Now available: The Design and Implementation of the FreeBSD Operating System (Second Edition)
[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]
FreeBSD/Linux Kernel Cross Reference
sys/kern/kern_sysctl.c
Version:
- FREEBSD - FREEBSD-13-STABLE - FREEBSD-13-0 - FREEBSD-12-STABLE - FREEBSD-12-0 - FREEBSD-11-STABLE - FREEBSD-11-0 - FREEBSD-10-STABLE - FREEBSD-10-0 - FREEBSD-9-STABLE - FREEBSD-9-0 - FREEBSD-8-STABLE - FREEBSD-8-0 - FREEBSD-7-STABLE - FREEBSD-7-0 - FREEBSD-6-STABLE - FREEBSD-6-0 - FREEBSD-5-STABLE - FREEBSD-5-0 - FREEBSD-4-STABLE - FREEBSD-3-STABLE - FREEBSD22 - l41 - OPENBSD - linux-2.6 - MK84 - PLAN9 - xnu-8792
SearchContext: - none - 3 - 10
SearchContext: - none - 3 - 10
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 * 4. 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 */ 37 38 #include <sys/cdefs.h> 39 __FBSDID("$FreeBSD$"); 40 41 #include "opt_capsicum.h" 42 #include "opt_compat.h" 43 #include "opt_ktrace.h" 44 45 #include <sys/param.h> 46 #include <sys/fail.h> 47 #include <sys/systm.h> 48 #include <sys/capability.h> 49 #include <sys/kernel.h> 50 #include <sys/sysctl.h> 51 #include <sys/malloc.h> 52 #include <sys/priv.h> 53 #include <sys/proc.h> 54 #include <sys/jail.h> 55 #include <sys/lock.h> 56 #include <sys/mutex.h> 57 #include <sys/sbuf.h> 58 #include <sys/sx.h> 59 #include <sys/sysproto.h> 60 #include <sys/uio.h> 61 #ifdef KTRACE 62 #include <sys/ktrace.h> 63 #endif 64 65 #include <net/vnet.h> 66 67 #include <security/mac/mac_framework.h> 68 69 #include <vm/vm.h> 70 #include <vm/vm_extern.h> 71 72 static MALLOC_DEFINE(M_SYSCTL, "sysctl", "sysctl internal magic"); 73 static MALLOC_DEFINE(M_SYSCTLOID, "sysctloid", "sysctl dynamic oids"); 74 static MALLOC_DEFINE(M_SYSCTLTMP, "sysctltmp", "sysctl temp output buffer"); 75 76 /* 77 * The sysctllock protects the MIB tree. It also protects sysctl 78 * contexts used with dynamic sysctls. The sysctl_register_oid() and 79 * sysctl_unregister_oid() routines require the sysctllock to already 80 * be held, so the sysctl_lock() and sysctl_unlock() routines are 81 * provided for the few places in the kernel which need to use that 82 * API rather than using the dynamic API. Use of the dynamic API is 83 * strongly encouraged for most code. 84 * 85 * The sysctlmemlock is used to limit the amount of user memory wired for 86 * sysctl requests. This is implemented by serializing any userland 87 * sysctl requests larger than a single page via an exclusive lock. 88 */ 89 static struct sx sysctllock; 90 static struct sx sysctlmemlock; 91 92 #define SYSCTL_XLOCK() sx_xlock(&sysctllock) 93 #define SYSCTL_XUNLOCK() sx_xunlock(&sysctllock) 94 #define SYSCTL_ASSERT_XLOCKED() sx_assert(&sysctllock, SA_XLOCKED) 95 #define SYSCTL_INIT() sx_init(&sysctllock, "sysctl lock") 96 #define SYSCTL_SLEEP(ch, wmesg, timo) \ 97 sx_sleep(ch, &sysctllock, 0, wmesg, timo) 98 99 static int sysctl_root(SYSCTL_HANDLER_ARGS); 100 101 struct sysctl_oid_list sysctl__children; /* root list */ 102 103 static int sysctl_remove_oid_locked(struct sysctl_oid *oidp, int del, 104 int recurse); 105 106 static struct sysctl_oid * 107 sysctl_find_oidname(const char *name, struct sysctl_oid_list *list) 108 { 109 struct sysctl_oid *oidp; 110 111 SYSCTL_ASSERT_XLOCKED(); 112 SLIST_FOREACH(oidp, list, oid_link) { 113 if (strcmp(oidp->oid_name, name) == 0) { 114 return (oidp); 115 } 116 } 117 return (NULL); 118 } 119 120 /* 121 * Initialization of the MIB tree. 122 * 123 * Order by number in each list. 124 */ 125 void 126 sysctl_lock(void) 127 { 128 129 SYSCTL_XLOCK(); 130 } 131 132 void 133 sysctl_unlock(void) 134 { 135 136 SYSCTL_XUNLOCK(); 137 } 138 139 void 140 sysctl_register_oid(struct sysctl_oid *oidp) 141 { 142 struct sysctl_oid_list *parent = oidp->oid_parent; 143 struct sysctl_oid *p; 144 struct sysctl_oid *q; 145 int oid_number; 146 int timeout = 2; 147 148 /* 149 * First check if another oid with the same name already 150 * exists in the parent's list. 151 */ 152 SYSCTL_ASSERT_XLOCKED(); 153 p = sysctl_find_oidname(oidp->oid_name, parent); 154 if (p != NULL) { 155 if ((p->oid_kind & CTLTYPE) == CTLTYPE_NODE) { 156 p->oid_refcnt++; 157 return; 158 } else { 159 printf("can't re-use a leaf (%s)!\n", p->oid_name); 160 return; 161 } 162 } 163 /* get current OID number */ 164 oid_number = oidp->oid_number; 165 166 #if (OID_AUTO >= 0) 167 #error "OID_AUTO is expected to be a negative value" 168 #endif 169 /* 170 * Any negative OID number qualifies as OID_AUTO. Valid OID 171 * numbers should always be positive. 172 * 173 * NOTE: DO NOT change the starting value here, change it in 174 * <sys/sysctl.h>, and make sure it is at least 256 to 175 * accomodate e.g. net.inet.raw as a static sysctl node. 176 */ 177 if (oid_number < 0) { 178 static int newoid; 179 180 /* 181 * By decrementing the next OID number we spend less 182 * time inserting the OIDs into a sorted list. 183 */ 184 if (--newoid < CTL_AUTO_START) 185 newoid = 0x7fffffff; 186 187 oid_number = newoid; 188 } 189 190 /* 191 * Insert the OID into the parent's list sorted by OID number. 192 */ 193 retry: 194 q = NULL; 195 SLIST_FOREACH(p, parent, oid_link) { 196 /* check if the current OID number is in use */ 197 if (oid_number == p->oid_number) { 198 /* get the next valid OID number */ 199 if (oid_number < CTL_AUTO_START || 200 oid_number == 0x7fffffff) { 201 /* wraparound - restart */ 202 oid_number = CTL_AUTO_START; 203 /* don't loop forever */ 204 if (!timeout--) 205 panic("sysctl: Out of OID numbers\n"); 206 goto retry; 207 } else { 208 oid_number++; 209 } 210 } else if (oid_number < p->oid_number) 211 break; 212 q = p; 213 } 214 /* check for non-auto OID number collision */ 215 if (oidp->oid_number >= 0 && oidp->oid_number < CTL_AUTO_START && 216 oid_number >= CTL_AUTO_START) { 217 printf("sysctl: OID number(%d) is already in use for '%s'\n", 218 oidp->oid_number, oidp->oid_name); 219 } 220 /* update the OID number, if any */ 221 oidp->oid_number = oid_number; 222 if (q != NULL) 223 SLIST_INSERT_AFTER(q, oidp, oid_link); 224 else 225 SLIST_INSERT_HEAD(parent, oidp, oid_link); 226 } 227 228 void 229 sysctl_unregister_oid(struct sysctl_oid *oidp) 230 { 231 struct sysctl_oid *p; 232 int error; 233 234 SYSCTL_ASSERT_XLOCKED(); 235 error = ENOENT; 236 if (oidp->oid_number == OID_AUTO) { 237 error = EINVAL; 238 } else { 239 SLIST_FOREACH(p, oidp->oid_parent, oid_link) { 240 if (p == oidp) { 241 SLIST_REMOVE(oidp->oid_parent, oidp, 242 sysctl_oid, oid_link); 243 error = 0; 244 break; 245 } 246 } 247 } 248 249 /* 250 * This can happen when a module fails to register and is 251 * being unloaded afterwards. It should not be a panic() 252 * for normal use. 253 */ 254 if (error) 255 printf("%s: failed to unregister sysctl\n", __func__); 256 } 257 258 /* Initialize a new context to keep track of dynamically added sysctls. */ 259 int 260 sysctl_ctx_init(struct sysctl_ctx_list *c) 261 { 262 263 if (c == NULL) { 264 return (EINVAL); 265 } 266 267 /* 268 * No locking here, the caller is responsible for not adding 269 * new nodes to a context until after this function has 270 * returned. 271 */ 272 TAILQ_INIT(c); 273 return (0); 274 } 275 276 /* Free the context, and destroy all dynamic oids registered in this context */ 277 int 278 sysctl_ctx_free(struct sysctl_ctx_list *clist) 279 { 280 struct sysctl_ctx_entry *e, *e1; 281 int error; 282 283 error = 0; 284 /* 285 * First perform a "dry run" to check if it's ok to remove oids. 286 * XXX FIXME 287 * XXX This algorithm is a hack. But I don't know any 288 * XXX better solution for now... 289 */ 290 SYSCTL_XLOCK(); 291 TAILQ_FOREACH(e, clist, link) { 292 error = sysctl_remove_oid_locked(e->entry, 0, 0); 293 if (error) 294 break; 295 } 296 /* 297 * Restore deregistered entries, either from the end, 298 * or from the place where error occured. 299 * e contains the entry that was not unregistered 300 */ 301 if (error) 302 e1 = TAILQ_PREV(e, sysctl_ctx_list, link); 303 else 304 e1 = TAILQ_LAST(clist, sysctl_ctx_list); 305 while (e1 != NULL) { 306 sysctl_register_oid(e1->entry); 307 e1 = TAILQ_PREV(e1, sysctl_ctx_list, link); 308 } 309 if (error) { 310 SYSCTL_XUNLOCK(); 311 return(EBUSY); 312 } 313 /* Now really delete the entries */ 314 e = TAILQ_FIRST(clist); 315 while (e != NULL) { 316 e1 = TAILQ_NEXT(e, link); 317 error = sysctl_remove_oid_locked(e->entry, 1, 0); 318 if (error) 319 panic("sysctl_remove_oid: corrupt tree, entry: %s", 320 e->entry->oid_name); 321 free(e, M_SYSCTLOID); 322 e = e1; 323 } 324 SYSCTL_XUNLOCK(); 325 return (error); 326 } 327 328 /* Add an entry to the context */ 329 struct sysctl_ctx_entry * 330 sysctl_ctx_entry_add(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp) 331 { 332 struct sysctl_ctx_entry *e; 333 334 SYSCTL_ASSERT_XLOCKED(); 335 if (clist == NULL || oidp == NULL) 336 return(NULL); 337 e = malloc(sizeof(struct sysctl_ctx_entry), M_SYSCTLOID, M_WAITOK); 338 e->entry = oidp; 339 TAILQ_INSERT_HEAD(clist, e, link); 340 return (e); 341 } 342 343 /* Find an entry in the context */ 344 struct sysctl_ctx_entry * 345 sysctl_ctx_entry_find(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp) 346 { 347 struct sysctl_ctx_entry *e; 348 349 SYSCTL_ASSERT_XLOCKED(); 350 if (clist == NULL || oidp == NULL) 351 return(NULL); 352 TAILQ_FOREACH(e, clist, link) { 353 if(e->entry == oidp) 354 return(e); 355 } 356 return (e); 357 } 358 359 /* 360 * Delete an entry from the context. 361 * NOTE: this function doesn't free oidp! You have to remove it 362 * with sysctl_remove_oid(). 363 */ 364 int 365 sysctl_ctx_entry_del(struct sysctl_ctx_list *clist, struct sysctl_oid *oidp) 366 { 367 struct sysctl_ctx_entry *e; 368 369 if (clist == NULL || oidp == NULL) 370 return (EINVAL); 371 SYSCTL_XLOCK(); 372 e = sysctl_ctx_entry_find(clist, oidp); 373 if (e != NULL) { 374 TAILQ_REMOVE(clist, e, link); 375 SYSCTL_XUNLOCK(); 376 free(e, M_SYSCTLOID); 377 return (0); 378 } else { 379 SYSCTL_XUNLOCK(); 380 return (ENOENT); 381 } 382 } 383 384 /* 385 * Remove dynamically created sysctl trees. 386 * oidp - top of the tree to be removed 387 * del - if 0 - just deregister, otherwise free up entries as well 388 * recurse - if != 0 traverse the subtree to be deleted 389 */ 390 int 391 sysctl_remove_oid(struct sysctl_oid *oidp, int del, int recurse) 392 { 393 int error; 394 395 SYSCTL_XLOCK(); 396 error = sysctl_remove_oid_locked(oidp, del, recurse); 397 SYSCTL_XUNLOCK(); 398 return (error); 399 } 400 401 int 402 sysctl_remove_name(struct sysctl_oid *parent, const char *name, 403 int del, int recurse) 404 { 405 struct sysctl_oid *p, *tmp; 406 int error; 407 408 error = ENOENT; 409 SYSCTL_XLOCK(); 410 SLIST_FOREACH_SAFE(p, SYSCTL_CHILDREN(parent), oid_link, tmp) { 411 if (strcmp(p->oid_name, name) == 0) { 412 error = sysctl_remove_oid_locked(p, del, recurse); 413 break; 414 } 415 } 416 SYSCTL_XUNLOCK(); 417 418 return (error); 419 } 420 421 422 static int 423 sysctl_remove_oid_locked(struct sysctl_oid *oidp, int del, int recurse) 424 { 425 struct sysctl_oid *p, *tmp; 426 int error; 427 428 SYSCTL_ASSERT_XLOCKED(); 429 if (oidp == NULL) 430 return(EINVAL); 431 if ((oidp->oid_kind & CTLFLAG_DYN) == 0) { 432 printf("can't remove non-dynamic nodes!\n"); 433 return (EINVAL); 434 } 435 /* 436 * WARNING: normal method to do this should be through 437 * sysctl_ctx_free(). Use recursing as the last resort 438 * method to purge your sysctl tree of leftovers... 439 * However, if some other code still references these nodes, 440 * it will panic. 441 */ 442 if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) { 443 if (oidp->oid_refcnt == 1) { 444 SLIST_FOREACH_SAFE(p, 445 SYSCTL_CHILDREN(oidp), oid_link, tmp) { 446 if (!recurse) 447 return (ENOTEMPTY); 448 error = sysctl_remove_oid_locked(p, del, 449 recurse); 450 if (error) 451 return (error); 452 } 453 if (del) 454 free(SYSCTL_CHILDREN(oidp), M_SYSCTLOID); 455 } 456 } 457 if (oidp->oid_refcnt > 1 ) { 458 oidp->oid_refcnt--; 459 } else { 460 if (oidp->oid_refcnt == 0) { 461 printf("Warning: bad oid_refcnt=%u (%s)!\n", 462 oidp->oid_refcnt, oidp->oid_name); 463 return (EINVAL); 464 } 465 sysctl_unregister_oid(oidp); 466 if (del) { 467 /* 468 * Wait for all threads running the handler to drain. 469 * This preserves the previous behavior when the 470 * sysctl lock was held across a handler invocation, 471 * and is necessary for module unload correctness. 472 */ 473 while (oidp->oid_running > 0) { 474 oidp->oid_kind |= CTLFLAG_DYING; 475 SYSCTL_SLEEP(&oidp->oid_running, "oidrm", 0); 476 } 477 if (oidp->oid_descr) 478 free(__DECONST(char *, oidp->oid_descr), 479 M_SYSCTLOID); 480 free(__DECONST(char *, oidp->oid_name), M_SYSCTLOID); 481 free(oidp, M_SYSCTLOID); 482 } 483 } 484 return (0); 485 } 486 /* 487 * Create new sysctls at run time. 488 * clist may point to a valid context initialized with sysctl_ctx_init(). 489 */ 490 struct sysctl_oid * 491 sysctl_add_oid(struct sysctl_ctx_list *clist, struct sysctl_oid_list *parent, 492 int number, const char *name, int kind, void *arg1, intptr_t arg2, 493 int (*handler)(SYSCTL_HANDLER_ARGS), const char *fmt, const char *descr) 494 { 495 struct sysctl_oid *oidp; 496 497 /* You have to hook up somewhere.. */ 498 if (parent == NULL) 499 return(NULL); 500 /* Check if the node already exists, otherwise create it */ 501 SYSCTL_XLOCK(); 502 oidp = sysctl_find_oidname(name, parent); 503 if (oidp != NULL) { 504 if ((oidp->oid_kind & CTLTYPE) == CTLTYPE_NODE) { 505 oidp->oid_refcnt++; 506 /* Update the context */ 507 if (clist != NULL) 508 sysctl_ctx_entry_add(clist, oidp); 509 SYSCTL_XUNLOCK(); 510 return (oidp); 511 } else { 512 SYSCTL_XUNLOCK(); 513 printf("can't re-use a leaf (%s)!\n", name); 514 return (NULL); 515 } 516 } 517 oidp = malloc(sizeof(struct sysctl_oid), M_SYSCTLOID, M_WAITOK|M_ZERO); 518 oidp->oid_parent = parent; 519 SLIST_NEXT(oidp, oid_link) = NULL; 520 oidp->oid_number = number; 521 oidp->oid_refcnt = 1; 522 oidp->oid_name = strdup(name, M_SYSCTLOID); 523 oidp->oid_handler = handler; 524 oidp->oid_kind = CTLFLAG_DYN | kind; 525 if ((kind & CTLTYPE) == CTLTYPE_NODE) { 526 /* Allocate space for children */ 527 SYSCTL_CHILDREN_SET(oidp, malloc(sizeof(struct sysctl_oid_list), 528 M_SYSCTLOID, M_WAITOK)); 529 SLIST_INIT(SYSCTL_CHILDREN(oidp)); 530 oidp->oid_arg2 = arg2; 531 } else { 532 oidp->oid_arg1 = arg1; 533 oidp->oid_arg2 = arg2; 534 } 535 oidp->oid_fmt = fmt; 536 if (descr) 537 oidp->oid_descr = strdup(descr, M_SYSCTLOID); 538 /* Update the context, if used */ 539 if (clist != NULL) 540 sysctl_ctx_entry_add(clist, oidp); 541 /* Register this oid */ 542 sysctl_register_oid(oidp); 543 SYSCTL_XUNLOCK(); 544 return (oidp); 545 } 546 547 /* 548 * Rename an existing oid. 549 */ 550 void 551 sysctl_rename_oid(struct sysctl_oid *oidp, const char *name) 552 { 553 char *newname; 554 char *oldname; 555 556 newname = strdup(name, M_SYSCTLOID); 557 SYSCTL_XLOCK(); 558 oldname = __DECONST(char *, oidp->oid_name); 559 oidp->oid_name = newname; 560 SYSCTL_XUNLOCK(); 561 free(oldname, M_SYSCTLOID); 562 } 563 564 /* 565 * Reparent an existing oid. 566 */ 567 int 568 sysctl_move_oid(struct sysctl_oid *oid, struct sysctl_oid_list *parent) 569 { 570 struct sysctl_oid *oidp; 571 572 SYSCTL_XLOCK(); 573 if (oid->oid_parent == parent) { 574 SYSCTL_XUNLOCK(); 575 return (0); 576 } 577 oidp = sysctl_find_oidname(oid->oid_name, parent); 578 if (oidp != NULL) { 579 SYSCTL_XUNLOCK(); 580 return (EEXIST); 581 } 582 sysctl_unregister_oid(oid); 583 oid->oid_parent = parent; 584 oid->oid_number = OID_AUTO; 585 sysctl_register_oid(oid); 586 SYSCTL_XUNLOCK(); 587 return (0); 588 } 589 590 /* 591 * Register the kernel's oids on startup. 592 */ 593 SET_DECLARE(sysctl_set, struct sysctl_oid); 594 595 static void 596 sysctl_register_all(void *arg) 597 { 598 struct sysctl_oid **oidp; 599 600 sx_init(&sysctlmemlock, "sysctl mem"); 601 SYSCTL_INIT(); 602 SYSCTL_XLOCK(); 603 SET_FOREACH(oidp, sysctl_set) 604 sysctl_register_oid(*oidp); 605 SYSCTL_XUNLOCK(); 606 } 607 SYSINIT(sysctl, SI_SUB_KMEM, SI_ORDER_ANY, sysctl_register_all, 0); 608 609 /* 610 * "Staff-functions" 611 * 612 * These functions implement a presently undocumented interface 613 * used by the sysctl program to walk the tree, and get the type 614 * so it can print the value. 615 * This interface is under work and consideration, and should probably 616 * be killed with a big axe by the first person who can find the time. 617 * (be aware though, that the proper interface isn't as obvious as it 618 * may seem, there are various conflicting requirements. 619 * 620 * {0,0} printf the entire MIB-tree. 621 * {0,1,...} return the name of the "..." OID. 622 * {0,2,...} return the next OID. 623 * {0,3} return the OID of the name in "new" 624 * {0,4,...} return the kind & format info for the "..." OID. 625 * {0,5,...} return the description the "..." OID. 626 */ 627 628 #ifdef SYSCTL_DEBUG 629 static void 630 sysctl_sysctl_debug_dump_node(struct sysctl_oid_list *l, int i) 631 { 632 int k; 633 struct sysctl_oid *oidp; 634 635 SYSCTL_ASSERT_XLOCKED(); 636 SLIST_FOREACH(oidp, l, oid_link) { 637 638 for (k=0; k<i; k++) 639 printf(" "); 640 641 printf("%d %s ", oidp->oid_number, oidp->oid_name); 642 643 printf("%c%c", 644 oidp->oid_kind & CTLFLAG_RD ? 'R':' ', 645 oidp->oid_kind & CTLFLAG_WR ? 'W':' '); 646 647 if (oidp->oid_handler) 648 printf(" *Handler"); 649 650 switch (oidp->oid_kind & CTLTYPE) { 651 case CTLTYPE_NODE: 652 printf(" Node\n"); 653 if (!oidp->oid_handler) { 654 sysctl_sysctl_debug_dump_node( 655 oidp->oid_arg1, i+2); 656 } 657 break; 658 case CTLTYPE_INT: printf(" Int\n"); break; 659 case CTLTYPE_UINT: printf(" u_int\n"); break; 660 case CTLTYPE_LONG: printf(" Long\n"); break; 661 case CTLTYPE_ULONG: printf(" u_long\n"); break; 662 case CTLTYPE_STRING: printf(" String\n"); break; 663 case CTLTYPE_U64: printf(" uint64_t\n"); break; 664 case CTLTYPE_S64: printf(" int64_t\n"); break; 665 case CTLTYPE_OPAQUE: printf(" Opaque/struct\n"); break; 666 default: printf("\n"); 667 } 668 669 } 670 } 671 672 static int 673 sysctl_sysctl_debug(SYSCTL_HANDLER_ARGS) 674 { 675 int error; 676 677 error = priv_check(req->td, PRIV_SYSCTL_DEBUG); 678 if (error) 679 return (error); 680 SYSCTL_XLOCK(); 681 sysctl_sysctl_debug_dump_node(&sysctl__children, 0); 682 SYSCTL_XUNLOCK(); 683 return (ENOENT); 684 } 685 686 SYSCTL_PROC(_sysctl, 0, debug, CTLTYPE_STRING|CTLFLAG_RD, 687 0, 0, sysctl_sysctl_debug, "-", ""); 688 #endif 689 690 static int 691 sysctl_sysctl_name(SYSCTL_HANDLER_ARGS) 692 { 693 int *name = (int *) arg1; 694 u_int namelen = arg2; 695 int error = 0; 696 struct sysctl_oid *oid; 697 struct sysctl_oid_list *lsp = &sysctl__children, *lsp2; 698 char buf[10]; 699 700 SYSCTL_XLOCK(); 701 while (namelen) { 702 if (!lsp) { 703 snprintf(buf,sizeof(buf),"%d",*name); 704 if (req->oldidx) 705 error = SYSCTL_OUT(req, ".", 1); 706 if (!error) 707 error = SYSCTL_OUT(req, buf, strlen(buf)); 708 if (error) 709 goto out; 710 namelen--; 711 name++; 712 continue; 713 } 714 lsp2 = 0; 715 SLIST_FOREACH(oid, lsp, oid_link) { 716 if (oid->oid_number != *name) 717 continue; 718 719 if (req->oldidx) 720 error = SYSCTL_OUT(req, ".", 1); 721 if (!error) 722 error = SYSCTL_OUT(req, oid->oid_name, 723 strlen(oid->oid_name)); 724 if (error) 725 goto out; 726 727 namelen--; 728 name++; 729 730 if ((oid->oid_kind & CTLTYPE) != CTLTYPE_NODE) 731 break; 732 733 if (oid->oid_handler) 734 break; 735 736 lsp2 = SYSCTL_CHILDREN(oid); 737 break; 738 } 739 lsp = lsp2; 740 } 741 error = SYSCTL_OUT(req, "", 1); 742 out: 743 SYSCTL_XUNLOCK(); 744 return (error); 745 } 746 747 /* 748 * XXXRW/JA: Shouldn't return name data for nodes that we don't permit in 749 * capability mode. 750 */ 751 static SYSCTL_NODE(_sysctl, 1, name, CTLFLAG_RD | CTLFLAG_CAPRD, 752 sysctl_sysctl_name, ""); 753 754 static int 755 sysctl_sysctl_next_ls(struct sysctl_oid_list *lsp, int *name, u_int namelen, 756 int *next, int *len, int level, struct sysctl_oid **oidpp) 757 { 758 struct sysctl_oid *oidp; 759 760 SYSCTL_ASSERT_XLOCKED(); 761 *len = level; 762 SLIST_FOREACH(oidp, lsp, oid_link) { 763 *next = oidp->oid_number; 764 *oidpp = oidp; 765 766 if (oidp->oid_kind & CTLFLAG_SKIP) 767 continue; 768 769 if (!namelen) { 770 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE) 771 return (0); 772 if (oidp->oid_handler) 773 /* We really should call the handler here...*/ 774 return (0); 775 lsp = SYSCTL_CHILDREN(oidp); 776 if (!sysctl_sysctl_next_ls(lsp, 0, 0, next+1, 777 len, level+1, oidpp)) 778 return (0); 779 goto emptynode; 780 } 781 782 if (oidp->oid_number < *name) 783 continue; 784 785 if (oidp->oid_number > *name) { 786 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE) 787 return (0); 788 if (oidp->oid_handler) 789 return (0); 790 lsp = SYSCTL_CHILDREN(oidp); 791 if (!sysctl_sysctl_next_ls(lsp, name+1, namelen-1, 792 next+1, len, level+1, oidpp)) 793 return (0); 794 goto next; 795 } 796 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE) 797 continue; 798 799 if (oidp->oid_handler) 800 continue; 801 802 lsp = SYSCTL_CHILDREN(oidp); 803 if (!sysctl_sysctl_next_ls(lsp, name+1, namelen-1, next+1, 804 len, level+1, oidpp)) 805 return (0); 806 next: 807 namelen = 1; 808 emptynode: 809 *len = level; 810 } 811 return (1); 812 } 813 814 static int 815 sysctl_sysctl_next(SYSCTL_HANDLER_ARGS) 816 { 817 int *name = (int *) arg1; 818 u_int namelen = arg2; 819 int i, j, error; 820 struct sysctl_oid *oid; 821 struct sysctl_oid_list *lsp = &sysctl__children; 822 int newoid[CTL_MAXNAME]; 823 824 SYSCTL_XLOCK(); 825 i = sysctl_sysctl_next_ls(lsp, name, namelen, newoid, &j, 1, &oid); 826 SYSCTL_XUNLOCK(); 827 if (i) 828 return (ENOENT); 829 error = SYSCTL_OUT(req, newoid, j * sizeof (int)); 830 return (error); 831 } 832 833 /* 834 * XXXRW/JA: Shouldn't return next data for nodes that we don't permit in 835 * capability mode. 836 */ 837 static SYSCTL_NODE(_sysctl, 2, next, CTLFLAG_RD | CTLFLAG_CAPRD, 838 sysctl_sysctl_next, ""); 839 840 static int 841 name2oid(char *name, int *oid, int *len, struct sysctl_oid **oidpp) 842 { 843 struct sysctl_oid *oidp; 844 struct sysctl_oid_list *lsp = &sysctl__children; 845 char *p; 846 847 SYSCTL_ASSERT_XLOCKED(); 848 849 for (*len = 0; *len < CTL_MAXNAME;) { 850 p = strsep(&name, "."); 851 852 oidp = SLIST_FIRST(lsp); 853 for (;; oidp = SLIST_NEXT(oidp, oid_link)) { 854 if (oidp == NULL) 855 return (ENOENT); 856 if (strcmp(p, oidp->oid_name) == 0) 857 break; 858 } 859 *oid++ = oidp->oid_number; 860 (*len)++; 861 862 if (name == NULL || *name == '\0') { 863 if (oidpp) 864 *oidpp = oidp; 865 return (0); 866 } 867 868 if ((oidp->oid_kind & CTLTYPE) != CTLTYPE_NODE) 869 break; 870 871 if (oidp->oid_handler) 872 break; 873 874 lsp = SYSCTL_CHILDREN(oidp); 875 } 876 return (ENOENT); 877 } 878 879 static int 880 sysctl_sysctl_name2oid(SYSCTL_HANDLER_ARGS) 881 { 882 char *p; 883 int error, oid[CTL_MAXNAME], len = 0; 884 struct sysctl_oid *op = 0; 885 886 if (!req->newlen) 887 return (ENOENT); 888 if (req->newlen >= MAXPATHLEN) /* XXX arbitrary, undocumented */ 889 return (ENAMETOOLONG); 890 891 p = malloc(req->newlen+1, M_SYSCTL, M_WAITOK); 892 893 error = SYSCTL_IN(req, p, req->newlen); 894 if (error) { 895 free(p, M_SYSCTL); 896 return (error); 897 } 898 899 p [req->newlen] = '\0'; 900 901 SYSCTL_XLOCK(); 902 error = name2oid(p, oid, &len, &op); 903 SYSCTL_XUNLOCK(); 904 905 free(p, M_SYSCTL); 906 907 if (error) 908 return (error); 909 910 error = SYSCTL_OUT(req, oid, len * sizeof *oid); 911 return (error); 912 } 913 914 /* 915 * XXXRW/JA: Shouldn't return name2oid data for nodes that we don't permit in 916 * capability mode. 917 */ 918 SYSCTL_PROC(_sysctl, 3, name2oid, 919 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE 920 | CTLFLAG_CAPRW, 0, 0, sysctl_sysctl_name2oid, "I", ""); 921 922 static int 923 sysctl_sysctl_oidfmt(SYSCTL_HANDLER_ARGS) 924 { 925 struct sysctl_oid *oid; 926 int error; 927 928 SYSCTL_XLOCK(); 929 error = sysctl_find_oid(arg1, arg2, &oid, NULL, req); 930 if (error) 931 goto out; 932 933 if (oid->oid_fmt == NULL) { 934 error = ENOENT; 935 goto out; 936 } 937 error = SYSCTL_OUT(req, &oid->oid_kind, sizeof(oid->oid_kind)); 938 if (error) 939 goto out; 940 error = SYSCTL_OUT(req, oid->oid_fmt, strlen(oid->oid_fmt) + 1); 941 out: 942 SYSCTL_XUNLOCK(); 943 return (error); 944 } 945 946 947 static SYSCTL_NODE(_sysctl, 4, oidfmt, CTLFLAG_RD|CTLFLAG_MPSAFE|CTLFLAG_CAPRD, 948 sysctl_sysctl_oidfmt, ""); 949 950 static int 951 sysctl_sysctl_oiddescr(SYSCTL_HANDLER_ARGS) 952 { 953 struct sysctl_oid *oid; 954 int error; 955 956 SYSCTL_XLOCK(); 957 error = sysctl_find_oid(arg1, arg2, &oid, NULL, req); 958 if (error) 959 goto out; 960 961 if (oid->oid_descr == NULL) { 962 error = ENOENT; 963 goto out; 964 } 965 error = SYSCTL_OUT(req, oid->oid_descr, strlen(oid->oid_descr) + 1); 966 out: 967 SYSCTL_XUNLOCK(); 968 return (error); 969 } 970 971 static SYSCTL_NODE(_sysctl, 5, oiddescr, CTLFLAG_RD|CTLFLAG_CAPRD, 972 sysctl_sysctl_oiddescr, ""); 973 974 /* 975 * Default "handler" functions. 976 */ 977 978 /* 979 * Handle an int, signed or unsigned. 980 * Two cases: 981 * a variable: point arg1 at it. 982 * a constant: pass it in arg2. 983 */ 984 985 int 986 sysctl_handle_int(SYSCTL_HANDLER_ARGS) 987 { 988 int tmpout, error = 0; 989 990 /* 991 * Attempt to get a coherent snapshot by making a copy of the data. 992 */ 993 if (arg1) 994 tmpout = *(int *)arg1; 995 else 996 tmpout = arg2; 997 error = SYSCTL_OUT(req, &tmpout, sizeof(int)); 998 999 if (error || !req->newptr) 1000 return (error); 1001 1002 if (!arg1) 1003 error = EPERM; 1004 else 1005 error = SYSCTL_IN(req, arg1, sizeof(int)); 1006 return (error); 1007 } 1008 1009 /* 1010 * Based on on sysctl_handle_int() convert milliseconds into ticks. 1011 * Note: this is used by TCP. 1012 */ 1013 1014 int 1015 sysctl_msec_to_ticks(SYSCTL_HANDLER_ARGS) 1016 { 1017 int error, s, tt; 1018 1019 tt = *(int *)arg1; 1020 s = (int)((int64_t)tt * 1000 / hz); 1021 1022 error = sysctl_handle_int(oidp, &s, 0, req); 1023 if (error || !req->newptr) 1024 return (error); 1025 1026 tt = (int)((int64_t)s * hz / 1000); 1027 if (tt < 1) 1028 return (EINVAL); 1029 1030 *(int *)arg1 = tt; 1031 return (0); 1032 } 1033 1034 1035 /* 1036 * Handle a long, signed or unsigned. 1037 * Two cases: 1038 * a variable: point arg1 at it. 1039 * a constant: pass it in arg2. 1040 */ 1041 1042 int 1043 sysctl_handle_long(SYSCTL_HANDLER_ARGS) 1044 { 1045 int error = 0; 1046 long tmplong; 1047 #ifdef SCTL_MASK32 1048 int tmpint; 1049 #endif 1050 1051 /* 1052 * Attempt to get a coherent snapshot by making a copy of the data. 1053 */ 1054 if (arg1) 1055 tmplong = *(long *)arg1; 1056 else 1057 tmplong = arg2; 1058 #ifdef SCTL_MASK32 1059 if (req->flags & SCTL_MASK32) { 1060 tmpint = tmplong; 1061 error = SYSCTL_OUT(req, &tmpint, sizeof(int)); 1062 } else 1063 #endif 1064 error = SYSCTL_OUT(req, &tmplong, sizeof(long)); 1065 1066 if (error || !req->newptr) 1067 return (error); 1068 1069 if (!arg1) 1070 error = EPERM; 1071 #ifdef SCTL_MASK32 1072 else if (req->flags & SCTL_MASK32) { 1073 error = SYSCTL_IN(req, &tmpint, sizeof(int)); 1074 *(long *)arg1 = (long)tmpint; 1075 } 1076 #endif 1077 else 1078 error = SYSCTL_IN(req, arg1, sizeof(long)); 1079 return (error); 1080 } 1081 1082 /* 1083 * Handle a 64 bit int, signed or unsigned. 1084 * Two cases: 1085 * a variable: point arg1 at it. 1086 * a constant: pass it in arg2. 1087 */ 1088 int 1089 sysctl_handle_64(SYSCTL_HANDLER_ARGS) 1090 { 1091 int error = 0; 1092 uint64_t tmpout; 1093 1094 /* 1095 * Attempt to get a coherent snapshot by making a copy of the data. 1096 */ 1097 if (arg1) 1098 tmpout = *(uint64_t *)arg1; 1099 else 1100 tmpout = arg2; 1101 error = SYSCTL_OUT(req, &tmpout, sizeof(uint64_t)); 1102 1103 if (error || !req->newptr) 1104 return (error); 1105 1106 if (!arg1) 1107 error = EPERM; 1108 else 1109 error = SYSCTL_IN(req, arg1, sizeof(uint64_t)); 1110 return (error); 1111 } 1112 1113 /* 1114 * Handle our generic '\0' terminated 'C' string. 1115 * Two cases: 1116 * a variable string: point arg1 at it, arg2 is max length. 1117 * a constant string: point arg1 at it, arg2 is zero. 1118 */ 1119 1120 int 1121 sysctl_handle_string(SYSCTL_HANDLER_ARGS) 1122 { 1123 int error=0; 1124 char *tmparg; 1125 size_t outlen; 1126 1127 /* 1128 * Attempt to get a coherent snapshot by copying to a 1129 * temporary kernel buffer. 1130 */ 1131 retry: 1132 outlen = strlen((char *)arg1)+1; 1133 tmparg = malloc(outlen, M_SYSCTLTMP, M_WAITOK); 1134 1135 if (strlcpy(tmparg, (char *)arg1, outlen) >= outlen) { 1136 free(tmparg, M_SYSCTLTMP); 1137 goto retry; 1138 } 1139 1140 error = SYSCTL_OUT(req, tmparg, outlen); 1141 free(tmparg, M_SYSCTLTMP); 1142 1143 if (error || !req->newptr) 1144 return (error); 1145 1146 if ((req->newlen - req->newidx) >= arg2) { 1147 error = EINVAL; 1148 } else { 1149 arg2 = (req->newlen - req->newidx); 1150 error = SYSCTL_IN(req, arg1, arg2); 1151 ((char *)arg1)[arg2] = '\0'; 1152 } 1153 1154 return (error); 1155 } 1156 1157 /* 1158 * Handle any kind of opaque data. 1159 * arg1 points to it, arg2 is the size. 1160 */ 1161 1162 int 1163 sysctl_handle_opaque(SYSCTL_HANDLER_ARGS) 1164 { 1165 int error, tries; 1166 u_int generation; 1167 struct sysctl_req req2; 1168 1169 /* 1170 * Attempt to get a coherent snapshot, by using the thread 1171 * pre-emption counter updated from within mi_switch() to 1172 * determine if we were pre-empted during a bcopy() or 1173 * copyout(). Make 3 attempts at doing this before giving up. 1174 * If we encounter an error, stop immediately. 1175 */ 1176 tries = 0; 1177 req2 = *req; 1178 retry: 1179 generation = curthread->td_generation; 1180 error = SYSCTL_OUT(req, arg1, arg2); 1181 if (error) 1182 return (error); 1183 tries++; 1184 if (generation != curthread->td_generation && tries < 3) { 1185 *req = req2; 1186 goto retry; 1187 } 1188 1189 error = SYSCTL_IN(req, arg1, arg2); 1190 1191 return (error); 1192 } 1193 1194 /* 1195 * Transfer functions to/from kernel space. 1196 * XXX: rather untested at this point 1197 */ 1198 static int 1199 sysctl_old_kernel(struct sysctl_req *req, const void *p, size_t l) 1200 { 1201 size_t i = 0; 1202 1203 if (req->oldptr) { 1204 i = l; 1205 if (req->oldlen <= req->oldidx) 1206 i = 0; 1207 else 1208 if (i > req->oldlen - req->oldidx) 1209 i = req->oldlen - req->oldidx; 1210 if (i > 0) 1211 bcopy(p, (char *)req->oldptr + req->oldidx, i); 1212 } 1213 req->oldidx += l; 1214 if (req->oldptr && i != l) 1215 return (ENOMEM); 1216 return (0); 1217 } 1218 1219 static int 1220 sysctl_new_kernel(struct sysctl_req *req, void *p, size_t l) 1221 { 1222 if (!req->newptr) 1223 return (0); 1224 if (req->newlen - req->newidx < l) 1225 return (EINVAL); 1226 bcopy((char *)req->newptr + req->newidx, p, l); 1227 req->newidx += l; 1228 return (0); 1229 } 1230 1231 int 1232 kernel_sysctl(struct thread *td, int *name, u_int namelen, void *old, 1233 size_t *oldlenp, void *new, size_t newlen, size_t *retval, int flags) 1234 { 1235 int error = 0; 1236 struct sysctl_req req; 1237 1238 bzero(&req, sizeof req); 1239 1240 req.td = td; 1241 req.flags = flags; 1242 1243 if (oldlenp) { 1244 req.oldlen = *oldlenp; 1245 } 1246 req.validlen = req.oldlen; 1247 1248 if (old) { 1249 req.oldptr= old; 1250 } 1251 1252 if (new != NULL) { 1253 req.newlen = newlen; 1254 req.newptr = new; 1255 } 1256 1257 req.oldfunc = sysctl_old_kernel; 1258 req.newfunc = sysctl_new_kernel; 1259 req.lock = REQ_UNWIRED; 1260 1261 SYSCTL_XLOCK(); 1262 error = sysctl_root(0, name, namelen, &req); 1263 SYSCTL_XUNLOCK(); 1264 1265 if (req.lock == REQ_WIRED && req.validlen > 0) 1266 vsunlock(req.oldptr, req.validlen); 1267 1268 if (error && error != ENOMEM) 1269 return (error); 1270 1271 if (retval) { 1272 if (req.oldptr && req.oldidx > req.validlen) 1273 *retval = req.validlen; 1274 else 1275 *retval = req.oldidx; 1276 } 1277 return (error); 1278 } 1279 1280 int 1281 kernel_sysctlbyname(struct thread *td, char *name, void *old, size_t *oldlenp, 1282 void *new, size_t newlen, size_t *retval, int flags) 1283 { 1284 int oid[CTL_MAXNAME]; 1285 size_t oidlen, plen; 1286 int error; 1287 1288 oid[0] = 0; /* sysctl internal magic */ 1289 oid[1] = 3; /* name2oid */ 1290 oidlen = sizeof(oid); 1291 1292 error = kernel_sysctl(td, oid, 2, oid, &oidlen, 1293 (void *)name, strlen(name), &plen, flags); 1294 if (error) 1295 return (error); 1296 1297 error = kernel_sysctl(td, oid, plen / sizeof(int), old, oldlenp, 1298 new, newlen, retval, flags); 1299 return (error); 1300 } 1301 1302 /* 1303 * Transfer function to/from user space. 1304 */ 1305 static int 1306 sysctl_old_user(struct sysctl_req *req, const void *p, size_t l) 1307 { 1308 size_t i, len, origidx; 1309 int error; 1310 1311 origidx = req->oldidx; 1312 req->oldidx += l; 1313 if (req->oldptr == NULL) 1314 return (0); 1315 /* 1316 * If we have not wired the user supplied buffer and we are currently 1317 * holding locks, drop a witness warning, as it's possible that 1318 * write operations to the user page can sleep. 1319 */ 1320 if (req->lock != REQ_WIRED) 1321 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL, 1322 "sysctl_old_user()"); 1323 i = l; 1324 len = req->validlen; 1325 if (len <= origidx) 1326 i = 0; 1327 else { 1328 if (i > len - origidx) 1329 i = len - origidx; 1330 if (req->lock == REQ_WIRED) { 1331 error = copyout_nofault(p, (char *)req->oldptr + 1332 origidx, i); 1333 } else 1334 error = copyout(p, (char *)req->oldptr + origidx, i); 1335 if (error != 0) 1336 return (error); 1337 } 1338 if (i < l) 1339 return (ENOMEM); 1340 return (0); 1341 } 1342 1343 static int 1344 sysctl_new_user(struct sysctl_req *req, void *p, size_t l) 1345 { 1346 int error; 1347 1348 if (!req->newptr) 1349 return (0); 1350 if (req->newlen - req->newidx < l) 1351 return (EINVAL); 1352 WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL, 1353 "sysctl_new_user()"); 1354 error = copyin((char *)req->newptr + req->newidx, p, l); 1355 req->newidx += l; 1356 return (error); 1357 } 1358 1359 /* 1360 * Wire the user space destination buffer. If set to a value greater than 1361 * zero, the len parameter limits the maximum amount of wired memory. 1362 */ 1363 int 1364 sysctl_wire_old_buffer(struct sysctl_req *req, size_t len) 1365 { 1366 int ret; 1367 size_t wiredlen; 1368 1369 wiredlen = (len > 0 && len < req->oldlen) ? len : req->oldlen; 1370 ret = 0; 1371 if (req->lock != REQ_WIRED && req->oldptr && 1372 req->oldfunc == sysctl_old_user) { 1373 if (wiredlen != 0) { 1374 ret = vslock(req->oldptr, wiredlen); 1375 if (ret != 0) { 1376 if (ret != ENOMEM) 1377 return (ret); 1378 wiredlen = 0; 1379 } 1380 } 1381 req->lock = REQ_WIRED; 1382 req->validlen = wiredlen; 1383 } 1384 return (0); 1385 } 1386 1387 int 1388 sysctl_find_oid(int *name, u_int namelen, struct sysctl_oid **noid, 1389 int *nindx, struct sysctl_req *req) 1390 { 1391 struct sysctl_oid_list *lsp; 1392 struct sysctl_oid *oid; 1393 int indx; 1394 1395 SYSCTL_ASSERT_XLOCKED(); 1396 lsp = &sysctl__children; 1397 indx = 0; 1398 while (indx < CTL_MAXNAME) { 1399 SLIST_FOREACH(oid, lsp, oid_link) { 1400 if (oid->oid_number == name[indx]) 1401 break; 1402 } 1403 if (oid == NULL) 1404 return (ENOENT); 1405 1406 indx++; 1407 if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) { 1408 if (oid->oid_handler != NULL || indx == namelen) { 1409 *noid = oid; 1410 if (nindx != NULL) 1411 *nindx = indx; 1412 KASSERT((oid->oid_kind & CTLFLAG_DYING) == 0, 1413 ("%s found DYING node %p", __func__, oid)); 1414 return (0); 1415 } 1416 lsp = SYSCTL_CHILDREN(oid); 1417 } else if (indx == namelen) { 1418 *noid = oid; 1419 if (nindx != NULL) 1420 *nindx = indx; 1421 KASSERT((oid->oid_kind & CTLFLAG_DYING) == 0, 1422 ("%s found DYING node %p", __func__, oid)); 1423 return (0); 1424 } else { 1425 return (ENOTDIR); 1426 } 1427 } 1428 return (ENOENT); 1429 } 1430 1431 /* 1432 * Traverse our tree, and find the right node, execute whatever it points 1433 * to, and return the resulting error code. 1434 */ 1435 1436 static int 1437 sysctl_root(SYSCTL_HANDLER_ARGS) 1438 { 1439 struct sysctl_oid *oid; 1440 int error, indx, lvl; 1441 1442 SYSCTL_ASSERT_XLOCKED(); 1443 1444 error = sysctl_find_oid(arg1, arg2, &oid, &indx, req); 1445 if (error) 1446 return (error); 1447 1448 if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) { 1449 /* 1450 * You can't call a sysctl when it's a node, but has 1451 * no handler. Inform the user that it's a node. 1452 * The indx may or may not be the same as namelen. 1453 */ 1454 if (oid->oid_handler == NULL) 1455 return (EISDIR); 1456 } 1457 1458 /* Is this sysctl writable? */ 1459 if (req->newptr && !(oid->oid_kind & CTLFLAG_WR)) 1460 return (EPERM); 1461 1462 KASSERT(req->td != NULL, ("sysctl_root(): req->td == NULL")); 1463 1464 #ifdef CAPABILITY_MODE 1465 /* 1466 * If the process is in capability mode, then don't permit reading or 1467 * writing unless specifically granted for the node. 1468 */ 1469 if (IN_CAPABILITY_MODE(req->td)) { 1470 if (req->oldptr && !(oid->oid_kind & CTLFLAG_CAPRD)) 1471 return (EPERM); 1472 if (req->newptr && !(oid->oid_kind & CTLFLAG_CAPWR)) 1473 return (EPERM); 1474 } 1475 #endif 1476 1477 /* Is this sysctl sensitive to securelevels? */ 1478 if (req->newptr && (oid->oid_kind & CTLFLAG_SECURE)) { 1479 lvl = (oid->oid_kind & CTLMASK_SECURE) >> CTLSHIFT_SECURE; 1480 error = securelevel_gt(req->td->td_ucred, lvl); 1481 if (error) 1482 return (error); 1483 } 1484 1485 /* Is this sysctl writable by only privileged users? */ 1486 if (req->newptr && !(oid->oid_kind & CTLFLAG_ANYBODY)) { 1487 int priv; 1488 1489 if (oid->oid_kind & CTLFLAG_PRISON) 1490 priv = PRIV_SYSCTL_WRITEJAIL; 1491 #ifdef VIMAGE 1492 else if ((oid->oid_kind & CTLFLAG_VNET) && 1493 prison_owns_vnet(req->td->td_ucred)) 1494 priv = PRIV_SYSCTL_WRITEJAIL; 1495 #endif 1496 else 1497 priv = PRIV_SYSCTL_WRITE; 1498 error = priv_check(req->td, priv); 1499 if (error) 1500 return (error); 1501 } 1502 1503 if (!oid->oid_handler) 1504 return (EINVAL); 1505 1506 if ((oid->oid_kind & CTLTYPE) == CTLTYPE_NODE) { 1507 arg1 = (int *)arg1 + indx; 1508 arg2 -= indx; 1509 } else { 1510 arg1 = oid->oid_arg1; 1511 arg2 = oid->oid_arg2; 1512 } 1513 #ifdef MAC 1514 error = mac_system_check_sysctl(req->td->td_ucred, oid, arg1, arg2, 1515 req); 1516 if (error != 0) 1517 return (error); 1518 #endif 1519 oid->oid_running++; 1520 SYSCTL_XUNLOCK(); 1521 1522 if (!(oid->oid_kind & CTLFLAG_MPSAFE)) 1523 mtx_lock(&Giant); 1524 error = oid->oid_handler(oid, arg1, arg2, req); 1525 if (!(oid->oid_kind & CTLFLAG_MPSAFE)) 1526 mtx_unlock(&Giant); 1527 1528 KFAIL_POINT_ERROR(_debug_fail_point, sysctl_running, error); 1529 1530 SYSCTL_XLOCK(); 1531 oid->oid_running--; 1532 if (oid->oid_running == 0 && (oid->oid_kind & CTLFLAG_DYING) != 0) 1533 wakeup(&oid->oid_running); 1534 return (error); 1535 } 1536 1537 #ifndef _SYS_SYSPROTO_H_ 1538 struct sysctl_args { 1539 int *name; 1540 u_int namelen; 1541 void *old; 1542 size_t *oldlenp; 1543 void *new; 1544 size_t newlen; 1545 }; 1546 #endif 1547 int 1548 sys___sysctl(struct thread *td, struct sysctl_args *uap) 1549 { 1550 int error, i, name[CTL_MAXNAME]; 1551 size_t j; 1552 1553 if (uap->namelen > CTL_MAXNAME || uap->namelen < 2) 1554 return (EINVAL); 1555 1556 error = copyin(uap->name, &name, uap->namelen * sizeof(int)); 1557 if (error) 1558 return (error); 1559 1560 error = userland_sysctl(td, name, uap->namelen, 1561 uap->old, uap->oldlenp, 0, 1562 uap->new, uap->newlen, &j, 0); 1563 if (error && error != ENOMEM) 1564 return (error); 1565 if (uap->oldlenp) { 1566 i = copyout(&j, uap->oldlenp, sizeof(j)); 1567 if (i) 1568 return (i); 1569 } 1570 return (error); 1571 } 1572 1573 /* 1574 * This is used from various compatibility syscalls too. That's why name 1575 * must be in kernel space. 1576 */ 1577 int 1578 userland_sysctl(struct thread *td, int *name, u_int namelen, void *old, 1579 size_t *oldlenp, int inkernel, void *new, size_t newlen, size_t *retval, 1580 int flags) 1581 { 1582 int error = 0, memlocked; 1583 struct sysctl_req req; 1584 1585 bzero(&req, sizeof req); 1586 1587 req.td = td; 1588 req.flags = flags; 1589 1590 if (oldlenp) { 1591 if (inkernel) { 1592 req.oldlen = *oldlenp; 1593 } else { 1594 error = copyin(oldlenp, &req.oldlen, sizeof(*oldlenp)); 1595 if (error) 1596 return (error); 1597 } 1598 } 1599 req.validlen = req.oldlen; 1600 1601 if (old) { 1602 if (!useracc(old, req.oldlen, VM_PROT_WRITE)) 1603 return (EFAULT); 1604 req.oldptr= old; 1605 } 1606 1607 if (new != NULL) { 1608 if (!useracc(new, newlen, VM_PROT_READ)) 1609 return (EFAULT); 1610 req.newlen = newlen; 1611 req.newptr = new; 1612 } 1613 1614 req.oldfunc = sysctl_old_user; 1615 req.newfunc = sysctl_new_user; 1616 req.lock = REQ_UNWIRED; 1617 1618 #ifdef KTRACE 1619 if (KTRPOINT(curthread, KTR_SYSCTL)) 1620 ktrsysctl(name, namelen); 1621 #endif 1622 1623 if (req.oldlen > PAGE_SIZE) { 1624 memlocked = 1; 1625 sx_xlock(&sysctlmemlock); 1626 } else 1627 memlocked = 0; 1628 CURVNET_SET(TD_TO_VNET(td)); 1629 1630 for (;;) { 1631 req.oldidx = 0; 1632 req.newidx = 0; 1633 SYSCTL_XLOCK(); 1634 error = sysctl_root(0, name, namelen, &req); 1635 SYSCTL_XUNLOCK(); 1636 if (error != EAGAIN) 1637 break; 1638 kern_yield(PRI_USER); 1639 } 1640 1641 CURVNET_RESTORE(); 1642 1643 if (req.lock == REQ_WIRED && req.validlen > 0) 1644 vsunlock(req.oldptr, req.validlen); 1645 if (memlocked) 1646 sx_xunlock(&sysctlmemlock); 1647 1648 if (error && error != ENOMEM) 1649 return (error); 1650 1651 if (retval) { 1652 if (req.oldptr && req.oldidx > req.validlen) 1653 *retval = req.validlen; 1654 else 1655 *retval = req.oldidx; 1656 } 1657 return (error); 1658 } 1659 1660 /* 1661 * Drain into a sysctl struct. The user buffer should be wired if a page 1662 * fault would cause issue. 1663 */ 1664 static int 1665 sbuf_sysctl_drain(void *arg, const char *data, int len) 1666 { 1667 struct sysctl_req *req = arg; 1668 int error; 1669 1670 error = SYSCTL_OUT(req, data, len); 1671 KASSERT(error >= 0, ("Got unexpected negative value %d", error)); 1672 return (error == 0 ? len : -error); 1673 } 1674 1675 struct sbuf * 1676 sbuf_new_for_sysctl(struct sbuf *s, char *buf, int length, 1677 struct sysctl_req *req) 1678 { 1679 1680 s = sbuf_new(s, buf, length, SBUF_FIXEDLEN); 1681 sbuf_set_drain(s, sbuf_sysctl_drain, req); 1682 return (s); 1683 }
Cache object: 71414afbbc21066797b8e4f8ba59d849
[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]