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> | |
33 | #include <linux/bootmem.h> | |
34 | #include <linux/hash.h> | |
61a58c6c | 35 | #include <linux/pid_namespace.h> |
820e45db | 36 | #include <linux/init_task.h> |
3eb07c8c | 37 | #include <linux/syscalls.h> |
1da177e4 | 38 | |
8ef047aa PE |
39 | #define pid_hashfn(nr, ns) \ |
40 | hash_long((unsigned long)nr + (unsigned long)ns, pidhash_shift) | |
92476d7f | 41 | static struct hlist_head *pid_hash; |
1da177e4 | 42 | static int pidhash_shift; |
820e45db | 43 | struct pid init_struct_pid = INIT_STRUCT_PID; |
1da177e4 LT |
44 | |
45 | int pid_max = PID_MAX_DEFAULT; | |
1da177e4 LT |
46 | |
47 | #define RESERVED_PIDS 300 | |
48 | ||
49 | int pid_max_min = RESERVED_PIDS + 1; | |
50 | int pid_max_max = PID_MAX_LIMIT; | |
51 | ||
1da177e4 LT |
52 | #define BITS_PER_PAGE (PAGE_SIZE*8) |
53 | #define BITS_PER_PAGE_MASK (BITS_PER_PAGE-1) | |
3fbc9648 | 54 | |
61a58c6c SB |
55 | static inline int mk_pid(struct pid_namespace *pid_ns, |
56 | struct pidmap *map, int off) | |
3fbc9648 | 57 | { |
61a58c6c | 58 | return (map - pid_ns->pidmap)*BITS_PER_PAGE + off; |
3fbc9648 SB |
59 | } |
60 | ||
1da177e4 LT |
61 | #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 | */ | |
61a58c6c | 70 | struct pid_namespace init_pid_ns = { |
9a575a92 CLG |
71 | .kref = { |
72 | .refcount = ATOMIC_INIT(2), | |
73 | }, | |
3fbc9648 SB |
74 | .pidmap = { |
75 | [ 0 ... PIDMAP_ENTRIES-1] = { ATOMIC_INIT(BITS_PER_PAGE), NULL } | |
76 | }, | |
84d73786 | 77 | .last_pid = 0, |
faacbfd3 PE |
78 | .level = 0, |
79 | .child_reaper = &init_task, | |
3fbc9648 | 80 | }; |
198fe21b | 81 | EXPORT_SYMBOL_GPL(init_pid_ns); |
1da177e4 | 82 | |
b461cc03 | 83 | int is_container_init(struct task_struct *tsk) |
b460cbc5 | 84 | { |
b461cc03 PE |
85 | 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; | |
b460cbc5 | 95 | } |
b461cc03 | 96 | EXPORT_SYMBOL(is_container_init); |
b460cbc5 | 97 | |
92476d7f EB |
98 | /* |
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 | */ | |
3fbc9648 | 111 | |
1da177e4 LT |
112 | static __cacheline_aligned_in_smp DEFINE_SPINLOCK(pidmap_lock); |
113 | ||
b7127aa4 | 114 | static void free_pidmap(struct upid *upid) |
1da177e4 | 115 | { |
b7127aa4 ON |
116 | int nr = upid->nr; |
117 | struct pidmap *map = upid->ns->pidmap + nr / BITS_PER_PAGE; | |
118 | int offset = nr & BITS_PER_PAGE_MASK; | |
1da177e4 LT |
119 | |
120 | clear_bit(offset, map->page); | |
121 | atomic_inc(&map->nr_free); | |
122 | } | |
123 | ||
61a58c6c | 124 | static int alloc_pidmap(struct pid_namespace *pid_ns) |
1da177e4 | 125 | { |
61a58c6c | 126 | int i, offset, max_scan, pid, last = pid_ns->last_pid; |
6a1f3b84 | 127 | struct pidmap *map; |
1da177e4 LT |
128 | |
129 | pid = last + 1; | |
130 | if (pid >= pid_max) | |
131 | pid = RESERVED_PIDS; | |
132 | offset = pid & BITS_PER_PAGE_MASK; | |
61a58c6c | 133 | map = &pid_ns->pidmap[pid/BITS_PER_PAGE]; |
1da177e4 LT |
134 | max_scan = (pid_max + BITS_PER_PAGE - 1)/BITS_PER_PAGE - !offset; |
135 | for (i = 0; i <= max_scan; ++i) { | |
136 | if (unlikely(!map->page)) { | |
3fbc9648 | 137 | void *page = kzalloc(PAGE_SIZE, GFP_KERNEL); |
1da177e4 LT |
138 | /* |
139 | * Free the page if someone raced with us | |
140 | * installing it: | |
141 | */ | |
92476d7f | 142 | spin_lock_irq(&pidmap_lock); |
1da177e4 | 143 | if (map->page) |
3fbc9648 | 144 | kfree(page); |
1da177e4 | 145 | else |
3fbc9648 | 146 | map->page = page; |
92476d7f | 147 | spin_unlock_irq(&pidmap_lock); |
1da177e4 LT |
148 | if (unlikely(!map->page)) |
149 | break; | |
150 | } | |
151 | if (likely(atomic_read(&map->nr_free))) { | |
152 | do { | |
153 | if (!test_and_set_bit(offset, map->page)) { | |
154 | atomic_dec(&map->nr_free); | |
61a58c6c | 155 | pid_ns->last_pid = pid; |
1da177e4 LT |
156 | return pid; |
157 | } | |
158 | offset = find_next_offset(map, offset); | |
61a58c6c | 159 | pid = mk_pid(pid_ns, map, offset); |
1da177e4 LT |
160 | /* |
161 | * find_next_offset() found a bit, the pid from it | |
162 | * is in-bounds, and if we fell back to the last | |
163 | * bitmap block and the final block was the same | |
164 | * as the starting point, pid is before last_pid. | |
165 | */ | |
166 | } while (offset < BITS_PER_PAGE && pid < pid_max && | |
167 | (i != max_scan || pid < last || | |
168 | !((last+1) & BITS_PER_PAGE_MASK))); | |
169 | } | |
61a58c6c | 170 | if (map < &pid_ns->pidmap[(pid_max-1)/BITS_PER_PAGE]) { |
1da177e4 LT |
171 | ++map; |
172 | offset = 0; | |
173 | } else { | |
61a58c6c | 174 | map = &pid_ns->pidmap[0]; |
1da177e4 LT |
175 | offset = RESERVED_PIDS; |
176 | if (unlikely(last == offset)) | |
177 | break; | |
178 | } | |
61a58c6c | 179 | pid = mk_pid(pid_ns, map, offset); |
1da177e4 LT |
180 | } |
181 | return -1; | |
182 | } | |
183 | ||
74bd59bb | 184 | int next_pidmap(struct pid_namespace *pid_ns, int last) |
0804ef4b EB |
185 | { |
186 | int offset; | |
f40f50d3 | 187 | struct pidmap *map, *end; |
0804ef4b EB |
188 | |
189 | offset = (last + 1) & BITS_PER_PAGE_MASK; | |
61a58c6c SB |
190 | map = &pid_ns->pidmap[(last + 1)/BITS_PER_PAGE]; |
191 | end = &pid_ns->pidmap[PIDMAP_ENTRIES]; | |
f40f50d3 | 192 | for (; map < end; map++, offset = 0) { |
0804ef4b EB |
193 | if (unlikely(!map->page)) |
194 | continue; | |
195 | offset = find_next_bit((map)->page, BITS_PER_PAGE, offset); | |
196 | if (offset < BITS_PER_PAGE) | |
61a58c6c | 197 | return mk_pid(pid_ns, map, offset); |
0804ef4b EB |
198 | } |
199 | return -1; | |
200 | } | |
201 | ||
7ad5b3a5 | 202 | void put_pid(struct pid *pid) |
92476d7f | 203 | { |
baf8f0f8 PE |
204 | struct pid_namespace *ns; |
205 | ||
92476d7f EB |
206 | if (!pid) |
207 | return; | |
baf8f0f8 | 208 | |
8ef047aa | 209 | ns = pid->numbers[pid->level].ns; |
92476d7f | 210 | if ((atomic_read(&pid->count) == 1) || |
8ef047aa | 211 | atomic_dec_and_test(&pid->count)) { |
baf8f0f8 | 212 | kmem_cache_free(ns->pid_cachep, pid); |
b461cc03 | 213 | put_pid_ns(ns); |
8ef047aa | 214 | } |
92476d7f | 215 | } |
bbf73147 | 216 | EXPORT_SYMBOL_GPL(put_pid); |
92476d7f EB |
217 | |
218 | static void delayed_put_pid(struct rcu_head *rhp) | |
219 | { | |
220 | struct pid *pid = container_of(rhp, struct pid, rcu); | |
221 | put_pid(pid); | |
222 | } | |
223 | ||
7ad5b3a5 | 224 | void free_pid(struct pid *pid) |
92476d7f EB |
225 | { |
226 | /* We can be called with write_lock_irq(&tasklist_lock) held */ | |
8ef047aa | 227 | int i; |
92476d7f EB |
228 | unsigned long flags; |
229 | ||
230 | spin_lock_irqsave(&pidmap_lock, flags); | |
198fe21b PE |
231 | for (i = 0; i <= pid->level; i++) |
232 | hlist_del_rcu(&pid->numbers[i].pid_chain); | |
92476d7f EB |
233 | spin_unlock_irqrestore(&pidmap_lock, flags); |
234 | ||
8ef047aa | 235 | for (i = 0; i <= pid->level; i++) |
b7127aa4 | 236 | free_pidmap(pid->numbers + i); |
8ef047aa | 237 | |
92476d7f EB |
238 | call_rcu(&pid->rcu, delayed_put_pid); |
239 | } | |
240 | ||
8ef047aa | 241 | struct pid *alloc_pid(struct pid_namespace *ns) |
92476d7f EB |
242 | { |
243 | struct pid *pid; | |
244 | enum pid_type type; | |
8ef047aa PE |
245 | int i, nr; |
246 | struct pid_namespace *tmp; | |
198fe21b | 247 | struct upid *upid; |
92476d7f | 248 | |
baf8f0f8 | 249 | pid = kmem_cache_alloc(ns->pid_cachep, GFP_KERNEL); |
92476d7f EB |
250 | if (!pid) |
251 | goto out; | |
252 | ||
8ef047aa PE |
253 | tmp = ns; |
254 | for (i = ns->level; i >= 0; i--) { | |
255 | nr = alloc_pidmap(tmp); | |
256 | if (nr < 0) | |
257 | goto out_free; | |
92476d7f | 258 | |
8ef047aa PE |
259 | pid->numbers[i].nr = nr; |
260 | pid->numbers[i].ns = tmp; | |
261 | tmp = tmp->parent; | |
262 | } | |
263 | ||
b461cc03 | 264 | get_pid_ns(ns); |
8ef047aa | 265 | pid->level = ns->level; |
92476d7f | 266 | atomic_set(&pid->count, 1); |
92476d7f EB |
267 | for (type = 0; type < PIDTYPE_MAX; ++type) |
268 | INIT_HLIST_HEAD(&pid->tasks[type]); | |
269 | ||
270 | spin_lock_irq(&pidmap_lock); | |
198fe21b PE |
271 | for (i = ns->level; i >= 0; i--) { |
272 | upid = &pid->numbers[i]; | |
273 | hlist_add_head_rcu(&upid->pid_chain, | |
274 | &pid_hash[pid_hashfn(upid->nr, upid->ns)]); | |
275 | } | |
92476d7f EB |
276 | spin_unlock_irq(&pidmap_lock); |
277 | ||
278 | out: | |
279 | return pid; | |
280 | ||
281 | out_free: | |
b7127aa4 ON |
282 | while (++i <= ns->level) |
283 | free_pidmap(pid->numbers + i); | |
8ef047aa | 284 | |
baf8f0f8 | 285 | kmem_cache_free(ns->pid_cachep, pid); |
92476d7f EB |
286 | pid = NULL; |
287 | goto out; | |
288 | } | |
289 | ||
7ad5b3a5 | 290 | struct pid *find_pid_ns(int nr, struct pid_namespace *ns) |
1da177e4 LT |
291 | { |
292 | struct hlist_node *elem; | |
198fe21b PE |
293 | struct upid *pnr; |
294 | ||
295 | hlist_for_each_entry_rcu(pnr, elem, | |
296 | &pid_hash[pid_hashfn(nr, ns)], pid_chain) | |
297 | if (pnr->nr == nr && pnr->ns == ns) | |
298 | return container_of(pnr, struct pid, | |
299 | numbers[ns->level]); | |
1da177e4 | 300 | |
1da177e4 LT |
301 | return NULL; |
302 | } | |
198fe21b | 303 | EXPORT_SYMBOL_GPL(find_pid_ns); |
1da177e4 | 304 | |
8990571e PE |
305 | struct pid *find_vpid(int nr) |
306 | { | |
307 | return find_pid_ns(nr, current->nsproxy->pid_ns); | |
308 | } | |
309 | EXPORT_SYMBOL_GPL(find_vpid); | |
310 | ||
311 | struct pid *find_pid(int nr) | |
312 | { | |
313 | return find_pid_ns(nr, &init_pid_ns); | |
314 | } | |
315 | EXPORT_SYMBOL_GPL(find_pid); | |
316 | ||
e713d0da SB |
317 | /* |
318 | * attach_pid() must be called with the tasklist_lock write-held. | |
319 | */ | |
24336eae | 320 | void attach_pid(struct task_struct *task, enum pid_type type, |
e713d0da | 321 | struct pid *pid) |
1da177e4 | 322 | { |
92476d7f | 323 | struct pid_link *link; |
92476d7f | 324 | |
92476d7f | 325 | link = &task->pids[type]; |
e713d0da | 326 | link->pid = pid; |
92476d7f | 327 | hlist_add_head_rcu(&link->node, &pid->tasks[type]); |
1da177e4 LT |
328 | } |
329 | ||
24336eae ON |
330 | static void __change_pid(struct task_struct *task, enum pid_type type, |
331 | struct pid *new) | |
1da177e4 | 332 | { |
92476d7f EB |
333 | struct pid_link *link; |
334 | struct pid *pid; | |
335 | int tmp; | |
1da177e4 | 336 | |
92476d7f EB |
337 | link = &task->pids[type]; |
338 | pid = link->pid; | |
1da177e4 | 339 | |
92476d7f | 340 | hlist_del_rcu(&link->node); |
24336eae | 341 | link->pid = new; |
1da177e4 | 342 | |
92476d7f EB |
343 | for (tmp = PIDTYPE_MAX; --tmp >= 0; ) |
344 | if (!hlist_empty(&pid->tasks[tmp])) | |
345 | return; | |
1da177e4 | 346 | |
92476d7f | 347 | free_pid(pid); |
1da177e4 LT |
348 | } |
349 | ||
24336eae ON |
350 | void detach_pid(struct task_struct *task, enum pid_type type) |
351 | { | |
352 | __change_pid(task, type, NULL); | |
353 | } | |
354 | ||
355 | void change_pid(struct task_struct *task, enum pid_type type, | |
356 | struct pid *pid) | |
357 | { | |
358 | __change_pid(task, type, pid); | |
359 | attach_pid(task, type, pid); | |
360 | } | |
361 | ||
c18258c6 | 362 | /* transfer_pid is an optimization of attach_pid(new), detach_pid(old) */ |
7ad5b3a5 | 363 | void transfer_pid(struct task_struct *old, struct task_struct *new, |
c18258c6 EB |
364 | enum pid_type type) |
365 | { | |
366 | new->pids[type].pid = old->pids[type].pid; | |
367 | hlist_replace_rcu(&old->pids[type].node, &new->pids[type].node); | |
c18258c6 EB |
368 | } |
369 | ||
7ad5b3a5 | 370 | struct task_struct *pid_task(struct pid *pid, enum pid_type type) |
1da177e4 | 371 | { |
92476d7f EB |
372 | struct task_struct *result = NULL; |
373 | if (pid) { | |
374 | struct hlist_node *first; | |
375 | first = rcu_dereference(pid->tasks[type].first); | |
376 | if (first) | |
377 | result = hlist_entry(first, struct task_struct, pids[(type)].node); | |
378 | } | |
379 | return result; | |
380 | } | |
eccba068 | 381 | EXPORT_SYMBOL(pid_task); |
1da177e4 | 382 | |
92476d7f EB |
383 | /* |
384 | * Must be called under rcu_read_lock() or with tasklist_lock read-held. | |
385 | */ | |
198fe21b PE |
386 | struct task_struct *find_task_by_pid_type_ns(int type, int nr, |
387 | struct pid_namespace *ns) | |
92476d7f | 388 | { |
198fe21b | 389 | return pid_task(find_pid_ns(nr, ns), type); |
92476d7f | 390 | } |
1da177e4 | 391 | |
198fe21b | 392 | EXPORT_SYMBOL(find_task_by_pid_type_ns); |
1da177e4 | 393 | |
228ebcbe PE |
394 | struct task_struct *find_task_by_vpid(pid_t vnr) |
395 | { | |
396 | return find_task_by_pid_type_ns(PIDTYPE_PID, vnr, | |
397 | current->nsproxy->pid_ns); | |
398 | } | |
399 | EXPORT_SYMBOL(find_task_by_vpid); | |
400 | ||
401 | struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns) | |
402 | { | |
403 | return find_task_by_pid_type_ns(PIDTYPE_PID, nr, ns); | |
404 | } | |
405 | EXPORT_SYMBOL(find_task_by_pid_ns); | |
406 | ||
1a657f78 ON |
407 | struct pid *get_task_pid(struct task_struct *task, enum pid_type type) |
408 | { | |
409 | struct pid *pid; | |
410 | rcu_read_lock(); | |
411 | pid = get_pid(task->pids[type].pid); | |
412 | rcu_read_unlock(); | |
413 | return pid; | |
414 | } | |
415 | ||
7ad5b3a5 | 416 | struct task_struct *get_pid_task(struct pid *pid, enum pid_type type) |
92476d7f EB |
417 | { |
418 | struct task_struct *result; | |
419 | rcu_read_lock(); | |
420 | result = pid_task(pid, type); | |
421 | if (result) | |
422 | get_task_struct(result); | |
423 | rcu_read_unlock(); | |
424 | return result; | |
1da177e4 LT |
425 | } |
426 | ||
92476d7f | 427 | struct pid *find_get_pid(pid_t nr) |
1da177e4 LT |
428 | { |
429 | struct pid *pid; | |
430 | ||
92476d7f | 431 | rcu_read_lock(); |
198fe21b | 432 | pid = get_pid(find_vpid(nr)); |
92476d7f | 433 | rcu_read_unlock(); |
1da177e4 | 434 | |
92476d7f | 435 | return pid; |
1da177e4 LT |
436 | } |
437 | ||
7af57294 PE |
438 | pid_t pid_nr_ns(struct pid *pid, struct pid_namespace *ns) |
439 | { | |
440 | struct upid *upid; | |
441 | pid_t nr = 0; | |
442 | ||
443 | if (pid && ns->level <= pid->level) { | |
444 | upid = &pid->numbers[ns->level]; | |
445 | if (upid->ns == ns) | |
446 | nr = upid->nr; | |
447 | } | |
448 | return nr; | |
449 | } | |
450 | ||
44c4e1b2 EB |
451 | pid_t pid_vnr(struct pid *pid) |
452 | { | |
453 | return pid_nr_ns(pid, current->nsproxy->pid_ns); | |
454 | } | |
455 | EXPORT_SYMBOL_GPL(pid_vnr); | |
456 | ||
2f2a3a46 PE |
457 | pid_t task_pid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns) |
458 | { | |
459 | return pid_nr_ns(task_pid(tsk), ns); | |
460 | } | |
461 | EXPORT_SYMBOL(task_pid_nr_ns); | |
462 | ||
463 | pid_t task_tgid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns) | |
464 | { | |
465 | return pid_nr_ns(task_tgid(tsk), ns); | |
466 | } | |
467 | EXPORT_SYMBOL(task_tgid_nr_ns); | |
468 | ||
469 | pid_t task_pgrp_nr_ns(struct task_struct *tsk, struct pid_namespace *ns) | |
470 | { | |
471 | return pid_nr_ns(task_pgrp(tsk), ns); | |
472 | } | |
473 | EXPORT_SYMBOL(task_pgrp_nr_ns); | |
474 | ||
475 | pid_t task_session_nr_ns(struct task_struct *tsk, struct pid_namespace *ns) | |
476 | { | |
477 | return pid_nr_ns(task_session(tsk), ns); | |
478 | } | |
479 | EXPORT_SYMBOL(task_session_nr_ns); | |
480 | ||
0804ef4b EB |
481 | /* |
482 | * Used by proc to find the first pid that is greater then or equal to nr. | |
483 | * | |
484 | * If there is a pid at nr this function is exactly the same as find_pid. | |
485 | */ | |
198fe21b | 486 | struct pid *find_ge_pid(int nr, struct pid_namespace *ns) |
0804ef4b EB |
487 | { |
488 | struct pid *pid; | |
489 | ||
490 | do { | |
198fe21b | 491 | pid = find_pid_ns(nr, ns); |
0804ef4b EB |
492 | if (pid) |
493 | break; | |
198fe21b | 494 | nr = next_pidmap(ns, nr); |
0804ef4b EB |
495 | } while (nr > 0); |
496 | ||
497 | return pid; | |
498 | } | |
bbf73147 | 499 | EXPORT_SYMBOL_GPL(find_get_pid); |
0804ef4b | 500 | |
1da177e4 LT |
501 | /* |
502 | * The pid hash table is scaled according to the amount of memory in the | |
503 | * machine. From a minimum of 16 slots up to 4096 slots at one gigabyte or | |
504 | * more. | |
505 | */ | |
506 | void __init pidhash_init(void) | |
507 | { | |
92476d7f | 508 | int i, pidhash_size; |
1da177e4 LT |
509 | unsigned long megabytes = nr_kernel_pages >> (20 - PAGE_SHIFT); |
510 | ||
511 | pidhash_shift = max(4, fls(megabytes * 4)); | |
512 | pidhash_shift = min(12, pidhash_shift); | |
513 | pidhash_size = 1 << pidhash_shift; | |
514 | ||
515 | printk("PID hash table entries: %d (order: %d, %Zd bytes)\n", | |
516 | pidhash_size, pidhash_shift, | |
92476d7f EB |
517 | pidhash_size * sizeof(struct hlist_head)); |
518 | ||
519 | pid_hash = alloc_bootmem(pidhash_size * sizeof(*(pid_hash))); | |
520 | if (!pid_hash) | |
521 | panic("Could not alloc pidhash!\n"); | |
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 | } |