kern_jail.c revision 298565
1/*- 2 * Copyright (c) 1999 Poul-Henning Kamp. 3 * Copyright (c) 2008 Bjoern A. Zeeb. 4 * Copyright (c) 2009 James Gritton. 5 * All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 16 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 17 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 18 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 19 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 20 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 21 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 22 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 23 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 24 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 25 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 26 * SUCH DAMAGE. 27 */ 28 29#include <sys/cdefs.h> 30__FBSDID("$FreeBSD: head/sys/kern/kern_jail.c 298565 2016-04-25 04:24:00Z 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#endif /* DDB */ 72 73#include <security/mac/mac_framework.h> 74 75#define DEFAULT_HOSTUUID "00000000-0000-0000-0000-000000000000" 76 77MALLOC_DEFINE(M_PRISON, "prison", "Prison structures"); 78static MALLOC_DEFINE(M_PRISON_RACCT, "prison_racct", "Prison racct structures"); 79 80/* Keep struct prison prison0 and some code in kern_jail_set() readable. */ 81#ifdef INET 82#ifdef INET6 83#define _PR_IP_SADDRSEL PR_IP4_SADDRSEL|PR_IP6_SADDRSEL 84#else 85#define _PR_IP_SADDRSEL PR_IP4_SADDRSEL 86#endif 87#else /* !INET */ 88#ifdef INET6 89#define _PR_IP_SADDRSEL PR_IP6_SADDRSEL 90#else 91#define _PR_IP_SADDRSEL 0 92#endif 93#endif 94 95/* prison0 describes what is "real" about the system. */ 96struct prison prison0 = { 97 .pr_id = 0, 98 .pr_name = "0", 99 .pr_ref = 1, 100 .pr_uref = 1, 101 .pr_path = "/", 102 .pr_securelevel = -1, 103 .pr_devfs_rsnum = 0, 104 .pr_childmax = JAIL_MAX, 105 .pr_hostuuid = DEFAULT_HOSTUUID, 106 .pr_children = LIST_HEAD_INITIALIZER(prison0.pr_children), 107#ifdef VIMAGE 108 .pr_flags = PR_HOST|PR_VNET|_PR_IP_SADDRSEL, 109#else 110 .pr_flags = PR_HOST|_PR_IP_SADDRSEL, 111#endif 112 .pr_allow = PR_ALLOW_ALL, 113}; 114MTX_SYSINIT(prison0, &prison0.pr_mtx, "jail mutex", MTX_DEF); 115 116/* allprison, allprison_racct and lastprid are protected by allprison_lock. */ 117struct sx allprison_lock; 118SX_SYSINIT(allprison_lock, &allprison_lock, "allprison"); 119struct prisonlist allprison = TAILQ_HEAD_INITIALIZER(allprison); 120LIST_HEAD(, prison_racct) allprison_racct; 121int lastprid = 0; 122 123static int do_jail_attach(struct thread *td, struct prison *pr); 124static void prison_complete(void *context, int pending); 125static void prison_deref(struct prison *pr, int flags); 126static char *prison_path(struct prison *pr1, struct prison *pr2); 127static void prison_remove_one(struct prison *pr); 128#ifdef RACCT 129static void prison_racct_attach(struct prison *pr); 130static void prison_racct_modify(struct prison *pr); 131static void prison_racct_detach(struct prison *pr); 132#endif 133#ifdef INET 134static int _prison_check_ip4(const struct prison *, const struct in_addr *); 135static int prison_restrict_ip4(struct prison *pr, struct in_addr *newip4); 136#endif 137#ifdef INET6 138static int _prison_check_ip6(struct prison *pr, struct in6_addr *ia6); 139static int prison_restrict_ip6(struct prison *pr, struct in6_addr *newip6); 140#endif 141 142/* Flags for prison_deref */ 143#define PD_DEREF 0x01 144#define PD_DEUREF 0x02 145#define PD_LOCKED 0x04 146#define PD_LIST_SLOCKED 0x08 147#define PD_LIST_XLOCKED 0x10 148 149/* 150 * Parameter names corresponding to PR_* flag values. Size values are for kvm 151 * as we cannot figure out the size of a sparse array, or an array without a 152 * terminating entry. 153 */ 154static char *pr_flag_names[] = { 155 [0] = "persist", 156#ifdef INET 157 [7] = "ip4.saddrsel", 158#endif 159#ifdef INET6 160 [8] = "ip6.saddrsel", 161#endif 162}; 163const size_t pr_flag_names_size = sizeof(pr_flag_names); 164 165static char *pr_flag_nonames[] = { 166 [0] = "nopersist", 167#ifdef INET 168 [7] = "ip4.nosaddrsel", 169#endif 170#ifdef INET6 171 [8] = "ip6.nosaddrsel", 172#endif 173}; 174const size_t pr_flag_nonames_size = sizeof(pr_flag_nonames); 175 176struct jailsys_flags { 177 const char *name; 178 unsigned disable; 179 unsigned new; 180} pr_flag_jailsys[] = { 181 { "host", 0, PR_HOST }, 182#ifdef VIMAGE 183 { "vnet", 0, PR_VNET }, 184#endif 185#ifdef INET 186 { "ip4", PR_IP4_USER, PR_IP4_USER }, 187#endif 188#ifdef INET6 189 { "ip6", PR_IP6_USER, PR_IP6_USER }, 190#endif 191}; 192const size_t pr_flag_jailsys_size = sizeof(pr_flag_jailsys); 193 194static char *pr_allow_names[] = { 195 "allow.set_hostname", 196 "allow.sysvipc", 197 "allow.raw_sockets", 198 "allow.chflags", 199 "allow.mount", 200 "allow.quotas", 201 "allow.socket_af", 202 "allow.mount.devfs", 203 "allow.mount.nullfs", 204 "allow.mount.zfs", 205 "allow.mount.procfs", 206 "allow.mount.tmpfs", 207 "allow.mount.fdescfs", 208 "allow.mount.linprocfs", 209 "allow.mount.linsysfs", 210}; 211const size_t pr_allow_names_size = sizeof(pr_allow_names); 212 213static char *pr_allow_nonames[] = { 214 "allow.noset_hostname", 215 "allow.nosysvipc", 216 "allow.noraw_sockets", 217 "allow.nochflags", 218 "allow.nomount", 219 "allow.noquotas", 220 "allow.nosocket_af", 221 "allow.mount.nodevfs", 222 "allow.mount.nonullfs", 223 "allow.mount.nozfs", 224 "allow.mount.noprocfs", 225 "allow.mount.notmpfs", 226 "allow.mount.nofdescfs", 227 "allow.mount.nolinprocfs", 228 "allow.mount.nolinsysfs", 229}; 230const size_t pr_allow_nonames_size = sizeof(pr_allow_nonames); 231 232#define JAIL_DEFAULT_ALLOW PR_ALLOW_SET_HOSTNAME 233#define JAIL_DEFAULT_ENFORCE_STATFS 2 234#define JAIL_DEFAULT_DEVFS_RSNUM 0 235static unsigned jail_default_allow = JAIL_DEFAULT_ALLOW; 236static int jail_default_enforce_statfs = JAIL_DEFAULT_ENFORCE_STATFS; 237static int jail_default_devfs_rsnum = JAIL_DEFAULT_DEVFS_RSNUM; 238#if defined(INET) || defined(INET6) 239static unsigned jail_max_af_ips = 255; 240#endif 241 242/* 243 * Initialize the parts of prison0 that can't be static-initialized with 244 * constants. This is called from proc0_init() after creating thread0 cpuset. 245 */ 246void 247prison0_init(void) 248{ 249 250 prison0.pr_cpuset = cpuset_ref(thread0.td_cpuset); 251 prison0.pr_osreldate = osreldate; 252 strlcpy(prison0.pr_osrelease, osrelease, sizeof(prison0.pr_osrelease)); 253} 254 255#ifdef INET 256static int 257qcmp_v4(const void *ip1, const void *ip2) 258{ 259 in_addr_t iaa, iab; 260 261 /* 262 * We need to compare in HBO here to get the list sorted as expected 263 * by the result of the code. Sorting NBO addresses gives you 264 * interesting results. If you do not understand, do not try. 265 */ 266 iaa = ntohl(((const struct in_addr *)ip1)->s_addr); 267 iab = ntohl(((const struct in_addr *)ip2)->s_addr); 268 269 /* 270 * Do not simply return the difference of the two numbers, the int is 271 * not wide enough. 272 */ 273 if (iaa > iab) 274 return (1); 275 else if (iaa < iab) 276 return (-1); 277 else 278 return (0); 279} 280#endif 281 282#ifdef INET6 283static int 284qcmp_v6(const void *ip1, const void *ip2) 285{ 286 const struct in6_addr *ia6a, *ia6b; 287 int i, rc; 288 289 ia6a = (const struct in6_addr *)ip1; 290 ia6b = (const struct in6_addr *)ip2; 291 292 rc = 0; 293 for (i = 0; rc == 0 && i < sizeof(struct in6_addr); i++) { 294 if (ia6a->s6_addr[i] > ia6b->s6_addr[i]) 295 rc = 1; 296 else if (ia6a->s6_addr[i] < ia6b->s6_addr[i]) 297 rc = -1; 298 } 299 return (rc); 300} 301#endif 302 303/* 304 * struct jail_args { 305 * struct jail *jail; 306 * }; 307 */ 308int 309sys_jail(struct thread *td, struct jail_args *uap) 310{ 311 uint32_t version; 312 int error; 313 struct jail j; 314 315 error = copyin(uap->jail, &version, sizeof(uint32_t)); 316 if (error) 317 return (error); 318 319 switch (version) { 320 case 0: 321 { 322 struct jail_v0 j0; 323 324 /* FreeBSD single IPv4 jails. */ 325 bzero(&j, sizeof(struct jail)); 326 error = copyin(uap->jail, &j0, sizeof(struct jail_v0)); 327 if (error) 328 return (error); 329 j.version = j0.version; 330 j.path = j0.path; 331 j.hostname = j0.hostname; 332 j.ip4s = htonl(j0.ip_number); /* jail_v0 is host order */ 333 break; 334 } 335 336 case 1: 337 /* 338 * Version 1 was used by multi-IPv4 jail implementations 339 * that never made it into the official kernel. 340 */ 341 return (EINVAL); 342 343 case 2: /* JAIL_API_VERSION */ 344 /* FreeBSD multi-IPv4/IPv6,noIP jails. */ 345 error = copyin(uap->jail, &j, sizeof(struct jail)); 346 if (error) 347 return (error); 348 break; 349 350 default: 351 /* Sci-Fi jails are not supported, sorry. */ 352 return (EINVAL); 353 } 354 return (kern_jail(td, &j)); 355} 356 357int 358kern_jail(struct thread *td, struct jail *j) 359{ 360 struct iovec optiov[2 * (4 + nitems(pr_allow_names) 361#ifdef INET 362 + 1 363#endif 364#ifdef INET6 365 + 1 366#endif 367 )]; 368 struct uio opt; 369 char *u_path, *u_hostname, *u_name; 370#ifdef INET 371 uint32_t ip4s; 372 struct in_addr *u_ip4; 373#endif 374#ifdef INET6 375 struct in6_addr *u_ip6; 376#endif 377 size_t tmplen; 378 int error, enforce_statfs, fi; 379 380 bzero(&optiov, sizeof(optiov)); 381 opt.uio_iov = optiov; 382 opt.uio_iovcnt = 0; 383 opt.uio_offset = -1; 384 opt.uio_resid = -1; 385 opt.uio_segflg = UIO_SYSSPACE; 386 opt.uio_rw = UIO_READ; 387 opt.uio_td = td; 388 389 /* Set permissions for top-level jails from sysctls. */ 390 if (!jailed(td->td_ucred)) { 391 for (fi = 0; fi < nitems(pr_allow_names); fi++) { 392 optiov[opt.uio_iovcnt].iov_base = 393 (jail_default_allow & (1 << fi)) 394 ? pr_allow_names[fi] : pr_allow_nonames[fi]; 395 optiov[opt.uio_iovcnt].iov_len = 396 strlen(optiov[opt.uio_iovcnt].iov_base) + 1; 397 opt.uio_iovcnt += 2; 398 } 399 optiov[opt.uio_iovcnt].iov_base = "enforce_statfs"; 400 optiov[opt.uio_iovcnt].iov_len = sizeof("enforce_statfs"); 401 opt.uio_iovcnt++; 402 enforce_statfs = jail_default_enforce_statfs; 403 optiov[opt.uio_iovcnt].iov_base = &enforce_statfs; 404 optiov[opt.uio_iovcnt].iov_len = sizeof(enforce_statfs); 405 opt.uio_iovcnt++; 406 } 407 408 tmplen = MAXPATHLEN + MAXHOSTNAMELEN + MAXHOSTNAMELEN; 409#ifdef INET 410 ip4s = (j->version == 0) ? 1 : j->ip4s; 411 if (ip4s > jail_max_af_ips) 412 return (EINVAL); 413 tmplen += ip4s * sizeof(struct in_addr); 414#else 415 if (j->ip4s > 0) 416 return (EINVAL); 417#endif 418#ifdef INET6 419 if (j->ip6s > jail_max_af_ips) 420 return (EINVAL); 421 tmplen += j->ip6s * sizeof(struct in6_addr); 422#else 423 if (j->ip6s > 0) 424 return (EINVAL); 425#endif 426 u_path = malloc(tmplen, M_TEMP, M_WAITOK); 427 u_hostname = u_path + MAXPATHLEN; 428 u_name = u_hostname + MAXHOSTNAMELEN; 429#ifdef INET 430 u_ip4 = (struct in_addr *)(u_name + MAXHOSTNAMELEN); 431#endif 432#ifdef INET6 433#ifdef INET 434 u_ip6 = (struct in6_addr *)(u_ip4 + ip4s); 435#else 436 u_ip6 = (struct in6_addr *)(u_name + MAXHOSTNAMELEN); 437#endif 438#endif 439 optiov[opt.uio_iovcnt].iov_base = "path"; 440 optiov[opt.uio_iovcnt].iov_len = sizeof("path"); 441 opt.uio_iovcnt++; 442 optiov[opt.uio_iovcnt].iov_base = u_path; 443 error = copyinstr(j->path, u_path, MAXPATHLEN, 444 &optiov[opt.uio_iovcnt].iov_len); 445 if (error) { 446 free(u_path, M_TEMP); 447 return (error); 448 } 449 opt.uio_iovcnt++; 450 optiov[opt.uio_iovcnt].iov_base = "host.hostname"; 451 optiov[opt.uio_iovcnt].iov_len = sizeof("host.hostname"); 452 opt.uio_iovcnt++; 453 optiov[opt.uio_iovcnt].iov_base = u_hostname; 454 error = copyinstr(j->hostname, u_hostname, MAXHOSTNAMELEN, 455 &optiov[opt.uio_iovcnt].iov_len); 456 if (error) { 457 free(u_path, M_TEMP); 458 return (error); 459 } 460 opt.uio_iovcnt++; 461 if (j->jailname != NULL) { 462 optiov[opt.uio_iovcnt].iov_base = "name"; 463 optiov[opt.uio_iovcnt].iov_len = sizeof("name"); 464 opt.uio_iovcnt++; 465 optiov[opt.uio_iovcnt].iov_base = u_name; 466 error = copyinstr(j->jailname, u_name, MAXHOSTNAMELEN, 467 &optiov[opt.uio_iovcnt].iov_len); 468 if (error) { 469 free(u_path, M_TEMP); 470 return (error); 471 } 472 opt.uio_iovcnt++; 473 } 474#ifdef INET 475 optiov[opt.uio_iovcnt].iov_base = "ip4.addr"; 476 optiov[opt.uio_iovcnt].iov_len = sizeof("ip4.addr"); 477 opt.uio_iovcnt++; 478 optiov[opt.uio_iovcnt].iov_base = u_ip4; 479 optiov[opt.uio_iovcnt].iov_len = ip4s * sizeof(struct in_addr); 480 if (j->version == 0) 481 u_ip4->s_addr = j->ip4s; 482 else { 483 error = copyin(j->ip4, u_ip4, optiov[opt.uio_iovcnt].iov_len); 484 if (error) { 485 free(u_path, M_TEMP); 486 return (error); 487 } 488 } 489 opt.uio_iovcnt++; 490#endif 491#ifdef INET6 492 optiov[opt.uio_iovcnt].iov_base = "ip6.addr"; 493 optiov[opt.uio_iovcnt].iov_len = sizeof("ip6.addr"); 494 opt.uio_iovcnt++; 495 optiov[opt.uio_iovcnt].iov_base = u_ip6; 496 optiov[opt.uio_iovcnt].iov_len = j->ip6s * sizeof(struct in6_addr); 497 error = copyin(j->ip6, u_ip6, optiov[opt.uio_iovcnt].iov_len); 498 if (error) { 499 free(u_path, M_TEMP); 500 return (error); 501 } 502 opt.uio_iovcnt++; 503#endif 504 KASSERT(opt.uio_iovcnt <= nitems(optiov), 505 ("kern_jail: too many iovecs (%d)", opt.uio_iovcnt)); 506 error = kern_jail_set(td, &opt, JAIL_CREATE | JAIL_ATTACH); 507 free(u_path, M_TEMP); 508 return (error); 509} 510 511 512/* 513 * struct jail_set_args { 514 * struct iovec *iovp; 515 * unsigned int iovcnt; 516 * int flags; 517 * }; 518 */ 519int 520sys_jail_set(struct thread *td, struct jail_set_args *uap) 521{ 522 struct uio *auio; 523 int error; 524 525 /* Check that we have an even number of iovecs. */ 526 if (uap->iovcnt & 1) 527 return (EINVAL); 528 529 error = copyinuio(uap->iovp, uap->iovcnt, &auio); 530 if (error) 531 return (error); 532 error = kern_jail_set(td, auio, uap->flags); 533 free(auio, M_IOV); 534 return (error); 535} 536 537int 538kern_jail_set(struct thread *td, struct uio *optuio, int flags) 539{ 540 struct nameidata nd; 541#ifdef INET 542 struct in_addr *ip4; 543#endif 544#ifdef INET6 545 struct in6_addr *ip6; 546#endif 547 struct vfsopt *opt; 548 struct vfsoptlist *opts; 549 struct prison *pr, *deadpr, *mypr, *ppr, *tpr; 550 struct vnode *root; 551 char *domain, *errmsg, *host, *name, *namelc, *p, *path, *uuid; 552 char *g_path, *osrelstr; 553#if defined(INET) || defined(INET6) 554 struct prison *tppr; 555 void *op; 556#endif 557 unsigned long hid; 558 size_t namelen, onamelen; 559 int born, created, cuflags, descend, enforce; 560 int error, errmsg_len, errmsg_pos; 561 int gotchildmax, gotenforce, gothid, gotrsnum, gotslevel; 562 int fi, jid, jsys, len, level; 563 int childmax, osreldt, rsnum, slevel; 564 int fullpath_disabled; 565#if defined(INET) || defined(INET6) 566 int ii, ij; 567#endif 568#ifdef INET 569 int ip4s, redo_ip4; 570#endif 571#ifdef INET6 572 int ip6s, redo_ip6; 573#endif 574 uint64_t pr_allow, ch_allow, pr_flags, ch_flags; 575 unsigned tallow; 576 char numbuf[12]; 577 578 error = priv_check(td, PRIV_JAIL_SET); 579 if (!error && (flags & JAIL_ATTACH)) 580 error = priv_check(td, PRIV_JAIL_ATTACH); 581 if (error) 582 return (error); 583 mypr = ppr = td->td_ucred->cr_prison; 584 if ((flags & JAIL_CREATE) && mypr->pr_childmax == 0) 585 return (EPERM); 586 if (flags & ~JAIL_SET_MASK) 587 return (EINVAL); 588 589 /* 590 * Check all the parameters before committing to anything. Not all 591 * errors can be caught early, but we may as well try. Also, this 592 * takes care of some expensive stuff (path lookup) before getting 593 * the allprison lock. 594 * 595 * XXX Jails are not filesystems, and jail parameters are not mount 596 * options. But it makes more sense to re-use the vfsopt code 597 * than duplicate it under a different name. 598 */ 599 error = vfs_buildopts(optuio, &opts); 600 if (error) 601 return (error); 602#ifdef INET 603 ip4 = NULL; 604#endif 605#ifdef INET6 606 ip6 = NULL; 607#endif 608 g_path = NULL; 609 610 error = vfs_copyopt(opts, "jid", &jid, sizeof(jid)); 611 if (error == ENOENT) 612 jid = 0; 613 else if (error != 0) 614 goto done_free; 615 616 error = vfs_copyopt(opts, "securelevel", &slevel, sizeof(slevel)); 617 if (error == ENOENT) 618 gotslevel = 0; 619 else if (error != 0) 620 goto done_free; 621 else 622 gotslevel = 1; 623 624 error = 625 vfs_copyopt(opts, "children.max", &childmax, sizeof(childmax)); 626 if (error == ENOENT) 627 gotchildmax = 0; 628 else if (error != 0) 629 goto done_free; 630 else 631 gotchildmax = 1; 632 633 error = vfs_copyopt(opts, "enforce_statfs", &enforce, sizeof(enforce)); 634 if (error == ENOENT) 635 gotenforce = 0; 636 else if (error != 0) 637 goto done_free; 638 else if (enforce < 0 || enforce > 2) { 639 error = EINVAL; 640 goto done_free; 641 } else 642 gotenforce = 1; 643 644 error = vfs_copyopt(opts, "devfs_ruleset", &rsnum, sizeof(rsnum)); 645 if (error == ENOENT) 646 gotrsnum = 0; 647 else if (error != 0) 648 goto done_free; 649 else 650 gotrsnum = 1; 651 652 pr_flags = ch_flags = 0; 653 for (fi = 0; fi < nitems(pr_flag_names); fi++) { 654 if (pr_flag_names[fi] == NULL) 655 continue; 656 vfs_flagopt(opts, pr_flag_names[fi], &pr_flags, 1 << fi); 657 vfs_flagopt(opts, pr_flag_nonames[fi], &ch_flags, 1 << fi); 658 } 659 ch_flags |= pr_flags; 660 for (fi = 0; fi < nitems(pr_flag_jailsys); fi++) { 661 error = vfs_copyopt(opts, pr_flag_jailsys[fi].name, &jsys, 662 sizeof(jsys)); 663 if (error == ENOENT) 664 continue; 665 if (error != 0) 666 goto done_free; 667 switch (jsys) { 668 case JAIL_SYS_DISABLE: 669 if (!pr_flag_jailsys[fi].disable) { 670 error = EINVAL; 671 goto done_free; 672 } 673 pr_flags |= pr_flag_jailsys[fi].disable; 674 break; 675 case JAIL_SYS_NEW: 676 pr_flags |= pr_flag_jailsys[fi].new; 677 break; 678 case JAIL_SYS_INHERIT: 679 break; 680 default: 681 error = EINVAL; 682 goto done_free; 683 } 684 ch_flags |= 685 pr_flag_jailsys[fi].new | pr_flag_jailsys[fi].disable; 686 } 687 if ((flags & (JAIL_CREATE | JAIL_UPDATE | JAIL_ATTACH)) == JAIL_CREATE 688 && !(pr_flags & PR_PERSIST)) { 689 error = EINVAL; 690 vfs_opterror(opts, "new jail must persist or attach"); 691 goto done_errmsg; 692 } 693#ifdef VIMAGE 694 if ((flags & JAIL_UPDATE) && (ch_flags & PR_VNET)) { 695 error = EINVAL; 696 vfs_opterror(opts, "vnet cannot be changed after creation"); 697 goto done_errmsg; 698 } 699#endif 700#ifdef INET 701 if ((flags & JAIL_UPDATE) && (ch_flags & PR_IP4_USER)) { 702 error = EINVAL; 703 vfs_opterror(opts, "ip4 cannot be changed after creation"); 704 goto done_errmsg; 705 } 706#endif 707#ifdef INET6 708 if ((flags & JAIL_UPDATE) && (ch_flags & PR_IP6_USER)) { 709 error = EINVAL; 710 vfs_opterror(opts, "ip6 cannot be changed after creation"); 711 goto done_errmsg; 712 } 713#endif 714 715 pr_allow = ch_allow = 0; 716 for (fi = 0; fi < nitems(pr_allow_names); fi++) { 717 vfs_flagopt(opts, pr_allow_names[fi], &pr_allow, 1 << fi); 718 vfs_flagopt(opts, pr_allow_nonames[fi], &ch_allow, 1 << fi); 719 } 720 ch_allow |= pr_allow; 721 722 error = vfs_getopt(opts, "name", (void **)&name, &len); 723 if (error == ENOENT) 724 name = NULL; 725 else if (error != 0) 726 goto done_free; 727 else { 728 if (len == 0 || name[len - 1] != '\0') { 729 error = EINVAL; 730 goto done_free; 731 } 732 if (len > MAXHOSTNAMELEN) { 733 error = ENAMETOOLONG; 734 goto done_free; 735 } 736 } 737 738 error = vfs_getopt(opts, "host.hostname", (void **)&host, &len); 739 if (error == ENOENT) 740 host = NULL; 741 else if (error != 0) 742 goto done_free; 743 else { 744 ch_flags |= PR_HOST; 745 pr_flags |= PR_HOST; 746 if (len == 0 || host[len - 1] != '\0') { 747 error = EINVAL; 748 goto done_free; 749 } 750 if (len > MAXHOSTNAMELEN) { 751 error = ENAMETOOLONG; 752 goto done_free; 753 } 754 } 755 756 error = vfs_getopt(opts, "host.domainname", (void **)&domain, &len); 757 if (error == ENOENT) 758 domain = NULL; 759 else if (error != 0) 760 goto done_free; 761 else { 762 ch_flags |= PR_HOST; 763 pr_flags |= PR_HOST; 764 if (len == 0 || domain[len - 1] != '\0') { 765 error = EINVAL; 766 goto done_free; 767 } 768 if (len > MAXHOSTNAMELEN) { 769 error = ENAMETOOLONG; 770 goto done_free; 771 } 772 } 773 774 error = vfs_getopt(opts, "host.hostuuid", (void **)&uuid, &len); 775 if (error == ENOENT) 776 uuid = NULL; 777 else if (error != 0) 778 goto done_free; 779 else { 780 ch_flags |= PR_HOST; 781 pr_flags |= PR_HOST; 782 if (len == 0 || uuid[len - 1] != '\0') { 783 error = EINVAL; 784 goto done_free; 785 } 786 if (len > HOSTUUIDLEN) { 787 error = ENAMETOOLONG; 788 goto done_free; 789 } 790 } 791 792#ifdef COMPAT_FREEBSD32 793 if (SV_PROC_FLAG(td->td_proc, SV_ILP32)) { 794 uint32_t hid32; 795 796 error = vfs_copyopt(opts, "host.hostid", &hid32, sizeof(hid32)); 797 hid = hid32; 798 } else 799#endif 800 error = vfs_copyopt(opts, "host.hostid", &hid, sizeof(hid)); 801 if (error == ENOENT) 802 gothid = 0; 803 else if (error != 0) 804 goto done_free; 805 else { 806 gothid = 1; 807 ch_flags |= PR_HOST; 808 pr_flags |= PR_HOST; 809 } 810 811#ifdef INET 812 error = vfs_getopt(opts, "ip4.addr", &op, &ip4s); 813 if (error == ENOENT) 814 ip4s = 0; 815 else if (error != 0) 816 goto done_free; 817 else if (ip4s & (sizeof(*ip4) - 1)) { 818 error = EINVAL; 819 goto done_free; 820 } else { 821 ch_flags |= PR_IP4_USER; 822 pr_flags |= PR_IP4_USER; 823 if (ip4s > 0) { 824 ip4s /= sizeof(*ip4); 825 if (ip4s > jail_max_af_ips) { 826 error = EINVAL; 827 vfs_opterror(opts, "too many IPv4 addresses"); 828 goto done_errmsg; 829 } 830 ip4 = malloc(ip4s * sizeof(*ip4), M_PRISON, M_WAITOK); 831 bcopy(op, ip4, ip4s * sizeof(*ip4)); 832 /* 833 * IP addresses are all sorted but ip[0] to preserve 834 * the primary IP address as given from userland. 835 * This special IP is used for unbound outgoing 836 * connections as well for "loopback" traffic in case 837 * source address selection cannot find any more fitting 838 * address to connect from. 839 */ 840 if (ip4s > 1) 841 qsort(ip4 + 1, ip4s - 1, sizeof(*ip4), qcmp_v4); 842 /* 843 * Check for duplicate addresses and do some simple 844 * zero and broadcast checks. If users give other bogus 845 * addresses it is their problem. 846 * 847 * We do not have to care about byte order for these 848 * checks so we will do them in NBO. 849 */ 850 for (ii = 0; ii < ip4s; ii++) { 851 if (ip4[ii].s_addr == INADDR_ANY || 852 ip4[ii].s_addr == INADDR_BROADCAST) { 853 error = EINVAL; 854 goto done_free; 855 } 856 if ((ii+1) < ip4s && 857 (ip4[0].s_addr == ip4[ii+1].s_addr || 858 ip4[ii].s_addr == ip4[ii+1].s_addr)) { 859 error = EINVAL; 860 goto done_free; 861 } 862 } 863 } 864 } 865#endif 866 867#ifdef INET6 868 error = vfs_getopt(opts, "ip6.addr", &op, &ip6s); 869 if (error == ENOENT) 870 ip6s = 0; 871 else if (error != 0) 872 goto done_free; 873 else if (ip6s & (sizeof(*ip6) - 1)) { 874 error = EINVAL; 875 goto done_free; 876 } else { 877 ch_flags |= PR_IP6_USER; 878 pr_flags |= PR_IP6_USER; 879 if (ip6s > 0) { 880 ip6s /= sizeof(*ip6); 881 if (ip6s > jail_max_af_ips) { 882 error = EINVAL; 883 vfs_opterror(opts, "too many IPv6 addresses"); 884 goto done_errmsg; 885 } 886 ip6 = malloc(ip6s * sizeof(*ip6), M_PRISON, M_WAITOK); 887 bcopy(op, ip6, ip6s * sizeof(*ip6)); 888 if (ip6s > 1) 889 qsort(ip6 + 1, ip6s - 1, sizeof(*ip6), qcmp_v6); 890 for (ii = 0; ii < ip6s; ii++) { 891 if (IN6_IS_ADDR_UNSPECIFIED(&ip6[ii])) { 892 error = EINVAL; 893 goto done_free; 894 } 895 if ((ii+1) < ip6s && 896 (IN6_ARE_ADDR_EQUAL(&ip6[0], &ip6[ii+1]) || 897 IN6_ARE_ADDR_EQUAL(&ip6[ii], &ip6[ii+1]))) 898 { 899 error = EINVAL; 900 goto done_free; 901 } 902 } 903 } 904 } 905#endif 906 907#if defined(VIMAGE) && (defined(INET) || defined(INET6)) 908 if ((ch_flags & PR_VNET) && (ch_flags & (PR_IP4_USER | PR_IP6_USER))) { 909 error = EINVAL; 910 vfs_opterror(opts, 911 "vnet jails cannot have IP address restrictions"); 912 goto done_errmsg; 913 } 914#endif 915 916 fullpath_disabled = 0; 917 root = NULL; 918 error = vfs_getopt(opts, "path", (void **)&path, &len); 919 if (error == ENOENT) 920 path = NULL; 921 else if (error != 0) 922 goto done_free; 923 else { 924 if (flags & JAIL_UPDATE) { 925 error = EINVAL; 926 vfs_opterror(opts, 927 "path cannot be changed after creation"); 928 goto done_errmsg; 929 } 930 if (len == 0 || path[len - 1] != '\0') { 931 error = EINVAL; 932 goto done_free; 933 } 934 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, 935 path, td); 936 error = namei(&nd); 937 if (error) 938 goto done_free; 939 root = nd.ni_vp; 940 NDFREE(&nd, NDF_ONLY_PNBUF); 941 g_path = malloc(MAXPATHLEN, M_TEMP, M_WAITOK); 942 strlcpy(g_path, path, MAXPATHLEN); 943 error = vn_path_to_global_path(td, root, g_path, MAXPATHLEN); 944 if (error == 0) 945 path = g_path; 946 else if (error == ENODEV) { 947 /* proceed if sysctl debug.disablefullpath == 1 */ 948 fullpath_disabled = 1; 949 if (len < 2 || (len == 2 && path[0] == '/')) 950 path = NULL; 951 } else { 952 /* exit on other errors */ 953 goto done_free; 954 } 955 if (root->v_type != VDIR) { 956 error = ENOTDIR; 957 vput(root); 958 goto done_free; 959 } 960 VOP_UNLOCK(root, 0); 961 if (fullpath_disabled) { 962 /* Leave room for a real-root full pathname. */ 963 if (len + (path[0] == '/' && strcmp(mypr->pr_path, "/") 964 ? strlen(mypr->pr_path) : 0) > MAXPATHLEN) { 965 error = ENAMETOOLONG; 966 goto done_free; 967 } 968 } 969 } 970 971 error = vfs_getopt(opts, "osrelease", (void **)&osrelstr, &len); 972 if (error == ENOENT) 973 osrelstr = NULL; 974 else if (error != 0) 975 goto done_free; 976 else { 977 if (flags & JAIL_UPDATE) { 978 error = EINVAL; 979 vfs_opterror(opts, 980 "osrelease cannot be changed after creation"); 981 goto done_errmsg; 982 } 983 if (len == 0 || len >= OSRELEASELEN) { 984 error = EINVAL; 985 vfs_opterror(opts, 986 "osrelease string must be 1-%d bytes long", 987 OSRELEASELEN - 1); 988 goto done_errmsg; 989 } 990 } 991 992 error = vfs_copyopt(opts, "osreldate", &osreldt, sizeof(osreldt)); 993 if (error == ENOENT) 994 osreldt = 0; 995 else if (error != 0) 996 goto done_free; 997 else { 998 if (flags & JAIL_UPDATE) { 999 error = EINVAL; 1000 vfs_opterror(opts, 1001 "osreldate cannot be changed after creation"); 1002 goto done_errmsg; 1003 } 1004 if (osreldt == 0) { 1005 error = EINVAL; 1006 vfs_opterror(opts, "osreldate cannot be 0"); 1007 goto done_errmsg; 1008 } 1009 } 1010 1011 /* 1012 * Grab the allprison lock before letting modules check their 1013 * parameters. Once we have it, do not let go so we'll have a 1014 * consistent view of the OSD list. 1015 */ 1016 sx_xlock(&allprison_lock); 1017 error = osd_jail_call(NULL, PR_METHOD_CHECK, opts); 1018 if (error) 1019 goto done_unlock_list; 1020 1021 /* By now, all parameters should have been noted. */ 1022 TAILQ_FOREACH(opt, opts, link) { 1023 if (!opt->seen && strcmp(opt->name, "errmsg")) { 1024 error = EINVAL; 1025 vfs_opterror(opts, "unknown parameter: %s", opt->name); 1026 goto done_unlock_list; 1027 } 1028 } 1029 1030 /* 1031 * See if we are creating a new record or updating an existing one. 1032 * This abuses the file error codes ENOENT and EEXIST. 1033 */ 1034 cuflags = flags & (JAIL_CREATE | JAIL_UPDATE); 1035 if (!cuflags) { 1036 error = EINVAL; 1037 vfs_opterror(opts, "no valid operation (create or update)"); 1038 goto done_unlock_list; 1039 } 1040 pr = NULL; 1041 namelc = NULL; 1042 if (cuflags == JAIL_CREATE && jid == 0 && name != NULL) { 1043 namelc = strrchr(name, '.'); 1044 jid = strtoul(namelc != NULL ? namelc + 1 : name, &p, 10); 1045 if (*p != '\0') 1046 jid = 0; 1047 } 1048 if (jid != 0) { 1049 /* 1050 * See if a requested jid already exists. There is an 1051 * information leak here if the jid exists but is not within 1052 * the caller's jail hierarchy. Jail creators will get EEXIST 1053 * even though they cannot see the jail, and CREATE | UPDATE 1054 * will return ENOENT which is not normally a valid error. 1055 */ 1056 if (jid < 0) { 1057 error = EINVAL; 1058 vfs_opterror(opts, "negative jid"); 1059 goto done_unlock_list; 1060 } 1061 pr = prison_find(jid); 1062 if (pr != NULL) { 1063 ppr = pr->pr_parent; 1064 /* Create: jid must not exist. */ 1065 if (cuflags == JAIL_CREATE) { 1066 mtx_unlock(&pr->pr_mtx); 1067 error = EEXIST; 1068 vfs_opterror(opts, "jail %d already exists", 1069 jid); 1070 goto done_unlock_list; 1071 } 1072 if (!prison_ischild(mypr, pr)) { 1073 mtx_unlock(&pr->pr_mtx); 1074 pr = NULL; 1075 } else if (pr->pr_uref == 0) { 1076 if (!(flags & JAIL_DYING)) { 1077 mtx_unlock(&pr->pr_mtx); 1078 error = ENOENT; 1079 vfs_opterror(opts, "jail %d is dying", 1080 jid); 1081 goto done_unlock_list; 1082 } else if ((flags & JAIL_ATTACH) || 1083 (pr_flags & PR_PERSIST)) { 1084 /* 1085 * A dying jail might be resurrected 1086 * (via attach or persist), but first 1087 * it must determine if another jail 1088 * has claimed its name. Accomplish 1089 * this by implicitly re-setting the 1090 * name. 1091 */ 1092 if (name == NULL) 1093 name = prison_name(mypr, pr); 1094 } 1095 } 1096 } 1097 if (pr == NULL) { 1098 /* Update: jid must exist. */ 1099 if (cuflags == JAIL_UPDATE) { 1100 error = ENOENT; 1101 vfs_opterror(opts, "jail %d not found", jid); 1102 goto done_unlock_list; 1103 } 1104 } 1105 } 1106 /* 1107 * If the caller provided a name, look for a jail by that name. 1108 * This has different semantics for creates and updates keyed by jid 1109 * (where the name must not already exist in a different jail), 1110 * and updates keyed by the name itself (where the name must exist 1111 * because that is the jail being updated). 1112 */ 1113 if (name != NULL) { 1114 namelc = strrchr(name, '.'); 1115 if (namelc == NULL) 1116 namelc = name; 1117 else { 1118 /* 1119 * This is a hierarchical name. Split it into the 1120 * parent and child names, and make sure the parent 1121 * exists or matches an already found jail. 1122 */ 1123 *namelc = '\0'; 1124 if (pr != NULL) { 1125 if (strncmp(name, ppr->pr_name, namelc - name) 1126 || ppr->pr_name[namelc - name] != '\0') { 1127 mtx_unlock(&pr->pr_mtx); 1128 error = EINVAL; 1129 vfs_opterror(opts, 1130 "cannot change jail's parent"); 1131 goto done_unlock_list; 1132 } 1133 } else { 1134 ppr = prison_find_name(mypr, name); 1135 if (ppr == NULL) { 1136 error = ENOENT; 1137 vfs_opterror(opts, 1138 "jail \"%s\" not found", name); 1139 goto done_unlock_list; 1140 } 1141 mtx_unlock(&ppr->pr_mtx); 1142 } 1143 name = ++namelc; 1144 } 1145 if (name[0] != '\0') { 1146 namelen = 1147 (ppr == &prison0) ? 0 : strlen(ppr->pr_name) + 1; 1148 name_again: 1149 deadpr = NULL; 1150 FOREACH_PRISON_CHILD(ppr, tpr) { 1151 if (tpr != pr && tpr->pr_ref > 0 && 1152 !strcmp(tpr->pr_name + namelen, name)) { 1153 if (pr == NULL && 1154 cuflags != JAIL_CREATE) { 1155 mtx_lock(&tpr->pr_mtx); 1156 if (tpr->pr_ref > 0) { 1157 /* 1158 * Use this jail 1159 * for updates. 1160 */ 1161 if (tpr->pr_uref > 0) { 1162 pr = tpr; 1163 break; 1164 } 1165 deadpr = tpr; 1166 } 1167 mtx_unlock(&tpr->pr_mtx); 1168 } else if (tpr->pr_uref > 0) { 1169 /* 1170 * Create, or update(jid): 1171 * name must not exist in an 1172 * active sibling jail. 1173 */ 1174 error = EEXIST; 1175 if (pr != NULL) 1176 mtx_unlock(&pr->pr_mtx); 1177 vfs_opterror(opts, 1178 "jail \"%s\" already exists", 1179 name); 1180 goto done_unlock_list; 1181 } 1182 } 1183 } 1184 /* If no active jail is found, use a dying one. */ 1185 if (deadpr != NULL && pr == NULL) { 1186 if (flags & JAIL_DYING) { 1187 mtx_lock(&deadpr->pr_mtx); 1188 if (deadpr->pr_ref == 0) { 1189 mtx_unlock(&deadpr->pr_mtx); 1190 goto name_again; 1191 } 1192 pr = deadpr; 1193 } else if (cuflags == JAIL_UPDATE) { 1194 error = ENOENT; 1195 vfs_opterror(opts, 1196 "jail \"%s\" is dying", name); 1197 goto done_unlock_list; 1198 } 1199 } 1200 /* Update: name must exist if no jid. */ 1201 else if (cuflags == JAIL_UPDATE && pr == NULL) { 1202 error = ENOENT; 1203 vfs_opterror(opts, "jail \"%s\" not found", 1204 name); 1205 goto done_unlock_list; 1206 } 1207 } 1208 } 1209 /* Update: must provide a jid or name. */ 1210 else if (cuflags == JAIL_UPDATE && pr == NULL) { 1211 error = ENOENT; 1212 vfs_opterror(opts, "update specified no jail"); 1213 goto done_unlock_list; 1214 } 1215 1216 /* If there's no prison to update, create a new one and link it in. */ 1217 if (pr == NULL) { 1218 for (tpr = mypr; tpr != NULL; tpr = tpr->pr_parent) 1219 if (tpr->pr_childcount >= tpr->pr_childmax) { 1220 error = EPERM; 1221 vfs_opterror(opts, "prison limit exceeded"); 1222 goto done_unlock_list; 1223 } 1224 created = 1; 1225 mtx_lock(&ppr->pr_mtx); 1226 if (ppr->pr_ref == 0) { 1227 mtx_unlock(&ppr->pr_mtx); 1228 error = ENOENT; 1229 vfs_opterror(opts, "parent jail went away!"); 1230 goto done_unlock_list; 1231 } 1232 ppr->pr_ref++; 1233 ppr->pr_uref++; 1234 mtx_unlock(&ppr->pr_mtx); 1235 pr = malloc(sizeof(*pr), M_PRISON, M_WAITOK | M_ZERO); 1236 if (jid == 0) { 1237 /* Find the next free jid. */ 1238 jid = lastprid + 1; 1239 findnext: 1240 if (jid == JAIL_MAX) 1241 jid = 1; 1242 TAILQ_FOREACH(tpr, &allprison, pr_list) { 1243 if (tpr->pr_id < jid) 1244 continue; 1245 if (tpr->pr_id > jid || tpr->pr_ref == 0) { 1246 TAILQ_INSERT_BEFORE(tpr, pr, pr_list); 1247 break; 1248 } 1249 if (jid == lastprid) { 1250 error = EAGAIN; 1251 vfs_opterror(opts, 1252 "no available jail IDs"); 1253 free(pr, M_PRISON); 1254 prison_deref(ppr, PD_DEREF | 1255 PD_DEUREF | PD_LIST_XLOCKED); 1256 goto done_releroot; 1257 } 1258 jid++; 1259 goto findnext; 1260 } 1261 lastprid = jid; 1262 } else { 1263 /* 1264 * The jail already has a jid (that did not yet exist), 1265 * so just find where to insert it. 1266 */ 1267 TAILQ_FOREACH(tpr, &allprison, pr_list) 1268 if (tpr->pr_id >= jid) { 1269 TAILQ_INSERT_BEFORE(tpr, pr, pr_list); 1270 break; 1271 } 1272 } 1273 if (tpr == NULL) 1274 TAILQ_INSERT_TAIL(&allprison, pr, pr_list); 1275 LIST_INSERT_HEAD(&ppr->pr_children, pr, pr_sibling); 1276 for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent) 1277 tpr->pr_childcount++; 1278 1279 pr->pr_parent = ppr; 1280 pr->pr_id = jid; 1281 1282 /* Set some default values, and inherit some from the parent. */ 1283 if (name == NULL) 1284 name = ""; 1285 if (path == NULL) { 1286 path = "/"; 1287 root = mypr->pr_root; 1288 vref(root); 1289 } 1290 strlcpy(pr->pr_hostuuid, DEFAULT_HOSTUUID, HOSTUUIDLEN); 1291 pr->pr_flags |= PR_HOST; 1292#if defined(INET) || defined(INET6) 1293#ifdef VIMAGE 1294 if (!(pr_flags & PR_VNET)) 1295#endif 1296 { 1297#ifdef INET 1298 if (!(ch_flags & PR_IP4_USER)) 1299 pr->pr_flags |= PR_IP4 | PR_IP4_USER; 1300 else if (!(pr_flags & PR_IP4_USER)) { 1301 pr->pr_flags |= ppr->pr_flags & PR_IP4; 1302 if (ppr->pr_ip4 != NULL) { 1303 pr->pr_ip4s = ppr->pr_ip4s; 1304 pr->pr_ip4 = malloc(pr->pr_ip4s * 1305 sizeof(struct in_addr), M_PRISON, 1306 M_WAITOK); 1307 bcopy(ppr->pr_ip4, pr->pr_ip4, 1308 pr->pr_ip4s * sizeof(*pr->pr_ip4)); 1309 } 1310 } 1311#endif 1312#ifdef INET6 1313 if (!(ch_flags & PR_IP6_USER)) 1314 pr->pr_flags |= PR_IP6 | PR_IP6_USER; 1315 else if (!(pr_flags & PR_IP6_USER)) { 1316 pr->pr_flags |= ppr->pr_flags & PR_IP6; 1317 if (ppr->pr_ip6 != NULL) { 1318 pr->pr_ip6s = ppr->pr_ip6s; 1319 pr->pr_ip6 = malloc(pr->pr_ip6s * 1320 sizeof(struct in6_addr), M_PRISON, 1321 M_WAITOK); 1322 bcopy(ppr->pr_ip6, pr->pr_ip6, 1323 pr->pr_ip6s * sizeof(*pr->pr_ip6)); 1324 } 1325 } 1326#endif 1327 } 1328#endif 1329 /* Source address selection is always on by default. */ 1330 pr->pr_flags |= _PR_IP_SADDRSEL; 1331 1332 pr->pr_securelevel = ppr->pr_securelevel; 1333 pr->pr_allow = JAIL_DEFAULT_ALLOW & ppr->pr_allow; 1334 pr->pr_enforce_statfs = jail_default_enforce_statfs; 1335 pr->pr_devfs_rsnum = ppr->pr_devfs_rsnum; 1336 1337 pr->pr_osreldate = osreldt ? osreldt : ppr->pr_osreldate; 1338 if (osrelstr == NULL) 1339 strcpy(pr->pr_osrelease, ppr->pr_osrelease); 1340 else 1341 strcpy(pr->pr_osrelease, osrelstr); 1342 1343 LIST_INIT(&pr->pr_children); 1344 mtx_init(&pr->pr_mtx, "jail mutex", NULL, MTX_DEF | MTX_DUPOK); 1345 1346#ifdef VIMAGE 1347 /* Allocate a new vnet if specified. */ 1348 pr->pr_vnet = (pr_flags & PR_VNET) 1349 ? vnet_alloc() : ppr->pr_vnet; 1350#endif 1351 /* 1352 * Allocate a dedicated cpuset for each jail. 1353 * Unlike other initial settings, this may return an erorr. 1354 */ 1355 error = cpuset_create_root(ppr, &pr->pr_cpuset); 1356 if (error) { 1357 prison_deref(pr, PD_LIST_XLOCKED); 1358 goto done_releroot; 1359 } 1360 1361 mtx_lock(&pr->pr_mtx); 1362 /* 1363 * New prisons do not yet have a reference, because we do not 1364 * want other to see the incomplete prison once the 1365 * allprison_lock is downgraded. 1366 */ 1367 } else { 1368 created = 0; 1369 /* 1370 * Grab a reference for existing prisons, to ensure they 1371 * continue to exist for the duration of the call. 1372 */ 1373 pr->pr_ref++; 1374#if defined(VIMAGE) && (defined(INET) || defined(INET6)) 1375 if ((pr->pr_flags & PR_VNET) && 1376 (ch_flags & (PR_IP4_USER | PR_IP6_USER))) { 1377 error = EINVAL; 1378 vfs_opterror(opts, 1379 "vnet jails cannot have IP address restrictions"); 1380 goto done_deref_locked; 1381 } 1382#endif 1383#ifdef INET 1384 if (PR_IP4_USER & ch_flags & (pr_flags ^ pr->pr_flags)) { 1385 error = EINVAL; 1386 vfs_opterror(opts, 1387 "ip4 cannot be changed after creation"); 1388 goto done_deref_locked; 1389 } 1390#endif 1391#ifdef INET6 1392 if (PR_IP6_USER & ch_flags & (pr_flags ^ pr->pr_flags)) { 1393 error = EINVAL; 1394 vfs_opterror(opts, 1395 "ip6 cannot be changed after creation"); 1396 goto done_deref_locked; 1397 } 1398#endif 1399 } 1400 1401 /* Do final error checking before setting anything. */ 1402 if (gotslevel) { 1403 if (slevel < ppr->pr_securelevel) { 1404 error = EPERM; 1405 goto done_deref_locked; 1406 } 1407 } 1408 if (gotchildmax) { 1409 if (childmax >= ppr->pr_childmax) { 1410 error = EPERM; 1411 goto done_deref_locked; 1412 } 1413 } 1414 if (gotenforce) { 1415 if (enforce < ppr->pr_enforce_statfs) { 1416 error = EPERM; 1417 goto done_deref_locked; 1418 } 1419 } 1420 if (gotrsnum) { 1421 /* 1422 * devfs_rsnum is a uint16_t 1423 */ 1424 if (rsnum < 0 || rsnum > 65535) { 1425 error = EINVAL; 1426 goto done_deref_locked; 1427 } 1428 /* 1429 * Nested jails always inherit parent's devfs ruleset 1430 */ 1431 if (jailed(td->td_ucred)) { 1432 if (rsnum > 0 && rsnum != ppr->pr_devfs_rsnum) { 1433 error = EPERM; 1434 goto done_deref_locked; 1435 } else 1436 rsnum = ppr->pr_devfs_rsnum; 1437 } 1438 } 1439#ifdef INET 1440 if (ip4s > 0) { 1441 if (ppr->pr_flags & PR_IP4) { 1442 /* 1443 * Make sure the new set of IP addresses is a 1444 * subset of the parent's list. Don't worry 1445 * about the parent being unlocked, as any 1446 * setting is done with allprison_lock held. 1447 */ 1448 for (ij = 0; ij < ppr->pr_ip4s; ij++) 1449 if (ip4[0].s_addr == ppr->pr_ip4[ij].s_addr) 1450 break; 1451 if (ij == ppr->pr_ip4s) { 1452 error = EPERM; 1453 goto done_deref_locked; 1454 } 1455 if (ip4s > 1) { 1456 for (ii = ij = 1; ii < ip4s; ii++) { 1457 if (ip4[ii].s_addr == 1458 ppr->pr_ip4[0].s_addr) 1459 continue; 1460 for (; ij < ppr->pr_ip4s; ij++) 1461 if (ip4[ii].s_addr == 1462 ppr->pr_ip4[ij].s_addr) 1463 break; 1464 if (ij == ppr->pr_ip4s) 1465 break; 1466 } 1467 if (ij == ppr->pr_ip4s) { 1468 error = EPERM; 1469 goto done_deref_locked; 1470 } 1471 } 1472 } 1473 /* 1474 * Check for conflicting IP addresses. We permit them 1475 * if there is no more than one IP on each jail. If 1476 * there is a duplicate on a jail with more than one 1477 * IP stop checking and return error. 1478 */ 1479 tppr = ppr; 1480#ifdef VIMAGE 1481 for (; tppr != &prison0; tppr = tppr->pr_parent) 1482 if (tppr->pr_flags & PR_VNET) 1483 break; 1484#endif 1485 FOREACH_PRISON_DESCENDANT(tppr, tpr, descend) { 1486 if (tpr == pr || 1487#ifdef VIMAGE 1488 (tpr != tppr && (tpr->pr_flags & PR_VNET)) || 1489#endif 1490 tpr->pr_uref == 0) { 1491 descend = 0; 1492 continue; 1493 } 1494 if (!(tpr->pr_flags & PR_IP4_USER)) 1495 continue; 1496 descend = 0; 1497 if (tpr->pr_ip4 == NULL || 1498 (ip4s == 1 && tpr->pr_ip4s == 1)) 1499 continue; 1500 for (ii = 0; ii < ip4s; ii++) { 1501 if (_prison_check_ip4(tpr, &ip4[ii]) == 0) { 1502 error = EADDRINUSE; 1503 vfs_opterror(opts, 1504 "IPv4 addresses clash"); 1505 goto done_deref_locked; 1506 } 1507 } 1508 } 1509 } 1510#endif 1511#ifdef INET6 1512 if (ip6s > 0) { 1513 if (ppr->pr_flags & PR_IP6) { 1514 /* 1515 * Make sure the new set of IP addresses is a 1516 * subset of the parent's list. 1517 */ 1518 for (ij = 0; ij < ppr->pr_ip6s; ij++) 1519 if (IN6_ARE_ADDR_EQUAL(&ip6[0], 1520 &ppr->pr_ip6[ij])) 1521 break; 1522 if (ij == ppr->pr_ip6s) { 1523 error = EPERM; 1524 goto done_deref_locked; 1525 } 1526 if (ip6s > 1) { 1527 for (ii = ij = 1; ii < ip6s; ii++) { 1528 if (IN6_ARE_ADDR_EQUAL(&ip6[ii], 1529 &ppr->pr_ip6[0])) 1530 continue; 1531 for (; ij < ppr->pr_ip6s; ij++) 1532 if (IN6_ARE_ADDR_EQUAL( 1533 &ip6[ii], &ppr->pr_ip6[ij])) 1534 break; 1535 if (ij == ppr->pr_ip6s) 1536 break; 1537 } 1538 if (ij == ppr->pr_ip6s) { 1539 error = EPERM; 1540 goto done_deref_locked; 1541 } 1542 } 1543 } 1544 /* Check for conflicting IP addresses. */ 1545 tppr = ppr; 1546#ifdef VIMAGE 1547 for (; tppr != &prison0; tppr = tppr->pr_parent) 1548 if (tppr->pr_flags & PR_VNET) 1549 break; 1550#endif 1551 FOREACH_PRISON_DESCENDANT(tppr, tpr, descend) { 1552 if (tpr == pr || 1553#ifdef VIMAGE 1554 (tpr != tppr && (tpr->pr_flags & PR_VNET)) || 1555#endif 1556 tpr->pr_uref == 0) { 1557 descend = 0; 1558 continue; 1559 } 1560 if (!(tpr->pr_flags & PR_IP6_USER)) 1561 continue; 1562 descend = 0; 1563 if (tpr->pr_ip6 == NULL || 1564 (ip6s == 1 && tpr->pr_ip6s == 1)) 1565 continue; 1566 for (ii = 0; ii < ip6s; ii++) { 1567 if (_prison_check_ip6(tpr, &ip6[ii]) == 0) { 1568 error = EADDRINUSE; 1569 vfs_opterror(opts, 1570 "IPv6 addresses clash"); 1571 goto done_deref_locked; 1572 } 1573 } 1574 } 1575 } 1576#endif 1577 onamelen = namelen = 0; 1578 if (name != NULL) { 1579 /* Give a default name of the jid. Also allow the name to be 1580 * explicitly the jid - but not any other number, and only in 1581 * normal form (no leading zero/etc). 1582 */ 1583 if (name[0] == '\0') 1584 snprintf(name = numbuf, sizeof(numbuf), "%d", jid); 1585 else if ((strtoul(namelc, &p, 10) != jid || 1586 namelc[0] < '1' || namelc[0] > '9') && *p == '\0') { 1587 error = EINVAL; 1588 vfs_opterror(opts, 1589 "name cannot be numeric (unless it is the jid)"); 1590 goto done_deref_locked; 1591 } 1592 /* 1593 * Make sure the name isn't too long for the prison or its 1594 * children. 1595 */ 1596 onamelen = strlen(pr->pr_name); 1597 namelen = strlen(name); 1598 if (strlen(ppr->pr_name) + namelen + 2 > sizeof(pr->pr_name)) { 1599 error = ENAMETOOLONG; 1600 goto done_deref_locked; 1601 } 1602 FOREACH_PRISON_DESCENDANT(pr, tpr, descend) { 1603 if (strlen(tpr->pr_name) + (namelen - onamelen) >= 1604 sizeof(pr->pr_name)) { 1605 error = ENAMETOOLONG; 1606 goto done_deref_locked; 1607 } 1608 } 1609 } 1610 if (pr_allow & ~ppr->pr_allow) { 1611 error = EPERM; 1612 goto done_deref_locked; 1613 } 1614 1615 /* Set the parameters of the prison. */ 1616#ifdef INET 1617 redo_ip4 = 0; 1618 if (pr_flags & PR_IP4_USER) { 1619 pr->pr_flags |= PR_IP4; 1620 free(pr->pr_ip4, M_PRISON); 1621 pr->pr_ip4s = ip4s; 1622 pr->pr_ip4 = ip4; 1623 ip4 = NULL; 1624 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) { 1625#ifdef VIMAGE 1626 if (tpr->pr_flags & PR_VNET) { 1627 descend = 0; 1628 continue; 1629 } 1630#endif 1631 if (prison_restrict_ip4(tpr, NULL)) { 1632 redo_ip4 = 1; 1633 descend = 0; 1634 } 1635 } 1636 } 1637#endif 1638#ifdef INET6 1639 redo_ip6 = 0; 1640 if (pr_flags & PR_IP6_USER) { 1641 pr->pr_flags |= PR_IP6; 1642 free(pr->pr_ip6, M_PRISON); 1643 pr->pr_ip6s = ip6s; 1644 pr->pr_ip6 = ip6; 1645 ip6 = NULL; 1646 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) { 1647#ifdef VIMAGE 1648 if (tpr->pr_flags & PR_VNET) { 1649 descend = 0; 1650 continue; 1651 } 1652#endif 1653 if (prison_restrict_ip6(tpr, NULL)) { 1654 redo_ip6 = 1; 1655 descend = 0; 1656 } 1657 } 1658 } 1659#endif 1660 if (gotslevel) { 1661 pr->pr_securelevel = slevel; 1662 /* Set all child jails to be at least this level. */ 1663 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) 1664 if (tpr->pr_securelevel < slevel) 1665 tpr->pr_securelevel = slevel; 1666 } 1667 if (gotchildmax) { 1668 pr->pr_childmax = childmax; 1669 /* Set all child jails to under this limit. */ 1670 FOREACH_PRISON_DESCENDANT_LOCKED_LEVEL(pr, tpr, descend, level) 1671 if (tpr->pr_childmax > childmax - level) 1672 tpr->pr_childmax = childmax > level 1673 ? childmax - level : 0; 1674 } 1675 if (gotenforce) { 1676 pr->pr_enforce_statfs = enforce; 1677 /* Pass this restriction on to the children. */ 1678 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) 1679 if (tpr->pr_enforce_statfs < enforce) 1680 tpr->pr_enforce_statfs = enforce; 1681 } 1682 if (gotrsnum) { 1683 pr->pr_devfs_rsnum = rsnum; 1684 /* Pass this restriction on to the children. */ 1685 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) 1686 tpr->pr_devfs_rsnum = rsnum; 1687 } 1688 if (name != NULL) { 1689 if (ppr == &prison0) 1690 strlcpy(pr->pr_name, name, sizeof(pr->pr_name)); 1691 else 1692 snprintf(pr->pr_name, sizeof(pr->pr_name), "%s.%s", 1693 ppr->pr_name, name); 1694 /* Change this component of child names. */ 1695 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) { 1696 bcopy(tpr->pr_name + onamelen, tpr->pr_name + namelen, 1697 strlen(tpr->pr_name + onamelen) + 1); 1698 bcopy(pr->pr_name, tpr->pr_name, namelen); 1699 } 1700 } 1701 if (path != NULL) { 1702 /* Try to keep a real-rooted full pathname. */ 1703 if (fullpath_disabled && path[0] == '/' && 1704 strcmp(mypr->pr_path, "/")) 1705 snprintf(pr->pr_path, sizeof(pr->pr_path), "%s%s", 1706 mypr->pr_path, path); 1707 else 1708 strlcpy(pr->pr_path, path, sizeof(pr->pr_path)); 1709 pr->pr_root = root; 1710 } 1711 if (PR_HOST & ch_flags & ~pr_flags) { 1712 if (pr->pr_flags & PR_HOST) { 1713 /* 1714 * Copy the parent's host info. As with pr_ip4 above, 1715 * the lack of a lock on the parent is not a problem; 1716 * it is always set with allprison_lock at least 1717 * shared, and is held exclusively here. 1718 */ 1719 strlcpy(pr->pr_hostname, pr->pr_parent->pr_hostname, 1720 sizeof(pr->pr_hostname)); 1721 strlcpy(pr->pr_domainname, pr->pr_parent->pr_domainname, 1722 sizeof(pr->pr_domainname)); 1723 strlcpy(pr->pr_hostuuid, pr->pr_parent->pr_hostuuid, 1724 sizeof(pr->pr_hostuuid)); 1725 pr->pr_hostid = pr->pr_parent->pr_hostid; 1726 } 1727 } else if (host != NULL || domain != NULL || uuid != NULL || gothid) { 1728 /* Set this prison, and any descendants without PR_HOST. */ 1729 if (host != NULL) 1730 strlcpy(pr->pr_hostname, host, sizeof(pr->pr_hostname)); 1731 if (domain != NULL) 1732 strlcpy(pr->pr_domainname, domain, 1733 sizeof(pr->pr_domainname)); 1734 if (uuid != NULL) 1735 strlcpy(pr->pr_hostuuid, uuid, sizeof(pr->pr_hostuuid)); 1736 if (gothid) 1737 pr->pr_hostid = hid; 1738 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) { 1739 if (tpr->pr_flags & PR_HOST) 1740 descend = 0; 1741 else { 1742 if (host != NULL) 1743 strlcpy(tpr->pr_hostname, 1744 pr->pr_hostname, 1745 sizeof(tpr->pr_hostname)); 1746 if (domain != NULL) 1747 strlcpy(tpr->pr_domainname, 1748 pr->pr_domainname, 1749 sizeof(tpr->pr_domainname)); 1750 if (uuid != NULL) 1751 strlcpy(tpr->pr_hostuuid, 1752 pr->pr_hostuuid, 1753 sizeof(tpr->pr_hostuuid)); 1754 if (gothid) 1755 tpr->pr_hostid = hid; 1756 } 1757 } 1758 } 1759 if ((tallow = ch_allow & ~pr_allow)) { 1760 /* Clear allow bits in all children. */ 1761 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) 1762 tpr->pr_allow &= ~tallow; 1763 } 1764 pr->pr_allow = (pr->pr_allow & ~ch_allow) | pr_allow; 1765 /* 1766 * Persistent prisons get an extra reference, and prisons losing their 1767 * persist flag lose that reference. Only do this for existing prisons 1768 * for now, so new ones will remain unseen until after the module 1769 * handlers have completed. 1770 */ 1771 born = pr->pr_uref == 0; 1772 if (!created && (ch_flags & PR_PERSIST & (pr_flags ^ pr->pr_flags))) { 1773 if (pr_flags & PR_PERSIST) { 1774 pr->pr_ref++; 1775 pr->pr_uref++; 1776 } else { 1777 pr->pr_ref--; 1778 pr->pr_uref--; 1779 } 1780 } 1781 pr->pr_flags = (pr->pr_flags & ~ch_flags) | pr_flags; 1782 mtx_unlock(&pr->pr_mtx); 1783 1784#ifdef RACCT 1785 if (racct_enable && created) 1786 prison_racct_attach(pr); 1787#endif 1788 1789 /* Locks may have prevented a complete restriction of child IP 1790 * addresses. If so, allocate some more memory and try again. 1791 */ 1792#ifdef INET 1793 while (redo_ip4) { 1794 ip4s = pr->pr_ip4s; 1795 ip4 = malloc(ip4s * sizeof(*ip4), M_PRISON, M_WAITOK); 1796 mtx_lock(&pr->pr_mtx); 1797 redo_ip4 = 0; 1798 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) { 1799#ifdef VIMAGE 1800 if (tpr->pr_flags & PR_VNET) { 1801 descend = 0; 1802 continue; 1803 } 1804#endif 1805 if (prison_restrict_ip4(tpr, ip4)) { 1806 if (ip4 != NULL) 1807 ip4 = NULL; 1808 else 1809 redo_ip4 = 1; 1810 } 1811 } 1812 mtx_unlock(&pr->pr_mtx); 1813 } 1814#endif 1815#ifdef INET6 1816 while (redo_ip6) { 1817 ip6s = pr->pr_ip6s; 1818 ip6 = malloc(ip6s * sizeof(*ip6), M_PRISON, M_WAITOK); 1819 mtx_lock(&pr->pr_mtx); 1820 redo_ip6 = 0; 1821 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) { 1822#ifdef VIMAGE 1823 if (tpr->pr_flags & PR_VNET) { 1824 descend = 0; 1825 continue; 1826 } 1827#endif 1828 if (prison_restrict_ip6(tpr, ip6)) { 1829 if (ip6 != NULL) 1830 ip6 = NULL; 1831 else 1832 redo_ip6 = 1; 1833 } 1834 } 1835 mtx_unlock(&pr->pr_mtx); 1836 } 1837#endif 1838 1839 /* Let the modules do their work. */ 1840 sx_downgrade(&allprison_lock); 1841 if (born) { 1842 error = osd_jail_call(pr, PR_METHOD_CREATE, opts); 1843 if (error) { 1844 (void)osd_jail_call(pr, PR_METHOD_REMOVE, NULL); 1845 prison_deref(pr, created 1846 ? PD_LIST_SLOCKED 1847 : PD_DEREF | PD_LIST_SLOCKED); 1848 goto done_errmsg; 1849 } 1850 } 1851 error = osd_jail_call(pr, PR_METHOD_SET, opts); 1852 if (error) { 1853 if (born) 1854 (void)osd_jail_call(pr, PR_METHOD_REMOVE, NULL); 1855 prison_deref(pr, created 1856 ? PD_LIST_SLOCKED 1857 : PD_DEREF | PD_LIST_SLOCKED); 1858 goto done_errmsg; 1859 } 1860 1861 /* Attach this process to the prison if requested. */ 1862 if (flags & JAIL_ATTACH) { 1863 mtx_lock(&pr->pr_mtx); 1864 error = do_jail_attach(td, pr); 1865 if (error) { 1866 vfs_opterror(opts, "attach failed"); 1867 if (!created) 1868 prison_deref(pr, PD_DEREF); 1869 goto done_errmsg; 1870 } 1871 } 1872 1873#ifdef RACCT 1874 if (racct_enable && !created) { 1875 if (!(flags & JAIL_ATTACH)) 1876 sx_sunlock(&allprison_lock); 1877 prison_racct_modify(pr); 1878 if (!(flags & JAIL_ATTACH)) 1879 sx_slock(&allprison_lock); 1880 } 1881#endif 1882 1883 td->td_retval[0] = pr->pr_id; 1884 1885 /* 1886 * Now that it is all there, drop the temporary reference from existing 1887 * prisons. Or add a reference to newly created persistent prisons 1888 * (which was not done earlier so that the prison would not be publicly 1889 * visible). 1890 */ 1891 if (!created) { 1892 prison_deref(pr, (flags & JAIL_ATTACH) 1893 ? PD_DEREF 1894 : PD_DEREF | PD_LIST_SLOCKED); 1895 } else { 1896 if (pr_flags & PR_PERSIST) { 1897 mtx_lock(&pr->pr_mtx); 1898 pr->pr_ref++; 1899 pr->pr_uref++; 1900 mtx_unlock(&pr->pr_mtx); 1901 } 1902 if (!(flags & JAIL_ATTACH)) 1903 sx_sunlock(&allprison_lock); 1904 } 1905 1906 goto done_free; 1907 1908 done_deref_locked: 1909 prison_deref(pr, created 1910 ? PD_LOCKED | PD_LIST_XLOCKED 1911 : PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED); 1912 goto done_releroot; 1913 done_unlock_list: 1914 sx_xunlock(&allprison_lock); 1915 done_releroot: 1916 if (root != NULL) 1917 vrele(root); 1918 done_errmsg: 1919 if (error) { 1920 vfs_getopt(opts, "errmsg", (void **)&errmsg, &errmsg_len); 1921 if (errmsg_len > 0) { 1922 errmsg_pos = 2 * vfs_getopt_pos(opts, "errmsg") + 1; 1923 if (errmsg_pos > 0) { 1924 if (optuio->uio_segflg == UIO_SYSSPACE) 1925 bcopy(errmsg, 1926 optuio->uio_iov[errmsg_pos].iov_base, 1927 errmsg_len); 1928 else 1929 copyout(errmsg, 1930 optuio->uio_iov[errmsg_pos].iov_base, 1931 errmsg_len); 1932 } 1933 } 1934 } 1935 done_free: 1936#ifdef INET 1937 free(ip4, M_PRISON); 1938#endif 1939#ifdef INET6 1940 free(ip6, M_PRISON); 1941#endif 1942 if (g_path != NULL) 1943 free(g_path, M_TEMP); 1944 vfs_freeopts(opts); 1945 return (error); 1946} 1947 1948 1949/* 1950 * struct jail_get_args { 1951 * struct iovec *iovp; 1952 * unsigned int iovcnt; 1953 * int flags; 1954 * }; 1955 */ 1956int 1957sys_jail_get(struct thread *td, struct jail_get_args *uap) 1958{ 1959 struct uio *auio; 1960 int error; 1961 1962 /* Check that we have an even number of iovecs. */ 1963 if (uap->iovcnt & 1) 1964 return (EINVAL); 1965 1966 error = copyinuio(uap->iovp, uap->iovcnt, &auio); 1967 if (error) 1968 return (error); 1969 error = kern_jail_get(td, auio, uap->flags); 1970 if (error == 0) 1971 error = copyout(auio->uio_iov, uap->iovp, 1972 uap->iovcnt * sizeof (struct iovec)); 1973 free(auio, M_IOV); 1974 return (error); 1975} 1976 1977int 1978kern_jail_get(struct thread *td, struct uio *optuio, int flags) 1979{ 1980 struct prison *pr, *mypr; 1981 struct vfsopt *opt; 1982 struct vfsoptlist *opts; 1983 char *errmsg, *name; 1984 int error, errmsg_len, errmsg_pos, fi, i, jid, len, locked, pos; 1985 1986 if (flags & ~JAIL_GET_MASK) 1987 return (EINVAL); 1988 1989 /* Get the parameter list. */ 1990 error = vfs_buildopts(optuio, &opts); 1991 if (error) 1992 return (error); 1993 errmsg_pos = vfs_getopt_pos(opts, "errmsg"); 1994 mypr = td->td_ucred->cr_prison; 1995 1996 /* 1997 * Find the prison specified by one of: lastjid, jid, name. 1998 */ 1999 sx_slock(&allprison_lock); 2000 error = vfs_copyopt(opts, "lastjid", &jid, sizeof(jid)); 2001 if (error == 0) { 2002 TAILQ_FOREACH(pr, &allprison, pr_list) { 2003 if (pr->pr_id > jid && prison_ischild(mypr, pr)) { 2004 mtx_lock(&pr->pr_mtx); 2005 if (pr->pr_ref > 0 && 2006 (pr->pr_uref > 0 || (flags & JAIL_DYING))) 2007 break; 2008 mtx_unlock(&pr->pr_mtx); 2009 } 2010 } 2011 if (pr != NULL) 2012 goto found_prison; 2013 error = ENOENT; 2014 vfs_opterror(opts, "no jail after %d", jid); 2015 goto done_unlock_list; 2016 } else if (error != ENOENT) 2017 goto done_unlock_list; 2018 2019 error = vfs_copyopt(opts, "jid", &jid, sizeof(jid)); 2020 if (error == 0) { 2021 if (jid != 0) { 2022 pr = prison_find_child(mypr, jid); 2023 if (pr != NULL) { 2024 if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) { 2025 mtx_unlock(&pr->pr_mtx); 2026 error = ENOENT; 2027 vfs_opterror(opts, "jail %d is dying", 2028 jid); 2029 goto done_unlock_list; 2030 } 2031 goto found_prison; 2032 } 2033 error = ENOENT; 2034 vfs_opterror(opts, "jail %d not found", jid); 2035 goto done_unlock_list; 2036 } 2037 } else if (error != ENOENT) 2038 goto done_unlock_list; 2039 2040 error = vfs_getopt(opts, "name", (void **)&name, &len); 2041 if (error == 0) { 2042 if (len == 0 || name[len - 1] != '\0') { 2043 error = EINVAL; 2044 goto done_unlock_list; 2045 } 2046 pr = prison_find_name(mypr, name); 2047 if (pr != NULL) { 2048 if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) { 2049 mtx_unlock(&pr->pr_mtx); 2050 error = ENOENT; 2051 vfs_opterror(opts, "jail \"%s\" is dying", 2052 name); 2053 goto done_unlock_list; 2054 } 2055 goto found_prison; 2056 } 2057 error = ENOENT; 2058 vfs_opterror(opts, "jail \"%s\" not found", name); 2059 goto done_unlock_list; 2060 } else if (error != ENOENT) 2061 goto done_unlock_list; 2062 2063 vfs_opterror(opts, "no jail specified"); 2064 error = ENOENT; 2065 goto done_unlock_list; 2066 2067 found_prison: 2068 /* Get the parameters of the prison. */ 2069 pr->pr_ref++; 2070 locked = PD_LOCKED; 2071 td->td_retval[0] = pr->pr_id; 2072 error = vfs_setopt(opts, "jid", &pr->pr_id, sizeof(pr->pr_id)); 2073 if (error != 0 && error != ENOENT) 2074 goto done_deref; 2075 i = (pr->pr_parent == mypr) ? 0 : pr->pr_parent->pr_id; 2076 error = vfs_setopt(opts, "parent", &i, sizeof(i)); 2077 if (error != 0 && error != ENOENT) 2078 goto done_deref; 2079 error = vfs_setopts(opts, "name", prison_name(mypr, pr)); 2080 if (error != 0 && error != ENOENT) 2081 goto done_deref; 2082 error = vfs_setopt(opts, "cpuset.id", &pr->pr_cpuset->cs_id, 2083 sizeof(pr->pr_cpuset->cs_id)); 2084 if (error != 0 && error != ENOENT) 2085 goto done_deref; 2086 error = vfs_setopts(opts, "path", prison_path(mypr, pr)); 2087 if (error != 0 && error != ENOENT) 2088 goto done_deref; 2089#ifdef INET 2090 error = vfs_setopt_part(opts, "ip4.addr", pr->pr_ip4, 2091 pr->pr_ip4s * sizeof(*pr->pr_ip4)); 2092 if (error != 0 && error != ENOENT) 2093 goto done_deref; 2094#endif 2095#ifdef INET6 2096 error = vfs_setopt_part(opts, "ip6.addr", pr->pr_ip6, 2097 pr->pr_ip6s * sizeof(*pr->pr_ip6)); 2098 if (error != 0 && error != ENOENT) 2099 goto done_deref; 2100#endif 2101 error = vfs_setopt(opts, "securelevel", &pr->pr_securelevel, 2102 sizeof(pr->pr_securelevel)); 2103 if (error != 0 && error != ENOENT) 2104 goto done_deref; 2105 error = vfs_setopt(opts, "children.cur", &pr->pr_childcount, 2106 sizeof(pr->pr_childcount)); 2107 if (error != 0 && error != ENOENT) 2108 goto done_deref; 2109 error = vfs_setopt(opts, "children.max", &pr->pr_childmax, 2110 sizeof(pr->pr_childmax)); 2111 if (error != 0 && error != ENOENT) 2112 goto done_deref; 2113 error = vfs_setopts(opts, "host.hostname", pr->pr_hostname); 2114 if (error != 0 && error != ENOENT) 2115 goto done_deref; 2116 error = vfs_setopts(opts, "host.domainname", pr->pr_domainname); 2117 if (error != 0 && error != ENOENT) 2118 goto done_deref; 2119 error = vfs_setopts(opts, "host.hostuuid", pr->pr_hostuuid); 2120 if (error != 0 && error != ENOENT) 2121 goto done_deref; 2122#ifdef COMPAT_FREEBSD32 2123 if (SV_PROC_FLAG(td->td_proc, SV_ILP32)) { 2124 uint32_t hid32 = pr->pr_hostid; 2125 2126 error = vfs_setopt(opts, "host.hostid", &hid32, sizeof(hid32)); 2127 } else 2128#endif 2129 error = vfs_setopt(opts, "host.hostid", &pr->pr_hostid, 2130 sizeof(pr->pr_hostid)); 2131 if (error != 0 && error != ENOENT) 2132 goto done_deref; 2133 error = vfs_setopt(opts, "enforce_statfs", &pr->pr_enforce_statfs, 2134 sizeof(pr->pr_enforce_statfs)); 2135 if (error != 0 && error != ENOENT) 2136 goto done_deref; 2137 error = vfs_setopt(opts, "devfs_ruleset", &pr->pr_devfs_rsnum, 2138 sizeof(pr->pr_devfs_rsnum)); 2139 if (error != 0 && error != ENOENT) 2140 goto done_deref; 2141 for (fi = 0; fi < nitems(pr_flag_names); fi++) { 2142 if (pr_flag_names[fi] == NULL) 2143 continue; 2144 i = (pr->pr_flags & (1 << fi)) ? 1 : 0; 2145 error = vfs_setopt(opts, pr_flag_names[fi], &i, sizeof(i)); 2146 if (error != 0 && error != ENOENT) 2147 goto done_deref; 2148 i = !i; 2149 error = vfs_setopt(opts, pr_flag_nonames[fi], &i, sizeof(i)); 2150 if (error != 0 && error != ENOENT) 2151 goto done_deref; 2152 } 2153 for (fi = 0; fi < nitems(pr_flag_jailsys); fi++) { 2154 i = pr->pr_flags & 2155 (pr_flag_jailsys[fi].disable | pr_flag_jailsys[fi].new); 2156 i = pr_flag_jailsys[fi].disable && 2157 (i == pr_flag_jailsys[fi].disable) ? JAIL_SYS_DISABLE 2158 : (i == pr_flag_jailsys[fi].new) ? JAIL_SYS_NEW 2159 : JAIL_SYS_INHERIT; 2160 error = 2161 vfs_setopt(opts, pr_flag_jailsys[fi].name, &i, sizeof(i)); 2162 if (error != 0 && error != ENOENT) 2163 goto done_deref; 2164 } 2165 for (fi = 0; fi < nitems(pr_allow_names); fi++) { 2166 if (pr_allow_names[fi] == NULL) 2167 continue; 2168 i = (pr->pr_allow & (1 << fi)) ? 1 : 0; 2169 error = vfs_setopt(opts, pr_allow_names[fi], &i, sizeof(i)); 2170 if (error != 0 && error != ENOENT) 2171 goto done_deref; 2172 i = !i; 2173 error = vfs_setopt(opts, pr_allow_nonames[fi], &i, sizeof(i)); 2174 if (error != 0 && error != ENOENT) 2175 goto done_deref; 2176 } 2177 i = (pr->pr_uref == 0); 2178 error = vfs_setopt(opts, "dying", &i, sizeof(i)); 2179 if (error != 0 && error != ENOENT) 2180 goto done_deref; 2181 i = !i; 2182 error = vfs_setopt(opts, "nodying", &i, sizeof(i)); 2183 if (error != 0 && error != ENOENT) 2184 goto done_deref; 2185 error = vfs_setopt(opts, "osreldate", &pr->pr_osreldate, 2186 sizeof(pr->pr_osreldate)); 2187 if (error != 0 && error != ENOENT) 2188 goto done_deref; 2189 error = vfs_setopts(opts, "osrelease", pr->pr_osrelease); 2190 if (error != 0 && error != ENOENT) 2191 goto done_deref; 2192 2193 /* Get the module parameters. */ 2194 mtx_unlock(&pr->pr_mtx); 2195 locked = 0; 2196 error = osd_jail_call(pr, PR_METHOD_GET, opts); 2197 if (error) 2198 goto done_deref; 2199 prison_deref(pr, PD_DEREF | PD_LIST_SLOCKED); 2200 2201 /* By now, all parameters should have been noted. */ 2202 TAILQ_FOREACH(opt, opts, link) { 2203 if (!opt->seen && strcmp(opt->name, "errmsg")) { 2204 error = EINVAL; 2205 vfs_opterror(opts, "unknown parameter: %s", opt->name); 2206 goto done_errmsg; 2207 } 2208 } 2209 2210 /* Write the fetched parameters back to userspace. */ 2211 error = 0; 2212 TAILQ_FOREACH(opt, opts, link) { 2213 if (opt->pos >= 0 && opt->pos != errmsg_pos) { 2214 pos = 2 * opt->pos + 1; 2215 optuio->uio_iov[pos].iov_len = opt->len; 2216 if (opt->value != NULL) { 2217 if (optuio->uio_segflg == UIO_SYSSPACE) { 2218 bcopy(opt->value, 2219 optuio->uio_iov[pos].iov_base, 2220 opt->len); 2221 } else { 2222 error = copyout(opt->value, 2223 optuio->uio_iov[pos].iov_base, 2224 opt->len); 2225 if (error) 2226 break; 2227 } 2228 } 2229 } 2230 } 2231 goto done_errmsg; 2232 2233 done_deref: 2234 prison_deref(pr, locked | PD_DEREF | PD_LIST_SLOCKED); 2235 goto done_errmsg; 2236 2237 done_unlock_list: 2238 sx_sunlock(&allprison_lock); 2239 done_errmsg: 2240 if (error && errmsg_pos >= 0) { 2241 vfs_getopt(opts, "errmsg", (void **)&errmsg, &errmsg_len); 2242 errmsg_pos = 2 * errmsg_pos + 1; 2243 if (errmsg_len > 0) { 2244 if (optuio->uio_segflg == UIO_SYSSPACE) 2245 bcopy(errmsg, 2246 optuio->uio_iov[errmsg_pos].iov_base, 2247 errmsg_len); 2248 else 2249 copyout(errmsg, 2250 optuio->uio_iov[errmsg_pos].iov_base, 2251 errmsg_len); 2252 } 2253 } 2254 vfs_freeopts(opts); 2255 return (error); 2256} 2257 2258 2259/* 2260 * struct jail_remove_args { 2261 * int jid; 2262 * }; 2263 */ 2264int 2265sys_jail_remove(struct thread *td, struct jail_remove_args *uap) 2266{ 2267 struct prison *pr, *cpr, *lpr, *tpr; 2268 int descend, error; 2269 2270 error = priv_check(td, PRIV_JAIL_REMOVE); 2271 if (error) 2272 return (error); 2273 2274 sx_xlock(&allprison_lock); 2275 pr = prison_find_child(td->td_ucred->cr_prison, uap->jid); 2276 if (pr == NULL) { 2277 sx_xunlock(&allprison_lock); 2278 return (EINVAL); 2279 } 2280 2281 /* Remove all descendants of this prison, then remove this prison. */ 2282 pr->pr_ref++; 2283 if (!LIST_EMPTY(&pr->pr_children)) { 2284 mtx_unlock(&pr->pr_mtx); 2285 lpr = NULL; 2286 FOREACH_PRISON_DESCENDANT(pr, cpr, descend) { 2287 mtx_lock(&cpr->pr_mtx); 2288 if (cpr->pr_ref > 0) { 2289 tpr = cpr; 2290 cpr->pr_ref++; 2291 } else { 2292 /* Already removed - do not do it again. */ 2293 tpr = NULL; 2294 } 2295 mtx_unlock(&cpr->pr_mtx); 2296 if (lpr != NULL) { 2297 mtx_lock(&lpr->pr_mtx); 2298 prison_remove_one(lpr); 2299 sx_xlock(&allprison_lock); 2300 } 2301 lpr = tpr; 2302 } 2303 if (lpr != NULL) { 2304 mtx_lock(&lpr->pr_mtx); 2305 prison_remove_one(lpr); 2306 sx_xlock(&allprison_lock); 2307 } 2308 mtx_lock(&pr->pr_mtx); 2309 } 2310 prison_remove_one(pr); 2311 return (0); 2312} 2313 2314static void 2315prison_remove_one(struct prison *pr) 2316{ 2317 struct proc *p; 2318 int deuref; 2319 2320 /* If the prison was persistent, it is not anymore. */ 2321 deuref = 0; 2322 if (pr->pr_flags & PR_PERSIST) { 2323 pr->pr_ref--; 2324 deuref = PD_DEUREF; 2325 pr->pr_flags &= ~PR_PERSIST; 2326 } 2327 2328 /* 2329 * jail_remove added a reference. If that's the only one, remove 2330 * the prison now. 2331 */ 2332 KASSERT(pr->pr_ref > 0, 2333 ("prison_remove_one removing a dead prison (jid=%d)", pr->pr_id)); 2334 if (pr->pr_ref == 1) { 2335 prison_deref(pr, 2336 deuref | PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED); 2337 return; 2338 } 2339 2340 mtx_unlock(&pr->pr_mtx); 2341 sx_xunlock(&allprison_lock); 2342 /* 2343 * Kill all processes unfortunate enough to be attached to this prison. 2344 */ 2345 sx_slock(&allproc_lock); 2346 LIST_FOREACH(p, &allproc, p_list) { 2347 PROC_LOCK(p); 2348 if (p->p_state != PRS_NEW && p->p_ucred && 2349 p->p_ucred->cr_prison == pr) 2350 kern_psignal(p, SIGKILL); 2351 PROC_UNLOCK(p); 2352 } 2353 sx_sunlock(&allproc_lock); 2354 /* Remove the temporary reference added by jail_remove. */ 2355 prison_deref(pr, deuref | PD_DEREF); 2356} 2357 2358 2359/* 2360 * struct jail_attach_args { 2361 * int jid; 2362 * }; 2363 */ 2364int 2365sys_jail_attach(struct thread *td, struct jail_attach_args *uap) 2366{ 2367 struct prison *pr; 2368 int error; 2369 2370 error = priv_check(td, PRIV_JAIL_ATTACH); 2371 if (error) 2372 return (error); 2373 2374 sx_slock(&allprison_lock); 2375 pr = prison_find_child(td->td_ucred->cr_prison, uap->jid); 2376 if (pr == NULL) { 2377 sx_sunlock(&allprison_lock); 2378 return (EINVAL); 2379 } 2380 2381 /* 2382 * Do not allow a process to attach to a prison that is not 2383 * considered to be "alive". 2384 */ 2385 if (pr->pr_uref == 0) { 2386 mtx_unlock(&pr->pr_mtx); 2387 sx_sunlock(&allprison_lock); 2388 return (EINVAL); 2389 } 2390 2391 return (do_jail_attach(td, pr)); 2392} 2393 2394static int 2395do_jail_attach(struct thread *td, struct prison *pr) 2396{ 2397 struct prison *ppr; 2398 struct proc *p; 2399 struct ucred *newcred, *oldcred; 2400 int error; 2401 2402 /* 2403 * XXX: Note that there is a slight race here if two threads 2404 * in the same privileged process attempt to attach to two 2405 * different jails at the same time. It is important for 2406 * user processes not to do this, or they might end up with 2407 * a process root from one prison, but attached to the jail 2408 * of another. 2409 */ 2410 pr->pr_ref++; 2411 pr->pr_uref++; 2412 mtx_unlock(&pr->pr_mtx); 2413 2414 /* Let modules do whatever they need to prepare for attaching. */ 2415 error = osd_jail_call(pr, PR_METHOD_ATTACH, td); 2416 if (error) { 2417 prison_deref(pr, PD_DEREF | PD_DEUREF | PD_LIST_SLOCKED); 2418 return (error); 2419 } 2420 sx_sunlock(&allprison_lock); 2421 2422 /* 2423 * Reparent the newly attached process to this jail. 2424 */ 2425 ppr = td->td_ucred->cr_prison; 2426 p = td->td_proc; 2427 error = cpuset_setproc_update_set(p, pr->pr_cpuset); 2428 if (error) 2429 goto e_revert_osd; 2430 2431 vn_lock(pr->pr_root, LK_EXCLUSIVE | LK_RETRY); 2432 if ((error = change_dir(pr->pr_root, td)) != 0) 2433 goto e_unlock; 2434#ifdef MAC 2435 if ((error = mac_vnode_check_chroot(td->td_ucred, pr->pr_root))) 2436 goto e_unlock; 2437#endif 2438 VOP_UNLOCK(pr->pr_root, 0); 2439 if ((error = pwd_chroot(td, pr->pr_root))) 2440 goto e_revert_osd; 2441 2442 newcred = crget(); 2443 PROC_LOCK(p); 2444 oldcred = p->p_ucred; 2445 setsugid(p); 2446 crcopy(newcred, oldcred); 2447 newcred->cr_prison = pr; 2448 proc_set_cred(p, newcred); 2449 PROC_UNLOCK(p); 2450#ifdef RACCT 2451 racct_proc_ucred_changed(p, oldcred, newcred); 2452#endif 2453 crfree(oldcred); 2454 prison_deref(ppr, PD_DEREF | PD_DEUREF); 2455 return (0); 2456 e_unlock: 2457 VOP_UNLOCK(pr->pr_root, 0); 2458 e_revert_osd: 2459 /* Tell modules this thread is still in its old jail after all. */ 2460 (void)osd_jail_call(ppr, PR_METHOD_ATTACH, td); 2461 prison_deref(pr, PD_DEREF | PD_DEUREF); 2462 return (error); 2463} 2464 2465 2466/* 2467 * Returns a locked prison instance, or NULL on failure. 2468 */ 2469struct prison * 2470prison_find(int prid) 2471{ 2472 struct prison *pr; 2473 2474 sx_assert(&allprison_lock, SX_LOCKED); 2475 TAILQ_FOREACH(pr, &allprison, pr_list) { 2476 if (pr->pr_id == prid) { 2477 mtx_lock(&pr->pr_mtx); 2478 if (pr->pr_ref > 0) 2479 return (pr); 2480 mtx_unlock(&pr->pr_mtx); 2481 } 2482 } 2483 return (NULL); 2484} 2485 2486/* 2487 * Find a prison that is a descendant of mypr. Returns a locked prison or NULL. 2488 */ 2489struct prison * 2490prison_find_child(struct prison *mypr, int prid) 2491{ 2492 struct prison *pr; 2493 int descend; 2494 2495 sx_assert(&allprison_lock, SX_LOCKED); 2496 FOREACH_PRISON_DESCENDANT(mypr, pr, descend) { 2497 if (pr->pr_id == prid) { 2498 mtx_lock(&pr->pr_mtx); 2499 if (pr->pr_ref > 0) 2500 return (pr); 2501 mtx_unlock(&pr->pr_mtx); 2502 } 2503 } 2504 return (NULL); 2505} 2506 2507/* 2508 * Look for the name relative to mypr. Returns a locked prison or NULL. 2509 */ 2510struct prison * 2511prison_find_name(struct prison *mypr, const char *name) 2512{ 2513 struct prison *pr, *deadpr; 2514 size_t mylen; 2515 int descend; 2516 2517 sx_assert(&allprison_lock, SX_LOCKED); 2518 mylen = (mypr == &prison0) ? 0 : strlen(mypr->pr_name) + 1; 2519 again: 2520 deadpr = NULL; 2521 FOREACH_PRISON_DESCENDANT(mypr, pr, descend) { 2522 if (!strcmp(pr->pr_name + mylen, name)) { 2523 mtx_lock(&pr->pr_mtx); 2524 if (pr->pr_ref > 0) { 2525 if (pr->pr_uref > 0) 2526 return (pr); 2527 deadpr = pr; 2528 } 2529 mtx_unlock(&pr->pr_mtx); 2530 } 2531 } 2532 /* There was no valid prison - perhaps there was a dying one. */ 2533 if (deadpr != NULL) { 2534 mtx_lock(&deadpr->pr_mtx); 2535 if (deadpr->pr_ref == 0) { 2536 mtx_unlock(&deadpr->pr_mtx); 2537 goto again; 2538 } 2539 } 2540 return (deadpr); 2541} 2542 2543/* 2544 * See if a prison has the specific flag set. 2545 */ 2546int 2547prison_flag(struct ucred *cred, unsigned flag) 2548{ 2549 2550 /* This is an atomic read, so no locking is necessary. */ 2551 return (cred->cr_prison->pr_flags & flag); 2552} 2553 2554int 2555prison_allow(struct ucred *cred, unsigned flag) 2556{ 2557 2558 /* This is an atomic read, so no locking is necessary. */ 2559 return (cred->cr_prison->pr_allow & flag); 2560} 2561 2562/* 2563 * Remove a prison reference. If that was the last reference, remove the 2564 * prison itself - but not in this context in case there are locks held. 2565 */ 2566void 2567prison_free_locked(struct prison *pr) 2568{ 2569 2570 mtx_assert(&pr->pr_mtx, MA_OWNED); 2571 pr->pr_ref--; 2572 if (pr->pr_ref == 0) { 2573 mtx_unlock(&pr->pr_mtx); 2574 TASK_INIT(&pr->pr_task, 0, prison_complete, pr); 2575 taskqueue_enqueue(taskqueue_thread, &pr->pr_task); 2576 return; 2577 } 2578 mtx_unlock(&pr->pr_mtx); 2579} 2580 2581void 2582prison_free(struct prison *pr) 2583{ 2584 2585 mtx_lock(&pr->pr_mtx); 2586 prison_free_locked(pr); 2587} 2588 2589static void 2590prison_complete(void *context, int pending) 2591{ 2592 2593 prison_deref((struct prison *)context, 0); 2594} 2595 2596/* 2597 * Remove a prison reference (usually). This internal version assumes no 2598 * mutexes are held, except perhaps the prison itself. If there are no more 2599 * references, release and delist the prison. On completion, the prison lock 2600 * and the allprison lock are both unlocked. 2601 */ 2602static void 2603prison_deref(struct prison *pr, int flags) 2604{ 2605 struct prison *ppr, *tpr; 2606 int ref, lasturef; 2607 2608 if (!(flags & PD_LOCKED)) 2609 mtx_lock(&pr->pr_mtx); 2610 for (;;) { 2611 if (flags & PD_DEUREF) { 2612 pr->pr_uref--; 2613 lasturef = pr->pr_uref == 0; 2614 if (lasturef) 2615 pr->pr_ref++; 2616 KASSERT(prison0.pr_uref != 0, ("prison0 pr_uref=0")); 2617 } else 2618 lasturef = 0; 2619 if (flags & PD_DEREF) 2620 pr->pr_ref--; 2621 ref = pr->pr_ref; 2622 mtx_unlock(&pr->pr_mtx); 2623 2624 /* 2625 * Tell the modules if the last user reference was removed 2626 * (even it sticks around in dying state). 2627 */ 2628 if (lasturef) { 2629 if (!(flags & (PD_LIST_SLOCKED | PD_LIST_XLOCKED))) { 2630 if (ref > 1) { 2631 sx_slock(&allprison_lock); 2632 flags |= PD_LIST_SLOCKED; 2633 } else { 2634 sx_xlock(&allprison_lock); 2635 flags |= PD_LIST_XLOCKED; 2636 } 2637 } 2638 (void)osd_jail_call(pr, PR_METHOD_REMOVE, NULL); 2639 mtx_lock(&pr->pr_mtx); 2640 ref = --pr->pr_ref; 2641 mtx_unlock(&pr->pr_mtx); 2642 } 2643 2644 /* If the prison still has references, nothing else to do. */ 2645 if (ref > 0) { 2646 if (flags & PD_LIST_SLOCKED) 2647 sx_sunlock(&allprison_lock); 2648 else if (flags & PD_LIST_XLOCKED) 2649 sx_xunlock(&allprison_lock); 2650 return; 2651 } 2652 2653 if (flags & PD_LIST_SLOCKED) { 2654 if (!sx_try_upgrade(&allprison_lock)) { 2655 sx_sunlock(&allprison_lock); 2656 sx_xlock(&allprison_lock); 2657 } 2658 } else if (!(flags & PD_LIST_XLOCKED)) 2659 sx_xlock(&allprison_lock); 2660 2661 TAILQ_REMOVE(&allprison, pr, pr_list); 2662 LIST_REMOVE(pr, pr_sibling); 2663 ppr = pr->pr_parent; 2664 for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent) 2665 tpr->pr_childcount--; 2666 sx_xunlock(&allprison_lock); 2667 2668#ifdef VIMAGE 2669 if (pr->pr_vnet != ppr->pr_vnet) 2670 vnet_destroy(pr->pr_vnet); 2671#endif 2672 if (pr->pr_root != NULL) 2673 vrele(pr->pr_root); 2674 mtx_destroy(&pr->pr_mtx); 2675#ifdef INET 2676 free(pr->pr_ip4, M_PRISON); 2677#endif 2678#ifdef INET6 2679 free(pr->pr_ip6, M_PRISON); 2680#endif 2681 if (pr->pr_cpuset != NULL) 2682 cpuset_rel(pr->pr_cpuset); 2683 osd_jail_exit(pr); 2684#ifdef RACCT 2685 if (racct_enable) 2686 prison_racct_detach(pr); 2687#endif 2688 free(pr, M_PRISON); 2689 2690 /* Removing a prison frees a reference on its parent. */ 2691 pr = ppr; 2692 mtx_lock(&pr->pr_mtx); 2693 flags = PD_DEREF | PD_DEUREF; 2694 } 2695} 2696 2697void 2698prison_hold_locked(struct prison *pr) 2699{ 2700 2701 mtx_assert(&pr->pr_mtx, MA_OWNED); 2702 KASSERT(pr->pr_ref > 0, 2703 ("Trying to hold dead prison (jid=%d).", pr->pr_id)); 2704 pr->pr_ref++; 2705} 2706 2707void 2708prison_hold(struct prison *pr) 2709{ 2710 2711 mtx_lock(&pr->pr_mtx); 2712 prison_hold_locked(pr); 2713 mtx_unlock(&pr->pr_mtx); 2714} 2715 2716void 2717prison_proc_hold(struct prison *pr) 2718{ 2719 2720 mtx_lock(&pr->pr_mtx); 2721 KASSERT(pr->pr_uref > 0, 2722 ("Cannot add a process to a non-alive prison (jid=%d)", pr->pr_id)); 2723 pr->pr_uref++; 2724 mtx_unlock(&pr->pr_mtx); 2725} 2726 2727void 2728prison_proc_free(struct prison *pr) 2729{ 2730 2731 mtx_lock(&pr->pr_mtx); 2732 KASSERT(pr->pr_uref > 0, 2733 ("Trying to kill a process in a dead prison (jid=%d)", pr->pr_id)); 2734 prison_deref(pr, PD_DEUREF | PD_LOCKED); 2735} 2736 2737 2738#ifdef INET 2739/* 2740 * Restrict a prison's IP address list with its parent's, possibly replacing 2741 * it. Return true if the replacement buffer was used (or would have been). 2742 */ 2743static int 2744prison_restrict_ip4(struct prison *pr, struct in_addr *newip4) 2745{ 2746 int ii, ij, used; 2747 struct prison *ppr; 2748 2749 ppr = pr->pr_parent; 2750 if (!(pr->pr_flags & PR_IP4_USER)) { 2751 /* This has no user settings, so just copy the parent's list. */ 2752 if (pr->pr_ip4s < ppr->pr_ip4s) { 2753 /* 2754 * There's no room for the parent's list. Use the 2755 * new list buffer, which is assumed to be big enough 2756 * (if it was passed). If there's no buffer, try to 2757 * allocate one. 2758 */ 2759 used = 1; 2760 if (newip4 == NULL) { 2761 newip4 = malloc(ppr->pr_ip4s * sizeof(*newip4), 2762 M_PRISON, M_NOWAIT); 2763 if (newip4 != NULL) 2764 used = 0; 2765 } 2766 if (newip4 != NULL) { 2767 bcopy(ppr->pr_ip4, newip4, 2768 ppr->pr_ip4s * sizeof(*newip4)); 2769 free(pr->pr_ip4, M_PRISON); 2770 pr->pr_ip4 = newip4; 2771 pr->pr_ip4s = ppr->pr_ip4s; 2772 } 2773 return (used); 2774 } 2775 pr->pr_ip4s = ppr->pr_ip4s; 2776 if (pr->pr_ip4s > 0) 2777 bcopy(ppr->pr_ip4, pr->pr_ip4, 2778 pr->pr_ip4s * sizeof(*newip4)); 2779 else if (pr->pr_ip4 != NULL) { 2780 free(pr->pr_ip4, M_PRISON); 2781 pr->pr_ip4 = NULL; 2782 } 2783 } else if (pr->pr_ip4s > 0) { 2784 /* Remove addresses that aren't in the parent. */ 2785 for (ij = 0; ij < ppr->pr_ip4s; ij++) 2786 if (pr->pr_ip4[0].s_addr == ppr->pr_ip4[ij].s_addr) 2787 break; 2788 if (ij < ppr->pr_ip4s) 2789 ii = 1; 2790 else { 2791 bcopy(pr->pr_ip4 + 1, pr->pr_ip4, 2792 --pr->pr_ip4s * sizeof(*pr->pr_ip4)); 2793 ii = 0; 2794 } 2795 for (ij = 1; ii < pr->pr_ip4s; ) { 2796 if (pr->pr_ip4[ii].s_addr == ppr->pr_ip4[0].s_addr) { 2797 ii++; 2798 continue; 2799 } 2800 switch (ij >= ppr->pr_ip4s ? -1 : 2801 qcmp_v4(&pr->pr_ip4[ii], &ppr->pr_ip4[ij])) { 2802 case -1: 2803 bcopy(pr->pr_ip4 + ii + 1, pr->pr_ip4 + ii, 2804 (--pr->pr_ip4s - ii) * sizeof(*pr->pr_ip4)); 2805 break; 2806 case 0: 2807 ii++; 2808 ij++; 2809 break; 2810 case 1: 2811 ij++; 2812 break; 2813 } 2814 } 2815 if (pr->pr_ip4s == 0) { 2816 free(pr->pr_ip4, M_PRISON); 2817 pr->pr_ip4 = NULL; 2818 } 2819 } 2820 return (0); 2821} 2822 2823/* 2824 * Pass back primary IPv4 address of this jail. 2825 * 2826 * If not restricted return success but do not alter the address. Caller has 2827 * to make sure to initialize it correctly (e.g. INADDR_ANY). 2828 * 2829 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv4. 2830 * Address returned in NBO. 2831 */ 2832int 2833prison_get_ip4(struct ucred *cred, struct in_addr *ia) 2834{ 2835 struct prison *pr; 2836 2837 KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); 2838 KASSERT(ia != NULL, ("%s: ia is NULL", __func__)); 2839 2840 pr = cred->cr_prison; 2841 if (!(pr->pr_flags & PR_IP4)) 2842 return (0); 2843 mtx_lock(&pr->pr_mtx); 2844 if (!(pr->pr_flags & PR_IP4)) { 2845 mtx_unlock(&pr->pr_mtx); 2846 return (0); 2847 } 2848 if (pr->pr_ip4 == NULL) { 2849 mtx_unlock(&pr->pr_mtx); 2850 return (EAFNOSUPPORT); 2851 } 2852 2853 ia->s_addr = pr->pr_ip4[0].s_addr; 2854 mtx_unlock(&pr->pr_mtx); 2855 return (0); 2856} 2857 2858/* 2859 * Return 1 if we should do proper source address selection or are not jailed. 2860 * We will return 0 if we should bypass source address selection in favour 2861 * of the primary jail IPv4 address. Only in this case *ia will be updated and 2862 * returned in NBO. 2863 * Return EAFNOSUPPORT, in case this jail does not allow IPv4. 2864 */ 2865int 2866prison_saddrsel_ip4(struct ucred *cred, struct in_addr *ia) 2867{ 2868 struct prison *pr; 2869 struct in_addr lia; 2870 int error; 2871 2872 KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); 2873 KASSERT(ia != NULL, ("%s: ia is NULL", __func__)); 2874 2875 if (!jailed(cred)) 2876 return (1); 2877 2878 pr = cred->cr_prison; 2879 if (pr->pr_flags & PR_IP4_SADDRSEL) 2880 return (1); 2881 2882 lia.s_addr = INADDR_ANY; 2883 error = prison_get_ip4(cred, &lia); 2884 if (error) 2885 return (error); 2886 if (lia.s_addr == INADDR_ANY) 2887 return (1); 2888 2889 ia->s_addr = lia.s_addr; 2890 return (0); 2891} 2892 2893/* 2894 * Return true if pr1 and pr2 have the same IPv4 address restrictions. 2895 */ 2896int 2897prison_equal_ip4(struct prison *pr1, struct prison *pr2) 2898{ 2899 2900 if (pr1 == pr2) 2901 return (1); 2902 2903 /* 2904 * No need to lock since the PR_IP4_USER flag can't be altered for 2905 * existing prisons. 2906 */ 2907 while (pr1 != &prison0 && 2908#ifdef VIMAGE 2909 !(pr1->pr_flags & PR_VNET) && 2910#endif 2911 !(pr1->pr_flags & PR_IP4_USER)) 2912 pr1 = pr1->pr_parent; 2913 while (pr2 != &prison0 && 2914#ifdef VIMAGE 2915 !(pr2->pr_flags & PR_VNET) && 2916#endif 2917 !(pr2->pr_flags & PR_IP4_USER)) 2918 pr2 = pr2->pr_parent; 2919 return (pr1 == pr2); 2920} 2921 2922/* 2923 * Make sure our (source) address is set to something meaningful to this 2924 * jail. 2925 * 2926 * Returns 0 if jail doesn't restrict IPv4 or if address belongs to jail, 2927 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail 2928 * doesn't allow IPv4. Address passed in in NBO and returned in NBO. 2929 */ 2930int 2931prison_local_ip4(struct ucred *cred, struct in_addr *ia) 2932{ 2933 struct prison *pr; 2934 struct in_addr ia0; 2935 int error; 2936 2937 KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); 2938 KASSERT(ia != NULL, ("%s: ia is NULL", __func__)); 2939 2940 pr = cred->cr_prison; 2941 if (!(pr->pr_flags & PR_IP4)) 2942 return (0); 2943 mtx_lock(&pr->pr_mtx); 2944 if (!(pr->pr_flags & PR_IP4)) { 2945 mtx_unlock(&pr->pr_mtx); 2946 return (0); 2947 } 2948 if (pr->pr_ip4 == NULL) { 2949 mtx_unlock(&pr->pr_mtx); 2950 return (EAFNOSUPPORT); 2951 } 2952 2953 ia0.s_addr = ntohl(ia->s_addr); 2954 if (ia0.s_addr == INADDR_LOOPBACK) { 2955 ia->s_addr = pr->pr_ip4[0].s_addr; 2956 mtx_unlock(&pr->pr_mtx); 2957 return (0); 2958 } 2959 2960 if (ia0.s_addr == INADDR_ANY) { 2961 /* 2962 * In case there is only 1 IPv4 address, bind directly. 2963 */ 2964 if (pr->pr_ip4s == 1) 2965 ia->s_addr = pr->pr_ip4[0].s_addr; 2966 mtx_unlock(&pr->pr_mtx); 2967 return (0); 2968 } 2969 2970 error = _prison_check_ip4(pr, ia); 2971 mtx_unlock(&pr->pr_mtx); 2972 return (error); 2973} 2974 2975/* 2976 * Rewrite destination address in case we will connect to loopback address. 2977 * 2978 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv4. 2979 * Address passed in in NBO and returned in NBO. 2980 */ 2981int 2982prison_remote_ip4(struct ucred *cred, struct in_addr *ia) 2983{ 2984 struct prison *pr; 2985 2986 KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); 2987 KASSERT(ia != NULL, ("%s: ia is NULL", __func__)); 2988 2989 pr = cred->cr_prison; 2990 if (!(pr->pr_flags & PR_IP4)) 2991 return (0); 2992 mtx_lock(&pr->pr_mtx); 2993 if (!(pr->pr_flags & PR_IP4)) { 2994 mtx_unlock(&pr->pr_mtx); 2995 return (0); 2996 } 2997 if (pr->pr_ip4 == NULL) { 2998 mtx_unlock(&pr->pr_mtx); 2999 return (EAFNOSUPPORT); 3000 } 3001 3002 if (ntohl(ia->s_addr) == INADDR_LOOPBACK) { 3003 ia->s_addr = pr->pr_ip4[0].s_addr; 3004 mtx_unlock(&pr->pr_mtx); 3005 return (0); 3006 } 3007 3008 /* 3009 * Return success because nothing had to be changed. 3010 */ 3011 mtx_unlock(&pr->pr_mtx); 3012 return (0); 3013} 3014 3015/* 3016 * Check if given address belongs to the jail referenced by cred/prison. 3017 * 3018 * Returns 0 if jail doesn't restrict IPv4 or if address belongs to jail, 3019 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail 3020 * doesn't allow IPv4. Address passed in in NBO. 3021 */ 3022static int 3023_prison_check_ip4(const struct prison *pr, const struct in_addr *ia) 3024{ 3025 int i, a, z, d; 3026 3027 /* 3028 * Check the primary IP. 3029 */ 3030 if (pr->pr_ip4[0].s_addr == ia->s_addr) 3031 return (0); 3032 3033 /* 3034 * All the other IPs are sorted so we can do a binary search. 3035 */ 3036 a = 0; 3037 z = pr->pr_ip4s - 2; 3038 while (a <= z) { 3039 i = (a + z) / 2; 3040 d = qcmp_v4(&pr->pr_ip4[i+1], ia); 3041 if (d > 0) 3042 z = i - 1; 3043 else if (d < 0) 3044 a = i + 1; 3045 else 3046 return (0); 3047 } 3048 3049 return (EADDRNOTAVAIL); 3050} 3051 3052int 3053prison_check_ip4(const struct ucred *cred, const struct in_addr *ia) 3054{ 3055 struct prison *pr; 3056 int error; 3057 3058 KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); 3059 KASSERT(ia != NULL, ("%s: ia is NULL", __func__)); 3060 3061 pr = cred->cr_prison; 3062 if (!(pr->pr_flags & PR_IP4)) 3063 return (0); 3064 mtx_lock(&pr->pr_mtx); 3065 if (!(pr->pr_flags & PR_IP4)) { 3066 mtx_unlock(&pr->pr_mtx); 3067 return (0); 3068 } 3069 if (pr->pr_ip4 == NULL) { 3070 mtx_unlock(&pr->pr_mtx); 3071 return (EAFNOSUPPORT); 3072 } 3073 3074 error = _prison_check_ip4(pr, ia); 3075 mtx_unlock(&pr->pr_mtx); 3076 return (error); 3077} 3078#endif 3079 3080#ifdef INET6 3081static int 3082prison_restrict_ip6(struct prison *pr, struct in6_addr *newip6) 3083{ 3084 int ii, ij, used; 3085 struct prison *ppr; 3086 3087 ppr = pr->pr_parent; 3088 if (!(pr->pr_flags & PR_IP6_USER)) { 3089 /* This has no user settings, so just copy the parent's list. */ 3090 if (pr->pr_ip6s < ppr->pr_ip6s) { 3091 /* 3092 * There's no room for the parent's list. Use the 3093 * new list buffer, which is assumed to be big enough 3094 * (if it was passed). If there's no buffer, try to 3095 * allocate one. 3096 */ 3097 used = 1; 3098 if (newip6 == NULL) { 3099 newip6 = malloc(ppr->pr_ip6s * sizeof(*newip6), 3100 M_PRISON, M_NOWAIT); 3101 if (newip6 != NULL) 3102 used = 0; 3103 } 3104 if (newip6 != NULL) { 3105 bcopy(ppr->pr_ip6, newip6, 3106 ppr->pr_ip6s * sizeof(*newip6)); 3107 free(pr->pr_ip6, M_PRISON); 3108 pr->pr_ip6 = newip6; 3109 pr->pr_ip6s = ppr->pr_ip6s; 3110 } 3111 return (used); 3112 } 3113 pr->pr_ip6s = ppr->pr_ip6s; 3114 if (pr->pr_ip6s > 0) 3115 bcopy(ppr->pr_ip6, pr->pr_ip6, 3116 pr->pr_ip6s * sizeof(*newip6)); 3117 else if (pr->pr_ip6 != NULL) { 3118 free(pr->pr_ip6, M_PRISON); 3119 pr->pr_ip6 = NULL; 3120 } 3121 } else if (pr->pr_ip6s > 0) { 3122 /* Remove addresses that aren't in the parent. */ 3123 for (ij = 0; ij < ppr->pr_ip6s; ij++) 3124 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[0], 3125 &ppr->pr_ip6[ij])) 3126 break; 3127 if (ij < ppr->pr_ip6s) 3128 ii = 1; 3129 else { 3130 bcopy(pr->pr_ip6 + 1, pr->pr_ip6, 3131 --pr->pr_ip6s * sizeof(*pr->pr_ip6)); 3132 ii = 0; 3133 } 3134 for (ij = 1; ii < pr->pr_ip6s; ) { 3135 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[ii], 3136 &ppr->pr_ip6[0])) { 3137 ii++; 3138 continue; 3139 } 3140 switch (ij >= ppr->pr_ip6s ? -1 : 3141 qcmp_v6(&pr->pr_ip6[ii], &ppr->pr_ip6[ij])) { 3142 case -1: 3143 bcopy(pr->pr_ip6 + ii + 1, pr->pr_ip6 + ii, 3144 (--pr->pr_ip6s - ii) * sizeof(*pr->pr_ip6)); 3145 break; 3146 case 0: 3147 ii++; 3148 ij++; 3149 break; 3150 case 1: 3151 ij++; 3152 break; 3153 } 3154 } 3155 if (pr->pr_ip6s == 0) { 3156 free(pr->pr_ip6, M_PRISON); 3157 pr->pr_ip6 = NULL; 3158 } 3159 } 3160 return 0; 3161} 3162 3163/* 3164 * Pass back primary IPv6 address for this jail. 3165 * 3166 * If not restricted return success but do not alter the address. Caller has 3167 * to make sure to initialize it correctly (e.g. IN6ADDR_ANY_INIT). 3168 * 3169 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv6. 3170 */ 3171int 3172prison_get_ip6(struct ucred *cred, struct in6_addr *ia6) 3173{ 3174 struct prison *pr; 3175 3176 KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); 3177 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__)); 3178 3179 pr = cred->cr_prison; 3180 if (!(pr->pr_flags & PR_IP6)) 3181 return (0); 3182 mtx_lock(&pr->pr_mtx); 3183 if (!(pr->pr_flags & PR_IP6)) { 3184 mtx_unlock(&pr->pr_mtx); 3185 return (0); 3186 } 3187 if (pr->pr_ip6 == NULL) { 3188 mtx_unlock(&pr->pr_mtx); 3189 return (EAFNOSUPPORT); 3190 } 3191 3192 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr)); 3193 mtx_unlock(&pr->pr_mtx); 3194 return (0); 3195} 3196 3197/* 3198 * Return 1 if we should do proper source address selection or are not jailed. 3199 * We will return 0 if we should bypass source address selection in favour 3200 * of the primary jail IPv6 address. Only in this case *ia will be updated and 3201 * returned in NBO. 3202 * Return EAFNOSUPPORT, in case this jail does not allow IPv6. 3203 */ 3204int 3205prison_saddrsel_ip6(struct ucred *cred, struct in6_addr *ia6) 3206{ 3207 struct prison *pr; 3208 struct in6_addr lia6; 3209 int error; 3210 3211 KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); 3212 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__)); 3213 3214 if (!jailed(cred)) 3215 return (1); 3216 3217 pr = cred->cr_prison; 3218 if (pr->pr_flags & PR_IP6_SADDRSEL) 3219 return (1); 3220 3221 lia6 = in6addr_any; 3222 error = prison_get_ip6(cred, &lia6); 3223 if (error) 3224 return (error); 3225 if (IN6_IS_ADDR_UNSPECIFIED(&lia6)) 3226 return (1); 3227 3228 bcopy(&lia6, ia6, sizeof(struct in6_addr)); 3229 return (0); 3230} 3231 3232/* 3233 * Return true if pr1 and pr2 have the same IPv6 address restrictions. 3234 */ 3235int 3236prison_equal_ip6(struct prison *pr1, struct prison *pr2) 3237{ 3238 3239 if (pr1 == pr2) 3240 return (1); 3241 3242 while (pr1 != &prison0 && 3243#ifdef VIMAGE 3244 !(pr1->pr_flags & PR_VNET) && 3245#endif 3246 !(pr1->pr_flags & PR_IP6_USER)) 3247 pr1 = pr1->pr_parent; 3248 while (pr2 != &prison0 && 3249#ifdef VIMAGE 3250 !(pr2->pr_flags & PR_VNET) && 3251#endif 3252 !(pr2->pr_flags & PR_IP6_USER)) 3253 pr2 = pr2->pr_parent; 3254 return (pr1 == pr2); 3255} 3256 3257/* 3258 * Make sure our (source) address is set to something meaningful to this jail. 3259 * 3260 * v6only should be set based on (inp->inp_flags & IN6P_IPV6_V6ONLY != 0) 3261 * when needed while binding. 3262 * 3263 * Returns 0 if jail doesn't restrict IPv6 or if address belongs to jail, 3264 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail 3265 * doesn't allow IPv6. 3266 */ 3267int 3268prison_local_ip6(struct ucred *cred, struct in6_addr *ia6, int v6only) 3269{ 3270 struct prison *pr; 3271 int error; 3272 3273 KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); 3274 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__)); 3275 3276 pr = cred->cr_prison; 3277 if (!(pr->pr_flags & PR_IP6)) 3278 return (0); 3279 mtx_lock(&pr->pr_mtx); 3280 if (!(pr->pr_flags & PR_IP6)) { 3281 mtx_unlock(&pr->pr_mtx); 3282 return (0); 3283 } 3284 if (pr->pr_ip6 == NULL) { 3285 mtx_unlock(&pr->pr_mtx); 3286 return (EAFNOSUPPORT); 3287 } 3288 3289 if (IN6_IS_ADDR_LOOPBACK(ia6)) { 3290 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr)); 3291 mtx_unlock(&pr->pr_mtx); 3292 return (0); 3293 } 3294 3295 if (IN6_IS_ADDR_UNSPECIFIED(ia6)) { 3296 /* 3297 * In case there is only 1 IPv6 address, and v6only is true, 3298 * then bind directly. 3299 */ 3300 if (v6only != 0 && pr->pr_ip6s == 1) 3301 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr)); 3302 mtx_unlock(&pr->pr_mtx); 3303 return (0); 3304 } 3305 3306 error = _prison_check_ip6(pr, ia6); 3307 mtx_unlock(&pr->pr_mtx); 3308 return (error); 3309} 3310 3311/* 3312 * Rewrite destination address in case we will connect to loopback address. 3313 * 3314 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv6. 3315 */ 3316int 3317prison_remote_ip6(struct ucred *cred, struct in6_addr *ia6) 3318{ 3319 struct prison *pr; 3320 3321 KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); 3322 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__)); 3323 3324 pr = cred->cr_prison; 3325 if (!(pr->pr_flags & PR_IP6)) 3326 return (0); 3327 mtx_lock(&pr->pr_mtx); 3328 if (!(pr->pr_flags & PR_IP6)) { 3329 mtx_unlock(&pr->pr_mtx); 3330 return (0); 3331 } 3332 if (pr->pr_ip6 == NULL) { 3333 mtx_unlock(&pr->pr_mtx); 3334 return (EAFNOSUPPORT); 3335 } 3336 3337 if (IN6_IS_ADDR_LOOPBACK(ia6)) { 3338 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr)); 3339 mtx_unlock(&pr->pr_mtx); 3340 return (0); 3341 } 3342 3343 /* 3344 * Return success because nothing had to be changed. 3345 */ 3346 mtx_unlock(&pr->pr_mtx); 3347 return (0); 3348} 3349 3350/* 3351 * Check if given address belongs to the jail referenced by cred/prison. 3352 * 3353 * Returns 0 if jail doesn't restrict IPv6 or if address belongs to jail, 3354 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail 3355 * doesn't allow IPv6. 3356 */ 3357static int 3358_prison_check_ip6(struct prison *pr, struct in6_addr *ia6) 3359{ 3360 int i, a, z, d; 3361 3362 /* 3363 * Check the primary IP. 3364 */ 3365 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[0], ia6)) 3366 return (0); 3367 3368 /* 3369 * All the other IPs are sorted so we can do a binary search. 3370 */ 3371 a = 0; 3372 z = pr->pr_ip6s - 2; 3373 while (a <= z) { 3374 i = (a + z) / 2; 3375 d = qcmp_v6(&pr->pr_ip6[i+1], ia6); 3376 if (d > 0) 3377 z = i - 1; 3378 else if (d < 0) 3379 a = i + 1; 3380 else 3381 return (0); 3382 } 3383 3384 return (EADDRNOTAVAIL); 3385} 3386 3387int 3388prison_check_ip6(struct ucred *cred, struct in6_addr *ia6) 3389{ 3390 struct prison *pr; 3391 int error; 3392 3393 KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); 3394 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__)); 3395 3396 pr = cred->cr_prison; 3397 if (!(pr->pr_flags & PR_IP6)) 3398 return (0); 3399 mtx_lock(&pr->pr_mtx); 3400 if (!(pr->pr_flags & PR_IP6)) { 3401 mtx_unlock(&pr->pr_mtx); 3402 return (0); 3403 } 3404 if (pr->pr_ip6 == NULL) { 3405 mtx_unlock(&pr->pr_mtx); 3406 return (EAFNOSUPPORT); 3407 } 3408 3409 error = _prison_check_ip6(pr, ia6); 3410 mtx_unlock(&pr->pr_mtx); 3411 return (error); 3412} 3413#endif 3414 3415/* 3416 * Check if a jail supports the given address family. 3417 * 3418 * Returns 0 if not jailed or the address family is supported, EAFNOSUPPORT 3419 * if not. 3420 */ 3421int 3422prison_check_af(struct ucred *cred, int af) 3423{ 3424 struct prison *pr; 3425 int error; 3426 3427 KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); 3428 3429 pr = cred->cr_prison; 3430#ifdef VIMAGE 3431 /* Prisons with their own network stack are not limited. */ 3432 if (prison_owns_vnet(cred)) 3433 return (0); 3434#endif 3435 3436 error = 0; 3437 switch (af) 3438 { 3439#ifdef INET 3440 case AF_INET: 3441 if (pr->pr_flags & PR_IP4) 3442 { 3443 mtx_lock(&pr->pr_mtx); 3444 if ((pr->pr_flags & PR_IP4) && pr->pr_ip4 == NULL) 3445 error = EAFNOSUPPORT; 3446 mtx_unlock(&pr->pr_mtx); 3447 } 3448 break; 3449#endif 3450#ifdef INET6 3451 case AF_INET6: 3452 if (pr->pr_flags & PR_IP6) 3453 { 3454 mtx_lock(&pr->pr_mtx); 3455 if ((pr->pr_flags & PR_IP6) && pr->pr_ip6 == NULL) 3456 error = EAFNOSUPPORT; 3457 mtx_unlock(&pr->pr_mtx); 3458 } 3459 break; 3460#endif 3461 case AF_LOCAL: 3462 case AF_ROUTE: 3463 break; 3464 default: 3465 if (!(pr->pr_allow & PR_ALLOW_SOCKET_AF)) 3466 error = EAFNOSUPPORT; 3467 } 3468 return (error); 3469} 3470 3471/* 3472 * Check if given address belongs to the jail referenced by cred (wrapper to 3473 * prison_check_ip[46]). 3474 * 3475 * Returns 0 if jail doesn't restrict the address family or if address belongs 3476 * to jail, EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if 3477 * the jail doesn't allow the address family. IPv4 Address passed in in NBO. 3478 */ 3479int 3480prison_if(struct ucred *cred, struct sockaddr *sa) 3481{ 3482#ifdef INET 3483 struct sockaddr_in *sai; 3484#endif 3485#ifdef INET6 3486 struct sockaddr_in6 *sai6; 3487#endif 3488 int error; 3489 3490 KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); 3491 KASSERT(sa != NULL, ("%s: sa is NULL", __func__)); 3492 3493#ifdef VIMAGE 3494 if (prison_owns_vnet(cred)) 3495 return (0); 3496#endif 3497 3498 error = 0; 3499 switch (sa->sa_family) 3500 { 3501#ifdef INET 3502 case AF_INET: 3503 sai = (struct sockaddr_in *)sa; 3504 error = prison_check_ip4(cred, &sai->sin_addr); 3505 break; 3506#endif 3507#ifdef INET6 3508 case AF_INET6: 3509 sai6 = (struct sockaddr_in6 *)sa; 3510 error = prison_check_ip6(cred, &sai6->sin6_addr); 3511 break; 3512#endif 3513 default: 3514 if (!(cred->cr_prison->pr_allow & PR_ALLOW_SOCKET_AF)) 3515 error = EAFNOSUPPORT; 3516 } 3517 return (error); 3518} 3519 3520/* 3521 * Return 0 if jails permit p1 to frob p2, otherwise ESRCH. 3522 */ 3523int 3524prison_check(struct ucred *cred1, struct ucred *cred2) 3525{ 3526 3527 return ((cred1->cr_prison == cred2->cr_prison || 3528 prison_ischild(cred1->cr_prison, cred2->cr_prison)) ? 0 : ESRCH); 3529} 3530 3531/* 3532 * Return 1 if p2 is a child of p1, otherwise 0. 3533 */ 3534int 3535prison_ischild(struct prison *pr1, struct prison *pr2) 3536{ 3537 3538 for (pr2 = pr2->pr_parent; pr2 != NULL; pr2 = pr2->pr_parent) 3539 if (pr1 == pr2) 3540 return (1); 3541 return (0); 3542} 3543 3544/* 3545 * Return 1 if the passed credential is in a jail, otherwise 0. 3546 */ 3547int 3548jailed(struct ucred *cred) 3549{ 3550 3551 return (cred->cr_prison != &prison0); 3552} 3553 3554/* 3555 * Return 1 if the passed credential is in a jail and that jail does not 3556 * have its own virtual network stack, otherwise 0. 3557 */ 3558int 3559jailed_without_vnet(struct ucred *cred) 3560{ 3561 3562 if (!jailed(cred)) 3563 return (0); 3564#ifdef VIMAGE 3565 if (prison_owns_vnet(cred)) 3566 return (0); 3567#endif 3568 3569 return (1); 3570} 3571 3572/* 3573 * Return the correct hostname (domainname, et al) for the passed credential. 3574 */ 3575void 3576getcredhostname(struct ucred *cred, char *buf, size_t size) 3577{ 3578 struct prison *pr; 3579 3580 /* 3581 * A NULL credential can be used to shortcut to the physical 3582 * system's hostname. 3583 */ 3584 pr = (cred != NULL) ? cred->cr_prison : &prison0; 3585 mtx_lock(&pr->pr_mtx); 3586 strlcpy(buf, pr->pr_hostname, size); 3587 mtx_unlock(&pr->pr_mtx); 3588} 3589 3590void 3591getcreddomainname(struct ucred *cred, char *buf, size_t size) 3592{ 3593 3594 mtx_lock(&cred->cr_prison->pr_mtx); 3595 strlcpy(buf, cred->cr_prison->pr_domainname, size); 3596 mtx_unlock(&cred->cr_prison->pr_mtx); 3597} 3598 3599void 3600getcredhostuuid(struct ucred *cred, char *buf, size_t size) 3601{ 3602 3603 mtx_lock(&cred->cr_prison->pr_mtx); 3604 strlcpy(buf, cred->cr_prison->pr_hostuuid, size); 3605 mtx_unlock(&cred->cr_prison->pr_mtx); 3606} 3607 3608void 3609getcredhostid(struct ucred *cred, unsigned long *hostid) 3610{ 3611 3612 mtx_lock(&cred->cr_prison->pr_mtx); 3613 *hostid = cred->cr_prison->pr_hostid; 3614 mtx_unlock(&cred->cr_prison->pr_mtx); 3615} 3616 3617#ifdef VIMAGE 3618/* 3619 * Determine whether the prison represented by cred owns 3620 * its vnet rather than having it inherited. 3621 * 3622 * Returns 1 in case the prison owns the vnet, 0 otherwise. 3623 */ 3624int 3625prison_owns_vnet(struct ucred *cred) 3626{ 3627 3628 /* 3629 * vnets cannot be added/removed after jail creation, 3630 * so no need to lock here. 3631 */ 3632 return (cred->cr_prison->pr_flags & PR_VNET ? 1 : 0); 3633} 3634#endif 3635 3636/* 3637 * Determine whether the subject represented by cred can "see" 3638 * status of a mount point. 3639 * Returns: 0 for permitted, ENOENT otherwise. 3640 * XXX: This function should be called cr_canseemount() and should be 3641 * placed in kern_prot.c. 3642 */ 3643int 3644prison_canseemount(struct ucred *cred, struct mount *mp) 3645{ 3646 struct prison *pr; 3647 struct statfs *sp; 3648 size_t len; 3649 3650 pr = cred->cr_prison; 3651 if (pr->pr_enforce_statfs == 0) 3652 return (0); 3653 if (pr->pr_root->v_mount == mp) 3654 return (0); 3655 if (pr->pr_enforce_statfs == 2) 3656 return (ENOENT); 3657 /* 3658 * If jail's chroot directory is set to "/" we should be able to see 3659 * all mount-points from inside a jail. 3660 * This is ugly check, but this is the only situation when jail's 3661 * directory ends with '/'. 3662 */ 3663 if (strcmp(pr->pr_path, "/") == 0) 3664 return (0); 3665 len = strlen(pr->pr_path); 3666 sp = &mp->mnt_stat; 3667 if (strncmp(pr->pr_path, sp->f_mntonname, len) != 0) 3668 return (ENOENT); 3669 /* 3670 * Be sure that we don't have situation where jail's root directory 3671 * is "/some/path" and mount point is "/some/pathpath". 3672 */ 3673 if (sp->f_mntonname[len] != '\0' && sp->f_mntonname[len] != '/') 3674 return (ENOENT); 3675 return (0); 3676} 3677 3678void 3679prison_enforce_statfs(struct ucred *cred, struct mount *mp, struct statfs *sp) 3680{ 3681 char jpath[MAXPATHLEN]; 3682 struct prison *pr; 3683 size_t len; 3684 3685 pr = cred->cr_prison; 3686 if (pr->pr_enforce_statfs == 0) 3687 return; 3688 if (prison_canseemount(cred, mp) != 0) { 3689 bzero(sp->f_mntonname, sizeof(sp->f_mntonname)); 3690 strlcpy(sp->f_mntonname, "[restricted]", 3691 sizeof(sp->f_mntonname)); 3692 return; 3693 } 3694 if (pr->pr_root->v_mount == mp) { 3695 /* 3696 * Clear current buffer data, so we are sure nothing from 3697 * the valid path left there. 3698 */ 3699 bzero(sp->f_mntonname, sizeof(sp->f_mntonname)); 3700 *sp->f_mntonname = '/'; 3701 return; 3702 } 3703 /* 3704 * If jail's chroot directory is set to "/" we should be able to see 3705 * all mount-points from inside a jail. 3706 */ 3707 if (strcmp(pr->pr_path, "/") == 0) 3708 return; 3709 len = strlen(pr->pr_path); 3710 strlcpy(jpath, sp->f_mntonname + len, sizeof(jpath)); 3711 /* 3712 * Clear current buffer data, so we are sure nothing from 3713 * the valid path left there. 3714 */ 3715 bzero(sp->f_mntonname, sizeof(sp->f_mntonname)); 3716 if (*jpath == '\0') { 3717 /* Should never happen. */ 3718 *sp->f_mntonname = '/'; 3719 } else { 3720 strlcpy(sp->f_mntonname, jpath, sizeof(sp->f_mntonname)); 3721 } 3722} 3723 3724/* 3725 * Check with permission for a specific privilege is granted within jail. We 3726 * have a specific list of accepted privileges; the rest are denied. 3727 */ 3728int 3729prison_priv_check(struct ucred *cred, int priv) 3730{ 3731 3732 if (!jailed(cred)) 3733 return (0); 3734 3735#ifdef VIMAGE 3736 /* 3737 * Privileges specific to prisons with a virtual network stack. 3738 * There might be a duplicate entry here in case the privilege 3739 * is only granted conditionally in the legacy jail case. 3740 */ 3741 switch (priv) { 3742#ifdef notyet 3743 /* 3744 * NFS-specific privileges. 3745 */ 3746 case PRIV_NFS_DAEMON: 3747 case PRIV_NFS_LOCKD: 3748#endif 3749 /* 3750 * Network stack privileges. 3751 */ 3752 case PRIV_NET_BRIDGE: 3753 case PRIV_NET_GRE: 3754 case PRIV_NET_BPF: 3755 case PRIV_NET_RAW: /* Dup, cond. in legacy jail case. */ 3756 case PRIV_NET_ROUTE: 3757 case PRIV_NET_TAP: 3758 case PRIV_NET_SETIFMTU: 3759 case PRIV_NET_SETIFFLAGS: 3760 case PRIV_NET_SETIFCAP: 3761 case PRIV_NET_SETIFDESCR: 3762 case PRIV_NET_SETIFNAME : 3763 case PRIV_NET_SETIFMETRIC: 3764 case PRIV_NET_SETIFPHYS: 3765 case PRIV_NET_SETIFMAC: 3766 case PRIV_NET_ADDMULTI: 3767 case PRIV_NET_DELMULTI: 3768 case PRIV_NET_HWIOCTL: 3769 case PRIV_NET_SETLLADDR: 3770 case PRIV_NET_ADDIFGROUP: 3771 case PRIV_NET_DELIFGROUP: 3772 case PRIV_NET_IFCREATE: 3773 case PRIV_NET_IFDESTROY: 3774 case PRIV_NET_ADDIFADDR: 3775 case PRIV_NET_DELIFADDR: 3776 case PRIV_NET_LAGG: 3777 case PRIV_NET_GIF: 3778 case PRIV_NET_SETIFVNET: 3779 case PRIV_NET_SETIFFIB: 3780 3781 /* 3782 * 802.11-related privileges. 3783 */ 3784 case PRIV_NET80211_GETKEY: 3785#ifdef notyet 3786 case PRIV_NET80211_MANAGE: /* XXX-BZ discuss with sam@ */ 3787#endif 3788 3789#ifdef notyet 3790 /* 3791 * ATM privileges. 3792 */ 3793 case PRIV_NETATM_CFG: 3794 case PRIV_NETATM_ADD: 3795 case PRIV_NETATM_DEL: 3796 case PRIV_NETATM_SET: 3797 3798 /* 3799 * Bluetooth privileges. 3800 */ 3801 case PRIV_NETBLUETOOTH_RAW: 3802#endif 3803 3804 /* 3805 * Netgraph and netgraph module privileges. 3806 */ 3807 case PRIV_NETGRAPH_CONTROL: 3808#ifdef notyet 3809 case PRIV_NETGRAPH_TTY: 3810#endif 3811 3812 /* 3813 * IPv4 and IPv6 privileges. 3814 */ 3815 case PRIV_NETINET_IPFW: 3816 case PRIV_NETINET_DIVERT: 3817 case PRIV_NETINET_PF: 3818 case PRIV_NETINET_DUMMYNET: 3819 case PRIV_NETINET_CARP: 3820 case PRIV_NETINET_MROUTE: 3821 case PRIV_NETINET_RAW: 3822 case PRIV_NETINET_ADDRCTRL6: 3823 case PRIV_NETINET_ND6: 3824 case PRIV_NETINET_SCOPE6: 3825 case PRIV_NETINET_ALIFETIME6: 3826 case PRIV_NETINET_IPSEC: 3827 case PRIV_NETINET_BINDANY: 3828 3829#ifdef notyet 3830 /* 3831 * NCP privileges. 3832 */ 3833 case PRIV_NETNCP: 3834 3835 /* 3836 * SMB privileges. 3837 */ 3838 case PRIV_NETSMB: 3839#endif 3840 3841 /* 3842 * No default: or deny here. 3843 * In case of no permit fall through to next switch(). 3844 */ 3845 if (cred->cr_prison->pr_flags & PR_VNET) 3846 return (0); 3847 } 3848#endif /* VIMAGE */ 3849 3850 switch (priv) { 3851 3852 /* 3853 * Allow ktrace privileges for root in jail. 3854 */ 3855 case PRIV_KTRACE: 3856 3857#if 0 3858 /* 3859 * Allow jailed processes to configure audit identity and 3860 * submit audit records (login, etc). In the future we may 3861 * want to further refine the relationship between audit and 3862 * jail. 3863 */ 3864 case PRIV_AUDIT_GETAUDIT: 3865 case PRIV_AUDIT_SETAUDIT: 3866 case PRIV_AUDIT_SUBMIT: 3867#endif 3868 3869 /* 3870 * Allow jailed processes to manipulate process UNIX 3871 * credentials in any way they see fit. 3872 */ 3873 case PRIV_CRED_SETUID: 3874 case PRIV_CRED_SETEUID: 3875 case PRIV_CRED_SETGID: 3876 case PRIV_CRED_SETEGID: 3877 case PRIV_CRED_SETGROUPS: 3878 case PRIV_CRED_SETREUID: 3879 case PRIV_CRED_SETREGID: 3880 case PRIV_CRED_SETRESUID: 3881 case PRIV_CRED_SETRESGID: 3882 3883 /* 3884 * Jail implements visibility constraints already, so allow 3885 * jailed root to override uid/gid-based constraints. 3886 */ 3887 case PRIV_SEEOTHERGIDS: 3888 case PRIV_SEEOTHERUIDS: 3889 3890 /* 3891 * Jail implements inter-process debugging limits already, so 3892 * allow jailed root various debugging privileges. 3893 */ 3894 case PRIV_DEBUG_DIFFCRED: 3895 case PRIV_DEBUG_SUGID: 3896 case PRIV_DEBUG_UNPRIV: 3897 3898 /* 3899 * Allow jail to set various resource limits and login 3900 * properties, and for now, exceed process resource limits. 3901 */ 3902 case PRIV_PROC_LIMIT: 3903 case PRIV_PROC_SETLOGIN: 3904 case PRIV_PROC_SETRLIMIT: 3905 3906 /* 3907 * System V and POSIX IPC privileges are granted in jail. 3908 */ 3909 case PRIV_IPC_READ: 3910 case PRIV_IPC_WRITE: 3911 case PRIV_IPC_ADMIN: 3912 case PRIV_IPC_MSGSIZE: 3913 case PRIV_MQ_ADMIN: 3914 3915 /* 3916 * Jail operations within a jail work on child jails. 3917 */ 3918 case PRIV_JAIL_ATTACH: 3919 case PRIV_JAIL_SET: 3920 case PRIV_JAIL_REMOVE: 3921 3922 /* 3923 * Jail implements its own inter-process limits, so allow 3924 * root processes in jail to change scheduling on other 3925 * processes in the same jail. Likewise for signalling. 3926 */ 3927 case PRIV_SCHED_DIFFCRED: 3928 case PRIV_SCHED_CPUSET: 3929 case PRIV_SIGNAL_DIFFCRED: 3930 case PRIV_SIGNAL_SUGID: 3931 3932 /* 3933 * Allow jailed processes to write to sysctls marked as jail 3934 * writable. 3935 */ 3936 case PRIV_SYSCTL_WRITEJAIL: 3937 3938 /* 3939 * Allow root in jail to manage a variety of quota 3940 * properties. These should likely be conditional on a 3941 * configuration option. 3942 */ 3943 case PRIV_VFS_GETQUOTA: 3944 case PRIV_VFS_SETQUOTA: 3945 3946 /* 3947 * Since Jail relies on chroot() to implement file system 3948 * protections, grant many VFS privileges to root in jail. 3949 * Be careful to exclude mount-related and NFS-related 3950 * privileges. 3951 */ 3952 case PRIV_VFS_READ: 3953 case PRIV_VFS_WRITE: 3954 case PRIV_VFS_ADMIN: 3955 case PRIV_VFS_EXEC: 3956 case PRIV_VFS_LOOKUP: 3957 case PRIV_VFS_BLOCKRESERVE: /* XXXRW: Slightly surprising. */ 3958 case PRIV_VFS_CHFLAGS_DEV: 3959 case PRIV_VFS_CHOWN: 3960 case PRIV_VFS_CHROOT: 3961 case PRIV_VFS_RETAINSUGID: 3962 case PRIV_VFS_FCHROOT: 3963 case PRIV_VFS_LINK: 3964 case PRIV_VFS_SETGID: 3965 case PRIV_VFS_STAT: 3966 case PRIV_VFS_STICKYFILE: 3967 3968 /* 3969 * As in the non-jail case, non-root users are expected to be 3970 * able to read kernel/phyiscal memory (provided /dev/[k]mem 3971 * exists in the jail and they have permission to access it). 3972 */ 3973 case PRIV_KMEM_READ: 3974 return (0); 3975 3976 /* 3977 * Depending on the global setting, allow privilege of 3978 * setting system flags. 3979 */ 3980 case PRIV_VFS_SYSFLAGS: 3981 if (cred->cr_prison->pr_allow & PR_ALLOW_CHFLAGS) 3982 return (0); 3983 else 3984 return (EPERM); 3985 3986 /* 3987 * Depending on the global setting, allow privilege of 3988 * mounting/unmounting file systems. 3989 */ 3990 case PRIV_VFS_MOUNT: 3991 case PRIV_VFS_UNMOUNT: 3992 case PRIV_VFS_MOUNT_NONUSER: 3993 case PRIV_VFS_MOUNT_OWNER: 3994 if (cred->cr_prison->pr_allow & PR_ALLOW_MOUNT && 3995 cred->cr_prison->pr_enforce_statfs < 2) 3996 return (0); 3997 else 3998 return (EPERM); 3999 4000 /* 4001 * Allow jailed root to bind reserved ports and reuse in-use 4002 * ports. 4003 */ 4004 case PRIV_NETINET_RESERVEDPORT: 4005 case PRIV_NETINET_REUSEPORT: 4006 return (0); 4007 4008 /* 4009 * Allow jailed root to set certian IPv4/6 (option) headers. 4010 */ 4011 case PRIV_NETINET_SETHDROPTS: 4012 return (0); 4013 4014 /* 4015 * Conditionally allow creating raw sockets in jail. 4016 */ 4017 case PRIV_NETINET_RAW: 4018 if (cred->cr_prison->pr_allow & PR_ALLOW_RAW_SOCKETS) 4019 return (0); 4020 else 4021 return (EPERM); 4022 4023 /* 4024 * Since jail implements its own visibility limits on netstat 4025 * sysctls, allow getcred. This allows identd to work in 4026 * jail. 4027 */ 4028 case PRIV_NETINET_GETCRED: 4029 return (0); 4030 4031 /* 4032 * Allow jailed root to set loginclass. 4033 */ 4034 case PRIV_PROC_SETLOGINCLASS: 4035 return (0); 4036 4037 default: 4038 /* 4039 * In all remaining cases, deny the privilege request. This 4040 * includes almost all network privileges, many system 4041 * configuration privileges. 4042 */ 4043 return (EPERM); 4044 } 4045} 4046 4047/* 4048 * Return the part of pr2's name that is relative to pr1, or the whole name 4049 * if it does not directly follow. 4050 */ 4051 4052char * 4053prison_name(struct prison *pr1, struct prison *pr2) 4054{ 4055 char *name; 4056 4057 /* Jails see themselves as "0" (if they see themselves at all). */ 4058 if (pr1 == pr2) 4059 return "0"; 4060 name = pr2->pr_name; 4061 if (prison_ischild(pr1, pr2)) { 4062 /* 4063 * pr1 isn't locked (and allprison_lock may not be either) 4064 * so its length can't be counted on. But the number of dots 4065 * can be counted on - and counted. 4066 */ 4067 for (; pr1 != &prison0; pr1 = pr1->pr_parent) 4068 name = strchr(name, '.') + 1; 4069 } 4070 return (name); 4071} 4072 4073/* 4074 * Return the part of pr2's path that is relative to pr1, or the whole path 4075 * if it does not directly follow. 4076 */ 4077static char * 4078prison_path(struct prison *pr1, struct prison *pr2) 4079{ 4080 char *path1, *path2; 4081 int len1; 4082 4083 path1 = pr1->pr_path; 4084 path2 = pr2->pr_path; 4085 if (!strcmp(path1, "/")) 4086 return (path2); 4087 len1 = strlen(path1); 4088 if (strncmp(path1, path2, len1)) 4089 return (path2); 4090 if (path2[len1] == '\0') 4091 return "/"; 4092 if (path2[len1] == '/') 4093 return (path2 + len1); 4094 return (path2); 4095} 4096 4097 4098/* 4099 * Jail-related sysctls. 4100 */ 4101static SYSCTL_NODE(_security, OID_AUTO, jail, CTLFLAG_RW, 0, 4102 "Jails"); 4103 4104static int 4105sysctl_jail_list(SYSCTL_HANDLER_ARGS) 4106{ 4107 struct xprison *xp; 4108 struct prison *pr, *cpr; 4109#ifdef INET 4110 struct in_addr *ip4 = NULL; 4111 int ip4s = 0; 4112#endif 4113#ifdef INET6 4114 struct in6_addr *ip6 = NULL; 4115 int ip6s = 0; 4116#endif 4117 int descend, error; 4118 4119 xp = malloc(sizeof(*xp), M_TEMP, M_WAITOK); 4120 pr = req->td->td_ucred->cr_prison; 4121 error = 0; 4122 sx_slock(&allprison_lock); 4123 FOREACH_PRISON_DESCENDANT(pr, cpr, descend) { 4124#if defined(INET) || defined(INET6) 4125 again: 4126#endif 4127 mtx_lock(&cpr->pr_mtx); 4128#ifdef INET 4129 if (cpr->pr_ip4s > 0) { 4130 if (ip4s < cpr->pr_ip4s) { 4131 ip4s = cpr->pr_ip4s; 4132 mtx_unlock(&cpr->pr_mtx); 4133 ip4 = realloc(ip4, ip4s * 4134 sizeof(struct in_addr), M_TEMP, M_WAITOK); 4135 goto again; 4136 } 4137 bcopy(cpr->pr_ip4, ip4, 4138 cpr->pr_ip4s * sizeof(struct in_addr)); 4139 } 4140#endif 4141#ifdef INET6 4142 if (cpr->pr_ip6s > 0) { 4143 if (ip6s < cpr->pr_ip6s) { 4144 ip6s = cpr->pr_ip6s; 4145 mtx_unlock(&cpr->pr_mtx); 4146 ip6 = realloc(ip6, ip6s * 4147 sizeof(struct in6_addr), M_TEMP, M_WAITOK); 4148 goto again; 4149 } 4150 bcopy(cpr->pr_ip6, ip6, 4151 cpr->pr_ip6s * sizeof(struct in6_addr)); 4152 } 4153#endif 4154 if (cpr->pr_ref == 0) { 4155 mtx_unlock(&cpr->pr_mtx); 4156 continue; 4157 } 4158 bzero(xp, sizeof(*xp)); 4159 xp->pr_version = XPRISON_VERSION; 4160 xp->pr_id = cpr->pr_id; 4161 xp->pr_state = cpr->pr_uref > 0 4162 ? PRISON_STATE_ALIVE : PRISON_STATE_DYING; 4163 strlcpy(xp->pr_path, prison_path(pr, cpr), sizeof(xp->pr_path)); 4164 strlcpy(xp->pr_host, cpr->pr_hostname, sizeof(xp->pr_host)); 4165 strlcpy(xp->pr_name, prison_name(pr, cpr), sizeof(xp->pr_name)); 4166#ifdef INET 4167 xp->pr_ip4s = cpr->pr_ip4s; 4168#endif 4169#ifdef INET6 4170 xp->pr_ip6s = cpr->pr_ip6s; 4171#endif 4172 mtx_unlock(&cpr->pr_mtx); 4173 error = SYSCTL_OUT(req, xp, sizeof(*xp)); 4174 if (error) 4175 break; 4176#ifdef INET 4177 if (xp->pr_ip4s > 0) { 4178 error = SYSCTL_OUT(req, ip4, 4179 xp->pr_ip4s * sizeof(struct in_addr)); 4180 if (error) 4181 break; 4182 } 4183#endif 4184#ifdef INET6 4185 if (xp->pr_ip6s > 0) { 4186 error = SYSCTL_OUT(req, ip6, 4187 xp->pr_ip6s * sizeof(struct in6_addr)); 4188 if (error) 4189 break; 4190 } 4191#endif 4192 } 4193 sx_sunlock(&allprison_lock); 4194 free(xp, M_TEMP); 4195#ifdef INET 4196 free(ip4, M_TEMP); 4197#endif 4198#ifdef INET6 4199 free(ip6, M_TEMP); 4200#endif 4201 return (error); 4202} 4203 4204SYSCTL_OID(_security_jail, OID_AUTO, list, 4205 CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0, 4206 sysctl_jail_list, "S", "List of active jails"); 4207 4208static int 4209sysctl_jail_jailed(SYSCTL_HANDLER_ARGS) 4210{ 4211 int error, injail; 4212 4213 injail = jailed(req->td->td_ucred); 4214 error = SYSCTL_OUT(req, &injail, sizeof(injail)); 4215 4216 return (error); 4217} 4218 4219SYSCTL_PROC(_security_jail, OID_AUTO, jailed, 4220 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0, 4221 sysctl_jail_jailed, "I", "Process in jail?"); 4222 4223static int 4224sysctl_jail_vnet(SYSCTL_HANDLER_ARGS) 4225{ 4226 int error, havevnet; 4227#ifdef VIMAGE 4228 struct ucred *cred = req->td->td_ucred; 4229 4230 havevnet = jailed(cred) && prison_owns_vnet(cred); 4231#else 4232 havevnet = 0; 4233#endif 4234 error = SYSCTL_OUT(req, &havevnet, sizeof(havevnet)); 4235 4236 return (error); 4237} 4238 4239SYSCTL_PROC(_security_jail, OID_AUTO, vnet, 4240 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0, 4241 sysctl_jail_vnet, "I", "Jail owns VNET?"); 4242 4243#if defined(INET) || defined(INET6) 4244SYSCTL_UINT(_security_jail, OID_AUTO, jail_max_af_ips, CTLFLAG_RW, 4245 &jail_max_af_ips, 0, 4246 "Number of IP addresses a jail may have at most per address family"); 4247#endif 4248 4249/* 4250 * Default parameters for jail(2) compatability. For historical reasons, 4251 * the sysctl names have varying similarity to the parameter names. Prisons 4252 * just see their own parameters, and can't change them. 4253 */ 4254static int 4255sysctl_jail_default_allow(SYSCTL_HANDLER_ARGS) 4256{ 4257 struct prison *pr; 4258 int allow, error, i; 4259 4260 pr = req->td->td_ucred->cr_prison; 4261 allow = (pr == &prison0) ? jail_default_allow : pr->pr_allow; 4262 4263 /* Get the current flag value, and convert it to a boolean. */ 4264 i = (allow & arg2) ? 1 : 0; 4265 if (arg1 != NULL) 4266 i = !i; 4267 error = sysctl_handle_int(oidp, &i, 0, req); 4268 if (error || !req->newptr) 4269 return (error); 4270 i = i ? arg2 : 0; 4271 if (arg1 != NULL) 4272 i ^= arg2; 4273 /* 4274 * The sysctls don't have CTLFLAGS_PRISON, so assume prison0 4275 * for writing. 4276 */ 4277 mtx_lock(&prison0.pr_mtx); 4278 jail_default_allow = (jail_default_allow & ~arg2) | i; 4279 mtx_unlock(&prison0.pr_mtx); 4280 return (0); 4281} 4282 4283SYSCTL_PROC(_security_jail, OID_AUTO, set_hostname_allowed, 4284 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 4285 NULL, PR_ALLOW_SET_HOSTNAME, sysctl_jail_default_allow, "I", 4286 "Processes in jail can set their hostnames"); 4287SYSCTL_PROC(_security_jail, OID_AUTO, socket_unixiproute_only, 4288 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 4289 (void *)1, PR_ALLOW_SOCKET_AF, sysctl_jail_default_allow, "I", 4290 "Processes in jail are limited to creating UNIX/IP/route sockets only"); 4291SYSCTL_PROC(_security_jail, OID_AUTO, sysvipc_allowed, 4292 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 4293 NULL, PR_ALLOW_SYSVIPC, sysctl_jail_default_allow, "I", 4294 "Processes in jail can use System V IPC primitives"); 4295SYSCTL_PROC(_security_jail, OID_AUTO, allow_raw_sockets, 4296 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 4297 NULL, PR_ALLOW_RAW_SOCKETS, sysctl_jail_default_allow, "I", 4298 "Prison root can create raw sockets"); 4299SYSCTL_PROC(_security_jail, OID_AUTO, chflags_allowed, 4300 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 4301 NULL, PR_ALLOW_CHFLAGS, sysctl_jail_default_allow, "I", 4302 "Processes in jail can alter system file flags"); 4303SYSCTL_PROC(_security_jail, OID_AUTO, mount_allowed, 4304 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 4305 NULL, PR_ALLOW_MOUNT, sysctl_jail_default_allow, "I", 4306 "Processes in jail can mount/unmount jail-friendly file systems"); 4307SYSCTL_PROC(_security_jail, OID_AUTO, mount_devfs_allowed, 4308 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 4309 NULL, PR_ALLOW_MOUNT_DEVFS, sysctl_jail_default_allow, "I", 4310 "Processes in jail can mount the devfs file system"); 4311SYSCTL_PROC(_security_jail, OID_AUTO, mount_fdescfs_allowed, 4312 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 4313 NULL, PR_ALLOW_MOUNT_FDESCFS, sysctl_jail_default_allow, "I", 4314 "Processes in jail can mount the fdescfs file system"); 4315SYSCTL_PROC(_security_jail, OID_AUTO, mount_nullfs_allowed, 4316 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 4317 NULL, PR_ALLOW_MOUNT_NULLFS, sysctl_jail_default_allow, "I", 4318 "Processes in jail can mount the nullfs file system"); 4319SYSCTL_PROC(_security_jail, OID_AUTO, mount_procfs_allowed, 4320 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 4321 NULL, PR_ALLOW_MOUNT_PROCFS, sysctl_jail_default_allow, "I", 4322 "Processes in jail can mount the procfs file system"); 4323SYSCTL_PROC(_security_jail, OID_AUTO, mount_linprocfs_allowed, 4324 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 4325 NULL, PR_ALLOW_MOUNT_LINPROCFS, sysctl_jail_default_allow, "I", 4326 "Processes in jail can mount the linprocfs file system"); 4327SYSCTL_PROC(_security_jail, OID_AUTO, mount_linsysfs_allowed, 4328 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 4329 NULL, PR_ALLOW_MOUNT_LINSYSFS, sysctl_jail_default_allow, "I", 4330 "Processes in jail can mount the linsysfs file system"); 4331SYSCTL_PROC(_security_jail, OID_AUTO, mount_tmpfs_allowed, 4332 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 4333 NULL, PR_ALLOW_MOUNT_TMPFS, sysctl_jail_default_allow, "I", 4334 "Processes in jail can mount the tmpfs file system"); 4335SYSCTL_PROC(_security_jail, OID_AUTO, mount_zfs_allowed, 4336 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 4337 NULL, PR_ALLOW_MOUNT_ZFS, sysctl_jail_default_allow, "I", 4338 "Processes in jail can mount the zfs file system"); 4339 4340static int 4341sysctl_jail_default_level(SYSCTL_HANDLER_ARGS) 4342{ 4343 struct prison *pr; 4344 int level, error; 4345 4346 pr = req->td->td_ucred->cr_prison; 4347 level = (pr == &prison0) ? *(int *)arg1 : *(int *)((char *)pr + arg2); 4348 error = sysctl_handle_int(oidp, &level, 0, req); 4349 if (error || !req->newptr) 4350 return (error); 4351 *(int *)arg1 = level; 4352 return (0); 4353} 4354 4355SYSCTL_PROC(_security_jail, OID_AUTO, enforce_statfs, 4356 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 4357 &jail_default_enforce_statfs, offsetof(struct prison, pr_enforce_statfs), 4358 sysctl_jail_default_level, "I", 4359 "Processes in jail cannot see all mounted file systems"); 4360 4361SYSCTL_PROC(_security_jail, OID_AUTO, devfs_ruleset, 4362 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, 4363 &jail_default_devfs_rsnum, offsetof(struct prison, pr_devfs_rsnum), 4364 sysctl_jail_default_level, "I", 4365 "Ruleset for the devfs filesystem in jail"); 4366 4367/* 4368 * Nodes to describe jail parameters. Maximum length of string parameters 4369 * is returned in the string itself, and the other parameters exist merely 4370 * to make themselves and their types known. 4371 */ 4372SYSCTL_NODE(_security_jail, OID_AUTO, param, CTLFLAG_RW, 0, 4373 "Jail parameters"); 4374 4375int 4376sysctl_jail_param(SYSCTL_HANDLER_ARGS) 4377{ 4378 int i; 4379 long l; 4380 size_t s; 4381 char numbuf[12]; 4382 4383 switch (oidp->oid_kind & CTLTYPE) 4384 { 4385 case CTLTYPE_LONG: 4386 case CTLTYPE_ULONG: 4387 l = 0; 4388#ifdef SCTL_MASK32 4389 if (!(req->flags & SCTL_MASK32)) 4390#endif 4391 return (SYSCTL_OUT(req, &l, sizeof(l))); 4392 case CTLTYPE_INT: 4393 case CTLTYPE_UINT: 4394 i = 0; 4395 return (SYSCTL_OUT(req, &i, sizeof(i))); 4396 case CTLTYPE_STRING: 4397 snprintf(numbuf, sizeof(numbuf), "%jd", (intmax_t)arg2); 4398 return 4399 (sysctl_handle_string(oidp, numbuf, sizeof(numbuf), req)); 4400 case CTLTYPE_STRUCT: 4401 s = (size_t)arg2; 4402 return (SYSCTL_OUT(req, &s, sizeof(s))); 4403 } 4404 return (0); 4405} 4406 4407/* 4408 * CTLFLAG_RDTUN in the following indicates jail parameters that can be set at 4409 * jail creation time but cannot be changed in an existing jail. 4410 */ 4411SYSCTL_JAIL_PARAM(, jid, CTLTYPE_INT | CTLFLAG_RDTUN, "I", "Jail ID"); 4412SYSCTL_JAIL_PARAM(, parent, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail parent ID"); 4413SYSCTL_JAIL_PARAM_STRING(, name, CTLFLAG_RW, MAXHOSTNAMELEN, "Jail name"); 4414SYSCTL_JAIL_PARAM_STRING(, path, CTLFLAG_RDTUN, MAXPATHLEN, "Jail root path"); 4415SYSCTL_JAIL_PARAM(, securelevel, CTLTYPE_INT | CTLFLAG_RW, 4416 "I", "Jail secure level"); 4417SYSCTL_JAIL_PARAM(, osreldate, CTLTYPE_INT | CTLFLAG_RDTUN, "I", 4418 "Jail value for kern.osreldate and uname -K"); 4419SYSCTL_JAIL_PARAM_STRING(, osrelease, CTLFLAG_RDTUN, OSRELEASELEN, 4420 "Jail value for kern.osrelease and uname -r"); 4421SYSCTL_JAIL_PARAM(, enforce_statfs, CTLTYPE_INT | CTLFLAG_RW, 4422 "I", "Jail cannot see all mounted file systems"); 4423SYSCTL_JAIL_PARAM(, devfs_ruleset, CTLTYPE_INT | CTLFLAG_RW, 4424 "I", "Ruleset for in-jail devfs mounts"); 4425SYSCTL_JAIL_PARAM(, persist, CTLTYPE_INT | CTLFLAG_RW, 4426 "B", "Jail persistence"); 4427#ifdef VIMAGE 4428SYSCTL_JAIL_PARAM(, vnet, CTLTYPE_INT | CTLFLAG_RDTUN, 4429 "E,jailsys", "Virtual network stack"); 4430#endif 4431SYSCTL_JAIL_PARAM(, dying, CTLTYPE_INT | CTLFLAG_RD, 4432 "B", "Jail is in the process of shutting down"); 4433 4434SYSCTL_JAIL_PARAM_NODE(children, "Number of child jails"); 4435SYSCTL_JAIL_PARAM(_children, cur, CTLTYPE_INT | CTLFLAG_RD, 4436 "I", "Current number of child jails"); 4437SYSCTL_JAIL_PARAM(_children, max, CTLTYPE_INT | CTLFLAG_RW, 4438 "I", "Maximum number of child jails"); 4439 4440SYSCTL_JAIL_PARAM_SYS_NODE(host, CTLFLAG_RW, "Jail host info"); 4441SYSCTL_JAIL_PARAM_STRING(_host, hostname, CTLFLAG_RW, MAXHOSTNAMELEN, 4442 "Jail hostname"); 4443SYSCTL_JAIL_PARAM_STRING(_host, domainname, CTLFLAG_RW, MAXHOSTNAMELEN, 4444 "Jail NIS domainname"); 4445SYSCTL_JAIL_PARAM_STRING(_host, hostuuid, CTLFLAG_RW, HOSTUUIDLEN, 4446 "Jail host UUID"); 4447SYSCTL_JAIL_PARAM(_host, hostid, CTLTYPE_ULONG | CTLFLAG_RW, 4448 "LU", "Jail host ID"); 4449 4450SYSCTL_JAIL_PARAM_NODE(cpuset, "Jail cpuset"); 4451SYSCTL_JAIL_PARAM(_cpuset, id, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail cpuset ID"); 4452 4453#ifdef INET 4454SYSCTL_JAIL_PARAM_SYS_NODE(ip4, CTLFLAG_RDTUN, 4455 "Jail IPv4 address virtualization"); 4456SYSCTL_JAIL_PARAM_STRUCT(_ip4, addr, CTLFLAG_RW, sizeof(struct in_addr), 4457 "S,in_addr,a", "Jail IPv4 addresses"); 4458SYSCTL_JAIL_PARAM(_ip4, saddrsel, CTLTYPE_INT | CTLFLAG_RW, 4459 "B", "Do (not) use IPv4 source address selection rather than the " 4460 "primary jail IPv4 address."); 4461#endif 4462#ifdef INET6 4463SYSCTL_JAIL_PARAM_SYS_NODE(ip6, CTLFLAG_RDTUN, 4464 "Jail IPv6 address virtualization"); 4465SYSCTL_JAIL_PARAM_STRUCT(_ip6, addr, CTLFLAG_RW, sizeof(struct in6_addr), 4466 "S,in6_addr,a", "Jail IPv6 addresses"); 4467SYSCTL_JAIL_PARAM(_ip6, saddrsel, CTLTYPE_INT | CTLFLAG_RW, 4468 "B", "Do (not) use IPv6 source address selection rather than the " 4469 "primary jail IPv6 address."); 4470#endif 4471 4472SYSCTL_JAIL_PARAM_NODE(allow, "Jail permission flags"); 4473SYSCTL_JAIL_PARAM(_allow, set_hostname, CTLTYPE_INT | CTLFLAG_RW, 4474 "B", "Jail may set hostname"); 4475SYSCTL_JAIL_PARAM(_allow, sysvipc, CTLTYPE_INT | CTLFLAG_RW, 4476 "B", "Jail may use SYSV IPC"); 4477SYSCTL_JAIL_PARAM(_allow, raw_sockets, CTLTYPE_INT | CTLFLAG_RW, 4478 "B", "Jail may create raw sockets"); 4479SYSCTL_JAIL_PARAM(_allow, chflags, CTLTYPE_INT | CTLFLAG_RW, 4480 "B", "Jail may alter system file flags"); 4481SYSCTL_JAIL_PARAM(_allow, quotas, CTLTYPE_INT | CTLFLAG_RW, 4482 "B", "Jail may set file quotas"); 4483SYSCTL_JAIL_PARAM(_allow, socket_af, CTLTYPE_INT | CTLFLAG_RW, 4484 "B", "Jail may create sockets other than just UNIX/IPv4/IPv6/route"); 4485 4486SYSCTL_JAIL_PARAM_SUBNODE(allow, mount, "Jail mount/unmount permission flags"); 4487SYSCTL_JAIL_PARAM(_allow_mount, , CTLTYPE_INT | CTLFLAG_RW, 4488 "B", "Jail may mount/unmount jail-friendly file systems in general"); 4489SYSCTL_JAIL_PARAM(_allow_mount, devfs, CTLTYPE_INT | CTLFLAG_RW, 4490 "B", "Jail may mount the devfs file system"); 4491SYSCTL_JAIL_PARAM(_allow_mount, fdescfs, CTLTYPE_INT | CTLFLAG_RW, 4492 "B", "Jail may mount the fdescfs file system"); 4493SYSCTL_JAIL_PARAM(_allow_mount, nullfs, CTLTYPE_INT | CTLFLAG_RW, 4494 "B", "Jail may mount the nullfs file system"); 4495SYSCTL_JAIL_PARAM(_allow_mount, procfs, CTLTYPE_INT | CTLFLAG_RW, 4496 "B", "Jail may mount the procfs file system"); 4497SYSCTL_JAIL_PARAM(_allow_mount, linprocfs, CTLTYPE_INT | CTLFLAG_RW, 4498 "B", "Jail may mount the linprocfs file system"); 4499SYSCTL_JAIL_PARAM(_allow_mount, linsysfs, CTLTYPE_INT | CTLFLAG_RW, 4500 "B", "Jail may mount the linsysfs file system"); 4501SYSCTL_JAIL_PARAM(_allow_mount, tmpfs, CTLTYPE_INT | CTLFLAG_RW, 4502 "B", "Jail may mount the tmpfs file system"); 4503SYSCTL_JAIL_PARAM(_allow_mount, zfs, CTLTYPE_INT | CTLFLAG_RW, 4504 "B", "Jail may mount the zfs file system"); 4505 4506#ifdef RACCT 4507void 4508prison_racct_foreach(void (*callback)(struct racct *racct, 4509 void *arg2, void *arg3), void (*pre)(void), void (*post)(void), 4510 void *arg2, void *arg3) 4511{ 4512 struct prison_racct *prr; 4513 4514 ASSERT_RACCT_ENABLED(); 4515 4516 sx_slock(&allprison_lock); 4517 if (pre != NULL) 4518 (pre)(); 4519 LIST_FOREACH(prr, &allprison_racct, prr_next) 4520 (callback)(prr->prr_racct, arg2, arg3); 4521 if (post != NULL) 4522 (post)(); 4523 sx_sunlock(&allprison_lock); 4524} 4525 4526static struct prison_racct * 4527prison_racct_find_locked(const char *name) 4528{ 4529 struct prison_racct *prr; 4530 4531 ASSERT_RACCT_ENABLED(); 4532 sx_assert(&allprison_lock, SA_XLOCKED); 4533 4534 if (name[0] == '\0' || strlen(name) >= MAXHOSTNAMELEN) 4535 return (NULL); 4536 4537 LIST_FOREACH(prr, &allprison_racct, prr_next) { 4538 if (strcmp(name, prr->prr_name) != 0) 4539 continue; 4540 4541 /* Found prison_racct with a matching name? */ 4542 prison_racct_hold(prr); 4543 return (prr); 4544 } 4545 4546 /* Add new prison_racct. */ 4547 prr = malloc(sizeof(*prr), M_PRISON_RACCT, M_ZERO | M_WAITOK); 4548 racct_create(&prr->prr_racct); 4549 4550 strcpy(prr->prr_name, name); 4551 refcount_init(&prr->prr_refcount, 1); 4552 LIST_INSERT_HEAD(&allprison_racct, prr, prr_next); 4553 4554 return (prr); 4555} 4556 4557struct prison_racct * 4558prison_racct_find(const char *name) 4559{ 4560 struct prison_racct *prr; 4561 4562 ASSERT_RACCT_ENABLED(); 4563 4564 sx_xlock(&allprison_lock); 4565 prr = prison_racct_find_locked(name); 4566 sx_xunlock(&allprison_lock); 4567 return (prr); 4568} 4569 4570void 4571prison_racct_hold(struct prison_racct *prr) 4572{ 4573 4574 ASSERT_RACCT_ENABLED(); 4575 4576 refcount_acquire(&prr->prr_refcount); 4577} 4578 4579static void 4580prison_racct_free_locked(struct prison_racct *prr) 4581{ 4582 4583 ASSERT_RACCT_ENABLED(); 4584 sx_assert(&allprison_lock, SA_XLOCKED); 4585 4586 if (refcount_release(&prr->prr_refcount)) { 4587 racct_destroy(&prr->prr_racct); 4588 LIST_REMOVE(prr, prr_next); 4589 free(prr, M_PRISON_RACCT); 4590 } 4591} 4592 4593void 4594prison_racct_free(struct prison_racct *prr) 4595{ 4596 int old; 4597 4598 ASSERT_RACCT_ENABLED(); 4599 sx_assert(&allprison_lock, SA_UNLOCKED); 4600 4601 old = prr->prr_refcount; 4602 if (old > 1 && atomic_cmpset_int(&prr->prr_refcount, old, old - 1)) 4603 return; 4604 4605 sx_xlock(&allprison_lock); 4606 prison_racct_free_locked(prr); 4607 sx_xunlock(&allprison_lock); 4608} 4609 4610static void 4611prison_racct_attach(struct prison *pr) 4612{ 4613 struct prison_racct *prr; 4614 4615 ASSERT_RACCT_ENABLED(); 4616 sx_assert(&allprison_lock, SA_XLOCKED); 4617 4618 prr = prison_racct_find_locked(pr->pr_name); 4619 KASSERT(prr != NULL, ("cannot find prison_racct")); 4620 4621 pr->pr_prison_racct = prr; 4622} 4623 4624/* 4625 * Handle jail renaming. From the racct point of view, renaming means 4626 * moving from one prison_racct to another. 4627 */ 4628static void 4629prison_racct_modify(struct prison *pr) 4630{ 4631 struct proc *p; 4632 struct ucred *cred; 4633 struct prison_racct *oldprr; 4634 4635 ASSERT_RACCT_ENABLED(); 4636 4637 sx_slock(&allproc_lock); 4638 sx_xlock(&allprison_lock); 4639 4640 if (strcmp(pr->pr_name, pr->pr_prison_racct->prr_name) == 0) { 4641 sx_xunlock(&allprison_lock); 4642 sx_sunlock(&allproc_lock); 4643 return; 4644 } 4645 4646 oldprr = pr->pr_prison_racct; 4647 pr->pr_prison_racct = NULL; 4648 4649 prison_racct_attach(pr); 4650 4651 /* 4652 * Move resource utilisation records. 4653 */ 4654 racct_move(pr->pr_prison_racct->prr_racct, oldprr->prr_racct); 4655 4656 /* 4657 * Force rctl to reattach rules to processes. 4658 */ 4659 FOREACH_PROC_IN_SYSTEM(p) { 4660 PROC_LOCK(p); 4661 cred = crhold(p->p_ucred); 4662 PROC_UNLOCK(p); 4663 racct_proc_ucred_changed(p, cred, cred); 4664 crfree(cred); 4665 } 4666 4667 sx_sunlock(&allproc_lock); 4668 prison_racct_free_locked(oldprr); 4669 sx_xunlock(&allprison_lock); 4670} 4671 4672static void 4673prison_racct_detach(struct prison *pr) 4674{ 4675 4676 ASSERT_RACCT_ENABLED(); 4677 sx_assert(&allprison_lock, SA_UNLOCKED); 4678 4679 if (pr->pr_prison_racct == NULL) 4680 return; 4681 prison_racct_free(pr->pr_prison_racct); 4682 pr->pr_prison_racct = NULL; 4683} 4684#endif /* RACCT */ 4685 4686#ifdef DDB 4687 4688static void 4689db_show_prison(struct prison *pr) 4690{ 4691 int fi; 4692#if defined(INET) || defined(INET6) 4693 int ii; 4694#endif 4695 unsigned jsf; 4696#ifdef INET6 4697 char ip6buf[INET6_ADDRSTRLEN]; 4698#endif 4699 4700 db_printf("prison %p:\n", pr); 4701 db_printf(" jid = %d\n", pr->pr_id); 4702 db_printf(" name = %s\n", pr->pr_name); 4703 db_printf(" parent = %p\n", pr->pr_parent); 4704 db_printf(" ref = %d\n", pr->pr_ref); 4705 db_printf(" uref = %d\n", pr->pr_uref); 4706 db_printf(" path = %s\n", pr->pr_path); 4707 db_printf(" cpuset = %d\n", pr->pr_cpuset 4708 ? pr->pr_cpuset->cs_id : -1); 4709#ifdef VIMAGE 4710 db_printf(" vnet = %p\n", pr->pr_vnet); 4711#endif 4712 db_printf(" root = %p\n", pr->pr_root); 4713 db_printf(" securelevel = %d\n", pr->pr_securelevel); 4714 db_printf(" devfs_rsnum = %d\n", pr->pr_devfs_rsnum); 4715 db_printf(" children.max = %d\n", pr->pr_childmax); 4716 db_printf(" children.cur = %d\n", pr->pr_childcount); 4717 db_printf(" child = %p\n", LIST_FIRST(&pr->pr_children)); 4718 db_printf(" sibling = %p\n", LIST_NEXT(pr, pr_sibling)); 4719 db_printf(" flags = 0x%x", pr->pr_flags); 4720 for (fi = 0; fi < nitems(pr_flag_names); fi++) 4721 if (pr_flag_names[fi] != NULL && (pr->pr_flags & (1 << fi))) 4722 db_printf(" %s", pr_flag_names[fi]); 4723 for (fi = 0; fi < nitems(pr_flag_jailsys); fi++) { 4724 jsf = pr->pr_flags & 4725 (pr_flag_jailsys[fi].disable | pr_flag_jailsys[fi].new); 4726 db_printf(" %-16s= %s\n", pr_flag_jailsys[fi].name, 4727 pr_flag_jailsys[fi].disable && 4728 (jsf == pr_flag_jailsys[fi].disable) ? "disable" 4729 : (jsf == pr_flag_jailsys[fi].new) ? "new" 4730 : "inherit"); 4731 } 4732 db_printf(" allow = 0x%x", pr->pr_allow); 4733 for (fi = 0; fi < nitems(pr_allow_names); fi++) 4734 if (pr_allow_names[fi] != NULL && (pr->pr_allow & (1 << fi))) 4735 db_printf(" %s", pr_allow_names[fi]); 4736 db_printf("\n"); 4737 db_printf(" enforce_statfs = %d\n", pr->pr_enforce_statfs); 4738 db_printf(" host.hostname = %s\n", pr->pr_hostname); 4739 db_printf(" host.domainname = %s\n", pr->pr_domainname); 4740 db_printf(" host.hostuuid = %s\n", pr->pr_hostuuid); 4741 db_printf(" host.hostid = %lu\n", pr->pr_hostid); 4742#ifdef INET 4743 db_printf(" ip4s = %d\n", pr->pr_ip4s); 4744 for (ii = 0; ii < pr->pr_ip4s; ii++) 4745 db_printf(" %s %s\n", 4746 ii == 0 ? "ip4.addr =" : " ", 4747 inet_ntoa(pr->pr_ip4[ii])); 4748#endif 4749#ifdef INET6 4750 db_printf(" ip6s = %d\n", pr->pr_ip6s); 4751 for (ii = 0; ii < pr->pr_ip6s; ii++) 4752 db_printf(" %s %s\n", 4753 ii == 0 ? "ip6.addr =" : " ", 4754 ip6_sprintf(ip6buf, &pr->pr_ip6[ii])); 4755#endif 4756} 4757 4758DB_SHOW_COMMAND(prison, db_show_prison_command) 4759{ 4760 struct prison *pr; 4761 4762 if (!have_addr) { 4763 /* 4764 * Show all prisons in the list, and prison0 which is not 4765 * listed. 4766 */ 4767 db_show_prison(&prison0); 4768 if (!db_pager_quit) { 4769 TAILQ_FOREACH(pr, &allprison, pr_list) { 4770 db_show_prison(pr); 4771 if (db_pager_quit) 4772 break; 4773 } 4774 } 4775 return; 4776 } 4777 4778 if (addr == 0) 4779 pr = &prison0; 4780 else { 4781 /* Look for a prison with the ID and with references. */ 4782 TAILQ_FOREACH(pr, &allprison, pr_list) 4783 if (pr->pr_id == addr && pr->pr_ref > 0) 4784 break; 4785 if (pr == NULL) 4786 /* Look again, without requiring a reference. */ 4787 TAILQ_FOREACH(pr, &allprison, pr_list) 4788 if (pr->pr_id == addr) 4789 break; 4790 if (pr == NULL) 4791 /* Assume address points to a valid prison. */ 4792 pr = (struct prison *)addr; 4793 } 4794 db_show_prison(pr); 4795} 4796 4797#endif /* DDB */ 4798