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. Neither the name of the University nor the names of its contributors 20 * may be used to endorse or promote products derived from this software 21 * without specific prior written permission. 22 * 23 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 24 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 25 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 26 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 27 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 28 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 29 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 30 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 31 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 32 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 33 * SUCH DAMAGE. 34 * 35 * @(#)kern_sysctl.c 8.4 (Berkeley) 4/14/94 36 * $FreeBSD: src/sys/kern/kern_sysctl.c,v 1.92.2.9 2003/05/01 22:48:09 trhodes Exp $ 37 */ 38 39 #include <sys/param.h> 40 #include <sys/systm.h> 41 #include <sys/kernel.h> 42 #include <sys/buf.h> 43 #include <sys/sysctl.h> 44 #include <sys/malloc.h> 45 #include <sys/proc.h> 46 #include <sys/priv.h> 47 #include <sys/sysproto.h> 48 #include <sys/lock.h> 49 #include <sys/sbuf.h> 50 51 #include <vm/vm.h> 52 #include <vm/vm_extern.h> 53 54 static MALLOC_DEFINE(M_SYSCTL, "sysctl", "sysctl internal magic"); 55 static MALLOC_DEFINE(M_SYSCTLOID, "sysctloid", "sysctl dynamic oids"); 56 57 /* 58 * The sysctllock protects the MIB tree. It also protects sysctl 59 * contexts used with dynamic sysctls. The sysctl_register_oid() and 60 * sysctl_unregister_oid() routines require the sysctllock to already 61 * be held, so the sysctl_lock() and sysctl_unlock() routines are 62 * provided for the few places in the kernel which need to use that 63 * API rather than using the dynamic API. Use of the dynamic API is 64 * strongly encouraged for most code. 65 * 66 * The sysctlmemlock is used to limit the amount of user memory wired for 67 * sysctl requests. This is implemented by serializing any userland 68 * sysctl requests larger than a single page via an exclusive lock. 69 */ 70 struct lock sysctllock; 71 static struct lock sysctlmemlock; 72 73 #define SYSCTL_INIT() lockinit(&sysctllock, \ 74 "sysctl lock", 0, LK_CANRECURSE) 75 #define SYSCTL_SLEEP(ch, wmesg, timo) \ 76 lksleep(ch, &sysctllock, 0, wmesg, timo) 77 78 static int sysctl_root(SYSCTL_HANDLER_ARGS); 79 static void sysctl_register_oid_int(struct sysctl_oid *oipd); 80 static void sysctl_unregister_oid_int(struct sysctl_oid *oipd); 81 82 struct sysctl_oid_list sysctl__children; /* root list */ 83 84 static int sysctl_remove_oid_locked(struct sysctl_oid *oidp, int del, 85 int recurse); 86 87 static struct sysctl_oid * 88 sysctl_find_oidname(const char *name, struct sysctl_oid_list *list, int lock) 89 { 90 struct sysctl_oid *oidp; 91 92 SLIST_FOREACH(oidp, list, oid_link) { 93 if (strcmp(oidp->oid_name, name) == 0) { 94 break; 95 } 96 } 97 return (oidp); 98 } 99 100 /* 101 * Initialization of the MIB tree. 102 * 103 * Order by number in each list. 104 */ 105 106 void 107 sysctl_register_oid(struct sysctl_oid *oidp) 108 { 109 SYSCTL_XLOCK(); 110 sysctl_register_oid_int(oidp); 111 SYSCTL_XUNLOCK(); 112 } 113 114 static void 115 sysctl_register_oid_int(struct sysctl_oid *oidp) 116 { 117 struct sysctl_oid_list *parent = oidp->oid_parent; 118 struct sysctl_oid *p; 119 struct sysctl_oid *q; 120 121 /* 122 * Finish initialization from sysctl_set or add. 123 */ 124 lockinit(&oidp->oid_lock, "oidlk", 0, LK_CANRECURSE); 125 126 /* 127 * First check if another oid with the same name already 128 * exists in the parent's list. 129 */ 130 p = sysctl_find_oidname(oidp->oid_name, parent, 0); 131 if (p != NULL) { 132 if ((p->oid_kind & CTLTYPE) == CTLTYPE_NODE) 133 p->oid_refcnt++; 134 else 135 kprintf("can't re-use a leaf (%s)!\n", p->oid_name); 136 return; 137 } 138 139 /* 140 * If this oid has a number OID_AUTO, give it a number which 141 * is greater than any current oid. Make sure it is at least 142 * 256 to leave space for pre-assigned oid numbers. 143 */ 144 if (oidp->oid_number == OID_AUTO) { 145 int newoid = 0x100; /* minimum AUTO oid */ 146 147 /* 148 * Adjust based on highest oid in parent list 149 */ 150 SLIST_FOREACH(p, parent, oid_link) { 151 if (newoid <= p->oid_number) 152 newoid = p->oid_number + 1; 153 } 154 oidp->oid_number = newoid; 155 } 156 157 /* 158 * Insert the oid into the parent's list in order. 159 */ 160 q = NULL; 161 SLIST_FOREACH(p, parent, oid_link) { 162 if (oidp->oid_number < p->oid_number) 163 break; 164 q = p; 165 } 166 if (q) 167 SLIST_INSERT_AFTER(q, oidp, oid_link); 168 else 169 SLIST_INSERT_HEAD(parent, oidp, oid_link); 170 } 171 172 void 173 sysctl_unregister_oid(struct sysctl_oid *oidp) 174 { 175 SYSCTL_XLOCK(); 176 sysctl_unregister_oid_int(oidp); 177 SYSCTL_XUNLOCK(); 178 } 179 180 static void 181 sysctl_unregister_oid_int(struct sysctl_oid *oidp) 182 { 183 struct sysctl_oid *p; 184 185 if (oidp->oid_number == OID_AUTO) 186 panic("Trying to unregister OID_AUTO entry: %p", oidp); 187 188 SLIST_FOREACH(p, oidp->oid_parent, oid_link) { 189 if (p != oidp) 190 continue; 191 SLIST_REMOVE(oidp->oid_parent, oidp, sysctl_oid, oid_link); 192 return; 193 } 194 195 /* 196 * This can happen when a module fails to register and is 197 * being unloaded afterwards. It should not be a panic() 198 * for normal use. 199 */ 200 kprintf("%s: failed to unregister sysctl\n", __func__); 201 } 202 203 /* Initialize a new context to keep track of dynamically added sysctls. */ 204 int 205 sysctl_ctx_init(struct sysctl_ctx_list *c) 206 { 207 if (c == NULL) 208 return(EINVAL); 209 TAILQ_INIT(c); 210 return(0); 211 } 212 213 /* Free the context, and destroy all dynamic oids registered in this context */ 214 int 215 sysctl_ctx_free(struct sysctl_ctx_list *clist) 216 { 217 struct sysctl_ctx_entry *e, *e1; 218 int error; 219 220 error = 0; 221 /* 222 * First perform a "dry run" to check if it's ok to remove oids. 223 * XXX FIXME 224 * XXX This algorithm is a hack. But I don't know any 225 * XXX better solution for now... 226 */ 227 SYSCTL_XLOCK(); 228 TAILQ_FOREACH(e, clist, link) { 229 error = sysctl_remove_oid_locked(e->entry, 0, 0); 230 if (error) 231 break; 232 } 233 /* 234 * Restore deregistered entries, either from the end, 235 * or from the place where error occured. 236 * e contains the entry that was not unregistered 237 */ 238 if (error) 239 e1 = TAILQ_PREV(e, sysctl_ctx_list, link); 240 else 241 e1 = TAILQ_LAST(clist, sysctl_ctx_list); 242 while (e1 != NULL) { 243 sysctl_register_oid(e1->entry); 244 e1 = TAILQ_PREV(e1, sysctl_ctx_list, link); 245 } 246 if (error) { 247 SYSCTL_XUNLOCK(); 248 return(EBUSY); 249 } 250 /* Now really delete the entries */ 251 e = TAILQ_FIRST(clist); 252 while (e != NULL) { 253 e1 = TAILQ_NEXT(e, link); 254 error = sysctl_remove_oid_locked(e->entry, 1, 0); 255 if (error) 256 panic("sysctl_remove_oid: corrupt tree, entry: %s", 257 e->entry->oid_name); 258 kfree(e, M_SYSCTLOID); 259 e = e1; 260 } 261 SYSCTL_XUNLOCK(); 262 return (error); 263 } 264 265 /* Add an entry to the context */ 266 struct sysctl_ctx_entry * 267 sysctl_ctx_entry_add(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp) 268 { 269 struct sysctl_ctx_entry *e; 270 271 SYSCTL_ASSERT_LOCKED(); 272 if (clist == NULL || oidp == NULL) 273 return(NULL); 274 e = kmalloc(sizeof(struct sysctl_ctx_entry), M_SYSCTLOID, M_WAITOK); 275 e->entry = oidp; 276 TAILQ_INSERT_HEAD(clist, e, link); 277 return (e); 278 } 279 280 /* Find an entry in the context */ 281 struct sysctl_ctx_entry * 282 sysctl_ctx_entry_find(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp) 283 { 284 struct sysctl_ctx_entry *e; 285 286 SYSCTL_ASSERT_LOCKED(); 287 if (clist == NULL || oidp == NULL) 288 return(NULL); 289 TAILQ_FOREACH(e, clist, link) { 290 if(e->entry == oidp) 291 return(e); 292 } 293 return (e); 294 } 295 296 /* 297 * Delete an entry from the context. 298 * NOTE: this function doesn't free oidp! You have to remove it 299 * with sysctl_remove_oid(). 300 */ 301 int 302 sysctl_ctx_entry_del(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp) 303 { 304 struct sysctl_ctx_entry *e; 305 306 if (clist == NULL || oidp == NULL) 307 return (EINVAL); 308 SYSCTL_XLOCK(); 309 e = sysctl_ctx_entry_find(clist, oidp); 310 if (e != NULL) { 311 TAILQ_REMOVE(clist, e, link); 312 SYSCTL_XUNLOCK(); 313 kfree(e, M_SYSCTLOID); 314 return (0); 315 } else { 316 SYSCTL_XUNLOCK(); 317 return (ENOENT); 318 } 319 } 320 321 /* 322 * Remove dynamically created sysctl trees. 323 * oidp - top of the tree to be removed 324 * del - if 0 - just deregister, otherwise free up entries as well 325 * recurse - if != 0 traverse the subtree to be deleted 326 */ 327 int 328 sysctl_remove_oid(struct sysctl_oid *oidp, int del, int recurse) 329 { 330 int error; 331 332 SYSCTL_XLOCK(); 333 error = sysctl_remove_oid_locked(oidp, del, recurse); 334 SYSCTL_XUNLOCK(); 335 return (error); 336 } 337 338 static int 339 sysctl_remove_oid_locked(struct sysctl_oid *oidp, int del, int recurse) 340 { 341 struct sysctl_oid *p, *tmp; 342 int error; 343 344 SYSCTL_ASSERT_LOCKED(); 345 if (oidp == NULL) 346 return(EINVAL); 347 if ((oidp->oid_kind & CTLFLAG_DYN) == 0) { 348 kprintf("can't remove non-dynamic nodes!\n"); 349 return (EINVAL); 350 } 351 /* 352 * WARNING: normal method to do this should be through 353 * sysctl_ctx_free(). Use recursing as the last resort 354 * method to purge your sysctl tree of leftovers... 355 * However, if some other code still references these nodes, 356 * it will panic. 357 */ 358 if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) { 359 if (oidp->oid_refcnt == 1) { 360 SLIST_FOREACH_MUTABLE(p, 361 SYSCTL_CHILDREN(oidp), oid_link, tmp) { 362 if (!recurse) { 363 kprintf("Warning: failed attempt to " 364 "remove oid %s with child %s\n", 365 oidp->oid_name, p->oid_name); 366 return (ENOTEMPTY); 367 } 368 error = sysctl_remove_oid_locked(p, del, 369 recurse); 370 if (error) 371 return (error); 372 } 373 if (del) 374 kfree(SYSCTL_CHILDREN(oidp), M_SYSCTLOID); 375 } 376 } 377 if (oidp->oid_refcnt > 1 ) { 378 oidp->oid_refcnt--; 379 } else { 380 if (oidp->oid_refcnt == 0) { 381 kprintf("Warning: bad oid_refcnt=%u (%s)!\n", 382 oidp->oid_refcnt, oidp->oid_name); 383 return (EINVAL); 384 } 385 sysctl_unregister_oid(oidp); 386 if (del) { 387 /* 388 * Wait for all threads running the handler to drain. 389 * This preserves the previous behavior when the 390 * sysctl lock was held across a handler invocation, 391 * and is necessary for module unload correctness. 392 */ 393 while (oidp->oid_running > 0) { 394 oidp->oid_kind |= CTLFLAG_DYING; 395 SYSCTL_SLEEP(&oidp->oid_running, "oidrm", 0); 396 } 397 if (oidp->oid_descr) 398 kfree(__DECONST(char *, oidp->oid_descr), 399 M_SYSCTLOID); 400 kfree(__DECONST(char *, oidp->oid_name), M_SYSCTLOID); 401 lockuninit(&oidp->oid_lock); 402 kfree(oidp, M_SYSCTLOID); 403 } 404 } 405 return (0); 406 } 407 408 int 409 sysctl_remove_name(struct sysctl_oid *parent, const char *name, 410 int del, int recurse) 411 { 412 struct sysctl_oid *p, *tmp; 413 int error; 414 415 error = ENOENT; 416 SYSCTL_XLOCK(); 417 SLIST_FOREACH_MUTABLE(p, SYSCTL_CHILDREN(parent), oid_link, tmp) { 418 if (strcmp(p->oid_name, name) == 0) { 419 error = sysctl_remove_oid_locked(p, del, recurse); 420 break; 421 } 422 } 423 SYSCTL_XUNLOCK(); 424 425 return (error); 426 } 427 428 /* 429 * Create new sysctls at run time. 430 * clist may point to a valid context initialized with sysctl_ctx_init(). 431 */ 432 struct sysctl_oid * 433 sysctl_add_oid(struct sysctl_ctx_list *clist, struct sysctl_oid_list *parent, 434 int number, const char *name, int kind, void *arg1, int arg2, 435 int (*handler)(SYSCTL_HANDLER_ARGS), const char *fmt, const char *descr) 436 { 437 struct sysctl_oid *oidp; 438 ssize_t len; 439 char *newname; 440 441 /* You have to hook up somewhere.. */ 442 if (parent == NULL) 443 return(NULL); 444 SYSCTL_XLOCK(); 445 /* Check if the node already exists, otherwise create it */ 446 oidp = sysctl_find_oidname(name, parent, 0); 447 if (oidp != NULL) { 448 if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) { 449 oidp->oid_refcnt++; 450 /* Update the context */ 451 if (clist != NULL) 452 sysctl_ctx_entry_add(clist, oidp); 453 SYSCTL_XUNLOCK(); 454 return (oidp); 455 } else { 456 kprintf("can't re-use a leaf (%s)!\n", name); 457 SYSCTL_XUNLOCK(); 458 return (NULL); 459 } 460 } 461 oidp = kmalloc(sizeof(struct sysctl_oid), M_SYSCTLOID, 462 M_WAITOK | M_ZERO); 463 oidp->oid_parent = parent; 464 SLIST_NEXT(oidp, oid_link) = NULL; 465 oidp->oid_number = number; 466 oidp->oid_refcnt = 1; 467 len = strlen(name); 468 newname = kmalloc(len + 1, M_SYSCTLOID, M_WAITOK); 469 bcopy(name, newname, len + 1); 470 newname[len] = '\0'; 471 oidp->oid_name = newname; 472 oidp->oid_handler = handler; 473 oidp->oid_kind = CTLFLAG_DYN | kind; 474 if ((kind & CTLTYPE) == CTLTYPE_NODE) { 475 struct sysctl_oid_list *children; 476 477 /* Allocate space for children */ 478 children = kmalloc(sizeof(*children), M_SYSCTLOID, M_WAITOK); 479 SYSCTL_SET_CHILDREN(oidp, children); 480 SLIST_INIT(children); 481 } else { 482 oidp->oid_arg1 = arg1; 483 oidp->oid_arg2 = arg2; 484 } 485 oidp->oid_fmt = fmt; 486 if (descr) { 487 int len = strlen(descr) + 1; 488 oidp->oid_descr = kmalloc(len, M_SYSCTLOID, M_WAITOK); 489 strcpy((char *)(uintptr_t)(const void *)oidp->oid_descr, descr); 490 }; 491 /* Update the context, if used */ 492 if (clist != NULL) 493 sysctl_ctx_entry_add(clist, oidp); 494 /* Register this oid */ 495 sysctl_register_oid_int(oidp); 496 SYSCTL_XUNLOCK(); 497 return (oidp); 498 } 499 500 /* 501 * Rename an existing oid. 502 */ 503 void 504 sysctl_rename_oid(struct sysctl_oid *oidp, const char *name) 505 { 506 char *newname; 507 char *oldname; 508 509 newname = kstrdup(name, M_SYSCTLOID); 510 SYSCTL_XLOCK(); 511 oldname = __DECONST(char *, oidp->oid_name); 512 oidp->oid_name = newname; 513 SYSCTL_XUNLOCK(); 514 kfree(oldname, M_SYSCTLOID); 515 } 516 517 /* 518 * Register the kernel's oids on startup. 519 */ 520 SET_DECLARE(sysctl_set, struct sysctl_oid); 521 522 static void 523 sysctl_register_all(void *arg) 524 { 525 struct sysctl_oid **oidp; 526 527 lockinit(&sysctlmemlock, "sysctl mem", 0, LK_CANRECURSE); 528 SYSCTL_INIT(); 529 SYSCTL_XLOCK(); 530 SET_FOREACH(oidp, sysctl_set) 531 sysctl_register_oid(*oidp); 532 SYSCTL_XUNLOCK(); 533 } 534 SYSINIT(sysctl, SI_BOOT1_POST, SI_ORDER_ANY, sysctl_register_all, 0); 535 536 /* 537 * "Staff-functions" 538 * 539 * These functions implement a presently undocumented interface 540 * used by the sysctl program to walk the tree, and get the type 541 * so it can print the value. 542 * This interface is under work and consideration, and should probably 543 * be killed with a big axe by the first person who can find the time. 544 * (be aware though, that the proper interface isn't as obvious as it 545 * may seem, there are various conflicting requirements. 546 * 547 * {0,0} kprintf the entire MIB-tree. 548 * {0,1,...} return the name of the "..." OID. 549 * {0,2,...} return the next OID. 550 * {0,3} return the OID of the name in "new" 551 * {0,4,...} return the kind & format info for the "..." OID. 552 */ 553 554 static void 555 sysctl_sysctl_debug_dump_node(struct sysctl_oid_list *l, int i) 556 { 557 int k; 558 struct sysctl_oid *oidp; 559 560 SLIST_FOREACH(oidp, l, oid_link) { 561 562 for (k=0; k<i; k++) 563 kprintf(" "); 564 565 kprintf("%d %s ", oidp->oid_number, oidp->oid_name); 566 567 kprintf("%c%c", 568 oidp->oid_kind & CTLFLAG_RD ? 'R':' ', 569 oidp->oid_kind & CTLFLAG_WR ? 'W':' '); 570 571 if (oidp->oid_handler) 572 kprintf(" *Handler"); 573 574 switch (oidp->oid_kind & CTLTYPE) { 575 case CTLTYPE_NODE: 576 kprintf(" Node\n"); 577 if (!oidp->oid_handler) { 578 sysctl_sysctl_debug_dump_node( 579 oidp->oid_arg1, i+2); 580 } 581 break; 582 case CTLTYPE_INT: kprintf(" Int\n"); break; 583 case CTLTYPE_STRING: kprintf(" String\n"); break; 584 case CTLTYPE_QUAD: kprintf(" Quad\n"); break; 585 case CTLTYPE_OPAQUE: kprintf(" Opaque/struct\n"); break; 586 default: kprintf("\n"); 587 } 588 589 } 590 } 591 592 static int 593 sysctl_sysctl_debug(SYSCTL_HANDLER_ARGS) 594 { 595 int error; 596 597 error = priv_check(req->td, PRIV_SYSCTL_DEBUG); 598 if (error) 599 return (error); 600 sysctl_sysctl_debug_dump_node(&sysctl__children, 0); 601 602 return (ENOENT); 603 } 604 605 SYSCTL_PROC(_sysctl, 0, debug, CTLTYPE_STRING|CTLFLAG_RD, 606 0, 0, sysctl_sysctl_debug, "-", ""); 607 608 static int 609 sysctl_sysctl_name(SYSCTL_HANDLER_ARGS) 610 { 611 int *name = (int *) arg1; 612 u_int namelen = arg2; 613 int error = 0; 614 struct sysctl_oid *oid; 615 struct sysctl_oid_list *lsp = &sysctl__children, *lsp2; 616 char buf[10]; 617 618 while (namelen) { 619 if (!lsp) { 620 ksnprintf(buf, sizeof(buf), "%d", *name); 621 if (req->oldidx) 622 error = SYSCTL_OUT(req, ".", 1); 623 if (!error) 624 error = SYSCTL_OUT(req, buf, strlen(buf)); 625 if (error) 626 goto out; 627 namelen--; 628 name++; 629 continue; 630 } 631 lsp2 = NULL; 632 SLIST_FOREACH(oid, lsp, oid_link) { 633 if (oid->oid_number != *name) 634 continue; 635 636 if (req->oldidx) 637 error = SYSCTL_OUT(req, ".", 1); 638 if (!error) 639 error = SYSCTL_OUT(req, oid->oid_name, 640 strlen(oid->oid_name)); 641 if (error) 642 goto out; 643 644 namelen--; 645 name++; 646 647 if ((oid->oid_kind & CTLTYPE) != CTLTYPE_NODE) 648 break; 649 650 if (oid->oid_handler) 651 break; 652 653 lsp2 = SYSCTL_CHILDREN(oid); 654 break; 655 } 656 lsp = lsp2; 657 } 658 error = SYSCTL_OUT(req, "", 1); 659 out: 660 return (error); 661 } 662 663 SYSCTL_NODE(_sysctl, 1, name, CTLFLAG_RD, sysctl_sysctl_name, ""); 664 665 static int 666 sysctl_sysctl_next_ls(struct sysctl_oid_list *lsp, int *name, u_int namelen, 667 int *next, int *len, int level, struct sysctl_oid **oidpp) 668 { 669 struct sysctl_oid *oidp; 670 671 *len = level; 672 SLIST_FOREACH(oidp, lsp, oid_link) { 673 *next = oidp->oid_number; 674 *oidpp = oidp; 675 676 if (oidp->oid_kind & CTLFLAG_SKIP) 677 continue; 678 679 if (!namelen) { 680 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE) 681 return (0); 682 if (oidp->oid_handler) 683 /* We really should call the handler here...*/ 684 return (0); 685 lsp = SYSCTL_CHILDREN(oidp); 686 if (!sysctl_sysctl_next_ls(lsp, 0, 0, next+1, 687 len, level+1, oidpp)) 688 return (0); 689 goto emptynode; 690 } 691 692 if (oidp->oid_number < *name) 693 continue; 694 695 if (oidp->oid_number > *name) { 696 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE) 697 return (0); 698 if (oidp->oid_handler) 699 return (0); 700 lsp = SYSCTL_CHILDREN(oidp); 701 if (!sysctl_sysctl_next_ls(lsp, name+1, namelen-1, 702 next+1, len, level+1, oidpp)) 703 return (0); 704 goto next; 705 } 706 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE) 707 continue; 708 709 if (oidp->oid_handler) 710 continue; 711 712 lsp = SYSCTL_CHILDREN(oidp); 713 if (!sysctl_sysctl_next_ls(lsp, name+1, namelen-1, next+1, 714 len, level+1, oidpp)) 715 return (0); 716 next: 717 namelen = 1; 718 emptynode: 719 *len = level; 720 } 721 return (1); 722 } 723 724 static int 725 sysctl_sysctl_next(SYSCTL_HANDLER_ARGS) 726 { 727 int *name = (int *) arg1; 728 u_int namelen = arg2; 729 int i, j, error; 730 struct sysctl_oid *oid; 731 struct sysctl_oid_list *lsp = &sysctl__children; 732 int newoid[CTL_MAXNAME]; 733 734 i = sysctl_sysctl_next_ls(lsp, name, namelen, newoid, &j, 1, &oid); 735 if (i) 736 return ENOENT; 737 error = SYSCTL_OUT(req, newoid, j * sizeof (int)); 738 739 return (error); 740 } 741 742 SYSCTL_NODE(_sysctl, 2, next, CTLFLAG_RD, sysctl_sysctl_next, ""); 743 744 static int 745 name2oid(char *name, int *oid, int *len, struct sysctl_oid **oidpp) 746 { 747 struct sysctl_oid *oidp; 748 struct sysctl_oid_list *lsp = &sysctl__children; 749 char *p; 750 751 SYSCTL_ASSERT_LOCKED(); 752 753 for (*len = 0; *len < CTL_MAXNAME;) { 754 p = strsep(&name, "."); 755 756 oidp = SLIST_FIRST(lsp); 757 for (;; oidp = SLIST_NEXT(oidp, oid_link)) { 758 if (oidp == NULL) 759 return (ENOENT); 760 if (strcmp(p, oidp->oid_name) == 0) 761 break; 762 } 763 *oid++ = oidp->oid_number; 764 (*len)++; 765 766 if (name == NULL || *name == '\0') { 767 if (oidpp) 768 *oidpp = oidp; 769 return (0); 770 } 771 772 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE) 773 break; 774 775 if (oidp->oid_handler) 776 break; 777 778 lsp = SYSCTL_CHILDREN(oidp); 779 } 780 return (ENOENT); 781 } 782 783 static int 784 sysctl_sysctl_name2oid(SYSCTL_HANDLER_ARGS) 785 { 786 char *p; 787 int error, oid[CTL_MAXNAME], len; 788 struct sysctl_oid *op = NULL; 789 790 if (!req->newlen) 791 return ENOENT; 792 if (req->newlen >= MAXPATHLEN) /* XXX arbitrary, undocumented */ 793 return (ENAMETOOLONG); 794 795 p = kmalloc(req->newlen+1, M_SYSCTL, M_WAITOK); 796 797 error = SYSCTL_IN(req, p, req->newlen); 798 if (error) { 799 kfree(p, M_SYSCTL); 800 return (error); 801 } 802 803 p [req->newlen] = '\0'; 804 805 error = name2oid(p, oid, &len, &op); 806 807 kfree(p, M_SYSCTL); 808 809 if (error) 810 return (error); 811 812 error = SYSCTL_OUT(req, oid, len * sizeof *oid); 813 return (error); 814 } 815 816 SYSCTL_PROC(_sysctl, 3, name2oid, CTLFLAG_RW|CTLFLAG_ANYBODY|CTLFLAG_SHLOCK, 817 0, 0, sysctl_sysctl_name2oid, "I", ""); 818 819 static int 820 sysctl_sysctl_oidfmt(SYSCTL_HANDLER_ARGS) 821 { 822 struct sysctl_oid *oid; 823 int error; 824 825 error = sysctl_find_oid(arg1, arg2, &oid, NULL, req); 826 if (error) 827 return (error); 828 829 if (!oid->oid_fmt) 830 return (ENOENT); 831 error = SYSCTL_OUT(req, &oid->oid_kind, sizeof(oid->oid_kind)); 832 if (error) 833 return (error); 834 error = SYSCTL_OUT(req, oid->oid_fmt, strlen(oid->oid_fmt) + 1); 835 return (error); 836 } 837 838 839 SYSCTL_NODE(_sysctl, 4, oidfmt, CTLFLAG_RD, sysctl_sysctl_oidfmt, ""); 840 841 static int 842 sysctl_sysctl_oiddescr(SYSCTL_HANDLER_ARGS) 843 { 844 struct sysctl_oid *oid; 845 int error; 846 847 error = sysctl_find_oid(arg1, arg2, &oid, NULL, req); 848 if (error) 849 return (error); 850 851 if (!oid->oid_descr) 852 return (ENOENT); 853 error = SYSCTL_OUT(req, oid->oid_descr, strlen(oid->oid_descr) + 1); 854 return (error); 855 } 856 857 SYSCTL_NODE(_sysctl, 5, oiddescr, CTLFLAG_RD, sysctl_sysctl_oiddescr, ""); 858 859 /* 860 * Default "handler" functions. 861 */ 862 863 /* 864 * Handle an int, signed or unsigned. 865 * Two cases: 866 * a variable: point arg1 at it. 867 * a constant: pass it in arg2. 868 */ 869 870 int 871 sysctl_handle_int(SYSCTL_HANDLER_ARGS) 872 { 873 int error = 0; 874 875 if (arg1) 876 error = SYSCTL_OUT(req, arg1, sizeof(int)); 877 else 878 error = SYSCTL_OUT(req, &arg2, sizeof(int)); 879 880 if (error || !req->newptr) 881 return (error); 882 883 if (!arg1) 884 error = EPERM; 885 else 886 error = SYSCTL_IN(req, arg1, sizeof(int)); 887 return (error); 888 } 889 890 /* 891 * Handle a long, signed or unsigned. arg1 points to it. 892 */ 893 894 int 895 sysctl_handle_long(SYSCTL_HANDLER_ARGS) 896 { 897 int error = 0; 898 899 if (!arg1) 900 return (EINVAL); 901 error = SYSCTL_OUT(req, arg1, sizeof(long)); 902 903 if (error || !req->newptr) 904 return (error); 905 906 error = SYSCTL_IN(req, arg1, sizeof(long)); 907 return (error); 908 } 909 910 /* 911 * Handle a quad, signed or unsigned. arg1 points to it. 912 */ 913 914 int 915 sysctl_handle_quad(SYSCTL_HANDLER_ARGS) 916 { 917 int error = 0; 918 919 if (!arg1) 920 return (EINVAL); 921 error = SYSCTL_OUT(req, arg1, sizeof(quad_t)); 922 923 if (error || !req->newptr) 924 return (error); 925 926 error = SYSCTL_IN(req, arg1, sizeof(quad_t)); 927 return (error); 928 } 929 930 /* 931 * Handle our generic '\0' terminated 'C' string. 932 * Two cases: 933 * a variable string: point arg1 at it, arg2 is max length. 934 * a constant string: point arg1 at it, arg2 is zero. 935 */ 936 937 int 938 sysctl_handle_string(SYSCTL_HANDLER_ARGS) 939 { 940 int error=0; 941 942 error = SYSCTL_OUT(req, arg1, strlen((char *)arg1)+1); 943 944 if (error || !req->newptr) 945 return (error); 946 947 if ((req->newlen - req->newidx) >= arg2) { 948 error = EINVAL; 949 } else { 950 arg2 = (req->newlen - req->newidx); 951 error = SYSCTL_IN(req, arg1, arg2); 952 ((char *)arg1)[arg2] = '\0'; 953 } 954 955 return (error); 956 } 957 958 /* 959 * Handle any kind of opaque data. 960 * arg1 points to it, arg2 is the size. 961 */ 962 963 int 964 sysctl_handle_opaque(SYSCTL_HANDLER_ARGS) 965 { 966 int error; 967 968 error = SYSCTL_OUT(req, arg1, arg2); 969 970 if (error || !req->newptr) 971 return (error); 972 973 error = SYSCTL_IN(req, arg1, arg2); 974 975 return (error); 976 } 977 978 /* 979 * Transfer functions to/from kernel space. 980 * XXX: rather untested at this point 981 */ 982 static int 983 sysctl_old_kernel(struct sysctl_req *req, const void *p, size_t l) 984 { 985 size_t i = 0; 986 987 if (req->oldptr) { 988 i = l; 989 if (i > req->oldlen - req->oldidx) 990 i = req->oldlen - req->oldidx; 991 if (i > 0) 992 bcopy(p, (char *)req->oldptr + req->oldidx, i); 993 } 994 req->oldidx += l; 995 if (req->oldptr && i != l) 996 return (ENOMEM); 997 return (0); 998 } 999 1000 static int 1001 sysctl_new_kernel(struct sysctl_req *req, void *p, size_t l) 1002 { 1003 1004 if (!req->newptr) 1005 return 0; 1006 if (req->newlen - req->newidx < l) 1007 return (EINVAL); 1008 bcopy((char *)req->newptr + req->newidx, p, l); 1009 req->newidx += l; 1010 return (0); 1011 } 1012 1013 int 1014 kernel_sysctl(int *name, u_int namelen, 1015 void *old, size_t *oldlenp, 1016 void *new, size_t newlen, size_t *retval) 1017 { 1018 int error = 0; 1019 struct sysctl_req req; 1020 1021 bzero(&req, sizeof req); 1022 1023 req.td = curthread; 1024 1025 if (oldlenp) { 1026 req.oldlen = *oldlenp; 1027 } 1028 req.validlen = req.oldlen; 1029 1030 if (old) { 1031 req.oldptr= old; 1032 } 1033 1034 if (new != NULL) { 1035 req.newlen = newlen; 1036 req.newptr = new; 1037 } 1038 1039 req.oldfunc = sysctl_old_kernel; 1040 req.newfunc = sysctl_new_kernel; 1041 #if 0 1042 req.lock = REQ_UNWIRED; 1043 #endif 1044 1045 SYSCTL_SLOCK(); 1046 error = sysctl_root(0, name, namelen, &req); 1047 SYSCTL_SUNLOCK(); 1048 1049 #if 0 1050 if (req.lock == REQ_WIRED && req.validlen > 0) 1051 vsunlock(req.oldptr, req.validlen); 1052 #endif 1053 1054 if (error && error != ENOMEM) 1055 return (error); 1056 1057 if (retval) { 1058 if (req.oldptr && req.oldidx > req.validlen) 1059 *retval = req.validlen; 1060 else 1061 *retval = req.oldidx; 1062 } 1063 return (error); 1064 } 1065 1066 int 1067 kernel_sysctlbyname(char *name, 1068 void *old, size_t *oldlenp, 1069 void *new, size_t newlen, size_t *retval) 1070 { 1071 int oid[CTL_MAXNAME]; 1072 size_t oidlen, plen; 1073 int error; 1074 1075 oid[0] = 0; /* sysctl internal magic */ 1076 oid[1] = 3; /* name2oid */ 1077 oidlen = sizeof(oid); 1078 1079 error = kernel_sysctl(oid, 2, oid, &oidlen, name, strlen(name), &plen); 1080 if (error) 1081 return (error); 1082 1083 error = kernel_sysctl(oid, plen / sizeof(int), old, oldlenp, 1084 new, newlen, retval); 1085 return (error); 1086 } 1087 1088 /* 1089 * Transfer function to/from user space. 1090 */ 1091 static int 1092 sysctl_old_user(struct sysctl_req *req, const void *p, size_t l) 1093 { 1094 int error = 0; 1095 size_t i = 0; 1096 1097 #if 0 1098 if (req->lock == 1 && req->oldptr) { 1099 vslock(req->oldptr, req->oldlen); 1100 req->lock = 2; 1101 } 1102 #endif 1103 if (req->oldptr) { 1104 i = l; 1105 if (i > req->oldlen - req->oldidx) 1106 i = req->oldlen - req->oldidx; 1107 if (i > 0) 1108 error = copyout(p, (char *)req->oldptr + req->oldidx, 1109 i); 1110 } 1111 req->oldidx += l; 1112 if (error) 1113 return (error); 1114 if (req->oldptr && i < l) 1115 return (ENOMEM); 1116 return (0); 1117 } 1118 1119 static int 1120 sysctl_new_user(struct sysctl_req *req, void *p, size_t l) 1121 { 1122 int error; 1123 1124 if (!req->newptr) 1125 return 0; 1126 if (req->newlen - req->newidx < l) 1127 return (EINVAL); 1128 error = copyin((char *)req->newptr + req->newidx, p, l); 1129 req->newidx += l; 1130 return (error); 1131 } 1132 1133 int 1134 sysctl_find_oid(int *name, u_int namelen, struct sysctl_oid **noid, 1135 int *nindx, struct sysctl_req *req) 1136 { 1137 struct sysctl_oid_list *lsp; 1138 struct sysctl_oid *oid; 1139 int indx; 1140 1141 lsp = &sysctl__children; 1142 indx = 0; 1143 while (indx < CTL_MAXNAME) { 1144 SLIST_FOREACH(oid, lsp, oid_link) { 1145 if (oid->oid_number == name[indx]) 1146 break; 1147 } 1148 if (oid == NULL) 1149 return (ENOENT); 1150 1151 indx++; 1152 if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) { 1153 if (oid->oid_handler != NULL || indx == namelen) { 1154 *noid = oid; 1155 if (nindx != NULL) 1156 *nindx = indx; 1157 KASSERT((oid->oid_kind & CTLFLAG_DYING) == 0, 1158 ("%s found DYING node %p", __func__, oid)); 1159 return (0); 1160 } 1161 lsp = SYSCTL_CHILDREN(oid); 1162 } else if (indx == namelen) { 1163 *noid = oid; 1164 if (nindx != NULL) 1165 *nindx = indx; 1166 KASSERT((oid->oid_kind & CTLFLAG_DYING) == 0, 1167 ("%s found DYING node %p", __func__, oid)); 1168 return (0); 1169 } else { 1170 return (ENOTDIR); 1171 } 1172 } 1173 return (ENOENT); 1174 } 1175 1176 /* 1177 * Traverse our tree, and find the right node, execute whatever it points 1178 * to, and return the resulting error code. 1179 */ 1180 static int 1181 sysctl_root(SYSCTL_HANDLER_ARGS) 1182 { 1183 struct thread *td = req->td; 1184 struct proc *p = td ? td->td_proc : NULL; 1185 struct sysctl_oid *oid; 1186 int error, indx; 1187 int lktype; 1188 1189 error = sysctl_find_oid(arg1, arg2, &oid, &indx, req); 1190 if (error) 1191 return (error); 1192 1193 if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) { 1194 /* 1195 * You can't call a sysctl when it's a node, but has 1196 * no handler. Inform the user that it's a node. 1197 * The indx may or may not be the same as namelen. 1198 */ 1199 if (oid->oid_handler == NULL) 1200 return (EISDIR); 1201 } 1202 1203 /* If writing isn't allowed */ 1204 if (req->newptr && (!(oid->oid_kind & CTLFLAG_WR) || 1205 ((oid->oid_kind & CTLFLAG_SECURE) && securelevel > 0))) 1206 return (EPERM); 1207 1208 /* Most likely only root can write */ 1209 if (!(oid->oid_kind & CTLFLAG_ANYBODY) && req->newptr && p && 1210 (error = priv_check_cred(td->td_ucred, 1211 (oid->oid_kind & CTLFLAG_PRISON) ? PRIV_SYSCTL_WRITEJAIL : 1212 PRIV_SYSCTL_WRITE, 0))) 1213 return (error); 1214 1215 if (oid->oid_handler == NULL) 1216 return EINVAL; 1217 1218 /* 1219 * Default oid locking is exclusive when modifying (newptr), 1220 * shared otherwise, unless overridden with a control flag. 1221 */ 1222 lktype = (req->newptr != NULL) ? LK_EXCLUSIVE : LK_SHARED; 1223 if (oid->oid_kind & CTLFLAG_SHLOCK) 1224 lktype = LK_SHARED; 1225 if (oid->oid_kind & CTLFLAG_EXLOCK) 1226 lktype = LK_EXCLUSIVE; 1227 lockmgr(&oid->oid_lock, lktype); 1228 1229 if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) 1230 error = oid->oid_handler(oid, (int *)arg1 + indx, arg2 - indx, 1231 req); 1232 else 1233 error = oid->oid_handler(oid, oid->oid_arg1, oid->oid_arg2, 1234 req); 1235 lockmgr(&oid->oid_lock, LK_RELEASE); 1236 1237 return (error); 1238 } 1239 1240 int 1241 sys___sysctl(struct sysctl_args *uap) 1242 { 1243 int error, i, name[CTL_MAXNAME]; 1244 size_t j; 1245 1246 if (uap->namelen > CTL_MAXNAME || uap->namelen < 2) 1247 return (EINVAL); 1248 1249 error = copyin(uap->name, &name, uap->namelen * sizeof(int)); 1250 if (error) 1251 return (error); 1252 1253 error = userland_sysctl(name, uap->namelen, 1254 uap->old, uap->oldlenp, 0, 1255 uap->new, uap->newlen, &j); 1256 if (error && error != ENOMEM) 1257 return (error); 1258 if (uap->oldlenp) { 1259 i = copyout(&j, uap->oldlenp, sizeof(j)); 1260 if (i) 1261 return (i); 1262 } 1263 return (error); 1264 } 1265 1266 /* 1267 * This is used from various compatibility syscalls too. That's why name 1268 * must be in kernel space. 1269 */ 1270 int 1271 userland_sysctl(int *name, u_int namelen, 1272 void *old, size_t *oldlenp, int inkernel, 1273 void *new, size_t newlen, size_t *retval) 1274 { 1275 int error = 0, memlocked; 1276 struct sysctl_req req; 1277 1278 bzero(&req, sizeof req); 1279 1280 req.td = curthread; 1281 req.flags = 0; 1282 1283 if (oldlenp) { 1284 if (inkernel) { 1285 req.oldlen = *oldlenp; 1286 } else { 1287 error = copyin(oldlenp, &req.oldlen, sizeof(*oldlenp)); 1288 if (error) 1289 return (error); 1290 } 1291 } 1292 req.validlen = req.oldlen; 1293 1294 if (old) { 1295 if (!useracc(old, req.oldlen, VM_PROT_WRITE)) 1296 return (EFAULT); 1297 req.oldptr= old; 1298 } 1299 1300 if (new != NULL) { 1301 if (!useracc(new, newlen, VM_PROT_READ)) 1302 return (EFAULT); 1303 req.newlen = newlen; 1304 req.newptr = new; 1305 } 1306 1307 req.oldfunc = sysctl_old_user; 1308 req.newfunc = sysctl_new_user; 1309 #if 0 1310 req.lock = REQ_UNWIRED; 1311 #endif 1312 1313 #ifdef KTRACE 1314 if (KTRPOINT(curthread, KTR_SYSCTL)) 1315 ktrsysctl(name, namelen); 1316 #endif 1317 1318 if (req.oldlen > PAGE_SIZE) { 1319 memlocked = 1; 1320 lockmgr(&sysctlmemlock, LK_EXCLUSIVE); 1321 } else 1322 memlocked = 0; 1323 1324 for (;;) { 1325 req.oldidx = 0; 1326 req.newidx = 0; 1327 SYSCTL_SLOCK(); 1328 error = sysctl_root(0, name, namelen, &req); 1329 SYSCTL_SUNLOCK(); 1330 if (error != EAGAIN) 1331 break; 1332 lwkt_yield(); 1333 } 1334 1335 #if 0 1336 if (req.lock == REQ_WIRED && req.validlen > 0) 1337 vsunlock(req.oldptr, req.validlen); 1338 #endif 1339 if (memlocked) 1340 lockmgr(&sysctlmemlock, LK_RELEASE); 1341 1342 if (error && error != ENOMEM) 1343 return (error); 1344 1345 if (retval) { 1346 if (req.oldptr && req.oldidx > req.validlen) 1347 *retval = req.validlen; 1348 else 1349 *retval = req.oldidx; 1350 } 1351 return (error); 1352 } 1353 1354 int 1355 sysctl_int_range(SYSCTL_HANDLER_ARGS, int low, int high) 1356 { 1357 int error, value; 1358 1359 value = *(int *)arg1; 1360 error = sysctl_handle_int(oidp, &value, 0, req); 1361 if (error || !req->newptr) 1362 return (error); 1363 if (value < low || value > high) 1364 return (EINVAL); 1365 *(int *)arg1 = value; 1366 return (0); 1367 } 1368 1369 /* 1370 * Drain into a sysctl struct. The user buffer should be wired if a page 1371 * fault would cause issue. 1372 */ 1373 static int 1374 sbuf_sysctl_drain(void *arg, const char *data, int len) 1375 { 1376 struct sysctl_req *req = arg; 1377 int error; 1378 1379 error = SYSCTL_OUT(req, data, len); 1380 KASSERT(error >= 0, ("Got unexpected negative value %d", error)); 1381 return (error == 0 ? len : -error); 1382 } 1383 1384 struct sbuf * 1385 sbuf_new_for_sysctl(struct sbuf *s, char *buf, int length, 1386 struct sysctl_req *req) 1387 { 1388 1389 s = sbuf_new(s, buf, length, SBUF_FIXEDLEN); 1390 sbuf_set_drain(s, sbuf_sysctl_drain, req); 1391 return (s); 1392 } 1393