blob: 2278e249141dbad205f598c5857e6ea83b159077 [file] [log] [blame]
Thomas Gleixner457c8992019-05-19 13:08:55 +01001// SPDX-License-Identifier: GPL-2.0-only
Linus Torvalds1da177e2005-04-16 15:20:36 -07002/*
3 * Generic pidhash and scalable, time-bounded PID allocator
4 *
Nadia Yvette Chambers6d49e352012-12-06 10:39:54 +01005 * (C) 2002-2003 Nadia Yvette Chambers, IBM
6 * (C) 2004 Nadia Yvette Chambers, Oracle
Linus Torvalds1da177e2005-04-16 15:20:36 -07007 * (C) 2002-2004 Ingo Molnar, Red Hat
8 *
9 * pid-structures are backing objects for tasks sharing a given ID to chain
10 * against. There is very little to them aside from hashing them and
11 * parking tasks using given ID's on a list.
12 *
13 * The hash is always changed with the tasklist_lock write-acquired,
14 * and the hash is only accessed with the tasklist_lock at least
15 * read-acquired, so there's no additional SMP locking needed here.
16 *
17 * We have a list of bitmap pages, which bitmaps represent the PID space.
18 * Allocating and freeing PIDs is completely lockless. The worst-case
19 * allocation scenario when all but one out of 1 million PIDs possible are
20 * allocated already: the scanning of 32 list entries and at most PAGE_SIZE
21 * bytes. The typical fastpath is a single successful setbit. Freeing is O(1).
Pavel Emelyanov30e49c22007-10-18 23:40:10 -070022 *
23 * Pid namespaces:
24 * (C) 2007 Pavel Emelyanov <xemul@openvz.org>, OpenVZ, SWsoft Inc.
25 * (C) 2007 Sukadev Bhattiprolu <sukadev@us.ibm.com>, IBM
26 * Many thanks to Oleg Nesterov for comments and help
27 *
Linus Torvalds1da177e2005-04-16 15:20:36 -070028 */
29
30#include <linux/mm.h>
Paul Gortmaker9984de12011-05-23 14:51:41 -040031#include <linux/export.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070032#include <linux/slab.h>
33#include <linux/init.h>
Franck Bui-Huu82524742008-05-12 21:21:05 +020034#include <linux/rculist.h>
Mike Rapoport57c8a662018-10-30 15:09:49 -070035#include <linux/memblock.h>
Sukadev Bhattiprolu61a58c62006-12-08 02:37:58 -080036#include <linux/pid_namespace.h>
Sukadev Bhattiprolu820e45d2007-05-10 22:23:00 -070037#include <linux/init_task.h>
Sukadev Bhattiprolu3eb07c82007-10-18 23:40:13 -070038#include <linux/syscalls.h>
David Howells0bb80f22013-04-12 01:50:06 +010039#include <linux/proc_ns.h>
Joel Fernandes (Google)f57e5152019-07-16 16:30:06 -070040#include <linux/refcount.h>
Christian Brauner32fcb422019-05-24 12:43:51 +020041#include <linux/anon_inodes.h>
42#include <linux/sched/signal.h>
Ingo Molnar29930022017-02-08 18:51:36 +010043#include <linux/sched/task.h>
Gargi Sharma95846ec2017-11-17 15:30:30 -080044#include <linux/idr.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070045
David Howellse1e871a2018-01-02 15:12:01 +000046struct pid init_struct_pid = {
Joel Fernandes (Google)f57e5152019-07-16 16:30:06 -070047 .count = REFCOUNT_INIT(1),
David Howellse1e871a2018-01-02 15:12:01 +000048 .tasks = {
49 { .first = NULL },
50 { .first = NULL },
51 { .first = NULL },
52 },
53 .level = 0,
54 .numbers = { {
55 .nr = 0,
56 .ns = &init_pid_ns,
57 }, }
58};
Linus Torvalds1da177e2005-04-16 15:20:36 -070059
60int pid_max = PID_MAX_DEFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -070061
62#define RESERVED_PIDS 300
63
64int pid_max_min = RESERVED_PIDS + 1;
65int pid_max_max = PID_MAX_LIMIT;
66
Linus Torvalds1da177e2005-04-16 15:20:36 -070067/*
68 * PID-map pages start out as NULL, they get allocated upon
69 * first use and are never deallocated. This way a low pid_max
70 * value does not cause lots of bitmaps to be allocated, but
71 * the scheme scales to up to 4 million PIDs, runtime.
72 */
Sukadev Bhattiprolu61a58c62006-12-08 02:37:58 -080073struct pid_namespace init_pid_ns = {
Peter Zijlstra1e24edc2016-11-14 17:12:23 +010074 .kref = KREF_INIT(2),
Matthew Wilcoxf6bb2a22018-04-10 16:36:52 -070075 .idr = IDR_INIT(init_pid_ns.idr),
Gargi Sharmae8cfbc22017-11-17 15:30:34 -080076 .pid_allocated = PIDNS_ADDING,
Pavel Emelyanovfaacbfd2007-10-18 23:40:04 -070077 .level = 0,
78 .child_reaper = &init_task,
Eric W. Biederman49f4d8b2012-08-02 04:25:10 -070079 .user_ns = &init_user_ns,
Al Viro435d5f42014-10-31 22:56:04 -040080 .ns.inum = PROC_PID_INIT_INO,
Al Viro33c42942014-11-01 02:32:53 -040081#ifdef CONFIG_PID_NS
82 .ns.ops = &pidns_operations,
83#endif
Sukadev Bhattiprolu3fbc9642006-10-02 02:17:24 -070084};
Pavel Emelyanov198fe212007-10-18 23:40:06 -070085EXPORT_SYMBOL_GPL(init_pid_ns);
Linus Torvalds1da177e2005-04-16 15:20:36 -070086
Eric W. Biederman92476d72006-03-31 02:31:42 -080087/*
88 * Note: disable interrupts while the pidmap_lock is held as an
89 * interrupt might come in and do read_lock(&tasklist_lock).
90 *
91 * If we don't disable interrupts there is a nasty deadlock between
92 * detach_pid()->free_pid() and another cpu that does
93 * spin_lock(&pidmap_lock) followed by an interrupt routine that does
94 * read_lock(&tasklist_lock);
95 *
96 * After we clean up the tasklist_lock and know there are no
97 * irq handlers that take it we can leave the interrupts enabled.
98 * For now it is easier to be safe than to prove it can't happen.
99 */
Sukadev Bhattiprolu3fbc9642006-10-02 02:17:24 -0700100
Linus Torvalds1da177e2005-04-16 15:20:36 -0700101static __cacheline_aligned_in_smp DEFINE_SPINLOCK(pidmap_lock);
102
Harvey Harrison7ad5b3a2008-02-08 04:19:53 -0800103void put_pid(struct pid *pid)
Eric W. Biederman92476d72006-03-31 02:31:42 -0800104{
Pavel Emelianovbaf8f0f2007-10-18 23:39:48 -0700105 struct pid_namespace *ns;
106
Eric W. Biederman92476d72006-03-31 02:31:42 -0800107 if (!pid)
108 return;
Pavel Emelianovbaf8f0f2007-10-18 23:39:48 -0700109
Pavel Emelyanov8ef047a2007-10-18 23:40:05 -0700110 ns = pid->numbers[pid->level].ns;
Joel Fernandes (Google)f57e5152019-07-16 16:30:06 -0700111 if (refcount_dec_and_test(&pid->count)) {
Pavel Emelianovbaf8f0f2007-10-18 23:39:48 -0700112 kmem_cache_free(ns->pid_cachep, pid);
Pavel Emelyanovb461cc02007-10-18 23:40:09 -0700113 put_pid_ns(ns);
Pavel Emelyanov8ef047a2007-10-18 23:40:05 -0700114 }
Eric W. Biederman92476d72006-03-31 02:31:42 -0800115}
Eric W. Biedermanbbf73142006-10-02 02:17:11 -0700116EXPORT_SYMBOL_GPL(put_pid);
Eric W. Biederman92476d72006-03-31 02:31:42 -0800117
118static void delayed_put_pid(struct rcu_head *rhp)
119{
120 struct pid *pid = container_of(rhp, struct pid, rcu);
121 put_pid(pid);
122}
123
Harvey Harrison7ad5b3a2008-02-08 04:19:53 -0800124void free_pid(struct pid *pid)
Eric W. Biederman92476d72006-03-31 02:31:42 -0800125{
126 /* We can be called with write_lock_irq(&tasklist_lock) held */
Pavel Emelyanov8ef047a2007-10-18 23:40:05 -0700127 int i;
Eric W. Biederman92476d72006-03-31 02:31:42 -0800128 unsigned long flags;
129
130 spin_lock_irqsave(&pidmap_lock, flags);
Eric W. Biederman0a01f2c2012-08-01 10:33:47 -0700131 for (i = 0; i <= pid->level; i++) {
132 struct upid *upid = pid->numbers + i;
Eric W. Biedermanaf4b8a82012-08-01 15:03:42 -0700133 struct pid_namespace *ns = upid->ns;
Gargi Sharmae8cfbc22017-11-17 15:30:34 -0800134 switch (--ns->pid_allocated) {
Eric W. Biedermana6064882013-08-29 13:56:50 -0700135 case 2:
Eric W. Biedermanaf4b8a82012-08-01 15:03:42 -0700136 case 1:
137 /* When all that is left in the pid namespace
138 * is the reaper wake up the reaper. The reaper
139 * may be sleeping in zap_pid_ns_processes().
140 */
141 wake_up_process(ns->child_reaper);
142 break;
Gargi Sharmae8cfbc22017-11-17 15:30:34 -0800143 case PIDNS_ADDING:
Oleg Nesterov314a8ad2013-09-30 13:45:27 -0700144 /* Handle a fork failure of the first process */
145 WARN_ON(ns->child_reaper);
Gargi Sharmae8cfbc22017-11-17 15:30:34 -0800146 ns->pid_allocated = 0;
Oleg Nesterov314a8ad2013-09-30 13:45:27 -0700147 /* fall through */
Eric W. Biedermanaf4b8a82012-08-01 15:03:42 -0700148 case 0:
Eric W. Biedermanaf4b8a82012-08-01 15:03:42 -0700149 schedule_work(&ns->proc_work);
150 break;
Eric W. Biederman5e1182deb2010-07-12 18:50:25 -0700151 }
Gargi Sharma95846ec2017-11-17 15:30:30 -0800152
153 idr_remove(&ns->idr, upid->nr);
Eric W. Biederman0a01f2c2012-08-01 10:33:47 -0700154 }
Eric W. Biederman92476d72006-03-31 02:31:42 -0800155 spin_unlock_irqrestore(&pidmap_lock, flags);
156
Eric W. Biederman92476d72006-03-31 02:31:42 -0800157 call_rcu(&pid->rcu, delayed_put_pid);
158}
159
Adrian Reber49cb2fc2019-11-15 13:36:20 +0100160struct pid *alloc_pid(struct pid_namespace *ns, pid_t *set_tid,
161 size_t set_tid_size)
Eric W. Biederman92476d72006-03-31 02:31:42 -0800162{
163 struct pid *pid;
164 enum pid_type type;
Pavel Emelyanov8ef047a2007-10-18 23:40:05 -0700165 int i, nr;
166 struct pid_namespace *tmp;
Pavel Emelyanov198fe212007-10-18 23:40:06 -0700167 struct upid *upid;
Michal Hocko35f71bc2015-04-16 12:47:38 -0700168 int retval = -ENOMEM;
Eric W. Biederman92476d72006-03-31 02:31:42 -0800169
Adrian Reber49cb2fc2019-11-15 13:36:20 +0100170 /*
171 * set_tid_size contains the size of the set_tid array. Starting at
172 * the most nested currently active PID namespace it tells alloc_pid()
173 * which PID to set for a process in that most nested PID namespace
174 * up to set_tid_size PID namespaces. It does not have to set the PID
175 * for a process in all nested PID namespaces but set_tid_size must
176 * never be greater than the current ns->level + 1.
177 */
178 if (set_tid_size > ns->level + 1)
179 return ERR_PTR(-EINVAL);
180
Pavel Emelianovbaf8f0f2007-10-18 23:39:48 -0700181 pid = kmem_cache_alloc(ns->pid_cachep, GFP_KERNEL);
Eric W. Biederman92476d72006-03-31 02:31:42 -0800182 if (!pid)
Michal Hocko35f71bc2015-04-16 12:47:38 -0700183 return ERR_PTR(retval);
Eric W. Biederman92476d72006-03-31 02:31:42 -0800184
Pavel Emelyanov8ef047a2007-10-18 23:40:05 -0700185 tmp = ns;
Eric W. Biederman0a01f2c2012-08-01 10:33:47 -0700186 pid->level = ns->level;
Gargi Sharma95846ec2017-11-17 15:30:30 -0800187
Pavel Emelyanov8ef047a2007-10-18 23:40:05 -0700188 for (i = ns->level; i >= 0; i--) {
Adrian Reber49cb2fc2019-11-15 13:36:20 +0100189 int tid = 0;
190
191 if (set_tid_size) {
192 tid = set_tid[ns->level - i];
193
194 retval = -EINVAL;
195 if (tid < 1 || tid >= pid_max)
196 goto out_free;
197 /*
198 * Also fail if a PID != 1 is requested and
199 * no PID 1 exists.
200 */
201 if (tid != 1 && !tmp->child_reaper)
202 goto out_free;
203 retval = -EPERM;
204 if (!ns_capable(tmp->user_ns, CAP_SYS_ADMIN))
205 goto out_free;
206 set_tid_size--;
207 }
Gargi Sharma95846ec2017-11-17 15:30:30 -0800208
209 idr_preload(GFP_KERNEL);
210 spin_lock_irq(&pidmap_lock);
211
Adrian Reber49cb2fc2019-11-15 13:36:20 +0100212 if (tid) {
213 nr = idr_alloc(&tmp->idr, NULL, tid,
214 tid + 1, GFP_ATOMIC);
215 /*
216 * If ENOSPC is returned it means that the PID is
217 * alreay in use. Return EEXIST in that case.
218 */
219 if (nr == -ENOSPC)
220 nr = -EEXIST;
221 } else {
222 int pid_min = 1;
223 /*
224 * init really needs pid 1, but after reaching the
225 * maximum wrap back to RESERVED_PIDS
226 */
227 if (idr_get_cursor(&tmp->idr) > RESERVED_PIDS)
228 pid_min = RESERVED_PIDS;
Gargi Sharma95846ec2017-11-17 15:30:30 -0800229
Adrian Reber49cb2fc2019-11-15 13:36:20 +0100230 /*
231 * Store a null pointer so find_pid_ns does not find
232 * a partially initialized PID (see below).
233 */
234 nr = idr_alloc_cyclic(&tmp->idr, NULL, pid_min,
235 pid_max, GFP_ATOMIC);
236 }
Gargi Sharma95846ec2017-11-17 15:30:30 -0800237 spin_unlock_irq(&pidmap_lock);
238 idr_preload_end();
239
Arnd Bergmann287980e2016-05-27 23:23:25 +0200240 if (nr < 0) {
KJ Tsanaktsidisf83606f2018-09-20 12:22:25 -0700241 retval = (nr == -ENOSPC) ? -EAGAIN : nr;
Pavel Emelyanov8ef047a2007-10-18 23:40:05 -0700242 goto out_free;
Michal Hocko35f71bc2015-04-16 12:47:38 -0700243 }
Eric W. Biederman92476d72006-03-31 02:31:42 -0800244
Pavel Emelyanov8ef047a2007-10-18 23:40:05 -0700245 pid->numbers[i].nr = nr;
246 pid->numbers[i].ns = tmp;
247 tmp = tmp->parent;
248 }
249
Eric W. Biederman0a01f2c2012-08-01 10:33:47 -0700250 if (unlikely(is_child_reaper(pid))) {
Eric W. Biedermanc0ee5542017-12-22 12:37:43 -0600251 if (pid_ns_prepare_proc(ns))
Eric W. Biederman0a01f2c2012-08-01 10:33:47 -0700252 goto out_free;
253 }
254
Pavel Emelyanovb461cc02007-10-18 23:40:09 -0700255 get_pid_ns(ns);
Joel Fernandes (Google)f57e5152019-07-16 16:30:06 -0700256 refcount_set(&pid->count, 1);
Eric W. Biederman92476d72006-03-31 02:31:42 -0800257 for (type = 0; type < PIDTYPE_MAX; ++type)
258 INIT_HLIST_HEAD(&pid->tasks[type]);
259
Joel Fernandes (Google)b53b0b92019-04-30 12:21:53 -0400260 init_waitqueue_head(&pid->wait_pidfd);
261
André Goddard Rosa417e3152009-12-15 16:47:40 -0800262 upid = pid->numbers + ns->level;
Eric W. Biederman92476d72006-03-31 02:31:42 -0800263 spin_lock_irq(&pidmap_lock);
Gargi Sharmae8cfbc22017-11-17 15:30:34 -0800264 if (!(ns->pid_allocated & PIDNS_ADDING))
Eric W. Biederman5e1182deb2010-07-12 18:50:25 -0700265 goto out_unlock;
Eric W. Biederman0a01f2c2012-08-01 10:33:47 -0700266 for ( ; upid >= pid->numbers; --upid) {
Gargi Sharma95846ec2017-11-17 15:30:30 -0800267 /* Make the PID visible to find_pid_ns. */
268 idr_replace(&upid->ns->idr, pid, upid->nr);
Gargi Sharmae8cfbc22017-11-17 15:30:34 -0800269 upid->ns->pid_allocated++;
Eric W. Biederman0a01f2c2012-08-01 10:33:47 -0700270 }
Eric W. Biederman92476d72006-03-31 02:31:42 -0800271 spin_unlock_irq(&pidmap_lock);
272
Eric W. Biederman92476d72006-03-31 02:31:42 -0800273 return pid;
274
Eric W. Biederman5e1182deb2010-07-12 18:50:25 -0700275out_unlock:
Eric W. Biederman6e666882013-02-12 13:46:23 -0800276 spin_unlock_irq(&pidmap_lock);
Oleg Nesterov24c037e2014-12-10 15:55:25 -0800277 put_pid_ns(ns);
278
Eric W. Biederman92476d72006-03-31 02:31:42 -0800279out_free:
Gargi Sharma95846ec2017-11-17 15:30:30 -0800280 spin_lock_irq(&pidmap_lock);
Matthew Wilcox1a80dad2018-12-28 07:22:26 -0800281 while (++i <= ns->level) {
282 upid = pid->numbers + i;
283 idr_remove(&upid->ns->idr, upid->nr);
284 }
Gargi Sharma95846ec2017-11-17 15:30:30 -0800285
Eric W. Biedermanc0ee5542017-12-22 12:37:43 -0600286 /* On failure to allocate the first pid, reset the state */
287 if (ns->pid_allocated == PIDNS_ADDING)
288 idr_set_cursor(&ns->idr, 0);
289
Gargi Sharma95846ec2017-11-17 15:30:30 -0800290 spin_unlock_irq(&pidmap_lock);
Pavel Emelyanov8ef047a2007-10-18 23:40:05 -0700291
Pavel Emelianovbaf8f0f2007-10-18 23:39:48 -0700292 kmem_cache_free(ns->pid_cachep, pid);
Michal Hocko35f71bc2015-04-16 12:47:38 -0700293 return ERR_PTR(retval);
Eric W. Biederman92476d72006-03-31 02:31:42 -0800294}
295
Eric W. Biedermanc876ad762012-12-21 20:27:12 -0800296void disable_pid_allocation(struct pid_namespace *ns)
297{
298 spin_lock_irq(&pidmap_lock);
Gargi Sharmae8cfbc22017-11-17 15:30:34 -0800299 ns->pid_allocated &= ~PIDNS_ADDING;
Eric W. Biedermanc876ad762012-12-21 20:27:12 -0800300 spin_unlock_irq(&pidmap_lock);
301}
302
Harvey Harrison7ad5b3a2008-02-08 04:19:53 -0800303struct pid *find_pid_ns(int nr, struct pid_namespace *ns)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700304{
Gargi Sharmae8cfbc22017-11-17 15:30:34 -0800305 return idr_find(&ns->idr, nr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700306}
Pavel Emelyanov198fe212007-10-18 23:40:06 -0700307EXPORT_SYMBOL_GPL(find_pid_ns);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700308
Pavel Emelyanov89905712007-10-18 23:40:19 -0700309struct pid *find_vpid(int nr)
310{
Eric W. Biederman17cf22c2010-03-02 14:51:53 -0800311 return find_pid_ns(nr, task_active_pid_ns(current));
Pavel Emelyanov89905712007-10-18 23:40:19 -0700312}
313EXPORT_SYMBOL_GPL(find_vpid);
314
Eric W. Biederman2c470472017-09-26 13:06:43 -0500315static struct pid **task_pid_ptr(struct task_struct *task, enum pid_type type)
316{
317 return (type == PIDTYPE_PID) ?
318 &task->thread_pid :
Eric W. Biederman2c470472017-09-26 13:06:43 -0500319 &task->signal->pids[type];
320}
321
Sukadev Bhattiprolue713d0d2007-05-10 22:22:58 -0700322/*
323 * attach_pid() must be called with the tasklist_lock write-held.
324 */
Oleg Nesterov81907732013-07-03 15:08:31 -0700325void attach_pid(struct task_struct *task, enum pid_type type)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700326{
Eric W. Biederman2c470472017-09-26 13:06:43 -0500327 struct pid *pid = *task_pid_ptr(task, type);
328 hlist_add_head_rcu(&task->pid_links[type], &pid->tasks[type]);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700329}
330
Oleg Nesterov24336eae2008-04-30 00:54:26 -0700331static void __change_pid(struct task_struct *task, enum pid_type type,
332 struct pid *new)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700333{
Eric W. Biederman2c470472017-09-26 13:06:43 -0500334 struct pid **pid_ptr = task_pid_ptr(task, type);
Eric W. Biederman92476d72006-03-31 02:31:42 -0800335 struct pid *pid;
336 int tmp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700337
Eric W. Biederman2c470472017-09-26 13:06:43 -0500338 pid = *pid_ptr;
Eric W. Biederman92476d72006-03-31 02:31:42 -0800339
Eric W. Biederman2c470472017-09-26 13:06:43 -0500340 hlist_del_rcu(&task->pid_links[type]);
341 *pid_ptr = new;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700342
343 for (tmp = PIDTYPE_MAX; --tmp >= 0; )
Christian Brauner1d416a12019-10-17 12:18:30 +0200344 if (pid_has_task(pid, tmp))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700345 return;
346
Eric W. Biederman92476d72006-03-31 02:31:42 -0800347 free_pid(pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700348}
349
Oleg Nesterov24336eae2008-04-30 00:54:26 -0700350void detach_pid(struct task_struct *task, enum pid_type type)
351{
352 __change_pid(task, type, NULL);
353}
354
355void change_pid(struct task_struct *task, enum pid_type type,
356 struct pid *pid)
357{
358 __change_pid(task, type, pid);
Oleg Nesterov81907732013-07-03 15:08:31 -0700359 attach_pid(task, type);
Oleg Nesterov24336eae2008-04-30 00:54:26 -0700360}
361
Eric W. Biedermanc18258c2006-09-27 01:51:06 -0700362/* transfer_pid is an optimization of attach_pid(new), detach_pid(old) */
Harvey Harrison7ad5b3a2008-02-08 04:19:53 -0800363void transfer_pid(struct task_struct *old, struct task_struct *new,
Eric W. Biedermanc18258c2006-09-27 01:51:06 -0700364 enum pid_type type)
365{
Eric W. Biederman2c470472017-09-26 13:06:43 -0500366 if (type == PIDTYPE_PID)
367 new->thread_pid = old->thread_pid;
368 hlist_replace_rcu(&old->pid_links[type], &new->pid_links[type]);
Eric W. Biedermanc18258c2006-09-27 01:51:06 -0700369}
370
Harvey Harrison7ad5b3a2008-02-08 04:19:53 -0800371struct task_struct *pid_task(struct pid *pid, enum pid_type type)
Eric W. Biederman92476d72006-03-31 02:31:42 -0800372{
373 struct task_struct *result = NULL;
374 if (pid) {
375 struct hlist_node *first;
Arnd Bergmann67bdbff2010-02-25 16:55:13 +0100376 first = rcu_dereference_check(hlist_first_rcu(&pid->tasks[type]),
Paul E. McKenneydb1466b2010-03-03 07:46:56 -0800377 lockdep_tasklist_lock_is_held());
Eric W. Biederman92476d72006-03-31 02:31:42 -0800378 if (first)
Eric W. Biederman2c470472017-09-26 13:06:43 -0500379 result = hlist_entry(first, struct task_struct, pid_links[(type)]);
Eric W. Biederman92476d72006-03-31 02:31:42 -0800380 }
381 return result;
382}
Pavel Emelyanoveccba062008-02-07 00:13:21 -0800383EXPORT_SYMBOL(pid_task);
Eric W. Biederman92476d72006-03-31 02:31:42 -0800384
385/*
Tetsuo Handa9728e5d2010-03-05 13:42:56 -0800386 * Must be called under rcu_read_lock().
Eric W. Biederman92476d72006-03-31 02:31:42 -0800387 */
Christoph Hellwig17f98dc2009-06-17 16:27:51 -0700388struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700389{
Paul E. McKenneyf78f5b92015-06-18 15:50:02 -0700390 RCU_LOCKDEP_WARN(!rcu_read_lock_held(),
391 "find_task_by_pid_ns() needs rcu_read_lock() protection");
Christoph Hellwig17f98dc2009-06-17 16:27:51 -0700392 return pid_task(find_pid_ns(nr, ns), PIDTYPE_PID);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700393}
394
Pavel Emelyanov228ebcb2007-10-18 23:40:16 -0700395struct task_struct *find_task_by_vpid(pid_t vnr)
396{
Eric W. Biederman17cf22c2010-03-02 14:51:53 -0800397 return find_task_by_pid_ns(vnr, task_active_pid_ns(current));
Pavel Emelyanov228ebcb2007-10-18 23:40:16 -0700398}
Pavel Emelyanov228ebcb2007-10-18 23:40:16 -0700399
Mike Rapoport2ee08262018-02-06 15:40:17 -0800400struct task_struct *find_get_task_by_vpid(pid_t nr)
401{
402 struct task_struct *task;
403
404 rcu_read_lock();
405 task = find_task_by_vpid(nr);
406 if (task)
407 get_task_struct(task);
408 rcu_read_unlock();
409
410 return task;
411}
412
Oleg Nesterov1a657f782006-10-02 02:18:59 -0700413struct pid *get_task_pid(struct task_struct *task, enum pid_type type)
414{
415 struct pid *pid;
416 rcu_read_lock();
Eric W. Biederman2c470472017-09-26 13:06:43 -0500417 pid = get_pid(rcu_dereference(*task_pid_ptr(task, type)));
Oleg Nesterov1a657f782006-10-02 02:18:59 -0700418 rcu_read_unlock();
419 return pid;
420}
Rik van Riel77c100c2011-02-01 09:51:46 -0500421EXPORT_SYMBOL_GPL(get_task_pid);
Oleg Nesterov1a657f782006-10-02 02:18:59 -0700422
Harvey Harrison7ad5b3a2008-02-08 04:19:53 -0800423struct task_struct *get_pid_task(struct pid *pid, enum pid_type type)
Eric W. Biederman92476d72006-03-31 02:31:42 -0800424{
425 struct task_struct *result;
426 rcu_read_lock();
427 result = pid_task(pid, type);
428 if (result)
429 get_task_struct(result);
430 rcu_read_unlock();
431 return result;
432}
Rik van Riel77c100c2011-02-01 09:51:46 -0500433EXPORT_SYMBOL_GPL(get_pid_task);
Eric W. Biederman92476d72006-03-31 02:31:42 -0800434
435struct pid *find_get_pid(pid_t nr)
436{
437 struct pid *pid;
438
439 rcu_read_lock();
Pavel Emelyanov198fe212007-10-18 23:40:06 -0700440 pid = get_pid(find_vpid(nr));
Eric W. Biederman92476d72006-03-31 02:31:42 -0800441 rcu_read_unlock();
442
443 return pid;
444}
David Sterba339caf22008-07-25 01:48:31 -0700445EXPORT_SYMBOL_GPL(find_get_pid);
Eric W. Biederman92476d72006-03-31 02:31:42 -0800446
Pavel Emelyanov7af57292007-10-18 23:40:06 -0700447pid_t pid_nr_ns(struct pid *pid, struct pid_namespace *ns)
448{
449 struct upid *upid;
450 pid_t nr = 0;
451
452 if (pid && ns->level <= pid->level) {
453 upid = &pid->numbers[ns->level];
454 if (upid->ns == ns)
455 nr = upid->nr;
456 }
457 return nr;
458}
Eric W. Biederman4f82f452012-05-24 10:37:59 -0600459EXPORT_SYMBOL_GPL(pid_nr_ns);
Pavel Emelyanov7af57292007-10-18 23:40:06 -0700460
Eric W. Biederman44c4e1b2008-02-08 04:19:15 -0800461pid_t pid_vnr(struct pid *pid)
462{
Eric W. Biederman17cf22c2010-03-02 14:51:53 -0800463 return pid_nr_ns(pid, task_active_pid_ns(current));
Eric W. Biederman44c4e1b2008-02-08 04:19:15 -0800464}
465EXPORT_SYMBOL_GPL(pid_vnr);
466
Oleg Nesterov52ee2df2009-04-02 16:58:38 -0700467pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type,
468 struct pid_namespace *ns)
Pavel Emelyanov2f2a3a42007-10-18 23:40:19 -0700469{
Oleg Nesterov52ee2df2009-04-02 16:58:38 -0700470 pid_t nr = 0;
471
472 rcu_read_lock();
473 if (!ns)
Eric W. Biederman17cf22c2010-03-02 14:51:53 -0800474 ns = task_active_pid_ns(current);
Eric W. Biederman2c470472017-09-26 13:06:43 -0500475 if (likely(pid_alive(task)))
476 nr = pid_nr_ns(rcu_dereference(*task_pid_ptr(task, type)), ns);
Oleg Nesterov52ee2df2009-04-02 16:58:38 -0700477 rcu_read_unlock();
478
479 return nr;
Pavel Emelyanov2f2a3a42007-10-18 23:40:19 -0700480}
Oleg Nesterov52ee2df2009-04-02 16:58:38 -0700481EXPORT_SYMBOL(__task_pid_nr_ns);
Pavel Emelyanov2f2a3a42007-10-18 23:40:19 -0700482
Eric W. Biederman61bce0f2009-01-07 18:08:49 -0800483struct pid_namespace *task_active_pid_ns(struct task_struct *tsk)
484{
485 return ns_of_pid(task_pid(tsk));
486}
487EXPORT_SYMBOL_GPL(task_active_pid_ns);
488
Linus Torvalds1da177e2005-04-16 15:20:36 -0700489/*
Frederik Schwarzer025dfda2008-10-16 19:02:37 +0200490 * Used by proc to find the first pid that is greater than or equal to nr.
Eric W. Biederman0804ef42006-10-02 02:17:04 -0700491 *
Pavel Emelyanove49859e2008-07-25 01:48:36 -0700492 * If there is a pid at nr this function is exactly the same as find_pid_ns.
Eric W. Biederman0804ef42006-10-02 02:17:04 -0700493 */
Pavel Emelyanov198fe212007-10-18 23:40:06 -0700494struct pid *find_ge_pid(int nr, struct pid_namespace *ns)
Eric W. Biederman0804ef42006-10-02 02:17:04 -0700495{
Gargi Sharma95846ec2017-11-17 15:30:30 -0800496 return idr_get_next(&ns->idr, &nr);
Eric W. Biederman0804ef42006-10-02 02:17:04 -0700497}
498
Christian Brauner32fcb422019-05-24 12:43:51 +0200499/**
500 * pidfd_create() - Create a new pid file descriptor.
501 *
502 * @pid: struct pid that the pidfd will reference
503 *
504 * This creates a new pid file descriptor with the O_CLOEXEC flag set.
505 *
506 * Note, that this function can only be called after the fd table has
507 * been unshared to avoid leaking the pidfd to the new process.
508 *
509 * Return: On success, a cloexec pidfd is returned.
510 * On error, a negative errno number will be returned.
511 */
512static int pidfd_create(struct pid *pid)
513{
514 int fd;
515
516 fd = anon_inode_getfd("[pidfd]", &pidfd_fops, get_pid(pid),
517 O_RDWR | O_CLOEXEC);
518 if (fd < 0)
519 put_pid(pid);
520
521 return fd;
522}
523
524/**
525 * pidfd_open() - Open new pid file descriptor.
526 *
527 * @pid: pid for which to retrieve a pidfd
528 * @flags: flags to pass
529 *
530 * This creates a new pid file descriptor with the O_CLOEXEC flag set for
531 * the process identified by @pid. Currently, the process identified by
532 * @pid must be a thread-group leader. This restriction currently exists
533 * for all aspects of pidfds including pidfd creation (CLONE_PIDFD cannot
534 * be used with CLONE_THREAD) and pidfd polling (only supports thread group
535 * leaders).
536 *
537 * Return: On success, a cloexec pidfd is returned.
538 * On error, a negative errno number will be returned.
539 */
540SYSCALL_DEFINE2(pidfd_open, pid_t, pid, unsigned int, flags)
541{
Christian Brauner1e1d0f02019-10-17 12:18:32 +0200542 int fd;
Christian Brauner32fcb422019-05-24 12:43:51 +0200543 struct pid *p;
544
545 if (flags)
546 return -EINVAL;
547
548 if (pid <= 0)
549 return -EINVAL;
550
551 p = find_get_pid(pid);
552 if (!p)
553 return -ESRCH;
554
Christian Brauner1e1d0f02019-10-17 12:18:32 +0200555 if (pid_has_task(p, PIDTYPE_TGID))
556 fd = pidfd_create(p);
557 else
558 fd = -EINVAL;
Christian Brauner32fcb422019-05-24 12:43:51 +0200559
Christian Brauner32fcb422019-05-24 12:43:51 +0200560 put_pid(p);
561 return fd;
562}
563
Gargi Sharma95846ec2017-11-17 15:30:30 -0800564void __init pid_idr_init(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700565{
Zhen Lei840d6fe2016-01-30 10:04:17 +0800566 /* Verify no one has done anything silly: */
Gargi Sharmae8cfbc22017-11-17 15:30:34 -0800567 BUILD_BUG_ON(PID_MAX_LIMIT >= PIDNS_ADDING);
Eric W. Biedermanc876ad762012-12-21 20:27:12 -0800568
Hedi Berriche72680a12010-05-26 14:44:06 -0700569 /* bump default and minimum pid_max based on number of cpus */
570 pid_max = min(pid_max_max, max_t(int, pid_max,
571 PIDS_PER_CPU_DEFAULT * num_possible_cpus()));
572 pid_max_min = max_t(int, pid_max_min,
573 PIDS_PER_CPU_MIN * num_possible_cpus());
574 pr_info("pid_max: default: %u minimum: %u\n", pid_max, pid_max_min);
575
Gargi Sharma95846ec2017-11-17 15:30:30 -0800576 idr_init(&init_pid_ns.idr);
Eric W. Biederman92476d72006-03-31 02:31:42 -0800577
Pavel Emelyanov74bd59b2008-02-08 04:18:24 -0800578 init_pid_ns.pid_cachep = KMEM_CACHE(pid,
Vladimir Davydov5d097052016-01-14 15:18:21 -0800579 SLAB_HWCACHE_ALIGN | SLAB_PANIC | SLAB_ACCOUNT);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700580}