Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
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). | |
30e49c26 PE |
21 | * |
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 | * | |
1da177e4 LT |
27 | */ |
28 | ||
29 | #include <linux/mm.h> | |
30 | #include <linux/module.h> | |
31 | #include <linux/slab.h> | |
32 | #include <linux/init.h> | |
82524746 | 33 | #include <linux/rculist.h> |
1da177e4 LT |
34 | #include <linux/bootmem.h> |
35 | #include <linux/hash.h> | |
61a58c6c | 36 | #include <linux/pid_namespace.h> |
820e45db | 37 | #include <linux/init_task.h> |
3eb07c8c | 38 | #include <linux/syscalls.h> |
12de38b1 | 39 | #include <linux/kmemleak.h> |
1da177e4 | 40 | |
8ef047aa PE |
41 | #define pid_hashfn(nr, ns) \ |
42 | hash_long((unsigned long)nr + (unsigned long)ns, pidhash_shift) | |
92476d7f | 43 | static struct hlist_head *pid_hash; |
1da177e4 | 44 | static int pidhash_shift; |
820e45db | 45 | struct pid init_struct_pid = INIT_STRUCT_PID; |
1da177e4 LT |
46 | |
47 | int pid_max = PID_MAX_DEFAULT; | |
1da177e4 LT |
48 | |
49 | #define RESERVED_PIDS 300 | |
50 | ||
51 | int pid_max_min = RESERVED_PIDS + 1; | |
52 | int pid_max_max = PID_MAX_LIMIT; | |
53 | ||
1da177e4 LT |
54 | #define BITS_PER_PAGE (PAGE_SIZE*8) |
55 | #define BITS_PER_PAGE_MASK (BITS_PER_PAGE-1) | |
3fbc9648 | 56 | |
61a58c6c SB |
57 | static inline int mk_pid(struct pid_namespace *pid_ns, |
58 | struct pidmap *map, int off) | |
3fbc9648 | 59 | { |
61a58c6c | 60 | return (map - pid_ns->pidmap)*BITS_PER_PAGE + off; |
3fbc9648 SB |
61 | } |
62 | ||
1da177e4 LT |
63 | #define find_next_offset(map, off) \ |
64 | find_next_zero_bit((map)->page, BITS_PER_PAGE, off) | |
65 | ||
66 | /* | |
67 | * PID-map pages start out as NULL, they get allocated upon | |
68 | * first use and are never deallocated. This way a low pid_max | |
69 | * value does not cause lots of bitmaps to be allocated, but | |
70 | * the scheme scales to up to 4 million PIDs, runtime. | |
71 | */ | |
61a58c6c | 72 | struct pid_namespace init_pid_ns = { |
9a575a92 CLG |
73 | .kref = { |
74 | .refcount = ATOMIC_INIT(2), | |
75 | }, | |
3fbc9648 SB |
76 | .pidmap = { |
77 | [ 0 ... PIDMAP_ENTRIES-1] = { ATOMIC_INIT(BITS_PER_PAGE), NULL } | |
78 | }, | |
84d73786 | 79 | .last_pid = 0, |
faacbfd3 PE |
80 | .level = 0, |
81 | .child_reaper = &init_task, | |
3fbc9648 | 82 | }; |
198fe21b | 83 | EXPORT_SYMBOL_GPL(init_pid_ns); |
1da177e4 | 84 | |
b461cc03 | 85 | int is_container_init(struct task_struct *tsk) |
b460cbc5 | 86 | { |
b461cc03 PE |
87 | int ret = 0; |
88 | struct pid *pid; | |
89 | ||
90 | rcu_read_lock(); | |
91 | pid = task_pid(tsk); | |
92 | if (pid != NULL && pid->numbers[pid->level].nr == 1) | |
93 | ret = 1; | |
94 | rcu_read_unlock(); | |
95 | ||
96 | return ret; | |
b460cbc5 | 97 | } |
b461cc03 | 98 | EXPORT_SYMBOL(is_container_init); |
b460cbc5 | 99 | |
92476d7f EB |
100 | /* |
101 | * Note: disable interrupts while the pidmap_lock is held as an | |
102 | * interrupt might come in and do read_lock(&tasklist_lock). | |
103 | * | |
104 | * If we don't disable interrupts there is a nasty deadlock between | |
105 | * detach_pid()->free_pid() and another cpu that does | |
106 | * spin_lock(&pidmap_lock) followed by an interrupt routine that does | |
107 | * read_lock(&tasklist_lock); | |
108 | * | |
109 | * After we clean up the tasklist_lock and know there are no | |
110 | * irq handlers that take it we can leave the interrupts enabled. | |
111 | * For now it is easier to be safe than to prove it can't happen. | |
112 | */ | |
3fbc9648 | 113 | |
1da177e4 LT |
114 | static __cacheline_aligned_in_smp DEFINE_SPINLOCK(pidmap_lock); |
115 | ||
b7127aa4 | 116 | static void free_pidmap(struct upid *upid) |
1da177e4 | 117 | { |
b7127aa4 ON |
118 | int nr = upid->nr; |
119 | struct pidmap *map = upid->ns->pidmap + nr / BITS_PER_PAGE; | |
120 | int offset = nr & BITS_PER_PAGE_MASK; | |
1da177e4 LT |
121 | |
122 | clear_bit(offset, map->page); | |
123 | atomic_inc(&map->nr_free); | |
124 | } | |
125 | ||
61a58c6c | 126 | static int alloc_pidmap(struct pid_namespace *pid_ns) |
1da177e4 | 127 | { |
61a58c6c | 128 | int i, offset, max_scan, pid, last = pid_ns->last_pid; |
6a1f3b84 | 129 | struct pidmap *map; |
1da177e4 LT |
130 | |
131 | pid = last + 1; | |
132 | if (pid >= pid_max) | |
133 | pid = RESERVED_PIDS; | |
134 | offset = pid & BITS_PER_PAGE_MASK; | |
61a58c6c | 135 | map = &pid_ns->pidmap[pid/BITS_PER_PAGE]; |
1da177e4 LT |
136 | max_scan = (pid_max + BITS_PER_PAGE - 1)/BITS_PER_PAGE - !offset; |
137 | for (i = 0; i <= max_scan; ++i) { | |
138 | if (unlikely(!map->page)) { | |
3fbc9648 | 139 | void *page = kzalloc(PAGE_SIZE, GFP_KERNEL); |
1da177e4 LT |
140 | /* |
141 | * Free the page if someone raced with us | |
142 | * installing it: | |
143 | */ | |
92476d7f | 144 | spin_lock_irq(&pidmap_lock); |
1da177e4 | 145 | if (map->page) |
3fbc9648 | 146 | kfree(page); |
1da177e4 | 147 | else |
3fbc9648 | 148 | map->page = page; |
92476d7f | 149 | spin_unlock_irq(&pidmap_lock); |
1da177e4 LT |
150 | 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); | |
61a58c6c | 157 | pid_ns->last_pid = pid; |
1da177e4 LT |
158 | return pid; |
159 | } | |
160 | offset = find_next_offset(map, offset); | |
61a58c6c | 161 | pid = mk_pid(pid_ns, map, offset); |
1da177e4 LT |
162 | /* |
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 | } | |
61a58c6c | 172 | if (map < &pid_ns->pidmap[(pid_max-1)/BITS_PER_PAGE]) { |
1da177e4 LT |
173 | ++map; |
174 | offset = 0; | |
175 | } else { | |
61a58c6c | 176 | map = &pid_ns->pidmap[0]; |
1da177e4 LT |
177 | offset = RESERVED_PIDS; |
178 | if (unlikely(last == offset)) | |
179 | break; | |
180 | } | |
61a58c6c | 181 | pid = mk_pid(pid_ns, map, offset); |
1da177e4 LT |
182 | } |
183 | return -1; | |
184 | } | |
185 | ||
74bd59bb | 186 | int next_pidmap(struct pid_namespace *pid_ns, int last) |
0804ef4b EB |
187 | { |
188 | int offset; | |
f40f50d3 | 189 | struct pidmap *map, *end; |
0804ef4b EB |
190 | |
191 | offset = (last + 1) & BITS_PER_PAGE_MASK; | |
61a58c6c SB |
192 | map = &pid_ns->pidmap[(last + 1)/BITS_PER_PAGE]; |
193 | end = &pid_ns->pidmap[PIDMAP_ENTRIES]; | |
f40f50d3 | 194 | for (; map < end; map++, offset = 0) { |
0804ef4b EB |
195 | if (unlikely(!map->page)) |
196 | continue; | |
197 | offset = find_next_bit((map)->page, BITS_PER_PAGE, offset); | |
198 | if (offset < BITS_PER_PAGE) | |
61a58c6c | 199 | return mk_pid(pid_ns, map, offset); |
0804ef4b EB |
200 | } |
201 | return -1; | |
202 | } | |
203 | ||
7ad5b3a5 | 204 | void put_pid(struct pid *pid) |
92476d7f | 205 | { |
baf8f0f8 PE |
206 | struct pid_namespace *ns; |
207 | ||
92476d7f EB |
208 | if (!pid) |
209 | return; | |
baf8f0f8 | 210 | |
8ef047aa | 211 | ns = pid->numbers[pid->level].ns; |
92476d7f | 212 | if ((atomic_read(&pid->count) == 1) || |
8ef047aa | 213 | atomic_dec_and_test(&pid->count)) { |
baf8f0f8 | 214 | kmem_cache_free(ns->pid_cachep, pid); |
b461cc03 | 215 | put_pid_ns(ns); |
8ef047aa | 216 | } |
92476d7f | 217 | } |
bbf73147 | 218 | EXPORT_SYMBOL_GPL(put_pid); |
92476d7f EB |
219 | |
220 | static 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 | ||
7ad5b3a5 | 226 | void free_pid(struct pid *pid) |
92476d7f EB |
227 | { |
228 | /* We can be called with write_lock_irq(&tasklist_lock) held */ | |
8ef047aa | 229 | int i; |
92476d7f EB |
230 | unsigned long flags; |
231 | ||
232 | spin_lock_irqsave(&pidmap_lock, flags); | |
198fe21b PE |
233 | for (i = 0; i <= pid->level; i++) |
234 | hlist_del_rcu(&pid->numbers[i].pid_chain); | |
92476d7f EB |
235 | spin_unlock_irqrestore(&pidmap_lock, flags); |
236 | ||
8ef047aa | 237 | for (i = 0; i <= pid->level; i++) |
b7127aa4 | 238 | free_pidmap(pid->numbers + i); |
8ef047aa | 239 | |
92476d7f EB |
240 | call_rcu(&pid->rcu, delayed_put_pid); |
241 | } | |
242 | ||
8ef047aa | 243 | struct pid *alloc_pid(struct pid_namespace *ns) |
92476d7f EB |
244 | { |
245 | struct pid *pid; | |
246 | enum pid_type type; | |
8ef047aa PE |
247 | int i, nr; |
248 | struct pid_namespace *tmp; | |
198fe21b | 249 | struct upid *upid; |
92476d7f | 250 | |
baf8f0f8 | 251 | pid = kmem_cache_alloc(ns->pid_cachep, GFP_KERNEL); |
92476d7f EB |
252 | if (!pid) |
253 | goto out; | |
254 | ||
8ef047aa PE |
255 | tmp = ns; |
256 | for (i = ns->level; i >= 0; i--) { | |
257 | nr = alloc_pidmap(tmp); | |
258 | if (nr < 0) | |
259 | goto out_free; | |
92476d7f | 260 | |
8ef047aa PE |
261 | pid->numbers[i].nr = nr; |
262 | pid->numbers[i].ns = tmp; | |
263 | tmp = tmp->parent; | |
264 | } | |
265 | ||
b461cc03 | 266 | get_pid_ns(ns); |
8ef047aa | 267 | pid->level = ns->level; |
92476d7f | 268 | atomic_set(&pid->count, 1); |
92476d7f EB |
269 | for (type = 0; type < PIDTYPE_MAX; ++type) |
270 | INIT_HLIST_HEAD(&pid->tasks[type]); | |
271 | ||
272 | spin_lock_irq(&pidmap_lock); | |
198fe21b PE |
273 | for (i = ns->level; i >= 0; i--) { |
274 | upid = &pid->numbers[i]; | |
275 | hlist_add_head_rcu(&upid->pid_chain, | |
276 | &pid_hash[pid_hashfn(upid->nr, upid->ns)]); | |
277 | } | |
92476d7f EB |
278 | spin_unlock_irq(&pidmap_lock); |
279 | ||
280 | out: | |
281 | return pid; | |
282 | ||
283 | out_free: | |
b7127aa4 ON |
284 | while (++i <= ns->level) |
285 | free_pidmap(pid->numbers + i); | |
8ef047aa | 286 | |
baf8f0f8 | 287 | kmem_cache_free(ns->pid_cachep, pid); |
92476d7f EB |
288 | pid = NULL; |
289 | goto out; | |
290 | } | |
291 | ||
7ad5b3a5 | 292 | struct pid *find_pid_ns(int nr, struct pid_namespace *ns) |
1da177e4 LT |
293 | { |
294 | struct hlist_node *elem; | |
198fe21b PE |
295 | struct upid *pnr; |
296 | ||
297 | hlist_for_each_entry_rcu(pnr, elem, | |
298 | &pid_hash[pid_hashfn(nr, ns)], pid_chain) | |
299 | if (pnr->nr == nr && pnr->ns == ns) | |
300 | return container_of(pnr, struct pid, | |
301 | numbers[ns->level]); | |
1da177e4 | 302 | |
1da177e4 LT |
303 | return NULL; |
304 | } | |
198fe21b | 305 | EXPORT_SYMBOL_GPL(find_pid_ns); |
1da177e4 | 306 | |
8990571e PE |
307 | struct pid *find_vpid(int nr) |
308 | { | |
309 | return find_pid_ns(nr, current->nsproxy->pid_ns); | |
310 | } | |
311 | EXPORT_SYMBOL_GPL(find_vpid); | |
312 | ||
e713d0da SB |
313 | /* |
314 | * attach_pid() must be called with the tasklist_lock write-held. | |
315 | */ | |
24336eae | 316 | void attach_pid(struct task_struct *task, enum pid_type type, |
e713d0da | 317 | struct pid *pid) |
1da177e4 | 318 | { |
92476d7f | 319 | struct pid_link *link; |
92476d7f | 320 | |
92476d7f | 321 | link = &task->pids[type]; |
e713d0da | 322 | link->pid = pid; |
92476d7f | 323 | hlist_add_head_rcu(&link->node, &pid->tasks[type]); |
1da177e4 LT |
324 | } |
325 | ||
24336eae ON |
326 | static void __change_pid(struct task_struct *task, enum pid_type type, |
327 | struct pid *new) | |
1da177e4 | 328 | { |
92476d7f EB |
329 | struct pid_link *link; |
330 | struct pid *pid; | |
331 | int tmp; | |
1da177e4 | 332 | |
92476d7f EB |
333 | link = &task->pids[type]; |
334 | pid = link->pid; | |
1da177e4 | 335 | |
92476d7f | 336 | hlist_del_rcu(&link->node); |
24336eae | 337 | link->pid = new; |
1da177e4 | 338 | |
92476d7f EB |
339 | for (tmp = PIDTYPE_MAX; --tmp >= 0; ) |
340 | if (!hlist_empty(&pid->tasks[tmp])) | |
341 | return; | |
1da177e4 | 342 | |
92476d7f | 343 | free_pid(pid); |
1da177e4 LT |
344 | } |
345 | ||
24336eae ON |
346 | void detach_pid(struct task_struct *task, enum pid_type type) |
347 | { | |
348 | __change_pid(task, type, NULL); | |
349 | } | |
350 | ||
351 | void change_pid(struct task_struct *task, enum pid_type type, | |
352 | struct pid *pid) | |
353 | { | |
354 | __change_pid(task, type, pid); | |
355 | attach_pid(task, type, pid); | |
356 | } | |
357 | ||
c18258c6 | 358 | /* transfer_pid is an optimization of attach_pid(new), detach_pid(old) */ |
7ad5b3a5 | 359 | void transfer_pid(struct task_struct *old, struct task_struct *new, |
c18258c6 EB |
360 | enum pid_type type) |
361 | { | |
362 | new->pids[type].pid = old->pids[type].pid; | |
363 | hlist_replace_rcu(&old->pids[type].node, &new->pids[type].node); | |
c18258c6 EB |
364 | } |
365 | ||
7ad5b3a5 | 366 | struct task_struct *pid_task(struct pid *pid, enum pid_type type) |
1da177e4 | 367 | { |
92476d7f EB |
368 | struct task_struct *result = NULL; |
369 | if (pid) { | |
370 | struct hlist_node *first; | |
371 | first = rcu_dereference(pid->tasks[type].first); | |
372 | if (first) | |
373 | result = hlist_entry(first, struct task_struct, pids[(type)].node); | |
374 | } | |
375 | return result; | |
376 | } | |
eccba068 | 377 | EXPORT_SYMBOL(pid_task); |
1da177e4 | 378 | |
92476d7f EB |
379 | /* |
380 | * Must be called under rcu_read_lock() or with tasklist_lock read-held. | |
381 | */ | |
17f98dcf | 382 | struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns) |
92476d7f | 383 | { |
17f98dcf | 384 | return pid_task(find_pid_ns(nr, ns), PIDTYPE_PID); |
92476d7f | 385 | } |
1da177e4 | 386 | |
228ebcbe PE |
387 | struct task_struct *find_task_by_vpid(pid_t vnr) |
388 | { | |
17f98dcf | 389 | return find_task_by_pid_ns(vnr, current->nsproxy->pid_ns); |
228ebcbe | 390 | } |
228ebcbe | 391 | |
1a657f78 ON |
392 | struct pid *get_task_pid(struct task_struct *task, enum pid_type type) |
393 | { | |
394 | struct pid *pid; | |
395 | rcu_read_lock(); | |
2ae448ef ON |
396 | if (type != PIDTYPE_PID) |
397 | task = task->group_leader; | |
1a657f78 ON |
398 | pid = get_pid(task->pids[type].pid); |
399 | rcu_read_unlock(); | |
400 | return pid; | |
401 | } | |
402 | ||
7ad5b3a5 | 403 | struct task_struct *get_pid_task(struct pid *pid, enum pid_type type) |
92476d7f EB |
404 | { |
405 | struct task_struct *result; | |
406 | rcu_read_lock(); | |
407 | result = pid_task(pid, type); | |
408 | if (result) | |
409 | get_task_struct(result); | |
410 | rcu_read_unlock(); | |
411 | return result; | |
1da177e4 LT |
412 | } |
413 | ||
92476d7f | 414 | struct pid *find_get_pid(pid_t nr) |
1da177e4 LT |
415 | { |
416 | struct pid *pid; | |
417 | ||
92476d7f | 418 | rcu_read_lock(); |
198fe21b | 419 | pid = get_pid(find_vpid(nr)); |
92476d7f | 420 | rcu_read_unlock(); |
1da177e4 | 421 | |
92476d7f | 422 | return pid; |
1da177e4 | 423 | } |
339caf2a | 424 | EXPORT_SYMBOL_GPL(find_get_pid); |
1da177e4 | 425 | |
7af57294 PE |
426 | pid_t pid_nr_ns(struct pid *pid, struct pid_namespace *ns) |
427 | { | |
428 | struct upid *upid; | |
429 | pid_t nr = 0; | |
430 | ||
431 | if (pid && ns->level <= pid->level) { | |
432 | upid = &pid->numbers[ns->level]; | |
433 | if (upid->ns == ns) | |
434 | nr = upid->nr; | |
435 | } | |
436 | return nr; | |
437 | } | |
438 | ||
44c4e1b2 EB |
439 | pid_t pid_vnr(struct pid *pid) |
440 | { | |
441 | return pid_nr_ns(pid, current->nsproxy->pid_ns); | |
442 | } | |
443 | EXPORT_SYMBOL_GPL(pid_vnr); | |
444 | ||
52ee2dfd ON |
445 | pid_t __task_pid_nr_ns(struct task_struct *task, enum pid_type type, |
446 | struct pid_namespace *ns) | |
2f2a3a46 | 447 | { |
52ee2dfd ON |
448 | pid_t nr = 0; |
449 | ||
450 | rcu_read_lock(); | |
451 | if (!ns) | |
452 | ns = current->nsproxy->pid_ns; | |
453 | if (likely(pid_alive(task))) { | |
454 | if (type != PIDTYPE_PID) | |
455 | task = task->group_leader; | |
456 | nr = pid_nr_ns(task->pids[type].pid, ns); | |
457 | } | |
458 | rcu_read_unlock(); | |
459 | ||
460 | return nr; | |
2f2a3a46 | 461 | } |
52ee2dfd | 462 | EXPORT_SYMBOL(__task_pid_nr_ns); |
2f2a3a46 PE |
463 | |
464 | pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns) | |
465 | { | |
466 | return pid_nr_ns(task_tgid(tsk), ns); | |
467 | } | |
468 | EXPORT_SYMBOL(task_tgid_nr_ns); | |
469 | ||
61bce0f1 EB |
470 | struct pid_namespace *task_active_pid_ns(struct task_struct *tsk) |
471 | { | |
472 | return ns_of_pid(task_pid(tsk)); | |
473 | } | |
474 | EXPORT_SYMBOL_GPL(task_active_pid_ns); | |
475 | ||
0804ef4b | 476 | /* |
025dfdaf | 477 | * Used by proc to find the first pid that is greater than or equal to nr. |
0804ef4b | 478 | * |
e49859e7 | 479 | * If there is a pid at nr this function is exactly the same as find_pid_ns. |
0804ef4b | 480 | */ |
198fe21b | 481 | struct pid *find_ge_pid(int nr, struct pid_namespace *ns) |
0804ef4b EB |
482 | { |
483 | struct pid *pid; | |
484 | ||
485 | do { | |
198fe21b | 486 | pid = find_pid_ns(nr, ns); |
0804ef4b EB |
487 | if (pid) |
488 | break; | |
198fe21b | 489 | nr = next_pidmap(ns, nr); |
0804ef4b EB |
490 | } while (nr > 0); |
491 | ||
492 | return pid; | |
493 | } | |
494 | ||
1da177e4 LT |
495 | /* |
496 | * The pid hash table is scaled according to the amount of memory in the | |
497 | * machine. From a minimum of 16 slots up to 4096 slots at one gigabyte or | |
498 | * more. | |
499 | */ | |
500 | void __init pidhash_init(void) | |
501 | { | |
92476d7f | 502 | int i, pidhash_size; |
1da177e4 LT |
503 | unsigned long megabytes = nr_kernel_pages >> (20 - PAGE_SHIFT); |
504 | ||
505 | pidhash_shift = max(4, fls(megabytes * 4)); | |
506 | pidhash_shift = min(12, pidhash_shift); | |
507 | pidhash_size = 1 << pidhash_shift; | |
508 | ||
509 | printk("PID hash table entries: %d (order: %d, %Zd bytes)\n", | |
510 | pidhash_size, pidhash_shift, | |
92476d7f EB |
511 | pidhash_size * sizeof(struct hlist_head)); |
512 | ||
513 | pid_hash = alloc_bootmem(pidhash_size * sizeof(*(pid_hash))); | |
514 | if (!pid_hash) | |
515 | panic("Could not alloc pidhash!\n"); | |
12de38b1 CM |
516 | /* |
517 | * pid_hash contains references to allocated struct pid objects and it | |
518 | * must be scanned by kmemleak to avoid false positives. | |
519 | */ | |
520 | kmemleak_alloc(pid_hash, pidhash_size * sizeof(*(pid_hash)), 0, | |
521 | GFP_KERNEL); | |
92476d7f EB |
522 | for (i = 0; i < pidhash_size; i++) |
523 | INIT_HLIST_HEAD(&pid_hash[i]); | |
1da177e4 LT |
524 | } |
525 | ||
526 | void __init pidmap_init(void) | |
527 | { | |
61a58c6c | 528 | init_pid_ns.pidmap[0].page = kzalloc(PAGE_SIZE, GFP_KERNEL); |
73b9ebfe | 529 | /* Reserve PID 0. We never call free_pidmap(0) */ |
61a58c6c SB |
530 | set_bit(0, init_pid_ns.pidmap[0].page); |
531 | atomic_dec(&init_pid_ns.pidmap[0].nr_free); | |
92476d7f | 532 | |
74bd59bb PE |
533 | init_pid_ns.pid_cachep = KMEM_CACHE(pid, |
534 | SLAB_HWCACHE_ALIGN | SLAB_PANIC); | |
1da177e4 | 535 | } |