1 // SPDX-License-Identifier: GPL-2.0-only 1 // SPDX-License-Identifier: GPL-2.0-only
2 /* 2 /*
3 * Pid namespaces 3 * Pid namespaces
4 * 4 *
5 * Authors: 5 * Authors:
6 * (C) 2007 Pavel Emelyanov <xemul@openvz.o 6 * (C) 2007 Pavel Emelyanov <xemul@openvz.org>, OpenVZ, SWsoft Inc.
7 * (C) 2007 Sukadev Bhattiprolu <sukadev@us 7 * (C) 2007 Sukadev Bhattiprolu <sukadev@us.ibm.com>, IBM
8 * Many thanks to Oleg Nesterov for commen 8 * Many thanks to Oleg Nesterov for comments and help
9 * 9 *
10 */ 10 */
11 11
12 #include <linux/pid.h> 12 #include <linux/pid.h>
13 #include <linux/pid_namespace.h> 13 #include <linux/pid_namespace.h>
14 #include <linux/user_namespace.h> 14 #include <linux/user_namespace.h>
15 #include <linux/syscalls.h> 15 #include <linux/syscalls.h>
16 #include <linux/cred.h> 16 #include <linux/cred.h>
17 #include <linux/err.h> 17 #include <linux/err.h>
18 #include <linux/acct.h> 18 #include <linux/acct.h>
19 #include <linux/slab.h> 19 #include <linux/slab.h>
20 #include <linux/proc_ns.h> 20 #include <linux/proc_ns.h>
21 #include <linux/reboot.h> 21 #include <linux/reboot.h>
22 #include <linux/export.h> 22 #include <linux/export.h>
23 #include <linux/sched/task.h> 23 #include <linux/sched/task.h>
24 #include <linux/sched/signal.h> 24 #include <linux/sched/signal.h>
25 #include <linux/idr.h> 25 #include <linux/idr.h>
26 #include <uapi/linux/wait.h> <<
27 #include "pid_sysctl.h" <<
28 26
29 static DEFINE_MUTEX(pid_caches_mutex); 27 static DEFINE_MUTEX(pid_caches_mutex);
30 static struct kmem_cache *pid_ns_cachep; 28 static struct kmem_cache *pid_ns_cachep;
31 /* Write once array, filled from the beginning 29 /* Write once array, filled from the beginning. */
32 static struct kmem_cache *pid_cache[MAX_PID_NS 30 static struct kmem_cache *pid_cache[MAX_PID_NS_LEVEL];
33 31
34 /* 32 /*
35 * creates the kmem cache to allocate pids fro 33 * creates the kmem cache to allocate pids from.
36 * @level: pid namespace level 34 * @level: pid namespace level
37 */ 35 */
38 36
39 static struct kmem_cache *create_pid_cachep(un 37 static struct kmem_cache *create_pid_cachep(unsigned int level)
40 { 38 {
41 /* Level 0 is init_pid_ns.pid_cachep * 39 /* Level 0 is init_pid_ns.pid_cachep */
42 struct kmem_cache **pkc = &pid_cache[l 40 struct kmem_cache **pkc = &pid_cache[level - 1];
43 struct kmem_cache *kc; 41 struct kmem_cache *kc;
44 char name[4 + 10 + 1]; 42 char name[4 + 10 + 1];
45 unsigned int len; 43 unsigned int len;
46 44
47 kc = READ_ONCE(*pkc); 45 kc = READ_ONCE(*pkc);
48 if (kc) 46 if (kc)
49 return kc; 47 return kc;
50 48
51 snprintf(name, sizeof(name), "pid_%u", 49 snprintf(name, sizeof(name), "pid_%u", level + 1);
52 len = struct_size_t(struct pid, number !! 50 len = sizeof(struct pid) + level * sizeof(struct upid);
53 mutex_lock(&pid_caches_mutex); 51 mutex_lock(&pid_caches_mutex);
54 /* Name collision forces to do allocat 52 /* Name collision forces to do allocation under mutex. */
55 if (!*pkc) 53 if (!*pkc)
56 *pkc = kmem_cache_create(name, 54 *pkc = kmem_cache_create(name, len, 0,
57 SLAB_ !! 55 SLAB_HWCACHE_ALIGN | SLAB_ACCOUNT, 0);
58 mutex_unlock(&pid_caches_mutex); 56 mutex_unlock(&pid_caches_mutex);
59 /* current can fail, but someone else 57 /* current can fail, but someone else can succeed. */
60 return READ_ONCE(*pkc); 58 return READ_ONCE(*pkc);
61 } 59 }
62 60
63 static struct ucounts *inc_pid_namespaces(stru 61 static struct ucounts *inc_pid_namespaces(struct user_namespace *ns)
64 { 62 {
65 return inc_ucount(ns, current_euid(), 63 return inc_ucount(ns, current_euid(), UCOUNT_PID_NAMESPACES);
66 } 64 }
67 65
68 static void dec_pid_namespaces(struct ucounts 66 static void dec_pid_namespaces(struct ucounts *ucounts)
69 { 67 {
70 dec_ucount(ucounts, UCOUNT_PID_NAMESPA 68 dec_ucount(ucounts, UCOUNT_PID_NAMESPACES);
71 } 69 }
72 70
73 static struct pid_namespace *create_pid_namesp 71 static struct pid_namespace *create_pid_namespace(struct user_namespace *user_ns,
74 struct pid_namespace *parent_pid_ns) 72 struct pid_namespace *parent_pid_ns)
75 { 73 {
76 struct pid_namespace *ns; 74 struct pid_namespace *ns;
77 unsigned int level = parent_pid_ns->le 75 unsigned int level = parent_pid_ns->level + 1;
78 struct ucounts *ucounts; 76 struct ucounts *ucounts;
79 int err; 77 int err;
80 78
81 err = -EINVAL; 79 err = -EINVAL;
82 if (!in_userns(parent_pid_ns->user_ns, 80 if (!in_userns(parent_pid_ns->user_ns, user_ns))
83 goto out; 81 goto out;
84 82
85 err = -ENOSPC; 83 err = -ENOSPC;
86 if (level > MAX_PID_NS_LEVEL) 84 if (level > MAX_PID_NS_LEVEL)
87 goto out; 85 goto out;
88 ucounts = inc_pid_namespaces(user_ns); 86 ucounts = inc_pid_namespaces(user_ns);
89 if (!ucounts) 87 if (!ucounts)
90 goto out; 88 goto out;
91 89
92 err = -ENOMEM; 90 err = -ENOMEM;
93 ns = kmem_cache_zalloc(pid_ns_cachep, 91 ns = kmem_cache_zalloc(pid_ns_cachep, GFP_KERNEL);
94 if (ns == NULL) 92 if (ns == NULL)
95 goto out_dec; 93 goto out_dec;
96 94
97 idr_init(&ns->idr); 95 idr_init(&ns->idr);
98 96
99 ns->pid_cachep = create_pid_cachep(lev 97 ns->pid_cachep = create_pid_cachep(level);
100 if (ns->pid_cachep == NULL) 98 if (ns->pid_cachep == NULL)
101 goto out_free_idr; 99 goto out_free_idr;
102 100
103 err = ns_alloc_inum(&ns->ns); 101 err = ns_alloc_inum(&ns->ns);
104 if (err) 102 if (err)
105 goto out_free_idr; 103 goto out_free_idr;
106 ns->ns.ops = &pidns_operations; 104 ns->ns.ops = &pidns_operations;
107 105
108 refcount_set(&ns->ns.count, 1); 106 refcount_set(&ns->ns.count, 1);
109 ns->level = level; 107 ns->level = level;
110 ns->parent = get_pid_ns(parent_pid_ns) 108 ns->parent = get_pid_ns(parent_pid_ns);
111 ns->user_ns = get_user_ns(user_ns); 109 ns->user_ns = get_user_ns(user_ns);
112 ns->ucounts = ucounts; 110 ns->ucounts = ucounts;
113 ns->pid_allocated = PIDNS_ADDING; 111 ns->pid_allocated = PIDNS_ADDING;
114 #if defined(CONFIG_SYSCTL) && defined(CONFIG_M !! 112
115 ns->memfd_noexec_scope = pidns_memfd_n <<
116 #endif <<
117 return ns; 113 return ns;
118 114
119 out_free_idr: 115 out_free_idr:
120 idr_destroy(&ns->idr); 116 idr_destroy(&ns->idr);
121 kmem_cache_free(pid_ns_cachep, ns); 117 kmem_cache_free(pid_ns_cachep, ns);
122 out_dec: 118 out_dec:
123 dec_pid_namespaces(ucounts); 119 dec_pid_namespaces(ucounts);
124 out: 120 out:
125 return ERR_PTR(err); 121 return ERR_PTR(err);
126 } 122 }
127 123
128 static void delayed_free_pidns(struct rcu_head 124 static void delayed_free_pidns(struct rcu_head *p)
129 { 125 {
130 struct pid_namespace *ns = container_o 126 struct pid_namespace *ns = container_of(p, struct pid_namespace, rcu);
131 127
132 dec_pid_namespaces(ns->ucounts); 128 dec_pid_namespaces(ns->ucounts);
133 put_user_ns(ns->user_ns); 129 put_user_ns(ns->user_ns);
134 130
135 kmem_cache_free(pid_ns_cachep, ns); 131 kmem_cache_free(pid_ns_cachep, ns);
136 } 132 }
137 133
138 static void destroy_pid_namespace(struct pid_n 134 static void destroy_pid_namespace(struct pid_namespace *ns)
139 { 135 {
140 ns_free_inum(&ns->ns); 136 ns_free_inum(&ns->ns);
141 137
142 idr_destroy(&ns->idr); 138 idr_destroy(&ns->idr);
143 call_rcu(&ns->rcu, delayed_free_pidns) 139 call_rcu(&ns->rcu, delayed_free_pidns);
144 } 140 }
145 141
146 struct pid_namespace *copy_pid_ns(unsigned lon 142 struct pid_namespace *copy_pid_ns(unsigned long flags,
147 struct user_namespace *user_ns, struct 143 struct user_namespace *user_ns, struct pid_namespace *old_ns)
148 { 144 {
149 if (!(flags & CLONE_NEWPID)) 145 if (!(flags & CLONE_NEWPID))
150 return get_pid_ns(old_ns); 146 return get_pid_ns(old_ns);
151 if (task_active_pid_ns(current) != old 147 if (task_active_pid_ns(current) != old_ns)
152 return ERR_PTR(-EINVAL); 148 return ERR_PTR(-EINVAL);
153 return create_pid_namespace(user_ns, o 149 return create_pid_namespace(user_ns, old_ns);
154 } 150 }
155 151
156 void put_pid_ns(struct pid_namespace *ns) 152 void put_pid_ns(struct pid_namespace *ns)
157 { 153 {
158 struct pid_namespace *parent; 154 struct pid_namespace *parent;
159 155
160 while (ns != &init_pid_ns) { 156 while (ns != &init_pid_ns) {
161 parent = ns->parent; 157 parent = ns->parent;
162 if (!refcount_dec_and_test(&ns 158 if (!refcount_dec_and_test(&ns->ns.count))
163 break; 159 break;
164 destroy_pid_namespace(ns); 160 destroy_pid_namespace(ns);
165 ns = parent; 161 ns = parent;
166 } 162 }
167 } 163 }
168 EXPORT_SYMBOL_GPL(put_pid_ns); 164 EXPORT_SYMBOL_GPL(put_pid_ns);
169 165
170 void zap_pid_ns_processes(struct pid_namespace 166 void zap_pid_ns_processes(struct pid_namespace *pid_ns)
171 { 167 {
172 int nr; 168 int nr;
173 int rc; 169 int rc;
174 struct task_struct *task, *me = curren 170 struct task_struct *task, *me = current;
175 int init_pids = thread_group_leader(me 171 int init_pids = thread_group_leader(me) ? 1 : 2;
176 struct pid *pid; 172 struct pid *pid;
177 173
178 /* Don't allow any more processes into 174 /* Don't allow any more processes into the pid namespace */
179 disable_pid_allocation(pid_ns); 175 disable_pid_allocation(pid_ns);
180 176
181 /* 177 /*
182 * Ignore SIGCHLD causing any terminat 178 * Ignore SIGCHLD causing any terminated children to autoreap.
183 * This speeds up the namespace shutdo 179 * This speeds up the namespace shutdown, plus see the comment
184 * below. 180 * below.
185 */ 181 */
186 spin_lock_irq(&me->sighand->siglock); 182 spin_lock_irq(&me->sighand->siglock);
187 me->sighand->action[SIGCHLD - 1].sa.sa 183 me->sighand->action[SIGCHLD - 1].sa.sa_handler = SIG_IGN;
188 spin_unlock_irq(&me->sighand->siglock) 184 spin_unlock_irq(&me->sighand->siglock);
189 185
190 /* 186 /*
191 * The last thread in the cgroup-init 187 * The last thread in the cgroup-init thread group is terminating.
192 * Find remaining pid_ts in the namesp 188 * Find remaining pid_ts in the namespace, signal and wait for them
193 * to exit. 189 * to exit.
194 * 190 *
195 * Note: This signals each threads in 191 * Note: This signals each threads in the namespace - even those that
196 * belong to the same thread gr 192 * belong to the same thread group, To avoid this, we would have
197 * to walk the entire tasklist 193 * to walk the entire tasklist looking a processes in this
198 * namespace, but that could be 194 * namespace, but that could be unnecessarily expensive if the
199 * pid namespace has just a few 195 * pid namespace has just a few processes. Or we need to
200 * maintain a tasklist for each 196 * maintain a tasklist for each pid namespace.
201 * 197 *
202 */ 198 */
203 rcu_read_lock(); 199 rcu_read_lock();
204 read_lock(&tasklist_lock); 200 read_lock(&tasklist_lock);
205 nr = 2; 201 nr = 2;
206 idr_for_each_entry_continue(&pid_ns->i 202 idr_for_each_entry_continue(&pid_ns->idr, pid, nr) {
207 task = pid_task(pid, PIDTYPE_P 203 task = pid_task(pid, PIDTYPE_PID);
208 if (task && !__fatal_signal_pe 204 if (task && !__fatal_signal_pending(task))
209 group_send_sig_info(SI 205 group_send_sig_info(SIGKILL, SEND_SIG_PRIV, task, PIDTYPE_MAX);
210 } 206 }
211 read_unlock(&tasklist_lock); 207 read_unlock(&tasklist_lock);
212 rcu_read_unlock(); 208 rcu_read_unlock();
213 209
214 /* 210 /*
215 * Reap the EXIT_ZOMBIE children we ha 211 * Reap the EXIT_ZOMBIE children we had before we ignored SIGCHLD.
216 * kernel_wait4() will also block unti 212 * kernel_wait4() will also block until our children traced from the
217 * parent namespace are detached and b 213 * parent namespace are detached and become EXIT_DEAD.
218 */ 214 */
219 do { 215 do {
220 clear_thread_flag(TIF_SIGPENDI 216 clear_thread_flag(TIF_SIGPENDING);
221 clear_thread_flag(TIF_NOTIFY_S <<
222 rc = kernel_wait4(-1, NULL, __ 217 rc = kernel_wait4(-1, NULL, __WALL, NULL);
223 } while (rc != -ECHILD); 218 } while (rc != -ECHILD);
224 219
225 /* 220 /*
226 * kernel_wait4() misses EXIT_DEAD chi 221 * kernel_wait4() misses EXIT_DEAD children, and EXIT_ZOMBIE
227 * process whose parents processes are 222 * process whose parents processes are outside of the pid
228 * namespace. Such processes are crea 223 * namespace. Such processes are created with setns()+fork().
229 * 224 *
230 * If those EXIT_ZOMBIE processes are 225 * If those EXIT_ZOMBIE processes are not reaped by their
231 * parents before their parents exit, 226 * parents before their parents exit, they will be reparented
232 * to pid_ns->child_reaper. Thus pidn 227 * to pid_ns->child_reaper. Thus pidns->child_reaper needs to
233 * stay valid until they all go away. 228 * stay valid until they all go away.
234 * 229 *
235 * The code relies on the pid_ns->chil 230 * The code relies on the pid_ns->child_reaper ignoring
236 * SIGCHILD to cause those EXIT_ZOMBIE 231 * SIGCHILD to cause those EXIT_ZOMBIE processes to be
237 * autoreaped if reparented. 232 * autoreaped if reparented.
238 * 233 *
239 * Semantically it is also desirable t 234 * Semantically it is also desirable to wait for EXIT_ZOMBIE
240 * processes before allowing the child 235 * processes before allowing the child_reaper to be reaped, as
241 * that gives the invariant that when 236 * that gives the invariant that when the init process of a
242 * pid namespace is reaped all of the 237 * pid namespace is reaped all of the processes in the pid
243 * namespace are gone. 238 * namespace are gone.
244 * 239 *
245 * Once all of the other tasks are gon 240 * Once all of the other tasks are gone from the pid_namespace
246 * free_pid() will awaken this task. 241 * free_pid() will awaken this task.
247 */ 242 */
248 for (;;) { 243 for (;;) {
249 set_current_state(TASK_INTERRU 244 set_current_state(TASK_INTERRUPTIBLE);
250 if (pid_ns->pid_allocated == i 245 if (pid_ns->pid_allocated == init_pids)
251 break; 246 break;
252 schedule(); 247 schedule();
253 } 248 }
254 __set_current_state(TASK_RUNNING); 249 __set_current_state(TASK_RUNNING);
255 250
256 if (pid_ns->reboot) 251 if (pid_ns->reboot)
257 current->signal->group_exit_co 252 current->signal->group_exit_code = pid_ns->reboot;
258 253
259 acct_exit_ns(pid_ns); 254 acct_exit_ns(pid_ns);
260 return; 255 return;
261 } 256 }
262 257
263 #ifdef CONFIG_CHECKPOINT_RESTORE 258 #ifdef CONFIG_CHECKPOINT_RESTORE
264 static int pid_ns_ctl_handler(const struct ctl !! 259 static int pid_ns_ctl_handler(struct ctl_table *table, int write,
265 void *buffer, size_t *lenp, lo 260 void *buffer, size_t *lenp, loff_t *ppos)
266 { 261 {
267 struct pid_namespace *pid_ns = task_ac 262 struct pid_namespace *pid_ns = task_active_pid_ns(current);
268 struct ctl_table tmp = *table; 263 struct ctl_table tmp = *table;
269 int ret, next; 264 int ret, next;
270 265
271 if (write && !checkpoint_restore_ns_ca 266 if (write && !checkpoint_restore_ns_capable(pid_ns->user_ns))
272 return -EPERM; 267 return -EPERM;
273 268
>> 269 /*
>> 270 * Writing directly to ns' last_pid field is OK, since this field
>> 271 * is volatile in a living namespace anyway and a code writing to
>> 272 * it should synchronize its usage with external means.
>> 273 */
>> 274
274 next = idr_get_cursor(&pid_ns->idr) - 275 next = idr_get_cursor(&pid_ns->idr) - 1;
275 276
276 tmp.data = &next; 277 tmp.data = &next;
277 ret = proc_dointvec_minmax(&tmp, write 278 ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
278 if (!ret && write) 279 if (!ret && write)
279 idr_set_cursor(&pid_ns->idr, n 280 idr_set_cursor(&pid_ns->idr, next + 1);
280 281
281 return ret; 282 return ret;
282 } 283 }
283 284
284 extern int pid_max; 285 extern int pid_max;
285 static struct ctl_table pid_ns_ctl_table[] = { 286 static struct ctl_table pid_ns_ctl_table[] = {
286 { 287 {
287 .procname = "ns_last_pid", 288 .procname = "ns_last_pid",
288 .maxlen = sizeof(int), 289 .maxlen = sizeof(int),
289 .mode = 0666, /* permissions a 290 .mode = 0666, /* permissions are checked in the handler */
290 .proc_handler = pid_ns_ctl_han 291 .proc_handler = pid_ns_ctl_handler,
291 .extra1 = SYSCTL_ZERO, 292 .extra1 = SYSCTL_ZERO,
292 .extra2 = &pid_max, 293 .extra2 = &pid_max,
293 }, 294 },
>> 295 { }
294 }; 296 };
>> 297 static struct ctl_path kern_path[] = { { .procname = "kernel", }, { } };
295 #endif /* CONFIG_CHECKPOINT_RESTORE */ 298 #endif /* CONFIG_CHECKPOINT_RESTORE */
296 299
297 int reboot_pid_ns(struct pid_namespace *pid_ns 300 int reboot_pid_ns(struct pid_namespace *pid_ns, int cmd)
298 { 301 {
299 if (pid_ns == &init_pid_ns) 302 if (pid_ns == &init_pid_ns)
300 return 0; 303 return 0;
301 304
302 switch (cmd) { 305 switch (cmd) {
303 case LINUX_REBOOT_CMD_RESTART2: 306 case LINUX_REBOOT_CMD_RESTART2:
304 case LINUX_REBOOT_CMD_RESTART: 307 case LINUX_REBOOT_CMD_RESTART:
305 pid_ns->reboot = SIGHUP; 308 pid_ns->reboot = SIGHUP;
306 break; 309 break;
307 310
308 case LINUX_REBOOT_CMD_POWER_OFF: 311 case LINUX_REBOOT_CMD_POWER_OFF:
309 case LINUX_REBOOT_CMD_HALT: 312 case LINUX_REBOOT_CMD_HALT:
310 pid_ns->reboot = SIGINT; 313 pid_ns->reboot = SIGINT;
311 break; 314 break;
312 default: 315 default:
313 return -EINVAL; 316 return -EINVAL;
314 } 317 }
315 318
316 read_lock(&tasklist_lock); 319 read_lock(&tasklist_lock);
317 send_sig(SIGKILL, pid_ns->child_reaper 320 send_sig(SIGKILL, pid_ns->child_reaper, 1);
318 read_unlock(&tasklist_lock); 321 read_unlock(&tasklist_lock);
319 322
320 do_exit(0); 323 do_exit(0);
321 324
322 /* Not reached */ 325 /* Not reached */
323 return 0; 326 return 0;
324 } 327 }
325 328
326 static inline struct pid_namespace *to_pid_ns( 329 static inline struct pid_namespace *to_pid_ns(struct ns_common *ns)
327 { 330 {
328 return container_of(ns, struct pid_nam 331 return container_of(ns, struct pid_namespace, ns);
329 } 332 }
330 333
331 static struct ns_common *pidns_get(struct task 334 static struct ns_common *pidns_get(struct task_struct *task)
332 { 335 {
333 struct pid_namespace *ns; 336 struct pid_namespace *ns;
334 337
335 rcu_read_lock(); 338 rcu_read_lock();
336 ns = task_active_pid_ns(task); 339 ns = task_active_pid_ns(task);
337 if (ns) 340 if (ns)
338 get_pid_ns(ns); 341 get_pid_ns(ns);
339 rcu_read_unlock(); 342 rcu_read_unlock();
340 343
341 return ns ? &ns->ns : NULL; 344 return ns ? &ns->ns : NULL;
342 } 345 }
343 346
344 static struct ns_common *pidns_for_children_ge 347 static struct ns_common *pidns_for_children_get(struct task_struct *task)
345 { 348 {
346 struct pid_namespace *ns = NULL; 349 struct pid_namespace *ns = NULL;
347 350
348 task_lock(task); 351 task_lock(task);
349 if (task->nsproxy) { 352 if (task->nsproxy) {
350 ns = task->nsproxy->pid_ns_for 353 ns = task->nsproxy->pid_ns_for_children;
351 get_pid_ns(ns); 354 get_pid_ns(ns);
352 } 355 }
353 task_unlock(task); 356 task_unlock(task);
354 357
355 if (ns) { 358 if (ns) {
356 read_lock(&tasklist_lock); 359 read_lock(&tasklist_lock);
357 if (!ns->child_reaper) { 360 if (!ns->child_reaper) {
358 put_pid_ns(ns); 361 put_pid_ns(ns);
359 ns = NULL; 362 ns = NULL;
360 } 363 }
361 read_unlock(&tasklist_lock); 364 read_unlock(&tasklist_lock);
362 } 365 }
363 366
364 return ns ? &ns->ns : NULL; 367 return ns ? &ns->ns : NULL;
365 } 368 }
366 369
367 static void pidns_put(struct ns_common *ns) 370 static void pidns_put(struct ns_common *ns)
368 { 371 {
369 put_pid_ns(to_pid_ns(ns)); 372 put_pid_ns(to_pid_ns(ns));
370 } 373 }
371 374
372 static int pidns_install(struct nsset *nsset, 375 static int pidns_install(struct nsset *nsset, struct ns_common *ns)
373 { 376 {
374 struct nsproxy *nsproxy = nsset->nspro 377 struct nsproxy *nsproxy = nsset->nsproxy;
375 struct pid_namespace *active = task_ac 378 struct pid_namespace *active = task_active_pid_ns(current);
376 struct pid_namespace *ancestor, *new = 379 struct pid_namespace *ancestor, *new = to_pid_ns(ns);
377 380
378 if (!ns_capable(new->user_ns, CAP_SYS_ 381 if (!ns_capable(new->user_ns, CAP_SYS_ADMIN) ||
379 !ns_capable(nsset->cred->user_ns, 382 !ns_capable(nsset->cred->user_ns, CAP_SYS_ADMIN))
380 return -EPERM; 383 return -EPERM;
381 384
382 /* 385 /*
383 * Only allow entering the current act 386 * Only allow entering the current active pid namespace
384 * or a child of the current active pi 387 * or a child of the current active pid namespace.
385 * 388 *
386 * This is required for fork to return 389 * This is required for fork to return a usable pid value and
387 * this maintains the property that pr 390 * this maintains the property that processes and their
388 * children can not escape their curre 391 * children can not escape their current pid namespace.
389 */ 392 */
390 if (new->level < active->level) 393 if (new->level < active->level)
391 return -EINVAL; 394 return -EINVAL;
392 395
393 ancestor = new; 396 ancestor = new;
394 while (ancestor->level > active->level 397 while (ancestor->level > active->level)
395 ancestor = ancestor->parent; 398 ancestor = ancestor->parent;
396 if (ancestor != active) 399 if (ancestor != active)
397 return -EINVAL; 400 return -EINVAL;
398 401
399 put_pid_ns(nsproxy->pid_ns_for_childre 402 put_pid_ns(nsproxy->pid_ns_for_children);
400 nsproxy->pid_ns_for_children = get_pid 403 nsproxy->pid_ns_for_children = get_pid_ns(new);
401 return 0; 404 return 0;
402 } 405 }
403 406
404 static struct ns_common *pidns_get_parent(stru 407 static struct ns_common *pidns_get_parent(struct ns_common *ns)
405 { 408 {
406 struct pid_namespace *active = task_ac 409 struct pid_namespace *active = task_active_pid_ns(current);
407 struct pid_namespace *pid_ns, *p; 410 struct pid_namespace *pid_ns, *p;
408 411
409 /* See if the parent is in the current 412 /* See if the parent is in the current namespace */
410 pid_ns = p = to_pid_ns(ns)->parent; 413 pid_ns = p = to_pid_ns(ns)->parent;
411 for (;;) { 414 for (;;) {
412 if (!p) 415 if (!p)
413 return ERR_PTR(-EPERM) 416 return ERR_PTR(-EPERM);
414 if (p == active) 417 if (p == active)
415 break; 418 break;
416 p = p->parent; 419 p = p->parent;
417 } 420 }
418 421
419 return &get_pid_ns(pid_ns)->ns; 422 return &get_pid_ns(pid_ns)->ns;
420 } 423 }
421 424
422 static struct user_namespace *pidns_owner(stru 425 static struct user_namespace *pidns_owner(struct ns_common *ns)
423 { 426 {
424 return to_pid_ns(ns)->user_ns; 427 return to_pid_ns(ns)->user_ns;
425 } 428 }
426 429
427 const struct proc_ns_operations pidns_operatio 430 const struct proc_ns_operations pidns_operations = {
428 .name = "pid", 431 .name = "pid",
429 .type = CLONE_NEWPID, 432 .type = CLONE_NEWPID,
430 .get = pidns_get, 433 .get = pidns_get,
431 .put = pidns_put, 434 .put = pidns_put,
432 .install = pidns_install, 435 .install = pidns_install,
433 .owner = pidns_owner, 436 .owner = pidns_owner,
434 .get_parent = pidns_get_parent, 437 .get_parent = pidns_get_parent,
435 }; 438 };
436 439
437 const struct proc_ns_operations pidns_for_chil 440 const struct proc_ns_operations pidns_for_children_operations = {
438 .name = "pid_for_children", 441 .name = "pid_for_children",
439 .real_ns_name = "pid", 442 .real_ns_name = "pid",
440 .type = CLONE_NEWPID, 443 .type = CLONE_NEWPID,
441 .get = pidns_for_children_g 444 .get = pidns_for_children_get,
442 .put = pidns_put, 445 .put = pidns_put,
443 .install = pidns_install, 446 .install = pidns_install,
444 .owner = pidns_owner, 447 .owner = pidns_owner,
445 .get_parent = pidns_get_parent, 448 .get_parent = pidns_get_parent,
446 }; 449 };
447 450
448 static __init int pid_namespaces_init(void) 451 static __init int pid_namespaces_init(void)
449 { 452 {
450 pid_ns_cachep = KMEM_CACHE(pid_namespa 453 pid_ns_cachep = KMEM_CACHE(pid_namespace, SLAB_PANIC | SLAB_ACCOUNT);
451 454
452 #ifdef CONFIG_CHECKPOINT_RESTORE 455 #ifdef CONFIG_CHECKPOINT_RESTORE
453 register_sysctl_init("kernel", pid_ns_ !! 456 register_sysctl_paths(kern_path, pid_ns_ctl_table);
454 #endif 457 #endif
455 <<
456 register_pid_ns_sysctl_table_vm(); <<
457 return 0; 458 return 0;
458 } 459 }
459 460
460 __initcall(pid_namespaces_init); 461 __initcall(pid_namespaces_init);
461 462
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.