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 static struct lock sysctllock; 73 static struct lock sysctlmemlock; 74 75 #define SYSCTL_XLOCK() lockmgr(&sysctllock, LK_EXCLUSIVE) 76 #define SYSCTL_XUNLOCK() lockmgr(&sysctllock, LK_RELEASE) 77 #define SYSCTL_ASSERT_XLOCKED() KKASSERT(lockstatus(&sysctllock, curthread) != 0) 78 #define SYSCTL_INIT() lockinit(&sysctllock, \ 79 "sysctl lock", 0, LK_CANRECURSE) 80 #define SYSCTL_SLEEP(ch, wmesg, timo) \ 81 lksleep(ch, &sysctllock, 0, wmesg, timo) 82 83 static int sysctl_root(SYSCTL_HANDLER_ARGS); 84 static void sysctl_register_oid_int(struct sysctl_oid *oipd); 85 static void sysctl_unregister_oid_int(struct sysctl_oid *oipd); 86 87 struct sysctl_oid_list sysctl__children; /* root list */ 88 89 static int sysctl_remove_oid_locked(struct sysctl_oid *oidp, int del, 90 int recurse); 91 92 static struct sysctl_oid * 93 sysctl_find_oidname(const char *name, struct sysctl_oid_list *list, int lock) 94 { 95 struct sysctl_oid *oidp; 96 97 SLIST_FOREACH(oidp, list, oid_link) { 98 if (strcmp(oidp->oid_name, name) == 0) { 99 break; 100 } 101 } 102 return (oidp); 103 } 104 105 /* 106 * Initialization of the MIB tree. 107 * 108 * Order by number in each list. 109 */ 110 111 void 112 sysctl_register_oid(struct sysctl_oid *oidp) 113 { 114 SYSCTL_XLOCK(); 115 sysctl_register_oid_int(oidp); 116 SYSCTL_XUNLOCK(); 117 } 118 119 static void 120 sysctl_register_oid_int(struct sysctl_oid *oidp) 121 { 122 struct sysctl_oid_list *parent = oidp->oid_parent; 123 struct sysctl_oid *p; 124 struct sysctl_oid *q; 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_XLOCKED(); 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_XLOCKED(); 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_XLOCKED(); 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 kfree(oidp, M_SYSCTLOID); 402 } 403 } 404 return (0); 405 } 406 407 int 408 sysctl_remove_name(struct sysctl_oid *parent, const char *name, 409 int del, int recurse) 410 { 411 struct sysctl_oid *p, *tmp; 412 int error; 413 414 error = ENOENT; 415 SYSCTL_XLOCK(); 416 SLIST_FOREACH_MUTABLE(p, SYSCTL_CHILDREN(parent), oid_link, tmp) { 417 if (strcmp(p->oid_name, name) == 0) { 418 error = sysctl_remove_oid_locked(p, del, recurse); 419 break; 420 } 421 } 422 SYSCTL_XUNLOCK(); 423 424 return (error); 425 } 426 427 /* 428 * Create new sysctls at run time. 429 * clist may point to a valid context initialized with sysctl_ctx_init(). 430 */ 431 struct sysctl_oid * 432 sysctl_add_oid(struct sysctl_ctx_list *clist, struct sysctl_oid_list *parent, 433 int number, const char *name, int kind, void *arg1, int arg2, 434 int (*handler)(SYSCTL_HANDLER_ARGS), const char *fmt, const char *descr) 435 { 436 struct sysctl_oid *oidp; 437 ssize_t len; 438 char *newname; 439 440 /* You have to hook up somewhere.. */ 441 if (parent == NULL) 442 return(NULL); 443 SYSCTL_XLOCK(); 444 /* Check if the node already exists, otherwise create it */ 445 oidp = sysctl_find_oidname(name, parent, 0); 446 if (oidp != NULL) { 447 if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) { 448 oidp->oid_refcnt++; 449 /* Update the context */ 450 if (clist != NULL) 451 sysctl_ctx_entry_add(clist, oidp); 452 SYSCTL_XUNLOCK(); 453 return (oidp); 454 } else { 455 kprintf("can't re-use a leaf (%s)!\n", name); 456 SYSCTL_XUNLOCK(); 457 return (NULL); 458 } 459 } 460 oidp = kmalloc(sizeof(struct sysctl_oid), M_SYSCTLOID, M_WAITOK | M_ZERO); 461 oidp->oid_parent = parent; 462 SLIST_NEXT(oidp, oid_link) = NULL; 463 oidp->oid_number = number; 464 oidp->oid_refcnt = 1; 465 len = strlen(name); 466 newname = kmalloc(len + 1, M_SYSCTLOID, M_WAITOK); 467 bcopy(name, newname, len + 1); 468 newname[len] = '\0'; 469 oidp->oid_name = newname; 470 oidp->oid_handler = handler; 471 oidp->oid_kind = CTLFLAG_DYN | kind; 472 if ((kind & CTLTYPE) == CTLTYPE_NODE) { 473 struct sysctl_oid_list *children; 474 475 /* Allocate space for children */ 476 children = kmalloc(sizeof(*children), M_SYSCTLOID, M_WAITOK); 477 SYSCTL_SET_CHILDREN(oidp, children); 478 SLIST_INIT(children); 479 } else { 480 oidp->oid_arg1 = arg1; 481 oidp->oid_arg2 = arg2; 482 } 483 oidp->oid_fmt = fmt; 484 if (descr) { 485 int len = strlen(descr) + 1; 486 oidp->oid_descr = kmalloc(len, M_SYSCTLOID, M_WAITOK); 487 strcpy((char *)(uintptr_t)(const void *)oidp->oid_descr, descr); 488 }; 489 /* Update the context, if used */ 490 if (clist != NULL) 491 sysctl_ctx_entry_add(clist, oidp); 492 /* Register this oid */ 493 sysctl_register_oid_int(oidp); 494 SYSCTL_XUNLOCK(); 495 return (oidp); 496 } 497 498 /* 499 * Rename an existing oid. 500 */ 501 void 502 sysctl_rename_oid(struct sysctl_oid *oidp, const char *name) 503 { 504 char *newname; 505 char *oldname; 506 507 newname = kstrdup(name, M_SYSCTLOID); 508 SYSCTL_XLOCK(); 509 oldname = __DECONST(char *, oidp->oid_name); 510 oidp->oid_name = newname; 511 SYSCTL_XUNLOCK(); 512 kfree(oldname, M_SYSCTLOID); 513 } 514 515 /* 516 * Register the kernel's oids on startup. 517 */ 518 SET_DECLARE(sysctl_set, struct sysctl_oid); 519 520 static void 521 sysctl_register_all(void *arg) 522 { 523 struct sysctl_oid **oidp; 524 525 lockinit(&sysctlmemlock, "sysctl mem", 0, LK_CANRECURSE); 526 SYSCTL_INIT(); 527 SYSCTL_XLOCK(); 528 SET_FOREACH(oidp, sysctl_set) 529 sysctl_register_oid(*oidp); 530 SYSCTL_XUNLOCK(); 531 } 532 SYSINIT(sysctl, SI_BOOT1_POST, SI_ORDER_ANY, sysctl_register_all, 0); 533 534 /* 535 * "Staff-functions" 536 * 537 * These functions implement a presently undocumented interface 538 * used by the sysctl program to walk the tree, and get the type 539 * so it can print the value. 540 * This interface is under work and consideration, and should probably 541 * be killed with a big axe by the first person who can find the time. 542 * (be aware though, that the proper interface isn't as obvious as it 543 * may seem, there are various conflicting requirements. 544 * 545 * {0,0} kprintf the entire MIB-tree. 546 * {0,1,...} return the name of the "..." OID. 547 * {0,2,...} return the next OID. 548 * {0,3} return the OID of the name in "new" 549 * {0,4,...} return the kind & format info for the "..." OID. 550 */ 551 552 static void 553 sysctl_sysctl_debug_dump_node(struct sysctl_oid_list *l, int i) 554 { 555 int k; 556 struct sysctl_oid *oidp; 557 558 SYSCTL_ASSERT_XLOCKED(); 559 SLIST_FOREACH(oidp, l, oid_link) { 560 561 for (k=0; k<i; k++) 562 kprintf(" "); 563 564 kprintf("%d %s ", oidp->oid_number, oidp->oid_name); 565 566 kprintf("%c%c", 567 oidp->oid_kind & CTLFLAG_RD ? 'R':' ', 568 oidp->oid_kind & CTLFLAG_WR ? 'W':' '); 569 570 if (oidp->oid_handler) 571 kprintf(" *Handler"); 572 573 switch (oidp->oid_kind & CTLTYPE) { 574 case CTLTYPE_NODE: 575 kprintf(" Node\n"); 576 if (!oidp->oid_handler) { 577 sysctl_sysctl_debug_dump_node( 578 oidp->oid_arg1, i+2); 579 } 580 break; 581 case CTLTYPE_INT: kprintf(" Int\n"); break; 582 case CTLTYPE_STRING: kprintf(" String\n"); break; 583 case CTLTYPE_QUAD: kprintf(" Quad\n"); break; 584 case CTLTYPE_OPAQUE: kprintf(" Opaque/struct\n"); break; 585 default: kprintf("\n"); 586 } 587 588 } 589 SYSCTL_XUNLOCK(); 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_XLOCK(); 601 sysctl_sysctl_debug_dump_node(&sysctl__children, 0); 602 SYSCTL_XUNLOCK(); 603 return (ENOENT); 604 } 605 606 SYSCTL_PROC(_sysctl, 0, debug, CTLTYPE_STRING|CTLFLAG_RD, 607 0, 0, sysctl_sysctl_debug, "-", ""); 608 609 static int 610 sysctl_sysctl_name(SYSCTL_HANDLER_ARGS) 611 { 612 int *name = (int *) arg1; 613 u_int namelen = arg2; 614 int error = 0; 615 struct sysctl_oid *oid; 616 struct sysctl_oid_list *lsp = &sysctl__children, *lsp2; 617 char buf[10]; 618 619 SYSCTL_XLOCK(); 620 while (namelen) { 621 if (!lsp) { 622 ksnprintf(buf, sizeof(buf), "%d", *name); 623 if (req->oldidx) 624 error = SYSCTL_OUT(req, ".", 1); 625 if (!error) 626 error = SYSCTL_OUT(req, buf, strlen(buf)); 627 if (error) 628 goto out; 629 namelen--; 630 name++; 631 continue; 632 } 633 lsp2 = NULL; 634 SLIST_FOREACH(oid, lsp, oid_link) { 635 if (oid->oid_number != *name) 636 continue; 637 638 if (req->oldidx) 639 error = SYSCTL_OUT(req, ".", 1); 640 if (!error) 641 error = SYSCTL_OUT(req, oid->oid_name, 642 strlen(oid->oid_name)); 643 if (error) 644 goto out; 645 646 namelen--; 647 name++; 648 649 if ((oid->oid_kind & CTLTYPE) != CTLTYPE_NODE) 650 break; 651 652 if (oid->oid_handler) 653 break; 654 655 lsp2 = SYSCTL_CHILDREN(oid); 656 break; 657 } 658 lsp = lsp2; 659 } 660 error = SYSCTL_OUT(req, "", 1); 661 out: 662 SYSCTL_XUNLOCK(); 663 return (error); 664 } 665 666 SYSCTL_NODE(_sysctl, 1, name, CTLFLAG_RD, sysctl_sysctl_name, ""); 667 668 static int 669 sysctl_sysctl_next_ls(struct sysctl_oid_list *lsp, int *name, u_int namelen, 670 int *next, int *len, int level, struct sysctl_oid **oidpp) 671 { 672 struct sysctl_oid *oidp; 673 674 SYSCTL_ASSERT_XLOCKED(); 675 *len = level; 676 SLIST_FOREACH(oidp, lsp, oid_link) { 677 *next = oidp->oid_number; 678 *oidpp = oidp; 679 680 if (oidp->oid_kind & CTLFLAG_SKIP) 681 continue; 682 683 if (!namelen) { 684 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE) 685 return (0); 686 if (oidp->oid_handler) 687 /* We really should call the handler here...*/ 688 return (0); 689 lsp = SYSCTL_CHILDREN(oidp); 690 if (!sysctl_sysctl_next_ls(lsp, 0, 0, next+1, 691 len, level+1, oidpp)) 692 return (0); 693 goto emptynode; 694 } 695 696 if (oidp->oid_number < *name) 697 continue; 698 699 if (oidp->oid_number > *name) { 700 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE) 701 return (0); 702 if (oidp->oid_handler) 703 return (0); 704 lsp = SYSCTL_CHILDREN(oidp); 705 if (!sysctl_sysctl_next_ls(lsp, name+1, namelen-1, 706 next+1, len, level+1, oidpp)) 707 return (0); 708 goto next; 709 } 710 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE) 711 continue; 712 713 if (oidp->oid_handler) 714 continue; 715 716 lsp = SYSCTL_CHILDREN(oidp); 717 if (!sysctl_sysctl_next_ls(lsp, name+1, namelen-1, next+1, 718 len, level+1, oidpp)) 719 return (0); 720 next: 721 namelen = 1; 722 emptynode: 723 *len = level; 724 } 725 return (1); 726 } 727 728 static int 729 sysctl_sysctl_next(SYSCTL_HANDLER_ARGS) 730 { 731 int *name = (int *) arg1; 732 u_int namelen = arg2; 733 int i, j, error; 734 struct sysctl_oid *oid; 735 struct sysctl_oid_list *lsp = &sysctl__children; 736 int newoid[CTL_MAXNAME]; 737 738 i = sysctl_sysctl_next_ls(lsp, name, namelen, newoid, &j, 1, &oid); 739 if (i) 740 return ENOENT; 741 error = SYSCTL_OUT(req, newoid, j * sizeof (int)); 742 return (error); 743 } 744 745 SYSCTL_NODE(_sysctl, 2, next, CTLFLAG_RD, sysctl_sysctl_next, ""); 746 747 static int 748 name2oid(char *name, int *oid, int *len, struct sysctl_oid **oidpp) 749 { 750 struct sysctl_oid *oidp; 751 struct sysctl_oid_list *lsp = &sysctl__children; 752 char *p; 753 754 SYSCTL_ASSERT_XLOCKED(); 755 756 for (*len = 0; *len < CTL_MAXNAME;) { 757 p = strsep(&name, "."); 758 759 oidp = SLIST_FIRST(lsp); 760 for (;; oidp = SLIST_NEXT(oidp, oid_link)) { 761 if (oidp == NULL) 762 return (ENOENT); 763 if (strcmp(p, oidp->oid_name) == 0) 764 break; 765 } 766 *oid++ = oidp->oid_number; 767 (*len)++; 768 769 if (name == NULL || *name == '\0') { 770 if (oidpp) 771 *oidpp = oidp; 772 return (0); 773 } 774 775 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE) 776 break; 777 778 if (oidp->oid_handler) 779 break; 780 781 lsp = SYSCTL_CHILDREN(oidp); 782 } 783 return (ENOENT); 784 } 785 786 static int 787 sysctl_sysctl_name2oid(SYSCTL_HANDLER_ARGS) 788 { 789 char *p; 790 int error, oid[CTL_MAXNAME], len; 791 struct sysctl_oid *op = NULL; 792 793 if (!req->newlen) 794 return ENOENT; 795 if (req->newlen >= MAXPATHLEN) /* XXX arbitrary, undocumented */ 796 return (ENAMETOOLONG); 797 798 p = kmalloc(req->newlen+1, M_SYSCTL, M_WAITOK); 799 800 error = SYSCTL_IN(req, p, req->newlen); 801 if (error) { 802 kfree(p, M_SYSCTL); 803 return (error); 804 } 805 806 p [req->newlen] = '\0'; 807 808 error = name2oid(p, oid, &len, &op); 809 810 kfree(p, M_SYSCTL); 811 812 if (error) 813 return (error); 814 815 error = SYSCTL_OUT(req, oid, len * sizeof *oid); 816 return (error); 817 } 818 819 SYSCTL_PROC(_sysctl, 3, name2oid, CTLFLAG_RW|CTLFLAG_ANYBODY, 0, 0, 820 sysctl_sysctl_name2oid, "I", ""); 821 822 static int 823 sysctl_sysctl_oidfmt(SYSCTL_HANDLER_ARGS) 824 { 825 struct sysctl_oid *oid; 826 int error; 827 828 error = sysctl_find_oid(arg1, arg2, &oid, NULL, req); 829 if (error) 830 return (error); 831 832 if (!oid->oid_fmt) 833 return (ENOENT); 834 error = SYSCTL_OUT(req, &oid->oid_kind, sizeof(oid->oid_kind)); 835 if (error) 836 return (error); 837 error = SYSCTL_OUT(req, oid->oid_fmt, strlen(oid->oid_fmt) + 1); 838 return (error); 839 } 840 841 842 SYSCTL_NODE(_sysctl, 4, oidfmt, CTLFLAG_RD, sysctl_sysctl_oidfmt, ""); 843 844 static int 845 sysctl_sysctl_oiddescr(SYSCTL_HANDLER_ARGS) 846 { 847 struct sysctl_oid *oid; 848 int error; 849 850 error = sysctl_find_oid(arg1, arg2, &oid, NULL, req); 851 if (error) 852 return (error); 853 854 if (!oid->oid_descr) 855 return (ENOENT); 856 error = SYSCTL_OUT(req, oid->oid_descr, strlen(oid->oid_descr) + 1); 857 return (error); 858 } 859 860 SYSCTL_NODE(_sysctl, 5, oiddescr, CTLFLAG_RD, sysctl_sysctl_oiddescr, ""); 861 862 /* 863 * Default "handler" functions. 864 */ 865 866 /* 867 * Handle an int, signed or unsigned. 868 * Two cases: 869 * a variable: point arg1 at it. 870 * a constant: pass it in arg2. 871 */ 872 873 int 874 sysctl_handle_int(SYSCTL_HANDLER_ARGS) 875 { 876 int error = 0; 877 878 if (arg1) 879 error = SYSCTL_OUT(req, arg1, sizeof(int)); 880 else 881 error = SYSCTL_OUT(req, &arg2, sizeof(int)); 882 883 if (error || !req->newptr) 884 return (error); 885 886 if (!arg1) 887 error = EPERM; 888 else 889 error = SYSCTL_IN(req, arg1, sizeof(int)); 890 return (error); 891 } 892 893 /* 894 * Handle a long, signed or unsigned. arg1 points to it. 895 */ 896 897 int 898 sysctl_handle_long(SYSCTL_HANDLER_ARGS) 899 { 900 int error = 0; 901 902 if (!arg1) 903 return (EINVAL); 904 error = SYSCTL_OUT(req, arg1, sizeof(long)); 905 906 if (error || !req->newptr) 907 return (error); 908 909 error = SYSCTL_IN(req, arg1, sizeof(long)); 910 return (error); 911 } 912 913 /* 914 * Handle a quad, signed or unsigned. arg1 points to it. 915 */ 916 917 int 918 sysctl_handle_quad(SYSCTL_HANDLER_ARGS) 919 { 920 int error = 0; 921 922 if (!arg1) 923 return (EINVAL); 924 error = SYSCTL_OUT(req, arg1, sizeof(quad_t)); 925 926 if (error || !req->newptr) 927 return (error); 928 929 error = SYSCTL_IN(req, arg1, sizeof(quad_t)); 930 return (error); 931 } 932 933 /* 934 * Handle our generic '\0' terminated 'C' string. 935 * Two cases: 936 * a variable string: point arg1 at it, arg2 is max length. 937 * a constant string: point arg1 at it, arg2 is zero. 938 */ 939 940 int 941 sysctl_handle_string(SYSCTL_HANDLER_ARGS) 942 { 943 int error=0; 944 945 error = SYSCTL_OUT(req, arg1, strlen((char *)arg1)+1); 946 947 if (error || !req->newptr) 948 return (error); 949 950 if ((req->newlen - req->newidx) >= arg2) { 951 error = EINVAL; 952 } else { 953 arg2 = (req->newlen - req->newidx); 954 error = SYSCTL_IN(req, arg1, arg2); 955 ((char *)arg1)[arg2] = '\0'; 956 } 957 958 return (error); 959 } 960 961 /* 962 * Handle any kind of opaque data. 963 * arg1 points to it, arg2 is the size. 964 */ 965 966 int 967 sysctl_handle_opaque(SYSCTL_HANDLER_ARGS) 968 { 969 int error; 970 971 error = SYSCTL_OUT(req, arg1, arg2); 972 973 if (error || !req->newptr) 974 return (error); 975 976 error = SYSCTL_IN(req, arg1, arg2); 977 978 return (error); 979 } 980 981 /* 982 * Transfer functions to/from kernel space. 983 * XXX: rather untested at this point 984 */ 985 static int 986 sysctl_old_kernel(struct sysctl_req *req, const void *p, size_t l) 987 { 988 size_t i = 0; 989 990 if (req->oldptr) { 991 i = l; 992 if (i > req->oldlen - req->oldidx) 993 i = req->oldlen - req->oldidx; 994 if (i > 0) 995 bcopy(p, (char *)req->oldptr + req->oldidx, i); 996 } 997 req->oldidx += l; 998 if (req->oldptr && i != l) 999 return (ENOMEM); 1000 return (0); 1001 } 1002 1003 static int 1004 sysctl_new_kernel(struct sysctl_req *req, void *p, size_t l) 1005 { 1006 1007 if (!req->newptr) 1008 return 0; 1009 if (req->newlen - req->newidx < l) 1010 return (EINVAL); 1011 bcopy((char *)req->newptr + req->newidx, p, l); 1012 req->newidx += l; 1013 return (0); 1014 } 1015 1016 int 1017 kernel_sysctl(int *name, u_int namelen, void *old, size_t *oldlenp, void *new, size_t newlen, size_t *retval) 1018 { 1019 int error = 0; 1020 struct sysctl_req req; 1021 1022 bzero(&req, sizeof req); 1023 1024 req.td = curthread; 1025 1026 if (oldlenp) { 1027 req.oldlen = *oldlenp; 1028 } 1029 req.validlen = req.oldlen; 1030 1031 if (old) { 1032 req.oldptr= old; 1033 } 1034 1035 if (new != NULL) { 1036 req.newlen = newlen; 1037 req.newptr = new; 1038 } 1039 1040 req.oldfunc = sysctl_old_kernel; 1041 req.newfunc = sysctl_new_kernel; 1042 #if 0 1043 req.lock = REQ_UNWIRED; 1044 #endif 1045 1046 SYSCTL_XLOCK(); 1047 error = sysctl_root(0, name, namelen, &req); 1048 SYSCTL_XUNLOCK(); 1049 1050 #if 0 1051 if (req.lock == REQ_WIRED && req.validlen > 0) 1052 vsunlock(req.oldptr, req.validlen); 1053 #endif 1054 1055 if (error && error != ENOMEM) 1056 return (error); 1057 1058 if (retval) { 1059 if (req.oldptr && req.oldidx > req.validlen) 1060 *retval = req.validlen; 1061 else 1062 *retval = req.oldidx; 1063 } 1064 return (error); 1065 } 1066 1067 int 1068 kernel_sysctlbyname(char *name, 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 SYSCTL_ASSERT_XLOCKED(); 1142 lsp = &sysctl__children; 1143 indx = 0; 1144 while (indx < CTL_MAXNAME) { 1145 SLIST_FOREACH(oid, lsp, oid_link) { 1146 if (oid->oid_number == name[indx]) 1147 break; 1148 } 1149 if (oid == NULL) 1150 return (ENOENT); 1151 1152 indx++; 1153 if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) { 1154 if (oid->oid_handler != NULL || indx == namelen) { 1155 *noid = oid; 1156 if (nindx != NULL) 1157 *nindx = indx; 1158 KASSERT((oid->oid_kind & CTLFLAG_DYING) == 0, 1159 ("%s found DYING node %p", __func__, oid)); 1160 return (0); 1161 } 1162 lsp = SYSCTL_CHILDREN(oid); 1163 } else if (indx == namelen) { 1164 *noid = oid; 1165 if (nindx != NULL) 1166 *nindx = indx; 1167 KASSERT((oid->oid_kind & CTLFLAG_DYING) == 0, 1168 ("%s found DYING node %p", __func__, oid)); 1169 return (0); 1170 } else { 1171 return (ENOTDIR); 1172 } 1173 } 1174 return (ENOENT); 1175 } 1176 1177 /* 1178 * Traverse our tree, and find the right node, execute whatever it points 1179 * to, and return the resulting error code. 1180 */ 1181 1182 int 1183 sysctl_root(SYSCTL_HANDLER_ARGS) 1184 { 1185 struct thread *td = req->td; 1186 struct proc *p = td ? td->td_proc : NULL; 1187 struct sysctl_oid *oid; 1188 int error, indx; 1189 1190 error = sysctl_find_oid(arg1, arg2, &oid, &indx, req); 1191 if (error) 1192 return (error); 1193 1194 if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) { 1195 /* 1196 * You can't call a sysctl when it's a node, but has 1197 * no handler. Inform the user that it's a node. 1198 * The indx may or may not be the same as namelen. 1199 */ 1200 if (oid->oid_handler == NULL) 1201 return (EISDIR); 1202 } 1203 1204 /* If writing isn't allowed */ 1205 if (req->newptr && (!(oid->oid_kind & CTLFLAG_WR) || 1206 ((oid->oid_kind & CTLFLAG_SECURE) && securelevel > 0))) 1207 return (EPERM); 1208 1209 /* Most likely only root can write */ 1210 if (!(oid->oid_kind & CTLFLAG_ANYBODY) && req->newptr && p && 1211 (error = priv_check_cred(td->td_ucred, 1212 (oid->oid_kind & CTLFLAG_PRISON) ? PRIV_SYSCTL_WRITEJAIL : 1213 PRIV_SYSCTL_WRITE, 0))) 1214 return (error); 1215 1216 if (!oid->oid_handler) 1217 return EINVAL; 1218 1219 if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) 1220 error = oid->oid_handler(oid, (int *)arg1 + indx, arg2 - indx, 1221 req); 1222 else 1223 error = oid->oid_handler(oid, oid->oid_arg1, oid->oid_arg2, 1224 req); 1225 return (error); 1226 } 1227 1228 int 1229 sys___sysctl(struct sysctl_args *uap) 1230 { 1231 int error, i, name[CTL_MAXNAME]; 1232 size_t j; 1233 1234 if (uap->namelen > CTL_MAXNAME || uap->namelen < 2) 1235 return (EINVAL); 1236 1237 error = copyin(uap->name, &name, uap->namelen * sizeof(int)); 1238 if (error) 1239 return (error); 1240 1241 error = userland_sysctl(name, uap->namelen, 1242 uap->old, uap->oldlenp, 0, 1243 uap->new, uap->newlen, &j); 1244 if (error && error != ENOMEM) 1245 return (error); 1246 if (uap->oldlenp) { 1247 i = copyout(&j, uap->oldlenp, sizeof(j)); 1248 if (i) 1249 return (i); 1250 } 1251 return (error); 1252 } 1253 1254 /* 1255 * This is used from various compatibility syscalls too. That's why name 1256 * must be in kernel space. 1257 */ 1258 int 1259 userland_sysctl(int *name, u_int namelen, void *old, 1260 size_t *oldlenp, int inkernel, void *new, size_t newlen, size_t *retval) 1261 { 1262 int error = 0, memlocked; 1263 struct sysctl_req req; 1264 1265 bzero(&req, sizeof req); 1266 1267 req.td = curthread; 1268 req.flags = 0; 1269 1270 if (oldlenp) { 1271 if (inkernel) { 1272 req.oldlen = *oldlenp; 1273 } else { 1274 error = copyin(oldlenp, &req.oldlen, sizeof(*oldlenp)); 1275 if (error) 1276 return (error); 1277 } 1278 } 1279 req.validlen = req.oldlen; 1280 1281 if (old) { 1282 if (!useracc(old, req.oldlen, VM_PROT_WRITE)) 1283 return (EFAULT); 1284 req.oldptr= old; 1285 } 1286 1287 if (new != NULL) { 1288 if (!useracc(new, newlen, VM_PROT_READ)) 1289 return (EFAULT); 1290 req.newlen = newlen; 1291 req.newptr = new; 1292 } 1293 1294 req.oldfunc = sysctl_old_user; 1295 req.newfunc = sysctl_new_user; 1296 #if 0 1297 req.lock = REQ_UNWIRED; 1298 #endif 1299 1300 #ifdef KTRACE 1301 if (KTRPOINT(curthread, KTR_SYSCTL)) 1302 ktrsysctl(name, namelen); 1303 #endif 1304 1305 if (req.oldlen > PAGE_SIZE) { 1306 memlocked = 1; 1307 lockmgr(&sysctlmemlock, LK_EXCLUSIVE); 1308 } else 1309 memlocked = 0; 1310 1311 for (;;) { 1312 req.oldidx = 0; 1313 req.newidx = 0; 1314 SYSCTL_XLOCK(); 1315 error = sysctl_root(0, name, namelen, &req); 1316 SYSCTL_XUNLOCK(); 1317 if (error != EAGAIN) 1318 break; 1319 lwkt_yield(); 1320 } 1321 1322 #if 0 1323 if (req.lock == REQ_WIRED && req.validlen > 0) 1324 vsunlock(req.oldptr, req.validlen); 1325 #endif 1326 if (memlocked) 1327 lockmgr(&sysctlmemlock, LK_RELEASE); 1328 1329 if (error && error != ENOMEM) 1330 return (error); 1331 1332 if (retval) { 1333 if (req.oldptr && req.oldidx > req.validlen) 1334 *retval = req.validlen; 1335 else 1336 *retval = req.oldidx; 1337 } 1338 return (error); 1339 } 1340 1341 int 1342 sysctl_int_range(SYSCTL_HANDLER_ARGS, int low, int high) 1343 { 1344 int error, value; 1345 1346 value = *(int *)arg1; 1347 error = sysctl_handle_int(oidp, &value, 0, req); 1348 if (error || !req->newptr) 1349 return (error); 1350 if (value < low || value > high) 1351 return (EINVAL); 1352 *(int *)arg1 = value; 1353 return (0); 1354 } 1355 1356 /* 1357 * Drain into a sysctl struct. The user buffer should be wired if a page 1358 * fault would cause issue. 1359 */ 1360 static int 1361 sbuf_sysctl_drain(void *arg, const char *data, int len) 1362 { 1363 struct sysctl_req *req = arg; 1364 int error; 1365 1366 error = SYSCTL_OUT(req, data, len); 1367 KASSERT(error >= 0, ("Got unexpected negative value %d", error)); 1368 return (error == 0 ? len : -error); 1369 } 1370 1371 struct sbuf * 1372 sbuf_new_for_sysctl(struct sbuf *s, char *buf, int length, 1373 struct sysctl_req *req) 1374 { 1375 1376 s = sbuf_new(s, buf, length, SBUF_FIXEDLEN); 1377 sbuf_set_drain(s, sbuf_sysctl_drain, req); 1378 return (s); 1379 } 1380