pid.c 11.1 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3
/*
 * Generic pidhash and scalable, time-bounded PID allocator
 *
4 5
 * (C) 2002-2003 Nadia Yvette Chambers, IBM
 * (C) 2004 Nadia Yvette Chambers, Oracle
L
Linus Torvalds 已提交
6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
 * (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
 */

#include <linux/mm.h>
30
#include <linux/export.h>
L
Linus Torvalds 已提交
31 32
#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>
39
#include <linux/proc_ns.h>
40
#include <linux/proc_fs.h>
41
#include <linux/sched/task.h>
42
#include <linux/idr.h>
L
Linus Torvalds 已提交
43

44 45 46 47 48 49 50 51 52 53 54 55 56
struct pid init_struct_pid = {
	.count 		= ATOMIC_INIT(1),
	.tasks		= {
		{ .first = NULL },
		{ .first = NULL },
		{ .first = NULL },
	},
	.level		= 0,
	.numbers	= { {
		.nr		= 0,
		.ns		= &init_pid_ns,
	}, }
};
L
Linus Torvalds 已提交
57 58 59 60 61 62 63 64 65 66 67 68 69 70

int pid_max = PID_MAX_DEFAULT;

#define RESERVED_PIDS		300

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

/*
 * 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 = {
72
	.kref = KREF_INIT(2),
73
	.idr = IDR_INIT(init_pid_ns.idr),
G
Gargi Sharma 已提交
74
	.pid_allocated = PIDNS_ADDING,
75 76
	.level = 0,
	.child_reaper = &init_task,
77
	.user_ns = &init_user_ns,
78
	.ns.inum = PROC_PID_INIT_INO,
79 80 81
#ifdef CONFIG_PID_NS
	.ns.ops = &pidns_operations,
#endif
S
Sukadev Bhattiprolu 已提交
82
};
83
EXPORT_SYMBOL_GPL(init_pid_ns);
L
Linus Torvalds 已提交
84

85 86 87 88 89 90 91 92 93 94 95 96 97
/*
 * 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
98

L
Linus Torvalds 已提交
99 100
static  __cacheline_aligned_in_smp DEFINE_SPINLOCK(pidmap_lock);

101
void put_pid(struct pid *pid)
102
{
103 104
	struct pid_namespace *ns;

105 106
	if (!pid)
		return;
107

108
	ns = pid->numbers[pid->level].ns;
109
	if ((atomic_read(&pid->count) == 1) ||
110
	     atomic_dec_and_test(&pid->count)) {
111
		kmem_cache_free(ns->pid_cachep, pid);
112
		put_pid_ns(ns);
113
	}
114
}
115
EXPORT_SYMBOL_GPL(put_pid);
116 117 118 119 120 121 122

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

123
void free_pid(struct pid *pid)
124 125
{
	/* We can be called with write_lock_irq(&tasklist_lock) held */
126
	int i;
127 128 129
	unsigned long flags;

	spin_lock_irqsave(&pidmap_lock, flags);
130 131
	for (i = 0; i <= pid->level; i++) {
		struct upid *upid = pid->numbers + i;
132
		struct pid_namespace *ns = upid->ns;
G
Gargi Sharma 已提交
133
		switch (--ns->pid_allocated) {
134
		case 2:
135 136 137 138 139 140 141
		case 1:
			/* When all that is left in the pid namespace
			 * is the reaper wake up the reaper.  The reaper
			 * may be sleeping in zap_pid_ns_processes().
			 */
			wake_up_process(ns->child_reaper);
			break;
G
Gargi Sharma 已提交
142
		case PIDNS_ADDING:
143 144
			/* Handle a fork failure of the first process */
			WARN_ON(ns->child_reaper);
G
Gargi Sharma 已提交
145
			ns->pid_allocated = 0;
146
			/* fall through */
147 148 149
		case 0:
			schedule_work(&ns->proc_work);
			break;
150
		}
151 152

		idr_remove(&ns->idr, upid->nr);
153
	}
154 155 156 157 158
	spin_unlock_irqrestore(&pidmap_lock, flags);

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

159
struct pid *alloc_pid(struct pid_namespace *ns)
160 161 162
{
	struct pid *pid;
	enum pid_type type;
163 164
	int i, nr;
	struct pid_namespace *tmp;
165
	struct upid *upid;
166
	int retval = -ENOMEM;
167

168
	pid = kmem_cache_alloc(ns->pid_cachep, GFP_KERNEL);
169
	if (!pid)
170
		return ERR_PTR(retval);
171

172
	tmp = ns;
173
	pid->level = ns->level;
174

175
	for (i = ns->level; i >= 0; i--) {
176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196
		int pid_min = 1;

		idr_preload(GFP_KERNEL);
		spin_lock_irq(&pidmap_lock);

		/*
		 * init really needs pid 1, but after reaching the maximum
		 * wrap back to RESERVED_PIDS
		 */
		if (idr_get_cursor(&tmp->idr) > RESERVED_PIDS)
			pid_min = RESERVED_PIDS;

		/*
		 * Store a null pointer so find_pid_ns does not find
		 * a partially initialized PID (see below).
		 */
		nr = idr_alloc_cyclic(&tmp->idr, NULL, pid_min,
				      pid_max, GFP_ATOMIC);
		spin_unlock_irq(&pidmap_lock);
		idr_preload_end();

197
		if (nr < 0) {
198
			retval = nr;
199
			goto out_free;
200
		}
201

202 203 204 205 206
		pid->numbers[i].nr = nr;
		pid->numbers[i].ns = tmp;
		tmp = tmp->parent;
	}

207
	if (unlikely(is_child_reaper(pid))) {
208
		if (pid_ns_prepare_proc(ns))
209 210 211
			goto out_free;
	}

212
	get_pid_ns(ns);
213 214 215 216
	atomic_set(&pid->count, 1);
	for (type = 0; type < PIDTYPE_MAX; ++type)
		INIT_HLIST_HEAD(&pid->tasks[type]);

217
	upid = pid->numbers + ns->level;
218
	spin_lock_irq(&pidmap_lock);
G
Gargi Sharma 已提交
219
	if (!(ns->pid_allocated & PIDNS_ADDING))
220
		goto out_unlock;
221
	for ( ; upid >= pid->numbers; --upid) {
222 223
		/* Make the PID visible to find_pid_ns. */
		idr_replace(&upid->ns->idr, pid, upid->nr);
G
Gargi Sharma 已提交
224
		upid->ns->pid_allocated++;
225
	}
226 227 228 229
	spin_unlock_irq(&pidmap_lock);

	return pid;

230
out_unlock:
231
	spin_unlock_irq(&pidmap_lock);
232 233
	put_pid_ns(ns);

234
out_free:
235
	spin_lock_irq(&pidmap_lock);
236
	while (++i <= ns->level)
237 238
		idr_remove(&ns->idr, (pid->numbers + i)->nr);

239 240 241 242
	/* On failure to allocate the first pid, reset the state */
	if (ns->pid_allocated == PIDNS_ADDING)
		idr_set_cursor(&ns->idr, 0);

243
	spin_unlock_irq(&pidmap_lock);
244

245
	kmem_cache_free(ns->pid_cachep, pid);
246
	return ERR_PTR(retval);
247 248
}

249 250 251
void disable_pid_allocation(struct pid_namespace *ns)
{
	spin_lock_irq(&pidmap_lock);
G
Gargi Sharma 已提交
252
	ns->pid_allocated &= ~PIDNS_ADDING;
253 254 255
	spin_unlock_irq(&pidmap_lock);
}

256
struct pid *find_pid_ns(int nr, struct pid_namespace *ns)
L
Linus Torvalds 已提交
257
{
G
Gargi Sharma 已提交
258
	return idr_find(&ns->idr, nr);
L
Linus Torvalds 已提交
259
}
260
EXPORT_SYMBOL_GPL(find_pid_ns);
L
Linus Torvalds 已提交
261

262 263
struct pid *find_vpid(int nr)
{
264
	return find_pid_ns(nr, task_active_pid_ns(current));
265 266 267
}
EXPORT_SYMBOL_GPL(find_vpid);

268 269 270
/*
 * attach_pid() must be called with the tasklist_lock write-held.
 */
271
void attach_pid(struct task_struct *task, enum pid_type type)
L
Linus Torvalds 已提交
272
{
273 274
	struct pid_link *link = &task->pids[type];
	hlist_add_head_rcu(&link->node, &link->pid->tasks[type]);
L
Linus Torvalds 已提交
275 276
}

277 278
static void __change_pid(struct task_struct *task, enum pid_type type,
			struct pid *new)
L
Linus Torvalds 已提交
279
{
280 281 282
	struct pid_link *link;
	struct pid *pid;
	int tmp;
L
Linus Torvalds 已提交
283

284 285
	link = &task->pids[type];
	pid = link->pid;
L
Linus Torvalds 已提交
286

287
	hlist_del_rcu(&link->node);
288
	link->pid = new;
L
Linus Torvalds 已提交
289

290 291 292
	for (tmp = PIDTYPE_MAX; --tmp >= 0; )
		if (!hlist_empty(&pid->tasks[tmp]))
			return;
L
Linus Torvalds 已提交
293

294
	free_pid(pid);
L
Linus Torvalds 已提交
295 296
}

297 298 299 300 301 302 303 304 305
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);
306
	attach_pid(task, type);
307 308
}

309
/* transfer_pid is an optimization of attach_pid(new), detach_pid(old) */
310
void transfer_pid(struct task_struct *old, struct task_struct *new,
311 312 313 314 315 316
			   enum pid_type type)
{
	new->pids[type].pid = old->pids[type].pid;
	hlist_replace_rcu(&old->pids[type].node, &new->pids[type].node);
}

317
struct task_struct *pid_task(struct pid *pid, enum pid_type type)
L
Linus Torvalds 已提交
318
{
319 320 321
	struct task_struct *result = NULL;
	if (pid) {
		struct hlist_node *first;
A
Arnd Bergmann 已提交
322
		first = rcu_dereference_check(hlist_first_rcu(&pid->tasks[type]),
323
					      lockdep_tasklist_lock_is_held());
324 325 326 327 328
		if (first)
			result = hlist_entry(first, struct task_struct, pids[(type)].node);
	}
	return result;
}
329
EXPORT_SYMBOL(pid_task);
L
Linus Torvalds 已提交
330

331
/*
332
 * Must be called under rcu_read_lock().
333
 */
334
struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns)
335
{
336 337
	RCU_LOCKDEP_WARN(!rcu_read_lock_held(),
			 "find_task_by_pid_ns() needs rcu_read_lock() protection");
338
	return pid_task(find_pid_ns(nr, ns), PIDTYPE_PID);
339
}
L
Linus Torvalds 已提交
340

341 342
struct task_struct *find_task_by_vpid(pid_t vnr)
{
343
	return find_task_by_pid_ns(vnr, task_active_pid_ns(current));
344 345
}

346 347 348 349 350 351 352 353 354 355 356 357 358
struct task_struct *find_get_task_by_vpid(pid_t nr)
{
	struct task_struct *task;

	rcu_read_lock();
	task = find_task_by_vpid(nr);
	if (task)
		get_task_struct(task);
	rcu_read_unlock();

	return task;
}

359 360 361 362
struct pid *get_task_pid(struct task_struct *task, enum pid_type type)
{
	struct pid *pid;
	rcu_read_lock();
363 364
	if (type != PIDTYPE_PID)
		task = task->group_leader;
365
	pid = get_pid(rcu_dereference(task->pids[type].pid));
366 367 368
	rcu_read_unlock();
	return pid;
}
369
EXPORT_SYMBOL_GPL(get_task_pid);
370

371
struct task_struct *get_pid_task(struct pid *pid, enum pid_type type)
372 373 374 375 376 377 378 379
{
	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
380
}
381
EXPORT_SYMBOL_GPL(get_pid_task);
L
Linus Torvalds 已提交
382

383
struct pid *find_get_pid(pid_t nr)
L
Linus Torvalds 已提交
384 385 386
{
	struct pid *pid;

387
	rcu_read_lock();
388
	pid = get_pid(find_vpid(nr));
389
	rcu_read_unlock();
L
Linus Torvalds 已提交
390

391
	return pid;
L
Linus Torvalds 已提交
392
}
393
EXPORT_SYMBOL_GPL(find_get_pid);
L
Linus Torvalds 已提交
394

395 396 397 398 399 400 401 402 403 404 405 406
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;
}
407
EXPORT_SYMBOL_GPL(pid_nr_ns);
408

E
Eric W. Biederman 已提交
409 410
pid_t pid_vnr(struct pid *pid)
{
411
	return pid_nr_ns(pid, task_active_pid_ns(current));
E
Eric W. Biederman 已提交
412 413 414
}
EXPORT_SYMBOL_GPL(pid_vnr);

415 416
pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type,
			struct pid_namespace *ns)
417
{
418 419 420 421
	pid_t nr = 0;

	rcu_read_lock();
	if (!ns)
422
		ns = task_active_pid_ns(current);
423
	if (likely(pid_alive(task))) {
424 425 426
		if (type != PIDTYPE_PID) {
			if (type == __PIDTYPE_TGID)
				type = PIDTYPE_PID;
G
Gargi Sharma 已提交
427

428
			task = task->group_leader;
429
		}
430
		nr = pid_nr_ns(rcu_dereference(task->pids[type].pid), ns);
431 432 433 434
	}
	rcu_read_unlock();

	return nr;
435
}
436
EXPORT_SYMBOL(__task_pid_nr_ns);
437

438 439 440 441 442 443
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);

444
/*
445
 * Used by proc to find the first pid that is greater than or equal to nr.
446
 *
447
 * If there is a pid at nr this function is exactly the same as find_pid_ns.
448
 */
449
struct pid *find_ge_pid(int nr, struct pid_namespace *ns)
450
{
451
	return idr_get_next(&ns->idr, &nr);
452 453
}

454
void __init pid_idr_init(void)
L
Linus Torvalds 已提交
455
{
Z
Zhen Lei 已提交
456
	/* Verify no one has done anything silly: */
G
Gargi Sharma 已提交
457
	BUILD_BUG_ON(PID_MAX_LIMIT >= PIDNS_ADDING);
458

459 460 461 462 463 464 465
	/* bump default and minimum pid_max based on number of cpus */
	pid_max = min(pid_max_max, max_t(int, pid_max,
				PIDS_PER_CPU_DEFAULT * num_possible_cpus()));
	pid_max_min = max_t(int, pid_max_min,
				PIDS_PER_CPU_MIN * num_possible_cpus());
	pr_info("pid_max: default: %u minimum: %u\n", pid_max, pid_max_min);

466
	idr_init(&init_pid_ns.idr);
467

468
	init_pid_ns.pid_cachep = KMEM_CACHE(pid,
469
			SLAB_HWCACHE_ALIGN | SLAB_PANIC | SLAB_ACCOUNT);
L
Linus Torvalds 已提交
470
}