blob: 2e17c9c92cbec3ff3ab5afbb9ec30ce0fbafcb21 [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
Sukadev Bhattiprolu61a58c62006-12-08 02:37:58 -0800125static int alloc_pidmap(struct pid_namespace *pid_ns)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700126{
Sukadev Bhattiprolu61a58c62006-12-08 02:37:58 -0800127 int i, offset, max_scan, pid, last = pid_ns->last_pid;
Sukadev Bhattiprolu6a1f3b82006-10-02 02:17:20 -0700128 struct pidmap *map;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700129
130 pid = last + 1;
131 if (pid >= pid_max)
132 pid = RESERVED_PIDS;
133 offset = pid & BITS_PER_PAGE_MASK;
Sukadev Bhattiprolu61a58c62006-12-08 02:37:58 -0800134 map = &pid_ns->pidmap[pid/BITS_PER_PAGE];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700135 max_scan = (pid_max + BITS_PER_PAGE - 1)/BITS_PER_PAGE - !offset;
136 for (i = 0; i <= max_scan; ++i) {
137 if (unlikely(!map->page)) {
Sukadev Bhattiprolu3fbc9642006-10-02 02:17:24 -0700138 void *page = kzalloc(PAGE_SIZE, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700139 /*
140 * Free the page if someone raced with us
141 * installing it:
142 */
Eric W. Biederman92476d72006-03-31 02:31:42 -0800143 spin_lock_irq(&pidmap_lock);
André Goddard Rosa7be6d992009-12-15 16:47:39 -0800144 if (!map->page) {
Sukadev Bhattiprolu3fbc9642006-10-02 02:17:24 -0700145 map->page = page;
André Goddard Rosa7be6d992009-12-15 16:47:39 -0800146 page = NULL;
147 }
Eric W. Biederman92476d72006-03-31 02:31:42 -0800148 spin_unlock_irq(&pidmap_lock);
André Goddard Rosa7be6d992009-12-15 16:47:39 -0800149 kfree(page);
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
André Goddard Rosa417e3152009-12-15 16:47:40 -0800272 upid = pid->numbers + ns->level;
Eric W. Biederman92476d72006-03-31 02:31:42 -0800273 spin_lock_irq(&pidmap_lock);
André Goddard Rosa417e3152009-12-15 16:47:40 -0800274 for ( ; upid >= pid->numbers; --upid)
Pavel Emelyanov198fe212007-10-18 23:40:06 -0700275 hlist_add_head_rcu(&upid->pid_chain,
276 &pid_hash[pid_hashfn(upid->nr, upid->ns)]);
Eric W. Biederman92476d72006-03-31 02:31:42 -0800277 spin_unlock_irq(&pidmap_lock);
278
279out:
280 return pid;
281
282out_free:
Oleg Nesterovb7127aa2008-04-30 00:54:22 -0700283 while (++i <= ns->level)
284 free_pidmap(pid->numbers + i);
Pavel Emelyanov8ef047a2007-10-18 23:40:05 -0700285
Pavel Emelianovbaf8f0f2007-10-18 23:39:48 -0700286 kmem_cache_free(ns->pid_cachep, pid);
Eric W. Biederman92476d72006-03-31 02:31:42 -0800287 pid = NULL;
288 goto out;
289}
290
Harvey Harrison7ad5b3a2008-02-08 04:19:53 -0800291struct pid *find_pid_ns(int nr, struct pid_namespace *ns)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700292{
293 struct hlist_node *elem;
Pavel Emelyanov198fe212007-10-18 23:40:06 -0700294 struct upid *pnr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700295
Pavel Emelyanov198fe212007-10-18 23:40:06 -0700296 hlist_for_each_entry_rcu(pnr, elem,
297 &pid_hash[pid_hashfn(nr, ns)], pid_chain)
298 if (pnr->nr == nr && pnr->ns == ns)
299 return container_of(pnr, struct pid,
300 numbers[ns->level]);
301
Linus Torvalds1da177e2005-04-16 15:20:36 -0700302 return NULL;
303}
Pavel Emelyanov198fe212007-10-18 23:40:06 -0700304EXPORT_SYMBOL_GPL(find_pid_ns);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700305
Pavel Emelyanov89905712007-10-18 23:40:19 -0700306struct pid *find_vpid(int nr)
307{
308 return find_pid_ns(nr, current->nsproxy->pid_ns);
309}
310EXPORT_SYMBOL_GPL(find_vpid);
311
Sukadev Bhattiprolue713d0d2007-05-10 22:22:58 -0700312/*
313 * attach_pid() must be called with the tasklist_lock write-held.
314 */
Oleg Nesterov24336ea2008-04-30 00:54:26 -0700315void attach_pid(struct task_struct *task, enum pid_type type,
Sukadev Bhattiprolue713d0d2007-05-10 22:22:58 -0700316 struct pid *pid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700317{
Eric W. Biederman92476d72006-03-31 02:31:42 -0800318 struct pid_link *link;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700319
Eric W. Biederman92476d72006-03-31 02:31:42 -0800320 link = &task->pids[type];
Sukadev Bhattiprolue713d0d2007-05-10 22:22:58 -0700321 link->pid = pid;
Eric W. Biederman92476d72006-03-31 02:31:42 -0800322 hlist_add_head_rcu(&link->node, &pid->tasks[type]);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700323}
324
Oleg Nesterov24336ea2008-04-30 00:54:26 -0700325static void __change_pid(struct task_struct *task, enum pid_type type,
326 struct pid *new)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700327{
Eric W. Biederman92476d72006-03-31 02:31:42 -0800328 struct pid_link *link;
329 struct pid *pid;
330 int tmp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700331
Eric W. Biederman92476d72006-03-31 02:31:42 -0800332 link = &task->pids[type];
333 pid = link->pid;
334
335 hlist_del_rcu(&link->node);
Oleg Nesterov24336ea2008-04-30 00:54:26 -0700336 link->pid = new;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700337
338 for (tmp = PIDTYPE_MAX; --tmp >= 0; )
Eric W. Biederman92476d72006-03-31 02:31:42 -0800339 if (!hlist_empty(&pid->tasks[tmp]))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700340 return;
341
Eric W. Biederman92476d72006-03-31 02:31:42 -0800342 free_pid(pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700343}
344
Oleg Nesterov24336ea2008-04-30 00:54:26 -0700345void detach_pid(struct task_struct *task, enum pid_type type)
346{
347 __change_pid(task, type, NULL);
348}
349
350void change_pid(struct task_struct *task, enum pid_type type,
351 struct pid *pid)
352{
353 __change_pid(task, type, pid);
354 attach_pid(task, type, pid);
355}
356
Eric W. Biedermanc18258c2006-09-27 01:51:06 -0700357/* transfer_pid is an optimization of attach_pid(new), detach_pid(old) */
Harvey Harrison7ad5b3a2008-02-08 04:19:53 -0800358void transfer_pid(struct task_struct *old, struct task_struct *new,
Eric W. Biedermanc18258c2006-09-27 01:51:06 -0700359 enum pid_type type)
360{
361 new->pids[type].pid = old->pids[type].pid;
362 hlist_replace_rcu(&old->pids[type].node, &new->pids[type].node);
Eric W. Biedermanc18258c2006-09-27 01:51:06 -0700363}
364
Harvey Harrison7ad5b3a2008-02-08 04:19:53 -0800365struct task_struct *pid_task(struct pid *pid, enum pid_type type)
Eric W. Biederman92476d72006-03-31 02:31:42 -0800366{
367 struct task_struct *result = NULL;
368 if (pid) {
369 struct hlist_node *first;
370 first = rcu_dereference(pid->tasks[type].first);
371 if (first)
372 result = hlist_entry(first, struct task_struct, pids[(type)].node);
373 }
374 return result;
375}
Pavel Emelyanoveccba062008-02-07 00:13:21 -0800376EXPORT_SYMBOL(pid_task);
Eric W. Biederman92476d72006-03-31 02:31:42 -0800377
378/*
379 * Must be called under rcu_read_lock() or with tasklist_lock read-held.
380 */
Christoph Hellwig17f98dc2009-06-17 16:27:51 -0700381struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700382{
Christoph Hellwig17f98dc2009-06-17 16:27:51 -0700383 return pid_task(find_pid_ns(nr, ns), PIDTYPE_PID);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700384}
385
Pavel Emelyanov228ebcb2007-10-18 23:40:16 -0700386struct task_struct *find_task_by_vpid(pid_t vnr)
387{
Christoph Hellwig17f98dc2009-06-17 16:27:51 -0700388 return find_task_by_pid_ns(vnr, current->nsproxy->pid_ns);
Pavel Emelyanov228ebcb2007-10-18 23:40:16 -0700389}
Pavel Emelyanov228ebcb2007-10-18 23:40:16 -0700390
Oleg Nesterov1a657f782006-10-02 02:18:59 -0700391struct pid *get_task_pid(struct task_struct *task, enum pid_type type)
392{
393 struct pid *pid;
394 rcu_read_lock();
Oleg Nesterov2ae448e2009-04-02 16:58:36 -0700395 if (type != PIDTYPE_PID)
396 task = task->group_leader;
Oleg Nesterov1a657f782006-10-02 02:18:59 -0700397 pid = get_pid(task->pids[type].pid);
398 rcu_read_unlock();
399 return pid;
400}
401
Harvey Harrison7ad5b3a2008-02-08 04:19:53 -0800402struct task_struct *get_pid_task(struct pid *pid, enum pid_type type)
Eric W. Biederman92476d72006-03-31 02:31:42 -0800403{
404 struct task_struct *result;
405 rcu_read_lock();
406 result = pid_task(pid, type);
407 if (result)
408 get_task_struct(result);
409 rcu_read_unlock();
410 return result;
411}
412
413struct pid *find_get_pid(pid_t nr)
414{
415 struct pid *pid;
416
417 rcu_read_lock();
Pavel Emelyanov198fe212007-10-18 23:40:06 -0700418 pid = get_pid(find_vpid(nr));
Eric W. Biederman92476d72006-03-31 02:31:42 -0800419 rcu_read_unlock();
420
421 return pid;
422}
David Sterba339caf22008-07-25 01:48:31 -0700423EXPORT_SYMBOL_GPL(find_get_pid);
Eric W. Biederman92476d72006-03-31 02:31:42 -0800424
Pavel Emelyanov7af57292007-10-18 23:40:06 -0700425pid_t pid_nr_ns(struct pid *pid, struct pid_namespace *ns)
426{
427 struct upid *upid;
428 pid_t nr = 0;
429
430 if (pid && ns->level <= pid->level) {
431 upid = &pid->numbers[ns->level];
432 if (upid->ns == ns)
433 nr = upid->nr;
434 }
435 return nr;
436}
437
Eric W. Biederman44c4e1b2008-02-08 04:19:15 -0800438pid_t pid_vnr(struct pid *pid)
439{
440 return pid_nr_ns(pid, current->nsproxy->pid_ns);
441}
442EXPORT_SYMBOL_GPL(pid_vnr);
443
Oleg Nesterov52ee2df2009-04-02 16:58:38 -0700444pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type,
445 struct pid_namespace *ns)
Pavel Emelyanov2f2a3a42007-10-18 23:40:19 -0700446{
Oleg Nesterov52ee2df2009-04-02 16:58:38 -0700447 pid_t nr = 0;
448
449 rcu_read_lock();
450 if (!ns)
451 ns = current->nsproxy->pid_ns;
452 if (likely(pid_alive(task))) {
453 if (type != PIDTYPE_PID)
454 task = task->group_leader;
455 nr = pid_nr_ns(task->pids[type].pid, ns);
456 }
457 rcu_read_unlock();
458
459 return nr;
Pavel Emelyanov2f2a3a42007-10-18 23:40:19 -0700460}
Oleg Nesterov52ee2df2009-04-02 16:58:38 -0700461EXPORT_SYMBOL(__task_pid_nr_ns);
Pavel Emelyanov2f2a3a42007-10-18 23:40:19 -0700462
463pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
464{
465 return pid_nr_ns(task_tgid(tsk), ns);
466}
467EXPORT_SYMBOL(task_tgid_nr_ns);
468
Eric W. Biederman61bce0f2009-01-07 18:08:49 -0800469struct pid_namespace *task_active_pid_ns(struct task_struct *tsk)
470{
471 return ns_of_pid(task_pid(tsk));
472}
473EXPORT_SYMBOL_GPL(task_active_pid_ns);
474
Linus Torvalds1da177e2005-04-16 15:20:36 -0700475/*
Frederik Schwarzer025dfda2008-10-16 19:02:37 +0200476 * Used by proc to find the first pid that is greater than or equal to nr.
Eric W. Biederman0804ef42006-10-02 02:17:04 -0700477 *
Pavel Emelyanove49859e2008-07-25 01:48:36 -0700478 * 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 -0700479 */
Pavel Emelyanov198fe212007-10-18 23:40:06 -0700480struct pid *find_ge_pid(int nr, struct pid_namespace *ns)
Eric W. Biederman0804ef42006-10-02 02:17:04 -0700481{
482 struct pid *pid;
483
484 do {
Pavel Emelyanov198fe212007-10-18 23:40:06 -0700485 pid = find_pid_ns(nr, ns);
Eric W. Biederman0804ef42006-10-02 02:17:04 -0700486 if (pid)
487 break;
Pavel Emelyanov198fe212007-10-18 23:40:06 -0700488 nr = next_pidmap(ns, nr);
Eric W. Biederman0804ef42006-10-02 02:17:04 -0700489 } while (nr > 0);
490
491 return pid;
492}
493
494/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700495 * The pid hash table is scaled according to the amount of memory in the
496 * machine. From a minimum of 16 slots up to 4096 slots at one gigabyte or
497 * more.
498 */
499void __init pidhash_init(void)
500{
Eric W. Biederman92476d72006-03-31 02:31:42 -0800501 int i, pidhash_size;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700502
Jan Beulich2c85f512009-09-21 17:03:07 -0700503 pid_hash = alloc_large_system_hash("PID", sizeof(*pid_hash), 0, 18,
504 HASH_EARLY | HASH_SMALL,
505 &pidhash_shift, NULL, 4096);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700506 pidhash_size = 1 << pidhash_shift;
507
Eric W. Biederman92476d72006-03-31 02:31:42 -0800508 for (i = 0; i < pidhash_size; i++)
509 INIT_HLIST_HEAD(&pid_hash[i]);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700510}
511
512void __init pidmap_init(void)
513{
Sukadev Bhattiprolu61a58c62006-12-08 02:37:58 -0800514 init_pid_ns.pidmap[0].page = kzalloc(PAGE_SIZE, GFP_KERNEL);
Oleg Nesterov73b9ebf2006-03-28 16:11:07 -0800515 /* Reserve PID 0. We never call free_pidmap(0) */
Sukadev Bhattiprolu61a58c62006-12-08 02:37:58 -0800516 set_bit(0, init_pid_ns.pidmap[0].page);
517 atomic_dec(&init_pid_ns.pidmap[0].nr_free);
Eric W. Biederman92476d72006-03-31 02:31:42 -0800518
Pavel Emelyanov74bd59b2008-02-08 04:18:24 -0800519 init_pid_ns.pid_cachep = KMEM_CACHE(pid,
520 SLAB_HWCACHE_ALIGN | SLAB_PANIC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700521}