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> |
e2cfabdf | 37 | #include <linux/seccomp.h> |
1da177e4 LT |
38 | #include <linux/swap.h> |
39 | #include <linux/syscalls.h> | |
40 | #include <linux/jiffies.h> | |
41 | #include <linux/futex.h> | |
8141c7f3 | 42 | #include <linux/compat.h> |
207205a2 | 43 | #include <linux/kthread.h> |
7c3ab738 | 44 | #include <linux/task_io_accounting_ops.h> |
ab2af1f5 | 45 | #include <linux/rcupdate.h> |
1da177e4 LT |
46 | #include <linux/ptrace.h> |
47 | #include <linux/mount.h> | |
48 | #include <linux/audit.h> | |
78fb7466 | 49 | #include <linux/memcontrol.h> |
f201ae23 | 50 | #include <linux/ftrace.h> |
5e2bf014 | 51 | #include <linux/proc_fs.h> |
1da177e4 LT |
52 | #include <linux/profile.h> |
53 | #include <linux/rmap.h> | |
f8af4da3 | 54 | #include <linux/ksm.h> |
1da177e4 | 55 | #include <linux/acct.h> |
8f0ab514 | 56 | #include <linux/tsacct_kern.h> |
9f46080c | 57 | #include <linux/cn_proc.h> |
ba96a0c8 | 58 | #include <linux/freezer.h> |
ca74e92b | 59 | #include <linux/delayacct.h> |
ad4ecbcb | 60 | #include <linux/taskstats_kern.h> |
0a425405 | 61 | #include <linux/random.h> |
522ed776 | 62 | #include <linux/tty.h> |
fd0928df | 63 | #include <linux/blkdev.h> |
5ad4e53b | 64 | #include <linux/fs_struct.h> |
7c9f8861 | 65 | #include <linux/magic.h> |
cdd6c482 | 66 | #include <linux/perf_event.h> |
42c4ab41 | 67 | #include <linux/posix-timers.h> |
8e7cac79 | 68 | #include <linux/user-return-notifier.h> |
3d5992d2 | 69 | #include <linux/oom.h> |
ba76149f | 70 | #include <linux/khugepaged.h> |
d80e731e | 71 | #include <linux/signalfd.h> |
0326f5a9 | 72 | #include <linux/uprobes.h> |
a27bb332 | 73 | #include <linux/aio.h> |
1da177e4 LT |
74 | |
75 | #include <asm/pgtable.h> | |
76 | #include <asm/pgalloc.h> | |
77 | #include <asm/uaccess.h> | |
78 | #include <asm/mmu_context.h> | |
79 | #include <asm/cacheflush.h> | |
80 | #include <asm/tlbflush.h> | |
81 | ||
ad8d75ff SR |
82 | #include <trace/events/sched.h> |
83 | ||
43d2b113 KH |
84 | #define CREATE_TRACE_POINTS |
85 | #include <trace/events/task.h> | |
86 | ||
6fa3eb70 S |
87 | #ifdef CONFIG_MT_PRIO_TRACER |
88 | # include <linux/prio_tracer.h> | |
89 | #endif | |
90 | ||
1da177e4 LT |
91 | /* |
92 | * Protected counters by write_lock_irq(&tasklist_lock) | |
93 | */ | |
94 | unsigned long total_forks; /* Handle normal Linux uptimes. */ | |
fb0a685c | 95 | int nr_threads; /* The idle threads do not count.. */ |
1da177e4 LT |
96 | |
97 | int max_threads; /* tunable limit on nr_threads */ | |
98 | ||
99 | DEFINE_PER_CPU(unsigned long, process_counts) = 0; | |
100 | ||
c59923a1 | 101 | __cacheline_aligned DEFINE_RWLOCK(tasklist_lock); /* outer */ |
db1466b3 PM |
102 | |
103 | #ifdef CONFIG_PROVE_RCU | |
104 | int lockdep_tasklist_lock_is_held(void) | |
105 | { | |
106 | return lockdep_is_held(&tasklist_lock); | |
107 | } | |
108 | EXPORT_SYMBOL_GPL(lockdep_tasklist_lock_is_held); | |
109 | #endif /* #ifdef CONFIG_PROVE_RCU */ | |
1da177e4 LT |
110 | |
111 | int nr_processes(void) | |
112 | { | |
113 | int cpu; | |
114 | int total = 0; | |
115 | ||
1d510750 | 116 | for_each_possible_cpu(cpu) |
1da177e4 LT |
117 | total += per_cpu(process_counts, cpu); |
118 | ||
119 | return total; | |
120 | } | |
121 | ||
f19b9f74 AM |
122 | void __weak arch_release_task_struct(struct task_struct *tsk) |
123 | { | |
124 | } | |
125 | ||
f5e10287 | 126 | #ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR |
e18b890b | 127 | static struct kmem_cache *task_struct_cachep; |
41101809 TG |
128 | |
129 | static inline struct task_struct *alloc_task_struct_node(int node) | |
130 | { | |
131 | return kmem_cache_alloc_node(task_struct_cachep, GFP_KERNEL, node); | |
132 | } | |
133 | ||
41101809 TG |
134 | static inline void free_task_struct(struct task_struct *tsk) |
135 | { | |
41101809 TG |
136 | kmem_cache_free(task_struct_cachep, tsk); |
137 | } | |
1da177e4 LT |
138 | #endif |
139 | ||
f19b9f74 AM |
140 | void __weak arch_release_thread_info(struct thread_info *ti) |
141 | { | |
142 | } | |
143 | ||
f5e10287 | 144 | #ifndef CONFIG_ARCH_THREAD_INFO_ALLOCATOR |
41101809 | 145 | |
0d15d74a TG |
146 | /* |
147 | * Allocate pages if THREAD_SIZE is >= PAGE_SIZE, otherwise use a | |
148 | * kmemcache based allocator. | |
149 | */ | |
150 | # if THREAD_SIZE >= PAGE_SIZE | |
b6a84016 ED |
151 | static struct thread_info *alloc_thread_info_node(struct task_struct *tsk, |
152 | int node) | |
b69c49b7 | 153 | { |
2ad306b1 | 154 | struct page *page = alloc_pages_node(node, THREADINFO_GFP_ACCOUNTED, |
2889f608 | 155 | THREAD_SIZE_ORDER); |
b6a84016 ED |
156 | |
157 | return page ? page_address(page) : NULL; | |
b69c49b7 FT |
158 | } |
159 | ||
160 | static inline void free_thread_info(struct thread_info *ti) | |
161 | { | |
2ad306b1 | 162 | free_memcg_kmem_pages((unsigned long)ti, THREAD_SIZE_ORDER); |
b69c49b7 | 163 | } |
0d15d74a TG |
164 | # else |
165 | static struct kmem_cache *thread_info_cache; | |
166 | ||
167 | static struct thread_info *alloc_thread_info_node(struct task_struct *tsk, | |
168 | int node) | |
169 | { | |
170 | return kmem_cache_alloc_node(thread_info_cache, THREADINFO_GFP, node); | |
171 | } | |
172 | ||
173 | static void free_thread_info(struct thread_info *ti) | |
174 | { | |
0d15d74a TG |
175 | kmem_cache_free(thread_info_cache, ti); |
176 | } | |
177 | ||
178 | void thread_info_cache_init(void) | |
179 | { | |
180 | thread_info_cache = kmem_cache_create("thread_info", THREAD_SIZE, | |
181 | THREAD_SIZE, 0, NULL); | |
182 | BUG_ON(thread_info_cache == NULL); | |
183 | } | |
184 | # endif | |
b69c49b7 FT |
185 | #endif |
186 | ||
1da177e4 | 187 | /* SLAB cache for signal_struct structures (tsk->signal) */ |
e18b890b | 188 | static struct kmem_cache *signal_cachep; |
1da177e4 LT |
189 | |
190 | /* SLAB cache for sighand_struct structures (tsk->sighand) */ | |
e18b890b | 191 | struct kmem_cache *sighand_cachep; |
1da177e4 LT |
192 | |
193 | /* SLAB cache for files_struct structures (tsk->files) */ | |
e18b890b | 194 | struct kmem_cache *files_cachep; |
1da177e4 LT |
195 | |
196 | /* SLAB cache for fs_struct structures (tsk->fs) */ | |
e18b890b | 197 | struct kmem_cache *fs_cachep; |
1da177e4 LT |
198 | |
199 | /* SLAB cache for vm_area_struct structures */ | |
e18b890b | 200 | struct kmem_cache *vm_area_cachep; |
1da177e4 LT |
201 | |
202 | /* SLAB cache for mm_struct structures (tsk->mm) */ | |
e18b890b | 203 | static struct kmem_cache *mm_cachep; |
1da177e4 | 204 | |
6fa3eb70 S |
205 | /* Notifier list called when a task struct is freed */ |
206 | static ATOMIC_NOTIFIER_HEAD(task_free_notifier); | |
207 | ||
c6a7f572 KM |
208 | static void account_kernel_stack(struct thread_info *ti, int account) |
209 | { | |
210 | struct zone *zone = page_zone(virt_to_page(ti)); | |
211 | ||
212 | mod_zone_page_state(zone, NR_KERNEL_STACK, account); | |
213 | } | |
214 | ||
1da177e4 LT |
215 | void free_task(struct task_struct *tsk) |
216 | { | |
c6a7f572 | 217 | account_kernel_stack(tsk->stack, -1); |
f19b9f74 | 218 | arch_release_thread_info(tsk->stack); |
f7e4217b | 219 | free_thread_info(tsk->stack); |
23f78d4a | 220 | rt_mutex_debug_task_free(tsk); |
fb52607a | 221 | ftrace_graph_exit_task(tsk); |
e2cfabdf | 222 | put_seccomp_filter(tsk); |
f19b9f74 | 223 | arch_release_task_struct(tsk); |
1da177e4 LT |
224 | free_task_struct(tsk); |
225 | } | |
226 | EXPORT_SYMBOL(free_task); | |
227 | ||
ea6d290c ON |
228 | static inline void free_signal_struct(struct signal_struct *sig) |
229 | { | |
97101eb4 | 230 | taskstats_tgid_free(sig); |
1c5354de | 231 | sched_autogroup_exit(sig); |
ea6d290c ON |
232 | kmem_cache_free(signal_cachep, sig); |
233 | } | |
234 | ||
235 | static inline void put_signal_struct(struct signal_struct *sig) | |
236 | { | |
1c5354de | 237 | if (atomic_dec_and_test(&sig->sigcnt)) |
ea6d290c ON |
238 | free_signal_struct(sig); |
239 | } | |
240 | ||
6fa3eb70 S |
241 | int task_free_register(struct notifier_block *n) |
242 | { | |
243 | return atomic_notifier_chain_register(&task_free_notifier, n); | |
244 | } | |
245 | EXPORT_SYMBOL(task_free_register); | |
246 | ||
247 | int task_free_unregister(struct notifier_block *n) | |
248 | { | |
249 | return atomic_notifier_chain_unregister(&task_free_notifier, n); | |
250 | } | |
251 | EXPORT_SYMBOL(task_free_unregister); | |
252 | ||
158d9ebd | 253 | void __put_task_struct(struct task_struct *tsk) |
1da177e4 | 254 | { |
270f722d | 255 | WARN_ON(!tsk->exit_state); |
1da177e4 LT |
256 | WARN_ON(atomic_read(&tsk->usage)); |
257 | WARN_ON(tsk == current); | |
258 | ||
1a2a4d06 | 259 | security_task_free(tsk); |
e0e81739 | 260 | exit_creds(tsk); |
35df17c5 | 261 | delayacct_tsk_free(tsk); |
ea6d290c | 262 | put_signal_struct(tsk->signal); |
1da177e4 | 263 | |
6fa3eb70 | 264 | atomic_notifier_call_chain(&task_free_notifier, 0, tsk); |
1da177e4 LT |
265 | if (!profile_handoff_task(tsk)) |
266 | free_task(tsk); | |
267 | } | |
77c100c8 | 268 | EXPORT_SYMBOL_GPL(__put_task_struct); |
1da177e4 | 269 | |
6c0a9fa6 | 270 | void __init __weak arch_task_cache_init(void) { } |
61c4628b | 271 | |
1da177e4 LT |
272 | void __init fork_init(unsigned long mempages) |
273 | { | |
f5e10287 | 274 | #ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR |
1da177e4 LT |
275 | #ifndef ARCH_MIN_TASKALIGN |
276 | #define ARCH_MIN_TASKALIGN L1_CACHE_BYTES | |
277 | #endif | |
278 | /* create a slab on which task_structs can be allocated */ | |
279 | task_struct_cachep = | |
280 | kmem_cache_create("task_struct", sizeof(struct task_struct), | |
2dff4405 | 281 | ARCH_MIN_TASKALIGN, SLAB_PANIC | SLAB_NOTRACK, NULL); |
1da177e4 LT |
282 | #endif |
283 | ||
61c4628b SS |
284 | /* do the arch specific task caches init */ |
285 | arch_task_cache_init(); | |
286 | ||
1da177e4 LT |
287 | /* |
288 | * The default maximum number of threads is set to a safe | |
289 | * value: the thread structures can take up at most half | |
290 | * of memory. | |
291 | */ | |
292 | max_threads = mempages / (8 * THREAD_SIZE / PAGE_SIZE); | |
293 | ||
294 | /* | |
295 | * we need to allow at least 20 threads to boot a system | |
296 | */ | |
fb0a685c | 297 | if (max_threads < 20) |
1da177e4 LT |
298 | max_threads = 20; |
299 | ||
300 | init_task.signal->rlim[RLIMIT_NPROC].rlim_cur = max_threads/2; | |
301 | init_task.signal->rlim[RLIMIT_NPROC].rlim_max = max_threads/2; | |
302 | init_task.signal->rlim[RLIMIT_SIGPENDING] = | |
303 | init_task.signal->rlim[RLIMIT_NPROC]; | |
304 | } | |
305 | ||
61c4628b SS |
306 | int __attribute__((weak)) arch_dup_task_struct(struct task_struct *dst, |
307 | struct task_struct *src) | |
308 | { | |
309 | *dst = *src; | |
310 | return 0; | |
311 | } | |
312 | ||
1da177e4 LT |
313 | static struct task_struct *dup_task_struct(struct task_struct *orig) |
314 | { | |
315 | struct task_struct *tsk; | |
316 | struct thread_info *ti; | |
7c9f8861 | 317 | unsigned long *stackend; |
207205a2 | 318 | int node = tsk_fork_get_node(orig); |
3e26c149 | 319 | int err; |
1da177e4 | 320 | |
504f52b5 | 321 | tsk = alloc_task_struct_node(node); |
6fa3eb70 S |
322 | if (!tsk){ |
323 | printk("[%d:%s] fork fail at alloc_tsk_node, please check kmem_cache_alloc_node()\n", current->pid, current->comm); | |
1da177e4 | 324 | return NULL; |
6fa3eb70 | 325 | } |
b6a84016 | 326 | ti = alloc_thread_info_node(tsk, node); |
6fa3eb70 S |
327 | if (!ti) { |
328 | printk("[%d:%s] fork fail at alloc_t_info_node, please check alloc_pages_node()\n", current->pid, current->comm); | |
f19b9f74 | 329 | goto free_tsk; |
6fa3eb70 | 330 | } |
1da177e4 | 331 | |
fb0a685c | 332 | err = arch_dup_task_struct(tsk, orig); |
6fa3eb70 S |
333 | if (err){ |
334 | printk("[%d:%s] fork fail at arch_dup_task_struct, err:%d \n", current->pid, current->comm, err); | |
f19b9f74 | 335 | goto free_ti; |
6fa3eb70 | 336 | } |
87bec58a | 337 | tsk->stack = ti; |
6fa3eb70 S |
338 | #ifdef CONFIG_SECCOMP |
339 | /* | |
340 | * We must handle setting up seccomp filters once we're under | |
341 | * the sighand lock in case orig has changed between now and | |
342 | * then. Until then, filter must be NULL to avoid messing up | |
343 | * the usage counts on the error path calling free_task. | |
344 | */ | |
345 | tsk->seccomp.filter = NULL; | |
346 | #endif | |
87bec58a AM |
347 | |
348 | setup_thread_stack(tsk, orig); | |
8e7cac79 | 349 | clear_user_return_notifier(tsk); |
f26f9aff | 350 | clear_tsk_need_resched(tsk); |
7c9f8861 ES |
351 | stackend = end_of_stack(tsk); |
352 | *stackend = STACK_END_MAGIC; /* for overflow detection */ | |
1da177e4 | 353 | |
0a425405 AV |
354 | #ifdef CONFIG_CC_STACKPROTECTOR |
355 | tsk->stack_canary = get_random_int(); | |
356 | #endif | |
357 | ||
fb0a685c DRO |
358 | /* |
359 | * One for us, one for whoever does the "release_task()" (usually | |
360 | * parent) | |
361 | */ | |
362 | atomic_set(&tsk->usage, 2); | |
6c5c9341 | 363 | #ifdef CONFIG_BLK_DEV_IO_TRACE |
2056a782 | 364 | tsk->btrace_seq = 0; |
6c5c9341 | 365 | #endif |
a0aa7f68 | 366 | tsk->splice_pipe = NULL; |
5640f768 | 367 | tsk->task_frag.page = NULL; |
c6a7f572 KM |
368 | |
369 | account_kernel_stack(ti, 1); | |
370 | ||
1da177e4 | 371 | return tsk; |
61c4628b | 372 | |
f19b9f74 | 373 | free_ti: |
61c4628b | 374 | free_thread_info(ti); |
f19b9f74 | 375 | free_tsk: |
61c4628b SS |
376 | free_task_struct(tsk); |
377 | return NULL; | |
1da177e4 LT |
378 | } |
379 | ||
380 | #ifdef CONFIG_MMU | |
a39bc516 | 381 | static int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm) |
1da177e4 | 382 | { |
297c5eee | 383 | struct vm_area_struct *mpnt, *tmp, *prev, **pprev; |
1da177e4 LT |
384 | struct rb_node **rb_link, *rb_parent; |
385 | int retval; | |
386 | unsigned long charge; | |
387 | struct mempolicy *pol; | |
388 | ||
32cdba1e | 389 | uprobe_start_dup_mmap(); |
1da177e4 | 390 | down_write(&oldmm->mmap_sem); |
ec8c0446 | 391 | flush_cache_dup_mm(oldmm); |
f8ac4ec9 | 392 | uprobe_dup_mmap(oldmm, mm); |
ad339451 IM |
393 | /* |
394 | * Not linked in yet - no deadlock potential: | |
395 | */ | |
396 | down_write_nested(&mm->mmap_sem, SINGLE_DEPTH_NESTING); | |
7ee78232 | 397 | |
1da177e4 LT |
398 | mm->locked_vm = 0; |
399 | mm->mmap = NULL; | |
400 | mm->mmap_cache = NULL; | |
401 | mm->free_area_cache = oldmm->mmap_base; | |
1363c3cd | 402 | mm->cached_hole_size = ~0UL; |
1da177e4 | 403 | mm->map_count = 0; |
94894244 | 404 | cpumask_clear(mm_cpumask(mm)); |
1da177e4 LT |
405 | mm->mm_rb = RB_ROOT; |
406 | rb_link = &mm->mm_rb.rb_node; | |
407 | rb_parent = NULL; | |
408 | pprev = &mm->mmap; | |
f8af4da3 | 409 | retval = ksm_fork(mm, oldmm); |
ba76149f AA |
410 | if (retval) |
411 | goto out; | |
412 | retval = khugepaged_fork(mm, oldmm); | |
f8af4da3 HD |
413 | if (retval) |
414 | goto out; | |
1da177e4 | 415 | |
297c5eee | 416 | prev = NULL; |
fd3e42fc | 417 | for (mpnt = oldmm->mmap; mpnt; mpnt = mpnt->vm_next) { |
1da177e4 LT |
418 | struct file *file; |
419 | ||
420 | if (mpnt->vm_flags & VM_DONTCOPY) { | |
ab50b8ed | 421 | vm_stat_account(mm, mpnt->vm_flags, mpnt->vm_file, |
44de9d0c | 422 | -vma_pages(mpnt)); |
1da177e4 LT |
423 | continue; |
424 | } | |
425 | charge = 0; | |
426 | if (mpnt->vm_flags & VM_ACCOUNT) { | |
b2412b7f HS |
427 | unsigned long len = vma_pages(mpnt); |
428 | ||
191c5424 | 429 | if (security_vm_enough_memory_mm(oldmm, len)) /* sic */ |
1da177e4 LT |
430 | goto fail_nomem; |
431 | charge = len; | |
432 | } | |
e94b1766 | 433 | tmp = kmem_cache_alloc(vm_area_cachep, GFP_KERNEL); |
1da177e4 LT |
434 | if (!tmp) |
435 | goto fail_nomem; | |
436 | *tmp = *mpnt; | |
5beb4930 | 437 | INIT_LIST_HEAD(&tmp->anon_vma_chain); |
846a16bf | 438 | pol = mpol_dup(vma_policy(mpnt)); |
1da177e4 LT |
439 | retval = PTR_ERR(pol); |
440 | if (IS_ERR(pol)) | |
441 | goto fail_nomem_policy; | |
442 | vma_set_policy(tmp, pol); | |
a247c3a9 | 443 | tmp->vm_mm = mm; |
5beb4930 RR |
444 | if (anon_vma_fork(tmp, mpnt)) |
445 | goto fail_nomem_anon_vma_fork; | |
1da177e4 | 446 | tmp->vm_flags &= ~VM_LOCKED; |
297c5eee | 447 | tmp->vm_next = tmp->vm_prev = NULL; |
1da177e4 LT |
448 | file = tmp->vm_file; |
449 | if (file) { | |
496ad9aa | 450 | struct inode *inode = file_inode(file); |
b88ed205 HD |
451 | struct address_space *mapping = file->f_mapping; |
452 | ||
1da177e4 LT |
453 | get_file(file); |
454 | if (tmp->vm_flags & VM_DENYWRITE) | |
455 | atomic_dec(&inode->i_writecount); | |
3d48ae45 | 456 | mutex_lock(&mapping->i_mmap_mutex); |
b88ed205 HD |
457 | if (tmp->vm_flags & VM_SHARED) |
458 | mapping->i_mmap_writable++; | |
b88ed205 HD |
459 | flush_dcache_mmap_lock(mapping); |
460 | /* insert tmp into the share list, just after mpnt */ | |
9826a516 ML |
461 | if (unlikely(tmp->vm_flags & VM_NONLINEAR)) |
462 | vma_nonlinear_insert(tmp, | |
463 | &mapping->i_mmap_nonlinear); | |
464 | else | |
465 | vma_interval_tree_insert_after(tmp, mpnt, | |
466 | &mapping->i_mmap); | |
b88ed205 | 467 | flush_dcache_mmap_unlock(mapping); |
3d48ae45 | 468 | mutex_unlock(&mapping->i_mmap_mutex); |
1da177e4 LT |
469 | } |
470 | ||
a1e78772 MG |
471 | /* |
472 | * Clear hugetlb-related page reserves for children. This only | |
473 | * affects MAP_PRIVATE mappings. Faults generated by the child | |
474 | * are not guaranteed to succeed, even if read-only | |
475 | */ | |
476 | if (is_vm_hugetlb_page(tmp)) | |
477 | reset_vma_resv_huge_pages(tmp); | |
478 | ||
1da177e4 | 479 | /* |
7ee78232 | 480 | * Link in the new vma and copy the page table entries. |
1da177e4 | 481 | */ |
1da177e4 LT |
482 | *pprev = tmp; |
483 | pprev = &tmp->vm_next; | |
297c5eee LT |
484 | tmp->vm_prev = prev; |
485 | prev = tmp; | |
1da177e4 LT |
486 | |
487 | __vma_link_rb(mm, tmp, rb_link, rb_parent); | |
488 | rb_link = &tmp->vm_rb.rb_right; | |
489 | rb_parent = &tmp->vm_rb; | |
490 | ||
491 | mm->map_count++; | |
0b0db14c | 492 | retval = copy_page_range(mm, oldmm, mpnt); |
1da177e4 LT |
493 | |
494 | if (tmp->vm_ops && tmp->vm_ops->open) | |
495 | tmp->vm_ops->open(tmp); | |
496 | ||
497 | if (retval) | |
498 | goto out; | |
499 | } | |
d6dd61c8 JF |
500 | /* a new mm has just been created */ |
501 | arch_dup_mmap(oldmm, mm); | |
1da177e4 | 502 | retval = 0; |
1da177e4 | 503 | out: |
7ee78232 | 504 | up_write(&mm->mmap_sem); |
fd3e42fc | 505 | flush_tlb_mm(oldmm); |
1da177e4 | 506 | up_write(&oldmm->mmap_sem); |
32cdba1e | 507 | uprobe_end_dup_mmap(); |
1da177e4 | 508 | return retval; |
5beb4930 RR |
509 | fail_nomem_anon_vma_fork: |
510 | mpol_put(pol); | |
1da177e4 LT |
511 | fail_nomem_policy: |
512 | kmem_cache_free(vm_area_cachep, tmp); | |
513 | fail_nomem: | |
514 | retval = -ENOMEM; | |
515 | vm_unacct_memory(charge); | |
516 | goto out; | |
517 | } | |
518 | ||
fb0a685c | 519 | static inline int mm_alloc_pgd(struct mm_struct *mm) |
1da177e4 LT |
520 | { |
521 | mm->pgd = pgd_alloc(mm); | |
522 | if (unlikely(!mm->pgd)) | |
523 | return -ENOMEM; | |
524 | return 0; | |
525 | } | |
526 | ||
fb0a685c | 527 | static inline void mm_free_pgd(struct mm_struct *mm) |
1da177e4 | 528 | { |
5e541973 | 529 | pgd_free(mm, mm->pgd); |
1da177e4 LT |
530 | } |
531 | #else | |
532 | #define dup_mmap(mm, oldmm) (0) | |
533 | #define mm_alloc_pgd(mm) (0) | |
534 | #define mm_free_pgd(mm) | |
535 | #endif /* CONFIG_MMU */ | |
536 | ||
23ff4440 | 537 | __cacheline_aligned_in_smp DEFINE_SPINLOCK(mmlist_lock); |
1da177e4 | 538 | |
e94b1766 | 539 | #define allocate_mm() (kmem_cache_alloc(mm_cachep, GFP_KERNEL)) |
1da177e4 LT |
540 | #define free_mm(mm) (kmem_cache_free(mm_cachep, (mm))) |
541 | ||
4cb0e11b HK |
542 | static unsigned long default_dump_filter = MMF_DUMP_FILTER_DEFAULT; |
543 | ||
544 | static int __init coredump_filter_setup(char *s) | |
545 | { | |
546 | default_dump_filter = | |
547 | (simple_strtoul(s, NULL, 0) << MMF_DUMP_FILTER_SHIFT) & | |
548 | MMF_DUMP_FILTER_MASK; | |
549 | return 1; | |
550 | } | |
551 | ||
552 | __setup("coredump_filter=", coredump_filter_setup); | |
553 | ||
1da177e4 LT |
554 | #include <linux/init_task.h> |
555 | ||
858f0993 AD |
556 | static void mm_init_aio(struct mm_struct *mm) |
557 | { | |
558 | #ifdef CONFIG_AIO | |
559 | spin_lock_init(&mm->ioctx_lock); | |
560 | INIT_HLIST_HEAD(&mm->ioctx_list); | |
561 | #endif | |
562 | } | |
563 | ||
fb0a685c | 564 | static struct mm_struct *mm_init(struct mm_struct *mm, struct task_struct *p) |
1da177e4 LT |
565 | { |
566 | atomic_set(&mm->mm_users, 1); | |
567 | atomic_set(&mm->mm_count, 1); | |
568 | init_rwsem(&mm->mmap_sem); | |
569 | INIT_LIST_HEAD(&mm->mmlist); | |
f8af4da3 HD |
570 | mm->flags = (current->mm) ? |
571 | (current->mm->flags & MMF_INIT_MASK) : default_dump_filter; | |
999d9fc1 | 572 | mm->core_state = NULL; |
1da177e4 | 573 | mm->nr_ptes = 0; |
d559db08 | 574 | memset(&mm->rss_stat, 0, sizeof(mm->rss_stat)); |
1da177e4 | 575 | spin_lock_init(&mm->page_table_lock); |
1da177e4 | 576 | mm->free_area_cache = TASK_UNMAPPED_BASE; |
1363c3cd | 577 | mm->cached_hole_size = ~0UL; |
858f0993 | 578 | mm_init_aio(mm); |
cf475ad2 | 579 | mm_init_owner(mm, p); |
d303cf46 | 580 | clear_tlb_flush_pending(mm); |
1da177e4 LT |
581 | |
582 | if (likely(!mm_alloc_pgd(mm))) { | |
583 | mm->def_flags = 0; | |
cddb8a5c | 584 | mmu_notifier_mm_init(mm); |
1da177e4 LT |
585 | return mm; |
586 | } | |
78fb7466 | 587 | |
1da177e4 LT |
588 | free_mm(mm); |
589 | return NULL; | |
590 | } | |
591 | ||
c3f0327f KK |
592 | static void check_mm(struct mm_struct *mm) |
593 | { | |
594 | int i; | |
595 | ||
596 | for (i = 0; i < NR_MM_COUNTERS; i++) { | |
597 | long x = atomic_long_read(&mm->rss_stat.count[i]); | |
598 | ||
599 | if (unlikely(x)) | |
600 | printk(KERN_ALERT "BUG: Bad rss-counter state " | |
601 | "mm:%p idx:%d val:%ld\n", mm, i, x); | |
602 | } | |
603 | ||
604 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE | |
605 | VM_BUG_ON(mm->pmd_huge_pte); | |
606 | #endif | |
607 | } | |
608 | ||
1da177e4 LT |
609 | /* |
610 | * Allocate and initialize an mm_struct. | |
611 | */ | |
fb0a685c | 612 | struct mm_struct *mm_alloc(void) |
1da177e4 | 613 | { |
fb0a685c | 614 | struct mm_struct *mm; |
1da177e4 LT |
615 | |
616 | mm = allocate_mm(); | |
de03c72c KM |
617 | if (!mm) |
618 | return NULL; | |
619 | ||
620 | memset(mm, 0, sizeof(*mm)); | |
6345d24d LT |
621 | mm_init_cpumask(mm); |
622 | return mm_init(mm, current); | |
1da177e4 LT |
623 | } |
624 | ||
625 | /* | |
626 | * Called when the last reference to the mm | |
627 | * is dropped: either by a lazy thread or by | |
628 | * mmput. Free the page directory and the mm. | |
629 | */ | |
7ad5b3a5 | 630 | void __mmdrop(struct mm_struct *mm) |
1da177e4 LT |
631 | { |
632 | BUG_ON(mm == &init_mm); | |
633 | mm_free_pgd(mm); | |
634 | destroy_context(mm); | |
cddb8a5c | 635 | mmu_notifier_mm_destroy(mm); |
c3f0327f | 636 | check_mm(mm); |
1da177e4 LT |
637 | free_mm(mm); |
638 | } | |
6d4e4c4f | 639 | EXPORT_SYMBOL_GPL(__mmdrop); |
1da177e4 LT |
640 | |
641 | /* | |
642 | * Decrement the use count and release all resources for an mm. | |
643 | */ | |
644 | void mmput(struct mm_struct *mm) | |
645 | { | |
0ae26f1b AM |
646 | might_sleep(); |
647 | ||
1da177e4 | 648 | if (atomic_dec_and_test(&mm->mm_users)) { |
d4b3b638 | 649 | uprobe_clear_state(mm); |
1da177e4 | 650 | exit_aio(mm); |
1c2fb7a4 | 651 | ksm_exit(mm); |
ba76149f | 652 | khugepaged_exit(mm); /* must run before exit_mmap */ |
1da177e4 | 653 | exit_mmap(mm); |
925d1c40 | 654 | set_mm_exe_file(mm, NULL); |
1da177e4 LT |
655 | if (!list_empty(&mm->mmlist)) { |
656 | spin_lock(&mmlist_lock); | |
657 | list_del(&mm->mmlist); | |
658 | spin_unlock(&mmlist_lock); | |
659 | } | |
801460d0 HS |
660 | if (mm->binfmt) |
661 | module_put(mm->binfmt->module); | |
1da177e4 LT |
662 | mmdrop(mm); |
663 | } | |
664 | } | |
665 | EXPORT_SYMBOL_GPL(mmput); | |
666 | ||
38646013 JS |
667 | void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file) |
668 | { | |
669 | if (new_exe_file) | |
670 | get_file(new_exe_file); | |
671 | if (mm->exe_file) | |
672 | fput(mm->exe_file); | |
673 | mm->exe_file = new_exe_file; | |
38646013 JS |
674 | } |
675 | ||
676 | struct file *get_mm_exe_file(struct mm_struct *mm) | |
677 | { | |
678 | struct file *exe_file; | |
679 | ||
2dd8ad81 | 680 | /* We need mmap_sem to protect against races with removal of exe_file */ |
38646013 JS |
681 | down_read(&mm->mmap_sem); |
682 | exe_file = mm->exe_file; | |
683 | if (exe_file) | |
684 | get_file(exe_file); | |
685 | up_read(&mm->mmap_sem); | |
686 | return exe_file; | |
687 | } | |
688 | ||
689 | static void dup_mm_exe_file(struct mm_struct *oldmm, struct mm_struct *newmm) | |
690 | { | |
691 | /* It's safe to write the exe_file pointer without exe_file_lock because | |
692 | * this is called during fork when the task is not yet in /proc */ | |
693 | newmm->exe_file = get_mm_exe_file(oldmm); | |
694 | } | |
695 | ||
1da177e4 LT |
696 | /** |
697 | * get_task_mm - acquire a reference to the task's mm | |
698 | * | |
246bb0b1 | 699 | * Returns %NULL if the task has no mm. Checks PF_KTHREAD (meaning |
1da177e4 LT |
700 | * this kernel workthread has transiently adopted a user mm with use_mm, |
701 | * to do its AIO) is not set and if so returns a reference to it, after | |
702 | * bumping up the use count. User must release the mm via mmput() | |
703 | * after use. Typically used by /proc and ptrace. | |
704 | */ | |
705 | struct mm_struct *get_task_mm(struct task_struct *task) | |
706 | { | |
707 | struct mm_struct *mm; | |
708 | ||
709 | task_lock(task); | |
710 | mm = task->mm; | |
711 | if (mm) { | |
246bb0b1 | 712 | if (task->flags & PF_KTHREAD) |
1da177e4 LT |
713 | mm = NULL; |
714 | else | |
715 | atomic_inc(&mm->mm_users); | |
716 | } | |
717 | task_unlock(task); | |
718 | return mm; | |
719 | } | |
720 | EXPORT_SYMBOL_GPL(get_task_mm); | |
721 | ||
8cdb878d CY |
722 | struct mm_struct *mm_access(struct task_struct *task, unsigned int mode) |
723 | { | |
724 | struct mm_struct *mm; | |
725 | int err; | |
726 | ||
727 | err = mutex_lock_killable(&task->signal->cred_guard_mutex); | |
728 | if (err) | |
729 | return ERR_PTR(err); | |
730 | ||
731 | mm = get_task_mm(task); | |
732 | if (mm && mm != current->mm && | |
6fa3eb70 S |
733 | !ptrace_may_access(task, mode) && |
734 | !capable(CAP_SYS_RESOURCE)) { | |
8cdb878d CY |
735 | mmput(mm); |
736 | mm = ERR_PTR(-EACCES); | |
737 | } | |
738 | mutex_unlock(&task->signal->cred_guard_mutex); | |
739 | ||
740 | return mm; | |
741 | } | |
742 | ||
57b59c4a | 743 | static void complete_vfork_done(struct task_struct *tsk) |
c415c3b4 | 744 | { |
d68b46fe | 745 | struct completion *vfork; |
c415c3b4 | 746 | |
d68b46fe ON |
747 | task_lock(tsk); |
748 | vfork = tsk->vfork_done; | |
749 | if (likely(vfork)) { | |
750 | tsk->vfork_done = NULL; | |
751 | complete(vfork); | |
752 | } | |
753 | task_unlock(tsk); | |
754 | } | |
755 | ||
756 | static int wait_for_vfork_done(struct task_struct *child, | |
757 | struct completion *vfork) | |
758 | { | |
759 | int killed; | |
760 | ||
761 | freezer_do_not_count(); | |
762 | killed = wait_for_completion_killable(vfork); | |
763 | freezer_count(); | |
764 | ||
765 | if (killed) { | |
766 | task_lock(child); | |
767 | child->vfork_done = NULL; | |
768 | task_unlock(child); | |
769 | } | |
770 | ||
771 | put_task_struct(child); | |
772 | return killed; | |
c415c3b4 ON |
773 | } |
774 | ||
1da177e4 LT |
775 | /* Please note the differences between mmput and mm_release. |
776 | * mmput is called whenever we stop holding onto a mm_struct, | |
777 | * error success whatever. | |
778 | * | |
779 | * mm_release is called after a mm_struct has been removed | |
780 | * from the current process. | |
781 | * | |
782 | * This difference is important for error handling, when we | |
783 | * only half set up a mm_struct for a new process and need to restore | |
784 | * the old one. Because we mmput the new mm_struct before | |
785 | * restoring the old one. . . | |
786 | * Eric Biederman 10 January 1998 | |
787 | */ | |
788 | void mm_release(struct task_struct *tsk, struct mm_struct *mm) | |
789 | { | |
8141c7f3 LT |
790 | /* Get rid of any futexes when releasing the mm */ |
791 | #ifdef CONFIG_FUTEX | |
fc6b177d | 792 | if (unlikely(tsk->robust_list)) { |
8141c7f3 | 793 | exit_robust_list(tsk); |
fc6b177d PZ |
794 | tsk->robust_list = NULL; |
795 | } | |
8141c7f3 | 796 | #ifdef CONFIG_COMPAT |
fc6b177d | 797 | if (unlikely(tsk->compat_robust_list)) { |
8141c7f3 | 798 | compat_exit_robust_list(tsk); |
fc6b177d PZ |
799 | tsk->compat_robust_list = NULL; |
800 | } | |
8141c7f3 | 801 | #endif |
322a2c10 TG |
802 | if (unlikely(!list_empty(&tsk->pi_state_list))) |
803 | exit_pi_state_list(tsk); | |
8141c7f3 LT |
804 | #endif |
805 | ||
0326f5a9 SD |
806 | uprobe_free_utask(tsk); |
807 | ||
1da177e4 LT |
808 | /* Get rid of any cached register state */ |
809 | deactivate_mm(tsk, mm); | |
810 | ||
fec1d011 | 811 | /* |
9c559b0f MH |
812 | * Signal userspace if we're not exiting with a core dump |
813 | * because we want to leave the value intact for debugging | |
814 | * purposes. | |
fec1d011 | 815 | */ |
9c8a8228 | 816 | if (tsk->clear_child_tid) { |
9c559b0f | 817 | if (!(tsk->signal->flags & SIGNAL_GROUP_COREDUMP) && |
9c8a8228 ED |
818 | atomic_read(&mm->mm_users) > 1) { |
819 | /* | |
820 | * We don't check the error code - if userspace has | |
821 | * not set up a proper pointer then tough luck. | |
822 | */ | |
823 | put_user(0, tsk->clear_child_tid); | |
824 | sys_futex(tsk->clear_child_tid, FUTEX_WAKE, | |
825 | 1, NULL, NULL, 0); | |
826 | } | |
1da177e4 | 827 | tsk->clear_child_tid = NULL; |
1da177e4 | 828 | } |
f7505d64 KK |
829 | |
830 | /* | |
831 | * All done, finally we can wake up parent and return this mm to him. | |
832 | * Also kthread_stop() uses this completion for synchronization. | |
833 | */ | |
834 | if (tsk->vfork_done) | |
835 | complete_vfork_done(tsk); | |
1da177e4 LT |
836 | } |
837 | ||
a0a7ec30 JD |
838 | /* |
839 | * Allocate a new mm structure and copy contents from the | |
840 | * mm structure of the passed in task structure. | |
841 | */ | |
402b0862 | 842 | struct mm_struct *dup_mm(struct task_struct *tsk) |
a0a7ec30 JD |
843 | { |
844 | struct mm_struct *mm, *oldmm = current->mm; | |
845 | int err; | |
846 | ||
847 | if (!oldmm) | |
848 | return NULL; | |
849 | ||
850 | mm = allocate_mm(); | |
851 | if (!mm) | |
852 | goto fail_nomem; | |
853 | ||
854 | memcpy(mm, oldmm, sizeof(*mm)); | |
6345d24d | 855 | mm_init_cpumask(mm); |
a0a7ec30 | 856 | |
e7a00c45 AA |
857 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE |
858 | mm->pmd_huge_pte = NULL; | |
5bca2303 MG |
859 | #endif |
860 | #ifdef CONFIG_NUMA_BALANCING | |
861 | mm->first_nid = NUMA_PTE_SCAN_INIT; | |
e7a00c45 | 862 | #endif |
78fb7466 | 863 | if (!mm_init(mm, tsk)) |
a0a7ec30 JD |
864 | goto fail_nomem; |
865 | ||
866 | if (init_new_context(tsk, mm)) | |
867 | goto fail_nocontext; | |
868 | ||
925d1c40 MH |
869 | dup_mm_exe_file(oldmm, mm); |
870 | ||
a0a7ec30 JD |
871 | err = dup_mmap(mm, oldmm); |
872 | if (err) | |
873 | goto free_pt; | |
874 | ||
875 | mm->hiwater_rss = get_mm_rss(mm); | |
876 | mm->hiwater_vm = mm->total_vm; | |
877 | ||
801460d0 HS |
878 | if (mm->binfmt && !try_module_get(mm->binfmt->module)) |
879 | goto free_pt; | |
880 | ||
a0a7ec30 JD |
881 | return mm; |
882 | ||
883 | free_pt: | |
801460d0 HS |
884 | /* don't put binfmt in mmput, we haven't got module yet */ |
885 | mm->binfmt = NULL; | |
a0a7ec30 JD |
886 | mmput(mm); |
887 | ||
888 | fail_nomem: | |
889 | return NULL; | |
890 | ||
891 | fail_nocontext: | |
892 | /* | |
893 | * If init_new_context() failed, we cannot use mmput() to free the mm | |
894 | * because it calls destroy_context() | |
895 | */ | |
896 | mm_free_pgd(mm); | |
897 | free_mm(mm); | |
898 | return NULL; | |
899 | } | |
900 | ||
fb0a685c | 901 | static int copy_mm(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 | 902 | { |
fb0a685c | 903 | struct mm_struct *mm, *oldmm; |
1da177e4 LT |
904 | int retval; |
905 | ||
906 | tsk->min_flt = tsk->maj_flt = 0; | |
6fa3eb70 S |
907 | #ifdef CONFIG_ZRAM |
908 | tsk->fm_flt = tsk->swap_in = tsk->swap_out = 0; | |
909 | #endif | |
1da177e4 | 910 | tsk->nvcsw = tsk->nivcsw = 0; |
17406b82 MSB |
911 | #ifdef CONFIG_DETECT_HUNG_TASK |
912 | tsk->last_switch_count = tsk->nvcsw + tsk->nivcsw; | |
913 | #endif | |
1da177e4 LT |
914 | |
915 | tsk->mm = NULL; | |
916 | tsk->active_mm = NULL; | |
917 | ||
918 | /* | |
919 | * Are we cloning a kernel thread? | |
920 | * | |
921 | * We need to steal a active VM for that.. | |
922 | */ | |
923 | oldmm = current->mm; | |
924 | if (!oldmm) | |
925 | return 0; | |
926 | ||
927 | if (clone_flags & CLONE_VM) { | |
928 | atomic_inc(&oldmm->mm_users); | |
929 | mm = oldmm; | |
1da177e4 LT |
930 | goto good_mm; |
931 | } | |
932 | ||
933 | retval = -ENOMEM; | |
a0a7ec30 | 934 | mm = dup_mm(tsk); |
1da177e4 LT |
935 | if (!mm) |
936 | goto fail_nomem; | |
937 | ||
1da177e4 LT |
938 | good_mm: |
939 | tsk->mm = mm; | |
940 | tsk->active_mm = mm; | |
941 | return 0; | |
942 | ||
1da177e4 LT |
943 | fail_nomem: |
944 | return retval; | |
1da177e4 LT |
945 | } |
946 | ||
a39bc516 | 947 | static int copy_fs(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 | 948 | { |
498052bb | 949 | struct fs_struct *fs = current->fs; |
1da177e4 | 950 | if (clone_flags & CLONE_FS) { |
498052bb | 951 | /* tsk->fs is already what we want */ |
2a4419b5 | 952 | spin_lock(&fs->lock); |
498052bb | 953 | if (fs->in_exec) { |
2a4419b5 | 954 | spin_unlock(&fs->lock); |
498052bb AV |
955 | return -EAGAIN; |
956 | } | |
957 | fs->users++; | |
2a4419b5 | 958 | spin_unlock(&fs->lock); |
1da177e4 LT |
959 | return 0; |
960 | } | |
498052bb | 961 | tsk->fs = copy_fs_struct(fs); |
1da177e4 LT |
962 | if (!tsk->fs) |
963 | return -ENOMEM; | |
964 | return 0; | |
965 | } | |
966 | ||
fb0a685c | 967 | static int copy_files(unsigned long clone_flags, struct task_struct *tsk) |
a016f338 JD |
968 | { |
969 | struct files_struct *oldf, *newf; | |
970 | int error = 0; | |
971 | ||
972 | /* | |
973 | * A background process may not have any files ... | |
974 | */ | |
975 | oldf = current->files; | |
976 | if (!oldf) | |
977 | goto out; | |
978 | ||
979 | if (clone_flags & CLONE_FILES) { | |
980 | atomic_inc(&oldf->count); | |
981 | goto out; | |
982 | } | |
983 | ||
a016f338 JD |
984 | newf = dup_fd(oldf, &error); |
985 | if (!newf) | |
986 | goto out; | |
987 | ||
988 | tsk->files = newf; | |
989 | error = 0; | |
990 | out: | |
991 | return error; | |
992 | } | |
993 | ||
fadad878 | 994 | static int copy_io(unsigned long clone_flags, struct task_struct *tsk) |
fd0928df JA |
995 | { |
996 | #ifdef CONFIG_BLOCK | |
997 | struct io_context *ioc = current->io_context; | |
6e736be7 | 998 | struct io_context *new_ioc; |
fd0928df JA |
999 | |
1000 | if (!ioc) | |
1001 | return 0; | |
fadad878 JA |
1002 | /* |
1003 | * Share io context with parent, if CLONE_IO is set | |
1004 | */ | |
1005 | if (clone_flags & CLONE_IO) { | |
3d48749d TH |
1006 | ioc_task_link(ioc); |
1007 | tsk->io_context = ioc; | |
fadad878 | 1008 | } else if (ioprio_valid(ioc->ioprio)) { |
6e736be7 TH |
1009 | new_ioc = get_task_io_context(tsk, GFP_KERNEL, NUMA_NO_NODE); |
1010 | if (unlikely(!new_ioc)) | |
fd0928df JA |
1011 | return -ENOMEM; |
1012 | ||
6e736be7 | 1013 | new_ioc->ioprio = ioc->ioprio; |
11a3122f | 1014 | put_io_context(new_ioc); |
fd0928df JA |
1015 | } |
1016 | #endif | |
1017 | return 0; | |
1018 | } | |
1019 | ||
a39bc516 | 1020 | static int copy_sighand(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 LT |
1021 | { |
1022 | struct sighand_struct *sig; | |
1023 | ||
60348802 | 1024 | if (clone_flags & CLONE_SIGHAND) { |
1da177e4 LT |
1025 | atomic_inc(¤t->sighand->count); |
1026 | return 0; | |
1027 | } | |
1028 | sig = kmem_cache_alloc(sighand_cachep, GFP_KERNEL); | |
e56d0903 | 1029 | rcu_assign_pointer(tsk->sighand, sig); |
1da177e4 LT |
1030 | if (!sig) |
1031 | return -ENOMEM; | |
1da177e4 LT |
1032 | atomic_set(&sig->count, 1); |
1033 | memcpy(sig->action, current->sighand->action, sizeof(sig->action)); | |
1034 | return 0; | |
1035 | } | |
1036 | ||
a7e5328a | 1037 | void __cleanup_sighand(struct sighand_struct *sighand) |
c81addc9 | 1038 | { |
d80e731e ON |
1039 | if (atomic_dec_and_test(&sighand->count)) { |
1040 | signalfd_cleanup(sighand); | |
c81addc9 | 1041 | kmem_cache_free(sighand_cachep, sighand); |
d80e731e | 1042 | } |
c81addc9 ON |
1043 | } |
1044 | ||
f06febc9 FM |
1045 | |
1046 | /* | |
1047 | * Initialize POSIX timer handling for a thread group. | |
1048 | */ | |
1049 | static void posix_cpu_timers_init_group(struct signal_struct *sig) | |
1050 | { | |
78d7d407 JS |
1051 | unsigned long cpu_limit; |
1052 | ||
f06febc9 FM |
1053 | /* Thread group counters. */ |
1054 | thread_group_cputime_init(sig); | |
1055 | ||
78d7d407 JS |
1056 | cpu_limit = ACCESS_ONCE(sig->rlim[RLIMIT_CPU].rlim_cur); |
1057 | if (cpu_limit != RLIM_INFINITY) { | |
1058 | sig->cputime_expires.prof_exp = secs_to_cputime(cpu_limit); | |
6279a751 ON |
1059 | sig->cputimer.running = 1; |
1060 | } | |
1061 | ||
f06febc9 FM |
1062 | /* The timer lists. */ |
1063 | INIT_LIST_HEAD(&sig->cpu_timers[0]); | |
1064 | INIT_LIST_HEAD(&sig->cpu_timers[1]); | |
1065 | INIT_LIST_HEAD(&sig->cpu_timers[2]); | |
1066 | } | |
1067 | ||
a39bc516 | 1068 | static int copy_signal(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 LT |
1069 | { |
1070 | struct signal_struct *sig; | |
1da177e4 | 1071 | |
4ab6c083 | 1072 | if (clone_flags & CLONE_THREAD) |
490dea45 | 1073 | return 0; |
490dea45 | 1074 | |
a56704ef | 1075 | sig = kmem_cache_zalloc(signal_cachep, GFP_KERNEL); |
1da177e4 LT |
1076 | tsk->signal = sig; |
1077 | if (!sig) | |
1078 | return -ENOMEM; | |
1079 | ||
b3ac022c | 1080 | sig->nr_threads = 1; |
1da177e4 | 1081 | atomic_set(&sig->live, 1); |
b3ac022c | 1082 | atomic_set(&sig->sigcnt, 1); |
6fa3eb70 S |
1083 | |
1084 | /* list_add(thread_node, thread_head) without INIT_LIST_HEAD() */ | |
1085 | sig->thread_head = (struct list_head)LIST_HEAD_INIT(tsk->thread_node); | |
1086 | tsk->thread_node = (struct list_head)LIST_HEAD_INIT(sig->thread_head); | |
1087 | ||
1da177e4 | 1088 | init_waitqueue_head(&sig->wait_chldexit); |
db51aecc | 1089 | sig->curr_target = tsk; |
1da177e4 LT |
1090 | init_sigpending(&sig->shared_pending); |
1091 | INIT_LIST_HEAD(&sig->posix_timers); | |
1092 | ||
c9cb2e3d | 1093 | hrtimer_init(&sig->real_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL); |
1da177e4 | 1094 | sig->real_timer.function = it_real_fn; |
1da177e4 | 1095 | |
1da177e4 LT |
1096 | task_lock(current->group_leader); |
1097 | memcpy(sig->rlim, current->signal->rlim, sizeof sig->rlim); | |
1098 | task_unlock(current->group_leader); | |
1099 | ||
6279a751 ON |
1100 | posix_cpu_timers_init_group(sig); |
1101 | ||
522ed776 | 1102 | tty_audit_fork(sig); |
5091faa4 | 1103 | sched_autogroup_fork(sig); |
522ed776 | 1104 | |
4714d1d3 | 1105 | #ifdef CONFIG_CGROUPS |
257058ae | 1106 | init_rwsem(&sig->group_rwsem); |
4714d1d3 BB |
1107 | #endif |
1108 | ||
a63d83f4 | 1109 | sig->oom_score_adj = current->signal->oom_score_adj; |
dabb16f6 | 1110 | sig->oom_score_adj_min = current->signal->oom_score_adj_min; |
28b83c51 | 1111 | |
ebec18a6 LP |
1112 | sig->has_child_subreaper = current->signal->has_child_subreaper || |
1113 | current->signal->is_child_subreaper; | |
1114 | ||
9b1bf12d KM |
1115 | mutex_init(&sig->cred_guard_mutex); |
1116 | ||
1da177e4 LT |
1117 | return 0; |
1118 | } | |
1119 | ||
a39bc516 | 1120 | static void copy_flags(unsigned long clone_flags, struct task_struct *p) |
1da177e4 LT |
1121 | { |
1122 | unsigned long new_flags = p->flags; | |
1123 | ||
21aa9af0 | 1124 | new_flags &= ~(PF_SUPERPRIV | PF_WQ_WORKER); |
1da177e4 | 1125 | new_flags |= PF_FORKNOEXEC; |
1da177e4 LT |
1126 | p->flags = new_flags; |
1127 | } | |
1128 | ||
6fa3eb70 S |
1129 | static void copy_seccomp(struct task_struct *p) |
1130 | { | |
1131 | #ifdef CONFIG_SECCOMP | |
1132 | /* | |
1133 | * Must be called with sighand->lock held, which is common to | |
1134 | * all threads in the group. Holding cred_guard_mutex is not | |
1135 | * needed because this new task is not yet running and cannot | |
1136 | * be racing exec. | |
1137 | */ | |
1138 | assert_spin_locked(¤t->sighand->siglock); | |
1139 | ||
1140 | /* Ref-count the new filter user, and assign it. */ | |
1141 | get_seccomp_filter(current); | |
1142 | p->seccomp = current->seccomp; | |
1143 | ||
1144 | /* | |
1145 | * Explicitly enable no_new_privs here in case it got set | |
1146 | * between the task_struct being duplicated and holding the | |
1147 | * sighand lock. The seccomp state and nnp must be in sync. | |
1148 | */ | |
1149 | if (task_no_new_privs(current)) | |
1150 | task_set_no_new_privs(p); | |
1151 | ||
1152 | /* | |
1153 | * If the parent gained a seccomp mode after copying thread | |
1154 | * flags and between before we held the sighand lock, we have | |
1155 | * to manually enable the seccomp thread flag here. | |
1156 | */ | |
1157 | if (p->seccomp.mode != SECCOMP_MODE_DISABLED) | |
1158 | set_tsk_thread_flag(p, TIF_SECCOMP); | |
1159 | #endif | |
1160 | } | |
1161 | ||
17da2bd9 | 1162 | SYSCALL_DEFINE1(set_tid_address, int __user *, tidptr) |
1da177e4 LT |
1163 | { |
1164 | current->clear_child_tid = tidptr; | |
1165 | ||
b488893a | 1166 | return task_pid_vnr(current); |
1da177e4 LT |
1167 | } |
1168 | ||
a39bc516 | 1169 | static void rt_mutex_init_task(struct task_struct *p) |
23f78d4a | 1170 | { |
1d615482 | 1171 | raw_spin_lock_init(&p->pi_lock); |
e29e175b | 1172 | #ifdef CONFIG_RT_MUTEXES |
732375c6 | 1173 | plist_head_init(&p->pi_waiters); |
23f78d4a | 1174 | p->pi_blocked_on = NULL; |
23f78d4a IM |
1175 | #endif |
1176 | } | |
1177 | ||
cf475ad2 BS |
1178 | #ifdef CONFIG_MM_OWNER |
1179 | void mm_init_owner(struct mm_struct *mm, struct task_struct *p) | |
1180 | { | |
1181 | mm->owner = p; | |
1182 | } | |
1183 | #endif /* CONFIG_MM_OWNER */ | |
1184 | ||
f06febc9 FM |
1185 | /* |
1186 | * Initialize POSIX timer handling for a single task. | |
1187 | */ | |
1188 | static void posix_cpu_timers_init(struct task_struct *tsk) | |
1189 | { | |
64861634 MS |
1190 | tsk->cputime_expires.prof_exp = 0; |
1191 | tsk->cputime_expires.virt_exp = 0; | |
f06febc9 FM |
1192 | tsk->cputime_expires.sched_exp = 0; |
1193 | INIT_LIST_HEAD(&tsk->cpu_timers[0]); | |
1194 | INIT_LIST_HEAD(&tsk->cpu_timers[1]); | |
1195 | INIT_LIST_HEAD(&tsk->cpu_timers[2]); | |
1196 | } | |
1197 | ||
6fa3eb70 S |
1198 | #ifdef CONFIG_MTK_SCHED_CMP_TGS |
1199 | static void mt_init_thread_group(struct task_struct *p){ | |
1200 | #ifdef CONFIG_MT_SCHED_INFO | |
1201 | struct task_struct *tg = p->group_leader; | |
1202 | #endif | |
1203 | ||
1204 | p->thread_group_info[0].cfs_nr_running = 0; | |
1205 | p->thread_group_info[0].nr_running = 0 ; | |
1206 | p->thread_group_info[0].load_avg_ratio = 0; | |
1207 | p->thread_group_info[1].cfs_nr_running = 0; | |
1208 | p->thread_group_info[1].nr_running = 0; | |
1209 | p->thread_group_info[1].load_avg_ratio = 0; | |
1210 | ||
1211 | #ifdef CONFIG_MT_SCHED_INFO | |
1212 | mt_sched_printf("fork %d:%s %d:%s %lu %lu %lu, %lu %lu %lu", | |
1213 | tg->pid, tg->comm, p->pid, p->comm, | |
1214 | tg->thread_group_info[0].nr_running, | |
1215 | tg->thread_group_info[0].cfs_nr_running, | |
1216 | tg->thread_group_info[0].load_avg_ratio, | |
1217 | tg->thread_group_info[1].cfs_nr_running, | |
1218 | tg->thread_group_info[1].nr_running, | |
1219 | tg->thread_group_info[1].load_avg_ratio); | |
1220 | #endif | |
1221 | } | |
1222 | #endif | |
1223 | ||
1da177e4 LT |
1224 | /* |
1225 | * This creates a new process as a copy of the old one, | |
1226 | * but does not actually start it yet. | |
1227 | * | |
1228 | * It copies the registers, and all the appropriate | |
1229 | * parts of the process environment (as per the clone | |
1230 | * flags). The actual kick-off is left to the caller. | |
1231 | */ | |
36c8b586 IM |
1232 | static struct task_struct *copy_process(unsigned long clone_flags, |
1233 | unsigned long stack_start, | |
36c8b586 | 1234 | unsigned long stack_size, |
36c8b586 | 1235 | int __user *child_tidptr, |
09a05394 RM |
1236 | struct pid *pid, |
1237 | int trace) | |
1da177e4 LT |
1238 | { |
1239 | int retval; | |
a24efe62 | 1240 | struct task_struct *p; |
1da177e4 | 1241 | |
6fa3eb70 S |
1242 | if ((clone_flags & (CLONE_NEWNS|CLONE_FS)) == (CLONE_NEWNS|CLONE_FS)){ |
1243 | printk("[%d:%s] fork fail at cpp 1, clone_flags:0x%x\n", current->pid, current->comm, (unsigned int)clone_flags); | |
1da177e4 | 1244 | return ERR_PTR(-EINVAL); |
6fa3eb70 | 1245 | } |
e66eded8 EB |
1246 | if ((clone_flags & (CLONE_NEWUSER|CLONE_FS)) == (CLONE_NEWUSER|CLONE_FS)) |
1247 | return ERR_PTR(-EINVAL); | |
1248 | ||
1da177e4 LT |
1249 | /* |
1250 | * Thread groups must share signals as well, and detached threads | |
1251 | * can only be started up within the thread group. | |
1252 | */ | |
6fa3eb70 S |
1253 | if ((clone_flags & CLONE_THREAD) && !(clone_flags & CLONE_SIGHAND)){ |
1254 | printk("[%d:%s] fork fail at cpp 2, clone_flags:0x%x\n", current->pid, current->comm, (unsigned int)clone_flags); | |
1da177e4 | 1255 | return ERR_PTR(-EINVAL); |
6fa3eb70 | 1256 | } |
1da177e4 LT |
1257 | /* |
1258 | * Shared signal handlers imply shared VM. By way of the above, | |
1259 | * thread groups also imply shared VM. Blocking this case allows | |
1260 | * for various simplifications in other code. | |
1261 | */ | |
6fa3eb70 S |
1262 | if ((clone_flags & CLONE_SIGHAND) && !(clone_flags & CLONE_VM)){ |
1263 | printk("[%d:%s] fork fail at cpp 3, clone_flags:0x%x\n", current->pid, current->comm, (unsigned int)clone_flags); | |
1da177e4 | 1264 | return ERR_PTR(-EINVAL); |
6fa3eb70 | 1265 | } |
123be07b SB |
1266 | /* |
1267 | * Siblings of global init remain as zombies on exit since they are | |
1268 | * not reaped by their parent (swapper). To solve this and to avoid | |
1269 | * multi-rooted process trees, prevent global and container-inits | |
1270 | * from creating siblings. | |
1271 | */ | |
1272 | if ((clone_flags & CLONE_PARENT) && | |
6fa3eb70 S |
1273 | current->signal->flags & SIGNAL_UNKILLABLE){ |
1274 | printk("[%d:%s] fork fail at cpp 4, clone_flags:0x%x\n", current->pid, current->comm, (unsigned int)clone_flags); | |
123be07b | 1275 | return ERR_PTR(-EINVAL); |
6fa3eb70 | 1276 | } |
8382fcac | 1277 | /* |
f608ebd7 ON |
1278 | * If the new process will be in a different pid namespace don't |
1279 | * allow it to share a thread group or signal handlers with the | |
1280 | * forking task. | |
8382fcac | 1281 | */ |
f608ebd7 | 1282 | if ((clone_flags & (CLONE_SIGHAND | CLONE_NEWPID)) && |
8382fcac EB |
1283 | (task_active_pid_ns(current) != current->nsproxy->pid_ns)) |
1284 | return ERR_PTR(-EINVAL); | |
1285 | ||
1da177e4 LT |
1286 | retval = security_task_create(clone_flags); |
1287 | if (retval) | |
1288 | goto fork_out; | |
1289 | ||
1290 | retval = -ENOMEM; | |
1291 | p = dup_task_struct(current); | |
6fa3eb70 S |
1292 | if (!p){ |
1293 | printk("[%d:%s] fork fail at dup_task_struc, p=%p\n", current->pid, current->comm, p); | |
1da177e4 | 1294 | goto fork_out; |
6fa3eb70 | 1295 | } |
1da177e4 | 1296 | |
f7e8b616 SR |
1297 | ftrace_graph_init_task(p); |
1298 | ||
bea493a0 | 1299 | rt_mutex_init_task(p); |
6fa3eb70 S |
1300 | #ifdef CONFIG_MTK_SCHED_CMP_TGS |
1301 | raw_spin_lock_init(&p->thread_group_info_lock); | |
1302 | #endif | |
bea493a0 | 1303 | |
d12c1a37 | 1304 | #ifdef CONFIG_PROVE_LOCKING |
de30a2b3 IM |
1305 | DEBUG_LOCKS_WARN_ON(!p->hardirqs_enabled); |
1306 | DEBUG_LOCKS_WARN_ON(!p->softirqs_enabled); | |
1307 | #endif | |
1da177e4 | 1308 | retval = -EAGAIN; |
3b11a1de | 1309 | if (atomic_read(&p->real_cred->user->processes) >= |
78d7d407 | 1310 | task_rlimit(p, RLIMIT_NPROC)) { |
1da177e4 | 1311 | if (!capable(CAP_SYS_ADMIN) && !capable(CAP_SYS_RESOURCE) && |
18b6e041 | 1312 | p->real_cred->user != INIT_USER) |
1da177e4 LT |
1313 | goto bad_fork_free; |
1314 | } | |
72fa5997 | 1315 | current->flags &= ~PF_NPROC_EXCEEDED; |
1da177e4 | 1316 | |
f1752eec DH |
1317 | retval = copy_creds(p, clone_flags); |
1318 | if (retval < 0) | |
1319 | goto bad_fork_free; | |
1da177e4 LT |
1320 | |
1321 | /* | |
1322 | * If multiple threads are within copy_process(), then this check | |
1323 | * triggers too late. This doesn't hurt, the check is only there | |
1324 | * to stop root fork bombs. | |
1325 | */ | |
04ec93fe | 1326 | retval = -EAGAIN; |
1da177e4 LT |
1327 | if (nr_threads >= max_threads) |
1328 | goto bad_fork_cleanup_count; | |
1329 | ||
a1261f54 | 1330 | if (!try_module_get(task_thread_info(p)->exec_domain->module)) |
1da177e4 LT |
1331 | goto bad_fork_cleanup_count; |
1332 | ||
1da177e4 | 1333 | p->did_exec = 0; |
ca74e92b | 1334 | delayacct_tsk_init(p); /* Must remain after dup_task_struct() */ |
1da177e4 | 1335 | copy_flags(clone_flags, p); |
1da177e4 LT |
1336 | INIT_LIST_HEAD(&p->children); |
1337 | INIT_LIST_HEAD(&p->sibling); | |
f41d911f | 1338 | rcu_copy_process(p); |
1da177e4 LT |
1339 | p->vfork_done = NULL; |
1340 | spin_lock_init(&p->alloc_lock); | |
1da177e4 | 1341 | |
1da177e4 LT |
1342 | init_sigpending(&p->pending); |
1343 | ||
64861634 MS |
1344 | p->utime = p->stime = p->gtime = 0; |
1345 | p->utimescaled = p->stimescaled = 0; | |
9fbc42ea | 1346 | #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE |
d37f761d | 1347 | p->prev_cputime.utime = p->prev_cputime.stime = 0; |
d99ca3b9 | 1348 | #endif |
6a61671b FW |
1349 | #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN |
1350 | seqlock_init(&p->vtime_seqlock); | |
1351 | p->vtime_snap = 0; | |
1352 | p->vtime_snap_whence = VTIME_SLEEPING; | |
1353 | #endif | |
1354 | ||
a3a2e76c KH |
1355 | #if defined(SPLIT_RSS_COUNTING) |
1356 | memset(&p->rss_stat, 0, sizeof(p->rss_stat)); | |
1357 | #endif | |
172ba844 | 1358 | |
6976675d AV |
1359 | p->default_timer_slack_ns = current->timer_slack_ns; |
1360 | ||
5995477a | 1361 | task_io_accounting_init(&p->ioac); |
1da177e4 LT |
1362 | acct_clear_integrals(p); |
1363 | ||
f06febc9 | 1364 | posix_cpu_timers_init(p); |
1da177e4 | 1365 | |
1da177e4 | 1366 | do_posix_clock_monotonic_gettime(&p->start_time); |
924b42d5 TJ |
1367 | p->real_start_time = p->start_time; |
1368 | monotonic_to_bootbased(&p->real_start_time); | |
1da177e4 | 1369 | p->io_context = NULL; |
1da177e4 | 1370 | p->audit_context = NULL; |
4714d1d3 | 1371 | if (clone_flags & CLONE_THREAD) |
257058ae | 1372 | threadgroup_change_begin(current); |
b4f48b63 | 1373 | cgroup_fork(p); |
1da177e4 | 1374 | #ifdef CONFIG_NUMA |
846a16bf | 1375 | p->mempolicy = mpol_dup(p->mempolicy); |
fb0a685c DRO |
1376 | if (IS_ERR(p->mempolicy)) { |
1377 | retval = PTR_ERR(p->mempolicy); | |
1378 | p->mempolicy = NULL; | |
1379 | goto bad_fork_cleanup_cgroup; | |
1380 | } | |
c61afb18 | 1381 | mpol_fix_fork_child_flag(p); |
1da177e4 | 1382 | #endif |
778d3b0f MH |
1383 | #ifdef CONFIG_CPUSETS |
1384 | p->cpuset_mem_spread_rotor = NUMA_NO_NODE; | |
1385 | p->cpuset_slab_spread_rotor = NUMA_NO_NODE; | |
cc9a6c87 | 1386 | seqcount_init(&p->mems_allowed_seq); |
778d3b0f | 1387 | #endif |
de30a2b3 IM |
1388 | #ifdef CONFIG_TRACE_IRQFLAGS |
1389 | p->irq_events = 0; | |
1390 | p->hardirqs_enabled = 0; | |
1391 | p->hardirq_enable_ip = 0; | |
1392 | p->hardirq_enable_event = 0; | |
1393 | p->hardirq_disable_ip = _THIS_IP_; | |
1394 | p->hardirq_disable_event = 0; | |
1395 | p->softirqs_enabled = 1; | |
1396 | p->softirq_enable_ip = _THIS_IP_; | |
1397 | p->softirq_enable_event = 0; | |
1398 | p->softirq_disable_ip = 0; | |
1399 | p->softirq_disable_event = 0; | |
1400 | p->hardirq_context = 0; | |
1401 | p->softirq_context = 0; | |
1402 | #endif | |
fbb9ce95 IM |
1403 | #ifdef CONFIG_LOCKDEP |
1404 | p->lockdep_depth = 0; /* no locks held yet */ | |
1405 | p->curr_chain_key = 0; | |
1406 | p->lockdep_recursion = 0; | |
1407 | #endif | |
1da177e4 | 1408 | |
408894ee IM |
1409 | #ifdef CONFIG_DEBUG_MUTEXES |
1410 | p->blocked_on = NULL; /* not blocked yet */ | |
1411 | #endif | |
c255a458 | 1412 | #ifdef CONFIG_MEMCG |
569b846d KH |
1413 | p->memcg_batch.do_batch = 0; |
1414 | p->memcg_batch.memcg = NULL; | |
1415 | #endif | |
cafe5635 KO |
1416 | #ifdef CONFIG_BCACHE |
1417 | p->sequential_io = 0; | |
1418 | p->sequential_io_avg = 0; | |
1419 | #endif | |
0f481406 | 1420 | |
3c90e6e9 | 1421 | /* Perform scheduler related setup. Assign this task to a CPU. */ |
3e51e3ed | 1422 | sched_fork(p); |
6ab423e0 | 1423 | |
cdd6c482 | 1424 | retval = perf_event_init_task(p); |
6ab423e0 PZ |
1425 | if (retval) |
1426 | goto bad_fork_cleanup_policy; | |
fb0a685c DRO |
1427 | retval = audit_alloc(p); |
1428 | if (retval) | |
bee870fc | 1429 | goto bad_fork_cleanup_perf; |
1da177e4 | 1430 | /* copy all the process information */ |
fb0a685c DRO |
1431 | retval = copy_semundo(clone_flags, p); |
1432 | if (retval) | |
1da177e4 | 1433 | goto bad_fork_cleanup_audit; |
fb0a685c DRO |
1434 | retval = copy_files(clone_flags, p); |
1435 | if (retval) | |
1da177e4 | 1436 | goto bad_fork_cleanup_semundo; |
fb0a685c DRO |
1437 | retval = copy_fs(clone_flags, p); |
1438 | if (retval) | |
1da177e4 | 1439 | goto bad_fork_cleanup_files; |
fb0a685c DRO |
1440 | retval = copy_sighand(clone_flags, p); |
1441 | if (retval) | |
1da177e4 | 1442 | goto bad_fork_cleanup_fs; |
fb0a685c DRO |
1443 | retval = copy_signal(clone_flags, p); |
1444 | if (retval) | |
1da177e4 | 1445 | goto bad_fork_cleanup_sighand; |
fb0a685c DRO |
1446 | retval = copy_mm(clone_flags, p); |
1447 | if (retval) | |
1da177e4 | 1448 | goto bad_fork_cleanup_signal; |
fb0a685c DRO |
1449 | retval = copy_namespaces(clone_flags, p); |
1450 | if (retval) | |
d84f4f99 | 1451 | goto bad_fork_cleanup_mm; |
fb0a685c DRO |
1452 | retval = copy_io(clone_flags, p); |
1453 | if (retval) | |
fd0928df | 1454 | goto bad_fork_cleanup_namespaces; |
afa86fc4 | 1455 | retval = copy_thread(clone_flags, stack_start, stack_size, p); |
1da177e4 | 1456 | if (retval) |
fd0928df | 1457 | goto bad_fork_cleanup_io; |
1da177e4 | 1458 | |
425fb2b4 PE |
1459 | if (pid != &init_struct_pid) { |
1460 | retval = -ENOMEM; | |
61bce0f1 | 1461 | pid = alloc_pid(p->nsproxy->pid_ns); |
425fb2b4 | 1462 | if (!pid) |
fd0928df | 1463 | goto bad_fork_cleanup_io; |
425fb2b4 PE |
1464 | } |
1465 | ||
1466 | p->pid = pid_nr(pid); | |
1467 | p->tgid = p->pid; | |
1468 | if (clone_flags & CLONE_THREAD) | |
1469 | p->tgid = current->tgid; | |
1470 | ||
1da177e4 LT |
1471 | p->set_child_tid = (clone_flags & CLONE_CHILD_SETTID) ? child_tidptr : NULL; |
1472 | /* | |
1473 | * Clear TID on mm_release()? | |
1474 | */ | |
fb0a685c | 1475 | p->clear_child_tid = (clone_flags & CLONE_CHILD_CLEARTID) ? child_tidptr : NULL; |
73c10101 JA |
1476 | #ifdef CONFIG_BLOCK |
1477 | p->plug = NULL; | |
1478 | #endif | |
42b2dd0a | 1479 | #ifdef CONFIG_FUTEX |
8f17d3a5 IM |
1480 | p->robust_list = NULL; |
1481 | #ifdef CONFIG_COMPAT | |
1482 | p->compat_robust_list = NULL; | |
1483 | #endif | |
c87e2837 IM |
1484 | INIT_LIST_HEAD(&p->pi_state_list); |
1485 | p->pi_state_cache = NULL; | |
42b2dd0a | 1486 | #endif |
0326f5a9 | 1487 | uprobe_copy_process(p); |
f9a3879a GM |
1488 | /* |
1489 | * sigaltstack should be cleared when sharing the same VM | |
1490 | */ | |
1491 | if ((clone_flags & (CLONE_VM|CLONE_VFORK)) == CLONE_VM) | |
1492 | p->sas_ss_sp = p->sas_ss_size = 0; | |
1493 | ||
1da177e4 | 1494 | /* |
6580807d ON |
1495 | * Syscall tracing and stepping should be turned off in the |
1496 | * child regardless of CLONE_PTRACE. | |
1da177e4 | 1497 | */ |
6580807d | 1498 | user_disable_single_step(p); |
1da177e4 | 1499 | clear_tsk_thread_flag(p, TIF_SYSCALL_TRACE); |
ed75e8d5 LV |
1500 | #ifdef TIF_SYSCALL_EMU |
1501 | clear_tsk_thread_flag(p, TIF_SYSCALL_EMU); | |
1502 | #endif | |
9745512c | 1503 | clear_all_latency_tracing(p); |
1da177e4 | 1504 | |
1da177e4 | 1505 | /* ok, now we should be set up.. */ |
5f8aadd8 ON |
1506 | if (clone_flags & CLONE_THREAD) |
1507 | p->exit_signal = -1; | |
1508 | else if (clone_flags & CLONE_PARENT) | |
1509 | p->exit_signal = current->group_leader->exit_signal; | |
1510 | else | |
1511 | p->exit_signal = (clone_flags & CSIGNAL); | |
1512 | ||
1da177e4 LT |
1513 | p->pdeath_signal = 0; |
1514 | p->exit_state = 0; | |
1515 | ||
9d823e8f WF |
1516 | p->nr_dirtied = 0; |
1517 | p->nr_dirtied_pause = 128 >> (PAGE_SHIFT - 10); | |
83712358 | 1518 | p->dirty_paused_when = 0; |
9d823e8f | 1519 | |
1da177e4 LT |
1520 | /* |
1521 | * Ok, make it visible to the rest of the system. | |
1522 | * We dont wake it up yet. | |
1523 | */ | |
1524 | p->group_leader = p; | |
47e65328 | 1525 | INIT_LIST_HEAD(&p->thread_group); |
6fa3eb70 S |
1526 | #ifdef CONFIG_MTK_SCHED_CMP_TGS |
1527 | mt_init_thread_group(p); | |
1528 | #endif | |
158e1645 | 1529 | p->task_works = NULL; |
1da177e4 LT |
1530 | |
1531 | /* Need tasklist lock for parent etc handling! */ | |
1532 | write_lock_irq(&tasklist_lock); | |
1533 | ||
1da177e4 | 1534 | /* CLONE_PARENT re-uses the old parent */ |
2d5516cb | 1535 | if (clone_flags & (CLONE_PARENT|CLONE_THREAD)) { |
1da177e4 | 1536 | p->real_parent = current->real_parent; |
2d5516cb ON |
1537 | p->parent_exec_id = current->parent_exec_id; |
1538 | } else { | |
1da177e4 | 1539 | p->real_parent = current; |
2d5516cb ON |
1540 | p->parent_exec_id = current->self_exec_id; |
1541 | } | |
1da177e4 | 1542 | |
3f17da69 | 1543 | spin_lock(¤t->sighand->siglock); |
4a2c7a78 | 1544 | |
6fa3eb70 S |
1545 | /* |
1546 | * Copy seccomp details explicitly here, in case they were changed | |
1547 | * before holding sighand lock. | |
1548 | */ | |
1549 | copy_seccomp(p); | |
1550 | ||
4a2c7a78 ON |
1551 | /* |
1552 | * Process group and session signals need to be delivered to just the | |
1553 | * parent before the fork or both the parent and the child after the | |
1554 | * fork. Restart if a signal comes in before we add the new process to | |
1555 | * it's process group. | |
1556 | * A fatal signal pending means that current will exit, so the new | |
1557 | * thread can't slip out of an OOM kill (or normal SIGKILL). | |
fb0a685c | 1558 | */ |
23ff4440 | 1559 | recalc_sigpending(); |
4a2c7a78 ON |
1560 | if (signal_pending(current)) { |
1561 | spin_unlock(¤t->sighand->siglock); | |
1562 | write_unlock_irq(&tasklist_lock); | |
1563 | retval = -ERESTARTNOINTR; | |
f7e8b616 | 1564 | goto bad_fork_free_pid; |
4a2c7a78 ON |
1565 | } |
1566 | ||
73b9ebfe | 1567 | if (likely(p->pid)) { |
4b9d33e6 | 1568 | ptrace_init_task(p, (clone_flags & CLONE_PTRACE) || trace); |
73b9ebfe ON |
1569 | |
1570 | if (thread_group_leader(p)) { | |
1c4042c2 | 1571 | if (is_child_reaper(pid)) { |
17cf22c3 | 1572 | ns_of_pid(pid)->child_reaper = p; |
1c4042c2 EB |
1573 | p->signal->flags |= SIGNAL_UNKILLABLE; |
1574 | } | |
73b9ebfe | 1575 | |
fea9d175 | 1576 | p->signal->leader_pid = pid; |
9c9f4ded | 1577 | p->signal->tty = tty_kref_get(current->signal->tty); |
5cd17569 EB |
1578 | attach_pid(p, PIDTYPE_PGID, task_pgrp(current)); |
1579 | attach_pid(p, PIDTYPE_SID, task_session(current)); | |
9cd80bbb | 1580 | list_add_tail(&p->sibling, &p->real_parent->children); |
5e85d4ab | 1581 | list_add_tail_rcu(&p->tasks, &init_task.tasks); |
909ea964 | 1582 | __this_cpu_inc(process_counts); |
6fa3eb70 | 1583 | } else { |
beed6106 ON |
1584 | current->signal->nr_threads++; |
1585 | atomic_inc(¤t->signal->live); | |
1586 | atomic_inc(¤t->signal->sigcnt); | |
1587 | p->group_leader = current->group_leader; | |
1588 | list_add_tail_rcu(&p->thread_group, | |
1589 | &p->group_leader->thread_group); | |
6fa3eb70 S |
1590 | list_add_tail_rcu(&p->thread_node, |
1591 | &p->signal->thread_head); | |
73b9ebfe | 1592 | } |
85868995 | 1593 | attach_pid(p, PIDTYPE_PID, pid); |
73b9ebfe | 1594 | nr_threads++; |
1da177e4 LT |
1595 | } |
1596 | ||
1da177e4 | 1597 | total_forks++; |
3f17da69 | 1598 | spin_unlock(¤t->sighand->siglock); |
e6bc60b8 | 1599 | syscall_tracepoint_update(p); |
1da177e4 | 1600 | write_unlock_irq(&tasklist_lock); |
e6bc60b8 | 1601 | |
c13cf856 | 1602 | proc_fork_connector(p); |
817929ec | 1603 | cgroup_post_fork(p); |
4714d1d3 | 1604 | if (clone_flags & CLONE_THREAD) |
257058ae | 1605 | threadgroup_change_end(current); |
cdd6c482 | 1606 | perf_event_fork(p); |
43d2b113 KH |
1607 | |
1608 | trace_task_newtask(p, clone_flags); | |
1609 | ||
1da177e4 LT |
1610 | return p; |
1611 | ||
425fb2b4 PE |
1612 | bad_fork_free_pid: |
1613 | if (pid != &init_struct_pid) | |
1614 | free_pid(pid); | |
fd0928df | 1615 | bad_fork_cleanup_io: |
b69f2292 LR |
1616 | if (p->io_context) |
1617 | exit_io_context(p); | |
ab516013 | 1618 | bad_fork_cleanup_namespaces: |
444f378b | 1619 | exit_task_namespaces(p); |
1da177e4 | 1620 | bad_fork_cleanup_mm: |
c9f01245 | 1621 | if (p->mm) |
1da177e4 LT |
1622 | mmput(p->mm); |
1623 | bad_fork_cleanup_signal: | |
4ab6c083 | 1624 | if (!(clone_flags & CLONE_THREAD)) |
1c5354de | 1625 | free_signal_struct(p->signal); |
1da177e4 | 1626 | bad_fork_cleanup_sighand: |
a7e5328a | 1627 | __cleanup_sighand(p->sighand); |
1da177e4 LT |
1628 | bad_fork_cleanup_fs: |
1629 | exit_fs(p); /* blocking */ | |
1630 | bad_fork_cleanup_files: | |
1631 | exit_files(p); /* blocking */ | |
1632 | bad_fork_cleanup_semundo: | |
1633 | exit_sem(p); | |
1634 | bad_fork_cleanup_audit: | |
1635 | audit_free(p); | |
bee870fc | 1636 | bad_fork_cleanup_perf: |
cdd6c482 | 1637 | perf_event_free_task(p); |
bee870fc | 1638 | bad_fork_cleanup_policy: |
1da177e4 | 1639 | #ifdef CONFIG_NUMA |
f0be3d32 | 1640 | mpol_put(p->mempolicy); |
b4f48b63 | 1641 | bad_fork_cleanup_cgroup: |
1da177e4 | 1642 | #endif |
4714d1d3 | 1643 | if (clone_flags & CLONE_THREAD) |
257058ae | 1644 | threadgroup_change_end(current); |
5edee61e | 1645 | cgroup_exit(p, 0); |
35df17c5 | 1646 | delayacct_tsk_free(p); |
a1261f54 | 1647 | module_put(task_thread_info(p)->exec_domain->module); |
1da177e4 | 1648 | bad_fork_cleanup_count: |
d84f4f99 | 1649 | atomic_dec(&p->cred->user->processes); |
e0e81739 | 1650 | exit_creds(p); |
1da177e4 LT |
1651 | bad_fork_free: |
1652 | free_task(p); | |
fe7d37d1 | 1653 | fork_out: |
6fa3eb70 | 1654 | printk("[%d:%s] fork fail retval:0x%x\n", current->pid, current->comm, retval); |
fe7d37d1 | 1655 | return ERR_PTR(retval); |
1da177e4 LT |
1656 | } |
1657 | ||
f106eee1 ON |
1658 | static inline void init_idle_pids(struct pid_link *links) |
1659 | { | |
1660 | enum pid_type type; | |
1661 | ||
1662 | for (type = PIDTYPE_PID; type < PIDTYPE_MAX; ++type) { | |
1663 | INIT_HLIST_NODE(&links[type].node); /* not really needed */ | |
1664 | links[type].pid = &init_struct_pid; | |
1665 | } | |
1666 | } | |
1667 | ||
9abcf40b | 1668 | struct task_struct * __cpuinit fork_idle(int cpu) |
1da177e4 | 1669 | { |
36c8b586 | 1670 | struct task_struct *task; |
62e791c1 | 1671 | task = copy_process(CLONE_VM, 0, 0, NULL, &init_struct_pid, 0); |
f106eee1 ON |
1672 | if (!IS_ERR(task)) { |
1673 | init_idle_pids(task->pids); | |
753ca4f3 | 1674 | init_idle(task, cpu); |
f106eee1 | 1675 | } |
73b9ebfe | 1676 | |
1da177e4 LT |
1677 | return task; |
1678 | } | |
1679 | ||
1da177e4 LT |
1680 | /* |
1681 | * Ok, this is the main fork-routine. | |
1682 | * | |
1683 | * It copies the process, and if successful kick-starts | |
1684 | * it and waits for it to finish using the VM if required. | |
1685 | */ | |
6fa3eb70 S |
1686 | #ifdef CONFIG_SCHEDSTATS |
1687 | /* mt shceduler profiling*/ | |
1688 | extern void save_mtproc_info(struct task_struct *p, unsigned long long ts); | |
1689 | #endif | |
1da177e4 LT |
1690 | long do_fork(unsigned long clone_flags, |
1691 | unsigned long stack_start, | |
1da177e4 LT |
1692 | unsigned long stack_size, |
1693 | int __user *parent_tidptr, | |
1694 | int __user *child_tidptr) | |
1695 | { | |
1696 | struct task_struct *p; | |
1697 | int trace = 0; | |
92476d7f | 1698 | long nr; |
1da177e4 | 1699 | |
18b6e041 SH |
1700 | /* |
1701 | * Do some preliminary argument and permissions checking before we | |
1702 | * actually start allocating stuff | |
1703 | */ | |
50804fe3 | 1704 | if (clone_flags & (CLONE_NEWUSER | CLONE_NEWPID)) { |
6fa3eb70 S |
1705 | if (clone_flags & (CLONE_THREAD|CLONE_PARENT)) { |
1706 | printk("[%d:%s] fork fail at clone_thread, flags:0x%x\n", current->pid, current->comm, (unsigned int)clone_flags); | |
18b6e041 | 1707 | return -EINVAL; |
6fa3eb70 | 1708 | } |
18b6e041 SH |
1709 | } |
1710 | ||
09a05394 | 1711 | /* |
4b9d33e6 TH |
1712 | * Determine whether and which event to report to ptracer. When |
1713 | * called from kernel_thread or CLONE_UNTRACED is explicitly | |
1714 | * requested, no event is reported; otherwise, report if the event | |
1715 | * for the type of forking is enabled. | |
09a05394 | 1716 | */ |
e80d6661 | 1717 | if (!(clone_flags & CLONE_UNTRACED)) { |
4b9d33e6 TH |
1718 | if (clone_flags & CLONE_VFORK) |
1719 | trace = PTRACE_EVENT_VFORK; | |
1720 | else if ((clone_flags & CSIGNAL) != SIGCHLD) | |
1721 | trace = PTRACE_EVENT_CLONE; | |
1722 | else | |
1723 | trace = PTRACE_EVENT_FORK; | |
1724 | ||
1725 | if (likely(!ptrace_event_enabled(current, trace))) | |
1726 | trace = 0; | |
1727 | } | |
1da177e4 | 1728 | |
62e791c1 | 1729 | p = copy_process(clone_flags, stack_start, stack_size, |
09a05394 | 1730 | child_tidptr, NULL, trace); |
1da177e4 LT |
1731 | /* |
1732 | * Do this prior waking up the new thread - the thread pointer | |
1733 | * might get invalid after that point, if the thread exits quickly. | |
1734 | */ | |
1735 | if (!IS_ERR(p)) { | |
1736 | struct completion vfork; | |
1a2d9732 | 1737 | struct pid *pid; |
1da177e4 | 1738 | |
0a16b607 MD |
1739 | trace_sched_process_fork(current, p); |
1740 | ||
1a2d9732 MD |
1741 | pid = get_task_pid(p, PIDTYPE_PID); |
1742 | nr = pid_vnr(pid); | |
30e49c26 PE |
1743 | |
1744 | if (clone_flags & CLONE_PARENT_SETTID) | |
1745 | put_user(nr, parent_tidptr); | |
a6f5e063 | 1746 | |
1da177e4 LT |
1747 | if (clone_flags & CLONE_VFORK) { |
1748 | p->vfork_done = &vfork; | |
1749 | init_completion(&vfork); | |
d68b46fe | 1750 | get_task_struct(p); |
1da177e4 LT |
1751 | } |
1752 | ||
6fa3eb70 S |
1753 | #ifdef CONFIG_SCHEDSTATS |
1754 | /* mt shceduler profiling*/ | |
1755 | save_mtproc_info(p, sched_clock()); | |
1756 | printk(KERN_DEBUG "[%d:%s] fork [%d:%s]\n", current->pid, current->comm, p->pid, p->comm); | |
1757 | #endif | |
3e51e3ed | 1758 | wake_up_new_task(p); |
1da177e4 | 1759 | |
4b9d33e6 TH |
1760 | /* forking complete and child started to run, tell ptracer */ |
1761 | if (unlikely(trace)) | |
1a2d9732 | 1762 | ptrace_event_pid(trace, pid); |
09a05394 | 1763 | |
1da177e4 | 1764 | if (clone_flags & CLONE_VFORK) { |
d68b46fe | 1765 | if (!wait_for_vfork_done(p, &vfork)) |
1a2d9732 | 1766 | ptrace_event_pid(PTRACE_EVENT_VFORK_DONE, pid); |
1da177e4 | 1767 | } |
1a2d9732 MD |
1768 | |
1769 | put_pid(pid); | |
6fa3eb70 S |
1770 | #ifdef CONFIG_MT_PRIO_TRACER |
1771 | create_prio_tracer(task_pid_nr(p)); | |
1772 | update_prio_tracer(task_pid_nr(p), p->prio, p->policy, PTS_KRNL); | |
1773 | #endif | |
1da177e4 | 1774 | } else { |
92476d7f | 1775 | nr = PTR_ERR(p); |
6fa3eb70 | 1776 | printk("[%d:%s] fork fail:[%p, %d]\n", current->pid, current->comm, p,(int) nr); |
1da177e4 | 1777 | } |
92476d7f | 1778 | return nr; |
1da177e4 LT |
1779 | } |
1780 | ||
2aa3a7f8 AV |
1781 | /* |
1782 | * Create a kernel thread. | |
1783 | */ | |
1784 | pid_t kernel_thread(int (*fn)(void *), void *arg, unsigned long flags) | |
1785 | { | |
e80d6661 | 1786 | return do_fork(flags|CLONE_VM|CLONE_UNTRACED, (unsigned long)fn, |
2aa3a7f8 AV |
1787 | (unsigned long)arg, NULL, NULL); |
1788 | } | |
2aa3a7f8 | 1789 | |
d2125043 AV |
1790 | #ifdef __ARCH_WANT_SYS_FORK |
1791 | SYSCALL_DEFINE0(fork) | |
1792 | { | |
1793 | #ifdef CONFIG_MMU | |
e80d6661 | 1794 | return do_fork(SIGCHLD, 0, 0, NULL, NULL); |
d2125043 AV |
1795 | #else |
1796 | /* can not support in nommu mode */ | |
1797 | return(-EINVAL); | |
1798 | #endif | |
1799 | } | |
1800 | #endif | |
1801 | ||
1802 | #ifdef __ARCH_WANT_SYS_VFORK | |
1803 | SYSCALL_DEFINE0(vfork) | |
1804 | { | |
e80d6661 | 1805 | return do_fork(CLONE_VFORK | CLONE_VM | SIGCHLD, 0, |
d2125043 AV |
1806 | 0, NULL, NULL); |
1807 | } | |
1808 | #endif | |
1809 | ||
1810 | #ifdef __ARCH_WANT_SYS_CLONE | |
1811 | #ifdef CONFIG_CLONE_BACKWARDS | |
1812 | SYSCALL_DEFINE5(clone, unsigned long, clone_flags, unsigned long, newsp, | |
1813 | int __user *, parent_tidptr, | |
1814 | int, tls_val, | |
1815 | int __user *, child_tidptr) | |
1816 | #elif defined(CONFIG_CLONE_BACKWARDS2) | |
1817 | SYSCALL_DEFINE5(clone, unsigned long, newsp, unsigned long, clone_flags, | |
1818 | int __user *, parent_tidptr, | |
1819 | int __user *, child_tidptr, | |
1820 | int, tls_val) | |
4f01c72e MS |
1821 | #elif defined(CONFIG_CLONE_BACKWARDS3) |
1822 | SYSCALL_DEFINE6(clone, unsigned long, clone_flags, unsigned long, newsp, | |
1823 | int, stack_size, | |
1824 | int __user *, parent_tidptr, | |
1825 | int __user *, child_tidptr, | |
1826 | int, tls_val) | |
d2125043 AV |
1827 | #else |
1828 | SYSCALL_DEFINE5(clone, unsigned long, clone_flags, unsigned long, newsp, | |
1829 | int __user *, parent_tidptr, | |
1830 | int __user *, child_tidptr, | |
1831 | int, tls_val) | |
1832 | #endif | |
1833 | { | |
2cf09666 | 1834 | return do_fork(clone_flags, newsp, 0, parent_tidptr, child_tidptr); |
d2125043 AV |
1835 | } |
1836 | #endif | |
1837 | ||
5fd63b30 RT |
1838 | #ifndef ARCH_MIN_MMSTRUCT_ALIGN |
1839 | #define ARCH_MIN_MMSTRUCT_ALIGN 0 | |
1840 | #endif | |
1841 | ||
51cc5068 | 1842 | static void sighand_ctor(void *data) |
aa1757f9 ON |
1843 | { |
1844 | struct sighand_struct *sighand = data; | |
1845 | ||
a35afb83 | 1846 | spin_lock_init(&sighand->siglock); |
b8fceee1 | 1847 | init_waitqueue_head(&sighand->signalfd_wqh); |
aa1757f9 ON |
1848 | } |
1849 | ||
1da177e4 LT |
1850 | void __init proc_caches_init(void) |
1851 | { | |
1852 | sighand_cachep = kmem_cache_create("sighand_cache", | |
1853 | sizeof(struct sighand_struct), 0, | |
2dff4405 VN |
1854 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_DESTROY_BY_RCU| |
1855 | SLAB_NOTRACK, sighand_ctor); | |
1da177e4 LT |
1856 | signal_cachep = kmem_cache_create("signal_cache", |
1857 | sizeof(struct signal_struct), 0, | |
2dff4405 | 1858 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL); |
20c2df83 | 1859 | files_cachep = kmem_cache_create("files_cache", |
1da177e4 | 1860 | sizeof(struct files_struct), 0, |
2dff4405 | 1861 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL); |
20c2df83 | 1862 | fs_cachep = kmem_cache_create("fs_cache", |
1da177e4 | 1863 | sizeof(struct fs_struct), 0, |
2dff4405 | 1864 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL); |
6345d24d LT |
1865 | /* |
1866 | * FIXME! The "sizeof(struct mm_struct)" currently includes the | |
1867 | * whole struct cpumask for the OFFSTACK case. We could change | |
1868 | * this to *only* allocate as much of it as required by the | |
1869 | * maximum number of CPU's we can ever have. The cpumask_allocation | |
1870 | * is at the end of the structure, exactly for that reason. | |
1871 | */ | |
1da177e4 | 1872 | mm_cachep = kmem_cache_create("mm_struct", |
5fd63b30 | 1873 | sizeof(struct mm_struct), ARCH_MIN_MMSTRUCT_ALIGN, |
2dff4405 | 1874 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL); |
33e5d769 | 1875 | vm_area_cachep = KMEM_CACHE(vm_area_struct, SLAB_PANIC); |
8feae131 | 1876 | mmap_init(); |
66577193 | 1877 | nsproxy_cache_init(); |
1da177e4 | 1878 | } |
cf2e340f | 1879 | |
cf2e340f | 1880 | /* |
9bfb23fc | 1881 | * Check constraints on flags passed to the unshare system call. |
cf2e340f | 1882 | */ |
9bfb23fc | 1883 | static int check_unshare_flags(unsigned long unshare_flags) |
cf2e340f | 1884 | { |
9bfb23fc ON |
1885 | if (unshare_flags & ~(CLONE_THREAD|CLONE_FS|CLONE_NEWNS|CLONE_SIGHAND| |
1886 | CLONE_VM|CLONE_FILES|CLONE_SYSVSEM| | |
50804fe3 | 1887 | CLONE_NEWUTS|CLONE_NEWIPC|CLONE_NEWNET| |
b2e0d987 | 1888 | CLONE_NEWUSER|CLONE_NEWPID)) |
9bfb23fc | 1889 | return -EINVAL; |
cf2e340f | 1890 | /* |
6b7d2f5b EB |
1891 | * Not implemented, but pretend it works if there is nothing |
1892 | * to unshare. Note that unsharing the address space or the | |
1893 | * signal handlers also need to unshare the signal queues (aka | |
1894 | * CLONE_THREAD). | |
cf2e340f | 1895 | */ |
9bfb23fc | 1896 | if (unshare_flags & (CLONE_THREAD | CLONE_SIGHAND | CLONE_VM)) { |
6b7d2f5b EB |
1897 | if (!thread_group_empty(current)) |
1898 | return -EINVAL; | |
1899 | } | |
1900 | if (unshare_flags & (CLONE_SIGHAND | CLONE_VM)) { | |
1901 | if (atomic_read(¤t->sighand->count) > 1) | |
1902 | return -EINVAL; | |
1903 | } | |
1904 | if (unshare_flags & CLONE_VM) { | |
1905 | if (!current_is_single_threaded()) | |
9bfb23fc ON |
1906 | return -EINVAL; |
1907 | } | |
cf2e340f JD |
1908 | |
1909 | return 0; | |
1910 | } | |
1911 | ||
1912 | /* | |
99d1419d | 1913 | * Unshare the filesystem structure if it is being shared |
cf2e340f JD |
1914 | */ |
1915 | static int unshare_fs(unsigned long unshare_flags, struct fs_struct **new_fsp) | |
1916 | { | |
1917 | struct fs_struct *fs = current->fs; | |
1918 | ||
498052bb AV |
1919 | if (!(unshare_flags & CLONE_FS) || !fs) |
1920 | return 0; | |
1921 | ||
1922 | /* don't need lock here; in the worst case we'll do useless copy */ | |
1923 | if (fs->users == 1) | |
1924 | return 0; | |
1925 | ||
1926 | *new_fsp = copy_fs_struct(fs); | |
1927 | if (!*new_fsp) | |
1928 | return -ENOMEM; | |
cf2e340f JD |
1929 | |
1930 | return 0; | |
1931 | } | |
1932 | ||
cf2e340f | 1933 | /* |
a016f338 | 1934 | * Unshare file descriptor table if it is being shared |
cf2e340f JD |
1935 | */ |
1936 | static int unshare_fd(unsigned long unshare_flags, struct files_struct **new_fdp) | |
1937 | { | |
1938 | struct files_struct *fd = current->files; | |
a016f338 | 1939 | int error = 0; |
cf2e340f JD |
1940 | |
1941 | if ((unshare_flags & CLONE_FILES) && | |
a016f338 JD |
1942 | (fd && atomic_read(&fd->count) > 1)) { |
1943 | *new_fdp = dup_fd(fd, &error); | |
1944 | if (!*new_fdp) | |
1945 | return error; | |
1946 | } | |
cf2e340f JD |
1947 | |
1948 | return 0; | |
1949 | } | |
1950 | ||
cf2e340f JD |
1951 | /* |
1952 | * unshare allows a process to 'unshare' part of the process | |
1953 | * context which was originally shared using clone. copy_* | |
1954 | * functions used by do_fork() cannot be used here directly | |
1955 | * because they modify an inactive task_struct that is being | |
1956 | * constructed. Here we are modifying the current, active, | |
1957 | * task_struct. | |
1958 | */ | |
6559eed8 | 1959 | SYSCALL_DEFINE1(unshare, unsigned long, unshare_flags) |
cf2e340f | 1960 | { |
cf2e340f | 1961 | struct fs_struct *fs, *new_fs = NULL; |
cf2e340f | 1962 | struct files_struct *fd, *new_fd = NULL; |
b2e0d987 | 1963 | struct cred *new_cred = NULL; |
cf7b708c | 1964 | struct nsproxy *new_nsproxy = NULL; |
9edff4ab | 1965 | int do_sysvsem = 0; |
9bfb23fc | 1966 | int err; |
cf2e340f | 1967 | |
b2e0d987 EB |
1968 | /* |
1969 | * If unsharing a user namespace must also unshare the thread. | |
1970 | */ | |
1971 | if (unshare_flags & CLONE_NEWUSER) | |
e66eded8 | 1972 | unshare_flags |= CLONE_THREAD | CLONE_FS; |
50804fe3 EB |
1973 | /* |
1974 | * If unsharing a pid namespace must also unshare the thread. | |
1975 | */ | |
1976 | if (unshare_flags & CLONE_NEWPID) | |
1977 | unshare_flags |= CLONE_THREAD; | |
50804fe3 EB |
1978 | /* |
1979 | * If unsharing vm, must also unshare signal handlers. | |
1980 | */ | |
1981 | if (unshare_flags & CLONE_VM) | |
1982 | unshare_flags |= CLONE_SIGHAND; | |
6b7d2f5b EB |
1983 | /* |
1984 | * If unsharing a signal handlers, must also unshare the signal queues. | |
1985 | */ | |
1986 | if (unshare_flags & CLONE_SIGHAND) | |
1987 | unshare_flags |= CLONE_THREAD; | |
9bfb23fc ON |
1988 | /* |
1989 | * If unsharing namespace, must also unshare filesystem information. | |
1990 | */ | |
1991 | if (unshare_flags & CLONE_NEWNS) | |
1992 | unshare_flags |= CLONE_FS; | |
50804fe3 EB |
1993 | |
1994 | err = check_unshare_flags(unshare_flags); | |
1995 | if (err) | |
1996 | goto bad_unshare_out; | |
6013f67f MS |
1997 | /* |
1998 | * CLONE_NEWIPC must also detach from the undolist: after switching | |
1999 | * to a new ipc namespace, the semaphore arrays from the old | |
2000 | * namespace are unreachable. | |
2001 | */ | |
2002 | if (unshare_flags & (CLONE_NEWIPC|CLONE_SYSVSEM)) | |
9edff4ab | 2003 | do_sysvsem = 1; |
fb0a685c DRO |
2004 | err = unshare_fs(unshare_flags, &new_fs); |
2005 | if (err) | |
9bfb23fc | 2006 | goto bad_unshare_out; |
fb0a685c DRO |
2007 | err = unshare_fd(unshare_flags, &new_fd); |
2008 | if (err) | |
9bfb23fc | 2009 | goto bad_unshare_cleanup_fs; |
b2e0d987 | 2010 | err = unshare_userns(unshare_flags, &new_cred); |
fb0a685c | 2011 | if (err) |
9edff4ab | 2012 | goto bad_unshare_cleanup_fd; |
b2e0d987 EB |
2013 | err = unshare_nsproxy_namespaces(unshare_flags, &new_nsproxy, |
2014 | new_cred, new_fs); | |
2015 | if (err) | |
2016 | goto bad_unshare_cleanup_cred; | |
c0b2fc31 | 2017 | |
b2e0d987 | 2018 | if (new_fs || new_fd || do_sysvsem || new_cred || new_nsproxy) { |
9edff4ab MS |
2019 | if (do_sysvsem) { |
2020 | /* | |
2021 | * CLONE_SYSVSEM is equivalent to sys_exit(). | |
2022 | */ | |
2023 | exit_sem(current); | |
2024 | } | |
ab516013 | 2025 | |
6f977e6b | 2026 | if (new_nsproxy) |
cf7b708c | 2027 | switch_task_namespaces(current, new_nsproxy); |
cf2e340f | 2028 | |
cf7b708c PE |
2029 | task_lock(current); |
2030 | ||
cf2e340f JD |
2031 | if (new_fs) { |
2032 | fs = current->fs; | |
2a4419b5 | 2033 | spin_lock(&fs->lock); |
cf2e340f | 2034 | current->fs = new_fs; |
498052bb AV |
2035 | if (--fs->users) |
2036 | new_fs = NULL; | |
2037 | else | |
2038 | new_fs = fs; | |
2a4419b5 | 2039 | spin_unlock(&fs->lock); |
cf2e340f JD |
2040 | } |
2041 | ||
cf2e340f JD |
2042 | if (new_fd) { |
2043 | fd = current->files; | |
2044 | current->files = new_fd; | |
2045 | new_fd = fd; | |
2046 | } | |
2047 | ||
2048 | task_unlock(current); | |
b2e0d987 EB |
2049 | |
2050 | if (new_cred) { | |
2051 | /* Install the new user namespace */ | |
2052 | commit_creds(new_cred); | |
2053 | new_cred = NULL; | |
2054 | } | |
cf2e340f JD |
2055 | } |
2056 | ||
b2e0d987 EB |
2057 | bad_unshare_cleanup_cred: |
2058 | if (new_cred) | |
2059 | put_cred(new_cred); | |
cf2e340f JD |
2060 | bad_unshare_cleanup_fd: |
2061 | if (new_fd) | |
2062 | put_files_struct(new_fd); | |
2063 | ||
cf2e340f JD |
2064 | bad_unshare_cleanup_fs: |
2065 | if (new_fs) | |
498052bb | 2066 | free_fs_struct(new_fs); |
cf2e340f | 2067 | |
cf2e340f JD |
2068 | bad_unshare_out: |
2069 | return err; | |
2070 | } | |
3b125388 AV |
2071 | |
2072 | /* | |
2073 | * Helper to unshare the files of the current task. | |
2074 | * We don't want to expose copy_files internals to | |
2075 | * the exec layer of the kernel. | |
2076 | */ | |
2077 | ||
2078 | int unshare_files(struct files_struct **displaced) | |
2079 | { | |
2080 | struct task_struct *task = current; | |
50704516 | 2081 | struct files_struct *copy = NULL; |
3b125388 AV |
2082 | int error; |
2083 | ||
2084 | error = unshare_fd(CLONE_FILES, ©); | |
2085 | if (error || !copy) { | |
2086 | *displaced = NULL; | |
2087 | return error; | |
2088 | } | |
2089 | *displaced = task->files; | |
2090 | task_lock(task); | |
2091 | task->files = copy; | |
2092 | task_unlock(task); | |
2093 | return 0; | |
2094 | } |