5 * (C) 2007 Pavel Emelyanov <xemul@openvz.org>, OpenVZ, SWsoft Inc.
6 * (C) 2007 Sukadev Bhattiprolu <sukadev@us.ibm.com>, IBM
7 * Many thanks to Oleg Nesterov for comments and help
11 #include <linux/pid.h>
12 #include <linux/pid_namespace.h>
13 #include <linux/syscalls.h>
14 #include <linux/err.h>
15 #include <linux/acct.h>
17 #define BITS_PER_PAGE (PAGE_SIZE*8)
22 struct kmem_cache *cachep;
23 struct list_head list;
26 static LIST_HEAD(pid_caches_lh);
27 static DEFINE_MUTEX(pid_caches_mutex);
28 static struct kmem_cache *pid_ns_cachep;
31 * creates the kmem cache to allocate pids from.
32 * @nr_ids: the number of numerical ids this pid will have to carry
35 static struct kmem_cache *create_pid_cachep(int nr_ids)
37 struct pid_cache *pcache;
38 struct kmem_cache *cachep;
40 mutex_lock(&pid_caches_mutex);
41 list_for_each_entry(pcache, &pid_caches_lh, list)
42 if (pcache->nr_ids == nr_ids)
45 pcache = kmalloc(sizeof(struct pid_cache), GFP_KERNEL);
49 snprintf(pcache->name, sizeof(pcache->name), "pid_%d", nr_ids);
50 cachep = kmem_cache_create(pcache->name,
51 sizeof(struct pid) + (nr_ids - 1) * sizeof(struct upid),
52 0, SLAB_HWCACHE_ALIGN, NULL);
56 pcache->nr_ids = nr_ids;
57 pcache->cachep = cachep;
58 list_add(&pcache->list, &pid_caches_lh);
60 mutex_unlock(&pid_caches_mutex);
61 return pcache->cachep;
66 mutex_unlock(&pid_caches_mutex);
70 static struct pid_namespace *create_pid_namespace(struct pid_namespace *parent_pid_ns)
72 struct pid_namespace *ns;
73 unsigned int level = parent_pid_ns->level + 1;
76 ns = kmem_cache_zalloc(pid_ns_cachep, GFP_KERNEL);
80 ns->pidmap[0].page = kzalloc(PAGE_SIZE, GFP_KERNEL);
81 if (!ns->pidmap[0].page)
84 ns->pid_cachep = create_pid_cachep(level + 1);
85 if (ns->pid_cachep == NULL)
90 ns->parent = get_pid_ns(parent_pid_ns);
92 set_bit(0, ns->pidmap[0].page);
93 atomic_set(&ns->pidmap[0].nr_free, BITS_PER_PAGE - 1);
95 for (i = 1; i < PIDMAP_ENTRIES; i++)
96 atomic_set(&ns->pidmap[i].nr_free, BITS_PER_PAGE);
101 kfree(ns->pidmap[0].page);
103 kmem_cache_free(pid_ns_cachep, ns);
105 return ERR_PTR(-ENOMEM);
108 static void destroy_pid_namespace(struct pid_namespace *ns)
112 for (i = 0; i < PIDMAP_ENTRIES; i++)
113 kfree(ns->pidmap[i].page);
114 kmem_cache_free(pid_ns_cachep, ns);
117 struct pid_namespace *copy_pid_ns(unsigned long flags, struct pid_namespace *old_ns)
119 if (!(flags & CLONE_NEWPID))
120 return get_pid_ns(old_ns);
121 if (flags & (CLONE_THREAD|CLONE_PARENT))
122 return ERR_PTR(-EINVAL);
123 return create_pid_namespace(old_ns);
126 void free_pid_ns(struct kref *kref)
128 struct pid_namespace *ns, *parent;
130 ns = container_of(kref, struct pid_namespace, kref);
133 destroy_pid_namespace(ns);
139 void zap_pid_ns_processes(struct pid_namespace *pid_ns)
143 struct task_struct *task;
146 * The last thread in the cgroup-init thread group is terminating.
147 * Find remaining pid_ts in the namespace, signal and wait for them
150 * Note: This signals each threads in the namespace - even those that
151 * belong to the same thread group, To avoid this, we would have
152 * to walk the entire tasklist looking a processes in this
153 * namespace, but that could be unnecessarily expensive if the
154 * pid namespace has just a few processes. Or we need to
155 * maintain a tasklist for each pid namespace.
158 read_lock(&tasklist_lock);
159 nr = next_pidmap(pid_ns, 1);
164 * Use force_sig() since it clears SIGNAL_UNKILLABLE ensuring
165 * any nested-container's init processes don't ignore the
168 task = pid_task(find_vpid(nr), PIDTYPE_PID);
170 force_sig(SIGKILL, task);
174 nr = next_pidmap(pid_ns, nr);
176 read_unlock(&tasklist_lock);
179 clear_thread_flag(TIF_SIGPENDING);
180 rc = sys_wait4(-1, NULL, __WALL, NULL);
181 } while (rc != -ECHILD);
183 acct_exit_ns(pid_ns);
187 static __init int pid_namespaces_init(void)
189 pid_ns_cachep = KMEM_CACHE(pid_namespace, SLAB_PANIC);
193 __initcall(pid_namespaces_init);