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> |
1da177e4 LT |
54 | |
55 | #include <asm/pgtable.h> | |
56 | #include <asm/pgalloc.h> | |
57 | #include <asm/uaccess.h> | |
58 | #include <asm/mmu_context.h> | |
59 | #include <asm/cacheflush.h> | |
60 | #include <asm/tlbflush.h> | |
61 | ||
62 | /* | |
63 | * Protected counters by write_lock_irq(&tasklist_lock) | |
64 | */ | |
65 | unsigned long total_forks; /* Handle normal Linux uptimes. */ | |
66 | int nr_threads; /* The idle threads do not count.. */ | |
67 | ||
68 | int max_threads; /* tunable limit on nr_threads */ | |
69 | ||
70 | DEFINE_PER_CPU(unsigned long, process_counts) = 0; | |
71 | ||
c59923a1 | 72 | __cacheline_aligned DEFINE_RWLOCK(tasklist_lock); /* outer */ |
1da177e4 LT |
73 | |
74 | int nr_processes(void) | |
75 | { | |
76 | int cpu; | |
77 | int total = 0; | |
78 | ||
79 | for_each_online_cpu(cpu) | |
80 | total += per_cpu(process_counts, cpu); | |
81 | ||
82 | return total; | |
83 | } | |
84 | ||
85 | #ifndef __HAVE_ARCH_TASK_STRUCT_ALLOCATOR | |
86 | # define alloc_task_struct() kmem_cache_alloc(task_struct_cachep, GFP_KERNEL) | |
87 | # define free_task_struct(tsk) kmem_cache_free(task_struct_cachep, (tsk)) | |
e18b890b | 88 | static struct kmem_cache *task_struct_cachep; |
1da177e4 LT |
89 | #endif |
90 | ||
91 | /* SLAB cache for signal_struct structures (tsk->signal) */ | |
e18b890b | 92 | static struct kmem_cache *signal_cachep; |
1da177e4 LT |
93 | |
94 | /* SLAB cache for sighand_struct structures (tsk->sighand) */ | |
e18b890b | 95 | struct kmem_cache *sighand_cachep; |
1da177e4 LT |
96 | |
97 | /* SLAB cache for files_struct structures (tsk->files) */ | |
e18b890b | 98 | struct kmem_cache *files_cachep; |
1da177e4 LT |
99 | |
100 | /* SLAB cache for fs_struct structures (tsk->fs) */ | |
e18b890b | 101 | struct kmem_cache *fs_cachep; |
1da177e4 LT |
102 | |
103 | /* SLAB cache for vm_area_struct structures */ | |
e18b890b | 104 | struct kmem_cache *vm_area_cachep; |
1da177e4 LT |
105 | |
106 | /* SLAB cache for mm_struct structures (tsk->mm) */ | |
e18b890b | 107 | static struct kmem_cache *mm_cachep; |
1da177e4 LT |
108 | |
109 | void free_task(struct task_struct *tsk) | |
110 | { | |
3e26c149 | 111 | prop_local_destroy_single(&tsk->dirties); |
f7e4217b | 112 | free_thread_info(tsk->stack); |
23f78d4a | 113 | rt_mutex_debug_task_free(tsk); |
1da177e4 LT |
114 | free_task_struct(tsk); |
115 | } | |
116 | EXPORT_SYMBOL(free_task); | |
117 | ||
158d9ebd | 118 | void __put_task_struct(struct task_struct *tsk) |
1da177e4 LT |
119 | { |
120 | WARN_ON(!(tsk->exit_state & (EXIT_DEAD | EXIT_ZOMBIE))); | |
121 | WARN_ON(atomic_read(&tsk->usage)); | |
122 | WARN_ON(tsk == current); | |
123 | ||
1da177e4 LT |
124 | security_task_free(tsk); |
125 | free_uid(tsk->user); | |
126 | put_group_info(tsk->group_info); | |
35df17c5 | 127 | delayacct_tsk_free(tsk); |
1da177e4 LT |
128 | |
129 | if (!profile_handoff_task(tsk)) | |
130 | free_task(tsk); | |
131 | } | |
132 | ||
133 | void __init fork_init(unsigned long mempages) | |
134 | { | |
135 | #ifndef __HAVE_ARCH_TASK_STRUCT_ALLOCATOR | |
136 | #ifndef ARCH_MIN_TASKALIGN | |
137 | #define ARCH_MIN_TASKALIGN L1_CACHE_BYTES | |
138 | #endif | |
139 | /* create a slab on which task_structs can be allocated */ | |
140 | task_struct_cachep = | |
141 | kmem_cache_create("task_struct", sizeof(struct task_struct), | |
20c2df83 | 142 | ARCH_MIN_TASKALIGN, SLAB_PANIC, NULL); |
1da177e4 LT |
143 | #endif |
144 | ||
145 | /* | |
146 | * The default maximum number of threads is set to a safe | |
147 | * value: the thread structures can take up at most half | |
148 | * of memory. | |
149 | */ | |
150 | max_threads = mempages / (8 * THREAD_SIZE / PAGE_SIZE); | |
151 | ||
152 | /* | |
153 | * we need to allow at least 20 threads to boot a system | |
154 | */ | |
155 | if(max_threads < 20) | |
156 | max_threads = 20; | |
157 | ||
158 | init_task.signal->rlim[RLIMIT_NPROC].rlim_cur = max_threads/2; | |
159 | init_task.signal->rlim[RLIMIT_NPROC].rlim_max = max_threads/2; | |
160 | init_task.signal->rlim[RLIMIT_SIGPENDING] = | |
161 | init_task.signal->rlim[RLIMIT_NPROC]; | |
162 | } | |
163 | ||
164 | static struct task_struct *dup_task_struct(struct task_struct *orig) | |
165 | { | |
166 | struct task_struct *tsk; | |
167 | struct thread_info *ti; | |
3e26c149 | 168 | int err; |
1da177e4 LT |
169 | |
170 | prepare_to_copy(orig); | |
171 | ||
172 | tsk = alloc_task_struct(); | |
173 | if (!tsk) | |
174 | return NULL; | |
175 | ||
176 | ti = alloc_thread_info(tsk); | |
177 | if (!ti) { | |
178 | free_task_struct(tsk); | |
179 | return NULL; | |
180 | } | |
181 | ||
1da177e4 | 182 | *tsk = *orig; |
f7e4217b | 183 | tsk->stack = ti; |
3e26c149 PZ |
184 | |
185 | err = prop_local_init_single(&tsk->dirties); | |
186 | if (err) { | |
187 | free_thread_info(ti); | |
188 | free_task_struct(tsk); | |
189 | return NULL; | |
190 | } | |
191 | ||
10ebffde | 192 | setup_thread_stack(tsk, orig); |
1da177e4 | 193 | |
0a425405 AV |
194 | #ifdef CONFIG_CC_STACKPROTECTOR |
195 | tsk->stack_canary = get_random_int(); | |
196 | #endif | |
197 | ||
1da177e4 LT |
198 | /* One for us, one for whoever does the "release_task()" (usually parent) */ |
199 | atomic_set(&tsk->usage,2); | |
4b5d37ac | 200 | atomic_set(&tsk->fs_excl, 0); |
6c5c9341 | 201 | #ifdef CONFIG_BLK_DEV_IO_TRACE |
2056a782 | 202 | tsk->btrace_seq = 0; |
6c5c9341 | 203 | #endif |
a0aa7f68 | 204 | tsk->splice_pipe = NULL; |
1da177e4 LT |
205 | return tsk; |
206 | } | |
207 | ||
208 | #ifdef CONFIG_MMU | |
fd3e42fc | 209 | static inline int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm) |
1da177e4 | 210 | { |
fd3e42fc | 211 | struct vm_area_struct *mpnt, *tmp, **pprev; |
1da177e4 LT |
212 | struct rb_node **rb_link, *rb_parent; |
213 | int retval; | |
214 | unsigned long charge; | |
215 | struct mempolicy *pol; | |
216 | ||
217 | down_write(&oldmm->mmap_sem); | |
ec8c0446 | 218 | flush_cache_dup_mm(oldmm); |
ad339451 IM |
219 | /* |
220 | * Not linked in yet - no deadlock potential: | |
221 | */ | |
222 | down_write_nested(&mm->mmap_sem, SINGLE_DEPTH_NESTING); | |
7ee78232 | 223 | |
1da177e4 LT |
224 | mm->locked_vm = 0; |
225 | mm->mmap = NULL; | |
226 | mm->mmap_cache = NULL; | |
227 | mm->free_area_cache = oldmm->mmap_base; | |
1363c3cd | 228 | mm->cached_hole_size = ~0UL; |
1da177e4 | 229 | mm->map_count = 0; |
1da177e4 LT |
230 | cpus_clear(mm->cpu_vm_mask); |
231 | mm->mm_rb = RB_ROOT; | |
232 | rb_link = &mm->mm_rb.rb_node; | |
233 | rb_parent = NULL; | |
234 | pprev = &mm->mmap; | |
235 | ||
fd3e42fc | 236 | for (mpnt = oldmm->mmap; mpnt; mpnt = mpnt->vm_next) { |
1da177e4 LT |
237 | struct file *file; |
238 | ||
239 | if (mpnt->vm_flags & VM_DONTCOPY) { | |
3b6bfcdb HD |
240 | long pages = vma_pages(mpnt); |
241 | mm->total_vm -= pages; | |
ab50b8ed | 242 | vm_stat_account(mm, mpnt->vm_flags, mpnt->vm_file, |
3b6bfcdb | 243 | -pages); |
1da177e4 LT |
244 | continue; |
245 | } | |
246 | charge = 0; | |
247 | if (mpnt->vm_flags & VM_ACCOUNT) { | |
248 | unsigned int len = (mpnt->vm_end - mpnt->vm_start) >> PAGE_SHIFT; | |
249 | if (security_vm_enough_memory(len)) | |
250 | goto fail_nomem; | |
251 | charge = len; | |
252 | } | |
e94b1766 | 253 | tmp = kmem_cache_alloc(vm_area_cachep, GFP_KERNEL); |
1da177e4 LT |
254 | if (!tmp) |
255 | goto fail_nomem; | |
256 | *tmp = *mpnt; | |
257 | pol = mpol_copy(vma_policy(mpnt)); | |
258 | retval = PTR_ERR(pol); | |
259 | if (IS_ERR(pol)) | |
260 | goto fail_nomem_policy; | |
261 | vma_set_policy(tmp, pol); | |
262 | tmp->vm_flags &= ~VM_LOCKED; | |
263 | tmp->vm_mm = mm; | |
264 | tmp->vm_next = NULL; | |
265 | anon_vma_link(tmp); | |
266 | file = tmp->vm_file; | |
267 | if (file) { | |
f3a43f3f | 268 | struct inode *inode = file->f_path.dentry->d_inode; |
1da177e4 LT |
269 | get_file(file); |
270 | if (tmp->vm_flags & VM_DENYWRITE) | |
271 | atomic_dec(&inode->i_writecount); | |
23ff4440 | 272 | |
1da177e4 LT |
273 | /* insert tmp into the share list, just after mpnt */ |
274 | spin_lock(&file->f_mapping->i_mmap_lock); | |
275 | tmp->vm_truncate_count = mpnt->vm_truncate_count; | |
276 | flush_dcache_mmap_lock(file->f_mapping); | |
277 | vma_prio_tree_add(tmp, mpnt); | |
278 | flush_dcache_mmap_unlock(file->f_mapping); | |
279 | spin_unlock(&file->f_mapping->i_mmap_lock); | |
280 | } | |
281 | ||
282 | /* | |
7ee78232 | 283 | * Link in the new vma and copy the page table entries. |
1da177e4 | 284 | */ |
1da177e4 LT |
285 | *pprev = tmp; |
286 | pprev = &tmp->vm_next; | |
287 | ||
288 | __vma_link_rb(mm, tmp, rb_link, rb_parent); | |
289 | rb_link = &tmp->vm_rb.rb_right; | |
290 | rb_parent = &tmp->vm_rb; | |
291 | ||
292 | mm->map_count++; | |
0b0db14c | 293 | retval = copy_page_range(mm, oldmm, mpnt); |
1da177e4 LT |
294 | |
295 | if (tmp->vm_ops && tmp->vm_ops->open) | |
296 | tmp->vm_ops->open(tmp); | |
297 | ||
298 | if (retval) | |
299 | goto out; | |
300 | } | |
d6dd61c8 JF |
301 | /* a new mm has just been created */ |
302 | arch_dup_mmap(oldmm, mm); | |
1da177e4 | 303 | retval = 0; |
1da177e4 | 304 | out: |
7ee78232 | 305 | up_write(&mm->mmap_sem); |
fd3e42fc | 306 | flush_tlb_mm(oldmm); |
1da177e4 LT |
307 | up_write(&oldmm->mmap_sem); |
308 | return retval; | |
309 | fail_nomem_policy: | |
310 | kmem_cache_free(vm_area_cachep, tmp); | |
311 | fail_nomem: | |
312 | retval = -ENOMEM; | |
313 | vm_unacct_memory(charge); | |
314 | goto out; | |
315 | } | |
316 | ||
317 | static inline int mm_alloc_pgd(struct mm_struct * mm) | |
318 | { | |
319 | mm->pgd = pgd_alloc(mm); | |
320 | if (unlikely(!mm->pgd)) | |
321 | return -ENOMEM; | |
322 | return 0; | |
323 | } | |
324 | ||
325 | static inline void mm_free_pgd(struct mm_struct * mm) | |
326 | { | |
327 | pgd_free(mm->pgd); | |
328 | } | |
329 | #else | |
330 | #define dup_mmap(mm, oldmm) (0) | |
331 | #define mm_alloc_pgd(mm) (0) | |
332 | #define mm_free_pgd(mm) | |
333 | #endif /* CONFIG_MMU */ | |
334 | ||
23ff4440 | 335 | __cacheline_aligned_in_smp DEFINE_SPINLOCK(mmlist_lock); |
1da177e4 | 336 | |
e94b1766 | 337 | #define allocate_mm() (kmem_cache_alloc(mm_cachep, GFP_KERNEL)) |
1da177e4 LT |
338 | #define free_mm(mm) (kmem_cache_free(mm_cachep, (mm))) |
339 | ||
340 | #include <linux/init_task.h> | |
341 | ||
342 | static struct mm_struct * mm_init(struct mm_struct * mm) | |
343 | { | |
344 | atomic_set(&mm->mm_users, 1); | |
345 | atomic_set(&mm->mm_count, 1); | |
346 | init_rwsem(&mm->mmap_sem); | |
347 | INIT_LIST_HEAD(&mm->mmlist); | |
3cb4a0bb KH |
348 | mm->flags = (current->mm) ? current->mm->flags |
349 | : MMF_DUMP_FILTER_DEFAULT; | |
1da177e4 LT |
350 | mm->core_waiters = 0; |
351 | mm->nr_ptes = 0; | |
4294621f | 352 | set_mm_counter(mm, file_rss, 0); |
404351e6 | 353 | set_mm_counter(mm, anon_rss, 0); |
1da177e4 LT |
354 | spin_lock_init(&mm->page_table_lock); |
355 | rwlock_init(&mm->ioctx_list_lock); | |
356 | mm->ioctx_list = NULL; | |
1da177e4 | 357 | mm->free_area_cache = TASK_UNMAPPED_BASE; |
1363c3cd | 358 | mm->cached_hole_size = ~0UL; |
1da177e4 LT |
359 | |
360 | if (likely(!mm_alloc_pgd(mm))) { | |
361 | mm->def_flags = 0; | |
362 | return mm; | |
363 | } | |
364 | free_mm(mm); | |
365 | return NULL; | |
366 | } | |
367 | ||
368 | /* | |
369 | * Allocate and initialize an mm_struct. | |
370 | */ | |
371 | struct mm_struct * mm_alloc(void) | |
372 | { | |
373 | struct mm_struct * mm; | |
374 | ||
375 | mm = allocate_mm(); | |
376 | if (mm) { | |
377 | memset(mm, 0, sizeof(*mm)); | |
378 | mm = mm_init(mm); | |
379 | } | |
380 | return mm; | |
381 | } | |
382 | ||
383 | /* | |
384 | * Called when the last reference to the mm | |
385 | * is dropped: either by a lazy thread or by | |
386 | * mmput. Free the page directory and the mm. | |
387 | */ | |
388 | void fastcall __mmdrop(struct mm_struct *mm) | |
389 | { | |
390 | BUG_ON(mm == &init_mm); | |
391 | mm_free_pgd(mm); | |
392 | destroy_context(mm); | |
393 | free_mm(mm); | |
394 | } | |
395 | ||
396 | /* | |
397 | * Decrement the use count and release all resources for an mm. | |
398 | */ | |
399 | void mmput(struct mm_struct *mm) | |
400 | { | |
0ae26f1b AM |
401 | might_sleep(); |
402 | ||
1da177e4 LT |
403 | if (atomic_dec_and_test(&mm->mm_users)) { |
404 | exit_aio(mm); | |
405 | exit_mmap(mm); | |
406 | if (!list_empty(&mm->mmlist)) { | |
407 | spin_lock(&mmlist_lock); | |
408 | list_del(&mm->mmlist); | |
409 | spin_unlock(&mmlist_lock); | |
410 | } | |
411 | put_swap_token(mm); | |
412 | mmdrop(mm); | |
413 | } | |
414 | } | |
415 | EXPORT_SYMBOL_GPL(mmput); | |
416 | ||
417 | /** | |
418 | * get_task_mm - acquire a reference to the task's mm | |
419 | * | |
420 | * Returns %NULL if the task has no mm. Checks PF_BORROWED_MM (meaning | |
421 | * this kernel workthread has transiently adopted a user mm with use_mm, | |
422 | * to do its AIO) is not set and if so returns a reference to it, after | |
423 | * bumping up the use count. User must release the mm via mmput() | |
424 | * after use. Typically used by /proc and ptrace. | |
425 | */ | |
426 | struct mm_struct *get_task_mm(struct task_struct *task) | |
427 | { | |
428 | struct mm_struct *mm; | |
429 | ||
430 | task_lock(task); | |
431 | mm = task->mm; | |
432 | if (mm) { | |
433 | if (task->flags & PF_BORROWED_MM) | |
434 | mm = NULL; | |
435 | else | |
436 | atomic_inc(&mm->mm_users); | |
437 | } | |
438 | task_unlock(task); | |
439 | return mm; | |
440 | } | |
441 | EXPORT_SYMBOL_GPL(get_task_mm); | |
442 | ||
443 | /* Please note the differences between mmput and mm_release. | |
444 | * mmput is called whenever we stop holding onto a mm_struct, | |
445 | * error success whatever. | |
446 | * | |
447 | * mm_release is called after a mm_struct has been removed | |
448 | * from the current process. | |
449 | * | |
450 | * This difference is important for error handling, when we | |
451 | * only half set up a mm_struct for a new process and need to restore | |
452 | * the old one. Because we mmput the new mm_struct before | |
453 | * restoring the old one. . . | |
454 | * Eric Biederman 10 January 1998 | |
455 | */ | |
456 | void mm_release(struct task_struct *tsk, struct mm_struct *mm) | |
457 | { | |
458 | struct completion *vfork_done = tsk->vfork_done; | |
459 | ||
460 | /* Get rid of any cached register state */ | |
461 | deactivate_mm(tsk, mm); | |
462 | ||
463 | /* notify parent sleeping on vfork() */ | |
464 | if (vfork_done) { | |
465 | tsk->vfork_done = NULL; | |
466 | complete(vfork_done); | |
467 | } | |
fec1d011 RM |
468 | |
469 | /* | |
470 | * If we're exiting normally, clear a user-space tid field if | |
471 | * requested. We leave this alone when dying by signal, to leave | |
472 | * the value intact in a core dump, and to save the unnecessary | |
473 | * trouble otherwise. Userland only wants this done for a sys_exit. | |
474 | */ | |
475 | if (tsk->clear_child_tid | |
476 | && !(tsk->flags & PF_SIGNALED) | |
477 | && atomic_read(&mm->mm_users) > 1) { | |
1da177e4 LT |
478 | u32 __user * tidptr = tsk->clear_child_tid; |
479 | tsk->clear_child_tid = NULL; | |
480 | ||
481 | /* | |
482 | * We don't check the error code - if userspace has | |
483 | * not set up a proper pointer then tough luck. | |
484 | */ | |
485 | put_user(0, tidptr); | |
486 | sys_futex(tidptr, FUTEX_WAKE, 1, NULL, NULL, 0); | |
487 | } | |
488 | } | |
489 | ||
a0a7ec30 JD |
490 | /* |
491 | * Allocate a new mm structure and copy contents from the | |
492 | * mm structure of the passed in task structure. | |
493 | */ | |
494 | static struct mm_struct *dup_mm(struct task_struct *tsk) | |
495 | { | |
496 | struct mm_struct *mm, *oldmm = current->mm; | |
497 | int err; | |
498 | ||
499 | if (!oldmm) | |
500 | return NULL; | |
501 | ||
502 | mm = allocate_mm(); | |
503 | if (!mm) | |
504 | goto fail_nomem; | |
505 | ||
506 | memcpy(mm, oldmm, sizeof(*mm)); | |
507 | ||
7602bdf2 AC |
508 | /* Initializing for Swap token stuff */ |
509 | mm->token_priority = 0; | |
510 | mm->last_interval = 0; | |
511 | ||
a0a7ec30 JD |
512 | if (!mm_init(mm)) |
513 | goto fail_nomem; | |
514 | ||
515 | if (init_new_context(tsk, mm)) | |
516 | goto fail_nocontext; | |
517 | ||
518 | err = dup_mmap(mm, oldmm); | |
519 | if (err) | |
520 | goto free_pt; | |
521 | ||
522 | mm->hiwater_rss = get_mm_rss(mm); | |
523 | mm->hiwater_vm = mm->total_vm; | |
524 | ||
525 | return mm; | |
526 | ||
527 | free_pt: | |
528 | mmput(mm); | |
529 | ||
530 | fail_nomem: | |
531 | return NULL; | |
532 | ||
533 | fail_nocontext: | |
534 | /* | |
535 | * If init_new_context() failed, we cannot use mmput() to free the mm | |
536 | * because it calls destroy_context() | |
537 | */ | |
538 | mm_free_pgd(mm); | |
539 | free_mm(mm); | |
540 | return NULL; | |
541 | } | |
542 | ||
1da177e4 LT |
543 | static int copy_mm(unsigned long clone_flags, struct task_struct * tsk) |
544 | { | |
545 | struct mm_struct * mm, *oldmm; | |
546 | int retval; | |
547 | ||
548 | tsk->min_flt = tsk->maj_flt = 0; | |
549 | tsk->nvcsw = tsk->nivcsw = 0; | |
550 | ||
551 | tsk->mm = NULL; | |
552 | tsk->active_mm = NULL; | |
553 | ||
554 | /* | |
555 | * Are we cloning a kernel thread? | |
556 | * | |
557 | * We need to steal a active VM for that.. | |
558 | */ | |
559 | oldmm = current->mm; | |
560 | if (!oldmm) | |
561 | return 0; | |
562 | ||
563 | if (clone_flags & CLONE_VM) { | |
564 | atomic_inc(&oldmm->mm_users); | |
565 | mm = oldmm; | |
1da177e4 LT |
566 | goto good_mm; |
567 | } | |
568 | ||
569 | retval = -ENOMEM; | |
a0a7ec30 | 570 | mm = dup_mm(tsk); |
1da177e4 LT |
571 | if (!mm) |
572 | goto fail_nomem; | |
573 | ||
1da177e4 | 574 | good_mm: |
7602bdf2 AC |
575 | /* Initializing for Swap token stuff */ |
576 | mm->token_priority = 0; | |
577 | mm->last_interval = 0; | |
578 | ||
1da177e4 LT |
579 | tsk->mm = mm; |
580 | tsk->active_mm = mm; | |
581 | return 0; | |
582 | ||
1da177e4 LT |
583 | fail_nomem: |
584 | return retval; | |
1da177e4 LT |
585 | } |
586 | ||
587 | static inline struct fs_struct *__copy_fs_struct(struct fs_struct *old) | |
588 | { | |
589 | struct fs_struct *fs = kmem_cache_alloc(fs_cachep, GFP_KERNEL); | |
590 | /* We don't need to lock fs - think why ;-) */ | |
591 | if (fs) { | |
592 | atomic_set(&fs->count, 1); | |
593 | rwlock_init(&fs->lock); | |
594 | fs->umask = old->umask; | |
595 | read_lock(&old->lock); | |
596 | fs->rootmnt = mntget(old->rootmnt); | |
597 | fs->root = dget(old->root); | |
598 | fs->pwdmnt = mntget(old->pwdmnt); | |
599 | fs->pwd = dget(old->pwd); | |
600 | if (old->altroot) { | |
601 | fs->altrootmnt = mntget(old->altrootmnt); | |
602 | fs->altroot = dget(old->altroot); | |
603 | } else { | |
604 | fs->altrootmnt = NULL; | |
605 | fs->altroot = NULL; | |
606 | } | |
607 | read_unlock(&old->lock); | |
608 | } | |
609 | return fs; | |
610 | } | |
611 | ||
612 | struct fs_struct *copy_fs_struct(struct fs_struct *old) | |
613 | { | |
614 | return __copy_fs_struct(old); | |
615 | } | |
616 | ||
617 | EXPORT_SYMBOL_GPL(copy_fs_struct); | |
618 | ||
619 | static inline int copy_fs(unsigned long clone_flags, struct task_struct * tsk) | |
620 | { | |
621 | if (clone_flags & CLONE_FS) { | |
622 | atomic_inc(¤t->fs->count); | |
623 | return 0; | |
624 | } | |
625 | tsk->fs = __copy_fs_struct(current->fs); | |
626 | if (!tsk->fs) | |
627 | return -ENOMEM; | |
628 | return 0; | |
629 | } | |
630 | ||
ab2af1f5 | 631 | static int count_open_files(struct fdtable *fdt) |
1da177e4 | 632 | { |
bbea9f69 | 633 | int size = fdt->max_fds; |
1da177e4 LT |
634 | int i; |
635 | ||
636 | /* Find the last open fd */ | |
637 | for (i = size/(8*sizeof(long)); i > 0; ) { | |
badf1662 | 638 | if (fdt->open_fds->fds_bits[--i]) |
1da177e4 LT |
639 | break; |
640 | } | |
641 | i = (i+1) * 8 * sizeof(long); | |
642 | return i; | |
643 | } | |
644 | ||
badf1662 DS |
645 | static struct files_struct *alloc_files(void) |
646 | { | |
647 | struct files_struct *newf; | |
648 | struct fdtable *fdt; | |
649 | ||
e94b1766 | 650 | newf = kmem_cache_alloc(files_cachep, GFP_KERNEL); |
badf1662 DS |
651 | if (!newf) |
652 | goto out; | |
653 | ||
654 | atomic_set(&newf->count, 1); | |
655 | ||
656 | spin_lock_init(&newf->file_lock); | |
0c9e63fd | 657 | newf->next_fd = 0; |
ab2af1f5 | 658 | fdt = &newf->fdtab; |
badf1662 | 659 | fdt->max_fds = NR_OPEN_DEFAULT; |
0c9e63fd ED |
660 | fdt->close_on_exec = (fd_set *)&newf->close_on_exec_init; |
661 | fdt->open_fds = (fd_set *)&newf->open_fds_init; | |
badf1662 | 662 | fdt->fd = &newf->fd_array[0]; |
ab2af1f5 | 663 | INIT_RCU_HEAD(&fdt->rcu); |
ab2af1f5 DS |
664 | fdt->next = NULL; |
665 | rcu_assign_pointer(newf->fdt, fdt); | |
badf1662 DS |
666 | out: |
667 | return newf; | |
668 | } | |
669 | ||
a016f338 JD |
670 | /* |
671 | * Allocate a new files structure and copy contents from the | |
672 | * passed in files structure. | |
6e667260 | 673 | * errorp will be valid only when the returned files_struct is NULL. |
a016f338 JD |
674 | */ |
675 | static struct files_struct *dup_fd(struct files_struct *oldf, int *errorp) | |
1da177e4 | 676 | { |
a016f338 | 677 | struct files_struct *newf; |
1da177e4 | 678 | struct file **old_fds, **new_fds; |
bbea9f69 | 679 | int open_files, size, i; |
badf1662 | 680 | struct fdtable *old_fdt, *new_fdt; |
1da177e4 | 681 | |
6e667260 | 682 | *errorp = -ENOMEM; |
badf1662 DS |
683 | newf = alloc_files(); |
684 | if (!newf) | |
1da177e4 LT |
685 | goto out; |
686 | ||
1da177e4 | 687 | spin_lock(&oldf->file_lock); |
badf1662 DS |
688 | old_fdt = files_fdtable(oldf); |
689 | new_fdt = files_fdtable(newf); | |
ab2af1f5 | 690 | open_files = count_open_files(old_fdt); |
1da177e4 LT |
691 | |
692 | /* | |
bbea9f69 VL |
693 | * Check whether we need to allocate a larger fd array and fd set. |
694 | * Note: we're not a clone task, so the open count won't change. | |
1da177e4 | 695 | */ |
badf1662 DS |
696 | if (open_files > new_fdt->max_fds) { |
697 | new_fdt->max_fds = 0; | |
1da177e4 LT |
698 | spin_unlock(&oldf->file_lock); |
699 | spin_lock(&newf->file_lock); | |
a016f338 | 700 | *errorp = expand_files(newf, open_files-1); |
1da177e4 | 701 | spin_unlock(&newf->file_lock); |
a016f338 | 702 | if (*errorp < 0) |
1da177e4 | 703 | goto out_release; |
ab2af1f5 DS |
704 | new_fdt = files_fdtable(newf); |
705 | /* | |
706 | * Reacquire the oldf lock and a pointer to its fd table | |
707 | * who knows it may have a new bigger fd table. We need | |
708 | * the latest pointer. | |
709 | */ | |
1da177e4 | 710 | spin_lock(&oldf->file_lock); |
ab2af1f5 | 711 | old_fdt = files_fdtable(oldf); |
1da177e4 LT |
712 | } |
713 | ||
badf1662 DS |
714 | old_fds = old_fdt->fd; |
715 | new_fds = new_fdt->fd; | |
1da177e4 | 716 | |
f3d19c90 VL |
717 | memcpy(new_fdt->open_fds->fds_bits, |
718 | old_fdt->open_fds->fds_bits, open_files/8); | |
719 | memcpy(new_fdt->close_on_exec->fds_bits, | |
720 | old_fdt->close_on_exec->fds_bits, open_files/8); | |
1da177e4 LT |
721 | |
722 | for (i = open_files; i != 0; i--) { | |
723 | struct file *f = *old_fds++; | |
724 | if (f) { | |
725 | get_file(f); | |
726 | } else { | |
727 | /* | |
728 | * The fd may be claimed in the fd bitmap but not yet | |
729 | * instantiated in the files array if a sibling thread | |
730 | * is partway through open(). So make sure that this | |
731 | * fd is available to the new process. | |
732 | */ | |
badf1662 | 733 | FD_CLR(open_files - i, new_fdt->open_fds); |
1da177e4 | 734 | } |
ab2af1f5 | 735 | rcu_assign_pointer(*new_fds++, f); |
1da177e4 LT |
736 | } |
737 | spin_unlock(&oldf->file_lock); | |
738 | ||
739 | /* compute the remainder to be cleared */ | |
badf1662 | 740 | size = (new_fdt->max_fds - open_files) * sizeof(struct file *); |
1da177e4 | 741 | |
23ff4440 DW |
742 | /* This is long word aligned thus could use a optimized version */ |
743 | memset(new_fds, 0, size); | |
1da177e4 | 744 | |
bbea9f69 VL |
745 | if (new_fdt->max_fds > open_files) { |
746 | int left = (new_fdt->max_fds-open_files)/8; | |
1da177e4 LT |
747 | int start = open_files / (8 * sizeof(unsigned long)); |
748 | ||
badf1662 DS |
749 | memset(&new_fdt->open_fds->fds_bits[start], 0, left); |
750 | memset(&new_fdt->close_on_exec->fds_bits[start], 0, left); | |
1da177e4 LT |
751 | } |
752 | ||
a016f338 | 753 | return newf; |
1da177e4 LT |
754 | |
755 | out_release: | |
1da177e4 | 756 | kmem_cache_free(files_cachep, newf); |
f3d19c90 | 757 | out: |
42862298 | 758 | return NULL; |
1da177e4 LT |
759 | } |
760 | ||
a016f338 JD |
761 | static int copy_files(unsigned long clone_flags, struct task_struct * tsk) |
762 | { | |
763 | struct files_struct *oldf, *newf; | |
764 | int error = 0; | |
765 | ||
766 | /* | |
767 | * A background process may not have any files ... | |
768 | */ | |
769 | oldf = current->files; | |
770 | if (!oldf) | |
771 | goto out; | |
772 | ||
773 | if (clone_flags & CLONE_FILES) { | |
774 | atomic_inc(&oldf->count); | |
775 | goto out; | |
776 | } | |
777 | ||
778 | /* | |
779 | * Note: we may be using current for both targets (See exec.c) | |
780 | * This works because we cache current->files (old) as oldf. Don't | |
781 | * break this. | |
782 | */ | |
783 | tsk->files = NULL; | |
a016f338 JD |
784 | newf = dup_fd(oldf, &error); |
785 | if (!newf) | |
786 | goto out; | |
787 | ||
788 | tsk->files = newf; | |
789 | error = 0; | |
790 | out: | |
791 | return error; | |
792 | } | |
793 | ||
1da177e4 LT |
794 | /* |
795 | * Helper to unshare the files of the current task. | |
796 | * We don't want to expose copy_files internals to | |
797 | * the exec layer of the kernel. | |
798 | */ | |
799 | ||
800 | int unshare_files(void) | |
801 | { | |
802 | struct files_struct *files = current->files; | |
803 | int rc; | |
804 | ||
910dea7f | 805 | BUG_ON(!files); |
1da177e4 LT |
806 | |
807 | /* This can race but the race causes us to copy when we don't | |
808 | need to and drop the copy */ | |
809 | if(atomic_read(&files->count) == 1) | |
810 | { | |
811 | atomic_inc(&files->count); | |
812 | return 0; | |
813 | } | |
814 | rc = copy_files(0, current); | |
815 | if(rc) | |
816 | current->files = files; | |
817 | return rc; | |
818 | } | |
819 | ||
820 | EXPORT_SYMBOL(unshare_files); | |
821 | ||
822 | static inline int copy_sighand(unsigned long clone_flags, struct task_struct * tsk) | |
823 | { | |
824 | struct sighand_struct *sig; | |
825 | ||
826 | if (clone_flags & (CLONE_SIGHAND | CLONE_THREAD)) { | |
827 | atomic_inc(¤t->sighand->count); | |
828 | return 0; | |
829 | } | |
830 | sig = kmem_cache_alloc(sighand_cachep, GFP_KERNEL); | |
e56d0903 | 831 | rcu_assign_pointer(tsk->sighand, sig); |
1da177e4 LT |
832 | if (!sig) |
833 | return -ENOMEM; | |
1da177e4 LT |
834 | atomic_set(&sig->count, 1); |
835 | memcpy(sig->action, current->sighand->action, sizeof(sig->action)); | |
836 | return 0; | |
837 | } | |
838 | ||
a7e5328a | 839 | void __cleanup_sighand(struct sighand_struct *sighand) |
c81addc9 | 840 | { |
c81addc9 ON |
841 | if (atomic_dec_and_test(&sighand->count)) |
842 | kmem_cache_free(sighand_cachep, sighand); | |
843 | } | |
844 | ||
1da177e4 LT |
845 | static inline int copy_signal(unsigned long clone_flags, struct task_struct * tsk) |
846 | { | |
847 | struct signal_struct *sig; | |
848 | int ret; | |
849 | ||
850 | if (clone_flags & CLONE_THREAD) { | |
851 | atomic_inc(¤t->signal->count); | |
852 | atomic_inc(¤t->signal->live); | |
853 | return 0; | |
854 | } | |
855 | sig = kmem_cache_alloc(signal_cachep, GFP_KERNEL); | |
856 | tsk->signal = sig; | |
857 | if (!sig) | |
858 | return -ENOMEM; | |
859 | ||
860 | ret = copy_thread_group_keys(tsk); | |
861 | if (ret < 0) { | |
862 | kmem_cache_free(signal_cachep, sig); | |
863 | return ret; | |
864 | } | |
865 | ||
866 | atomic_set(&sig->count, 1); | |
867 | atomic_set(&sig->live, 1); | |
868 | init_waitqueue_head(&sig->wait_chldexit); | |
869 | sig->flags = 0; | |
870 | sig->group_exit_code = 0; | |
871 | sig->group_exit_task = NULL; | |
872 | sig->group_stop_count = 0; | |
873 | sig->curr_target = NULL; | |
874 | init_sigpending(&sig->shared_pending); | |
875 | INIT_LIST_HEAD(&sig->posix_timers); | |
876 | ||
c9cb2e3d | 877 | hrtimer_init(&sig->real_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL); |
2ff678b8 | 878 | sig->it_real_incr.tv64 = 0; |
1da177e4 | 879 | sig->real_timer.function = it_real_fn; |
05cfb614 | 880 | sig->tsk = tsk; |
1da177e4 LT |
881 | |
882 | sig->it_virt_expires = cputime_zero; | |
883 | sig->it_virt_incr = cputime_zero; | |
884 | sig->it_prof_expires = cputime_zero; | |
885 | sig->it_prof_incr = cputime_zero; | |
886 | ||
1da177e4 | 887 | sig->leader = 0; /* session leadership doesn't inherit */ |
ab521dc0 | 888 | sig->tty_old_pgrp = NULL; |
1da177e4 LT |
889 | |
890 | sig->utime = sig->stime = sig->cutime = sig->cstime = cputime_zero; | |
9ac52315 LV |
891 | sig->gtime = cputime_zero; |
892 | sig->cgtime = cputime_zero; | |
1da177e4 LT |
893 | sig->nvcsw = sig->nivcsw = sig->cnvcsw = sig->cnivcsw = 0; |
894 | sig->min_flt = sig->maj_flt = sig->cmin_flt = sig->cmaj_flt = 0; | |
6eaeeaba | 895 | sig->inblock = sig->oublock = sig->cinblock = sig->coublock = 0; |
172ba844 | 896 | sig->sum_sched_runtime = 0; |
1da177e4 LT |
897 | INIT_LIST_HEAD(&sig->cpu_timers[0]); |
898 | INIT_LIST_HEAD(&sig->cpu_timers[1]); | |
899 | INIT_LIST_HEAD(&sig->cpu_timers[2]); | |
ad4ecbcb | 900 | taskstats_tgid_init(sig); |
1da177e4 LT |
901 | |
902 | task_lock(current->group_leader); | |
903 | memcpy(sig->rlim, current->signal->rlim, sizeof sig->rlim); | |
904 | task_unlock(current->group_leader); | |
905 | ||
906 | if (sig->rlim[RLIMIT_CPU].rlim_cur != RLIM_INFINITY) { | |
907 | /* | |
908 | * New sole thread in the process gets an expiry time | |
909 | * of the whole CPU time limit. | |
910 | */ | |
911 | tsk->it_prof_expires = | |
912 | secs_to_cputime(sig->rlim[RLIMIT_CPU].rlim_cur); | |
913 | } | |
0e464814 | 914 | acct_init_pacct(&sig->pacct); |
1da177e4 | 915 | |
522ed776 MT |
916 | tty_audit_fork(sig); |
917 | ||
1da177e4 LT |
918 | return 0; |
919 | } | |
920 | ||
6b3934ef ON |
921 | void __cleanup_signal(struct signal_struct *sig) |
922 | { | |
923 | exit_thread_group_keys(sig); | |
924 | kmem_cache_free(signal_cachep, sig); | |
925 | } | |
926 | ||
927 | static inline void cleanup_signal(struct task_struct *tsk) | |
928 | { | |
929 | struct signal_struct *sig = tsk->signal; | |
930 | ||
931 | atomic_dec(&sig->live); | |
932 | ||
933 | if (atomic_dec_and_test(&sig->count)) | |
934 | __cleanup_signal(sig); | |
935 | } | |
936 | ||
1da177e4 LT |
937 | static inline void copy_flags(unsigned long clone_flags, struct task_struct *p) |
938 | { | |
939 | unsigned long new_flags = p->flags; | |
940 | ||
83144186 | 941 | new_flags &= ~PF_SUPERPRIV; |
1da177e4 LT |
942 | new_flags |= PF_FORKNOEXEC; |
943 | if (!(clone_flags & CLONE_PTRACE)) | |
944 | p->ptrace = 0; | |
945 | p->flags = new_flags; | |
2e131895 | 946 | clear_freeze_flag(p); |
1da177e4 LT |
947 | } |
948 | ||
949 | asmlinkage long sys_set_tid_address(int __user *tidptr) | |
950 | { | |
951 | current->clear_child_tid = tidptr; | |
952 | ||
953 | return current->pid; | |
954 | } | |
955 | ||
23f78d4a IM |
956 | static inline void rt_mutex_init_task(struct task_struct *p) |
957 | { | |
23f78d4a | 958 | spin_lock_init(&p->pi_lock); |
e29e175b | 959 | #ifdef CONFIG_RT_MUTEXES |
23f78d4a IM |
960 | plist_head_init(&p->pi_waiters, &p->pi_lock); |
961 | p->pi_blocked_on = NULL; | |
23f78d4a IM |
962 | #endif |
963 | } | |
964 | ||
1da177e4 LT |
965 | /* |
966 | * This creates a new process as a copy of the old one, | |
967 | * but does not actually start it yet. | |
968 | * | |
969 | * It copies the registers, and all the appropriate | |
970 | * parts of the process environment (as per the clone | |
971 | * flags). The actual kick-off is left to the caller. | |
972 | */ | |
36c8b586 IM |
973 | static struct task_struct *copy_process(unsigned long clone_flags, |
974 | unsigned long stack_start, | |
975 | struct pt_regs *regs, | |
976 | unsigned long stack_size, | |
36c8b586 | 977 | int __user *child_tidptr, |
85868995 | 978 | struct pid *pid) |
1da177e4 LT |
979 | { |
980 | int retval; | |
981 | struct task_struct *p = NULL; | |
b4f48b63 | 982 | int cgroup_callbacks_done = 0; |
1da177e4 LT |
983 | |
984 | if ((clone_flags & (CLONE_NEWNS|CLONE_FS)) == (CLONE_NEWNS|CLONE_FS)) | |
985 | return ERR_PTR(-EINVAL); | |
986 | ||
987 | /* | |
988 | * Thread groups must share signals as well, and detached threads | |
989 | * can only be started up within the thread group. | |
990 | */ | |
991 | if ((clone_flags & CLONE_THREAD) && !(clone_flags & CLONE_SIGHAND)) | |
992 | return ERR_PTR(-EINVAL); | |
993 | ||
994 | /* | |
995 | * Shared signal handlers imply shared VM. By way of the above, | |
996 | * thread groups also imply shared VM. Blocking this case allows | |
997 | * for various simplifications in other code. | |
998 | */ | |
999 | if ((clone_flags & CLONE_SIGHAND) && !(clone_flags & CLONE_VM)) | |
1000 | return ERR_PTR(-EINVAL); | |
1001 | ||
1002 | retval = security_task_create(clone_flags); | |
1003 | if (retval) | |
1004 | goto fork_out; | |
1005 | ||
1006 | retval = -ENOMEM; | |
1007 | p = dup_task_struct(current); | |
1008 | if (!p) | |
1009 | goto fork_out; | |
1010 | ||
bea493a0 PZ |
1011 | rt_mutex_init_task(p); |
1012 | ||
de30a2b3 IM |
1013 | #ifdef CONFIG_TRACE_IRQFLAGS |
1014 | DEBUG_LOCKS_WARN_ON(!p->hardirqs_enabled); | |
1015 | DEBUG_LOCKS_WARN_ON(!p->softirqs_enabled); | |
1016 | #endif | |
1da177e4 LT |
1017 | retval = -EAGAIN; |
1018 | if (atomic_read(&p->user->processes) >= | |
1019 | p->signal->rlim[RLIMIT_NPROC].rlim_cur) { | |
1020 | if (!capable(CAP_SYS_ADMIN) && !capable(CAP_SYS_RESOURCE) && | |
acce292c | 1021 | p->user != current->nsproxy->user_ns->root_user) |
1da177e4 LT |
1022 | goto bad_fork_free; |
1023 | } | |
1024 | ||
1025 | atomic_inc(&p->user->__count); | |
1026 | atomic_inc(&p->user->processes); | |
1027 | get_group_info(p->group_info); | |
1028 | ||
1029 | /* | |
1030 | * If multiple threads are within copy_process(), then this check | |
1031 | * triggers too late. This doesn't hurt, the check is only there | |
1032 | * to stop root fork bombs. | |
1033 | */ | |
1034 | if (nr_threads >= max_threads) | |
1035 | goto bad_fork_cleanup_count; | |
1036 | ||
a1261f54 | 1037 | if (!try_module_get(task_thread_info(p)->exec_domain->module)) |
1da177e4 LT |
1038 | goto bad_fork_cleanup_count; |
1039 | ||
1040 | if (p->binfmt && !try_module_get(p->binfmt->module)) | |
1041 | goto bad_fork_cleanup_put_domain; | |
1042 | ||
1043 | p->did_exec = 0; | |
ca74e92b | 1044 | delayacct_tsk_init(p); /* Must remain after dup_task_struct() */ |
1da177e4 | 1045 | copy_flags(clone_flags, p); |
1da177e4 LT |
1046 | INIT_LIST_HEAD(&p->children); |
1047 | INIT_LIST_HEAD(&p->sibling); | |
1048 | p->vfork_done = NULL; | |
1049 | spin_lock_init(&p->alloc_lock); | |
1da177e4 LT |
1050 | |
1051 | clear_tsk_thread_flag(p, TIF_SIGPENDING); | |
1052 | init_sigpending(&p->pending); | |
1053 | ||
1054 | p->utime = cputime_zero; | |
1055 | p->stime = cputime_zero; | |
9ac52315 | 1056 | p->gtime = cputime_zero; |
c66f08be MN |
1057 | p->utimescaled = cputime_zero; |
1058 | p->stimescaled = cputime_zero; | |
172ba844 | 1059 | |
4b98d11b | 1060 | #ifdef CONFIG_TASK_XACCT |
1da177e4 LT |
1061 | p->rchar = 0; /* I/O counter: bytes read */ |
1062 | p->wchar = 0; /* I/O counter: bytes written */ | |
1063 | p->syscr = 0; /* I/O counter: read syscalls */ | |
1064 | p->syscw = 0; /* I/O counter: write syscalls */ | |
4b98d11b | 1065 | #endif |
7c3ab738 | 1066 | task_io_accounting_init(p); |
1da177e4 LT |
1067 | acct_clear_integrals(p); |
1068 | ||
23ff4440 | 1069 | p->it_virt_expires = cputime_zero; |
1da177e4 | 1070 | p->it_prof_expires = cputime_zero; |
23ff4440 DW |
1071 | p->it_sched_expires = 0; |
1072 | INIT_LIST_HEAD(&p->cpu_timers[0]); | |
1073 | INIT_LIST_HEAD(&p->cpu_timers[1]); | |
1074 | INIT_LIST_HEAD(&p->cpu_timers[2]); | |
1da177e4 LT |
1075 | |
1076 | p->lock_depth = -1; /* -1 = no lock */ | |
1077 | do_posix_clock_monotonic_gettime(&p->start_time); | |
924b42d5 TJ |
1078 | p->real_start_time = p->start_time; |
1079 | monotonic_to_bootbased(&p->real_start_time); | |
57c521ce | 1080 | #ifdef CONFIG_SECURITY |
1da177e4 | 1081 | p->security = NULL; |
57c521ce | 1082 | #endif |
1da177e4 | 1083 | p->io_context = NULL; |
1da177e4 | 1084 | p->audit_context = NULL; |
b4f48b63 | 1085 | cgroup_fork(p); |
1da177e4 LT |
1086 | #ifdef CONFIG_NUMA |
1087 | p->mempolicy = mpol_copy(p->mempolicy); | |
1088 | if (IS_ERR(p->mempolicy)) { | |
1089 | retval = PTR_ERR(p->mempolicy); | |
1090 | p->mempolicy = NULL; | |
b4f48b63 | 1091 | goto bad_fork_cleanup_cgroup; |
1da177e4 | 1092 | } |
c61afb18 | 1093 | mpol_fix_fork_child_flag(p); |
1da177e4 | 1094 | #endif |
de30a2b3 IM |
1095 | #ifdef CONFIG_TRACE_IRQFLAGS |
1096 | p->irq_events = 0; | |
b36e4758 RK |
1097 | #ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW |
1098 | p->hardirqs_enabled = 1; | |
1099 | #else | |
de30a2b3 | 1100 | p->hardirqs_enabled = 0; |
b36e4758 | 1101 | #endif |
de30a2b3 IM |
1102 | p->hardirq_enable_ip = 0; |
1103 | p->hardirq_enable_event = 0; | |
1104 | p->hardirq_disable_ip = _THIS_IP_; | |
1105 | p->hardirq_disable_event = 0; | |
1106 | p->softirqs_enabled = 1; | |
1107 | p->softirq_enable_ip = _THIS_IP_; | |
1108 | p->softirq_enable_event = 0; | |
1109 | p->softirq_disable_ip = 0; | |
1110 | p->softirq_disable_event = 0; | |
1111 | p->hardirq_context = 0; | |
1112 | p->softirq_context = 0; | |
1113 | #endif | |
fbb9ce95 IM |
1114 | #ifdef CONFIG_LOCKDEP |
1115 | p->lockdep_depth = 0; /* no locks held yet */ | |
1116 | p->curr_chain_key = 0; | |
1117 | p->lockdep_recursion = 0; | |
1118 | #endif | |
1da177e4 | 1119 | |
408894ee IM |
1120 | #ifdef CONFIG_DEBUG_MUTEXES |
1121 | p->blocked_on = NULL; /* not blocked yet */ | |
1122 | #endif | |
1123 | ||
1da177e4 LT |
1124 | if ((retval = security_task_alloc(p))) |
1125 | goto bad_fork_cleanup_policy; | |
1126 | if ((retval = audit_alloc(p))) | |
1127 | goto bad_fork_cleanup_security; | |
1128 | /* copy all the process information */ | |
1129 | if ((retval = copy_semundo(clone_flags, p))) | |
1130 | goto bad_fork_cleanup_audit; | |
1131 | if ((retval = copy_files(clone_flags, p))) | |
1132 | goto bad_fork_cleanup_semundo; | |
1133 | if ((retval = copy_fs(clone_flags, p))) | |
1134 | goto bad_fork_cleanup_files; | |
1135 | if ((retval = copy_sighand(clone_flags, p))) | |
1136 | goto bad_fork_cleanup_fs; | |
1137 | if ((retval = copy_signal(clone_flags, p))) | |
1138 | goto bad_fork_cleanup_sighand; | |
1139 | if ((retval = copy_mm(clone_flags, p))) | |
1140 | goto bad_fork_cleanup_signal; | |
1141 | if ((retval = copy_keys(clone_flags, p))) | |
1142 | goto bad_fork_cleanup_mm; | |
ab516013 | 1143 | if ((retval = copy_namespaces(clone_flags, p))) |
1da177e4 LT |
1144 | goto bad_fork_cleanup_keys; |
1145 | retval = copy_thread(0, clone_flags, stack_start, stack_size, p, regs); | |
1146 | if (retval) | |
1651e14e | 1147 | goto bad_fork_cleanup_namespaces; |
1da177e4 | 1148 | |
425fb2b4 PE |
1149 | if (pid != &init_struct_pid) { |
1150 | retval = -ENOMEM; | |
1151 | pid = alloc_pid(task_active_pid_ns(p)); | |
1152 | if (!pid) | |
1153 | goto bad_fork_cleanup_namespaces; | |
6f4e6433 PE |
1154 | |
1155 | if (clone_flags & CLONE_NEWPID) { | |
1156 | retval = pid_ns_prepare_proc(task_active_pid_ns(p)); | |
1157 | if (retval < 0) | |
1158 | goto bad_fork_free_pid; | |
1159 | } | |
425fb2b4 PE |
1160 | } |
1161 | ||
1162 | p->pid = pid_nr(pid); | |
1163 | p->tgid = p->pid; | |
1164 | if (clone_flags & CLONE_THREAD) | |
1165 | p->tgid = current->tgid; | |
1166 | ||
1da177e4 LT |
1167 | p->set_child_tid = (clone_flags & CLONE_CHILD_SETTID) ? child_tidptr : NULL; |
1168 | /* | |
1169 | * Clear TID on mm_release()? | |
1170 | */ | |
1171 | p->clear_child_tid = (clone_flags & CLONE_CHILD_CLEARTID) ? child_tidptr: NULL; | |
42b2dd0a | 1172 | #ifdef CONFIG_FUTEX |
8f17d3a5 IM |
1173 | p->robust_list = NULL; |
1174 | #ifdef CONFIG_COMPAT | |
1175 | p->compat_robust_list = NULL; | |
1176 | #endif | |
c87e2837 IM |
1177 | INIT_LIST_HEAD(&p->pi_state_list); |
1178 | p->pi_state_cache = NULL; | |
42b2dd0a | 1179 | #endif |
f9a3879a GM |
1180 | /* |
1181 | * sigaltstack should be cleared when sharing the same VM | |
1182 | */ | |
1183 | if ((clone_flags & (CLONE_VM|CLONE_VFORK)) == CLONE_VM) | |
1184 | p->sas_ss_sp = p->sas_ss_size = 0; | |
1185 | ||
1da177e4 LT |
1186 | /* |
1187 | * Syscall tracing should be turned off in the child regardless | |
1188 | * of CLONE_PTRACE. | |
1189 | */ | |
1190 | clear_tsk_thread_flag(p, TIF_SYSCALL_TRACE); | |
ed75e8d5 LV |
1191 | #ifdef TIF_SYSCALL_EMU |
1192 | clear_tsk_thread_flag(p, TIF_SYSCALL_EMU); | |
1193 | #endif | |
1da177e4 LT |
1194 | |
1195 | /* Our parent execution domain becomes current domain | |
1196 | These must match for thread signalling to apply */ | |
1da177e4 LT |
1197 | p->parent_exec_id = p->self_exec_id; |
1198 | ||
1199 | /* ok, now we should be set up.. */ | |
1200 | p->exit_signal = (clone_flags & CLONE_THREAD) ? -1 : (clone_flags & CSIGNAL); | |
1201 | p->pdeath_signal = 0; | |
1202 | p->exit_state = 0; | |
1203 | ||
1da177e4 LT |
1204 | /* |
1205 | * Ok, make it visible to the rest of the system. | |
1206 | * We dont wake it up yet. | |
1207 | */ | |
1208 | p->group_leader = p; | |
47e65328 | 1209 | INIT_LIST_HEAD(&p->thread_group); |
1da177e4 LT |
1210 | INIT_LIST_HEAD(&p->ptrace_children); |
1211 | INIT_LIST_HEAD(&p->ptrace_list); | |
1212 | ||
476d139c NP |
1213 | /* Perform scheduler related setup. Assign this task to a CPU. */ |
1214 | sched_fork(p, clone_flags); | |
1215 | ||
b4f48b63 PM |
1216 | /* Now that the task is set up, run cgroup callbacks if |
1217 | * necessary. We need to run them before the task is visible | |
1218 | * on the tasklist. */ | |
1219 | cgroup_fork_callbacks(p); | |
1220 | cgroup_callbacks_done = 1; | |
1221 | ||
1da177e4 LT |
1222 | /* Need tasklist lock for parent etc handling! */ |
1223 | write_lock_irq(&tasklist_lock); | |
1224 | ||
5b160f5e ON |
1225 | /* for sys_ioprio_set(IOPRIO_WHO_PGRP) */ |
1226 | p->ioprio = current->ioprio; | |
1227 | ||
1da177e4 | 1228 | /* |
476d139c NP |
1229 | * The task hasn't been attached yet, so its cpus_allowed mask will |
1230 | * not be changed, nor will its assigned CPU. | |
1231 | * | |
1232 | * The cpus_allowed mask of the parent may have changed after it was | |
1233 | * copied first time - so re-copy it here, then check the child's CPU | |
1234 | * to ensure it is on a valid CPU (and if not, just force it back to | |
1235 | * parent's CPU). This avoids alot of nasty races. | |
1da177e4 LT |
1236 | */ |
1237 | p->cpus_allowed = current->cpus_allowed; | |
26ff6ad9 SV |
1238 | if (unlikely(!cpu_isset(task_cpu(p), p->cpus_allowed) || |
1239 | !cpu_online(task_cpu(p)))) | |
476d139c | 1240 | set_task_cpu(p, smp_processor_id()); |
1da177e4 | 1241 | |
1da177e4 LT |
1242 | /* CLONE_PARENT re-uses the old parent */ |
1243 | if (clone_flags & (CLONE_PARENT|CLONE_THREAD)) | |
1244 | p->real_parent = current->real_parent; | |
1245 | else | |
1246 | p->real_parent = current; | |
1247 | p->parent = p->real_parent; | |
1248 | ||
3f17da69 | 1249 | spin_lock(¤t->sighand->siglock); |
4a2c7a78 ON |
1250 | |
1251 | /* | |
1252 | * Process group and session signals need to be delivered to just the | |
1253 | * parent before the fork or both the parent and the child after the | |
1254 | * fork. Restart if a signal comes in before we add the new process to | |
1255 | * it's process group. | |
1256 | * A fatal signal pending means that current will exit, so the new | |
1257 | * thread can't slip out of an OOM kill (or normal SIGKILL). | |
1258 | */ | |
23ff4440 | 1259 | recalc_sigpending(); |
4a2c7a78 ON |
1260 | if (signal_pending(current)) { |
1261 | spin_unlock(¤t->sighand->siglock); | |
1262 | write_unlock_irq(&tasklist_lock); | |
1263 | retval = -ERESTARTNOINTR; | |
425fb2b4 | 1264 | goto bad_fork_free_pid; |
4a2c7a78 ON |
1265 | } |
1266 | ||
1da177e4 | 1267 | if (clone_flags & CLONE_THREAD) { |
1da177e4 | 1268 | p->group_leader = current->group_leader; |
47e65328 | 1269 | list_add_tail_rcu(&p->thread_group, &p->group_leader->thread_group); |
1da177e4 | 1270 | |
1da177e4 LT |
1271 | if (!cputime_eq(current->signal->it_virt_expires, |
1272 | cputime_zero) || | |
1273 | !cputime_eq(current->signal->it_prof_expires, | |
1274 | cputime_zero) || | |
1275 | current->signal->rlim[RLIMIT_CPU].rlim_cur != RLIM_INFINITY || | |
1276 | !list_empty(¤t->signal->cpu_timers[0]) || | |
1277 | !list_empty(¤t->signal->cpu_timers[1]) || | |
1278 | !list_empty(¤t->signal->cpu_timers[2])) { | |
1279 | /* | |
1280 | * Have child wake up on its first tick to check | |
1281 | * for process CPU timers. | |
1282 | */ | |
1283 | p->it_prof_expires = jiffies_to_cputime(1); | |
1284 | } | |
1da177e4 LT |
1285 | } |
1286 | ||
73b9ebfe ON |
1287 | if (likely(p->pid)) { |
1288 | add_parent(p); | |
1289 | if (unlikely(p->ptrace & PT_PTRACED)) | |
1290 | __ptrace_link(p, current->parent); | |
1291 | ||
1292 | if (thread_group_leader(p)) { | |
30e49c26 PE |
1293 | if (clone_flags & CLONE_NEWPID) { |
1294 | p->nsproxy->pid_ns->child_reaper = p; | |
1295 | p->signal->tty = NULL; | |
1296 | p->signal->pgrp = p->pid; | |
1297 | set_task_session(p, p->pid); | |
1298 | attach_pid(p, PIDTYPE_PGID, pid); | |
1299 | attach_pid(p, PIDTYPE_SID, pid); | |
1300 | } else { | |
1301 | p->signal->tty = current->signal->tty; | |
1302 | p->signal->pgrp = task_pgrp_nr(current); | |
1303 | set_task_session(p, task_session_nr(current)); | |
1304 | attach_pid(p, PIDTYPE_PGID, | |
1305 | task_pgrp(current)); | |
1306 | attach_pid(p, PIDTYPE_SID, | |
1307 | task_session(current)); | |
1308 | } | |
73b9ebfe | 1309 | |
5e85d4ab | 1310 | list_add_tail_rcu(&p->tasks, &init_task.tasks); |
1da177e4 | 1311 | __get_cpu_var(process_counts)++; |
73b9ebfe | 1312 | } |
85868995 | 1313 | attach_pid(p, PIDTYPE_PID, pid); |
73b9ebfe | 1314 | nr_threads++; |
1da177e4 LT |
1315 | } |
1316 | ||
1da177e4 | 1317 | total_forks++; |
3f17da69 | 1318 | spin_unlock(¤t->sighand->siglock); |
1da177e4 | 1319 | write_unlock_irq(&tasklist_lock); |
c13cf856 | 1320 | proc_fork_connector(p); |
817929ec | 1321 | cgroup_post_fork(p); |
1da177e4 LT |
1322 | return p; |
1323 | ||
425fb2b4 PE |
1324 | bad_fork_free_pid: |
1325 | if (pid != &init_struct_pid) | |
1326 | free_pid(pid); | |
ab516013 | 1327 | bad_fork_cleanup_namespaces: |
444f378b | 1328 | exit_task_namespaces(p); |
1da177e4 LT |
1329 | bad_fork_cleanup_keys: |
1330 | exit_keys(p); | |
1331 | bad_fork_cleanup_mm: | |
1332 | if (p->mm) | |
1333 | mmput(p->mm); | |
1334 | bad_fork_cleanup_signal: | |
6b3934ef | 1335 | cleanup_signal(p); |
1da177e4 | 1336 | bad_fork_cleanup_sighand: |
a7e5328a | 1337 | __cleanup_sighand(p->sighand); |
1da177e4 LT |
1338 | bad_fork_cleanup_fs: |
1339 | exit_fs(p); /* blocking */ | |
1340 | bad_fork_cleanup_files: | |
1341 | exit_files(p); /* blocking */ | |
1342 | bad_fork_cleanup_semundo: | |
1343 | exit_sem(p); | |
1344 | bad_fork_cleanup_audit: | |
1345 | audit_free(p); | |
1346 | bad_fork_cleanup_security: | |
1347 | security_task_free(p); | |
1348 | bad_fork_cleanup_policy: | |
1349 | #ifdef CONFIG_NUMA | |
1350 | mpol_free(p->mempolicy); | |
b4f48b63 | 1351 | bad_fork_cleanup_cgroup: |
1da177e4 | 1352 | #endif |
b4f48b63 | 1353 | cgroup_exit(p, cgroup_callbacks_done); |
35df17c5 | 1354 | delayacct_tsk_free(p); |
1da177e4 LT |
1355 | if (p->binfmt) |
1356 | module_put(p->binfmt->module); | |
1357 | bad_fork_cleanup_put_domain: | |
a1261f54 | 1358 | module_put(task_thread_info(p)->exec_domain->module); |
1da177e4 LT |
1359 | bad_fork_cleanup_count: |
1360 | put_group_info(p->group_info); | |
1361 | atomic_dec(&p->user->processes); | |
1362 | free_uid(p->user); | |
1363 | bad_fork_free: | |
1364 | free_task(p); | |
fe7d37d1 ON |
1365 | fork_out: |
1366 | return ERR_PTR(retval); | |
1da177e4 LT |
1367 | } |
1368 | ||
f95d47ca | 1369 | noinline struct pt_regs * __devinit __attribute__((weak)) idle_regs(struct pt_regs *regs) |
1da177e4 LT |
1370 | { |
1371 | memset(regs, 0, sizeof(struct pt_regs)); | |
1372 | return regs; | |
1373 | } | |
1374 | ||
9abcf40b | 1375 | struct task_struct * __cpuinit fork_idle(int cpu) |
1da177e4 | 1376 | { |
36c8b586 | 1377 | struct task_struct *task; |
1da177e4 LT |
1378 | struct pt_regs regs; |
1379 | ||
30e49c26 | 1380 | task = copy_process(CLONE_VM, 0, idle_regs(®s), 0, NULL, |
85868995 | 1381 | &init_struct_pid); |
753ca4f3 AM |
1382 | if (!IS_ERR(task)) |
1383 | init_idle(task, cpu); | |
73b9ebfe | 1384 | |
1da177e4 LT |
1385 | return task; |
1386 | } | |
1387 | ||
1388 | static inline int fork_traceflag (unsigned clone_flags) | |
1389 | { | |
1390 | if (clone_flags & CLONE_UNTRACED) | |
1391 | return 0; | |
1392 | else if (clone_flags & CLONE_VFORK) { | |
1393 | if (current->ptrace & PT_TRACE_VFORK) | |
1394 | return PTRACE_EVENT_VFORK; | |
1395 | } else if ((clone_flags & CSIGNAL) != SIGCHLD) { | |
1396 | if (current->ptrace & PT_TRACE_CLONE) | |
1397 | return PTRACE_EVENT_CLONE; | |
1398 | } else if (current->ptrace & PT_TRACE_FORK) | |
1399 | return PTRACE_EVENT_FORK; | |
1400 | ||
1401 | return 0; | |
1402 | } | |
1403 | ||
1404 | /* | |
1405 | * Ok, this is the main fork-routine. | |
1406 | * | |
1407 | * It copies the process, and if successful kick-starts | |
1408 | * it and waits for it to finish using the VM if required. | |
1409 | */ | |
1410 | long do_fork(unsigned long clone_flags, | |
1411 | unsigned long stack_start, | |
1412 | struct pt_regs *regs, | |
1413 | unsigned long stack_size, | |
1414 | int __user *parent_tidptr, | |
1415 | int __user *child_tidptr) | |
1416 | { | |
1417 | struct task_struct *p; | |
1418 | int trace = 0; | |
92476d7f | 1419 | long nr; |
1da177e4 | 1420 | |
1da177e4 LT |
1421 | if (unlikely(current->ptrace)) { |
1422 | trace = fork_traceflag (clone_flags); | |
1423 | if (trace) | |
1424 | clone_flags |= CLONE_PTRACE; | |
1425 | } | |
1426 | ||
a6f5e063 | 1427 | p = copy_process(clone_flags, stack_start, regs, stack_size, |
30e49c26 | 1428 | child_tidptr, NULL); |
1da177e4 LT |
1429 | /* |
1430 | * Do this prior waking up the new thread - the thread pointer | |
1431 | * might get invalid after that point, if the thread exits quickly. | |
1432 | */ | |
1433 | if (!IS_ERR(p)) { | |
1434 | struct completion vfork; | |
1435 | ||
30e49c26 PE |
1436 | /* |
1437 | * this is enough to call pid_nr_ns here, but this if | |
1438 | * improves optimisation of regular fork() | |
1439 | */ | |
1440 | nr = (clone_flags & CLONE_NEWPID) ? | |
1441 | task_pid_nr_ns(p, current->nsproxy->pid_ns) : | |
1442 | task_pid_vnr(p); | |
1443 | ||
1444 | if (clone_flags & CLONE_PARENT_SETTID) | |
1445 | put_user(nr, parent_tidptr); | |
a6f5e063 | 1446 | |
1da177e4 LT |
1447 | if (clone_flags & CLONE_VFORK) { |
1448 | p->vfork_done = &vfork; | |
1449 | init_completion(&vfork); | |
1450 | } | |
1451 | ||
1452 | if ((p->ptrace & PT_PTRACED) || (clone_flags & CLONE_STOPPED)) { | |
1453 | /* | |
1454 | * We'll start up with an immediate SIGSTOP. | |
1455 | */ | |
1456 | sigaddset(&p->pending.signal, SIGSTOP); | |
1457 | set_tsk_thread_flag(p, TIF_SIGPENDING); | |
1458 | } | |
1459 | ||
1460 | if (!(clone_flags & CLONE_STOPPED)) | |
1461 | wake_up_new_task(p, clone_flags); | |
1462 | else | |
1463 | p->state = TASK_STOPPED; | |
1464 | ||
1465 | if (unlikely (trace)) { | |
92476d7f | 1466 | current->ptrace_message = nr; |
1da177e4 LT |
1467 | ptrace_notify ((trace << 8) | SIGTRAP); |
1468 | } | |
1469 | ||
1470 | if (clone_flags & CLONE_VFORK) { | |
ba96a0c8 | 1471 | freezer_do_not_count(); |
1da177e4 | 1472 | wait_for_completion(&vfork); |
ba96a0c8 | 1473 | freezer_count(); |
9f59ce5d CE |
1474 | if (unlikely (current->ptrace & PT_TRACE_VFORK_DONE)) { |
1475 | current->ptrace_message = nr; | |
1da177e4 | 1476 | ptrace_notify ((PTRACE_EVENT_VFORK_DONE << 8) | SIGTRAP); |
9f59ce5d | 1477 | } |
1da177e4 LT |
1478 | } |
1479 | } else { | |
92476d7f | 1480 | nr = PTR_ERR(p); |
1da177e4 | 1481 | } |
92476d7f | 1482 | return nr; |
1da177e4 LT |
1483 | } |
1484 | ||
5fd63b30 RT |
1485 | #ifndef ARCH_MIN_MMSTRUCT_ALIGN |
1486 | #define ARCH_MIN_MMSTRUCT_ALIGN 0 | |
1487 | #endif | |
1488 | ||
4ba9b9d0 | 1489 | static void sighand_ctor(struct kmem_cache *cachep, void *data) |
aa1757f9 ON |
1490 | { |
1491 | struct sighand_struct *sighand = data; | |
1492 | ||
a35afb83 | 1493 | spin_lock_init(&sighand->siglock); |
b8fceee1 | 1494 | init_waitqueue_head(&sighand->signalfd_wqh); |
aa1757f9 ON |
1495 | } |
1496 | ||
1da177e4 LT |
1497 | void __init proc_caches_init(void) |
1498 | { | |
1499 | sighand_cachep = kmem_cache_create("sighand_cache", | |
1500 | sizeof(struct sighand_struct), 0, | |
aa1757f9 | 1501 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_DESTROY_BY_RCU, |
20c2df83 | 1502 | sighand_ctor); |
1da177e4 LT |
1503 | signal_cachep = kmem_cache_create("signal_cache", |
1504 | sizeof(struct signal_struct), 0, | |
20c2df83 PM |
1505 | SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL); |
1506 | files_cachep = kmem_cache_create("files_cache", | |
1da177e4 | 1507 | sizeof(struct files_struct), 0, |
20c2df83 PM |
1508 | SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL); |
1509 | fs_cachep = kmem_cache_create("fs_cache", | |
1da177e4 | 1510 | sizeof(struct fs_struct), 0, |
20c2df83 | 1511 | SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL); |
1da177e4 LT |
1512 | vm_area_cachep = kmem_cache_create("vm_area_struct", |
1513 | sizeof(struct vm_area_struct), 0, | |
20c2df83 | 1514 | SLAB_PANIC, NULL); |
1da177e4 | 1515 | mm_cachep = kmem_cache_create("mm_struct", |
5fd63b30 | 1516 | sizeof(struct mm_struct), ARCH_MIN_MMSTRUCT_ALIGN, |
20c2df83 | 1517 | SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL); |
1da177e4 | 1518 | } |
cf2e340f | 1519 | |
cf2e340f JD |
1520 | /* |
1521 | * Check constraints on flags passed to the unshare system call and | |
1522 | * force unsharing of additional process context as appropriate. | |
1523 | */ | |
1524 | static inline void check_unshare_flags(unsigned long *flags_ptr) | |
1525 | { | |
1526 | /* | |
1527 | * If unsharing a thread from a thread group, must also | |
1528 | * unshare vm. | |
1529 | */ | |
1530 | if (*flags_ptr & CLONE_THREAD) | |
1531 | *flags_ptr |= CLONE_VM; | |
1532 | ||
1533 | /* | |
1534 | * If unsharing vm, must also unshare signal handlers. | |
1535 | */ | |
1536 | if (*flags_ptr & CLONE_VM) | |
1537 | *flags_ptr |= CLONE_SIGHAND; | |
1538 | ||
1539 | /* | |
1540 | * If unsharing signal handlers and the task was created | |
1541 | * using CLONE_THREAD, then must unshare the thread | |
1542 | */ | |
1543 | if ((*flags_ptr & CLONE_SIGHAND) && | |
1544 | (atomic_read(¤t->signal->count) > 1)) | |
1545 | *flags_ptr |= CLONE_THREAD; | |
1546 | ||
1547 | /* | |
1548 | * If unsharing namespace, must also unshare filesystem information. | |
1549 | */ | |
1550 | if (*flags_ptr & CLONE_NEWNS) | |
1551 | *flags_ptr |= CLONE_FS; | |
1552 | } | |
1553 | ||
1554 | /* | |
1555 | * Unsharing of tasks created with CLONE_THREAD is not supported yet | |
1556 | */ | |
1557 | static int unshare_thread(unsigned long unshare_flags) | |
1558 | { | |
1559 | if (unshare_flags & CLONE_THREAD) | |
1560 | return -EINVAL; | |
1561 | ||
1562 | return 0; | |
1563 | } | |
1564 | ||
1565 | /* | |
99d1419d | 1566 | * Unshare the filesystem structure if it is being shared |
cf2e340f JD |
1567 | */ |
1568 | static int unshare_fs(unsigned long unshare_flags, struct fs_struct **new_fsp) | |
1569 | { | |
1570 | struct fs_struct *fs = current->fs; | |
1571 | ||
1572 | if ((unshare_flags & CLONE_FS) && | |
99d1419d JD |
1573 | (fs && atomic_read(&fs->count) > 1)) { |
1574 | *new_fsp = __copy_fs_struct(current->fs); | |
1575 | if (!*new_fsp) | |
1576 | return -ENOMEM; | |
1577 | } | |
cf2e340f JD |
1578 | |
1579 | return 0; | |
1580 | } | |
1581 | ||
cf2e340f | 1582 | /* |
dae3c5a0 | 1583 | * Unsharing of sighand is not supported yet |
cf2e340f JD |
1584 | */ |
1585 | static int unshare_sighand(unsigned long unshare_flags, struct sighand_struct **new_sighp) | |
1586 | { | |
1587 | struct sighand_struct *sigh = current->sighand; | |
1588 | ||
dae3c5a0 | 1589 | if ((unshare_flags & CLONE_SIGHAND) && atomic_read(&sigh->count) > 1) |
cf2e340f JD |
1590 | return -EINVAL; |
1591 | else | |
1592 | return 0; | |
1593 | } | |
1594 | ||
1595 | /* | |
a0a7ec30 | 1596 | * Unshare vm if it is being shared |
cf2e340f JD |
1597 | */ |
1598 | static int unshare_vm(unsigned long unshare_flags, struct mm_struct **new_mmp) | |
1599 | { | |
1600 | struct mm_struct *mm = current->mm; | |
1601 | ||
1602 | if ((unshare_flags & CLONE_VM) && | |
a0a7ec30 | 1603 | (mm && atomic_read(&mm->mm_users) > 1)) { |
2d61b867 | 1604 | return -EINVAL; |
a0a7ec30 | 1605 | } |
cf2e340f JD |
1606 | |
1607 | return 0; | |
cf2e340f JD |
1608 | } |
1609 | ||
1610 | /* | |
a016f338 | 1611 | * Unshare file descriptor table if it is being shared |
cf2e340f JD |
1612 | */ |
1613 | static int unshare_fd(unsigned long unshare_flags, struct files_struct **new_fdp) | |
1614 | { | |
1615 | struct files_struct *fd = current->files; | |
a016f338 | 1616 | int error = 0; |
cf2e340f JD |
1617 | |
1618 | if ((unshare_flags & CLONE_FILES) && | |
a016f338 JD |
1619 | (fd && atomic_read(&fd->count) > 1)) { |
1620 | *new_fdp = dup_fd(fd, &error); | |
1621 | if (!*new_fdp) | |
1622 | return error; | |
1623 | } | |
cf2e340f JD |
1624 | |
1625 | return 0; | |
1626 | } | |
1627 | ||
1628 | /* | |
1629 | * Unsharing of semundo for tasks created with CLONE_SYSVSEM is not | |
1630 | * supported yet | |
1631 | */ | |
1632 | static int unshare_semundo(unsigned long unshare_flags, struct sem_undo_list **new_ulistp) | |
1633 | { | |
1634 | if (unshare_flags & CLONE_SYSVSEM) | |
1635 | return -EINVAL; | |
1636 | ||
1637 | return 0; | |
1638 | } | |
1639 | ||
1640 | /* | |
1641 | * unshare allows a process to 'unshare' part of the process | |
1642 | * context which was originally shared using clone. copy_* | |
1643 | * functions used by do_fork() cannot be used here directly | |
1644 | * because they modify an inactive task_struct that is being | |
1645 | * constructed. Here we are modifying the current, active, | |
1646 | * task_struct. | |
1647 | */ | |
1648 | asmlinkage long sys_unshare(unsigned long unshare_flags) | |
1649 | { | |
1650 | int err = 0; | |
1651 | struct fs_struct *fs, *new_fs = NULL; | |
dae3c5a0 | 1652 | struct sighand_struct *new_sigh = NULL; |
cf2e340f JD |
1653 | struct mm_struct *mm, *new_mm = NULL, *active_mm = NULL; |
1654 | struct files_struct *fd, *new_fd = NULL; | |
1655 | struct sem_undo_list *new_ulist = NULL; | |
cf7b708c | 1656 | struct nsproxy *new_nsproxy = NULL; |
cf2e340f JD |
1657 | |
1658 | check_unshare_flags(&unshare_flags); | |
1659 | ||
06f9d4f9 EB |
1660 | /* Return -EINVAL for all unsupported flags */ |
1661 | err = -EINVAL; | |
1662 | if (unshare_flags & ~(CLONE_THREAD|CLONE_FS|CLONE_NEWNS|CLONE_SIGHAND| | |
25b21cb2 | 1663 | CLONE_VM|CLONE_FILES|CLONE_SYSVSEM| |
9dd776b6 EB |
1664 | CLONE_NEWUTS|CLONE_NEWIPC|CLONE_NEWUSER| |
1665 | CLONE_NEWNET)) | |
06f9d4f9 EB |
1666 | goto bad_unshare_out; |
1667 | ||
cf2e340f JD |
1668 | if ((err = unshare_thread(unshare_flags))) |
1669 | goto bad_unshare_out; | |
1670 | if ((err = unshare_fs(unshare_flags, &new_fs))) | |
1671 | goto bad_unshare_cleanup_thread; | |
cf2e340f | 1672 | if ((err = unshare_sighand(unshare_flags, &new_sigh))) |
e3222c4e | 1673 | goto bad_unshare_cleanup_fs; |
cf2e340f JD |
1674 | if ((err = unshare_vm(unshare_flags, &new_mm))) |
1675 | goto bad_unshare_cleanup_sigh; | |
1676 | if ((err = unshare_fd(unshare_flags, &new_fd))) | |
1677 | goto bad_unshare_cleanup_vm; | |
1678 | if ((err = unshare_semundo(unshare_flags, &new_ulist))) | |
1679 | goto bad_unshare_cleanup_fd; | |
e3222c4e BP |
1680 | if ((err = unshare_nsproxy_namespaces(unshare_flags, &new_nsproxy, |
1681 | new_fs))) | |
071df104 | 1682 | goto bad_unshare_cleanup_semundo; |
c0b2fc31 | 1683 | |
e3222c4e | 1684 | if (new_fs || new_mm || new_fd || new_ulist || new_nsproxy) { |
ab516013 | 1685 | |
c0b2fc31 | 1686 | if (new_nsproxy) { |
cf7b708c PE |
1687 | switch_task_namespaces(current, new_nsproxy); |
1688 | new_nsproxy = NULL; | |
c0b2fc31 | 1689 | } |
cf2e340f | 1690 | |
cf7b708c PE |
1691 | task_lock(current); |
1692 | ||
cf2e340f JD |
1693 | if (new_fs) { |
1694 | fs = current->fs; | |
1695 | current->fs = new_fs; | |
1696 | new_fs = fs; | |
1697 | } | |
1698 | ||
cf2e340f JD |
1699 | if (new_mm) { |
1700 | mm = current->mm; | |
1701 | active_mm = current->active_mm; | |
1702 | current->mm = new_mm; | |
1703 | current->active_mm = new_mm; | |
1704 | activate_mm(active_mm, new_mm); | |
1705 | new_mm = mm; | |
1706 | } | |
1707 | ||
1708 | if (new_fd) { | |
1709 | fd = current->files; | |
1710 | current->files = new_fd; | |
1711 | new_fd = fd; | |
1712 | } | |
1713 | ||
1714 | task_unlock(current); | |
1715 | } | |
1716 | ||
c0b2fc31 | 1717 | if (new_nsproxy) |
444f378b | 1718 | put_nsproxy(new_nsproxy); |
c0b2fc31 | 1719 | |
ab516013 | 1720 | bad_unshare_cleanup_semundo: |
cf2e340f JD |
1721 | bad_unshare_cleanup_fd: |
1722 | if (new_fd) | |
1723 | put_files_struct(new_fd); | |
1724 | ||
1725 | bad_unshare_cleanup_vm: | |
1726 | if (new_mm) | |
1727 | mmput(new_mm); | |
1728 | ||
1729 | bad_unshare_cleanup_sigh: | |
1730 | if (new_sigh) | |
1731 | if (atomic_dec_and_test(&new_sigh->count)) | |
1732 | kmem_cache_free(sighand_cachep, new_sigh); | |
1733 | ||
cf2e340f JD |
1734 | bad_unshare_cleanup_fs: |
1735 | if (new_fs) | |
1736 | put_fs_struct(new_fs); | |
1737 | ||
1738 | bad_unshare_cleanup_thread: | |
1739 | bad_unshare_out: | |
1740 | return err; | |
1741 | } |