Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * linux/kernel/fork.c | |
3 | * | |
4 | * Copyright (C) 1991, 1992 Linus Torvalds | |
5 | */ | |
6 | ||
7 | /* | |
8 | * 'fork.c' contains the help-routines for the 'fork' system call | |
9 | * (see also entry.S and others). | |
10 | * Fork is rather simple, once you get the hang of it, but the memory | |
11 | * management can be a bitch. See 'mm/memory.c': 'copy_page_range()' | |
12 | */ | |
13 | ||
1da177e4 LT |
14 | #include <linux/slab.h> |
15 | #include <linux/init.h> | |
16 | #include <linux/unistd.h> | |
1da177e4 LT |
17 | #include <linux/module.h> |
18 | #include <linux/vmalloc.h> | |
19 | #include <linux/completion.h> | |
6b3286ed | 20 | #include <linux/mnt_namespace.h> |
1da177e4 LT |
21 | #include <linux/personality.h> |
22 | #include <linux/mempolicy.h> | |
23 | #include <linux/sem.h> | |
24 | #include <linux/file.h> | |
25 | #include <linux/key.h> | |
26 | #include <linux/binfmts.h> | |
27 | #include <linux/mman.h> | |
28 | #include <linux/fs.h> | |
ab516013 | 29 | #include <linux/nsproxy.h> |
c59ede7b | 30 | #include <linux/capability.h> |
1da177e4 | 31 | #include <linux/cpu.h> |
b4f48b63 | 32 | #include <linux/cgroup.h> |
1da177e4 LT |
33 | #include <linux/security.h> |
34 | #include <linux/swap.h> | |
35 | #include <linux/syscalls.h> | |
36 | #include <linux/jiffies.h> | |
37 | #include <linux/futex.h> | |
7c3ab738 | 38 | #include <linux/task_io_accounting_ops.h> |
ab2af1f5 | 39 | #include <linux/rcupdate.h> |
1da177e4 LT |
40 | #include <linux/ptrace.h> |
41 | #include <linux/mount.h> | |
42 | #include <linux/audit.h> | |
43 | #include <linux/profile.h> | |
44 | #include <linux/rmap.h> | |
45 | #include <linux/acct.h> | |
8f0ab514 | 46 | #include <linux/tsacct_kern.h> |
9f46080c | 47 | #include <linux/cn_proc.h> |
ba96a0c8 | 48 | #include <linux/freezer.h> |
ca74e92b | 49 | #include <linux/delayacct.h> |
ad4ecbcb | 50 | #include <linux/taskstats_kern.h> |
0a425405 | 51 | #include <linux/random.h> |
522ed776 | 52 | #include <linux/tty.h> |
6f4e6433 | 53 | #include <linux/proc_fs.h> |
fd0928df | 54 | #include <linux/blkdev.h> |
1da177e4 LT |
55 | |
56 | #include <asm/pgtable.h> | |
57 | #include <asm/pgalloc.h> | |
58 | #include <asm/uaccess.h> | |
59 | #include <asm/mmu_context.h> | |
60 | #include <asm/cacheflush.h> | |
61 | #include <asm/tlbflush.h> | |
62 | ||
63 | /* | |
64 | * Protected counters by write_lock_irq(&tasklist_lock) | |
65 | */ | |
66 | unsigned long total_forks; /* Handle normal Linux uptimes. */ | |
67 | int nr_threads; /* The idle threads do not count.. */ | |
68 | ||
69 | int max_threads; /* tunable limit on nr_threads */ | |
70 | ||
71 | DEFINE_PER_CPU(unsigned long, process_counts) = 0; | |
72 | ||
c59923a1 | 73 | __cacheline_aligned DEFINE_RWLOCK(tasklist_lock); /* outer */ |
1da177e4 LT |
74 | |
75 | int nr_processes(void) | |
76 | { | |
77 | int cpu; | |
78 | int total = 0; | |
79 | ||
80 | for_each_online_cpu(cpu) | |
81 | total += per_cpu(process_counts, cpu); | |
82 | ||
83 | return total; | |
84 | } | |
85 | ||
86 | #ifndef __HAVE_ARCH_TASK_STRUCT_ALLOCATOR | |
87 | # define alloc_task_struct() kmem_cache_alloc(task_struct_cachep, GFP_KERNEL) | |
88 | # define free_task_struct(tsk) kmem_cache_free(task_struct_cachep, (tsk)) | |
e18b890b | 89 | static struct kmem_cache *task_struct_cachep; |
1da177e4 LT |
90 | #endif |
91 | ||
92 | /* SLAB cache for signal_struct structures (tsk->signal) */ | |
e18b890b | 93 | static struct kmem_cache *signal_cachep; |
1da177e4 LT |
94 | |
95 | /* SLAB cache for sighand_struct structures (tsk->sighand) */ | |
e18b890b | 96 | struct kmem_cache *sighand_cachep; |
1da177e4 LT |
97 | |
98 | /* SLAB cache for files_struct structures (tsk->files) */ | |
e18b890b | 99 | struct kmem_cache *files_cachep; |
1da177e4 LT |
100 | |
101 | /* SLAB cache for fs_struct structures (tsk->fs) */ | |
e18b890b | 102 | struct kmem_cache *fs_cachep; |
1da177e4 LT |
103 | |
104 | /* SLAB cache for vm_area_struct structures */ | |
e18b890b | 105 | struct kmem_cache *vm_area_cachep; |
1da177e4 LT |
106 | |
107 | /* SLAB cache for mm_struct structures (tsk->mm) */ | |
e18b890b | 108 | static struct kmem_cache *mm_cachep; |
1da177e4 LT |
109 | |
110 | void free_task(struct task_struct *tsk) | |
111 | { | |
3e26c149 | 112 | prop_local_destroy_single(&tsk->dirties); |
f7e4217b | 113 | free_thread_info(tsk->stack); |
23f78d4a | 114 | rt_mutex_debug_task_free(tsk); |
1da177e4 LT |
115 | free_task_struct(tsk); |
116 | } | |
117 | EXPORT_SYMBOL(free_task); | |
118 | ||
158d9ebd | 119 | void __put_task_struct(struct task_struct *tsk) |
1da177e4 | 120 | { |
270f722d | 121 | WARN_ON(!tsk->exit_state); |
1da177e4 LT |
122 | WARN_ON(atomic_read(&tsk->usage)); |
123 | WARN_ON(tsk == current); | |
124 | ||
1da177e4 LT |
125 | security_task_free(tsk); |
126 | free_uid(tsk->user); | |
127 | put_group_info(tsk->group_info); | |
35df17c5 | 128 | delayacct_tsk_free(tsk); |
1da177e4 LT |
129 | |
130 | if (!profile_handoff_task(tsk)) | |
131 | free_task(tsk); | |
132 | } | |
133 | ||
134 | void __init fork_init(unsigned long mempages) | |
135 | { | |
136 | #ifndef __HAVE_ARCH_TASK_STRUCT_ALLOCATOR | |
137 | #ifndef ARCH_MIN_TASKALIGN | |
138 | #define ARCH_MIN_TASKALIGN L1_CACHE_BYTES | |
139 | #endif | |
140 | /* create a slab on which task_structs can be allocated */ | |
141 | task_struct_cachep = | |
142 | kmem_cache_create("task_struct", sizeof(struct task_struct), | |
20c2df83 | 143 | ARCH_MIN_TASKALIGN, SLAB_PANIC, NULL); |
1da177e4 LT |
144 | #endif |
145 | ||
146 | /* | |
147 | * The default maximum number of threads is set to a safe | |
148 | * value: the thread structures can take up at most half | |
149 | * of memory. | |
150 | */ | |
151 | max_threads = mempages / (8 * THREAD_SIZE / PAGE_SIZE); | |
152 | ||
153 | /* | |
154 | * we need to allow at least 20 threads to boot a system | |
155 | */ | |
156 | if(max_threads < 20) | |
157 | max_threads = 20; | |
158 | ||
159 | init_task.signal->rlim[RLIMIT_NPROC].rlim_cur = max_threads/2; | |
160 | init_task.signal->rlim[RLIMIT_NPROC].rlim_max = max_threads/2; | |
161 | init_task.signal->rlim[RLIMIT_SIGPENDING] = | |
162 | init_task.signal->rlim[RLIMIT_NPROC]; | |
163 | } | |
164 | ||
165 | static struct task_struct *dup_task_struct(struct task_struct *orig) | |
166 | { | |
167 | struct task_struct *tsk; | |
168 | struct thread_info *ti; | |
3e26c149 | 169 | int err; |
1da177e4 LT |
170 | |
171 | prepare_to_copy(orig); | |
172 | ||
173 | tsk = alloc_task_struct(); | |
174 | if (!tsk) | |
175 | return NULL; | |
176 | ||
177 | ti = alloc_thread_info(tsk); | |
178 | if (!ti) { | |
179 | free_task_struct(tsk); | |
180 | return NULL; | |
181 | } | |
182 | ||
1da177e4 | 183 | *tsk = *orig; |
f7e4217b | 184 | tsk->stack = ti; |
3e26c149 PZ |
185 | |
186 | err = prop_local_init_single(&tsk->dirties); | |
187 | if (err) { | |
188 | free_thread_info(ti); | |
189 | free_task_struct(tsk); | |
190 | return NULL; | |
191 | } | |
192 | ||
10ebffde | 193 | setup_thread_stack(tsk, orig); |
1da177e4 | 194 | |
0a425405 AV |
195 | #ifdef CONFIG_CC_STACKPROTECTOR |
196 | tsk->stack_canary = get_random_int(); | |
197 | #endif | |
198 | ||
1da177e4 LT |
199 | /* One for us, one for whoever does the "release_task()" (usually parent) */ |
200 | atomic_set(&tsk->usage,2); | |
4b5d37ac | 201 | atomic_set(&tsk->fs_excl, 0); |
6c5c9341 | 202 | #ifdef CONFIG_BLK_DEV_IO_TRACE |
2056a782 | 203 | tsk->btrace_seq = 0; |
6c5c9341 | 204 | #endif |
a0aa7f68 | 205 | tsk->splice_pipe = NULL; |
1da177e4 LT |
206 | return tsk; |
207 | } | |
208 | ||
209 | #ifdef CONFIG_MMU | |
a39bc516 | 210 | static int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm) |
1da177e4 | 211 | { |
fd3e42fc | 212 | struct vm_area_struct *mpnt, *tmp, **pprev; |
1da177e4 LT |
213 | struct rb_node **rb_link, *rb_parent; |
214 | int retval; | |
215 | unsigned long charge; | |
216 | struct mempolicy *pol; | |
217 | ||
218 | down_write(&oldmm->mmap_sem); | |
ec8c0446 | 219 | flush_cache_dup_mm(oldmm); |
ad339451 IM |
220 | /* |
221 | * Not linked in yet - no deadlock potential: | |
222 | */ | |
223 | down_write_nested(&mm->mmap_sem, SINGLE_DEPTH_NESTING); | |
7ee78232 | 224 | |
1da177e4 LT |
225 | mm->locked_vm = 0; |
226 | mm->mmap = NULL; | |
227 | mm->mmap_cache = NULL; | |
228 | mm->free_area_cache = oldmm->mmap_base; | |
1363c3cd | 229 | mm->cached_hole_size = ~0UL; |
1da177e4 | 230 | mm->map_count = 0; |
1da177e4 LT |
231 | cpus_clear(mm->cpu_vm_mask); |
232 | mm->mm_rb = RB_ROOT; | |
233 | rb_link = &mm->mm_rb.rb_node; | |
234 | rb_parent = NULL; | |
235 | pprev = &mm->mmap; | |
236 | ||
fd3e42fc | 237 | for (mpnt = oldmm->mmap; mpnt; mpnt = mpnt->vm_next) { |
1da177e4 LT |
238 | struct file *file; |
239 | ||
240 | if (mpnt->vm_flags & VM_DONTCOPY) { | |
3b6bfcdb HD |
241 | long pages = vma_pages(mpnt); |
242 | mm->total_vm -= pages; | |
ab50b8ed | 243 | vm_stat_account(mm, mpnt->vm_flags, mpnt->vm_file, |
3b6bfcdb | 244 | -pages); |
1da177e4 LT |
245 | continue; |
246 | } | |
247 | charge = 0; | |
248 | if (mpnt->vm_flags & VM_ACCOUNT) { | |
249 | unsigned int len = (mpnt->vm_end - mpnt->vm_start) >> PAGE_SHIFT; | |
250 | if (security_vm_enough_memory(len)) | |
251 | goto fail_nomem; | |
252 | charge = len; | |
253 | } | |
e94b1766 | 254 | tmp = kmem_cache_alloc(vm_area_cachep, GFP_KERNEL); |
1da177e4 LT |
255 | if (!tmp) |
256 | goto fail_nomem; | |
257 | *tmp = *mpnt; | |
258 | pol = mpol_copy(vma_policy(mpnt)); | |
259 | retval = PTR_ERR(pol); | |
260 | if (IS_ERR(pol)) | |
261 | goto fail_nomem_policy; | |
262 | vma_set_policy(tmp, pol); | |
263 | tmp->vm_flags &= ~VM_LOCKED; | |
264 | tmp->vm_mm = mm; | |
265 | tmp->vm_next = NULL; | |
266 | anon_vma_link(tmp); | |
267 | file = tmp->vm_file; | |
268 | if (file) { | |
f3a43f3f | 269 | struct inode *inode = file->f_path.dentry->d_inode; |
1da177e4 LT |
270 | get_file(file); |
271 | if (tmp->vm_flags & VM_DENYWRITE) | |
272 | atomic_dec(&inode->i_writecount); | |
23ff4440 | 273 | |
1da177e4 LT |
274 | /* insert tmp into the share list, just after mpnt */ |
275 | spin_lock(&file->f_mapping->i_mmap_lock); | |
276 | tmp->vm_truncate_count = mpnt->vm_truncate_count; | |
277 | flush_dcache_mmap_lock(file->f_mapping); | |
278 | vma_prio_tree_add(tmp, mpnt); | |
279 | flush_dcache_mmap_unlock(file->f_mapping); | |
280 | spin_unlock(&file->f_mapping->i_mmap_lock); | |
281 | } | |
282 | ||
283 | /* | |
7ee78232 | 284 | * Link in the new vma and copy the page table entries. |
1da177e4 | 285 | */ |
1da177e4 LT |
286 | *pprev = tmp; |
287 | pprev = &tmp->vm_next; | |
288 | ||
289 | __vma_link_rb(mm, tmp, rb_link, rb_parent); | |
290 | rb_link = &tmp->vm_rb.rb_right; | |
291 | rb_parent = &tmp->vm_rb; | |
292 | ||
293 | mm->map_count++; | |
0b0db14c | 294 | retval = copy_page_range(mm, oldmm, mpnt); |
1da177e4 LT |
295 | |
296 | if (tmp->vm_ops && tmp->vm_ops->open) | |
297 | tmp->vm_ops->open(tmp); | |
298 | ||
299 | if (retval) | |
300 | goto out; | |
301 | } | |
d6dd61c8 JF |
302 | /* a new mm has just been created */ |
303 | arch_dup_mmap(oldmm, mm); | |
1da177e4 | 304 | retval = 0; |
1da177e4 | 305 | out: |
7ee78232 | 306 | up_write(&mm->mmap_sem); |
fd3e42fc | 307 | flush_tlb_mm(oldmm); |
1da177e4 LT |
308 | up_write(&oldmm->mmap_sem); |
309 | return retval; | |
310 | fail_nomem_policy: | |
311 | kmem_cache_free(vm_area_cachep, tmp); | |
312 | fail_nomem: | |
313 | retval = -ENOMEM; | |
314 | vm_unacct_memory(charge); | |
315 | goto out; | |
316 | } | |
317 | ||
318 | static inline int mm_alloc_pgd(struct mm_struct * mm) | |
319 | { | |
320 | mm->pgd = pgd_alloc(mm); | |
321 | if (unlikely(!mm->pgd)) | |
322 | return -ENOMEM; | |
323 | return 0; | |
324 | } | |
325 | ||
326 | static inline void mm_free_pgd(struct mm_struct * mm) | |
327 | { | |
328 | pgd_free(mm->pgd); | |
329 | } | |
330 | #else | |
331 | #define dup_mmap(mm, oldmm) (0) | |
332 | #define mm_alloc_pgd(mm) (0) | |
333 | #define mm_free_pgd(mm) | |
334 | #endif /* CONFIG_MMU */ | |
335 | ||
23ff4440 | 336 | __cacheline_aligned_in_smp DEFINE_SPINLOCK(mmlist_lock); |
1da177e4 | 337 | |
e94b1766 | 338 | #define allocate_mm() (kmem_cache_alloc(mm_cachep, GFP_KERNEL)) |
1da177e4 LT |
339 | #define free_mm(mm) (kmem_cache_free(mm_cachep, (mm))) |
340 | ||
341 | #include <linux/init_task.h> | |
342 | ||
343 | static struct mm_struct * mm_init(struct mm_struct * mm) | |
344 | { | |
345 | atomic_set(&mm->mm_users, 1); | |
346 | atomic_set(&mm->mm_count, 1); | |
347 | init_rwsem(&mm->mmap_sem); | |
348 | INIT_LIST_HEAD(&mm->mmlist); | |
3cb4a0bb KH |
349 | mm->flags = (current->mm) ? current->mm->flags |
350 | : MMF_DUMP_FILTER_DEFAULT; | |
1da177e4 LT |
351 | mm->core_waiters = 0; |
352 | mm->nr_ptes = 0; | |
4294621f | 353 | set_mm_counter(mm, file_rss, 0); |
404351e6 | 354 | set_mm_counter(mm, anon_rss, 0); |
1da177e4 LT |
355 | spin_lock_init(&mm->page_table_lock); |
356 | rwlock_init(&mm->ioctx_list_lock); | |
357 | mm->ioctx_list = NULL; | |
1da177e4 | 358 | mm->free_area_cache = TASK_UNMAPPED_BASE; |
1363c3cd | 359 | mm->cached_hole_size = ~0UL; |
1da177e4 LT |
360 | |
361 | if (likely(!mm_alloc_pgd(mm))) { | |
362 | mm->def_flags = 0; | |
363 | return mm; | |
364 | } | |
365 | free_mm(mm); | |
366 | return NULL; | |
367 | } | |
368 | ||
369 | /* | |
370 | * Allocate and initialize an mm_struct. | |
371 | */ | |
372 | struct mm_struct * mm_alloc(void) | |
373 | { | |
374 | struct mm_struct * mm; | |
375 | ||
376 | mm = allocate_mm(); | |
377 | if (mm) { | |
378 | memset(mm, 0, sizeof(*mm)); | |
379 | mm = mm_init(mm); | |
380 | } | |
381 | return mm; | |
382 | } | |
383 | ||
384 | /* | |
385 | * Called when the last reference to the mm | |
386 | * is dropped: either by a lazy thread or by | |
387 | * mmput. Free the page directory and the mm. | |
388 | */ | |
389 | void fastcall __mmdrop(struct mm_struct *mm) | |
390 | { | |
391 | BUG_ON(mm == &init_mm); | |
392 | mm_free_pgd(mm); | |
393 | destroy_context(mm); | |
394 | free_mm(mm); | |
395 | } | |
6d4e4c4f | 396 | EXPORT_SYMBOL_GPL(__mmdrop); |
1da177e4 LT |
397 | |
398 | /* | |
399 | * Decrement the use count and release all resources for an mm. | |
400 | */ | |
401 | void mmput(struct mm_struct *mm) | |
402 | { | |
0ae26f1b AM |
403 | might_sleep(); |
404 | ||
1da177e4 LT |
405 | if (atomic_dec_and_test(&mm->mm_users)) { |
406 | exit_aio(mm); | |
407 | exit_mmap(mm); | |
408 | if (!list_empty(&mm->mmlist)) { | |
409 | spin_lock(&mmlist_lock); | |
410 | list_del(&mm->mmlist); | |
411 | spin_unlock(&mmlist_lock); | |
412 | } | |
413 | put_swap_token(mm); | |
414 | mmdrop(mm); | |
415 | } | |
416 | } | |
417 | EXPORT_SYMBOL_GPL(mmput); | |
418 | ||
419 | /** | |
420 | * get_task_mm - acquire a reference to the task's mm | |
421 | * | |
422 | * Returns %NULL if the task has no mm. Checks PF_BORROWED_MM (meaning | |
423 | * this kernel workthread has transiently adopted a user mm with use_mm, | |
424 | * to do its AIO) is not set and if so returns a reference to it, after | |
425 | * bumping up the use count. User must release the mm via mmput() | |
426 | * after use. Typically used by /proc and ptrace. | |
427 | */ | |
428 | struct mm_struct *get_task_mm(struct task_struct *task) | |
429 | { | |
430 | struct mm_struct *mm; | |
431 | ||
432 | task_lock(task); | |
433 | mm = task->mm; | |
434 | if (mm) { | |
435 | if (task->flags & PF_BORROWED_MM) | |
436 | mm = NULL; | |
437 | else | |
438 | atomic_inc(&mm->mm_users); | |
439 | } | |
440 | task_unlock(task); | |
441 | return mm; | |
442 | } | |
443 | EXPORT_SYMBOL_GPL(get_task_mm); | |
444 | ||
445 | /* Please note the differences between mmput and mm_release. | |
446 | * mmput is called whenever we stop holding onto a mm_struct, | |
447 | * error success whatever. | |
448 | * | |
449 | * mm_release is called after a mm_struct has been removed | |
450 | * from the current process. | |
451 | * | |
452 | * This difference is important for error handling, when we | |
453 | * only half set up a mm_struct for a new process and need to restore | |
454 | * the old one. Because we mmput the new mm_struct before | |
455 | * restoring the old one. . . | |
456 | * Eric Biederman 10 January 1998 | |
457 | */ | |
458 | void mm_release(struct task_struct *tsk, struct mm_struct *mm) | |
459 | { | |
460 | struct completion *vfork_done = tsk->vfork_done; | |
461 | ||
462 | /* Get rid of any cached register state */ | |
463 | deactivate_mm(tsk, mm); | |
464 | ||
465 | /* notify parent sleeping on vfork() */ | |
466 | if (vfork_done) { | |
467 | tsk->vfork_done = NULL; | |
468 | complete(vfork_done); | |
469 | } | |
fec1d011 RM |
470 | |
471 | /* | |
472 | * If we're exiting normally, clear a user-space tid field if | |
473 | * requested. We leave this alone when dying by signal, to leave | |
474 | * the value intact in a core dump, and to save the unnecessary | |
475 | * trouble otherwise. Userland only wants this done for a sys_exit. | |
476 | */ | |
477 | if (tsk->clear_child_tid | |
478 | && !(tsk->flags & PF_SIGNALED) | |
479 | && atomic_read(&mm->mm_users) > 1) { | |
1da177e4 LT |
480 | u32 __user * tidptr = tsk->clear_child_tid; |
481 | tsk->clear_child_tid = NULL; | |
482 | ||
483 | /* | |
484 | * We don't check the error code - if userspace has | |
485 | * not set up a proper pointer then tough luck. | |
486 | */ | |
487 | put_user(0, tidptr); | |
488 | sys_futex(tidptr, FUTEX_WAKE, 1, NULL, NULL, 0); | |
489 | } | |
490 | } | |
491 | ||
a0a7ec30 JD |
492 | /* |
493 | * Allocate a new mm structure and copy contents from the | |
494 | * mm structure of the passed in task structure. | |
495 | */ | |
496 | static struct mm_struct *dup_mm(struct task_struct *tsk) | |
497 | { | |
498 | struct mm_struct *mm, *oldmm = current->mm; | |
499 | int err; | |
500 | ||
501 | if (!oldmm) | |
502 | return NULL; | |
503 | ||
504 | mm = allocate_mm(); | |
505 | if (!mm) | |
506 | goto fail_nomem; | |
507 | ||
508 | memcpy(mm, oldmm, sizeof(*mm)); | |
509 | ||
7602bdf2 AC |
510 | /* Initializing for Swap token stuff */ |
511 | mm->token_priority = 0; | |
512 | mm->last_interval = 0; | |
513 | ||
a0a7ec30 JD |
514 | if (!mm_init(mm)) |
515 | goto fail_nomem; | |
516 | ||
517 | if (init_new_context(tsk, mm)) | |
518 | goto fail_nocontext; | |
519 | ||
520 | err = dup_mmap(mm, oldmm); | |
521 | if (err) | |
522 | goto free_pt; | |
523 | ||
524 | mm->hiwater_rss = get_mm_rss(mm); | |
525 | mm->hiwater_vm = mm->total_vm; | |
526 | ||
527 | return mm; | |
528 | ||
529 | free_pt: | |
530 | mmput(mm); | |
531 | ||
532 | fail_nomem: | |
533 | return NULL; | |
534 | ||
535 | fail_nocontext: | |
536 | /* | |
537 | * If init_new_context() failed, we cannot use mmput() to free the mm | |
538 | * because it calls destroy_context() | |
539 | */ | |
540 | mm_free_pgd(mm); | |
541 | free_mm(mm); | |
542 | return NULL; | |
543 | } | |
544 | ||
1da177e4 LT |
545 | static int copy_mm(unsigned long clone_flags, struct task_struct * tsk) |
546 | { | |
547 | struct mm_struct * mm, *oldmm; | |
548 | int retval; | |
549 | ||
550 | tsk->min_flt = tsk->maj_flt = 0; | |
551 | tsk->nvcsw = tsk->nivcsw = 0; | |
552 | ||
553 | tsk->mm = NULL; | |
554 | tsk->active_mm = NULL; | |
555 | ||
556 | /* | |
557 | * Are we cloning a kernel thread? | |
558 | * | |
559 | * We need to steal a active VM for that.. | |
560 | */ | |
561 | oldmm = current->mm; | |
562 | if (!oldmm) | |
563 | return 0; | |
564 | ||
565 | if (clone_flags & CLONE_VM) { | |
566 | atomic_inc(&oldmm->mm_users); | |
567 | mm = oldmm; | |
1da177e4 LT |
568 | goto good_mm; |
569 | } | |
570 | ||
571 | retval = -ENOMEM; | |
a0a7ec30 | 572 | mm = dup_mm(tsk); |
1da177e4 LT |
573 | if (!mm) |
574 | goto fail_nomem; | |
575 | ||
1da177e4 | 576 | good_mm: |
7602bdf2 AC |
577 | /* Initializing for Swap token stuff */ |
578 | mm->token_priority = 0; | |
579 | mm->last_interval = 0; | |
580 | ||
1da177e4 LT |
581 | tsk->mm = mm; |
582 | tsk->active_mm = mm; | |
583 | return 0; | |
584 | ||
1da177e4 LT |
585 | fail_nomem: |
586 | return retval; | |
1da177e4 LT |
587 | } |
588 | ||
a39bc516 | 589 | static struct fs_struct *__copy_fs_struct(struct fs_struct *old) |
1da177e4 LT |
590 | { |
591 | struct fs_struct *fs = kmem_cache_alloc(fs_cachep, GFP_KERNEL); | |
592 | /* We don't need to lock fs - think why ;-) */ | |
593 | if (fs) { | |
594 | atomic_set(&fs->count, 1); | |
595 | rwlock_init(&fs->lock); | |
596 | fs->umask = old->umask; | |
597 | read_lock(&old->lock); | |
598 | fs->rootmnt = mntget(old->rootmnt); | |
599 | fs->root = dget(old->root); | |
600 | fs->pwdmnt = mntget(old->pwdmnt); | |
601 | fs->pwd = dget(old->pwd); | |
602 | if (old->altroot) { | |
603 | fs->altrootmnt = mntget(old->altrootmnt); | |
604 | fs->altroot = dget(old->altroot); | |
605 | } else { | |
606 | fs->altrootmnt = NULL; | |
607 | fs->altroot = NULL; | |
608 | } | |
609 | read_unlock(&old->lock); | |
610 | } | |
611 | return fs; | |
612 | } | |
613 | ||
614 | struct fs_struct *copy_fs_struct(struct fs_struct *old) | |
615 | { | |
616 | return __copy_fs_struct(old); | |
617 | } | |
618 | ||
619 | EXPORT_SYMBOL_GPL(copy_fs_struct); | |
620 | ||
a39bc516 | 621 | static int copy_fs(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 LT |
622 | { |
623 | if (clone_flags & CLONE_FS) { | |
624 | atomic_inc(¤t->fs->count); | |
625 | return 0; | |
626 | } | |
627 | tsk->fs = __copy_fs_struct(current->fs); | |
628 | if (!tsk->fs) | |
629 | return -ENOMEM; | |
630 | return 0; | |
631 | } | |
632 | ||
ab2af1f5 | 633 | static int count_open_files(struct fdtable *fdt) |
1da177e4 | 634 | { |
bbea9f69 | 635 | int size = fdt->max_fds; |
1da177e4 LT |
636 | int i; |
637 | ||
638 | /* Find the last open fd */ | |
639 | for (i = size/(8*sizeof(long)); i > 0; ) { | |
badf1662 | 640 | if (fdt->open_fds->fds_bits[--i]) |
1da177e4 LT |
641 | break; |
642 | } | |
643 | i = (i+1) * 8 * sizeof(long); | |
644 | return i; | |
645 | } | |
646 | ||
badf1662 DS |
647 | static struct files_struct *alloc_files(void) |
648 | { | |
649 | struct files_struct *newf; | |
650 | struct fdtable *fdt; | |
651 | ||
e94b1766 | 652 | newf = kmem_cache_alloc(files_cachep, GFP_KERNEL); |
badf1662 DS |
653 | if (!newf) |
654 | goto out; | |
655 | ||
656 | atomic_set(&newf->count, 1); | |
657 | ||
658 | spin_lock_init(&newf->file_lock); | |
0c9e63fd | 659 | newf->next_fd = 0; |
ab2af1f5 | 660 | fdt = &newf->fdtab; |
badf1662 | 661 | fdt->max_fds = NR_OPEN_DEFAULT; |
0c9e63fd ED |
662 | fdt->close_on_exec = (fd_set *)&newf->close_on_exec_init; |
663 | fdt->open_fds = (fd_set *)&newf->open_fds_init; | |
badf1662 | 664 | fdt->fd = &newf->fd_array[0]; |
ab2af1f5 | 665 | INIT_RCU_HEAD(&fdt->rcu); |
ab2af1f5 DS |
666 | fdt->next = NULL; |
667 | rcu_assign_pointer(newf->fdt, fdt); | |
badf1662 DS |
668 | out: |
669 | return newf; | |
670 | } | |
671 | ||
a016f338 JD |
672 | /* |
673 | * Allocate a new files structure and copy contents from the | |
674 | * passed in files structure. | |
6e667260 | 675 | * errorp will be valid only when the returned files_struct is NULL. |
a016f338 JD |
676 | */ |
677 | static struct files_struct *dup_fd(struct files_struct *oldf, int *errorp) | |
1da177e4 | 678 | { |
a016f338 | 679 | struct files_struct *newf; |
1da177e4 | 680 | struct file **old_fds, **new_fds; |
bbea9f69 | 681 | int open_files, size, i; |
badf1662 | 682 | struct fdtable *old_fdt, *new_fdt; |
1da177e4 | 683 | |
6e667260 | 684 | *errorp = -ENOMEM; |
badf1662 DS |
685 | newf = alloc_files(); |
686 | if (!newf) | |
1da177e4 LT |
687 | goto out; |
688 | ||
1da177e4 | 689 | spin_lock(&oldf->file_lock); |
badf1662 DS |
690 | old_fdt = files_fdtable(oldf); |
691 | new_fdt = files_fdtable(newf); | |
ab2af1f5 | 692 | open_files = count_open_files(old_fdt); |
1da177e4 LT |
693 | |
694 | /* | |
bbea9f69 VL |
695 | * Check whether we need to allocate a larger fd array and fd set. |
696 | * Note: we're not a clone task, so the open count won't change. | |
1da177e4 | 697 | */ |
badf1662 DS |
698 | if (open_files > new_fdt->max_fds) { |
699 | new_fdt->max_fds = 0; | |
1da177e4 LT |
700 | spin_unlock(&oldf->file_lock); |
701 | spin_lock(&newf->file_lock); | |
a016f338 | 702 | *errorp = expand_files(newf, open_files-1); |
1da177e4 | 703 | spin_unlock(&newf->file_lock); |
a016f338 | 704 | if (*errorp < 0) |
1da177e4 | 705 | goto out_release; |
ab2af1f5 DS |
706 | new_fdt = files_fdtable(newf); |
707 | /* | |
708 | * Reacquire the oldf lock and a pointer to its fd table | |
709 | * who knows it may have a new bigger fd table. We need | |
710 | * the latest pointer. | |
711 | */ | |
1da177e4 | 712 | spin_lock(&oldf->file_lock); |
ab2af1f5 | 713 | old_fdt = files_fdtable(oldf); |
1da177e4 LT |
714 | } |
715 | ||
badf1662 DS |
716 | old_fds = old_fdt->fd; |
717 | new_fds = new_fdt->fd; | |
1da177e4 | 718 | |
f3d19c90 VL |
719 | memcpy(new_fdt->open_fds->fds_bits, |
720 | old_fdt->open_fds->fds_bits, open_files/8); | |
721 | memcpy(new_fdt->close_on_exec->fds_bits, | |
722 | old_fdt->close_on_exec->fds_bits, open_files/8); | |
1da177e4 LT |
723 | |
724 | for (i = open_files; i != 0; i--) { | |
725 | struct file *f = *old_fds++; | |
726 | if (f) { | |
727 | get_file(f); | |
728 | } else { | |
729 | /* | |
730 | * The fd may be claimed in the fd bitmap but not yet | |
731 | * instantiated in the files array if a sibling thread | |
732 | * is partway through open(). So make sure that this | |
733 | * fd is available to the new process. | |
734 | */ | |
badf1662 | 735 | FD_CLR(open_files - i, new_fdt->open_fds); |
1da177e4 | 736 | } |
ab2af1f5 | 737 | rcu_assign_pointer(*new_fds++, f); |
1da177e4 LT |
738 | } |
739 | spin_unlock(&oldf->file_lock); | |
740 | ||
741 | /* compute the remainder to be cleared */ | |
badf1662 | 742 | size = (new_fdt->max_fds - open_files) * sizeof(struct file *); |
1da177e4 | 743 | |
23ff4440 DW |
744 | /* This is long word aligned thus could use a optimized version */ |
745 | memset(new_fds, 0, size); | |
1da177e4 | 746 | |
bbea9f69 VL |
747 | if (new_fdt->max_fds > open_files) { |
748 | int left = (new_fdt->max_fds-open_files)/8; | |
1da177e4 LT |
749 | int start = open_files / (8 * sizeof(unsigned long)); |
750 | ||
badf1662 DS |
751 | memset(&new_fdt->open_fds->fds_bits[start], 0, left); |
752 | memset(&new_fdt->close_on_exec->fds_bits[start], 0, left); | |
1da177e4 LT |
753 | } |
754 | ||
a016f338 | 755 | return newf; |
1da177e4 LT |
756 | |
757 | out_release: | |
1da177e4 | 758 | kmem_cache_free(files_cachep, newf); |
f3d19c90 | 759 | out: |
42862298 | 760 | return NULL; |
1da177e4 LT |
761 | } |
762 | ||
a016f338 JD |
763 | static int copy_files(unsigned long clone_flags, struct task_struct * tsk) |
764 | { | |
765 | struct files_struct *oldf, *newf; | |
766 | int error = 0; | |
767 | ||
768 | /* | |
769 | * A background process may not have any files ... | |
770 | */ | |
771 | oldf = current->files; | |
772 | if (!oldf) | |
773 | goto out; | |
774 | ||
775 | if (clone_flags & CLONE_FILES) { | |
776 | atomic_inc(&oldf->count); | |
777 | goto out; | |
778 | } | |
779 | ||
780 | /* | |
781 | * Note: we may be using current for both targets (See exec.c) | |
782 | * This works because we cache current->files (old) as oldf. Don't | |
783 | * break this. | |
784 | */ | |
785 | tsk->files = NULL; | |
a016f338 JD |
786 | newf = dup_fd(oldf, &error); |
787 | if (!newf) | |
788 | goto out; | |
789 | ||
790 | tsk->files = newf; | |
791 | error = 0; | |
792 | out: | |
793 | return error; | |
794 | } | |
795 | ||
fadad878 | 796 | static int copy_io(unsigned long clone_flags, struct task_struct *tsk) |
fd0928df JA |
797 | { |
798 | #ifdef CONFIG_BLOCK | |
799 | struct io_context *ioc = current->io_context; | |
800 | ||
801 | if (!ioc) | |
802 | return 0; | |
fadad878 JA |
803 | /* |
804 | * Share io context with parent, if CLONE_IO is set | |
805 | */ | |
806 | if (clone_flags & CLONE_IO) { | |
807 | tsk->io_context = ioc_task_link(ioc); | |
808 | if (unlikely(!tsk->io_context)) | |
809 | return -ENOMEM; | |
810 | } else if (ioprio_valid(ioc->ioprio)) { | |
fd0928df JA |
811 | tsk->io_context = alloc_io_context(GFP_KERNEL, -1); |
812 | if (unlikely(!tsk->io_context)) | |
813 | return -ENOMEM; | |
814 | ||
fd0928df JA |
815 | tsk->io_context->ioprio = ioc->ioprio; |
816 | } | |
817 | #endif | |
818 | return 0; | |
819 | } | |
820 | ||
1da177e4 LT |
821 | /* |
822 | * Helper to unshare the files of the current task. | |
823 | * We don't want to expose copy_files internals to | |
824 | * the exec layer of the kernel. | |
825 | */ | |
826 | ||
827 | int unshare_files(void) | |
828 | { | |
829 | struct files_struct *files = current->files; | |
830 | int rc; | |
831 | ||
910dea7f | 832 | BUG_ON(!files); |
1da177e4 LT |
833 | |
834 | /* This can race but the race causes us to copy when we don't | |
835 | need to and drop the copy */ | |
836 | if(atomic_read(&files->count) == 1) | |
837 | { | |
838 | atomic_inc(&files->count); | |
839 | return 0; | |
840 | } | |
841 | rc = copy_files(0, current); | |
842 | if(rc) | |
843 | current->files = files; | |
844 | return rc; | |
845 | } | |
846 | ||
847 | EXPORT_SYMBOL(unshare_files); | |
848 | ||
a39bc516 | 849 | static int copy_sighand(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 LT |
850 | { |
851 | struct sighand_struct *sig; | |
852 | ||
853 | if (clone_flags & (CLONE_SIGHAND | CLONE_THREAD)) { | |
854 | atomic_inc(¤t->sighand->count); | |
855 | return 0; | |
856 | } | |
857 | sig = kmem_cache_alloc(sighand_cachep, GFP_KERNEL); | |
e56d0903 | 858 | rcu_assign_pointer(tsk->sighand, sig); |
1da177e4 LT |
859 | if (!sig) |
860 | return -ENOMEM; | |
1da177e4 LT |
861 | atomic_set(&sig->count, 1); |
862 | memcpy(sig->action, current->sighand->action, sizeof(sig->action)); | |
863 | return 0; | |
864 | } | |
865 | ||
a7e5328a | 866 | void __cleanup_sighand(struct sighand_struct *sighand) |
c81addc9 | 867 | { |
c81addc9 ON |
868 | if (atomic_dec_and_test(&sighand->count)) |
869 | kmem_cache_free(sighand_cachep, sighand); | |
870 | } | |
871 | ||
a39bc516 | 872 | static int copy_signal(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 LT |
873 | { |
874 | struct signal_struct *sig; | |
875 | int ret; | |
876 | ||
877 | if (clone_flags & CLONE_THREAD) { | |
878 | atomic_inc(¤t->signal->count); | |
879 | atomic_inc(¤t->signal->live); | |
880 | return 0; | |
881 | } | |
882 | sig = kmem_cache_alloc(signal_cachep, GFP_KERNEL); | |
883 | tsk->signal = sig; | |
884 | if (!sig) | |
885 | return -ENOMEM; | |
886 | ||
887 | ret = copy_thread_group_keys(tsk); | |
888 | if (ret < 0) { | |
889 | kmem_cache_free(signal_cachep, sig); | |
890 | return ret; | |
891 | } | |
892 | ||
893 | atomic_set(&sig->count, 1); | |
894 | atomic_set(&sig->live, 1); | |
895 | init_waitqueue_head(&sig->wait_chldexit); | |
896 | sig->flags = 0; | |
897 | sig->group_exit_code = 0; | |
898 | sig->group_exit_task = NULL; | |
899 | sig->group_stop_count = 0; | |
900 | sig->curr_target = NULL; | |
901 | init_sigpending(&sig->shared_pending); | |
902 | INIT_LIST_HEAD(&sig->posix_timers); | |
903 | ||
c9cb2e3d | 904 | hrtimer_init(&sig->real_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL); |
2ff678b8 | 905 | sig->it_real_incr.tv64 = 0; |
1da177e4 | 906 | sig->real_timer.function = it_real_fn; |
05cfb614 | 907 | sig->tsk = tsk; |
1da177e4 LT |
908 | |
909 | sig->it_virt_expires = cputime_zero; | |
910 | sig->it_virt_incr = cputime_zero; | |
911 | sig->it_prof_expires = cputime_zero; | |
912 | sig->it_prof_incr = cputime_zero; | |
913 | ||
1da177e4 | 914 | sig->leader = 0; /* session leadership doesn't inherit */ |
ab521dc0 | 915 | sig->tty_old_pgrp = NULL; |
1da177e4 LT |
916 | |
917 | sig->utime = sig->stime = sig->cutime = sig->cstime = cputime_zero; | |
9ac52315 LV |
918 | sig->gtime = cputime_zero; |
919 | sig->cgtime = cputime_zero; | |
1da177e4 LT |
920 | sig->nvcsw = sig->nivcsw = sig->cnvcsw = sig->cnivcsw = 0; |
921 | sig->min_flt = sig->maj_flt = sig->cmin_flt = sig->cmaj_flt = 0; | |
6eaeeaba | 922 | sig->inblock = sig->oublock = sig->cinblock = sig->coublock = 0; |
172ba844 | 923 | sig->sum_sched_runtime = 0; |
1da177e4 LT |
924 | INIT_LIST_HEAD(&sig->cpu_timers[0]); |
925 | INIT_LIST_HEAD(&sig->cpu_timers[1]); | |
926 | INIT_LIST_HEAD(&sig->cpu_timers[2]); | |
ad4ecbcb | 927 | taskstats_tgid_init(sig); |
1da177e4 LT |
928 | |
929 | task_lock(current->group_leader); | |
930 | memcpy(sig->rlim, current->signal->rlim, sizeof sig->rlim); | |
931 | task_unlock(current->group_leader); | |
932 | ||
933 | if (sig->rlim[RLIMIT_CPU].rlim_cur != RLIM_INFINITY) { | |
934 | /* | |
935 | * New sole thread in the process gets an expiry time | |
936 | * of the whole CPU time limit. | |
937 | */ | |
938 | tsk->it_prof_expires = | |
939 | secs_to_cputime(sig->rlim[RLIMIT_CPU].rlim_cur); | |
940 | } | |
0e464814 | 941 | acct_init_pacct(&sig->pacct); |
1da177e4 | 942 | |
522ed776 MT |
943 | tty_audit_fork(sig); |
944 | ||
1da177e4 LT |
945 | return 0; |
946 | } | |
947 | ||
6b3934ef ON |
948 | void __cleanup_signal(struct signal_struct *sig) |
949 | { | |
950 | exit_thread_group_keys(sig); | |
951 | kmem_cache_free(signal_cachep, sig); | |
952 | } | |
953 | ||
a39bc516 | 954 | static void cleanup_signal(struct task_struct *tsk) |
6b3934ef ON |
955 | { |
956 | struct signal_struct *sig = tsk->signal; | |
957 | ||
958 | atomic_dec(&sig->live); | |
959 | ||
960 | if (atomic_dec_and_test(&sig->count)) | |
961 | __cleanup_signal(sig); | |
962 | } | |
963 | ||
a39bc516 | 964 | static void copy_flags(unsigned long clone_flags, struct task_struct *p) |
1da177e4 LT |
965 | { |
966 | unsigned long new_flags = p->flags; | |
967 | ||
83144186 | 968 | new_flags &= ~PF_SUPERPRIV; |
1da177e4 LT |
969 | new_flags |= PF_FORKNOEXEC; |
970 | if (!(clone_flags & CLONE_PTRACE)) | |
971 | p->ptrace = 0; | |
972 | p->flags = new_flags; | |
2e131895 | 973 | clear_freeze_flag(p); |
1da177e4 LT |
974 | } |
975 | ||
976 | asmlinkage long sys_set_tid_address(int __user *tidptr) | |
977 | { | |
978 | current->clear_child_tid = tidptr; | |
979 | ||
b488893a | 980 | return task_pid_vnr(current); |
1da177e4 LT |
981 | } |
982 | ||
a39bc516 | 983 | static void rt_mutex_init_task(struct task_struct *p) |
23f78d4a | 984 | { |
23f78d4a | 985 | spin_lock_init(&p->pi_lock); |
e29e175b | 986 | #ifdef CONFIG_RT_MUTEXES |
23f78d4a IM |
987 | plist_head_init(&p->pi_waiters, &p->pi_lock); |
988 | p->pi_blocked_on = NULL; | |
23f78d4a IM |
989 | #endif |
990 | } | |
991 | ||
1da177e4 LT |
992 | /* |
993 | * This creates a new process as a copy of the old one, | |
994 | * but does not actually start it yet. | |
995 | * | |
996 | * It copies the registers, and all the appropriate | |
997 | * parts of the process environment (as per the clone | |
998 | * flags). The actual kick-off is left to the caller. | |
999 | */ | |
36c8b586 IM |
1000 | static struct task_struct *copy_process(unsigned long clone_flags, |
1001 | unsigned long stack_start, | |
1002 | struct pt_regs *regs, | |
1003 | unsigned long stack_size, | |
36c8b586 | 1004 | int __user *child_tidptr, |
85868995 | 1005 | struct pid *pid) |
1da177e4 LT |
1006 | { |
1007 | int retval; | |
a24efe62 | 1008 | struct task_struct *p; |
b4f48b63 | 1009 | int cgroup_callbacks_done = 0; |
1da177e4 LT |
1010 | |
1011 | if ((clone_flags & (CLONE_NEWNS|CLONE_FS)) == (CLONE_NEWNS|CLONE_FS)) | |
1012 | return ERR_PTR(-EINVAL); | |
1013 | ||
1014 | /* | |
1015 | * Thread groups must share signals as well, and detached threads | |
1016 | * can only be started up within the thread group. | |
1017 | */ | |
1018 | if ((clone_flags & CLONE_THREAD) && !(clone_flags & CLONE_SIGHAND)) | |
1019 | return ERR_PTR(-EINVAL); | |
1020 | ||
1021 | /* | |
1022 | * Shared signal handlers imply shared VM. By way of the above, | |
1023 | * thread groups also imply shared VM. Blocking this case allows | |
1024 | * for various simplifications in other code. | |
1025 | */ | |
1026 | if ((clone_flags & CLONE_SIGHAND) && !(clone_flags & CLONE_VM)) | |
1027 | return ERR_PTR(-EINVAL); | |
1028 | ||
1029 | retval = security_task_create(clone_flags); | |
1030 | if (retval) | |
1031 | goto fork_out; | |
1032 | ||
1033 | retval = -ENOMEM; | |
1034 | p = dup_task_struct(current); | |
1035 | if (!p) | |
1036 | goto fork_out; | |
1037 | ||
bea493a0 PZ |
1038 | rt_mutex_init_task(p); |
1039 | ||
de30a2b3 IM |
1040 | #ifdef CONFIG_TRACE_IRQFLAGS |
1041 | DEBUG_LOCKS_WARN_ON(!p->hardirqs_enabled); | |
1042 | DEBUG_LOCKS_WARN_ON(!p->softirqs_enabled); | |
1043 | #endif | |
1da177e4 LT |
1044 | retval = -EAGAIN; |
1045 | if (atomic_read(&p->user->processes) >= | |
1046 | p->signal->rlim[RLIMIT_NPROC].rlim_cur) { | |
1047 | if (!capable(CAP_SYS_ADMIN) && !capable(CAP_SYS_RESOURCE) && | |
acce292c | 1048 | p->user != current->nsproxy->user_ns->root_user) |
1da177e4 LT |
1049 | goto bad_fork_free; |
1050 | } | |
1051 | ||
1052 | atomic_inc(&p->user->__count); | |
1053 | atomic_inc(&p->user->processes); | |
1054 | get_group_info(p->group_info); | |
1055 | ||
1056 | /* | |
1057 | * If multiple threads are within copy_process(), then this check | |
1058 | * triggers too late. This doesn't hurt, the check is only there | |
1059 | * to stop root fork bombs. | |
1060 | */ | |
1061 | if (nr_threads >= max_threads) | |
1062 | goto bad_fork_cleanup_count; | |
1063 | ||
a1261f54 | 1064 | if (!try_module_get(task_thread_info(p)->exec_domain->module)) |
1da177e4 LT |
1065 | goto bad_fork_cleanup_count; |
1066 | ||
1067 | if (p->binfmt && !try_module_get(p->binfmt->module)) | |
1068 | goto bad_fork_cleanup_put_domain; | |
1069 | ||
1070 | p->did_exec = 0; | |
ca74e92b | 1071 | delayacct_tsk_init(p); /* Must remain after dup_task_struct() */ |
1da177e4 | 1072 | copy_flags(clone_flags, p); |
1da177e4 LT |
1073 | INIT_LIST_HEAD(&p->children); |
1074 | INIT_LIST_HEAD(&p->sibling); | |
e260be67 PM |
1075 | #ifdef CONFIG_PREEMPT_RCU |
1076 | p->rcu_read_lock_nesting = 0; | |
1077 | p->rcu_flipctr_idx = 0; | |
1078 | #endif /* #ifdef CONFIG_PREEMPT_RCU */ | |
1da177e4 LT |
1079 | p->vfork_done = NULL; |
1080 | spin_lock_init(&p->alloc_lock); | |
1da177e4 LT |
1081 | |
1082 | clear_tsk_thread_flag(p, TIF_SIGPENDING); | |
1083 | init_sigpending(&p->pending); | |
1084 | ||
1085 | p->utime = cputime_zero; | |
1086 | p->stime = cputime_zero; | |
9ac52315 | 1087 | p->gtime = cputime_zero; |
c66f08be MN |
1088 | p->utimescaled = cputime_zero; |
1089 | p->stimescaled = cputime_zero; | |
73a2bcb0 | 1090 | p->prev_utime = cputime_zero; |
9301899b | 1091 | p->prev_stime = cputime_zero; |
172ba844 | 1092 | |
82a1fcb9 IM |
1093 | #ifdef CONFIG_DETECT_SOFTLOCKUP |
1094 | p->last_switch_count = 0; | |
1095 | p->last_switch_timestamp = 0; | |
1096 | #endif | |
1097 | ||
4b98d11b | 1098 | #ifdef CONFIG_TASK_XACCT |
1da177e4 LT |
1099 | p->rchar = 0; /* I/O counter: bytes read */ |
1100 | p->wchar = 0; /* I/O counter: bytes written */ | |
1101 | p->syscr = 0; /* I/O counter: read syscalls */ | |
1102 | p->syscw = 0; /* I/O counter: write syscalls */ | |
4b98d11b | 1103 | #endif |
7c3ab738 | 1104 | task_io_accounting_init(p); |
1da177e4 LT |
1105 | acct_clear_integrals(p); |
1106 | ||
23ff4440 | 1107 | p->it_virt_expires = cputime_zero; |
1da177e4 | 1108 | p->it_prof_expires = cputime_zero; |
23ff4440 DW |
1109 | p->it_sched_expires = 0; |
1110 | INIT_LIST_HEAD(&p->cpu_timers[0]); | |
1111 | INIT_LIST_HEAD(&p->cpu_timers[1]); | |
1112 | INIT_LIST_HEAD(&p->cpu_timers[2]); | |
1da177e4 LT |
1113 | |
1114 | p->lock_depth = -1; /* -1 = no lock */ | |
1115 | do_posix_clock_monotonic_gettime(&p->start_time); | |
924b42d5 TJ |
1116 | p->real_start_time = p->start_time; |
1117 | monotonic_to_bootbased(&p->real_start_time); | |
57c521ce | 1118 | #ifdef CONFIG_SECURITY |
1da177e4 | 1119 | p->security = NULL; |
57c521ce | 1120 | #endif |
1da177e4 | 1121 | p->io_context = NULL; |
1da177e4 | 1122 | p->audit_context = NULL; |
b4f48b63 | 1123 | cgroup_fork(p); |
1da177e4 LT |
1124 | #ifdef CONFIG_NUMA |
1125 | p->mempolicy = mpol_copy(p->mempolicy); | |
1126 | if (IS_ERR(p->mempolicy)) { | |
1127 | retval = PTR_ERR(p->mempolicy); | |
1128 | p->mempolicy = NULL; | |
b4f48b63 | 1129 | goto bad_fork_cleanup_cgroup; |
1da177e4 | 1130 | } |
c61afb18 | 1131 | mpol_fix_fork_child_flag(p); |
1da177e4 | 1132 | #endif |
de30a2b3 IM |
1133 | #ifdef CONFIG_TRACE_IRQFLAGS |
1134 | p->irq_events = 0; | |
b36e4758 RK |
1135 | #ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW |
1136 | p->hardirqs_enabled = 1; | |
1137 | #else | |
de30a2b3 | 1138 | p->hardirqs_enabled = 0; |
b36e4758 | 1139 | #endif |
de30a2b3 IM |
1140 | p->hardirq_enable_ip = 0; |
1141 | p->hardirq_enable_event = 0; | |
1142 | p->hardirq_disable_ip = _THIS_IP_; | |
1143 | p->hardirq_disable_event = 0; | |
1144 | p->softirqs_enabled = 1; | |
1145 | p->softirq_enable_ip = _THIS_IP_; | |
1146 | p->softirq_enable_event = 0; | |
1147 | p->softirq_disable_ip = 0; | |
1148 | p->softirq_disable_event = 0; | |
1149 | p->hardirq_context = 0; | |
1150 | p->softirq_context = 0; | |
1151 | #endif | |
fbb9ce95 IM |
1152 | #ifdef CONFIG_LOCKDEP |
1153 | p->lockdep_depth = 0; /* no locks held yet */ | |
1154 | p->curr_chain_key = 0; | |
1155 | p->lockdep_recursion = 0; | |
1156 | #endif | |
1da177e4 | 1157 | |
408894ee IM |
1158 | #ifdef CONFIG_DEBUG_MUTEXES |
1159 | p->blocked_on = NULL; /* not blocked yet */ | |
1160 | #endif | |
1161 | ||
3c90e6e9 SV |
1162 | /* Perform scheduler related setup. Assign this task to a CPU. */ |
1163 | sched_fork(p, clone_flags); | |
1164 | ||
1da177e4 LT |
1165 | if ((retval = security_task_alloc(p))) |
1166 | goto bad_fork_cleanup_policy; | |
1167 | if ((retval = audit_alloc(p))) | |
1168 | goto bad_fork_cleanup_security; | |
1169 | /* copy all the process information */ | |
1170 | if ((retval = copy_semundo(clone_flags, p))) | |
1171 | goto bad_fork_cleanup_audit; | |
1172 | if ((retval = copy_files(clone_flags, p))) | |
1173 | goto bad_fork_cleanup_semundo; | |
1174 | if ((retval = copy_fs(clone_flags, p))) | |
1175 | goto bad_fork_cleanup_files; | |
1176 | if ((retval = copy_sighand(clone_flags, p))) | |
1177 | goto bad_fork_cleanup_fs; | |
1178 | if ((retval = copy_signal(clone_flags, p))) | |
1179 | goto bad_fork_cleanup_sighand; | |
1180 | if ((retval = copy_mm(clone_flags, p))) | |
1181 | goto bad_fork_cleanup_signal; | |
1182 | if ((retval = copy_keys(clone_flags, p))) | |
1183 | goto bad_fork_cleanup_mm; | |
ab516013 | 1184 | if ((retval = copy_namespaces(clone_flags, p))) |
1da177e4 | 1185 | goto bad_fork_cleanup_keys; |
fadad878 | 1186 | if ((retval = copy_io(clone_flags, p))) |
fd0928df | 1187 | goto bad_fork_cleanup_namespaces; |
1da177e4 LT |
1188 | retval = copy_thread(0, clone_flags, stack_start, stack_size, p, regs); |
1189 | if (retval) | |
fd0928df | 1190 | goto bad_fork_cleanup_io; |
1da177e4 | 1191 | |
425fb2b4 PE |
1192 | if (pid != &init_struct_pid) { |
1193 | retval = -ENOMEM; | |
1194 | pid = alloc_pid(task_active_pid_ns(p)); | |
1195 | if (!pid) | |
fd0928df | 1196 | goto bad_fork_cleanup_io; |
6f4e6433 PE |
1197 | |
1198 | if (clone_flags & CLONE_NEWPID) { | |
1199 | retval = pid_ns_prepare_proc(task_active_pid_ns(p)); | |
1200 | if (retval < 0) | |
1201 | goto bad_fork_free_pid; | |
1202 | } | |
425fb2b4 PE |
1203 | } |
1204 | ||
1205 | p->pid = pid_nr(pid); | |
1206 | p->tgid = p->pid; | |
1207 | if (clone_flags & CLONE_THREAD) | |
1208 | p->tgid = current->tgid; | |
1209 | ||
1da177e4 LT |
1210 | p->set_child_tid = (clone_flags & CLONE_CHILD_SETTID) ? child_tidptr : NULL; |
1211 | /* | |
1212 | * Clear TID on mm_release()? | |
1213 | */ | |
1214 | p->clear_child_tid = (clone_flags & CLONE_CHILD_CLEARTID) ? child_tidptr: NULL; | |
42b2dd0a | 1215 | #ifdef CONFIG_FUTEX |
8f17d3a5 IM |
1216 | p->robust_list = NULL; |
1217 | #ifdef CONFIG_COMPAT | |
1218 | p->compat_robust_list = NULL; | |
1219 | #endif | |
c87e2837 IM |
1220 | INIT_LIST_HEAD(&p->pi_state_list); |
1221 | p->pi_state_cache = NULL; | |
42b2dd0a | 1222 | #endif |
f9a3879a GM |
1223 | /* |
1224 | * sigaltstack should be cleared when sharing the same VM | |
1225 | */ | |
1226 | if ((clone_flags & (CLONE_VM|CLONE_VFORK)) == CLONE_VM) | |
1227 | p->sas_ss_sp = p->sas_ss_size = 0; | |
1228 | ||
1da177e4 LT |
1229 | /* |
1230 | * Syscall tracing should be turned off in the child regardless | |
1231 | * of CLONE_PTRACE. | |
1232 | */ | |
1233 | clear_tsk_thread_flag(p, TIF_SYSCALL_TRACE); | |
ed75e8d5 LV |
1234 | #ifdef TIF_SYSCALL_EMU |
1235 | clear_tsk_thread_flag(p, TIF_SYSCALL_EMU); | |
1236 | #endif | |
9745512c | 1237 | clear_all_latency_tracing(p); |
1da177e4 LT |
1238 | |
1239 | /* Our parent execution domain becomes current domain | |
1240 | These must match for thread signalling to apply */ | |
1da177e4 LT |
1241 | p->parent_exec_id = p->self_exec_id; |
1242 | ||
1243 | /* ok, now we should be set up.. */ | |
1244 | p->exit_signal = (clone_flags & CLONE_THREAD) ? -1 : (clone_flags & CSIGNAL); | |
1245 | p->pdeath_signal = 0; | |
1246 | p->exit_state = 0; | |
1247 | ||
1da177e4 LT |
1248 | /* |
1249 | * Ok, make it visible to the rest of the system. | |
1250 | * We dont wake it up yet. | |
1251 | */ | |
1252 | p->group_leader = p; | |
47e65328 | 1253 | INIT_LIST_HEAD(&p->thread_group); |
1da177e4 LT |
1254 | INIT_LIST_HEAD(&p->ptrace_children); |
1255 | INIT_LIST_HEAD(&p->ptrace_list); | |
1256 | ||
b4f48b63 PM |
1257 | /* Now that the task is set up, run cgroup callbacks if |
1258 | * necessary. We need to run them before the task is visible | |
1259 | * on the tasklist. */ | |
1260 | cgroup_fork_callbacks(p); | |
1261 | cgroup_callbacks_done = 1; | |
1262 | ||
1da177e4 LT |
1263 | /* Need tasklist lock for parent etc handling! */ |
1264 | write_lock_irq(&tasklist_lock); | |
1265 | ||
1266 | /* | |
476d139c NP |
1267 | * The task hasn't been attached yet, so its cpus_allowed mask will |
1268 | * not be changed, nor will its assigned CPU. | |
1269 | * | |
1270 | * The cpus_allowed mask of the parent may have changed after it was | |
1271 | * copied first time - so re-copy it here, then check the child's CPU | |
1272 | * to ensure it is on a valid CPU (and if not, just force it back to | |
1273 | * parent's CPU). This avoids alot of nasty races. | |
1da177e4 LT |
1274 | */ |
1275 | p->cpus_allowed = current->cpus_allowed; | |
6f505b16 | 1276 | p->rt.nr_cpus_allowed = current->rt.nr_cpus_allowed; |
26ff6ad9 SV |
1277 | if (unlikely(!cpu_isset(task_cpu(p), p->cpus_allowed) || |
1278 | !cpu_online(task_cpu(p)))) | |
476d139c | 1279 | set_task_cpu(p, smp_processor_id()); |
1da177e4 | 1280 | |
1da177e4 LT |
1281 | /* CLONE_PARENT re-uses the old parent */ |
1282 | if (clone_flags & (CLONE_PARENT|CLONE_THREAD)) | |
1283 | p->real_parent = current->real_parent; | |
1284 | else | |
1285 | p->real_parent = current; | |
1286 | p->parent = p->real_parent; | |
1287 | ||
3f17da69 | 1288 | spin_lock(¤t->sighand->siglock); |
4a2c7a78 ON |
1289 | |
1290 | /* | |
1291 | * Process group and session signals need to be delivered to just the | |
1292 | * parent before the fork or both the parent and the child after the | |
1293 | * fork. Restart if a signal comes in before we add the new process to | |
1294 | * it's process group. | |
1295 | * A fatal signal pending means that current will exit, so the new | |
1296 | * thread can't slip out of an OOM kill (or normal SIGKILL). | |
1297 | */ | |
23ff4440 | 1298 | recalc_sigpending(); |
4a2c7a78 ON |
1299 | if (signal_pending(current)) { |
1300 | spin_unlock(¤t->sighand->siglock); | |
1301 | write_unlock_irq(&tasklist_lock); | |
1302 | retval = -ERESTARTNOINTR; | |
425fb2b4 | 1303 | goto bad_fork_free_pid; |
4a2c7a78 ON |
1304 | } |
1305 | ||
1da177e4 | 1306 | if (clone_flags & CLONE_THREAD) { |
1da177e4 | 1307 | p->group_leader = current->group_leader; |
47e65328 | 1308 | list_add_tail_rcu(&p->thread_group, &p->group_leader->thread_group); |
1da177e4 | 1309 | |
1da177e4 LT |
1310 | if (!cputime_eq(current->signal->it_virt_expires, |
1311 | cputime_zero) || | |
1312 | !cputime_eq(current->signal->it_prof_expires, | |
1313 | cputime_zero) || | |
1314 | current->signal->rlim[RLIMIT_CPU].rlim_cur != RLIM_INFINITY || | |
1315 | !list_empty(¤t->signal->cpu_timers[0]) || | |
1316 | !list_empty(¤t->signal->cpu_timers[1]) || | |
1317 | !list_empty(¤t->signal->cpu_timers[2])) { | |
1318 | /* | |
1319 | * Have child wake up on its first tick to check | |
1320 | * for process CPU timers. | |
1321 | */ | |
1322 | p->it_prof_expires = jiffies_to_cputime(1); | |
1323 | } | |
1da177e4 LT |
1324 | } |
1325 | ||
73b9ebfe ON |
1326 | if (likely(p->pid)) { |
1327 | add_parent(p); | |
1328 | if (unlikely(p->ptrace & PT_PTRACED)) | |
1329 | __ptrace_link(p, current->parent); | |
1330 | ||
1331 | if (thread_group_leader(p)) { | |
5cd17569 | 1332 | if (clone_flags & CLONE_NEWPID) |
30e49c26 | 1333 | p->nsproxy->pid_ns->child_reaper = p; |
73b9ebfe | 1334 | |
5cd17569 EB |
1335 | p->signal->tty = current->signal->tty; |
1336 | set_task_pgrp(p, task_pgrp_nr(current)); | |
1337 | set_task_session(p, task_session_nr(current)); | |
1338 | attach_pid(p, PIDTYPE_PGID, task_pgrp(current)); | |
1339 | attach_pid(p, PIDTYPE_SID, task_session(current)); | |
5e85d4ab | 1340 | list_add_tail_rcu(&p->tasks, &init_task.tasks); |
1da177e4 | 1341 | __get_cpu_var(process_counts)++; |
73b9ebfe | 1342 | } |
85868995 | 1343 | attach_pid(p, PIDTYPE_PID, pid); |
73b9ebfe | 1344 | nr_threads++; |
1da177e4 LT |
1345 | } |
1346 | ||
1da177e4 | 1347 | total_forks++; |
3f17da69 | 1348 | spin_unlock(¤t->sighand->siglock); |
1da177e4 | 1349 | write_unlock_irq(&tasklist_lock); |
c13cf856 | 1350 | proc_fork_connector(p); |
817929ec | 1351 | cgroup_post_fork(p); |
1da177e4 LT |
1352 | return p; |
1353 | ||
425fb2b4 PE |
1354 | bad_fork_free_pid: |
1355 | if (pid != &init_struct_pid) | |
1356 | free_pid(pid); | |
fd0928df JA |
1357 | bad_fork_cleanup_io: |
1358 | put_io_context(p->io_context); | |
ab516013 | 1359 | bad_fork_cleanup_namespaces: |
444f378b | 1360 | exit_task_namespaces(p); |
1da177e4 LT |
1361 | bad_fork_cleanup_keys: |
1362 | exit_keys(p); | |
1363 | bad_fork_cleanup_mm: | |
1364 | if (p->mm) | |
1365 | mmput(p->mm); | |
1366 | bad_fork_cleanup_signal: | |
6b3934ef | 1367 | cleanup_signal(p); |
1da177e4 | 1368 | bad_fork_cleanup_sighand: |
a7e5328a | 1369 | __cleanup_sighand(p->sighand); |
1da177e4 LT |
1370 | bad_fork_cleanup_fs: |
1371 | exit_fs(p); /* blocking */ | |
1372 | bad_fork_cleanup_files: | |
1373 | exit_files(p); /* blocking */ | |
1374 | bad_fork_cleanup_semundo: | |
1375 | exit_sem(p); | |
1376 | bad_fork_cleanup_audit: | |
1377 | audit_free(p); | |
1378 | bad_fork_cleanup_security: | |
1379 | security_task_free(p); | |
1380 | bad_fork_cleanup_policy: | |
1381 | #ifdef CONFIG_NUMA | |
1382 | mpol_free(p->mempolicy); | |
b4f48b63 | 1383 | bad_fork_cleanup_cgroup: |
1da177e4 | 1384 | #endif |
b4f48b63 | 1385 | cgroup_exit(p, cgroup_callbacks_done); |
35df17c5 | 1386 | delayacct_tsk_free(p); |
1da177e4 LT |
1387 | if (p->binfmt) |
1388 | module_put(p->binfmt->module); | |
1389 | bad_fork_cleanup_put_domain: | |
a1261f54 | 1390 | module_put(task_thread_info(p)->exec_domain->module); |
1da177e4 LT |
1391 | bad_fork_cleanup_count: |
1392 | put_group_info(p->group_info); | |
1393 | atomic_dec(&p->user->processes); | |
1394 | free_uid(p->user); | |
1395 | bad_fork_free: | |
1396 | free_task(p); | |
fe7d37d1 ON |
1397 | fork_out: |
1398 | return ERR_PTR(retval); | |
1da177e4 LT |
1399 | } |
1400 | ||
f95d47ca | 1401 | noinline struct pt_regs * __devinit __attribute__((weak)) idle_regs(struct pt_regs *regs) |
1da177e4 LT |
1402 | { |
1403 | memset(regs, 0, sizeof(struct pt_regs)); | |
1404 | return regs; | |
1405 | } | |
1406 | ||
9abcf40b | 1407 | struct task_struct * __cpuinit fork_idle(int cpu) |
1da177e4 | 1408 | { |
36c8b586 | 1409 | struct task_struct *task; |
1da177e4 LT |
1410 | struct pt_regs regs; |
1411 | ||
30e49c26 | 1412 | task = copy_process(CLONE_VM, 0, idle_regs(®s), 0, NULL, |
85868995 | 1413 | &init_struct_pid); |
753ca4f3 AM |
1414 | if (!IS_ERR(task)) |
1415 | init_idle(task, cpu); | |
73b9ebfe | 1416 | |
1da177e4 LT |
1417 | return task; |
1418 | } | |
1419 | ||
a39bc516 | 1420 | static int fork_traceflag(unsigned clone_flags) |
1da177e4 LT |
1421 | { |
1422 | if (clone_flags & CLONE_UNTRACED) | |
1423 | return 0; | |
1424 | else if (clone_flags & CLONE_VFORK) { | |
1425 | if (current->ptrace & PT_TRACE_VFORK) | |
1426 | return PTRACE_EVENT_VFORK; | |
1427 | } else if ((clone_flags & CSIGNAL) != SIGCHLD) { | |
1428 | if (current->ptrace & PT_TRACE_CLONE) | |
1429 | return PTRACE_EVENT_CLONE; | |
1430 | } else if (current->ptrace & PT_TRACE_FORK) | |
1431 | return PTRACE_EVENT_FORK; | |
1432 | ||
1433 | return 0; | |
1434 | } | |
1435 | ||
1436 | /* | |
1437 | * Ok, this is the main fork-routine. | |
1438 | * | |
1439 | * It copies the process, and if successful kick-starts | |
1440 | * it and waits for it to finish using the VM if required. | |
1441 | */ | |
1442 | long do_fork(unsigned long clone_flags, | |
1443 | unsigned long stack_start, | |
1444 | struct pt_regs *regs, | |
1445 | unsigned long stack_size, | |
1446 | int __user *parent_tidptr, | |
1447 | int __user *child_tidptr) | |
1448 | { | |
1449 | struct task_struct *p; | |
1450 | int trace = 0; | |
92476d7f | 1451 | long nr; |
1da177e4 | 1452 | |
1da177e4 LT |
1453 | if (unlikely(current->ptrace)) { |
1454 | trace = fork_traceflag (clone_flags); | |
1455 | if (trace) | |
1456 | clone_flags |= CLONE_PTRACE; | |
1457 | } | |
1458 | ||
a6f5e063 | 1459 | p = copy_process(clone_flags, stack_start, regs, stack_size, |
30e49c26 | 1460 | child_tidptr, NULL); |
1da177e4 LT |
1461 | /* |
1462 | * Do this prior waking up the new thread - the thread pointer | |
1463 | * might get invalid after that point, if the thread exits quickly. | |
1464 | */ | |
1465 | if (!IS_ERR(p)) { | |
1466 | struct completion vfork; | |
1467 | ||
30e49c26 PE |
1468 | /* |
1469 | * this is enough to call pid_nr_ns here, but this if | |
1470 | * improves optimisation of regular fork() | |
1471 | */ | |
1472 | nr = (clone_flags & CLONE_NEWPID) ? | |
1473 | task_pid_nr_ns(p, current->nsproxy->pid_ns) : | |
1474 | task_pid_vnr(p); | |
1475 | ||
1476 | if (clone_flags & CLONE_PARENT_SETTID) | |
1477 | put_user(nr, parent_tidptr); | |
a6f5e063 | 1478 | |
1da177e4 LT |
1479 | if (clone_flags & CLONE_VFORK) { |
1480 | p->vfork_done = &vfork; | |
1481 | init_completion(&vfork); | |
1482 | } | |
1483 | ||
1484 | if ((p->ptrace & PT_PTRACED) || (clone_flags & CLONE_STOPPED)) { | |
1485 | /* | |
1486 | * We'll start up with an immediate SIGSTOP. | |
1487 | */ | |
1488 | sigaddset(&p->pending.signal, SIGSTOP); | |
1489 | set_tsk_thread_flag(p, TIF_SIGPENDING); | |
1490 | } | |
1491 | ||
1492 | if (!(clone_flags & CLONE_STOPPED)) | |
1493 | wake_up_new_task(p, clone_flags); | |
1494 | else | |
1495 | p->state = TASK_STOPPED; | |
1496 | ||
1497 | if (unlikely (trace)) { | |
92476d7f | 1498 | current->ptrace_message = nr; |
1da177e4 LT |
1499 | ptrace_notify ((trace << 8) | SIGTRAP); |
1500 | } | |
1501 | ||
1502 | if (clone_flags & CLONE_VFORK) { | |
ba96a0c8 | 1503 | freezer_do_not_count(); |
1da177e4 | 1504 | wait_for_completion(&vfork); |
ba96a0c8 | 1505 | freezer_count(); |
9f59ce5d CE |
1506 | if (unlikely (current->ptrace & PT_TRACE_VFORK_DONE)) { |
1507 | current->ptrace_message = nr; | |
1da177e4 | 1508 | ptrace_notify ((PTRACE_EVENT_VFORK_DONE << 8) | SIGTRAP); |
9f59ce5d | 1509 | } |
1da177e4 LT |
1510 | } |
1511 | } else { | |
92476d7f | 1512 | nr = PTR_ERR(p); |
1da177e4 | 1513 | } |
92476d7f | 1514 | return nr; |
1da177e4 LT |
1515 | } |
1516 | ||
5fd63b30 RT |
1517 | #ifndef ARCH_MIN_MMSTRUCT_ALIGN |
1518 | #define ARCH_MIN_MMSTRUCT_ALIGN 0 | |
1519 | #endif | |
1520 | ||
4ba9b9d0 | 1521 | static void sighand_ctor(struct kmem_cache *cachep, void *data) |
aa1757f9 ON |
1522 | { |
1523 | struct sighand_struct *sighand = data; | |
1524 | ||
a35afb83 | 1525 | spin_lock_init(&sighand->siglock); |
b8fceee1 | 1526 | init_waitqueue_head(&sighand->signalfd_wqh); |
aa1757f9 ON |
1527 | } |
1528 | ||
1da177e4 LT |
1529 | void __init proc_caches_init(void) |
1530 | { | |
1531 | sighand_cachep = kmem_cache_create("sighand_cache", | |
1532 | sizeof(struct sighand_struct), 0, | |
aa1757f9 | 1533 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_DESTROY_BY_RCU, |
20c2df83 | 1534 | sighand_ctor); |
1da177e4 LT |
1535 | signal_cachep = kmem_cache_create("signal_cache", |
1536 | sizeof(struct signal_struct), 0, | |
20c2df83 PM |
1537 | SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL); |
1538 | files_cachep = kmem_cache_create("files_cache", | |
1da177e4 | 1539 | sizeof(struct files_struct), 0, |
20c2df83 PM |
1540 | SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL); |
1541 | fs_cachep = kmem_cache_create("fs_cache", | |
1da177e4 | 1542 | sizeof(struct fs_struct), 0, |
20c2df83 | 1543 | SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL); |
1da177e4 LT |
1544 | vm_area_cachep = kmem_cache_create("vm_area_struct", |
1545 | sizeof(struct vm_area_struct), 0, | |
20c2df83 | 1546 | SLAB_PANIC, NULL); |
1da177e4 | 1547 | mm_cachep = kmem_cache_create("mm_struct", |
5fd63b30 | 1548 | sizeof(struct mm_struct), ARCH_MIN_MMSTRUCT_ALIGN, |
20c2df83 | 1549 | SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL); |
1da177e4 | 1550 | } |
cf2e340f | 1551 | |
cf2e340f JD |
1552 | /* |
1553 | * Check constraints on flags passed to the unshare system call and | |
1554 | * force unsharing of additional process context as appropriate. | |
1555 | */ | |
a39bc516 | 1556 | static void check_unshare_flags(unsigned long *flags_ptr) |
cf2e340f JD |
1557 | { |
1558 | /* | |
1559 | * If unsharing a thread from a thread group, must also | |
1560 | * unshare vm. | |
1561 | */ | |
1562 | if (*flags_ptr & CLONE_THREAD) | |
1563 | *flags_ptr |= CLONE_VM; | |
1564 | ||
1565 | /* | |
1566 | * If unsharing vm, must also unshare signal handlers. | |
1567 | */ | |
1568 | if (*flags_ptr & CLONE_VM) | |
1569 | *flags_ptr |= CLONE_SIGHAND; | |
1570 | ||
1571 | /* | |
1572 | * If unsharing signal handlers and the task was created | |
1573 | * using CLONE_THREAD, then must unshare the thread | |
1574 | */ | |
1575 | if ((*flags_ptr & CLONE_SIGHAND) && | |
1576 | (atomic_read(¤t->signal->count) > 1)) | |
1577 | *flags_ptr |= CLONE_THREAD; | |
1578 | ||
1579 | /* | |
1580 | * If unsharing namespace, must also unshare filesystem information. | |
1581 | */ | |
1582 | if (*flags_ptr & CLONE_NEWNS) | |
1583 | *flags_ptr |= CLONE_FS; | |
1584 | } | |
1585 | ||
1586 | /* | |
1587 | * Unsharing of tasks created with CLONE_THREAD is not supported yet | |
1588 | */ | |
1589 | static int unshare_thread(unsigned long unshare_flags) | |
1590 | { | |
1591 | if (unshare_flags & CLONE_THREAD) | |
1592 | return -EINVAL; | |
1593 | ||
1594 | return 0; | |
1595 | } | |
1596 | ||
1597 | /* | |
99d1419d | 1598 | * Unshare the filesystem structure if it is being shared |
cf2e340f JD |
1599 | */ |
1600 | static int unshare_fs(unsigned long unshare_flags, struct fs_struct **new_fsp) | |
1601 | { | |
1602 | struct fs_struct *fs = current->fs; | |
1603 | ||
1604 | if ((unshare_flags & CLONE_FS) && | |
99d1419d JD |
1605 | (fs && atomic_read(&fs->count) > 1)) { |
1606 | *new_fsp = __copy_fs_struct(current->fs); | |
1607 | if (!*new_fsp) | |
1608 | return -ENOMEM; | |
1609 | } | |
cf2e340f JD |
1610 | |
1611 | return 0; | |
1612 | } | |
1613 | ||
cf2e340f | 1614 | /* |
dae3c5a0 | 1615 | * Unsharing of sighand is not supported yet |
cf2e340f JD |
1616 | */ |
1617 | static int unshare_sighand(unsigned long unshare_flags, struct sighand_struct **new_sighp) | |
1618 | { | |
1619 | struct sighand_struct *sigh = current->sighand; | |
1620 | ||
dae3c5a0 | 1621 | if ((unshare_flags & CLONE_SIGHAND) && atomic_read(&sigh->count) > 1) |
cf2e340f JD |
1622 | return -EINVAL; |
1623 | else | |
1624 | return 0; | |
1625 | } | |
1626 | ||
1627 | /* | |
a0a7ec30 | 1628 | * Unshare vm if it is being shared |
cf2e340f JD |
1629 | */ |
1630 | static int unshare_vm(unsigned long unshare_flags, struct mm_struct **new_mmp) | |
1631 | { | |
1632 | struct mm_struct *mm = current->mm; | |
1633 | ||
1634 | if ((unshare_flags & CLONE_VM) && | |
a0a7ec30 | 1635 | (mm && atomic_read(&mm->mm_users) > 1)) { |
2d61b867 | 1636 | return -EINVAL; |
a0a7ec30 | 1637 | } |
cf2e340f JD |
1638 | |
1639 | return 0; | |
cf2e340f JD |
1640 | } |
1641 | ||
1642 | /* | |
a016f338 | 1643 | * Unshare file descriptor table if it is being shared |
cf2e340f JD |
1644 | */ |
1645 | static int unshare_fd(unsigned long unshare_flags, struct files_struct **new_fdp) | |
1646 | { | |
1647 | struct files_struct *fd = current->files; | |
a016f338 | 1648 | int error = 0; |
cf2e340f JD |
1649 | |
1650 | if ((unshare_flags & CLONE_FILES) && | |
a016f338 JD |
1651 | (fd && atomic_read(&fd->count) > 1)) { |
1652 | *new_fdp = dup_fd(fd, &error); | |
1653 | if (!*new_fdp) | |
1654 | return error; | |
1655 | } | |
cf2e340f JD |
1656 | |
1657 | return 0; | |
1658 | } | |
1659 | ||
1660 | /* | |
1661 | * Unsharing of semundo for tasks created with CLONE_SYSVSEM is not | |
1662 | * supported yet | |
1663 | */ | |
1664 | static int unshare_semundo(unsigned long unshare_flags, struct sem_undo_list **new_ulistp) | |
1665 | { | |
1666 | if (unshare_flags & CLONE_SYSVSEM) | |
1667 | return -EINVAL; | |
1668 | ||
1669 | return 0; | |
1670 | } | |
1671 | ||
1672 | /* | |
1673 | * unshare allows a process to 'unshare' part of the process | |
1674 | * context which was originally shared using clone. copy_* | |
1675 | * functions used by do_fork() cannot be used here directly | |
1676 | * because they modify an inactive task_struct that is being | |
1677 | * constructed. Here we are modifying the current, active, | |
1678 | * task_struct. | |
1679 | */ | |
1680 | asmlinkage long sys_unshare(unsigned long unshare_flags) | |
1681 | { | |
1682 | int err = 0; | |
1683 | struct fs_struct *fs, *new_fs = NULL; | |
dae3c5a0 | 1684 | struct sighand_struct *new_sigh = NULL; |
cf2e340f JD |
1685 | struct mm_struct *mm, *new_mm = NULL, *active_mm = NULL; |
1686 | struct files_struct *fd, *new_fd = NULL; | |
1687 | struct sem_undo_list *new_ulist = NULL; | |
cf7b708c | 1688 | struct nsproxy *new_nsproxy = NULL; |
cf2e340f JD |
1689 | |
1690 | check_unshare_flags(&unshare_flags); | |
1691 | ||
06f9d4f9 EB |
1692 | /* Return -EINVAL for all unsupported flags */ |
1693 | err = -EINVAL; | |
1694 | if (unshare_flags & ~(CLONE_THREAD|CLONE_FS|CLONE_NEWNS|CLONE_SIGHAND| | |
25b21cb2 | 1695 | CLONE_VM|CLONE_FILES|CLONE_SYSVSEM| |
9dd776b6 EB |
1696 | CLONE_NEWUTS|CLONE_NEWIPC|CLONE_NEWUSER| |
1697 | CLONE_NEWNET)) | |
06f9d4f9 EB |
1698 | goto bad_unshare_out; |
1699 | ||
cf2e340f JD |
1700 | if ((err = unshare_thread(unshare_flags))) |
1701 | goto bad_unshare_out; | |
1702 | if ((err = unshare_fs(unshare_flags, &new_fs))) | |
1703 | goto bad_unshare_cleanup_thread; | |
cf2e340f | 1704 | if ((err = unshare_sighand(unshare_flags, &new_sigh))) |
e3222c4e | 1705 | goto bad_unshare_cleanup_fs; |
cf2e340f JD |
1706 | if ((err = unshare_vm(unshare_flags, &new_mm))) |
1707 | goto bad_unshare_cleanup_sigh; | |
1708 | if ((err = unshare_fd(unshare_flags, &new_fd))) | |
1709 | goto bad_unshare_cleanup_vm; | |
1710 | if ((err = unshare_semundo(unshare_flags, &new_ulist))) | |
1711 | goto bad_unshare_cleanup_fd; | |
e3222c4e BP |
1712 | if ((err = unshare_nsproxy_namespaces(unshare_flags, &new_nsproxy, |
1713 | new_fs))) | |
071df104 | 1714 | goto bad_unshare_cleanup_semundo; |
c0b2fc31 | 1715 | |
e3222c4e | 1716 | if (new_fs || new_mm || new_fd || new_ulist || new_nsproxy) { |
ab516013 | 1717 | |
c0b2fc31 | 1718 | if (new_nsproxy) { |
cf7b708c PE |
1719 | switch_task_namespaces(current, new_nsproxy); |
1720 | new_nsproxy = NULL; | |
c0b2fc31 | 1721 | } |
cf2e340f | 1722 | |
cf7b708c PE |
1723 | task_lock(current); |
1724 | ||
cf2e340f JD |
1725 | if (new_fs) { |
1726 | fs = current->fs; | |
1727 | current->fs = new_fs; | |
1728 | new_fs = fs; | |
1729 | } | |
1730 | ||
cf2e340f JD |
1731 | if (new_mm) { |
1732 | mm = current->mm; | |
1733 | active_mm = current->active_mm; | |
1734 | current->mm = new_mm; | |
1735 | current->active_mm = new_mm; | |
1736 | activate_mm(active_mm, new_mm); | |
1737 | new_mm = mm; | |
1738 | } | |
1739 | ||
1740 | if (new_fd) { | |
1741 | fd = current->files; | |
1742 | current->files = new_fd; | |
1743 | new_fd = fd; | |
1744 | } | |
1745 | ||
1746 | task_unlock(current); | |
1747 | } | |
1748 | ||
c0b2fc31 | 1749 | if (new_nsproxy) |
444f378b | 1750 | put_nsproxy(new_nsproxy); |
c0b2fc31 | 1751 | |
ab516013 | 1752 | bad_unshare_cleanup_semundo: |
cf2e340f JD |
1753 | bad_unshare_cleanup_fd: |
1754 | if (new_fd) | |
1755 | put_files_struct(new_fd); | |
1756 | ||
1757 | bad_unshare_cleanup_vm: | |
1758 | if (new_mm) | |
1759 | mmput(new_mm); | |
1760 | ||
1761 | bad_unshare_cleanup_sigh: | |
1762 | if (new_sigh) | |
1763 | if (atomic_dec_and_test(&new_sigh->count)) | |
1764 | kmem_cache_free(sighand_cachep, new_sigh); | |
1765 | ||
cf2e340f JD |
1766 | bad_unshare_cleanup_fs: |
1767 | if (new_fs) | |
1768 | put_fs_struct(new_fs); | |
1769 | ||
1770 | bad_unshare_cleanup_thread: | |
1771 | bad_unshare_out: | |
1772 | return err; | |
1773 | } |