kern_jail.c revision 258929
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: stable/10/sys/kern/kern_jail.c 258929 2013-12-04 18:25:04Z peter $"); 31 32#include "opt_compat.h" 33#include "opt_ddb.h" 34#include "opt_inet.h" 35#include "opt_inet6.h" 36 37#include <sys/param.h> 38#include <sys/types.h> 39#include <sys/kernel.h> 40#include <sys/systm.h> 41#include <sys/errno.h> 42#include <sys/sysproto.h> 43#include <sys/malloc.h> 44#include <sys/osd.h> 45#include <sys/priv.h> 46#include <sys/proc.h> 47#include <sys/taskqueue.h> 48#include <sys/fcntl.h> 49#include <sys/jail.h> 50#include <sys/lock.h> 51#include <sys/mutex.h> 52#include <sys/racct.h> 53#include <sys/refcount.h> 54#include <sys/sx.h> 55#include <sys/sysent.h> 56#include <sys/namei.h> 57#include <sys/mount.h> 58#include <sys/queue.h> 59#include <sys/socket.h> 60#include <sys/syscallsubr.h> 61#include <sys/sysctl.h> 62#include <sys/vnode.h> 63 64#include <net/if.h> 65#include <net/vnet.h> 66 67#include <netinet/in.h> 68 69#ifdef DDB 70#include <ddb/ddb.h> 71#ifdef INET6 72#include <netinet6/in6_var.h> 73#endif /* INET6 */ 74#endif /* DDB */ 75 76#include <security/mac/mac_framework.h> 77 78#define DEFAULT_HOSTUUID "00000000-0000-0000-0000-000000000000" 79 80MALLOC_DEFINE(M_PRISON, "prison", "Prison structures"); 81static MALLOC_DEFINE(M_PRISON_RACCT, "prison_racct", "Prison racct structures"); 82 83/* Keep struct prison prison0 and some code in kern_jail_set() readable. */ 84#ifdef INET 85#ifdef INET6 86#define _PR_IP_SADDRSEL PR_IP4_SADDRSEL|PR_IP6_SADDRSEL 87#else 88#define _PR_IP_SADDRSEL PR_IP4_SADDRSEL 89#endif 90#else /* !INET */ 91#ifdef INET6 92#define _PR_IP_SADDRSEL PR_IP6_SADDRSEL 93#else 94#define _PR_IP_SADDRSEL 0 95#endif 96#endif 97 98/* prison0 describes what is "real" about the system. */ 99struct prison prison0 = { 100 .pr_id = 0, 101 .pr_name = "0", 102 .pr_ref = 1, 103 .pr_uref = 1, 104 .pr_path = "/", 105 .pr_securelevel = -1, 106 .pr_devfs_rsnum = 0, 107 .pr_childmax = JAIL_MAX, 108 .pr_hostuuid = DEFAULT_HOSTUUID, 109 .pr_children = LIST_HEAD_INITIALIZER(prison0.pr_children), 110#ifdef VIMAGE 111 .pr_flags = PR_HOST|PR_VNET|_PR_IP_SADDRSEL, 112#else 113 .pr_flags = PR_HOST|_PR_IP_SADDRSEL, 114#endif 115 .pr_allow = PR_ALLOW_ALL, 116}; 117MTX_SYSINIT(prison0, &prison0.pr_mtx, "jail mutex", MTX_DEF); 118 119/* allprison, allprison_racct and lastprid are protected by allprison_lock. */ 120struct sx allprison_lock; 121SX_SYSINIT(allprison_lock, &allprison_lock, "allprison"); 122struct prisonlist allprison = TAILQ_HEAD_INITIALIZER(allprison); 123LIST_HEAD(, prison_racct) allprison_racct; 124int lastprid = 0; 125 126static int do_jail_attach(struct thread *td, struct prison *pr); 127static void prison_complete(void *context, int pending); 128static void prison_deref(struct prison *pr, int flags); 129static char *prison_path(struct prison *pr1, struct prison *pr2); 130static void prison_remove_one(struct prison *pr); 131#ifdef RACCT 132static void prison_racct_attach(struct prison *pr); 133static void prison_racct_modify(struct prison *pr); 134static void prison_racct_detach(struct prison *pr); 135#endif 136#ifdef INET 137static int _prison_check_ip4(struct prison *pr, struct in_addr *ia); 138static int prison_restrict_ip4(struct prison *pr, struct in_addr *newip4); 139#endif 140#ifdef INET6 141static int _prison_check_ip6(struct prison *pr, struct in6_addr *ia6); 142static int prison_restrict_ip6(struct prison *pr, struct in6_addr *newip6); 143#endif 144 145/* Flags for prison_deref */ 146#define PD_DEREF 0x01 147#define PD_DEUREF 0x02 148#define PD_LOCKED 0x04 149#define PD_LIST_SLOCKED 0x08 150#define PD_LIST_XLOCKED 0x10 151 152/* 153 * Parameter names corresponding to PR_* flag values. Size values are for kvm 154 * as we cannot figure out the size of a sparse array, or an array without a 155 * terminating entry. 156 */ 157static char *pr_flag_names[] = { 158 [0] = "persist", 159#ifdef INET 160 [7] = "ip4.saddrsel", 161#endif 162#ifdef INET6 163 [8] = "ip6.saddrsel", 164#endif 165}; 166const size_t pr_flag_names_size = sizeof(pr_flag_names); 167 168static char *pr_flag_nonames[] = { 169 [0] = "nopersist", 170#ifdef INET 171 [7] = "ip4.nosaddrsel", 172#endif 173#ifdef INET6 174 [8] = "ip6.nosaddrsel", 175#endif 176}; 177const size_t pr_flag_nonames_size = sizeof(pr_flag_nonames); 178 179struct jailsys_flags { 180 const char *name; 181 unsigned disable; 182 unsigned new; 183} pr_flag_jailsys[] = { 184 { "host", 0, PR_HOST }, 185#ifdef VIMAGE 186 { "vnet", 0, PR_VNET }, 187#endif 188#ifdef INET 189 { "ip4", PR_IP4_USER | PR_IP4_DISABLE, PR_IP4_USER }, 190#endif 191#ifdef INET6 192 { "ip6", PR_IP6_USER | PR_IP6_DISABLE, PR_IP6_USER }, 193#endif 194}; 195const size_t pr_flag_jailsys_size = sizeof(pr_flag_jailsys); 196 197static char *pr_allow_names[] = { 198 "allow.set_hostname", 199 "allow.sysvipc", 200 "allow.raw_sockets", 201 "allow.chflags", 202 "allow.mount", 203 "allow.quotas", 204 "allow.socket_af", 205 "allow.mount.devfs", 206 "allow.mount.nullfs", 207 "allow.mount.zfs", 208 "allow.mount.procfs", 209 "allow.mount.tmpfs", 210}; 211const size_t pr_allow_names_size = sizeof(pr_allow_names); 212 213static char *pr_allow_nonames[] = { 214 "allow.noset_hostname", 215 "allow.nosysvipc", 216 "allow.noraw_sockets", 217 "allow.nochflags", 218 "allow.nomount", 219 "allow.noquotas", 220 "allow.nosocket_af", 221 "allow.mount.nodevfs", 222 "allow.mount.nonullfs", 223 "allow.mount.nozfs", 224 "allow.mount.noprocfs", 225 "allow.mount.notmpfs", 226}; 227const size_t pr_allow_nonames_size = sizeof(pr_allow_nonames); 228 229#define JAIL_DEFAULT_ALLOW PR_ALLOW_SET_HOSTNAME 230#define JAIL_DEFAULT_ENFORCE_STATFS 2 231#define JAIL_DEFAULT_DEVFS_RSNUM 0 232static unsigned jail_default_allow = JAIL_DEFAULT_ALLOW; 233static int jail_default_enforce_statfs = JAIL_DEFAULT_ENFORCE_STATFS; 234static int jail_default_devfs_rsnum = JAIL_DEFAULT_DEVFS_RSNUM; 235#if defined(INET) || defined(INET6) 236static unsigned jail_max_af_ips = 255; 237#endif 238 239#ifdef INET 240static int 241qcmp_v4(const void *ip1, const void *ip2) 242{ 243 in_addr_t iaa, iab; 244 245 /* 246 * We need to compare in HBO here to get the list sorted as expected 247 * by the result of the code. Sorting NBO addresses gives you 248 * interesting results. If you do not understand, do not try. 249 */ 250 iaa = ntohl(((const struct in_addr *)ip1)->s_addr); 251 iab = ntohl(((const struct in_addr *)ip2)->s_addr); 252 253 /* 254 * Do not simply return the difference of the two numbers, the int is 255 * not wide enough. 256 */ 257 if (iaa > iab) 258 return (1); 259 else if (iaa < iab) 260 return (-1); 261 else 262 return (0); 263} 264#endif 265 266#ifdef INET6 267static int 268qcmp_v6(const void *ip1, const void *ip2) 269{ 270 const struct in6_addr *ia6a, *ia6b; 271 int i, rc; 272 273 ia6a = (const struct in6_addr *)ip1; 274 ia6b = (const struct in6_addr *)ip2; 275 276 rc = 0; 277 for (i = 0; rc == 0 && i < sizeof(struct in6_addr); i++) { 278 if (ia6a->s6_addr[i] > ia6b->s6_addr[i]) 279 rc = 1; 280 else if (ia6a->s6_addr[i] < ia6b->s6_addr[i]) 281 rc = -1; 282 } 283 return (rc); 284} 285#endif 286 287/* 288 * struct jail_args { 289 * struct jail *jail; 290 * }; 291 */ 292int 293sys_jail(struct thread *td, struct jail_args *uap) 294{ 295 uint32_t version; 296 int error; 297 struct jail j; 298 299 error = copyin(uap->jail, &version, sizeof(uint32_t)); 300 if (error) 301 return (error); 302 303 switch (version) { 304 case 0: 305 { 306 struct jail_v0 j0; 307 308 /* FreeBSD single IPv4 jails. */ 309 bzero(&j, sizeof(struct jail)); 310 error = copyin(uap->jail, &j0, sizeof(struct jail_v0)); 311 if (error) 312 return (error); 313 j.version = j0.version; 314 j.path = j0.path; 315 j.hostname = j0.hostname; 316 j.ip4s = htonl(j0.ip_number); /* jail_v0 is host order */ 317 break; 318 } 319 320 case 1: 321 /* 322 * Version 1 was used by multi-IPv4 jail implementations 323 * that never made it into the official kernel. 324 */ 325 return (EINVAL); 326 327 case 2: /* JAIL_API_VERSION */ 328 /* FreeBSD multi-IPv4/IPv6,noIP jails. */ 329 error = copyin(uap->jail, &j, sizeof(struct jail)); 330 if (error) 331 return (error); 332 break; 333 334 default: 335 /* Sci-Fi jails are not supported, sorry. */ 336 return (EINVAL); 337 } 338 return (kern_jail(td, &j)); 339} 340 341int 342kern_jail(struct thread *td, struct jail *j) 343{ 344 struct iovec optiov[2 * (4 345 + sizeof(pr_allow_names) / sizeof(pr_allow_names[0]) 346#ifdef INET 347 + 1 348#endif 349#ifdef INET6 350 + 1 351#endif 352 )]; 353 struct uio opt; 354 char *u_path, *u_hostname, *u_name; 355#ifdef INET 356 uint32_t ip4s; 357 struct in_addr *u_ip4; 358#endif 359#ifdef INET6 360 struct in6_addr *u_ip6; 361#endif 362 size_t tmplen; 363 int error, enforce_statfs, fi; 364 365 bzero(&optiov, sizeof(optiov)); 366 opt.uio_iov = optiov; 367 opt.uio_iovcnt = 0; 368 opt.uio_offset = -1; 369 opt.uio_resid = -1; 370 opt.uio_segflg = UIO_SYSSPACE; 371 opt.uio_rw = UIO_READ; 372 opt.uio_td = td; 373 374 /* Set permissions for top-level jails from sysctls. */ 375 if (!jailed(td->td_ucred)) { 376 for (fi = 0; fi < sizeof(pr_allow_names) / 377 sizeof(pr_allow_names[0]); fi++) { 378 optiov[opt.uio_iovcnt].iov_base = 379 (jail_default_allow & (1 << fi)) 380 ? pr_allow_names[fi] : pr_allow_nonames[fi]; 381 optiov[opt.uio_iovcnt].iov_len = 382 strlen(optiov[opt.uio_iovcnt].iov_base) + 1; 383 opt.uio_iovcnt += 2; 384 } 385 optiov[opt.uio_iovcnt].iov_base = "enforce_statfs"; 386 optiov[opt.uio_iovcnt].iov_len = sizeof("enforce_statfs"); 387 opt.uio_iovcnt++; 388 enforce_statfs = jail_default_enforce_statfs; 389 optiov[opt.uio_iovcnt].iov_base = &enforce_statfs; 390 optiov[opt.uio_iovcnt].iov_len = sizeof(enforce_statfs); 391 opt.uio_iovcnt++; 392 } 393 394 tmplen = MAXPATHLEN + MAXHOSTNAMELEN + MAXHOSTNAMELEN; 395#ifdef INET 396 ip4s = (j->version == 0) ? 1 : j->ip4s; 397 if (ip4s > jail_max_af_ips) 398 return (EINVAL); 399 tmplen += ip4s * sizeof(struct in_addr); 400#else 401 if (j->ip4s > 0) 402 return (EINVAL); 403#endif 404#ifdef INET6 405 if (j->ip6s > jail_max_af_ips) 406 return (EINVAL); 407 tmplen += j->ip6s * sizeof(struct in6_addr); 408#else 409 if (j->ip6s > 0) 410 return (EINVAL); 411#endif 412 u_path = malloc(tmplen, M_TEMP, M_WAITOK); 413 u_hostname = u_path + MAXPATHLEN; 414 u_name = u_hostname + MAXHOSTNAMELEN; 415#ifdef INET 416 u_ip4 = (struct in_addr *)(u_name + MAXHOSTNAMELEN); 417#endif 418#ifdef INET6 419#ifdef INET 420 u_ip6 = (struct in6_addr *)(u_ip4 + ip4s); 421#else 422 u_ip6 = (struct in6_addr *)(u_name + MAXHOSTNAMELEN); 423#endif 424#endif 425 optiov[opt.uio_iovcnt].iov_base = "path"; 426 optiov[opt.uio_iovcnt].iov_len = sizeof("path"); 427 opt.uio_iovcnt++; 428 optiov[opt.uio_iovcnt].iov_base = u_path; 429 error = copyinstr(j->path, u_path, MAXPATHLEN, 430 &optiov[opt.uio_iovcnt].iov_len); 431 if (error) { 432 free(u_path, M_TEMP); 433 return (error); 434 } 435 opt.uio_iovcnt++; 436 optiov[opt.uio_iovcnt].iov_base = "host.hostname"; 437 optiov[opt.uio_iovcnt].iov_len = sizeof("host.hostname"); 438 opt.uio_iovcnt++; 439 optiov[opt.uio_iovcnt].iov_base = u_hostname; 440 error = copyinstr(j->hostname, u_hostname, MAXHOSTNAMELEN, 441 &optiov[opt.uio_iovcnt].iov_len); 442 if (error) { 443 free(u_path, M_TEMP); 444 return (error); 445 } 446 opt.uio_iovcnt++; 447 if (j->jailname != NULL) { 448 optiov[opt.uio_iovcnt].iov_base = "name"; 449 optiov[opt.uio_iovcnt].iov_len = sizeof("name"); 450 opt.uio_iovcnt++; 451 optiov[opt.uio_iovcnt].iov_base = u_name; 452 error = copyinstr(j->jailname, u_name, MAXHOSTNAMELEN, 453 &optiov[opt.uio_iovcnt].iov_len); 454 if (error) { 455 free(u_path, M_TEMP); 456 return (error); 457 } 458 opt.uio_iovcnt++; 459 } 460#ifdef INET 461 optiov[opt.uio_iovcnt].iov_base = "ip4.addr"; 462 optiov[opt.uio_iovcnt].iov_len = sizeof("ip4.addr"); 463 opt.uio_iovcnt++; 464 optiov[opt.uio_iovcnt].iov_base = u_ip4; 465 optiov[opt.uio_iovcnt].iov_len = ip4s * sizeof(struct in_addr); 466 if (j->version == 0) 467 u_ip4->s_addr = j->ip4s; 468 else { 469 error = copyin(j->ip4, u_ip4, optiov[opt.uio_iovcnt].iov_len); 470 if (error) { 471 free(u_path, M_TEMP); 472 return (error); 473 } 474 } 475 opt.uio_iovcnt++; 476#endif 477#ifdef INET6 478 optiov[opt.uio_iovcnt].iov_base = "ip6.addr"; 479 optiov[opt.uio_iovcnt].iov_len = sizeof("ip6.addr"); 480 opt.uio_iovcnt++; 481 optiov[opt.uio_iovcnt].iov_base = u_ip6; 482 optiov[opt.uio_iovcnt].iov_len = j->ip6s * sizeof(struct in6_addr); 483 error = copyin(j->ip6, u_ip6, optiov[opt.uio_iovcnt].iov_len); 484 if (error) { 485 free(u_path, M_TEMP); 486 return (error); 487 } 488 opt.uio_iovcnt++; 489#endif 490 KASSERT(opt.uio_iovcnt <= sizeof(optiov) / sizeof(optiov[0]), 491 ("kern_jail: too many iovecs (%d)", opt.uio_iovcnt)); 492 error = kern_jail_set(td, &opt, JAIL_CREATE | JAIL_ATTACH); 493 free(u_path, M_TEMP); 494 return (error); 495} 496 497 498/* 499 * struct jail_set_args { 500 * struct iovec *iovp; 501 * unsigned int iovcnt; 502 * int flags; 503 * }; 504 */ 505int 506sys_jail_set(struct thread *td, struct jail_set_args *uap) 507{ 508 struct uio *auio; 509 int error; 510 511 /* Check that we have an even number of iovecs. */ 512 if (uap->iovcnt & 1) 513 return (EINVAL); 514 515 error = copyinuio(uap->iovp, uap->iovcnt, &auio); 516 if (error) 517 return (error); 518 error = kern_jail_set(td, auio, uap->flags); 519 free(auio, M_IOV); 520 return (error); 521} 522 523int 524kern_jail_set(struct thread *td, struct uio *optuio, int flags) 525{ 526 struct nameidata nd; 527#ifdef INET 528 struct in_addr *ip4; 529#endif 530#ifdef INET6 531 struct in6_addr *ip6; 532#endif 533 struct vfsopt *opt; 534 struct vfsoptlist *opts; 535 struct prison *pr, *deadpr, *mypr, *ppr, *tpr; 536 struct vnode *root; 537 char *domain, *errmsg, *host, *name, *namelc, *p, *path, *uuid; 538 char *g_path; 539#if defined(INET) || defined(INET6) 540 struct prison *tppr; 541 void *op; 542#endif 543 unsigned long hid; 544 size_t namelen, onamelen; 545 int created, cuflags, descend, enforce, error, errmsg_len, errmsg_pos; 546 int gotchildmax, gotenforce, gothid, gotrsnum, gotslevel; 547 int fi, jid, jsys, len, level; 548 int childmax, rsnum, slevel; 549 int fullpath_disabled; 550#if defined(INET) || defined(INET6) 551 int ii, ij; 552#endif 553#ifdef INET 554 int ip4s, redo_ip4; 555#endif 556#ifdef INET6 557 int ip6s, redo_ip6; 558#endif 559 uint64_t pr_allow, ch_allow, pr_flags, ch_flags; 560 unsigned tallow; 561 char numbuf[12]; 562 563 error = priv_check(td, PRIV_JAIL_SET); 564 if (!error && (flags & JAIL_ATTACH)) 565 error = priv_check(td, PRIV_JAIL_ATTACH); 566 if (error) 567 return (error); 568 mypr = ppr = td->td_ucred->cr_prison; 569 if ((flags & JAIL_CREATE) && mypr->pr_childmax == 0) 570 return (EPERM); 571 if (flags & ~JAIL_SET_MASK) 572 return (EINVAL); 573 574 /* 575 * Check all the parameters before committing to anything. Not all 576 * errors can be caught early, but we may as well try. Also, this 577 * takes care of some expensive stuff (path lookup) before getting 578 * the allprison lock. 579 * 580 * XXX Jails are not filesystems, and jail parameters are not mount 581 * options. But it makes more sense to re-use the vfsopt code 582 * than duplicate it under a different name. 583 */ 584 error = vfs_buildopts(optuio, &opts); 585 if (error) 586 return (error); 587#ifdef INET 588 ip4 = NULL; 589#endif 590#ifdef INET6 591 ip6 = NULL; 592#endif 593 g_path = NULL; 594 595 error = vfs_copyopt(opts, "jid", &jid, sizeof(jid)); 596 if (error == ENOENT) 597 jid = 0; 598 else if (error != 0) 599 goto done_free; 600 601 error = vfs_copyopt(opts, "securelevel", &slevel, sizeof(slevel)); 602 if (error == ENOENT) 603 gotslevel = 0; 604 else if (error != 0) 605 goto done_free; 606 else 607 gotslevel = 1; 608 609 error = 610 vfs_copyopt(opts, "children.max", &childmax, sizeof(childmax)); 611 if (error == ENOENT) 612 gotchildmax = 0; 613 else if (error != 0) 614 goto done_free; 615 else 616 gotchildmax = 1; 617 618 error = vfs_copyopt(opts, "enforce_statfs", &enforce, sizeof(enforce)); 619 if (error == ENOENT) 620 gotenforce = 0; 621 else if (error != 0) 622 goto done_free; 623 else if (enforce < 0 || enforce > 2) { 624 error = EINVAL; 625 goto done_free; 626 } else 627 gotenforce = 1; 628 629 error = vfs_copyopt(opts, "devfs_ruleset", &rsnum, sizeof(rsnum)); 630 if (error == ENOENT) 631 gotrsnum = 0; 632 else if (error != 0) 633 goto done_free; 634 else 635 gotrsnum = 1; 636 637 pr_flags = ch_flags = 0; 638 for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]); 639 fi++) { 640 if (pr_flag_names[fi] == NULL) 641 continue; 642 vfs_flagopt(opts, pr_flag_names[fi], &pr_flags, 1 << fi); 643 vfs_flagopt(opts, pr_flag_nonames[fi], &ch_flags, 1 << fi); 644 } 645 ch_flags |= pr_flags; 646 for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]); 647 fi++) { 648 error = vfs_copyopt(opts, pr_flag_jailsys[fi].name, &jsys, 649 sizeof(jsys)); 650 if (error == ENOENT) 651 continue; 652 if (error != 0) 653 goto done_free; 654 switch (jsys) { 655 case JAIL_SYS_DISABLE: 656 if (!pr_flag_jailsys[fi].disable) { 657 error = EINVAL; 658 goto done_free; 659 } 660 pr_flags |= pr_flag_jailsys[fi].disable; 661 break; 662 case JAIL_SYS_NEW: 663 pr_flags |= pr_flag_jailsys[fi].new; 664 break; 665 case JAIL_SYS_INHERIT: 666 break; 667 default: 668 error = EINVAL; 669 goto done_free; 670 } 671 ch_flags |= 672 pr_flag_jailsys[fi].new | pr_flag_jailsys[fi].disable; 673 } 674 if ((flags & (JAIL_CREATE | JAIL_UPDATE | JAIL_ATTACH)) == JAIL_CREATE 675 && !(pr_flags & PR_PERSIST)) { 676 error = EINVAL; 677 vfs_opterror(opts, "new jail must persist or attach"); 678 goto done_errmsg; 679 } 680#ifdef VIMAGE 681 if ((flags & JAIL_UPDATE) && (ch_flags & PR_VNET)) { 682 error = EINVAL; 683 vfs_opterror(opts, "vnet cannot be changed after creation"); 684 goto done_errmsg; 685 } 686#endif 687#ifdef INET 688 if ((flags & JAIL_UPDATE) && (ch_flags & PR_IP4_USER)) { 689 error = EINVAL; 690 vfs_opterror(opts, "ip4 cannot be changed after creation"); 691 goto done_errmsg; 692 } 693#endif 694#ifdef INET6 695 if ((flags & JAIL_UPDATE) && (ch_flags & PR_IP6_USER)) { 696 error = EINVAL; 697 vfs_opterror(opts, "ip6 cannot be changed after creation"); 698 goto done_errmsg; 699 } 700#endif 701 702 pr_allow = ch_allow = 0; 703 for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]); 704 fi++) { 705 vfs_flagopt(opts, pr_allow_names[fi], &pr_allow, 1 << fi); 706 vfs_flagopt(opts, pr_allow_nonames[fi], &ch_allow, 1 << fi); 707 } 708 ch_allow |= pr_allow; 709 710 error = vfs_getopt(opts, "name", (void **)&name, &len); 711 if (error == ENOENT) 712 name = NULL; 713 else if (error != 0) 714 goto done_free; 715 else { 716 if (len == 0 || name[len - 1] != '\0') { 717 error = EINVAL; 718 goto done_free; 719 } 720 if (len > MAXHOSTNAMELEN) { 721 error = ENAMETOOLONG; 722 goto done_free; 723 } 724 } 725 726 error = vfs_getopt(opts, "host.hostname", (void **)&host, &len); 727 if (error == ENOENT) 728 host = NULL; 729 else if (error != 0) 730 goto done_free; 731 else { 732 ch_flags |= PR_HOST; 733 pr_flags |= PR_HOST; 734 if (len == 0 || host[len - 1] != '\0') { 735 error = EINVAL; 736 goto done_free; 737 } 738 if (len > MAXHOSTNAMELEN) { 739 error = ENAMETOOLONG; 740 goto done_free; 741 } 742 } 743 744 error = vfs_getopt(opts, "host.domainname", (void **)&domain, &len); 745 if (error == ENOENT) 746 domain = NULL; 747 else if (error != 0) 748 goto done_free; 749 else { 750 ch_flags |= PR_HOST; 751 pr_flags |= PR_HOST; 752 if (len == 0 || domain[len - 1] != '\0') { 753 error = EINVAL; 754 goto done_free; 755 } 756 if (len > MAXHOSTNAMELEN) { 757 error = ENAMETOOLONG; 758 goto done_free; 759 } 760 } 761 762 error = vfs_getopt(opts, "host.hostuuid", (void **)&uuid, &len); 763 if (error == ENOENT) 764 uuid = NULL; 765 else if (error != 0) 766 goto done_free; 767 else { 768 ch_flags |= PR_HOST; 769 pr_flags |= PR_HOST; 770 if (len == 0 || uuid[len - 1] != '\0') { 771 error = EINVAL; 772 goto done_free; 773 } 774 if (len > HOSTUUIDLEN) { 775 error = ENAMETOOLONG; 776 goto done_free; 777 } 778 } 779 780#ifdef COMPAT_FREEBSD32 781 if (SV_PROC_FLAG(td->td_proc, SV_ILP32)) { 782 uint32_t hid32; 783 784 error = vfs_copyopt(opts, "host.hostid", &hid32, sizeof(hid32)); 785 hid = hid32; 786 } else 787#endif 788 error = vfs_copyopt(opts, "host.hostid", &hid, sizeof(hid)); 789 if (error == ENOENT) 790 gothid = 0; 791 else if (error != 0) 792 goto done_free; 793 else { 794 gothid = 1; 795 ch_flags |= PR_HOST; 796 pr_flags |= PR_HOST; 797 } 798 799#ifdef INET 800 error = vfs_getopt(opts, "ip4.addr", &op, &ip4s); 801 if (error == ENOENT) 802 ip4s = (pr_flags & PR_IP4_DISABLE) ? 0 : -1; 803 else if (error != 0) 804 goto done_free; 805 else if (ip4s & (sizeof(*ip4) - 1)) { 806 error = EINVAL; 807 goto done_free; 808 } else { 809 ch_flags |= PR_IP4_USER | PR_IP4_DISABLE; 810 if (ip4s == 0) 811 pr_flags |= PR_IP4_USER | PR_IP4_DISABLE; 812 else { 813 pr_flags = (pr_flags & ~PR_IP4_DISABLE) | PR_IP4_USER; 814 ip4s /= sizeof(*ip4); 815 if (ip4s > jail_max_af_ips) { 816 error = EINVAL; 817 vfs_opterror(opts, "too many IPv4 addresses"); 818 goto done_errmsg; 819 } 820 ip4 = malloc(ip4s * sizeof(*ip4), M_PRISON, M_WAITOK); 821 bcopy(op, ip4, ip4s * sizeof(*ip4)); 822 /* 823 * IP addresses are all sorted but ip[0] to preserve 824 * the primary IP address as given from userland. 825 * This special IP is used for unbound outgoing 826 * connections as well for "loopback" traffic in case 827 * source address selection cannot find any more fitting 828 * address to connect from. 829 */ 830 if (ip4s > 1) 831 qsort(ip4 + 1, ip4s - 1, sizeof(*ip4), qcmp_v4); 832 /* 833 * Check for duplicate addresses and do some simple 834 * zero and broadcast checks. If users give other bogus 835 * addresses it is their problem. 836 * 837 * We do not have to care about byte order for these 838 * checks so we will do them in NBO. 839 */ 840 for (ii = 0; ii < ip4s; ii++) { 841 if (ip4[ii].s_addr == INADDR_ANY || 842 ip4[ii].s_addr == INADDR_BROADCAST) { 843 error = EINVAL; 844 goto done_free; 845 } 846 if ((ii+1) < ip4s && 847 (ip4[0].s_addr == ip4[ii+1].s_addr || 848 ip4[ii].s_addr == ip4[ii+1].s_addr)) { 849 error = EINVAL; 850 goto done_free; 851 } 852 } 853 } 854 } 855#endif 856 857#ifdef INET6 858 error = vfs_getopt(opts, "ip6.addr", &op, &ip6s); 859 if (error == ENOENT) 860 ip6s = (pr_flags & PR_IP6_DISABLE) ? 0 : -1; 861 else if (error != 0) 862 goto done_free; 863 else if (ip6s & (sizeof(*ip6) - 1)) { 864 error = EINVAL; 865 goto done_free; 866 } else { 867 ch_flags |= PR_IP6_USER | PR_IP6_DISABLE; 868 if (ip6s == 0) 869 pr_flags |= PR_IP6_USER | PR_IP6_DISABLE; 870 else { 871 pr_flags = (pr_flags & ~PR_IP6_DISABLE) | PR_IP6_USER; 872 ip6s /= sizeof(*ip6); 873 if (ip6s > jail_max_af_ips) { 874 error = EINVAL; 875 vfs_opterror(opts, "too many IPv6 addresses"); 876 goto done_errmsg; 877 } 878 ip6 = malloc(ip6s * sizeof(*ip6), M_PRISON, M_WAITOK); 879 bcopy(op, ip6, ip6s * sizeof(*ip6)); 880 if (ip6s > 1) 881 qsort(ip6 + 1, ip6s - 1, sizeof(*ip6), qcmp_v6); 882 for (ii = 0; ii < ip6s; ii++) { 883 if (IN6_IS_ADDR_UNSPECIFIED(&ip6[ii])) { 884 error = EINVAL; 885 goto done_free; 886 } 887 if ((ii+1) < ip6s && 888 (IN6_ARE_ADDR_EQUAL(&ip6[0], &ip6[ii+1]) || 889 IN6_ARE_ADDR_EQUAL(&ip6[ii], &ip6[ii+1]))) 890 { 891 error = EINVAL; 892 goto done_free; 893 } 894 } 895 } 896 } 897#endif 898 899#if defined(VIMAGE) && (defined(INET) || defined(INET6)) 900 if ((ch_flags & PR_VNET) && (ch_flags & (PR_IP4_USER | PR_IP6_USER))) { 901 error = EINVAL; 902 vfs_opterror(opts, 903 "vnet jails cannot have IP address restrictions"); 904 goto done_errmsg; 905 } 906#endif 907 908 fullpath_disabled = 0; 909 root = NULL; 910 error = vfs_getopt(opts, "path", (void **)&path, &len); 911 if (error == ENOENT) 912 path = NULL; 913 else if (error != 0) 914 goto done_free; 915 else { 916 if (flags & JAIL_UPDATE) { 917 error = EINVAL; 918 vfs_opterror(opts, 919 "path cannot be changed after creation"); 920 goto done_errmsg; 921 } 922 if (len == 0 || path[len - 1] != '\0') { 923 error = EINVAL; 924 goto done_free; 925 } 926 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, 927 path, td); 928 error = namei(&nd); 929 if (error) 930 goto done_free; 931 root = nd.ni_vp; 932 NDFREE(&nd, NDF_ONLY_PNBUF); 933 g_path = malloc(MAXPATHLEN, M_TEMP, M_WAITOK); 934 strlcpy(g_path, path, MAXPATHLEN); 935 error = vn_path_to_global_path(td, root, g_path, MAXPATHLEN); 936 if (error == 0) 937 path = g_path; 938 else if (error == ENODEV) { 939 /* proceed if sysctl debug.disablefullpath == 1 */ 940 fullpath_disabled = 1; 941 if (len < 2 || (len == 2 && path[0] == '/')) 942 path = NULL; 943 } else { 944 /* exit on other errors */ 945 goto done_free; 946 } 947 if (root->v_type != VDIR) { 948 error = ENOTDIR; 949 vput(root); 950 goto done_free; 951 } 952 VOP_UNLOCK(root, 0); 953 if (fullpath_disabled) { 954 /* Leave room for a real-root full pathname. */ 955 if (len + (path[0] == '/' && strcmp(mypr->pr_path, "/") 956 ? strlen(mypr->pr_path) : 0) > MAXPATHLEN) { 957 error = ENAMETOOLONG; 958 goto done_free; 959 } 960 } 961 } 962 963 /* 964 * Grab the allprison lock before letting modules check their 965 * parameters. Once we have it, do not let go so we'll have a 966 * consistent view of the OSD list. 967 */ 968 sx_xlock(&allprison_lock); 969 error = osd_jail_call(NULL, PR_METHOD_CHECK, opts); 970 if (error) 971 goto done_unlock_list; 972 973 /* By now, all parameters should have been noted. */ 974 TAILQ_FOREACH(opt, opts, link) { 975 if (!opt->seen && strcmp(opt->name, "errmsg")) { 976 error = EINVAL; 977 vfs_opterror(opts, "unknown parameter: %s", opt->name); 978 goto done_unlock_list; 979 } 980 } 981 982 /* 983 * See if we are creating a new record or updating an existing one. 984 * This abuses the file error codes ENOENT and EEXIST. 985 */ 986 cuflags = flags & (JAIL_CREATE | JAIL_UPDATE); 987 if (!cuflags) { 988 error = EINVAL; 989 vfs_opterror(opts, "no valid operation (create or update)"); 990 goto done_unlock_list; 991 } 992 pr = NULL; 993 namelc = NULL; 994 if (cuflags == JAIL_CREATE && jid == 0 && name != NULL) { 995 namelc = strrchr(name, '.'); 996 jid = strtoul(namelc != NULL ? namelc + 1 : name, &p, 10); 997 if (*p != '\0') 998 jid = 0; 999 } 1000 if (jid != 0) { 1001 /* 1002 * See if a requested jid already exists. There is an 1003 * information leak here if the jid exists but is not within 1004 * the caller's jail hierarchy. Jail creators will get EEXIST 1005 * even though they cannot see the jail, and CREATE | UPDATE 1006 * will return ENOENT which is not normally a valid error. 1007 */ 1008 if (jid < 0) { 1009 error = EINVAL; 1010 vfs_opterror(opts, "negative jid"); 1011 goto done_unlock_list; 1012 } 1013 pr = prison_find(jid); 1014 if (pr != NULL) { 1015 ppr = pr->pr_parent; 1016 /* Create: jid must not exist. */ 1017 if (cuflags == JAIL_CREATE) { 1018 mtx_unlock(&pr->pr_mtx); 1019 error = EEXIST; 1020 vfs_opterror(opts, "jail %d already exists", 1021 jid); 1022 goto done_unlock_list; 1023 } 1024 if (!prison_ischild(mypr, pr)) { 1025 mtx_unlock(&pr->pr_mtx); 1026 pr = NULL; 1027 } else if (pr->pr_uref == 0) { 1028 if (!(flags & JAIL_DYING)) { 1029 mtx_unlock(&pr->pr_mtx); 1030 error = ENOENT; 1031 vfs_opterror(opts, "jail %d is dying", 1032 jid); 1033 goto done_unlock_list; 1034 } else if ((flags & JAIL_ATTACH) || 1035 (pr_flags & PR_PERSIST)) { 1036 /* 1037 * A dying jail might be resurrected 1038 * (via attach or persist), but first 1039 * it must determine if another jail 1040 * has claimed its name. Accomplish 1041 * this by implicitly re-setting the 1042 * name. 1043 */ 1044 if (name == NULL) 1045 name = prison_name(mypr, pr); 1046 } 1047 } 1048 } 1049 if (pr == NULL) { 1050 /* Update: jid must exist. */ 1051 if (cuflags == JAIL_UPDATE) { 1052 error = ENOENT; 1053 vfs_opterror(opts, "jail %d not found", jid); 1054 goto done_unlock_list; 1055 } 1056 } 1057 } 1058 /* 1059 * If the caller provided a name, look for a jail by that name. 1060 * This has different semantics for creates and updates keyed by jid 1061 * (where the name must not already exist in a different jail), 1062 * and updates keyed by the name itself (where the name must exist 1063 * because that is the jail being updated). 1064 */ 1065 if (name != NULL) { 1066 namelc = strrchr(name, '.'); 1067 if (namelc == NULL) 1068 namelc = name; 1069 else { 1070 /* 1071 * This is a hierarchical name. Split it into the 1072 * parent and child names, and make sure the parent 1073 * exists or matches an already found jail. 1074 */ 1075 *namelc = '\0'; 1076 if (pr != NULL) { 1077 if (strncmp(name, ppr->pr_name, namelc - name) 1078 || ppr->pr_name[namelc - name] != '\0') { 1079 mtx_unlock(&pr->pr_mtx); 1080 error = EINVAL; 1081 vfs_opterror(opts, 1082 "cannot change jail's parent"); 1083 goto done_unlock_list; 1084 } 1085 } else { 1086 ppr = prison_find_name(mypr, name); 1087 if (ppr == NULL) { 1088 error = ENOENT; 1089 vfs_opterror(opts, 1090 "jail \"%s\" not found", name); 1091 goto done_unlock_list; 1092 } 1093 mtx_unlock(&ppr->pr_mtx); 1094 } 1095 name = ++namelc; 1096 } 1097 if (name[0] != '\0') { 1098 namelen = 1099 (ppr == &prison0) ? 0 : strlen(ppr->pr_name) + 1; 1100 name_again: 1101 deadpr = NULL; 1102 FOREACH_PRISON_CHILD(ppr, tpr) { 1103 if (tpr != pr && tpr->pr_ref > 0 && 1104 !strcmp(tpr->pr_name + namelen, name)) { 1105 if (pr == NULL && 1106 cuflags != JAIL_CREATE) { 1107 mtx_lock(&tpr->pr_mtx); 1108 if (tpr->pr_ref > 0) { 1109 /* 1110 * Use this jail 1111 * for updates. 1112 */ 1113 if (tpr->pr_uref > 0) { 1114 pr = tpr; 1115 break; 1116 } 1117 deadpr = tpr; 1118 } 1119 mtx_unlock(&tpr->pr_mtx); 1120 } else if (tpr->pr_uref > 0) { 1121 /* 1122 * Create, or update(jid): 1123 * name must not exist in an 1124 * active sibling jail. 1125 */ 1126 error = EEXIST; 1127 if (pr != NULL) 1128 mtx_unlock(&pr->pr_mtx); 1129 vfs_opterror(opts, 1130 "jail \"%s\" already exists", 1131 name); 1132 goto done_unlock_list; 1133 } 1134 } 1135 } 1136 /* If no active jail is found, use a dying one. */ 1137 if (deadpr != NULL && pr == NULL) { 1138 if (flags & JAIL_DYING) { 1139 mtx_lock(&deadpr->pr_mtx); 1140 if (deadpr->pr_ref == 0) { 1141 mtx_unlock(&deadpr->pr_mtx); 1142 goto name_again; 1143 } 1144 pr = deadpr; 1145 } else if (cuflags == JAIL_UPDATE) { 1146 error = ENOENT; 1147 vfs_opterror(opts, 1148 "jail \"%s\" is dying", name); 1149 goto done_unlock_list; 1150 } 1151 } 1152 /* Update: name must exist if no jid. */ 1153 else if (cuflags == JAIL_UPDATE && pr == NULL) { 1154 error = ENOENT; 1155 vfs_opterror(opts, "jail \"%s\" not found", 1156 name); 1157 goto done_unlock_list; 1158 } 1159 } 1160 } 1161 /* Update: must provide a jid or name. */ 1162 else if (cuflags == JAIL_UPDATE && pr == NULL) { 1163 error = ENOENT; 1164 vfs_opterror(opts, "update specified no jail"); 1165 goto done_unlock_list; 1166 } 1167 1168 /* If there's no prison to update, create a new one and link it in. */ 1169 if (pr == NULL) { 1170 for (tpr = mypr; tpr != NULL; tpr = tpr->pr_parent) 1171 if (tpr->pr_childcount >= tpr->pr_childmax) { 1172 error = EPERM; 1173 vfs_opterror(opts, "prison limit exceeded"); 1174 goto done_unlock_list; 1175 } 1176 created = 1; 1177 mtx_lock(&ppr->pr_mtx); 1178 if (ppr->pr_ref == 0 || (ppr->pr_flags & PR_REMOVE)) { 1179 mtx_unlock(&ppr->pr_mtx); 1180 error = ENOENT; 1181 vfs_opterror(opts, "parent jail went away!"); 1182 goto done_unlock_list; 1183 } 1184 ppr->pr_ref++; 1185 ppr->pr_uref++; 1186 mtx_unlock(&ppr->pr_mtx); 1187 pr = malloc(sizeof(*pr), M_PRISON, M_WAITOK | M_ZERO); 1188 if (jid == 0) { 1189 /* Find the next free jid. */ 1190 jid = lastprid + 1; 1191 findnext: 1192 if (jid == JAIL_MAX) 1193 jid = 1; 1194 TAILQ_FOREACH(tpr, &allprison, pr_list) { 1195 if (tpr->pr_id < jid) 1196 continue; 1197 if (tpr->pr_id > jid || tpr->pr_ref == 0) { 1198 TAILQ_INSERT_BEFORE(tpr, pr, pr_list); 1199 break; 1200 } 1201 if (jid == lastprid) { 1202 error = EAGAIN; 1203 vfs_opterror(opts, 1204 "no available jail IDs"); 1205 free(pr, M_PRISON); 1206 prison_deref(ppr, PD_DEREF | 1207 PD_DEUREF | PD_LIST_XLOCKED); 1208 goto done_releroot; 1209 } 1210 jid++; 1211 goto findnext; 1212 } 1213 lastprid = jid; 1214 } else { 1215 /* 1216 * The jail already has a jid (that did not yet exist), 1217 * so just find where to insert it. 1218 */ 1219 TAILQ_FOREACH(tpr, &allprison, pr_list) 1220 if (tpr->pr_id >= jid) { 1221 TAILQ_INSERT_BEFORE(tpr, pr, pr_list); 1222 break; 1223 } 1224 } 1225 if (tpr == NULL) 1226 TAILQ_INSERT_TAIL(&allprison, pr, pr_list); 1227 LIST_INSERT_HEAD(&ppr->pr_children, pr, pr_sibling); 1228 for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent) 1229 tpr->pr_childcount++; 1230 1231 pr->pr_parent = ppr; 1232 pr->pr_id = jid; 1233 1234 /* Set some default values, and inherit some from the parent. */ 1235 if (name == NULL) 1236 name = ""; 1237 if (path == NULL) { 1238 path = "/"; 1239 root = mypr->pr_root; 1240 vref(root); 1241 } 1242 strlcpy(pr->pr_hostuuid, DEFAULT_HOSTUUID, HOSTUUIDLEN); 1243 pr->pr_flags |= PR_HOST; 1244#if defined(INET) || defined(INET6) 1245#ifdef VIMAGE 1246 if (!(pr_flags & PR_VNET)) 1247#endif 1248 { 1249#ifdef INET 1250 if (!(ch_flags & PR_IP4_USER)) 1251 pr->pr_flags |= 1252 PR_IP4 | PR_IP4_USER | PR_IP4_DISABLE; 1253 else if (!(pr_flags & PR_IP4_USER)) { 1254 pr->pr_flags |= ppr->pr_flags & PR_IP4; 1255 if (ppr->pr_ip4 != NULL) { 1256 pr->pr_ip4s = ppr->pr_ip4s; 1257 pr->pr_ip4 = malloc(pr->pr_ip4s * 1258 sizeof(struct in_addr), M_PRISON, 1259 M_WAITOK); 1260 bcopy(ppr->pr_ip4, pr->pr_ip4, 1261 pr->pr_ip4s * sizeof(*pr->pr_ip4)); 1262 } 1263 } 1264#endif 1265#ifdef INET6 1266 if (!(ch_flags & PR_IP6_USER)) 1267 pr->pr_flags |= 1268 PR_IP6 | PR_IP6_USER | PR_IP6_DISABLE; 1269 else if (!(pr_flags & PR_IP6_USER)) { 1270 pr->pr_flags |= ppr->pr_flags & PR_IP6; 1271 if (ppr->pr_ip6 != NULL) { 1272 pr->pr_ip6s = ppr->pr_ip6s; 1273 pr->pr_ip6 = malloc(pr->pr_ip6s * 1274 sizeof(struct in6_addr), M_PRISON, 1275 M_WAITOK); 1276 bcopy(ppr->pr_ip6, pr->pr_ip6, 1277 pr->pr_ip6s * sizeof(*pr->pr_ip6)); 1278 } 1279 } 1280#endif 1281 } 1282#endif 1283 /* Source address selection is always on by default. */ 1284 pr->pr_flags |= _PR_IP_SADDRSEL; 1285 1286 pr->pr_securelevel = ppr->pr_securelevel; 1287 pr->pr_allow = JAIL_DEFAULT_ALLOW & ppr->pr_allow; 1288 pr->pr_enforce_statfs = JAIL_DEFAULT_ENFORCE_STATFS; 1289 pr->pr_devfs_rsnum = ppr->pr_devfs_rsnum; 1290 1291 LIST_INIT(&pr->pr_children); 1292 mtx_init(&pr->pr_mtx, "jail mutex", NULL, MTX_DEF | MTX_DUPOK); 1293 1294#ifdef VIMAGE 1295 /* Allocate a new vnet if specified. */ 1296 pr->pr_vnet = (pr_flags & PR_VNET) 1297 ? vnet_alloc() : ppr->pr_vnet; 1298#endif 1299 /* 1300 * Allocate a dedicated cpuset for each jail. 1301 * Unlike other initial settings, this may return an erorr. 1302 */ 1303 error = cpuset_create_root(ppr, &pr->pr_cpuset); 1304 if (error) { 1305 prison_deref(pr, PD_LIST_XLOCKED); 1306 goto done_releroot; 1307 } 1308 1309 mtx_lock(&pr->pr_mtx); 1310 /* 1311 * New prisons do not yet have a reference, because we do not 1312 * want other to see the incomplete prison once the 1313 * allprison_lock is downgraded. 1314 */ 1315 } else { 1316 created = 0; 1317 /* 1318 * Grab a reference for existing prisons, to ensure they 1319 * continue to exist for the duration of the call. 1320 */ 1321 pr->pr_ref++; 1322#if defined(VIMAGE) && (defined(INET) || defined(INET6)) 1323 if ((pr->pr_flags & PR_VNET) && 1324 (ch_flags & (PR_IP4_USER | PR_IP6_USER))) { 1325 error = EINVAL; 1326 vfs_opterror(opts, 1327 "vnet jails cannot have IP address restrictions"); 1328 goto done_deref_locked; 1329 } 1330#endif 1331#ifdef INET 1332 if (PR_IP4_USER & ch_flags & (pr_flags ^ pr->pr_flags)) { 1333 error = EINVAL; 1334 vfs_opterror(opts, 1335 "ip4 cannot be changed after creation"); 1336 goto done_deref_locked; 1337 } 1338#endif 1339#ifdef INET6 1340 if (PR_IP6_USER & ch_flags & (pr_flags ^ pr->pr_flags)) { 1341 error = EINVAL; 1342 vfs_opterror(opts, 1343 "ip6 cannot be changed after creation"); 1344 goto done_deref_locked; 1345 } 1346#endif 1347 } 1348 1349 /* Do final error checking before setting anything. */ 1350 if (gotslevel) { 1351 if (slevel < ppr->pr_securelevel) { 1352 error = EPERM; 1353 goto done_deref_locked; 1354 } 1355 } 1356 if (gotchildmax) { 1357 if (childmax >= ppr->pr_childmax) { 1358 error = EPERM; 1359 goto done_deref_locked; 1360 } 1361 } 1362 if (gotenforce) { 1363 if (enforce < ppr->pr_enforce_statfs) { 1364 error = EPERM; 1365 goto done_deref_locked; 1366 } 1367 } 1368 if (gotrsnum) { 1369 /* 1370 * devfs_rsnum is a uint16_t 1371 */ 1372 if (rsnum < 0 || rsnum > 65535) { 1373 error = EINVAL; 1374 goto done_deref_locked; 1375 } 1376 /* 1377 * Nested jails always inherit parent's devfs ruleset 1378 */ 1379 if (jailed(td->td_ucred)) { 1380 if (rsnum > 0 && rsnum != ppr->pr_devfs_rsnum) { 1381 error = EPERM; 1382 goto done_deref_locked; 1383 } else 1384 rsnum = ppr->pr_devfs_rsnum; 1385 } 1386 } 1387#ifdef INET 1388 if (ip4s > 0) { 1389 if (ppr->pr_flags & PR_IP4) { 1390 /* 1391 * Make sure the new set of IP addresses is a 1392 * subset of the parent's list. Don't worry 1393 * about the parent being unlocked, as any 1394 * setting is done with allprison_lock held. 1395 */ 1396 for (ij = 0; ij < ppr->pr_ip4s; ij++) 1397 if (ip4[0].s_addr == ppr->pr_ip4[ij].s_addr) 1398 break; 1399 if (ij == ppr->pr_ip4s) { 1400 error = EPERM; 1401 goto done_deref_locked; 1402 } 1403 if (ip4s > 1) { 1404 for (ii = ij = 1; ii < ip4s; ii++) { 1405 if (ip4[ii].s_addr == 1406 ppr->pr_ip4[0].s_addr) 1407 continue; 1408 for (; ij < ppr->pr_ip4s; ij++) 1409 if (ip4[ii].s_addr == 1410 ppr->pr_ip4[ij].s_addr) 1411 break; 1412 if (ij == ppr->pr_ip4s) 1413 break; 1414 } 1415 if (ij == ppr->pr_ip4s) { 1416 error = EPERM; 1417 goto done_deref_locked; 1418 } 1419 } 1420 } 1421 /* 1422 * Check for conflicting IP addresses. We permit them 1423 * if there is no more than one IP on each jail. If 1424 * there is a duplicate on a jail with more than one 1425 * IP stop checking and return error. 1426 */ 1427 tppr = ppr; 1428#ifdef VIMAGE 1429 for (; tppr != &prison0; tppr = tppr->pr_parent) 1430 if (tppr->pr_flags & PR_VNET) 1431 break; 1432#endif 1433 FOREACH_PRISON_DESCENDANT(tppr, tpr, descend) { 1434 if (tpr == pr || 1435#ifdef VIMAGE 1436 (tpr != tppr && (tpr->pr_flags & PR_VNET)) || 1437#endif 1438 tpr->pr_uref == 0) { 1439 descend = 0; 1440 continue; 1441 } 1442 if (!(tpr->pr_flags & PR_IP4_USER)) 1443 continue; 1444 descend = 0; 1445 if (tpr->pr_ip4 == NULL || 1446 (ip4s == 1 && tpr->pr_ip4s == 1)) 1447 continue; 1448 for (ii = 0; ii < ip4s; ii++) { 1449 if (_prison_check_ip4(tpr, &ip4[ii]) == 0) { 1450 error = EADDRINUSE; 1451 vfs_opterror(opts, 1452 "IPv4 addresses clash"); 1453 goto done_deref_locked; 1454 } 1455 } 1456 } 1457 } 1458#endif 1459#ifdef INET6 1460 if (ip6s > 0) { 1461 if (ppr->pr_flags & PR_IP6) { 1462 /* 1463 * Make sure the new set of IP addresses is a 1464 * subset of the parent's list. 1465 */ 1466 for (ij = 0; ij < ppr->pr_ip6s; ij++) 1467 if (IN6_ARE_ADDR_EQUAL(&ip6[0], 1468 &ppr->pr_ip6[ij])) 1469 break; 1470 if (ij == ppr->pr_ip6s) { 1471 error = EPERM; 1472 goto done_deref_locked; 1473 } 1474 if (ip6s > 1) { 1475 for (ii = ij = 1; ii < ip6s; ii++) { 1476 if (IN6_ARE_ADDR_EQUAL(&ip6[ii], 1477 &ppr->pr_ip6[0])) 1478 continue; 1479 for (; ij < ppr->pr_ip6s; ij++) 1480 if (IN6_ARE_ADDR_EQUAL( 1481 &ip6[ii], &ppr->pr_ip6[ij])) 1482 break; 1483 if (ij == ppr->pr_ip6s) 1484 break; 1485 } 1486 if (ij == ppr->pr_ip6s) { 1487 error = EPERM; 1488 goto done_deref_locked; 1489 } 1490 } 1491 } 1492 /* Check for conflicting IP addresses. */ 1493 tppr = ppr; 1494#ifdef VIMAGE 1495 for (; tppr != &prison0; tppr = tppr->pr_parent) 1496 if (tppr->pr_flags & PR_VNET) 1497 break; 1498#endif 1499 FOREACH_PRISON_DESCENDANT(tppr, tpr, descend) { 1500 if (tpr == pr || 1501#ifdef VIMAGE 1502 (tpr != tppr && (tpr->pr_flags & PR_VNET)) || 1503#endif 1504 tpr->pr_uref == 0) { 1505 descend = 0; 1506 continue; 1507 } 1508 if (!(tpr->pr_flags & PR_IP6_USER)) 1509 continue; 1510 descend = 0; 1511 if (tpr->pr_ip6 == NULL || 1512 (ip6s == 1 && tpr->pr_ip6s == 1)) 1513 continue; 1514 for (ii = 0; ii < ip6s; ii++) { 1515 if (_prison_check_ip6(tpr, &ip6[ii]) == 0) { 1516 error = EADDRINUSE; 1517 vfs_opterror(opts, 1518 "IPv6 addresses clash"); 1519 goto done_deref_locked; 1520 } 1521 } 1522 } 1523 } 1524#endif 1525 onamelen = namelen = 0; 1526 if (name != NULL) { 1527 /* Give a default name of the jid. */ 1528 if (name[0] == '\0') 1529 snprintf(name = numbuf, sizeof(numbuf), "%d", jid); 1530 else if (*namelc == '0' || (strtoul(namelc, &p, 10) != jid && 1531 *p == '\0')) { 1532 error = EINVAL; 1533 vfs_opterror(opts, 1534 "name cannot be numeric (unless it is the jid)"); 1535 goto done_deref_locked; 1536 } 1537 /* 1538 * Make sure the name isn't too long for the prison or its 1539 * children. 1540 */ 1541 onamelen = strlen(pr->pr_name); 1542 namelen = strlen(name); 1543 if (strlen(ppr->pr_name) + namelen + 2 > sizeof(pr->pr_name)) { 1544 error = ENAMETOOLONG; 1545 goto done_deref_locked; 1546 } 1547 FOREACH_PRISON_DESCENDANT(pr, tpr, descend) { 1548 if (strlen(tpr->pr_name) + (namelen - onamelen) >= 1549 sizeof(pr->pr_name)) { 1550 error = ENAMETOOLONG; 1551 goto done_deref_locked; 1552 } 1553 } 1554 } 1555 if (pr_allow & ~ppr->pr_allow) { 1556 error = EPERM; 1557 goto done_deref_locked; 1558 } 1559 1560 /* Set the parameters of the prison. */ 1561#ifdef INET 1562 redo_ip4 = 0; 1563 if (pr_flags & PR_IP4_USER) { 1564 pr->pr_flags |= PR_IP4; 1565 free(pr->pr_ip4, M_PRISON); 1566 pr->pr_ip4s = ip4s; 1567 pr->pr_ip4 = ip4; 1568 ip4 = NULL; 1569 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) { 1570#ifdef VIMAGE 1571 if (tpr->pr_flags & PR_VNET) { 1572 descend = 0; 1573 continue; 1574 } 1575#endif 1576 if (prison_restrict_ip4(tpr, NULL)) { 1577 redo_ip4 = 1; 1578 descend = 0; 1579 } 1580 } 1581 } 1582#endif 1583#ifdef INET6 1584 redo_ip6 = 0; 1585 if (pr_flags & PR_IP6_USER) { 1586 pr->pr_flags |= PR_IP6; 1587 free(pr->pr_ip6, M_PRISON); 1588 pr->pr_ip6s = ip6s; 1589 pr->pr_ip6 = ip6; 1590 ip6 = NULL; 1591 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) { 1592#ifdef VIMAGE 1593 if (tpr->pr_flags & PR_VNET) { 1594 descend = 0; 1595 continue; 1596 } 1597#endif 1598 if (prison_restrict_ip6(tpr, NULL)) { 1599 redo_ip6 = 1; 1600 descend = 0; 1601 } 1602 } 1603 } 1604#endif 1605 if (gotslevel) { 1606 pr->pr_securelevel = slevel; 1607 /* Set all child jails to be at least this level. */ 1608 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) 1609 if (tpr->pr_securelevel < slevel) 1610 tpr->pr_securelevel = slevel; 1611 } 1612 if (gotchildmax) { 1613 pr->pr_childmax = childmax; 1614 /* Set all child jails to under this limit. */ 1615 FOREACH_PRISON_DESCENDANT_LOCKED_LEVEL(pr, tpr, descend, level) 1616 if (tpr->pr_childmax > childmax - level) 1617 tpr->pr_childmax = childmax > level 1618 ? childmax - level : 0; 1619 } 1620 if (gotenforce) { 1621 pr->pr_enforce_statfs = enforce; 1622 /* Pass this restriction on to the children. */ 1623 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) 1624 if (tpr->pr_enforce_statfs < enforce) 1625 tpr->pr_enforce_statfs = enforce; 1626 } 1627 if (gotrsnum) { 1628 pr->pr_devfs_rsnum = rsnum; 1629 /* Pass this restriction on to the children. */ 1630 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) 1631 tpr->pr_devfs_rsnum = rsnum; 1632 } 1633 if (name != NULL) { 1634 if (ppr == &prison0) 1635 strlcpy(pr->pr_name, name, sizeof(pr->pr_name)); 1636 else 1637 snprintf(pr->pr_name, sizeof(pr->pr_name), "%s.%s", 1638 ppr->pr_name, name); 1639 /* Change this component of child names. */ 1640 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) { 1641 bcopy(tpr->pr_name + onamelen, tpr->pr_name + namelen, 1642 strlen(tpr->pr_name + onamelen) + 1); 1643 bcopy(pr->pr_name, tpr->pr_name, namelen); 1644 } 1645 } 1646 if (path != NULL) { 1647 /* Try to keep a real-rooted full pathname. */ 1648 if (fullpath_disabled && path[0] == '/' && 1649 strcmp(mypr->pr_path, "/")) 1650 snprintf(pr->pr_path, sizeof(pr->pr_path), "%s%s", 1651 mypr->pr_path, path); 1652 else 1653 strlcpy(pr->pr_path, path, sizeof(pr->pr_path)); 1654 pr->pr_root = root; 1655 } 1656 if (PR_HOST & ch_flags & ~pr_flags) { 1657 if (pr->pr_flags & PR_HOST) { 1658 /* 1659 * Copy the parent's host info. As with pr_ip4 above, 1660 * the lack of a lock on the parent is not a problem; 1661 * it is always set with allprison_lock at least 1662 * shared, and is held exclusively here. 1663 */ 1664 strlcpy(pr->pr_hostname, pr->pr_parent->pr_hostname, 1665 sizeof(pr->pr_hostname)); 1666 strlcpy(pr->pr_domainname, pr->pr_parent->pr_domainname, 1667 sizeof(pr->pr_domainname)); 1668 strlcpy(pr->pr_hostuuid, pr->pr_parent->pr_hostuuid, 1669 sizeof(pr->pr_hostuuid)); 1670 pr->pr_hostid = pr->pr_parent->pr_hostid; 1671 } 1672 } else if (host != NULL || domain != NULL || uuid != NULL || gothid) { 1673 /* Set this prison, and any descendants without PR_HOST. */ 1674 if (host != NULL) 1675 strlcpy(pr->pr_hostname, host, sizeof(pr->pr_hostname)); 1676 if (domain != NULL) 1677 strlcpy(pr->pr_domainname, domain, 1678 sizeof(pr->pr_domainname)); 1679 if (uuid != NULL) 1680 strlcpy(pr->pr_hostuuid, uuid, sizeof(pr->pr_hostuuid)); 1681 if (gothid) 1682 pr->pr_hostid = hid; 1683 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) { 1684 if (tpr->pr_flags & PR_HOST) 1685 descend = 0; 1686 else { 1687 if (host != NULL) 1688 strlcpy(tpr->pr_hostname, 1689 pr->pr_hostname, 1690 sizeof(tpr->pr_hostname)); 1691 if (domain != NULL) 1692 strlcpy(tpr->pr_domainname, 1693 pr->pr_domainname, 1694 sizeof(tpr->pr_domainname)); 1695 if (uuid != NULL) 1696 strlcpy(tpr->pr_hostuuid, 1697 pr->pr_hostuuid, 1698 sizeof(tpr->pr_hostuuid)); 1699 if (gothid) 1700 tpr->pr_hostid = hid; 1701 } 1702 } 1703 } 1704 if ((tallow = ch_allow & ~pr_allow)) { 1705 /* Clear allow bits in all children. */ 1706 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) 1707 tpr->pr_allow &= ~tallow; 1708 } 1709 pr->pr_allow = (pr->pr_allow & ~ch_allow) | pr_allow; 1710 /* 1711 * Persistent prisons get an extra reference, and prisons losing their 1712 * persist flag lose that reference. Only do this for existing prisons 1713 * for now, so new ones will remain unseen until after the module 1714 * handlers have completed. 1715 */ 1716 if (!created && (ch_flags & PR_PERSIST & (pr_flags ^ pr->pr_flags))) { 1717 if (pr_flags & PR_PERSIST) { 1718 pr->pr_ref++; 1719 pr->pr_uref++; 1720 } else { 1721 pr->pr_ref--; 1722 pr->pr_uref--; 1723 } 1724 } 1725 pr->pr_flags = (pr->pr_flags & ~ch_flags) | pr_flags; 1726 mtx_unlock(&pr->pr_mtx); 1727 1728#ifdef RACCT 1729 if (created) 1730 prison_racct_attach(pr); 1731#endif 1732 1733 /* Locks may have prevented a complete restriction of child IP 1734 * addresses. If so, allocate some more memory and try again. 1735 */ 1736#ifdef INET 1737 while (redo_ip4) { 1738 ip4s = pr->pr_ip4s; 1739 ip4 = malloc(ip4s * sizeof(*ip4), M_PRISON, M_WAITOK); 1740 mtx_lock(&pr->pr_mtx); 1741 redo_ip4 = 0; 1742 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) { 1743#ifdef VIMAGE 1744 if (tpr->pr_flags & PR_VNET) { 1745 descend = 0; 1746 continue; 1747 } 1748#endif 1749 if (prison_restrict_ip4(tpr, ip4)) { 1750 if (ip4 != NULL) 1751 ip4 = NULL; 1752 else 1753 redo_ip4 = 1; 1754 } 1755 } 1756 mtx_unlock(&pr->pr_mtx); 1757 } 1758#endif 1759#ifdef INET6 1760 while (redo_ip6) { 1761 ip6s = pr->pr_ip6s; 1762 ip6 = malloc(ip6s * sizeof(*ip6), M_PRISON, M_WAITOK); 1763 mtx_lock(&pr->pr_mtx); 1764 redo_ip6 = 0; 1765 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) { 1766#ifdef VIMAGE 1767 if (tpr->pr_flags & PR_VNET) { 1768 descend = 0; 1769 continue; 1770 } 1771#endif 1772 if (prison_restrict_ip6(tpr, ip6)) { 1773 if (ip6 != NULL) 1774 ip6 = NULL; 1775 else 1776 redo_ip6 = 1; 1777 } 1778 } 1779 mtx_unlock(&pr->pr_mtx); 1780 } 1781#endif 1782 1783 /* Let the modules do their work. */ 1784 sx_downgrade(&allprison_lock); 1785 if (created) { 1786 error = osd_jail_call(pr, PR_METHOD_CREATE, opts); 1787 if (error) { 1788 prison_deref(pr, PD_LIST_SLOCKED); 1789 goto done_errmsg; 1790 } 1791 } 1792 error = osd_jail_call(pr, PR_METHOD_SET, opts); 1793 if (error) { 1794 prison_deref(pr, created 1795 ? PD_LIST_SLOCKED 1796 : PD_DEREF | PD_LIST_SLOCKED); 1797 goto done_errmsg; 1798 } 1799 1800 /* Attach this process to the prison if requested. */ 1801 if (flags & JAIL_ATTACH) { 1802 mtx_lock(&pr->pr_mtx); 1803 error = do_jail_attach(td, pr); 1804 if (error) { 1805 vfs_opterror(opts, "attach failed"); 1806 if (!created) 1807 prison_deref(pr, PD_DEREF); 1808 goto done_errmsg; 1809 } 1810 } 1811 1812#ifdef RACCT 1813 if (!created) { 1814 sx_sunlock(&allprison_lock); 1815 prison_racct_modify(pr); 1816 sx_slock(&allprison_lock); 1817 } 1818#endif 1819 1820 td->td_retval[0] = pr->pr_id; 1821 1822 /* 1823 * Now that it is all there, drop the temporary reference from existing 1824 * prisons. Or add a reference to newly created persistent prisons 1825 * (which was not done earlier so that the prison would not be publicly 1826 * visible). 1827 */ 1828 if (!created) { 1829 prison_deref(pr, (flags & JAIL_ATTACH) 1830 ? PD_DEREF 1831 : PD_DEREF | PD_LIST_SLOCKED); 1832 } else { 1833 if (pr_flags & PR_PERSIST) { 1834 mtx_lock(&pr->pr_mtx); 1835 pr->pr_ref++; 1836 pr->pr_uref++; 1837 mtx_unlock(&pr->pr_mtx); 1838 } 1839 if (!(flags & JAIL_ATTACH)) 1840 sx_sunlock(&allprison_lock); 1841 } 1842 1843 goto done_errmsg; 1844 1845 done_deref_locked: 1846 prison_deref(pr, created 1847 ? PD_LOCKED | PD_LIST_XLOCKED 1848 : PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED); 1849 goto done_releroot; 1850 done_unlock_list: 1851 sx_xunlock(&allprison_lock); 1852 done_releroot: 1853 if (root != NULL) 1854 vrele(root); 1855 done_errmsg: 1856 if (error) { 1857 vfs_getopt(opts, "errmsg", (void **)&errmsg, &errmsg_len); 1858 if (errmsg_len > 0) { 1859 errmsg_pos = 2 * vfs_getopt_pos(opts, "errmsg") + 1; 1860 if (errmsg_pos > 0) { 1861 if (optuio->uio_segflg == UIO_SYSSPACE) 1862 bcopy(errmsg, 1863 optuio->uio_iov[errmsg_pos].iov_base, 1864 errmsg_len); 1865 else 1866 copyout(errmsg, 1867 optuio->uio_iov[errmsg_pos].iov_base, 1868 errmsg_len); 1869 } 1870 } 1871 } 1872 done_free: 1873#ifdef INET 1874 free(ip4, M_PRISON); 1875#endif 1876#ifdef INET6 1877 free(ip6, M_PRISON); 1878#endif 1879 if (g_path != NULL) 1880 free(g_path, M_TEMP); 1881 vfs_freeopts(opts); 1882 return (error); 1883} 1884 1885 1886/* 1887 * struct jail_get_args { 1888 * struct iovec *iovp; 1889 * unsigned int iovcnt; 1890 * int flags; 1891 * }; 1892 */ 1893int 1894sys_jail_get(struct thread *td, struct jail_get_args *uap) 1895{ 1896 struct uio *auio; 1897 int error; 1898 1899 /* Check that we have an even number of iovecs. */ 1900 if (uap->iovcnt & 1) 1901 return (EINVAL); 1902 1903 error = copyinuio(uap->iovp, uap->iovcnt, &auio); 1904 if (error) 1905 return (error); 1906 error = kern_jail_get(td, auio, uap->flags); 1907 if (error == 0) 1908 error = copyout(auio->uio_iov, uap->iovp, 1909 uap->iovcnt * sizeof (struct iovec)); 1910 free(auio, M_IOV); 1911 return (error); 1912} 1913 1914int 1915kern_jail_get(struct thread *td, struct uio *optuio, int flags) 1916{ 1917 struct prison *pr, *mypr; 1918 struct vfsopt *opt; 1919 struct vfsoptlist *opts; 1920 char *errmsg, *name; 1921 int error, errmsg_len, errmsg_pos, fi, i, jid, len, locked, pos; 1922 1923 if (flags & ~JAIL_GET_MASK) 1924 return (EINVAL); 1925 1926 /* Get the parameter list. */ 1927 error = vfs_buildopts(optuio, &opts); 1928 if (error) 1929 return (error); 1930 errmsg_pos = vfs_getopt_pos(opts, "errmsg"); 1931 mypr = td->td_ucred->cr_prison; 1932 1933 /* 1934 * Find the prison specified by one of: lastjid, jid, name. 1935 */ 1936 sx_slock(&allprison_lock); 1937 error = vfs_copyopt(opts, "lastjid", &jid, sizeof(jid)); 1938 if (error == 0) { 1939 TAILQ_FOREACH(pr, &allprison, pr_list) { 1940 if (pr->pr_id > jid && prison_ischild(mypr, pr)) { 1941 mtx_lock(&pr->pr_mtx); 1942 if (pr->pr_ref > 0 && 1943 (pr->pr_uref > 0 || (flags & JAIL_DYING))) 1944 break; 1945 mtx_unlock(&pr->pr_mtx); 1946 } 1947 } 1948 if (pr != NULL) 1949 goto found_prison; 1950 error = ENOENT; 1951 vfs_opterror(opts, "no jail after %d", jid); 1952 goto done_unlock_list; 1953 } else if (error != ENOENT) 1954 goto done_unlock_list; 1955 1956 error = vfs_copyopt(opts, "jid", &jid, sizeof(jid)); 1957 if (error == 0) { 1958 if (jid != 0) { 1959 pr = prison_find_child(mypr, jid); 1960 if (pr != NULL) { 1961 if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) { 1962 mtx_unlock(&pr->pr_mtx); 1963 error = ENOENT; 1964 vfs_opterror(opts, "jail %d is dying", 1965 jid); 1966 goto done_unlock_list; 1967 } 1968 goto found_prison; 1969 } 1970 error = ENOENT; 1971 vfs_opterror(opts, "jail %d not found", jid); 1972 goto done_unlock_list; 1973 } 1974 } else if (error != ENOENT) 1975 goto done_unlock_list; 1976 1977 error = vfs_getopt(opts, "name", (void **)&name, &len); 1978 if (error == 0) { 1979 if (len == 0 || name[len - 1] != '\0') { 1980 error = EINVAL; 1981 goto done_unlock_list; 1982 } 1983 pr = prison_find_name(mypr, name); 1984 if (pr != NULL) { 1985 if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) { 1986 mtx_unlock(&pr->pr_mtx); 1987 error = ENOENT; 1988 vfs_opterror(opts, "jail \"%s\" is dying", 1989 name); 1990 goto done_unlock_list; 1991 } 1992 goto found_prison; 1993 } 1994 error = ENOENT; 1995 vfs_opterror(opts, "jail \"%s\" not found", name); 1996 goto done_unlock_list; 1997 } else if (error != ENOENT) 1998 goto done_unlock_list; 1999 2000 vfs_opterror(opts, "no jail specified"); 2001 error = ENOENT; 2002 goto done_unlock_list; 2003 2004 found_prison: 2005 /* Get the parameters of the prison. */ 2006 pr->pr_ref++; 2007 locked = PD_LOCKED; 2008 td->td_retval[0] = pr->pr_id; 2009 error = vfs_setopt(opts, "jid", &pr->pr_id, sizeof(pr->pr_id)); 2010 if (error != 0 && error != ENOENT) 2011 goto done_deref; 2012 i = (pr->pr_parent == mypr) ? 0 : pr->pr_parent->pr_id; 2013 error = vfs_setopt(opts, "parent", &i, sizeof(i)); 2014 if (error != 0 && error != ENOENT) 2015 goto done_deref; 2016 error = vfs_setopts(opts, "name", prison_name(mypr, pr)); 2017 if (error != 0 && error != ENOENT) 2018 goto done_deref; 2019 error = vfs_setopt(opts, "cpuset.id", &pr->pr_cpuset->cs_id, 2020 sizeof(pr->pr_cpuset->cs_id)); 2021 if (error != 0 && error != ENOENT) 2022 goto done_deref; 2023 error = vfs_setopts(opts, "path", prison_path(mypr, pr)); 2024 if (error != 0 && error != ENOENT) 2025 goto done_deref; 2026#ifdef INET 2027 error = vfs_setopt_part(opts, "ip4.addr", pr->pr_ip4, 2028 pr->pr_ip4s * sizeof(*pr->pr_ip4)); 2029 if (error != 0 && error != ENOENT) 2030 goto done_deref; 2031#endif 2032#ifdef INET6 2033 error = vfs_setopt_part(opts, "ip6.addr", pr->pr_ip6, 2034 pr->pr_ip6s * sizeof(*pr->pr_ip6)); 2035 if (error != 0 && error != ENOENT) 2036 goto done_deref; 2037#endif 2038 error = vfs_setopt(opts, "securelevel", &pr->pr_securelevel, 2039 sizeof(pr->pr_securelevel)); 2040 if (error != 0 && error != ENOENT) 2041 goto done_deref; 2042 error = vfs_setopt(opts, "children.cur", &pr->pr_childcount, 2043 sizeof(pr->pr_childcount)); 2044 if (error != 0 && error != ENOENT) 2045 goto done_deref; 2046 error = vfs_setopt(opts, "children.max", &pr->pr_childmax, 2047 sizeof(pr->pr_childmax)); 2048 if (error != 0 && error != ENOENT) 2049 goto done_deref; 2050 error = vfs_setopts(opts, "host.hostname", pr->pr_hostname); 2051 if (error != 0 && error != ENOENT) 2052 goto done_deref; 2053 error = vfs_setopts(opts, "host.domainname", pr->pr_domainname); 2054 if (error != 0 && error != ENOENT) 2055 goto done_deref; 2056 error = vfs_setopts(opts, "host.hostuuid", pr->pr_hostuuid); 2057 if (error != 0 && error != ENOENT) 2058 goto done_deref; 2059#ifdef COMPAT_FREEBSD32 2060 if (SV_PROC_FLAG(td->td_proc, SV_ILP32)) { 2061 uint32_t hid32 = pr->pr_hostid; 2062 2063 error = vfs_setopt(opts, "host.hostid", &hid32, sizeof(hid32)); 2064 } else 2065#endif 2066 error = vfs_setopt(opts, "host.hostid", &pr->pr_hostid, 2067 sizeof(pr->pr_hostid)); 2068 if (error != 0 && error != ENOENT) 2069 goto done_deref; 2070 error = vfs_setopt(opts, "enforce_statfs", &pr->pr_enforce_statfs, 2071 sizeof(pr->pr_enforce_statfs)); 2072 if (error != 0 && error != ENOENT) 2073 goto done_deref; 2074 error = vfs_setopt(opts, "devfs_ruleset", &pr->pr_devfs_rsnum, 2075 sizeof(pr->pr_devfs_rsnum)); 2076 if (error != 0 && error != ENOENT) 2077 goto done_deref; 2078 for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]); 2079 fi++) { 2080 if (pr_flag_names[fi] == NULL) 2081 continue; 2082 i = (pr->pr_flags & (1 << fi)) ? 1 : 0; 2083 error = vfs_setopt(opts, pr_flag_names[fi], &i, sizeof(i)); 2084 if (error != 0 && error != ENOENT) 2085 goto done_deref; 2086 i = !i; 2087 error = vfs_setopt(opts, pr_flag_nonames[fi], &i, sizeof(i)); 2088 if (error != 0 && error != ENOENT) 2089 goto done_deref; 2090 } 2091 for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]); 2092 fi++) { 2093 i = pr->pr_flags & 2094 (pr_flag_jailsys[fi].disable | pr_flag_jailsys[fi].new); 2095 i = pr_flag_jailsys[fi].disable && 2096 (i == pr_flag_jailsys[fi].disable) ? JAIL_SYS_DISABLE 2097 : (i == pr_flag_jailsys[fi].new) ? JAIL_SYS_NEW 2098 : JAIL_SYS_INHERIT; 2099 error = 2100 vfs_setopt(opts, pr_flag_jailsys[fi].name, &i, sizeof(i)); 2101 if (error != 0 && error != ENOENT) 2102 goto done_deref; 2103 } 2104 for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]); 2105 fi++) { 2106 if (pr_allow_names[fi] == NULL) 2107 continue; 2108 i = (pr->pr_allow & (1 << fi)) ? 1 : 0; 2109 error = vfs_setopt(opts, pr_allow_names[fi], &i, sizeof(i)); 2110 if (error != 0 && error != ENOENT) 2111 goto done_deref; 2112 i = !i; 2113 error = vfs_setopt(opts, pr_allow_nonames[fi], &i, sizeof(i)); 2114 if (error != 0 && error != ENOENT) 2115 goto done_deref; 2116 } 2117 i = (pr->pr_uref == 0); 2118 error = vfs_setopt(opts, "dying", &i, sizeof(i)); 2119 if (error != 0 && error != ENOENT) 2120 goto done_deref; 2121 i = !i; 2122 error = vfs_setopt(opts, "nodying", &i, sizeof(i)); 2123 if (error != 0 && error != ENOENT) 2124 goto done_deref; 2125 2126 /* Get the module parameters. */ 2127 mtx_unlock(&pr->pr_mtx); 2128 locked = 0; 2129 error = osd_jail_call(pr, PR_METHOD_GET, opts); 2130 if (error) 2131 goto done_deref; 2132 prison_deref(pr, PD_DEREF | PD_LIST_SLOCKED); 2133 2134 /* By now, all parameters should have been noted. */ 2135 TAILQ_FOREACH(opt, opts, link) { 2136 if (!opt->seen && strcmp(opt->name, "errmsg")) { 2137 error = EINVAL; 2138 vfs_opterror(opts, "unknown parameter: %s", opt->name); 2139 goto done_errmsg; 2140 } 2141 } 2142 2143 /* Write the fetched parameters back to userspace. */ 2144 error = 0; 2145 TAILQ_FOREACH(opt, opts, link) { 2146 if (opt->pos >= 0 && opt->pos != errmsg_pos) { 2147 pos = 2 * opt->pos + 1; 2148 optuio->uio_iov[pos].iov_len = opt->len; 2149 if (opt->value != NULL) { 2150 if (optuio->uio_segflg == UIO_SYSSPACE) { 2151 bcopy(opt->value, 2152 optuio->uio_iov[pos].iov_base, 2153 opt->len); 2154 } else { 2155 error = copyout(opt->value, 2156 optuio->uio_iov[pos].iov_base, 2157 opt->len); 2158 if (error) 2159 break; 2160 } 2161 } 2162 } 2163 } 2164 goto done_errmsg; 2165 2166 done_deref: 2167 prison_deref(pr, locked | PD_DEREF | PD_LIST_SLOCKED); 2168 goto done_errmsg; 2169 2170 done_unlock_list: 2171 sx_sunlock(&allprison_lock); 2172 done_errmsg: 2173 if (error && errmsg_pos >= 0) { 2174 vfs_getopt(opts, "errmsg", (void **)&errmsg, &errmsg_len); 2175 errmsg_pos = 2 * errmsg_pos + 1; 2176 if (errmsg_len > 0) { 2177 if (optuio->uio_segflg == UIO_SYSSPACE) 2178 bcopy(errmsg, 2179 optuio->uio_iov[errmsg_pos].iov_base, 2180 errmsg_len); 2181 else 2182 copyout(errmsg, 2183 optuio->uio_iov[errmsg_pos].iov_base, 2184 errmsg_len); 2185 } 2186 } 2187 vfs_freeopts(opts); 2188 return (error); 2189} 2190 2191 2192/* 2193 * struct jail_remove_args { 2194 * int jid; 2195 * }; 2196 */ 2197int 2198sys_jail_remove(struct thread *td, struct jail_remove_args *uap) 2199{ 2200 struct prison *pr, *cpr, *lpr, *tpr; 2201 int descend, error; 2202 2203 error = priv_check(td, PRIV_JAIL_REMOVE); 2204 if (error) 2205 return (error); 2206 2207 sx_xlock(&allprison_lock); 2208 pr = prison_find_child(td->td_ucred->cr_prison, uap->jid); 2209 if (pr == NULL) { 2210 sx_xunlock(&allprison_lock); 2211 return (EINVAL); 2212 } 2213 2214 /* Remove all descendants of this prison, then remove this prison. */ 2215 pr->pr_ref++; 2216 pr->pr_flags |= PR_REMOVE; 2217 if (!LIST_EMPTY(&pr->pr_children)) { 2218 mtx_unlock(&pr->pr_mtx); 2219 lpr = NULL; 2220 FOREACH_PRISON_DESCENDANT(pr, cpr, descend) { 2221 mtx_lock(&cpr->pr_mtx); 2222 if (cpr->pr_ref > 0) { 2223 tpr = cpr; 2224 cpr->pr_ref++; 2225 cpr->pr_flags |= PR_REMOVE; 2226 } else { 2227 /* Already removed - do not do it again. */ 2228 tpr = NULL; 2229 } 2230 mtx_unlock(&cpr->pr_mtx); 2231 if (lpr != NULL) { 2232 mtx_lock(&lpr->pr_mtx); 2233 prison_remove_one(lpr); 2234 sx_xlock(&allprison_lock); 2235 } 2236 lpr = tpr; 2237 } 2238 if (lpr != NULL) { 2239 mtx_lock(&lpr->pr_mtx); 2240 prison_remove_one(lpr); 2241 sx_xlock(&allprison_lock); 2242 } 2243 mtx_lock(&pr->pr_mtx); 2244 } 2245 prison_remove_one(pr); 2246 return (0); 2247} 2248 2249static void 2250prison_remove_one(struct prison *pr) 2251{ 2252 struct proc *p; 2253 int deuref; 2254 2255 /* If the prison was persistent, it is not anymore. */ 2256 deuref = 0; 2257 if (pr->pr_flags & PR_PERSIST) { 2258 pr->pr_ref--; 2259 deuref = PD_DEUREF; 2260 pr->pr_flags &= ~PR_PERSIST; 2261 } 2262 2263 /* 2264 * jail_remove added a reference. If that's the only one, remove 2265 * the prison now. 2266 */ 2267 KASSERT(pr->pr_ref > 0, 2268 ("prison_remove_one removing a dead prison (jid=%d)", pr->pr_id)); 2269 if (pr->pr_ref == 1) { 2270 prison_deref(pr, 2271 deuref | PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED); 2272 return; 2273 } 2274 2275 mtx_unlock(&pr->pr_mtx); 2276 sx_xunlock(&allprison_lock); 2277 /* 2278 * Kill all processes unfortunate enough to be attached to this prison. 2279 */ 2280 sx_slock(&allproc_lock); 2281 LIST_FOREACH(p, &allproc, p_list) { 2282 PROC_LOCK(p); 2283 if (p->p_state != PRS_NEW && p->p_ucred && 2284 p->p_ucred->cr_prison == pr) 2285 kern_psignal(p, SIGKILL); 2286 PROC_UNLOCK(p); 2287 } 2288 sx_sunlock(&allproc_lock); 2289 /* Remove the temporary reference added by jail_remove. */ 2290 prison_deref(pr, deuref | PD_DEREF); 2291} 2292 2293 2294/* 2295 * struct jail_attach_args { 2296 * int jid; 2297 * }; 2298 */ 2299int 2300sys_jail_attach(struct thread *td, struct jail_attach_args *uap) 2301{ 2302 struct prison *pr; 2303 int error; 2304 2305 error = priv_check(td, PRIV_JAIL_ATTACH); 2306 if (error) 2307 return (error); 2308 2309 sx_slock(&allprison_lock); 2310 pr = prison_find_child(td->td_ucred->cr_prison, uap->jid); 2311 if (pr == NULL) { 2312 sx_sunlock(&allprison_lock); 2313 return (EINVAL); 2314 } 2315 2316 /* 2317 * Do not allow a process to attach to a prison that is not 2318 * considered to be "alive". 2319 */ 2320 if (pr->pr_uref == 0) { 2321 mtx_unlock(&pr->pr_mtx); 2322 sx_sunlock(&allprison_lock); 2323 return (EINVAL); 2324 } 2325 2326 return (do_jail_attach(td, pr)); 2327} 2328 2329static int 2330do_jail_attach(struct thread *td, struct prison *pr) 2331{ 2332 struct prison *ppr; 2333 struct proc *p; 2334 struct ucred *newcred, *oldcred; 2335 int error; 2336 2337 /* 2338 * XXX: Note that there is a slight race here if two threads 2339 * in the same privileged process attempt to attach to two 2340 * different jails at the same time. It is important for 2341 * user processes not to do this, or they might end up with 2342 * a process root from one prison, but attached to the jail 2343 * of another. 2344 */ 2345 pr->pr_ref++; 2346 pr->pr_uref++; 2347 mtx_unlock(&pr->pr_mtx); 2348 2349 /* Let modules do whatever they need to prepare for attaching. */ 2350 error = osd_jail_call(pr, PR_METHOD_ATTACH, td); 2351 if (error) { 2352 prison_deref(pr, PD_DEREF | PD_DEUREF | PD_LIST_SLOCKED); 2353 return (error); 2354 } 2355 sx_sunlock(&allprison_lock); 2356 2357 /* 2358 * Reparent the newly attached process to this jail. 2359 */ 2360 ppr = td->td_ucred->cr_prison; 2361 p = td->td_proc; 2362 error = cpuset_setproc_update_set(p, pr->pr_cpuset); 2363 if (error) 2364 goto e_revert_osd; 2365 2366 vn_lock(pr->pr_root, LK_EXCLUSIVE | LK_RETRY); 2367 if ((error = change_dir(pr->pr_root, td)) != 0) 2368 goto e_unlock; 2369#ifdef MAC 2370 if ((error = mac_vnode_check_chroot(td->td_ucred, pr->pr_root))) 2371 goto e_unlock; 2372#endif 2373 VOP_UNLOCK(pr->pr_root, 0); 2374 if ((error = change_root(pr->pr_root, td))) 2375 goto e_revert_osd; 2376 2377 newcred = crget(); 2378 PROC_LOCK(p); 2379 oldcred = p->p_ucred; 2380 setsugid(p); 2381 crcopy(newcred, oldcred); 2382 newcred->cr_prison = pr; 2383 p->p_ucred = newcred; 2384 PROC_UNLOCK(p); 2385#ifdef RACCT 2386 racct_proc_ucred_changed(p, oldcred, newcred); 2387#endif 2388 crfree(oldcred); 2389 prison_deref(ppr, PD_DEREF | PD_DEUREF); 2390 return (0); 2391 e_unlock: 2392 VOP_UNLOCK(pr->pr_root, 0); 2393 e_revert_osd: 2394 /* Tell modules this thread is still in its old jail after all. */ 2395 (void)osd_jail_call(ppr, PR_METHOD_ATTACH, td); 2396 prison_deref(pr, PD_DEREF | PD_DEUREF); 2397 return (error); 2398} 2399 2400 2401/* 2402 * Returns a locked prison instance, or NULL on failure. 2403 */ 2404struct prison * 2405prison_find(int prid) 2406{ 2407 struct prison *pr; 2408 2409 sx_assert(&allprison_lock, SX_LOCKED); 2410 TAILQ_FOREACH(pr, &allprison, pr_list) { 2411 if (pr->pr_id == prid) { 2412 mtx_lock(&pr->pr_mtx); 2413 if (pr->pr_ref > 0) 2414 return (pr); 2415 mtx_unlock(&pr->pr_mtx); 2416 } 2417 } 2418 return (NULL); 2419} 2420 2421/* 2422 * Find a prison that is a descendant of mypr. Returns a locked prison or NULL. 2423 */ 2424struct prison * 2425prison_find_child(struct prison *mypr, int prid) 2426{ 2427 struct prison *pr; 2428 int descend; 2429 2430 sx_assert(&allprison_lock, SX_LOCKED); 2431 FOREACH_PRISON_DESCENDANT(mypr, pr, descend) { 2432 if (pr->pr_id == prid) { 2433 mtx_lock(&pr->pr_mtx); 2434 if (pr->pr_ref > 0) 2435 return (pr); 2436 mtx_unlock(&pr->pr_mtx); 2437 } 2438 } 2439 return (NULL); 2440} 2441 2442/* 2443 * Look for the name relative to mypr. Returns a locked prison or NULL. 2444 */ 2445struct prison * 2446prison_find_name(struct prison *mypr, const char *name) 2447{ 2448 struct prison *pr, *deadpr; 2449 size_t mylen; 2450 int descend; 2451 2452 sx_assert(&allprison_lock, SX_LOCKED); 2453 mylen = (mypr == &prison0) ? 0 : strlen(mypr->pr_name) + 1; 2454 again: 2455 deadpr = NULL; 2456 FOREACH_PRISON_DESCENDANT(mypr, pr, descend) { 2457 if (!strcmp(pr->pr_name + mylen, name)) { 2458 mtx_lock(&pr->pr_mtx); 2459 if (pr->pr_ref > 0) { 2460 if (pr->pr_uref > 0) 2461 return (pr); 2462 deadpr = pr; 2463 } 2464 mtx_unlock(&pr->pr_mtx); 2465 } 2466 } 2467 /* There was no valid prison - perhaps there was a dying one. */ 2468 if (deadpr != NULL) { 2469 mtx_lock(&deadpr->pr_mtx); 2470 if (deadpr->pr_ref == 0) { 2471 mtx_unlock(&deadpr->pr_mtx); 2472 goto again; 2473 } 2474 } 2475 return (deadpr); 2476} 2477 2478/* 2479 * See if a prison has the specific flag set. 2480 */ 2481int 2482prison_flag(struct ucred *cred, unsigned flag) 2483{ 2484 2485 /* This is an atomic read, so no locking is necessary. */ 2486 return (cred->cr_prison->pr_flags & flag); 2487} 2488 2489int 2490prison_allow(struct ucred *cred, unsigned flag) 2491{ 2492 2493 /* This is an atomic read, so no locking is necessary. */ 2494 return (cred->cr_prison->pr_allow & flag); 2495} 2496 2497/* 2498 * Remove a prison reference. If that was the last reference, remove the 2499 * prison itself - but not in this context in case there are locks held. 2500 */ 2501void 2502prison_free_locked(struct prison *pr) 2503{ 2504 2505 mtx_assert(&pr->pr_mtx, MA_OWNED); 2506 pr->pr_ref--; 2507 if (pr->pr_ref == 0) { 2508 mtx_unlock(&pr->pr_mtx); 2509 TASK_INIT(&pr->pr_task, 0, prison_complete, pr); 2510 taskqueue_enqueue(taskqueue_thread, &pr->pr_task); 2511 return; 2512 } 2513 mtx_unlock(&pr->pr_mtx); 2514} 2515 2516void 2517prison_free(struct prison *pr) 2518{ 2519 2520 mtx_lock(&pr->pr_mtx); 2521 prison_free_locked(pr); 2522} 2523 2524static void 2525prison_complete(void *context, int pending) 2526{ 2527 2528 prison_deref((struct prison *)context, 0); 2529} 2530 2531/* 2532 * Remove a prison reference (usually). This internal version assumes no 2533 * mutexes are held, except perhaps the prison itself. If there are no more 2534 * references, release and delist the prison. On completion, the prison lock 2535 * and the allprison lock are both unlocked. 2536 */ 2537static void 2538prison_deref(struct prison *pr, int flags) 2539{ 2540 struct prison *ppr, *tpr; 2541 2542 if (!(flags & PD_LOCKED)) 2543 mtx_lock(&pr->pr_mtx); 2544 for (;;) { 2545 if (flags & PD_DEUREF) { 2546 pr->pr_uref--; 2547 KASSERT(prison0.pr_uref != 0, ("prison0 pr_uref=0")); 2548 } 2549 if (flags & PD_DEREF) 2550 pr->pr_ref--; 2551 /* If the prison still has references, nothing else to do. */ 2552 if (pr->pr_ref > 0) { 2553 mtx_unlock(&pr->pr_mtx); 2554 if (flags & PD_LIST_SLOCKED) 2555 sx_sunlock(&allprison_lock); 2556 else if (flags & PD_LIST_XLOCKED) 2557 sx_xunlock(&allprison_lock); 2558 return; 2559 } 2560 2561 mtx_unlock(&pr->pr_mtx); 2562 if (flags & PD_LIST_SLOCKED) { 2563 if (!sx_try_upgrade(&allprison_lock)) { 2564 sx_sunlock(&allprison_lock); 2565 sx_xlock(&allprison_lock); 2566 } 2567 } else if (!(flags & PD_LIST_XLOCKED)) 2568 sx_xlock(&allprison_lock); 2569 2570 TAILQ_REMOVE(&allprison, pr, pr_list); 2571 LIST_REMOVE(pr, pr_sibling); 2572 ppr = pr->pr_parent; 2573 for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent) 2574 tpr->pr_childcount--; 2575 sx_xunlock(&allprison_lock); 2576 2577#ifdef VIMAGE 2578 if (pr->pr_vnet != ppr->pr_vnet) 2579 vnet_destroy(pr->pr_vnet); 2580#endif 2581 if (pr->pr_root != NULL) 2582 vrele(pr->pr_root); 2583 mtx_destroy(&pr->pr_mtx); 2584#ifdef INET 2585 free(pr->pr_ip4, M_PRISON); 2586#endif 2587#ifdef INET6 2588 free(pr->pr_ip6, M_PRISON); 2589#endif 2590 if (pr->pr_cpuset != NULL) 2591 cpuset_rel(pr->pr_cpuset); 2592 osd_jail_exit(pr); 2593#ifdef RACCT 2594 prison_racct_detach(pr); 2595#endif 2596 free(pr, M_PRISON); 2597 2598 /* Removing a prison frees a reference on its parent. */ 2599 pr = ppr; 2600 mtx_lock(&pr->pr_mtx); 2601 flags = PD_DEREF | PD_DEUREF; 2602 } 2603} 2604 2605void 2606prison_hold_locked(struct prison *pr) 2607{ 2608 2609 mtx_assert(&pr->pr_mtx, MA_OWNED); 2610 KASSERT(pr->pr_ref > 0, 2611 ("Trying to hold dead prison (jid=%d).", pr->pr_id)); 2612 pr->pr_ref++; 2613} 2614 2615void 2616prison_hold(struct prison *pr) 2617{ 2618 2619 mtx_lock(&pr->pr_mtx); 2620 prison_hold_locked(pr); 2621 mtx_unlock(&pr->pr_mtx); 2622} 2623 2624void 2625prison_proc_hold(struct prison *pr) 2626{ 2627 2628 mtx_lock(&pr->pr_mtx); 2629 KASSERT(pr->pr_uref > 0, 2630 ("Cannot add a process to a non-alive prison (jid=%d)", pr->pr_id)); 2631 pr->pr_uref++; 2632 mtx_unlock(&pr->pr_mtx); 2633} 2634 2635void 2636prison_proc_free(struct prison *pr) 2637{ 2638 2639 mtx_lock(&pr->pr_mtx); 2640 KASSERT(pr->pr_uref > 0, 2641 ("Trying to kill a process in a dead prison (jid=%d)", pr->pr_id)); 2642 prison_deref(pr, PD_DEUREF | PD_LOCKED); 2643} 2644 2645 2646#ifdef INET 2647/* 2648 * Restrict a prison's IP address list with its parent's, possibly replacing 2649 * it. Return true if the replacement buffer was used (or would have been). 2650 */ 2651static int 2652prison_restrict_ip4(struct prison *pr, struct in_addr *newip4) 2653{ 2654 int ii, ij, used; 2655 struct prison *ppr; 2656 2657 ppr = pr->pr_parent; 2658 if (!(pr->pr_flags & PR_IP4_USER)) { 2659 /* This has no user settings, so just copy the parent's list. */ 2660 if (pr->pr_ip4s < ppr->pr_ip4s) { 2661 /* 2662 * There's no room for the parent's list. Use the 2663 * new list buffer, which is assumed to be big enough 2664 * (if it was passed). If there's no buffer, try to 2665 * allocate one. 2666 */ 2667 used = 1; 2668 if (newip4 == NULL) { 2669 newip4 = malloc(ppr->pr_ip4s * sizeof(*newip4), 2670 M_PRISON, M_NOWAIT); 2671 if (newip4 != NULL) 2672 used = 0; 2673 } 2674 if (newip4 != NULL) { 2675 bcopy(ppr->pr_ip4, newip4, 2676 ppr->pr_ip4s * sizeof(*newip4)); 2677 free(pr->pr_ip4, M_PRISON); 2678 pr->pr_ip4 = newip4; 2679 pr->pr_ip4s = ppr->pr_ip4s; 2680 } 2681 return (used); 2682 } 2683 pr->pr_ip4s = ppr->pr_ip4s; 2684 if (pr->pr_ip4s > 0) 2685 bcopy(ppr->pr_ip4, pr->pr_ip4, 2686 pr->pr_ip4s * sizeof(*newip4)); 2687 else if (pr->pr_ip4 != NULL) { 2688 free(pr->pr_ip4, M_PRISON); 2689 pr->pr_ip4 = NULL; 2690 } 2691 } else if (pr->pr_ip4s > 0) { 2692 /* Remove addresses that aren't in the parent. */ 2693 for (ij = 0; ij < ppr->pr_ip4s; ij++) 2694 if (pr->pr_ip4[0].s_addr == ppr->pr_ip4[ij].s_addr) 2695 break; 2696 if (ij < ppr->pr_ip4s) 2697 ii = 1; 2698 else { 2699 bcopy(pr->pr_ip4 + 1, pr->pr_ip4, 2700 --pr->pr_ip4s * sizeof(*pr->pr_ip4)); 2701 ii = 0; 2702 } 2703 for (ij = 1; ii < pr->pr_ip4s; ) { 2704 if (pr->pr_ip4[ii].s_addr == ppr->pr_ip4[0].s_addr) { 2705 ii++; 2706 continue; 2707 } 2708 switch (ij >= ppr->pr_ip4s ? -1 : 2709 qcmp_v4(&pr->pr_ip4[ii], &ppr->pr_ip4[ij])) { 2710 case -1: 2711 bcopy(pr->pr_ip4 + ii + 1, pr->pr_ip4 + ii, 2712 (--pr->pr_ip4s - ii) * sizeof(*pr->pr_ip4)); 2713 break; 2714 case 0: 2715 ii++; 2716 ij++; 2717 break; 2718 case 1: 2719 ij++; 2720 break; 2721 } 2722 } 2723 if (pr->pr_ip4s == 0) { 2724 pr->pr_flags |= PR_IP4_DISABLE; 2725 free(pr->pr_ip4, M_PRISON); 2726 pr->pr_ip4 = NULL; 2727 } 2728 } 2729 return (0); 2730} 2731 2732/* 2733 * Pass back primary IPv4 address of this jail. 2734 * 2735 * If not restricted return success but do not alter the address. Caller has 2736 * to make sure to initialize it correctly (e.g. INADDR_ANY). 2737 * 2738 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv4. 2739 * Address returned in NBO. 2740 */ 2741int 2742prison_get_ip4(struct ucred *cred, struct in_addr *ia) 2743{ 2744 struct prison *pr; 2745 2746 KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); 2747 KASSERT(ia != NULL, ("%s: ia is NULL", __func__)); 2748 2749 pr = cred->cr_prison; 2750 if (!(pr->pr_flags & PR_IP4)) 2751 return (0); 2752 mtx_lock(&pr->pr_mtx); 2753 if (!(pr->pr_flags & PR_IP4)) { 2754 mtx_unlock(&pr->pr_mtx); 2755 return (0); 2756 } 2757 if (pr->pr_ip4 == NULL) { 2758 mtx_unlock(&pr->pr_mtx); 2759 return (EAFNOSUPPORT); 2760 } 2761 2762 ia->s_addr = pr->pr_ip4[0].s_addr; 2763 mtx_unlock(&pr->pr_mtx); 2764 return (0); 2765} 2766 2767/* 2768 * Return 1 if we should do proper source address selection or are not jailed. 2769 * We will return 0 if we should bypass source address selection in favour 2770 * of the primary jail IPv4 address. Only in this case *ia will be updated and 2771 * returned in NBO. 2772 * Return EAFNOSUPPORT, in case this jail does not allow IPv4. 2773 */ 2774int 2775prison_saddrsel_ip4(struct ucred *cred, struct in_addr *ia) 2776{ 2777 struct prison *pr; 2778 struct in_addr lia; 2779 int error; 2780 2781 KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); 2782 KASSERT(ia != NULL, ("%s: ia is NULL", __func__)); 2783 2784 if (!jailed(cred)) 2785 return (1); 2786 2787 pr = cred->cr_prison; 2788 if (pr->pr_flags & PR_IP4_SADDRSEL) 2789 return (1); 2790 2791 lia.s_addr = INADDR_ANY; 2792 error = prison_get_ip4(cred, &lia); 2793 if (error) 2794 return (error); 2795 if (lia.s_addr == INADDR_ANY) 2796 return (1); 2797 2798 ia->s_addr = lia.s_addr; 2799 return (0); 2800} 2801 2802/* 2803 * Return true if pr1 and pr2 have the same IPv4 address restrictions. 2804 */ 2805int 2806prison_equal_ip4(struct prison *pr1, struct prison *pr2) 2807{ 2808 2809 if (pr1 == pr2) 2810 return (1); 2811 2812 /* 2813 * No need to lock since the PR_IP4_USER flag can't be altered for 2814 * existing prisons. 2815 */ 2816 while (pr1 != &prison0 && 2817#ifdef VIMAGE 2818 !(pr1->pr_flags & PR_VNET) && 2819#endif 2820 !(pr1->pr_flags & PR_IP4_USER)) 2821 pr1 = pr1->pr_parent; 2822 while (pr2 != &prison0 && 2823#ifdef VIMAGE 2824 !(pr2->pr_flags & PR_VNET) && 2825#endif 2826 !(pr2->pr_flags & PR_IP4_USER)) 2827 pr2 = pr2->pr_parent; 2828 return (pr1 == pr2); 2829} 2830 2831/* 2832 * Make sure our (source) address is set to something meaningful to this 2833 * jail. 2834 * 2835 * Returns 0 if jail doesn't restrict IPv4 or if address belongs to jail, 2836 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail 2837 * doesn't allow IPv4. Address passed in in NBO and returned in NBO. 2838 */ 2839int 2840prison_local_ip4(struct ucred *cred, struct in_addr *ia) 2841{ 2842 struct prison *pr; 2843 struct in_addr ia0; 2844 int error; 2845 2846 KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); 2847 KASSERT(ia != NULL, ("%s: ia is NULL", __func__)); 2848 2849 pr = cred->cr_prison; 2850 if (!(pr->pr_flags & PR_IP4)) 2851 return (0); 2852 mtx_lock(&pr->pr_mtx); 2853 if (!(pr->pr_flags & PR_IP4)) { 2854 mtx_unlock(&pr->pr_mtx); 2855 return (0); 2856 } 2857 if (pr->pr_ip4 == NULL) { 2858 mtx_unlock(&pr->pr_mtx); 2859 return (EAFNOSUPPORT); 2860 } 2861 2862 ia0.s_addr = ntohl(ia->s_addr); 2863 if (ia0.s_addr == INADDR_LOOPBACK) { 2864 ia->s_addr = pr->pr_ip4[0].s_addr; 2865 mtx_unlock(&pr->pr_mtx); 2866 return (0); 2867 } 2868 2869 if (ia0.s_addr == INADDR_ANY) { 2870 /* 2871 * In case there is only 1 IPv4 address, bind directly. 2872 */ 2873 if (pr->pr_ip4s == 1) 2874 ia->s_addr = pr->pr_ip4[0].s_addr; 2875 mtx_unlock(&pr->pr_mtx); 2876 return (0); 2877 } 2878 2879 error = _prison_check_ip4(pr, ia); 2880 mtx_unlock(&pr->pr_mtx); 2881 return (error); 2882} 2883 2884/* 2885 * Rewrite destination address in case we will connect to loopback address. 2886 * 2887 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv4. 2888 * Address passed in in NBO and returned in NBO. 2889 */ 2890int 2891prison_remote_ip4(struct ucred *cred, struct in_addr *ia) 2892{ 2893 struct prison *pr; 2894 2895 KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); 2896 KASSERT(ia != NULL, ("%s: ia is NULL", __func__)); 2897 2898 pr = cred->cr_prison; 2899 if (!(pr->pr_flags & PR_IP4)) 2900 return (0); 2901 mtx_lock(&pr->pr_mtx); 2902 if (!(pr->pr_flags & PR_IP4)) { 2903 mtx_unlock(&pr->pr_mtx); 2904 return (0); 2905 } 2906 if (pr->pr_ip4 == NULL) { 2907 mtx_unlock(&pr->pr_mtx); 2908 return (EAFNOSUPPORT); 2909 } 2910 2911 if (ntohl(ia->s_addr) == INADDR_LOOPBACK) { 2912 ia->s_addr = pr->pr_ip4[0].s_addr; 2913 mtx_unlock(&pr->pr_mtx); 2914 return (0); 2915 } 2916 2917 /* 2918 * Return success because nothing had to be changed. 2919 */ 2920 mtx_unlock(&pr->pr_mtx); 2921 return (0); 2922} 2923 2924/* 2925 * Check if given address belongs to the jail referenced by cred/prison. 2926 * 2927 * Returns 0 if jail doesn't restrict IPv4 or if address belongs to jail, 2928 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail 2929 * doesn't allow IPv4. Address passed in in NBO. 2930 */ 2931static int 2932_prison_check_ip4(struct prison *pr, struct in_addr *ia) 2933{ 2934 int i, a, z, d; 2935 2936 /* 2937 * Check the primary IP. 2938 */ 2939 if (pr->pr_ip4[0].s_addr == ia->s_addr) 2940 return (0); 2941 2942 /* 2943 * All the other IPs are sorted so we can do a binary search. 2944 */ 2945 a = 0; 2946 z = pr->pr_ip4s - 2; 2947 while (a <= z) { 2948 i = (a + z) / 2; 2949 d = qcmp_v4(&pr->pr_ip4[i+1], ia); 2950 if (d > 0) 2951 z = i - 1; 2952 else if (d < 0) 2953 a = i + 1; 2954 else 2955 return (0); 2956 } 2957 2958 return (EADDRNOTAVAIL); 2959} 2960 2961int 2962prison_check_ip4(struct ucred *cred, struct in_addr *ia) 2963{ 2964 struct prison *pr; 2965 int error; 2966 2967 KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); 2968 KASSERT(ia != NULL, ("%s: ia is NULL", __func__)); 2969 2970 pr = cred->cr_prison; 2971 if (!(pr->pr_flags & PR_IP4)) 2972 return (0); 2973 mtx_lock(&pr->pr_mtx); 2974 if (!(pr->pr_flags & PR_IP4)) { 2975 mtx_unlock(&pr->pr_mtx); 2976 return (0); 2977 } 2978 if (pr->pr_ip4 == NULL) { 2979 mtx_unlock(&pr->pr_mtx); 2980 return (EAFNOSUPPORT); 2981 } 2982 2983 error = _prison_check_ip4(pr, ia); 2984 mtx_unlock(&pr->pr_mtx); 2985 return (error); 2986} 2987#endif 2988 2989#ifdef INET6 2990static int 2991prison_restrict_ip6(struct prison *pr, struct in6_addr *newip6) 2992{ 2993 int ii, ij, used; 2994 struct prison *ppr; 2995 2996 ppr = pr->pr_parent; 2997 if (!(pr->pr_flags & PR_IP6_USER)) { 2998 /* This has no user settings, so just copy the parent's list. */ 2999 if (pr->pr_ip6s < ppr->pr_ip6s) { 3000 /* 3001 * There's no room for the parent's list. Use the 3002 * new list buffer, which is assumed to be big enough 3003 * (if it was passed). If there's no buffer, try to 3004 * allocate one. 3005 */ 3006 used = 1; 3007 if (newip6 == NULL) { 3008 newip6 = malloc(ppr->pr_ip6s * sizeof(*newip6), 3009 M_PRISON, M_NOWAIT); 3010 if (newip6 != NULL) 3011 used = 0; 3012 } 3013 if (newip6 != NULL) { 3014 bcopy(ppr->pr_ip6, newip6, 3015 ppr->pr_ip6s * sizeof(*newip6)); 3016 free(pr->pr_ip6, M_PRISON); 3017 pr->pr_ip6 = newip6; 3018 pr->pr_ip6s = ppr->pr_ip6s; 3019 } 3020 return (used); 3021 } 3022 pr->pr_ip6s = ppr->pr_ip6s; 3023 if (pr->pr_ip6s > 0) 3024 bcopy(ppr->pr_ip6, pr->pr_ip6, 3025 pr->pr_ip6s * sizeof(*newip6)); 3026 else if (pr->pr_ip6 != NULL) { 3027 free(pr->pr_ip6, M_PRISON); 3028 pr->pr_ip6 = NULL; 3029 } 3030 } else if (pr->pr_ip6s > 0) { 3031 /* Remove addresses that aren't in the parent. */ 3032 for (ij = 0; ij < ppr->pr_ip6s; ij++) 3033 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[0], 3034 &ppr->pr_ip6[ij])) 3035 break; 3036 if (ij < ppr->pr_ip6s) 3037 ii = 1; 3038 else { 3039 bcopy(pr->pr_ip6 + 1, pr->pr_ip6, 3040 --pr->pr_ip6s * sizeof(*pr->pr_ip6)); 3041 ii = 0; 3042 } 3043 for (ij = 1; ii < pr->pr_ip6s; ) { 3044 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[ii], 3045 &ppr->pr_ip6[0])) { 3046 ii++; 3047 continue; 3048 } 3049 switch (ij >= ppr->pr_ip4s ? -1 : 3050 qcmp_v6(&pr->pr_ip6[ii], &ppr->pr_ip6[ij])) { 3051 case -1: 3052 bcopy(pr->pr_ip6 + ii + 1, pr->pr_ip6 + ii, 3053 (--pr->pr_ip6s - ii) * sizeof(*pr->pr_ip6)); 3054 break; 3055 case 0: 3056 ii++; 3057 ij++; 3058 break; 3059 case 1: 3060 ij++; 3061 break; 3062 } 3063 } 3064 if (pr->pr_ip6s == 0) { 3065 pr->pr_flags |= PR_IP6_DISABLE; 3066 free(pr->pr_ip6, M_PRISON); 3067 pr->pr_ip6 = NULL; 3068 } 3069 } 3070 return 0; 3071} 3072 3073/* 3074 * Pass back primary IPv6 address for this jail. 3075 * 3076 * If not restricted return success but do not alter the address. Caller has 3077 * to make sure to initialize it correctly (e.g. IN6ADDR_ANY_INIT). 3078 * 3079 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv6. 3080 */ 3081int 3082prison_get_ip6(struct ucred *cred, struct in6_addr *ia6) 3083{ 3084 struct prison *pr; 3085 3086 KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); 3087 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__)); 3088 3089 pr = cred->cr_prison; 3090 if (!(pr->pr_flags & PR_IP6)) 3091 return (0); 3092 mtx_lock(&pr->pr_mtx); 3093 if (!(pr->pr_flags & PR_IP6)) { 3094 mtx_unlock(&pr->pr_mtx); 3095 return (0); 3096 } 3097 if (pr->pr_ip6 == NULL) { 3098 mtx_unlock(&pr->pr_mtx); 3099 return (EAFNOSUPPORT); 3100 } 3101 3102 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr)); 3103 mtx_unlock(&pr->pr_mtx); 3104 return (0); 3105} 3106 3107/* 3108 * Return 1 if we should do proper source address selection or are not jailed. 3109 * We will return 0 if we should bypass source address selection in favour 3110 * of the primary jail IPv6 address. Only in this case *ia will be updated and 3111 * returned in NBO. 3112 * Return EAFNOSUPPORT, in case this jail does not allow IPv6. 3113 */ 3114int 3115prison_saddrsel_ip6(struct ucred *cred, struct in6_addr *ia6) 3116{ 3117 struct prison *pr; 3118 struct in6_addr lia6; 3119 int error; 3120 3121 KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); 3122 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__)); 3123 3124 if (!jailed(cred)) 3125 return (1); 3126 3127 pr = cred->cr_prison; 3128 if (pr->pr_flags & PR_IP6_SADDRSEL) 3129 return (1); 3130 3131 lia6 = in6addr_any; 3132 error = prison_get_ip6(cred, &lia6); 3133 if (error) 3134 return (error); 3135 if (IN6_IS_ADDR_UNSPECIFIED(&lia6)) 3136 return (1); 3137 3138 bcopy(&lia6, ia6, sizeof(struct in6_addr)); 3139 return (0); 3140} 3141 3142/* 3143 * Return true if pr1 and pr2 have the same IPv6 address restrictions. 3144 */ 3145int 3146prison_equal_ip6(struct prison *pr1, struct prison *pr2) 3147{ 3148 3149 if (pr1 == pr2) 3150 return (1); 3151 3152 while (pr1 != &prison0 && 3153#ifdef VIMAGE 3154 !(pr1->pr_flags & PR_VNET) && 3155#endif 3156 !(pr1->pr_flags & PR_IP6_USER)) 3157 pr1 = pr1->pr_parent; 3158 while (pr2 != &prison0 && 3159#ifdef VIMAGE 3160 !(pr2->pr_flags & PR_VNET) && 3161#endif 3162 !(pr2->pr_flags & PR_IP6_USER)) 3163 pr2 = pr2->pr_parent; 3164 return (pr1 == pr2); 3165} 3166 3167/* 3168 * Make sure our (source) address is set to something meaningful to this jail. 3169 * 3170 * v6only should be set based on (inp->inp_flags & IN6P_IPV6_V6ONLY != 0) 3171 * when needed while binding. 3172 * 3173 * Returns 0 if jail doesn't restrict IPv6 or if address belongs to jail, 3174 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail 3175 * doesn't allow IPv6. 3176 */ 3177int 3178prison_local_ip6(struct ucred *cred, struct in6_addr *ia6, int v6only) 3179{ 3180 struct prison *pr; 3181 int error; 3182 3183 KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); 3184 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__)); 3185 3186 pr = cred->cr_prison; 3187 if (!(pr->pr_flags & PR_IP6)) 3188 return (0); 3189 mtx_lock(&pr->pr_mtx); 3190 if (!(pr->pr_flags & PR_IP6)) { 3191 mtx_unlock(&pr->pr_mtx); 3192 return (0); 3193 } 3194 if (pr->pr_ip6 == NULL) { 3195 mtx_unlock(&pr->pr_mtx); 3196 return (EAFNOSUPPORT); 3197 } 3198 3199 if (IN6_IS_ADDR_LOOPBACK(ia6)) { 3200 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr)); 3201 mtx_unlock(&pr->pr_mtx); 3202 return (0); 3203 } 3204 3205 if (IN6_IS_ADDR_UNSPECIFIED(ia6)) { 3206 /* 3207 * In case there is only 1 IPv6 address, and v6only is true, 3208 * then bind directly. 3209 */ 3210 if (v6only != 0 && pr->pr_ip6s == 1) 3211 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr)); 3212 mtx_unlock(&pr->pr_mtx); 3213 return (0); 3214 } 3215 3216 error = _prison_check_ip6(pr, ia6); 3217 mtx_unlock(&pr->pr_mtx); 3218 return (error); 3219} 3220 3221/* 3222 * Rewrite destination address in case we will connect to loopback address. 3223 * 3224 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv6. 3225 */ 3226int 3227prison_remote_ip6(struct ucred *cred, struct in6_addr *ia6) 3228{ 3229 struct prison *pr; 3230 3231 KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); 3232 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__)); 3233 3234 pr = cred->cr_prison; 3235 if (!(pr->pr_flags & PR_IP6)) 3236 return (0); 3237 mtx_lock(&pr->pr_mtx); 3238 if (!(pr->pr_flags & PR_IP6)) { 3239 mtx_unlock(&pr->pr_mtx); 3240 return (0); 3241 } 3242 if (pr->pr_ip6 == NULL) { 3243 mtx_unlock(&pr->pr_mtx); 3244 return (EAFNOSUPPORT); 3245 } 3246 3247 if (IN6_IS_ADDR_LOOPBACK(ia6)) { 3248 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr)); 3249 mtx_unlock(&pr->pr_mtx); 3250 return (0); 3251 } 3252 3253 /* 3254 * Return success because nothing had to be changed. 3255 */ 3256 mtx_unlock(&pr->pr_mtx); 3257 return (0); 3258} 3259 3260/* 3261 * Check if given address belongs to the jail referenced by cred/prison. 3262 * 3263 * Returns 0 if jail doesn't restrict IPv6 or if address belongs to jail, 3264 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail 3265 * doesn't allow IPv6. 3266 */ 3267static int 3268_prison_check_ip6(struct prison *pr, struct in6_addr *ia6) 3269{ 3270 int i, a, z, d; 3271 3272 /* 3273 * Check the primary IP. 3274 */ 3275 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[0], ia6)) 3276 return (0); 3277 3278 /* 3279 * All the other IPs are sorted so we can do a binary search. 3280 */ 3281 a = 0; 3282 z = pr->pr_ip6s - 2; 3283 while (a <= z) { 3284 i = (a + z) / 2; 3285 d = qcmp_v6(&pr->pr_ip6[i+1], ia6); 3286 if (d > 0) 3287 z = i - 1; 3288 else if (d < 0) 3289 a = i + 1; 3290 else 3291 return (0); 3292 } 3293 3294 return (EADDRNOTAVAIL); 3295} 3296 3297int 3298prison_check_ip6(struct ucred *cred, struct in6_addr *ia6) 3299{ 3300 struct prison *pr; 3301 int error; 3302 3303 KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); 3304 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__)); 3305 3306 pr = cred->cr_prison; 3307 if (!(pr->pr_flags & PR_IP6)) 3308 return (0); 3309 mtx_lock(&pr->pr_mtx); 3310 if (!(pr->pr_flags & PR_IP6)) { 3311 mtx_unlock(&pr->pr_mtx); 3312 return (0); 3313 } 3314 if (pr->pr_ip6 == NULL) { 3315 mtx_unlock(&pr->pr_mtx); 3316 return (EAFNOSUPPORT); 3317 } 3318 3319 error = _prison_check_ip6(pr, ia6); 3320 mtx_unlock(&pr->pr_mtx); 3321 return (error); 3322} 3323#endif 3324 3325/* 3326 * Check if a jail supports the given address family. 3327 * 3328 * Returns 0 if not jailed or the address family is supported, EAFNOSUPPORT 3329 * if not. 3330 */ 3331int 3332prison_check_af(struct ucred *cred, int af) 3333{ 3334 struct prison *pr; 3335 int error; 3336 3337 KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); 3338 3339 pr = cred->cr_prison; 3340#ifdef VIMAGE 3341 /* Prisons with their own network stack are not limited. */ 3342 if (prison_owns_vnet(cred)) 3343 return (0); 3344#endif 3345 3346 error = 0; 3347 switch (af) 3348 { 3349#ifdef INET 3350 case AF_INET: 3351 if (pr->pr_flags & PR_IP4) 3352 { 3353 mtx_lock(&pr->pr_mtx); 3354 if ((pr->pr_flags & PR_IP4) && pr->pr_ip4 == NULL) 3355 error = EAFNOSUPPORT; 3356 mtx_unlock(&pr->pr_mtx); 3357 } 3358 break; 3359#endif 3360#ifdef INET6 3361 case AF_INET6: 3362 if (pr->pr_flags & PR_IP6) 3363 { 3364 mtx_lock(&pr->pr_mtx); 3365 if ((pr->pr_flags & PR_IP6) && pr->pr_ip6 == NULL) 3366 error = EAFNOSUPPORT; 3367 mtx_unlock(&pr->pr_mtx); 3368 } 3369 break; 3370#endif 3371 case AF_LOCAL: 3372 case AF_ROUTE: 3373 break; 3374 default: 3375 if (!(pr->pr_allow & PR_ALLOW_SOCKET_AF)) 3376 error = EAFNOSUPPORT; 3377 } 3378 return (error); 3379} 3380 3381/* 3382 * Check if given address belongs to the jail referenced by cred (wrapper to 3383 * prison_check_ip[46]). 3384 * 3385 * Returns 0 if jail doesn't restrict the address family or if address belongs 3386 * to jail, EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if 3387 * the jail doesn't allow the address family. IPv4 Address passed in in NBO. 3388 */ 3389int 3390prison_if(struct ucred *cred, struct sockaddr *sa) 3391{ 3392#ifdef INET 3393 struct sockaddr_in *sai; 3394#endif 3395#ifdef INET6 3396 struct sockaddr_in6 *sai6; 3397#endif 3398 int error; 3399 3400 KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); 3401 KASSERT(sa != NULL, ("%s: sa is NULL", __func__)); 3402 3403#ifdef VIMAGE 3404 if (prison_owns_vnet(cred)) 3405 return (0); 3406#endif 3407 3408 error = 0; 3409 switch (sa->sa_family) 3410 { 3411#ifdef INET 3412 case AF_INET: 3413 sai = (struct sockaddr_in *)sa; 3414 error = prison_check_ip4(cred, &sai->sin_addr); 3415 break; 3416#endif 3417#ifdef INET6 3418 case AF_INET6: 3419 sai6 = (struct sockaddr_in6 *)sa; 3420 error = prison_check_ip6(cred, &sai6->sin6_addr); 3421 break; 3422#endif 3423 default: 3424 if (!(cred->cr_prison->pr_allow & PR_ALLOW_SOCKET_AF)) 3425 error = EAFNOSUPPORT; 3426 } 3427 return (error); 3428} 3429 3430/* 3431 * Return 0 if jails permit p1 to frob p2, otherwise ESRCH. 3432 */ 3433int 3434prison_check(struct ucred *cred1, struct ucred *cred2) 3435{ 3436 3437 return ((cred1->cr_prison == cred2->cr_prison || 3438 prison_ischild(cred1->cr_prison, cred2->cr_prison)) ? 0 : ESRCH); 3439} 3440 3441/* 3442 * Return 1 if p2 is a child of p1, otherwise 0. 3443 */ 3444int 3445prison_ischild(struct prison *pr1, struct prison *pr2) 3446{ 3447 3448 for (pr2 = pr2->pr_parent; pr2 != NULL; pr2 = pr2->pr_parent) 3449 if (pr1 == pr2) 3450 return (1); 3451 return (0); 3452} 3453 3454/* 3455 * Return 1 if the passed credential is in a jail, otherwise 0. 3456 */ 3457int 3458jailed(struct ucred *cred) 3459{ 3460 3461 return (cred->cr_prison != &prison0); 3462} 3463 3464/* 3465 * Return 1 if the passed credential is in a jail and that jail does not 3466 * have its own virtual network stack, otherwise 0. 3467 */ 3468int 3469jailed_without_vnet(struct ucred *cred) 3470{ 3471 3472 if (!jailed(cred)) 3473 return (0); 3474#ifdef VIMAGE 3475 if (prison_owns_vnet(cred)) 3476 return (0); 3477#endif 3478 3479 return (1); 3480} 3481 3482/* 3483 * Return the correct hostname (domainname, et al) for the passed credential. 3484 */ 3485void 3486getcredhostname(struct ucred *cred, char *buf, size_t size) 3487{ 3488 struct prison *pr; 3489 3490 /* 3491 * A NULL credential can be used to shortcut to the physical 3492 * system's hostname. 3493 */ 3494 pr = (cred != NULL) ? cred->cr_prison : &prison0; 3495 mtx_lock(&pr->pr_mtx); 3496 strlcpy(buf, pr->pr_hostname, size); 3497 mtx_unlock(&pr->pr_mtx); 3498} 3499 3500void 3501getcreddomainname(struct ucred *cred, char *buf, size_t size) 3502{ 3503 3504 mtx_lock(&cred->cr_prison->pr_mtx); 3505 strlcpy(buf, cred->cr_prison->pr_domainname, size); 3506 mtx_unlock(&cred->cr_prison->pr_mtx); 3507} 3508 3509void 3510getcredhostuuid(struct ucred *cred, char *buf, size_t size) 3511{ 3512 3513 mtx_lock(&cred->cr_prison->pr_mtx); 3514 strlcpy(buf, cred->cr_prison->pr_hostuuid, size); 3515 mtx_unlock(&cred->cr_prison->pr_mtx); 3516} 3517 3518void 3519getcredhostid(struct ucred *cred, unsigned long *hostid) 3520{ 3521 3522 mtx_lock(&cred->cr_prison->pr_mtx); 3523 *hostid = cred->cr_prison->pr_hostid; 3524 mtx_unlock(&cred->cr_prison->pr_mtx); 3525} 3526 3527#ifdef VIMAGE 3528/* 3529 * Determine whether the prison represented by cred owns 3530 * its vnet rather than having it inherited. 3531 * 3532 * Returns 1 in case the prison owns the vnet, 0 otherwise. 3533 */ 3534int 3535prison_owns_vnet(struct ucred *cred) 3536{ 3537 3538 /* 3539 * vnets cannot be added/removed after jail creation, 3540 * so no need to lock here. 3541 */ 3542 return (cred->cr_prison->pr_flags & PR_VNET ? 1 : 0); 3543} 3544#endif 3545 3546/* 3547 * Determine whether the subject represented by cred can "see" 3548 * status of a mount point. 3549 * Returns: 0 for permitted, ENOENT otherwise. 3550 * XXX: This function should be called cr_canseemount() and should be 3551 * placed in kern_prot.c. 3552 */ 3553int 3554prison_canseemount(struct ucred *cred, struct mount *mp) 3555{ 3556 struct prison *pr; 3557 struct statfs *sp; 3558 size_t len; 3559 3560 pr = cred->cr_prison; 3561 if (pr->pr_enforce_statfs == 0) 3562 return (0); 3563 if (pr->pr_root->v_mount == mp) 3564 return (0); 3565 if (pr->pr_enforce_statfs == 2) 3566 return (ENOENT); 3567 /* 3568 * If jail's chroot directory is set to "/" we should be able to see 3569 * all mount-points from inside a jail. 3570 * This is ugly check, but this is the only situation when jail's 3571 * directory ends with '/'. 3572 */ 3573 if (strcmp(pr->pr_path, "/") == 0) 3574 return (0); 3575 len = strlen(pr->pr_path); 3576 sp = &mp->mnt_stat; 3577 if (strncmp(pr->pr_path, sp->f_mntonname, len) != 0) 3578 return (ENOENT); 3579 /* 3580 * Be sure that we don't have situation where jail's root directory 3581 * is "/some/path" and mount point is "/some/pathpath". 3582 */ 3583 if (sp->f_mntonname[len] != '\0' && sp->f_mntonname[len] != '/') 3584 return (ENOENT); 3585 return (0); 3586} 3587 3588void 3589prison_enforce_statfs(struct ucred *cred, struct mount *mp, struct statfs *sp) 3590{ 3591 char jpath[MAXPATHLEN]; 3592 struct prison *pr; 3593 size_t len; 3594 3595 pr = cred->cr_prison; 3596 if (pr->pr_enforce_statfs == 0) 3597 return; 3598 if (prison_canseemount(cred, mp) != 0) { 3599 bzero(sp->f_mntonname, sizeof(sp->f_mntonname)); 3600 strlcpy(sp->f_mntonname, "[restricted]", 3601 sizeof(sp->f_mntonname)); 3602 return; 3603 } 3604 if (pr->pr_root->v_mount == mp) { 3605 /* 3606 * Clear current buffer data, so we are sure nothing from 3607 * the valid path left there. 3608 */ 3609 bzero(sp->f_mntonname, sizeof(sp->f_mntonname)); 3610 *sp->f_mntonname = '/'; 3611 return; 3612 } 3613 /* 3614 * If jail's chroot directory is set to "/" we should be able to see 3615 * all mount-points from inside a jail. 3616 */ 3617 if (strcmp(pr->pr_path, "/") == 0) 3618 return; 3619 len = strlen(pr->pr_path); 3620 strlcpy(jpath, sp->f_mntonname + len, sizeof(jpath)); 3621 /* 3622 * Clear current buffer data, so we are sure nothing from 3623 * the valid path left there. 3624 */ 3625 bzero(sp->f_mntonname, sizeof(sp->f_mntonname)); 3626 if (*jpath == '\0') { 3627 /* Should never happen. */ 3628 *sp->f_mntonname = '/'; 3629 } else { 3630 strlcpy(sp->f_mntonname, jpath, sizeof(sp->f_mntonname)); 3631 } 3632} 3633 3634/* 3635 * Check with permission for a specific privilege is granted within jail. We 3636 * have a specific list of accepted privileges; the rest are denied. 3637 */ 3638int 3639prison_priv_check(struct ucred *cred, int priv) 3640{ 3641 3642 if (!jailed(cred)) 3643 return (0); 3644 3645#ifdef VIMAGE 3646 /* 3647 * Privileges specific to prisons with a virtual network stack. 3648 * There might be a duplicate entry here in case the privilege 3649 * is only granted conditionally in the legacy jail case. 3650 */ 3651 switch (priv) { 3652#ifdef notyet 3653 /* 3654 * NFS-specific privileges. 3655 */ 3656 case PRIV_NFS_DAEMON: 3657 case PRIV_NFS_LOCKD: 3658#endif 3659 /* 3660 * Network stack privileges. 3661 */ 3662 case PRIV_NET_BRIDGE: 3663 case PRIV_NET_GRE: 3664 case PRIV_NET_BPF: 3665 case PRIV_NET_RAW: /* Dup, cond. in legacy jail case. */ 3666 case PRIV_NET_ROUTE: 3667 case PRIV_NET_TAP: 3668 case PRIV_NET_SETIFMTU: 3669 case PRIV_NET_SETIFFLAGS: 3670 case PRIV_NET_SETIFCAP: 3671 case PRIV_NET_SETIFDESCR: 3672 case PRIV_NET_SETIFNAME : 3673 case PRIV_NET_SETIFMETRIC: 3674 case PRIV_NET_SETIFPHYS: 3675 case PRIV_NET_SETIFMAC: 3676 case PRIV_NET_ADDMULTI: 3677 case PRIV_NET_DELMULTI: 3678 case PRIV_NET_HWIOCTL: 3679 case PRIV_NET_SETLLADDR: 3680 case PRIV_NET_ADDIFGROUP: 3681 case PRIV_NET_DELIFGROUP: 3682 case PRIV_NET_IFCREATE: 3683 case PRIV_NET_IFDESTROY: 3684 case PRIV_NET_ADDIFADDR: 3685 case PRIV_NET_DELIFADDR: 3686 case PRIV_NET_LAGG: 3687 case PRIV_NET_GIF: 3688 case PRIV_NET_SETIFVNET: 3689 case PRIV_NET_SETIFFIB: 3690 3691 /* 3692 * 802.11-related privileges. 3693 */ 3694 case PRIV_NET80211_GETKEY: 3695#ifdef notyet 3696 case PRIV_NET80211_MANAGE: /* XXX-BZ discuss with sam@ */ 3697#endif 3698 3699#ifdef notyet 3700 /* 3701 * AppleTalk privileges. 3702 */ 3703 case PRIV_NETATALK_RESERVEDPORT: 3704 3705 /* 3706 * ATM privileges. 3707 */ 3708 case PRIV_NETATM_CFG: 3709 case PRIV_NETATM_ADD: 3710 case PRIV_NETATM_DEL: 3711 case PRIV_NETATM_SET: 3712 3713 /* 3714 * Bluetooth privileges. 3715 */ 3716 case PRIV_NETBLUETOOTH_RAW: 3717#endif 3718 3719 /* 3720 * Netgraph and netgraph module privileges. 3721 */ 3722 case PRIV_NETGRAPH_CONTROL: 3723#ifdef notyet 3724 case PRIV_NETGRAPH_TTY: 3725#endif 3726 3727 /* 3728 * IPv4 and IPv6 privileges. 3729 */ 3730 case PRIV_NETINET_IPFW: 3731 case PRIV_NETINET_DIVERT: 3732 case PRIV_NETINET_PF: 3733 case PRIV_NETINET_DUMMYNET: 3734 case PRIV_NETINET_CARP: 3735 case PRIV_NETINET_MROUTE: 3736 case PRIV_NETINET_RAW: 3737 case PRIV_NETINET_ADDRCTRL6: 3738 case PRIV_NETINET_ND6: 3739 case PRIV_NETINET_SCOPE6: 3740 case PRIV_NETINET_ALIFETIME6: 3741 case PRIV_NETINET_IPSEC: 3742 case PRIV_NETINET_BINDANY: 3743 3744#ifdef notyet 3745 /* 3746 * IPX/SPX privileges. 3747 */ 3748 case PRIV_NETIPX_RESERVEDPORT: 3749 case PRIV_NETIPX_RAW: 3750 3751 /* 3752 * NCP privileges. 3753 */ 3754 case PRIV_NETNCP: 3755 3756 /* 3757 * SMB privileges. 3758 */ 3759 case PRIV_NETSMB: 3760#endif 3761 3762 /* 3763 * No default: or deny here. 3764 * In case of no permit fall through to next switch(). 3765 */ 3766 if (cred->cr_prison->pr_flags & PR_VNET) 3767 return (0); 3768 } 3769#endif /* VIMAGE */ 3770 3771 switch (priv) { 3772 3773 /* 3774 * Allow ktrace privileges for root in jail. 3775 */ 3776 case PRIV_KTRACE: 3777 3778#if 0 3779 /* 3780 * Allow jailed processes to configure audit identity and 3781 * submit audit records (login, etc). In the future we may 3782 * want to further refine the relationship between audit and 3783 * jail. 3784 */ 3785 case PRIV_AUDIT_GETAUDIT: 3786 case PRIV_AUDIT_SETAUDIT: 3787 case PRIV_AUDIT_SUBMIT: 3788#endif 3789 3790 /* 3791 * Allow jailed processes to manipulate process UNIX 3792 * credentials in any way they see fit. 3793 */ 3794 case PRIV_CRED_SETUID: 3795 case PRIV_CRED_SETEUID: 3796 case PRIV_CRED_SETGID: 3797 case PRIV_CRED_SETEGID: 3798 case PRIV_CRED_SETGROUPS: 3799 case PRIV_CRED_SETREUID: 3800 case PRIV_CRED_SETREGID: 3801 case PRIV_CRED_SETRESUID: 3802 case PRIV_CRED_SETRESGID: 3803 3804 /* 3805 * Jail implements visibility constraints already, so allow 3806 * jailed root to override uid/gid-based constraints. 3807 */ 3808 case PRIV_SEEOTHERGIDS: 3809 case PRIV_SEEOTHERUIDS: 3810 3811 /* 3812 * Jail implements inter-process debugging limits already, so 3813 * allow jailed root various debugging privileges. 3814 */ 3815 case PRIV_DEBUG_DIFFCRED: 3816 case PRIV_DEBUG_SUGID: 3817 case PRIV_DEBUG_UNPRIV: 3818 3819 /* 3820 * Allow jail to set various resource limits and login 3821 * properties, and for now, exceed process resource limits. 3822 */ 3823 case PRIV_PROC_LIMIT: 3824 case PRIV_PROC_SETLOGIN: 3825 case PRIV_PROC_SETRLIMIT: 3826 3827 /* 3828 * System V and POSIX IPC privileges are granted in jail. 3829 */ 3830 case PRIV_IPC_READ: 3831 case PRIV_IPC_WRITE: 3832 case PRIV_IPC_ADMIN: 3833 case PRIV_IPC_MSGSIZE: 3834 case PRIV_MQ_ADMIN: 3835 3836 /* 3837 * Jail operations within a jail work on child jails. 3838 */ 3839 case PRIV_JAIL_ATTACH: 3840 case PRIV_JAIL_SET: 3841 case PRIV_JAIL_REMOVE: 3842 3843 /* 3844 * Jail implements its own inter-process limits, so allow 3845 * root processes in jail to change scheduling on other 3846 * processes in the same jail. Likewise for signalling. 3847 */ 3848 case PRIV_SCHED_DIFFCRED: 3849 case PRIV_SCHED_CPUSET: 3850 case PRIV_SIGNAL_DIFFCRED: 3851 case PRIV_SIGNAL_SUGID: 3852 3853 /* 3854 * Allow jailed processes to write to sysctls marked as jail 3855 * writable. 3856 */ 3857 case PRIV_SYSCTL_WRITEJAIL: 3858 3859 /* 3860 * Allow root in jail to manage a variety of quota 3861 * properties. These should likely be conditional on a 3862 * configuration option. 3863 */ 3864 case PRIV_VFS_GETQUOTA: 3865 case PRIV_VFS_SETQUOTA: 3866 3867 /* 3868 * Since Jail relies on chroot() to implement file system 3869 * protections, grant many VFS privileges to root in jail. 3870 * Be careful to exclude mount-related and NFS-related 3871 * privileges. 3872 */ 3873 case PRIV_VFS_READ: 3874 case PRIV_VFS_WRITE: 3875 case PRIV_VFS_ADMIN: 3876 case PRIV_VFS_EXEC: 3877 case PRIV_VFS_LOOKUP: 3878 case PRIV_VFS_BLOCKRESERVE: /* XXXRW: Slightly surprising. */ 3879 case PRIV_VFS_CHFLAGS_DEV: 3880 case PRIV_VFS_CHOWN: 3881 case PRIV_VFS_CHROOT: 3882 case PRIV_VFS_RETAINSUGID: 3883 case PRIV_VFS_FCHROOT: 3884 case PRIV_VFS_LINK: 3885 case PRIV_VFS_SETGID: 3886 case PRIV_VFS_STAT: 3887 case PRIV_VFS_STICKYFILE: 3888 3889 /* 3890 * As in the non-jail case, non-root users are expected to be 3891 * able to read kernel/phyiscal memory (provided /dev/[k]mem 3892 * exists in the jail and they have permission to access it). 3893 */ 3894 case PRIV_KMEM_READ: 3895 return (0); 3896 3897 /* 3898 * Depending on the global setting, allow privilege of 3899 * setting system flags. 3900 */ 3901 case PRIV_VFS_SYSFLAGS: 3902 if (cred->cr_prison->pr_allow & PR_ALLOW_CHFLAGS) 3903 return (0); 3904 else 3905 return (EPERM); 3906 3907 /* 3908 * Depending on the global setting, allow privilege of 3909 * mounting/unmounting file systems. 3910 */ 3911 case PRIV_VFS_MOUNT: 3912 case PRIV_VFS_UNMOUNT: 3913 case PRIV_VFS_MOUNT_NONUSER: 3914 case PRIV_VFS_MOUNT_OWNER: 3915 if (cred->cr_prison->pr_allow & PR_ALLOW_MOUNT && 3916 cred->cr_prison->pr_enforce_statfs < 2) 3917 return (0); 3918 else 3919 return (EPERM); 3920 3921 /* 3922 * Allow jailed root to bind reserved ports and reuse in-use 3923 * ports. 3924 */ 3925 case PRIV_NETINET_RESERVEDPORT: 3926 case PRIV_NETINET_REUSEPORT: 3927 return (0); 3928 3929 /* 3930 * Allow jailed root to set certian IPv4/6 (option) headers. 3931 */ 3932 case PRIV_NETINET_SETHDROPTS: 3933 return (0); 3934 3935 /* 3936 * Conditionally allow creating raw sockets in jail. 3937 */ 3938 case PRIV_NETINET_RAW: 3939 if (cred->cr_prison->pr_allow & PR_ALLOW_RAW_SOCKETS) 3940 return (0); 3941 else 3942 return (EPERM); 3943 3944 /* 3945 * Since jail implements its own visibility limits on netstat 3946 * sysctls, allow getcred. This allows identd to work in 3947 * jail. 3948 */ 3949 case PRIV_NETINET_GETCRED: 3950 return (0); 3951 3952 /* 3953 * Allow jailed root to set loginclass. 3954 */ 3955 case PRIV_PROC_SETLOGINCLASS: 3956 return (0); 3957 3958 default: 3959 /* 3960 * In all remaining cases, deny the privilege request. This 3961 * includes almost all network privileges, many system 3962 * configuration privileges. 3963 */ 3964 return (EPERM); 3965 } 3966} 3967 3968/* 3969 * Return the part of pr2's name that is relative to pr1, or the whole name 3970 * if it does not directly follow. 3971 */ 3972 3973char * 3974prison_name(struct prison *pr1, struct prison *pr2) 3975{ 3976 char *name; 3977 3978 /* Jails see themselves as "0" (if they see themselves at all). */ 3979 if (pr1 == pr2) 3980 return "0"; 3981 name = pr2->pr_name; 3982 if (prison_ischild(pr1, pr2)) { 3983 /* 3984 * pr1 isn't locked (and allprison_lock may not be either) 3985 * so its length can't be counted on. But the number of dots 3986 * can be counted on - and counted. 3987 */ 3988 for (; pr1 != &prison0; pr1 = pr1->pr_parent) 3989 name = strchr(name, '.') + 1; 3990 } 3991 return (name); 3992} 3993 3994/* 3995 * Return the part of pr2's path that is relative to pr1, or the whole path 3996 * if it does not directly follow. 3997 */ 3998static char * 3999prison_path(struct prison *pr1, struct prison *pr2) 4000{ 4001 char *path1, *path2; 4002 int len1; 4003 4004 path1 = pr1->pr_path; 4005 path2 = pr2->pr_path; 4006 if (!strcmp(path1, "/")) 4007 return (path2); 4008 len1 = strlen(path1); 4009 if (strncmp(path1, path2, len1)) 4010 return (path2); 4011 if (path2[len1] == '\0') 4012 return "/"; 4013 if (path2[len1] == '/') 4014 return (path2 + len1); 4015 return (path2); 4016} 4017 4018 4019/* 4020 * Jail-related sysctls. 4021 */ 4022static SYSCTL_NODE(_security, OID_AUTO, jail, CTLFLAG_RW, 0, 4023 "Jails"); 4024 4025static int 4026sysctl_jail_list(SYSCTL_HANDLER_ARGS) 4027{ 4028 struct xprison *xp; 4029 struct prison *pr, *cpr; 4030#ifdef INET 4031 struct in_addr *ip4 = NULL; 4032 int ip4s = 0; 4033#endif 4034#ifdef INET6 4035 struct in6_addr *ip6 = NULL; 4036 int ip6s = 0; 4037#endif 4038 int descend, error; 4039 4040 xp = malloc(sizeof(*xp), M_TEMP, M_WAITOK); 4041 pr = req->td->td_ucred->cr_prison; 4042 error = 0; 4043 sx_slock(&allprison_lock); 4044 FOREACH_PRISON_DESCENDANT(pr, cpr, descend) { 4045#if defined(INET) || defined(INET6) 4046 again: 4047#endif 4048 mtx_lock(&cpr->pr_mtx); 4049#ifdef INET 4050 if (cpr->pr_ip4s > 0) { 4051 if (ip4s < cpr->pr_ip4s) { 4052 ip4s = cpr->pr_ip4s; 4053 mtx_unlock(&cpr->pr_mtx); 4054 ip4 = realloc(ip4, ip4s * 4055 sizeof(struct in_addr), M_TEMP, M_WAITOK); 4056 goto again; 4057 } 4058 bcopy(cpr->pr_ip4, ip4, 4059 cpr->pr_ip4s * sizeof(struct in_addr)); 4060 } 4061#endif 4062#ifdef INET6 4063 if (cpr->pr_ip6s > 0) { 4064 if (ip6s < cpr->pr_ip6s) { 4065 ip6s = cpr->pr_ip6s; 4066 mtx_unlock(&cpr->pr_mtx); 4067 ip6 = realloc(ip6, ip6s * 4068 sizeof(struct in6_addr), M_TEMP, M_WAITOK); 4069 goto again; 4070 } 4071 bcopy(cpr->pr_ip6, ip6, 4072 cpr->pr_ip6s * sizeof(struct in6_addr)); 4073 } 4074#endif 4075 if (cpr->pr_ref == 0) { 4076 mtx_unlock(&cpr->pr_mtx); 4077 continue; 4078 } 4079 bzero(xp, sizeof(*xp)); 4080 xp->pr_version = XPRISON_VERSION; 4081 xp->pr_id = cpr->pr_id; 4082 xp->pr_state = cpr->pr_uref > 0 4083 ? PRISON_STATE_ALIVE : PRISON_STATE_DYING; 4084 strlcpy(xp->pr_path, prison_path(pr, cpr), sizeof(xp->pr_path)); 4085 strlcpy(xp->pr_host, cpr->pr_hostname, sizeof(xp->pr_host)); 4086 strlcpy(xp->pr_name, prison_name(pr, cpr), sizeof(xp->pr_name)); 4087#ifdef INET 4088 xp->pr_ip4s = cpr->pr_ip4s; 4089#endif 4090#ifdef INET6 4091 xp->pr_ip6s = cpr->pr_ip6s; 4092#endif 4093 mtx_unlock(&cpr->pr_mtx); 4094 error = SYSCTL_OUT(req, xp, sizeof(*xp)); 4095 if (error) 4096 break; 4097#ifdef INET 4098 if (xp->pr_ip4s > 0) { 4099 error = SYSCTL_OUT(req, ip4, 4100 xp->pr_ip4s * sizeof(struct in_addr)); 4101 if (error) 4102 break; 4103 } 4104#endif 4105#ifdef INET6 4106 if (xp->pr_ip6s > 0) { 4107 error = SYSCTL_OUT(req, ip6, 4108 xp->pr_ip6s * sizeof(struct in6_addr)); 4109 if (error) 4110 break; 4111 } 4112#endif 4113 } 4114 sx_sunlock(&allprison_lock); 4115 free(xp, M_TEMP); 4116#ifdef INET 4117 free(ip4, M_TEMP); 4118#endif 4119#ifdef INET6 4120 free(ip6, M_TEMP); 4121#endif 4122 return (error); 4123} 4124 4125SYSCTL_OID(_security_jail, OID_AUTO, list, 4126 CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0, 4127 sysctl_jail_list, "S", "List of active jails"); 4128 4129static int 4130sysctl_jail_jailed(SYSCTL_HANDLER_ARGS) 4131{ 4132 int error, injail; 4133 4134 injail = jailed(req->td->td_ucred); 4135 error = SYSCTL_OUT(req, &injail, sizeof(injail)); 4136 4137 return (error); 4138} 4139 4140SYSCTL_PROC(_security_jail, OID_AUTO, jailed, 4141 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0, 4142 sysctl_jail_jailed, "I", "Process in jail?"); 4143 4144static int 4145sysctl_jail_vnet(SYSCTL_HANDLER_ARGS) 4146{ 4147 int error, havevnet; 4148#ifdef VIMAGE 4149 struct ucred *cred = req->td->td_ucred; 4150 4151 havevnet = jailed(cred) && prison_owns_vnet(cred); 4152#else 4153 havevnet = 0; 4154#endif 4155 error = SYSCTL_OUT(req, &havevnet, sizeof(havevnet)); 4156 4157 return (error); 4158} 4159 4160SYSCTL_PROC(_security_jail, OID_AUTO, vnet, 4161 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0, 4162 sysctl_jail_vnet, "I", "Jail owns VNET?"); 4163 4164#if defined(INET) || defined(INET6) 4165SYSCTL_UINT(_security_jail, OID_AUTO, jail_max_af_ips, CTLFLAG_RW, 4166 &jail_max_af_ips, 0, 4167 "Number of IP addresses a jail may have at most per address family"); 4168#endif 4169 4170/* 4171 * Default parameters for jail(2) compatability. For historical reasons, 4172 * the sysctl names have varying similarity to the parameter names. Prisons 4173 * just see their own parameters, and can't change them. 4174 */ 4175static int 4176sysctl_jail_default_allow(SYSCTL_HANDLER_ARGS) 4177{ 4178 struct prison *pr; 4179 int allow, error, i; 4180 4181 pr = req->td->td_ucred->cr_prison; 4182 allow = (pr == &prison0) ? jail_default_allow : pr->pr_allow; 4183 4184 /* Get the current flag value, and convert it to a boolean. */ 4185 i = (allow & arg2) ? 1 : 0; 4186 if (arg1 != NULL) 4187 i = !i; 4188 error = sysctl_handle_int(oidp, &i, 0, req); 4189 if (error || !req->newptr) 4190 return (error); 4191 i = i ? arg2 : 0; 4192 if (arg1 != NULL) 4193 i ^= arg2; 4194 /* 4195 * The sysctls don't have CTLFLAGS_PRISON, so assume prison0 4196 * for writing. 4197 */ 4198 mtx_lock(&prison0.pr_mtx); 4199 jail_default_allow = (jail_default_allow & ~arg2) | i; 4200 mtx_unlock(&prison0.pr_mtx); 4201 return (0); 4202} 4203 4204SYSCTL_PROC(_security_jail, OID_AUTO, set_hostname_allowed, 4205 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 4206 NULL, PR_ALLOW_SET_HOSTNAME, sysctl_jail_default_allow, "I", 4207 "Processes in jail can set their hostnames"); 4208SYSCTL_PROC(_security_jail, OID_AUTO, socket_unixiproute_only, 4209 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 4210 (void *)1, PR_ALLOW_SOCKET_AF, sysctl_jail_default_allow, "I", 4211 "Processes in jail are limited to creating UNIX/IP/route sockets only"); 4212SYSCTL_PROC(_security_jail, OID_AUTO, sysvipc_allowed, 4213 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 4214 NULL, PR_ALLOW_SYSVIPC, sysctl_jail_default_allow, "I", 4215 "Processes in jail can use System V IPC primitives"); 4216SYSCTL_PROC(_security_jail, OID_AUTO, allow_raw_sockets, 4217 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 4218 NULL, PR_ALLOW_RAW_SOCKETS, sysctl_jail_default_allow, "I", 4219 "Prison root can create raw sockets"); 4220SYSCTL_PROC(_security_jail, OID_AUTO, chflags_allowed, 4221 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 4222 NULL, PR_ALLOW_CHFLAGS, sysctl_jail_default_allow, "I", 4223 "Processes in jail can alter system file flags"); 4224SYSCTL_PROC(_security_jail, OID_AUTO, mount_allowed, 4225 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 4226 NULL, PR_ALLOW_MOUNT, sysctl_jail_default_allow, "I", 4227 "Processes in jail can mount/unmount jail-friendly file systems"); 4228SYSCTL_PROC(_security_jail, OID_AUTO, mount_devfs_allowed, 4229 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 4230 NULL, PR_ALLOW_MOUNT_DEVFS, sysctl_jail_default_allow, "I", 4231 "Processes in jail can mount the devfs file system"); 4232SYSCTL_PROC(_security_jail, OID_AUTO, mount_nullfs_allowed, 4233 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 4234 NULL, PR_ALLOW_MOUNT_NULLFS, sysctl_jail_default_allow, "I", 4235 "Processes in jail can mount the nullfs file system"); 4236SYSCTL_PROC(_security_jail, OID_AUTO, mount_procfs_allowed, 4237 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 4238 NULL, PR_ALLOW_MOUNT_PROCFS, sysctl_jail_default_allow, "I", 4239 "Processes in jail can mount the procfs file system"); 4240SYSCTL_PROC(_security_jail, OID_AUTO, mount_tmpfs_allowed, 4241 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 4242 NULL, PR_ALLOW_MOUNT_TMPFS, sysctl_jail_default_allow, "I", 4243 "Processes in jail can mount the tmpfs file system"); 4244SYSCTL_PROC(_security_jail, OID_AUTO, mount_zfs_allowed, 4245 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 4246 NULL, PR_ALLOW_MOUNT_ZFS, sysctl_jail_default_allow, "I", 4247 "Processes in jail can mount the zfs file system"); 4248 4249static int 4250sysctl_jail_default_level(SYSCTL_HANDLER_ARGS) 4251{ 4252 struct prison *pr; 4253 int level, error; 4254 4255 pr = req->td->td_ucred->cr_prison; 4256 level = (pr == &prison0) ? *(int *)arg1 : *(int *)((char *)pr + arg2); 4257 error = sysctl_handle_int(oidp, &level, 0, req); 4258 if (error || !req->newptr) 4259 return (error); 4260 *(int *)arg1 = level; 4261 return (0); 4262} 4263 4264SYSCTL_PROC(_security_jail, OID_AUTO, enforce_statfs, 4265 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 4266 &jail_default_enforce_statfs, offsetof(struct prison, pr_enforce_statfs), 4267 sysctl_jail_default_level, "I", 4268 "Processes in jail cannot see all mounted file systems"); 4269 4270SYSCTL_PROC(_security_jail, OID_AUTO, devfs_ruleset, 4271 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, 4272 &jail_default_devfs_rsnum, offsetof(struct prison, pr_devfs_rsnum), 4273 sysctl_jail_default_level, "I", 4274 "Ruleset for the devfs filesystem in jail"); 4275 4276/* 4277 * Nodes to describe jail parameters. Maximum length of string parameters 4278 * is returned in the string itself, and the other parameters exist merely 4279 * to make themselves and their types known. 4280 */ 4281SYSCTL_NODE(_security_jail, OID_AUTO, param, CTLFLAG_RW, 0, 4282 "Jail parameters"); 4283 4284int 4285sysctl_jail_param(SYSCTL_HANDLER_ARGS) 4286{ 4287 int i; 4288 long l; 4289 size_t s; 4290 char numbuf[12]; 4291 4292 switch (oidp->oid_kind & CTLTYPE) 4293 { 4294 case CTLTYPE_LONG: 4295 case CTLTYPE_ULONG: 4296 l = 0; 4297#ifdef SCTL_MASK32 4298 if (!(req->flags & SCTL_MASK32)) 4299#endif 4300 return (SYSCTL_OUT(req, &l, sizeof(l))); 4301 case CTLTYPE_INT: 4302 case CTLTYPE_UINT: 4303 i = 0; 4304 return (SYSCTL_OUT(req, &i, sizeof(i))); 4305 case CTLTYPE_STRING: 4306 snprintf(numbuf, sizeof(numbuf), "%jd", (intmax_t)arg2); 4307 return 4308 (sysctl_handle_string(oidp, numbuf, sizeof(numbuf), req)); 4309 case CTLTYPE_STRUCT: 4310 s = (size_t)arg2; 4311 return (SYSCTL_OUT(req, &s, sizeof(s))); 4312 } 4313 return (0); 4314} 4315 4316SYSCTL_JAIL_PARAM(, jid, CTLTYPE_INT | CTLFLAG_RDTUN, "I", "Jail ID"); 4317SYSCTL_JAIL_PARAM(, parent, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail parent ID"); 4318SYSCTL_JAIL_PARAM_STRING(, name, CTLFLAG_RW, MAXHOSTNAMELEN, "Jail name"); 4319SYSCTL_JAIL_PARAM_STRING(, path, CTLFLAG_RDTUN, MAXPATHLEN, "Jail root path"); 4320SYSCTL_JAIL_PARAM(, securelevel, CTLTYPE_INT | CTLFLAG_RW, 4321 "I", "Jail secure level"); 4322SYSCTL_JAIL_PARAM(, enforce_statfs, CTLTYPE_INT | CTLFLAG_RW, 4323 "I", "Jail cannot see all mounted file systems"); 4324SYSCTL_JAIL_PARAM(, devfs_ruleset, CTLTYPE_INT | CTLFLAG_RW, 4325 "I", "Ruleset for in-jail devfs mounts"); 4326SYSCTL_JAIL_PARAM(, persist, CTLTYPE_INT | CTLFLAG_RW, 4327 "B", "Jail persistence"); 4328#ifdef VIMAGE 4329SYSCTL_JAIL_PARAM(, vnet, CTLTYPE_INT | CTLFLAG_RDTUN, 4330 "E,jailsys", "Virtual network stack"); 4331#endif 4332SYSCTL_JAIL_PARAM(, dying, CTLTYPE_INT | CTLFLAG_RD, 4333 "B", "Jail is in the process of shutting down"); 4334 4335SYSCTL_JAIL_PARAM_NODE(children, "Number of child jails"); 4336SYSCTL_JAIL_PARAM(_children, cur, CTLTYPE_INT | CTLFLAG_RD, 4337 "I", "Current number of child jails"); 4338SYSCTL_JAIL_PARAM(_children, max, CTLTYPE_INT | CTLFLAG_RW, 4339 "I", "Maximum number of child jails"); 4340 4341SYSCTL_JAIL_PARAM_SYS_NODE(host, CTLFLAG_RW, "Jail host info"); 4342SYSCTL_JAIL_PARAM_STRING(_host, hostname, CTLFLAG_RW, MAXHOSTNAMELEN, 4343 "Jail hostname"); 4344SYSCTL_JAIL_PARAM_STRING(_host, domainname, CTLFLAG_RW, MAXHOSTNAMELEN, 4345 "Jail NIS domainname"); 4346SYSCTL_JAIL_PARAM_STRING(_host, hostuuid, CTLFLAG_RW, HOSTUUIDLEN, 4347 "Jail host UUID"); 4348SYSCTL_JAIL_PARAM(_host, hostid, CTLTYPE_ULONG | CTLFLAG_RW, 4349 "LU", "Jail host ID"); 4350 4351SYSCTL_JAIL_PARAM_NODE(cpuset, "Jail cpuset"); 4352SYSCTL_JAIL_PARAM(_cpuset, id, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail cpuset ID"); 4353 4354#ifdef INET 4355SYSCTL_JAIL_PARAM_SYS_NODE(ip4, CTLFLAG_RDTUN, 4356 "Jail IPv4 address virtualization"); 4357SYSCTL_JAIL_PARAM_STRUCT(_ip4, addr, CTLFLAG_RW, sizeof(struct in_addr), 4358 "S,in_addr,a", "Jail IPv4 addresses"); 4359SYSCTL_JAIL_PARAM(_ip4, saddrsel, CTLTYPE_INT | CTLFLAG_RW, 4360 "B", "Do (not) use IPv4 source address selection rather than the " 4361 "primary jail IPv4 address."); 4362#endif 4363#ifdef INET6 4364SYSCTL_JAIL_PARAM_SYS_NODE(ip6, CTLFLAG_RDTUN, 4365 "Jail IPv6 address virtualization"); 4366SYSCTL_JAIL_PARAM_STRUCT(_ip6, addr, CTLFLAG_RW, sizeof(struct in6_addr), 4367 "S,in6_addr,a", "Jail IPv6 addresses"); 4368SYSCTL_JAIL_PARAM(_ip6, saddrsel, CTLTYPE_INT | CTLFLAG_RW, 4369 "B", "Do (not) use IPv6 source address selection rather than the " 4370 "primary jail IPv6 address."); 4371#endif 4372 4373SYSCTL_JAIL_PARAM_NODE(allow, "Jail permission flags"); 4374SYSCTL_JAIL_PARAM(_allow, set_hostname, CTLTYPE_INT | CTLFLAG_RW, 4375 "B", "Jail may set hostname"); 4376SYSCTL_JAIL_PARAM(_allow, sysvipc, CTLTYPE_INT | CTLFLAG_RW, 4377 "B", "Jail may use SYSV IPC"); 4378SYSCTL_JAIL_PARAM(_allow, raw_sockets, CTLTYPE_INT | CTLFLAG_RW, 4379 "B", "Jail may create raw sockets"); 4380SYSCTL_JAIL_PARAM(_allow, chflags, CTLTYPE_INT | CTLFLAG_RW, 4381 "B", "Jail may alter system file flags"); 4382SYSCTL_JAIL_PARAM(_allow, quotas, CTLTYPE_INT | CTLFLAG_RW, 4383 "B", "Jail may set file quotas"); 4384SYSCTL_JAIL_PARAM(_allow, socket_af, CTLTYPE_INT | CTLFLAG_RW, 4385 "B", "Jail may create sockets other than just UNIX/IPv4/IPv6/route"); 4386 4387SYSCTL_JAIL_PARAM_SUBNODE(allow, mount, "Jail mount/unmount permission flags"); 4388SYSCTL_JAIL_PARAM(_allow_mount, , CTLTYPE_INT | CTLFLAG_RW, 4389 "B", "Jail may mount/unmount jail-friendly file systems in general"); 4390SYSCTL_JAIL_PARAM(_allow_mount, devfs, CTLTYPE_INT | CTLFLAG_RW, 4391 "B", "Jail may mount the devfs file system"); 4392SYSCTL_JAIL_PARAM(_allow_mount, nullfs, CTLTYPE_INT | CTLFLAG_RW, 4393 "B", "Jail may mount the nullfs file system"); 4394SYSCTL_JAIL_PARAM(_allow_mount, procfs, CTLTYPE_INT | CTLFLAG_RW, 4395 "B", "Jail may mount the procfs file system"); 4396SYSCTL_JAIL_PARAM(_allow_mount, tmpfs, CTLTYPE_INT | CTLFLAG_RW, 4397 "B", "Jail may mount the tmpfs file system"); 4398SYSCTL_JAIL_PARAM(_allow_mount, zfs, CTLTYPE_INT | CTLFLAG_RW, 4399 "B", "Jail may mount the zfs file system"); 4400 4401void 4402prison_racct_foreach(void (*callback)(struct racct *racct, 4403 void *arg2, void *arg3), void *arg2, void *arg3) 4404{ 4405 struct prison_racct *prr; 4406 4407 sx_slock(&allprison_lock); 4408 LIST_FOREACH(prr, &allprison_racct, prr_next) 4409 (callback)(prr->prr_racct, arg2, arg3); 4410 sx_sunlock(&allprison_lock); 4411} 4412 4413static struct prison_racct * 4414prison_racct_find_locked(const char *name) 4415{ 4416 struct prison_racct *prr; 4417 4418 sx_assert(&allprison_lock, SA_XLOCKED); 4419 4420 if (name[0] == '\0' || strlen(name) >= MAXHOSTNAMELEN) 4421 return (NULL); 4422 4423 LIST_FOREACH(prr, &allprison_racct, prr_next) { 4424 if (strcmp(name, prr->prr_name) != 0) 4425 continue; 4426 4427 /* Found prison_racct with a matching name? */ 4428 prison_racct_hold(prr); 4429 return (prr); 4430 } 4431 4432 /* Add new prison_racct. */ 4433 prr = malloc(sizeof(*prr), M_PRISON_RACCT, M_ZERO | M_WAITOK); 4434 racct_create(&prr->prr_racct); 4435 4436 strcpy(prr->prr_name, name); 4437 refcount_init(&prr->prr_refcount, 1); 4438 LIST_INSERT_HEAD(&allprison_racct, prr, prr_next); 4439 4440 return (prr); 4441} 4442 4443struct prison_racct * 4444prison_racct_find(const char *name) 4445{ 4446 struct prison_racct *prr; 4447 4448 sx_xlock(&allprison_lock); 4449 prr = prison_racct_find_locked(name); 4450 sx_xunlock(&allprison_lock); 4451 return (prr); 4452} 4453 4454void 4455prison_racct_hold(struct prison_racct *prr) 4456{ 4457 4458 refcount_acquire(&prr->prr_refcount); 4459} 4460 4461static void 4462prison_racct_free_locked(struct prison_racct *prr) 4463{ 4464 4465 sx_assert(&allprison_lock, SA_XLOCKED); 4466 4467 if (refcount_release(&prr->prr_refcount)) { 4468 racct_destroy(&prr->prr_racct); 4469 LIST_REMOVE(prr, prr_next); 4470 free(prr, M_PRISON_RACCT); 4471 } 4472} 4473 4474void 4475prison_racct_free(struct prison_racct *prr) 4476{ 4477 int old; 4478 4479 sx_assert(&allprison_lock, SA_UNLOCKED); 4480 4481 old = prr->prr_refcount; 4482 if (old > 1 && atomic_cmpset_int(&prr->prr_refcount, old, old - 1)) 4483 return; 4484 4485 sx_xlock(&allprison_lock); 4486 prison_racct_free_locked(prr); 4487 sx_xunlock(&allprison_lock); 4488} 4489 4490#ifdef RACCT 4491static void 4492prison_racct_attach(struct prison *pr) 4493{ 4494 struct prison_racct *prr; 4495 4496 sx_assert(&allprison_lock, SA_XLOCKED); 4497 4498 prr = prison_racct_find_locked(pr->pr_name); 4499 KASSERT(prr != NULL, ("cannot find prison_racct")); 4500 4501 pr->pr_prison_racct = prr; 4502} 4503 4504/* 4505 * Handle jail renaming. From the racct point of view, renaming means 4506 * moving from one prison_racct to another. 4507 */ 4508static void 4509prison_racct_modify(struct prison *pr) 4510{ 4511 struct proc *p; 4512 struct ucred *cred; 4513 struct prison_racct *oldprr; 4514 4515 sx_slock(&allproc_lock); 4516 sx_xlock(&allprison_lock); 4517 4518 if (strcmp(pr->pr_name, pr->pr_prison_racct->prr_name) == 0) { 4519 sx_xunlock(&allprison_lock); 4520 sx_sunlock(&allproc_lock); 4521 return; 4522 } 4523 4524 oldprr = pr->pr_prison_racct; 4525 pr->pr_prison_racct = NULL; 4526 4527 prison_racct_attach(pr); 4528 4529 /* 4530 * Move resource utilisation records. 4531 */ 4532 racct_move(pr->pr_prison_racct->prr_racct, oldprr->prr_racct); 4533 4534 /* 4535 * Force rctl to reattach rules to processes. 4536 */ 4537 FOREACH_PROC_IN_SYSTEM(p) { 4538 PROC_LOCK(p); 4539 cred = crhold(p->p_ucred); 4540 PROC_UNLOCK(p); 4541 racct_proc_ucred_changed(p, cred, cred); 4542 crfree(cred); 4543 } 4544 4545 sx_sunlock(&allproc_lock); 4546 prison_racct_free_locked(oldprr); 4547 sx_xunlock(&allprison_lock); 4548} 4549 4550static void 4551prison_racct_detach(struct prison *pr) 4552{ 4553 4554 sx_assert(&allprison_lock, SA_UNLOCKED); 4555 4556 if (pr->pr_prison_racct == NULL) 4557 return; 4558 prison_racct_free(pr->pr_prison_racct); 4559 pr->pr_prison_racct = NULL; 4560} 4561#endif /* RACCT */ 4562 4563#ifdef DDB 4564 4565static void 4566db_show_prison(struct prison *pr) 4567{ 4568 int fi; 4569#if defined(INET) || defined(INET6) 4570 int ii; 4571#endif 4572 unsigned jsf; 4573#ifdef INET6 4574 char ip6buf[INET6_ADDRSTRLEN]; 4575#endif 4576 4577 db_printf("prison %p:\n", pr); 4578 db_printf(" jid = %d\n", pr->pr_id); 4579 db_printf(" name = %s\n", pr->pr_name); 4580 db_printf(" parent = %p\n", pr->pr_parent); 4581 db_printf(" ref = %d\n", pr->pr_ref); 4582 db_printf(" uref = %d\n", pr->pr_uref); 4583 db_printf(" path = %s\n", pr->pr_path); 4584 db_printf(" cpuset = %d\n", pr->pr_cpuset 4585 ? pr->pr_cpuset->cs_id : -1); 4586#ifdef VIMAGE 4587 db_printf(" vnet = %p\n", pr->pr_vnet); 4588#endif 4589 db_printf(" root = %p\n", pr->pr_root); 4590 db_printf(" securelevel = %d\n", pr->pr_securelevel); 4591 db_printf(" devfs_rsnum = %d\n", pr->pr_devfs_rsnum); 4592 db_printf(" children.max = %d\n", pr->pr_childmax); 4593 db_printf(" children.cur = %d\n", pr->pr_childcount); 4594 db_printf(" child = %p\n", LIST_FIRST(&pr->pr_children)); 4595 db_printf(" sibling = %p\n", LIST_NEXT(pr, pr_sibling)); 4596 db_printf(" flags = 0x%x", pr->pr_flags); 4597 for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]); 4598 fi++) 4599 if (pr_flag_names[fi] != NULL && (pr->pr_flags & (1 << fi))) 4600 db_printf(" %s", pr_flag_names[fi]); 4601 for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]); 4602 fi++) { 4603 jsf = pr->pr_flags & 4604 (pr_flag_jailsys[fi].disable | pr_flag_jailsys[fi].new); 4605 db_printf(" %-16s= %s\n", pr_flag_jailsys[fi].name, 4606 pr_flag_jailsys[fi].disable && 4607 (jsf == pr_flag_jailsys[fi].disable) ? "disable" 4608 : (jsf == pr_flag_jailsys[fi].new) ? "new" 4609 : "inherit"); 4610 } 4611 db_printf(" allow = 0x%x", pr->pr_allow); 4612 for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]); 4613 fi++) 4614 if (pr_allow_names[fi] != NULL && (pr->pr_allow & (1 << fi))) 4615 db_printf(" %s", pr_allow_names[fi]); 4616 db_printf("\n"); 4617 db_printf(" enforce_statfs = %d\n", pr->pr_enforce_statfs); 4618 db_printf(" host.hostname = %s\n", pr->pr_hostname); 4619 db_printf(" host.domainname = %s\n", pr->pr_domainname); 4620 db_printf(" host.hostuuid = %s\n", pr->pr_hostuuid); 4621 db_printf(" host.hostid = %lu\n", pr->pr_hostid); 4622#ifdef INET 4623 db_printf(" ip4s = %d\n", pr->pr_ip4s); 4624 for (ii = 0; ii < pr->pr_ip4s; ii++) 4625 db_printf(" %s %s\n", 4626 ii == 0 ? "ip4.addr =" : " ", 4627 inet_ntoa(pr->pr_ip4[ii])); 4628#endif 4629#ifdef INET6 4630 db_printf(" ip6s = %d\n", pr->pr_ip6s); 4631 for (ii = 0; ii < pr->pr_ip6s; ii++) 4632 db_printf(" %s %s\n", 4633 ii == 0 ? "ip6.addr =" : " ", 4634 ip6_sprintf(ip6buf, &pr->pr_ip6[ii])); 4635#endif 4636} 4637 4638DB_SHOW_COMMAND(prison, db_show_prison_command) 4639{ 4640 struct prison *pr; 4641 4642 if (!have_addr) { 4643 /* 4644 * Show all prisons in the list, and prison0 which is not 4645 * listed. 4646 */ 4647 db_show_prison(&prison0); 4648 if (!db_pager_quit) { 4649 TAILQ_FOREACH(pr, &allprison, pr_list) { 4650 db_show_prison(pr); 4651 if (db_pager_quit) 4652 break; 4653 } 4654 } 4655 return; 4656 } 4657 4658 if (addr == 0) 4659 pr = &prison0; 4660 else { 4661 /* Look for a prison with the ID and with references. */ 4662 TAILQ_FOREACH(pr, &allprison, pr_list) 4663 if (pr->pr_id == addr && pr->pr_ref > 0) 4664 break; 4665 if (pr == NULL) 4666 /* Look again, without requiring a reference. */ 4667 TAILQ_FOREACH(pr, &allprison, pr_list) 4668 if (pr->pr_id == addr) 4669 break; 4670 if (pr == NULL) 4671 /* Assume address points to a valid prison. */ 4672 pr = (struct prison *)addr; 4673 } 4674 db_show_prison(pr); 4675} 4676 4677#endif /* DDB */ 4678