blob: a32859c4a3cd725cebb02f43e11e04a6cdef6888 [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>
33#include <linux/bootmem.h>
34#include <linux/hash.h>
Sukadev Bhattiprolu61a58c62006-12-08 02:37:58 -080035#include <linux/pid_namespace.h>
Sukadev Bhattiprolu820e45d2007-05-10 22:23:00 -070036#include <linux/init_task.h>
Sukadev Bhattiprolu3eb07c82007-10-18 23:40:13 -070037#include <linux/syscalls.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070038
Pavel Emelyanov8ef047a2007-10-18 23:40:05 -070039#define pid_hashfn(nr, ns) \
40 hash_long((unsigned long)nr + (unsigned long)ns, pidhash_shift)
Eric W. Biederman92476d72006-03-31 02:31:42 -080041static struct hlist_head *pid_hash;
Linus Torvalds1da177e2005-04-16 15:20:36 -070042static int pidhash_shift;
Sukadev Bhattiprolu820e45d2007-05-10 22:23:00 -070043struct pid init_struct_pid = INIT_STRUCT_PID;
Linus Torvalds1da177e2005-04-16 15:20:36 -070044
45int pid_max = PID_MAX_DEFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -070046
47#define RESERVED_PIDS 300
48
49int pid_max_min = RESERVED_PIDS + 1;
50int pid_max_max = PID_MAX_LIMIT;
51
Linus Torvalds1da177e2005-04-16 15:20:36 -070052#define BITS_PER_PAGE (PAGE_SIZE*8)
53#define BITS_PER_PAGE_MASK (BITS_PER_PAGE-1)
Sukadev Bhattiprolu3fbc9642006-10-02 02:17:24 -070054
Sukadev Bhattiprolu61a58c62006-12-08 02:37:58 -080055static inline int mk_pid(struct pid_namespace *pid_ns,
56 struct pidmap *map, int off)
Sukadev Bhattiprolu3fbc9642006-10-02 02:17:24 -070057{
Sukadev Bhattiprolu61a58c62006-12-08 02:37:58 -080058 return (map - pid_ns->pidmap)*BITS_PER_PAGE + off;
Sukadev Bhattiprolu3fbc9642006-10-02 02:17:24 -070059}
60
Linus Torvalds1da177e2005-04-16 15:20:36 -070061#define find_next_offset(map, off) \
62 find_next_zero_bit((map)->page, BITS_PER_PAGE, off)
63
64/*
65 * PID-map pages start out as NULL, they get allocated upon
66 * first use and are never deallocated. This way a low pid_max
67 * value does not cause lots of bitmaps to be allocated, but
68 * the scheme scales to up to 4 million PIDs, runtime.
69 */
Sukadev Bhattiprolu61a58c62006-12-08 02:37:58 -080070struct pid_namespace init_pid_ns = {
Cedric Le Goater9a575a92006-12-08 02:37:59 -080071 .kref = {
72 .refcount = ATOMIC_INIT(2),
73 },
Sukadev Bhattiprolu3fbc9642006-10-02 02:17:24 -070074 .pidmap = {
75 [ 0 ... PIDMAP_ENTRIES-1] = { ATOMIC_INIT(BITS_PER_PAGE), NULL }
76 },
Sukadev Bhattiprolu84d73782006-12-08 02:38:01 -080077 .last_pid = 0,
Pavel Emelyanovfaacbfd2007-10-18 23:40:04 -070078 .level = 0,
79 .child_reaper = &init_task,
Sukadev Bhattiprolu3fbc9642006-10-02 02:17:24 -070080};
Pavel Emelyanov198fe212007-10-18 23:40:06 -070081EXPORT_SYMBOL_GPL(init_pid_ns);
Linus Torvalds1da177e2005-04-16 15:20:36 -070082
Pavel Emelyanovb461cc02007-10-18 23:40:09 -070083int is_container_init(struct task_struct *tsk)
Serge E. Hallynb460cbc2007-10-18 23:39:52 -070084{
Pavel Emelyanovb461cc02007-10-18 23:40:09 -070085 int ret = 0;
86 struct pid *pid;
87
88 rcu_read_lock();
89 pid = task_pid(tsk);
90 if (pid != NULL && pid->numbers[pid->level].nr == 1)
91 ret = 1;
92 rcu_read_unlock();
93
94 return ret;
Serge E. Hallynb460cbc2007-10-18 23:39:52 -070095}
Pavel Emelyanovb461cc02007-10-18 23:40:09 -070096EXPORT_SYMBOL(is_container_init);
Serge E. Hallynb460cbc2007-10-18 23:39:52 -070097
Eric W. Biederman92476d72006-03-31 02:31:42 -080098/*
99 * Note: disable interrupts while the pidmap_lock is held as an
100 * interrupt might come in and do read_lock(&tasklist_lock).
101 *
102 * If we don't disable interrupts there is a nasty deadlock between
103 * detach_pid()->free_pid() and another cpu that does
104 * spin_lock(&pidmap_lock) followed by an interrupt routine that does
105 * read_lock(&tasklist_lock);
106 *
107 * After we clean up the tasklist_lock and know there are no
108 * irq handlers that take it we can leave the interrupts enabled.
109 * For now it is easier to be safe than to prove it can't happen.
110 */
Sukadev Bhattiprolu3fbc9642006-10-02 02:17:24 -0700111
Linus Torvalds1da177e2005-04-16 15:20:36 -0700112static __cacheline_aligned_in_smp DEFINE_SPINLOCK(pidmap_lock);
113
Sukadev Bhattiprolu61a58c62006-12-08 02:37:58 -0800114static fastcall void free_pidmap(struct pid_namespace *pid_ns, int pid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700115{
Sukadev Bhattiprolu61a58c62006-12-08 02:37:58 -0800116 struct pidmap *map = pid_ns->pidmap + pid / BITS_PER_PAGE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700117 int offset = pid & BITS_PER_PAGE_MASK;
118
119 clear_bit(offset, map->page);
120 atomic_inc(&map->nr_free);
121}
122
Sukadev Bhattiprolu61a58c62006-12-08 02:37:58 -0800123static int alloc_pidmap(struct pid_namespace *pid_ns)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700124{
Sukadev Bhattiprolu61a58c62006-12-08 02:37:58 -0800125 int i, offset, max_scan, pid, last = pid_ns->last_pid;
Sukadev Bhattiprolu6a1f3b82006-10-02 02:17:20 -0700126 struct pidmap *map;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700127
128 pid = last + 1;
129 if (pid >= pid_max)
130 pid = RESERVED_PIDS;
131 offset = pid & BITS_PER_PAGE_MASK;
Sukadev Bhattiprolu61a58c62006-12-08 02:37:58 -0800132 map = &pid_ns->pidmap[pid/BITS_PER_PAGE];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700133 max_scan = (pid_max + BITS_PER_PAGE - 1)/BITS_PER_PAGE - !offset;
134 for (i = 0; i <= max_scan; ++i) {
135 if (unlikely(!map->page)) {
Sukadev Bhattiprolu3fbc9642006-10-02 02:17:24 -0700136 void *page = kzalloc(PAGE_SIZE, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700137 /*
138 * Free the page if someone raced with us
139 * installing it:
140 */
Eric W. Biederman92476d72006-03-31 02:31:42 -0800141 spin_lock_irq(&pidmap_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700142 if (map->page)
Sukadev Bhattiprolu3fbc9642006-10-02 02:17:24 -0700143 kfree(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700144 else
Sukadev Bhattiprolu3fbc9642006-10-02 02:17:24 -0700145 map->page = page;
Eric W. Biederman92476d72006-03-31 02:31:42 -0800146 spin_unlock_irq(&pidmap_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700147 if (unlikely(!map->page))
148 break;
149 }
150 if (likely(atomic_read(&map->nr_free))) {
151 do {
152 if (!test_and_set_bit(offset, map->page)) {
153 atomic_dec(&map->nr_free);
Sukadev Bhattiprolu61a58c62006-12-08 02:37:58 -0800154 pid_ns->last_pid = pid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700155 return pid;
156 }
157 offset = find_next_offset(map, offset);
Sukadev Bhattiprolu61a58c62006-12-08 02:37:58 -0800158 pid = mk_pid(pid_ns, map, offset);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700159 /*
160 * find_next_offset() found a bit, the pid from it
161 * is in-bounds, and if we fell back to the last
162 * bitmap block and the final block was the same
163 * as the starting point, pid is before last_pid.
164 */
165 } while (offset < BITS_PER_PAGE && pid < pid_max &&
166 (i != max_scan || pid < last ||
167 !((last+1) & BITS_PER_PAGE_MASK)));
168 }
Sukadev Bhattiprolu61a58c62006-12-08 02:37:58 -0800169 if (map < &pid_ns->pidmap[(pid_max-1)/BITS_PER_PAGE]) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700170 ++map;
171 offset = 0;
172 } else {
Sukadev Bhattiprolu61a58c62006-12-08 02:37:58 -0800173 map = &pid_ns->pidmap[0];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700174 offset = RESERVED_PIDS;
175 if (unlikely(last == offset))
176 break;
177 }
Sukadev Bhattiprolu61a58c62006-12-08 02:37:58 -0800178 pid = mk_pid(pid_ns, map, offset);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700179 }
180 return -1;
181}
182
Pavel Emelyanov74bd59b2008-02-08 04:18:24 -0800183int next_pidmap(struct pid_namespace *pid_ns, int last)
Eric W. Biederman0804ef42006-10-02 02:17:04 -0700184{
185 int offset;
Eric W. Biedermanf40f50d2006-10-02 02:17:25 -0700186 struct pidmap *map, *end;
Eric W. Biederman0804ef42006-10-02 02:17:04 -0700187
188 offset = (last + 1) & BITS_PER_PAGE_MASK;
Sukadev Bhattiprolu61a58c62006-12-08 02:37:58 -0800189 map = &pid_ns->pidmap[(last + 1)/BITS_PER_PAGE];
190 end = &pid_ns->pidmap[PIDMAP_ENTRIES];
Eric W. Biedermanf40f50d2006-10-02 02:17:25 -0700191 for (; map < end; map++, offset = 0) {
Eric W. Biederman0804ef42006-10-02 02:17:04 -0700192 if (unlikely(!map->page))
193 continue;
194 offset = find_next_bit((map)->page, BITS_PER_PAGE, offset);
195 if (offset < BITS_PER_PAGE)
Sukadev Bhattiprolu61a58c62006-12-08 02:37:58 -0800196 return mk_pid(pid_ns, map, offset);
Eric W. Biederman0804ef42006-10-02 02:17:04 -0700197 }
198 return -1;
199}
200
Eric W. Biederman92476d72006-03-31 02:31:42 -0800201fastcall void put_pid(struct pid *pid)
202{
Pavel Emelianovbaf8f0f2007-10-18 23:39:48 -0700203 struct pid_namespace *ns;
204
Eric W. Biederman92476d72006-03-31 02:31:42 -0800205 if (!pid)
206 return;
Pavel Emelianovbaf8f0f2007-10-18 23:39:48 -0700207
Pavel Emelyanov8ef047a2007-10-18 23:40:05 -0700208 ns = pid->numbers[pid->level].ns;
Eric W. Biederman92476d72006-03-31 02:31:42 -0800209 if ((atomic_read(&pid->count) == 1) ||
Pavel Emelyanov8ef047a2007-10-18 23:40:05 -0700210 atomic_dec_and_test(&pid->count)) {
Pavel Emelianovbaf8f0f2007-10-18 23:39:48 -0700211 kmem_cache_free(ns->pid_cachep, pid);
Pavel Emelyanovb461cc02007-10-18 23:40:09 -0700212 put_pid_ns(ns);
Pavel Emelyanov8ef047a2007-10-18 23:40:05 -0700213 }
Eric W. Biederman92476d72006-03-31 02:31:42 -0800214}
Eric W. Biedermanbbf73142006-10-02 02:17:11 -0700215EXPORT_SYMBOL_GPL(put_pid);
Eric W. Biederman92476d72006-03-31 02:31:42 -0800216
217static void delayed_put_pid(struct rcu_head *rhp)
218{
219 struct pid *pid = container_of(rhp, struct pid, rcu);
220 put_pid(pid);
221}
222
223fastcall void free_pid(struct pid *pid)
224{
225 /* We can be called with write_lock_irq(&tasklist_lock) held */
Pavel Emelyanov8ef047a2007-10-18 23:40:05 -0700226 int i;
Eric W. Biederman92476d72006-03-31 02:31:42 -0800227 unsigned long flags;
228
229 spin_lock_irqsave(&pidmap_lock, flags);
Pavel Emelyanov198fe212007-10-18 23:40:06 -0700230 for (i = 0; i <= pid->level; i++)
231 hlist_del_rcu(&pid->numbers[i].pid_chain);
Eric W. Biederman92476d72006-03-31 02:31:42 -0800232 spin_unlock_irqrestore(&pidmap_lock, flags);
233
Pavel Emelyanov8ef047a2007-10-18 23:40:05 -0700234 for (i = 0; i <= pid->level; i++)
235 free_pidmap(pid->numbers[i].ns, pid->numbers[i].nr);
236
Eric W. Biederman92476d72006-03-31 02:31:42 -0800237 call_rcu(&pid->rcu, delayed_put_pid);
238}
239
Pavel Emelyanov8ef047a2007-10-18 23:40:05 -0700240struct pid *alloc_pid(struct pid_namespace *ns)
Eric W. Biederman92476d72006-03-31 02:31:42 -0800241{
242 struct pid *pid;
243 enum pid_type type;
Pavel Emelyanov8ef047a2007-10-18 23:40:05 -0700244 int i, nr;
245 struct pid_namespace *tmp;
Pavel Emelyanov198fe212007-10-18 23:40:06 -0700246 struct upid *upid;
Eric W. Biederman92476d72006-03-31 02:31:42 -0800247
Pavel Emelianovbaf8f0f2007-10-18 23:39:48 -0700248 pid = kmem_cache_alloc(ns->pid_cachep, GFP_KERNEL);
Eric W. Biederman92476d72006-03-31 02:31:42 -0800249 if (!pid)
250 goto out;
251
Pavel Emelyanov8ef047a2007-10-18 23:40:05 -0700252 tmp = ns;
253 for (i = ns->level; i >= 0; i--) {
254 nr = alloc_pidmap(tmp);
255 if (nr < 0)
256 goto out_free;
Eric W. Biederman92476d72006-03-31 02:31:42 -0800257
Pavel Emelyanov8ef047a2007-10-18 23:40:05 -0700258 pid->numbers[i].nr = nr;
259 pid->numbers[i].ns = tmp;
260 tmp = tmp->parent;
261 }
262
Pavel Emelyanovb461cc02007-10-18 23:40:09 -0700263 get_pid_ns(ns);
Pavel Emelyanov8ef047a2007-10-18 23:40:05 -0700264 pid->level = ns->level;
Eric W. Biederman92476d72006-03-31 02:31:42 -0800265 atomic_set(&pid->count, 1);
Eric W. Biederman92476d72006-03-31 02:31:42 -0800266 for (type = 0; type < PIDTYPE_MAX; ++type)
267 INIT_HLIST_HEAD(&pid->tasks[type]);
268
269 spin_lock_irq(&pidmap_lock);
Pavel Emelyanov198fe212007-10-18 23:40:06 -0700270 for (i = ns->level; i >= 0; i--) {
271 upid = &pid->numbers[i];
272 hlist_add_head_rcu(&upid->pid_chain,
273 &pid_hash[pid_hashfn(upid->nr, upid->ns)]);
274 }
Eric W. Biederman92476d72006-03-31 02:31:42 -0800275 spin_unlock_irq(&pidmap_lock);
276
277out:
278 return pid;
279
280out_free:
Pavel Emelyanov8ef047a2007-10-18 23:40:05 -0700281 for (i++; i <= ns->level; i++)
282 free_pidmap(pid->numbers[i].ns, pid->numbers[i].nr);
283
Pavel Emelianovbaf8f0f2007-10-18 23:39:48 -0700284 kmem_cache_free(ns->pid_cachep, pid);
Eric W. Biederman92476d72006-03-31 02:31:42 -0800285 pid = NULL;
286 goto out;
287}
288
Pavel Emelyanov198fe212007-10-18 23:40:06 -0700289struct pid * fastcall find_pid_ns(int nr, struct pid_namespace *ns)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700290{
291 struct hlist_node *elem;
Pavel Emelyanov198fe212007-10-18 23:40:06 -0700292 struct upid *pnr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700293
Pavel Emelyanov198fe212007-10-18 23:40:06 -0700294 hlist_for_each_entry_rcu(pnr, elem,
295 &pid_hash[pid_hashfn(nr, ns)], pid_chain)
296 if (pnr->nr == nr && pnr->ns == ns)
297 return container_of(pnr, struct pid,
298 numbers[ns->level]);
299
Linus Torvalds1da177e2005-04-16 15:20:36 -0700300 return NULL;
301}
Pavel Emelyanov198fe212007-10-18 23:40:06 -0700302EXPORT_SYMBOL_GPL(find_pid_ns);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700303
Pavel Emelyanov89905712007-10-18 23:40:19 -0700304struct pid *find_vpid(int nr)
305{
306 return find_pid_ns(nr, current->nsproxy->pid_ns);
307}
308EXPORT_SYMBOL_GPL(find_vpid);
309
310struct pid *find_pid(int nr)
311{
312 return find_pid_ns(nr, &init_pid_ns);
313}
314EXPORT_SYMBOL_GPL(find_pid);
315
Sukadev Bhattiprolue713d0d2007-05-10 22:22:58 -0700316/*
317 * attach_pid() must be called with the tasklist_lock write-held.
318 */
319int fastcall attach_pid(struct task_struct *task, enum pid_type type,
320 struct pid *pid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700321{
Eric W. Biederman92476d72006-03-31 02:31:42 -0800322 struct pid_link *link;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700323
Eric W. Biederman92476d72006-03-31 02:31:42 -0800324 link = &task->pids[type];
Sukadev Bhattiprolue713d0d2007-05-10 22:22:58 -0700325 link->pid = pid;
Eric W. Biederman92476d72006-03-31 02:31:42 -0800326 hlist_add_head_rcu(&link->node, &pid->tasks[type]);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700327
328 return 0;
329}
330
Ingo Molnar36c8b582006-07-03 00:25:41 -0700331void fastcall detach_pid(struct task_struct *task, enum pid_type type)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700332{
Eric W. Biederman92476d72006-03-31 02:31:42 -0800333 struct pid_link *link;
334 struct pid *pid;
335 int tmp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700336
Eric W. Biederman92476d72006-03-31 02:31:42 -0800337 link = &task->pids[type];
338 pid = link->pid;
339
340 hlist_del_rcu(&link->node);
341 link->pid = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700342
343 for (tmp = PIDTYPE_MAX; --tmp >= 0; )
Eric W. Biederman92476d72006-03-31 02:31:42 -0800344 if (!hlist_empty(&pid->tasks[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
Eric W. Biedermanc18258c2006-09-27 01:51:06 -0700350/* transfer_pid is an optimization of attach_pid(new), detach_pid(old) */
351void fastcall transfer_pid(struct task_struct *old, struct task_struct *new,
352 enum pid_type type)
353{
354 new->pids[type].pid = old->pids[type].pid;
355 hlist_replace_rcu(&old->pids[type].node, &new->pids[type].node);
356 old->pids[type].pid = NULL;
357}
358
Eric W. Biederman92476d72006-03-31 02:31:42 -0800359struct task_struct * fastcall pid_task(struct pid *pid, enum pid_type type)
360{
361 struct task_struct *result = NULL;
362 if (pid) {
363 struct hlist_node *first;
364 first = rcu_dereference(pid->tasks[type].first);
365 if (first)
366 result = hlist_entry(first, struct task_struct, pids[(type)].node);
367 }
368 return result;
369}
Pavel Emelyanoveccba062008-02-07 00:13:21 -0800370EXPORT_SYMBOL(pid_task);
Eric W. Biederman92476d72006-03-31 02:31:42 -0800371
372/*
373 * Must be called under rcu_read_lock() or with tasklist_lock read-held.
374 */
Pavel Emelyanov198fe212007-10-18 23:40:06 -0700375struct task_struct *find_task_by_pid_type_ns(int type, int nr,
376 struct pid_namespace *ns)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700377{
Pavel Emelyanov198fe212007-10-18 23:40:06 -0700378 return pid_task(find_pid_ns(nr, ns), type);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700379}
380
Pavel Emelyanov198fe212007-10-18 23:40:06 -0700381EXPORT_SYMBOL(find_task_by_pid_type_ns);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700382
Pavel Emelyanov228ebcb2007-10-18 23:40:16 -0700383struct task_struct *find_task_by_pid(pid_t nr)
384{
385 return find_task_by_pid_type_ns(PIDTYPE_PID, nr, &init_pid_ns);
386}
387EXPORT_SYMBOL(find_task_by_pid);
388
389struct task_struct *find_task_by_vpid(pid_t vnr)
390{
391 return find_task_by_pid_type_ns(PIDTYPE_PID, vnr,
392 current->nsproxy->pid_ns);
393}
394EXPORT_SYMBOL(find_task_by_vpid);
395
396struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns)
397{
398 return find_task_by_pid_type_ns(PIDTYPE_PID, nr, ns);
399}
400EXPORT_SYMBOL(find_task_by_pid_ns);
401
Oleg Nesterov1a657f782006-10-02 02:18:59 -0700402struct pid *get_task_pid(struct task_struct *task, enum pid_type type)
403{
404 struct pid *pid;
405 rcu_read_lock();
406 pid = get_pid(task->pids[type].pid);
407 rcu_read_unlock();
408 return pid;
409}
410
Eric W. Biederman92476d72006-03-31 02:31:42 -0800411struct task_struct *fastcall get_pid_task(struct pid *pid, enum pid_type type)
412{
413 struct task_struct *result;
414 rcu_read_lock();
415 result = pid_task(pid, type);
416 if (result)
417 get_task_struct(result);
418 rcu_read_unlock();
419 return result;
420}
421
422struct pid *find_get_pid(pid_t nr)
423{
424 struct pid *pid;
425
426 rcu_read_lock();
Pavel Emelyanov198fe212007-10-18 23:40:06 -0700427 pid = get_pid(find_vpid(nr));
Eric W. Biederman92476d72006-03-31 02:31:42 -0800428 rcu_read_unlock();
429
430 return pid;
431}
432
Pavel Emelyanov7af57292007-10-18 23:40:06 -0700433pid_t pid_nr_ns(struct pid *pid, struct pid_namespace *ns)
434{
435 struct upid *upid;
436 pid_t nr = 0;
437
438 if (pid && ns->level <= pid->level) {
439 upid = &pid->numbers[ns->level];
440 if (upid->ns == ns)
441 nr = upid->nr;
442 }
443 return nr;
444}
445
Eric W. Biederman44c4e1b2008-02-08 04:19:15 -0800446pid_t pid_vnr(struct pid *pid)
447{
448 return pid_nr_ns(pid, current->nsproxy->pid_ns);
449}
450EXPORT_SYMBOL_GPL(pid_vnr);
451
Pavel Emelyanov2f2a3a42007-10-18 23:40:19 -0700452pid_t task_pid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
453{
454 return pid_nr_ns(task_pid(tsk), ns);
455}
456EXPORT_SYMBOL(task_pid_nr_ns);
457
458pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
459{
460 return pid_nr_ns(task_tgid(tsk), ns);
461}
462EXPORT_SYMBOL(task_tgid_nr_ns);
463
464pid_t task_pgrp_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
465{
466 return pid_nr_ns(task_pgrp(tsk), ns);
467}
468EXPORT_SYMBOL(task_pgrp_nr_ns);
469
470pid_t task_session_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
471{
472 return pid_nr_ns(task_session(tsk), ns);
473}
474EXPORT_SYMBOL(task_session_nr_ns);
475
Linus Torvalds1da177e2005-04-16 15:20:36 -0700476/*
Eric W. Biederman0804ef42006-10-02 02:17:04 -0700477 * Used by proc to find the first pid that is greater then or equal to nr.
478 *
479 * If there is a pid at nr this function is exactly the same as find_pid.
480 */
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}
Eric W. Biedermanbbf73142006-10-02 02:17:11 -0700494EXPORT_SYMBOL_GPL(find_get_pid);
Eric W. Biederman0804ef42006-10-02 02:17:04 -0700495
496/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700497 * The pid hash table is scaled according to the amount of memory in the
498 * machine. From a minimum of 16 slots up to 4096 slots at one gigabyte or
499 * more.
500 */
501void __init pidhash_init(void)
502{
Eric W. Biederman92476d72006-03-31 02:31:42 -0800503 int i, pidhash_size;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700504 unsigned long megabytes = nr_kernel_pages >> (20 - PAGE_SHIFT);
505
506 pidhash_shift = max(4, fls(megabytes * 4));
507 pidhash_shift = min(12, pidhash_shift);
508 pidhash_size = 1 << pidhash_shift;
509
510 printk("PID hash table entries: %d (order: %d, %Zd bytes)\n",
511 pidhash_size, pidhash_shift,
Eric W. Biederman92476d72006-03-31 02:31:42 -0800512 pidhash_size * sizeof(struct hlist_head));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700513
Eric W. Biederman92476d72006-03-31 02:31:42 -0800514 pid_hash = alloc_bootmem(pidhash_size * sizeof(*(pid_hash)));
515 if (!pid_hash)
516 panic("Could not alloc pidhash!\n");
517 for (i = 0; i < pidhash_size; i++)
518 INIT_HLIST_HEAD(&pid_hash[i]);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700519}
520
521void __init pidmap_init(void)
522{
Sukadev Bhattiprolu61a58c62006-12-08 02:37:58 -0800523 init_pid_ns.pidmap[0].page = kzalloc(PAGE_SIZE, GFP_KERNEL);
Oleg Nesterov73b9ebf2006-03-28 16:11:07 -0800524 /* Reserve PID 0. We never call free_pidmap(0) */
Sukadev Bhattiprolu61a58c62006-12-08 02:37:58 -0800525 set_bit(0, init_pid_ns.pidmap[0].page);
526 atomic_dec(&init_pid_ns.pidmap[0].nr_free);
Eric W. Biederman92476d72006-03-31 02:31:42 -0800527
Pavel Emelyanov74bd59b2008-02-08 04:18:24 -0800528 init_pid_ns.pid_cachep = KMEM_CACHE(pid,
529 SLAB_HWCACHE_ALIGN | SLAB_PANIC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700530}