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 <sys/mplock2.h> 52 53 #include <vm/vm.h> 54 #include <vm/vm_extern.h> 55 56 static MALLOC_DEFINE(M_SYSCTL, "sysctl", "sysctl internal magic"); 57 static MALLOC_DEFINE(M_SYSCTLOID, "sysctloid", "sysctl dynamic oids"); 58 59 /* 60 * The sysctllock protects the MIB tree. It also protects sysctl 61 * contexts used with dynamic sysctls. The sysctl_register_oid() and 62 * sysctl_unregister_oid() routines require the sysctllock to already 63 * be held, so the sysctl_lock() and sysctl_unlock() routines are 64 * provided for the few places in the kernel which need to use that 65 * API rather than using the dynamic API. Use of the dynamic API is 66 * strongly encouraged for most code. 67 * 68 * The sysctlmemlock is used to limit the amount of user memory wired for 69 * sysctl requests. This is implemented by serializing any userland 70 * sysctl requests larger than a single page via an exclusive lock. 71 */ 72 struct lock sysctllock; 73 static struct lock sysctlmemlock; 74 75 #define SYSCTL_INIT() lockinit(&sysctllock, \ 76 "sysctl lock", 0, LK_CANRECURSE) 77 #define SYSCTL_SLEEP(ch, wmesg, timo) \ 78 lksleep(ch, &sysctllock, 0, wmesg, timo) 79 80 static int sysctl_root(SYSCTL_HANDLER_ARGS); 81 static void sysctl_register_oid_int(struct sysctl_oid *oipd); 82 static void sysctl_unregister_oid_int(struct sysctl_oid *oipd); 83 84 struct sysctl_oid_list sysctl__children; /* root list */ 85 86 static int sysctl_remove_oid_locked(struct sysctl_oid *oidp, int del, 87 int recurse); 88 89 static struct sysctl_oid * 90 sysctl_find_oidname(const char *name, struct sysctl_oid_list *list, int lock) 91 { 92 struct sysctl_oid *oidp; 93 94 SLIST_FOREACH(oidp, list, oid_link) { 95 if (strcmp(oidp->oid_name, name) == 0) { 96 break; 97 } 98 } 99 return (oidp); 100 } 101 102 /* 103 * Initialization of the MIB tree. 104 * 105 * Order by number in each list. 106 */ 107 108 void 109 sysctl_register_oid(struct sysctl_oid *oidp) 110 { 111 SYSCTL_XLOCK(); 112 sysctl_register_oid_int(oidp); 113 SYSCTL_XUNLOCK(); 114 } 115 116 static void 117 sysctl_register_oid_int(struct sysctl_oid *oidp) 118 { 119 struct sysctl_oid_list *parent = oidp->oid_parent; 120 struct sysctl_oid *p; 121 struct sysctl_oid *q; 122 123 /* 124 * First check if another oid with the same name already 125 * exists in the parent's list. 126 */ 127 p = sysctl_find_oidname(oidp->oid_name, parent, 0); 128 if (p != NULL) { 129 if ((p->oid_kind & CTLTYPE) == CTLTYPE_NODE) 130 p->oid_refcnt++; 131 else 132 kprintf("can't re-use a leaf (%s)!\n", p->oid_name); 133 return; 134 } 135 136 /* 137 * If this oid has a number OID_AUTO, give it a number which 138 * is greater than any current oid. Make sure it is at least 139 * 256 to leave space for pre-assigned oid numbers. 140 */ 141 if (oidp->oid_number == OID_AUTO) { 142 int newoid = 0x100; /* minimum AUTO oid */ 143 144 /* 145 * Adjust based on highest oid in parent list 146 */ 147 SLIST_FOREACH(p, parent, oid_link) { 148 if (newoid <= p->oid_number) 149 newoid = p->oid_number + 1; 150 } 151 oidp->oid_number = newoid; 152 } 153 154 /* 155 * Insert the oid into the parent's list in order. 156 */ 157 q = NULL; 158 SLIST_FOREACH(p, parent, oid_link) { 159 if (oidp->oid_number < p->oid_number) 160 break; 161 q = p; 162 } 163 if (q) 164 SLIST_INSERT_AFTER(q, oidp, oid_link); 165 else 166 SLIST_INSERT_HEAD(parent, oidp, oid_link); 167 } 168 169 void 170 sysctl_unregister_oid(struct sysctl_oid *oidp) 171 { 172 SYSCTL_XLOCK(); 173 sysctl_unregister_oid_int(oidp); 174 SYSCTL_XUNLOCK(); 175 } 176 177 static void 178 sysctl_unregister_oid_int(struct sysctl_oid *oidp) 179 { 180 struct sysctl_oid *p; 181 182 if (oidp->oid_number == OID_AUTO) 183 panic("Trying to unregister OID_AUTO entry: %p", oidp); 184 185 SLIST_FOREACH(p, oidp->oid_parent, oid_link) { 186 if (p != oidp) 187 continue; 188 SLIST_REMOVE(oidp->oid_parent, oidp, sysctl_oid, oid_link); 189 return; 190 } 191 192 /* 193 * This can happen when a module fails to register and is 194 * being unloaded afterwards. It should not be a panic() 195 * for normal use. 196 */ 197 kprintf("%s: failed to unregister sysctl\n", __func__); 198 } 199 200 /* Initialize a new context to keep track of dynamically added sysctls. */ 201 int 202 sysctl_ctx_init(struct sysctl_ctx_list *c) 203 { 204 if (c == NULL) 205 return(EINVAL); 206 TAILQ_INIT(c); 207 return(0); 208 } 209 210 /* Free the context, and destroy all dynamic oids registered in this context */ 211 int 212 sysctl_ctx_free(struct sysctl_ctx_list *clist) 213 { 214 struct sysctl_ctx_entry *e, *e1; 215 int error; 216 217 error = 0; 218 /* 219 * First perform a "dry run" to check if it's ok to remove oids. 220 * XXX FIXME 221 * XXX This algorithm is a hack. But I don't know any 222 * XXX better solution for now... 223 */ 224 SYSCTL_XLOCK(); 225 TAILQ_FOREACH(e, clist, link) { 226 error = sysctl_remove_oid_locked(e->entry, 0, 0); 227 if (error) 228 break; 229 } 230 /* 231 * Restore deregistered entries, either from the end, 232 * or from the place where error occured. 233 * e contains the entry that was not unregistered 234 */ 235 if (error) 236 e1 = TAILQ_PREV(e, sysctl_ctx_list, link); 237 else 238 e1 = TAILQ_LAST(clist, sysctl_ctx_list); 239 while (e1 != NULL) { 240 sysctl_register_oid(e1->entry); 241 e1 = TAILQ_PREV(e1, sysctl_ctx_list, link); 242 } 243 if (error) { 244 SYSCTL_XUNLOCK(); 245 return(EBUSY); 246 } 247 /* Now really delete the entries */ 248 e = TAILQ_FIRST(clist); 249 while (e != NULL) { 250 e1 = TAILQ_NEXT(e, link); 251 error = sysctl_remove_oid_locked(e->entry, 1, 0); 252 if (error) 253 panic("sysctl_remove_oid: corrupt tree, entry: %s", 254 e->entry->oid_name); 255 kfree(e, M_SYSCTLOID); 256 e = e1; 257 } 258 SYSCTL_XUNLOCK(); 259 return (error); 260 } 261 262 /* Add an entry to the context */ 263 struct sysctl_ctx_entry * 264 sysctl_ctx_entry_add(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp) 265 { 266 struct sysctl_ctx_entry *e; 267 268 SYSCTL_ASSERT_XLOCKED(); 269 if (clist == NULL || oidp == NULL) 270 return(NULL); 271 e = kmalloc(sizeof(struct sysctl_ctx_entry), M_SYSCTLOID, M_WAITOK); 272 e->entry = oidp; 273 TAILQ_INSERT_HEAD(clist, e, link); 274 return (e); 275 } 276 277 /* Find an entry in the context */ 278 struct sysctl_ctx_entry * 279 sysctl_ctx_entry_find(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp) 280 { 281 struct sysctl_ctx_entry *e; 282 283 SYSCTL_ASSERT_XLOCKED(); 284 if (clist == NULL || oidp == NULL) 285 return(NULL); 286 TAILQ_FOREACH(e, clist, link) { 287 if(e->entry == oidp) 288 return(e); 289 } 290 return (e); 291 } 292 293 /* 294 * Delete an entry from the context. 295 * NOTE: this function doesn't free oidp! You have to remove it 296 * with sysctl_remove_oid(). 297 */ 298 int 299 sysctl_ctx_entry_del(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp) 300 { 301 struct sysctl_ctx_entry *e; 302 303 if (clist == NULL || oidp == NULL) 304 return (EINVAL); 305 SYSCTL_XLOCK(); 306 e = sysctl_ctx_entry_find(clist, oidp); 307 if (e != NULL) { 308 TAILQ_REMOVE(clist, e, link); 309 SYSCTL_XUNLOCK(); 310 kfree(e, M_SYSCTLOID); 311 return (0); 312 } else { 313 SYSCTL_XUNLOCK(); 314 return (ENOENT); 315 } 316 } 317 318 /* 319 * Remove dynamically created sysctl trees. 320 * oidp - top of the tree to be removed 321 * del - if 0 - just deregister, otherwise free up entries as well 322 * recurse - if != 0 traverse the subtree to be deleted 323 */ 324 int 325 sysctl_remove_oid(struct sysctl_oid *oidp, int del, int recurse) 326 { 327 int error; 328 329 SYSCTL_XLOCK(); 330 error = sysctl_remove_oid_locked(oidp, del, recurse); 331 SYSCTL_XUNLOCK(); 332 return (error); 333 } 334 335 static int 336 sysctl_remove_oid_locked(struct sysctl_oid *oidp, int del, int recurse) 337 { 338 struct sysctl_oid *p, *tmp; 339 int error; 340 341 SYSCTL_ASSERT_XLOCKED(); 342 if (oidp == NULL) 343 return(EINVAL); 344 if ((oidp->oid_kind & CTLFLAG_DYN) == 0) { 345 kprintf("can't remove non-dynamic nodes!\n"); 346 return (EINVAL); 347 } 348 /* 349 * WARNING: normal method to do this should be through 350 * sysctl_ctx_free(). Use recursing as the last resort 351 * method to purge your sysctl tree of leftovers... 352 * However, if some other code still references these nodes, 353 * it will panic. 354 */ 355 if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) { 356 if (oidp->oid_refcnt == 1) { 357 SLIST_FOREACH_MUTABLE(p, 358 SYSCTL_CHILDREN(oidp), oid_link, tmp) { 359 if (!recurse) { 360 kprintf("Warning: failed attempt to " 361 "remove oid %s with child %s\n", 362 oidp->oid_name, p->oid_name); 363 return (ENOTEMPTY); 364 } 365 error = sysctl_remove_oid_locked(p, del, 366 recurse); 367 if (error) 368 return (error); 369 } 370 if (del) 371 kfree(SYSCTL_CHILDREN(oidp), M_SYSCTLOID); 372 } 373 } 374 if (oidp->oid_refcnt > 1 ) { 375 oidp->oid_refcnt--; 376 } else { 377 if (oidp->oid_refcnt == 0) { 378 kprintf("Warning: bad oid_refcnt=%u (%s)!\n", 379 oidp->oid_refcnt, oidp->oid_name); 380 return (EINVAL); 381 } 382 sysctl_unregister_oid(oidp); 383 if (del) { 384 /* 385 * Wait for all threads running the handler to drain. 386 * This preserves the previous behavior when the 387 * sysctl lock was held across a handler invocation, 388 * and is necessary for module unload correctness. 389 */ 390 while (oidp->oid_running > 0) { 391 oidp->oid_kind |= CTLFLAG_DYING; 392 SYSCTL_SLEEP(&oidp->oid_running, "oidrm", 0); 393 } 394 if (oidp->oid_descr) 395 kfree(__DECONST(char *, oidp->oid_descr), 396 M_SYSCTLOID); 397 kfree(__DECONST(char *, oidp->oid_name), M_SYSCTLOID); 398 kfree(oidp, M_SYSCTLOID); 399 } 400 } 401 return (0); 402 } 403 404 int 405 sysctl_remove_name(struct sysctl_oid *parent, const char *name, 406 int del, int recurse) 407 { 408 struct sysctl_oid *p, *tmp; 409 int error; 410 411 error = ENOENT; 412 SYSCTL_XLOCK(); 413 SLIST_FOREACH_MUTABLE(p, SYSCTL_CHILDREN(parent), oid_link, tmp) { 414 if (strcmp(p->oid_name, name) == 0) { 415 error = sysctl_remove_oid_locked(p, del, recurse); 416 break; 417 } 418 } 419 SYSCTL_XUNLOCK(); 420 421 return (error); 422 } 423 424 /* 425 * Create new sysctls at run time. 426 * clist may point to a valid context initialized with sysctl_ctx_init(). 427 */ 428 struct sysctl_oid * 429 sysctl_add_oid(struct sysctl_ctx_list *clist, struct sysctl_oid_list *parent, 430 int number, const char *name, int kind, void *arg1, int arg2, 431 int (*handler)(SYSCTL_HANDLER_ARGS), const char *fmt, const char *descr) 432 { 433 struct sysctl_oid *oidp; 434 ssize_t len; 435 char *newname; 436 437 /* You have to hook up somewhere.. */ 438 if (parent == NULL) 439 return(NULL); 440 SYSCTL_XLOCK(); 441 /* Check if the node already exists, otherwise create it */ 442 oidp = sysctl_find_oidname(name, parent, 0); 443 if (oidp != NULL) { 444 if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) { 445 oidp->oid_refcnt++; 446 /* Update the context */ 447 if (clist != NULL) 448 sysctl_ctx_entry_add(clist, oidp); 449 SYSCTL_XUNLOCK(); 450 return (oidp); 451 } else { 452 kprintf("can't re-use a leaf (%s)!\n", name); 453 SYSCTL_XUNLOCK(); 454 return (NULL); 455 } 456 } 457 oidp = kmalloc(sizeof(struct sysctl_oid), M_SYSCTLOID, M_WAITOK | M_ZERO); 458 oidp->oid_parent = parent; 459 SLIST_NEXT(oidp, oid_link) = NULL; 460 oidp->oid_number = number; 461 oidp->oid_refcnt = 1; 462 len = strlen(name); 463 newname = kmalloc(len + 1, M_SYSCTLOID, M_WAITOK); 464 bcopy(name, newname, len + 1); 465 newname[len] = '\0'; 466 oidp->oid_name = newname; 467 oidp->oid_handler = handler; 468 oidp->oid_kind = CTLFLAG_DYN | kind; 469 if ((kind & CTLTYPE) == CTLTYPE_NODE) { 470 struct sysctl_oid_list *children; 471 472 /* Allocate space for children */ 473 children = kmalloc(sizeof(*children), M_SYSCTLOID, M_WAITOK); 474 SYSCTL_SET_CHILDREN(oidp, children); 475 SLIST_INIT(children); 476 } else { 477 oidp->oid_arg1 = arg1; 478 oidp->oid_arg2 = arg2; 479 } 480 oidp->oid_fmt = fmt; 481 if (descr) { 482 int len = strlen(descr) + 1; 483 oidp->oid_descr = kmalloc(len, M_SYSCTLOID, M_WAITOK); 484 strcpy((char *)(uintptr_t)(const void *)oidp->oid_descr, descr); 485 }; 486 /* Update the context, if used */ 487 if (clist != NULL) 488 sysctl_ctx_entry_add(clist, oidp); 489 /* Register this oid */ 490 sysctl_register_oid_int(oidp); 491 SYSCTL_XUNLOCK(); 492 return (oidp); 493 } 494 495 /* 496 * Rename an existing oid. 497 */ 498 void 499 sysctl_rename_oid(struct sysctl_oid *oidp, const char *name) 500 { 501 char *newname; 502 char *oldname; 503 504 newname = kstrdup(name, M_SYSCTLOID); 505 SYSCTL_XLOCK(); 506 oldname = __DECONST(char *, oidp->oid_name); 507 oidp->oid_name = newname; 508 SYSCTL_XUNLOCK(); 509 kfree(oldname, M_SYSCTLOID); 510 } 511 512 /* 513 * Register the kernel's oids on startup. 514 */ 515 SET_DECLARE(sysctl_set, struct sysctl_oid); 516 517 static void 518 sysctl_register_all(void *arg) 519 { 520 struct sysctl_oid **oidp; 521 522 lockinit(&sysctlmemlock, "sysctl mem", 0, LK_CANRECURSE); 523 SYSCTL_INIT(); 524 SYSCTL_XLOCK(); 525 SET_FOREACH(oidp, sysctl_set) 526 sysctl_register_oid(*oidp); 527 SYSCTL_XUNLOCK(); 528 } 529 SYSINIT(sysctl, SI_BOOT1_POST, SI_ORDER_ANY, sysctl_register_all, 0); 530 531 /* 532 * "Staff-functions" 533 * 534 * These functions implement a presently undocumented interface 535 * used by the sysctl program to walk the tree, and get the type 536 * so it can print the value. 537 * This interface is under work and consideration, and should probably 538 * be killed with a big axe by the first person who can find the time. 539 * (be aware though, that the proper interface isn't as obvious as it 540 * may seem, there are various conflicting requirements. 541 * 542 * {0,0} kprintf the entire MIB-tree. 543 * {0,1,...} return the name of the "..." OID. 544 * {0,2,...} return the next OID. 545 * {0,3} return the OID of the name in "new" 546 * {0,4,...} return the kind & format info for the "..." OID. 547 */ 548 549 static void 550 sysctl_sysctl_debug_dump_node(struct sysctl_oid_list *l, int i) 551 { 552 int k; 553 struct sysctl_oid *oidp; 554 555 SYSCTL_ASSERT_XLOCKED(); 556 SLIST_FOREACH(oidp, l, oid_link) { 557 558 for (k=0; k<i; k++) 559 kprintf(" "); 560 561 kprintf("%d %s ", oidp->oid_number, oidp->oid_name); 562 563 kprintf("%c%c", 564 oidp->oid_kind & CTLFLAG_RD ? 'R':' ', 565 oidp->oid_kind & CTLFLAG_WR ? 'W':' '); 566 567 if (oidp->oid_handler) 568 kprintf(" *Handler"); 569 570 switch (oidp->oid_kind & CTLTYPE) { 571 case CTLTYPE_NODE: 572 kprintf(" Node\n"); 573 if (!oidp->oid_handler) { 574 sysctl_sysctl_debug_dump_node( 575 oidp->oid_arg1, i+2); 576 } 577 break; 578 case CTLTYPE_INT: kprintf(" Int\n"); break; 579 case CTLTYPE_STRING: kprintf(" String\n"); break; 580 case CTLTYPE_QUAD: kprintf(" Quad\n"); break; 581 case CTLTYPE_OPAQUE: kprintf(" Opaque/struct\n"); break; 582 default: kprintf("\n"); 583 } 584 585 } 586 } 587 588 static int 589 sysctl_sysctl_debug(SYSCTL_HANDLER_ARGS) 590 { 591 int error; 592 593 error = priv_check(req->td, PRIV_SYSCTL_DEBUG); 594 if (error) 595 return (error); 596 SYSCTL_XLOCK(); 597 sysctl_sysctl_debug_dump_node(&sysctl__children, 0); 598 SYSCTL_XUNLOCK(); 599 return (ENOENT); 600 } 601 602 SYSCTL_PROC(_sysctl, 0, debug, CTLTYPE_STRING|CTLFLAG_RD, 603 0, 0, sysctl_sysctl_debug, "-", ""); 604 605 static int 606 sysctl_sysctl_name(SYSCTL_HANDLER_ARGS) 607 { 608 int *name = (int *) arg1; 609 u_int namelen = arg2; 610 int error = 0; 611 struct sysctl_oid *oid; 612 struct sysctl_oid_list *lsp = &sysctl__children, *lsp2; 613 char buf[10]; 614 615 SYSCTL_XLOCK(); 616 while (namelen) { 617 if (!lsp) { 618 ksnprintf(buf, sizeof(buf), "%d", *name); 619 if (req->oldidx) 620 error = SYSCTL_OUT(req, ".", 1); 621 if (!error) 622 error = SYSCTL_OUT(req, buf, strlen(buf)); 623 if (error) 624 goto out; 625 namelen--; 626 name++; 627 continue; 628 } 629 lsp2 = NULL; 630 SLIST_FOREACH(oid, lsp, oid_link) { 631 if (oid->oid_number != *name) 632 continue; 633 634 if (req->oldidx) 635 error = SYSCTL_OUT(req, ".", 1); 636 if (!error) 637 error = SYSCTL_OUT(req, oid->oid_name, 638 strlen(oid->oid_name)); 639 if (error) 640 goto out; 641 642 namelen--; 643 name++; 644 645 if ((oid->oid_kind & CTLTYPE) != CTLTYPE_NODE) 646 break; 647 648 if (oid->oid_handler) 649 break; 650 651 lsp2 = SYSCTL_CHILDREN(oid); 652 break; 653 } 654 lsp = lsp2; 655 } 656 error = SYSCTL_OUT(req, "", 1); 657 out: 658 SYSCTL_XUNLOCK(); 659 return (error); 660 } 661 662 SYSCTL_NODE(_sysctl, 1, name, CTLFLAG_RD, sysctl_sysctl_name, ""); 663 664 static int 665 sysctl_sysctl_next_ls(struct sysctl_oid_list *lsp, int *name, u_int namelen, 666 int *next, int *len, int level, struct sysctl_oid **oidpp) 667 { 668 struct sysctl_oid *oidp; 669 670 SYSCTL_ASSERT_XLOCKED(); 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 return (error); 739 } 740 741 SYSCTL_NODE(_sysctl, 2, next, CTLFLAG_RD, sysctl_sysctl_next, ""); 742 743 static int 744 name2oid(char *name, int *oid, int *len, struct sysctl_oid **oidpp) 745 { 746 struct sysctl_oid *oidp; 747 struct sysctl_oid_list *lsp = &sysctl__children; 748 char *p; 749 750 SYSCTL_ASSERT_XLOCKED(); 751 752 for (*len = 0; *len < CTL_MAXNAME;) { 753 p = strsep(&name, "."); 754 755 oidp = SLIST_FIRST(lsp); 756 for (;; oidp = SLIST_NEXT(oidp, oid_link)) { 757 if (oidp == NULL) 758 return (ENOENT); 759 if (strcmp(p, oidp->oid_name) == 0) 760 break; 761 } 762 *oid++ = oidp->oid_number; 763 (*len)++; 764 765 if (name == NULL || *name == '\0') { 766 if (oidpp) 767 *oidpp = oidp; 768 return (0); 769 } 770 771 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE) 772 break; 773 774 if (oidp->oid_handler) 775 break; 776 777 lsp = SYSCTL_CHILDREN(oidp); 778 } 779 return (ENOENT); 780 } 781 782 static int 783 sysctl_sysctl_name2oid(SYSCTL_HANDLER_ARGS) 784 { 785 char *p; 786 int error, oid[CTL_MAXNAME], len; 787 struct sysctl_oid *op = NULL; 788 789 if (!req->newlen) 790 return ENOENT; 791 if (req->newlen >= MAXPATHLEN) /* XXX arbitrary, undocumented */ 792 return (ENAMETOOLONG); 793 794 p = kmalloc(req->newlen+1, M_SYSCTL, M_WAITOK); 795 796 error = SYSCTL_IN(req, p, req->newlen); 797 if (error) { 798 kfree(p, M_SYSCTL); 799 return (error); 800 } 801 802 p [req->newlen] = '\0'; 803 804 error = name2oid(p, oid, &len, &op); 805 806 kfree(p, M_SYSCTL); 807 808 if (error) 809 return (error); 810 811 error = SYSCTL_OUT(req, oid, len * sizeof *oid); 812 return (error); 813 } 814 815 SYSCTL_PROC(_sysctl, 3, name2oid, CTLFLAG_RW|CTLFLAG_ANYBODY, 0, 0, 816 sysctl_sysctl_name2oid, "I", ""); 817 818 static int 819 sysctl_sysctl_oidfmt(SYSCTL_HANDLER_ARGS) 820 { 821 struct sysctl_oid *oid; 822 int error; 823 824 error = sysctl_find_oid(arg1, arg2, &oid, NULL, req); 825 if (error) 826 return (error); 827 828 if (!oid->oid_fmt) 829 return (ENOENT); 830 error = SYSCTL_OUT(req, &oid->oid_kind, sizeof(oid->oid_kind)); 831 if (error) 832 return (error); 833 error = SYSCTL_OUT(req, oid->oid_fmt, strlen(oid->oid_fmt) + 1); 834 return (error); 835 } 836 837 838 SYSCTL_NODE(_sysctl, 4, oidfmt, CTLFLAG_RD, sysctl_sysctl_oidfmt, ""); 839 840 static int 841 sysctl_sysctl_oiddescr(SYSCTL_HANDLER_ARGS) 842 { 843 struct sysctl_oid *oid; 844 int error; 845 846 error = sysctl_find_oid(arg1, arg2, &oid, NULL, req); 847 if (error) 848 return (error); 849 850 if (!oid->oid_descr) 851 return (ENOENT); 852 error = SYSCTL_OUT(req, oid->oid_descr, strlen(oid->oid_descr) + 1); 853 return (error); 854 } 855 856 SYSCTL_NODE(_sysctl, 5, oiddescr, CTLFLAG_RD, sysctl_sysctl_oiddescr, ""); 857 858 /* 859 * Default "handler" functions. 860 */ 861 862 /* 863 * Handle an int, signed or unsigned. 864 * Two cases: 865 * a variable: point arg1 at it. 866 * a constant: pass it in arg2. 867 */ 868 869 int 870 sysctl_handle_int(SYSCTL_HANDLER_ARGS) 871 { 872 int error = 0; 873 874 if (arg1) 875 error = SYSCTL_OUT(req, arg1, sizeof(int)); 876 else 877 error = SYSCTL_OUT(req, &arg2, sizeof(int)); 878 879 if (error || !req->newptr) 880 return (error); 881 882 if (!arg1) 883 error = EPERM; 884 else 885 error = SYSCTL_IN(req, arg1, sizeof(int)); 886 return (error); 887 } 888 889 /* 890 * Handle a long, signed or unsigned. arg1 points to it. 891 */ 892 893 int 894 sysctl_handle_long(SYSCTL_HANDLER_ARGS) 895 { 896 int error = 0; 897 898 if (!arg1) 899 return (EINVAL); 900 error = SYSCTL_OUT(req, arg1, sizeof(long)); 901 902 if (error || !req->newptr) 903 return (error); 904 905 error = SYSCTL_IN(req, arg1, sizeof(long)); 906 return (error); 907 } 908 909 /* 910 * Handle a quad, signed or unsigned. arg1 points to it. 911 */ 912 913 int 914 sysctl_handle_quad(SYSCTL_HANDLER_ARGS) 915 { 916 int error = 0; 917 918 if (!arg1) 919 return (EINVAL); 920 error = SYSCTL_OUT(req, arg1, sizeof(quad_t)); 921 922 if (error || !req->newptr) 923 return (error); 924 925 error = SYSCTL_IN(req, arg1, sizeof(quad_t)); 926 return (error); 927 } 928 929 /* 930 * Handle our generic '\0' terminated 'C' string. 931 * Two cases: 932 * a variable string: point arg1 at it, arg2 is max length. 933 * a constant string: point arg1 at it, arg2 is zero. 934 */ 935 936 int 937 sysctl_handle_string(SYSCTL_HANDLER_ARGS) 938 { 939 int error=0; 940 941 error = SYSCTL_OUT(req, arg1, strlen((char *)arg1)+1); 942 943 if (error || !req->newptr) 944 return (error); 945 946 if ((req->newlen - req->newidx) >= arg2) { 947 error = EINVAL; 948 } else { 949 arg2 = (req->newlen - req->newidx); 950 error = SYSCTL_IN(req, arg1, arg2); 951 ((char *)arg1)[arg2] = '\0'; 952 } 953 954 return (error); 955 } 956 957 /* 958 * Handle any kind of opaque data. 959 * arg1 points to it, arg2 is the size. 960 */ 961 962 int 963 sysctl_handle_opaque(SYSCTL_HANDLER_ARGS) 964 { 965 int error; 966 967 error = SYSCTL_OUT(req, arg1, arg2); 968 969 if (error || !req->newptr) 970 return (error); 971 972 error = SYSCTL_IN(req, arg1, arg2); 973 974 return (error); 975 } 976 977 /* 978 * Transfer functions to/from kernel space. 979 * XXX: rather untested at this point 980 */ 981 static int 982 sysctl_old_kernel(struct sysctl_req *req, const void *p, size_t l) 983 { 984 size_t i = 0; 985 986 if (req->oldptr) { 987 i = l; 988 if (i > req->oldlen - req->oldidx) 989 i = req->oldlen - req->oldidx; 990 if (i > 0) 991 bcopy(p, (char *)req->oldptr + req->oldidx, i); 992 } 993 req->oldidx += l; 994 if (req->oldptr && i != l) 995 return (ENOMEM); 996 return (0); 997 } 998 999 static int 1000 sysctl_new_kernel(struct sysctl_req *req, void *p, size_t l) 1001 { 1002 1003 if (!req->newptr) 1004 return 0; 1005 if (req->newlen - req->newidx < l) 1006 return (EINVAL); 1007 bcopy((char *)req->newptr + req->newidx, p, l); 1008 req->newidx += l; 1009 return (0); 1010 } 1011 1012 int 1013 kernel_sysctl(int *name, u_int namelen, void *old, size_t *oldlenp, void *new, size_t newlen, size_t *retval) 1014 { 1015 int error = 0; 1016 struct sysctl_req req; 1017 1018 bzero(&req, sizeof req); 1019 1020 req.td = curthread; 1021 1022 if (oldlenp) { 1023 req.oldlen = *oldlenp; 1024 } 1025 req.validlen = req.oldlen; 1026 1027 if (old) { 1028 req.oldptr= old; 1029 } 1030 1031 if (new != NULL) { 1032 req.newlen = newlen; 1033 req.newptr = new; 1034 } 1035 1036 req.oldfunc = sysctl_old_kernel; 1037 req.newfunc = sysctl_new_kernel; 1038 #if 0 1039 req.lock = REQ_UNWIRED; 1040 #endif 1041 1042 SYSCTL_XLOCK(); 1043 error = sysctl_root(0, name, namelen, &req); 1044 SYSCTL_XUNLOCK(); 1045 1046 #if 0 1047 if (req.lock == REQ_WIRED && req.validlen > 0) 1048 vsunlock(req.oldptr, req.validlen); 1049 #endif 1050 1051 if (error && error != ENOMEM) 1052 return (error); 1053 1054 if (retval) { 1055 if (req.oldptr && req.oldidx > req.validlen) 1056 *retval = req.validlen; 1057 else 1058 *retval = req.oldidx; 1059 } 1060 return (error); 1061 } 1062 1063 int 1064 kernel_sysctlbyname(char *name, void *old, size_t *oldlenp, 1065 void *new, size_t newlen, size_t *retval) 1066 { 1067 int oid[CTL_MAXNAME]; 1068 size_t oidlen, plen; 1069 int error; 1070 1071 oid[0] = 0; /* sysctl internal magic */ 1072 oid[1] = 3; /* name2oid */ 1073 oidlen = sizeof(oid); 1074 1075 error = kernel_sysctl(oid, 2, oid, &oidlen, name, strlen(name), &plen); 1076 if (error) 1077 return (error); 1078 1079 error = kernel_sysctl(oid, plen / sizeof(int), old, oldlenp, 1080 new, newlen, retval); 1081 return (error); 1082 } 1083 1084 /* 1085 * Transfer function to/from user space. 1086 */ 1087 static int 1088 sysctl_old_user(struct sysctl_req *req, const void *p, size_t l) 1089 { 1090 int error = 0; 1091 size_t i = 0; 1092 1093 #if 0 1094 if (req->lock == 1 && req->oldptr) { 1095 vslock(req->oldptr, req->oldlen); 1096 req->lock = 2; 1097 } 1098 #endif 1099 if (req->oldptr) { 1100 i = l; 1101 if (i > req->oldlen - req->oldidx) 1102 i = req->oldlen - req->oldidx; 1103 if (i > 0) 1104 error = copyout(p, (char *)req->oldptr + req->oldidx, 1105 i); 1106 } 1107 req->oldidx += l; 1108 if (error) 1109 return (error); 1110 if (req->oldptr && i < l) 1111 return (ENOMEM); 1112 return (0); 1113 } 1114 1115 static int 1116 sysctl_new_user(struct sysctl_req *req, void *p, size_t l) 1117 { 1118 int error; 1119 1120 if (!req->newptr) 1121 return 0; 1122 if (req->newlen - req->newidx < l) 1123 return (EINVAL); 1124 error = copyin((char *)req->newptr + req->newidx, p, l); 1125 req->newidx += l; 1126 return (error); 1127 } 1128 1129 int 1130 sysctl_find_oid(int *name, u_int namelen, struct sysctl_oid **noid, 1131 int *nindx, struct sysctl_req *req) 1132 { 1133 struct sysctl_oid_list *lsp; 1134 struct sysctl_oid *oid; 1135 int indx; 1136 1137 SYSCTL_ASSERT_XLOCKED(); 1138 lsp = &sysctl__children; 1139 indx = 0; 1140 while (indx < CTL_MAXNAME) { 1141 SLIST_FOREACH(oid, lsp, oid_link) { 1142 if (oid->oid_number == name[indx]) 1143 break; 1144 } 1145 if (oid == NULL) 1146 return (ENOENT); 1147 1148 indx++; 1149 if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) { 1150 if (oid->oid_handler != NULL || indx == namelen) { 1151 *noid = oid; 1152 if (nindx != NULL) 1153 *nindx = indx; 1154 KASSERT((oid->oid_kind & CTLFLAG_DYING) == 0, 1155 ("%s found DYING node %p", __func__, oid)); 1156 return (0); 1157 } 1158 lsp = SYSCTL_CHILDREN(oid); 1159 } else if (indx == namelen) { 1160 *noid = oid; 1161 if (nindx != NULL) 1162 *nindx = indx; 1163 KASSERT((oid->oid_kind & CTLFLAG_DYING) == 0, 1164 ("%s found DYING node %p", __func__, oid)); 1165 return (0); 1166 } else { 1167 return (ENOTDIR); 1168 } 1169 } 1170 return (ENOENT); 1171 } 1172 1173 /* 1174 * Traverse our tree, and find the right node, execute whatever it points 1175 * to, and return the resulting error code. 1176 */ 1177 1178 static int 1179 sysctl_root(SYSCTL_HANDLER_ARGS) 1180 { 1181 struct thread *td = req->td; 1182 struct proc *p = td ? td->td_proc : NULL; 1183 struct sysctl_oid *oid; 1184 int error, indx; 1185 1186 error = sysctl_find_oid(arg1, arg2, &oid, &indx, req); 1187 if (error) 1188 return (error); 1189 1190 if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) { 1191 /* 1192 * You can't call a sysctl when it's a node, but has 1193 * no handler. Inform the user that it's a node. 1194 * The indx may or may not be the same as namelen. 1195 */ 1196 if (oid->oid_handler == NULL) 1197 return (EISDIR); 1198 } 1199 1200 /* If writing isn't allowed */ 1201 if (req->newptr && (!(oid->oid_kind & CTLFLAG_WR) || 1202 ((oid->oid_kind & CTLFLAG_SECURE) && securelevel > 0))) 1203 return (EPERM); 1204 1205 /* Most likely only root can write */ 1206 if (!(oid->oid_kind & CTLFLAG_ANYBODY) && req->newptr && p && 1207 (error = priv_check_cred(td->td_ucred, 1208 (oid->oid_kind & CTLFLAG_PRISON) ? PRIV_SYSCTL_WRITEJAIL : 1209 PRIV_SYSCTL_WRITE, 0))) 1210 return (error); 1211 1212 if (!oid->oid_handler) 1213 return EINVAL; 1214 1215 if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) 1216 error = oid->oid_handler(oid, (int *)arg1 + indx, arg2 - indx, 1217 req); 1218 else 1219 error = oid->oid_handler(oid, oid->oid_arg1, oid->oid_arg2, 1220 req); 1221 return (error); 1222 } 1223 1224 int 1225 sys___sysctl(struct sysctl_args *uap) 1226 { 1227 int error, i, name[CTL_MAXNAME]; 1228 size_t j; 1229 1230 if (uap->namelen > CTL_MAXNAME || uap->namelen < 2) 1231 return (EINVAL); 1232 1233 error = copyin(uap->name, &name, uap->namelen * sizeof(int)); 1234 if (error) 1235 return (error); 1236 1237 error = userland_sysctl(name, uap->namelen, 1238 uap->old, uap->oldlenp, 0, 1239 uap->new, uap->newlen, &j); 1240 if (error && error != ENOMEM) 1241 return (error); 1242 if (uap->oldlenp) { 1243 i = copyout(&j, uap->oldlenp, sizeof(j)); 1244 if (i) 1245 return (i); 1246 } 1247 return (error); 1248 } 1249 1250 /* 1251 * This is used from various compatibility syscalls too. That's why name 1252 * must be in kernel space. 1253 */ 1254 int 1255 userland_sysctl(int *name, u_int namelen, void *old, 1256 size_t *oldlenp, int inkernel, void *new, size_t newlen, size_t *retval) 1257 { 1258 int error = 0, memlocked; 1259 struct sysctl_req req; 1260 1261 bzero(&req, sizeof req); 1262 1263 req.td = curthread; 1264 req.flags = 0; 1265 1266 if (oldlenp) { 1267 if (inkernel) { 1268 req.oldlen = *oldlenp; 1269 } else { 1270 error = copyin(oldlenp, &req.oldlen, sizeof(*oldlenp)); 1271 if (error) 1272 return (error); 1273 } 1274 } 1275 req.validlen = req.oldlen; 1276 1277 if (old) { 1278 if (!useracc(old, req.oldlen, VM_PROT_WRITE)) 1279 return (EFAULT); 1280 req.oldptr= old; 1281 } 1282 1283 if (new != NULL) { 1284 if (!useracc(new, newlen, VM_PROT_READ)) 1285 return (EFAULT); 1286 req.newlen = newlen; 1287 req.newptr = new; 1288 } 1289 1290 req.oldfunc = sysctl_old_user; 1291 req.newfunc = sysctl_new_user; 1292 #if 0 1293 req.lock = REQ_UNWIRED; 1294 #endif 1295 1296 #ifdef KTRACE 1297 if (KTRPOINT(curthread, KTR_SYSCTL)) 1298 ktrsysctl(name, namelen); 1299 #endif 1300 1301 if (req.oldlen > PAGE_SIZE) { 1302 memlocked = 1; 1303 lockmgr(&sysctlmemlock, LK_EXCLUSIVE); 1304 } else 1305 memlocked = 0; 1306 1307 for (;;) { 1308 req.oldidx = 0; 1309 req.newidx = 0; 1310 SYSCTL_XLOCK(); 1311 error = sysctl_root(0, name, namelen, &req); 1312 SYSCTL_XUNLOCK(); 1313 if (error != EAGAIN) 1314 break; 1315 lwkt_yield(); 1316 } 1317 1318 #if 0 1319 if (req.lock == REQ_WIRED && req.validlen > 0) 1320 vsunlock(req.oldptr, req.validlen); 1321 #endif 1322 if (memlocked) 1323 lockmgr(&sysctlmemlock, LK_RELEASE); 1324 1325 if (error && error != ENOMEM) 1326 return (error); 1327 1328 if (retval) { 1329 if (req.oldptr && req.oldidx > req.validlen) 1330 *retval = req.validlen; 1331 else 1332 *retval = req.oldidx; 1333 } 1334 return (error); 1335 } 1336 1337 int 1338 sysctl_int_range(SYSCTL_HANDLER_ARGS, int low, int high) 1339 { 1340 int error, value; 1341 1342 value = *(int *)arg1; 1343 error = sysctl_handle_int(oidp, &value, 0, req); 1344 if (error || !req->newptr) 1345 return (error); 1346 if (value < low || value > high) 1347 return (EINVAL); 1348 *(int *)arg1 = value; 1349 return (0); 1350 } 1351 1352 /* 1353 * Drain into a sysctl struct. The user buffer should be wired if a page 1354 * fault would cause issue. 1355 */ 1356 static int 1357 sbuf_sysctl_drain(void *arg, const char *data, int len) 1358 { 1359 struct sysctl_req *req = arg; 1360 int error; 1361 1362 error = SYSCTL_OUT(req, data, len); 1363 KASSERT(error >= 0, ("Got unexpected negative value %d", error)); 1364 return (error == 0 ? len : -error); 1365 } 1366 1367 struct sbuf * 1368 sbuf_new_for_sysctl(struct sbuf *s, char *buf, int length, 1369 struct sysctl_req *req) 1370 { 1371 1372 s = sbuf_new(s, buf, length, SBUF_FIXEDLEN); 1373 sbuf_set_drain(s, sbuf_sysctl_drain, req); 1374 return (s); 1375 } 1376