1 /*- 2 * Copyright (c) 1999 Poul-Henning Kamp. 3 * Copyright (c) 2008 Bjoern A. Zeeb. 4 * Copyright (c) 2009 James Gritton. 5 * All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 16 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 17 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 18 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 19 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 20 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 21 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 22 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 23 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 24 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 25 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 26 * SUCH DAMAGE. 27 */ 28 29 #include <sys/cdefs.h> 30 __FBSDID("$FreeBSD$"); 31 32 #include "opt_compat.h" 33 #include "opt_ddb.h" 34 #include "opt_inet.h" 35 #include "opt_inet6.h" 36 37 #include <sys/param.h> 38 #include <sys/types.h> 39 #include <sys/kernel.h> 40 #include <sys/systm.h> 41 #include <sys/errno.h> 42 #include <sys/sysproto.h> 43 #include <sys/malloc.h> 44 #include <sys/osd.h> 45 #include <sys/priv.h> 46 #include <sys/proc.h> 47 #include <sys/taskqueue.h> 48 #include <sys/fcntl.h> 49 #include <sys/jail.h> 50 #include <sys/lock.h> 51 #include <sys/mutex.h> 52 #include <sys/racct.h> 53 #include <sys/refcount.h> 54 #include <sys/sx.h> 55 #include <sys/sysent.h> 56 #include <sys/namei.h> 57 #include <sys/mount.h> 58 #include <sys/queue.h> 59 #include <sys/socket.h> 60 #include <sys/syscallsubr.h> 61 #include <sys/sysctl.h> 62 #include <sys/vnode.h> 63 64 #include <net/if.h> 65 #include <net/vnet.h> 66 67 #include <netinet/in.h> 68 69 #ifdef DDB 70 #include <ddb/ddb.h> 71 #endif /* DDB */ 72 73 #include <security/mac/mac_framework.h> 74 75 #define DEFAULT_HOSTUUID "00000000-0000-0000-0000-000000000000" 76 77 MALLOC_DEFINE(M_PRISON, "prison", "Prison structures"); 78 static MALLOC_DEFINE(M_PRISON_RACCT, "prison_racct", "Prison racct structures"); 79 80 /* Keep struct prison prison0 and some code in kern_jail_set() readable. */ 81 #ifdef INET 82 #ifdef INET6 83 #define _PR_IP_SADDRSEL PR_IP4_SADDRSEL|PR_IP6_SADDRSEL 84 #else 85 #define _PR_IP_SADDRSEL PR_IP4_SADDRSEL 86 #endif 87 #else /* !INET */ 88 #ifdef INET6 89 #define _PR_IP_SADDRSEL PR_IP6_SADDRSEL 90 #else 91 #define _PR_IP_SADDRSEL 0 92 #endif 93 #endif 94 95 /* prison0 describes what is "real" about the system. */ 96 struct prison prison0 = { 97 .pr_id = 0, 98 .pr_name = "0", 99 .pr_ref = 1, 100 .pr_uref = 1, 101 .pr_path = "/", 102 .pr_securelevel = -1, 103 .pr_devfs_rsnum = 0, 104 .pr_childmax = JAIL_MAX, 105 .pr_hostuuid = DEFAULT_HOSTUUID, 106 .pr_children = LIST_HEAD_INITIALIZER(prison0.pr_children), 107 #ifdef VIMAGE 108 .pr_flags = PR_HOST|PR_VNET|_PR_IP_SADDRSEL, 109 #else 110 .pr_flags = PR_HOST|_PR_IP_SADDRSEL, 111 #endif 112 .pr_allow = PR_ALLOW_ALL, 113 }; 114 MTX_SYSINIT(prison0, &prison0.pr_mtx, "jail mutex", MTX_DEF); 115 116 /* allprison, allprison_racct and lastprid are protected by allprison_lock. */ 117 struct sx allprison_lock; 118 SX_SYSINIT(allprison_lock, &allprison_lock, "allprison"); 119 struct prisonlist allprison = TAILQ_HEAD_INITIALIZER(allprison); 120 LIST_HEAD(, prison_racct) allprison_racct; 121 int lastprid = 0; 122 123 static int do_jail_attach(struct thread *td, struct prison *pr); 124 static void prison_complete(void *context, int pending); 125 static void prison_deref(struct prison *pr, int flags); 126 static char *prison_path(struct prison *pr1, struct prison *pr2); 127 static void prison_remove_one(struct prison *pr); 128 #ifdef RACCT 129 static void prison_racct_attach(struct prison *pr); 130 static void prison_racct_modify(struct prison *pr); 131 static void prison_racct_detach(struct prison *pr); 132 #endif 133 134 /* Flags for prison_deref */ 135 #define PD_DEREF 0x01 136 #define PD_DEUREF 0x02 137 #define PD_LOCKED 0x04 138 #define PD_LIST_SLOCKED 0x08 139 #define PD_LIST_XLOCKED 0x10 140 141 /* 142 * Parameter names corresponding to PR_* flag values. Size values are for kvm 143 * as we cannot figure out the size of a sparse array, or an array without a 144 * terminating entry. 145 */ 146 static char *pr_flag_names[] = { 147 [0] = "persist", 148 #ifdef INET 149 [7] = "ip4.saddrsel", 150 #endif 151 #ifdef INET6 152 [8] = "ip6.saddrsel", 153 #endif 154 }; 155 const size_t pr_flag_names_size = sizeof(pr_flag_names); 156 157 static char *pr_flag_nonames[] = { 158 [0] = "nopersist", 159 #ifdef INET 160 [7] = "ip4.nosaddrsel", 161 #endif 162 #ifdef INET6 163 [8] = "ip6.nosaddrsel", 164 #endif 165 }; 166 const size_t pr_flag_nonames_size = sizeof(pr_flag_nonames); 167 168 struct jailsys_flags { 169 const char *name; 170 unsigned disable; 171 unsigned new; 172 } pr_flag_jailsys[] = { 173 { "host", 0, PR_HOST }, 174 #ifdef VIMAGE 175 { "vnet", 0, PR_VNET }, 176 #endif 177 #ifdef INET 178 { "ip4", PR_IP4_USER, PR_IP4_USER }, 179 #endif 180 #ifdef INET6 181 { "ip6", PR_IP6_USER, PR_IP6_USER }, 182 #endif 183 }; 184 const size_t pr_flag_jailsys_size = sizeof(pr_flag_jailsys); 185 186 static char *pr_allow_names[] = { 187 "allow.set_hostname", 188 "allow.sysvipc", 189 "allow.raw_sockets", 190 "allow.chflags", 191 "allow.mount", 192 "allow.quotas", 193 "allow.socket_af", 194 "allow.mount.devfs", 195 "allow.mount.nullfs", 196 "allow.mount.zfs", 197 "allow.mount.procfs", 198 "allow.mount.tmpfs", 199 "allow.mount.fdescfs", 200 "allow.mount.linprocfs", 201 "allow.mount.linsysfs", 202 "allow.reserved_ports", 203 }; 204 const size_t pr_allow_names_size = sizeof(pr_allow_names); 205 206 static char *pr_allow_nonames[] = { 207 "allow.noset_hostname", 208 "allow.nosysvipc", 209 "allow.noraw_sockets", 210 "allow.nochflags", 211 "allow.nomount", 212 "allow.noquotas", 213 "allow.nosocket_af", 214 "allow.mount.nodevfs", 215 "allow.mount.nonullfs", 216 "allow.mount.nozfs", 217 "allow.mount.noprocfs", 218 "allow.mount.notmpfs", 219 "allow.mount.nofdescfs", 220 "allow.mount.nolinprocfs", 221 "allow.mount.nolinsysfs", 222 "allow.noreserved_ports", 223 }; 224 const size_t pr_allow_nonames_size = sizeof(pr_allow_nonames); 225 226 #define JAIL_DEFAULT_ALLOW (PR_ALLOW_SET_HOSTNAME | PR_ALLOW_RESERVED_PORTS) 227 #define JAIL_DEFAULT_ENFORCE_STATFS 2 228 #define JAIL_DEFAULT_DEVFS_RSNUM 0 229 static unsigned jail_default_allow = JAIL_DEFAULT_ALLOW; 230 static int jail_default_enforce_statfs = JAIL_DEFAULT_ENFORCE_STATFS; 231 static int jail_default_devfs_rsnum = JAIL_DEFAULT_DEVFS_RSNUM; 232 #if defined(INET) || defined(INET6) 233 static unsigned jail_max_af_ips = 255; 234 #endif 235 236 /* 237 * Initialize the parts of prison0 that can't be static-initialized with 238 * constants. This is called from proc0_init() after creating thread0 cpuset. 239 */ 240 void 241 prison0_init(void) 242 { 243 244 prison0.pr_cpuset = cpuset_ref(thread0.td_cpuset); 245 prison0.pr_osreldate = osreldate; 246 strlcpy(prison0.pr_osrelease, osrelease, sizeof(prison0.pr_osrelease)); 247 } 248 249 /* 250 * struct jail_args { 251 * struct jail *jail; 252 * }; 253 */ 254 int 255 sys_jail(struct thread *td, struct jail_args *uap) 256 { 257 uint32_t version; 258 int error; 259 struct jail j; 260 261 error = copyin(uap->jail, &version, sizeof(uint32_t)); 262 if (error) 263 return (error); 264 265 switch (version) { 266 case 0: 267 { 268 struct jail_v0 j0; 269 270 /* FreeBSD single IPv4 jails. */ 271 bzero(&j, sizeof(struct jail)); 272 error = copyin(uap->jail, &j0, sizeof(struct jail_v0)); 273 if (error) 274 return (error); 275 j.version = j0.version; 276 j.path = j0.path; 277 j.hostname = j0.hostname; 278 j.ip4s = htonl(j0.ip_number); /* jail_v0 is host order */ 279 break; 280 } 281 282 case 1: 283 /* 284 * Version 1 was used by multi-IPv4 jail implementations 285 * that never made it into the official kernel. 286 */ 287 return (EINVAL); 288 289 case 2: /* JAIL_API_VERSION */ 290 /* FreeBSD multi-IPv4/IPv6,noIP jails. */ 291 error = copyin(uap->jail, &j, sizeof(struct jail)); 292 if (error) 293 return (error); 294 break; 295 296 default: 297 /* Sci-Fi jails are not supported, sorry. */ 298 return (EINVAL); 299 } 300 return (kern_jail(td, &j)); 301 } 302 303 int 304 kern_jail(struct thread *td, struct jail *j) 305 { 306 struct iovec optiov[2 * (4 + nitems(pr_allow_names) 307 #ifdef INET 308 + 1 309 #endif 310 #ifdef INET6 311 + 1 312 #endif 313 )]; 314 struct uio opt; 315 char *u_path, *u_hostname, *u_name; 316 #ifdef INET 317 uint32_t ip4s; 318 struct in_addr *u_ip4; 319 #endif 320 #ifdef INET6 321 struct in6_addr *u_ip6; 322 #endif 323 size_t tmplen; 324 int error, enforce_statfs, fi; 325 326 bzero(&optiov, sizeof(optiov)); 327 opt.uio_iov = optiov; 328 opt.uio_iovcnt = 0; 329 opt.uio_offset = -1; 330 opt.uio_resid = -1; 331 opt.uio_segflg = UIO_SYSSPACE; 332 opt.uio_rw = UIO_READ; 333 opt.uio_td = td; 334 335 /* Set permissions for top-level jails from sysctls. */ 336 if (!jailed(td->td_ucred)) { 337 for (fi = 0; fi < nitems(pr_allow_names); fi++) { 338 optiov[opt.uio_iovcnt].iov_base = 339 (jail_default_allow & (1 << fi)) 340 ? pr_allow_names[fi] : pr_allow_nonames[fi]; 341 optiov[opt.uio_iovcnt].iov_len = 342 strlen(optiov[opt.uio_iovcnt].iov_base) + 1; 343 opt.uio_iovcnt += 2; 344 } 345 optiov[opt.uio_iovcnt].iov_base = "enforce_statfs"; 346 optiov[opt.uio_iovcnt].iov_len = sizeof("enforce_statfs"); 347 opt.uio_iovcnt++; 348 enforce_statfs = jail_default_enforce_statfs; 349 optiov[opt.uio_iovcnt].iov_base = &enforce_statfs; 350 optiov[opt.uio_iovcnt].iov_len = sizeof(enforce_statfs); 351 opt.uio_iovcnt++; 352 } 353 354 tmplen = MAXPATHLEN + MAXHOSTNAMELEN + MAXHOSTNAMELEN; 355 #ifdef INET 356 ip4s = (j->version == 0) ? 1 : j->ip4s; 357 if (ip4s > jail_max_af_ips) 358 return (EINVAL); 359 tmplen += ip4s * sizeof(struct in_addr); 360 #else 361 if (j->ip4s > 0) 362 return (EINVAL); 363 #endif 364 #ifdef INET6 365 if (j->ip6s > jail_max_af_ips) 366 return (EINVAL); 367 tmplen += j->ip6s * sizeof(struct in6_addr); 368 #else 369 if (j->ip6s > 0) 370 return (EINVAL); 371 #endif 372 u_path = malloc(tmplen, M_TEMP, M_WAITOK); 373 u_hostname = u_path + MAXPATHLEN; 374 u_name = u_hostname + MAXHOSTNAMELEN; 375 #ifdef INET 376 u_ip4 = (struct in_addr *)(u_name + MAXHOSTNAMELEN); 377 #endif 378 #ifdef INET6 379 #ifdef INET 380 u_ip6 = (struct in6_addr *)(u_ip4 + ip4s); 381 #else 382 u_ip6 = (struct in6_addr *)(u_name + MAXHOSTNAMELEN); 383 #endif 384 #endif 385 optiov[opt.uio_iovcnt].iov_base = "path"; 386 optiov[opt.uio_iovcnt].iov_len = sizeof("path"); 387 opt.uio_iovcnt++; 388 optiov[opt.uio_iovcnt].iov_base = u_path; 389 error = copyinstr(j->path, u_path, MAXPATHLEN, 390 &optiov[opt.uio_iovcnt].iov_len); 391 if (error) { 392 free(u_path, M_TEMP); 393 return (error); 394 } 395 opt.uio_iovcnt++; 396 optiov[opt.uio_iovcnt].iov_base = "host.hostname"; 397 optiov[opt.uio_iovcnt].iov_len = sizeof("host.hostname"); 398 opt.uio_iovcnt++; 399 optiov[opt.uio_iovcnt].iov_base = u_hostname; 400 error = copyinstr(j->hostname, u_hostname, MAXHOSTNAMELEN, 401 &optiov[opt.uio_iovcnt].iov_len); 402 if (error) { 403 free(u_path, M_TEMP); 404 return (error); 405 } 406 opt.uio_iovcnt++; 407 if (j->jailname != NULL) { 408 optiov[opt.uio_iovcnt].iov_base = "name"; 409 optiov[opt.uio_iovcnt].iov_len = sizeof("name"); 410 opt.uio_iovcnt++; 411 optiov[opt.uio_iovcnt].iov_base = u_name; 412 error = copyinstr(j->jailname, u_name, MAXHOSTNAMELEN, 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 } 420 #ifdef INET 421 optiov[opt.uio_iovcnt].iov_base = "ip4.addr"; 422 optiov[opt.uio_iovcnt].iov_len = sizeof("ip4.addr"); 423 opt.uio_iovcnt++; 424 optiov[opt.uio_iovcnt].iov_base = u_ip4; 425 optiov[opt.uio_iovcnt].iov_len = ip4s * sizeof(struct in_addr); 426 if (j->version == 0) 427 u_ip4->s_addr = j->ip4s; 428 else { 429 error = copyin(j->ip4, u_ip4, optiov[opt.uio_iovcnt].iov_len); 430 if (error) { 431 free(u_path, M_TEMP); 432 return (error); 433 } 434 } 435 opt.uio_iovcnt++; 436 #endif 437 #ifdef INET6 438 optiov[opt.uio_iovcnt].iov_base = "ip6.addr"; 439 optiov[opt.uio_iovcnt].iov_len = sizeof("ip6.addr"); 440 opt.uio_iovcnt++; 441 optiov[opt.uio_iovcnt].iov_base = u_ip6; 442 optiov[opt.uio_iovcnt].iov_len = j->ip6s * sizeof(struct in6_addr); 443 error = copyin(j->ip6, u_ip6, optiov[opt.uio_iovcnt].iov_len); 444 if (error) { 445 free(u_path, M_TEMP); 446 return (error); 447 } 448 opt.uio_iovcnt++; 449 #endif 450 KASSERT(opt.uio_iovcnt <= nitems(optiov), 451 ("kern_jail: too many iovecs (%d)", opt.uio_iovcnt)); 452 error = kern_jail_set(td, &opt, JAIL_CREATE | JAIL_ATTACH); 453 free(u_path, M_TEMP); 454 return (error); 455 } 456 457 458 /* 459 * struct jail_set_args { 460 * struct iovec *iovp; 461 * unsigned int iovcnt; 462 * int flags; 463 * }; 464 */ 465 int 466 sys_jail_set(struct thread *td, struct jail_set_args *uap) 467 { 468 struct uio *auio; 469 int error; 470 471 /* Check that we have an even number of iovecs. */ 472 if (uap->iovcnt & 1) 473 return (EINVAL); 474 475 error = copyinuio(uap->iovp, uap->iovcnt, &auio); 476 if (error) 477 return (error); 478 error = kern_jail_set(td, auio, uap->flags); 479 free(auio, M_IOV); 480 return (error); 481 } 482 483 int 484 kern_jail_set(struct thread *td, struct uio *optuio, int flags) 485 { 486 struct nameidata nd; 487 #ifdef INET 488 struct in_addr *ip4; 489 #endif 490 #ifdef INET6 491 struct in6_addr *ip6; 492 #endif 493 struct vfsopt *opt; 494 struct vfsoptlist *opts; 495 struct prison *pr, *deadpr, *mypr, *ppr, *tpr; 496 struct vnode *root; 497 char *domain, *errmsg, *host, *name, *namelc, *p, *path, *uuid; 498 char *g_path, *osrelstr; 499 #if defined(INET) || defined(INET6) 500 struct prison *tppr; 501 void *op; 502 #endif 503 unsigned long hid; 504 size_t namelen, onamelen, pnamelen; 505 int born, created, cuflags, descend, enforce; 506 int error, errmsg_len, errmsg_pos; 507 int gotchildmax, gotenforce, gothid, gotrsnum, gotslevel; 508 int fi, jid, jsys, len, level; 509 int childmax, osreldt, rsnum, slevel; 510 int fullpath_disabled; 511 #if defined(INET) || defined(INET6) 512 int ii, ij; 513 #endif 514 #ifdef INET 515 int ip4s, redo_ip4; 516 #endif 517 #ifdef INET6 518 int ip6s, redo_ip6; 519 #endif 520 uint64_t pr_allow, ch_allow, pr_flags, ch_flags; 521 unsigned tallow; 522 char numbuf[12]; 523 524 error = priv_check(td, PRIV_JAIL_SET); 525 if (!error && (flags & JAIL_ATTACH)) 526 error = priv_check(td, PRIV_JAIL_ATTACH); 527 if (error) 528 return (error); 529 mypr = td->td_ucred->cr_prison; 530 if ((flags & JAIL_CREATE) && mypr->pr_childmax == 0) 531 return (EPERM); 532 if (flags & ~JAIL_SET_MASK) 533 return (EINVAL); 534 535 /* 536 * Check all the parameters before committing to anything. Not all 537 * errors can be caught early, but we may as well try. Also, this 538 * takes care of some expensive stuff (path lookup) before getting 539 * the allprison lock. 540 * 541 * XXX Jails are not filesystems, and jail parameters are not mount 542 * options. But it makes more sense to re-use the vfsopt code 543 * than duplicate it under a different name. 544 */ 545 error = vfs_buildopts(optuio, &opts); 546 if (error) 547 return (error); 548 #ifdef INET 549 ip4 = NULL; 550 #endif 551 #ifdef INET6 552 ip6 = NULL; 553 #endif 554 g_path = NULL; 555 556 cuflags = flags & (JAIL_CREATE | JAIL_UPDATE); 557 if (!cuflags) { 558 error = EINVAL; 559 vfs_opterror(opts, "no valid operation (create or update)"); 560 goto done_errmsg; 561 } 562 563 error = vfs_copyopt(opts, "jid", &jid, sizeof(jid)); 564 if (error == ENOENT) 565 jid = 0; 566 else if (error != 0) 567 goto done_free; 568 569 error = vfs_copyopt(opts, "securelevel", &slevel, sizeof(slevel)); 570 if (error == ENOENT) 571 gotslevel = 0; 572 else if (error != 0) 573 goto done_free; 574 else 575 gotslevel = 1; 576 577 error = 578 vfs_copyopt(opts, "children.max", &childmax, sizeof(childmax)); 579 if (error == ENOENT) 580 gotchildmax = 0; 581 else if (error != 0) 582 goto done_free; 583 else 584 gotchildmax = 1; 585 586 error = vfs_copyopt(opts, "enforce_statfs", &enforce, sizeof(enforce)); 587 if (error == ENOENT) 588 gotenforce = 0; 589 else if (error != 0) 590 goto done_free; 591 else if (enforce < 0 || enforce > 2) { 592 error = EINVAL; 593 goto done_free; 594 } else 595 gotenforce = 1; 596 597 error = vfs_copyopt(opts, "devfs_ruleset", &rsnum, sizeof(rsnum)); 598 if (error == ENOENT) 599 gotrsnum = 0; 600 else if (error != 0) 601 goto done_free; 602 else 603 gotrsnum = 1; 604 605 pr_flags = ch_flags = 0; 606 for (fi = 0; fi < nitems(pr_flag_names); fi++) { 607 if (pr_flag_names[fi] == NULL) 608 continue; 609 vfs_flagopt(opts, pr_flag_names[fi], &pr_flags, 1 << fi); 610 vfs_flagopt(opts, pr_flag_nonames[fi], &ch_flags, 1 << fi); 611 } 612 ch_flags |= pr_flags; 613 for (fi = 0; fi < nitems(pr_flag_jailsys); fi++) { 614 error = vfs_copyopt(opts, pr_flag_jailsys[fi].name, &jsys, 615 sizeof(jsys)); 616 if (error == ENOENT) 617 continue; 618 if (error != 0) 619 goto done_free; 620 switch (jsys) { 621 case JAIL_SYS_DISABLE: 622 if (!pr_flag_jailsys[fi].disable) { 623 error = EINVAL; 624 goto done_free; 625 } 626 pr_flags |= pr_flag_jailsys[fi].disable; 627 break; 628 case JAIL_SYS_NEW: 629 pr_flags |= pr_flag_jailsys[fi].new; 630 break; 631 case JAIL_SYS_INHERIT: 632 break; 633 default: 634 error = EINVAL; 635 goto done_free; 636 } 637 ch_flags |= 638 pr_flag_jailsys[fi].new | pr_flag_jailsys[fi].disable; 639 } 640 if ((flags & (JAIL_CREATE | JAIL_UPDATE | JAIL_ATTACH)) == JAIL_CREATE 641 && !(pr_flags & PR_PERSIST)) { 642 error = EINVAL; 643 vfs_opterror(opts, "new jail must persist or attach"); 644 goto done_errmsg; 645 } 646 #ifdef VIMAGE 647 if ((flags & JAIL_UPDATE) && (ch_flags & PR_VNET)) { 648 error = EINVAL; 649 vfs_opterror(opts, "vnet cannot be changed after creation"); 650 goto done_errmsg; 651 } 652 #endif 653 #ifdef INET 654 if ((flags & JAIL_UPDATE) && (ch_flags & PR_IP4_USER)) { 655 error = EINVAL; 656 vfs_opterror(opts, "ip4 cannot be changed after creation"); 657 goto done_errmsg; 658 } 659 #endif 660 #ifdef INET6 661 if ((flags & JAIL_UPDATE) && (ch_flags & PR_IP6_USER)) { 662 error = EINVAL; 663 vfs_opterror(opts, "ip6 cannot be changed after creation"); 664 goto done_errmsg; 665 } 666 #endif 667 668 pr_allow = ch_allow = 0; 669 for (fi = 0; fi < nitems(pr_allow_names); fi++) { 670 vfs_flagopt(opts, pr_allow_names[fi], &pr_allow, 1 << fi); 671 vfs_flagopt(opts, pr_allow_nonames[fi], &ch_allow, 1 << fi); 672 } 673 ch_allow |= pr_allow; 674 675 error = vfs_getopt(opts, "name", (void **)&name, &len); 676 if (error == ENOENT) 677 name = NULL; 678 else if (error != 0) 679 goto done_free; 680 else { 681 if (len == 0 || name[len - 1] != '\0') { 682 error = EINVAL; 683 goto done_free; 684 } 685 if (len > MAXHOSTNAMELEN) { 686 error = ENAMETOOLONG; 687 goto done_free; 688 } 689 } 690 691 error = vfs_getopt(opts, "host.hostname", (void **)&host, &len); 692 if (error == ENOENT) 693 host = NULL; 694 else if (error != 0) 695 goto done_free; 696 else { 697 ch_flags |= PR_HOST; 698 pr_flags |= PR_HOST; 699 if (len == 0 || host[len - 1] != '\0') { 700 error = EINVAL; 701 goto done_free; 702 } 703 if (len > MAXHOSTNAMELEN) { 704 error = ENAMETOOLONG; 705 goto done_free; 706 } 707 } 708 709 error = vfs_getopt(opts, "host.domainname", (void **)&domain, &len); 710 if (error == ENOENT) 711 domain = NULL; 712 else if (error != 0) 713 goto done_free; 714 else { 715 ch_flags |= PR_HOST; 716 pr_flags |= PR_HOST; 717 if (len == 0 || domain[len - 1] != '\0') { 718 error = EINVAL; 719 goto done_free; 720 } 721 if (len > MAXHOSTNAMELEN) { 722 error = ENAMETOOLONG; 723 goto done_free; 724 } 725 } 726 727 error = vfs_getopt(opts, "host.hostuuid", (void **)&uuid, &len); 728 if (error == ENOENT) 729 uuid = NULL; 730 else if (error != 0) 731 goto done_free; 732 else { 733 ch_flags |= PR_HOST; 734 pr_flags |= PR_HOST; 735 if (len == 0 || uuid[len - 1] != '\0') { 736 error = EINVAL; 737 goto done_free; 738 } 739 if (len > HOSTUUIDLEN) { 740 error = ENAMETOOLONG; 741 goto done_free; 742 } 743 } 744 745 #ifdef COMPAT_FREEBSD32 746 if (SV_PROC_FLAG(td->td_proc, SV_ILP32)) { 747 uint32_t hid32; 748 749 error = vfs_copyopt(opts, "host.hostid", &hid32, sizeof(hid32)); 750 hid = hid32; 751 } else 752 #endif 753 error = vfs_copyopt(opts, "host.hostid", &hid, sizeof(hid)); 754 if (error == ENOENT) 755 gothid = 0; 756 else if (error != 0) 757 goto done_free; 758 else { 759 gothid = 1; 760 ch_flags |= PR_HOST; 761 pr_flags |= PR_HOST; 762 } 763 764 #ifdef INET 765 error = vfs_getopt(opts, "ip4.addr", &op, &ip4s); 766 if (error == ENOENT) 767 ip4s = 0; 768 else if (error != 0) 769 goto done_free; 770 else if (ip4s & (sizeof(*ip4) - 1)) { 771 error = EINVAL; 772 goto done_free; 773 } else { 774 ch_flags |= PR_IP4_USER; 775 pr_flags |= PR_IP4_USER; 776 if (ip4s > 0) { 777 ip4s /= sizeof(*ip4); 778 if (ip4s > jail_max_af_ips) { 779 error = EINVAL; 780 vfs_opterror(opts, "too many IPv4 addresses"); 781 goto done_errmsg; 782 } 783 ip4 = malloc(ip4s * sizeof(*ip4), M_PRISON, M_WAITOK); 784 bcopy(op, ip4, ip4s * sizeof(*ip4)); 785 /* 786 * IP addresses are all sorted but ip[0] to preserve 787 * the primary IP address as given from userland. 788 * This special IP is used for unbound outgoing 789 * connections as well for "loopback" traffic in case 790 * source address selection cannot find any more fitting 791 * address to connect from. 792 */ 793 if (ip4s > 1) 794 qsort(ip4 + 1, ip4s - 1, sizeof(*ip4), 795 prison_qcmp_v4); 796 /* 797 * Check for duplicate addresses and do some simple 798 * zero and broadcast checks. If users give other bogus 799 * addresses it is their problem. 800 * 801 * We do not have to care about byte order for these 802 * checks so we will do them in NBO. 803 */ 804 for (ii = 0; ii < ip4s; ii++) { 805 if (ip4[ii].s_addr == INADDR_ANY || 806 ip4[ii].s_addr == INADDR_BROADCAST) { 807 error = EINVAL; 808 goto done_free; 809 } 810 if ((ii+1) < ip4s && 811 (ip4[0].s_addr == ip4[ii+1].s_addr || 812 ip4[ii].s_addr == ip4[ii+1].s_addr)) { 813 error = EINVAL; 814 goto done_free; 815 } 816 } 817 } 818 } 819 #endif 820 821 #ifdef INET6 822 error = vfs_getopt(opts, "ip6.addr", &op, &ip6s); 823 if (error == ENOENT) 824 ip6s = 0; 825 else if (error != 0) 826 goto done_free; 827 else if (ip6s & (sizeof(*ip6) - 1)) { 828 error = EINVAL; 829 goto done_free; 830 } else { 831 ch_flags |= PR_IP6_USER; 832 pr_flags |= PR_IP6_USER; 833 if (ip6s > 0) { 834 ip6s /= sizeof(*ip6); 835 if (ip6s > jail_max_af_ips) { 836 error = EINVAL; 837 vfs_opterror(opts, "too many IPv6 addresses"); 838 goto done_errmsg; 839 } 840 ip6 = malloc(ip6s * sizeof(*ip6), M_PRISON, M_WAITOK); 841 bcopy(op, ip6, ip6s * sizeof(*ip6)); 842 if (ip6s > 1) 843 qsort(ip6 + 1, ip6s - 1, sizeof(*ip6), 844 prison_qcmp_v6); 845 for (ii = 0; ii < ip6s; ii++) { 846 if (IN6_IS_ADDR_UNSPECIFIED(&ip6[ii])) { 847 error = EINVAL; 848 goto done_free; 849 } 850 if ((ii+1) < ip6s && 851 (IN6_ARE_ADDR_EQUAL(&ip6[0], &ip6[ii+1]) || 852 IN6_ARE_ADDR_EQUAL(&ip6[ii], &ip6[ii+1]))) 853 { 854 error = EINVAL; 855 goto done_free; 856 } 857 } 858 } 859 } 860 #endif 861 862 #if defined(VIMAGE) && (defined(INET) || defined(INET6)) 863 if ((ch_flags & PR_VNET) && (ch_flags & (PR_IP4_USER | PR_IP6_USER))) { 864 error = EINVAL; 865 vfs_opterror(opts, 866 "vnet jails cannot have IP address restrictions"); 867 goto done_errmsg; 868 } 869 #endif 870 871 error = vfs_getopt(opts, "osrelease", (void **)&osrelstr, &len); 872 if (error == ENOENT) 873 osrelstr = NULL; 874 else if (error != 0) 875 goto done_free; 876 else { 877 if (flags & JAIL_UPDATE) { 878 error = EINVAL; 879 vfs_opterror(opts, 880 "osrelease cannot be changed after creation"); 881 goto done_errmsg; 882 } 883 if (len == 0 || len >= OSRELEASELEN) { 884 error = EINVAL; 885 vfs_opterror(opts, 886 "osrelease string must be 1-%d bytes long", 887 OSRELEASELEN - 1); 888 goto done_errmsg; 889 } 890 } 891 892 error = vfs_copyopt(opts, "osreldate", &osreldt, sizeof(osreldt)); 893 if (error == ENOENT) 894 osreldt = 0; 895 else if (error != 0) 896 goto done_free; 897 else { 898 if (flags & JAIL_UPDATE) { 899 error = EINVAL; 900 vfs_opterror(opts, 901 "osreldate cannot be changed after creation"); 902 goto done_errmsg; 903 } 904 if (osreldt == 0) { 905 error = EINVAL; 906 vfs_opterror(opts, "osreldate cannot be 0"); 907 goto done_errmsg; 908 } 909 } 910 911 fullpath_disabled = 0; 912 root = NULL; 913 error = vfs_getopt(opts, "path", (void **)&path, &len); 914 if (error == ENOENT) 915 path = NULL; 916 else if (error != 0) 917 goto done_free; 918 else { 919 if (flags & JAIL_UPDATE) { 920 error = EINVAL; 921 vfs_opterror(opts, 922 "path cannot be changed after creation"); 923 goto done_errmsg; 924 } 925 if (len == 0 || path[len - 1] != '\0') { 926 error = EINVAL; 927 goto done_free; 928 } 929 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, 930 path, td); 931 error = namei(&nd); 932 if (error) 933 goto done_free; 934 root = nd.ni_vp; 935 NDFREE(&nd, NDF_ONLY_PNBUF); 936 g_path = malloc(MAXPATHLEN, M_TEMP, M_WAITOK); 937 strlcpy(g_path, path, MAXPATHLEN); 938 error = vn_path_to_global_path(td, root, g_path, MAXPATHLEN); 939 if (error == 0) 940 path = g_path; 941 else if (error == ENODEV) { 942 /* proceed if sysctl debug.disablefullpath == 1 */ 943 fullpath_disabled = 1; 944 if (len < 2 || (len == 2 && path[0] == '/')) 945 path = NULL; 946 } else { 947 /* exit on other errors */ 948 goto done_free; 949 } 950 if (root->v_type != VDIR) { 951 error = ENOTDIR; 952 vput(root); 953 goto done_free; 954 } 955 VOP_UNLOCK(root, 0); 956 if (fullpath_disabled) { 957 /* Leave room for a real-root full pathname. */ 958 if (len + (path[0] == '/' && strcmp(mypr->pr_path, "/") 959 ? strlen(mypr->pr_path) : 0) > MAXPATHLEN) { 960 error = ENAMETOOLONG; 961 vrele(root); 962 goto done_free; 963 } 964 } 965 } 966 967 /* 968 * Find the specified jail, or at least its parent. 969 * This abuses the file error codes ENOENT and EEXIST. 970 */ 971 pr = NULL; 972 ppr = mypr; 973 if (cuflags == JAIL_CREATE && jid == 0 && name != NULL) { 974 namelc = strrchr(name, '.'); 975 jid = strtoul(namelc != NULL ? namelc + 1 : name, &p, 10); 976 if (*p != '\0') 977 jid = 0; 978 } 979 sx_xlock(&allprison_lock); 980 if (jid != 0) { 981 /* 982 * See if a requested jid already exists. There is an 983 * information leak here if the jid exists but is not within 984 * the caller's jail hierarchy. Jail creators will get EEXIST 985 * even though they cannot see the jail, and CREATE | UPDATE 986 * will return ENOENT which is not normally a valid error. 987 */ 988 if (jid < 0) { 989 error = EINVAL; 990 vfs_opterror(opts, "negative jid"); 991 goto done_unlock_list; 992 } 993 pr = prison_find(jid); 994 if (pr != NULL) { 995 ppr = pr->pr_parent; 996 /* Create: jid must not exist. */ 997 if (cuflags == JAIL_CREATE) { 998 mtx_unlock(&pr->pr_mtx); 999 error = EEXIST; 1000 vfs_opterror(opts, "jail %d already exists", 1001 jid); 1002 goto done_unlock_list; 1003 } 1004 if (!prison_ischild(mypr, pr)) { 1005 mtx_unlock(&pr->pr_mtx); 1006 pr = NULL; 1007 } else if (pr->pr_uref == 0) { 1008 if (!(flags & JAIL_DYING)) { 1009 mtx_unlock(&pr->pr_mtx); 1010 error = ENOENT; 1011 vfs_opterror(opts, "jail %d is dying", 1012 jid); 1013 goto done_unlock_list; 1014 } else if ((flags & JAIL_ATTACH) || 1015 (pr_flags & PR_PERSIST)) { 1016 /* 1017 * A dying jail might be resurrected 1018 * (via attach or persist), but first 1019 * it must determine if another jail 1020 * has claimed its name. Accomplish 1021 * this by implicitly re-setting the 1022 * name. 1023 */ 1024 if (name == NULL) 1025 name = prison_name(mypr, pr); 1026 } 1027 } 1028 } 1029 if (pr == NULL) { 1030 /* Update: jid must exist. */ 1031 if (cuflags == JAIL_UPDATE) { 1032 error = ENOENT; 1033 vfs_opterror(opts, "jail %d not found", jid); 1034 goto done_unlock_list; 1035 } 1036 } 1037 } 1038 /* 1039 * If the caller provided a name, look for a jail by that name. 1040 * This has different semantics for creates and updates keyed by jid 1041 * (where the name must not already exist in a different jail), 1042 * and updates keyed by the name itself (where the name must exist 1043 * because that is the jail being updated). 1044 */ 1045 namelc = NULL; 1046 if (name != NULL) { 1047 namelc = strrchr(name, '.'); 1048 if (namelc == NULL) 1049 namelc = name; 1050 else { 1051 /* 1052 * This is a hierarchical name. Split it into the 1053 * parent and child names, and make sure the parent 1054 * exists or matches an already found jail. 1055 */ 1056 if (pr != NULL) { 1057 if (strncmp(name, ppr->pr_name, namelc - name) 1058 || ppr->pr_name[namelc - name] != '\0') { 1059 mtx_unlock(&pr->pr_mtx); 1060 error = EINVAL; 1061 vfs_opterror(opts, 1062 "cannot change jail's parent"); 1063 goto done_unlock_list; 1064 } 1065 } else { 1066 *namelc = '\0'; 1067 ppr = prison_find_name(mypr, name); 1068 if (ppr == NULL) { 1069 error = ENOENT; 1070 vfs_opterror(opts, 1071 "jail \"%s\" not found", name); 1072 goto done_unlock_list; 1073 } 1074 mtx_unlock(&ppr->pr_mtx); 1075 *namelc = '.'; 1076 } 1077 namelc++; 1078 } 1079 if (namelc[0] != '\0') { 1080 pnamelen = 1081 (ppr == &prison0) ? 0 : strlen(ppr->pr_name) + 1; 1082 name_again: 1083 deadpr = NULL; 1084 FOREACH_PRISON_CHILD(ppr, tpr) { 1085 if (tpr != pr && tpr->pr_ref > 0 && 1086 !strcmp(tpr->pr_name + pnamelen, namelc)) { 1087 if (pr == NULL && 1088 cuflags != JAIL_CREATE) { 1089 mtx_lock(&tpr->pr_mtx); 1090 if (tpr->pr_ref > 0) { 1091 /* 1092 * Use this jail 1093 * for updates. 1094 */ 1095 if (tpr->pr_uref > 0) { 1096 pr = tpr; 1097 break; 1098 } 1099 deadpr = tpr; 1100 } 1101 mtx_unlock(&tpr->pr_mtx); 1102 } else if (tpr->pr_uref > 0) { 1103 /* 1104 * Create, or update(jid): 1105 * name must not exist in an 1106 * active sibling jail. 1107 */ 1108 error = EEXIST; 1109 if (pr != NULL) 1110 mtx_unlock(&pr->pr_mtx); 1111 vfs_opterror(opts, 1112 "jail \"%s\" already exists", 1113 name); 1114 goto done_unlock_list; 1115 } 1116 } 1117 } 1118 /* If no active jail is found, use a dying one. */ 1119 if (deadpr != NULL && pr == NULL) { 1120 if (flags & JAIL_DYING) { 1121 mtx_lock(&deadpr->pr_mtx); 1122 if (deadpr->pr_ref == 0) { 1123 mtx_unlock(&deadpr->pr_mtx); 1124 goto name_again; 1125 } 1126 pr = deadpr; 1127 } else if (cuflags == JAIL_UPDATE) { 1128 error = ENOENT; 1129 vfs_opterror(opts, 1130 "jail \"%s\" is dying", name); 1131 goto done_unlock_list; 1132 } 1133 } 1134 /* Update: name must exist if no jid. */ 1135 else if (cuflags == JAIL_UPDATE && pr == NULL) { 1136 error = ENOENT; 1137 vfs_opterror(opts, "jail \"%s\" not found", 1138 name); 1139 goto done_unlock_list; 1140 } 1141 } 1142 } 1143 /* Update: must provide a jid or name. */ 1144 else if (cuflags == JAIL_UPDATE && pr == NULL) { 1145 error = ENOENT; 1146 vfs_opterror(opts, "update specified no jail"); 1147 goto done_unlock_list; 1148 } 1149 1150 /* If there's no prison to update, create a new one and link it in. */ 1151 if (pr == NULL) { 1152 for (tpr = mypr; tpr != NULL; tpr = tpr->pr_parent) 1153 if (tpr->pr_childcount >= tpr->pr_childmax) { 1154 error = EPERM; 1155 vfs_opterror(opts, "prison limit exceeded"); 1156 goto done_unlock_list; 1157 } 1158 created = 1; 1159 mtx_lock(&ppr->pr_mtx); 1160 if (ppr->pr_ref == 0) { 1161 mtx_unlock(&ppr->pr_mtx); 1162 error = ENOENT; 1163 vfs_opterror(opts, "jail \"%s\" not found", 1164 prison_name(mypr, ppr)); 1165 goto done_unlock_list; 1166 } 1167 ppr->pr_ref++; 1168 ppr->pr_uref++; 1169 mtx_unlock(&ppr->pr_mtx); 1170 pr = malloc(sizeof(*pr), M_PRISON, M_WAITOK | M_ZERO); 1171 if (jid == 0) { 1172 /* Find the next free jid. */ 1173 jid = lastprid + 1; 1174 findnext: 1175 if (jid == JAIL_MAX) 1176 jid = 1; 1177 TAILQ_FOREACH(tpr, &allprison, pr_list) { 1178 if (tpr->pr_id < jid) 1179 continue; 1180 if (tpr->pr_id > jid || tpr->pr_ref == 0) { 1181 TAILQ_INSERT_BEFORE(tpr, pr, pr_list); 1182 break; 1183 } 1184 if (jid == lastprid) { 1185 error = EAGAIN; 1186 vfs_opterror(opts, 1187 "no available jail IDs"); 1188 free(pr, M_PRISON); 1189 prison_deref(ppr, PD_DEREF | 1190 PD_DEUREF | PD_LIST_XLOCKED); 1191 goto done_releroot; 1192 } 1193 jid++; 1194 goto findnext; 1195 } 1196 lastprid = jid; 1197 } else { 1198 /* 1199 * The jail already has a jid (that did not yet exist), 1200 * so just find where to insert it. 1201 */ 1202 TAILQ_FOREACH(tpr, &allprison, pr_list) 1203 if (tpr->pr_id >= jid) { 1204 TAILQ_INSERT_BEFORE(tpr, pr, pr_list); 1205 break; 1206 } 1207 } 1208 if (tpr == NULL) 1209 TAILQ_INSERT_TAIL(&allprison, pr, pr_list); 1210 LIST_INSERT_HEAD(&ppr->pr_children, pr, pr_sibling); 1211 for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent) 1212 tpr->pr_childcount++; 1213 1214 pr->pr_parent = ppr; 1215 pr->pr_id = jid; 1216 1217 /* Set some default values, and inherit some from the parent. */ 1218 if (namelc == NULL) 1219 namelc = ""; 1220 if (path == NULL) { 1221 path = "/"; 1222 root = mypr->pr_root; 1223 vref(root); 1224 } 1225 strlcpy(pr->pr_hostuuid, DEFAULT_HOSTUUID, HOSTUUIDLEN); 1226 pr->pr_flags |= PR_HOST; 1227 #if defined(INET) || defined(INET6) 1228 #ifdef VIMAGE 1229 if (!(pr_flags & PR_VNET)) 1230 #endif 1231 { 1232 #ifdef INET 1233 if (!(ch_flags & PR_IP4_USER)) 1234 pr->pr_flags |= PR_IP4 | PR_IP4_USER; 1235 else if (!(pr_flags & PR_IP4_USER)) { 1236 pr->pr_flags |= ppr->pr_flags & PR_IP4; 1237 if (ppr->pr_ip4 != NULL) { 1238 pr->pr_ip4s = ppr->pr_ip4s; 1239 pr->pr_ip4 = malloc(pr->pr_ip4s * 1240 sizeof(struct in_addr), M_PRISON, 1241 M_WAITOK); 1242 bcopy(ppr->pr_ip4, pr->pr_ip4, 1243 pr->pr_ip4s * sizeof(*pr->pr_ip4)); 1244 } 1245 } 1246 #endif 1247 #ifdef INET6 1248 if (!(ch_flags & PR_IP6_USER)) 1249 pr->pr_flags |= PR_IP6 | PR_IP6_USER; 1250 else if (!(pr_flags & PR_IP6_USER)) { 1251 pr->pr_flags |= ppr->pr_flags & PR_IP6; 1252 if (ppr->pr_ip6 != NULL) { 1253 pr->pr_ip6s = ppr->pr_ip6s; 1254 pr->pr_ip6 = malloc(pr->pr_ip6s * 1255 sizeof(struct in6_addr), M_PRISON, 1256 M_WAITOK); 1257 bcopy(ppr->pr_ip6, pr->pr_ip6, 1258 pr->pr_ip6s * sizeof(*pr->pr_ip6)); 1259 } 1260 } 1261 #endif 1262 } 1263 #endif 1264 /* Source address selection is always on by default. */ 1265 pr->pr_flags |= _PR_IP_SADDRSEL; 1266 1267 pr->pr_securelevel = ppr->pr_securelevel; 1268 pr->pr_allow = JAIL_DEFAULT_ALLOW & ppr->pr_allow; 1269 pr->pr_enforce_statfs = jail_default_enforce_statfs; 1270 pr->pr_devfs_rsnum = ppr->pr_devfs_rsnum; 1271 1272 pr->pr_osreldate = osreldt ? osreldt : ppr->pr_osreldate; 1273 if (osrelstr == NULL) 1274 strcpy(pr->pr_osrelease, ppr->pr_osrelease); 1275 else 1276 strcpy(pr->pr_osrelease, osrelstr); 1277 1278 LIST_INIT(&pr->pr_children); 1279 mtx_init(&pr->pr_mtx, "jail mutex", NULL, MTX_DEF | MTX_DUPOK); 1280 TASK_INIT(&pr->pr_task, 0, prison_complete, pr); 1281 1282 #ifdef VIMAGE 1283 /* Allocate a new vnet if specified. */ 1284 pr->pr_vnet = (pr_flags & PR_VNET) 1285 ? vnet_alloc() : ppr->pr_vnet; 1286 #endif 1287 /* 1288 * Allocate a dedicated cpuset for each jail. 1289 * Unlike other initial settings, this may return an erorr. 1290 */ 1291 error = cpuset_create_root(ppr, &pr->pr_cpuset); 1292 if (error) { 1293 prison_deref(pr, PD_LIST_XLOCKED); 1294 goto done_releroot; 1295 } 1296 1297 mtx_lock(&pr->pr_mtx); 1298 /* 1299 * New prisons do not yet have a reference, because we do not 1300 * want others to see the incomplete prison once the 1301 * allprison_lock is downgraded. 1302 */ 1303 } else { 1304 created = 0; 1305 /* 1306 * Grab a reference for existing prisons, to ensure they 1307 * continue to exist for the duration of the call. 1308 */ 1309 pr->pr_ref++; 1310 #if defined(VIMAGE) && (defined(INET) || defined(INET6)) 1311 if ((pr->pr_flags & PR_VNET) && 1312 (ch_flags & (PR_IP4_USER | PR_IP6_USER))) { 1313 error = EINVAL; 1314 vfs_opterror(opts, 1315 "vnet jails cannot have IP address restrictions"); 1316 goto done_deref_locked; 1317 } 1318 #endif 1319 #ifdef INET 1320 if (PR_IP4_USER & ch_flags & (pr_flags ^ pr->pr_flags)) { 1321 error = EINVAL; 1322 vfs_opterror(opts, 1323 "ip4 cannot be changed after creation"); 1324 goto done_deref_locked; 1325 } 1326 #endif 1327 #ifdef INET6 1328 if (PR_IP6_USER & ch_flags & (pr_flags ^ pr->pr_flags)) { 1329 error = EINVAL; 1330 vfs_opterror(opts, 1331 "ip6 cannot be changed after creation"); 1332 goto done_deref_locked; 1333 } 1334 #endif 1335 } 1336 1337 /* Do final error checking before setting anything. */ 1338 if (gotslevel) { 1339 if (slevel < ppr->pr_securelevel) { 1340 error = EPERM; 1341 goto done_deref_locked; 1342 } 1343 } 1344 if (gotchildmax) { 1345 if (childmax >= ppr->pr_childmax) { 1346 error = EPERM; 1347 goto done_deref_locked; 1348 } 1349 } 1350 if (gotenforce) { 1351 if (enforce < ppr->pr_enforce_statfs) { 1352 error = EPERM; 1353 goto done_deref_locked; 1354 } 1355 } 1356 if (gotrsnum) { 1357 /* 1358 * devfs_rsnum is a uint16_t 1359 */ 1360 if (rsnum < 0 || rsnum > 65535) { 1361 error = EINVAL; 1362 goto done_deref_locked; 1363 } 1364 /* 1365 * Nested jails always inherit parent's devfs ruleset 1366 */ 1367 if (jailed(td->td_ucred)) { 1368 if (rsnum > 0 && rsnum != ppr->pr_devfs_rsnum) { 1369 error = EPERM; 1370 goto done_deref_locked; 1371 } else 1372 rsnum = ppr->pr_devfs_rsnum; 1373 } 1374 } 1375 #ifdef INET 1376 if (ip4s > 0) { 1377 if (ppr->pr_flags & PR_IP4) { 1378 /* 1379 * Make sure the new set of IP addresses is a 1380 * subset of the parent's list. Don't worry 1381 * about the parent being unlocked, as any 1382 * setting is done with allprison_lock held. 1383 */ 1384 for (ij = 0; ij < ppr->pr_ip4s; ij++) 1385 if (ip4[0].s_addr == ppr->pr_ip4[ij].s_addr) 1386 break; 1387 if (ij == ppr->pr_ip4s) { 1388 error = EPERM; 1389 goto done_deref_locked; 1390 } 1391 if (ip4s > 1) { 1392 for (ii = ij = 1; ii < ip4s; ii++) { 1393 if (ip4[ii].s_addr == 1394 ppr->pr_ip4[0].s_addr) 1395 continue; 1396 for (; ij < ppr->pr_ip4s; ij++) 1397 if (ip4[ii].s_addr == 1398 ppr->pr_ip4[ij].s_addr) 1399 break; 1400 if (ij == ppr->pr_ip4s) 1401 break; 1402 } 1403 if (ij == ppr->pr_ip4s) { 1404 error = EPERM; 1405 goto done_deref_locked; 1406 } 1407 } 1408 } 1409 /* 1410 * Check for conflicting IP addresses. We permit them 1411 * if there is no more than one IP on each jail. If 1412 * there is a duplicate on a jail with more than one 1413 * IP stop checking and return error. 1414 */ 1415 tppr = ppr; 1416 #ifdef VIMAGE 1417 for (; tppr != &prison0; tppr = tppr->pr_parent) 1418 if (tppr->pr_flags & PR_VNET) 1419 break; 1420 #endif 1421 FOREACH_PRISON_DESCENDANT(tppr, tpr, descend) { 1422 if (tpr == pr || 1423 #ifdef VIMAGE 1424 (tpr != tppr && (tpr->pr_flags & PR_VNET)) || 1425 #endif 1426 tpr->pr_uref == 0) { 1427 descend = 0; 1428 continue; 1429 } 1430 if (!(tpr->pr_flags & PR_IP4_USER)) 1431 continue; 1432 descend = 0; 1433 if (tpr->pr_ip4 == NULL || 1434 (ip4s == 1 && tpr->pr_ip4s == 1)) 1435 continue; 1436 for (ii = 0; ii < ip4s; ii++) { 1437 if (prison_check_ip4_locked(tpr, &ip4[ii]) == 1438 0) { 1439 error = EADDRINUSE; 1440 vfs_opterror(opts, 1441 "IPv4 addresses clash"); 1442 goto done_deref_locked; 1443 } 1444 } 1445 } 1446 } 1447 #endif 1448 #ifdef INET6 1449 if (ip6s > 0) { 1450 if (ppr->pr_flags & PR_IP6) { 1451 /* 1452 * Make sure the new set of IP addresses is a 1453 * subset of the parent's list. 1454 */ 1455 for (ij = 0; ij < ppr->pr_ip6s; ij++) 1456 if (IN6_ARE_ADDR_EQUAL(&ip6[0], 1457 &ppr->pr_ip6[ij])) 1458 break; 1459 if (ij == ppr->pr_ip6s) { 1460 error = EPERM; 1461 goto done_deref_locked; 1462 } 1463 if (ip6s > 1) { 1464 for (ii = ij = 1; ii < ip6s; ii++) { 1465 if (IN6_ARE_ADDR_EQUAL(&ip6[ii], 1466 &ppr->pr_ip6[0])) 1467 continue; 1468 for (; ij < ppr->pr_ip6s; ij++) 1469 if (IN6_ARE_ADDR_EQUAL( 1470 &ip6[ii], &ppr->pr_ip6[ij])) 1471 break; 1472 if (ij == ppr->pr_ip6s) 1473 break; 1474 } 1475 if (ij == ppr->pr_ip6s) { 1476 error = EPERM; 1477 goto done_deref_locked; 1478 } 1479 } 1480 } 1481 /* Check for conflicting IP addresses. */ 1482 tppr = ppr; 1483 #ifdef VIMAGE 1484 for (; tppr != &prison0; tppr = tppr->pr_parent) 1485 if (tppr->pr_flags & PR_VNET) 1486 break; 1487 #endif 1488 FOREACH_PRISON_DESCENDANT(tppr, tpr, descend) { 1489 if (tpr == pr || 1490 #ifdef VIMAGE 1491 (tpr != tppr && (tpr->pr_flags & PR_VNET)) || 1492 #endif 1493 tpr->pr_uref == 0) { 1494 descend = 0; 1495 continue; 1496 } 1497 if (!(tpr->pr_flags & PR_IP6_USER)) 1498 continue; 1499 descend = 0; 1500 if (tpr->pr_ip6 == NULL || 1501 (ip6s == 1 && tpr->pr_ip6s == 1)) 1502 continue; 1503 for (ii = 0; ii < ip6s; ii++) { 1504 if (prison_check_ip6_locked(tpr, &ip6[ii]) == 1505 0) { 1506 error = EADDRINUSE; 1507 vfs_opterror(opts, 1508 "IPv6 addresses clash"); 1509 goto done_deref_locked; 1510 } 1511 } 1512 } 1513 } 1514 #endif 1515 onamelen = namelen = 0; 1516 if (namelc != NULL) { 1517 /* Give a default name of the jid. Also allow the name to be 1518 * explicitly the jid - but not any other number, and only in 1519 * normal form (no leading zero/etc). 1520 */ 1521 if (namelc[0] == '\0') 1522 snprintf(namelc = numbuf, sizeof(numbuf), "%d", jid); 1523 else if ((strtoul(namelc, &p, 10) != jid || 1524 namelc[0] < '1' || namelc[0] > '9') && *p == '\0') { 1525 error = EINVAL; 1526 vfs_opterror(opts, 1527 "name cannot be numeric (unless it is the jid)"); 1528 goto done_deref_locked; 1529 } 1530 /* 1531 * Make sure the name isn't too long for the prison or its 1532 * children. 1533 */ 1534 pnamelen = (ppr == &prison0) ? 0 : strlen(ppr->pr_name) + 1; 1535 onamelen = strlen(pr->pr_name + pnamelen); 1536 namelen = strlen(namelc); 1537 if (pnamelen + namelen + 1 > sizeof(pr->pr_name)) { 1538 error = ENAMETOOLONG; 1539 goto done_deref_locked; 1540 } 1541 FOREACH_PRISON_DESCENDANT(pr, tpr, descend) { 1542 if (strlen(tpr->pr_name) + (namelen - onamelen) >= 1543 sizeof(pr->pr_name)) { 1544 error = ENAMETOOLONG; 1545 goto done_deref_locked; 1546 } 1547 } 1548 } 1549 if (pr_allow & ~ppr->pr_allow) { 1550 error = EPERM; 1551 goto done_deref_locked; 1552 } 1553 1554 /* 1555 * Let modules check their parameters. This requires unlocking and 1556 * then re-locking the prison, but this is still a valid state as long 1557 * as allprison_lock remains xlocked. 1558 */ 1559 mtx_unlock(&pr->pr_mtx); 1560 error = osd_jail_call(pr, PR_METHOD_CHECK, opts); 1561 if (error != 0) { 1562 prison_deref(pr, created 1563 ? PD_LIST_XLOCKED 1564 : PD_DEREF | PD_LIST_XLOCKED); 1565 goto done_releroot; 1566 } 1567 mtx_lock(&pr->pr_mtx); 1568 1569 /* At this point, all valid parameters should have been noted. */ 1570 TAILQ_FOREACH(opt, opts, link) { 1571 if (!opt->seen && strcmp(opt->name, "errmsg")) { 1572 error = EINVAL; 1573 vfs_opterror(opts, "unknown parameter: %s", opt->name); 1574 goto done_deref_locked; 1575 } 1576 } 1577 1578 /* Set the parameters of the prison. */ 1579 #ifdef INET 1580 redo_ip4 = 0; 1581 if (pr_flags & PR_IP4_USER) { 1582 pr->pr_flags |= PR_IP4; 1583 free(pr->pr_ip4, M_PRISON); 1584 pr->pr_ip4s = ip4s; 1585 pr->pr_ip4 = ip4; 1586 ip4 = NULL; 1587 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) { 1588 #ifdef VIMAGE 1589 if (tpr->pr_flags & PR_VNET) { 1590 descend = 0; 1591 continue; 1592 } 1593 #endif 1594 if (prison_restrict_ip4(tpr, NULL)) { 1595 redo_ip4 = 1; 1596 descend = 0; 1597 } 1598 } 1599 } 1600 #endif 1601 #ifdef INET6 1602 redo_ip6 = 0; 1603 if (pr_flags & PR_IP6_USER) { 1604 pr->pr_flags |= PR_IP6; 1605 free(pr->pr_ip6, M_PRISON); 1606 pr->pr_ip6s = ip6s; 1607 pr->pr_ip6 = ip6; 1608 ip6 = NULL; 1609 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) { 1610 #ifdef VIMAGE 1611 if (tpr->pr_flags & PR_VNET) { 1612 descend = 0; 1613 continue; 1614 } 1615 #endif 1616 if (prison_restrict_ip6(tpr, NULL)) { 1617 redo_ip6 = 1; 1618 descend = 0; 1619 } 1620 } 1621 } 1622 #endif 1623 if (gotslevel) { 1624 pr->pr_securelevel = slevel; 1625 /* Set all child jails to be at least this level. */ 1626 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) 1627 if (tpr->pr_securelevel < slevel) 1628 tpr->pr_securelevel = slevel; 1629 } 1630 if (gotchildmax) { 1631 pr->pr_childmax = childmax; 1632 /* Set all child jails to under this limit. */ 1633 FOREACH_PRISON_DESCENDANT_LOCKED_LEVEL(pr, tpr, descend, level) 1634 if (tpr->pr_childmax > childmax - level) 1635 tpr->pr_childmax = childmax > level 1636 ? childmax - level : 0; 1637 } 1638 if (gotenforce) { 1639 pr->pr_enforce_statfs = enforce; 1640 /* Pass this restriction on to the children. */ 1641 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) 1642 if (tpr->pr_enforce_statfs < enforce) 1643 tpr->pr_enforce_statfs = enforce; 1644 } 1645 if (gotrsnum) { 1646 pr->pr_devfs_rsnum = rsnum; 1647 /* Pass this restriction on to the children. */ 1648 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) 1649 tpr->pr_devfs_rsnum = rsnum; 1650 } 1651 if (namelc != NULL) { 1652 if (ppr == &prison0) 1653 strlcpy(pr->pr_name, namelc, sizeof(pr->pr_name)); 1654 else 1655 snprintf(pr->pr_name, sizeof(pr->pr_name), "%s.%s", 1656 ppr->pr_name, namelc); 1657 /* Change this component of child names. */ 1658 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) { 1659 bcopy(tpr->pr_name + onamelen, tpr->pr_name + namelen, 1660 strlen(tpr->pr_name + onamelen) + 1); 1661 bcopy(pr->pr_name, tpr->pr_name, namelen); 1662 } 1663 } 1664 if (path != NULL) { 1665 /* Try to keep a real-rooted full pathname. */ 1666 if (fullpath_disabled && path[0] == '/' && 1667 strcmp(mypr->pr_path, "/")) 1668 snprintf(pr->pr_path, sizeof(pr->pr_path), "%s%s", 1669 mypr->pr_path, path); 1670 else 1671 strlcpy(pr->pr_path, path, sizeof(pr->pr_path)); 1672 pr->pr_root = root; 1673 } 1674 if (PR_HOST & ch_flags & ~pr_flags) { 1675 if (pr->pr_flags & PR_HOST) { 1676 /* 1677 * Copy the parent's host info. As with pr_ip4 above, 1678 * the lack of a lock on the parent is not a problem; 1679 * it is always set with allprison_lock at least 1680 * shared, and is held exclusively here. 1681 */ 1682 strlcpy(pr->pr_hostname, pr->pr_parent->pr_hostname, 1683 sizeof(pr->pr_hostname)); 1684 strlcpy(pr->pr_domainname, pr->pr_parent->pr_domainname, 1685 sizeof(pr->pr_domainname)); 1686 strlcpy(pr->pr_hostuuid, pr->pr_parent->pr_hostuuid, 1687 sizeof(pr->pr_hostuuid)); 1688 pr->pr_hostid = pr->pr_parent->pr_hostid; 1689 } 1690 } else if (host != NULL || domain != NULL || uuid != NULL || gothid) { 1691 /* Set this prison, and any descendants without PR_HOST. */ 1692 if (host != NULL) 1693 strlcpy(pr->pr_hostname, host, sizeof(pr->pr_hostname)); 1694 if (domain != NULL) 1695 strlcpy(pr->pr_domainname, domain, 1696 sizeof(pr->pr_domainname)); 1697 if (uuid != NULL) 1698 strlcpy(pr->pr_hostuuid, uuid, sizeof(pr->pr_hostuuid)); 1699 if (gothid) 1700 pr->pr_hostid = hid; 1701 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) { 1702 if (tpr->pr_flags & PR_HOST) 1703 descend = 0; 1704 else { 1705 if (host != NULL) 1706 strlcpy(tpr->pr_hostname, 1707 pr->pr_hostname, 1708 sizeof(tpr->pr_hostname)); 1709 if (domain != NULL) 1710 strlcpy(tpr->pr_domainname, 1711 pr->pr_domainname, 1712 sizeof(tpr->pr_domainname)); 1713 if (uuid != NULL) 1714 strlcpy(tpr->pr_hostuuid, 1715 pr->pr_hostuuid, 1716 sizeof(tpr->pr_hostuuid)); 1717 if (gothid) 1718 tpr->pr_hostid = hid; 1719 } 1720 } 1721 } 1722 if ((tallow = ch_allow & ~pr_allow)) { 1723 /* Clear allow bits in all children. */ 1724 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) 1725 tpr->pr_allow &= ~tallow; 1726 } 1727 pr->pr_allow = (pr->pr_allow & ~ch_allow) | pr_allow; 1728 /* 1729 * Persistent prisons get an extra reference, and prisons losing their 1730 * persist flag lose that reference. Only do this for existing prisons 1731 * for now, so new ones will remain unseen until after the module 1732 * handlers have completed. 1733 */ 1734 born = pr->pr_uref == 0; 1735 if (!created && (ch_flags & PR_PERSIST & (pr_flags ^ pr->pr_flags))) { 1736 if (pr_flags & PR_PERSIST) { 1737 pr->pr_ref++; 1738 pr->pr_uref++; 1739 } else { 1740 pr->pr_ref--; 1741 pr->pr_uref--; 1742 } 1743 } 1744 pr->pr_flags = (pr->pr_flags & ~ch_flags) | pr_flags; 1745 mtx_unlock(&pr->pr_mtx); 1746 1747 #ifdef RACCT 1748 if (racct_enable && created) 1749 prison_racct_attach(pr); 1750 #endif 1751 1752 /* Locks may have prevented a complete restriction of child IP 1753 * addresses. If so, allocate some more memory and try again. 1754 */ 1755 #ifdef INET 1756 while (redo_ip4) { 1757 ip4s = pr->pr_ip4s; 1758 ip4 = malloc(ip4s * sizeof(*ip4), M_PRISON, M_WAITOK); 1759 mtx_lock(&pr->pr_mtx); 1760 redo_ip4 = 0; 1761 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) { 1762 #ifdef VIMAGE 1763 if (tpr->pr_flags & PR_VNET) { 1764 descend = 0; 1765 continue; 1766 } 1767 #endif 1768 if (prison_restrict_ip4(tpr, ip4)) { 1769 if (ip4 != NULL) 1770 ip4 = NULL; 1771 else 1772 redo_ip4 = 1; 1773 } 1774 } 1775 mtx_unlock(&pr->pr_mtx); 1776 } 1777 #endif 1778 #ifdef INET6 1779 while (redo_ip6) { 1780 ip6s = pr->pr_ip6s; 1781 ip6 = malloc(ip6s * sizeof(*ip6), M_PRISON, M_WAITOK); 1782 mtx_lock(&pr->pr_mtx); 1783 redo_ip6 = 0; 1784 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) { 1785 #ifdef VIMAGE 1786 if (tpr->pr_flags & PR_VNET) { 1787 descend = 0; 1788 continue; 1789 } 1790 #endif 1791 if (prison_restrict_ip6(tpr, ip6)) { 1792 if (ip6 != NULL) 1793 ip6 = NULL; 1794 else 1795 redo_ip6 = 1; 1796 } 1797 } 1798 mtx_unlock(&pr->pr_mtx); 1799 } 1800 #endif 1801 1802 /* Let the modules do their work. */ 1803 sx_downgrade(&allprison_lock); 1804 if (born) { 1805 error = osd_jail_call(pr, PR_METHOD_CREATE, opts); 1806 if (error) { 1807 (void)osd_jail_call(pr, PR_METHOD_REMOVE, NULL); 1808 prison_deref(pr, created 1809 ? PD_LIST_SLOCKED 1810 : PD_DEREF | PD_LIST_SLOCKED); 1811 goto done_errmsg; 1812 } 1813 } 1814 error = osd_jail_call(pr, PR_METHOD_SET, opts); 1815 if (error) { 1816 if (born) 1817 (void)osd_jail_call(pr, PR_METHOD_REMOVE, NULL); 1818 prison_deref(pr, created 1819 ? PD_LIST_SLOCKED 1820 : PD_DEREF | PD_LIST_SLOCKED); 1821 goto done_errmsg; 1822 } 1823 1824 /* Attach this process to the prison if requested. */ 1825 if (flags & JAIL_ATTACH) { 1826 mtx_lock(&pr->pr_mtx); 1827 error = do_jail_attach(td, pr); 1828 if (error) { 1829 vfs_opterror(opts, "attach failed"); 1830 if (!created) 1831 prison_deref(pr, PD_DEREF); 1832 goto done_errmsg; 1833 } 1834 } 1835 1836 #ifdef RACCT 1837 if (racct_enable && !created) { 1838 if (!(flags & JAIL_ATTACH)) 1839 sx_sunlock(&allprison_lock); 1840 prison_racct_modify(pr); 1841 if (!(flags & JAIL_ATTACH)) 1842 sx_slock(&allprison_lock); 1843 } 1844 #endif 1845 1846 td->td_retval[0] = pr->pr_id; 1847 1848 /* 1849 * Now that it is all there, drop the temporary reference from existing 1850 * prisons. Or add a reference to newly created persistent prisons 1851 * (which was not done earlier so that the prison would not be publicly 1852 * visible). 1853 */ 1854 if (!created) { 1855 prison_deref(pr, (flags & JAIL_ATTACH) 1856 ? PD_DEREF 1857 : PD_DEREF | PD_LIST_SLOCKED); 1858 } else { 1859 if (pr_flags & PR_PERSIST) { 1860 mtx_lock(&pr->pr_mtx); 1861 pr->pr_ref++; 1862 pr->pr_uref++; 1863 mtx_unlock(&pr->pr_mtx); 1864 } 1865 if (!(flags & JAIL_ATTACH)) 1866 sx_sunlock(&allprison_lock); 1867 } 1868 1869 goto done_free; 1870 1871 done_deref_locked: 1872 prison_deref(pr, created 1873 ? PD_LOCKED | PD_LIST_XLOCKED 1874 : PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED); 1875 goto done_releroot; 1876 done_unlock_list: 1877 sx_xunlock(&allprison_lock); 1878 done_releroot: 1879 if (root != NULL) 1880 vrele(root); 1881 done_errmsg: 1882 if (error) { 1883 if (vfs_getopt(opts, "errmsg", (void **)&errmsg, 1884 &errmsg_len) == 0 && errmsg_len > 0) { 1885 errmsg_pos = 2 * vfs_getopt_pos(opts, "errmsg") + 1; 1886 if (optuio->uio_segflg == UIO_SYSSPACE) 1887 bcopy(errmsg, 1888 optuio->uio_iov[errmsg_pos].iov_base, 1889 errmsg_len); 1890 else 1891 copyout(errmsg, 1892 optuio->uio_iov[errmsg_pos].iov_base, 1893 errmsg_len); 1894 } 1895 } 1896 done_free: 1897 #ifdef INET 1898 free(ip4, M_PRISON); 1899 #endif 1900 #ifdef INET6 1901 free(ip6, M_PRISON); 1902 #endif 1903 if (g_path != NULL) 1904 free(g_path, M_TEMP); 1905 vfs_freeopts(opts); 1906 return (error); 1907 } 1908 1909 1910 /* 1911 * struct jail_get_args { 1912 * struct iovec *iovp; 1913 * unsigned int iovcnt; 1914 * int flags; 1915 * }; 1916 */ 1917 int 1918 sys_jail_get(struct thread *td, struct jail_get_args *uap) 1919 { 1920 struct uio *auio; 1921 int error; 1922 1923 /* Check that we have an even number of iovecs. */ 1924 if (uap->iovcnt & 1) 1925 return (EINVAL); 1926 1927 error = copyinuio(uap->iovp, uap->iovcnt, &auio); 1928 if (error) 1929 return (error); 1930 error = kern_jail_get(td, auio, uap->flags); 1931 if (error == 0) 1932 error = copyout(auio->uio_iov, uap->iovp, 1933 uap->iovcnt * sizeof (struct iovec)); 1934 free(auio, M_IOV); 1935 return (error); 1936 } 1937 1938 int 1939 kern_jail_get(struct thread *td, struct uio *optuio, int flags) 1940 { 1941 struct prison *pr, *mypr; 1942 struct vfsopt *opt; 1943 struct vfsoptlist *opts; 1944 char *errmsg, *name; 1945 int error, errmsg_len, errmsg_pos, fi, i, jid, len, locked, pos; 1946 1947 if (flags & ~JAIL_GET_MASK) 1948 return (EINVAL); 1949 1950 /* Get the parameter list. */ 1951 error = vfs_buildopts(optuio, &opts); 1952 if (error) 1953 return (error); 1954 errmsg_pos = vfs_getopt_pos(opts, "errmsg"); 1955 mypr = td->td_ucred->cr_prison; 1956 1957 /* 1958 * Find the prison specified by one of: lastjid, jid, name. 1959 */ 1960 sx_slock(&allprison_lock); 1961 error = vfs_copyopt(opts, "lastjid", &jid, sizeof(jid)); 1962 if (error == 0) { 1963 TAILQ_FOREACH(pr, &allprison, pr_list) { 1964 if (pr->pr_id > jid && prison_ischild(mypr, pr)) { 1965 mtx_lock(&pr->pr_mtx); 1966 if (pr->pr_ref > 0 && 1967 (pr->pr_uref > 0 || (flags & JAIL_DYING))) 1968 break; 1969 mtx_unlock(&pr->pr_mtx); 1970 } 1971 } 1972 if (pr != NULL) 1973 goto found_prison; 1974 error = ENOENT; 1975 vfs_opterror(opts, "no jail after %d", jid); 1976 goto done_unlock_list; 1977 } else if (error != ENOENT) 1978 goto done_unlock_list; 1979 1980 error = vfs_copyopt(opts, "jid", &jid, sizeof(jid)); 1981 if (error == 0) { 1982 if (jid != 0) { 1983 pr = prison_find_child(mypr, jid); 1984 if (pr != NULL) { 1985 if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) { 1986 mtx_unlock(&pr->pr_mtx); 1987 error = ENOENT; 1988 vfs_opterror(opts, "jail %d is dying", 1989 jid); 1990 goto done_unlock_list; 1991 } 1992 goto found_prison; 1993 } 1994 error = ENOENT; 1995 vfs_opterror(opts, "jail %d not found", jid); 1996 goto done_unlock_list; 1997 } 1998 } else if (error != ENOENT) 1999 goto done_unlock_list; 2000 2001 error = vfs_getopt(opts, "name", (void **)&name, &len); 2002 if (error == 0) { 2003 if (len == 0 || name[len - 1] != '\0') { 2004 error = EINVAL; 2005 goto done_unlock_list; 2006 } 2007 pr = prison_find_name(mypr, name); 2008 if (pr != NULL) { 2009 if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) { 2010 mtx_unlock(&pr->pr_mtx); 2011 error = ENOENT; 2012 vfs_opterror(opts, "jail \"%s\" is dying", 2013 name); 2014 goto done_unlock_list; 2015 } 2016 goto found_prison; 2017 } 2018 error = ENOENT; 2019 vfs_opterror(opts, "jail \"%s\" not found", name); 2020 goto done_unlock_list; 2021 } else if (error != ENOENT) 2022 goto done_unlock_list; 2023 2024 vfs_opterror(opts, "no jail specified"); 2025 error = ENOENT; 2026 goto done_unlock_list; 2027 2028 found_prison: 2029 /* Get the parameters of the prison. */ 2030 pr->pr_ref++; 2031 locked = PD_LOCKED; 2032 td->td_retval[0] = pr->pr_id; 2033 error = vfs_setopt(opts, "jid", &pr->pr_id, sizeof(pr->pr_id)); 2034 if (error != 0 && error != ENOENT) 2035 goto done_deref; 2036 i = (pr->pr_parent == mypr) ? 0 : pr->pr_parent->pr_id; 2037 error = vfs_setopt(opts, "parent", &i, sizeof(i)); 2038 if (error != 0 && error != ENOENT) 2039 goto done_deref; 2040 error = vfs_setopts(opts, "name", prison_name(mypr, pr)); 2041 if (error != 0 && error != ENOENT) 2042 goto done_deref; 2043 error = vfs_setopt(opts, "cpuset.id", &pr->pr_cpuset->cs_id, 2044 sizeof(pr->pr_cpuset->cs_id)); 2045 if (error != 0 && error != ENOENT) 2046 goto done_deref; 2047 error = vfs_setopts(opts, "path", prison_path(mypr, pr)); 2048 if (error != 0 && error != ENOENT) 2049 goto done_deref; 2050 #ifdef INET 2051 error = vfs_setopt_part(opts, "ip4.addr", pr->pr_ip4, 2052 pr->pr_ip4s * sizeof(*pr->pr_ip4)); 2053 if (error != 0 && error != ENOENT) 2054 goto done_deref; 2055 #endif 2056 #ifdef INET6 2057 error = vfs_setopt_part(opts, "ip6.addr", pr->pr_ip6, 2058 pr->pr_ip6s * sizeof(*pr->pr_ip6)); 2059 if (error != 0 && error != ENOENT) 2060 goto done_deref; 2061 #endif 2062 error = vfs_setopt(opts, "securelevel", &pr->pr_securelevel, 2063 sizeof(pr->pr_securelevel)); 2064 if (error != 0 && error != ENOENT) 2065 goto done_deref; 2066 error = vfs_setopt(opts, "children.cur", &pr->pr_childcount, 2067 sizeof(pr->pr_childcount)); 2068 if (error != 0 && error != ENOENT) 2069 goto done_deref; 2070 error = vfs_setopt(opts, "children.max", &pr->pr_childmax, 2071 sizeof(pr->pr_childmax)); 2072 if (error != 0 && error != ENOENT) 2073 goto done_deref; 2074 error = vfs_setopts(opts, "host.hostname", pr->pr_hostname); 2075 if (error != 0 && error != ENOENT) 2076 goto done_deref; 2077 error = vfs_setopts(opts, "host.domainname", pr->pr_domainname); 2078 if (error != 0 && error != ENOENT) 2079 goto done_deref; 2080 error = vfs_setopts(opts, "host.hostuuid", pr->pr_hostuuid); 2081 if (error != 0 && error != ENOENT) 2082 goto done_deref; 2083 #ifdef COMPAT_FREEBSD32 2084 if (SV_PROC_FLAG(td->td_proc, SV_ILP32)) { 2085 uint32_t hid32 = pr->pr_hostid; 2086 2087 error = vfs_setopt(opts, "host.hostid", &hid32, sizeof(hid32)); 2088 } else 2089 #endif 2090 error = vfs_setopt(opts, "host.hostid", &pr->pr_hostid, 2091 sizeof(pr->pr_hostid)); 2092 if (error != 0 && error != ENOENT) 2093 goto done_deref; 2094 error = vfs_setopt(opts, "enforce_statfs", &pr->pr_enforce_statfs, 2095 sizeof(pr->pr_enforce_statfs)); 2096 if (error != 0 && error != ENOENT) 2097 goto done_deref; 2098 error = vfs_setopt(opts, "devfs_ruleset", &pr->pr_devfs_rsnum, 2099 sizeof(pr->pr_devfs_rsnum)); 2100 if (error != 0 && error != ENOENT) 2101 goto done_deref; 2102 for (fi = 0; fi < nitems(pr_flag_names); fi++) { 2103 if (pr_flag_names[fi] == NULL) 2104 continue; 2105 i = (pr->pr_flags & (1 << fi)) ? 1 : 0; 2106 error = vfs_setopt(opts, pr_flag_names[fi], &i, sizeof(i)); 2107 if (error != 0 && error != ENOENT) 2108 goto done_deref; 2109 i = !i; 2110 error = vfs_setopt(opts, pr_flag_nonames[fi], &i, sizeof(i)); 2111 if (error != 0 && error != ENOENT) 2112 goto done_deref; 2113 } 2114 for (fi = 0; fi < nitems(pr_flag_jailsys); fi++) { 2115 i = pr->pr_flags & 2116 (pr_flag_jailsys[fi].disable | pr_flag_jailsys[fi].new); 2117 i = pr_flag_jailsys[fi].disable && 2118 (i == pr_flag_jailsys[fi].disable) ? JAIL_SYS_DISABLE 2119 : (i == pr_flag_jailsys[fi].new) ? JAIL_SYS_NEW 2120 : JAIL_SYS_INHERIT; 2121 error = 2122 vfs_setopt(opts, pr_flag_jailsys[fi].name, &i, sizeof(i)); 2123 if (error != 0 && error != ENOENT) 2124 goto done_deref; 2125 } 2126 for (fi = 0; fi < nitems(pr_allow_names); fi++) { 2127 if (pr_allow_names[fi] == NULL) 2128 continue; 2129 i = (pr->pr_allow & (1 << fi)) ? 1 : 0; 2130 error = vfs_setopt(opts, pr_allow_names[fi], &i, sizeof(i)); 2131 if (error != 0 && error != ENOENT) 2132 goto done_deref; 2133 i = !i; 2134 error = vfs_setopt(opts, pr_allow_nonames[fi], &i, sizeof(i)); 2135 if (error != 0 && error != ENOENT) 2136 goto done_deref; 2137 } 2138 i = (pr->pr_uref == 0); 2139 error = vfs_setopt(opts, "dying", &i, sizeof(i)); 2140 if (error != 0 && error != ENOENT) 2141 goto done_deref; 2142 i = !i; 2143 error = vfs_setopt(opts, "nodying", &i, sizeof(i)); 2144 if (error != 0 && error != ENOENT) 2145 goto done_deref; 2146 error = vfs_setopt(opts, "osreldate", &pr->pr_osreldate, 2147 sizeof(pr->pr_osreldate)); 2148 if (error != 0 && error != ENOENT) 2149 goto done_deref; 2150 error = vfs_setopts(opts, "osrelease", pr->pr_osrelease); 2151 if (error != 0 && error != ENOENT) 2152 goto done_deref; 2153 2154 /* Get the module parameters. */ 2155 mtx_unlock(&pr->pr_mtx); 2156 locked = 0; 2157 error = osd_jail_call(pr, PR_METHOD_GET, opts); 2158 if (error) 2159 goto done_deref; 2160 prison_deref(pr, PD_DEREF | PD_LIST_SLOCKED); 2161 2162 /* By now, all parameters should have been noted. */ 2163 TAILQ_FOREACH(opt, opts, link) { 2164 if (!opt->seen && strcmp(opt->name, "errmsg")) { 2165 error = EINVAL; 2166 vfs_opterror(opts, "unknown parameter: %s", opt->name); 2167 goto done_errmsg; 2168 } 2169 } 2170 2171 /* Write the fetched parameters back to userspace. */ 2172 error = 0; 2173 TAILQ_FOREACH(opt, opts, link) { 2174 if (opt->pos >= 0 && opt->pos != errmsg_pos) { 2175 pos = 2 * opt->pos + 1; 2176 optuio->uio_iov[pos].iov_len = opt->len; 2177 if (opt->value != NULL) { 2178 if (optuio->uio_segflg == UIO_SYSSPACE) { 2179 bcopy(opt->value, 2180 optuio->uio_iov[pos].iov_base, 2181 opt->len); 2182 } else { 2183 error = copyout(opt->value, 2184 optuio->uio_iov[pos].iov_base, 2185 opt->len); 2186 if (error) 2187 break; 2188 } 2189 } 2190 } 2191 } 2192 goto done_errmsg; 2193 2194 done_deref: 2195 prison_deref(pr, locked | PD_DEREF | PD_LIST_SLOCKED); 2196 goto done_errmsg; 2197 2198 done_unlock_list: 2199 sx_sunlock(&allprison_lock); 2200 done_errmsg: 2201 if (error && errmsg_pos >= 0) { 2202 vfs_getopt(opts, "errmsg", (void **)&errmsg, &errmsg_len); 2203 errmsg_pos = 2 * errmsg_pos + 1; 2204 if (errmsg_len > 0) { 2205 if (optuio->uio_segflg == UIO_SYSSPACE) 2206 bcopy(errmsg, 2207 optuio->uio_iov[errmsg_pos].iov_base, 2208 errmsg_len); 2209 else 2210 copyout(errmsg, 2211 optuio->uio_iov[errmsg_pos].iov_base, 2212 errmsg_len); 2213 } 2214 } 2215 vfs_freeopts(opts); 2216 return (error); 2217 } 2218 2219 2220 /* 2221 * struct jail_remove_args { 2222 * int jid; 2223 * }; 2224 */ 2225 int 2226 sys_jail_remove(struct thread *td, struct jail_remove_args *uap) 2227 { 2228 struct prison *pr, *cpr, *lpr, *tpr; 2229 int descend, error; 2230 2231 error = priv_check(td, PRIV_JAIL_REMOVE); 2232 if (error) 2233 return (error); 2234 2235 sx_xlock(&allprison_lock); 2236 pr = prison_find_child(td->td_ucred->cr_prison, uap->jid); 2237 if (pr == NULL) { 2238 sx_xunlock(&allprison_lock); 2239 return (EINVAL); 2240 } 2241 2242 /* Remove all descendants of this prison, then remove this prison. */ 2243 pr->pr_ref++; 2244 if (!LIST_EMPTY(&pr->pr_children)) { 2245 mtx_unlock(&pr->pr_mtx); 2246 lpr = NULL; 2247 FOREACH_PRISON_DESCENDANT(pr, cpr, descend) { 2248 mtx_lock(&cpr->pr_mtx); 2249 if (cpr->pr_ref > 0) { 2250 tpr = cpr; 2251 cpr->pr_ref++; 2252 } else { 2253 /* Already removed - do not do it again. */ 2254 tpr = NULL; 2255 } 2256 mtx_unlock(&cpr->pr_mtx); 2257 if (lpr != NULL) { 2258 mtx_lock(&lpr->pr_mtx); 2259 prison_remove_one(lpr); 2260 sx_xlock(&allprison_lock); 2261 } 2262 lpr = tpr; 2263 } 2264 if (lpr != NULL) { 2265 mtx_lock(&lpr->pr_mtx); 2266 prison_remove_one(lpr); 2267 sx_xlock(&allprison_lock); 2268 } 2269 mtx_lock(&pr->pr_mtx); 2270 } 2271 prison_remove_one(pr); 2272 return (0); 2273 } 2274 2275 static void 2276 prison_remove_one(struct prison *pr) 2277 { 2278 struct proc *p; 2279 int deuref; 2280 2281 /* If the prison was persistent, it is not anymore. */ 2282 deuref = 0; 2283 if (pr->pr_flags & PR_PERSIST) { 2284 pr->pr_ref--; 2285 deuref = PD_DEUREF; 2286 pr->pr_flags &= ~PR_PERSIST; 2287 } 2288 2289 /* 2290 * jail_remove added a reference. If that's the only one, remove 2291 * the prison now. 2292 */ 2293 KASSERT(pr->pr_ref > 0, 2294 ("prison_remove_one removing a dead prison (jid=%d)", pr->pr_id)); 2295 if (pr->pr_ref == 1) { 2296 prison_deref(pr, 2297 deuref | PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED); 2298 return; 2299 } 2300 2301 mtx_unlock(&pr->pr_mtx); 2302 sx_xunlock(&allprison_lock); 2303 /* 2304 * Kill all processes unfortunate enough to be attached to this prison. 2305 */ 2306 sx_slock(&allproc_lock); 2307 LIST_FOREACH(p, &allproc, p_list) { 2308 PROC_LOCK(p); 2309 if (p->p_state != PRS_NEW && p->p_ucred && 2310 p->p_ucred->cr_prison == pr) 2311 kern_psignal(p, SIGKILL); 2312 PROC_UNLOCK(p); 2313 } 2314 sx_sunlock(&allproc_lock); 2315 /* Remove the temporary reference added by jail_remove. */ 2316 prison_deref(pr, deuref | PD_DEREF); 2317 } 2318 2319 2320 /* 2321 * struct jail_attach_args { 2322 * int jid; 2323 * }; 2324 */ 2325 int 2326 sys_jail_attach(struct thread *td, struct jail_attach_args *uap) 2327 { 2328 struct prison *pr; 2329 int error; 2330 2331 error = priv_check(td, PRIV_JAIL_ATTACH); 2332 if (error) 2333 return (error); 2334 2335 /* 2336 * Start with exclusive hold on allprison_lock to ensure that a possible 2337 * PR_METHOD_REMOVE call isn't concurrent with jail_set or jail_remove. 2338 * But then immediately downgrade it since we don't need to stop 2339 * readers. 2340 */ 2341 sx_xlock(&allprison_lock); 2342 sx_downgrade(&allprison_lock); 2343 pr = prison_find_child(td->td_ucred->cr_prison, uap->jid); 2344 if (pr == NULL) { 2345 sx_sunlock(&allprison_lock); 2346 return (EINVAL); 2347 } 2348 2349 /* 2350 * Do not allow a process to attach to a prison that is not 2351 * considered to be "alive". 2352 */ 2353 if (pr->pr_uref == 0) { 2354 mtx_unlock(&pr->pr_mtx); 2355 sx_sunlock(&allprison_lock); 2356 return (EINVAL); 2357 } 2358 2359 return (do_jail_attach(td, pr)); 2360 } 2361 2362 static int 2363 do_jail_attach(struct thread *td, struct prison *pr) 2364 { 2365 struct proc *p; 2366 struct ucred *newcred, *oldcred; 2367 int error; 2368 2369 /* 2370 * XXX: Note that there is a slight race here if two threads 2371 * in the same privileged process attempt to attach to two 2372 * different jails at the same time. It is important for 2373 * user processes not to do this, or they might end up with 2374 * a process root from one prison, but attached to the jail 2375 * of another. 2376 */ 2377 pr->pr_ref++; 2378 pr->pr_uref++; 2379 mtx_unlock(&pr->pr_mtx); 2380 2381 /* Let modules do whatever they need to prepare for attaching. */ 2382 error = osd_jail_call(pr, PR_METHOD_ATTACH, td); 2383 if (error) { 2384 prison_deref(pr, PD_DEREF | PD_DEUREF | PD_LIST_SLOCKED); 2385 return (error); 2386 } 2387 sx_sunlock(&allprison_lock); 2388 2389 /* 2390 * Reparent the newly attached process to this jail. 2391 */ 2392 p = td->td_proc; 2393 error = cpuset_setproc_update_set(p, pr->pr_cpuset); 2394 if (error) 2395 goto e_revert_osd; 2396 2397 vn_lock(pr->pr_root, LK_EXCLUSIVE | LK_RETRY); 2398 if ((error = change_dir(pr->pr_root, td)) != 0) 2399 goto e_unlock; 2400 #ifdef MAC 2401 if ((error = mac_vnode_check_chroot(td->td_ucred, pr->pr_root))) 2402 goto e_unlock; 2403 #endif 2404 VOP_UNLOCK(pr->pr_root, 0); 2405 if ((error = pwd_chroot(td, pr->pr_root))) 2406 goto e_revert_osd; 2407 2408 newcred = crget(); 2409 PROC_LOCK(p); 2410 oldcred = crcopysafe(p, newcred); 2411 newcred->cr_prison = pr; 2412 proc_set_cred(p, newcred); 2413 setsugid(p); 2414 PROC_UNLOCK(p); 2415 #ifdef RACCT 2416 racct_proc_ucred_changed(p, oldcred, newcred); 2417 #endif 2418 prison_deref(oldcred->cr_prison, PD_DEREF | PD_DEUREF); 2419 crfree(oldcred); 2420 return (0); 2421 2422 e_unlock: 2423 VOP_UNLOCK(pr->pr_root, 0); 2424 e_revert_osd: 2425 /* Tell modules this thread is still in its old jail after all. */ 2426 (void)osd_jail_call(td->td_ucred->cr_prison, PR_METHOD_ATTACH, td); 2427 prison_deref(pr, PD_DEREF | PD_DEUREF); 2428 return (error); 2429 } 2430 2431 2432 /* 2433 * Returns a locked prison instance, or NULL on failure. 2434 */ 2435 struct prison * 2436 prison_find(int prid) 2437 { 2438 struct prison *pr; 2439 2440 sx_assert(&allprison_lock, SX_LOCKED); 2441 TAILQ_FOREACH(pr, &allprison, pr_list) { 2442 if (pr->pr_id == prid) { 2443 mtx_lock(&pr->pr_mtx); 2444 if (pr->pr_ref > 0) 2445 return (pr); 2446 mtx_unlock(&pr->pr_mtx); 2447 } 2448 } 2449 return (NULL); 2450 } 2451 2452 /* 2453 * Find a prison that is a descendant of mypr. Returns a locked prison or NULL. 2454 */ 2455 struct prison * 2456 prison_find_child(struct prison *mypr, int prid) 2457 { 2458 struct prison *pr; 2459 int descend; 2460 2461 sx_assert(&allprison_lock, SX_LOCKED); 2462 FOREACH_PRISON_DESCENDANT(mypr, pr, descend) { 2463 if (pr->pr_id == prid) { 2464 mtx_lock(&pr->pr_mtx); 2465 if (pr->pr_ref > 0) 2466 return (pr); 2467 mtx_unlock(&pr->pr_mtx); 2468 } 2469 } 2470 return (NULL); 2471 } 2472 2473 /* 2474 * Look for the name relative to mypr. Returns a locked prison or NULL. 2475 */ 2476 struct prison * 2477 prison_find_name(struct prison *mypr, const char *name) 2478 { 2479 struct prison *pr, *deadpr; 2480 size_t mylen; 2481 int descend; 2482 2483 sx_assert(&allprison_lock, SX_LOCKED); 2484 mylen = (mypr == &prison0) ? 0 : strlen(mypr->pr_name) + 1; 2485 again: 2486 deadpr = NULL; 2487 FOREACH_PRISON_DESCENDANT(mypr, pr, descend) { 2488 if (!strcmp(pr->pr_name + mylen, name)) { 2489 mtx_lock(&pr->pr_mtx); 2490 if (pr->pr_ref > 0) { 2491 if (pr->pr_uref > 0) 2492 return (pr); 2493 deadpr = pr; 2494 } 2495 mtx_unlock(&pr->pr_mtx); 2496 } 2497 } 2498 /* There was no valid prison - perhaps there was a dying one. */ 2499 if (deadpr != NULL) { 2500 mtx_lock(&deadpr->pr_mtx); 2501 if (deadpr->pr_ref == 0) { 2502 mtx_unlock(&deadpr->pr_mtx); 2503 goto again; 2504 } 2505 } 2506 return (deadpr); 2507 } 2508 2509 /* 2510 * See if a prison has the specific flag set. 2511 */ 2512 int 2513 prison_flag(struct ucred *cred, unsigned flag) 2514 { 2515 2516 /* This is an atomic read, so no locking is necessary. */ 2517 return (cred->cr_prison->pr_flags & flag); 2518 } 2519 2520 int 2521 prison_allow(struct ucred *cred, unsigned flag) 2522 { 2523 2524 /* This is an atomic read, so no locking is necessary. */ 2525 return (cred->cr_prison->pr_allow & flag); 2526 } 2527 2528 /* 2529 * Remove a prison reference. If that was the last reference, remove the 2530 * prison itself - but not in this context in case there are locks held. 2531 */ 2532 void 2533 prison_free_locked(struct prison *pr) 2534 { 2535 int ref; 2536 2537 mtx_assert(&pr->pr_mtx, MA_OWNED); 2538 ref = --pr->pr_ref; 2539 mtx_unlock(&pr->pr_mtx); 2540 if (ref == 0) 2541 taskqueue_enqueue(taskqueue_thread, &pr->pr_task); 2542 } 2543 2544 void 2545 prison_free(struct prison *pr) 2546 { 2547 2548 mtx_lock(&pr->pr_mtx); 2549 prison_free_locked(pr); 2550 } 2551 2552 /* 2553 * Complete a call to either prison_free or prison_proc_free. 2554 */ 2555 static void 2556 prison_complete(void *context, int pending) 2557 { 2558 struct prison *pr = context; 2559 2560 sx_xlock(&allprison_lock); 2561 mtx_lock(&pr->pr_mtx); 2562 prison_deref(pr, pr->pr_uref 2563 ? PD_DEREF | PD_DEUREF | PD_LOCKED | PD_LIST_XLOCKED 2564 : PD_LOCKED | PD_LIST_XLOCKED); 2565 } 2566 2567 /* 2568 * Remove a prison reference (usually). This internal version assumes no 2569 * mutexes are held, except perhaps the prison itself. If there are no more 2570 * references, release and delist the prison. On completion, the prison lock 2571 * and the allprison lock are both unlocked. 2572 */ 2573 static void 2574 prison_deref(struct prison *pr, int flags) 2575 { 2576 struct prison *ppr, *tpr; 2577 int ref, lasturef; 2578 2579 if (!(flags & PD_LOCKED)) 2580 mtx_lock(&pr->pr_mtx); 2581 for (;;) { 2582 if (flags & PD_DEUREF) { 2583 KASSERT(pr->pr_uref > 0, 2584 ("prison_deref PD_DEUREF on a dead prison (jid=%d)", 2585 pr->pr_id)); 2586 pr->pr_uref--; 2587 lasturef = pr->pr_uref == 0; 2588 if (lasturef) 2589 pr->pr_ref++; 2590 KASSERT(prison0.pr_uref != 0, ("prison0 pr_uref=0")); 2591 } else 2592 lasturef = 0; 2593 if (flags & PD_DEREF) { 2594 KASSERT(pr->pr_ref > 0, 2595 ("prison_deref PD_DEREF on a dead prison (jid=%d)", 2596 pr->pr_id)); 2597 pr->pr_ref--; 2598 } 2599 ref = pr->pr_ref; 2600 mtx_unlock(&pr->pr_mtx); 2601 2602 /* 2603 * Tell the modules if the last user reference was removed 2604 * (even it sticks around in dying state). 2605 */ 2606 if (lasturef) { 2607 if (!(flags & (PD_LIST_SLOCKED | PD_LIST_XLOCKED))) { 2608 sx_xlock(&allprison_lock); 2609 flags |= PD_LIST_XLOCKED; 2610 } 2611 (void)osd_jail_call(pr, PR_METHOD_REMOVE, NULL); 2612 mtx_lock(&pr->pr_mtx); 2613 ref = --pr->pr_ref; 2614 mtx_unlock(&pr->pr_mtx); 2615 } 2616 2617 /* If the prison still has references, nothing else to do. */ 2618 if (ref > 0) { 2619 if (flags & PD_LIST_SLOCKED) 2620 sx_sunlock(&allprison_lock); 2621 else if (flags & PD_LIST_XLOCKED) 2622 sx_xunlock(&allprison_lock); 2623 return; 2624 } 2625 2626 if (flags & PD_LIST_SLOCKED) { 2627 if (!sx_try_upgrade(&allprison_lock)) { 2628 sx_sunlock(&allprison_lock); 2629 sx_xlock(&allprison_lock); 2630 } 2631 } else if (!(flags & PD_LIST_XLOCKED)) 2632 sx_xlock(&allprison_lock); 2633 2634 TAILQ_REMOVE(&allprison, pr, pr_list); 2635 LIST_REMOVE(pr, pr_sibling); 2636 ppr = pr->pr_parent; 2637 for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent) 2638 tpr->pr_childcount--; 2639 sx_xunlock(&allprison_lock); 2640 2641 #ifdef VIMAGE 2642 if (pr->pr_vnet != ppr->pr_vnet) 2643 vnet_destroy(pr->pr_vnet); 2644 #endif 2645 if (pr->pr_root != NULL) 2646 vrele(pr->pr_root); 2647 mtx_destroy(&pr->pr_mtx); 2648 #ifdef INET 2649 free(pr->pr_ip4, M_PRISON); 2650 #endif 2651 #ifdef INET6 2652 free(pr->pr_ip6, M_PRISON); 2653 #endif 2654 if (pr->pr_cpuset != NULL) 2655 cpuset_rel(pr->pr_cpuset); 2656 osd_jail_exit(pr); 2657 #ifdef RACCT 2658 if (racct_enable) 2659 prison_racct_detach(pr); 2660 #endif 2661 free(pr, M_PRISON); 2662 2663 /* Removing a prison frees a reference on its parent. */ 2664 pr = ppr; 2665 mtx_lock(&pr->pr_mtx); 2666 flags = PD_DEREF | PD_DEUREF; 2667 } 2668 } 2669 2670 void 2671 prison_hold_locked(struct prison *pr) 2672 { 2673 2674 mtx_assert(&pr->pr_mtx, MA_OWNED); 2675 KASSERT(pr->pr_ref > 0, 2676 ("Trying to hold dead prison (jid=%d).", pr->pr_id)); 2677 pr->pr_ref++; 2678 } 2679 2680 void 2681 prison_hold(struct prison *pr) 2682 { 2683 2684 mtx_lock(&pr->pr_mtx); 2685 prison_hold_locked(pr); 2686 mtx_unlock(&pr->pr_mtx); 2687 } 2688 2689 void 2690 prison_proc_hold(struct prison *pr) 2691 { 2692 2693 mtx_lock(&pr->pr_mtx); 2694 KASSERT(pr->pr_uref > 0, 2695 ("Cannot add a process to a non-alive prison (jid=%d)", pr->pr_id)); 2696 pr->pr_uref++; 2697 mtx_unlock(&pr->pr_mtx); 2698 } 2699 2700 void 2701 prison_proc_free(struct prison *pr) 2702 { 2703 2704 mtx_lock(&pr->pr_mtx); 2705 KASSERT(pr->pr_uref > 0, 2706 ("Trying to kill a process in a dead prison (jid=%d)", pr->pr_id)); 2707 if (pr->pr_uref > 1) 2708 pr->pr_uref--; 2709 else { 2710 /* 2711 * Don't remove the last user reference in this context, which 2712 * is expected to be a process that is not only locked, but 2713 * also half dead. 2714 */ 2715 pr->pr_ref++; 2716 mtx_unlock(&pr->pr_mtx); 2717 taskqueue_enqueue(taskqueue_thread, &pr->pr_task); 2718 return; 2719 } 2720 mtx_unlock(&pr->pr_mtx); 2721 } 2722 2723 /* 2724 * Check if a jail supports the given address family. 2725 * 2726 * Returns 0 if not jailed or the address family is supported, EAFNOSUPPORT 2727 * if not. 2728 */ 2729 int 2730 prison_check_af(struct ucred *cred, int af) 2731 { 2732 struct prison *pr; 2733 int error; 2734 2735 KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); 2736 2737 pr = cred->cr_prison; 2738 #ifdef VIMAGE 2739 /* Prisons with their own network stack are not limited. */ 2740 if (prison_owns_vnet(cred)) 2741 return (0); 2742 #endif 2743 2744 error = 0; 2745 switch (af) 2746 { 2747 #ifdef INET 2748 case AF_INET: 2749 if (pr->pr_flags & PR_IP4) 2750 { 2751 mtx_lock(&pr->pr_mtx); 2752 if ((pr->pr_flags & PR_IP4) && pr->pr_ip4 == NULL) 2753 error = EAFNOSUPPORT; 2754 mtx_unlock(&pr->pr_mtx); 2755 } 2756 break; 2757 #endif 2758 #ifdef INET6 2759 case AF_INET6: 2760 if (pr->pr_flags & PR_IP6) 2761 { 2762 mtx_lock(&pr->pr_mtx); 2763 if ((pr->pr_flags & PR_IP6) && pr->pr_ip6 == NULL) 2764 error = EAFNOSUPPORT; 2765 mtx_unlock(&pr->pr_mtx); 2766 } 2767 break; 2768 #endif 2769 case AF_LOCAL: 2770 case AF_ROUTE: 2771 break; 2772 default: 2773 if (!(pr->pr_allow & PR_ALLOW_SOCKET_AF)) 2774 error = EAFNOSUPPORT; 2775 } 2776 return (error); 2777 } 2778 2779 /* 2780 * Check if given address belongs to the jail referenced by cred (wrapper to 2781 * prison_check_ip[46]). 2782 * 2783 * Returns 0 if jail doesn't restrict the address family or if address belongs 2784 * to jail, EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if 2785 * the jail doesn't allow the address family. IPv4 Address passed in in NBO. 2786 */ 2787 int 2788 prison_if(struct ucred *cred, struct sockaddr *sa) 2789 { 2790 #ifdef INET 2791 struct sockaddr_in *sai; 2792 #endif 2793 #ifdef INET6 2794 struct sockaddr_in6 *sai6; 2795 #endif 2796 int error; 2797 2798 KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); 2799 KASSERT(sa != NULL, ("%s: sa is NULL", __func__)); 2800 2801 #ifdef VIMAGE 2802 if (prison_owns_vnet(cred)) 2803 return (0); 2804 #endif 2805 2806 error = 0; 2807 switch (sa->sa_family) 2808 { 2809 #ifdef INET 2810 case AF_INET: 2811 sai = (struct sockaddr_in *)sa; 2812 error = prison_check_ip4(cred, &sai->sin_addr); 2813 break; 2814 #endif 2815 #ifdef INET6 2816 case AF_INET6: 2817 sai6 = (struct sockaddr_in6 *)sa; 2818 error = prison_check_ip6(cred, &sai6->sin6_addr); 2819 break; 2820 #endif 2821 default: 2822 if (!(cred->cr_prison->pr_allow & PR_ALLOW_SOCKET_AF)) 2823 error = EAFNOSUPPORT; 2824 } 2825 return (error); 2826 } 2827 2828 /* 2829 * Return 0 if jails permit p1 to frob p2, otherwise ESRCH. 2830 */ 2831 int 2832 prison_check(struct ucred *cred1, struct ucred *cred2) 2833 { 2834 2835 return ((cred1->cr_prison == cred2->cr_prison || 2836 prison_ischild(cred1->cr_prison, cred2->cr_prison)) ? 0 : ESRCH); 2837 } 2838 2839 /* 2840 * Return 1 if p2 is a child of p1, otherwise 0. 2841 */ 2842 int 2843 prison_ischild(struct prison *pr1, struct prison *pr2) 2844 { 2845 2846 for (pr2 = pr2->pr_parent; pr2 != NULL; pr2 = pr2->pr_parent) 2847 if (pr1 == pr2) 2848 return (1); 2849 return (0); 2850 } 2851 2852 /* 2853 * Return 1 if the passed credential is in a jail, otherwise 0. 2854 */ 2855 int 2856 jailed(struct ucred *cred) 2857 { 2858 2859 return (cred->cr_prison != &prison0); 2860 } 2861 2862 /* 2863 * Return 1 if the passed credential is in a jail and that jail does not 2864 * have its own virtual network stack, otherwise 0. 2865 */ 2866 int 2867 jailed_without_vnet(struct ucred *cred) 2868 { 2869 2870 if (!jailed(cred)) 2871 return (0); 2872 #ifdef VIMAGE 2873 if (prison_owns_vnet(cred)) 2874 return (0); 2875 #endif 2876 2877 return (1); 2878 } 2879 2880 /* 2881 * Return the correct hostname (domainname, et al) for the passed credential. 2882 */ 2883 void 2884 getcredhostname(struct ucred *cred, char *buf, size_t size) 2885 { 2886 struct prison *pr; 2887 2888 /* 2889 * A NULL credential can be used to shortcut to the physical 2890 * system's hostname. 2891 */ 2892 pr = (cred != NULL) ? cred->cr_prison : &prison0; 2893 mtx_lock(&pr->pr_mtx); 2894 strlcpy(buf, pr->pr_hostname, size); 2895 mtx_unlock(&pr->pr_mtx); 2896 } 2897 2898 void 2899 getcreddomainname(struct ucred *cred, char *buf, size_t size) 2900 { 2901 2902 mtx_lock(&cred->cr_prison->pr_mtx); 2903 strlcpy(buf, cred->cr_prison->pr_domainname, size); 2904 mtx_unlock(&cred->cr_prison->pr_mtx); 2905 } 2906 2907 void 2908 getcredhostuuid(struct ucred *cred, char *buf, size_t size) 2909 { 2910 2911 mtx_lock(&cred->cr_prison->pr_mtx); 2912 strlcpy(buf, cred->cr_prison->pr_hostuuid, size); 2913 mtx_unlock(&cred->cr_prison->pr_mtx); 2914 } 2915 2916 void 2917 getcredhostid(struct ucred *cred, unsigned long *hostid) 2918 { 2919 2920 mtx_lock(&cred->cr_prison->pr_mtx); 2921 *hostid = cred->cr_prison->pr_hostid; 2922 mtx_unlock(&cred->cr_prison->pr_mtx); 2923 } 2924 2925 #ifdef VIMAGE 2926 /* 2927 * Determine whether the prison represented by cred owns 2928 * its vnet rather than having it inherited. 2929 * 2930 * Returns 1 in case the prison owns the vnet, 0 otherwise. 2931 */ 2932 int 2933 prison_owns_vnet(struct ucred *cred) 2934 { 2935 2936 /* 2937 * vnets cannot be added/removed after jail creation, 2938 * so no need to lock here. 2939 */ 2940 return (cred->cr_prison->pr_flags & PR_VNET ? 1 : 0); 2941 } 2942 #endif 2943 2944 /* 2945 * Determine whether the subject represented by cred can "see" 2946 * status of a mount point. 2947 * Returns: 0 for permitted, ENOENT otherwise. 2948 * XXX: This function should be called cr_canseemount() and should be 2949 * placed in kern_prot.c. 2950 */ 2951 int 2952 prison_canseemount(struct ucred *cred, struct mount *mp) 2953 { 2954 struct prison *pr; 2955 struct statfs *sp; 2956 size_t len; 2957 2958 pr = cred->cr_prison; 2959 if (pr->pr_enforce_statfs == 0) 2960 return (0); 2961 if (pr->pr_root->v_mount == mp) 2962 return (0); 2963 if (pr->pr_enforce_statfs == 2) 2964 return (ENOENT); 2965 /* 2966 * If jail's chroot directory is set to "/" we should be able to see 2967 * all mount-points from inside a jail. 2968 * This is ugly check, but this is the only situation when jail's 2969 * directory ends with '/'. 2970 */ 2971 if (strcmp(pr->pr_path, "/") == 0) 2972 return (0); 2973 len = strlen(pr->pr_path); 2974 sp = &mp->mnt_stat; 2975 if (strncmp(pr->pr_path, sp->f_mntonname, len) != 0) 2976 return (ENOENT); 2977 /* 2978 * Be sure that we don't have situation where jail's root directory 2979 * is "/some/path" and mount point is "/some/pathpath". 2980 */ 2981 if (sp->f_mntonname[len] != '\0' && sp->f_mntonname[len] != '/') 2982 return (ENOENT); 2983 return (0); 2984 } 2985 2986 void 2987 prison_enforce_statfs(struct ucred *cred, struct mount *mp, struct statfs *sp) 2988 { 2989 char jpath[MAXPATHLEN]; 2990 struct prison *pr; 2991 size_t len; 2992 2993 pr = cred->cr_prison; 2994 if (pr->pr_enforce_statfs == 0) 2995 return; 2996 if (prison_canseemount(cred, mp) != 0) { 2997 bzero(sp->f_mntonname, sizeof(sp->f_mntonname)); 2998 strlcpy(sp->f_mntonname, "[restricted]", 2999 sizeof(sp->f_mntonname)); 3000 return; 3001 } 3002 if (pr->pr_root->v_mount == mp) { 3003 /* 3004 * Clear current buffer data, so we are sure nothing from 3005 * the valid path left there. 3006 */ 3007 bzero(sp->f_mntonname, sizeof(sp->f_mntonname)); 3008 *sp->f_mntonname = '/'; 3009 return; 3010 } 3011 /* 3012 * If jail's chroot directory is set to "/" we should be able to see 3013 * all mount-points from inside a jail. 3014 */ 3015 if (strcmp(pr->pr_path, "/") == 0) 3016 return; 3017 len = strlen(pr->pr_path); 3018 strlcpy(jpath, sp->f_mntonname + len, sizeof(jpath)); 3019 /* 3020 * Clear current buffer data, so we are sure nothing from 3021 * the valid path left there. 3022 */ 3023 bzero(sp->f_mntonname, sizeof(sp->f_mntonname)); 3024 if (*jpath == '\0') { 3025 /* Should never happen. */ 3026 *sp->f_mntonname = '/'; 3027 } else { 3028 strlcpy(sp->f_mntonname, jpath, sizeof(sp->f_mntonname)); 3029 } 3030 } 3031 3032 /* 3033 * Check with permission for a specific privilege is granted within jail. We 3034 * have a specific list of accepted privileges; the rest are denied. 3035 */ 3036 int 3037 prison_priv_check(struct ucred *cred, int priv) 3038 { 3039 3040 if (!jailed(cred)) 3041 return (0); 3042 3043 #ifdef VIMAGE 3044 /* 3045 * Privileges specific to prisons with a virtual network stack. 3046 * There might be a duplicate entry here in case the privilege 3047 * is only granted conditionally in the legacy jail case. 3048 */ 3049 switch (priv) { 3050 #ifdef notyet 3051 /* 3052 * NFS-specific privileges. 3053 */ 3054 case PRIV_NFS_DAEMON: 3055 case PRIV_NFS_LOCKD: 3056 #endif 3057 /* 3058 * Network stack privileges. 3059 */ 3060 case PRIV_NET_BRIDGE: 3061 case PRIV_NET_GRE: 3062 case PRIV_NET_BPF: 3063 case PRIV_NET_RAW: /* Dup, cond. in legacy jail case. */ 3064 case PRIV_NET_ROUTE: 3065 case PRIV_NET_TAP: 3066 case PRIV_NET_SETIFMTU: 3067 case PRIV_NET_SETIFFLAGS: 3068 case PRIV_NET_SETIFCAP: 3069 case PRIV_NET_SETIFDESCR: 3070 case PRIV_NET_SETIFNAME : 3071 case PRIV_NET_SETIFMETRIC: 3072 case PRIV_NET_SETIFPHYS: 3073 case PRIV_NET_SETIFMAC: 3074 case PRIV_NET_ADDMULTI: 3075 case PRIV_NET_DELMULTI: 3076 case PRIV_NET_HWIOCTL: 3077 case PRIV_NET_SETLLADDR: 3078 case PRIV_NET_ADDIFGROUP: 3079 case PRIV_NET_DELIFGROUP: 3080 case PRIV_NET_IFCREATE: 3081 case PRIV_NET_IFDESTROY: 3082 case PRIV_NET_ADDIFADDR: 3083 case PRIV_NET_DELIFADDR: 3084 case PRIV_NET_LAGG: 3085 case PRIV_NET_GIF: 3086 case PRIV_NET_SETIFVNET: 3087 case PRIV_NET_SETIFFIB: 3088 3089 /* 3090 * 802.11-related privileges. 3091 */ 3092 case PRIV_NET80211_GETKEY: 3093 #ifdef notyet 3094 case PRIV_NET80211_MANAGE: /* XXX-BZ discuss with sam@ */ 3095 #endif 3096 3097 #ifdef notyet 3098 /* 3099 * ATM privileges. 3100 */ 3101 case PRIV_NETATM_CFG: 3102 case PRIV_NETATM_ADD: 3103 case PRIV_NETATM_DEL: 3104 case PRIV_NETATM_SET: 3105 3106 /* 3107 * Bluetooth privileges. 3108 */ 3109 case PRIV_NETBLUETOOTH_RAW: 3110 #endif 3111 3112 /* 3113 * Netgraph and netgraph module privileges. 3114 */ 3115 case PRIV_NETGRAPH_CONTROL: 3116 #ifdef notyet 3117 case PRIV_NETGRAPH_TTY: 3118 #endif 3119 3120 /* 3121 * IPv4 and IPv6 privileges. 3122 */ 3123 case PRIV_NETINET_IPFW: 3124 case PRIV_NETINET_DIVERT: 3125 case PRIV_NETINET_PF: 3126 case PRIV_NETINET_DUMMYNET: 3127 case PRIV_NETINET_CARP: 3128 case PRIV_NETINET_MROUTE: 3129 case PRIV_NETINET_RAW: 3130 case PRIV_NETINET_ADDRCTRL6: 3131 case PRIV_NETINET_ND6: 3132 case PRIV_NETINET_SCOPE6: 3133 case PRIV_NETINET_ALIFETIME6: 3134 case PRIV_NETINET_IPSEC: 3135 case PRIV_NETINET_BINDANY: 3136 3137 #ifdef notyet 3138 /* 3139 * NCP privileges. 3140 */ 3141 case PRIV_NETNCP: 3142 3143 /* 3144 * SMB privileges. 3145 */ 3146 case PRIV_NETSMB: 3147 #endif 3148 3149 /* 3150 * No default: or deny here. 3151 * In case of no permit fall through to next switch(). 3152 */ 3153 if (cred->cr_prison->pr_flags & PR_VNET) 3154 return (0); 3155 } 3156 #endif /* VIMAGE */ 3157 3158 switch (priv) { 3159 3160 /* 3161 * Allow ktrace privileges for root in jail. 3162 */ 3163 case PRIV_KTRACE: 3164 3165 #if 0 3166 /* 3167 * Allow jailed processes to configure audit identity and 3168 * submit audit records (login, etc). In the future we may 3169 * want to further refine the relationship between audit and 3170 * jail. 3171 */ 3172 case PRIV_AUDIT_GETAUDIT: 3173 case PRIV_AUDIT_SETAUDIT: 3174 case PRIV_AUDIT_SUBMIT: 3175 #endif 3176 3177 /* 3178 * Allow jailed processes to manipulate process UNIX 3179 * credentials in any way they see fit. 3180 */ 3181 case PRIV_CRED_SETUID: 3182 case PRIV_CRED_SETEUID: 3183 case PRIV_CRED_SETGID: 3184 case PRIV_CRED_SETEGID: 3185 case PRIV_CRED_SETGROUPS: 3186 case PRIV_CRED_SETREUID: 3187 case PRIV_CRED_SETREGID: 3188 case PRIV_CRED_SETRESUID: 3189 case PRIV_CRED_SETRESGID: 3190 3191 /* 3192 * Jail implements visibility constraints already, so allow 3193 * jailed root to override uid/gid-based constraints. 3194 */ 3195 case PRIV_SEEOTHERGIDS: 3196 case PRIV_SEEOTHERUIDS: 3197 3198 /* 3199 * Jail implements inter-process debugging limits already, so 3200 * allow jailed root various debugging privileges. 3201 */ 3202 case PRIV_DEBUG_DIFFCRED: 3203 case PRIV_DEBUG_SUGID: 3204 case PRIV_DEBUG_UNPRIV: 3205 3206 /* 3207 * Allow jail to set various resource limits and login 3208 * properties, and for now, exceed process resource limits. 3209 */ 3210 case PRIV_PROC_LIMIT: 3211 case PRIV_PROC_SETLOGIN: 3212 case PRIV_PROC_SETRLIMIT: 3213 3214 /* 3215 * System V and POSIX IPC privileges are granted in jail. 3216 */ 3217 case PRIV_IPC_READ: 3218 case PRIV_IPC_WRITE: 3219 case PRIV_IPC_ADMIN: 3220 case PRIV_IPC_MSGSIZE: 3221 case PRIV_MQ_ADMIN: 3222 3223 /* 3224 * Jail operations within a jail work on child jails. 3225 */ 3226 case PRIV_JAIL_ATTACH: 3227 case PRIV_JAIL_SET: 3228 case PRIV_JAIL_REMOVE: 3229 3230 /* 3231 * Jail implements its own inter-process limits, so allow 3232 * root processes in jail to change scheduling on other 3233 * processes in the same jail. Likewise for signalling. 3234 */ 3235 case PRIV_SCHED_DIFFCRED: 3236 case PRIV_SCHED_CPUSET: 3237 case PRIV_SIGNAL_DIFFCRED: 3238 case PRIV_SIGNAL_SUGID: 3239 3240 /* 3241 * Allow jailed processes to write to sysctls marked as jail 3242 * writable. 3243 */ 3244 case PRIV_SYSCTL_WRITEJAIL: 3245 3246 /* 3247 * Allow root in jail to manage a variety of quota 3248 * properties. These should likely be conditional on a 3249 * configuration option. 3250 */ 3251 case PRIV_VFS_GETQUOTA: 3252 case PRIV_VFS_SETQUOTA: 3253 3254 /* 3255 * Since Jail relies on chroot() to implement file system 3256 * protections, grant many VFS privileges to root in jail. 3257 * Be careful to exclude mount-related and NFS-related 3258 * privileges. 3259 */ 3260 case PRIV_VFS_READ: 3261 case PRIV_VFS_WRITE: 3262 case PRIV_VFS_ADMIN: 3263 case PRIV_VFS_EXEC: 3264 case PRIV_VFS_LOOKUP: 3265 case PRIV_VFS_BLOCKRESERVE: /* XXXRW: Slightly surprising. */ 3266 case PRIV_VFS_CHFLAGS_DEV: 3267 case PRIV_VFS_CHOWN: 3268 case PRIV_VFS_CHROOT: 3269 case PRIV_VFS_RETAINSUGID: 3270 case PRIV_VFS_FCHROOT: 3271 case PRIV_VFS_LINK: 3272 case PRIV_VFS_SETGID: 3273 case PRIV_VFS_STAT: 3274 case PRIV_VFS_STICKYFILE: 3275 3276 /* 3277 * As in the non-jail case, non-root users are expected to be 3278 * able to read kernel/phyiscal memory (provided /dev/[k]mem 3279 * exists in the jail and they have permission to access it). 3280 */ 3281 case PRIV_KMEM_READ: 3282 return (0); 3283 3284 /* 3285 * Depending on the global setting, allow privilege of 3286 * setting system flags. 3287 */ 3288 case PRIV_VFS_SYSFLAGS: 3289 if (cred->cr_prison->pr_allow & PR_ALLOW_CHFLAGS) 3290 return (0); 3291 else 3292 return (EPERM); 3293 3294 /* 3295 * Depending on the global setting, allow privilege of 3296 * mounting/unmounting file systems. 3297 */ 3298 case PRIV_VFS_MOUNT: 3299 case PRIV_VFS_UNMOUNT: 3300 case PRIV_VFS_MOUNT_NONUSER: 3301 case PRIV_VFS_MOUNT_OWNER: 3302 if (cred->cr_prison->pr_allow & PR_ALLOW_MOUNT && 3303 cred->cr_prison->pr_enforce_statfs < 2) 3304 return (0); 3305 else 3306 return (EPERM); 3307 3308 /* 3309 * Conditionally allow jailed root to bind reserved ports. 3310 */ 3311 case PRIV_NETINET_RESERVEDPORT: 3312 if (cred->cr_prison->pr_allow & PR_ALLOW_RESERVED_PORTS) 3313 return (0); 3314 else 3315 return (EPERM); 3316 3317 /* 3318 * Allow jailed root to reuse in-use ports. 3319 */ 3320 case PRIV_NETINET_REUSEPORT: 3321 return (0); 3322 3323 /* 3324 * Allow jailed root to set certain IPv4/6 (option) headers. 3325 */ 3326 case PRIV_NETINET_SETHDROPTS: 3327 return (0); 3328 3329 /* 3330 * Conditionally allow creating raw sockets in jail. 3331 */ 3332 case PRIV_NETINET_RAW: 3333 if (cred->cr_prison->pr_allow & PR_ALLOW_RAW_SOCKETS) 3334 return (0); 3335 else 3336 return (EPERM); 3337 3338 /* 3339 * Since jail implements its own visibility limits on netstat 3340 * sysctls, allow getcred. This allows identd to work in 3341 * jail. 3342 */ 3343 case PRIV_NETINET_GETCRED: 3344 return (0); 3345 3346 /* 3347 * Allow jailed root to set loginclass. 3348 */ 3349 case PRIV_PROC_SETLOGINCLASS: 3350 return (0); 3351 3352 default: 3353 /* 3354 * In all remaining cases, deny the privilege request. This 3355 * includes almost all network privileges, many system 3356 * configuration privileges. 3357 */ 3358 return (EPERM); 3359 } 3360 } 3361 3362 /* 3363 * Return the part of pr2's name that is relative to pr1, or the whole name 3364 * if it does not directly follow. 3365 */ 3366 3367 char * 3368 prison_name(struct prison *pr1, struct prison *pr2) 3369 { 3370 char *name; 3371 3372 /* Jails see themselves as "0" (if they see themselves at all). */ 3373 if (pr1 == pr2) 3374 return "0"; 3375 name = pr2->pr_name; 3376 if (prison_ischild(pr1, pr2)) { 3377 /* 3378 * pr1 isn't locked (and allprison_lock may not be either) 3379 * so its length can't be counted on. But the number of dots 3380 * can be counted on - and counted. 3381 */ 3382 for (; pr1 != &prison0; pr1 = pr1->pr_parent) 3383 name = strchr(name, '.') + 1; 3384 } 3385 return (name); 3386 } 3387 3388 /* 3389 * Return the part of pr2's path that is relative to pr1, or the whole path 3390 * if it does not directly follow. 3391 */ 3392 static char * 3393 prison_path(struct prison *pr1, struct prison *pr2) 3394 { 3395 char *path1, *path2; 3396 int len1; 3397 3398 path1 = pr1->pr_path; 3399 path2 = pr2->pr_path; 3400 if (!strcmp(path1, "/")) 3401 return (path2); 3402 len1 = strlen(path1); 3403 if (strncmp(path1, path2, len1)) 3404 return (path2); 3405 if (path2[len1] == '\0') 3406 return "/"; 3407 if (path2[len1] == '/') 3408 return (path2 + len1); 3409 return (path2); 3410 } 3411 3412 3413 /* 3414 * Jail-related sysctls. 3415 */ 3416 static SYSCTL_NODE(_security, OID_AUTO, jail, CTLFLAG_RW, 0, 3417 "Jails"); 3418 3419 static int 3420 sysctl_jail_list(SYSCTL_HANDLER_ARGS) 3421 { 3422 struct xprison *xp; 3423 struct prison *pr, *cpr; 3424 #ifdef INET 3425 struct in_addr *ip4 = NULL; 3426 int ip4s = 0; 3427 #endif 3428 #ifdef INET6 3429 struct in6_addr *ip6 = NULL; 3430 int ip6s = 0; 3431 #endif 3432 int descend, error; 3433 3434 xp = malloc(sizeof(*xp), M_TEMP, M_WAITOK); 3435 pr = req->td->td_ucred->cr_prison; 3436 error = 0; 3437 sx_slock(&allprison_lock); 3438 FOREACH_PRISON_DESCENDANT(pr, cpr, descend) { 3439 #if defined(INET) || defined(INET6) 3440 again: 3441 #endif 3442 mtx_lock(&cpr->pr_mtx); 3443 #ifdef INET 3444 if (cpr->pr_ip4s > 0) { 3445 if (ip4s < cpr->pr_ip4s) { 3446 ip4s = cpr->pr_ip4s; 3447 mtx_unlock(&cpr->pr_mtx); 3448 ip4 = realloc(ip4, ip4s * 3449 sizeof(struct in_addr), M_TEMP, M_WAITOK); 3450 goto again; 3451 } 3452 bcopy(cpr->pr_ip4, ip4, 3453 cpr->pr_ip4s * sizeof(struct in_addr)); 3454 } 3455 #endif 3456 #ifdef INET6 3457 if (cpr->pr_ip6s > 0) { 3458 if (ip6s < cpr->pr_ip6s) { 3459 ip6s = cpr->pr_ip6s; 3460 mtx_unlock(&cpr->pr_mtx); 3461 ip6 = realloc(ip6, ip6s * 3462 sizeof(struct in6_addr), M_TEMP, M_WAITOK); 3463 goto again; 3464 } 3465 bcopy(cpr->pr_ip6, ip6, 3466 cpr->pr_ip6s * sizeof(struct in6_addr)); 3467 } 3468 #endif 3469 if (cpr->pr_ref == 0) { 3470 mtx_unlock(&cpr->pr_mtx); 3471 continue; 3472 } 3473 bzero(xp, sizeof(*xp)); 3474 xp->pr_version = XPRISON_VERSION; 3475 xp->pr_id = cpr->pr_id; 3476 xp->pr_state = cpr->pr_uref > 0 3477 ? PRISON_STATE_ALIVE : PRISON_STATE_DYING; 3478 strlcpy(xp->pr_path, prison_path(pr, cpr), sizeof(xp->pr_path)); 3479 strlcpy(xp->pr_host, cpr->pr_hostname, sizeof(xp->pr_host)); 3480 strlcpy(xp->pr_name, prison_name(pr, cpr), sizeof(xp->pr_name)); 3481 #ifdef INET 3482 xp->pr_ip4s = cpr->pr_ip4s; 3483 #endif 3484 #ifdef INET6 3485 xp->pr_ip6s = cpr->pr_ip6s; 3486 #endif 3487 mtx_unlock(&cpr->pr_mtx); 3488 error = SYSCTL_OUT(req, xp, sizeof(*xp)); 3489 if (error) 3490 break; 3491 #ifdef INET 3492 if (xp->pr_ip4s > 0) { 3493 error = SYSCTL_OUT(req, ip4, 3494 xp->pr_ip4s * sizeof(struct in_addr)); 3495 if (error) 3496 break; 3497 } 3498 #endif 3499 #ifdef INET6 3500 if (xp->pr_ip6s > 0) { 3501 error = SYSCTL_OUT(req, ip6, 3502 xp->pr_ip6s * sizeof(struct in6_addr)); 3503 if (error) 3504 break; 3505 } 3506 #endif 3507 } 3508 sx_sunlock(&allprison_lock); 3509 free(xp, M_TEMP); 3510 #ifdef INET 3511 free(ip4, M_TEMP); 3512 #endif 3513 #ifdef INET6 3514 free(ip6, M_TEMP); 3515 #endif 3516 return (error); 3517 } 3518 3519 SYSCTL_OID(_security_jail, OID_AUTO, list, 3520 CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0, 3521 sysctl_jail_list, "S", "List of active jails"); 3522 3523 static int 3524 sysctl_jail_jailed(SYSCTL_HANDLER_ARGS) 3525 { 3526 int error, injail; 3527 3528 injail = jailed(req->td->td_ucred); 3529 error = SYSCTL_OUT(req, &injail, sizeof(injail)); 3530 3531 return (error); 3532 } 3533 3534 SYSCTL_PROC(_security_jail, OID_AUTO, jailed, 3535 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0, 3536 sysctl_jail_jailed, "I", "Process in jail?"); 3537 3538 static int 3539 sysctl_jail_vnet(SYSCTL_HANDLER_ARGS) 3540 { 3541 int error, havevnet; 3542 #ifdef VIMAGE 3543 struct ucred *cred = req->td->td_ucred; 3544 3545 havevnet = jailed(cred) && prison_owns_vnet(cred); 3546 #else 3547 havevnet = 0; 3548 #endif 3549 error = SYSCTL_OUT(req, &havevnet, sizeof(havevnet)); 3550 3551 return (error); 3552 } 3553 3554 SYSCTL_PROC(_security_jail, OID_AUTO, vnet, 3555 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0, 3556 sysctl_jail_vnet, "I", "Jail owns VNET?"); 3557 3558 #if defined(INET) || defined(INET6) 3559 SYSCTL_UINT(_security_jail, OID_AUTO, jail_max_af_ips, CTLFLAG_RW, 3560 &jail_max_af_ips, 0, 3561 "Number of IP addresses a jail may have at most per address family (deprecated)"); 3562 #endif 3563 3564 /* 3565 * Default parameters for jail(2) compatibility. For historical reasons, 3566 * the sysctl names have varying similarity to the parameter names. Prisons 3567 * just see their own parameters, and can't change them. 3568 */ 3569 static int 3570 sysctl_jail_default_allow(SYSCTL_HANDLER_ARGS) 3571 { 3572 struct prison *pr; 3573 int allow, error, i; 3574 3575 pr = req->td->td_ucred->cr_prison; 3576 allow = (pr == &prison0) ? jail_default_allow : pr->pr_allow; 3577 3578 /* Get the current flag value, and convert it to a boolean. */ 3579 i = (allow & arg2) ? 1 : 0; 3580 if (arg1 != NULL) 3581 i = !i; 3582 error = sysctl_handle_int(oidp, &i, 0, req); 3583 if (error || !req->newptr) 3584 return (error); 3585 i = i ? arg2 : 0; 3586 if (arg1 != NULL) 3587 i ^= arg2; 3588 /* 3589 * The sysctls don't have CTLFLAGS_PRISON, so assume prison0 3590 * for writing. 3591 */ 3592 mtx_lock(&prison0.pr_mtx); 3593 jail_default_allow = (jail_default_allow & ~arg2) | i; 3594 mtx_unlock(&prison0.pr_mtx); 3595 return (0); 3596 } 3597 3598 SYSCTL_PROC(_security_jail, OID_AUTO, set_hostname_allowed, 3599 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 3600 NULL, PR_ALLOW_SET_HOSTNAME, sysctl_jail_default_allow, "I", 3601 "Processes in jail can set their hostnames (deprecated)"); 3602 SYSCTL_PROC(_security_jail, OID_AUTO, socket_unixiproute_only, 3603 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 3604 (void *)1, PR_ALLOW_SOCKET_AF, sysctl_jail_default_allow, "I", 3605 "Processes in jail are limited to creating UNIX/IP/route sockets only (deprecated)"); 3606 SYSCTL_PROC(_security_jail, OID_AUTO, sysvipc_allowed, 3607 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 3608 NULL, PR_ALLOW_SYSVIPC, sysctl_jail_default_allow, "I", 3609 "Processes in jail can use System V IPC primitives (deprecated)"); 3610 SYSCTL_PROC(_security_jail, OID_AUTO, allow_raw_sockets, 3611 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 3612 NULL, PR_ALLOW_RAW_SOCKETS, sysctl_jail_default_allow, "I", 3613 "Prison root can create raw sockets (deprecated)"); 3614 SYSCTL_PROC(_security_jail, OID_AUTO, chflags_allowed, 3615 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 3616 NULL, PR_ALLOW_CHFLAGS, sysctl_jail_default_allow, "I", 3617 "Processes in jail can alter system file flags (deprecated)"); 3618 SYSCTL_PROC(_security_jail, OID_AUTO, mount_allowed, 3619 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 3620 NULL, PR_ALLOW_MOUNT, sysctl_jail_default_allow, "I", 3621 "Processes in jail can mount/unmount jail-friendly file systems (deprecated)"); 3622 SYSCTL_PROC(_security_jail, OID_AUTO, mount_devfs_allowed, 3623 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 3624 NULL, PR_ALLOW_MOUNT_DEVFS, sysctl_jail_default_allow, "I", 3625 "Processes in jail can mount the devfs file system (deprecated)"); 3626 SYSCTL_PROC(_security_jail, OID_AUTO, mount_fdescfs_allowed, 3627 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 3628 NULL, PR_ALLOW_MOUNT_FDESCFS, sysctl_jail_default_allow, "I", 3629 "Processes in jail can mount the fdescfs file system (deprecated)"); 3630 SYSCTL_PROC(_security_jail, OID_AUTO, mount_nullfs_allowed, 3631 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 3632 NULL, PR_ALLOW_MOUNT_NULLFS, sysctl_jail_default_allow, "I", 3633 "Processes in jail can mount the nullfs file system (deprecated)"); 3634 SYSCTL_PROC(_security_jail, OID_AUTO, mount_procfs_allowed, 3635 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 3636 NULL, PR_ALLOW_MOUNT_PROCFS, sysctl_jail_default_allow, "I", 3637 "Processes in jail can mount the procfs file system (deprecated)"); 3638 SYSCTL_PROC(_security_jail, OID_AUTO, mount_linprocfs_allowed, 3639 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 3640 NULL, PR_ALLOW_MOUNT_LINPROCFS, sysctl_jail_default_allow, "I", 3641 "Processes in jail can mount the linprocfs file system (deprecated)"); 3642 SYSCTL_PROC(_security_jail, OID_AUTO, mount_linsysfs_allowed, 3643 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 3644 NULL, PR_ALLOW_MOUNT_LINSYSFS, sysctl_jail_default_allow, "I", 3645 "Processes in jail can mount the linsysfs file system (deprecated)"); 3646 SYSCTL_PROC(_security_jail, OID_AUTO, mount_tmpfs_allowed, 3647 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 3648 NULL, PR_ALLOW_MOUNT_TMPFS, sysctl_jail_default_allow, "I", 3649 "Processes in jail can mount the tmpfs file system (deprecated)"); 3650 SYSCTL_PROC(_security_jail, OID_AUTO, mount_zfs_allowed, 3651 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 3652 NULL, PR_ALLOW_MOUNT_ZFS, sysctl_jail_default_allow, "I", 3653 "Processes in jail can mount the zfs file system (deprecated)"); 3654 3655 static int 3656 sysctl_jail_default_level(SYSCTL_HANDLER_ARGS) 3657 { 3658 struct prison *pr; 3659 int level, error; 3660 3661 pr = req->td->td_ucred->cr_prison; 3662 level = (pr == &prison0) ? *(int *)arg1 : *(int *)((char *)pr + arg2); 3663 error = sysctl_handle_int(oidp, &level, 0, req); 3664 if (error || !req->newptr) 3665 return (error); 3666 *(int *)arg1 = level; 3667 return (0); 3668 } 3669 3670 SYSCTL_PROC(_security_jail, OID_AUTO, enforce_statfs, 3671 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 3672 &jail_default_enforce_statfs, offsetof(struct prison, pr_enforce_statfs), 3673 sysctl_jail_default_level, "I", 3674 "Processes in jail cannot see all mounted file systems (deprecated)"); 3675 3676 SYSCTL_PROC(_security_jail, OID_AUTO, devfs_ruleset, 3677 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, 3678 &jail_default_devfs_rsnum, offsetof(struct prison, pr_devfs_rsnum), 3679 sysctl_jail_default_level, "I", 3680 "Ruleset for the devfs filesystem in jail (deprecated)"); 3681 3682 /* 3683 * Nodes to describe jail parameters. Maximum length of string parameters 3684 * is returned in the string itself, and the other parameters exist merely 3685 * to make themselves and their types known. 3686 */ 3687 SYSCTL_NODE(_security_jail, OID_AUTO, param, CTLFLAG_RW, 0, 3688 "Jail parameters"); 3689 3690 int 3691 sysctl_jail_param(SYSCTL_HANDLER_ARGS) 3692 { 3693 int i; 3694 long l; 3695 size_t s; 3696 char numbuf[12]; 3697 3698 switch (oidp->oid_kind & CTLTYPE) 3699 { 3700 case CTLTYPE_LONG: 3701 case CTLTYPE_ULONG: 3702 l = 0; 3703 #ifdef SCTL_MASK32 3704 if (!(req->flags & SCTL_MASK32)) 3705 #endif 3706 return (SYSCTL_OUT(req, &l, sizeof(l))); 3707 case CTLTYPE_INT: 3708 case CTLTYPE_UINT: 3709 i = 0; 3710 return (SYSCTL_OUT(req, &i, sizeof(i))); 3711 case CTLTYPE_STRING: 3712 snprintf(numbuf, sizeof(numbuf), "%jd", (intmax_t)arg2); 3713 return 3714 (sysctl_handle_string(oidp, numbuf, sizeof(numbuf), req)); 3715 case CTLTYPE_STRUCT: 3716 s = (size_t)arg2; 3717 return (SYSCTL_OUT(req, &s, sizeof(s))); 3718 } 3719 return (0); 3720 } 3721 3722 /* 3723 * CTLFLAG_RDTUN in the following indicates jail parameters that can be set at 3724 * jail creation time but cannot be changed in an existing jail. 3725 */ 3726 SYSCTL_JAIL_PARAM(, jid, CTLTYPE_INT | CTLFLAG_RDTUN, "I", "Jail ID"); 3727 SYSCTL_JAIL_PARAM(, parent, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail parent ID"); 3728 SYSCTL_JAIL_PARAM_STRING(, name, CTLFLAG_RW, MAXHOSTNAMELEN, "Jail name"); 3729 SYSCTL_JAIL_PARAM_STRING(, path, CTLFLAG_RDTUN, MAXPATHLEN, "Jail root path"); 3730 SYSCTL_JAIL_PARAM(, securelevel, CTLTYPE_INT | CTLFLAG_RW, 3731 "I", "Jail secure level"); 3732 SYSCTL_JAIL_PARAM(, osreldate, CTLTYPE_INT | CTLFLAG_RDTUN, "I", 3733 "Jail value for kern.osreldate and uname -K"); 3734 SYSCTL_JAIL_PARAM_STRING(, osrelease, CTLFLAG_RDTUN, OSRELEASELEN, 3735 "Jail value for kern.osrelease and uname -r"); 3736 SYSCTL_JAIL_PARAM(, enforce_statfs, CTLTYPE_INT | CTLFLAG_RW, 3737 "I", "Jail cannot see all mounted file systems"); 3738 SYSCTL_JAIL_PARAM(, devfs_ruleset, CTLTYPE_INT | CTLFLAG_RW, 3739 "I", "Ruleset for in-jail devfs mounts"); 3740 SYSCTL_JAIL_PARAM(, persist, CTLTYPE_INT | CTLFLAG_RW, 3741 "B", "Jail persistence"); 3742 #ifdef VIMAGE 3743 SYSCTL_JAIL_PARAM(, vnet, CTLTYPE_INT | CTLFLAG_RDTUN, 3744 "E,jailsys", "Virtual network stack"); 3745 #endif 3746 SYSCTL_JAIL_PARAM(, dying, CTLTYPE_INT | CTLFLAG_RD, 3747 "B", "Jail is in the process of shutting down"); 3748 3749 SYSCTL_JAIL_PARAM_NODE(children, "Number of child jails"); 3750 SYSCTL_JAIL_PARAM(_children, cur, CTLTYPE_INT | CTLFLAG_RD, 3751 "I", "Current number of child jails"); 3752 SYSCTL_JAIL_PARAM(_children, max, CTLTYPE_INT | CTLFLAG_RW, 3753 "I", "Maximum number of child jails"); 3754 3755 SYSCTL_JAIL_PARAM_SYS_NODE(host, CTLFLAG_RW, "Jail host info"); 3756 SYSCTL_JAIL_PARAM_STRING(_host, hostname, CTLFLAG_RW, MAXHOSTNAMELEN, 3757 "Jail hostname"); 3758 SYSCTL_JAIL_PARAM_STRING(_host, domainname, CTLFLAG_RW, MAXHOSTNAMELEN, 3759 "Jail NIS domainname"); 3760 SYSCTL_JAIL_PARAM_STRING(_host, hostuuid, CTLFLAG_RW, HOSTUUIDLEN, 3761 "Jail host UUID"); 3762 SYSCTL_JAIL_PARAM(_host, hostid, CTLTYPE_ULONG | CTLFLAG_RW, 3763 "LU", "Jail host ID"); 3764 3765 SYSCTL_JAIL_PARAM_NODE(cpuset, "Jail cpuset"); 3766 SYSCTL_JAIL_PARAM(_cpuset, id, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail cpuset ID"); 3767 3768 #ifdef INET 3769 SYSCTL_JAIL_PARAM_SYS_NODE(ip4, CTLFLAG_RDTUN, 3770 "Jail IPv4 address virtualization"); 3771 SYSCTL_JAIL_PARAM_STRUCT(_ip4, addr, CTLFLAG_RW, sizeof(struct in_addr), 3772 "S,in_addr,a", "Jail IPv4 addresses"); 3773 SYSCTL_JAIL_PARAM(_ip4, saddrsel, CTLTYPE_INT | CTLFLAG_RW, 3774 "B", "Do (not) use IPv4 source address selection rather than the " 3775 "primary jail IPv4 address."); 3776 #endif 3777 #ifdef INET6 3778 SYSCTL_JAIL_PARAM_SYS_NODE(ip6, CTLFLAG_RDTUN, 3779 "Jail IPv6 address virtualization"); 3780 SYSCTL_JAIL_PARAM_STRUCT(_ip6, addr, CTLFLAG_RW, sizeof(struct in6_addr), 3781 "S,in6_addr,a", "Jail IPv6 addresses"); 3782 SYSCTL_JAIL_PARAM(_ip6, saddrsel, CTLTYPE_INT | CTLFLAG_RW, 3783 "B", "Do (not) use IPv6 source address selection rather than the " 3784 "primary jail IPv6 address."); 3785 #endif 3786 3787 SYSCTL_JAIL_PARAM_NODE(allow, "Jail permission flags"); 3788 SYSCTL_JAIL_PARAM(_allow, set_hostname, CTLTYPE_INT | CTLFLAG_RW, 3789 "B", "Jail may set hostname"); 3790 SYSCTL_JAIL_PARAM(_allow, sysvipc, CTLTYPE_INT | CTLFLAG_RW, 3791 "B", "Jail may use SYSV IPC"); 3792 SYSCTL_JAIL_PARAM(_allow, raw_sockets, CTLTYPE_INT | CTLFLAG_RW, 3793 "B", "Jail may create raw sockets"); 3794 SYSCTL_JAIL_PARAM(_allow, chflags, CTLTYPE_INT | CTLFLAG_RW, 3795 "B", "Jail may alter system file flags"); 3796 SYSCTL_JAIL_PARAM(_allow, quotas, CTLTYPE_INT | CTLFLAG_RW, 3797 "B", "Jail may set file quotas"); 3798 SYSCTL_JAIL_PARAM(_allow, socket_af, CTLTYPE_INT | CTLFLAG_RW, 3799 "B", "Jail may create sockets other than just UNIX/IPv4/IPv6/route"); 3800 SYSCTL_JAIL_PARAM(_allow, reserved_ports, CTLTYPE_INT | CTLFLAG_RW, 3801 "B", "Jail may bind sockets to reserved ports"); 3802 3803 SYSCTL_JAIL_PARAM_SUBNODE(allow, mount, "Jail mount/unmount permission flags"); 3804 SYSCTL_JAIL_PARAM(_allow_mount, , CTLTYPE_INT | CTLFLAG_RW, 3805 "B", "Jail may mount/unmount jail-friendly file systems in general"); 3806 SYSCTL_JAIL_PARAM(_allow_mount, devfs, CTLTYPE_INT | CTLFLAG_RW, 3807 "B", "Jail may mount the devfs file system"); 3808 SYSCTL_JAIL_PARAM(_allow_mount, fdescfs, CTLTYPE_INT | CTLFLAG_RW, 3809 "B", "Jail may mount the fdescfs file system"); 3810 SYSCTL_JAIL_PARAM(_allow_mount, nullfs, CTLTYPE_INT | CTLFLAG_RW, 3811 "B", "Jail may mount the nullfs file system"); 3812 SYSCTL_JAIL_PARAM(_allow_mount, procfs, CTLTYPE_INT | CTLFLAG_RW, 3813 "B", "Jail may mount the procfs file system"); 3814 SYSCTL_JAIL_PARAM(_allow_mount, linprocfs, CTLTYPE_INT | CTLFLAG_RW, 3815 "B", "Jail may mount the linprocfs file system"); 3816 SYSCTL_JAIL_PARAM(_allow_mount, linsysfs, CTLTYPE_INT | CTLFLAG_RW, 3817 "B", "Jail may mount the linsysfs file system"); 3818 SYSCTL_JAIL_PARAM(_allow_mount, tmpfs, CTLTYPE_INT | CTLFLAG_RW, 3819 "B", "Jail may mount the tmpfs file system"); 3820 SYSCTL_JAIL_PARAM(_allow_mount, zfs, CTLTYPE_INT | CTLFLAG_RW, 3821 "B", "Jail may mount the zfs file system"); 3822 3823 #ifdef RACCT 3824 void 3825 prison_racct_foreach(void (*callback)(struct racct *racct, 3826 void *arg2, void *arg3), void (*pre)(void), void (*post)(void), 3827 void *arg2, void *arg3) 3828 { 3829 struct prison_racct *prr; 3830 3831 ASSERT_RACCT_ENABLED(); 3832 3833 sx_slock(&allprison_lock); 3834 if (pre != NULL) 3835 (pre)(); 3836 LIST_FOREACH(prr, &allprison_racct, prr_next) 3837 (callback)(prr->prr_racct, arg2, arg3); 3838 if (post != NULL) 3839 (post)(); 3840 sx_sunlock(&allprison_lock); 3841 } 3842 3843 static struct prison_racct * 3844 prison_racct_find_locked(const char *name) 3845 { 3846 struct prison_racct *prr; 3847 3848 ASSERT_RACCT_ENABLED(); 3849 sx_assert(&allprison_lock, SA_XLOCKED); 3850 3851 if (name[0] == '\0' || strlen(name) >= MAXHOSTNAMELEN) 3852 return (NULL); 3853 3854 LIST_FOREACH(prr, &allprison_racct, prr_next) { 3855 if (strcmp(name, prr->prr_name) != 0) 3856 continue; 3857 3858 /* Found prison_racct with a matching name? */ 3859 prison_racct_hold(prr); 3860 return (prr); 3861 } 3862 3863 /* Add new prison_racct. */ 3864 prr = malloc(sizeof(*prr), M_PRISON_RACCT, M_ZERO | M_WAITOK); 3865 racct_create(&prr->prr_racct); 3866 3867 strcpy(prr->prr_name, name); 3868 refcount_init(&prr->prr_refcount, 1); 3869 LIST_INSERT_HEAD(&allprison_racct, prr, prr_next); 3870 3871 return (prr); 3872 } 3873 3874 struct prison_racct * 3875 prison_racct_find(const char *name) 3876 { 3877 struct prison_racct *prr; 3878 3879 ASSERT_RACCT_ENABLED(); 3880 3881 sx_xlock(&allprison_lock); 3882 prr = prison_racct_find_locked(name); 3883 sx_xunlock(&allprison_lock); 3884 return (prr); 3885 } 3886 3887 void 3888 prison_racct_hold(struct prison_racct *prr) 3889 { 3890 3891 ASSERT_RACCT_ENABLED(); 3892 3893 refcount_acquire(&prr->prr_refcount); 3894 } 3895 3896 static void 3897 prison_racct_free_locked(struct prison_racct *prr) 3898 { 3899 3900 ASSERT_RACCT_ENABLED(); 3901 sx_assert(&allprison_lock, SA_XLOCKED); 3902 3903 if (refcount_release(&prr->prr_refcount)) { 3904 racct_destroy(&prr->prr_racct); 3905 LIST_REMOVE(prr, prr_next); 3906 free(prr, M_PRISON_RACCT); 3907 } 3908 } 3909 3910 void 3911 prison_racct_free(struct prison_racct *prr) 3912 { 3913 int old; 3914 3915 ASSERT_RACCT_ENABLED(); 3916 sx_assert(&allprison_lock, SA_UNLOCKED); 3917 3918 old = prr->prr_refcount; 3919 if (old > 1 && atomic_cmpset_int(&prr->prr_refcount, old, old - 1)) 3920 return; 3921 3922 sx_xlock(&allprison_lock); 3923 prison_racct_free_locked(prr); 3924 sx_xunlock(&allprison_lock); 3925 } 3926 3927 static void 3928 prison_racct_attach(struct prison *pr) 3929 { 3930 struct prison_racct *prr; 3931 3932 ASSERT_RACCT_ENABLED(); 3933 sx_assert(&allprison_lock, SA_XLOCKED); 3934 3935 prr = prison_racct_find_locked(pr->pr_name); 3936 KASSERT(prr != NULL, ("cannot find prison_racct")); 3937 3938 pr->pr_prison_racct = prr; 3939 } 3940 3941 /* 3942 * Handle jail renaming. From the racct point of view, renaming means 3943 * moving from one prison_racct to another. 3944 */ 3945 static void 3946 prison_racct_modify(struct prison *pr) 3947 { 3948 struct proc *p; 3949 struct ucred *cred; 3950 struct prison_racct *oldprr; 3951 3952 ASSERT_RACCT_ENABLED(); 3953 3954 sx_slock(&allproc_lock); 3955 sx_xlock(&allprison_lock); 3956 3957 if (strcmp(pr->pr_name, pr->pr_prison_racct->prr_name) == 0) { 3958 sx_xunlock(&allprison_lock); 3959 sx_sunlock(&allproc_lock); 3960 return; 3961 } 3962 3963 oldprr = pr->pr_prison_racct; 3964 pr->pr_prison_racct = NULL; 3965 3966 prison_racct_attach(pr); 3967 3968 /* 3969 * Move resource utilisation records. 3970 */ 3971 racct_move(pr->pr_prison_racct->prr_racct, oldprr->prr_racct); 3972 3973 /* 3974 * Force rctl to reattach rules to processes. 3975 */ 3976 FOREACH_PROC_IN_SYSTEM(p) { 3977 PROC_LOCK(p); 3978 cred = crhold(p->p_ucred); 3979 PROC_UNLOCK(p); 3980 racct_proc_ucred_changed(p, cred, cred); 3981 crfree(cred); 3982 } 3983 3984 sx_sunlock(&allproc_lock); 3985 prison_racct_free_locked(oldprr); 3986 sx_xunlock(&allprison_lock); 3987 } 3988 3989 static void 3990 prison_racct_detach(struct prison *pr) 3991 { 3992 3993 ASSERT_RACCT_ENABLED(); 3994 sx_assert(&allprison_lock, SA_UNLOCKED); 3995 3996 if (pr->pr_prison_racct == NULL) 3997 return; 3998 prison_racct_free(pr->pr_prison_racct); 3999 pr->pr_prison_racct = NULL; 4000 } 4001 #endif /* RACCT */ 4002 4003 #ifdef DDB 4004 4005 static void 4006 db_show_prison(struct prison *pr) 4007 { 4008 int fi; 4009 #if defined(INET) || defined(INET6) 4010 int ii; 4011 #endif 4012 unsigned jsf; 4013 #ifdef INET 4014 char ip4buf[INET_ADDRSTRLEN]; 4015 #endif 4016 #ifdef INET6 4017 char ip6buf[INET6_ADDRSTRLEN]; 4018 #endif 4019 4020 db_printf("prison %p:\n", pr); 4021 db_printf(" jid = %d\n", pr->pr_id); 4022 db_printf(" name = %s\n", pr->pr_name); 4023 db_printf(" parent = %p\n", pr->pr_parent); 4024 db_printf(" ref = %d\n", pr->pr_ref); 4025 db_printf(" uref = %d\n", pr->pr_uref); 4026 db_printf(" path = %s\n", pr->pr_path); 4027 db_printf(" cpuset = %d\n", pr->pr_cpuset 4028 ? pr->pr_cpuset->cs_id : -1); 4029 #ifdef VIMAGE 4030 db_printf(" vnet = %p\n", pr->pr_vnet); 4031 #endif 4032 db_printf(" root = %p\n", pr->pr_root); 4033 db_printf(" securelevel = %d\n", pr->pr_securelevel); 4034 db_printf(" devfs_rsnum = %d\n", pr->pr_devfs_rsnum); 4035 db_printf(" children.max = %d\n", pr->pr_childmax); 4036 db_printf(" children.cur = %d\n", pr->pr_childcount); 4037 db_printf(" child = %p\n", LIST_FIRST(&pr->pr_children)); 4038 db_printf(" sibling = %p\n", LIST_NEXT(pr, pr_sibling)); 4039 db_printf(" flags = 0x%x", pr->pr_flags); 4040 for (fi = 0; fi < nitems(pr_flag_names); fi++) 4041 if (pr_flag_names[fi] != NULL && (pr->pr_flags & (1 << fi))) 4042 db_printf(" %s", pr_flag_names[fi]); 4043 for (fi = 0; fi < nitems(pr_flag_jailsys); fi++) { 4044 jsf = pr->pr_flags & 4045 (pr_flag_jailsys[fi].disable | pr_flag_jailsys[fi].new); 4046 db_printf(" %-16s= %s\n", pr_flag_jailsys[fi].name, 4047 pr_flag_jailsys[fi].disable && 4048 (jsf == pr_flag_jailsys[fi].disable) ? "disable" 4049 : (jsf == pr_flag_jailsys[fi].new) ? "new" 4050 : "inherit"); 4051 } 4052 db_printf(" allow = 0x%x", pr->pr_allow); 4053 for (fi = 0; fi < nitems(pr_allow_names); fi++) 4054 if (pr_allow_names[fi] != NULL && (pr->pr_allow & (1 << fi))) 4055 db_printf(" %s", pr_allow_names[fi]); 4056 db_printf("\n"); 4057 db_printf(" enforce_statfs = %d\n", pr->pr_enforce_statfs); 4058 db_printf(" host.hostname = %s\n", pr->pr_hostname); 4059 db_printf(" host.domainname = %s\n", pr->pr_domainname); 4060 db_printf(" host.hostuuid = %s\n", pr->pr_hostuuid); 4061 db_printf(" host.hostid = %lu\n", pr->pr_hostid); 4062 #ifdef INET 4063 db_printf(" ip4s = %d\n", pr->pr_ip4s); 4064 for (ii = 0; ii < pr->pr_ip4s; ii++) 4065 db_printf(" %s %s\n", 4066 ii == 0 ? "ip4.addr =" : " ", 4067 inet_ntoa_r(pr->pr_ip4[ii], ip4buf)); 4068 #endif 4069 #ifdef INET6 4070 db_printf(" ip6s = %d\n", pr->pr_ip6s); 4071 for (ii = 0; ii < pr->pr_ip6s; ii++) 4072 db_printf(" %s %s\n", 4073 ii == 0 ? "ip6.addr =" : " ", 4074 ip6_sprintf(ip6buf, &pr->pr_ip6[ii])); 4075 #endif 4076 } 4077 4078 DB_SHOW_COMMAND(prison, db_show_prison_command) 4079 { 4080 struct prison *pr; 4081 4082 if (!have_addr) { 4083 /* 4084 * Show all prisons in the list, and prison0 which is not 4085 * listed. 4086 */ 4087 db_show_prison(&prison0); 4088 if (!db_pager_quit) { 4089 TAILQ_FOREACH(pr, &allprison, pr_list) { 4090 db_show_prison(pr); 4091 if (db_pager_quit) 4092 break; 4093 } 4094 } 4095 return; 4096 } 4097 4098 if (addr == 0) 4099 pr = &prison0; 4100 else { 4101 /* Look for a prison with the ID and with references. */ 4102 TAILQ_FOREACH(pr, &allprison, pr_list) 4103 if (pr->pr_id == addr && pr->pr_ref > 0) 4104 break; 4105 if (pr == NULL) 4106 /* Look again, without requiring a reference. */ 4107 TAILQ_FOREACH(pr, &allprison, pr_list) 4108 if (pr->pr_id == addr) 4109 break; 4110 if (pr == NULL) 4111 /* Assume address points to a valid prison. */ 4112 pr = (struct prison *)addr; 4113 } 4114 db_show_prison(pr); 4115 } 4116 4117 #endif /* DDB */ 4118