kern_jail.c revision 298832
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 298832 2016-04-30 03:05:36Z jamie $"); 31 32#include "opt_compat.h" 33#include "opt_ddb.h" 34#include "opt_inet.h" 35#include "opt_inet6.h" 36 37#include <sys/param.h> 38#include <sys/types.h> 39#include <sys/kernel.h> 40#include <sys/systm.h> 41#include <sys/errno.h> 42#include <sys/sysproto.h> 43#include <sys/malloc.h> 44#include <sys/osd.h> 45#include <sys/priv.h> 46#include <sys/proc.h> 47#include <sys/taskqueue.h> 48#include <sys/fcntl.h> 49#include <sys/jail.h> 50#include <sys/lock.h> 51#include <sys/mutex.h> 52#include <sys/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 "allow.mount.fdescfs", 211 "allow.mount.linprocfs", 212 "allow.mount.linsysfs", 213}; 214const size_t pr_allow_names_size = sizeof(pr_allow_names); 215 216static char *pr_allow_nonames[] = { 217 "allow.noset_hostname", 218 "allow.nosysvipc", 219 "allow.noraw_sockets", 220 "allow.nochflags", 221 "allow.nomount", 222 "allow.noquotas", 223 "allow.nosocket_af", 224 "allow.mount.nodevfs", 225 "allow.mount.nonullfs", 226 "allow.mount.nozfs", 227 "allow.mount.noprocfs", 228 "allow.mount.notmpfs", 229 "allow.mount.nofdescfs", 230 "allow.mount.nolinprocfs", 231 "allow.mount.nolinsysfs", 232}; 233const size_t pr_allow_nonames_size = sizeof(pr_allow_nonames); 234 235#define JAIL_DEFAULT_ALLOW PR_ALLOW_SET_HOSTNAME 236#define JAIL_DEFAULT_ENFORCE_STATFS 2 237#define JAIL_DEFAULT_DEVFS_RSNUM 0 238static unsigned jail_default_allow = JAIL_DEFAULT_ALLOW; 239static int jail_default_enforce_statfs = JAIL_DEFAULT_ENFORCE_STATFS; 240static int jail_default_devfs_rsnum = JAIL_DEFAULT_DEVFS_RSNUM; 241#if defined(INET) || defined(INET6) 242static unsigned jail_max_af_ips = 255; 243#endif 244 245/* 246 * Initialize the parts of prison0 that can't be static-initialized with 247 * constants. This is called from proc0_init() after creating thread0 cpuset. 248 */ 249void 250prison0_init(void) 251{ 252 253 prison0.pr_cpuset = cpuset_ref(thread0.td_cpuset); 254 prison0.pr_osreldate = osreldate; 255 strlcpy(prison0.pr_osrelease, osrelease, sizeof(prison0.pr_osrelease)); 256} 257 258#ifdef INET 259static int 260qcmp_v4(const void *ip1, const void *ip2) 261{ 262 in_addr_t iaa, iab; 263 264 /* 265 * We need to compare in HBO here to get the list sorted as expected 266 * by the result of the code. Sorting NBO addresses gives you 267 * interesting results. If you do not understand, do not try. 268 */ 269 iaa = ntohl(((const struct in_addr *)ip1)->s_addr); 270 iab = ntohl(((const struct in_addr *)ip2)->s_addr); 271 272 /* 273 * Do not simply return the difference of the two numbers, the int is 274 * not wide enough. 275 */ 276 if (iaa > iab) 277 return (1); 278 else if (iaa < iab) 279 return (-1); 280 else 281 return (0); 282} 283#endif 284 285#ifdef INET6 286static int 287qcmp_v6(const void *ip1, const void *ip2) 288{ 289 const struct in6_addr *ia6a, *ia6b; 290 int i, rc; 291 292 ia6a = (const struct in6_addr *)ip1; 293 ia6b = (const struct in6_addr *)ip2; 294 295 rc = 0; 296 for (i = 0; rc == 0 && i < sizeof(struct in6_addr); i++) { 297 if (ia6a->s6_addr[i] > ia6b->s6_addr[i]) 298 rc = 1; 299 else if (ia6a->s6_addr[i] < ia6b->s6_addr[i]) 300 rc = -1; 301 } 302 return (rc); 303} 304#endif 305 306/* 307 * struct jail_args { 308 * struct jail *jail; 309 * }; 310 */ 311int 312sys_jail(struct thread *td, struct jail_args *uap) 313{ 314 uint32_t version; 315 int error; 316 struct jail j; 317 318 error = copyin(uap->jail, &version, sizeof(uint32_t)); 319 if (error) 320 return (error); 321 322 switch (version) { 323 case 0: 324 { 325 struct jail_v0 j0; 326 327 /* FreeBSD single IPv4 jails. */ 328 bzero(&j, sizeof(struct jail)); 329 error = copyin(uap->jail, &j0, sizeof(struct jail_v0)); 330 if (error) 331 return (error); 332 j.version = j0.version; 333 j.path = j0.path; 334 j.hostname = j0.hostname; 335 j.ip4s = htonl(j0.ip_number); /* jail_v0 is host order */ 336 break; 337 } 338 339 case 1: 340 /* 341 * Version 1 was used by multi-IPv4 jail implementations 342 * that never made it into the official kernel. 343 */ 344 return (EINVAL); 345 346 case 2: /* JAIL_API_VERSION */ 347 /* FreeBSD multi-IPv4/IPv6,noIP jails. */ 348 error = copyin(uap->jail, &j, sizeof(struct jail)); 349 if (error) 350 return (error); 351 break; 352 353 default: 354 /* Sci-Fi jails are not supported, sorry. */ 355 return (EINVAL); 356 } 357 return (kern_jail(td, &j)); 358} 359 360int 361kern_jail(struct thread *td, struct jail *j) 362{ 363 struct iovec optiov[2 * (4 364 + sizeof(pr_allow_names) / sizeof(pr_allow_names[0]) 365#ifdef INET 366 + 1 367#endif 368#ifdef INET6 369 + 1 370#endif 371 )]; 372 struct uio opt; 373 char *u_path, *u_hostname, *u_name; 374#ifdef INET 375 uint32_t ip4s; 376 struct in_addr *u_ip4; 377#endif 378#ifdef INET6 379 struct in6_addr *u_ip6; 380#endif 381 size_t tmplen; 382 int error, enforce_statfs, fi; 383 384 bzero(&optiov, sizeof(optiov)); 385 opt.uio_iov = optiov; 386 opt.uio_iovcnt = 0; 387 opt.uio_offset = -1; 388 opt.uio_resid = -1; 389 opt.uio_segflg = UIO_SYSSPACE; 390 opt.uio_rw = UIO_READ; 391 opt.uio_td = td; 392 393 /* Set permissions for top-level jails from sysctls. */ 394 if (!jailed(td->td_ucred)) { 395 for (fi = 0; fi < sizeof(pr_allow_names) / 396 sizeof(pr_allow_names[0]); fi++) { 397 optiov[opt.uio_iovcnt].iov_base = 398 (jail_default_allow & (1 << fi)) 399 ? pr_allow_names[fi] : pr_allow_nonames[fi]; 400 optiov[opt.uio_iovcnt].iov_len = 401 strlen(optiov[opt.uio_iovcnt].iov_base) + 1; 402 opt.uio_iovcnt += 2; 403 } 404 optiov[opt.uio_iovcnt].iov_base = "enforce_statfs"; 405 optiov[opt.uio_iovcnt].iov_len = sizeof("enforce_statfs"); 406 opt.uio_iovcnt++; 407 enforce_statfs = jail_default_enforce_statfs; 408 optiov[opt.uio_iovcnt].iov_base = &enforce_statfs; 409 optiov[opt.uio_iovcnt].iov_len = sizeof(enforce_statfs); 410 opt.uio_iovcnt++; 411 } 412 413 tmplen = MAXPATHLEN + MAXHOSTNAMELEN + MAXHOSTNAMELEN; 414#ifdef INET 415 ip4s = (j->version == 0) ? 1 : j->ip4s; 416 if (ip4s > jail_max_af_ips) 417 return (EINVAL); 418 tmplen += ip4s * sizeof(struct in_addr); 419#else 420 if (j->ip4s > 0) 421 return (EINVAL); 422#endif 423#ifdef INET6 424 if (j->ip6s > jail_max_af_ips) 425 return (EINVAL); 426 tmplen += j->ip6s * sizeof(struct in6_addr); 427#else 428 if (j->ip6s > 0) 429 return (EINVAL); 430#endif 431 u_path = malloc(tmplen, M_TEMP, M_WAITOK); 432 u_hostname = u_path + MAXPATHLEN; 433 u_name = u_hostname + MAXHOSTNAMELEN; 434#ifdef INET 435 u_ip4 = (struct in_addr *)(u_name + MAXHOSTNAMELEN); 436#endif 437#ifdef INET6 438#ifdef INET 439 u_ip6 = (struct in6_addr *)(u_ip4 + ip4s); 440#else 441 u_ip6 = (struct in6_addr *)(u_name + MAXHOSTNAMELEN); 442#endif 443#endif 444 optiov[opt.uio_iovcnt].iov_base = "path"; 445 optiov[opt.uio_iovcnt].iov_len = sizeof("path"); 446 opt.uio_iovcnt++; 447 optiov[opt.uio_iovcnt].iov_base = u_path; 448 error = copyinstr(j->path, u_path, MAXPATHLEN, 449 &optiov[opt.uio_iovcnt].iov_len); 450 if (error) { 451 free(u_path, M_TEMP); 452 return (error); 453 } 454 opt.uio_iovcnt++; 455 optiov[opt.uio_iovcnt].iov_base = "host.hostname"; 456 optiov[opt.uio_iovcnt].iov_len = sizeof("host.hostname"); 457 opt.uio_iovcnt++; 458 optiov[opt.uio_iovcnt].iov_base = u_hostname; 459 error = copyinstr(j->hostname, u_hostname, MAXHOSTNAMELEN, 460 &optiov[opt.uio_iovcnt].iov_len); 461 if (error) { 462 free(u_path, M_TEMP); 463 return (error); 464 } 465 opt.uio_iovcnt++; 466 if (j->jailname != NULL) { 467 optiov[opt.uio_iovcnt].iov_base = "name"; 468 optiov[opt.uio_iovcnt].iov_len = sizeof("name"); 469 opt.uio_iovcnt++; 470 optiov[opt.uio_iovcnt].iov_base = u_name; 471 error = copyinstr(j->jailname, u_name, MAXHOSTNAMELEN, 472 &optiov[opt.uio_iovcnt].iov_len); 473 if (error) { 474 free(u_path, M_TEMP); 475 return (error); 476 } 477 opt.uio_iovcnt++; 478 } 479#ifdef INET 480 optiov[opt.uio_iovcnt].iov_base = "ip4.addr"; 481 optiov[opt.uio_iovcnt].iov_len = sizeof("ip4.addr"); 482 opt.uio_iovcnt++; 483 optiov[opt.uio_iovcnt].iov_base = u_ip4; 484 optiov[opt.uio_iovcnt].iov_len = ip4s * sizeof(struct in_addr); 485 if (j->version == 0) 486 u_ip4->s_addr = j->ip4s; 487 else { 488 error = copyin(j->ip4, u_ip4, optiov[opt.uio_iovcnt].iov_len); 489 if (error) { 490 free(u_path, M_TEMP); 491 return (error); 492 } 493 } 494 opt.uio_iovcnt++; 495#endif 496#ifdef INET6 497 optiov[opt.uio_iovcnt].iov_base = "ip6.addr"; 498 optiov[opt.uio_iovcnt].iov_len = sizeof("ip6.addr"); 499 opt.uio_iovcnt++; 500 optiov[opt.uio_iovcnt].iov_base = u_ip6; 501 optiov[opt.uio_iovcnt].iov_len = j->ip6s * sizeof(struct in6_addr); 502 error = copyin(j->ip6, u_ip6, optiov[opt.uio_iovcnt].iov_len); 503 if (error) { 504 free(u_path, M_TEMP); 505 return (error); 506 } 507 opt.uio_iovcnt++; 508#endif 509 KASSERT(opt.uio_iovcnt <= sizeof(optiov) / sizeof(optiov[0]), 510 ("kern_jail: too many iovecs (%d)", opt.uio_iovcnt)); 511 error = kern_jail_set(td, &opt, JAIL_CREATE | JAIL_ATTACH); 512 free(u_path, M_TEMP); 513 return (error); 514} 515 516 517/* 518 * struct jail_set_args { 519 * struct iovec *iovp; 520 * unsigned int iovcnt; 521 * int flags; 522 * }; 523 */ 524int 525sys_jail_set(struct thread *td, struct jail_set_args *uap) 526{ 527 struct uio *auio; 528 int error; 529 530 /* Check that we have an even number of iovecs. */ 531 if (uap->iovcnt & 1) 532 return (EINVAL); 533 534 error = copyinuio(uap->iovp, uap->iovcnt, &auio); 535 if (error) 536 return (error); 537 error = kern_jail_set(td, auio, uap->flags); 538 free(auio, M_IOV); 539 return (error); 540} 541 542int 543kern_jail_set(struct thread *td, struct uio *optuio, int flags) 544{ 545 struct nameidata nd; 546#ifdef INET 547 struct in_addr *ip4; 548#endif 549#ifdef INET6 550 struct in6_addr *ip6; 551#endif 552 struct vfsopt *opt; 553 struct vfsoptlist *opts; 554 struct prison *pr, *deadpr, *mypr, *ppr, *tpr; 555 struct vnode *root; 556 char *domain, *errmsg, *host, *name, *namelc, *p, *path, *uuid; 557 char *g_path, *osrelstr; 558#if defined(INET) || defined(INET6) 559 struct prison *tppr; 560 void *op; 561#endif 562 unsigned long hid; 563 size_t namelen, onamelen; 564 int created, cuflags, descend, enforce, error, errmsg_len, errmsg_pos; 565 int gotchildmax, gotenforce, gothid, gotrsnum, gotslevel; 566 int fi, jid, jsys, len, level; 567 int childmax, osreldt, rsnum, slevel; 568 int fullpath_disabled; 569#if defined(INET) || defined(INET6) 570 int ii, ij; 571#endif 572#ifdef INET 573 int ip4s, redo_ip4; 574#endif 575#ifdef INET6 576 int ip6s, redo_ip6; 577#endif 578 uint64_t pr_allow, ch_allow, pr_flags, ch_flags; 579 unsigned tallow; 580 char numbuf[12]; 581 582 error = priv_check(td, PRIV_JAIL_SET); 583 if (!error && (flags & JAIL_ATTACH)) 584 error = priv_check(td, PRIV_JAIL_ATTACH); 585 if (error) 586 return (error); 587 mypr = ppr = td->td_ucred->cr_prison; 588 if ((flags & JAIL_CREATE) && mypr->pr_childmax == 0) 589 return (EPERM); 590 if (flags & ~JAIL_SET_MASK) 591 return (EINVAL); 592 593 /* 594 * Check all the parameters before committing to anything. Not all 595 * errors can be caught early, but we may as well try. Also, this 596 * takes care of some expensive stuff (path lookup) before getting 597 * the allprison lock. 598 * 599 * XXX Jails are not filesystems, and jail parameters are not mount 600 * options. But it makes more sense to re-use the vfsopt code 601 * than duplicate it under a different name. 602 */ 603 error = vfs_buildopts(optuio, &opts); 604 if (error) 605 return (error); 606#ifdef INET 607 ip4 = NULL; 608#endif 609#ifdef INET6 610 ip6 = NULL; 611#endif 612 g_path = NULL; 613 614 error = vfs_copyopt(opts, "jid", &jid, sizeof(jid)); 615 if (error == ENOENT) 616 jid = 0; 617 else if (error != 0) 618 goto done_free; 619 620 error = vfs_copyopt(opts, "securelevel", &slevel, sizeof(slevel)); 621 if (error == ENOENT) 622 gotslevel = 0; 623 else if (error != 0) 624 goto done_free; 625 else 626 gotslevel = 1; 627 628 error = 629 vfs_copyopt(opts, "children.max", &childmax, sizeof(childmax)); 630 if (error == ENOENT) 631 gotchildmax = 0; 632 else if (error != 0) 633 goto done_free; 634 else 635 gotchildmax = 1; 636 637 error = vfs_copyopt(opts, "enforce_statfs", &enforce, sizeof(enforce)); 638 if (error == ENOENT) 639 gotenforce = 0; 640 else if (error != 0) 641 goto done_free; 642 else if (enforce < 0 || enforce > 2) { 643 error = EINVAL; 644 goto done_free; 645 } else 646 gotenforce = 1; 647 648 error = vfs_copyopt(opts, "devfs_ruleset", &rsnum, sizeof(rsnum)); 649 if (error == ENOENT) 650 gotrsnum = 0; 651 else if (error != 0) 652 goto done_free; 653 else 654 gotrsnum = 1; 655 656 pr_flags = ch_flags = 0; 657 for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]); 658 fi++) { 659 if (pr_flag_names[fi] == NULL) 660 continue; 661 vfs_flagopt(opts, pr_flag_names[fi], &pr_flags, 1 << fi); 662 vfs_flagopt(opts, pr_flag_nonames[fi], &ch_flags, 1 << fi); 663 } 664 ch_flags |= pr_flags; 665 for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]); 666 fi++) { 667 error = vfs_copyopt(opts, pr_flag_jailsys[fi].name, &jsys, 668 sizeof(jsys)); 669 if (error == ENOENT) 670 continue; 671 if (error != 0) 672 goto done_free; 673 switch (jsys) { 674 case JAIL_SYS_DISABLE: 675 if (!pr_flag_jailsys[fi].disable) { 676 error = EINVAL; 677 goto done_free; 678 } 679 pr_flags |= pr_flag_jailsys[fi].disable; 680 break; 681 case JAIL_SYS_NEW: 682 pr_flags |= pr_flag_jailsys[fi].new; 683 break; 684 case JAIL_SYS_INHERIT: 685 break; 686 default: 687 error = EINVAL; 688 goto done_free; 689 } 690 ch_flags |= 691 pr_flag_jailsys[fi].new | pr_flag_jailsys[fi].disable; 692 } 693 if ((flags & (JAIL_CREATE | JAIL_UPDATE | JAIL_ATTACH)) == JAIL_CREATE 694 && !(pr_flags & PR_PERSIST)) { 695 error = EINVAL; 696 vfs_opterror(opts, "new jail must persist or attach"); 697 goto done_errmsg; 698 } 699#ifdef VIMAGE 700 if ((flags & JAIL_UPDATE) && (ch_flags & PR_VNET)) { 701 error = EINVAL; 702 vfs_opterror(opts, "vnet cannot be changed after creation"); 703 goto done_errmsg; 704 } 705#endif 706#ifdef INET 707 if ((flags & JAIL_UPDATE) && (ch_flags & PR_IP4_USER)) { 708 error = EINVAL; 709 vfs_opterror(opts, "ip4 cannot be changed after creation"); 710 goto done_errmsg; 711 } 712#endif 713#ifdef INET6 714 if ((flags & JAIL_UPDATE) && (ch_flags & PR_IP6_USER)) { 715 error = EINVAL; 716 vfs_opterror(opts, "ip6 cannot be changed after creation"); 717 goto done_errmsg; 718 } 719#endif 720 721 pr_allow = ch_allow = 0; 722 for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]); 723 fi++) { 724 vfs_flagopt(opts, pr_allow_names[fi], &pr_allow, 1 << fi); 725 vfs_flagopt(opts, pr_allow_nonames[fi], &ch_allow, 1 << fi); 726 } 727 ch_allow |= pr_allow; 728 729 error = vfs_getopt(opts, "name", (void **)&name, &len); 730 if (error == ENOENT) 731 name = NULL; 732 else if (error != 0) 733 goto done_free; 734 else { 735 if (len == 0 || name[len - 1] != '\0') { 736 error = EINVAL; 737 goto done_free; 738 } 739 if (len > MAXHOSTNAMELEN) { 740 error = ENAMETOOLONG; 741 goto done_free; 742 } 743 } 744 745 error = vfs_getopt(opts, "host.hostname", (void **)&host, &len); 746 if (error == ENOENT) 747 host = NULL; 748 else if (error != 0) 749 goto done_free; 750 else { 751 ch_flags |= PR_HOST; 752 pr_flags |= PR_HOST; 753 if (len == 0 || host[len - 1] != '\0') { 754 error = EINVAL; 755 goto done_free; 756 } 757 if (len > MAXHOSTNAMELEN) { 758 error = ENAMETOOLONG; 759 goto done_free; 760 } 761 } 762 763 error = vfs_getopt(opts, "host.domainname", (void **)&domain, &len); 764 if (error == ENOENT) 765 domain = NULL; 766 else if (error != 0) 767 goto done_free; 768 else { 769 ch_flags |= PR_HOST; 770 pr_flags |= PR_HOST; 771 if (len == 0 || domain[len - 1] != '\0') { 772 error = EINVAL; 773 goto done_free; 774 } 775 if (len > MAXHOSTNAMELEN) { 776 error = ENAMETOOLONG; 777 goto done_free; 778 } 779 } 780 781 error = vfs_getopt(opts, "host.hostuuid", (void **)&uuid, &len); 782 if (error == ENOENT) 783 uuid = NULL; 784 else if (error != 0) 785 goto done_free; 786 else { 787 ch_flags |= PR_HOST; 788 pr_flags |= PR_HOST; 789 if (len == 0 || uuid[len - 1] != '\0') { 790 error = EINVAL; 791 goto done_free; 792 } 793 if (len > HOSTUUIDLEN) { 794 error = ENAMETOOLONG; 795 goto done_free; 796 } 797 } 798 799#ifdef COMPAT_FREEBSD32 800 if (SV_PROC_FLAG(td->td_proc, SV_ILP32)) { 801 uint32_t hid32; 802 803 error = vfs_copyopt(opts, "host.hostid", &hid32, sizeof(hid32)); 804 hid = hid32; 805 } else 806#endif 807 error = vfs_copyopt(opts, "host.hostid", &hid, sizeof(hid)); 808 if (error == ENOENT) 809 gothid = 0; 810 else if (error != 0) 811 goto done_free; 812 else { 813 gothid = 1; 814 ch_flags |= PR_HOST; 815 pr_flags |= PR_HOST; 816 } 817 818#ifdef INET 819 error = vfs_getopt(opts, "ip4.addr", &op, &ip4s); 820 if (error == ENOENT) 821 ip4s = 0; 822 else if (error != 0) 823 goto done_free; 824 else if (ip4s & (sizeof(*ip4) - 1)) { 825 error = EINVAL; 826 goto done_free; 827 } else { 828 ch_flags |= PR_IP4_USER | PR_IP4_DISABLE; 829 if (ip4s == 0) 830 pr_flags |= PR_IP4_USER | PR_IP4_DISABLE; 831 else { 832 pr_flags = (pr_flags & ~PR_IP4_DISABLE) | PR_IP4_USER; 833 ip4s /= sizeof(*ip4); 834 if (ip4s > jail_max_af_ips) { 835 error = EINVAL; 836 vfs_opterror(opts, "too many IPv4 addresses"); 837 goto done_errmsg; 838 } 839 ip4 = malloc(ip4s * sizeof(*ip4), M_PRISON, M_WAITOK); 840 bcopy(op, ip4, ip4s * sizeof(*ip4)); 841 /* 842 * IP addresses are all sorted but ip[0] to preserve 843 * the primary IP address as given from userland. 844 * This special IP is used for unbound outgoing 845 * connections as well for "loopback" traffic in case 846 * source address selection cannot find any more fitting 847 * address to connect from. 848 */ 849 if (ip4s > 1) 850 qsort(ip4 + 1, ip4s - 1, sizeof(*ip4), qcmp_v4); 851 /* 852 * Check for duplicate addresses and do some simple 853 * zero and broadcast checks. If users give other bogus 854 * addresses it is their problem. 855 * 856 * We do not have to care about byte order for these 857 * checks so we will do them in NBO. 858 */ 859 for (ii = 0; ii < ip4s; ii++) { 860 if (ip4[ii].s_addr == INADDR_ANY || 861 ip4[ii].s_addr == INADDR_BROADCAST) { 862 error = EINVAL; 863 goto done_free; 864 } 865 if ((ii+1) < ip4s && 866 (ip4[0].s_addr == ip4[ii+1].s_addr || 867 ip4[ii].s_addr == ip4[ii+1].s_addr)) { 868 error = EINVAL; 869 goto done_free; 870 } 871 } 872 } 873 } 874#endif 875 876#ifdef INET6 877 error = vfs_getopt(opts, "ip6.addr", &op, &ip6s); 878 if (error == ENOENT) 879 ip6s = 0; 880 else if (error != 0) 881 goto done_free; 882 else if (ip6s & (sizeof(*ip6) - 1)) { 883 error = EINVAL; 884 goto done_free; 885 } else { 886 ch_flags |= PR_IP6_USER | PR_IP6_DISABLE; 887 if (ip6s == 0) 888 pr_flags |= PR_IP6_USER | PR_IP6_DISABLE; 889 else { 890 pr_flags = (pr_flags & ~PR_IP6_DISABLE) | PR_IP6_USER; 891 ip6s /= sizeof(*ip6); 892 if (ip6s > jail_max_af_ips) { 893 error = EINVAL; 894 vfs_opterror(opts, "too many IPv6 addresses"); 895 goto done_errmsg; 896 } 897 ip6 = malloc(ip6s * sizeof(*ip6), M_PRISON, M_WAITOK); 898 bcopy(op, ip6, ip6s * sizeof(*ip6)); 899 if (ip6s > 1) 900 qsort(ip6 + 1, ip6s - 1, sizeof(*ip6), qcmp_v6); 901 for (ii = 0; ii < ip6s; ii++) { 902 if (IN6_IS_ADDR_UNSPECIFIED(&ip6[ii])) { 903 error = EINVAL; 904 goto done_free; 905 } 906 if ((ii+1) < ip6s && 907 (IN6_ARE_ADDR_EQUAL(&ip6[0], &ip6[ii+1]) || 908 IN6_ARE_ADDR_EQUAL(&ip6[ii], &ip6[ii+1]))) 909 { 910 error = EINVAL; 911 goto done_free; 912 } 913 } 914 } 915 } 916#endif 917 918#if defined(VIMAGE) && (defined(INET) || defined(INET6)) 919 if ((ch_flags & PR_VNET) && (ch_flags & (PR_IP4_USER | PR_IP6_USER))) { 920 error = EINVAL; 921 vfs_opterror(opts, 922 "vnet jails cannot have IP address restrictions"); 923 goto done_errmsg; 924 } 925#endif 926 927 fullpath_disabled = 0; 928 root = NULL; 929 error = vfs_getopt(opts, "path", (void **)&path, &len); 930 if (error == ENOENT) 931 path = NULL; 932 else if (error != 0) 933 goto done_free; 934 else { 935 if (flags & JAIL_UPDATE) { 936 error = EINVAL; 937 vfs_opterror(opts, 938 "path cannot be changed after creation"); 939 goto done_errmsg; 940 } 941 if (len == 0 || path[len - 1] != '\0') { 942 error = EINVAL; 943 goto done_free; 944 } 945 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, 946 path, td); 947 error = namei(&nd); 948 if (error) 949 goto done_free; 950 root = nd.ni_vp; 951 NDFREE(&nd, NDF_ONLY_PNBUF); 952 g_path = malloc(MAXPATHLEN, M_TEMP, M_WAITOK); 953 strlcpy(g_path, path, MAXPATHLEN); 954 error = vn_path_to_global_path(td, root, g_path, MAXPATHLEN); 955 if (error == 0) 956 path = g_path; 957 else if (error == ENODEV) { 958 /* proceed if sysctl debug.disablefullpath == 1 */ 959 fullpath_disabled = 1; 960 if (len < 2 || (len == 2 && path[0] == '/')) 961 path = NULL; 962 } else { 963 /* exit on other errors */ 964 goto done_free; 965 } 966 if (root->v_type != VDIR) { 967 error = ENOTDIR; 968 vput(root); 969 goto done_free; 970 } 971 VOP_UNLOCK(root, 0); 972 if (fullpath_disabled) { 973 /* Leave room for a real-root full pathname. */ 974 if (len + (path[0] == '/' && strcmp(mypr->pr_path, "/") 975 ? strlen(mypr->pr_path) : 0) > MAXPATHLEN) { 976 error = ENAMETOOLONG; 977 goto done_free; 978 } 979 } 980 } 981 982 error = vfs_getopt(opts, "osrelease", (void **)&osrelstr, &len); 983 if (error == ENOENT) 984 osrelstr = NULL; 985 else if (error != 0) 986 goto done_free; 987 else { 988 if (flags & JAIL_UPDATE) { 989 error = EINVAL; 990 vfs_opterror(opts, 991 "osrelease cannot be changed after creation"); 992 goto done_errmsg; 993 } 994 if (len == 0 || len >= OSRELEASELEN) { 995 error = EINVAL; 996 vfs_opterror(opts, 997 "osrelease string must be 1-%d bytes long", 998 OSRELEASELEN - 1); 999 goto done_errmsg; 1000 } 1001 } 1002 1003 error = vfs_copyopt(opts, "osreldate", &osreldt, sizeof(osreldt)); 1004 if (error == ENOENT) 1005 osreldt = 0; 1006 else if (error != 0) 1007 goto done_free; 1008 else { 1009 if (flags & JAIL_UPDATE) { 1010 error = EINVAL; 1011 vfs_opterror(opts, 1012 "osreldate cannot be changed after creation"); 1013 goto done_errmsg; 1014 } 1015 if (osreldt == 0) { 1016 error = EINVAL; 1017 vfs_opterror(opts, "osreldate cannot be 0"); 1018 goto done_errmsg; 1019 } 1020 } 1021 1022 /* 1023 * Grab the allprison lock before letting modules check their 1024 * parameters. Once we have it, do not let go so we'll have a 1025 * consistent view of the OSD list. 1026 */ 1027 sx_xlock(&allprison_lock); 1028 error = osd_jail_call(NULL, PR_METHOD_CHECK, opts); 1029 if (error) 1030 goto done_unlock_list; 1031 1032 /* By now, all parameters should have been noted. */ 1033 TAILQ_FOREACH(opt, opts, link) { 1034 if (!opt->seen && strcmp(opt->name, "errmsg")) { 1035 error = EINVAL; 1036 vfs_opterror(opts, "unknown parameter: %s", opt->name); 1037 goto done_unlock_list; 1038 } 1039 } 1040 1041 /* 1042 * See if we are creating a new record or updating an existing one. 1043 * This abuses the file error codes ENOENT and EEXIST. 1044 */ 1045 cuflags = flags & (JAIL_CREATE | JAIL_UPDATE); 1046 if (!cuflags) { 1047 error = EINVAL; 1048 vfs_opterror(opts, "no valid operation (create or update)"); 1049 goto done_unlock_list; 1050 } 1051 pr = NULL; 1052 namelc = NULL; 1053 if (cuflags == JAIL_CREATE && jid == 0 && name != NULL) { 1054 namelc = strrchr(name, '.'); 1055 jid = strtoul(namelc != NULL ? namelc + 1 : name, &p, 10); 1056 if (*p != '\0') 1057 jid = 0; 1058 } 1059 if (jid != 0) { 1060 /* 1061 * See if a requested jid already exists. There is an 1062 * information leak here if the jid exists but is not within 1063 * the caller's jail hierarchy. Jail creators will get EEXIST 1064 * even though they cannot see the jail, and CREATE | UPDATE 1065 * will return ENOENT which is not normally a valid error. 1066 */ 1067 if (jid < 0) { 1068 error = EINVAL; 1069 vfs_opterror(opts, "negative jid"); 1070 goto done_unlock_list; 1071 } 1072 pr = prison_find(jid); 1073 if (pr != NULL) { 1074 ppr = pr->pr_parent; 1075 /* Create: jid must not exist. */ 1076 if (cuflags == JAIL_CREATE) { 1077 mtx_unlock(&pr->pr_mtx); 1078 error = EEXIST; 1079 vfs_opterror(opts, "jail %d already exists", 1080 jid); 1081 goto done_unlock_list; 1082 } 1083 if (!prison_ischild(mypr, pr)) { 1084 mtx_unlock(&pr->pr_mtx); 1085 pr = NULL; 1086 } else if (pr->pr_uref == 0) { 1087 if (!(flags & JAIL_DYING)) { 1088 mtx_unlock(&pr->pr_mtx); 1089 error = ENOENT; 1090 vfs_opterror(opts, "jail %d is dying", 1091 jid); 1092 goto done_unlock_list; 1093 } else if ((flags & JAIL_ATTACH) || 1094 (pr_flags & PR_PERSIST)) { 1095 /* 1096 * A dying jail might be resurrected 1097 * (via attach or persist), but first 1098 * it must determine if another jail 1099 * has claimed its name. Accomplish 1100 * this by implicitly re-setting the 1101 * name. 1102 */ 1103 if (name == NULL) 1104 name = prison_name(mypr, pr); 1105 } 1106 } 1107 } 1108 if (pr == NULL) { 1109 /* Update: jid must exist. */ 1110 if (cuflags == JAIL_UPDATE) { 1111 error = ENOENT; 1112 vfs_opterror(opts, "jail %d not found", jid); 1113 goto done_unlock_list; 1114 } 1115 } 1116 } 1117 /* 1118 * If the caller provided a name, look for a jail by that name. 1119 * This has different semantics for creates and updates keyed by jid 1120 * (where the name must not already exist in a different jail), 1121 * and updates keyed by the name itself (where the name must exist 1122 * because that is the jail being updated). 1123 */ 1124 if (name != NULL) { 1125 namelc = strrchr(name, '.'); 1126 if (namelc == NULL) 1127 namelc = name; 1128 else { 1129 /* 1130 * This is a hierarchical name. Split it into the 1131 * parent and child names, and make sure the parent 1132 * exists or matches an already found jail. 1133 */ 1134 *namelc = '\0'; 1135 if (pr != NULL) { 1136 if (strncmp(name, ppr->pr_name, namelc - name) 1137 || ppr->pr_name[namelc - name] != '\0') { 1138 mtx_unlock(&pr->pr_mtx); 1139 error = EINVAL; 1140 vfs_opterror(opts, 1141 "cannot change jail's parent"); 1142 goto done_unlock_list; 1143 } 1144 } else { 1145 ppr = prison_find_name(mypr, name); 1146 if (ppr == NULL) { 1147 error = ENOENT; 1148 vfs_opterror(opts, 1149 "jail \"%s\" not found", name); 1150 goto done_unlock_list; 1151 } 1152 mtx_unlock(&ppr->pr_mtx); 1153 } 1154 name = ++namelc; 1155 } 1156 if (name[0] != '\0') { 1157 namelen = 1158 (ppr == &prison0) ? 0 : strlen(ppr->pr_name) + 1; 1159 name_again: 1160 deadpr = NULL; 1161 FOREACH_PRISON_CHILD(ppr, tpr) { 1162 if (tpr != pr && tpr->pr_ref > 0 && 1163 !strcmp(tpr->pr_name + namelen, name)) { 1164 if (pr == NULL && 1165 cuflags != JAIL_CREATE) { 1166 mtx_lock(&tpr->pr_mtx); 1167 if (tpr->pr_ref > 0) { 1168 /* 1169 * Use this jail 1170 * for updates. 1171 */ 1172 if (tpr->pr_uref > 0) { 1173 pr = tpr; 1174 break; 1175 } 1176 deadpr = tpr; 1177 } 1178 mtx_unlock(&tpr->pr_mtx); 1179 } else if (tpr->pr_uref > 0) { 1180 /* 1181 * Create, or update(jid): 1182 * name must not exist in an 1183 * active sibling jail. 1184 */ 1185 error = EEXIST; 1186 if (pr != NULL) 1187 mtx_unlock(&pr->pr_mtx); 1188 vfs_opterror(opts, 1189 "jail \"%s\" already exists", 1190 name); 1191 goto done_unlock_list; 1192 } 1193 } 1194 } 1195 /* If no active jail is found, use a dying one. */ 1196 if (deadpr != NULL && pr == NULL) { 1197 if (flags & JAIL_DYING) { 1198 mtx_lock(&deadpr->pr_mtx); 1199 if (deadpr->pr_ref == 0) { 1200 mtx_unlock(&deadpr->pr_mtx); 1201 goto name_again; 1202 } 1203 pr = deadpr; 1204 } else if (cuflags == JAIL_UPDATE) { 1205 error = ENOENT; 1206 vfs_opterror(opts, 1207 "jail \"%s\" is dying", name); 1208 goto done_unlock_list; 1209 } 1210 } 1211 /* Update: name must exist if no jid. */ 1212 else if (cuflags == JAIL_UPDATE && pr == NULL) { 1213 error = ENOENT; 1214 vfs_opterror(opts, "jail \"%s\" not found", 1215 name); 1216 goto done_unlock_list; 1217 } 1218 } 1219 } 1220 /* Update: must provide a jid or name. */ 1221 else if (cuflags == JAIL_UPDATE && pr == NULL) { 1222 error = ENOENT; 1223 vfs_opterror(opts, "update specified no jail"); 1224 goto done_unlock_list; 1225 } 1226 1227 /* If there's no prison to update, create a new one and link it in. */ 1228 if (pr == NULL) { 1229 for (tpr = mypr; tpr != NULL; tpr = tpr->pr_parent) 1230 if (tpr->pr_childcount >= tpr->pr_childmax) { 1231 error = EPERM; 1232 vfs_opterror(opts, "prison limit exceeded"); 1233 goto done_unlock_list; 1234 } 1235 created = 1; 1236 mtx_lock(&ppr->pr_mtx); 1237 if (ppr->pr_ref == 0) { 1238 mtx_unlock(&ppr->pr_mtx); 1239 error = ENOENT; 1240 vfs_opterror(opts, "parent jail went away!"); 1241 goto done_unlock_list; 1242 } 1243 ppr->pr_ref++; 1244 ppr->pr_uref++; 1245 mtx_unlock(&ppr->pr_mtx); 1246 pr = malloc(sizeof(*pr), M_PRISON, M_WAITOK | M_ZERO); 1247 if (jid == 0) { 1248 /* Find the next free jid. */ 1249 jid = lastprid + 1; 1250 findnext: 1251 if (jid == JAIL_MAX) 1252 jid = 1; 1253 TAILQ_FOREACH(tpr, &allprison, pr_list) { 1254 if (tpr->pr_id < jid) 1255 continue; 1256 if (tpr->pr_id > jid || tpr->pr_ref == 0) { 1257 TAILQ_INSERT_BEFORE(tpr, pr, pr_list); 1258 break; 1259 } 1260 if (jid == lastprid) { 1261 error = EAGAIN; 1262 vfs_opterror(opts, 1263 "no available jail IDs"); 1264 free(pr, M_PRISON); 1265 prison_deref(ppr, PD_DEREF | 1266 PD_DEUREF | PD_LIST_XLOCKED); 1267 goto done_releroot; 1268 } 1269 jid++; 1270 goto findnext; 1271 } 1272 lastprid = jid; 1273 } else { 1274 /* 1275 * The jail already has a jid (that did not yet exist), 1276 * so just find where to insert it. 1277 */ 1278 TAILQ_FOREACH(tpr, &allprison, pr_list) 1279 if (tpr->pr_id >= jid) { 1280 TAILQ_INSERT_BEFORE(tpr, pr, pr_list); 1281 break; 1282 } 1283 } 1284 if (tpr == NULL) 1285 TAILQ_INSERT_TAIL(&allprison, pr, pr_list); 1286 LIST_INSERT_HEAD(&ppr->pr_children, pr, pr_sibling); 1287 for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent) 1288 tpr->pr_childcount++; 1289 1290 pr->pr_parent = ppr; 1291 pr->pr_id = jid; 1292 1293 /* Set some default values, and inherit some from the parent. */ 1294 if (name == NULL) 1295 name = ""; 1296 if (path == NULL) { 1297 path = "/"; 1298 root = mypr->pr_root; 1299 vref(root); 1300 } 1301 strlcpy(pr->pr_hostuuid, DEFAULT_HOSTUUID, HOSTUUIDLEN); 1302 pr->pr_flags |= PR_HOST; 1303#if defined(INET) || defined(INET6) 1304#ifdef VIMAGE 1305 if (!(pr_flags & PR_VNET)) 1306#endif 1307 { 1308#ifdef INET 1309 if (!(ch_flags & PR_IP4_USER)) 1310 pr->pr_flags |= 1311 PR_IP4 | PR_IP4_USER | PR_IP4_DISABLE; 1312 else if (!(pr_flags & PR_IP4_USER)) { 1313 pr->pr_flags |= ppr->pr_flags & PR_IP4; 1314 if (ppr->pr_ip4 != NULL) { 1315 pr->pr_ip4s = ppr->pr_ip4s; 1316 pr->pr_ip4 = malloc(pr->pr_ip4s * 1317 sizeof(struct in_addr), M_PRISON, 1318 M_WAITOK); 1319 bcopy(ppr->pr_ip4, pr->pr_ip4, 1320 pr->pr_ip4s * sizeof(*pr->pr_ip4)); 1321 } 1322 } 1323#endif 1324#ifdef INET6 1325 if (!(ch_flags & PR_IP6_USER)) 1326 pr->pr_flags |= 1327 PR_IP6 | PR_IP6_USER | PR_IP6_DISABLE; 1328 else if (!(pr_flags & PR_IP6_USER)) { 1329 pr->pr_flags |= ppr->pr_flags & PR_IP6; 1330 if (ppr->pr_ip6 != NULL) { 1331 pr->pr_ip6s = ppr->pr_ip6s; 1332 pr->pr_ip6 = malloc(pr->pr_ip6s * 1333 sizeof(struct in6_addr), M_PRISON, 1334 M_WAITOK); 1335 bcopy(ppr->pr_ip6, pr->pr_ip6, 1336 pr->pr_ip6s * sizeof(*pr->pr_ip6)); 1337 } 1338 } 1339#endif 1340 } 1341#endif 1342 /* Source address selection is always on by default. */ 1343 pr->pr_flags |= _PR_IP_SADDRSEL; 1344 1345 pr->pr_securelevel = ppr->pr_securelevel; 1346 pr->pr_allow = JAIL_DEFAULT_ALLOW & ppr->pr_allow; 1347 pr->pr_enforce_statfs = JAIL_DEFAULT_ENFORCE_STATFS; 1348 pr->pr_devfs_rsnum = ppr->pr_devfs_rsnum; 1349 1350 pr->pr_osreldate = osreldt ? osreldt : ppr->pr_osreldate; 1351 if (osrelstr == NULL) 1352 strcpy(pr->pr_osrelease, ppr->pr_osrelease); 1353 else 1354 strcpy(pr->pr_osrelease, osrelstr); 1355 1356 LIST_INIT(&pr->pr_children); 1357 mtx_init(&pr->pr_mtx, "jail mutex", NULL, MTX_DEF | MTX_DUPOK); 1358 1359#ifdef VIMAGE 1360 /* Allocate a new vnet if specified. */ 1361 pr->pr_vnet = (pr_flags & PR_VNET) 1362 ? vnet_alloc() : ppr->pr_vnet; 1363#endif 1364 /* 1365 * Allocate a dedicated cpuset for each jail. 1366 * Unlike other initial settings, this may return an erorr. 1367 */ 1368 error = cpuset_create_root(ppr, &pr->pr_cpuset); 1369 if (error) { 1370 prison_deref(pr, PD_LIST_XLOCKED); 1371 goto done_releroot; 1372 } 1373 1374 mtx_lock(&pr->pr_mtx); 1375 /* 1376 * New prisons do not yet have a reference, because we do not 1377 * want other to see the incomplete prison once the 1378 * allprison_lock is downgraded. 1379 */ 1380 } else { 1381 created = 0; 1382 /* 1383 * Grab a reference for existing prisons, to ensure they 1384 * continue to exist for the duration of the call. 1385 */ 1386 pr->pr_ref++; 1387#if defined(VIMAGE) && (defined(INET) || defined(INET6)) 1388 if ((pr->pr_flags & PR_VNET) && 1389 (ch_flags & (PR_IP4_USER | PR_IP6_USER))) { 1390 error = EINVAL; 1391 vfs_opterror(opts, 1392 "vnet jails cannot have IP address restrictions"); 1393 goto done_deref_locked; 1394 } 1395#endif 1396#ifdef INET 1397 if (PR_IP4_USER & ch_flags & (pr_flags ^ pr->pr_flags)) { 1398 error = EINVAL; 1399 vfs_opterror(opts, 1400 "ip4 cannot be changed after creation"); 1401 goto done_deref_locked; 1402 } 1403#endif 1404#ifdef INET6 1405 if (PR_IP6_USER & ch_flags & (pr_flags ^ pr->pr_flags)) { 1406 error = EINVAL; 1407 vfs_opterror(opts, 1408 "ip6 cannot be changed after creation"); 1409 goto done_deref_locked; 1410 } 1411#endif 1412 } 1413 1414 /* Do final error checking before setting anything. */ 1415 if (gotslevel) { 1416 if (slevel < ppr->pr_securelevel) { 1417 error = EPERM; 1418 goto done_deref_locked; 1419 } 1420 } 1421 if (gotchildmax) { 1422 if (childmax >= ppr->pr_childmax) { 1423 error = EPERM; 1424 goto done_deref_locked; 1425 } 1426 } 1427 if (gotenforce) { 1428 if (enforce < ppr->pr_enforce_statfs) { 1429 error = EPERM; 1430 goto done_deref_locked; 1431 } 1432 } 1433 if (gotrsnum) { 1434 /* 1435 * devfs_rsnum is a uint16_t 1436 */ 1437 if (rsnum < 0 || rsnum > 65535) { 1438 error = EINVAL; 1439 goto done_deref_locked; 1440 } 1441 /* 1442 * Nested jails always inherit parent's devfs ruleset 1443 */ 1444 if (jailed(td->td_ucred)) { 1445 if (rsnum > 0 && rsnum != ppr->pr_devfs_rsnum) { 1446 error = EPERM; 1447 goto done_deref_locked; 1448 } else 1449 rsnum = ppr->pr_devfs_rsnum; 1450 } 1451 } 1452#ifdef INET 1453 if (ip4s > 0) { 1454 if (ppr->pr_flags & PR_IP4) { 1455 /* 1456 * Make sure the new set of IP addresses is a 1457 * subset of the parent's list. Don't worry 1458 * about the parent being unlocked, as any 1459 * setting is done with allprison_lock held. 1460 */ 1461 for (ij = 0; ij < ppr->pr_ip4s; ij++) 1462 if (ip4[0].s_addr == ppr->pr_ip4[ij].s_addr) 1463 break; 1464 if (ij == ppr->pr_ip4s) { 1465 error = EPERM; 1466 goto done_deref_locked; 1467 } 1468 if (ip4s > 1) { 1469 for (ii = ij = 1; ii < ip4s; ii++) { 1470 if (ip4[ii].s_addr == 1471 ppr->pr_ip4[0].s_addr) 1472 continue; 1473 for (; ij < ppr->pr_ip4s; ij++) 1474 if (ip4[ii].s_addr == 1475 ppr->pr_ip4[ij].s_addr) 1476 break; 1477 if (ij == ppr->pr_ip4s) 1478 break; 1479 } 1480 if (ij == ppr->pr_ip4s) { 1481 error = EPERM; 1482 goto done_deref_locked; 1483 } 1484 } 1485 } 1486 /* 1487 * Check for conflicting IP addresses. We permit them 1488 * if there is no more than one IP on each jail. If 1489 * there is a duplicate on a jail with more than one 1490 * IP stop checking and return error. 1491 */ 1492 tppr = ppr; 1493#ifdef VIMAGE 1494 for (; tppr != &prison0; tppr = tppr->pr_parent) 1495 if (tppr->pr_flags & PR_VNET) 1496 break; 1497#endif 1498 FOREACH_PRISON_DESCENDANT(tppr, tpr, descend) { 1499 if (tpr == pr || 1500#ifdef VIMAGE 1501 (tpr != tppr && (tpr->pr_flags & PR_VNET)) || 1502#endif 1503 tpr->pr_uref == 0) { 1504 descend = 0; 1505 continue; 1506 } 1507 if (!(tpr->pr_flags & PR_IP4_USER)) 1508 continue; 1509 descend = 0; 1510 if (tpr->pr_ip4 == NULL || 1511 (ip4s == 1 && tpr->pr_ip4s == 1)) 1512 continue; 1513 for (ii = 0; ii < ip4s; ii++) { 1514 if (_prison_check_ip4(tpr, &ip4[ii]) == 0) { 1515 error = EADDRINUSE; 1516 vfs_opterror(opts, 1517 "IPv4 addresses clash"); 1518 goto done_deref_locked; 1519 } 1520 } 1521 } 1522 } 1523#endif 1524#ifdef INET6 1525 if (ip6s > 0) { 1526 if (ppr->pr_flags & PR_IP6) { 1527 /* 1528 * Make sure the new set of IP addresses is a 1529 * subset of the parent's list. 1530 */ 1531 for (ij = 0; ij < ppr->pr_ip6s; ij++) 1532 if (IN6_ARE_ADDR_EQUAL(&ip6[0], 1533 &ppr->pr_ip6[ij])) 1534 break; 1535 if (ij == ppr->pr_ip6s) { 1536 error = EPERM; 1537 goto done_deref_locked; 1538 } 1539 if (ip6s > 1) { 1540 for (ii = ij = 1; ii < ip6s; ii++) { 1541 if (IN6_ARE_ADDR_EQUAL(&ip6[ii], 1542 &ppr->pr_ip6[0])) 1543 continue; 1544 for (; ij < ppr->pr_ip6s; ij++) 1545 if (IN6_ARE_ADDR_EQUAL( 1546 &ip6[ii], &ppr->pr_ip6[ij])) 1547 break; 1548 if (ij == ppr->pr_ip6s) 1549 break; 1550 } 1551 if (ij == ppr->pr_ip6s) { 1552 error = EPERM; 1553 goto done_deref_locked; 1554 } 1555 } 1556 } 1557 /* Check for conflicting IP addresses. */ 1558 tppr = ppr; 1559#ifdef VIMAGE 1560 for (; tppr != &prison0; tppr = tppr->pr_parent) 1561 if (tppr->pr_flags & PR_VNET) 1562 break; 1563#endif 1564 FOREACH_PRISON_DESCENDANT(tppr, tpr, descend) { 1565 if (tpr == pr || 1566#ifdef VIMAGE 1567 (tpr != tppr && (tpr->pr_flags & PR_VNET)) || 1568#endif 1569 tpr->pr_uref == 0) { 1570 descend = 0; 1571 continue; 1572 } 1573 if (!(tpr->pr_flags & PR_IP6_USER)) 1574 continue; 1575 descend = 0; 1576 if (tpr->pr_ip6 == NULL || 1577 (ip6s == 1 && tpr->pr_ip6s == 1)) 1578 continue; 1579 for (ii = 0; ii < ip6s; ii++) { 1580 if (_prison_check_ip6(tpr, &ip6[ii]) == 0) { 1581 error = EADDRINUSE; 1582 vfs_opterror(opts, 1583 "IPv6 addresses clash"); 1584 goto done_deref_locked; 1585 } 1586 } 1587 } 1588 } 1589#endif 1590 onamelen = namelen = 0; 1591 if (name != NULL) { 1592 /* Give a default name of the jid. Also allow the name to be 1593 * explicitly the jid - but not any other number, and only in 1594 * normal form (no leading zero/etc). 1595 */ 1596 if (name[0] == '\0') 1597 snprintf(name = numbuf, sizeof(numbuf), "%d", jid); 1598 else if ((strtoul(namelc, &p, 10) != jid || 1599 namelc[0] < '1' || namelc[0] > '9') && *p == '\0') { 1600 error = EINVAL; 1601 vfs_opterror(opts, 1602 "name cannot be numeric (unless it is the jid)"); 1603 goto done_deref_locked; 1604 } 1605 /* 1606 * Make sure the name isn't too long for the prison or its 1607 * children. 1608 */ 1609 onamelen = strlen(pr->pr_name); 1610 namelen = strlen(name); 1611 if (strlen(ppr->pr_name) + namelen + 2 > sizeof(pr->pr_name)) { 1612 error = ENAMETOOLONG; 1613 goto done_deref_locked; 1614 } 1615 FOREACH_PRISON_DESCENDANT(pr, tpr, descend) { 1616 if (strlen(tpr->pr_name) + (namelen - onamelen) >= 1617 sizeof(pr->pr_name)) { 1618 error = ENAMETOOLONG; 1619 goto done_deref_locked; 1620 } 1621 } 1622 } 1623 if (pr_allow & ~ppr->pr_allow) { 1624 error = EPERM; 1625 goto done_deref_locked; 1626 } 1627 1628 /* Set the parameters of the prison. */ 1629#ifdef INET 1630 redo_ip4 = 0; 1631 if (pr_flags & PR_IP4_USER) { 1632 pr->pr_flags |= PR_IP4; 1633 free(pr->pr_ip4, M_PRISON); 1634 pr->pr_ip4s = ip4s; 1635 pr->pr_ip4 = ip4; 1636 ip4 = NULL; 1637 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) { 1638#ifdef VIMAGE 1639 if (tpr->pr_flags & PR_VNET) { 1640 descend = 0; 1641 continue; 1642 } 1643#endif 1644 if (prison_restrict_ip4(tpr, NULL)) { 1645 redo_ip4 = 1; 1646 descend = 0; 1647 } 1648 } 1649 } 1650#endif 1651#ifdef INET6 1652 redo_ip6 = 0; 1653 if (pr_flags & PR_IP6_USER) { 1654 pr->pr_flags |= PR_IP6; 1655 free(pr->pr_ip6, M_PRISON); 1656 pr->pr_ip6s = ip6s; 1657 pr->pr_ip6 = ip6; 1658 ip6 = NULL; 1659 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) { 1660#ifdef VIMAGE 1661 if (tpr->pr_flags & PR_VNET) { 1662 descend = 0; 1663 continue; 1664 } 1665#endif 1666 if (prison_restrict_ip6(tpr, NULL)) { 1667 redo_ip6 = 1; 1668 descend = 0; 1669 } 1670 } 1671 } 1672#endif 1673 if (gotslevel) { 1674 pr->pr_securelevel = slevel; 1675 /* Set all child jails to be at least this level. */ 1676 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) 1677 if (tpr->pr_securelevel < slevel) 1678 tpr->pr_securelevel = slevel; 1679 } 1680 if (gotchildmax) { 1681 pr->pr_childmax = childmax; 1682 /* Set all child jails to under this limit. */ 1683 FOREACH_PRISON_DESCENDANT_LOCKED_LEVEL(pr, tpr, descend, level) 1684 if (tpr->pr_childmax > childmax - level) 1685 tpr->pr_childmax = childmax > level 1686 ? childmax - level : 0; 1687 } 1688 if (gotenforce) { 1689 pr->pr_enforce_statfs = enforce; 1690 /* Pass this restriction on to the children. */ 1691 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) 1692 if (tpr->pr_enforce_statfs < enforce) 1693 tpr->pr_enforce_statfs = enforce; 1694 } 1695 if (gotrsnum) { 1696 pr->pr_devfs_rsnum = rsnum; 1697 /* Pass this restriction on to the children. */ 1698 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) 1699 tpr->pr_devfs_rsnum = rsnum; 1700 } 1701 if (name != NULL) { 1702 if (ppr == &prison0) 1703 strlcpy(pr->pr_name, name, sizeof(pr->pr_name)); 1704 else 1705 snprintf(pr->pr_name, sizeof(pr->pr_name), "%s.%s", 1706 ppr->pr_name, name); 1707 /* Change this component of child names. */ 1708 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) { 1709 bcopy(tpr->pr_name + onamelen, tpr->pr_name + namelen, 1710 strlen(tpr->pr_name + onamelen) + 1); 1711 bcopy(pr->pr_name, tpr->pr_name, namelen); 1712 } 1713 } 1714 if (path != NULL) { 1715 /* Try to keep a real-rooted full pathname. */ 1716 if (fullpath_disabled && path[0] == '/' && 1717 strcmp(mypr->pr_path, "/")) 1718 snprintf(pr->pr_path, sizeof(pr->pr_path), "%s%s", 1719 mypr->pr_path, path); 1720 else 1721 strlcpy(pr->pr_path, path, sizeof(pr->pr_path)); 1722 pr->pr_root = root; 1723 } 1724 if (PR_HOST & ch_flags & ~pr_flags) { 1725 if (pr->pr_flags & PR_HOST) { 1726 /* 1727 * Copy the parent's host info. As with pr_ip4 above, 1728 * the lack of a lock on the parent is not a problem; 1729 * it is always set with allprison_lock at least 1730 * shared, and is held exclusively here. 1731 */ 1732 strlcpy(pr->pr_hostname, pr->pr_parent->pr_hostname, 1733 sizeof(pr->pr_hostname)); 1734 strlcpy(pr->pr_domainname, pr->pr_parent->pr_domainname, 1735 sizeof(pr->pr_domainname)); 1736 strlcpy(pr->pr_hostuuid, pr->pr_parent->pr_hostuuid, 1737 sizeof(pr->pr_hostuuid)); 1738 pr->pr_hostid = pr->pr_parent->pr_hostid; 1739 } 1740 } else if (host != NULL || domain != NULL || uuid != NULL || gothid) { 1741 /* Set this prison, and any descendants without PR_HOST. */ 1742 if (host != NULL) 1743 strlcpy(pr->pr_hostname, host, sizeof(pr->pr_hostname)); 1744 if (domain != NULL) 1745 strlcpy(pr->pr_domainname, domain, 1746 sizeof(pr->pr_domainname)); 1747 if (uuid != NULL) 1748 strlcpy(pr->pr_hostuuid, uuid, sizeof(pr->pr_hostuuid)); 1749 if (gothid) 1750 pr->pr_hostid = hid; 1751 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) { 1752 if (tpr->pr_flags & PR_HOST) 1753 descend = 0; 1754 else { 1755 if (host != NULL) 1756 strlcpy(tpr->pr_hostname, 1757 pr->pr_hostname, 1758 sizeof(tpr->pr_hostname)); 1759 if (domain != NULL) 1760 strlcpy(tpr->pr_domainname, 1761 pr->pr_domainname, 1762 sizeof(tpr->pr_domainname)); 1763 if (uuid != NULL) 1764 strlcpy(tpr->pr_hostuuid, 1765 pr->pr_hostuuid, 1766 sizeof(tpr->pr_hostuuid)); 1767 if (gothid) 1768 tpr->pr_hostid = hid; 1769 } 1770 } 1771 } 1772 if ((tallow = ch_allow & ~pr_allow)) { 1773 /* Clear allow bits in all children. */ 1774 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) 1775 tpr->pr_allow &= ~tallow; 1776 } 1777 pr->pr_allow = (pr->pr_allow & ~ch_allow) | pr_allow; 1778 /* 1779 * Persistent prisons get an extra reference, and prisons losing their 1780 * persist flag lose that reference. Only do this for existing prisons 1781 * for now, so new ones will remain unseen until after the module 1782 * handlers have completed. 1783 */ 1784 if (!created && (ch_flags & PR_PERSIST & (pr_flags ^ pr->pr_flags))) { 1785 if (pr_flags & PR_PERSIST) { 1786 pr->pr_ref++; 1787 pr->pr_uref++; 1788 } else { 1789 pr->pr_ref--; 1790 pr->pr_uref--; 1791 } 1792 } 1793 pr->pr_flags = (pr->pr_flags & ~ch_flags) | pr_flags; 1794 mtx_unlock(&pr->pr_mtx); 1795 1796#ifdef RACCT 1797 if (racct_enable && created) 1798 prison_racct_attach(pr); 1799#endif 1800 1801 /* Locks may have prevented a complete restriction of child IP 1802 * addresses. If so, allocate some more memory and try again. 1803 */ 1804#ifdef INET 1805 while (redo_ip4) { 1806 ip4s = pr->pr_ip4s; 1807 ip4 = malloc(ip4s * sizeof(*ip4), M_PRISON, M_WAITOK); 1808 mtx_lock(&pr->pr_mtx); 1809 redo_ip4 = 0; 1810 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) { 1811#ifdef VIMAGE 1812 if (tpr->pr_flags & PR_VNET) { 1813 descend = 0; 1814 continue; 1815 } 1816#endif 1817 if (prison_restrict_ip4(tpr, ip4)) { 1818 if (ip4 != NULL) 1819 ip4 = NULL; 1820 else 1821 redo_ip4 = 1; 1822 } 1823 } 1824 mtx_unlock(&pr->pr_mtx); 1825 } 1826#endif 1827#ifdef INET6 1828 while (redo_ip6) { 1829 ip6s = pr->pr_ip6s; 1830 ip6 = malloc(ip6s * sizeof(*ip6), M_PRISON, M_WAITOK); 1831 mtx_lock(&pr->pr_mtx); 1832 redo_ip6 = 0; 1833 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) { 1834#ifdef VIMAGE 1835 if (tpr->pr_flags & PR_VNET) { 1836 descend = 0; 1837 continue; 1838 } 1839#endif 1840 if (prison_restrict_ip6(tpr, ip6)) { 1841 if (ip6 != NULL) 1842 ip6 = NULL; 1843 else 1844 redo_ip6 = 1; 1845 } 1846 } 1847 mtx_unlock(&pr->pr_mtx); 1848 } 1849#endif 1850 1851 /* Let the modules do their work. */ 1852 sx_downgrade(&allprison_lock); 1853 if (created) { 1854 error = osd_jail_call(pr, PR_METHOD_CREATE, opts); 1855 if (error) { 1856 prison_deref(pr, PD_LIST_SLOCKED); 1857 goto done_errmsg; 1858 } 1859 } 1860 error = osd_jail_call(pr, PR_METHOD_SET, opts); 1861 if (error) { 1862 prison_deref(pr, created 1863 ? PD_LIST_SLOCKED 1864 : PD_DEREF | PD_LIST_SLOCKED); 1865 goto done_errmsg; 1866 } 1867 1868 /* Attach this process to the prison if requested. */ 1869 if (flags & JAIL_ATTACH) { 1870 mtx_lock(&pr->pr_mtx); 1871 error = do_jail_attach(td, pr); 1872 if (error) { 1873 vfs_opterror(opts, "attach failed"); 1874 if (!created) 1875 prison_deref(pr, PD_DEREF); 1876 goto done_errmsg; 1877 } 1878 } 1879 1880#ifdef RACCT 1881 if (racct_enable && !created) { 1882 if (!(flags & JAIL_ATTACH)) 1883 sx_sunlock(&allprison_lock); 1884 prison_racct_modify(pr); 1885 if (!(flags & JAIL_ATTACH)) 1886 sx_slock(&allprison_lock); 1887 } 1888#endif 1889 1890 td->td_retval[0] = pr->pr_id; 1891 1892 /* 1893 * Now that it is all there, drop the temporary reference from existing 1894 * prisons. Or add a reference to newly created persistent prisons 1895 * (which was not done earlier so that the prison would not be publicly 1896 * visible). 1897 */ 1898 if (!created) { 1899 prison_deref(pr, (flags & JAIL_ATTACH) 1900 ? PD_DEREF 1901 : PD_DEREF | PD_LIST_SLOCKED); 1902 } else { 1903 if (pr_flags & PR_PERSIST) { 1904 mtx_lock(&pr->pr_mtx); 1905 pr->pr_ref++; 1906 pr->pr_uref++; 1907 mtx_unlock(&pr->pr_mtx); 1908 } 1909 if (!(flags & JAIL_ATTACH)) 1910 sx_sunlock(&allprison_lock); 1911 } 1912 1913 goto done_errmsg; 1914 1915 done_deref_locked: 1916 prison_deref(pr, created 1917 ? PD_LOCKED | PD_LIST_XLOCKED 1918 : PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED); 1919 goto done_releroot; 1920 done_unlock_list: 1921 sx_xunlock(&allprison_lock); 1922 done_releroot: 1923 if (root != NULL) 1924 vrele(root); 1925 done_errmsg: 1926 if (error) { 1927 vfs_getopt(opts, "errmsg", (void **)&errmsg, &errmsg_len); 1928 if (errmsg_len > 0) { 1929 errmsg_pos = 2 * vfs_getopt_pos(opts, "errmsg") + 1; 1930 if (errmsg_pos > 0) { 1931 if (optuio->uio_segflg == UIO_SYSSPACE) 1932 bcopy(errmsg, 1933 optuio->uio_iov[errmsg_pos].iov_base, 1934 errmsg_len); 1935 else 1936 copyout(errmsg, 1937 optuio->uio_iov[errmsg_pos].iov_base, 1938 errmsg_len); 1939 } 1940 } 1941 } 1942 done_free: 1943#ifdef INET 1944 free(ip4, M_PRISON); 1945#endif 1946#ifdef INET6 1947 free(ip6, M_PRISON); 1948#endif 1949 if (g_path != NULL) 1950 free(g_path, M_TEMP); 1951 vfs_freeopts(opts); 1952 return (error); 1953} 1954 1955 1956/* 1957 * struct jail_get_args { 1958 * struct iovec *iovp; 1959 * unsigned int iovcnt; 1960 * int flags; 1961 * }; 1962 */ 1963int 1964sys_jail_get(struct thread *td, struct jail_get_args *uap) 1965{ 1966 struct uio *auio; 1967 int error; 1968 1969 /* Check that we have an even number of iovecs. */ 1970 if (uap->iovcnt & 1) 1971 return (EINVAL); 1972 1973 error = copyinuio(uap->iovp, uap->iovcnt, &auio); 1974 if (error) 1975 return (error); 1976 error = kern_jail_get(td, auio, uap->flags); 1977 if (error == 0) 1978 error = copyout(auio->uio_iov, uap->iovp, 1979 uap->iovcnt * sizeof (struct iovec)); 1980 free(auio, M_IOV); 1981 return (error); 1982} 1983 1984int 1985kern_jail_get(struct thread *td, struct uio *optuio, int flags) 1986{ 1987 struct prison *pr, *mypr; 1988 struct vfsopt *opt; 1989 struct vfsoptlist *opts; 1990 char *errmsg, *name; 1991 int error, errmsg_len, errmsg_pos, fi, i, jid, len, locked, pos; 1992 1993 if (flags & ~JAIL_GET_MASK) 1994 return (EINVAL); 1995 1996 /* Get the parameter list. */ 1997 error = vfs_buildopts(optuio, &opts); 1998 if (error) 1999 return (error); 2000 errmsg_pos = vfs_getopt_pos(opts, "errmsg"); 2001 mypr = td->td_ucred->cr_prison; 2002 2003 /* 2004 * Find the prison specified by one of: lastjid, jid, name. 2005 */ 2006 sx_slock(&allprison_lock); 2007 error = vfs_copyopt(opts, "lastjid", &jid, sizeof(jid)); 2008 if (error == 0) { 2009 TAILQ_FOREACH(pr, &allprison, pr_list) { 2010 if (pr->pr_id > jid && prison_ischild(mypr, pr)) { 2011 mtx_lock(&pr->pr_mtx); 2012 if (pr->pr_ref > 0 && 2013 (pr->pr_uref > 0 || (flags & JAIL_DYING))) 2014 break; 2015 mtx_unlock(&pr->pr_mtx); 2016 } 2017 } 2018 if (pr != NULL) 2019 goto found_prison; 2020 error = ENOENT; 2021 vfs_opterror(opts, "no jail after %d", jid); 2022 goto done_unlock_list; 2023 } else if (error != ENOENT) 2024 goto done_unlock_list; 2025 2026 error = vfs_copyopt(opts, "jid", &jid, sizeof(jid)); 2027 if (error == 0) { 2028 if (jid != 0) { 2029 pr = prison_find_child(mypr, jid); 2030 if (pr != NULL) { 2031 if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) { 2032 mtx_unlock(&pr->pr_mtx); 2033 error = ENOENT; 2034 vfs_opterror(opts, "jail %d is dying", 2035 jid); 2036 goto done_unlock_list; 2037 } 2038 goto found_prison; 2039 } 2040 error = ENOENT; 2041 vfs_opterror(opts, "jail %d not found", jid); 2042 goto done_unlock_list; 2043 } 2044 } else if (error != ENOENT) 2045 goto done_unlock_list; 2046 2047 error = vfs_getopt(opts, "name", (void **)&name, &len); 2048 if (error == 0) { 2049 if (len == 0 || name[len - 1] != '\0') { 2050 error = EINVAL; 2051 goto done_unlock_list; 2052 } 2053 pr = prison_find_name(mypr, name); 2054 if (pr != NULL) { 2055 if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) { 2056 mtx_unlock(&pr->pr_mtx); 2057 error = ENOENT; 2058 vfs_opterror(opts, "jail \"%s\" is dying", 2059 name); 2060 goto done_unlock_list; 2061 } 2062 goto found_prison; 2063 } 2064 error = ENOENT; 2065 vfs_opterror(opts, "jail \"%s\" not found", name); 2066 goto done_unlock_list; 2067 } else if (error != ENOENT) 2068 goto done_unlock_list; 2069 2070 vfs_opterror(opts, "no jail specified"); 2071 error = ENOENT; 2072 goto done_unlock_list; 2073 2074 found_prison: 2075 /* Get the parameters of the prison. */ 2076 pr->pr_ref++; 2077 locked = PD_LOCKED; 2078 td->td_retval[0] = pr->pr_id; 2079 error = vfs_setopt(opts, "jid", &pr->pr_id, sizeof(pr->pr_id)); 2080 if (error != 0 && error != ENOENT) 2081 goto done_deref; 2082 i = (pr->pr_parent == mypr) ? 0 : pr->pr_parent->pr_id; 2083 error = vfs_setopt(opts, "parent", &i, sizeof(i)); 2084 if (error != 0 && error != ENOENT) 2085 goto done_deref; 2086 error = vfs_setopts(opts, "name", prison_name(mypr, pr)); 2087 if (error != 0 && error != ENOENT) 2088 goto done_deref; 2089 error = vfs_setopt(opts, "cpuset.id", &pr->pr_cpuset->cs_id, 2090 sizeof(pr->pr_cpuset->cs_id)); 2091 if (error != 0 && error != ENOENT) 2092 goto done_deref; 2093 error = vfs_setopts(opts, "path", prison_path(mypr, pr)); 2094 if (error != 0 && error != ENOENT) 2095 goto done_deref; 2096#ifdef INET 2097 error = vfs_setopt_part(opts, "ip4.addr", pr->pr_ip4, 2098 pr->pr_ip4s * sizeof(*pr->pr_ip4)); 2099 if (error != 0 && error != ENOENT) 2100 goto done_deref; 2101#endif 2102#ifdef INET6 2103 error = vfs_setopt_part(opts, "ip6.addr", pr->pr_ip6, 2104 pr->pr_ip6s * sizeof(*pr->pr_ip6)); 2105 if (error != 0 && error != ENOENT) 2106 goto done_deref; 2107#endif 2108 error = vfs_setopt(opts, "securelevel", &pr->pr_securelevel, 2109 sizeof(pr->pr_securelevel)); 2110 if (error != 0 && error != ENOENT) 2111 goto done_deref; 2112 error = vfs_setopt(opts, "children.cur", &pr->pr_childcount, 2113 sizeof(pr->pr_childcount)); 2114 if (error != 0 && error != ENOENT) 2115 goto done_deref; 2116 error = vfs_setopt(opts, "children.max", &pr->pr_childmax, 2117 sizeof(pr->pr_childmax)); 2118 if (error != 0 && error != ENOENT) 2119 goto done_deref; 2120 error = vfs_setopts(opts, "host.hostname", pr->pr_hostname); 2121 if (error != 0 && error != ENOENT) 2122 goto done_deref; 2123 error = vfs_setopts(opts, "host.domainname", pr->pr_domainname); 2124 if (error != 0 && error != ENOENT) 2125 goto done_deref; 2126 error = vfs_setopts(opts, "host.hostuuid", pr->pr_hostuuid); 2127 if (error != 0 && error != ENOENT) 2128 goto done_deref; 2129#ifdef COMPAT_FREEBSD32 2130 if (SV_PROC_FLAG(td->td_proc, SV_ILP32)) { 2131 uint32_t hid32 = pr->pr_hostid; 2132 2133 error = vfs_setopt(opts, "host.hostid", &hid32, sizeof(hid32)); 2134 } else 2135#endif 2136 error = vfs_setopt(opts, "host.hostid", &pr->pr_hostid, 2137 sizeof(pr->pr_hostid)); 2138 if (error != 0 && error != ENOENT) 2139 goto done_deref; 2140 error = vfs_setopt(opts, "enforce_statfs", &pr->pr_enforce_statfs, 2141 sizeof(pr->pr_enforce_statfs)); 2142 if (error != 0 && error != ENOENT) 2143 goto done_deref; 2144 error = vfs_setopt(opts, "devfs_ruleset", &pr->pr_devfs_rsnum, 2145 sizeof(pr->pr_devfs_rsnum)); 2146 if (error != 0 && error != ENOENT) 2147 goto done_deref; 2148 for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]); 2149 fi++) { 2150 if (pr_flag_names[fi] == NULL) 2151 continue; 2152 i = (pr->pr_flags & (1 << fi)) ? 1 : 0; 2153 error = vfs_setopt(opts, pr_flag_names[fi], &i, sizeof(i)); 2154 if (error != 0 && error != ENOENT) 2155 goto done_deref; 2156 i = !i; 2157 error = vfs_setopt(opts, pr_flag_nonames[fi], &i, sizeof(i)); 2158 if (error != 0 && error != ENOENT) 2159 goto done_deref; 2160 } 2161 for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]); 2162 fi++) { 2163 i = pr->pr_flags & 2164 (pr_flag_jailsys[fi].disable | pr_flag_jailsys[fi].new); 2165 i = pr_flag_jailsys[fi].disable && 2166 (i == pr_flag_jailsys[fi].disable) ? JAIL_SYS_DISABLE 2167 : (i == pr_flag_jailsys[fi].new) ? JAIL_SYS_NEW 2168 : JAIL_SYS_INHERIT; 2169 error = 2170 vfs_setopt(opts, pr_flag_jailsys[fi].name, &i, sizeof(i)); 2171 if (error != 0 && error != ENOENT) 2172 goto done_deref; 2173 } 2174 for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]); 2175 fi++) { 2176 if (pr_allow_names[fi] == NULL) 2177 continue; 2178 i = (pr->pr_allow & (1 << fi)) ? 1 : 0; 2179 error = vfs_setopt(opts, pr_allow_names[fi], &i, sizeof(i)); 2180 if (error != 0 && error != ENOENT) 2181 goto done_deref; 2182 i = !i; 2183 error = vfs_setopt(opts, pr_allow_nonames[fi], &i, sizeof(i)); 2184 if (error != 0 && error != ENOENT) 2185 goto done_deref; 2186 } 2187 i = (pr->pr_uref == 0); 2188 error = vfs_setopt(opts, "dying", &i, sizeof(i)); 2189 if (error != 0 && error != ENOENT) 2190 goto done_deref; 2191 i = !i; 2192 error = vfs_setopt(opts, "nodying", &i, sizeof(i)); 2193 if (error != 0 && error != ENOENT) 2194 goto done_deref; 2195 error = vfs_setopt(opts, "osreldate", &pr->pr_osreldate, 2196 sizeof(pr->pr_osreldate)); 2197 if (error != 0 && error != ENOENT) 2198 goto done_deref; 2199 error = vfs_setopts(opts, "osrelease", pr->pr_osrelease); 2200 if (error != 0 && error != ENOENT) 2201 goto done_deref; 2202 2203 /* Get the module parameters. */ 2204 mtx_unlock(&pr->pr_mtx); 2205 locked = 0; 2206 error = osd_jail_call(pr, PR_METHOD_GET, opts); 2207 if (error) 2208 goto done_deref; 2209 prison_deref(pr, PD_DEREF | PD_LIST_SLOCKED); 2210 2211 /* By now, all parameters should have been noted. */ 2212 TAILQ_FOREACH(opt, opts, link) { 2213 if (!opt->seen && strcmp(opt->name, "errmsg")) { 2214 error = EINVAL; 2215 vfs_opterror(opts, "unknown parameter: %s", opt->name); 2216 goto done_errmsg; 2217 } 2218 } 2219 2220 /* Write the fetched parameters back to userspace. */ 2221 error = 0; 2222 TAILQ_FOREACH(opt, opts, link) { 2223 if (opt->pos >= 0 && opt->pos != errmsg_pos) { 2224 pos = 2 * opt->pos + 1; 2225 optuio->uio_iov[pos].iov_len = opt->len; 2226 if (opt->value != NULL) { 2227 if (optuio->uio_segflg == UIO_SYSSPACE) { 2228 bcopy(opt->value, 2229 optuio->uio_iov[pos].iov_base, 2230 opt->len); 2231 } else { 2232 error = copyout(opt->value, 2233 optuio->uio_iov[pos].iov_base, 2234 opt->len); 2235 if (error) 2236 break; 2237 } 2238 } 2239 } 2240 } 2241 goto done_errmsg; 2242 2243 done_deref: 2244 prison_deref(pr, locked | PD_DEREF | PD_LIST_SLOCKED); 2245 goto done_errmsg; 2246 2247 done_unlock_list: 2248 sx_sunlock(&allprison_lock); 2249 done_errmsg: 2250 if (error && errmsg_pos >= 0) { 2251 vfs_getopt(opts, "errmsg", (void **)&errmsg, &errmsg_len); 2252 errmsg_pos = 2 * errmsg_pos + 1; 2253 if (errmsg_len > 0) { 2254 if (optuio->uio_segflg == UIO_SYSSPACE) 2255 bcopy(errmsg, 2256 optuio->uio_iov[errmsg_pos].iov_base, 2257 errmsg_len); 2258 else 2259 copyout(errmsg, 2260 optuio->uio_iov[errmsg_pos].iov_base, 2261 errmsg_len); 2262 } 2263 } 2264 vfs_freeopts(opts); 2265 return (error); 2266} 2267 2268 2269/* 2270 * struct jail_remove_args { 2271 * int jid; 2272 * }; 2273 */ 2274int 2275sys_jail_remove(struct thread *td, struct jail_remove_args *uap) 2276{ 2277 struct prison *pr, *cpr, *lpr, *tpr; 2278 int descend, error; 2279 2280 error = priv_check(td, PRIV_JAIL_REMOVE); 2281 if (error) 2282 return (error); 2283 2284 sx_xlock(&allprison_lock); 2285 pr = prison_find_child(td->td_ucred->cr_prison, uap->jid); 2286 if (pr == NULL) { 2287 sx_xunlock(&allprison_lock); 2288 return (EINVAL); 2289 } 2290 2291 /* Remove all descendants of this prison, then remove this prison. */ 2292 pr->pr_ref++; 2293 if (!LIST_EMPTY(&pr->pr_children)) { 2294 mtx_unlock(&pr->pr_mtx); 2295 lpr = NULL; 2296 FOREACH_PRISON_DESCENDANT(pr, cpr, descend) { 2297 mtx_lock(&cpr->pr_mtx); 2298 if (cpr->pr_ref > 0) { 2299 tpr = cpr; 2300 cpr->pr_ref++; 2301 } else { 2302 /* Already removed - do not do it again. */ 2303 tpr = NULL; 2304 } 2305 mtx_unlock(&cpr->pr_mtx); 2306 if (lpr != NULL) { 2307 mtx_lock(&lpr->pr_mtx); 2308 prison_remove_one(lpr); 2309 sx_xlock(&allprison_lock); 2310 } 2311 lpr = tpr; 2312 } 2313 if (lpr != NULL) { 2314 mtx_lock(&lpr->pr_mtx); 2315 prison_remove_one(lpr); 2316 sx_xlock(&allprison_lock); 2317 } 2318 mtx_lock(&pr->pr_mtx); 2319 } 2320 prison_remove_one(pr); 2321 return (0); 2322} 2323 2324static void 2325prison_remove_one(struct prison *pr) 2326{ 2327 struct proc *p; 2328 int deuref; 2329 2330 /* If the prison was persistent, it is not anymore. */ 2331 deuref = 0; 2332 if (pr->pr_flags & PR_PERSIST) { 2333 pr->pr_ref--; 2334 deuref = PD_DEUREF; 2335 pr->pr_flags &= ~PR_PERSIST; 2336 } 2337 2338 /* 2339 * jail_remove added a reference. If that's the only one, remove 2340 * the prison now. 2341 */ 2342 KASSERT(pr->pr_ref > 0, 2343 ("prison_remove_one removing a dead prison (jid=%d)", pr->pr_id)); 2344 if (pr->pr_ref == 1) { 2345 prison_deref(pr, 2346 deuref | PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED); 2347 return; 2348 } 2349 2350 mtx_unlock(&pr->pr_mtx); 2351 sx_xunlock(&allprison_lock); 2352 /* 2353 * Kill all processes unfortunate enough to be attached to this prison. 2354 */ 2355 sx_slock(&allproc_lock); 2356 LIST_FOREACH(p, &allproc, p_list) { 2357 PROC_LOCK(p); 2358 if (p->p_state != PRS_NEW && p->p_ucred && 2359 p->p_ucred->cr_prison == pr) 2360 kern_psignal(p, SIGKILL); 2361 PROC_UNLOCK(p); 2362 } 2363 sx_sunlock(&allproc_lock); 2364 /* Remove the temporary reference added by jail_remove. */ 2365 prison_deref(pr, deuref | PD_DEREF); 2366} 2367 2368 2369/* 2370 * struct jail_attach_args { 2371 * int jid; 2372 * }; 2373 */ 2374int 2375sys_jail_attach(struct thread *td, struct jail_attach_args *uap) 2376{ 2377 struct prison *pr; 2378 int error; 2379 2380 error = priv_check(td, PRIV_JAIL_ATTACH); 2381 if (error) 2382 return (error); 2383 2384 sx_slock(&allprison_lock); 2385 pr = prison_find_child(td->td_ucred->cr_prison, uap->jid); 2386 if (pr == NULL) { 2387 sx_sunlock(&allprison_lock); 2388 return (EINVAL); 2389 } 2390 2391 /* 2392 * Do not allow a process to attach to a prison that is not 2393 * considered to be "alive". 2394 */ 2395 if (pr->pr_uref == 0) { 2396 mtx_unlock(&pr->pr_mtx); 2397 sx_sunlock(&allprison_lock); 2398 return (EINVAL); 2399 } 2400 2401 return (do_jail_attach(td, pr)); 2402} 2403 2404static int 2405do_jail_attach(struct thread *td, struct prison *pr) 2406{ 2407 struct prison *ppr; 2408 struct proc *p; 2409 struct ucred *newcred, *oldcred; 2410 int error; 2411 2412 /* 2413 * XXX: Note that there is a slight race here if two threads 2414 * in the same privileged process attempt to attach to two 2415 * different jails at the same time. It is important for 2416 * user processes not to do this, or they might end up with 2417 * a process root from one prison, but attached to the jail 2418 * of another. 2419 */ 2420 pr->pr_ref++; 2421 pr->pr_uref++; 2422 mtx_unlock(&pr->pr_mtx); 2423 2424 /* Let modules do whatever they need to prepare for attaching. */ 2425 error = osd_jail_call(pr, PR_METHOD_ATTACH, td); 2426 if (error) { 2427 prison_deref(pr, PD_DEREF | PD_DEUREF | PD_LIST_SLOCKED); 2428 return (error); 2429 } 2430 sx_sunlock(&allprison_lock); 2431 2432 /* 2433 * Reparent the newly attached process to this jail. 2434 */ 2435 ppr = td->td_ucred->cr_prison; 2436 p = td->td_proc; 2437 error = cpuset_setproc_update_set(p, pr->pr_cpuset); 2438 if (error) 2439 goto e_revert_osd; 2440 2441 vn_lock(pr->pr_root, LK_EXCLUSIVE | LK_RETRY); 2442 if ((error = change_dir(pr->pr_root, td)) != 0) 2443 goto e_unlock; 2444#ifdef MAC 2445 if ((error = mac_vnode_check_chroot(td->td_ucred, pr->pr_root))) 2446 goto e_unlock; 2447#endif 2448 VOP_UNLOCK(pr->pr_root, 0); 2449 if ((error = change_root(pr->pr_root, td))) 2450 goto e_revert_osd; 2451 2452 newcred = crget(); 2453 PROC_LOCK(p); 2454 oldcred = p->p_ucred; 2455 setsugid(p); 2456 crcopy(newcred, oldcred); 2457 newcred->cr_prison = pr; 2458 p->p_ucred = newcred; 2459 PROC_UNLOCK(p); 2460#ifdef RACCT 2461 racct_proc_ucred_changed(p, oldcred, newcred); 2462#endif 2463 crfree(oldcred); 2464 prison_deref(ppr, PD_DEREF | PD_DEUREF); 2465 return (0); 2466 e_unlock: 2467 VOP_UNLOCK(pr->pr_root, 0); 2468 e_revert_osd: 2469 /* Tell modules this thread is still in its old jail after all. */ 2470 (void)osd_jail_call(ppr, PR_METHOD_ATTACH, td); 2471 prison_deref(pr, PD_DEREF | PD_DEUREF); 2472 return (error); 2473} 2474 2475 2476/* 2477 * Returns a locked prison instance, or NULL on failure. 2478 */ 2479struct prison * 2480prison_find(int prid) 2481{ 2482 struct prison *pr; 2483 2484 sx_assert(&allprison_lock, SX_LOCKED); 2485 TAILQ_FOREACH(pr, &allprison, pr_list) { 2486 if (pr->pr_id == prid) { 2487 mtx_lock(&pr->pr_mtx); 2488 if (pr->pr_ref > 0) 2489 return (pr); 2490 mtx_unlock(&pr->pr_mtx); 2491 } 2492 } 2493 return (NULL); 2494} 2495 2496/* 2497 * Find a prison that is a descendant of mypr. Returns a locked prison or NULL. 2498 */ 2499struct prison * 2500prison_find_child(struct prison *mypr, int prid) 2501{ 2502 struct prison *pr; 2503 int descend; 2504 2505 sx_assert(&allprison_lock, SX_LOCKED); 2506 FOREACH_PRISON_DESCENDANT(mypr, pr, descend) { 2507 if (pr->pr_id == prid) { 2508 mtx_lock(&pr->pr_mtx); 2509 if (pr->pr_ref > 0) 2510 return (pr); 2511 mtx_unlock(&pr->pr_mtx); 2512 } 2513 } 2514 return (NULL); 2515} 2516 2517/* 2518 * Look for the name relative to mypr. Returns a locked prison or NULL. 2519 */ 2520struct prison * 2521prison_find_name(struct prison *mypr, const char *name) 2522{ 2523 struct prison *pr, *deadpr; 2524 size_t mylen; 2525 int descend; 2526 2527 sx_assert(&allprison_lock, SX_LOCKED); 2528 mylen = (mypr == &prison0) ? 0 : strlen(mypr->pr_name) + 1; 2529 again: 2530 deadpr = NULL; 2531 FOREACH_PRISON_DESCENDANT(mypr, pr, descend) { 2532 if (!strcmp(pr->pr_name + mylen, name)) { 2533 mtx_lock(&pr->pr_mtx); 2534 if (pr->pr_ref > 0) { 2535 if (pr->pr_uref > 0) 2536 return (pr); 2537 deadpr = pr; 2538 } 2539 mtx_unlock(&pr->pr_mtx); 2540 } 2541 } 2542 /* There was no valid prison - perhaps there was a dying one. */ 2543 if (deadpr != NULL) { 2544 mtx_lock(&deadpr->pr_mtx); 2545 if (deadpr->pr_ref == 0) { 2546 mtx_unlock(&deadpr->pr_mtx); 2547 goto again; 2548 } 2549 } 2550 return (deadpr); 2551} 2552 2553/* 2554 * See if a prison has the specific flag set. 2555 */ 2556int 2557prison_flag(struct ucred *cred, unsigned flag) 2558{ 2559 2560 /* This is an atomic read, so no locking is necessary. */ 2561 return (cred->cr_prison->pr_flags & flag); 2562} 2563 2564int 2565prison_allow(struct ucred *cred, unsigned flag) 2566{ 2567 2568 /* This is an atomic read, so no locking is necessary. */ 2569 return (cred->cr_prison->pr_allow & flag); 2570} 2571 2572/* 2573 * Remove a prison reference. If that was the last reference, remove the 2574 * prison itself - but not in this context in case there are locks held. 2575 */ 2576void 2577prison_free_locked(struct prison *pr) 2578{ 2579 2580 mtx_assert(&pr->pr_mtx, MA_OWNED); 2581 pr->pr_ref--; 2582 if (pr->pr_ref == 0) { 2583 mtx_unlock(&pr->pr_mtx); 2584 TASK_INIT(&pr->pr_task, 0, prison_complete, pr); 2585 taskqueue_enqueue(taskqueue_thread, &pr->pr_task); 2586 return; 2587 } 2588 mtx_unlock(&pr->pr_mtx); 2589} 2590 2591void 2592prison_free(struct prison *pr) 2593{ 2594 2595 mtx_lock(&pr->pr_mtx); 2596 prison_free_locked(pr); 2597} 2598 2599static void 2600prison_complete(void *context, int pending) 2601{ 2602 2603 prison_deref((struct prison *)context, 0); 2604} 2605 2606/* 2607 * Remove a prison reference (usually). This internal version assumes no 2608 * mutexes are held, except perhaps the prison itself. If there are no more 2609 * references, release and delist the prison. On completion, the prison lock 2610 * and the allprison lock are both unlocked. 2611 */ 2612static void 2613prison_deref(struct prison *pr, int flags) 2614{ 2615 struct prison *ppr, *tpr; 2616 2617 if (!(flags & PD_LOCKED)) 2618 mtx_lock(&pr->pr_mtx); 2619 for (;;) { 2620 if (flags & PD_DEUREF) { 2621 pr->pr_uref--; 2622 KASSERT(prison0.pr_uref != 0, ("prison0 pr_uref=0")); 2623 } 2624 if (flags & PD_DEREF) 2625 pr->pr_ref--; 2626 /* If the prison still has references, nothing else to do. */ 2627 if (pr->pr_ref > 0) { 2628 mtx_unlock(&pr->pr_mtx); 2629 if (flags & PD_LIST_SLOCKED) 2630 sx_sunlock(&allprison_lock); 2631 else if (flags & PD_LIST_XLOCKED) 2632 sx_xunlock(&allprison_lock); 2633 return; 2634 } 2635 2636 mtx_unlock(&pr->pr_mtx); 2637 if (flags & PD_LIST_SLOCKED) { 2638 if (!sx_try_upgrade(&allprison_lock)) { 2639 sx_sunlock(&allprison_lock); 2640 sx_xlock(&allprison_lock); 2641 } 2642 } else if (!(flags & PD_LIST_XLOCKED)) 2643 sx_xlock(&allprison_lock); 2644 2645 TAILQ_REMOVE(&allprison, pr, pr_list); 2646 LIST_REMOVE(pr, pr_sibling); 2647 ppr = pr->pr_parent; 2648 for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent) 2649 tpr->pr_childcount--; 2650 sx_xunlock(&allprison_lock); 2651 2652#ifdef VIMAGE 2653 if (pr->pr_vnet != ppr->pr_vnet) 2654 vnet_destroy(pr->pr_vnet); 2655#endif 2656 if (pr->pr_root != NULL) 2657 vrele(pr->pr_root); 2658 mtx_destroy(&pr->pr_mtx); 2659#ifdef INET 2660 free(pr->pr_ip4, M_PRISON); 2661#endif 2662#ifdef INET6 2663 free(pr->pr_ip6, M_PRISON); 2664#endif 2665 if (pr->pr_cpuset != NULL) 2666 cpuset_rel(pr->pr_cpuset); 2667 osd_jail_exit(pr); 2668#ifdef RACCT 2669 if (racct_enable) 2670 prison_racct_detach(pr); 2671#endif 2672 free(pr, M_PRISON); 2673 2674 /* Removing a prison frees a reference on its parent. */ 2675 pr = ppr; 2676 mtx_lock(&pr->pr_mtx); 2677 flags = PD_DEREF | PD_DEUREF; 2678 } 2679} 2680 2681void 2682prison_hold_locked(struct prison *pr) 2683{ 2684 2685 mtx_assert(&pr->pr_mtx, MA_OWNED); 2686 KASSERT(pr->pr_ref > 0, 2687 ("Trying to hold dead prison (jid=%d).", pr->pr_id)); 2688 pr->pr_ref++; 2689} 2690 2691void 2692prison_hold(struct prison *pr) 2693{ 2694 2695 mtx_lock(&pr->pr_mtx); 2696 prison_hold_locked(pr); 2697 mtx_unlock(&pr->pr_mtx); 2698} 2699 2700void 2701prison_proc_hold(struct prison *pr) 2702{ 2703 2704 mtx_lock(&pr->pr_mtx); 2705 KASSERT(pr->pr_uref > 0, 2706 ("Cannot add a process to a non-alive prison (jid=%d)", pr->pr_id)); 2707 pr->pr_uref++; 2708 mtx_unlock(&pr->pr_mtx); 2709} 2710 2711void 2712prison_proc_free(struct prison *pr) 2713{ 2714 2715 mtx_lock(&pr->pr_mtx); 2716 KASSERT(pr->pr_uref > 0, 2717 ("Trying to kill a process in a dead prison (jid=%d)", pr->pr_id)); 2718 prison_deref(pr, PD_DEUREF | PD_LOCKED); 2719} 2720 2721 2722#ifdef INET 2723/* 2724 * Restrict a prison's IP address list with its parent's, possibly replacing 2725 * it. Return true if the replacement buffer was used (or would have been). 2726 */ 2727static int 2728prison_restrict_ip4(struct prison *pr, struct in_addr *newip4) 2729{ 2730 int ii, ij, used; 2731 struct prison *ppr; 2732 2733 ppr = pr->pr_parent; 2734 if (!(pr->pr_flags & PR_IP4_USER)) { 2735 /* This has no user settings, so just copy the parent's list. */ 2736 if (pr->pr_ip4s < ppr->pr_ip4s) { 2737 /* 2738 * There's no room for the parent's list. Use the 2739 * new list buffer, which is assumed to be big enough 2740 * (if it was passed). If there's no buffer, try to 2741 * allocate one. 2742 */ 2743 used = 1; 2744 if (newip4 == NULL) { 2745 newip4 = malloc(ppr->pr_ip4s * sizeof(*newip4), 2746 M_PRISON, M_NOWAIT); 2747 if (newip4 != NULL) 2748 used = 0; 2749 } 2750 if (newip4 != NULL) { 2751 bcopy(ppr->pr_ip4, newip4, 2752 ppr->pr_ip4s * sizeof(*newip4)); 2753 free(pr->pr_ip4, M_PRISON); 2754 pr->pr_ip4 = newip4; 2755 pr->pr_ip4s = ppr->pr_ip4s; 2756 } 2757 return (used); 2758 } 2759 pr->pr_ip4s = ppr->pr_ip4s; 2760 if (pr->pr_ip4s > 0) 2761 bcopy(ppr->pr_ip4, pr->pr_ip4, 2762 pr->pr_ip4s * sizeof(*newip4)); 2763 else if (pr->pr_ip4 != NULL) { 2764 free(pr->pr_ip4, M_PRISON); 2765 pr->pr_ip4 = NULL; 2766 } 2767 } else if (pr->pr_ip4s > 0) { 2768 /* Remove addresses that aren't in the parent. */ 2769 for (ij = 0; ij < ppr->pr_ip4s; ij++) 2770 if (pr->pr_ip4[0].s_addr == ppr->pr_ip4[ij].s_addr) 2771 break; 2772 if (ij < ppr->pr_ip4s) 2773 ii = 1; 2774 else { 2775 bcopy(pr->pr_ip4 + 1, pr->pr_ip4, 2776 --pr->pr_ip4s * sizeof(*pr->pr_ip4)); 2777 ii = 0; 2778 } 2779 for (ij = 1; ii < pr->pr_ip4s; ) { 2780 if (pr->pr_ip4[ii].s_addr == ppr->pr_ip4[0].s_addr) { 2781 ii++; 2782 continue; 2783 } 2784 switch (ij >= ppr->pr_ip4s ? -1 : 2785 qcmp_v4(&pr->pr_ip4[ii], &ppr->pr_ip4[ij])) { 2786 case -1: 2787 bcopy(pr->pr_ip4 + ii + 1, pr->pr_ip4 + ii, 2788 (--pr->pr_ip4s - ii) * sizeof(*pr->pr_ip4)); 2789 break; 2790 case 0: 2791 ii++; 2792 ij++; 2793 break; 2794 case 1: 2795 ij++; 2796 break; 2797 } 2798 } 2799 if (pr->pr_ip4s == 0) { 2800 pr->pr_flags |= PR_IP4_DISABLE; 2801 free(pr->pr_ip4, M_PRISON); 2802 pr->pr_ip4 = NULL; 2803 } 2804 } 2805 return (0); 2806} 2807 2808/* 2809 * Pass back primary IPv4 address of this jail. 2810 * 2811 * If not restricted return success but do not alter the address. Caller has 2812 * to make sure to initialize it correctly (e.g. INADDR_ANY). 2813 * 2814 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv4. 2815 * Address returned in NBO. 2816 */ 2817int 2818prison_get_ip4(struct ucred *cred, struct in_addr *ia) 2819{ 2820 struct prison *pr; 2821 2822 KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); 2823 KASSERT(ia != NULL, ("%s: ia is NULL", __func__)); 2824 2825 pr = cred->cr_prison; 2826 if (!(pr->pr_flags & PR_IP4)) 2827 return (0); 2828 mtx_lock(&pr->pr_mtx); 2829 if (!(pr->pr_flags & PR_IP4)) { 2830 mtx_unlock(&pr->pr_mtx); 2831 return (0); 2832 } 2833 if (pr->pr_ip4 == NULL) { 2834 mtx_unlock(&pr->pr_mtx); 2835 return (EAFNOSUPPORT); 2836 } 2837 2838 ia->s_addr = pr->pr_ip4[0].s_addr; 2839 mtx_unlock(&pr->pr_mtx); 2840 return (0); 2841} 2842 2843/* 2844 * Return 1 if we should do proper source address selection or are not jailed. 2845 * We will return 0 if we should bypass source address selection in favour 2846 * of the primary jail IPv4 address. Only in this case *ia will be updated and 2847 * returned in NBO. 2848 * Return EAFNOSUPPORT, in case this jail does not allow IPv4. 2849 */ 2850int 2851prison_saddrsel_ip4(struct ucred *cred, struct in_addr *ia) 2852{ 2853 struct prison *pr; 2854 struct in_addr lia; 2855 int error; 2856 2857 KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); 2858 KASSERT(ia != NULL, ("%s: ia is NULL", __func__)); 2859 2860 if (!jailed(cred)) 2861 return (1); 2862 2863 pr = cred->cr_prison; 2864 if (pr->pr_flags & PR_IP4_SADDRSEL) 2865 return (1); 2866 2867 lia.s_addr = INADDR_ANY; 2868 error = prison_get_ip4(cred, &lia); 2869 if (error) 2870 return (error); 2871 if (lia.s_addr == INADDR_ANY) 2872 return (1); 2873 2874 ia->s_addr = lia.s_addr; 2875 return (0); 2876} 2877 2878/* 2879 * Return true if pr1 and pr2 have the same IPv4 address restrictions. 2880 */ 2881int 2882prison_equal_ip4(struct prison *pr1, struct prison *pr2) 2883{ 2884 2885 if (pr1 == pr2) 2886 return (1); 2887 2888 /* 2889 * No need to lock since the PR_IP4_USER flag can't be altered for 2890 * existing prisons. 2891 */ 2892 while (pr1 != &prison0 && 2893#ifdef VIMAGE 2894 !(pr1->pr_flags & PR_VNET) && 2895#endif 2896 !(pr1->pr_flags & PR_IP4_USER)) 2897 pr1 = pr1->pr_parent; 2898 while (pr2 != &prison0 && 2899#ifdef VIMAGE 2900 !(pr2->pr_flags & PR_VNET) && 2901#endif 2902 !(pr2->pr_flags & PR_IP4_USER)) 2903 pr2 = pr2->pr_parent; 2904 return (pr1 == pr2); 2905} 2906 2907/* 2908 * Make sure our (source) address is set to something meaningful to this 2909 * jail. 2910 * 2911 * Returns 0 if jail doesn't restrict IPv4 or if address belongs to jail, 2912 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail 2913 * doesn't allow IPv4. Address passed in in NBO and returned in NBO. 2914 */ 2915int 2916prison_local_ip4(struct ucred *cred, struct in_addr *ia) 2917{ 2918 struct prison *pr; 2919 struct in_addr ia0; 2920 int error; 2921 2922 KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); 2923 KASSERT(ia != NULL, ("%s: ia is NULL", __func__)); 2924 2925 pr = cred->cr_prison; 2926 if (!(pr->pr_flags & PR_IP4)) 2927 return (0); 2928 mtx_lock(&pr->pr_mtx); 2929 if (!(pr->pr_flags & PR_IP4)) { 2930 mtx_unlock(&pr->pr_mtx); 2931 return (0); 2932 } 2933 if (pr->pr_ip4 == NULL) { 2934 mtx_unlock(&pr->pr_mtx); 2935 return (EAFNOSUPPORT); 2936 } 2937 2938 ia0.s_addr = ntohl(ia->s_addr); 2939 if (ia0.s_addr == INADDR_LOOPBACK) { 2940 ia->s_addr = pr->pr_ip4[0].s_addr; 2941 mtx_unlock(&pr->pr_mtx); 2942 return (0); 2943 } 2944 2945 if (ia0.s_addr == INADDR_ANY) { 2946 /* 2947 * In case there is only 1 IPv4 address, bind directly. 2948 */ 2949 if (pr->pr_ip4s == 1) 2950 ia->s_addr = pr->pr_ip4[0].s_addr; 2951 mtx_unlock(&pr->pr_mtx); 2952 return (0); 2953 } 2954 2955 error = _prison_check_ip4(pr, ia); 2956 mtx_unlock(&pr->pr_mtx); 2957 return (error); 2958} 2959 2960/* 2961 * Rewrite destination address in case we will connect to loopback address. 2962 * 2963 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv4. 2964 * Address passed in in NBO and returned in NBO. 2965 */ 2966int 2967prison_remote_ip4(struct ucred *cred, struct in_addr *ia) 2968{ 2969 struct prison *pr; 2970 2971 KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); 2972 KASSERT(ia != NULL, ("%s: ia is NULL", __func__)); 2973 2974 pr = cred->cr_prison; 2975 if (!(pr->pr_flags & PR_IP4)) 2976 return (0); 2977 mtx_lock(&pr->pr_mtx); 2978 if (!(pr->pr_flags & PR_IP4)) { 2979 mtx_unlock(&pr->pr_mtx); 2980 return (0); 2981 } 2982 if (pr->pr_ip4 == NULL) { 2983 mtx_unlock(&pr->pr_mtx); 2984 return (EAFNOSUPPORT); 2985 } 2986 2987 if (ntohl(ia->s_addr) == INADDR_LOOPBACK) { 2988 ia->s_addr = pr->pr_ip4[0].s_addr; 2989 mtx_unlock(&pr->pr_mtx); 2990 return (0); 2991 } 2992 2993 /* 2994 * Return success because nothing had to be changed. 2995 */ 2996 mtx_unlock(&pr->pr_mtx); 2997 return (0); 2998} 2999 3000/* 3001 * Check if given address belongs to the jail referenced by cred/prison. 3002 * 3003 * Returns 0 if jail doesn't restrict IPv4 or if address belongs to jail, 3004 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail 3005 * doesn't allow IPv4. Address passed in in NBO. 3006 */ 3007static int 3008_prison_check_ip4(struct prison *pr, struct in_addr *ia) 3009{ 3010 int i, a, z, d; 3011 3012 /* 3013 * Check the primary IP. 3014 */ 3015 if (pr->pr_ip4[0].s_addr == ia->s_addr) 3016 return (0); 3017 3018 /* 3019 * All the other IPs are sorted so we can do a binary search. 3020 */ 3021 a = 0; 3022 z = pr->pr_ip4s - 2; 3023 while (a <= z) { 3024 i = (a + z) / 2; 3025 d = qcmp_v4(&pr->pr_ip4[i+1], ia); 3026 if (d > 0) 3027 z = i - 1; 3028 else if (d < 0) 3029 a = i + 1; 3030 else 3031 return (0); 3032 } 3033 3034 return (EADDRNOTAVAIL); 3035} 3036 3037int 3038prison_check_ip4(struct ucred *cred, struct in_addr *ia) 3039{ 3040 struct prison *pr; 3041 int error; 3042 3043 KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); 3044 KASSERT(ia != NULL, ("%s: ia is NULL", __func__)); 3045 3046 pr = cred->cr_prison; 3047 if (!(pr->pr_flags & PR_IP4)) 3048 return (0); 3049 mtx_lock(&pr->pr_mtx); 3050 if (!(pr->pr_flags & PR_IP4)) { 3051 mtx_unlock(&pr->pr_mtx); 3052 return (0); 3053 } 3054 if (pr->pr_ip4 == NULL) { 3055 mtx_unlock(&pr->pr_mtx); 3056 return (EAFNOSUPPORT); 3057 } 3058 3059 error = _prison_check_ip4(pr, ia); 3060 mtx_unlock(&pr->pr_mtx); 3061 return (error); 3062} 3063#endif 3064 3065#ifdef INET6 3066static int 3067prison_restrict_ip6(struct prison *pr, struct in6_addr *newip6) 3068{ 3069 int ii, ij, used; 3070 struct prison *ppr; 3071 3072 ppr = pr->pr_parent; 3073 if (!(pr->pr_flags & PR_IP6_USER)) { 3074 /* This has no user settings, so just copy the parent's list. */ 3075 if (pr->pr_ip6s < ppr->pr_ip6s) { 3076 /* 3077 * There's no room for the parent's list. Use the 3078 * new list buffer, which is assumed to be big enough 3079 * (if it was passed). If there's no buffer, try to 3080 * allocate one. 3081 */ 3082 used = 1; 3083 if (newip6 == NULL) { 3084 newip6 = malloc(ppr->pr_ip6s * sizeof(*newip6), 3085 M_PRISON, M_NOWAIT); 3086 if (newip6 != NULL) 3087 used = 0; 3088 } 3089 if (newip6 != NULL) { 3090 bcopy(ppr->pr_ip6, newip6, 3091 ppr->pr_ip6s * sizeof(*newip6)); 3092 free(pr->pr_ip6, M_PRISON); 3093 pr->pr_ip6 = newip6; 3094 pr->pr_ip6s = ppr->pr_ip6s; 3095 } 3096 return (used); 3097 } 3098 pr->pr_ip6s = ppr->pr_ip6s; 3099 if (pr->pr_ip6s > 0) 3100 bcopy(ppr->pr_ip6, pr->pr_ip6, 3101 pr->pr_ip6s * sizeof(*newip6)); 3102 else if (pr->pr_ip6 != NULL) { 3103 free(pr->pr_ip6, M_PRISON); 3104 pr->pr_ip6 = NULL; 3105 } 3106 } else if (pr->pr_ip6s > 0) { 3107 /* Remove addresses that aren't in the parent. */ 3108 for (ij = 0; ij < ppr->pr_ip6s; ij++) 3109 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[0], 3110 &ppr->pr_ip6[ij])) 3111 break; 3112 if (ij < ppr->pr_ip6s) 3113 ii = 1; 3114 else { 3115 bcopy(pr->pr_ip6 + 1, pr->pr_ip6, 3116 --pr->pr_ip6s * sizeof(*pr->pr_ip6)); 3117 ii = 0; 3118 } 3119 for (ij = 1; ii < pr->pr_ip6s; ) { 3120 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[ii], 3121 &ppr->pr_ip6[0])) { 3122 ii++; 3123 continue; 3124 } 3125 switch (ij >= ppr->pr_ip6s ? -1 : 3126 qcmp_v6(&pr->pr_ip6[ii], &ppr->pr_ip6[ij])) { 3127 case -1: 3128 bcopy(pr->pr_ip6 + ii + 1, pr->pr_ip6 + ii, 3129 (--pr->pr_ip6s - ii) * sizeof(*pr->pr_ip6)); 3130 break; 3131 case 0: 3132 ii++; 3133 ij++; 3134 break; 3135 case 1: 3136 ij++; 3137 break; 3138 } 3139 } 3140 if (pr->pr_ip6s == 0) { 3141 pr->pr_flags |= PR_IP6_DISABLE; 3142 free(pr->pr_ip6, M_PRISON); 3143 pr->pr_ip6 = NULL; 3144 } 3145 } 3146 return 0; 3147} 3148 3149/* 3150 * Pass back primary IPv6 address for this jail. 3151 * 3152 * If not restricted return success but do not alter the address. Caller has 3153 * to make sure to initialize it correctly (e.g. IN6ADDR_ANY_INIT). 3154 * 3155 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv6. 3156 */ 3157int 3158prison_get_ip6(struct ucred *cred, struct in6_addr *ia6) 3159{ 3160 struct prison *pr; 3161 3162 KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); 3163 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__)); 3164 3165 pr = cred->cr_prison; 3166 if (!(pr->pr_flags & PR_IP6)) 3167 return (0); 3168 mtx_lock(&pr->pr_mtx); 3169 if (!(pr->pr_flags & PR_IP6)) { 3170 mtx_unlock(&pr->pr_mtx); 3171 return (0); 3172 } 3173 if (pr->pr_ip6 == NULL) { 3174 mtx_unlock(&pr->pr_mtx); 3175 return (EAFNOSUPPORT); 3176 } 3177 3178 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr)); 3179 mtx_unlock(&pr->pr_mtx); 3180 return (0); 3181} 3182 3183/* 3184 * Return 1 if we should do proper source address selection or are not jailed. 3185 * We will return 0 if we should bypass source address selection in favour 3186 * of the primary jail IPv6 address. Only in this case *ia will be updated and 3187 * returned in NBO. 3188 * Return EAFNOSUPPORT, in case this jail does not allow IPv6. 3189 */ 3190int 3191prison_saddrsel_ip6(struct ucred *cred, struct in6_addr *ia6) 3192{ 3193 struct prison *pr; 3194 struct in6_addr lia6; 3195 int error; 3196 3197 KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); 3198 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__)); 3199 3200 if (!jailed(cred)) 3201 return (1); 3202 3203 pr = cred->cr_prison; 3204 if (pr->pr_flags & PR_IP6_SADDRSEL) 3205 return (1); 3206 3207 lia6 = in6addr_any; 3208 error = prison_get_ip6(cred, &lia6); 3209 if (error) 3210 return (error); 3211 if (IN6_IS_ADDR_UNSPECIFIED(&lia6)) 3212 return (1); 3213 3214 bcopy(&lia6, ia6, sizeof(struct in6_addr)); 3215 return (0); 3216} 3217 3218/* 3219 * Return true if pr1 and pr2 have the same IPv6 address restrictions. 3220 */ 3221int 3222prison_equal_ip6(struct prison *pr1, struct prison *pr2) 3223{ 3224 3225 if (pr1 == pr2) 3226 return (1); 3227 3228 while (pr1 != &prison0 && 3229#ifdef VIMAGE 3230 !(pr1->pr_flags & PR_VNET) && 3231#endif 3232 !(pr1->pr_flags & PR_IP6_USER)) 3233 pr1 = pr1->pr_parent; 3234 while (pr2 != &prison0 && 3235#ifdef VIMAGE 3236 !(pr2->pr_flags & PR_VNET) && 3237#endif 3238 !(pr2->pr_flags & PR_IP6_USER)) 3239 pr2 = pr2->pr_parent; 3240 return (pr1 == pr2); 3241} 3242 3243/* 3244 * Make sure our (source) address is set to something meaningful to this jail. 3245 * 3246 * v6only should be set based on (inp->inp_flags & IN6P_IPV6_V6ONLY != 0) 3247 * when needed while binding. 3248 * 3249 * Returns 0 if jail doesn't restrict IPv6 or if address belongs to jail, 3250 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail 3251 * doesn't allow IPv6. 3252 */ 3253int 3254prison_local_ip6(struct ucred *cred, struct in6_addr *ia6, int v6only) 3255{ 3256 struct prison *pr; 3257 int error; 3258 3259 KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); 3260 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__)); 3261 3262 pr = cred->cr_prison; 3263 if (!(pr->pr_flags & PR_IP6)) 3264 return (0); 3265 mtx_lock(&pr->pr_mtx); 3266 if (!(pr->pr_flags & PR_IP6)) { 3267 mtx_unlock(&pr->pr_mtx); 3268 return (0); 3269 } 3270 if (pr->pr_ip6 == NULL) { 3271 mtx_unlock(&pr->pr_mtx); 3272 return (EAFNOSUPPORT); 3273 } 3274 3275 if (IN6_IS_ADDR_LOOPBACK(ia6)) { 3276 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr)); 3277 mtx_unlock(&pr->pr_mtx); 3278 return (0); 3279 } 3280 3281 if (IN6_IS_ADDR_UNSPECIFIED(ia6)) { 3282 /* 3283 * In case there is only 1 IPv6 address, and v6only is true, 3284 * then bind directly. 3285 */ 3286 if (v6only != 0 && pr->pr_ip6s == 1) 3287 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr)); 3288 mtx_unlock(&pr->pr_mtx); 3289 return (0); 3290 } 3291 3292 error = _prison_check_ip6(pr, ia6); 3293 mtx_unlock(&pr->pr_mtx); 3294 return (error); 3295} 3296 3297/* 3298 * Rewrite destination address in case we will connect to loopback address. 3299 * 3300 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv6. 3301 */ 3302int 3303prison_remote_ip6(struct ucred *cred, struct in6_addr *ia6) 3304{ 3305 struct prison *pr; 3306 3307 KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); 3308 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__)); 3309 3310 pr = cred->cr_prison; 3311 if (!(pr->pr_flags & PR_IP6)) 3312 return (0); 3313 mtx_lock(&pr->pr_mtx); 3314 if (!(pr->pr_flags & PR_IP6)) { 3315 mtx_unlock(&pr->pr_mtx); 3316 return (0); 3317 } 3318 if (pr->pr_ip6 == NULL) { 3319 mtx_unlock(&pr->pr_mtx); 3320 return (EAFNOSUPPORT); 3321 } 3322 3323 if (IN6_IS_ADDR_LOOPBACK(ia6)) { 3324 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr)); 3325 mtx_unlock(&pr->pr_mtx); 3326 return (0); 3327 } 3328 3329 /* 3330 * Return success because nothing had to be changed. 3331 */ 3332 mtx_unlock(&pr->pr_mtx); 3333 return (0); 3334} 3335 3336/* 3337 * Check if given address belongs to the jail referenced by cred/prison. 3338 * 3339 * Returns 0 if jail doesn't restrict IPv6 or if address belongs to jail, 3340 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail 3341 * doesn't allow IPv6. 3342 */ 3343static int 3344_prison_check_ip6(struct prison *pr, struct in6_addr *ia6) 3345{ 3346 int i, a, z, d; 3347 3348 /* 3349 * Check the primary IP. 3350 */ 3351 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[0], ia6)) 3352 return (0); 3353 3354 /* 3355 * All the other IPs are sorted so we can do a binary search. 3356 */ 3357 a = 0; 3358 z = pr->pr_ip6s - 2; 3359 while (a <= z) { 3360 i = (a + z) / 2; 3361 d = qcmp_v6(&pr->pr_ip6[i+1], ia6); 3362 if (d > 0) 3363 z = i - 1; 3364 else if (d < 0) 3365 a = i + 1; 3366 else 3367 return (0); 3368 } 3369 3370 return (EADDRNOTAVAIL); 3371} 3372 3373int 3374prison_check_ip6(struct ucred *cred, struct in6_addr *ia6) 3375{ 3376 struct prison *pr; 3377 int error; 3378 3379 KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); 3380 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__)); 3381 3382 pr = cred->cr_prison; 3383 if (!(pr->pr_flags & PR_IP6)) 3384 return (0); 3385 mtx_lock(&pr->pr_mtx); 3386 if (!(pr->pr_flags & PR_IP6)) { 3387 mtx_unlock(&pr->pr_mtx); 3388 return (0); 3389 } 3390 if (pr->pr_ip6 == NULL) { 3391 mtx_unlock(&pr->pr_mtx); 3392 return (EAFNOSUPPORT); 3393 } 3394 3395 error = _prison_check_ip6(pr, ia6); 3396 mtx_unlock(&pr->pr_mtx); 3397 return (error); 3398} 3399#endif 3400 3401/* 3402 * Check if a jail supports the given address family. 3403 * 3404 * Returns 0 if not jailed or the address family is supported, EAFNOSUPPORT 3405 * if not. 3406 */ 3407int 3408prison_check_af(struct ucred *cred, int af) 3409{ 3410 struct prison *pr; 3411 int error; 3412 3413 KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); 3414 3415 pr = cred->cr_prison; 3416#ifdef VIMAGE 3417 /* Prisons with their own network stack are not limited. */ 3418 if (prison_owns_vnet(cred)) 3419 return (0); 3420#endif 3421 3422 error = 0; 3423 switch (af) 3424 { 3425#ifdef INET 3426 case AF_INET: 3427 if (pr->pr_flags & PR_IP4) 3428 { 3429 mtx_lock(&pr->pr_mtx); 3430 if ((pr->pr_flags & PR_IP4) && pr->pr_ip4 == NULL) 3431 error = EAFNOSUPPORT; 3432 mtx_unlock(&pr->pr_mtx); 3433 } 3434 break; 3435#endif 3436#ifdef INET6 3437 case AF_INET6: 3438 if (pr->pr_flags & PR_IP6) 3439 { 3440 mtx_lock(&pr->pr_mtx); 3441 if ((pr->pr_flags & PR_IP6) && pr->pr_ip6 == NULL) 3442 error = EAFNOSUPPORT; 3443 mtx_unlock(&pr->pr_mtx); 3444 } 3445 break; 3446#endif 3447 case AF_LOCAL: 3448 case AF_ROUTE: 3449 break; 3450 default: 3451 if (!(pr->pr_allow & PR_ALLOW_SOCKET_AF)) 3452 error = EAFNOSUPPORT; 3453 } 3454 return (error); 3455} 3456 3457/* 3458 * Check if given address belongs to the jail referenced by cred (wrapper to 3459 * prison_check_ip[46]). 3460 * 3461 * Returns 0 if jail doesn't restrict the address family or if address belongs 3462 * to jail, EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if 3463 * the jail doesn't allow the address family. IPv4 Address passed in in NBO. 3464 */ 3465int 3466prison_if(struct ucred *cred, struct sockaddr *sa) 3467{ 3468#ifdef INET 3469 struct sockaddr_in *sai; 3470#endif 3471#ifdef INET6 3472 struct sockaddr_in6 *sai6; 3473#endif 3474 int error; 3475 3476 KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); 3477 KASSERT(sa != NULL, ("%s: sa is NULL", __func__)); 3478 3479#ifdef VIMAGE 3480 if (prison_owns_vnet(cred)) 3481 return (0); 3482#endif 3483 3484 error = 0; 3485 switch (sa->sa_family) 3486 { 3487#ifdef INET 3488 case AF_INET: 3489 sai = (struct sockaddr_in *)sa; 3490 error = prison_check_ip4(cred, &sai->sin_addr); 3491 break; 3492#endif 3493#ifdef INET6 3494 case AF_INET6: 3495 sai6 = (struct sockaddr_in6 *)sa; 3496 error = prison_check_ip6(cred, &sai6->sin6_addr); 3497 break; 3498#endif 3499 default: 3500 if (!(cred->cr_prison->pr_allow & PR_ALLOW_SOCKET_AF)) 3501 error = EAFNOSUPPORT; 3502 } 3503 return (error); 3504} 3505 3506/* 3507 * Return 0 if jails permit p1 to frob p2, otherwise ESRCH. 3508 */ 3509int 3510prison_check(struct ucred *cred1, struct ucred *cred2) 3511{ 3512 3513 return ((cred1->cr_prison == cred2->cr_prison || 3514 prison_ischild(cred1->cr_prison, cred2->cr_prison)) ? 0 : ESRCH); 3515} 3516 3517/* 3518 * Return 1 if p2 is a child of p1, otherwise 0. 3519 */ 3520int 3521prison_ischild(struct prison *pr1, struct prison *pr2) 3522{ 3523 3524 for (pr2 = pr2->pr_parent; pr2 != NULL; pr2 = pr2->pr_parent) 3525 if (pr1 == pr2) 3526 return (1); 3527 return (0); 3528} 3529 3530/* 3531 * Return 1 if the passed credential is in a jail, otherwise 0. 3532 */ 3533int 3534jailed(struct ucred *cred) 3535{ 3536 3537 return (cred->cr_prison != &prison0); 3538} 3539 3540/* 3541 * Return 1 if the passed credential is in a jail and that jail does not 3542 * have its own virtual network stack, otherwise 0. 3543 */ 3544int 3545jailed_without_vnet(struct ucred *cred) 3546{ 3547 3548 if (!jailed(cred)) 3549 return (0); 3550#ifdef VIMAGE 3551 if (prison_owns_vnet(cred)) 3552 return (0); 3553#endif 3554 3555 return (1); 3556} 3557 3558/* 3559 * Return the correct hostname (domainname, et al) for the passed credential. 3560 */ 3561void 3562getcredhostname(struct ucred *cred, char *buf, size_t size) 3563{ 3564 struct prison *pr; 3565 3566 /* 3567 * A NULL credential can be used to shortcut to the physical 3568 * system's hostname. 3569 */ 3570 pr = (cred != NULL) ? cred->cr_prison : &prison0; 3571 mtx_lock(&pr->pr_mtx); 3572 strlcpy(buf, pr->pr_hostname, size); 3573 mtx_unlock(&pr->pr_mtx); 3574} 3575 3576void 3577getcreddomainname(struct ucred *cred, char *buf, size_t size) 3578{ 3579 3580 mtx_lock(&cred->cr_prison->pr_mtx); 3581 strlcpy(buf, cred->cr_prison->pr_domainname, size); 3582 mtx_unlock(&cred->cr_prison->pr_mtx); 3583} 3584 3585void 3586getcredhostuuid(struct ucred *cred, char *buf, size_t size) 3587{ 3588 3589 mtx_lock(&cred->cr_prison->pr_mtx); 3590 strlcpy(buf, cred->cr_prison->pr_hostuuid, size); 3591 mtx_unlock(&cred->cr_prison->pr_mtx); 3592} 3593 3594void 3595getcredhostid(struct ucred *cred, unsigned long *hostid) 3596{ 3597 3598 mtx_lock(&cred->cr_prison->pr_mtx); 3599 *hostid = cred->cr_prison->pr_hostid; 3600 mtx_unlock(&cred->cr_prison->pr_mtx); 3601} 3602 3603#ifdef VIMAGE 3604/* 3605 * Determine whether the prison represented by cred owns 3606 * its vnet rather than having it inherited. 3607 * 3608 * Returns 1 in case the prison owns the vnet, 0 otherwise. 3609 */ 3610int 3611prison_owns_vnet(struct ucred *cred) 3612{ 3613 3614 /* 3615 * vnets cannot be added/removed after jail creation, 3616 * so no need to lock here. 3617 */ 3618 return (cred->cr_prison->pr_flags & PR_VNET ? 1 : 0); 3619} 3620#endif 3621 3622/* 3623 * Determine whether the subject represented by cred can "see" 3624 * status of a mount point. 3625 * Returns: 0 for permitted, ENOENT otherwise. 3626 * XXX: This function should be called cr_canseemount() and should be 3627 * placed in kern_prot.c. 3628 */ 3629int 3630prison_canseemount(struct ucred *cred, struct mount *mp) 3631{ 3632 struct prison *pr; 3633 struct statfs *sp; 3634 size_t len; 3635 3636 pr = cred->cr_prison; 3637 if (pr->pr_enforce_statfs == 0) 3638 return (0); 3639 if (pr->pr_root->v_mount == mp) 3640 return (0); 3641 if (pr->pr_enforce_statfs == 2) 3642 return (ENOENT); 3643 /* 3644 * If jail's chroot directory is set to "/" we should be able to see 3645 * all mount-points from inside a jail. 3646 * This is ugly check, but this is the only situation when jail's 3647 * directory ends with '/'. 3648 */ 3649 if (strcmp(pr->pr_path, "/") == 0) 3650 return (0); 3651 len = strlen(pr->pr_path); 3652 sp = &mp->mnt_stat; 3653 if (strncmp(pr->pr_path, sp->f_mntonname, len) != 0) 3654 return (ENOENT); 3655 /* 3656 * Be sure that we don't have situation where jail's root directory 3657 * is "/some/path" and mount point is "/some/pathpath". 3658 */ 3659 if (sp->f_mntonname[len] != '\0' && sp->f_mntonname[len] != '/') 3660 return (ENOENT); 3661 return (0); 3662} 3663 3664void 3665prison_enforce_statfs(struct ucred *cred, struct mount *mp, struct statfs *sp) 3666{ 3667 char jpath[MAXPATHLEN]; 3668 struct prison *pr; 3669 size_t len; 3670 3671 pr = cred->cr_prison; 3672 if (pr->pr_enforce_statfs == 0) 3673 return; 3674 if (prison_canseemount(cred, mp) != 0) { 3675 bzero(sp->f_mntonname, sizeof(sp->f_mntonname)); 3676 strlcpy(sp->f_mntonname, "[restricted]", 3677 sizeof(sp->f_mntonname)); 3678 return; 3679 } 3680 if (pr->pr_root->v_mount == mp) { 3681 /* 3682 * Clear current buffer data, so we are sure nothing from 3683 * the valid path left there. 3684 */ 3685 bzero(sp->f_mntonname, sizeof(sp->f_mntonname)); 3686 *sp->f_mntonname = '/'; 3687 return; 3688 } 3689 /* 3690 * If jail's chroot directory is set to "/" we should be able to see 3691 * all mount-points from inside a jail. 3692 */ 3693 if (strcmp(pr->pr_path, "/") == 0) 3694 return; 3695 len = strlen(pr->pr_path); 3696 strlcpy(jpath, sp->f_mntonname + len, sizeof(jpath)); 3697 /* 3698 * Clear current buffer data, so we are sure nothing from 3699 * the valid path left there. 3700 */ 3701 bzero(sp->f_mntonname, sizeof(sp->f_mntonname)); 3702 if (*jpath == '\0') { 3703 /* Should never happen. */ 3704 *sp->f_mntonname = '/'; 3705 } else { 3706 strlcpy(sp->f_mntonname, jpath, sizeof(sp->f_mntonname)); 3707 } 3708} 3709 3710/* 3711 * Check with permission for a specific privilege is granted within jail. We 3712 * have a specific list of accepted privileges; the rest are denied. 3713 */ 3714int 3715prison_priv_check(struct ucred *cred, int priv) 3716{ 3717 3718 if (!jailed(cred)) 3719 return (0); 3720 3721#ifdef VIMAGE 3722 /* 3723 * Privileges specific to prisons with a virtual network stack. 3724 * There might be a duplicate entry here in case the privilege 3725 * is only granted conditionally in the legacy jail case. 3726 */ 3727 switch (priv) { 3728#ifdef notyet 3729 /* 3730 * NFS-specific privileges. 3731 */ 3732 case PRIV_NFS_DAEMON: 3733 case PRIV_NFS_LOCKD: 3734#endif 3735 /* 3736 * Network stack privileges. 3737 */ 3738 case PRIV_NET_BRIDGE: 3739 case PRIV_NET_GRE: 3740 case PRIV_NET_BPF: 3741 case PRIV_NET_RAW: /* Dup, cond. in legacy jail case. */ 3742 case PRIV_NET_ROUTE: 3743 case PRIV_NET_TAP: 3744 case PRIV_NET_SETIFMTU: 3745 case PRIV_NET_SETIFFLAGS: 3746 case PRIV_NET_SETIFCAP: 3747 case PRIV_NET_SETIFDESCR: 3748 case PRIV_NET_SETIFNAME : 3749 case PRIV_NET_SETIFMETRIC: 3750 case PRIV_NET_SETIFPHYS: 3751 case PRIV_NET_SETIFMAC: 3752 case PRIV_NET_ADDMULTI: 3753 case PRIV_NET_DELMULTI: 3754 case PRIV_NET_HWIOCTL: 3755 case PRIV_NET_SETLLADDR: 3756 case PRIV_NET_ADDIFGROUP: 3757 case PRIV_NET_DELIFGROUP: 3758 case PRIV_NET_IFCREATE: 3759 case PRIV_NET_IFDESTROY: 3760 case PRIV_NET_ADDIFADDR: 3761 case PRIV_NET_DELIFADDR: 3762 case PRIV_NET_LAGG: 3763 case PRIV_NET_GIF: 3764 case PRIV_NET_SETIFVNET: 3765 case PRIV_NET_SETIFFIB: 3766 3767 /* 3768 * 802.11-related privileges. 3769 */ 3770 case PRIV_NET80211_GETKEY: 3771#ifdef notyet 3772 case PRIV_NET80211_MANAGE: /* XXX-BZ discuss with sam@ */ 3773#endif 3774 3775#ifdef notyet 3776 /* 3777 * AppleTalk privileges. 3778 */ 3779 case PRIV_NETATALK_RESERVEDPORT: 3780 3781 /* 3782 * ATM privileges. 3783 */ 3784 case PRIV_NETATM_CFG: 3785 case PRIV_NETATM_ADD: 3786 case PRIV_NETATM_DEL: 3787 case PRIV_NETATM_SET: 3788 3789 /* 3790 * Bluetooth privileges. 3791 */ 3792 case PRIV_NETBLUETOOTH_RAW: 3793#endif 3794 3795 /* 3796 * Netgraph and netgraph module privileges. 3797 */ 3798 case PRIV_NETGRAPH_CONTROL: 3799#ifdef notyet 3800 case PRIV_NETGRAPH_TTY: 3801#endif 3802 3803 /* 3804 * IPv4 and IPv6 privileges. 3805 */ 3806 case PRIV_NETINET_IPFW: 3807 case PRIV_NETINET_DIVERT: 3808 case PRIV_NETINET_PF: 3809 case PRIV_NETINET_DUMMYNET: 3810 case PRIV_NETINET_CARP: 3811 case PRIV_NETINET_MROUTE: 3812 case PRIV_NETINET_RAW: 3813 case PRIV_NETINET_ADDRCTRL6: 3814 case PRIV_NETINET_ND6: 3815 case PRIV_NETINET_SCOPE6: 3816 case PRIV_NETINET_ALIFETIME6: 3817 case PRIV_NETINET_IPSEC: 3818 case PRIV_NETINET_BINDANY: 3819 3820#ifdef notyet 3821 /* 3822 * IPX/SPX privileges. 3823 */ 3824 case PRIV_NETIPX_RESERVEDPORT: 3825 case PRIV_NETIPX_RAW: 3826 3827 /* 3828 * NCP privileges. 3829 */ 3830 case PRIV_NETNCP: 3831 3832 /* 3833 * SMB privileges. 3834 */ 3835 case PRIV_NETSMB: 3836#endif 3837 3838 /* 3839 * No default: or deny here. 3840 * In case of no permit fall through to next switch(). 3841 */ 3842 if (cred->cr_prison->pr_flags & PR_VNET) 3843 return (0); 3844 } 3845#endif /* VIMAGE */ 3846 3847 switch (priv) { 3848 3849 /* 3850 * Allow ktrace privileges for root in jail. 3851 */ 3852 case PRIV_KTRACE: 3853 3854#if 0 3855 /* 3856 * Allow jailed processes to configure audit identity and 3857 * submit audit records (login, etc). In the future we may 3858 * want to further refine the relationship between audit and 3859 * jail. 3860 */ 3861 case PRIV_AUDIT_GETAUDIT: 3862 case PRIV_AUDIT_SETAUDIT: 3863 case PRIV_AUDIT_SUBMIT: 3864#endif 3865 3866 /* 3867 * Allow jailed processes to manipulate process UNIX 3868 * credentials in any way they see fit. 3869 */ 3870 case PRIV_CRED_SETUID: 3871 case PRIV_CRED_SETEUID: 3872 case PRIV_CRED_SETGID: 3873 case PRIV_CRED_SETEGID: 3874 case PRIV_CRED_SETGROUPS: 3875 case PRIV_CRED_SETREUID: 3876 case PRIV_CRED_SETREGID: 3877 case PRIV_CRED_SETRESUID: 3878 case PRIV_CRED_SETRESGID: 3879 3880 /* 3881 * Jail implements visibility constraints already, so allow 3882 * jailed root to override uid/gid-based constraints. 3883 */ 3884 case PRIV_SEEOTHERGIDS: 3885 case PRIV_SEEOTHERUIDS: 3886 3887 /* 3888 * Jail implements inter-process debugging limits already, so 3889 * allow jailed root various debugging privileges. 3890 */ 3891 case PRIV_DEBUG_DIFFCRED: 3892 case PRIV_DEBUG_SUGID: 3893 case PRIV_DEBUG_UNPRIV: 3894 3895 /* 3896 * Allow jail to set various resource limits and login 3897 * properties, and for now, exceed process resource limits. 3898 */ 3899 case PRIV_PROC_LIMIT: 3900 case PRIV_PROC_SETLOGIN: 3901 case PRIV_PROC_SETRLIMIT: 3902 3903 /* 3904 * System V and POSIX IPC privileges are granted in jail. 3905 */ 3906 case PRIV_IPC_READ: 3907 case PRIV_IPC_WRITE: 3908 case PRIV_IPC_ADMIN: 3909 case PRIV_IPC_MSGSIZE: 3910 case PRIV_MQ_ADMIN: 3911 3912 /* 3913 * Jail operations within a jail work on child jails. 3914 */ 3915 case PRIV_JAIL_ATTACH: 3916 case PRIV_JAIL_SET: 3917 case PRIV_JAIL_REMOVE: 3918 3919 /* 3920 * Jail implements its own inter-process limits, so allow 3921 * root processes in jail to change scheduling on other 3922 * processes in the same jail. Likewise for signalling. 3923 */ 3924 case PRIV_SCHED_DIFFCRED: 3925 case PRIV_SCHED_CPUSET: 3926 case PRIV_SIGNAL_DIFFCRED: 3927 case PRIV_SIGNAL_SUGID: 3928 3929 /* 3930 * Allow jailed processes to write to sysctls marked as jail 3931 * writable. 3932 */ 3933 case PRIV_SYSCTL_WRITEJAIL: 3934 3935 /* 3936 * Allow root in jail to manage a variety of quota 3937 * properties. These should likely be conditional on a 3938 * configuration option. 3939 */ 3940 case PRIV_VFS_GETQUOTA: 3941 case PRIV_VFS_SETQUOTA: 3942 3943 /* 3944 * Since Jail relies on chroot() to implement file system 3945 * protections, grant many VFS privileges to root in jail. 3946 * Be careful to exclude mount-related and NFS-related 3947 * privileges. 3948 */ 3949 case PRIV_VFS_READ: 3950 case PRIV_VFS_WRITE: 3951 case PRIV_VFS_ADMIN: 3952 case PRIV_VFS_EXEC: 3953 case PRIV_VFS_LOOKUP: 3954 case PRIV_VFS_BLOCKRESERVE: /* XXXRW: Slightly surprising. */ 3955 case PRIV_VFS_CHFLAGS_DEV: 3956 case PRIV_VFS_CHOWN: 3957 case PRIV_VFS_CHROOT: 3958 case PRIV_VFS_RETAINSUGID: 3959 case PRIV_VFS_FCHROOT: 3960 case PRIV_VFS_LINK: 3961 case PRIV_VFS_SETGID: 3962 case PRIV_VFS_STAT: 3963 case PRIV_VFS_STICKYFILE: 3964 3965 /* 3966 * As in the non-jail case, non-root users are expected to be 3967 * able to read kernel/phyiscal memory (provided /dev/[k]mem 3968 * exists in the jail and they have permission to access it). 3969 */ 3970 case PRIV_KMEM_READ: 3971 return (0); 3972 3973 /* 3974 * Depending on the global setting, allow privilege of 3975 * setting system flags. 3976 */ 3977 case PRIV_VFS_SYSFLAGS: 3978 if (cred->cr_prison->pr_allow & PR_ALLOW_CHFLAGS) 3979 return (0); 3980 else 3981 return (EPERM); 3982 3983 /* 3984 * Depending on the global setting, allow privilege of 3985 * mounting/unmounting file systems. 3986 */ 3987 case PRIV_VFS_MOUNT: 3988 case PRIV_VFS_UNMOUNT: 3989 case PRIV_VFS_MOUNT_NONUSER: 3990 case PRIV_VFS_MOUNT_OWNER: 3991 if (cred->cr_prison->pr_allow & PR_ALLOW_MOUNT && 3992 cred->cr_prison->pr_enforce_statfs < 2) 3993 return (0); 3994 else 3995 return (EPERM); 3996 3997 /* 3998 * Allow jailed root to bind reserved ports and reuse in-use 3999 * ports. 4000 */ 4001 case PRIV_NETINET_RESERVEDPORT: 4002 case PRIV_NETINET_REUSEPORT: 4003 return (0); 4004 4005 /* 4006 * Allow jailed root to set certian IPv4/6 (option) headers. 4007 */ 4008 case PRIV_NETINET_SETHDROPTS: 4009 return (0); 4010 4011 /* 4012 * Conditionally allow creating raw sockets in jail. 4013 */ 4014 case PRIV_NETINET_RAW: 4015 if (cred->cr_prison->pr_allow & PR_ALLOW_RAW_SOCKETS) 4016 return (0); 4017 else 4018 return (EPERM); 4019 4020 /* 4021 * Since jail implements its own visibility limits on netstat 4022 * sysctls, allow getcred. This allows identd to work in 4023 * jail. 4024 */ 4025 case PRIV_NETINET_GETCRED: 4026 return (0); 4027 4028 /* 4029 * Allow jailed root to set loginclass. 4030 */ 4031 case PRIV_PROC_SETLOGINCLASS: 4032 return (0); 4033 4034 default: 4035 /* 4036 * In all remaining cases, deny the privilege request. This 4037 * includes almost all network privileges, many system 4038 * configuration privileges. 4039 */ 4040 return (EPERM); 4041 } 4042} 4043 4044/* 4045 * Return the part of pr2's name that is relative to pr1, or the whole name 4046 * if it does not directly follow. 4047 */ 4048 4049char * 4050prison_name(struct prison *pr1, struct prison *pr2) 4051{ 4052 char *name; 4053 4054 /* Jails see themselves as "0" (if they see themselves at all). */ 4055 if (pr1 == pr2) 4056 return "0"; 4057 name = pr2->pr_name; 4058 if (prison_ischild(pr1, pr2)) { 4059 /* 4060 * pr1 isn't locked (and allprison_lock may not be either) 4061 * so its length can't be counted on. But the number of dots 4062 * can be counted on - and counted. 4063 */ 4064 for (; pr1 != &prison0; pr1 = pr1->pr_parent) 4065 name = strchr(name, '.') + 1; 4066 } 4067 return (name); 4068} 4069 4070/* 4071 * Return the part of pr2's path that is relative to pr1, or the whole path 4072 * if it does not directly follow. 4073 */ 4074static char * 4075prison_path(struct prison *pr1, struct prison *pr2) 4076{ 4077 char *path1, *path2; 4078 int len1; 4079 4080 path1 = pr1->pr_path; 4081 path2 = pr2->pr_path; 4082 if (!strcmp(path1, "/")) 4083 return (path2); 4084 len1 = strlen(path1); 4085 if (strncmp(path1, path2, len1)) 4086 return (path2); 4087 if (path2[len1] == '\0') 4088 return "/"; 4089 if (path2[len1] == '/') 4090 return (path2 + len1); 4091 return (path2); 4092} 4093 4094 4095/* 4096 * Jail-related sysctls. 4097 */ 4098static SYSCTL_NODE(_security, OID_AUTO, jail, CTLFLAG_RW, 0, 4099 "Jails"); 4100 4101static int 4102sysctl_jail_list(SYSCTL_HANDLER_ARGS) 4103{ 4104 struct xprison *xp; 4105 struct prison *pr, *cpr; 4106#ifdef INET 4107 struct in_addr *ip4 = NULL; 4108 int ip4s = 0; 4109#endif 4110#ifdef INET6 4111 struct in6_addr *ip6 = NULL; 4112 int ip6s = 0; 4113#endif 4114 int descend, error; 4115 4116 xp = malloc(sizeof(*xp), M_TEMP, M_WAITOK); 4117 pr = req->td->td_ucred->cr_prison; 4118 error = 0; 4119 sx_slock(&allprison_lock); 4120 FOREACH_PRISON_DESCENDANT(pr, cpr, descend) { 4121#if defined(INET) || defined(INET6) 4122 again: 4123#endif 4124 mtx_lock(&cpr->pr_mtx); 4125#ifdef INET 4126 if (cpr->pr_ip4s > 0) { 4127 if (ip4s < cpr->pr_ip4s) { 4128 ip4s = cpr->pr_ip4s; 4129 mtx_unlock(&cpr->pr_mtx); 4130 ip4 = realloc(ip4, ip4s * 4131 sizeof(struct in_addr), M_TEMP, M_WAITOK); 4132 goto again; 4133 } 4134 bcopy(cpr->pr_ip4, ip4, 4135 cpr->pr_ip4s * sizeof(struct in_addr)); 4136 } 4137#endif 4138#ifdef INET6 4139 if (cpr->pr_ip6s > 0) { 4140 if (ip6s < cpr->pr_ip6s) { 4141 ip6s = cpr->pr_ip6s; 4142 mtx_unlock(&cpr->pr_mtx); 4143 ip6 = realloc(ip6, ip6s * 4144 sizeof(struct in6_addr), M_TEMP, M_WAITOK); 4145 goto again; 4146 } 4147 bcopy(cpr->pr_ip6, ip6, 4148 cpr->pr_ip6s * sizeof(struct in6_addr)); 4149 } 4150#endif 4151 if (cpr->pr_ref == 0) { 4152 mtx_unlock(&cpr->pr_mtx); 4153 continue; 4154 } 4155 bzero(xp, sizeof(*xp)); 4156 xp->pr_version = XPRISON_VERSION; 4157 xp->pr_id = cpr->pr_id; 4158 xp->pr_state = cpr->pr_uref > 0 4159 ? PRISON_STATE_ALIVE : PRISON_STATE_DYING; 4160 strlcpy(xp->pr_path, prison_path(pr, cpr), sizeof(xp->pr_path)); 4161 strlcpy(xp->pr_host, cpr->pr_hostname, sizeof(xp->pr_host)); 4162 strlcpy(xp->pr_name, prison_name(pr, cpr), sizeof(xp->pr_name)); 4163#ifdef INET 4164 xp->pr_ip4s = cpr->pr_ip4s; 4165#endif 4166#ifdef INET6 4167 xp->pr_ip6s = cpr->pr_ip6s; 4168#endif 4169 mtx_unlock(&cpr->pr_mtx); 4170 error = SYSCTL_OUT(req, xp, sizeof(*xp)); 4171 if (error) 4172 break; 4173#ifdef INET 4174 if (xp->pr_ip4s > 0) { 4175 error = SYSCTL_OUT(req, ip4, 4176 xp->pr_ip4s * sizeof(struct in_addr)); 4177 if (error) 4178 break; 4179 } 4180#endif 4181#ifdef INET6 4182 if (xp->pr_ip6s > 0) { 4183 error = SYSCTL_OUT(req, ip6, 4184 xp->pr_ip6s * sizeof(struct in6_addr)); 4185 if (error) 4186 break; 4187 } 4188#endif 4189 } 4190 sx_sunlock(&allprison_lock); 4191 free(xp, M_TEMP); 4192#ifdef INET 4193 free(ip4, M_TEMP); 4194#endif 4195#ifdef INET6 4196 free(ip6, M_TEMP); 4197#endif 4198 return (error); 4199} 4200 4201SYSCTL_OID(_security_jail, OID_AUTO, list, 4202 CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0, 4203 sysctl_jail_list, "S", "List of active jails"); 4204 4205static int 4206sysctl_jail_jailed(SYSCTL_HANDLER_ARGS) 4207{ 4208 int error, injail; 4209 4210 injail = jailed(req->td->td_ucred); 4211 error = SYSCTL_OUT(req, &injail, sizeof(injail)); 4212 4213 return (error); 4214} 4215 4216SYSCTL_PROC(_security_jail, OID_AUTO, jailed, 4217 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0, 4218 sysctl_jail_jailed, "I", "Process in jail?"); 4219 4220static int 4221sysctl_jail_vnet(SYSCTL_HANDLER_ARGS) 4222{ 4223 int error, havevnet; 4224#ifdef VIMAGE 4225 struct ucred *cred = req->td->td_ucred; 4226 4227 havevnet = jailed(cred) && prison_owns_vnet(cred); 4228#else 4229 havevnet = 0; 4230#endif 4231 error = SYSCTL_OUT(req, &havevnet, sizeof(havevnet)); 4232 4233 return (error); 4234} 4235 4236SYSCTL_PROC(_security_jail, OID_AUTO, vnet, 4237 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0, 4238 sysctl_jail_vnet, "I", "Jail owns VNET?"); 4239 4240#if defined(INET) || defined(INET6) 4241SYSCTL_UINT(_security_jail, OID_AUTO, jail_max_af_ips, CTLFLAG_RW, 4242 &jail_max_af_ips, 0, 4243 "Number of IP addresses a jail may have at most per address family"); 4244#endif 4245 4246/* 4247 * Default parameters for jail(2) compatability. For historical reasons, 4248 * the sysctl names have varying similarity to the parameter names. Prisons 4249 * just see their own parameters, and can't change them. 4250 */ 4251static int 4252sysctl_jail_default_allow(SYSCTL_HANDLER_ARGS) 4253{ 4254 struct prison *pr; 4255 int allow, error, i; 4256 4257 pr = req->td->td_ucred->cr_prison; 4258 allow = (pr == &prison0) ? jail_default_allow : pr->pr_allow; 4259 4260 /* Get the current flag value, and convert it to a boolean. */ 4261 i = (allow & arg2) ? 1 : 0; 4262 if (arg1 != NULL) 4263 i = !i; 4264 error = sysctl_handle_int(oidp, &i, 0, req); 4265 if (error || !req->newptr) 4266 return (error); 4267 i = i ? arg2 : 0; 4268 if (arg1 != NULL) 4269 i ^= arg2; 4270 /* 4271 * The sysctls don't have CTLFLAGS_PRISON, so assume prison0 4272 * for writing. 4273 */ 4274 mtx_lock(&prison0.pr_mtx); 4275 jail_default_allow = (jail_default_allow & ~arg2) | i; 4276 mtx_unlock(&prison0.pr_mtx); 4277 return (0); 4278} 4279 4280SYSCTL_PROC(_security_jail, OID_AUTO, set_hostname_allowed, 4281 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 4282 NULL, PR_ALLOW_SET_HOSTNAME, sysctl_jail_default_allow, "I", 4283 "Processes in jail can set their hostnames"); 4284SYSCTL_PROC(_security_jail, OID_AUTO, socket_unixiproute_only, 4285 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 4286 (void *)1, PR_ALLOW_SOCKET_AF, sysctl_jail_default_allow, "I", 4287 "Processes in jail are limited to creating UNIX/IP/route sockets only"); 4288SYSCTL_PROC(_security_jail, OID_AUTO, sysvipc_allowed, 4289 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 4290 NULL, PR_ALLOW_SYSVIPC, sysctl_jail_default_allow, "I", 4291 "Processes in jail can use System V IPC primitives"); 4292SYSCTL_PROC(_security_jail, OID_AUTO, allow_raw_sockets, 4293 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 4294 NULL, PR_ALLOW_RAW_SOCKETS, sysctl_jail_default_allow, "I", 4295 "Prison root can create raw sockets"); 4296SYSCTL_PROC(_security_jail, OID_AUTO, chflags_allowed, 4297 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 4298 NULL, PR_ALLOW_CHFLAGS, sysctl_jail_default_allow, "I", 4299 "Processes in jail can alter system file flags"); 4300SYSCTL_PROC(_security_jail, OID_AUTO, mount_allowed, 4301 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 4302 NULL, PR_ALLOW_MOUNT, sysctl_jail_default_allow, "I", 4303 "Processes in jail can mount/unmount jail-friendly file systems"); 4304SYSCTL_PROC(_security_jail, OID_AUTO, mount_devfs_allowed, 4305 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 4306 NULL, PR_ALLOW_MOUNT_DEVFS, sysctl_jail_default_allow, "I", 4307 "Processes in jail can mount the devfs file system"); 4308SYSCTL_PROC(_security_jail, OID_AUTO, mount_fdescfs_allowed, 4309 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 4310 NULL, PR_ALLOW_MOUNT_FDESCFS, sysctl_jail_default_allow, "I", 4311 "Processes in jail can mount the fdescfs file system"); 4312SYSCTL_PROC(_security_jail, OID_AUTO, mount_nullfs_allowed, 4313 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 4314 NULL, PR_ALLOW_MOUNT_NULLFS, sysctl_jail_default_allow, "I", 4315 "Processes in jail can mount the nullfs file system"); 4316SYSCTL_PROC(_security_jail, OID_AUTO, mount_procfs_allowed, 4317 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 4318 NULL, PR_ALLOW_MOUNT_PROCFS, sysctl_jail_default_allow, "I", 4319 "Processes in jail can mount the procfs file system"); 4320SYSCTL_PROC(_security_jail, OID_AUTO, mount_linprocfs_allowed, 4321 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 4322 NULL, PR_ALLOW_MOUNT_LINPROCFS, sysctl_jail_default_allow, "I", 4323 "Processes in jail can mount the linprocfs file system"); 4324SYSCTL_PROC(_security_jail, OID_AUTO, mount_linsysfs_allowed, 4325 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 4326 NULL, PR_ALLOW_MOUNT_LINSYSFS, sysctl_jail_default_allow, "I", 4327 "Processes in jail can mount the linsysfs file system"); 4328SYSCTL_PROC(_security_jail, OID_AUTO, mount_tmpfs_allowed, 4329 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 4330 NULL, PR_ALLOW_MOUNT_TMPFS, sysctl_jail_default_allow, "I", 4331 "Processes in jail can mount the tmpfs file system"); 4332SYSCTL_PROC(_security_jail, OID_AUTO, mount_zfs_allowed, 4333 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 4334 NULL, PR_ALLOW_MOUNT_ZFS, sysctl_jail_default_allow, "I", 4335 "Processes in jail can mount the zfs file system"); 4336 4337static int 4338sysctl_jail_default_level(SYSCTL_HANDLER_ARGS) 4339{ 4340 struct prison *pr; 4341 int level, error; 4342 4343 pr = req->td->td_ucred->cr_prison; 4344 level = (pr == &prison0) ? *(int *)arg1 : *(int *)((char *)pr + arg2); 4345 error = sysctl_handle_int(oidp, &level, 0, req); 4346 if (error || !req->newptr) 4347 return (error); 4348 *(int *)arg1 = level; 4349 return (0); 4350} 4351 4352SYSCTL_PROC(_security_jail, OID_AUTO, enforce_statfs, 4353 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 4354 &jail_default_enforce_statfs, offsetof(struct prison, pr_enforce_statfs), 4355 sysctl_jail_default_level, "I", 4356 "Processes in jail cannot see all mounted file systems"); 4357 4358SYSCTL_PROC(_security_jail, OID_AUTO, devfs_ruleset, 4359 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, 4360 &jail_default_devfs_rsnum, offsetof(struct prison, pr_devfs_rsnum), 4361 sysctl_jail_default_level, "I", 4362 "Ruleset for the devfs filesystem in jail"); 4363 4364/* 4365 * Nodes to describe jail parameters. Maximum length of string parameters 4366 * is returned in the string itself, and the other parameters exist merely 4367 * to make themselves and their types known. 4368 */ 4369SYSCTL_NODE(_security_jail, OID_AUTO, param, CTLFLAG_RW, 0, 4370 "Jail parameters"); 4371 4372int 4373sysctl_jail_param(SYSCTL_HANDLER_ARGS) 4374{ 4375 int i; 4376 long l; 4377 size_t s; 4378 char numbuf[12]; 4379 4380 switch (oidp->oid_kind & CTLTYPE) 4381 { 4382 case CTLTYPE_LONG: 4383 case CTLTYPE_ULONG: 4384 l = 0; 4385#ifdef SCTL_MASK32 4386 if (!(req->flags & SCTL_MASK32)) 4387#endif 4388 return (SYSCTL_OUT(req, &l, sizeof(l))); 4389 case CTLTYPE_INT: 4390 case CTLTYPE_UINT: 4391 i = 0; 4392 return (SYSCTL_OUT(req, &i, sizeof(i))); 4393 case CTLTYPE_STRING: 4394 snprintf(numbuf, sizeof(numbuf), "%jd", (intmax_t)arg2); 4395 return 4396 (sysctl_handle_string(oidp, numbuf, sizeof(numbuf), req)); 4397 case CTLTYPE_STRUCT: 4398 s = (size_t)arg2; 4399 return (SYSCTL_OUT(req, &s, sizeof(s))); 4400 } 4401 return (0); 4402} 4403 4404/* 4405 * CTLFLAG_RDTUN in the following indicates jail parameters that can be set at 4406 * jail creation time but cannot be changed in an existing jail. 4407 */ 4408SYSCTL_JAIL_PARAM(, jid, CTLTYPE_INT | CTLFLAG_RDTUN, "I", "Jail ID"); 4409SYSCTL_JAIL_PARAM(, parent, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail parent ID"); 4410SYSCTL_JAIL_PARAM_STRING(, name, CTLFLAG_RW, MAXHOSTNAMELEN, "Jail name"); 4411SYSCTL_JAIL_PARAM_STRING(, path, CTLFLAG_RDTUN, MAXPATHLEN, "Jail root path"); 4412SYSCTL_JAIL_PARAM(, securelevel, CTLTYPE_INT | CTLFLAG_RW, 4413 "I", "Jail secure level"); 4414SYSCTL_JAIL_PARAM(, osreldate, CTLTYPE_INT | CTLFLAG_RDTUN, "I", 4415 "Jail value for kern.osreldate and uname -K"); 4416SYSCTL_JAIL_PARAM_STRING(, osrelease, CTLFLAG_RDTUN, OSRELEASELEN, 4417 "Jail value for kern.osrelease and uname -r"); 4418SYSCTL_JAIL_PARAM(, enforce_statfs, CTLTYPE_INT | CTLFLAG_RW, 4419 "I", "Jail cannot see all mounted file systems"); 4420SYSCTL_JAIL_PARAM(, devfs_ruleset, CTLTYPE_INT | CTLFLAG_RW, 4421 "I", "Ruleset for in-jail devfs mounts"); 4422SYSCTL_JAIL_PARAM(, persist, CTLTYPE_INT | CTLFLAG_RW, 4423 "B", "Jail persistence"); 4424#ifdef VIMAGE 4425SYSCTL_JAIL_PARAM(, vnet, CTLTYPE_INT | CTLFLAG_RDTUN, 4426 "E,jailsys", "Virtual network stack"); 4427#endif 4428SYSCTL_JAIL_PARAM(, dying, CTLTYPE_INT | CTLFLAG_RD, 4429 "B", "Jail is in the process of shutting down"); 4430 4431SYSCTL_JAIL_PARAM_NODE(children, "Number of child jails"); 4432SYSCTL_JAIL_PARAM(_children, cur, CTLTYPE_INT | CTLFLAG_RD, 4433 "I", "Current number of child jails"); 4434SYSCTL_JAIL_PARAM(_children, max, CTLTYPE_INT | CTLFLAG_RW, 4435 "I", "Maximum number of child jails"); 4436 4437SYSCTL_JAIL_PARAM_SYS_NODE(host, CTLFLAG_RW, "Jail host info"); 4438SYSCTL_JAIL_PARAM_STRING(_host, hostname, CTLFLAG_RW, MAXHOSTNAMELEN, 4439 "Jail hostname"); 4440SYSCTL_JAIL_PARAM_STRING(_host, domainname, CTLFLAG_RW, MAXHOSTNAMELEN, 4441 "Jail NIS domainname"); 4442SYSCTL_JAIL_PARAM_STRING(_host, hostuuid, CTLFLAG_RW, HOSTUUIDLEN, 4443 "Jail host UUID"); 4444SYSCTL_JAIL_PARAM(_host, hostid, CTLTYPE_ULONG | CTLFLAG_RW, 4445 "LU", "Jail host ID"); 4446 4447SYSCTL_JAIL_PARAM_NODE(cpuset, "Jail cpuset"); 4448SYSCTL_JAIL_PARAM(_cpuset, id, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail cpuset ID"); 4449 4450#ifdef INET 4451SYSCTL_JAIL_PARAM_SYS_NODE(ip4, CTLFLAG_RDTUN, 4452 "Jail IPv4 address virtualization"); 4453SYSCTL_JAIL_PARAM_STRUCT(_ip4, addr, CTLFLAG_RW, sizeof(struct in_addr), 4454 "S,in_addr,a", "Jail IPv4 addresses"); 4455SYSCTL_JAIL_PARAM(_ip4, saddrsel, CTLTYPE_INT | CTLFLAG_RW, 4456 "B", "Do (not) use IPv4 source address selection rather than the " 4457 "primary jail IPv4 address."); 4458#endif 4459#ifdef INET6 4460SYSCTL_JAIL_PARAM_SYS_NODE(ip6, CTLFLAG_RDTUN, 4461 "Jail IPv6 address virtualization"); 4462SYSCTL_JAIL_PARAM_STRUCT(_ip6, addr, CTLFLAG_RW, sizeof(struct in6_addr), 4463 "S,in6_addr,a", "Jail IPv6 addresses"); 4464SYSCTL_JAIL_PARAM(_ip6, saddrsel, CTLTYPE_INT | CTLFLAG_RW, 4465 "B", "Do (not) use IPv6 source address selection rather than the " 4466 "primary jail IPv6 address."); 4467#endif 4468 4469SYSCTL_JAIL_PARAM_NODE(allow, "Jail permission flags"); 4470SYSCTL_JAIL_PARAM(_allow, set_hostname, CTLTYPE_INT | CTLFLAG_RW, 4471 "B", "Jail may set hostname"); 4472SYSCTL_JAIL_PARAM(_allow, sysvipc, CTLTYPE_INT | CTLFLAG_RW, 4473 "B", "Jail may use SYSV IPC"); 4474SYSCTL_JAIL_PARAM(_allow, raw_sockets, CTLTYPE_INT | CTLFLAG_RW, 4475 "B", "Jail may create raw sockets"); 4476SYSCTL_JAIL_PARAM(_allow, chflags, CTLTYPE_INT | CTLFLAG_RW, 4477 "B", "Jail may alter system file flags"); 4478SYSCTL_JAIL_PARAM(_allow, quotas, CTLTYPE_INT | CTLFLAG_RW, 4479 "B", "Jail may set file quotas"); 4480SYSCTL_JAIL_PARAM(_allow, socket_af, CTLTYPE_INT | CTLFLAG_RW, 4481 "B", "Jail may create sockets other than just UNIX/IPv4/IPv6/route"); 4482 4483SYSCTL_JAIL_PARAM_SUBNODE(allow, mount, "Jail mount/unmount permission flags"); 4484SYSCTL_JAIL_PARAM(_allow_mount, , CTLTYPE_INT | CTLFLAG_RW, 4485 "B", "Jail may mount/unmount jail-friendly file systems in general"); 4486SYSCTL_JAIL_PARAM(_allow_mount, devfs, CTLTYPE_INT | CTLFLAG_RW, 4487 "B", "Jail may mount the devfs file system"); 4488SYSCTL_JAIL_PARAM(_allow_mount, fdescfs, CTLTYPE_INT | CTLFLAG_RW, 4489 "B", "Jail may mount the fdescfs file system"); 4490SYSCTL_JAIL_PARAM(_allow_mount, nullfs, CTLTYPE_INT | CTLFLAG_RW, 4491 "B", "Jail may mount the nullfs file system"); 4492SYSCTL_JAIL_PARAM(_allow_mount, procfs, CTLTYPE_INT | CTLFLAG_RW, 4493 "B", "Jail may mount the procfs file system"); 4494SYSCTL_JAIL_PARAM(_allow_mount, linprocfs, CTLTYPE_INT | CTLFLAG_RW, 4495 "B", "Jail may mount the linprocfs file system"); 4496SYSCTL_JAIL_PARAM(_allow_mount, linsysfs, CTLTYPE_INT | CTLFLAG_RW, 4497 "B", "Jail may mount the linsysfs file system"); 4498SYSCTL_JAIL_PARAM(_allow_mount, tmpfs, CTLTYPE_INT | CTLFLAG_RW, 4499 "B", "Jail may mount the tmpfs file system"); 4500SYSCTL_JAIL_PARAM(_allow_mount, zfs, CTLTYPE_INT | CTLFLAG_RW, 4501 "B", "Jail may mount the zfs file system"); 4502 4503#ifdef RACCT 4504void 4505prison_racct_foreach(void (*callback)(struct racct *racct, 4506 void *arg2, void *arg3), void *arg2, void *arg3) 4507{ 4508 struct prison_racct *prr; 4509 4510 ASSERT_RACCT_ENABLED(); 4511 4512 sx_slock(&allprison_lock); 4513 LIST_FOREACH(prr, &allprison_racct, prr_next) 4514 (callback)(prr->prr_racct, arg2, arg3); 4515 sx_sunlock(&allprison_lock); 4516} 4517 4518static struct prison_racct * 4519prison_racct_find_locked(const char *name) 4520{ 4521 struct prison_racct *prr; 4522 4523 ASSERT_RACCT_ENABLED(); 4524 sx_assert(&allprison_lock, SA_XLOCKED); 4525 4526 if (name[0] == '\0' || strlen(name) >= MAXHOSTNAMELEN) 4527 return (NULL); 4528 4529 LIST_FOREACH(prr, &allprison_racct, prr_next) { 4530 if (strcmp(name, prr->prr_name) != 0) 4531 continue; 4532 4533 /* Found prison_racct with a matching name? */ 4534 prison_racct_hold(prr); 4535 return (prr); 4536 } 4537 4538 /* Add new prison_racct. */ 4539 prr = malloc(sizeof(*prr), M_PRISON_RACCT, M_ZERO | M_WAITOK); 4540 racct_create(&prr->prr_racct); 4541 4542 strcpy(prr->prr_name, name); 4543 refcount_init(&prr->prr_refcount, 1); 4544 LIST_INSERT_HEAD(&allprison_racct, prr, prr_next); 4545 4546 return (prr); 4547} 4548 4549struct prison_racct * 4550prison_racct_find(const char *name) 4551{ 4552 struct prison_racct *prr; 4553 4554 ASSERT_RACCT_ENABLED(); 4555 4556 sx_xlock(&allprison_lock); 4557 prr = prison_racct_find_locked(name); 4558 sx_xunlock(&allprison_lock); 4559 return (prr); 4560} 4561 4562void 4563prison_racct_hold(struct prison_racct *prr) 4564{ 4565 4566 ASSERT_RACCT_ENABLED(); 4567 4568 refcount_acquire(&prr->prr_refcount); 4569} 4570 4571static void 4572prison_racct_free_locked(struct prison_racct *prr) 4573{ 4574 4575 ASSERT_RACCT_ENABLED(); 4576 sx_assert(&allprison_lock, SA_XLOCKED); 4577 4578 if (refcount_release(&prr->prr_refcount)) { 4579 racct_destroy(&prr->prr_racct); 4580 LIST_REMOVE(prr, prr_next); 4581 free(prr, M_PRISON_RACCT); 4582 } 4583} 4584 4585void 4586prison_racct_free(struct prison_racct *prr) 4587{ 4588 int old; 4589 4590 ASSERT_RACCT_ENABLED(); 4591 sx_assert(&allprison_lock, SA_UNLOCKED); 4592 4593 old = prr->prr_refcount; 4594 if (old > 1 && atomic_cmpset_int(&prr->prr_refcount, old, old - 1)) 4595 return; 4596 4597 sx_xlock(&allprison_lock); 4598 prison_racct_free_locked(prr); 4599 sx_xunlock(&allprison_lock); 4600} 4601 4602static void 4603prison_racct_attach(struct prison *pr) 4604{ 4605 struct prison_racct *prr; 4606 4607 ASSERT_RACCT_ENABLED(); 4608 sx_assert(&allprison_lock, SA_XLOCKED); 4609 4610 prr = prison_racct_find_locked(pr->pr_name); 4611 KASSERT(prr != NULL, ("cannot find prison_racct")); 4612 4613 pr->pr_prison_racct = prr; 4614} 4615 4616/* 4617 * Handle jail renaming. From the racct point of view, renaming means 4618 * moving from one prison_racct to another. 4619 */ 4620static void 4621prison_racct_modify(struct prison *pr) 4622{ 4623 struct proc *p; 4624 struct ucred *cred; 4625 struct prison_racct *oldprr; 4626 4627 ASSERT_RACCT_ENABLED(); 4628 4629 sx_slock(&allproc_lock); 4630 sx_xlock(&allprison_lock); 4631 4632 if (strcmp(pr->pr_name, pr->pr_prison_racct->prr_name) == 0) { 4633 sx_xunlock(&allprison_lock); 4634 sx_sunlock(&allproc_lock); 4635 return; 4636 } 4637 4638 oldprr = pr->pr_prison_racct; 4639 pr->pr_prison_racct = NULL; 4640 4641 prison_racct_attach(pr); 4642 4643 /* 4644 * Move resource utilisation records. 4645 */ 4646 racct_move(pr->pr_prison_racct->prr_racct, oldprr->prr_racct); 4647 4648 /* 4649 * Force rctl to reattach rules to processes. 4650 */ 4651 FOREACH_PROC_IN_SYSTEM(p) { 4652 PROC_LOCK(p); 4653 cred = crhold(p->p_ucred); 4654 PROC_UNLOCK(p); 4655 racct_proc_ucred_changed(p, cred, cred); 4656 crfree(cred); 4657 } 4658 4659 sx_sunlock(&allproc_lock); 4660 prison_racct_free_locked(oldprr); 4661 sx_xunlock(&allprison_lock); 4662} 4663 4664static void 4665prison_racct_detach(struct prison *pr) 4666{ 4667 4668 ASSERT_RACCT_ENABLED(); 4669 sx_assert(&allprison_lock, SA_UNLOCKED); 4670 4671 if (pr->pr_prison_racct == NULL) 4672 return; 4673 prison_racct_free(pr->pr_prison_racct); 4674 pr->pr_prison_racct = NULL; 4675} 4676#endif /* RACCT */ 4677 4678#ifdef DDB 4679 4680static void 4681db_show_prison(struct prison *pr) 4682{ 4683 int fi; 4684#if defined(INET) || defined(INET6) 4685 int ii; 4686#endif 4687 unsigned jsf; 4688#ifdef INET6 4689 char ip6buf[INET6_ADDRSTRLEN]; 4690#endif 4691 4692 db_printf("prison %p:\n", pr); 4693 db_printf(" jid = %d\n", pr->pr_id); 4694 db_printf(" name = %s\n", pr->pr_name); 4695 db_printf(" parent = %p\n", pr->pr_parent); 4696 db_printf(" ref = %d\n", pr->pr_ref); 4697 db_printf(" uref = %d\n", pr->pr_uref); 4698 db_printf(" path = %s\n", pr->pr_path); 4699 db_printf(" cpuset = %d\n", pr->pr_cpuset 4700 ? pr->pr_cpuset->cs_id : -1); 4701#ifdef VIMAGE 4702 db_printf(" vnet = %p\n", pr->pr_vnet); 4703#endif 4704 db_printf(" root = %p\n", pr->pr_root); 4705 db_printf(" securelevel = %d\n", pr->pr_securelevel); 4706 db_printf(" devfs_rsnum = %d\n", pr->pr_devfs_rsnum); 4707 db_printf(" children.max = %d\n", pr->pr_childmax); 4708 db_printf(" children.cur = %d\n", pr->pr_childcount); 4709 db_printf(" child = %p\n", LIST_FIRST(&pr->pr_children)); 4710 db_printf(" sibling = %p\n", LIST_NEXT(pr, pr_sibling)); 4711 db_printf(" flags = 0x%x", pr->pr_flags); 4712 for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]); 4713 fi++) 4714 if (pr_flag_names[fi] != NULL && (pr->pr_flags & (1 << fi))) 4715 db_printf(" %s", pr_flag_names[fi]); 4716 for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]); 4717 fi++) { 4718 jsf = pr->pr_flags & 4719 (pr_flag_jailsys[fi].disable | pr_flag_jailsys[fi].new); 4720 db_printf(" %-16s= %s\n", pr_flag_jailsys[fi].name, 4721 pr_flag_jailsys[fi].disable && 4722 (jsf == pr_flag_jailsys[fi].disable) ? "disable" 4723 : (jsf == pr_flag_jailsys[fi].new) ? "new" 4724 : "inherit"); 4725 } 4726 db_printf(" allow = 0x%x", pr->pr_allow); 4727 for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]); 4728 fi++) 4729 if (pr_allow_names[fi] != NULL && (pr->pr_allow & (1 << fi))) 4730 db_printf(" %s", pr_allow_names[fi]); 4731 db_printf("\n"); 4732 db_printf(" enforce_statfs = %d\n", pr->pr_enforce_statfs); 4733 db_printf(" host.hostname = %s\n", pr->pr_hostname); 4734 db_printf(" host.domainname = %s\n", pr->pr_domainname); 4735 db_printf(" host.hostuuid = %s\n", pr->pr_hostuuid); 4736 db_printf(" host.hostid = %lu\n", pr->pr_hostid); 4737#ifdef INET 4738 db_printf(" ip4s = %d\n", pr->pr_ip4s); 4739 for (ii = 0; ii < pr->pr_ip4s; ii++) 4740 db_printf(" %s %s\n", 4741 ii == 0 ? "ip4.addr =" : " ", 4742 inet_ntoa(pr->pr_ip4[ii])); 4743#endif 4744#ifdef INET6 4745 db_printf(" ip6s = %d\n", pr->pr_ip6s); 4746 for (ii = 0; ii < pr->pr_ip6s; ii++) 4747 db_printf(" %s %s\n", 4748 ii == 0 ? "ip6.addr =" : " ", 4749 ip6_sprintf(ip6buf, &pr->pr_ip6[ii])); 4750#endif 4751} 4752 4753DB_SHOW_COMMAND(prison, db_show_prison_command) 4754{ 4755 struct prison *pr; 4756 4757 if (!have_addr) { 4758 /* 4759 * Show all prisons in the list, and prison0 which is not 4760 * listed. 4761 */ 4762 db_show_prison(&prison0); 4763 if (!db_pager_quit) { 4764 TAILQ_FOREACH(pr, &allprison, pr_list) { 4765 db_show_prison(pr); 4766 if (db_pager_quit) 4767 break; 4768 } 4769 } 4770 return; 4771 } 4772 4773 if (addr == 0) 4774 pr = &prison0; 4775 else { 4776 /* Look for a prison with the ID and with references. */ 4777 TAILQ_FOREACH(pr, &allprison, pr_list) 4778 if (pr->pr_id == addr && pr->pr_ref > 0) 4779 break; 4780 if (pr == NULL) 4781 /* Look again, without requiring a reference. */ 4782 TAILQ_FOREACH(pr, &allprison, pr_list) 4783 if (pr->pr_id == addr) 4784 break; 4785 if (pr == NULL) 4786 /* Assume address points to a valid prison. */ 4787 pr = (struct prison *)addr; 4788 } 4789 db_show_prison(pr); 4790} 4791 4792#endif /* DDB */ 4793