kern_sysctl.c revision 106025
1/*- 2 * Copyright (c) 1982, 1986, 1989, 1993 3 * The Regents of the University of California. All rights reserved. 4 * 5 * This code is derived from software contributed to Berkeley by 6 * Mike Karels at Berkeley Software Design, Inc. 7 * 8 * Quite extensively rewritten by Poul-Henning Kamp of the FreeBSD 9 * project, to make these variables more userfriendly. 10 * 11 * Redistribution and use in source and binary forms, with or without 12 * modification, are permitted provided that the following conditions 13 * are met: 14 * 1. Redistributions of source code must retain the above copyright 15 * notice, this list of conditions and the following disclaimer. 16 * 2. Redistributions in binary form must reproduce the above copyright 17 * notice, this list of conditions and the following disclaimer in the 18 * documentation and/or other materials provided with the distribution. 19 * 3. All advertising materials mentioning features or use of this software 20 * must display the following acknowledgement: 21 * This product includes software developed by the University of 22 * California, Berkeley and its contributors. 23 * 4. Neither the name of the University nor the names of its contributors 24 * may be used to endorse or promote products derived from this software 25 * without specific prior written permission. 26 * 27 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 28 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 29 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 30 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 31 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 32 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 33 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 34 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 35 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 36 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 37 * SUCH DAMAGE. 38 * 39 * @(#)kern_sysctl.c 8.4 (Berkeley) 4/14/94 40 * $FreeBSD: head/sys/kern/kern_sysctl.c 106025 2002-10-27 07:12:34Z rwatson $ 41 */ 42 43#include "opt_compat.h" 44#include "opt_mac.h" 45 46#include <sys/param.h> 47#include <sys/systm.h> 48#include <sys/kernel.h> 49#include <sys/sysctl.h> 50#include <sys/mac.h> 51#include <sys/malloc.h> 52#include <sys/proc.h> 53#include <sys/lock.h> 54#include <sys/mutex.h> 55#include <sys/sx.h> 56#include <sys/sysproto.h> 57#include <vm/vm.h> 58#include <vm/vm_extern.h> 59 60static MALLOC_DEFINE(M_SYSCTL, "sysctl", "sysctl internal magic"); 61static MALLOC_DEFINE(M_SYSCTLOID, "sysctloid", "sysctl dynamic oids"); 62static MALLOC_DEFINE(M_SYSCTLTMP, "sysctltmp", "sysctl temp output buffer"); 63 64/* 65 * Locking - this locks the sysctl tree in memory. 66 */ 67static struct sx sysctllock; 68 69#define SYSCTL_LOCK() sx_xlock(&sysctllock) 70#define SYSCTL_UNLOCK() sx_xunlock(&sysctllock) 71#define SYSCTL_INIT() sx_init(&sysctllock, "sysctl sysctllock") 72 73static int sysctl_root(SYSCTL_HANDLER_ARGS); 74 75struct sysctl_oid_list sysctl__children; /* root list */ 76 77static struct sysctl_oid * 78sysctl_find_oidname(const char *name, struct sysctl_oid_list *list) 79{ 80 struct sysctl_oid *oidp; 81 82 SLIST_FOREACH(oidp, list, oid_link) { 83 if (strcmp(oidp->oid_name, name) == 0) { 84 return (oidp); 85 } 86 } 87 return (NULL); 88} 89 90/* 91 * Initialization of the MIB tree. 92 * 93 * Order by number in each list. 94 */ 95 96void 97sysctl_register_oid(struct sysctl_oid *oidp) 98{ 99 struct sysctl_oid_list *parent = oidp->oid_parent; 100 struct sysctl_oid *p; 101 struct sysctl_oid *q; 102 103 /* 104 * First check if another oid with the same name already 105 * exists in the parent's list. 106 */ 107 p = sysctl_find_oidname(oidp->oid_name, parent); 108 if (p != NULL) { 109 if ((p->oid_kind & CTLTYPE) == CTLTYPE_NODE) { 110 p->oid_refcnt++; 111 return; 112 } else { 113 printf("can't re-use a leaf (%s)!\n", p->oid_name); 114 return; 115 } 116 } 117 /* 118 * If this oid has a number OID_AUTO, give it a number which 119 * is greater than any current oid. 120 * NOTE: DO NOT change the starting value here, change it in 121 * <sys/sysctl.h>, and make sure it is at least 256 to 122 * accomodate e.g. net.inet.raw as a static sysctl node. 123 */ 124 if (oidp->oid_number == OID_AUTO) { 125 static int newoid = CTL_AUTO_START; 126 127 oidp->oid_number = newoid++; 128 if (newoid == 0x7fffffff) 129 panic("out of oids"); 130 } 131#if 0 132 else if (oidp->oid_number >= CTL_AUTO_START) { 133 /* do not panic; this happens when unregistering sysctl sets */ 134 printf("static sysctl oid too high: %d", oidp->oid_number); 135 } 136#endif 137 138 /* 139 * Insert the oid into the parent's list in order. 140 */ 141 q = NULL; 142 SLIST_FOREACH(p, parent, oid_link) { 143 if (oidp->oid_number < p->oid_number) 144 break; 145 q = p; 146 } 147 if (q) 148 SLIST_INSERT_AFTER(q, oidp, oid_link); 149 else 150 SLIST_INSERT_HEAD(parent, oidp, oid_link); 151} 152 153void 154sysctl_unregister_oid(struct sysctl_oid *oidp) 155{ 156 SLIST_REMOVE(oidp->oid_parent, oidp, sysctl_oid, oid_link); 157} 158 159/* Initialize a new context to keep track of dynamically added sysctls. */ 160int 161sysctl_ctx_init(struct sysctl_ctx_list *c) 162{ 163 164 if (c == NULL) { 165 return (EINVAL); 166 } 167 TAILQ_INIT(c); 168 return (0); 169} 170 171/* Free the context, and destroy all dynamic oids registered in this context */ 172int 173sysctl_ctx_free(struct sysctl_ctx_list *clist) 174{ 175 struct sysctl_ctx_entry *e, *e1; 176 int error; 177 178 error = 0; 179 /* 180 * First perform a "dry run" to check if it's ok to remove oids. 181 * XXX FIXME 182 * XXX This algorithm is a hack. But I don't know any 183 * XXX better solution for now... 184 */ 185 TAILQ_FOREACH(e, clist, link) { 186 error = sysctl_remove_oid(e->entry, 0, 0); 187 if (error) 188 break; 189 } 190 /* 191 * Restore deregistered entries, either from the end, 192 * or from the place where error occured. 193 * e contains the entry that was not unregistered 194 */ 195 if (error) 196 e1 = TAILQ_PREV(e, sysctl_ctx_list, link); 197 else 198 e1 = TAILQ_LAST(clist, sysctl_ctx_list); 199 while (e1 != NULL) { 200 sysctl_register_oid(e1->entry); 201 e1 = TAILQ_PREV(e1, sysctl_ctx_list, link); 202 } 203 if (error) 204 return(EBUSY); 205 /* Now really delete the entries */ 206 e = TAILQ_FIRST(clist); 207 while (e != NULL) { 208 e1 = TAILQ_NEXT(e, link); 209 error = sysctl_remove_oid(e->entry, 1, 0); 210 if (error) 211 panic("sysctl_remove_oid: corrupt tree, entry: %s", 212 e->entry->oid_name); 213 free(e, M_SYSCTLOID); 214 e = e1; 215 } 216 return (error); 217} 218 219/* Add an entry to the context */ 220struct sysctl_ctx_entry * 221sysctl_ctx_entry_add(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp) 222{ 223 struct sysctl_ctx_entry *e; 224 225 if (clist == NULL || oidp == NULL) 226 return(NULL); 227 e = malloc(sizeof(struct sysctl_ctx_entry), M_SYSCTLOID, M_WAITOK); 228 e->entry = oidp; 229 TAILQ_INSERT_HEAD(clist, e, link); 230 return (e); 231} 232 233/* Find an entry in the context */ 234struct sysctl_ctx_entry * 235sysctl_ctx_entry_find(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp) 236{ 237 struct sysctl_ctx_entry *e; 238 239 if (clist == NULL || oidp == NULL) 240 return(NULL); 241 TAILQ_FOREACH(e, clist, link) { 242 if(e->entry == oidp) 243 return(e); 244 } 245 return (e); 246} 247 248/* 249 * Delete an entry from the context. 250 * NOTE: this function doesn't free oidp! You have to remove it 251 * with sysctl_remove_oid(). 252 */ 253int 254sysctl_ctx_entry_del(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp) 255{ 256 struct sysctl_ctx_entry *e; 257 258 if (clist == NULL || oidp == NULL) 259 return (EINVAL); 260 e = sysctl_ctx_entry_find(clist, oidp); 261 if (e != NULL) { 262 TAILQ_REMOVE(clist, e, link); 263 free(e, M_SYSCTLOID); 264 return (0); 265 } else 266 return (ENOENT); 267} 268 269/* 270 * Remove dynamically created sysctl trees. 271 * oidp - top of the tree to be removed 272 * del - if 0 - just deregister, otherwise free up entries as well 273 * recurse - if != 0 traverse the subtree to be deleted 274 */ 275int 276sysctl_remove_oid(struct sysctl_oid *oidp, int del, int recurse) 277{ 278 struct sysctl_oid *p; 279 int error; 280 281 if (oidp == NULL) 282 return(EINVAL); 283 if ((oidp->oid_kind & CTLFLAG_DYN) == 0) { 284 printf("can't remove non-dynamic nodes!\n"); 285 return (EINVAL); 286 } 287 /* 288 * WARNING: normal method to do this should be through 289 * sysctl_ctx_free(). Use recursing as the last resort 290 * method to purge your sysctl tree of leftovers... 291 * However, if some other code still references these nodes, 292 * it will panic. 293 */ 294 if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) { 295 if (oidp->oid_refcnt == 1) { 296 SLIST_FOREACH(p, SYSCTL_CHILDREN(oidp), oid_link) { 297 if (!recurse) 298 return (ENOTEMPTY); 299 error = sysctl_remove_oid(p, del, recurse); 300 if (error) 301 return (error); 302 } 303 if (del) 304 free(SYSCTL_CHILDREN(oidp), M_SYSCTLOID); 305 } 306 } 307 if (oidp->oid_refcnt > 1 ) { 308 oidp->oid_refcnt--; 309 } else { 310 if (oidp->oid_refcnt == 0) { 311 printf("Warning: bad oid_refcnt=%u (%s)!\n", 312 oidp->oid_refcnt, oidp->oid_name); 313 return (EINVAL); 314 } 315 sysctl_unregister_oid(oidp); 316 if (del) { 317 if (oidp->descr) 318 free((void *)(uintptr_t)(const void *)oidp->descr, M_SYSCTLOID); 319 free((void *)(uintptr_t)(const void *)oidp->oid_name, 320 M_SYSCTLOID); 321 free(oidp, M_SYSCTLOID); 322 } 323 } 324 return (0); 325} 326 327/* 328 * Create new sysctls at run time. 329 * clist may point to a valid context initialized with sysctl_ctx_init(). 330 */ 331struct sysctl_oid * 332sysctl_add_oid(struct sysctl_ctx_list *clist, struct sysctl_oid_list *parent, 333 int number, const char *name, int kind, void *arg1, int arg2, 334 int (*handler)(SYSCTL_HANDLER_ARGS), const char *fmt, const char *descr) 335{ 336 struct sysctl_oid *oidp; 337 ssize_t len; 338 char *newname; 339 340 /* You have to hook up somewhere.. */ 341 if (parent == NULL) 342 return(NULL); 343 /* Check if the node already exists, otherwise create it */ 344 oidp = sysctl_find_oidname(name, parent); 345 if (oidp != NULL) { 346 if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) { 347 oidp->oid_refcnt++; 348 /* Update the context */ 349 if (clist != NULL) 350 sysctl_ctx_entry_add(clist, oidp); 351 return (oidp); 352 } else { 353 printf("can't re-use a leaf (%s)!\n", name); 354 return (NULL); 355 } 356 } 357 oidp = malloc(sizeof(struct sysctl_oid), M_SYSCTLOID, M_WAITOK|M_ZERO); 358 oidp->oid_parent = parent; 359 SLIST_NEXT(oidp, oid_link) = NULL; 360 oidp->oid_number = number; 361 oidp->oid_refcnt = 1; 362 len = strlen(name); 363 newname = malloc(len + 1, M_SYSCTLOID, M_WAITOK); 364 bcopy(name, newname, len + 1); 365 newname[len] = '\0'; 366 oidp->oid_name = newname; 367 oidp->oid_handler = handler; 368 oidp->oid_kind = CTLFLAG_DYN | kind; 369 if ((kind & CTLTYPE) == CTLTYPE_NODE) { 370 /* Allocate space for children */ 371 SYSCTL_CHILDREN(oidp) = malloc(sizeof(struct sysctl_oid_list), 372 M_SYSCTLOID, M_WAITOK); 373 SLIST_INIT(SYSCTL_CHILDREN(oidp)); 374 } else { 375 oidp->oid_arg1 = arg1; 376 oidp->oid_arg2 = arg2; 377 } 378 oidp->oid_fmt = fmt; 379 if (descr) { 380 int len = strlen(descr) + 1; 381 oidp->descr = malloc(len, M_SYSCTLOID, M_WAITOK); 382 if (oidp->descr) 383 strcpy((char *)(uintptr_t)(const void *)oidp->descr, descr); 384 } 385 /* Update the context, if used */ 386 if (clist != NULL) 387 sysctl_ctx_entry_add(clist, oidp); 388 /* Register this oid */ 389 sysctl_register_oid(oidp); 390 return (oidp); 391} 392 393/* 394 * Register the kernel's oids on startup. 395 */ 396SET_DECLARE(sysctl_set, struct sysctl_oid); 397 398static void 399sysctl_register_all(void *arg) 400{ 401 struct sysctl_oid **oidp; 402 403 SYSCTL_INIT(); 404 SET_FOREACH(oidp, sysctl_set) 405 sysctl_register_oid(*oidp); 406} 407SYSINIT(sysctl, SI_SUB_KMEM, SI_ORDER_ANY, sysctl_register_all, 0); 408 409/* 410 * "Staff-functions" 411 * 412 * These functions implement a presently undocumented interface 413 * used by the sysctl program to walk the tree, and get the type 414 * so it can print the value. 415 * This interface is under work and consideration, and should probably 416 * be killed with a big axe by the first person who can find the time. 417 * (be aware though, that the proper interface isn't as obvious as it 418 * may seem, there are various conflicting requirements. 419 * 420 * {0,0} printf the entire MIB-tree. 421 * {0,1,...} return the name of the "..." OID. 422 * {0,2,...} return the next OID. 423 * {0,3} return the OID of the name in "new" 424 * {0,4,...} return the kind & format info for the "..." OID. 425 * {0,5,...} return the description the "..." OID. 426 */ 427 428static void 429sysctl_sysctl_debug_dump_node(struct sysctl_oid_list *l, int i) 430{ 431 int k; 432 struct sysctl_oid *oidp; 433 434 SLIST_FOREACH(oidp, l, oid_link) { 435 436 for (k=0; k<i; k++) 437 printf(" "); 438 439 printf("%d %s ", oidp->oid_number, oidp->oid_name); 440 441 printf("%c%c", 442 oidp->oid_kind & CTLFLAG_RD ? 'R':' ', 443 oidp->oid_kind & CTLFLAG_WR ? 'W':' '); 444 445 if (oidp->oid_handler) 446 printf(" *Handler"); 447 448 switch (oidp->oid_kind & CTLTYPE) { 449 case CTLTYPE_NODE: 450 printf(" Node\n"); 451 if (!oidp->oid_handler) { 452 sysctl_sysctl_debug_dump_node( 453 oidp->oid_arg1, i+2); 454 } 455 break; 456 case CTLTYPE_INT: printf(" Int\n"); break; 457 case CTLTYPE_STRING: printf(" String\n"); break; 458 case CTLTYPE_QUAD: printf(" Quad\n"); break; 459 case CTLTYPE_OPAQUE: printf(" Opaque/struct\n"); break; 460 default: printf("\n"); 461 } 462 463 } 464} 465 466static int 467sysctl_sysctl_debug(SYSCTL_HANDLER_ARGS) 468{ 469 int error; 470 471 error = suser(req->td); 472 if (error) 473 return error; 474 sysctl_sysctl_debug_dump_node(&sysctl__children, 0); 475 return ENOENT; 476} 477 478SYSCTL_PROC(_sysctl, 0, debug, CTLTYPE_STRING|CTLFLAG_RD, 479 0, 0, sysctl_sysctl_debug, "-", ""); 480 481static int 482sysctl_sysctl_name(SYSCTL_HANDLER_ARGS) 483{ 484 int *name = (int *) arg1; 485 u_int namelen = arg2; 486 int error = 0; 487 struct sysctl_oid *oid; 488 struct sysctl_oid_list *lsp = &sysctl__children, *lsp2; 489 char buf[10]; 490 491 while (namelen) { 492 if (!lsp) { 493 snprintf(buf,sizeof(buf),"%d",*name); 494 if (req->oldidx) 495 error = SYSCTL_OUT(req, ".", 1); 496 if (!error) 497 error = SYSCTL_OUT(req, buf, strlen(buf)); 498 if (error) 499 return (error); 500 namelen--; 501 name++; 502 continue; 503 } 504 lsp2 = 0; 505 SLIST_FOREACH(oid, lsp, oid_link) { 506 if (oid->oid_number != *name) 507 continue; 508 509 if (req->oldidx) 510 error = SYSCTL_OUT(req, ".", 1); 511 if (!error) 512 error = SYSCTL_OUT(req, oid->oid_name, 513 strlen(oid->oid_name)); 514 if (error) 515 return (error); 516 517 namelen--; 518 name++; 519 520 if ((oid->oid_kind & CTLTYPE) != CTLTYPE_NODE) 521 break; 522 523 if (oid->oid_handler) 524 break; 525 526 lsp2 = (struct sysctl_oid_list *)oid->oid_arg1; 527 break; 528 } 529 lsp = lsp2; 530 } 531 return (SYSCTL_OUT(req, "", 1)); 532} 533 534SYSCTL_NODE(_sysctl, 1, name, CTLFLAG_RD, sysctl_sysctl_name, ""); 535 536static int 537sysctl_sysctl_next_ls(struct sysctl_oid_list *lsp, int *name, u_int namelen, 538 int *next, int *len, int level, struct sysctl_oid **oidpp) 539{ 540 struct sysctl_oid *oidp; 541 542 *len = level; 543 SLIST_FOREACH(oidp, lsp, oid_link) { 544 *next = oidp->oid_number; 545 *oidpp = oidp; 546 547 if (oidp->oid_kind & CTLFLAG_SKIP) 548 continue; 549 550 if (!namelen) { 551 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE) 552 return 0; 553 if (oidp->oid_handler) 554 /* We really should call the handler here...*/ 555 return 0; 556 lsp = (struct sysctl_oid_list *)oidp->oid_arg1; 557 if (!sysctl_sysctl_next_ls(lsp, 0, 0, next+1, 558 len, level+1, oidpp)) 559 return 0; 560 goto next; 561 } 562 563 if (oidp->oid_number < *name) 564 continue; 565 566 if (oidp->oid_number > *name) { 567 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE) 568 return 0; 569 if (oidp->oid_handler) 570 return 0; 571 lsp = (struct sysctl_oid_list *)oidp->oid_arg1; 572 if (!sysctl_sysctl_next_ls(lsp, name+1, namelen-1, 573 next+1, len, level+1, oidpp)) 574 return (0); 575 goto next; 576 } 577 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE) 578 continue; 579 580 if (oidp->oid_handler) 581 continue; 582 583 lsp = (struct sysctl_oid_list *)oidp->oid_arg1; 584 if (!sysctl_sysctl_next_ls(lsp, name+1, namelen-1, next+1, 585 len, level+1, oidpp)) 586 return (0); 587 next: 588 namelen = 1; 589 *len = level; 590 } 591 return 1; 592} 593 594static int 595sysctl_sysctl_next(SYSCTL_HANDLER_ARGS) 596{ 597 int *name = (int *) arg1; 598 u_int namelen = arg2; 599 int i, j, error; 600 struct sysctl_oid *oid; 601 struct sysctl_oid_list *lsp = &sysctl__children; 602 int newoid[CTL_MAXNAME]; 603 604 i = sysctl_sysctl_next_ls(lsp, name, namelen, newoid, &j, 1, &oid); 605 if (i) 606 return ENOENT; 607 error = SYSCTL_OUT(req, newoid, j * sizeof (int)); 608 return (error); 609} 610 611SYSCTL_NODE(_sysctl, 2, next, CTLFLAG_RD, sysctl_sysctl_next, ""); 612 613static int 614name2oid (char *name, int *oid, int *len, struct sysctl_oid **oidpp) 615{ 616 int i; 617 struct sysctl_oid *oidp; 618 struct sysctl_oid_list *lsp = &sysctl__children; 619 char *p; 620 621 if (!*name) 622 return ENOENT; 623 624 p = name + strlen(name) - 1 ; 625 if (*p == '.') 626 *p = '\0'; 627 628 *len = 0; 629 630 for (p = name; *p && *p != '.'; p++) 631 ; 632 i = *p; 633 if (i == '.') 634 *p = '\0'; 635 636 oidp = SLIST_FIRST(lsp); 637 638 while (oidp && *len < CTL_MAXNAME) { 639 if (strcmp(name, oidp->oid_name)) { 640 oidp = SLIST_NEXT(oidp, oid_link); 641 continue; 642 } 643 *oid++ = oidp->oid_number; 644 (*len)++; 645 646 if (!i) { 647 if (oidpp) 648 *oidpp = oidp; 649 return (0); 650 } 651 652 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE) 653 break; 654 655 if (oidp->oid_handler) 656 break; 657 658 lsp = (struct sysctl_oid_list *)oidp->oid_arg1; 659 oidp = SLIST_FIRST(lsp); 660 name = p+1; 661 for (p = name; *p && *p != '.'; p++) 662 ; 663 i = *p; 664 if (i == '.') 665 *p = '\0'; 666 } 667 return ENOENT; 668} 669 670static int 671sysctl_sysctl_name2oid(SYSCTL_HANDLER_ARGS) 672{ 673 char *p; 674 int error, oid[CTL_MAXNAME], len; 675 struct sysctl_oid *op = 0; 676 677 if (!req->newlen) 678 return ENOENT; 679 if (req->newlen >= MAXPATHLEN) /* XXX arbitrary, undocumented */ 680 return (ENAMETOOLONG); 681 682 p = malloc(req->newlen+1, M_SYSCTL, M_WAITOK); 683 684 error = SYSCTL_IN(req, p, req->newlen); 685 if (error) { 686 free(p, M_SYSCTL); 687 return (error); 688 } 689 690 p [req->newlen] = '\0'; 691 692 error = name2oid(p, oid, &len, &op); 693 694 free(p, M_SYSCTL); 695 696 if (error) 697 return (error); 698 699 error = SYSCTL_OUT(req, oid, len * sizeof *oid); 700 return (error); 701} 702 703SYSCTL_PROC(_sysctl, 3, name2oid, CTLFLAG_RW|CTLFLAG_ANYBODY, 0, 0, 704 sysctl_sysctl_name2oid, "I", ""); 705 706static int 707sysctl_sysctl_oidfmt(SYSCTL_HANDLER_ARGS) 708{ 709 struct sysctl_oid *oid; 710 int error; 711 712 error = sysctl_find_oid(arg1, arg2, &oid, NULL, req); 713 if (error) 714 return (error); 715 716 if (!oid->oid_fmt) 717 return (ENOENT); 718 error = SYSCTL_OUT(req, &oid->oid_kind, sizeof(oid->oid_kind)); 719 if (error) 720 return (error); 721 error = SYSCTL_OUT(req, oid->oid_fmt, strlen(oid->oid_fmt) + 1); 722 return (error); 723} 724 725 726SYSCTL_NODE(_sysctl, 4, oidfmt, CTLFLAG_RD, sysctl_sysctl_oidfmt, ""); 727 728static int 729sysctl_sysctl_oiddescr(SYSCTL_HANDLER_ARGS) 730{ 731 struct sysctl_oid *oid; 732 int error; 733 734 error = sysctl_find_oid(arg1, arg2, &oid, NULL, req); 735 if (error) 736 return (error); 737 738 if (!oid->descr) 739 return (ENOENT); 740 error = SYSCTL_OUT(req, oid->descr, strlen(oid->descr) + 1); 741 return (error); 742} 743 744SYSCTL_NODE(_sysctl, 5, oiddescr, CTLFLAG_RD, sysctl_sysctl_oiddescr, ""); 745 746/* 747 * Default "handler" functions. 748 */ 749 750/* 751 * Handle an int, signed or unsigned. 752 * Two cases: 753 * a variable: point arg1 at it. 754 * a constant: pass it in arg2. 755 */ 756 757int 758sysctl_handle_int(SYSCTL_HANDLER_ARGS) 759{ 760 int tmpout, error = 0; 761 762 /* 763 * Attempt to get a coherent snapshot by making a copy of the data. 764 */ 765 if (arg1) 766 tmpout = *(int *)arg1; 767 else 768 tmpout = arg2; 769 error = SYSCTL_OUT(req, &tmpout, sizeof(int)); 770 771 if (error || !req->newptr) 772 return (error); 773 774 if (!arg1) 775 error = EPERM; 776 else 777 error = SYSCTL_IN(req, arg1, sizeof(int)); 778 return (error); 779} 780 781/* 782 * Handle a long, signed or unsigned. arg1 points to it. 783 */ 784 785int 786sysctl_handle_long(SYSCTL_HANDLER_ARGS) 787{ 788 int error = 0; 789 long tmpout; 790 791 /* 792 * Attempt to get a coherent snapshot by making a copy of the data. 793 */ 794 if (!arg1) 795 return (EINVAL); 796 tmpout = *(long *)arg1; 797 error = SYSCTL_OUT(req, &tmpout, sizeof(long)); 798 799 if (error || !req->newptr) 800 return (error); 801 802 error = SYSCTL_IN(req, arg1, sizeof(long)); 803 return (error); 804} 805 806/* 807 * Handle our generic '\0' terminated 'C' string. 808 * Two cases: 809 * a variable string: point arg1 at it, arg2 is max length. 810 * a constant string: point arg1 at it, arg2 is zero. 811 */ 812 813int 814sysctl_handle_string(SYSCTL_HANDLER_ARGS) 815{ 816 int error=0; 817 char *tmparg; 818 size_t outlen; 819 820 /* 821 * Attempt to get a coherent snapshot by copying to a 822 * temporary kernel buffer. 823 */ 824retry: 825 outlen = strlen((char *)arg1)+1; 826 tmparg = malloc(outlen, M_SYSCTLTMP, M_WAITOK); 827 828 if (strlcpy(tmparg, (char *)arg1, outlen) >= outlen) { 829 free(tmparg, M_SYSCTLTMP); 830 goto retry; 831 } 832 833 error = SYSCTL_OUT(req, tmparg, outlen); 834 free(tmparg, M_SYSCTLTMP); 835 836 if (error || !req->newptr) 837 return (error); 838 839 if ((req->newlen - req->newidx) >= arg2) { 840 error = EINVAL; 841 } else { 842 arg2 = (req->newlen - req->newidx); 843 error = SYSCTL_IN(req, arg1, arg2); 844 ((char *)arg1)[arg2] = '\0'; 845 } 846 847 return (error); 848} 849 850/* 851 * Handle any kind of opaque data. 852 * arg1 points to it, arg2 is the size. 853 */ 854 855int 856sysctl_handle_opaque(SYSCTL_HANDLER_ARGS) 857{ 858 int error; 859 void *tmparg; 860 861 /* 862 * Attempt to get a coherent snapshot, either by wiring the 863 * user space buffer or copying to a temporary kernel buffer 864 * depending on the size of the data. 865 */ 866 if (arg2 > PAGE_SIZE) { 867 sysctl_wire_old_buffer(req, arg2); 868 error = SYSCTL_OUT(req, arg1, arg2); 869 } else { 870 tmparg = malloc(arg2, M_SYSCTLTMP, M_WAITOK); 871 bcopy(arg1, tmparg, arg2); 872 error = SYSCTL_OUT(req, tmparg, arg2); 873 free(tmparg, M_SYSCTLTMP); 874 } 875 876 if (error || !req->newptr) 877 return (error); 878 879 error = SYSCTL_IN(req, arg1, arg2); 880 881 return (error); 882} 883 884/* 885 * Transfer functions to/from kernel space. 886 * XXX: rather untested at this point 887 */ 888static int 889sysctl_old_kernel(struct sysctl_req *req, const void *p, size_t l) 890{ 891 size_t i = 0; 892 893 if (req->oldptr) { 894 i = l; 895 if (req->oldlen <= req->oldidx) 896 i = 0; 897 else 898 if (i > req->oldlen - req->oldidx) 899 i = req->oldlen - req->oldidx; 900 if (i > 0) 901 bcopy(p, (char *)req->oldptr + req->oldidx, i); 902 } 903 req->oldidx += l; 904 if (req->oldptr && i != l) 905 return (ENOMEM); 906 return (0); 907} 908 909static int 910sysctl_new_kernel(struct sysctl_req *req, void *p, size_t l) 911{ 912 if (!req->newptr) 913 return 0; 914 if (req->newlen - req->newidx < l) 915 return (EINVAL); 916 bcopy((char *)req->newptr + req->newidx, p, l); 917 req->newidx += l; 918 return (0); 919} 920 921int 922kernel_sysctl(struct thread *td, int *name, u_int namelen, void *old, 923 size_t *oldlenp, void *new, size_t newlen, size_t *retval) 924{ 925 int error = 0; 926 struct sysctl_req req; 927 928 bzero(&req, sizeof req); 929 930 req.td = td; 931 932 if (oldlenp) { 933 req.oldlen = *oldlenp; 934 } 935 936 if (old) { 937 req.oldptr= old; 938 } 939 940 if (new != NULL) { 941 req.newlen = newlen; 942 req.newptr = new; 943 } 944 945 req.oldfunc = sysctl_old_kernel; 946 req.newfunc = sysctl_new_kernel; 947 req.lock = 1; 948 949 SYSCTL_LOCK(); 950 951 error = sysctl_root(0, name, namelen, &req); 952 953 if (req.lock == 2) 954 vsunlock(req.oldptr, req.oldlen); 955 956 SYSCTL_UNLOCK(); 957 958 if (error && error != ENOMEM) 959 return (error); 960 961 if (retval) { 962 if (req.oldptr && req.oldidx > req.oldlen) 963 *retval = req.oldlen; 964 else 965 *retval = req.oldidx; 966 } 967 return (error); 968} 969 970int 971kernel_sysctlbyname(struct thread *td, char *name, void *old, size_t *oldlenp, 972 void *new, size_t newlen, size_t *retval) 973{ 974 int oid[CTL_MAXNAME]; 975 size_t oidlen, plen; 976 int error; 977 978 oid[0] = 0; /* sysctl internal magic */ 979 oid[1] = 3; /* name2oid */ 980 oidlen = sizeof(oid); 981 982 error = kernel_sysctl(td, oid, 2, oid, &oidlen, 983 (void *)name, strlen(name), &plen); 984 if (error) 985 return (error); 986 987 error = kernel_sysctl(td, oid, plen / sizeof(int), old, oldlenp, 988 new, newlen, retval); 989 return (error); 990} 991 992/* 993 * Transfer function to/from user space. 994 */ 995static int 996sysctl_old_user(struct sysctl_req *req, const void *p, size_t l) 997{ 998 int error = 0; 999 size_t i = 0; 1000 1001 if (req->lock == 1 && req->oldptr) 1002 WITNESS_SLEEP(1, NULL); 1003 if (req->oldptr) { 1004 i = l; 1005 if (req->oldlen <= req->oldidx) 1006 i = 0; 1007 else 1008 if (i > req->oldlen - req->oldidx) 1009 i = req->oldlen - req->oldidx; 1010 if (i > 0) 1011 error = copyout(p, (char *)req->oldptr + req->oldidx, 1012 i); 1013 } 1014 req->oldidx += l; 1015 if (error) 1016 return (error); 1017 if (req->oldptr && i < l) 1018 return (ENOMEM); 1019 return (0); 1020} 1021 1022static int 1023sysctl_new_user(struct sysctl_req *req, void *p, size_t l) 1024{ 1025 int error; 1026 1027 if (!req->newptr) 1028 return 0; 1029 if (req->newlen - req->newidx < l) 1030 return (EINVAL); 1031 error = copyin((char *)req->newptr + req->newidx, p, l); 1032 req->newidx += l; 1033 return (error); 1034} 1035 1036/* 1037 * Wire the user space destination buffer. If set to a value greater than 1038 * zero, the len parameter limits the maximum amount of wired memory. 1039 * 1040 * XXX - The len parameter is currently ignored due to the lack of 1041 * a place to save it in the sysctl_req structure so that the matching 1042 * amount of memory can be unwired in the sysctl exit code. 1043 */ 1044void 1045sysctl_wire_old_buffer(struct sysctl_req *req, size_t len) 1046{ 1047 if (req->lock == 1 && req->oldptr && req->oldfunc == sysctl_old_user) { 1048 vslock(req->oldptr, req->oldlen); 1049 req->lock = 2; 1050 } 1051} 1052 1053int 1054sysctl_find_oid(int *name, u_int namelen, struct sysctl_oid **noid, 1055 int *nindx, struct sysctl_req *req) 1056{ 1057 struct sysctl_oid *oid; 1058 int indx; 1059 1060 oid = SLIST_FIRST(&sysctl__children); 1061 indx = 0; 1062 while (oid && indx < CTL_MAXNAME) { 1063 if (oid->oid_number == name[indx]) { 1064 indx++; 1065 if (oid->oid_kind & CTLFLAG_NOLOCK) 1066 req->lock = 0; 1067 if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) { 1068 if (oid->oid_handler != NULL || 1069 indx == namelen) { 1070 *noid = oid; 1071 if (nindx != NULL) 1072 *nindx = indx; 1073 return (0); 1074 } 1075 oid = SLIST_FIRST( 1076 (struct sysctl_oid_list *)oid->oid_arg1); 1077 } else if (indx == namelen) { 1078 *noid = oid; 1079 if (nindx != NULL) 1080 *nindx = indx; 1081 return (0); 1082 } else { 1083 return (ENOTDIR); 1084 } 1085 } else { 1086 oid = SLIST_NEXT(oid, oid_link); 1087 } 1088 } 1089 return (ENOENT); 1090} 1091 1092/* 1093 * Traverse our tree, and find the right node, execute whatever it points 1094 * to, and return the resulting error code. 1095 */ 1096 1097static int 1098sysctl_root(SYSCTL_HANDLER_ARGS) 1099{ 1100 struct sysctl_oid *oid; 1101 int error, indx; 1102 1103 error = sysctl_find_oid(arg1, arg2, &oid, &indx, req); 1104 if (error) 1105 return (error); 1106 1107 if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) { 1108 /* 1109 * You can't call a sysctl when it's a node, but has 1110 * no handler. Inform the user that it's a node. 1111 * The indx may or may not be the same as namelen. 1112 */ 1113 if (oid->oid_handler == NULL) 1114 return (EISDIR); 1115 } 1116 1117 /* Is this sysctl writable? */ 1118 if (req->newptr && !(oid->oid_kind & CTLFLAG_WR)) 1119 return (EPERM); 1120 1121 KASSERT(req->td != NULL, ("sysctl_root(): req->td == NULL")); 1122 1123 /* Is this sysctl sensitive to securelevels? */ 1124 if (req->newptr && (oid->oid_kind & CTLFLAG_SECURE)) { 1125 error = securelevel_gt(req->td->td_ucred, 0); 1126 if (error) 1127 return (error); 1128 } 1129 1130 /* Is this sysctl writable by only privileged users? */ 1131 if (req->newptr && !(oid->oid_kind & CTLFLAG_ANYBODY)) { 1132 int flags; 1133 1134 if (oid->oid_kind & CTLFLAG_PRISON) 1135 flags = PRISON_ROOT; 1136 else 1137 flags = 0; 1138 error = suser_cred(req->td->td_ucred, flags); 1139 if (error) 1140 return (error); 1141 } 1142 1143 if (!oid->oid_handler) 1144 return EINVAL; 1145 1146 if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) 1147 error = oid->oid_handler(oid, (int *)arg1 + indx, arg2 - indx, 1148 req); 1149 else 1150 error = oid->oid_handler(oid, oid->oid_arg1, oid->oid_arg2, 1151 req); 1152 return (error); 1153} 1154 1155#ifndef _SYS_SYSPROTO_H_ 1156struct sysctl_args { 1157 int *name; 1158 u_int namelen; 1159 void *old; 1160 size_t *oldlenp; 1161 void *new; 1162 size_t newlen; 1163}; 1164#endif 1165 1166/* 1167 * MPSAFE 1168 */ 1169int 1170__sysctl(struct thread *td, struct sysctl_args *uap) 1171{ 1172 int error, name[CTL_MAXNAME]; 1173 size_t j; 1174 1175 if (uap->namelen > CTL_MAXNAME || uap->namelen < 2) 1176 return (EINVAL); 1177 1178 error = copyin(uap->name, &name, uap->namelen * sizeof(int)); 1179 if (error) 1180 return (error); 1181 1182 mtx_lock(&Giant); 1183 1184 error = userland_sysctl(td, name, uap->namelen, 1185 uap->old, uap->oldlenp, 0, 1186 uap->new, uap->newlen, &j); 1187 if (error && error != ENOMEM) 1188 goto done2; 1189 if (uap->oldlenp) { 1190 int i = copyout(&j, uap->oldlenp, sizeof(j)); 1191 if (i) 1192 error = i; 1193 } 1194done2: 1195 mtx_unlock(&Giant); 1196 return (error); 1197} 1198 1199/* 1200 * This is used from various compatibility syscalls too. That's why name 1201 * must be in kernel space. 1202 */ 1203int 1204userland_sysctl(struct thread *td, int *name, u_int namelen, void *old, 1205 size_t *oldlenp, int inkernel, void *new, size_t newlen, size_t *retval) 1206{ 1207 int error = 0; 1208 struct sysctl_req req, req2; 1209 1210 bzero(&req, sizeof req); 1211 1212 req.td = td; 1213 1214 if (oldlenp) { 1215 if (inkernel) { 1216 req.oldlen = *oldlenp; 1217 } else { 1218 error = copyin(oldlenp, &req.oldlen, sizeof(*oldlenp)); 1219 if (error) 1220 return (error); 1221 } 1222 } 1223 1224 if (old) { 1225 if (!useracc(old, req.oldlen, VM_PROT_WRITE)) 1226 return (EFAULT); 1227 req.oldptr= old; 1228 } 1229 1230 if (new != NULL) { 1231 if (!useracc(new, req.newlen, VM_PROT_READ)) 1232 return (EFAULT); 1233 req.newlen = newlen; 1234 req.newptr = new; 1235 } 1236 1237 req.oldfunc = sysctl_old_user; 1238 req.newfunc = sysctl_new_user; 1239 req.lock = 1; 1240 1241 SYSCTL_LOCK(); 1242 1243#ifdef MAC 1244 error = mac_check_system_sysctl(td->td_ucred, name, namelen, old, 1245 oldlenp, inkernel, new, newlen); 1246 if (error) { 1247 SYSCTL_UNLOCK(); 1248 return (error); 1249 } 1250#endif 1251 1252 do { 1253 req2 = req; 1254 error = sysctl_root(0, name, namelen, &req2); 1255 } while (error == EAGAIN); 1256 1257 req = req2; 1258 if (req.lock == 2) 1259 vsunlock(req.oldptr, req.oldlen); 1260 1261 SYSCTL_UNLOCK(); 1262 1263 if (error && error != ENOMEM) 1264 return (error); 1265 1266 if (retval) { 1267 if (req.oldptr && req.oldidx > req.oldlen) 1268 *retval = req.oldlen; 1269 else 1270 *retval = req.oldidx; 1271 } 1272 return (error); 1273} 1274 1275#ifdef COMPAT_43 1276#include <sys/socket.h> 1277#include <vm/vm_param.h> 1278 1279#define KINFO_PROC (0<<8) 1280#define KINFO_RT (1<<8) 1281#define KINFO_VNODE (2<<8) 1282#define KINFO_FILE (3<<8) 1283#define KINFO_METER (4<<8) 1284#define KINFO_LOADAVG (5<<8) 1285#define KINFO_CLOCKRATE (6<<8) 1286 1287/* Non-standard BSDI extension - only present on their 4.3 net-2 releases */ 1288#define KINFO_BSDI_SYSINFO (101<<8) 1289 1290/* 1291 * XXX this is bloat, but I hope it's better here than on the potentially 1292 * limited kernel stack... -Peter 1293 */ 1294 1295static struct { 1296 int bsdi_machine; /* "i386" on BSD/386 */ 1297/* ^^^ this is an offset to the string, relative to the struct start */ 1298 char *pad0; 1299 long pad1; 1300 long pad2; 1301 long pad3; 1302 u_long pad4; 1303 u_long pad5; 1304 u_long pad6; 1305 1306 int bsdi_ostype; /* "BSD/386" on BSD/386 */ 1307 int bsdi_osrelease; /* "1.1" on BSD/386 */ 1308 long pad7; 1309 long pad8; 1310 char *pad9; 1311 1312 long pad10; 1313 long pad11; 1314 int pad12; 1315 long pad13; 1316 quad_t pad14; 1317 long pad15; 1318 1319 struct timeval pad16; 1320 /* we dont set this, because BSDI's uname used gethostname() instead */ 1321 int bsdi_hostname; /* hostname on BSD/386 */ 1322 1323 /* the actual string data is appended here */ 1324 1325} bsdi_si; 1326/* 1327 * this data is appended to the end of the bsdi_si structure during copyout. 1328 * The "char *" offsets are relative to the base of the bsdi_si struct. 1329 * This contains "FreeBSD\02.0-BUILT-nnnnnn\0i386\0", and these strings 1330 * should not exceed the length of the buffer here... (or else!! :-) 1331 */ 1332static char bsdi_strings[80]; /* It had better be less than this! */ 1333 1334#ifndef _SYS_SYSPROTO_H_ 1335struct getkerninfo_args { 1336 int op; 1337 char *where; 1338 size_t *size; 1339 int arg; 1340}; 1341#endif 1342 1343/* 1344 * MPSAFE 1345 */ 1346int 1347ogetkerninfo(struct thread *td, struct getkerninfo_args *uap) 1348{ 1349 int error, name[6]; 1350 size_t size; 1351 u_int needed = 0; 1352 1353 mtx_lock(&Giant); 1354 1355 switch (uap->op & 0xff00) { 1356 1357 case KINFO_RT: 1358 name[0] = CTL_NET; 1359 name[1] = PF_ROUTE; 1360 name[2] = 0; 1361 name[3] = (uap->op & 0xff0000) >> 16; 1362 name[4] = uap->op & 0xff; 1363 name[5] = uap->arg; 1364 error = userland_sysctl(td, name, 6, uap->where, uap->size, 1365 0, 0, 0, &size); 1366 break; 1367 1368 case KINFO_VNODE: 1369 name[0] = CTL_KERN; 1370 name[1] = KERN_VNODE; 1371 error = userland_sysctl(td, name, 2, uap->where, uap->size, 1372 0, 0, 0, &size); 1373 break; 1374 1375 case KINFO_PROC: 1376 name[0] = CTL_KERN; 1377 name[1] = KERN_PROC; 1378 name[2] = uap->op & 0xff; 1379 name[3] = uap->arg; 1380 error = userland_sysctl(td, name, 4, uap->where, uap->size, 1381 0, 0, 0, &size); 1382 break; 1383 1384 case KINFO_FILE: 1385 name[0] = CTL_KERN; 1386 name[1] = KERN_FILE; 1387 error = userland_sysctl(td, name, 2, uap->where, uap->size, 1388 0, 0, 0, &size); 1389 break; 1390 1391 case KINFO_METER: 1392 name[0] = CTL_VM; 1393 name[1] = VM_METER; 1394 error = userland_sysctl(td, name, 2, uap->where, uap->size, 1395 0, 0, 0, &size); 1396 break; 1397 1398 case KINFO_LOADAVG: 1399 name[0] = CTL_VM; 1400 name[1] = VM_LOADAVG; 1401 error = userland_sysctl(td, name, 2, uap->where, uap->size, 1402 0, 0, 0, &size); 1403 break; 1404 1405 case KINFO_CLOCKRATE: 1406 name[0] = CTL_KERN; 1407 name[1] = KERN_CLOCKRATE; 1408 error = userland_sysctl(td, name, 2, uap->where, uap->size, 1409 0, 0, 0, &size); 1410 break; 1411 1412 case KINFO_BSDI_SYSINFO: { 1413 /* 1414 * this is pretty crude, but it's just enough for uname() 1415 * from BSDI's 1.x libc to work. 1416 * 1417 * *size gives the size of the buffer before the call, and 1418 * the amount of data copied after a successful call. 1419 * If successful, the return value is the amount of data 1420 * available, which can be larger than *size. 1421 * 1422 * BSDI's 2.x product apparently fails with ENOMEM if *size 1423 * is too small. 1424 */ 1425 1426 u_int left; 1427 char *s; 1428 1429 bzero((char *)&bsdi_si, sizeof(bsdi_si)); 1430 bzero(bsdi_strings, sizeof(bsdi_strings)); 1431 1432 s = bsdi_strings; 1433 1434 bsdi_si.bsdi_ostype = (s - bsdi_strings) + sizeof(bsdi_si); 1435 strcpy(s, ostype); 1436 s += strlen(s) + 1; 1437 1438 bsdi_si.bsdi_osrelease = (s - bsdi_strings) + sizeof(bsdi_si); 1439 strcpy(s, osrelease); 1440 s += strlen(s) + 1; 1441 1442 bsdi_si.bsdi_machine = (s - bsdi_strings) + sizeof(bsdi_si); 1443 strcpy(s, machine); 1444 s += strlen(s) + 1; 1445 1446 needed = sizeof(bsdi_si) + (s - bsdi_strings); 1447 1448 if ((uap->where == NULL) || (uap->size == NULL)) { 1449 /* process is asking how much buffer to supply.. */ 1450 size = needed; 1451 error = 0; 1452 break; 1453 } 1454 1455 if ((error = copyin(uap->size, &size, sizeof(size))) != 0) 1456 break; 1457 1458 /* if too much buffer supplied, trim it down */ 1459 if (size > needed) 1460 size = needed; 1461 1462 /* how much of the buffer is remaining */ 1463 left = size; 1464 1465 if ((error = copyout((char *)&bsdi_si, uap->where, left)) != 0) 1466 break; 1467 1468 /* is there any point in continuing? */ 1469 if (left > sizeof(bsdi_si)) { 1470 left -= sizeof(bsdi_si); 1471 error = copyout(&bsdi_strings, 1472 uap->where + sizeof(bsdi_si), left); 1473 } 1474 break; 1475 } 1476 1477 default: 1478 error = EOPNOTSUPP; 1479 break; 1480 } 1481 if (error == 0) { 1482 td->td_retval[0] = needed ? needed : size; 1483 if (uap->size) { 1484 error = copyout(&size, uap->size, sizeof(size)); 1485 } 1486 } 1487 mtx_unlock(&Giant); 1488 return (error); 1489} 1490#endif /* COMPAT_43 */ 1491