Commit | Line | Data |
---|---|---|
74bd59bb PE |
1 | /* |
2 | * Pid namespaces | |
3 | * | |
4 | * Authors: | |
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 | |
8 | * | |
9 | */ | |
10 | ||
11 | #include <linux/pid.h> | |
12 | #include <linux/pid_namespace.h> | |
13 | #include <linux/syscalls.h> | |
14 | #include <linux/err.h> | |
0b6b030f | 15 | #include <linux/acct.h> |
5a0e3ad6 | 16 | #include <linux/slab.h> |
4308eebb | 17 | #include <linux/proc_fs.h> |
cf3f8921 | 18 | #include <linux/reboot.h> |
74bd59bb PE |
19 | |
20 | #define BITS_PER_PAGE (PAGE_SIZE*8) | |
21 | ||
22 | struct pid_cache { | |
23 | int nr_ids; | |
24 | char name[16]; | |
25 | struct kmem_cache *cachep; | |
26 | struct list_head list; | |
27 | }; | |
28 | ||
29 | static LIST_HEAD(pid_caches_lh); | |
30 | static DEFINE_MUTEX(pid_caches_mutex); | |
31 | static struct kmem_cache *pid_ns_cachep; | |
32 | ||
33 | /* | |
34 | * creates the kmem cache to allocate pids from. | |
35 | * @nr_ids: the number of numerical ids this pid will have to carry | |
36 | */ | |
37 | ||
38 | static struct kmem_cache *create_pid_cachep(int nr_ids) | |
39 | { | |
40 | struct pid_cache *pcache; | |
41 | struct kmem_cache *cachep; | |
42 | ||
43 | mutex_lock(&pid_caches_mutex); | |
44 | list_for_each_entry(pcache, &pid_caches_lh, list) | |
45 | if (pcache->nr_ids == nr_ids) | |
46 | goto out; | |
47 | ||
48 | pcache = kmalloc(sizeof(struct pid_cache), GFP_KERNEL); | |
49 | if (pcache == NULL) | |
50 | goto err_alloc; | |
51 | ||
52 | snprintf(pcache->name, sizeof(pcache->name), "pid_%d", nr_ids); | |
53 | cachep = kmem_cache_create(pcache->name, | |
54 | sizeof(struct pid) + (nr_ids - 1) * sizeof(struct upid), | |
55 | 0, SLAB_HWCACHE_ALIGN, NULL); | |
56 | if (cachep == NULL) | |
57 | goto err_cachep; | |
58 | ||
59 | pcache->nr_ids = nr_ids; | |
60 | pcache->cachep = cachep; | |
61 | list_add(&pcache->list, &pid_caches_lh); | |
62 | out: | |
63 | mutex_unlock(&pid_caches_mutex); | |
64 | return pcache->cachep; | |
65 | ||
66 | err_cachep: | |
67 | kfree(pcache); | |
68 | err_alloc: | |
69 | mutex_unlock(&pid_caches_mutex); | |
70 | return NULL; | |
71 | } | |
72 | ||
ed469a63 | 73 | static struct pid_namespace *create_pid_namespace(struct pid_namespace *parent_pid_ns) |
74bd59bb PE |
74 | { |
75 | struct pid_namespace *ns; | |
ed469a63 | 76 | unsigned int level = parent_pid_ns->level + 1; |
4308eebb | 77 | int i, err = -ENOMEM; |
74bd59bb | 78 | |
84406c15 | 79 | ns = kmem_cache_zalloc(pid_ns_cachep, GFP_KERNEL); |
74bd59bb PE |
80 | if (ns == NULL) |
81 | goto out; | |
82 | ||
83 | ns->pidmap[0].page = kzalloc(PAGE_SIZE, GFP_KERNEL); | |
84 | if (!ns->pidmap[0].page) | |
85 | goto out_free; | |
86 | ||
87 | ns->pid_cachep = create_pid_cachep(level + 1); | |
88 | if (ns->pid_cachep == NULL) | |
89 | goto out_free_map; | |
90 | ||
91 | kref_init(&ns->kref); | |
74bd59bb | 92 | ns->level = level; |
ed469a63 | 93 | ns->parent = get_pid_ns(parent_pid_ns); |
74bd59bb PE |
94 | |
95 | set_bit(0, ns->pidmap[0].page); | |
96 | atomic_set(&ns->pidmap[0].nr_free, BITS_PER_PAGE - 1); | |
97 | ||
84406c15 | 98 | for (i = 1; i < PIDMAP_ENTRIES; i++) |
74bd59bb | 99 | atomic_set(&ns->pidmap[i].nr_free, BITS_PER_PAGE); |
74bd59bb | 100 | |
4308eebb EB |
101 | err = pid_ns_prepare_proc(ns); |
102 | if (err) | |
103 | goto out_put_parent_pid_ns; | |
104 | ||
74bd59bb PE |
105 | return ns; |
106 | ||
4308eebb EB |
107 | out_put_parent_pid_ns: |
108 | put_pid_ns(parent_pid_ns); | |
74bd59bb PE |
109 | out_free_map: |
110 | kfree(ns->pidmap[0].page); | |
111 | out_free: | |
112 | kmem_cache_free(pid_ns_cachep, ns); | |
113 | out: | |
4308eebb | 114 | return ERR_PTR(err); |
74bd59bb PE |
115 | } |
116 | ||
117 | static void destroy_pid_namespace(struct pid_namespace *ns) | |
118 | { | |
119 | int i; | |
120 | ||
121 | for (i = 0; i < PIDMAP_ENTRIES; i++) | |
122 | kfree(ns->pidmap[i].page); | |
123 | kmem_cache_free(pid_ns_cachep, ns); | |
124 | } | |
125 | ||
126 | struct pid_namespace *copy_pid_ns(unsigned long flags, struct pid_namespace *old_ns) | |
127 | { | |
74bd59bb | 128 | if (!(flags & CLONE_NEWPID)) |
dca4a979 | 129 | return get_pid_ns(old_ns); |
e5a47386 | 130 | if (flags & (CLONE_THREAD|CLONE_PARENT)) |
dca4a979 AD |
131 | return ERR_PTR(-EINVAL); |
132 | return create_pid_namespace(old_ns); | |
74bd59bb PE |
133 | } |
134 | ||
135 | void free_pid_ns(struct kref *kref) | |
136 | { | |
137 | struct pid_namespace *ns, *parent; | |
138 | ||
139 | ns = container_of(kref, struct pid_namespace, kref); | |
140 | ||
141 | parent = ns->parent; | |
142 | destroy_pid_namespace(ns); | |
143 | ||
144 | if (parent != NULL) | |
145 | put_pid_ns(parent); | |
146 | } | |
147 | ||
148 | void zap_pid_ns_processes(struct pid_namespace *pid_ns) | |
149 | { | |
150 | int nr; | |
151 | int rc; | |
e4da026f | 152 | struct task_struct *task; |
74bd59bb PE |
153 | |
154 | /* | |
155 | * The last thread in the cgroup-init thread group is terminating. | |
156 | * Find remaining pid_ts in the namespace, signal and wait for them | |
157 | * to exit. | |
158 | * | |
159 | * Note: This signals each threads in the namespace - even those that | |
160 | * belong to the same thread group, To avoid this, we would have | |
161 | * to walk the entire tasklist looking a processes in this | |
162 | * namespace, but that could be unnecessarily expensive if the | |
163 | * pid namespace has just a few processes. Or we need to | |
164 | * maintain a tasklist for each pid namespace. | |
165 | * | |
166 | */ | |
167 | read_lock(&tasklist_lock); | |
168 | nr = next_pidmap(pid_ns, 1); | |
169 | while (nr > 0) { | |
e4da026f SB |
170 | rcu_read_lock(); |
171 | ||
e4da026f | 172 | task = pid_task(find_vpid(nr), PIDTYPE_PID); |
a02d6fd6 ON |
173 | if (task && !__fatal_signal_pending(task)) |
174 | send_sig_info(SIGKILL, SEND_SIG_FORCED, task); | |
e4da026f SB |
175 | |
176 | rcu_read_unlock(); | |
177 | ||
74bd59bb PE |
178 | nr = next_pidmap(pid_ns, nr); |
179 | } | |
180 | read_unlock(&tasklist_lock); | |
181 | ||
182 | do { | |
183 | clear_thread_flag(TIF_SIGPENDING); | |
184 | rc = sys_wait4(-1, NULL, __WALL, NULL); | |
185 | } while (rc != -ECHILD); | |
186 | ||
cf3f8921 DL |
187 | if (pid_ns->reboot) |
188 | current->signal->group_exit_code = pid_ns->reboot; | |
189 | ||
0b6b030f | 190 | acct_exit_ns(pid_ns); |
74bd59bb PE |
191 | return; |
192 | } | |
193 | ||
b8f566b0 PE |
194 | static int pid_ns_ctl_handler(struct ctl_table *table, int write, |
195 | void __user *buffer, size_t *lenp, loff_t *ppos) | |
196 | { | |
197 | struct ctl_table tmp = *table; | |
198 | ||
199 | if (write && !capable(CAP_SYS_ADMIN)) | |
200 | return -EPERM; | |
201 | ||
202 | /* | |
203 | * Writing directly to ns' last_pid field is OK, since this field | |
204 | * is volatile in a living namespace anyway and a code writing to | |
205 | * it should synchronize its usage with external means. | |
206 | */ | |
207 | ||
208 | tmp.data = ¤t->nsproxy->pid_ns->last_pid; | |
209 | return proc_dointvec(&tmp, write, buffer, lenp, ppos); | |
210 | } | |
211 | ||
212 | static struct ctl_table pid_ns_ctl_table[] = { | |
213 | { | |
214 | .procname = "ns_last_pid", | |
215 | .maxlen = sizeof(int), | |
216 | .mode = 0666, /* permissions are checked in the handler */ | |
217 | .proc_handler = pid_ns_ctl_handler, | |
218 | }, | |
219 | { } | |
220 | }; | |
221 | ||
222 | static struct ctl_path kern_path[] = { { .procname = "kernel", }, { } }; | |
223 | ||
cf3f8921 DL |
224 | int reboot_pid_ns(struct pid_namespace *pid_ns, int cmd) |
225 | { | |
226 | if (pid_ns == &init_pid_ns) | |
227 | return 0; | |
228 | ||
229 | switch (cmd) { | |
230 | case LINUX_REBOOT_CMD_RESTART2: | |
231 | case LINUX_REBOOT_CMD_RESTART: | |
232 | pid_ns->reboot = SIGHUP; | |
233 | break; | |
234 | ||
235 | case LINUX_REBOOT_CMD_POWER_OFF: | |
236 | case LINUX_REBOOT_CMD_HALT: | |
237 | pid_ns->reboot = SIGINT; | |
238 | break; | |
239 | default: | |
240 | return -EINVAL; | |
241 | } | |
242 | ||
243 | read_lock(&tasklist_lock); | |
244 | force_sig(SIGKILL, pid_ns->child_reaper); | |
245 | read_unlock(&tasklist_lock); | |
246 | ||
247 | do_exit(0); | |
248 | ||
249 | /* Not reached */ | |
250 | return 0; | |
251 | } | |
252 | ||
74bd59bb PE |
253 | static __init int pid_namespaces_init(void) |
254 | { | |
255 | pid_ns_cachep = KMEM_CACHE(pid_namespace, SLAB_PANIC); | |
b8f566b0 | 256 | register_sysctl_paths(kern_path, pid_ns_ctl_table); |
74bd59bb PE |
257 | return 0; |
258 | } | |
259 | ||
260 | __initcall(pid_namespaces_init); |