]>
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> | |
1da177e4 LT |
20 | #include <linux/personality.h> |
21 | #include <linux/mempolicy.h> | |
22 | #include <linux/sem.h> | |
23 | #include <linux/file.h> | |
9f3acc31 | 24 | #include <linux/fdtable.h> |
da9cbc87 | 25 | #include <linux/iocontext.h> |
1da177e4 LT |
26 | #include <linux/key.h> |
27 | #include <linux/binfmts.h> | |
28 | #include <linux/mman.h> | |
cddb8a5c | 29 | #include <linux/mmu_notifier.h> |
1da177e4 | 30 | #include <linux/fs.h> |
ab516013 | 31 | #include <linux/nsproxy.h> |
c59ede7b | 32 | #include <linux/capability.h> |
1da177e4 | 33 | #include <linux/cpu.h> |
b4f48b63 | 34 | #include <linux/cgroup.h> |
1da177e4 | 35 | #include <linux/security.h> |
a1e78772 | 36 | #include <linux/hugetlb.h> |
1da177e4 LT |
37 | #include <linux/swap.h> |
38 | #include <linux/syscalls.h> | |
39 | #include <linux/jiffies.h> | |
40 | #include <linux/futex.h> | |
8141c7f3 | 41 | #include <linux/compat.h> |
207205a2 | 42 | #include <linux/kthread.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> |
f201ae23 | 49 | #include <linux/ftrace.h> |
1da177e4 LT |
50 | #include <linux/profile.h> |
51 | #include <linux/rmap.h> | |
f8af4da3 | 52 | #include <linux/ksm.h> |
1da177e4 | 53 | #include <linux/acct.h> |
8f0ab514 | 54 | #include <linux/tsacct_kern.h> |
9f46080c | 55 | #include <linux/cn_proc.h> |
ba96a0c8 | 56 | #include <linux/freezer.h> |
ca74e92b | 57 | #include <linux/delayacct.h> |
ad4ecbcb | 58 | #include <linux/taskstats_kern.h> |
0a425405 | 59 | #include <linux/random.h> |
522ed776 | 60 | #include <linux/tty.h> |
fd0928df | 61 | #include <linux/blkdev.h> |
5ad4e53b | 62 | #include <linux/fs_struct.h> |
7c9f8861 | 63 | #include <linux/magic.h> |
cdd6c482 | 64 | #include <linux/perf_event.h> |
42c4ab41 | 65 | #include <linux/posix-timers.h> |
8e7cac79 | 66 | #include <linux/user-return-notifier.h> |
3d5992d2 | 67 | #include <linux/oom.h> |
ba76149f | 68 | #include <linux/khugepaged.h> |
d80e731e | 69 | #include <linux/signalfd.h> |
1da177e4 LT |
70 | |
71 | #include <asm/pgtable.h> | |
72 | #include <asm/pgalloc.h> | |
73 | #include <asm/uaccess.h> | |
74 | #include <asm/mmu_context.h> | |
75 | #include <asm/cacheflush.h> | |
76 | #include <asm/tlbflush.h> | |
77 | ||
ad8d75ff SR |
78 | #include <trace/events/sched.h> |
79 | ||
43d2b113 KH |
80 | #define CREATE_TRACE_POINTS |
81 | #include <trace/events/task.h> | |
82 | ||
1da177e4 LT |
83 | /* |
84 | * Protected counters by write_lock_irq(&tasklist_lock) | |
85 | */ | |
86 | unsigned long total_forks; /* Handle normal Linux uptimes. */ | |
fb0a685c | 87 | int nr_threads; /* The idle threads do not count.. */ |
1da177e4 LT |
88 | |
89 | int max_threads; /* tunable limit on nr_threads */ | |
90 | ||
91 | DEFINE_PER_CPU(unsigned long, process_counts) = 0; | |
92 | ||
c59923a1 | 93 | __cacheline_aligned DEFINE_RWLOCK(tasklist_lock); /* outer */ |
db1466b3 PM |
94 | |
95 | #ifdef CONFIG_PROVE_RCU | |
96 | int lockdep_tasklist_lock_is_held(void) | |
97 | { | |
98 | return lockdep_is_held(&tasklist_lock); | |
99 | } | |
100 | EXPORT_SYMBOL_GPL(lockdep_tasklist_lock_is_held); | |
101 | #endif /* #ifdef CONFIG_PROVE_RCU */ | |
1da177e4 LT |
102 | |
103 | int nr_processes(void) | |
104 | { | |
105 | int cpu; | |
106 | int total = 0; | |
107 | ||
1d510750 | 108 | for_each_possible_cpu(cpu) |
1da177e4 LT |
109 | total += per_cpu(process_counts, cpu); |
110 | ||
111 | return total; | |
112 | } | |
113 | ||
114 | #ifndef __HAVE_ARCH_TASK_STRUCT_ALLOCATOR | |
e18b890b | 115 | static struct kmem_cache *task_struct_cachep; |
41101809 TG |
116 | |
117 | static inline struct task_struct *alloc_task_struct_node(int node) | |
118 | { | |
119 | return kmem_cache_alloc_node(task_struct_cachep, GFP_KERNEL, node); | |
120 | } | |
121 | ||
122 | void __weak arch_release_task_struct(struct task_struct *tsk) { } | |
123 | ||
124 | static inline void free_task_struct(struct task_struct *tsk) | |
125 | { | |
126 | arch_release_task_struct(tsk); | |
127 | kmem_cache_free(task_struct_cachep, tsk); | |
128 | } | |
1da177e4 LT |
129 | #endif |
130 | ||
b69c49b7 | 131 | #ifndef __HAVE_ARCH_THREAD_INFO_ALLOCATOR |
41101809 TG |
132 | |
133 | void __weak arch_release_thread_info(struct thread_info *ti) { } | |
134 | ||
b6a84016 ED |
135 | static struct thread_info *alloc_thread_info_node(struct task_struct *tsk, |
136 | int node) | |
b69c49b7 | 137 | { |
2889f608 TG |
138 | struct page *page = alloc_pages_node(node, THREADINFO_GFP, |
139 | THREAD_SIZE_ORDER); | |
b6a84016 ED |
140 | |
141 | return page ? page_address(page) : NULL; | |
b69c49b7 FT |
142 | } |
143 | ||
144 | static inline void free_thread_info(struct thread_info *ti) | |
145 | { | |
41101809 | 146 | arch_release_thread_info(ti); |
b69c49b7 FT |
147 | free_pages((unsigned long)ti, THREAD_SIZE_ORDER); |
148 | } | |
149 | #endif | |
150 | ||
1da177e4 | 151 | /* SLAB cache for signal_struct structures (tsk->signal) */ |
e18b890b | 152 | static struct kmem_cache *signal_cachep; |
1da177e4 LT |
153 | |
154 | /* SLAB cache for sighand_struct structures (tsk->sighand) */ | |
e18b890b | 155 | struct kmem_cache *sighand_cachep; |
1da177e4 LT |
156 | |
157 | /* SLAB cache for files_struct structures (tsk->files) */ | |
e18b890b | 158 | struct kmem_cache *files_cachep; |
1da177e4 LT |
159 | |
160 | /* SLAB cache for fs_struct structures (tsk->fs) */ | |
e18b890b | 161 | struct kmem_cache *fs_cachep; |
1da177e4 LT |
162 | |
163 | /* SLAB cache for vm_area_struct structures */ | |
e18b890b | 164 | struct kmem_cache *vm_area_cachep; |
1da177e4 LT |
165 | |
166 | /* SLAB cache for mm_struct structures (tsk->mm) */ | |
e18b890b | 167 | static struct kmem_cache *mm_cachep; |
1da177e4 | 168 | |
c6a7f572 KM |
169 | static void account_kernel_stack(struct thread_info *ti, int account) |
170 | { | |
171 | struct zone *zone = page_zone(virt_to_page(ti)); | |
172 | ||
173 | mod_zone_page_state(zone, NR_KERNEL_STACK, account); | |
174 | } | |
175 | ||
1da177e4 LT |
176 | void free_task(struct task_struct *tsk) |
177 | { | |
c6a7f572 | 178 | account_kernel_stack(tsk->stack, -1); |
f7e4217b | 179 | free_thread_info(tsk->stack); |
23f78d4a | 180 | rt_mutex_debug_task_free(tsk); |
fb52607a | 181 | ftrace_graph_exit_task(tsk); |
1da177e4 LT |
182 | free_task_struct(tsk); |
183 | } | |
184 | EXPORT_SYMBOL(free_task); | |
185 | ||
ea6d290c ON |
186 | static inline void free_signal_struct(struct signal_struct *sig) |
187 | { | |
97101eb4 | 188 | taskstats_tgid_free(sig); |
1c5354de | 189 | sched_autogroup_exit(sig); |
ea6d290c ON |
190 | kmem_cache_free(signal_cachep, sig); |
191 | } | |
192 | ||
193 | static inline void put_signal_struct(struct signal_struct *sig) | |
194 | { | |
1c5354de | 195 | if (atomic_dec_and_test(&sig->sigcnt)) |
ea6d290c ON |
196 | free_signal_struct(sig); |
197 | } | |
198 | ||
158d9ebd | 199 | void __put_task_struct(struct task_struct *tsk) |
1da177e4 | 200 | { |
270f722d | 201 | WARN_ON(!tsk->exit_state); |
1da177e4 LT |
202 | WARN_ON(atomic_read(&tsk->usage)); |
203 | WARN_ON(tsk == current); | |
204 | ||
1a2a4d06 | 205 | security_task_free(tsk); |
e0e81739 | 206 | exit_creds(tsk); |
35df17c5 | 207 | delayacct_tsk_free(tsk); |
ea6d290c | 208 | put_signal_struct(tsk->signal); |
1da177e4 LT |
209 | |
210 | if (!profile_handoff_task(tsk)) | |
211 | free_task(tsk); | |
212 | } | |
77c100c8 | 213 | EXPORT_SYMBOL_GPL(__put_task_struct); |
1da177e4 | 214 | |
6c0a9fa6 | 215 | void __init __weak arch_task_cache_init(void) { } |
61c4628b | 216 | |
1da177e4 LT |
217 | void __init fork_init(unsigned long mempages) |
218 | { | |
219 | #ifndef __HAVE_ARCH_TASK_STRUCT_ALLOCATOR | |
220 | #ifndef ARCH_MIN_TASKALIGN | |
221 | #define ARCH_MIN_TASKALIGN L1_CACHE_BYTES | |
222 | #endif | |
223 | /* create a slab on which task_structs can be allocated */ | |
224 | task_struct_cachep = | |
225 | kmem_cache_create("task_struct", sizeof(struct task_struct), | |
2dff4405 | 226 | ARCH_MIN_TASKALIGN, SLAB_PANIC | SLAB_NOTRACK, NULL); |
1da177e4 LT |
227 | #endif |
228 | ||
61c4628b SS |
229 | /* do the arch specific task caches init */ |
230 | arch_task_cache_init(); | |
231 | ||
1da177e4 LT |
232 | /* |
233 | * The default maximum number of threads is set to a safe | |
234 | * value: the thread structures can take up at most half | |
235 | * of memory. | |
236 | */ | |
237 | max_threads = mempages / (8 * THREAD_SIZE / PAGE_SIZE); | |
238 | ||
239 | /* | |
240 | * we need to allow at least 20 threads to boot a system | |
241 | */ | |
fb0a685c | 242 | if (max_threads < 20) |
1da177e4 LT |
243 | max_threads = 20; |
244 | ||
245 | init_task.signal->rlim[RLIMIT_NPROC].rlim_cur = max_threads/2; | |
246 | init_task.signal->rlim[RLIMIT_NPROC].rlim_max = max_threads/2; | |
247 | init_task.signal->rlim[RLIMIT_SIGPENDING] = | |
248 | init_task.signal->rlim[RLIMIT_NPROC]; | |
249 | } | |
250 | ||
61c4628b SS |
251 | int __attribute__((weak)) arch_dup_task_struct(struct task_struct *dst, |
252 | struct task_struct *src) | |
253 | { | |
254 | *dst = *src; | |
255 | return 0; | |
256 | } | |
257 | ||
1da177e4 LT |
258 | static struct task_struct *dup_task_struct(struct task_struct *orig) |
259 | { | |
260 | struct task_struct *tsk; | |
261 | struct thread_info *ti; | |
7c9f8861 | 262 | unsigned long *stackend; |
207205a2 | 263 | int node = tsk_fork_get_node(orig); |
3e26c149 | 264 | int err; |
1da177e4 LT |
265 | |
266 | prepare_to_copy(orig); | |
267 | ||
504f52b5 | 268 | tsk = alloc_task_struct_node(node); |
1da177e4 LT |
269 | if (!tsk) |
270 | return NULL; | |
271 | ||
b6a84016 | 272 | ti = alloc_thread_info_node(tsk, node); |
1da177e4 LT |
273 | if (!ti) { |
274 | free_task_struct(tsk); | |
275 | return NULL; | |
276 | } | |
277 | ||
fb0a685c | 278 | err = arch_dup_task_struct(tsk, orig); |
61c4628b SS |
279 | if (err) |
280 | goto out; | |
281 | ||
f7e4217b | 282 | tsk->stack = ti; |
3e26c149 | 283 | |
10ebffde | 284 | setup_thread_stack(tsk, orig); |
8e7cac79 | 285 | clear_user_return_notifier(tsk); |
f26f9aff | 286 | clear_tsk_need_resched(tsk); |
7c9f8861 ES |
287 | stackend = end_of_stack(tsk); |
288 | *stackend = STACK_END_MAGIC; /* for overflow detection */ | |
1da177e4 | 289 | |
0a425405 AV |
290 | #ifdef CONFIG_CC_STACKPROTECTOR |
291 | tsk->stack_canary = get_random_int(); | |
292 | #endif | |
293 | ||
fb0a685c DRO |
294 | /* |
295 | * One for us, one for whoever does the "release_task()" (usually | |
296 | * parent) | |
297 | */ | |
298 | atomic_set(&tsk->usage, 2); | |
6c5c9341 | 299 | #ifdef CONFIG_BLK_DEV_IO_TRACE |
2056a782 | 300 | tsk->btrace_seq = 0; |
6c5c9341 | 301 | #endif |
a0aa7f68 | 302 | tsk->splice_pipe = NULL; |
c6a7f572 KM |
303 | |
304 | account_kernel_stack(ti, 1); | |
305 | ||
1da177e4 | 306 | return tsk; |
61c4628b SS |
307 | |
308 | out: | |
309 | free_thread_info(ti); | |
310 | free_task_struct(tsk); | |
311 | return NULL; | |
1da177e4 LT |
312 | } |
313 | ||
314 | #ifdef CONFIG_MMU | |
a39bc516 | 315 | static int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm) |
1da177e4 | 316 | { |
297c5eee | 317 | struct vm_area_struct *mpnt, *tmp, *prev, **pprev; |
1da177e4 LT |
318 | struct rb_node **rb_link, *rb_parent; |
319 | int retval; | |
320 | unsigned long charge; | |
321 | struct mempolicy *pol; | |
322 | ||
323 | down_write(&oldmm->mmap_sem); | |
ec8c0446 | 324 | flush_cache_dup_mm(oldmm); |
ad339451 IM |
325 | /* |
326 | * Not linked in yet - no deadlock potential: | |
327 | */ | |
328 | down_write_nested(&mm->mmap_sem, SINGLE_DEPTH_NESTING); | |
7ee78232 | 329 | |
1da177e4 LT |
330 | mm->locked_vm = 0; |
331 | mm->mmap = NULL; | |
332 | mm->mmap_cache = NULL; | |
333 | mm->free_area_cache = oldmm->mmap_base; | |
1363c3cd | 334 | mm->cached_hole_size = ~0UL; |
1da177e4 | 335 | mm->map_count = 0; |
94894244 | 336 | cpumask_clear(mm_cpumask(mm)); |
1da177e4 LT |
337 | mm->mm_rb = RB_ROOT; |
338 | rb_link = &mm->mm_rb.rb_node; | |
339 | rb_parent = NULL; | |
340 | pprev = &mm->mmap; | |
f8af4da3 | 341 | retval = ksm_fork(mm, oldmm); |
ba76149f AA |
342 | if (retval) |
343 | goto out; | |
344 | retval = khugepaged_fork(mm, oldmm); | |
f8af4da3 HD |
345 | if (retval) |
346 | goto out; | |
1da177e4 | 347 | |
297c5eee | 348 | prev = NULL; |
fd3e42fc | 349 | for (mpnt = oldmm->mmap; mpnt; mpnt = mpnt->vm_next) { |
1da177e4 LT |
350 | struct file *file; |
351 | ||
352 | if (mpnt->vm_flags & VM_DONTCOPY) { | |
3b6bfcdb HD |
353 | long pages = vma_pages(mpnt); |
354 | mm->total_vm -= pages; | |
ab50b8ed | 355 | vm_stat_account(mm, mpnt->vm_flags, mpnt->vm_file, |
3b6bfcdb | 356 | -pages); |
1da177e4 LT |
357 | continue; |
358 | } | |
359 | charge = 0; | |
360 | if (mpnt->vm_flags & VM_ACCOUNT) { | |
361 | unsigned int len = (mpnt->vm_end - mpnt->vm_start) >> PAGE_SHIFT; | |
191c5424 | 362 | if (security_vm_enough_memory_mm(oldmm, len)) /* sic */ |
1da177e4 LT |
363 | goto fail_nomem; |
364 | charge = len; | |
365 | } | |
e94b1766 | 366 | tmp = kmem_cache_alloc(vm_area_cachep, GFP_KERNEL); |
1da177e4 LT |
367 | if (!tmp) |
368 | goto fail_nomem; | |
369 | *tmp = *mpnt; | |
5beb4930 | 370 | INIT_LIST_HEAD(&tmp->anon_vma_chain); |
846a16bf | 371 | pol = mpol_dup(vma_policy(mpnt)); |
1da177e4 LT |
372 | retval = PTR_ERR(pol); |
373 | if (IS_ERR(pol)) | |
374 | goto fail_nomem_policy; | |
375 | vma_set_policy(tmp, pol); | |
a247c3a9 | 376 | tmp->vm_mm = mm; |
5beb4930 RR |
377 | if (anon_vma_fork(tmp, mpnt)) |
378 | goto fail_nomem_anon_vma_fork; | |
1da177e4 | 379 | tmp->vm_flags &= ~VM_LOCKED; |
297c5eee | 380 | tmp->vm_next = tmp->vm_prev = NULL; |
1da177e4 LT |
381 | file = tmp->vm_file; |
382 | if (file) { | |
f3a43f3f | 383 | struct inode *inode = file->f_path.dentry->d_inode; |
b88ed205 HD |
384 | struct address_space *mapping = file->f_mapping; |
385 | ||
1da177e4 LT |
386 | get_file(file); |
387 | if (tmp->vm_flags & VM_DENYWRITE) | |
388 | atomic_dec(&inode->i_writecount); | |
3d48ae45 | 389 | mutex_lock(&mapping->i_mmap_mutex); |
b88ed205 HD |
390 | if (tmp->vm_flags & VM_SHARED) |
391 | mapping->i_mmap_writable++; | |
b88ed205 HD |
392 | flush_dcache_mmap_lock(mapping); |
393 | /* insert tmp into the share list, just after mpnt */ | |
1da177e4 | 394 | vma_prio_tree_add(tmp, mpnt); |
b88ed205 | 395 | flush_dcache_mmap_unlock(mapping); |
3d48ae45 | 396 | mutex_unlock(&mapping->i_mmap_mutex); |
1da177e4 LT |
397 | } |
398 | ||
a1e78772 MG |
399 | /* |
400 | * Clear hugetlb-related page reserves for children. This only | |
401 | * affects MAP_PRIVATE mappings. Faults generated by the child | |
402 | * are not guaranteed to succeed, even if read-only | |
403 | */ | |
404 | if (is_vm_hugetlb_page(tmp)) | |
405 | reset_vma_resv_huge_pages(tmp); | |
406 | ||
1da177e4 | 407 | /* |
7ee78232 | 408 | * Link in the new vma and copy the page table entries. |
1da177e4 | 409 | */ |
1da177e4 LT |
410 | *pprev = tmp; |
411 | pprev = &tmp->vm_next; | |
297c5eee LT |
412 | tmp->vm_prev = prev; |
413 | prev = tmp; | |
1da177e4 LT |
414 | |
415 | __vma_link_rb(mm, tmp, rb_link, rb_parent); | |
416 | rb_link = &tmp->vm_rb.rb_right; | |
417 | rb_parent = &tmp->vm_rb; | |
418 | ||
419 | mm->map_count++; | |
0b0db14c | 420 | retval = copy_page_range(mm, oldmm, mpnt); |
1da177e4 LT |
421 | |
422 | if (tmp->vm_ops && tmp->vm_ops->open) | |
423 | tmp->vm_ops->open(tmp); | |
424 | ||
425 | if (retval) | |
426 | goto out; | |
427 | } | |
d6dd61c8 JF |
428 | /* a new mm has just been created */ |
429 | arch_dup_mmap(oldmm, mm); | |
1da177e4 | 430 | retval = 0; |
1da177e4 | 431 | out: |
7ee78232 | 432 | up_write(&mm->mmap_sem); |
fd3e42fc | 433 | flush_tlb_mm(oldmm); |
1da177e4 LT |
434 | up_write(&oldmm->mmap_sem); |
435 | return retval; | |
5beb4930 RR |
436 | fail_nomem_anon_vma_fork: |
437 | mpol_put(pol); | |
1da177e4 LT |
438 | fail_nomem_policy: |
439 | kmem_cache_free(vm_area_cachep, tmp); | |
440 | fail_nomem: | |
441 | retval = -ENOMEM; | |
442 | vm_unacct_memory(charge); | |
443 | goto out; | |
444 | } | |
445 | ||
fb0a685c | 446 | static inline int mm_alloc_pgd(struct mm_struct *mm) |
1da177e4 LT |
447 | { |
448 | mm->pgd = pgd_alloc(mm); | |
449 | if (unlikely(!mm->pgd)) | |
450 | return -ENOMEM; | |
451 | return 0; | |
452 | } | |
453 | ||
fb0a685c | 454 | static inline void mm_free_pgd(struct mm_struct *mm) |
1da177e4 | 455 | { |
5e541973 | 456 | pgd_free(mm, mm->pgd); |
1da177e4 LT |
457 | } |
458 | #else | |
459 | #define dup_mmap(mm, oldmm) (0) | |
460 | #define mm_alloc_pgd(mm) (0) | |
461 | #define mm_free_pgd(mm) | |
462 | #endif /* CONFIG_MMU */ | |
463 | ||
23ff4440 | 464 | __cacheline_aligned_in_smp DEFINE_SPINLOCK(mmlist_lock); |
1da177e4 | 465 | |
e94b1766 | 466 | #define allocate_mm() (kmem_cache_alloc(mm_cachep, GFP_KERNEL)) |
1da177e4 LT |
467 | #define free_mm(mm) (kmem_cache_free(mm_cachep, (mm))) |
468 | ||
4cb0e11b HK |
469 | static unsigned long default_dump_filter = MMF_DUMP_FILTER_DEFAULT; |
470 | ||
471 | static int __init coredump_filter_setup(char *s) | |
472 | { | |
473 | default_dump_filter = | |
474 | (simple_strtoul(s, NULL, 0) << MMF_DUMP_FILTER_SHIFT) & | |
475 | MMF_DUMP_FILTER_MASK; | |
476 | return 1; | |
477 | } | |
478 | ||
479 | __setup("coredump_filter=", coredump_filter_setup); | |
480 | ||
1da177e4 LT |
481 | #include <linux/init_task.h> |
482 | ||
858f0993 AD |
483 | static void mm_init_aio(struct mm_struct *mm) |
484 | { | |
485 | #ifdef CONFIG_AIO | |
486 | spin_lock_init(&mm->ioctx_lock); | |
487 | INIT_HLIST_HEAD(&mm->ioctx_list); | |
488 | #endif | |
489 | } | |
490 | ||
fb0a685c | 491 | static struct mm_struct *mm_init(struct mm_struct *mm, struct task_struct *p) |
1da177e4 LT |
492 | { |
493 | atomic_set(&mm->mm_users, 1); | |
494 | atomic_set(&mm->mm_count, 1); | |
495 | init_rwsem(&mm->mmap_sem); | |
496 | INIT_LIST_HEAD(&mm->mmlist); | |
f8af4da3 HD |
497 | mm->flags = (current->mm) ? |
498 | (current->mm->flags & MMF_INIT_MASK) : default_dump_filter; | |
999d9fc1 | 499 | mm->core_state = NULL; |
1da177e4 | 500 | mm->nr_ptes = 0; |
d559db08 | 501 | memset(&mm->rss_stat, 0, sizeof(mm->rss_stat)); |
1da177e4 | 502 | spin_lock_init(&mm->page_table_lock); |
1da177e4 | 503 | mm->free_area_cache = TASK_UNMAPPED_BASE; |
1363c3cd | 504 | mm->cached_hole_size = ~0UL; |
858f0993 | 505 | mm_init_aio(mm); |
cf475ad2 | 506 | mm_init_owner(mm, p); |
1da177e4 LT |
507 | |
508 | if (likely(!mm_alloc_pgd(mm))) { | |
509 | mm->def_flags = 0; | |
cddb8a5c | 510 | mmu_notifier_mm_init(mm); |
1da177e4 LT |
511 | return mm; |
512 | } | |
78fb7466 | 513 | |
1da177e4 LT |
514 | free_mm(mm); |
515 | return NULL; | |
516 | } | |
517 | ||
c3f0327f KK |
518 | static void check_mm(struct mm_struct *mm) |
519 | { | |
520 | int i; | |
521 | ||
522 | for (i = 0; i < NR_MM_COUNTERS; i++) { | |
523 | long x = atomic_long_read(&mm->rss_stat.count[i]); | |
524 | ||
525 | if (unlikely(x)) | |
526 | printk(KERN_ALERT "BUG: Bad rss-counter state " | |
527 | "mm:%p idx:%d val:%ld\n", mm, i, x); | |
528 | } | |
529 | ||
530 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE | |
531 | VM_BUG_ON(mm->pmd_huge_pte); | |
532 | #endif | |
533 | } | |
534 | ||
1da177e4 LT |
535 | /* |
536 | * Allocate and initialize an mm_struct. | |
537 | */ | |
fb0a685c | 538 | struct mm_struct *mm_alloc(void) |
1da177e4 | 539 | { |
fb0a685c | 540 | struct mm_struct *mm; |
1da177e4 LT |
541 | |
542 | mm = allocate_mm(); | |
de03c72c KM |
543 | if (!mm) |
544 | return NULL; | |
545 | ||
546 | memset(mm, 0, sizeof(*mm)); | |
6345d24d LT |
547 | mm_init_cpumask(mm); |
548 | return mm_init(mm, current); | |
1da177e4 LT |
549 | } |
550 | ||
551 | /* | |
552 | * Called when the last reference to the mm | |
553 | * is dropped: either by a lazy thread or by | |
554 | * mmput. Free the page directory and the mm. | |
555 | */ | |
7ad5b3a5 | 556 | void __mmdrop(struct mm_struct *mm) |
1da177e4 LT |
557 | { |
558 | BUG_ON(mm == &init_mm); | |
559 | mm_free_pgd(mm); | |
560 | destroy_context(mm); | |
cddb8a5c | 561 | mmu_notifier_mm_destroy(mm); |
c3f0327f | 562 | check_mm(mm); |
1da177e4 LT |
563 | free_mm(mm); |
564 | } | |
6d4e4c4f | 565 | EXPORT_SYMBOL_GPL(__mmdrop); |
1da177e4 LT |
566 | |
567 | /* | |
568 | * Decrement the use count and release all resources for an mm. | |
569 | */ | |
570 | void mmput(struct mm_struct *mm) | |
571 | { | |
0ae26f1b AM |
572 | might_sleep(); |
573 | ||
1da177e4 LT |
574 | if (atomic_dec_and_test(&mm->mm_users)) { |
575 | exit_aio(mm); | |
1c2fb7a4 | 576 | ksm_exit(mm); |
ba76149f | 577 | khugepaged_exit(mm); /* must run before exit_mmap */ |
1da177e4 | 578 | exit_mmap(mm); |
925d1c40 | 579 | set_mm_exe_file(mm, NULL); |
1da177e4 LT |
580 | if (!list_empty(&mm->mmlist)) { |
581 | spin_lock(&mmlist_lock); | |
582 | list_del(&mm->mmlist); | |
583 | spin_unlock(&mmlist_lock); | |
584 | } | |
585 | put_swap_token(mm); | |
801460d0 HS |
586 | if (mm->binfmt) |
587 | module_put(mm->binfmt->module); | |
1da177e4 LT |
588 | mmdrop(mm); |
589 | } | |
590 | } | |
591 | EXPORT_SYMBOL_GPL(mmput); | |
592 | ||
38646013 JS |
593 | /* |
594 | * We added or removed a vma mapping the executable. The vmas are only mapped | |
595 | * during exec and are not mapped with the mmap system call. | |
596 | * Callers must hold down_write() on the mm's mmap_sem for these | |
597 | */ | |
598 | void added_exe_file_vma(struct mm_struct *mm) | |
599 | { | |
600 | mm->num_exe_file_vmas++; | |
601 | } | |
602 | ||
603 | void removed_exe_file_vma(struct mm_struct *mm) | |
604 | { | |
605 | mm->num_exe_file_vmas--; | |
fb0a685c | 606 | if ((mm->num_exe_file_vmas == 0) && mm->exe_file) { |
38646013 JS |
607 | fput(mm->exe_file); |
608 | mm->exe_file = NULL; | |
609 | } | |
610 | ||
611 | } | |
612 | ||
613 | void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file) | |
614 | { | |
615 | if (new_exe_file) | |
616 | get_file(new_exe_file); | |
617 | if (mm->exe_file) | |
618 | fput(mm->exe_file); | |
619 | mm->exe_file = new_exe_file; | |
620 | mm->num_exe_file_vmas = 0; | |
621 | } | |
622 | ||
623 | struct file *get_mm_exe_file(struct mm_struct *mm) | |
624 | { | |
625 | struct file *exe_file; | |
626 | ||
627 | /* We need mmap_sem to protect against races with removal of | |
628 | * VM_EXECUTABLE vmas */ | |
629 | down_read(&mm->mmap_sem); | |
630 | exe_file = mm->exe_file; | |
631 | if (exe_file) | |
632 | get_file(exe_file); | |
633 | up_read(&mm->mmap_sem); | |
634 | return exe_file; | |
635 | } | |
636 | ||
637 | static void dup_mm_exe_file(struct mm_struct *oldmm, struct mm_struct *newmm) | |
638 | { | |
639 | /* It's safe to write the exe_file pointer without exe_file_lock because | |
640 | * this is called during fork when the task is not yet in /proc */ | |
641 | newmm->exe_file = get_mm_exe_file(oldmm); | |
642 | } | |
643 | ||
1da177e4 LT |
644 | /** |
645 | * get_task_mm - acquire a reference to the task's mm | |
646 | * | |
246bb0b1 | 647 | * Returns %NULL if the task has no mm. Checks PF_KTHREAD (meaning |
1da177e4 LT |
648 | * this kernel workthread has transiently adopted a user mm with use_mm, |
649 | * to do its AIO) is not set and if so returns a reference to it, after | |
650 | * bumping up the use count. User must release the mm via mmput() | |
651 | * after use. Typically used by /proc and ptrace. | |
652 | */ | |
653 | struct mm_struct *get_task_mm(struct task_struct *task) | |
654 | { | |
655 | struct mm_struct *mm; | |
656 | ||
657 | task_lock(task); | |
658 | mm = task->mm; | |
659 | if (mm) { | |
246bb0b1 | 660 | if (task->flags & PF_KTHREAD) |
1da177e4 LT |
661 | mm = NULL; |
662 | else | |
663 | atomic_inc(&mm->mm_users); | |
664 | } | |
665 | task_unlock(task); | |
666 | return mm; | |
667 | } | |
668 | EXPORT_SYMBOL_GPL(get_task_mm); | |
669 | ||
8cdb878d CY |
670 | struct mm_struct *mm_access(struct task_struct *task, unsigned int mode) |
671 | { | |
672 | struct mm_struct *mm; | |
673 | int err; | |
674 | ||
675 | err = mutex_lock_killable(&task->signal->cred_guard_mutex); | |
676 | if (err) | |
677 | return ERR_PTR(err); | |
678 | ||
679 | mm = get_task_mm(task); | |
680 | if (mm && mm != current->mm && | |
681 | !ptrace_may_access(task, mode)) { | |
682 | mmput(mm); | |
683 | mm = ERR_PTR(-EACCES); | |
684 | } | |
685 | mutex_unlock(&task->signal->cred_guard_mutex); | |
686 | ||
687 | return mm; | |
688 | } | |
689 | ||
57b59c4a | 690 | static void complete_vfork_done(struct task_struct *tsk) |
c415c3b4 | 691 | { |
d68b46fe | 692 | struct completion *vfork; |
c415c3b4 | 693 | |
d68b46fe ON |
694 | task_lock(tsk); |
695 | vfork = tsk->vfork_done; | |
696 | if (likely(vfork)) { | |
697 | tsk->vfork_done = NULL; | |
698 | complete(vfork); | |
699 | } | |
700 | task_unlock(tsk); | |
701 | } | |
702 | ||
703 | static int wait_for_vfork_done(struct task_struct *child, | |
704 | struct completion *vfork) | |
705 | { | |
706 | int killed; | |
707 | ||
708 | freezer_do_not_count(); | |
709 | killed = wait_for_completion_killable(vfork); | |
710 | freezer_count(); | |
711 | ||
712 | if (killed) { | |
713 | task_lock(child); | |
714 | child->vfork_done = NULL; | |
715 | task_unlock(child); | |
716 | } | |
717 | ||
718 | put_task_struct(child); | |
719 | return killed; | |
c415c3b4 ON |
720 | } |
721 | ||
1da177e4 LT |
722 | /* Please note the differences between mmput and mm_release. |
723 | * mmput is called whenever we stop holding onto a mm_struct, | |
724 | * error success whatever. | |
725 | * | |
726 | * mm_release is called after a mm_struct has been removed | |
727 | * from the current process. | |
728 | * | |
729 | * This difference is important for error handling, when we | |
730 | * only half set up a mm_struct for a new process and need to restore | |
731 | * the old one. Because we mmput the new mm_struct before | |
732 | * restoring the old one. . . | |
733 | * Eric Biederman 10 January 1998 | |
734 | */ | |
735 | void mm_release(struct task_struct *tsk, struct mm_struct *mm) | |
736 | { | |
8141c7f3 LT |
737 | /* Get rid of any futexes when releasing the mm */ |
738 | #ifdef CONFIG_FUTEX | |
fc6b177d | 739 | if (unlikely(tsk->robust_list)) { |
8141c7f3 | 740 | exit_robust_list(tsk); |
fc6b177d PZ |
741 | tsk->robust_list = NULL; |
742 | } | |
8141c7f3 | 743 | #ifdef CONFIG_COMPAT |
fc6b177d | 744 | if (unlikely(tsk->compat_robust_list)) { |
8141c7f3 | 745 | compat_exit_robust_list(tsk); |
fc6b177d PZ |
746 | tsk->compat_robust_list = NULL; |
747 | } | |
8141c7f3 | 748 | #endif |
322a2c10 TG |
749 | if (unlikely(!list_empty(&tsk->pi_state_list))) |
750 | exit_pi_state_list(tsk); | |
8141c7f3 LT |
751 | #endif |
752 | ||
1da177e4 LT |
753 | /* Get rid of any cached register state */ |
754 | deactivate_mm(tsk, mm); | |
755 | ||
c415c3b4 ON |
756 | if (tsk->vfork_done) |
757 | complete_vfork_done(tsk); | |
fec1d011 RM |
758 | |
759 | /* | |
760 | * If we're exiting normally, clear a user-space tid field if | |
761 | * requested. We leave this alone when dying by signal, to leave | |
762 | * the value intact in a core dump, and to save the unnecessary | |
d68b46fe ON |
763 | * trouble, say, a killed vfork parent shouldn't touch this mm. |
764 | * Userland only wants this done for a sys_exit. | |
fec1d011 | 765 | */ |
9c8a8228 ED |
766 | if (tsk->clear_child_tid) { |
767 | if (!(tsk->flags & PF_SIGNALED) && | |
768 | atomic_read(&mm->mm_users) > 1) { | |
769 | /* | |
770 | * We don't check the error code - if userspace has | |
771 | * not set up a proper pointer then tough luck. | |
772 | */ | |
773 | put_user(0, tsk->clear_child_tid); | |
774 | sys_futex(tsk->clear_child_tid, FUTEX_WAKE, | |
775 | 1, NULL, NULL, 0); | |
776 | } | |
1da177e4 | 777 | tsk->clear_child_tid = NULL; |
1da177e4 LT |
778 | } |
779 | } | |
780 | ||
a0a7ec30 JD |
781 | /* |
782 | * Allocate a new mm structure and copy contents from the | |
783 | * mm structure of the passed in task structure. | |
784 | */ | |
402b0862 | 785 | struct mm_struct *dup_mm(struct task_struct *tsk) |
a0a7ec30 JD |
786 | { |
787 | struct mm_struct *mm, *oldmm = current->mm; | |
788 | int err; | |
789 | ||
790 | if (!oldmm) | |
791 | return NULL; | |
792 | ||
793 | mm = allocate_mm(); | |
794 | if (!mm) | |
795 | goto fail_nomem; | |
796 | ||
797 | memcpy(mm, oldmm, sizeof(*mm)); | |
6345d24d | 798 | mm_init_cpumask(mm); |
a0a7ec30 | 799 | |
7602bdf2 AC |
800 | /* Initializing for Swap token stuff */ |
801 | mm->token_priority = 0; | |
802 | mm->last_interval = 0; | |
803 | ||
e7a00c45 AA |
804 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE |
805 | mm->pmd_huge_pte = NULL; | |
806 | #endif | |
807 | ||
78fb7466 | 808 | if (!mm_init(mm, tsk)) |
a0a7ec30 JD |
809 | goto fail_nomem; |
810 | ||
811 | if (init_new_context(tsk, mm)) | |
812 | goto fail_nocontext; | |
813 | ||
925d1c40 MH |
814 | dup_mm_exe_file(oldmm, mm); |
815 | ||
a0a7ec30 JD |
816 | err = dup_mmap(mm, oldmm); |
817 | if (err) | |
818 | goto free_pt; | |
819 | ||
820 | mm->hiwater_rss = get_mm_rss(mm); | |
821 | mm->hiwater_vm = mm->total_vm; | |
822 | ||
801460d0 HS |
823 | if (mm->binfmt && !try_module_get(mm->binfmt->module)) |
824 | goto free_pt; | |
825 | ||
a0a7ec30 JD |
826 | return mm; |
827 | ||
828 | free_pt: | |
801460d0 HS |
829 | /* don't put binfmt in mmput, we haven't got module yet */ |
830 | mm->binfmt = NULL; | |
a0a7ec30 JD |
831 | mmput(mm); |
832 | ||
833 | fail_nomem: | |
834 | return NULL; | |
835 | ||
836 | fail_nocontext: | |
837 | /* | |
838 | * If init_new_context() failed, we cannot use mmput() to free the mm | |
839 | * because it calls destroy_context() | |
840 | */ | |
841 | mm_free_pgd(mm); | |
842 | free_mm(mm); | |
843 | return NULL; | |
844 | } | |
845 | ||
fb0a685c | 846 | static int copy_mm(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 | 847 | { |
fb0a685c | 848 | struct mm_struct *mm, *oldmm; |
1da177e4 LT |
849 | int retval; |
850 | ||
851 | tsk->min_flt = tsk->maj_flt = 0; | |
852 | tsk->nvcsw = tsk->nivcsw = 0; | |
17406b82 MSB |
853 | #ifdef CONFIG_DETECT_HUNG_TASK |
854 | tsk->last_switch_count = tsk->nvcsw + tsk->nivcsw; | |
855 | #endif | |
1da177e4 LT |
856 | |
857 | tsk->mm = NULL; | |
858 | tsk->active_mm = NULL; | |
859 | ||
860 | /* | |
861 | * Are we cloning a kernel thread? | |
862 | * | |
863 | * We need to steal a active VM for that.. | |
864 | */ | |
865 | oldmm = current->mm; | |
866 | if (!oldmm) | |
867 | return 0; | |
868 | ||
869 | if (clone_flags & CLONE_VM) { | |
870 | atomic_inc(&oldmm->mm_users); | |
871 | mm = oldmm; | |
1da177e4 LT |
872 | goto good_mm; |
873 | } | |
874 | ||
875 | retval = -ENOMEM; | |
a0a7ec30 | 876 | mm = dup_mm(tsk); |
1da177e4 LT |
877 | if (!mm) |
878 | goto fail_nomem; | |
879 | ||
1da177e4 | 880 | good_mm: |
7602bdf2 AC |
881 | /* Initializing for Swap token stuff */ |
882 | mm->token_priority = 0; | |
883 | mm->last_interval = 0; | |
884 | ||
1da177e4 LT |
885 | tsk->mm = mm; |
886 | tsk->active_mm = mm; | |
887 | return 0; | |
888 | ||
1da177e4 LT |
889 | fail_nomem: |
890 | return retval; | |
1da177e4 LT |
891 | } |
892 | ||
a39bc516 | 893 | static int copy_fs(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 | 894 | { |
498052bb | 895 | struct fs_struct *fs = current->fs; |
1da177e4 | 896 | if (clone_flags & CLONE_FS) { |
498052bb | 897 | /* tsk->fs is already what we want */ |
2a4419b5 | 898 | spin_lock(&fs->lock); |
498052bb | 899 | if (fs->in_exec) { |
2a4419b5 | 900 | spin_unlock(&fs->lock); |
498052bb AV |
901 | return -EAGAIN; |
902 | } | |
903 | fs->users++; | |
2a4419b5 | 904 | spin_unlock(&fs->lock); |
1da177e4 LT |
905 | return 0; |
906 | } | |
498052bb | 907 | tsk->fs = copy_fs_struct(fs); |
1da177e4 LT |
908 | if (!tsk->fs) |
909 | return -ENOMEM; | |
910 | return 0; | |
911 | } | |
912 | ||
fb0a685c | 913 | static int copy_files(unsigned long clone_flags, struct task_struct *tsk) |
a016f338 JD |
914 | { |
915 | struct files_struct *oldf, *newf; | |
916 | int error = 0; | |
917 | ||
918 | /* | |
919 | * A background process may not have any files ... | |
920 | */ | |
921 | oldf = current->files; | |
922 | if (!oldf) | |
923 | goto out; | |
924 | ||
925 | if (clone_flags & CLONE_FILES) { | |
926 | atomic_inc(&oldf->count); | |
927 | goto out; | |
928 | } | |
929 | ||
a016f338 JD |
930 | newf = dup_fd(oldf, &error); |
931 | if (!newf) | |
932 | goto out; | |
933 | ||
934 | tsk->files = newf; | |
935 | error = 0; | |
936 | out: | |
937 | return error; | |
938 | } | |
939 | ||
fadad878 | 940 | static int copy_io(unsigned long clone_flags, struct task_struct *tsk) |
fd0928df JA |
941 | { |
942 | #ifdef CONFIG_BLOCK | |
943 | struct io_context *ioc = current->io_context; | |
6e736be7 | 944 | struct io_context *new_ioc; |
fd0928df JA |
945 | |
946 | if (!ioc) | |
947 | return 0; | |
fadad878 JA |
948 | /* |
949 | * Share io context with parent, if CLONE_IO is set | |
950 | */ | |
951 | if (clone_flags & CLONE_IO) { | |
952 | tsk->io_context = ioc_task_link(ioc); | |
953 | if (unlikely(!tsk->io_context)) | |
954 | return -ENOMEM; | |
955 | } else if (ioprio_valid(ioc->ioprio)) { | |
6e736be7 TH |
956 | new_ioc = get_task_io_context(tsk, GFP_KERNEL, NUMA_NO_NODE); |
957 | if (unlikely(!new_ioc)) | |
fd0928df JA |
958 | return -ENOMEM; |
959 | ||
6e736be7 | 960 | new_ioc->ioprio = ioc->ioprio; |
11a3122f | 961 | put_io_context(new_ioc); |
fd0928df JA |
962 | } |
963 | #endif | |
964 | return 0; | |
965 | } | |
966 | ||
a39bc516 | 967 | static int copy_sighand(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 LT |
968 | { |
969 | struct sighand_struct *sig; | |
970 | ||
60348802 | 971 | if (clone_flags & CLONE_SIGHAND) { |
1da177e4 LT |
972 | atomic_inc(¤t->sighand->count); |
973 | return 0; | |
974 | } | |
975 | sig = kmem_cache_alloc(sighand_cachep, GFP_KERNEL); | |
e56d0903 | 976 | rcu_assign_pointer(tsk->sighand, sig); |
1da177e4 LT |
977 | if (!sig) |
978 | return -ENOMEM; | |
1da177e4 LT |
979 | atomic_set(&sig->count, 1); |
980 | memcpy(sig->action, current->sighand->action, sizeof(sig->action)); | |
981 | return 0; | |
982 | } | |
983 | ||
a7e5328a | 984 | void __cleanup_sighand(struct sighand_struct *sighand) |
c81addc9 | 985 | { |
d80e731e ON |
986 | if (atomic_dec_and_test(&sighand->count)) { |
987 | signalfd_cleanup(sighand); | |
c81addc9 | 988 | kmem_cache_free(sighand_cachep, sighand); |
d80e731e | 989 | } |
c81addc9 ON |
990 | } |
991 | ||
f06febc9 FM |
992 | |
993 | /* | |
994 | * Initialize POSIX timer handling for a thread group. | |
995 | */ | |
996 | static void posix_cpu_timers_init_group(struct signal_struct *sig) | |
997 | { | |
78d7d407 JS |
998 | unsigned long cpu_limit; |
999 | ||
f06febc9 FM |
1000 | /* Thread group counters. */ |
1001 | thread_group_cputime_init(sig); | |
1002 | ||
78d7d407 JS |
1003 | cpu_limit = ACCESS_ONCE(sig->rlim[RLIMIT_CPU].rlim_cur); |
1004 | if (cpu_limit != RLIM_INFINITY) { | |
1005 | sig->cputime_expires.prof_exp = secs_to_cputime(cpu_limit); | |
6279a751 ON |
1006 | sig->cputimer.running = 1; |
1007 | } | |
1008 | ||
f06febc9 FM |
1009 | /* The timer lists. */ |
1010 | INIT_LIST_HEAD(&sig->cpu_timers[0]); | |
1011 | INIT_LIST_HEAD(&sig->cpu_timers[1]); | |
1012 | INIT_LIST_HEAD(&sig->cpu_timers[2]); | |
1013 | } | |
1014 | ||
a39bc516 | 1015 | static int copy_signal(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 LT |
1016 | { |
1017 | struct signal_struct *sig; | |
1da177e4 | 1018 | |
4ab6c083 | 1019 | if (clone_flags & CLONE_THREAD) |
490dea45 | 1020 | return 0; |
490dea45 | 1021 | |
a56704ef | 1022 | sig = kmem_cache_zalloc(signal_cachep, GFP_KERNEL); |
1da177e4 LT |
1023 | tsk->signal = sig; |
1024 | if (!sig) | |
1025 | return -ENOMEM; | |
1026 | ||
b3ac022c | 1027 | sig->nr_threads = 1; |
1da177e4 | 1028 | atomic_set(&sig->live, 1); |
b3ac022c | 1029 | atomic_set(&sig->sigcnt, 1); |
1da177e4 | 1030 | init_waitqueue_head(&sig->wait_chldexit); |
b3bfa0cb SB |
1031 | if (clone_flags & CLONE_NEWPID) |
1032 | sig->flags |= SIGNAL_UNKILLABLE; | |
db51aecc | 1033 | sig->curr_target = tsk; |
1da177e4 LT |
1034 | init_sigpending(&sig->shared_pending); |
1035 | INIT_LIST_HEAD(&sig->posix_timers); | |
1036 | ||
c9cb2e3d | 1037 | hrtimer_init(&sig->real_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL); |
1da177e4 | 1038 | sig->real_timer.function = it_real_fn; |
1da177e4 | 1039 | |
1da177e4 LT |
1040 | task_lock(current->group_leader); |
1041 | memcpy(sig->rlim, current->signal->rlim, sizeof sig->rlim); | |
1042 | task_unlock(current->group_leader); | |
1043 | ||
6279a751 ON |
1044 | posix_cpu_timers_init_group(sig); |
1045 | ||
522ed776 | 1046 | tty_audit_fork(sig); |
5091faa4 | 1047 | sched_autogroup_fork(sig); |
522ed776 | 1048 | |
4714d1d3 | 1049 | #ifdef CONFIG_CGROUPS |
257058ae | 1050 | init_rwsem(&sig->group_rwsem); |
4714d1d3 BB |
1051 | #endif |
1052 | ||
28b83c51 | 1053 | sig->oom_adj = current->signal->oom_adj; |
a63d83f4 | 1054 | sig->oom_score_adj = current->signal->oom_score_adj; |
dabb16f6 | 1055 | sig->oom_score_adj_min = current->signal->oom_score_adj_min; |
28b83c51 | 1056 | |
ebec18a6 LP |
1057 | sig->has_child_subreaper = current->signal->has_child_subreaper || |
1058 | current->signal->is_child_subreaper; | |
1059 | ||
9b1bf12d KM |
1060 | mutex_init(&sig->cred_guard_mutex); |
1061 | ||
1da177e4 LT |
1062 | return 0; |
1063 | } | |
1064 | ||
a39bc516 | 1065 | static void copy_flags(unsigned long clone_flags, struct task_struct *p) |
1da177e4 LT |
1066 | { |
1067 | unsigned long new_flags = p->flags; | |
1068 | ||
21aa9af0 | 1069 | new_flags &= ~(PF_SUPERPRIV | PF_WQ_WORKER); |
1da177e4 | 1070 | new_flags |= PF_FORKNOEXEC; |
1da177e4 LT |
1071 | p->flags = new_flags; |
1072 | } | |
1073 | ||
17da2bd9 | 1074 | SYSCALL_DEFINE1(set_tid_address, int __user *, tidptr) |
1da177e4 LT |
1075 | { |
1076 | current->clear_child_tid = tidptr; | |
1077 | ||
b488893a | 1078 | return task_pid_vnr(current); |
1da177e4 LT |
1079 | } |
1080 | ||
a39bc516 | 1081 | static void rt_mutex_init_task(struct task_struct *p) |
23f78d4a | 1082 | { |
1d615482 | 1083 | raw_spin_lock_init(&p->pi_lock); |
e29e175b | 1084 | #ifdef CONFIG_RT_MUTEXES |
732375c6 | 1085 | plist_head_init(&p->pi_waiters); |
23f78d4a | 1086 | p->pi_blocked_on = NULL; |
23f78d4a IM |
1087 | #endif |
1088 | } | |
1089 | ||
cf475ad2 BS |
1090 | #ifdef CONFIG_MM_OWNER |
1091 | void mm_init_owner(struct mm_struct *mm, struct task_struct *p) | |
1092 | { | |
1093 | mm->owner = p; | |
1094 | } | |
1095 | #endif /* CONFIG_MM_OWNER */ | |
1096 | ||
f06febc9 FM |
1097 | /* |
1098 | * Initialize POSIX timer handling for a single task. | |
1099 | */ | |
1100 | static void posix_cpu_timers_init(struct task_struct *tsk) | |
1101 | { | |
64861634 MS |
1102 | tsk->cputime_expires.prof_exp = 0; |
1103 | tsk->cputime_expires.virt_exp = 0; | |
f06febc9 FM |
1104 | tsk->cputime_expires.sched_exp = 0; |
1105 | INIT_LIST_HEAD(&tsk->cpu_timers[0]); | |
1106 | INIT_LIST_HEAD(&tsk->cpu_timers[1]); | |
1107 | INIT_LIST_HEAD(&tsk->cpu_timers[2]); | |
1108 | } | |
1109 | ||
1da177e4 LT |
1110 | /* |
1111 | * This creates a new process as a copy of the old one, | |
1112 | * but does not actually start it yet. | |
1113 | * | |
1114 | * It copies the registers, and all the appropriate | |
1115 | * parts of the process environment (as per the clone | |
1116 | * flags). The actual kick-off is left to the caller. | |
1117 | */ | |
36c8b586 IM |
1118 | static struct task_struct *copy_process(unsigned long clone_flags, |
1119 | unsigned long stack_start, | |
1120 | struct pt_regs *regs, | |
1121 | unsigned long stack_size, | |
36c8b586 | 1122 | int __user *child_tidptr, |
09a05394 RM |
1123 | struct pid *pid, |
1124 | int trace) | |
1da177e4 LT |
1125 | { |
1126 | int retval; | |
a24efe62 | 1127 | struct task_struct *p; |
b4f48b63 | 1128 | int cgroup_callbacks_done = 0; |
1da177e4 LT |
1129 | |
1130 | if ((clone_flags & (CLONE_NEWNS|CLONE_FS)) == (CLONE_NEWNS|CLONE_FS)) | |
1131 | return ERR_PTR(-EINVAL); | |
1132 | ||
1133 | /* | |
1134 | * Thread groups must share signals as well, and detached threads | |
1135 | * can only be started up within the thread group. | |
1136 | */ | |
1137 | if ((clone_flags & CLONE_THREAD) && !(clone_flags & CLONE_SIGHAND)) | |
1138 | return ERR_PTR(-EINVAL); | |
1139 | ||
1140 | /* | |
1141 | * Shared signal handlers imply shared VM. By way of the above, | |
1142 | * thread groups also imply shared VM. Blocking this case allows | |
1143 | * for various simplifications in other code. | |
1144 | */ | |
1145 | if ((clone_flags & CLONE_SIGHAND) && !(clone_flags & CLONE_VM)) | |
1146 | return ERR_PTR(-EINVAL); | |
1147 | ||
123be07b SB |
1148 | /* |
1149 | * Siblings of global init remain as zombies on exit since they are | |
1150 | * not reaped by their parent (swapper). To solve this and to avoid | |
1151 | * multi-rooted process trees, prevent global and container-inits | |
1152 | * from creating siblings. | |
1153 | */ | |
1154 | if ((clone_flags & CLONE_PARENT) && | |
1155 | current->signal->flags & SIGNAL_UNKILLABLE) | |
1156 | return ERR_PTR(-EINVAL); | |
1157 | ||
1da177e4 LT |
1158 | retval = security_task_create(clone_flags); |
1159 | if (retval) | |
1160 | goto fork_out; | |
1161 | ||
1162 | retval = -ENOMEM; | |
1163 | p = dup_task_struct(current); | |
1164 | if (!p) | |
1165 | goto fork_out; | |
1166 | ||
f7e8b616 SR |
1167 | ftrace_graph_init_task(p); |
1168 | ||
bea493a0 PZ |
1169 | rt_mutex_init_task(p); |
1170 | ||
d12c1a37 | 1171 | #ifdef CONFIG_PROVE_LOCKING |
de30a2b3 IM |
1172 | DEBUG_LOCKS_WARN_ON(!p->hardirqs_enabled); |
1173 | DEBUG_LOCKS_WARN_ON(!p->softirqs_enabled); | |
1174 | #endif | |
1da177e4 | 1175 | retval = -EAGAIN; |
3b11a1de | 1176 | if (atomic_read(&p->real_cred->user->processes) >= |
78d7d407 | 1177 | task_rlimit(p, RLIMIT_NPROC)) { |
1da177e4 | 1178 | if (!capable(CAP_SYS_ADMIN) && !capable(CAP_SYS_RESOURCE) && |
18b6e041 | 1179 | p->real_cred->user != INIT_USER) |
1da177e4 LT |
1180 | goto bad_fork_free; |
1181 | } | |
72fa5997 | 1182 | current->flags &= ~PF_NPROC_EXCEEDED; |
1da177e4 | 1183 | |
f1752eec DH |
1184 | retval = copy_creds(p, clone_flags); |
1185 | if (retval < 0) | |
1186 | goto bad_fork_free; | |
1da177e4 LT |
1187 | |
1188 | /* | |
1189 | * If multiple threads are within copy_process(), then this check | |
1190 | * triggers too late. This doesn't hurt, the check is only there | |
1191 | * to stop root fork bombs. | |
1192 | */ | |
04ec93fe | 1193 | retval = -EAGAIN; |
1da177e4 LT |
1194 | if (nr_threads >= max_threads) |
1195 | goto bad_fork_cleanup_count; | |
1196 | ||
a1261f54 | 1197 | if (!try_module_get(task_thread_info(p)->exec_domain->module)) |
1da177e4 LT |
1198 | goto bad_fork_cleanup_count; |
1199 | ||
1da177e4 | 1200 | p->did_exec = 0; |
ca74e92b | 1201 | delayacct_tsk_init(p); /* Must remain after dup_task_struct() */ |
1da177e4 | 1202 | copy_flags(clone_flags, p); |
1da177e4 LT |
1203 | INIT_LIST_HEAD(&p->children); |
1204 | INIT_LIST_HEAD(&p->sibling); | |
f41d911f | 1205 | rcu_copy_process(p); |
1da177e4 LT |
1206 | p->vfork_done = NULL; |
1207 | spin_lock_init(&p->alloc_lock); | |
1da177e4 | 1208 | |
1da177e4 LT |
1209 | init_sigpending(&p->pending); |
1210 | ||
64861634 MS |
1211 | p->utime = p->stime = p->gtime = 0; |
1212 | p->utimescaled = p->stimescaled = 0; | |
d99ca3b9 | 1213 | #ifndef CONFIG_VIRT_CPU_ACCOUNTING |
64861634 | 1214 | p->prev_utime = p->prev_stime = 0; |
d99ca3b9 | 1215 | #endif |
a3a2e76c KH |
1216 | #if defined(SPLIT_RSS_COUNTING) |
1217 | memset(&p->rss_stat, 0, sizeof(p->rss_stat)); | |
1218 | #endif | |
172ba844 | 1219 | |
6976675d AV |
1220 | p->default_timer_slack_ns = current->timer_slack_ns; |
1221 | ||
5995477a | 1222 | task_io_accounting_init(&p->ioac); |
1da177e4 LT |
1223 | acct_clear_integrals(p); |
1224 | ||
f06febc9 | 1225 | posix_cpu_timers_init(p); |
1da177e4 | 1226 | |
1da177e4 | 1227 | do_posix_clock_monotonic_gettime(&p->start_time); |
924b42d5 TJ |
1228 | p->real_start_time = p->start_time; |
1229 | monotonic_to_bootbased(&p->real_start_time); | |
1da177e4 | 1230 | p->io_context = NULL; |
1da177e4 | 1231 | p->audit_context = NULL; |
4714d1d3 | 1232 | if (clone_flags & CLONE_THREAD) |
257058ae | 1233 | threadgroup_change_begin(current); |
b4f48b63 | 1234 | cgroup_fork(p); |
1da177e4 | 1235 | #ifdef CONFIG_NUMA |
846a16bf | 1236 | p->mempolicy = mpol_dup(p->mempolicy); |
fb0a685c DRO |
1237 | if (IS_ERR(p->mempolicy)) { |
1238 | retval = PTR_ERR(p->mempolicy); | |
1239 | p->mempolicy = NULL; | |
1240 | goto bad_fork_cleanup_cgroup; | |
1241 | } | |
c61afb18 | 1242 | mpol_fix_fork_child_flag(p); |
1da177e4 | 1243 | #endif |
778d3b0f MH |
1244 | #ifdef CONFIG_CPUSETS |
1245 | p->cpuset_mem_spread_rotor = NUMA_NO_NODE; | |
1246 | p->cpuset_slab_spread_rotor = NUMA_NO_NODE; | |
cc9a6c87 | 1247 | seqcount_init(&p->mems_allowed_seq); |
778d3b0f | 1248 | #endif |
de30a2b3 IM |
1249 | #ifdef CONFIG_TRACE_IRQFLAGS |
1250 | p->irq_events = 0; | |
b36e4758 RK |
1251 | #ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW |
1252 | p->hardirqs_enabled = 1; | |
1253 | #else | |
de30a2b3 | 1254 | p->hardirqs_enabled = 0; |
b36e4758 | 1255 | #endif |
de30a2b3 IM |
1256 | p->hardirq_enable_ip = 0; |
1257 | p->hardirq_enable_event = 0; | |
1258 | p->hardirq_disable_ip = _THIS_IP_; | |
1259 | p->hardirq_disable_event = 0; | |
1260 | p->softirqs_enabled = 1; | |
1261 | p->softirq_enable_ip = _THIS_IP_; | |
1262 | p->softirq_enable_event = 0; | |
1263 | p->softirq_disable_ip = 0; | |
1264 | p->softirq_disable_event = 0; | |
1265 | p->hardirq_context = 0; | |
1266 | p->softirq_context = 0; | |
1267 | #endif | |
fbb9ce95 IM |
1268 | #ifdef CONFIG_LOCKDEP |
1269 | p->lockdep_depth = 0; /* no locks held yet */ | |
1270 | p->curr_chain_key = 0; | |
1271 | p->lockdep_recursion = 0; | |
1272 | #endif | |
1da177e4 | 1273 | |
408894ee IM |
1274 | #ifdef CONFIG_DEBUG_MUTEXES |
1275 | p->blocked_on = NULL; /* not blocked yet */ | |
1276 | #endif | |
569b846d KH |
1277 | #ifdef CONFIG_CGROUP_MEM_RES_CTLR |
1278 | p->memcg_batch.do_batch = 0; | |
1279 | p->memcg_batch.memcg = NULL; | |
1280 | #endif | |
0f481406 | 1281 | |
3c90e6e9 | 1282 | /* Perform scheduler related setup. Assign this task to a CPU. */ |
3e51e3ed | 1283 | sched_fork(p); |
6ab423e0 | 1284 | |
cdd6c482 | 1285 | retval = perf_event_init_task(p); |
6ab423e0 PZ |
1286 | if (retval) |
1287 | goto bad_fork_cleanup_policy; | |
fb0a685c DRO |
1288 | retval = audit_alloc(p); |
1289 | if (retval) | |
f1752eec | 1290 | goto bad_fork_cleanup_policy; |
1da177e4 | 1291 | /* copy all the process information */ |
fb0a685c DRO |
1292 | retval = copy_semundo(clone_flags, p); |
1293 | if (retval) | |
1da177e4 | 1294 | goto bad_fork_cleanup_audit; |
fb0a685c DRO |
1295 | retval = copy_files(clone_flags, p); |
1296 | if (retval) | |
1da177e4 | 1297 | goto bad_fork_cleanup_semundo; |
fb0a685c DRO |
1298 | retval = copy_fs(clone_flags, p); |
1299 | if (retval) | |
1da177e4 | 1300 | goto bad_fork_cleanup_files; |
fb0a685c DRO |
1301 | retval = copy_sighand(clone_flags, p); |
1302 | if (retval) | |
1da177e4 | 1303 | goto bad_fork_cleanup_fs; |
fb0a685c DRO |
1304 | retval = copy_signal(clone_flags, p); |
1305 | if (retval) | |
1da177e4 | 1306 | goto bad_fork_cleanup_sighand; |
fb0a685c DRO |
1307 | retval = copy_mm(clone_flags, p); |
1308 | if (retval) | |
1da177e4 | 1309 | goto bad_fork_cleanup_signal; |
fb0a685c DRO |
1310 | retval = copy_namespaces(clone_flags, p); |
1311 | if (retval) | |
d84f4f99 | 1312 | goto bad_fork_cleanup_mm; |
fb0a685c DRO |
1313 | retval = copy_io(clone_flags, p); |
1314 | if (retval) | |
fd0928df | 1315 | goto bad_fork_cleanup_namespaces; |
6f2c55b8 | 1316 | retval = copy_thread(clone_flags, stack_start, stack_size, p, regs); |
1da177e4 | 1317 | if (retval) |
fd0928df | 1318 | goto bad_fork_cleanup_io; |
1da177e4 | 1319 | |
425fb2b4 PE |
1320 | if (pid != &init_struct_pid) { |
1321 | retval = -ENOMEM; | |
61bce0f1 | 1322 | pid = alloc_pid(p->nsproxy->pid_ns); |
425fb2b4 | 1323 | if (!pid) |
fd0928df | 1324 | goto bad_fork_cleanup_io; |
425fb2b4 PE |
1325 | } |
1326 | ||
1327 | p->pid = pid_nr(pid); | |
1328 | p->tgid = p->pid; | |
1329 | if (clone_flags & CLONE_THREAD) | |
1330 | p->tgid = current->tgid; | |
1331 | ||
1da177e4 LT |
1332 | p->set_child_tid = (clone_flags & CLONE_CHILD_SETTID) ? child_tidptr : NULL; |
1333 | /* | |
1334 | * Clear TID on mm_release()? | |
1335 | */ | |
fb0a685c | 1336 | p->clear_child_tid = (clone_flags & CLONE_CHILD_CLEARTID) ? child_tidptr : NULL; |
73c10101 JA |
1337 | #ifdef CONFIG_BLOCK |
1338 | p->plug = NULL; | |
1339 | #endif | |
42b2dd0a | 1340 | #ifdef CONFIG_FUTEX |
8f17d3a5 IM |
1341 | p->robust_list = NULL; |
1342 | #ifdef CONFIG_COMPAT | |
1343 | p->compat_robust_list = NULL; | |
1344 | #endif | |
c87e2837 IM |
1345 | INIT_LIST_HEAD(&p->pi_state_list); |
1346 | p->pi_state_cache = NULL; | |
42b2dd0a | 1347 | #endif |
f9a3879a GM |
1348 | /* |
1349 | * sigaltstack should be cleared when sharing the same VM | |
1350 | */ | |
1351 | if ((clone_flags & (CLONE_VM|CLONE_VFORK)) == CLONE_VM) | |
1352 | p->sas_ss_sp = p->sas_ss_size = 0; | |
1353 | ||
1da177e4 | 1354 | /* |
6580807d ON |
1355 | * Syscall tracing and stepping should be turned off in the |
1356 | * child regardless of CLONE_PTRACE. | |
1da177e4 | 1357 | */ |
6580807d | 1358 | user_disable_single_step(p); |
1da177e4 | 1359 | clear_tsk_thread_flag(p, TIF_SYSCALL_TRACE); |
ed75e8d5 LV |
1360 | #ifdef TIF_SYSCALL_EMU |
1361 | clear_tsk_thread_flag(p, TIF_SYSCALL_EMU); | |
1362 | #endif | |
9745512c | 1363 | clear_all_latency_tracing(p); |
1da177e4 | 1364 | |
1da177e4 | 1365 | /* ok, now we should be set up.. */ |
5f8aadd8 ON |
1366 | if (clone_flags & CLONE_THREAD) |
1367 | p->exit_signal = -1; | |
1368 | else if (clone_flags & CLONE_PARENT) | |
1369 | p->exit_signal = current->group_leader->exit_signal; | |
1370 | else | |
1371 | p->exit_signal = (clone_flags & CSIGNAL); | |
1372 | ||
1da177e4 LT |
1373 | p->pdeath_signal = 0; |
1374 | p->exit_state = 0; | |
1375 | ||
9d823e8f WF |
1376 | p->nr_dirtied = 0; |
1377 | p->nr_dirtied_pause = 128 >> (PAGE_SHIFT - 10); | |
83712358 | 1378 | p->dirty_paused_when = 0; |
9d823e8f | 1379 | |
1da177e4 LT |
1380 | /* |
1381 | * Ok, make it visible to the rest of the system. | |
1382 | * We dont wake it up yet. | |
1383 | */ | |
1384 | p->group_leader = p; | |
47e65328 | 1385 | INIT_LIST_HEAD(&p->thread_group); |
1da177e4 | 1386 | |
b4f48b63 PM |
1387 | /* Now that the task is set up, run cgroup callbacks if |
1388 | * necessary. We need to run them before the task is visible | |
1389 | * on the tasklist. */ | |
1390 | cgroup_fork_callbacks(p); | |
1391 | cgroup_callbacks_done = 1; | |
1392 | ||
1da177e4 LT |
1393 | /* Need tasklist lock for parent etc handling! */ |
1394 | write_lock_irq(&tasklist_lock); | |
1395 | ||
1da177e4 | 1396 | /* CLONE_PARENT re-uses the old parent */ |
2d5516cb | 1397 | if (clone_flags & (CLONE_PARENT|CLONE_THREAD)) { |
1da177e4 | 1398 | p->real_parent = current->real_parent; |
2d5516cb ON |
1399 | p->parent_exec_id = current->parent_exec_id; |
1400 | } else { | |
1da177e4 | 1401 | p->real_parent = current; |
2d5516cb ON |
1402 | p->parent_exec_id = current->self_exec_id; |
1403 | } | |
1da177e4 | 1404 | |
3f17da69 | 1405 | spin_lock(¤t->sighand->siglock); |
4a2c7a78 ON |
1406 | |
1407 | /* | |
1408 | * Process group and session signals need to be delivered to just the | |
1409 | * parent before the fork or both the parent and the child after the | |
1410 | * fork. Restart if a signal comes in before we add the new process to | |
1411 | * it's process group. | |
1412 | * A fatal signal pending means that current will exit, so the new | |
1413 | * thread can't slip out of an OOM kill (or normal SIGKILL). | |
fb0a685c | 1414 | */ |
23ff4440 | 1415 | recalc_sigpending(); |
4a2c7a78 ON |
1416 | if (signal_pending(current)) { |
1417 | spin_unlock(¤t->sighand->siglock); | |
1418 | write_unlock_irq(&tasklist_lock); | |
1419 | retval = -ERESTARTNOINTR; | |
f7e8b616 | 1420 | goto bad_fork_free_pid; |
4a2c7a78 ON |
1421 | } |
1422 | ||
1da177e4 | 1423 | if (clone_flags & CLONE_THREAD) { |
b3ac022c | 1424 | current->signal->nr_threads++; |
4ab6c083 | 1425 | atomic_inc(¤t->signal->live); |
b3ac022c | 1426 | atomic_inc(¤t->signal->sigcnt); |
1da177e4 | 1427 | p->group_leader = current->group_leader; |
47e65328 | 1428 | list_add_tail_rcu(&p->thread_group, &p->group_leader->thread_group); |
1da177e4 LT |
1429 | } |
1430 | ||
73b9ebfe | 1431 | if (likely(p->pid)) { |
4b9d33e6 | 1432 | ptrace_init_task(p, (clone_flags & CLONE_PTRACE) || trace); |
73b9ebfe ON |
1433 | |
1434 | if (thread_group_leader(p)) { | |
45a68628 | 1435 | if (is_child_reaper(pid)) |
30e49c26 | 1436 | p->nsproxy->pid_ns->child_reaper = p; |
73b9ebfe | 1437 | |
fea9d175 | 1438 | p->signal->leader_pid = pid; |
9c9f4ded | 1439 | p->signal->tty = tty_kref_get(current->signal->tty); |
5cd17569 EB |
1440 | attach_pid(p, PIDTYPE_PGID, task_pgrp(current)); |
1441 | attach_pid(p, PIDTYPE_SID, task_session(current)); | |
9cd80bbb | 1442 | list_add_tail(&p->sibling, &p->real_parent->children); |
5e85d4ab | 1443 | list_add_tail_rcu(&p->tasks, &init_task.tasks); |
909ea964 | 1444 | __this_cpu_inc(process_counts); |
73b9ebfe | 1445 | } |
85868995 | 1446 | attach_pid(p, PIDTYPE_PID, pid); |
73b9ebfe | 1447 | nr_threads++; |
1da177e4 LT |
1448 | } |
1449 | ||
1da177e4 | 1450 | total_forks++; |
3f17da69 | 1451 | spin_unlock(¤t->sighand->siglock); |
1da177e4 | 1452 | write_unlock_irq(&tasklist_lock); |
c13cf856 | 1453 | proc_fork_connector(p); |
817929ec | 1454 | cgroup_post_fork(p); |
4714d1d3 | 1455 | if (clone_flags & CLONE_THREAD) |
257058ae | 1456 | threadgroup_change_end(current); |
cdd6c482 | 1457 | perf_event_fork(p); |
43d2b113 KH |
1458 | |
1459 | trace_task_newtask(p, clone_flags); | |
1460 | ||
1da177e4 LT |
1461 | return p; |
1462 | ||
425fb2b4 PE |
1463 | bad_fork_free_pid: |
1464 | if (pid != &init_struct_pid) | |
1465 | free_pid(pid); | |
fd0928df | 1466 | bad_fork_cleanup_io: |
b69f2292 LR |
1467 | if (p->io_context) |
1468 | exit_io_context(p); | |
ab516013 | 1469 | bad_fork_cleanup_namespaces: |
444f378b | 1470 | exit_task_namespaces(p); |
1da177e4 | 1471 | bad_fork_cleanup_mm: |
c9f01245 | 1472 | if (p->mm) |
1da177e4 LT |
1473 | mmput(p->mm); |
1474 | bad_fork_cleanup_signal: | |
4ab6c083 | 1475 | if (!(clone_flags & CLONE_THREAD)) |
1c5354de | 1476 | free_signal_struct(p->signal); |
1da177e4 | 1477 | bad_fork_cleanup_sighand: |
a7e5328a | 1478 | __cleanup_sighand(p->sighand); |
1da177e4 LT |
1479 | bad_fork_cleanup_fs: |
1480 | exit_fs(p); /* blocking */ | |
1481 | bad_fork_cleanup_files: | |
1482 | exit_files(p); /* blocking */ | |
1483 | bad_fork_cleanup_semundo: | |
1484 | exit_sem(p); | |
1485 | bad_fork_cleanup_audit: | |
1486 | audit_free(p); | |
1da177e4 | 1487 | bad_fork_cleanup_policy: |
cdd6c482 | 1488 | perf_event_free_task(p); |
1da177e4 | 1489 | #ifdef CONFIG_NUMA |
f0be3d32 | 1490 | mpol_put(p->mempolicy); |
b4f48b63 | 1491 | bad_fork_cleanup_cgroup: |
1da177e4 | 1492 | #endif |
4714d1d3 | 1493 | if (clone_flags & CLONE_THREAD) |
257058ae | 1494 | threadgroup_change_end(current); |
b4f48b63 | 1495 | cgroup_exit(p, cgroup_callbacks_done); |
35df17c5 | 1496 | delayacct_tsk_free(p); |
a1261f54 | 1497 | module_put(task_thread_info(p)->exec_domain->module); |
1da177e4 | 1498 | bad_fork_cleanup_count: |
d84f4f99 | 1499 | atomic_dec(&p->cred->user->processes); |
e0e81739 | 1500 | exit_creds(p); |
1da177e4 LT |
1501 | bad_fork_free: |
1502 | free_task(p); | |
fe7d37d1 ON |
1503 | fork_out: |
1504 | return ERR_PTR(retval); | |
1da177e4 LT |
1505 | } |
1506 | ||
6b2fb3c6 | 1507 | noinline struct pt_regs * __cpuinit __attribute__((weak)) idle_regs(struct pt_regs *regs) |
1da177e4 LT |
1508 | { |
1509 | memset(regs, 0, sizeof(struct pt_regs)); | |
1510 | return regs; | |
1511 | } | |
1512 | ||
f106eee1 ON |
1513 | static inline void init_idle_pids(struct pid_link *links) |
1514 | { | |
1515 | enum pid_type type; | |
1516 | ||
1517 | for (type = PIDTYPE_PID; type < PIDTYPE_MAX; ++type) { | |
1518 | INIT_HLIST_NODE(&links[type].node); /* not really needed */ | |
1519 | links[type].pid = &init_struct_pid; | |
1520 | } | |
1521 | } | |
1522 | ||
9abcf40b | 1523 | struct task_struct * __cpuinit fork_idle(int cpu) |
1da177e4 | 1524 | { |
36c8b586 | 1525 | struct task_struct *task; |
1da177e4 LT |
1526 | struct pt_regs regs; |
1527 | ||
30e49c26 | 1528 | task = copy_process(CLONE_VM, 0, idle_regs(®s), 0, NULL, |
09a05394 | 1529 | &init_struct_pid, 0); |
f106eee1 ON |
1530 | if (!IS_ERR(task)) { |
1531 | init_idle_pids(task->pids); | |
753ca4f3 | 1532 | init_idle(task, cpu); |
f106eee1 | 1533 | } |
73b9ebfe | 1534 | |
1da177e4 LT |
1535 | return task; |
1536 | } | |
1537 | ||
1da177e4 LT |
1538 | /* |
1539 | * Ok, this is the main fork-routine. | |
1540 | * | |
1541 | * It copies the process, and if successful kick-starts | |
1542 | * it and waits for it to finish using the VM if required. | |
1543 | */ | |
1544 | long do_fork(unsigned long clone_flags, | |
1545 | unsigned long stack_start, | |
1546 | struct pt_regs *regs, | |
1547 | unsigned long stack_size, | |
1548 | int __user *parent_tidptr, | |
1549 | int __user *child_tidptr) | |
1550 | { | |
1551 | struct task_struct *p; | |
1552 | int trace = 0; | |
92476d7f | 1553 | long nr; |
1da177e4 | 1554 | |
18b6e041 SH |
1555 | /* |
1556 | * Do some preliminary argument and permissions checking before we | |
1557 | * actually start allocating stuff | |
1558 | */ | |
1559 | if (clone_flags & CLONE_NEWUSER) { | |
1560 | if (clone_flags & CLONE_THREAD) | |
1561 | return -EINVAL; | |
1562 | /* hopefully this check will go away when userns support is | |
1563 | * complete | |
1564 | */ | |
7657d904 SH |
1565 | if (!capable(CAP_SYS_ADMIN) || !capable(CAP_SETUID) || |
1566 | !capable(CAP_SETGID)) | |
18b6e041 SH |
1567 | return -EPERM; |
1568 | } | |
1569 | ||
09a05394 | 1570 | /* |
4b9d33e6 TH |
1571 | * Determine whether and which event to report to ptracer. When |
1572 | * called from kernel_thread or CLONE_UNTRACED is explicitly | |
1573 | * requested, no event is reported; otherwise, report if the event | |
1574 | * for the type of forking is enabled. | |
09a05394 | 1575 | */ |
4b9d33e6 TH |
1576 | if (likely(user_mode(regs)) && !(clone_flags & CLONE_UNTRACED)) { |
1577 | if (clone_flags & CLONE_VFORK) | |
1578 | trace = PTRACE_EVENT_VFORK; | |
1579 | else if ((clone_flags & CSIGNAL) != SIGCHLD) | |
1580 | trace = PTRACE_EVENT_CLONE; | |
1581 | else | |
1582 | trace = PTRACE_EVENT_FORK; | |
1583 | ||
1584 | if (likely(!ptrace_event_enabled(current, trace))) | |
1585 | trace = 0; | |
1586 | } | |
1da177e4 | 1587 | |
a6f5e063 | 1588 | p = copy_process(clone_flags, stack_start, regs, stack_size, |
09a05394 | 1589 | child_tidptr, NULL, trace); |
1da177e4 LT |
1590 | /* |
1591 | * Do this prior waking up the new thread - the thread pointer | |
1592 | * might get invalid after that point, if the thread exits quickly. | |
1593 | */ | |
1594 | if (!IS_ERR(p)) { | |
1595 | struct completion vfork; | |
1596 | ||
0a16b607 MD |
1597 | trace_sched_process_fork(current, p); |
1598 | ||
6c5f3e7b | 1599 | nr = task_pid_vnr(p); |
30e49c26 PE |
1600 | |
1601 | if (clone_flags & CLONE_PARENT_SETTID) | |
1602 | put_user(nr, parent_tidptr); | |
a6f5e063 | 1603 | |
1da177e4 LT |
1604 | if (clone_flags & CLONE_VFORK) { |
1605 | p->vfork_done = &vfork; | |
1606 | init_completion(&vfork); | |
d68b46fe | 1607 | get_task_struct(p); |
1da177e4 LT |
1608 | } |
1609 | ||
3e51e3ed | 1610 | wake_up_new_task(p); |
1da177e4 | 1611 | |
4b9d33e6 TH |
1612 | /* forking complete and child started to run, tell ptracer */ |
1613 | if (unlikely(trace)) | |
1614 | ptrace_event(trace, nr); | |
09a05394 | 1615 | |
1da177e4 | 1616 | if (clone_flags & CLONE_VFORK) { |
d68b46fe ON |
1617 | if (!wait_for_vfork_done(p, &vfork)) |
1618 | ptrace_event(PTRACE_EVENT_VFORK_DONE, nr); | |
1da177e4 LT |
1619 | } |
1620 | } else { | |
92476d7f | 1621 | nr = PTR_ERR(p); |
1da177e4 | 1622 | } |
92476d7f | 1623 | return nr; |
1da177e4 LT |
1624 | } |
1625 | ||
5fd63b30 RT |
1626 | #ifndef ARCH_MIN_MMSTRUCT_ALIGN |
1627 | #define ARCH_MIN_MMSTRUCT_ALIGN 0 | |
1628 | #endif | |
1629 | ||
51cc5068 | 1630 | static void sighand_ctor(void *data) |
aa1757f9 ON |
1631 | { |
1632 | struct sighand_struct *sighand = data; | |
1633 | ||
a35afb83 | 1634 | spin_lock_init(&sighand->siglock); |
b8fceee1 | 1635 | init_waitqueue_head(&sighand->signalfd_wqh); |
aa1757f9 ON |
1636 | } |
1637 | ||
1da177e4 LT |
1638 | void __init proc_caches_init(void) |
1639 | { | |
1640 | sighand_cachep = kmem_cache_create("sighand_cache", | |
1641 | sizeof(struct sighand_struct), 0, | |
2dff4405 VN |
1642 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_DESTROY_BY_RCU| |
1643 | SLAB_NOTRACK, sighand_ctor); | |
1da177e4 LT |
1644 | signal_cachep = kmem_cache_create("signal_cache", |
1645 | sizeof(struct signal_struct), 0, | |
2dff4405 | 1646 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL); |
20c2df83 | 1647 | files_cachep = kmem_cache_create("files_cache", |
1da177e4 | 1648 | sizeof(struct files_struct), 0, |
2dff4405 | 1649 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL); |
20c2df83 | 1650 | fs_cachep = kmem_cache_create("fs_cache", |
1da177e4 | 1651 | sizeof(struct fs_struct), 0, |
2dff4405 | 1652 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL); |
6345d24d LT |
1653 | /* |
1654 | * FIXME! The "sizeof(struct mm_struct)" currently includes the | |
1655 | * whole struct cpumask for the OFFSTACK case. We could change | |
1656 | * this to *only* allocate as much of it as required by the | |
1657 | * maximum number of CPU's we can ever have. The cpumask_allocation | |
1658 | * is at the end of the structure, exactly for that reason. | |
1659 | */ | |
1da177e4 | 1660 | mm_cachep = kmem_cache_create("mm_struct", |
5fd63b30 | 1661 | sizeof(struct mm_struct), ARCH_MIN_MMSTRUCT_ALIGN, |
2dff4405 | 1662 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL); |
33e5d769 | 1663 | vm_area_cachep = KMEM_CACHE(vm_area_struct, SLAB_PANIC); |
8feae131 | 1664 | mmap_init(); |
66577193 | 1665 | nsproxy_cache_init(); |
1da177e4 | 1666 | } |
cf2e340f | 1667 | |
cf2e340f | 1668 | /* |
9bfb23fc | 1669 | * Check constraints on flags passed to the unshare system call. |
cf2e340f | 1670 | */ |
9bfb23fc | 1671 | static int check_unshare_flags(unsigned long unshare_flags) |
cf2e340f | 1672 | { |
9bfb23fc ON |
1673 | if (unshare_flags & ~(CLONE_THREAD|CLONE_FS|CLONE_NEWNS|CLONE_SIGHAND| |
1674 | CLONE_VM|CLONE_FILES|CLONE_SYSVSEM| | |
1675 | CLONE_NEWUTS|CLONE_NEWIPC|CLONE_NEWNET)) | |
1676 | return -EINVAL; | |
cf2e340f | 1677 | /* |
9bfb23fc ON |
1678 | * Not implemented, but pretend it works if there is nothing to |
1679 | * unshare. Note that unsharing CLONE_THREAD or CLONE_SIGHAND | |
1680 | * needs to unshare vm. | |
cf2e340f | 1681 | */ |
9bfb23fc ON |
1682 | if (unshare_flags & (CLONE_THREAD | CLONE_SIGHAND | CLONE_VM)) { |
1683 | /* FIXME: get_task_mm() increments ->mm_users */ | |
1684 | if (atomic_read(¤t->mm->mm_users) > 1) | |
1685 | return -EINVAL; | |
1686 | } | |
cf2e340f JD |
1687 | |
1688 | return 0; | |
1689 | } | |
1690 | ||
1691 | /* | |
99d1419d | 1692 | * Unshare the filesystem structure if it is being shared |
cf2e340f JD |
1693 | */ |
1694 | static int unshare_fs(unsigned long unshare_flags, struct fs_struct **new_fsp) | |
1695 | { | |
1696 | struct fs_struct *fs = current->fs; | |
1697 | ||
498052bb AV |
1698 | if (!(unshare_flags & CLONE_FS) || !fs) |
1699 | return 0; | |
1700 | ||
1701 | /* don't need lock here; in the worst case we'll do useless copy */ | |
1702 | if (fs->users == 1) | |
1703 | return 0; | |
1704 | ||
1705 | *new_fsp = copy_fs_struct(fs); | |
1706 | if (!*new_fsp) | |
1707 | return -ENOMEM; | |
cf2e340f JD |
1708 | |
1709 | return 0; | |
1710 | } | |
1711 | ||
cf2e340f | 1712 | /* |
a016f338 | 1713 | * Unshare file descriptor table if it is being shared |
cf2e340f JD |
1714 | */ |
1715 | static int unshare_fd(unsigned long unshare_flags, struct files_struct **new_fdp) | |
1716 | { | |
1717 | struct files_struct *fd = current->files; | |
a016f338 | 1718 | int error = 0; |
cf2e340f JD |
1719 | |
1720 | if ((unshare_flags & CLONE_FILES) && | |
a016f338 JD |
1721 | (fd && atomic_read(&fd->count) > 1)) { |
1722 | *new_fdp = dup_fd(fd, &error); | |
1723 | if (!*new_fdp) | |
1724 | return error; | |
1725 | } | |
cf2e340f JD |
1726 | |
1727 | return 0; | |
1728 | } | |
1729 | ||
cf2e340f JD |
1730 | /* |
1731 | * unshare allows a process to 'unshare' part of the process | |
1732 | * context which was originally shared using clone. copy_* | |
1733 | * functions used by do_fork() cannot be used here directly | |
1734 | * because they modify an inactive task_struct that is being | |
1735 | * constructed. Here we are modifying the current, active, | |
1736 | * task_struct. | |
1737 | */ | |
6559eed8 | 1738 | SYSCALL_DEFINE1(unshare, unsigned long, unshare_flags) |
cf2e340f | 1739 | { |
cf2e340f | 1740 | struct fs_struct *fs, *new_fs = NULL; |
cf2e340f | 1741 | struct files_struct *fd, *new_fd = NULL; |
cf7b708c | 1742 | struct nsproxy *new_nsproxy = NULL; |
9edff4ab | 1743 | int do_sysvsem = 0; |
9bfb23fc | 1744 | int err; |
cf2e340f | 1745 | |
9bfb23fc ON |
1746 | err = check_unshare_flags(unshare_flags); |
1747 | if (err) | |
06f9d4f9 EB |
1748 | goto bad_unshare_out; |
1749 | ||
9bfb23fc ON |
1750 | /* |
1751 | * If unsharing namespace, must also unshare filesystem information. | |
1752 | */ | |
1753 | if (unshare_flags & CLONE_NEWNS) | |
1754 | unshare_flags |= CLONE_FS; | |
6013f67f MS |
1755 | /* |
1756 | * CLONE_NEWIPC must also detach from the undolist: after switching | |
1757 | * to a new ipc namespace, the semaphore arrays from the old | |
1758 | * namespace are unreachable. | |
1759 | */ | |
1760 | if (unshare_flags & (CLONE_NEWIPC|CLONE_SYSVSEM)) | |
9edff4ab | 1761 | do_sysvsem = 1; |
fb0a685c DRO |
1762 | err = unshare_fs(unshare_flags, &new_fs); |
1763 | if (err) | |
9bfb23fc | 1764 | goto bad_unshare_out; |
fb0a685c DRO |
1765 | err = unshare_fd(unshare_flags, &new_fd); |
1766 | if (err) | |
9bfb23fc | 1767 | goto bad_unshare_cleanup_fs; |
fb0a685c DRO |
1768 | err = unshare_nsproxy_namespaces(unshare_flags, &new_nsproxy, new_fs); |
1769 | if (err) | |
9edff4ab | 1770 | goto bad_unshare_cleanup_fd; |
c0b2fc31 | 1771 | |
9bfb23fc | 1772 | if (new_fs || new_fd || do_sysvsem || new_nsproxy) { |
9edff4ab MS |
1773 | if (do_sysvsem) { |
1774 | /* | |
1775 | * CLONE_SYSVSEM is equivalent to sys_exit(). | |
1776 | */ | |
1777 | exit_sem(current); | |
1778 | } | |
ab516013 | 1779 | |
c0b2fc31 | 1780 | if (new_nsproxy) { |
cf7b708c PE |
1781 | switch_task_namespaces(current, new_nsproxy); |
1782 | new_nsproxy = NULL; | |
c0b2fc31 | 1783 | } |
cf2e340f | 1784 | |
cf7b708c PE |
1785 | task_lock(current); |
1786 | ||
cf2e340f JD |
1787 | if (new_fs) { |
1788 | fs = current->fs; | |
2a4419b5 | 1789 | spin_lock(&fs->lock); |
cf2e340f | 1790 | current->fs = new_fs; |
498052bb AV |
1791 | if (--fs->users) |
1792 | new_fs = NULL; | |
1793 | else | |
1794 | new_fs = fs; | |
2a4419b5 | 1795 | spin_unlock(&fs->lock); |
cf2e340f JD |
1796 | } |
1797 | ||
cf2e340f JD |
1798 | if (new_fd) { |
1799 | fd = current->files; | |
1800 | current->files = new_fd; | |
1801 | new_fd = fd; | |
1802 | } | |
1803 | ||
1804 | task_unlock(current); | |
1805 | } | |
1806 | ||
c0b2fc31 | 1807 | if (new_nsproxy) |
444f378b | 1808 | put_nsproxy(new_nsproxy); |
c0b2fc31 | 1809 | |
cf2e340f JD |
1810 | bad_unshare_cleanup_fd: |
1811 | if (new_fd) | |
1812 | put_files_struct(new_fd); | |
1813 | ||
cf2e340f JD |
1814 | bad_unshare_cleanup_fs: |
1815 | if (new_fs) | |
498052bb | 1816 | free_fs_struct(new_fs); |
cf2e340f | 1817 | |
cf2e340f JD |
1818 | bad_unshare_out: |
1819 | return err; | |
1820 | } | |
3b125388 AV |
1821 | |
1822 | /* | |
1823 | * Helper to unshare the files of the current task. | |
1824 | * We don't want to expose copy_files internals to | |
1825 | * the exec layer of the kernel. | |
1826 | */ | |
1827 | ||
1828 | int unshare_files(struct files_struct **displaced) | |
1829 | { | |
1830 | struct task_struct *task = current; | |
50704516 | 1831 | struct files_struct *copy = NULL; |
3b125388 AV |
1832 | int error; |
1833 | ||
1834 | error = unshare_fd(CLONE_FILES, ©); | |
1835 | if (error || !copy) { | |
1836 | *displaced = NULL; | |
1837 | return error; | |
1838 | } | |
1839 | *displaced = task->files; | |
1840 | task_lock(task); | |
1841 | task->files = copy; | |
1842 | task_unlock(task); | |
1843 | return 0; | |
1844 | } |