Commit | Line | Data |
---|---|---|
457c8996 | 1 | // SPDX-License-Identifier: GPL-2.0-only |
1da177e4 LT |
2 | /* |
3 | * linux/kernel/fork.c | |
4 | * | |
5 | * Copyright (C) 1991, 1992 Linus Torvalds | |
6 | */ | |
7 | ||
8 | /* | |
9 | * 'fork.c' contains the help-routines for the 'fork' system call | |
10 | * (see also entry.S and others). | |
11 | * Fork is rather simple, once you get the hang of it, but the memory | |
12 | * management can be a bitch. See 'mm/memory.c': 'copy_page_range()' | |
13 | */ | |
14 | ||
b3e58382 | 15 | #include <linux/anon_inodes.h> |
1da177e4 | 16 | #include <linux/slab.h> |
4eb5aaa3 | 17 | #include <linux/sched/autogroup.h> |
6e84f315 | 18 | #include <linux/sched/mm.h> |
f7ccbae4 | 19 | #include <linux/sched/coredump.h> |
8703e8a4 | 20 | #include <linux/sched/user.h> |
6a3827d7 | 21 | #include <linux/sched/numa_balancing.h> |
03441a34 | 22 | #include <linux/sched/stat.h> |
29930025 | 23 | #include <linux/sched/task.h> |
68db0cf1 | 24 | #include <linux/sched/task_stack.h> |
32ef5517 | 25 | #include <linux/sched/cputime.h> |
b3e58382 | 26 | #include <linux/seq_file.h> |
037741a6 | 27 | #include <linux/rtmutex.h> |
1da177e4 LT |
28 | #include <linux/init.h> |
29 | #include <linux/unistd.h> | |
1da177e4 LT |
30 | #include <linux/module.h> |
31 | #include <linux/vmalloc.h> | |
32 | #include <linux/completion.h> | |
1da177e4 LT |
33 | #include <linux/personality.h> |
34 | #include <linux/mempolicy.h> | |
35 | #include <linux/sem.h> | |
36 | #include <linux/file.h> | |
9f3acc31 | 37 | #include <linux/fdtable.h> |
da9cbc87 | 38 | #include <linux/iocontext.h> |
1da177e4 | 39 | #include <linux/key.h> |
50b5e49c | 40 | #include <linux/kmsan.h> |
1da177e4 LT |
41 | #include <linux/binfmts.h> |
42 | #include <linux/mman.h> | |
cddb8a5c | 43 | #include <linux/mmu_notifier.h> |
1da177e4 | 44 | #include <linux/fs.h> |
615d6e87 | 45 | #include <linux/mm.h> |
17fca131 | 46 | #include <linux/mm_inline.h> |
ab516013 | 47 | #include <linux/nsproxy.h> |
c59ede7b | 48 | #include <linux/capability.h> |
1da177e4 | 49 | #include <linux/cpu.h> |
b4f48b63 | 50 | #include <linux/cgroup.h> |
1da177e4 | 51 | #include <linux/security.h> |
a1e78772 | 52 | #include <linux/hugetlb.h> |
e2cfabdf | 53 | #include <linux/seccomp.h> |
1da177e4 LT |
54 | #include <linux/swap.h> |
55 | #include <linux/syscalls.h> | |
56 | #include <linux/jiffies.h> | |
57 | #include <linux/futex.h> | |
8141c7f3 | 58 | #include <linux/compat.h> |
207205a2 | 59 | #include <linux/kthread.h> |
7c3ab738 | 60 | #include <linux/task_io_accounting_ops.h> |
ab2af1f5 | 61 | #include <linux/rcupdate.h> |
1da177e4 LT |
62 | #include <linux/ptrace.h> |
63 | #include <linux/mount.h> | |
64 | #include <linux/audit.h> | |
78fb7466 | 65 | #include <linux/memcontrol.h> |
f201ae23 | 66 | #include <linux/ftrace.h> |
5e2bf014 | 67 | #include <linux/proc_fs.h> |
1da177e4 LT |
68 | #include <linux/profile.h> |
69 | #include <linux/rmap.h> | |
f8af4da3 | 70 | #include <linux/ksm.h> |
1da177e4 | 71 | #include <linux/acct.h> |
893e26e6 | 72 | #include <linux/userfaultfd_k.h> |
8f0ab514 | 73 | #include <linux/tsacct_kern.h> |
9f46080c | 74 | #include <linux/cn_proc.h> |
ba96a0c8 | 75 | #include <linux/freezer.h> |
ca74e92b | 76 | #include <linux/delayacct.h> |
ad4ecbcb | 77 | #include <linux/taskstats_kern.h> |
522ed776 | 78 | #include <linux/tty.h> |
5ad4e53b | 79 | #include <linux/fs_struct.h> |
7c9f8861 | 80 | #include <linux/magic.h> |
cdd6c482 | 81 | #include <linux/perf_event.h> |
42c4ab41 | 82 | #include <linux/posix-timers.h> |
8e7cac79 | 83 | #include <linux/user-return-notifier.h> |
3d5992d2 | 84 | #include <linux/oom.h> |
ba76149f | 85 | #include <linux/khugepaged.h> |
d80e731e | 86 | #include <linux/signalfd.h> |
0326f5a9 | 87 | #include <linux/uprobes.h> |
a27bb332 | 88 | #include <linux/aio.h> |
52f5684c | 89 | #include <linux/compiler.h> |
16db3d3f | 90 | #include <linux/sysctl.h> |
5c9a8750 | 91 | #include <linux/kcov.h> |
d83a7cb3 | 92 | #include <linux/livepatch.h> |
48ac3c18 | 93 | #include <linux/thread_info.h> |
afaef01c | 94 | #include <linux/stackleak.h> |
eafb149e | 95 | #include <linux/kasan.h> |
d08b9f0c | 96 | #include <linux/scs.h> |
0f212204 | 97 | #include <linux/io_uring.h> |
a10787e6 | 98 | #include <linux/bpf.h> |
b3883a9a | 99 | #include <linux/stackprotector.h> |
1da177e4 | 100 | |
1da177e4 | 101 | #include <asm/pgalloc.h> |
7c0f6ba6 | 102 | #include <linux/uaccess.h> |
1da177e4 LT |
103 | #include <asm/mmu_context.h> |
104 | #include <asm/cacheflush.h> | |
105 | #include <asm/tlbflush.h> | |
106 | ||
ad8d75ff SR |
107 | #include <trace/events/sched.h> |
108 | ||
43d2b113 KH |
109 | #define CREATE_TRACE_POINTS |
110 | #include <trace/events/task.h> | |
111 | ||
ac1b398d HS |
112 | /* |
113 | * Minimum number of threads to boot the kernel | |
114 | */ | |
115 | #define MIN_THREADS 20 | |
116 | ||
117 | /* | |
118 | * Maximum number of threads | |
119 | */ | |
120 | #define MAX_THREADS FUTEX_TID_MASK | |
121 | ||
1da177e4 LT |
122 | /* |
123 | * Protected counters by write_lock_irq(&tasklist_lock) | |
124 | */ | |
125 | unsigned long total_forks; /* Handle normal Linux uptimes. */ | |
fb0a685c | 126 | int nr_threads; /* The idle threads do not count.. */ |
1da177e4 | 127 | |
8856ae4d | 128 | static int max_threads; /* tunable limit on nr_threads */ |
1da177e4 | 129 | |
8495f7e6 SPP |
130 | #define NAMED_ARRAY_INDEX(x) [x] = __stringify(x) |
131 | ||
132 | static const char * const resident_page_types[] = { | |
133 | NAMED_ARRAY_INDEX(MM_FILEPAGES), | |
134 | NAMED_ARRAY_INDEX(MM_ANONPAGES), | |
135 | NAMED_ARRAY_INDEX(MM_SWAPENTS), | |
136 | NAMED_ARRAY_INDEX(MM_SHMEMPAGES), | |
137 | }; | |
138 | ||
1da177e4 LT |
139 | DEFINE_PER_CPU(unsigned long, process_counts) = 0; |
140 | ||
c59923a1 | 141 | __cacheline_aligned DEFINE_RWLOCK(tasklist_lock); /* outer */ |
db1466b3 PM |
142 | |
143 | #ifdef CONFIG_PROVE_RCU | |
144 | int lockdep_tasklist_lock_is_held(void) | |
145 | { | |
146 | return lockdep_is_held(&tasklist_lock); | |
147 | } | |
148 | EXPORT_SYMBOL_GPL(lockdep_tasklist_lock_is_held); | |
149 | #endif /* #ifdef CONFIG_PROVE_RCU */ | |
1da177e4 LT |
150 | |
151 | int nr_processes(void) | |
152 | { | |
153 | int cpu; | |
154 | int total = 0; | |
155 | ||
1d510750 | 156 | for_each_possible_cpu(cpu) |
1da177e4 LT |
157 | total += per_cpu(process_counts, cpu); |
158 | ||
159 | return total; | |
160 | } | |
161 | ||
f19b9f74 AM |
162 | void __weak arch_release_task_struct(struct task_struct *tsk) |
163 | { | |
164 | } | |
165 | ||
f5e10287 | 166 | #ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR |
e18b890b | 167 | static struct kmem_cache *task_struct_cachep; |
41101809 TG |
168 | |
169 | static inline struct task_struct *alloc_task_struct_node(int node) | |
170 | { | |
171 | return kmem_cache_alloc_node(task_struct_cachep, GFP_KERNEL, node); | |
172 | } | |
173 | ||
41101809 TG |
174 | static inline void free_task_struct(struct task_struct *tsk) |
175 | { | |
41101809 TG |
176 | kmem_cache_free(task_struct_cachep, tsk); |
177 | } | |
1da177e4 LT |
178 | #endif |
179 | ||
b235beea | 180 | #ifndef CONFIG_ARCH_THREAD_STACK_ALLOCATOR |
41101809 | 181 | |
0d15d74a TG |
182 | /* |
183 | * Allocate pages if THREAD_SIZE is >= PAGE_SIZE, otherwise use a | |
184 | * kmemcache based allocator. | |
185 | */ | |
ba14a194 | 186 | # if THREAD_SIZE >= PAGE_SIZE || defined(CONFIG_VMAP_STACK) |
ac496bf4 | 187 | |
be9a2277 | 188 | # ifdef CONFIG_VMAP_STACK |
ac496bf4 AL |
189 | /* |
190 | * vmalloc() is a bit slow, and calling vfree() enough times will force a TLB | |
191 | * flush. Try to minimize the number of calls by caching stacks. | |
192 | */ | |
193 | #define NR_CACHED_STACKS 2 | |
194 | static DEFINE_PER_CPU(struct vm_struct *, cached_stacks[NR_CACHED_STACKS]); | |
19659c59 | 195 | |
e540bf31 SAS |
196 | struct vm_stack { |
197 | struct rcu_head rcu; | |
198 | struct vm_struct *stack_vm_area; | |
199 | }; | |
200 | ||
201 | static bool try_release_thread_stack_to_cache(struct vm_struct *vm) | |
202 | { | |
203 | unsigned int i; | |
204 | ||
205 | for (i = 0; i < NR_CACHED_STACKS; i++) { | |
206 | if (this_cpu_cmpxchg(cached_stacks[i], NULL, vm) != NULL) | |
207 | continue; | |
208 | return true; | |
209 | } | |
210 | return false; | |
211 | } | |
212 | ||
213 | static void thread_stack_free_rcu(struct rcu_head *rh) | |
214 | { | |
215 | struct vm_stack *vm_stack = container_of(rh, struct vm_stack, rcu); | |
216 | ||
217 | if (try_release_thread_stack_to_cache(vm_stack->stack_vm_area)) | |
218 | return; | |
219 | ||
220 | vfree(vm_stack); | |
221 | } | |
222 | ||
223 | static void thread_stack_delayed_free(struct task_struct *tsk) | |
224 | { | |
225 | struct vm_stack *vm_stack = tsk->stack; | |
226 | ||
227 | vm_stack->stack_vm_area = tsk->stack_vm_area; | |
228 | call_rcu(&vm_stack->rcu, thread_stack_free_rcu); | |
229 | } | |
230 | ||
19659c59 HR |
231 | static int free_vm_stack_cache(unsigned int cpu) |
232 | { | |
233 | struct vm_struct **cached_vm_stacks = per_cpu_ptr(cached_stacks, cpu); | |
234 | int i; | |
235 | ||
236 | for (i = 0; i < NR_CACHED_STACKS; i++) { | |
237 | struct vm_struct *vm_stack = cached_vm_stacks[i]; | |
238 | ||
239 | if (!vm_stack) | |
240 | continue; | |
241 | ||
242 | vfree(vm_stack->addr); | |
243 | cached_vm_stacks[i] = NULL; | |
244 | } | |
245 | ||
246 | return 0; | |
247 | } | |
ac496bf4 | 248 | |
1a03d3f1 | 249 | static int memcg_charge_kernel_stack(struct vm_struct *vm) |
b69c49b7 | 250 | { |
f1c1a9ee SAS |
251 | int i; |
252 | int ret; | |
253 | ||
254 | BUILD_BUG_ON(IS_ENABLED(CONFIG_VMAP_STACK) && PAGE_SIZE % 1024 != 0); | |
255 | BUG_ON(vm->nr_pages != THREAD_SIZE / PAGE_SIZE); | |
256 | ||
257 | for (i = 0; i < THREAD_SIZE / PAGE_SIZE; i++) { | |
258 | ret = memcg_kmem_charge_page(vm->pages[i], GFP_KERNEL, 0); | |
259 | if (ret) | |
260 | goto err; | |
261 | } | |
262 | return 0; | |
263 | err: | |
264 | /* | |
265 | * If memcg_kmem_charge_page() fails, page's memory cgroup pointer is | |
266 | * NULL, and memcg_kmem_uncharge_page() in free_thread_stack() will | |
267 | * ignore this page. | |
268 | */ | |
269 | for (i = 0; i < THREAD_SIZE / PAGE_SIZE; i++) | |
270 | memcg_kmem_uncharge_page(vm->pages[i], 0); | |
271 | return ret; | |
272 | } | |
273 | ||
7865aba3 | 274 | static int alloc_thread_stack_node(struct task_struct *tsk, int node) |
b69c49b7 | 275 | { |
1a03d3f1 | 276 | struct vm_struct *vm; |
ac496bf4 AL |
277 | void *stack; |
278 | int i; | |
279 | ||
ac496bf4 | 280 | for (i = 0; i < NR_CACHED_STACKS; i++) { |
112166f8 CL |
281 | struct vm_struct *s; |
282 | ||
283 | s = this_cpu_xchg(cached_stacks[i], NULL); | |
ac496bf4 AL |
284 | |
285 | if (!s) | |
286 | continue; | |
ac496bf4 | 287 | |
51fb34de | 288 | /* Reset stack metadata. */ |
cebd0eb2 | 289 | kasan_unpoison_range(s->addr, THREAD_SIZE); |
eafb149e | 290 | |
51fb34de AK |
291 | stack = kasan_reset_tag(s->addr); |
292 | ||
ca182551 | 293 | /* Clear stale pointers from reused stack. */ |
51fb34de | 294 | memset(stack, 0, THREAD_SIZE); |
e01e8063 | 295 | |
1a03d3f1 | 296 | if (memcg_charge_kernel_stack(s)) { |
f1c1a9ee SAS |
297 | vfree(s->addr); |
298 | return -ENOMEM; | |
299 | } | |
300 | ||
ac496bf4 | 301 | tsk->stack_vm_area = s; |
51fb34de | 302 | tsk->stack = stack; |
7865aba3 | 303 | return 0; |
ac496bf4 | 304 | } |
ac496bf4 | 305 | |
9b6f7e16 RG |
306 | /* |
307 | * Allocated stacks are cached and later reused by new threads, | |
308 | * so memcg accounting is performed manually on assigning/releasing | |
309 | * stacks to tasks. Drop __GFP_ACCOUNT. | |
310 | */ | |
48ac3c18 | 311 | stack = __vmalloc_node_range(THREAD_SIZE, THREAD_ALIGN, |
ac496bf4 | 312 | VMALLOC_START, VMALLOC_END, |
9b6f7e16 | 313 | THREADINFO_GFP & ~__GFP_ACCOUNT, |
ac496bf4 AL |
314 | PAGE_KERNEL, |
315 | 0, node, __builtin_return_address(0)); | |
7865aba3 SAS |
316 | if (!stack) |
317 | return -ENOMEM; | |
ba14a194 | 318 | |
1a03d3f1 SAS |
319 | vm = find_vm_area(stack); |
320 | if (memcg_charge_kernel_stack(vm)) { | |
f1c1a9ee SAS |
321 | vfree(stack); |
322 | return -ENOMEM; | |
323 | } | |
ba14a194 AL |
324 | /* |
325 | * We can't call find_vm_area() in interrupt context, and | |
326 | * free_thread_stack() can be called in interrupt context, | |
327 | * so cache the vm_struct. | |
328 | */ | |
1a03d3f1 | 329 | tsk->stack_vm_area = vm; |
51fb34de | 330 | stack = kasan_reset_tag(stack); |
7865aba3 SAS |
331 | tsk->stack = stack; |
332 | return 0; | |
b69c49b7 FT |
333 | } |
334 | ||
be9a2277 | 335 | static void free_thread_stack(struct task_struct *tsk) |
b69c49b7 | 336 | { |
e540bf31 SAS |
337 | if (!try_release_thread_stack_to_cache(tsk->stack_vm_area)) |
338 | thread_stack_delayed_free(tsk); | |
9b6f7e16 | 339 | |
be9a2277 SAS |
340 | tsk->stack = NULL; |
341 | tsk->stack_vm_area = NULL; | |
342 | } | |
ac496bf4 | 343 | |
be9a2277 | 344 | # else /* !CONFIG_VMAP_STACK */ |
ac496bf4 | 345 | |
e540bf31 SAS |
346 | static void thread_stack_free_rcu(struct rcu_head *rh) |
347 | { | |
348 | __free_pages(virt_to_page(rh), THREAD_SIZE_ORDER); | |
349 | } | |
350 | ||
351 | static void thread_stack_delayed_free(struct task_struct *tsk) | |
352 | { | |
353 | struct rcu_head *rh = tsk->stack; | |
354 | ||
355 | call_rcu(rh, thread_stack_free_rcu); | |
356 | } | |
357 | ||
7865aba3 | 358 | static int alloc_thread_stack_node(struct task_struct *tsk, int node) |
be9a2277 | 359 | { |
4949148a VD |
360 | struct page *page = alloc_pages_node(node, THREADINFO_GFP, |
361 | THREAD_SIZE_ORDER); | |
b6a84016 | 362 | |
1bf4580e | 363 | if (likely(page)) { |
8dcc1d34 | 364 | tsk->stack = kasan_reset_tag(page_address(page)); |
7865aba3 | 365 | return 0; |
1bf4580e | 366 | } |
7865aba3 | 367 | return -ENOMEM; |
b69c49b7 FT |
368 | } |
369 | ||
be9a2277 | 370 | static void free_thread_stack(struct task_struct *tsk) |
b69c49b7 | 371 | { |
e540bf31 | 372 | thread_stack_delayed_free(tsk); |
be9a2277 | 373 | tsk->stack = NULL; |
b69c49b7 | 374 | } |
ac496bf4 | 375 | |
be9a2277 SAS |
376 | # endif /* CONFIG_VMAP_STACK */ |
377 | # else /* !(THREAD_SIZE >= PAGE_SIZE || defined(CONFIG_VMAP_STACK)) */ | |
9b6f7e16 | 378 | |
b235beea | 379 | static struct kmem_cache *thread_stack_cache; |
ac496bf4 | 380 | |
e540bf31 SAS |
381 | static void thread_stack_free_rcu(struct rcu_head *rh) |
382 | { | |
383 | kmem_cache_free(thread_stack_cache, rh); | |
384 | } | |
ac496bf4 | 385 | |
e540bf31 SAS |
386 | static void thread_stack_delayed_free(struct task_struct *tsk) |
387 | { | |
388 | struct rcu_head *rh = tsk->stack; | |
ac496bf4 | 389 | |
e540bf31 | 390 | call_rcu(rh, thread_stack_free_rcu); |
b69c49b7 | 391 | } |
0d15d74a | 392 | |
7865aba3 | 393 | static int alloc_thread_stack_node(struct task_struct *tsk, int node) |
0d15d74a | 394 | { |
5eed6f1d RR |
395 | unsigned long *stack; |
396 | stack = kmem_cache_alloc_node(thread_stack_cache, THREADINFO_GFP, node); | |
8dcc1d34 | 397 | stack = kasan_reset_tag(stack); |
5eed6f1d | 398 | tsk->stack = stack; |
7865aba3 | 399 | return stack ? 0 : -ENOMEM; |
0d15d74a TG |
400 | } |
401 | ||
ba14a194 | 402 | static void free_thread_stack(struct task_struct *tsk) |
0d15d74a | 403 | { |
e540bf31 | 404 | thread_stack_delayed_free(tsk); |
be9a2277 | 405 | tsk->stack = NULL; |
0d15d74a TG |
406 | } |
407 | ||
b235beea | 408 | void thread_stack_cache_init(void) |
0d15d74a | 409 | { |
f9d29946 DW |
410 | thread_stack_cache = kmem_cache_create_usercopy("thread_stack", |
411 | THREAD_SIZE, THREAD_SIZE, 0, 0, | |
412 | THREAD_SIZE, NULL); | |
b235beea | 413 | BUG_ON(thread_stack_cache == NULL); |
0d15d74a | 414 | } |
be9a2277 SAS |
415 | |
416 | # endif /* THREAD_SIZE >= PAGE_SIZE || defined(CONFIG_VMAP_STACK) */ | |
2bb0529c SAS |
417 | #else /* CONFIG_ARCH_THREAD_STACK_ALLOCATOR */ |
418 | ||
7865aba3 | 419 | static int alloc_thread_stack_node(struct task_struct *tsk, int node) |
2bb0529c SAS |
420 | { |
421 | unsigned long *stack; | |
422 | ||
423 | stack = arch_alloc_thread_stack_node(tsk, node); | |
424 | tsk->stack = stack; | |
7865aba3 | 425 | return stack ? 0 : -ENOMEM; |
2bb0529c SAS |
426 | } |
427 | ||
428 | static void free_thread_stack(struct task_struct *tsk) | |
429 | { | |
430 | arch_free_thread_stack(tsk); | |
431 | tsk->stack = NULL; | |
432 | } | |
433 | ||
be9a2277 | 434 | #endif /* !CONFIG_ARCH_THREAD_STACK_ALLOCATOR */ |
b69c49b7 | 435 | |
1da177e4 | 436 | /* SLAB cache for signal_struct structures (tsk->signal) */ |
e18b890b | 437 | static struct kmem_cache *signal_cachep; |
1da177e4 LT |
438 | |
439 | /* SLAB cache for sighand_struct structures (tsk->sighand) */ | |
e18b890b | 440 | struct kmem_cache *sighand_cachep; |
1da177e4 LT |
441 | |
442 | /* SLAB cache for files_struct structures (tsk->files) */ | |
e18b890b | 443 | struct kmem_cache *files_cachep; |
1da177e4 LT |
444 | |
445 | /* SLAB cache for fs_struct structures (tsk->fs) */ | |
e18b890b | 446 | struct kmem_cache *fs_cachep; |
1da177e4 LT |
447 | |
448 | /* SLAB cache for vm_area_struct structures */ | |
3928d4f5 | 449 | static struct kmem_cache *vm_area_cachep; |
1da177e4 LT |
450 | |
451 | /* SLAB cache for mm_struct structures (tsk->mm) */ | |
e18b890b | 452 | static struct kmem_cache *mm_cachep; |
1da177e4 | 453 | |
490fc053 | 454 | struct vm_area_struct *vm_area_alloc(struct mm_struct *mm) |
3928d4f5 | 455 | { |
a670468f | 456 | struct vm_area_struct *vma; |
490fc053 | 457 | |
a670468f | 458 | vma = kmem_cache_alloc(vm_area_cachep, GFP_KERNEL); |
027232da KS |
459 | if (vma) |
460 | vma_init(vma, mm); | |
490fc053 | 461 | return vma; |
3928d4f5 LT |
462 | } |
463 | ||
464 | struct vm_area_struct *vm_area_dup(struct vm_area_struct *orig) | |
465 | { | |
95faf699 LT |
466 | struct vm_area_struct *new = kmem_cache_alloc(vm_area_cachep, GFP_KERNEL); |
467 | ||
468 | if (new) { | |
cda099b3 QC |
469 | ASSERT_EXCLUSIVE_WRITER(orig->vm_flags); |
470 | ASSERT_EXCLUSIVE_WRITER(orig->vm_file); | |
471 | /* | |
472 | * orig->shared.rb may be modified concurrently, but the clone | |
473 | * will be reinitialized. | |
474 | */ | |
475 | *new = data_race(*orig); | |
95faf699 | 476 | INIT_LIST_HEAD(&new->anon_vma_chain); |
5c26f6ac | 477 | dup_anon_vma_name(orig, new); |
95faf699 LT |
478 | } |
479 | return new; | |
3928d4f5 LT |
480 | } |
481 | ||
482 | void vm_area_free(struct vm_area_struct *vma) | |
483 | { | |
5c26f6ac | 484 | free_anon_vma_name(vma); |
3928d4f5 LT |
485 | kmem_cache_free(vm_area_cachep, vma); |
486 | } | |
487 | ||
ba14a194 | 488 | static void account_kernel_stack(struct task_struct *tsk, int account) |
c6a7f572 | 489 | { |
0ce055f8 SAS |
490 | if (IS_ENABLED(CONFIG_VMAP_STACK)) { |
491 | struct vm_struct *vm = task_stack_vm_area(tsk); | |
27faca83 | 492 | int i; |
ba14a194 | 493 | |
27faca83 MS |
494 | for (i = 0; i < THREAD_SIZE / PAGE_SIZE; i++) |
495 | mod_lruvec_page_state(vm->pages[i], NR_KERNEL_STACK_KB, | |
496 | account * (PAGE_SIZE / 1024)); | |
497 | } else { | |
0ce055f8 SAS |
498 | void *stack = task_stack_page(tsk); |
499 | ||
27faca83 | 500 | /* All stack pages are in the same node. */ |
da3ceeff | 501 | mod_lruvec_kmem_state(stack, NR_KERNEL_STACK_KB, |
991e7673 | 502 | account * (THREAD_SIZE / 1024)); |
27faca83 | 503 | } |
c6a7f572 KM |
504 | } |
505 | ||
1a03d3f1 | 506 | void exit_task_stack_account(struct task_struct *tsk) |
9b6f7e16 | 507 | { |
1a03d3f1 | 508 | account_kernel_stack(tsk, -1); |
991e7673 | 509 | |
1a03d3f1 SAS |
510 | if (IS_ENABLED(CONFIG_VMAP_STACK)) { |
511 | struct vm_struct *vm; | |
9b6f7e16 RG |
512 | int i; |
513 | ||
1a03d3f1 SAS |
514 | vm = task_stack_vm_area(tsk); |
515 | for (i = 0; i < THREAD_SIZE / PAGE_SIZE; i++) | |
516 | memcg_kmem_uncharge_page(vm->pages[i], 0); | |
9b6f7e16 | 517 | } |
9b6f7e16 RG |
518 | } |
519 | ||
68f24b08 | 520 | static void release_task_stack(struct task_struct *tsk) |
1da177e4 | 521 | { |
2f064a59 | 522 | if (WARN_ON(READ_ONCE(tsk->__state) != TASK_DEAD)) |
405c0759 AL |
523 | return; /* Better to leak the stack than to free prematurely */ |
524 | ||
ba14a194 | 525 | free_thread_stack(tsk); |
68f24b08 AL |
526 | } |
527 | ||
528 | #ifdef CONFIG_THREAD_INFO_IN_TASK | |
529 | void put_task_stack(struct task_struct *tsk) | |
530 | { | |
f0b89d39 | 531 | if (refcount_dec_and_test(&tsk->stack_refcount)) |
68f24b08 AL |
532 | release_task_stack(tsk); |
533 | } | |
534 | #endif | |
535 | ||
536 | void free_task(struct task_struct *tsk) | |
537 | { | |
a1140cb2 KI |
538 | #ifdef CONFIG_SECCOMP |
539 | WARN_ON_ONCE(tsk->seccomp.filter); | |
540 | #endif | |
b90ca8ba | 541 | release_user_cpus_ptr(tsk); |
d08b9f0c ST |
542 | scs_release(tsk); |
543 | ||
68f24b08 AL |
544 | #ifndef CONFIG_THREAD_INFO_IN_TASK |
545 | /* | |
546 | * The task is finally done with both the stack and thread_info, | |
547 | * so free both. | |
548 | */ | |
549 | release_task_stack(tsk); | |
550 | #else | |
551 | /* | |
552 | * If the task had a separate stack allocation, it should be gone | |
553 | * by now. | |
554 | */ | |
f0b89d39 | 555 | WARN_ON_ONCE(refcount_read(&tsk->stack_refcount) != 0); |
68f24b08 | 556 | #endif |
23f78d4a | 557 | rt_mutex_debug_task_free(tsk); |
fb52607a | 558 | ftrace_graph_exit_task(tsk); |
f19b9f74 | 559 | arch_release_task_struct(tsk); |
1da5c46f ON |
560 | if (tsk->flags & PF_KTHREAD) |
561 | free_kthread_struct(tsk); | |
1da177e4 LT |
562 | free_task_struct(tsk); |
563 | } | |
564 | EXPORT_SYMBOL(free_task); | |
565 | ||
fe69d560 DH |
566 | static void dup_mm_exe_file(struct mm_struct *mm, struct mm_struct *oldmm) |
567 | { | |
568 | struct file *exe_file; | |
569 | ||
570 | exe_file = get_mm_exe_file(oldmm); | |
571 | RCU_INIT_POINTER(mm->exe_file, exe_file); | |
572 | /* | |
573 | * We depend on the oldmm having properly denied write access to the | |
574 | * exe_file already. | |
575 | */ | |
576 | if (exe_file && deny_write_access(exe_file)) | |
577 | pr_warn_once("deny_write_access() failed in %s\n", __func__); | |
578 | } | |
579 | ||
d70f2a14 AM |
580 | #ifdef CONFIG_MMU |
581 | static __latent_entropy int dup_mmap(struct mm_struct *mm, | |
582 | struct mm_struct *oldmm) | |
583 | { | |
763ecb03 | 584 | struct vm_area_struct *mpnt, *tmp; |
d70f2a14 | 585 | int retval; |
c9dbe82c | 586 | unsigned long charge = 0; |
d70f2a14 | 587 | LIST_HEAD(uf); |
c9dbe82c | 588 | MA_STATE(old_mas, &oldmm->mm_mt, 0, 0); |
d4af56c5 | 589 | MA_STATE(mas, &mm->mm_mt, 0, 0); |
d70f2a14 AM |
590 | |
591 | uprobe_start_dup_mmap(); | |
d8ed45c5 | 592 | if (mmap_write_lock_killable(oldmm)) { |
d70f2a14 AM |
593 | retval = -EINTR; |
594 | goto fail_uprobe_end; | |
595 | } | |
596 | flush_cache_dup_mm(oldmm); | |
597 | uprobe_dup_mmap(oldmm, mm); | |
598 | /* | |
599 | * Not linked in yet - no deadlock potential: | |
600 | */ | |
aaa2cc56 | 601 | mmap_write_lock_nested(mm, SINGLE_DEPTH_NESTING); |
d70f2a14 AM |
602 | |
603 | /* No ordering required: file already has been exposed. */ | |
fe69d560 | 604 | dup_mm_exe_file(mm, oldmm); |
d70f2a14 AM |
605 | |
606 | mm->total_vm = oldmm->total_vm; | |
607 | mm->data_vm = oldmm->data_vm; | |
608 | mm->exec_vm = oldmm->exec_vm; | |
609 | mm->stack_vm = oldmm->stack_vm; | |
610 | ||
d70f2a14 AM |
611 | retval = ksm_fork(mm, oldmm); |
612 | if (retval) | |
613 | goto out; | |
d2081b2b | 614 | khugepaged_fork(mm, oldmm); |
d70f2a14 | 615 | |
c9dbe82c LH |
616 | retval = mas_expected_entries(&mas, oldmm->map_count); |
617 | if (retval) | |
618 | goto out; | |
619 | ||
620 | mas_for_each(&old_mas, mpnt, ULONG_MAX) { | |
d70f2a14 AM |
621 | struct file *file; |
622 | ||
623 | if (mpnt->vm_flags & VM_DONTCOPY) { | |
624 | vm_stat_account(mm, mpnt->vm_flags, -vma_pages(mpnt)); | |
625 | continue; | |
626 | } | |
627 | charge = 0; | |
655c79bb TH |
628 | /* |
629 | * Don't duplicate many vmas if we've been oom-killed (for | |
630 | * example) | |
631 | */ | |
632 | if (fatal_signal_pending(current)) { | |
633 | retval = -EINTR; | |
d4af56c5 | 634 | goto loop_out; |
655c79bb | 635 | } |
d70f2a14 AM |
636 | if (mpnt->vm_flags & VM_ACCOUNT) { |
637 | unsigned long len = vma_pages(mpnt); | |
638 | ||
639 | if (security_vm_enough_memory_mm(oldmm, len)) /* sic */ | |
640 | goto fail_nomem; | |
641 | charge = len; | |
642 | } | |
3928d4f5 | 643 | tmp = vm_area_dup(mpnt); |
d70f2a14 AM |
644 | if (!tmp) |
645 | goto fail_nomem; | |
d70f2a14 AM |
646 | retval = vma_dup_policy(mpnt, tmp); |
647 | if (retval) | |
648 | goto fail_nomem_policy; | |
649 | tmp->vm_mm = mm; | |
650 | retval = dup_userfaultfd(tmp, &uf); | |
651 | if (retval) | |
652 | goto fail_nomem_anon_vma_fork; | |
653 | if (tmp->vm_flags & VM_WIPEONFORK) { | |
93949bb2 LX |
654 | /* |
655 | * VM_WIPEONFORK gets a clean slate in the child. | |
656 | * Don't prepare anon_vma until fault since we don't | |
657 | * copy page for current vma. | |
658 | */ | |
d70f2a14 | 659 | tmp->anon_vma = NULL; |
d70f2a14 AM |
660 | } else if (anon_vma_fork(tmp, mpnt)) |
661 | goto fail_nomem_anon_vma_fork; | |
662 | tmp->vm_flags &= ~(VM_LOCKED | VM_LOCKONFAULT); | |
d70f2a14 AM |
663 | file = tmp->vm_file; |
664 | if (file) { | |
d70f2a14 AM |
665 | struct address_space *mapping = file->f_mapping; |
666 | ||
667 | get_file(file); | |
d70f2a14 AM |
668 | i_mmap_lock_write(mapping); |
669 | if (tmp->vm_flags & VM_SHARED) | |
cf508b58 | 670 | mapping_allow_writable(mapping); |
d70f2a14 AM |
671 | flush_dcache_mmap_lock(mapping); |
672 | /* insert tmp into the share list, just after mpnt */ | |
673 | vma_interval_tree_insert_after(tmp, mpnt, | |
674 | &mapping->i_mmap); | |
675 | flush_dcache_mmap_unlock(mapping); | |
676 | i_mmap_unlock_write(mapping); | |
677 | } | |
678 | ||
679 | /* | |
8d9bfb26 | 680 | * Copy/update hugetlb private vma information. |
d70f2a14 AM |
681 | */ |
682 | if (is_vm_hugetlb_page(tmp)) | |
8d9bfb26 | 683 | hugetlb_dup_vma_private(tmp); |
d70f2a14 | 684 | |
d4af56c5 LH |
685 | /* Link the vma into the MT */ |
686 | mas.index = tmp->vm_start; | |
687 | mas.last = tmp->vm_end - 1; | |
688 | mas_store(&mas, tmp); | |
c9dbe82c LH |
689 | if (mas_is_err(&mas)) |
690 | goto fail_nomem_mas_store; | |
d70f2a14 AM |
691 | |
692 | mm->map_count++; | |
693 | if (!(tmp->vm_flags & VM_WIPEONFORK)) | |
c78f4636 | 694 | retval = copy_page_range(tmp, mpnt); |
d70f2a14 AM |
695 | |
696 | if (tmp->vm_ops && tmp->vm_ops->open) | |
697 | tmp->vm_ops->open(tmp); | |
698 | ||
699 | if (retval) | |
d4af56c5 | 700 | goto loop_out; |
d70f2a14 AM |
701 | } |
702 | /* a new mm has just been created */ | |
1ed0cc5a | 703 | retval = arch_dup_mmap(oldmm, mm); |
d4af56c5 LH |
704 | loop_out: |
705 | mas_destroy(&mas); | |
d70f2a14 | 706 | out: |
d8ed45c5 | 707 | mmap_write_unlock(mm); |
d70f2a14 | 708 | flush_tlb_mm(oldmm); |
d8ed45c5 | 709 | mmap_write_unlock(oldmm); |
d70f2a14 AM |
710 | dup_userfaultfd_complete(&uf); |
711 | fail_uprobe_end: | |
712 | uprobe_end_dup_mmap(); | |
713 | return retval; | |
c9dbe82c LH |
714 | |
715 | fail_nomem_mas_store: | |
716 | unlink_anon_vmas(tmp); | |
d70f2a14 AM |
717 | fail_nomem_anon_vma_fork: |
718 | mpol_put(vma_policy(tmp)); | |
719 | fail_nomem_policy: | |
3928d4f5 | 720 | vm_area_free(tmp); |
d70f2a14 AM |
721 | fail_nomem: |
722 | retval = -ENOMEM; | |
723 | vm_unacct_memory(charge); | |
d4af56c5 | 724 | goto loop_out; |
d70f2a14 AM |
725 | } |
726 | ||
727 | static inline int mm_alloc_pgd(struct mm_struct *mm) | |
728 | { | |
729 | mm->pgd = pgd_alloc(mm); | |
730 | if (unlikely(!mm->pgd)) | |
731 | return -ENOMEM; | |
732 | return 0; | |
733 | } | |
734 | ||
735 | static inline void mm_free_pgd(struct mm_struct *mm) | |
736 | { | |
737 | pgd_free(mm, mm->pgd); | |
738 | } | |
739 | #else | |
740 | static int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm) | |
741 | { | |
d8ed45c5 | 742 | mmap_write_lock(oldmm); |
fe69d560 | 743 | dup_mm_exe_file(mm, oldmm); |
d8ed45c5 | 744 | mmap_write_unlock(oldmm); |
d70f2a14 AM |
745 | return 0; |
746 | } | |
747 | #define mm_alloc_pgd(mm) (0) | |
748 | #define mm_free_pgd(mm) | |
749 | #endif /* CONFIG_MMU */ | |
750 | ||
751 | static void check_mm(struct mm_struct *mm) | |
752 | { | |
753 | int i; | |
754 | ||
8495f7e6 SPP |
755 | BUILD_BUG_ON_MSG(ARRAY_SIZE(resident_page_types) != NR_MM_COUNTERS, |
756 | "Please make sure 'struct resident_page_types[]' is updated as well"); | |
757 | ||
d70f2a14 | 758 | for (i = 0; i < NR_MM_COUNTERS; i++) { |
f1a79412 | 759 | long x = percpu_counter_sum(&mm->rss_stat[i]); |
d70f2a14 | 760 | |
f689054a SB |
761 | if (likely(!x)) |
762 | continue; | |
763 | ||
764 | /* Making sure this is not due to race with CPU offlining. */ | |
765 | x = percpu_counter_sum_all(&mm->rss_stat[i]); | |
d70f2a14 | 766 | if (unlikely(x)) |
8495f7e6 SPP |
767 | pr_alert("BUG: Bad rss-counter state mm:%p type:%s val:%ld\n", |
768 | mm, resident_page_types[i], x); | |
d70f2a14 AM |
769 | } |
770 | ||
771 | if (mm_pgtables_bytes(mm)) | |
772 | pr_alert("BUG: non-zero pgtables_bytes on freeing mm: %ld\n", | |
773 | mm_pgtables_bytes(mm)); | |
774 | ||
775 | #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS | |
776 | VM_BUG_ON_MM(mm->pmd_huge_pte, mm); | |
777 | #endif | |
778 | } | |
779 | ||
780 | #define allocate_mm() (kmem_cache_alloc(mm_cachep, GFP_KERNEL)) | |
781 | #define free_mm(mm) (kmem_cache_free(mm_cachep, (mm))) | |
782 | ||
783 | /* | |
784 | * Called when the last reference to the mm | |
785 | * is dropped: either by a lazy thread or by | |
786 | * mmput. Free the page directory and the mm. | |
787 | */ | |
d34bc48f | 788 | void __mmdrop(struct mm_struct *mm) |
d70f2a14 | 789 | { |
f1a79412 SB |
790 | int i; |
791 | ||
d70f2a14 | 792 | BUG_ON(mm == &init_mm); |
3eda69c9 MR |
793 | WARN_ON_ONCE(mm == current->mm); |
794 | WARN_ON_ONCE(mm == current->active_mm); | |
d70f2a14 AM |
795 | mm_free_pgd(mm); |
796 | destroy_context(mm); | |
984cfe4e | 797 | mmu_notifier_subscriptions_destroy(mm); |
d70f2a14 AM |
798 | check_mm(mm); |
799 | put_user_ns(mm->user_ns); | |
2667ed10 | 800 | mm_pasid_drop(mm); |
f1a79412 SB |
801 | |
802 | for (i = 0; i < NR_MM_COUNTERS; i++) | |
803 | percpu_counter_destroy(&mm->rss_stat[i]); | |
d70f2a14 AM |
804 | free_mm(mm); |
805 | } | |
d34bc48f | 806 | EXPORT_SYMBOL_GPL(__mmdrop); |
d70f2a14 AM |
807 | |
808 | static void mmdrop_async_fn(struct work_struct *work) | |
809 | { | |
810 | struct mm_struct *mm; | |
811 | ||
812 | mm = container_of(work, struct mm_struct, async_put_work); | |
813 | __mmdrop(mm); | |
814 | } | |
815 | ||
816 | static void mmdrop_async(struct mm_struct *mm) | |
817 | { | |
818 | if (unlikely(atomic_dec_and_test(&mm->mm_count))) { | |
819 | INIT_WORK(&mm->async_put_work, mmdrop_async_fn); | |
820 | schedule_work(&mm->async_put_work); | |
821 | } | |
822 | } | |
823 | ||
ea6d290c ON |
824 | static inline void free_signal_struct(struct signal_struct *sig) |
825 | { | |
97101eb4 | 826 | taskstats_tgid_free(sig); |
1c5354de | 827 | sched_autogroup_exit(sig); |
7283094e MH |
828 | /* |
829 | * __mmdrop is not safe to call from softirq context on x86 due to | |
830 | * pgd_dtor so postpone it to the async context | |
831 | */ | |
26db62f1 | 832 | if (sig->oom_mm) |
7283094e | 833 | mmdrop_async(sig->oom_mm); |
ea6d290c ON |
834 | kmem_cache_free(signal_cachep, sig); |
835 | } | |
836 | ||
837 | static inline void put_signal_struct(struct signal_struct *sig) | |
838 | { | |
60d4de3f | 839 | if (refcount_dec_and_test(&sig->sigcnt)) |
ea6d290c ON |
840 | free_signal_struct(sig); |
841 | } | |
842 | ||
158d9ebd | 843 | void __put_task_struct(struct task_struct *tsk) |
1da177e4 | 844 | { |
270f722d | 845 | WARN_ON(!tsk->exit_state); |
ec1d2819 | 846 | WARN_ON(refcount_read(&tsk->usage)); |
1da177e4 LT |
847 | WARN_ON(tsk == current); |
848 | ||
0f212204 | 849 | io_uring_free(tsk); |
2e91fa7f | 850 | cgroup_free(tsk); |
16d51a59 | 851 | task_numa_free(tsk, true); |
1a2a4d06 | 852 | security_task_free(tsk); |
a10787e6 | 853 | bpf_task_storage_free(tsk); |
e0e81739 | 854 | exit_creds(tsk); |
35df17c5 | 855 | delayacct_tsk_free(tsk); |
ea6d290c | 856 | put_signal_struct(tsk->signal); |
6e33cad0 | 857 | sched_core_free(tsk); |
2873cd31 | 858 | free_task(tsk); |
1da177e4 | 859 | } |
77c100c8 | 860 | EXPORT_SYMBOL_GPL(__put_task_struct); |
1da177e4 | 861 | |
6c0a9fa6 | 862 | void __init __weak arch_task_cache_init(void) { } |
61c4628b | 863 | |
ff691f6e HS |
864 | /* |
865 | * set_max_threads | |
866 | */ | |
16db3d3f | 867 | static void set_max_threads(unsigned int max_threads_suggested) |
ff691f6e | 868 | { |
ac1b398d | 869 | u64 threads; |
ca79b0c2 | 870 | unsigned long nr_pages = totalram_pages(); |
ff691f6e HS |
871 | |
872 | /* | |
ac1b398d HS |
873 | * The number of threads shall be limited such that the thread |
874 | * structures may only consume a small part of the available memory. | |
ff691f6e | 875 | */ |
3d6357de | 876 | if (fls64(nr_pages) + fls64(PAGE_SIZE) > 64) |
ac1b398d HS |
877 | threads = MAX_THREADS; |
878 | else | |
3d6357de | 879 | threads = div64_u64((u64) nr_pages * (u64) PAGE_SIZE, |
ac1b398d HS |
880 | (u64) THREAD_SIZE * 8UL); |
881 | ||
16db3d3f HS |
882 | if (threads > max_threads_suggested) |
883 | threads = max_threads_suggested; | |
884 | ||
ac1b398d | 885 | max_threads = clamp_t(u64, threads, MIN_THREADS, MAX_THREADS); |
ff691f6e HS |
886 | } |
887 | ||
5aaeb5c0 IM |
888 | #ifdef CONFIG_ARCH_WANTS_DYNAMIC_TASK_STRUCT |
889 | /* Initialized by the architecture: */ | |
890 | int arch_task_struct_size __read_mostly; | |
891 | #endif | |
0c8c0f03 | 892 | |
4189ff23 | 893 | #ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR |
5905429a KC |
894 | static void task_struct_whitelist(unsigned long *offset, unsigned long *size) |
895 | { | |
896 | /* Fetch thread_struct whitelist for the architecture. */ | |
897 | arch_thread_struct_whitelist(offset, size); | |
898 | ||
899 | /* | |
900 | * Handle zero-sized whitelist or empty thread_struct, otherwise | |
901 | * adjust offset to position of thread_struct in task_struct. | |
902 | */ | |
903 | if (unlikely(*size == 0)) | |
904 | *offset = 0; | |
905 | else | |
906 | *offset += offsetof(struct task_struct, thread); | |
907 | } | |
4189ff23 | 908 | #endif /* CONFIG_ARCH_TASK_STRUCT_ALLOCATOR */ |
5905429a | 909 | |
ff691f6e | 910 | void __init fork_init(void) |
1da177e4 | 911 | { |
25f9c081 | 912 | int i; |
f5e10287 | 913 | #ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR |
1da177e4 | 914 | #ifndef ARCH_MIN_TASKALIGN |
e274795e | 915 | #define ARCH_MIN_TASKALIGN 0 |
1da177e4 | 916 | #endif |
95cb64c1 | 917 | int align = max_t(int, L1_CACHE_BYTES, ARCH_MIN_TASKALIGN); |
5905429a | 918 | unsigned long useroffset, usersize; |
e274795e | 919 | |
1da177e4 | 920 | /* create a slab on which task_structs can be allocated */ |
5905429a KC |
921 | task_struct_whitelist(&useroffset, &usersize); |
922 | task_struct_cachep = kmem_cache_create_usercopy("task_struct", | |
e274795e | 923 | arch_task_struct_size, align, |
5905429a KC |
924 | SLAB_PANIC|SLAB_ACCOUNT, |
925 | useroffset, usersize, NULL); | |
1da177e4 LT |
926 | #endif |
927 | ||
61c4628b SS |
928 | /* do the arch specific task caches init */ |
929 | arch_task_cache_init(); | |
930 | ||
16db3d3f | 931 | set_max_threads(MAX_THREADS); |
1da177e4 LT |
932 | |
933 | init_task.signal->rlim[RLIMIT_NPROC].rlim_cur = max_threads/2; | |
934 | init_task.signal->rlim[RLIMIT_NPROC].rlim_max = max_threads/2; | |
935 | init_task.signal->rlim[RLIMIT_SIGPENDING] = | |
936 | init_task.signal->rlim[RLIMIT_NPROC]; | |
b376c3e1 | 937 | |
de399236 | 938 | for (i = 0; i < UCOUNT_COUNTS; i++) |
25f9c081 | 939 | init_user_ns.ucount_max[i] = max_threads/2; |
19659c59 | 940 | |
de399236 AG |
941 | set_userns_rlimit_max(&init_user_ns, UCOUNT_RLIMIT_NPROC, RLIM_INFINITY); |
942 | set_userns_rlimit_max(&init_user_ns, UCOUNT_RLIMIT_MSGQUEUE, RLIM_INFINITY); | |
943 | set_userns_rlimit_max(&init_user_ns, UCOUNT_RLIMIT_SIGPENDING, RLIM_INFINITY); | |
944 | set_userns_rlimit_max(&init_user_ns, UCOUNT_RLIMIT_MEMLOCK, RLIM_INFINITY); | |
21d1c5e3 | 945 | |
19659c59 HR |
946 | #ifdef CONFIG_VMAP_STACK |
947 | cpuhp_setup_state(CPUHP_BP_PREPARE_DYN, "fork:vm_stack_cache", | |
948 | NULL, free_vm_stack_cache); | |
949 | #endif | |
b09be676 | 950 | |
d08b9f0c ST |
951 | scs_init(); |
952 | ||
b09be676 | 953 | lockdep_init_task(&init_task); |
aad42dd4 | 954 | uprobes_init(); |
1da177e4 LT |
955 | } |
956 | ||
52f5684c | 957 | int __weak arch_dup_task_struct(struct task_struct *dst, |
61c4628b SS |
958 | struct task_struct *src) |
959 | { | |
960 | *dst = *src; | |
961 | return 0; | |
962 | } | |
963 | ||
d4311ff1 AT |
964 | void set_task_stack_end_magic(struct task_struct *tsk) |
965 | { | |
966 | unsigned long *stackend; | |
967 | ||
968 | stackend = end_of_stack(tsk); | |
969 | *stackend = STACK_END_MAGIC; /* for overflow detection */ | |
970 | } | |
971 | ||
725fc629 | 972 | static struct task_struct *dup_task_struct(struct task_struct *orig, int node) |
1da177e4 LT |
973 | { |
974 | struct task_struct *tsk; | |
3e26c149 | 975 | int err; |
1da177e4 | 976 | |
725fc629 AK |
977 | if (node == NUMA_NO_NODE) |
978 | node = tsk_fork_get_node(orig); | |
504f52b5 | 979 | tsk = alloc_task_struct_node(node); |
1da177e4 LT |
980 | if (!tsk) |
981 | return NULL; | |
982 | ||
546c42b2 SAS |
983 | err = arch_dup_task_struct(tsk, orig); |
984 | if (err) | |
f19b9f74 | 985 | goto free_tsk; |
1da177e4 | 986 | |
7865aba3 SAS |
987 | err = alloc_thread_stack_node(tsk, node); |
988 | if (err) | |
f19b9f74 | 989 | goto free_tsk; |
ba14a194 | 990 | |
68f24b08 | 991 | #ifdef CONFIG_THREAD_INFO_IN_TASK |
f0b89d39 | 992 | refcount_set(&tsk->stack_refcount, 1); |
68f24b08 | 993 | #endif |
1a03d3f1 | 994 | account_kernel_stack(tsk, 1); |
164c33c6 | 995 | |
d08b9f0c ST |
996 | err = scs_prepare(tsk, node); |
997 | if (err) | |
998 | goto free_stack; | |
999 | ||
dbd95212 KC |
1000 | #ifdef CONFIG_SECCOMP |
1001 | /* | |
1002 | * We must handle setting up seccomp filters once we're under | |
1003 | * the sighand lock in case orig has changed between now and | |
1004 | * then. Until then, filter must be NULL to avoid messing up | |
1005 | * the usage counts on the error path calling free_task. | |
1006 | */ | |
1007 | tsk->seccomp.filter = NULL; | |
1008 | #endif | |
87bec58a AM |
1009 | |
1010 | setup_thread_stack(tsk, orig); | |
8e7cac79 | 1011 | clear_user_return_notifier(tsk); |
f26f9aff | 1012 | clear_tsk_need_resched(tsk); |
d4311ff1 | 1013 | set_task_stack_end_magic(tsk); |
1446e1df | 1014 | clear_syscall_work_syscall_user_dispatch(tsk); |
1da177e4 | 1015 | |
050e9baa | 1016 | #ifdef CONFIG_STACKPROTECTOR |
7cd815bc | 1017 | tsk->stack_canary = get_random_canary(); |
0a425405 | 1018 | #endif |
3bd37062 SAS |
1019 | if (orig->cpus_ptr == &orig->cpus_mask) |
1020 | tsk->cpus_ptr = &tsk->cpus_mask; | |
b90ca8ba | 1021 | dup_user_cpus_ptr(tsk, orig, node); |
0a425405 | 1022 | |
fb0a685c | 1023 | /* |
0ff7b2cf EB |
1024 | * One for the user space visible state that goes away when reaped. |
1025 | * One for the scheduler. | |
fb0a685c | 1026 | */ |
0ff7b2cf EB |
1027 | refcount_set(&tsk->rcu_users, 2); |
1028 | /* One for the rcu users */ | |
1029 | refcount_set(&tsk->usage, 1); | |
6c5c9341 | 1030 | #ifdef CONFIG_BLK_DEV_IO_TRACE |
2056a782 | 1031 | tsk->btrace_seq = 0; |
6c5c9341 | 1032 | #endif |
a0aa7f68 | 1033 | tsk->splice_pipe = NULL; |
5640f768 | 1034 | tsk->task_frag.page = NULL; |
093e5840 | 1035 | tsk->wake_q.next = NULL; |
e32cf5df | 1036 | tsk->worker_private = NULL; |
c6a7f572 | 1037 | |
5c9a8750 | 1038 | kcov_task_init(tsk); |
50b5e49c | 1039 | kmsan_task_create(tsk); |
5fbda3ec | 1040 | kmap_local_fork(tsk); |
5c9a8750 | 1041 | |
e41d5818 DV |
1042 | #ifdef CONFIG_FAULT_INJECTION |
1043 | tsk->fail_nth = 0; | |
1044 | #endif | |
1045 | ||
2c323017 JB |
1046 | #ifdef CONFIG_BLK_CGROUP |
1047 | tsk->throttle_queue = NULL; | |
1048 | tsk->use_memdelay = 0; | |
1049 | #endif | |
1050 | ||
a3d29e82 PZ |
1051 | #ifdef CONFIG_IOMMU_SVA |
1052 | tsk->pasid_activated = 0; | |
1053 | #endif | |
1054 | ||
d46eb14b SB |
1055 | #ifdef CONFIG_MEMCG |
1056 | tsk->active_memcg = NULL; | |
1057 | #endif | |
b041b525 TL |
1058 | |
1059 | #ifdef CONFIG_CPU_SUP_INTEL | |
1060 | tsk->reported_split_lock = 0; | |
1061 | #endif | |
1062 | ||
1da177e4 | 1063 | return tsk; |
61c4628b | 1064 | |
b235beea | 1065 | free_stack: |
1a03d3f1 | 1066 | exit_task_stack_account(tsk); |
ba14a194 | 1067 | free_thread_stack(tsk); |
f19b9f74 | 1068 | free_tsk: |
61c4628b SS |
1069 | free_task_struct(tsk); |
1070 | return NULL; | |
1da177e4 LT |
1071 | } |
1072 | ||
23ff4440 | 1073 | __cacheline_aligned_in_smp DEFINE_SPINLOCK(mmlist_lock); |
1da177e4 | 1074 | |
4cb0e11b HK |
1075 | static unsigned long default_dump_filter = MMF_DUMP_FILTER_DEFAULT; |
1076 | ||
1077 | static int __init coredump_filter_setup(char *s) | |
1078 | { | |
1079 | default_dump_filter = | |
1080 | (simple_strtoul(s, NULL, 0) << MMF_DUMP_FILTER_SHIFT) & | |
1081 | MMF_DUMP_FILTER_MASK; | |
1082 | return 1; | |
1083 | } | |
1084 | ||
1085 | __setup("coredump_filter=", coredump_filter_setup); | |
1086 | ||
1da177e4 LT |
1087 | #include <linux/init_task.h> |
1088 | ||
858f0993 AD |
1089 | static void mm_init_aio(struct mm_struct *mm) |
1090 | { | |
1091 | #ifdef CONFIG_AIO | |
1092 | spin_lock_init(&mm->ioctx_lock); | |
db446a08 | 1093 | mm->ioctx_table = NULL; |
858f0993 AD |
1094 | #endif |
1095 | } | |
1096 | ||
c3f3ce04 AA |
1097 | static __always_inline void mm_clear_owner(struct mm_struct *mm, |
1098 | struct task_struct *p) | |
1099 | { | |
1100 | #ifdef CONFIG_MEMCG | |
1101 | if (mm->owner == p) | |
1102 | WRITE_ONCE(mm->owner, NULL); | |
1103 | #endif | |
1104 | } | |
1105 | ||
33144e84 VD |
1106 | static void mm_init_owner(struct mm_struct *mm, struct task_struct *p) |
1107 | { | |
1108 | #ifdef CONFIG_MEMCG | |
1109 | mm->owner = p; | |
1110 | #endif | |
1111 | } | |
1112 | ||
355627f5 EB |
1113 | static void mm_init_uprobes_state(struct mm_struct *mm) |
1114 | { | |
1115 | #ifdef CONFIG_UPROBES | |
1116 | mm->uprobes_state.xol_area = NULL; | |
1117 | #endif | |
1118 | } | |
1119 | ||
bfedb589 EB |
1120 | static struct mm_struct *mm_init(struct mm_struct *mm, struct task_struct *p, |
1121 | struct user_namespace *user_ns) | |
1da177e4 | 1122 | { |
f1a79412 SB |
1123 | int i; |
1124 | ||
d4af56c5 LH |
1125 | mt_init_flags(&mm->mm_mt, MM_MT_FLAGS); |
1126 | mt_set_external_lock(&mm->mm_mt, &mm->mmap_lock); | |
1da177e4 LT |
1127 | atomic_set(&mm->mm_users, 1); |
1128 | atomic_set(&mm->mm_count, 1); | |
57efa1fe | 1129 | seqcount_init(&mm->write_protect_seq); |
d8ed45c5 | 1130 | mmap_init_lock(mm); |
1da177e4 | 1131 | INIT_LIST_HEAD(&mm->mmlist); |
af5b0f6a | 1132 | mm_pgtables_bytes_init(mm); |
41f727fd VD |
1133 | mm->map_count = 0; |
1134 | mm->locked_vm = 0; | |
70f8a3ca | 1135 | atomic64_set(&mm->pinned_vm, 0); |
d559db08 | 1136 | memset(&mm->rss_stat, 0, sizeof(mm->rss_stat)); |
1da177e4 | 1137 | spin_lock_init(&mm->page_table_lock); |
88aa7cc6 | 1138 | spin_lock_init(&mm->arg_lock); |
41f727fd | 1139 | mm_init_cpumask(mm); |
858f0993 | 1140 | mm_init_aio(mm); |
cf475ad2 | 1141 | mm_init_owner(mm, p); |
a6cbd440 | 1142 | mm_pasid_init(mm); |
2b7e8665 | 1143 | RCU_INIT_POINTER(mm->exe_file, NULL); |
984cfe4e | 1144 | mmu_notifier_subscriptions_init(mm); |
16af97dc | 1145 | init_tlb_flush_pending(mm); |
41f727fd VD |
1146 | #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS |
1147 | mm->pmd_huge_pte = NULL; | |
1148 | #endif | |
355627f5 | 1149 | mm_init_uprobes_state(mm); |
13db8c50 | 1150 | hugetlb_count_init(mm); |
1da177e4 | 1151 | |
a0715cc2 AT |
1152 | if (current->mm) { |
1153 | mm->flags = current->mm->flags & MMF_INIT_MASK; | |
1154 | mm->def_flags = current->mm->def_flags & VM_INIT_DEF_MASK; | |
1155 | } else { | |
1156 | mm->flags = default_dump_filter; | |
1da177e4 | 1157 | mm->def_flags = 0; |
a0715cc2 AT |
1158 | } |
1159 | ||
41f727fd VD |
1160 | if (mm_alloc_pgd(mm)) |
1161 | goto fail_nopgd; | |
1162 | ||
1163 | if (init_new_context(p, mm)) | |
1164 | goto fail_nocontext; | |
78fb7466 | 1165 | |
f1a79412 SB |
1166 | for (i = 0; i < NR_MM_COUNTERS; i++) |
1167 | if (percpu_counter_init(&mm->rss_stat[i], 0, GFP_KERNEL_ACCOUNT)) | |
1168 | goto fail_pcpu; | |
1169 | ||
bfedb589 | 1170 | mm->user_ns = get_user_ns(user_ns); |
bd74fdae | 1171 | lru_gen_init_mm(mm); |
41f727fd VD |
1172 | return mm; |
1173 | ||
f1a79412 SB |
1174 | fail_pcpu: |
1175 | while (i > 0) | |
1176 | percpu_counter_destroy(&mm->rss_stat[--i]); | |
41f727fd VD |
1177 | fail_nocontext: |
1178 | mm_free_pgd(mm); | |
1179 | fail_nopgd: | |
1da177e4 LT |
1180 | free_mm(mm); |
1181 | return NULL; | |
1182 | } | |
1183 | ||
1184 | /* | |
1185 | * Allocate and initialize an mm_struct. | |
1186 | */ | |
fb0a685c | 1187 | struct mm_struct *mm_alloc(void) |
1da177e4 | 1188 | { |
fb0a685c | 1189 | struct mm_struct *mm; |
1da177e4 LT |
1190 | |
1191 | mm = allocate_mm(); | |
de03c72c KM |
1192 | if (!mm) |
1193 | return NULL; | |
1194 | ||
1195 | memset(mm, 0, sizeof(*mm)); | |
bfedb589 | 1196 | return mm_init(mm, current, current_user_ns()); |
1da177e4 LT |
1197 | } |
1198 | ||
ec8d7c14 MH |
1199 | static inline void __mmput(struct mm_struct *mm) |
1200 | { | |
1201 | VM_BUG_ON(atomic_read(&mm->mm_users)); | |
1202 | ||
1203 | uprobe_clear_state(mm); | |
1204 | exit_aio(mm); | |
1205 | ksm_exit(mm); | |
1206 | khugepaged_exit(mm); /* must run before exit_mmap */ | |
1207 | exit_mmap(mm); | |
6fcb52a5 | 1208 | mm_put_huge_zero_page(mm); |
ec8d7c14 MH |
1209 | set_mm_exe_file(mm, NULL); |
1210 | if (!list_empty(&mm->mmlist)) { | |
1211 | spin_lock(&mmlist_lock); | |
1212 | list_del(&mm->mmlist); | |
1213 | spin_unlock(&mmlist_lock); | |
1214 | } | |
1215 | if (mm->binfmt) | |
1216 | module_put(mm->binfmt->module); | |
bd74fdae | 1217 | lru_gen_del_mm(mm); |
ec8d7c14 MH |
1218 | mmdrop(mm); |
1219 | } | |
1220 | ||
1da177e4 LT |
1221 | /* |
1222 | * Decrement the use count and release all resources for an mm. | |
1223 | */ | |
1224 | void mmput(struct mm_struct *mm) | |
1225 | { | |
0ae26f1b AM |
1226 | might_sleep(); |
1227 | ||
ec8d7c14 MH |
1228 | if (atomic_dec_and_test(&mm->mm_users)) |
1229 | __mmput(mm); | |
1230 | } | |
1231 | EXPORT_SYMBOL_GPL(mmput); | |
1232 | ||
a1b2289c SY |
1233 | #ifdef CONFIG_MMU |
1234 | static void mmput_async_fn(struct work_struct *work) | |
1235 | { | |
1236 | struct mm_struct *mm = container_of(work, struct mm_struct, | |
1237 | async_put_work); | |
1238 | ||
1239 | __mmput(mm); | |
1240 | } | |
1241 | ||
1242 | void mmput_async(struct mm_struct *mm) | |
1243 | { | |
1244 | if (atomic_dec_and_test(&mm->mm_users)) { | |
1245 | INIT_WORK(&mm->async_put_work, mmput_async_fn); | |
1246 | schedule_work(&mm->async_put_work); | |
1247 | } | |
1248 | } | |
85eaeb50 | 1249 | EXPORT_SYMBOL_GPL(mmput_async); |
a1b2289c SY |
1250 | #endif |
1251 | ||
90f31d0e KK |
1252 | /** |
1253 | * set_mm_exe_file - change a reference to the mm's executable file | |
1254 | * | |
1255 | * This changes mm's executable file (shown as symlink /proc/[pid]/exe). | |
1256 | * | |
6e399cd1 DB |
1257 | * Main users are mmput() and sys_execve(). Callers prevent concurrent |
1258 | * invocations: in mmput() nobody alive left, in execve task is single | |
35d7bdc8 | 1259 | * threaded. |
fe69d560 DH |
1260 | * |
1261 | * Can only fail if new_exe_file != NULL. | |
90f31d0e | 1262 | */ |
fe69d560 | 1263 | int set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file) |
38646013 | 1264 | { |
6e399cd1 DB |
1265 | struct file *old_exe_file; |
1266 | ||
1267 | /* | |
1268 | * It is safe to dereference the exe_file without RCU as | |
1269 | * this function is only called if nobody else can access | |
1270 | * this mm -- see comment above for justification. | |
1271 | */ | |
1272 | old_exe_file = rcu_dereference_raw(mm->exe_file); | |
90f31d0e | 1273 | |
fe69d560 DH |
1274 | if (new_exe_file) { |
1275 | /* | |
1276 | * We expect the caller (i.e., sys_execve) to already denied | |
1277 | * write access, so this is unlikely to fail. | |
1278 | */ | |
1279 | if (unlikely(deny_write_access(new_exe_file))) | |
1280 | return -EACCES; | |
38646013 | 1281 | get_file(new_exe_file); |
fe69d560 | 1282 | } |
90f31d0e | 1283 | rcu_assign_pointer(mm->exe_file, new_exe_file); |
fe69d560 DH |
1284 | if (old_exe_file) { |
1285 | allow_write_access(old_exe_file); | |
90f31d0e | 1286 | fput(old_exe_file); |
fe69d560 DH |
1287 | } |
1288 | return 0; | |
38646013 JS |
1289 | } |
1290 | ||
35d7bdc8 DH |
1291 | /** |
1292 | * replace_mm_exe_file - replace a reference to the mm's executable file | |
1293 | * | |
1294 | * This changes mm's executable file (shown as symlink /proc/[pid]/exe), | |
1295 | * dealing with concurrent invocation and without grabbing the mmap lock in | |
1296 | * write mode. | |
1297 | * | |
1298 | * Main user is sys_prctl(PR_SET_MM_MAP/EXE_FILE). | |
1299 | */ | |
1300 | int replace_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file) | |
1301 | { | |
1302 | struct vm_area_struct *vma; | |
1303 | struct file *old_exe_file; | |
1304 | int ret = 0; | |
1305 | ||
1306 | /* Forbid mm->exe_file change if old file still mapped. */ | |
1307 | old_exe_file = get_mm_exe_file(mm); | |
1308 | if (old_exe_file) { | |
fa5e5876 | 1309 | VMA_ITERATOR(vmi, mm, 0); |
35d7bdc8 | 1310 | mmap_read_lock(mm); |
fa5e5876 | 1311 | for_each_vma(vmi, vma) { |
35d7bdc8 DH |
1312 | if (!vma->vm_file) |
1313 | continue; | |
1314 | if (path_equal(&vma->vm_file->f_path, | |
fa5e5876 | 1315 | &old_exe_file->f_path)) { |
35d7bdc8 | 1316 | ret = -EBUSY; |
fa5e5876 MWO |
1317 | break; |
1318 | } | |
35d7bdc8 DH |
1319 | } |
1320 | mmap_read_unlock(mm); | |
1321 | fput(old_exe_file); | |
1322 | if (ret) | |
1323 | return ret; | |
1324 | } | |
1325 | ||
1326 | /* set the new file, lockless */ | |
fe69d560 DH |
1327 | ret = deny_write_access(new_exe_file); |
1328 | if (ret) | |
1329 | return -EACCES; | |
35d7bdc8 | 1330 | get_file(new_exe_file); |
fe69d560 | 1331 | |
35d7bdc8 | 1332 | old_exe_file = xchg(&mm->exe_file, new_exe_file); |
fe69d560 DH |
1333 | if (old_exe_file) { |
1334 | /* | |
1335 | * Don't race with dup_mmap() getting the file and disallowing | |
1336 | * write access while someone might open the file writable. | |
1337 | */ | |
1338 | mmap_read_lock(mm); | |
1339 | allow_write_access(old_exe_file); | |
35d7bdc8 | 1340 | fput(old_exe_file); |
fe69d560 DH |
1341 | mmap_read_unlock(mm); |
1342 | } | |
35d7bdc8 | 1343 | return 0; |
38646013 JS |
1344 | } |
1345 | ||
90f31d0e KK |
1346 | /** |
1347 | * get_mm_exe_file - acquire a reference to the mm's executable file | |
1348 | * | |
1349 | * Returns %NULL if mm has no associated executable file. | |
1350 | * User must release file via fput(). | |
1351 | */ | |
38646013 JS |
1352 | struct file *get_mm_exe_file(struct mm_struct *mm) |
1353 | { | |
1354 | struct file *exe_file; | |
1355 | ||
90f31d0e KK |
1356 | rcu_read_lock(); |
1357 | exe_file = rcu_dereference(mm->exe_file); | |
1358 | if (exe_file && !get_file_rcu(exe_file)) | |
1359 | exe_file = NULL; | |
1360 | rcu_read_unlock(); | |
38646013 JS |
1361 | return exe_file; |
1362 | } | |
1363 | ||
cd81a917 MG |
1364 | /** |
1365 | * get_task_exe_file - acquire a reference to the task's executable file | |
1366 | * | |
1367 | * Returns %NULL if task's mm (if any) has no associated executable file or | |
1368 | * this is a kernel thread with borrowed mm (see the comment above get_task_mm). | |
1369 | * User must release file via fput(). | |
1370 | */ | |
1371 | struct file *get_task_exe_file(struct task_struct *task) | |
1372 | { | |
1373 | struct file *exe_file = NULL; | |
1374 | struct mm_struct *mm; | |
1375 | ||
1376 | task_lock(task); | |
1377 | mm = task->mm; | |
1378 | if (mm) { | |
1379 | if (!(task->flags & PF_KTHREAD)) | |
1380 | exe_file = get_mm_exe_file(mm); | |
1381 | } | |
1382 | task_unlock(task); | |
1383 | return exe_file; | |
1384 | } | |
38646013 | 1385 | |
1da177e4 LT |
1386 | /** |
1387 | * get_task_mm - acquire a reference to the task's mm | |
1388 | * | |
246bb0b1 | 1389 | * Returns %NULL if the task has no mm. Checks PF_KTHREAD (meaning |
1da177e4 LT |
1390 | * this kernel workthread has transiently adopted a user mm with use_mm, |
1391 | * to do its AIO) is not set and if so returns a reference to it, after | |
1392 | * bumping up the use count. User must release the mm via mmput() | |
1393 | * after use. Typically used by /proc and ptrace. | |
1394 | */ | |
1395 | struct mm_struct *get_task_mm(struct task_struct *task) | |
1396 | { | |
1397 | struct mm_struct *mm; | |
1398 | ||
1399 | task_lock(task); | |
1400 | mm = task->mm; | |
1401 | if (mm) { | |
246bb0b1 | 1402 | if (task->flags & PF_KTHREAD) |
1da177e4 LT |
1403 | mm = NULL; |
1404 | else | |
3fce371b | 1405 | mmget(mm); |
1da177e4 LT |
1406 | } |
1407 | task_unlock(task); | |
1408 | return mm; | |
1409 | } | |
1410 | EXPORT_SYMBOL_GPL(get_task_mm); | |
1411 | ||
8cdb878d CY |
1412 | struct mm_struct *mm_access(struct task_struct *task, unsigned int mode) |
1413 | { | |
1414 | struct mm_struct *mm; | |
1415 | int err; | |
1416 | ||
f7cfd871 | 1417 | err = down_read_killable(&task->signal->exec_update_lock); |
8cdb878d CY |
1418 | if (err) |
1419 | return ERR_PTR(err); | |
1420 | ||
1421 | mm = get_task_mm(task); | |
1422 | if (mm && mm != current->mm && | |
1423 | !ptrace_may_access(task, mode)) { | |
1424 | mmput(mm); | |
1425 | mm = ERR_PTR(-EACCES); | |
1426 | } | |
f7cfd871 | 1427 | up_read(&task->signal->exec_update_lock); |
8cdb878d CY |
1428 | |
1429 | return mm; | |
1430 | } | |
1431 | ||
57b59c4a | 1432 | static void complete_vfork_done(struct task_struct *tsk) |
c415c3b4 | 1433 | { |
d68b46fe | 1434 | struct completion *vfork; |
c415c3b4 | 1435 | |
d68b46fe ON |
1436 | task_lock(tsk); |
1437 | vfork = tsk->vfork_done; | |
1438 | if (likely(vfork)) { | |
1439 | tsk->vfork_done = NULL; | |
1440 | complete(vfork); | |
1441 | } | |
1442 | task_unlock(tsk); | |
1443 | } | |
1444 | ||
1445 | static int wait_for_vfork_done(struct task_struct *child, | |
1446 | struct completion *vfork) | |
1447 | { | |
f5d39b02 | 1448 | unsigned int state = TASK_UNINTERRUPTIBLE|TASK_KILLABLE|TASK_FREEZABLE; |
d68b46fe ON |
1449 | int killed; |
1450 | ||
76f969e8 | 1451 | cgroup_enter_frozen(); |
f5d39b02 | 1452 | killed = wait_for_completion_state(vfork, state); |
76f969e8 | 1453 | cgroup_leave_frozen(false); |
d68b46fe ON |
1454 | |
1455 | if (killed) { | |
1456 | task_lock(child); | |
1457 | child->vfork_done = NULL; | |
1458 | task_unlock(child); | |
1459 | } | |
1460 | ||
1461 | put_task_struct(child); | |
1462 | return killed; | |
c415c3b4 ON |
1463 | } |
1464 | ||
1da177e4 LT |
1465 | /* Please note the differences between mmput and mm_release. |
1466 | * mmput is called whenever we stop holding onto a mm_struct, | |
1467 | * error success whatever. | |
1468 | * | |
1469 | * mm_release is called after a mm_struct has been removed | |
1470 | * from the current process. | |
1471 | * | |
1472 | * This difference is important for error handling, when we | |
1473 | * only half set up a mm_struct for a new process and need to restore | |
1474 | * the old one. Because we mmput the new mm_struct before | |
1475 | * restoring the old one. . . | |
1476 | * Eric Biederman 10 January 1998 | |
1477 | */ | |
4610ba7a | 1478 | static void mm_release(struct task_struct *tsk, struct mm_struct *mm) |
1da177e4 | 1479 | { |
0326f5a9 SD |
1480 | uprobe_free_utask(tsk); |
1481 | ||
1da177e4 LT |
1482 | /* Get rid of any cached register state */ |
1483 | deactivate_mm(tsk, mm); | |
1484 | ||
fec1d011 | 1485 | /* |
735f2770 MH |
1486 | * Signal userspace if we're not exiting with a core dump |
1487 | * because we want to leave the value intact for debugging | |
1488 | * purposes. | |
fec1d011 | 1489 | */ |
9c8a8228 | 1490 | if (tsk->clear_child_tid) { |
92307383 | 1491 | if (atomic_read(&mm->mm_users) > 1) { |
9c8a8228 ED |
1492 | /* |
1493 | * We don't check the error code - if userspace has | |
1494 | * not set up a proper pointer then tough luck. | |
1495 | */ | |
1496 | put_user(0, tsk->clear_child_tid); | |
2de0db99 DB |
1497 | do_futex(tsk->clear_child_tid, FUTEX_WAKE, |
1498 | 1, NULL, NULL, 0, 0); | |
9c8a8228 | 1499 | } |
1da177e4 | 1500 | tsk->clear_child_tid = NULL; |
1da177e4 | 1501 | } |
f7505d64 KK |
1502 | |
1503 | /* | |
1504 | * All done, finally we can wake up parent and return this mm to him. | |
1505 | * Also kthread_stop() uses this completion for synchronization. | |
1506 | */ | |
1507 | if (tsk->vfork_done) | |
1508 | complete_vfork_done(tsk); | |
1da177e4 LT |
1509 | } |
1510 | ||
4610ba7a TG |
1511 | void exit_mm_release(struct task_struct *tsk, struct mm_struct *mm) |
1512 | { | |
150d7158 | 1513 | futex_exit_release(tsk); |
4610ba7a TG |
1514 | mm_release(tsk, mm); |
1515 | } | |
1516 | ||
1517 | void exec_mm_release(struct task_struct *tsk, struct mm_struct *mm) | |
1518 | { | |
150d7158 | 1519 | futex_exec_release(tsk); |
4610ba7a TG |
1520 | mm_release(tsk, mm); |
1521 | } | |
1522 | ||
13585fa0 NA |
1523 | /** |
1524 | * dup_mm() - duplicates an existing mm structure | |
1525 | * @tsk: the task_struct with which the new mm will be associated. | |
1526 | * @oldmm: the mm to duplicate. | |
1527 | * | |
1528 | * Allocates a new mm structure and duplicates the provided @oldmm structure | |
1529 | * content into it. | |
1530 | * | |
1531 | * Return: the duplicated mm or NULL on failure. | |
a0a7ec30 | 1532 | */ |
13585fa0 NA |
1533 | static struct mm_struct *dup_mm(struct task_struct *tsk, |
1534 | struct mm_struct *oldmm) | |
a0a7ec30 | 1535 | { |
13585fa0 | 1536 | struct mm_struct *mm; |
a0a7ec30 JD |
1537 | int err; |
1538 | ||
a0a7ec30 JD |
1539 | mm = allocate_mm(); |
1540 | if (!mm) | |
1541 | goto fail_nomem; | |
1542 | ||
1543 | memcpy(mm, oldmm, sizeof(*mm)); | |
1544 | ||
bfedb589 | 1545 | if (!mm_init(mm, tsk, mm->user_ns)) |
a0a7ec30 JD |
1546 | goto fail_nomem; |
1547 | ||
a0a7ec30 JD |
1548 | err = dup_mmap(mm, oldmm); |
1549 | if (err) | |
1550 | goto free_pt; | |
1551 | ||
1552 | mm->hiwater_rss = get_mm_rss(mm); | |
1553 | mm->hiwater_vm = mm->total_vm; | |
1554 | ||
801460d0 HS |
1555 | if (mm->binfmt && !try_module_get(mm->binfmt->module)) |
1556 | goto free_pt; | |
1557 | ||
a0a7ec30 JD |
1558 | return mm; |
1559 | ||
1560 | free_pt: | |
801460d0 HS |
1561 | /* don't put binfmt in mmput, we haven't got module yet */ |
1562 | mm->binfmt = NULL; | |
c3f3ce04 | 1563 | mm_init_owner(mm, NULL); |
a0a7ec30 JD |
1564 | mmput(mm); |
1565 | ||
1566 | fail_nomem: | |
1567 | return NULL; | |
a0a7ec30 JD |
1568 | } |
1569 | ||
fb0a685c | 1570 | static int copy_mm(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 | 1571 | { |
fb0a685c | 1572 | struct mm_struct *mm, *oldmm; |
1da177e4 LT |
1573 | |
1574 | tsk->min_flt = tsk->maj_flt = 0; | |
1575 | tsk->nvcsw = tsk->nivcsw = 0; | |
17406b82 MSB |
1576 | #ifdef CONFIG_DETECT_HUNG_TASK |
1577 | tsk->last_switch_count = tsk->nvcsw + tsk->nivcsw; | |
a2e51445 | 1578 | tsk->last_switch_time = 0; |
17406b82 | 1579 | #endif |
1da177e4 LT |
1580 | |
1581 | tsk->mm = NULL; | |
1582 | tsk->active_mm = NULL; | |
1583 | ||
1584 | /* | |
1585 | * Are we cloning a kernel thread? | |
1586 | * | |
1587 | * We need to steal a active VM for that.. | |
1588 | */ | |
1589 | oldmm = current->mm; | |
1590 | if (!oldmm) | |
1591 | return 0; | |
1592 | ||
1593 | if (clone_flags & CLONE_VM) { | |
3fce371b | 1594 | mmget(oldmm); |
1da177e4 | 1595 | mm = oldmm; |
a6895399 REB |
1596 | } else { |
1597 | mm = dup_mm(tsk, current->mm); | |
1598 | if (!mm) | |
1599 | return -ENOMEM; | |
1da177e4 LT |
1600 | } |
1601 | ||
1da177e4 LT |
1602 | tsk->mm = mm; |
1603 | tsk->active_mm = mm; | |
1604 | return 0; | |
1da177e4 LT |
1605 | } |
1606 | ||
a39bc516 | 1607 | static int copy_fs(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 | 1608 | { |
498052bb | 1609 | struct fs_struct *fs = current->fs; |
1da177e4 | 1610 | if (clone_flags & CLONE_FS) { |
498052bb | 1611 | /* tsk->fs is already what we want */ |
2a4419b5 | 1612 | spin_lock(&fs->lock); |
498052bb | 1613 | if (fs->in_exec) { |
2a4419b5 | 1614 | spin_unlock(&fs->lock); |
498052bb AV |
1615 | return -EAGAIN; |
1616 | } | |
1617 | fs->users++; | |
2a4419b5 | 1618 | spin_unlock(&fs->lock); |
1da177e4 LT |
1619 | return 0; |
1620 | } | |
498052bb | 1621 | tsk->fs = copy_fs_struct(fs); |
1da177e4 LT |
1622 | if (!tsk->fs) |
1623 | return -ENOMEM; | |
1624 | return 0; | |
1625 | } | |
1626 | ||
fb0a685c | 1627 | static int copy_files(unsigned long clone_flags, struct task_struct *tsk) |
a016f338 JD |
1628 | { |
1629 | struct files_struct *oldf, *newf; | |
1630 | int error = 0; | |
1631 | ||
1632 | /* | |
1633 | * A background process may not have any files ... | |
1634 | */ | |
1635 | oldf = current->files; | |
1636 | if (!oldf) | |
1637 | goto out; | |
1638 | ||
1639 | if (clone_flags & CLONE_FILES) { | |
1640 | atomic_inc(&oldf->count); | |
1641 | goto out; | |
1642 | } | |
1643 | ||
60997c3d | 1644 | newf = dup_fd(oldf, NR_OPEN_MAX, &error); |
a016f338 JD |
1645 | if (!newf) |
1646 | goto out; | |
1647 | ||
1648 | tsk->files = newf; | |
1649 | error = 0; | |
1650 | out: | |
1651 | return error; | |
1652 | } | |
1653 | ||
a39bc516 | 1654 | static int copy_sighand(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 LT |
1655 | { |
1656 | struct sighand_struct *sig; | |
1657 | ||
60348802 | 1658 | if (clone_flags & CLONE_SIGHAND) { |
d036bda7 | 1659 | refcount_inc(¤t->sighand->count); |
1da177e4 LT |
1660 | return 0; |
1661 | } | |
1662 | sig = kmem_cache_alloc(sighand_cachep, GFP_KERNEL); | |
0c282b06 | 1663 | RCU_INIT_POINTER(tsk->sighand, sig); |
1da177e4 LT |
1664 | if (!sig) |
1665 | return -ENOMEM; | |
9d7fb042 | 1666 | |
d036bda7 | 1667 | refcount_set(&sig->count, 1); |
06e62a46 | 1668 | spin_lock_irq(¤t->sighand->siglock); |
1da177e4 | 1669 | memcpy(sig->action, current->sighand->action, sizeof(sig->action)); |
06e62a46 | 1670 | spin_unlock_irq(¤t->sighand->siglock); |
b612e5df CB |
1671 | |
1672 | /* Reset all signal handler not set to SIG_IGN to SIG_DFL. */ | |
1673 | if (clone_flags & CLONE_CLEAR_SIGHAND) | |
1674 | flush_signal_handlers(tsk, 0); | |
1675 | ||
1da177e4 LT |
1676 | return 0; |
1677 | } | |
1678 | ||
a7e5328a | 1679 | void __cleanup_sighand(struct sighand_struct *sighand) |
c81addc9 | 1680 | { |
d036bda7 | 1681 | if (refcount_dec_and_test(&sighand->count)) { |
d80e731e | 1682 | signalfd_cleanup(sighand); |
392809b2 | 1683 | /* |
5f0d5a3a | 1684 | * sighand_cachep is SLAB_TYPESAFE_BY_RCU so we can free it |
392809b2 ON |
1685 | * without an RCU grace period, see __lock_task_sighand(). |
1686 | */ | |
c81addc9 | 1687 | kmem_cache_free(sighand_cachep, sighand); |
d80e731e | 1688 | } |
c81addc9 ON |
1689 | } |
1690 | ||
f06febc9 FM |
1691 | /* |
1692 | * Initialize POSIX timer handling for a thread group. | |
1693 | */ | |
1694 | static void posix_cpu_timers_init_group(struct signal_struct *sig) | |
1695 | { | |
2b69942f | 1696 | struct posix_cputimers *pct = &sig->posix_cputimers; |
78d7d407 JS |
1697 | unsigned long cpu_limit; |
1698 | ||
316c1608 | 1699 | cpu_limit = READ_ONCE(sig->rlim[RLIMIT_CPU].rlim_cur); |
3a245c0f | 1700 | posix_cputimers_group_init(pct, cpu_limit); |
f06febc9 FM |
1701 | } |
1702 | ||
a39bc516 | 1703 | static int copy_signal(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 LT |
1704 | { |
1705 | struct signal_struct *sig; | |
1da177e4 | 1706 | |
4ab6c083 | 1707 | if (clone_flags & CLONE_THREAD) |
490dea45 | 1708 | return 0; |
490dea45 | 1709 | |
a56704ef | 1710 | sig = kmem_cache_zalloc(signal_cachep, GFP_KERNEL); |
1da177e4 LT |
1711 | tsk->signal = sig; |
1712 | if (!sig) | |
1713 | return -ENOMEM; | |
1714 | ||
b3ac022c | 1715 | sig->nr_threads = 1; |
d80f7d7b | 1716 | sig->quick_threads = 1; |
1da177e4 | 1717 | atomic_set(&sig->live, 1); |
60d4de3f | 1718 | refcount_set(&sig->sigcnt, 1); |
0c740d0a ON |
1719 | |
1720 | /* list_add(thread_node, thread_head) without INIT_LIST_HEAD() */ | |
1721 | sig->thread_head = (struct list_head)LIST_HEAD_INIT(tsk->thread_node); | |
1722 | tsk->thread_node = (struct list_head)LIST_HEAD_INIT(sig->thread_head); | |
1723 | ||
1da177e4 | 1724 | init_waitqueue_head(&sig->wait_chldexit); |
db51aecc | 1725 | sig->curr_target = tsk; |
1da177e4 | 1726 | init_sigpending(&sig->shared_pending); |
c3ad2c3b | 1727 | INIT_HLIST_HEAD(&sig->multiprocess); |
e78c3496 | 1728 | seqlock_init(&sig->stats_lock); |
9d7fb042 | 1729 | prev_cputime_init(&sig->prev_cputime); |
1da177e4 | 1730 | |
baa73d9e | 1731 | #ifdef CONFIG_POSIX_TIMERS |
b18b6a9c | 1732 | INIT_LIST_HEAD(&sig->posix_timers); |
c9cb2e3d | 1733 | hrtimer_init(&sig->real_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL); |
1da177e4 | 1734 | sig->real_timer.function = it_real_fn; |
baa73d9e | 1735 | #endif |
1da177e4 | 1736 | |
1da177e4 LT |
1737 | task_lock(current->group_leader); |
1738 | memcpy(sig->rlim, current->signal->rlim, sizeof sig->rlim); | |
1739 | task_unlock(current->group_leader); | |
1740 | ||
6279a751 ON |
1741 | posix_cpu_timers_init_group(sig); |
1742 | ||
522ed776 | 1743 | tty_audit_fork(sig); |
5091faa4 | 1744 | sched_autogroup_fork(sig); |
522ed776 | 1745 | |
a63d83f4 | 1746 | sig->oom_score_adj = current->signal->oom_score_adj; |
dabb16f6 | 1747 | sig->oom_score_adj_min = current->signal->oom_score_adj_min; |
28b83c51 | 1748 | |
9b1bf12d | 1749 | mutex_init(&sig->cred_guard_mutex); |
f7cfd871 | 1750 | init_rwsem(&sig->exec_update_lock); |
9b1bf12d | 1751 | |
1da177e4 LT |
1752 | return 0; |
1753 | } | |
1754 | ||
dbd95212 KC |
1755 | static void copy_seccomp(struct task_struct *p) |
1756 | { | |
1757 | #ifdef CONFIG_SECCOMP | |
1758 | /* | |
1759 | * Must be called with sighand->lock held, which is common to | |
1760 | * all threads in the group. Holding cred_guard_mutex is not | |
1761 | * needed because this new task is not yet running and cannot | |
1762 | * be racing exec. | |
1763 | */ | |
69f6a34b | 1764 | assert_spin_locked(¤t->sighand->siglock); |
dbd95212 KC |
1765 | |
1766 | /* Ref-count the new filter user, and assign it. */ | |
1767 | get_seccomp_filter(current); | |
1768 | p->seccomp = current->seccomp; | |
1769 | ||
1770 | /* | |
1771 | * Explicitly enable no_new_privs here in case it got set | |
1772 | * between the task_struct being duplicated and holding the | |
1773 | * sighand lock. The seccomp state and nnp must be in sync. | |
1774 | */ | |
1775 | if (task_no_new_privs(current)) | |
1776 | task_set_no_new_privs(p); | |
1777 | ||
1778 | /* | |
1779 | * If the parent gained a seccomp mode after copying thread | |
1780 | * flags and between before we held the sighand lock, we have | |
1781 | * to manually enable the seccomp thread flag here. | |
1782 | */ | |
1783 | if (p->seccomp.mode != SECCOMP_MODE_DISABLED) | |
23d67a54 | 1784 | set_task_syscall_work(p, SECCOMP); |
dbd95212 KC |
1785 | #endif |
1786 | } | |
1787 | ||
17da2bd9 | 1788 | SYSCALL_DEFINE1(set_tid_address, int __user *, tidptr) |
1da177e4 LT |
1789 | { |
1790 | current->clear_child_tid = tidptr; | |
1791 | ||
b488893a | 1792 | return task_pid_vnr(current); |
1da177e4 LT |
1793 | } |
1794 | ||
a39bc516 | 1795 | static void rt_mutex_init_task(struct task_struct *p) |
23f78d4a | 1796 | { |
1d615482 | 1797 | raw_spin_lock_init(&p->pi_lock); |
e29e175b | 1798 | #ifdef CONFIG_RT_MUTEXES |
a23ba907 | 1799 | p->pi_waiters = RB_ROOT_CACHED; |
e96a7705 | 1800 | p->pi_top_task = NULL; |
23f78d4a | 1801 | p->pi_blocked_on = NULL; |
23f78d4a IM |
1802 | #endif |
1803 | } | |
1804 | ||
2c470475 EB |
1805 | static inline void init_task_pid_links(struct task_struct *task) |
1806 | { | |
1807 | enum pid_type type; | |
1808 | ||
96e1e984 | 1809 | for (type = PIDTYPE_PID; type < PIDTYPE_MAX; ++type) |
2c470475 | 1810 | INIT_HLIST_NODE(&task->pid_links[type]); |
2c470475 EB |
1811 | } |
1812 | ||
81907739 ON |
1813 | static inline void |
1814 | init_task_pid(struct task_struct *task, enum pid_type type, struct pid *pid) | |
1815 | { | |
2c470475 EB |
1816 | if (type == PIDTYPE_PID) |
1817 | task->thread_pid = pid; | |
1818 | else | |
1819 | task->signal->pids[type] = pid; | |
81907739 ON |
1820 | } |
1821 | ||
6bfbaa51 IM |
1822 | static inline void rcu_copy_process(struct task_struct *p) |
1823 | { | |
1824 | #ifdef CONFIG_PREEMPT_RCU | |
1825 | p->rcu_read_lock_nesting = 0; | |
1826 | p->rcu_read_unlock_special.s = 0; | |
1827 | p->rcu_blocked_node = NULL; | |
1828 | INIT_LIST_HEAD(&p->rcu_node_entry); | |
1829 | #endif /* #ifdef CONFIG_PREEMPT_RCU */ | |
1830 | #ifdef CONFIG_TASKS_RCU | |
1831 | p->rcu_tasks_holdout = false; | |
1832 | INIT_LIST_HEAD(&p->rcu_tasks_holdout_list); | |
1833 | p->rcu_tasks_idle_cpu = -1; | |
1834 | #endif /* #ifdef CONFIG_TASKS_RCU */ | |
d5f177d3 PM |
1835 | #ifdef CONFIG_TASKS_TRACE_RCU |
1836 | p->trc_reader_nesting = 0; | |
276c4104 | 1837 | p->trc_reader_special.s = 0; |
d5f177d3 | 1838 | INIT_LIST_HEAD(&p->trc_holdout_list); |
434c9eef | 1839 | INIT_LIST_HEAD(&p->trc_blkd_node); |
d5f177d3 | 1840 | #endif /* #ifdef CONFIG_TASKS_TRACE_RCU */ |
6bfbaa51 IM |
1841 | } |
1842 | ||
3695eae5 CB |
1843 | struct pid *pidfd_pid(const struct file *file) |
1844 | { | |
1845 | if (file->f_op == &pidfd_fops) | |
1846 | return file->private_data; | |
1847 | ||
1848 | return ERR_PTR(-EBADF); | |
1849 | } | |
1850 | ||
b3e58382 CB |
1851 | static int pidfd_release(struct inode *inode, struct file *file) |
1852 | { | |
1853 | struct pid *pid = file->private_data; | |
1854 | ||
1855 | file->private_data = NULL; | |
1856 | put_pid(pid); | |
1857 | return 0; | |
1858 | } | |
1859 | ||
1860 | #ifdef CONFIG_PROC_FS | |
15d42eb2 CK |
1861 | /** |
1862 | * pidfd_show_fdinfo - print information about a pidfd | |
1863 | * @m: proc fdinfo file | |
1864 | * @f: file referencing a pidfd | |
1865 | * | |
1866 | * Pid: | |
1867 | * This function will print the pid that a given pidfd refers to in the | |
1868 | * pid namespace of the procfs instance. | |
1869 | * If the pid namespace of the process is not a descendant of the pid | |
1870 | * namespace of the procfs instance 0 will be shown as its pid. This is | |
1871 | * similar to calling getppid() on a process whose parent is outside of | |
1872 | * its pid namespace. | |
1873 | * | |
1874 | * NSpid: | |
1875 | * If pid namespaces are supported then this function will also print | |
1876 | * the pid of a given pidfd refers to for all descendant pid namespaces | |
1877 | * starting from the current pid namespace of the instance, i.e. the | |
1878 | * Pid field and the first entry in the NSpid field will be identical. | |
1879 | * If the pid namespace of the process is not a descendant of the pid | |
1880 | * namespace of the procfs instance 0 will be shown as its first NSpid | |
1881 | * entry and no others will be shown. | |
1882 | * Note that this differs from the Pid and NSpid fields in | |
1883 | * /proc/<pid>/status where Pid and NSpid are always shown relative to | |
1884 | * the pid namespace of the procfs instance. The difference becomes | |
1885 | * obvious when sending around a pidfd between pid namespaces from a | |
a8ca6b13 | 1886 | * different branch of the tree, i.e. where no ancestral relation is |
15d42eb2 CK |
1887 | * present between the pid namespaces: |
1888 | * - create two new pid namespaces ns1 and ns2 in the initial pid | |
1889 | * namespace (also take care to create new mount namespaces in the | |
1890 | * new pid namespace and mount procfs) | |
1891 | * - create a process with a pidfd in ns1 | |
1892 | * - send pidfd from ns1 to ns2 | |
1893 | * - read /proc/self/fdinfo/<pidfd> and observe that both Pid and NSpid | |
1894 | * have exactly one entry, which is 0 | |
1895 | */ | |
b3e58382 CB |
1896 | static void pidfd_show_fdinfo(struct seq_file *m, struct file *f) |
1897 | { | |
b3e58382 | 1898 | struct pid *pid = f->private_data; |
3d6d8da4 CB |
1899 | struct pid_namespace *ns; |
1900 | pid_t nr = -1; | |
15d42eb2 | 1901 | |
3d6d8da4 | 1902 | if (likely(pid_has_task(pid, PIDTYPE_PID))) { |
9d78edea | 1903 | ns = proc_pid_ns(file_inode(m->file)->i_sb); |
3d6d8da4 CB |
1904 | nr = pid_nr_ns(pid, ns); |
1905 | } | |
1906 | ||
1907 | seq_put_decimal_ll(m, "Pid:\t", nr); | |
b3e58382 | 1908 | |
15d42eb2 | 1909 | #ifdef CONFIG_PID_NS |
3d6d8da4 CB |
1910 | seq_put_decimal_ll(m, "\nNSpid:\t", nr); |
1911 | if (nr > 0) { | |
15d42eb2 | 1912 | int i; |
b3e58382 | 1913 | |
15d42eb2 CK |
1914 | /* If nr is non-zero it means that 'pid' is valid and that |
1915 | * ns, i.e. the pid namespace associated with the procfs | |
1916 | * instance, is in the pid namespace hierarchy of pid. | |
1917 | * Start at one below the already printed level. | |
1918 | */ | |
1919 | for (i = ns->level + 1; i <= pid->level; i++) | |
3d6d8da4 | 1920 | seq_put_decimal_ll(m, "\t", pid->numbers[i].nr); |
15d42eb2 CK |
1921 | } |
1922 | #endif | |
b3e58382 CB |
1923 | seq_putc(m, '\n'); |
1924 | } | |
1925 | #endif | |
1926 | ||
b53b0b9d JFG |
1927 | /* |
1928 | * Poll support for process exit notification. | |
1929 | */ | |
9e77716a | 1930 | static __poll_t pidfd_poll(struct file *file, struct poll_table_struct *pts) |
b53b0b9d | 1931 | { |
b53b0b9d | 1932 | struct pid *pid = file->private_data; |
9e77716a | 1933 | __poll_t poll_flags = 0; |
b53b0b9d JFG |
1934 | |
1935 | poll_wait(file, &pid->wait_pidfd, pts); | |
1936 | ||
b53b0b9d JFG |
1937 | /* |
1938 | * Inform pollers only when the whole thread group exits. | |
1939 | * If the thread group leader exits before all other threads in the | |
1940 | * group, then poll(2) should block, similar to the wait(2) family. | |
1941 | */ | |
38fd525a | 1942 | if (thread_group_exited(pid)) |
9e77716a | 1943 | poll_flags = EPOLLIN | EPOLLRDNORM; |
b53b0b9d JFG |
1944 | |
1945 | return poll_flags; | |
1946 | } | |
1947 | ||
b3e58382 CB |
1948 | const struct file_operations pidfd_fops = { |
1949 | .release = pidfd_release, | |
b53b0b9d | 1950 | .poll = pidfd_poll, |
b3e58382 CB |
1951 | #ifdef CONFIG_PROC_FS |
1952 | .show_fdinfo = pidfd_show_fdinfo, | |
1953 | #endif | |
1954 | }; | |
1955 | ||
c3f3ce04 AA |
1956 | static void __delayed_free_task(struct rcu_head *rhp) |
1957 | { | |
1958 | struct task_struct *tsk = container_of(rhp, struct task_struct, rcu); | |
1959 | ||
1960 | free_task(tsk); | |
1961 | } | |
1962 | ||
1963 | static __always_inline void delayed_free_task(struct task_struct *tsk) | |
1964 | { | |
1965 | if (IS_ENABLED(CONFIG_MEMCG)) | |
1966 | call_rcu(&tsk->rcu, __delayed_free_task); | |
1967 | else | |
1968 | free_task(tsk); | |
1969 | } | |
1970 | ||
67197a4f SB |
1971 | static void copy_oom_score_adj(u64 clone_flags, struct task_struct *tsk) |
1972 | { | |
1973 | /* Skip if kernel thread */ | |
1974 | if (!tsk->mm) | |
1975 | return; | |
1976 | ||
1977 | /* Skip if spawning a thread or using vfork */ | |
1978 | if ((clone_flags & (CLONE_VM | CLONE_THREAD | CLONE_VFORK)) != CLONE_VM) | |
1979 | return; | |
1980 | ||
1981 | /* We need to synchronize with __set_oom_adj */ | |
1982 | mutex_lock(&oom_adj_mutex); | |
1983 | set_bit(MMF_MULTIPROCESS, &tsk->mm->flags); | |
1984 | /* Update the values in case they were changed after copy_signal */ | |
1985 | tsk->signal->oom_score_adj = current->signal->oom_score_adj; | |
1986 | tsk->signal->oom_score_adj_min = current->signal->oom_score_adj_min; | |
1987 | mutex_unlock(&oom_adj_mutex); | |
1988 | } | |
1989 | ||
79257534 DBO |
1990 | #ifdef CONFIG_RV |
1991 | static void rv_task_fork(struct task_struct *p) | |
1992 | { | |
1993 | int i; | |
1994 | ||
1995 | for (i = 0; i < RV_PER_TASK_MONITORS; i++) | |
1996 | p->rv[i].da_mon.monitoring = false; | |
1997 | } | |
1998 | #else | |
1999 | #define rv_task_fork(p) do {} while (0) | |
2000 | #endif | |
2001 | ||
1da177e4 LT |
2002 | /* |
2003 | * This creates a new process as a copy of the old one, | |
2004 | * but does not actually start it yet. | |
2005 | * | |
2006 | * It copies the registers, and all the appropriate | |
2007 | * parts of the process environment (as per the clone | |
2008 | * flags). The actual kick-off is left to the caller. | |
2009 | */ | |
0766f788 | 2010 | static __latent_entropy struct task_struct *copy_process( |
09a05394 | 2011 | struct pid *pid, |
3033f14a | 2012 | int trace, |
7f192e3c CB |
2013 | int node, |
2014 | struct kernel_clone_args *args) | |
1da177e4 | 2015 | { |
b3e58382 | 2016 | int pidfd = -1, retval; |
a24efe62 | 2017 | struct task_struct *p; |
c3ad2c3b | 2018 | struct multiprocess_signals delayed; |
6fd2fe49 | 2019 | struct file *pidfile = NULL; |
c5febea0 | 2020 | const u64 clone_flags = args->flags; |
769071ac | 2021 | struct nsproxy *nsp = current->nsproxy; |
1da177e4 | 2022 | |
667b6094 MPS |
2023 | /* |
2024 | * Don't allow sharing the root directory with processes in a different | |
2025 | * namespace | |
2026 | */ | |
1da177e4 LT |
2027 | if ((clone_flags & (CLONE_NEWNS|CLONE_FS)) == (CLONE_NEWNS|CLONE_FS)) |
2028 | return ERR_PTR(-EINVAL); | |
2029 | ||
e66eded8 EB |
2030 | if ((clone_flags & (CLONE_NEWUSER|CLONE_FS)) == (CLONE_NEWUSER|CLONE_FS)) |
2031 | return ERR_PTR(-EINVAL); | |
2032 | ||
1da177e4 LT |
2033 | /* |
2034 | * Thread groups must share signals as well, and detached threads | |
2035 | * can only be started up within the thread group. | |
2036 | */ | |
2037 | if ((clone_flags & CLONE_THREAD) && !(clone_flags & CLONE_SIGHAND)) | |
2038 | return ERR_PTR(-EINVAL); | |
2039 | ||
2040 | /* | |
2041 | * Shared signal handlers imply shared VM. By way of the above, | |
2042 | * thread groups also imply shared VM. Blocking this case allows | |
2043 | * for various simplifications in other code. | |
2044 | */ | |
2045 | if ((clone_flags & CLONE_SIGHAND) && !(clone_flags & CLONE_VM)) | |
2046 | return ERR_PTR(-EINVAL); | |
2047 | ||
123be07b SB |
2048 | /* |
2049 | * Siblings of global init remain as zombies on exit since they are | |
2050 | * not reaped by their parent (swapper). To solve this and to avoid | |
2051 | * multi-rooted process trees, prevent global and container-inits | |
2052 | * from creating siblings. | |
2053 | */ | |
2054 | if ((clone_flags & CLONE_PARENT) && | |
2055 | current->signal->flags & SIGNAL_UNKILLABLE) | |
2056 | return ERR_PTR(-EINVAL); | |
2057 | ||
8382fcac | 2058 | /* |
40a0d32d | 2059 | * If the new process will be in a different pid or user namespace |
faf00da5 | 2060 | * do not allow it to share a thread group with the forking task. |
8382fcac | 2061 | */ |
faf00da5 | 2062 | if (clone_flags & CLONE_THREAD) { |
40a0d32d | 2063 | if ((clone_flags & (CLONE_NEWUSER | CLONE_NEWPID)) || |
769071ac AV |
2064 | (task_active_pid_ns(current) != nsp->pid_ns_for_children)) |
2065 | return ERR_PTR(-EINVAL); | |
2066 | } | |
2067 | ||
b3e58382 | 2068 | if (clone_flags & CLONE_PIDFD) { |
b3e58382 | 2069 | /* |
b3e58382 CB |
2070 | * - CLONE_DETACHED is blocked so that we can potentially |
2071 | * reuse it later for CLONE_PIDFD. | |
2072 | * - CLONE_THREAD is blocked until someone really needs it. | |
2073 | */ | |
7f192e3c | 2074 | if (clone_flags & (CLONE_DETACHED | CLONE_THREAD)) |
b3e58382 | 2075 | return ERR_PTR(-EINVAL); |
b3e58382 CB |
2076 | } |
2077 | ||
c3ad2c3b EB |
2078 | /* |
2079 | * Force any signals received before this point to be delivered | |
2080 | * before the fork happens. Collect up signals sent to multiple | |
2081 | * processes that happen during the fork and delay them so that | |
2082 | * they appear to happen after the fork. | |
2083 | */ | |
2084 | sigemptyset(&delayed.signal); | |
2085 | INIT_HLIST_NODE(&delayed.node); | |
2086 | ||
2087 | spin_lock_irq(¤t->sighand->siglock); | |
2088 | if (!(clone_flags & CLONE_THREAD)) | |
2089 | hlist_add_head(&delayed.node, ¤t->signal->multiprocess); | |
2090 | recalc_sigpending(); | |
2091 | spin_unlock_irq(¤t->sighand->siglock); | |
2092 | retval = -ERESTARTNOINTR; | |
66ae0d1e | 2093 | if (task_sigpending(current)) |
c3ad2c3b EB |
2094 | goto fork_out; |
2095 | ||
1da177e4 | 2096 | retval = -ENOMEM; |
725fc629 | 2097 | p = dup_task_struct(current, node); |
1da177e4 LT |
2098 | if (!p) |
2099 | goto fork_out; | |
753550eb EB |
2100 | p->flags &= ~PF_KTHREAD; |
2101 | if (args->kthread) | |
2102 | p->flags |= PF_KTHREAD; | |
b16b3855 JA |
2103 | if (args->io_thread) { |
2104 | /* | |
2105 | * Mark us an IO worker, and block any signal that isn't | |
2106 | * fatal or STOP | |
2107 | */ | |
cc440e87 | 2108 | p->flags |= PF_IO_WORKER; |
b16b3855 JA |
2109 | siginitsetinv(&p->blocked, sigmask(SIGKILL)|sigmask(SIGSTOP)); |
2110 | } | |
1da177e4 | 2111 | |
7f192e3c | 2112 | p->set_child_tid = (clone_flags & CLONE_CHILD_SETTID) ? args->child_tid : NULL; |
4d6501dc VN |
2113 | /* |
2114 | * Clear TID on mm_release()? | |
2115 | */ | |
7f192e3c | 2116 | p->clear_child_tid = (clone_flags & CLONE_CHILD_CLEARTID) ? args->child_tid : NULL; |
4d6501dc | 2117 | |
f7e8b616 SR |
2118 | ftrace_graph_init_task(p); |
2119 | ||
bea493a0 PZ |
2120 | rt_mutex_init_task(p); |
2121 | ||
a21ee605 | 2122 | lockdep_assert_irqs_enabled(); |
d12c1a37 | 2123 | #ifdef CONFIG_PROVE_LOCKING |
de30a2b3 IM |
2124 | DEBUG_LOCKS_WARN_ON(!p->softirqs_enabled); |
2125 | #endif | |
8f2f9c4d EB |
2126 | retval = copy_creds(p, clone_flags); |
2127 | if (retval < 0) | |
2128 | goto bad_fork_free; | |
2129 | ||
1da177e4 | 2130 | retval = -EAGAIN; |
de399236 | 2131 | if (is_rlimit_overlimit(task_ucounts(p), UCOUNT_RLIMIT_NPROC, rlimit(RLIMIT_NPROC))) { |
b57922b6 EP |
2132 | if (p->real_cred->user != INIT_USER && |
2133 | !capable(CAP_SYS_RESOURCE) && !capable(CAP_SYS_ADMIN)) | |
8f2f9c4d | 2134 | goto bad_fork_cleanup_count; |
1da177e4 | 2135 | } |
72fa5997 | 2136 | current->flags &= ~PF_NPROC_EXCEEDED; |
1da177e4 | 2137 | |
1da177e4 LT |
2138 | /* |
2139 | * If multiple threads are within copy_process(), then this check | |
2140 | * triggers too late. This doesn't hurt, the check is only there | |
2141 | * to stop root fork bombs. | |
2142 | */ | |
04ec93fe | 2143 | retval = -EAGAIN; |
c17d1a3a | 2144 | if (data_race(nr_threads >= max_threads)) |
1da177e4 LT |
2145 | goto bad_fork_cleanup_count; |
2146 | ||
ca74e92b | 2147 | delayacct_tsk_init(p); /* Must remain after dup_task_struct() */ |
a8ea6fc9 | 2148 | p->flags &= ~(PF_SUPERPRIV | PF_WQ_WORKER | PF_IDLE | PF_NO_SETAFFINITY); |
514ddb44 | 2149 | p->flags |= PF_FORKNOEXEC; |
1da177e4 LT |
2150 | INIT_LIST_HEAD(&p->children); |
2151 | INIT_LIST_HEAD(&p->sibling); | |
f41d911f | 2152 | rcu_copy_process(p); |
1da177e4 LT |
2153 | p->vfork_done = NULL; |
2154 | spin_lock_init(&p->alloc_lock); | |
1da177e4 | 2155 | |
1da177e4 LT |
2156 | init_sigpending(&p->pending); |
2157 | ||
64861634 | 2158 | p->utime = p->stime = p->gtime = 0; |
40565b5a | 2159 | #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME |
64861634 | 2160 | p->utimescaled = p->stimescaled = 0; |
40565b5a | 2161 | #endif |
9d7fb042 PZ |
2162 | prev_cputime_init(&p->prev_cputime); |
2163 | ||
6a61671b | 2164 | #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN |
bac5b6b6 FW |
2165 | seqcount_init(&p->vtime.seqcount); |
2166 | p->vtime.starttime = 0; | |
2167 | p->vtime.state = VTIME_INACTIVE; | |
6a61671b FW |
2168 | #endif |
2169 | ||
0f212204 JA |
2170 | #ifdef CONFIG_IO_URING |
2171 | p->io_uring = NULL; | |
2172 | #endif | |
2173 | ||
a3a2e76c KH |
2174 | #if defined(SPLIT_RSS_COUNTING) |
2175 | memset(&p->rss_stat, 0, sizeof(p->rss_stat)); | |
2176 | #endif | |
172ba844 | 2177 | |
6976675d AV |
2178 | p->default_timer_slack_ns = current->timer_slack_ns; |
2179 | ||
eb414681 JW |
2180 | #ifdef CONFIG_PSI |
2181 | p->psi_flags = 0; | |
2182 | #endif | |
2183 | ||
5995477a | 2184 | task_io_accounting_init(&p->ioac); |
1da177e4 LT |
2185 | acct_clear_integrals(p); |
2186 | ||
3a245c0f | 2187 | posix_cputimers_init(&p->posix_cputimers); |
1da177e4 | 2188 | |
1da177e4 | 2189 | p->io_context = NULL; |
c0b0ae8a | 2190 | audit_set_context(p, NULL); |
b4f48b63 | 2191 | cgroup_fork(p); |
343f4c49 | 2192 | if (args->kthread) { |
40966e31 | 2193 | if (!set_kthread_struct(p)) |
ff8288ff | 2194 | goto bad_fork_cleanup_delayacct; |
40966e31 | 2195 | } |
1da177e4 | 2196 | #ifdef CONFIG_NUMA |
846a16bf | 2197 | p->mempolicy = mpol_dup(p->mempolicy); |
fb0a685c DRO |
2198 | if (IS_ERR(p->mempolicy)) { |
2199 | retval = PTR_ERR(p->mempolicy); | |
2200 | p->mempolicy = NULL; | |
ff8288ff | 2201 | goto bad_fork_cleanup_delayacct; |
fb0a685c | 2202 | } |
1da177e4 | 2203 | #endif |
778d3b0f MH |
2204 | #ifdef CONFIG_CPUSETS |
2205 | p->cpuset_mem_spread_rotor = NUMA_NO_NODE; | |
2206 | p->cpuset_slab_spread_rotor = NUMA_NO_NODE; | |
b7505861 | 2207 | seqcount_spinlock_init(&p->mems_allowed_seq, &p->alloc_lock); |
778d3b0f | 2208 | #endif |
de30a2b3 | 2209 | #ifdef CONFIG_TRACE_IRQFLAGS |
0584df9c ME |
2210 | memset(&p->irqtrace, 0, sizeof(p->irqtrace)); |
2211 | p->irqtrace.hardirq_disable_ip = _THIS_IP_; | |
2212 | p->irqtrace.softirq_enable_ip = _THIS_IP_; | |
2213 | p->softirqs_enabled = 1; | |
2214 | p->softirq_context = 0; | |
de30a2b3 | 2215 | #endif |
8bcbde54 DH |
2216 | |
2217 | p->pagefault_disabled = 0; | |
2218 | ||
fbb9ce95 | 2219 | #ifdef CONFIG_LOCKDEP |
b09be676 | 2220 | lockdep_init_task(p); |
fbb9ce95 | 2221 | #endif |
1da177e4 | 2222 | |
408894ee IM |
2223 | #ifdef CONFIG_DEBUG_MUTEXES |
2224 | p->blocked_on = NULL; /* not blocked yet */ | |
2225 | #endif | |
cafe5635 KO |
2226 | #ifdef CONFIG_BCACHE |
2227 | p->sequential_io = 0; | |
2228 | p->sequential_io_avg = 0; | |
2229 | #endif | |
a10787e6 SL |
2230 | #ifdef CONFIG_BPF_SYSCALL |
2231 | RCU_INIT_POINTER(p->bpf_storage, NULL); | |
c7603cfa | 2232 | p->bpf_ctx = NULL; |
a10787e6 | 2233 | #endif |
0f481406 | 2234 | |
3c90e6e9 | 2235 | /* Perform scheduler related setup. Assign this task to a CPU. */ |
aab03e05 DF |
2236 | retval = sched_fork(clone_flags, p); |
2237 | if (retval) | |
2238 | goto bad_fork_cleanup_policy; | |
6ab423e0 | 2239 | |
2b26f0aa | 2240 | retval = perf_event_init_task(p, clone_flags); |
6ab423e0 PZ |
2241 | if (retval) |
2242 | goto bad_fork_cleanup_policy; | |
fb0a685c DRO |
2243 | retval = audit_alloc(p); |
2244 | if (retval) | |
6c72e350 | 2245 | goto bad_fork_cleanup_perf; |
1da177e4 | 2246 | /* copy all the process information */ |
ab602f79 | 2247 | shm_init_task(p); |
e4e55b47 | 2248 | retval = security_task_alloc(p, clone_flags); |
fb0a685c | 2249 | if (retval) |
1da177e4 | 2250 | goto bad_fork_cleanup_audit; |
e4e55b47 TH |
2251 | retval = copy_semundo(clone_flags, p); |
2252 | if (retval) | |
2253 | goto bad_fork_cleanup_security; | |
fb0a685c DRO |
2254 | retval = copy_files(clone_flags, p); |
2255 | if (retval) | |
1da177e4 | 2256 | goto bad_fork_cleanup_semundo; |
fb0a685c DRO |
2257 | retval = copy_fs(clone_flags, p); |
2258 | if (retval) | |
1da177e4 | 2259 | goto bad_fork_cleanup_files; |
fb0a685c DRO |
2260 | retval = copy_sighand(clone_flags, p); |
2261 | if (retval) | |
1da177e4 | 2262 | goto bad_fork_cleanup_fs; |
fb0a685c DRO |
2263 | retval = copy_signal(clone_flags, p); |
2264 | if (retval) | |
1da177e4 | 2265 | goto bad_fork_cleanup_sighand; |
fb0a685c DRO |
2266 | retval = copy_mm(clone_flags, p); |
2267 | if (retval) | |
1da177e4 | 2268 | goto bad_fork_cleanup_signal; |
fb0a685c DRO |
2269 | retval = copy_namespaces(clone_flags, p); |
2270 | if (retval) | |
d84f4f99 | 2271 | goto bad_fork_cleanup_mm; |
fb0a685c DRO |
2272 | retval = copy_io(clone_flags, p); |
2273 | if (retval) | |
fd0928df | 2274 | goto bad_fork_cleanup_namespaces; |
c5febea0 | 2275 | retval = copy_thread(p, args); |
1da177e4 | 2276 | if (retval) |
fd0928df | 2277 | goto bad_fork_cleanup_io; |
1da177e4 | 2278 | |
afaef01c AP |
2279 | stackleak_task_init(p); |
2280 | ||
425fb2b4 | 2281 | if (pid != &init_struct_pid) { |
49cb2fc4 AR |
2282 | pid = alloc_pid(p->nsproxy->pid_ns_for_children, args->set_tid, |
2283 | args->set_tid_size); | |
35f71bc0 MH |
2284 | if (IS_ERR(pid)) { |
2285 | retval = PTR_ERR(pid); | |
0740aa5f | 2286 | goto bad_fork_cleanup_thread; |
35f71bc0 | 2287 | } |
425fb2b4 PE |
2288 | } |
2289 | ||
b3e58382 CB |
2290 | /* |
2291 | * This has to happen after we've potentially unshared the file | |
2292 | * descriptor table (so that the pidfd doesn't leak into the child | |
2293 | * if the fd table isn't shared). | |
2294 | */ | |
2295 | if (clone_flags & CLONE_PIDFD) { | |
6fd2fe49 | 2296 | retval = get_unused_fd_flags(O_RDWR | O_CLOEXEC); |
b3e58382 CB |
2297 | if (retval < 0) |
2298 | goto bad_fork_free_pid; | |
2299 | ||
2300 | pidfd = retval; | |
6fd2fe49 AV |
2301 | |
2302 | pidfile = anon_inode_getfile("[pidfd]", &pidfd_fops, pid, | |
2303 | O_RDWR | O_CLOEXEC); | |
2304 | if (IS_ERR(pidfile)) { | |
2305 | put_unused_fd(pidfd); | |
28dd29c0 | 2306 | retval = PTR_ERR(pidfile); |
6fd2fe49 AV |
2307 | goto bad_fork_free_pid; |
2308 | } | |
2309 | get_pid(pid); /* held by pidfile now */ | |
2310 | ||
7f192e3c | 2311 | retval = put_user(pidfd, args->pidfd); |
b3e58382 CB |
2312 | if (retval) |
2313 | goto bad_fork_put_pidfd; | |
2314 | } | |
2315 | ||
73c10101 JA |
2316 | #ifdef CONFIG_BLOCK |
2317 | p->plug = NULL; | |
2318 | #endif | |
ba31c1a4 TG |
2319 | futex_init_task(p); |
2320 | ||
f9a3879a GM |
2321 | /* |
2322 | * sigaltstack should be cleared when sharing the same VM | |
2323 | */ | |
2324 | if ((clone_flags & (CLONE_VM|CLONE_VFORK)) == CLONE_VM) | |
2a742138 | 2325 | sas_ss_reset(p); |
f9a3879a | 2326 | |
1da177e4 | 2327 | /* |
6580807d ON |
2328 | * Syscall tracing and stepping should be turned off in the |
2329 | * child regardless of CLONE_PTRACE. | |
1da177e4 | 2330 | */ |
6580807d | 2331 | user_disable_single_step(p); |
64c19ba2 | 2332 | clear_task_syscall_work(p, SYSCALL_TRACE); |
64eb35f7 GKB |
2333 | #if defined(CONFIG_GENERIC_ENTRY) || defined(TIF_SYSCALL_EMU) |
2334 | clear_task_syscall_work(p, SYSCALL_EMU); | |
ed75e8d5 | 2335 | #endif |
e02c9b0d | 2336 | clear_tsk_latency_tracing(p); |
1da177e4 | 2337 | |
1da177e4 | 2338 | /* ok, now we should be set up.. */ |
18c830df ON |
2339 | p->pid = pid_nr(pid); |
2340 | if (clone_flags & CLONE_THREAD) { | |
18c830df ON |
2341 | p->group_leader = current->group_leader; |
2342 | p->tgid = current->tgid; | |
2343 | } else { | |
18c830df ON |
2344 | p->group_leader = p; |
2345 | p->tgid = p->pid; | |
2346 | } | |
5f8aadd8 | 2347 | |
9d823e8f WF |
2348 | p->nr_dirtied = 0; |
2349 | p->nr_dirtied_pause = 128 >> (PAGE_SHIFT - 10); | |
83712358 | 2350 | p->dirty_paused_when = 0; |
9d823e8f | 2351 | |
bb8cbbfe | 2352 | p->pdeath_signal = 0; |
47e65328 | 2353 | INIT_LIST_HEAD(&p->thread_group); |
158e1645 | 2354 | p->task_works = NULL; |
ca7752ca | 2355 | clear_posix_cputimers_work(p); |
1da177e4 | 2356 | |
d741bf41 PZ |
2357 | #ifdef CONFIG_KRETPROBES |
2358 | p->kretprobe_instances.first = NULL; | |
2359 | #endif | |
54ecbe6f MH |
2360 | #ifdef CONFIG_RETHOOK |
2361 | p->rethooks.first = NULL; | |
2362 | #endif | |
d741bf41 | 2363 | |
7e47682e AS |
2364 | /* |
2365 | * Ensure that the cgroup subsystem policies allow the new process to be | |
7b7b8a2c | 2366 | * forked. It should be noted that the new process's css_set can be changed |
7e47682e AS |
2367 | * between here and cgroup_post_fork() if an organisation operation is in |
2368 | * progress. | |
2369 | */ | |
ef2c41cf | 2370 | retval = cgroup_can_fork(p, args); |
7e47682e | 2371 | if (retval) |
5a5cf5cb | 2372 | goto bad_fork_put_pidfd; |
7e47682e | 2373 | |
b1e82065 PZ |
2374 | /* |
2375 | * Now that the cgroups are pinned, re-clone the parent cgroup and put | |
2376 | * the new task on the correct runqueue. All this *before* the task | |
2377 | * becomes visible. | |
2378 | * | |
2379 | * This isn't part of ->can_fork() because while the re-cloning is | |
2380 | * cgroup specific, it unconditionally needs to place the task on a | |
2381 | * runqueue. | |
2382 | */ | |
2383 | sched_cgroup_fork(p, args); | |
2384 | ||
7b558513 DH |
2385 | /* |
2386 | * From this point on we must avoid any synchronous user-space | |
2387 | * communication until we take the tasklist-lock. In particular, we do | |
2388 | * not want user-space to be able to predict the process start-time by | |
2389 | * stalling fork(2) after we recorded the start_time but before it is | |
2390 | * visible to the system. | |
2391 | */ | |
2392 | ||
2393 | p->start_time = ktime_get_ns(); | |
cf25e24d | 2394 | p->start_boottime = ktime_get_boottime_ns(); |
7b558513 | 2395 | |
18c830df ON |
2396 | /* |
2397 | * Make it visible to the rest of the system, but dont wake it up yet. | |
2398 | * Need tasklist lock for parent etc handling! | |
2399 | */ | |
1da177e4 LT |
2400 | write_lock_irq(&tasklist_lock); |
2401 | ||
1da177e4 | 2402 | /* CLONE_PARENT re-uses the old parent */ |
2d5516cb | 2403 | if (clone_flags & (CLONE_PARENT|CLONE_THREAD)) { |
1da177e4 | 2404 | p->real_parent = current->real_parent; |
2d5516cb | 2405 | p->parent_exec_id = current->parent_exec_id; |
b4e00444 EW |
2406 | if (clone_flags & CLONE_THREAD) |
2407 | p->exit_signal = -1; | |
2408 | else | |
2409 | p->exit_signal = current->group_leader->exit_signal; | |
2d5516cb | 2410 | } else { |
1da177e4 | 2411 | p->real_parent = current; |
2d5516cb | 2412 | p->parent_exec_id = current->self_exec_id; |
b4e00444 | 2413 | p->exit_signal = args->exit_signal; |
2d5516cb | 2414 | } |
1da177e4 | 2415 | |
d83a7cb3 JP |
2416 | klp_copy_process(p); |
2417 | ||
85dd3f61 PZ |
2418 | sched_core_fork(p); |
2419 | ||
3f17da69 | 2420 | spin_lock(¤t->sighand->siglock); |
4a2c7a78 | 2421 | |
79257534 DBO |
2422 | rv_task_fork(p); |
2423 | ||
d7822b1e MD |
2424 | rseq_fork(p, clone_flags); |
2425 | ||
4ca1d3ee | 2426 | /* Don't start children in a dying pid namespace */ |
e8cfbc24 | 2427 | if (unlikely(!(ns_of_pid(pid)->pid_allocated & PIDNS_ADDING))) { |
3fd37226 KT |
2428 | retval = -ENOMEM; |
2429 | goto bad_fork_cancel_cgroup; | |
2430 | } | |
4a2c7a78 | 2431 | |
7673bf55 EB |
2432 | /* Let kill terminate clone/fork in the middle */ |
2433 | if (fatal_signal_pending(current)) { | |
2434 | retval = -EINTR; | |
2435 | goto bad_fork_cancel_cgroup; | |
2436 | } | |
2437 | ||
a1140cb2 KI |
2438 | /* No more failure paths after this point. */ |
2439 | ||
2440 | /* | |
2441 | * Copy seccomp details explicitly here, in case they were changed | |
2442 | * before holding sighand lock. | |
2443 | */ | |
2444 | copy_seccomp(p); | |
2445 | ||
2c470475 | 2446 | init_task_pid_links(p); |
73b9ebfe | 2447 | if (likely(p->pid)) { |
4b9d33e6 | 2448 | ptrace_init_task(p, (clone_flags & CLONE_PTRACE) || trace); |
73b9ebfe | 2449 | |
81907739 | 2450 | init_task_pid(p, PIDTYPE_PID, pid); |
73b9ebfe | 2451 | if (thread_group_leader(p)) { |
6883f81a | 2452 | init_task_pid(p, PIDTYPE_TGID, pid); |
81907739 ON |
2453 | init_task_pid(p, PIDTYPE_PGID, task_pgrp(current)); |
2454 | init_task_pid(p, PIDTYPE_SID, task_session(current)); | |
2455 | ||
1c4042c2 | 2456 | if (is_child_reaper(pid)) { |
17cf22c3 | 2457 | ns_of_pid(pid)->child_reaper = p; |
1c4042c2 EB |
2458 | p->signal->flags |= SIGNAL_UNKILLABLE; |
2459 | } | |
c3ad2c3b | 2460 | p->signal->shared_pending.signal = delayed.signal; |
9c9f4ded | 2461 | p->signal->tty = tty_kref_get(current->signal->tty); |
749860ce PT |
2462 | /* |
2463 | * Inherit has_child_subreaper flag under the same | |
2464 | * tasklist_lock with adding child to the process tree | |
2465 | * for propagate_has_child_subreaper optimization. | |
2466 | */ | |
2467 | p->signal->has_child_subreaper = p->real_parent->signal->has_child_subreaper || | |
2468 | p->real_parent->signal->is_child_subreaper; | |
9cd80bbb | 2469 | list_add_tail(&p->sibling, &p->real_parent->children); |
5e85d4ab | 2470 | list_add_tail_rcu(&p->tasks, &init_task.tasks); |
6883f81a | 2471 | attach_pid(p, PIDTYPE_TGID); |
81907739 ON |
2472 | attach_pid(p, PIDTYPE_PGID); |
2473 | attach_pid(p, PIDTYPE_SID); | |
909ea964 | 2474 | __this_cpu_inc(process_counts); |
80628ca0 ON |
2475 | } else { |
2476 | current->signal->nr_threads++; | |
d80f7d7b | 2477 | current->signal->quick_threads++; |
80628ca0 | 2478 | atomic_inc(¤t->signal->live); |
60d4de3f | 2479 | refcount_inc(¤t->signal->sigcnt); |
924de3b8 | 2480 | task_join_group_stop(p); |
80628ca0 ON |
2481 | list_add_tail_rcu(&p->thread_group, |
2482 | &p->group_leader->thread_group); | |
0c740d0a ON |
2483 | list_add_tail_rcu(&p->thread_node, |
2484 | &p->signal->thread_head); | |
73b9ebfe | 2485 | } |
81907739 | 2486 | attach_pid(p, PIDTYPE_PID); |
73b9ebfe | 2487 | nr_threads++; |
1da177e4 | 2488 | } |
1da177e4 | 2489 | total_forks++; |
c3ad2c3b | 2490 | hlist_del_init(&delayed.node); |
3f17da69 | 2491 | spin_unlock(¤t->sighand->siglock); |
4af4206b | 2492 | syscall_tracepoint_update(p); |
1da177e4 | 2493 | write_unlock_irq(&tasklist_lock); |
4af4206b | 2494 | |
ddc204b5 WL |
2495 | if (pidfile) |
2496 | fd_install(pidfd, pidfile); | |
2497 | ||
c13cf856 | 2498 | proc_fork_connector(p); |
b1e82065 | 2499 | sched_post_fork(p); |
ef2c41cf | 2500 | cgroup_post_fork(p, args); |
cdd6c482 | 2501 | perf_event_fork(p); |
43d2b113 KH |
2502 | |
2503 | trace_task_newtask(p, clone_flags); | |
3ab67966 | 2504 | uprobe_copy_process(p, clone_flags); |
43d2b113 | 2505 | |
67197a4f SB |
2506 | copy_oom_score_adj(clone_flags, p); |
2507 | ||
1da177e4 LT |
2508 | return p; |
2509 | ||
7e47682e | 2510 | bad_fork_cancel_cgroup: |
85dd3f61 | 2511 | sched_core_free(p); |
3fd37226 KT |
2512 | spin_unlock(¤t->sighand->siglock); |
2513 | write_unlock_irq(&tasklist_lock); | |
ef2c41cf | 2514 | cgroup_cancel_fork(p, args); |
b3e58382 | 2515 | bad_fork_put_pidfd: |
6fd2fe49 AV |
2516 | if (clone_flags & CLONE_PIDFD) { |
2517 | fput(pidfile); | |
2518 | put_unused_fd(pidfd); | |
2519 | } | |
425fb2b4 PE |
2520 | bad_fork_free_pid: |
2521 | if (pid != &init_struct_pid) | |
2522 | free_pid(pid); | |
0740aa5f JS |
2523 | bad_fork_cleanup_thread: |
2524 | exit_thread(p); | |
fd0928df | 2525 | bad_fork_cleanup_io: |
b69f2292 LR |
2526 | if (p->io_context) |
2527 | exit_io_context(p); | |
ab516013 | 2528 | bad_fork_cleanup_namespaces: |
444f378b | 2529 | exit_task_namespaces(p); |
1da177e4 | 2530 | bad_fork_cleanup_mm: |
c3f3ce04 AA |
2531 | if (p->mm) { |
2532 | mm_clear_owner(p->mm, p); | |
1da177e4 | 2533 | mmput(p->mm); |
c3f3ce04 | 2534 | } |
1da177e4 | 2535 | bad_fork_cleanup_signal: |
4ab6c083 | 2536 | if (!(clone_flags & CLONE_THREAD)) |
1c5354de | 2537 | free_signal_struct(p->signal); |
1da177e4 | 2538 | bad_fork_cleanup_sighand: |
a7e5328a | 2539 | __cleanup_sighand(p->sighand); |
1da177e4 LT |
2540 | bad_fork_cleanup_fs: |
2541 | exit_fs(p); /* blocking */ | |
2542 | bad_fork_cleanup_files: | |
2543 | exit_files(p); /* blocking */ | |
2544 | bad_fork_cleanup_semundo: | |
2545 | exit_sem(p); | |
e4e55b47 TH |
2546 | bad_fork_cleanup_security: |
2547 | security_task_free(p); | |
1da177e4 LT |
2548 | bad_fork_cleanup_audit: |
2549 | audit_free(p); | |
6c72e350 | 2550 | bad_fork_cleanup_perf: |
cdd6c482 | 2551 | perf_event_free_task(p); |
6c72e350 | 2552 | bad_fork_cleanup_policy: |
b09be676 | 2553 | lockdep_free_task(p); |
1da177e4 | 2554 | #ifdef CONFIG_NUMA |
f0be3d32 | 2555 | mpol_put(p->mempolicy); |
1da177e4 | 2556 | #endif |
ff8288ff | 2557 | bad_fork_cleanup_delayacct: |
35df17c5 | 2558 | delayacct_tsk_free(p); |
1da177e4 | 2559 | bad_fork_cleanup_count: |
21d1c5e3 | 2560 | dec_rlimit_ucounts(task_ucounts(p), UCOUNT_RLIMIT_NPROC, 1); |
e0e81739 | 2561 | exit_creds(p); |
1da177e4 | 2562 | bad_fork_free: |
2f064a59 | 2563 | WRITE_ONCE(p->__state, TASK_DEAD); |
1a03d3f1 | 2564 | exit_task_stack_account(p); |
68f24b08 | 2565 | put_task_stack(p); |
c3f3ce04 | 2566 | delayed_free_task(p); |
fe7d37d1 | 2567 | fork_out: |
c3ad2c3b EB |
2568 | spin_lock_irq(¤t->sighand->siglock); |
2569 | hlist_del_init(&delayed.node); | |
2570 | spin_unlock_irq(¤t->sighand->siglock); | |
fe7d37d1 | 2571 | return ERR_PTR(retval); |
1da177e4 LT |
2572 | } |
2573 | ||
2c470475 | 2574 | static inline void init_idle_pids(struct task_struct *idle) |
f106eee1 ON |
2575 | { |
2576 | enum pid_type type; | |
2577 | ||
2578 | for (type = PIDTYPE_PID; type < PIDTYPE_MAX; ++type) { | |
2c470475 EB |
2579 | INIT_HLIST_NODE(&idle->pid_links[type]); /* not really needed */ |
2580 | init_task_pid(idle, type, &init_struct_pid); | |
f106eee1 ON |
2581 | } |
2582 | } | |
2583 | ||
36cb0e1c EB |
2584 | static int idle_dummy(void *dummy) |
2585 | { | |
2586 | /* This function is never called */ | |
2587 | return 0; | |
2588 | } | |
2589 | ||
f1a0a376 | 2590 | struct task_struct * __init fork_idle(int cpu) |
1da177e4 | 2591 | { |
36c8b586 | 2592 | struct task_struct *task; |
7f192e3c | 2593 | struct kernel_clone_args args = { |
343f4c49 | 2594 | .flags = CLONE_VM, |
5bd2e97c EB |
2595 | .fn = &idle_dummy, |
2596 | .fn_arg = NULL, | |
343f4c49 | 2597 | .kthread = 1, |
36cb0e1c | 2598 | .idle = 1, |
7f192e3c CB |
2599 | }; |
2600 | ||
2601 | task = copy_process(&init_struct_pid, 0, cpu_to_node(cpu), &args); | |
f106eee1 | 2602 | if (!IS_ERR(task)) { |
2c470475 | 2603 | init_idle_pids(task); |
753ca4f3 | 2604 | init_idle(task, cpu); |
f106eee1 | 2605 | } |
73b9ebfe | 2606 | |
1da177e4 LT |
2607 | return task; |
2608 | } | |
2609 | ||
13585fa0 NA |
2610 | struct mm_struct *copy_init_mm(void) |
2611 | { | |
2612 | return dup_mm(NULL, &init_mm); | |
2613 | } | |
2614 | ||
cc440e87 JA |
2615 | /* |
2616 | * This is like kernel_clone(), but shaved down and tailored to just | |
2617 | * creating io_uring workers. It returns a created task, or an error pointer. | |
2618 | * The returned task is inactive, and the caller must fire it up through | |
2619 | * wake_up_new_task(p). All signals are blocked in the created task. | |
2620 | */ | |
2621 | struct task_struct *create_io_thread(int (*fn)(void *), void *arg, int node) | |
2622 | { | |
2623 | unsigned long flags = CLONE_FS|CLONE_FILES|CLONE_SIGHAND|CLONE_THREAD| | |
2624 | CLONE_IO; | |
2625 | struct kernel_clone_args args = { | |
2626 | .flags = ((lower_32_bits(flags) | CLONE_VM | | |
2627 | CLONE_UNTRACED) & ~CSIGNAL), | |
2628 | .exit_signal = (lower_32_bits(flags) & CSIGNAL), | |
5bd2e97c EB |
2629 | .fn = fn, |
2630 | .fn_arg = arg, | |
cc440e87 JA |
2631 | .io_thread = 1, |
2632 | }; | |
cc440e87 | 2633 | |
b16b3855 | 2634 | return copy_process(NULL, 0, node, &args); |
cc440e87 JA |
2635 | } |
2636 | ||
1da177e4 LT |
2637 | /* |
2638 | * Ok, this is the main fork-routine. | |
2639 | * | |
2640 | * It copies the process, and if successful kick-starts | |
2641 | * it and waits for it to finish using the VM if required. | |
a0eb9abd ES |
2642 | * |
2643 | * args->exit_signal is expected to be checked for sanity by the caller. | |
1da177e4 | 2644 | */ |
cad6967a | 2645 | pid_t kernel_clone(struct kernel_clone_args *args) |
1da177e4 | 2646 | { |
7f192e3c | 2647 | u64 clone_flags = args->flags; |
9f5325aa MPS |
2648 | struct completion vfork; |
2649 | struct pid *pid; | |
1da177e4 LT |
2650 | struct task_struct *p; |
2651 | int trace = 0; | |
cad6967a | 2652 | pid_t nr; |
1da177e4 | 2653 | |
3af8588c CB |
2654 | /* |
2655 | * For legacy clone() calls, CLONE_PIDFD uses the parent_tid argument | |
2656 | * to return the pidfd. Hence, CLONE_PIDFD and CLONE_PARENT_SETTID are | |
2657 | * mutually exclusive. With clone3() CLONE_PIDFD has grown a separate | |
2658 | * field in struct clone_args and it still doesn't make sense to have | |
2659 | * them both point at the same memory location. Performing this check | |
2660 | * here has the advantage that we don't need to have a separate helper | |
2661 | * to check for legacy clone(). | |
2662 | */ | |
2663 | if ((args->flags & CLONE_PIDFD) && | |
2664 | (args->flags & CLONE_PARENT_SETTID) && | |
2665 | (args->pidfd == args->parent_tid)) | |
2666 | return -EINVAL; | |
2667 | ||
09a05394 | 2668 | /* |
4b9d33e6 TH |
2669 | * Determine whether and which event to report to ptracer. When |
2670 | * called from kernel_thread or CLONE_UNTRACED is explicitly | |
2671 | * requested, no event is reported; otherwise, report if the event | |
2672 | * for the type of forking is enabled. | |
09a05394 | 2673 | */ |
e80d6661 | 2674 | if (!(clone_flags & CLONE_UNTRACED)) { |
4b9d33e6 TH |
2675 | if (clone_flags & CLONE_VFORK) |
2676 | trace = PTRACE_EVENT_VFORK; | |
7f192e3c | 2677 | else if (args->exit_signal != SIGCHLD) |
4b9d33e6 TH |
2678 | trace = PTRACE_EVENT_CLONE; |
2679 | else | |
2680 | trace = PTRACE_EVENT_FORK; | |
2681 | ||
2682 | if (likely(!ptrace_event_enabled(current, trace))) | |
2683 | trace = 0; | |
2684 | } | |
1da177e4 | 2685 | |
7f192e3c | 2686 | p = copy_process(NULL, trace, NUMA_NO_NODE, args); |
38addce8 | 2687 | add_latent_entropy(); |
9f5325aa MPS |
2688 | |
2689 | if (IS_ERR(p)) | |
2690 | return PTR_ERR(p); | |
2691 | ||
1da177e4 LT |
2692 | /* |
2693 | * Do this prior waking up the new thread - the thread pointer | |
2694 | * might get invalid after that point, if the thread exits quickly. | |
2695 | */ | |
9f5325aa | 2696 | trace_sched_process_fork(current, p); |
0a16b607 | 2697 | |
9f5325aa MPS |
2698 | pid = get_task_pid(p, PIDTYPE_PID); |
2699 | nr = pid_vnr(pid); | |
30e49c26 | 2700 | |
9f5325aa | 2701 | if (clone_flags & CLONE_PARENT_SETTID) |
7f192e3c | 2702 | put_user(nr, args->parent_tid); |
a6f5e063 | 2703 | |
9f5325aa MPS |
2704 | if (clone_flags & CLONE_VFORK) { |
2705 | p->vfork_done = &vfork; | |
2706 | init_completion(&vfork); | |
2707 | get_task_struct(p); | |
2708 | } | |
1da177e4 | 2709 | |
bd74fdae YZ |
2710 | if (IS_ENABLED(CONFIG_LRU_GEN) && !(clone_flags & CLONE_VM)) { |
2711 | /* lock the task to synchronize with memcg migration */ | |
2712 | task_lock(p); | |
2713 | lru_gen_add_mm(p->mm); | |
2714 | task_unlock(p); | |
2715 | } | |
2716 | ||
9f5325aa | 2717 | wake_up_new_task(p); |
09a05394 | 2718 | |
9f5325aa MPS |
2719 | /* forking complete and child started to run, tell ptracer */ |
2720 | if (unlikely(trace)) | |
2721 | ptrace_event_pid(trace, pid); | |
4e52365f | 2722 | |
9f5325aa MPS |
2723 | if (clone_flags & CLONE_VFORK) { |
2724 | if (!wait_for_vfork_done(p, &vfork)) | |
2725 | ptrace_event_pid(PTRACE_EVENT_VFORK_DONE, pid); | |
1da177e4 | 2726 | } |
9f5325aa MPS |
2727 | |
2728 | put_pid(pid); | |
92476d7f | 2729 | return nr; |
1da177e4 LT |
2730 | } |
2731 | ||
2aa3a7f8 AV |
2732 | /* |
2733 | * Create a kernel thread. | |
2734 | */ | |
2735 | pid_t kernel_thread(int (*fn)(void *), void *arg, unsigned long flags) | |
2736 | { | |
7f192e3c | 2737 | struct kernel_clone_args args = { |
3f2c788a CB |
2738 | .flags = ((lower_32_bits(flags) | CLONE_VM | |
2739 | CLONE_UNTRACED) & ~CSIGNAL), | |
2740 | .exit_signal = (lower_32_bits(flags) & CSIGNAL), | |
5bd2e97c EB |
2741 | .fn = fn, |
2742 | .fn_arg = arg, | |
343f4c49 EB |
2743 | .kthread = 1, |
2744 | }; | |
2745 | ||
2746 | return kernel_clone(&args); | |
2747 | } | |
2748 | ||
2749 | /* | |
2750 | * Create a user mode thread. | |
2751 | */ | |
2752 | pid_t user_mode_thread(int (*fn)(void *), void *arg, unsigned long flags) | |
2aa3a7f8 | 2753 | { |
7f192e3c | 2754 | struct kernel_clone_args args = { |
3f2c788a CB |
2755 | .flags = ((lower_32_bits(flags) | CLONE_VM | |
2756 | CLONE_UNTRACED) & ~CSIGNAL), | |
2757 | .exit_signal = (lower_32_bits(flags) & CSIGNAL), | |
5bd2e97c EB |
2758 | .fn = fn, |
2759 | .fn_arg = arg, | |
7f192e3c CB |
2760 | }; |
2761 | ||
cad6967a | 2762 | return kernel_clone(&args); |
2aa3a7f8 | 2763 | } |
2aa3a7f8 | 2764 | |
d2125043 AV |
2765 | #ifdef __ARCH_WANT_SYS_FORK |
2766 | SYSCALL_DEFINE0(fork) | |
2767 | { | |
2768 | #ifdef CONFIG_MMU | |
7f192e3c CB |
2769 | struct kernel_clone_args args = { |
2770 | .exit_signal = SIGCHLD, | |
2771 | }; | |
2772 | ||
cad6967a | 2773 | return kernel_clone(&args); |
d2125043 AV |
2774 | #else |
2775 | /* can not support in nommu mode */ | |
5d59e182 | 2776 | return -EINVAL; |
d2125043 AV |
2777 | #endif |
2778 | } | |
2779 | #endif | |
2780 | ||
2781 | #ifdef __ARCH_WANT_SYS_VFORK | |
2782 | SYSCALL_DEFINE0(vfork) | |
2783 | { | |
7f192e3c CB |
2784 | struct kernel_clone_args args = { |
2785 | .flags = CLONE_VFORK | CLONE_VM, | |
2786 | .exit_signal = SIGCHLD, | |
2787 | }; | |
2788 | ||
cad6967a | 2789 | return kernel_clone(&args); |
d2125043 AV |
2790 | } |
2791 | #endif | |
2792 | ||
2793 | #ifdef __ARCH_WANT_SYS_CLONE | |
2794 | #ifdef CONFIG_CLONE_BACKWARDS | |
2795 | SYSCALL_DEFINE5(clone, unsigned long, clone_flags, unsigned long, newsp, | |
2796 | int __user *, parent_tidptr, | |
3033f14a | 2797 | unsigned long, tls, |
d2125043 AV |
2798 | int __user *, child_tidptr) |
2799 | #elif defined(CONFIG_CLONE_BACKWARDS2) | |
2800 | SYSCALL_DEFINE5(clone, unsigned long, newsp, unsigned long, clone_flags, | |
2801 | int __user *, parent_tidptr, | |
2802 | int __user *, child_tidptr, | |
3033f14a | 2803 | unsigned long, tls) |
dfa9771a MS |
2804 | #elif defined(CONFIG_CLONE_BACKWARDS3) |
2805 | SYSCALL_DEFINE6(clone, unsigned long, clone_flags, unsigned long, newsp, | |
2806 | int, stack_size, | |
2807 | int __user *, parent_tidptr, | |
2808 | int __user *, child_tidptr, | |
3033f14a | 2809 | unsigned long, tls) |
d2125043 AV |
2810 | #else |
2811 | SYSCALL_DEFINE5(clone, unsigned long, clone_flags, unsigned long, newsp, | |
2812 | int __user *, parent_tidptr, | |
2813 | int __user *, child_tidptr, | |
3033f14a | 2814 | unsigned long, tls) |
d2125043 AV |
2815 | #endif |
2816 | { | |
7f192e3c | 2817 | struct kernel_clone_args args = { |
3f2c788a | 2818 | .flags = (lower_32_bits(clone_flags) & ~CSIGNAL), |
7f192e3c CB |
2819 | .pidfd = parent_tidptr, |
2820 | .child_tid = child_tidptr, | |
2821 | .parent_tid = parent_tidptr, | |
3f2c788a | 2822 | .exit_signal = (lower_32_bits(clone_flags) & CSIGNAL), |
7f192e3c CB |
2823 | .stack = newsp, |
2824 | .tls = tls, | |
2825 | }; | |
2826 | ||
cad6967a | 2827 | return kernel_clone(&args); |
7f192e3c | 2828 | } |
d68dbb0c | 2829 | #endif |
7f192e3c | 2830 | |
d68dbb0c | 2831 | #ifdef __ARCH_WANT_SYS_CLONE3 |
dd499f7a | 2832 | |
7f192e3c CB |
2833 | noinline static int copy_clone_args_from_user(struct kernel_clone_args *kargs, |
2834 | struct clone_args __user *uargs, | |
f14c234b | 2835 | size_t usize) |
7f192e3c | 2836 | { |
f14c234b | 2837 | int err; |
7f192e3c | 2838 | struct clone_args args; |
49cb2fc4 | 2839 | pid_t *kset_tid = kargs->set_tid; |
7f192e3c | 2840 | |
a966dcfe ES |
2841 | BUILD_BUG_ON(offsetofend(struct clone_args, tls) != |
2842 | CLONE_ARGS_SIZE_VER0); | |
2843 | BUILD_BUG_ON(offsetofend(struct clone_args, set_tid_size) != | |
2844 | CLONE_ARGS_SIZE_VER1); | |
2845 | BUILD_BUG_ON(offsetofend(struct clone_args, cgroup) != | |
2846 | CLONE_ARGS_SIZE_VER2); | |
2847 | BUILD_BUG_ON(sizeof(struct clone_args) != CLONE_ARGS_SIZE_VER2); | |
2848 | ||
f14c234b | 2849 | if (unlikely(usize > PAGE_SIZE)) |
7f192e3c | 2850 | return -E2BIG; |
f14c234b | 2851 | if (unlikely(usize < CLONE_ARGS_SIZE_VER0)) |
7f192e3c CB |
2852 | return -EINVAL; |
2853 | ||
f14c234b AS |
2854 | err = copy_struct_from_user(&args, sizeof(args), uargs, usize); |
2855 | if (err) | |
2856 | return err; | |
7f192e3c | 2857 | |
49cb2fc4 AR |
2858 | if (unlikely(args.set_tid_size > MAX_PID_NS_LEVEL)) |
2859 | return -EINVAL; | |
2860 | ||
2861 | if (unlikely(!args.set_tid && args.set_tid_size > 0)) | |
2862 | return -EINVAL; | |
2863 | ||
2864 | if (unlikely(args.set_tid && args.set_tid_size == 0)) | |
2865 | return -EINVAL; | |
2866 | ||
a0eb9abd ES |
2867 | /* |
2868 | * Verify that higher 32bits of exit_signal are unset and that | |
2869 | * it is a valid signal | |
2870 | */ | |
2871 | if (unlikely((args.exit_signal & ~((u64)CSIGNAL)) || | |
2872 | !valid_signal(args.exit_signal))) | |
2873 | return -EINVAL; | |
2874 | ||
62173872 ES |
2875 | if ((args.flags & CLONE_INTO_CGROUP) && |
2876 | (args.cgroup > INT_MAX || usize < CLONE_ARGS_SIZE_VER2)) | |
ef2c41cf CB |
2877 | return -EINVAL; |
2878 | ||
7f192e3c CB |
2879 | *kargs = (struct kernel_clone_args){ |
2880 | .flags = args.flags, | |
2881 | .pidfd = u64_to_user_ptr(args.pidfd), | |
2882 | .child_tid = u64_to_user_ptr(args.child_tid), | |
2883 | .parent_tid = u64_to_user_ptr(args.parent_tid), | |
2884 | .exit_signal = args.exit_signal, | |
2885 | .stack = args.stack, | |
2886 | .stack_size = args.stack_size, | |
2887 | .tls = args.tls, | |
49cb2fc4 | 2888 | .set_tid_size = args.set_tid_size, |
ef2c41cf | 2889 | .cgroup = args.cgroup, |
7f192e3c CB |
2890 | }; |
2891 | ||
49cb2fc4 AR |
2892 | if (args.set_tid && |
2893 | copy_from_user(kset_tid, u64_to_user_ptr(args.set_tid), | |
2894 | (kargs->set_tid_size * sizeof(pid_t)))) | |
2895 | return -EFAULT; | |
2896 | ||
2897 | kargs->set_tid = kset_tid; | |
2898 | ||
7f192e3c CB |
2899 | return 0; |
2900 | } | |
2901 | ||
fa729c4d CB |
2902 | /** |
2903 | * clone3_stack_valid - check and prepare stack | |
2904 | * @kargs: kernel clone args | |
2905 | * | |
2906 | * Verify that the stack arguments userspace gave us are sane. | |
2907 | * In addition, set the stack direction for userspace since it's easy for us to | |
2908 | * determine. | |
2909 | */ | |
2910 | static inline bool clone3_stack_valid(struct kernel_clone_args *kargs) | |
2911 | { | |
2912 | if (kargs->stack == 0) { | |
2913 | if (kargs->stack_size > 0) | |
2914 | return false; | |
2915 | } else { | |
2916 | if (kargs->stack_size == 0) | |
2917 | return false; | |
2918 | ||
2919 | if (!access_ok((void __user *)kargs->stack, kargs->stack_size)) | |
2920 | return false; | |
2921 | ||
2922 | #if !defined(CONFIG_STACK_GROWSUP) && !defined(CONFIG_IA64) | |
2923 | kargs->stack += kargs->stack_size; | |
2924 | #endif | |
2925 | } | |
2926 | ||
2927 | return true; | |
2928 | } | |
2929 | ||
2930 | static bool clone3_args_valid(struct kernel_clone_args *kargs) | |
7f192e3c | 2931 | { |
b612e5df | 2932 | /* Verify that no unknown flags are passed along. */ |
ef2c41cf CB |
2933 | if (kargs->flags & |
2934 | ~(CLONE_LEGACY_FLAGS | CLONE_CLEAR_SIGHAND | CLONE_INTO_CGROUP)) | |
7f192e3c CB |
2935 | return false; |
2936 | ||
2937 | /* | |
a8ca6b13 XC |
2938 | * - make the CLONE_DETACHED bit reusable for clone3 |
2939 | * - make the CSIGNAL bits reusable for clone3 | |
7f192e3c CB |
2940 | */ |
2941 | if (kargs->flags & (CLONE_DETACHED | CSIGNAL)) | |
2942 | return false; | |
2943 | ||
b612e5df CB |
2944 | if ((kargs->flags & (CLONE_SIGHAND | CLONE_CLEAR_SIGHAND)) == |
2945 | (CLONE_SIGHAND | CLONE_CLEAR_SIGHAND)) | |
2946 | return false; | |
2947 | ||
7f192e3c CB |
2948 | if ((kargs->flags & (CLONE_THREAD | CLONE_PARENT)) && |
2949 | kargs->exit_signal) | |
2950 | return false; | |
2951 | ||
fa729c4d CB |
2952 | if (!clone3_stack_valid(kargs)) |
2953 | return false; | |
2954 | ||
7f192e3c CB |
2955 | return true; |
2956 | } | |
2957 | ||
501bd016 CB |
2958 | /** |
2959 | * clone3 - create a new process with specific properties | |
2960 | * @uargs: argument structure | |
2961 | * @size: size of @uargs | |
2962 | * | |
2963 | * clone3() is the extensible successor to clone()/clone2(). | |
2964 | * It takes a struct as argument that is versioned by its size. | |
2965 | * | |
2966 | * Return: On success, a positive PID for the child process. | |
2967 | * On error, a negative errno number. | |
2968 | */ | |
7f192e3c CB |
2969 | SYSCALL_DEFINE2(clone3, struct clone_args __user *, uargs, size_t, size) |
2970 | { | |
2971 | int err; | |
2972 | ||
2973 | struct kernel_clone_args kargs; | |
49cb2fc4 AR |
2974 | pid_t set_tid[MAX_PID_NS_LEVEL]; |
2975 | ||
2976 | kargs.set_tid = set_tid; | |
7f192e3c CB |
2977 | |
2978 | err = copy_clone_args_from_user(&kargs, uargs, size); | |
2979 | if (err) | |
2980 | return err; | |
2981 | ||
2982 | if (!clone3_args_valid(&kargs)) | |
2983 | return -EINVAL; | |
2984 | ||
cad6967a | 2985 | return kernel_clone(&kargs); |
d2125043 AV |
2986 | } |
2987 | #endif | |
2988 | ||
0f1b92cb ON |
2989 | void walk_process_tree(struct task_struct *top, proc_visitor visitor, void *data) |
2990 | { | |
2991 | struct task_struct *leader, *parent, *child; | |
2992 | int res; | |
2993 | ||
2994 | read_lock(&tasklist_lock); | |
2995 | leader = top = top->group_leader; | |
2996 | down: | |
2997 | for_each_thread(leader, parent) { | |
2998 | list_for_each_entry(child, &parent->children, sibling) { | |
2999 | res = visitor(child, data); | |
3000 | if (res) { | |
3001 | if (res < 0) | |
3002 | goto out; | |
3003 | leader = child; | |
3004 | goto down; | |
3005 | } | |
3006 | up: | |
3007 | ; | |
3008 | } | |
3009 | } | |
3010 | ||
3011 | if (leader != top) { | |
3012 | child = leader; | |
3013 | parent = child->real_parent; | |
3014 | leader = parent->group_leader; | |
3015 | goto up; | |
3016 | } | |
3017 | out: | |
3018 | read_unlock(&tasklist_lock); | |
3019 | } | |
3020 | ||
5fd63b30 RT |
3021 | #ifndef ARCH_MIN_MMSTRUCT_ALIGN |
3022 | #define ARCH_MIN_MMSTRUCT_ALIGN 0 | |
3023 | #endif | |
3024 | ||
51cc5068 | 3025 | static void sighand_ctor(void *data) |
aa1757f9 ON |
3026 | { |
3027 | struct sighand_struct *sighand = data; | |
3028 | ||
a35afb83 | 3029 | spin_lock_init(&sighand->siglock); |
b8fceee1 | 3030 | init_waitqueue_head(&sighand->signalfd_wqh); |
aa1757f9 ON |
3031 | } |
3032 | ||
1da177e4 LT |
3033 | void __init proc_caches_init(void) |
3034 | { | |
c1a2f7f0 RR |
3035 | unsigned int mm_size; |
3036 | ||
1da177e4 LT |
3037 | sighand_cachep = kmem_cache_create("sighand_cache", |
3038 | sizeof(struct sighand_struct), 0, | |
5f0d5a3a | 3039 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_TYPESAFE_BY_RCU| |
75f296d9 | 3040 | SLAB_ACCOUNT, sighand_ctor); |
1da177e4 LT |
3041 | signal_cachep = kmem_cache_create("signal_cache", |
3042 | sizeof(struct signal_struct), 0, | |
75f296d9 | 3043 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_ACCOUNT, |
5d097056 | 3044 | NULL); |
20c2df83 | 3045 | files_cachep = kmem_cache_create("files_cache", |
1da177e4 | 3046 | sizeof(struct files_struct), 0, |
75f296d9 | 3047 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_ACCOUNT, |
5d097056 | 3048 | NULL); |
20c2df83 | 3049 | fs_cachep = kmem_cache_create("fs_cache", |
1da177e4 | 3050 | sizeof(struct fs_struct), 0, |
75f296d9 | 3051 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_ACCOUNT, |
5d097056 | 3052 | NULL); |
c1a2f7f0 | 3053 | |
6345d24d | 3054 | /* |
c1a2f7f0 RR |
3055 | * The mm_cpumask is located at the end of mm_struct, and is |
3056 | * dynamically sized based on the maximum CPU number this system | |
3057 | * can have, taking hotplug into account (nr_cpu_ids). | |
6345d24d | 3058 | */ |
c1a2f7f0 RR |
3059 | mm_size = sizeof(struct mm_struct) + cpumask_size(); |
3060 | ||
07dcd7fe | 3061 | mm_cachep = kmem_cache_create_usercopy("mm_struct", |
c1a2f7f0 | 3062 | mm_size, ARCH_MIN_MMSTRUCT_ALIGN, |
75f296d9 | 3063 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_ACCOUNT, |
07dcd7fe DW |
3064 | offsetof(struct mm_struct, saved_auxv), |
3065 | sizeof_field(struct mm_struct, saved_auxv), | |
5d097056 VD |
3066 | NULL); |
3067 | vm_area_cachep = KMEM_CACHE(vm_area_struct, SLAB_PANIC|SLAB_ACCOUNT); | |
8feae131 | 3068 | mmap_init(); |
66577193 | 3069 | nsproxy_cache_init(); |
1da177e4 | 3070 | } |
cf2e340f | 3071 | |
cf2e340f | 3072 | /* |
9bfb23fc | 3073 | * Check constraints on flags passed to the unshare system call. |
cf2e340f | 3074 | */ |
9bfb23fc | 3075 | static int check_unshare_flags(unsigned long unshare_flags) |
cf2e340f | 3076 | { |
9bfb23fc ON |
3077 | if (unshare_flags & ~(CLONE_THREAD|CLONE_FS|CLONE_NEWNS|CLONE_SIGHAND| |
3078 | CLONE_VM|CLONE_FILES|CLONE_SYSVSEM| | |
50804fe3 | 3079 | CLONE_NEWUTS|CLONE_NEWIPC|CLONE_NEWNET| |
769071ac AV |
3080 | CLONE_NEWUSER|CLONE_NEWPID|CLONE_NEWCGROUP| |
3081 | CLONE_NEWTIME)) | |
9bfb23fc | 3082 | return -EINVAL; |
cf2e340f | 3083 | /* |
12c641ab EB |
3084 | * Not implemented, but pretend it works if there is nothing |
3085 | * to unshare. Note that unsharing the address space or the | |
3086 | * signal handlers also need to unshare the signal queues (aka | |
3087 | * CLONE_THREAD). | |
cf2e340f | 3088 | */ |
9bfb23fc | 3089 | if (unshare_flags & (CLONE_THREAD | CLONE_SIGHAND | CLONE_VM)) { |
12c641ab EB |
3090 | if (!thread_group_empty(current)) |
3091 | return -EINVAL; | |
3092 | } | |
3093 | if (unshare_flags & (CLONE_SIGHAND | CLONE_VM)) { | |
d036bda7 | 3094 | if (refcount_read(¤t->sighand->count) > 1) |
12c641ab EB |
3095 | return -EINVAL; |
3096 | } | |
3097 | if (unshare_flags & CLONE_VM) { | |
3098 | if (!current_is_single_threaded()) | |
9bfb23fc ON |
3099 | return -EINVAL; |
3100 | } | |
cf2e340f JD |
3101 | |
3102 | return 0; | |
3103 | } | |
3104 | ||
3105 | /* | |
99d1419d | 3106 | * Unshare the filesystem structure if it is being shared |
cf2e340f JD |
3107 | */ |
3108 | static int unshare_fs(unsigned long unshare_flags, struct fs_struct **new_fsp) | |
3109 | { | |
3110 | struct fs_struct *fs = current->fs; | |
3111 | ||
498052bb AV |
3112 | if (!(unshare_flags & CLONE_FS) || !fs) |
3113 | return 0; | |
3114 | ||
3115 | /* don't need lock here; in the worst case we'll do useless copy */ | |
3116 | if (fs->users == 1) | |
3117 | return 0; | |
3118 | ||
3119 | *new_fsp = copy_fs_struct(fs); | |
3120 | if (!*new_fsp) | |
3121 | return -ENOMEM; | |
cf2e340f JD |
3122 | |
3123 | return 0; | |
3124 | } | |
3125 | ||
cf2e340f | 3126 | /* |
a016f338 | 3127 | * Unshare file descriptor table if it is being shared |
cf2e340f | 3128 | */ |
60997c3d CB |
3129 | int unshare_fd(unsigned long unshare_flags, unsigned int max_fds, |
3130 | struct files_struct **new_fdp) | |
cf2e340f JD |
3131 | { |
3132 | struct files_struct *fd = current->files; | |
a016f338 | 3133 | int error = 0; |
cf2e340f JD |
3134 | |
3135 | if ((unshare_flags & CLONE_FILES) && | |
a016f338 | 3136 | (fd && atomic_read(&fd->count) > 1)) { |
60997c3d | 3137 | *new_fdp = dup_fd(fd, max_fds, &error); |
a016f338 JD |
3138 | if (!*new_fdp) |
3139 | return error; | |
3140 | } | |
cf2e340f JD |
3141 | |
3142 | return 0; | |
3143 | } | |
3144 | ||
cf2e340f JD |
3145 | /* |
3146 | * unshare allows a process to 'unshare' part of the process | |
3147 | * context which was originally shared using clone. copy_* | |
cad6967a | 3148 | * functions used by kernel_clone() cannot be used here directly |
cf2e340f JD |
3149 | * because they modify an inactive task_struct that is being |
3150 | * constructed. Here we are modifying the current, active, | |
3151 | * task_struct. | |
3152 | */ | |
9b32105e | 3153 | int ksys_unshare(unsigned long unshare_flags) |
cf2e340f | 3154 | { |
cf2e340f | 3155 | struct fs_struct *fs, *new_fs = NULL; |
ba1f70dd | 3156 | struct files_struct *new_fd = NULL; |
b2e0d987 | 3157 | struct cred *new_cred = NULL; |
cf7b708c | 3158 | struct nsproxy *new_nsproxy = NULL; |
9edff4ab | 3159 | int do_sysvsem = 0; |
9bfb23fc | 3160 | int err; |
cf2e340f | 3161 | |
b2e0d987 | 3162 | /* |
faf00da5 EB |
3163 | * If unsharing a user namespace must also unshare the thread group |
3164 | * and unshare the filesystem root and working directories. | |
b2e0d987 EB |
3165 | */ |
3166 | if (unshare_flags & CLONE_NEWUSER) | |
e66eded8 | 3167 | unshare_flags |= CLONE_THREAD | CLONE_FS; |
50804fe3 EB |
3168 | /* |
3169 | * If unsharing vm, must also unshare signal handlers. | |
3170 | */ | |
3171 | if (unshare_flags & CLONE_VM) | |
3172 | unshare_flags |= CLONE_SIGHAND; | |
12c641ab EB |
3173 | /* |
3174 | * If unsharing a signal handlers, must also unshare the signal queues. | |
3175 | */ | |
3176 | if (unshare_flags & CLONE_SIGHAND) | |
3177 | unshare_flags |= CLONE_THREAD; | |
9bfb23fc ON |
3178 | /* |
3179 | * If unsharing namespace, must also unshare filesystem information. | |
3180 | */ | |
3181 | if (unshare_flags & CLONE_NEWNS) | |
3182 | unshare_flags |= CLONE_FS; | |
50804fe3 EB |
3183 | |
3184 | err = check_unshare_flags(unshare_flags); | |
3185 | if (err) | |
3186 | goto bad_unshare_out; | |
6013f67f MS |
3187 | /* |
3188 | * CLONE_NEWIPC must also detach from the undolist: after switching | |
3189 | * to a new ipc namespace, the semaphore arrays from the old | |
3190 | * namespace are unreachable. | |
3191 | */ | |
3192 | if (unshare_flags & (CLONE_NEWIPC|CLONE_SYSVSEM)) | |
9edff4ab | 3193 | do_sysvsem = 1; |
fb0a685c DRO |
3194 | err = unshare_fs(unshare_flags, &new_fs); |
3195 | if (err) | |
9bfb23fc | 3196 | goto bad_unshare_out; |
60997c3d | 3197 | err = unshare_fd(unshare_flags, NR_OPEN_MAX, &new_fd); |
fb0a685c | 3198 | if (err) |
9bfb23fc | 3199 | goto bad_unshare_cleanup_fs; |
b2e0d987 | 3200 | err = unshare_userns(unshare_flags, &new_cred); |
fb0a685c | 3201 | if (err) |
9edff4ab | 3202 | goto bad_unshare_cleanup_fd; |
b2e0d987 EB |
3203 | err = unshare_nsproxy_namespaces(unshare_flags, &new_nsproxy, |
3204 | new_cred, new_fs); | |
3205 | if (err) | |
3206 | goto bad_unshare_cleanup_cred; | |
c0b2fc31 | 3207 | |
905ae01c AG |
3208 | if (new_cred) { |
3209 | err = set_cred_ucounts(new_cred); | |
3210 | if (err) | |
3211 | goto bad_unshare_cleanup_cred; | |
3212 | } | |
3213 | ||
b2e0d987 | 3214 | if (new_fs || new_fd || do_sysvsem || new_cred || new_nsproxy) { |
9edff4ab MS |
3215 | if (do_sysvsem) { |
3216 | /* | |
3217 | * CLONE_SYSVSEM is equivalent to sys_exit(). | |
3218 | */ | |
3219 | exit_sem(current); | |
3220 | } | |
ab602f79 JM |
3221 | if (unshare_flags & CLONE_NEWIPC) { |
3222 | /* Orphan segments in old ns (see sem above). */ | |
3223 | exit_shm(current); | |
3224 | shm_init_task(current); | |
3225 | } | |
ab516013 | 3226 | |
6f977e6b | 3227 | if (new_nsproxy) |
cf7b708c | 3228 | switch_task_namespaces(current, new_nsproxy); |
cf2e340f | 3229 | |
cf7b708c PE |
3230 | task_lock(current); |
3231 | ||
cf2e340f JD |
3232 | if (new_fs) { |
3233 | fs = current->fs; | |
2a4419b5 | 3234 | spin_lock(&fs->lock); |
cf2e340f | 3235 | current->fs = new_fs; |
498052bb AV |
3236 | if (--fs->users) |
3237 | new_fs = NULL; | |
3238 | else | |
3239 | new_fs = fs; | |
2a4419b5 | 3240 | spin_unlock(&fs->lock); |
cf2e340f JD |
3241 | } |
3242 | ||
ba1f70dd RX |
3243 | if (new_fd) |
3244 | swap(current->files, new_fd); | |
cf2e340f JD |
3245 | |
3246 | task_unlock(current); | |
b2e0d987 EB |
3247 | |
3248 | if (new_cred) { | |
3249 | /* Install the new user namespace */ | |
3250 | commit_creds(new_cred); | |
3251 | new_cred = NULL; | |
3252 | } | |
cf2e340f JD |
3253 | } |
3254 | ||
e4222673 HB |
3255 | perf_event_namespaces(current); |
3256 | ||
b2e0d987 EB |
3257 | bad_unshare_cleanup_cred: |
3258 | if (new_cred) | |
3259 | put_cred(new_cred); | |
cf2e340f JD |
3260 | bad_unshare_cleanup_fd: |
3261 | if (new_fd) | |
3262 | put_files_struct(new_fd); | |
3263 | ||
cf2e340f JD |
3264 | bad_unshare_cleanup_fs: |
3265 | if (new_fs) | |
498052bb | 3266 | free_fs_struct(new_fs); |
cf2e340f | 3267 | |
cf2e340f JD |
3268 | bad_unshare_out: |
3269 | return err; | |
3270 | } | |
3b125388 | 3271 | |
9b32105e DB |
3272 | SYSCALL_DEFINE1(unshare, unsigned long, unshare_flags) |
3273 | { | |
3274 | return ksys_unshare(unshare_flags); | |
3275 | } | |
3276 | ||
3b125388 AV |
3277 | /* |
3278 | * Helper to unshare the files of the current task. | |
3279 | * We don't want to expose copy_files internals to | |
3280 | * the exec layer of the kernel. | |
3281 | */ | |
3282 | ||
1f702603 | 3283 | int unshare_files(void) |
3b125388 AV |
3284 | { |
3285 | struct task_struct *task = current; | |
1f702603 | 3286 | struct files_struct *old, *copy = NULL; |
3b125388 AV |
3287 | int error; |
3288 | ||
60997c3d | 3289 | error = unshare_fd(CLONE_FILES, NR_OPEN_MAX, ©); |
1f702603 | 3290 | if (error || !copy) |
3b125388 | 3291 | return error; |
1f702603 EB |
3292 | |
3293 | old = task->files; | |
3b125388 AV |
3294 | task_lock(task); |
3295 | task->files = copy; | |
3296 | task_unlock(task); | |
1f702603 | 3297 | put_files_struct(old); |
3b125388 AV |
3298 | return 0; |
3299 | } | |
16db3d3f HS |
3300 | |
3301 | int sysctl_max_threads(struct ctl_table *table, int write, | |
b0daa2c7 | 3302 | void *buffer, size_t *lenp, loff_t *ppos) |
16db3d3f HS |
3303 | { |
3304 | struct ctl_table t; | |
3305 | int ret; | |
3306 | int threads = max_threads; | |
b0f53dbc | 3307 | int min = 1; |
16db3d3f HS |
3308 | int max = MAX_THREADS; |
3309 | ||
3310 | t = *table; | |
3311 | t.data = &threads; | |
3312 | t.extra1 = &min; | |
3313 | t.extra2 = &max; | |
3314 | ||
3315 | ret = proc_dointvec_minmax(&t, write, buffer, lenp, ppos); | |
3316 | if (ret || !write) | |
3317 | return ret; | |
3318 | ||
b0f53dbc | 3319 | max_threads = threads; |
16db3d3f HS |
3320 | |
3321 | return 0; | |
3322 | } |