1 /*- 2 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD 3 * 4 * Copyright (c) 1999 Poul-Henning Kamp. 5 * Copyright (c) 2008 Bjoern A. Zeeb. 6 * Copyright (c) 2009 James Gritton. 7 * All rights reserved. 8 * 9 * Redistribution and use in source and binary forms, with or without 10 * modification, are permitted provided that the following conditions 11 * are met: 12 * 1. Redistributions of source code must retain the above copyright 13 * notice, this list of conditions and the following disclaimer. 14 * 2. Redistributions in binary form must reproduce the above copyright 15 * notice, this list of conditions and the following disclaimer in the 16 * documentation and/or other materials provided with the distribution. 17 * 18 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 19 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 20 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 21 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 22 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 23 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 24 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 25 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 26 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 27 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 28 * SUCH DAMAGE. 29 */ 30 31 #include <sys/cdefs.h> 32 __FBSDID("$FreeBSD$"); 33 34 #include "opt_ddb.h" 35 #include "opt_inet.h" 36 #include "opt_inet6.h" 37 38 #include <sys/param.h> 39 #include <sys/types.h> 40 #include <sys/kernel.h> 41 #include <sys/systm.h> 42 #include <sys/errno.h> 43 #include <sys/sysproto.h> 44 #include <sys/malloc.h> 45 #include <sys/osd.h> 46 #include <sys/priv.h> 47 #include <sys/proc.h> 48 #include <sys/taskqueue.h> 49 #include <sys/fcntl.h> 50 #include <sys/jail.h> 51 #include <sys/linker.h> 52 #include <sys/lock.h> 53 #include <sys/mutex.h> 54 #include <sys/racct.h> 55 #include <sys/rctl.h> 56 #include <sys/refcount.h> 57 #include <sys/sx.h> 58 #include <sys/sysent.h> 59 #include <sys/namei.h> 60 #include <sys/mount.h> 61 #include <sys/queue.h> 62 #include <sys/socket.h> 63 #include <sys/syscallsubr.h> 64 #include <sys/sysctl.h> 65 #include <sys/uuid.h> 66 #include <sys/vnode.h> 67 68 #include <net/if.h> 69 #include <net/vnet.h> 70 71 #include <netinet/in.h> 72 73 #ifdef DDB 74 #include <ddb/ddb.h> 75 #endif /* DDB */ 76 77 #include <security/mac/mac_framework.h> 78 79 #define DEFAULT_HOSTUUID "00000000-0000-0000-0000-000000000000" 80 #define PRISON0_HOSTUUID_MODULE "hostuuid" 81 82 MALLOC_DEFINE(M_PRISON, "prison", "Prison structures"); 83 static MALLOC_DEFINE(M_PRISON_RACCT, "prison_racct", "Prison racct structures"); 84 85 /* Keep struct prison prison0 and some code in kern_jail_set() readable. */ 86 #ifdef INET 87 #ifdef INET6 88 #define _PR_IP_SADDRSEL PR_IP4_SADDRSEL|PR_IP6_SADDRSEL 89 #else 90 #define _PR_IP_SADDRSEL PR_IP4_SADDRSEL 91 #endif 92 #else /* !INET */ 93 #ifdef INET6 94 #define _PR_IP_SADDRSEL PR_IP6_SADDRSEL 95 #else 96 #define _PR_IP_SADDRSEL 0 97 #endif 98 #endif 99 100 /* prison0 describes what is "real" about the system. */ 101 struct prison prison0 = { 102 .pr_id = 0, 103 .pr_name = "0", 104 .pr_ref = 1, 105 .pr_uref = 1, 106 .pr_path = "/", 107 .pr_securelevel = -1, 108 .pr_devfs_rsnum = 0, 109 .pr_childmax = JAIL_MAX, 110 .pr_hostuuid = DEFAULT_HOSTUUID, 111 .pr_children = LIST_HEAD_INITIALIZER(prison0.pr_children), 112 #ifdef VIMAGE 113 .pr_flags = PR_HOST|PR_VNET|_PR_IP_SADDRSEL, 114 #else 115 .pr_flags = PR_HOST|_PR_IP_SADDRSEL, 116 #endif 117 .pr_allow = PR_ALLOW_ALL_STATIC, 118 }; 119 MTX_SYSINIT(prison0, &prison0.pr_mtx, "jail mutex", MTX_DEF); 120 121 struct bool_flags { 122 const char *name; 123 const char *noname; 124 volatile u_int flag; 125 }; 126 struct jailsys_flags { 127 const char *name; 128 unsigned disable; 129 unsigned new; 130 }; 131 132 /* allprison, allprison_racct and lastprid are protected by allprison_lock. */ 133 struct sx allprison_lock; 134 SX_SYSINIT(allprison_lock, &allprison_lock, "allprison"); 135 struct prisonlist allprison = TAILQ_HEAD_INITIALIZER(allprison); 136 LIST_HEAD(, prison_racct) allprison_racct; 137 int lastprid = 0; 138 139 static int get_next_prid(struct prison **insprp); 140 static int do_jail_attach(struct thread *td, struct prison *pr); 141 static void prison_complete(void *context, int pending); 142 static void prison_deref(struct prison *pr, int flags); 143 static void prison_set_allow_locked(struct prison *pr, unsigned flag, 144 int enable); 145 static char *prison_path(struct prison *pr1, struct prison *pr2); 146 static void prison_remove_one(struct prison *pr); 147 #ifdef RACCT 148 static void prison_racct_attach(struct prison *pr); 149 static void prison_racct_modify(struct prison *pr); 150 static void prison_racct_detach(struct prison *pr); 151 #endif 152 153 /* Flags for prison_deref */ 154 #define PD_DEREF 0x01 /* Decrement pr_ref */ 155 #define PD_DEUREF 0x02 /* Decrement pr_uref */ 156 #define PD_LOCKED 0x04 /* pr_mtx is held */ 157 #define PD_LIST_SLOCKED 0x08 /* allprison_lock is held shared */ 158 #define PD_LIST_XLOCKED 0x10 /* allprison_lock is held exclusive */ 159 160 /* 161 * Parameter names corresponding to PR_* flag values. Size values are for kvm 162 * as we cannot figure out the size of a sparse array, or an array without a 163 * terminating entry. 164 */ 165 static struct bool_flags pr_flag_bool[] = { 166 {"persist", "nopersist", PR_PERSIST}, 167 #ifdef INET 168 {"ip4.saddrsel", "ip4.nosaddrsel", PR_IP4_SADDRSEL}, 169 #endif 170 #ifdef INET6 171 {"ip6.saddrsel", "ip6.nosaddrsel", PR_IP6_SADDRSEL}, 172 #endif 173 }; 174 const size_t pr_flag_bool_size = sizeof(pr_flag_bool); 175 176 static struct jailsys_flags pr_flag_jailsys[] = { 177 {"host", 0, PR_HOST}, 178 #ifdef VIMAGE 179 {"vnet", 0, PR_VNET}, 180 #endif 181 #ifdef INET 182 {"ip4", PR_IP4_USER, PR_IP4_USER}, 183 #endif 184 #ifdef INET6 185 {"ip6", PR_IP6_USER, PR_IP6_USER}, 186 #endif 187 }; 188 const size_t pr_flag_jailsys_size = sizeof(pr_flag_jailsys); 189 190 /* 191 * Make this array full-size so dynamic parameters can be added. 192 * It is protected by prison0.mtx, but lockless reading is allowed 193 * with an atomic check of the flag values. 194 */ 195 static struct bool_flags pr_flag_allow[NBBY * NBPW] = { 196 {"allow.set_hostname", "allow.noset_hostname", PR_ALLOW_SET_HOSTNAME}, 197 {"allow.sysvipc", "allow.nosysvipc", PR_ALLOW_SYSVIPC}, 198 {"allow.raw_sockets", "allow.noraw_sockets", PR_ALLOW_RAW_SOCKETS}, 199 {"allow.chflags", "allow.nochflags", PR_ALLOW_CHFLAGS}, 200 {"allow.mount", "allow.nomount", PR_ALLOW_MOUNT}, 201 {"allow.quotas", "allow.noquotas", PR_ALLOW_QUOTAS}, 202 {"allow.socket_af", "allow.nosocket_af", PR_ALLOW_SOCKET_AF}, 203 {"allow.mlock", "allow.nomlock", PR_ALLOW_MLOCK}, 204 {"allow.reserved_ports", "allow.noreserved_ports", 205 PR_ALLOW_RESERVED_PORTS}, 206 {"allow.read_msgbuf", "allow.noread_msgbuf", PR_ALLOW_READ_MSGBUF}, 207 {"allow.unprivileged_proc_debug", "allow.nounprivileged_proc_debug", 208 PR_ALLOW_UNPRIV_DEBUG}, 209 {"allow.suser", "allow.nosuser", PR_ALLOW_SUSER}, 210 }; 211 static unsigned pr_allow_all = PR_ALLOW_ALL_STATIC; 212 const size_t pr_flag_allow_size = sizeof(pr_flag_allow); 213 214 #define JAIL_DEFAULT_ALLOW (PR_ALLOW_SET_HOSTNAME | \ 215 PR_ALLOW_RESERVED_PORTS | \ 216 PR_ALLOW_UNPRIV_DEBUG | \ 217 PR_ALLOW_SUSER) 218 #define JAIL_DEFAULT_ENFORCE_STATFS 2 219 #define JAIL_DEFAULT_DEVFS_RSNUM 0 220 static unsigned jail_default_allow = JAIL_DEFAULT_ALLOW; 221 static int jail_default_enforce_statfs = JAIL_DEFAULT_ENFORCE_STATFS; 222 static int jail_default_devfs_rsnum = JAIL_DEFAULT_DEVFS_RSNUM; 223 #if defined(INET) || defined(INET6) 224 static unsigned jail_max_af_ips = 255; 225 #endif 226 227 /* 228 * Initialize the parts of prison0 that can't be static-initialized with 229 * constants. This is called from proc0_init() after creating thread0 cpuset. 230 */ 231 void 232 prison0_init(void) 233 { 234 uint8_t *file, *data; 235 size_t size; 236 237 prison0.pr_cpuset = cpuset_ref(thread0.td_cpuset); 238 prison0.pr_osreldate = osreldate; 239 strlcpy(prison0.pr_osrelease, osrelease, sizeof(prison0.pr_osrelease)); 240 241 /* If we have a preloaded hostuuid, use it. */ 242 file = preload_search_by_type(PRISON0_HOSTUUID_MODULE); 243 if (file != NULL) { 244 data = preload_fetch_addr(file); 245 size = preload_fetch_size(file); 246 if (data != NULL) { 247 /* 248 * The preloaded data may include trailing whitespace, almost 249 * certainly a newline; skip over any whitespace or 250 * non-printable characters to be safe. 251 */ 252 while (size > 0 && data[size - 1] <= 0x20) { 253 data[size--] = '\0'; 254 } 255 if (validate_uuid(data, size, NULL, 0) == 0) { 256 (void)strlcpy(prison0.pr_hostuuid, data, 257 size + 1); 258 } else if (bootverbose) { 259 printf("hostuuid: preload data malformed: '%s'", 260 data); 261 } 262 } 263 } 264 if (bootverbose) 265 printf("hostuuid: using %s\n", prison0.pr_hostuuid); 266 } 267 268 /* 269 * struct jail_args { 270 * struct jail *jail; 271 * }; 272 */ 273 int 274 sys_jail(struct thread *td, struct jail_args *uap) 275 { 276 uint32_t version; 277 int error; 278 struct jail j; 279 280 error = copyin(uap->jail, &version, sizeof(uint32_t)); 281 if (error) 282 return (error); 283 284 switch (version) { 285 case 0: 286 { 287 struct jail_v0 j0; 288 289 /* FreeBSD single IPv4 jails. */ 290 bzero(&j, sizeof(struct jail)); 291 error = copyin(uap->jail, &j0, sizeof(struct jail_v0)); 292 if (error) 293 return (error); 294 j.version = j0.version; 295 j.path = j0.path; 296 j.hostname = j0.hostname; 297 j.ip4s = htonl(j0.ip_number); /* jail_v0 is host order */ 298 break; 299 } 300 301 case 1: 302 /* 303 * Version 1 was used by multi-IPv4 jail implementations 304 * that never made it into the official kernel. 305 */ 306 return (EINVAL); 307 308 case 2: /* JAIL_API_VERSION */ 309 /* FreeBSD multi-IPv4/IPv6,noIP jails. */ 310 error = copyin(uap->jail, &j, sizeof(struct jail)); 311 if (error) 312 return (error); 313 break; 314 315 default: 316 /* Sci-Fi jails are not supported, sorry. */ 317 return (EINVAL); 318 } 319 return (kern_jail(td, &j)); 320 } 321 322 int 323 kern_jail(struct thread *td, struct jail *j) 324 { 325 struct iovec optiov[2 * (4 + nitems(pr_flag_allow) 326 #ifdef INET 327 + 1 328 #endif 329 #ifdef INET6 330 + 1 331 #endif 332 )]; 333 struct uio opt; 334 char *u_path, *u_hostname, *u_name; 335 struct bool_flags *bf; 336 #ifdef INET 337 uint32_t ip4s; 338 struct in_addr *u_ip4; 339 #endif 340 #ifdef INET6 341 struct in6_addr *u_ip6; 342 #endif 343 size_t tmplen; 344 int error, enforce_statfs; 345 346 bzero(&optiov, sizeof(optiov)); 347 opt.uio_iov = optiov; 348 opt.uio_iovcnt = 0; 349 opt.uio_offset = -1; 350 opt.uio_resid = -1; 351 opt.uio_segflg = UIO_SYSSPACE; 352 opt.uio_rw = UIO_READ; 353 opt.uio_td = td; 354 355 /* Set permissions for top-level jails from sysctls. */ 356 if (!jailed(td->td_ucred)) { 357 for (bf = pr_flag_allow; 358 bf < pr_flag_allow + nitems(pr_flag_allow) && 359 atomic_load_int(&bf->flag) != 0; 360 bf++) { 361 optiov[opt.uio_iovcnt].iov_base = __DECONST(char *, 362 (jail_default_allow & bf->flag) 363 ? bf->name : bf->noname); 364 optiov[opt.uio_iovcnt].iov_len = 365 strlen(optiov[opt.uio_iovcnt].iov_base) + 1; 366 opt.uio_iovcnt += 2; 367 } 368 optiov[opt.uio_iovcnt].iov_base = "enforce_statfs"; 369 optiov[opt.uio_iovcnt].iov_len = sizeof("enforce_statfs"); 370 opt.uio_iovcnt++; 371 enforce_statfs = jail_default_enforce_statfs; 372 optiov[opt.uio_iovcnt].iov_base = &enforce_statfs; 373 optiov[opt.uio_iovcnt].iov_len = sizeof(enforce_statfs); 374 opt.uio_iovcnt++; 375 } 376 377 tmplen = MAXPATHLEN + MAXHOSTNAMELEN + MAXHOSTNAMELEN; 378 #ifdef INET 379 ip4s = (j->version == 0) ? 1 : j->ip4s; 380 if (ip4s > jail_max_af_ips) 381 return (EINVAL); 382 tmplen += ip4s * sizeof(struct in_addr); 383 #else 384 if (j->ip4s > 0) 385 return (EINVAL); 386 #endif 387 #ifdef INET6 388 if (j->ip6s > jail_max_af_ips) 389 return (EINVAL); 390 tmplen += j->ip6s * sizeof(struct in6_addr); 391 #else 392 if (j->ip6s > 0) 393 return (EINVAL); 394 #endif 395 u_path = malloc(tmplen, M_TEMP, M_WAITOK); 396 u_hostname = u_path + MAXPATHLEN; 397 u_name = u_hostname + MAXHOSTNAMELEN; 398 #ifdef INET 399 u_ip4 = (struct in_addr *)(u_name + MAXHOSTNAMELEN); 400 #endif 401 #ifdef INET6 402 #ifdef INET 403 u_ip6 = (struct in6_addr *)(u_ip4 + ip4s); 404 #else 405 u_ip6 = (struct in6_addr *)(u_name + MAXHOSTNAMELEN); 406 #endif 407 #endif 408 optiov[opt.uio_iovcnt].iov_base = "path"; 409 optiov[opt.uio_iovcnt].iov_len = sizeof("path"); 410 opt.uio_iovcnt++; 411 optiov[opt.uio_iovcnt].iov_base = u_path; 412 error = copyinstr(j->path, u_path, MAXPATHLEN, 413 &optiov[opt.uio_iovcnt].iov_len); 414 if (error) { 415 free(u_path, M_TEMP); 416 return (error); 417 } 418 opt.uio_iovcnt++; 419 optiov[opt.uio_iovcnt].iov_base = "host.hostname"; 420 optiov[opt.uio_iovcnt].iov_len = sizeof("host.hostname"); 421 opt.uio_iovcnt++; 422 optiov[opt.uio_iovcnt].iov_base = u_hostname; 423 error = copyinstr(j->hostname, u_hostname, MAXHOSTNAMELEN, 424 &optiov[opt.uio_iovcnt].iov_len); 425 if (error) { 426 free(u_path, M_TEMP); 427 return (error); 428 } 429 opt.uio_iovcnt++; 430 if (j->jailname != NULL) { 431 optiov[opt.uio_iovcnt].iov_base = "name"; 432 optiov[opt.uio_iovcnt].iov_len = sizeof("name"); 433 opt.uio_iovcnt++; 434 optiov[opt.uio_iovcnt].iov_base = u_name; 435 error = copyinstr(j->jailname, u_name, MAXHOSTNAMELEN, 436 &optiov[opt.uio_iovcnt].iov_len); 437 if (error) { 438 free(u_path, M_TEMP); 439 return (error); 440 } 441 opt.uio_iovcnt++; 442 } 443 #ifdef INET 444 optiov[opt.uio_iovcnt].iov_base = "ip4.addr"; 445 optiov[opt.uio_iovcnt].iov_len = sizeof("ip4.addr"); 446 opt.uio_iovcnt++; 447 optiov[opt.uio_iovcnt].iov_base = u_ip4; 448 optiov[opt.uio_iovcnt].iov_len = ip4s * sizeof(struct in_addr); 449 if (j->version == 0) 450 u_ip4->s_addr = j->ip4s; 451 else { 452 error = copyin(j->ip4, u_ip4, optiov[opt.uio_iovcnt].iov_len); 453 if (error) { 454 free(u_path, M_TEMP); 455 return (error); 456 } 457 } 458 opt.uio_iovcnt++; 459 #endif 460 #ifdef INET6 461 optiov[opt.uio_iovcnt].iov_base = "ip6.addr"; 462 optiov[opt.uio_iovcnt].iov_len = sizeof("ip6.addr"); 463 opt.uio_iovcnt++; 464 optiov[opt.uio_iovcnt].iov_base = u_ip6; 465 optiov[opt.uio_iovcnt].iov_len = j->ip6s * sizeof(struct in6_addr); 466 error = copyin(j->ip6, u_ip6, optiov[opt.uio_iovcnt].iov_len); 467 if (error) { 468 free(u_path, M_TEMP); 469 return (error); 470 } 471 opt.uio_iovcnt++; 472 #endif 473 KASSERT(opt.uio_iovcnt <= nitems(optiov), 474 ("kern_jail: too many iovecs (%d)", opt.uio_iovcnt)); 475 error = kern_jail_set(td, &opt, JAIL_CREATE | JAIL_ATTACH); 476 free(u_path, M_TEMP); 477 return (error); 478 } 479 480 /* 481 * struct jail_set_args { 482 * struct iovec *iovp; 483 * unsigned int iovcnt; 484 * int flags; 485 * }; 486 */ 487 int 488 sys_jail_set(struct thread *td, struct jail_set_args *uap) 489 { 490 struct uio *auio; 491 int error; 492 493 /* Check that we have an even number of iovecs. */ 494 if (uap->iovcnt & 1) 495 return (EINVAL); 496 497 error = copyinuio(uap->iovp, uap->iovcnt, &auio); 498 if (error) 499 return (error); 500 error = kern_jail_set(td, auio, uap->flags); 501 free(auio, M_IOV); 502 return (error); 503 } 504 505 int 506 kern_jail_set(struct thread *td, struct uio *optuio, int flags) 507 { 508 struct nameidata nd; 509 #ifdef INET 510 struct in_addr *ip4; 511 #endif 512 #ifdef INET6 513 struct in6_addr *ip6; 514 #endif 515 struct vfsopt *opt; 516 struct vfsoptlist *opts; 517 struct prison *pr, *deadpr, *inspr, *mypr, *ppr, *tpr; 518 struct vnode *root; 519 char *domain, *errmsg, *host, *name, *namelc, *p, *path, *uuid; 520 char *g_path, *osrelstr; 521 struct bool_flags *bf; 522 struct jailsys_flags *jsf; 523 #if defined(INET) || defined(INET6) 524 struct prison *tppr; 525 void *op; 526 #endif 527 unsigned long hid; 528 size_t namelen, onamelen, pnamelen; 529 int born, created, cuflags, descend, drflags, enforce; 530 int error, errmsg_len, errmsg_pos; 531 int gotchildmax, gotenforce, gothid, gotrsnum, gotslevel; 532 int jid, jsys, len, level; 533 int childmax, osreldt, rsnum, slevel; 534 #if defined(INET) || defined(INET6) 535 int ii, ij; 536 #endif 537 #ifdef INET 538 int ip4s, redo_ip4; 539 #endif 540 #ifdef INET6 541 int ip6s, redo_ip6; 542 #endif 543 uint64_t pr_allow, ch_allow, pr_flags, ch_flags; 544 uint64_t pr_allow_diff; 545 unsigned tallow; 546 char numbuf[12]; 547 548 error = priv_check(td, PRIV_JAIL_SET); 549 if (!error && (flags & JAIL_ATTACH)) 550 error = priv_check(td, PRIV_JAIL_ATTACH); 551 if (error) 552 return (error); 553 mypr = td->td_ucred->cr_prison; 554 if ((flags & JAIL_CREATE) && mypr->pr_childmax == 0) 555 return (EPERM); 556 if (flags & ~JAIL_SET_MASK) 557 return (EINVAL); 558 559 /* 560 * Check all the parameters before committing to anything. Not all 561 * errors can be caught early, but we may as well try. Also, this 562 * takes care of some expensive stuff (path lookup) before getting 563 * the allprison lock. 564 * 565 * XXX Jails are not filesystems, and jail parameters are not mount 566 * options. But it makes more sense to re-use the vfsopt code 567 * than duplicate it under a different name. 568 */ 569 error = vfs_buildopts(optuio, &opts); 570 if (error) 571 return (error); 572 #ifdef INET 573 ip4 = NULL; 574 #endif 575 #ifdef INET6 576 ip6 = NULL; 577 #endif 578 g_path = NULL; 579 580 cuflags = flags & (JAIL_CREATE | JAIL_UPDATE); 581 if (!cuflags) { 582 error = EINVAL; 583 vfs_opterror(opts, "no valid operation (create or update)"); 584 goto done_errmsg; 585 } 586 587 error = vfs_copyopt(opts, "jid", &jid, sizeof(jid)); 588 if (error == ENOENT) 589 jid = 0; 590 else if (error != 0) 591 goto done_free; 592 593 error = vfs_copyopt(opts, "securelevel", &slevel, sizeof(slevel)); 594 if (error == ENOENT) 595 gotslevel = 0; 596 else if (error != 0) 597 goto done_free; 598 else 599 gotslevel = 1; 600 601 error = 602 vfs_copyopt(opts, "children.max", &childmax, sizeof(childmax)); 603 if (error == ENOENT) 604 gotchildmax = 0; 605 else if (error != 0) 606 goto done_free; 607 else 608 gotchildmax = 1; 609 610 error = vfs_copyopt(opts, "enforce_statfs", &enforce, sizeof(enforce)); 611 if (error == ENOENT) 612 gotenforce = 0; 613 else if (error != 0) 614 goto done_free; 615 else if (enforce < 0 || enforce > 2) { 616 error = EINVAL; 617 goto done_free; 618 } else 619 gotenforce = 1; 620 621 error = vfs_copyopt(opts, "devfs_ruleset", &rsnum, sizeof(rsnum)); 622 if (error == ENOENT) 623 gotrsnum = 0; 624 else if (error != 0) 625 goto done_free; 626 else 627 gotrsnum = 1; 628 629 pr_flags = ch_flags = 0; 630 for (bf = pr_flag_bool; 631 bf < pr_flag_bool + nitems(pr_flag_bool); 632 bf++) { 633 vfs_flagopt(opts, bf->name, &pr_flags, bf->flag); 634 vfs_flagopt(opts, bf->noname, &ch_flags, bf->flag); 635 } 636 ch_flags |= pr_flags; 637 for (jsf = pr_flag_jailsys; 638 jsf < pr_flag_jailsys + nitems(pr_flag_jailsys); 639 jsf++) { 640 error = vfs_copyopt(opts, jsf->name, &jsys, sizeof(jsys)); 641 if (error == ENOENT) 642 continue; 643 if (error != 0) 644 goto done_free; 645 switch (jsys) { 646 case JAIL_SYS_DISABLE: 647 if (!jsf->disable) { 648 error = EINVAL; 649 goto done_free; 650 } 651 pr_flags |= jsf->disable; 652 break; 653 case JAIL_SYS_NEW: 654 pr_flags |= jsf->new; 655 break; 656 case JAIL_SYS_INHERIT: 657 break; 658 default: 659 error = EINVAL; 660 goto done_free; 661 } 662 ch_flags |= jsf->new | jsf->disable; 663 } 664 if ((flags & (JAIL_CREATE | JAIL_UPDATE | JAIL_ATTACH)) == JAIL_CREATE 665 && !(pr_flags & PR_PERSIST)) { 666 error = EINVAL; 667 vfs_opterror(opts, "new jail must persist or attach"); 668 goto done_errmsg; 669 } 670 #ifdef VIMAGE 671 if ((flags & JAIL_UPDATE) && (ch_flags & PR_VNET)) { 672 error = EINVAL; 673 vfs_opterror(opts, "vnet cannot be changed after creation"); 674 goto done_errmsg; 675 } 676 #endif 677 #ifdef INET 678 if ((flags & JAIL_UPDATE) && (ch_flags & PR_IP4_USER)) { 679 error = EINVAL; 680 vfs_opterror(opts, "ip4 cannot be changed after creation"); 681 goto done_errmsg; 682 } 683 #endif 684 #ifdef INET6 685 if ((flags & JAIL_UPDATE) && (ch_flags & PR_IP6_USER)) { 686 error = EINVAL; 687 vfs_opterror(opts, "ip6 cannot be changed after creation"); 688 goto done_errmsg; 689 } 690 #endif 691 692 pr_allow = ch_allow = 0; 693 for (bf = pr_flag_allow; 694 bf < pr_flag_allow + nitems(pr_flag_allow) && 695 atomic_load_int(&bf->flag) != 0; 696 bf++) { 697 vfs_flagopt(opts, bf->name, &pr_allow, bf->flag); 698 vfs_flagopt(opts, bf->noname, &ch_allow, bf->flag); 699 } 700 ch_allow |= pr_allow; 701 702 error = vfs_getopt(opts, "name", (void **)&name, &len); 703 if (error == ENOENT) 704 name = NULL; 705 else if (error != 0) 706 goto done_free; 707 else { 708 if (len == 0 || name[len - 1] != '\0') { 709 error = EINVAL; 710 goto done_free; 711 } 712 if (len > MAXHOSTNAMELEN) { 713 error = ENAMETOOLONG; 714 goto done_free; 715 } 716 } 717 718 error = vfs_getopt(opts, "host.hostname", (void **)&host, &len); 719 if (error == ENOENT) 720 host = NULL; 721 else if (error != 0) 722 goto done_free; 723 else { 724 ch_flags |= PR_HOST; 725 pr_flags |= PR_HOST; 726 if (len == 0 || host[len - 1] != '\0') { 727 error = EINVAL; 728 goto done_free; 729 } 730 if (len > MAXHOSTNAMELEN) { 731 error = ENAMETOOLONG; 732 goto done_free; 733 } 734 } 735 736 error = vfs_getopt(opts, "host.domainname", (void **)&domain, &len); 737 if (error == ENOENT) 738 domain = NULL; 739 else if (error != 0) 740 goto done_free; 741 else { 742 ch_flags |= PR_HOST; 743 pr_flags |= PR_HOST; 744 if (len == 0 || domain[len - 1] != '\0') { 745 error = EINVAL; 746 goto done_free; 747 } 748 if (len > MAXHOSTNAMELEN) { 749 error = ENAMETOOLONG; 750 goto done_free; 751 } 752 } 753 754 error = vfs_getopt(opts, "host.hostuuid", (void **)&uuid, &len); 755 if (error == ENOENT) 756 uuid = NULL; 757 else if (error != 0) 758 goto done_free; 759 else { 760 ch_flags |= PR_HOST; 761 pr_flags |= PR_HOST; 762 if (len == 0 || uuid[len - 1] != '\0') { 763 error = EINVAL; 764 goto done_free; 765 } 766 if (len > HOSTUUIDLEN) { 767 error = ENAMETOOLONG; 768 goto done_free; 769 } 770 } 771 772 #ifdef COMPAT_FREEBSD32 773 if (SV_PROC_FLAG(td->td_proc, SV_ILP32)) { 774 uint32_t hid32; 775 776 error = vfs_copyopt(opts, "host.hostid", &hid32, sizeof(hid32)); 777 hid = hid32; 778 } else 779 #endif 780 error = vfs_copyopt(opts, "host.hostid", &hid, sizeof(hid)); 781 if (error == ENOENT) 782 gothid = 0; 783 else if (error != 0) 784 goto done_free; 785 else { 786 gothid = 1; 787 ch_flags |= PR_HOST; 788 pr_flags |= PR_HOST; 789 } 790 791 #ifdef INET 792 error = vfs_getopt(opts, "ip4.addr", &op, &ip4s); 793 if (error == ENOENT) 794 ip4s = 0; 795 else if (error != 0) 796 goto done_free; 797 else if (ip4s & (sizeof(*ip4) - 1)) { 798 error = EINVAL; 799 goto done_free; 800 } else { 801 ch_flags |= PR_IP4_USER; 802 pr_flags |= PR_IP4_USER; 803 if (ip4s > 0) { 804 ip4s /= sizeof(*ip4); 805 if (ip4s > jail_max_af_ips) { 806 error = EINVAL; 807 vfs_opterror(opts, "too many IPv4 addresses"); 808 goto done_errmsg; 809 } 810 ip4 = malloc(ip4s * sizeof(*ip4), M_PRISON, M_WAITOK); 811 bcopy(op, ip4, ip4s * sizeof(*ip4)); 812 /* 813 * IP addresses are all sorted but ip[0] to preserve 814 * the primary IP address as given from userland. 815 * This special IP is used for unbound outgoing 816 * connections as well for "loopback" traffic in case 817 * source address selection cannot find any more fitting 818 * address to connect from. 819 */ 820 if (ip4s > 1) 821 qsort(ip4 + 1, ip4s - 1, sizeof(*ip4), 822 prison_qcmp_v4); 823 /* 824 * Check for duplicate addresses and do some simple 825 * zero and broadcast checks. If users give other bogus 826 * addresses it is their problem. 827 * 828 * We do not have to care about byte order for these 829 * checks so we will do them in NBO. 830 */ 831 for (ii = 0; ii < ip4s; ii++) { 832 if (ip4[ii].s_addr == INADDR_ANY || 833 ip4[ii].s_addr == INADDR_BROADCAST) { 834 error = EINVAL; 835 goto done_free; 836 } 837 if ((ii+1) < ip4s && 838 (ip4[0].s_addr == ip4[ii+1].s_addr || 839 ip4[ii].s_addr == ip4[ii+1].s_addr)) { 840 error = EINVAL; 841 goto done_free; 842 } 843 } 844 } 845 } 846 #endif 847 848 #ifdef INET6 849 error = vfs_getopt(opts, "ip6.addr", &op, &ip6s); 850 if (error == ENOENT) 851 ip6s = 0; 852 else if (error != 0) 853 goto done_free; 854 else if (ip6s & (sizeof(*ip6) - 1)) { 855 error = EINVAL; 856 goto done_free; 857 } else { 858 ch_flags |= PR_IP6_USER; 859 pr_flags |= PR_IP6_USER; 860 if (ip6s > 0) { 861 ip6s /= sizeof(*ip6); 862 if (ip6s > jail_max_af_ips) { 863 error = EINVAL; 864 vfs_opterror(opts, "too many IPv6 addresses"); 865 goto done_errmsg; 866 } 867 ip6 = malloc(ip6s * sizeof(*ip6), M_PRISON, M_WAITOK); 868 bcopy(op, ip6, ip6s * sizeof(*ip6)); 869 if (ip6s > 1) 870 qsort(ip6 + 1, ip6s - 1, sizeof(*ip6), 871 prison_qcmp_v6); 872 for (ii = 0; ii < ip6s; ii++) { 873 if (IN6_IS_ADDR_UNSPECIFIED(&ip6[ii])) { 874 error = EINVAL; 875 goto done_free; 876 } 877 if ((ii+1) < ip6s && 878 (IN6_ARE_ADDR_EQUAL(&ip6[0], &ip6[ii+1]) || 879 IN6_ARE_ADDR_EQUAL(&ip6[ii], &ip6[ii+1]))) 880 { 881 error = EINVAL; 882 goto done_free; 883 } 884 } 885 } 886 } 887 #endif 888 889 #if defined(VIMAGE) && (defined(INET) || defined(INET6)) 890 if ((ch_flags & PR_VNET) && (ch_flags & (PR_IP4_USER | PR_IP6_USER))) { 891 error = EINVAL; 892 vfs_opterror(opts, 893 "vnet jails cannot have IP address restrictions"); 894 goto done_errmsg; 895 } 896 #endif 897 898 error = vfs_getopt(opts, "osrelease", (void **)&osrelstr, &len); 899 if (error == ENOENT) 900 osrelstr = NULL; 901 else if (error != 0) 902 goto done_free; 903 else { 904 if (flags & JAIL_UPDATE) { 905 error = EINVAL; 906 vfs_opterror(opts, 907 "osrelease cannot be changed after creation"); 908 goto done_errmsg; 909 } 910 if (len == 0 || osrelstr[len - 1] != '\0') { 911 error = EINVAL; 912 goto done_free; 913 } 914 if (len >= OSRELEASELEN) { 915 error = ENAMETOOLONG; 916 vfs_opterror(opts, 917 "osrelease string must be 1-%d bytes long", 918 OSRELEASELEN - 1); 919 goto done_errmsg; 920 } 921 } 922 923 error = vfs_copyopt(opts, "osreldate", &osreldt, sizeof(osreldt)); 924 if (error == ENOENT) 925 osreldt = 0; 926 else if (error != 0) 927 goto done_free; 928 else { 929 if (flags & JAIL_UPDATE) { 930 error = EINVAL; 931 vfs_opterror(opts, 932 "osreldate cannot be changed after creation"); 933 goto done_errmsg; 934 } 935 if (osreldt == 0) { 936 error = EINVAL; 937 vfs_opterror(opts, "osreldate cannot be 0"); 938 goto done_errmsg; 939 } 940 } 941 942 root = NULL; 943 error = vfs_getopt(opts, "path", (void **)&path, &len); 944 if (error == ENOENT) 945 path = NULL; 946 else if (error != 0) 947 goto done_free; 948 else { 949 if (flags & JAIL_UPDATE) { 950 error = EINVAL; 951 vfs_opterror(opts, 952 "path cannot be changed after creation"); 953 goto done_errmsg; 954 } 955 if (len == 0 || path[len - 1] != '\0') { 956 error = EINVAL; 957 goto done_free; 958 } 959 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, 960 path, td); 961 error = namei(&nd); 962 if (error) 963 goto done_free; 964 root = nd.ni_vp; 965 NDFREE(&nd, NDF_ONLY_PNBUF); 966 g_path = malloc(MAXPATHLEN, M_TEMP, M_WAITOK); 967 strlcpy(g_path, path, MAXPATHLEN); 968 error = vn_path_to_global_path(td, root, g_path, MAXPATHLEN); 969 if (error == 0) { 970 path = g_path; 971 } else { 972 /* exit on other errors */ 973 goto done_free; 974 } 975 if (root->v_type != VDIR) { 976 error = ENOTDIR; 977 vput(root); 978 goto done_free; 979 } 980 VOP_UNLOCK(root); 981 } 982 983 /* 984 * Find the specified jail, or at least its parent. 985 * This abuses the file error codes ENOENT and EEXIST. 986 */ 987 pr = NULL; 988 ppr = mypr; 989 inspr = NULL; 990 if (cuflags == JAIL_CREATE && jid == 0 && name != NULL) { 991 namelc = strrchr(name, '.'); 992 jid = strtoul(namelc != NULL ? namelc + 1 : name, &p, 10); 993 if (*p != '\0') 994 jid = 0; 995 } 996 sx_xlock(&allprison_lock); 997 drflags = PD_LIST_XLOCKED; 998 if (jid != 0) { 999 if (jid < 0) { 1000 error = EINVAL; 1001 vfs_opterror(opts, "negative jid"); 1002 goto done_deref; 1003 } 1004 /* 1005 * See if a requested jid already exists. Keep track of 1006 * where it can be inserted later. 1007 */ 1008 TAILQ_FOREACH(inspr, &allprison, pr_list) { 1009 if (inspr->pr_id == jid) { 1010 mtx_lock(&inspr->pr_mtx); 1011 if (prison_isvalid(inspr)) { 1012 pr = inspr; 1013 drflags |= PD_LOCKED; 1014 inspr = NULL; 1015 } else 1016 mtx_unlock(&inspr->pr_mtx); 1017 break; 1018 } 1019 if (inspr->pr_id > jid) 1020 break; 1021 } 1022 if (pr != NULL) { 1023 ppr = pr->pr_parent; 1024 /* Create: jid must not exist. */ 1025 if (cuflags == JAIL_CREATE) { 1026 /* 1027 * Even creators that cannot see the jail will 1028 * get EEXIST. 1029 */ 1030 error = EEXIST; 1031 vfs_opterror(opts, "jail %d already exists", 1032 jid); 1033 goto done_deref; 1034 } 1035 if (!prison_ischild(mypr, pr)) { 1036 /* 1037 * Updaters get ENOENT if they cannot see the 1038 * jail. This is true even for CREATE | UPDATE, 1039 * which normally cannot give this error. 1040 */ 1041 error = ENOENT; 1042 vfs_opterror(opts, "jail %d not found", jid); 1043 goto done_deref; 1044 } else if (!prison_isalive(pr)) { 1045 if (!(flags & JAIL_DYING)) { 1046 error = ENOENT; 1047 vfs_opterror(opts, "jail %d is dying", 1048 jid); 1049 goto done_deref; 1050 } else if ((flags & JAIL_ATTACH) || 1051 (pr_flags & PR_PERSIST)) { 1052 /* 1053 * A dying jail might be resurrected 1054 * (via attach or persist), but first 1055 * it must determine if another jail 1056 * has claimed its name. Accomplish 1057 * this by implicitly re-setting the 1058 * name. 1059 */ 1060 if (name == NULL) 1061 name = prison_name(mypr, pr); 1062 } 1063 } 1064 } else { 1065 /* Update: jid must exist. */ 1066 if (cuflags == JAIL_UPDATE) { 1067 error = ENOENT; 1068 vfs_opterror(opts, "jail %d not found", jid); 1069 goto done_deref; 1070 } 1071 } 1072 } 1073 /* 1074 * If the caller provided a name, look for a jail by that name. 1075 * This has different semantics for creates and updates keyed by jid 1076 * (where the name must not already exist in a different jail), 1077 * and updates keyed by the name itself (where the name must exist 1078 * because that is the jail being updated). 1079 */ 1080 namelc = NULL; 1081 if (name != NULL) { 1082 namelc = strrchr(name, '.'); 1083 if (namelc == NULL) 1084 namelc = name; 1085 else { 1086 /* 1087 * This is a hierarchical name. Split it into the 1088 * parent and child names, and make sure the parent 1089 * exists or matches an already found jail. 1090 */ 1091 if (pr != NULL) { 1092 if (strncmp(name, ppr->pr_name, namelc - name) 1093 || ppr->pr_name[namelc - name] != '\0') { 1094 error = EINVAL; 1095 vfs_opterror(opts, 1096 "cannot change jail's parent"); 1097 goto done_deref; 1098 } 1099 } else { 1100 *namelc = '\0'; 1101 ppr = prison_find_name(mypr, name); 1102 if (ppr == NULL) { 1103 error = ENOENT; 1104 vfs_opterror(opts, 1105 "jail \"%s\" not found", name); 1106 goto done_deref; 1107 } 1108 if (!(flags & JAIL_DYING) && 1109 !prison_isalive(ppr)) { 1110 mtx_unlock(&ppr->pr_mtx); 1111 error = ENOENT; 1112 vfs_opterror(opts, 1113 "jail \"%s\" is dying", name); 1114 goto done_deref; 1115 } 1116 mtx_unlock(&ppr->pr_mtx); 1117 *namelc = '.'; 1118 } 1119 namelc++; 1120 } 1121 if (namelc[0] != '\0') { 1122 pnamelen = 1123 (ppr == &prison0) ? 0 : strlen(ppr->pr_name) + 1; 1124 name_again: 1125 deadpr = NULL; 1126 FOREACH_PRISON_CHILD(ppr, tpr) { 1127 if (tpr != pr && 1128 !strcmp(tpr->pr_name + pnamelen, namelc)) { 1129 mtx_lock(&tpr->pr_mtx); 1130 if (prison_isalive(tpr)) { 1131 if (pr == NULL && 1132 cuflags != JAIL_CREATE) { 1133 /* 1134 * Use this jail 1135 * for updates. 1136 */ 1137 pr = tpr; 1138 drflags |= PD_LOCKED; 1139 break; 1140 } 1141 /* 1142 * Create, or update(jid): 1143 * name must not exist in an 1144 * active sibling jail. 1145 */ 1146 error = EEXIST; 1147 mtx_unlock(&tpr->pr_mtx); 1148 vfs_opterror(opts, 1149 "jail \"%s\" already exists", 1150 name); 1151 goto done_deref; 1152 } 1153 if (pr == NULL && 1154 cuflags != JAIL_CREATE && 1155 prison_isvalid(tpr)) 1156 deadpr = tpr; 1157 mtx_unlock(&tpr->pr_mtx); 1158 } 1159 } 1160 /* If no active jail is found, use a dying one. */ 1161 if (deadpr != NULL && pr == NULL) { 1162 if (flags & JAIL_DYING) { 1163 mtx_lock(&deadpr->pr_mtx); 1164 if (!prison_isvalid(deadpr)) { 1165 mtx_unlock(&deadpr->pr_mtx); 1166 goto name_again; 1167 } 1168 pr = deadpr; 1169 drflags |= PD_LOCKED; 1170 } else if (cuflags == JAIL_UPDATE) { 1171 error = ENOENT; 1172 vfs_opterror(opts, 1173 "jail \"%s\" is dying", name); 1174 goto done_deref; 1175 } 1176 } 1177 /* Update: name must exist if no jid. */ 1178 else if (cuflags == JAIL_UPDATE && pr == NULL) { 1179 error = ENOENT; 1180 vfs_opterror(opts, "jail \"%s\" not found", 1181 name); 1182 goto done_deref; 1183 } 1184 } 1185 } 1186 /* Update: must provide a jid or name. */ 1187 else if (cuflags == JAIL_UPDATE && pr == NULL) { 1188 error = ENOENT; 1189 vfs_opterror(opts, "update specified no jail"); 1190 goto done_deref; 1191 } 1192 1193 /* If there's no prison to update, create a new one and link it in. */ 1194 created = pr == NULL; 1195 if (created) { 1196 for (tpr = mypr; tpr != NULL; tpr = tpr->pr_parent) 1197 if (tpr->pr_childcount >= tpr->pr_childmax) { 1198 error = EPERM; 1199 vfs_opterror(opts, "prison limit exceeded"); 1200 goto done_deref; 1201 } 1202 mtx_lock(&ppr->pr_mtx); 1203 if (!prison_isvalid(ppr)) { 1204 mtx_unlock(&ppr->pr_mtx); 1205 error = ENOENT; 1206 vfs_opterror(opts, "jail \"%s\" not found", 1207 prison_name(mypr, ppr)); 1208 goto done_deref; 1209 } 1210 prison_hold(ppr); 1211 if (refcount_acquire(&ppr->pr_uref)) 1212 mtx_unlock(&ppr->pr_mtx); 1213 else { 1214 /* This brings the parent back to life. */ 1215 mtx_unlock(&ppr->pr_mtx); 1216 error = osd_jail_call(ppr, PR_METHOD_CREATE, opts); 1217 if (error) { 1218 pr = ppr; 1219 drflags |= PD_DEREF | PD_DEUREF; 1220 goto done_deref; 1221 } 1222 } 1223 1224 if (jid == 0 && (jid = get_next_prid(&inspr)) == 0) { 1225 error = EAGAIN; 1226 vfs_opterror(opts, "no available jail IDs"); 1227 pr = ppr; 1228 drflags |= PD_DEREF | PD_DEUREF; 1229 goto done_deref; 1230 } 1231 1232 pr = malloc(sizeof(*pr), M_PRISON, M_WAITOK | M_ZERO); 1233 LIST_INIT(&pr->pr_children); 1234 mtx_init(&pr->pr_mtx, "jail mutex", NULL, MTX_DEF | MTX_DUPOK); 1235 TASK_INIT(&pr->pr_task, 0, prison_complete, pr); 1236 1237 pr->pr_id = jid; 1238 if (inspr != NULL) 1239 TAILQ_INSERT_BEFORE(inspr, pr, pr_list); 1240 else 1241 TAILQ_INSERT_TAIL(&allprison, pr, pr_list); 1242 1243 pr->pr_parent = ppr; 1244 LIST_INSERT_HEAD(&ppr->pr_children, pr, pr_sibling); 1245 for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent) 1246 tpr->pr_childcount++; 1247 1248 /* Set some default values, and inherit some from the parent. */ 1249 if (namelc == NULL) 1250 namelc = ""; 1251 if (path == NULL) { 1252 path = "/"; 1253 root = mypr->pr_root; 1254 vref(root); 1255 } 1256 strlcpy(pr->pr_hostuuid, DEFAULT_HOSTUUID, HOSTUUIDLEN); 1257 pr->pr_flags |= PR_HOST; 1258 #if defined(INET) || defined(INET6) 1259 #ifdef VIMAGE 1260 if (!(pr_flags & PR_VNET)) 1261 #endif 1262 { 1263 #ifdef INET 1264 if (!(ch_flags & PR_IP4_USER)) 1265 pr->pr_flags |= PR_IP4 | PR_IP4_USER; 1266 else if (!(pr_flags & PR_IP4_USER)) { 1267 pr->pr_flags |= ppr->pr_flags & PR_IP4; 1268 if (ppr->pr_ip4 != NULL) { 1269 pr->pr_ip4s = ppr->pr_ip4s; 1270 pr->pr_ip4 = malloc(pr->pr_ip4s * 1271 sizeof(struct in_addr), M_PRISON, 1272 M_WAITOK); 1273 bcopy(ppr->pr_ip4, pr->pr_ip4, 1274 pr->pr_ip4s * sizeof(*pr->pr_ip4)); 1275 } 1276 } 1277 #endif 1278 #ifdef INET6 1279 if (!(ch_flags & PR_IP6_USER)) 1280 pr->pr_flags |= PR_IP6 | PR_IP6_USER; 1281 else if (!(pr_flags & PR_IP6_USER)) { 1282 pr->pr_flags |= ppr->pr_flags & PR_IP6; 1283 if (ppr->pr_ip6 != NULL) { 1284 pr->pr_ip6s = ppr->pr_ip6s; 1285 pr->pr_ip6 = malloc(pr->pr_ip6s * 1286 sizeof(struct in6_addr), M_PRISON, 1287 M_WAITOK); 1288 bcopy(ppr->pr_ip6, pr->pr_ip6, 1289 pr->pr_ip6s * sizeof(*pr->pr_ip6)); 1290 } 1291 } 1292 #endif 1293 } 1294 #endif 1295 /* Source address selection is always on by default. */ 1296 pr->pr_flags |= _PR_IP_SADDRSEL; 1297 1298 pr->pr_securelevel = ppr->pr_securelevel; 1299 pr->pr_allow = JAIL_DEFAULT_ALLOW & ppr->pr_allow; 1300 pr->pr_enforce_statfs = jail_default_enforce_statfs; 1301 pr->pr_devfs_rsnum = ppr->pr_devfs_rsnum; 1302 1303 pr->pr_osreldate = osreldt ? osreldt : ppr->pr_osreldate; 1304 if (osrelstr == NULL) 1305 strlcpy(pr->pr_osrelease, ppr->pr_osrelease, 1306 sizeof(pr->pr_osrelease)); 1307 else 1308 strlcpy(pr->pr_osrelease, osrelstr, 1309 sizeof(pr->pr_osrelease)); 1310 1311 #ifdef VIMAGE 1312 /* Allocate a new vnet if specified. */ 1313 pr->pr_vnet = (pr_flags & PR_VNET) 1314 ? vnet_alloc() : ppr->pr_vnet; 1315 #endif 1316 /* 1317 * Allocate a dedicated cpuset for each jail. 1318 * Unlike other initial settings, this may return an erorr. 1319 */ 1320 error = cpuset_create_root(ppr, &pr->pr_cpuset); 1321 if (error) 1322 goto done_deref; 1323 1324 mtx_lock(&pr->pr_mtx); 1325 drflags |= PD_LOCKED; 1326 /* 1327 * New prisons do not yet have a reference, because we do not 1328 * want others to see the incomplete prison once the 1329 * allprison_lock is downgraded. 1330 */ 1331 } else { 1332 /* 1333 * Grab a reference for existing prisons, to ensure they 1334 * continue to exist for the duration of the call. 1335 */ 1336 prison_hold(pr); 1337 drflags |= PD_DEREF; 1338 #if defined(VIMAGE) && (defined(INET) || defined(INET6)) 1339 if ((pr->pr_flags & PR_VNET) && 1340 (ch_flags & (PR_IP4_USER | PR_IP6_USER))) { 1341 error = EINVAL; 1342 vfs_opterror(opts, 1343 "vnet jails cannot have IP address restrictions"); 1344 goto done_deref; 1345 } 1346 #endif 1347 #ifdef INET 1348 if (PR_IP4_USER & ch_flags & (pr_flags ^ pr->pr_flags)) { 1349 error = EINVAL; 1350 vfs_opterror(opts, 1351 "ip4 cannot be changed after creation"); 1352 goto done_deref; 1353 } 1354 #endif 1355 #ifdef INET6 1356 if (PR_IP6_USER & ch_flags & (pr_flags ^ pr->pr_flags)) { 1357 error = EINVAL; 1358 vfs_opterror(opts, 1359 "ip6 cannot be changed after creation"); 1360 goto done_deref; 1361 } 1362 #endif 1363 } 1364 1365 /* Do final error checking before setting anything. */ 1366 if (gotslevel) { 1367 if (slevel < ppr->pr_securelevel) { 1368 error = EPERM; 1369 goto done_deref; 1370 } 1371 } 1372 if (gotchildmax) { 1373 if (childmax >= ppr->pr_childmax) { 1374 error = EPERM; 1375 goto done_deref; 1376 } 1377 } 1378 if (gotenforce) { 1379 if (enforce < ppr->pr_enforce_statfs) { 1380 error = EPERM; 1381 goto done_deref; 1382 } 1383 } 1384 if (gotrsnum) { 1385 /* 1386 * devfs_rsnum is a uint16_t 1387 */ 1388 if (rsnum < 0 || rsnum > 65535) { 1389 error = EINVAL; 1390 goto done_deref; 1391 } 1392 /* 1393 * Nested jails always inherit parent's devfs ruleset 1394 */ 1395 if (jailed(td->td_ucred)) { 1396 if (rsnum > 0 && rsnum != ppr->pr_devfs_rsnum) { 1397 error = EPERM; 1398 goto done_deref; 1399 } else 1400 rsnum = ppr->pr_devfs_rsnum; 1401 } 1402 } 1403 #ifdef INET 1404 if (ip4s > 0) { 1405 if (ppr->pr_flags & PR_IP4) { 1406 /* 1407 * Make sure the new set of IP addresses is a 1408 * subset of the parent's list. Don't worry 1409 * about the parent being unlocked, as any 1410 * setting is done with allprison_lock held. 1411 */ 1412 for (ij = 0; ij < ppr->pr_ip4s; ij++) 1413 if (ip4[0].s_addr == ppr->pr_ip4[ij].s_addr) 1414 break; 1415 if (ij == ppr->pr_ip4s) { 1416 error = EPERM; 1417 goto done_deref; 1418 } 1419 if (ip4s > 1) { 1420 for (ii = ij = 1; ii < ip4s; ii++) { 1421 if (ip4[ii].s_addr == 1422 ppr->pr_ip4[0].s_addr) 1423 continue; 1424 for (; ij < ppr->pr_ip4s; ij++) 1425 if (ip4[ii].s_addr == 1426 ppr->pr_ip4[ij].s_addr) 1427 break; 1428 if (ij == ppr->pr_ip4s) 1429 break; 1430 } 1431 if (ij == ppr->pr_ip4s) { 1432 error = EPERM; 1433 goto done_deref; 1434 } 1435 } 1436 } 1437 /* 1438 * Check for conflicting IP addresses. We permit them 1439 * if there is no more than one IP on each jail. If 1440 * there is a duplicate on a jail with more than one 1441 * IP stop checking and return error. 1442 */ 1443 #ifdef VIMAGE 1444 for (tppr = ppr; tppr != &prison0; tppr = tppr->pr_parent) 1445 if (tppr->pr_flags & PR_VNET) 1446 break; 1447 #else 1448 tppr = &prison0; 1449 #endif 1450 FOREACH_PRISON_DESCENDANT(tppr, tpr, descend) { 1451 if (tpr == pr || 1452 #ifdef VIMAGE 1453 (tpr != tppr && (tpr->pr_flags & PR_VNET)) || 1454 #endif 1455 refcount_load(&tpr->pr_uref) == 0) { 1456 descend = 0; 1457 continue; 1458 } 1459 if (!(tpr->pr_flags & PR_IP4_USER)) 1460 continue; 1461 descend = 0; 1462 if (tpr->pr_ip4 == NULL || 1463 (ip4s == 1 && tpr->pr_ip4s == 1)) 1464 continue; 1465 for (ii = 0; ii < ip4s; ii++) { 1466 if (prison_check_ip4_locked(tpr, &ip4[ii]) == 1467 0) { 1468 error = EADDRINUSE; 1469 vfs_opterror(opts, 1470 "IPv4 addresses clash"); 1471 goto done_deref; 1472 } 1473 } 1474 } 1475 } 1476 #endif 1477 #ifdef INET6 1478 if (ip6s > 0) { 1479 if (ppr->pr_flags & PR_IP6) { 1480 /* 1481 * Make sure the new set of IP addresses is a 1482 * subset of the parent's list. 1483 */ 1484 for (ij = 0; ij < ppr->pr_ip6s; ij++) 1485 if (IN6_ARE_ADDR_EQUAL(&ip6[0], 1486 &ppr->pr_ip6[ij])) 1487 break; 1488 if (ij == ppr->pr_ip6s) { 1489 error = EPERM; 1490 goto done_deref; 1491 } 1492 if (ip6s > 1) { 1493 for (ii = ij = 1; ii < ip6s; ii++) { 1494 if (IN6_ARE_ADDR_EQUAL(&ip6[ii], 1495 &ppr->pr_ip6[0])) 1496 continue; 1497 for (; ij < ppr->pr_ip6s; ij++) 1498 if (IN6_ARE_ADDR_EQUAL( 1499 &ip6[ii], &ppr->pr_ip6[ij])) 1500 break; 1501 if (ij == ppr->pr_ip6s) 1502 break; 1503 } 1504 if (ij == ppr->pr_ip6s) { 1505 error = EPERM; 1506 goto done_deref; 1507 } 1508 } 1509 } 1510 /* Check for conflicting IP addresses. */ 1511 #ifdef VIMAGE 1512 for (tppr = ppr; tppr != &prison0; tppr = tppr->pr_parent) 1513 if (tppr->pr_flags & PR_VNET) 1514 break; 1515 #else 1516 tppr = &prison0; 1517 #endif 1518 FOREACH_PRISON_DESCENDANT(tppr, tpr, descend) { 1519 if (tpr == pr || 1520 #ifdef VIMAGE 1521 (tpr != tppr && (tpr->pr_flags & PR_VNET)) || 1522 #endif 1523 refcount_load(&tpr->pr_uref) == 0) { 1524 descend = 0; 1525 continue; 1526 } 1527 if (!(tpr->pr_flags & PR_IP6_USER)) 1528 continue; 1529 descend = 0; 1530 if (tpr->pr_ip6 == NULL || 1531 (ip6s == 1 && tpr->pr_ip6s == 1)) 1532 continue; 1533 for (ii = 0; ii < ip6s; ii++) { 1534 if (prison_check_ip6_locked(tpr, &ip6[ii]) == 1535 0) { 1536 error = EADDRINUSE; 1537 vfs_opterror(opts, 1538 "IPv6 addresses clash"); 1539 goto done_deref; 1540 } 1541 } 1542 } 1543 } 1544 #endif 1545 onamelen = namelen = 0; 1546 if (namelc != NULL) { 1547 /* Give a default name of the jid. Also allow the name to be 1548 * explicitly the jid - but not any other number, and only in 1549 * normal form (no leading zero/etc). 1550 */ 1551 if (namelc[0] == '\0') 1552 snprintf(namelc = numbuf, sizeof(numbuf), "%d", jid); 1553 else if ((strtoul(namelc, &p, 10) != jid || 1554 namelc[0] < '1' || namelc[0] > '9') && *p == '\0') { 1555 error = EINVAL; 1556 vfs_opterror(opts, 1557 "name cannot be numeric (unless it is the jid)"); 1558 goto done_deref; 1559 } 1560 /* 1561 * Make sure the name isn't too long for the prison or its 1562 * children. 1563 */ 1564 pnamelen = (ppr == &prison0) ? 0 : strlen(ppr->pr_name) + 1; 1565 onamelen = strlen(pr->pr_name + pnamelen); 1566 namelen = strlen(namelc); 1567 if (pnamelen + namelen + 1 > sizeof(pr->pr_name)) { 1568 error = ENAMETOOLONG; 1569 goto done_deref; 1570 } 1571 FOREACH_PRISON_DESCENDANT(pr, tpr, descend) { 1572 if (strlen(tpr->pr_name) + (namelen - onamelen) >= 1573 sizeof(pr->pr_name)) { 1574 error = ENAMETOOLONG; 1575 goto done_deref; 1576 } 1577 } 1578 } 1579 pr_allow_diff = pr_allow & ~ppr->pr_allow; 1580 if (pr_allow_diff & ~PR_ALLOW_DIFFERENCES) { 1581 error = EPERM; 1582 goto done_deref; 1583 } 1584 1585 /* 1586 * Let modules check their parameters. This requires unlocking and 1587 * then re-locking the prison, but this is still a valid state as long 1588 * as allprison_lock remains xlocked. 1589 */ 1590 mtx_unlock(&pr->pr_mtx); 1591 drflags &= ~PD_LOCKED; 1592 error = osd_jail_call(pr, PR_METHOD_CHECK, opts); 1593 if (error != 0) 1594 goto done_deref; 1595 mtx_lock(&pr->pr_mtx); 1596 drflags |= PD_LOCKED; 1597 1598 /* At this point, all valid parameters should have been noted. */ 1599 TAILQ_FOREACH(opt, opts, link) { 1600 if (!opt->seen && strcmp(opt->name, "errmsg")) { 1601 error = EINVAL; 1602 vfs_opterror(opts, "unknown parameter: %s", opt->name); 1603 goto done_deref; 1604 } 1605 } 1606 1607 /* Set the parameters of the prison. */ 1608 #ifdef INET 1609 redo_ip4 = 0; 1610 if (pr_flags & PR_IP4_USER) { 1611 pr->pr_flags |= PR_IP4; 1612 free(pr->pr_ip4, M_PRISON); 1613 pr->pr_ip4s = ip4s; 1614 pr->pr_ip4 = ip4; 1615 ip4 = NULL; 1616 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) { 1617 #ifdef VIMAGE 1618 if (tpr->pr_flags & PR_VNET) { 1619 descend = 0; 1620 continue; 1621 } 1622 #endif 1623 if (prison_restrict_ip4(tpr, NULL)) { 1624 redo_ip4 = 1; 1625 descend = 0; 1626 } 1627 } 1628 } 1629 #endif 1630 #ifdef INET6 1631 redo_ip6 = 0; 1632 if (pr_flags & PR_IP6_USER) { 1633 pr->pr_flags |= PR_IP6; 1634 free(pr->pr_ip6, M_PRISON); 1635 pr->pr_ip6s = ip6s; 1636 pr->pr_ip6 = ip6; 1637 ip6 = NULL; 1638 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) { 1639 #ifdef VIMAGE 1640 if (tpr->pr_flags & PR_VNET) { 1641 descend = 0; 1642 continue; 1643 } 1644 #endif 1645 if (prison_restrict_ip6(tpr, NULL)) { 1646 redo_ip6 = 1; 1647 descend = 0; 1648 } 1649 } 1650 } 1651 #endif 1652 if (gotslevel) { 1653 pr->pr_securelevel = slevel; 1654 /* Set all child jails to be at least this level. */ 1655 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) 1656 if (tpr->pr_securelevel < slevel) 1657 tpr->pr_securelevel = slevel; 1658 } 1659 if (gotchildmax) { 1660 pr->pr_childmax = childmax; 1661 /* Set all child jails to under this limit. */ 1662 FOREACH_PRISON_DESCENDANT_LOCKED_LEVEL(pr, tpr, descend, level) 1663 if (tpr->pr_childmax > childmax - level) 1664 tpr->pr_childmax = childmax > level 1665 ? childmax - level : 0; 1666 } 1667 if (gotenforce) { 1668 pr->pr_enforce_statfs = enforce; 1669 /* Pass this restriction on to the children. */ 1670 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) 1671 if (tpr->pr_enforce_statfs < enforce) 1672 tpr->pr_enforce_statfs = enforce; 1673 } 1674 if (gotrsnum) { 1675 pr->pr_devfs_rsnum = rsnum; 1676 /* Pass this restriction on to the children. */ 1677 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) 1678 tpr->pr_devfs_rsnum = rsnum; 1679 } 1680 if (namelc != NULL) { 1681 if (ppr == &prison0) 1682 strlcpy(pr->pr_name, namelc, sizeof(pr->pr_name)); 1683 else 1684 snprintf(pr->pr_name, sizeof(pr->pr_name), "%s.%s", 1685 ppr->pr_name, namelc); 1686 /* Change this component of child names. */ 1687 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) { 1688 bcopy(tpr->pr_name + onamelen, tpr->pr_name + namelen, 1689 strlen(tpr->pr_name + onamelen) + 1); 1690 bcopy(pr->pr_name, tpr->pr_name, namelen); 1691 } 1692 } 1693 if (path != NULL) { 1694 /* Try to keep a real-rooted full pathname. */ 1695 strlcpy(pr->pr_path, path, sizeof(pr->pr_path)); 1696 pr->pr_root = root; 1697 root = NULL; 1698 } 1699 if (PR_HOST & ch_flags & ~pr_flags) { 1700 if (pr->pr_flags & PR_HOST) { 1701 /* 1702 * Copy the parent's host info. As with pr_ip4 above, 1703 * the lack of a lock on the parent is not a problem; 1704 * it is always set with allprison_lock at least 1705 * shared, and is held exclusively here. 1706 */ 1707 strlcpy(pr->pr_hostname, pr->pr_parent->pr_hostname, 1708 sizeof(pr->pr_hostname)); 1709 strlcpy(pr->pr_domainname, pr->pr_parent->pr_domainname, 1710 sizeof(pr->pr_domainname)); 1711 strlcpy(pr->pr_hostuuid, pr->pr_parent->pr_hostuuid, 1712 sizeof(pr->pr_hostuuid)); 1713 pr->pr_hostid = pr->pr_parent->pr_hostid; 1714 } 1715 } else if (host != NULL || domain != NULL || uuid != NULL || gothid) { 1716 /* Set this prison, and any descendants without PR_HOST. */ 1717 if (host != NULL) 1718 strlcpy(pr->pr_hostname, host, sizeof(pr->pr_hostname)); 1719 if (domain != NULL) 1720 strlcpy(pr->pr_domainname, domain, 1721 sizeof(pr->pr_domainname)); 1722 if (uuid != NULL) 1723 strlcpy(pr->pr_hostuuid, uuid, sizeof(pr->pr_hostuuid)); 1724 if (gothid) 1725 pr->pr_hostid = hid; 1726 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) { 1727 if (tpr->pr_flags & PR_HOST) 1728 descend = 0; 1729 else { 1730 if (host != NULL) 1731 strlcpy(tpr->pr_hostname, 1732 pr->pr_hostname, 1733 sizeof(tpr->pr_hostname)); 1734 if (domain != NULL) 1735 strlcpy(tpr->pr_domainname, 1736 pr->pr_domainname, 1737 sizeof(tpr->pr_domainname)); 1738 if (uuid != NULL) 1739 strlcpy(tpr->pr_hostuuid, 1740 pr->pr_hostuuid, 1741 sizeof(tpr->pr_hostuuid)); 1742 if (gothid) 1743 tpr->pr_hostid = hid; 1744 } 1745 } 1746 } 1747 pr->pr_allow = (pr->pr_allow & ~ch_allow) | pr_allow; 1748 if ((tallow = ch_allow & ~pr_allow)) 1749 prison_set_allow_locked(pr, tallow, 0); 1750 /* 1751 * Persistent prisons get an extra reference, and prisons losing their 1752 * persist flag lose that reference. Only do this for existing prisons 1753 * for now, so new ones will remain unseen until after the module 1754 * handlers have completed. 1755 */ 1756 born = !prison_isalive(pr); 1757 if (!created && (ch_flags & PR_PERSIST & (pr_flags ^ pr->pr_flags))) { 1758 if (pr_flags & PR_PERSIST) { 1759 prison_hold(pr); 1760 refcount_acquire(&pr->pr_uref); 1761 } else { 1762 refcount_release(&pr->pr_ref); 1763 drflags |= PD_DEUREF; 1764 } 1765 } 1766 pr->pr_flags = (pr->pr_flags & ~ch_flags) | pr_flags; 1767 pr->pr_flags &= ~PR_REMOVE; 1768 mtx_unlock(&pr->pr_mtx); 1769 drflags &= ~PD_LOCKED; 1770 1771 #ifdef RACCT 1772 if (racct_enable && created) 1773 prison_racct_attach(pr); 1774 #endif 1775 1776 /* Locks may have prevented a complete restriction of child IP 1777 * addresses. If so, allocate some more memory and try again. 1778 */ 1779 #ifdef INET 1780 while (redo_ip4) { 1781 ip4s = pr->pr_ip4s; 1782 ip4 = malloc(ip4s * sizeof(*ip4), M_PRISON, M_WAITOK); 1783 mtx_lock(&pr->pr_mtx); 1784 redo_ip4 = 0; 1785 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) { 1786 #ifdef VIMAGE 1787 if (tpr->pr_flags & PR_VNET) { 1788 descend = 0; 1789 continue; 1790 } 1791 #endif 1792 if (prison_restrict_ip4(tpr, ip4)) { 1793 if (ip4 != NULL) 1794 ip4 = NULL; 1795 else 1796 redo_ip4 = 1; 1797 } 1798 } 1799 mtx_unlock(&pr->pr_mtx); 1800 } 1801 #endif 1802 #ifdef INET6 1803 while (redo_ip6) { 1804 ip6s = pr->pr_ip6s; 1805 ip6 = malloc(ip6s * sizeof(*ip6), M_PRISON, M_WAITOK); 1806 mtx_lock(&pr->pr_mtx); 1807 redo_ip6 = 0; 1808 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) { 1809 #ifdef VIMAGE 1810 if (tpr->pr_flags & PR_VNET) { 1811 descend = 0; 1812 continue; 1813 } 1814 #endif 1815 if (prison_restrict_ip6(tpr, ip6)) { 1816 if (ip6 != NULL) 1817 ip6 = NULL; 1818 else 1819 redo_ip6 = 1; 1820 } 1821 } 1822 mtx_unlock(&pr->pr_mtx); 1823 } 1824 #endif 1825 1826 /* Let the modules do their work. */ 1827 sx_downgrade(&allprison_lock); 1828 drflags = (drflags & ~PD_LIST_XLOCKED) | PD_LIST_SLOCKED; 1829 if (born) { 1830 error = osd_jail_call(pr, PR_METHOD_CREATE, opts); 1831 if (error) { 1832 (void)osd_jail_call(pr, PR_METHOD_REMOVE, NULL); 1833 goto done_deref; 1834 } 1835 } 1836 error = osd_jail_call(pr, PR_METHOD_SET, opts); 1837 if (error) { 1838 if (born) 1839 (void)osd_jail_call(pr, PR_METHOD_REMOVE, NULL); 1840 goto done_deref; 1841 } 1842 1843 /* Attach this process to the prison if requested. */ 1844 if (flags & JAIL_ATTACH) { 1845 mtx_lock(&pr->pr_mtx); 1846 error = do_jail_attach(td, pr); 1847 drflags &= ~PD_LIST_SLOCKED; 1848 if (error) { 1849 if (created) { 1850 /* do_jail_attach has removed the prison. */ 1851 pr = NULL; 1852 } 1853 vfs_opterror(opts, "attach failed"); 1854 goto done_deref; 1855 } 1856 } 1857 1858 #ifdef RACCT 1859 if (racct_enable && !created) { 1860 if (drflags & PD_LIST_SLOCKED) { 1861 sx_sunlock(&allprison_lock); 1862 drflags &= ~PD_LIST_SLOCKED; 1863 } 1864 prison_racct_modify(pr); 1865 } 1866 #endif 1867 1868 td->td_retval[0] = pr->pr_id; 1869 1870 if (created) { 1871 /* 1872 * Add a reference to newly created persistent prisons 1873 * (which was not done earlier so that the prison would 1874 * not be publicly visible). 1875 */ 1876 if (pr_flags & PR_PERSIST) { 1877 mtx_lock(&pr->pr_mtx); 1878 drflags |= PD_LOCKED; 1879 refcount_acquire(&pr->pr_ref); 1880 refcount_acquire(&pr->pr_uref); 1881 } else { 1882 /* Non-persistent jails need no further changes. */ 1883 pr = NULL; 1884 } 1885 } 1886 1887 done_deref: 1888 /* Release any temporary prison holds and/or locks. */ 1889 if (pr != NULL) 1890 prison_deref(pr, drflags); 1891 else if (drflags & PD_LIST_SLOCKED) 1892 sx_sunlock(&allprison_lock); 1893 else if (drflags & PD_LIST_XLOCKED) 1894 sx_xunlock(&allprison_lock); 1895 if (root != NULL) 1896 vrele(root); 1897 done_errmsg: 1898 if (error) { 1899 /* Write the error message back to userspace. */ 1900 if (vfs_getopt(opts, "errmsg", (void **)&errmsg, 1901 &errmsg_len) == 0 && errmsg_len > 0) { 1902 errmsg_pos = 2 * vfs_getopt_pos(opts, "errmsg") + 1; 1903 if (optuio->uio_segflg == UIO_SYSSPACE) 1904 bcopy(errmsg, 1905 optuio->uio_iov[errmsg_pos].iov_base, 1906 errmsg_len); 1907 else 1908 copyout(errmsg, 1909 optuio->uio_iov[errmsg_pos].iov_base, 1910 errmsg_len); 1911 } 1912 } 1913 done_free: 1914 #ifdef INET 1915 free(ip4, M_PRISON); 1916 #endif 1917 #ifdef INET6 1918 free(ip6, M_PRISON); 1919 #endif 1920 if (g_path != NULL) 1921 free(g_path, M_TEMP); 1922 vfs_freeopts(opts); 1923 return (error); 1924 } 1925 1926 /* 1927 * Find the next available prison ID. Return the ID on success, or zero 1928 * on failure. Also set a pointer to the allprison list entry the prison 1929 * should be inserted before. 1930 */ 1931 static int 1932 get_next_prid(struct prison **insprp) 1933 { 1934 struct prison *inspr; 1935 int jid, maxid; 1936 1937 jid = lastprid % JAIL_MAX + 1; 1938 if (TAILQ_EMPTY(&allprison) || 1939 TAILQ_LAST(&allprison, prisonlist)->pr_id < jid) { 1940 /* 1941 * A common case is for all jails to be implicitly numbered, 1942 * which means they'll go on the end of the list, at least 1943 * for the first JAIL_MAX times. 1944 */ 1945 inspr = NULL; 1946 } else { 1947 /* 1948 * Take two passes through the allprison list: first starting 1949 * with the proposed jid, then ending with it. 1950 */ 1951 for (maxid = JAIL_MAX; maxid != 0; ) { 1952 TAILQ_FOREACH(inspr, &allprison, pr_list) { 1953 if (inspr->pr_id < jid) 1954 continue; 1955 if (inspr->pr_id > jid || 1956 refcount_load(&inspr->pr_ref) == 0) { 1957 /* 1958 * Found an opening. This may be a gap 1959 * in the list, or a dead jail with the 1960 * same ID. 1961 */ 1962 maxid = 0; 1963 break; 1964 } 1965 if (++jid > maxid) { 1966 if (lastprid == maxid || lastprid == 0) 1967 { 1968 /* 1969 * The entire legal range 1970 * has been traversed 1971 */ 1972 return 0; 1973 } 1974 /* Try again from the start. */ 1975 jid = 1; 1976 maxid = lastprid; 1977 break; 1978 } 1979 } 1980 if (inspr == NULL) { 1981 /* Found room at the end of the list. */ 1982 break; 1983 } 1984 } 1985 } 1986 *insprp = inspr; 1987 lastprid = jid; 1988 return (jid); 1989 } 1990 1991 /* 1992 * struct jail_get_args { 1993 * struct iovec *iovp; 1994 * unsigned int iovcnt; 1995 * int flags; 1996 * }; 1997 */ 1998 int 1999 sys_jail_get(struct thread *td, struct jail_get_args *uap) 2000 { 2001 struct uio *auio; 2002 int error; 2003 2004 /* Check that we have an even number of iovecs. */ 2005 if (uap->iovcnt & 1) 2006 return (EINVAL); 2007 2008 error = copyinuio(uap->iovp, uap->iovcnt, &auio); 2009 if (error) 2010 return (error); 2011 error = kern_jail_get(td, auio, uap->flags); 2012 if (error == 0) 2013 error = copyout(auio->uio_iov, uap->iovp, 2014 uap->iovcnt * sizeof (struct iovec)); 2015 free(auio, M_IOV); 2016 return (error); 2017 } 2018 2019 int 2020 kern_jail_get(struct thread *td, struct uio *optuio, int flags) 2021 { 2022 struct bool_flags *bf; 2023 struct jailsys_flags *jsf; 2024 struct prison *pr, *mypr; 2025 struct vfsopt *opt; 2026 struct vfsoptlist *opts; 2027 char *errmsg, *name; 2028 int drflags, error, errmsg_len, errmsg_pos, i, jid, len, pos; 2029 unsigned f; 2030 2031 if (flags & ~JAIL_GET_MASK) 2032 return (EINVAL); 2033 2034 /* Get the parameter list. */ 2035 error = vfs_buildopts(optuio, &opts); 2036 if (error) 2037 return (error); 2038 errmsg_pos = vfs_getopt_pos(opts, "errmsg"); 2039 mypr = td->td_ucred->cr_prison; 2040 pr = NULL; 2041 2042 /* 2043 * Find the prison specified by one of: lastjid, jid, name. 2044 */ 2045 sx_slock(&allprison_lock); 2046 drflags = PD_LIST_SLOCKED; 2047 error = vfs_copyopt(opts, "lastjid", &jid, sizeof(jid)); 2048 if (error == 0) { 2049 TAILQ_FOREACH(pr, &allprison, pr_list) { 2050 if (pr->pr_id > jid && prison_ischild(mypr, pr)) { 2051 mtx_lock(&pr->pr_mtx); 2052 if ((flags & JAIL_DYING) 2053 ? prison_isvalid(pr) : prison_isalive(pr)) 2054 break; 2055 mtx_unlock(&pr->pr_mtx); 2056 } 2057 } 2058 if (pr != NULL) { 2059 drflags |= PD_LOCKED; 2060 goto found_prison; 2061 } 2062 error = ENOENT; 2063 vfs_opterror(opts, "no jail after %d", jid); 2064 goto done; 2065 } else if (error != ENOENT) 2066 goto done; 2067 2068 error = vfs_copyopt(opts, "jid", &jid, sizeof(jid)); 2069 if (error == 0) { 2070 if (jid != 0) { 2071 pr = prison_find_child(mypr, jid); 2072 if (pr != NULL) { 2073 drflags |= PD_LOCKED; 2074 if (!(prison_isalive(pr) || 2075 (flags & JAIL_DYING))) { 2076 error = ENOENT; 2077 vfs_opterror(opts, "jail %d is dying", 2078 jid); 2079 goto done; 2080 } 2081 goto found_prison; 2082 } 2083 error = ENOENT; 2084 vfs_opterror(opts, "jail %d not found", jid); 2085 goto done; 2086 } 2087 } else if (error != ENOENT) 2088 goto done; 2089 2090 error = vfs_getopt(opts, "name", (void **)&name, &len); 2091 if (error == 0) { 2092 if (len == 0 || name[len - 1] != '\0') { 2093 error = EINVAL; 2094 goto done; 2095 } 2096 pr = prison_find_name(mypr, name); 2097 if (pr != NULL) { 2098 drflags |= PD_LOCKED; 2099 if (!(prison_isalive(pr) || (flags & JAIL_DYING))) { 2100 error = ENOENT; 2101 vfs_opterror(opts, "jail \"%s\" is dying", 2102 name); 2103 goto done; 2104 } 2105 goto found_prison; 2106 } 2107 error = ENOENT; 2108 vfs_opterror(opts, "jail \"%s\" not found", name); 2109 goto done; 2110 } else if (error != ENOENT) 2111 goto done; 2112 2113 vfs_opterror(opts, "no jail specified"); 2114 error = ENOENT; 2115 goto done; 2116 2117 found_prison: 2118 /* Get the parameters of the prison. */ 2119 prison_hold(pr); 2120 drflags |= PD_DEREF; 2121 td->td_retval[0] = pr->pr_id; 2122 error = vfs_setopt(opts, "jid", &pr->pr_id, sizeof(pr->pr_id)); 2123 if (error != 0 && error != ENOENT) 2124 goto done; 2125 i = (pr->pr_parent == mypr) ? 0 : pr->pr_parent->pr_id; 2126 error = vfs_setopt(opts, "parent", &i, sizeof(i)); 2127 if (error != 0 && error != ENOENT) 2128 goto done; 2129 error = vfs_setopts(opts, "name", prison_name(mypr, pr)); 2130 if (error != 0 && error != ENOENT) 2131 goto done; 2132 error = vfs_setopt(opts, "cpuset.id", &pr->pr_cpuset->cs_id, 2133 sizeof(pr->pr_cpuset->cs_id)); 2134 if (error != 0 && error != ENOENT) 2135 goto done; 2136 error = vfs_setopts(opts, "path", prison_path(mypr, pr)); 2137 if (error != 0 && error != ENOENT) 2138 goto done; 2139 #ifdef INET 2140 error = vfs_setopt_part(opts, "ip4.addr", pr->pr_ip4, 2141 pr->pr_ip4s * sizeof(*pr->pr_ip4)); 2142 if (error != 0 && error != ENOENT) 2143 goto done; 2144 #endif 2145 #ifdef INET6 2146 error = vfs_setopt_part(opts, "ip6.addr", pr->pr_ip6, 2147 pr->pr_ip6s * sizeof(*pr->pr_ip6)); 2148 if (error != 0 && error != ENOENT) 2149 goto done; 2150 #endif 2151 error = vfs_setopt(opts, "securelevel", &pr->pr_securelevel, 2152 sizeof(pr->pr_securelevel)); 2153 if (error != 0 && error != ENOENT) 2154 goto done; 2155 error = vfs_setopt(opts, "children.cur", &pr->pr_childcount, 2156 sizeof(pr->pr_childcount)); 2157 if (error != 0 && error != ENOENT) 2158 goto done; 2159 error = vfs_setopt(opts, "children.max", &pr->pr_childmax, 2160 sizeof(pr->pr_childmax)); 2161 if (error != 0 && error != ENOENT) 2162 goto done; 2163 error = vfs_setopts(opts, "host.hostname", pr->pr_hostname); 2164 if (error != 0 && error != ENOENT) 2165 goto done; 2166 error = vfs_setopts(opts, "host.domainname", pr->pr_domainname); 2167 if (error != 0 && error != ENOENT) 2168 goto done; 2169 error = vfs_setopts(opts, "host.hostuuid", pr->pr_hostuuid); 2170 if (error != 0 && error != ENOENT) 2171 goto done; 2172 #ifdef COMPAT_FREEBSD32 2173 if (SV_PROC_FLAG(td->td_proc, SV_ILP32)) { 2174 uint32_t hid32 = pr->pr_hostid; 2175 2176 error = vfs_setopt(opts, "host.hostid", &hid32, sizeof(hid32)); 2177 } else 2178 #endif 2179 error = vfs_setopt(opts, "host.hostid", &pr->pr_hostid, 2180 sizeof(pr->pr_hostid)); 2181 if (error != 0 && error != ENOENT) 2182 goto done; 2183 error = vfs_setopt(opts, "enforce_statfs", &pr->pr_enforce_statfs, 2184 sizeof(pr->pr_enforce_statfs)); 2185 if (error != 0 && error != ENOENT) 2186 goto done; 2187 error = vfs_setopt(opts, "devfs_ruleset", &pr->pr_devfs_rsnum, 2188 sizeof(pr->pr_devfs_rsnum)); 2189 if (error != 0 && error != ENOENT) 2190 goto done; 2191 for (bf = pr_flag_bool; 2192 bf < pr_flag_bool + nitems(pr_flag_bool); 2193 bf++) { 2194 i = (pr->pr_flags & bf->flag) ? 1 : 0; 2195 error = vfs_setopt(opts, bf->name, &i, sizeof(i)); 2196 if (error != 0 && error != ENOENT) 2197 goto done; 2198 i = !i; 2199 error = vfs_setopt(opts, bf->noname, &i, sizeof(i)); 2200 if (error != 0 && error != ENOENT) 2201 goto done; 2202 } 2203 for (jsf = pr_flag_jailsys; 2204 jsf < pr_flag_jailsys + nitems(pr_flag_jailsys); 2205 jsf++) { 2206 f = pr->pr_flags & (jsf->disable | jsf->new); 2207 i = (f != 0 && f == jsf->disable) ? JAIL_SYS_DISABLE 2208 : (f == jsf->new) ? JAIL_SYS_NEW 2209 : JAIL_SYS_INHERIT; 2210 error = vfs_setopt(opts, jsf->name, &i, sizeof(i)); 2211 if (error != 0 && error != ENOENT) 2212 goto done; 2213 } 2214 for (bf = pr_flag_allow; 2215 bf < pr_flag_allow + nitems(pr_flag_allow) && 2216 atomic_load_int(&bf->flag) != 0; 2217 bf++) { 2218 i = (pr->pr_allow & bf->flag) ? 1 : 0; 2219 error = vfs_setopt(opts, bf->name, &i, sizeof(i)); 2220 if (error != 0 && error != ENOENT) 2221 goto done; 2222 i = !i; 2223 error = vfs_setopt(opts, bf->noname, &i, sizeof(i)); 2224 if (error != 0 && error != ENOENT) 2225 goto done; 2226 } 2227 i = !prison_isalive(pr); 2228 error = vfs_setopt(opts, "dying", &i, sizeof(i)); 2229 if (error != 0 && error != ENOENT) 2230 goto done; 2231 i = !i; 2232 error = vfs_setopt(opts, "nodying", &i, sizeof(i)); 2233 if (error != 0 && error != ENOENT) 2234 goto done; 2235 error = vfs_setopt(opts, "osreldate", &pr->pr_osreldate, 2236 sizeof(pr->pr_osreldate)); 2237 if (error != 0 && error != ENOENT) 2238 goto done; 2239 error = vfs_setopts(opts, "osrelease", pr->pr_osrelease); 2240 if (error != 0 && error != ENOENT) 2241 goto done; 2242 2243 /* Get the module parameters. */ 2244 mtx_unlock(&pr->pr_mtx); 2245 drflags &= ~PD_LOCKED; 2246 error = osd_jail_call(pr, PR_METHOD_GET, opts); 2247 if (error) 2248 goto done; 2249 prison_deref(pr, drflags); 2250 pr = NULL; 2251 drflags = 0; 2252 2253 /* By now, all parameters should have been noted. */ 2254 TAILQ_FOREACH(opt, opts, link) { 2255 if (!opt->seen && strcmp(opt->name, "errmsg")) { 2256 error = EINVAL; 2257 vfs_opterror(opts, "unknown parameter: %s", opt->name); 2258 goto done; 2259 } 2260 } 2261 2262 /* Write the fetched parameters back to userspace. */ 2263 error = 0; 2264 TAILQ_FOREACH(opt, opts, link) { 2265 if (opt->pos >= 0 && opt->pos != errmsg_pos) { 2266 pos = 2 * opt->pos + 1; 2267 optuio->uio_iov[pos].iov_len = opt->len; 2268 if (opt->value != NULL) { 2269 if (optuio->uio_segflg == UIO_SYSSPACE) { 2270 bcopy(opt->value, 2271 optuio->uio_iov[pos].iov_base, 2272 opt->len); 2273 } else { 2274 error = copyout(opt->value, 2275 optuio->uio_iov[pos].iov_base, 2276 opt->len); 2277 if (error) 2278 break; 2279 } 2280 } 2281 } 2282 } 2283 2284 done: 2285 /* Release any temporary prison holds and/or locks. */ 2286 if (pr != NULL) 2287 prison_deref(pr, drflags); 2288 else if (drflags & PD_LIST_SLOCKED) 2289 sx_sunlock(&allprison_lock); 2290 if (error && errmsg_pos >= 0) { 2291 /* Write the error message back to userspace. */ 2292 vfs_getopt(opts, "errmsg", (void **)&errmsg, &errmsg_len); 2293 errmsg_pos = 2 * errmsg_pos + 1; 2294 if (errmsg_len > 0) { 2295 if (optuio->uio_segflg == UIO_SYSSPACE) 2296 bcopy(errmsg, 2297 optuio->uio_iov[errmsg_pos].iov_base, 2298 errmsg_len); 2299 else 2300 copyout(errmsg, 2301 optuio->uio_iov[errmsg_pos].iov_base, 2302 errmsg_len); 2303 } 2304 } 2305 vfs_freeopts(opts); 2306 return (error); 2307 } 2308 2309 /* 2310 * struct jail_remove_args { 2311 * int jid; 2312 * }; 2313 */ 2314 int 2315 sys_jail_remove(struct thread *td, struct jail_remove_args *uap) 2316 { 2317 struct prison *pr, *cpr, *lpr, *tpr; 2318 int descend, error; 2319 2320 error = priv_check(td, PRIV_JAIL_REMOVE); 2321 if (error) 2322 return (error); 2323 2324 sx_xlock(&allprison_lock); 2325 pr = prison_find_child(td->td_ucred->cr_prison, uap->jid); 2326 if (pr == NULL) { 2327 sx_xunlock(&allprison_lock); 2328 return (EINVAL); 2329 } 2330 2331 /* Remove all descendants of this prison, then remove this prison. */ 2332 prison_hold(pr); 2333 if (!LIST_EMPTY(&pr->pr_children)) { 2334 mtx_unlock(&pr->pr_mtx); 2335 lpr = NULL; 2336 FOREACH_PRISON_DESCENDANT(pr, cpr, descend) { 2337 mtx_lock(&cpr->pr_mtx); 2338 if (prison_isvalid(cpr)) { 2339 tpr = cpr; 2340 prison_hold(cpr); 2341 } else { 2342 /* Already removed - do not do it again. */ 2343 tpr = NULL; 2344 } 2345 mtx_unlock(&cpr->pr_mtx); 2346 if (lpr != NULL) { 2347 mtx_lock(&lpr->pr_mtx); 2348 prison_remove_one(lpr); 2349 sx_xlock(&allprison_lock); 2350 } 2351 lpr = tpr; 2352 } 2353 if (lpr != NULL) { 2354 mtx_lock(&lpr->pr_mtx); 2355 prison_remove_one(lpr); 2356 sx_xlock(&allprison_lock); 2357 } 2358 mtx_lock(&pr->pr_mtx); 2359 } 2360 prison_remove_one(pr); 2361 return (0); 2362 } 2363 2364 static void 2365 prison_remove_one(struct prison *pr) 2366 { 2367 struct proc *p; 2368 int drflags; 2369 2370 drflags = PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED; 2371 2372 /* 2373 * Mark the prison as doomed, so it doesn't accidentally come back 2374 * to life. It may still be explicitly brought back by jail_set(2). 2375 */ 2376 pr->pr_flags |= PR_REMOVE; 2377 2378 /* If the prison was persistent, it is not anymore. */ 2379 if (pr->pr_flags & PR_PERSIST) { 2380 refcount_release(&pr->pr_ref); 2381 drflags |= PD_DEUREF; 2382 pr->pr_flags &= ~PR_PERSIST; 2383 } 2384 2385 /* 2386 * jail_remove added a reference. If that's the only one, remove 2387 * the prison now. refcount(9) doesn't guarantee the cache coherence 2388 * of non-zero counters, so force it here. 2389 */ 2390 KASSERT(refcount_load(&pr->pr_ref) > 0, 2391 ("prison_remove_one removing a dead prison (jid=%d)", pr->pr_id)); 2392 if (atomic_load_acq_int(&pr->pr_ref) == 1) { 2393 prison_deref(pr, drflags); 2394 return; 2395 } 2396 2397 mtx_unlock(&pr->pr_mtx); 2398 sx_xunlock(&allprison_lock); 2399 drflags &= ~(PD_LOCKED | PD_LIST_XLOCKED); 2400 /* 2401 * Kill all processes unfortunate enough to be attached to this prison. 2402 */ 2403 sx_slock(&allproc_lock); 2404 FOREACH_PROC_IN_SYSTEM(p) { 2405 PROC_LOCK(p); 2406 if (p->p_state != PRS_NEW && p->p_ucred && 2407 p->p_ucred->cr_prison == pr) 2408 kern_psignal(p, SIGKILL); 2409 PROC_UNLOCK(p); 2410 } 2411 sx_sunlock(&allproc_lock); 2412 /* Remove the temporary reference added by jail_remove. */ 2413 prison_deref(pr, drflags); 2414 } 2415 2416 /* 2417 * struct jail_attach_args { 2418 * int jid; 2419 * }; 2420 */ 2421 int 2422 sys_jail_attach(struct thread *td, struct jail_attach_args *uap) 2423 { 2424 struct prison *pr; 2425 int error; 2426 2427 error = priv_check(td, PRIV_JAIL_ATTACH); 2428 if (error) 2429 return (error); 2430 2431 /* 2432 * Start with exclusive hold on allprison_lock to ensure that a possible 2433 * PR_METHOD_REMOVE call isn't concurrent with jail_set or jail_remove. 2434 * But then immediately downgrade it since we don't need to stop 2435 * readers. 2436 */ 2437 sx_xlock(&allprison_lock); 2438 sx_downgrade(&allprison_lock); 2439 pr = prison_find_child(td->td_ucred->cr_prison, uap->jid); 2440 if (pr == NULL) { 2441 sx_sunlock(&allprison_lock); 2442 return (EINVAL); 2443 } 2444 2445 /* Do not allow a process to attach to a prison that is not alive. */ 2446 if (!prison_isalive(pr)) { 2447 mtx_unlock(&pr->pr_mtx); 2448 sx_sunlock(&allprison_lock); 2449 return (EINVAL); 2450 } 2451 2452 return (do_jail_attach(td, pr)); 2453 } 2454 2455 static int 2456 do_jail_attach(struct thread *td, struct prison *pr) 2457 { 2458 struct proc *p; 2459 struct ucred *newcred, *oldcred; 2460 int error; 2461 2462 /* 2463 * XXX: Note that there is a slight race here if two threads 2464 * in the same privileged process attempt to attach to two 2465 * different jails at the same time. It is important for 2466 * user processes not to do this, or they might end up with 2467 * a process root from one prison, but attached to the jail 2468 * of another. 2469 */ 2470 refcount_acquire(&pr->pr_ref); 2471 refcount_acquire(&pr->pr_uref); 2472 mtx_unlock(&pr->pr_mtx); 2473 2474 /* Let modules do whatever they need to prepare for attaching. */ 2475 error = osd_jail_call(pr, PR_METHOD_ATTACH, td); 2476 if (error) { 2477 prison_deref(pr, PD_DEREF | PD_DEUREF | PD_LIST_SLOCKED); 2478 return (error); 2479 } 2480 sx_sunlock(&allprison_lock); 2481 2482 /* 2483 * Reparent the newly attached process to this jail. 2484 */ 2485 p = td->td_proc; 2486 error = cpuset_setproc_update_set(p, pr->pr_cpuset); 2487 if (error) 2488 goto e_revert_osd; 2489 2490 vn_lock(pr->pr_root, LK_EXCLUSIVE | LK_RETRY); 2491 if ((error = change_dir(pr->pr_root, td)) != 0) 2492 goto e_unlock; 2493 #ifdef MAC 2494 if ((error = mac_vnode_check_chroot(td->td_ucred, pr->pr_root))) 2495 goto e_unlock; 2496 #endif 2497 VOP_UNLOCK(pr->pr_root); 2498 if ((error = pwd_chroot_chdir(td, pr->pr_root))) 2499 goto e_revert_osd; 2500 2501 newcred = crget(); 2502 PROC_LOCK(p); 2503 oldcred = crcopysafe(p, newcred); 2504 newcred->cr_prison = pr; 2505 proc_set_cred(p, newcred); 2506 setsugid(p); 2507 #ifdef RACCT 2508 racct_proc_ucred_changed(p, oldcred, newcred); 2509 crhold(newcred); 2510 #endif 2511 PROC_UNLOCK(p); 2512 #ifdef RCTL 2513 rctl_proc_ucred_changed(p, newcred); 2514 crfree(newcred); 2515 #endif 2516 prison_deref(oldcred->cr_prison, PD_DEREF | PD_DEUREF); 2517 crfree(oldcred); 2518 2519 /* 2520 * If the prison was killed while changing credentials, die along 2521 * with it. 2522 */ 2523 if (!prison_isalive(pr)) { 2524 PROC_LOCK(p); 2525 kern_psignal(p, SIGKILL); 2526 PROC_UNLOCK(p); 2527 } 2528 2529 return (0); 2530 2531 e_unlock: 2532 VOP_UNLOCK(pr->pr_root); 2533 e_revert_osd: 2534 /* Tell modules this thread is still in its old jail after all. */ 2535 sx_slock(&allprison_lock); 2536 (void)osd_jail_call(td->td_ucred->cr_prison, PR_METHOD_ATTACH, td); 2537 prison_deref(pr, PD_DEREF | PD_DEUREF | PD_LIST_SLOCKED); 2538 return (error); 2539 } 2540 2541 /* 2542 * Returns a locked prison instance, or NULL on failure. 2543 */ 2544 struct prison * 2545 prison_find(int prid) 2546 { 2547 struct prison *pr; 2548 2549 sx_assert(&allprison_lock, SX_LOCKED); 2550 TAILQ_FOREACH(pr, &allprison, pr_list) { 2551 if (pr->pr_id == prid) { 2552 mtx_lock(&pr->pr_mtx); 2553 if (prison_isvalid(pr)) 2554 return (pr); 2555 /* 2556 * Any active prison with the same ID would have 2557 * been inserted before a dead one. 2558 */ 2559 mtx_unlock(&pr->pr_mtx); 2560 break; 2561 } 2562 if (pr->pr_id > prid) 2563 break; 2564 } 2565 return (NULL); 2566 } 2567 2568 /* 2569 * Find a prison that is a descendant of mypr. Returns a locked prison or NULL. 2570 */ 2571 struct prison * 2572 prison_find_child(struct prison *mypr, int prid) 2573 { 2574 struct prison *pr; 2575 int descend; 2576 2577 sx_assert(&allprison_lock, SX_LOCKED); 2578 FOREACH_PRISON_DESCENDANT(mypr, pr, descend) { 2579 if (pr->pr_id == prid) { 2580 mtx_lock(&pr->pr_mtx); 2581 if (prison_isvalid(pr)) 2582 return (pr); 2583 mtx_unlock(&pr->pr_mtx); 2584 } 2585 } 2586 return (NULL); 2587 } 2588 2589 /* 2590 * Look for the name relative to mypr. Returns a locked prison or NULL. 2591 */ 2592 struct prison * 2593 prison_find_name(struct prison *mypr, const char *name) 2594 { 2595 struct prison *pr, *deadpr; 2596 size_t mylen; 2597 int descend; 2598 2599 sx_assert(&allprison_lock, SX_LOCKED); 2600 mylen = (mypr == &prison0) ? 0 : strlen(mypr->pr_name) + 1; 2601 again: 2602 deadpr = NULL; 2603 FOREACH_PRISON_DESCENDANT(mypr, pr, descend) { 2604 if (!strcmp(pr->pr_name + mylen, name)) { 2605 mtx_lock(&pr->pr_mtx); 2606 if (prison_isalive(pr)) 2607 return (pr); 2608 if (prison_isvalid(pr)) 2609 deadpr = pr; 2610 mtx_unlock(&pr->pr_mtx); 2611 } 2612 } 2613 /* There was no valid prison - perhaps there was a dying one. */ 2614 if (deadpr != NULL) { 2615 mtx_lock(&deadpr->pr_mtx); 2616 if (!prison_isvalid(deadpr)) { 2617 mtx_unlock(&deadpr->pr_mtx); 2618 goto again; 2619 } 2620 } 2621 return (deadpr); 2622 } 2623 2624 /* 2625 * See if a prison has the specific flag set. The prison should be locked, 2626 * unless checking for flags that are only set at jail creation (such as 2627 * PR_IP4 and PR_IP6), or only the single bit is examined, without regard 2628 * to any other prison data. 2629 */ 2630 int 2631 prison_flag(struct ucred *cred, unsigned flag) 2632 { 2633 2634 return (cred->cr_prison->pr_flags & flag); 2635 } 2636 2637 int 2638 prison_allow(struct ucred *cred, unsigned flag) 2639 { 2640 2641 return ((cred->cr_prison->pr_allow & flag) != 0); 2642 } 2643 2644 /* 2645 * Hold a prison reference, by incrementing pr_ref. It is generally 2646 * an error to hold a prison that does not already have a reference. 2647 * A prison record will remain valid as long as it has at least one 2648 * reference, and will not be removed as long as either the prison 2649 * mutex or the allprison lock is held (allprison_lock may be shared). 2650 */ 2651 void 2652 prison_hold_locked(struct prison *pr) 2653 { 2654 2655 /* Locking is no longer required. */ 2656 prison_hold(pr); 2657 } 2658 2659 void 2660 prison_hold(struct prison *pr) 2661 { 2662 #ifdef INVARIANTS 2663 int was_valid = refcount_acquire_if_not_zero(&pr->pr_ref); 2664 2665 KASSERT(was_valid, 2666 ("Trying to hold dead prison %p (jid=%d).", pr, pr->pr_id)); 2667 #else 2668 refcount_acquire(&pr->pr_ref); 2669 #endif 2670 } 2671 2672 /* 2673 * Remove a prison reference. If that was the last reference, the 2674 * prison will be removed (at a later time). Return with the prison 2675 * unlocked. 2676 */ 2677 void 2678 prison_free_locked(struct prison *pr) 2679 { 2680 int lastref; 2681 2682 mtx_assert(&pr->pr_mtx, MA_OWNED); 2683 KASSERT(refcount_load(&pr->pr_ref) > 0, 2684 ("Trying to free dead prison %p (jid=%d).", 2685 pr, pr->pr_id)); 2686 lastref = refcount_release(&pr->pr_ref); 2687 mtx_unlock(&pr->pr_mtx); 2688 if (lastref) { 2689 /* 2690 * Don't remove the prison itself in this context, 2691 * in case there are locks held. 2692 */ 2693 taskqueue_enqueue(taskqueue_thread, &pr->pr_task); 2694 } 2695 } 2696 2697 void 2698 prison_free(struct prison *pr) 2699 { 2700 2701 /* 2702 * Locking is only required when releasing the last reference. 2703 * This allows assurance that a locked prison will remain valid 2704 * until it is unlocked. 2705 */ 2706 KASSERT(refcount_load(&pr->pr_ref) > 0, 2707 ("Trying to free dead prison %p (jid=%d).", 2708 pr, pr->pr_id)); 2709 if (refcount_release_if_not_last(&pr->pr_ref)) 2710 return; 2711 mtx_lock(&pr->pr_mtx); 2712 prison_free_locked(pr); 2713 } 2714 2715 /* 2716 * Hold a a prison for user visibility, by incrementing pr_uref. 2717 * It is generally an error to hold a prison that isn't already 2718 * user-visible, except through the the jail system calls. It is also 2719 * an error to hold an invalid prison. A prison record will remain 2720 * alive as long as it has at least one user reference, and will not 2721 * be set to the dying state was long as the prison mutex is held. 2722 */ 2723 void 2724 prison_proc_hold(struct prison *pr) 2725 { 2726 #ifdef INVARIANTS 2727 int was_alive = refcount_acquire_if_not_zero(&pr->pr_uref); 2728 2729 KASSERT(was_alive, 2730 ("Cannot add a process to a non-alive prison (jid=%d)", pr->pr_id)); 2731 #else 2732 refcount_acquire(&pr->pr_uref); 2733 #endif 2734 } 2735 2736 /* 2737 * Remove a prison user reference. If it was the last reference, the 2738 * prison will be considered "dying", and may be removed once all of 2739 * its references are dropped. 2740 */ 2741 void 2742 prison_proc_free(struct prison *pr) 2743 { 2744 2745 /* 2746 * Locking is only required when releasing the last reference. 2747 * This allows assurance that a locked prison will remain alive 2748 * until it is unlocked. 2749 */ 2750 KASSERT(refcount_load(&pr->pr_uref) > 0, 2751 ("Trying to kill a process in a dead prison (jid=%d)", pr->pr_id)); 2752 if (!refcount_release_if_not_last(&pr->pr_uref)) { 2753 /* 2754 * Don't remove the last user reference in this context, 2755 * which is expected to be a process that is not only locked, 2756 * but also half dead. Add a reference so any calls to 2757 * prison_free() won't re-submit the task. 2758 */ 2759 refcount_acquire(&pr->pr_ref); 2760 taskqueue_enqueue(taskqueue_thread, &pr->pr_task); 2761 } 2762 } 2763 2764 /* 2765 * Complete a call to either prison_free or prison_proc_free. 2766 */ 2767 static void 2768 prison_complete(void *context, int pending) 2769 { 2770 struct prison *pr = context; 2771 2772 sx_xlock(&allprison_lock); 2773 mtx_lock(&pr->pr_mtx); 2774 /* 2775 * If this is completing a call to prison_proc_free, there will still 2776 * be a user reference held; clear that as well as the reference that 2777 * was added. No references are expected if this is completing a call 2778 * to prison_free, but prison_deref is still called for the cleanup. 2779 */ 2780 prison_deref(pr, refcount_load(&pr->pr_uref) > 0 2781 ? PD_DEREF | PD_DEUREF | PD_LOCKED | PD_LIST_XLOCKED 2782 : PD_LOCKED | PD_LIST_XLOCKED); 2783 } 2784 2785 /* 2786 * Remove a prison reference and/or user reference (usually). 2787 * This assumes context that allows sleeping (for allprison_lock), 2788 * with no non-sleeping locks held, except perhaps the prison itself. 2789 * If there are no more references, release and delist the prison. 2790 * On completion, the prison lock and the allprison lock are both 2791 * unlocked. 2792 */ 2793 static void 2794 prison_deref(struct prison *pr, int flags) 2795 { 2796 struct prisonlist freeprison; 2797 struct prison *rpr, *tpr; 2798 int lastref, lasturef; 2799 2800 TAILQ_INIT(&freeprison); 2801 if (!(flags & PD_LOCKED)) 2802 mtx_lock(&pr->pr_mtx); 2803 /* 2804 * Release this prison as requested, which may cause its parent 2805 * to be released, and then maybe its grandparent, etc. 2806 */ 2807 for (;;) { 2808 if (flags & PD_DEUREF) { 2809 KASSERT(refcount_load(&pr->pr_uref) > 0, 2810 ("prison_deref PD_DEUREF on a dead prison (jid=%d)", 2811 pr->pr_id)); 2812 lasturef = refcount_release(&pr->pr_uref); 2813 if (lasturef) 2814 refcount_acquire(&pr->pr_ref); 2815 KASSERT(refcount_load(&prison0.pr_uref) > 0, 2816 ("prison0 pr_uref=0")); 2817 } else 2818 lasturef = 0; 2819 if (flags & PD_DEREF) { 2820 KASSERT(refcount_load(&pr->pr_ref) > 0, 2821 ("prison_deref PD_DEREF on a dead prison (jid=%d)", 2822 pr->pr_id)); 2823 lastref = refcount_release(&pr->pr_ref); 2824 } 2825 else 2826 lastref = refcount_load(&pr->pr_ref) == 0; 2827 mtx_unlock(&pr->pr_mtx); 2828 2829 /* 2830 * Tell the modules if the last user reference was removed 2831 * (even it sticks around in dying state). 2832 */ 2833 if (lasturef) { 2834 if (!(flags & (PD_LIST_SLOCKED | PD_LIST_XLOCKED))) { 2835 if (atomic_load_acq_int(&pr->pr_ref) > 1) { 2836 sx_slock(&allprison_lock); 2837 flags |= PD_LIST_SLOCKED; 2838 } else { 2839 sx_xlock(&allprison_lock); 2840 flags |= PD_LIST_XLOCKED; 2841 } 2842 } 2843 (void)osd_jail_call(pr, PR_METHOD_REMOVE, NULL); 2844 mtx_lock(&pr->pr_mtx); 2845 lastref = refcount_release(&pr->pr_ref); 2846 mtx_unlock(&pr->pr_mtx); 2847 } 2848 2849 if (!lastref) 2850 break; 2851 2852 if (flags & PD_LIST_SLOCKED) { 2853 if (!sx_try_upgrade(&allprison_lock)) { 2854 sx_sunlock(&allprison_lock); 2855 sx_xlock(&allprison_lock); 2856 } 2857 flags &= ~PD_LIST_SLOCKED; 2858 } else if (!(flags & PD_LIST_XLOCKED)) 2859 sx_xlock(&allprison_lock); 2860 flags |= PD_LIST_XLOCKED; 2861 2862 TAILQ_REMOVE(&allprison, pr, pr_list); 2863 LIST_REMOVE(pr, pr_sibling); 2864 TAILQ_INSERT_TAIL(&freeprison, pr, pr_list); 2865 for (tpr = pr->pr_parent; tpr != NULL; tpr = tpr->pr_parent) 2866 tpr->pr_childcount--; 2867 2868 /* Removing a prison frees a reference on its parent. */ 2869 pr = pr->pr_parent; 2870 mtx_lock(&pr->pr_mtx); 2871 flags |= PD_DEREF | PD_DEUREF; 2872 } 2873 2874 /* Release all the prison locks. */ 2875 if (flags & PD_LIST_SLOCKED) 2876 sx_sunlock(&allprison_lock); 2877 else if (flags & PD_LIST_XLOCKED) 2878 sx_xunlock(&allprison_lock); 2879 2880 /* 2881 * Finish removing any unreferenced prisons, which couldn't happen 2882 * while allprison_lock was held (to avoid a LOR on vrele). 2883 */ 2884 TAILQ_FOREACH_SAFE(rpr, &freeprison, pr_list, tpr) { 2885 #ifdef VIMAGE 2886 if (rpr->pr_vnet != rpr->pr_parent->pr_vnet) 2887 vnet_destroy(rpr->pr_vnet); 2888 #endif 2889 if (rpr->pr_root != NULL) 2890 vrele(rpr->pr_root); 2891 mtx_destroy(&rpr->pr_mtx); 2892 #ifdef INET 2893 free(rpr->pr_ip4, M_PRISON); 2894 #endif 2895 #ifdef INET6 2896 free(rpr->pr_ip6, M_PRISON); 2897 #endif 2898 if (rpr->pr_cpuset != NULL) 2899 cpuset_rel(rpr->pr_cpuset); 2900 osd_jail_exit(rpr); 2901 #ifdef RACCT 2902 if (racct_enable) 2903 prison_racct_detach(rpr); 2904 #endif 2905 free(rpr, M_PRISON); 2906 } 2907 } 2908 2909 /* 2910 * Set or clear a permission bit in the pr_allow field, passing restrictions 2911 * (cleared permission) down to child jails. 2912 */ 2913 void 2914 prison_set_allow(struct ucred *cred, unsigned flag, int enable) 2915 { 2916 struct prison *pr; 2917 2918 pr = cred->cr_prison; 2919 sx_slock(&allprison_lock); 2920 mtx_lock(&pr->pr_mtx); 2921 prison_set_allow_locked(pr, flag, enable); 2922 mtx_unlock(&pr->pr_mtx); 2923 sx_sunlock(&allprison_lock); 2924 } 2925 2926 static void 2927 prison_set_allow_locked(struct prison *pr, unsigned flag, int enable) 2928 { 2929 struct prison *cpr; 2930 int descend; 2931 2932 if (enable != 0) 2933 pr->pr_allow |= flag; 2934 else { 2935 pr->pr_allow &= ~flag; 2936 FOREACH_PRISON_DESCENDANT_LOCKED(pr, cpr, descend) 2937 cpr->pr_allow &= ~flag; 2938 } 2939 } 2940 2941 /* 2942 * Check if a jail supports the given address family. 2943 * 2944 * Returns 0 if not jailed or the address family is supported, EAFNOSUPPORT 2945 * if not. 2946 */ 2947 int 2948 prison_check_af(struct ucred *cred, int af) 2949 { 2950 struct prison *pr; 2951 int error; 2952 2953 KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); 2954 2955 pr = cred->cr_prison; 2956 #ifdef VIMAGE 2957 /* Prisons with their own network stack are not limited. */ 2958 if (prison_owns_vnet(cred)) 2959 return (0); 2960 #endif 2961 2962 error = 0; 2963 switch (af) 2964 { 2965 #ifdef INET 2966 case AF_INET: 2967 if (pr->pr_flags & PR_IP4) 2968 { 2969 mtx_lock(&pr->pr_mtx); 2970 if ((pr->pr_flags & PR_IP4) && pr->pr_ip4 == NULL) 2971 error = EAFNOSUPPORT; 2972 mtx_unlock(&pr->pr_mtx); 2973 } 2974 break; 2975 #endif 2976 #ifdef INET6 2977 case AF_INET6: 2978 if (pr->pr_flags & PR_IP6) 2979 { 2980 mtx_lock(&pr->pr_mtx); 2981 if ((pr->pr_flags & PR_IP6) && pr->pr_ip6 == NULL) 2982 error = EAFNOSUPPORT; 2983 mtx_unlock(&pr->pr_mtx); 2984 } 2985 break; 2986 #endif 2987 case AF_LOCAL: 2988 case AF_ROUTE: 2989 break; 2990 default: 2991 if (!(pr->pr_allow & PR_ALLOW_SOCKET_AF)) 2992 error = EAFNOSUPPORT; 2993 } 2994 return (error); 2995 } 2996 2997 /* 2998 * Check if given address belongs to the jail referenced by cred (wrapper to 2999 * prison_check_ip[46]). 3000 * 3001 * Returns 0 if jail doesn't restrict the address family or if address belongs 3002 * to jail, EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if 3003 * the jail doesn't allow the address family. IPv4 Address passed in in NBO. 3004 */ 3005 int 3006 prison_if(struct ucred *cred, const struct sockaddr *sa) 3007 { 3008 #ifdef INET 3009 const struct sockaddr_in *sai; 3010 #endif 3011 #ifdef INET6 3012 const struct sockaddr_in6 *sai6; 3013 #endif 3014 int error; 3015 3016 KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); 3017 KASSERT(sa != NULL, ("%s: sa is NULL", __func__)); 3018 3019 #ifdef VIMAGE 3020 if (prison_owns_vnet(cred)) 3021 return (0); 3022 #endif 3023 3024 error = 0; 3025 switch (sa->sa_family) 3026 { 3027 #ifdef INET 3028 case AF_INET: 3029 sai = (const struct sockaddr_in *)sa; 3030 error = prison_check_ip4(cred, &sai->sin_addr); 3031 break; 3032 #endif 3033 #ifdef INET6 3034 case AF_INET6: 3035 sai6 = (const struct sockaddr_in6 *)sa; 3036 error = prison_check_ip6(cred, &sai6->sin6_addr); 3037 break; 3038 #endif 3039 default: 3040 if (!(cred->cr_prison->pr_allow & PR_ALLOW_SOCKET_AF)) 3041 error = EAFNOSUPPORT; 3042 } 3043 return (error); 3044 } 3045 3046 /* 3047 * Return 0 if jails permit p1 to frob p2, otherwise ESRCH. 3048 */ 3049 int 3050 prison_check(struct ucred *cred1, struct ucred *cred2) 3051 { 3052 3053 return ((cred1->cr_prison == cred2->cr_prison || 3054 prison_ischild(cred1->cr_prison, cred2->cr_prison)) ? 0 : ESRCH); 3055 } 3056 3057 /* 3058 * Return 1 if p2 is a child of p1, otherwise 0. 3059 */ 3060 int 3061 prison_ischild(struct prison *pr1, struct prison *pr2) 3062 { 3063 3064 for (pr2 = pr2->pr_parent; pr2 != NULL; pr2 = pr2->pr_parent) 3065 if (pr1 == pr2) 3066 return (1); 3067 return (0); 3068 } 3069 3070 /* 3071 * Return true if the prison is currently alive. A prison is alive if it is 3072 * valid and holds user references, and it isn't being removed. 3073 */ 3074 bool 3075 prison_isalive(struct prison *pr) 3076 { 3077 3078 if (__predict_false(refcount_load(&pr->pr_ref) == 0)) 3079 return (false); 3080 if (__predict_false(refcount_load(&pr->pr_uref) == 0)) 3081 return (false); 3082 if (__predict_false(pr->pr_flags & PR_REMOVE)) 3083 return (false); 3084 return (true); 3085 } 3086 3087 /* 3088 * Return true if the prison is currently valid. A prison is valid if it has 3089 * been fully created, and is not being destroyed. Note that dying prisons 3090 * are still considered valid. 3091 */ 3092 bool 3093 prison_isvalid(struct prison *pr) 3094 { 3095 3096 if (__predict_false(refcount_load(&pr->pr_ref) == 0)) 3097 return (false); 3098 return (true); 3099 } 3100 3101 /* 3102 * Return 1 if the passed credential is in a jail and that jail does not 3103 * have its own virtual network stack, otherwise 0. 3104 */ 3105 int 3106 jailed_without_vnet(struct ucred *cred) 3107 { 3108 3109 if (!jailed(cred)) 3110 return (0); 3111 #ifdef VIMAGE 3112 if (prison_owns_vnet(cred)) 3113 return (0); 3114 #endif 3115 3116 return (1); 3117 } 3118 3119 /* 3120 * Return the correct hostname (domainname, et al) for the passed credential. 3121 */ 3122 void 3123 getcredhostname(struct ucred *cred, char *buf, size_t size) 3124 { 3125 struct prison *pr; 3126 3127 /* 3128 * A NULL credential can be used to shortcut to the physical 3129 * system's hostname. 3130 */ 3131 pr = (cred != NULL) ? cred->cr_prison : &prison0; 3132 mtx_lock(&pr->pr_mtx); 3133 strlcpy(buf, pr->pr_hostname, size); 3134 mtx_unlock(&pr->pr_mtx); 3135 } 3136 3137 void 3138 getcreddomainname(struct ucred *cred, char *buf, size_t size) 3139 { 3140 3141 mtx_lock(&cred->cr_prison->pr_mtx); 3142 strlcpy(buf, cred->cr_prison->pr_domainname, size); 3143 mtx_unlock(&cred->cr_prison->pr_mtx); 3144 } 3145 3146 void 3147 getcredhostuuid(struct ucred *cred, char *buf, size_t size) 3148 { 3149 3150 mtx_lock(&cred->cr_prison->pr_mtx); 3151 strlcpy(buf, cred->cr_prison->pr_hostuuid, size); 3152 mtx_unlock(&cred->cr_prison->pr_mtx); 3153 } 3154 3155 void 3156 getcredhostid(struct ucred *cred, unsigned long *hostid) 3157 { 3158 3159 mtx_lock(&cred->cr_prison->pr_mtx); 3160 *hostid = cred->cr_prison->pr_hostid; 3161 mtx_unlock(&cred->cr_prison->pr_mtx); 3162 } 3163 3164 void 3165 getjailname(struct ucred *cred, char *name, size_t len) 3166 { 3167 3168 mtx_lock(&cred->cr_prison->pr_mtx); 3169 strlcpy(name, cred->cr_prison->pr_name, len); 3170 mtx_unlock(&cred->cr_prison->pr_mtx); 3171 } 3172 3173 #ifdef VIMAGE 3174 /* 3175 * Determine whether the prison represented by cred owns 3176 * its vnet rather than having it inherited. 3177 * 3178 * Returns 1 in case the prison owns the vnet, 0 otherwise. 3179 */ 3180 int 3181 prison_owns_vnet(struct ucred *cred) 3182 { 3183 3184 /* 3185 * vnets cannot be added/removed after jail creation, 3186 * so no need to lock here. 3187 */ 3188 return (cred->cr_prison->pr_flags & PR_VNET ? 1 : 0); 3189 } 3190 #endif 3191 3192 /* 3193 * Determine whether the subject represented by cred can "see" 3194 * status of a mount point. 3195 * Returns: 0 for permitted, ENOENT otherwise. 3196 * XXX: This function should be called cr_canseemount() and should be 3197 * placed in kern_prot.c. 3198 */ 3199 int 3200 prison_canseemount(struct ucred *cred, struct mount *mp) 3201 { 3202 struct prison *pr; 3203 struct statfs *sp; 3204 size_t len; 3205 3206 pr = cred->cr_prison; 3207 if (pr->pr_enforce_statfs == 0) 3208 return (0); 3209 if (pr->pr_root->v_mount == mp) 3210 return (0); 3211 if (pr->pr_enforce_statfs == 2) 3212 return (ENOENT); 3213 /* 3214 * If jail's chroot directory is set to "/" we should be able to see 3215 * all mount-points from inside a jail. 3216 * This is ugly check, but this is the only situation when jail's 3217 * directory ends with '/'. 3218 */ 3219 if (strcmp(pr->pr_path, "/") == 0) 3220 return (0); 3221 len = strlen(pr->pr_path); 3222 sp = &mp->mnt_stat; 3223 if (strncmp(pr->pr_path, sp->f_mntonname, len) != 0) 3224 return (ENOENT); 3225 /* 3226 * Be sure that we don't have situation where jail's root directory 3227 * is "/some/path" and mount point is "/some/pathpath". 3228 */ 3229 if (sp->f_mntonname[len] != '\0' && sp->f_mntonname[len] != '/') 3230 return (ENOENT); 3231 return (0); 3232 } 3233 3234 void 3235 prison_enforce_statfs(struct ucred *cred, struct mount *mp, struct statfs *sp) 3236 { 3237 char jpath[MAXPATHLEN]; 3238 struct prison *pr; 3239 size_t len; 3240 3241 pr = cred->cr_prison; 3242 if (pr->pr_enforce_statfs == 0) 3243 return; 3244 if (prison_canseemount(cred, mp) != 0) { 3245 bzero(sp->f_mntonname, sizeof(sp->f_mntonname)); 3246 strlcpy(sp->f_mntonname, "[restricted]", 3247 sizeof(sp->f_mntonname)); 3248 return; 3249 } 3250 if (pr->pr_root->v_mount == mp) { 3251 /* 3252 * Clear current buffer data, so we are sure nothing from 3253 * the valid path left there. 3254 */ 3255 bzero(sp->f_mntonname, sizeof(sp->f_mntonname)); 3256 *sp->f_mntonname = '/'; 3257 return; 3258 } 3259 /* 3260 * If jail's chroot directory is set to "/" we should be able to see 3261 * all mount-points from inside a jail. 3262 */ 3263 if (strcmp(pr->pr_path, "/") == 0) 3264 return; 3265 len = strlen(pr->pr_path); 3266 strlcpy(jpath, sp->f_mntonname + len, sizeof(jpath)); 3267 /* 3268 * Clear current buffer data, so we are sure nothing from 3269 * the valid path left there. 3270 */ 3271 bzero(sp->f_mntonname, sizeof(sp->f_mntonname)); 3272 if (*jpath == '\0') { 3273 /* Should never happen. */ 3274 *sp->f_mntonname = '/'; 3275 } else { 3276 strlcpy(sp->f_mntonname, jpath, sizeof(sp->f_mntonname)); 3277 } 3278 } 3279 3280 /* 3281 * Check with permission for a specific privilege is granted within jail. We 3282 * have a specific list of accepted privileges; the rest are denied. 3283 */ 3284 int 3285 prison_priv_check(struct ucred *cred, int priv) 3286 { 3287 struct prison *pr; 3288 int error; 3289 3290 /* 3291 * Some policies have custom handlers. This routine should not be 3292 * called for them. See priv_check_cred(). 3293 */ 3294 switch (priv) { 3295 case PRIV_VFS_LOOKUP: 3296 case PRIV_VFS_GENERATION: 3297 KASSERT(0, ("prison_priv_check instead of a custom handler " 3298 "called for %d\n", priv)); 3299 } 3300 3301 if (!jailed(cred)) 3302 return (0); 3303 3304 #ifdef VIMAGE 3305 /* 3306 * Privileges specific to prisons with a virtual network stack. 3307 * There might be a duplicate entry here in case the privilege 3308 * is only granted conditionally in the legacy jail case. 3309 */ 3310 switch (priv) { 3311 #ifdef notyet 3312 /* 3313 * NFS-specific privileges. 3314 */ 3315 case PRIV_NFS_DAEMON: 3316 case PRIV_NFS_LOCKD: 3317 #endif 3318 /* 3319 * Network stack privileges. 3320 */ 3321 case PRIV_NET_BRIDGE: 3322 case PRIV_NET_GRE: 3323 case PRIV_NET_BPF: 3324 case PRIV_NET_RAW: /* Dup, cond. in legacy jail case. */ 3325 case PRIV_NET_ROUTE: 3326 case PRIV_NET_TAP: 3327 case PRIV_NET_SETIFMTU: 3328 case PRIV_NET_SETIFFLAGS: 3329 case PRIV_NET_SETIFCAP: 3330 case PRIV_NET_SETIFDESCR: 3331 case PRIV_NET_SETIFNAME : 3332 case PRIV_NET_SETIFMETRIC: 3333 case PRIV_NET_SETIFPHYS: 3334 case PRIV_NET_SETIFMAC: 3335 case PRIV_NET_SETLANPCP: 3336 case PRIV_NET_ADDMULTI: 3337 case PRIV_NET_DELMULTI: 3338 case PRIV_NET_HWIOCTL: 3339 case PRIV_NET_SETLLADDR: 3340 case PRIV_NET_ADDIFGROUP: 3341 case PRIV_NET_DELIFGROUP: 3342 case PRIV_NET_IFCREATE: 3343 case PRIV_NET_IFDESTROY: 3344 case PRIV_NET_ADDIFADDR: 3345 case PRIV_NET_DELIFADDR: 3346 case PRIV_NET_LAGG: 3347 case PRIV_NET_GIF: 3348 case PRIV_NET_SETIFVNET: 3349 case PRIV_NET_SETIFFIB: 3350 3351 /* 3352 * 802.11-related privileges. 3353 */ 3354 case PRIV_NET80211_VAP_GETKEY: 3355 case PRIV_NET80211_VAP_MANAGE: 3356 3357 #ifdef notyet 3358 /* 3359 * ATM privileges. 3360 */ 3361 case PRIV_NETATM_CFG: 3362 case PRIV_NETATM_ADD: 3363 case PRIV_NETATM_DEL: 3364 case PRIV_NETATM_SET: 3365 3366 /* 3367 * Bluetooth privileges. 3368 */ 3369 case PRIV_NETBLUETOOTH_RAW: 3370 #endif 3371 3372 /* 3373 * Netgraph and netgraph module privileges. 3374 */ 3375 case PRIV_NETGRAPH_CONTROL: 3376 #ifdef notyet 3377 case PRIV_NETGRAPH_TTY: 3378 #endif 3379 3380 /* 3381 * IPv4 and IPv6 privileges. 3382 */ 3383 case PRIV_NETINET_IPFW: 3384 case PRIV_NETINET_DIVERT: 3385 case PRIV_NETINET_PF: 3386 case PRIV_NETINET_DUMMYNET: 3387 case PRIV_NETINET_CARP: 3388 case PRIV_NETINET_MROUTE: 3389 case PRIV_NETINET_RAW: 3390 case PRIV_NETINET_ADDRCTRL6: 3391 case PRIV_NETINET_ND6: 3392 case PRIV_NETINET_SCOPE6: 3393 case PRIV_NETINET_ALIFETIME6: 3394 case PRIV_NETINET_IPSEC: 3395 case PRIV_NETINET_BINDANY: 3396 3397 #ifdef notyet 3398 /* 3399 * NCP privileges. 3400 */ 3401 case PRIV_NETNCP: 3402 3403 /* 3404 * SMB privileges. 3405 */ 3406 case PRIV_NETSMB: 3407 #endif 3408 3409 /* 3410 * No default: or deny here. 3411 * In case of no permit fall through to next switch(). 3412 */ 3413 if (cred->cr_prison->pr_flags & PR_VNET) 3414 return (0); 3415 } 3416 #endif /* VIMAGE */ 3417 3418 switch (priv) { 3419 /* 3420 * Allow ktrace privileges for root in jail. 3421 */ 3422 case PRIV_KTRACE: 3423 3424 #if 0 3425 /* 3426 * Allow jailed processes to configure audit identity and 3427 * submit audit records (login, etc). In the future we may 3428 * want to further refine the relationship between audit and 3429 * jail. 3430 */ 3431 case PRIV_AUDIT_GETAUDIT: 3432 case PRIV_AUDIT_SETAUDIT: 3433 case PRIV_AUDIT_SUBMIT: 3434 #endif 3435 3436 /* 3437 * Allow jailed processes to manipulate process UNIX 3438 * credentials in any way they see fit. 3439 */ 3440 case PRIV_CRED_SETUID: 3441 case PRIV_CRED_SETEUID: 3442 case PRIV_CRED_SETGID: 3443 case PRIV_CRED_SETEGID: 3444 case PRIV_CRED_SETGROUPS: 3445 case PRIV_CRED_SETREUID: 3446 case PRIV_CRED_SETREGID: 3447 case PRIV_CRED_SETRESUID: 3448 case PRIV_CRED_SETRESGID: 3449 3450 /* 3451 * Jail implements visibility constraints already, so allow 3452 * jailed root to override uid/gid-based constraints. 3453 */ 3454 case PRIV_SEEOTHERGIDS: 3455 case PRIV_SEEOTHERUIDS: 3456 3457 /* 3458 * Jail implements inter-process debugging limits already, so 3459 * allow jailed root various debugging privileges. 3460 */ 3461 case PRIV_DEBUG_DIFFCRED: 3462 case PRIV_DEBUG_SUGID: 3463 case PRIV_DEBUG_UNPRIV: 3464 3465 /* 3466 * Allow jail to set various resource limits and login 3467 * properties, and for now, exceed process resource limits. 3468 */ 3469 case PRIV_PROC_LIMIT: 3470 case PRIV_PROC_SETLOGIN: 3471 case PRIV_PROC_SETRLIMIT: 3472 3473 /* 3474 * System V and POSIX IPC privileges are granted in jail. 3475 */ 3476 case PRIV_IPC_READ: 3477 case PRIV_IPC_WRITE: 3478 case PRIV_IPC_ADMIN: 3479 case PRIV_IPC_MSGSIZE: 3480 case PRIV_MQ_ADMIN: 3481 3482 /* 3483 * Jail operations within a jail work on child jails. 3484 */ 3485 case PRIV_JAIL_ATTACH: 3486 case PRIV_JAIL_SET: 3487 case PRIV_JAIL_REMOVE: 3488 3489 /* 3490 * Jail implements its own inter-process limits, so allow 3491 * root processes in jail to change scheduling on other 3492 * processes in the same jail. Likewise for signalling. 3493 */ 3494 case PRIV_SCHED_DIFFCRED: 3495 case PRIV_SCHED_CPUSET: 3496 case PRIV_SIGNAL_DIFFCRED: 3497 case PRIV_SIGNAL_SUGID: 3498 3499 /* 3500 * Allow jailed processes to write to sysctls marked as jail 3501 * writable. 3502 */ 3503 case PRIV_SYSCTL_WRITEJAIL: 3504 3505 /* 3506 * Allow root in jail to manage a variety of quota 3507 * properties. These should likely be conditional on a 3508 * configuration option. 3509 */ 3510 case PRIV_VFS_GETQUOTA: 3511 case PRIV_VFS_SETQUOTA: 3512 3513 /* 3514 * Since Jail relies on chroot() to implement file system 3515 * protections, grant many VFS privileges to root in jail. 3516 * Be careful to exclude mount-related and NFS-related 3517 * privileges. 3518 */ 3519 case PRIV_VFS_READ: 3520 case PRIV_VFS_WRITE: 3521 case PRIV_VFS_ADMIN: 3522 case PRIV_VFS_EXEC: 3523 case PRIV_VFS_BLOCKRESERVE: /* XXXRW: Slightly surprising. */ 3524 case PRIV_VFS_CHFLAGS_DEV: 3525 case PRIV_VFS_CHOWN: 3526 case PRIV_VFS_CHROOT: 3527 case PRIV_VFS_RETAINSUGID: 3528 case PRIV_VFS_FCHROOT: 3529 case PRIV_VFS_LINK: 3530 case PRIV_VFS_SETGID: 3531 case PRIV_VFS_STAT: 3532 case PRIV_VFS_STICKYFILE: 3533 3534 /* 3535 * As in the non-jail case, non-root users are expected to be 3536 * able to read kernel/phyiscal memory (provided /dev/[k]mem 3537 * exists in the jail and they have permission to access it). 3538 */ 3539 case PRIV_KMEM_READ: 3540 return (0); 3541 3542 /* 3543 * Depending on the global setting, allow privilege of 3544 * setting system flags. 3545 */ 3546 case PRIV_VFS_SYSFLAGS: 3547 if (cred->cr_prison->pr_allow & PR_ALLOW_CHFLAGS) 3548 return (0); 3549 else 3550 return (EPERM); 3551 3552 /* 3553 * Depending on the global setting, allow privilege of 3554 * mounting/unmounting file systems. 3555 */ 3556 case PRIV_VFS_MOUNT: 3557 case PRIV_VFS_UNMOUNT: 3558 case PRIV_VFS_MOUNT_NONUSER: 3559 case PRIV_VFS_MOUNT_OWNER: 3560 pr = cred->cr_prison; 3561 prison_lock(pr); 3562 if (pr->pr_allow & PR_ALLOW_MOUNT && pr->pr_enforce_statfs < 2) 3563 error = 0; 3564 else 3565 error = EPERM; 3566 prison_unlock(pr); 3567 return (error); 3568 3569 /* 3570 * Jails should hold no disposition on the PRIV_VFS_READ_DIR 3571 * policy. priv_check_cred will not specifically allow it, and 3572 * we may want a MAC policy to allow it. 3573 */ 3574 case PRIV_VFS_READ_DIR: 3575 return (0); 3576 3577 /* 3578 * Conditionnaly allow locking (unlocking) physical pages 3579 * in memory. 3580 */ 3581 case PRIV_VM_MLOCK: 3582 case PRIV_VM_MUNLOCK: 3583 if (cred->cr_prison->pr_allow & PR_ALLOW_MLOCK) 3584 return (0); 3585 else 3586 return (EPERM); 3587 3588 /* 3589 * Conditionally allow jailed root to bind reserved ports. 3590 */ 3591 case PRIV_NETINET_RESERVEDPORT: 3592 if (cred->cr_prison->pr_allow & PR_ALLOW_RESERVED_PORTS) 3593 return (0); 3594 else 3595 return (EPERM); 3596 3597 /* 3598 * Allow jailed root to reuse in-use ports. 3599 */ 3600 case PRIV_NETINET_REUSEPORT: 3601 return (0); 3602 3603 /* 3604 * Allow jailed root to set certain IPv4/6 (option) headers. 3605 */ 3606 case PRIV_NETINET_SETHDROPTS: 3607 return (0); 3608 3609 /* 3610 * Conditionally allow creating raw sockets in jail. 3611 */ 3612 case PRIV_NETINET_RAW: 3613 if (cred->cr_prison->pr_allow & PR_ALLOW_RAW_SOCKETS) 3614 return (0); 3615 else 3616 return (EPERM); 3617 3618 /* 3619 * Since jail implements its own visibility limits on netstat 3620 * sysctls, allow getcred. This allows identd to work in 3621 * jail. 3622 */ 3623 case PRIV_NETINET_GETCRED: 3624 return (0); 3625 3626 /* 3627 * Allow jailed root to set loginclass. 3628 */ 3629 case PRIV_PROC_SETLOGINCLASS: 3630 return (0); 3631 3632 /* 3633 * Do not allow a process inside a jail to read the kernel 3634 * message buffer unless explicitly permitted. 3635 */ 3636 case PRIV_MSGBUF: 3637 if (cred->cr_prison->pr_allow & PR_ALLOW_READ_MSGBUF) 3638 return (0); 3639 return (EPERM); 3640 3641 default: 3642 /* 3643 * In all remaining cases, deny the privilege request. This 3644 * includes almost all network privileges, many system 3645 * configuration privileges. 3646 */ 3647 return (EPERM); 3648 } 3649 } 3650 3651 /* 3652 * Return the part of pr2's name that is relative to pr1, or the whole name 3653 * if it does not directly follow. 3654 */ 3655 3656 char * 3657 prison_name(struct prison *pr1, struct prison *pr2) 3658 { 3659 char *name; 3660 3661 /* Jails see themselves as "0" (if they see themselves at all). */ 3662 if (pr1 == pr2) 3663 return "0"; 3664 name = pr2->pr_name; 3665 if (prison_ischild(pr1, pr2)) { 3666 /* 3667 * pr1 isn't locked (and allprison_lock may not be either) 3668 * so its length can't be counted on. But the number of dots 3669 * can be counted on - and counted. 3670 */ 3671 for (; pr1 != &prison0; pr1 = pr1->pr_parent) 3672 name = strchr(name, '.') + 1; 3673 } 3674 return (name); 3675 } 3676 3677 /* 3678 * Return the part of pr2's path that is relative to pr1, or the whole path 3679 * if it does not directly follow. 3680 */ 3681 static char * 3682 prison_path(struct prison *pr1, struct prison *pr2) 3683 { 3684 char *path1, *path2; 3685 int len1; 3686 3687 path1 = pr1->pr_path; 3688 path2 = pr2->pr_path; 3689 if (!strcmp(path1, "/")) 3690 return (path2); 3691 len1 = strlen(path1); 3692 if (strncmp(path1, path2, len1)) 3693 return (path2); 3694 if (path2[len1] == '\0') 3695 return "/"; 3696 if (path2[len1] == '/') 3697 return (path2 + len1); 3698 return (path2); 3699 } 3700 3701 /* 3702 * Jail-related sysctls. 3703 */ 3704 static SYSCTL_NODE(_security, OID_AUTO, jail, CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 3705 "Jails"); 3706 3707 static int 3708 sysctl_jail_list(SYSCTL_HANDLER_ARGS) 3709 { 3710 struct xprison *xp; 3711 struct prison *pr, *cpr; 3712 #ifdef INET 3713 struct in_addr *ip4 = NULL; 3714 int ip4s = 0; 3715 #endif 3716 #ifdef INET6 3717 struct in6_addr *ip6 = NULL; 3718 int ip6s = 0; 3719 #endif 3720 int descend, error; 3721 3722 xp = malloc(sizeof(*xp), M_TEMP, M_WAITOK); 3723 pr = req->td->td_ucred->cr_prison; 3724 error = 0; 3725 sx_slock(&allprison_lock); 3726 FOREACH_PRISON_DESCENDANT(pr, cpr, descend) { 3727 #if defined(INET) || defined(INET6) 3728 again: 3729 #endif 3730 mtx_lock(&cpr->pr_mtx); 3731 #ifdef INET 3732 if (cpr->pr_ip4s > 0) { 3733 if (ip4s < cpr->pr_ip4s) { 3734 ip4s = cpr->pr_ip4s; 3735 mtx_unlock(&cpr->pr_mtx); 3736 ip4 = realloc(ip4, ip4s * 3737 sizeof(struct in_addr), M_TEMP, M_WAITOK); 3738 goto again; 3739 } 3740 bcopy(cpr->pr_ip4, ip4, 3741 cpr->pr_ip4s * sizeof(struct in_addr)); 3742 } 3743 #endif 3744 #ifdef INET6 3745 if (cpr->pr_ip6s > 0) { 3746 if (ip6s < cpr->pr_ip6s) { 3747 ip6s = cpr->pr_ip6s; 3748 mtx_unlock(&cpr->pr_mtx); 3749 ip6 = realloc(ip6, ip6s * 3750 sizeof(struct in6_addr), M_TEMP, M_WAITOK); 3751 goto again; 3752 } 3753 bcopy(cpr->pr_ip6, ip6, 3754 cpr->pr_ip6s * sizeof(struct in6_addr)); 3755 } 3756 #endif 3757 if (!prison_isvalid(cpr)) { 3758 mtx_unlock(&cpr->pr_mtx); 3759 continue; 3760 } 3761 bzero(xp, sizeof(*xp)); 3762 xp->pr_version = XPRISON_VERSION; 3763 xp->pr_id = cpr->pr_id; 3764 xp->pr_state = prison_isalive(cpr) 3765 ? PRISON_STATE_ALIVE : PRISON_STATE_DYING; 3766 strlcpy(xp->pr_path, prison_path(pr, cpr), sizeof(xp->pr_path)); 3767 strlcpy(xp->pr_host, cpr->pr_hostname, sizeof(xp->pr_host)); 3768 strlcpy(xp->pr_name, prison_name(pr, cpr), sizeof(xp->pr_name)); 3769 #ifdef INET 3770 xp->pr_ip4s = cpr->pr_ip4s; 3771 #endif 3772 #ifdef INET6 3773 xp->pr_ip6s = cpr->pr_ip6s; 3774 #endif 3775 mtx_unlock(&cpr->pr_mtx); 3776 error = SYSCTL_OUT(req, xp, sizeof(*xp)); 3777 if (error) 3778 break; 3779 #ifdef INET 3780 if (xp->pr_ip4s > 0) { 3781 error = SYSCTL_OUT(req, ip4, 3782 xp->pr_ip4s * sizeof(struct in_addr)); 3783 if (error) 3784 break; 3785 } 3786 #endif 3787 #ifdef INET6 3788 if (xp->pr_ip6s > 0) { 3789 error = SYSCTL_OUT(req, ip6, 3790 xp->pr_ip6s * sizeof(struct in6_addr)); 3791 if (error) 3792 break; 3793 } 3794 #endif 3795 } 3796 sx_sunlock(&allprison_lock); 3797 free(xp, M_TEMP); 3798 #ifdef INET 3799 free(ip4, M_TEMP); 3800 #endif 3801 #ifdef INET6 3802 free(ip6, M_TEMP); 3803 #endif 3804 return (error); 3805 } 3806 3807 SYSCTL_OID(_security_jail, OID_AUTO, list, 3808 CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0, 3809 sysctl_jail_list, "S", "List of active jails"); 3810 3811 static int 3812 sysctl_jail_jailed(SYSCTL_HANDLER_ARGS) 3813 { 3814 int error, injail; 3815 3816 injail = jailed(req->td->td_ucred); 3817 error = SYSCTL_OUT(req, &injail, sizeof(injail)); 3818 3819 return (error); 3820 } 3821 3822 SYSCTL_PROC(_security_jail, OID_AUTO, jailed, 3823 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0, 3824 sysctl_jail_jailed, "I", "Process in jail?"); 3825 3826 static int 3827 sysctl_jail_vnet(SYSCTL_HANDLER_ARGS) 3828 { 3829 int error, havevnet; 3830 #ifdef VIMAGE 3831 struct ucred *cred = req->td->td_ucred; 3832 3833 havevnet = jailed(cred) && prison_owns_vnet(cred); 3834 #else 3835 havevnet = 0; 3836 #endif 3837 error = SYSCTL_OUT(req, &havevnet, sizeof(havevnet)); 3838 3839 return (error); 3840 } 3841 3842 SYSCTL_PROC(_security_jail, OID_AUTO, vnet, 3843 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0, 3844 sysctl_jail_vnet, "I", "Jail owns vnet?"); 3845 3846 #if defined(INET) || defined(INET6) 3847 SYSCTL_UINT(_security_jail, OID_AUTO, jail_max_af_ips, CTLFLAG_RW, 3848 &jail_max_af_ips, 0, 3849 "Number of IP addresses a jail may have at most per address family (deprecated)"); 3850 #endif 3851 3852 /* 3853 * Default parameters for jail(2) compatibility. For historical reasons, 3854 * the sysctl names have varying similarity to the parameter names. Prisons 3855 * just see their own parameters, and can't change them. 3856 */ 3857 static int 3858 sysctl_jail_default_allow(SYSCTL_HANDLER_ARGS) 3859 { 3860 int error, i; 3861 3862 /* Get the current flag value, and convert it to a boolean. */ 3863 if (req->td->td_ucred->cr_prison == &prison0) { 3864 mtx_lock(&prison0.pr_mtx); 3865 i = (jail_default_allow & arg2) != 0; 3866 mtx_unlock(&prison0.pr_mtx); 3867 } else 3868 i = prison_allow(req->td->td_ucred, arg2); 3869 3870 if (arg1 != NULL) 3871 i = !i; 3872 error = sysctl_handle_int(oidp, &i, 0, req); 3873 if (error || !req->newptr) 3874 return (error); 3875 i = i ? arg2 : 0; 3876 if (arg1 != NULL) 3877 i ^= arg2; 3878 /* 3879 * The sysctls don't have CTLFLAGS_PRISON, so assume prison0 3880 * for writing. 3881 */ 3882 mtx_lock(&prison0.pr_mtx); 3883 jail_default_allow = (jail_default_allow & ~arg2) | i; 3884 mtx_unlock(&prison0.pr_mtx); 3885 return (0); 3886 } 3887 3888 SYSCTL_PROC(_security_jail, OID_AUTO, set_hostname_allowed, 3889 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 3890 NULL, PR_ALLOW_SET_HOSTNAME, sysctl_jail_default_allow, "I", 3891 "Processes in jail can set their hostnames (deprecated)"); 3892 SYSCTL_PROC(_security_jail, OID_AUTO, socket_unixiproute_only, 3893 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 3894 (void *)1, PR_ALLOW_SOCKET_AF, sysctl_jail_default_allow, "I", 3895 "Processes in jail are limited to creating UNIX/IP/route sockets only (deprecated)"); 3896 SYSCTL_PROC(_security_jail, OID_AUTO, sysvipc_allowed, 3897 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 3898 NULL, PR_ALLOW_SYSVIPC, sysctl_jail_default_allow, "I", 3899 "Processes in jail can use System V IPC primitives (deprecated)"); 3900 SYSCTL_PROC(_security_jail, OID_AUTO, allow_raw_sockets, 3901 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 3902 NULL, PR_ALLOW_RAW_SOCKETS, sysctl_jail_default_allow, "I", 3903 "Prison root can create raw sockets (deprecated)"); 3904 SYSCTL_PROC(_security_jail, OID_AUTO, chflags_allowed, 3905 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 3906 NULL, PR_ALLOW_CHFLAGS, sysctl_jail_default_allow, "I", 3907 "Processes in jail can alter system file flags (deprecated)"); 3908 SYSCTL_PROC(_security_jail, OID_AUTO, mount_allowed, 3909 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 3910 NULL, PR_ALLOW_MOUNT, sysctl_jail_default_allow, "I", 3911 "Processes in jail can mount/unmount jail-friendly file systems (deprecated)"); 3912 3913 static int 3914 sysctl_jail_default_level(SYSCTL_HANDLER_ARGS) 3915 { 3916 struct prison *pr; 3917 int level, error; 3918 3919 pr = req->td->td_ucred->cr_prison; 3920 level = (pr == &prison0) ? *(int *)arg1 : *(int *)((char *)pr + arg2); 3921 error = sysctl_handle_int(oidp, &level, 0, req); 3922 if (error || !req->newptr) 3923 return (error); 3924 *(int *)arg1 = level; 3925 return (0); 3926 } 3927 3928 SYSCTL_PROC(_security_jail, OID_AUTO, enforce_statfs, 3929 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 3930 &jail_default_enforce_statfs, offsetof(struct prison, pr_enforce_statfs), 3931 sysctl_jail_default_level, "I", 3932 "Processes in jail cannot see all mounted file systems (deprecated)"); 3933 3934 SYSCTL_PROC(_security_jail, OID_AUTO, devfs_ruleset, 3935 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, 3936 &jail_default_devfs_rsnum, offsetof(struct prison, pr_devfs_rsnum), 3937 sysctl_jail_default_level, "I", 3938 "Ruleset for the devfs filesystem in jail (deprecated)"); 3939 3940 /* 3941 * Nodes to describe jail parameters. Maximum length of string parameters 3942 * is returned in the string itself, and the other parameters exist merely 3943 * to make themselves and their types known. 3944 */ 3945 SYSCTL_NODE(_security_jail, OID_AUTO, param, CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 3946 "Jail parameters"); 3947 3948 int 3949 sysctl_jail_param(SYSCTL_HANDLER_ARGS) 3950 { 3951 int i; 3952 long l; 3953 size_t s; 3954 char numbuf[12]; 3955 3956 switch (oidp->oid_kind & CTLTYPE) 3957 { 3958 case CTLTYPE_LONG: 3959 case CTLTYPE_ULONG: 3960 l = 0; 3961 #ifdef SCTL_MASK32 3962 if (!(req->flags & SCTL_MASK32)) 3963 #endif 3964 return (SYSCTL_OUT(req, &l, sizeof(l))); 3965 case CTLTYPE_INT: 3966 case CTLTYPE_UINT: 3967 i = 0; 3968 return (SYSCTL_OUT(req, &i, sizeof(i))); 3969 case CTLTYPE_STRING: 3970 snprintf(numbuf, sizeof(numbuf), "%jd", (intmax_t)arg2); 3971 return 3972 (sysctl_handle_string(oidp, numbuf, sizeof(numbuf), req)); 3973 case CTLTYPE_STRUCT: 3974 s = (size_t)arg2; 3975 return (SYSCTL_OUT(req, &s, sizeof(s))); 3976 } 3977 return (0); 3978 } 3979 3980 /* 3981 * CTLFLAG_RDTUN in the following indicates jail parameters that can be set at 3982 * jail creation time but cannot be changed in an existing jail. 3983 */ 3984 SYSCTL_JAIL_PARAM(, jid, CTLTYPE_INT | CTLFLAG_RDTUN, "I", "Jail ID"); 3985 SYSCTL_JAIL_PARAM(, parent, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail parent ID"); 3986 SYSCTL_JAIL_PARAM_STRING(, name, CTLFLAG_RW, MAXHOSTNAMELEN, "Jail name"); 3987 SYSCTL_JAIL_PARAM_STRING(, path, CTLFLAG_RDTUN, MAXPATHLEN, "Jail root path"); 3988 SYSCTL_JAIL_PARAM(, securelevel, CTLTYPE_INT | CTLFLAG_RW, 3989 "I", "Jail secure level"); 3990 SYSCTL_JAIL_PARAM(, osreldate, CTLTYPE_INT | CTLFLAG_RDTUN, "I", 3991 "Jail value for kern.osreldate and uname -K"); 3992 SYSCTL_JAIL_PARAM_STRING(, osrelease, CTLFLAG_RDTUN, OSRELEASELEN, 3993 "Jail value for kern.osrelease and uname -r"); 3994 SYSCTL_JAIL_PARAM(, enforce_statfs, CTLTYPE_INT | CTLFLAG_RW, 3995 "I", "Jail cannot see all mounted file systems"); 3996 SYSCTL_JAIL_PARAM(, devfs_ruleset, CTLTYPE_INT | CTLFLAG_RW, 3997 "I", "Ruleset for in-jail devfs mounts"); 3998 SYSCTL_JAIL_PARAM(, persist, CTLTYPE_INT | CTLFLAG_RW, 3999 "B", "Jail persistence"); 4000 #ifdef VIMAGE 4001 SYSCTL_JAIL_PARAM(, vnet, CTLTYPE_INT | CTLFLAG_RDTUN, 4002 "E,jailsys", "Virtual network stack"); 4003 #endif 4004 SYSCTL_JAIL_PARAM(, dying, CTLTYPE_INT | CTLFLAG_RD, 4005 "B", "Jail is in the process of shutting down"); 4006 4007 SYSCTL_JAIL_PARAM_NODE(children, "Number of child jails"); 4008 SYSCTL_JAIL_PARAM(_children, cur, CTLTYPE_INT | CTLFLAG_RD, 4009 "I", "Current number of child jails"); 4010 SYSCTL_JAIL_PARAM(_children, max, CTLTYPE_INT | CTLFLAG_RW, 4011 "I", "Maximum number of child jails"); 4012 4013 SYSCTL_JAIL_PARAM_SYS_NODE(host, CTLFLAG_RW, "Jail host info"); 4014 SYSCTL_JAIL_PARAM_STRING(_host, hostname, CTLFLAG_RW, MAXHOSTNAMELEN, 4015 "Jail hostname"); 4016 SYSCTL_JAIL_PARAM_STRING(_host, domainname, CTLFLAG_RW, MAXHOSTNAMELEN, 4017 "Jail NIS domainname"); 4018 SYSCTL_JAIL_PARAM_STRING(_host, hostuuid, CTLFLAG_RW, HOSTUUIDLEN, 4019 "Jail host UUID"); 4020 SYSCTL_JAIL_PARAM(_host, hostid, CTLTYPE_ULONG | CTLFLAG_RW, 4021 "LU", "Jail host ID"); 4022 4023 SYSCTL_JAIL_PARAM_NODE(cpuset, "Jail cpuset"); 4024 SYSCTL_JAIL_PARAM(_cpuset, id, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail cpuset ID"); 4025 4026 #ifdef INET 4027 SYSCTL_JAIL_PARAM_SYS_NODE(ip4, CTLFLAG_RDTUN, 4028 "Jail IPv4 address virtualization"); 4029 SYSCTL_JAIL_PARAM_STRUCT(_ip4, addr, CTLFLAG_RW, sizeof(struct in_addr), 4030 "S,in_addr,a", "Jail IPv4 addresses"); 4031 SYSCTL_JAIL_PARAM(_ip4, saddrsel, CTLTYPE_INT | CTLFLAG_RW, 4032 "B", "Do (not) use IPv4 source address selection rather than the " 4033 "primary jail IPv4 address."); 4034 #endif 4035 #ifdef INET6 4036 SYSCTL_JAIL_PARAM_SYS_NODE(ip6, CTLFLAG_RDTUN, 4037 "Jail IPv6 address virtualization"); 4038 SYSCTL_JAIL_PARAM_STRUCT(_ip6, addr, CTLFLAG_RW, sizeof(struct in6_addr), 4039 "S,in6_addr,a", "Jail IPv6 addresses"); 4040 SYSCTL_JAIL_PARAM(_ip6, saddrsel, CTLTYPE_INT | CTLFLAG_RW, 4041 "B", "Do (not) use IPv6 source address selection rather than the " 4042 "primary jail IPv6 address."); 4043 #endif 4044 4045 SYSCTL_JAIL_PARAM_NODE(allow, "Jail permission flags"); 4046 SYSCTL_JAIL_PARAM(_allow, set_hostname, CTLTYPE_INT | CTLFLAG_RW, 4047 "B", "Jail may set hostname"); 4048 SYSCTL_JAIL_PARAM(_allow, sysvipc, CTLTYPE_INT | CTLFLAG_RW, 4049 "B", "Jail may use SYSV IPC"); 4050 SYSCTL_JAIL_PARAM(_allow, raw_sockets, CTLTYPE_INT | CTLFLAG_RW, 4051 "B", "Jail may create raw sockets"); 4052 SYSCTL_JAIL_PARAM(_allow, chflags, CTLTYPE_INT | CTLFLAG_RW, 4053 "B", "Jail may alter system file flags"); 4054 SYSCTL_JAIL_PARAM(_allow, quotas, CTLTYPE_INT | CTLFLAG_RW, 4055 "B", "Jail may set file quotas"); 4056 SYSCTL_JAIL_PARAM(_allow, socket_af, CTLTYPE_INT | CTLFLAG_RW, 4057 "B", "Jail may create sockets other than just UNIX/IPv4/IPv6/route"); 4058 SYSCTL_JAIL_PARAM(_allow, mlock, CTLTYPE_INT | CTLFLAG_RW, 4059 "B", "Jail may lock (unlock) physical pages in memory"); 4060 SYSCTL_JAIL_PARAM(_allow, reserved_ports, CTLTYPE_INT | CTLFLAG_RW, 4061 "B", "Jail may bind sockets to reserved ports"); 4062 SYSCTL_JAIL_PARAM(_allow, read_msgbuf, CTLTYPE_INT | CTLFLAG_RW, 4063 "B", "Jail may read the kernel message buffer"); 4064 SYSCTL_JAIL_PARAM(_allow, unprivileged_proc_debug, CTLTYPE_INT | CTLFLAG_RW, 4065 "B", "Unprivileged processes may use process debugging facilities"); 4066 SYSCTL_JAIL_PARAM(_allow, suser, CTLTYPE_INT | CTLFLAG_RW, 4067 "B", "Processes in jail with uid 0 have privilege"); 4068 4069 SYSCTL_JAIL_PARAM_SUBNODE(allow, mount, "Jail mount/unmount permission flags"); 4070 SYSCTL_JAIL_PARAM(_allow_mount, , CTLTYPE_INT | CTLFLAG_RW, 4071 "B", "Jail may mount/unmount jail-friendly file systems in general"); 4072 4073 /* 4074 * Add a dynamic parameter allow.<name>, or allow.<prefix>.<name>. Return 4075 * its associated bit in the pr_allow bitmask, or zero if the parameter was 4076 * not created. 4077 */ 4078 unsigned 4079 prison_add_allow(const char *prefix, const char *name, const char *prefix_descr, 4080 const char *descr) 4081 { 4082 struct bool_flags *bf; 4083 struct sysctl_oid *parent; 4084 char *allow_name, *allow_noname, *allowed; 4085 #ifndef NO_SYSCTL_DESCR 4086 char *descr_deprecated; 4087 #endif 4088 u_int allow_flag; 4089 4090 if (prefix 4091 ? asprintf(&allow_name, M_PRISON, "allow.%s.%s", prefix, name) 4092 < 0 || 4093 asprintf(&allow_noname, M_PRISON, "allow.%s.no%s", prefix, name) 4094 < 0 4095 : asprintf(&allow_name, M_PRISON, "allow.%s", name) < 0 || 4096 asprintf(&allow_noname, M_PRISON, "allow.no%s", name) < 0) { 4097 free(allow_name, M_PRISON); 4098 return 0; 4099 } 4100 4101 /* 4102 * See if this parameter has already beed added, i.e. a module was 4103 * previously loaded/unloaded. 4104 */ 4105 mtx_lock(&prison0.pr_mtx); 4106 for (bf = pr_flag_allow; 4107 bf < pr_flag_allow + nitems(pr_flag_allow) && 4108 atomic_load_int(&bf->flag) != 0; 4109 bf++) { 4110 if (strcmp(bf->name, allow_name) == 0) { 4111 allow_flag = bf->flag; 4112 goto no_add; 4113 } 4114 } 4115 4116 /* 4117 * Find a free bit in pr_allow_all, failing if there are none 4118 * (which shouldn't happen as long as we keep track of how many 4119 * potential dynamic flags exist). 4120 */ 4121 for (allow_flag = 1;; allow_flag <<= 1) { 4122 if (allow_flag == 0) 4123 goto no_add; 4124 if ((pr_allow_all & allow_flag) == 0) 4125 break; 4126 } 4127 4128 /* Note the parameter in the next open slot in pr_flag_allow. */ 4129 for (bf = pr_flag_allow; ; bf++) { 4130 if (bf == pr_flag_allow + nitems(pr_flag_allow)) { 4131 /* This should never happen, but is not fatal. */ 4132 allow_flag = 0; 4133 goto no_add; 4134 } 4135 if (atomic_load_int(&bf->flag) == 0) 4136 break; 4137 } 4138 bf->name = allow_name; 4139 bf->noname = allow_noname; 4140 pr_allow_all |= allow_flag; 4141 /* 4142 * prison0 always has permission for the new parameter. 4143 * Other jails must have it granted to them. 4144 */ 4145 prison0.pr_allow |= allow_flag; 4146 /* The flag indicates a valid entry, so make sure it is set last. */ 4147 atomic_store_rel_int(&bf->flag, allow_flag); 4148 mtx_unlock(&prison0.pr_mtx); 4149 4150 /* 4151 * Create sysctls for the paramter, and the back-compat global 4152 * permission. 4153 */ 4154 parent = prefix 4155 ? SYSCTL_ADD_NODE(NULL, 4156 SYSCTL_CHILDREN(&sysctl___security_jail_param_allow), 4157 OID_AUTO, prefix, CTLFLAG_MPSAFE, 0, prefix_descr) 4158 : &sysctl___security_jail_param_allow; 4159 (void)SYSCTL_ADD_PROC(NULL, SYSCTL_CHILDREN(parent), OID_AUTO, 4160 name, CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 4161 NULL, 0, sysctl_jail_param, "B", descr); 4162 if ((prefix 4163 ? asprintf(&allowed, M_TEMP, "%s_%s_allowed", prefix, name) 4164 : asprintf(&allowed, M_TEMP, "%s_allowed", name)) >= 0) { 4165 #ifndef NO_SYSCTL_DESCR 4166 (void)asprintf(&descr_deprecated, M_TEMP, "%s (deprecated)", 4167 descr); 4168 #endif 4169 (void)SYSCTL_ADD_PROC(NULL, 4170 SYSCTL_CHILDREN(&sysctl___security_jail), OID_AUTO, allowed, 4171 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, NULL, allow_flag, 4172 sysctl_jail_default_allow, "I", descr_deprecated); 4173 #ifndef NO_SYSCTL_DESCR 4174 free(descr_deprecated, M_TEMP); 4175 #endif 4176 free(allowed, M_TEMP); 4177 } 4178 return allow_flag; 4179 4180 no_add: 4181 mtx_unlock(&prison0.pr_mtx); 4182 free(allow_name, M_PRISON); 4183 free(allow_noname, M_PRISON); 4184 return allow_flag; 4185 } 4186 4187 /* 4188 * The VFS system will register jail-aware filesystems here. They each get 4189 * a parameter allow.mount.xxxfs and a flag to check when a jailed user 4190 * attempts to mount. 4191 */ 4192 void 4193 prison_add_vfs(struct vfsconf *vfsp) 4194 { 4195 #ifdef NO_SYSCTL_DESCR 4196 4197 vfsp->vfc_prison_flag = prison_add_allow("mount", vfsp->vfc_name, 4198 NULL, NULL); 4199 #else 4200 char *descr; 4201 4202 (void)asprintf(&descr, M_TEMP, "Jail may mount the %s file system", 4203 vfsp->vfc_name); 4204 vfsp->vfc_prison_flag = prison_add_allow("mount", vfsp->vfc_name, 4205 NULL, descr); 4206 free(descr, M_TEMP); 4207 #endif 4208 } 4209 4210 #ifdef RACCT 4211 void 4212 prison_racct_foreach(void (*callback)(struct racct *racct, 4213 void *arg2, void *arg3), void (*pre)(void), void (*post)(void), 4214 void *arg2, void *arg3) 4215 { 4216 struct prison_racct *prr; 4217 4218 ASSERT_RACCT_ENABLED(); 4219 4220 sx_slock(&allprison_lock); 4221 if (pre != NULL) 4222 (pre)(); 4223 LIST_FOREACH(prr, &allprison_racct, prr_next) 4224 (callback)(prr->prr_racct, arg2, arg3); 4225 if (post != NULL) 4226 (post)(); 4227 sx_sunlock(&allprison_lock); 4228 } 4229 4230 static struct prison_racct * 4231 prison_racct_find_locked(const char *name) 4232 { 4233 struct prison_racct *prr; 4234 4235 ASSERT_RACCT_ENABLED(); 4236 sx_assert(&allprison_lock, SA_XLOCKED); 4237 4238 if (name[0] == '\0' || strlen(name) >= MAXHOSTNAMELEN) 4239 return (NULL); 4240 4241 LIST_FOREACH(prr, &allprison_racct, prr_next) { 4242 if (strcmp(name, prr->prr_name) != 0) 4243 continue; 4244 4245 /* Found prison_racct with a matching name? */ 4246 prison_racct_hold(prr); 4247 return (prr); 4248 } 4249 4250 /* Add new prison_racct. */ 4251 prr = malloc(sizeof(*prr), M_PRISON_RACCT, M_ZERO | M_WAITOK); 4252 racct_create(&prr->prr_racct); 4253 4254 strcpy(prr->prr_name, name); 4255 refcount_init(&prr->prr_refcount, 1); 4256 LIST_INSERT_HEAD(&allprison_racct, prr, prr_next); 4257 4258 return (prr); 4259 } 4260 4261 struct prison_racct * 4262 prison_racct_find(const char *name) 4263 { 4264 struct prison_racct *prr; 4265 4266 ASSERT_RACCT_ENABLED(); 4267 4268 sx_xlock(&allprison_lock); 4269 prr = prison_racct_find_locked(name); 4270 sx_xunlock(&allprison_lock); 4271 return (prr); 4272 } 4273 4274 void 4275 prison_racct_hold(struct prison_racct *prr) 4276 { 4277 4278 ASSERT_RACCT_ENABLED(); 4279 4280 refcount_acquire(&prr->prr_refcount); 4281 } 4282 4283 static void 4284 prison_racct_free_locked(struct prison_racct *prr) 4285 { 4286 4287 ASSERT_RACCT_ENABLED(); 4288 sx_assert(&allprison_lock, SA_XLOCKED); 4289 4290 if (refcount_release(&prr->prr_refcount)) { 4291 racct_destroy(&prr->prr_racct); 4292 LIST_REMOVE(prr, prr_next); 4293 free(prr, M_PRISON_RACCT); 4294 } 4295 } 4296 4297 void 4298 prison_racct_free(struct prison_racct *prr) 4299 { 4300 4301 ASSERT_RACCT_ENABLED(); 4302 sx_assert(&allprison_lock, SA_UNLOCKED); 4303 4304 if (refcount_release_if_not_last(&prr->prr_refcount)) 4305 return; 4306 4307 sx_xlock(&allprison_lock); 4308 prison_racct_free_locked(prr); 4309 sx_xunlock(&allprison_lock); 4310 } 4311 4312 static void 4313 prison_racct_attach(struct prison *pr) 4314 { 4315 struct prison_racct *prr; 4316 4317 ASSERT_RACCT_ENABLED(); 4318 sx_assert(&allprison_lock, SA_XLOCKED); 4319 4320 prr = prison_racct_find_locked(pr->pr_name); 4321 KASSERT(prr != NULL, ("cannot find prison_racct")); 4322 4323 pr->pr_prison_racct = prr; 4324 } 4325 4326 /* 4327 * Handle jail renaming. From the racct point of view, renaming means 4328 * moving from one prison_racct to another. 4329 */ 4330 static void 4331 prison_racct_modify(struct prison *pr) 4332 { 4333 #ifdef RCTL 4334 struct proc *p; 4335 struct ucred *cred; 4336 #endif 4337 struct prison_racct *oldprr; 4338 4339 ASSERT_RACCT_ENABLED(); 4340 4341 sx_slock(&allproc_lock); 4342 sx_xlock(&allprison_lock); 4343 4344 if (strcmp(pr->pr_name, pr->pr_prison_racct->prr_name) == 0) { 4345 sx_xunlock(&allprison_lock); 4346 sx_sunlock(&allproc_lock); 4347 return; 4348 } 4349 4350 oldprr = pr->pr_prison_racct; 4351 pr->pr_prison_racct = NULL; 4352 4353 prison_racct_attach(pr); 4354 4355 /* 4356 * Move resource utilisation records. 4357 */ 4358 racct_move(pr->pr_prison_racct->prr_racct, oldprr->prr_racct); 4359 4360 #ifdef RCTL 4361 /* 4362 * Force rctl to reattach rules to processes. 4363 */ 4364 FOREACH_PROC_IN_SYSTEM(p) { 4365 PROC_LOCK(p); 4366 cred = crhold(p->p_ucred); 4367 PROC_UNLOCK(p); 4368 rctl_proc_ucred_changed(p, cred); 4369 crfree(cred); 4370 } 4371 #endif 4372 4373 sx_sunlock(&allproc_lock); 4374 prison_racct_free_locked(oldprr); 4375 sx_xunlock(&allprison_lock); 4376 } 4377 4378 static void 4379 prison_racct_detach(struct prison *pr) 4380 { 4381 4382 ASSERT_RACCT_ENABLED(); 4383 sx_assert(&allprison_lock, SA_UNLOCKED); 4384 4385 if (pr->pr_prison_racct == NULL) 4386 return; 4387 prison_racct_free(pr->pr_prison_racct); 4388 pr->pr_prison_racct = NULL; 4389 } 4390 #endif /* RACCT */ 4391 4392 #ifdef DDB 4393 4394 static void 4395 db_show_prison(struct prison *pr) 4396 { 4397 struct bool_flags *bf; 4398 struct jailsys_flags *jsf; 4399 #if defined(INET) || defined(INET6) 4400 int ii; 4401 #endif 4402 unsigned f; 4403 #ifdef INET 4404 char ip4buf[INET_ADDRSTRLEN]; 4405 #endif 4406 #ifdef INET6 4407 char ip6buf[INET6_ADDRSTRLEN]; 4408 #endif 4409 4410 db_printf("prison %p:\n", pr); 4411 db_printf(" jid = %d\n", pr->pr_id); 4412 db_printf(" name = %s\n", pr->pr_name); 4413 db_printf(" parent = %p\n", pr->pr_parent); 4414 db_printf(" ref = %d\n", pr->pr_ref); 4415 db_printf(" uref = %d\n", pr->pr_uref); 4416 db_printf(" path = %s\n", pr->pr_path); 4417 db_printf(" cpuset = %d\n", pr->pr_cpuset 4418 ? pr->pr_cpuset->cs_id : -1); 4419 #ifdef VIMAGE 4420 db_printf(" vnet = %p\n", pr->pr_vnet); 4421 #endif 4422 db_printf(" root = %p\n", pr->pr_root); 4423 db_printf(" securelevel = %d\n", pr->pr_securelevel); 4424 db_printf(" devfs_rsnum = %d\n", pr->pr_devfs_rsnum); 4425 db_printf(" children.max = %d\n", pr->pr_childmax); 4426 db_printf(" children.cur = %d\n", pr->pr_childcount); 4427 db_printf(" child = %p\n", LIST_FIRST(&pr->pr_children)); 4428 db_printf(" sibling = %p\n", LIST_NEXT(pr, pr_sibling)); 4429 db_printf(" flags = 0x%x", pr->pr_flags); 4430 for (bf = pr_flag_bool; bf < pr_flag_bool + nitems(pr_flag_bool); bf++) 4431 if (pr->pr_flags & bf->flag) 4432 db_printf(" %s", bf->name); 4433 for (jsf = pr_flag_jailsys; 4434 jsf < pr_flag_jailsys + nitems(pr_flag_jailsys); 4435 jsf++) { 4436 f = pr->pr_flags & (jsf->disable | jsf->new); 4437 db_printf(" %-16s= %s\n", jsf->name, 4438 (f != 0 && f == jsf->disable) ? "disable" 4439 : (f == jsf->new) ? "new" 4440 : "inherit"); 4441 } 4442 db_printf(" allow = 0x%x", pr->pr_allow); 4443 for (bf = pr_flag_allow; 4444 bf < pr_flag_allow + nitems(pr_flag_allow) && 4445 atomic_load_int(&bf->flag) != 0; 4446 bf++) 4447 if (pr->pr_allow & bf->flag) 4448 db_printf(" %s", bf->name); 4449 db_printf("\n"); 4450 db_printf(" enforce_statfs = %d\n", pr->pr_enforce_statfs); 4451 db_printf(" host.hostname = %s\n", pr->pr_hostname); 4452 db_printf(" host.domainname = %s\n", pr->pr_domainname); 4453 db_printf(" host.hostuuid = %s\n", pr->pr_hostuuid); 4454 db_printf(" host.hostid = %lu\n", pr->pr_hostid); 4455 #ifdef INET 4456 db_printf(" ip4s = %d\n", pr->pr_ip4s); 4457 for (ii = 0; ii < pr->pr_ip4s; ii++) 4458 db_printf(" %s %s\n", 4459 ii == 0 ? "ip4.addr =" : " ", 4460 inet_ntoa_r(pr->pr_ip4[ii], ip4buf)); 4461 #endif 4462 #ifdef INET6 4463 db_printf(" ip6s = %d\n", pr->pr_ip6s); 4464 for (ii = 0; ii < pr->pr_ip6s; ii++) 4465 db_printf(" %s %s\n", 4466 ii == 0 ? "ip6.addr =" : " ", 4467 ip6_sprintf(ip6buf, &pr->pr_ip6[ii])); 4468 #endif 4469 } 4470 4471 DB_SHOW_COMMAND(prison, db_show_prison_command) 4472 { 4473 struct prison *pr; 4474 4475 if (!have_addr) { 4476 /* 4477 * Show all prisons in the list, and prison0 which is not 4478 * listed. 4479 */ 4480 db_show_prison(&prison0); 4481 if (!db_pager_quit) { 4482 TAILQ_FOREACH(pr, &allprison, pr_list) { 4483 db_show_prison(pr); 4484 if (db_pager_quit) 4485 break; 4486 } 4487 } 4488 return; 4489 } 4490 4491 if (addr == 0) 4492 pr = &prison0; 4493 else { 4494 /* Look for a prison with the ID and with references. */ 4495 TAILQ_FOREACH(pr, &allprison, pr_list) 4496 if (pr->pr_id == addr && pr->pr_ref > 0) 4497 break; 4498 if (pr == NULL) 4499 /* Look again, without requiring a reference. */ 4500 TAILQ_FOREACH(pr, &allprison, pr_list) 4501 if (pr->pr_id == addr) 4502 break; 4503 if (pr == NULL) 4504 /* Assume address points to a valid prison. */ 4505 pr = (struct prison *)addr; 4506 } 4507 db_show_prison(pr); 4508 } 4509 4510 #endif /* DDB */ 4511