.oldidx > req.validlen)
 1771                         *retval = req.validlen;
 1772                 else
 1773                         *retval = req.oldidx;
 1774         }
 1775         return (error);
 1776 }
 1777 
 1778 int
 1779 kernel_sysctlbyname(struct thread *td, char *name, void *old, size_t *oldlenp,
 1780     void *new, size_t newlen, size_t *retval, int flags)
 1781 {
 1782         int oid[CTL_MAXNAME];
 1783         size_t oidlen, plen;
 1784         int error;
 1785 
 1786         oid[0] = 0;             /* sysctl internal magic */
 1787         oid[1] = 3;             /* name2oid */
 1788         oidlen = sizeof(oid);
 1789 
 1790         error = kernel_sysctl(td, oid, 2, oid, &oidlen,
 1791             (void *)name, strlen(name), &plen, flags);
 1792         if (error)
 1793                 return (error);
 1794 
 1795         error = kernel_sysctl(td, oid, plen / sizeof(int), old, oldlenp,
 1796             new, newlen, retval, flags);
 1797         return (error);
 1798 }
 1799 
 1800 /*
 1801  * Transfer function to/from user space.
 1802  */
 1803 static int
 1804 sysctl_old_user(struct sysctl_req *req, const void *p, size_t l)
 1805 {
 1806         size_t i, len, origidx;
 1807         int error;
 1808 
 1809         origidx = req->oldidx;
 1810         req->oldidx += l;
 1811         if (req->oldptr == NULL)
 1812                 return (0);
 1813         /*
 1814          * If we have not wired the user supplied buffer and we are currently
 1815          * holding locks, drop a witness warning, as it's possible that
 1816          * write operations to the user page can sleep.
 1817          */
 1818         if (req->lock != REQ_WIRED)
 1819                 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,
 1820                     "sysctl_old_user()");
 1821         i = l;
 1822         len = req->validlen;
 1823         if (len <= origidx)
 1824                 i = 0;
 1825         else {
 1826                 if (i > len - origidx)
 1827                         i = len - origidx;
 1828                 if (req->lock == REQ_WIRED) {
 1829                         error = copyout_nofault(p, (char *)req->oldptr +
 1830                             origidx, i);
 1831                 } else
 1832                         error = copyout(p, (char *)req->oldptr + origidx, i);
 1833                 if (error != 0)
 1834                         return (error);
 1835         }
 1836         if (i < l)
 1837                 return (ENOMEM);
 1838         return (0);
 1839 }
 1840 
 1841 static int
 1842 sysctl_new_user(struct sysctl_req *req, void *p, size_t l)
 1843 {
 1844         int error;
 1845 
 1846         if (!req->newptr)
 1847                 return (0);
 1848         if (req->newlen - req->newidx < l)
 1849                 return (EINVAL);
 1850         WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,
 1851             "sysctl_new_user()");
 1852         error = copyin((char *)req->newptr + req->newidx, p, l);
 1853         req->newidx += l;
 1854         return (error);
 1855 }
 1856 
 1857 /*
 1858  * Wire the user space destination buffer.  If set to a value greater than
 1859  * zero, the len parameter limits the maximum amount of wired memory.
 1860  */
 1861 int
 1862 sysctl_wire_old_buffer(struct sysctl_req *req, size_t len)
 1863 {
 1864         int ret;
 1865         size_t wiredlen;
 1866 
 1867         wiredlen = (len > 0 && len < req->oldlen) ? len : req->oldlen;
 1868         ret = 0;
 1869         if (req->lock != REQ_WIRED && req->oldptr &&
 1870             req->oldfunc == sysctl_old_user) {
 1871                 if (wiredlen != 0) {
 1872                         ret = vslock(req->oldptr, wiredlen);
 1873                         if (ret != 0) {
 1874                                 if (ret != ENOMEM)
 1875                                         return (ret);
 1876                                 wiredlen = 0;
 1877                         }
 1878                 }
 1879                 req->lock = REQ_WIRED;
 1880                 req->validlen = wiredlen;
 1881         }
 1882         return (0);
 1883 }
 1884 
 1885 int
 1886 sysctl_find_oid(int *name, u_int namelen, struct sysctl_oid **noid,
 1887     int *nindx, struct sysctl_req *req)
 1888 {
 1889         struct sysctl_oid_list *lsp;
 1890         struct sysctl_oid *oid;
 1891         int indx;
 1892 
 1893         SYSCTL_ASSERT_LOCKED();
 1894         lsp = &sysctl__children;
 1895         indx = 0;
 1896         while (indx < CTL_MAXNAME) {
 1897                 SLIST_FOREACH(oid, lsp, oid_link) {
 1898                         if (oid->oid_number == name[indx])
 1899                                 break;
 1900                 }
 1901                 if (oid == NULL)
 1902                         return (ENOENT);
 1903 
 1904                 indx++;
 1905                 if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
 1906                         if (oid->oid_handler != NULL || indx == namelen) {
 1907                                 *noid = oid;
 1908                                 if (nindx != NULL)
 1909                                         *nindx = indx;
 1910                                 KASSERT((oid->oid_kind & CTLFLAG_DYING) == 0,
 1911                                     ("%s found DYING node %p", __func__, oid));
 1912                                 return (0);
 1913                         }
 1914                         lsp = SYSCTL_CHILDREN(oid);
 1915                 } else if (indx == namelen) {
 1916                         if ((oid->oid_kind & CTLFLAG_DORMANT) != 0)
 1917                                 return (ENOENT);
 1918                         *noid = oid;
 1919                         if (nindx != NULL)
 1920                                 *nindx = indx;
 1921                         KASSERT((oid->oid_kind & CTLFLAG_DYING) == 0,
 1922                             ("%s found DYING node %p", __func__, oid));
 1923                         return (0);
 1924                 } else {
 1925                         return (ENOTDIR);
 1926                 }
 1927         }
 1928         return (ENOENT);
 1929 }
 1930 
 1931 /*
 1932  * Traverse our tree, and find the right node, execute whatever it points
 1933  * to, and return the resulting error code.
 1934  */
 1935 
 1936 static int
 1937 sysctl_root(SYSCTL_HANDLER_ARGS)
 1938 {
 1939         struct sysctl_oid *oid;
 1940         struct rm_priotracker tracker;
 1941         int error, indx, lvl;
 1942 
 1943         SYSCTL_RLOCK(&tracker);
 1944 
 1945         error = sysctl_find_oid(arg1, arg2, &oid, &indx, req);
 1946         if (error)
 1947                 goto out;
 1948 
 1949         if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
 1950                 /*
 1951                  * You can't call a sysctl when it's a node, but has
 1952                  * no handler.  Inform the user that it's a node.
 1953                  * The indx may or may not be the same as namelen.
 1954                  */
 1955                 if (oid->oid_handler == NULL) {
 1956                         error = EISDIR;
 1957                         goto out;
 1958                 }
 1959         }
 1960 
 1961         /* Is this sysctl writable? */
 1962         if (req->newptr && !(oid->oid_kind & CTLFLAG_WR)) {
 1963                 error = EPERM;
 1964                 goto out;
 1965         }
 1966 
 1967         KASSERT(req->td != NULL, ("sysctl_root(): req->td == NULL"));
 1968 
 1969 #ifdef CAPABILITY_MODE
 1970         /*
 1971          * If the process is in capability mode, then don't permit reading or
 1972          * writing unless specifically granted for the node.
 1973          */
 1974         if (IN_CAPABILITY_MODE(req->td)) {
 1975                 if ((req->oldptr && !(oid->oid_kind & CTLFLAG_CAPRD)) ||
 1976                     (req->newptr && !(oid->oid_kind & CTLFLAG_CAPWR))) {
 1977                         error = EPERM;
 1978                         goto out;
 1979                 }
 1980         }
 1981 #endif
 1982 
 1983         /* Is this sysctl sensitive to securelevels? */
 1984         if (req->newptr && (oid->oid_kind & CTLFLAG_SECURE)) {
 1985                 lvl = (oid->oid_kind & CTLMASK_SECURE) >> CTLSHIFT_SECURE;
 1986                 error = securelevel_gt(req->td->td_ucred, lvl);
 1987                 if (error)
 1988                         goto out;
 1989         }
 1990 
 1991         /* Is this sysctl writable by only privileged users? */
 1992         if (req->newptr && !(oid->oid_kind & CTLFLAG_ANYBODY)) {
 1993                 int priv;
 1994 
 1995                 if (oid->oid_kind & CTLFLAG_PRISON)
 1996                         priv = PRIV_SYSCTL_WRITEJAIL;
 1997 #ifdef VIMAGE
 1998                 else if ((oid->oid_kind & CTLFLAG_VNET) &&
 1999                      prison_owns_vnet(req->td->td_ucred))
 2000                         priv = PRIV_SYSCTL_WRITEJAIL;
 2001 #endif
 2002                 else
 2003                         priv = PRIV_SYSCTL_WRITE;
 2004                 error = priv_check(req->td, priv);
 2005                 if (error)
 2006                         goto out;
 2007         }
 2008 
 2009         if (!oid->oid_handler) {
 2010                 error = EINVAL;
 2011                 goto out;
 2012         }
 2013 
 2014         if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) {
 2015                 arg1 = (int *)arg1 + indx;
 2016                 arg2 -= indx;
 2017         } else {
 2018                 arg1 = oid->oid_arg1;
 2019                 arg2 = oid->oid_arg2;
 2020         }
 2021 #ifdef MAC
 2022         error = mac_system_check_sysctl(req->td->td_ucred, oid, arg1, arg2,
 2023             req);
 2024         if (error != 0)
 2025                 goto out;
 2026 #endif
 2027 #ifdef VIMAGE
 2028         if ((oid->oid_kind & CTLFLAG_VNET) && arg1 != NULL)
 2029                 arg1 = (void *)(curvnet->vnet_data_base + (uintptr_t)arg1);
 2030 #endif
 2031         error = sysctl_root_handler_locked(oid, arg1, arg2, req, &tracker);
 2032 
 2033 out:
 2034         SYSCTL_RUNLOCK(&tracker);
 2035         return (error);
 2036 }
 2037 
 2038 #ifndef _SYS_SYSPROTO_H_
 2039 struct sysctl_args {
 2040         int     *name;
 2041         u_int   namelen;
 2042         void    *old;
 2043         size_t  *oldlenp;
 2044         void    *new;
 2045         size_t  newlen;
 2046 };
 2047 #endif
 2048 int
 2049 sys___sysctl(struct thread *td, struct sysctl_args *uap)
 2050 {
 2051         int error, i, name[CTL_MAXNAME];
 2052         size_t j;
 2053 
 2054         if (uap->namelen > CTL_MAXNAME || uap->namelen < 2)
 2055                 return (EINVAL);
 2056 
 2057         error = copyin(uap->name, &name, uap->namelen * sizeof(int));
 2058         if (error)
 2059                 return (error);
 2060 
 2061         error = userland_sysctl(td, name, uap->namelen,
 2062                 uap->old, uap->oldlenp, 0,
 2063                 uap->new, uap->newlen, &j, 0);
 2064         if (error && error != ENOMEM)
 2065                 return (error);
 2066         if (uap->oldlenp) {
 2067                 i = copyout(&j, uap->oldlenp, sizeof(j));
 2068                 if (i)
 2069                         return (i);
 2070         }
 2071         return (error);
 2072 }
 2073 
 2074 /*
 2075  * This is used from various compatibility syscalls too.  That's why name
 2076  * must be in kernel space.
 2077  */
 2078 int
 2079 userland_sysctl(struct thread *td, int *name, u_int namelen, void *old,
 2080     size_t *oldlenp, int inkernel, void *new, size_t newlen, size_t *retval,
 2081     int flags)
 2082 {
 2083         int error = 0, memlocked;
 2084         struct sysctl_req req;
 2085 
 2086         bzero(&req, sizeof req);
 2087 
 2088         req.td = td;
 2089         req.flags = flags;
 2090 
 2091         if (oldlenp) {
 2092                 if (inkernel) {
 2093                         req.oldlen = *oldlenp;
 2094                 } else {
 2095                         error = copyin(oldlenp, &req.oldlen, sizeof(*oldlenp));
 2096                         if (error)
 2097                                 return (error);
 2098                 }
 2099         }
 2100         req.validlen = req.oldlen;
 2101         req.oldptr = old;
 2102 
 2103         if (new != NULL) {
 2104                 req.newlen = newlen;
 2105                 req.newptr = new;
 2106         }
 2107 
 2108         req.oldfunc = sysctl_old_user;
 2109         req.newfunc = sysctl_new_user;
 2110         req.lock = REQ_UNWIRED;
 2111 
 2112 #ifdef KTRACE
 2113         if (KTRPOINT(curthread, KTR_SYSCTL))
 2114                 ktrsysctl(name, namelen);
 2115 #endif
 2116         memlocked = 0;
 2117         if (req.oldptr && req.oldlen > 4 * PAGE_SIZE) {
 2118                 memlocked = 1;
 2119                 sx_xlock(&sysctlmemlock);
 2120         }
 2121         CURVNET_SET(TD_TO_VNET(td));
 2122 
 2123         for (;;) {
 2124                 req.oldidx = 0;
 2125                 req.newidx = 0;
 2126                 error = sysctl_root(0, name, namelen, &req);
 2127                 if (error != EAGAIN)
 2128                         break;
 2129                 kern_yield(PRI_USER);
 2130         }
 2131 
 2132         CURVNET_RESTORE();
 2133 
 2134         if (req.lock == REQ_WIRED && req.validlen > 0)
 2135                 vsunlock(req.oldptr, req.validlen);
 2136         if (memlocked)
 2137                 sx_xunlock(&sysctlmemlock);
 2138 
 2139         if (error && error != ENOMEM)
 2140                 return (error);
 2141 
 2142         if (retval) {
 2143                 if (req.oldptr && req.oldidx > req.validlen)
 2144                         *retval = req.validlen;
 2145                 else
 2146                         *retval = req.oldidx;
 2147         }
 2148         return (error);
 2149 }
 2150 
 2151 /*
 2152  * Drain into a sysctl struct.  The user buffer should be wired if a page
 2153  * fault would cause issue.
 2154  */
 2155 static int
 2156 sbuf_sysctl_drain(void *arg, const char *data, int len)
 2157 {
 2158         struct sysctl_req *req = arg;
 2159         int error;
 2160 
 2161         error = SYSCTL_OUT(req, data, len);
 2162         KASSERT(error >= 0, ("Got unexpected negative value %d", error));
 2163         return (error == 0 ? len : -error);
 2164 }
 2165 
 2166 struct sbuf *
 2167 sbuf_new_for_sysctl(struct sbuf *s, char *buf, int length,
 2168     struct sysctl_req *req)
 2169 {
 2170 
 2171         /* Supply a default buffer size if none given. */
 2172         if (buf == NULL && length == 0)
 2173                 length = 64;
 2174         s = sbuf_new(s, buf, length, SBUF_FIXEDLEN | SBUF_INCLUDENUL);
 2175         sbuf_set_drain(s, sbuf_sysctl_drain, req);
 2176         return (s);
 2177 }
Cache object: 30c256db8765e9b11b93880cd7961ff8
[ 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
