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 | */ | |
06e78b61 | 475 | data_race(memcpy(new, orig, sizeof(*new))); |
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); |
3b9dbd5e LH |
588 | VMA_ITERATOR(old_vmi, oldmm, 0); |
589 | VMA_ITERATOR(vmi, mm, 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 | |
3b9dbd5e | 616 | retval = vma_iter_bulk_alloc(&vmi, oldmm->map_count); |
c9dbe82c LH |
617 | if (retval) |
618 | goto out; | |
619 | ||
3b9dbd5e | 620 | for_each_vma(old_vmi, mpnt) { |
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; | |
e430a95a | 662 | vm_flags_clear(tmp, VM_LOCKED_MASK); |
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 | 685 | /* Link the vma into the MT */ |
3b9dbd5e LH |
686 | if (vma_iter_bulk_store(&vmi, tmp)) |
687 | goto fail_nomem_vmi_store; | |
d70f2a14 AM |
688 | |
689 | mm->map_count++; | |
690 | if (!(tmp->vm_flags & VM_WIPEONFORK)) | |
c78f4636 | 691 | retval = copy_page_range(tmp, mpnt); |
d70f2a14 AM |
692 | |
693 | if (tmp->vm_ops && tmp->vm_ops->open) | |
694 | tmp->vm_ops->open(tmp); | |
695 | ||
696 | if (retval) | |
d4af56c5 | 697 | goto loop_out; |
d70f2a14 AM |
698 | } |
699 | /* a new mm has just been created */ | |
1ed0cc5a | 700 | retval = arch_dup_mmap(oldmm, mm); |
d4af56c5 | 701 | loop_out: |
3b9dbd5e | 702 | vma_iter_free(&vmi); |
d70f2a14 | 703 | out: |
d8ed45c5 | 704 | mmap_write_unlock(mm); |
d70f2a14 | 705 | flush_tlb_mm(oldmm); |
d8ed45c5 | 706 | mmap_write_unlock(oldmm); |
d70f2a14 AM |
707 | dup_userfaultfd_complete(&uf); |
708 | fail_uprobe_end: | |
709 | uprobe_end_dup_mmap(); | |
710 | return retval; | |
c9dbe82c | 711 | |
3b9dbd5e | 712 | fail_nomem_vmi_store: |
c9dbe82c | 713 | unlink_anon_vmas(tmp); |
d70f2a14 AM |
714 | fail_nomem_anon_vma_fork: |
715 | mpol_put(vma_policy(tmp)); | |
716 | fail_nomem_policy: | |
3928d4f5 | 717 | vm_area_free(tmp); |
d70f2a14 AM |
718 | fail_nomem: |
719 | retval = -ENOMEM; | |
720 | vm_unacct_memory(charge); | |
d4af56c5 | 721 | goto loop_out; |
d70f2a14 AM |
722 | } |
723 | ||
724 | static inline int mm_alloc_pgd(struct mm_struct *mm) | |
725 | { | |
726 | mm->pgd = pgd_alloc(mm); | |
727 | if (unlikely(!mm->pgd)) | |
728 | return -ENOMEM; | |
729 | return 0; | |
730 | } | |
731 | ||
732 | static inline void mm_free_pgd(struct mm_struct *mm) | |
733 | { | |
734 | pgd_free(mm, mm->pgd); | |
735 | } | |
736 | #else | |
737 | static int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm) | |
738 | { | |
d8ed45c5 | 739 | mmap_write_lock(oldmm); |
fe69d560 | 740 | dup_mm_exe_file(mm, oldmm); |
d8ed45c5 | 741 | mmap_write_unlock(oldmm); |
d70f2a14 AM |
742 | return 0; |
743 | } | |
744 | #define mm_alloc_pgd(mm) (0) | |
745 | #define mm_free_pgd(mm) | |
746 | #endif /* CONFIG_MMU */ | |
747 | ||
748 | static void check_mm(struct mm_struct *mm) | |
749 | { | |
750 | int i; | |
751 | ||
8495f7e6 SPP |
752 | BUILD_BUG_ON_MSG(ARRAY_SIZE(resident_page_types) != NR_MM_COUNTERS, |
753 | "Please make sure 'struct resident_page_types[]' is updated as well"); | |
754 | ||
d70f2a14 | 755 | for (i = 0; i < NR_MM_COUNTERS; i++) { |
f1a79412 | 756 | long x = percpu_counter_sum(&mm->rss_stat[i]); |
d70f2a14 AM |
757 | |
758 | if (unlikely(x)) | |
8495f7e6 SPP |
759 | pr_alert("BUG: Bad rss-counter state mm:%p type:%s val:%ld\n", |
760 | mm, resident_page_types[i], x); | |
d70f2a14 AM |
761 | } |
762 | ||
763 | if (mm_pgtables_bytes(mm)) | |
764 | pr_alert("BUG: non-zero pgtables_bytes on freeing mm: %ld\n", | |
765 | mm_pgtables_bytes(mm)); | |
766 | ||
767 | #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS | |
768 | VM_BUG_ON_MM(mm->pmd_huge_pte, mm); | |
769 | #endif | |
770 | } | |
771 | ||
772 | #define allocate_mm() (kmem_cache_alloc(mm_cachep, GFP_KERNEL)) | |
773 | #define free_mm(mm) (kmem_cache_free(mm_cachep, (mm))) | |
774 | ||
2655421a NP |
775 | static void do_check_lazy_tlb(void *arg) |
776 | { | |
777 | struct mm_struct *mm = arg; | |
778 | ||
779 | WARN_ON_ONCE(current->active_mm == mm); | |
780 | } | |
781 | ||
782 | static void do_shoot_lazy_tlb(void *arg) | |
783 | { | |
784 | struct mm_struct *mm = arg; | |
785 | ||
786 | if (current->active_mm == mm) { | |
787 | WARN_ON_ONCE(current->mm); | |
788 | current->active_mm = &init_mm; | |
789 | switch_mm(mm, &init_mm, current); | |
790 | } | |
791 | } | |
792 | ||
793 | static void cleanup_lazy_tlbs(struct mm_struct *mm) | |
794 | { | |
795 | if (!IS_ENABLED(CONFIG_MMU_LAZY_TLB_SHOOTDOWN)) { | |
796 | /* | |
797 | * In this case, lazy tlb mms are refounted and would not reach | |
798 | * __mmdrop until all CPUs have switched away and mmdrop()ed. | |
799 | */ | |
800 | return; | |
801 | } | |
802 | ||
803 | /* | |
804 | * Lazy mm shootdown does not refcount "lazy tlb mm" usage, rather it | |
805 | * requires lazy mm users to switch to another mm when the refcount | |
806 | * drops to zero, before the mm is freed. This requires IPIs here to | |
807 | * switch kernel threads to init_mm. | |
808 | * | |
809 | * archs that use IPIs to flush TLBs can piggy-back that lazy tlb mm | |
810 | * switch with the final userspace teardown TLB flush which leaves the | |
811 | * mm lazy on this CPU but no others, reducing the need for additional | |
812 | * IPIs here. There are cases where a final IPI is still required here, | |
813 | * such as the final mmdrop being performed on a different CPU than the | |
814 | * one exiting, or kernel threads using the mm when userspace exits. | |
815 | * | |
816 | * IPI overheads have not found to be expensive, but they could be | |
817 | * reduced in a number of possible ways, for example (roughly | |
818 | * increasing order of complexity): | |
819 | * - The last lazy reference created by exit_mm() could instead switch | |
820 | * to init_mm, however it's probable this will run on the same CPU | |
821 | * immediately afterwards, so this may not reduce IPIs much. | |
822 | * - A batch of mms requiring IPIs could be gathered and freed at once. | |
823 | * - CPUs store active_mm where it can be remotely checked without a | |
824 | * lock, to filter out false-positives in the cpumask. | |
825 | * - After mm_users or mm_count reaches zero, switching away from the | |
826 | * mm could clear mm_cpumask to reduce some IPIs, perhaps together | |
827 | * with some batching or delaying of the final IPIs. | |
828 | * - A delayed freeing and RCU-like quiescing sequence based on mm | |
829 | * switching to avoid IPIs completely. | |
830 | */ | |
831 | on_each_cpu_mask(mm_cpumask(mm), do_shoot_lazy_tlb, (void *)mm, 1); | |
832 | if (IS_ENABLED(CONFIG_DEBUG_VM_SHOOT_LAZIES)) | |
833 | on_each_cpu(do_check_lazy_tlb, (void *)mm, 1); | |
834 | } | |
835 | ||
d70f2a14 AM |
836 | /* |
837 | * Called when the last reference to the mm | |
838 | * is dropped: either by a lazy thread or by | |
839 | * mmput. Free the page directory and the mm. | |
840 | */ | |
d34bc48f | 841 | void __mmdrop(struct mm_struct *mm) |
d70f2a14 | 842 | { |
f1a79412 SB |
843 | int i; |
844 | ||
d70f2a14 | 845 | BUG_ON(mm == &init_mm); |
3eda69c9 | 846 | WARN_ON_ONCE(mm == current->mm); |
2655421a NP |
847 | |
848 | /* Ensure no CPUs are using this as their lazy tlb mm */ | |
849 | cleanup_lazy_tlbs(mm); | |
850 | ||
3eda69c9 | 851 | WARN_ON_ONCE(mm == current->active_mm); |
d70f2a14 AM |
852 | mm_free_pgd(mm); |
853 | destroy_context(mm); | |
984cfe4e | 854 | mmu_notifier_subscriptions_destroy(mm); |
d70f2a14 AM |
855 | check_mm(mm); |
856 | put_user_ns(mm->user_ns); | |
2667ed10 | 857 | mm_pasid_drop(mm); |
f1a79412 SB |
858 | |
859 | for (i = 0; i < NR_MM_COUNTERS; i++) | |
860 | percpu_counter_destroy(&mm->rss_stat[i]); | |
d70f2a14 AM |
861 | free_mm(mm); |
862 | } | |
d34bc48f | 863 | EXPORT_SYMBOL_GPL(__mmdrop); |
d70f2a14 AM |
864 | |
865 | static void mmdrop_async_fn(struct work_struct *work) | |
866 | { | |
867 | struct mm_struct *mm; | |
868 | ||
869 | mm = container_of(work, struct mm_struct, async_put_work); | |
870 | __mmdrop(mm); | |
871 | } | |
872 | ||
873 | static void mmdrop_async(struct mm_struct *mm) | |
874 | { | |
875 | if (unlikely(atomic_dec_and_test(&mm->mm_count))) { | |
876 | INIT_WORK(&mm->async_put_work, mmdrop_async_fn); | |
877 | schedule_work(&mm->async_put_work); | |
878 | } | |
879 | } | |
880 | ||
ea6d290c ON |
881 | static inline void free_signal_struct(struct signal_struct *sig) |
882 | { | |
97101eb4 | 883 | taskstats_tgid_free(sig); |
1c5354de | 884 | sched_autogroup_exit(sig); |
7283094e MH |
885 | /* |
886 | * __mmdrop is not safe to call from softirq context on x86 due to | |
887 | * pgd_dtor so postpone it to the async context | |
888 | */ | |
26db62f1 | 889 | if (sig->oom_mm) |
7283094e | 890 | mmdrop_async(sig->oom_mm); |
ea6d290c ON |
891 | kmem_cache_free(signal_cachep, sig); |
892 | } | |
893 | ||
894 | static inline void put_signal_struct(struct signal_struct *sig) | |
895 | { | |
60d4de3f | 896 | if (refcount_dec_and_test(&sig->sigcnt)) |
ea6d290c ON |
897 | free_signal_struct(sig); |
898 | } | |
899 | ||
158d9ebd | 900 | void __put_task_struct(struct task_struct *tsk) |
1da177e4 | 901 | { |
270f722d | 902 | WARN_ON(!tsk->exit_state); |
ec1d2819 | 903 | WARN_ON(refcount_read(&tsk->usage)); |
1da177e4 LT |
904 | WARN_ON(tsk == current); |
905 | ||
0f212204 | 906 | io_uring_free(tsk); |
2e91fa7f | 907 | cgroup_free(tsk); |
16d51a59 | 908 | task_numa_free(tsk, true); |
1a2a4d06 | 909 | security_task_free(tsk); |
a10787e6 | 910 | bpf_task_storage_free(tsk); |
e0e81739 | 911 | exit_creds(tsk); |
35df17c5 | 912 | delayacct_tsk_free(tsk); |
ea6d290c | 913 | put_signal_struct(tsk->signal); |
6e33cad0 | 914 | sched_core_free(tsk); |
2873cd31 | 915 | free_task(tsk); |
1da177e4 | 916 | } |
77c100c8 | 917 | EXPORT_SYMBOL_GPL(__put_task_struct); |
1da177e4 | 918 | |
6c0a9fa6 | 919 | void __init __weak arch_task_cache_init(void) { } |
61c4628b | 920 | |
ff691f6e HS |
921 | /* |
922 | * set_max_threads | |
923 | */ | |
16db3d3f | 924 | static void set_max_threads(unsigned int max_threads_suggested) |
ff691f6e | 925 | { |
ac1b398d | 926 | u64 threads; |
ca79b0c2 | 927 | unsigned long nr_pages = totalram_pages(); |
ff691f6e HS |
928 | |
929 | /* | |
ac1b398d HS |
930 | * The number of threads shall be limited such that the thread |
931 | * structures may only consume a small part of the available memory. | |
ff691f6e | 932 | */ |
3d6357de | 933 | if (fls64(nr_pages) + fls64(PAGE_SIZE) > 64) |
ac1b398d HS |
934 | threads = MAX_THREADS; |
935 | else | |
3d6357de | 936 | threads = div64_u64((u64) nr_pages * (u64) PAGE_SIZE, |
ac1b398d HS |
937 | (u64) THREAD_SIZE * 8UL); |
938 | ||
16db3d3f HS |
939 | if (threads > max_threads_suggested) |
940 | threads = max_threads_suggested; | |
941 | ||
ac1b398d | 942 | max_threads = clamp_t(u64, threads, MIN_THREADS, MAX_THREADS); |
ff691f6e HS |
943 | } |
944 | ||
5aaeb5c0 IM |
945 | #ifdef CONFIG_ARCH_WANTS_DYNAMIC_TASK_STRUCT |
946 | /* Initialized by the architecture: */ | |
947 | int arch_task_struct_size __read_mostly; | |
948 | #endif | |
0c8c0f03 | 949 | |
4189ff23 | 950 | #ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR |
5905429a KC |
951 | static void task_struct_whitelist(unsigned long *offset, unsigned long *size) |
952 | { | |
953 | /* Fetch thread_struct whitelist for the architecture. */ | |
954 | arch_thread_struct_whitelist(offset, size); | |
955 | ||
956 | /* | |
957 | * Handle zero-sized whitelist or empty thread_struct, otherwise | |
958 | * adjust offset to position of thread_struct in task_struct. | |
959 | */ | |
960 | if (unlikely(*size == 0)) | |
961 | *offset = 0; | |
962 | else | |
963 | *offset += offsetof(struct task_struct, thread); | |
964 | } | |
4189ff23 | 965 | #endif /* CONFIG_ARCH_TASK_STRUCT_ALLOCATOR */ |
5905429a | 966 | |
ff691f6e | 967 | void __init fork_init(void) |
1da177e4 | 968 | { |
25f9c081 | 969 | int i; |
f5e10287 | 970 | #ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR |
1da177e4 | 971 | #ifndef ARCH_MIN_TASKALIGN |
e274795e | 972 | #define ARCH_MIN_TASKALIGN 0 |
1da177e4 | 973 | #endif |
95cb64c1 | 974 | int align = max_t(int, L1_CACHE_BYTES, ARCH_MIN_TASKALIGN); |
5905429a | 975 | unsigned long useroffset, usersize; |
e274795e | 976 | |
1da177e4 | 977 | /* create a slab on which task_structs can be allocated */ |
5905429a KC |
978 | task_struct_whitelist(&useroffset, &usersize); |
979 | task_struct_cachep = kmem_cache_create_usercopy("task_struct", | |
e274795e | 980 | arch_task_struct_size, align, |
5905429a KC |
981 | SLAB_PANIC|SLAB_ACCOUNT, |
982 | useroffset, usersize, NULL); | |
1da177e4 LT |
983 | #endif |
984 | ||
61c4628b SS |
985 | /* do the arch specific task caches init */ |
986 | arch_task_cache_init(); | |
987 | ||
16db3d3f | 988 | set_max_threads(MAX_THREADS); |
1da177e4 LT |
989 | |
990 | init_task.signal->rlim[RLIMIT_NPROC].rlim_cur = max_threads/2; | |
991 | init_task.signal->rlim[RLIMIT_NPROC].rlim_max = max_threads/2; | |
992 | init_task.signal->rlim[RLIMIT_SIGPENDING] = | |
993 | init_task.signal->rlim[RLIMIT_NPROC]; | |
b376c3e1 | 994 | |
de399236 | 995 | for (i = 0; i < UCOUNT_COUNTS; i++) |
25f9c081 | 996 | init_user_ns.ucount_max[i] = max_threads/2; |
19659c59 | 997 | |
de399236 AG |
998 | set_userns_rlimit_max(&init_user_ns, UCOUNT_RLIMIT_NPROC, RLIM_INFINITY); |
999 | set_userns_rlimit_max(&init_user_ns, UCOUNT_RLIMIT_MSGQUEUE, RLIM_INFINITY); | |
1000 | set_userns_rlimit_max(&init_user_ns, UCOUNT_RLIMIT_SIGPENDING, RLIM_INFINITY); | |
1001 | set_userns_rlimit_max(&init_user_ns, UCOUNT_RLIMIT_MEMLOCK, RLIM_INFINITY); | |
21d1c5e3 | 1002 | |
19659c59 HR |
1003 | #ifdef CONFIG_VMAP_STACK |
1004 | cpuhp_setup_state(CPUHP_BP_PREPARE_DYN, "fork:vm_stack_cache", | |
1005 | NULL, free_vm_stack_cache); | |
1006 | #endif | |
b09be676 | 1007 | |
d08b9f0c ST |
1008 | scs_init(); |
1009 | ||
b09be676 | 1010 | lockdep_init_task(&init_task); |
aad42dd4 | 1011 | uprobes_init(); |
1da177e4 LT |
1012 | } |
1013 | ||
52f5684c | 1014 | int __weak arch_dup_task_struct(struct task_struct *dst, |
61c4628b SS |
1015 | struct task_struct *src) |
1016 | { | |
1017 | *dst = *src; | |
1018 | return 0; | |
1019 | } | |
1020 | ||
d4311ff1 AT |
1021 | void set_task_stack_end_magic(struct task_struct *tsk) |
1022 | { | |
1023 | unsigned long *stackend; | |
1024 | ||
1025 | stackend = end_of_stack(tsk); | |
1026 | *stackend = STACK_END_MAGIC; /* for overflow detection */ | |
1027 | } | |
1028 | ||
725fc629 | 1029 | static struct task_struct *dup_task_struct(struct task_struct *orig, int node) |
1da177e4 LT |
1030 | { |
1031 | struct task_struct *tsk; | |
3e26c149 | 1032 | int err; |
1da177e4 | 1033 | |
725fc629 AK |
1034 | if (node == NUMA_NO_NODE) |
1035 | node = tsk_fork_get_node(orig); | |
504f52b5 | 1036 | tsk = alloc_task_struct_node(node); |
1da177e4 LT |
1037 | if (!tsk) |
1038 | return NULL; | |
1039 | ||
546c42b2 SAS |
1040 | err = arch_dup_task_struct(tsk, orig); |
1041 | if (err) | |
f19b9f74 | 1042 | goto free_tsk; |
1da177e4 | 1043 | |
7865aba3 SAS |
1044 | err = alloc_thread_stack_node(tsk, node); |
1045 | if (err) | |
f19b9f74 | 1046 | goto free_tsk; |
ba14a194 | 1047 | |
68f24b08 | 1048 | #ifdef CONFIG_THREAD_INFO_IN_TASK |
f0b89d39 | 1049 | refcount_set(&tsk->stack_refcount, 1); |
68f24b08 | 1050 | #endif |
1a03d3f1 | 1051 | account_kernel_stack(tsk, 1); |
164c33c6 | 1052 | |
d08b9f0c ST |
1053 | err = scs_prepare(tsk, node); |
1054 | if (err) | |
1055 | goto free_stack; | |
1056 | ||
dbd95212 KC |
1057 | #ifdef CONFIG_SECCOMP |
1058 | /* | |
1059 | * We must handle setting up seccomp filters once we're under | |
1060 | * the sighand lock in case orig has changed between now and | |
1061 | * then. Until then, filter must be NULL to avoid messing up | |
1062 | * the usage counts on the error path calling free_task. | |
1063 | */ | |
1064 | tsk->seccomp.filter = NULL; | |
1065 | #endif | |
87bec58a AM |
1066 | |
1067 | setup_thread_stack(tsk, orig); | |
8e7cac79 | 1068 | clear_user_return_notifier(tsk); |
f26f9aff | 1069 | clear_tsk_need_resched(tsk); |
d4311ff1 | 1070 | set_task_stack_end_magic(tsk); |
1446e1df | 1071 | clear_syscall_work_syscall_user_dispatch(tsk); |
1da177e4 | 1072 | |
050e9baa | 1073 | #ifdef CONFIG_STACKPROTECTOR |
7cd815bc | 1074 | tsk->stack_canary = get_random_canary(); |
0a425405 | 1075 | #endif |
3bd37062 SAS |
1076 | if (orig->cpus_ptr == &orig->cpus_mask) |
1077 | tsk->cpus_ptr = &tsk->cpus_mask; | |
b90ca8ba | 1078 | dup_user_cpus_ptr(tsk, orig, node); |
0a425405 | 1079 | |
fb0a685c | 1080 | /* |
0ff7b2cf EB |
1081 | * One for the user space visible state that goes away when reaped. |
1082 | * One for the scheduler. | |
fb0a685c | 1083 | */ |
0ff7b2cf EB |
1084 | refcount_set(&tsk->rcu_users, 2); |
1085 | /* One for the rcu users */ | |
1086 | refcount_set(&tsk->usage, 1); | |
6c5c9341 | 1087 | #ifdef CONFIG_BLK_DEV_IO_TRACE |
2056a782 | 1088 | tsk->btrace_seq = 0; |
6c5c9341 | 1089 | #endif |
a0aa7f68 | 1090 | tsk->splice_pipe = NULL; |
5640f768 | 1091 | tsk->task_frag.page = NULL; |
093e5840 | 1092 | tsk->wake_q.next = NULL; |
e32cf5df | 1093 | tsk->worker_private = NULL; |
c6a7f572 | 1094 | |
5c9a8750 | 1095 | kcov_task_init(tsk); |
50b5e49c | 1096 | kmsan_task_create(tsk); |
5fbda3ec | 1097 | kmap_local_fork(tsk); |
5c9a8750 | 1098 | |
e41d5818 DV |
1099 | #ifdef CONFIG_FAULT_INJECTION |
1100 | tsk->fail_nth = 0; | |
1101 | #endif | |
1102 | ||
2c323017 | 1103 | #ifdef CONFIG_BLK_CGROUP |
f05837ed | 1104 | tsk->throttle_disk = NULL; |
2c323017 JB |
1105 | tsk->use_memdelay = 0; |
1106 | #endif | |
1107 | ||
a3d29e82 PZ |
1108 | #ifdef CONFIG_IOMMU_SVA |
1109 | tsk->pasid_activated = 0; | |
1110 | #endif | |
1111 | ||
d46eb14b SB |
1112 | #ifdef CONFIG_MEMCG |
1113 | tsk->active_memcg = NULL; | |
1114 | #endif | |
b041b525 TL |
1115 | |
1116 | #ifdef CONFIG_CPU_SUP_INTEL | |
1117 | tsk->reported_split_lock = 0; | |
1118 | #endif | |
1119 | ||
af7f588d MD |
1120 | #ifdef CONFIG_SCHED_MM_CID |
1121 | tsk->mm_cid = -1; | |
1122 | tsk->mm_cid_active = 0; | |
1123 | #endif | |
1da177e4 | 1124 | return tsk; |
61c4628b | 1125 | |
b235beea | 1126 | free_stack: |
1a03d3f1 | 1127 | exit_task_stack_account(tsk); |
ba14a194 | 1128 | free_thread_stack(tsk); |
f19b9f74 | 1129 | free_tsk: |
61c4628b SS |
1130 | free_task_struct(tsk); |
1131 | return NULL; | |
1da177e4 LT |
1132 | } |
1133 | ||
23ff4440 | 1134 | __cacheline_aligned_in_smp DEFINE_SPINLOCK(mmlist_lock); |
1da177e4 | 1135 | |
4cb0e11b HK |
1136 | static unsigned long default_dump_filter = MMF_DUMP_FILTER_DEFAULT; |
1137 | ||
1138 | static int __init coredump_filter_setup(char *s) | |
1139 | { | |
1140 | default_dump_filter = | |
1141 | (simple_strtoul(s, NULL, 0) << MMF_DUMP_FILTER_SHIFT) & | |
1142 | MMF_DUMP_FILTER_MASK; | |
1143 | return 1; | |
1144 | } | |
1145 | ||
1146 | __setup("coredump_filter=", coredump_filter_setup); | |
1147 | ||
1da177e4 LT |
1148 | #include <linux/init_task.h> |
1149 | ||
858f0993 AD |
1150 | static void mm_init_aio(struct mm_struct *mm) |
1151 | { | |
1152 | #ifdef CONFIG_AIO | |
1153 | spin_lock_init(&mm->ioctx_lock); | |
db446a08 | 1154 | mm->ioctx_table = NULL; |
858f0993 AD |
1155 | #endif |
1156 | } | |
1157 | ||
c3f3ce04 AA |
1158 | static __always_inline void mm_clear_owner(struct mm_struct *mm, |
1159 | struct task_struct *p) | |
1160 | { | |
1161 | #ifdef CONFIG_MEMCG | |
1162 | if (mm->owner == p) | |
1163 | WRITE_ONCE(mm->owner, NULL); | |
1164 | #endif | |
1165 | } | |
1166 | ||
33144e84 VD |
1167 | static void mm_init_owner(struct mm_struct *mm, struct task_struct *p) |
1168 | { | |
1169 | #ifdef CONFIG_MEMCG | |
1170 | mm->owner = p; | |
1171 | #endif | |
1172 | } | |
1173 | ||
355627f5 EB |
1174 | static void mm_init_uprobes_state(struct mm_struct *mm) |
1175 | { | |
1176 | #ifdef CONFIG_UPROBES | |
1177 | mm->uprobes_state.xol_area = NULL; | |
1178 | #endif | |
1179 | } | |
1180 | ||
bfedb589 EB |
1181 | static struct mm_struct *mm_init(struct mm_struct *mm, struct task_struct *p, |
1182 | struct user_namespace *user_ns) | |
1da177e4 | 1183 | { |
f1a79412 SB |
1184 | int i; |
1185 | ||
d4af56c5 LH |
1186 | mt_init_flags(&mm->mm_mt, MM_MT_FLAGS); |
1187 | mt_set_external_lock(&mm->mm_mt, &mm->mmap_lock); | |
1da177e4 LT |
1188 | atomic_set(&mm->mm_users, 1); |
1189 | atomic_set(&mm->mm_count, 1); | |
57efa1fe | 1190 | seqcount_init(&mm->write_protect_seq); |
d8ed45c5 | 1191 | mmap_init_lock(mm); |
1da177e4 | 1192 | INIT_LIST_HEAD(&mm->mmlist); |
af5b0f6a | 1193 | mm_pgtables_bytes_init(mm); |
41f727fd VD |
1194 | mm->map_count = 0; |
1195 | mm->locked_vm = 0; | |
70f8a3ca | 1196 | atomic64_set(&mm->pinned_vm, 0); |
d559db08 | 1197 | memset(&mm->rss_stat, 0, sizeof(mm->rss_stat)); |
1da177e4 | 1198 | spin_lock_init(&mm->page_table_lock); |
88aa7cc6 | 1199 | spin_lock_init(&mm->arg_lock); |
41f727fd | 1200 | mm_init_cpumask(mm); |
858f0993 | 1201 | mm_init_aio(mm); |
cf475ad2 | 1202 | mm_init_owner(mm, p); |
a6cbd440 | 1203 | mm_pasid_init(mm); |
2b7e8665 | 1204 | RCU_INIT_POINTER(mm->exe_file, NULL); |
984cfe4e | 1205 | mmu_notifier_subscriptions_init(mm); |
16af97dc | 1206 | init_tlb_flush_pending(mm); |
41f727fd VD |
1207 | #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS |
1208 | mm->pmd_huge_pte = NULL; | |
1209 | #endif | |
355627f5 | 1210 | mm_init_uprobes_state(mm); |
13db8c50 | 1211 | hugetlb_count_init(mm); |
1da177e4 | 1212 | |
a0715cc2 AT |
1213 | if (current->mm) { |
1214 | mm->flags = current->mm->flags & MMF_INIT_MASK; | |
1215 | mm->def_flags = current->mm->def_flags & VM_INIT_DEF_MASK; | |
1216 | } else { | |
1217 | mm->flags = default_dump_filter; | |
1da177e4 | 1218 | mm->def_flags = 0; |
a0715cc2 AT |
1219 | } |
1220 | ||
41f727fd VD |
1221 | if (mm_alloc_pgd(mm)) |
1222 | goto fail_nopgd; | |
1223 | ||
1224 | if (init_new_context(p, mm)) | |
1225 | goto fail_nocontext; | |
78fb7466 | 1226 | |
f1a79412 SB |
1227 | for (i = 0; i < NR_MM_COUNTERS; i++) |
1228 | if (percpu_counter_init(&mm->rss_stat[i], 0, GFP_KERNEL_ACCOUNT)) | |
1229 | goto fail_pcpu; | |
1230 | ||
bfedb589 | 1231 | mm->user_ns = get_user_ns(user_ns); |
bd74fdae | 1232 | lru_gen_init_mm(mm); |
af7f588d | 1233 | mm_init_cid(mm); |
41f727fd VD |
1234 | return mm; |
1235 | ||
f1a79412 SB |
1236 | fail_pcpu: |
1237 | while (i > 0) | |
1238 | percpu_counter_destroy(&mm->rss_stat[--i]); | |
41f727fd VD |
1239 | fail_nocontext: |
1240 | mm_free_pgd(mm); | |
1241 | fail_nopgd: | |
1da177e4 LT |
1242 | free_mm(mm); |
1243 | return NULL; | |
1244 | } | |
1245 | ||
1246 | /* | |
1247 | * Allocate and initialize an mm_struct. | |
1248 | */ | |
fb0a685c | 1249 | struct mm_struct *mm_alloc(void) |
1da177e4 | 1250 | { |
fb0a685c | 1251 | struct mm_struct *mm; |
1da177e4 LT |
1252 | |
1253 | mm = allocate_mm(); | |
de03c72c KM |
1254 | if (!mm) |
1255 | return NULL; | |
1256 | ||
1257 | memset(mm, 0, sizeof(*mm)); | |
bfedb589 | 1258 | return mm_init(mm, current, current_user_ns()); |
1da177e4 LT |
1259 | } |
1260 | ||
ec8d7c14 MH |
1261 | static inline void __mmput(struct mm_struct *mm) |
1262 | { | |
1263 | VM_BUG_ON(atomic_read(&mm->mm_users)); | |
1264 | ||
1265 | uprobe_clear_state(mm); | |
1266 | exit_aio(mm); | |
1267 | ksm_exit(mm); | |
1268 | khugepaged_exit(mm); /* must run before exit_mmap */ | |
1269 | exit_mmap(mm); | |
6fcb52a5 | 1270 | mm_put_huge_zero_page(mm); |
ec8d7c14 MH |
1271 | set_mm_exe_file(mm, NULL); |
1272 | if (!list_empty(&mm->mmlist)) { | |
1273 | spin_lock(&mmlist_lock); | |
1274 | list_del(&mm->mmlist); | |
1275 | spin_unlock(&mmlist_lock); | |
1276 | } | |
1277 | if (mm->binfmt) | |
1278 | module_put(mm->binfmt->module); | |
bd74fdae | 1279 | lru_gen_del_mm(mm); |
ec8d7c14 MH |
1280 | mmdrop(mm); |
1281 | } | |
1282 | ||
1da177e4 LT |
1283 | /* |
1284 | * Decrement the use count and release all resources for an mm. | |
1285 | */ | |
1286 | void mmput(struct mm_struct *mm) | |
1287 | { | |
0ae26f1b AM |
1288 | might_sleep(); |
1289 | ||
ec8d7c14 MH |
1290 | if (atomic_dec_and_test(&mm->mm_users)) |
1291 | __mmput(mm); | |
1292 | } | |
1293 | EXPORT_SYMBOL_GPL(mmput); | |
1294 | ||
a1b2289c SY |
1295 | #ifdef CONFIG_MMU |
1296 | static void mmput_async_fn(struct work_struct *work) | |
1297 | { | |
1298 | struct mm_struct *mm = container_of(work, struct mm_struct, | |
1299 | async_put_work); | |
1300 | ||
1301 | __mmput(mm); | |
1302 | } | |
1303 | ||
1304 | void mmput_async(struct mm_struct *mm) | |
1305 | { | |
1306 | if (atomic_dec_and_test(&mm->mm_users)) { | |
1307 | INIT_WORK(&mm->async_put_work, mmput_async_fn); | |
1308 | schedule_work(&mm->async_put_work); | |
1309 | } | |
1310 | } | |
85eaeb50 | 1311 | EXPORT_SYMBOL_GPL(mmput_async); |
a1b2289c SY |
1312 | #endif |
1313 | ||
90f31d0e KK |
1314 | /** |
1315 | * set_mm_exe_file - change a reference to the mm's executable file | |
1316 | * | |
1317 | * This changes mm's executable file (shown as symlink /proc/[pid]/exe). | |
1318 | * | |
6e399cd1 DB |
1319 | * Main users are mmput() and sys_execve(). Callers prevent concurrent |
1320 | * invocations: in mmput() nobody alive left, in execve task is single | |
35d7bdc8 | 1321 | * threaded. |
fe69d560 DH |
1322 | * |
1323 | * Can only fail if new_exe_file != NULL. | |
90f31d0e | 1324 | */ |
fe69d560 | 1325 | int set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file) |
38646013 | 1326 | { |
6e399cd1 DB |
1327 | struct file *old_exe_file; |
1328 | ||
1329 | /* | |
1330 | * It is safe to dereference the exe_file without RCU as | |
1331 | * this function is only called if nobody else can access | |
1332 | * this mm -- see comment above for justification. | |
1333 | */ | |
1334 | old_exe_file = rcu_dereference_raw(mm->exe_file); | |
90f31d0e | 1335 | |
fe69d560 DH |
1336 | if (new_exe_file) { |
1337 | /* | |
1338 | * We expect the caller (i.e., sys_execve) to already denied | |
1339 | * write access, so this is unlikely to fail. | |
1340 | */ | |
1341 | if (unlikely(deny_write_access(new_exe_file))) | |
1342 | return -EACCES; | |
38646013 | 1343 | get_file(new_exe_file); |
fe69d560 | 1344 | } |
90f31d0e | 1345 | rcu_assign_pointer(mm->exe_file, new_exe_file); |
fe69d560 DH |
1346 | if (old_exe_file) { |
1347 | allow_write_access(old_exe_file); | |
90f31d0e | 1348 | fput(old_exe_file); |
fe69d560 DH |
1349 | } |
1350 | return 0; | |
38646013 JS |
1351 | } |
1352 | ||
35d7bdc8 DH |
1353 | /** |
1354 | * replace_mm_exe_file - replace a reference to the mm's executable file | |
1355 | * | |
1356 | * This changes mm's executable file (shown as symlink /proc/[pid]/exe), | |
1357 | * dealing with concurrent invocation and without grabbing the mmap lock in | |
1358 | * write mode. | |
1359 | * | |
1360 | * Main user is sys_prctl(PR_SET_MM_MAP/EXE_FILE). | |
1361 | */ | |
1362 | int replace_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file) | |
1363 | { | |
1364 | struct vm_area_struct *vma; | |
1365 | struct file *old_exe_file; | |
1366 | int ret = 0; | |
1367 | ||
1368 | /* Forbid mm->exe_file change if old file still mapped. */ | |
1369 | old_exe_file = get_mm_exe_file(mm); | |
1370 | if (old_exe_file) { | |
fa5e5876 | 1371 | VMA_ITERATOR(vmi, mm, 0); |
35d7bdc8 | 1372 | mmap_read_lock(mm); |
fa5e5876 | 1373 | for_each_vma(vmi, vma) { |
35d7bdc8 DH |
1374 | if (!vma->vm_file) |
1375 | continue; | |
1376 | if (path_equal(&vma->vm_file->f_path, | |
fa5e5876 | 1377 | &old_exe_file->f_path)) { |
35d7bdc8 | 1378 | ret = -EBUSY; |
fa5e5876 MWO |
1379 | break; |
1380 | } | |
35d7bdc8 DH |
1381 | } |
1382 | mmap_read_unlock(mm); | |
1383 | fput(old_exe_file); | |
1384 | if (ret) | |
1385 | return ret; | |
1386 | } | |
1387 | ||
1388 | /* set the new file, lockless */ | |
fe69d560 DH |
1389 | ret = deny_write_access(new_exe_file); |
1390 | if (ret) | |
1391 | return -EACCES; | |
35d7bdc8 | 1392 | get_file(new_exe_file); |
fe69d560 | 1393 | |
35d7bdc8 | 1394 | old_exe_file = xchg(&mm->exe_file, new_exe_file); |
fe69d560 DH |
1395 | if (old_exe_file) { |
1396 | /* | |
1397 | * Don't race with dup_mmap() getting the file and disallowing | |
1398 | * write access while someone might open the file writable. | |
1399 | */ | |
1400 | mmap_read_lock(mm); | |
1401 | allow_write_access(old_exe_file); | |
35d7bdc8 | 1402 | fput(old_exe_file); |
fe69d560 DH |
1403 | mmap_read_unlock(mm); |
1404 | } | |
35d7bdc8 | 1405 | return 0; |
38646013 JS |
1406 | } |
1407 | ||
90f31d0e KK |
1408 | /** |
1409 | * get_mm_exe_file - acquire a reference to the mm's executable file | |
1410 | * | |
1411 | * Returns %NULL if mm has no associated executable file. | |
1412 | * User must release file via fput(). | |
1413 | */ | |
38646013 JS |
1414 | struct file *get_mm_exe_file(struct mm_struct *mm) |
1415 | { | |
1416 | struct file *exe_file; | |
1417 | ||
90f31d0e KK |
1418 | rcu_read_lock(); |
1419 | exe_file = rcu_dereference(mm->exe_file); | |
1420 | if (exe_file && !get_file_rcu(exe_file)) | |
1421 | exe_file = NULL; | |
1422 | rcu_read_unlock(); | |
38646013 JS |
1423 | return exe_file; |
1424 | } | |
1425 | ||
cd81a917 MG |
1426 | /** |
1427 | * get_task_exe_file - acquire a reference to the task's executable file | |
1428 | * | |
1429 | * Returns %NULL if task's mm (if any) has no associated executable file or | |
1430 | * this is a kernel thread with borrowed mm (see the comment above get_task_mm). | |
1431 | * User must release file via fput(). | |
1432 | */ | |
1433 | struct file *get_task_exe_file(struct task_struct *task) | |
1434 | { | |
1435 | struct file *exe_file = NULL; | |
1436 | struct mm_struct *mm; | |
1437 | ||
1438 | task_lock(task); | |
1439 | mm = task->mm; | |
1440 | if (mm) { | |
1441 | if (!(task->flags & PF_KTHREAD)) | |
1442 | exe_file = get_mm_exe_file(mm); | |
1443 | } | |
1444 | task_unlock(task); | |
1445 | return exe_file; | |
1446 | } | |
38646013 | 1447 | |
1da177e4 LT |
1448 | /** |
1449 | * get_task_mm - acquire a reference to the task's mm | |
1450 | * | |
246bb0b1 | 1451 | * Returns %NULL if the task has no mm. Checks PF_KTHREAD (meaning |
1da177e4 LT |
1452 | * this kernel workthread has transiently adopted a user mm with use_mm, |
1453 | * to do its AIO) is not set and if so returns a reference to it, after | |
1454 | * bumping up the use count. User must release the mm via mmput() | |
1455 | * after use. Typically used by /proc and ptrace. | |
1456 | */ | |
1457 | struct mm_struct *get_task_mm(struct task_struct *task) | |
1458 | { | |
1459 | struct mm_struct *mm; | |
1460 | ||
1461 | task_lock(task); | |
1462 | mm = task->mm; | |
1463 | if (mm) { | |
246bb0b1 | 1464 | if (task->flags & PF_KTHREAD) |
1da177e4 LT |
1465 | mm = NULL; |
1466 | else | |
3fce371b | 1467 | mmget(mm); |
1da177e4 LT |
1468 | } |
1469 | task_unlock(task); | |
1470 | return mm; | |
1471 | } | |
1472 | EXPORT_SYMBOL_GPL(get_task_mm); | |
1473 | ||
8cdb878d CY |
1474 | struct mm_struct *mm_access(struct task_struct *task, unsigned int mode) |
1475 | { | |
1476 | struct mm_struct *mm; | |
1477 | int err; | |
1478 | ||
f7cfd871 | 1479 | err = down_read_killable(&task->signal->exec_update_lock); |
8cdb878d CY |
1480 | if (err) |
1481 | return ERR_PTR(err); | |
1482 | ||
1483 | mm = get_task_mm(task); | |
1484 | if (mm && mm != current->mm && | |
1485 | !ptrace_may_access(task, mode)) { | |
1486 | mmput(mm); | |
1487 | mm = ERR_PTR(-EACCES); | |
1488 | } | |
f7cfd871 | 1489 | up_read(&task->signal->exec_update_lock); |
8cdb878d CY |
1490 | |
1491 | return mm; | |
1492 | } | |
1493 | ||
57b59c4a | 1494 | static void complete_vfork_done(struct task_struct *tsk) |
c415c3b4 | 1495 | { |
d68b46fe | 1496 | struct completion *vfork; |
c415c3b4 | 1497 | |
d68b46fe ON |
1498 | task_lock(tsk); |
1499 | vfork = tsk->vfork_done; | |
1500 | if (likely(vfork)) { | |
1501 | tsk->vfork_done = NULL; | |
1502 | complete(vfork); | |
1503 | } | |
1504 | task_unlock(tsk); | |
1505 | } | |
1506 | ||
1507 | static int wait_for_vfork_done(struct task_struct *child, | |
1508 | struct completion *vfork) | |
1509 | { | |
f5d39b02 | 1510 | unsigned int state = TASK_UNINTERRUPTIBLE|TASK_KILLABLE|TASK_FREEZABLE; |
d68b46fe ON |
1511 | int killed; |
1512 | ||
76f969e8 | 1513 | cgroup_enter_frozen(); |
f5d39b02 | 1514 | killed = wait_for_completion_state(vfork, state); |
76f969e8 | 1515 | cgroup_leave_frozen(false); |
d68b46fe ON |
1516 | |
1517 | if (killed) { | |
1518 | task_lock(child); | |
1519 | child->vfork_done = NULL; | |
1520 | task_unlock(child); | |
1521 | } | |
1522 | ||
1523 | put_task_struct(child); | |
1524 | return killed; | |
c415c3b4 ON |
1525 | } |
1526 | ||
1da177e4 LT |
1527 | /* Please note the differences between mmput and mm_release. |
1528 | * mmput is called whenever we stop holding onto a mm_struct, | |
1529 | * error success whatever. | |
1530 | * | |
1531 | * mm_release is called after a mm_struct has been removed | |
1532 | * from the current process. | |
1533 | * | |
1534 | * This difference is important for error handling, when we | |
1535 | * only half set up a mm_struct for a new process and need to restore | |
1536 | * the old one. Because we mmput the new mm_struct before | |
1537 | * restoring the old one. . . | |
1538 | * Eric Biederman 10 January 1998 | |
1539 | */ | |
4610ba7a | 1540 | static void mm_release(struct task_struct *tsk, struct mm_struct *mm) |
1da177e4 | 1541 | { |
0326f5a9 SD |
1542 | uprobe_free_utask(tsk); |
1543 | ||
1da177e4 LT |
1544 | /* Get rid of any cached register state */ |
1545 | deactivate_mm(tsk, mm); | |
1546 | ||
fec1d011 | 1547 | /* |
735f2770 MH |
1548 | * Signal userspace if we're not exiting with a core dump |
1549 | * because we want to leave the value intact for debugging | |
1550 | * purposes. | |
fec1d011 | 1551 | */ |
9c8a8228 | 1552 | if (tsk->clear_child_tid) { |
92307383 | 1553 | if (atomic_read(&mm->mm_users) > 1) { |
9c8a8228 ED |
1554 | /* |
1555 | * We don't check the error code - if userspace has | |
1556 | * not set up a proper pointer then tough luck. | |
1557 | */ | |
1558 | put_user(0, tsk->clear_child_tid); | |
2de0db99 DB |
1559 | do_futex(tsk->clear_child_tid, FUTEX_WAKE, |
1560 | 1, NULL, NULL, 0, 0); | |
9c8a8228 | 1561 | } |
1da177e4 | 1562 | tsk->clear_child_tid = NULL; |
1da177e4 | 1563 | } |
f7505d64 KK |
1564 | |
1565 | /* | |
1566 | * All done, finally we can wake up parent and return this mm to him. | |
1567 | * Also kthread_stop() uses this completion for synchronization. | |
1568 | */ | |
1569 | if (tsk->vfork_done) | |
1570 | complete_vfork_done(tsk); | |
1da177e4 LT |
1571 | } |
1572 | ||
4610ba7a TG |
1573 | void exit_mm_release(struct task_struct *tsk, struct mm_struct *mm) |
1574 | { | |
150d7158 | 1575 | futex_exit_release(tsk); |
4610ba7a TG |
1576 | mm_release(tsk, mm); |
1577 | } | |
1578 | ||
1579 | void exec_mm_release(struct task_struct *tsk, struct mm_struct *mm) | |
1580 | { | |
150d7158 | 1581 | futex_exec_release(tsk); |
4610ba7a TG |
1582 | mm_release(tsk, mm); |
1583 | } | |
1584 | ||
13585fa0 NA |
1585 | /** |
1586 | * dup_mm() - duplicates an existing mm structure | |
1587 | * @tsk: the task_struct with which the new mm will be associated. | |
1588 | * @oldmm: the mm to duplicate. | |
1589 | * | |
1590 | * Allocates a new mm structure and duplicates the provided @oldmm structure | |
1591 | * content into it. | |
1592 | * | |
1593 | * Return: the duplicated mm or NULL on failure. | |
a0a7ec30 | 1594 | */ |
13585fa0 NA |
1595 | static struct mm_struct *dup_mm(struct task_struct *tsk, |
1596 | struct mm_struct *oldmm) | |
a0a7ec30 | 1597 | { |
13585fa0 | 1598 | struct mm_struct *mm; |
a0a7ec30 JD |
1599 | int err; |
1600 | ||
a0a7ec30 JD |
1601 | mm = allocate_mm(); |
1602 | if (!mm) | |
1603 | goto fail_nomem; | |
1604 | ||
1605 | memcpy(mm, oldmm, sizeof(*mm)); | |
1606 | ||
bfedb589 | 1607 | if (!mm_init(mm, tsk, mm->user_ns)) |
a0a7ec30 JD |
1608 | goto fail_nomem; |
1609 | ||
a0a7ec30 JD |
1610 | err = dup_mmap(mm, oldmm); |
1611 | if (err) | |
1612 | goto free_pt; | |
1613 | ||
1614 | mm->hiwater_rss = get_mm_rss(mm); | |
1615 | mm->hiwater_vm = mm->total_vm; | |
1616 | ||
801460d0 HS |
1617 | if (mm->binfmt && !try_module_get(mm->binfmt->module)) |
1618 | goto free_pt; | |
1619 | ||
a0a7ec30 JD |
1620 | return mm; |
1621 | ||
1622 | free_pt: | |
801460d0 HS |
1623 | /* don't put binfmt in mmput, we haven't got module yet */ |
1624 | mm->binfmt = NULL; | |
c3f3ce04 | 1625 | mm_init_owner(mm, NULL); |
a0a7ec30 JD |
1626 | mmput(mm); |
1627 | ||
1628 | fail_nomem: | |
1629 | return NULL; | |
a0a7ec30 JD |
1630 | } |
1631 | ||
fb0a685c | 1632 | static int copy_mm(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 | 1633 | { |
fb0a685c | 1634 | struct mm_struct *mm, *oldmm; |
1da177e4 LT |
1635 | |
1636 | tsk->min_flt = tsk->maj_flt = 0; | |
1637 | tsk->nvcsw = tsk->nivcsw = 0; | |
17406b82 MSB |
1638 | #ifdef CONFIG_DETECT_HUNG_TASK |
1639 | tsk->last_switch_count = tsk->nvcsw + tsk->nivcsw; | |
a2e51445 | 1640 | tsk->last_switch_time = 0; |
17406b82 | 1641 | #endif |
1da177e4 LT |
1642 | |
1643 | tsk->mm = NULL; | |
1644 | tsk->active_mm = NULL; | |
1645 | ||
1646 | /* | |
1647 | * Are we cloning a kernel thread? | |
1648 | * | |
1649 | * We need to steal a active VM for that.. | |
1650 | */ | |
1651 | oldmm = current->mm; | |
1652 | if (!oldmm) | |
1653 | return 0; | |
1654 | ||
1655 | if (clone_flags & CLONE_VM) { | |
3fce371b | 1656 | mmget(oldmm); |
1da177e4 | 1657 | mm = oldmm; |
a6895399 REB |
1658 | } else { |
1659 | mm = dup_mm(tsk, current->mm); | |
1660 | if (!mm) | |
1661 | return -ENOMEM; | |
1da177e4 LT |
1662 | } |
1663 | ||
1da177e4 LT |
1664 | tsk->mm = mm; |
1665 | tsk->active_mm = mm; | |
af7f588d | 1666 | sched_mm_cid_fork(tsk); |
1da177e4 | 1667 | return 0; |
1da177e4 LT |
1668 | } |
1669 | ||
a39bc516 | 1670 | static int copy_fs(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 | 1671 | { |
498052bb | 1672 | struct fs_struct *fs = current->fs; |
1da177e4 | 1673 | if (clone_flags & CLONE_FS) { |
498052bb | 1674 | /* tsk->fs is already what we want */ |
2a4419b5 | 1675 | spin_lock(&fs->lock); |
498052bb | 1676 | if (fs->in_exec) { |
2a4419b5 | 1677 | spin_unlock(&fs->lock); |
498052bb AV |
1678 | return -EAGAIN; |
1679 | } | |
1680 | fs->users++; | |
2a4419b5 | 1681 | spin_unlock(&fs->lock); |
1da177e4 LT |
1682 | return 0; |
1683 | } | |
498052bb | 1684 | tsk->fs = copy_fs_struct(fs); |
1da177e4 LT |
1685 | if (!tsk->fs) |
1686 | return -ENOMEM; | |
1687 | return 0; | |
1688 | } | |
1689 | ||
fb0a685c | 1690 | static int copy_files(unsigned long clone_flags, struct task_struct *tsk) |
a016f338 JD |
1691 | { |
1692 | struct files_struct *oldf, *newf; | |
1693 | int error = 0; | |
1694 | ||
1695 | /* | |
1696 | * A background process may not have any files ... | |
1697 | */ | |
1698 | oldf = current->files; | |
1699 | if (!oldf) | |
1700 | goto out; | |
1701 | ||
1702 | if (clone_flags & CLONE_FILES) { | |
1703 | atomic_inc(&oldf->count); | |
1704 | goto out; | |
1705 | } | |
1706 | ||
60997c3d | 1707 | newf = dup_fd(oldf, NR_OPEN_MAX, &error); |
a016f338 JD |
1708 | if (!newf) |
1709 | goto out; | |
1710 | ||
1711 | tsk->files = newf; | |
1712 | error = 0; | |
1713 | out: | |
1714 | return error; | |
1715 | } | |
1716 | ||
a39bc516 | 1717 | static int copy_sighand(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 LT |
1718 | { |
1719 | struct sighand_struct *sig; | |
1720 | ||
60348802 | 1721 | if (clone_flags & CLONE_SIGHAND) { |
d036bda7 | 1722 | refcount_inc(¤t->sighand->count); |
1da177e4 LT |
1723 | return 0; |
1724 | } | |
1725 | sig = kmem_cache_alloc(sighand_cachep, GFP_KERNEL); | |
0c282b06 | 1726 | RCU_INIT_POINTER(tsk->sighand, sig); |
1da177e4 LT |
1727 | if (!sig) |
1728 | return -ENOMEM; | |
9d7fb042 | 1729 | |
d036bda7 | 1730 | refcount_set(&sig->count, 1); |
06e62a46 | 1731 | spin_lock_irq(¤t->sighand->siglock); |
1da177e4 | 1732 | memcpy(sig->action, current->sighand->action, sizeof(sig->action)); |
06e62a46 | 1733 | spin_unlock_irq(¤t->sighand->siglock); |
b612e5df CB |
1734 | |
1735 | /* Reset all signal handler not set to SIG_IGN to SIG_DFL. */ | |
1736 | if (clone_flags & CLONE_CLEAR_SIGHAND) | |
1737 | flush_signal_handlers(tsk, 0); | |
1738 | ||
1da177e4 LT |
1739 | return 0; |
1740 | } | |
1741 | ||
a7e5328a | 1742 | void __cleanup_sighand(struct sighand_struct *sighand) |
c81addc9 | 1743 | { |
d036bda7 | 1744 | if (refcount_dec_and_test(&sighand->count)) { |
d80e731e | 1745 | signalfd_cleanup(sighand); |
392809b2 | 1746 | /* |
5f0d5a3a | 1747 | * sighand_cachep is SLAB_TYPESAFE_BY_RCU so we can free it |
392809b2 ON |
1748 | * without an RCU grace period, see __lock_task_sighand(). |
1749 | */ | |
c81addc9 | 1750 | kmem_cache_free(sighand_cachep, sighand); |
d80e731e | 1751 | } |
c81addc9 ON |
1752 | } |
1753 | ||
f06febc9 FM |
1754 | /* |
1755 | * Initialize POSIX timer handling for a thread group. | |
1756 | */ | |
1757 | static void posix_cpu_timers_init_group(struct signal_struct *sig) | |
1758 | { | |
2b69942f | 1759 | struct posix_cputimers *pct = &sig->posix_cputimers; |
78d7d407 JS |
1760 | unsigned long cpu_limit; |
1761 | ||
316c1608 | 1762 | cpu_limit = READ_ONCE(sig->rlim[RLIMIT_CPU].rlim_cur); |
3a245c0f | 1763 | posix_cputimers_group_init(pct, cpu_limit); |
f06febc9 FM |
1764 | } |
1765 | ||
a39bc516 | 1766 | static int copy_signal(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 LT |
1767 | { |
1768 | struct signal_struct *sig; | |
1da177e4 | 1769 | |
4ab6c083 | 1770 | if (clone_flags & CLONE_THREAD) |
490dea45 | 1771 | return 0; |
490dea45 | 1772 | |
a56704ef | 1773 | sig = kmem_cache_zalloc(signal_cachep, GFP_KERNEL); |
1da177e4 LT |
1774 | tsk->signal = sig; |
1775 | if (!sig) | |
1776 | return -ENOMEM; | |
1777 | ||
b3ac022c | 1778 | sig->nr_threads = 1; |
d80f7d7b | 1779 | sig->quick_threads = 1; |
1da177e4 | 1780 | atomic_set(&sig->live, 1); |
60d4de3f | 1781 | refcount_set(&sig->sigcnt, 1); |
0c740d0a ON |
1782 | |
1783 | /* list_add(thread_node, thread_head) without INIT_LIST_HEAD() */ | |
1784 | sig->thread_head = (struct list_head)LIST_HEAD_INIT(tsk->thread_node); | |
1785 | tsk->thread_node = (struct list_head)LIST_HEAD_INIT(sig->thread_head); | |
1786 | ||
1da177e4 | 1787 | init_waitqueue_head(&sig->wait_chldexit); |
db51aecc | 1788 | sig->curr_target = tsk; |
1da177e4 | 1789 | init_sigpending(&sig->shared_pending); |
c3ad2c3b | 1790 | INIT_HLIST_HEAD(&sig->multiprocess); |
e78c3496 | 1791 | seqlock_init(&sig->stats_lock); |
9d7fb042 | 1792 | prev_cputime_init(&sig->prev_cputime); |
1da177e4 | 1793 | |
baa73d9e | 1794 | #ifdef CONFIG_POSIX_TIMERS |
b18b6a9c | 1795 | INIT_LIST_HEAD(&sig->posix_timers); |
c9cb2e3d | 1796 | hrtimer_init(&sig->real_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL); |
1da177e4 | 1797 | sig->real_timer.function = it_real_fn; |
baa73d9e | 1798 | #endif |
1da177e4 | 1799 | |
1da177e4 LT |
1800 | task_lock(current->group_leader); |
1801 | memcpy(sig->rlim, current->signal->rlim, sizeof sig->rlim); | |
1802 | task_unlock(current->group_leader); | |
1803 | ||
6279a751 ON |
1804 | posix_cpu_timers_init_group(sig); |
1805 | ||
522ed776 | 1806 | tty_audit_fork(sig); |
5091faa4 | 1807 | sched_autogroup_fork(sig); |
522ed776 | 1808 | |
a63d83f4 | 1809 | sig->oom_score_adj = current->signal->oom_score_adj; |
dabb16f6 | 1810 | sig->oom_score_adj_min = current->signal->oom_score_adj_min; |
28b83c51 | 1811 | |
9b1bf12d | 1812 | mutex_init(&sig->cred_guard_mutex); |
f7cfd871 | 1813 | init_rwsem(&sig->exec_update_lock); |
9b1bf12d | 1814 | |
1da177e4 LT |
1815 | return 0; |
1816 | } | |
1817 | ||
dbd95212 KC |
1818 | static void copy_seccomp(struct task_struct *p) |
1819 | { | |
1820 | #ifdef CONFIG_SECCOMP | |
1821 | /* | |
1822 | * Must be called with sighand->lock held, which is common to | |
1823 | * all threads in the group. Holding cred_guard_mutex is not | |
1824 | * needed because this new task is not yet running and cannot | |
1825 | * be racing exec. | |
1826 | */ | |
69f6a34b | 1827 | assert_spin_locked(¤t->sighand->siglock); |
dbd95212 KC |
1828 | |
1829 | /* Ref-count the new filter user, and assign it. */ | |
1830 | get_seccomp_filter(current); | |
1831 | p->seccomp = current->seccomp; | |
1832 | ||
1833 | /* | |
1834 | * Explicitly enable no_new_privs here in case it got set | |
1835 | * between the task_struct being duplicated and holding the | |
1836 | * sighand lock. The seccomp state and nnp must be in sync. | |
1837 | */ | |
1838 | if (task_no_new_privs(current)) | |
1839 | task_set_no_new_privs(p); | |
1840 | ||
1841 | /* | |
1842 | * If the parent gained a seccomp mode after copying thread | |
1843 | * flags and between before we held the sighand lock, we have | |
1844 | * to manually enable the seccomp thread flag here. | |
1845 | */ | |
1846 | if (p->seccomp.mode != SECCOMP_MODE_DISABLED) | |
23d67a54 | 1847 | set_task_syscall_work(p, SECCOMP); |
dbd95212 KC |
1848 | #endif |
1849 | } | |
1850 | ||
17da2bd9 | 1851 | SYSCALL_DEFINE1(set_tid_address, int __user *, tidptr) |
1da177e4 LT |
1852 | { |
1853 | current->clear_child_tid = tidptr; | |
1854 | ||
b488893a | 1855 | return task_pid_vnr(current); |
1da177e4 LT |
1856 | } |
1857 | ||
a39bc516 | 1858 | static void rt_mutex_init_task(struct task_struct *p) |
23f78d4a | 1859 | { |
1d615482 | 1860 | raw_spin_lock_init(&p->pi_lock); |
e29e175b | 1861 | #ifdef CONFIG_RT_MUTEXES |
a23ba907 | 1862 | p->pi_waiters = RB_ROOT_CACHED; |
e96a7705 | 1863 | p->pi_top_task = NULL; |
23f78d4a | 1864 | p->pi_blocked_on = NULL; |
23f78d4a IM |
1865 | #endif |
1866 | } | |
1867 | ||
2c470475 EB |
1868 | static inline void init_task_pid_links(struct task_struct *task) |
1869 | { | |
1870 | enum pid_type type; | |
1871 | ||
96e1e984 | 1872 | for (type = PIDTYPE_PID; type < PIDTYPE_MAX; ++type) |
2c470475 | 1873 | INIT_HLIST_NODE(&task->pid_links[type]); |
2c470475 EB |
1874 | } |
1875 | ||
81907739 ON |
1876 | static inline void |
1877 | init_task_pid(struct task_struct *task, enum pid_type type, struct pid *pid) | |
1878 | { | |
2c470475 EB |
1879 | if (type == PIDTYPE_PID) |
1880 | task->thread_pid = pid; | |
1881 | else | |
1882 | task->signal->pids[type] = pid; | |
81907739 ON |
1883 | } |
1884 | ||
6bfbaa51 IM |
1885 | static inline void rcu_copy_process(struct task_struct *p) |
1886 | { | |
1887 | #ifdef CONFIG_PREEMPT_RCU | |
1888 | p->rcu_read_lock_nesting = 0; | |
1889 | p->rcu_read_unlock_special.s = 0; | |
1890 | p->rcu_blocked_node = NULL; | |
1891 | INIT_LIST_HEAD(&p->rcu_node_entry); | |
1892 | #endif /* #ifdef CONFIG_PREEMPT_RCU */ | |
1893 | #ifdef CONFIG_TASKS_RCU | |
1894 | p->rcu_tasks_holdout = false; | |
1895 | INIT_LIST_HEAD(&p->rcu_tasks_holdout_list); | |
1896 | p->rcu_tasks_idle_cpu = -1; | |
1897 | #endif /* #ifdef CONFIG_TASKS_RCU */ | |
d5f177d3 PM |
1898 | #ifdef CONFIG_TASKS_TRACE_RCU |
1899 | p->trc_reader_nesting = 0; | |
276c4104 | 1900 | p->trc_reader_special.s = 0; |
d5f177d3 | 1901 | INIT_LIST_HEAD(&p->trc_holdout_list); |
434c9eef | 1902 | INIT_LIST_HEAD(&p->trc_blkd_node); |
d5f177d3 | 1903 | #endif /* #ifdef CONFIG_TASKS_TRACE_RCU */ |
6bfbaa51 IM |
1904 | } |
1905 | ||
3695eae5 CB |
1906 | struct pid *pidfd_pid(const struct file *file) |
1907 | { | |
1908 | if (file->f_op == &pidfd_fops) | |
1909 | return file->private_data; | |
1910 | ||
1911 | return ERR_PTR(-EBADF); | |
1912 | } | |
1913 | ||
b3e58382 CB |
1914 | static int pidfd_release(struct inode *inode, struct file *file) |
1915 | { | |
1916 | struct pid *pid = file->private_data; | |
1917 | ||
1918 | file->private_data = NULL; | |
1919 | put_pid(pid); | |
1920 | return 0; | |
1921 | } | |
1922 | ||
1923 | #ifdef CONFIG_PROC_FS | |
15d42eb2 CK |
1924 | /** |
1925 | * pidfd_show_fdinfo - print information about a pidfd | |
1926 | * @m: proc fdinfo file | |
1927 | * @f: file referencing a pidfd | |
1928 | * | |
1929 | * Pid: | |
1930 | * This function will print the pid that a given pidfd refers to in the | |
1931 | * pid namespace of the procfs instance. | |
1932 | * If the pid namespace of the process is not a descendant of the pid | |
1933 | * namespace of the procfs instance 0 will be shown as its pid. This is | |
1934 | * similar to calling getppid() on a process whose parent is outside of | |
1935 | * its pid namespace. | |
1936 | * | |
1937 | * NSpid: | |
1938 | * If pid namespaces are supported then this function will also print | |
1939 | * the pid of a given pidfd refers to for all descendant pid namespaces | |
1940 | * starting from the current pid namespace of the instance, i.e. the | |
1941 | * Pid field and the first entry in the NSpid field will be identical. | |
1942 | * If the pid namespace of the process is not a descendant of the pid | |
1943 | * namespace of the procfs instance 0 will be shown as its first NSpid | |
1944 | * entry and no others will be shown. | |
1945 | * Note that this differs from the Pid and NSpid fields in | |
1946 | * /proc/<pid>/status where Pid and NSpid are always shown relative to | |
1947 | * the pid namespace of the procfs instance. The difference becomes | |
1948 | * obvious when sending around a pidfd between pid namespaces from a | |
a8ca6b13 | 1949 | * different branch of the tree, i.e. where no ancestral relation is |
15d42eb2 CK |
1950 | * present between the pid namespaces: |
1951 | * - create two new pid namespaces ns1 and ns2 in the initial pid | |
1952 | * namespace (also take care to create new mount namespaces in the | |
1953 | * new pid namespace and mount procfs) | |
1954 | * - create a process with a pidfd in ns1 | |
1955 | * - send pidfd from ns1 to ns2 | |
1956 | * - read /proc/self/fdinfo/<pidfd> and observe that both Pid and NSpid | |
1957 | * have exactly one entry, which is 0 | |
1958 | */ | |
b3e58382 CB |
1959 | static void pidfd_show_fdinfo(struct seq_file *m, struct file *f) |
1960 | { | |
b3e58382 | 1961 | struct pid *pid = f->private_data; |
3d6d8da4 CB |
1962 | struct pid_namespace *ns; |
1963 | pid_t nr = -1; | |
15d42eb2 | 1964 | |
3d6d8da4 | 1965 | if (likely(pid_has_task(pid, PIDTYPE_PID))) { |
9d78edea | 1966 | ns = proc_pid_ns(file_inode(m->file)->i_sb); |
3d6d8da4 CB |
1967 | nr = pid_nr_ns(pid, ns); |
1968 | } | |
1969 | ||
1970 | seq_put_decimal_ll(m, "Pid:\t", nr); | |
b3e58382 | 1971 | |
15d42eb2 | 1972 | #ifdef CONFIG_PID_NS |
3d6d8da4 CB |
1973 | seq_put_decimal_ll(m, "\nNSpid:\t", nr); |
1974 | if (nr > 0) { | |
15d42eb2 | 1975 | int i; |
b3e58382 | 1976 | |
15d42eb2 CK |
1977 | /* If nr is non-zero it means that 'pid' is valid and that |
1978 | * ns, i.e. the pid namespace associated with the procfs | |
1979 | * instance, is in the pid namespace hierarchy of pid. | |
1980 | * Start at one below the already printed level. | |
1981 | */ | |
1982 | for (i = ns->level + 1; i <= pid->level; i++) | |
3d6d8da4 | 1983 | seq_put_decimal_ll(m, "\t", pid->numbers[i].nr); |
15d42eb2 CK |
1984 | } |
1985 | #endif | |
b3e58382 CB |
1986 | seq_putc(m, '\n'); |
1987 | } | |
1988 | #endif | |
1989 | ||
b53b0b9d JFG |
1990 | /* |
1991 | * Poll support for process exit notification. | |
1992 | */ | |
9e77716a | 1993 | static __poll_t pidfd_poll(struct file *file, struct poll_table_struct *pts) |
b53b0b9d | 1994 | { |
b53b0b9d | 1995 | struct pid *pid = file->private_data; |
9e77716a | 1996 | __poll_t poll_flags = 0; |
b53b0b9d JFG |
1997 | |
1998 | poll_wait(file, &pid->wait_pidfd, pts); | |
1999 | ||
b53b0b9d JFG |
2000 | /* |
2001 | * Inform pollers only when the whole thread group exits. | |
2002 | * If the thread group leader exits before all other threads in the | |
2003 | * group, then poll(2) should block, similar to the wait(2) family. | |
2004 | */ | |
38fd525a | 2005 | if (thread_group_exited(pid)) |
9e77716a | 2006 | poll_flags = EPOLLIN | EPOLLRDNORM; |
b53b0b9d JFG |
2007 | |
2008 | return poll_flags; | |
2009 | } | |
2010 | ||
b3e58382 CB |
2011 | const struct file_operations pidfd_fops = { |
2012 | .release = pidfd_release, | |
b53b0b9d | 2013 | .poll = pidfd_poll, |
b3e58382 CB |
2014 | #ifdef CONFIG_PROC_FS |
2015 | .show_fdinfo = pidfd_show_fdinfo, | |
2016 | #endif | |
2017 | }; | |
2018 | ||
c3f3ce04 AA |
2019 | static void __delayed_free_task(struct rcu_head *rhp) |
2020 | { | |
2021 | struct task_struct *tsk = container_of(rhp, struct task_struct, rcu); | |
2022 | ||
2023 | free_task(tsk); | |
2024 | } | |
2025 | ||
2026 | static __always_inline void delayed_free_task(struct task_struct *tsk) | |
2027 | { | |
2028 | if (IS_ENABLED(CONFIG_MEMCG)) | |
2029 | call_rcu(&tsk->rcu, __delayed_free_task); | |
2030 | else | |
2031 | free_task(tsk); | |
2032 | } | |
2033 | ||
67197a4f SB |
2034 | static void copy_oom_score_adj(u64 clone_flags, struct task_struct *tsk) |
2035 | { | |
2036 | /* Skip if kernel thread */ | |
2037 | if (!tsk->mm) | |
2038 | return; | |
2039 | ||
2040 | /* Skip if spawning a thread or using vfork */ | |
2041 | if ((clone_flags & (CLONE_VM | CLONE_THREAD | CLONE_VFORK)) != CLONE_VM) | |
2042 | return; | |
2043 | ||
2044 | /* We need to synchronize with __set_oom_adj */ | |
2045 | mutex_lock(&oom_adj_mutex); | |
2046 | set_bit(MMF_MULTIPROCESS, &tsk->mm->flags); | |
2047 | /* Update the values in case they were changed after copy_signal */ | |
2048 | tsk->signal->oom_score_adj = current->signal->oom_score_adj; | |
2049 | tsk->signal->oom_score_adj_min = current->signal->oom_score_adj_min; | |
2050 | mutex_unlock(&oom_adj_mutex); | |
2051 | } | |
2052 | ||
79257534 DBO |
2053 | #ifdef CONFIG_RV |
2054 | static void rv_task_fork(struct task_struct *p) | |
2055 | { | |
2056 | int i; | |
2057 | ||
2058 | for (i = 0; i < RV_PER_TASK_MONITORS; i++) | |
2059 | p->rv[i].da_mon.monitoring = false; | |
2060 | } | |
2061 | #else | |
2062 | #define rv_task_fork(p) do {} while (0) | |
2063 | #endif | |
2064 | ||
1da177e4 LT |
2065 | /* |
2066 | * This creates a new process as a copy of the old one, | |
2067 | * but does not actually start it yet. | |
2068 | * | |
2069 | * It copies the registers, and all the appropriate | |
2070 | * parts of the process environment (as per the clone | |
2071 | * flags). The actual kick-off is left to the caller. | |
2072 | */ | |
0766f788 | 2073 | static __latent_entropy struct task_struct *copy_process( |
09a05394 | 2074 | struct pid *pid, |
3033f14a | 2075 | int trace, |
7f192e3c CB |
2076 | int node, |
2077 | struct kernel_clone_args *args) | |
1da177e4 | 2078 | { |
b3e58382 | 2079 | int pidfd = -1, retval; |
a24efe62 | 2080 | struct task_struct *p; |
c3ad2c3b | 2081 | struct multiprocess_signals delayed; |
6fd2fe49 | 2082 | struct file *pidfile = NULL; |
c5febea0 | 2083 | const u64 clone_flags = args->flags; |
769071ac | 2084 | struct nsproxy *nsp = current->nsproxy; |
1da177e4 | 2085 | |
667b6094 MPS |
2086 | /* |
2087 | * Don't allow sharing the root directory with processes in a different | |
2088 | * namespace | |
2089 | */ | |
1da177e4 LT |
2090 | if ((clone_flags & (CLONE_NEWNS|CLONE_FS)) == (CLONE_NEWNS|CLONE_FS)) |
2091 | return ERR_PTR(-EINVAL); | |
2092 | ||
e66eded8 EB |
2093 | if ((clone_flags & (CLONE_NEWUSER|CLONE_FS)) == (CLONE_NEWUSER|CLONE_FS)) |
2094 | return ERR_PTR(-EINVAL); | |
2095 | ||
1da177e4 LT |
2096 | /* |
2097 | * Thread groups must share signals as well, and detached threads | |
2098 | * can only be started up within the thread group. | |
2099 | */ | |
2100 | if ((clone_flags & CLONE_THREAD) && !(clone_flags & CLONE_SIGHAND)) | |
2101 | return ERR_PTR(-EINVAL); | |
2102 | ||
2103 | /* | |
2104 | * Shared signal handlers imply shared VM. By way of the above, | |
2105 | * thread groups also imply shared VM. Blocking this case allows | |
2106 | * for various simplifications in other code. | |
2107 | */ | |
2108 | if ((clone_flags & CLONE_SIGHAND) && !(clone_flags & CLONE_VM)) | |
2109 | return ERR_PTR(-EINVAL); | |
2110 | ||
123be07b SB |
2111 | /* |
2112 | * Siblings of global init remain as zombies on exit since they are | |
2113 | * not reaped by their parent (swapper). To solve this and to avoid | |
2114 | * multi-rooted process trees, prevent global and container-inits | |
2115 | * from creating siblings. | |
2116 | */ | |
2117 | if ((clone_flags & CLONE_PARENT) && | |
2118 | current->signal->flags & SIGNAL_UNKILLABLE) | |
2119 | return ERR_PTR(-EINVAL); | |
2120 | ||
8382fcac | 2121 | /* |
40a0d32d | 2122 | * If the new process will be in a different pid or user namespace |
faf00da5 | 2123 | * do not allow it to share a thread group with the forking task. |
8382fcac | 2124 | */ |
faf00da5 | 2125 | if (clone_flags & CLONE_THREAD) { |
40a0d32d | 2126 | if ((clone_flags & (CLONE_NEWUSER | CLONE_NEWPID)) || |
769071ac AV |
2127 | (task_active_pid_ns(current) != nsp->pid_ns_for_children)) |
2128 | return ERR_PTR(-EINVAL); | |
2129 | } | |
2130 | ||
b3e58382 | 2131 | if (clone_flags & CLONE_PIDFD) { |
b3e58382 | 2132 | /* |
b3e58382 CB |
2133 | * - CLONE_DETACHED is blocked so that we can potentially |
2134 | * reuse it later for CLONE_PIDFD. | |
2135 | * - CLONE_THREAD is blocked until someone really needs it. | |
2136 | */ | |
7f192e3c | 2137 | if (clone_flags & (CLONE_DETACHED | CLONE_THREAD)) |
b3e58382 | 2138 | return ERR_PTR(-EINVAL); |
b3e58382 CB |
2139 | } |
2140 | ||
c3ad2c3b EB |
2141 | /* |
2142 | * Force any signals received before this point to be delivered | |
2143 | * before the fork happens. Collect up signals sent to multiple | |
2144 | * processes that happen during the fork and delay them so that | |
2145 | * they appear to happen after the fork. | |
2146 | */ | |
2147 | sigemptyset(&delayed.signal); | |
2148 | INIT_HLIST_NODE(&delayed.node); | |
2149 | ||
2150 | spin_lock_irq(¤t->sighand->siglock); | |
2151 | if (!(clone_flags & CLONE_THREAD)) | |
2152 | hlist_add_head(&delayed.node, ¤t->signal->multiprocess); | |
2153 | recalc_sigpending(); | |
2154 | spin_unlock_irq(¤t->sighand->siglock); | |
2155 | retval = -ERESTARTNOINTR; | |
66ae0d1e | 2156 | if (task_sigpending(current)) |
c3ad2c3b EB |
2157 | goto fork_out; |
2158 | ||
1da177e4 | 2159 | retval = -ENOMEM; |
725fc629 | 2160 | p = dup_task_struct(current, node); |
1da177e4 LT |
2161 | if (!p) |
2162 | goto fork_out; | |
753550eb EB |
2163 | p->flags &= ~PF_KTHREAD; |
2164 | if (args->kthread) | |
2165 | p->flags |= PF_KTHREAD; | |
b16b3855 JA |
2166 | if (args->io_thread) { |
2167 | /* | |
2168 | * Mark us an IO worker, and block any signal that isn't | |
2169 | * fatal or STOP | |
2170 | */ | |
cc440e87 | 2171 | p->flags |= PF_IO_WORKER; |
b16b3855 JA |
2172 | siginitsetinv(&p->blocked, sigmask(SIGKILL)|sigmask(SIGSTOP)); |
2173 | } | |
1da177e4 | 2174 | |
7f192e3c | 2175 | p->set_child_tid = (clone_flags & CLONE_CHILD_SETTID) ? args->child_tid : NULL; |
4d6501dc VN |
2176 | /* |
2177 | * Clear TID on mm_release()? | |
2178 | */ | |
7f192e3c | 2179 | p->clear_child_tid = (clone_flags & CLONE_CHILD_CLEARTID) ? args->child_tid : NULL; |
4d6501dc | 2180 | |
f7e8b616 SR |
2181 | ftrace_graph_init_task(p); |
2182 | ||
bea493a0 PZ |
2183 | rt_mutex_init_task(p); |
2184 | ||
a21ee605 | 2185 | lockdep_assert_irqs_enabled(); |
d12c1a37 | 2186 | #ifdef CONFIG_PROVE_LOCKING |
de30a2b3 IM |
2187 | DEBUG_LOCKS_WARN_ON(!p->softirqs_enabled); |
2188 | #endif | |
8f2f9c4d EB |
2189 | retval = copy_creds(p, clone_flags); |
2190 | if (retval < 0) | |
2191 | goto bad_fork_free; | |
2192 | ||
1da177e4 | 2193 | retval = -EAGAIN; |
de399236 | 2194 | if (is_rlimit_overlimit(task_ucounts(p), UCOUNT_RLIMIT_NPROC, rlimit(RLIMIT_NPROC))) { |
b57922b6 EP |
2195 | if (p->real_cred->user != INIT_USER && |
2196 | !capable(CAP_SYS_RESOURCE) && !capable(CAP_SYS_ADMIN)) | |
8f2f9c4d | 2197 | goto bad_fork_cleanup_count; |
1da177e4 | 2198 | } |
72fa5997 | 2199 | current->flags &= ~PF_NPROC_EXCEEDED; |
1da177e4 | 2200 | |
1da177e4 LT |
2201 | /* |
2202 | * If multiple threads are within copy_process(), then this check | |
2203 | * triggers too late. This doesn't hurt, the check is only there | |
2204 | * to stop root fork bombs. | |
2205 | */ | |
04ec93fe | 2206 | retval = -EAGAIN; |
c17d1a3a | 2207 | if (data_race(nr_threads >= max_threads)) |
1da177e4 LT |
2208 | goto bad_fork_cleanup_count; |
2209 | ||
ca74e92b | 2210 | delayacct_tsk_init(p); /* Must remain after dup_task_struct() */ |
a8ea6fc9 | 2211 | p->flags &= ~(PF_SUPERPRIV | PF_WQ_WORKER | PF_IDLE | PF_NO_SETAFFINITY); |
514ddb44 | 2212 | p->flags |= PF_FORKNOEXEC; |
1da177e4 LT |
2213 | INIT_LIST_HEAD(&p->children); |
2214 | INIT_LIST_HEAD(&p->sibling); | |
f41d911f | 2215 | rcu_copy_process(p); |
1da177e4 LT |
2216 | p->vfork_done = NULL; |
2217 | spin_lock_init(&p->alloc_lock); | |
1da177e4 | 2218 | |
1da177e4 LT |
2219 | init_sigpending(&p->pending); |
2220 | ||
64861634 | 2221 | p->utime = p->stime = p->gtime = 0; |
40565b5a | 2222 | #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME |
64861634 | 2223 | p->utimescaled = p->stimescaled = 0; |
40565b5a | 2224 | #endif |
9d7fb042 PZ |
2225 | prev_cputime_init(&p->prev_cputime); |
2226 | ||
6a61671b | 2227 | #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN |
bac5b6b6 FW |
2228 | seqcount_init(&p->vtime.seqcount); |
2229 | p->vtime.starttime = 0; | |
2230 | p->vtime.state = VTIME_INACTIVE; | |
6a61671b FW |
2231 | #endif |
2232 | ||
0f212204 JA |
2233 | #ifdef CONFIG_IO_URING |
2234 | p->io_uring = NULL; | |
2235 | #endif | |
2236 | ||
a3a2e76c KH |
2237 | #if defined(SPLIT_RSS_COUNTING) |
2238 | memset(&p->rss_stat, 0, sizeof(p->rss_stat)); | |
2239 | #endif | |
172ba844 | 2240 | |
6976675d AV |
2241 | p->default_timer_slack_ns = current->timer_slack_ns; |
2242 | ||
eb414681 JW |
2243 | #ifdef CONFIG_PSI |
2244 | p->psi_flags = 0; | |
2245 | #endif | |
2246 | ||
5995477a | 2247 | task_io_accounting_init(&p->ioac); |
1da177e4 LT |
2248 | acct_clear_integrals(p); |
2249 | ||
3a245c0f | 2250 | posix_cputimers_init(&p->posix_cputimers); |
1da177e4 | 2251 | |
1da177e4 | 2252 | p->io_context = NULL; |
c0b0ae8a | 2253 | audit_set_context(p, NULL); |
b4f48b63 | 2254 | cgroup_fork(p); |
343f4c49 | 2255 | if (args->kthread) { |
40966e31 | 2256 | if (!set_kthread_struct(p)) |
ff8288ff | 2257 | goto bad_fork_cleanup_delayacct; |
40966e31 | 2258 | } |
1da177e4 | 2259 | #ifdef CONFIG_NUMA |
846a16bf | 2260 | p->mempolicy = mpol_dup(p->mempolicy); |
fb0a685c DRO |
2261 | if (IS_ERR(p->mempolicy)) { |
2262 | retval = PTR_ERR(p->mempolicy); | |
2263 | p->mempolicy = NULL; | |
ff8288ff | 2264 | goto bad_fork_cleanup_delayacct; |
fb0a685c | 2265 | } |
1da177e4 | 2266 | #endif |
778d3b0f MH |
2267 | #ifdef CONFIG_CPUSETS |
2268 | p->cpuset_mem_spread_rotor = NUMA_NO_NODE; | |
2269 | p->cpuset_slab_spread_rotor = NUMA_NO_NODE; | |
b7505861 | 2270 | seqcount_spinlock_init(&p->mems_allowed_seq, &p->alloc_lock); |
778d3b0f | 2271 | #endif |
de30a2b3 | 2272 | #ifdef CONFIG_TRACE_IRQFLAGS |
0584df9c ME |
2273 | memset(&p->irqtrace, 0, sizeof(p->irqtrace)); |
2274 | p->irqtrace.hardirq_disable_ip = _THIS_IP_; | |
2275 | p->irqtrace.softirq_enable_ip = _THIS_IP_; | |
2276 | p->softirqs_enabled = 1; | |
2277 | p->softirq_context = 0; | |
de30a2b3 | 2278 | #endif |
8bcbde54 DH |
2279 | |
2280 | p->pagefault_disabled = 0; | |
2281 | ||
fbb9ce95 | 2282 | #ifdef CONFIG_LOCKDEP |
b09be676 | 2283 | lockdep_init_task(p); |
fbb9ce95 | 2284 | #endif |
1da177e4 | 2285 | |
408894ee IM |
2286 | #ifdef CONFIG_DEBUG_MUTEXES |
2287 | p->blocked_on = NULL; /* not blocked yet */ | |
2288 | #endif | |
cafe5635 KO |
2289 | #ifdef CONFIG_BCACHE |
2290 | p->sequential_io = 0; | |
2291 | p->sequential_io_avg = 0; | |
2292 | #endif | |
a10787e6 SL |
2293 | #ifdef CONFIG_BPF_SYSCALL |
2294 | RCU_INIT_POINTER(p->bpf_storage, NULL); | |
c7603cfa | 2295 | p->bpf_ctx = NULL; |
a10787e6 | 2296 | #endif |
0f481406 | 2297 | |
3c90e6e9 | 2298 | /* Perform scheduler related setup. Assign this task to a CPU. */ |
aab03e05 DF |
2299 | retval = sched_fork(clone_flags, p); |
2300 | if (retval) | |
2301 | goto bad_fork_cleanup_policy; | |
6ab423e0 | 2302 | |
2b26f0aa | 2303 | retval = perf_event_init_task(p, clone_flags); |
6ab423e0 PZ |
2304 | if (retval) |
2305 | goto bad_fork_cleanup_policy; | |
fb0a685c DRO |
2306 | retval = audit_alloc(p); |
2307 | if (retval) | |
6c72e350 | 2308 | goto bad_fork_cleanup_perf; |
1da177e4 | 2309 | /* copy all the process information */ |
ab602f79 | 2310 | shm_init_task(p); |
e4e55b47 | 2311 | retval = security_task_alloc(p, clone_flags); |
fb0a685c | 2312 | if (retval) |
1da177e4 | 2313 | goto bad_fork_cleanup_audit; |
e4e55b47 TH |
2314 | retval = copy_semundo(clone_flags, p); |
2315 | if (retval) | |
2316 | goto bad_fork_cleanup_security; | |
fb0a685c DRO |
2317 | retval = copy_files(clone_flags, p); |
2318 | if (retval) | |
1da177e4 | 2319 | goto bad_fork_cleanup_semundo; |
fb0a685c DRO |
2320 | retval = copy_fs(clone_flags, p); |
2321 | if (retval) | |
1da177e4 | 2322 | goto bad_fork_cleanup_files; |
fb0a685c DRO |
2323 | retval = copy_sighand(clone_flags, p); |
2324 | if (retval) | |
1da177e4 | 2325 | goto bad_fork_cleanup_fs; |
fb0a685c DRO |
2326 | retval = copy_signal(clone_flags, p); |
2327 | if (retval) | |
1da177e4 | 2328 | goto bad_fork_cleanup_sighand; |
fb0a685c DRO |
2329 | retval = copy_mm(clone_flags, p); |
2330 | if (retval) | |
1da177e4 | 2331 | goto bad_fork_cleanup_signal; |
fb0a685c DRO |
2332 | retval = copy_namespaces(clone_flags, p); |
2333 | if (retval) | |
d84f4f99 | 2334 | goto bad_fork_cleanup_mm; |
fb0a685c DRO |
2335 | retval = copy_io(clone_flags, p); |
2336 | if (retval) | |
fd0928df | 2337 | goto bad_fork_cleanup_namespaces; |
c5febea0 | 2338 | retval = copy_thread(p, args); |
1da177e4 | 2339 | if (retval) |
fd0928df | 2340 | goto bad_fork_cleanup_io; |
1da177e4 | 2341 | |
afaef01c AP |
2342 | stackleak_task_init(p); |
2343 | ||
425fb2b4 | 2344 | if (pid != &init_struct_pid) { |
49cb2fc4 AR |
2345 | pid = alloc_pid(p->nsproxy->pid_ns_for_children, args->set_tid, |
2346 | args->set_tid_size); | |
35f71bc0 MH |
2347 | if (IS_ERR(pid)) { |
2348 | retval = PTR_ERR(pid); | |
0740aa5f | 2349 | goto bad_fork_cleanup_thread; |
35f71bc0 | 2350 | } |
425fb2b4 PE |
2351 | } |
2352 | ||
b3e58382 CB |
2353 | /* |
2354 | * This has to happen after we've potentially unshared the file | |
2355 | * descriptor table (so that the pidfd doesn't leak into the child | |
2356 | * if the fd table isn't shared). | |
2357 | */ | |
2358 | if (clone_flags & CLONE_PIDFD) { | |
6fd2fe49 | 2359 | retval = get_unused_fd_flags(O_RDWR | O_CLOEXEC); |
b3e58382 CB |
2360 | if (retval < 0) |
2361 | goto bad_fork_free_pid; | |
2362 | ||
2363 | pidfd = retval; | |
6fd2fe49 AV |
2364 | |
2365 | pidfile = anon_inode_getfile("[pidfd]", &pidfd_fops, pid, | |
2366 | O_RDWR | O_CLOEXEC); | |
2367 | if (IS_ERR(pidfile)) { | |
2368 | put_unused_fd(pidfd); | |
28dd29c0 | 2369 | retval = PTR_ERR(pidfile); |
6fd2fe49 AV |
2370 | goto bad_fork_free_pid; |
2371 | } | |
2372 | get_pid(pid); /* held by pidfile now */ | |
2373 | ||
7f192e3c | 2374 | retval = put_user(pidfd, args->pidfd); |
b3e58382 CB |
2375 | if (retval) |
2376 | goto bad_fork_put_pidfd; | |
2377 | } | |
2378 | ||
73c10101 JA |
2379 | #ifdef CONFIG_BLOCK |
2380 | p->plug = NULL; | |
2381 | #endif | |
ba31c1a4 TG |
2382 | futex_init_task(p); |
2383 | ||
f9a3879a GM |
2384 | /* |
2385 | * sigaltstack should be cleared when sharing the same VM | |
2386 | */ | |
2387 | if ((clone_flags & (CLONE_VM|CLONE_VFORK)) == CLONE_VM) | |
2a742138 | 2388 | sas_ss_reset(p); |
f9a3879a | 2389 | |
1da177e4 | 2390 | /* |
6580807d ON |
2391 | * Syscall tracing and stepping should be turned off in the |
2392 | * child regardless of CLONE_PTRACE. | |
1da177e4 | 2393 | */ |
6580807d | 2394 | user_disable_single_step(p); |
64c19ba2 | 2395 | clear_task_syscall_work(p, SYSCALL_TRACE); |
64eb35f7 GKB |
2396 | #if defined(CONFIG_GENERIC_ENTRY) || defined(TIF_SYSCALL_EMU) |
2397 | clear_task_syscall_work(p, SYSCALL_EMU); | |
ed75e8d5 | 2398 | #endif |
e02c9b0d | 2399 | clear_tsk_latency_tracing(p); |
1da177e4 | 2400 | |
1da177e4 | 2401 | /* ok, now we should be set up.. */ |
18c830df ON |
2402 | p->pid = pid_nr(pid); |
2403 | if (clone_flags & CLONE_THREAD) { | |
18c830df ON |
2404 | p->group_leader = current->group_leader; |
2405 | p->tgid = current->tgid; | |
2406 | } else { | |
18c830df ON |
2407 | p->group_leader = p; |
2408 | p->tgid = p->pid; | |
2409 | } | |
5f8aadd8 | 2410 | |
9d823e8f WF |
2411 | p->nr_dirtied = 0; |
2412 | p->nr_dirtied_pause = 128 >> (PAGE_SHIFT - 10); | |
83712358 | 2413 | p->dirty_paused_when = 0; |
9d823e8f | 2414 | |
bb8cbbfe | 2415 | p->pdeath_signal = 0; |
47e65328 | 2416 | INIT_LIST_HEAD(&p->thread_group); |
158e1645 | 2417 | p->task_works = NULL; |
ca7752ca | 2418 | clear_posix_cputimers_work(p); |
1da177e4 | 2419 | |
d741bf41 PZ |
2420 | #ifdef CONFIG_KRETPROBES |
2421 | p->kretprobe_instances.first = NULL; | |
2422 | #endif | |
54ecbe6f MH |
2423 | #ifdef CONFIG_RETHOOK |
2424 | p->rethooks.first = NULL; | |
2425 | #endif | |
d741bf41 | 2426 | |
7e47682e AS |
2427 | /* |
2428 | * Ensure that the cgroup subsystem policies allow the new process to be | |
7b7b8a2c | 2429 | * forked. It should be noted that the new process's css_set can be changed |
7e47682e AS |
2430 | * between here and cgroup_post_fork() if an organisation operation is in |
2431 | * progress. | |
2432 | */ | |
ef2c41cf | 2433 | retval = cgroup_can_fork(p, args); |
7e47682e | 2434 | if (retval) |
5a5cf5cb | 2435 | goto bad_fork_put_pidfd; |
7e47682e | 2436 | |
b1e82065 PZ |
2437 | /* |
2438 | * Now that the cgroups are pinned, re-clone the parent cgroup and put | |
2439 | * the new task on the correct runqueue. All this *before* the task | |
2440 | * becomes visible. | |
2441 | * | |
2442 | * This isn't part of ->can_fork() because while the re-cloning is | |
2443 | * cgroup specific, it unconditionally needs to place the task on a | |
2444 | * runqueue. | |
2445 | */ | |
2446 | sched_cgroup_fork(p, args); | |
2447 | ||
7b558513 DH |
2448 | /* |
2449 | * From this point on we must avoid any synchronous user-space | |
2450 | * communication until we take the tasklist-lock. In particular, we do | |
2451 | * not want user-space to be able to predict the process start-time by | |
2452 | * stalling fork(2) after we recorded the start_time but before it is | |
2453 | * visible to the system. | |
2454 | */ | |
2455 | ||
2456 | p->start_time = ktime_get_ns(); | |
cf25e24d | 2457 | p->start_boottime = ktime_get_boottime_ns(); |
7b558513 | 2458 | |
18c830df ON |
2459 | /* |
2460 | * Make it visible to the rest of the system, but dont wake it up yet. | |
2461 | * Need tasklist lock for parent etc handling! | |
2462 | */ | |
1da177e4 LT |
2463 | write_lock_irq(&tasklist_lock); |
2464 | ||
1da177e4 | 2465 | /* CLONE_PARENT re-uses the old parent */ |
2d5516cb | 2466 | if (clone_flags & (CLONE_PARENT|CLONE_THREAD)) { |
1da177e4 | 2467 | p->real_parent = current->real_parent; |
2d5516cb | 2468 | p->parent_exec_id = current->parent_exec_id; |
b4e00444 EW |
2469 | if (clone_flags & CLONE_THREAD) |
2470 | p->exit_signal = -1; | |
2471 | else | |
2472 | p->exit_signal = current->group_leader->exit_signal; | |
2d5516cb | 2473 | } else { |
1da177e4 | 2474 | p->real_parent = current; |
2d5516cb | 2475 | p->parent_exec_id = current->self_exec_id; |
b4e00444 | 2476 | p->exit_signal = args->exit_signal; |
2d5516cb | 2477 | } |
1da177e4 | 2478 | |
d83a7cb3 JP |
2479 | klp_copy_process(p); |
2480 | ||
85dd3f61 PZ |
2481 | sched_core_fork(p); |
2482 | ||
3f17da69 | 2483 | spin_lock(¤t->sighand->siglock); |
4a2c7a78 | 2484 | |
79257534 DBO |
2485 | rv_task_fork(p); |
2486 | ||
d7822b1e MD |
2487 | rseq_fork(p, clone_flags); |
2488 | ||
4ca1d3ee | 2489 | /* Don't start children in a dying pid namespace */ |
e8cfbc24 | 2490 | if (unlikely(!(ns_of_pid(pid)->pid_allocated & PIDNS_ADDING))) { |
3fd37226 KT |
2491 | retval = -ENOMEM; |
2492 | goto bad_fork_cancel_cgroup; | |
2493 | } | |
4a2c7a78 | 2494 | |
7673bf55 EB |
2495 | /* Let kill terminate clone/fork in the middle */ |
2496 | if (fatal_signal_pending(current)) { | |
2497 | retval = -EINTR; | |
2498 | goto bad_fork_cancel_cgroup; | |
2499 | } | |
2500 | ||
a1140cb2 KI |
2501 | /* No more failure paths after this point. */ |
2502 | ||
2503 | /* | |
2504 | * Copy seccomp details explicitly here, in case they were changed | |
2505 | * before holding sighand lock. | |
2506 | */ | |
2507 | copy_seccomp(p); | |
2508 | ||
2c470475 | 2509 | init_task_pid_links(p); |
73b9ebfe | 2510 | if (likely(p->pid)) { |
4b9d33e6 | 2511 | ptrace_init_task(p, (clone_flags & CLONE_PTRACE) || trace); |
73b9ebfe | 2512 | |
81907739 | 2513 | init_task_pid(p, PIDTYPE_PID, pid); |
73b9ebfe | 2514 | if (thread_group_leader(p)) { |
6883f81a | 2515 | init_task_pid(p, PIDTYPE_TGID, pid); |
81907739 ON |
2516 | init_task_pid(p, PIDTYPE_PGID, task_pgrp(current)); |
2517 | init_task_pid(p, PIDTYPE_SID, task_session(current)); | |
2518 | ||
1c4042c2 | 2519 | if (is_child_reaper(pid)) { |
17cf22c3 | 2520 | ns_of_pid(pid)->child_reaper = p; |
1c4042c2 EB |
2521 | p->signal->flags |= SIGNAL_UNKILLABLE; |
2522 | } | |
c3ad2c3b | 2523 | p->signal->shared_pending.signal = delayed.signal; |
9c9f4ded | 2524 | p->signal->tty = tty_kref_get(current->signal->tty); |
749860ce PT |
2525 | /* |
2526 | * Inherit has_child_subreaper flag under the same | |
2527 | * tasklist_lock with adding child to the process tree | |
2528 | * for propagate_has_child_subreaper optimization. | |
2529 | */ | |
2530 | p->signal->has_child_subreaper = p->real_parent->signal->has_child_subreaper || | |
2531 | p->real_parent->signal->is_child_subreaper; | |
9cd80bbb | 2532 | list_add_tail(&p->sibling, &p->real_parent->children); |
5e85d4ab | 2533 | list_add_tail_rcu(&p->tasks, &init_task.tasks); |
6883f81a | 2534 | attach_pid(p, PIDTYPE_TGID); |
81907739 ON |
2535 | attach_pid(p, PIDTYPE_PGID); |
2536 | attach_pid(p, PIDTYPE_SID); | |
909ea964 | 2537 | __this_cpu_inc(process_counts); |
80628ca0 ON |
2538 | } else { |
2539 | current->signal->nr_threads++; | |
d80f7d7b | 2540 | current->signal->quick_threads++; |
80628ca0 | 2541 | atomic_inc(¤t->signal->live); |
60d4de3f | 2542 | refcount_inc(¤t->signal->sigcnt); |
924de3b8 | 2543 | task_join_group_stop(p); |
80628ca0 ON |
2544 | list_add_tail_rcu(&p->thread_group, |
2545 | &p->group_leader->thread_group); | |
0c740d0a ON |
2546 | list_add_tail_rcu(&p->thread_node, |
2547 | &p->signal->thread_head); | |
73b9ebfe | 2548 | } |
81907739 | 2549 | attach_pid(p, PIDTYPE_PID); |
73b9ebfe | 2550 | nr_threads++; |
1da177e4 | 2551 | } |
1da177e4 | 2552 | total_forks++; |
c3ad2c3b | 2553 | hlist_del_init(&delayed.node); |
3f17da69 | 2554 | spin_unlock(¤t->sighand->siglock); |
4af4206b | 2555 | syscall_tracepoint_update(p); |
1da177e4 | 2556 | write_unlock_irq(&tasklist_lock); |
4af4206b | 2557 | |
ddc204b5 WL |
2558 | if (pidfile) |
2559 | fd_install(pidfd, pidfile); | |
2560 | ||
c13cf856 | 2561 | proc_fork_connector(p); |
b1e82065 | 2562 | sched_post_fork(p); |
ef2c41cf | 2563 | cgroup_post_fork(p, args); |
cdd6c482 | 2564 | perf_event_fork(p); |
43d2b113 KH |
2565 | |
2566 | trace_task_newtask(p, clone_flags); | |
3ab67966 | 2567 | uprobe_copy_process(p, clone_flags); |
43d2b113 | 2568 | |
67197a4f SB |
2569 | copy_oom_score_adj(clone_flags, p); |
2570 | ||
1da177e4 LT |
2571 | return p; |
2572 | ||
7e47682e | 2573 | bad_fork_cancel_cgroup: |
85dd3f61 | 2574 | sched_core_free(p); |
3fd37226 KT |
2575 | spin_unlock(¤t->sighand->siglock); |
2576 | write_unlock_irq(&tasklist_lock); | |
ef2c41cf | 2577 | cgroup_cancel_fork(p, args); |
b3e58382 | 2578 | bad_fork_put_pidfd: |
6fd2fe49 AV |
2579 | if (clone_flags & CLONE_PIDFD) { |
2580 | fput(pidfile); | |
2581 | put_unused_fd(pidfd); | |
2582 | } | |
425fb2b4 PE |
2583 | bad_fork_free_pid: |
2584 | if (pid != &init_struct_pid) | |
2585 | free_pid(pid); | |
0740aa5f JS |
2586 | bad_fork_cleanup_thread: |
2587 | exit_thread(p); | |
fd0928df | 2588 | bad_fork_cleanup_io: |
b69f2292 LR |
2589 | if (p->io_context) |
2590 | exit_io_context(p); | |
ab516013 | 2591 | bad_fork_cleanup_namespaces: |
444f378b | 2592 | exit_task_namespaces(p); |
1da177e4 | 2593 | bad_fork_cleanup_mm: |
c3f3ce04 AA |
2594 | if (p->mm) { |
2595 | mm_clear_owner(p->mm, p); | |
1da177e4 | 2596 | mmput(p->mm); |
c3f3ce04 | 2597 | } |
1da177e4 | 2598 | bad_fork_cleanup_signal: |
4ab6c083 | 2599 | if (!(clone_flags & CLONE_THREAD)) |
1c5354de | 2600 | free_signal_struct(p->signal); |
1da177e4 | 2601 | bad_fork_cleanup_sighand: |
a7e5328a | 2602 | __cleanup_sighand(p->sighand); |
1da177e4 LT |
2603 | bad_fork_cleanup_fs: |
2604 | exit_fs(p); /* blocking */ | |
2605 | bad_fork_cleanup_files: | |
2606 | exit_files(p); /* blocking */ | |
2607 | bad_fork_cleanup_semundo: | |
2608 | exit_sem(p); | |
e4e55b47 TH |
2609 | bad_fork_cleanup_security: |
2610 | security_task_free(p); | |
1da177e4 LT |
2611 | bad_fork_cleanup_audit: |
2612 | audit_free(p); | |
6c72e350 | 2613 | bad_fork_cleanup_perf: |
cdd6c482 | 2614 | perf_event_free_task(p); |
6c72e350 | 2615 | bad_fork_cleanup_policy: |
b09be676 | 2616 | lockdep_free_task(p); |
1da177e4 | 2617 | #ifdef CONFIG_NUMA |
f0be3d32 | 2618 | mpol_put(p->mempolicy); |
1da177e4 | 2619 | #endif |
ff8288ff | 2620 | bad_fork_cleanup_delayacct: |
35df17c5 | 2621 | delayacct_tsk_free(p); |
1da177e4 | 2622 | bad_fork_cleanup_count: |
21d1c5e3 | 2623 | dec_rlimit_ucounts(task_ucounts(p), UCOUNT_RLIMIT_NPROC, 1); |
e0e81739 | 2624 | exit_creds(p); |
1da177e4 | 2625 | bad_fork_free: |
2f064a59 | 2626 | WRITE_ONCE(p->__state, TASK_DEAD); |
1a03d3f1 | 2627 | exit_task_stack_account(p); |
68f24b08 | 2628 | put_task_stack(p); |
c3f3ce04 | 2629 | delayed_free_task(p); |
fe7d37d1 | 2630 | fork_out: |
c3ad2c3b EB |
2631 | spin_lock_irq(¤t->sighand->siglock); |
2632 | hlist_del_init(&delayed.node); | |
2633 | spin_unlock_irq(¤t->sighand->siglock); | |
fe7d37d1 | 2634 | return ERR_PTR(retval); |
1da177e4 LT |
2635 | } |
2636 | ||
2c470475 | 2637 | static inline void init_idle_pids(struct task_struct *idle) |
f106eee1 ON |
2638 | { |
2639 | enum pid_type type; | |
2640 | ||
2641 | for (type = PIDTYPE_PID; type < PIDTYPE_MAX; ++type) { | |
2c470475 EB |
2642 | INIT_HLIST_NODE(&idle->pid_links[type]); /* not really needed */ |
2643 | init_task_pid(idle, type, &init_struct_pid); | |
f106eee1 ON |
2644 | } |
2645 | } | |
2646 | ||
36cb0e1c EB |
2647 | static int idle_dummy(void *dummy) |
2648 | { | |
2649 | /* This function is never called */ | |
2650 | return 0; | |
2651 | } | |
2652 | ||
f1a0a376 | 2653 | struct task_struct * __init fork_idle(int cpu) |
1da177e4 | 2654 | { |
36c8b586 | 2655 | struct task_struct *task; |
7f192e3c | 2656 | struct kernel_clone_args args = { |
343f4c49 | 2657 | .flags = CLONE_VM, |
5bd2e97c EB |
2658 | .fn = &idle_dummy, |
2659 | .fn_arg = NULL, | |
343f4c49 | 2660 | .kthread = 1, |
36cb0e1c | 2661 | .idle = 1, |
7f192e3c CB |
2662 | }; |
2663 | ||
2664 | task = copy_process(&init_struct_pid, 0, cpu_to_node(cpu), &args); | |
f106eee1 | 2665 | if (!IS_ERR(task)) { |
2c470475 | 2666 | init_idle_pids(task); |
753ca4f3 | 2667 | init_idle(task, cpu); |
f106eee1 | 2668 | } |
73b9ebfe | 2669 | |
1da177e4 LT |
2670 | return task; |
2671 | } | |
2672 | ||
cc440e87 JA |
2673 | /* |
2674 | * This is like kernel_clone(), but shaved down and tailored to just | |
2675 | * creating io_uring workers. It returns a created task, or an error pointer. | |
2676 | * The returned task is inactive, and the caller must fire it up through | |
2677 | * wake_up_new_task(p). All signals are blocked in the created task. | |
2678 | */ | |
2679 | struct task_struct *create_io_thread(int (*fn)(void *), void *arg, int node) | |
2680 | { | |
2681 | unsigned long flags = CLONE_FS|CLONE_FILES|CLONE_SIGHAND|CLONE_THREAD| | |
2682 | CLONE_IO; | |
2683 | struct kernel_clone_args args = { | |
2684 | .flags = ((lower_32_bits(flags) | CLONE_VM | | |
2685 | CLONE_UNTRACED) & ~CSIGNAL), | |
2686 | .exit_signal = (lower_32_bits(flags) & CSIGNAL), | |
5bd2e97c EB |
2687 | .fn = fn, |
2688 | .fn_arg = arg, | |
cc440e87 JA |
2689 | .io_thread = 1, |
2690 | }; | |
cc440e87 | 2691 | |
b16b3855 | 2692 | return copy_process(NULL, 0, node, &args); |
cc440e87 JA |
2693 | } |
2694 | ||
1da177e4 LT |
2695 | /* |
2696 | * Ok, this is the main fork-routine. | |
2697 | * | |
2698 | * It copies the process, and if successful kick-starts | |
2699 | * it and waits for it to finish using the VM if required. | |
a0eb9abd ES |
2700 | * |
2701 | * args->exit_signal is expected to be checked for sanity by the caller. | |
1da177e4 | 2702 | */ |
cad6967a | 2703 | pid_t kernel_clone(struct kernel_clone_args *args) |
1da177e4 | 2704 | { |
7f192e3c | 2705 | u64 clone_flags = args->flags; |
9f5325aa MPS |
2706 | struct completion vfork; |
2707 | struct pid *pid; | |
1da177e4 LT |
2708 | struct task_struct *p; |
2709 | int trace = 0; | |
cad6967a | 2710 | pid_t nr; |
1da177e4 | 2711 | |
3af8588c CB |
2712 | /* |
2713 | * For legacy clone() calls, CLONE_PIDFD uses the parent_tid argument | |
2714 | * to return the pidfd. Hence, CLONE_PIDFD and CLONE_PARENT_SETTID are | |
2715 | * mutually exclusive. With clone3() CLONE_PIDFD has grown a separate | |
2716 | * field in struct clone_args and it still doesn't make sense to have | |
2717 | * them both point at the same memory location. Performing this check | |
2718 | * here has the advantage that we don't need to have a separate helper | |
2719 | * to check for legacy clone(). | |
2720 | */ | |
2721 | if ((args->flags & CLONE_PIDFD) && | |
2722 | (args->flags & CLONE_PARENT_SETTID) && | |
2723 | (args->pidfd == args->parent_tid)) | |
2724 | return -EINVAL; | |
2725 | ||
09a05394 | 2726 | /* |
4b9d33e6 TH |
2727 | * Determine whether and which event to report to ptracer. When |
2728 | * called from kernel_thread or CLONE_UNTRACED is explicitly | |
2729 | * requested, no event is reported; otherwise, report if the event | |
2730 | * for the type of forking is enabled. | |
09a05394 | 2731 | */ |
e80d6661 | 2732 | if (!(clone_flags & CLONE_UNTRACED)) { |
4b9d33e6 TH |
2733 | if (clone_flags & CLONE_VFORK) |
2734 | trace = PTRACE_EVENT_VFORK; | |
7f192e3c | 2735 | else if (args->exit_signal != SIGCHLD) |
4b9d33e6 TH |
2736 | trace = PTRACE_EVENT_CLONE; |
2737 | else | |
2738 | trace = PTRACE_EVENT_FORK; | |
2739 | ||
2740 | if (likely(!ptrace_event_enabled(current, trace))) | |
2741 | trace = 0; | |
2742 | } | |
1da177e4 | 2743 | |
7f192e3c | 2744 | p = copy_process(NULL, trace, NUMA_NO_NODE, args); |
38addce8 | 2745 | add_latent_entropy(); |
9f5325aa MPS |
2746 | |
2747 | if (IS_ERR(p)) | |
2748 | return PTR_ERR(p); | |
2749 | ||
1da177e4 LT |
2750 | /* |
2751 | * Do this prior waking up the new thread - the thread pointer | |
2752 | * might get invalid after that point, if the thread exits quickly. | |
2753 | */ | |
9f5325aa | 2754 | trace_sched_process_fork(current, p); |
0a16b607 | 2755 | |
9f5325aa MPS |
2756 | pid = get_task_pid(p, PIDTYPE_PID); |
2757 | nr = pid_vnr(pid); | |
30e49c26 | 2758 | |
9f5325aa | 2759 | if (clone_flags & CLONE_PARENT_SETTID) |
7f192e3c | 2760 | put_user(nr, args->parent_tid); |
a6f5e063 | 2761 | |
9f5325aa MPS |
2762 | if (clone_flags & CLONE_VFORK) { |
2763 | p->vfork_done = &vfork; | |
2764 | init_completion(&vfork); | |
2765 | get_task_struct(p); | |
2766 | } | |
1da177e4 | 2767 | |
bd74fdae YZ |
2768 | if (IS_ENABLED(CONFIG_LRU_GEN) && !(clone_flags & CLONE_VM)) { |
2769 | /* lock the task to synchronize with memcg migration */ | |
2770 | task_lock(p); | |
2771 | lru_gen_add_mm(p->mm); | |
2772 | task_unlock(p); | |
2773 | } | |
2774 | ||
9f5325aa | 2775 | wake_up_new_task(p); |
09a05394 | 2776 | |
9f5325aa MPS |
2777 | /* forking complete and child started to run, tell ptracer */ |
2778 | if (unlikely(trace)) | |
2779 | ptrace_event_pid(trace, pid); | |
4e52365f | 2780 | |
9f5325aa MPS |
2781 | if (clone_flags & CLONE_VFORK) { |
2782 | if (!wait_for_vfork_done(p, &vfork)) | |
2783 | ptrace_event_pid(PTRACE_EVENT_VFORK_DONE, pid); | |
1da177e4 | 2784 | } |
9f5325aa MPS |
2785 | |
2786 | put_pid(pid); | |
92476d7f | 2787 | return nr; |
1da177e4 LT |
2788 | } |
2789 | ||
2aa3a7f8 AV |
2790 | /* |
2791 | * Create a kernel thread. | |
2792 | */ | |
2793 | pid_t kernel_thread(int (*fn)(void *), void *arg, unsigned long flags) | |
2794 | { | |
7f192e3c | 2795 | struct kernel_clone_args args = { |
3f2c788a CB |
2796 | .flags = ((lower_32_bits(flags) | CLONE_VM | |
2797 | CLONE_UNTRACED) & ~CSIGNAL), | |
2798 | .exit_signal = (lower_32_bits(flags) & CSIGNAL), | |
5bd2e97c EB |
2799 | .fn = fn, |
2800 | .fn_arg = arg, | |
343f4c49 EB |
2801 | .kthread = 1, |
2802 | }; | |
2803 | ||
2804 | return kernel_clone(&args); | |
2805 | } | |
2806 | ||
2807 | /* | |
2808 | * Create a user mode thread. | |
2809 | */ | |
2810 | pid_t user_mode_thread(int (*fn)(void *), void *arg, unsigned long flags) | |
2aa3a7f8 | 2811 | { |
7f192e3c | 2812 | struct kernel_clone_args args = { |
3f2c788a CB |
2813 | .flags = ((lower_32_bits(flags) | CLONE_VM | |
2814 | CLONE_UNTRACED) & ~CSIGNAL), | |
2815 | .exit_signal = (lower_32_bits(flags) & CSIGNAL), | |
5bd2e97c EB |
2816 | .fn = fn, |
2817 | .fn_arg = arg, | |
7f192e3c CB |
2818 | }; |
2819 | ||
cad6967a | 2820 | return kernel_clone(&args); |
2aa3a7f8 | 2821 | } |
2aa3a7f8 | 2822 | |
d2125043 AV |
2823 | #ifdef __ARCH_WANT_SYS_FORK |
2824 | SYSCALL_DEFINE0(fork) | |
2825 | { | |
2826 | #ifdef CONFIG_MMU | |
7f192e3c CB |
2827 | struct kernel_clone_args args = { |
2828 | .exit_signal = SIGCHLD, | |
2829 | }; | |
2830 | ||
cad6967a | 2831 | return kernel_clone(&args); |
d2125043 AV |
2832 | #else |
2833 | /* can not support in nommu mode */ | |
5d59e182 | 2834 | return -EINVAL; |
d2125043 AV |
2835 | #endif |
2836 | } | |
2837 | #endif | |
2838 | ||
2839 | #ifdef __ARCH_WANT_SYS_VFORK | |
2840 | SYSCALL_DEFINE0(vfork) | |
2841 | { | |
7f192e3c CB |
2842 | struct kernel_clone_args args = { |
2843 | .flags = CLONE_VFORK | CLONE_VM, | |
2844 | .exit_signal = SIGCHLD, | |
2845 | }; | |
2846 | ||
cad6967a | 2847 | return kernel_clone(&args); |
d2125043 AV |
2848 | } |
2849 | #endif | |
2850 | ||
2851 | #ifdef __ARCH_WANT_SYS_CLONE | |
2852 | #ifdef CONFIG_CLONE_BACKWARDS | |
2853 | SYSCALL_DEFINE5(clone, unsigned long, clone_flags, unsigned long, newsp, | |
2854 | int __user *, parent_tidptr, | |
3033f14a | 2855 | unsigned long, tls, |
d2125043 AV |
2856 | int __user *, child_tidptr) |
2857 | #elif defined(CONFIG_CLONE_BACKWARDS2) | |
2858 | SYSCALL_DEFINE5(clone, unsigned long, newsp, unsigned long, clone_flags, | |
2859 | int __user *, parent_tidptr, | |
2860 | int __user *, child_tidptr, | |
3033f14a | 2861 | unsigned long, tls) |
dfa9771a MS |
2862 | #elif defined(CONFIG_CLONE_BACKWARDS3) |
2863 | SYSCALL_DEFINE6(clone, unsigned long, clone_flags, unsigned long, newsp, | |
2864 | int, stack_size, | |
2865 | int __user *, parent_tidptr, | |
2866 | int __user *, child_tidptr, | |
3033f14a | 2867 | unsigned long, tls) |
d2125043 AV |
2868 | #else |
2869 | SYSCALL_DEFINE5(clone, unsigned long, clone_flags, unsigned long, newsp, | |
2870 | int __user *, parent_tidptr, | |
2871 | int __user *, child_tidptr, | |
3033f14a | 2872 | unsigned long, tls) |
d2125043 AV |
2873 | #endif |
2874 | { | |
7f192e3c | 2875 | struct kernel_clone_args args = { |
3f2c788a | 2876 | .flags = (lower_32_bits(clone_flags) & ~CSIGNAL), |
7f192e3c CB |
2877 | .pidfd = parent_tidptr, |
2878 | .child_tid = child_tidptr, | |
2879 | .parent_tid = parent_tidptr, | |
3f2c788a | 2880 | .exit_signal = (lower_32_bits(clone_flags) & CSIGNAL), |
7f192e3c CB |
2881 | .stack = newsp, |
2882 | .tls = tls, | |
2883 | }; | |
2884 | ||
cad6967a | 2885 | return kernel_clone(&args); |
7f192e3c | 2886 | } |
d68dbb0c | 2887 | #endif |
7f192e3c | 2888 | |
d68dbb0c | 2889 | #ifdef __ARCH_WANT_SYS_CLONE3 |
dd499f7a | 2890 | |
7f192e3c CB |
2891 | noinline static int copy_clone_args_from_user(struct kernel_clone_args *kargs, |
2892 | struct clone_args __user *uargs, | |
f14c234b | 2893 | size_t usize) |
7f192e3c | 2894 | { |
f14c234b | 2895 | int err; |
7f192e3c | 2896 | struct clone_args args; |
49cb2fc4 | 2897 | pid_t *kset_tid = kargs->set_tid; |
7f192e3c | 2898 | |
a966dcfe ES |
2899 | BUILD_BUG_ON(offsetofend(struct clone_args, tls) != |
2900 | CLONE_ARGS_SIZE_VER0); | |
2901 | BUILD_BUG_ON(offsetofend(struct clone_args, set_tid_size) != | |
2902 | CLONE_ARGS_SIZE_VER1); | |
2903 | BUILD_BUG_ON(offsetofend(struct clone_args, cgroup) != | |
2904 | CLONE_ARGS_SIZE_VER2); | |
2905 | BUILD_BUG_ON(sizeof(struct clone_args) != CLONE_ARGS_SIZE_VER2); | |
2906 | ||
f14c234b | 2907 | if (unlikely(usize > PAGE_SIZE)) |
7f192e3c | 2908 | return -E2BIG; |
f14c234b | 2909 | if (unlikely(usize < CLONE_ARGS_SIZE_VER0)) |
7f192e3c CB |
2910 | return -EINVAL; |
2911 | ||
f14c234b AS |
2912 | err = copy_struct_from_user(&args, sizeof(args), uargs, usize); |
2913 | if (err) | |
2914 | return err; | |
7f192e3c | 2915 | |
49cb2fc4 AR |
2916 | if (unlikely(args.set_tid_size > MAX_PID_NS_LEVEL)) |
2917 | return -EINVAL; | |
2918 | ||
2919 | if (unlikely(!args.set_tid && args.set_tid_size > 0)) | |
2920 | return -EINVAL; | |
2921 | ||
2922 | if (unlikely(args.set_tid && args.set_tid_size == 0)) | |
2923 | return -EINVAL; | |
2924 | ||
a0eb9abd ES |
2925 | /* |
2926 | * Verify that higher 32bits of exit_signal are unset and that | |
2927 | * it is a valid signal | |
2928 | */ | |
2929 | if (unlikely((args.exit_signal & ~((u64)CSIGNAL)) || | |
2930 | !valid_signal(args.exit_signal))) | |
2931 | return -EINVAL; | |
2932 | ||
62173872 ES |
2933 | if ((args.flags & CLONE_INTO_CGROUP) && |
2934 | (args.cgroup > INT_MAX || usize < CLONE_ARGS_SIZE_VER2)) | |
ef2c41cf CB |
2935 | return -EINVAL; |
2936 | ||
7f192e3c CB |
2937 | *kargs = (struct kernel_clone_args){ |
2938 | .flags = args.flags, | |
2939 | .pidfd = u64_to_user_ptr(args.pidfd), | |
2940 | .child_tid = u64_to_user_ptr(args.child_tid), | |
2941 | .parent_tid = u64_to_user_ptr(args.parent_tid), | |
2942 | .exit_signal = args.exit_signal, | |
2943 | .stack = args.stack, | |
2944 | .stack_size = args.stack_size, | |
2945 | .tls = args.tls, | |
49cb2fc4 | 2946 | .set_tid_size = args.set_tid_size, |
ef2c41cf | 2947 | .cgroup = args.cgroup, |
7f192e3c CB |
2948 | }; |
2949 | ||
49cb2fc4 AR |
2950 | if (args.set_tid && |
2951 | copy_from_user(kset_tid, u64_to_user_ptr(args.set_tid), | |
2952 | (kargs->set_tid_size * sizeof(pid_t)))) | |
2953 | return -EFAULT; | |
2954 | ||
2955 | kargs->set_tid = kset_tid; | |
2956 | ||
7f192e3c CB |
2957 | return 0; |
2958 | } | |
2959 | ||
fa729c4d CB |
2960 | /** |
2961 | * clone3_stack_valid - check and prepare stack | |
2962 | * @kargs: kernel clone args | |
2963 | * | |
2964 | * Verify that the stack arguments userspace gave us are sane. | |
2965 | * In addition, set the stack direction for userspace since it's easy for us to | |
2966 | * determine. | |
2967 | */ | |
2968 | static inline bool clone3_stack_valid(struct kernel_clone_args *kargs) | |
2969 | { | |
2970 | if (kargs->stack == 0) { | |
2971 | if (kargs->stack_size > 0) | |
2972 | return false; | |
2973 | } else { | |
2974 | if (kargs->stack_size == 0) | |
2975 | return false; | |
2976 | ||
2977 | if (!access_ok((void __user *)kargs->stack, kargs->stack_size)) | |
2978 | return false; | |
2979 | ||
2980 | #if !defined(CONFIG_STACK_GROWSUP) && !defined(CONFIG_IA64) | |
2981 | kargs->stack += kargs->stack_size; | |
2982 | #endif | |
2983 | } | |
2984 | ||
2985 | return true; | |
2986 | } | |
2987 | ||
2988 | static bool clone3_args_valid(struct kernel_clone_args *kargs) | |
7f192e3c | 2989 | { |
b612e5df | 2990 | /* Verify that no unknown flags are passed along. */ |
ef2c41cf CB |
2991 | if (kargs->flags & |
2992 | ~(CLONE_LEGACY_FLAGS | CLONE_CLEAR_SIGHAND | CLONE_INTO_CGROUP)) | |
7f192e3c CB |
2993 | return false; |
2994 | ||
2995 | /* | |
a8ca6b13 XC |
2996 | * - make the CLONE_DETACHED bit reusable for clone3 |
2997 | * - make the CSIGNAL bits reusable for clone3 | |
7f192e3c | 2998 | */ |
a402f1e3 | 2999 | if (kargs->flags & (CLONE_DETACHED | (CSIGNAL & (~CLONE_NEWTIME)))) |
7f192e3c CB |
3000 | return false; |
3001 | ||
b612e5df CB |
3002 | if ((kargs->flags & (CLONE_SIGHAND | CLONE_CLEAR_SIGHAND)) == |
3003 | (CLONE_SIGHAND | CLONE_CLEAR_SIGHAND)) | |
3004 | return false; | |
3005 | ||
7f192e3c CB |
3006 | if ((kargs->flags & (CLONE_THREAD | CLONE_PARENT)) && |
3007 | kargs->exit_signal) | |
3008 | return false; | |
3009 | ||
fa729c4d CB |
3010 | if (!clone3_stack_valid(kargs)) |
3011 | return false; | |
3012 | ||
7f192e3c CB |
3013 | return true; |
3014 | } | |
3015 | ||
501bd016 CB |
3016 | /** |
3017 | * clone3 - create a new process with specific properties | |
3018 | * @uargs: argument structure | |
3019 | * @size: size of @uargs | |
3020 | * | |
3021 | * clone3() is the extensible successor to clone()/clone2(). | |
3022 | * It takes a struct as argument that is versioned by its size. | |
3023 | * | |
3024 | * Return: On success, a positive PID for the child process. | |
3025 | * On error, a negative errno number. | |
3026 | */ | |
7f192e3c CB |
3027 | SYSCALL_DEFINE2(clone3, struct clone_args __user *, uargs, size_t, size) |
3028 | { | |
3029 | int err; | |
3030 | ||
3031 | struct kernel_clone_args kargs; | |
49cb2fc4 AR |
3032 | pid_t set_tid[MAX_PID_NS_LEVEL]; |
3033 | ||
3034 | kargs.set_tid = set_tid; | |
7f192e3c CB |
3035 | |
3036 | err = copy_clone_args_from_user(&kargs, uargs, size); | |
3037 | if (err) | |
3038 | return err; | |
3039 | ||
3040 | if (!clone3_args_valid(&kargs)) | |
3041 | return -EINVAL; | |
3042 | ||
cad6967a | 3043 | return kernel_clone(&kargs); |
d2125043 AV |
3044 | } |
3045 | #endif | |
3046 | ||
0f1b92cb ON |
3047 | void walk_process_tree(struct task_struct *top, proc_visitor visitor, void *data) |
3048 | { | |
3049 | struct task_struct *leader, *parent, *child; | |
3050 | int res; | |
3051 | ||
3052 | read_lock(&tasklist_lock); | |
3053 | leader = top = top->group_leader; | |
3054 | down: | |
3055 | for_each_thread(leader, parent) { | |
3056 | list_for_each_entry(child, &parent->children, sibling) { | |
3057 | res = visitor(child, data); | |
3058 | if (res) { | |
3059 | if (res < 0) | |
3060 | goto out; | |
3061 | leader = child; | |
3062 | goto down; | |
3063 | } | |
3064 | up: | |
3065 | ; | |
3066 | } | |
3067 | } | |
3068 | ||
3069 | if (leader != top) { | |
3070 | child = leader; | |
3071 | parent = child->real_parent; | |
3072 | leader = parent->group_leader; | |
3073 | goto up; | |
3074 | } | |
3075 | out: | |
3076 | read_unlock(&tasklist_lock); | |
3077 | } | |
3078 | ||
5fd63b30 RT |
3079 | #ifndef ARCH_MIN_MMSTRUCT_ALIGN |
3080 | #define ARCH_MIN_MMSTRUCT_ALIGN 0 | |
3081 | #endif | |
3082 | ||
51cc5068 | 3083 | static void sighand_ctor(void *data) |
aa1757f9 ON |
3084 | { |
3085 | struct sighand_struct *sighand = data; | |
3086 | ||
a35afb83 | 3087 | spin_lock_init(&sighand->siglock); |
b8fceee1 | 3088 | init_waitqueue_head(&sighand->signalfd_wqh); |
aa1757f9 ON |
3089 | } |
3090 | ||
af806027 | 3091 | void __init mm_cache_init(void) |
1da177e4 | 3092 | { |
c1a2f7f0 RR |
3093 | unsigned int mm_size; |
3094 | ||
af806027 PZ |
3095 | /* |
3096 | * The mm_cpumask is located at the end of mm_struct, and is | |
3097 | * dynamically sized based on the maximum CPU number this system | |
3098 | * can have, taking hotplug into account (nr_cpu_ids). | |
3099 | */ | |
af7f588d | 3100 | mm_size = sizeof(struct mm_struct) + cpumask_size() + mm_cid_size(); |
af806027 PZ |
3101 | |
3102 | mm_cachep = kmem_cache_create_usercopy("mm_struct", | |
3103 | mm_size, ARCH_MIN_MMSTRUCT_ALIGN, | |
3104 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_ACCOUNT, | |
3105 | offsetof(struct mm_struct, saved_auxv), | |
3106 | sizeof_field(struct mm_struct, saved_auxv), | |
3107 | NULL); | |
3108 | } | |
3109 | ||
3110 | void __init proc_caches_init(void) | |
3111 | { | |
1da177e4 LT |
3112 | sighand_cachep = kmem_cache_create("sighand_cache", |
3113 | sizeof(struct sighand_struct), 0, | |
5f0d5a3a | 3114 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_TYPESAFE_BY_RCU| |
75f296d9 | 3115 | SLAB_ACCOUNT, sighand_ctor); |
1da177e4 LT |
3116 | signal_cachep = kmem_cache_create("signal_cache", |
3117 | sizeof(struct signal_struct), 0, | |
75f296d9 | 3118 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_ACCOUNT, |
5d097056 | 3119 | NULL); |
20c2df83 | 3120 | files_cachep = kmem_cache_create("files_cache", |
1da177e4 | 3121 | sizeof(struct files_struct), 0, |
75f296d9 | 3122 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_ACCOUNT, |
5d097056 | 3123 | NULL); |
20c2df83 | 3124 | fs_cachep = kmem_cache_create("fs_cache", |
1da177e4 | 3125 | sizeof(struct fs_struct), 0, |
75f296d9 | 3126 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_ACCOUNT, |
5d097056 | 3127 | NULL); |
c1a2f7f0 | 3128 | |
5d097056 | 3129 | vm_area_cachep = KMEM_CACHE(vm_area_struct, SLAB_PANIC|SLAB_ACCOUNT); |
8feae131 | 3130 | mmap_init(); |
66577193 | 3131 | nsproxy_cache_init(); |
1da177e4 | 3132 | } |
cf2e340f | 3133 | |
cf2e340f | 3134 | /* |
9bfb23fc | 3135 | * Check constraints on flags passed to the unshare system call. |
cf2e340f | 3136 | */ |
9bfb23fc | 3137 | static int check_unshare_flags(unsigned long unshare_flags) |
cf2e340f | 3138 | { |
9bfb23fc ON |
3139 | if (unshare_flags & ~(CLONE_THREAD|CLONE_FS|CLONE_NEWNS|CLONE_SIGHAND| |
3140 | CLONE_VM|CLONE_FILES|CLONE_SYSVSEM| | |
50804fe3 | 3141 | CLONE_NEWUTS|CLONE_NEWIPC|CLONE_NEWNET| |
769071ac AV |
3142 | CLONE_NEWUSER|CLONE_NEWPID|CLONE_NEWCGROUP| |
3143 | CLONE_NEWTIME)) | |
9bfb23fc | 3144 | return -EINVAL; |
cf2e340f | 3145 | /* |
12c641ab EB |
3146 | * Not implemented, but pretend it works if there is nothing |
3147 | * to unshare. Note that unsharing the address space or the | |
3148 | * signal handlers also need to unshare the signal queues (aka | |
3149 | * CLONE_THREAD). | |
cf2e340f | 3150 | */ |
9bfb23fc | 3151 | if (unshare_flags & (CLONE_THREAD | CLONE_SIGHAND | CLONE_VM)) { |
12c641ab EB |
3152 | if (!thread_group_empty(current)) |
3153 | return -EINVAL; | |
3154 | } | |
3155 | if (unshare_flags & (CLONE_SIGHAND | CLONE_VM)) { | |
d036bda7 | 3156 | if (refcount_read(¤t->sighand->count) > 1) |
12c641ab EB |
3157 | return -EINVAL; |
3158 | } | |
3159 | if (unshare_flags & CLONE_VM) { | |
3160 | if (!current_is_single_threaded()) | |
9bfb23fc ON |
3161 | return -EINVAL; |
3162 | } | |
cf2e340f JD |
3163 | |
3164 | return 0; | |
3165 | } | |
3166 | ||
3167 | /* | |
99d1419d | 3168 | * Unshare the filesystem structure if it is being shared |
cf2e340f JD |
3169 | */ |
3170 | static int unshare_fs(unsigned long unshare_flags, struct fs_struct **new_fsp) | |
3171 | { | |
3172 | struct fs_struct *fs = current->fs; | |
3173 | ||
498052bb AV |
3174 | if (!(unshare_flags & CLONE_FS) || !fs) |
3175 | return 0; | |
3176 | ||
3177 | /* don't need lock here; in the worst case we'll do useless copy */ | |
3178 | if (fs->users == 1) | |
3179 | return 0; | |
3180 | ||
3181 | *new_fsp = copy_fs_struct(fs); | |
3182 | if (!*new_fsp) | |
3183 | return -ENOMEM; | |
cf2e340f JD |
3184 | |
3185 | return 0; | |
3186 | } | |
3187 | ||
cf2e340f | 3188 | /* |
a016f338 | 3189 | * Unshare file descriptor table if it is being shared |
cf2e340f | 3190 | */ |
60997c3d CB |
3191 | int unshare_fd(unsigned long unshare_flags, unsigned int max_fds, |
3192 | struct files_struct **new_fdp) | |
cf2e340f JD |
3193 | { |
3194 | struct files_struct *fd = current->files; | |
a016f338 | 3195 | int error = 0; |
cf2e340f JD |
3196 | |
3197 | if ((unshare_flags & CLONE_FILES) && | |
a016f338 | 3198 | (fd && atomic_read(&fd->count) > 1)) { |
60997c3d | 3199 | *new_fdp = dup_fd(fd, max_fds, &error); |
a016f338 JD |
3200 | if (!*new_fdp) |
3201 | return error; | |
3202 | } | |
cf2e340f JD |
3203 | |
3204 | return 0; | |
3205 | } | |
3206 | ||
cf2e340f JD |
3207 | /* |
3208 | * unshare allows a process to 'unshare' part of the process | |
3209 | * context which was originally shared using clone. copy_* | |
cad6967a | 3210 | * functions used by kernel_clone() cannot be used here directly |
cf2e340f JD |
3211 | * because they modify an inactive task_struct that is being |
3212 | * constructed. Here we are modifying the current, active, | |
3213 | * task_struct. | |
3214 | */ | |
9b32105e | 3215 | int ksys_unshare(unsigned long unshare_flags) |
cf2e340f | 3216 | { |
cf2e340f | 3217 | struct fs_struct *fs, *new_fs = NULL; |
ba1f70dd | 3218 | struct files_struct *new_fd = NULL; |
b2e0d987 | 3219 | struct cred *new_cred = NULL; |
cf7b708c | 3220 | struct nsproxy *new_nsproxy = NULL; |
9edff4ab | 3221 | int do_sysvsem = 0; |
9bfb23fc | 3222 | int err; |
cf2e340f | 3223 | |
b2e0d987 | 3224 | /* |
faf00da5 EB |
3225 | * If unsharing a user namespace must also unshare the thread group |
3226 | * and unshare the filesystem root and working directories. | |
b2e0d987 EB |
3227 | */ |
3228 | if (unshare_flags & CLONE_NEWUSER) | |
e66eded8 | 3229 | unshare_flags |= CLONE_THREAD | CLONE_FS; |
50804fe3 EB |
3230 | /* |
3231 | * If unsharing vm, must also unshare signal handlers. | |
3232 | */ | |
3233 | if (unshare_flags & CLONE_VM) | |
3234 | unshare_flags |= CLONE_SIGHAND; | |
12c641ab EB |
3235 | /* |
3236 | * If unsharing a signal handlers, must also unshare the signal queues. | |
3237 | */ | |
3238 | if (unshare_flags & CLONE_SIGHAND) | |
3239 | unshare_flags |= CLONE_THREAD; | |
9bfb23fc ON |
3240 | /* |
3241 | * If unsharing namespace, must also unshare filesystem information. | |
3242 | */ | |
3243 | if (unshare_flags & CLONE_NEWNS) | |
3244 | unshare_flags |= CLONE_FS; | |
50804fe3 EB |
3245 | |
3246 | err = check_unshare_flags(unshare_flags); | |
3247 | if (err) | |
3248 | goto bad_unshare_out; | |
6013f67f MS |
3249 | /* |
3250 | * CLONE_NEWIPC must also detach from the undolist: after switching | |
3251 | * to a new ipc namespace, the semaphore arrays from the old | |
3252 | * namespace are unreachable. | |
3253 | */ | |
3254 | if (unshare_flags & (CLONE_NEWIPC|CLONE_SYSVSEM)) | |
9edff4ab | 3255 | do_sysvsem = 1; |
fb0a685c DRO |
3256 | err = unshare_fs(unshare_flags, &new_fs); |
3257 | if (err) | |
9bfb23fc | 3258 | goto bad_unshare_out; |
60997c3d | 3259 | err = unshare_fd(unshare_flags, NR_OPEN_MAX, &new_fd); |
fb0a685c | 3260 | if (err) |
9bfb23fc | 3261 | goto bad_unshare_cleanup_fs; |
b2e0d987 | 3262 | err = unshare_userns(unshare_flags, &new_cred); |
fb0a685c | 3263 | if (err) |
9edff4ab | 3264 | goto bad_unshare_cleanup_fd; |
b2e0d987 EB |
3265 | err = unshare_nsproxy_namespaces(unshare_flags, &new_nsproxy, |
3266 | new_cred, new_fs); | |
3267 | if (err) | |
3268 | goto bad_unshare_cleanup_cred; | |
c0b2fc31 | 3269 | |
905ae01c AG |
3270 | if (new_cred) { |
3271 | err = set_cred_ucounts(new_cred); | |
3272 | if (err) | |
3273 | goto bad_unshare_cleanup_cred; | |
3274 | } | |
3275 | ||
b2e0d987 | 3276 | if (new_fs || new_fd || do_sysvsem || new_cred || new_nsproxy) { |
9edff4ab MS |
3277 | if (do_sysvsem) { |
3278 | /* | |
3279 | * CLONE_SYSVSEM is equivalent to sys_exit(). | |
3280 | */ | |
3281 | exit_sem(current); | |
3282 | } | |
ab602f79 JM |
3283 | if (unshare_flags & CLONE_NEWIPC) { |
3284 | /* Orphan segments in old ns (see sem above). */ | |
3285 | exit_shm(current); | |
3286 | shm_init_task(current); | |
3287 | } | |
ab516013 | 3288 | |
6f977e6b | 3289 | if (new_nsproxy) |
cf7b708c | 3290 | switch_task_namespaces(current, new_nsproxy); |
cf2e340f | 3291 | |
cf7b708c PE |
3292 | task_lock(current); |
3293 | ||
cf2e340f JD |
3294 | if (new_fs) { |
3295 | fs = current->fs; | |
2a4419b5 | 3296 | spin_lock(&fs->lock); |
cf2e340f | 3297 | current->fs = new_fs; |
498052bb AV |
3298 | if (--fs->users) |
3299 | new_fs = NULL; | |
3300 | else | |
3301 | new_fs = fs; | |
2a4419b5 | 3302 | spin_unlock(&fs->lock); |
cf2e340f JD |
3303 | } |
3304 | ||
ba1f70dd RX |
3305 | if (new_fd) |
3306 | swap(current->files, new_fd); | |
cf2e340f JD |
3307 | |
3308 | task_unlock(current); | |
b2e0d987 EB |
3309 | |
3310 | if (new_cred) { | |
3311 | /* Install the new user namespace */ | |
3312 | commit_creds(new_cred); | |
3313 | new_cred = NULL; | |
3314 | } | |
cf2e340f JD |
3315 | } |
3316 | ||
e4222673 HB |
3317 | perf_event_namespaces(current); |
3318 | ||
b2e0d987 EB |
3319 | bad_unshare_cleanup_cred: |
3320 | if (new_cred) | |
3321 | put_cred(new_cred); | |
cf2e340f JD |
3322 | bad_unshare_cleanup_fd: |
3323 | if (new_fd) | |
3324 | put_files_struct(new_fd); | |
3325 | ||
cf2e340f JD |
3326 | bad_unshare_cleanup_fs: |
3327 | if (new_fs) | |
498052bb | 3328 | free_fs_struct(new_fs); |
cf2e340f | 3329 | |
cf2e340f JD |
3330 | bad_unshare_out: |
3331 | return err; | |
3332 | } | |
3b125388 | 3333 | |
9b32105e DB |
3334 | SYSCALL_DEFINE1(unshare, unsigned long, unshare_flags) |
3335 | { | |
3336 | return ksys_unshare(unshare_flags); | |
3337 | } | |
3338 | ||
3b125388 AV |
3339 | /* |
3340 | * Helper to unshare the files of the current task. | |
3341 | * We don't want to expose copy_files internals to | |
3342 | * the exec layer of the kernel. | |
3343 | */ | |
3344 | ||
1f702603 | 3345 | int unshare_files(void) |
3b125388 AV |
3346 | { |
3347 | struct task_struct *task = current; | |
1f702603 | 3348 | struct files_struct *old, *copy = NULL; |
3b125388 AV |
3349 | int error; |
3350 | ||
60997c3d | 3351 | error = unshare_fd(CLONE_FILES, NR_OPEN_MAX, ©); |
1f702603 | 3352 | if (error || !copy) |
3b125388 | 3353 | return error; |
1f702603 EB |
3354 | |
3355 | old = task->files; | |
3b125388 AV |
3356 | task_lock(task); |
3357 | task->files = copy; | |
3358 | task_unlock(task); | |
1f702603 | 3359 | put_files_struct(old); |
3b125388 AV |
3360 | return 0; |
3361 | } | |
16db3d3f HS |
3362 | |
3363 | int sysctl_max_threads(struct ctl_table *table, int write, | |
b0daa2c7 | 3364 | void *buffer, size_t *lenp, loff_t *ppos) |
16db3d3f HS |
3365 | { |
3366 | struct ctl_table t; | |
3367 | int ret; | |
3368 | int threads = max_threads; | |
b0f53dbc | 3369 | int min = 1; |
16db3d3f HS |
3370 | int max = MAX_THREADS; |
3371 | ||
3372 | t = *table; | |
3373 | t.data = &threads; | |
3374 | t.extra1 = &min; | |
3375 | t.extra2 = &max; | |
3376 | ||
3377 | ret = proc_dointvec_minmax(&t, write, buffer, lenp, ppos); | |
3378 | if (ret || !write) | |
3379 | return ret; | |
3380 | ||
b0f53dbc | 3381 | max_threads = threads; |
16db3d3f HS |
3382 | |
3383 | return 0; | |
3384 | } |