pid.c 12.3 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
/*
 * Generic pidhash and scalable, time-bounded PID allocator
 *
 * (C) 2002-2003 William Irwin, IBM
 * (C) 2004 William Irwin, Oracle
 * (C) 2002-2004 Ingo Molnar, Red Hat
 *
 * pid-structures are backing objects for tasks sharing a given ID to chain
 * against. There is very little to them aside from hashing them and
 * parking tasks using given ID's on a list.
 *
 * The hash is always changed with the tasklist_lock write-acquired,
 * and the hash is only accessed with the tasklist_lock at least
 * read-acquired, so there's no additional SMP locking needed here.
 *
 * We have a list of bitmap pages, which bitmaps represent the PID space.
 * Allocating and freeing PIDs is completely lockless. The worst-case
 * allocation scenario when all but one out of 1 million PIDs possible are
 * allocated already: the scanning of 32 list entries and at most PAGE_SIZE
 * bytes. The typical fastpath is a single successful setbit. Freeing is O(1).
21 22 23 24 25 26
 *
 * Pid namespaces:
 *    (C) 2007 Pavel Emelyanov <xemul@openvz.org>, OpenVZ, SWsoft Inc.
 *    (C) 2007 Sukadev Bhattiprolu <sukadev@us.ibm.com>, IBM
 *     Many thanks to Oleg Nesterov for comments and help
 *
L
Linus Torvalds 已提交
27 28 29 30 31 32
 */

#include <linux/mm.h>
#include <linux/module.h>
#include <linux/slab.h>
#include <linux/init.h>
33
#include <linux/rculist.h>
L
Linus Torvalds 已提交
34 35
#include <linux/bootmem.h>
#include <linux/hash.h>
36
#include <linux/pid_namespace.h>
37
#include <linux/init_task.h>
38
#include <linux/syscalls.h>
L
Linus Torvalds 已提交
39

40 41
#define pid_hashfn(nr, ns)	\
	hash_long((unsigned long)nr + (unsigned long)ns, pidhash_shift)
42
static struct hlist_head *pid_hash;
43
static unsigned int pidhash_shift = 4;
44
struct pid init_struct_pid = INIT_STRUCT_PID;
L
Linus Torvalds 已提交
45 46 47 48 49 50 51 52 53 54

int pid_max = PID_MAX_DEFAULT;

#define RESERVED_PIDS		300

int pid_max_min = RESERVED_PIDS + 1;
int pid_max_max = PID_MAX_LIMIT;

#define BITS_PER_PAGE		(PAGE_SIZE*8)
#define BITS_PER_PAGE_MASK	(BITS_PER_PAGE-1)
S
Sukadev Bhattiprolu 已提交
55

56 57
static inline int mk_pid(struct pid_namespace *pid_ns,
		struct pidmap *map, int off)
S
Sukadev Bhattiprolu 已提交
58
{
59
	return (map - pid_ns->pidmap)*BITS_PER_PAGE + off;
S
Sukadev Bhattiprolu 已提交
60 61
}

L
Linus Torvalds 已提交
62 63 64 65 66 67 68 69 70
#define find_next_offset(map, off)					\
		find_next_zero_bit((map)->page, BITS_PER_PAGE, off)

/*
 * PID-map pages start out as NULL, they get allocated upon
 * first use and are never deallocated. This way a low pid_max
 * value does not cause lots of bitmaps to be allocated, but
 * the scheme scales to up to 4 million PIDs, runtime.
 */
71
struct pid_namespace init_pid_ns = {
C
Cedric Le Goater 已提交
72 73 74
	.kref = {
		.refcount       = ATOMIC_INIT(2),
	},
S
Sukadev Bhattiprolu 已提交
75 76 77
	.pidmap = {
		[ 0 ... PIDMAP_ENTRIES-1] = { ATOMIC_INIT(BITS_PER_PAGE), NULL }
	},
78
	.last_pid = 0,
79 80
	.level = 0,
	.child_reaper = &init_task,
S
Sukadev Bhattiprolu 已提交
81
};
82
EXPORT_SYMBOL_GPL(init_pid_ns);
L
Linus Torvalds 已提交
83

84
int is_container_init(struct task_struct *tsk)
85
{
86 87 88 89 90 91 92 93 94 95
	int ret = 0;
	struct pid *pid;

	rcu_read_lock();
	pid = task_pid(tsk);
	if (pid != NULL && pid->numbers[pid->level].nr == 1)
		ret = 1;
	rcu_read_unlock();

	return ret;
96
}
97
EXPORT_SYMBOL(is_container_init);
98

99 100 101 102 103 104 105 106 107 108 109 110 111
/*
 * Note: disable interrupts while the pidmap_lock is held as an
 * interrupt might come in and do read_lock(&tasklist_lock).
 *
 * If we don't disable interrupts there is a nasty deadlock between
 * detach_pid()->free_pid() and another cpu that does
 * spin_lock(&pidmap_lock) followed by an interrupt routine that does
 * read_lock(&tasklist_lock);
 *
 * After we clean up the tasklist_lock and know there are no
 * irq handlers that take it we can leave the interrupts enabled.
 * For now it is easier to be safe than to prove it can't happen.
 */
S
Sukadev Bhattiprolu 已提交
112

L
Linus Torvalds 已提交
113 114
static  __cacheline_aligned_in_smp DEFINE_SPINLOCK(pidmap_lock);

115
static void free_pidmap(struct upid *upid)
L
Linus Torvalds 已提交
116
{
117 118 119
	int nr = upid->nr;
	struct pidmap *map = upid->ns->pidmap + nr / BITS_PER_PAGE;
	int offset = nr & BITS_PER_PAGE_MASK;
L
Linus Torvalds 已提交
120 121 122 123 124

	clear_bit(offset, map->page);
	atomic_inc(&map->nr_free);
}

125
static int alloc_pidmap(struct pid_namespace *pid_ns)
L
Linus Torvalds 已提交
126
{
127
	int i, offset, max_scan, pid, last = pid_ns->last_pid;
128
	struct pidmap *map;
L
Linus Torvalds 已提交
129 130 131 132 133

	pid = last + 1;
	if (pid >= pid_max)
		pid = RESERVED_PIDS;
	offset = pid & BITS_PER_PAGE_MASK;
134
	map = &pid_ns->pidmap[pid/BITS_PER_PAGE];
L
Linus Torvalds 已提交
135 136 137
	max_scan = (pid_max + BITS_PER_PAGE - 1)/BITS_PER_PAGE - !offset;
	for (i = 0; i <= max_scan; ++i) {
		if (unlikely(!map->page)) {
S
Sukadev Bhattiprolu 已提交
138
			void *page = kzalloc(PAGE_SIZE, GFP_KERNEL);
L
Linus Torvalds 已提交
139 140 141 142
			/*
			 * Free the page if someone raced with us
			 * installing it:
			 */
143
			spin_lock_irq(&pidmap_lock);
144
			if (!map->page) {
S
Sukadev Bhattiprolu 已提交
145
				map->page = page;
146 147
				page = NULL;
			}
148
			spin_unlock_irq(&pidmap_lock);
149
			kfree(page);
L
Linus Torvalds 已提交
150 151 152 153 154 155 156
			if (unlikely(!map->page))
				break;
		}
		if (likely(atomic_read(&map->nr_free))) {
			do {
				if (!test_and_set_bit(offset, map->page)) {
					atomic_dec(&map->nr_free);
157
					pid_ns->last_pid = pid;
L
Linus Torvalds 已提交
158 159 160
					return pid;
				}
				offset = find_next_offset(map, offset);
161
				pid = mk_pid(pid_ns, map, offset);
L
Linus Torvalds 已提交
162 163 164 165 166 167 168 169 170 171
			/*
			 * find_next_offset() found a bit, the pid from it
			 * is in-bounds, and if we fell back to the last
			 * bitmap block and the final block was the same
			 * as the starting point, pid is before last_pid.
			 */
			} while (offset < BITS_PER_PAGE && pid < pid_max &&
					(i != max_scan || pid < last ||
					    !((last+1) & BITS_PER_PAGE_MASK)));
		}
172
		if (map < &pid_ns->pidmap[(pid_max-1)/BITS_PER_PAGE]) {
L
Linus Torvalds 已提交
173 174 175
			++map;
			offset = 0;
		} else {
176
			map = &pid_ns->pidmap[0];
L
Linus Torvalds 已提交
177 178 179 180
			offset = RESERVED_PIDS;
			if (unlikely(last == offset))
				break;
		}
181
		pid = mk_pid(pid_ns, map, offset);
L
Linus Torvalds 已提交
182 183 184 185
	}
	return -1;
}

186
int next_pidmap(struct pid_namespace *pid_ns, int last)
187 188
{
	int offset;
189
	struct pidmap *map, *end;
190 191

	offset = (last + 1) & BITS_PER_PAGE_MASK;
192 193
	map = &pid_ns->pidmap[(last + 1)/BITS_PER_PAGE];
	end = &pid_ns->pidmap[PIDMAP_ENTRIES];
194
	for (; map < end; map++, offset = 0) {
195 196 197 198
		if (unlikely(!map->page))
			continue;
		offset = find_next_bit((map)->page, BITS_PER_PAGE, offset);
		if (offset < BITS_PER_PAGE)
199
			return mk_pid(pid_ns, map, offset);
200 201 202 203
	}
	return -1;
}

204
void put_pid(struct pid *pid)
205
{
206 207
	struct pid_namespace *ns;

208 209
	if (!pid)
		return;
210

211
	ns = pid->numbers[pid->level].ns;
212
	if ((atomic_read(&pid->count) == 1) ||
213
	     atomic_dec_and_test(&pid->count)) {
214
		kmem_cache_free(ns->pid_cachep, pid);
215
		put_pid_ns(ns);
216
	}
217
}
218
EXPORT_SYMBOL_GPL(put_pid);
219 220 221 222 223 224 225

static void delayed_put_pid(struct rcu_head *rhp)
{
	struct pid *pid = container_of(rhp, struct pid, rcu);
	put_pid(pid);
}

226
void free_pid(struct pid *pid)
227 228
{
	/* We can be called with write_lock_irq(&tasklist_lock) held */
229
	int i;
230 231 232
	unsigned long flags;

	spin_lock_irqsave(&pidmap_lock, flags);
233 234
	for (i = 0; i <= pid->level; i++)
		hlist_del_rcu(&pid->numbers[i].pid_chain);
235 236
	spin_unlock_irqrestore(&pidmap_lock, flags);

237
	for (i = 0; i <= pid->level; i++)
238
		free_pidmap(pid->numbers + i);
239

240 241 242
	call_rcu(&pid->rcu, delayed_put_pid);
}

243
struct pid *alloc_pid(struct pid_namespace *ns)
244 245 246
{
	struct pid *pid;
	enum pid_type type;
247 248
	int i, nr;
	struct pid_namespace *tmp;
249
	struct upid *upid;
250

251
	pid = kmem_cache_alloc(ns->pid_cachep, GFP_KERNEL);
252 253 254
	if (!pid)
		goto out;

255 256 257 258 259
	tmp = ns;
	for (i = ns->level; i >= 0; i--) {
		nr = alloc_pidmap(tmp);
		if (nr < 0)
			goto out_free;
260

261 262 263 264 265
		pid->numbers[i].nr = nr;
		pid->numbers[i].ns = tmp;
		tmp = tmp->parent;
	}

266
	get_pid_ns(ns);
267
	pid->level = ns->level;
268 269 270 271 272
	atomic_set(&pid->count, 1);
	for (type = 0; type < PIDTYPE_MAX; ++type)
		INIT_HLIST_HEAD(&pid->tasks[type]);

	spin_lock_irq(&pidmap_lock);
273 274 275 276 277
	for (i = ns->level; i >= 0; i--) {
		upid = &pid->numbers[i];
		hlist_add_head_rcu(&upid->pid_chain,
				&pid_hash[pid_hashfn(upid->nr, upid->ns)]);
	}
278 279 280 281 282 283
	spin_unlock_irq(&pidmap_lock);

out:
	return pid;

out_free:
284 285
	while (++i <= ns->level)
		free_pidmap(pid->numbers + i);
286

287
	kmem_cache_free(ns->pid_cachep, pid);
288 289 290 291
	pid = NULL;
	goto out;
}

292
struct pid *find_pid_ns(int nr, struct pid_namespace *ns)
L
Linus Torvalds 已提交
293 294
{
	struct hlist_node *elem;
295 296 297 298 299 300 301
	struct upid *pnr;

	hlist_for_each_entry_rcu(pnr, elem,
			&pid_hash[pid_hashfn(nr, ns)], pid_chain)
		if (pnr->nr == nr && pnr->ns == ns)
			return container_of(pnr, struct pid,
					numbers[ns->level]);
L
Linus Torvalds 已提交
302 303 304

	return NULL;
}
305
EXPORT_SYMBOL_GPL(find_pid_ns);
L
Linus Torvalds 已提交
306

307 308 309 310 311 312
struct pid *find_vpid(int nr)
{
	return find_pid_ns(nr, current->nsproxy->pid_ns);
}
EXPORT_SYMBOL_GPL(find_vpid);

313 314 315
/*
 * attach_pid() must be called with the tasklist_lock write-held.
 */
316
void attach_pid(struct task_struct *task, enum pid_type type,
317
		struct pid *pid)
L
Linus Torvalds 已提交
318
{
319 320 321
	struct pid_link *link;

	link = &task->pids[type];
322
	link->pid = pid;
323
	hlist_add_head_rcu(&link->node, &pid->tasks[type]);
L
Linus Torvalds 已提交
324 325
}

326 327
static void __change_pid(struct task_struct *task, enum pid_type type,
			struct pid *new)
L
Linus Torvalds 已提交
328
{
329 330 331
	struct pid_link *link;
	struct pid *pid;
	int tmp;
L
Linus Torvalds 已提交
332

333 334
	link = &task->pids[type];
	pid = link->pid;
L
Linus Torvalds 已提交
335

336
	hlist_del_rcu(&link->node);
337
	link->pid = new;
L
Linus Torvalds 已提交
338

339 340 341
	for (tmp = PIDTYPE_MAX; --tmp >= 0; )
		if (!hlist_empty(&pid->tasks[tmp]))
			return;
L
Linus Torvalds 已提交
342

343
	free_pid(pid);
L
Linus Torvalds 已提交
344 345
}

346 347 348 349 350 351 352 353 354 355 356 357
void detach_pid(struct task_struct *task, enum pid_type type)
{
	__change_pid(task, type, NULL);
}

void change_pid(struct task_struct *task, enum pid_type type,
		struct pid *pid)
{
	__change_pid(task, type, pid);
	attach_pid(task, type, pid);
}

358
/* transfer_pid is an optimization of attach_pid(new), detach_pid(old) */
359
void transfer_pid(struct task_struct *old, struct task_struct *new,
360 361 362 363 364 365
			   enum pid_type type)
{
	new->pids[type].pid = old->pids[type].pid;
	hlist_replace_rcu(&old->pids[type].node, &new->pids[type].node);
}

366
struct task_struct *pid_task(struct pid *pid, enum pid_type type)
L
Linus Torvalds 已提交
367
{
368 369 370 371 372 373 374 375 376
	struct task_struct *result = NULL;
	if (pid) {
		struct hlist_node *first;
		first = rcu_dereference(pid->tasks[type].first);
		if (first)
			result = hlist_entry(first, struct task_struct, pids[(type)].node);
	}
	return result;
}
377
EXPORT_SYMBOL(pid_task);
L
Linus Torvalds 已提交
378

379 380 381
/*
 * Must be called under rcu_read_lock() or with tasklist_lock read-held.
 */
382
struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns)
383
{
384
	return pid_task(find_pid_ns(nr, ns), PIDTYPE_PID);
385
}
L
Linus Torvalds 已提交
386

387 388
struct task_struct *find_task_by_vpid(pid_t vnr)
{
389
	return find_task_by_pid_ns(vnr, current->nsproxy->pid_ns);
390 391
}

392 393 394 395
struct pid *get_task_pid(struct task_struct *task, enum pid_type type)
{
	struct pid *pid;
	rcu_read_lock();
396 397
	if (type != PIDTYPE_PID)
		task = task->group_leader;
398 399 400 401 402
	pid = get_pid(task->pids[type].pid);
	rcu_read_unlock();
	return pid;
}

403
struct task_struct *get_pid_task(struct pid *pid, enum pid_type type)
404 405 406 407 408 409 410 411
{
	struct task_struct *result;
	rcu_read_lock();
	result = pid_task(pid, type);
	if (result)
		get_task_struct(result);
	rcu_read_unlock();
	return result;
L
Linus Torvalds 已提交
412 413
}

414
struct pid *find_get_pid(pid_t nr)
L
Linus Torvalds 已提交
415 416 417
{
	struct pid *pid;

418
	rcu_read_lock();
419
	pid = get_pid(find_vpid(nr));
420
	rcu_read_unlock();
L
Linus Torvalds 已提交
421

422
	return pid;
L
Linus Torvalds 已提交
423
}
424
EXPORT_SYMBOL_GPL(find_get_pid);
L
Linus Torvalds 已提交
425

426 427 428 429 430 431 432 433 434 435 436 437 438
pid_t pid_nr_ns(struct pid *pid, struct pid_namespace *ns)
{
	struct upid *upid;
	pid_t nr = 0;

	if (pid && ns->level <= pid->level) {
		upid = &pid->numbers[ns->level];
		if (upid->ns == ns)
			nr = upid->nr;
	}
	return nr;
}

E
Eric W. Biederman 已提交
439 440 441 442 443 444
pid_t pid_vnr(struct pid *pid)
{
	return pid_nr_ns(pid, current->nsproxy->pid_ns);
}
EXPORT_SYMBOL_GPL(pid_vnr);

445 446
pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type,
			struct pid_namespace *ns)
447
{
448 449 450 451 452 453 454 455 456 457 458 459 460
	pid_t nr = 0;

	rcu_read_lock();
	if (!ns)
		ns = current->nsproxy->pid_ns;
	if (likely(pid_alive(task))) {
		if (type != PIDTYPE_PID)
			task = task->group_leader;
		nr = pid_nr_ns(task->pids[type].pid, ns);
	}
	rcu_read_unlock();

	return nr;
461
}
462
EXPORT_SYMBOL(__task_pid_nr_ns);
463 464 465 466 467 468 469

pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns)
{
	return pid_nr_ns(task_tgid(tsk), ns);
}
EXPORT_SYMBOL(task_tgid_nr_ns);

470 471 472 473 474 475
struct pid_namespace *task_active_pid_ns(struct task_struct *tsk)
{
	return ns_of_pid(task_pid(tsk));
}
EXPORT_SYMBOL_GPL(task_active_pid_ns);

476
/*
477
 * Used by proc to find the first pid that is greater than or equal to nr.
478
 *
479
 * If there is a pid at nr this function is exactly the same as find_pid_ns.
480
 */
481
struct pid *find_ge_pid(int nr, struct pid_namespace *ns)
482 483 484 485
{
	struct pid *pid;

	do {
486
		pid = find_pid_ns(nr, ns);
487 488
		if (pid)
			break;
489
		nr = next_pidmap(ns, nr);
490 491 492 493 494
	} while (nr > 0);

	return pid;
}

L
Linus Torvalds 已提交
495 496 497 498 499 500 501
/*
 * The pid hash table is scaled according to the amount of memory in the
 * machine.  From a minimum of 16 slots up to 4096 slots at one gigabyte or
 * more.
 */
void __init pidhash_init(void)
{
502
	int i, pidhash_size;
L
Linus Torvalds 已提交
503

504 505 506
	pid_hash = alloc_large_system_hash("PID", sizeof(*pid_hash), 0, 18,
					   HASH_EARLY | HASH_SMALL,
					   &pidhash_shift, NULL, 4096);
L
Linus Torvalds 已提交
507 508
	pidhash_size = 1 << pidhash_shift;

509 510
	for (i = 0; i < pidhash_size; i++)
		INIT_HLIST_HEAD(&pid_hash[i]);
L
Linus Torvalds 已提交
511 512 513 514
}

void __init pidmap_init(void)
{
515
	init_pid_ns.pidmap[0].page = kzalloc(PAGE_SIZE, GFP_KERNEL);
516
	/* Reserve PID 0. We never call free_pidmap(0) */
517 518
	set_bit(0, init_pid_ns.pidmap[0].page);
	atomic_dec(&init_pid_ns.pidmap[0].nr_free);
519

520 521
	init_pid_ns.pid_cachep = KMEM_CACHE(pid,
			SLAB_HWCACHE_ALIGN | SLAB_PANIC);
L
Linus Torvalds 已提交
522
}