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> |
615d6e87 DB |
31 | #include <linux/mm.h> |
32 | #include <linux/vmacache.h> | |
ab516013 | 33 | #include <linux/nsproxy.h> |
c59ede7b | 34 | #include <linux/capability.h> |
1da177e4 | 35 | #include <linux/cpu.h> |
b4f48b63 | 36 | #include <linux/cgroup.h> |
1da177e4 | 37 | #include <linux/security.h> |
a1e78772 | 38 | #include <linux/hugetlb.h> |
e2cfabdf | 39 | #include <linux/seccomp.h> |
1da177e4 LT |
40 | #include <linux/swap.h> |
41 | #include <linux/syscalls.h> | |
42 | #include <linux/jiffies.h> | |
43 | #include <linux/futex.h> | |
8141c7f3 | 44 | #include <linux/compat.h> |
207205a2 | 45 | #include <linux/kthread.h> |
7c3ab738 | 46 | #include <linux/task_io_accounting_ops.h> |
ab2af1f5 | 47 | #include <linux/rcupdate.h> |
1da177e4 LT |
48 | #include <linux/ptrace.h> |
49 | #include <linux/mount.h> | |
50 | #include <linux/audit.h> | |
78fb7466 | 51 | #include <linux/memcontrol.h> |
f201ae23 | 52 | #include <linux/ftrace.h> |
5e2bf014 | 53 | #include <linux/proc_fs.h> |
1da177e4 LT |
54 | #include <linux/profile.h> |
55 | #include <linux/rmap.h> | |
f8af4da3 | 56 | #include <linux/ksm.h> |
1da177e4 | 57 | #include <linux/acct.h> |
8f0ab514 | 58 | #include <linux/tsacct_kern.h> |
9f46080c | 59 | #include <linux/cn_proc.h> |
ba96a0c8 | 60 | #include <linux/freezer.h> |
ca74e92b | 61 | #include <linux/delayacct.h> |
ad4ecbcb | 62 | #include <linux/taskstats_kern.h> |
0a425405 | 63 | #include <linux/random.h> |
522ed776 | 64 | #include <linux/tty.h> |
fd0928df | 65 | #include <linux/blkdev.h> |
5ad4e53b | 66 | #include <linux/fs_struct.h> |
7c9f8861 | 67 | #include <linux/magic.h> |
cdd6c482 | 68 | #include <linux/perf_event.h> |
42c4ab41 | 69 | #include <linux/posix-timers.h> |
8e7cac79 | 70 | #include <linux/user-return-notifier.h> |
3d5992d2 | 71 | #include <linux/oom.h> |
ba76149f | 72 | #include <linux/khugepaged.h> |
d80e731e | 73 | #include <linux/signalfd.h> |
0326f5a9 | 74 | #include <linux/uprobes.h> |
a27bb332 | 75 | #include <linux/aio.h> |
52f5684c | 76 | #include <linux/compiler.h> |
16db3d3f | 77 | #include <linux/sysctl.h> |
5c9a8750 | 78 | #include <linux/kcov.h> |
1da177e4 LT |
79 | |
80 | #include <asm/pgtable.h> | |
81 | #include <asm/pgalloc.h> | |
82 | #include <asm/uaccess.h> | |
83 | #include <asm/mmu_context.h> | |
84 | #include <asm/cacheflush.h> | |
85 | #include <asm/tlbflush.h> | |
86 | ||
ad8d75ff SR |
87 | #include <trace/events/sched.h> |
88 | ||
43d2b113 KH |
89 | #define CREATE_TRACE_POINTS |
90 | #include <trace/events/task.h> | |
91 | ||
ac1b398d HS |
92 | /* |
93 | * Minimum number of threads to boot the kernel | |
94 | */ | |
95 | #define MIN_THREADS 20 | |
96 | ||
97 | /* | |
98 | * Maximum number of threads | |
99 | */ | |
100 | #define MAX_THREADS FUTEX_TID_MASK | |
101 | ||
1da177e4 LT |
102 | /* |
103 | * Protected counters by write_lock_irq(&tasklist_lock) | |
104 | */ | |
105 | unsigned long total_forks; /* Handle normal Linux uptimes. */ | |
fb0a685c | 106 | int nr_threads; /* The idle threads do not count.. */ |
1da177e4 LT |
107 | |
108 | int max_threads; /* tunable limit on nr_threads */ | |
109 | ||
110 | DEFINE_PER_CPU(unsigned long, process_counts) = 0; | |
111 | ||
c59923a1 | 112 | __cacheline_aligned DEFINE_RWLOCK(tasklist_lock); /* outer */ |
db1466b3 PM |
113 | |
114 | #ifdef CONFIG_PROVE_RCU | |
115 | int lockdep_tasklist_lock_is_held(void) | |
116 | { | |
117 | return lockdep_is_held(&tasklist_lock); | |
118 | } | |
119 | EXPORT_SYMBOL_GPL(lockdep_tasklist_lock_is_held); | |
120 | #endif /* #ifdef CONFIG_PROVE_RCU */ | |
1da177e4 LT |
121 | |
122 | int nr_processes(void) | |
123 | { | |
124 | int cpu; | |
125 | int total = 0; | |
126 | ||
1d510750 | 127 | for_each_possible_cpu(cpu) |
1da177e4 LT |
128 | total += per_cpu(process_counts, cpu); |
129 | ||
130 | return total; | |
131 | } | |
132 | ||
f19b9f74 AM |
133 | void __weak arch_release_task_struct(struct task_struct *tsk) |
134 | { | |
135 | } | |
136 | ||
f5e10287 | 137 | #ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR |
e18b890b | 138 | static struct kmem_cache *task_struct_cachep; |
41101809 TG |
139 | |
140 | static inline struct task_struct *alloc_task_struct_node(int node) | |
141 | { | |
142 | return kmem_cache_alloc_node(task_struct_cachep, GFP_KERNEL, node); | |
143 | } | |
144 | ||
41101809 TG |
145 | static inline void free_task_struct(struct task_struct *tsk) |
146 | { | |
41101809 TG |
147 | kmem_cache_free(task_struct_cachep, tsk); |
148 | } | |
1da177e4 LT |
149 | #endif |
150 | ||
b235beea | 151 | void __weak arch_release_thread_stack(unsigned long *stack) |
f19b9f74 AM |
152 | { |
153 | } | |
154 | ||
b235beea | 155 | #ifndef CONFIG_ARCH_THREAD_STACK_ALLOCATOR |
41101809 | 156 | |
0d15d74a TG |
157 | /* |
158 | * Allocate pages if THREAD_SIZE is >= PAGE_SIZE, otherwise use a | |
159 | * kmemcache based allocator. | |
160 | */ | |
ba14a194 AL |
161 | # if THREAD_SIZE >= PAGE_SIZE || defined(CONFIG_VMAP_STACK) |
162 | static unsigned long *alloc_thread_stack_node(struct task_struct *tsk, int node) | |
b69c49b7 | 163 | { |
ba14a194 AL |
164 | #ifdef CONFIG_VMAP_STACK |
165 | void *stack = __vmalloc_node_range(THREAD_SIZE, THREAD_SIZE, | |
166 | VMALLOC_START, VMALLOC_END, | |
167 | THREADINFO_GFP | __GFP_HIGHMEM, | |
168 | PAGE_KERNEL, | |
169 | 0, node, | |
170 | __builtin_return_address(0)); | |
171 | ||
172 | /* | |
173 | * We can't call find_vm_area() in interrupt context, and | |
174 | * free_thread_stack() can be called in interrupt context, | |
175 | * so cache the vm_struct. | |
176 | */ | |
177 | if (stack) | |
178 | tsk->stack_vm_area = find_vm_area(stack); | |
179 | return stack; | |
180 | #else | |
4949148a VD |
181 | struct page *page = alloc_pages_node(node, THREADINFO_GFP, |
182 | THREAD_SIZE_ORDER); | |
b6a84016 ED |
183 | |
184 | return page ? page_address(page) : NULL; | |
ba14a194 | 185 | #endif |
b69c49b7 FT |
186 | } |
187 | ||
ba14a194 | 188 | static inline void free_thread_stack(struct task_struct *tsk) |
b69c49b7 | 189 | { |
ba14a194 AL |
190 | if (task_stack_vm_area(tsk)) |
191 | vfree(tsk->stack); | |
192 | else | |
193 | __free_pages(virt_to_page(tsk->stack), THREAD_SIZE_ORDER); | |
b69c49b7 | 194 | } |
0d15d74a | 195 | # else |
b235beea | 196 | static struct kmem_cache *thread_stack_cache; |
0d15d74a | 197 | |
9521d399 | 198 | static unsigned long *alloc_thread_stack_node(struct task_struct *tsk, |
0d15d74a TG |
199 | int node) |
200 | { | |
b235beea | 201 | return kmem_cache_alloc_node(thread_stack_cache, THREADINFO_GFP, node); |
0d15d74a TG |
202 | } |
203 | ||
ba14a194 | 204 | static void free_thread_stack(struct task_struct *tsk) |
0d15d74a | 205 | { |
ba14a194 | 206 | kmem_cache_free(thread_stack_cache, tsk->stack); |
0d15d74a TG |
207 | } |
208 | ||
b235beea | 209 | void thread_stack_cache_init(void) |
0d15d74a | 210 | { |
b235beea | 211 | thread_stack_cache = kmem_cache_create("thread_stack", THREAD_SIZE, |
0d15d74a | 212 | THREAD_SIZE, 0, NULL); |
b235beea | 213 | BUG_ON(thread_stack_cache == NULL); |
0d15d74a TG |
214 | } |
215 | # endif | |
b69c49b7 FT |
216 | #endif |
217 | ||
1da177e4 | 218 | /* SLAB cache for signal_struct structures (tsk->signal) */ |
e18b890b | 219 | static struct kmem_cache *signal_cachep; |
1da177e4 LT |
220 | |
221 | /* SLAB cache for sighand_struct structures (tsk->sighand) */ | |
e18b890b | 222 | struct kmem_cache *sighand_cachep; |
1da177e4 LT |
223 | |
224 | /* SLAB cache for files_struct structures (tsk->files) */ | |
e18b890b | 225 | struct kmem_cache *files_cachep; |
1da177e4 LT |
226 | |
227 | /* SLAB cache for fs_struct structures (tsk->fs) */ | |
e18b890b | 228 | struct kmem_cache *fs_cachep; |
1da177e4 LT |
229 | |
230 | /* SLAB cache for vm_area_struct structures */ | |
e18b890b | 231 | struct kmem_cache *vm_area_cachep; |
1da177e4 LT |
232 | |
233 | /* SLAB cache for mm_struct structures (tsk->mm) */ | |
e18b890b | 234 | static struct kmem_cache *mm_cachep; |
1da177e4 | 235 | |
ba14a194 | 236 | static void account_kernel_stack(struct task_struct *tsk, int account) |
c6a7f572 | 237 | { |
ba14a194 AL |
238 | void *stack = task_stack_page(tsk); |
239 | struct vm_struct *vm = task_stack_vm_area(tsk); | |
240 | ||
241 | BUILD_BUG_ON(IS_ENABLED(CONFIG_VMAP_STACK) && PAGE_SIZE % 1024 != 0); | |
242 | ||
243 | if (vm) { | |
244 | int i; | |
245 | ||
246 | BUG_ON(vm->nr_pages != THREAD_SIZE / PAGE_SIZE); | |
247 | ||
248 | for (i = 0; i < THREAD_SIZE / PAGE_SIZE; i++) { | |
249 | mod_zone_page_state(page_zone(vm->pages[i]), | |
250 | NR_KERNEL_STACK_KB, | |
251 | PAGE_SIZE / 1024 * account); | |
252 | } | |
c6a7f572 | 253 | |
ba14a194 AL |
254 | /* All stack pages belong to the same memcg. */ |
255 | memcg_kmem_update_page_stat(vm->pages[0], MEMCG_KERNEL_STACK_KB, | |
256 | account * (THREAD_SIZE / 1024)); | |
257 | } else { | |
258 | /* | |
259 | * All stack pages are in the same zone and belong to the | |
260 | * same memcg. | |
261 | */ | |
262 | struct page *first_page = virt_to_page(stack); | |
efdc9490 | 263 | |
ba14a194 AL |
264 | mod_zone_page_state(page_zone(first_page), NR_KERNEL_STACK_KB, |
265 | THREAD_SIZE / 1024 * account); | |
266 | ||
267 | memcg_kmem_update_page_stat(first_page, MEMCG_KERNEL_STACK_KB, | |
268 | account * (THREAD_SIZE / 1024)); | |
269 | } | |
c6a7f572 KM |
270 | } |
271 | ||
1da177e4 LT |
272 | void free_task(struct task_struct *tsk) |
273 | { | |
ba14a194 | 274 | account_kernel_stack(tsk, -1); |
b235beea | 275 | arch_release_thread_stack(tsk->stack); |
ba14a194 | 276 | free_thread_stack(tsk); |
23f78d4a | 277 | rt_mutex_debug_task_free(tsk); |
fb52607a | 278 | ftrace_graph_exit_task(tsk); |
e2cfabdf | 279 | put_seccomp_filter(tsk); |
f19b9f74 | 280 | arch_release_task_struct(tsk); |
1da177e4 LT |
281 | free_task_struct(tsk); |
282 | } | |
283 | EXPORT_SYMBOL(free_task); | |
284 | ||
ea6d290c ON |
285 | static inline void free_signal_struct(struct signal_struct *sig) |
286 | { | |
97101eb4 | 287 | taskstats_tgid_free(sig); |
1c5354de | 288 | sched_autogroup_exit(sig); |
ea6d290c ON |
289 | kmem_cache_free(signal_cachep, sig); |
290 | } | |
291 | ||
292 | static inline void put_signal_struct(struct signal_struct *sig) | |
293 | { | |
1c5354de | 294 | if (atomic_dec_and_test(&sig->sigcnt)) |
ea6d290c ON |
295 | free_signal_struct(sig); |
296 | } | |
297 | ||
158d9ebd | 298 | void __put_task_struct(struct task_struct *tsk) |
1da177e4 | 299 | { |
270f722d | 300 | WARN_ON(!tsk->exit_state); |
1da177e4 LT |
301 | WARN_ON(atomic_read(&tsk->usage)); |
302 | WARN_ON(tsk == current); | |
303 | ||
2e91fa7f | 304 | cgroup_free(tsk); |
156654f4 | 305 | task_numa_free(tsk); |
1a2a4d06 | 306 | security_task_free(tsk); |
e0e81739 | 307 | exit_creds(tsk); |
35df17c5 | 308 | delayacct_tsk_free(tsk); |
ea6d290c | 309 | put_signal_struct(tsk->signal); |
1da177e4 LT |
310 | |
311 | if (!profile_handoff_task(tsk)) | |
312 | free_task(tsk); | |
313 | } | |
77c100c8 | 314 | EXPORT_SYMBOL_GPL(__put_task_struct); |
1da177e4 | 315 | |
6c0a9fa6 | 316 | void __init __weak arch_task_cache_init(void) { } |
61c4628b | 317 | |
ff691f6e HS |
318 | /* |
319 | * set_max_threads | |
320 | */ | |
16db3d3f | 321 | static void set_max_threads(unsigned int max_threads_suggested) |
ff691f6e | 322 | { |
ac1b398d | 323 | u64 threads; |
ff691f6e HS |
324 | |
325 | /* | |
ac1b398d HS |
326 | * The number of threads shall be limited such that the thread |
327 | * structures may only consume a small part of the available memory. | |
ff691f6e | 328 | */ |
ac1b398d HS |
329 | if (fls64(totalram_pages) + fls64(PAGE_SIZE) > 64) |
330 | threads = MAX_THREADS; | |
331 | else | |
332 | threads = div64_u64((u64) totalram_pages * (u64) PAGE_SIZE, | |
333 | (u64) THREAD_SIZE * 8UL); | |
334 | ||
16db3d3f HS |
335 | if (threads > max_threads_suggested) |
336 | threads = max_threads_suggested; | |
337 | ||
ac1b398d | 338 | max_threads = clamp_t(u64, threads, MIN_THREADS, MAX_THREADS); |
ff691f6e HS |
339 | } |
340 | ||
5aaeb5c0 IM |
341 | #ifdef CONFIG_ARCH_WANTS_DYNAMIC_TASK_STRUCT |
342 | /* Initialized by the architecture: */ | |
343 | int arch_task_struct_size __read_mostly; | |
344 | #endif | |
0c8c0f03 | 345 | |
ff691f6e | 346 | void __init fork_init(void) |
1da177e4 | 347 | { |
f5e10287 | 348 | #ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR |
1da177e4 LT |
349 | #ifndef ARCH_MIN_TASKALIGN |
350 | #define ARCH_MIN_TASKALIGN L1_CACHE_BYTES | |
351 | #endif | |
352 | /* create a slab on which task_structs can be allocated */ | |
5d097056 VD |
353 | task_struct_cachep = kmem_cache_create("task_struct", |
354 | arch_task_struct_size, ARCH_MIN_TASKALIGN, | |
355 | SLAB_PANIC|SLAB_NOTRACK|SLAB_ACCOUNT, NULL); | |
1da177e4 LT |
356 | #endif |
357 | ||
61c4628b SS |
358 | /* do the arch specific task caches init */ |
359 | arch_task_cache_init(); | |
360 | ||
16db3d3f | 361 | set_max_threads(MAX_THREADS); |
1da177e4 LT |
362 | |
363 | init_task.signal->rlim[RLIMIT_NPROC].rlim_cur = max_threads/2; | |
364 | init_task.signal->rlim[RLIMIT_NPROC].rlim_max = max_threads/2; | |
365 | init_task.signal->rlim[RLIMIT_SIGPENDING] = | |
366 | init_task.signal->rlim[RLIMIT_NPROC]; | |
367 | } | |
368 | ||
52f5684c | 369 | int __weak arch_dup_task_struct(struct task_struct *dst, |
61c4628b SS |
370 | struct task_struct *src) |
371 | { | |
372 | *dst = *src; | |
373 | return 0; | |
374 | } | |
375 | ||
d4311ff1 AT |
376 | void set_task_stack_end_magic(struct task_struct *tsk) |
377 | { | |
378 | unsigned long *stackend; | |
379 | ||
380 | stackend = end_of_stack(tsk); | |
381 | *stackend = STACK_END_MAGIC; /* for overflow detection */ | |
382 | } | |
383 | ||
725fc629 | 384 | static struct task_struct *dup_task_struct(struct task_struct *orig, int node) |
1da177e4 LT |
385 | { |
386 | struct task_struct *tsk; | |
b235beea | 387 | unsigned long *stack; |
ba14a194 | 388 | struct vm_struct *stack_vm_area; |
3e26c149 | 389 | int err; |
1da177e4 | 390 | |
725fc629 AK |
391 | if (node == NUMA_NO_NODE) |
392 | node = tsk_fork_get_node(orig); | |
504f52b5 | 393 | tsk = alloc_task_struct_node(node); |
1da177e4 LT |
394 | if (!tsk) |
395 | return NULL; | |
396 | ||
b235beea LT |
397 | stack = alloc_thread_stack_node(tsk, node); |
398 | if (!stack) | |
f19b9f74 | 399 | goto free_tsk; |
1da177e4 | 400 | |
ba14a194 AL |
401 | stack_vm_area = task_stack_vm_area(tsk); |
402 | ||
fb0a685c | 403 | err = arch_dup_task_struct(tsk, orig); |
ba14a194 AL |
404 | |
405 | /* | |
406 | * arch_dup_task_struct() clobbers the stack-related fields. Make | |
407 | * sure they're properly initialized before using any stack-related | |
408 | * functions again. | |
409 | */ | |
410 | tsk->stack = stack; | |
411 | #ifdef CONFIG_VMAP_STACK | |
412 | tsk->stack_vm_area = stack_vm_area; | |
413 | #endif | |
414 | ||
164c33c6 | 415 | if (err) |
b235beea | 416 | goto free_stack; |
164c33c6 | 417 | |
dbd95212 KC |
418 | #ifdef CONFIG_SECCOMP |
419 | /* | |
420 | * We must handle setting up seccomp filters once we're under | |
421 | * the sighand lock in case orig has changed between now and | |
422 | * then. Until then, filter must be NULL to avoid messing up | |
423 | * the usage counts on the error path calling free_task. | |
424 | */ | |
425 | tsk->seccomp.filter = NULL; | |
426 | #endif | |
87bec58a AM |
427 | |
428 | setup_thread_stack(tsk, orig); | |
8e7cac79 | 429 | clear_user_return_notifier(tsk); |
f26f9aff | 430 | clear_tsk_need_resched(tsk); |
d4311ff1 | 431 | set_task_stack_end_magic(tsk); |
1da177e4 | 432 | |
0a425405 AV |
433 | #ifdef CONFIG_CC_STACKPROTECTOR |
434 | tsk->stack_canary = get_random_int(); | |
435 | #endif | |
436 | ||
fb0a685c DRO |
437 | /* |
438 | * One for us, one for whoever does the "release_task()" (usually | |
439 | * parent) | |
440 | */ | |
441 | atomic_set(&tsk->usage, 2); | |
6c5c9341 | 442 | #ifdef CONFIG_BLK_DEV_IO_TRACE |
2056a782 | 443 | tsk->btrace_seq = 0; |
6c5c9341 | 444 | #endif |
a0aa7f68 | 445 | tsk->splice_pipe = NULL; |
5640f768 | 446 | tsk->task_frag.page = NULL; |
093e5840 | 447 | tsk->wake_q.next = NULL; |
c6a7f572 | 448 | |
ba14a194 | 449 | account_kernel_stack(tsk, 1); |
c6a7f572 | 450 | |
5c9a8750 DV |
451 | kcov_task_init(tsk); |
452 | ||
1da177e4 | 453 | return tsk; |
61c4628b | 454 | |
b235beea | 455 | free_stack: |
ba14a194 | 456 | free_thread_stack(tsk); |
f19b9f74 | 457 | free_tsk: |
61c4628b SS |
458 | free_task_struct(tsk); |
459 | return NULL; | |
1da177e4 LT |
460 | } |
461 | ||
462 | #ifdef CONFIG_MMU | |
09dd109d ER |
463 | static __latent_entropy int dup_mmap(struct mm_struct *mm, |
464 | struct mm_struct *oldmm) | |
1da177e4 | 465 | { |
297c5eee | 466 | struct vm_area_struct *mpnt, *tmp, *prev, **pprev; |
1da177e4 LT |
467 | struct rb_node **rb_link, *rb_parent; |
468 | int retval; | |
469 | unsigned long charge; | |
1da177e4 | 470 | |
32cdba1e | 471 | uprobe_start_dup_mmap(); |
7c051267 MH |
472 | if (down_write_killable(&oldmm->mmap_sem)) { |
473 | retval = -EINTR; | |
474 | goto fail_uprobe_end; | |
475 | } | |
ec8c0446 | 476 | flush_cache_dup_mm(oldmm); |
f8ac4ec9 | 477 | uprobe_dup_mmap(oldmm, mm); |
ad339451 IM |
478 | /* |
479 | * Not linked in yet - no deadlock potential: | |
480 | */ | |
481 | down_write_nested(&mm->mmap_sem, SINGLE_DEPTH_NESTING); | |
7ee78232 | 482 | |
90f31d0e KK |
483 | /* No ordering required: file already has been exposed. */ |
484 | RCU_INIT_POINTER(mm->exe_file, get_mm_exe_file(oldmm)); | |
485 | ||
4f7d4614 | 486 | mm->total_vm = oldmm->total_vm; |
84638335 | 487 | mm->data_vm = oldmm->data_vm; |
4f7d4614 VD |
488 | mm->exec_vm = oldmm->exec_vm; |
489 | mm->stack_vm = oldmm->stack_vm; | |
490 | ||
1da177e4 LT |
491 | rb_link = &mm->mm_rb.rb_node; |
492 | rb_parent = NULL; | |
493 | pprev = &mm->mmap; | |
f8af4da3 | 494 | retval = ksm_fork(mm, oldmm); |
ba76149f AA |
495 | if (retval) |
496 | goto out; | |
497 | retval = khugepaged_fork(mm, oldmm); | |
f8af4da3 HD |
498 | if (retval) |
499 | goto out; | |
1da177e4 | 500 | |
297c5eee | 501 | prev = NULL; |
fd3e42fc | 502 | for (mpnt = oldmm->mmap; mpnt; mpnt = mpnt->vm_next) { |
1da177e4 LT |
503 | struct file *file; |
504 | ||
505 | if (mpnt->vm_flags & VM_DONTCOPY) { | |
84638335 | 506 | vm_stat_account(mm, mpnt->vm_flags, -vma_pages(mpnt)); |
1da177e4 LT |
507 | continue; |
508 | } | |
509 | charge = 0; | |
510 | if (mpnt->vm_flags & VM_ACCOUNT) { | |
b2412b7f HS |
511 | unsigned long len = vma_pages(mpnt); |
512 | ||
191c5424 | 513 | if (security_vm_enough_memory_mm(oldmm, len)) /* sic */ |
1da177e4 LT |
514 | goto fail_nomem; |
515 | charge = len; | |
516 | } | |
e94b1766 | 517 | tmp = kmem_cache_alloc(vm_area_cachep, GFP_KERNEL); |
1da177e4 LT |
518 | if (!tmp) |
519 | goto fail_nomem; | |
520 | *tmp = *mpnt; | |
5beb4930 | 521 | INIT_LIST_HEAD(&tmp->anon_vma_chain); |
ef0855d3 ON |
522 | retval = vma_dup_policy(mpnt, tmp); |
523 | if (retval) | |
1da177e4 | 524 | goto fail_nomem_policy; |
a247c3a9 | 525 | tmp->vm_mm = mm; |
5beb4930 RR |
526 | if (anon_vma_fork(tmp, mpnt)) |
527 | goto fail_nomem_anon_vma_fork; | |
de60f5f1 EM |
528 | tmp->vm_flags &= |
529 | ~(VM_LOCKED|VM_LOCKONFAULT|VM_UFFD_MISSING|VM_UFFD_WP); | |
297c5eee | 530 | tmp->vm_next = tmp->vm_prev = NULL; |
745f234b | 531 | tmp->vm_userfaultfd_ctx = NULL_VM_UFFD_CTX; |
1da177e4 LT |
532 | file = tmp->vm_file; |
533 | if (file) { | |
496ad9aa | 534 | struct inode *inode = file_inode(file); |
b88ed205 HD |
535 | struct address_space *mapping = file->f_mapping; |
536 | ||
1da177e4 LT |
537 | get_file(file); |
538 | if (tmp->vm_flags & VM_DENYWRITE) | |
539 | atomic_dec(&inode->i_writecount); | |
83cde9e8 | 540 | i_mmap_lock_write(mapping); |
b88ed205 | 541 | if (tmp->vm_flags & VM_SHARED) |
4bb5f5d9 | 542 | atomic_inc(&mapping->i_mmap_writable); |
b88ed205 HD |
543 | flush_dcache_mmap_lock(mapping); |
544 | /* insert tmp into the share list, just after mpnt */ | |
27ba0644 KS |
545 | vma_interval_tree_insert_after(tmp, mpnt, |
546 | &mapping->i_mmap); | |
b88ed205 | 547 | flush_dcache_mmap_unlock(mapping); |
83cde9e8 | 548 | i_mmap_unlock_write(mapping); |
1da177e4 LT |
549 | } |
550 | ||
a1e78772 MG |
551 | /* |
552 | * Clear hugetlb-related page reserves for children. This only | |
553 | * affects MAP_PRIVATE mappings. Faults generated by the child | |
554 | * are not guaranteed to succeed, even if read-only | |
555 | */ | |
556 | if (is_vm_hugetlb_page(tmp)) | |
557 | reset_vma_resv_huge_pages(tmp); | |
558 | ||
1da177e4 | 559 | /* |
7ee78232 | 560 | * Link in the new vma and copy the page table entries. |
1da177e4 | 561 | */ |
1da177e4 LT |
562 | *pprev = tmp; |
563 | pprev = &tmp->vm_next; | |
297c5eee LT |
564 | tmp->vm_prev = prev; |
565 | prev = tmp; | |
1da177e4 LT |
566 | |
567 | __vma_link_rb(mm, tmp, rb_link, rb_parent); | |
568 | rb_link = &tmp->vm_rb.rb_right; | |
569 | rb_parent = &tmp->vm_rb; | |
570 | ||
571 | mm->map_count++; | |
0b0db14c | 572 | retval = copy_page_range(mm, oldmm, mpnt); |
1da177e4 LT |
573 | |
574 | if (tmp->vm_ops && tmp->vm_ops->open) | |
575 | tmp->vm_ops->open(tmp); | |
576 | ||
577 | if (retval) | |
578 | goto out; | |
579 | } | |
d6dd61c8 JF |
580 | /* a new mm has just been created */ |
581 | arch_dup_mmap(oldmm, mm); | |
1da177e4 | 582 | retval = 0; |
1da177e4 | 583 | out: |
7ee78232 | 584 | up_write(&mm->mmap_sem); |
fd3e42fc | 585 | flush_tlb_mm(oldmm); |
1da177e4 | 586 | up_write(&oldmm->mmap_sem); |
7c051267 | 587 | fail_uprobe_end: |
32cdba1e | 588 | uprobe_end_dup_mmap(); |
1da177e4 | 589 | return retval; |
5beb4930 | 590 | fail_nomem_anon_vma_fork: |
ef0855d3 | 591 | mpol_put(vma_policy(tmp)); |
1da177e4 LT |
592 | fail_nomem_policy: |
593 | kmem_cache_free(vm_area_cachep, tmp); | |
594 | fail_nomem: | |
595 | retval = -ENOMEM; | |
596 | vm_unacct_memory(charge); | |
597 | goto out; | |
598 | } | |
599 | ||
fb0a685c | 600 | static inline int mm_alloc_pgd(struct mm_struct *mm) |
1da177e4 LT |
601 | { |
602 | mm->pgd = pgd_alloc(mm); | |
603 | if (unlikely(!mm->pgd)) | |
604 | return -ENOMEM; | |
605 | return 0; | |
606 | } | |
607 | ||
fb0a685c | 608 | static inline void mm_free_pgd(struct mm_struct *mm) |
1da177e4 | 609 | { |
5e541973 | 610 | pgd_free(mm, mm->pgd); |
1da177e4 LT |
611 | } |
612 | #else | |
90f31d0e KK |
613 | static int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm) |
614 | { | |
615 | down_write(&oldmm->mmap_sem); | |
616 | RCU_INIT_POINTER(mm->exe_file, get_mm_exe_file(oldmm)); | |
617 | up_write(&oldmm->mmap_sem); | |
618 | return 0; | |
619 | } | |
1da177e4 LT |
620 | #define mm_alloc_pgd(mm) (0) |
621 | #define mm_free_pgd(mm) | |
622 | #endif /* CONFIG_MMU */ | |
623 | ||
23ff4440 | 624 | __cacheline_aligned_in_smp DEFINE_SPINLOCK(mmlist_lock); |
1da177e4 | 625 | |
e94b1766 | 626 | #define allocate_mm() (kmem_cache_alloc(mm_cachep, GFP_KERNEL)) |
1da177e4 LT |
627 | #define free_mm(mm) (kmem_cache_free(mm_cachep, (mm))) |
628 | ||
4cb0e11b HK |
629 | static unsigned long default_dump_filter = MMF_DUMP_FILTER_DEFAULT; |
630 | ||
631 | static int __init coredump_filter_setup(char *s) | |
632 | { | |
633 | default_dump_filter = | |
634 | (simple_strtoul(s, NULL, 0) << MMF_DUMP_FILTER_SHIFT) & | |
635 | MMF_DUMP_FILTER_MASK; | |
636 | return 1; | |
637 | } | |
638 | ||
639 | __setup("coredump_filter=", coredump_filter_setup); | |
640 | ||
1da177e4 LT |
641 | #include <linux/init_task.h> |
642 | ||
858f0993 AD |
643 | static void mm_init_aio(struct mm_struct *mm) |
644 | { | |
645 | #ifdef CONFIG_AIO | |
646 | spin_lock_init(&mm->ioctx_lock); | |
db446a08 | 647 | mm->ioctx_table = NULL; |
858f0993 AD |
648 | #endif |
649 | } | |
650 | ||
33144e84 VD |
651 | static void mm_init_owner(struct mm_struct *mm, struct task_struct *p) |
652 | { | |
653 | #ifdef CONFIG_MEMCG | |
654 | mm->owner = p; | |
655 | #endif | |
656 | } | |
657 | ||
fb0a685c | 658 | static struct mm_struct *mm_init(struct mm_struct *mm, struct task_struct *p) |
1da177e4 | 659 | { |
41f727fd VD |
660 | mm->mmap = NULL; |
661 | mm->mm_rb = RB_ROOT; | |
662 | mm->vmacache_seqnum = 0; | |
1da177e4 LT |
663 | atomic_set(&mm->mm_users, 1); |
664 | atomic_set(&mm->mm_count, 1); | |
665 | init_rwsem(&mm->mmap_sem); | |
666 | INIT_LIST_HEAD(&mm->mmlist); | |
999d9fc1 | 667 | mm->core_state = NULL; |
e1f56c89 | 668 | atomic_long_set(&mm->nr_ptes, 0); |
2d2f5119 | 669 | mm_nr_pmds_init(mm); |
41f727fd VD |
670 | mm->map_count = 0; |
671 | mm->locked_vm = 0; | |
ce65cefa | 672 | mm->pinned_vm = 0; |
d559db08 | 673 | memset(&mm->rss_stat, 0, sizeof(mm->rss_stat)); |
1da177e4 | 674 | spin_lock_init(&mm->page_table_lock); |
41f727fd | 675 | mm_init_cpumask(mm); |
858f0993 | 676 | mm_init_aio(mm); |
cf475ad2 | 677 | mm_init_owner(mm, p); |
41f727fd | 678 | mmu_notifier_mm_init(mm); |
20841405 | 679 | clear_tlb_flush_pending(mm); |
41f727fd VD |
680 | #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS |
681 | mm->pmd_huge_pte = NULL; | |
682 | #endif | |
1da177e4 | 683 | |
a0715cc2 AT |
684 | if (current->mm) { |
685 | mm->flags = current->mm->flags & MMF_INIT_MASK; | |
686 | mm->def_flags = current->mm->def_flags & VM_INIT_DEF_MASK; | |
687 | } else { | |
688 | mm->flags = default_dump_filter; | |
1da177e4 | 689 | mm->def_flags = 0; |
a0715cc2 AT |
690 | } |
691 | ||
41f727fd VD |
692 | if (mm_alloc_pgd(mm)) |
693 | goto fail_nopgd; | |
694 | ||
695 | if (init_new_context(p, mm)) | |
696 | goto fail_nocontext; | |
78fb7466 | 697 | |
41f727fd VD |
698 | return mm; |
699 | ||
700 | fail_nocontext: | |
701 | mm_free_pgd(mm); | |
702 | fail_nopgd: | |
1da177e4 LT |
703 | free_mm(mm); |
704 | return NULL; | |
705 | } | |
706 | ||
c3f0327f KK |
707 | static void check_mm(struct mm_struct *mm) |
708 | { | |
709 | int i; | |
710 | ||
711 | for (i = 0; i < NR_MM_COUNTERS; i++) { | |
712 | long x = atomic_long_read(&mm->rss_stat.count[i]); | |
713 | ||
714 | if (unlikely(x)) | |
715 | printk(KERN_ALERT "BUG: Bad rss-counter state " | |
716 | "mm:%p idx:%d val:%ld\n", mm, i, x); | |
717 | } | |
b30fe6c7 KS |
718 | |
719 | if (atomic_long_read(&mm->nr_ptes)) | |
720 | pr_alert("BUG: non-zero nr_ptes on freeing mm: %ld\n", | |
721 | atomic_long_read(&mm->nr_ptes)); | |
722 | if (mm_nr_pmds(mm)) | |
723 | pr_alert("BUG: non-zero nr_pmds on freeing mm: %ld\n", | |
724 | mm_nr_pmds(mm)); | |
725 | ||
e009bb30 | 726 | #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS |
96dad67f | 727 | VM_BUG_ON_MM(mm->pmd_huge_pte, mm); |
c3f0327f KK |
728 | #endif |
729 | } | |
730 | ||
1da177e4 LT |
731 | /* |
732 | * Allocate and initialize an mm_struct. | |
733 | */ | |
fb0a685c | 734 | struct mm_struct *mm_alloc(void) |
1da177e4 | 735 | { |
fb0a685c | 736 | struct mm_struct *mm; |
1da177e4 LT |
737 | |
738 | mm = allocate_mm(); | |
de03c72c KM |
739 | if (!mm) |
740 | return NULL; | |
741 | ||
742 | memset(mm, 0, sizeof(*mm)); | |
6345d24d | 743 | return mm_init(mm, current); |
1da177e4 LT |
744 | } |
745 | ||
746 | /* | |
747 | * Called when the last reference to the mm | |
748 | * is dropped: either by a lazy thread or by | |
749 | * mmput. Free the page directory and the mm. | |
750 | */ | |
7ad5b3a5 | 751 | void __mmdrop(struct mm_struct *mm) |
1da177e4 LT |
752 | { |
753 | BUG_ON(mm == &init_mm); | |
754 | mm_free_pgd(mm); | |
755 | destroy_context(mm); | |
cddb8a5c | 756 | mmu_notifier_mm_destroy(mm); |
c3f0327f | 757 | check_mm(mm); |
1da177e4 LT |
758 | free_mm(mm); |
759 | } | |
6d4e4c4f | 760 | EXPORT_SYMBOL_GPL(__mmdrop); |
1da177e4 | 761 | |
ec8d7c14 MH |
762 | static inline void __mmput(struct mm_struct *mm) |
763 | { | |
764 | VM_BUG_ON(atomic_read(&mm->mm_users)); | |
765 | ||
766 | uprobe_clear_state(mm); | |
767 | exit_aio(mm); | |
768 | ksm_exit(mm); | |
769 | khugepaged_exit(mm); /* must run before exit_mmap */ | |
770 | exit_mmap(mm); | |
771 | set_mm_exe_file(mm, NULL); | |
772 | if (!list_empty(&mm->mmlist)) { | |
773 | spin_lock(&mmlist_lock); | |
774 | list_del(&mm->mmlist); | |
775 | spin_unlock(&mmlist_lock); | |
776 | } | |
777 | if (mm->binfmt) | |
778 | module_put(mm->binfmt->module); | |
779 | mmdrop(mm); | |
780 | } | |
781 | ||
1da177e4 LT |
782 | /* |
783 | * Decrement the use count and release all resources for an mm. | |
784 | */ | |
785 | void mmput(struct mm_struct *mm) | |
786 | { | |
0ae26f1b AM |
787 | might_sleep(); |
788 | ||
ec8d7c14 MH |
789 | if (atomic_dec_and_test(&mm->mm_users)) |
790 | __mmput(mm); | |
791 | } | |
792 | EXPORT_SYMBOL_GPL(mmput); | |
793 | ||
7ef949d7 | 794 | #ifdef CONFIG_MMU |
ec8d7c14 MH |
795 | static void mmput_async_fn(struct work_struct *work) |
796 | { | |
797 | struct mm_struct *mm = container_of(work, struct mm_struct, async_put_work); | |
798 | __mmput(mm); | |
799 | } | |
800 | ||
801 | void mmput_async(struct mm_struct *mm) | |
802 | { | |
1da177e4 | 803 | if (atomic_dec_and_test(&mm->mm_users)) { |
ec8d7c14 MH |
804 | INIT_WORK(&mm->async_put_work, mmput_async_fn); |
805 | schedule_work(&mm->async_put_work); | |
1da177e4 LT |
806 | } |
807 | } | |
7ef949d7 | 808 | #endif |
1da177e4 | 809 | |
90f31d0e KK |
810 | /** |
811 | * set_mm_exe_file - change a reference to the mm's executable file | |
812 | * | |
813 | * This changes mm's executable file (shown as symlink /proc/[pid]/exe). | |
814 | * | |
6e399cd1 DB |
815 | * Main users are mmput() and sys_execve(). Callers prevent concurrent |
816 | * invocations: in mmput() nobody alive left, in execve task is single | |
817 | * threaded. sys_prctl(PR_SET_MM_MAP/EXE_FILE) also needs to set the | |
818 | * mm->exe_file, but does so without using set_mm_exe_file() in order | |
819 | * to do avoid the need for any locks. | |
90f31d0e | 820 | */ |
38646013 JS |
821 | void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file) |
822 | { | |
6e399cd1 DB |
823 | struct file *old_exe_file; |
824 | ||
825 | /* | |
826 | * It is safe to dereference the exe_file without RCU as | |
827 | * this function is only called if nobody else can access | |
828 | * this mm -- see comment above for justification. | |
829 | */ | |
830 | old_exe_file = rcu_dereference_raw(mm->exe_file); | |
90f31d0e | 831 | |
38646013 JS |
832 | if (new_exe_file) |
833 | get_file(new_exe_file); | |
90f31d0e KK |
834 | rcu_assign_pointer(mm->exe_file, new_exe_file); |
835 | if (old_exe_file) | |
836 | fput(old_exe_file); | |
38646013 JS |
837 | } |
838 | ||
90f31d0e KK |
839 | /** |
840 | * get_mm_exe_file - acquire a reference to the mm's executable file | |
841 | * | |
842 | * Returns %NULL if mm has no associated executable file. | |
843 | * User must release file via fput(). | |
844 | */ | |
38646013 JS |
845 | struct file *get_mm_exe_file(struct mm_struct *mm) |
846 | { | |
847 | struct file *exe_file; | |
848 | ||
90f31d0e KK |
849 | rcu_read_lock(); |
850 | exe_file = rcu_dereference(mm->exe_file); | |
851 | if (exe_file && !get_file_rcu(exe_file)) | |
852 | exe_file = NULL; | |
853 | rcu_read_unlock(); | |
38646013 JS |
854 | return exe_file; |
855 | } | |
11163348 | 856 | EXPORT_SYMBOL(get_mm_exe_file); |
38646013 | 857 | |
cd81a917 MG |
858 | /** |
859 | * get_task_exe_file - acquire a reference to the task's executable file | |
860 | * | |
861 | * Returns %NULL if task's mm (if any) has no associated executable file or | |
862 | * this is a kernel thread with borrowed mm (see the comment above get_task_mm). | |
863 | * User must release file via fput(). | |
864 | */ | |
865 | struct file *get_task_exe_file(struct task_struct *task) | |
866 | { | |
867 | struct file *exe_file = NULL; | |
868 | struct mm_struct *mm; | |
869 | ||
870 | task_lock(task); | |
871 | mm = task->mm; | |
872 | if (mm) { | |
873 | if (!(task->flags & PF_KTHREAD)) | |
874 | exe_file = get_mm_exe_file(mm); | |
875 | } | |
876 | task_unlock(task); | |
877 | return exe_file; | |
878 | } | |
879 | EXPORT_SYMBOL(get_task_exe_file); | |
38646013 | 880 | |
1da177e4 LT |
881 | /** |
882 | * get_task_mm - acquire a reference to the task's mm | |
883 | * | |
246bb0b1 | 884 | * Returns %NULL if the task has no mm. Checks PF_KTHREAD (meaning |
1da177e4 LT |
885 | * this kernel workthread has transiently adopted a user mm with use_mm, |
886 | * to do its AIO) is not set and if so returns a reference to it, after | |
887 | * bumping up the use count. User must release the mm via mmput() | |
888 | * after use. Typically used by /proc and ptrace. | |
889 | */ | |
890 | struct mm_struct *get_task_mm(struct task_struct *task) | |
891 | { | |
892 | struct mm_struct *mm; | |
893 | ||
894 | task_lock(task); | |
895 | mm = task->mm; | |
896 | if (mm) { | |
246bb0b1 | 897 | if (task->flags & PF_KTHREAD) |
1da177e4 LT |
898 | mm = NULL; |
899 | else | |
900 | atomic_inc(&mm->mm_users); | |
901 | } | |
902 | task_unlock(task); | |
903 | return mm; | |
904 | } | |
905 | EXPORT_SYMBOL_GPL(get_task_mm); | |
906 | ||
8cdb878d CY |
907 | struct mm_struct *mm_access(struct task_struct *task, unsigned int mode) |
908 | { | |
909 | struct mm_struct *mm; | |
910 | int err; | |
911 | ||
912 | err = mutex_lock_killable(&task->signal->cred_guard_mutex); | |
913 | if (err) | |
914 | return ERR_PTR(err); | |
915 | ||
916 | mm = get_task_mm(task); | |
917 | if (mm && mm != current->mm && | |
918 | !ptrace_may_access(task, mode)) { | |
919 | mmput(mm); | |
920 | mm = ERR_PTR(-EACCES); | |
921 | } | |
922 | mutex_unlock(&task->signal->cred_guard_mutex); | |
923 | ||
924 | return mm; | |
925 | } | |
926 | ||
57b59c4a | 927 | static void complete_vfork_done(struct task_struct *tsk) |
c415c3b4 | 928 | { |
d68b46fe | 929 | struct completion *vfork; |
c415c3b4 | 930 | |
d68b46fe ON |
931 | task_lock(tsk); |
932 | vfork = tsk->vfork_done; | |
933 | if (likely(vfork)) { | |
934 | tsk->vfork_done = NULL; | |
935 | complete(vfork); | |
936 | } | |
937 | task_unlock(tsk); | |
938 | } | |
939 | ||
940 | static int wait_for_vfork_done(struct task_struct *child, | |
941 | struct completion *vfork) | |
942 | { | |
943 | int killed; | |
944 | ||
945 | freezer_do_not_count(); | |
946 | killed = wait_for_completion_killable(vfork); | |
947 | freezer_count(); | |
948 | ||
949 | if (killed) { | |
950 | task_lock(child); | |
951 | child->vfork_done = NULL; | |
952 | task_unlock(child); | |
953 | } | |
954 | ||
955 | put_task_struct(child); | |
956 | return killed; | |
c415c3b4 ON |
957 | } |
958 | ||
1da177e4 LT |
959 | /* Please note the differences between mmput and mm_release. |
960 | * mmput is called whenever we stop holding onto a mm_struct, | |
961 | * error success whatever. | |
962 | * | |
963 | * mm_release is called after a mm_struct has been removed | |
964 | * from the current process. | |
965 | * | |
966 | * This difference is important for error handling, when we | |
967 | * only half set up a mm_struct for a new process and need to restore | |
968 | * the old one. Because we mmput the new mm_struct before | |
969 | * restoring the old one. . . | |
970 | * Eric Biederman 10 January 1998 | |
971 | */ | |
972 | void mm_release(struct task_struct *tsk, struct mm_struct *mm) | |
973 | { | |
8141c7f3 LT |
974 | /* Get rid of any futexes when releasing the mm */ |
975 | #ifdef CONFIG_FUTEX | |
fc6b177d | 976 | if (unlikely(tsk->robust_list)) { |
8141c7f3 | 977 | exit_robust_list(tsk); |
fc6b177d PZ |
978 | tsk->robust_list = NULL; |
979 | } | |
8141c7f3 | 980 | #ifdef CONFIG_COMPAT |
fc6b177d | 981 | if (unlikely(tsk->compat_robust_list)) { |
8141c7f3 | 982 | compat_exit_robust_list(tsk); |
fc6b177d PZ |
983 | tsk->compat_robust_list = NULL; |
984 | } | |
8141c7f3 | 985 | #endif |
322a2c10 TG |
986 | if (unlikely(!list_empty(&tsk->pi_state_list))) |
987 | exit_pi_state_list(tsk); | |
8141c7f3 LT |
988 | #endif |
989 | ||
0326f5a9 SD |
990 | uprobe_free_utask(tsk); |
991 | ||
1da177e4 LT |
992 | /* Get rid of any cached register state */ |
993 | deactivate_mm(tsk, mm); | |
994 | ||
fec1d011 | 995 | /* |
735f2770 MH |
996 | * Signal userspace if we're not exiting with a core dump |
997 | * because we want to leave the value intact for debugging | |
998 | * purposes. | |
fec1d011 | 999 | */ |
9c8a8228 | 1000 | if (tsk->clear_child_tid) { |
735f2770 | 1001 | if (!(tsk->signal->flags & SIGNAL_GROUP_COREDUMP) && |
9c8a8228 ED |
1002 | atomic_read(&mm->mm_users) > 1) { |
1003 | /* | |
1004 | * We don't check the error code - if userspace has | |
1005 | * not set up a proper pointer then tough luck. | |
1006 | */ | |
1007 | put_user(0, tsk->clear_child_tid); | |
1008 | sys_futex(tsk->clear_child_tid, FUTEX_WAKE, | |
1009 | 1, NULL, NULL, 0); | |
1010 | } | |
1da177e4 | 1011 | tsk->clear_child_tid = NULL; |
1da177e4 | 1012 | } |
f7505d64 KK |
1013 | |
1014 | /* | |
1015 | * All done, finally we can wake up parent and return this mm to him. | |
1016 | * Also kthread_stop() uses this completion for synchronization. | |
1017 | */ | |
1018 | if (tsk->vfork_done) | |
1019 | complete_vfork_done(tsk); | |
1da177e4 LT |
1020 | } |
1021 | ||
a0a7ec30 JD |
1022 | /* |
1023 | * Allocate a new mm structure and copy contents from the | |
1024 | * mm structure of the passed in task structure. | |
1025 | */ | |
ff252c1f | 1026 | static struct mm_struct *dup_mm(struct task_struct *tsk) |
a0a7ec30 JD |
1027 | { |
1028 | struct mm_struct *mm, *oldmm = current->mm; | |
1029 | int err; | |
1030 | ||
a0a7ec30 JD |
1031 | mm = allocate_mm(); |
1032 | if (!mm) | |
1033 | goto fail_nomem; | |
1034 | ||
1035 | memcpy(mm, oldmm, sizeof(*mm)); | |
1036 | ||
78fb7466 | 1037 | if (!mm_init(mm, tsk)) |
a0a7ec30 JD |
1038 | goto fail_nomem; |
1039 | ||
a0a7ec30 JD |
1040 | err = dup_mmap(mm, oldmm); |
1041 | if (err) | |
1042 | goto free_pt; | |
1043 | ||
1044 | mm->hiwater_rss = get_mm_rss(mm); | |
1045 | mm->hiwater_vm = mm->total_vm; | |
1046 | ||
801460d0 HS |
1047 | if (mm->binfmt && !try_module_get(mm->binfmt->module)) |
1048 | goto free_pt; | |
1049 | ||
a0a7ec30 JD |
1050 | return mm; |
1051 | ||
1052 | free_pt: | |
801460d0 HS |
1053 | /* don't put binfmt in mmput, we haven't got module yet */ |
1054 | mm->binfmt = NULL; | |
a0a7ec30 JD |
1055 | mmput(mm); |
1056 | ||
1057 | fail_nomem: | |
1058 | return NULL; | |
a0a7ec30 JD |
1059 | } |
1060 | ||
fb0a685c | 1061 | static int copy_mm(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 | 1062 | { |
fb0a685c | 1063 | struct mm_struct *mm, *oldmm; |
1da177e4 LT |
1064 | int retval; |
1065 | ||
1066 | tsk->min_flt = tsk->maj_flt = 0; | |
1067 | tsk->nvcsw = tsk->nivcsw = 0; | |
17406b82 MSB |
1068 | #ifdef CONFIG_DETECT_HUNG_TASK |
1069 | tsk->last_switch_count = tsk->nvcsw + tsk->nivcsw; | |
1070 | #endif | |
1da177e4 LT |
1071 | |
1072 | tsk->mm = NULL; | |
1073 | tsk->active_mm = NULL; | |
1074 | ||
1075 | /* | |
1076 | * Are we cloning a kernel thread? | |
1077 | * | |
1078 | * We need to steal a active VM for that.. | |
1079 | */ | |
1080 | oldmm = current->mm; | |
1081 | if (!oldmm) | |
1082 | return 0; | |
1083 | ||
615d6e87 DB |
1084 | /* initialize the new vmacache entries */ |
1085 | vmacache_flush(tsk); | |
1086 | ||
1da177e4 LT |
1087 | if (clone_flags & CLONE_VM) { |
1088 | atomic_inc(&oldmm->mm_users); | |
1089 | mm = oldmm; | |
1da177e4 LT |
1090 | goto good_mm; |
1091 | } | |
1092 | ||
1093 | retval = -ENOMEM; | |
a0a7ec30 | 1094 | mm = dup_mm(tsk); |
1da177e4 LT |
1095 | if (!mm) |
1096 | goto fail_nomem; | |
1097 | ||
1da177e4 LT |
1098 | good_mm: |
1099 | tsk->mm = mm; | |
1100 | tsk->active_mm = mm; | |
1101 | return 0; | |
1102 | ||
1da177e4 LT |
1103 | fail_nomem: |
1104 | return retval; | |
1da177e4 LT |
1105 | } |
1106 | ||
a39bc516 | 1107 | static int copy_fs(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 | 1108 | { |
498052bb | 1109 | struct fs_struct *fs = current->fs; |
1da177e4 | 1110 | if (clone_flags & CLONE_FS) { |
498052bb | 1111 | /* tsk->fs is already what we want */ |
2a4419b5 | 1112 | spin_lock(&fs->lock); |
498052bb | 1113 | if (fs->in_exec) { |
2a4419b5 | 1114 | spin_unlock(&fs->lock); |
498052bb AV |
1115 | return -EAGAIN; |
1116 | } | |
1117 | fs->users++; | |
2a4419b5 | 1118 | spin_unlock(&fs->lock); |
1da177e4 LT |
1119 | return 0; |
1120 | } | |
498052bb | 1121 | tsk->fs = copy_fs_struct(fs); |
1da177e4 LT |
1122 | if (!tsk->fs) |
1123 | return -ENOMEM; | |
1124 | return 0; | |
1125 | } | |
1126 | ||
fb0a685c | 1127 | static int copy_files(unsigned long clone_flags, struct task_struct *tsk) |
a016f338 JD |
1128 | { |
1129 | struct files_struct *oldf, *newf; | |
1130 | int error = 0; | |
1131 | ||
1132 | /* | |
1133 | * A background process may not have any files ... | |
1134 | */ | |
1135 | oldf = current->files; | |
1136 | if (!oldf) | |
1137 | goto out; | |
1138 | ||
1139 | if (clone_flags & CLONE_FILES) { | |
1140 | atomic_inc(&oldf->count); | |
1141 | goto out; | |
1142 | } | |
1143 | ||
a016f338 JD |
1144 | newf = dup_fd(oldf, &error); |
1145 | if (!newf) | |
1146 | goto out; | |
1147 | ||
1148 | tsk->files = newf; | |
1149 | error = 0; | |
1150 | out: | |
1151 | return error; | |
1152 | } | |
1153 | ||
fadad878 | 1154 | static int copy_io(unsigned long clone_flags, struct task_struct *tsk) |
fd0928df JA |
1155 | { |
1156 | #ifdef CONFIG_BLOCK | |
1157 | struct io_context *ioc = current->io_context; | |
6e736be7 | 1158 | struct io_context *new_ioc; |
fd0928df JA |
1159 | |
1160 | if (!ioc) | |
1161 | return 0; | |
fadad878 JA |
1162 | /* |
1163 | * Share io context with parent, if CLONE_IO is set | |
1164 | */ | |
1165 | if (clone_flags & CLONE_IO) { | |
3d48749d TH |
1166 | ioc_task_link(ioc); |
1167 | tsk->io_context = ioc; | |
fadad878 | 1168 | } else if (ioprio_valid(ioc->ioprio)) { |
6e736be7 TH |
1169 | new_ioc = get_task_io_context(tsk, GFP_KERNEL, NUMA_NO_NODE); |
1170 | if (unlikely(!new_ioc)) | |
fd0928df JA |
1171 | return -ENOMEM; |
1172 | ||
6e736be7 | 1173 | new_ioc->ioprio = ioc->ioprio; |
11a3122f | 1174 | put_io_context(new_ioc); |
fd0928df JA |
1175 | } |
1176 | #endif | |
1177 | return 0; | |
1178 | } | |
1179 | ||
a39bc516 | 1180 | static int copy_sighand(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 LT |
1181 | { |
1182 | struct sighand_struct *sig; | |
1183 | ||
60348802 | 1184 | if (clone_flags & CLONE_SIGHAND) { |
1da177e4 LT |
1185 | atomic_inc(¤t->sighand->count); |
1186 | return 0; | |
1187 | } | |
1188 | sig = kmem_cache_alloc(sighand_cachep, GFP_KERNEL); | |
e56d0903 | 1189 | rcu_assign_pointer(tsk->sighand, sig); |
1da177e4 LT |
1190 | if (!sig) |
1191 | return -ENOMEM; | |
9d7fb042 | 1192 | |
1da177e4 LT |
1193 | atomic_set(&sig->count, 1); |
1194 | memcpy(sig->action, current->sighand->action, sizeof(sig->action)); | |
1195 | return 0; | |
1196 | } | |
1197 | ||
a7e5328a | 1198 | void __cleanup_sighand(struct sighand_struct *sighand) |
c81addc9 | 1199 | { |
d80e731e ON |
1200 | if (atomic_dec_and_test(&sighand->count)) { |
1201 | signalfd_cleanup(sighand); | |
392809b2 ON |
1202 | /* |
1203 | * sighand_cachep is SLAB_DESTROY_BY_RCU so we can free it | |
1204 | * without an RCU grace period, see __lock_task_sighand(). | |
1205 | */ | |
c81addc9 | 1206 | kmem_cache_free(sighand_cachep, sighand); |
d80e731e | 1207 | } |
c81addc9 ON |
1208 | } |
1209 | ||
f06febc9 FM |
1210 | /* |
1211 | * Initialize POSIX timer handling for a thread group. | |
1212 | */ | |
1213 | static void posix_cpu_timers_init_group(struct signal_struct *sig) | |
1214 | { | |
78d7d407 JS |
1215 | unsigned long cpu_limit; |
1216 | ||
316c1608 | 1217 | cpu_limit = READ_ONCE(sig->rlim[RLIMIT_CPU].rlim_cur); |
78d7d407 JS |
1218 | if (cpu_limit != RLIM_INFINITY) { |
1219 | sig->cputime_expires.prof_exp = secs_to_cputime(cpu_limit); | |
d5c373eb | 1220 | sig->cputimer.running = true; |
6279a751 ON |
1221 | } |
1222 | ||
f06febc9 FM |
1223 | /* The timer lists. */ |
1224 | INIT_LIST_HEAD(&sig->cpu_timers[0]); | |
1225 | INIT_LIST_HEAD(&sig->cpu_timers[1]); | |
1226 | INIT_LIST_HEAD(&sig->cpu_timers[2]); | |
1227 | } | |
1228 | ||
a39bc516 | 1229 | static int copy_signal(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 LT |
1230 | { |
1231 | struct signal_struct *sig; | |
1da177e4 | 1232 | |
4ab6c083 | 1233 | if (clone_flags & CLONE_THREAD) |
490dea45 | 1234 | return 0; |
490dea45 | 1235 | |
a56704ef | 1236 | sig = kmem_cache_zalloc(signal_cachep, GFP_KERNEL); |
1da177e4 LT |
1237 | tsk->signal = sig; |
1238 | if (!sig) | |
1239 | return -ENOMEM; | |
1240 | ||
b3ac022c | 1241 | sig->nr_threads = 1; |
1da177e4 | 1242 | atomic_set(&sig->live, 1); |
b3ac022c | 1243 | atomic_set(&sig->sigcnt, 1); |
0c740d0a ON |
1244 | |
1245 | /* list_add(thread_node, thread_head) without INIT_LIST_HEAD() */ | |
1246 | sig->thread_head = (struct list_head)LIST_HEAD_INIT(tsk->thread_node); | |
1247 | tsk->thread_node = (struct list_head)LIST_HEAD_INIT(sig->thread_head); | |
1248 | ||
1da177e4 | 1249 | init_waitqueue_head(&sig->wait_chldexit); |
db51aecc | 1250 | sig->curr_target = tsk; |
1da177e4 LT |
1251 | init_sigpending(&sig->shared_pending); |
1252 | INIT_LIST_HEAD(&sig->posix_timers); | |
e78c3496 | 1253 | seqlock_init(&sig->stats_lock); |
9d7fb042 | 1254 | prev_cputime_init(&sig->prev_cputime); |
1da177e4 | 1255 | |
c9cb2e3d | 1256 | hrtimer_init(&sig->real_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL); |
1da177e4 | 1257 | sig->real_timer.function = it_real_fn; |
1da177e4 | 1258 | |
1da177e4 LT |
1259 | task_lock(current->group_leader); |
1260 | memcpy(sig->rlim, current->signal->rlim, sizeof sig->rlim); | |
1261 | task_unlock(current->group_leader); | |
1262 | ||
6279a751 ON |
1263 | posix_cpu_timers_init_group(sig); |
1264 | ||
522ed776 | 1265 | tty_audit_fork(sig); |
5091faa4 | 1266 | sched_autogroup_fork(sig); |
522ed776 | 1267 | |
a63d83f4 | 1268 | sig->oom_score_adj = current->signal->oom_score_adj; |
dabb16f6 | 1269 | sig->oom_score_adj_min = current->signal->oom_score_adj_min; |
28b83c51 | 1270 | |
ebec18a6 LP |
1271 | sig->has_child_subreaper = current->signal->has_child_subreaper || |
1272 | current->signal->is_child_subreaper; | |
1273 | ||
9b1bf12d KM |
1274 | mutex_init(&sig->cred_guard_mutex); |
1275 | ||
1da177e4 LT |
1276 | return 0; |
1277 | } | |
1278 | ||
dbd95212 KC |
1279 | static void copy_seccomp(struct task_struct *p) |
1280 | { | |
1281 | #ifdef CONFIG_SECCOMP | |
1282 | /* | |
1283 | * Must be called with sighand->lock held, which is common to | |
1284 | * all threads in the group. Holding cred_guard_mutex is not | |
1285 | * needed because this new task is not yet running and cannot | |
1286 | * be racing exec. | |
1287 | */ | |
69f6a34b | 1288 | assert_spin_locked(¤t->sighand->siglock); |
dbd95212 KC |
1289 | |
1290 | /* Ref-count the new filter user, and assign it. */ | |
1291 | get_seccomp_filter(current); | |
1292 | p->seccomp = current->seccomp; | |
1293 | ||
1294 | /* | |
1295 | * Explicitly enable no_new_privs here in case it got set | |
1296 | * between the task_struct being duplicated and holding the | |
1297 | * sighand lock. The seccomp state and nnp must be in sync. | |
1298 | */ | |
1299 | if (task_no_new_privs(current)) | |
1300 | task_set_no_new_privs(p); | |
1301 | ||
1302 | /* | |
1303 | * If the parent gained a seccomp mode after copying thread | |
1304 | * flags and between before we held the sighand lock, we have | |
1305 | * to manually enable the seccomp thread flag here. | |
1306 | */ | |
1307 | if (p->seccomp.mode != SECCOMP_MODE_DISABLED) | |
1308 | set_tsk_thread_flag(p, TIF_SECCOMP); | |
1309 | #endif | |
1310 | } | |
1311 | ||
17da2bd9 | 1312 | SYSCALL_DEFINE1(set_tid_address, int __user *, tidptr) |
1da177e4 LT |
1313 | { |
1314 | current->clear_child_tid = tidptr; | |
1315 | ||
b488893a | 1316 | return task_pid_vnr(current); |
1da177e4 LT |
1317 | } |
1318 | ||
a39bc516 | 1319 | static void rt_mutex_init_task(struct task_struct *p) |
23f78d4a | 1320 | { |
1d615482 | 1321 | raw_spin_lock_init(&p->pi_lock); |
e29e175b | 1322 | #ifdef CONFIG_RT_MUTEXES |
fb00aca4 PZ |
1323 | p->pi_waiters = RB_ROOT; |
1324 | p->pi_waiters_leftmost = NULL; | |
23f78d4a | 1325 | p->pi_blocked_on = NULL; |
23f78d4a IM |
1326 | #endif |
1327 | } | |
1328 | ||
f06febc9 FM |
1329 | /* |
1330 | * Initialize POSIX timer handling for a single task. | |
1331 | */ | |
1332 | static void posix_cpu_timers_init(struct task_struct *tsk) | |
1333 | { | |
64861634 MS |
1334 | tsk->cputime_expires.prof_exp = 0; |
1335 | tsk->cputime_expires.virt_exp = 0; | |
f06febc9 FM |
1336 | tsk->cputime_expires.sched_exp = 0; |
1337 | INIT_LIST_HEAD(&tsk->cpu_timers[0]); | |
1338 | INIT_LIST_HEAD(&tsk->cpu_timers[1]); | |
1339 | INIT_LIST_HEAD(&tsk->cpu_timers[2]); | |
1340 | } | |
1341 | ||
81907739 ON |
1342 | static inline void |
1343 | init_task_pid(struct task_struct *task, enum pid_type type, struct pid *pid) | |
1344 | { | |
1345 | task->pids[type].pid = pid; | |
1346 | } | |
1347 | ||
1da177e4 LT |
1348 | /* |
1349 | * This creates a new process as a copy of the old one, | |
1350 | * but does not actually start it yet. | |
1351 | * | |
1352 | * It copies the registers, and all the appropriate | |
1353 | * parts of the process environment (as per the clone | |
1354 | * flags). The actual kick-off is left to the caller. | |
1355 | */ | |
09dd109d ER |
1356 | static __latent_entropy struct task_struct *copy_process( |
1357 | unsigned long clone_flags, | |
36c8b586 | 1358 | unsigned long stack_start, |
36c8b586 | 1359 | unsigned long stack_size, |
36c8b586 | 1360 | int __user *child_tidptr, |
09a05394 | 1361 | struct pid *pid, |
3033f14a | 1362 | int trace, |
725fc629 AK |
1363 | unsigned long tls, |
1364 | int node) | |
1da177e4 LT |
1365 | { |
1366 | int retval; | |
a24efe62 | 1367 | struct task_struct *p; |
1da177e4 LT |
1368 | |
1369 | if ((clone_flags & (CLONE_NEWNS|CLONE_FS)) == (CLONE_NEWNS|CLONE_FS)) | |
1370 | return ERR_PTR(-EINVAL); | |
1371 | ||
e66eded8 EB |
1372 | if ((clone_flags & (CLONE_NEWUSER|CLONE_FS)) == (CLONE_NEWUSER|CLONE_FS)) |
1373 | return ERR_PTR(-EINVAL); | |
1374 | ||
1da177e4 LT |
1375 | /* |
1376 | * Thread groups must share signals as well, and detached threads | |
1377 | * can only be started up within the thread group. | |
1378 | */ | |
1379 | if ((clone_flags & CLONE_THREAD) && !(clone_flags & CLONE_SIGHAND)) | |
1380 | return ERR_PTR(-EINVAL); | |
1381 | ||
1382 | /* | |
1383 | * Shared signal handlers imply shared VM. By way of the above, | |
1384 | * thread groups also imply shared VM. Blocking this case allows | |
1385 | * for various simplifications in other code. | |
1386 | */ | |
1387 | if ((clone_flags & CLONE_SIGHAND) && !(clone_flags & CLONE_VM)) | |
1388 | return ERR_PTR(-EINVAL); | |
1389 | ||
123be07b SB |
1390 | /* |
1391 | * Siblings of global init remain as zombies on exit since they are | |
1392 | * not reaped by their parent (swapper). To solve this and to avoid | |
1393 | * multi-rooted process trees, prevent global and container-inits | |
1394 | * from creating siblings. | |
1395 | */ | |
1396 | if ((clone_flags & CLONE_PARENT) && | |
1397 | current->signal->flags & SIGNAL_UNKILLABLE) | |
1398 | return ERR_PTR(-EINVAL); | |
1399 | ||
8382fcac | 1400 | /* |
40a0d32d | 1401 | * If the new process will be in a different pid or user namespace |
faf00da5 | 1402 | * do not allow it to share a thread group with the forking task. |
8382fcac | 1403 | */ |
faf00da5 | 1404 | if (clone_flags & CLONE_THREAD) { |
40a0d32d ON |
1405 | if ((clone_flags & (CLONE_NEWUSER | CLONE_NEWPID)) || |
1406 | (task_active_pid_ns(current) != | |
1407 | current->nsproxy->pid_ns_for_children)) | |
1408 | return ERR_PTR(-EINVAL); | |
1409 | } | |
8382fcac | 1410 | |
1da177e4 LT |
1411 | retval = security_task_create(clone_flags); |
1412 | if (retval) | |
1413 | goto fork_out; | |
1414 | ||
1415 | retval = -ENOMEM; | |
725fc629 | 1416 | p = dup_task_struct(current, node); |
1da177e4 LT |
1417 | if (!p) |
1418 | goto fork_out; | |
1419 | ||
f7e8b616 SR |
1420 | ftrace_graph_init_task(p); |
1421 | ||
bea493a0 PZ |
1422 | rt_mutex_init_task(p); |
1423 | ||
d12c1a37 | 1424 | #ifdef CONFIG_PROVE_LOCKING |
de30a2b3 IM |
1425 | DEBUG_LOCKS_WARN_ON(!p->hardirqs_enabled); |
1426 | DEBUG_LOCKS_WARN_ON(!p->softirqs_enabled); | |
1427 | #endif | |
1da177e4 | 1428 | retval = -EAGAIN; |
3b11a1de | 1429 | if (atomic_read(&p->real_cred->user->processes) >= |
78d7d407 | 1430 | task_rlimit(p, RLIMIT_NPROC)) { |
b57922b6 EP |
1431 | if (p->real_cred->user != INIT_USER && |
1432 | !capable(CAP_SYS_RESOURCE) && !capable(CAP_SYS_ADMIN)) | |
1da177e4 LT |
1433 | goto bad_fork_free; |
1434 | } | |
72fa5997 | 1435 | current->flags &= ~PF_NPROC_EXCEEDED; |
1da177e4 | 1436 | |
f1752eec DH |
1437 | retval = copy_creds(p, clone_flags); |
1438 | if (retval < 0) | |
1439 | goto bad_fork_free; | |
1da177e4 LT |
1440 | |
1441 | /* | |
1442 | * If multiple threads are within copy_process(), then this check | |
1443 | * triggers too late. This doesn't hurt, the check is only there | |
1444 | * to stop root fork bombs. | |
1445 | */ | |
04ec93fe | 1446 | retval = -EAGAIN; |
1da177e4 LT |
1447 | if (nr_threads >= max_threads) |
1448 | goto bad_fork_cleanup_count; | |
1449 | ||
ca74e92b | 1450 | delayacct_tsk_init(p); /* Must remain after dup_task_struct() */ |
514ddb44 DR |
1451 | p->flags &= ~(PF_SUPERPRIV | PF_WQ_WORKER); |
1452 | p->flags |= PF_FORKNOEXEC; | |
1da177e4 LT |
1453 | INIT_LIST_HEAD(&p->children); |
1454 | INIT_LIST_HEAD(&p->sibling); | |
f41d911f | 1455 | rcu_copy_process(p); |
1da177e4 LT |
1456 | p->vfork_done = NULL; |
1457 | spin_lock_init(&p->alloc_lock); | |
1da177e4 | 1458 | |
1da177e4 LT |
1459 | init_sigpending(&p->pending); |
1460 | ||
64861634 MS |
1461 | p->utime = p->stime = p->gtime = 0; |
1462 | p->utimescaled = p->stimescaled = 0; | |
9d7fb042 PZ |
1463 | prev_cputime_init(&p->prev_cputime); |
1464 | ||
6a61671b | 1465 | #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN |
b7ce2277 | 1466 | seqcount_init(&p->vtime_seqcount); |
6a61671b | 1467 | p->vtime_snap = 0; |
7098c1ea | 1468 | p->vtime_snap_whence = VTIME_INACTIVE; |
6a61671b FW |
1469 | #endif |
1470 | ||
a3a2e76c KH |
1471 | #if defined(SPLIT_RSS_COUNTING) |
1472 | memset(&p->rss_stat, 0, sizeof(p->rss_stat)); | |
1473 | #endif | |
172ba844 | 1474 | |
6976675d AV |
1475 | p->default_timer_slack_ns = current->timer_slack_ns; |
1476 | ||
5995477a | 1477 | task_io_accounting_init(&p->ioac); |
1da177e4 LT |
1478 | acct_clear_integrals(p); |
1479 | ||
f06febc9 | 1480 | posix_cpu_timers_init(p); |
1da177e4 | 1481 | |
ccbf62d8 | 1482 | p->start_time = ktime_get_ns(); |
57e0be04 | 1483 | p->real_start_time = ktime_get_boot_ns(); |
1da177e4 | 1484 | p->io_context = NULL; |
1da177e4 | 1485 | p->audit_context = NULL; |
b4f48b63 | 1486 | cgroup_fork(p); |
1da177e4 | 1487 | #ifdef CONFIG_NUMA |
846a16bf | 1488 | p->mempolicy = mpol_dup(p->mempolicy); |
fb0a685c DRO |
1489 | if (IS_ERR(p->mempolicy)) { |
1490 | retval = PTR_ERR(p->mempolicy); | |
1491 | p->mempolicy = NULL; | |
e8604cb4 | 1492 | goto bad_fork_cleanup_threadgroup_lock; |
fb0a685c | 1493 | } |
1da177e4 | 1494 | #endif |
778d3b0f MH |
1495 | #ifdef CONFIG_CPUSETS |
1496 | p->cpuset_mem_spread_rotor = NUMA_NO_NODE; | |
1497 | p->cpuset_slab_spread_rotor = NUMA_NO_NODE; | |
cc9a6c87 | 1498 | seqcount_init(&p->mems_allowed_seq); |
778d3b0f | 1499 | #endif |
de30a2b3 IM |
1500 | #ifdef CONFIG_TRACE_IRQFLAGS |
1501 | p->irq_events = 0; | |
1502 | p->hardirqs_enabled = 0; | |
1503 | p->hardirq_enable_ip = 0; | |
1504 | p->hardirq_enable_event = 0; | |
1505 | p->hardirq_disable_ip = _THIS_IP_; | |
1506 | p->hardirq_disable_event = 0; | |
1507 | p->softirqs_enabled = 1; | |
1508 | p->softirq_enable_ip = _THIS_IP_; | |
1509 | p->softirq_enable_event = 0; | |
1510 | p->softirq_disable_ip = 0; | |
1511 | p->softirq_disable_event = 0; | |
1512 | p->hardirq_context = 0; | |
1513 | p->softirq_context = 0; | |
1514 | #endif | |
8bcbde54 DH |
1515 | |
1516 | p->pagefault_disabled = 0; | |
1517 | ||
fbb9ce95 IM |
1518 | #ifdef CONFIG_LOCKDEP |
1519 | p->lockdep_depth = 0; /* no locks held yet */ | |
1520 | p->curr_chain_key = 0; | |
1521 | p->lockdep_recursion = 0; | |
1522 | #endif | |
1da177e4 | 1523 | |
408894ee IM |
1524 | #ifdef CONFIG_DEBUG_MUTEXES |
1525 | p->blocked_on = NULL; /* not blocked yet */ | |
1526 | #endif | |
cafe5635 KO |
1527 | #ifdef CONFIG_BCACHE |
1528 | p->sequential_io = 0; | |
1529 | p->sequential_io_avg = 0; | |
1530 | #endif | |
0f481406 | 1531 | |
3c90e6e9 | 1532 | /* Perform scheduler related setup. Assign this task to a CPU. */ |
aab03e05 DF |
1533 | retval = sched_fork(clone_flags, p); |
1534 | if (retval) | |
1535 | goto bad_fork_cleanup_policy; | |
6ab423e0 | 1536 | |
cdd6c482 | 1537 | retval = perf_event_init_task(p); |
6ab423e0 PZ |
1538 | if (retval) |
1539 | goto bad_fork_cleanup_policy; | |
fb0a685c DRO |
1540 | retval = audit_alloc(p); |
1541 | if (retval) | |
6c72e350 | 1542 | goto bad_fork_cleanup_perf; |
1da177e4 | 1543 | /* copy all the process information */ |
ab602f79 | 1544 | shm_init_task(p); |
fb0a685c DRO |
1545 | retval = copy_semundo(clone_flags, p); |
1546 | if (retval) | |
1da177e4 | 1547 | goto bad_fork_cleanup_audit; |
fb0a685c DRO |
1548 | retval = copy_files(clone_flags, p); |
1549 | if (retval) | |
1da177e4 | 1550 | goto bad_fork_cleanup_semundo; |
fb0a685c DRO |
1551 | retval = copy_fs(clone_flags, p); |
1552 | if (retval) | |
1da177e4 | 1553 | goto bad_fork_cleanup_files; |
fb0a685c DRO |
1554 | retval = copy_sighand(clone_flags, p); |
1555 | if (retval) | |
1da177e4 | 1556 | goto bad_fork_cleanup_fs; |
fb0a685c DRO |
1557 | retval = copy_signal(clone_flags, p); |
1558 | if (retval) | |
1da177e4 | 1559 | goto bad_fork_cleanup_sighand; |
fb0a685c DRO |
1560 | retval = copy_mm(clone_flags, p); |
1561 | if (retval) | |
1da177e4 | 1562 | goto bad_fork_cleanup_signal; |
fb0a685c DRO |
1563 | retval = copy_namespaces(clone_flags, p); |
1564 | if (retval) | |
d84f4f99 | 1565 | goto bad_fork_cleanup_mm; |
fb0a685c DRO |
1566 | retval = copy_io(clone_flags, p); |
1567 | if (retval) | |
fd0928df | 1568 | goto bad_fork_cleanup_namespaces; |
3033f14a | 1569 | retval = copy_thread_tls(clone_flags, stack_start, stack_size, p, tls); |
1da177e4 | 1570 | if (retval) |
fd0928df | 1571 | goto bad_fork_cleanup_io; |
1da177e4 | 1572 | |
425fb2b4 | 1573 | if (pid != &init_struct_pid) { |
c2b1df2e | 1574 | pid = alloc_pid(p->nsproxy->pid_ns_for_children); |
35f71bc0 MH |
1575 | if (IS_ERR(pid)) { |
1576 | retval = PTR_ERR(pid); | |
0740aa5f | 1577 | goto bad_fork_cleanup_thread; |
35f71bc0 | 1578 | } |
425fb2b4 PE |
1579 | } |
1580 | ||
1da177e4 LT |
1581 | p->set_child_tid = (clone_flags & CLONE_CHILD_SETTID) ? child_tidptr : NULL; |
1582 | /* | |
1583 | * Clear TID on mm_release()? | |
1584 | */ | |
fb0a685c | 1585 | p->clear_child_tid = (clone_flags & CLONE_CHILD_CLEARTID) ? child_tidptr : NULL; |
73c10101 JA |
1586 | #ifdef CONFIG_BLOCK |
1587 | p->plug = NULL; | |
1588 | #endif | |
42b2dd0a | 1589 | #ifdef CONFIG_FUTEX |
8f17d3a5 IM |
1590 | p->robust_list = NULL; |
1591 | #ifdef CONFIG_COMPAT | |
1592 | p->compat_robust_list = NULL; | |
1593 | #endif | |
c87e2837 IM |
1594 | INIT_LIST_HEAD(&p->pi_state_list); |
1595 | p->pi_state_cache = NULL; | |
42b2dd0a | 1596 | #endif |
f9a3879a GM |
1597 | /* |
1598 | * sigaltstack should be cleared when sharing the same VM | |
1599 | */ | |
1600 | if ((clone_flags & (CLONE_VM|CLONE_VFORK)) == CLONE_VM) | |
2a742138 | 1601 | sas_ss_reset(p); |
f9a3879a | 1602 | |
1da177e4 | 1603 | /* |
6580807d ON |
1604 | * Syscall tracing and stepping should be turned off in the |
1605 | * child regardless of CLONE_PTRACE. | |
1da177e4 | 1606 | */ |
6580807d | 1607 | user_disable_single_step(p); |
1da177e4 | 1608 | clear_tsk_thread_flag(p, TIF_SYSCALL_TRACE); |
ed75e8d5 LV |
1609 | #ifdef TIF_SYSCALL_EMU |
1610 | clear_tsk_thread_flag(p, TIF_SYSCALL_EMU); | |
1611 | #endif | |
9745512c | 1612 | clear_all_latency_tracing(p); |
1da177e4 | 1613 | |
1da177e4 | 1614 | /* ok, now we should be set up.. */ |
18c830df ON |
1615 | p->pid = pid_nr(pid); |
1616 | if (clone_flags & CLONE_THREAD) { | |
5f8aadd8 | 1617 | p->exit_signal = -1; |
18c830df ON |
1618 | p->group_leader = current->group_leader; |
1619 | p->tgid = current->tgid; | |
1620 | } else { | |
1621 | if (clone_flags & CLONE_PARENT) | |
1622 | p->exit_signal = current->group_leader->exit_signal; | |
1623 | else | |
1624 | p->exit_signal = (clone_flags & CSIGNAL); | |
1625 | p->group_leader = p; | |
1626 | p->tgid = p->pid; | |
1627 | } | |
5f8aadd8 | 1628 | |
9d823e8f WF |
1629 | p->nr_dirtied = 0; |
1630 | p->nr_dirtied_pause = 128 >> (PAGE_SHIFT - 10); | |
83712358 | 1631 | p->dirty_paused_when = 0; |
9d823e8f | 1632 | |
bb8cbbfe | 1633 | p->pdeath_signal = 0; |
47e65328 | 1634 | INIT_LIST_HEAD(&p->thread_group); |
158e1645 | 1635 | p->task_works = NULL; |
1da177e4 | 1636 | |
568ac888 | 1637 | threadgroup_change_begin(current); |
7e47682e AS |
1638 | /* |
1639 | * Ensure that the cgroup subsystem policies allow the new process to be | |
1640 | * forked. It should be noted the the new process's css_set can be changed | |
1641 | * between here and cgroup_post_fork() if an organisation operation is in | |
1642 | * progress. | |
1643 | */ | |
b53202e6 | 1644 | retval = cgroup_can_fork(p); |
7e47682e AS |
1645 | if (retval) |
1646 | goto bad_fork_free_pid; | |
1647 | ||
18c830df ON |
1648 | /* |
1649 | * Make it visible to the rest of the system, but dont wake it up yet. | |
1650 | * Need tasklist lock for parent etc handling! | |
1651 | */ | |
1da177e4 LT |
1652 | write_lock_irq(&tasklist_lock); |
1653 | ||
1da177e4 | 1654 | /* CLONE_PARENT re-uses the old parent */ |
2d5516cb | 1655 | if (clone_flags & (CLONE_PARENT|CLONE_THREAD)) { |
1da177e4 | 1656 | p->real_parent = current->real_parent; |
2d5516cb ON |
1657 | p->parent_exec_id = current->parent_exec_id; |
1658 | } else { | |
1da177e4 | 1659 | p->real_parent = current; |
2d5516cb ON |
1660 | p->parent_exec_id = current->self_exec_id; |
1661 | } | |
1da177e4 | 1662 | |
3f17da69 | 1663 | spin_lock(¤t->sighand->siglock); |
4a2c7a78 | 1664 | |
dbd95212 KC |
1665 | /* |
1666 | * Copy seccomp details explicitly here, in case they were changed | |
1667 | * before holding sighand lock. | |
1668 | */ | |
1669 | copy_seccomp(p); | |
1670 | ||
4a2c7a78 ON |
1671 | /* |
1672 | * Process group and session signals need to be delivered to just the | |
1673 | * parent before the fork or both the parent and the child after the | |
1674 | * fork. Restart if a signal comes in before we add the new process to | |
1675 | * it's process group. | |
1676 | * A fatal signal pending means that current will exit, so the new | |
1677 | * thread can't slip out of an OOM kill (or normal SIGKILL). | |
fb0a685c | 1678 | */ |
23ff4440 | 1679 | recalc_sigpending(); |
4a2c7a78 ON |
1680 | if (signal_pending(current)) { |
1681 | spin_unlock(¤t->sighand->siglock); | |
1682 | write_unlock_irq(&tasklist_lock); | |
1683 | retval = -ERESTARTNOINTR; | |
7e47682e | 1684 | goto bad_fork_cancel_cgroup; |
4a2c7a78 ON |
1685 | } |
1686 | ||
73b9ebfe | 1687 | if (likely(p->pid)) { |
4b9d33e6 | 1688 | ptrace_init_task(p, (clone_flags & CLONE_PTRACE) || trace); |
73b9ebfe | 1689 | |
81907739 | 1690 | init_task_pid(p, PIDTYPE_PID, pid); |
73b9ebfe | 1691 | if (thread_group_leader(p)) { |
81907739 ON |
1692 | init_task_pid(p, PIDTYPE_PGID, task_pgrp(current)); |
1693 | init_task_pid(p, PIDTYPE_SID, task_session(current)); | |
1694 | ||
1c4042c2 | 1695 | if (is_child_reaper(pid)) { |
17cf22c3 | 1696 | ns_of_pid(pid)->child_reaper = p; |
1c4042c2 EB |
1697 | p->signal->flags |= SIGNAL_UNKILLABLE; |
1698 | } | |
73b9ebfe | 1699 | |
fea9d175 | 1700 | p->signal->leader_pid = pid; |
9c9f4ded | 1701 | p->signal->tty = tty_kref_get(current->signal->tty); |
9cd80bbb | 1702 | list_add_tail(&p->sibling, &p->real_parent->children); |
5e85d4ab | 1703 | list_add_tail_rcu(&p->tasks, &init_task.tasks); |
81907739 ON |
1704 | attach_pid(p, PIDTYPE_PGID); |
1705 | attach_pid(p, PIDTYPE_SID); | |
909ea964 | 1706 | __this_cpu_inc(process_counts); |
80628ca0 ON |
1707 | } else { |
1708 | current->signal->nr_threads++; | |
1709 | atomic_inc(¤t->signal->live); | |
1710 | atomic_inc(¤t->signal->sigcnt); | |
80628ca0 ON |
1711 | list_add_tail_rcu(&p->thread_group, |
1712 | &p->group_leader->thread_group); | |
0c740d0a ON |
1713 | list_add_tail_rcu(&p->thread_node, |
1714 | &p->signal->thread_head); | |
73b9ebfe | 1715 | } |
81907739 | 1716 | attach_pid(p, PIDTYPE_PID); |
73b9ebfe | 1717 | nr_threads++; |
1da177e4 LT |
1718 | } |
1719 | ||
1da177e4 | 1720 | total_forks++; |
3f17da69 | 1721 | spin_unlock(¤t->sighand->siglock); |
4af4206b | 1722 | syscall_tracepoint_update(p); |
1da177e4 | 1723 | write_unlock_irq(&tasklist_lock); |
4af4206b | 1724 | |
c13cf856 | 1725 | proc_fork_connector(p); |
b53202e6 | 1726 | cgroup_post_fork(p); |
c9e75f04 | 1727 | threadgroup_change_end(current); |
cdd6c482 | 1728 | perf_event_fork(p); |
43d2b113 KH |
1729 | |
1730 | trace_task_newtask(p, clone_flags); | |
3ab67966 | 1731 | uprobe_copy_process(p, clone_flags); |
43d2b113 | 1732 | |
1da177e4 LT |
1733 | return p; |
1734 | ||
7e47682e | 1735 | bad_fork_cancel_cgroup: |
b53202e6 | 1736 | cgroup_cancel_fork(p); |
425fb2b4 | 1737 | bad_fork_free_pid: |
568ac888 | 1738 | threadgroup_change_end(current); |
425fb2b4 PE |
1739 | if (pid != &init_struct_pid) |
1740 | free_pid(pid); | |
0740aa5f JS |
1741 | bad_fork_cleanup_thread: |
1742 | exit_thread(p); | |
fd0928df | 1743 | bad_fork_cleanup_io: |
b69f2292 LR |
1744 | if (p->io_context) |
1745 | exit_io_context(p); | |
ab516013 | 1746 | bad_fork_cleanup_namespaces: |
444f378b | 1747 | exit_task_namespaces(p); |
1da177e4 | 1748 | bad_fork_cleanup_mm: |
c9f01245 | 1749 | if (p->mm) |
1da177e4 LT |
1750 | mmput(p->mm); |
1751 | bad_fork_cleanup_signal: | |
4ab6c083 | 1752 | if (!(clone_flags & CLONE_THREAD)) |
1c5354de | 1753 | free_signal_struct(p->signal); |
1da177e4 | 1754 | bad_fork_cleanup_sighand: |
a7e5328a | 1755 | __cleanup_sighand(p->sighand); |
1da177e4 LT |
1756 | bad_fork_cleanup_fs: |
1757 | exit_fs(p); /* blocking */ | |
1758 | bad_fork_cleanup_files: | |
1759 | exit_files(p); /* blocking */ | |
1760 | bad_fork_cleanup_semundo: | |
1761 | exit_sem(p); | |
1762 | bad_fork_cleanup_audit: | |
1763 | audit_free(p); | |
6c72e350 | 1764 | bad_fork_cleanup_perf: |
cdd6c482 | 1765 | perf_event_free_task(p); |
6c72e350 | 1766 | bad_fork_cleanup_policy: |
1da177e4 | 1767 | #ifdef CONFIG_NUMA |
f0be3d32 | 1768 | mpol_put(p->mempolicy); |
e8604cb4 | 1769 | bad_fork_cleanup_threadgroup_lock: |
1da177e4 | 1770 | #endif |
35df17c5 | 1771 | delayacct_tsk_free(p); |
1da177e4 | 1772 | bad_fork_cleanup_count: |
d84f4f99 | 1773 | atomic_dec(&p->cred->user->processes); |
e0e81739 | 1774 | exit_creds(p); |
1da177e4 LT |
1775 | bad_fork_free: |
1776 | free_task(p); | |
fe7d37d1 ON |
1777 | fork_out: |
1778 | return ERR_PTR(retval); | |
1da177e4 LT |
1779 | } |
1780 | ||
f106eee1 ON |
1781 | static inline void init_idle_pids(struct pid_link *links) |
1782 | { | |
1783 | enum pid_type type; | |
1784 | ||
1785 | for (type = PIDTYPE_PID; type < PIDTYPE_MAX; ++type) { | |
1786 | INIT_HLIST_NODE(&links[type].node); /* not really needed */ | |
1787 | links[type].pid = &init_struct_pid; | |
1788 | } | |
1789 | } | |
1790 | ||
0db0628d | 1791 | struct task_struct *fork_idle(int cpu) |
1da177e4 | 1792 | { |
36c8b586 | 1793 | struct task_struct *task; |
725fc629 AK |
1794 | task = copy_process(CLONE_VM, 0, 0, NULL, &init_struct_pid, 0, 0, |
1795 | cpu_to_node(cpu)); | |
f106eee1 ON |
1796 | if (!IS_ERR(task)) { |
1797 | init_idle_pids(task->pids); | |
753ca4f3 | 1798 | init_idle(task, cpu); |
f106eee1 | 1799 | } |
73b9ebfe | 1800 | |
1da177e4 LT |
1801 | return task; |
1802 | } | |
1803 | ||
1da177e4 LT |
1804 | /* |
1805 | * Ok, this is the main fork-routine. | |
1806 | * | |
1807 | * It copies the process, and if successful kick-starts | |
1808 | * it and waits for it to finish using the VM if required. | |
1809 | */ | |
3033f14a | 1810 | long _do_fork(unsigned long clone_flags, |
1da177e4 | 1811 | unsigned long stack_start, |
1da177e4 LT |
1812 | unsigned long stack_size, |
1813 | int __user *parent_tidptr, | |
3033f14a JT |
1814 | int __user *child_tidptr, |
1815 | unsigned long tls) | |
1da177e4 LT |
1816 | { |
1817 | struct task_struct *p; | |
1818 | int trace = 0; | |
92476d7f | 1819 | long nr; |
1da177e4 | 1820 | |
09a05394 | 1821 | /* |
4b9d33e6 TH |
1822 | * Determine whether and which event to report to ptracer. When |
1823 | * called from kernel_thread or CLONE_UNTRACED is explicitly | |
1824 | * requested, no event is reported; otherwise, report if the event | |
1825 | * for the type of forking is enabled. | |
09a05394 | 1826 | */ |
e80d6661 | 1827 | if (!(clone_flags & CLONE_UNTRACED)) { |
4b9d33e6 TH |
1828 | if (clone_flags & CLONE_VFORK) |
1829 | trace = PTRACE_EVENT_VFORK; | |
1830 | else if ((clone_flags & CSIGNAL) != SIGCHLD) | |
1831 | trace = PTRACE_EVENT_CLONE; | |
1832 | else | |
1833 | trace = PTRACE_EVENT_FORK; | |
1834 | ||
1835 | if (likely(!ptrace_event_enabled(current, trace))) | |
1836 | trace = 0; | |
1837 | } | |
1da177e4 | 1838 | |
62e791c1 | 1839 | p = copy_process(clone_flags, stack_start, stack_size, |
725fc629 | 1840 | child_tidptr, NULL, trace, tls, NUMA_NO_NODE); |
2a544866 | 1841 | add_latent_entropy(); |
1da177e4 LT |
1842 | /* |
1843 | * Do this prior waking up the new thread - the thread pointer | |
1844 | * might get invalid after that point, if the thread exits quickly. | |
1845 | */ | |
1846 | if (!IS_ERR(p)) { | |
1847 | struct completion vfork; | |
4e52365f | 1848 | struct pid *pid; |
1da177e4 | 1849 | |
0a16b607 MD |
1850 | trace_sched_process_fork(current, p); |
1851 | ||
4e52365f MD |
1852 | pid = get_task_pid(p, PIDTYPE_PID); |
1853 | nr = pid_vnr(pid); | |
30e49c26 PE |
1854 | |
1855 | if (clone_flags & CLONE_PARENT_SETTID) | |
1856 | put_user(nr, parent_tidptr); | |
a6f5e063 | 1857 | |
1da177e4 LT |
1858 | if (clone_flags & CLONE_VFORK) { |
1859 | p->vfork_done = &vfork; | |
1860 | init_completion(&vfork); | |
d68b46fe | 1861 | get_task_struct(p); |
1da177e4 LT |
1862 | } |
1863 | ||
3e51e3ed | 1864 | wake_up_new_task(p); |
1da177e4 | 1865 | |
4b9d33e6 TH |
1866 | /* forking complete and child started to run, tell ptracer */ |
1867 | if (unlikely(trace)) | |
4e52365f | 1868 | ptrace_event_pid(trace, pid); |
09a05394 | 1869 | |
1da177e4 | 1870 | if (clone_flags & CLONE_VFORK) { |
d68b46fe | 1871 | if (!wait_for_vfork_done(p, &vfork)) |
4e52365f | 1872 | ptrace_event_pid(PTRACE_EVENT_VFORK_DONE, pid); |
1da177e4 | 1873 | } |
4e52365f MD |
1874 | |
1875 | put_pid(pid); | |
1da177e4 | 1876 | } else { |
92476d7f | 1877 | nr = PTR_ERR(p); |
1da177e4 | 1878 | } |
92476d7f | 1879 | return nr; |
1da177e4 LT |
1880 | } |
1881 | ||
3033f14a JT |
1882 | #ifndef CONFIG_HAVE_COPY_THREAD_TLS |
1883 | /* For compatibility with architectures that call do_fork directly rather than | |
1884 | * using the syscall entry points below. */ | |
1885 | long do_fork(unsigned long clone_flags, | |
1886 | unsigned long stack_start, | |
1887 | unsigned long stack_size, | |
1888 | int __user *parent_tidptr, | |
1889 | int __user *child_tidptr) | |
1890 | { | |
1891 | return _do_fork(clone_flags, stack_start, stack_size, | |
1892 | parent_tidptr, child_tidptr, 0); | |
1893 | } | |
1894 | #endif | |
1895 | ||
2aa3a7f8 AV |
1896 | /* |
1897 | * Create a kernel thread. | |
1898 | */ | |
1899 | pid_t kernel_thread(int (*fn)(void *), void *arg, unsigned long flags) | |
1900 | { | |
3033f14a JT |
1901 | return _do_fork(flags|CLONE_VM|CLONE_UNTRACED, (unsigned long)fn, |
1902 | (unsigned long)arg, NULL, NULL, 0); | |
2aa3a7f8 | 1903 | } |
2aa3a7f8 | 1904 | |
d2125043 AV |
1905 | #ifdef __ARCH_WANT_SYS_FORK |
1906 | SYSCALL_DEFINE0(fork) | |
1907 | { | |
1908 | #ifdef CONFIG_MMU | |
3033f14a | 1909 | return _do_fork(SIGCHLD, 0, 0, NULL, NULL, 0); |
d2125043 AV |
1910 | #else |
1911 | /* can not support in nommu mode */ | |
5d59e182 | 1912 | return -EINVAL; |
d2125043 AV |
1913 | #endif |
1914 | } | |
1915 | #endif | |
1916 | ||
1917 | #ifdef __ARCH_WANT_SYS_VFORK | |
1918 | SYSCALL_DEFINE0(vfork) | |
1919 | { | |
3033f14a JT |
1920 | return _do_fork(CLONE_VFORK | CLONE_VM | SIGCHLD, 0, |
1921 | 0, NULL, NULL, 0); | |
d2125043 AV |
1922 | } |
1923 | #endif | |
1924 | ||
1925 | #ifdef __ARCH_WANT_SYS_CLONE | |
1926 | #ifdef CONFIG_CLONE_BACKWARDS | |
1927 | SYSCALL_DEFINE5(clone, unsigned long, clone_flags, unsigned long, newsp, | |
1928 | int __user *, parent_tidptr, | |
3033f14a | 1929 | unsigned long, tls, |
d2125043 AV |
1930 | int __user *, child_tidptr) |
1931 | #elif defined(CONFIG_CLONE_BACKWARDS2) | |
1932 | SYSCALL_DEFINE5(clone, unsigned long, newsp, unsigned long, clone_flags, | |
1933 | int __user *, parent_tidptr, | |
1934 | int __user *, child_tidptr, | |
3033f14a | 1935 | unsigned long, tls) |
dfa9771a MS |
1936 | #elif defined(CONFIG_CLONE_BACKWARDS3) |
1937 | SYSCALL_DEFINE6(clone, unsigned long, clone_flags, unsigned long, newsp, | |
1938 | int, stack_size, | |
1939 | int __user *, parent_tidptr, | |
1940 | int __user *, child_tidptr, | |
3033f14a | 1941 | unsigned long, tls) |
d2125043 AV |
1942 | #else |
1943 | SYSCALL_DEFINE5(clone, unsigned long, clone_flags, unsigned long, newsp, | |
1944 | int __user *, parent_tidptr, | |
1945 | int __user *, child_tidptr, | |
3033f14a | 1946 | unsigned long, tls) |
d2125043 AV |
1947 | #endif |
1948 | { | |
3033f14a | 1949 | return _do_fork(clone_flags, newsp, 0, parent_tidptr, child_tidptr, tls); |
d2125043 AV |
1950 | } |
1951 | #endif | |
1952 | ||
5fd63b30 RT |
1953 | #ifndef ARCH_MIN_MMSTRUCT_ALIGN |
1954 | #define ARCH_MIN_MMSTRUCT_ALIGN 0 | |
1955 | #endif | |
1956 | ||
51cc5068 | 1957 | static void sighand_ctor(void *data) |
aa1757f9 ON |
1958 | { |
1959 | struct sighand_struct *sighand = data; | |
1960 | ||
a35afb83 | 1961 | spin_lock_init(&sighand->siglock); |
b8fceee1 | 1962 | init_waitqueue_head(&sighand->signalfd_wqh); |
aa1757f9 ON |
1963 | } |
1964 | ||
1da177e4 LT |
1965 | void __init proc_caches_init(void) |
1966 | { | |
1967 | sighand_cachep = kmem_cache_create("sighand_cache", | |
1968 | sizeof(struct sighand_struct), 0, | |
2dff4405 | 1969 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_DESTROY_BY_RCU| |
5d097056 | 1970 | SLAB_NOTRACK|SLAB_ACCOUNT, sighand_ctor); |
1da177e4 LT |
1971 | signal_cachep = kmem_cache_create("signal_cache", |
1972 | sizeof(struct signal_struct), 0, | |
5d097056 VD |
1973 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK|SLAB_ACCOUNT, |
1974 | NULL); | |
20c2df83 | 1975 | files_cachep = kmem_cache_create("files_cache", |
1da177e4 | 1976 | sizeof(struct files_struct), 0, |
5d097056 VD |
1977 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK|SLAB_ACCOUNT, |
1978 | NULL); | |
20c2df83 | 1979 | fs_cachep = kmem_cache_create("fs_cache", |
1da177e4 | 1980 | sizeof(struct fs_struct), 0, |
5d097056 VD |
1981 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK|SLAB_ACCOUNT, |
1982 | NULL); | |
6345d24d LT |
1983 | /* |
1984 | * FIXME! The "sizeof(struct mm_struct)" currently includes the | |
1985 | * whole struct cpumask for the OFFSTACK case. We could change | |
1986 | * this to *only* allocate as much of it as required by the | |
1987 | * maximum number of CPU's we can ever have. The cpumask_allocation | |
1988 | * is at the end of the structure, exactly for that reason. | |
1989 | */ | |
1da177e4 | 1990 | mm_cachep = kmem_cache_create("mm_struct", |
5fd63b30 | 1991 | sizeof(struct mm_struct), ARCH_MIN_MMSTRUCT_ALIGN, |
5d097056 VD |
1992 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK|SLAB_ACCOUNT, |
1993 | NULL); | |
1994 | vm_area_cachep = KMEM_CACHE(vm_area_struct, SLAB_PANIC|SLAB_ACCOUNT); | |
8feae131 | 1995 | mmap_init(); |
66577193 | 1996 | nsproxy_cache_init(); |
1da177e4 | 1997 | } |
cf2e340f | 1998 | |
cf2e340f | 1999 | /* |
9bfb23fc | 2000 | * Check constraints on flags passed to the unshare system call. |
cf2e340f | 2001 | */ |
9bfb23fc | 2002 | static int check_unshare_flags(unsigned long unshare_flags) |
cf2e340f | 2003 | { |
9bfb23fc ON |
2004 | if (unshare_flags & ~(CLONE_THREAD|CLONE_FS|CLONE_NEWNS|CLONE_SIGHAND| |
2005 | CLONE_VM|CLONE_FILES|CLONE_SYSVSEM| | |
50804fe3 | 2006 | CLONE_NEWUTS|CLONE_NEWIPC|CLONE_NEWNET| |
a79a908f | 2007 | CLONE_NEWUSER|CLONE_NEWPID|CLONE_NEWCGROUP)) |
9bfb23fc | 2008 | return -EINVAL; |
cf2e340f | 2009 | /* |
12c641ab EB |
2010 | * Not implemented, but pretend it works if there is nothing |
2011 | * to unshare. Note that unsharing the address space or the | |
2012 | * signal handlers also need to unshare the signal queues (aka | |
2013 | * CLONE_THREAD). | |
cf2e340f | 2014 | */ |
9bfb23fc | 2015 | if (unshare_flags & (CLONE_THREAD | CLONE_SIGHAND | CLONE_VM)) { |
12c641ab EB |
2016 | if (!thread_group_empty(current)) |
2017 | return -EINVAL; | |
2018 | } | |
2019 | if (unshare_flags & (CLONE_SIGHAND | CLONE_VM)) { | |
2020 | if (atomic_read(¤t->sighand->count) > 1) | |
2021 | return -EINVAL; | |
2022 | } | |
2023 | if (unshare_flags & CLONE_VM) { | |
2024 | if (!current_is_single_threaded()) | |
9bfb23fc ON |
2025 | return -EINVAL; |
2026 | } | |
cf2e340f JD |
2027 | |
2028 | return 0; | |
2029 | } | |
2030 | ||
2031 | /* | |
99d1419d | 2032 | * Unshare the filesystem structure if it is being shared |
cf2e340f JD |
2033 | */ |
2034 | static int unshare_fs(unsigned long unshare_flags, struct fs_struct **new_fsp) | |
2035 | { | |
2036 | struct fs_struct *fs = current->fs; | |
2037 | ||
498052bb AV |
2038 | if (!(unshare_flags & CLONE_FS) || !fs) |
2039 | return 0; | |
2040 | ||
2041 | /* don't need lock here; in the worst case we'll do useless copy */ | |
2042 | if (fs->users == 1) | |
2043 | return 0; | |
2044 | ||
2045 | *new_fsp = copy_fs_struct(fs); | |
2046 | if (!*new_fsp) | |
2047 | return -ENOMEM; | |
cf2e340f JD |
2048 | |
2049 | return 0; | |
2050 | } | |
2051 | ||
cf2e340f | 2052 | /* |
a016f338 | 2053 | * Unshare file descriptor table if it is being shared |
cf2e340f JD |
2054 | */ |
2055 | static int unshare_fd(unsigned long unshare_flags, struct files_struct **new_fdp) | |
2056 | { | |
2057 | struct files_struct *fd = current->files; | |
a016f338 | 2058 | int error = 0; |
cf2e340f JD |
2059 | |
2060 | if ((unshare_flags & CLONE_FILES) && | |
a016f338 JD |
2061 | (fd && atomic_read(&fd->count) > 1)) { |
2062 | *new_fdp = dup_fd(fd, &error); | |
2063 | if (!*new_fdp) | |
2064 | return error; | |
2065 | } | |
cf2e340f JD |
2066 | |
2067 | return 0; | |
2068 | } | |
2069 | ||
cf2e340f JD |
2070 | /* |
2071 | * unshare allows a process to 'unshare' part of the process | |
2072 | * context which was originally shared using clone. copy_* | |
2073 | * functions used by do_fork() cannot be used here directly | |
2074 | * because they modify an inactive task_struct that is being | |
2075 | * constructed. Here we are modifying the current, active, | |
2076 | * task_struct. | |
2077 | */ | |
6559eed8 | 2078 | SYSCALL_DEFINE1(unshare, unsigned long, unshare_flags) |
cf2e340f | 2079 | { |
cf2e340f | 2080 | struct fs_struct *fs, *new_fs = NULL; |
cf2e340f | 2081 | struct files_struct *fd, *new_fd = NULL; |
b2e0d987 | 2082 | struct cred *new_cred = NULL; |
cf7b708c | 2083 | struct nsproxy *new_nsproxy = NULL; |
9edff4ab | 2084 | int do_sysvsem = 0; |
9bfb23fc | 2085 | int err; |
cf2e340f | 2086 | |
b2e0d987 | 2087 | /* |
faf00da5 EB |
2088 | * If unsharing a user namespace must also unshare the thread group |
2089 | * and unshare the filesystem root and working directories. | |
b2e0d987 EB |
2090 | */ |
2091 | if (unshare_flags & CLONE_NEWUSER) | |
e66eded8 | 2092 | unshare_flags |= CLONE_THREAD | CLONE_FS; |
50804fe3 EB |
2093 | /* |
2094 | * If unsharing vm, must also unshare signal handlers. | |
2095 | */ | |
2096 | if (unshare_flags & CLONE_VM) | |
2097 | unshare_flags |= CLONE_SIGHAND; | |
12c641ab EB |
2098 | /* |
2099 | * If unsharing a signal handlers, must also unshare the signal queues. | |
2100 | */ | |
2101 | if (unshare_flags & CLONE_SIGHAND) | |
2102 | unshare_flags |= CLONE_THREAD; | |
9bfb23fc ON |
2103 | /* |
2104 | * If unsharing namespace, must also unshare filesystem information. | |
2105 | */ | |
2106 | if (unshare_flags & CLONE_NEWNS) | |
2107 | unshare_flags |= CLONE_FS; | |
50804fe3 EB |
2108 | |
2109 | err = check_unshare_flags(unshare_flags); | |
2110 | if (err) | |
2111 | goto bad_unshare_out; | |
6013f67f MS |
2112 | /* |
2113 | * CLONE_NEWIPC must also detach from the undolist: after switching | |
2114 | * to a new ipc namespace, the semaphore arrays from the old | |
2115 | * namespace are unreachable. | |
2116 | */ | |
2117 | if (unshare_flags & (CLONE_NEWIPC|CLONE_SYSVSEM)) | |
9edff4ab | 2118 | do_sysvsem = 1; |
fb0a685c DRO |
2119 | err = unshare_fs(unshare_flags, &new_fs); |
2120 | if (err) | |
9bfb23fc | 2121 | goto bad_unshare_out; |
fb0a685c DRO |
2122 | err = unshare_fd(unshare_flags, &new_fd); |
2123 | if (err) | |
9bfb23fc | 2124 | goto bad_unshare_cleanup_fs; |
b2e0d987 | 2125 | err = unshare_userns(unshare_flags, &new_cred); |
fb0a685c | 2126 | if (err) |
9edff4ab | 2127 | goto bad_unshare_cleanup_fd; |
b2e0d987 EB |
2128 | err = unshare_nsproxy_namespaces(unshare_flags, &new_nsproxy, |
2129 | new_cred, new_fs); | |
2130 | if (err) | |
2131 | goto bad_unshare_cleanup_cred; | |
c0b2fc31 | 2132 | |
b2e0d987 | 2133 | if (new_fs || new_fd || do_sysvsem || new_cred || new_nsproxy) { |
9edff4ab MS |
2134 | if (do_sysvsem) { |
2135 | /* | |
2136 | * CLONE_SYSVSEM is equivalent to sys_exit(). | |
2137 | */ | |
2138 | exit_sem(current); | |
2139 | } | |
ab602f79 JM |
2140 | if (unshare_flags & CLONE_NEWIPC) { |
2141 | /* Orphan segments in old ns (see sem above). */ | |
2142 | exit_shm(current); | |
2143 | shm_init_task(current); | |
2144 | } | |
ab516013 | 2145 | |
6f977e6b | 2146 | if (new_nsproxy) |
cf7b708c | 2147 | switch_task_namespaces(current, new_nsproxy); |
cf2e340f | 2148 | |
cf7b708c PE |
2149 | task_lock(current); |
2150 | ||
cf2e340f JD |
2151 | if (new_fs) { |
2152 | fs = current->fs; | |
2a4419b5 | 2153 | spin_lock(&fs->lock); |
cf2e340f | 2154 | current->fs = new_fs; |
498052bb AV |
2155 | if (--fs->users) |
2156 | new_fs = NULL; | |
2157 | else | |
2158 | new_fs = fs; | |
2a4419b5 | 2159 | spin_unlock(&fs->lock); |
cf2e340f JD |
2160 | } |
2161 | ||
cf2e340f JD |
2162 | if (new_fd) { |
2163 | fd = current->files; | |
2164 | current->files = new_fd; | |
2165 | new_fd = fd; | |
2166 | } | |
2167 | ||
2168 | task_unlock(current); | |
b2e0d987 EB |
2169 | |
2170 | if (new_cred) { | |
2171 | /* Install the new user namespace */ | |
2172 | commit_creds(new_cred); | |
2173 | new_cred = NULL; | |
2174 | } | |
cf2e340f JD |
2175 | } |
2176 | ||
b2e0d987 EB |
2177 | bad_unshare_cleanup_cred: |
2178 | if (new_cred) | |
2179 | put_cred(new_cred); | |
cf2e340f JD |
2180 | bad_unshare_cleanup_fd: |
2181 | if (new_fd) | |
2182 | put_files_struct(new_fd); | |
2183 | ||
cf2e340f JD |
2184 | bad_unshare_cleanup_fs: |
2185 | if (new_fs) | |
498052bb | 2186 | free_fs_struct(new_fs); |
cf2e340f | 2187 | |
cf2e340f JD |
2188 | bad_unshare_out: |
2189 | return err; | |
2190 | } | |
3b125388 AV |
2191 | |
2192 | /* | |
2193 | * Helper to unshare the files of the current task. | |
2194 | * We don't want to expose copy_files internals to | |
2195 | * the exec layer of the kernel. | |
2196 | */ | |
2197 | ||
2198 | int unshare_files(struct files_struct **displaced) | |
2199 | { | |
2200 | struct task_struct *task = current; | |
50704516 | 2201 | struct files_struct *copy = NULL; |
3b125388 AV |
2202 | int error; |
2203 | ||
2204 | error = unshare_fd(CLONE_FILES, ©); | |
2205 | if (error || !copy) { | |
2206 | *displaced = NULL; | |
2207 | return error; | |
2208 | } | |
2209 | *displaced = task->files; | |
2210 | task_lock(task); | |
2211 | task->files = copy; | |
2212 | task_unlock(task); | |
2213 | return 0; | |
2214 | } | |
16db3d3f HS |
2215 | |
2216 | int sysctl_max_threads(struct ctl_table *table, int write, | |
2217 | void __user *buffer, size_t *lenp, loff_t *ppos) | |
2218 | { | |
2219 | struct ctl_table t; | |
2220 | int ret; | |
2221 | int threads = max_threads; | |
2222 | int min = MIN_THREADS; | |
2223 | int max = MAX_THREADS; | |
2224 | ||
2225 | t = *table; | |
2226 | t.data = &threads; | |
2227 | t.extra1 = &min; | |
2228 | t.extra2 = &max; | |
2229 | ||
2230 | ret = proc_dointvec_minmax(&t, write, buffer, lenp, ppos); | |
2231 | if (ret || !write) | |
2232 | return ret; | |
2233 | ||
2234 | set_max_threads(threads); | |
2235 | ||
2236 | return 0; | |
2237 | } |