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> | |
9f3acc31 | 25 | #include <linux/fdtable.h> |
da9cbc87 | 26 | #include <linux/iocontext.h> |
1da177e4 LT |
27 | #include <linux/key.h> |
28 | #include <linux/binfmts.h> | |
29 | #include <linux/mman.h> | |
cddb8a5c | 30 | #include <linux/mmu_notifier.h> |
1da177e4 | 31 | #include <linux/fs.h> |
ab516013 | 32 | #include <linux/nsproxy.h> |
c59ede7b | 33 | #include <linux/capability.h> |
1da177e4 | 34 | #include <linux/cpu.h> |
b4f48b63 | 35 | #include <linux/cgroup.h> |
1da177e4 | 36 | #include <linux/security.h> |
a1e78772 | 37 | #include <linux/hugetlb.h> |
1da177e4 LT |
38 | #include <linux/swap.h> |
39 | #include <linux/syscalls.h> | |
40 | #include <linux/jiffies.h> | |
09a05394 | 41 | #include <linux/tracehook.h> |
1da177e4 | 42 | #include <linux/futex.h> |
7c3ab738 | 43 | #include <linux/task_io_accounting_ops.h> |
ab2af1f5 | 44 | #include <linux/rcupdate.h> |
1da177e4 LT |
45 | #include <linux/ptrace.h> |
46 | #include <linux/mount.h> | |
47 | #include <linux/audit.h> | |
78fb7466 | 48 | #include <linux/memcontrol.h> |
1da177e4 LT |
49 | #include <linux/profile.h> |
50 | #include <linux/rmap.h> | |
51 | #include <linux/acct.h> | |
8f0ab514 | 52 | #include <linux/tsacct_kern.h> |
9f46080c | 53 | #include <linux/cn_proc.h> |
ba96a0c8 | 54 | #include <linux/freezer.h> |
ca74e92b | 55 | #include <linux/delayacct.h> |
ad4ecbcb | 56 | #include <linux/taskstats_kern.h> |
0a425405 | 57 | #include <linux/random.h> |
522ed776 | 58 | #include <linux/tty.h> |
6f4e6433 | 59 | #include <linux/proc_fs.h> |
fd0928df | 60 | #include <linux/blkdev.h> |
0a16b607 | 61 | #include <trace/sched.h> |
1da177e4 LT |
62 | |
63 | #include <asm/pgtable.h> | |
64 | #include <asm/pgalloc.h> | |
65 | #include <asm/uaccess.h> | |
66 | #include <asm/mmu_context.h> | |
67 | #include <asm/cacheflush.h> | |
68 | #include <asm/tlbflush.h> | |
69 | ||
70 | /* | |
71 | * Protected counters by write_lock_irq(&tasklist_lock) | |
72 | */ | |
73 | unsigned long total_forks; /* Handle normal Linux uptimes. */ | |
74 | int nr_threads; /* The idle threads do not count.. */ | |
75 | ||
76 | int max_threads; /* tunable limit on nr_threads */ | |
77 | ||
78 | DEFINE_PER_CPU(unsigned long, process_counts) = 0; | |
79 | ||
c59923a1 | 80 | __cacheline_aligned DEFINE_RWLOCK(tasklist_lock); /* outer */ |
1da177e4 LT |
81 | |
82 | int nr_processes(void) | |
83 | { | |
84 | int cpu; | |
85 | int total = 0; | |
86 | ||
87 | for_each_online_cpu(cpu) | |
88 | total += per_cpu(process_counts, cpu); | |
89 | ||
90 | return total; | |
91 | } | |
92 | ||
93 | #ifndef __HAVE_ARCH_TASK_STRUCT_ALLOCATOR | |
94 | # define alloc_task_struct() kmem_cache_alloc(task_struct_cachep, GFP_KERNEL) | |
95 | # define free_task_struct(tsk) kmem_cache_free(task_struct_cachep, (tsk)) | |
e18b890b | 96 | static struct kmem_cache *task_struct_cachep; |
1da177e4 LT |
97 | #endif |
98 | ||
b69c49b7 FT |
99 | #ifndef __HAVE_ARCH_THREAD_INFO_ALLOCATOR |
100 | static inline struct thread_info *alloc_thread_info(struct task_struct *tsk) | |
101 | { | |
102 | #ifdef CONFIG_DEBUG_STACK_USAGE | |
103 | gfp_t mask = GFP_KERNEL | __GFP_ZERO; | |
104 | #else | |
105 | gfp_t mask = GFP_KERNEL; | |
106 | #endif | |
107 | return (struct thread_info *)__get_free_pages(mask, THREAD_SIZE_ORDER); | |
108 | } | |
109 | ||
110 | static inline void free_thread_info(struct thread_info *ti) | |
111 | { | |
112 | free_pages((unsigned long)ti, THREAD_SIZE_ORDER); | |
113 | } | |
114 | #endif | |
115 | ||
1da177e4 | 116 | /* SLAB cache for signal_struct structures (tsk->signal) */ |
e18b890b | 117 | static struct kmem_cache *signal_cachep; |
1da177e4 LT |
118 | |
119 | /* SLAB cache for sighand_struct structures (tsk->sighand) */ | |
e18b890b | 120 | struct kmem_cache *sighand_cachep; |
1da177e4 LT |
121 | |
122 | /* SLAB cache for files_struct structures (tsk->files) */ | |
e18b890b | 123 | struct kmem_cache *files_cachep; |
1da177e4 LT |
124 | |
125 | /* SLAB cache for fs_struct structures (tsk->fs) */ | |
e18b890b | 126 | struct kmem_cache *fs_cachep; |
1da177e4 LT |
127 | |
128 | /* SLAB cache for vm_area_struct structures */ | |
e18b890b | 129 | struct kmem_cache *vm_area_cachep; |
1da177e4 LT |
130 | |
131 | /* SLAB cache for mm_struct structures (tsk->mm) */ | |
e18b890b | 132 | static struct kmem_cache *mm_cachep; |
1da177e4 LT |
133 | |
134 | void free_task(struct task_struct *tsk) | |
135 | { | |
3e26c149 | 136 | prop_local_destroy_single(&tsk->dirties); |
f7e4217b | 137 | free_thread_info(tsk->stack); |
23f78d4a | 138 | rt_mutex_debug_task_free(tsk); |
1da177e4 LT |
139 | free_task_struct(tsk); |
140 | } | |
141 | EXPORT_SYMBOL(free_task); | |
142 | ||
158d9ebd | 143 | void __put_task_struct(struct task_struct *tsk) |
1da177e4 | 144 | { |
270f722d | 145 | WARN_ON(!tsk->exit_state); |
1da177e4 LT |
146 | WARN_ON(atomic_read(&tsk->usage)); |
147 | WARN_ON(tsk == current); | |
148 | ||
f1752eec | 149 | put_cred(tsk->cred); |
35df17c5 | 150 | delayacct_tsk_free(tsk); |
1da177e4 LT |
151 | |
152 | if (!profile_handoff_task(tsk)) | |
153 | free_task(tsk); | |
154 | } | |
155 | ||
2adee9b3 SS |
156 | /* |
157 | * macro override instead of weak attribute alias, to workaround | |
158 | * gcc 4.1.0 and 4.1.1 bugs with weak attribute and empty functions. | |
159 | */ | |
160 | #ifndef arch_task_cache_init | |
161 | #define arch_task_cache_init() | |
162 | #endif | |
61c4628b | 163 | |
1da177e4 LT |
164 | void __init fork_init(unsigned long mempages) |
165 | { | |
166 | #ifndef __HAVE_ARCH_TASK_STRUCT_ALLOCATOR | |
167 | #ifndef ARCH_MIN_TASKALIGN | |
168 | #define ARCH_MIN_TASKALIGN L1_CACHE_BYTES | |
169 | #endif | |
170 | /* create a slab on which task_structs can be allocated */ | |
171 | task_struct_cachep = | |
172 | kmem_cache_create("task_struct", sizeof(struct task_struct), | |
20c2df83 | 173 | ARCH_MIN_TASKALIGN, SLAB_PANIC, NULL); |
1da177e4 LT |
174 | #endif |
175 | ||
61c4628b SS |
176 | /* do the arch specific task caches init */ |
177 | arch_task_cache_init(); | |
178 | ||
1da177e4 LT |
179 | /* |
180 | * The default maximum number of threads is set to a safe | |
181 | * value: the thread structures can take up at most half | |
182 | * of memory. | |
183 | */ | |
184 | max_threads = mempages / (8 * THREAD_SIZE / PAGE_SIZE); | |
185 | ||
186 | /* | |
187 | * we need to allow at least 20 threads to boot a system | |
188 | */ | |
189 | if(max_threads < 20) | |
190 | max_threads = 20; | |
191 | ||
192 | init_task.signal->rlim[RLIMIT_NPROC].rlim_cur = max_threads/2; | |
193 | init_task.signal->rlim[RLIMIT_NPROC].rlim_max = max_threads/2; | |
194 | init_task.signal->rlim[RLIMIT_SIGPENDING] = | |
195 | init_task.signal->rlim[RLIMIT_NPROC]; | |
196 | } | |
197 | ||
61c4628b SS |
198 | int __attribute__((weak)) arch_dup_task_struct(struct task_struct *dst, |
199 | struct task_struct *src) | |
200 | { | |
201 | *dst = *src; | |
202 | return 0; | |
203 | } | |
204 | ||
1da177e4 LT |
205 | static struct task_struct *dup_task_struct(struct task_struct *orig) |
206 | { | |
207 | struct task_struct *tsk; | |
208 | struct thread_info *ti; | |
3e26c149 | 209 | int err; |
1da177e4 LT |
210 | |
211 | prepare_to_copy(orig); | |
212 | ||
213 | tsk = alloc_task_struct(); | |
214 | if (!tsk) | |
215 | return NULL; | |
216 | ||
217 | ti = alloc_thread_info(tsk); | |
218 | if (!ti) { | |
219 | free_task_struct(tsk); | |
220 | return NULL; | |
221 | } | |
222 | ||
61c4628b SS |
223 | err = arch_dup_task_struct(tsk, orig); |
224 | if (err) | |
225 | goto out; | |
226 | ||
f7e4217b | 227 | tsk->stack = ti; |
3e26c149 PZ |
228 | |
229 | err = prop_local_init_single(&tsk->dirties); | |
61c4628b SS |
230 | if (err) |
231 | goto out; | |
3e26c149 | 232 | |
10ebffde | 233 | setup_thread_stack(tsk, orig); |
1da177e4 | 234 | |
0a425405 AV |
235 | #ifdef CONFIG_CC_STACKPROTECTOR |
236 | tsk->stack_canary = get_random_int(); | |
237 | #endif | |
238 | ||
1da177e4 LT |
239 | /* One for us, one for whoever does the "release_task()" (usually parent) */ |
240 | atomic_set(&tsk->usage,2); | |
4b5d37ac | 241 | atomic_set(&tsk->fs_excl, 0); |
6c5c9341 | 242 | #ifdef CONFIG_BLK_DEV_IO_TRACE |
2056a782 | 243 | tsk->btrace_seq = 0; |
6c5c9341 | 244 | #endif |
a0aa7f68 | 245 | tsk->splice_pipe = NULL; |
1da177e4 | 246 | return tsk; |
61c4628b SS |
247 | |
248 | out: | |
249 | free_thread_info(ti); | |
250 | free_task_struct(tsk); | |
251 | return NULL; | |
1da177e4 LT |
252 | } |
253 | ||
254 | #ifdef CONFIG_MMU | |
a39bc516 | 255 | static int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm) |
1da177e4 | 256 | { |
fd3e42fc | 257 | struct vm_area_struct *mpnt, *tmp, **pprev; |
1da177e4 LT |
258 | struct rb_node **rb_link, *rb_parent; |
259 | int retval; | |
260 | unsigned long charge; | |
261 | struct mempolicy *pol; | |
262 | ||
263 | down_write(&oldmm->mmap_sem); | |
ec8c0446 | 264 | flush_cache_dup_mm(oldmm); |
ad339451 IM |
265 | /* |
266 | * Not linked in yet - no deadlock potential: | |
267 | */ | |
268 | down_write_nested(&mm->mmap_sem, SINGLE_DEPTH_NESTING); | |
7ee78232 | 269 | |
1da177e4 LT |
270 | mm->locked_vm = 0; |
271 | mm->mmap = NULL; | |
272 | mm->mmap_cache = NULL; | |
273 | mm->free_area_cache = oldmm->mmap_base; | |
1363c3cd | 274 | mm->cached_hole_size = ~0UL; |
1da177e4 | 275 | mm->map_count = 0; |
1da177e4 LT |
276 | cpus_clear(mm->cpu_vm_mask); |
277 | mm->mm_rb = RB_ROOT; | |
278 | rb_link = &mm->mm_rb.rb_node; | |
279 | rb_parent = NULL; | |
280 | pprev = &mm->mmap; | |
281 | ||
fd3e42fc | 282 | for (mpnt = oldmm->mmap; mpnt; mpnt = mpnt->vm_next) { |
1da177e4 LT |
283 | struct file *file; |
284 | ||
285 | if (mpnt->vm_flags & VM_DONTCOPY) { | |
3b6bfcdb HD |
286 | long pages = vma_pages(mpnt); |
287 | mm->total_vm -= pages; | |
ab50b8ed | 288 | vm_stat_account(mm, mpnt->vm_flags, mpnt->vm_file, |
3b6bfcdb | 289 | -pages); |
1da177e4 LT |
290 | continue; |
291 | } | |
292 | charge = 0; | |
293 | if (mpnt->vm_flags & VM_ACCOUNT) { | |
294 | unsigned int len = (mpnt->vm_end - mpnt->vm_start) >> PAGE_SHIFT; | |
295 | if (security_vm_enough_memory(len)) | |
296 | goto fail_nomem; | |
297 | charge = len; | |
298 | } | |
e94b1766 | 299 | tmp = kmem_cache_alloc(vm_area_cachep, GFP_KERNEL); |
1da177e4 LT |
300 | if (!tmp) |
301 | goto fail_nomem; | |
302 | *tmp = *mpnt; | |
846a16bf | 303 | pol = mpol_dup(vma_policy(mpnt)); |
1da177e4 LT |
304 | retval = PTR_ERR(pol); |
305 | if (IS_ERR(pol)) | |
306 | goto fail_nomem_policy; | |
307 | vma_set_policy(tmp, pol); | |
308 | tmp->vm_flags &= ~VM_LOCKED; | |
309 | tmp->vm_mm = mm; | |
310 | tmp->vm_next = NULL; | |
311 | anon_vma_link(tmp); | |
312 | file = tmp->vm_file; | |
313 | if (file) { | |
f3a43f3f | 314 | struct inode *inode = file->f_path.dentry->d_inode; |
1da177e4 LT |
315 | get_file(file); |
316 | if (tmp->vm_flags & VM_DENYWRITE) | |
317 | atomic_dec(&inode->i_writecount); | |
23ff4440 | 318 | |
1da177e4 LT |
319 | /* insert tmp into the share list, just after mpnt */ |
320 | spin_lock(&file->f_mapping->i_mmap_lock); | |
321 | tmp->vm_truncate_count = mpnt->vm_truncate_count; | |
322 | flush_dcache_mmap_lock(file->f_mapping); | |
323 | vma_prio_tree_add(tmp, mpnt); | |
324 | flush_dcache_mmap_unlock(file->f_mapping); | |
325 | spin_unlock(&file->f_mapping->i_mmap_lock); | |
326 | } | |
327 | ||
a1e78772 MG |
328 | /* |
329 | * Clear hugetlb-related page reserves for children. This only | |
330 | * affects MAP_PRIVATE mappings. Faults generated by the child | |
331 | * are not guaranteed to succeed, even if read-only | |
332 | */ | |
333 | if (is_vm_hugetlb_page(tmp)) | |
334 | reset_vma_resv_huge_pages(tmp); | |
335 | ||
1da177e4 | 336 | /* |
7ee78232 | 337 | * Link in the new vma and copy the page table entries. |
1da177e4 | 338 | */ |
1da177e4 LT |
339 | *pprev = tmp; |
340 | pprev = &tmp->vm_next; | |
341 | ||
342 | __vma_link_rb(mm, tmp, rb_link, rb_parent); | |
343 | rb_link = &tmp->vm_rb.rb_right; | |
344 | rb_parent = &tmp->vm_rb; | |
345 | ||
346 | mm->map_count++; | |
0b0db14c | 347 | retval = copy_page_range(mm, oldmm, mpnt); |
1da177e4 LT |
348 | |
349 | if (tmp->vm_ops && tmp->vm_ops->open) | |
350 | tmp->vm_ops->open(tmp); | |
351 | ||
352 | if (retval) | |
353 | goto out; | |
354 | } | |
d6dd61c8 JF |
355 | /* a new mm has just been created */ |
356 | arch_dup_mmap(oldmm, mm); | |
1da177e4 | 357 | retval = 0; |
1da177e4 | 358 | out: |
7ee78232 | 359 | up_write(&mm->mmap_sem); |
fd3e42fc | 360 | flush_tlb_mm(oldmm); |
1da177e4 LT |
361 | up_write(&oldmm->mmap_sem); |
362 | return retval; | |
363 | fail_nomem_policy: | |
364 | kmem_cache_free(vm_area_cachep, tmp); | |
365 | fail_nomem: | |
366 | retval = -ENOMEM; | |
367 | vm_unacct_memory(charge); | |
368 | goto out; | |
369 | } | |
370 | ||
371 | static inline int mm_alloc_pgd(struct mm_struct * mm) | |
372 | { | |
373 | mm->pgd = pgd_alloc(mm); | |
374 | if (unlikely(!mm->pgd)) | |
375 | return -ENOMEM; | |
376 | return 0; | |
377 | } | |
378 | ||
379 | static inline void mm_free_pgd(struct mm_struct * mm) | |
380 | { | |
5e541973 | 381 | pgd_free(mm, mm->pgd); |
1da177e4 LT |
382 | } |
383 | #else | |
384 | #define dup_mmap(mm, oldmm) (0) | |
385 | #define mm_alloc_pgd(mm) (0) | |
386 | #define mm_free_pgd(mm) | |
387 | #endif /* CONFIG_MMU */ | |
388 | ||
23ff4440 | 389 | __cacheline_aligned_in_smp DEFINE_SPINLOCK(mmlist_lock); |
1da177e4 | 390 | |
e94b1766 | 391 | #define allocate_mm() (kmem_cache_alloc(mm_cachep, GFP_KERNEL)) |
1da177e4 LT |
392 | #define free_mm(mm) (kmem_cache_free(mm_cachep, (mm))) |
393 | ||
394 | #include <linux/init_task.h> | |
395 | ||
78fb7466 | 396 | static struct mm_struct * mm_init(struct mm_struct * mm, struct task_struct *p) |
1da177e4 LT |
397 | { |
398 | atomic_set(&mm->mm_users, 1); | |
399 | atomic_set(&mm->mm_count, 1); | |
400 | init_rwsem(&mm->mmap_sem); | |
401 | INIT_LIST_HEAD(&mm->mmlist); | |
3cb4a0bb KH |
402 | mm->flags = (current->mm) ? current->mm->flags |
403 | : MMF_DUMP_FILTER_DEFAULT; | |
999d9fc1 | 404 | mm->core_state = NULL; |
1da177e4 | 405 | mm->nr_ptes = 0; |
4294621f | 406 | set_mm_counter(mm, file_rss, 0); |
404351e6 | 407 | set_mm_counter(mm, anon_rss, 0); |
1da177e4 LT |
408 | spin_lock_init(&mm->page_table_lock); |
409 | rwlock_init(&mm->ioctx_list_lock); | |
410 | mm->ioctx_list = NULL; | |
1da177e4 | 411 | mm->free_area_cache = TASK_UNMAPPED_BASE; |
1363c3cd | 412 | mm->cached_hole_size = ~0UL; |
cf475ad2 | 413 | mm_init_owner(mm, p); |
1da177e4 LT |
414 | |
415 | if (likely(!mm_alloc_pgd(mm))) { | |
416 | mm->def_flags = 0; | |
cddb8a5c | 417 | mmu_notifier_mm_init(mm); |
1da177e4 LT |
418 | return mm; |
419 | } | |
78fb7466 | 420 | |
1da177e4 LT |
421 | free_mm(mm); |
422 | return NULL; | |
423 | } | |
424 | ||
425 | /* | |
426 | * Allocate and initialize an mm_struct. | |
427 | */ | |
428 | struct mm_struct * mm_alloc(void) | |
429 | { | |
430 | struct mm_struct * mm; | |
431 | ||
432 | mm = allocate_mm(); | |
433 | if (mm) { | |
434 | memset(mm, 0, sizeof(*mm)); | |
78fb7466 | 435 | mm = mm_init(mm, current); |
1da177e4 LT |
436 | } |
437 | return mm; | |
438 | } | |
439 | ||
440 | /* | |
441 | * Called when the last reference to the mm | |
442 | * is dropped: either by a lazy thread or by | |
443 | * mmput. Free the page directory and the mm. | |
444 | */ | |
7ad5b3a5 | 445 | void __mmdrop(struct mm_struct *mm) |
1da177e4 LT |
446 | { |
447 | BUG_ON(mm == &init_mm); | |
448 | mm_free_pgd(mm); | |
449 | destroy_context(mm); | |
cddb8a5c | 450 | mmu_notifier_mm_destroy(mm); |
1da177e4 LT |
451 | free_mm(mm); |
452 | } | |
6d4e4c4f | 453 | EXPORT_SYMBOL_GPL(__mmdrop); |
1da177e4 LT |
454 | |
455 | /* | |
456 | * Decrement the use count and release all resources for an mm. | |
457 | */ | |
458 | void mmput(struct mm_struct *mm) | |
459 | { | |
0ae26f1b AM |
460 | might_sleep(); |
461 | ||
1da177e4 LT |
462 | if (atomic_dec_and_test(&mm->mm_users)) { |
463 | exit_aio(mm); | |
464 | exit_mmap(mm); | |
925d1c40 | 465 | set_mm_exe_file(mm, NULL); |
1da177e4 LT |
466 | if (!list_empty(&mm->mmlist)) { |
467 | spin_lock(&mmlist_lock); | |
468 | list_del(&mm->mmlist); | |
469 | spin_unlock(&mmlist_lock); | |
470 | } | |
471 | put_swap_token(mm); | |
472 | mmdrop(mm); | |
473 | } | |
474 | } | |
475 | EXPORT_SYMBOL_GPL(mmput); | |
476 | ||
477 | /** | |
478 | * get_task_mm - acquire a reference to the task's mm | |
479 | * | |
246bb0b1 | 480 | * Returns %NULL if the task has no mm. Checks PF_KTHREAD (meaning |
1da177e4 LT |
481 | * this kernel workthread has transiently adopted a user mm with use_mm, |
482 | * to do its AIO) is not set and if so returns a reference to it, after | |
483 | * bumping up the use count. User must release the mm via mmput() | |
484 | * after use. Typically used by /proc and ptrace. | |
485 | */ | |
486 | struct mm_struct *get_task_mm(struct task_struct *task) | |
487 | { | |
488 | struct mm_struct *mm; | |
489 | ||
490 | task_lock(task); | |
491 | mm = task->mm; | |
492 | if (mm) { | |
246bb0b1 | 493 | if (task->flags & PF_KTHREAD) |
1da177e4 LT |
494 | mm = NULL; |
495 | else | |
496 | atomic_inc(&mm->mm_users); | |
497 | } | |
498 | task_unlock(task); | |
499 | return mm; | |
500 | } | |
501 | EXPORT_SYMBOL_GPL(get_task_mm); | |
502 | ||
503 | /* Please note the differences between mmput and mm_release. | |
504 | * mmput is called whenever we stop holding onto a mm_struct, | |
505 | * error success whatever. | |
506 | * | |
507 | * mm_release is called after a mm_struct has been removed | |
508 | * from the current process. | |
509 | * | |
510 | * This difference is important for error handling, when we | |
511 | * only half set up a mm_struct for a new process and need to restore | |
512 | * the old one. Because we mmput the new mm_struct before | |
513 | * restoring the old one. . . | |
514 | * Eric Biederman 10 January 1998 | |
515 | */ | |
516 | void mm_release(struct task_struct *tsk, struct mm_struct *mm) | |
517 | { | |
518 | struct completion *vfork_done = tsk->vfork_done; | |
519 | ||
520 | /* Get rid of any cached register state */ | |
521 | deactivate_mm(tsk, mm); | |
522 | ||
523 | /* notify parent sleeping on vfork() */ | |
524 | if (vfork_done) { | |
525 | tsk->vfork_done = NULL; | |
526 | complete(vfork_done); | |
527 | } | |
fec1d011 RM |
528 | |
529 | /* | |
530 | * If we're exiting normally, clear a user-space tid field if | |
531 | * requested. We leave this alone when dying by signal, to leave | |
532 | * the value intact in a core dump, and to save the unnecessary | |
533 | * trouble otherwise. Userland only wants this done for a sys_exit. | |
534 | */ | |
535 | if (tsk->clear_child_tid | |
536 | && !(tsk->flags & PF_SIGNALED) | |
537 | && atomic_read(&mm->mm_users) > 1) { | |
1da177e4 LT |
538 | u32 __user * tidptr = tsk->clear_child_tid; |
539 | tsk->clear_child_tid = NULL; | |
540 | ||
541 | /* | |
542 | * We don't check the error code - if userspace has | |
543 | * not set up a proper pointer then tough luck. | |
544 | */ | |
545 | put_user(0, tidptr); | |
546 | sys_futex(tidptr, FUTEX_WAKE, 1, NULL, NULL, 0); | |
547 | } | |
548 | } | |
549 | ||
a0a7ec30 JD |
550 | /* |
551 | * Allocate a new mm structure and copy contents from the | |
552 | * mm structure of the passed in task structure. | |
553 | */ | |
402b0862 | 554 | struct mm_struct *dup_mm(struct task_struct *tsk) |
a0a7ec30 JD |
555 | { |
556 | struct mm_struct *mm, *oldmm = current->mm; | |
557 | int err; | |
558 | ||
559 | if (!oldmm) | |
560 | return NULL; | |
561 | ||
562 | mm = allocate_mm(); | |
563 | if (!mm) | |
564 | goto fail_nomem; | |
565 | ||
566 | memcpy(mm, oldmm, sizeof(*mm)); | |
567 | ||
7602bdf2 AC |
568 | /* Initializing for Swap token stuff */ |
569 | mm->token_priority = 0; | |
570 | mm->last_interval = 0; | |
571 | ||
78fb7466 | 572 | if (!mm_init(mm, tsk)) |
a0a7ec30 JD |
573 | goto fail_nomem; |
574 | ||
575 | if (init_new_context(tsk, mm)) | |
576 | goto fail_nocontext; | |
577 | ||
925d1c40 MH |
578 | dup_mm_exe_file(oldmm, mm); |
579 | ||
a0a7ec30 JD |
580 | err = dup_mmap(mm, oldmm); |
581 | if (err) | |
582 | goto free_pt; | |
583 | ||
584 | mm->hiwater_rss = get_mm_rss(mm); | |
585 | mm->hiwater_vm = mm->total_vm; | |
586 | ||
587 | return mm; | |
588 | ||
589 | free_pt: | |
590 | mmput(mm); | |
591 | ||
592 | fail_nomem: | |
593 | return NULL; | |
594 | ||
595 | fail_nocontext: | |
596 | /* | |
597 | * If init_new_context() failed, we cannot use mmput() to free the mm | |
598 | * because it calls destroy_context() | |
599 | */ | |
600 | mm_free_pgd(mm); | |
601 | free_mm(mm); | |
602 | return NULL; | |
603 | } | |
604 | ||
1da177e4 LT |
605 | static int copy_mm(unsigned long clone_flags, struct task_struct * tsk) |
606 | { | |
607 | struct mm_struct * mm, *oldmm; | |
608 | int retval; | |
609 | ||
610 | tsk->min_flt = tsk->maj_flt = 0; | |
611 | tsk->nvcsw = tsk->nivcsw = 0; | |
612 | ||
613 | tsk->mm = NULL; | |
614 | tsk->active_mm = NULL; | |
615 | ||
616 | /* | |
617 | * Are we cloning a kernel thread? | |
618 | * | |
619 | * We need to steal a active VM for that.. | |
620 | */ | |
621 | oldmm = current->mm; | |
622 | if (!oldmm) | |
623 | return 0; | |
624 | ||
625 | if (clone_flags & CLONE_VM) { | |
626 | atomic_inc(&oldmm->mm_users); | |
627 | mm = oldmm; | |
1da177e4 LT |
628 | goto good_mm; |
629 | } | |
630 | ||
631 | retval = -ENOMEM; | |
a0a7ec30 | 632 | mm = dup_mm(tsk); |
1da177e4 LT |
633 | if (!mm) |
634 | goto fail_nomem; | |
635 | ||
1da177e4 | 636 | good_mm: |
7602bdf2 AC |
637 | /* Initializing for Swap token stuff */ |
638 | mm->token_priority = 0; | |
639 | mm->last_interval = 0; | |
640 | ||
1da177e4 LT |
641 | tsk->mm = mm; |
642 | tsk->active_mm = mm; | |
643 | return 0; | |
644 | ||
1da177e4 LT |
645 | fail_nomem: |
646 | return retval; | |
1da177e4 LT |
647 | } |
648 | ||
a39bc516 | 649 | static struct fs_struct *__copy_fs_struct(struct fs_struct *old) |
1da177e4 LT |
650 | { |
651 | struct fs_struct *fs = kmem_cache_alloc(fs_cachep, GFP_KERNEL); | |
652 | /* We don't need to lock fs - think why ;-) */ | |
653 | if (fs) { | |
654 | atomic_set(&fs->count, 1); | |
655 | rwlock_init(&fs->lock); | |
656 | fs->umask = old->umask; | |
657 | read_lock(&old->lock); | |
6ac08c39 JB |
658 | fs->root = old->root; |
659 | path_get(&old->root); | |
660 | fs->pwd = old->pwd; | |
661 | path_get(&old->pwd); | |
1da177e4 LT |
662 | read_unlock(&old->lock); |
663 | } | |
664 | return fs; | |
665 | } | |
666 | ||
667 | struct fs_struct *copy_fs_struct(struct fs_struct *old) | |
668 | { | |
669 | return __copy_fs_struct(old); | |
670 | } | |
671 | ||
672 | EXPORT_SYMBOL_GPL(copy_fs_struct); | |
673 | ||
a39bc516 | 674 | static int copy_fs(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 LT |
675 | { |
676 | if (clone_flags & CLONE_FS) { | |
677 | atomic_inc(¤t->fs->count); | |
678 | return 0; | |
679 | } | |
680 | tsk->fs = __copy_fs_struct(current->fs); | |
681 | if (!tsk->fs) | |
682 | return -ENOMEM; | |
683 | return 0; | |
684 | } | |
685 | ||
a016f338 JD |
686 | static int copy_files(unsigned long clone_flags, struct task_struct * tsk) |
687 | { | |
688 | struct files_struct *oldf, *newf; | |
689 | int error = 0; | |
690 | ||
691 | /* | |
692 | * A background process may not have any files ... | |
693 | */ | |
694 | oldf = current->files; | |
695 | if (!oldf) | |
696 | goto out; | |
697 | ||
698 | if (clone_flags & CLONE_FILES) { | |
699 | atomic_inc(&oldf->count); | |
700 | goto out; | |
701 | } | |
702 | ||
a016f338 JD |
703 | newf = dup_fd(oldf, &error); |
704 | if (!newf) | |
705 | goto out; | |
706 | ||
707 | tsk->files = newf; | |
708 | error = 0; | |
709 | out: | |
710 | return error; | |
711 | } | |
712 | ||
fadad878 | 713 | static int copy_io(unsigned long clone_flags, struct task_struct *tsk) |
fd0928df JA |
714 | { |
715 | #ifdef CONFIG_BLOCK | |
716 | struct io_context *ioc = current->io_context; | |
717 | ||
718 | if (!ioc) | |
719 | return 0; | |
fadad878 JA |
720 | /* |
721 | * Share io context with parent, if CLONE_IO is set | |
722 | */ | |
723 | if (clone_flags & CLONE_IO) { | |
724 | tsk->io_context = ioc_task_link(ioc); | |
725 | if (unlikely(!tsk->io_context)) | |
726 | return -ENOMEM; | |
727 | } else if (ioprio_valid(ioc->ioprio)) { | |
fd0928df JA |
728 | tsk->io_context = alloc_io_context(GFP_KERNEL, -1); |
729 | if (unlikely(!tsk->io_context)) | |
730 | return -ENOMEM; | |
731 | ||
fd0928df JA |
732 | tsk->io_context->ioprio = ioc->ioprio; |
733 | } | |
734 | #endif | |
735 | return 0; | |
736 | } | |
737 | ||
a39bc516 | 738 | static int copy_sighand(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 LT |
739 | { |
740 | struct sighand_struct *sig; | |
741 | ||
742 | if (clone_flags & (CLONE_SIGHAND | CLONE_THREAD)) { | |
743 | atomic_inc(¤t->sighand->count); | |
744 | return 0; | |
745 | } | |
746 | sig = kmem_cache_alloc(sighand_cachep, GFP_KERNEL); | |
e56d0903 | 747 | rcu_assign_pointer(tsk->sighand, sig); |
1da177e4 LT |
748 | if (!sig) |
749 | return -ENOMEM; | |
1da177e4 LT |
750 | atomic_set(&sig->count, 1); |
751 | memcpy(sig->action, current->sighand->action, sizeof(sig->action)); | |
752 | return 0; | |
753 | } | |
754 | ||
a7e5328a | 755 | void __cleanup_sighand(struct sighand_struct *sighand) |
c81addc9 | 756 | { |
c81addc9 ON |
757 | if (atomic_dec_and_test(&sighand->count)) |
758 | kmem_cache_free(sighand_cachep, sighand); | |
759 | } | |
760 | ||
f06febc9 FM |
761 | |
762 | /* | |
763 | * Initialize POSIX timer handling for a thread group. | |
764 | */ | |
765 | static void posix_cpu_timers_init_group(struct signal_struct *sig) | |
766 | { | |
767 | /* Thread group counters. */ | |
768 | thread_group_cputime_init(sig); | |
769 | ||
770 | /* Expiration times and increments. */ | |
771 | sig->it_virt_expires = cputime_zero; | |
772 | sig->it_virt_incr = cputime_zero; | |
773 | sig->it_prof_expires = cputime_zero; | |
774 | sig->it_prof_incr = cputime_zero; | |
775 | ||
776 | /* Cached expiration times. */ | |
777 | sig->cputime_expires.prof_exp = cputime_zero; | |
778 | sig->cputime_expires.virt_exp = cputime_zero; | |
779 | sig->cputime_expires.sched_exp = 0; | |
780 | ||
781 | /* The timer lists. */ | |
782 | INIT_LIST_HEAD(&sig->cpu_timers[0]); | |
783 | INIT_LIST_HEAD(&sig->cpu_timers[1]); | |
784 | INIT_LIST_HEAD(&sig->cpu_timers[2]); | |
785 | } | |
786 | ||
a39bc516 | 787 | static int copy_signal(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 LT |
788 | { |
789 | struct signal_struct *sig; | |
790 | int ret; | |
791 | ||
792 | if (clone_flags & CLONE_THREAD) { | |
bb34d92f | 793 | ret = thread_group_cputime_clone_thread(current); |
f06febc9 FM |
794 | if (likely(!ret)) { |
795 | atomic_inc(¤t->signal->count); | |
796 | atomic_inc(¤t->signal->live); | |
797 | } | |
798 | return ret; | |
1da177e4 LT |
799 | } |
800 | sig = kmem_cache_alloc(signal_cachep, GFP_KERNEL); | |
801 | tsk->signal = sig; | |
802 | if (!sig) | |
803 | return -ENOMEM; | |
804 | ||
1da177e4 LT |
805 | atomic_set(&sig->count, 1); |
806 | atomic_set(&sig->live, 1); | |
807 | init_waitqueue_head(&sig->wait_chldexit); | |
808 | sig->flags = 0; | |
809 | sig->group_exit_code = 0; | |
810 | sig->group_exit_task = NULL; | |
811 | sig->group_stop_count = 0; | |
db51aecc | 812 | sig->curr_target = tsk; |
1da177e4 LT |
813 | init_sigpending(&sig->shared_pending); |
814 | INIT_LIST_HEAD(&sig->posix_timers); | |
815 | ||
c9cb2e3d | 816 | hrtimer_init(&sig->real_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL); |
2ff678b8 | 817 | sig->it_real_incr.tv64 = 0; |
1da177e4 | 818 | sig->real_timer.function = it_real_fn; |
1da177e4 | 819 | |
1da177e4 | 820 | sig->leader = 0; /* session leadership doesn't inherit */ |
ab521dc0 | 821 | sig->tty_old_pgrp = NULL; |
9c9f4ded | 822 | sig->tty = NULL; |
1da177e4 | 823 | |
f06febc9 | 824 | sig->cutime = sig->cstime = cputime_zero; |
9ac52315 LV |
825 | sig->gtime = cputime_zero; |
826 | sig->cgtime = cputime_zero; | |
1da177e4 LT |
827 | sig->nvcsw = sig->nivcsw = sig->cnvcsw = sig->cnivcsw = 0; |
828 | sig->min_flt = sig->maj_flt = sig->cmin_flt = sig->cmaj_flt = 0; | |
6eaeeaba | 829 | sig->inblock = sig->oublock = sig->cinblock = sig->coublock = 0; |
5995477a | 830 | task_io_accounting_init(&sig->ioac); |
ad4ecbcb | 831 | taskstats_tgid_init(sig); |
1da177e4 LT |
832 | |
833 | task_lock(current->group_leader); | |
834 | memcpy(sig->rlim, current->signal->rlim, sizeof sig->rlim); | |
835 | task_unlock(current->group_leader); | |
836 | ||
f06febc9 FM |
837 | posix_cpu_timers_init_group(sig); |
838 | ||
0e464814 | 839 | acct_init_pacct(&sig->pacct); |
1da177e4 | 840 | |
522ed776 MT |
841 | tty_audit_fork(sig); |
842 | ||
1da177e4 LT |
843 | return 0; |
844 | } | |
845 | ||
6b3934ef ON |
846 | void __cleanup_signal(struct signal_struct *sig) |
847 | { | |
f06febc9 | 848 | thread_group_cputime_free(sig); |
9c9f4ded | 849 | tty_kref_put(sig->tty); |
6b3934ef ON |
850 | kmem_cache_free(signal_cachep, sig); |
851 | } | |
852 | ||
a39bc516 | 853 | static void cleanup_signal(struct task_struct *tsk) |
6b3934ef ON |
854 | { |
855 | struct signal_struct *sig = tsk->signal; | |
856 | ||
857 | atomic_dec(&sig->live); | |
858 | ||
859 | if (atomic_dec_and_test(&sig->count)) | |
860 | __cleanup_signal(sig); | |
861 | } | |
862 | ||
a39bc516 | 863 | static void copy_flags(unsigned long clone_flags, struct task_struct *p) |
1da177e4 LT |
864 | { |
865 | unsigned long new_flags = p->flags; | |
866 | ||
83144186 | 867 | new_flags &= ~PF_SUPERPRIV; |
1da177e4 | 868 | new_flags |= PF_FORKNOEXEC; |
09a05394 | 869 | new_flags |= PF_STARTING; |
1da177e4 | 870 | p->flags = new_flags; |
2e131895 | 871 | clear_freeze_flag(p); |
1da177e4 LT |
872 | } |
873 | ||
874 | asmlinkage long sys_set_tid_address(int __user *tidptr) | |
875 | { | |
876 | current->clear_child_tid = tidptr; | |
877 | ||
b488893a | 878 | return task_pid_vnr(current); |
1da177e4 LT |
879 | } |
880 | ||
a39bc516 | 881 | static void rt_mutex_init_task(struct task_struct *p) |
23f78d4a | 882 | { |
23f78d4a | 883 | spin_lock_init(&p->pi_lock); |
e29e175b | 884 | #ifdef CONFIG_RT_MUTEXES |
23f78d4a IM |
885 | plist_head_init(&p->pi_waiters, &p->pi_lock); |
886 | p->pi_blocked_on = NULL; | |
23f78d4a IM |
887 | #endif |
888 | } | |
889 | ||
cf475ad2 BS |
890 | #ifdef CONFIG_MM_OWNER |
891 | void mm_init_owner(struct mm_struct *mm, struct task_struct *p) | |
892 | { | |
893 | mm->owner = p; | |
894 | } | |
895 | #endif /* CONFIG_MM_OWNER */ | |
896 | ||
f06febc9 FM |
897 | /* |
898 | * Initialize POSIX timer handling for a single task. | |
899 | */ | |
900 | static void posix_cpu_timers_init(struct task_struct *tsk) | |
901 | { | |
902 | tsk->cputime_expires.prof_exp = cputime_zero; | |
903 | tsk->cputime_expires.virt_exp = cputime_zero; | |
904 | tsk->cputime_expires.sched_exp = 0; | |
905 | INIT_LIST_HEAD(&tsk->cpu_timers[0]); | |
906 | INIT_LIST_HEAD(&tsk->cpu_timers[1]); | |
907 | INIT_LIST_HEAD(&tsk->cpu_timers[2]); | |
908 | } | |
909 | ||
1da177e4 LT |
910 | /* |
911 | * This creates a new process as a copy of the old one, | |
912 | * but does not actually start it yet. | |
913 | * | |
914 | * It copies the registers, and all the appropriate | |
915 | * parts of the process environment (as per the clone | |
916 | * flags). The actual kick-off is left to the caller. | |
917 | */ | |
36c8b586 IM |
918 | static struct task_struct *copy_process(unsigned long clone_flags, |
919 | unsigned long stack_start, | |
920 | struct pt_regs *regs, | |
921 | unsigned long stack_size, | |
36c8b586 | 922 | int __user *child_tidptr, |
09a05394 RM |
923 | struct pid *pid, |
924 | int trace) | |
1da177e4 LT |
925 | { |
926 | int retval; | |
a24efe62 | 927 | struct task_struct *p; |
b4f48b63 | 928 | int cgroup_callbacks_done = 0; |
1da177e4 LT |
929 | |
930 | if ((clone_flags & (CLONE_NEWNS|CLONE_FS)) == (CLONE_NEWNS|CLONE_FS)) | |
931 | return ERR_PTR(-EINVAL); | |
932 | ||
933 | /* | |
934 | * Thread groups must share signals as well, and detached threads | |
935 | * can only be started up within the thread group. | |
936 | */ | |
937 | if ((clone_flags & CLONE_THREAD) && !(clone_flags & CLONE_SIGHAND)) | |
938 | return ERR_PTR(-EINVAL); | |
939 | ||
940 | /* | |
941 | * Shared signal handlers imply shared VM. By way of the above, | |
942 | * thread groups also imply shared VM. Blocking this case allows | |
943 | * for various simplifications in other code. | |
944 | */ | |
945 | if ((clone_flags & CLONE_SIGHAND) && !(clone_flags & CLONE_VM)) | |
946 | return ERR_PTR(-EINVAL); | |
947 | ||
948 | retval = security_task_create(clone_flags); | |
949 | if (retval) | |
950 | goto fork_out; | |
951 | ||
952 | retval = -ENOMEM; | |
953 | p = dup_task_struct(current); | |
954 | if (!p) | |
955 | goto fork_out; | |
956 | ||
bea493a0 PZ |
957 | rt_mutex_init_task(p); |
958 | ||
d12c1a37 | 959 | #ifdef CONFIG_PROVE_LOCKING |
de30a2b3 IM |
960 | DEBUG_LOCKS_WARN_ON(!p->hardirqs_enabled); |
961 | DEBUG_LOCKS_WARN_ON(!p->softirqs_enabled); | |
962 | #endif | |
1da177e4 | 963 | retval = -EAGAIN; |
b6dff3ec | 964 | if (atomic_read(&p->cred->user->processes) >= |
1da177e4 LT |
965 | p->signal->rlim[RLIMIT_NPROC].rlim_cur) { |
966 | if (!capable(CAP_SYS_ADMIN) && !capable(CAP_SYS_RESOURCE) && | |
b6dff3ec | 967 | p->cred->user != current->nsproxy->user_ns->root_user) |
1da177e4 LT |
968 | goto bad_fork_free; |
969 | } | |
970 | ||
f1752eec DH |
971 | retval = copy_creds(p, clone_flags); |
972 | if (retval < 0) | |
973 | goto bad_fork_free; | |
1da177e4 LT |
974 | |
975 | /* | |
976 | * If multiple threads are within copy_process(), then this check | |
977 | * triggers too late. This doesn't hurt, the check is only there | |
978 | * to stop root fork bombs. | |
979 | */ | |
980 | if (nr_threads >= max_threads) | |
981 | goto bad_fork_cleanup_count; | |
982 | ||
a1261f54 | 983 | if (!try_module_get(task_thread_info(p)->exec_domain->module)) |
1da177e4 LT |
984 | goto bad_fork_cleanup_count; |
985 | ||
986 | if (p->binfmt && !try_module_get(p->binfmt->module)) | |
987 | goto bad_fork_cleanup_put_domain; | |
988 | ||
989 | p->did_exec = 0; | |
ca74e92b | 990 | delayacct_tsk_init(p); /* Must remain after dup_task_struct() */ |
1da177e4 | 991 | copy_flags(clone_flags, p); |
1da177e4 LT |
992 | INIT_LIST_HEAD(&p->children); |
993 | INIT_LIST_HEAD(&p->sibling); | |
e260be67 PM |
994 | #ifdef CONFIG_PREEMPT_RCU |
995 | p->rcu_read_lock_nesting = 0; | |
996 | p->rcu_flipctr_idx = 0; | |
997 | #endif /* #ifdef CONFIG_PREEMPT_RCU */ | |
1da177e4 LT |
998 | p->vfork_done = NULL; |
999 | spin_lock_init(&p->alloc_lock); | |
1da177e4 LT |
1000 | |
1001 | clear_tsk_thread_flag(p, TIF_SIGPENDING); | |
1002 | init_sigpending(&p->pending); | |
1003 | ||
1004 | p->utime = cputime_zero; | |
1005 | p->stime = cputime_zero; | |
9ac52315 | 1006 | p->gtime = cputime_zero; |
c66f08be MN |
1007 | p->utimescaled = cputime_zero; |
1008 | p->stimescaled = cputime_zero; | |
73a2bcb0 | 1009 | p->prev_utime = cputime_zero; |
9301899b | 1010 | p->prev_stime = cputime_zero; |
172ba844 | 1011 | |
6976675d AV |
1012 | p->default_timer_slack_ns = current->timer_slack_ns; |
1013 | ||
82a1fcb9 IM |
1014 | #ifdef CONFIG_DETECT_SOFTLOCKUP |
1015 | p->last_switch_count = 0; | |
1016 | p->last_switch_timestamp = 0; | |
1017 | #endif | |
1018 | ||
5995477a | 1019 | task_io_accounting_init(&p->ioac); |
1da177e4 LT |
1020 | acct_clear_integrals(p); |
1021 | ||
f06febc9 | 1022 | posix_cpu_timers_init(p); |
1da177e4 LT |
1023 | |
1024 | p->lock_depth = -1; /* -1 = no lock */ | |
1025 | do_posix_clock_monotonic_gettime(&p->start_time); | |
924b42d5 TJ |
1026 | p->real_start_time = p->start_time; |
1027 | monotonic_to_bootbased(&p->real_start_time); | |
1da177e4 | 1028 | p->io_context = NULL; |
1da177e4 | 1029 | p->audit_context = NULL; |
b4f48b63 | 1030 | cgroup_fork(p); |
1da177e4 | 1031 | #ifdef CONFIG_NUMA |
846a16bf | 1032 | p->mempolicy = mpol_dup(p->mempolicy); |
1da177e4 LT |
1033 | if (IS_ERR(p->mempolicy)) { |
1034 | retval = PTR_ERR(p->mempolicy); | |
1035 | p->mempolicy = NULL; | |
b4f48b63 | 1036 | goto bad_fork_cleanup_cgroup; |
1da177e4 | 1037 | } |
c61afb18 | 1038 | mpol_fix_fork_child_flag(p); |
1da177e4 | 1039 | #endif |
de30a2b3 IM |
1040 | #ifdef CONFIG_TRACE_IRQFLAGS |
1041 | p->irq_events = 0; | |
b36e4758 RK |
1042 | #ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW |
1043 | p->hardirqs_enabled = 1; | |
1044 | #else | |
de30a2b3 | 1045 | p->hardirqs_enabled = 0; |
b36e4758 | 1046 | #endif |
de30a2b3 IM |
1047 | p->hardirq_enable_ip = 0; |
1048 | p->hardirq_enable_event = 0; | |
1049 | p->hardirq_disable_ip = _THIS_IP_; | |
1050 | p->hardirq_disable_event = 0; | |
1051 | p->softirqs_enabled = 1; | |
1052 | p->softirq_enable_ip = _THIS_IP_; | |
1053 | p->softirq_enable_event = 0; | |
1054 | p->softirq_disable_ip = 0; | |
1055 | p->softirq_disable_event = 0; | |
1056 | p->hardirq_context = 0; | |
1057 | p->softirq_context = 0; | |
1058 | #endif | |
fbb9ce95 IM |
1059 | #ifdef CONFIG_LOCKDEP |
1060 | p->lockdep_depth = 0; /* no locks held yet */ | |
1061 | p->curr_chain_key = 0; | |
1062 | p->lockdep_recursion = 0; | |
1063 | #endif | |
1da177e4 | 1064 | |
408894ee IM |
1065 | #ifdef CONFIG_DEBUG_MUTEXES |
1066 | p->blocked_on = NULL; /* not blocked yet */ | |
1067 | #endif | |
1068 | ||
3c90e6e9 SV |
1069 | /* Perform scheduler related setup. Assign this task to a CPU. */ |
1070 | sched_fork(p, clone_flags); | |
1071 | ||
1da177e4 | 1072 | if ((retval = audit_alloc(p))) |
f1752eec | 1073 | goto bad_fork_cleanup_policy; |
1da177e4 LT |
1074 | /* copy all the process information */ |
1075 | if ((retval = copy_semundo(clone_flags, p))) | |
1076 | goto bad_fork_cleanup_audit; | |
1077 | if ((retval = copy_files(clone_flags, p))) | |
1078 | goto bad_fork_cleanup_semundo; | |
1079 | if ((retval = copy_fs(clone_flags, p))) | |
1080 | goto bad_fork_cleanup_files; | |
1081 | if ((retval = copy_sighand(clone_flags, p))) | |
1082 | goto bad_fork_cleanup_fs; | |
1083 | if ((retval = copy_signal(clone_flags, p))) | |
1084 | goto bad_fork_cleanup_sighand; | |
1085 | if ((retval = copy_mm(clone_flags, p))) | |
1086 | goto bad_fork_cleanup_signal; | |
ab516013 | 1087 | if ((retval = copy_namespaces(clone_flags, p))) |
d84f4f99 | 1088 | goto bad_fork_cleanup_mm; |
fadad878 | 1089 | if ((retval = copy_io(clone_flags, p))) |
fd0928df | 1090 | goto bad_fork_cleanup_namespaces; |
1da177e4 LT |
1091 | retval = copy_thread(0, clone_flags, stack_start, stack_size, p, regs); |
1092 | if (retval) | |
fd0928df | 1093 | goto bad_fork_cleanup_io; |
1da177e4 | 1094 | |
425fb2b4 PE |
1095 | if (pid != &init_struct_pid) { |
1096 | retval = -ENOMEM; | |
1097 | pid = alloc_pid(task_active_pid_ns(p)); | |
1098 | if (!pid) | |
fd0928df | 1099 | goto bad_fork_cleanup_io; |
6f4e6433 PE |
1100 | |
1101 | if (clone_flags & CLONE_NEWPID) { | |
1102 | retval = pid_ns_prepare_proc(task_active_pid_ns(p)); | |
1103 | if (retval < 0) | |
1104 | goto bad_fork_free_pid; | |
1105 | } | |
425fb2b4 PE |
1106 | } |
1107 | ||
1108 | p->pid = pid_nr(pid); | |
1109 | p->tgid = p->pid; | |
1110 | if (clone_flags & CLONE_THREAD) | |
1111 | p->tgid = current->tgid; | |
1112 | ||
e885dcde SH |
1113 | if (current->nsproxy != p->nsproxy) { |
1114 | retval = ns_cgroup_clone(p, pid); | |
1115 | if (retval) | |
1116 | goto bad_fork_free_pid; | |
1117 | } | |
1118 | ||
1da177e4 LT |
1119 | p->set_child_tid = (clone_flags & CLONE_CHILD_SETTID) ? child_tidptr : NULL; |
1120 | /* | |
1121 | * Clear TID on mm_release()? | |
1122 | */ | |
1123 | p->clear_child_tid = (clone_flags & CLONE_CHILD_CLEARTID) ? child_tidptr: NULL; | |
42b2dd0a | 1124 | #ifdef CONFIG_FUTEX |
8f17d3a5 IM |
1125 | p->robust_list = NULL; |
1126 | #ifdef CONFIG_COMPAT | |
1127 | p->compat_robust_list = NULL; | |
1128 | #endif | |
c87e2837 IM |
1129 | INIT_LIST_HEAD(&p->pi_state_list); |
1130 | p->pi_state_cache = NULL; | |
42b2dd0a | 1131 | #endif |
f9a3879a GM |
1132 | /* |
1133 | * sigaltstack should be cleared when sharing the same VM | |
1134 | */ | |
1135 | if ((clone_flags & (CLONE_VM|CLONE_VFORK)) == CLONE_VM) | |
1136 | p->sas_ss_sp = p->sas_ss_size = 0; | |
1137 | ||
1da177e4 LT |
1138 | /* |
1139 | * Syscall tracing should be turned off in the child regardless | |
1140 | * of CLONE_PTRACE. | |
1141 | */ | |
1142 | clear_tsk_thread_flag(p, TIF_SYSCALL_TRACE); | |
ed75e8d5 LV |
1143 | #ifdef TIF_SYSCALL_EMU |
1144 | clear_tsk_thread_flag(p, TIF_SYSCALL_EMU); | |
1145 | #endif | |
9745512c | 1146 | clear_all_latency_tracing(p); |
1da177e4 LT |
1147 | |
1148 | /* Our parent execution domain becomes current domain | |
1149 | These must match for thread signalling to apply */ | |
1da177e4 LT |
1150 | p->parent_exec_id = p->self_exec_id; |
1151 | ||
1152 | /* ok, now we should be set up.. */ | |
1153 | p->exit_signal = (clone_flags & CLONE_THREAD) ? -1 : (clone_flags & CSIGNAL); | |
1154 | p->pdeath_signal = 0; | |
1155 | p->exit_state = 0; | |
1156 | ||
1da177e4 LT |
1157 | /* |
1158 | * Ok, make it visible to the rest of the system. | |
1159 | * We dont wake it up yet. | |
1160 | */ | |
1161 | p->group_leader = p; | |
47e65328 | 1162 | INIT_LIST_HEAD(&p->thread_group); |
1da177e4 | 1163 | |
b4f48b63 PM |
1164 | /* Now that the task is set up, run cgroup callbacks if |
1165 | * necessary. We need to run them before the task is visible | |
1166 | * on the tasklist. */ | |
1167 | cgroup_fork_callbacks(p); | |
1168 | cgroup_callbacks_done = 1; | |
1169 | ||
1da177e4 LT |
1170 | /* Need tasklist lock for parent etc handling! */ |
1171 | write_lock_irq(&tasklist_lock); | |
1172 | ||
1173 | /* | |
476d139c NP |
1174 | * The task hasn't been attached yet, so its cpus_allowed mask will |
1175 | * not be changed, nor will its assigned CPU. | |
1176 | * | |
1177 | * The cpus_allowed mask of the parent may have changed after it was | |
1178 | * copied first time - so re-copy it here, then check the child's CPU | |
1179 | * to ensure it is on a valid CPU (and if not, just force it back to | |
1180 | * parent's CPU). This avoids alot of nasty races. | |
1da177e4 LT |
1181 | */ |
1182 | p->cpus_allowed = current->cpus_allowed; | |
6f505b16 | 1183 | p->rt.nr_cpus_allowed = current->rt.nr_cpus_allowed; |
26ff6ad9 SV |
1184 | if (unlikely(!cpu_isset(task_cpu(p), p->cpus_allowed) || |
1185 | !cpu_online(task_cpu(p)))) | |
476d139c | 1186 | set_task_cpu(p, smp_processor_id()); |
1da177e4 | 1187 | |
1da177e4 LT |
1188 | /* CLONE_PARENT re-uses the old parent */ |
1189 | if (clone_flags & (CLONE_PARENT|CLONE_THREAD)) | |
1190 | p->real_parent = current->real_parent; | |
1191 | else | |
1192 | p->real_parent = current; | |
1da177e4 | 1193 | |
3f17da69 | 1194 | spin_lock(¤t->sighand->siglock); |
4a2c7a78 ON |
1195 | |
1196 | /* | |
1197 | * Process group and session signals need to be delivered to just the | |
1198 | * parent before the fork or both the parent and the child after the | |
1199 | * fork. Restart if a signal comes in before we add the new process to | |
1200 | * it's process group. | |
1201 | * A fatal signal pending means that current will exit, so the new | |
1202 | * thread can't slip out of an OOM kill (or normal SIGKILL). | |
1203 | */ | |
23ff4440 | 1204 | recalc_sigpending(); |
4a2c7a78 ON |
1205 | if (signal_pending(current)) { |
1206 | spin_unlock(¤t->sighand->siglock); | |
1207 | write_unlock_irq(&tasklist_lock); | |
1208 | retval = -ERESTARTNOINTR; | |
425fb2b4 | 1209 | goto bad_fork_free_pid; |
4a2c7a78 ON |
1210 | } |
1211 | ||
1da177e4 | 1212 | if (clone_flags & CLONE_THREAD) { |
1da177e4 | 1213 | p->group_leader = current->group_leader; |
47e65328 | 1214 | list_add_tail_rcu(&p->thread_group, &p->group_leader->thread_group); |
1da177e4 LT |
1215 | } |
1216 | ||
73b9ebfe | 1217 | if (likely(p->pid)) { |
f470021a | 1218 | list_add_tail(&p->sibling, &p->real_parent->children); |
09a05394 | 1219 | tracehook_finish_clone(p, clone_flags, trace); |
73b9ebfe ON |
1220 | |
1221 | if (thread_group_leader(p)) { | |
5cd17569 | 1222 | if (clone_flags & CLONE_NEWPID) |
30e49c26 | 1223 | p->nsproxy->pid_ns->child_reaper = p; |
73b9ebfe | 1224 | |
fea9d175 | 1225 | p->signal->leader_pid = pid; |
9c9f4ded AC |
1226 | tty_kref_put(p->signal->tty); |
1227 | p->signal->tty = tty_kref_get(current->signal->tty); | |
5cd17569 EB |
1228 | set_task_pgrp(p, task_pgrp_nr(current)); |
1229 | set_task_session(p, task_session_nr(current)); | |
1230 | attach_pid(p, PIDTYPE_PGID, task_pgrp(current)); | |
1231 | attach_pid(p, PIDTYPE_SID, task_session(current)); | |
5e85d4ab | 1232 | list_add_tail_rcu(&p->tasks, &init_task.tasks); |
1da177e4 | 1233 | __get_cpu_var(process_counts)++; |
73b9ebfe | 1234 | } |
85868995 | 1235 | attach_pid(p, PIDTYPE_PID, pid); |
73b9ebfe | 1236 | nr_threads++; |
1da177e4 LT |
1237 | } |
1238 | ||
1da177e4 | 1239 | total_forks++; |
3f17da69 | 1240 | spin_unlock(¤t->sighand->siglock); |
1da177e4 | 1241 | write_unlock_irq(&tasklist_lock); |
c13cf856 | 1242 | proc_fork_connector(p); |
817929ec | 1243 | cgroup_post_fork(p); |
1da177e4 LT |
1244 | return p; |
1245 | ||
425fb2b4 PE |
1246 | bad_fork_free_pid: |
1247 | if (pid != &init_struct_pid) | |
1248 | free_pid(pid); | |
fd0928df JA |
1249 | bad_fork_cleanup_io: |
1250 | put_io_context(p->io_context); | |
ab516013 | 1251 | bad_fork_cleanup_namespaces: |
444f378b | 1252 | exit_task_namespaces(p); |
1da177e4 LT |
1253 | bad_fork_cleanup_mm: |
1254 | if (p->mm) | |
1255 | mmput(p->mm); | |
1256 | bad_fork_cleanup_signal: | |
6b3934ef | 1257 | cleanup_signal(p); |
1da177e4 | 1258 | bad_fork_cleanup_sighand: |
a7e5328a | 1259 | __cleanup_sighand(p->sighand); |
1da177e4 LT |
1260 | bad_fork_cleanup_fs: |
1261 | exit_fs(p); /* blocking */ | |
1262 | bad_fork_cleanup_files: | |
1263 | exit_files(p); /* blocking */ | |
1264 | bad_fork_cleanup_semundo: | |
1265 | exit_sem(p); | |
1266 | bad_fork_cleanup_audit: | |
1267 | audit_free(p); | |
1da177e4 LT |
1268 | bad_fork_cleanup_policy: |
1269 | #ifdef CONFIG_NUMA | |
f0be3d32 | 1270 | mpol_put(p->mempolicy); |
b4f48b63 | 1271 | bad_fork_cleanup_cgroup: |
1da177e4 | 1272 | #endif |
b4f48b63 | 1273 | cgroup_exit(p, cgroup_callbacks_done); |
35df17c5 | 1274 | delayacct_tsk_free(p); |
1da177e4 LT |
1275 | if (p->binfmt) |
1276 | module_put(p->binfmt->module); | |
1277 | bad_fork_cleanup_put_domain: | |
a1261f54 | 1278 | module_put(task_thread_info(p)->exec_domain->module); |
1da177e4 | 1279 | bad_fork_cleanup_count: |
d84f4f99 | 1280 | atomic_dec(&p->cred->user->processes); |
f1752eec | 1281 | put_cred(p->cred); |
1da177e4 LT |
1282 | bad_fork_free: |
1283 | free_task(p); | |
fe7d37d1 ON |
1284 | fork_out: |
1285 | return ERR_PTR(retval); | |
1da177e4 LT |
1286 | } |
1287 | ||
6b2fb3c6 | 1288 | noinline struct pt_regs * __cpuinit __attribute__((weak)) idle_regs(struct pt_regs *regs) |
1da177e4 LT |
1289 | { |
1290 | memset(regs, 0, sizeof(struct pt_regs)); | |
1291 | return regs; | |
1292 | } | |
1293 | ||
9abcf40b | 1294 | struct task_struct * __cpuinit fork_idle(int cpu) |
1da177e4 | 1295 | { |
36c8b586 | 1296 | struct task_struct *task; |
1da177e4 LT |
1297 | struct pt_regs regs; |
1298 | ||
30e49c26 | 1299 | task = copy_process(CLONE_VM, 0, idle_regs(®s), 0, NULL, |
09a05394 | 1300 | &init_struct_pid, 0); |
753ca4f3 AM |
1301 | if (!IS_ERR(task)) |
1302 | init_idle(task, cpu); | |
73b9ebfe | 1303 | |
1da177e4 LT |
1304 | return task; |
1305 | } | |
1306 | ||
1da177e4 LT |
1307 | /* |
1308 | * Ok, this is the main fork-routine. | |
1309 | * | |
1310 | * It copies the process, and if successful kick-starts | |
1311 | * it and waits for it to finish using the VM if required. | |
1312 | */ | |
1313 | long do_fork(unsigned long clone_flags, | |
1314 | unsigned long stack_start, | |
1315 | struct pt_regs *regs, | |
1316 | unsigned long stack_size, | |
1317 | int __user *parent_tidptr, | |
1318 | int __user *child_tidptr) | |
1319 | { | |
1320 | struct task_struct *p; | |
1321 | int trace = 0; | |
92476d7f | 1322 | long nr; |
1da177e4 | 1323 | |
bdff746a AM |
1324 | /* |
1325 | * We hope to recycle these flags after 2.6.26 | |
1326 | */ | |
1327 | if (unlikely(clone_flags & CLONE_STOPPED)) { | |
1328 | static int __read_mostly count = 100; | |
1329 | ||
1330 | if (count > 0 && printk_ratelimit()) { | |
1331 | char comm[TASK_COMM_LEN]; | |
1332 | ||
1333 | count--; | |
1334 | printk(KERN_INFO "fork(): process `%s' used deprecated " | |
1335 | "clone flags 0x%lx\n", | |
1336 | get_task_comm(comm, current), | |
1337 | clone_flags & CLONE_STOPPED); | |
1338 | } | |
1339 | } | |
1340 | ||
09a05394 RM |
1341 | /* |
1342 | * When called from kernel_thread, don't do user tracing stuff. | |
1343 | */ | |
1344 | if (likely(user_mode(regs))) | |
1345 | trace = tracehook_prepare_clone(clone_flags); | |
1da177e4 | 1346 | |
a6f5e063 | 1347 | p = copy_process(clone_flags, stack_start, regs, stack_size, |
09a05394 | 1348 | child_tidptr, NULL, trace); |
1da177e4 LT |
1349 | /* |
1350 | * Do this prior waking up the new thread - the thread pointer | |
1351 | * might get invalid after that point, if the thread exits quickly. | |
1352 | */ | |
1353 | if (!IS_ERR(p)) { | |
1354 | struct completion vfork; | |
1355 | ||
0a16b607 MD |
1356 | trace_sched_process_fork(current, p); |
1357 | ||
6c5f3e7b | 1358 | nr = task_pid_vnr(p); |
30e49c26 PE |
1359 | |
1360 | if (clone_flags & CLONE_PARENT_SETTID) | |
1361 | put_user(nr, parent_tidptr); | |
a6f5e063 | 1362 | |
1da177e4 LT |
1363 | if (clone_flags & CLONE_VFORK) { |
1364 | p->vfork_done = &vfork; | |
1365 | init_completion(&vfork); | |
1366 | } | |
1367 | ||
09a05394 RM |
1368 | tracehook_report_clone(trace, regs, clone_flags, nr, p); |
1369 | ||
1370 | /* | |
1371 | * We set PF_STARTING at creation in case tracing wants to | |
1372 | * use this to distinguish a fully live task from one that | |
1373 | * hasn't gotten to tracehook_report_clone() yet. Now we | |
1374 | * clear it and set the child going. | |
1375 | */ | |
1376 | p->flags &= ~PF_STARTING; | |
1377 | ||
1378 | if (unlikely(clone_flags & CLONE_STOPPED)) { | |
1da177e4 LT |
1379 | /* |
1380 | * We'll start up with an immediate SIGSTOP. | |
1381 | */ | |
1382 | sigaddset(&p->pending.signal, SIGSTOP); | |
1383 | set_tsk_thread_flag(p, TIF_SIGPENDING); | |
d9ae90ac | 1384 | __set_task_state(p, TASK_STOPPED); |
09a05394 RM |
1385 | } else { |
1386 | wake_up_new_task(p, clone_flags); | |
1da177e4 LT |
1387 | } |
1388 | ||
09a05394 RM |
1389 | tracehook_report_clone_complete(trace, regs, |
1390 | clone_flags, nr, p); | |
1391 | ||
1da177e4 | 1392 | if (clone_flags & CLONE_VFORK) { |
ba96a0c8 | 1393 | freezer_do_not_count(); |
1da177e4 | 1394 | wait_for_completion(&vfork); |
ba96a0c8 | 1395 | freezer_count(); |
daded34b | 1396 | tracehook_report_vfork_done(p, nr); |
1da177e4 LT |
1397 | } |
1398 | } else { | |
92476d7f | 1399 | nr = PTR_ERR(p); |
1da177e4 | 1400 | } |
92476d7f | 1401 | return nr; |
1da177e4 LT |
1402 | } |
1403 | ||
5fd63b30 RT |
1404 | #ifndef ARCH_MIN_MMSTRUCT_ALIGN |
1405 | #define ARCH_MIN_MMSTRUCT_ALIGN 0 | |
1406 | #endif | |
1407 | ||
51cc5068 | 1408 | static void sighand_ctor(void *data) |
aa1757f9 ON |
1409 | { |
1410 | struct sighand_struct *sighand = data; | |
1411 | ||
a35afb83 | 1412 | spin_lock_init(&sighand->siglock); |
b8fceee1 | 1413 | init_waitqueue_head(&sighand->signalfd_wqh); |
aa1757f9 ON |
1414 | } |
1415 | ||
1da177e4 LT |
1416 | void __init proc_caches_init(void) |
1417 | { | |
1418 | sighand_cachep = kmem_cache_create("sighand_cache", | |
1419 | sizeof(struct sighand_struct), 0, | |
aa1757f9 | 1420 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_DESTROY_BY_RCU, |
20c2df83 | 1421 | sighand_ctor); |
1da177e4 LT |
1422 | signal_cachep = kmem_cache_create("signal_cache", |
1423 | sizeof(struct signal_struct), 0, | |
20c2df83 PM |
1424 | SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL); |
1425 | files_cachep = kmem_cache_create("files_cache", | |
1da177e4 | 1426 | sizeof(struct files_struct), 0, |
20c2df83 PM |
1427 | SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL); |
1428 | fs_cachep = kmem_cache_create("fs_cache", | |
1da177e4 | 1429 | sizeof(struct fs_struct), 0, |
20c2df83 | 1430 | SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL); |
1da177e4 LT |
1431 | vm_area_cachep = kmem_cache_create("vm_area_struct", |
1432 | sizeof(struct vm_area_struct), 0, | |
20c2df83 | 1433 | SLAB_PANIC, NULL); |
1da177e4 | 1434 | mm_cachep = kmem_cache_create("mm_struct", |
5fd63b30 | 1435 | sizeof(struct mm_struct), ARCH_MIN_MMSTRUCT_ALIGN, |
20c2df83 | 1436 | SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL); |
1da177e4 | 1437 | } |
cf2e340f | 1438 | |
cf2e340f JD |
1439 | /* |
1440 | * Check constraints on flags passed to the unshare system call and | |
1441 | * force unsharing of additional process context as appropriate. | |
1442 | */ | |
a39bc516 | 1443 | static void check_unshare_flags(unsigned long *flags_ptr) |
cf2e340f JD |
1444 | { |
1445 | /* | |
1446 | * If unsharing a thread from a thread group, must also | |
1447 | * unshare vm. | |
1448 | */ | |
1449 | if (*flags_ptr & CLONE_THREAD) | |
1450 | *flags_ptr |= CLONE_VM; | |
1451 | ||
1452 | /* | |
1453 | * If unsharing vm, must also unshare signal handlers. | |
1454 | */ | |
1455 | if (*flags_ptr & CLONE_VM) | |
1456 | *flags_ptr |= CLONE_SIGHAND; | |
1457 | ||
1458 | /* | |
1459 | * If unsharing signal handlers and the task was created | |
1460 | * using CLONE_THREAD, then must unshare the thread | |
1461 | */ | |
1462 | if ((*flags_ptr & CLONE_SIGHAND) && | |
1463 | (atomic_read(¤t->signal->count) > 1)) | |
1464 | *flags_ptr |= CLONE_THREAD; | |
1465 | ||
1466 | /* | |
1467 | * If unsharing namespace, must also unshare filesystem information. | |
1468 | */ | |
1469 | if (*flags_ptr & CLONE_NEWNS) | |
1470 | *flags_ptr |= CLONE_FS; | |
1471 | } | |
1472 | ||
1473 | /* | |
1474 | * Unsharing of tasks created with CLONE_THREAD is not supported yet | |
1475 | */ | |
1476 | static int unshare_thread(unsigned long unshare_flags) | |
1477 | { | |
1478 | if (unshare_flags & CLONE_THREAD) | |
1479 | return -EINVAL; | |
1480 | ||
1481 | return 0; | |
1482 | } | |
1483 | ||
1484 | /* | |
99d1419d | 1485 | * Unshare the filesystem structure if it is being shared |
cf2e340f JD |
1486 | */ |
1487 | static int unshare_fs(unsigned long unshare_flags, struct fs_struct **new_fsp) | |
1488 | { | |
1489 | struct fs_struct *fs = current->fs; | |
1490 | ||
1491 | if ((unshare_flags & CLONE_FS) && | |
99d1419d JD |
1492 | (fs && atomic_read(&fs->count) > 1)) { |
1493 | *new_fsp = __copy_fs_struct(current->fs); | |
1494 | if (!*new_fsp) | |
1495 | return -ENOMEM; | |
1496 | } | |
cf2e340f JD |
1497 | |
1498 | return 0; | |
1499 | } | |
1500 | ||
cf2e340f | 1501 | /* |
dae3c5a0 | 1502 | * Unsharing of sighand is not supported yet |
cf2e340f JD |
1503 | */ |
1504 | static int unshare_sighand(unsigned long unshare_flags, struct sighand_struct **new_sighp) | |
1505 | { | |
1506 | struct sighand_struct *sigh = current->sighand; | |
1507 | ||
dae3c5a0 | 1508 | if ((unshare_flags & CLONE_SIGHAND) && atomic_read(&sigh->count) > 1) |
cf2e340f JD |
1509 | return -EINVAL; |
1510 | else | |
1511 | return 0; | |
1512 | } | |
1513 | ||
1514 | /* | |
a0a7ec30 | 1515 | * Unshare vm if it is being shared |
cf2e340f JD |
1516 | */ |
1517 | static int unshare_vm(unsigned long unshare_flags, struct mm_struct **new_mmp) | |
1518 | { | |
1519 | struct mm_struct *mm = current->mm; | |
1520 | ||
1521 | if ((unshare_flags & CLONE_VM) && | |
a0a7ec30 | 1522 | (mm && atomic_read(&mm->mm_users) > 1)) { |
2d61b867 | 1523 | return -EINVAL; |
a0a7ec30 | 1524 | } |
cf2e340f JD |
1525 | |
1526 | return 0; | |
cf2e340f JD |
1527 | } |
1528 | ||
1529 | /* | |
a016f338 | 1530 | * Unshare file descriptor table if it is being shared |
cf2e340f JD |
1531 | */ |
1532 | static int unshare_fd(unsigned long unshare_flags, struct files_struct **new_fdp) | |
1533 | { | |
1534 | struct files_struct *fd = current->files; | |
a016f338 | 1535 | int error = 0; |
cf2e340f JD |
1536 | |
1537 | if ((unshare_flags & CLONE_FILES) && | |
a016f338 JD |
1538 | (fd && atomic_read(&fd->count) > 1)) { |
1539 | *new_fdp = dup_fd(fd, &error); | |
1540 | if (!*new_fdp) | |
1541 | return error; | |
1542 | } | |
cf2e340f JD |
1543 | |
1544 | return 0; | |
1545 | } | |
1546 | ||
cf2e340f JD |
1547 | /* |
1548 | * unshare allows a process to 'unshare' part of the process | |
1549 | * context which was originally shared using clone. copy_* | |
1550 | * functions used by do_fork() cannot be used here directly | |
1551 | * because they modify an inactive task_struct that is being | |
1552 | * constructed. Here we are modifying the current, active, | |
1553 | * task_struct. | |
1554 | */ | |
1555 | asmlinkage long sys_unshare(unsigned long unshare_flags) | |
1556 | { | |
1557 | int err = 0; | |
1558 | struct fs_struct *fs, *new_fs = NULL; | |
dae3c5a0 | 1559 | struct sighand_struct *new_sigh = NULL; |
cf2e340f JD |
1560 | struct mm_struct *mm, *new_mm = NULL, *active_mm = NULL; |
1561 | struct files_struct *fd, *new_fd = NULL; | |
cf7b708c | 1562 | struct nsproxy *new_nsproxy = NULL; |
9edff4ab | 1563 | int do_sysvsem = 0; |
cf2e340f JD |
1564 | |
1565 | check_unshare_flags(&unshare_flags); | |
1566 | ||
06f9d4f9 EB |
1567 | /* Return -EINVAL for all unsupported flags */ |
1568 | err = -EINVAL; | |
1569 | if (unshare_flags & ~(CLONE_THREAD|CLONE_FS|CLONE_NEWNS|CLONE_SIGHAND| | |
25b21cb2 | 1570 | CLONE_VM|CLONE_FILES|CLONE_SYSVSEM| |
9dd776b6 EB |
1571 | CLONE_NEWUTS|CLONE_NEWIPC|CLONE_NEWUSER| |
1572 | CLONE_NEWNET)) | |
06f9d4f9 EB |
1573 | goto bad_unshare_out; |
1574 | ||
6013f67f MS |
1575 | /* |
1576 | * CLONE_NEWIPC must also detach from the undolist: after switching | |
1577 | * to a new ipc namespace, the semaphore arrays from the old | |
1578 | * namespace are unreachable. | |
1579 | */ | |
1580 | if (unshare_flags & (CLONE_NEWIPC|CLONE_SYSVSEM)) | |
9edff4ab | 1581 | do_sysvsem = 1; |
cf2e340f JD |
1582 | if ((err = unshare_thread(unshare_flags))) |
1583 | goto bad_unshare_out; | |
1584 | if ((err = unshare_fs(unshare_flags, &new_fs))) | |
1585 | goto bad_unshare_cleanup_thread; | |
cf2e340f | 1586 | if ((err = unshare_sighand(unshare_flags, &new_sigh))) |
e3222c4e | 1587 | goto bad_unshare_cleanup_fs; |
cf2e340f JD |
1588 | if ((err = unshare_vm(unshare_flags, &new_mm))) |
1589 | goto bad_unshare_cleanup_sigh; | |
1590 | if ((err = unshare_fd(unshare_flags, &new_fd))) | |
1591 | goto bad_unshare_cleanup_vm; | |
e3222c4e BP |
1592 | if ((err = unshare_nsproxy_namespaces(unshare_flags, &new_nsproxy, |
1593 | new_fs))) | |
9edff4ab | 1594 | goto bad_unshare_cleanup_fd; |
c0b2fc31 | 1595 | |
9edff4ab MS |
1596 | if (new_fs || new_mm || new_fd || do_sysvsem || new_nsproxy) { |
1597 | if (do_sysvsem) { | |
1598 | /* | |
1599 | * CLONE_SYSVSEM is equivalent to sys_exit(). | |
1600 | */ | |
1601 | exit_sem(current); | |
1602 | } | |
ab516013 | 1603 | |
c0b2fc31 | 1604 | if (new_nsproxy) { |
cf7b708c PE |
1605 | switch_task_namespaces(current, new_nsproxy); |
1606 | new_nsproxy = NULL; | |
c0b2fc31 | 1607 | } |
cf2e340f | 1608 | |
cf7b708c PE |
1609 | task_lock(current); |
1610 | ||
cf2e340f JD |
1611 | if (new_fs) { |
1612 | fs = current->fs; | |
1613 | current->fs = new_fs; | |
1614 | new_fs = fs; | |
1615 | } | |
1616 | ||
cf2e340f JD |
1617 | if (new_mm) { |
1618 | mm = current->mm; | |
1619 | active_mm = current->active_mm; | |
1620 | current->mm = new_mm; | |
1621 | current->active_mm = new_mm; | |
1622 | activate_mm(active_mm, new_mm); | |
1623 | new_mm = mm; | |
1624 | } | |
1625 | ||
1626 | if (new_fd) { | |
1627 | fd = current->files; | |
1628 | current->files = new_fd; | |
1629 | new_fd = fd; | |
1630 | } | |
1631 | ||
1632 | task_unlock(current); | |
1633 | } | |
1634 | ||
c0b2fc31 | 1635 | if (new_nsproxy) |
444f378b | 1636 | put_nsproxy(new_nsproxy); |
c0b2fc31 | 1637 | |
cf2e340f JD |
1638 | bad_unshare_cleanup_fd: |
1639 | if (new_fd) | |
1640 | put_files_struct(new_fd); | |
1641 | ||
1642 | bad_unshare_cleanup_vm: | |
1643 | if (new_mm) | |
1644 | mmput(new_mm); | |
1645 | ||
1646 | bad_unshare_cleanup_sigh: | |
1647 | if (new_sigh) | |
1648 | if (atomic_dec_and_test(&new_sigh->count)) | |
1649 | kmem_cache_free(sighand_cachep, new_sigh); | |
1650 | ||
cf2e340f JD |
1651 | bad_unshare_cleanup_fs: |
1652 | if (new_fs) | |
1653 | put_fs_struct(new_fs); | |
1654 | ||
1655 | bad_unshare_cleanup_thread: | |
1656 | bad_unshare_out: | |
1657 | return err; | |
1658 | } | |
3b125388 AV |
1659 | |
1660 | /* | |
1661 | * Helper to unshare the files of the current task. | |
1662 | * We don't want to expose copy_files internals to | |
1663 | * the exec layer of the kernel. | |
1664 | */ | |
1665 | ||
1666 | int unshare_files(struct files_struct **displaced) | |
1667 | { | |
1668 | struct task_struct *task = current; | |
50704516 | 1669 | struct files_struct *copy = NULL; |
3b125388 AV |
1670 | int error; |
1671 | ||
1672 | error = unshare_fd(CLONE_FILES, ©); | |
1673 | if (error || !copy) { | |
1674 | *displaced = NULL; | |
1675 | return error; | |
1676 | } | |
1677 | *displaced = task->files; | |
1678 | task_lock(task); | |
1679 | task->files = copy; | |
1680 | task_unlock(task); | |
1681 | return 0; | |
1682 | } |