diff options
Diffstat (limited to 'kernel/pid.c')
-rw-r--r-- | kernel/pid.c | 208 |
1 files changed, 18 insertions, 190 deletions
diff --git a/kernel/pid.c b/kernel/pid.c index 3b30bccdfcdc..477691576b33 100644 --- a/kernel/pid.c +++ b/kernel/pid.c | |||
@@ -41,7 +41,6 @@ | |||
41 | static struct hlist_head *pid_hash; | 41 | static struct hlist_head *pid_hash; |
42 | static int pidhash_shift; | 42 | static int pidhash_shift; |
43 | struct pid init_struct_pid = INIT_STRUCT_PID; | 43 | struct pid init_struct_pid = INIT_STRUCT_PID; |
44 | static struct kmem_cache *pid_ns_cachep; | ||
45 | 44 | ||
46 | int pid_max = PID_MAX_DEFAULT; | 45 | int pid_max = PID_MAX_DEFAULT; |
47 | 46 | ||
@@ -112,7 +111,7 @@ EXPORT_SYMBOL(is_container_init); | |||
112 | 111 | ||
113 | static __cacheline_aligned_in_smp DEFINE_SPINLOCK(pidmap_lock); | 112 | static __cacheline_aligned_in_smp DEFINE_SPINLOCK(pidmap_lock); |
114 | 113 | ||
115 | static fastcall void free_pidmap(struct pid_namespace *pid_ns, int pid) | 114 | static void free_pidmap(struct pid_namespace *pid_ns, int pid) |
116 | { | 115 | { |
117 | struct pidmap *map = pid_ns->pidmap + pid / BITS_PER_PAGE; | 116 | struct pidmap *map = pid_ns->pidmap + pid / BITS_PER_PAGE; |
118 | int offset = pid & BITS_PER_PAGE_MASK; | 117 | int offset = pid & BITS_PER_PAGE_MASK; |
@@ -181,7 +180,7 @@ static int alloc_pidmap(struct pid_namespace *pid_ns) | |||
181 | return -1; | 180 | return -1; |
182 | } | 181 | } |
183 | 182 | ||
184 | static int next_pidmap(struct pid_namespace *pid_ns, int last) | 183 | int next_pidmap(struct pid_namespace *pid_ns, int last) |
185 | { | 184 | { |
186 | int offset; | 185 | int offset; |
187 | struct pidmap *map, *end; | 186 | struct pidmap *map, *end; |
@@ -199,7 +198,7 @@ static int next_pidmap(struct pid_namespace *pid_ns, int last) | |||
199 | return -1; | 198 | return -1; |
200 | } | 199 | } |
201 | 200 | ||
202 | fastcall void put_pid(struct pid *pid) | 201 | void put_pid(struct pid *pid) |
203 | { | 202 | { |
204 | struct pid_namespace *ns; | 203 | struct pid_namespace *ns; |
205 | 204 | ||
@@ -221,7 +220,7 @@ static void delayed_put_pid(struct rcu_head *rhp) | |||
221 | put_pid(pid); | 220 | put_pid(pid); |
222 | } | 221 | } |
223 | 222 | ||
224 | fastcall void free_pid(struct pid *pid) | 223 | void free_pid(struct pid *pid) |
225 | { | 224 | { |
226 | /* We can be called with write_lock_irq(&tasklist_lock) held */ | 225 | /* We can be called with write_lock_irq(&tasklist_lock) held */ |
227 | int i; | 226 | int i; |
@@ -287,7 +286,7 @@ out_free: | |||
287 | goto out; | 286 | goto out; |
288 | } | 287 | } |
289 | 288 | ||
290 | struct pid * fastcall find_pid_ns(int nr, struct pid_namespace *ns) | 289 | struct pid *find_pid_ns(int nr, struct pid_namespace *ns) |
291 | { | 290 | { |
292 | struct hlist_node *elem; | 291 | struct hlist_node *elem; |
293 | struct upid *pnr; | 292 | struct upid *pnr; |
@@ -317,7 +316,7 @@ EXPORT_SYMBOL_GPL(find_pid); | |||
317 | /* | 316 | /* |
318 | * attach_pid() must be called with the tasklist_lock write-held. | 317 | * attach_pid() must be called with the tasklist_lock write-held. |
319 | */ | 318 | */ |
320 | int fastcall attach_pid(struct task_struct *task, enum pid_type type, | 319 | int attach_pid(struct task_struct *task, enum pid_type type, |
321 | struct pid *pid) | 320 | struct pid *pid) |
322 | { | 321 | { |
323 | struct pid_link *link; | 322 | struct pid_link *link; |
@@ -329,7 +328,7 @@ int fastcall attach_pid(struct task_struct *task, enum pid_type type, | |||
329 | return 0; | 328 | return 0; |
330 | } | 329 | } |
331 | 330 | ||
332 | void fastcall detach_pid(struct task_struct *task, enum pid_type type) | 331 | void detach_pid(struct task_struct *task, enum pid_type type) |
333 | { | 332 | { |
334 | struct pid_link *link; | 333 | struct pid_link *link; |
335 | struct pid *pid; | 334 | struct pid *pid; |
@@ -349,7 +348,7 @@ void fastcall detach_pid(struct task_struct *task, enum pid_type type) | |||
349 | } | 348 | } |
350 | 349 | ||
351 | /* transfer_pid is an optimization of attach_pid(new), detach_pid(old) */ | 350 | /* transfer_pid is an optimization of attach_pid(new), detach_pid(old) */ |
352 | void fastcall transfer_pid(struct task_struct *old, struct task_struct *new, | 351 | void transfer_pid(struct task_struct *old, struct task_struct *new, |
353 | enum pid_type type) | 352 | enum pid_type type) |
354 | { | 353 | { |
355 | new->pids[type].pid = old->pids[type].pid; | 354 | new->pids[type].pid = old->pids[type].pid; |
@@ -357,7 +356,7 @@ void fastcall transfer_pid(struct task_struct *old, struct task_struct *new, | |||
357 | old->pids[type].pid = NULL; | 356 | old->pids[type].pid = NULL; |
358 | } | 357 | } |
359 | 358 | ||
360 | struct task_struct * fastcall pid_task(struct pid *pid, enum pid_type type) | 359 | struct task_struct *pid_task(struct pid *pid, enum pid_type type) |
361 | { | 360 | { |
362 | struct task_struct *result = NULL; | 361 | struct task_struct *result = NULL; |
363 | if (pid) { | 362 | if (pid) { |
@@ -409,7 +408,7 @@ struct pid *get_task_pid(struct task_struct *task, enum pid_type type) | |||
409 | return pid; | 408 | return pid; |
410 | } | 409 | } |
411 | 410 | ||
412 | struct task_struct *fastcall get_pid_task(struct pid *pid, enum pid_type type) | 411 | struct task_struct *get_pid_task(struct pid *pid, enum pid_type type) |
413 | { | 412 | { |
414 | struct task_struct *result; | 413 | struct task_struct *result; |
415 | rcu_read_lock(); | 414 | rcu_read_lock(); |
@@ -444,6 +443,12 @@ pid_t pid_nr_ns(struct pid *pid, struct pid_namespace *ns) | |||
444 | return nr; | 443 | return nr; |
445 | } | 444 | } |
446 | 445 | ||
446 | pid_t pid_vnr(struct pid *pid) | ||
447 | { | ||
448 | return pid_nr_ns(pid, current->nsproxy->pid_ns); | ||
449 | } | ||
450 | EXPORT_SYMBOL_GPL(pid_vnr); | ||
451 | |||
447 | pid_t task_pid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns) | 452 | pid_t task_pid_nr_ns(struct task_struct *tsk, struct pid_namespace *ns) |
448 | { | 453 | { |
449 | return pid_nr_ns(task_pid(tsk), ns); | 454 | return pid_nr_ns(task_pid(tsk), ns); |
@@ -488,180 +493,6 @@ struct pid *find_ge_pid(int nr, struct pid_namespace *ns) | |||
488 | } | 493 | } |
489 | EXPORT_SYMBOL_GPL(find_get_pid); | 494 | EXPORT_SYMBOL_GPL(find_get_pid); |
490 | 495 | ||
491 | struct pid_cache { | ||
492 | int nr_ids; | ||
493 | char name[16]; | ||
494 | struct kmem_cache *cachep; | ||
495 | struct list_head list; | ||
496 | }; | ||
497 | |||
498 | static LIST_HEAD(pid_caches_lh); | ||
499 | static DEFINE_MUTEX(pid_caches_mutex); | ||
500 | |||
501 | /* | ||
502 | * creates the kmem cache to allocate pids from. | ||
503 | * @nr_ids: the number of numerical ids this pid will have to carry | ||
504 | */ | ||
505 | |||
506 | static struct kmem_cache *create_pid_cachep(int nr_ids) | ||
507 | { | ||
508 | struct pid_cache *pcache; | ||
509 | struct kmem_cache *cachep; | ||
510 | |||
511 | mutex_lock(&pid_caches_mutex); | ||
512 | list_for_each_entry (pcache, &pid_caches_lh, list) | ||
513 | if (pcache->nr_ids == nr_ids) | ||
514 | goto out; | ||
515 | |||
516 | pcache = kmalloc(sizeof(struct pid_cache), GFP_KERNEL); | ||
517 | if (pcache == NULL) | ||
518 | goto err_alloc; | ||
519 | |||
520 | snprintf(pcache->name, sizeof(pcache->name), "pid_%d", nr_ids); | ||
521 | cachep = kmem_cache_create(pcache->name, | ||
522 | sizeof(struct pid) + (nr_ids - 1) * sizeof(struct upid), | ||
523 | 0, SLAB_HWCACHE_ALIGN, NULL); | ||
524 | if (cachep == NULL) | ||
525 | goto err_cachep; | ||
526 | |||
527 | pcache->nr_ids = nr_ids; | ||
528 | pcache->cachep = cachep; | ||
529 | list_add(&pcache->list, &pid_caches_lh); | ||
530 | out: | ||
531 | mutex_unlock(&pid_caches_mutex); | ||
532 | return pcache->cachep; | ||
533 | |||
534 | err_cachep: | ||
535 | kfree(pcache); | ||
536 | err_alloc: | ||
537 | mutex_unlock(&pid_caches_mutex); | ||
538 | return NULL; | ||
539 | } | ||
540 | |||
541 | #ifdef CONFIG_PID_NS | ||
542 | static struct pid_namespace *create_pid_namespace(int level) | ||
543 | { | ||
544 | struct pid_namespace *ns; | ||
545 | int i; | ||
546 | |||
547 | ns = kmem_cache_alloc(pid_ns_cachep, GFP_KERNEL); | ||
548 | if (ns == NULL) | ||
549 | goto out; | ||
550 | |||
551 | ns->pidmap[0].page = kzalloc(PAGE_SIZE, GFP_KERNEL); | ||
552 | if (!ns->pidmap[0].page) | ||
553 | goto out_free; | ||
554 | |||
555 | ns->pid_cachep = create_pid_cachep(level + 1); | ||
556 | if (ns->pid_cachep == NULL) | ||
557 | goto out_free_map; | ||
558 | |||
559 | kref_init(&ns->kref); | ||
560 | ns->last_pid = 0; | ||
561 | ns->child_reaper = NULL; | ||
562 | ns->level = level; | ||
563 | |||
564 | set_bit(0, ns->pidmap[0].page); | ||
565 | atomic_set(&ns->pidmap[0].nr_free, BITS_PER_PAGE - 1); | ||
566 | |||
567 | for (i = 1; i < PIDMAP_ENTRIES; i++) { | ||
568 | ns->pidmap[i].page = 0; | ||
569 | atomic_set(&ns->pidmap[i].nr_free, BITS_PER_PAGE); | ||
570 | } | ||
571 | |||
572 | return ns; | ||
573 | |||
574 | out_free_map: | ||
575 | kfree(ns->pidmap[0].page); | ||
576 | out_free: | ||
577 | kmem_cache_free(pid_ns_cachep, ns); | ||
578 | out: | ||
579 | return ERR_PTR(-ENOMEM); | ||
580 | } | ||
581 | |||
582 | static void destroy_pid_namespace(struct pid_namespace *ns) | ||
583 | { | ||
584 | int i; | ||
585 | |||
586 | for (i = 0; i < PIDMAP_ENTRIES; i++) | ||
587 | kfree(ns->pidmap[i].page); | ||
588 | kmem_cache_free(pid_ns_cachep, ns); | ||
589 | } | ||
590 | |||
591 | struct pid_namespace *copy_pid_ns(unsigned long flags, struct pid_namespace *old_ns) | ||
592 | { | ||
593 | struct pid_namespace *new_ns; | ||
594 | |||
595 | BUG_ON(!old_ns); | ||
596 | new_ns = get_pid_ns(old_ns); | ||
597 | if (!(flags & CLONE_NEWPID)) | ||
598 | goto out; | ||
599 | |||
600 | new_ns = ERR_PTR(-EINVAL); | ||
601 | if (flags & CLONE_THREAD) | ||
602 | goto out_put; | ||
603 | |||
604 | new_ns = create_pid_namespace(old_ns->level + 1); | ||
605 | if (!IS_ERR(new_ns)) | ||
606 | new_ns->parent = get_pid_ns(old_ns); | ||
607 | |||
608 | out_put: | ||
609 | put_pid_ns(old_ns); | ||
610 | out: | ||
611 | return new_ns; | ||
612 | } | ||
613 | |||
614 | void free_pid_ns(struct kref *kref) | ||
615 | { | ||
616 | struct pid_namespace *ns, *parent; | ||
617 | |||
618 | ns = container_of(kref, struct pid_namespace, kref); | ||
619 | |||
620 | parent = ns->parent; | ||
621 | destroy_pid_namespace(ns); | ||
622 | |||
623 | if (parent != NULL) | ||
624 | put_pid_ns(parent); | ||
625 | } | ||
626 | #endif /* CONFIG_PID_NS */ | ||
627 | |||
628 | void zap_pid_ns_processes(struct pid_namespace *pid_ns) | ||
629 | { | ||
630 | int nr; | ||
631 | int rc; | ||
632 | |||
633 | /* | ||
634 | * The last thread in the cgroup-init thread group is terminating. | ||
635 | * Find remaining pid_ts in the namespace, signal and wait for them | ||
636 | * to exit. | ||
637 | * | ||
638 | * Note: This signals each threads in the namespace - even those that | ||
639 | * belong to the same thread group, To avoid this, we would have | ||
640 | * to walk the entire tasklist looking a processes in this | ||
641 | * namespace, but that could be unnecessarily expensive if the | ||
642 | * pid namespace has just a few processes. Or we need to | ||
643 | * maintain a tasklist for each pid namespace. | ||
644 | * | ||
645 | */ | ||
646 | read_lock(&tasklist_lock); | ||
647 | nr = next_pidmap(pid_ns, 1); | ||
648 | while (nr > 0) { | ||
649 | kill_proc_info(SIGKILL, SEND_SIG_PRIV, nr); | ||
650 | nr = next_pidmap(pid_ns, nr); | ||
651 | } | ||
652 | read_unlock(&tasklist_lock); | ||
653 | |||
654 | do { | ||
655 | clear_thread_flag(TIF_SIGPENDING); | ||
656 | rc = sys_wait4(-1, NULL, __WALL, NULL); | ||
657 | } while (rc != -ECHILD); | ||
658 | |||
659 | |||
660 | /* Child reaper for the pid namespace is going away */ | ||
661 | pid_ns->child_reaper = NULL; | ||
662 | return; | ||
663 | } | ||
664 | |||
665 | /* | 496 | /* |
666 | * The pid hash table is scaled according to the amount of memory in the | 497 | * The pid hash table is scaled according to the amount of memory in the |
667 | * machine. From a minimum of 16 slots up to 4096 slots at one gigabyte or | 498 | * machine. From a minimum of 16 slots up to 4096 slots at one gigabyte or |
@@ -694,9 +525,6 @@ void __init pidmap_init(void) | |||
694 | set_bit(0, init_pid_ns.pidmap[0].page); | 525 | set_bit(0, init_pid_ns.pidmap[0].page); |
695 | atomic_dec(&init_pid_ns.pidmap[0].nr_free); | 526 | atomic_dec(&init_pid_ns.pidmap[0].nr_free); |
696 | 527 | ||
697 | init_pid_ns.pid_cachep = create_pid_cachep(1); | 528 | init_pid_ns.pid_cachep = KMEM_CACHE(pid, |
698 | if (init_pid_ns.pid_cachep == NULL) | 529 | SLAB_HWCACHE_ALIGN | SLAB_PANIC); |
699 | panic("Can't create pid_1 cachep\n"); | ||
700 | |||
701 | pid_ns_cachep = KMEM_CACHE(pid_namespace, SLAB_PANIC); | ||
702 | } | 530 | } |