blob: 5fa1db48d8b70c686462fc7315b833efe503d101 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * Generic pidhash and scalable, time-bounded PID allocator
3 *
4 * (C) 2002-2003 William Irwin, IBM
5 * (C) 2004 William Irwin, Oracle
6 * (C) 2002-2004 Ingo Molnar, Red Hat
7 *
8 * pid-structures are backing objects for tasks sharing a given ID to chain
9 * against. There is very little to them aside from hashing them and
10 * parking tasks using given ID's on a list.
11 *
12 * The hash is always changed with the tasklist_lock write-acquired,
13 * and the hash is only accessed with the tasklist_lock at least
14 * read-acquired, so there's no additional SMP locking needed here.
15 *
16 * We have a list of bitmap pages, which bitmaps represent the PID space.
17 * Allocating and freeing PIDs is completely lockless. The worst-case
18 * allocation scenario when all but one out of 1 million PIDs possible are
19 * allocated already: the scanning of 32 list entries and at most PAGE_SIZE
20 * bytes. The typical fastpath is a single successful setbit. Freeing is O(1).
Pavel Emelyanov30e49c22007-10-18 23:40:10 -070021 *
22 * Pid namespaces:
23 * (C) 2007 Pavel Emelyanov <xemul@openvz.org>, OpenVZ, SWsoft Inc.
24 * (C) 2007 Sukadev Bhattiprolu <sukadev@us.ibm.com>, IBM
25 * Many thanks to Oleg Nesterov for comments and help
26 *
Linus Torvalds1da177e2005-04-16 15:20:36 -070027 */
28
29#include <linux/mm.h>
30#include <linux/module.h>
31#include <linux/slab.h>
32#include <linux/init.h>
Franck Bui-Huu82524742008-05-12 21:21:05 +020033#include <linux/rculist.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070034#include <linux/bootmem.h>
35#include <linux/hash.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>
Catalin Marinas12de38b2009-06-29 17:13:55 +010039#include <linux/kmemleak.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070040
Pavel Emelyanov8ef047a2007-10-18 23:40:05 -070041#define pid_hashfn(nr, ns) \
42 hash_long((unsigned long)nr + (unsigned long)ns, pidhash_shift)
Eric W. Biederman92476d72006-03-31 02:31:42 -080043static struct hlist_head *pid_hash;
Linus Torvalds1da177e2005-04-16 15:20:36 -070044static int pidhash_shift;
Sukadev Bhattiprolu820e45d2007-05-10 22:23:00 -070045struct pid init_struct_pid = INIT_STRUCT_PID;
Linus Torvalds1da177e2005-04-16 15:20:36 -070046
47int pid_max = PID_MAX_DEFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -070048
49#define RESERVED_PIDS 300
50
51int pid_max_min = RESERVED_PIDS + 1;
52int pid_max_max = PID_MAX_LIMIT;
53
Linus Torvalds1da177e2005-04-16 15:20:36 -070054#define BITS_PER_PAGE (PAGE_SIZE*8)
55#define BITS_PER_PAGE_MASK (BITS_PER_PAGE-1)
Sukadev Bhattiprolu3fbc9642006-10-02 02:17:24 -070056
Sukadev Bhattiprolu61a58c62006-12-08 02:37:58 -080057static inline int mk_pid(struct pid_namespace *pid_ns,
58 struct pidmap *map, int off)
Sukadev Bhattiprolu3fbc9642006-10-02 02:17:24 -070059{
Sukadev Bhattiprolu61a58c62006-12-08 02:37:58 -080060 return (map - pid_ns->pidmap)*BITS_PER_PAGE + off;
Sukadev Bhattiprolu3fbc9642006-10-02 02:17:24 -070061}
62
Linus Torvalds1da177e2005-04-16 15:20:36 -070063#define find_next_offset(map, off) \
64 find_next_zero_bit((map)->page, BITS_PER_PAGE, off)
65
66/*
67 * PID-map pages start out as NULL, they get allocated upon
68 * first use and are never deallocated. This way a low pid_max
69 * value does not cause lots of bitmaps to be allocated, but
70 * the scheme scales to up to 4 million PIDs, runtime.
71 */
Sukadev Bhattiprolu61a58c62006-12-08 02:37:58 -080072struct pid_namespace init_pid_ns = {
Cedric Le Goater9a575a92006-12-08 02:37:59 -080073 .kref = {
74 .refcount = ATOMIC_INIT(2),
75 },
Sukadev Bhattiprolu3fbc9642006-10-02 02:17:24 -070076 .pidmap = {
77 [ 0 ... PIDMAP_ENTRIES-1] = { ATOMIC_INIT(BITS_PER_PAGE), NULL }
78 },
Sukadev Bhattiprolu84d73782006-12-08 02:38:01 -080079 .last_pid = 0,
Pavel Emelyanovfaacbfd2007-10-18 23:40:04 -070080 .level = 0,
81 .child_reaper = &init_task,
Sukadev Bhattiprolu3fbc9642006-10-02 02:17:24 -070082};
Pavel Emelyanov198fe212007-10-18 23:40:06 -070083EXPORT_SYMBOL_GPL(init_pid_ns);
Linus Torvalds1da177e2005-04-16 15:20:36 -070084
Pavel Emelyanovb461cc02007-10-18 23:40:09 -070085int is_container_init(struct task_struct *tsk)
Serge E. Hallynb460cbc2007-10-18 23:39:52 -070086{
Pavel Emelyanovb461cc02007-10-18 23:40:09 -070087 int ret = 0;
88 struct pid *pid;
89
90 rcu_read_lock();
91 pid = task_pid(tsk);
92 if (pid != NULL && pid->numbers[pid->level].nr == 1)
93 ret = 1;
94 rcu_read_unlock();
95
96 return ret;
Serge E. Hallynb460cbc2007-10-18 23:39:52 -070097}
Pavel Emelyanovb461cc02007-10-18 23:40:09 -070098EXPORT_SYMBOL(is_container_init);
Serge E. Hallynb460cbc2007-10-18 23:39:52 -070099
Eric W. Biederman92476d72006-03-31 02:31:42 -0800100/*
101 * Note: disable interrupts while the pidmap_lock is held as an
102 * interrupt might come in and do read_lock(&tasklist_lock).
103 *
104 * If we don't disable interrupts there is a nasty deadlock between
105 * detach_pid()->free_pid() and another cpu that does
106 * spin_lock(&pidmap_lock) followed by an interrupt routine that does
107 * read_lock(&tasklist_lock);
108 *
109 * After we clean up the tasklist_lock and know there are no
110 * irq handlers that take it we can leave the interrupts enabled.
111 * For now it is easier to be safe than to prove it can't happen.
112 */
Sukadev Bhattiprolu3fbc9642006-10-02 02:17:24 -0700113
Linus Torvalds1da177e2005-04-16 15:20:36 -0700114static __cacheline_aligned_in_smp DEFINE_SPINLOCK(pidmap_lock);
115
Oleg Nesterovb7127aa2008-04-30 00:54:22 -0700116static void free_pidmap(struct upid *upid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700117{
Oleg Nesterovb7127aa2008-04-30 00:54:22 -0700118 int nr = upid->nr;
119 struct pidmap *map = upid->ns->pidmap + nr / BITS_PER_PAGE;
120 int offset = nr & BITS_PER_PAGE_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700121
122 clear_bit(offset, map->page);
123 atomic_inc(&map->nr_free);
124}
125
Sukadev Bhattiprolu61a58c62006-12-08 02:37:58 -0800126static int alloc_pidmap(struct pid_namespace *pid_ns)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700127{
Sukadev Bhattiprolu61a58c62006-12-08 02:37:58 -0800128 int i, offset, max_scan, pid, last = pid_ns->last_pid;
Sukadev Bhattiprolu6a1f3b82006-10-02 02:17:20 -0700129 struct pidmap *map;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700130
131 pid = last + 1;
132 if (pid >= pid_max)
133 pid = RESERVED_PIDS;
134 offset = pid & BITS_PER_PAGE_MASK;
Sukadev Bhattiprolu61a58c62006-12-08 02:37:58 -0800135 map = &pid_ns->pidmap[pid/BITS_PER_PAGE];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700136 max_scan = (pid_max + BITS_PER_PAGE - 1)/BITS_PER_PAGE - !offset;
137 for (i = 0; i <= max_scan; ++i) {
138 if (unlikely(!map->page)) {
Sukadev Bhattiprolu3fbc9642006-10-02 02:17:24 -0700139 void *page = kzalloc(PAGE_SIZE, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700140 /*
141 * Free the page if someone raced with us
142 * installing it:
143 */
Eric W. Biederman92476d72006-03-31 02:31:42 -0800144 spin_lock_irq(&pidmap_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700145 if (map->page)
Sukadev Bhattiprolu3fbc9642006-10-02 02:17:24 -0700146 kfree(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700147 else
Sukadev Bhattiprolu3fbc9642006-10-02 02:17:24 -0700148 map->page = page;
Eric W. Biederman92476d72006-03-31 02:31:42 -0800149 spin_unlock_irq(&pidmap_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700150 if (unlikely(!map->page))
151 break;
152 }
153 if (likely(atomic_read(&map->nr_free))) {
154 do {
155 if (!test_and_set_bit(offset, map->page)) {
156 atomic_dec(&map->nr_free);
Sukadev Bhattiprolu61a58c62006-12-08 02:37:58 -0800157 pid_ns->last_pid = pid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700158 return pid;
159 }
160 offset = find_next_offset(map, offset);
Sukadev Bhattiprolu61a58c62006-12-08 02:37:58 -0800161 pid = mk_pid(pid_ns, map, offset);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700162 /*
163 * find_next_offset() found a bit, the pid from it
164 * is in-bounds, and if we fell back to the last
165 * bitmap block and the final block was the same
166 * as the starting point, pid is before last_pid.
167 */
168 } while (offset < BITS_PER_PAGE && pid < pid_max &&
169 (i != max_scan || pid < last ||
170 !((last+1) & BITS_PER_PAGE_MASK)));
171 }
Sukadev Bhattiprolu61a58c62006-12-08 02:37:58 -0800172 if (map < &pid_ns->pidmap[(pid_max-1)/BITS_PER_PAGE]) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700173 ++map;
174 offset = 0;
175 } else {
Sukadev Bhattiprolu61a58c62006-12-08 02:37:58 -0800176 map = &pid_ns->pidmap[0];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700177 offset = RESERVED_PIDS;
178 if (unlikely(last == offset))
179 break;
180 }
Sukadev Bhattiprolu61a58c62006-12-08 02:37:58 -0800181 pid = mk_pid(pid_ns, map, offset);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700182 }
183 return -1;
184}
185
Pavel Emelyanov74bd59b2008-02-08 04:18:24 -0800186int next_pidmap(struct pid_namespace *pid_ns, int last)
Eric W. Biederman0804ef42006-10-02 02:17:04 -0700187{
188 int offset;
Eric W. Biedermanf40f50d2006-10-02 02:17:25 -0700189 struct pidmap *map, *end;
Eric W. Biederman0804ef42006-10-02 02:17:04 -0700190
191 offset = (last + 1) & BITS_PER_PAGE_MASK;
Sukadev Bhattiprolu61a58c62006-12-08 02:37:58 -0800192 map = &pid_ns->pidmap[(last + 1)/BITS_PER_PAGE];
193 end = &pid_ns->pidmap[PIDMAP_ENTRIES];
Eric W. Biedermanf40f50d2006-10-02 02:17:25 -0700194 for (; map < end; map++, offset = 0) {
Eric W. Biederman0804ef42006-10-02 02:17:04 -0700195 if (unlikely(!map->page))
196 continue;
197 offset = find_next_bit((map)->page, BITS_PER_PAGE, offset);
198 if (offset < BITS_PER_PAGE)
Sukadev Bhattiprolu61a58c62006-12-08 02:37:58 -0800199 return mk_pid(pid_ns, map, offset);
Eric W. Biederman0804ef42006-10-02 02:17:04 -0700200 }
201 return -1;
202}
203
Harvey Harrison7ad5b3a2008-02-08 04:19:53 -0800204void put_pid(struct pid *pid)
Eric W. Biederman92476d72006-03-31 02:31:42 -0800205{
Pavel Emelianovbaf8f0f2007-10-18 23:39:48 -0700206 struct pid_namespace *ns;
207
Eric W. Biederman92476d72006-03-31 02:31:42 -0800208 if (!pid)
209 return;
Pavel Emelianovbaf8f0f2007-10-18 23:39:48 -0700210
Pavel Emelyanov8ef047a2007-10-18 23:40:05 -0700211 ns = pid->numbers[pid->level].ns;
Eric W. Biederman92476d72006-03-31 02:31:42 -0800212 if ((atomic_read(&pid->count) == 1) ||
Pavel Emelyanov8ef047a2007-10-18 23:40:05 -0700213 atomic_dec_and_test(&pid->count)) {
Pavel Emelianovbaf8f0f2007-10-18 23:39:48 -0700214 kmem_cache_free(ns->pid_cachep, pid);
Pavel Emelyanovb461cc02007-10-18 23:40:09 -0700215 put_pid_ns(ns);
Pavel Emelyanov8ef047a2007-10-18 23:40:05 -0700216 }
Eric W. Biederman92476d72006-03-31 02:31:42 -0800217}
Eric W. Biedermanbbf73142006-10-02 02:17:11 -0700218EXPORT_SYMBOL_GPL(put_pid);
Eric W. Biederman92476d72006-03-31 02:31:42 -0800219
220static void delayed_put_pid(struct rcu_head *rhp)
221{
222 struct pid *pid = container_of(rhp, struct pid, rcu);
223 put_pid(pid);
224}
225
Harvey Harrison7ad5b3a2008-02-08 04:19:53 -0800226void free_pid(struct pid *pid)
Eric W. Biederman92476d72006-03-31 02:31:42 -0800227{
228 /* We can be called with write_lock_irq(&tasklist_lock) held */
Pavel Emelyanov8ef047a2007-10-18 23:40:05 -0700229 int i;
Eric W. Biederman92476d72006-03-31 02:31:42 -0800230 unsigned long flags;
231
232 spin_lock_irqsave(&pidmap_lock, flags);
Pavel Emelyanov198fe212007-10-18 23:40:06 -0700233 for (i = 0; i <= pid->level; i++)
234 hlist_del_rcu(&pid->numbers[i].pid_chain);
Eric W. Biederman92476d72006-03-31 02:31:42 -0800235 spin_unlock_irqrestore(&pidmap_lock, flags);
236
Pavel Emelyanov8ef047a2007-10-18 23:40:05 -0700237 for (i = 0; i <= pid->level; i++)
Oleg Nesterovb7127aa2008-04-30 00:54:22 -0700238 free_pidmap(pid->numbers + i);
Pavel Emelyanov8ef047a2007-10-18 23:40:05 -0700239
Eric W. Biederman92476d72006-03-31 02:31:42 -0800240 call_rcu(&pid->rcu, delayed_put_pid);
241}
242
Pavel Emelyanov8ef047a2007-10-18 23:40:05 -0700243struct pid *alloc_pid(struct pid_namespace *ns)
Eric W. Biederman92476d72006-03-31 02:31:42 -0800244{
245 struct pid *pid;
246 enum pid_type type;
Pavel Emelyanov8ef047a2007-10-18 23:40:05 -0700247 int i, nr;
248 struct pid_namespace *tmp;
Pavel Emelyanov198fe212007-10-18 23:40:06 -0700249 struct upid *upid;
Eric W. Biederman92476d72006-03-31 02:31:42 -0800250
Pavel Emelianovbaf8f0f2007-10-18 23:39:48 -0700251 pid = kmem_cache_alloc(ns->pid_cachep, GFP_KERNEL);
Eric W. Biederman92476d72006-03-31 02:31:42 -0800252 if (!pid)
253 goto out;
254
Pavel Emelyanov8ef047a2007-10-18 23:40:05 -0700255 tmp = ns;
256 for (i = ns->level; i >= 0; i--) {
257 nr = alloc_pidmap(tmp);
258 if (nr < 0)
259 goto out_free;
Eric W. Biederman92476d72006-03-31 02:31:42 -0800260
Pavel Emelyanov8ef047a2007-10-18 23:40:05 -0700261 pid->numbers[i].nr = nr;
262 pid->numbers[i].ns = tmp;
263 tmp = tmp->parent;
264 }
265
Pavel Emelyanovb461cc02007-10-18 23:40:09 -0700266 get_pid_ns(ns);
Pavel Emelyanov8ef047a2007-10-18 23:40:05 -0700267 pid->level = ns->level;
Eric W. Biederman92476d72006-03-31 02:31:42 -0800268 atomic_set(&pid->count, 1);
Eric W. Biederman92476d72006-03-31 02:31:42 -0800269 for (type = 0; type < PIDTYPE_MAX; ++type)
270 INIT_HLIST_HEAD(&pid->tasks[type]);
271
272 spin_lock_irq(&pidmap_lock);
Pavel Emelyanov198fe212007-10-18 23:40:06 -0700273 for (i = ns->level; i >= 0; i--) {
274 upid = &pid->numbers[i];
275 hlist_add_head_rcu(&upid->pid_chain,
276 &pid_hash[pid_hashfn(upid->nr, upid->ns)]);
277 }
Eric W. Biederman92476d72006-03-31 02:31:42 -0800278 spin_unlock_irq(&pidmap_lock);
279
280out:
281 return pid;
282
283out_free:
Oleg Nesterovb7127aa2008-04-30 00:54:22 -0700284 while (++i <= ns->level)
285 free_pidmap(pid->numbers + i);
Pavel Emelyanov8ef047a2007-10-18 23:40:05 -0700286
Pavel Emelianovbaf8f0f2007-10-18 23:39:48 -0700287 kmem_cache_free(ns->pid_cachep, pid);
Eric W. Biederman92476d72006-03-31 02:31:42 -0800288 pid = NULL;
289 goto out;
290}
291
Harvey Harrison7ad5b3a2008-02-08 04:19:53 -0800292struct pid *find_pid_ns(int nr, struct pid_namespace *ns)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700293{
294 struct hlist_node *elem;
Pavel Emelyanov198fe212007-10-18 23:40:06 -0700295 struct upid *pnr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700296
Pavel Emelyanov198fe212007-10-18 23:40:06 -0700297 hlist_for_each_entry_rcu(pnr, elem,
298 &pid_hash[pid_hashfn(nr, ns)], pid_chain)
299 if (pnr->nr == nr && pnr->ns == ns)
300 return container_of(pnr, struct pid,
301 numbers[ns->level]);
302
Linus Torvalds1da177e2005-04-16 15:20:36 -0700303 return NULL;
304}
Pavel Emelyanov198fe212007-10-18 23:40:06 -0700305EXPORT_SYMBOL_GPL(find_pid_ns);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700306
Pavel Emelyanov89905712007-10-18 23:40:19 -0700307struct pid *find_vpid(int nr)
308{
309 return find_pid_ns(nr, current->nsproxy->pid_ns);
310}
311EXPORT_SYMBOL_GPL(find_vpid);
312
Sukadev Bhattiprolue713d0d2007-05-10 22:22:58 -0700313/*
314 * attach_pid() must be called with the tasklist_lock write-held.
315 */
Oleg Nesterov24336ea2008-04-30 00:54:26 -0700316void attach_pid(struct task_struct *task, enum pid_type type,
Sukadev Bhattiprolue713d0d2007-05-10 22:22:58 -0700317 struct pid *pid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700318{
Eric W. Biederman92476d72006-03-31 02:31:42 -0800319 struct pid_link *link;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700320
Eric W. Biederman92476d72006-03-31 02:31:42 -0800321 link = &task->pids[type];
Sukadev Bhattiprolue713d0d2007-05-10 22:22:58 -0700322 link->pid = pid;
Eric W. Biederman92476d72006-03-31 02:31:42 -0800323 hlist_add_head_rcu(&link->node, &pid->tasks[type]);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700324}
325
Oleg Nesterov24336ea2008-04-30 00:54:26 -0700326static void __change_pid(struct task_struct *task, enum pid_type type,
327 struct pid *new)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700328{
Eric W. Biederman92476d72006-03-31 02:31:42 -0800329 struct pid_link *link;
330 struct pid *pid;
331 int tmp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700332
Eric W. Biederman92476d72006-03-31 02:31:42 -0800333 link = &task->pids[type];
334 pid = link->pid;
335
336 hlist_del_rcu(&link->node);
Oleg Nesterov24336ea2008-04-30 00:54:26 -0700337 link->pid = new;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700338
339 for (tmp = PIDTYPE_MAX; --tmp >= 0; )
Eric W. Biederman92476d72006-03-31 02:31:42 -0800340 if (!hlist_empty(&pid->tasks[tmp]))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700341 return;
342
Eric W. Biederman92476d72006-03-31 02:31:42 -0800343 free_pid(pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700344}
345
Oleg Nesterov24336ea2008-04-30 00:54:26 -0700346void detach_pid(struct task_struct *task, enum pid_type type)
347{
348 __change_pid(task, type, NULL);
349}
350
351void change_pid(struct task_struct *task, enum pid_type type,
352 struct pid *pid)
353{
354 __change_pid(task, type, pid);
355 attach_pid(task, type, pid);
356}
357
Eric W. Biedermanc18258c2006-09-27 01:51:06 -0700358/* transfer_pid is an optimization of attach_pid(new), detach_pid(old) */
Harvey Harrison7ad5b3a2008-02-08 04:19:53 -0800359void transfer_pid(struct task_struct *old, struct task_struct *new,
Eric W. Biedermanc18258c2006-09-27 01:51:06 -0700360 enum pid_type type)
361{
362 new->pids[type].pid = old->pids[type].pid;
363 hlist_replace_rcu(&old->pids[type].node, &new->pids[type].node);
Eric W. Biedermanc18258c2006-09-27 01:51:06 -0700364}
365
Harvey Harrison7ad5b3a2008-02-08 04:19:53 -0800366struct task_struct *pid_task(struct pid *pid, enum pid_type type)
Eric W. Biederman92476d72006-03-31 02:31:42 -0800367{
368 struct task_struct *result = NULL;
369 if (pid) {
370 struct hlist_node *first;
371 first = rcu_dereference(pid->tasks[type].first);
372 if (first)
373 result = hlist_entry(first, struct task_struct, pids[(type)].node);
374 }
375 return result;
376}
Pavel Emelyanoveccba062008-02-07 00:13:21 -0800377EXPORT_SYMBOL(pid_task);
Eric W. Biederman92476d72006-03-31 02:31:42 -0800378
379/*
380 * Must be called under rcu_read_lock() or with tasklist_lock read-held.
381 */
Christoph Hellwig17f98dc2009-06-17 16:27:51 -0700382struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700383{
Christoph Hellwig17f98dc2009-06-17 16:27:51 -0700384 return pid_task(find_pid_ns(nr, ns), PIDTYPE_PID);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700385}
386
Pavel Emelyanov228ebcb2007-10-18 23:40:16 -0700387struct task_struct *find_task_by_vpid(pid_t vnr)
388{
Christoph Hellwig17f98dc2009-06-17 16:27:51 -0700389 return find_task_by_pid_ns(vnr, current->nsproxy->pid_ns);
Pavel Emelyanov228ebcb2007-10-18 23:40:16 -0700390}
Pavel Emelyanov228ebcb2007-10-18 23:40:16 -0700391
Oleg Nesterov1a657f782006-10-02 02:18:59 -0700392struct pid *get_task_pid(struct task_struct *task, enum pid_type type)
393{
394 struct pid *pid;
395 rcu_read_lock();
Oleg Nesterov2ae448e2009-04-02 16:58:36 -0700396 if (type != PIDTYPE_PID)
397 task = task->group_leader;
Oleg Nesterov1a657f782006-10-02 02:18:59 -0700398 pid = get_pid(task->pids[type].pid);
399 rcu_read_unlock();
400 return pid;
401}
402
Harvey Harrison7ad5b3a2008-02-08 04:19:53 -0800403struct task_struct *get_pid_task(struct pid *pid, enum pid_type type)
Eric W. Biederman92476d72006-03-31 02:31:42 -0800404{
405 struct task_struct *result;
406 rcu_read_lock();
407 result = pid_task(pid, type);
408 if (result)
409 get_task_struct(result);
410 rcu_read_unlock();
411 return result;
412}
413
414struct pid *find_get_pid(pid_t nr)
415{
416 struct pid *pid;
417
418 rcu_read_lock();
Pavel Emelyanov198fe212007-10-18 23:40:06 -0700419 pid = get_pid(find_vpid(nr));
Eric W. Biederman92476d72006-03-31 02:31:42 -0800420 rcu_read_unlock();
421
422 return pid;
423}
David Sterba339caf22008-07-25 01:48:31 -0700424EXPORT_SYMBOL_GPL(find_get_pid);
Eric W. Biederman92476d72006-03-31 02:31:42 -0800425
Pavel Emelyanov7af57292007-10-18 23:40:06 -0700426pid_t pid_nr_ns(struct pid *pid, struct pid_namespace *ns)
427{
428 struct upid *upid;
429 pid_t nr = 0;
430
431 if (pid && ns->level <= pid->level) {
432 upid = &pid->numbers[ns->level];
433 if (upid->ns == ns)
434 nr = upid->nr;
435 }
436 return nr;
437}
438
Eric W. Biederman44c4e1b2008-02-08 04:19:15 -0800439pid_t pid_vnr(struct pid *pid)
440{
441 return pid_nr_ns(pid, current->nsproxy->pid_ns);
442}
443EXPORT_SYMBOL_GPL(pid_vnr);
444
Oleg Nesterov52ee2df2009-04-02 16:58:38 -0700445pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type,
446 struct pid_namespace *ns)
Pavel Emelyanov2f2a3a42007-10-18 23:40:19 -0700447{
Oleg Nesterov52ee2df2009-04-02 16:58:38 -0700448 pid_t nr = 0;
449
450 rcu_read_lock();
451 if (!ns)
452 ns = current->nsproxy->pid_ns;
453 if (likely(pid_alive(task))) {
454 if (type != PIDTYPE_PID)
455 task = task->group_leader;
456 nr = pid_nr_ns(task->pids[type].pid, ns);
457 }
458 rcu_read_unlock();
459
460 return nr;
Pavel Emelyanov2f2a3a42007-10-18 23:40:19 -0700461}
Oleg Nesterov52ee2df2009-04-02 16:58:38 -0700462EXPORT_SYMBOL(__task_pid_nr_ns);
Pavel Emelyanov2f2a3a42007-10-18 23:40:19 -0700463
464pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
465{
466 return pid_nr_ns(task_tgid(tsk), ns);
467}
468EXPORT_SYMBOL(task_tgid_nr_ns);
469
Eric W. Biederman61bce0f2009-01-07 18:08:49 -0800470struct pid_namespace *task_active_pid_ns(struct task_struct *tsk)
471{
472 return ns_of_pid(task_pid(tsk));
473}
474EXPORT_SYMBOL_GPL(task_active_pid_ns);
475
Linus Torvalds1da177e2005-04-16 15:20:36 -0700476/*
Frederik Schwarzer025dfda2008-10-16 19:02:37 +0200477 * Used by proc to find the first pid that is greater than or equal to nr.
Eric W. Biederman0804ef42006-10-02 02:17:04 -0700478 *
Pavel Emelyanove49859e2008-07-25 01:48:36 -0700479 * 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 -0700480 */
Pavel Emelyanov198fe212007-10-18 23:40:06 -0700481struct pid *find_ge_pid(int nr, struct pid_namespace *ns)
Eric W. Biederman0804ef42006-10-02 02:17:04 -0700482{
483 struct pid *pid;
484
485 do {
Pavel Emelyanov198fe212007-10-18 23:40:06 -0700486 pid = find_pid_ns(nr, ns);
Eric W. Biederman0804ef42006-10-02 02:17:04 -0700487 if (pid)
488 break;
Pavel Emelyanov198fe212007-10-18 23:40:06 -0700489 nr = next_pidmap(ns, nr);
Eric W. Biederman0804ef42006-10-02 02:17:04 -0700490 } while (nr > 0);
491
492 return pid;
493}
494
495/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700496 * The pid hash table is scaled according to the amount of memory in the
497 * machine. From a minimum of 16 slots up to 4096 slots at one gigabyte or
498 * more.
499 */
500void __init pidhash_init(void)
501{
Eric W. Biederman92476d72006-03-31 02:31:42 -0800502 int i, pidhash_size;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700503 unsigned long megabytes = nr_kernel_pages >> (20 - PAGE_SHIFT);
504
505 pidhash_shift = max(4, fls(megabytes * 4));
506 pidhash_shift = min(12, pidhash_shift);
507 pidhash_size = 1 << pidhash_shift;
508
509 printk("PID hash table entries: %d (order: %d, %Zd bytes)\n",
510 pidhash_size, pidhash_shift,
Eric W. Biederman92476d72006-03-31 02:31:42 -0800511 pidhash_size * sizeof(struct hlist_head));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700512
Eric W. Biederman92476d72006-03-31 02:31:42 -0800513 pid_hash = alloc_bootmem(pidhash_size * sizeof(*(pid_hash)));
514 if (!pid_hash)
515 panic("Could not alloc pidhash!\n");
Catalin Marinas12de38b2009-06-29 17:13:55 +0100516 /*
517 * pid_hash contains references to allocated struct pid objects and it
518 * must be scanned by kmemleak to avoid false positives.
519 */
520 kmemleak_alloc(pid_hash, pidhash_size * sizeof(*(pid_hash)), 0,
521 GFP_KERNEL);
Eric W. Biederman92476d72006-03-31 02:31:42 -0800522 for (i = 0; i < pidhash_size; i++)
523 INIT_HLIST_HEAD(&pid_hash[i]);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700524}
525
526void __init pidmap_init(void)
527{
Sukadev Bhattiprolu61a58c62006-12-08 02:37:58 -0800528 init_pid_ns.pidmap[0].page = kzalloc(PAGE_SIZE, GFP_KERNEL);
Oleg Nesterov73b9ebf2006-03-28 16:11:07 -0800529 /* Reserve PID 0. We never call free_pidmap(0) */
Sukadev Bhattiprolu61a58c62006-12-08 02:37:58 -0800530 set_bit(0, init_pid_ns.pidmap[0].page);
531 atomic_dec(&init_pid_ns.pidmap[0].nr_free);
Eric W. Biederman92476d72006-03-31 02:31:42 -0800532
Pavel Emelyanov74bd59b2008-02-08 04:18:24 -0800533 init_pid_ns.pid_cachep = KMEM_CACHE(pid,
534 SLAB_HWCACHE_ALIGN | SLAB_PANIC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700535}