174bd59bbSPavel Emelyanov /* 274bd59bbSPavel Emelyanov * Pid namespaces 374bd59bbSPavel Emelyanov * 474bd59bbSPavel Emelyanov * Authors: 574bd59bbSPavel Emelyanov * (C) 2007 Pavel Emelyanov <xemul@openvz.org>, OpenVZ, SWsoft Inc. 674bd59bbSPavel Emelyanov * (C) 2007 Sukadev Bhattiprolu <sukadev@us.ibm.com>, IBM 774bd59bbSPavel Emelyanov * Many thanks to Oleg Nesterov for comments and help 874bd59bbSPavel Emelyanov * 974bd59bbSPavel Emelyanov */ 1074bd59bbSPavel Emelyanov 1174bd59bbSPavel Emelyanov #include <linux/pid.h> 1274bd59bbSPavel Emelyanov #include <linux/pid_namespace.h> 1349f4d8b9SEric W. Biederman #include <linux/user_namespace.h> 1474bd59bbSPavel Emelyanov #include <linux/syscalls.h> 1574bd59bbSPavel Emelyanov #include <linux/err.h> 160b6b030fSPavel Emelyanov #include <linux/acct.h> 175a0e3ad6STejun Heo #include <linux/slab.h> 180bb80f24SDavid Howells #include <linux/proc_ns.h> 19cf3f8921SDaniel Lezcano #include <linux/reboot.h> 20523a6a94SEric W. Biederman #include <linux/export.h> 2174bd59bbSPavel Emelyanov 2274bd59bbSPavel Emelyanov struct pid_cache { 2374bd59bbSPavel Emelyanov int nr_ids; 2474bd59bbSPavel Emelyanov char name[16]; 2574bd59bbSPavel Emelyanov struct kmem_cache *cachep; 2674bd59bbSPavel Emelyanov struct list_head list; 2774bd59bbSPavel Emelyanov }; 2874bd59bbSPavel Emelyanov 2974bd59bbSPavel Emelyanov static LIST_HEAD(pid_caches_lh); 3074bd59bbSPavel Emelyanov static DEFINE_MUTEX(pid_caches_mutex); 3174bd59bbSPavel Emelyanov static struct kmem_cache *pid_ns_cachep; 3274bd59bbSPavel Emelyanov 3374bd59bbSPavel Emelyanov /* 3474bd59bbSPavel Emelyanov * creates the kmem cache to allocate pids from. 3574bd59bbSPavel Emelyanov * @nr_ids: the number of numerical ids this pid will have to carry 3674bd59bbSPavel Emelyanov */ 3774bd59bbSPavel Emelyanov 3874bd59bbSPavel Emelyanov static struct kmem_cache *create_pid_cachep(int nr_ids) 3974bd59bbSPavel Emelyanov { 4074bd59bbSPavel Emelyanov struct pid_cache *pcache; 4174bd59bbSPavel Emelyanov struct kmem_cache *cachep; 4274bd59bbSPavel Emelyanov 4374bd59bbSPavel Emelyanov mutex_lock(&pid_caches_mutex); 4474bd59bbSPavel Emelyanov list_for_each_entry(pcache, &pid_caches_lh, list) 4574bd59bbSPavel Emelyanov if (pcache->nr_ids == nr_ids) 4674bd59bbSPavel Emelyanov goto out; 4774bd59bbSPavel Emelyanov 4874bd59bbSPavel Emelyanov pcache = kmalloc(sizeof(struct pid_cache), GFP_KERNEL); 4974bd59bbSPavel Emelyanov if (pcache == NULL) 5074bd59bbSPavel Emelyanov goto err_alloc; 5174bd59bbSPavel Emelyanov 5274bd59bbSPavel Emelyanov snprintf(pcache->name, sizeof(pcache->name), "pid_%d", nr_ids); 5374bd59bbSPavel Emelyanov cachep = kmem_cache_create(pcache->name, 5474bd59bbSPavel Emelyanov sizeof(struct pid) + (nr_ids - 1) * sizeof(struct upid), 5574bd59bbSPavel Emelyanov 0, SLAB_HWCACHE_ALIGN, NULL); 5674bd59bbSPavel Emelyanov if (cachep == NULL) 5774bd59bbSPavel Emelyanov goto err_cachep; 5874bd59bbSPavel Emelyanov 5974bd59bbSPavel Emelyanov pcache->nr_ids = nr_ids; 6074bd59bbSPavel Emelyanov pcache->cachep = cachep; 6174bd59bbSPavel Emelyanov list_add(&pcache->list, &pid_caches_lh); 6274bd59bbSPavel Emelyanov out: 6374bd59bbSPavel Emelyanov mutex_unlock(&pid_caches_mutex); 6474bd59bbSPavel Emelyanov return pcache->cachep; 6574bd59bbSPavel Emelyanov 6674bd59bbSPavel Emelyanov err_cachep: 6774bd59bbSPavel Emelyanov kfree(pcache); 6874bd59bbSPavel Emelyanov err_alloc: 6974bd59bbSPavel Emelyanov mutex_unlock(&pid_caches_mutex); 7074bd59bbSPavel Emelyanov return NULL; 7174bd59bbSPavel Emelyanov } 7274bd59bbSPavel Emelyanov 730a01f2ccSEric W. Biederman static void proc_cleanup_work(struct work_struct *work) 740a01f2ccSEric W. Biederman { 750a01f2ccSEric W. Biederman struct pid_namespace *ns = container_of(work, struct pid_namespace, proc_work); 760a01f2ccSEric W. Biederman pid_ns_release_proc(ns); 770a01f2ccSEric W. Biederman } 780a01f2ccSEric W. Biederman 79f2302505SAndrew Vagin /* MAX_PID_NS_LEVEL is needed for limiting size of 'struct pid' */ 80f2302505SAndrew Vagin #define MAX_PID_NS_LEVEL 32 81f2302505SAndrew Vagin 82f333c700SEric W. Biederman static struct ucounts *inc_pid_namespaces(struct user_namespace *ns) 83f333c700SEric W. Biederman { 84f333c700SEric W. Biederman return inc_ucount(ns, current_euid(), UCOUNT_PID_NAMESPACES); 85f333c700SEric W. Biederman } 86f333c700SEric W. Biederman 87f333c700SEric W. Biederman static void dec_pid_namespaces(struct ucounts *ucounts) 88f333c700SEric W. Biederman { 89f333c700SEric W. Biederman dec_ucount(ucounts, UCOUNT_PID_NAMESPACES); 90f333c700SEric W. Biederman } 91f333c700SEric W. Biederman 9249f4d8b9SEric W. Biederman static struct pid_namespace *create_pid_namespace(struct user_namespace *user_ns, 9349f4d8b9SEric W. Biederman struct pid_namespace *parent_pid_ns) 9474bd59bbSPavel Emelyanov { 9574bd59bbSPavel Emelyanov struct pid_namespace *ns; 96ed469a63SAlexey Dobriyan unsigned int level = parent_pid_ns->level + 1; 97f333c700SEric W. Biederman struct ucounts *ucounts; 98f2302505SAndrew Vagin int i; 99f2302505SAndrew Vagin int err; 10074bd59bbSPavel Emelyanov 101*df75e774SEric W. Biederman err = -ENOSPC; 102f333c700SEric W. Biederman if (level > MAX_PID_NS_LEVEL) 103f2302505SAndrew Vagin goto out; 104f333c700SEric W. Biederman ucounts = inc_pid_namespaces(user_ns); 105f333c700SEric W. Biederman if (!ucounts) 106f333c700SEric W. Biederman goto out; 107f2302505SAndrew Vagin 108f2302505SAndrew Vagin err = -ENOMEM; 10984406c15SPavel Emelyanov ns = kmem_cache_zalloc(pid_ns_cachep, GFP_KERNEL); 11074bd59bbSPavel Emelyanov if (ns == NULL) 111f333c700SEric W. Biederman goto out_dec; 11274bd59bbSPavel Emelyanov 11374bd59bbSPavel Emelyanov ns->pidmap[0].page = kzalloc(PAGE_SIZE, GFP_KERNEL); 11474bd59bbSPavel Emelyanov if (!ns->pidmap[0].page) 11574bd59bbSPavel Emelyanov goto out_free; 11674bd59bbSPavel Emelyanov 11774bd59bbSPavel Emelyanov ns->pid_cachep = create_pid_cachep(level + 1); 11874bd59bbSPavel Emelyanov if (ns->pid_cachep == NULL) 11974bd59bbSPavel Emelyanov goto out_free_map; 12074bd59bbSPavel Emelyanov 1216344c433SAl Viro err = ns_alloc_inum(&ns->ns); 12298f842e6SEric W. Biederman if (err) 12398f842e6SEric W. Biederman goto out_free_map; 12433c42940SAl Viro ns->ns.ops = &pidns_operations; 12598f842e6SEric W. Biederman 12674bd59bbSPavel Emelyanov kref_init(&ns->kref); 12774bd59bbSPavel Emelyanov ns->level = level; 128ed469a63SAlexey Dobriyan ns->parent = get_pid_ns(parent_pid_ns); 12949f4d8b9SEric W. Biederman ns->user_ns = get_user_ns(user_ns); 130f333c700SEric W. Biederman ns->ucounts = ucounts; 131c876ad76SEric W. Biederman ns->nr_hashed = PIDNS_HASH_ADDING; 1320a01f2ccSEric W. Biederman INIT_WORK(&ns->proc_work, proc_cleanup_work); 13374bd59bbSPavel Emelyanov 13474bd59bbSPavel Emelyanov set_bit(0, ns->pidmap[0].page); 13574bd59bbSPavel Emelyanov atomic_set(&ns->pidmap[0].nr_free, BITS_PER_PAGE - 1); 13674bd59bbSPavel Emelyanov 13784406c15SPavel Emelyanov for (i = 1; i < PIDMAP_ENTRIES; i++) 13874bd59bbSPavel Emelyanov atomic_set(&ns->pidmap[i].nr_free, BITS_PER_PAGE); 13974bd59bbSPavel Emelyanov 14074bd59bbSPavel Emelyanov return ns; 14174bd59bbSPavel Emelyanov 14274bd59bbSPavel Emelyanov out_free_map: 14374bd59bbSPavel Emelyanov kfree(ns->pidmap[0].page); 14474bd59bbSPavel Emelyanov out_free: 14574bd59bbSPavel Emelyanov kmem_cache_free(pid_ns_cachep, ns); 146f333c700SEric W. Biederman out_dec: 147f333c700SEric W. Biederman dec_pid_namespaces(ucounts); 14874bd59bbSPavel Emelyanov out: 1494308eebbSEric W. Biederman return ERR_PTR(err); 15074bd59bbSPavel Emelyanov } 15174bd59bbSPavel Emelyanov 1521adfcb03SAl Viro static void delayed_free_pidns(struct rcu_head *p) 1531adfcb03SAl Viro { 1541adfcb03SAl Viro kmem_cache_free(pid_ns_cachep, 1551adfcb03SAl Viro container_of(p, struct pid_namespace, rcu)); 1561adfcb03SAl Viro } 1571adfcb03SAl Viro 15874bd59bbSPavel Emelyanov static void destroy_pid_namespace(struct pid_namespace *ns) 15974bd59bbSPavel Emelyanov { 16074bd59bbSPavel Emelyanov int i; 16174bd59bbSPavel Emelyanov 1626344c433SAl Viro ns_free_inum(&ns->ns); 16374bd59bbSPavel Emelyanov for (i = 0; i < PIDMAP_ENTRIES; i++) 16474bd59bbSPavel Emelyanov kfree(ns->pidmap[i].page); 165f333c700SEric W. Biederman dec_pid_namespaces(ns->ucounts); 16649f4d8b9SEric W. Biederman put_user_ns(ns->user_ns); 1671adfcb03SAl Viro call_rcu(&ns->rcu, delayed_free_pidns); 16874bd59bbSPavel Emelyanov } 16974bd59bbSPavel Emelyanov 17049f4d8b9SEric W. Biederman struct pid_namespace *copy_pid_ns(unsigned long flags, 17149f4d8b9SEric W. Biederman struct user_namespace *user_ns, struct pid_namespace *old_ns) 17274bd59bbSPavel Emelyanov { 17374bd59bbSPavel Emelyanov if (!(flags & CLONE_NEWPID)) 174dca4a979SAlexey Dobriyan return get_pid_ns(old_ns); 175225778d6SEric W. Biederman if (task_active_pid_ns(current) != old_ns) 176225778d6SEric W. Biederman return ERR_PTR(-EINVAL); 17749f4d8b9SEric W. Biederman return create_pid_namespace(user_ns, old_ns); 17874bd59bbSPavel Emelyanov } 17974bd59bbSPavel Emelyanov 180bbc2e3efSCyrill Gorcunov static void free_pid_ns(struct kref *kref) 18174bd59bbSPavel Emelyanov { 182bbc2e3efSCyrill Gorcunov struct pid_namespace *ns; 18374bd59bbSPavel Emelyanov 18474bd59bbSPavel Emelyanov ns = container_of(kref, struct pid_namespace, kref); 18574bd59bbSPavel Emelyanov destroy_pid_namespace(ns); 18674bd59bbSPavel Emelyanov } 187bbc2e3efSCyrill Gorcunov 188bbc2e3efSCyrill Gorcunov void put_pid_ns(struct pid_namespace *ns) 189bbc2e3efSCyrill Gorcunov { 190bbc2e3efSCyrill Gorcunov struct pid_namespace *parent; 191bbc2e3efSCyrill Gorcunov 192bbc2e3efSCyrill Gorcunov while (ns != &init_pid_ns) { 193bbc2e3efSCyrill Gorcunov parent = ns->parent; 194bbc2e3efSCyrill Gorcunov if (!kref_put(&ns->kref, free_pid_ns)) 195bbc2e3efSCyrill Gorcunov break; 196bbc2e3efSCyrill Gorcunov ns = parent; 197bbc2e3efSCyrill Gorcunov } 198bbc2e3efSCyrill Gorcunov } 199bbc2e3efSCyrill Gorcunov EXPORT_SYMBOL_GPL(put_pid_ns); 20074bd59bbSPavel Emelyanov 20174bd59bbSPavel Emelyanov void zap_pid_ns_processes(struct pid_namespace *pid_ns) 20274bd59bbSPavel Emelyanov { 20374bd59bbSPavel Emelyanov int nr; 20474bd59bbSPavel Emelyanov int rc; 20500c10bc1SEric W. Biederman struct task_struct *task, *me = current; 206751c644bSEric W. Biederman int init_pids = thread_group_leader(me) ? 1 : 2; 20700c10bc1SEric W. Biederman 208c876ad76SEric W. Biederman /* Don't allow any more processes into the pid namespace */ 209c876ad76SEric W. Biederman disable_pid_allocation(pid_ns); 210c876ad76SEric W. Biederman 211a53b8315SOleg Nesterov /* 212a53b8315SOleg Nesterov * Ignore SIGCHLD causing any terminated children to autoreap. 213a53b8315SOleg Nesterov * This speeds up the namespace shutdown, plus see the comment 214a53b8315SOleg Nesterov * below. 215a53b8315SOleg Nesterov */ 21600c10bc1SEric W. Biederman spin_lock_irq(&me->sighand->siglock); 21700c10bc1SEric W. Biederman me->sighand->action[SIGCHLD - 1].sa.sa_handler = SIG_IGN; 21800c10bc1SEric W. Biederman spin_unlock_irq(&me->sighand->siglock); 21974bd59bbSPavel Emelyanov 22074bd59bbSPavel Emelyanov /* 22174bd59bbSPavel Emelyanov * The last thread in the cgroup-init thread group is terminating. 22274bd59bbSPavel Emelyanov * Find remaining pid_ts in the namespace, signal and wait for them 22374bd59bbSPavel Emelyanov * to exit. 22474bd59bbSPavel Emelyanov * 22574bd59bbSPavel Emelyanov * Note: This signals each threads in the namespace - even those that 22674bd59bbSPavel Emelyanov * belong to the same thread group, To avoid this, we would have 22774bd59bbSPavel Emelyanov * to walk the entire tasklist looking a processes in this 22874bd59bbSPavel Emelyanov * namespace, but that could be unnecessarily expensive if the 22974bd59bbSPavel Emelyanov * pid namespace has just a few processes. Or we need to 23074bd59bbSPavel Emelyanov * maintain a tasklist for each pid namespace. 23174bd59bbSPavel Emelyanov * 23274bd59bbSPavel Emelyanov */ 23374bd59bbSPavel Emelyanov read_lock(&tasklist_lock); 23474bd59bbSPavel Emelyanov nr = next_pidmap(pid_ns, 1); 23574bd59bbSPavel Emelyanov while (nr > 0) { 236e4da026fSSukadev Bhattiprolu rcu_read_lock(); 237e4da026fSSukadev Bhattiprolu 238e4da026fSSukadev Bhattiprolu task = pid_task(find_vpid(nr), PIDTYPE_PID); 239a02d6fd6SOleg Nesterov if (task && !__fatal_signal_pending(task)) 240a02d6fd6SOleg Nesterov send_sig_info(SIGKILL, SEND_SIG_FORCED, task); 241e4da026fSSukadev Bhattiprolu 242e4da026fSSukadev Bhattiprolu rcu_read_unlock(); 243e4da026fSSukadev Bhattiprolu 24474bd59bbSPavel Emelyanov nr = next_pidmap(pid_ns, nr); 24574bd59bbSPavel Emelyanov } 24674bd59bbSPavel Emelyanov read_unlock(&tasklist_lock); 24774bd59bbSPavel Emelyanov 248a53b8315SOleg Nesterov /* 249a53b8315SOleg Nesterov * Reap the EXIT_ZOMBIE children we had before we ignored SIGCHLD. 250a53b8315SOleg Nesterov * sys_wait4() will also block until our children traced from the 251a53b8315SOleg Nesterov * parent namespace are detached and become EXIT_DEAD. 252a53b8315SOleg Nesterov */ 25374bd59bbSPavel Emelyanov do { 25474bd59bbSPavel Emelyanov clear_thread_flag(TIF_SIGPENDING); 25574bd59bbSPavel Emelyanov rc = sys_wait4(-1, NULL, __WALL, NULL); 25674bd59bbSPavel Emelyanov } while (rc != -ECHILD); 25774bd59bbSPavel Emelyanov 2586347e900SEric W. Biederman /* 259a53b8315SOleg Nesterov * sys_wait4() above can't reap the EXIT_DEAD children but we do not 260a53b8315SOleg Nesterov * really care, we could reparent them to the global init. We could 261a53b8315SOleg Nesterov * exit and reap ->child_reaper even if it is not the last thread in 262a53b8315SOleg Nesterov * this pid_ns, free_pid(nr_hashed == 0) calls proc_cleanup_work(), 263a53b8315SOleg Nesterov * pid_ns can not go away until proc_kill_sb() drops the reference. 264a53b8315SOleg Nesterov * 265a53b8315SOleg Nesterov * But this ns can also have other tasks injected by setns()+fork(). 266a53b8315SOleg Nesterov * Again, ignoring the user visible semantics we do not really need 267a53b8315SOleg Nesterov * to wait until they are all reaped, but they can be reparented to 268a53b8315SOleg Nesterov * us and thus we need to ensure that pid->child_reaper stays valid 269a53b8315SOleg Nesterov * until they all go away. See free_pid()->wake_up_process(). 270a53b8315SOleg Nesterov * 271a53b8315SOleg Nesterov * We rely on ignored SIGCHLD, an injected zombie must be autoreaped 272a53b8315SOleg Nesterov * if reparented. 2736347e900SEric W. Biederman */ 2746347e900SEric W. Biederman for (;;) { 275af4b8a83SEric W. Biederman set_current_state(TASK_UNINTERRUPTIBLE); 276751c644bSEric W. Biederman if (pid_ns->nr_hashed == init_pids) 2776347e900SEric W. Biederman break; 2786347e900SEric W. Biederman schedule(); 2796347e900SEric W. Biederman } 280af4b8a83SEric W. Biederman __set_current_state(TASK_RUNNING); 2816347e900SEric W. Biederman 282cf3f8921SDaniel Lezcano if (pid_ns->reboot) 283cf3f8921SDaniel Lezcano current->signal->group_exit_code = pid_ns->reboot; 284cf3f8921SDaniel Lezcano 2850b6b030fSPavel Emelyanov acct_exit_ns(pid_ns); 28674bd59bbSPavel Emelyanov return; 28774bd59bbSPavel Emelyanov } 28874bd59bbSPavel Emelyanov 28998ed57eeSCyrill Gorcunov #ifdef CONFIG_CHECKPOINT_RESTORE 290b8f566b0SPavel Emelyanov static int pid_ns_ctl_handler(struct ctl_table *table, int write, 291b8f566b0SPavel Emelyanov void __user *buffer, size_t *lenp, loff_t *ppos) 292b8f566b0SPavel Emelyanov { 29349f4d8b9SEric W. Biederman struct pid_namespace *pid_ns = task_active_pid_ns(current); 294b8f566b0SPavel Emelyanov struct ctl_table tmp = *table; 295b8f566b0SPavel Emelyanov 29649f4d8b9SEric W. Biederman if (write && !ns_capable(pid_ns->user_ns, CAP_SYS_ADMIN)) 297b8f566b0SPavel Emelyanov return -EPERM; 298b8f566b0SPavel Emelyanov 299b8f566b0SPavel Emelyanov /* 300b8f566b0SPavel Emelyanov * Writing directly to ns' last_pid field is OK, since this field 301b8f566b0SPavel Emelyanov * is volatile in a living namespace anyway and a code writing to 302b8f566b0SPavel Emelyanov * it should synchronize its usage with external means. 303b8f566b0SPavel Emelyanov */ 304b8f566b0SPavel Emelyanov 30549f4d8b9SEric W. Biederman tmp.data = &pid_ns->last_pid; 306579035dcSAndrew Vagin return proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos); 307b8f566b0SPavel Emelyanov } 308b8f566b0SPavel Emelyanov 309579035dcSAndrew Vagin extern int pid_max; 310579035dcSAndrew Vagin static int zero = 0; 311b8f566b0SPavel Emelyanov static struct ctl_table pid_ns_ctl_table[] = { 312b8f566b0SPavel Emelyanov { 313b8f566b0SPavel Emelyanov .procname = "ns_last_pid", 314b8f566b0SPavel Emelyanov .maxlen = sizeof(int), 315b8f566b0SPavel Emelyanov .mode = 0666, /* permissions are checked in the handler */ 316b8f566b0SPavel Emelyanov .proc_handler = pid_ns_ctl_handler, 317579035dcSAndrew Vagin .extra1 = &zero, 318579035dcSAndrew Vagin .extra2 = &pid_max, 319b8f566b0SPavel Emelyanov }, 320b8f566b0SPavel Emelyanov { } 321b8f566b0SPavel Emelyanov }; 322b8f566b0SPavel Emelyanov static struct ctl_path kern_path[] = { { .procname = "kernel", }, { } }; 32398ed57eeSCyrill Gorcunov #endif /* CONFIG_CHECKPOINT_RESTORE */ 324b8f566b0SPavel Emelyanov 325cf3f8921SDaniel Lezcano int reboot_pid_ns(struct pid_namespace *pid_ns, int cmd) 326cf3f8921SDaniel Lezcano { 327cf3f8921SDaniel Lezcano if (pid_ns == &init_pid_ns) 328cf3f8921SDaniel Lezcano return 0; 329cf3f8921SDaniel Lezcano 330cf3f8921SDaniel Lezcano switch (cmd) { 331cf3f8921SDaniel Lezcano case LINUX_REBOOT_CMD_RESTART2: 332cf3f8921SDaniel Lezcano case LINUX_REBOOT_CMD_RESTART: 333cf3f8921SDaniel Lezcano pid_ns->reboot = SIGHUP; 334cf3f8921SDaniel Lezcano break; 335cf3f8921SDaniel Lezcano 336cf3f8921SDaniel Lezcano case LINUX_REBOOT_CMD_POWER_OFF: 337cf3f8921SDaniel Lezcano case LINUX_REBOOT_CMD_HALT: 338cf3f8921SDaniel Lezcano pid_ns->reboot = SIGINT; 339cf3f8921SDaniel Lezcano break; 340cf3f8921SDaniel Lezcano default: 341cf3f8921SDaniel Lezcano return -EINVAL; 342cf3f8921SDaniel Lezcano } 343cf3f8921SDaniel Lezcano 344cf3f8921SDaniel Lezcano read_lock(&tasklist_lock); 345cf3f8921SDaniel Lezcano force_sig(SIGKILL, pid_ns->child_reaper); 346cf3f8921SDaniel Lezcano read_unlock(&tasklist_lock); 347cf3f8921SDaniel Lezcano 348cf3f8921SDaniel Lezcano do_exit(0); 349cf3f8921SDaniel Lezcano 350cf3f8921SDaniel Lezcano /* Not reached */ 351cf3f8921SDaniel Lezcano return 0; 352cf3f8921SDaniel Lezcano } 353cf3f8921SDaniel Lezcano 3543c041184SAl Viro static inline struct pid_namespace *to_pid_ns(struct ns_common *ns) 3553c041184SAl Viro { 3563c041184SAl Viro return container_of(ns, struct pid_namespace, ns); 3573c041184SAl Viro } 3583c041184SAl Viro 35964964528SAl Viro static struct ns_common *pidns_get(struct task_struct *task) 36057e8391dSEric W. Biederman { 36157e8391dSEric W. Biederman struct pid_namespace *ns; 36257e8391dSEric W. Biederman 36357e8391dSEric W. Biederman rcu_read_lock(); 364d2308225SOleg Nesterov ns = task_active_pid_ns(task); 365d2308225SOleg Nesterov if (ns) 366d2308225SOleg Nesterov get_pid_ns(ns); 36757e8391dSEric W. Biederman rcu_read_unlock(); 36857e8391dSEric W. Biederman 3693c041184SAl Viro return ns ? &ns->ns : NULL; 37057e8391dSEric W. Biederman } 37157e8391dSEric W. Biederman 37264964528SAl Viro static void pidns_put(struct ns_common *ns) 37357e8391dSEric W. Biederman { 3743c041184SAl Viro put_pid_ns(to_pid_ns(ns)); 37557e8391dSEric W. Biederman } 37657e8391dSEric W. Biederman 37764964528SAl Viro static int pidns_install(struct nsproxy *nsproxy, struct ns_common *ns) 37857e8391dSEric W. Biederman { 37957e8391dSEric W. Biederman struct pid_namespace *active = task_active_pid_ns(current); 3803c041184SAl Viro struct pid_namespace *ancestor, *new = to_pid_ns(ns); 38157e8391dSEric W. Biederman 3825e4a0847SEric W. Biederman if (!ns_capable(new->user_ns, CAP_SYS_ADMIN) || 383c7b96acfSEric W. Biederman !ns_capable(current_user_ns(), CAP_SYS_ADMIN)) 38457e8391dSEric W. Biederman return -EPERM; 38557e8391dSEric W. Biederman 38657e8391dSEric W. Biederman /* 38757e8391dSEric W. Biederman * Only allow entering the current active pid namespace 38857e8391dSEric W. Biederman * or a child of the current active pid namespace. 38957e8391dSEric W. Biederman * 39057e8391dSEric W. Biederman * This is required for fork to return a usable pid value and 39157e8391dSEric W. Biederman * this maintains the property that processes and their 39257e8391dSEric W. Biederman * children can not escape their current pid namespace. 39357e8391dSEric W. Biederman */ 39457e8391dSEric W. Biederman if (new->level < active->level) 39557e8391dSEric W. Biederman return -EINVAL; 39657e8391dSEric W. Biederman 39757e8391dSEric W. Biederman ancestor = new; 39857e8391dSEric W. Biederman while (ancestor->level > active->level) 39957e8391dSEric W. Biederman ancestor = ancestor->parent; 40057e8391dSEric W. Biederman if (ancestor != active) 40157e8391dSEric W. Biederman return -EINVAL; 40257e8391dSEric W. Biederman 403c2b1df2eSAndy Lutomirski put_pid_ns(nsproxy->pid_ns_for_children); 404c2b1df2eSAndy Lutomirski nsproxy->pid_ns_for_children = get_pid_ns(new); 40557e8391dSEric W. Biederman return 0; 40657e8391dSEric W. Biederman } 40757e8391dSEric W. Biederman 40857e8391dSEric W. Biederman const struct proc_ns_operations pidns_operations = { 40957e8391dSEric W. Biederman .name = "pid", 41057e8391dSEric W. Biederman .type = CLONE_NEWPID, 41157e8391dSEric W. Biederman .get = pidns_get, 41257e8391dSEric W. Biederman .put = pidns_put, 41357e8391dSEric W. Biederman .install = pidns_install, 41457e8391dSEric W. Biederman }; 41557e8391dSEric W. Biederman 41674bd59bbSPavel Emelyanov static __init int pid_namespaces_init(void) 41774bd59bbSPavel Emelyanov { 41874bd59bbSPavel Emelyanov pid_ns_cachep = KMEM_CACHE(pid_namespace, SLAB_PANIC); 41998ed57eeSCyrill Gorcunov 42098ed57eeSCyrill Gorcunov #ifdef CONFIG_CHECKPOINT_RESTORE 421b8f566b0SPavel Emelyanov register_sysctl_paths(kern_path, pid_ns_ctl_table); 42298ed57eeSCyrill Gorcunov #endif 42374bd59bbSPavel Emelyanov return 0; 42474bd59bbSPavel Emelyanov } 42574bd59bbSPavel Emelyanov 42674bd59bbSPavel Emelyanov __initcall(pid_namespaces_init); 427