diff options
author | bz <bz@FreeBSD.org> | 2008-11-29 14:32:14 +0000 |
---|---|---|
committer | bz <bz@FreeBSD.org> | 2008-11-29 14:32:14 +0000 |
commit | d2730d5b27273f2e5a9b9f2703b896b5194496ee (patch) | |
tree | 2b8f4835032d12a0e61cc40dad151e279cf7a49f /sys/kern | |
parent | 3b10ad8d71152b89fda1bef42dc519fc460aba1b (diff) | |
download | FreeBSD-src-d2730d5b27273f2e5a9b9f2703b896b5194496ee.zip FreeBSD-src-d2730d5b27273f2e5a9b9f2703b896b5194496ee.tar.gz |
MFp4:
Bring in updated jail support from bz_jail branch.
This enhances the current jail implementation to permit multiple
addresses per jail. In addtion to IPv4, IPv6 is supported as well.
Due to updated checks it is even possible to have jails without
an IP address at all, which basically gives one a chroot with
restricted process view, no networking,..
SCTP support was updated and supports IPv6 in jails as well.
Cpuset support permits jails to be bound to specific processor
sets after creation.
Jails can have an unrestricted (no duplicate protection, etc.) name
in addition to the hostname. The jail name cannot be changed from
within a jail and is considered to be used for management purposes
or as audit-token in the future.
DDB 'show jails' command was added to aid debugging.
Proper compat support permits 32bit jail binaries to be used on 64bit
systems to manage jails. Also backward compatibility was preserved where
possible: for jail v1 syscalls, as well as with user space management
utilities.
Both jail as well as prison version were updated for the new features.
A gap was intentionally left as the intermediate versions had been
used by various patches floating around the last years.
Bump __FreeBSD_version for the afore mentioned and in kernel changes.
Special thanks to:
- Pawel Jakub Dawidek (pjd) for his multi-IPv4 patches
and Olivier Houchard (cognet) for initial single-IPv6 patches.
- Jeff Roberson (jeff) and Randall Stewart (rrs) for their
help, ideas and review on cpuset and SCTP support.
- Robert Watson (rwatson) for lots and lots of help, discussions,
suggestions and review of most of the patch at various stages.
- John Baldwin (jhb) for his help.
- Simon L. Nielsen (simon) as early adopter testing changes
on cluster machines as well as all the testers and people
who provided feedback the last months on freebsd-jail and
other channels.
- My employer, CK Software GmbH, for the support so I could work on this.
Reviewed by: (see above)
MFC after: 3 months (this is just so that I get the mail)
X-MFC Before: 7.2-RELEASE if possible
Diffstat (limited to 'sys/kern')
-rw-r--r-- | sys/kern/kern_cpuset.c | 115 | ||||
-rw-r--r-- | sys/kern/kern_exit.c | 5 | ||||
-rw-r--r-- | sys/kern/kern_fork.c | 6 | ||||
-rw-r--r-- | sys/kern/kern_jail.c | 908 | ||||
-rw-r--r-- | sys/kern/uipc_socket.c | 4 |
5 files changed, 974 insertions, 64 deletions
diff --git a/sys/kern/kern_cpuset.c b/sys/kern/kern_cpuset.c index 2a4c248..07e5b24 100644 --- a/sys/kern/kern_cpuset.c +++ b/sys/kern/kern_cpuset.c @@ -53,6 +53,7 @@ __FBSDID("$FreeBSD$"); #include <sys/limits.h> #include <sys/bus.h> #include <sys/interrupt.h> +#include <sys/jail.h> /* Must come after sys/proc.h */ #include <vm/uma.h> @@ -208,7 +209,7 @@ cpuset_rel_complete(struct cpuset *set) * Find a set based on an id. Returns it with a ref. */ static struct cpuset * -cpuset_lookup(cpusetid_t setid) +cpuset_lookup(cpusetid_t setid, struct thread *td) { struct cpuset *set; @@ -221,6 +222,28 @@ cpuset_lookup(cpusetid_t setid) if (set) cpuset_ref(set); mtx_unlock_spin(&cpuset_lock); + + KASSERT(td != NULL, ("[%s:%d] td is NULL", __func__, __LINE__)); + if (set != NULL && jailed(td->td_ucred)) { + struct cpuset *rset, *jset; + struct prison *pr; + + rset = cpuset_refroot(set); + + pr = td->td_ucred->cr_prison; + mtx_lock(&pr->pr_mtx); + cpuset_ref(pr->pr_cpuset); + jset = pr->pr_cpuset; + mtx_unlock(&pr->pr_mtx); + + if (jset->cs_id != rset->cs_id) { + cpuset_rel(set); + set = NULL; + } + cpuset_rel(jset); + cpuset_rel(rset); + } + return (set); } @@ -412,12 +435,38 @@ cpuset_which(cpuwhich_t which, id_t id, struct proc **pp, struct thread **tdp, set = cpuset_refbase(curthread->td_cpuset); thread_unlock(curthread); } else - set = cpuset_lookup(id); + set = cpuset_lookup(id, curthread); if (set) { *setp = set; return (0); } return (ESRCH); + case CPU_WHICH_JAIL: + { + /* Find `set' for prison with given id. */ + struct prison *pr; + + sx_slock(&allprison_lock); + pr = prison_find(id); + sx_sunlock(&allprison_lock); + if (pr == NULL) + return (ESRCH); + if (jailed(curthread->td_ucred)) { + if (curthread->td_ucred->cr_prison == pr) { + cpuset_ref(pr->pr_cpuset); + set = pr->pr_cpuset; + } + } else { + cpuset_ref(pr->pr_cpuset); + set = pr->pr_cpuset; + } + mtx_unlock(&pr->pr_mtx); + if (set) { + *setp = set; + return (0); + } + return (ESRCH); + } case CPU_WHICH_IRQ: return (0); default: @@ -664,6 +713,59 @@ cpuset_thread0(void) } /* + * Create a cpuset, which would be cpuset_create() but + * mark the new 'set' as root. + * + * We are not going to reparent the td to it. Use cpuset_reparentproc() for that. + * + * In case of no error, returns the set in *setp locked with a reference. + */ +int +cpuset_create_root(struct thread *td, struct cpuset **setp) +{ + struct cpuset *root; + struct cpuset *set; + int error; + + KASSERT(td != NULL, ("[%s:%d] invalid td", __func__, __LINE__)); + KASSERT(setp != NULL, ("[%s:%d] invalid setp", __func__, __LINE__)); + + thread_lock(td); + root = cpuset_refroot(td->td_cpuset); + thread_unlock(td); + + error = cpuset_create(setp, td->td_cpuset, &root->cs_mask); + cpuset_rel(root); + if (error) + return (error); + + KASSERT(*setp != NULL, ("[%s:%d] cpuset_create returned invalid data", + __func__, __LINE__)); + + /* Mark the set as root. */ + set = *setp; + set->cs_flags |= CPU_SET_ROOT; + + return (0); +} + +int +cpuset_setproc_update_set(struct proc *p, struct cpuset *set) +{ + int error; + + KASSERT(p != NULL, ("[%s:%d] invalid proc", __func__, __LINE__)); + KASSERT(set != NULL, ("[%s:%d] invalid set", __func__, __LINE__)); + + cpuset_ref(set); + error = cpuset_setproc(p->p_pid, set, NULL); + if (error) + return (error); + cpuset_rel(set); + return (0); +} + +/* * This is called once the final set of system cpus is known. Modifies * the root set and all children and mark the root readonly. */ @@ -728,7 +830,7 @@ cpuset_setid(struct thread *td, struct cpuset_setid_args *uap) */ if (uap->which != CPU_WHICH_PID) return (EINVAL); - set = cpuset_lookup(uap->setid); + set = cpuset_lookup(uap->setid, td); if (set == NULL) return (ESRCH); error = cpuset_setproc(uap->id, set, NULL); @@ -767,6 +869,7 @@ cpuset_getid(struct thread *td, struct cpuset_getid_args *uap) PROC_UNLOCK(p); break; case CPU_WHICH_CPUSET: + case CPU_WHICH_JAIL: break; case CPU_WHICH_IRQ: return (EINVAL); @@ -829,6 +932,7 @@ cpuset_getaffinity(struct thread *td, struct cpuset_getaffinity_args *uap) thread_unlock(ttd); break; case CPU_WHICH_CPUSET: + case CPU_WHICH_JAIL: break; case CPU_WHICH_IRQ: error = EINVAL; @@ -856,6 +960,7 @@ cpuset_getaffinity(struct thread *td, struct cpuset_getaffinity_args *uap) } break; case CPU_WHICH_CPUSET: + case CPU_WHICH_JAIL: CPU_COPY(&set->cs_mask, mask); break; case CPU_WHICH_IRQ: @@ -936,6 +1041,7 @@ cpuset_setaffinity(struct thread *td, struct cpuset_setaffinity_args *uap) PROC_UNLOCK(p); break; case CPU_WHICH_CPUSET: + case CPU_WHICH_JAIL: break; case CPU_WHICH_IRQ: error = EINVAL; @@ -958,7 +1064,8 @@ cpuset_setaffinity(struct thread *td, struct cpuset_setaffinity_args *uap) error = cpuset_setproc(uap->id, NULL, mask); break; case CPU_WHICH_CPUSET: - error = cpuset_which(CPU_WHICH_CPUSET, uap->id, &p, + case CPU_WHICH_JAIL: + error = cpuset_which(uap->which, uap->id, &p, &ttd, &set); if (error == 0) { error = cpuset_modify(set, mask); diff --git a/sys/kern/kern_exit.c b/sys/kern/kern_exit.c index 6ebe032..3b45233 100644 --- a/sys/kern/kern_exit.c +++ b/sys/kern/kern_exit.c @@ -52,6 +52,7 @@ __FBSDID("$FreeBSD$"); #include <sys/mutex.h> #include <sys/proc.h> #include <sys/pioctl.h> +#include <sys/jail.h> #include <sys/tty.h> #include <sys/wait.h> #include <sys/vmmeter.h> @@ -453,6 +454,10 @@ exit1(struct thread *td, int rv) p->p_xstat = rv; p->p_xthread = td; + /* In case we are jailed tell the prison that we are gone. */ + if (jailed(p->p_ucred)) + prison_proc_free(p->p_ucred->cr_prison); + #ifdef KDTRACE_HOOKS /* * Tell the DTrace fasttrap provider about the exit if it diff --git a/sys/kern/kern_fork.c b/sys/kern/kern_fork.c index 673913b..96dd9c8 100644 --- a/sys/kern/kern_fork.c +++ b/sys/kern/kern_fork.c @@ -54,6 +54,7 @@ __FBSDID("$FreeBSD$"); #include <sys/mutex.h> #include <sys/priv.h> #include <sys/proc.h> +#include <sys/jail.h> #include <sys/pioctl.h> #include <sys/resourcevar.h> #include <sys/sched.h> @@ -452,6 +453,11 @@ again: __rangeof(struct proc, p_startzero, p_endzero)); p2->p_ucred = crhold(td->td_ucred); + + /* In case we are jailed tell the prison that we exist. */ + if (jailed(p2->p_ucred)) + prison_proc_hold(p2->p_ucred->cr_prison); + PROC_UNLOCK(p2); /* diff --git a/sys/kern/kern_jail.c b/sys/kern/kern_jail.c index aae9178..b453d58 100644 --- a/sys/kern/kern_jail.c +++ b/sys/kern/kern_jail.c @@ -1,5 +1,7 @@ /*- - * Copyright (c) 1999 Poul-Henning Kamp. All rights reserved. + * Copyright (c) 1999 Poul-Henning Kamp. + * Copyright (c) 2008 Bjoern A. Zeeb. + * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions @@ -26,6 +28,9 @@ #include <sys/cdefs.h> __FBSDID("$FreeBSD$"); +#include "opt_ddb.h" +#include "opt_inet.h" +#include "opt_inet6.h" #include "opt_mac.h" #include <sys/param.h> @@ -54,6 +59,12 @@ __FBSDID("$FreeBSD$"); #include <sys/osd.h> #include <net/if.h> #include <netinet/in.h> +#ifdef DDB +#include <ddb/ddb.h> +#ifdef INET6 +#include <netinet6/in6_var.h> +#endif /* INET6 */ +#endif /* DDB */ #include <security/mac/mac_framework.h> @@ -70,7 +81,7 @@ SYSCTL_INT(_security_jail, OID_AUTO, set_hostname_allowed, CTLFLAG_RW, int jail_socket_unixiproute_only = 1; SYSCTL_INT(_security_jail, OID_AUTO, socket_unixiproute_only, CTLFLAG_RW, &jail_socket_unixiproute_only, 0, - "Processes in jail are limited to creating UNIX/IPv4/route sockets only"); + "Processes in jail are limited to creating UNIX/IP/route sockets only"); int jail_sysvipc_allowed = 0; SYSCTL_INT(_security_jail, OID_AUTO, sysvipc_allowed, CTLFLAG_RW, @@ -97,6 +108,11 @@ SYSCTL_INT(_security_jail, OID_AUTO, mount_allowed, CTLFLAG_RW, &jail_mount_allowed, 0, "Processes in jail can mount/unmount jail-friendly file systems"); +int jail_max_af_ips = 255; +SYSCTL_INT(_security_jail, OID_AUTO, jail_max_af_ips, CTLFLAG_RW, + &jail_max_af_ips, 0, + "Number of IP addresses a jail may have at most per address family"); + /* allprison, lastprid, and prisoncount are protected by allprison_lock. */ struct prisonlist allprison; struct sx allprison_lock; @@ -106,6 +122,12 @@ int prisoncount = 0; static void init_prison(void *); static void prison_complete(void *context, int pending); static int sysctl_jail_list(SYSCTL_HANDLER_ARGS); +#ifdef INET +static int _prison_check_ip4(struct prison *, struct in_addr *); +#endif +#ifdef INET6 +static int _prison_check_ip6(struct prison *, struct in6_addr *); +#endif static void init_prison(void *data __unused) @@ -117,6 +139,276 @@ init_prison(void *data __unused) SYSINIT(prison, SI_SUB_INTRINSIC, SI_ORDER_ANY, init_prison, NULL); +#ifdef INET +static int +qcmp_v4(const void *ip1, const void *ip2) +{ + in_addr_t iaa, iab; + + /* + * We need to compare in HBO here to get the list sorted as expected + * by the result of the code. Sorting NBO addresses gives you + * interesting results. If you do not understand, do not try. + */ + iaa = ntohl(((const struct in_addr *)ip1)->s_addr); + iab = ntohl(((const struct in_addr *)ip2)->s_addr); + + /* + * Do not simply return the difference of the two numbers, the int is + * not wide enough. + */ + if (iaa > iab) + return (1); + else if (iaa < iab) + return (-1); + else + return (0); +} +#endif + +#ifdef INET6 +static int +qcmp_v6(const void *ip1, const void *ip2) +{ + const struct in6_addr *ia6a, *ia6b; + int i, rc; + + ia6a = (const struct in6_addr *)ip1; + ia6b = (const struct in6_addr *)ip2; + + rc = 0; + for (i=0; rc == 0 && i < sizeof(struct in6_addr); i++) { + if (ia6a->s6_addr[i] > ia6b->s6_addr[i]) + rc = 1; + else if (ia6a->s6_addr[i] < ia6b->s6_addr[i]) + rc = -1; + } + return (rc); +} +#endif + +#if defined(INET) || defined(INET6) +static int +prison_check_conflicting_ips(struct prison *p) +{ + struct prison *pr; + int i; + + sx_assert(&allprison_lock, SX_LOCKED); + + if (p->pr_ip4s == 0 && p->pr_ip6s == 0) + return (0); + + LIST_FOREACH(pr, &allprison, pr_list) { + /* + * Skip 'dying' prisons to avoid problems when + * restarting multi-IP jails. + */ + if (pr->pr_state == PRISON_STATE_DYING) + continue; + + /* + * We permit conflicting IPs if there is no + * more than 1 IP on eeach jail. + * In case there is one duplicate on a jail with + * more than one IP stop checking and return error. + */ +#ifdef INET + if ((p->pr_ip4s >= 1 && pr->pr_ip4s > 1) || + (p->pr_ip4s > 1 && pr->pr_ip4s >= 1)) { + for (i = 0; i < p->pr_ip4s; i++) { + if (_prison_check_ip4(pr, &p->pr_ip4[i])) + return (EINVAL); + } + } +#endif +#ifdef INET6 + if ((p->pr_ip6s >= 1 && pr->pr_ip6s > 1) || + (p->pr_ip6s > 1 && pr->pr_ip6s >= 1)) { + for (i = 0; i < p->pr_ip6s; i++) { + if (_prison_check_ip6(pr, &p->pr_ip6[i])) + return (EINVAL); + } + } +#endif + } + + return (0); +} + +static int +jail_copyin_ips(struct jail *j) +{ +#ifdef INET + struct in_addr *ip4; +#endif +#ifdef INET6 + struct in6_addr *ip6; +#endif + int error, i; + + /* + * Copy in addresses, check for duplicate addresses and do some + * simple 0 and broadcast checks. If users give other bogus addresses + * it is their problem. + * + * IP addresses are all sorted but ip[0] to preserve the primary IP + * address as given from userland. This special IP is used for + * unbound outgoing connections as well for "loopback" traffic. + */ +#ifdef INET + ip4 = NULL; +#endif +#ifdef INET6 + ip6 = NULL; +#endif +#ifdef INET + if (j->ip4s > 0) { + ip4 = (struct in_addr *)malloc(j->ip4s * sizeof(struct in_addr), + M_PRISON, M_WAITOK | M_ZERO); + error = copyin(j->ip4, ip4, j->ip4s * sizeof(struct in_addr)); + if (error) + goto e_free_ip; + /* Sort all but the first IPv4 address. */ + if (j->ip4s > 1) + qsort((ip4 + 1), j->ip4s - 1, + sizeof(struct in_addr), qcmp_v4); + + /* + * We do not have to care about byte order for these checks + * so we will do them in NBO. + */ + for (i=0; i<j->ip4s; i++) { + if (ip4[i].s_addr == htonl(INADDR_ANY) || + ip4[i].s_addr == htonl(INADDR_BROADCAST)) { + error = EINVAL; + goto e_free_ip; + } + if ((i+1) < j->ip4s && + (ip4[0].s_addr == ip4[i+1].s_addr || + ip4[i].s_addr == ip4[i+1].s_addr)) { + error = EINVAL; + goto e_free_ip; + } + } + + j->ip4 = ip4; + } +#endif +#ifdef INET6 + if (j->ip6s > 0) { + ip6 = (struct in6_addr *)malloc(j->ip6s * sizeof(struct in6_addr), + M_PRISON, M_WAITOK | M_ZERO); + error = copyin(j->ip6, ip6, j->ip6s * sizeof(struct in6_addr)); + if (error) + goto e_free_ip; + /* Sort all but the first IPv6 address. */ + if (j->ip6s > 1) + qsort((ip6 + 1), j->ip6s - 1, + sizeof(struct in6_addr), qcmp_v6); + for (i=0; i<j->ip6s; i++) { + if (IN6_IS_ADDR_UNSPECIFIED(&ip6[i])) { + error = EINVAL; + goto e_free_ip; + } + if ((i+1) < j->ip6s && + (IN6_ARE_ADDR_EQUAL(&ip6[0], &ip6[i+1]) || + IN6_ARE_ADDR_EQUAL(&ip6[i], &ip6[i+1]))) { + error = EINVAL; + goto e_free_ip; + } + } + + j->ip6 = ip6; + } +#endif + return (0); + +e_free_ip: +#ifdef INET6 + free(ip6, M_PRISON); +#endif +#ifdef INET + free(ip4, M_PRISON); +#endif + return (error); +} +#endif /* INET || INET6 */ + +static int +jail_handle_ips(struct jail *j) +{ +#if defined(INET) || defined(INET6) + int error; +#endif + + /* + * Finish conversion for older versions, copyin and setup IPs. + */ + switch (j->version) { + case 0: + { +#ifdef INET + /* FreeBSD single IPv4 jails. */ + struct in_addr *ip4; + + if (j->ip4s == INADDR_ANY || j->ip4s == INADDR_BROADCAST) + return (EINVAL); + ip4 = (struct in_addr *)malloc(sizeof(struct in_addr), + M_PRISON, M_WAITOK | M_ZERO); + + /* + * Jail version 0 still used HBO for the IPv4 address. + */ + ip4->s_addr = htonl(j->ip4s); + j->ip4s = 1; + j->ip4 = ip4; + break; +#else + return (EINVAL); +#endif + } + + case 1: + /* + * Version 1 was used by multi-IPv4 jail implementations + * that never made it into the official kernel. + * We should never hit this here; jail() should catch it. + */ + return (EINVAL); + + case 2: /* JAIL_API_VERSION */ + /* FreeBSD multi-IPv4/IPv6,noIP jails. */ +#if defined(INET) || defined(INET6) +#ifdef INET + if (j->ip4s > jail_max_af_ips) + return (EINVAL); +#else + if (j->ip4s != 0) + return (EINVAL); +#endif +#ifdef INET6 + if (j->ip6s > jail_max_af_ips) + return (EINVAL); +#else + if (j->ip6s != 0) + return (EINVAL); +#endif + error = jail_copyin_ips(j); + if (error) + return (error); +#endif + break; + + default: + /* Sci-Fi jails are not supported, sorry. */ + return (EINVAL); + } + + return (0); +} + + /* * struct jail_args { * struct jail *jail; @@ -125,22 +417,72 @@ SYSINIT(prison, SI_SUB_INTRINSIC, SI_ORDER_ANY, init_prison, NULL); int jail(struct thread *td, struct jail_args *uap) { + uint32_t version; + int error; + struct jail j; + + error = copyin(uap->jail, &version, sizeof(uint32_t)); + if (error) + return (error); + + switch (version) { + case 0: + /* FreeBSD single IPv4 jails. */ + { + struct jail_v0 j0; + + bzero(&j, sizeof(struct jail)); + error = copyin(uap->jail, &j0, sizeof(struct jail_v0)); + if (error) + return (error); + j.version = j0.version; + j.path = j0.path; + j.hostname = j0.hostname; + j.ip4s = j0.ip_number; + break; + } + + case 1: + /* + * Version 1 was used by multi-IPv4 jail implementations + * that never made it into the official kernel. + */ + return (EINVAL); + + case 2: /* JAIL_API_VERSION */ + /* FreeBSD multi-IPv4/IPv6,noIP jails. */ + error = copyin(uap->jail, &j, sizeof(struct jail)); + if (error) + return (error); + break; + + default: + /* Sci-Fi jails are not supported, sorry. */ + return (EINVAL); + } + return (kern_jail(td, &j)); +} + +int +kern_jail(struct thread *td, struct jail *j) +{ struct nameidata nd; struct prison *pr, *tpr; - struct jail j; struct jail_attach_args jaa; int vfslocked, error, tryprid; - error = copyin(uap->jail, &j, sizeof(j)); + KASSERT(j != NULL, ("%s: j is NULL", __func__)); + + /* Handle addresses - convert old structs, copyin, check IPs. */ + error = jail_handle_ips(j); if (error) return (error); - if (j.version != 0) - return (EINVAL); + /* Allocate struct prison and fill it with life. */ pr = malloc(sizeof(*pr), M_PRISON, M_WAITOK | M_ZERO); mtx_init(&pr->pr_mtx, "jail mutex", NULL, MTX_DEF); pr->pr_ref = 1; - error = copyinstr(j.path, &pr->pr_path, sizeof(pr->pr_path), 0); + error = copyinstr(j->path, &pr->pr_path, sizeof(pr->pr_path), NULL); if (error) goto e_killmtx; NDINIT(&nd, LOOKUP, MPSAFE | FOLLOW | LOCKLEAF, UIO_SYSSPACE, @@ -153,16 +495,50 @@ jail(struct thread *td, struct jail_args *uap) VOP_UNLOCK(nd.ni_vp, 0); NDFREE(&nd, NDF_ONLY_PNBUF); VFS_UNLOCK_GIANT(vfslocked); - error = copyinstr(j.hostname, &pr->pr_host, sizeof(pr->pr_host), 0); + error = copyinstr(j->hostname, &pr->pr_host, sizeof(pr->pr_host), NULL); if (error) goto e_dropvnref; - pr->pr_ip = j.ip_number; + if (j->jailname != NULL) { + error = copyinstr(j->jailname, &pr->pr_name, + sizeof(pr->pr_name), NULL); + if (error) + goto e_dropvnref; + } + if (j->ip4s > 0) { + pr->pr_ip4 = j->ip4; + pr->pr_ip4s = j->ip4s; + } +#ifdef INET6 + if (j->ip6s > 0) { + pr->pr_ip6 = j->ip6; + pr->pr_ip6s = j->ip6s; + } +#endif pr->pr_linux = NULL; pr->pr_securelevel = securelevel; bzero(&pr->pr_osd, sizeof(pr->pr_osd)); - /* Determine next pr_id and add prison to allprison list. */ + /* + * Pre-set prison state to ALIVE upon cration. This is needed so we + * can later attach the process to it, etc (avoiding another extra + * state for ther process of creation, complicating things). + */ + pr->pr_state = PRISON_STATE_ALIVE; + + /* Allocate a dedicated cpuset for each jail. */ + error = cpuset_create_root(td, &pr->pr_cpuset); + if (error) + goto e_dropvnref; + sx_xlock(&allprison_lock); + /* Make sure we cannot run into problems with ambiguous bind()ings. */ + error = prison_check_conflicting_ips(pr); + if (error) { + sx_xunlock(&allprison_lock); + goto e_dropcpuset; + } + + /* Determine next pr_id and add prison to allprison list. */ tryprid = lastprid + 1; if (tryprid == JAIL_MAX) tryprid = 1; @@ -173,7 +549,7 @@ next: if (tryprid == JAIL_MAX) { sx_xunlock(&allprison_lock); error = EAGAIN; - goto e_dropvnref; + goto e_dropcpuset; } goto next; } @@ -196,6 +572,8 @@ e_dropprref: LIST_REMOVE(pr, pr_list); prisoncount--; sx_xunlock(&allprison_lock); +e_dropcpuset: + cpuset_rel(pr->pr_cpuset); e_dropvnref: vfslocked = VFS_LOCK_GIANT(pr->pr_root->v_mount); vrele(pr->pr_root); @@ -203,6 +581,12 @@ e_dropvnref: e_killmtx: mtx_destroy(&pr->pr_mtx); free(pr, M_PRISON); +#ifdef INET6 + free(j->ip6, M_PRISON); +#endif +#ifdef INET + free(j->ip4, M_PRISON); +#endif return (error); } @@ -238,10 +622,27 @@ jail_attach(struct thread *td, struct jail_attach_args *uap) sx_sunlock(&allprison_lock); return (EINVAL); } + + /* + * Do not allow a process to attach to a prison that is not + * considered to be "ALIVE". + */ + if (pr->pr_state != PRISON_STATE_ALIVE) { + mtx_unlock(&pr->pr_mtx); + sx_sunlock(&allprison_lock); + return (EINVAL); + } pr->pr_ref++; mtx_unlock(&pr->pr_mtx); sx_sunlock(&allprison_lock); + /* + * Reparent the newly attached process to this jail. + */ + error = cpuset_setproc_update_set(p, pr->pr_cpuset); + if (error) + goto e_unref; + vfslocked = VFS_LOCK_GIANT(pr->pr_root->v_mount); vn_lock(pr->pr_root, LK_EXCLUSIVE | LK_RETRY); if ((error = change_dir(pr->pr_root, td)) != 0) @@ -261,12 +662,14 @@ jail_attach(struct thread *td, struct jail_attach_args *uap) crcopy(newcred, oldcred); newcred->cr_prison = pr; p->p_ucred = newcred; + prison_proc_hold(pr); PROC_UNLOCK(p); crfree(oldcred); return (0); e_unlock: VOP_UNLOCK(pr->pr_root, 0); VFS_UNLOCK_GIANT(vfslocked); +e_unref: mtx_lock(&pr->pr_mtx); pr->pr_ref--; mtx_unlock(&pr->pr_mtx); @@ -331,6 +734,8 @@ prison_complete(void *context, int pending) prisoncount--; sx_xunlock(&allprison_lock); + cpuset_rel(pr->pr_cpuset); + /* Free all OSD associated to this jail. */ osd_jail_exit(pr); @@ -339,8 +744,13 @@ prison_complete(void *context, int pending) VFS_UNLOCK_GIANT(vfslocked); mtx_destroy(&pr->pr_mtx); - if (pr->pr_linux != NULL) - free(pr->pr_linux, M_PRISON); + free(pr->pr_linux, M_PRISON); +#ifdef INET6 + free(pr->pr_ip6, M_PRISON); +#endif +#ifdef INET + free(pr->pr_ip4, M_PRISON); +#endif free(pr, M_PRISON); } @@ -363,79 +773,364 @@ prison_hold(struct prison *pr) mtx_unlock(&pr->pr_mtx); } -u_int32_t -prison_getip(struct ucred *cred) +void +prison_proc_hold(struct prison *pr) { - return (cred->cr_prison->pr_ip); + mtx_lock(&pr->pr_mtx); + KASSERT(pr->pr_state == PRISON_STATE_ALIVE, + ("Cannot add a process to a non-alive prison (id=%d).", pr->pr_id)); + pr->pr_nprocs++; + mtx_unlock(&pr->pr_mtx); } +void +prison_proc_free(struct prison *pr) +{ + + mtx_lock(&pr->pr_mtx); + KASSERT(pr->pr_state == PRISON_STATE_ALIVE && pr->pr_nprocs > 0, + ("Trying to kill a process in a dead prison (id=%d).", pr->pr_id)); + pr->pr_nprocs--; + if (pr->pr_nprocs == 0) + pr->pr_state = PRISON_STATE_DYING; + mtx_unlock(&pr->pr_mtx); +} + + +#ifdef INET +/* + * Pass back primary IPv4 address of this jail. + * + * If not jailed return success but do not alter the address. Caller has to + * make sure to intialize it correctly (INADDR_ANY). + * + * Returns 0 on success, 1 on error. Address returned in NBO. + */ int -prison_ip(struct ucred *cred, int flag, u_int32_t *ip) +prison_getip4(struct ucred *cred, struct in_addr *ia) { - u_int32_t tmp; + + KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); + KASSERT(ia != NULL, ("%s: ia is NULL", __func__)); if (!jailed(cred)) + /* Do not change address passed in. */ return (0); - if (flag) - tmp = *ip; - else - tmp = ntohl(*ip); - if (tmp == INADDR_ANY) { - if (flag) - *ip = cred->cr_prison->pr_ip; - else - *ip = htonl(cred->cr_prison->pr_ip); + + if (cred->cr_prison->pr_ip4 == NULL) + return (1); + + ia->s_addr = cred->cr_prison->pr_ip4[0].s_addr; + return (0); +} + +/* + * Make sure our (source) address is set to something meaningful to this + * jail. + * + * Returns 0 on success, 1 on error. Address passed in in NBO and returned + * in NBO. + */ +int +prison_local_ip4(struct ucred *cred, struct in_addr *ia) +{ + struct in_addr ia0; + + KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); + KASSERT(ia != NULL, ("%s: ia is NULL", __func__)); + + if (!jailed(cred)) + return (0); + if (cred->cr_prison->pr_ip4 == NULL) + return (1); + + ia0.s_addr = ntohl(ia->s_addr); + if (ia0.s_addr == INADDR_LOOPBACK) { + ia->s_addr = cred->cr_prison->pr_ip4[0].s_addr; return (0); } - if (tmp == INADDR_LOOPBACK) { - if (flag) - *ip = cred->cr_prison->pr_ip; - else - *ip = htonl(cred->cr_prison->pr_ip); + + /* + * In case there is only 1 IPv4 address, bind directly. + */ + if (ia0.s_addr == INADDR_ANY && cred->cr_prison->pr_ip4s == 1) { + ia->s_addr = cred->cr_prison->pr_ip4[0].s_addr; return (0); } - if (cred->cr_prison->pr_ip != tmp) + + if (ia0.s_addr == INADDR_ANY || prison_check_ip4(cred, ia)) + return (0); + + return (1); +} + +/* + * Rewrite destination address in case we will connect to loopback address. + * + * Returns 0 on success, 1 on error. Address passed in in NBO and returned + * in NBO. + */ +int +prison_remote_ip4(struct ucred *cred, struct in_addr *ia) +{ + + KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); + KASSERT(ia != NULL, ("%s: ia is NULL", __func__)); + + if (!jailed(cred)) + return (0); + if (cred->cr_prison->pr_ip4 == NULL) return (1); + if (ntohl(ia->s_addr) == INADDR_LOOPBACK) { + ia->s_addr = cred->cr_prison->pr_ip4[0].s_addr; + return (0); + } + + /* + * Return success because nothing had to be changed. + */ return (0); } -void -prison_remote_ip(struct ucred *cred, int flag, u_int32_t *ip) +/* + * Check if given address belongs to the jail referenced by cred. + * + * Returns 1 if address belongs to jail, 0 if not. Address passed in in NBO. + */ +static int +_prison_check_ip4(struct prison *pr, struct in_addr *ia) { - u_int32_t tmp; + int i, a, z, d; + + if (pr->pr_ip4 == NULL) + return (0); + + /* + * Check the primary IP. + */ + if (pr->pr_ip4[0].s_addr == ia->s_addr) + return (1); + + /* + * All the other IPs are sorted so we can do a binary search. + */ + a = 0; + z = pr->pr_ip4s - 2; + while (a <= z) { + i = (a + z) / 2; + d = qcmp_v4(&pr->pr_ip4[i+1], ia); + if (d > 0) + z = i - 1; + else if (d < 0) + a = i + 1; + else + return (1); + } + return (0); +} + +int +prison_check_ip4(struct ucred *cred, struct in_addr *ia) +{ + + KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); + KASSERT(ia != NULL, ("%s: ia is NULL", __func__)); if (!jailed(cred)) - return; - if (flag) - tmp = *ip; - else - tmp = ntohl(*ip); - if (tmp == INADDR_LOOPBACK) { - if (flag) - *ip = cred->cr_prison->pr_ip; + return (1); + + return (_prison_check_ip4(cred->cr_prison, ia)); +} +#endif + +#ifdef INET6 +/* + * Pass back primary IPv6 address for this jail. + * + * If not jailed return success but do not alter the address. Caller has to + * make sure to intialize it correctly (IN6ADDR_ANY_INIT). + * + * Returns 0 on success, 1 on error. + */ +int +prison_getip6(struct ucred *cred, struct in6_addr *ia6) +{ + + KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); + KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__)); + + if (!jailed(cred)) + return (0); + if (cred->cr_prison->pr_ip6 == NULL) + return (1); + bcopy(&cred->cr_prison->pr_ip6[0], ia6, sizeof(struct in6_addr)); + return (0); +} + +/* + * Make sure our (source) address is set to something meaningful to this jail. + * + * v6only should be set based on (inp->inp_flags & IN6P_IPV6_V6ONLY != 0) + * when needed while binding. + * + * Returns 0 on success, 1 on error. + */ +int +prison_local_ip6(struct ucred *cred, struct in6_addr *ia6, int v6only) +{ + + KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); + KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__)); + + if (!jailed(cred)) + return (0); + if (cred->cr_prison->pr_ip6 == NULL) + return (1); + if (IN6_IS_ADDR_LOOPBACK(ia6)) { + bcopy(&cred->cr_prison->pr_ip6[0], ia6, + sizeof(struct in6_addr)); + return (0); + } + + /* + * In case there is only 1 IPv6 address, and v6only is true, then + * bind directly. + */ + if (v6only != 0 && IN6_IS_ADDR_UNSPECIFIED(ia6) && + cred->cr_prison->pr_ip6s == 1) { + bcopy(&cred->cr_prison->pr_ip6[0], ia6, + sizeof(struct in6_addr)); + return (0); + } + if (IN6_IS_ADDR_UNSPECIFIED(ia6) || prison_check_ip6(cred, ia6)) + return (0); + return (1); +} + +/* + * Rewrite destination address in case we will connect to loopback address. + * + * Returns 0 on success, 1 on error. + */ +int +prison_remote_ip6(struct ucred *cred, struct in6_addr *ia6) +{ + + KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); + KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__)); + + if (!jailed(cred)) + return (0); + if (cred->cr_prison->pr_ip6 == NULL) + return (1); + if (IN6_IS_ADDR_LOOPBACK(ia6)) { + bcopy(&cred->cr_prison->pr_ip6[0], ia6, + sizeof(struct in6_addr)); + return (0); + } + + /* + * Return success because nothing had to be changed. + */ + return (0); +} + +/* + * Check if given address belongs to the jail referenced by cred. + * + * Returns 1 if address belongs to jail, 0 if not. + */ +static int +_prison_check_ip6(struct prison *pr, struct in6_addr *ia6) +{ + int i, a, z, d; + + if (pr->pr_ip6 == NULL) + return (0); + + /* + * Check the primary IP. + */ + if (IN6_ARE_ADDR_EQUAL(&pr->pr_ip6[0], ia6)) + return (1); + + /* + * All the other IPs are sorted so we can do a binary search. + */ + a = 0; + z = pr->pr_ip6s - 2; + while (a <= z) { + i = (a + z) / 2; + d = qcmp_v6(&pr->pr_ip6[i+1], ia6); + if (d > 0) + z = i - 1; + else if (d < 0) + a = i + 1; else - *ip = htonl(cred->cr_prison->pr_ip); - return; + return (1); } - return; + return (0); } int +prison_check_ip6(struct ucred *cred, struct in6_addr *ia6) +{ + + KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); + KASSERT(ia6 != NULL, ("%s: ia6 is NULL", __func__)); + + if (!jailed(cred)) + return (1); + + return (_prison_check_ip6(cred->cr_prison, ia6)); +} +#endif + +/* + * Check if given address belongs to the jail referenced by cred (wrapper to + * prison_check_ip[46]). + * + * Returns 1 if address belongs to jail, 0 if not. IPv4 Address passed in in + * NBO. + */ +int prison_if(struct ucred *cred, struct sockaddr *sa) { +#ifdef INET struct sockaddr_in *sai; +#endif +#ifdef INET6 + struct sockaddr_in6 *sai6; +#endif int ok; - sai = (struct sockaddr_in *)sa; - if ((sai->sin_family != AF_INET) && jail_socket_unixiproute_only) - ok = 1; - else if (sai->sin_family != AF_INET) - ok = 0; - else if (cred->cr_prison->pr_ip != ntohl(sai->sin_addr.s_addr)) - ok = 1; - else - ok = 0; + KASSERT(cred != NULL, ("%s: cred is NULL", __func__)); + KASSERT(sa != NULL, ("%s: sa is NULL", __func__)); + + ok = 0; + switch(sa->sa_family) + { +#ifdef INET + case AF_INET: + sai = (struct sockaddr_in *)sa; + if (prison_check_ip4(cred, &sai->sin_addr)) + ok = 1; + break; + +#endif +#ifdef INET6 + case AF_INET6: + sai6 = (struct sockaddr_in6 *)sa; + if (prison_check_ip6(cred, (struct in6_addr *)&sai6->sin6_addr)) + ok = 1; + break; + +#endif + default: + if (!jail_socket_unixiproute_only) + ok = 1; + } return (ok); } @@ -655,6 +1350,7 @@ prison_priv_check(struct ucred *cred, int priv) * processes in the same jail. Likewise for signalling. */ case PRIV_SCHED_DIFFCRED: + case PRIV_SCHED_CPUSET: case PRIV_SIGNAL_DIFFCRED: case PRIV_SIGNAL_SUGID: @@ -764,6 +1460,8 @@ sysctl_jail_list(SYSCTL_HANDLER_ARGS) { struct xprison *xp, *sxp; struct prison *pr; + char *p; + size_t len; int count, error; if (jailed(req->td->td_ucred)) @@ -775,21 +1473,54 @@ sysctl_jail_list(SYSCTL_HANDLER_ARGS) return (0); } - sxp = xp = malloc(sizeof(*xp) * count, M_TEMP, M_WAITOK | M_ZERO); + len = sizeof(*xp) * count; + LIST_FOREACH(pr, &allprison, pr_list) { +#ifdef INET + len += pr->pr_ip4s * sizeof(struct in_addr); +#endif +#ifdef INET6 + len += pr->pr_ip6s * sizeof(struct in6_addr); +#endif + } + + sxp = xp = malloc(len, M_TEMP, M_WAITOK | M_ZERO); LIST_FOREACH(pr, &allprison, pr_list) { xp->pr_version = XPRISON_VERSION; xp->pr_id = pr->pr_id; - xp->pr_ip = pr->pr_ip; + xp->pr_state = pr->pr_state; + xp->pr_cpusetid = pr->pr_cpuset->cs_id; strlcpy(xp->pr_path, pr->pr_path, sizeof(xp->pr_path)); mtx_lock(&pr->pr_mtx); strlcpy(xp->pr_host, pr->pr_host, sizeof(xp->pr_host)); + strlcpy(xp->pr_name, pr->pr_name, sizeof(xp->pr_name)); mtx_unlock(&pr->pr_mtx); - xp++; +#ifdef INET + xp->pr_ip4s = pr->pr_ip4s; +#endif +#ifdef INET6 + xp->pr_ip6s = pr->pr_ip6s; +#endif + p = (char *)(xp + 1); +#ifdef INET + if (pr->pr_ip4s > 0) { + bcopy(pr->pr_ip4, (struct in_addr *)p, + pr->pr_ip4s * sizeof(struct in_addr)); + p += (pr->pr_ip4s * sizeof(struct in_addr)); + } +#endif +#ifdef INET6 + if (pr->pr_ip6s > 0) { + bcopy(pr->pr_ip6, (struct in6_addr *)p, + pr->pr_ip6s * sizeof(struct in6_addr)); + p += (pr->pr_ip6s * sizeof(struct in6_addr)); + } +#endif + xp = (struct xprison *)p; } sx_sunlock(&allprison_lock); - error = SYSCTL_OUT(req, sxp, sizeof(*sxp) * count); + error = SYSCTL_OUT(req, sxp, len); free(sxp, M_TEMP); return (error); } @@ -809,3 +1540,60 @@ sysctl_jail_jailed(SYSCTL_HANDLER_ARGS) } SYSCTL_PROC(_security_jail, OID_AUTO, jailed, CTLTYPE_INT | CTLFLAG_RD, NULL, 0, sysctl_jail_jailed, "I", "Process in jail?"); + +#ifdef DDB +DB_SHOW_COMMAND(jails, db_show_jails) +{ + struct prison *pr; +#ifdef INET + struct in_addr ia; +#endif +#ifdef INET6 + char ip6buf[INET6_ADDRSTRLEN]; +#endif + const char *state; +#if defined(INET) || defined(INET6) + int i; +#endif + + db_printf( + " JID pr_ref pr_nprocs pr_ip4s pr_ip6s\n"); + db_printf( + " Hostname Path\n"); + db_printf( + " Name State\n"); + db_printf( + " Cpusetid\n"); + db_printf( + " IP Address(es)\n"); + LIST_FOREACH(pr, &allprison, pr_list) { + db_printf("%6d %6d %9d %7d %7d\n", + pr->pr_id, pr->pr_ref, pr->pr_nprocs, + pr->pr_ip4s, pr->pr_ip6s); + db_printf("%6s %-29.29s %.74s\n", + "", pr->pr_host, pr->pr_path); + if (pr->pr_state < 0 || pr->pr_state > (int)((sizeof( + prison_states) / sizeof(struct prison_state)))) + state = "(bogus)"; + else + state = prison_states[pr->pr_state].state_name; + db_printf("%6s %-29.29s %.74s\n", + "", (pr->pr_name != NULL) ? pr->pr_name : "", state); + db_printf("%6s %-6d\n", + "", pr->pr_cpuset->cs_id); +#ifdef INET + for (i=0; i < pr->pr_ip4s; i++) { + ia.s_addr = pr->pr_ip4[i].s_addr; + db_printf("%6s %s\n", "", inet_ntoa(ia)); + } +#endif +#ifdef INET6 + for (i=0; i < pr->pr_ip6s; i++) + db_printf("%6s %s\n", + "", ip6_sprintf(ip6buf, &pr->pr_ip6[i])); +#endif /* INET6 */ + if (db_pager_quit) + break; + } +} +#endif /* DDB */ diff --git a/sys/kern/uipc_socket.c b/sys/kern/uipc_socket.c index 53c800e..2aa63e8 100644 --- a/sys/kern/uipc_socket.c +++ b/sys/kern/uipc_socket.c @@ -98,6 +98,7 @@ __FBSDID("$FreeBSD$"); #include "opt_inet.h" +#include "opt_inet6.h" #include "opt_mac.h" #include "opt_zero.h" #include "opt_compat.h" @@ -346,6 +347,9 @@ socreate(int dom, struct socket **aso, int type, int proto, if (jailed(cred) && jail_socket_unixiproute_only && prp->pr_domain->dom_family != PF_LOCAL && prp->pr_domain->dom_family != PF_INET && +#ifdef INET6 + prp->pr_domain->dom_family != PF_INET6 && +#endif prp->pr_domain->dom_family != PF_ROUTE) { return (EPROTONOSUPPORT); } |