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 LT |
31 | #include <linux/cpu.h> |
32 | #include <linux/cpuset.h> | |
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> |
1da177e4 LT |
52 | |
53 | #include <asm/pgtable.h> | |
54 | #include <asm/pgalloc.h> | |
55 | #include <asm/uaccess.h> | |
56 | #include <asm/mmu_context.h> | |
57 | #include <asm/cacheflush.h> | |
58 | #include <asm/tlbflush.h> | |
59 | ||
60 | /* | |
61 | * Protected counters by write_lock_irq(&tasklist_lock) | |
62 | */ | |
63 | unsigned long total_forks; /* Handle normal Linux uptimes. */ | |
64 | int nr_threads; /* The idle threads do not count.. */ | |
65 | ||
66 | int max_threads; /* tunable limit on nr_threads */ | |
67 | ||
68 | DEFINE_PER_CPU(unsigned long, process_counts) = 0; | |
69 | ||
c59923a1 | 70 | __cacheline_aligned DEFINE_RWLOCK(tasklist_lock); /* outer */ |
1da177e4 LT |
71 | |
72 | int nr_processes(void) | |
73 | { | |
74 | int cpu; | |
75 | int total = 0; | |
76 | ||
77 | for_each_online_cpu(cpu) | |
78 | total += per_cpu(process_counts, cpu); | |
79 | ||
80 | return total; | |
81 | } | |
82 | ||
83 | #ifndef __HAVE_ARCH_TASK_STRUCT_ALLOCATOR | |
84 | # define alloc_task_struct() kmem_cache_alloc(task_struct_cachep, GFP_KERNEL) | |
85 | # define free_task_struct(tsk) kmem_cache_free(task_struct_cachep, (tsk)) | |
e18b890b | 86 | static struct kmem_cache *task_struct_cachep; |
1da177e4 LT |
87 | #endif |
88 | ||
89 | /* SLAB cache for signal_struct structures (tsk->signal) */ | |
e18b890b | 90 | static struct kmem_cache *signal_cachep; |
1da177e4 LT |
91 | |
92 | /* SLAB cache for sighand_struct structures (tsk->sighand) */ | |
e18b890b | 93 | struct kmem_cache *sighand_cachep; |
1da177e4 LT |
94 | |
95 | /* SLAB cache for files_struct structures (tsk->files) */ | |
e18b890b | 96 | struct kmem_cache *files_cachep; |
1da177e4 LT |
97 | |
98 | /* SLAB cache for fs_struct structures (tsk->fs) */ | |
e18b890b | 99 | struct kmem_cache *fs_cachep; |
1da177e4 LT |
100 | |
101 | /* SLAB cache for vm_area_struct structures */ | |
e18b890b | 102 | struct kmem_cache *vm_area_cachep; |
1da177e4 LT |
103 | |
104 | /* SLAB cache for mm_struct structures (tsk->mm) */ | |
e18b890b | 105 | static struct kmem_cache *mm_cachep; |
1da177e4 LT |
106 | |
107 | void free_task(struct task_struct *tsk) | |
108 | { | |
f7e4217b | 109 | free_thread_info(tsk->stack); |
23f78d4a | 110 | rt_mutex_debug_task_free(tsk); |
1da177e4 LT |
111 | free_task_struct(tsk); |
112 | } | |
113 | EXPORT_SYMBOL(free_task); | |
114 | ||
158d9ebd | 115 | void __put_task_struct(struct task_struct *tsk) |
1da177e4 LT |
116 | { |
117 | WARN_ON(!(tsk->exit_state & (EXIT_DEAD | EXIT_ZOMBIE))); | |
118 | WARN_ON(atomic_read(&tsk->usage)); | |
119 | WARN_ON(tsk == current); | |
120 | ||
1da177e4 LT |
121 | security_task_free(tsk); |
122 | free_uid(tsk->user); | |
123 | put_group_info(tsk->group_info); | |
35df17c5 | 124 | delayacct_tsk_free(tsk); |
1da177e4 LT |
125 | |
126 | if (!profile_handoff_task(tsk)) | |
127 | free_task(tsk); | |
128 | } | |
129 | ||
130 | void __init fork_init(unsigned long mempages) | |
131 | { | |
132 | #ifndef __HAVE_ARCH_TASK_STRUCT_ALLOCATOR | |
133 | #ifndef ARCH_MIN_TASKALIGN | |
134 | #define ARCH_MIN_TASKALIGN L1_CACHE_BYTES | |
135 | #endif | |
136 | /* create a slab on which task_structs can be allocated */ | |
137 | task_struct_cachep = | |
138 | kmem_cache_create("task_struct", sizeof(struct task_struct), | |
139 | ARCH_MIN_TASKALIGN, SLAB_PANIC, NULL, NULL); | |
140 | #endif | |
141 | ||
142 | /* | |
143 | * The default maximum number of threads is set to a safe | |
144 | * value: the thread structures can take up at most half | |
145 | * of memory. | |
146 | */ | |
147 | max_threads = mempages / (8 * THREAD_SIZE / PAGE_SIZE); | |
148 | ||
149 | /* | |
150 | * we need to allow at least 20 threads to boot a system | |
151 | */ | |
152 | if(max_threads < 20) | |
153 | max_threads = 20; | |
154 | ||
155 | init_task.signal->rlim[RLIMIT_NPROC].rlim_cur = max_threads/2; | |
156 | init_task.signal->rlim[RLIMIT_NPROC].rlim_max = max_threads/2; | |
157 | init_task.signal->rlim[RLIMIT_SIGPENDING] = | |
158 | init_task.signal->rlim[RLIMIT_NPROC]; | |
159 | } | |
160 | ||
161 | static struct task_struct *dup_task_struct(struct task_struct *orig) | |
162 | { | |
163 | struct task_struct *tsk; | |
164 | struct thread_info *ti; | |
165 | ||
166 | prepare_to_copy(orig); | |
167 | ||
168 | tsk = alloc_task_struct(); | |
169 | if (!tsk) | |
170 | return NULL; | |
171 | ||
172 | ti = alloc_thread_info(tsk); | |
173 | if (!ti) { | |
174 | free_task_struct(tsk); | |
175 | return NULL; | |
176 | } | |
177 | ||
1da177e4 | 178 | *tsk = *orig; |
f7e4217b | 179 | tsk->stack = ti; |
10ebffde | 180 | setup_thread_stack(tsk, orig); |
1da177e4 | 181 | |
0a425405 AV |
182 | #ifdef CONFIG_CC_STACKPROTECTOR |
183 | tsk->stack_canary = get_random_int(); | |
184 | #endif | |
185 | ||
1da177e4 LT |
186 | /* One for us, one for whoever does the "release_task()" (usually parent) */ |
187 | atomic_set(&tsk->usage,2); | |
4b5d37ac | 188 | atomic_set(&tsk->fs_excl, 0); |
6c5c9341 | 189 | #ifdef CONFIG_BLK_DEV_IO_TRACE |
2056a782 | 190 | tsk->btrace_seq = 0; |
6c5c9341 | 191 | #endif |
a0aa7f68 | 192 | tsk->splice_pipe = NULL; |
1da177e4 LT |
193 | return tsk; |
194 | } | |
195 | ||
196 | #ifdef CONFIG_MMU | |
fd3e42fc | 197 | static inline int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm) |
1da177e4 | 198 | { |
fd3e42fc | 199 | struct vm_area_struct *mpnt, *tmp, **pprev; |
1da177e4 LT |
200 | struct rb_node **rb_link, *rb_parent; |
201 | int retval; | |
202 | unsigned long charge; | |
203 | struct mempolicy *pol; | |
204 | ||
205 | down_write(&oldmm->mmap_sem); | |
ec8c0446 | 206 | flush_cache_dup_mm(oldmm); |
ad339451 IM |
207 | /* |
208 | * Not linked in yet - no deadlock potential: | |
209 | */ | |
210 | down_write_nested(&mm->mmap_sem, SINGLE_DEPTH_NESTING); | |
7ee78232 | 211 | |
1da177e4 LT |
212 | mm->locked_vm = 0; |
213 | mm->mmap = NULL; | |
214 | mm->mmap_cache = NULL; | |
215 | mm->free_area_cache = oldmm->mmap_base; | |
1363c3cd | 216 | mm->cached_hole_size = ~0UL; |
1da177e4 | 217 | mm->map_count = 0; |
1da177e4 LT |
218 | cpus_clear(mm->cpu_vm_mask); |
219 | mm->mm_rb = RB_ROOT; | |
220 | rb_link = &mm->mm_rb.rb_node; | |
221 | rb_parent = NULL; | |
222 | pprev = &mm->mmap; | |
223 | ||
fd3e42fc | 224 | for (mpnt = oldmm->mmap; mpnt; mpnt = mpnt->vm_next) { |
1da177e4 LT |
225 | struct file *file; |
226 | ||
227 | if (mpnt->vm_flags & VM_DONTCOPY) { | |
3b6bfcdb HD |
228 | long pages = vma_pages(mpnt); |
229 | mm->total_vm -= pages; | |
ab50b8ed | 230 | vm_stat_account(mm, mpnt->vm_flags, mpnt->vm_file, |
3b6bfcdb | 231 | -pages); |
1da177e4 LT |
232 | continue; |
233 | } | |
234 | charge = 0; | |
235 | if (mpnt->vm_flags & VM_ACCOUNT) { | |
236 | unsigned int len = (mpnt->vm_end - mpnt->vm_start) >> PAGE_SHIFT; | |
237 | if (security_vm_enough_memory(len)) | |
238 | goto fail_nomem; | |
239 | charge = len; | |
240 | } | |
e94b1766 | 241 | tmp = kmem_cache_alloc(vm_area_cachep, GFP_KERNEL); |
1da177e4 LT |
242 | if (!tmp) |
243 | goto fail_nomem; | |
244 | *tmp = *mpnt; | |
245 | pol = mpol_copy(vma_policy(mpnt)); | |
246 | retval = PTR_ERR(pol); | |
247 | if (IS_ERR(pol)) | |
248 | goto fail_nomem_policy; | |
249 | vma_set_policy(tmp, pol); | |
250 | tmp->vm_flags &= ~VM_LOCKED; | |
251 | tmp->vm_mm = mm; | |
252 | tmp->vm_next = NULL; | |
253 | anon_vma_link(tmp); | |
254 | file = tmp->vm_file; | |
255 | if (file) { | |
f3a43f3f | 256 | struct inode *inode = file->f_path.dentry->d_inode; |
1da177e4 LT |
257 | get_file(file); |
258 | if (tmp->vm_flags & VM_DENYWRITE) | |
259 | atomic_dec(&inode->i_writecount); | |
260 | ||
261 | /* insert tmp into the share list, just after mpnt */ | |
262 | spin_lock(&file->f_mapping->i_mmap_lock); | |
263 | tmp->vm_truncate_count = mpnt->vm_truncate_count; | |
264 | flush_dcache_mmap_lock(file->f_mapping); | |
265 | vma_prio_tree_add(tmp, mpnt); | |
266 | flush_dcache_mmap_unlock(file->f_mapping); | |
267 | spin_unlock(&file->f_mapping->i_mmap_lock); | |
268 | } | |
269 | ||
270 | /* | |
7ee78232 | 271 | * Link in the new vma and copy the page table entries. |
1da177e4 | 272 | */ |
1da177e4 LT |
273 | *pprev = tmp; |
274 | pprev = &tmp->vm_next; | |
275 | ||
276 | __vma_link_rb(mm, tmp, rb_link, rb_parent); | |
277 | rb_link = &tmp->vm_rb.rb_right; | |
278 | rb_parent = &tmp->vm_rb; | |
279 | ||
280 | mm->map_count++; | |
0b0db14c | 281 | retval = copy_page_range(mm, oldmm, mpnt); |
1da177e4 LT |
282 | |
283 | if (tmp->vm_ops && tmp->vm_ops->open) | |
284 | tmp->vm_ops->open(tmp); | |
285 | ||
286 | if (retval) | |
287 | goto out; | |
288 | } | |
d6dd61c8 JF |
289 | /* a new mm has just been created */ |
290 | arch_dup_mmap(oldmm, mm); | |
1da177e4 | 291 | retval = 0; |
1da177e4 | 292 | out: |
7ee78232 | 293 | up_write(&mm->mmap_sem); |
fd3e42fc | 294 | flush_tlb_mm(oldmm); |
1da177e4 LT |
295 | up_write(&oldmm->mmap_sem); |
296 | return retval; | |
297 | fail_nomem_policy: | |
298 | kmem_cache_free(vm_area_cachep, tmp); | |
299 | fail_nomem: | |
300 | retval = -ENOMEM; | |
301 | vm_unacct_memory(charge); | |
302 | goto out; | |
303 | } | |
304 | ||
305 | static inline int mm_alloc_pgd(struct mm_struct * mm) | |
306 | { | |
307 | mm->pgd = pgd_alloc(mm); | |
308 | if (unlikely(!mm->pgd)) | |
309 | return -ENOMEM; | |
310 | return 0; | |
311 | } | |
312 | ||
313 | static inline void mm_free_pgd(struct mm_struct * mm) | |
314 | { | |
315 | pgd_free(mm->pgd); | |
316 | } | |
317 | #else | |
318 | #define dup_mmap(mm, oldmm) (0) | |
319 | #define mm_alloc_pgd(mm) (0) | |
320 | #define mm_free_pgd(mm) | |
321 | #endif /* CONFIG_MMU */ | |
322 | ||
323 | __cacheline_aligned_in_smp DEFINE_SPINLOCK(mmlist_lock); | |
324 | ||
e94b1766 | 325 | #define allocate_mm() (kmem_cache_alloc(mm_cachep, GFP_KERNEL)) |
1da177e4 LT |
326 | #define free_mm(mm) (kmem_cache_free(mm_cachep, (mm))) |
327 | ||
328 | #include <linux/init_task.h> | |
329 | ||
330 | static struct mm_struct * mm_init(struct mm_struct * mm) | |
331 | { | |
332 | atomic_set(&mm->mm_users, 1); | |
333 | atomic_set(&mm->mm_count, 1); | |
334 | init_rwsem(&mm->mmap_sem); | |
335 | INIT_LIST_HEAD(&mm->mmlist); | |
336 | mm->core_waiters = 0; | |
337 | mm->nr_ptes = 0; | |
4294621f | 338 | set_mm_counter(mm, file_rss, 0); |
404351e6 | 339 | set_mm_counter(mm, anon_rss, 0); |
1da177e4 LT |
340 | spin_lock_init(&mm->page_table_lock); |
341 | rwlock_init(&mm->ioctx_list_lock); | |
342 | mm->ioctx_list = NULL; | |
1da177e4 | 343 | mm->free_area_cache = TASK_UNMAPPED_BASE; |
1363c3cd | 344 | mm->cached_hole_size = ~0UL; |
1da177e4 LT |
345 | |
346 | if (likely(!mm_alloc_pgd(mm))) { | |
347 | mm->def_flags = 0; | |
348 | return mm; | |
349 | } | |
350 | free_mm(mm); | |
351 | return NULL; | |
352 | } | |
353 | ||
354 | /* | |
355 | * Allocate and initialize an mm_struct. | |
356 | */ | |
357 | struct mm_struct * mm_alloc(void) | |
358 | { | |
359 | struct mm_struct * mm; | |
360 | ||
361 | mm = allocate_mm(); | |
362 | if (mm) { | |
363 | memset(mm, 0, sizeof(*mm)); | |
364 | mm = mm_init(mm); | |
365 | } | |
366 | return mm; | |
367 | } | |
368 | ||
369 | /* | |
370 | * Called when the last reference to the mm | |
371 | * is dropped: either by a lazy thread or by | |
372 | * mmput. Free the page directory and the mm. | |
373 | */ | |
374 | void fastcall __mmdrop(struct mm_struct *mm) | |
375 | { | |
376 | BUG_ON(mm == &init_mm); | |
377 | mm_free_pgd(mm); | |
378 | destroy_context(mm); | |
379 | free_mm(mm); | |
380 | } | |
381 | ||
382 | /* | |
383 | * Decrement the use count and release all resources for an mm. | |
384 | */ | |
385 | void mmput(struct mm_struct *mm) | |
386 | { | |
0ae26f1b AM |
387 | might_sleep(); |
388 | ||
1da177e4 LT |
389 | if (atomic_dec_and_test(&mm->mm_users)) { |
390 | exit_aio(mm); | |
391 | exit_mmap(mm); | |
392 | if (!list_empty(&mm->mmlist)) { | |
393 | spin_lock(&mmlist_lock); | |
394 | list_del(&mm->mmlist); | |
395 | spin_unlock(&mmlist_lock); | |
396 | } | |
397 | put_swap_token(mm); | |
398 | mmdrop(mm); | |
399 | } | |
400 | } | |
401 | EXPORT_SYMBOL_GPL(mmput); | |
402 | ||
403 | /** | |
404 | * get_task_mm - acquire a reference to the task's mm | |
405 | * | |
406 | * Returns %NULL if the task has no mm. Checks PF_BORROWED_MM (meaning | |
407 | * this kernel workthread has transiently adopted a user mm with use_mm, | |
408 | * to do its AIO) is not set and if so returns a reference to it, after | |
409 | * bumping up the use count. User must release the mm via mmput() | |
410 | * after use. Typically used by /proc and ptrace. | |
411 | */ | |
412 | struct mm_struct *get_task_mm(struct task_struct *task) | |
413 | { | |
414 | struct mm_struct *mm; | |
415 | ||
416 | task_lock(task); | |
417 | mm = task->mm; | |
418 | if (mm) { | |
419 | if (task->flags & PF_BORROWED_MM) | |
420 | mm = NULL; | |
421 | else | |
422 | atomic_inc(&mm->mm_users); | |
423 | } | |
424 | task_unlock(task); | |
425 | return mm; | |
426 | } | |
427 | EXPORT_SYMBOL_GPL(get_task_mm); | |
428 | ||
429 | /* Please note the differences between mmput and mm_release. | |
430 | * mmput is called whenever we stop holding onto a mm_struct, | |
431 | * error success whatever. | |
432 | * | |
433 | * mm_release is called after a mm_struct has been removed | |
434 | * from the current process. | |
435 | * | |
436 | * This difference is important for error handling, when we | |
437 | * only half set up a mm_struct for a new process and need to restore | |
438 | * the old one. Because we mmput the new mm_struct before | |
439 | * restoring the old one. . . | |
440 | * Eric Biederman 10 January 1998 | |
441 | */ | |
442 | void mm_release(struct task_struct *tsk, struct mm_struct *mm) | |
443 | { | |
444 | struct completion *vfork_done = tsk->vfork_done; | |
445 | ||
446 | /* Get rid of any cached register state */ | |
447 | deactivate_mm(tsk, mm); | |
448 | ||
449 | /* notify parent sleeping on vfork() */ | |
450 | if (vfork_done) { | |
451 | tsk->vfork_done = NULL; | |
452 | complete(vfork_done); | |
453 | } | |
fec1d011 RM |
454 | |
455 | /* | |
456 | * If we're exiting normally, clear a user-space tid field if | |
457 | * requested. We leave this alone when dying by signal, to leave | |
458 | * the value intact in a core dump, and to save the unnecessary | |
459 | * trouble otherwise. Userland only wants this done for a sys_exit. | |
460 | */ | |
461 | if (tsk->clear_child_tid | |
462 | && !(tsk->flags & PF_SIGNALED) | |
463 | && atomic_read(&mm->mm_users) > 1) { | |
1da177e4 LT |
464 | u32 __user * tidptr = tsk->clear_child_tid; |
465 | tsk->clear_child_tid = NULL; | |
466 | ||
467 | /* | |
468 | * We don't check the error code - if userspace has | |
469 | * not set up a proper pointer then tough luck. | |
470 | */ | |
471 | put_user(0, tidptr); | |
472 | sys_futex(tidptr, FUTEX_WAKE, 1, NULL, NULL, 0); | |
473 | } | |
474 | } | |
475 | ||
a0a7ec30 JD |
476 | /* |
477 | * Allocate a new mm structure and copy contents from the | |
478 | * mm structure of the passed in task structure. | |
479 | */ | |
480 | static struct mm_struct *dup_mm(struct task_struct *tsk) | |
481 | { | |
482 | struct mm_struct *mm, *oldmm = current->mm; | |
483 | int err; | |
484 | ||
485 | if (!oldmm) | |
486 | return NULL; | |
487 | ||
488 | mm = allocate_mm(); | |
489 | if (!mm) | |
490 | goto fail_nomem; | |
491 | ||
492 | memcpy(mm, oldmm, sizeof(*mm)); | |
493 | ||
7602bdf2 AC |
494 | /* Initializing for Swap token stuff */ |
495 | mm->token_priority = 0; | |
496 | mm->last_interval = 0; | |
497 | ||
a0a7ec30 JD |
498 | if (!mm_init(mm)) |
499 | goto fail_nomem; | |
500 | ||
501 | if (init_new_context(tsk, mm)) | |
502 | goto fail_nocontext; | |
503 | ||
504 | err = dup_mmap(mm, oldmm); | |
505 | if (err) | |
506 | goto free_pt; | |
507 | ||
508 | mm->hiwater_rss = get_mm_rss(mm); | |
509 | mm->hiwater_vm = mm->total_vm; | |
510 | ||
511 | return mm; | |
512 | ||
513 | free_pt: | |
514 | mmput(mm); | |
515 | ||
516 | fail_nomem: | |
517 | return NULL; | |
518 | ||
519 | fail_nocontext: | |
520 | /* | |
521 | * If init_new_context() failed, we cannot use mmput() to free the mm | |
522 | * because it calls destroy_context() | |
523 | */ | |
524 | mm_free_pgd(mm); | |
525 | free_mm(mm); | |
526 | return NULL; | |
527 | } | |
528 | ||
1da177e4 LT |
529 | static int copy_mm(unsigned long clone_flags, struct task_struct * tsk) |
530 | { | |
531 | struct mm_struct * mm, *oldmm; | |
532 | int retval; | |
533 | ||
534 | tsk->min_flt = tsk->maj_flt = 0; | |
535 | tsk->nvcsw = tsk->nivcsw = 0; | |
536 | ||
537 | tsk->mm = NULL; | |
538 | tsk->active_mm = NULL; | |
539 | ||
540 | /* | |
541 | * Are we cloning a kernel thread? | |
542 | * | |
543 | * We need to steal a active VM for that.. | |
544 | */ | |
545 | oldmm = current->mm; | |
546 | if (!oldmm) | |
547 | return 0; | |
548 | ||
549 | if (clone_flags & CLONE_VM) { | |
550 | atomic_inc(&oldmm->mm_users); | |
551 | mm = oldmm; | |
1da177e4 LT |
552 | goto good_mm; |
553 | } | |
554 | ||
555 | retval = -ENOMEM; | |
a0a7ec30 | 556 | mm = dup_mm(tsk); |
1da177e4 LT |
557 | if (!mm) |
558 | goto fail_nomem; | |
559 | ||
1da177e4 | 560 | good_mm: |
7602bdf2 AC |
561 | /* Initializing for Swap token stuff */ |
562 | mm->token_priority = 0; | |
563 | mm->last_interval = 0; | |
564 | ||
1da177e4 LT |
565 | tsk->mm = mm; |
566 | tsk->active_mm = mm; | |
567 | return 0; | |
568 | ||
1da177e4 LT |
569 | fail_nomem: |
570 | return retval; | |
1da177e4 LT |
571 | } |
572 | ||
573 | static inline struct fs_struct *__copy_fs_struct(struct fs_struct *old) | |
574 | { | |
575 | struct fs_struct *fs = kmem_cache_alloc(fs_cachep, GFP_KERNEL); | |
576 | /* We don't need to lock fs - think why ;-) */ | |
577 | if (fs) { | |
578 | atomic_set(&fs->count, 1); | |
579 | rwlock_init(&fs->lock); | |
580 | fs->umask = old->umask; | |
581 | read_lock(&old->lock); | |
582 | fs->rootmnt = mntget(old->rootmnt); | |
583 | fs->root = dget(old->root); | |
584 | fs->pwdmnt = mntget(old->pwdmnt); | |
585 | fs->pwd = dget(old->pwd); | |
586 | if (old->altroot) { | |
587 | fs->altrootmnt = mntget(old->altrootmnt); | |
588 | fs->altroot = dget(old->altroot); | |
589 | } else { | |
590 | fs->altrootmnt = NULL; | |
591 | fs->altroot = NULL; | |
592 | } | |
593 | read_unlock(&old->lock); | |
594 | } | |
595 | return fs; | |
596 | } | |
597 | ||
598 | struct fs_struct *copy_fs_struct(struct fs_struct *old) | |
599 | { | |
600 | return __copy_fs_struct(old); | |
601 | } | |
602 | ||
603 | EXPORT_SYMBOL_GPL(copy_fs_struct); | |
604 | ||
605 | static inline int copy_fs(unsigned long clone_flags, struct task_struct * tsk) | |
606 | { | |
607 | if (clone_flags & CLONE_FS) { | |
608 | atomic_inc(¤t->fs->count); | |
609 | return 0; | |
610 | } | |
611 | tsk->fs = __copy_fs_struct(current->fs); | |
612 | if (!tsk->fs) | |
613 | return -ENOMEM; | |
614 | return 0; | |
615 | } | |
616 | ||
ab2af1f5 | 617 | static int count_open_files(struct fdtable *fdt) |
1da177e4 | 618 | { |
bbea9f69 | 619 | int size = fdt->max_fds; |
1da177e4 LT |
620 | int i; |
621 | ||
622 | /* Find the last open fd */ | |
623 | for (i = size/(8*sizeof(long)); i > 0; ) { | |
badf1662 | 624 | if (fdt->open_fds->fds_bits[--i]) |
1da177e4 LT |
625 | break; |
626 | } | |
627 | i = (i+1) * 8 * sizeof(long); | |
628 | return i; | |
629 | } | |
630 | ||
badf1662 DS |
631 | static struct files_struct *alloc_files(void) |
632 | { | |
633 | struct files_struct *newf; | |
634 | struct fdtable *fdt; | |
635 | ||
e94b1766 | 636 | newf = kmem_cache_alloc(files_cachep, GFP_KERNEL); |
badf1662 DS |
637 | if (!newf) |
638 | goto out; | |
639 | ||
640 | atomic_set(&newf->count, 1); | |
641 | ||
642 | spin_lock_init(&newf->file_lock); | |
0c9e63fd | 643 | newf->next_fd = 0; |
ab2af1f5 | 644 | fdt = &newf->fdtab; |
badf1662 | 645 | fdt->max_fds = NR_OPEN_DEFAULT; |
0c9e63fd ED |
646 | fdt->close_on_exec = (fd_set *)&newf->close_on_exec_init; |
647 | fdt->open_fds = (fd_set *)&newf->open_fds_init; | |
badf1662 | 648 | fdt->fd = &newf->fd_array[0]; |
ab2af1f5 | 649 | INIT_RCU_HEAD(&fdt->rcu); |
ab2af1f5 DS |
650 | fdt->next = NULL; |
651 | rcu_assign_pointer(newf->fdt, fdt); | |
badf1662 DS |
652 | out: |
653 | return newf; | |
654 | } | |
655 | ||
a016f338 JD |
656 | /* |
657 | * Allocate a new files structure and copy contents from the | |
658 | * passed in files structure. | |
6e667260 | 659 | * errorp will be valid only when the returned files_struct is NULL. |
a016f338 JD |
660 | */ |
661 | static struct files_struct *dup_fd(struct files_struct *oldf, int *errorp) | |
1da177e4 | 662 | { |
a016f338 | 663 | struct files_struct *newf; |
1da177e4 | 664 | struct file **old_fds, **new_fds; |
bbea9f69 | 665 | int open_files, size, i; |
badf1662 | 666 | struct fdtable *old_fdt, *new_fdt; |
1da177e4 | 667 | |
6e667260 | 668 | *errorp = -ENOMEM; |
badf1662 DS |
669 | newf = alloc_files(); |
670 | if (!newf) | |
1da177e4 LT |
671 | goto out; |
672 | ||
1da177e4 | 673 | spin_lock(&oldf->file_lock); |
badf1662 DS |
674 | old_fdt = files_fdtable(oldf); |
675 | new_fdt = files_fdtable(newf); | |
ab2af1f5 | 676 | open_files = count_open_files(old_fdt); |
1da177e4 LT |
677 | |
678 | /* | |
bbea9f69 VL |
679 | * Check whether we need to allocate a larger fd array and fd set. |
680 | * Note: we're not a clone task, so the open count won't change. | |
1da177e4 | 681 | */ |
badf1662 DS |
682 | if (open_files > new_fdt->max_fds) { |
683 | new_fdt->max_fds = 0; | |
1da177e4 LT |
684 | spin_unlock(&oldf->file_lock); |
685 | spin_lock(&newf->file_lock); | |
a016f338 | 686 | *errorp = expand_files(newf, open_files-1); |
1da177e4 | 687 | spin_unlock(&newf->file_lock); |
a016f338 | 688 | if (*errorp < 0) |
1da177e4 | 689 | goto out_release; |
ab2af1f5 DS |
690 | new_fdt = files_fdtable(newf); |
691 | /* | |
692 | * Reacquire the oldf lock and a pointer to its fd table | |
693 | * who knows it may have a new bigger fd table. We need | |
694 | * the latest pointer. | |
695 | */ | |
1da177e4 | 696 | spin_lock(&oldf->file_lock); |
ab2af1f5 | 697 | old_fdt = files_fdtable(oldf); |
1da177e4 LT |
698 | } |
699 | ||
badf1662 DS |
700 | old_fds = old_fdt->fd; |
701 | new_fds = new_fdt->fd; | |
1da177e4 | 702 | |
f3d19c90 VL |
703 | memcpy(new_fdt->open_fds->fds_bits, |
704 | old_fdt->open_fds->fds_bits, open_files/8); | |
705 | memcpy(new_fdt->close_on_exec->fds_bits, | |
706 | old_fdt->close_on_exec->fds_bits, open_files/8); | |
1da177e4 LT |
707 | |
708 | for (i = open_files; i != 0; i--) { | |
709 | struct file *f = *old_fds++; | |
710 | if (f) { | |
711 | get_file(f); | |
712 | } else { | |
713 | /* | |
714 | * The fd may be claimed in the fd bitmap but not yet | |
715 | * instantiated in the files array if a sibling thread | |
716 | * is partway through open(). So make sure that this | |
717 | * fd is available to the new process. | |
718 | */ | |
badf1662 | 719 | FD_CLR(open_files - i, new_fdt->open_fds); |
1da177e4 | 720 | } |
ab2af1f5 | 721 | rcu_assign_pointer(*new_fds++, f); |
1da177e4 LT |
722 | } |
723 | spin_unlock(&oldf->file_lock); | |
724 | ||
725 | /* compute the remainder to be cleared */ | |
badf1662 | 726 | size = (new_fdt->max_fds - open_files) * sizeof(struct file *); |
1da177e4 LT |
727 | |
728 | /* This is long word aligned thus could use a optimized version */ | |
729 | memset(new_fds, 0, size); | |
730 | ||
bbea9f69 VL |
731 | if (new_fdt->max_fds > open_files) { |
732 | int left = (new_fdt->max_fds-open_files)/8; | |
1da177e4 LT |
733 | int start = open_files / (8 * sizeof(unsigned long)); |
734 | ||
badf1662 DS |
735 | memset(&new_fdt->open_fds->fds_bits[start], 0, left); |
736 | memset(&new_fdt->close_on_exec->fds_bits[start], 0, left); | |
1da177e4 LT |
737 | } |
738 | ||
a016f338 | 739 | return newf; |
1da177e4 LT |
740 | |
741 | out_release: | |
1da177e4 | 742 | kmem_cache_free(files_cachep, newf); |
f3d19c90 | 743 | out: |
42862298 | 744 | return NULL; |
1da177e4 LT |
745 | } |
746 | ||
a016f338 JD |
747 | static int copy_files(unsigned long clone_flags, struct task_struct * tsk) |
748 | { | |
749 | struct files_struct *oldf, *newf; | |
750 | int error = 0; | |
751 | ||
752 | /* | |
753 | * A background process may not have any files ... | |
754 | */ | |
755 | oldf = current->files; | |
756 | if (!oldf) | |
757 | goto out; | |
758 | ||
759 | if (clone_flags & CLONE_FILES) { | |
760 | atomic_inc(&oldf->count); | |
761 | goto out; | |
762 | } | |
763 | ||
764 | /* | |
765 | * Note: we may be using current for both targets (See exec.c) | |
766 | * This works because we cache current->files (old) as oldf. Don't | |
767 | * break this. | |
768 | */ | |
769 | tsk->files = NULL; | |
a016f338 JD |
770 | newf = dup_fd(oldf, &error); |
771 | if (!newf) | |
772 | goto out; | |
773 | ||
774 | tsk->files = newf; | |
775 | error = 0; | |
776 | out: | |
777 | return error; | |
778 | } | |
779 | ||
1da177e4 LT |
780 | /* |
781 | * Helper to unshare the files of the current task. | |
782 | * We don't want to expose copy_files internals to | |
783 | * the exec layer of the kernel. | |
784 | */ | |
785 | ||
786 | int unshare_files(void) | |
787 | { | |
788 | struct files_struct *files = current->files; | |
789 | int rc; | |
790 | ||
910dea7f | 791 | BUG_ON(!files); |
1da177e4 LT |
792 | |
793 | /* This can race but the race causes us to copy when we don't | |
794 | need to and drop the copy */ | |
795 | if(atomic_read(&files->count) == 1) | |
796 | { | |
797 | atomic_inc(&files->count); | |
798 | return 0; | |
799 | } | |
800 | rc = copy_files(0, current); | |
801 | if(rc) | |
802 | current->files = files; | |
803 | return rc; | |
804 | } | |
805 | ||
806 | EXPORT_SYMBOL(unshare_files); | |
807 | ||
808 | static inline int copy_sighand(unsigned long clone_flags, struct task_struct * tsk) | |
809 | { | |
810 | struct sighand_struct *sig; | |
811 | ||
812 | if (clone_flags & (CLONE_SIGHAND | CLONE_THREAD)) { | |
813 | atomic_inc(¤t->sighand->count); | |
814 | return 0; | |
815 | } | |
816 | sig = kmem_cache_alloc(sighand_cachep, GFP_KERNEL); | |
e56d0903 | 817 | rcu_assign_pointer(tsk->sighand, sig); |
1da177e4 LT |
818 | if (!sig) |
819 | return -ENOMEM; | |
1da177e4 LT |
820 | atomic_set(&sig->count, 1); |
821 | memcpy(sig->action, current->sighand->action, sizeof(sig->action)); | |
822 | return 0; | |
823 | } | |
824 | ||
a7e5328a | 825 | void __cleanup_sighand(struct sighand_struct *sighand) |
c81addc9 | 826 | { |
c81addc9 ON |
827 | if (atomic_dec_and_test(&sighand->count)) |
828 | kmem_cache_free(sighand_cachep, sighand); | |
829 | } | |
830 | ||
1da177e4 LT |
831 | static inline int copy_signal(unsigned long clone_flags, struct task_struct * tsk) |
832 | { | |
833 | struct signal_struct *sig; | |
834 | int ret; | |
835 | ||
836 | if (clone_flags & CLONE_THREAD) { | |
837 | atomic_inc(¤t->signal->count); | |
838 | atomic_inc(¤t->signal->live); | |
839 | return 0; | |
840 | } | |
841 | sig = kmem_cache_alloc(signal_cachep, GFP_KERNEL); | |
842 | tsk->signal = sig; | |
843 | if (!sig) | |
844 | return -ENOMEM; | |
845 | ||
846 | ret = copy_thread_group_keys(tsk); | |
847 | if (ret < 0) { | |
848 | kmem_cache_free(signal_cachep, sig); | |
849 | return ret; | |
850 | } | |
851 | ||
852 | atomic_set(&sig->count, 1); | |
853 | atomic_set(&sig->live, 1); | |
854 | init_waitqueue_head(&sig->wait_chldexit); | |
855 | sig->flags = 0; | |
856 | sig->group_exit_code = 0; | |
857 | sig->group_exit_task = NULL; | |
858 | sig->group_stop_count = 0; | |
859 | sig->curr_target = NULL; | |
860 | init_sigpending(&sig->shared_pending); | |
861 | INIT_LIST_HEAD(&sig->posix_timers); | |
862 | ||
c9cb2e3d | 863 | hrtimer_init(&sig->real_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL); |
2ff678b8 | 864 | sig->it_real_incr.tv64 = 0; |
1da177e4 | 865 | sig->real_timer.function = it_real_fn; |
05cfb614 | 866 | sig->tsk = tsk; |
1da177e4 LT |
867 | |
868 | sig->it_virt_expires = cputime_zero; | |
869 | sig->it_virt_incr = cputime_zero; | |
870 | sig->it_prof_expires = cputime_zero; | |
871 | sig->it_prof_incr = cputime_zero; | |
872 | ||
1da177e4 | 873 | sig->leader = 0; /* session leadership doesn't inherit */ |
ab521dc0 | 874 | sig->tty_old_pgrp = NULL; |
1da177e4 LT |
875 | |
876 | sig->utime = sig->stime = sig->cutime = sig->cstime = cputime_zero; | |
877 | sig->nvcsw = sig->nivcsw = sig->cnvcsw = sig->cnivcsw = 0; | |
878 | sig->min_flt = sig->maj_flt = sig->cmin_flt = sig->cmaj_flt = 0; | |
6eaeeaba | 879 | sig->inblock = sig->oublock = sig->cinblock = sig->coublock = 0; |
1da177e4 LT |
880 | sig->sched_time = 0; |
881 | INIT_LIST_HEAD(&sig->cpu_timers[0]); | |
882 | INIT_LIST_HEAD(&sig->cpu_timers[1]); | |
883 | INIT_LIST_HEAD(&sig->cpu_timers[2]); | |
ad4ecbcb | 884 | taskstats_tgid_init(sig); |
1da177e4 LT |
885 | |
886 | task_lock(current->group_leader); | |
887 | memcpy(sig->rlim, current->signal->rlim, sizeof sig->rlim); | |
888 | task_unlock(current->group_leader); | |
889 | ||
890 | if (sig->rlim[RLIMIT_CPU].rlim_cur != RLIM_INFINITY) { | |
891 | /* | |
892 | * New sole thread in the process gets an expiry time | |
893 | * of the whole CPU time limit. | |
894 | */ | |
895 | tsk->it_prof_expires = | |
896 | secs_to_cputime(sig->rlim[RLIMIT_CPU].rlim_cur); | |
897 | } | |
0e464814 | 898 | acct_init_pacct(&sig->pacct); |
1da177e4 LT |
899 | |
900 | return 0; | |
901 | } | |
902 | ||
6b3934ef ON |
903 | void __cleanup_signal(struct signal_struct *sig) |
904 | { | |
905 | exit_thread_group_keys(sig); | |
906 | kmem_cache_free(signal_cachep, sig); | |
907 | } | |
908 | ||
909 | static inline void cleanup_signal(struct task_struct *tsk) | |
910 | { | |
911 | struct signal_struct *sig = tsk->signal; | |
912 | ||
913 | atomic_dec(&sig->live); | |
914 | ||
915 | if (atomic_dec_and_test(&sig->count)) | |
916 | __cleanup_signal(sig); | |
917 | } | |
918 | ||
1da177e4 LT |
919 | static inline void copy_flags(unsigned long clone_flags, struct task_struct *p) |
920 | { | |
921 | unsigned long new_flags = p->flags; | |
922 | ||
d1209d04 | 923 | new_flags &= ~(PF_SUPERPRIV | PF_NOFREEZE); |
1da177e4 LT |
924 | new_flags |= PF_FORKNOEXEC; |
925 | if (!(clone_flags & CLONE_PTRACE)) | |
926 | p->ptrace = 0; | |
927 | p->flags = new_flags; | |
928 | } | |
929 | ||
930 | asmlinkage long sys_set_tid_address(int __user *tidptr) | |
931 | { | |
932 | current->clear_child_tid = tidptr; | |
933 | ||
934 | return current->pid; | |
935 | } | |
936 | ||
23f78d4a IM |
937 | static inline void rt_mutex_init_task(struct task_struct *p) |
938 | { | |
23f78d4a | 939 | spin_lock_init(&p->pi_lock); |
e29e175b | 940 | #ifdef CONFIG_RT_MUTEXES |
23f78d4a IM |
941 | plist_head_init(&p->pi_waiters, &p->pi_lock); |
942 | p->pi_blocked_on = NULL; | |
23f78d4a IM |
943 | #endif |
944 | } | |
945 | ||
1da177e4 LT |
946 | /* |
947 | * This creates a new process as a copy of the old one, | |
948 | * but does not actually start it yet. | |
949 | * | |
950 | * It copies the registers, and all the appropriate | |
951 | * parts of the process environment (as per the clone | |
952 | * flags). The actual kick-off is left to the caller. | |
953 | */ | |
36c8b586 IM |
954 | static struct task_struct *copy_process(unsigned long clone_flags, |
955 | unsigned long stack_start, | |
956 | struct pt_regs *regs, | |
957 | unsigned long stack_size, | |
958 | int __user *parent_tidptr, | |
959 | int __user *child_tidptr, | |
85868995 | 960 | struct pid *pid) |
1da177e4 LT |
961 | { |
962 | int retval; | |
963 | struct task_struct *p = NULL; | |
964 | ||
965 | if ((clone_flags & (CLONE_NEWNS|CLONE_FS)) == (CLONE_NEWNS|CLONE_FS)) | |
966 | return ERR_PTR(-EINVAL); | |
967 | ||
968 | /* | |
969 | * Thread groups must share signals as well, and detached threads | |
970 | * can only be started up within the thread group. | |
971 | */ | |
972 | if ((clone_flags & CLONE_THREAD) && !(clone_flags & CLONE_SIGHAND)) | |
973 | return ERR_PTR(-EINVAL); | |
974 | ||
975 | /* | |
976 | * Shared signal handlers imply shared VM. By way of the above, | |
977 | * thread groups also imply shared VM. Blocking this case allows | |
978 | * for various simplifications in other code. | |
979 | */ | |
980 | if ((clone_flags & CLONE_SIGHAND) && !(clone_flags & CLONE_VM)) | |
981 | return ERR_PTR(-EINVAL); | |
982 | ||
983 | retval = security_task_create(clone_flags); | |
984 | if (retval) | |
985 | goto fork_out; | |
986 | ||
987 | retval = -ENOMEM; | |
988 | p = dup_task_struct(current); | |
989 | if (!p) | |
990 | goto fork_out; | |
991 | ||
bea493a0 PZ |
992 | rt_mutex_init_task(p); |
993 | ||
de30a2b3 IM |
994 | #ifdef CONFIG_TRACE_IRQFLAGS |
995 | DEBUG_LOCKS_WARN_ON(!p->hardirqs_enabled); | |
996 | DEBUG_LOCKS_WARN_ON(!p->softirqs_enabled); | |
997 | #endif | |
1da177e4 LT |
998 | retval = -EAGAIN; |
999 | if (atomic_read(&p->user->processes) >= | |
1000 | p->signal->rlim[RLIMIT_NPROC].rlim_cur) { | |
1001 | if (!capable(CAP_SYS_ADMIN) && !capable(CAP_SYS_RESOURCE) && | |
1002 | p->user != &root_user) | |
1003 | goto bad_fork_free; | |
1004 | } | |
1005 | ||
1006 | atomic_inc(&p->user->__count); | |
1007 | atomic_inc(&p->user->processes); | |
1008 | get_group_info(p->group_info); | |
1009 | ||
1010 | /* | |
1011 | * If multiple threads are within copy_process(), then this check | |
1012 | * triggers too late. This doesn't hurt, the check is only there | |
1013 | * to stop root fork bombs. | |
1014 | */ | |
1015 | if (nr_threads >= max_threads) | |
1016 | goto bad_fork_cleanup_count; | |
1017 | ||
a1261f54 | 1018 | if (!try_module_get(task_thread_info(p)->exec_domain->module)) |
1da177e4 LT |
1019 | goto bad_fork_cleanup_count; |
1020 | ||
1021 | if (p->binfmt && !try_module_get(p->binfmt->module)) | |
1022 | goto bad_fork_cleanup_put_domain; | |
1023 | ||
1024 | p->did_exec = 0; | |
ca74e92b | 1025 | delayacct_tsk_init(p); /* Must remain after dup_task_struct() */ |
1da177e4 | 1026 | copy_flags(clone_flags, p); |
85868995 | 1027 | p->pid = pid_nr(pid); |
1da177e4 LT |
1028 | retval = -EFAULT; |
1029 | if (clone_flags & CLONE_PARENT_SETTID) | |
1030 | if (put_user(p->pid, parent_tidptr)) | |
35df17c5 | 1031 | goto bad_fork_cleanup_delays_binfmt; |
1da177e4 | 1032 | |
1da177e4 LT |
1033 | INIT_LIST_HEAD(&p->children); |
1034 | INIT_LIST_HEAD(&p->sibling); | |
1035 | p->vfork_done = NULL; | |
1036 | spin_lock_init(&p->alloc_lock); | |
1da177e4 LT |
1037 | |
1038 | clear_tsk_thread_flag(p, TIF_SIGPENDING); | |
1039 | init_sigpending(&p->pending); | |
1040 | ||
1041 | p->utime = cputime_zero; | |
1042 | p->stime = cputime_zero; | |
1043 | p->sched_time = 0; | |
4b98d11b | 1044 | #ifdef CONFIG_TASK_XACCT |
1da177e4 LT |
1045 | p->rchar = 0; /* I/O counter: bytes read */ |
1046 | p->wchar = 0; /* I/O counter: bytes written */ | |
1047 | p->syscr = 0; /* I/O counter: read syscalls */ | |
1048 | p->syscw = 0; /* I/O counter: write syscalls */ | |
4b98d11b | 1049 | #endif |
7c3ab738 | 1050 | task_io_accounting_init(p); |
1da177e4 LT |
1051 | acct_clear_integrals(p); |
1052 | ||
1053 | p->it_virt_expires = cputime_zero; | |
1054 | p->it_prof_expires = cputime_zero; | |
1055 | p->it_sched_expires = 0; | |
1056 | INIT_LIST_HEAD(&p->cpu_timers[0]); | |
1057 | INIT_LIST_HEAD(&p->cpu_timers[1]); | |
1058 | INIT_LIST_HEAD(&p->cpu_timers[2]); | |
1059 | ||
1060 | p->lock_depth = -1; /* -1 = no lock */ | |
1061 | do_posix_clock_monotonic_gettime(&p->start_time); | |
1062 | p->security = NULL; | |
1063 | p->io_context = NULL; | |
1064 | p->io_wait = NULL; | |
1065 | p->audit_context = NULL; | |
b4b26418 | 1066 | cpuset_fork(p); |
1da177e4 LT |
1067 | #ifdef CONFIG_NUMA |
1068 | p->mempolicy = mpol_copy(p->mempolicy); | |
1069 | if (IS_ERR(p->mempolicy)) { | |
1070 | retval = PTR_ERR(p->mempolicy); | |
1071 | p->mempolicy = NULL; | |
b4b26418 | 1072 | goto bad_fork_cleanup_cpuset; |
1da177e4 | 1073 | } |
c61afb18 | 1074 | mpol_fix_fork_child_flag(p); |
1da177e4 | 1075 | #endif |
de30a2b3 IM |
1076 | #ifdef CONFIG_TRACE_IRQFLAGS |
1077 | p->irq_events = 0; | |
b36e4758 RK |
1078 | #ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW |
1079 | p->hardirqs_enabled = 1; | |
1080 | #else | |
de30a2b3 | 1081 | p->hardirqs_enabled = 0; |
b36e4758 | 1082 | #endif |
de30a2b3 IM |
1083 | p->hardirq_enable_ip = 0; |
1084 | p->hardirq_enable_event = 0; | |
1085 | p->hardirq_disable_ip = _THIS_IP_; | |
1086 | p->hardirq_disable_event = 0; | |
1087 | p->softirqs_enabled = 1; | |
1088 | p->softirq_enable_ip = _THIS_IP_; | |
1089 | p->softirq_enable_event = 0; | |
1090 | p->softirq_disable_ip = 0; | |
1091 | p->softirq_disable_event = 0; | |
1092 | p->hardirq_context = 0; | |
1093 | p->softirq_context = 0; | |
1094 | #endif | |
fbb9ce95 IM |
1095 | #ifdef CONFIG_LOCKDEP |
1096 | p->lockdep_depth = 0; /* no locks held yet */ | |
1097 | p->curr_chain_key = 0; | |
1098 | p->lockdep_recursion = 0; | |
1099 | #endif | |
1da177e4 | 1100 | |
408894ee IM |
1101 | #ifdef CONFIG_DEBUG_MUTEXES |
1102 | p->blocked_on = NULL; /* not blocked yet */ | |
1103 | #endif | |
1104 | ||
1da177e4 LT |
1105 | p->tgid = p->pid; |
1106 | if (clone_flags & CLONE_THREAD) | |
1107 | p->tgid = current->tgid; | |
1108 | ||
1109 | if ((retval = security_task_alloc(p))) | |
1110 | goto bad_fork_cleanup_policy; | |
1111 | if ((retval = audit_alloc(p))) | |
1112 | goto bad_fork_cleanup_security; | |
1113 | /* copy all the process information */ | |
1114 | if ((retval = copy_semundo(clone_flags, p))) | |
1115 | goto bad_fork_cleanup_audit; | |
1116 | if ((retval = copy_files(clone_flags, p))) | |
1117 | goto bad_fork_cleanup_semundo; | |
1118 | if ((retval = copy_fs(clone_flags, p))) | |
1119 | goto bad_fork_cleanup_files; | |
1120 | if ((retval = copy_sighand(clone_flags, p))) | |
1121 | goto bad_fork_cleanup_fs; | |
1122 | if ((retval = copy_signal(clone_flags, p))) | |
1123 | goto bad_fork_cleanup_sighand; | |
1124 | if ((retval = copy_mm(clone_flags, p))) | |
1125 | goto bad_fork_cleanup_signal; | |
1126 | if ((retval = copy_keys(clone_flags, p))) | |
1127 | goto bad_fork_cleanup_mm; | |
ab516013 | 1128 | if ((retval = copy_namespaces(clone_flags, p))) |
1da177e4 LT |
1129 | goto bad_fork_cleanup_keys; |
1130 | retval = copy_thread(0, clone_flags, stack_start, stack_size, p, regs); | |
1131 | if (retval) | |
1651e14e | 1132 | goto bad_fork_cleanup_namespaces; |
1da177e4 LT |
1133 | |
1134 | p->set_child_tid = (clone_flags & CLONE_CHILD_SETTID) ? child_tidptr : NULL; | |
1135 | /* | |
1136 | * Clear TID on mm_release()? | |
1137 | */ | |
1138 | p->clear_child_tid = (clone_flags & CLONE_CHILD_CLEARTID) ? child_tidptr: NULL; | |
8f17d3a5 IM |
1139 | p->robust_list = NULL; |
1140 | #ifdef CONFIG_COMPAT | |
1141 | p->compat_robust_list = NULL; | |
1142 | #endif | |
c87e2837 IM |
1143 | INIT_LIST_HEAD(&p->pi_state_list); |
1144 | p->pi_state_cache = NULL; | |
1145 | ||
f9a3879a GM |
1146 | /* |
1147 | * sigaltstack should be cleared when sharing the same VM | |
1148 | */ | |
1149 | if ((clone_flags & (CLONE_VM|CLONE_VFORK)) == CLONE_VM) | |
1150 | p->sas_ss_sp = p->sas_ss_size = 0; | |
1151 | ||
1da177e4 LT |
1152 | /* |
1153 | * Syscall tracing should be turned off in the child regardless | |
1154 | * of CLONE_PTRACE. | |
1155 | */ | |
1156 | clear_tsk_thread_flag(p, TIF_SYSCALL_TRACE); | |
ed75e8d5 LV |
1157 | #ifdef TIF_SYSCALL_EMU |
1158 | clear_tsk_thread_flag(p, TIF_SYSCALL_EMU); | |
1159 | #endif | |
1da177e4 LT |
1160 | |
1161 | /* Our parent execution domain becomes current domain | |
1162 | These must match for thread signalling to apply */ | |
1da177e4 LT |
1163 | p->parent_exec_id = p->self_exec_id; |
1164 | ||
1165 | /* ok, now we should be set up.. */ | |
1166 | p->exit_signal = (clone_flags & CLONE_THREAD) ? -1 : (clone_flags & CSIGNAL); | |
1167 | p->pdeath_signal = 0; | |
1168 | p->exit_state = 0; | |
1169 | ||
1da177e4 LT |
1170 | /* |
1171 | * Ok, make it visible to the rest of the system. | |
1172 | * We dont wake it up yet. | |
1173 | */ | |
1174 | p->group_leader = p; | |
47e65328 | 1175 | INIT_LIST_HEAD(&p->thread_group); |
1da177e4 LT |
1176 | INIT_LIST_HEAD(&p->ptrace_children); |
1177 | INIT_LIST_HEAD(&p->ptrace_list); | |
1178 | ||
476d139c NP |
1179 | /* Perform scheduler related setup. Assign this task to a CPU. */ |
1180 | sched_fork(p, clone_flags); | |
1181 | ||
1da177e4 LT |
1182 | /* Need tasklist lock for parent etc handling! */ |
1183 | write_lock_irq(&tasklist_lock); | |
1184 | ||
5b160f5e ON |
1185 | /* for sys_ioprio_set(IOPRIO_WHO_PGRP) */ |
1186 | p->ioprio = current->ioprio; | |
1187 | ||
1da177e4 | 1188 | /* |
476d139c NP |
1189 | * The task hasn't been attached yet, so its cpus_allowed mask will |
1190 | * not be changed, nor will its assigned CPU. | |
1191 | * | |
1192 | * The cpus_allowed mask of the parent may have changed after it was | |
1193 | * copied first time - so re-copy it here, then check the child's CPU | |
1194 | * to ensure it is on a valid CPU (and if not, just force it back to | |
1195 | * parent's CPU). This avoids alot of nasty races. | |
1da177e4 LT |
1196 | */ |
1197 | p->cpus_allowed = current->cpus_allowed; | |
26ff6ad9 SV |
1198 | if (unlikely(!cpu_isset(task_cpu(p), p->cpus_allowed) || |
1199 | !cpu_online(task_cpu(p)))) | |
476d139c | 1200 | set_task_cpu(p, smp_processor_id()); |
1da177e4 | 1201 | |
1da177e4 LT |
1202 | /* CLONE_PARENT re-uses the old parent */ |
1203 | if (clone_flags & (CLONE_PARENT|CLONE_THREAD)) | |
1204 | p->real_parent = current->real_parent; | |
1205 | else | |
1206 | p->real_parent = current; | |
1207 | p->parent = p->real_parent; | |
1208 | ||
3f17da69 | 1209 | spin_lock(¤t->sighand->siglock); |
4a2c7a78 ON |
1210 | |
1211 | /* | |
1212 | * Process group and session signals need to be delivered to just the | |
1213 | * parent before the fork or both the parent and the child after the | |
1214 | * fork. Restart if a signal comes in before we add the new process to | |
1215 | * it's process group. | |
1216 | * A fatal signal pending means that current will exit, so the new | |
1217 | * thread can't slip out of an OOM kill (or normal SIGKILL). | |
1218 | */ | |
1219 | recalc_sigpending(); | |
1220 | if (signal_pending(current)) { | |
1221 | spin_unlock(¤t->sighand->siglock); | |
1222 | write_unlock_irq(&tasklist_lock); | |
1223 | retval = -ERESTARTNOINTR; | |
1651e14e | 1224 | goto bad_fork_cleanup_namespaces; |
4a2c7a78 ON |
1225 | } |
1226 | ||
1da177e4 | 1227 | if (clone_flags & CLONE_THREAD) { |
1da177e4 | 1228 | p->group_leader = current->group_leader; |
47e65328 | 1229 | list_add_tail_rcu(&p->thread_group, &p->group_leader->thread_group); |
1da177e4 | 1230 | |
1da177e4 LT |
1231 | if (!cputime_eq(current->signal->it_virt_expires, |
1232 | cputime_zero) || | |
1233 | !cputime_eq(current->signal->it_prof_expires, | |
1234 | cputime_zero) || | |
1235 | current->signal->rlim[RLIMIT_CPU].rlim_cur != RLIM_INFINITY || | |
1236 | !list_empty(¤t->signal->cpu_timers[0]) || | |
1237 | !list_empty(¤t->signal->cpu_timers[1]) || | |
1238 | !list_empty(¤t->signal->cpu_timers[2])) { | |
1239 | /* | |
1240 | * Have child wake up on its first tick to check | |
1241 | * for process CPU timers. | |
1242 | */ | |
1243 | p->it_prof_expires = jiffies_to_cputime(1); | |
1244 | } | |
1da177e4 LT |
1245 | } |
1246 | ||
73b9ebfe ON |
1247 | if (likely(p->pid)) { |
1248 | add_parent(p); | |
1249 | if (unlikely(p->ptrace & PT_PTRACED)) | |
1250 | __ptrace_link(p, current->parent); | |
1251 | ||
1252 | if (thread_group_leader(p)) { | |
1253 | p->signal->tty = current->signal->tty; | |
0800d308 | 1254 | p->signal->pgrp = process_group(current); |
1ec320af | 1255 | set_signal_session(p->signal, process_session(current)); |
0800d308 SB |
1256 | attach_pid(p, PIDTYPE_PGID, task_pgrp(current)); |
1257 | attach_pid(p, PIDTYPE_SID, task_session(current)); | |
73b9ebfe | 1258 | |
5e85d4ab | 1259 | list_add_tail_rcu(&p->tasks, &init_task.tasks); |
1da177e4 | 1260 | __get_cpu_var(process_counts)++; |
73b9ebfe | 1261 | } |
85868995 | 1262 | attach_pid(p, PIDTYPE_PID, pid); |
73b9ebfe | 1263 | nr_threads++; |
1da177e4 LT |
1264 | } |
1265 | ||
1da177e4 | 1266 | total_forks++; |
3f17da69 | 1267 | spin_unlock(¤t->sighand->siglock); |
1da177e4 | 1268 | write_unlock_irq(&tasklist_lock); |
c13cf856 | 1269 | proc_fork_connector(p); |
1da177e4 LT |
1270 | return p; |
1271 | ||
ab516013 | 1272 | bad_fork_cleanup_namespaces: |
444f378b | 1273 | exit_task_namespaces(p); |
1da177e4 LT |
1274 | bad_fork_cleanup_keys: |
1275 | exit_keys(p); | |
1276 | bad_fork_cleanup_mm: | |
1277 | if (p->mm) | |
1278 | mmput(p->mm); | |
1279 | bad_fork_cleanup_signal: | |
6b3934ef | 1280 | cleanup_signal(p); |
1da177e4 | 1281 | bad_fork_cleanup_sighand: |
a7e5328a | 1282 | __cleanup_sighand(p->sighand); |
1da177e4 LT |
1283 | bad_fork_cleanup_fs: |
1284 | exit_fs(p); /* blocking */ | |
1285 | bad_fork_cleanup_files: | |
1286 | exit_files(p); /* blocking */ | |
1287 | bad_fork_cleanup_semundo: | |
1288 | exit_sem(p); | |
1289 | bad_fork_cleanup_audit: | |
1290 | audit_free(p); | |
1291 | bad_fork_cleanup_security: | |
1292 | security_task_free(p); | |
1293 | bad_fork_cleanup_policy: | |
1294 | #ifdef CONFIG_NUMA | |
1295 | mpol_free(p->mempolicy); | |
b4b26418 | 1296 | bad_fork_cleanup_cpuset: |
1da177e4 | 1297 | #endif |
b4b26418 | 1298 | cpuset_exit(p); |
35df17c5 SN |
1299 | bad_fork_cleanup_delays_binfmt: |
1300 | delayacct_tsk_free(p); | |
1da177e4 LT |
1301 | if (p->binfmt) |
1302 | module_put(p->binfmt->module); | |
1303 | bad_fork_cleanup_put_domain: | |
a1261f54 | 1304 | module_put(task_thread_info(p)->exec_domain->module); |
1da177e4 LT |
1305 | bad_fork_cleanup_count: |
1306 | put_group_info(p->group_info); | |
1307 | atomic_dec(&p->user->processes); | |
1308 | free_uid(p->user); | |
1309 | bad_fork_free: | |
1310 | free_task(p); | |
fe7d37d1 ON |
1311 | fork_out: |
1312 | return ERR_PTR(retval); | |
1da177e4 LT |
1313 | } |
1314 | ||
f95d47ca | 1315 | noinline struct pt_regs * __devinit __attribute__((weak)) idle_regs(struct pt_regs *regs) |
1da177e4 LT |
1316 | { |
1317 | memset(regs, 0, sizeof(struct pt_regs)); | |
1318 | return regs; | |
1319 | } | |
1320 | ||
9abcf40b | 1321 | struct task_struct * __cpuinit fork_idle(int cpu) |
1da177e4 | 1322 | { |
36c8b586 | 1323 | struct task_struct *task; |
1da177e4 LT |
1324 | struct pt_regs regs; |
1325 | ||
85868995 SB |
1326 | task = copy_process(CLONE_VM, 0, idle_regs(®s), 0, NULL, NULL, |
1327 | &init_struct_pid); | |
753ca4f3 AM |
1328 | if (!IS_ERR(task)) |
1329 | init_idle(task, cpu); | |
73b9ebfe | 1330 | |
1da177e4 LT |
1331 | return task; |
1332 | } | |
1333 | ||
1334 | static inline int fork_traceflag (unsigned clone_flags) | |
1335 | { | |
1336 | if (clone_flags & CLONE_UNTRACED) | |
1337 | return 0; | |
1338 | else if (clone_flags & CLONE_VFORK) { | |
1339 | if (current->ptrace & PT_TRACE_VFORK) | |
1340 | return PTRACE_EVENT_VFORK; | |
1341 | } else if ((clone_flags & CSIGNAL) != SIGCHLD) { | |
1342 | if (current->ptrace & PT_TRACE_CLONE) | |
1343 | return PTRACE_EVENT_CLONE; | |
1344 | } else if (current->ptrace & PT_TRACE_FORK) | |
1345 | return PTRACE_EVENT_FORK; | |
1346 | ||
1347 | return 0; | |
1348 | } | |
1349 | ||
1350 | /* | |
1351 | * Ok, this is the main fork-routine. | |
1352 | * | |
1353 | * It copies the process, and if successful kick-starts | |
1354 | * it and waits for it to finish using the VM if required. | |
1355 | */ | |
1356 | long do_fork(unsigned long clone_flags, | |
1357 | unsigned long stack_start, | |
1358 | struct pt_regs *regs, | |
1359 | unsigned long stack_size, | |
1360 | int __user *parent_tidptr, | |
1361 | int __user *child_tidptr) | |
1362 | { | |
1363 | struct task_struct *p; | |
1364 | int trace = 0; | |
92476d7f EB |
1365 | struct pid *pid = alloc_pid(); |
1366 | long nr; | |
1da177e4 | 1367 | |
92476d7f | 1368 | if (!pid) |
1da177e4 | 1369 | return -EAGAIN; |
92476d7f | 1370 | nr = pid->nr; |
1da177e4 LT |
1371 | if (unlikely(current->ptrace)) { |
1372 | trace = fork_traceflag (clone_flags); | |
1373 | if (trace) | |
1374 | clone_flags |= CLONE_PTRACE; | |
1375 | } | |
1376 | ||
85868995 | 1377 | p = copy_process(clone_flags, stack_start, regs, stack_size, parent_tidptr, child_tidptr, pid); |
1da177e4 LT |
1378 | /* |
1379 | * Do this prior waking up the new thread - the thread pointer | |
1380 | * might get invalid after that point, if the thread exits quickly. | |
1381 | */ | |
1382 | if (!IS_ERR(p)) { | |
1383 | struct completion vfork; | |
1384 | ||
1385 | if (clone_flags & CLONE_VFORK) { | |
1386 | p->vfork_done = &vfork; | |
1387 | init_completion(&vfork); | |
1388 | } | |
1389 | ||
1390 | if ((p->ptrace & PT_PTRACED) || (clone_flags & CLONE_STOPPED)) { | |
1391 | /* | |
1392 | * We'll start up with an immediate SIGSTOP. | |
1393 | */ | |
1394 | sigaddset(&p->pending.signal, SIGSTOP); | |
1395 | set_tsk_thread_flag(p, TIF_SIGPENDING); | |
1396 | } | |
1397 | ||
1398 | if (!(clone_flags & CLONE_STOPPED)) | |
1399 | wake_up_new_task(p, clone_flags); | |
1400 | else | |
1401 | p->state = TASK_STOPPED; | |
1402 | ||
1403 | if (unlikely (trace)) { | |
92476d7f | 1404 | current->ptrace_message = nr; |
1da177e4 LT |
1405 | ptrace_notify ((trace << 8) | SIGTRAP); |
1406 | } | |
1407 | ||
1408 | if (clone_flags & CLONE_VFORK) { | |
ba96a0c8 | 1409 | freezer_do_not_count(); |
1da177e4 | 1410 | wait_for_completion(&vfork); |
ba96a0c8 | 1411 | freezer_count(); |
9f59ce5d CE |
1412 | if (unlikely (current->ptrace & PT_TRACE_VFORK_DONE)) { |
1413 | current->ptrace_message = nr; | |
1da177e4 | 1414 | ptrace_notify ((PTRACE_EVENT_VFORK_DONE << 8) | SIGTRAP); |
9f59ce5d | 1415 | } |
1da177e4 LT |
1416 | } |
1417 | } else { | |
92476d7f EB |
1418 | free_pid(pid); |
1419 | nr = PTR_ERR(p); | |
1da177e4 | 1420 | } |
92476d7f | 1421 | return nr; |
1da177e4 LT |
1422 | } |
1423 | ||
5fd63b30 RT |
1424 | #ifndef ARCH_MIN_MMSTRUCT_ALIGN |
1425 | #define ARCH_MIN_MMSTRUCT_ALIGN 0 | |
1426 | #endif | |
1427 | ||
fba2afaa DL |
1428 | static void sighand_ctor(void *data, struct kmem_cache *cachep, |
1429 | unsigned long flags) | |
aa1757f9 ON |
1430 | { |
1431 | struct sighand_struct *sighand = data; | |
1432 | ||
a35afb83 CL |
1433 | spin_lock_init(&sighand->siglock); |
1434 | INIT_LIST_HEAD(&sighand->signalfd_list); | |
aa1757f9 ON |
1435 | } |
1436 | ||
1da177e4 LT |
1437 | void __init proc_caches_init(void) |
1438 | { | |
1439 | sighand_cachep = kmem_cache_create("sighand_cache", | |
1440 | sizeof(struct sighand_struct), 0, | |
aa1757f9 ON |
1441 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_DESTROY_BY_RCU, |
1442 | sighand_ctor, NULL); | |
1da177e4 LT |
1443 | signal_cachep = kmem_cache_create("signal_cache", |
1444 | sizeof(struct signal_struct), 0, | |
1445 | SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL, NULL); | |
1446 | files_cachep = kmem_cache_create("files_cache", | |
1447 | sizeof(struct files_struct), 0, | |
1448 | SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL, NULL); | |
1449 | fs_cachep = kmem_cache_create("fs_cache", | |
1450 | sizeof(struct fs_struct), 0, | |
1451 | SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL, NULL); | |
1452 | vm_area_cachep = kmem_cache_create("vm_area_struct", | |
1453 | sizeof(struct vm_area_struct), 0, | |
1454 | SLAB_PANIC, NULL, NULL); | |
1455 | mm_cachep = kmem_cache_create("mm_struct", | |
5fd63b30 | 1456 | sizeof(struct mm_struct), ARCH_MIN_MMSTRUCT_ALIGN, |
1da177e4 LT |
1457 | SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL, NULL); |
1458 | } | |
cf2e340f | 1459 | |
cf2e340f JD |
1460 | /* |
1461 | * Check constraints on flags passed to the unshare system call and | |
1462 | * force unsharing of additional process context as appropriate. | |
1463 | */ | |
1464 | static inline void check_unshare_flags(unsigned long *flags_ptr) | |
1465 | { | |
1466 | /* | |
1467 | * If unsharing a thread from a thread group, must also | |
1468 | * unshare vm. | |
1469 | */ | |
1470 | if (*flags_ptr & CLONE_THREAD) | |
1471 | *flags_ptr |= CLONE_VM; | |
1472 | ||
1473 | /* | |
1474 | * If unsharing vm, must also unshare signal handlers. | |
1475 | */ | |
1476 | if (*flags_ptr & CLONE_VM) | |
1477 | *flags_ptr |= CLONE_SIGHAND; | |
1478 | ||
1479 | /* | |
1480 | * If unsharing signal handlers and the task was created | |
1481 | * using CLONE_THREAD, then must unshare the thread | |
1482 | */ | |
1483 | if ((*flags_ptr & CLONE_SIGHAND) && | |
1484 | (atomic_read(¤t->signal->count) > 1)) | |
1485 | *flags_ptr |= CLONE_THREAD; | |
1486 | ||
1487 | /* | |
1488 | * If unsharing namespace, must also unshare filesystem information. | |
1489 | */ | |
1490 | if (*flags_ptr & CLONE_NEWNS) | |
1491 | *flags_ptr |= CLONE_FS; | |
1492 | } | |
1493 | ||
1494 | /* | |
1495 | * Unsharing of tasks created with CLONE_THREAD is not supported yet | |
1496 | */ | |
1497 | static int unshare_thread(unsigned long unshare_flags) | |
1498 | { | |
1499 | if (unshare_flags & CLONE_THREAD) | |
1500 | return -EINVAL; | |
1501 | ||
1502 | return 0; | |
1503 | } | |
1504 | ||
1505 | /* | |
99d1419d | 1506 | * Unshare the filesystem structure if it is being shared |
cf2e340f JD |
1507 | */ |
1508 | static int unshare_fs(unsigned long unshare_flags, struct fs_struct **new_fsp) | |
1509 | { | |
1510 | struct fs_struct *fs = current->fs; | |
1511 | ||
1512 | if ((unshare_flags & CLONE_FS) && | |
99d1419d JD |
1513 | (fs && atomic_read(&fs->count) > 1)) { |
1514 | *new_fsp = __copy_fs_struct(current->fs); | |
1515 | if (!*new_fsp) | |
1516 | return -ENOMEM; | |
1517 | } | |
cf2e340f JD |
1518 | |
1519 | return 0; | |
1520 | } | |
1521 | ||
cf2e340f | 1522 | /* |
dae3c5a0 | 1523 | * Unsharing of sighand is not supported yet |
cf2e340f JD |
1524 | */ |
1525 | static int unshare_sighand(unsigned long unshare_flags, struct sighand_struct **new_sighp) | |
1526 | { | |
1527 | struct sighand_struct *sigh = current->sighand; | |
1528 | ||
dae3c5a0 | 1529 | if ((unshare_flags & CLONE_SIGHAND) && atomic_read(&sigh->count) > 1) |
cf2e340f JD |
1530 | return -EINVAL; |
1531 | else | |
1532 | return 0; | |
1533 | } | |
1534 | ||
1535 | /* | |
a0a7ec30 | 1536 | * Unshare vm if it is being shared |
cf2e340f JD |
1537 | */ |
1538 | static int unshare_vm(unsigned long unshare_flags, struct mm_struct **new_mmp) | |
1539 | { | |
1540 | struct mm_struct *mm = current->mm; | |
1541 | ||
1542 | if ((unshare_flags & CLONE_VM) && | |
a0a7ec30 | 1543 | (mm && atomic_read(&mm->mm_users) > 1)) { |
2d61b867 | 1544 | return -EINVAL; |
a0a7ec30 | 1545 | } |
cf2e340f JD |
1546 | |
1547 | return 0; | |
cf2e340f JD |
1548 | } |
1549 | ||
1550 | /* | |
a016f338 | 1551 | * Unshare file descriptor table if it is being shared |
cf2e340f JD |
1552 | */ |
1553 | static int unshare_fd(unsigned long unshare_flags, struct files_struct **new_fdp) | |
1554 | { | |
1555 | struct files_struct *fd = current->files; | |
a016f338 | 1556 | int error = 0; |
cf2e340f JD |
1557 | |
1558 | if ((unshare_flags & CLONE_FILES) && | |
a016f338 JD |
1559 | (fd && atomic_read(&fd->count) > 1)) { |
1560 | *new_fdp = dup_fd(fd, &error); | |
1561 | if (!*new_fdp) | |
1562 | return error; | |
1563 | } | |
cf2e340f JD |
1564 | |
1565 | return 0; | |
1566 | } | |
1567 | ||
1568 | /* | |
1569 | * Unsharing of semundo for tasks created with CLONE_SYSVSEM is not | |
1570 | * supported yet | |
1571 | */ | |
1572 | static int unshare_semundo(unsigned long unshare_flags, struct sem_undo_list **new_ulistp) | |
1573 | { | |
1574 | if (unshare_flags & CLONE_SYSVSEM) | |
1575 | return -EINVAL; | |
1576 | ||
1577 | return 0; | |
1578 | } | |
1579 | ||
1580 | /* | |
1581 | * unshare allows a process to 'unshare' part of the process | |
1582 | * context which was originally shared using clone. copy_* | |
1583 | * functions used by do_fork() cannot be used here directly | |
1584 | * because they modify an inactive task_struct that is being | |
1585 | * constructed. Here we are modifying the current, active, | |
1586 | * task_struct. | |
1587 | */ | |
1588 | asmlinkage long sys_unshare(unsigned long unshare_flags) | |
1589 | { | |
1590 | int err = 0; | |
1591 | struct fs_struct *fs, *new_fs = NULL; | |
dae3c5a0 | 1592 | struct sighand_struct *new_sigh = NULL; |
cf2e340f JD |
1593 | struct mm_struct *mm, *new_mm = NULL, *active_mm = NULL; |
1594 | struct files_struct *fd, *new_fd = NULL; | |
1595 | struct sem_undo_list *new_ulist = NULL; | |
c0b2fc31 | 1596 | struct nsproxy *new_nsproxy = NULL, *old_nsproxy = NULL; |
cf2e340f JD |
1597 | |
1598 | check_unshare_flags(&unshare_flags); | |
1599 | ||
06f9d4f9 EB |
1600 | /* Return -EINVAL for all unsupported flags */ |
1601 | err = -EINVAL; | |
1602 | if (unshare_flags & ~(CLONE_THREAD|CLONE_FS|CLONE_NEWNS|CLONE_SIGHAND| | |
25b21cb2 KK |
1603 | CLONE_VM|CLONE_FILES|CLONE_SYSVSEM| |
1604 | CLONE_NEWUTS|CLONE_NEWIPC)) | |
06f9d4f9 EB |
1605 | goto bad_unshare_out; |
1606 | ||
cf2e340f JD |
1607 | if ((err = unshare_thread(unshare_flags))) |
1608 | goto bad_unshare_out; | |
1609 | if ((err = unshare_fs(unshare_flags, &new_fs))) | |
1610 | goto bad_unshare_cleanup_thread; | |
cf2e340f | 1611 | if ((err = unshare_sighand(unshare_flags, &new_sigh))) |
e3222c4e | 1612 | goto bad_unshare_cleanup_fs; |
cf2e340f JD |
1613 | if ((err = unshare_vm(unshare_flags, &new_mm))) |
1614 | goto bad_unshare_cleanup_sigh; | |
1615 | if ((err = unshare_fd(unshare_flags, &new_fd))) | |
1616 | goto bad_unshare_cleanup_vm; | |
1617 | if ((err = unshare_semundo(unshare_flags, &new_ulist))) | |
1618 | goto bad_unshare_cleanup_fd; | |
e3222c4e BP |
1619 | if ((err = unshare_nsproxy_namespaces(unshare_flags, &new_nsproxy, |
1620 | new_fs))) | |
071df104 | 1621 | goto bad_unshare_cleanup_semundo; |
c0b2fc31 | 1622 | |
e3222c4e | 1623 | if (new_fs || new_mm || new_fd || new_ulist || new_nsproxy) { |
ab516013 | 1624 | |
cf2e340f | 1625 | task_lock(current); |
c0b2fc31 SH |
1626 | |
1627 | if (new_nsproxy) { | |
e3222c4e | 1628 | old_nsproxy = current->nsproxy; |
c0b2fc31 SH |
1629 | current->nsproxy = new_nsproxy; |
1630 | new_nsproxy = old_nsproxy; | |
1631 | } | |
cf2e340f JD |
1632 | |
1633 | if (new_fs) { | |
1634 | fs = current->fs; | |
1635 | current->fs = new_fs; | |
1636 | new_fs = fs; | |
1637 | } | |
1638 | ||
cf2e340f JD |
1639 | if (new_mm) { |
1640 | mm = current->mm; | |
1641 | active_mm = current->active_mm; | |
1642 | current->mm = new_mm; | |
1643 | current->active_mm = new_mm; | |
1644 | activate_mm(active_mm, new_mm); | |
1645 | new_mm = mm; | |
1646 | } | |
1647 | ||
1648 | if (new_fd) { | |
1649 | fd = current->files; | |
1650 | current->files = new_fd; | |
1651 | new_fd = fd; | |
1652 | } | |
1653 | ||
1654 | task_unlock(current); | |
1655 | } | |
1656 | ||
c0b2fc31 | 1657 | if (new_nsproxy) |
444f378b | 1658 | put_nsproxy(new_nsproxy); |
c0b2fc31 | 1659 | |
ab516013 | 1660 | bad_unshare_cleanup_semundo: |
cf2e340f JD |
1661 | bad_unshare_cleanup_fd: |
1662 | if (new_fd) | |
1663 | put_files_struct(new_fd); | |
1664 | ||
1665 | bad_unshare_cleanup_vm: | |
1666 | if (new_mm) | |
1667 | mmput(new_mm); | |
1668 | ||
1669 | bad_unshare_cleanup_sigh: | |
1670 | if (new_sigh) | |
1671 | if (atomic_dec_and_test(&new_sigh->count)) | |
1672 | kmem_cache_free(sighand_cachep, new_sigh); | |
1673 | ||
cf2e340f JD |
1674 | bad_unshare_cleanup_fs: |
1675 | if (new_fs) | |
1676 | put_fs_struct(new_fs); | |
1677 | ||
1678 | bad_unshare_cleanup_thread: | |
1679 | bad_unshare_out: | |
1680 | return err; | |
1681 | } |