]>
Commit | Line | Data |
---|---|---|
457c8996 | 1 | // SPDX-License-Identifier: GPL-2.0-only |
1da177e4 LT |
2 | /* |
3 | * Generic pidhash and scalable, time-bounded PID allocator | |
4 | * | |
6d49e352 NYC |
5 | * (C) 2002-2003 Nadia Yvette Chambers, IBM |
6 | * (C) 2004 Nadia Yvette Chambers, Oracle | |
1da177e4 LT |
7 | * (C) 2002-2004 Ingo Molnar, Red Hat |
8 | * | |
9 | * pid-structures are backing objects for tasks sharing a given ID to chain | |
10 | * against. There is very little to them aside from hashing them and | |
11 | * parking tasks using given ID's on a list. | |
12 | * | |
13 | * The hash is always changed with the tasklist_lock write-acquired, | |
14 | * and the hash is only accessed with the tasklist_lock at least | |
15 | * read-acquired, so there's no additional SMP locking needed here. | |
16 | * | |
17 | * We have a list of bitmap pages, which bitmaps represent the PID space. | |
18 | * Allocating and freeing PIDs is completely lockless. The worst-case | |
19 | * allocation scenario when all but one out of 1 million PIDs possible are | |
20 | * allocated already: the scanning of 32 list entries and at most PAGE_SIZE | |
21 | * bytes. The typical fastpath is a single successful setbit. Freeing is O(1). | |
30e49c26 PE |
22 | * |
23 | * Pid namespaces: | |
24 | * (C) 2007 Pavel Emelyanov <xemul@openvz.org>, OpenVZ, SWsoft Inc. | |
25 | * (C) 2007 Sukadev Bhattiprolu <sukadev@us.ibm.com>, IBM | |
26 | * Many thanks to Oleg Nesterov for comments and help | |
27 | * | |
1da177e4 LT |
28 | */ |
29 | ||
30 | #include <linux/mm.h> | |
9984de1a | 31 | #include <linux/export.h> |
1da177e4 LT |
32 | #include <linux/slab.h> |
33 | #include <linux/init.h> | |
82524746 | 34 | #include <linux/rculist.h> |
57c8a661 | 35 | #include <linux/memblock.h> |
61a58c6c | 36 | #include <linux/pid_namespace.h> |
820e45db | 37 | #include <linux/init_task.h> |
3eb07c8c | 38 | #include <linux/syscalls.h> |
0bb80f24 | 39 | #include <linux/proc_ns.h> |
0a01f2cc | 40 | #include <linux/proc_fs.h> |
32fcb426 CB |
41 | #include <linux/anon_inodes.h> |
42 | #include <linux/sched/signal.h> | |
29930025 | 43 | #include <linux/sched/task.h> |
95846ecf | 44 | #include <linux/idr.h> |
1da177e4 | 45 | |
e1e871af DH |
46 | struct pid init_struct_pid = { |
47 | .count = ATOMIC_INIT(1), | |
48 | .tasks = { | |
49 | { .first = NULL }, | |
50 | { .first = NULL }, | |
51 | { .first = NULL }, | |
52 | }, | |
53 | .level = 0, | |
54 | .numbers = { { | |
55 | .nr = 0, | |
56 | .ns = &init_pid_ns, | |
57 | }, } | |
58 | }; | |
1da177e4 LT |
59 | |
60 | int pid_max = PID_MAX_DEFAULT; | |
1da177e4 LT |
61 | |
62 | #define RESERVED_PIDS 300 | |
63 | ||
64 | int pid_max_min = RESERVED_PIDS + 1; | |
65 | int pid_max_max = PID_MAX_LIMIT; | |
66 | ||
1da177e4 LT |
67 | /* |
68 | * PID-map pages start out as NULL, they get allocated upon | |
69 | * first use and are never deallocated. This way a low pid_max | |
70 | * value does not cause lots of bitmaps to be allocated, but | |
71 | * the scheme scales to up to 4 million PIDs, runtime. | |
72 | */ | |
61a58c6c | 73 | struct pid_namespace init_pid_ns = { |
1e24edca | 74 | .kref = KREF_INIT(2), |
f6bb2a2c | 75 | .idr = IDR_INIT(init_pid_ns.idr), |
e8cfbc24 | 76 | .pid_allocated = PIDNS_ADDING, |
faacbfd3 PE |
77 | .level = 0, |
78 | .child_reaper = &init_task, | |
49f4d8b9 | 79 | .user_ns = &init_user_ns, |
435d5f4b | 80 | .ns.inum = PROC_PID_INIT_INO, |
33c42940 AV |
81 | #ifdef CONFIG_PID_NS |
82 | .ns.ops = &pidns_operations, | |
83 | #endif | |
3fbc9648 | 84 | }; |
198fe21b | 85 | EXPORT_SYMBOL_GPL(init_pid_ns); |
1da177e4 | 86 | |
92476d7f EB |
87 | /* |
88 | * Note: disable interrupts while the pidmap_lock is held as an | |
89 | * interrupt might come in and do read_lock(&tasklist_lock). | |
90 | * | |
91 | * If we don't disable interrupts there is a nasty deadlock between | |
92 | * detach_pid()->free_pid() and another cpu that does | |
93 | * spin_lock(&pidmap_lock) followed by an interrupt routine that does | |
94 | * read_lock(&tasklist_lock); | |
95 | * | |
96 | * After we clean up the tasklist_lock and know there are no | |
97 | * irq handlers that take it we can leave the interrupts enabled. | |
98 | * For now it is easier to be safe than to prove it can't happen. | |
99 | */ | |
3fbc9648 | 100 | |
1da177e4 LT |
101 | static __cacheline_aligned_in_smp DEFINE_SPINLOCK(pidmap_lock); |
102 | ||
7ad5b3a5 | 103 | void put_pid(struct pid *pid) |
92476d7f | 104 | { |
baf8f0f8 PE |
105 | struct pid_namespace *ns; |
106 | ||
92476d7f EB |
107 | if (!pid) |
108 | return; | |
baf8f0f8 | 109 | |
8ef047aa | 110 | ns = pid->numbers[pid->level].ns; |
92476d7f | 111 | if ((atomic_read(&pid->count) == 1) || |
8ef047aa | 112 | atomic_dec_and_test(&pid->count)) { |
baf8f0f8 | 113 | kmem_cache_free(ns->pid_cachep, pid); |
b461cc03 | 114 | put_pid_ns(ns); |
8ef047aa | 115 | } |
92476d7f | 116 | } |
bbf73147 | 117 | EXPORT_SYMBOL_GPL(put_pid); |
92476d7f EB |
118 | |
119 | static void delayed_put_pid(struct rcu_head *rhp) | |
120 | { | |
121 | struct pid *pid = container_of(rhp, struct pid, rcu); | |
122 | put_pid(pid); | |
123 | } | |
124 | ||
7ad5b3a5 | 125 | void free_pid(struct pid *pid) |
92476d7f EB |
126 | { |
127 | /* We can be called with write_lock_irq(&tasklist_lock) held */ | |
8ef047aa | 128 | int i; |
92476d7f EB |
129 | unsigned long flags; |
130 | ||
131 | spin_lock_irqsave(&pidmap_lock, flags); | |
0a01f2cc EB |
132 | for (i = 0; i <= pid->level; i++) { |
133 | struct upid *upid = pid->numbers + i; | |
af4b8a83 | 134 | struct pid_namespace *ns = upid->ns; |
e8cfbc24 | 135 | switch (--ns->pid_allocated) { |
a6064885 | 136 | case 2: |
af4b8a83 EB |
137 | case 1: |
138 | /* When all that is left in the pid namespace | |
139 | * is the reaper wake up the reaper. The reaper | |
140 | * may be sleeping in zap_pid_ns_processes(). | |
141 | */ | |
142 | wake_up_process(ns->child_reaper); | |
143 | break; | |
e8cfbc24 | 144 | case PIDNS_ADDING: |
314a8ad0 ON |
145 | /* Handle a fork failure of the first process */ |
146 | WARN_ON(ns->child_reaper); | |
e8cfbc24 | 147 | ns->pid_allocated = 0; |
314a8ad0 | 148 | /* fall through */ |
af4b8a83 | 149 | case 0: |
af4b8a83 EB |
150 | schedule_work(&ns->proc_work); |
151 | break; | |
5e1182de | 152 | } |
95846ecf GS |
153 | |
154 | idr_remove(&ns->idr, upid->nr); | |
0a01f2cc | 155 | } |
92476d7f EB |
156 | spin_unlock_irqrestore(&pidmap_lock, flags); |
157 | ||
92476d7f EB |
158 | call_rcu(&pid->rcu, delayed_put_pid); |
159 | } | |
160 | ||
8ef047aa | 161 | struct pid *alloc_pid(struct pid_namespace *ns) |
92476d7f EB |
162 | { |
163 | struct pid *pid; | |
164 | enum pid_type type; | |
8ef047aa PE |
165 | int i, nr; |
166 | struct pid_namespace *tmp; | |
198fe21b | 167 | struct upid *upid; |
35f71bc0 | 168 | int retval = -ENOMEM; |
92476d7f | 169 | |
baf8f0f8 | 170 | pid = kmem_cache_alloc(ns->pid_cachep, GFP_KERNEL); |
92476d7f | 171 | if (!pid) |
35f71bc0 | 172 | return ERR_PTR(retval); |
92476d7f | 173 | |
8ef047aa | 174 | tmp = ns; |
0a01f2cc | 175 | pid->level = ns->level; |
95846ecf | 176 | |
8ef047aa | 177 | for (i = ns->level; i >= 0; i--) { |
95846ecf GS |
178 | int pid_min = 1; |
179 | ||
180 | idr_preload(GFP_KERNEL); | |
181 | spin_lock_irq(&pidmap_lock); | |
182 | ||
183 | /* | |
184 | * init really needs pid 1, but after reaching the maximum | |
185 | * wrap back to RESERVED_PIDS | |
186 | */ | |
187 | if (idr_get_cursor(&tmp->idr) > RESERVED_PIDS) | |
188 | pid_min = RESERVED_PIDS; | |
189 | ||
190 | /* | |
191 | * Store a null pointer so find_pid_ns does not find | |
192 | * a partially initialized PID (see below). | |
193 | */ | |
194 | nr = idr_alloc_cyclic(&tmp->idr, NULL, pid_min, | |
195 | pid_max, GFP_ATOMIC); | |
196 | spin_unlock_irq(&pidmap_lock); | |
197 | idr_preload_end(); | |
198 | ||
287980e4 | 199 | if (nr < 0) { |
f83606f5 | 200 | retval = (nr == -ENOSPC) ? -EAGAIN : nr; |
8ef047aa | 201 | goto out_free; |
35f71bc0 | 202 | } |
92476d7f | 203 | |
8ef047aa PE |
204 | pid->numbers[i].nr = nr; |
205 | pid->numbers[i].ns = tmp; | |
206 | tmp = tmp->parent; | |
207 | } | |
208 | ||
0a01f2cc | 209 | if (unlikely(is_child_reaper(pid))) { |
c0ee5549 | 210 | if (pid_ns_prepare_proc(ns)) |
0a01f2cc EB |
211 | goto out_free; |
212 | } | |
213 | ||
b461cc03 | 214 | get_pid_ns(ns); |
92476d7f | 215 | atomic_set(&pid->count, 1); |
92476d7f EB |
216 | for (type = 0; type < PIDTYPE_MAX; ++type) |
217 | INIT_HLIST_HEAD(&pid->tasks[type]); | |
218 | ||
b53b0b9d JFG |
219 | init_waitqueue_head(&pid->wait_pidfd); |
220 | ||
417e3152 | 221 | upid = pid->numbers + ns->level; |
92476d7f | 222 | spin_lock_irq(&pidmap_lock); |
e8cfbc24 | 223 | if (!(ns->pid_allocated & PIDNS_ADDING)) |
5e1182de | 224 | goto out_unlock; |
0a01f2cc | 225 | for ( ; upid >= pid->numbers; --upid) { |
95846ecf GS |
226 | /* Make the PID visible to find_pid_ns. */ |
227 | idr_replace(&upid->ns->idr, pid, upid->nr); | |
e8cfbc24 | 228 | upid->ns->pid_allocated++; |
0a01f2cc | 229 | } |
92476d7f EB |
230 | spin_unlock_irq(&pidmap_lock); |
231 | ||
92476d7f EB |
232 | return pid; |
233 | ||
5e1182de | 234 | out_unlock: |
6e666884 | 235 | spin_unlock_irq(&pidmap_lock); |
24c037eb ON |
236 | put_pid_ns(ns); |
237 | ||
92476d7f | 238 | out_free: |
95846ecf | 239 | spin_lock_irq(&pidmap_lock); |
1a80dade MW |
240 | while (++i <= ns->level) { |
241 | upid = pid->numbers + i; | |
242 | idr_remove(&upid->ns->idr, upid->nr); | |
243 | } | |
95846ecf | 244 | |
c0ee5549 EB |
245 | /* On failure to allocate the first pid, reset the state */ |
246 | if (ns->pid_allocated == PIDNS_ADDING) | |
247 | idr_set_cursor(&ns->idr, 0); | |
248 | ||
95846ecf | 249 | spin_unlock_irq(&pidmap_lock); |
8ef047aa | 250 | |
baf8f0f8 | 251 | kmem_cache_free(ns->pid_cachep, pid); |
35f71bc0 | 252 | return ERR_PTR(retval); |
92476d7f EB |
253 | } |
254 | ||
c876ad76 EB |
255 | void disable_pid_allocation(struct pid_namespace *ns) |
256 | { | |
257 | spin_lock_irq(&pidmap_lock); | |
e8cfbc24 | 258 | ns->pid_allocated &= ~PIDNS_ADDING; |
c876ad76 EB |
259 | spin_unlock_irq(&pidmap_lock); |
260 | } | |
261 | ||
7ad5b3a5 | 262 | struct pid *find_pid_ns(int nr, struct pid_namespace *ns) |
1da177e4 | 263 | { |
e8cfbc24 | 264 | return idr_find(&ns->idr, nr); |
1da177e4 | 265 | } |
198fe21b | 266 | EXPORT_SYMBOL_GPL(find_pid_ns); |
1da177e4 | 267 | |
8990571e PE |
268 | struct pid *find_vpid(int nr) |
269 | { | |
17cf22c3 | 270 | return find_pid_ns(nr, task_active_pid_ns(current)); |
8990571e PE |
271 | } |
272 | EXPORT_SYMBOL_GPL(find_vpid); | |
273 | ||
2c470475 EB |
274 | static struct pid **task_pid_ptr(struct task_struct *task, enum pid_type type) |
275 | { | |
276 | return (type == PIDTYPE_PID) ? | |
277 | &task->thread_pid : | |
2c470475 EB |
278 | &task->signal->pids[type]; |
279 | } | |
280 | ||
e713d0da SB |
281 | /* |
282 | * attach_pid() must be called with the tasklist_lock write-held. | |
283 | */ | |
81907739 | 284 | void attach_pid(struct task_struct *task, enum pid_type type) |
1da177e4 | 285 | { |
2c470475 EB |
286 | struct pid *pid = *task_pid_ptr(task, type); |
287 | hlist_add_head_rcu(&task->pid_links[type], &pid->tasks[type]); | |
1da177e4 LT |
288 | } |
289 | ||
24336eae ON |
290 | static void __change_pid(struct task_struct *task, enum pid_type type, |
291 | struct pid *new) | |
1da177e4 | 292 | { |
2c470475 | 293 | struct pid **pid_ptr = task_pid_ptr(task, type); |
92476d7f EB |
294 | struct pid *pid; |
295 | int tmp; | |
1da177e4 | 296 | |
2c470475 | 297 | pid = *pid_ptr; |
1da177e4 | 298 | |
2c470475 EB |
299 | hlist_del_rcu(&task->pid_links[type]); |
300 | *pid_ptr = new; | |
1da177e4 | 301 | |
92476d7f EB |
302 | for (tmp = PIDTYPE_MAX; --tmp >= 0; ) |
303 | if (!hlist_empty(&pid->tasks[tmp])) | |
304 | return; | |
1da177e4 | 305 | |
92476d7f | 306 | free_pid(pid); |
1da177e4 LT |
307 | } |
308 | ||
24336eae ON |
309 | void detach_pid(struct task_struct *task, enum pid_type type) |
310 | { | |
311 | __change_pid(task, type, NULL); | |
312 | } | |
313 | ||
314 | void change_pid(struct task_struct *task, enum pid_type type, | |
315 | struct pid *pid) | |
316 | { | |
317 | __change_pid(task, type, pid); | |
81907739 | 318 | attach_pid(task, type); |
24336eae ON |
319 | } |
320 | ||
c18258c6 | 321 | /* transfer_pid is an optimization of attach_pid(new), detach_pid(old) */ |
7ad5b3a5 | 322 | void transfer_pid(struct task_struct *old, struct task_struct *new, |
c18258c6 EB |
323 | enum pid_type type) |
324 | { | |
2c470475 EB |
325 | if (type == PIDTYPE_PID) |
326 | new->thread_pid = old->thread_pid; | |
327 | hlist_replace_rcu(&old->pid_links[type], &new->pid_links[type]); | |
c18258c6 EB |
328 | } |
329 | ||
7ad5b3a5 | 330 | struct task_struct *pid_task(struct pid *pid, enum pid_type type) |
1da177e4 | 331 | { |
92476d7f EB |
332 | struct task_struct *result = NULL; |
333 | if (pid) { | |
334 | struct hlist_node *first; | |
67bdbffd | 335 | first = rcu_dereference_check(hlist_first_rcu(&pid->tasks[type]), |
db1466b3 | 336 | lockdep_tasklist_lock_is_held()); |
92476d7f | 337 | if (first) |
2c470475 | 338 | result = hlist_entry(first, struct task_struct, pid_links[(type)]); |
92476d7f EB |
339 | } |
340 | return result; | |
341 | } | |
eccba068 | 342 | EXPORT_SYMBOL(pid_task); |
1da177e4 | 343 | |
92476d7f | 344 | /* |
9728e5d6 | 345 | * Must be called under rcu_read_lock(). |
92476d7f | 346 | */ |
17f98dcf | 347 | struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns) |
92476d7f | 348 | { |
f78f5b90 PM |
349 | RCU_LOCKDEP_WARN(!rcu_read_lock_held(), |
350 | "find_task_by_pid_ns() needs rcu_read_lock() protection"); | |
17f98dcf | 351 | return pid_task(find_pid_ns(nr, ns), PIDTYPE_PID); |
92476d7f | 352 | } |
1da177e4 | 353 | |
228ebcbe PE |
354 | struct task_struct *find_task_by_vpid(pid_t vnr) |
355 | { | |
17cf22c3 | 356 | return find_task_by_pid_ns(vnr, task_active_pid_ns(current)); |
228ebcbe | 357 | } |
228ebcbe | 358 | |
2ee08260 MR |
359 | struct task_struct *find_get_task_by_vpid(pid_t nr) |
360 | { | |
361 | struct task_struct *task; | |
362 | ||
363 | rcu_read_lock(); | |
364 | task = find_task_by_vpid(nr); | |
365 | if (task) | |
366 | get_task_struct(task); | |
367 | rcu_read_unlock(); | |
368 | ||
369 | return task; | |
370 | } | |
371 | ||
1a657f78 ON |
372 | struct pid *get_task_pid(struct task_struct *task, enum pid_type type) |
373 | { | |
374 | struct pid *pid; | |
375 | rcu_read_lock(); | |
2c470475 | 376 | pid = get_pid(rcu_dereference(*task_pid_ptr(task, type))); |
1a657f78 ON |
377 | rcu_read_unlock(); |
378 | return pid; | |
379 | } | |
77c100c8 | 380 | EXPORT_SYMBOL_GPL(get_task_pid); |
1a657f78 | 381 | |
7ad5b3a5 | 382 | struct task_struct *get_pid_task(struct pid *pid, enum pid_type type) |
92476d7f EB |
383 | { |
384 | struct task_struct *result; | |
385 | rcu_read_lock(); | |
386 | result = pid_task(pid, type); | |
387 | if (result) | |
388 | get_task_struct(result); | |
389 | rcu_read_unlock(); | |
390 | return result; | |
1da177e4 | 391 | } |
77c100c8 | 392 | EXPORT_SYMBOL_GPL(get_pid_task); |
1da177e4 | 393 | |
92476d7f | 394 | struct pid *find_get_pid(pid_t nr) |
1da177e4 LT |
395 | { |
396 | struct pid *pid; | |
397 | ||
92476d7f | 398 | rcu_read_lock(); |
198fe21b | 399 | pid = get_pid(find_vpid(nr)); |
92476d7f | 400 | rcu_read_unlock(); |
1da177e4 | 401 | |
92476d7f | 402 | return pid; |
1da177e4 | 403 | } |
339caf2a | 404 | EXPORT_SYMBOL_GPL(find_get_pid); |
1da177e4 | 405 | |
7af57294 PE |
406 | pid_t pid_nr_ns(struct pid *pid, struct pid_namespace *ns) |
407 | { | |
408 | struct upid *upid; | |
409 | pid_t nr = 0; | |
410 | ||
411 | if (pid && ns->level <= pid->level) { | |
412 | upid = &pid->numbers[ns->level]; | |
413 | if (upid->ns == ns) | |
414 | nr = upid->nr; | |
415 | } | |
416 | return nr; | |
417 | } | |
4f82f457 | 418 | EXPORT_SYMBOL_GPL(pid_nr_ns); |
7af57294 | 419 | |
44c4e1b2 EB |
420 | pid_t pid_vnr(struct pid *pid) |
421 | { | |
17cf22c3 | 422 | return pid_nr_ns(pid, task_active_pid_ns(current)); |
44c4e1b2 EB |
423 | } |
424 | EXPORT_SYMBOL_GPL(pid_vnr); | |
425 | ||
52ee2dfd ON |
426 | pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type, |
427 | struct pid_namespace *ns) | |
2f2a3a46 | 428 | { |
52ee2dfd ON |
429 | pid_t nr = 0; |
430 | ||
431 | rcu_read_lock(); | |
432 | if (!ns) | |
17cf22c3 | 433 | ns = task_active_pid_ns(current); |
2c470475 EB |
434 | if (likely(pid_alive(task))) |
435 | nr = pid_nr_ns(rcu_dereference(*task_pid_ptr(task, type)), ns); | |
52ee2dfd ON |
436 | rcu_read_unlock(); |
437 | ||
438 | return nr; | |
2f2a3a46 | 439 | } |
52ee2dfd | 440 | EXPORT_SYMBOL(__task_pid_nr_ns); |
2f2a3a46 | 441 | |
61bce0f1 EB |
442 | struct pid_namespace *task_active_pid_ns(struct task_struct *tsk) |
443 | { | |
444 | return ns_of_pid(task_pid(tsk)); | |
445 | } | |
446 | EXPORT_SYMBOL_GPL(task_active_pid_ns); | |
447 | ||
0804ef4b | 448 | /* |
025dfdaf | 449 | * Used by proc to find the first pid that is greater than or equal to nr. |
0804ef4b | 450 | * |
e49859e7 | 451 | * If there is a pid at nr this function is exactly the same as find_pid_ns. |
0804ef4b | 452 | */ |
198fe21b | 453 | struct pid *find_ge_pid(int nr, struct pid_namespace *ns) |
0804ef4b | 454 | { |
95846ecf | 455 | return idr_get_next(&ns->idr, &nr); |
0804ef4b EB |
456 | } |
457 | ||
32fcb426 CB |
458 | /** |
459 | * pidfd_create() - Create a new pid file descriptor. | |
460 | * | |
461 | * @pid: struct pid that the pidfd will reference | |
462 | * | |
463 | * This creates a new pid file descriptor with the O_CLOEXEC flag set. | |
464 | * | |
465 | * Note, that this function can only be called after the fd table has | |
466 | * been unshared to avoid leaking the pidfd to the new process. | |
467 | * | |
468 | * Return: On success, a cloexec pidfd is returned. | |
469 | * On error, a negative errno number will be returned. | |
470 | */ | |
471 | static int pidfd_create(struct pid *pid) | |
472 | { | |
473 | int fd; | |
474 | ||
475 | fd = anon_inode_getfd("[pidfd]", &pidfd_fops, get_pid(pid), | |
476 | O_RDWR | O_CLOEXEC); | |
477 | if (fd < 0) | |
478 | put_pid(pid); | |
479 | ||
480 | return fd; | |
481 | } | |
482 | ||
483 | /** | |
484 | * pidfd_open() - Open new pid file descriptor. | |
485 | * | |
486 | * @pid: pid for which to retrieve a pidfd | |
487 | * @flags: flags to pass | |
488 | * | |
489 | * This creates a new pid file descriptor with the O_CLOEXEC flag set for | |
490 | * the process identified by @pid. Currently, the process identified by | |
491 | * @pid must be a thread-group leader. This restriction currently exists | |
492 | * for all aspects of pidfds including pidfd creation (CLONE_PIDFD cannot | |
493 | * be used with CLONE_THREAD) and pidfd polling (only supports thread group | |
494 | * leaders). | |
495 | * | |
496 | * Return: On success, a cloexec pidfd is returned. | |
497 | * On error, a negative errno number will be returned. | |
498 | */ | |
499 | SYSCALL_DEFINE2(pidfd_open, pid_t, pid, unsigned int, flags) | |
500 | { | |
501 | int fd, ret; | |
502 | struct pid *p; | |
503 | ||
504 | if (flags) | |
505 | return -EINVAL; | |
506 | ||
507 | if (pid <= 0) | |
508 | return -EINVAL; | |
509 | ||
510 | p = find_get_pid(pid); | |
511 | if (!p) | |
512 | return -ESRCH; | |
513 | ||
514 | ret = 0; | |
515 | rcu_read_lock(); | |
516 | if (!pid_task(p, PIDTYPE_TGID)) | |
517 | ret = -EINVAL; | |
518 | rcu_read_unlock(); | |
519 | ||
520 | fd = ret ?: pidfd_create(p); | |
521 | put_pid(p); | |
522 | return fd; | |
523 | } | |
524 | ||
95846ecf | 525 | void __init pid_idr_init(void) |
1da177e4 | 526 | { |
840d6fe7 | 527 | /* Verify no one has done anything silly: */ |
e8cfbc24 | 528 | BUILD_BUG_ON(PID_MAX_LIMIT >= PIDNS_ADDING); |
c876ad76 | 529 | |
72680a19 HB |
530 | /* bump default and minimum pid_max based on number of cpus */ |
531 | pid_max = min(pid_max_max, max_t(int, pid_max, | |
532 | PIDS_PER_CPU_DEFAULT * num_possible_cpus())); | |
533 | pid_max_min = max_t(int, pid_max_min, | |
534 | PIDS_PER_CPU_MIN * num_possible_cpus()); | |
535 | pr_info("pid_max: default: %u minimum: %u\n", pid_max, pid_max_min); | |
536 | ||
95846ecf | 537 | idr_init(&init_pid_ns.idr); |
92476d7f | 538 | |
74bd59bb | 539 | init_pid_ns.pid_cachep = KMEM_CACHE(pid, |
5d097056 | 540 | SLAB_HWCACHE_ALIGN | SLAB_PANIC | SLAB_ACCOUNT); |
1da177e4 | 541 | } |