blob: 39b65b69584f5b0f373e360d6ecc3118751840a9 [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>
Linus Torvalds1da177e2005-04-16 15:20:36 -070039
Pavel Emelyanov8ef047a2007-10-18 23:40:05 -070040#define pid_hashfn(nr, ns) \
41 hash_long((unsigned long)nr + (unsigned long)ns, pidhash_shift)
Eric W. Biederman92476d72006-03-31 02:31:42 -080042static struct hlist_head *pid_hash;
Jan Beulich2c85f512009-09-21 17:03:07 -070043static unsigned int pidhash_shift = 4;
Sukadev Bhattiprolu820e45d2007-05-10 22:23:00 -070044struct pid init_struct_pid = INIT_STRUCT_PID;
Linus Torvalds1da177e2005-04-16 15:20:36 -070045
46int pid_max = PID_MAX_DEFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -070047
48#define RESERVED_PIDS 300
49
50int pid_max_min = RESERVED_PIDS + 1;
51int pid_max_max = PID_MAX_LIMIT;
52
Linus Torvalds1da177e2005-04-16 15:20:36 -070053#define BITS_PER_PAGE (PAGE_SIZE*8)
54#define BITS_PER_PAGE_MASK (BITS_PER_PAGE-1)
Sukadev Bhattiprolu3fbc9642006-10-02 02:17:24 -070055
Sukadev Bhattiprolu61a58c62006-12-08 02:37:58 -080056static inline int mk_pid(struct pid_namespace *pid_ns,
57 struct pidmap *map, int off)
Sukadev Bhattiprolu3fbc9642006-10-02 02:17:24 -070058{
Sukadev Bhattiprolu61a58c62006-12-08 02:37:58 -080059 return (map - pid_ns->pidmap)*BITS_PER_PAGE + off;
Sukadev Bhattiprolu3fbc9642006-10-02 02:17:24 -070060}
61
Linus Torvalds1da177e2005-04-16 15:20:36 -070062#define find_next_offset(map, off) \
63 find_next_zero_bit((map)->page, BITS_PER_PAGE, off)
64
65/*
66 * PID-map pages start out as NULL, they get allocated upon
67 * first use and are never deallocated. This way a low pid_max
68 * value does not cause lots of bitmaps to be allocated, but
69 * the scheme scales to up to 4 million PIDs, runtime.
70 */
Sukadev Bhattiprolu61a58c62006-12-08 02:37:58 -080071struct pid_namespace init_pid_ns = {
Cedric Le Goater9a575a92006-12-08 02:37:59 -080072 .kref = {
73 .refcount = ATOMIC_INIT(2),
74 },
Sukadev Bhattiprolu3fbc9642006-10-02 02:17:24 -070075 .pidmap = {
76 [ 0 ... PIDMAP_ENTRIES-1] = { ATOMIC_INIT(BITS_PER_PAGE), NULL }
77 },
Sukadev Bhattiprolu84d73782006-12-08 02:38:01 -080078 .last_pid = 0,
Pavel Emelyanovfaacbfd2007-10-18 23:40:04 -070079 .level = 0,
80 .child_reaper = &init_task,
Sukadev Bhattiprolu3fbc9642006-10-02 02:17:24 -070081};
Pavel Emelyanov198fe212007-10-18 23:40:06 -070082EXPORT_SYMBOL_GPL(init_pid_ns);
Linus Torvalds1da177e2005-04-16 15:20:36 -070083
Pavel Emelyanovb461cc02007-10-18 23:40:09 -070084int is_container_init(struct task_struct *tsk)
Serge E. Hallynb460cbc2007-10-18 23:39:52 -070085{
Pavel Emelyanovb461cc02007-10-18 23:40:09 -070086 int ret = 0;
87 struct pid *pid;
88
89 rcu_read_lock();
90 pid = task_pid(tsk);
91 if (pid != NULL && pid->numbers[pid->level].nr == 1)
92 ret = 1;
93 rcu_read_unlock();
94
95 return ret;
Serge E. Hallynb460cbc2007-10-18 23:39:52 -070096}
Pavel Emelyanovb461cc02007-10-18 23:40:09 -070097EXPORT_SYMBOL(is_container_init);
Serge E. Hallynb460cbc2007-10-18 23:39:52 -070098
Eric W. Biederman92476d72006-03-31 02:31:42 -080099/*
100 * Note: disable interrupts while the pidmap_lock is held as an
101 * interrupt might come in and do read_lock(&tasklist_lock).
102 *
103 * If we don't disable interrupts there is a nasty deadlock between
104 * detach_pid()->free_pid() and another cpu that does
105 * spin_lock(&pidmap_lock) followed by an interrupt routine that does
106 * read_lock(&tasklist_lock);
107 *
108 * After we clean up the tasklist_lock and know there are no
109 * irq handlers that take it we can leave the interrupts enabled.
110 * For now it is easier to be safe than to prove it can't happen.
111 */
Sukadev Bhattiprolu3fbc9642006-10-02 02:17:24 -0700112
Linus Torvalds1da177e2005-04-16 15:20:36 -0700113static __cacheline_aligned_in_smp DEFINE_SPINLOCK(pidmap_lock);
114
Oleg Nesterovb7127aa2008-04-30 00:54:22 -0700115static void free_pidmap(struct upid *upid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700116{
Oleg Nesterovb7127aa2008-04-30 00:54:22 -0700117 int nr = upid->nr;
118 struct pidmap *map = upid->ns->pidmap + nr / BITS_PER_PAGE;
119 int offset = nr & BITS_PER_PAGE_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700120
121 clear_bit(offset, map->page);
122 atomic_inc(&map->nr_free);
123}
124
Salman5fdee8c2010-08-10 18:03:16 -0700125/*
126 * If we started walking pids at 'base', is 'a' seen before 'b'?
127 */
128static int pid_before(int base, int a, int b)
129{
130 /*
131 * This is the same as saying
132 *
133 * (a - base + MAXUINT) % MAXUINT < (b - base + MAXUINT) % MAXUINT
134 * and that mapping orders 'a' and 'b' with respect to 'base'.
135 */
136 return (unsigned)(a - base) < (unsigned)(b - base);
137}
138
139/*
140 * We might be racing with someone else trying to set pid_ns->last_pid.
141 * We want the winner to have the "later" value, because if the
142 * "earlier" value prevails, then a pid may get reused immediately.
143 *
144 * Since pids rollover, it is not sufficient to just pick the bigger
145 * value. We have to consider where we started counting from.
146 *
147 * 'base' is the value of pid_ns->last_pid that we observed when
148 * we started looking for a pid.
149 *
150 * 'pid' is the pid that we eventually found.
151 */
152static void set_last_pid(struct pid_namespace *pid_ns, int base, int pid)
153{
154 int prev;
155 int last_write = base;
156 do {
157 prev = last_write;
158 last_write = cmpxchg(&pid_ns->last_pid, prev, pid);
159 } while ((prev != last_write) && (pid_before(base, last_write, pid)));
160}
161
Sukadev Bhattiprolu61a58c62006-12-08 02:37:58 -0800162static int alloc_pidmap(struct pid_namespace *pid_ns)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700163{
Sukadev Bhattiprolu61a58c62006-12-08 02:37:58 -0800164 int i, offset, max_scan, pid, last = pid_ns->last_pid;
Sukadev Bhattiprolu6a1f3b82006-10-02 02:17:20 -0700165 struct pidmap *map;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700166
167 pid = last + 1;
168 if (pid >= pid_max)
169 pid = RESERVED_PIDS;
170 offset = pid & BITS_PER_PAGE_MASK;
Sukadev Bhattiprolu61a58c62006-12-08 02:37:58 -0800171 map = &pid_ns->pidmap[pid/BITS_PER_PAGE];
Oleg Nesterovc52b0b92010-08-10 18:03:17 -0700172 /*
173 * If last_pid points into the middle of the map->page we
174 * want to scan this bitmap block twice, the second time
175 * we start with offset == 0 (or RESERVED_PIDS).
176 */
177 max_scan = DIV_ROUND_UP(pid_max, BITS_PER_PAGE) - !offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700178 for (i = 0; i <= max_scan; ++i) {
179 if (unlikely(!map->page)) {
Sukadev Bhattiprolu3fbc9642006-10-02 02:17:24 -0700180 void *page = kzalloc(PAGE_SIZE, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700181 /*
182 * Free the page if someone raced with us
183 * installing it:
184 */
Eric W. Biederman92476d72006-03-31 02:31:42 -0800185 spin_lock_irq(&pidmap_lock);
André Goddard Rosa7be6d992009-12-15 16:47:39 -0800186 if (!map->page) {
Sukadev Bhattiprolu3fbc9642006-10-02 02:17:24 -0700187 map->page = page;
André Goddard Rosa7be6d992009-12-15 16:47:39 -0800188 page = NULL;
189 }
Eric W. Biederman92476d72006-03-31 02:31:42 -0800190 spin_unlock_irq(&pidmap_lock);
André Goddard Rosa7be6d992009-12-15 16:47:39 -0800191 kfree(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700192 if (unlikely(!map->page))
193 break;
194 }
195 if (likely(atomic_read(&map->nr_free))) {
196 do {
197 if (!test_and_set_bit(offset, map->page)) {
198 atomic_dec(&map->nr_free);
Salman5fdee8c2010-08-10 18:03:16 -0700199 set_last_pid(pid_ns, last, pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700200 return pid;
201 }
202 offset = find_next_offset(map, offset);
Sukadev Bhattiprolu61a58c62006-12-08 02:37:58 -0800203 pid = mk_pid(pid_ns, map, offset);
Oleg Nesterovc52b0b92010-08-10 18:03:17 -0700204 } while (offset < BITS_PER_PAGE && pid < pid_max);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700205 }
Sukadev Bhattiprolu61a58c62006-12-08 02:37:58 -0800206 if (map < &pid_ns->pidmap[(pid_max-1)/BITS_PER_PAGE]) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700207 ++map;
208 offset = 0;
209 } else {
Sukadev Bhattiprolu61a58c62006-12-08 02:37:58 -0800210 map = &pid_ns->pidmap[0];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700211 offset = RESERVED_PIDS;
212 if (unlikely(last == offset))
213 break;
214 }
Sukadev Bhattiprolu61a58c62006-12-08 02:37:58 -0800215 pid = mk_pid(pid_ns, map, offset);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700216 }
217 return -1;
218}
219
Pavel Emelyanov74bd59b2008-02-08 04:18:24 -0800220int next_pidmap(struct pid_namespace *pid_ns, int last)
Eric W. Biederman0804ef42006-10-02 02:17:04 -0700221{
222 int offset;
Eric W. Biedermanf40f50d2006-10-02 02:17:25 -0700223 struct pidmap *map, *end;
Eric W. Biederman0804ef42006-10-02 02:17:04 -0700224
225 offset = (last + 1) & BITS_PER_PAGE_MASK;
Sukadev Bhattiprolu61a58c62006-12-08 02:37:58 -0800226 map = &pid_ns->pidmap[(last + 1)/BITS_PER_PAGE];
227 end = &pid_ns->pidmap[PIDMAP_ENTRIES];
Eric W. Biedermanf40f50d2006-10-02 02:17:25 -0700228 for (; map < end; map++, offset = 0) {
Eric W. Biederman0804ef42006-10-02 02:17:04 -0700229 if (unlikely(!map->page))
230 continue;
231 offset = find_next_bit((map)->page, BITS_PER_PAGE, offset);
232 if (offset < BITS_PER_PAGE)
Sukadev Bhattiprolu61a58c62006-12-08 02:37:58 -0800233 return mk_pid(pid_ns, map, offset);
Eric W. Biederman0804ef42006-10-02 02:17:04 -0700234 }
235 return -1;
236}
237
Harvey Harrison7ad5b3a2008-02-08 04:19:53 -0800238void put_pid(struct pid *pid)
Eric W. Biederman92476d72006-03-31 02:31:42 -0800239{
Pavel Emelianovbaf8f0f2007-10-18 23:39:48 -0700240 struct pid_namespace *ns;
241
Eric W. Biederman92476d72006-03-31 02:31:42 -0800242 if (!pid)
243 return;
Pavel Emelianovbaf8f0f2007-10-18 23:39:48 -0700244
Pavel Emelyanov8ef047a2007-10-18 23:40:05 -0700245 ns = pid->numbers[pid->level].ns;
Eric W. Biederman92476d72006-03-31 02:31:42 -0800246 if ((atomic_read(&pid->count) == 1) ||
Pavel Emelyanov8ef047a2007-10-18 23:40:05 -0700247 atomic_dec_and_test(&pid->count)) {
Pavel Emelianovbaf8f0f2007-10-18 23:39:48 -0700248 kmem_cache_free(ns->pid_cachep, pid);
Pavel Emelyanovb461cc02007-10-18 23:40:09 -0700249 put_pid_ns(ns);
Pavel Emelyanov8ef047a2007-10-18 23:40:05 -0700250 }
Eric W. Biederman92476d72006-03-31 02:31:42 -0800251}
Eric W. Biedermanbbf73142006-10-02 02:17:11 -0700252EXPORT_SYMBOL_GPL(put_pid);
Eric W. Biederman92476d72006-03-31 02:31:42 -0800253
254static void delayed_put_pid(struct rcu_head *rhp)
255{
256 struct pid *pid = container_of(rhp, struct pid, rcu);
257 put_pid(pid);
258}
259
Harvey Harrison7ad5b3a2008-02-08 04:19:53 -0800260void free_pid(struct pid *pid)
Eric W. Biederman92476d72006-03-31 02:31:42 -0800261{
262 /* We can be called with write_lock_irq(&tasklist_lock) held */
Pavel Emelyanov8ef047a2007-10-18 23:40:05 -0700263 int i;
Eric W. Biederman92476d72006-03-31 02:31:42 -0800264 unsigned long flags;
265
266 spin_lock_irqsave(&pidmap_lock, flags);
Pavel Emelyanov198fe212007-10-18 23:40:06 -0700267 for (i = 0; i <= pid->level; i++)
268 hlist_del_rcu(&pid->numbers[i].pid_chain);
Eric W. Biederman92476d72006-03-31 02:31:42 -0800269 spin_unlock_irqrestore(&pidmap_lock, flags);
270
Pavel Emelyanov8ef047a2007-10-18 23:40:05 -0700271 for (i = 0; i <= pid->level; i++)
Oleg Nesterovb7127aa2008-04-30 00:54:22 -0700272 free_pidmap(pid->numbers + i);
Pavel Emelyanov8ef047a2007-10-18 23:40:05 -0700273
Eric W. Biederman92476d72006-03-31 02:31:42 -0800274 call_rcu(&pid->rcu, delayed_put_pid);
275}
276
Pavel Emelyanov8ef047a2007-10-18 23:40:05 -0700277struct pid *alloc_pid(struct pid_namespace *ns)
Eric W. Biederman92476d72006-03-31 02:31:42 -0800278{
279 struct pid *pid;
280 enum pid_type type;
Pavel Emelyanov8ef047a2007-10-18 23:40:05 -0700281 int i, nr;
282 struct pid_namespace *tmp;
Pavel Emelyanov198fe212007-10-18 23:40:06 -0700283 struct upid *upid;
Eric W. Biederman92476d72006-03-31 02:31:42 -0800284
Pavel Emelianovbaf8f0f2007-10-18 23:39:48 -0700285 pid = kmem_cache_alloc(ns->pid_cachep, GFP_KERNEL);
Eric W. Biederman92476d72006-03-31 02:31:42 -0800286 if (!pid)
287 goto out;
288
Pavel Emelyanov8ef047a2007-10-18 23:40:05 -0700289 tmp = ns;
290 for (i = ns->level; i >= 0; i--) {
291 nr = alloc_pidmap(tmp);
292 if (nr < 0)
293 goto out_free;
Eric W. Biederman92476d72006-03-31 02:31:42 -0800294
Pavel Emelyanov8ef047a2007-10-18 23:40:05 -0700295 pid->numbers[i].nr = nr;
296 pid->numbers[i].ns = tmp;
297 tmp = tmp->parent;
298 }
299
Pavel Emelyanovb461cc02007-10-18 23:40:09 -0700300 get_pid_ns(ns);
Pavel Emelyanov8ef047a2007-10-18 23:40:05 -0700301 pid->level = ns->level;
Eric W. Biederman92476d72006-03-31 02:31:42 -0800302 atomic_set(&pid->count, 1);
Eric W. Biederman92476d72006-03-31 02:31:42 -0800303 for (type = 0; type < PIDTYPE_MAX; ++type)
304 INIT_HLIST_HEAD(&pid->tasks[type]);
305
André Goddard Rosa417e3152009-12-15 16:47:40 -0800306 upid = pid->numbers + ns->level;
Eric W. Biederman92476d72006-03-31 02:31:42 -0800307 spin_lock_irq(&pidmap_lock);
André Goddard Rosa417e3152009-12-15 16:47:40 -0800308 for ( ; upid >= pid->numbers; --upid)
Pavel Emelyanov198fe212007-10-18 23:40:06 -0700309 hlist_add_head_rcu(&upid->pid_chain,
310 &pid_hash[pid_hashfn(upid->nr, upid->ns)]);
Eric W. Biederman92476d72006-03-31 02:31:42 -0800311 spin_unlock_irq(&pidmap_lock);
312
313out:
314 return pid;
315
316out_free:
Oleg Nesterovb7127aa2008-04-30 00:54:22 -0700317 while (++i <= ns->level)
318 free_pidmap(pid->numbers + i);
Pavel Emelyanov8ef047a2007-10-18 23:40:05 -0700319
Pavel Emelianovbaf8f0f2007-10-18 23:39:48 -0700320 kmem_cache_free(ns->pid_cachep, pid);
Eric W. Biederman92476d72006-03-31 02:31:42 -0800321 pid = NULL;
322 goto out;
323}
324
Harvey Harrison7ad5b3a2008-02-08 04:19:53 -0800325struct pid *find_pid_ns(int nr, struct pid_namespace *ns)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700326{
327 struct hlist_node *elem;
Pavel Emelyanov198fe212007-10-18 23:40:06 -0700328 struct upid *pnr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700329
Pavel Emelyanov198fe212007-10-18 23:40:06 -0700330 hlist_for_each_entry_rcu(pnr, elem,
331 &pid_hash[pid_hashfn(nr, ns)], pid_chain)
332 if (pnr->nr == nr && pnr->ns == ns)
333 return container_of(pnr, struct pid,
334 numbers[ns->level]);
335
Linus Torvalds1da177e2005-04-16 15:20:36 -0700336 return NULL;
337}
Pavel Emelyanov198fe212007-10-18 23:40:06 -0700338EXPORT_SYMBOL_GPL(find_pid_ns);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700339
Pavel Emelyanov89905712007-10-18 23:40:19 -0700340struct pid *find_vpid(int nr)
341{
342 return find_pid_ns(nr, current->nsproxy->pid_ns);
343}
344EXPORT_SYMBOL_GPL(find_vpid);
345
Sukadev Bhattiprolue713d0d2007-05-10 22:22:58 -0700346/*
347 * attach_pid() must be called with the tasklist_lock write-held.
348 */
Oleg Nesterov24336ea2008-04-30 00:54:26 -0700349void attach_pid(struct task_struct *task, enum pid_type type,
Sukadev Bhattiprolue713d0d2007-05-10 22:22:58 -0700350 struct pid *pid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700351{
Eric W. Biederman92476d72006-03-31 02:31:42 -0800352 struct pid_link *link;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700353
Eric W. Biederman92476d72006-03-31 02:31:42 -0800354 link = &task->pids[type];
Sukadev Bhattiprolue713d0d2007-05-10 22:22:58 -0700355 link->pid = pid;
Eric W. Biederman92476d72006-03-31 02:31:42 -0800356 hlist_add_head_rcu(&link->node, &pid->tasks[type]);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700357}
358
Oleg Nesterov24336ea2008-04-30 00:54:26 -0700359static void __change_pid(struct task_struct *task, enum pid_type type,
360 struct pid *new)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700361{
Eric W. Biederman92476d72006-03-31 02:31:42 -0800362 struct pid_link *link;
363 struct pid *pid;
364 int tmp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700365
Eric W. Biederman92476d72006-03-31 02:31:42 -0800366 link = &task->pids[type];
367 pid = link->pid;
368
369 hlist_del_rcu(&link->node);
Oleg Nesterov24336ea2008-04-30 00:54:26 -0700370 link->pid = new;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700371
372 for (tmp = PIDTYPE_MAX; --tmp >= 0; )
Eric W. Biederman92476d72006-03-31 02:31:42 -0800373 if (!hlist_empty(&pid->tasks[tmp]))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700374 return;
375
Eric W. Biederman92476d72006-03-31 02:31:42 -0800376 free_pid(pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700377}
378
Oleg Nesterov24336ea2008-04-30 00:54:26 -0700379void detach_pid(struct task_struct *task, enum pid_type type)
380{
381 __change_pid(task, type, NULL);
382}
383
384void change_pid(struct task_struct *task, enum pid_type type,
385 struct pid *pid)
386{
387 __change_pid(task, type, pid);
388 attach_pid(task, type, pid);
389}
390
Eric W. Biedermanc18258c2006-09-27 01:51:06 -0700391/* transfer_pid is an optimization of attach_pid(new), detach_pid(old) */
Harvey Harrison7ad5b3a2008-02-08 04:19:53 -0800392void transfer_pid(struct task_struct *old, struct task_struct *new,
Eric W. Biedermanc18258c2006-09-27 01:51:06 -0700393 enum pid_type type)
394{
395 new->pids[type].pid = old->pids[type].pid;
396 hlist_replace_rcu(&old->pids[type].node, &new->pids[type].node);
Eric W. Biedermanc18258c2006-09-27 01:51:06 -0700397}
398
Harvey Harrison7ad5b3a2008-02-08 04:19:53 -0800399struct task_struct *pid_task(struct pid *pid, enum pid_type type)
Eric W. Biederman92476d72006-03-31 02:31:42 -0800400{
401 struct task_struct *result = NULL;
402 if (pid) {
403 struct hlist_node *first;
Arnd Bergmann67bdbff2010-02-25 16:55:13 +0100404 first = rcu_dereference_check(hlist_first_rcu(&pid->tasks[type]),
Paul E. McKenneydb1466b2010-03-03 07:46:56 -0800405 rcu_read_lock_held() ||
406 lockdep_tasklist_lock_is_held());
Eric W. Biederman92476d72006-03-31 02:31:42 -0800407 if (first)
408 result = hlist_entry(first, struct task_struct, pids[(type)].node);
409 }
410 return result;
411}
Pavel Emelyanoveccba062008-02-07 00:13:21 -0800412EXPORT_SYMBOL(pid_task);
Eric W. Biederman92476d72006-03-31 02:31:42 -0800413
414/*
Tetsuo Handa9728e5d2010-03-05 13:42:56 -0800415 * Must be called under rcu_read_lock().
Eric W. Biederman92476d72006-03-31 02:31:42 -0800416 */
Christoph Hellwig17f98dc2009-06-17 16:27:51 -0700417struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700418{
Tetsuo Handa4221a992010-06-26 01:08:19 +0900419 rcu_lockdep_assert(rcu_read_lock_held());
Christoph Hellwig17f98dc2009-06-17 16:27:51 -0700420 return pid_task(find_pid_ns(nr, ns), PIDTYPE_PID);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700421}
422
Pavel Emelyanov228ebcb2007-10-18 23:40:16 -0700423struct task_struct *find_task_by_vpid(pid_t vnr)
424{
Christoph Hellwig17f98dc2009-06-17 16:27:51 -0700425 return find_task_by_pid_ns(vnr, current->nsproxy->pid_ns);
Pavel Emelyanov228ebcb2007-10-18 23:40:16 -0700426}
Pavel Emelyanov228ebcb2007-10-18 23:40:16 -0700427
Oleg Nesterov1a657f782006-10-02 02:18:59 -0700428struct pid *get_task_pid(struct task_struct *task, enum pid_type type)
429{
430 struct pid *pid;
431 rcu_read_lock();
Oleg Nesterov2ae448e2009-04-02 16:58:36 -0700432 if (type != PIDTYPE_PID)
433 task = task->group_leader;
Oleg Nesterov1a657f782006-10-02 02:18:59 -0700434 pid = get_pid(task->pids[type].pid);
435 rcu_read_unlock();
436 return pid;
437}
438
Harvey Harrison7ad5b3a2008-02-08 04:19:53 -0800439struct task_struct *get_pid_task(struct pid *pid, enum pid_type type)
Eric W. Biederman92476d72006-03-31 02:31:42 -0800440{
441 struct task_struct *result;
442 rcu_read_lock();
443 result = pid_task(pid, type);
444 if (result)
445 get_task_struct(result);
446 rcu_read_unlock();
447 return result;
448}
449
450struct pid *find_get_pid(pid_t nr)
451{
452 struct pid *pid;
453
454 rcu_read_lock();
Pavel Emelyanov198fe212007-10-18 23:40:06 -0700455 pid = get_pid(find_vpid(nr));
Eric W. Biederman92476d72006-03-31 02:31:42 -0800456 rcu_read_unlock();
457
458 return pid;
459}
David Sterba339caf22008-07-25 01:48:31 -0700460EXPORT_SYMBOL_GPL(find_get_pid);
Eric W. Biederman92476d72006-03-31 02:31:42 -0800461
Pavel Emelyanov7af57292007-10-18 23:40:06 -0700462pid_t pid_nr_ns(struct pid *pid, struct pid_namespace *ns)
463{
464 struct upid *upid;
465 pid_t nr = 0;
466
467 if (pid && ns->level <= pid->level) {
468 upid = &pid->numbers[ns->level];
469 if (upid->ns == ns)
470 nr = upid->nr;
471 }
472 return nr;
473}
474
Eric W. Biederman44c4e1b2008-02-08 04:19:15 -0800475pid_t pid_vnr(struct pid *pid)
476{
477 return pid_nr_ns(pid, current->nsproxy->pid_ns);
478}
479EXPORT_SYMBOL_GPL(pid_vnr);
480
Oleg Nesterov52ee2df2009-04-02 16:58:38 -0700481pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type,
482 struct pid_namespace *ns)
Pavel Emelyanov2f2a3a42007-10-18 23:40:19 -0700483{
Oleg Nesterov52ee2df2009-04-02 16:58:38 -0700484 pid_t nr = 0;
485
486 rcu_read_lock();
487 if (!ns)
488 ns = current->nsproxy->pid_ns;
489 if (likely(pid_alive(task))) {
490 if (type != PIDTYPE_PID)
491 task = task->group_leader;
492 nr = pid_nr_ns(task->pids[type].pid, ns);
493 }
494 rcu_read_unlock();
495
496 return nr;
Pavel Emelyanov2f2a3a42007-10-18 23:40:19 -0700497}
Oleg Nesterov52ee2df2009-04-02 16:58:38 -0700498EXPORT_SYMBOL(__task_pid_nr_ns);
Pavel Emelyanov2f2a3a42007-10-18 23:40:19 -0700499
500pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
501{
502 return pid_nr_ns(task_tgid(tsk), ns);
503}
504EXPORT_SYMBOL(task_tgid_nr_ns);
505
Eric W. Biederman61bce0f2009-01-07 18:08:49 -0800506struct pid_namespace *task_active_pid_ns(struct task_struct *tsk)
507{
508 return ns_of_pid(task_pid(tsk));
509}
510EXPORT_SYMBOL_GPL(task_active_pid_ns);
511
Linus Torvalds1da177e2005-04-16 15:20:36 -0700512/*
Frederik Schwarzer025dfda2008-10-16 19:02:37 +0200513 * Used by proc to find the first pid that is greater than or equal to nr.
Eric W. Biederman0804ef42006-10-02 02:17:04 -0700514 *
Pavel Emelyanove49859e2008-07-25 01:48:36 -0700515 * 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 -0700516 */
Pavel Emelyanov198fe212007-10-18 23:40:06 -0700517struct pid *find_ge_pid(int nr, struct pid_namespace *ns)
Eric W. Biederman0804ef42006-10-02 02:17:04 -0700518{
519 struct pid *pid;
520
521 do {
Pavel Emelyanov198fe212007-10-18 23:40:06 -0700522 pid = find_pid_ns(nr, ns);
Eric W. Biederman0804ef42006-10-02 02:17:04 -0700523 if (pid)
524 break;
Pavel Emelyanov198fe212007-10-18 23:40:06 -0700525 nr = next_pidmap(ns, nr);
Eric W. Biederman0804ef42006-10-02 02:17:04 -0700526 } while (nr > 0);
527
528 return pid;
529}
530
531/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700532 * The pid hash table is scaled according to the amount of memory in the
533 * machine. From a minimum of 16 slots up to 4096 slots at one gigabyte or
534 * more.
535 */
536void __init pidhash_init(void)
537{
Eric W. Biederman92476d72006-03-31 02:31:42 -0800538 int i, pidhash_size;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700539
Jan Beulich2c85f512009-09-21 17:03:07 -0700540 pid_hash = alloc_large_system_hash("PID", sizeof(*pid_hash), 0, 18,
541 HASH_EARLY | HASH_SMALL,
542 &pidhash_shift, NULL, 4096);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700543 pidhash_size = 1 << pidhash_shift;
544
Eric W. Biederman92476d72006-03-31 02:31:42 -0800545 for (i = 0; i < pidhash_size; i++)
546 INIT_HLIST_HEAD(&pid_hash[i]);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700547}
548
549void __init pidmap_init(void)
550{
Hedi Berriche72680a12010-05-26 14:44:06 -0700551 /* bump default and minimum pid_max based on number of cpus */
552 pid_max = min(pid_max_max, max_t(int, pid_max,
553 PIDS_PER_CPU_DEFAULT * num_possible_cpus()));
554 pid_max_min = max_t(int, pid_max_min,
555 PIDS_PER_CPU_MIN * num_possible_cpus());
556 pr_info("pid_max: default: %u minimum: %u\n", pid_max, pid_max_min);
557
Sukadev Bhattiprolu61a58c62006-12-08 02:37:58 -0800558 init_pid_ns.pidmap[0].page = kzalloc(PAGE_SIZE, GFP_KERNEL);
Oleg Nesterov73b9ebf2006-03-28 16:11:07 -0800559 /* Reserve PID 0. We never call free_pidmap(0) */
Sukadev Bhattiprolu61a58c62006-12-08 02:37:58 -0800560 set_bit(0, init_pid_ns.pidmap[0].page);
561 atomic_dec(&init_pid_ns.pidmap[0].nr_free);
Eric W. Biederman92476d72006-03-31 02:31:42 -0800562
Pavel Emelyanov74bd59b2008-02-08 04:18:24 -0800563 init_pid_ns.pid_cachep = KMEM_CACHE(pid,
564 SLAB_HWCACHE_ALIGN | SLAB_PANIC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700565}