kern_jail.c revision 232598
160812Sps/*- 2238730Sdelphij * Copyright (c) 1999 Poul-Henning Kamp. 3330571Sdelphij * Copyright (c) 2008 Bjoern A. Zeeb. 4238730Sdelphij * Copyright (c) 2009 James Gritton. 5238730Sdelphij * All rights reserved. 6238730Sdelphij * 7238730Sdelphij * Redistribution and use in source and binary forms, with or without 8238730Sdelphij * modification, are permitted provided that the following conditions 9238730Sdelphij * are met: 1060786Sps * 1. Redistributions of source code must retain the above copyright 1160786Sps * notice, this list of conditions and the following disclaimer. 1260786Sps * 2. Redistributions in binary form must reproduce the above copyright 1360786Sps * notice, this list of conditions and the following disclaimer in the 1460786Sps * documentation and/or other materials provided with the distribution. 1560786Sps * 1660786Sps * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 1789022Sps * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 1889022Sps * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 1989022Sps * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 2060786Sps * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 2160786Sps * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 2260786Sps * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 2360786Sps * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 2460786Sps * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 2560786Sps * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 2660786Sps * SUCH DAMAGE. 2760786Sps */ 2860786Sps 2960786Sps#include <sys/cdefs.h> 3060786Sps__FBSDID("$FreeBSD: head/sys/kern/kern_jail.c 232598 2012-03-06 11:05:50Z trasz $"); 3160786Sps 3260786Sps#include "opt_compat.h" 3360786Sps#include "opt_ddb.h" 3460786Sps#include "opt_inet.h" 3560786Sps#include "opt_inet6.h" 3660786Sps 3760786Sps#include <sys/param.h> 3860786Sps#include <sys/types.h> 3960786Sps#include <sys/kernel.h> 4060786Sps#include <sys/systm.h> 4160786Sps#include <sys/errno.h> 4260786Sps#include <sys/sysproto.h> 4360786Sps#include <sys/malloc.h> 4460786Sps#include <sys/osd.h> 4560786Sps#include <sys/priv.h> 4660786Sps#include <sys/proc.h> 4760786Sps#include <sys/taskqueue.h> 4889022Sps#include <sys/fcntl.h> 4960786Sps#include <sys/jail.h> 5060786Sps#include <sys/lock.h> 5160786Sps#include <sys/mutex.h> 5260786Sps#include <sys/racct.h> 5389022Sps#include <sys/refcount.h> 5489022Sps#include <sys/sx.h> 5589022Sps#include <sys/sysent.h> 5689022Sps#include <sys/namei.h> 57330571Sdelphij#include <sys/mount.h> 58221715Sdelphij#include <sys/queue.h> 5960786Sps#include <sys/socket.h> 6060786Sps#include <sys/syscallsubr.h> 61171009Sdelphij#include <sys/sysctl.h> 62170259Sdelphij#include <sys/vnode.h> 63330571Sdelphij 6460786Sps#include <net/if.h> 6560786Sps#include <net/vnet.h> 6660786Sps 6760786Sps#include <netinet/in.h> 6860786Sps 6960786Sps#ifdef DDB 7060786Sps#include <ddb/ddb.h> 7160786Sps#ifdef INET6 7260786Sps#include <netinet6/in6_var.h> 7360786Sps#endif /* INET6 */ 7460786Sps#endif /* DDB */ 7560786Sps 7660812Sps#include <security/mac/mac_framework.h> 7760786Sps 7860786Sps#define DEFAULT_HOSTUUID "00000000-0000-0000-0000-000000000000" 7960786Sps 8060786SpsMALLOC_DEFINE(M_PRISON, "prison", "Prison structures"); 8160786Spsstatic MALLOC_DEFINE(M_PRISON_RACCT, "prison_racct", "Prison racct structures"); 8260786Sps 8360786Sps/* Keep struct prison prison0 and some code in kern_jail_set() readable. */ 8460786Sps#ifdef INET 8560786Sps#ifdef INET6 8660786Sps#define _PR_IP_SADDRSEL PR_IP4_SADDRSEL|PR_IP6_SADDRSEL 8760786Sps#else 8860786Sps#define _PR_IP_SADDRSEL PR_IP4_SADDRSEL 8960786Sps#endif 9060786Sps#else /* !INET */ 9160786Sps#ifdef INET6 9260786Sps#define _PR_IP_SADDRSEL PR_IP6_SADDRSEL 9360786Sps#else 9460786Sps#define _PR_IP_SADDRSEL 0 9560786Sps#endif 9660786Sps#endif 9760786Sps 9860786Sps/* prison0 describes what is "real" about the system. */ 9960786Spsstruct prison prison0 = { 10060786Sps .pr_id = 0, 10160786Sps .pr_name = "0", 10260786Sps .pr_ref = 1, 10360786Sps .pr_uref = 1, 10460786Sps .pr_path = "/", 10560786Sps .pr_securelevel = -1, 10660786Sps .pr_devfs_rsnum = 0, 10760786Sps .pr_childmax = JAIL_MAX, 10860786Sps .pr_hostuuid = DEFAULT_HOSTUUID, 10960786Sps .pr_children = LIST_HEAD_INITIALIZER(prison0.pr_children), 11089022Sps#ifdef VIMAGE 11160786Sps .pr_flags = PR_HOST|PR_VNET|_PR_IP_SADDRSEL, 11260786Sps#else 11360786Sps .pr_flags = PR_HOST|_PR_IP_SADDRSEL, 11460786Sps#endif 11560786Sps .pr_allow = PR_ALLOW_ALL, 11660786Sps}; 11760786SpsMTX_SYSINIT(prison0, &prison0.pr_mtx, "jail mutex", MTX_DEF); 118330571Sdelphij 11960786Sps/* allprison, allprison_racct and lastprid are protected by allprison_lock. */ 120330571Sdelphijstruct sx allprison_lock; 12160786SpsSX_SYSINIT(allprison_lock, &allprison_lock, "allprison"); 12260786Spsstruct prisonlist allprison = TAILQ_HEAD_INITIALIZER(allprison); 123161478SdelphijLIST_HEAD(, prison_racct) allprison_racct; 12460786Spsint lastprid = 0; 125195941Sdelphij 126170259Sdelphijstatic int do_jail_attach(struct thread *td, struct prison *pr); 127170259Sdelphijstatic void prison_complete(void *context, int pending); 128170259Sdelphijstatic void prison_deref(struct prison *pr, int flags); 129170259Sdelphijstatic char *prison_path(struct prison *pr1, struct prison *pr2); 130170259Sdelphijstatic void prison_remove_one(struct prison *pr); 131170259Sdelphij#ifdef RACCT 132170259Sdelphijstatic void prison_racct_attach(struct prison *pr); 133170259Sdelphijstatic void prison_racct_modify(struct prison *pr); 134170259Sdelphijstatic void prison_racct_detach(struct prison *pr); 13560812Sps#endif 136170259Sdelphij#ifdef INET 137170259Sdelphijstatic int _prison_check_ip4(struct prison *pr, struct in_addr *ia); 138170259Sdelphijstatic int prison_restrict_ip4(struct prison *pr, struct in_addr *newip4); 139170259Sdelphij#endif 140170259Sdelphij#ifdef INET6 141170812Sdelphijstatic int _prison_check_ip6(struct prison *pr, struct in6_addr *ia6); 142170812Sdelphijstatic int prison_restrict_ip6(struct prison *pr, struct in6_addr *newip6); 143170812Sdelphij#endif 144170259Sdelphij 14560786Sps/* Flags for prison_deref */ 14660786Sps#define PD_DEREF 0x01 14760786Sps#define PD_DEUREF 0x02 148170963Sdelphij#define PD_LOCKED 0x04 149170963Sdelphij#define PD_LIST_SLOCKED 0x08 15060786Sps#define PD_LIST_XLOCKED 0x10 15160786Sps 15260786Sps/* 15360786Sps * Parameter names corresponding to PR_* flag values. Size values are for kvm 15460786Sps * as we cannot figure out the size of a sparse array, or an array without a 15560786Sps * terminating entry. 15660786Sps */ 15760786Spsstatic char *pr_flag_names[] = { 15860786Sps [0] = "persist", 15960786Sps#ifdef INET 16060786Sps [7] = "ip4.saddrsel", 16160786Sps#endif 16260786Sps#ifdef INET6 16360786Sps [8] = "ip6.saddrsel", 16460786Sps#endif 16560786Sps}; 16660786Spsconst size_t pr_flag_names_size = sizeof(pr_flag_names); 16760786Sps 16860786Spsstatic char *pr_flag_nonames[] = { 16960786Sps [0] = "nopersist", 170171817Sdelphij#ifdef INET 171171817Sdelphij [7] = "ip4.nosaddrsel", 172170259Sdelphij#endif 17360786Sps#ifdef INET6 17460786Sps [8] = "ip6.nosaddrsel", 17560786Sps#endif 17660786Sps}; 17760786Spsconst size_t pr_flag_nonames_size = sizeof(pr_flag_nonames); 17860786Sps 17960786Spsstruct jailsys_flags { 18060786Sps const char *name; 18160786Sps unsigned disable; 18260786Sps unsigned new; 18360786Sps} pr_flag_jailsys[] = { 18460786Sps { "host", 0, PR_HOST }, 18560786Sps#ifdef VIMAGE 18660786Sps { "vnet", 0, PR_VNET }, 18760786Sps#endif 18860786Sps#ifdef INET 18960786Sps { "ip4", PR_IP4_USER | PR_IP4_DISABLE, PR_IP4_USER }, 19060786Sps#endif 19160786Sps#ifdef INET6 19260786Sps { "ip6", PR_IP6_USER | PR_IP6_DISABLE, PR_IP6_USER }, 19360786Sps#endif 19460786Sps}; 19589022Spsconst size_t pr_flag_jailsys_size = sizeof(pr_flag_jailsys); 19660786Sps 19760786Spsstatic char *pr_allow_names[] = { 19860786Sps "allow.set_hostname", 19960786Sps "allow.sysvipc", 20060786Sps "allow.raw_sockets", 20160786Sps "allow.chflags", 20260786Sps "allow.mount", 203330571Sdelphij "allow.quotas", 20460786Sps "allow.socket_af", 205330571Sdelphij "allow.mount.devfs", 20660786Sps "allow.mount.nullfs", 20760786Sps "allow.mount.zfs", 20860786Sps "allow.mount.procfs", 20960786Sps}; 21060786Spsconst size_t pr_allow_names_size = sizeof(pr_allow_names); 211128348Stjr 212330571Sdelphijstatic char *pr_allow_nonames[] = { 213330571Sdelphij "allow.noset_hostname", 214330571Sdelphij "allow.nosysvipc", 215128348Stjr "allow.noraw_sockets", 216128348Stjr "allow.nochflags", 21760786Sps "allow.nomount", 21860786Sps "allow.noquotas", 219330571Sdelphij "allow.nosocket_af", 220128348Stjr "allow.mount.nodevfs", 22160786Sps "allow.mount.nonullfs", 22260786Sps "allow.mount.nozfs", 22360786Sps "allow.mount.noprocfs", 22460786Sps}; 22560786Spsconst size_t pr_allow_nonames_size = sizeof(pr_allow_nonames); 22660786Sps 22760786Sps#define JAIL_DEFAULT_ALLOW PR_ALLOW_SET_HOSTNAME 22860786Sps#define JAIL_DEFAULT_ENFORCE_STATFS 2 22960786Sps#define JAIL_DEFAULT_DEVFS_RSNUM 0 23060786Spsstatic unsigned jail_default_allow = JAIL_DEFAULT_ALLOW; 23160786Spsstatic int jail_default_enforce_statfs = JAIL_DEFAULT_ENFORCE_STATFS; 23260786Spsstatic int jail_default_devfs_rsnum = JAIL_DEFAULT_DEVFS_RSNUM; 23360786Sps#if defined(INET) || defined(INET6) 23460786Spsstatic unsigned jail_max_af_ips = 255; 23560786Sps#endif 23660786Sps 23760786Sps#ifdef INET 23860786Spsstatic int 23960786Spsqcmp_v4(const void *ip1, const void *ip2) 24060786Sps{ 24160786Sps in_addr_t iaa, iab; 24260786Sps 24360786Sps /* 24460786Sps * We need to compare in HBO here to get the list sorted as expected 24560786Sps * by the result of the code. Sorting NBO addresses gives you 246172045Sdelphij * interesting results. If you do not understand, do not try. 24760786Sps */ 24860786Sps iaa = ntohl(((const struct in_addr *)ip1)->s_addr); 249128348Stjr iab = ntohl(((const struct in_addr *)ip2)->s_addr); 25060786Sps 25160786Sps /* 25260786Sps * Do not simply return the difference of the two numbers, the int is 25360786Sps * not wide enough. 25460786Sps */ 25560786Sps if (iaa > iab) 25660786Sps return (1); 25789022Sps else if (iaa < iab) 25860786Sps return (-1); 25960786Sps else 26060786Sps return (0); 26160786Sps} 26260786Sps#endif 26360786Sps 26460786Sps#ifdef INET6 26560786Spsstatic int 26660786Spsqcmp_v6(const void *ip1, const void *ip2) 26760786Sps{ 26860786Sps const struct in6_addr *ia6a, *ia6b; 26960786Sps int i, rc; 27060786Sps 27160786Sps ia6a = (const struct in6_addr *)ip1; 27260786Sps ia6b = (const struct in6_addr *)ip2; 27360786Sps 27460786Sps rc = 0; 27560786Sps for (i = 0; rc == 0 && i < sizeof(struct in6_addr); i++) { 27660786Sps if (ia6a->s6_addr[i] > ia6b->s6_addr[i]) 27760786Sps rc = 1; 27860786Sps else if (ia6a->s6_addr[i] < ia6b->s6_addr[i]) 27960786Sps rc = -1; 28060786Sps } 28160786Sps return (rc); 28260786Sps} 28360786Sps#endif 28460786Sps 28560786Sps/* 28660786Sps * struct jail_args { 287330571Sdelphij * struct jail *jail; 288330571Sdelphij * }; 28960786Sps */ 29060786Spsint 29160786Spssys_jail(struct thread *td, struct jail_args *uap) 29260786Sps{ 293330571Sdelphij uint32_t version; 294330571Sdelphij int error; 295330571Sdelphij struct jail j; 296330571Sdelphij 297330571Sdelphij error = copyin(uap->jail, &version, sizeof(uint32_t)); 298330571Sdelphij if (error) 299330571Sdelphij return (error); 30060786Sps 30160786Sps switch (version) { 30260786Sps case 0: 30360786Sps { 30460786Sps struct jail_v0 j0; 30560786Sps 306128348Stjr /* FreeBSD single IPv4 jails. */ 30760786Sps bzero(&j, sizeof(struct jail)); 30860786Sps error = copyin(uap->jail, &j0, sizeof(struct jail_v0)); 30960786Sps if (error) 31060786Sps return (error); 31160786Sps j.version = j0.version; 31260786Sps j.path = j0.path; 31360786Sps j.hostname = j0.hostname; 31460786Sps j.ip4s = j0.ip_number; 315330571Sdelphij break; 31660786Sps } 317330571Sdelphij 31860786Sps case 1: 31960786Sps /* 32060786Sps * Version 1 was used by multi-IPv4 jail implementations 32160786Sps * that never made it into the official kernel. 32260786Sps */ 32360786Sps return (EINVAL); 32460786Sps 32560786Sps case 2: /* JAIL_API_VERSION */ 32660786Sps /* FreeBSD multi-IPv4/IPv6,noIP jails. */ 32760786Sps error = copyin(uap->jail, &j, sizeof(struct jail)); 32860786Sps if (error) 32960786Sps return (error); 33060786Sps break; 33160786Sps 33260786Sps default: 333330571Sdelphij /* Sci-Fi jails are not supported, sorry. */ 33460786Sps return (EINVAL); 33560786Sps } 33660786Sps return (kern_jail(td, &j)); 33760786Sps} 33860786Sps 33960786Spsint 34060786Spskern_jail(struct thread *td, struct jail *j) 341128348Stjr{ 34260786Sps struct iovec optiov[2 * (4 34360786Sps + sizeof(pr_allow_names) / sizeof(pr_allow_names[0]) 34460786Sps#ifdef INET 34560786Sps + 1 34660786Sps#endif 34760786Sps#ifdef INET6 34860786Sps + 1 349330571Sdelphij#endif 35060786Sps )]; 35160786Sps struct uio opt; 35260786Sps char *u_path, *u_hostname, *u_name; 35360786Sps#ifdef INET 35460786Sps uint32_t ip4s; 35560786Sps struct in_addr *u_ip4; 35660786Sps#endif 35760786Sps#ifdef INET6 35860786Sps struct in6_addr *u_ip6; 35960786Sps#endif 36060786Sps size_t tmplen; 36160786Sps int error, enforce_statfs, fi; 36260786Sps 36360786Sps bzero(&optiov, sizeof(optiov)); 36460786Sps opt.uio_iov = optiov; 36560786Sps opt.uio_iovcnt = 0; 36660786Sps opt.uio_offset = -1; 367330571Sdelphij opt.uio_resid = -1; 368330571Sdelphij opt.uio_segflg = UIO_SYSSPACE; 369330571Sdelphij opt.uio_rw = UIO_READ; 37060786Sps opt.uio_td = td; 37160786Sps 37260786Sps /* Set permissions for top-level jails from sysctls. */ 37360786Sps if (!jailed(td->td_ucred)) { 37460786Sps for (fi = 0; fi < sizeof(pr_allow_names) / 37560786Sps sizeof(pr_allow_names[0]); fi++) { 376161478Sdelphij optiov[opt.uio_iovcnt].iov_base = 37760786Sps (jail_default_allow & (1 << fi)) 378161478Sdelphij ? pr_allow_names[fi] : pr_allow_nonames[fi]; 379161478Sdelphij optiov[opt.uio_iovcnt].iov_len = 38060786Sps strlen(optiov[opt.uio_iovcnt].iov_base) + 1; 38160786Sps opt.uio_iovcnt += 2; 382161478Sdelphij } 383161478Sdelphij optiov[opt.uio_iovcnt].iov_base = "enforce_statfs"; 38460786Sps optiov[opt.uio_iovcnt].iov_len = sizeof("enforce_statfs"); 38560786Sps opt.uio_iovcnt++; 38660786Sps enforce_statfs = jail_default_enforce_statfs; 38760786Sps optiov[opt.uio_iovcnt].iov_base = &enforce_statfs; 38860786Sps optiov[opt.uio_iovcnt].iov_len = sizeof(enforce_statfs); 38960786Sps opt.uio_iovcnt++; 39060786Sps } 39160786Sps 39260786Sps tmplen = MAXPATHLEN + MAXHOSTNAMELEN + MAXHOSTNAMELEN; 39360786Sps#ifdef INET 39460786Sps ip4s = (j->version == 0) ? 1 : j->ip4s; 39560786Sps if (ip4s > jail_max_af_ips) 39660786Sps return (EINVAL); 39760786Sps tmplen += ip4s * sizeof(struct in_addr); 39860786Sps#else 39960786Sps if (j->ip4s > 0) 40060786Sps return (EINVAL); 40160786Sps#endif 40260786Sps#ifdef INET6 40360786Sps if (j->ip6s > jail_max_af_ips) 40460786Sps return (EINVAL); 40560786Sps tmplen += j->ip6s * sizeof(struct in6_addr); 40660786Sps#else 40760786Sps if (j->ip6s > 0) 40860786Sps return (EINVAL); 40960786Sps#endif 410161478Sdelphij u_path = malloc(tmplen, M_TEMP, M_WAITOK); 41160786Sps u_hostname = u_path + MAXPATHLEN; 41260786Sps u_name = u_hostname + MAXHOSTNAMELEN; 41360786Sps#ifdef INET 41460786Sps u_ip4 = (struct in_addr *)(u_name + MAXHOSTNAMELEN); 41560786Sps#endif 41660786Sps#ifdef INET6 41760786Sps#ifdef INET 41860786Sps u_ip6 = (struct in6_addr *)(u_ip4 + ip4s); 41960786Sps#else 42060786Sps u_ip6 = (struct in6_addr *)(u_name + MAXHOSTNAMELEN); 42160786Sps#endif 42260786Sps#endif 42360786Sps optiov[opt.uio_iovcnt].iov_base = "path"; 42460786Sps optiov[opt.uio_iovcnt].iov_len = sizeof("path"); 425221715Sdelphij opt.uio_iovcnt++; 42689022Sps optiov[opt.uio_iovcnt].iov_base = u_path; 42789022Sps error = copyinstr(j->path, u_path, MAXPATHLEN, 42860786Sps &optiov[opt.uio_iovcnt].iov_len); 42960786Sps if (error) { 43060786Sps free(u_path, M_TEMP); 431 return (error); 432 } 433 opt.uio_iovcnt++; 434 optiov[opt.uio_iovcnt].iov_base = "host.hostname"; 435 optiov[opt.uio_iovcnt].iov_len = sizeof("host.hostname"); 436 opt.uio_iovcnt++; 437 optiov[opt.uio_iovcnt].iov_base = u_hostname; 438 error = copyinstr(j->hostname, u_hostname, MAXHOSTNAMELEN, 439 &optiov[opt.uio_iovcnt].iov_len); 440 if (error) { 441 free(u_path, M_TEMP); 442 return (error); 443 } 444 opt.uio_iovcnt++; 445 if (j->jailname != NULL) { 446 optiov[opt.uio_iovcnt].iov_base = "name"; 447 optiov[opt.uio_iovcnt].iov_len = sizeof("name"); 448 opt.uio_iovcnt++; 449 optiov[opt.uio_iovcnt].iov_base = u_name; 450 error = copyinstr(j->jailname, u_name, MAXHOSTNAMELEN, 451 &optiov[opt.uio_iovcnt].iov_len); 452 if (error) { 453 free(u_path, M_TEMP); 454 return (error); 455 } 456 opt.uio_iovcnt++; 457 } 458#ifdef INET 459 optiov[opt.uio_iovcnt].iov_base = "ip4.addr"; 460 optiov[opt.uio_iovcnt].iov_len = sizeof("ip4.addr"); 461 opt.uio_iovcnt++; 462 optiov[opt.uio_iovcnt].iov_base = u_ip4; 463 optiov[opt.uio_iovcnt].iov_len = ip4s * sizeof(struct in_addr); 464 if (j->version == 0) 465 u_ip4->s_addr = j->ip4s; 466 else { 467 error = copyin(j->ip4, u_ip4, optiov[opt.uio_iovcnt].iov_len); 468 if (error) { 469 free(u_path, M_TEMP); 470 return (error); 471 } 472 } 473 opt.uio_iovcnt++; 474#endif 475#ifdef INET6 476 optiov[opt.uio_iovcnt].iov_base = "ip6.addr"; 477 optiov[opt.uio_iovcnt].iov_len = sizeof("ip6.addr"); 478 opt.uio_iovcnt++; 479 optiov[opt.uio_iovcnt].iov_base = u_ip6; 480 optiov[opt.uio_iovcnt].iov_len = j->ip6s * sizeof(struct in6_addr); 481 error = copyin(j->ip6, u_ip6, optiov[opt.uio_iovcnt].iov_len); 482 if (error) { 483 free(u_path, M_TEMP); 484 return (error); 485 } 486 opt.uio_iovcnt++; 487#endif 488 KASSERT(opt.uio_iovcnt <= sizeof(optiov) / sizeof(optiov[0]), 489 ("kern_jail: too many iovecs (%d)", opt.uio_iovcnt)); 490 error = kern_jail_set(td, &opt, JAIL_CREATE | JAIL_ATTACH); 491 free(u_path, M_TEMP); 492 return (error); 493} 494 495 496/* 497 * struct jail_set_args { 498 * struct iovec *iovp; 499 * unsigned int iovcnt; 500 * int flags; 501 * }; 502 */ 503int 504sys_jail_set(struct thread *td, struct jail_set_args *uap) 505{ 506 struct uio *auio; 507 int error; 508 509 /* Check that we have an even number of iovecs. */ 510 if (uap->iovcnt & 1) 511 return (EINVAL); 512 513 error = copyinuio(uap->iovp, uap->iovcnt, &auio); 514 if (error) 515 return (error); 516 error = kern_jail_set(td, auio, uap->flags); 517 free(auio, M_IOV); 518 return (error); 519} 520 521int 522kern_jail_set(struct thread *td, struct uio *optuio, int flags) 523{ 524 struct nameidata nd; 525#ifdef INET 526 struct in_addr *ip4; 527#endif 528#ifdef INET6 529 struct in6_addr *ip6; 530#endif 531 struct vfsopt *opt; 532 struct vfsoptlist *opts; 533 struct prison *pr, *deadpr, *mypr, *ppr, *tpr; 534 struct vnode *root; 535 char *domain, *errmsg, *host, *name, *namelc, *p, *path, *uuid; 536 char *g_path; 537#if defined(INET) || defined(INET6) 538 struct prison *tppr; 539 void *op; 540#endif 541 unsigned long hid; 542 size_t namelen, onamelen; 543 int created, cuflags, descend, enforce, error, errmsg_len, errmsg_pos; 544 int gotchildmax, gotenforce, gothid, gotrsnum, gotslevel; 545 int fi, jid, jsys, len, level; 546 int childmax, rsnum, slevel, vfslocked; 547 int fullpath_disabled; 548#if defined(INET) || defined(INET6) 549 int ii, ij; 550#endif 551#ifdef INET 552 int ip4s, redo_ip4; 553#endif 554#ifdef INET6 555 int ip6s, redo_ip6; 556#endif 557 uint64_t pr_allow, ch_allow, pr_flags, ch_flags; 558 unsigned tallow; 559 char numbuf[12]; 560 561 error = priv_check(td, PRIV_JAIL_SET); 562 if (!error && (flags & JAIL_ATTACH)) 563 error = priv_check(td, PRIV_JAIL_ATTACH); 564 if (error) 565 return (error); 566 mypr = ppr = td->td_ucred->cr_prison; 567 if ((flags & JAIL_CREATE) && mypr->pr_childmax == 0) 568 return (EPERM); 569 if (flags & ~JAIL_SET_MASK) 570 return (EINVAL); 571 572 /* 573 * Check all the parameters before committing to anything. Not all 574 * errors can be caught early, but we may as well try. Also, this 575 * takes care of some expensive stuff (path lookup) before getting 576 * the allprison lock. 577 * 578 * XXX Jails are not filesystems, and jail parameters are not mount 579 * options. But it makes more sense to re-use the vfsopt code 580 * than duplicate it under a different name. 581 */ 582 error = vfs_buildopts(optuio, &opts); 583 if (error) 584 return (error); 585#ifdef INET 586 ip4 = NULL; 587#endif 588#ifdef INET6 589 ip6 = NULL; 590#endif 591 g_path = NULL; 592 593 error = vfs_copyopt(opts, "jid", &jid, sizeof(jid)); 594 if (error == ENOENT) 595 jid = 0; 596 else if (error != 0) 597 goto done_free; 598 599 error = vfs_copyopt(opts, "securelevel", &slevel, sizeof(slevel)); 600 if (error == ENOENT) 601 gotslevel = 0; 602 else if (error != 0) 603 goto done_free; 604 else 605 gotslevel = 1; 606 607 error = 608 vfs_copyopt(opts, "children.max", &childmax, sizeof(childmax)); 609 if (error == ENOENT) 610 gotchildmax = 0; 611 else if (error != 0) 612 goto done_free; 613 else 614 gotchildmax = 1; 615 616 error = vfs_copyopt(opts, "enforce_statfs", &enforce, sizeof(enforce)); 617 if (error == ENOENT) 618 gotenforce = 0; 619 else if (error != 0) 620 goto done_free; 621 else if (enforce < 0 || enforce > 2) { 622 error = EINVAL; 623 goto done_free; 624 } else 625 gotenforce = 1; 626 627 error = vfs_copyopt(opts, "devfs_ruleset", &rsnum, sizeof(rsnum)); 628 if (error == ENOENT) 629 gotrsnum = 0; 630 else if (error != 0) 631 goto done_free; 632 else 633 gotrsnum = 1; 634 635 pr_flags = ch_flags = 0; 636 for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]); 637 fi++) { 638 if (pr_flag_names[fi] == NULL) 639 continue; 640 vfs_flagopt(opts, pr_flag_names[fi], &pr_flags, 1 << fi); 641 vfs_flagopt(opts, pr_flag_nonames[fi], &ch_flags, 1 << fi); 642 } 643 ch_flags |= pr_flags; 644 for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]); 645 fi++) { 646 error = vfs_copyopt(opts, pr_flag_jailsys[fi].name, &jsys, 647 sizeof(jsys)); 648 if (error == ENOENT) 649 continue; 650 if (error != 0) 651 goto done_free; 652 switch (jsys) { 653 case JAIL_SYS_DISABLE: 654 if (!pr_flag_jailsys[fi].disable) { 655 error = EINVAL; 656 goto done_free; 657 } 658 pr_flags |= pr_flag_jailsys[fi].disable; 659 break; 660 case JAIL_SYS_NEW: 661 pr_flags |= pr_flag_jailsys[fi].new; 662 break; 663 case JAIL_SYS_INHERIT: 664 break; 665 default: 666 error = EINVAL; 667 goto done_free; 668 } 669 ch_flags |= 670 pr_flag_jailsys[fi].new | pr_flag_jailsys[fi].disable; 671 } 672 if ((flags & (JAIL_CREATE | JAIL_UPDATE | JAIL_ATTACH)) == JAIL_CREATE 673 && !(pr_flags & PR_PERSIST)) { 674 error = EINVAL; 675 vfs_opterror(opts, "new jail must persist or attach"); 676 goto done_errmsg; 677 } 678#ifdef VIMAGE 679 if ((flags & JAIL_UPDATE) && (ch_flags & PR_VNET)) { 680 error = EINVAL; 681 vfs_opterror(opts, "vnet cannot be changed after creation"); 682 goto done_errmsg; 683 } 684#endif 685#ifdef INET 686 if ((flags & JAIL_UPDATE) && (ch_flags & PR_IP4_USER)) { 687 error = EINVAL; 688 vfs_opterror(opts, "ip4 cannot be changed after creation"); 689 goto done_errmsg; 690 } 691#endif 692#ifdef INET6 693 if ((flags & JAIL_UPDATE) && (ch_flags & PR_IP6_USER)) { 694 error = EINVAL; 695 vfs_opterror(opts, "ip6 cannot be changed after creation"); 696 goto done_errmsg; 697 } 698#endif 699 700 pr_allow = ch_allow = 0; 701 for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]); 702 fi++) { 703 vfs_flagopt(opts, pr_allow_names[fi], &pr_allow, 1 << fi); 704 vfs_flagopt(opts, pr_allow_nonames[fi], &ch_allow, 1 << fi); 705 } 706 ch_allow |= pr_allow; 707 708 error = vfs_getopt(opts, "name", (void **)&name, &len); 709 if (error == ENOENT) 710 name = NULL; 711 else if (error != 0) 712 goto done_free; 713 else { 714 if (len == 0 || name[len - 1] != '\0') { 715 error = EINVAL; 716 goto done_free; 717 } 718 if (len > MAXHOSTNAMELEN) { 719 error = ENAMETOOLONG; 720 goto done_free; 721 } 722 } 723 724 error = vfs_getopt(opts, "host.hostname", (void **)&host, &len); 725 if (error == ENOENT) 726 host = NULL; 727 else if (error != 0) 728 goto done_free; 729 else { 730 ch_flags |= PR_HOST; 731 pr_flags |= PR_HOST; 732 if (len == 0 || host[len - 1] != '\0') { 733 error = EINVAL; 734 goto done_free; 735 } 736 if (len > MAXHOSTNAMELEN) { 737 error = ENAMETOOLONG; 738 goto done_free; 739 } 740 } 741 742 error = vfs_getopt(opts, "host.domainname", (void **)&domain, &len); 743 if (error == ENOENT) 744 domain = NULL; 745 else if (error != 0) 746 goto done_free; 747 else { 748 ch_flags |= PR_HOST; 749 pr_flags |= PR_HOST; 750 if (len == 0 || domain[len - 1] != '\0') { 751 error = EINVAL; 752 goto done_free; 753 } 754 if (len > MAXHOSTNAMELEN) { 755 error = ENAMETOOLONG; 756 goto done_free; 757 } 758 } 759 760 error = vfs_getopt(opts, "host.hostuuid", (void **)&uuid, &len); 761 if (error == ENOENT) 762 uuid = NULL; 763 else if (error != 0) 764 goto done_free; 765 else { 766 ch_flags |= PR_HOST; 767 pr_flags |= PR_HOST; 768 if (len == 0 || uuid[len - 1] != '\0') { 769 error = EINVAL; 770 goto done_free; 771 } 772 if (len > HOSTUUIDLEN) { 773 error = ENAMETOOLONG; 774 goto done_free; 775 } 776 } 777 778#ifdef COMPAT_FREEBSD32 779 if (SV_PROC_FLAG(td->td_proc, SV_ILP32)) { 780 uint32_t hid32; 781 782 error = vfs_copyopt(opts, "host.hostid", &hid32, sizeof(hid32)); 783 hid = hid32; 784 } else 785#endif 786 error = vfs_copyopt(opts, "host.hostid", &hid, sizeof(hid)); 787 if (error == ENOENT) 788 gothid = 0; 789 else if (error != 0) 790 goto done_free; 791 else { 792 gothid = 1; 793 ch_flags |= PR_HOST; 794 pr_flags |= PR_HOST; 795 } 796 797#ifdef INET 798 error = vfs_getopt(opts, "ip4.addr", &op, &ip4s); 799 if (error == ENOENT) 800 ip4s = (pr_flags & PR_IP4_DISABLE) ? 0 : -1; 801 else if (error != 0) 802 goto done_free; 803 else if (ip4s & (sizeof(*ip4) - 1)) { 804 error = EINVAL; 805 goto done_free; 806 } else { 807 ch_flags |= PR_IP4_USER | PR_IP4_DISABLE; 808 if (ip4s == 0) 809 pr_flags |= PR_IP4_USER | PR_IP4_DISABLE; 810 else { 811 pr_flags = (pr_flags & ~PR_IP4_DISABLE) | PR_IP4_USER; 812 ip4s /= sizeof(*ip4); 813 if (ip4s > jail_max_af_ips) { 814 error = EINVAL; 815 vfs_opterror(opts, "too many IPv4 addresses"); 816 goto done_errmsg; 817 } 818 ip4 = malloc(ip4s * sizeof(*ip4), M_PRISON, M_WAITOK); 819 bcopy(op, ip4, ip4s * sizeof(*ip4)); 820 /* 821 * IP addresses are all sorted but ip[0] to preserve 822 * the primary IP address as given from userland. 823 * This special IP is used for unbound outgoing 824 * connections as well for "loopback" traffic in case 825 * source address selection cannot find any more fitting 826 * address to connect from. 827 */ 828 if (ip4s > 1) 829 qsort(ip4 + 1, ip4s - 1, sizeof(*ip4), qcmp_v4); 830 /* 831 * Check for duplicate addresses and do some simple 832 * zero and broadcast checks. If users give other bogus 833 * addresses it is their problem. 834 * 835 * We do not have to care about byte order for these 836 * checks so we will do them in NBO. 837 */ 838 for (ii = 0; ii < ip4s; ii++) { 839 if (ip4[ii].s_addr == INADDR_ANY || 840 ip4[ii].s_addr == INADDR_BROADCAST) { 841 error = EINVAL; 842 goto done_free; 843 } 844 if ((ii+1) < ip4s && 845 (ip4[0].s_addr == ip4[ii+1].s_addr || 846 ip4[ii].s_addr == ip4[ii+1].s_addr)) { 847 error = EINVAL; 848 goto done_free; 849 } 850 } 851 } 852 } 853#endif 854 855#ifdef INET6 856 error = vfs_getopt(opts, "ip6.addr", &op, &ip6s); 857 if (error == ENOENT) 858 ip6s = (pr_flags & PR_IP6_DISABLE) ? 0 : -1; 859 else if (error != 0) 860 goto done_free; 861 else if (ip6s & (sizeof(*ip6) - 1)) { 862 error = EINVAL; 863 goto done_free; 864 } else { 865 ch_flags |= PR_IP6_USER | PR_IP6_DISABLE; 866 if (ip6s == 0) 867 pr_flags |= PR_IP6_USER | PR_IP6_DISABLE; 868 else { 869 pr_flags = (pr_flags & ~PR_IP6_DISABLE) | PR_IP6_USER; 870 ip6s /= sizeof(*ip6); 871 if (ip6s > jail_max_af_ips) { 872 error = EINVAL; 873 vfs_opterror(opts, "too many IPv6 addresses"); 874 goto done_errmsg; 875 } 876 ip6 = malloc(ip6s * sizeof(*ip6), M_PRISON, M_WAITOK); 877 bcopy(op, ip6, ip6s * sizeof(*ip6)); 878 if (ip6s > 1) 879 qsort(ip6 + 1, ip6s - 1, sizeof(*ip6), qcmp_v6); 880 for (ii = 0; ii < ip6s; ii++) { 881 if (IN6_IS_ADDR_UNSPECIFIED(&ip6[ii])) { 882 error = EINVAL; 883 goto done_free; 884 } 885 if ((ii+1) < ip6s && 886 (IN6_ARE_ADDR_EQUAL(&ip6[0], &ip6[ii+1]) || 887 IN6_ARE_ADDR_EQUAL(&ip6[ii], &ip6[ii+1]))) 888 { 889 error = EINVAL; 890 goto done_free; 891 } 892 } 893 } 894 } 895#endif 896 897#if defined(VIMAGE) && (defined(INET) || defined(INET6)) 898 if ((ch_flags & PR_VNET) && (ch_flags & (PR_IP4_USER | PR_IP6_USER))) { 899 error = EINVAL; 900 vfs_opterror(opts, 901 "vnet jails cannot have IP address restrictions"); 902 goto done_errmsg; 903 } 904#endif 905 906 fullpath_disabled = 0; 907 root = NULL; 908 error = vfs_getopt(opts, "path", (void **)&path, &len); 909 if (error == ENOENT) 910 path = NULL; 911 else if (error != 0) 912 goto done_free; 913 else { 914 if (flags & JAIL_UPDATE) { 915 error = EINVAL; 916 vfs_opterror(opts, 917 "path cannot be changed after creation"); 918 goto done_errmsg; 919 } 920 if (len == 0 || path[len - 1] != '\0') { 921 error = EINVAL; 922 goto done_free; 923 } 924 NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF | MPSAFE, UIO_SYSSPACE, 925 path, td); 926 error = namei(&nd); 927 if (error) 928 goto done_free; 929 vfslocked = NDHASGIANT(&nd); 930 root = nd.ni_vp; 931 NDFREE(&nd, NDF_ONLY_PNBUF); 932 g_path = malloc(MAXPATHLEN, M_TEMP, M_WAITOK); 933 strlcpy(g_path, path, MAXPATHLEN); 934 error = vn_path_to_global_path(td, root, g_path, MAXPATHLEN); 935 if (error == 0) 936 path = g_path; 937 else if (error == ENODEV) { 938 /* proceed if sysctl debug.disablefullpath == 1 */ 939 fullpath_disabled = 1; 940 if (len < 2 || (len == 2 && path[0] == '/')) 941 path = NULL; 942 } else { 943 /* exit on other errors */ 944 VFS_UNLOCK_GIANT(vfslocked); 945 goto done_free; 946 } 947 if (root->v_type != VDIR) { 948 error = ENOTDIR; 949 vput(root); 950 VFS_UNLOCK_GIANT(vfslocked); 951 goto done_free; 952 } 953 VOP_UNLOCK(root, 0); 954 VFS_UNLOCK_GIANT(vfslocked); 955 if (fullpath_disabled) { 956 /* Leave room for a real-root full pathname. */ 957 if (len + (path[0] == '/' && strcmp(mypr->pr_path, "/") 958 ? strlen(mypr->pr_path) : 0) > MAXPATHLEN) { 959 error = ENAMETOOLONG; 960 goto done_free; 961 } 962 } 963 } 964 965 /* 966 * Grab the allprison lock before letting modules check their 967 * parameters. Once we have it, do not let go so we'll have a 968 * consistent view of the OSD list. 969 */ 970 sx_xlock(&allprison_lock); 971 error = osd_jail_call(NULL, PR_METHOD_CHECK, opts); 972 if (error) 973 goto done_unlock_list; 974 975 /* By now, all parameters should have been noted. */ 976 TAILQ_FOREACH(opt, opts, link) { 977 if (!opt->seen && strcmp(opt->name, "errmsg")) { 978 error = EINVAL; 979 vfs_opterror(opts, "unknown parameter: %s", opt->name); 980 goto done_unlock_list; 981 } 982 } 983 984 /* 985 * See if we are creating a new record or updating an existing one. 986 * This abuses the file error codes ENOENT and EEXIST. 987 */ 988 cuflags = flags & (JAIL_CREATE | JAIL_UPDATE); 989 if (!cuflags) { 990 error = EINVAL; 991 vfs_opterror(opts, "no valid operation (create or update)"); 992 goto done_unlock_list; 993 } 994 pr = NULL; 995 namelc = NULL; 996 if (cuflags == JAIL_CREATE && jid == 0 && name != NULL) { 997 namelc = strrchr(name, '.'); 998 jid = strtoul(namelc != NULL ? namelc + 1 : name, &p, 10); 999 if (*p != '\0') 1000 jid = 0; 1001 } 1002 if (jid != 0) { 1003 /* 1004 * See if a requested jid already exists. There is an 1005 * information leak here if the jid exists but is not within 1006 * the caller's jail hierarchy. Jail creators will get EEXIST 1007 * even though they cannot see the jail, and CREATE | UPDATE 1008 * will return ENOENT which is not normally a valid error. 1009 */ 1010 if (jid < 0) { 1011 error = EINVAL; 1012 vfs_opterror(opts, "negative jid"); 1013 goto done_unlock_list; 1014 } 1015 pr = prison_find(jid); 1016 if (pr != NULL) { 1017 ppr = pr->pr_parent; 1018 /* Create: jid must not exist. */ 1019 if (cuflags == JAIL_CREATE) { 1020 mtx_unlock(&pr->pr_mtx); 1021 error = EEXIST; 1022 vfs_opterror(opts, "jail %d already exists", 1023 jid); 1024 goto done_unlock_list; 1025 } 1026 if (!prison_ischild(mypr, pr)) { 1027 mtx_unlock(&pr->pr_mtx); 1028 pr = NULL; 1029 } else if (pr->pr_uref == 0) { 1030 if (!(flags & JAIL_DYING)) { 1031 mtx_unlock(&pr->pr_mtx); 1032 error = ENOENT; 1033 vfs_opterror(opts, "jail %d is dying", 1034 jid); 1035 goto done_unlock_list; 1036 } else if ((flags & JAIL_ATTACH) || 1037 (pr_flags & PR_PERSIST)) { 1038 /* 1039 * A dying jail might be resurrected 1040 * (via attach or persist), but first 1041 * it must determine if another jail 1042 * has claimed its name. Accomplish 1043 * this by implicitly re-setting the 1044 * name. 1045 */ 1046 if (name == NULL) 1047 name = prison_name(mypr, pr); 1048 } 1049 } 1050 } 1051 if (pr == NULL) { 1052 /* Update: jid must exist. */ 1053 if (cuflags == JAIL_UPDATE) { 1054 error = ENOENT; 1055 vfs_opterror(opts, "jail %d not found", jid); 1056 goto done_unlock_list; 1057 } 1058 } 1059 } 1060 /* 1061 * If the caller provided a name, look for a jail by that name. 1062 * This has different semantics for creates and updates keyed by jid 1063 * (where the name must not already exist in a different jail), 1064 * and updates keyed by the name itself (where the name must exist 1065 * because that is the jail being updated). 1066 */ 1067 if (name != NULL) { 1068 namelc = strrchr(name, '.'); 1069 if (namelc == NULL) 1070 namelc = name; 1071 else { 1072 /* 1073 * This is a hierarchical name. Split it into the 1074 * parent and child names, and make sure the parent 1075 * exists or matches an already found jail. 1076 */ 1077 *namelc = '\0'; 1078 if (pr != NULL) { 1079 if (strncmp(name, ppr->pr_name, namelc - name) 1080 || ppr->pr_name[namelc - name] != '\0') { 1081 mtx_unlock(&pr->pr_mtx); 1082 error = EINVAL; 1083 vfs_opterror(opts, 1084 "cannot change jail's parent"); 1085 goto done_unlock_list; 1086 } 1087 } else { 1088 ppr = prison_find_name(mypr, name); 1089 if (ppr == NULL) { 1090 error = ENOENT; 1091 vfs_opterror(opts, 1092 "jail \"%s\" not found", name); 1093 goto done_unlock_list; 1094 } 1095 mtx_unlock(&ppr->pr_mtx); 1096 } 1097 name = ++namelc; 1098 } 1099 if (name[0] != '\0') { 1100 namelen = 1101 (ppr == &prison0) ? 0 : strlen(ppr->pr_name) + 1; 1102 name_again: 1103 deadpr = NULL; 1104 FOREACH_PRISON_CHILD(ppr, tpr) { 1105 if (tpr != pr && tpr->pr_ref > 0 && 1106 !strcmp(tpr->pr_name + namelen, name)) { 1107 if (pr == NULL && 1108 cuflags != JAIL_CREATE) { 1109 mtx_lock(&tpr->pr_mtx); 1110 if (tpr->pr_ref > 0) { 1111 /* 1112 * Use this jail 1113 * for updates. 1114 */ 1115 if (tpr->pr_uref > 0) { 1116 pr = tpr; 1117 break; 1118 } 1119 deadpr = tpr; 1120 } 1121 mtx_unlock(&tpr->pr_mtx); 1122 } else if (tpr->pr_uref > 0) { 1123 /* 1124 * Create, or update(jid): 1125 * name must not exist in an 1126 * active sibling jail. 1127 */ 1128 error = EEXIST; 1129 if (pr != NULL) 1130 mtx_unlock(&pr->pr_mtx); 1131 vfs_opterror(opts, 1132 "jail \"%s\" already exists", 1133 name); 1134 goto done_unlock_list; 1135 } 1136 } 1137 } 1138 /* If no active jail is found, use a dying one. */ 1139 if (deadpr != NULL && pr == NULL) { 1140 if (flags & JAIL_DYING) { 1141 mtx_lock(&deadpr->pr_mtx); 1142 if (deadpr->pr_ref == 0) { 1143 mtx_unlock(&deadpr->pr_mtx); 1144 goto name_again; 1145 } 1146 pr = deadpr; 1147 } else if (cuflags == JAIL_UPDATE) { 1148 error = ENOENT; 1149 vfs_opterror(opts, 1150 "jail \"%s\" is dying", name); 1151 goto done_unlock_list; 1152 } 1153 } 1154 /* Update: name must exist if no jid. */ 1155 else if (cuflags == JAIL_UPDATE && pr == NULL) { 1156 error = ENOENT; 1157 vfs_opterror(opts, "jail \"%s\" not found", 1158 name); 1159 goto done_unlock_list; 1160 } 1161 } 1162 } 1163 /* Update: must provide a jid or name. */ 1164 else if (cuflags == JAIL_UPDATE && pr == NULL) { 1165 error = ENOENT; 1166 vfs_opterror(opts, "update specified no jail"); 1167 goto done_unlock_list; 1168 } 1169 1170 /* If there's no prison to update, create a new one and link it in. */ 1171 if (pr == NULL) { 1172 for (tpr = mypr; tpr != NULL; tpr = tpr->pr_parent) 1173 if (tpr->pr_childcount >= tpr->pr_childmax) { 1174 error = EPERM; 1175 vfs_opterror(opts, "prison limit exceeded"); 1176 goto done_unlock_list; 1177 } 1178 created = 1; 1179 mtx_lock(&ppr->pr_mtx); 1180 if (ppr->pr_ref == 0 || (ppr->pr_flags & PR_REMOVE)) { 1181 mtx_unlock(&ppr->pr_mtx); 1182 error = ENOENT; 1183 vfs_opterror(opts, "parent jail went away!"); 1184 goto done_unlock_list; 1185 } 1186 ppr->pr_ref++; 1187 ppr->pr_uref++; 1188 mtx_unlock(&ppr->pr_mtx); 1189 pr = malloc(sizeof(*pr), M_PRISON, M_WAITOK | M_ZERO); 1190 if (jid == 0) { 1191 /* Find the next free jid. */ 1192 jid = lastprid + 1; 1193 findnext: 1194 if (jid == JAIL_MAX) 1195 jid = 1; 1196 TAILQ_FOREACH(tpr, &allprison, pr_list) { 1197 if (tpr->pr_id < jid) 1198 continue; 1199 if (tpr->pr_id > jid || tpr->pr_ref == 0) { 1200 TAILQ_INSERT_BEFORE(tpr, pr, pr_list); 1201 break; 1202 } 1203 if (jid == lastprid) { 1204 error = EAGAIN; 1205 vfs_opterror(opts, 1206 "no available jail IDs"); 1207 free(pr, M_PRISON); 1208 prison_deref(ppr, PD_DEREF | 1209 PD_DEUREF | PD_LIST_XLOCKED); 1210 goto done_releroot; 1211 } 1212 jid++; 1213 goto findnext; 1214 } 1215 lastprid = jid; 1216 } else { 1217 /* 1218 * The jail already has a jid (that did not yet exist), 1219 * so just find where to insert it. 1220 */ 1221 TAILQ_FOREACH(tpr, &allprison, pr_list) 1222 if (tpr->pr_id >= jid) { 1223 TAILQ_INSERT_BEFORE(tpr, pr, pr_list); 1224 break; 1225 } 1226 } 1227 if (tpr == NULL) 1228 TAILQ_INSERT_TAIL(&allprison, pr, pr_list); 1229 LIST_INSERT_HEAD(&ppr->pr_children, pr, pr_sibling); 1230 for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent) 1231 tpr->pr_childcount++; 1232 1233 pr->pr_parent = ppr; 1234 pr->pr_id = jid; 1235 1236 /* Set some default values, and inherit some from the parent. */ 1237 if (name == NULL) 1238 name = ""; 1239 if (path == NULL) { 1240 path = "/"; 1241 root = mypr->pr_root; 1242 vref(root); 1243 } 1244 strlcpy(pr->pr_hostuuid, DEFAULT_HOSTUUID, HOSTUUIDLEN); 1245 pr->pr_flags |= PR_HOST; 1246#if defined(INET) || defined(INET6) 1247#ifdef VIMAGE 1248 if (!(pr_flags & PR_VNET)) 1249#endif 1250 { 1251#ifdef INET 1252 if (!(ch_flags & PR_IP4_USER)) 1253 pr->pr_flags |= 1254 PR_IP4 | PR_IP4_USER | PR_IP4_DISABLE; 1255 else if (!(pr_flags & PR_IP4_USER)) { 1256 pr->pr_flags |= ppr->pr_flags & PR_IP4; 1257 if (ppr->pr_ip4 != NULL) { 1258 pr->pr_ip4s = ppr->pr_ip4s; 1259 pr->pr_ip4 = malloc(pr->pr_ip4s * 1260 sizeof(struct in_addr), M_PRISON, 1261 M_WAITOK); 1262 bcopy(ppr->pr_ip4, pr->pr_ip4, 1263 pr->pr_ip4s * sizeof(*pr->pr_ip4)); 1264 } 1265 } 1266#endif 1267#ifdef INET6 1268 if (!(ch_flags & PR_IP6_USER)) 1269 pr->pr_flags |= 1270 PR_IP6 | PR_IP6_USER | PR_IP6_DISABLE; 1271 else if (!(pr_flags & PR_IP6_USER)) { 1272 pr->pr_flags |= ppr->pr_flags & PR_IP6; 1273 if (ppr->pr_ip6 != NULL) { 1274 pr->pr_ip6s = ppr->pr_ip6s; 1275 pr->pr_ip6 = malloc(pr->pr_ip6s * 1276 sizeof(struct in6_addr), M_PRISON, 1277 M_WAITOK); 1278 bcopy(ppr->pr_ip6, pr->pr_ip6, 1279 pr->pr_ip6s * sizeof(*pr->pr_ip6)); 1280 } 1281 } 1282#endif 1283 } 1284#endif 1285 /* Source address selection is always on by default. */ 1286 pr->pr_flags |= _PR_IP_SADDRSEL; 1287 1288 pr->pr_securelevel = ppr->pr_securelevel; 1289 pr->pr_allow = JAIL_DEFAULT_ALLOW & ppr->pr_allow; 1290 pr->pr_enforce_statfs = JAIL_DEFAULT_ENFORCE_STATFS; 1291 pr->pr_devfs_rsnum = ppr->pr_devfs_rsnum; 1292 1293 LIST_INIT(&pr->pr_children); 1294 mtx_init(&pr->pr_mtx, "jail mutex", NULL, MTX_DEF | MTX_DUPOK); 1295 1296#ifdef VIMAGE 1297 /* Allocate a new vnet if specified. */ 1298 pr->pr_vnet = (pr_flags & PR_VNET) 1299 ? vnet_alloc() : ppr->pr_vnet; 1300#endif 1301 /* 1302 * Allocate a dedicated cpuset for each jail. 1303 * Unlike other initial settings, this may return an erorr. 1304 */ 1305 error = cpuset_create_root(ppr, &pr->pr_cpuset); 1306 if (error) { 1307 prison_deref(pr, PD_LIST_XLOCKED); 1308 goto done_releroot; 1309 } 1310 1311 mtx_lock(&pr->pr_mtx); 1312 /* 1313 * New prisons do not yet have a reference, because we do not 1314 * want other to see the incomplete prison once the 1315 * allprison_lock is downgraded. 1316 */ 1317 } else { 1318 created = 0; 1319 /* 1320 * Grab a reference for existing prisons, to ensure they 1321 * continue to exist for the duration of the call. 1322 */ 1323 pr->pr_ref++; 1324#if defined(VIMAGE) && (defined(INET) || defined(INET6)) 1325 if ((pr->pr_flags & PR_VNET) && 1326 (ch_flags & (PR_IP4_USER | PR_IP6_USER))) { 1327 error = EINVAL; 1328 vfs_opterror(opts, 1329 "vnet jails cannot have IP address restrictions"); 1330 goto done_deref_locked; 1331 } 1332#endif 1333#ifdef INET 1334 if (PR_IP4_USER & ch_flags & (pr_flags ^ pr->pr_flags)) { 1335 error = EINVAL; 1336 vfs_opterror(opts, 1337 "ip4 cannot be changed after creation"); 1338 goto done_deref_locked; 1339 } 1340#endif 1341#ifdef INET6 1342 if (PR_IP6_USER & ch_flags & (pr_flags ^ pr->pr_flags)) { 1343 error = EINVAL; 1344 vfs_opterror(opts, 1345 "ip6 cannot be changed after creation"); 1346 goto done_deref_locked; 1347 } 1348#endif 1349 } 1350 1351 /* Do final error checking before setting anything. */ 1352 if (gotslevel) { 1353 if (slevel < ppr->pr_securelevel) { 1354 error = EPERM; 1355 goto done_deref_locked; 1356 } 1357 } 1358 if (gotchildmax) { 1359 if (childmax >= ppr->pr_childmax) { 1360 error = EPERM; 1361 goto done_deref_locked; 1362 } 1363 } 1364 if (gotenforce) { 1365 if (enforce < ppr->pr_enforce_statfs) { 1366 error = EPERM; 1367 goto done_deref_locked; 1368 } 1369 } 1370 if (gotrsnum) { 1371 /* 1372 * devfs_rsnum is a uint16_t 1373 */ 1374 if (rsnum < 0 || rsnum > 65535) { 1375 error = EINVAL; 1376 goto done_deref_locked; 1377 } 1378 /* 1379 * Nested jails always inherit parent's devfs ruleset 1380 */ 1381 if (jailed(td->td_ucred)) { 1382 if (rsnum > 0 && rsnum != ppr->pr_devfs_rsnum) { 1383 error = EPERM; 1384 goto done_deref_locked; 1385 } else 1386 rsnum = ppr->pr_devfs_rsnum; 1387 } 1388 } 1389#ifdef INET 1390 if (ip4s > 0) { 1391 if (ppr->pr_flags & PR_IP4) { 1392 /* 1393 * Make sure the new set of IP addresses is a 1394 * subset of the parent's list. Don't worry 1395 * about the parent being unlocked, as any 1396 * setting is done with allprison_lock held. 1397 */ 1398 for (ij = 0; ij < ppr->pr_ip4s; ij++) 1399 if (ip4[0].s_addr == ppr->pr_ip4[ij].s_addr) 1400 break; 1401 if (ij == ppr->pr_ip4s) { 1402 error = EPERM; 1403 goto done_deref_locked; 1404 } 1405 if (ip4s > 1) { 1406 for (ii = ij = 1; ii < ip4s; ii++) { 1407 if (ip4[ii].s_addr == 1408 ppr->pr_ip4[0].s_addr) 1409 continue; 1410 for (; ij < ppr->pr_ip4s; ij++) 1411 if (ip4[ii].s_addr == 1412 ppr->pr_ip4[ij].s_addr) 1413 break; 1414 if (ij == ppr->pr_ip4s) 1415 break; 1416 } 1417 if (ij == ppr->pr_ip4s) { 1418 error = EPERM; 1419 goto done_deref_locked; 1420 } 1421 } 1422 } 1423 /* 1424 * Check for conflicting IP addresses. We permit them 1425 * if there is no more than one IP on each jail. If 1426 * there is a duplicate on a jail with more than one 1427 * IP stop checking and return error. 1428 */ 1429 tppr = ppr; 1430#ifdef VIMAGE 1431 for (; tppr != &prison0; tppr = tppr->pr_parent) 1432 if (tppr->pr_flags & PR_VNET) 1433 break; 1434#endif 1435 FOREACH_PRISON_DESCENDANT(tppr, tpr, descend) { 1436 if (tpr == pr || 1437#ifdef VIMAGE 1438 (tpr != tppr && (tpr->pr_flags & PR_VNET)) || 1439#endif 1440 tpr->pr_uref == 0) { 1441 descend = 0; 1442 continue; 1443 } 1444 if (!(tpr->pr_flags & PR_IP4_USER)) 1445 continue; 1446 descend = 0; 1447 if (tpr->pr_ip4 == NULL || 1448 (ip4s == 1 && tpr->pr_ip4s == 1)) 1449 continue; 1450 for (ii = 0; ii < ip4s; ii++) { 1451 if (_prison_check_ip4(tpr, &ip4[ii]) == 0) { 1452 error = EADDRINUSE; 1453 vfs_opterror(opts, 1454 "IPv4 addresses clash"); 1455 goto done_deref_locked; 1456 } 1457 } 1458 } 1459 } 1460#endif 1461#ifdef INET6 1462 if (ip6s > 0) { 1463 if (ppr->pr_flags & PR_IP6) { 1464 /* 1465 * Make sure the new set of IP addresses is a 1466 * subset of the parent's list. 1467 */ 1468 for (ij = 0; ij < ppr->pr_ip6s; ij++) 1469 if (IN6_ARE_ADDR_EQUAL(&ip6[0], 1470 &ppr->pr_ip6[ij])) 1471 break; 1472 if (ij == ppr->pr_ip6s) { 1473 error = EPERM; 1474 goto done_deref_locked; 1475 } 1476 if (ip6s > 1) { 1477 for (ii = ij = 1; ii < ip6s; ii++) { 1478 if (IN6_ARE_ADDR_EQUAL(&ip6[ii], 1479 &ppr->pr_ip6[0])) 1480 continue; 1481 for (; ij < ppr->pr_ip6s; ij++) 1482 if (IN6_ARE_ADDR_EQUAL( 1483 &ip6[ii], &ppr->pr_ip6[ij])) 1484 break; 1485 if (ij == ppr->pr_ip6s) 1486 break; 1487 } 1488 if (ij == ppr->pr_ip6s) { 1489 error = EPERM; 1490 goto done_deref_locked; 1491 } 1492 } 1493 } 1494 /* Check for conflicting IP addresses. */ 1495 tppr = ppr; 1496#ifdef VIMAGE 1497 for (; tppr != &prison0; tppr = tppr->pr_parent) 1498 if (tppr->pr_flags & PR_VNET) 1499 break; 1500#endif 1501 FOREACH_PRISON_DESCENDANT(tppr, tpr, descend) { 1502 if (tpr == pr || 1503#ifdef VIMAGE 1504 (tpr != tppr && (tpr->pr_flags & PR_VNET)) || 1505#endif 1506 tpr->pr_uref == 0) { 1507 descend = 0; 1508 continue; 1509 } 1510 if (!(tpr->pr_flags & PR_IP6_USER)) 1511 continue; 1512 descend = 0; 1513 if (tpr->pr_ip6 == NULL || 1514 (ip6s == 1 && tpr->pr_ip6s == 1)) 1515 continue; 1516 for (ii = 0; ii < ip6s; ii++) { 1517 if (_prison_check_ip6(tpr, &ip6[ii]) == 0) { 1518 error = EADDRINUSE; 1519 vfs_opterror(opts, 1520 "IPv6 addresses clash"); 1521 goto done_deref_locked; 1522 } 1523 } 1524 } 1525 } 1526#endif 1527 onamelen = namelen = 0; 1528 if (name != NULL) { 1529 /* Give a default name of the jid. */ 1530 if (name[0] == '\0') 1531 snprintf(name = numbuf, sizeof(numbuf), "%d", jid); 1532 else if (*namelc == '0' || (strtoul(namelc, &p, 10) != jid && 1533 *p == '\0')) { 1534 error = EINVAL; 1535 vfs_opterror(opts, 1536 "name cannot be numeric (unless it is the jid)"); 1537 goto done_deref_locked; 1538 } 1539 /* 1540 * Make sure the name isn't too long for the prison or its 1541 * children. 1542 */ 1543 onamelen = strlen(pr->pr_name); 1544 namelen = strlen(name); 1545 if (strlen(ppr->pr_name) + namelen + 2 > sizeof(pr->pr_name)) { 1546 error = ENAMETOOLONG; 1547 goto done_deref_locked; 1548 } 1549 FOREACH_PRISON_DESCENDANT(pr, tpr, descend) { 1550 if (strlen(tpr->pr_name) + (namelen - onamelen) >= 1551 sizeof(pr->pr_name)) { 1552 error = ENAMETOOLONG; 1553 goto done_deref_locked; 1554 } 1555 } 1556 } 1557 if (pr_allow & ~ppr->pr_allow) { 1558 error = EPERM; 1559 goto done_deref_locked; 1560 } 1561 1562 /* Set the parameters of the prison. */ 1563#ifdef INET 1564 redo_ip4 = 0; 1565 if (pr_flags & PR_IP4_USER) { 1566 pr->pr_flags |= PR_IP4; 1567 free(pr->pr_ip4, M_PRISON); 1568 pr->pr_ip4s = ip4s; 1569 pr->pr_ip4 = ip4; 1570 ip4 = NULL; 1571 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) { 1572#ifdef VIMAGE 1573 if (tpr->pr_flags & PR_VNET) { 1574 descend = 0; 1575 continue; 1576 } 1577#endif 1578 if (prison_restrict_ip4(tpr, NULL)) { 1579 redo_ip4 = 1; 1580 descend = 0; 1581 } 1582 } 1583 } 1584#endif 1585#ifdef INET6 1586 redo_ip6 = 0; 1587 if (pr_flags & PR_IP6_USER) { 1588 pr->pr_flags |= PR_IP6; 1589 free(pr->pr_ip6, M_PRISON); 1590 pr->pr_ip6s = ip6s; 1591 pr->pr_ip6 = ip6; 1592 ip6 = NULL; 1593 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) { 1594#ifdef VIMAGE 1595 if (tpr->pr_flags & PR_VNET) { 1596 descend = 0; 1597 continue; 1598 } 1599#endif 1600 if (prison_restrict_ip6(tpr, NULL)) { 1601 redo_ip6 = 1; 1602 descend = 0; 1603 } 1604 } 1605 } 1606#endif 1607 if (gotslevel) { 1608 pr->pr_securelevel = slevel; 1609 /* Set all child jails to be at least this level. */ 1610 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) 1611 if (tpr->pr_securelevel < slevel) 1612 tpr->pr_securelevel = slevel; 1613 } 1614 if (gotchildmax) { 1615 pr->pr_childmax = childmax; 1616 /* Set all child jails to under this limit. */ 1617 FOREACH_PRISON_DESCENDANT_LOCKED_LEVEL(pr, tpr, descend, level) 1618 if (tpr->pr_childmax > childmax - level) 1619 tpr->pr_childmax = childmax > level 1620 ? childmax - level : 0; 1621 } 1622 if (gotenforce) { 1623 pr->pr_enforce_statfs = enforce; 1624 /* Pass this restriction on to the children. */ 1625 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) 1626 if (tpr->pr_enforce_statfs < enforce) 1627 tpr->pr_enforce_statfs = enforce; 1628 } 1629 if (gotrsnum) { 1630 pr->pr_devfs_rsnum = rsnum; 1631 /* Pass this restriction on to the children. */ 1632 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) 1633 tpr->pr_devfs_rsnum = rsnum; 1634 } 1635 if (name != NULL) { 1636 if (ppr == &prison0) 1637 strlcpy(pr->pr_name, name, sizeof(pr->pr_name)); 1638 else 1639 snprintf(pr->pr_name, sizeof(pr->pr_name), "%s.%s", 1640 ppr->pr_name, name); 1641 /* Change this component of child names. */ 1642 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) { 1643 bcopy(tpr->pr_name + onamelen, tpr->pr_name + namelen, 1644 strlen(tpr->pr_name + onamelen) + 1); 1645 bcopy(pr->pr_name, tpr->pr_name, namelen); 1646 } 1647 } 1648 if (path != NULL) { 1649 /* Try to keep a real-rooted full pathname. */ 1650 if (fullpath_disabled && path[0] == '/' && 1651 strcmp(mypr->pr_path, "/")) 1652 snprintf(pr->pr_path, sizeof(pr->pr_path), "%s%s", 1653 mypr->pr_path, path); 1654 else 1655 strlcpy(pr->pr_path, path, sizeof(pr->pr_path)); 1656 pr->pr_root = root; 1657 } 1658 if (PR_HOST & ch_flags & ~pr_flags) { 1659 if (pr->pr_flags & PR_HOST) { 1660 /* 1661 * Copy the parent's host info. As with pr_ip4 above, 1662 * the lack of a lock on the parent is not a problem; 1663 * it is always set with allprison_lock at least 1664 * shared, and is held exclusively here. 1665 */ 1666 strlcpy(pr->pr_hostname, pr->pr_parent->pr_hostname, 1667 sizeof(pr->pr_hostname)); 1668 strlcpy(pr->pr_domainname, pr->pr_parent->pr_domainname, 1669 sizeof(pr->pr_domainname)); 1670 strlcpy(pr->pr_hostuuid, pr->pr_parent->pr_hostuuid, 1671 sizeof(pr->pr_hostuuid)); 1672 pr->pr_hostid = pr->pr_parent->pr_hostid; 1673 } 1674 } else if (host != NULL || domain != NULL || uuid != NULL || gothid) { 1675 /* Set this prison, and any descendants without PR_HOST. */ 1676 if (host != NULL) 1677 strlcpy(pr->pr_hostname, host, sizeof(pr->pr_hostname)); 1678 if (domain != NULL) 1679 strlcpy(pr->pr_domainname, domain, 1680 sizeof(pr->pr_domainname)); 1681 if (uuid != NULL) 1682 strlcpy(pr->pr_hostuuid, uuid, sizeof(pr->pr_hostuuid)); 1683 if (gothid) 1684 pr->pr_hostid = hid; 1685 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) { 1686 if (tpr->pr_flags & PR_HOST) 1687 descend = 0; 1688 else { 1689 if (host != NULL) 1690 strlcpy(tpr->pr_hostname, 1691 pr->pr_hostname, 1692 sizeof(tpr->pr_hostname)); 1693 if (domain != NULL) 1694 strlcpy(tpr->pr_domainname, 1695 pr->pr_domainname, 1696 sizeof(tpr->pr_domainname)); 1697 if (uuid != NULL) 1698 strlcpy(tpr->pr_hostuuid, 1699 pr->pr_hostuuid, 1700 sizeof(tpr->pr_hostuuid)); 1701 if (gothid) 1702 tpr->pr_hostid = hid; 1703 } 1704 } 1705 } 1706 if ((tallow = ch_allow & ~pr_allow)) { 1707 /* Clear allow bits in all children. */ 1708 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) 1709 tpr->pr_allow &= ~tallow; 1710 } 1711 pr->pr_allow = (pr->pr_allow & ~ch_allow) | pr_allow; 1712 /* 1713 * Persistent prisons get an extra reference, and prisons losing their 1714 * persist flag lose that reference. Only do this for existing prisons 1715 * for now, so new ones will remain unseen until after the module 1716 * handlers have completed. 1717 */ 1718 if (!created && (ch_flags & PR_PERSIST & (pr_flags ^ pr->pr_flags))) { 1719 if (pr_flags & PR_PERSIST) { 1720 pr->pr_ref++; 1721 pr->pr_uref++; 1722 } else { 1723 pr->pr_ref--; 1724 pr->pr_uref--; 1725 } 1726 } 1727 pr->pr_flags = (pr->pr_flags & ~ch_flags) | pr_flags; 1728 mtx_unlock(&pr->pr_mtx); 1729 1730#ifdef RACCT 1731 if (created) 1732 prison_racct_attach(pr); 1733#endif 1734 1735 /* Locks may have prevented a complete restriction of child IP 1736 * addresses. If so, allocate some more memory and try again. 1737 */ 1738#ifdef INET 1739 while (redo_ip4) { 1740 ip4s = pr->pr_ip4s; 1741 ip4 = malloc(ip4s * sizeof(*ip4), M_PRISON, M_WAITOK); 1742 mtx_lock(&pr->pr_mtx); 1743 redo_ip4 = 0; 1744 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) { 1745#ifdef VIMAGE 1746 if (tpr->pr_flags & PR_VNET) { 1747 descend = 0; 1748 continue; 1749 } 1750#endif 1751 if (prison_restrict_ip4(tpr, ip4)) { 1752 if (ip4 != NULL) 1753 ip4 = NULL; 1754 else 1755 redo_ip4 = 1; 1756 } 1757 } 1758 mtx_unlock(&pr->pr_mtx); 1759 } 1760#endif 1761#ifdef INET6 1762 while (redo_ip6) { 1763 ip6s = pr->pr_ip6s; 1764 ip6 = malloc(ip6s * sizeof(*ip6), M_PRISON, M_WAITOK); 1765 mtx_lock(&pr->pr_mtx); 1766 redo_ip6 = 0; 1767 FOREACH_PRISON_DESCENDANT_LOCKED(pr, tpr, descend) { 1768#ifdef VIMAGE 1769 if (tpr->pr_flags & PR_VNET) { 1770 descend = 0; 1771 continue; 1772 } 1773#endif 1774 if (prison_restrict_ip6(tpr, ip6)) { 1775 if (ip6 != NULL) 1776 ip6 = NULL; 1777 else 1778 redo_ip6 = 1; 1779 } 1780 } 1781 mtx_unlock(&pr->pr_mtx); 1782 } 1783#endif 1784 1785 /* Let the modules do their work. */ 1786 sx_downgrade(&allprison_lock); 1787 if (created) { 1788 error = osd_jail_call(pr, PR_METHOD_CREATE, opts); 1789 if (error) { 1790 prison_deref(pr, PD_LIST_SLOCKED); 1791 goto done_errmsg; 1792 } 1793 } 1794 error = osd_jail_call(pr, PR_METHOD_SET, opts); 1795 if (error) { 1796 prison_deref(pr, created 1797 ? PD_LIST_SLOCKED 1798 : PD_DEREF | PD_LIST_SLOCKED); 1799 goto done_errmsg; 1800 } 1801 1802 /* Attach this process to the prison if requested. */ 1803 if (flags & JAIL_ATTACH) { 1804 mtx_lock(&pr->pr_mtx); 1805 error = do_jail_attach(td, pr); 1806 if (error) { 1807 vfs_opterror(opts, "attach failed"); 1808 if (!created) 1809 prison_deref(pr, PD_DEREF); 1810 goto done_errmsg; 1811 } 1812 } 1813 1814 /* 1815 * Now that it is all there, drop the temporary reference from existing 1816 * prisons. Or add a reference to newly created persistent prisons 1817 * (which was not done earlier so that the prison would not be publicly 1818 * visible). 1819 */ 1820 if (!created) { 1821 prison_deref(pr, (flags & JAIL_ATTACH) 1822 ? PD_DEREF 1823 : PD_DEREF | PD_LIST_SLOCKED); 1824 } else { 1825 if (pr_flags & PR_PERSIST) { 1826 mtx_lock(&pr->pr_mtx); 1827 pr->pr_ref++; 1828 pr->pr_uref++; 1829 mtx_unlock(&pr->pr_mtx); 1830 } 1831 if (!(flags & JAIL_ATTACH)) 1832 sx_sunlock(&allprison_lock); 1833 } 1834 1835#ifdef RACCT 1836 if (!created) 1837 prison_racct_modify(pr); 1838#endif 1839 1840 td->td_retval[0] = pr->pr_id; 1841 goto done_errmsg; 1842 1843 done_deref_locked: 1844 prison_deref(pr, created 1845 ? PD_LOCKED | PD_LIST_XLOCKED 1846 : PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED); 1847 goto done_releroot; 1848 done_unlock_list: 1849 sx_xunlock(&allprison_lock); 1850 done_releroot: 1851 if (root != NULL) { 1852 vfslocked = VFS_LOCK_GIANT(root->v_mount); 1853 vrele(root); 1854 VFS_UNLOCK_GIANT(vfslocked); 1855 } 1856 done_errmsg: 1857 if (error) { 1858 vfs_getopt(opts, "errmsg", (void **)&errmsg, &errmsg_len); 1859 if (errmsg_len > 0) { 1860 errmsg_pos = 2 * vfs_getopt_pos(opts, "errmsg") + 1; 1861 if (errmsg_pos > 0) { 1862 if (optuio->uio_segflg == UIO_SYSSPACE) 1863 bcopy(errmsg, 1864 optuio->uio_iov[errmsg_pos].iov_base, 1865 errmsg_len); 1866 else 1867 copyout(errmsg, 1868 optuio->uio_iov[errmsg_pos].iov_base, 1869 errmsg_len); 1870 } 1871 } 1872 } 1873 done_free: 1874#ifdef INET 1875 free(ip4, M_PRISON); 1876#endif 1877#ifdef INET6 1878 free(ip6, M_PRISON); 1879#endif 1880 if (g_path != NULL) 1881 free(g_path, M_TEMP); 1882 vfs_freeopts(opts); 1883 return (error); 1884} 1885 1886 1887/* 1888 * struct jail_get_args { 1889 * struct iovec *iovp; 1890 * unsigned int iovcnt; 1891 * int flags; 1892 * }; 1893 */ 1894int 1895sys_jail_get(struct thread *td, struct jail_get_args *uap) 1896{ 1897 struct uio *auio; 1898 int error; 1899 1900 /* Check that we have an even number of iovecs. */ 1901 if (uap->iovcnt & 1) 1902 return (EINVAL); 1903 1904 error = copyinuio(uap->iovp, uap->iovcnt, &auio); 1905 if (error) 1906 return (error); 1907 error = kern_jail_get(td, auio, uap->flags); 1908 if (error == 0) 1909 error = copyout(auio->uio_iov, uap->iovp, 1910 uap->iovcnt * sizeof (struct iovec)); 1911 free(auio, M_IOV); 1912 return (error); 1913} 1914 1915int 1916kern_jail_get(struct thread *td, struct uio *optuio, int flags) 1917{ 1918 struct prison *pr, *mypr; 1919 struct vfsopt *opt; 1920 struct vfsoptlist *opts; 1921 char *errmsg, *name; 1922 int error, errmsg_len, errmsg_pos, fi, i, jid, len, locked, pos; 1923 1924 if (flags & ~JAIL_GET_MASK) 1925 return (EINVAL); 1926 1927 /* Get the parameter list. */ 1928 error = vfs_buildopts(optuio, &opts); 1929 if (error) 1930 return (error); 1931 errmsg_pos = vfs_getopt_pos(opts, "errmsg"); 1932 mypr = td->td_ucred->cr_prison; 1933 1934 /* 1935 * Find the prison specified by one of: lastjid, jid, name. 1936 */ 1937 sx_slock(&allprison_lock); 1938 error = vfs_copyopt(opts, "lastjid", &jid, sizeof(jid)); 1939 if (error == 0) { 1940 TAILQ_FOREACH(pr, &allprison, pr_list) { 1941 if (pr->pr_id > jid && prison_ischild(mypr, pr)) { 1942 mtx_lock(&pr->pr_mtx); 1943 if (pr->pr_ref > 0 && 1944 (pr->pr_uref > 0 || (flags & JAIL_DYING))) 1945 break; 1946 mtx_unlock(&pr->pr_mtx); 1947 } 1948 } 1949 if (pr != NULL) 1950 goto found_prison; 1951 error = ENOENT; 1952 vfs_opterror(opts, "no jail after %d", jid); 1953 goto done_unlock_list; 1954 } else if (error != ENOENT) 1955 goto done_unlock_list; 1956 1957 error = vfs_copyopt(opts, "jid", &jid, sizeof(jid)); 1958 if (error == 0) { 1959 if (jid != 0) { 1960 pr = prison_find_child(mypr, jid); 1961 if (pr != NULL) { 1962 if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) { 1963 mtx_unlock(&pr->pr_mtx); 1964 error = ENOENT; 1965 vfs_opterror(opts, "jail %d is dying", 1966 jid); 1967 goto done_unlock_list; 1968 } 1969 goto found_prison; 1970 } 1971 error = ENOENT; 1972 vfs_opterror(opts, "jail %d not found", jid); 1973 goto done_unlock_list; 1974 } 1975 } else if (error != ENOENT) 1976 goto done_unlock_list; 1977 1978 error = vfs_getopt(opts, "name", (void **)&name, &len); 1979 if (error == 0) { 1980 if (len == 0 || name[len - 1] != '\0') { 1981 error = EINVAL; 1982 goto done_unlock_list; 1983 } 1984 pr = prison_find_name(mypr, name); 1985 if (pr != NULL) { 1986 if (pr->pr_uref == 0 && !(flags & JAIL_DYING)) { 1987 mtx_unlock(&pr->pr_mtx); 1988 error = ENOENT; 1989 vfs_opterror(opts, "jail \"%s\" is dying", 1990 name); 1991 goto done_unlock_list; 1992 } 1993 goto found_prison; 1994 } 1995 error = ENOENT; 1996 vfs_opterror(opts, "jail \"%s\" not found", name); 1997 goto done_unlock_list; 1998 } else if (error != ENOENT) 1999 goto done_unlock_list; 2000 2001 vfs_opterror(opts, "no jail specified"); 2002 error = ENOENT; 2003 goto done_unlock_list; 2004 2005 found_prison: 2006 /* Get the parameters of the prison. */ 2007 pr->pr_ref++; 2008 locked = PD_LOCKED; 2009 td->td_retval[0] = pr->pr_id; 2010 error = vfs_setopt(opts, "jid", &pr->pr_id, sizeof(pr->pr_id)); 2011 if (error != 0 && error != ENOENT) 2012 goto done_deref; 2013 i = (pr->pr_parent == mypr) ? 0 : pr->pr_parent->pr_id; 2014 error = vfs_setopt(opts, "parent", &i, sizeof(i)); 2015 if (error != 0 && error != ENOENT) 2016 goto done_deref; 2017 error = vfs_setopts(opts, "name", prison_name(mypr, pr)); 2018 if (error != 0 && error != ENOENT) 2019 goto done_deref; 2020 error = vfs_setopt(opts, "cpuset.id", &pr->pr_cpuset->cs_id, 2021 sizeof(pr->pr_cpuset->cs_id)); 2022 if (error != 0 && error != ENOENT) 2023 goto done_deref; 2024 error = vfs_setopts(opts, "path", prison_path(mypr, pr)); 2025 if (error != 0 && error != ENOENT) 2026 goto done_deref; 2027#ifdef INET 2028 error = vfs_setopt_part(opts, "ip4.addr", pr->pr_ip4, 2029 pr->pr_ip4s * sizeof(*pr->pr_ip4)); 2030 if (error != 0 && error != ENOENT) 2031 goto done_deref; 2032#endif 2033#ifdef INET6 2034 error = vfs_setopt_part(opts, "ip6.addr", pr->pr_ip6, 2035 pr->pr_ip6s * sizeof(*pr->pr_ip6)); 2036 if (error != 0 && error != ENOENT) 2037 goto done_deref; 2038#endif 2039 error = vfs_setopt(opts, "securelevel", &pr->pr_securelevel, 2040 sizeof(pr->pr_securelevel)); 2041 if (error != 0 && error != ENOENT) 2042 goto done_deref; 2043 error = vfs_setopt(opts, "children.cur", &pr->pr_childcount, 2044 sizeof(pr->pr_childcount)); 2045 if (error != 0 && error != ENOENT) 2046 goto done_deref; 2047 error = vfs_setopt(opts, "children.max", &pr->pr_childmax, 2048 sizeof(pr->pr_childmax)); 2049 if (error != 0 && error != ENOENT) 2050 goto done_deref; 2051 error = vfs_setopts(opts, "host.hostname", pr->pr_hostname); 2052 if (error != 0 && error != ENOENT) 2053 goto done_deref; 2054 error = vfs_setopts(opts, "host.domainname", pr->pr_domainname); 2055 if (error != 0 && error != ENOENT) 2056 goto done_deref; 2057 error = vfs_setopts(opts, "host.hostuuid", pr->pr_hostuuid); 2058 if (error != 0 && error != ENOENT) 2059 goto done_deref; 2060#ifdef COMPAT_FREEBSD32 2061 if (SV_PROC_FLAG(td->td_proc, SV_ILP32)) { 2062 uint32_t hid32 = pr->pr_hostid; 2063 2064 error = vfs_setopt(opts, "host.hostid", &hid32, sizeof(hid32)); 2065 } else 2066#endif 2067 error = vfs_setopt(opts, "host.hostid", &pr->pr_hostid, 2068 sizeof(pr->pr_hostid)); 2069 if (error != 0 && error != ENOENT) 2070 goto done_deref; 2071 error = vfs_setopt(opts, "enforce_statfs", &pr->pr_enforce_statfs, 2072 sizeof(pr->pr_enforce_statfs)); 2073 if (error != 0 && error != ENOENT) 2074 goto done_deref; 2075 error = vfs_setopt(opts, "devfs_ruleset", &pr->pr_devfs_rsnum, 2076 sizeof(pr->pr_devfs_rsnum)); 2077 if (error != 0 && error != ENOENT) 2078 goto done_deref; 2079 for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]); 2080 fi++) { 2081 if (pr_flag_names[fi] == NULL) 2082 continue; 2083 i = (pr->pr_flags & (1 << fi)) ? 1 : 0; 2084 error = vfs_setopt(opts, pr_flag_names[fi], &i, sizeof(i)); 2085 if (error != 0 && error != ENOENT) 2086 goto done_deref; 2087 i = !i; 2088 error = vfs_setopt(opts, pr_flag_nonames[fi], &i, sizeof(i)); 2089 if (error != 0 && error != ENOENT) 2090 goto done_deref; 2091 } 2092 for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]); 2093 fi++) { 2094 i = pr->pr_flags & 2095 (pr_flag_jailsys[fi].disable | pr_flag_jailsys[fi].new); 2096 i = pr_flag_jailsys[fi].disable && 2097 (i == pr_flag_jailsys[fi].disable) ? JAIL_SYS_DISABLE 2098 : (i == pr_flag_jailsys[fi].new) ? JAIL_SYS_NEW 2099 : JAIL_SYS_INHERIT; 2100 error = 2101 vfs_setopt(opts, pr_flag_jailsys[fi].name, &i, sizeof(i)); 2102 if (error != 0 && error != ENOENT) 2103 goto done_deref; 2104 } 2105 for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]); 2106 fi++) { 2107 if (pr_allow_names[fi] == NULL) 2108 continue; 2109 i = (pr->pr_allow & (1 << fi)) ? 1 : 0; 2110 error = vfs_setopt(opts, pr_allow_names[fi], &i, sizeof(i)); 2111 if (error != 0 && error != ENOENT) 2112 goto done_deref; 2113 i = !i; 2114 error = vfs_setopt(opts, pr_allow_nonames[fi], &i, sizeof(i)); 2115 if (error != 0 && error != ENOENT) 2116 goto done_deref; 2117 } 2118 i = (pr->pr_uref == 0); 2119 error = vfs_setopt(opts, "dying", &i, sizeof(i)); 2120 if (error != 0 && error != ENOENT) 2121 goto done_deref; 2122 i = !i; 2123 error = vfs_setopt(opts, "nodying", &i, sizeof(i)); 2124 if (error != 0 && error != ENOENT) 2125 goto done_deref; 2126 2127 /* Get the module parameters. */ 2128 mtx_unlock(&pr->pr_mtx); 2129 locked = 0; 2130 error = osd_jail_call(pr, PR_METHOD_GET, opts); 2131 if (error) 2132 goto done_deref; 2133 prison_deref(pr, PD_DEREF | PD_LIST_SLOCKED); 2134 2135 /* By now, all parameters should have been noted. */ 2136 TAILQ_FOREACH(opt, opts, link) { 2137 if (!opt->seen && strcmp(opt->name, "errmsg")) { 2138 error = EINVAL; 2139 vfs_opterror(opts, "unknown parameter: %s", opt->name); 2140 goto done_errmsg; 2141 } 2142 } 2143 2144 /* Write the fetched parameters back to userspace. */ 2145 error = 0; 2146 TAILQ_FOREACH(opt, opts, link) { 2147 if (opt->pos >= 0 && opt->pos != errmsg_pos) { 2148 pos = 2 * opt->pos + 1; 2149 optuio->uio_iov[pos].iov_len = opt->len; 2150 if (opt->value != NULL) { 2151 if (optuio->uio_segflg == UIO_SYSSPACE) { 2152 bcopy(opt->value, 2153 optuio->uio_iov[pos].iov_base, 2154 opt->len); 2155 } else { 2156 error = copyout(opt->value, 2157 optuio->uio_iov[pos].iov_base, 2158 opt->len); 2159 if (error) 2160 break; 2161 } 2162 } 2163 } 2164 } 2165 goto done_errmsg; 2166 2167 done_deref: 2168 prison_deref(pr, locked | PD_DEREF | PD_LIST_SLOCKED); 2169 goto done_errmsg; 2170 2171 done_unlock_list: 2172 sx_sunlock(&allprison_lock); 2173 done_errmsg: 2174 if (error && errmsg_pos >= 0) { 2175 vfs_getopt(opts, "errmsg", (void **)&errmsg, &errmsg_len); 2176 errmsg_pos = 2 * errmsg_pos + 1; 2177 if (errmsg_len > 0) { 2178 if (optuio->uio_segflg == UIO_SYSSPACE) 2179 bcopy(errmsg, 2180 optuio->uio_iov[errmsg_pos].iov_base, 2181 errmsg_len); 2182 else 2183 copyout(errmsg, 2184 optuio->uio_iov[errmsg_pos].iov_base, 2185 errmsg_len); 2186 } 2187 } 2188 vfs_freeopts(opts); 2189 return (error); 2190} 2191 2192 2193/* 2194 * struct jail_remove_args { 2195 * int jid; 2196 * }; 2197 */ 2198int 2199sys_jail_remove(struct thread *td, struct jail_remove_args *uap) 2200{ 2201 struct prison *pr, *cpr, *lpr, *tpr; 2202 int descend, error; 2203 2204 error = priv_check(td, PRIV_JAIL_REMOVE); 2205 if (error) 2206 return (error); 2207 2208 sx_xlock(&allprison_lock); 2209 pr = prison_find_child(td->td_ucred->cr_prison, uap->jid); 2210 if (pr == NULL) { 2211 sx_xunlock(&allprison_lock); 2212 return (EINVAL); 2213 } 2214 2215 /* Remove all descendants of this prison, then remove this prison. */ 2216 pr->pr_ref++; 2217 pr->pr_flags |= PR_REMOVE; 2218 if (!LIST_EMPTY(&pr->pr_children)) { 2219 mtx_unlock(&pr->pr_mtx); 2220 lpr = NULL; 2221 FOREACH_PRISON_DESCENDANT(pr, cpr, descend) { 2222 mtx_lock(&cpr->pr_mtx); 2223 if (cpr->pr_ref > 0) { 2224 tpr = cpr; 2225 cpr->pr_ref++; 2226 cpr->pr_flags |= PR_REMOVE; 2227 } else { 2228 /* Already removed - do not do it again. */ 2229 tpr = NULL; 2230 } 2231 mtx_unlock(&cpr->pr_mtx); 2232 if (lpr != NULL) { 2233 mtx_lock(&lpr->pr_mtx); 2234 prison_remove_one(lpr); 2235 sx_xlock(&allprison_lock); 2236 } 2237 lpr = tpr; 2238 } 2239 if (lpr != NULL) { 2240 mtx_lock(&lpr->pr_mtx); 2241 prison_remove_one(lpr); 2242 sx_xlock(&allprison_lock); 2243 } 2244 mtx_lock(&pr->pr_mtx); 2245 } 2246 prison_remove_one(pr); 2247 return (0); 2248} 2249 2250static void 2251prison_remove_one(struct prison *pr) 2252{ 2253 struct proc *p; 2254 int deuref; 2255 2256 /* If the prison was persistent, it is not anymore. */ 2257 deuref = 0; 2258 if (pr->pr_flags & PR_PERSIST) { 2259 pr->pr_ref--; 2260 deuref = PD_DEUREF; 2261 pr->pr_flags &= ~PR_PERSIST; 2262 } 2263 2264 /* 2265 * jail_remove added a reference. If that's the only one, remove 2266 * the prison now. 2267 */ 2268 KASSERT(pr->pr_ref > 0, 2269 ("prison_remove_one removing a dead prison (jid=%d)", pr->pr_id)); 2270 if (pr->pr_ref == 1) { 2271 prison_deref(pr, 2272 deuref | PD_DEREF | PD_LOCKED | PD_LIST_XLOCKED); 2273 return; 2274 } 2275 2276 mtx_unlock(&pr->pr_mtx); 2277 sx_xunlock(&allprison_lock); 2278 /* 2279 * Kill all processes unfortunate enough to be attached to this prison. 2280 */ 2281 sx_slock(&allproc_lock); 2282 LIST_FOREACH(p, &allproc, p_list) { 2283 PROC_LOCK(p); 2284 if (p->p_state != PRS_NEW && p->p_ucred && 2285 p->p_ucred->cr_prison == pr) 2286 kern_psignal(p, SIGKILL); 2287 PROC_UNLOCK(p); 2288 } 2289 sx_sunlock(&allproc_lock); 2290 /* Remove the temporary reference added by jail_remove. */ 2291 prison_deref(pr, deuref | PD_DEREF); 2292} 2293 2294 2295/* 2296 * struct jail_attach_args { 2297 * int jid; 2298 * }; 2299 */ 2300int 2301sys_jail_attach(struct thread *td, struct jail_attach_args *uap) 2302{ 2303 struct prison *pr; 2304 int error; 2305 2306 error = priv_check(td, PRIV_JAIL_ATTACH); 2307 if (error) 2308 return (error); 2309 2310 sx_slock(&allprison_lock); 2311 pr = prison_find_child(td->td_ucred->cr_prison, uap->jid); 2312 if (pr == NULL) { 2313 sx_sunlock(&allprison_lock); 2314 return (EINVAL); 2315 } 2316 2317 /* 2318 * Do not allow a process to attach to a prison that is not 2319 * considered to be "alive". 2320 */ 2321 if (pr->pr_uref == 0) { 2322 mtx_unlock(&pr->pr_mtx); 2323 sx_sunlock(&allprison_lock); 2324 return (EINVAL); 2325 } 2326 2327 return (do_jail_attach(td, pr)); 2328} 2329 2330static int 2331do_jail_attach(struct thread *td, struct prison *pr) 2332{ 2333 struct prison *ppr; 2334 struct proc *p; 2335 struct ucred *newcred, *oldcred; 2336 int vfslocked, error; 2337 2338 /* 2339 * XXX: Note that there is a slight race here if two threads 2340 * in the same privileged process attempt to attach to two 2341 * different jails at the same time. It is important for 2342 * user processes not to do this, or they might end up with 2343 * a process root from one prison, but attached to the jail 2344 * of another. 2345 */ 2346 pr->pr_ref++; 2347 pr->pr_uref++; 2348 mtx_unlock(&pr->pr_mtx); 2349 2350 /* Let modules do whatever they need to prepare for attaching. */ 2351 error = osd_jail_call(pr, PR_METHOD_ATTACH, td); 2352 if (error) { 2353 prison_deref(pr, PD_DEREF | PD_DEUREF | PD_LIST_SLOCKED); 2354 return (error); 2355 } 2356 sx_sunlock(&allprison_lock); 2357 2358 /* 2359 * Reparent the newly attached process to this jail. 2360 */ 2361 ppr = td->td_ucred->cr_prison; 2362 p = td->td_proc; 2363 error = cpuset_setproc_update_set(p, pr->pr_cpuset); 2364 if (error) 2365 goto e_revert_osd; 2366 2367 vfslocked = VFS_LOCK_GIANT(pr->pr_root->v_mount); 2368 vn_lock(pr->pr_root, LK_EXCLUSIVE | LK_RETRY); 2369 if ((error = change_dir(pr->pr_root, td)) != 0) 2370 goto e_unlock; 2371#ifdef MAC 2372 if ((error = mac_vnode_check_chroot(td->td_ucred, pr->pr_root))) 2373 goto e_unlock; 2374#endif 2375 VOP_UNLOCK(pr->pr_root, 0); 2376 if ((error = change_root(pr->pr_root, td))) 2377 goto e_unlock_giant; 2378 VFS_UNLOCK_GIANT(vfslocked); 2379 2380 newcred = crget(); 2381 PROC_LOCK(p); 2382 oldcred = p->p_ucred; 2383 setsugid(p); 2384 crcopy(newcred, oldcred); 2385 newcred->cr_prison = pr; 2386 p->p_ucred = newcred; 2387 PROC_UNLOCK(p); 2388#ifdef RACCT 2389 racct_proc_ucred_changed(p, oldcred, newcred); 2390#endif 2391 crfree(oldcred); 2392 prison_deref(ppr, PD_DEREF | PD_DEUREF); 2393 return (0); 2394 e_unlock: 2395 VOP_UNLOCK(pr->pr_root, 0); 2396 e_unlock_giant: 2397 VFS_UNLOCK_GIANT(vfslocked); 2398 e_revert_osd: 2399 /* Tell modules this thread is still in its old jail after all. */ 2400 (void)osd_jail_call(ppr, PR_METHOD_ATTACH, td); 2401 prison_deref(pr, PD_DEREF | PD_DEUREF); 2402 return (error); 2403} 2404 2405 2406/* 2407 * Returns a locked prison instance, or NULL on failure. 2408 */ 2409struct prison * 2410prison_find(int prid) 2411{ 2412 struct prison *pr; 2413 2414 sx_assert(&allprison_lock, SX_LOCKED); 2415 TAILQ_FOREACH(pr, &allprison, pr_list) { 2416 if (pr->pr_id == prid) { 2417 mtx_lock(&pr->pr_mtx); 2418 if (pr->pr_ref > 0) 2419 return (pr); 2420 mtx_unlock(&pr->pr_mtx); 2421 } 2422 } 2423 return (NULL); 2424} 2425 2426/* 2427 * Find a prison that is a descendant of mypr. Returns a locked prison or NULL. 2428 */ 2429struct prison * 2430prison_find_child(struct prison *mypr, int prid) 2431{ 2432 struct prison *pr; 2433 int descend; 2434 2435 sx_assert(&allprison_lock, SX_LOCKED); 2436 FOREACH_PRISON_DESCENDANT(mypr, pr, descend) { 2437 if (pr->pr_id == prid) { 2438 mtx_lock(&pr->pr_mtx); 2439 if (pr->pr_ref > 0) 2440 return (pr); 2441 mtx_unlock(&pr->pr_mtx); 2442 } 2443 } 2444 return (NULL); 2445} 2446 2447/* 2448 * Look for the name relative to mypr. Returns a locked prison or NULL. 2449 */ 2450struct prison * 2451prison_find_name(struct prison *mypr, const char *name) 2452{ 2453 struct prison *pr, *deadpr; 2454 size_t mylen; 2455 int descend; 2456 2457 sx_assert(&allprison_lock, SX_LOCKED); 2458 mylen = (mypr == &prison0) ? 0 : strlen(mypr->pr_name) + 1; 2459 again: 2460 deadpr = NULL; 2461 FOREACH_PRISON_DESCENDANT(mypr, pr, descend) { 2462 if (!strcmp(pr->pr_name + mylen, name)) { 2463 mtx_lock(&pr->pr_mtx); 2464 if (pr->pr_ref > 0) { 2465 if (pr->pr_uref > 0) 2466 return (pr); 2467 deadpr = pr; 2468 } 2469 mtx_unlock(&pr->pr_mtx); 2470 } 2471 } 2472 /* There was no valid prison - perhaps there was a dying one. */ 2473 if (deadpr != NULL) { 2474 mtx_lock(&deadpr->pr_mtx); 2475 if (deadpr->pr_ref == 0) { 2476 mtx_unlock(&deadpr->pr_mtx); 2477 goto again; 2478 } 2479 } 2480 return (deadpr); 2481} 2482 2483/* 2484 * See if a prison has the specific flag set. 2485 */ 2486int 2487prison_flag(struct ucred *cred, unsigned flag) 2488{ 2489 2490 /* This is an atomic read, so no locking is necessary. */ 2491 return (cred->cr_prison->pr_flags & flag); 2492} 2493 2494int 2495prison_allow(struct ucred *cred, unsigned flag) 2496{ 2497 2498 /* This is an atomic read, so no locking is necessary. */ 2499 return (cred->cr_prison->pr_allow & flag); 2500} 2501 2502/* 2503 * Remove a prison reference. If that was the last reference, remove the 2504 * prison itself - but not in this context in case there are locks held. 2505 */ 2506void 2507prison_free_locked(struct prison *pr) 2508{ 2509 2510 mtx_assert(&pr->pr_mtx, MA_OWNED); 2511 pr->pr_ref--; 2512 if (pr->pr_ref == 0) { 2513 mtx_unlock(&pr->pr_mtx); 2514 TASK_INIT(&pr->pr_task, 0, prison_complete, pr); 2515 taskqueue_enqueue(taskqueue_thread, &pr->pr_task); 2516 return; 2517 } 2518 mtx_unlock(&pr->pr_mtx); 2519} 2520 2521void 2522prison_free(struct prison *pr) 2523{ 2524 2525 mtx_lock(&pr->pr_mtx); 2526 prison_free_locked(pr); 2527} 2528 2529static void 2530prison_complete(void *context, int pending) 2531{ 2532 2533 prison_deref((struct prison *)context, 0); 2534} 2535 2536/* 2537 * Remove a prison reference (usually). This internal version assumes no 2538 * mutexes are held, except perhaps the prison itself. If there are no more 2539 * references, release and delist the prison. On completion, the prison lock 2540 * and the allprison lock are both unlocked. 2541 */ 2542static void 2543prison_deref(struct prison *pr, int flags) 2544{ 2545 struct prison *ppr, *tpr; 2546 int vfslocked; 2547 2548 if (!(flags & PD_LOCKED)) 2549 mtx_lock(&pr->pr_mtx); 2550 for (;;) { 2551 if (flags & PD_DEUREF) { 2552 pr->pr_uref--; 2553 KASSERT(prison0.pr_uref != 0, ("prison0 pr_uref=0")); 2554 } 2555 if (flags & PD_DEREF) 2556 pr->pr_ref--; 2557 /* If the prison still has references, nothing else to do. */ 2558 if (pr->pr_ref > 0) { 2559 mtx_unlock(&pr->pr_mtx); 2560 if (flags & PD_LIST_SLOCKED) 2561 sx_sunlock(&allprison_lock); 2562 else if (flags & PD_LIST_XLOCKED) 2563 sx_xunlock(&allprison_lock); 2564 return; 2565 } 2566 2567 mtx_unlock(&pr->pr_mtx); 2568 if (flags & PD_LIST_SLOCKED) { 2569 if (!sx_try_upgrade(&allprison_lock)) { 2570 sx_sunlock(&allprison_lock); 2571 sx_xlock(&allprison_lock); 2572 } 2573 } else if (!(flags & PD_LIST_XLOCKED)) 2574 sx_xlock(&allprison_lock); 2575 2576 TAILQ_REMOVE(&allprison, pr, pr_list); 2577 LIST_REMOVE(pr, pr_sibling); 2578 ppr = pr->pr_parent; 2579 for (tpr = ppr; tpr != NULL; tpr = tpr->pr_parent) 2580 tpr->pr_childcount--; 2581 sx_xunlock(&allprison_lock); 2582 2583#ifdef VIMAGE 2584 if (pr->pr_vnet != ppr->pr_vnet) 2585 vnet_destroy(pr->pr_vnet); 2586#endif 2587 if (pr->pr_root != NULL) { 2588 vfslocked = VFS_LOCK_GIANT(pr->pr_root->v_mount); 2589 vrele(pr->pr_root); 2590 VFS_UNLOCK_GIANT(vfslocked); 2591 } 2592 mtx_destroy(&pr->pr_mtx); 2593#ifdef INET 2594 free(pr->pr_ip4, M_PRISON); 2595#endif 2596#ifdef INET6 2597 free(pr->pr_ip6, M_PRISON); 2598#endif 2599 if (pr->pr_cpuset != NULL) 2600 cpuset_rel(pr->pr_cpuset); 2601 osd_jail_exit(pr); 2602#ifdef RACCT 2603 prison_racct_detach(pr); 2604#endif 2605 free(pr, M_PRISON); 2606 2607 /* Removing a prison frees a reference on its parent. */ 2608 pr = ppr; 2609 mtx_lock(&pr->pr_mtx); 2610 flags = PD_DEREF | PD_DEUREF; 2611 } 2612} 2613 2614void 2615prison_hold_locked(struct prison *pr) 2616{ 2617 2618 mtx_assert(&pr->pr_mtx, MA_OWNED); 2619 KASSERT(pr->pr_ref > 0, 2620 ("Trying to hold dead prison (jid=%d).", pr->pr_id)); 2621 pr->pr_ref++; 2622} 2623 2624void 2625prison_hold(struct prison *pr) 2626{ 2627 2628 mtx_lock(&pr->pr_mtx); 2629 prison_hold_locked(pr); 2630 mtx_unlock(&pr->pr_mtx); 2631} 2632 2633void 2634prison_proc_hold(struct prison *pr) 2635{ 2636 2637 mtx_lock(&pr->pr_mtx); 2638 KASSERT(pr->pr_uref > 0, 2639 ("Cannot add a process to a non-alive prison (jid=%d)", pr->pr_id)); 2640 pr->pr_uref++; 2641 mtx_unlock(&pr->pr_mtx); 2642} 2643 2644void 2645prison_proc_free(struct prison *pr) 2646{ 2647 2648 mtx_lock(&pr->pr_mtx); 2649 KASSERT(pr->pr_uref > 0, 2650 ("Trying to kill a process in a dead prison (jid=%d)", pr->pr_id)); 2651 prison_deref(pr, PD_DEUREF | PD_LOCKED); 2652} 2653 2654 2655#ifdef INET 2656/* 2657 * Restrict a prison's IP address list with its parent's, possibly replacing 2658 * it. Return true if the replacement buffer was used (or would have been). 2659 */ 2660static int 2661prison_restrict_ip4(struct prison *pr, struct in_addr *newip4) 2662{ 2663 int ii, ij, used; 2664 struct prison *ppr; 2665 2666 ppr = pr->pr_parent; 2667 if (!(pr->pr_flags & PR_IP4_USER)) { 2668 /* This has no user settings, so just copy the parent's list. */ 2669 if (pr->pr_ip4s < ppr->pr_ip4s) { 2670 /* 2671 * There's no room for the parent's list. Use the 2672 * new list buffer, which is assumed to be big enough 2673 * (if it was passed). If there's no buffer, try to 2674 * allocate one. 2675 */ 2676 used = 1; 2677 if (newip4 == NULL) { 2678 newip4 = malloc(ppr->pr_ip4s * sizeof(*newip4), 2679 M_PRISON, M_NOWAIT); 2680 if (newip4 != NULL) 2681 used = 0; 2682 } 2683 if (newip4 != NULL) { 2684 bcopy(ppr->pr_ip4, newip4, 2685 ppr->pr_ip4s * sizeof(*newip4)); 2686 free(pr->pr_ip4, M_PRISON); 2687 pr->pr_ip4 = newip4; 2688 pr->pr_ip4s = ppr->pr_ip4s; 2689 } 2690 return (used); 2691 } 2692 pr->pr_ip4s = ppr->pr_ip4s; 2693 if (pr->pr_ip4s > 0) 2694 bcopy(ppr->pr_ip4, pr->pr_ip4, 2695 pr->pr_ip4s * sizeof(*newip4)); 2696 else if (pr->pr_ip4 != NULL) { 2697 free(pr->pr_ip4, M_PRISON); 2698 pr->pr_ip4 = NULL; 2699 } 2700 } else if (pr->pr_ip4s > 0) { 2701 /* Remove addresses that aren't in the parent. */ 2702 for (ij = 0; ij < ppr->pr_ip4s; ij++) 2703 if (pr->pr_ip4[0].s_addr == ppr->pr_ip4[ij].s_addr) 2704 break; 2705 if (ij < ppr->pr_ip4s) 2706 ii = 1; 2707 else { 2708 bcopy(pr->pr_ip4 + 1, pr->pr_ip4, 2709 --pr->pr_ip4s * sizeof(*pr->pr_ip4)); 2710 ii = 0; 2711 } 2712 for (ij = 1; ii < pr->pr_ip4s; ) { 2713 if (pr->pr_ip4[ii].s_addr == ppr->pr_ip4[0].s_addr) { 2714 ii++; 2715 continue; 2716 } 2717 switch (ij >= ppr->pr_ip4s ? -1 : 2718 qcmp_v4(&pr->pr_ip4[ii], &ppr->pr_ip4[ij])) { 2719 case -1: 2720 bcopy(pr->pr_ip4 + ii + 1, pr->pr_ip4 + ii, 2721 (--pr->pr_ip4s - ii) * sizeof(*pr->pr_ip4)); 2722 break; 2723 case 0: 2724 ii++; 2725 ij++; 2726 break; 2727 case 1: 2728 ij++; 2729 break; 2730 } 2731 } 2732 if (pr->pr_ip4s == 0) { 2733 pr->pr_flags |= PR_IP4_DISABLE; 2734 free(pr->pr_ip4, M_PRISON); 2735 pr->pr_ip4 = NULL; 2736 } 2737 } 2738 return (0); 2739} 2740 2741/* 2742 * Pass back primary IPv4 address of this jail. 2743 * 2744 * If not restricted return success but do not alter the address. Caller has 2745 * to make sure to initialize it correctly (e.g. INADDR_ANY). 2746 * 2747 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv4. 2748 * Address returned in NBO. 2749 */ 2750int 2751prison_get_ip4(struct ucred *cred, struct in_addr *ia) 2752{ 2753 struct prison *pr; 2754 2755 KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); 2756 KASSERT(ia != NULL, ("%s: ia is NULL", __func__)); 2757 2758 pr = cred->cr_prison; 2759 if (!(pr->pr_flags & PR_IP4)) 2760 return (0); 2761 mtx_lock(&pr->pr_mtx); 2762 if (!(pr->pr_flags & PR_IP4)) { 2763 mtx_unlock(&pr->pr_mtx); 2764 return (0); 2765 } 2766 if (pr->pr_ip4 == NULL) { 2767 mtx_unlock(&pr->pr_mtx); 2768 return (EAFNOSUPPORT); 2769 } 2770 2771 ia->s_addr = pr->pr_ip4[0].s_addr; 2772 mtx_unlock(&pr->pr_mtx); 2773 return (0); 2774} 2775 2776/* 2777 * Return 1 if we should do proper source address selection or are not jailed. 2778 * We will return 0 if we should bypass source address selection in favour 2779 * of the primary jail IPv4 address. Only in this case *ia will be updated and 2780 * returned in NBO. 2781 * Return EAFNOSUPPORT, in case this jail does not allow IPv4. 2782 */ 2783int 2784prison_saddrsel_ip4(struct ucred *cred, struct in_addr *ia) 2785{ 2786 struct prison *pr; 2787 struct in_addr lia; 2788 int error; 2789 2790 KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); 2791 KASSERT(ia != NULL, ("%s: ia is NULL", __func__)); 2792 2793 if (!jailed(cred)) 2794 return (1); 2795 2796 pr = cred->cr_prison; 2797 if (pr->pr_flags & PR_IP4_SADDRSEL) 2798 return (1); 2799 2800 lia.s_addr = INADDR_ANY; 2801 error = prison_get_ip4(cred, &lia); 2802 if (error) 2803 return (error); 2804 if (lia.s_addr == INADDR_ANY) 2805 return (1); 2806 2807 ia->s_addr = lia.s_addr; 2808 return (0); 2809} 2810 2811/* 2812 * Return true if pr1 and pr2 have the same IPv4 address restrictions. 2813 */ 2814int 2815prison_equal_ip4(struct prison *pr1, struct prison *pr2) 2816{ 2817 2818 if (pr1 == pr2) 2819 return (1); 2820 2821 /* 2822 * No need to lock since the PR_IP4_USER flag can't be altered for 2823 * existing prisons. 2824 */ 2825 while (pr1 != &prison0 && 2826#ifdef VIMAGE 2827 !(pr1->pr_flags & PR_VNET) && 2828#endif 2829 !(pr1->pr_flags & PR_IP4_USER)) 2830 pr1 = pr1->pr_parent; 2831 while (pr2 != &prison0 && 2832#ifdef VIMAGE 2833 !(pr2->pr_flags & PR_VNET) && 2834#endif 2835 !(pr2->pr_flags & PR_IP4_USER)) 2836 pr2 = pr2->pr_parent; 2837 return (pr1 == pr2); 2838} 2839 2840/* 2841 * Make sure our (source) address is set to something meaningful to this 2842 * jail. 2843 * 2844 * Returns 0 if jail doesn't restrict IPv4 or if address belongs to jail, 2845 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail 2846 * doesn't allow IPv4. Address passed in in NBO and returned in NBO. 2847 */ 2848int 2849prison_local_ip4(struct ucred *cred, struct in_addr *ia) 2850{ 2851 struct prison *pr; 2852 struct in_addr ia0; 2853 int error; 2854 2855 KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); 2856 KASSERT(ia != NULL, ("%s: ia is NULL", __func__)); 2857 2858 pr = cred->cr_prison; 2859 if (!(pr->pr_flags & PR_IP4)) 2860 return (0); 2861 mtx_lock(&pr->pr_mtx); 2862 if (!(pr->pr_flags & PR_IP4)) { 2863 mtx_unlock(&pr->pr_mtx); 2864 return (0); 2865 } 2866 if (pr->pr_ip4 == NULL) { 2867 mtx_unlock(&pr->pr_mtx); 2868 return (EAFNOSUPPORT); 2869 } 2870 2871 ia0.s_addr = ntohl(ia->s_addr); 2872 if (ia0.s_addr == INADDR_LOOPBACK) { 2873 ia->s_addr = pr->pr_ip4[0].s_addr; 2874 mtx_unlock(&pr->pr_mtx); 2875 return (0); 2876 } 2877 2878 if (ia0.s_addr == INADDR_ANY) { 2879 /* 2880 * In case there is only 1 IPv4 address, bind directly. 2881 */ 2882 if (pr->pr_ip4s == 1) 2883 ia->s_addr = pr->pr_ip4[0].s_addr; 2884 mtx_unlock(&pr->pr_mtx); 2885 return (0); 2886 } 2887 2888 error = _prison_check_ip4(pr, ia); 2889 mtx_unlock(&pr->pr_mtx); 2890 return (error); 2891} 2892 2893/* 2894 * Rewrite destination address in case we will connect to loopback address. 2895 * 2896 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv4. 2897 * Address passed in in NBO and returned in NBO. 2898 */ 2899int 2900prison_remote_ip4(struct ucred *cred, struct in_addr *ia) 2901{ 2902 struct prison *pr; 2903 2904 KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); 2905 KASSERT(ia != NULL, ("%s: ia is NULL", __func__)); 2906 2907 pr = cred->cr_prison; 2908 if (!(pr->pr_flags & PR_IP4)) 2909 return (0); 2910 mtx_lock(&pr->pr_mtx); 2911 if (!(pr->pr_flags & PR_IP4)) { 2912 mtx_unlock(&pr->pr_mtx); 2913 return (0); 2914 } 2915 if (pr->pr_ip4 == NULL) { 2916 mtx_unlock(&pr->pr_mtx); 2917 return (EAFNOSUPPORT); 2918 } 2919 2920 if (ntohl(ia->s_addr) == INADDR_LOOPBACK) { 2921 ia->s_addr = pr->pr_ip4[0].s_addr; 2922 mtx_unlock(&pr->pr_mtx); 2923 return (0); 2924 } 2925 2926 /* 2927 * Return success because nothing had to be changed. 2928 */ 2929 mtx_unlock(&pr->pr_mtx); 2930 return (0); 2931} 2932 2933/* 2934 * Check if given address belongs to the jail referenced by cred/prison. 2935 * 2936 * Returns 0 if jail doesn't restrict IPv4 or if address belongs to jail, 2937 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail 2938 * doesn't allow IPv4. Address passed in in NBO. 2939 */ 2940static int 2941_prison_check_ip4(struct prison *pr, struct in_addr *ia) 2942{ 2943 int i, a, z, d; 2944 2945 /* 2946 * Check the primary IP. 2947 */ 2948 if (pr->pr_ip4[0].s_addr == ia->s_addr) 2949 return (0); 2950 2951 /* 2952 * All the other IPs are sorted so we can do a binary search. 2953 */ 2954 a = 0; 2955 z = pr->pr_ip4s - 2; 2956 while (a <= z) { 2957 i = (a + z) / 2; 2958 d = qcmp_v4(&pr->pr_ip4[i+1], ia); 2959 if (d > 0) 2960 z = i - 1; 2961 else if (d < 0) 2962 a = i + 1; 2963 else 2964 return (0); 2965 } 2966 2967 return (EADDRNOTAVAIL); 2968} 2969 2970int 2971prison_check_ip4(struct ucred *cred, struct in_addr *ia) 2972{ 2973 struct prison *pr; 2974 int error; 2975 2976 KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); 2977 KASSERT(ia != NULL, ("%s: ia is NULL", __func__)); 2978 2979 pr = cred->cr_prison; 2980 if (!(pr->pr_flags & PR_IP4)) 2981 return (0); 2982 mtx_lock(&pr->pr_mtx); 2983 if (!(pr->pr_flags & PR_IP4)) { 2984 mtx_unlock(&pr->pr_mtx); 2985 return (0); 2986 } 2987 if (pr->pr_ip4 == NULL) { 2988 mtx_unlock(&pr->pr_mtx); 2989 return (EAFNOSUPPORT); 2990 } 2991 2992 error = _prison_check_ip4(pr, ia); 2993 mtx_unlock(&pr->pr_mtx); 2994 return (error); 2995} 2996#endif 2997 2998#ifdef INET6 2999static int 3000prison_restrict_ip6(struct prison *pr, struct in6_addr *newip6) 3001{ 3002 int ii, ij, used; 3003 struct prison *ppr; 3004 3005 ppr = pr->pr_parent; 3006 if (!(pr->pr_flags & PR_IP6_USER)) { 3007 /* This has no user settings, so just copy the parent's list. */ 3008 if (pr->pr_ip6s < ppr->pr_ip6s) { 3009 /* 3010 * There's no room for the parent's list. Use the 3011 * new list buffer, which is assumed to be big enough 3012 * (if it was passed). If there's no buffer, try to 3013 * allocate one. 3014 */ 3015 used = 1; 3016 if (newip6 == NULL) { 3017 newip6 = malloc(ppr->pr_ip6s * sizeof(*newip6), 3018 M_PRISON, M_NOWAIT); 3019 if (newip6 != NULL) 3020 used = 0; 3021 } 3022 if (newip6 != NULL) { 3023 bcopy(ppr->pr_ip6, newip6, 3024 ppr->pr_ip6s * sizeof(*newip6)); 3025 free(pr->pr_ip6, M_PRISON); 3026 pr->pr_ip6 = newip6; 3027 pr->pr_ip6s = ppr->pr_ip6s; 3028 } 3029 return (used); 3030 } 3031 pr->pr_ip6s = ppr->pr_ip6s; 3032 if (pr->pr_ip6s > 0) 3033 bcopy(ppr->pr_ip6, pr->pr_ip6, 3034 pr->pr_ip6s * sizeof(*newip6)); 3035 else if (pr->pr_ip6 != NULL) { 3036 free(pr->pr_ip6, M_PRISON); 3037 pr->pr_ip6 = NULL; 3038 } 3039 } else if (pr->pr_ip6s > 0) { 3040 /* Remove addresses that aren't in the parent. */ 3041 for (ij = 0; ij < ppr->pr_ip6s; ij++) 3042 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[0], 3043 &ppr->pr_ip6[ij])) 3044 break; 3045 if (ij < ppr->pr_ip6s) 3046 ii = 1; 3047 else { 3048 bcopy(pr->pr_ip6 + 1, pr->pr_ip6, 3049 --pr->pr_ip6s * sizeof(*pr->pr_ip6)); 3050 ii = 0; 3051 } 3052 for (ij = 1; ii < pr->pr_ip6s; ) { 3053 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[ii], 3054 &ppr->pr_ip6[0])) { 3055 ii++; 3056 continue; 3057 } 3058 switch (ij >= ppr->pr_ip4s ? -1 : 3059 qcmp_v6(&pr->pr_ip6[ii], &ppr->pr_ip6[ij])) { 3060 case -1: 3061 bcopy(pr->pr_ip6 + ii + 1, pr->pr_ip6 + ii, 3062 (--pr->pr_ip6s - ii) * sizeof(*pr->pr_ip6)); 3063 break; 3064 case 0: 3065 ii++; 3066 ij++; 3067 break; 3068 case 1: 3069 ij++; 3070 break; 3071 } 3072 } 3073 if (pr->pr_ip6s == 0) { 3074 pr->pr_flags |= PR_IP6_DISABLE; 3075 free(pr->pr_ip6, M_PRISON); 3076 pr->pr_ip6 = NULL; 3077 } 3078 } 3079 return 0; 3080} 3081 3082/* 3083 * Pass back primary IPv6 address for this jail. 3084 * 3085 * If not restricted return success but do not alter the address. Caller has 3086 * to make sure to initialize it correctly (e.g. IN6ADDR_ANY_INIT). 3087 * 3088 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv6. 3089 */ 3090int 3091prison_get_ip6(struct ucred *cred, struct in6_addr *ia6) 3092{ 3093 struct prison *pr; 3094 3095 KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); 3096 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__)); 3097 3098 pr = cred->cr_prison; 3099 if (!(pr->pr_flags & PR_IP6)) 3100 return (0); 3101 mtx_lock(&pr->pr_mtx); 3102 if (!(pr->pr_flags & PR_IP6)) { 3103 mtx_unlock(&pr->pr_mtx); 3104 return (0); 3105 } 3106 if (pr->pr_ip6 == NULL) { 3107 mtx_unlock(&pr->pr_mtx); 3108 return (EAFNOSUPPORT); 3109 } 3110 3111 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr)); 3112 mtx_unlock(&pr->pr_mtx); 3113 return (0); 3114} 3115 3116/* 3117 * Return 1 if we should do proper source address selection or are not jailed. 3118 * We will return 0 if we should bypass source address selection in favour 3119 * of the primary jail IPv6 address. Only in this case *ia will be updated and 3120 * returned in NBO. 3121 * Return EAFNOSUPPORT, in case this jail does not allow IPv6. 3122 */ 3123int 3124prison_saddrsel_ip6(struct ucred *cred, struct in6_addr *ia6) 3125{ 3126 struct prison *pr; 3127 struct in6_addr lia6; 3128 int error; 3129 3130 KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); 3131 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__)); 3132 3133 if (!jailed(cred)) 3134 return (1); 3135 3136 pr = cred->cr_prison; 3137 if (pr->pr_flags & PR_IP6_SADDRSEL) 3138 return (1); 3139 3140 lia6 = in6addr_any; 3141 error = prison_get_ip6(cred, &lia6); 3142 if (error) 3143 return (error); 3144 if (IN6_IS_ADDR_UNSPECIFIED(&lia6)) 3145 return (1); 3146 3147 bcopy(&lia6, ia6, sizeof(struct in6_addr)); 3148 return (0); 3149} 3150 3151/* 3152 * Return true if pr1 and pr2 have the same IPv6 address restrictions. 3153 */ 3154int 3155prison_equal_ip6(struct prison *pr1, struct prison *pr2) 3156{ 3157 3158 if (pr1 == pr2) 3159 return (1); 3160 3161 while (pr1 != &prison0 && 3162#ifdef VIMAGE 3163 !(pr1->pr_flags & PR_VNET) && 3164#endif 3165 !(pr1->pr_flags & PR_IP6_USER)) 3166 pr1 = pr1->pr_parent; 3167 while (pr2 != &prison0 && 3168#ifdef VIMAGE 3169 !(pr2->pr_flags & PR_VNET) && 3170#endif 3171 !(pr2->pr_flags & PR_IP6_USER)) 3172 pr2 = pr2->pr_parent; 3173 return (pr1 == pr2); 3174} 3175 3176/* 3177 * Make sure our (source) address is set to something meaningful to this jail. 3178 * 3179 * v6only should be set based on (inp->inp_flags & IN6P_IPV6_V6ONLY != 0) 3180 * when needed while binding. 3181 * 3182 * Returns 0 if jail doesn't restrict IPv6 or if address belongs to jail, 3183 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail 3184 * doesn't allow IPv6. 3185 */ 3186int 3187prison_local_ip6(struct ucred *cred, struct in6_addr *ia6, int v6only) 3188{ 3189 struct prison *pr; 3190 int error; 3191 3192 KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); 3193 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__)); 3194 3195 pr = cred->cr_prison; 3196 if (!(pr->pr_flags & PR_IP6)) 3197 return (0); 3198 mtx_lock(&pr->pr_mtx); 3199 if (!(pr->pr_flags & PR_IP6)) { 3200 mtx_unlock(&pr->pr_mtx); 3201 return (0); 3202 } 3203 if (pr->pr_ip6 == NULL) { 3204 mtx_unlock(&pr->pr_mtx); 3205 return (EAFNOSUPPORT); 3206 } 3207 3208 if (IN6_IS_ADDR_LOOPBACK(ia6)) { 3209 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr)); 3210 mtx_unlock(&pr->pr_mtx); 3211 return (0); 3212 } 3213 3214 if (IN6_IS_ADDR_UNSPECIFIED(ia6)) { 3215 /* 3216 * In case there is only 1 IPv6 address, and v6only is true, 3217 * then bind directly. 3218 */ 3219 if (v6only != 0 && pr->pr_ip6s == 1) 3220 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr)); 3221 mtx_unlock(&pr->pr_mtx); 3222 return (0); 3223 } 3224 3225 error = _prison_check_ip6(pr, ia6); 3226 mtx_unlock(&pr->pr_mtx); 3227 return (error); 3228} 3229 3230/* 3231 * Rewrite destination address in case we will connect to loopback address. 3232 * 3233 * Returns 0 on success, EAFNOSUPPORT if the jail doesn't allow IPv6. 3234 */ 3235int 3236prison_remote_ip6(struct ucred *cred, struct in6_addr *ia6) 3237{ 3238 struct prison *pr; 3239 3240 KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); 3241 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__)); 3242 3243 pr = cred->cr_prison; 3244 if (!(pr->pr_flags & PR_IP6)) 3245 return (0); 3246 mtx_lock(&pr->pr_mtx); 3247 if (!(pr->pr_flags & PR_IP6)) { 3248 mtx_unlock(&pr->pr_mtx); 3249 return (0); 3250 } 3251 if (pr->pr_ip6 == NULL) { 3252 mtx_unlock(&pr->pr_mtx); 3253 return (EAFNOSUPPORT); 3254 } 3255 3256 if (IN6_IS_ADDR_LOOPBACK(ia6)) { 3257 bcopy(&pr->pr_ip6[0], ia6, sizeof(struct in6_addr)); 3258 mtx_unlock(&pr->pr_mtx); 3259 return (0); 3260 } 3261 3262 /* 3263 * Return success because nothing had to be changed. 3264 */ 3265 mtx_unlock(&pr->pr_mtx); 3266 return (0); 3267} 3268 3269/* 3270 * Check if given address belongs to the jail referenced by cred/prison. 3271 * 3272 * Returns 0 if jail doesn't restrict IPv6 or if address belongs to jail, 3273 * EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if the jail 3274 * doesn't allow IPv6. 3275 */ 3276static int 3277_prison_check_ip6(struct prison *pr, struct in6_addr *ia6) 3278{ 3279 int i, a, z, d; 3280 3281 /* 3282 * Check the primary IP. 3283 */ 3284 if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[0], ia6)) 3285 return (0); 3286 3287 /* 3288 * All the other IPs are sorted so we can do a binary search. 3289 */ 3290 a = 0; 3291 z = pr->pr_ip6s - 2; 3292 while (a <= z) { 3293 i = (a + z) / 2; 3294 d = qcmp_v6(&pr->pr_ip6[i+1], ia6); 3295 if (d > 0) 3296 z = i - 1; 3297 else if (d < 0) 3298 a = i + 1; 3299 else 3300 return (0); 3301 } 3302 3303 return (EADDRNOTAVAIL); 3304} 3305 3306int 3307prison_check_ip6(struct ucred *cred, struct in6_addr *ia6) 3308{ 3309 struct prison *pr; 3310 int error; 3311 3312 KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); 3313 KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__)); 3314 3315 pr = cred->cr_prison; 3316 if (!(pr->pr_flags & PR_IP6)) 3317 return (0); 3318 mtx_lock(&pr->pr_mtx); 3319 if (!(pr->pr_flags & PR_IP6)) { 3320 mtx_unlock(&pr->pr_mtx); 3321 return (0); 3322 } 3323 if (pr->pr_ip6 == NULL) { 3324 mtx_unlock(&pr->pr_mtx); 3325 return (EAFNOSUPPORT); 3326 } 3327 3328 error = _prison_check_ip6(pr, ia6); 3329 mtx_unlock(&pr->pr_mtx); 3330 return (error); 3331} 3332#endif 3333 3334/* 3335 * Check if a jail supports the given address family. 3336 * 3337 * Returns 0 if not jailed or the address family is supported, EAFNOSUPPORT 3338 * if not. 3339 */ 3340int 3341prison_check_af(struct ucred *cred, int af) 3342{ 3343 struct prison *pr; 3344 int error; 3345 3346 KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); 3347 3348 pr = cred->cr_prison; 3349#ifdef VIMAGE 3350 /* Prisons with their own network stack are not limited. */ 3351 if (prison_owns_vnet(cred)) 3352 return (0); 3353#endif 3354 3355 error = 0; 3356 switch (af) 3357 { 3358#ifdef INET 3359 case AF_INET: 3360 if (pr->pr_flags & PR_IP4) 3361 { 3362 mtx_lock(&pr->pr_mtx); 3363 if ((pr->pr_flags & PR_IP4) && pr->pr_ip4 == NULL) 3364 error = EAFNOSUPPORT; 3365 mtx_unlock(&pr->pr_mtx); 3366 } 3367 break; 3368#endif 3369#ifdef INET6 3370 case AF_INET6: 3371 if (pr->pr_flags & PR_IP6) 3372 { 3373 mtx_lock(&pr->pr_mtx); 3374 if ((pr->pr_flags & PR_IP6) && pr->pr_ip6 == NULL) 3375 error = EAFNOSUPPORT; 3376 mtx_unlock(&pr->pr_mtx); 3377 } 3378 break; 3379#endif 3380 case AF_LOCAL: 3381 case AF_ROUTE: 3382 break; 3383 default: 3384 if (!(pr->pr_allow & PR_ALLOW_SOCKET_AF)) 3385 error = EAFNOSUPPORT; 3386 } 3387 return (error); 3388} 3389 3390/* 3391 * Check if given address belongs to the jail referenced by cred (wrapper to 3392 * prison_check_ip[46]). 3393 * 3394 * Returns 0 if jail doesn't restrict the address family or if address belongs 3395 * to jail, EADDRNOTAVAIL if the address doesn't belong, or EAFNOSUPPORT if 3396 * the jail doesn't allow the address family. IPv4 Address passed in in NBO. 3397 */ 3398int 3399prison_if(struct ucred *cred, struct sockaddr *sa) 3400{ 3401#ifdef INET 3402 struct sockaddr_in *sai; 3403#endif 3404#ifdef INET6 3405 struct sockaddr_in6 *sai6; 3406#endif 3407 int error; 3408 3409 KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); 3410 KASSERT(sa != NULL, ("%s: sa is NULL", __func__)); 3411 3412#ifdef VIMAGE 3413 if (prison_owns_vnet(cred)) 3414 return (0); 3415#endif 3416 3417 error = 0; 3418 switch (sa->sa_family) 3419 { 3420#ifdef INET 3421 case AF_INET: 3422 sai = (struct sockaddr_in *)sa; 3423 error = prison_check_ip4(cred, &sai->sin_addr); 3424 break; 3425#endif 3426#ifdef INET6 3427 case AF_INET6: 3428 sai6 = (struct sockaddr_in6 *)sa; 3429 error = prison_check_ip6(cred, &sai6->sin6_addr); 3430 break; 3431#endif 3432 default: 3433 if (!(cred->cr_prison->pr_allow & PR_ALLOW_SOCKET_AF)) 3434 error = EAFNOSUPPORT; 3435 } 3436 return (error); 3437} 3438 3439/* 3440 * Return 0 if jails permit p1 to frob p2, otherwise ESRCH. 3441 */ 3442int 3443prison_check(struct ucred *cred1, struct ucred *cred2) 3444{ 3445 3446 return ((cred1->cr_prison == cred2->cr_prison || 3447 prison_ischild(cred1->cr_prison, cred2->cr_prison)) ? 0 : ESRCH); 3448} 3449 3450/* 3451 * Return 1 if p2 is a child of p1, otherwise 0. 3452 */ 3453int 3454prison_ischild(struct prison *pr1, struct prison *pr2) 3455{ 3456 3457 for (pr2 = pr2->pr_parent; pr2 != NULL; pr2 = pr2->pr_parent) 3458 if (pr1 == pr2) 3459 return (1); 3460 return (0); 3461} 3462 3463/* 3464 * Return 1 if the passed credential is in a jail, otherwise 0. 3465 */ 3466int 3467jailed(struct ucred *cred) 3468{ 3469 3470 return (cred->cr_prison != &prison0); 3471} 3472 3473/* 3474 * Return 1 if the passed credential is in a jail and that jail does not 3475 * have its own virtual network stack, otherwise 0. 3476 */ 3477int 3478jailed_without_vnet(struct ucred *cred) 3479{ 3480 3481 if (!jailed(cred)) 3482 return (0); 3483#ifdef VIMAGE 3484 if (prison_owns_vnet(cred)) 3485 return (0); 3486#endif 3487 3488 return (1); 3489} 3490 3491/* 3492 * Return the correct hostname (domainname, et al) for the passed credential. 3493 */ 3494void 3495getcredhostname(struct ucred *cred, char *buf, size_t size) 3496{ 3497 struct prison *pr; 3498 3499 /* 3500 * A NULL credential can be used to shortcut to the physical 3501 * system's hostname. 3502 */ 3503 pr = (cred != NULL) ? cred->cr_prison : &prison0; 3504 mtx_lock(&pr->pr_mtx); 3505 strlcpy(buf, pr->pr_hostname, size); 3506 mtx_unlock(&pr->pr_mtx); 3507} 3508 3509void 3510getcreddomainname(struct ucred *cred, char *buf, size_t size) 3511{ 3512 3513 mtx_lock(&cred->cr_prison->pr_mtx); 3514 strlcpy(buf, cred->cr_prison->pr_domainname, size); 3515 mtx_unlock(&cred->cr_prison->pr_mtx); 3516} 3517 3518void 3519getcredhostuuid(struct ucred *cred, char *buf, size_t size) 3520{ 3521 3522 mtx_lock(&cred->cr_prison->pr_mtx); 3523 strlcpy(buf, cred->cr_prison->pr_hostuuid, size); 3524 mtx_unlock(&cred->cr_prison->pr_mtx); 3525} 3526 3527void 3528getcredhostid(struct ucred *cred, unsigned long *hostid) 3529{ 3530 3531 mtx_lock(&cred->cr_prison->pr_mtx); 3532 *hostid = cred->cr_prison->pr_hostid; 3533 mtx_unlock(&cred->cr_prison->pr_mtx); 3534} 3535 3536#ifdef VIMAGE 3537/* 3538 * Determine whether the prison represented by cred owns 3539 * its vnet rather than having it inherited. 3540 * 3541 * Returns 1 in case the prison owns the vnet, 0 otherwise. 3542 */ 3543int 3544prison_owns_vnet(struct ucred *cred) 3545{ 3546 3547 /* 3548 * vnets cannot be added/removed after jail creation, 3549 * so no need to lock here. 3550 */ 3551 return (cred->cr_prison->pr_flags & PR_VNET ? 1 : 0); 3552} 3553#endif 3554 3555/* 3556 * Determine whether the subject represented by cred can "see" 3557 * status of a mount point. 3558 * Returns: 0 for permitted, ENOENT otherwise. 3559 * XXX: This function should be called cr_canseemount() and should be 3560 * placed in kern_prot.c. 3561 */ 3562int 3563prison_canseemount(struct ucred *cred, struct mount *mp) 3564{ 3565 struct prison *pr; 3566 struct statfs *sp; 3567 size_t len; 3568 3569 pr = cred->cr_prison; 3570 if (pr->pr_enforce_statfs == 0) 3571 return (0); 3572 if (pr->pr_root->v_mount == mp) 3573 return (0); 3574 if (pr->pr_enforce_statfs == 2) 3575 return (ENOENT); 3576 /* 3577 * If jail's chroot directory is set to "/" we should be able to see 3578 * all mount-points from inside a jail. 3579 * This is ugly check, but this is the only situation when jail's 3580 * directory ends with '/'. 3581 */ 3582 if (strcmp(pr->pr_path, "/") == 0) 3583 return (0); 3584 len = strlen(pr->pr_path); 3585 sp = &mp->mnt_stat; 3586 if (strncmp(pr->pr_path, sp->f_mntonname, len) != 0) 3587 return (ENOENT); 3588 /* 3589 * Be sure that we don't have situation where jail's root directory 3590 * is "/some/path" and mount point is "/some/pathpath". 3591 */ 3592 if (sp->f_mntonname[len] != '\0' && sp->f_mntonname[len] != '/') 3593 return (ENOENT); 3594 return (0); 3595} 3596 3597void 3598prison_enforce_statfs(struct ucred *cred, struct mount *mp, struct statfs *sp) 3599{ 3600 char jpath[MAXPATHLEN]; 3601 struct prison *pr; 3602 size_t len; 3603 3604 pr = cred->cr_prison; 3605 if (pr->pr_enforce_statfs == 0) 3606 return; 3607 if (prison_canseemount(cred, mp) != 0) { 3608 bzero(sp->f_mntonname, sizeof(sp->f_mntonname)); 3609 strlcpy(sp->f_mntonname, "[restricted]", 3610 sizeof(sp->f_mntonname)); 3611 return; 3612 } 3613 if (pr->pr_root->v_mount == mp) { 3614 /* 3615 * Clear current buffer data, so we are sure nothing from 3616 * the valid path left there. 3617 */ 3618 bzero(sp->f_mntonname, sizeof(sp->f_mntonname)); 3619 *sp->f_mntonname = '/'; 3620 return; 3621 } 3622 /* 3623 * If jail's chroot directory is set to "/" we should be able to see 3624 * all mount-points from inside a jail. 3625 */ 3626 if (strcmp(pr->pr_path, "/") == 0) 3627 return; 3628 len = strlen(pr->pr_path); 3629 strlcpy(jpath, sp->f_mntonname + len, sizeof(jpath)); 3630 /* 3631 * Clear current buffer data, so we are sure nothing from 3632 * the valid path left there. 3633 */ 3634 bzero(sp->f_mntonname, sizeof(sp->f_mntonname)); 3635 if (*jpath == '\0') { 3636 /* Should never happen. */ 3637 *sp->f_mntonname = '/'; 3638 } else { 3639 strlcpy(sp->f_mntonname, jpath, sizeof(sp->f_mntonname)); 3640 } 3641} 3642 3643/* 3644 * Check with permission for a specific privilege is granted within jail. We 3645 * have a specific list of accepted privileges; the rest are denied. 3646 */ 3647int 3648prison_priv_check(struct ucred *cred, int priv) 3649{ 3650 3651 if (!jailed(cred)) 3652 return (0); 3653 3654#ifdef VIMAGE 3655 /* 3656 * Privileges specific to prisons with a virtual network stack. 3657 * There might be a duplicate entry here in case the privilege 3658 * is only granted conditionally in the legacy jail case. 3659 */ 3660 switch (priv) { 3661#ifdef notyet 3662 /* 3663 * NFS-specific privileges. 3664 */ 3665 case PRIV_NFS_DAEMON: 3666 case PRIV_NFS_LOCKD: 3667#endif 3668 /* 3669 * Network stack privileges. 3670 */ 3671 case PRIV_NET_BRIDGE: 3672 case PRIV_NET_GRE: 3673 case PRIV_NET_BPF: 3674 case PRIV_NET_RAW: /* Dup, cond. in legacy jail case. */ 3675 case PRIV_NET_ROUTE: 3676 case PRIV_NET_TAP: 3677 case PRIV_NET_SETIFMTU: 3678 case PRIV_NET_SETIFFLAGS: 3679 case PRIV_NET_SETIFCAP: 3680 case PRIV_NET_SETIFDESCR: 3681 case PRIV_NET_SETIFNAME : 3682 case PRIV_NET_SETIFMETRIC: 3683 case PRIV_NET_SETIFPHYS: 3684 case PRIV_NET_SETIFMAC: 3685 case PRIV_NET_ADDMULTI: 3686 case PRIV_NET_DELMULTI: 3687 case PRIV_NET_HWIOCTL: 3688 case PRIV_NET_SETLLADDR: 3689 case PRIV_NET_ADDIFGROUP: 3690 case PRIV_NET_DELIFGROUP: 3691 case PRIV_NET_IFCREATE: 3692 case PRIV_NET_IFDESTROY: 3693 case PRIV_NET_ADDIFADDR: 3694 case PRIV_NET_DELIFADDR: 3695 case PRIV_NET_LAGG: 3696 case PRIV_NET_GIF: 3697 case PRIV_NET_SETIFVNET: 3698 case PRIV_NET_SETIFFIB: 3699 3700 /* 3701 * 802.11-related privileges. 3702 */ 3703 case PRIV_NET80211_GETKEY: 3704#ifdef notyet 3705 case PRIV_NET80211_MANAGE: /* XXX-BZ discuss with sam@ */ 3706#endif 3707 3708#ifdef notyet 3709 /* 3710 * AppleTalk privileges. 3711 */ 3712 case PRIV_NETATALK_RESERVEDPORT: 3713 3714 /* 3715 * ATM privileges. 3716 */ 3717 case PRIV_NETATM_CFG: 3718 case PRIV_NETATM_ADD: 3719 case PRIV_NETATM_DEL: 3720 case PRIV_NETATM_SET: 3721 3722 /* 3723 * Bluetooth privileges. 3724 */ 3725 case PRIV_NETBLUETOOTH_RAW: 3726#endif 3727 3728 /* 3729 * Netgraph and netgraph module privileges. 3730 */ 3731 case PRIV_NETGRAPH_CONTROL: 3732#ifdef notyet 3733 case PRIV_NETGRAPH_TTY: 3734#endif 3735 3736 /* 3737 * IPv4 and IPv6 privileges. 3738 */ 3739 case PRIV_NETINET_IPFW: 3740 case PRIV_NETINET_DIVERT: 3741 case PRIV_NETINET_PF: 3742 case PRIV_NETINET_DUMMYNET: 3743 case PRIV_NETINET_CARP: 3744 case PRIV_NETINET_MROUTE: 3745 case PRIV_NETINET_RAW: 3746 case PRIV_NETINET_ADDRCTRL6: 3747 case PRIV_NETINET_ND6: 3748 case PRIV_NETINET_SCOPE6: 3749 case PRIV_NETINET_ALIFETIME6: 3750 case PRIV_NETINET_IPSEC: 3751 case PRIV_NETINET_BINDANY: 3752 3753#ifdef notyet 3754 /* 3755 * IPX/SPX privileges. 3756 */ 3757 case PRIV_NETIPX_RESERVEDPORT: 3758 case PRIV_NETIPX_RAW: 3759 3760 /* 3761 * NCP privileges. 3762 */ 3763 case PRIV_NETNCP: 3764 3765 /* 3766 * SMB privileges. 3767 */ 3768 case PRIV_NETSMB: 3769#endif 3770 3771 /* 3772 * No default: or deny here. 3773 * In case of no permit fall through to next switch(). 3774 */ 3775 if (cred->cr_prison->pr_flags & PR_VNET) 3776 return (0); 3777 } 3778#endif /* VIMAGE */ 3779 3780 switch (priv) { 3781 3782 /* 3783 * Allow ktrace privileges for root in jail. 3784 */ 3785 case PRIV_KTRACE: 3786 3787#if 0 3788 /* 3789 * Allow jailed processes to configure audit identity and 3790 * submit audit records (login, etc). In the future we may 3791 * want to further refine the relationship between audit and 3792 * jail. 3793 */ 3794 case PRIV_AUDIT_GETAUDIT: 3795 case PRIV_AUDIT_SETAUDIT: 3796 case PRIV_AUDIT_SUBMIT: 3797#endif 3798 3799 /* 3800 * Allow jailed processes to manipulate process UNIX 3801 * credentials in any way they see fit. 3802 */ 3803 case PRIV_CRED_SETUID: 3804 case PRIV_CRED_SETEUID: 3805 case PRIV_CRED_SETGID: 3806 case PRIV_CRED_SETEGID: 3807 case PRIV_CRED_SETGROUPS: 3808 case PRIV_CRED_SETREUID: 3809 case PRIV_CRED_SETREGID: 3810 case PRIV_CRED_SETRESUID: 3811 case PRIV_CRED_SETRESGID: 3812 3813 /* 3814 * Jail implements visibility constraints already, so allow 3815 * jailed root to override uid/gid-based constraints. 3816 */ 3817 case PRIV_SEEOTHERGIDS: 3818 case PRIV_SEEOTHERUIDS: 3819 3820 /* 3821 * Jail implements inter-process debugging limits already, so 3822 * allow jailed root various debugging privileges. 3823 */ 3824 case PRIV_DEBUG_DIFFCRED: 3825 case PRIV_DEBUG_SUGID: 3826 case PRIV_DEBUG_UNPRIV: 3827 3828 /* 3829 * Allow jail to set various resource limits and login 3830 * properties, and for now, exceed process resource limits. 3831 */ 3832 case PRIV_PROC_LIMIT: 3833 case PRIV_PROC_SETLOGIN: 3834 case PRIV_PROC_SETRLIMIT: 3835 3836 /* 3837 * System V and POSIX IPC privileges are granted in jail. 3838 */ 3839 case PRIV_IPC_READ: 3840 case PRIV_IPC_WRITE: 3841 case PRIV_IPC_ADMIN: 3842 case PRIV_IPC_MSGSIZE: 3843 case PRIV_MQ_ADMIN: 3844 3845 /* 3846 * Jail operations within a jail work on child jails. 3847 */ 3848 case PRIV_JAIL_ATTACH: 3849 case PRIV_JAIL_SET: 3850 case PRIV_JAIL_REMOVE: 3851 3852 /* 3853 * Jail implements its own inter-process limits, so allow 3854 * root processes in jail to change scheduling on other 3855 * processes in the same jail. Likewise for signalling. 3856 */ 3857 case PRIV_SCHED_DIFFCRED: 3858 case PRIV_SCHED_CPUSET: 3859 case PRIV_SIGNAL_DIFFCRED: 3860 case PRIV_SIGNAL_SUGID: 3861 3862 /* 3863 * Allow jailed processes to write to sysctls marked as jail 3864 * writable. 3865 */ 3866 case PRIV_SYSCTL_WRITEJAIL: 3867 3868 /* 3869 * Allow root in jail to manage a variety of quota 3870 * properties. These should likely be conditional on a 3871 * configuration option. 3872 */ 3873 case PRIV_VFS_GETQUOTA: 3874 case PRIV_VFS_SETQUOTA: 3875 3876 /* 3877 * Since Jail relies on chroot() to implement file system 3878 * protections, grant many VFS privileges to root in jail. 3879 * Be careful to exclude mount-related and NFS-related 3880 * privileges. 3881 */ 3882 case PRIV_VFS_READ: 3883 case PRIV_VFS_WRITE: 3884 case PRIV_VFS_ADMIN: 3885 case PRIV_VFS_EXEC: 3886 case PRIV_VFS_LOOKUP: 3887 case PRIV_VFS_BLOCKRESERVE: /* XXXRW: Slightly surprising. */ 3888 case PRIV_VFS_CHFLAGS_DEV: 3889 case PRIV_VFS_CHOWN: 3890 case PRIV_VFS_CHROOT: 3891 case PRIV_VFS_RETAINSUGID: 3892 case PRIV_VFS_FCHROOT: 3893 case PRIV_VFS_LINK: 3894 case PRIV_VFS_SETGID: 3895 case PRIV_VFS_STAT: 3896 case PRIV_VFS_STICKYFILE: 3897 return (0); 3898 3899 /* 3900 * Depending on the global setting, allow privilege of 3901 * setting system flags. 3902 */ 3903 case PRIV_VFS_SYSFLAGS: 3904 if (cred->cr_prison->pr_allow & PR_ALLOW_CHFLAGS) 3905 return (0); 3906 else 3907 return (EPERM); 3908 3909 /* 3910 * Depending on the global setting, allow privilege of 3911 * mounting/unmounting file systems. 3912 */ 3913 case PRIV_VFS_MOUNT: 3914 case PRIV_VFS_UNMOUNT: 3915 case PRIV_VFS_MOUNT_NONUSER: 3916 case PRIV_VFS_MOUNT_OWNER: 3917 if (cred->cr_prison->pr_allow & PR_ALLOW_MOUNT && 3918 cred->cr_prison->pr_enforce_statfs < 2) 3919 return (0); 3920 else 3921 return (EPERM); 3922 3923 /* 3924 * Allow jailed root to bind reserved ports and reuse in-use 3925 * ports. 3926 */ 3927 case PRIV_NETINET_RESERVEDPORT: 3928 case PRIV_NETINET_REUSEPORT: 3929 return (0); 3930 3931 /* 3932 * Allow jailed root to set certian IPv4/6 (option) headers. 3933 */ 3934 case PRIV_NETINET_SETHDROPTS: 3935 return (0); 3936 3937 /* 3938 * Conditionally allow creating raw sockets in jail. 3939 */ 3940 case PRIV_NETINET_RAW: 3941 if (cred->cr_prison->pr_allow & PR_ALLOW_RAW_SOCKETS) 3942 return (0); 3943 else 3944 return (EPERM); 3945 3946 /* 3947 * Since jail implements its own visibility limits on netstat 3948 * sysctls, allow getcred. This allows identd to work in 3949 * jail. 3950 */ 3951 case PRIV_NETINET_GETCRED: 3952 return (0); 3953 3954 /* 3955 * Allow jailed root to set loginclass. 3956 */ 3957 case PRIV_PROC_SETLOGINCLASS: 3958 return (0); 3959 3960 default: 3961 /* 3962 * In all remaining cases, deny the privilege request. This 3963 * includes almost all network privileges, many system 3964 * configuration privileges. 3965 */ 3966 return (EPERM); 3967 } 3968} 3969 3970/* 3971 * Return the part of pr2's name that is relative to pr1, or the whole name 3972 * if it does not directly follow. 3973 */ 3974 3975char * 3976prison_name(struct prison *pr1, struct prison *pr2) 3977{ 3978 char *name; 3979 3980 /* Jails see themselves as "0" (if they see themselves at all). */ 3981 if (pr1 == pr2) 3982 return "0"; 3983 name = pr2->pr_name; 3984 if (prison_ischild(pr1, pr2)) { 3985 /* 3986 * pr1 isn't locked (and allprison_lock may not be either) 3987 * so its length can't be counted on. But the number of dots 3988 * can be counted on - and counted. 3989 */ 3990 for (; pr1 != &prison0; pr1 = pr1->pr_parent) 3991 name = strchr(name, '.') + 1; 3992 } 3993 return (name); 3994} 3995 3996/* 3997 * Return the part of pr2's path that is relative to pr1, or the whole path 3998 * if it does not directly follow. 3999 */ 4000static char * 4001prison_path(struct prison *pr1, struct prison *pr2) 4002{ 4003 char *path1, *path2; 4004 int len1; 4005 4006 path1 = pr1->pr_path; 4007 path2 = pr2->pr_path; 4008 if (!strcmp(path1, "/")) 4009 return (path2); 4010 len1 = strlen(path1); 4011 if (strncmp(path1, path2, len1)) 4012 return (path2); 4013 if (path2[len1] == '\0') 4014 return "/"; 4015 if (path2[len1] == '/') 4016 return (path2 + len1); 4017 return (path2); 4018} 4019 4020 4021/* 4022 * Jail-related sysctls. 4023 */ 4024static SYSCTL_NODE(_security, OID_AUTO, jail, CTLFLAG_RW, 0, 4025 "Jails"); 4026 4027static int 4028sysctl_jail_list(SYSCTL_HANDLER_ARGS) 4029{ 4030 struct xprison *xp; 4031 struct prison *pr, *cpr; 4032#ifdef INET 4033 struct in_addr *ip4 = NULL; 4034 int ip4s = 0; 4035#endif 4036#ifdef INET6 4037 struct in6_addr *ip6 = NULL; 4038 int ip6s = 0; 4039#endif 4040 int descend, error; 4041 4042 xp = malloc(sizeof(*xp), M_TEMP, M_WAITOK); 4043 pr = req->td->td_ucred->cr_prison; 4044 error = 0; 4045 sx_slock(&allprison_lock); 4046 FOREACH_PRISON_DESCENDANT(pr, cpr, descend) { 4047#if defined(INET) || defined(INET6) 4048 again: 4049#endif 4050 mtx_lock(&cpr->pr_mtx); 4051#ifdef INET 4052 if (cpr->pr_ip4s > 0) { 4053 if (ip4s < cpr->pr_ip4s) { 4054 ip4s = cpr->pr_ip4s; 4055 mtx_unlock(&cpr->pr_mtx); 4056 ip4 = realloc(ip4, ip4s * 4057 sizeof(struct in_addr), M_TEMP, M_WAITOK); 4058 goto again; 4059 } 4060 bcopy(cpr->pr_ip4, ip4, 4061 cpr->pr_ip4s * sizeof(struct in_addr)); 4062 } 4063#endif 4064#ifdef INET6 4065 if (cpr->pr_ip6s > 0) { 4066 if (ip6s < cpr->pr_ip6s) { 4067 ip6s = cpr->pr_ip6s; 4068 mtx_unlock(&cpr->pr_mtx); 4069 ip6 = realloc(ip6, ip6s * 4070 sizeof(struct in6_addr), M_TEMP, M_WAITOK); 4071 goto again; 4072 } 4073 bcopy(cpr->pr_ip6, ip6, 4074 cpr->pr_ip6s * sizeof(struct in6_addr)); 4075 } 4076#endif 4077 if (cpr->pr_ref == 0) { 4078 mtx_unlock(&cpr->pr_mtx); 4079 continue; 4080 } 4081 bzero(xp, sizeof(*xp)); 4082 xp->pr_version = XPRISON_VERSION; 4083 xp->pr_id = cpr->pr_id; 4084 xp->pr_state = cpr->pr_uref > 0 4085 ? PRISON_STATE_ALIVE : PRISON_STATE_DYING; 4086 strlcpy(xp->pr_path, prison_path(pr, cpr), sizeof(xp->pr_path)); 4087 strlcpy(xp->pr_host, cpr->pr_hostname, sizeof(xp->pr_host)); 4088 strlcpy(xp->pr_name, prison_name(pr, cpr), sizeof(xp->pr_name)); 4089#ifdef INET 4090 xp->pr_ip4s = cpr->pr_ip4s; 4091#endif 4092#ifdef INET6 4093 xp->pr_ip6s = cpr->pr_ip6s; 4094#endif 4095 mtx_unlock(&cpr->pr_mtx); 4096 error = SYSCTL_OUT(req, xp, sizeof(*xp)); 4097 if (error) 4098 break; 4099#ifdef INET 4100 if (xp->pr_ip4s > 0) { 4101 error = SYSCTL_OUT(req, ip4, 4102 xp->pr_ip4s * sizeof(struct in_addr)); 4103 if (error) 4104 break; 4105 } 4106#endif 4107#ifdef INET6 4108 if (xp->pr_ip6s > 0) { 4109 error = SYSCTL_OUT(req, ip6, 4110 xp->pr_ip6s * sizeof(struct in6_addr)); 4111 if (error) 4112 break; 4113 } 4114#endif 4115 } 4116 sx_sunlock(&allprison_lock); 4117 free(xp, M_TEMP); 4118#ifdef INET 4119 free(ip4, M_TEMP); 4120#endif 4121#ifdef INET6 4122 free(ip6, M_TEMP); 4123#endif 4124 return (error); 4125} 4126 4127SYSCTL_OID(_security_jail, OID_AUTO, list, 4128 CTLTYPE_STRUCT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0, 4129 sysctl_jail_list, "S", "List of active jails"); 4130 4131static int 4132sysctl_jail_jailed(SYSCTL_HANDLER_ARGS) 4133{ 4134 int error, injail; 4135 4136 injail = jailed(req->td->td_ucred); 4137 error = SYSCTL_OUT(req, &injail, sizeof(injail)); 4138 4139 return (error); 4140} 4141 4142SYSCTL_PROC(_security_jail, OID_AUTO, jailed, 4143 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0, 4144 sysctl_jail_jailed, "I", "Process in jail?"); 4145 4146#if defined(INET) || defined(INET6) 4147SYSCTL_UINT(_security_jail, OID_AUTO, jail_max_af_ips, CTLFLAG_RW, 4148 &jail_max_af_ips, 0, 4149 "Number of IP addresses a jail may have at most per address family"); 4150#endif 4151 4152/* 4153 * Default parameters for jail(2) compatability. For historical reasons, 4154 * the sysctl names have varying similarity to the parameter names. Prisons 4155 * just see their own parameters, and can't change them. 4156 */ 4157static int 4158sysctl_jail_default_allow(SYSCTL_HANDLER_ARGS) 4159{ 4160 struct prison *pr; 4161 int allow, error, i; 4162 4163 pr = req->td->td_ucred->cr_prison; 4164 allow = (pr == &prison0) ? jail_default_allow : pr->pr_allow; 4165 4166 /* Get the current flag value, and convert it to a boolean. */ 4167 i = (allow & arg2) ? 1 : 0; 4168 if (arg1 != NULL) 4169 i = !i; 4170 error = sysctl_handle_int(oidp, &i, 0, req); 4171 if (error || !req->newptr) 4172 return (error); 4173 i = i ? arg2 : 0; 4174 if (arg1 != NULL) 4175 i ^= arg2; 4176 /* 4177 * The sysctls don't have CTLFLAGS_PRISON, so assume prison0 4178 * for writing. 4179 */ 4180 mtx_lock(&prison0.pr_mtx); 4181 jail_default_allow = (jail_default_allow & ~arg2) | i; 4182 mtx_unlock(&prison0.pr_mtx); 4183 return (0); 4184} 4185 4186SYSCTL_PROC(_security_jail, OID_AUTO, set_hostname_allowed, 4187 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 4188 NULL, PR_ALLOW_SET_HOSTNAME, sysctl_jail_default_allow, "I", 4189 "Processes in jail can set their hostnames"); 4190SYSCTL_PROC(_security_jail, OID_AUTO, socket_unixiproute_only, 4191 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 4192 (void *)1, PR_ALLOW_SOCKET_AF, sysctl_jail_default_allow, "I", 4193 "Processes in jail are limited to creating UNIX/IP/route sockets only"); 4194SYSCTL_PROC(_security_jail, OID_AUTO, sysvipc_allowed, 4195 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 4196 NULL, PR_ALLOW_SYSVIPC, sysctl_jail_default_allow, "I", 4197 "Processes in jail can use System V IPC primitives"); 4198SYSCTL_PROC(_security_jail, OID_AUTO, allow_raw_sockets, 4199 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 4200 NULL, PR_ALLOW_RAW_SOCKETS, sysctl_jail_default_allow, "I", 4201 "Prison root can create raw sockets"); 4202SYSCTL_PROC(_security_jail, OID_AUTO, chflags_allowed, 4203 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 4204 NULL, PR_ALLOW_CHFLAGS, sysctl_jail_default_allow, "I", 4205 "Processes in jail can alter system file flags"); 4206SYSCTL_PROC(_security_jail, OID_AUTO, mount_allowed, 4207 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 4208 NULL, PR_ALLOW_MOUNT, sysctl_jail_default_allow, "I", 4209 "Processes in jail can mount/unmount jail-friendly file systems"); 4210SYSCTL_PROC(_security_jail, OID_AUTO, mount_devfs_allowed, 4211 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 4212 NULL, PR_ALLOW_MOUNT_DEVFS, sysctl_jail_default_allow, "I", 4213 "Processes in jail can mount the devfs file system"); 4214SYSCTL_PROC(_security_jail, OID_AUTO, mount_nullfs_allowed, 4215 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 4216 NULL, PR_ALLOW_MOUNT_NULLFS, sysctl_jail_default_allow, "I", 4217 "Processes in jail can mount the nullfs file system"); 4218SYSCTL_PROC(_security_jail, OID_AUTO, mount_procfs_allowed, 4219 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 4220 NULL, PR_ALLOW_MOUNT_PROCFS, sysctl_jail_default_allow, "I", 4221 "Processes in jail can mount the procfs file system"); 4222SYSCTL_PROC(_security_jail, OID_AUTO, mount_zfs_allowed, 4223 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 4224 NULL, PR_ALLOW_MOUNT_ZFS, sysctl_jail_default_allow, "I", 4225 "Processes in jail can mount the zfs file system"); 4226 4227static int 4228sysctl_jail_default_level(SYSCTL_HANDLER_ARGS) 4229{ 4230 struct prison *pr; 4231 int level, error; 4232 4233 pr = req->td->td_ucred->cr_prison; 4234 level = (pr == &prison0) ? *(int *)arg1 : *(int *)((char *)pr + arg2); 4235 error = sysctl_handle_int(oidp, &level, 0, req); 4236 if (error || !req->newptr) 4237 return (error); 4238 *(int *)arg1 = level; 4239 return (0); 4240} 4241 4242SYSCTL_PROC(_security_jail, OID_AUTO, enforce_statfs, 4243 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 4244 &jail_default_enforce_statfs, offsetof(struct prison, pr_enforce_statfs), 4245 sysctl_jail_default_level, "I", 4246 "Processes in jail cannot see all mounted file systems"); 4247 4248SYSCTL_PROC(_security_jail, OID_AUTO, devfs_ruleset, 4249 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, 4250 &jail_default_devfs_rsnum, offsetof(struct prison, pr_devfs_rsnum), 4251 sysctl_jail_default_level, "I", 4252 "Ruleset for the devfs filesystem in jail"); 4253 4254/* 4255 * Nodes to describe jail parameters. Maximum length of string parameters 4256 * is returned in the string itself, and the other parameters exist merely 4257 * to make themselves and their types known. 4258 */ 4259SYSCTL_NODE(_security_jail, OID_AUTO, param, CTLFLAG_RW, 0, 4260 "Jail parameters"); 4261 4262int 4263sysctl_jail_param(SYSCTL_HANDLER_ARGS) 4264{ 4265 int i; 4266 long l; 4267 size_t s; 4268 char numbuf[12]; 4269 4270 switch (oidp->oid_kind & CTLTYPE) 4271 { 4272 case CTLTYPE_LONG: 4273 case CTLTYPE_ULONG: 4274 l = 0; 4275#ifdef SCTL_MASK32 4276 if (!(req->flags & SCTL_MASK32)) 4277#endif 4278 return (SYSCTL_OUT(req, &l, sizeof(l))); 4279 case CTLTYPE_INT: 4280 case CTLTYPE_UINT: 4281 i = 0; 4282 return (SYSCTL_OUT(req, &i, sizeof(i))); 4283 case CTLTYPE_STRING: 4284 snprintf(numbuf, sizeof(numbuf), "%jd", (intmax_t)arg2); 4285 return 4286 (sysctl_handle_string(oidp, numbuf, sizeof(numbuf), req)); 4287 case CTLTYPE_STRUCT: 4288 s = (size_t)arg2; 4289 return (SYSCTL_OUT(req, &s, sizeof(s))); 4290 } 4291 return (0); 4292} 4293 4294SYSCTL_JAIL_PARAM(, jid, CTLTYPE_INT | CTLFLAG_RDTUN, "I", "Jail ID"); 4295SYSCTL_JAIL_PARAM(, parent, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail parent ID"); 4296SYSCTL_JAIL_PARAM_STRING(, name, CTLFLAG_RW, MAXHOSTNAMELEN, "Jail name"); 4297SYSCTL_JAIL_PARAM_STRING(, path, CTLFLAG_RDTUN, MAXPATHLEN, "Jail root path"); 4298SYSCTL_JAIL_PARAM(, securelevel, CTLTYPE_INT | CTLFLAG_RW, 4299 "I", "Jail secure level"); 4300SYSCTL_JAIL_PARAM(, enforce_statfs, CTLTYPE_INT | CTLFLAG_RW, 4301 "I", "Jail cannot see all mounted file systems"); 4302SYSCTL_JAIL_PARAM(, devfs_ruleset, CTLTYPE_INT | CTLFLAG_RW, 4303 "I", "Ruleset for in-jail devfs mounts"); 4304SYSCTL_JAIL_PARAM(, persist, CTLTYPE_INT | CTLFLAG_RW, 4305 "B", "Jail persistence"); 4306#ifdef VIMAGE 4307SYSCTL_JAIL_PARAM(, vnet, CTLTYPE_INT | CTLFLAG_RDTUN, 4308 "E,jailsys", "Virtual network stack"); 4309#endif 4310SYSCTL_JAIL_PARAM(, dying, CTLTYPE_INT | CTLFLAG_RD, 4311 "B", "Jail is in the process of shutting down"); 4312 4313SYSCTL_JAIL_PARAM_NODE(children, "Number of child jails"); 4314SYSCTL_JAIL_PARAM(_children, cur, CTLTYPE_INT | CTLFLAG_RD, 4315 "I", "Current number of child jails"); 4316SYSCTL_JAIL_PARAM(_children, max, CTLTYPE_INT | CTLFLAG_RW, 4317 "I", "Maximum number of child jails"); 4318 4319SYSCTL_JAIL_PARAM_SYS_NODE(host, CTLFLAG_RW, "Jail host info"); 4320SYSCTL_JAIL_PARAM_STRING(_host, hostname, CTLFLAG_RW, MAXHOSTNAMELEN, 4321 "Jail hostname"); 4322SYSCTL_JAIL_PARAM_STRING(_host, domainname, CTLFLAG_RW, MAXHOSTNAMELEN, 4323 "Jail NIS domainname"); 4324SYSCTL_JAIL_PARAM_STRING(_host, hostuuid, CTLFLAG_RW, HOSTUUIDLEN, 4325 "Jail host UUID"); 4326SYSCTL_JAIL_PARAM(_host, hostid, CTLTYPE_ULONG | CTLFLAG_RW, 4327 "LU", "Jail host ID"); 4328 4329SYSCTL_JAIL_PARAM_NODE(cpuset, "Jail cpuset"); 4330SYSCTL_JAIL_PARAM(_cpuset, id, CTLTYPE_INT | CTLFLAG_RD, "I", "Jail cpuset ID"); 4331 4332#ifdef INET 4333SYSCTL_JAIL_PARAM_SYS_NODE(ip4, CTLFLAG_RDTUN, 4334 "Jail IPv4 address virtualization"); 4335SYSCTL_JAIL_PARAM_STRUCT(_ip4, addr, CTLFLAG_RW, sizeof(struct in_addr), 4336 "S,in_addr,a", "Jail IPv4 addresses"); 4337SYSCTL_JAIL_PARAM(_ip4, saddrsel, CTLTYPE_INT | CTLFLAG_RW, 4338 "B", "Do (not) use IPv4 source address selection rather than the " 4339 "primary jail IPv4 address."); 4340#endif 4341#ifdef INET6 4342SYSCTL_JAIL_PARAM_SYS_NODE(ip6, CTLFLAG_RDTUN, 4343 "Jail IPv6 address virtualization"); 4344SYSCTL_JAIL_PARAM_STRUCT(_ip6, addr, CTLFLAG_RW, sizeof(struct in6_addr), 4345 "S,in6_addr,a", "Jail IPv6 addresses"); 4346SYSCTL_JAIL_PARAM(_ip6, saddrsel, CTLTYPE_INT | CTLFLAG_RW, 4347 "B", "Do (not) use IPv6 source address selection rather than the " 4348 "primary jail IPv6 address."); 4349#endif 4350 4351SYSCTL_JAIL_PARAM_NODE(allow, "Jail permission flags"); 4352SYSCTL_JAIL_PARAM(_allow, set_hostname, CTLTYPE_INT | CTLFLAG_RW, 4353 "B", "Jail may set hostname"); 4354SYSCTL_JAIL_PARAM(_allow, sysvipc, CTLTYPE_INT | CTLFLAG_RW, 4355 "B", "Jail may use SYSV IPC"); 4356SYSCTL_JAIL_PARAM(_allow, raw_sockets, CTLTYPE_INT | CTLFLAG_RW, 4357 "B", "Jail may create raw sockets"); 4358SYSCTL_JAIL_PARAM(_allow, chflags, CTLTYPE_INT | CTLFLAG_RW, 4359 "B", "Jail may alter system file flags"); 4360SYSCTL_JAIL_PARAM(_allow, quotas, CTLTYPE_INT | CTLFLAG_RW, 4361 "B", "Jail may set file quotas"); 4362SYSCTL_JAIL_PARAM(_allow, socket_af, CTLTYPE_INT | CTLFLAG_RW, 4363 "B", "Jail may create sockets other than just UNIX/IPv4/IPv6/route"); 4364 4365SYSCTL_JAIL_PARAM_SUBNODE(allow, mount, "Jail mount/unmount permission flags"); 4366SYSCTL_JAIL_PARAM(_allow_mount, , CTLTYPE_INT | CTLFLAG_RW, 4367 "B", "Jail may mount/unmount jail-friendly file systems in general"); 4368SYSCTL_JAIL_PARAM(_allow_mount, devfs, CTLTYPE_INT | CTLFLAG_RW, 4369 "B", "Jail may mount the devfs file system"); 4370SYSCTL_JAIL_PARAM(_allow_mount, nullfs, CTLTYPE_INT | CTLFLAG_RW, 4371 "B", "Jail may mount the nullfs file system"); 4372SYSCTL_JAIL_PARAM(_allow_mount, procfs, CTLTYPE_INT | CTLFLAG_RW, 4373 "B", "Jail may mount the procfs file system"); 4374SYSCTL_JAIL_PARAM(_allow_mount, zfs, CTLTYPE_INT | CTLFLAG_RW, 4375 "B", "Jail may mount the zfs file system"); 4376 4377void 4378prison_racct_foreach(void (*callback)(struct racct *racct, 4379 void *arg2, void *arg3), void *arg2, void *arg3) 4380{ 4381 struct prison_racct *prr; 4382 4383 sx_slock(&allprison_lock); 4384 LIST_FOREACH(prr, &allprison_racct, prr_next) 4385 (callback)(prr->prr_racct, arg2, arg3); 4386 sx_sunlock(&allprison_lock); 4387} 4388 4389static struct prison_racct * 4390prison_racct_find_locked(const char *name) 4391{ 4392 struct prison_racct *prr; 4393 4394 sx_assert(&allprison_lock, SA_XLOCKED); 4395 4396 if (name[0] == '\0' || strlen(name) >= MAXHOSTNAMELEN) 4397 return (NULL); 4398 4399 LIST_FOREACH(prr, &allprison_racct, prr_next) { 4400 if (strcmp(name, prr->prr_name) != 0) 4401 continue; 4402 4403 /* Found prison_racct with a matching name? */ 4404 prison_racct_hold(prr); 4405 return (prr); 4406 } 4407 4408 /* Add new prison_racct. */ 4409 prr = malloc(sizeof(*prr), M_PRISON_RACCT, M_ZERO | M_WAITOK); 4410 racct_create(&prr->prr_racct); 4411 4412 strcpy(prr->prr_name, name); 4413 refcount_init(&prr->prr_refcount, 1); 4414 LIST_INSERT_HEAD(&allprison_racct, prr, prr_next); 4415 4416 return (prr); 4417} 4418 4419struct prison_racct * 4420prison_racct_find(const char *name) 4421{ 4422 struct prison_racct *prr; 4423 4424 sx_xlock(&allprison_lock); 4425 prr = prison_racct_find_locked(name); 4426 sx_xunlock(&allprison_lock); 4427 return (prr); 4428} 4429 4430void 4431prison_racct_hold(struct prison_racct *prr) 4432{ 4433 4434 refcount_acquire(&prr->prr_refcount); 4435} 4436 4437static void 4438prison_racct_free_locked(struct prison_racct *prr) 4439{ 4440 4441 sx_assert(&allprison_lock, SA_XLOCKED); 4442 4443 if (refcount_release(&prr->prr_refcount)) { 4444 racct_destroy(&prr->prr_racct); 4445 LIST_REMOVE(prr, prr_next); 4446 free(prr, M_PRISON_RACCT); 4447 } 4448} 4449 4450void 4451prison_racct_free(struct prison_racct *prr) 4452{ 4453 int old; 4454 4455 sx_assert(&allprison_lock, SA_UNLOCKED); 4456 4457 old = prr->prr_refcount; 4458 if (old > 1 && atomic_cmpset_int(&prr->prr_refcount, old, old - 1)) 4459 return; 4460 4461 sx_xlock(&allprison_lock); 4462 prison_racct_free_locked(prr); 4463 sx_xunlock(&allprison_lock); 4464} 4465 4466#ifdef RACCT 4467static void 4468prison_racct_attach(struct prison *pr) 4469{ 4470 struct prison_racct *prr; 4471 4472 sx_assert(&allprison_lock, SA_XLOCKED); 4473 4474 prr = prison_racct_find_locked(pr->pr_name); 4475 KASSERT(prr != NULL, ("cannot find prison_racct")); 4476 4477 pr->pr_prison_racct = prr; 4478} 4479 4480/* 4481 * Handle jail renaming. From the racct point of view, renaming means 4482 * moving from one prison_racct to another. 4483 */ 4484static void 4485prison_racct_modify(struct prison *pr) 4486{ 4487 struct proc *p; 4488 struct ucred *cred; 4489 struct prison_racct *oldprr; 4490 4491 sx_slock(&allproc_lock); 4492 sx_xlock(&allprison_lock); 4493 4494 if (strcmp(pr->pr_name, pr->pr_prison_racct->prr_name) == 0) 4495 return; 4496 4497 oldprr = pr->pr_prison_racct; 4498 pr->pr_prison_racct = NULL; 4499 4500 prison_racct_attach(pr); 4501 4502 /* 4503 * Move resource utilisation records. 4504 */ 4505 racct_move(pr->pr_prison_racct->prr_racct, oldprr->prr_racct); 4506 4507 /* 4508 * Force rctl to reattach rules to processes. 4509 */ 4510 FOREACH_PROC_IN_SYSTEM(p) { 4511 PROC_LOCK(p); 4512 cred = crhold(p->p_ucred); 4513 PROC_UNLOCK(p); 4514 racct_proc_ucred_changed(p, cred, cred); 4515 crfree(cred); 4516 } 4517 4518 sx_sunlock(&allproc_lock); 4519 prison_racct_free_locked(oldprr); 4520 sx_xunlock(&allprison_lock); 4521} 4522 4523static void 4524prison_racct_detach(struct prison *pr) 4525{ 4526 4527 sx_assert(&allprison_lock, SA_UNLOCKED); 4528 4529 prison_racct_free(pr->pr_prison_racct); 4530 pr->pr_prison_racct = NULL; 4531} 4532#endif /* RACCT */ 4533 4534#ifdef DDB 4535 4536static void 4537db_show_prison(struct prison *pr) 4538{ 4539 int fi; 4540#if defined(INET) || defined(INET6) 4541 int ii; 4542#endif 4543 unsigned jsf; 4544#ifdef INET6 4545 char ip6buf[INET6_ADDRSTRLEN]; 4546#endif 4547 4548 db_printf("prison %p:\n", pr); 4549 db_printf(" jid = %d\n", pr->pr_id); 4550 db_printf(" name = %s\n", pr->pr_name); 4551 db_printf(" parent = %p\n", pr->pr_parent); 4552 db_printf(" ref = %d\n", pr->pr_ref); 4553 db_printf(" uref = %d\n", pr->pr_uref); 4554 db_printf(" path = %s\n", pr->pr_path); 4555 db_printf(" cpuset = %d\n", pr->pr_cpuset 4556 ? pr->pr_cpuset->cs_id : -1); 4557#ifdef VIMAGE 4558 db_printf(" vnet = %p\n", pr->pr_vnet); 4559#endif 4560 db_printf(" root = %p\n", pr->pr_root); 4561 db_printf(" securelevel = %d\n", pr->pr_securelevel); 4562 db_printf(" devfs_rsnum = %d\n", pr->pr_devfs_rsnum); 4563 db_printf(" children.max = %d\n", pr->pr_childmax); 4564 db_printf(" children.cur = %d\n", pr->pr_childcount); 4565 db_printf(" child = %p\n", LIST_FIRST(&pr->pr_children)); 4566 db_printf(" sibling = %p\n", LIST_NEXT(pr, pr_sibling)); 4567 db_printf(" flags = 0x%x", pr->pr_flags); 4568 for (fi = 0; fi < sizeof(pr_flag_names) / sizeof(pr_flag_names[0]); 4569 fi++) 4570 if (pr_flag_names[fi] != NULL && (pr->pr_flags & (1 << fi))) 4571 db_printf(" %s", pr_flag_names[fi]); 4572 for (fi = 0; fi < sizeof(pr_flag_jailsys) / sizeof(pr_flag_jailsys[0]); 4573 fi++) { 4574 jsf = pr->pr_flags & 4575 (pr_flag_jailsys[fi].disable | pr_flag_jailsys[fi].new); 4576 db_printf(" %-16s= %s\n", pr_flag_jailsys[fi].name, 4577 pr_flag_jailsys[fi].disable && 4578 (jsf == pr_flag_jailsys[fi].disable) ? "disable" 4579 : (jsf == pr_flag_jailsys[fi].new) ? "new" 4580 : "inherit"); 4581 } 4582 db_printf(" allow = 0x%x", pr->pr_allow); 4583 for (fi = 0; fi < sizeof(pr_allow_names) / sizeof(pr_allow_names[0]); 4584 fi++) 4585 if (pr_allow_names[fi] != NULL && (pr->pr_allow & (1 << fi))) 4586 db_printf(" %s", pr_allow_names[fi]); 4587 db_printf("\n"); 4588 db_printf(" enforce_statfs = %d\n", pr->pr_enforce_statfs); 4589 db_printf(" host.hostname = %s\n", pr->pr_hostname); 4590 db_printf(" host.domainname = %s\n", pr->pr_domainname); 4591 db_printf(" host.hostuuid = %s\n", pr->pr_hostuuid); 4592 db_printf(" host.hostid = %lu\n", pr->pr_hostid); 4593#ifdef INET 4594 db_printf(" ip4s = %d\n", pr->pr_ip4s); 4595 for (ii = 0; ii < pr->pr_ip4s; ii++) 4596 db_printf(" %s %s\n", 4597 ii == 0 ? "ip4.addr =" : " ", 4598 inet_ntoa(pr->pr_ip4[ii])); 4599#endif 4600#ifdef INET6 4601 db_printf(" ip6s = %d\n", pr->pr_ip6s); 4602 for (ii = 0; ii < pr->pr_ip6s; ii++) 4603 db_printf(" %s %s\n", 4604 ii == 0 ? "ip6.addr =" : " ", 4605 ip6_sprintf(ip6buf, &pr->pr_ip6[ii])); 4606#endif 4607} 4608 4609DB_SHOW_COMMAND(prison, db_show_prison_command) 4610{ 4611 struct prison *pr; 4612 4613 if (!have_addr) { 4614 /* 4615 * Show all prisons in the list, and prison0 which is not 4616 * listed. 4617 */ 4618 db_show_prison(&prison0); 4619 if (!db_pager_quit) { 4620 TAILQ_FOREACH(pr, &allprison, pr_list) { 4621 db_show_prison(pr); 4622 if (db_pager_quit) 4623 break; 4624 } 4625 } 4626 return; 4627 } 4628 4629 if (addr == 0) 4630 pr = &prison0; 4631 else { 4632 /* Look for a prison with the ID and with references. */ 4633 TAILQ_FOREACH(pr, &allprison, pr_list) 4634 if (pr->pr_id == addr && pr->pr_ref > 0) 4635 break; 4636 if (pr == NULL) 4637 /* Look again, without requiring a reference. */ 4638 TAILQ_FOREACH(pr, &allprison, pr_list) 4639 if (pr->pr_id == addr) 4640 break; 4641 if (pr == NULL) 4642 /* Assume address points to a valid prison. */ 4643 pr = (struct prison *)addr; 4644 } 4645 db_show_prison(pr); 4646} 4647 4648#endif /* DDB */ 4649