kern_jail.c revision 196002
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: head/sys/kern/kern_jail.c 196002 2009-07-31 16:00:41Z jamie $"); 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/sx.h> 53#include <sys/sysent.h> 54#include <sys/namei.h> 55#include <sys/mount.h> 56#include <sys/queue.h> 57#include <sys/socket.h> 58#include <sys/syscallsubr.h> 59#include <sys/sysctl.h> 60#include <sys/vnode.h> 61#include <sys/vimage.h> 62#include <net/if.h> 63#include <netinet/in.h> 64#ifdef DDB 65#include <ddb/ddb.h> 66#ifdef INET6 67#include <netinet6/in6_var.h> 68#endif /* INET6 */ 69#endif /* DDB */ 70 71#include <security/mac/mac_framework.h> 72 73#define DEFAULT_HOSTUUID "00000000-0000-0000-0000-000000000000" 74 75MALLOC_DEFINE(M_PRISON, "prison", "Prison structures"); 76 77/* prison0 describes what is "real" about the system. */ 78struct prison prison0 = { 79 .pr_id = 0, 80 .pr_name = "0", 81 .pr_ref = 1, 82 .pr_uref = 1, 83 .pr_path = "/", 84 .pr_securelevel = -1, 85 .pr_childmax = JAIL_MAX, 86 .pr_hostuuid = DEFAULT_HOSTUUID, 87 .pr_children = LIST_HEAD_INITIALIZER(&prison0.pr_children), 88 .pr_flags = PR_HOST, 89 .pr_allow = PR_ALLOW_ALL, 90}; 91MTX_SYSINIT(prison0, &prison0.pr_mtx, "jail mutex", MTX_DEF); 92 93/* allprison and lastprid are protected by allprison_lock. */ 94struct sx allprison_lock; 95SX_SYSINIT(allprison_lock, &allprison_lock, "allprison"); 96struct prisonlist allprison = TAILQ_HEAD_INITIALIZER(allprison); 97int lastprid = 0; 98 99static int do_jail_attach(struct thread *td, struct prison *pr); 100static void prison_complete(void *context, int pending); 101static void prison_deref(struct prison *pr, int flags); 102static char *prison_path(struct prison *pr1, struct prison *pr2); 103static void prison_remove_one(struct prison *pr); 104#ifdef INET 105static int _prison_check_ip4(struct prison *pr, struct in_addr *ia); 106static int prison_restrict_ip4(struct prison *pr, struct in_addr *newip4); 107#endif 108#ifdef INET6 109static int _prison_check_ip6(struct prison *pr, struct in6_addr *ia6); 110static int prison_restrict_ip6(struct prison *pr, struct in6_addr *newip6); 111#endif 112 113/* Flags for prison_deref */ 114#define PD_DEREF 0x01 115#define PD_DEUREF 0x02 116#define PD_LOCKED 0x04 117#define PD_LIST_SLOCKED 0x08 118#define PD_LIST_XLOCKED 0x10 119 120/* 121 * Parameter names corresponding to PR_* flag values 122 */ 123static char *pr_flag_names[] = { 124 [0] = "persist", 125}; 126 127static char *pr_flag_nonames[] = { 128 [0] = "nopersist", 129}; 130 131struct jailsys_flags { 132 const char *name; 133 unsigned disable; 134 unsigned new; 135} pr_flag_jailsys[] = { 136 { "host", 0, PR_HOST }, 137#ifdef VIMAGE 138 { "vnet", 0, PR_VNET }, 139#endif 140#ifdef INET 141 { "ip4", PR_IP4_USER | PR_IP4_DISABLE, PR_IP4_USER }, 142#endif 143#ifdef INET6 144 { "ip6", PR_IP6_USER | PR_IP6_DISABLE, PR_IP6_USER }, 145#endif 146}; 147 148static char *pr_allow_names[] = { 149 "allow.set_hostname", 150 "allow.sysvipc", 151 "allow.raw_sockets", 152 "allow.chflags", 153 "allow.mount", 154 "allow.quotas", 155 "allow.socket_af", 156}; 157 158static char *pr_allow_nonames[] = { 159 "allow.noset_hostname", 160 "allow.nosysvipc", 161 "allow.noraw_sockets", 162 "allow.nochflags", 163 "allow.nomount", 164 "allow.noquotas", 165 "allow.nosocket_af", 166}; 167 168#define JAIL_DEFAULT_ALLOW PR_ALLOW_SET_HOSTNAME 169#define JAIL_DEFAULT_ENFORCE_STATFS 2 170static unsigned jail_default_allow = JAIL_DEFAULT_ALLOW; 171static int jail_default_enforce_statfs = JAIL_DEFAULT_ENFORCE_STATFS; 172#if defined(INET) || defined(INET6) 173static unsigned jail_max_af_ips = 255; 174#endif 175 176#ifdef INET 177static int 178qcmp_v4(const void *ip1, const void *ip2) 179{ 180 in_addr_t iaa, iab; 181 182 /* 183 * We need to compare in HBO here to get the list sorted as expected 184 * by the result of the code. Sorting NBO addresses gives you 185 * interesting results. If you do not understand, do not try. 186 */ 187 iaa = ntohl(((const struct in_addr *)ip1)->s_addr); 188 iab = ntohl(((const struct in_addr *)ip2)->s_addr); 189 190 /* 191 * Do not simply return the difference of the two numbers, the int is 192 * not wide enough. 193 */ 194 if (iaa > iab) 195 return (1); 196 else if (iaa < iab) 197 return (-1); 198 else 199 return (0); 200} 201#endif 202 203#ifdef INET6 204static int 205qcmp_v6(const void *ip1, const void *ip2) 206{ 207 const struct in6_addr *ia6a, *ia6b; 208 int i, rc; 209 210 ia6a = (const struct in6_addr *)ip1; 211 ia6b = (const struct in6_addr *)ip2; 212 213 rc = 0; 214 for (i = 0; rc == 0 && i < sizeof(struct in6_addr); i++) { 215 if (ia6a->s6_addr[i] > ia6b->s6_addr[i]) 216 rc = 1; 217 else if (ia6a->s6_addr[i] < ia6b->s6_addr[i]) 218 rc = -1; 219 } 220 return (rc); 221} 222#endif 223 224/* 225 * struct jail_args { 226 * struct jail *jail; 227 * }; 228 */ 229int 230jail(struct thread *td, struct jail_args *uap) 231{ 232 uint32_t version; 233 int error; 234 struct jail j; 235 236 error = copyin(uap->jail, &version, sizeof(uint32_t)); 237 if (error) 238 return (error); 239 240 switch (version) { 241 case 0: 242 { 243 struct jail_v0 j0; 244 245 /* FreeBSD single IPv4 jails. */ 246 bzero(&j, sizeof(struct jail)); 247 error = copyin(uap->jail, &j0, sizeof(struct jail_v0)); 248 if (error) 249 return (error); 250 j.version = j0.version; 251 j.path = j0.path; 252 j.hostname = j0.hostname; 253 j.ip4s = j0.ip_number; 254 break; 255 } 256 257 case 1: 258 /* 259 * Version 1 was used by multi-IPv4 jail implementations 260 * that never made it into the official kernel. 261 */ 262 return (EINVAL); 263 264 case 2: /* JAIL_API_VERSION */ 265 /* FreeBSD multi-IPv4/IPv6,noIP jails. */ 266 error = copyin(uap->jail, &j, sizeof(struct jail)); 267 if (error) 268 return (error); 269 break; 270 271 default: 272 /* Sci-Fi jails are not supported, sorry. */ 273 return (EINVAL); 274 } 275 return (kern_jail(td, &j)); 276} 277 278int 279kern_jail(struct thread *td, struct jail *j) 280{ 281 struct iovec optiov[2 * (4 282 + sizeof(pr_allow_names) / sizeof(pr_allow_names[0]) 283#ifdef INET 284 + 1 285#endif 286#ifdef INET6 287 + 1 288#endif 289 )]; 290 struct uio opt; 291 char *u_path, *u_hostname, *u_name; 292#ifdef INET 293 uint32_t ip4s; 294 struct in_addr *u_ip4; 295#endif 296#ifdef INET6 297 struct in6_addr *u_ip6; 298#endif 299 size_t tmplen; 300 int error, enforce_statfs, fi; 301 302 bzero(&optiov, sizeof(optiov)); 303 opt.uio_iov = optiov; 304 opt.uio_iovcnt = 0; 305 opt.uio_offset = -1; 306 opt.uio_resid = -1; 307 opt.uio_segflg = UIO_SYSSPACE; 308 opt.uio_rw = UIO_READ; 309 opt.uio_td = td; 310 311 /* Set permissions for top-level jails from sysctls. */ 312 if (!jailed(td->td_ucred)) { 313 for (fi = 0; fi < sizeof(pr_allow_names) / 314 sizeof(pr_allow_names[0]); fi++) { 315 optiov[opt.uio_iovcnt].iov_base = 316 (jail_default_allow & (1 << fi)) 317 ? pr_allow_names[fi] : pr_allow_nonames[fi]; 318 optiov[opt.uio_iovcnt].iov_len = 319 strlen(optiov[opt.uio_iovcnt].iov_base) + 1; 320 opt.uio_iovcnt += 2; 321 } 322 optiov[opt.uio_iovcnt].iov_base = "enforce_statfs"; 323 optiov[opt.uio_iovcnt].iov_len = sizeof("enforce_statfs"); 324 opt.uio_iovcnt++; 325 enforce_statfs = jail_default_enforce_statfs; 326 optiov[opt.uio_iovcnt].iov_base = &enforce_statfs; 327 optiov[opt.uio_iovcnt].iov_len = sizeof(enforce_statfs); 328 opt.uio_iovcnt++; 329 } 330 331 tmplen = MAXPATHLEN + MAXHOSTNAMELEN + MAXHOSTNAMELEN; 332#ifdef INET 333 ip4s = (j->version == 0) ? 1 : j->ip4s; 334 if (ip4s > jail_max_af_ips) 335 return (EINVAL); 336 tmplen += ip4s * sizeof(struct in_addr); 337#else 338 if (j->ip4s > 0) 339 return (EINVAL); 340#endif 341#ifdef INET6 342 if (j->ip6s > jail_max_af_ips) 343 return (EINVAL); 344 tmplen += j->ip6s * sizeof(struct in6_addr); 345#else 346 if (j->ip6s > 0) 347 return (EINVAL); 348#endif 349 u_path = malloc(tmplen, M_TEMP, M_WAITOK); 350 u_hostname = u_path + MAXPATHLEN; 351 u_name = u_hostname + MAXHOSTNAMELEN; 352#ifdef INET 353 u_ip4 = (struct in_addr *)(u_name + MAXHOSTNAMELEN); 354#endif 355#ifdef INET6 356#ifdef INET 357 u_ip6 = (struct in6_addr *)(u_ip4 + ip4s); 358#else 359 u_ip6 = (struct in6_addr *)(u_name + MAXHOSTNAMELEN); 360#endif 361#endif 362 optiov[opt.uio_iovcnt].iov_base = "path"; 363 optiov[opt.uio_iovcnt].iov_len = sizeof("path"); 364 opt.uio_iovcnt++; 365 optiov[opt.uio_iovcnt].iov_base = u_path; 366 error = copyinstr(j->path, u_path, MAXPATHLEN, 367 &optiov[opt.uio_iovcnt].iov_len); 368 if (error) { 369 free(u_path, M_TEMP); 370 return (error); 371 } 372 opt.uio_iovcnt++; 373 optiov[opt.uio_iovcnt].iov_base = "host.hostname"; 374 optiov[opt.uio_iovcnt].iov_len = sizeof("host.hostname"); 375 opt.uio_iovcnt++; 376 optiov[opt.uio_iovcnt].iov_base = u_hostname; 377 error = copyinstr(j->hostname, u_hostname, MAXHOSTNAMELEN, 378 &optiov[opt.uio_iovcnt].iov_len); 379 if (error) { 380 free(u_path, M_TEMP); 381 return (error); 382 } 383 opt.uio_iovcnt++; 384 if (j->jailname != NULL) { 385 optiov[opt.uio_iovcnt].iov_base = "name"; 386 optiov[opt.uio_iovcnt].iov_len = sizeof("name"); 387 opt.uio_iovcnt++; 388 optiov[opt.uio_iovcnt].iov_base = u_name; 389 error = copyinstr(j->jailname, u_name, MAXHOSTNAMELEN, 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 } 397#ifdef INET 398 optiov[opt.uio_iovcnt].iov_base = "ip4.addr"; 399 optiov[opt.uio_iovcnt].iov_len = sizeof("ip4.addr"); 400 opt.uio_iovcnt++; 401 optiov[opt.uio_iovcnt].iov_base = u_ip4; 402 optiov[opt.uio_iovcnt].iov_len = ip4s * sizeof(struct in_addr); 403 if (j->version == 0) 404 u_ip4->s_addr = j->ip4s; 405 else { 406 error = copyin(j->ip4, u_ip4, optiov[opt.uio_iovcnt].iov_len); 407 if (error) { 408 free(u_path, M_TEMP); 409 return (error); 410 } 411 } 412 opt.uio_iovcnt++; 413#endif 414#ifdef INET6 415 optiov[opt.uio_iovcnt].iov_base = "ip6.addr"; 416 optiov[opt.uio_iovcnt].iov_len = sizeof("ip6.addr"); 417 opt.uio_iovcnt++; 418 optiov[opt.uio_iovcnt].iov_base = u_ip6; 419 optiov[opt.uio_iovcnt].iov_len = j->ip6s * sizeof(struct in6_addr); 420 error = copyin(j->ip6, u_ip6, optiov[opt.uio_iovcnt].iov_len); 421 if (error) { 422 free(u_path, M_TEMP); 423 return (error); 424 } 425 opt.uio_iovcnt++; 426#endif 427 KASSERT(opt.uio_iovcnt <= sizeof(optiov) / sizeof(optiov[0]), 428 ("kern_jail: too many iovecs (%d)", opt.uio_iovcnt)); 429 error = kern_jail_set(td, &opt, JAIL_CREATE | JAIL_ATTACH); 430 free(u_path, M_TEMP); 431 return (error); 432} 433 434 435/* 436 * struct jail_set_args { 437 * struct iovec *iovp; 438 * unsigned int iovcnt; 439 * int flags; 440 * }; 441 */ 442int 443jail_set(struct thread *td, struct jail_set_args *uap) 444{ 445 struct uio *auio; 446 int error; 447 448 /* Check that we have an even number of iovecs. */ 449 if (uap->iovcnt & 1) 450 return (EINVAL); 451 452 error = copyinuio(uap->iovp, uap->iovcnt, &auio); 453 if (error) 454 return (error); 455 error = kern_jail_set(td, auio, uap->flags); 456 free(auio, M_IOV); 457 return (error); 458} 459 460int 461kern_jail_set(struct thread *td, struct uio *optuio, int flags) 462{ 463 struct nameidata nd; 464#ifdef INET 465 struct in_addr *ip4; 466#endif 467#ifdef INET6 468 struct in6_addr *ip6; 469#endif 470 struct vfsopt *opt; 471 struct vfsoptlist *opts; 472 struct prison *pr, *deadpr, *mypr, *ppr, *tpr, *tppr; 473 struct vnode *root; 474 char *domain, *errmsg, *host, *name, *p, *path, *uuid; 475#if defined(INET) || defined(INET6) 476 void *op; 477#endif 478 unsigned long hid; 479 size_t namelen, onamelen; 480 int created, cuflags, descend, enforce, error, errmsg_len, errmsg_pos; 481 int gotchildmax, gotenforce, gothid, gotslevel; 482 int fi, jid, jsys, len, level; 483 int childmax, slevel, vfslocked; 484#if defined(INET) || defined(INET6) 485 int ii, ij; 486#endif 487#ifdef INET 488 int ip4s, redo_ip4; 489#endif 490#ifdef INET6 491 int ip6s, redo_ip6; 492#endif 493 unsigned pr_flags, ch_flags; 494 unsigned pr_allow, ch_allow, tallow; 495 char numbuf[12]; 496 497 error = priv_check(td, PRIV_JAIL_SET); 498 if (!error && (flags & JAIL_ATTACH)) 499 error = priv_check(td, PRIV_JAIL_ATTACH); 500 if (error) 501 return (error); 502 mypr = ppr = td->td_ucred->cr_prison; 503 if ((flags & JAIL_CREATE) && mypr->pr_childmax == 0) 504 return (EPERM); 505 if (flags & ~JAIL_SET_MASK) 506 return (EINVAL); 507 508 /* 509 * Check all the parameters before committing to anything. Not all 510 * errors can be caught early, but we may as well try. Also, this 511 * takes care of some expensive stuff (path lookup) before getting 512 * the allprison lock. 513 * 514 * XXX Jails are not filesystems, and jail parameters are not mount 515 * options. But it makes more sense to re-use the vfsopt code 516 * than duplicate it under a different name. 517 */ 518 error = vfs_buildopts(optuio, &opts); 519 if (error) 520 return (error); 521#ifdef INET 522 ip4 = NULL; 523#endif 524#ifdef INET6 525 ip6 = NULL; 526#endif 527 528 error = vfs_copyopt(opts, "jid", &jid, sizeof(jid)); 529 if (error == ENOENT) 530 jid = 0; 531 else if (error != 0) 532 goto done_free; 533 534 error = vfs_copyopt(opts, "securelevel", &slevel, sizeof(slevel)); 535 if (error == ENOENT) 536 gotslevel = 0; 537 else if (error != 0) 538 goto done_free; 539 else 540 gotslevel = 1; 541 542 error = 543 vfs_copyopt(opts, "children.max", &childmax, sizeof(childmax)); 544 if (error == ENOENT) 545 gotchildmax = 0; 546 else if (error != 0) 547 goto done_free; 548 else 549 gotchildmax = 1; 550 551 error = vfs_copyopt(opts, "enforce_statfs", &enforce, sizeof(enforce)); 552 gotenforce = (error == 0); 553 if (gotenforce) { 554 if (enforce < 0 || enforce > 2) 555 return (EINVAL); 556 } else if (error != ENOENT) 557 goto done_free; 558 559 pr_flags = ch_flags = 0; 560 for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]); 561 fi++) { 562 if (pr_flag_names[fi] == NULL) 563 continue; 564 vfs_flagopt(opts, pr_flag_names[fi], &pr_flags, 1 << fi); 565 vfs_flagopt(opts, pr_flag_nonames[fi], &ch_flags, 1 << fi); 566 } 567 ch_flags |= pr_flags; 568 for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]); 569 fi++) { 570 error = vfs_copyopt(opts, pr_flag_jailsys[fi].name, &jsys, 571 sizeof(jsys)); 572 if (error == ENOENT) 573 continue; 574 if (error != 0) 575 goto done_free; 576 switch (jsys) { 577 case JAIL_SYS_DISABLE: 578 if (!pr_flag_jailsys[fi].disable) { 579 error = EINVAL; 580 goto done_free; 581 } 582 pr_flags |= pr_flag_jailsys[fi].disable; 583 break; 584 case JAIL_SYS_NEW: 585 pr_flags |= pr_flag_jailsys[fi].new; 586 break; 587 case JAIL_SYS_INHERIT: 588 break; 589 default: 590 error = EINVAL; 591 goto done_free; 592 } 593 ch_flags |= 594 pr_flag_jailsys[fi].new | pr_flag_jailsys[fi].disable; 595 } 596 if ((flags & (JAIL_CREATE | JAIL_UPDATE | JAIL_ATTACH)) == JAIL_CREATE 597 && !(pr_flags & PR_PERSIST)) { 598 error = EINVAL; 599 vfs_opterror(opts, "new jail must persist or attach"); 600 goto done_errmsg; 601 } 602#ifdef VIMAGE 603 if ((flags & JAIL_UPDATE) && (ch_flags & PR_VNET)) { 604 error = EINVAL; 605 vfs_opterror(opts, "vnet cannot be changed after creation"); 606 goto done_errmsg; 607 } 608#endif 609#ifdef INET 610 if ((flags & JAIL_UPDATE) && (ch_flags & PR_IP4_USER)) { 611 error = EINVAL; 612 vfs_opterror(opts, "ip4 cannot be changed after creation"); 613 goto done_errmsg; 614 } 615#endif 616#ifdef INET6 617 if ((flags & JAIL_UPDATE) && (ch_flags & PR_IP6_USER)) { 618 error = EINVAL; 619 vfs_opterror(opts, "ip6 cannot be changed after creation"); 620 goto done_errmsg; 621 } 622#endif 623 624 pr_allow = ch_allow = 0; 625 for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]); 626 fi++) { 627 vfs_flagopt(opts, pr_allow_names[fi], &pr_allow, 1 << fi); 628 vfs_flagopt(opts, pr_allow_nonames[fi], &ch_allow, 1 << fi); 629 } 630 ch_allow |= pr_allow; 631 632 error = vfs_getopt(opts, "name", (void **)&name, &len); 633 if (error == ENOENT) 634 name = NULL; 635 else if (error != 0) 636 goto done_free; 637 else { 638 if (len == 0 || name[len - 1] != '\0') { 639 error = EINVAL; 640 goto done_free; 641 } 642 if (len > MAXHOSTNAMELEN) { 643 error = ENAMETOOLONG; 644 goto done_free; 645 } 646 } 647 648 error = vfs_getopt(opts, "host.hostname", (void **)&host, &len); 649 if (error == ENOENT) 650 host = NULL; 651 else if (error != 0) 652 goto done_free; 653 else { 654 ch_flags |= PR_HOST; 655 pr_flags |= PR_HOST; 656 if (len == 0 || host[len - 1] != '\0') { 657 error = EINVAL; 658 goto done_free; 659 } 660 if (len > MAXHOSTNAMELEN) { 661 error = ENAMETOOLONG; 662 goto done_free; 663 } 664 } 665 666 error = vfs_getopt(opts, "host.domainname", (void **)&domain, &len); 667 if (error == ENOENT) 668 domain = NULL; 669 else if (error != 0) 670 goto done_free; 671 else { 672 ch_flags |= PR_HOST; 673 pr_flags |= PR_HOST; 674 if (len == 0 || domain[len - 1] != '\0') { 675 error = EINVAL; 676 goto done_free; 677 } 678 if (len > MAXHOSTNAMELEN) { 679 error = ENAMETOOLONG; 680 goto done_free; 681 } 682 } 683 684 error = vfs_getopt(opts, "host.hostuuid", (void **)&uuid, &len); 685 if (error == ENOENT) 686 uuid = NULL; 687 else if (error != 0) 688 goto done_free; 689 else { 690 ch_flags |= PR_HOST; 691 pr_flags |= PR_HOST; 692 if (len == 0 || uuid[len - 1] != '\0') { 693 error = EINVAL; 694 goto done_free; 695 } 696 if (len > HOSTUUIDLEN) { 697 error = ENAMETOOLONG; 698 goto done_free; 699 } 700 } 701 702#ifdef COMPAT_IA32 703 if (td->td_proc->p_sysent->sv_flags & SV_IA32) { 704 uint32_t hid32; 705 706 error = vfs_copyopt(opts, "host.hostid", &hid32, sizeof(hid32)); 707 hid = hid32; 708 } else 709#endif 710 error = vfs_copyopt(opts, "host.hostid", &hid, sizeof(hid)); 711 if (error == ENOENT) 712 gothid = 0; 713 else if (error != 0) 714 goto done_free; 715 else { 716 gothid = 1; 717 ch_flags |= PR_HOST; 718 pr_flags |= PR_HOST; 719 } 720 721#ifdef INET 722 error = vfs_getopt(opts, "ip4.addr", &op, &ip4s); 723 if (error == ENOENT) 724 ip4s = (pr_flags & PR_IP4_DISABLE) ? 0 : -1; 725 else if (error != 0) 726 goto done_free; 727 else if (ip4s & (sizeof(*ip4) - 1)) { 728 error = EINVAL; 729 goto done_free; 730 } else { 731 ch_flags |= PR_IP4_USER | PR_IP4_DISABLE; 732 if (ip4s == 0) 733 pr_flags |= PR_IP4_USER | PR_IP4_DISABLE; 734 else { 735 pr_flags = (pr_flags & ~PR_IP4_DISABLE) | PR_IP4_USER; 736 ip4s /= sizeof(*ip4); 737 if (ip4s > jail_max_af_ips) { 738 error = EINVAL; 739 vfs_opterror(opts, "too many IPv4 addresses"); 740 goto done_errmsg; 741 } 742 ip4 = malloc(ip4s * sizeof(*ip4), M_PRISON, M_WAITOK); 743 bcopy(op, ip4, ip4s * sizeof(*ip4)); 744 /* 745 * IP addresses are all sorted but ip[0] to preserve 746 * the primary IP address as given from userland. 747 * This special IP is used for unbound outgoing 748 * connections as well for "loopback" traffic. 749 */ 750 if (ip4s > 1) 751 qsort(ip4 + 1, ip4s - 1, sizeof(*ip4), qcmp_v4); 752 /* 753 * Check for duplicate addresses and do some simple 754 * zero and broadcast checks. If users give other bogus 755 * addresses it is their problem. 756 * 757 * We do not have to care about byte order for these 758 * checks so we will do them in NBO. 759 */ 760 for (ii = 0; ii < ip4s; ii++) { 761 if (ip4[ii].s_addr == INADDR_ANY || 762 ip4[ii].s_addr == INADDR_BROADCAST) { 763 error = EINVAL; 764 goto done_free; 765 } 766 if ((ii+1) < ip4s && 767 (ip4[0].s_addr == ip4[ii+1].s_addr || 768 ip4[ii].s_addr == ip4[ii+1].s_addr)) { 769 error = EINVAL; 770 goto done_free; 771 } 772 } 773 } 774 } 775#endif 776 777#ifdef INET6 778 error = vfs_getopt(opts, "ip6.addr", &op, &ip6s); 779 if (error == ENOENT) 780 ip6s = (pr_flags & PR_IP6_DISABLE) ? 0 : -1; 781 else if (error != 0) 782 goto done_free; 783 else if (ip6s & (sizeof(*ip6) - 1)) { 784 error = EINVAL; 785 goto done_free; 786 } else { 787 ch_flags |= PR_IP6_USER | PR_IP6_DISABLE; 788 if (ip6s == 0) 789 pr_flags |= PR_IP6_USER | PR_IP6_DISABLE; 790 else { 791 pr_flags = (pr_flags & ~PR_IP6_DISABLE) | PR_IP6_USER; 792 ip6s /= sizeof(*ip6); 793 if (ip6s > jail_max_af_ips) { 794 error = EINVAL; 795 vfs_opterror(opts, "too many IPv6 addresses"); 796 goto done_errmsg; 797 } 798 ip6 = malloc(ip6s * sizeof(*ip6), M_PRISON, M_WAITOK); 799 bcopy(op, ip6, ip6s * sizeof(*ip6)); 800 if (ip6s > 1) 801 qsort(ip6 + 1, ip6s - 1, sizeof(*ip6), qcmp_v6); 802 for (ii = 0; ii < ip6s; ii++) { 803 if (IN6_IS_ADDR_UNSPECIFIED(&ip6[ii])) { 804 error = EINVAL; 805 goto done_free; 806 } 807 if ((ii+1) < ip6s && 808 (IN6_ARE_ADDR_EQUAL(&ip6[0], &ip6[ii+1]) || 809 IN6_ARE_ADDR_EQUAL(&ip6[ii], &ip6[ii+1]))) 810 { 811 error = EINVAL; 812 goto done_free; 813 } 814 } 815 } 816 } 817#endif 818 819#if defined(VIMAGE) && (defined(INET) || defined(INET6)) 820 if ((ch_flags & PR_VNET) && (ch_flags & (PR_IP4_USER | PR_IP6_USER))) { 821 error = EINVAL; 822 vfs_opterror(opts, 823 "vnet jails cannot have IP address restrictions"); 824 goto done_errmsg; 825 } 826#endif 827 828 root = NULL; 829 error = vfs_getopt(opts, "path", (void **)&path, &len); 830 if (error == ENOENT) 831 path = NULL; 832 else if (error != 0) 833 goto done_free; 834 else { 835 if (flags & JAIL_UPDATE) { 836 error = EINVAL; 837 vfs_opterror(opts, 838 "path cannot be changed after creation"); 839 goto done_errmsg; 840 } 841 if (len == 0 || path[len - 1] != '\0') { 842 error = EINVAL; 843 goto done_free; 844 } 845 if (len < 2 || (len == 2 && path[0] == '/')) 846 path = NULL; 847 else { 848 /* Leave room for a real-root full pathname. */ 849 if (len + (path[0] == '/' && strcmp(mypr->pr_path, "/") 850 ? strlen(mypr->pr_path) : 0) > MAXPATHLEN) { 851 error = ENAMETOOLONG; 852 goto done_free; 853 } 854 NDINIT(&nd, LOOKUP, MPSAFE | FOLLOW, UIO_SYSSPACE, 855 path, td); 856 error = namei(&nd); 857 if (error) 858 goto done_free; 859 vfslocked = NDHASGIANT(&nd); 860 root = nd.ni_vp; 861 NDFREE(&nd, NDF_ONLY_PNBUF); 862 if (root->v_type != VDIR) { 863 error = ENOTDIR; 864 vrele(root); 865 VFS_UNLOCK_GIANT(vfslocked); 866 goto done_free; 867 } 868 VFS_UNLOCK_GIANT(vfslocked); 869 } 870 } 871 872 /* 873 * Grab the allprison lock before letting modules check their 874 * parameters. Once we have it, do not let go so we'll have a 875 * consistent view of the OSD list. 876 */ 877 sx_xlock(&allprison_lock); 878 error = osd_jail_call(NULL, PR_METHOD_CHECK, opts); 879 if (error) 880 goto done_unlock_list; 881 882 /* By now, all parameters should have been noted. */ 883 TAILQ_FOREACH(opt, opts, link) { 884 if (!opt->seen && strcmp(opt->name, "errmsg")) { 885 error = EINVAL; 886 vfs_opterror(opts, "unknown parameter: %s", opt->name); 887 goto done_unlock_list; 888 } 889 } 890 891 /* 892 * See if we are creating a new record or updating an existing one. 893 * This abuses the file error codes ENOENT and EEXIST. 894 */ 895 cuflags = flags & (JAIL_CREATE | JAIL_UPDATE); 896 if (!cuflags) { 897 error = EINVAL; 898 vfs_opterror(opts, "no valid operation (create or update)"); 899 goto done_unlock_list; 900 } 901 pr = NULL; 902 if (jid != 0) { 903 /* 904 * See if a requested jid already exists. There is an 905 * information leak here if the jid exists but is not within 906 * the caller's jail hierarchy. Jail creators will get EEXIST 907 * even though they cannot see the jail, and CREATE | UPDATE 908 * will return ENOENT which is not normally a valid error. 909 */ 910 if (jid < 0) { 911 error = EINVAL; 912 vfs_opterror(opts, "negative jid"); 913 goto done_unlock_list; 914 } 915 pr = prison_find(jid); 916 if (pr != NULL) { 917 ppr = pr->pr_parent; 918 /* Create: jid must not exist. */ 919 if (cuflags == JAIL_CREATE) { 920 mtx_unlock(&pr->pr_mtx); 921 error = EEXIST; 922 vfs_opterror(opts, "jail %d already exists", 923 jid); 924 goto done_unlock_list; 925 } 926 if (!prison_ischild(mypr, pr)) { 927 mtx_unlock(&pr->pr_mtx); 928 pr = NULL; 929 } else if (pr->pr_uref == 0) { 930 if (!(flags & JAIL_DYING)) { 931 mtx_unlock(&pr->pr_mtx); 932 error = ENOENT; 933 vfs_opterror(opts, "jail %d is dying", 934 jid); 935 goto done_unlock_list; 936 } else if ((flags & JAIL_ATTACH) || 937 (pr_flags & PR_PERSIST)) { 938 /* 939 * A dying jail might be resurrected 940 * (via attach or persist), but first 941 * it must determine if another jail 942 * has claimed its name. Accomplish 943 * this by implicitly re-setting the 944 * name. 945 */ 946 if (name == NULL) 947 name = prison_name(mypr, pr); 948 } 949 } 950 } 951 if (pr == NULL) { 952 /* Update: jid must exist. */ 953 if (cuflags == JAIL_UPDATE) { 954 error = ENOENT; 955 vfs_opterror(opts, "jail %d not found", jid); 956 goto done_unlock_list; 957 } 958 } 959 } 960 /* 961 * If the caller provided a name, look for a jail by that name. 962 * This has different semantics for creates and updates keyed by jid 963 * (where the name must not already exist in a different jail), 964 * and updates keyed by the name itself (where the name must exist 965 * because that is the jail being updated). 966 */ 967 if (name != NULL) { 968 p = strrchr(name, '.'); 969 if (p != NULL) { 970 /* 971 * This is a hierarchical name. Split it into the 972 * parent and child names, and make sure the parent 973 * exists or matches an already found jail. 974 */ 975 *p = '\0'; 976 if (pr != NULL) { 977 if (strncmp(name, ppr->pr_name, p - name) || 978 ppr->pr_name[p - name] != '\0') { 979 mtx_unlock(&pr->pr_mtx); 980 error = EINVAL; 981 vfs_opterror(opts, 982 "cannot change jail's parent"); 983 goto done_unlock_list; 984 } 985 } else { 986 ppr = prison_find_name(mypr, name); 987 if (ppr == NULL) { 988 error = ENOENT; 989 vfs_opterror(opts, 990 "jail \"%s\" not found", name); 991 goto done_unlock_list; 992 } 993 mtx_unlock(&ppr->pr_mtx); 994 } 995 name = p + 1; 996 } 997 if (name[0] != '\0') { 998 namelen = 999 (ppr == &prison0) ? 0 : strlen(ppr->pr_name) + 1; 1000 name_again: 1001 deadpr = NULL; 1002 FOREACH_PRISON_CHILD(ppr, tpr) { 1003 if (tpr != pr && tpr->pr_ref > 0 && 1004 !strcmp(tpr->pr_name + namelen, name)) { 1005 if (pr == NULL && 1006 cuflags != JAIL_CREATE) { 1007 mtx_lock(&tpr->pr_mtx); 1008 if (tpr->pr_ref > 0) { 1009 /* 1010 * Use this jail 1011 * for updates. 1012 */ 1013 if (tpr->pr_uref > 0) { 1014 pr = tpr; 1015 break; 1016 } 1017 deadpr = tpr; 1018 } 1019 mtx_unlock(&tpr->pr_mtx); 1020 } else if (tpr->pr_uref > 0) { 1021 /* 1022 * Create, or update(jid): 1023 * name must not exist in an 1024 * active sibling jail. 1025 */ 1026 error = EEXIST; 1027 if (pr != NULL) 1028 mtx_unlock(&pr->pr_mtx); 1029 vfs_opterror(opts, 1030 "jail \"%s\" already exists", 1031 name); 1032 goto done_unlock_list; 1033 } 1034 } 1035 } 1036 /* If no active jail is found, use a dying one. */ 1037 if (deadpr != NULL && pr == NULL) { 1038 if (flags & JAIL_DYING) { 1039 mtx_lock(&deadpr->pr_mtx); 1040 if (deadpr->pr_ref == 0) { 1041 mtx_unlock(&deadpr->pr_mtx); 1042 goto name_again; 1043 } 1044 pr = deadpr; 1045 } else if (cuflags == JAIL_UPDATE) { 1046 error = ENOENT; 1047 vfs_opterror(opts, 1048 "jail \"%s\" is dying", name); 1049 goto done_unlock_list; 1050 } 1051 } 1052 /* Update: name must exist if no jid. */ 1053 else if (cuflags == JAIL_UPDATE && pr == NULL) { 1054 error = ENOENT; 1055 vfs_opterror(opts, "jail \"%s\" not found", 1056 name); 1057 goto done_unlock_list; 1058 } 1059 } 1060 } 1061 /* Update: must provide a jid or name. */ 1062 else if (cuflags == JAIL_UPDATE && pr == NULL) { 1063 error = ENOENT; 1064 vfs_opterror(opts, "update specified no jail"); 1065 goto done_unlock_list; 1066 } 1067 1068 /* If there's no prison to update, create a new one and link it in. */ 1069 if (pr == NULL) { 1070 for (tpr = mypr; tpr != NULL; tpr = tpr->pr_parent) 1071 if (tpr->pr_childcount >= tpr->pr_childmax) { 1072 error = EPERM; 1073 vfs_opterror(opts, "prison limit exceeded"); 1074 goto done_unlock_list; 1075 } 1076 created = 1; 1077 mtx_lock(&ppr->pr_mtx); 1078 if (ppr->pr_ref == 0 || (ppr->pr_flags & PR_REMOVE)) { 1079 mtx_unlock(&ppr->pr_mtx); 1080 error = ENOENT; 1081 vfs_opterror(opts, "parent jail went away!"); 1082 goto done_unlock_list; 1083 } 1084 ppr->pr_ref++; 1085 ppr->pr_uref++; 1086 mtx_unlock(&ppr->pr_mtx); 1087 pr = malloc(sizeof(*pr), M_PRISON, M_WAITOK | M_ZERO); 1088 if (jid == 0) { 1089 /* Find the next free jid. */ 1090 jid = lastprid + 1; 1091 findnext: 1092 if (jid == JAIL_MAX) 1093 jid = 1; 1094 TAILQ_FOREACH(tpr, &allprison, pr_list) { 1095 if (tpr->pr_id < jid) 1096 continue; 1097 if (tpr->pr_id > jid || tpr->pr_ref == 0) { 1098 TAILQ_INSERT_BEFORE(tpr, pr, pr_list); 1099 break; 1100 } 1101 if (jid == lastprid) { 1102 error = EAGAIN; 1103 vfs_opterror(opts, 1104 "no available jail IDs"); 1105 free(pr, M_PRISON); 1106 prison_deref(ppr, PD_DEREF | 1107 PD_DEUREF | PD_LIST_XLOCKED); 1108 goto done_releroot; 1109 } 1110 jid++; 1111 goto findnext; 1112 } 1113 lastprid = jid; 1114 } else { 1115 /* 1116 * The jail already has a jid (that did not yet exist), 1117 * so just find where to insert it. 1118 */ 1119 TAILQ_FOREACH(tpr, &allprison, pr_list) 1120 if (tpr->pr_id >= jid) { 1121 TAILQ_INSERT_BEFORE(tpr, pr, pr_list); 1122 break; 1123 } 1124 } 1125 if (tpr == NULL) 1126 TAILQ_INSERT_TAIL(&allprison, pr, pr_list); 1127 LIST_INSERT_HEAD(&ppr->pr_children, pr, pr_sibling); 1128 for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent) 1129 tpr->pr_childcount++; 1130 1131 pr->pr_parent = ppr; 1132 pr->pr_id = jid; 1133 1134 /* Set some default values, and inherit some from the parent. */ 1135 if (name == NULL) 1136 name = ""; 1137 if (path == NULL) { 1138 path = "/"; 1139 root = mypr->pr_root; 1140 vref(root); 1141 } 1142 strlcpy(pr->pr_hostuuid, DEFAULT_HOSTUUID, HOSTUUIDLEN); 1143 pr->pr_flags |= PR_HOST; 1144#if defined(INET) || defined(INET6) 1145#ifdef VIMAGE 1146 if (!(pr_flags & PR_VNET)) 1147#endif 1148 { 1149#ifdef INET 1150 if (!(ch_flags & PR_IP4_USER)) 1151 pr->pr_flags |= 1152 PR_IP4 | PR_IP4_USER | PR_IP4_DISABLE; 1153 else if (!(pr_flags & PR_IP4_USER)) { 1154 pr->pr_flags |= ppr->pr_flags & PR_IP4; 1155 if (ppr->pr_ip4 != NULL) { 1156 pr->pr_ip4s = ppr->pr_ip4s; 1157 pr->pr_ip4 = malloc(pr->pr_ip4s * 1158 sizeof(struct in_addr), M_PRISON, 1159 M_WAITOK); 1160 bcopy(ppr->pr_ip4, pr->pr_ip4, 1161 pr->pr_ip4s * sizeof(*pr->pr_ip4)); 1162 } 1163 } 1164#endif 1165#ifdef INET6 1166 if (!(ch_flags & PR_IP6_USER)) 1167 pr->pr_flags |= 1168 PR_IP6 | PR_IP6_USER | PR_IP6_DISABLE; 1169 else if (!(pr_flags & PR_IP6_USER)) { 1170 pr->pr_flags |= ppr->pr_flags & PR_IP6; 1171 if (ppr->pr_ip6 != NULL) { 1172 pr->pr_ip6s = ppr->pr_ip6s; 1173 pr->pr_ip6 = malloc(pr->pr_ip6s * 1174 sizeof(struct in6_addr), M_PRISON, 1175 M_WAITOK); 1176 bcopy(ppr->pr_ip6, pr->pr_ip6, 1177 pr->pr_ip6s * sizeof(*pr->pr_ip6)); 1178 } 1179 } 1180#endif 1181 } 1182#endif 1183 pr->pr_securelevel = ppr->pr_securelevel; 1184 pr->pr_allow = JAIL_DEFAULT_ALLOW & ppr->pr_allow; 1185 pr->pr_enforce_statfs = JAIL_DEFAULT_ENFORCE_STATFS; 1186 1187 LIST_INIT(&pr->pr_children); 1188 mtx_init(&pr->pr_mtx, "jail mutex", NULL, MTX_DEF | MTX_DUPOK); 1189 1190#ifdef VIMAGE 1191 /* Allocate a new vnet if specified. */ 1192 pr->pr_vnet = (pr_flags & PR_VNET) 1193 ? vnet_alloc() : ppr->pr_vnet; 1194#endif 1195 /* 1196 * Allocate a dedicated cpuset for each jail. 1197 * Unlike other initial settings, this may return an erorr. 1198 */ 1199 error = cpuset_create_root(ppr, &pr->pr_cpuset); 1200 if (error) { 1201 prison_deref(pr, PD_LIST_XLOCKED); 1202 goto done_releroot; 1203 } 1204 1205 mtx_lock(&pr->pr_mtx); 1206 /* 1207 * New prisons do not yet have a reference, because we do not 1208 * want other to see the incomplete prison once the 1209 * allprison_lock is downgraded. 1210 */ 1211 } else { 1212 created = 0; 1213 /* 1214 * Grab a reference for existing prisons, to ensure they 1215 * continue to exist for the duration of the call. 1216 */ 1217 pr->pr_ref++; 1218#if defined(VIMAGE) && (defined(INET) || defined(INET6)) 1219 if ((pr->pr_flags & PR_VNET) && 1220 (ch_flags & (PR_IP4_USER | PR_IP6_USER))) { 1221 error = EINVAL; 1222 vfs_opterror(opts, 1223 "vnet jails cannot have IP address restrictions"); 1224 goto done_deref_locked; 1225 } 1226#endif 1227#ifdef INET 1228 if (PR_IP4_USER & ch_flags & (pr_flags ^ pr->pr_flags)) { 1229 error = EINVAL; 1230 vfs_opterror(opts, 1231 "ip4 cannot be changed after creation"); 1232 goto done_deref_locked; 1233 } 1234#endif 1235#ifdef INET6 1236 if (PR_IP6_USER & ch_flags & (pr_flags ^ pr->pr_flags)) { 1237 error = EINVAL; 1238 vfs_opterror(opts, 1239 "ip6 cannot be changed after creation"); 1240 goto done_deref_locked; 1241 } 1242#endif 1243 } 1244 1245 /* Do final error checking before setting anything. */ 1246 if (gotslevel) { 1247 if (slevel < ppr->pr_securelevel) { 1248 error = EPERM; 1249 goto done_deref_locked; 1250 } 1251 } 1252 if (gotchildmax) { 1253 if (childmax >= ppr->pr_childmax) { 1254 error = EPERM; 1255 goto done_deref_locked; 1256 } 1257 } 1258 if (gotenforce) { 1259 if (enforce < ppr->pr_enforce_statfs) { 1260 error = EPERM; 1261 goto done_deref_locked; 1262 } 1263 } 1264#ifdef INET 1265 if (ip4s > 0) { 1266 if (ppr->pr_flags & PR_IP4) { 1267 /* 1268 * Make sure the new set of IP addresses is a 1269 * subset of the parent's list. Don't worry 1270 * about the parent being unlocked, as any 1271 * setting is done with allprison_lock held. 1272 */ 1273 for (ij = 0; ij < ppr->pr_ip4s; ij++) 1274 if (ip4[0].s_addr == ppr->pr_ip4[ij].s_addr) 1275 break; 1276 if (ij == ppr->pr_ip4s) { 1277 error = EPERM; 1278 goto done_deref_locked; 1279 } 1280 if (ip4s > 1) { 1281 for (ii = ij = 1; ii < ip4s; ii++) { 1282 if (ip4[ii].s_addr == 1283 ppr->pr_ip4[0].s_addr) 1284 continue; 1285 for (; ij < ppr->pr_ip4s; ij++) 1286 if (ip4[ii].s_addr == 1287 ppr->pr_ip4[ij].s_addr) 1288 break; 1289 if (ij == ppr->pr_ip4s) 1290 break; 1291 } 1292 if (ij == ppr->pr_ip4s) { 1293 error = EPERM; 1294 goto done_deref_locked; 1295 } 1296 } 1297 } 1298 /* 1299 * Check for conflicting IP addresses. We permit them 1300 * if there is no more than one IP on each jail. If 1301 * there is a duplicate on a jail with more than one 1302 * IP stop checking and return error. 1303 */ 1304 tppr = ppr; 1305#ifdef VIMAGE 1306 for (; tppr != &prison0; tppr = tppr->pr_parent) 1307 if (tppr->pr_flags & PR_VNET) 1308 break; 1309#endif 1310 FOREACH_PRISON_DESCENDANT(tppr, tpr, descend) { 1311 if (tpr == pr || 1312#ifdef VIMAGE 1313 (tpr != tppr && (tpr->pr_flags & PR_VNET)) || 1314#endif 1315 tpr->pr_uref == 0) { 1316 descend = 0; 1317 continue; 1318 } 1319 if (!(tpr->pr_flags & PR_IP4_USER)) 1320 continue; 1321 descend = 0; 1322 if (tpr->pr_ip4 == NULL || 1323 (ip4s == 1 && tpr->pr_ip4s == 1)) 1324 continue; 1325 for (ii = 0; ii < ip4s; ii++) { 1326 if (_prison_check_ip4(tpr, &ip4[ii]) == 0) { 1327 error = EADDRINUSE; 1328 vfs_opterror(opts, 1329 "IPv4 addresses clash"); 1330 goto done_deref_locked; 1331 } 1332 } 1333 } 1334 } 1335#endif 1336#ifdef INET6 1337 if (ip6s > 0) { 1338 if (ppr->pr_flags & PR_IP6) { 1339 /* 1340 * Make sure the new set of IP addresses is a 1341 * subset of the parent's list. 1342 */ 1343 for (ij = 0; ij < ppr->pr_ip6s; ij++) 1344 if (IN6_ARE_ADDR_EQUAL(&ip6[0], 1345 &ppr->pr_ip6[ij])) 1346 break; 1347 if (ij == ppr->pr_ip6s) { 1348 error = EPERM; 1349 goto done_deref_locked; 1350 } 1351 if (ip6s > 1) { 1352 for (ii = ij = 1; ii < ip6s; ii++) { 1353 if (IN6_ARE_ADDR_EQUAL(&ip6[ii], 1354 &ppr->pr_ip6[0])) 1355 continue; 1356 for (; ij < ppr->pr_ip6s; ij++) 1357 if (IN6_ARE_ADDR_EQUAL( 1358 &ip6[ii], &ppr->pr_ip6[ij])) 1359 break; 1360 if (ij == ppr->pr_ip6s) 1361 break; 1362 } 1363 if (ij == ppr->pr_ip6s) { 1364 error = EPERM; 1365 goto done_deref_locked; 1366 } 1367 } 1368 } 1369 /* Check for conflicting IP addresses. */ 1370 tppr = ppr; 1371#ifdef VIMAGE 1372 for (; tppr != &prison0; tppr = tppr->pr_parent) 1373 if (tppr->pr_flags & PR_VNET) 1374 break; 1375#endif 1376 FOREACH_PRISON_DESCENDANT(tppr, tpr, descend) { 1377 if (tpr == pr || 1378#ifdef VIMAGE 1379 (tpr != tppr && (tpr->pr_flags & PR_VNET)) || 1380#endif 1381 tpr->pr_uref == 0) { 1382 descend = 0; 1383 continue; 1384 } 1385 if (!(tpr->pr_flags & PR_IP6_USER)) 1386 continue; 1387 descend = 0; 1388 if (tpr->pr_ip6 == NULL || 1389 (ip6s == 1 && tpr->pr_ip6s == 1)) 1390 continue; 1391 for (ii = 0; ii < ip6s; ii++) { 1392 if (_prison_check_ip6(tpr, &ip6[ii]) == 0) { 1393 error = EADDRINUSE; 1394 vfs_opterror(opts, 1395 "IPv6 addresses clash"); 1396 goto done_deref_locked; 1397 } 1398 } 1399 } 1400 } 1401#endif 1402 onamelen = namelen = 0; 1403 if (name != NULL) { 1404 /* Give a default name of the jid. */ 1405 if (name[0] == '\0') 1406 snprintf(name = numbuf, sizeof(numbuf), "%d", jid); 1407 else if (strtoul(name, &p, 10) != jid && *p == '\0') { 1408 error = EINVAL; 1409 vfs_opterror(opts, "name cannot be numeric"); 1410 goto done_deref_locked; 1411 } 1412 /* 1413 * Make sure the name isn't too long for the prison or its 1414 * children. 1415 */ 1416 onamelen = strlen(pr->pr_name); 1417 namelen = strlen(name); 1418 if (strlen(ppr->pr_name) + namelen + 2 > sizeof(pr->pr_name)) { 1419 error = ENAMETOOLONG; 1420 goto done_deref_locked; 1421 } 1422 FOREACH_PRISON_DESCENDANT(pr, tpr, descend) { 1423 if (strlen(tpr->pr_name) + (namelen - onamelen) >= 1424 sizeof(pr->pr_name)) { 1425 error = ENAMETOOLONG; 1426 goto done_deref_locked; 1427 } 1428 } 1429 } 1430 if (pr_allow & ~ppr->pr_allow) { 1431 error = EPERM; 1432 goto done_deref_locked; 1433 } 1434 1435 /* Set the parameters of the prison. */ 1436#ifdef INET 1437 redo_ip4 = 0; 1438 if (pr_flags & PR_IP4_USER) { 1439 pr->pr_flags |= PR_IP4; 1440 free(pr->pr_ip4, M_PRISON); 1441 pr->pr_ip4s = ip4s; 1442 pr->pr_ip4 = ip4; 1443 ip4 = NULL; 1444 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) { 1445#ifdef VIMAGE 1446 if (tpr->pr_flags & PR_VNET) { 1447 descend = 0; 1448 continue; 1449 } 1450#endif 1451 if (prison_restrict_ip4(tpr, NULL)) { 1452 redo_ip4 = 1; 1453 descend = 0; 1454 } 1455 } 1456 } 1457#endif 1458#ifdef INET6 1459 redo_ip6 = 0; 1460 if (pr_flags & PR_IP6_USER) { 1461 pr->pr_flags |= PR_IP6; 1462 free(pr->pr_ip6, M_PRISON); 1463 pr->pr_ip6s = ip6s; 1464 pr->pr_ip6 = ip6; 1465 ip6 = NULL; 1466 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) { 1467#ifdef VIMAGE 1468 if (tpr->pr_flags & PR_VNET) { 1469 descend = 0; 1470 continue; 1471 } 1472#endif 1473 if (prison_restrict_ip6(tpr, NULL)) { 1474 redo_ip6 = 1; 1475 descend = 0; 1476 } 1477 } 1478 } 1479#endif 1480 if (gotslevel) { 1481 pr->pr_securelevel = slevel; 1482 /* Set all child jails to be at least this level. */ 1483 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) 1484 if (tpr->pr_securelevel < slevel) 1485 tpr->pr_securelevel = slevel; 1486 } 1487 if (gotchildmax) { 1488 pr->pr_childmax = childmax; 1489 /* Set all child jails to under this limit. */ 1490 FOREACH_PRISON_DESCENDANT_LOCKED_LEVEL(pr, tpr, descend, level) 1491 if (tpr->pr_childmax > childmax - level) 1492 tpr->pr_childmax = childmax > level 1493 ? childmax - level : 0; 1494 } 1495 if (gotenforce) { 1496 pr->pr_enforce_statfs = enforce; 1497 /* Pass this restriction on to the children. */ 1498 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) 1499 if (tpr->pr_enforce_statfs < enforce) 1500 tpr->pr_enforce_statfs = enforce; 1501 } 1502 if (name != NULL) { 1503 if (ppr == &prison0) 1504 strlcpy(pr->pr_name, name, sizeof(pr->pr_name)); 1505 else 1506 snprintf(pr->pr_name, sizeof(pr->pr_name), "%s.%s", 1507 ppr->pr_name, name); 1508 /* Change this component of child names. */ 1509 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) { 1510 bcopy(tpr->pr_name + onamelen, tpr->pr_name + namelen, 1511 strlen(tpr->pr_name + onamelen) + 1); 1512 bcopy(pr->pr_name, tpr->pr_name, namelen); 1513 } 1514 } 1515 if (path != NULL) { 1516 /* Try to keep a real-rooted full pathname. */ 1517 if (path[0] == '/' && strcmp(mypr->pr_path, "/")) 1518 snprintf(pr->pr_path, sizeof(pr->pr_path), "%s%s", 1519 mypr->pr_path, path); 1520 else 1521 strlcpy(pr->pr_path, path, sizeof(pr->pr_path)); 1522 pr->pr_root = root; 1523 } 1524 if (PR_HOST & ch_flags & ~pr_flags) { 1525 if (pr->pr_flags & PR_HOST) { 1526 /* 1527 * Copy the parent's host info. As with pr_ip4 above, 1528 * the lack of a lock on the parent is not a problem; 1529 * it is always set with allprison_lock at least 1530 * shared, and is held exclusively here. 1531 */ 1532 strlcpy(pr->pr_hostname, pr->pr_parent->pr_hostname, 1533 sizeof(pr->pr_hostname)); 1534 strlcpy(pr->pr_domainname, pr->pr_parent->pr_domainname, 1535 sizeof(pr->pr_domainname)); 1536 strlcpy(pr->pr_hostuuid, pr->pr_parent->pr_hostuuid, 1537 sizeof(pr->pr_hostuuid)); 1538 pr->pr_hostid = pr->pr_parent->pr_hostid; 1539 } 1540 } else if (host != NULL || domain != NULL || uuid != NULL || gothid) { 1541 /* Set this prison, and any descendants without PR_HOST. */ 1542 if (host != NULL) 1543 strlcpy(pr->pr_hostname, host, sizeof(pr->pr_hostname)); 1544 if (domain != NULL) 1545 strlcpy(pr->pr_domainname, domain, 1546 sizeof(pr->pr_domainname)); 1547 if (uuid != NULL) 1548 strlcpy(pr->pr_hostuuid, uuid, sizeof(pr->pr_hostuuid)); 1549 if (gothid) 1550 pr->pr_hostid = hid; 1551 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) { 1552 if (tpr->pr_flags & PR_HOST) 1553 descend = 0; 1554 else { 1555 if (host != NULL) 1556 strlcpy(tpr->pr_hostname, 1557 pr->pr_hostname, 1558 sizeof(tpr->pr_hostname)); 1559 if (domain != NULL) 1560 strlcpy(tpr->pr_domainname, 1561 pr->pr_domainname, 1562 sizeof(tpr->pr_domainname)); 1563 if (uuid != NULL) 1564 strlcpy(tpr->pr_hostuuid, 1565 pr->pr_hostuuid, 1566 sizeof(tpr->pr_hostuuid)); 1567 if (gothid) 1568 tpr->pr_hostid = hid; 1569 } 1570 } 1571 } 1572 if ((tallow = ch_allow & ~pr_allow)) { 1573 /* Clear allow bits in all children. */ 1574 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) 1575 tpr->pr_allow &= ~tallow; 1576 } 1577 pr->pr_allow = (pr->pr_allow & ~ch_allow) | pr_allow; 1578 /* 1579 * Persistent prisons get an extra reference, and prisons losing their 1580 * persist flag lose that reference. Only do this for existing prisons 1581 * for now, so new ones will remain unseen until after the module 1582 * handlers have completed. 1583 */ 1584 if (!created && (ch_flags & PR_PERSIST & (pr_flags ^ pr->pr_flags))) { 1585 if (pr_flags & PR_PERSIST) { 1586 pr->pr_ref++; 1587 pr->pr_uref++; 1588 } else { 1589 pr->pr_ref--; 1590 pr->pr_uref--; 1591 } 1592 } 1593 pr->pr_flags = (pr->pr_flags & ~ch_flags) | pr_flags; 1594 mtx_unlock(&pr->pr_mtx); 1595 1596 /* Locks may have prevented a complete restriction of child IP 1597 * addresses. If so, allocate some more memory and try again. 1598 */ 1599#ifdef INET 1600 while (redo_ip4) { 1601 ip4s = pr->pr_ip4s; 1602 ip4 = malloc(ip4s * sizeof(*ip4), M_PRISON, M_WAITOK); 1603 mtx_lock(&pr->pr_mtx); 1604 redo_ip4 = 0; 1605 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) { 1606#ifdef VIMAGE 1607 if (tpr->pr_flags & PR_VNET) { 1608 descend = 0; 1609 continue; 1610 } 1611#endif 1612 if (prison_restrict_ip4(tpr, ip4)) { 1613 if (ip4 != NULL) 1614 ip4 = NULL; 1615 else 1616 redo_ip4 = 1; 1617 } 1618 } 1619 mtx_unlock(&pr->pr_mtx); 1620 } 1621#endif 1622#ifdef INET6 1623 while (redo_ip6) { 1624 ip6s = pr->pr_ip6s; 1625 ip6 = malloc(ip6s * sizeof(*ip6), M_PRISON, M_WAITOK); 1626 mtx_lock(&pr->pr_mtx); 1627 redo_ip6 = 0; 1628 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) { 1629#ifdef VIMAGE 1630 if (tpr->pr_flags & PR_VNET) { 1631 descend = 0; 1632 continue; 1633 } 1634#endif 1635 if (prison_restrict_ip6(tpr, ip6)) { 1636 if (ip6 != NULL) 1637 ip6 = NULL; 1638 else 1639 redo_ip6 = 1; 1640 } 1641 } 1642 mtx_unlock(&pr->pr_mtx); 1643 } 1644#endif 1645 1646 /* Let the modules do their work. */ 1647 sx_downgrade(&allprison_lock); 1648 if (created) { 1649 error = osd_jail_call(pr, PR_METHOD_CREATE, opts); 1650 if (error) { 1651 prison_deref(pr, PD_LIST_SLOCKED); 1652 goto done_errmsg; 1653 } 1654 } 1655 error = osd_jail_call(pr, PR_METHOD_SET, opts); 1656 if (error) { 1657 prison_deref(pr, created 1658 ? PD_LIST_SLOCKED 1659 : PD_DEREF | PD_LIST_SLOCKED); 1660 goto done_errmsg; 1661 } 1662 1663 /* Attach this process to the prison if requested. */ 1664 if (flags & JAIL_ATTACH) { 1665 mtx_lock(&pr->pr_mtx); 1666 error = do_jail_attach(td, pr); 1667 if (error) { 1668 vfs_opterror(opts, "attach failed"); 1669 if (!created) 1670 prison_deref(pr, PD_DEREF); 1671 goto done_errmsg; 1672 } 1673 } 1674 1675 /* 1676 * Now that it is all there, drop the temporary reference from existing 1677 * prisons. Or add a reference to newly created persistent prisons 1678 * (which was not done earlier so that the prison would not be publicly 1679 * visible). 1680 */ 1681 if (!created) { 1682 prison_deref(pr, (flags & JAIL_ATTACH) 1683 ? PD_DEREF 1684 : PD_DEREF | PD_LIST_SLOCKED); 1685 } else { 1686 if (pr_flags & PR_PERSIST) { 1687 mtx_lock(&pr->pr_mtx); 1688 pr->pr_ref++; 1689 pr->pr_uref++; 1690 mtx_unlock(&pr->pr_mtx); 1691 } 1692 if (!(flags & JAIL_ATTACH)) 1693 sx_sunlock(&allprison_lock); 1694 } 1695 td->td_retval[0] = pr->pr_id; 1696 goto done_errmsg; 1697 1698 done_deref_locked: 1699 prison_deref(pr, created 1700 ? PD_LOCKED | PD_LIST_XLOCKED 1701 : PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED); 1702 goto done_releroot; 1703 done_unlock_list: 1704 sx_xunlock(&allprison_lock); 1705 done_releroot: 1706 if (root != NULL) { 1707 vfslocked = VFS_LOCK_GIANT(root->v_mount); 1708 vrele(root); 1709 VFS_UNLOCK_GIANT(vfslocked); 1710 } 1711 done_errmsg: 1712 if (error) { 1713 vfs_getopt(opts, "errmsg", (void **)&errmsg, &errmsg_len); 1714 if (errmsg_len > 0) { 1715 errmsg_pos = 2 * vfs_getopt_pos(opts, "errmsg") + 1; 1716 if (errmsg_pos > 0) { 1717 if (optuio->uio_segflg == UIO_SYSSPACE) 1718 bcopy(errmsg, 1719 optuio->uio_iov[errmsg_pos].iov_base, 1720 errmsg_len); 1721 else 1722 copyout(errmsg, 1723 optuio->uio_iov[errmsg_pos].iov_base, 1724 errmsg_len); 1725 } 1726 } 1727 } 1728 done_free: 1729#ifdef INET 1730 free(ip4, M_PRISON); 1731#endif 1732#ifdef INET6 1733 free(ip6, M_PRISON); 1734#endif 1735 vfs_freeopts(opts); 1736 return (error); 1737} 1738 1739 1740/* 1741 * struct jail_get_args { 1742 * struct iovec *iovp; 1743 * unsigned int iovcnt; 1744 * int flags; 1745 * }; 1746 */ 1747int 1748jail_get(struct thread *td, struct jail_get_args *uap) 1749{ 1750 struct uio *auio; 1751 int error; 1752 1753 /* Check that we have an even number of iovecs. */ 1754 if (uap->iovcnt & 1) 1755 return (EINVAL); 1756 1757 error = copyinuio(uap->iovp, uap->iovcnt, &auio); 1758 if (error) 1759 return (error); 1760 error = kern_jail_get(td, auio, uap->flags); 1761 if (error == 0) 1762 error = copyout(auio->uio_iov, uap->iovp, 1763 uap->iovcnt * sizeof (struct iovec)); 1764 free(auio, M_IOV); 1765 return (error); 1766} 1767 1768int 1769kern_jail_get(struct thread *td, struct uio *optuio, int flags) 1770{ 1771 struct prison *pr, *mypr; 1772 struct vfsopt *opt; 1773 struct vfsoptlist *opts; 1774 char *errmsg, *name; 1775 int error, errmsg_len, errmsg_pos, fi, i, jid, len, locked, pos; 1776 1777 if (flags & ~JAIL_GET_MASK) 1778 return (EINVAL); 1779 1780 /* Get the parameter list. */ 1781 error = vfs_buildopts(optuio, &opts); 1782 if (error) 1783 return (error); 1784 errmsg_pos = vfs_getopt_pos(opts, "errmsg"); 1785 mypr = td->td_ucred->cr_prison; 1786 1787 /* 1788 * Find the prison specified by one of: lastjid, jid, name. 1789 */ 1790 sx_slock(&allprison_lock); 1791 error = vfs_copyopt(opts, "lastjid", &jid, sizeof(jid)); 1792 if (error == 0) { 1793 TAILQ_FOREACH(pr, &allprison, pr_list) { 1794 if (pr->pr_id > jid && prison_ischild(mypr, pr)) { 1795 mtx_lock(&pr->pr_mtx); 1796 if (pr->pr_ref > 0 && 1797 (pr->pr_uref > 0 || (flags & JAIL_DYING))) 1798 break; 1799 mtx_unlock(&pr->pr_mtx); 1800 } 1801 } 1802 if (pr != NULL) 1803 goto found_prison; 1804 error = ENOENT; 1805 vfs_opterror(opts, "no jail after %d", jid); 1806 goto done_unlock_list; 1807 } else if (error != ENOENT) 1808 goto done_unlock_list; 1809 1810 error = vfs_copyopt(opts, "jid", &jid, sizeof(jid)); 1811 if (error == 0) { 1812 if (jid != 0) { 1813 pr = prison_find_child(mypr, jid); 1814 if (pr != NULL) { 1815 if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) { 1816 mtx_unlock(&pr->pr_mtx); 1817 error = ENOENT; 1818 vfs_opterror(opts, "jail %d is dying", 1819 jid); 1820 goto done_unlock_list; 1821 } 1822 goto found_prison; 1823 } 1824 error = ENOENT; 1825 vfs_opterror(opts, "jail %d not found", jid); 1826 goto done_unlock_list; 1827 } 1828 } else if (error != ENOENT) 1829 goto done_unlock_list; 1830 1831 error = vfs_getopt(opts, "name", (void **)&name, &len); 1832 if (error == 0) { 1833 if (len == 0 || name[len - 1] != '\0') { 1834 error = EINVAL; 1835 goto done_unlock_list; 1836 } 1837 pr = prison_find_name(mypr, name); 1838 if (pr != NULL) { 1839 if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) { 1840 mtx_unlock(&pr->pr_mtx); 1841 error = ENOENT; 1842 vfs_opterror(opts, "jail \"%s\" is dying", 1843 name); 1844 goto done_unlock_list; 1845 } 1846 goto found_prison; 1847 } 1848 error = ENOENT; 1849 vfs_opterror(opts, "jail \"%s\" not found", name); 1850 goto done_unlock_list; 1851 } else if (error != ENOENT) 1852 goto done_unlock_list; 1853 1854 vfs_opterror(opts, "no jail specified"); 1855 error = ENOENT; 1856 goto done_unlock_list; 1857 1858 found_prison: 1859 /* Get the parameters of the prison. */ 1860 pr->pr_ref++; 1861 locked = PD_LOCKED; 1862 td->td_retval[0] = pr->pr_id; 1863 error = vfs_setopt(opts, "jid", &pr->pr_id, sizeof(pr->pr_id)); 1864 if (error != 0 && error != ENOENT) 1865 goto done_deref; 1866 i = (pr->pr_parent == mypr) ? 0 : pr->pr_parent->pr_id; 1867 error = vfs_setopt(opts, "parent", &i, sizeof(i)); 1868 if (error != 0 && error != ENOENT) 1869 goto done_deref; 1870 error = vfs_setopts(opts, "name", prison_name(mypr, pr)); 1871 if (error != 0 && error != ENOENT) 1872 goto done_deref; 1873 error = vfs_setopt(opts, "cpuset.id", &pr->pr_cpuset->cs_id, 1874 sizeof(pr->pr_cpuset->cs_id)); 1875 if (error != 0 && error != ENOENT) 1876 goto done_deref; 1877 error = vfs_setopts(opts, "path", prison_path(mypr, pr)); 1878 if (error != 0 && error != ENOENT) 1879 goto done_deref; 1880#ifdef INET 1881 error = vfs_setopt_part(opts, "ip4.addr", pr->pr_ip4, 1882 pr->pr_ip4s * sizeof(*pr->pr_ip4)); 1883 if (error != 0 && error != ENOENT) 1884 goto done_deref; 1885#endif 1886#ifdef INET6 1887 error = vfs_setopt_part(opts, "ip6.addr", pr->pr_ip6, 1888 pr->pr_ip6s * sizeof(*pr->pr_ip6)); 1889 if (error != 0 && error != ENOENT) 1890 goto done_deref; 1891#endif 1892 error = vfs_setopt(opts, "securelevel", &pr->pr_securelevel, 1893 sizeof(pr->pr_securelevel)); 1894 if (error != 0 && error != ENOENT) 1895 goto done_deref; 1896 error = vfs_setopt(opts, "children.cur", &pr->pr_childcount, 1897 sizeof(pr->pr_childcount)); 1898 if (error != 0 && error != ENOENT) 1899 goto done_deref; 1900 error = vfs_setopt(opts, "children.max", &pr->pr_childmax, 1901 sizeof(pr->pr_childmax)); 1902 if (error != 0 && error != ENOENT) 1903 goto done_deref; 1904 error = vfs_setopts(opts, "host.hostname", pr->pr_hostname); 1905 if (error != 0 && error != ENOENT) 1906 goto done_deref; 1907 error = vfs_setopts(opts, "host.domainname", pr->pr_domainname); 1908 if (error != 0 && error != ENOENT) 1909 goto done_deref; 1910 error = vfs_setopts(opts, "host.hostuuid", pr->pr_hostuuid); 1911 if (error != 0 && error != ENOENT) 1912 goto done_deref; 1913#ifdef COMPAT_IA32 1914 if (td->td_proc->p_sysent->sv_flags & SV_IA32) { 1915 uint32_t hid32 = pr->pr_hostid; 1916 1917 error = vfs_setopt(opts, "host.hostid", &hid32, sizeof(hid32)); 1918 } else 1919#endif 1920 error = vfs_setopt(opts, "host.hostid", &pr->pr_hostid, 1921 sizeof(pr->pr_hostid)); 1922 if (error != 0 && error != ENOENT) 1923 goto done_deref; 1924 error = vfs_setopt(opts, "enforce_statfs", &pr->pr_enforce_statfs, 1925 sizeof(pr->pr_enforce_statfs)); 1926 if (error != 0 && error != ENOENT) 1927 goto done_deref; 1928 for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]); 1929 fi++) { 1930 if (pr_flag_names[fi] == NULL) 1931 continue; 1932 i = (pr->pr_flags & (1 << fi)) ? 1 : 0; 1933 error = vfs_setopt(opts, pr_flag_names[fi], &i, sizeof(i)); 1934 if (error != 0 && error != ENOENT) 1935 goto done_deref; 1936 i = !i; 1937 error = vfs_setopt(opts, pr_flag_nonames[fi], &i, sizeof(i)); 1938 if (error != 0 && error != ENOENT) 1939 goto done_deref; 1940 } 1941 for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]); 1942 fi++) { 1943 i = pr->pr_flags & 1944 (pr_flag_jailsys[fi].disable | pr_flag_jailsys[fi].new); 1945 i = pr_flag_jailsys[fi].disable && 1946 (i == pr_flag_jailsys[fi].disable) ? JAIL_SYS_DISABLE 1947 : (i == pr_flag_jailsys[fi].new) ? JAIL_SYS_NEW 1948 : JAIL_SYS_INHERIT; 1949 error = 1950 vfs_setopt(opts, pr_flag_jailsys[fi].name, &i, sizeof(i)); 1951 if (error != 0 && error != ENOENT) 1952 goto done_deref; 1953 } 1954 for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]); 1955 fi++) { 1956 if (pr_allow_names[fi] == NULL) 1957 continue; 1958 i = (pr->pr_allow & (1 << fi)) ? 1 : 0; 1959 error = vfs_setopt(opts, pr_allow_names[fi], &i, sizeof(i)); 1960 if (error != 0 && error != ENOENT) 1961 goto done_deref; 1962 i = !i; 1963 error = vfs_setopt(opts, pr_allow_nonames[fi], &i, sizeof(i)); 1964 if (error != 0 && error != ENOENT) 1965 goto done_deref; 1966 } 1967 i = (pr->pr_uref == 0); 1968 error = vfs_setopt(opts, "dying", &i, sizeof(i)); 1969 if (error != 0 && error != ENOENT) 1970 goto done_deref; 1971 i = !i; 1972 error = vfs_setopt(opts, "nodying", &i, sizeof(i)); 1973 if (error != 0 && error != ENOENT) 1974 goto done_deref; 1975 1976 /* Get the module parameters. */ 1977 mtx_unlock(&pr->pr_mtx); 1978 locked = 0; 1979 error = osd_jail_call(pr, PR_METHOD_GET, opts); 1980 if (error) 1981 goto done_deref; 1982 prison_deref(pr, PD_DEREF | PD_LIST_SLOCKED); 1983 1984 /* By now, all parameters should have been noted. */ 1985 TAILQ_FOREACH(opt, opts, link) { 1986 if (!opt->seen && strcmp(opt->name, "errmsg")) { 1987 error = EINVAL; 1988 vfs_opterror(opts, "unknown parameter: %s", opt->name); 1989 goto done_errmsg; 1990 } 1991 } 1992 1993 /* Write the fetched parameters back to userspace. */ 1994 error = 0; 1995 TAILQ_FOREACH(opt, opts, link) { 1996 if (opt->pos >= 0 && opt->pos != errmsg_pos) { 1997 pos = 2 * opt->pos + 1; 1998 optuio->uio_iov[pos].iov_len = opt->len; 1999 if (opt->value != NULL) { 2000 if (optuio->uio_segflg == UIO_SYSSPACE) { 2001 bcopy(opt->value, 2002 optuio->uio_iov[pos].iov_base, 2003 opt->len); 2004 } else { 2005 error = copyout(opt->value, 2006 optuio->uio_iov[pos].iov_base, 2007 opt->len); 2008 if (error) 2009 break; 2010 } 2011 } 2012 } 2013 } 2014 goto done_errmsg; 2015 2016 done_deref: 2017 prison_deref(pr, locked | PD_DEREF | PD_LIST_SLOCKED); 2018 goto done_errmsg; 2019 2020 done_unlock_list: 2021 sx_sunlock(&allprison_lock); 2022 done_errmsg: 2023 if (error && errmsg_pos >= 0) { 2024 vfs_getopt(opts, "errmsg", (void **)&errmsg, &errmsg_len); 2025 errmsg_pos = 2 * errmsg_pos + 1; 2026 if (errmsg_len > 0) { 2027 if (optuio->uio_segflg == UIO_SYSSPACE) 2028 bcopy(errmsg, 2029 optuio->uio_iov[errmsg_pos].iov_base, 2030 errmsg_len); 2031 else 2032 copyout(errmsg, 2033 optuio->uio_iov[errmsg_pos].iov_base, 2034 errmsg_len); 2035 } 2036 } 2037 vfs_freeopts(opts); 2038 return (error); 2039} 2040 2041 2042/* 2043 * struct jail_remove_args { 2044 * int jid; 2045 * }; 2046 */ 2047int 2048jail_remove(struct thread *td, struct jail_remove_args *uap) 2049{ 2050 struct prison *pr, *cpr, *lpr, *tpr; 2051 int descend, error; 2052 2053 error = priv_check(td, PRIV_JAIL_REMOVE); 2054 if (error) 2055 return (error); 2056 2057 sx_xlock(&allprison_lock); 2058 pr = prison_find_child(td->td_ucred->cr_prison, uap->jid); 2059 if (pr == NULL) { 2060 sx_xunlock(&allprison_lock); 2061 return (EINVAL); 2062 } 2063 2064 /* Remove all descendants of this prison, then remove this prison. */ 2065 pr->pr_ref++; 2066 pr->pr_flags |= PR_REMOVE; 2067 if (!LIST_EMPTY(&pr->pr_children)) { 2068 mtx_unlock(&pr->pr_mtx); 2069 lpr = NULL; 2070 FOREACH_PRISON_DESCENDANT(pr, cpr, descend) { 2071 mtx_lock(&cpr->pr_mtx); 2072 if (cpr->pr_ref > 0) { 2073 tpr = cpr; 2074 cpr->pr_ref++; 2075 cpr->pr_flags |= PR_REMOVE; 2076 } else { 2077 /* Already removed - do not do it again. */ 2078 tpr = NULL; 2079 } 2080 mtx_unlock(&cpr->pr_mtx); 2081 if (lpr != NULL) { 2082 mtx_lock(&lpr->pr_mtx); 2083 prison_remove_one(lpr); 2084 sx_xlock(&allprison_lock); 2085 } 2086 lpr = tpr; 2087 } 2088 if (lpr != NULL) { 2089 mtx_lock(&lpr->pr_mtx); 2090 prison_remove_one(lpr); 2091 sx_xlock(&allprison_lock); 2092 } 2093 mtx_lock(&pr->pr_mtx); 2094 } 2095 prison_remove_one(pr); 2096 return (0); 2097} 2098 2099static void 2100prison_remove_one(struct prison *pr) 2101{ 2102 struct proc *p; 2103 int deuref; 2104 2105 /* If the prison was persistent, it is not anymore. */ 2106 deuref = 0; 2107 if (pr->pr_flags & PR_PERSIST) { 2108 pr->pr_ref--; 2109 deuref = PD_DEUREF; 2110 pr->pr_flags &= ~PR_PERSIST; 2111 } 2112 2113 /* 2114 * jail_remove added a reference. If that's the only one, remove 2115 * the prison now. 2116 */ 2117 KASSERT(pr->pr_ref > 0, 2118 ("prison_remove_one removing a dead prison (jid=%d)", pr->pr_id)); 2119 if (pr->pr_ref == 1) { 2120 prison_deref(pr, 2121 deuref | PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED); 2122 return; 2123 } 2124 2125 mtx_unlock(&pr->pr_mtx); 2126 sx_xunlock(&allprison_lock); 2127 /* 2128 * Kill all processes unfortunate enough to be attached to this prison. 2129 */ 2130 sx_slock(&allproc_lock); 2131 LIST_FOREACH(p, &allproc, p_list) { 2132 PROC_LOCK(p); 2133 if (p->p_state != PRS_NEW && p->p_ucred && 2134 p->p_ucred->cr_prison == pr) 2135 psignal(p, SIGKILL); 2136 PROC_UNLOCK(p); 2137 } 2138 sx_sunlock(&allproc_lock); 2139 /* Remove the temporary reference added by jail_remove. */ 2140 prison_deref(pr, deuref | PD_DEREF); 2141} 2142 2143 2144/* 2145 * struct jail_attach_args { 2146 * int jid; 2147 * }; 2148 */ 2149int 2150jail_attach(struct thread *td, struct jail_attach_args *uap) 2151{ 2152 struct prison *pr; 2153 int error; 2154 2155 error = priv_check(td, PRIV_JAIL_ATTACH); 2156 if (error) 2157 return (error); 2158 2159 sx_slock(&allprison_lock); 2160 pr = prison_find_child(td->td_ucred->cr_prison, uap->jid); 2161 if (pr == NULL) { 2162 sx_sunlock(&allprison_lock); 2163 return (EINVAL); 2164 } 2165 2166 /* 2167 * Do not allow a process to attach to a prison that is not 2168 * considered to be "alive". 2169 */ 2170 if (pr->pr_uref == 0) { 2171 mtx_unlock(&pr->pr_mtx); 2172 sx_sunlock(&allprison_lock); 2173 return (EINVAL); 2174 } 2175 2176 return (do_jail_attach(td, pr)); 2177} 2178 2179static int 2180do_jail_attach(struct thread *td, struct prison *pr) 2181{ 2182 struct prison *ppr; 2183 struct proc *p; 2184 struct ucred *newcred, *oldcred; 2185 int vfslocked, error; 2186 2187 /* 2188 * XXX: Note that there is a slight race here if two threads 2189 * in the same privileged process attempt to attach to two 2190 * different jails at the same time. It is important for 2191 * user processes not to do this, or they might end up with 2192 * a process root from one prison, but attached to the jail 2193 * of another. 2194 */ 2195 pr->pr_ref++; 2196 pr->pr_uref++; 2197 mtx_unlock(&pr->pr_mtx); 2198 2199 /* Let modules do whatever they need to prepare for attaching. */ 2200 error = osd_jail_call(pr, PR_METHOD_ATTACH, td); 2201 if (error) { 2202 prison_deref(pr, PD_DEREF | PD_DEUREF | PD_LIST_SLOCKED); 2203 return (error); 2204 } 2205 sx_sunlock(&allprison_lock); 2206 2207 /* 2208 * Reparent the newly attached process to this jail. 2209 */ 2210 ppr = td->td_ucred->cr_prison; 2211 p = td->td_proc; 2212 error = cpuset_setproc_update_set(p, pr->pr_cpuset); 2213 if (error) 2214 goto e_revert_osd; 2215 2216 vfslocked = VFS_LOCK_GIANT(pr->pr_root->v_mount); 2217 vn_lock(pr->pr_root, LK_EXCLUSIVE | LK_RETRY); 2218 if ((error = change_dir(pr->pr_root, td)) != 0) 2219 goto e_unlock; 2220#ifdef MAC 2221 if ((error = mac_vnode_check_chroot(td->td_ucred, pr->pr_root))) 2222 goto e_unlock; 2223#endif 2224 VOP_UNLOCK(pr->pr_root, 0); 2225 if ((error = change_root(pr->pr_root, td))) 2226 goto e_unlock_giant; 2227 VFS_UNLOCK_GIANT(vfslocked); 2228 2229 newcred = crget(); 2230 PROC_LOCK(p); 2231 oldcred = p->p_ucred; 2232 setsugid(p); 2233 crcopy(newcred, oldcred); 2234 newcred->cr_prison = pr; 2235 p->p_ucred = newcred; 2236 PROC_UNLOCK(p); 2237 crfree(oldcred); 2238 prison_deref(ppr, PD_DEREF | PD_DEUREF); 2239 return (0); 2240 e_unlock: 2241 VOP_UNLOCK(pr->pr_root, 0); 2242 e_unlock_giant: 2243 VFS_UNLOCK_GIANT(vfslocked); 2244 e_revert_osd: 2245 /* Tell modules this thread is still in its old jail after all. */ 2246 (void)osd_jail_call(ppr, PR_METHOD_ATTACH, td); 2247 prison_deref(pr, PD_DEREF | PD_DEUREF); 2248 return (error); 2249} 2250 2251 2252/* 2253 * Returns a locked prison instance, or NULL on failure. 2254 */ 2255struct prison * 2256prison_find(int prid) 2257{ 2258 struct prison *pr; 2259 2260 sx_assert(&allprison_lock, SX_LOCKED); 2261 TAILQ_FOREACH(pr, &allprison, pr_list) { 2262 if (pr->pr_id == prid) { 2263 mtx_lock(&pr->pr_mtx); 2264 if (pr->pr_ref > 0) 2265 return (pr); 2266 mtx_unlock(&pr->pr_mtx); 2267 } 2268 } 2269 return (NULL); 2270} 2271 2272/* 2273 * Find a prison that is a descendant of mypr. Returns a locked prison or NULL. 2274 */ 2275struct prison * 2276prison_find_child(struct prison *mypr, int prid) 2277{ 2278 struct prison *pr; 2279 int descend; 2280 2281 sx_assert(&allprison_lock, SX_LOCKED); 2282 FOREACH_PRISON_DESCENDANT(mypr, pr, descend) { 2283 if (pr->pr_id == prid) { 2284 mtx_lock(&pr->pr_mtx); 2285 if (pr->pr_ref > 0) 2286 return (pr); 2287 mtx_unlock(&pr->pr_mtx); 2288 } 2289 } 2290 return (NULL); 2291} 2292 2293/* 2294 * Look for the name relative to mypr. Returns a locked prison or NULL. 2295 */ 2296struct prison * 2297prison_find_name(struct prison *mypr, const char *name) 2298{ 2299 struct prison *pr, *deadpr; 2300 size_t mylen; 2301 int descend; 2302 2303 sx_assert(&allprison_lock, SX_LOCKED); 2304 mylen = (mypr == &prison0) ? 0 : strlen(mypr->pr_name) + 1; 2305 again: 2306 deadpr = NULL; 2307 FOREACH_PRISON_DESCENDANT(mypr, pr, descend) { 2308 if (!strcmp(pr->pr_name + mylen, name)) { 2309 mtx_lock(&pr->pr_mtx); 2310 if (pr->pr_ref > 0) { 2311 if (pr->pr_uref > 0) 2312 return (pr); 2313 deadpr = pr; 2314 } 2315 mtx_unlock(&pr->pr_mtx); 2316 } 2317 } 2318 /* There was no valid prison - perhaps there was a dying one. */ 2319 if (deadpr != NULL) { 2320 mtx_lock(&deadpr->pr_mtx); 2321 if (deadpr->pr_ref == 0) { 2322 mtx_unlock(&deadpr->pr_mtx); 2323 goto again; 2324 } 2325 } 2326 return (deadpr); 2327} 2328 2329/* 2330 * See if a prison has the specific flag set. 2331 */ 2332int 2333prison_flag(struct ucred *cred, unsigned flag) 2334{ 2335 2336 /* This is an atomic read, so no locking is necessary. */ 2337 return (cred->cr_prison->pr_flags & flag); 2338} 2339 2340int 2341prison_allow(struct ucred *cred, unsigned flag) 2342{ 2343 2344 /* This is an atomic read, so no locking is necessary. */ 2345 return (cred->cr_prison->pr_allow & flag); 2346} 2347 2348/* 2349 * Remove a prison reference. If that was the last reference, remove the 2350 * prison itself - but not in this context in case there are locks held. 2351 */ 2352void 2353prison_free_locked(struct prison *pr) 2354{ 2355 2356 mtx_assert(&pr->pr_mtx, MA_OWNED); 2357 pr->pr_ref--; 2358 if (pr->pr_ref == 0) { 2359 mtx_unlock(&pr->pr_mtx); 2360 TASK_INIT(&pr->pr_task, 0, prison_complete, pr); 2361 taskqueue_enqueue(taskqueue_thread, &pr->pr_task); 2362 return; 2363 } 2364 mtx_unlock(&pr->pr_mtx); 2365} 2366 2367void 2368prison_free(struct prison *pr) 2369{ 2370 2371 mtx_lock(&pr->pr_mtx); 2372 prison_free_locked(pr); 2373} 2374 2375static void 2376prison_complete(void *context, int pending) 2377{ 2378 2379 prison_deref((struct prison *)context, 0); 2380} 2381 2382/* 2383 * Remove a prison reference (usually). This internal version assumes no 2384 * mutexes are held, except perhaps the prison itself. If there are no more 2385 * references, release and delist the prison. On completion, the prison lock 2386 * and the allprison lock are both unlocked. 2387 */ 2388static void 2389prison_deref(struct prison *pr, int flags) 2390{ 2391 struct prison *ppr, *tpr; 2392 int vfslocked; 2393 2394 if (!(flags & PD_LOCKED)) 2395 mtx_lock(&pr->pr_mtx); 2396 /* Decrement the user references in a separate loop. */ 2397 if (flags & PD_DEUREF) { 2398 for (tpr = pr;; tpr = tpr->pr_parent) { 2399 if (tpr != pr) 2400 mtx_lock(&tpr->pr_mtx); 2401 if (--tpr->pr_uref > 0) 2402 break; 2403 KASSERT(tpr != &prison0, ("prison0 pr_uref=0")); 2404 mtx_unlock(&tpr->pr_mtx); 2405 } 2406 /* Done if there were only user references to remove. */ 2407 if (!(flags & PD_DEREF)) { 2408 mtx_unlock(&tpr->pr_mtx); 2409 if (flags & PD_LIST_SLOCKED) 2410 sx_sunlock(&allprison_lock); 2411 else if (flags & PD_LIST_XLOCKED) 2412 sx_xunlock(&allprison_lock); 2413 return; 2414 } 2415 if (tpr != pr) { 2416 mtx_unlock(&tpr->pr_mtx); 2417 mtx_lock(&pr->pr_mtx); 2418 } 2419 } 2420 2421 for (;;) { 2422 if (flags & PD_DEREF) 2423 pr->pr_ref--; 2424 /* If the prison still has references, nothing else to do. */ 2425 if (pr->pr_ref > 0) { 2426 mtx_unlock(&pr->pr_mtx); 2427 if (flags & PD_LIST_SLOCKED) 2428 sx_sunlock(&allprison_lock); 2429 else if (flags & PD_LIST_XLOCKED) 2430 sx_xunlock(&allprison_lock); 2431 return; 2432 } 2433 2434 mtx_unlock(&pr->pr_mtx); 2435 if (flags & PD_LIST_SLOCKED) { 2436 if (!sx_try_upgrade(&allprison_lock)) { 2437 sx_sunlock(&allprison_lock); 2438 sx_xlock(&allprison_lock); 2439 } 2440 } else if (!(flags & PD_LIST_XLOCKED)) 2441 sx_xlock(&allprison_lock); 2442 2443 TAILQ_REMOVE(&allprison, pr, pr_list); 2444 LIST_REMOVE(pr, pr_sibling); 2445 ppr = pr->pr_parent; 2446 for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent) 2447 tpr->pr_childcount--; 2448 sx_downgrade(&allprison_lock); 2449 2450#ifdef VIMAGE 2451 if (pr->pr_flags & PR_VNET) 2452 vnet_destroy(pr->pr_vnet); 2453#endif 2454 if (pr->pr_root != NULL) { 2455 vfslocked = VFS_LOCK_GIANT(pr->pr_root->v_mount); 2456 vrele(pr->pr_root); 2457 VFS_UNLOCK_GIANT(vfslocked); 2458 } 2459 mtx_destroy(&pr->pr_mtx); 2460#ifdef INET 2461 free(pr->pr_ip4, M_PRISON); 2462#endif 2463#ifdef INET6 2464 free(pr->pr_ip6, M_PRISON); 2465#endif 2466 if (pr->pr_cpuset != NULL) 2467 cpuset_rel(pr->pr_cpuset); 2468 osd_jail_exit(pr); 2469 free(pr, M_PRISON); 2470 2471 /* Removing a prison frees a reference on its parent. */ 2472 pr = ppr; 2473 mtx_lock(&pr->pr_mtx); 2474 flags = PD_DEREF | PD_LIST_SLOCKED; 2475 } 2476} 2477 2478void 2479prison_hold_locked(struct prison *pr) 2480{ 2481 2482 mtx_assert(&pr->pr_mtx, MA_OWNED); 2483 KASSERT(pr->pr_ref > 0, 2484 ("Trying to hold dead prison (jid=%d).", pr->pr_id)); 2485 pr->pr_ref++; 2486} 2487 2488void 2489prison_hold(struct prison *pr) 2490{ 2491 2492 mtx_lock(&pr->pr_mtx); 2493 prison_hold_locked(pr); 2494 mtx_unlock(&pr->pr_mtx); 2495} 2496 2497void 2498prison_proc_hold(struct prison *pr) 2499{ 2500 2501 mtx_lock(&pr->pr_mtx); 2502 KASSERT(pr->pr_uref > 0, 2503 ("Cannot add a process to a non-alive prison (jid=%d)", pr->pr_id)); 2504 pr->pr_uref++; 2505 mtx_unlock(&pr->pr_mtx); 2506} 2507 2508void 2509prison_proc_free(struct prison *pr) 2510{ 2511 2512 mtx_lock(&pr->pr_mtx); 2513 KASSERT(pr->pr_uref > 0, 2514 ("Trying to kill a process in a dead prison (jid=%d)", pr->pr_id)); 2515 prison_deref(pr, PD_DEUREF | PD_LOCKED); 2516} 2517 2518 2519#ifdef INET 2520/* 2521 * Restrict a prison's IP address list with its parent's, possibly replacing 2522 * it. Return true if the replacement buffer was used (or would have been). 2523 */ 2524static int 2525prison_restrict_ip4(struct prison *pr, struct in_addr *newip4) 2526{ 2527 int ii, ij, used; 2528 struct prison *ppr; 2529 2530 ppr = pr->pr_parent; 2531 if (!(pr->pr_flags & PR_IP4_USER)) { 2532 /* This has no user settings, so just copy the parent's list. */ 2533 if (pr->pr_ip4s < ppr->pr_ip4s) { 2534 /* 2535 * There's no room for the parent's list. Use the 2536 * new list buffer, which is assumed to be big enough 2537 * (if it was passed). If there's no buffer, try to 2538 * allocate one. 2539 */ 2540 used = 1; 2541 if (newip4 == NULL) { 2542 newip4 = malloc(ppr->pr_ip4s * sizeof(*newip4), 2543 M_PRISON, M_NOWAIT); 2544 if (newip4 != NULL) 2545 used = 0; 2546 } 2547 if (newip4 != NULL) { 2548 bcopy(ppr->pr_ip4, newip4, 2549 ppr->pr_ip4s * sizeof(*newip4)); 2550 free(pr->pr_ip4, M_PRISON); 2551 pr->pr_ip4 = newip4; 2552 pr->pr_ip4s = ppr->pr_ip4s; 2553 } 2554 return (used); 2555 } 2556 pr->pr_ip4s = ppr->pr_ip4s; 2557 if (pr->pr_ip4s > 0) 2558 bcopy(ppr->pr_ip4, pr->pr_ip4, 2559 pr->pr_ip4s * sizeof(*newip4)); 2560 else if (pr->pr_ip4 != NULL) { 2561 free(pr->pr_ip4, M_PRISON); 2562 pr->pr_ip4 = NULL; 2563 } 2564 } else if (pr->pr_ip4s > 0) { 2565 /* Remove addresses that aren't in the parent. */ 2566 for (ij = 0; ij < ppr->pr_ip4s; ij++) 2567 if (pr->pr_ip4[0].s_addr == ppr->pr_ip4[ij].s_addr) 2568 break; 2569 if (ij < ppr->pr_ip4s) 2570 ii = 1; 2571 else { 2572 bcopy(pr->pr_ip4 + 1, pr->pr_ip4, 2573 --pr->pr_ip4s * sizeof(*pr->pr_ip4)); 2574 ii = 0; 2575 } 2576 for (ij = 1; ii < pr->pr_ip4s; ) { 2577 if (pr->pr_ip4[ii].s_addr == ppr->pr_ip4[0].s_addr) { 2578 ii++; 2579 continue; 2580 } 2581 switch (ij >= ppr->pr_ip4s ? -1 : 2582 qcmp_v4(&pr->pr_ip4[ii], &ppr->pr_ip4[ij])) { 2583 case -1: 2584 bcopy(pr->pr_ip4 + ii + 1, pr->pr_ip4 + ii, 2585 (--pr->pr_ip4s - ii) * sizeof(*pr->pr_ip4)); 2586 break; 2587 case 0: 2588 ii++; 2589 ij++; 2590 break; 2591 case 1: 2592 ij++; 2593 break; 2594 } 2595 } 2596 if (pr->pr_ip4s == 0) { 2597 pr->pr_flags |= PR_IP4_DISABLE; 2598 free(pr->pr_ip4, M_PRISON); 2599 pr->pr_ip4 = NULL; 2600 } 2601 } 2602 return (0); 2603} 2604 2605/* 2606 * Pass back primary IPv4 address of this jail. 2607 * 2608 * If not restricted return success but do not alter the address. Caller has 2609 * to make sure to initialize it correctly (e.g. INADDR_ANY). 2610 * 2611 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv4. 2612 * Address returned in NBO. 2613 */ 2614int 2615prison_get_ip4(struct ucred *cred, struct in_addr *ia) 2616{ 2617 struct prison *pr; 2618 2619 KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); 2620 KASSERT(ia != NULL, ("%s: ia is NULL", __func__)); 2621 2622 pr = cred->cr_prison; 2623 if (!(pr->pr_flags & PR_IP4)) 2624 return (0); 2625 mtx_lock(&pr->pr_mtx); 2626 if (!(pr->pr_flags & PR_IP4)) { 2627 mtx_unlock(&pr->pr_mtx); 2628 return (0); 2629 } 2630 if (pr->pr_ip4 == NULL) { 2631 mtx_unlock(&pr->pr_mtx); 2632 return (EAFNOSUPPORT); 2633 } 2634 2635 ia->s_addr = pr->pr_ip4[0].s_addr; 2636 mtx_unlock(&pr->pr_mtx); 2637 return (0); 2638} 2639 2640/* 2641 * Return true if pr1 and pr2 have the same IPv4 address restrictions. 2642 */ 2643int 2644prison_equal_ip4(struct prison *pr1, struct prison *pr2) 2645{ 2646 2647 if (pr1 == pr2) 2648 return (1); 2649 2650 /* 2651 * No need to lock since the PR_IP4_USER flag can't be altered for 2652 * existing prisons. 2653 */ 2654 while (pr1 != &prison0 && 2655#ifdef VIMAGE 2656 !(pr1->pr_flags & PR_VNET) && 2657#endif 2658 !(pr1->pr_flags & PR_IP4_USER)) 2659 pr1 = pr1->pr_parent; 2660 while (pr2 != &prison0 && 2661#ifdef VIMAGE 2662 !(pr2->pr_flags & PR_VNET) && 2663#endif 2664 !(pr2->pr_flags & PR_IP4_USER)) 2665 pr2 = pr2->pr_parent; 2666 return (pr1 == pr2); 2667} 2668 2669/* 2670 * Make sure our (source) address is set to something meaningful to this 2671 * jail. 2672 * 2673 * Returns 0 if jail doesn't restrict IPv4 or if address belongs to jail, 2674 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail 2675 * doesn't allow IPv4. Address passed in in NBO and returned in NBO. 2676 */ 2677int 2678prison_local_ip4(struct ucred *cred, struct in_addr *ia) 2679{ 2680 struct prison *pr; 2681 struct in_addr ia0; 2682 int error; 2683 2684 KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); 2685 KASSERT(ia != NULL, ("%s: ia is NULL", __func__)); 2686 2687 pr = cred->cr_prison; 2688 if (!(pr->pr_flags & PR_IP4)) 2689 return (0); 2690 mtx_lock(&pr->pr_mtx); 2691 if (!(pr->pr_flags & PR_IP4)) { 2692 mtx_unlock(&pr->pr_mtx); 2693 return (0); 2694 } 2695 if (pr->pr_ip4 == NULL) { 2696 mtx_unlock(&pr->pr_mtx); 2697 return (EAFNOSUPPORT); 2698 } 2699 2700 ia0.s_addr = ntohl(ia->s_addr); 2701 if (ia0.s_addr == INADDR_LOOPBACK) { 2702 ia->s_addr = pr->pr_ip4[0].s_addr; 2703 mtx_unlock(&pr->pr_mtx); 2704 return (0); 2705 } 2706 2707 if (ia0.s_addr == INADDR_ANY) { 2708 /* 2709 * In case there is only 1 IPv4 address, bind directly. 2710 */ 2711 if (pr->pr_ip4s == 1) 2712 ia->s_addr = pr->pr_ip4[0].s_addr; 2713 mtx_unlock(&pr->pr_mtx); 2714 return (0); 2715 } 2716 2717 error = _prison_check_ip4(pr, ia); 2718 mtx_unlock(&pr->pr_mtx); 2719 return (error); 2720} 2721 2722/* 2723 * Rewrite destination address in case we will connect to loopback address. 2724 * 2725 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv4. 2726 * Address passed in in NBO and returned in NBO. 2727 */ 2728int 2729prison_remote_ip4(struct ucred *cred, struct in_addr *ia) 2730{ 2731 struct prison *pr; 2732 2733 KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); 2734 KASSERT(ia != NULL, ("%s: ia is NULL", __func__)); 2735 2736 pr = cred->cr_prison; 2737 if (!(pr->pr_flags & PR_IP4)) 2738 return (0); 2739 mtx_lock(&pr->pr_mtx); 2740 if (!(pr->pr_flags & PR_IP4)) { 2741 mtx_unlock(&pr->pr_mtx); 2742 return (0); 2743 } 2744 if (pr->pr_ip4 == NULL) { 2745 mtx_unlock(&pr->pr_mtx); 2746 return (EAFNOSUPPORT); 2747 } 2748 2749 if (ntohl(ia->s_addr) == INADDR_LOOPBACK) { 2750 ia->s_addr = pr->pr_ip4[0].s_addr; 2751 mtx_unlock(&pr->pr_mtx); 2752 return (0); 2753 } 2754 2755 /* 2756 * Return success because nothing had to be changed. 2757 */ 2758 mtx_unlock(&pr->pr_mtx); 2759 return (0); 2760} 2761 2762/* 2763 * Check if given address belongs to the jail referenced by cred/prison. 2764 * 2765 * Returns 0 if jail doesn't restrict IPv4 or if address belongs to jail, 2766 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail 2767 * doesn't allow IPv4. Address passed in in NBO. 2768 */ 2769static int 2770_prison_check_ip4(struct prison *pr, struct in_addr *ia) 2771{ 2772 int i, a, z, d; 2773 2774 /* 2775 * Check the primary IP. 2776 */ 2777 if (pr->pr_ip4[0].s_addr == ia->s_addr) 2778 return (0); 2779 2780 /* 2781 * All the other IPs are sorted so we can do a binary search. 2782 */ 2783 a = 0; 2784 z = pr->pr_ip4s - 2; 2785 while (a <= z) { 2786 i = (a + z) / 2; 2787 d = qcmp_v4(&pr->pr_ip4[i+1], ia); 2788 if (d > 0) 2789 z = i - 1; 2790 else if (d < 0) 2791 a = i + 1; 2792 else 2793 return (0); 2794 } 2795 2796 return (EADDRNOTAVAIL); 2797} 2798 2799int 2800prison_check_ip4(struct ucred *cred, struct in_addr *ia) 2801{ 2802 struct prison *pr; 2803 int error; 2804 2805 KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); 2806 KASSERT(ia != NULL, ("%s: ia is NULL", __func__)); 2807 2808 pr = cred->cr_prison; 2809 if (!(pr->pr_flags & PR_IP4)) 2810 return (0); 2811 mtx_lock(&pr->pr_mtx); 2812 if (!(pr->pr_flags & PR_IP4)) { 2813 mtx_unlock(&pr->pr_mtx); 2814 return (0); 2815 } 2816 if (pr->pr_ip4 == NULL) { 2817 mtx_unlock(&pr->pr_mtx); 2818 return (EAFNOSUPPORT); 2819 } 2820 2821 error = _prison_check_ip4(pr, ia); 2822 mtx_unlock(&pr->pr_mtx); 2823 return (error); 2824} 2825#endif 2826 2827#ifdef INET6 2828static int 2829prison_restrict_ip6(struct prison *pr, struct in6_addr *newip6) 2830{ 2831 int ii, ij, used; 2832 struct prison *ppr; 2833 2834 ppr = pr->pr_parent; 2835 if (!(pr->pr_flags & PR_IP6_USER)) { 2836 /* This has no user settings, so just copy the parent's list. */ 2837 if (pr->pr_ip6s < ppr->pr_ip6s) { 2838 /* 2839 * There's no room for the parent's list. Use the 2840 * new list buffer, which is assumed to be big enough 2841 * (if it was passed). If there's no buffer, try to 2842 * allocate one. 2843 */ 2844 used = 1; 2845 if (newip6 == NULL) { 2846 newip6 = malloc(ppr->pr_ip6s * sizeof(*newip6), 2847 M_PRISON, M_NOWAIT); 2848 if (newip6 != NULL) 2849 used = 0; 2850 } 2851 if (newip6 != NULL) { 2852 bcopy(ppr->pr_ip6, newip6, 2853 ppr->pr_ip6s * sizeof(*newip6)); 2854 free(pr->pr_ip6, M_PRISON); 2855 pr->pr_ip6 = newip6; 2856 pr->pr_ip6s = ppr->pr_ip6s; 2857 } 2858 return (used); 2859 } 2860 pr->pr_ip6s = ppr->pr_ip6s; 2861 if (pr->pr_ip6s > 0) 2862 bcopy(ppr->pr_ip6, pr->pr_ip6, 2863 pr->pr_ip6s * sizeof(*newip6)); 2864 else if (pr->pr_ip6 != NULL) { 2865 free(pr->pr_ip6, M_PRISON); 2866 pr->pr_ip6 = NULL; 2867 } 2868 } else if (pr->pr_ip6s > 0) { 2869 /* Remove addresses that aren't in the parent. */ 2870 for (ij = 0; ij < ppr->pr_ip6s; ij++) 2871 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[0], 2872 &ppr->pr_ip6[ij])) 2873 break; 2874 if (ij < ppr->pr_ip6s) 2875 ii = 1; 2876 else { 2877 bcopy(pr->pr_ip6 + 1, pr->pr_ip6, 2878 --pr->pr_ip6s * sizeof(*pr->pr_ip6)); 2879 ii = 0; 2880 } 2881 for (ij = 1; ii < pr->pr_ip6s; ) { 2882 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[ii], 2883 &ppr->pr_ip6[0])) { 2884 ii++; 2885 continue; 2886 } 2887 switch (ij >= ppr->pr_ip4s ? -1 : 2888 qcmp_v6(&pr->pr_ip6[ii], &ppr->pr_ip6[ij])) { 2889 case -1: 2890 bcopy(pr->pr_ip6 + ii + 1, pr->pr_ip6 + ii, 2891 (--pr->pr_ip6s - ii) * sizeof(*pr->pr_ip6)); 2892 break; 2893 case 0: 2894 ii++; 2895 ij++; 2896 break; 2897 case 1: 2898 ij++; 2899 break; 2900 } 2901 } 2902 if (pr->pr_ip6s == 0) { 2903 pr->pr_flags |= PR_IP6_DISABLE; 2904 free(pr->pr_ip6, M_PRISON); 2905 pr->pr_ip6 = NULL; 2906 } 2907 } 2908 return 0; 2909} 2910 2911/* 2912 * Pass back primary IPv6 address for this jail. 2913 * 2914 * If not restricted return success but do not alter the address. Caller has 2915 * to make sure to initialize it correctly (e.g. IN6ADDR_ANY_INIT). 2916 * 2917 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv6. 2918 */ 2919int 2920prison_get_ip6(struct ucred *cred, struct in6_addr *ia6) 2921{ 2922 struct prison *pr; 2923 2924 KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); 2925 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__)); 2926 2927 pr = cred->cr_prison; 2928 if (!(pr->pr_flags & PR_IP6)) 2929 return (0); 2930 mtx_lock(&pr->pr_mtx); 2931 if (!(pr->pr_flags & PR_IP6)) { 2932 mtx_unlock(&pr->pr_mtx); 2933 return (0); 2934 } 2935 if (pr->pr_ip6 == NULL) { 2936 mtx_unlock(&pr->pr_mtx); 2937 return (EAFNOSUPPORT); 2938 } 2939 2940 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr)); 2941 mtx_unlock(&pr->pr_mtx); 2942 return (0); 2943} 2944 2945/* 2946 * Return true if pr1 and pr2 have the same IPv6 address restrictions. 2947 */ 2948int 2949prison_equal_ip6(struct prison *pr1, struct prison *pr2) 2950{ 2951 2952 if (pr1 == pr2) 2953 return (1); 2954 2955 while (pr1 != &prison0 && 2956#ifdef VIMAGE 2957 !(pr1->pr_flags & PR_VNET) && 2958#endif 2959 !(pr1->pr_flags & PR_IP6_USER)) 2960 pr1 = pr1->pr_parent; 2961 while (pr2 != &prison0 && 2962#ifdef VIMAGE 2963 !(pr2->pr_flags & PR_VNET) && 2964#endif 2965 !(pr2->pr_flags & PR_IP6_USER)) 2966 pr2 = pr2->pr_parent; 2967 return (pr1 == pr2); 2968} 2969 2970/* 2971 * Make sure our (source) address is set to something meaningful to this jail. 2972 * 2973 * v6only should be set based on (inp->inp_flags & IN6P_IPV6_V6ONLY != 0) 2974 * when needed while binding. 2975 * 2976 * Returns 0 if jail doesn't restrict IPv6 or if address belongs to jail, 2977 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail 2978 * doesn't allow IPv6. 2979 */ 2980int 2981prison_local_ip6(struct ucred *cred, struct in6_addr *ia6, int v6only) 2982{ 2983 struct prison *pr; 2984 int error; 2985 2986 KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); 2987 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__)); 2988 2989 pr = cred->cr_prison; 2990 if (!(pr->pr_flags & PR_IP6)) 2991 return (0); 2992 mtx_lock(&pr->pr_mtx); 2993 if (!(pr->pr_flags & PR_IP6)) { 2994 mtx_unlock(&pr->pr_mtx); 2995 return (0); 2996 } 2997 if (pr->pr_ip6 == NULL) { 2998 mtx_unlock(&pr->pr_mtx); 2999 return (EAFNOSUPPORT); 3000 } 3001 3002 if (IN6_IS_ADDR_LOOPBACK(ia6)) { 3003 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr)); 3004 mtx_unlock(&pr->pr_mtx); 3005 return (0); 3006 } 3007 3008 if (IN6_IS_ADDR_UNSPECIFIED(ia6)) { 3009 /* 3010 * In case there is only 1 IPv6 address, and v6only is true, 3011 * then bind directly. 3012 */ 3013 if (v6only != 0 && pr->pr_ip6s == 1) 3014 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr)); 3015 mtx_unlock(&pr->pr_mtx); 3016 return (0); 3017 } 3018 3019 error = _prison_check_ip6(pr, ia6); 3020 mtx_unlock(&pr->pr_mtx); 3021 return (error); 3022} 3023 3024/* 3025 * Rewrite destination address in case we will connect to loopback address. 3026 * 3027 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv6. 3028 */ 3029int 3030prison_remote_ip6(struct ucred *cred, struct in6_addr *ia6) 3031{ 3032 struct prison *pr; 3033 3034 KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); 3035 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__)); 3036 3037 pr = cred->cr_prison; 3038 if (!(pr->pr_flags & PR_IP6)) 3039 return (0); 3040 mtx_lock(&pr->pr_mtx); 3041 if (!(pr->pr_flags & PR_IP6)) { 3042 mtx_unlock(&pr->pr_mtx); 3043 return (0); 3044 } 3045 if (pr->pr_ip6 == NULL) { 3046 mtx_unlock(&pr->pr_mtx); 3047 return (EAFNOSUPPORT); 3048 } 3049 3050 if (IN6_IS_ADDR_LOOPBACK(ia6)) { 3051 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr)); 3052 mtx_unlock(&pr->pr_mtx); 3053 return (0); 3054 } 3055 3056 /* 3057 * Return success because nothing had to be changed. 3058 */ 3059 mtx_unlock(&pr->pr_mtx); 3060 return (0); 3061} 3062 3063/* 3064 * Check if given address belongs to the jail referenced by cred/prison. 3065 * 3066 * Returns 0 if jail doesn't restrict IPv6 or if address belongs to jail, 3067 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail 3068 * doesn't allow IPv6. 3069 */ 3070static int 3071_prison_check_ip6(struct prison *pr, struct in6_addr *ia6) 3072{ 3073 int i, a, z, d; 3074 3075 /* 3076 * Check the primary IP. 3077 */ 3078 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[0], ia6)) 3079 return (0); 3080 3081 /* 3082 * All the other IPs are sorted so we can do a binary search. 3083 */ 3084 a = 0; 3085 z = pr->pr_ip6s - 2; 3086 while (a <= z) { 3087 i = (a + z) / 2; 3088 d = qcmp_v6(&pr->pr_ip6[i+1], ia6); 3089 if (d > 0) 3090 z = i - 1; 3091 else if (d < 0) 3092 a = i + 1; 3093 else 3094 return (0); 3095 } 3096 3097 return (EADDRNOTAVAIL); 3098} 3099 3100int 3101prison_check_ip6(struct ucred *cred, struct in6_addr *ia6) 3102{ 3103 struct prison *pr; 3104 int error; 3105 3106 KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); 3107 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__)); 3108 3109 pr = cred->cr_prison; 3110 if (!(pr->pr_flags & PR_IP6)) 3111 return (0); 3112 mtx_lock(&pr->pr_mtx); 3113 if (!(pr->pr_flags & PR_IP6)) { 3114 mtx_unlock(&pr->pr_mtx); 3115 return (0); 3116 } 3117 if (pr->pr_ip6 == NULL) { 3118 mtx_unlock(&pr->pr_mtx); 3119 return (EAFNOSUPPORT); 3120 } 3121 3122 error = _prison_check_ip6(pr, ia6); 3123 mtx_unlock(&pr->pr_mtx); 3124 return (error); 3125} 3126#endif 3127 3128/* 3129 * Check if a jail supports the given address family. 3130 * 3131 * Returns 0 if not jailed or the address family is supported, EAFNOSUPPORT 3132 * if not. 3133 */ 3134int 3135prison_check_af(struct ucred *cred, int af) 3136{ 3137 struct prison *pr; 3138 int error; 3139 3140 KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); 3141 3142 pr = cred->cr_prison; 3143#ifdef VIMAGE 3144 /* Prisons with their own network stack are not limited. */ 3145 if (pr->pr_flags & PR_VNET) 3146 return (0); 3147#endif 3148 3149 error = 0; 3150 switch (af) 3151 { 3152#ifdef INET 3153 case AF_INET: 3154 if (pr->pr_flags & PR_IP4) 3155 { 3156 mtx_lock(&pr->pr_mtx); 3157 if ((pr->pr_flags & PR_IP4) && pr->pr_ip4 == NULL) 3158 error = EAFNOSUPPORT; 3159 mtx_unlock(&pr->pr_mtx); 3160 } 3161 break; 3162#endif 3163#ifdef INET6 3164 case AF_INET6: 3165 if (pr->pr_flags & PR_IP6) 3166 { 3167 mtx_lock(&pr->pr_mtx); 3168 if ((pr->pr_flags & PR_IP6) && pr->pr_ip6 == NULL) 3169 error = EAFNOSUPPORT; 3170 mtx_unlock(&pr->pr_mtx); 3171 } 3172 break; 3173#endif 3174 case AF_LOCAL: 3175 case AF_ROUTE: 3176 break; 3177 default: 3178 if (!(pr->pr_allow & PR_ALLOW_SOCKET_AF)) 3179 error = EAFNOSUPPORT; 3180 } 3181 return (error); 3182} 3183 3184/* 3185 * Check if given address belongs to the jail referenced by cred (wrapper to 3186 * prison_check_ip[46]). 3187 * 3188 * Returns 0 if jail doesn't restrict the address family or if address belongs 3189 * to jail, EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if 3190 * the jail doesn't allow the address family. IPv4 Address passed in in NBO. 3191 */ 3192int 3193prison_if(struct ucred *cred, struct sockaddr *sa) 3194{ 3195#ifdef INET 3196 struct sockaddr_in *sai; 3197#endif 3198#ifdef INET6 3199 struct sockaddr_in6 *sai6; 3200#endif 3201 int error; 3202 3203 KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); 3204 KASSERT(sa != NULL, ("%s: sa is NULL", __func__)); 3205 3206 error = 0; 3207 switch (sa->sa_family) 3208 { 3209#ifdef INET 3210 case AF_INET: 3211 sai = (struct sockaddr_in *)sa; 3212 error = prison_check_ip4(cred, &sai->sin_addr); 3213 break; 3214#endif 3215#ifdef INET6 3216 case AF_INET6: 3217 sai6 = (struct sockaddr_in6 *)sa; 3218 error = prison_check_ip6(cred, &sai6->sin6_addr); 3219 break; 3220#endif 3221 default: 3222 if (!(cred->cr_prison->pr_allow & PR_ALLOW_SOCKET_AF)) 3223 error = EAFNOSUPPORT; 3224 } 3225 return (error); 3226} 3227 3228/* 3229 * Return 0 if jails permit p1 to frob p2, otherwise ESRCH. 3230 */ 3231int 3232prison_check(struct ucred *cred1, struct ucred *cred2) 3233{ 3234 3235 return ((cred1->cr_prison == cred2->cr_prison || 3236 prison_ischild(cred1->cr_prison, cred2->cr_prison)) ? 0 : ESRCH); 3237} 3238 3239/* 3240 * Return 1 if p2 is a child of p1, otherwise 0. 3241 */ 3242int 3243prison_ischild(struct prison *pr1, struct prison *pr2) 3244{ 3245 3246 for (pr2 = pr2->pr_parent; pr2 != NULL; pr2 = pr2->pr_parent) 3247 if (pr1 == pr2) 3248 return (1); 3249 return (0); 3250} 3251 3252/* 3253 * Return 1 if the passed credential is in a jail, otherwise 0. 3254 */ 3255int 3256jailed(struct ucred *cred) 3257{ 3258 3259 return (cred->cr_prison != &prison0); 3260} 3261 3262/* 3263 * Return the correct hostname (domainname, et al) for the passed credential. 3264 */ 3265void 3266getcredhostname(struct ucred *cred, char *buf, size_t size) 3267{ 3268 struct prison *pr; 3269 3270 /* 3271 * A NULL credential can be used to shortcut to the physical 3272 * system's hostname. 3273 */ 3274 pr = (cred != NULL) ? cred->cr_prison : &prison0; 3275 mtx_lock(&pr->pr_mtx); 3276 strlcpy(buf, pr->pr_hostname, size); 3277 mtx_unlock(&pr->pr_mtx); 3278} 3279 3280void 3281getcreddomainname(struct ucred *cred, char *buf, size_t size) 3282{ 3283 3284 mtx_lock(&cred->cr_prison->pr_mtx); 3285 strlcpy(buf, cred->cr_prison->pr_domainname, size); 3286 mtx_unlock(&cred->cr_prison->pr_mtx); 3287} 3288 3289void 3290getcredhostuuid(struct ucred *cred, char *buf, size_t size) 3291{ 3292 3293 mtx_lock(&cred->cr_prison->pr_mtx); 3294 strlcpy(buf, cred->cr_prison->pr_hostuuid, size); 3295 mtx_unlock(&cred->cr_prison->pr_mtx); 3296} 3297 3298void 3299getcredhostid(struct ucred *cred, unsigned long *hostid) 3300{ 3301 3302 mtx_lock(&cred->cr_prison->pr_mtx); 3303 *hostid = cred->cr_prison->pr_hostid; 3304 mtx_unlock(&cred->cr_prison->pr_mtx); 3305} 3306 3307/* 3308 * Determine whether the subject represented by cred can "see" 3309 * status of a mount point. 3310 * Returns: 0 for permitted, ENOENT otherwise. 3311 * XXX: This function should be called cr_canseemount() and should be 3312 * placed in kern_prot.c. 3313 */ 3314int 3315prison_canseemount(struct ucred *cred, struct mount *mp) 3316{ 3317 struct prison *pr; 3318 struct statfs *sp; 3319 size_t len; 3320 3321 pr = cred->cr_prison; 3322 if (pr->pr_enforce_statfs == 0) 3323 return (0); 3324 if (pr->pr_root->v_mount == mp) 3325 return (0); 3326 if (pr->pr_enforce_statfs == 2) 3327 return (ENOENT); 3328 /* 3329 * If jail's chroot directory is set to "/" we should be able to see 3330 * all mount-points from inside a jail. 3331 * This is ugly check, but this is the only situation when jail's 3332 * directory ends with '/'. 3333 */ 3334 if (strcmp(pr->pr_path, "/") == 0) 3335 return (0); 3336 len = strlen(pr->pr_path); 3337 sp = &mp->mnt_stat; 3338 if (strncmp(pr->pr_path, sp->f_mntonname, len) != 0) 3339 return (ENOENT); 3340 /* 3341 * Be sure that we don't have situation where jail's root directory 3342 * is "/some/path" and mount point is "/some/pathpath". 3343 */ 3344 if (sp->f_mntonname[len] != '\0' && sp->f_mntonname[len] != '/') 3345 return (ENOENT); 3346 return (0); 3347} 3348 3349void 3350prison_enforce_statfs(struct ucred *cred, struct mount *mp, struct statfs *sp) 3351{ 3352 char jpath[MAXPATHLEN]; 3353 struct prison *pr; 3354 size_t len; 3355 3356 pr = cred->cr_prison; 3357 if (pr->pr_enforce_statfs == 0) 3358 return; 3359 if (prison_canseemount(cred, mp) != 0) { 3360 bzero(sp->f_mntonname, sizeof(sp->f_mntonname)); 3361 strlcpy(sp->f_mntonname, "[restricted]", 3362 sizeof(sp->f_mntonname)); 3363 return; 3364 } 3365 if (pr->pr_root->v_mount == mp) { 3366 /* 3367 * Clear current buffer data, so we are sure nothing from 3368 * the valid path left there. 3369 */ 3370 bzero(sp->f_mntonname, sizeof(sp->f_mntonname)); 3371 *sp->f_mntonname = '/'; 3372 return; 3373 } 3374 /* 3375 * If jail's chroot directory is set to "/" we should be able to see 3376 * all mount-points from inside a jail. 3377 */ 3378 if (strcmp(pr->pr_path, "/") == 0) 3379 return; 3380 len = strlen(pr->pr_path); 3381 strlcpy(jpath, sp->f_mntonname + len, sizeof(jpath)); 3382 /* 3383 * Clear current buffer data, so we are sure nothing from 3384 * the valid path left there. 3385 */ 3386 bzero(sp->f_mntonname, sizeof(sp->f_mntonname)); 3387 if (*jpath == '\0') { 3388 /* Should never happen. */ 3389 *sp->f_mntonname = '/'; 3390 } else { 3391 strlcpy(sp->f_mntonname, jpath, sizeof(sp->f_mntonname)); 3392 } 3393} 3394 3395/* 3396 * Check with permission for a specific privilege is granted within jail. We 3397 * have a specific list of accepted privileges; the rest are denied. 3398 */ 3399int 3400prison_priv_check(struct ucred *cred, int priv) 3401{ 3402 3403 if (!jailed(cred)) 3404 return (0); 3405 3406#ifdef VIMAGE 3407 /* 3408 * Privileges specific to prisons with a virtual network stack. 3409 * There might be a duplicate entry here in case the privilege 3410 * is only granted conditionally in the legacy jail case. 3411 */ 3412 switch (priv) { 3413#ifdef notyet 3414 /* 3415 * NFS-specific privileges. 3416 */ 3417 case PRIV_NFS_DAEMON: 3418 case PRIV_NFS_LOCKD: 3419#endif 3420 /* 3421 * Network stack privileges. 3422 */ 3423 case PRIV_NET_BRIDGE: 3424 case PRIV_NET_GRE: 3425 case PRIV_NET_BPF: 3426 case PRIV_NET_RAW: /* Dup, cond. in legacy jail case. */ 3427 case PRIV_NET_ROUTE: 3428 case PRIV_NET_TAP: 3429 case PRIV_NET_SETIFMTU: 3430 case PRIV_NET_SETIFFLAGS: 3431 case PRIV_NET_SETIFCAP: 3432 case PRIV_NET_SETIFNAME : 3433 case PRIV_NET_SETIFMETRIC: 3434 case PRIV_NET_SETIFPHYS: 3435 case PRIV_NET_SETIFMAC: 3436 case PRIV_NET_ADDMULTI: 3437 case PRIV_NET_DELMULTI: 3438 case PRIV_NET_HWIOCTL: 3439 case PRIV_NET_SETLLADDR: 3440 case PRIV_NET_ADDIFGROUP: 3441 case PRIV_NET_DELIFGROUP: 3442 case PRIV_NET_IFCREATE: 3443 case PRIV_NET_IFDESTROY: 3444 case PRIV_NET_ADDIFADDR: 3445 case PRIV_NET_DELIFADDR: 3446 case PRIV_NET_LAGG: 3447 case PRIV_NET_GIF: 3448 case PRIV_NET_SETIFVNET: 3449 3450 /* 3451 * 802.11-related privileges. 3452 */ 3453 case PRIV_NET80211_GETKEY: 3454#ifdef notyet 3455 case PRIV_NET80211_MANAGE: /* XXX-BZ discuss with sam@ */ 3456#endif 3457 3458#ifdef notyet 3459 /* 3460 * AppleTalk privileges. 3461 */ 3462 case PRIV_NETATALK_RESERVEDPORT: 3463 3464 /* 3465 * ATM privileges. 3466 */ 3467 case PRIV_NETATM_CFG: 3468 case PRIV_NETATM_ADD: 3469 case PRIV_NETATM_DEL: 3470 case PRIV_NETATM_SET: 3471 3472 /* 3473 * Bluetooth privileges. 3474 */ 3475 case PRIV_NETBLUETOOTH_RAW: 3476#endif 3477 3478 /* 3479 * Netgraph and netgraph module privileges. 3480 */ 3481 case PRIV_NETGRAPH_CONTROL: 3482#ifdef notyet 3483 case PRIV_NETGRAPH_TTY: 3484#endif 3485 3486 /* 3487 * IPv4 and IPv6 privileges. 3488 */ 3489 case PRIV_NETINET_IPFW: 3490 case PRIV_NETINET_DIVERT: 3491 case PRIV_NETINET_PF: 3492 case PRIV_NETINET_DUMMYNET: 3493 case PRIV_NETINET_CARP: 3494 case PRIV_NETINET_MROUTE: 3495 case PRIV_NETINET_RAW: 3496 case PRIV_NETINET_ADDRCTRL6: 3497 case PRIV_NETINET_ND6: 3498 case PRIV_NETINET_SCOPE6: 3499 case PRIV_NETINET_ALIFETIME6: 3500 case PRIV_NETINET_IPSEC: 3501 case PRIV_NETINET_BINDANY: 3502 3503#ifdef notyet 3504 /* 3505 * IPX/SPX privileges. 3506 */ 3507 case PRIV_NETIPX_RESERVEDPORT: 3508 case PRIV_NETIPX_RAW: 3509 3510 /* 3511 * NCP privileges. 3512 */ 3513 case PRIV_NETNCP: 3514 3515 /* 3516 * SMB privileges. 3517 */ 3518 case PRIV_NETSMB: 3519#endif 3520 3521 /* 3522 * No default: or deny here. 3523 * In case of no permit fall through to next switch(). 3524 */ 3525 if (cred->cr_prison->pr_flags & PR_VNET) 3526 return (0); 3527 } 3528#endif /* VIMAGE */ 3529 3530 switch (priv) { 3531 3532 /* 3533 * Allow ktrace privileges for root in jail. 3534 */ 3535 case PRIV_KTRACE: 3536 3537#if 0 3538 /* 3539 * Allow jailed processes to configure audit identity and 3540 * submit audit records (login, etc). In the future we may 3541 * want to further refine the relationship between audit and 3542 * jail. 3543 */ 3544 case PRIV_AUDIT_GETAUDIT: 3545 case PRIV_AUDIT_SETAUDIT: 3546 case PRIV_AUDIT_SUBMIT: 3547#endif 3548 3549 /* 3550 * Allow jailed processes to manipulate process UNIX 3551 * credentials in any way they see fit. 3552 */ 3553 case PRIV_CRED_SETUID: 3554 case PRIV_CRED_SETEUID: 3555 case PRIV_CRED_SETGID: 3556 case PRIV_CRED_SETEGID: 3557 case PRIV_CRED_SETGROUPS: 3558 case PRIV_CRED_SETREUID: 3559 case PRIV_CRED_SETREGID: 3560 case PRIV_CRED_SETRESUID: 3561 case PRIV_CRED_SETRESGID: 3562 3563 /* 3564 * Jail implements visibility constraints already, so allow 3565 * jailed root to override uid/gid-based constraints. 3566 */ 3567 case PRIV_SEEOTHERGIDS: 3568 case PRIV_SEEOTHERUIDS: 3569 3570 /* 3571 * Jail implements inter-process debugging limits already, so 3572 * allow jailed root various debugging privileges. 3573 */ 3574 case PRIV_DEBUG_DIFFCRED: 3575 case PRIV_DEBUG_SUGID: 3576 case PRIV_DEBUG_UNPRIV: 3577 3578 /* 3579 * Allow jail to set various resource limits and login 3580 * properties, and for now, exceed process resource limits. 3581 */ 3582 case PRIV_PROC_LIMIT: 3583 case PRIV_PROC_SETLOGIN: 3584 case PRIV_PROC_SETRLIMIT: 3585 3586 /* 3587 * System V and POSIX IPC privileges are granted in jail. 3588 */ 3589 case PRIV_IPC_READ: 3590 case PRIV_IPC_WRITE: 3591 case PRIV_IPC_ADMIN: 3592 case PRIV_IPC_MSGSIZE: 3593 case PRIV_MQ_ADMIN: 3594 3595 /* 3596 * Jail operations within a jail work on child jails. 3597 */ 3598 case PRIV_JAIL_ATTACH: 3599 case PRIV_JAIL_SET: 3600 case PRIV_JAIL_REMOVE: 3601 3602 /* 3603 * Jail implements its own inter-process limits, so allow 3604 * root processes in jail to change scheduling on other 3605 * processes in the same jail. Likewise for signalling. 3606 */ 3607 case PRIV_SCHED_DIFFCRED: 3608 case PRIV_SCHED_CPUSET: 3609 case PRIV_SIGNAL_DIFFCRED: 3610 case PRIV_SIGNAL_SUGID: 3611 3612 /* 3613 * Allow jailed processes to write to sysctls marked as jail 3614 * writable. 3615 */ 3616 case PRIV_SYSCTL_WRITEJAIL: 3617 3618 /* 3619 * Allow root in jail to manage a variety of quota 3620 * properties. These should likely be conditional on a 3621 * configuration option. 3622 */ 3623 case PRIV_VFS_GETQUOTA: 3624 case PRIV_VFS_SETQUOTA: 3625 3626 /* 3627 * Since Jail relies on chroot() to implement file system 3628 * protections, grant many VFS privileges to root in jail. 3629 * Be careful to exclude mount-related and NFS-related 3630 * privileges. 3631 */ 3632 case PRIV_VFS_READ: 3633 case PRIV_VFS_WRITE: 3634 case PRIV_VFS_ADMIN: 3635 case PRIV_VFS_EXEC: 3636 case PRIV_VFS_LOOKUP: 3637 case PRIV_VFS_BLOCKRESERVE: /* XXXRW: Slightly surprising. */ 3638 case PRIV_VFS_CHFLAGS_DEV: 3639 case PRIV_VFS_CHOWN: 3640 case PRIV_VFS_CHROOT: 3641 case PRIV_VFS_RETAINSUGID: 3642 case PRIV_VFS_FCHROOT: 3643 case PRIV_VFS_LINK: 3644 case PRIV_VFS_SETGID: 3645 case PRIV_VFS_STAT: 3646 case PRIV_VFS_STICKYFILE: 3647 return (0); 3648 3649 /* 3650 * Depending on the global setting, allow privilege of 3651 * setting system flags. 3652 */ 3653 case PRIV_VFS_SYSFLAGS: 3654 if (cred->cr_prison->pr_allow & PR_ALLOW_CHFLAGS) 3655 return (0); 3656 else 3657 return (EPERM); 3658 3659 /* 3660 * Depending on the global setting, allow privilege of 3661 * mounting/unmounting file systems. 3662 */ 3663 case PRIV_VFS_MOUNT: 3664 case PRIV_VFS_UNMOUNT: 3665 case PRIV_VFS_MOUNT_NONUSER: 3666 case PRIV_VFS_MOUNT_OWNER: 3667 if (cred->cr_prison->pr_allow & PR_ALLOW_MOUNT) 3668 return (0); 3669 else 3670 return (EPERM); 3671 3672 /* 3673 * Allow jailed root to bind reserved ports and reuse in-use 3674 * ports. 3675 */ 3676 case PRIV_NETINET_RESERVEDPORT: 3677 case PRIV_NETINET_REUSEPORT: 3678 return (0); 3679 3680 /* 3681 * Allow jailed root to set certian IPv4/6 (option) headers. 3682 */ 3683 case PRIV_NETINET_SETHDROPTS: 3684 return (0); 3685 3686 /* 3687 * Conditionally allow creating raw sockets in jail. 3688 */ 3689 case PRIV_NETINET_RAW: 3690 if (cred->cr_prison->pr_allow & PR_ALLOW_RAW_SOCKETS) 3691 return (0); 3692 else 3693 return (EPERM); 3694 3695 /* 3696 * Since jail implements its own visibility limits on netstat 3697 * sysctls, allow getcred. This allows identd to work in 3698 * jail. 3699 */ 3700 case PRIV_NETINET_GETCRED: 3701 return (0); 3702 3703 default: 3704 /* 3705 * In all remaining cases, deny the privilege request. This 3706 * includes almost all network privileges, many system 3707 * configuration privileges. 3708 */ 3709 return (EPERM); 3710 } 3711} 3712 3713/* 3714 * Return the part of pr2's name that is relative to pr1, or the whole name 3715 * if it does not directly follow. 3716 */ 3717 3718char * 3719prison_name(struct prison *pr1, struct prison *pr2) 3720{ 3721 char *name; 3722 3723 /* Jails see themselves as "0" (if they see themselves at all). */ 3724 if (pr1 == pr2) 3725 return "0"; 3726 name = pr2->pr_name; 3727 if (prison_ischild(pr1, pr2)) { 3728 /* 3729 * pr1 isn't locked (and allprison_lock may not be either) 3730 * so its length can't be counted on. But the number of dots 3731 * can be counted on - and counted. 3732 */ 3733 for (; pr1 != &prison0; pr1 = pr1->pr_parent) 3734 name = strchr(name, '.') + 1; 3735 } 3736 return (name); 3737} 3738 3739/* 3740 * Return the part of pr2's path that is relative to pr1, or the whole path 3741 * if it does not directly follow. 3742 */ 3743static char * 3744prison_path(struct prison *pr1, struct prison *pr2) 3745{ 3746 char *path1, *path2; 3747 int len1; 3748 3749 path1 = pr1->pr_path; 3750 path2 = pr2->pr_path; 3751 if (!strcmp(path1, "/")) 3752 return (path2); 3753 len1 = strlen(path1); 3754 if (strncmp(path1, path2, len1)) 3755 return (path2); 3756 if (path2[len1] == '\0') 3757 return "/"; 3758 if (path2[len1] == '/') 3759 return (path2 + len1); 3760 return (path2); 3761} 3762 3763 3764/* 3765 * Jail-related sysctls. 3766 */ 3767SYSCTL_NODE(_security, OID_AUTO, jail, CTLFLAG_RW, 0, 3768 "Jails"); 3769 3770static int 3771sysctl_jail_list(SYSCTL_HANDLER_ARGS) 3772{ 3773 struct xprison *xp; 3774 struct prison *pr, *cpr; 3775#ifdef INET 3776 struct in_addr *ip4 = NULL; 3777 int ip4s = 0; 3778#endif 3779#ifdef INET6 3780 struct in_addr *ip6 = NULL; 3781 int ip6s = 0; 3782#endif 3783 int descend, error; 3784 3785 xp = malloc(sizeof(*xp), M_TEMP, M_WAITOK); 3786 pr = req->td->td_ucred->cr_prison; 3787 error = 0; 3788 sx_slock(&allprison_lock); 3789 FOREACH_PRISON_DESCENDANT(pr, cpr, descend) { 3790#if defined(INET) || defined(INET6) 3791 again: 3792#endif 3793 mtx_lock(&cpr->pr_mtx); 3794#ifdef INET 3795 if (cpr->pr_ip4s > 0) { 3796 if (ip4s < cpr->pr_ip4s) { 3797 ip4s = cpr->pr_ip4s; 3798 mtx_unlock(&cpr->pr_mtx); 3799 ip4 = realloc(ip4, ip4s * 3800 sizeof(struct in_addr), M_TEMP, M_WAITOK); 3801 goto again; 3802 } 3803 bcopy(cpr->pr_ip4, ip4, 3804 cpr->pr_ip4s * sizeof(struct in_addr)); 3805 } 3806#endif 3807#ifdef INET6 3808 if (cpr->pr_ip6s > 0) { 3809 if (ip6s < cpr->pr_ip6s) { 3810 ip6s = cpr->pr_ip6s; 3811 mtx_unlock(&cpr->pr_mtx); 3812 ip6 = realloc(ip6, ip6s * 3813 sizeof(struct in6_addr), M_TEMP, M_WAITOK); 3814 goto again; 3815 } 3816 bcopy(cpr->pr_ip6, ip6, 3817 cpr->pr_ip6s * sizeof(struct in6_addr)); 3818 } 3819#endif 3820 if (cpr->pr_ref == 0) { 3821 mtx_unlock(&cpr->pr_mtx); 3822 continue; 3823 } 3824 bzero(xp, sizeof(*xp)); 3825 xp->pr_version = XPRISON_VERSION; 3826 xp->pr_id = cpr->pr_id; 3827 xp->pr_state = cpr->pr_uref > 0 3828 ? PRISON_STATE_ALIVE : PRISON_STATE_DYING; 3829 strlcpy(xp->pr_path, prison_path(pr, cpr), sizeof(xp->pr_path)); 3830 strlcpy(xp->pr_host, cpr->pr_hostname, sizeof(xp->pr_host)); 3831 strlcpy(xp->pr_name, prison_name(pr, cpr), sizeof(xp->pr_name)); 3832#ifdef INET 3833 xp->pr_ip4s = cpr->pr_ip4s; 3834#endif 3835#ifdef INET6 3836 xp->pr_ip6s = cpr->pr_ip6s; 3837#endif 3838 mtx_unlock(&cpr->pr_mtx); 3839 error = SYSCTL_OUT(req, xp, sizeof(*xp)); 3840 if (error) 3841 break; 3842#ifdef INET 3843 if (xp->pr_ip4s > 0) { 3844 error = SYSCTL_OUT(req, ip4, 3845 xp->pr_ip4s * sizeof(struct in_addr)); 3846 if (error) 3847 break; 3848 } 3849#endif 3850#ifdef INET6 3851 if (xp->pr_ip6s > 0) { 3852 error = SYSCTL_OUT(req, ip6, 3853 xp->pr_ip6s * sizeof(struct in6_addr)); 3854 if (error) 3855 break; 3856 } 3857#endif 3858 } 3859 sx_sunlock(&allprison_lock); 3860 free(xp, M_TEMP); 3861#ifdef INET 3862 free(ip4, M_TEMP); 3863#endif 3864#ifdef INET6 3865 free(ip6, M_TEMP); 3866#endif 3867 return (error); 3868} 3869 3870SYSCTL_OID(_security_jail, OID_AUTO, list, 3871 CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0, 3872 sysctl_jail_list, "S", "List of active jails"); 3873 3874static int 3875sysctl_jail_jailed(SYSCTL_HANDLER_ARGS) 3876{ 3877 int error, injail; 3878 3879 injail = jailed(req->td->td_ucred); 3880 error = SYSCTL_OUT(req, &injail, sizeof(injail)); 3881 3882 return (error); 3883} 3884 3885SYSCTL_PROC(_security_jail, OID_AUTO, jailed, 3886 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0, 3887 sysctl_jail_jailed, "I", "Process in jail?"); 3888 3889#if defined(INET) || defined(INET6) 3890SYSCTL_UINT(_security_jail, OID_AUTO, jail_max_af_ips, CTLFLAG_RW, 3891 &jail_max_af_ips, 0, 3892 "Number of IP addresses a jail may have at most per address family"); 3893#endif 3894 3895/* 3896 * Default parameters for jail(2) compatability. For historical reasons, 3897 * the sysctl names have varying similarity to the parameter names. Prisons 3898 * just see their own parameters, and can't change them. 3899 */ 3900static int 3901sysctl_jail_default_allow(SYSCTL_HANDLER_ARGS) 3902{ 3903 struct prison *pr; 3904 int allow, error, i; 3905 3906 pr = req->td->td_ucred->cr_prison; 3907 allow = (pr == &prison0) ? jail_default_allow : pr->pr_allow; 3908 3909 /* Get the current flag value, and convert it to a boolean. */ 3910 i = (allow & arg2) ? 1 : 0; 3911 if (arg1 != NULL) 3912 i = !i; 3913 error = sysctl_handle_int(oidp, &i, 0, req); 3914 if (error || !req->newptr) 3915 return (error); 3916 i = i ? arg2 : 0; 3917 if (arg1 != NULL) 3918 i ^= arg2; 3919 /* 3920 * The sysctls don't have CTLFLAGS_PRISON, so assume prison0 3921 * for writing. 3922 */ 3923 mtx_lock(&prison0.pr_mtx); 3924 jail_default_allow = (jail_default_allow & ~arg2) | i; 3925 mtx_unlock(&prison0.pr_mtx); 3926 return (0); 3927} 3928 3929SYSCTL_PROC(_security_jail, OID_AUTO, set_hostname_allowed, 3930 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 3931 NULL, PR_ALLOW_SET_HOSTNAME, sysctl_jail_default_allow, "I", 3932 "Processes in jail can set their hostnames"); 3933SYSCTL_PROC(_security_jail, OID_AUTO, socket_unixiproute_only, 3934 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 3935 (void *)1, PR_ALLOW_SOCKET_AF, sysctl_jail_default_allow, "I", 3936 "Processes in jail are limited to creating UNIX/IP/route sockets only"); 3937SYSCTL_PROC(_security_jail, OID_AUTO, sysvipc_allowed, 3938 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 3939 NULL, PR_ALLOW_SYSVIPC, sysctl_jail_default_allow, "I", 3940 "Processes in jail can use System V IPC primitives"); 3941SYSCTL_PROC(_security_jail, OID_AUTO, allow_raw_sockets, 3942 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 3943 NULL, PR_ALLOW_RAW_SOCKETS, sysctl_jail_default_allow, "I", 3944 "Prison root can create raw sockets"); 3945SYSCTL_PROC(_security_jail, OID_AUTO, chflags_allowed, 3946 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 3947 NULL, PR_ALLOW_CHFLAGS, sysctl_jail_default_allow, "I", 3948 "Processes in jail can alter system file flags"); 3949SYSCTL_PROC(_security_jail, OID_AUTO, mount_allowed, 3950 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 3951 NULL, PR_ALLOW_MOUNT, sysctl_jail_default_allow, "I", 3952 "Processes in jail can mount/unmount jail-friendly file systems"); 3953 3954static int 3955sysctl_jail_default_level(SYSCTL_HANDLER_ARGS) 3956{ 3957 struct prison *pr; 3958 int level, error; 3959 3960 pr = req->td->td_ucred->cr_prison; 3961 level = (pr == &prison0) ? *(int *)arg1 : *(int *)((char *)pr + arg2); 3962 error = sysctl_handle_int(oidp, &level, 0, req); 3963 if (error || !req->newptr) 3964 return (error); 3965 *(int *)arg1 = level; 3966 return (0); 3967} 3968 3969SYSCTL_PROC(_security_jail, OID_AUTO, enforce_statfs, 3970 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 3971 &jail_default_enforce_statfs, offsetof(struct prison, pr_enforce_statfs), 3972 sysctl_jail_default_level, "I", 3973 "Processes in jail cannot see all mounted file systems"); 3974 3975/* 3976 * Nodes to describe jail parameters. Maximum length of string parameters 3977 * is returned in the string itself, and the other parameters exist merely 3978 * to make themselves and their types known. 3979 */ 3980SYSCTL_NODE(_security_jail, OID_AUTO, param, CTLFLAG_RW, 0, 3981 "Jail parameters"); 3982 3983int 3984sysctl_jail_param(SYSCTL_HANDLER_ARGS) 3985{ 3986 int i; 3987 long l; 3988 size_t s; 3989 char numbuf[12]; 3990 3991 switch (oidp->oid_kind & CTLTYPE) 3992 { 3993 case CTLTYPE_LONG: 3994 case CTLTYPE_ULONG: 3995 l = 0; 3996#ifdef SCTL_MASK32 3997 if (!(req->flags & SCTL_MASK32)) 3998#endif 3999 return (SYSCTL_OUT(req, &l, sizeof(l))); 4000 case CTLTYPE_INT: 4001 case CTLTYPE_UINT: 4002 i = 0; 4003 return (SYSCTL_OUT(req, &i, sizeof(i))); 4004 case CTLTYPE_STRING: 4005 snprintf(numbuf, sizeof(numbuf), "%d", arg2); 4006 return 4007 (sysctl_handle_string(oidp, numbuf, sizeof(numbuf), req)); 4008 case CTLTYPE_STRUCT: 4009 s = (size_t)arg2; 4010 return (SYSCTL_OUT(req, &s, sizeof(s))); 4011 } 4012 return (0); 4013} 4014 4015SYSCTL_JAIL_PARAM(, jid, CTLTYPE_INT | CTLFLAG_RDTUN, "I", "Jail ID"); 4016SYSCTL_JAIL_PARAM(, parent, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail parent ID"); 4017SYSCTL_JAIL_PARAM_STRING(, name, CTLFLAG_RW, MAXHOSTNAMELEN, "Jail name"); 4018SYSCTL_JAIL_PARAM_STRING(, path, CTLFLAG_RDTUN, MAXPATHLEN, "Jail root path"); 4019SYSCTL_JAIL_PARAM(, securelevel, CTLTYPE_INT | CTLFLAG_RW, 4020 "I", "Jail secure level"); 4021SYSCTL_JAIL_PARAM(, enforce_statfs, CTLTYPE_INT | CTLFLAG_RW, 4022 "I", "Jail cannot see all mounted file systems"); 4023SYSCTL_JAIL_PARAM(, persist, CTLTYPE_INT | CTLFLAG_RW, 4024 "B", "Jail persistence"); 4025#ifdef VIMAGE 4026SYSCTL_JAIL_PARAM(, vnet, CTLTYPE_INT | CTLFLAG_RDTUN, 4027 "E,jailsys", "Virtual network stack"); 4028#endif 4029SYSCTL_JAIL_PARAM(, dying, CTLTYPE_INT | CTLFLAG_RD, 4030 "B", "Jail is in the process of shutting down"); 4031 4032SYSCTL_JAIL_PARAM_NODE(children, "Number of child jails"); 4033SYSCTL_JAIL_PARAM(_children, cur, CTLTYPE_INT | CTLFLAG_RD, 4034 "I", "Current number of child jails"); 4035SYSCTL_JAIL_PARAM(_children, max, CTLTYPE_INT | CTLFLAG_RW, 4036 "I", "Maximum number of child jails"); 4037 4038SYSCTL_JAIL_PARAM_SYS_NODE(host, CTLFLAG_RW, "Jail host info"); 4039SYSCTL_JAIL_PARAM_STRING(_host, hostname, CTLFLAG_RW, MAXHOSTNAMELEN, 4040 "Jail hostname"); 4041SYSCTL_JAIL_PARAM_STRING(_host, domainname, CTLFLAG_RW, MAXHOSTNAMELEN, 4042 "Jail NIS domainname"); 4043SYSCTL_JAIL_PARAM_STRING(_host, hostuuid, CTLFLAG_RW, HOSTUUIDLEN, 4044 "Jail host UUID"); 4045SYSCTL_JAIL_PARAM(_host, hostid, CTLTYPE_ULONG | CTLFLAG_RW, 4046 "LU", "Jail host ID"); 4047 4048SYSCTL_JAIL_PARAM_NODE(cpuset, "Jail cpuset"); 4049SYSCTL_JAIL_PARAM(_cpuset, id, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail cpuset ID"); 4050 4051#ifdef INET 4052SYSCTL_JAIL_PARAM_SYS_NODE(ip4, CTLFLAG_RDTUN, 4053 "Jail IPv4 address virtualization"); 4054SYSCTL_JAIL_PARAM_STRUCT(_ip4, addr, CTLFLAG_RW, sizeof(struct in_addr), 4055 "S,in_addr,a", "Jail IPv4 addresses"); 4056#endif 4057#ifdef INET6 4058SYSCTL_JAIL_PARAM_SYS_NODE(ip6, CTLFLAG_RDTUN, 4059 "Jail IPv6 address virtualization"); 4060SYSCTL_JAIL_PARAM_STRUCT(_ip6, addr, CTLFLAG_RW, sizeof(struct in6_addr), 4061 "S,in6_addr,a", "Jail IPv6 addresses"); 4062#endif 4063 4064SYSCTL_JAIL_PARAM_NODE(allow, "Jail permission flags"); 4065SYSCTL_JAIL_PARAM(_allow, set_hostname, CTLTYPE_INT | CTLFLAG_RW, 4066 "B", "Jail may set hostname"); 4067SYSCTL_JAIL_PARAM(_allow, sysvipc, CTLTYPE_INT | CTLFLAG_RW, 4068 "B", "Jail may use SYSV IPC"); 4069SYSCTL_JAIL_PARAM(_allow, raw_sockets, CTLTYPE_INT | CTLFLAG_RW, 4070 "B", "Jail may create raw sockets"); 4071SYSCTL_JAIL_PARAM(_allow, chflags, CTLTYPE_INT | CTLFLAG_RW, 4072 "B", "Jail may alter system file flags"); 4073SYSCTL_JAIL_PARAM(_allow, mount, CTLTYPE_INT | CTLFLAG_RW, 4074 "B", "Jail may mount/unmount jail-friendly file systems"); 4075SYSCTL_JAIL_PARAM(_allow, quotas, CTLTYPE_INT | CTLFLAG_RW, 4076 "B", "Jail may set file quotas"); 4077SYSCTL_JAIL_PARAM(_allow, socket_af, CTLTYPE_INT | CTLFLAG_RW, 4078 "B", "Jail may create sockets other than just UNIX/IPv4/IPv6/route"); 4079 4080 4081#ifdef DDB 4082 4083static void 4084db_show_prison(struct prison *pr) 4085{ 4086 int fi; 4087#if defined(INET) || defined(INET6) 4088 int ii; 4089#endif 4090 unsigned jsf; 4091#ifdef INET6 4092 char ip6buf[INET6_ADDRSTRLEN]; 4093#endif 4094 4095 db_printf("prison %p:\n", pr); 4096 db_printf(" jid = %d\n", pr->pr_id); 4097 db_printf(" name = %s\n", pr->pr_name); 4098 db_printf(" parent = %p\n", pr->pr_parent); 4099 db_printf(" ref = %d\n", pr->pr_ref); 4100 db_printf(" uref = %d\n", pr->pr_uref); 4101 db_printf(" path = %s\n", pr->pr_path); 4102 db_printf(" cpuset = %d\n", pr->pr_cpuset 4103 ? pr->pr_cpuset->cs_id : -1); 4104#ifdef VIMAGE 4105 db_printf(" vnet = %p\n", pr->pr_vnet); 4106#endif 4107 db_printf(" root = %p\n", pr->pr_root); 4108 db_printf(" securelevel = %d\n", pr->pr_securelevel); 4109 db_printf(" childcount = %d\n", pr->pr_childcount); 4110 db_printf(" child = %p\n", LIST_FIRST(&pr->pr_children)); 4111 db_printf(" sibling = %p\n", LIST_NEXT(pr, pr_sibling)); 4112 db_printf(" flags = %x", pr->pr_flags); 4113 for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]); 4114 fi++) 4115 if (pr_flag_names[fi] != NULL && (pr->pr_flags & (1 << fi))) 4116 db_printf(" %s", pr_flag_names[fi]); 4117 for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]); 4118 fi++) { 4119 jsf = pr->pr_flags & 4120 (pr_flag_jailsys[fi].disable | pr_flag_jailsys[fi].new); 4121 db_printf(" %-16s= %s\n", pr_flag_jailsys[fi].name, 4122 pr_flag_jailsys[fi].disable && 4123 (jsf == pr_flag_jailsys[fi].disable) ? "disable" 4124 : (jsf == pr_flag_jailsys[fi].new) ? "new" 4125 : "inherit"); 4126 } 4127 db_printf(" allow = %x", pr->pr_allow); 4128 for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]); 4129 fi++) 4130 if (pr_allow_names[fi] != NULL && (pr->pr_allow & (1 << fi))) 4131 db_printf(" %s", pr_allow_names[fi]); 4132 db_printf("\n"); 4133 db_printf(" enforce_statfs = %d\n", pr->pr_enforce_statfs); 4134 db_printf(" host.hostname = %s\n", pr->pr_hostname); 4135 db_printf(" host.domainname = %s\n", pr->pr_domainname); 4136 db_printf(" host.hostuuid = %s\n", pr->pr_hostuuid); 4137 db_printf(" host.hostid = %lu\n", pr->pr_hostid); 4138#ifdef INET 4139 db_printf(" ip4s = %d\n", pr->pr_ip4s); 4140 for (ii = 0; ii < pr->pr_ip4s; ii++) 4141 db_printf(" %s %s\n", 4142 ii == 0 ? "ip4 =" : " ", 4143 inet_ntoa(pr->pr_ip4[ii])); 4144#endif 4145#ifdef INET6 4146 db_printf(" ip6s = %d\n", pr->pr_ip6s); 4147 for (ii = 0; ii < pr->pr_ip6s; ii++) 4148 db_printf(" %s %s\n", 4149 ii == 0 ? "ip6 =" : " ", 4150 ip6_sprintf(ip6buf, &pr->pr_ip6[ii])); 4151#endif 4152} 4153 4154DB_SHOW_COMMAND(prison, db_show_prison_command) 4155{ 4156 struct prison *pr; 4157 4158 if (!have_addr) { 4159 /* 4160 * Show all prisons in the list, and prison0 which is not 4161 * listed. 4162 */ 4163 db_show_prison(&prison0); 4164 if (!db_pager_quit) { 4165 TAILQ_FOREACH(pr, &allprison, pr_list) { 4166 db_show_prison(pr); 4167 if (db_pager_quit) 4168 break; 4169 } 4170 } 4171 return; 4172 } 4173 4174 if (addr == 0) 4175 pr = &prison0; 4176 else { 4177 /* Look for a prison with the ID and with references. */ 4178 TAILQ_FOREACH(pr, &allprison, pr_list) 4179 if (pr->pr_id == addr && pr->pr_ref > 0) 4180 break; 4181 if (pr == NULL) 4182 /* Look again, without requiring a reference. */ 4183 TAILQ_FOREACH(pr, &allprison, pr_list) 4184 if (pr->pr_id == addr) 4185 break; 4186 if (pr == NULL) 4187 /* Assume address points to a valid prison. */ 4188 pr = (struct prison *)addr; 4189 } 4190 db_show_prison(pr); 4191} 4192 4193#endif /* DDB */ 4194