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> | |
a2bef835 | 56 | #include <linux/syscall_user_dispatch.h> |
1da177e4 LT |
57 | #include <linux/jiffies.h> |
58 | #include <linux/futex.h> | |
8141c7f3 | 59 | #include <linux/compat.h> |
207205a2 | 60 | #include <linux/kthread.h> |
7c3ab738 | 61 | #include <linux/task_io_accounting_ops.h> |
ab2af1f5 | 62 | #include <linux/rcupdate.h> |
1da177e4 LT |
63 | #include <linux/ptrace.h> |
64 | #include <linux/mount.h> | |
65 | #include <linux/audit.h> | |
78fb7466 | 66 | #include <linux/memcontrol.h> |
f201ae23 | 67 | #include <linux/ftrace.h> |
5e2bf014 | 68 | #include <linux/proc_fs.h> |
1da177e4 LT |
69 | #include <linux/profile.h> |
70 | #include <linux/rmap.h> | |
f8af4da3 | 71 | #include <linux/ksm.h> |
1da177e4 | 72 | #include <linux/acct.h> |
893e26e6 | 73 | #include <linux/userfaultfd_k.h> |
8f0ab514 | 74 | #include <linux/tsacct_kern.h> |
9f46080c | 75 | #include <linux/cn_proc.h> |
ba96a0c8 | 76 | #include <linux/freezer.h> |
ca74e92b | 77 | #include <linux/delayacct.h> |
ad4ecbcb | 78 | #include <linux/taskstats_kern.h> |
522ed776 | 79 | #include <linux/tty.h> |
5ad4e53b | 80 | #include <linux/fs_struct.h> |
7c9f8861 | 81 | #include <linux/magic.h> |
cdd6c482 | 82 | #include <linux/perf_event.h> |
42c4ab41 | 83 | #include <linux/posix-timers.h> |
8e7cac79 | 84 | #include <linux/user-return-notifier.h> |
3d5992d2 | 85 | #include <linux/oom.h> |
ba76149f | 86 | #include <linux/khugepaged.h> |
d80e731e | 87 | #include <linux/signalfd.h> |
0326f5a9 | 88 | #include <linux/uprobes.h> |
a27bb332 | 89 | #include <linux/aio.h> |
52f5684c | 90 | #include <linux/compiler.h> |
16db3d3f | 91 | #include <linux/sysctl.h> |
5c9a8750 | 92 | #include <linux/kcov.h> |
d83a7cb3 | 93 | #include <linux/livepatch.h> |
48ac3c18 | 94 | #include <linux/thread_info.h> |
afaef01c | 95 | #include <linux/stackleak.h> |
eafb149e | 96 | #include <linux/kasan.h> |
d08b9f0c | 97 | #include <linux/scs.h> |
0f212204 | 98 | #include <linux/io_uring.h> |
a10787e6 | 99 | #include <linux/bpf.h> |
b3883a9a | 100 | #include <linux/stackprotector.h> |
fd593511 | 101 | #include <linux/user_events.h> |
cd389115 | 102 | #include <linux/iommu.h> |
932562a6 | 103 | #include <linux/rseq.h> |
1da177e4 | 104 | |
1da177e4 | 105 | #include <asm/pgalloc.h> |
7c0f6ba6 | 106 | #include <linux/uaccess.h> |
1da177e4 LT |
107 | #include <asm/mmu_context.h> |
108 | #include <asm/cacheflush.h> | |
109 | #include <asm/tlbflush.h> | |
110 | ||
ad8d75ff SR |
111 | #include <trace/events/sched.h> |
112 | ||
43d2b113 KH |
113 | #define CREATE_TRACE_POINTS |
114 | #include <trace/events/task.h> | |
115 | ||
ac1b398d HS |
116 | /* |
117 | * Minimum number of threads to boot the kernel | |
118 | */ | |
119 | #define MIN_THREADS 20 | |
120 | ||
121 | /* | |
122 | * Maximum number of threads | |
123 | */ | |
124 | #define MAX_THREADS FUTEX_TID_MASK | |
125 | ||
1da177e4 LT |
126 | /* |
127 | * Protected counters by write_lock_irq(&tasklist_lock) | |
128 | */ | |
129 | unsigned long total_forks; /* Handle normal Linux uptimes. */ | |
fb0a685c | 130 | int nr_threads; /* The idle threads do not count.. */ |
1da177e4 | 131 | |
8856ae4d | 132 | static int max_threads; /* tunable limit on nr_threads */ |
1da177e4 | 133 | |
8495f7e6 SPP |
134 | #define NAMED_ARRAY_INDEX(x) [x] = __stringify(x) |
135 | ||
136 | static const char * const resident_page_types[] = { | |
137 | NAMED_ARRAY_INDEX(MM_FILEPAGES), | |
138 | NAMED_ARRAY_INDEX(MM_ANONPAGES), | |
139 | NAMED_ARRAY_INDEX(MM_SWAPENTS), | |
140 | NAMED_ARRAY_INDEX(MM_SHMEMPAGES), | |
141 | }; | |
142 | ||
1da177e4 LT |
143 | DEFINE_PER_CPU(unsigned long, process_counts) = 0; |
144 | ||
c59923a1 | 145 | __cacheline_aligned DEFINE_RWLOCK(tasklist_lock); /* outer */ |
db1466b3 PM |
146 | |
147 | #ifdef CONFIG_PROVE_RCU | |
148 | int lockdep_tasklist_lock_is_held(void) | |
149 | { | |
150 | return lockdep_is_held(&tasklist_lock); | |
151 | } | |
152 | EXPORT_SYMBOL_GPL(lockdep_tasklist_lock_is_held); | |
153 | #endif /* #ifdef CONFIG_PROVE_RCU */ | |
1da177e4 LT |
154 | |
155 | int nr_processes(void) | |
156 | { | |
157 | int cpu; | |
158 | int total = 0; | |
159 | ||
1d510750 | 160 | for_each_possible_cpu(cpu) |
1da177e4 LT |
161 | total += per_cpu(process_counts, cpu); |
162 | ||
163 | return total; | |
164 | } | |
165 | ||
f19b9f74 AM |
166 | void __weak arch_release_task_struct(struct task_struct *tsk) |
167 | { | |
168 | } | |
169 | ||
e18b890b | 170 | static struct kmem_cache *task_struct_cachep; |
41101809 TG |
171 | |
172 | static inline struct task_struct *alloc_task_struct_node(int node) | |
173 | { | |
174 | return kmem_cache_alloc_node(task_struct_cachep, GFP_KERNEL, node); | |
175 | } | |
176 | ||
41101809 TG |
177 | static inline void free_task_struct(struct task_struct *tsk) |
178 | { | |
41101809 TG |
179 | kmem_cache_free(task_struct_cachep, tsk); |
180 | } | |
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; | |
4e2f6342 | 253 | int nr_charged = 0; |
f1c1a9ee | 254 | |
f1c1a9ee SAS |
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; | |
4e2f6342 | 261 | nr_charged++; |
f1c1a9ee SAS |
262 | } |
263 | return 0; | |
264 | err: | |
4e2f6342 | 265 | for (i = 0; i < nr_charged; i++) |
f1c1a9ee SAS |
266 | memcg_kmem_uncharge_page(vm->pages[i], 0); |
267 | return ret; | |
268 | } | |
269 | ||
7865aba3 | 270 | static int alloc_thread_stack_node(struct task_struct *tsk, int node) |
b69c49b7 | 271 | { |
1a03d3f1 | 272 | struct vm_struct *vm; |
ac496bf4 AL |
273 | void *stack; |
274 | int i; | |
275 | ||
ac496bf4 | 276 | for (i = 0; i < NR_CACHED_STACKS; i++) { |
112166f8 CL |
277 | struct vm_struct *s; |
278 | ||
279 | s = this_cpu_xchg(cached_stacks[i], NULL); | |
ac496bf4 AL |
280 | |
281 | if (!s) | |
282 | continue; | |
ac496bf4 | 283 | |
51fb34de | 284 | /* Reset stack metadata. */ |
cebd0eb2 | 285 | kasan_unpoison_range(s->addr, THREAD_SIZE); |
eafb149e | 286 | |
51fb34de AK |
287 | stack = kasan_reset_tag(s->addr); |
288 | ||
ca182551 | 289 | /* Clear stale pointers from reused stack. */ |
51fb34de | 290 | memset(stack, 0, THREAD_SIZE); |
e01e8063 | 291 | |
1a03d3f1 | 292 | if (memcg_charge_kernel_stack(s)) { |
f1c1a9ee SAS |
293 | vfree(s->addr); |
294 | return -ENOMEM; | |
295 | } | |
296 | ||
ac496bf4 | 297 | tsk->stack_vm_area = s; |
51fb34de | 298 | tsk->stack = stack; |
7865aba3 | 299 | return 0; |
ac496bf4 | 300 | } |
ac496bf4 | 301 | |
9b6f7e16 RG |
302 | /* |
303 | * Allocated stacks are cached and later reused by new threads, | |
304 | * so memcg accounting is performed manually on assigning/releasing | |
305 | * stacks to tasks. Drop __GFP_ACCOUNT. | |
306 | */ | |
48ac3c18 | 307 | stack = __vmalloc_node_range(THREAD_SIZE, THREAD_ALIGN, |
ac496bf4 | 308 | VMALLOC_START, VMALLOC_END, |
9b6f7e16 | 309 | THREADINFO_GFP & ~__GFP_ACCOUNT, |
ac496bf4 AL |
310 | PAGE_KERNEL, |
311 | 0, node, __builtin_return_address(0)); | |
7865aba3 SAS |
312 | if (!stack) |
313 | return -ENOMEM; | |
ba14a194 | 314 | |
1a03d3f1 SAS |
315 | vm = find_vm_area(stack); |
316 | if (memcg_charge_kernel_stack(vm)) { | |
f1c1a9ee SAS |
317 | vfree(stack); |
318 | return -ENOMEM; | |
319 | } | |
ba14a194 AL |
320 | /* |
321 | * We can't call find_vm_area() in interrupt context, and | |
322 | * free_thread_stack() can be called in interrupt context, | |
323 | * so cache the vm_struct. | |
324 | */ | |
1a03d3f1 | 325 | tsk->stack_vm_area = vm; |
51fb34de | 326 | stack = kasan_reset_tag(stack); |
7865aba3 SAS |
327 | tsk->stack = stack; |
328 | return 0; | |
b69c49b7 FT |
329 | } |
330 | ||
be9a2277 | 331 | static void free_thread_stack(struct task_struct *tsk) |
b69c49b7 | 332 | { |
e540bf31 SAS |
333 | if (!try_release_thread_stack_to_cache(tsk->stack_vm_area)) |
334 | thread_stack_delayed_free(tsk); | |
9b6f7e16 | 335 | |
be9a2277 SAS |
336 | tsk->stack = NULL; |
337 | tsk->stack_vm_area = NULL; | |
338 | } | |
ac496bf4 | 339 | |
be9a2277 | 340 | # else /* !CONFIG_VMAP_STACK */ |
ac496bf4 | 341 | |
e540bf31 SAS |
342 | static void thread_stack_free_rcu(struct rcu_head *rh) |
343 | { | |
344 | __free_pages(virt_to_page(rh), THREAD_SIZE_ORDER); | |
345 | } | |
346 | ||
347 | static void thread_stack_delayed_free(struct task_struct *tsk) | |
348 | { | |
349 | struct rcu_head *rh = tsk->stack; | |
350 | ||
351 | call_rcu(rh, thread_stack_free_rcu); | |
352 | } | |
353 | ||
7865aba3 | 354 | static int alloc_thread_stack_node(struct task_struct *tsk, int node) |
be9a2277 | 355 | { |
4949148a VD |
356 | struct page *page = alloc_pages_node(node, THREADINFO_GFP, |
357 | THREAD_SIZE_ORDER); | |
b6a84016 | 358 | |
1bf4580e | 359 | if (likely(page)) { |
8dcc1d34 | 360 | tsk->stack = kasan_reset_tag(page_address(page)); |
7865aba3 | 361 | return 0; |
1bf4580e | 362 | } |
7865aba3 | 363 | return -ENOMEM; |
b69c49b7 FT |
364 | } |
365 | ||
be9a2277 | 366 | static void free_thread_stack(struct task_struct *tsk) |
b69c49b7 | 367 | { |
e540bf31 | 368 | thread_stack_delayed_free(tsk); |
be9a2277 | 369 | tsk->stack = NULL; |
b69c49b7 | 370 | } |
ac496bf4 | 371 | |
be9a2277 SAS |
372 | # endif /* CONFIG_VMAP_STACK */ |
373 | # else /* !(THREAD_SIZE >= PAGE_SIZE || defined(CONFIG_VMAP_STACK)) */ | |
9b6f7e16 | 374 | |
b235beea | 375 | static struct kmem_cache *thread_stack_cache; |
ac496bf4 | 376 | |
e540bf31 SAS |
377 | static void thread_stack_free_rcu(struct rcu_head *rh) |
378 | { | |
379 | kmem_cache_free(thread_stack_cache, rh); | |
380 | } | |
ac496bf4 | 381 | |
e540bf31 SAS |
382 | static void thread_stack_delayed_free(struct task_struct *tsk) |
383 | { | |
384 | struct rcu_head *rh = tsk->stack; | |
ac496bf4 | 385 | |
e540bf31 | 386 | call_rcu(rh, thread_stack_free_rcu); |
b69c49b7 | 387 | } |
0d15d74a | 388 | |
7865aba3 | 389 | static int alloc_thread_stack_node(struct task_struct *tsk, int node) |
0d15d74a | 390 | { |
5eed6f1d RR |
391 | unsigned long *stack; |
392 | stack = kmem_cache_alloc_node(thread_stack_cache, THREADINFO_GFP, node); | |
8dcc1d34 | 393 | stack = kasan_reset_tag(stack); |
5eed6f1d | 394 | tsk->stack = stack; |
7865aba3 | 395 | return stack ? 0 : -ENOMEM; |
0d15d74a TG |
396 | } |
397 | ||
ba14a194 | 398 | static void free_thread_stack(struct task_struct *tsk) |
0d15d74a | 399 | { |
e540bf31 | 400 | thread_stack_delayed_free(tsk); |
be9a2277 | 401 | tsk->stack = NULL; |
0d15d74a TG |
402 | } |
403 | ||
b235beea | 404 | void thread_stack_cache_init(void) |
0d15d74a | 405 | { |
f9d29946 DW |
406 | thread_stack_cache = kmem_cache_create_usercopy("thread_stack", |
407 | THREAD_SIZE, THREAD_SIZE, 0, 0, | |
408 | THREAD_SIZE, NULL); | |
b235beea | 409 | BUG_ON(thread_stack_cache == NULL); |
0d15d74a | 410 | } |
be9a2277 SAS |
411 | |
412 | # endif /* THREAD_SIZE >= PAGE_SIZE || defined(CONFIG_VMAP_STACK) */ | |
b69c49b7 | 413 | |
1da177e4 | 414 | /* SLAB cache for signal_struct structures (tsk->signal) */ |
e18b890b | 415 | static struct kmem_cache *signal_cachep; |
1da177e4 LT |
416 | |
417 | /* SLAB cache for sighand_struct structures (tsk->sighand) */ | |
e18b890b | 418 | struct kmem_cache *sighand_cachep; |
1da177e4 LT |
419 | |
420 | /* SLAB cache for files_struct structures (tsk->files) */ | |
e18b890b | 421 | struct kmem_cache *files_cachep; |
1da177e4 LT |
422 | |
423 | /* SLAB cache for fs_struct structures (tsk->fs) */ | |
e18b890b | 424 | struct kmem_cache *fs_cachep; |
1da177e4 LT |
425 | |
426 | /* SLAB cache for vm_area_struct structures */ | |
3928d4f5 | 427 | static struct kmem_cache *vm_area_cachep; |
1da177e4 LT |
428 | |
429 | /* SLAB cache for mm_struct structures (tsk->mm) */ | |
e18b890b | 430 | static struct kmem_cache *mm_cachep; |
1da177e4 | 431 | |
c7f8f31c SB |
432 | #ifdef CONFIG_PER_VMA_LOCK |
433 | ||
434 | /* SLAB cache for vm_area_struct.lock */ | |
435 | static struct kmem_cache *vma_lock_cachep; | |
436 | ||
437 | static bool vma_lock_alloc(struct vm_area_struct *vma) | |
438 | { | |
439 | vma->vm_lock = kmem_cache_alloc(vma_lock_cachep, GFP_KERNEL); | |
440 | if (!vma->vm_lock) | |
441 | return false; | |
442 | ||
443 | init_rwsem(&vma->vm_lock->lock); | |
444 | vma->vm_lock_seq = -1; | |
445 | ||
446 | return true; | |
447 | } | |
448 | ||
449 | static inline void vma_lock_free(struct vm_area_struct *vma) | |
450 | { | |
451 | kmem_cache_free(vma_lock_cachep, vma->vm_lock); | |
452 | } | |
453 | ||
454 | #else /* CONFIG_PER_VMA_LOCK */ | |
455 | ||
456 | static inline bool vma_lock_alloc(struct vm_area_struct *vma) { return true; } | |
457 | static inline void vma_lock_free(struct vm_area_struct *vma) {} | |
458 | ||
459 | #endif /* CONFIG_PER_VMA_LOCK */ | |
460 | ||
490fc053 | 461 | struct vm_area_struct *vm_area_alloc(struct mm_struct *mm) |
3928d4f5 | 462 | { |
a670468f | 463 | struct vm_area_struct *vma; |
490fc053 | 464 | |
a670468f | 465 | vma = kmem_cache_alloc(vm_area_cachep, GFP_KERNEL); |
c7f8f31c SB |
466 | if (!vma) |
467 | return NULL; | |
468 | ||
469 | vma_init(vma, mm); | |
470 | if (!vma_lock_alloc(vma)) { | |
471 | kmem_cache_free(vm_area_cachep, vma); | |
472 | return NULL; | |
473 | } | |
474 | ||
490fc053 | 475 | return vma; |
3928d4f5 LT |
476 | } |
477 | ||
478 | struct vm_area_struct *vm_area_dup(struct vm_area_struct *orig) | |
479 | { | |
95faf699 LT |
480 | struct vm_area_struct *new = kmem_cache_alloc(vm_area_cachep, GFP_KERNEL); |
481 | ||
c7f8f31c SB |
482 | if (!new) |
483 | return NULL; | |
484 | ||
485 | ASSERT_EXCLUSIVE_WRITER(orig->vm_flags); | |
486 | ASSERT_EXCLUSIVE_WRITER(orig->vm_file); | |
487 | /* | |
488 | * orig->shared.rb may be modified concurrently, but the clone | |
489 | * will be reinitialized. | |
490 | */ | |
491 | data_race(memcpy(new, orig, sizeof(*new))); | |
492 | if (!vma_lock_alloc(new)) { | |
493 | kmem_cache_free(vm_area_cachep, new); | |
494 | return NULL; | |
95faf699 | 495 | } |
c7f8f31c | 496 | INIT_LIST_HEAD(&new->anon_vma_chain); |
ef6a22b7 | 497 | vma_numab_state_init(new); |
c7f8f31c SB |
498 | dup_anon_vma_name(orig, new); |
499 | ||
95faf699 | 500 | return new; |
3928d4f5 LT |
501 | } |
502 | ||
0d2ebf9c | 503 | void __vm_area_free(struct vm_area_struct *vma) |
3928d4f5 | 504 | { |
ef6a22b7 | 505 | vma_numab_state_free(vma); |
5c26f6ac | 506 | free_anon_vma_name(vma); |
c7f8f31c | 507 | vma_lock_free(vma); |
3928d4f5 LT |
508 | kmem_cache_free(vm_area_cachep, vma); |
509 | } | |
510 | ||
20cce633 ML |
511 | #ifdef CONFIG_PER_VMA_LOCK |
512 | static void vm_area_free_rcu_cb(struct rcu_head *head) | |
513 | { | |
514 | struct vm_area_struct *vma = container_of(head, struct vm_area_struct, | |
515 | vm_rcu); | |
f2e13784 SB |
516 | |
517 | /* The vma should not be locked while being destroyed. */ | |
c7f8f31c | 518 | VM_BUG_ON_VMA(rwsem_is_locked(&vma->vm_lock->lock), vma); |
20cce633 ML |
519 | __vm_area_free(vma); |
520 | } | |
521 | #endif | |
522 | ||
523 | void vm_area_free(struct vm_area_struct *vma) | |
524 | { | |
525 | #ifdef CONFIG_PER_VMA_LOCK | |
526 | call_rcu(&vma->vm_rcu, vm_area_free_rcu_cb); | |
527 | #else | |
528 | __vm_area_free(vma); | |
529 | #endif | |
530 | } | |
531 | ||
ba14a194 | 532 | static void account_kernel_stack(struct task_struct *tsk, int account) |
c6a7f572 | 533 | { |
0ce055f8 SAS |
534 | if (IS_ENABLED(CONFIG_VMAP_STACK)) { |
535 | struct vm_struct *vm = task_stack_vm_area(tsk); | |
27faca83 | 536 | int i; |
ba14a194 | 537 | |
27faca83 MS |
538 | for (i = 0; i < THREAD_SIZE / PAGE_SIZE; i++) |
539 | mod_lruvec_page_state(vm->pages[i], NR_KERNEL_STACK_KB, | |
540 | account * (PAGE_SIZE / 1024)); | |
541 | } else { | |
0ce055f8 SAS |
542 | void *stack = task_stack_page(tsk); |
543 | ||
27faca83 | 544 | /* All stack pages are in the same node. */ |
da3ceeff | 545 | mod_lruvec_kmem_state(stack, NR_KERNEL_STACK_KB, |
991e7673 | 546 | account * (THREAD_SIZE / 1024)); |
27faca83 | 547 | } |
c6a7f572 KM |
548 | } |
549 | ||
1a03d3f1 | 550 | void exit_task_stack_account(struct task_struct *tsk) |
9b6f7e16 | 551 | { |
1a03d3f1 | 552 | account_kernel_stack(tsk, -1); |
991e7673 | 553 | |
1a03d3f1 SAS |
554 | if (IS_ENABLED(CONFIG_VMAP_STACK)) { |
555 | struct vm_struct *vm; | |
9b6f7e16 RG |
556 | int i; |
557 | ||
1a03d3f1 SAS |
558 | vm = task_stack_vm_area(tsk); |
559 | for (i = 0; i < THREAD_SIZE / PAGE_SIZE; i++) | |
560 | memcg_kmem_uncharge_page(vm->pages[i], 0); | |
9b6f7e16 | 561 | } |
9b6f7e16 RG |
562 | } |
563 | ||
68f24b08 | 564 | static void release_task_stack(struct task_struct *tsk) |
1da177e4 | 565 | { |
2f064a59 | 566 | if (WARN_ON(READ_ONCE(tsk->__state) != TASK_DEAD)) |
405c0759 AL |
567 | return; /* Better to leak the stack than to free prematurely */ |
568 | ||
ba14a194 | 569 | free_thread_stack(tsk); |
68f24b08 AL |
570 | } |
571 | ||
572 | #ifdef CONFIG_THREAD_INFO_IN_TASK | |
573 | void put_task_stack(struct task_struct *tsk) | |
574 | { | |
f0b89d39 | 575 | if (refcount_dec_and_test(&tsk->stack_refcount)) |
68f24b08 AL |
576 | release_task_stack(tsk); |
577 | } | |
578 | #endif | |
579 | ||
580 | void free_task(struct task_struct *tsk) | |
581 | { | |
a1140cb2 KI |
582 | #ifdef CONFIG_SECCOMP |
583 | WARN_ON_ONCE(tsk->seccomp.filter); | |
584 | #endif | |
b90ca8ba | 585 | release_user_cpus_ptr(tsk); |
d08b9f0c ST |
586 | scs_release(tsk); |
587 | ||
68f24b08 AL |
588 | #ifndef CONFIG_THREAD_INFO_IN_TASK |
589 | /* | |
590 | * The task is finally done with both the stack and thread_info, | |
591 | * so free both. | |
592 | */ | |
593 | release_task_stack(tsk); | |
594 | #else | |
595 | /* | |
596 | * If the task had a separate stack allocation, it should be gone | |
597 | * by now. | |
598 | */ | |
f0b89d39 | 599 | WARN_ON_ONCE(refcount_read(&tsk->stack_refcount) != 0); |
68f24b08 | 600 | #endif |
23f78d4a | 601 | rt_mutex_debug_task_free(tsk); |
fb52607a | 602 | ftrace_graph_exit_task(tsk); |
f19b9f74 | 603 | arch_release_task_struct(tsk); |
1da5c46f ON |
604 | if (tsk->flags & PF_KTHREAD) |
605 | free_kthread_struct(tsk); | |
b0fd1852 | 606 | bpf_task_storage_free(tsk); |
1da177e4 LT |
607 | free_task_struct(tsk); |
608 | } | |
609 | EXPORT_SYMBOL(free_task); | |
610 | ||
fe69d560 DH |
611 | static void dup_mm_exe_file(struct mm_struct *mm, struct mm_struct *oldmm) |
612 | { | |
613 | struct file *exe_file; | |
614 | ||
615 | exe_file = get_mm_exe_file(oldmm); | |
616 | RCU_INIT_POINTER(mm->exe_file, exe_file); | |
617 | /* | |
618 | * We depend on the oldmm having properly denied write access to the | |
619 | * exe_file already. | |
620 | */ | |
621 | if (exe_file && deny_write_access(exe_file)) | |
622 | pr_warn_once("deny_write_access() failed in %s\n", __func__); | |
623 | } | |
624 | ||
d70f2a14 AM |
625 | #ifdef CONFIG_MMU |
626 | static __latent_entropy int dup_mmap(struct mm_struct *mm, | |
627 | struct mm_struct *oldmm) | |
628 | { | |
763ecb03 | 629 | struct vm_area_struct *mpnt, *tmp; |
d70f2a14 | 630 | int retval; |
c9dbe82c | 631 | unsigned long charge = 0; |
d70f2a14 | 632 | LIST_HEAD(uf); |
3b9dbd5e | 633 | VMA_ITERATOR(vmi, mm, 0); |
d70f2a14 AM |
634 | |
635 | uprobe_start_dup_mmap(); | |
d8ed45c5 | 636 | if (mmap_write_lock_killable(oldmm)) { |
d70f2a14 AM |
637 | retval = -EINTR; |
638 | goto fail_uprobe_end; | |
639 | } | |
640 | flush_cache_dup_mm(oldmm); | |
641 | uprobe_dup_mmap(oldmm, mm); | |
642 | /* | |
643 | * Not linked in yet - no deadlock potential: | |
644 | */ | |
aaa2cc56 | 645 | mmap_write_lock_nested(mm, SINGLE_DEPTH_NESTING); |
d70f2a14 AM |
646 | |
647 | /* No ordering required: file already has been exposed. */ | |
fe69d560 | 648 | dup_mm_exe_file(mm, oldmm); |
d70f2a14 AM |
649 | |
650 | mm->total_vm = oldmm->total_vm; | |
651 | mm->data_vm = oldmm->data_vm; | |
652 | mm->exec_vm = oldmm->exec_vm; | |
653 | mm->stack_vm = oldmm->stack_vm; | |
654 | ||
d70f2a14 AM |
655 | retval = ksm_fork(mm, oldmm); |
656 | if (retval) | |
657 | goto out; | |
d2081b2b | 658 | khugepaged_fork(mm, oldmm); |
d70f2a14 | 659 | |
d2406291 PZ |
660 | /* Use __mt_dup() to efficiently build an identical maple tree. */ |
661 | retval = __mt_dup(&oldmm->mm_mt, &mm->mm_mt, GFP_KERNEL); | |
662 | if (unlikely(retval)) | |
c9dbe82c LH |
663 | goto out; |
664 | ||
3dd44325 | 665 | mt_clear_in_rcu(vmi.mas.tree); |
d2406291 | 666 | for_each_vma(vmi, mpnt) { |
d70f2a14 AM |
667 | struct file *file; |
668 | ||
fb49c455 | 669 | vma_start_write(mpnt); |
d70f2a14 | 670 | if (mpnt->vm_flags & VM_DONTCOPY) { |
d2406291 PZ |
671 | retval = vma_iter_clear_gfp(&vmi, mpnt->vm_start, |
672 | mpnt->vm_end, GFP_KERNEL); | |
673 | if (retval) | |
674 | goto loop_out; | |
675 | ||
d70f2a14 AM |
676 | vm_stat_account(mm, mpnt->vm_flags, -vma_pages(mpnt)); |
677 | continue; | |
678 | } | |
679 | charge = 0; | |
655c79bb TH |
680 | /* |
681 | * Don't duplicate many vmas if we've been oom-killed (for | |
682 | * example) | |
683 | */ | |
684 | if (fatal_signal_pending(current)) { | |
685 | retval = -EINTR; | |
d4af56c5 | 686 | goto loop_out; |
655c79bb | 687 | } |
d70f2a14 AM |
688 | if (mpnt->vm_flags & VM_ACCOUNT) { |
689 | unsigned long len = vma_pages(mpnt); | |
690 | ||
691 | if (security_vm_enough_memory_mm(oldmm, len)) /* sic */ | |
692 | goto fail_nomem; | |
693 | charge = len; | |
694 | } | |
3928d4f5 | 695 | tmp = vm_area_dup(mpnt); |
d70f2a14 AM |
696 | if (!tmp) |
697 | goto fail_nomem; | |
d70f2a14 AM |
698 | retval = vma_dup_policy(mpnt, tmp); |
699 | if (retval) | |
700 | goto fail_nomem_policy; | |
701 | tmp->vm_mm = mm; | |
702 | retval = dup_userfaultfd(tmp, &uf); | |
703 | if (retval) | |
704 | goto fail_nomem_anon_vma_fork; | |
705 | if (tmp->vm_flags & VM_WIPEONFORK) { | |
93949bb2 LX |
706 | /* |
707 | * VM_WIPEONFORK gets a clean slate in the child. | |
708 | * Don't prepare anon_vma until fault since we don't | |
709 | * copy page for current vma. | |
710 | */ | |
d70f2a14 | 711 | tmp->anon_vma = NULL; |
d70f2a14 AM |
712 | } else if (anon_vma_fork(tmp, mpnt)) |
713 | goto fail_nomem_anon_vma_fork; | |
e430a95a | 714 | vm_flags_clear(tmp, VM_LOCKED_MASK); |
d70f2a14 AM |
715 | file = tmp->vm_file; |
716 | if (file) { | |
d70f2a14 AM |
717 | struct address_space *mapping = file->f_mapping; |
718 | ||
719 | get_file(file); | |
d70f2a14 | 720 | i_mmap_lock_write(mapping); |
e8e17ee9 | 721 | if (vma_is_shared_maywrite(tmp)) |
cf508b58 | 722 | mapping_allow_writable(mapping); |
d70f2a14 AM |
723 | flush_dcache_mmap_lock(mapping); |
724 | /* insert tmp into the share list, just after mpnt */ | |
725 | vma_interval_tree_insert_after(tmp, mpnt, | |
726 | &mapping->i_mmap); | |
727 | flush_dcache_mmap_unlock(mapping); | |
728 | i_mmap_unlock_write(mapping); | |
729 | } | |
730 | ||
731 | /* | |
8d9bfb26 | 732 | * Copy/update hugetlb private vma information. |
d70f2a14 AM |
733 | */ |
734 | if (is_vm_hugetlb_page(tmp)) | |
8d9bfb26 | 735 | hugetlb_dup_vma_private(tmp); |
d70f2a14 | 736 | |
d2406291 PZ |
737 | /* |
738 | * Link the vma into the MT. After using __mt_dup(), memory | |
739 | * allocation is not necessary here, so it cannot fail. | |
740 | */ | |
741 | vma_iter_bulk_store(&vmi, tmp); | |
d70f2a14 AM |
742 | |
743 | mm->map_count++; | |
744 | if (!(tmp->vm_flags & VM_WIPEONFORK)) | |
c78f4636 | 745 | retval = copy_page_range(tmp, mpnt); |
d70f2a14 AM |
746 | |
747 | if (tmp->vm_ops && tmp->vm_ops->open) | |
748 | tmp->vm_ops->open(tmp); | |
749 | ||
d2406291 PZ |
750 | if (retval) { |
751 | mpnt = vma_next(&vmi); | |
d4af56c5 | 752 | goto loop_out; |
d2406291 | 753 | } |
d70f2a14 AM |
754 | } |
755 | /* a new mm has just been created */ | |
1ed0cc5a | 756 | retval = arch_dup_mmap(oldmm, mm); |
d4af56c5 | 757 | loop_out: |
3b9dbd5e | 758 | vma_iter_free(&vmi); |
d2406291 | 759 | if (!retval) { |
3dd44325 | 760 | mt_set_in_rcu(vmi.mas.tree); |
d2406291 PZ |
761 | } else if (mpnt) { |
762 | /* | |
763 | * The entire maple tree has already been duplicated. If the | |
764 | * mmap duplication fails, mark the failure point with | |
765 | * XA_ZERO_ENTRY. In exit_mmap(), if this marker is encountered, | |
766 | * stop releasing VMAs that have not been duplicated after this | |
767 | * point. | |
768 | */ | |
769 | mas_set_range(&vmi.mas, mpnt->vm_start, mpnt->vm_end - 1); | |
770 | mas_store(&vmi.mas, XA_ZERO_ENTRY); | |
771 | } | |
d70f2a14 | 772 | out: |
d8ed45c5 | 773 | mmap_write_unlock(mm); |
d70f2a14 | 774 | flush_tlb_mm(oldmm); |
d8ed45c5 | 775 | mmap_write_unlock(oldmm); |
d70f2a14 AM |
776 | dup_userfaultfd_complete(&uf); |
777 | fail_uprobe_end: | |
778 | uprobe_end_dup_mmap(); | |
779 | return retval; | |
c9dbe82c | 780 | |
d70f2a14 AM |
781 | fail_nomem_anon_vma_fork: |
782 | mpol_put(vma_policy(tmp)); | |
783 | fail_nomem_policy: | |
3928d4f5 | 784 | vm_area_free(tmp); |
d70f2a14 AM |
785 | fail_nomem: |
786 | retval = -ENOMEM; | |
787 | vm_unacct_memory(charge); | |
d4af56c5 | 788 | goto loop_out; |
d70f2a14 AM |
789 | } |
790 | ||
791 | static inline int mm_alloc_pgd(struct mm_struct *mm) | |
792 | { | |
793 | mm->pgd = pgd_alloc(mm); | |
794 | if (unlikely(!mm->pgd)) | |
795 | return -ENOMEM; | |
796 | return 0; | |
797 | } | |
798 | ||
799 | static inline void mm_free_pgd(struct mm_struct *mm) | |
800 | { | |
801 | pgd_free(mm, mm->pgd); | |
802 | } | |
803 | #else | |
804 | static int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm) | |
805 | { | |
d8ed45c5 | 806 | mmap_write_lock(oldmm); |
fe69d560 | 807 | dup_mm_exe_file(mm, oldmm); |
d8ed45c5 | 808 | mmap_write_unlock(oldmm); |
d70f2a14 AM |
809 | return 0; |
810 | } | |
811 | #define mm_alloc_pgd(mm) (0) | |
812 | #define mm_free_pgd(mm) | |
813 | #endif /* CONFIG_MMU */ | |
814 | ||
815 | static void check_mm(struct mm_struct *mm) | |
816 | { | |
817 | int i; | |
818 | ||
8495f7e6 SPP |
819 | BUILD_BUG_ON_MSG(ARRAY_SIZE(resident_page_types) != NR_MM_COUNTERS, |
820 | "Please make sure 'struct resident_page_types[]' is updated as well"); | |
821 | ||
d70f2a14 | 822 | for (i = 0; i < NR_MM_COUNTERS; i++) { |
f1a79412 | 823 | long x = percpu_counter_sum(&mm->rss_stat[i]); |
d70f2a14 AM |
824 | |
825 | if (unlikely(x)) | |
8495f7e6 SPP |
826 | pr_alert("BUG: Bad rss-counter state mm:%p type:%s val:%ld\n", |
827 | mm, resident_page_types[i], x); | |
d70f2a14 AM |
828 | } |
829 | ||
830 | if (mm_pgtables_bytes(mm)) | |
831 | pr_alert("BUG: non-zero pgtables_bytes on freeing mm: %ld\n", | |
832 | mm_pgtables_bytes(mm)); | |
833 | ||
834 | #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS | |
835 | VM_BUG_ON_MM(mm->pmd_huge_pte, mm); | |
836 | #endif | |
837 | } | |
838 | ||
839 | #define allocate_mm() (kmem_cache_alloc(mm_cachep, GFP_KERNEL)) | |
840 | #define free_mm(mm) (kmem_cache_free(mm_cachep, (mm))) | |
841 | ||
2655421a NP |
842 | static void do_check_lazy_tlb(void *arg) |
843 | { | |
844 | struct mm_struct *mm = arg; | |
845 | ||
846 | WARN_ON_ONCE(current->active_mm == mm); | |
847 | } | |
848 | ||
849 | static void do_shoot_lazy_tlb(void *arg) | |
850 | { | |
851 | struct mm_struct *mm = arg; | |
852 | ||
853 | if (current->active_mm == mm) { | |
854 | WARN_ON_ONCE(current->mm); | |
855 | current->active_mm = &init_mm; | |
856 | switch_mm(mm, &init_mm, current); | |
857 | } | |
858 | } | |
859 | ||
860 | static void cleanup_lazy_tlbs(struct mm_struct *mm) | |
861 | { | |
862 | if (!IS_ENABLED(CONFIG_MMU_LAZY_TLB_SHOOTDOWN)) { | |
863 | /* | |
864 | * In this case, lazy tlb mms are refounted and would not reach | |
865 | * __mmdrop until all CPUs have switched away and mmdrop()ed. | |
866 | */ | |
867 | return; | |
868 | } | |
869 | ||
870 | /* | |
871 | * Lazy mm shootdown does not refcount "lazy tlb mm" usage, rather it | |
872 | * requires lazy mm users to switch to another mm when the refcount | |
873 | * drops to zero, before the mm is freed. This requires IPIs here to | |
874 | * switch kernel threads to init_mm. | |
875 | * | |
876 | * archs that use IPIs to flush TLBs can piggy-back that lazy tlb mm | |
877 | * switch with the final userspace teardown TLB flush which leaves the | |
878 | * mm lazy on this CPU but no others, reducing the need for additional | |
879 | * IPIs here. There are cases where a final IPI is still required here, | |
880 | * such as the final mmdrop being performed on a different CPU than the | |
881 | * one exiting, or kernel threads using the mm when userspace exits. | |
882 | * | |
883 | * IPI overheads have not found to be expensive, but they could be | |
884 | * reduced in a number of possible ways, for example (roughly | |
885 | * increasing order of complexity): | |
886 | * - The last lazy reference created by exit_mm() could instead switch | |
887 | * to init_mm, however it's probable this will run on the same CPU | |
888 | * immediately afterwards, so this may not reduce IPIs much. | |
889 | * - A batch of mms requiring IPIs could be gathered and freed at once. | |
890 | * - CPUs store active_mm where it can be remotely checked without a | |
891 | * lock, to filter out false-positives in the cpumask. | |
892 | * - After mm_users or mm_count reaches zero, switching away from the | |
893 | * mm could clear mm_cpumask to reduce some IPIs, perhaps together | |
894 | * with some batching or delaying of the final IPIs. | |
895 | * - A delayed freeing and RCU-like quiescing sequence based on mm | |
896 | * switching to avoid IPIs completely. | |
897 | */ | |
898 | on_each_cpu_mask(mm_cpumask(mm), do_shoot_lazy_tlb, (void *)mm, 1); | |
899 | if (IS_ENABLED(CONFIG_DEBUG_VM_SHOOT_LAZIES)) | |
900 | on_each_cpu(do_check_lazy_tlb, (void *)mm, 1); | |
901 | } | |
902 | ||
d70f2a14 AM |
903 | /* |
904 | * Called when the last reference to the mm | |
905 | * is dropped: either by a lazy thread or by | |
906 | * mmput. Free the page directory and the mm. | |
907 | */ | |
d34bc48f | 908 | void __mmdrop(struct mm_struct *mm) |
d70f2a14 AM |
909 | { |
910 | BUG_ON(mm == &init_mm); | |
3eda69c9 | 911 | WARN_ON_ONCE(mm == current->mm); |
2655421a NP |
912 | |
913 | /* Ensure no CPUs are using this as their lazy tlb mm */ | |
914 | cleanup_lazy_tlbs(mm); | |
915 | ||
3eda69c9 | 916 | WARN_ON_ONCE(mm == current->active_mm); |
d70f2a14 AM |
917 | mm_free_pgd(mm); |
918 | destroy_context(mm); | |
984cfe4e | 919 | mmu_notifier_subscriptions_destroy(mm); |
d70f2a14 AM |
920 | check_mm(mm); |
921 | put_user_ns(mm->user_ns); | |
2667ed10 | 922 | mm_pasid_drop(mm); |
223baf9d | 923 | mm_destroy_cid(mm); |
14ef95be | 924 | percpu_counter_destroy_many(mm->rss_stat, NR_MM_COUNTERS); |
f1a79412 | 925 | |
d70f2a14 AM |
926 | free_mm(mm); |
927 | } | |
d34bc48f | 928 | EXPORT_SYMBOL_GPL(__mmdrop); |
d70f2a14 AM |
929 | |
930 | static void mmdrop_async_fn(struct work_struct *work) | |
931 | { | |
932 | struct mm_struct *mm; | |
933 | ||
934 | mm = container_of(work, struct mm_struct, async_put_work); | |
935 | __mmdrop(mm); | |
936 | } | |
937 | ||
938 | static void mmdrop_async(struct mm_struct *mm) | |
939 | { | |
940 | if (unlikely(atomic_dec_and_test(&mm->mm_count))) { | |
941 | INIT_WORK(&mm->async_put_work, mmdrop_async_fn); | |
942 | schedule_work(&mm->async_put_work); | |
943 | } | |
944 | } | |
945 | ||
ea6d290c ON |
946 | static inline void free_signal_struct(struct signal_struct *sig) |
947 | { | |
97101eb4 | 948 | taskstats_tgid_free(sig); |
1c5354de | 949 | sched_autogroup_exit(sig); |
7283094e MH |
950 | /* |
951 | * __mmdrop is not safe to call from softirq context on x86 due to | |
952 | * pgd_dtor so postpone it to the async context | |
953 | */ | |
26db62f1 | 954 | if (sig->oom_mm) |
7283094e | 955 | mmdrop_async(sig->oom_mm); |
ea6d290c ON |
956 | kmem_cache_free(signal_cachep, sig); |
957 | } | |
958 | ||
959 | static inline void put_signal_struct(struct signal_struct *sig) | |
960 | { | |
60d4de3f | 961 | if (refcount_dec_and_test(&sig->sigcnt)) |
ea6d290c ON |
962 | free_signal_struct(sig); |
963 | } | |
964 | ||
158d9ebd | 965 | void __put_task_struct(struct task_struct *tsk) |
1da177e4 | 966 | { |
270f722d | 967 | WARN_ON(!tsk->exit_state); |
ec1d2819 | 968 | WARN_ON(refcount_read(&tsk->usage)); |
1da177e4 LT |
969 | WARN_ON(tsk == current); |
970 | ||
0f212204 | 971 | io_uring_free(tsk); |
2e91fa7f | 972 | cgroup_free(tsk); |
16d51a59 | 973 | task_numa_free(tsk, true); |
1a2a4d06 | 974 | security_task_free(tsk); |
e0e81739 | 975 | exit_creds(tsk); |
35df17c5 | 976 | delayacct_tsk_free(tsk); |
ea6d290c | 977 | put_signal_struct(tsk->signal); |
6e33cad0 | 978 | sched_core_free(tsk); |
2873cd31 | 979 | free_task(tsk); |
1da177e4 | 980 | } |
77c100c8 | 981 | EXPORT_SYMBOL_GPL(__put_task_struct); |
1da177e4 | 982 | |
d243b344 WLC |
983 | void __put_task_struct_rcu_cb(struct rcu_head *rhp) |
984 | { | |
985 | struct task_struct *task = container_of(rhp, struct task_struct, rcu); | |
986 | ||
987 | __put_task_struct(task); | |
988 | } | |
989 | EXPORT_SYMBOL_GPL(__put_task_struct_rcu_cb); | |
990 | ||
6c0a9fa6 | 991 | void __init __weak arch_task_cache_init(void) { } |
61c4628b | 992 | |
ff691f6e HS |
993 | /* |
994 | * set_max_threads | |
995 | */ | |
16db3d3f | 996 | static void set_max_threads(unsigned int max_threads_suggested) |
ff691f6e | 997 | { |
ac1b398d | 998 | u64 threads; |
ca79b0c2 | 999 | unsigned long nr_pages = totalram_pages(); |
ff691f6e HS |
1000 | |
1001 | /* | |
ac1b398d HS |
1002 | * The number of threads shall be limited such that the thread |
1003 | * structures may only consume a small part of the available memory. | |
ff691f6e | 1004 | */ |
3d6357de | 1005 | if (fls64(nr_pages) + fls64(PAGE_SIZE) > 64) |
ac1b398d HS |
1006 | threads = MAX_THREADS; |
1007 | else | |
3d6357de | 1008 | threads = div64_u64((u64) nr_pages * (u64) PAGE_SIZE, |
ac1b398d HS |
1009 | (u64) THREAD_SIZE * 8UL); |
1010 | ||
16db3d3f HS |
1011 | if (threads > max_threads_suggested) |
1012 | threads = max_threads_suggested; | |
1013 | ||
ac1b398d | 1014 | max_threads = clamp_t(u64, threads, MIN_THREADS, MAX_THREADS); |
ff691f6e HS |
1015 | } |
1016 | ||
5aaeb5c0 IM |
1017 | #ifdef CONFIG_ARCH_WANTS_DYNAMIC_TASK_STRUCT |
1018 | /* Initialized by the architecture: */ | |
1019 | int arch_task_struct_size __read_mostly; | |
1020 | #endif | |
0c8c0f03 | 1021 | |
5905429a KC |
1022 | static void task_struct_whitelist(unsigned long *offset, unsigned long *size) |
1023 | { | |
1024 | /* Fetch thread_struct whitelist for the architecture. */ | |
1025 | arch_thread_struct_whitelist(offset, size); | |
1026 | ||
1027 | /* | |
1028 | * Handle zero-sized whitelist or empty thread_struct, otherwise | |
1029 | * adjust offset to position of thread_struct in task_struct. | |
1030 | */ | |
1031 | if (unlikely(*size == 0)) | |
1032 | *offset = 0; | |
1033 | else | |
1034 | *offset += offsetof(struct task_struct, thread); | |
1035 | } | |
1036 | ||
ff691f6e | 1037 | void __init fork_init(void) |
1da177e4 | 1038 | { |
25f9c081 | 1039 | int i; |
1da177e4 | 1040 | #ifndef ARCH_MIN_TASKALIGN |
e274795e | 1041 | #define ARCH_MIN_TASKALIGN 0 |
1da177e4 | 1042 | #endif |
95cb64c1 | 1043 | int align = max_t(int, L1_CACHE_BYTES, ARCH_MIN_TASKALIGN); |
5905429a | 1044 | unsigned long useroffset, usersize; |
e274795e | 1045 | |
1da177e4 | 1046 | /* create a slab on which task_structs can be allocated */ |
5905429a KC |
1047 | task_struct_whitelist(&useroffset, &usersize); |
1048 | task_struct_cachep = kmem_cache_create_usercopy("task_struct", | |
e274795e | 1049 | arch_task_struct_size, align, |
5905429a KC |
1050 | SLAB_PANIC|SLAB_ACCOUNT, |
1051 | useroffset, usersize, NULL); | |
1da177e4 | 1052 | |
61c4628b SS |
1053 | /* do the arch specific task caches init */ |
1054 | arch_task_cache_init(); | |
1055 | ||
16db3d3f | 1056 | set_max_threads(MAX_THREADS); |
1da177e4 LT |
1057 | |
1058 | init_task.signal->rlim[RLIMIT_NPROC].rlim_cur = max_threads/2; | |
1059 | init_task.signal->rlim[RLIMIT_NPROC].rlim_max = max_threads/2; | |
1060 | init_task.signal->rlim[RLIMIT_SIGPENDING] = | |
1061 | init_task.signal->rlim[RLIMIT_NPROC]; | |
b376c3e1 | 1062 | |
de399236 | 1063 | for (i = 0; i < UCOUNT_COUNTS; i++) |
25f9c081 | 1064 | init_user_ns.ucount_max[i] = max_threads/2; |
19659c59 | 1065 | |
de399236 AG |
1066 | set_userns_rlimit_max(&init_user_ns, UCOUNT_RLIMIT_NPROC, RLIM_INFINITY); |
1067 | set_userns_rlimit_max(&init_user_ns, UCOUNT_RLIMIT_MSGQUEUE, RLIM_INFINITY); | |
1068 | set_userns_rlimit_max(&init_user_ns, UCOUNT_RLIMIT_SIGPENDING, RLIM_INFINITY); | |
1069 | set_userns_rlimit_max(&init_user_ns, UCOUNT_RLIMIT_MEMLOCK, RLIM_INFINITY); | |
21d1c5e3 | 1070 | |
19659c59 HR |
1071 | #ifdef CONFIG_VMAP_STACK |
1072 | cpuhp_setup_state(CPUHP_BP_PREPARE_DYN, "fork:vm_stack_cache", | |
1073 | NULL, free_vm_stack_cache); | |
1074 | #endif | |
b09be676 | 1075 | |
d08b9f0c ST |
1076 | scs_init(); |
1077 | ||
b09be676 | 1078 | lockdep_init_task(&init_task); |
aad42dd4 | 1079 | uprobes_init(); |
1da177e4 LT |
1080 | } |
1081 | ||
52f5684c | 1082 | int __weak arch_dup_task_struct(struct task_struct *dst, |
61c4628b SS |
1083 | struct task_struct *src) |
1084 | { | |
1085 | *dst = *src; | |
1086 | return 0; | |
1087 | } | |
1088 | ||
d4311ff1 AT |
1089 | void set_task_stack_end_magic(struct task_struct *tsk) |
1090 | { | |
1091 | unsigned long *stackend; | |
1092 | ||
1093 | stackend = end_of_stack(tsk); | |
1094 | *stackend = STACK_END_MAGIC; /* for overflow detection */ | |
1095 | } | |
1096 | ||
725fc629 | 1097 | static struct task_struct *dup_task_struct(struct task_struct *orig, int node) |
1da177e4 LT |
1098 | { |
1099 | struct task_struct *tsk; | |
3e26c149 | 1100 | int err; |
1da177e4 | 1101 | |
725fc629 AK |
1102 | if (node == NUMA_NO_NODE) |
1103 | node = tsk_fork_get_node(orig); | |
504f52b5 | 1104 | tsk = alloc_task_struct_node(node); |
1da177e4 LT |
1105 | if (!tsk) |
1106 | return NULL; | |
1107 | ||
546c42b2 SAS |
1108 | err = arch_dup_task_struct(tsk, orig); |
1109 | if (err) | |
f19b9f74 | 1110 | goto free_tsk; |
1da177e4 | 1111 | |
7865aba3 SAS |
1112 | err = alloc_thread_stack_node(tsk, node); |
1113 | if (err) | |
f19b9f74 | 1114 | goto free_tsk; |
ba14a194 | 1115 | |
68f24b08 | 1116 | #ifdef CONFIG_THREAD_INFO_IN_TASK |
f0b89d39 | 1117 | refcount_set(&tsk->stack_refcount, 1); |
68f24b08 | 1118 | #endif |
1a03d3f1 | 1119 | account_kernel_stack(tsk, 1); |
164c33c6 | 1120 | |
d08b9f0c ST |
1121 | err = scs_prepare(tsk, node); |
1122 | if (err) | |
1123 | goto free_stack; | |
1124 | ||
dbd95212 KC |
1125 | #ifdef CONFIG_SECCOMP |
1126 | /* | |
1127 | * We must handle setting up seccomp filters once we're under | |
1128 | * the sighand lock in case orig has changed between now and | |
1129 | * then. Until then, filter must be NULL to avoid messing up | |
1130 | * the usage counts on the error path calling free_task. | |
1131 | */ | |
1132 | tsk->seccomp.filter = NULL; | |
1133 | #endif | |
87bec58a AM |
1134 | |
1135 | setup_thread_stack(tsk, orig); | |
8e7cac79 | 1136 | clear_user_return_notifier(tsk); |
f26f9aff | 1137 | clear_tsk_need_resched(tsk); |
d4311ff1 | 1138 | set_task_stack_end_magic(tsk); |
1446e1df | 1139 | clear_syscall_work_syscall_user_dispatch(tsk); |
1da177e4 | 1140 | |
050e9baa | 1141 | #ifdef CONFIG_STACKPROTECTOR |
7cd815bc | 1142 | tsk->stack_canary = get_random_canary(); |
0a425405 | 1143 | #endif |
3bd37062 SAS |
1144 | if (orig->cpus_ptr == &orig->cpus_mask) |
1145 | tsk->cpus_ptr = &tsk->cpus_mask; | |
b90ca8ba | 1146 | dup_user_cpus_ptr(tsk, orig, node); |
0a425405 | 1147 | |
fb0a685c | 1148 | /* |
0ff7b2cf EB |
1149 | * One for the user space visible state that goes away when reaped. |
1150 | * One for the scheduler. | |
fb0a685c | 1151 | */ |
0ff7b2cf EB |
1152 | refcount_set(&tsk->rcu_users, 2); |
1153 | /* One for the rcu users */ | |
1154 | refcount_set(&tsk->usage, 1); | |
6c5c9341 | 1155 | #ifdef CONFIG_BLK_DEV_IO_TRACE |
2056a782 | 1156 | tsk->btrace_seq = 0; |
6c5c9341 | 1157 | #endif |
a0aa7f68 | 1158 | tsk->splice_pipe = NULL; |
5640f768 | 1159 | tsk->task_frag.page = NULL; |
093e5840 | 1160 | tsk->wake_q.next = NULL; |
e32cf5df | 1161 | tsk->worker_private = NULL; |
c6a7f572 | 1162 | |
5c9a8750 | 1163 | kcov_task_init(tsk); |
50b5e49c | 1164 | kmsan_task_create(tsk); |
5fbda3ec | 1165 | kmap_local_fork(tsk); |
5c9a8750 | 1166 | |
e41d5818 DV |
1167 | #ifdef CONFIG_FAULT_INJECTION |
1168 | tsk->fail_nth = 0; | |
1169 | #endif | |
1170 | ||
2c323017 | 1171 | #ifdef CONFIG_BLK_CGROUP |
f05837ed | 1172 | tsk->throttle_disk = NULL; |
2c323017 JB |
1173 | tsk->use_memdelay = 0; |
1174 | #endif | |
1175 | ||
8f23f5db | 1176 | #ifdef CONFIG_ARCH_HAS_CPU_PASID |
a3d29e82 PZ |
1177 | tsk->pasid_activated = 0; |
1178 | #endif | |
1179 | ||
d46eb14b SB |
1180 | #ifdef CONFIG_MEMCG |
1181 | tsk->active_memcg = NULL; | |
1182 | #endif | |
b041b525 TL |
1183 | |
1184 | #ifdef CONFIG_CPU_SUP_INTEL | |
1185 | tsk->reported_split_lock = 0; | |
1186 | #endif | |
1187 | ||
af7f588d MD |
1188 | #ifdef CONFIG_SCHED_MM_CID |
1189 | tsk->mm_cid = -1; | |
223baf9d | 1190 | tsk->last_mm_cid = -1; |
af7f588d | 1191 | tsk->mm_cid_active = 0; |
223baf9d | 1192 | tsk->migrate_from_cpu = -1; |
af7f588d | 1193 | #endif |
1da177e4 | 1194 | return tsk; |
61c4628b | 1195 | |
b235beea | 1196 | free_stack: |
1a03d3f1 | 1197 | exit_task_stack_account(tsk); |
ba14a194 | 1198 | free_thread_stack(tsk); |
f19b9f74 | 1199 | free_tsk: |
61c4628b SS |
1200 | free_task_struct(tsk); |
1201 | return NULL; | |
1da177e4 LT |
1202 | } |
1203 | ||
23ff4440 | 1204 | __cacheline_aligned_in_smp DEFINE_SPINLOCK(mmlist_lock); |
1da177e4 | 1205 | |
4cb0e11b HK |
1206 | static unsigned long default_dump_filter = MMF_DUMP_FILTER_DEFAULT; |
1207 | ||
1208 | static int __init coredump_filter_setup(char *s) | |
1209 | { | |
1210 | default_dump_filter = | |
1211 | (simple_strtoul(s, NULL, 0) << MMF_DUMP_FILTER_SHIFT) & | |
1212 | MMF_DUMP_FILTER_MASK; | |
1213 | return 1; | |
1214 | } | |
1215 | ||
1216 | __setup("coredump_filter=", coredump_filter_setup); | |
1217 | ||
1da177e4 LT |
1218 | #include <linux/init_task.h> |
1219 | ||
858f0993 AD |
1220 | static void mm_init_aio(struct mm_struct *mm) |
1221 | { | |
1222 | #ifdef CONFIG_AIO | |
1223 | spin_lock_init(&mm->ioctx_lock); | |
db446a08 | 1224 | mm->ioctx_table = NULL; |
858f0993 AD |
1225 | #endif |
1226 | } | |
1227 | ||
c3f3ce04 AA |
1228 | static __always_inline void mm_clear_owner(struct mm_struct *mm, |
1229 | struct task_struct *p) | |
1230 | { | |
1231 | #ifdef CONFIG_MEMCG | |
1232 | if (mm->owner == p) | |
1233 | WRITE_ONCE(mm->owner, NULL); | |
1234 | #endif | |
1235 | } | |
1236 | ||
33144e84 VD |
1237 | static void mm_init_owner(struct mm_struct *mm, struct task_struct *p) |
1238 | { | |
1239 | #ifdef CONFIG_MEMCG | |
1240 | mm->owner = p; | |
1241 | #endif | |
1242 | } | |
1243 | ||
355627f5 EB |
1244 | static void mm_init_uprobes_state(struct mm_struct *mm) |
1245 | { | |
1246 | #ifdef CONFIG_UPROBES | |
1247 | mm->uprobes_state.xol_area = NULL; | |
1248 | #endif | |
1249 | } | |
1250 | ||
bfedb589 EB |
1251 | static struct mm_struct *mm_init(struct mm_struct *mm, struct task_struct *p, |
1252 | struct user_namespace *user_ns) | |
1da177e4 | 1253 | { |
d4af56c5 LH |
1254 | mt_init_flags(&mm->mm_mt, MM_MT_FLAGS); |
1255 | mt_set_external_lock(&mm->mm_mt, &mm->mmap_lock); | |
1da177e4 LT |
1256 | atomic_set(&mm->mm_users, 1); |
1257 | atomic_set(&mm->mm_count, 1); | |
57efa1fe | 1258 | seqcount_init(&mm->write_protect_seq); |
d8ed45c5 | 1259 | mmap_init_lock(mm); |
1da177e4 | 1260 | INIT_LIST_HEAD(&mm->mmlist); |
5e31275c SB |
1261 | #ifdef CONFIG_PER_VMA_LOCK |
1262 | mm->mm_lock_seq = 0; | |
1263 | #endif | |
af5b0f6a | 1264 | mm_pgtables_bytes_init(mm); |
41f727fd VD |
1265 | mm->map_count = 0; |
1266 | mm->locked_vm = 0; | |
70f8a3ca | 1267 | atomic64_set(&mm->pinned_vm, 0); |
d559db08 | 1268 | memset(&mm->rss_stat, 0, sizeof(mm->rss_stat)); |
1da177e4 | 1269 | spin_lock_init(&mm->page_table_lock); |
88aa7cc6 | 1270 | spin_lock_init(&mm->arg_lock); |
41f727fd | 1271 | mm_init_cpumask(mm); |
858f0993 | 1272 | mm_init_aio(mm); |
cf475ad2 | 1273 | mm_init_owner(mm, p); |
a6cbd440 | 1274 | mm_pasid_init(mm); |
2b7e8665 | 1275 | RCU_INIT_POINTER(mm->exe_file, NULL); |
984cfe4e | 1276 | mmu_notifier_subscriptions_init(mm); |
16af97dc | 1277 | init_tlb_flush_pending(mm); |
41f727fd VD |
1278 | #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS |
1279 | mm->pmd_huge_pte = NULL; | |
1280 | #endif | |
355627f5 | 1281 | mm_init_uprobes_state(mm); |
13db8c50 | 1282 | hugetlb_count_init(mm); |
1da177e4 | 1283 | |
a0715cc2 | 1284 | if (current->mm) { |
24e41bf8 | 1285 | mm->flags = mmf_init_flags(current->mm->flags); |
a0715cc2 AT |
1286 | mm->def_flags = current->mm->def_flags & VM_INIT_DEF_MASK; |
1287 | } else { | |
1288 | mm->flags = default_dump_filter; | |
1da177e4 | 1289 | mm->def_flags = 0; |
a0715cc2 AT |
1290 | } |
1291 | ||
41f727fd VD |
1292 | if (mm_alloc_pgd(mm)) |
1293 | goto fail_nopgd; | |
1294 | ||
1295 | if (init_new_context(p, mm)) | |
1296 | goto fail_nocontext; | |
78fb7466 | 1297 | |
223baf9d MD |
1298 | if (mm_alloc_cid(mm)) |
1299 | goto fail_cid; | |
1300 | ||
14ef95be MG |
1301 | if (percpu_counter_init_many(mm->rss_stat, 0, GFP_KERNEL_ACCOUNT, |
1302 | NR_MM_COUNTERS)) | |
1303 | goto fail_pcpu; | |
f1a79412 | 1304 | |
bfedb589 | 1305 | mm->user_ns = get_user_ns(user_ns); |
bd74fdae | 1306 | lru_gen_init_mm(mm); |
41f727fd VD |
1307 | return mm; |
1308 | ||
f1a79412 | 1309 | fail_pcpu: |
223baf9d MD |
1310 | mm_destroy_cid(mm); |
1311 | fail_cid: | |
b20b0368 | 1312 | destroy_context(mm); |
41f727fd VD |
1313 | fail_nocontext: |
1314 | mm_free_pgd(mm); | |
1315 | fail_nopgd: | |
1da177e4 LT |
1316 | free_mm(mm); |
1317 | return NULL; | |
1318 | } | |
1319 | ||
1320 | /* | |
1321 | * Allocate and initialize an mm_struct. | |
1322 | */ | |
fb0a685c | 1323 | struct mm_struct *mm_alloc(void) |
1da177e4 | 1324 | { |
fb0a685c | 1325 | struct mm_struct *mm; |
1da177e4 LT |
1326 | |
1327 | mm = allocate_mm(); | |
de03c72c KM |
1328 | if (!mm) |
1329 | return NULL; | |
1330 | ||
1331 | memset(mm, 0, sizeof(*mm)); | |
bfedb589 | 1332 | return mm_init(mm, current, current_user_ns()); |
1da177e4 LT |
1333 | } |
1334 | ||
ec8d7c14 MH |
1335 | static inline void __mmput(struct mm_struct *mm) |
1336 | { | |
1337 | VM_BUG_ON(atomic_read(&mm->mm_users)); | |
1338 | ||
1339 | uprobe_clear_state(mm); | |
1340 | exit_aio(mm); | |
1341 | ksm_exit(mm); | |
1342 | khugepaged_exit(mm); /* must run before exit_mmap */ | |
1343 | exit_mmap(mm); | |
6fcb52a5 | 1344 | mm_put_huge_zero_page(mm); |
ec8d7c14 MH |
1345 | set_mm_exe_file(mm, NULL); |
1346 | if (!list_empty(&mm->mmlist)) { | |
1347 | spin_lock(&mmlist_lock); | |
1348 | list_del(&mm->mmlist); | |
1349 | spin_unlock(&mmlist_lock); | |
1350 | } | |
1351 | if (mm->binfmt) | |
1352 | module_put(mm->binfmt->module); | |
bd74fdae | 1353 | lru_gen_del_mm(mm); |
ec8d7c14 MH |
1354 | mmdrop(mm); |
1355 | } | |
1356 | ||
1da177e4 LT |
1357 | /* |
1358 | * Decrement the use count and release all resources for an mm. | |
1359 | */ | |
1360 | void mmput(struct mm_struct *mm) | |
1361 | { | |
0ae26f1b AM |
1362 | might_sleep(); |
1363 | ||
ec8d7c14 MH |
1364 | if (atomic_dec_and_test(&mm->mm_users)) |
1365 | __mmput(mm); | |
1366 | } | |
1367 | EXPORT_SYMBOL_GPL(mmput); | |
1368 | ||
a1b2289c SY |
1369 | #ifdef CONFIG_MMU |
1370 | static void mmput_async_fn(struct work_struct *work) | |
1371 | { | |
1372 | struct mm_struct *mm = container_of(work, struct mm_struct, | |
1373 | async_put_work); | |
1374 | ||
1375 | __mmput(mm); | |
1376 | } | |
1377 | ||
1378 | void mmput_async(struct mm_struct *mm) | |
1379 | { | |
1380 | if (atomic_dec_and_test(&mm->mm_users)) { | |
1381 | INIT_WORK(&mm->async_put_work, mmput_async_fn); | |
1382 | schedule_work(&mm->async_put_work); | |
1383 | } | |
1384 | } | |
85eaeb50 | 1385 | EXPORT_SYMBOL_GPL(mmput_async); |
a1b2289c SY |
1386 | #endif |
1387 | ||
90f31d0e KK |
1388 | /** |
1389 | * set_mm_exe_file - change a reference to the mm's executable file | |
ff0712ea MWO |
1390 | * @mm: The mm to change. |
1391 | * @new_exe_file: The new file to use. | |
90f31d0e KK |
1392 | * |
1393 | * This changes mm's executable file (shown as symlink /proc/[pid]/exe). | |
1394 | * | |
6e399cd1 | 1395 | * Main users are mmput() and sys_execve(). Callers prevent concurrent |
a7031f14 MG |
1396 | * invocations: in mmput() nobody alive left, in execve it happens before |
1397 | * the new mm is made visible to anyone. | |
fe69d560 DH |
1398 | * |
1399 | * Can only fail if new_exe_file != NULL. | |
90f31d0e | 1400 | */ |
fe69d560 | 1401 | int set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file) |
38646013 | 1402 | { |
6e399cd1 DB |
1403 | struct file *old_exe_file; |
1404 | ||
1405 | /* | |
1406 | * It is safe to dereference the exe_file without RCU as | |
1407 | * this function is only called if nobody else can access | |
1408 | * this mm -- see comment above for justification. | |
1409 | */ | |
1410 | old_exe_file = rcu_dereference_raw(mm->exe_file); | |
90f31d0e | 1411 | |
fe69d560 DH |
1412 | if (new_exe_file) { |
1413 | /* | |
1414 | * We expect the caller (i.e., sys_execve) to already denied | |
1415 | * write access, so this is unlikely to fail. | |
1416 | */ | |
1417 | if (unlikely(deny_write_access(new_exe_file))) | |
1418 | return -EACCES; | |
38646013 | 1419 | get_file(new_exe_file); |
fe69d560 | 1420 | } |
90f31d0e | 1421 | rcu_assign_pointer(mm->exe_file, new_exe_file); |
fe69d560 DH |
1422 | if (old_exe_file) { |
1423 | allow_write_access(old_exe_file); | |
90f31d0e | 1424 | fput(old_exe_file); |
fe69d560 DH |
1425 | } |
1426 | return 0; | |
38646013 JS |
1427 | } |
1428 | ||
35d7bdc8 DH |
1429 | /** |
1430 | * replace_mm_exe_file - replace a reference to the mm's executable file | |
ff0712ea MWO |
1431 | * @mm: The mm to change. |
1432 | * @new_exe_file: The new file to use. | |
35d7bdc8 | 1433 | * |
a7031f14 | 1434 | * This changes mm's executable file (shown as symlink /proc/[pid]/exe). |
35d7bdc8 DH |
1435 | * |
1436 | * Main user is sys_prctl(PR_SET_MM_MAP/EXE_FILE). | |
1437 | */ | |
1438 | int replace_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file) | |
1439 | { | |
1440 | struct vm_area_struct *vma; | |
1441 | struct file *old_exe_file; | |
1442 | int ret = 0; | |
1443 | ||
1444 | /* Forbid mm->exe_file change if old file still mapped. */ | |
1445 | old_exe_file = get_mm_exe_file(mm); | |
1446 | if (old_exe_file) { | |
fa5e5876 | 1447 | VMA_ITERATOR(vmi, mm, 0); |
35d7bdc8 | 1448 | mmap_read_lock(mm); |
fa5e5876 | 1449 | for_each_vma(vmi, vma) { |
35d7bdc8 DH |
1450 | if (!vma->vm_file) |
1451 | continue; | |
1452 | if (path_equal(&vma->vm_file->f_path, | |
fa5e5876 | 1453 | &old_exe_file->f_path)) { |
35d7bdc8 | 1454 | ret = -EBUSY; |
fa5e5876 MWO |
1455 | break; |
1456 | } | |
35d7bdc8 DH |
1457 | } |
1458 | mmap_read_unlock(mm); | |
1459 | fput(old_exe_file); | |
1460 | if (ret) | |
1461 | return ret; | |
1462 | } | |
1463 | ||
fe69d560 DH |
1464 | ret = deny_write_access(new_exe_file); |
1465 | if (ret) | |
1466 | return -EACCES; | |
35d7bdc8 | 1467 | get_file(new_exe_file); |
fe69d560 | 1468 | |
a7031f14 MG |
1469 | /* set the new file */ |
1470 | mmap_write_lock(mm); | |
1471 | old_exe_file = rcu_dereference_raw(mm->exe_file); | |
1472 | rcu_assign_pointer(mm->exe_file, new_exe_file); | |
1473 | mmap_write_unlock(mm); | |
1474 | ||
fe69d560 | 1475 | if (old_exe_file) { |
fe69d560 | 1476 | allow_write_access(old_exe_file); |
35d7bdc8 | 1477 | fput(old_exe_file); |
fe69d560 | 1478 | } |
35d7bdc8 | 1479 | return 0; |
38646013 JS |
1480 | } |
1481 | ||
90f31d0e KK |
1482 | /** |
1483 | * get_mm_exe_file - acquire a reference to the mm's executable file | |
ff0712ea | 1484 | * @mm: The mm of interest. |
90f31d0e KK |
1485 | * |
1486 | * Returns %NULL if mm has no associated executable file. | |
1487 | * User must release file via fput(). | |
1488 | */ | |
38646013 JS |
1489 | struct file *get_mm_exe_file(struct mm_struct *mm) |
1490 | { | |
1491 | struct file *exe_file; | |
1492 | ||
90f31d0e | 1493 | rcu_read_lock(); |
0ede61d8 | 1494 | exe_file = get_file_rcu(&mm->exe_file); |
90f31d0e | 1495 | rcu_read_unlock(); |
38646013 JS |
1496 | return exe_file; |
1497 | } | |
1498 | ||
cd81a917 MG |
1499 | /** |
1500 | * get_task_exe_file - acquire a reference to the task's executable file | |
ff0712ea | 1501 | * @task: The task. |
cd81a917 MG |
1502 | * |
1503 | * Returns %NULL if task's mm (if any) has no associated executable file or | |
1504 | * this is a kernel thread with borrowed mm (see the comment above get_task_mm). | |
1505 | * User must release file via fput(). | |
1506 | */ | |
1507 | struct file *get_task_exe_file(struct task_struct *task) | |
1508 | { | |
1509 | struct file *exe_file = NULL; | |
1510 | struct mm_struct *mm; | |
1511 | ||
1512 | task_lock(task); | |
1513 | mm = task->mm; | |
1514 | if (mm) { | |
1515 | if (!(task->flags & PF_KTHREAD)) | |
1516 | exe_file = get_mm_exe_file(mm); | |
1517 | } | |
1518 | task_unlock(task); | |
1519 | return exe_file; | |
1520 | } | |
38646013 | 1521 | |
1da177e4 LT |
1522 | /** |
1523 | * get_task_mm - acquire a reference to the task's mm | |
ff0712ea | 1524 | * @task: The task. |
1da177e4 | 1525 | * |
246bb0b1 | 1526 | * Returns %NULL if the task has no mm. Checks PF_KTHREAD (meaning |
1da177e4 LT |
1527 | * this kernel workthread has transiently adopted a user mm with use_mm, |
1528 | * to do its AIO) is not set and if so returns a reference to it, after | |
1529 | * bumping up the use count. User must release the mm via mmput() | |
1530 | * after use. Typically used by /proc and ptrace. | |
1531 | */ | |
1532 | struct mm_struct *get_task_mm(struct task_struct *task) | |
1533 | { | |
1534 | struct mm_struct *mm; | |
1535 | ||
1536 | task_lock(task); | |
1537 | mm = task->mm; | |
1538 | if (mm) { | |
246bb0b1 | 1539 | if (task->flags & PF_KTHREAD) |
1da177e4 LT |
1540 | mm = NULL; |
1541 | else | |
3fce371b | 1542 | mmget(mm); |
1da177e4 LT |
1543 | } |
1544 | task_unlock(task); | |
1545 | return mm; | |
1546 | } | |
1547 | EXPORT_SYMBOL_GPL(get_task_mm); | |
1548 | ||
8cdb878d CY |
1549 | struct mm_struct *mm_access(struct task_struct *task, unsigned int mode) |
1550 | { | |
1551 | struct mm_struct *mm; | |
1552 | int err; | |
1553 | ||
f7cfd871 | 1554 | err = down_read_killable(&task->signal->exec_update_lock); |
8cdb878d CY |
1555 | if (err) |
1556 | return ERR_PTR(err); | |
1557 | ||
1558 | mm = get_task_mm(task); | |
1559 | if (mm && mm != current->mm && | |
1560 | !ptrace_may_access(task, mode)) { | |
1561 | mmput(mm); | |
1562 | mm = ERR_PTR(-EACCES); | |
1563 | } | |
f7cfd871 | 1564 | up_read(&task->signal->exec_update_lock); |
8cdb878d CY |
1565 | |
1566 | return mm; | |
1567 | } | |
1568 | ||
57b59c4a | 1569 | static void complete_vfork_done(struct task_struct *tsk) |
c415c3b4 | 1570 | { |
d68b46fe | 1571 | struct completion *vfork; |
c415c3b4 | 1572 | |
d68b46fe ON |
1573 | task_lock(tsk); |
1574 | vfork = tsk->vfork_done; | |
1575 | if (likely(vfork)) { | |
1576 | tsk->vfork_done = NULL; | |
1577 | complete(vfork); | |
1578 | } | |
1579 | task_unlock(tsk); | |
1580 | } | |
1581 | ||
1582 | static int wait_for_vfork_done(struct task_struct *child, | |
1583 | struct completion *vfork) | |
1584 | { | |
a903904c | 1585 | unsigned int state = TASK_KILLABLE|TASK_FREEZABLE; |
d68b46fe ON |
1586 | int killed; |
1587 | ||
76f969e8 | 1588 | cgroup_enter_frozen(); |
f5d39b02 | 1589 | killed = wait_for_completion_state(vfork, state); |
76f969e8 | 1590 | cgroup_leave_frozen(false); |
d68b46fe ON |
1591 | |
1592 | if (killed) { | |
1593 | task_lock(child); | |
1594 | child->vfork_done = NULL; | |
1595 | task_unlock(child); | |
1596 | } | |
1597 | ||
1598 | put_task_struct(child); | |
1599 | return killed; | |
c415c3b4 ON |
1600 | } |
1601 | ||
1da177e4 LT |
1602 | /* Please note the differences between mmput and mm_release. |
1603 | * mmput is called whenever we stop holding onto a mm_struct, | |
1604 | * error success whatever. | |
1605 | * | |
1606 | * mm_release is called after a mm_struct has been removed | |
1607 | * from the current process. | |
1608 | * | |
1609 | * This difference is important for error handling, when we | |
1610 | * only half set up a mm_struct for a new process and need to restore | |
1611 | * the old one. Because we mmput the new mm_struct before | |
1612 | * restoring the old one. . . | |
1613 | * Eric Biederman 10 January 1998 | |
1614 | */ | |
4610ba7a | 1615 | static void mm_release(struct task_struct *tsk, struct mm_struct *mm) |
1da177e4 | 1616 | { |
0326f5a9 SD |
1617 | uprobe_free_utask(tsk); |
1618 | ||
1da177e4 LT |
1619 | /* Get rid of any cached register state */ |
1620 | deactivate_mm(tsk, mm); | |
1621 | ||
fec1d011 | 1622 | /* |
735f2770 MH |
1623 | * Signal userspace if we're not exiting with a core dump |
1624 | * because we want to leave the value intact for debugging | |
1625 | * purposes. | |
fec1d011 | 1626 | */ |
9c8a8228 | 1627 | if (tsk->clear_child_tid) { |
92307383 | 1628 | if (atomic_read(&mm->mm_users) > 1) { |
9c8a8228 ED |
1629 | /* |
1630 | * We don't check the error code - if userspace has | |
1631 | * not set up a proper pointer then tough luck. | |
1632 | */ | |
1633 | put_user(0, tsk->clear_child_tid); | |
2de0db99 DB |
1634 | do_futex(tsk->clear_child_tid, FUTEX_WAKE, |
1635 | 1, NULL, NULL, 0, 0); | |
9c8a8228 | 1636 | } |
1da177e4 | 1637 | tsk->clear_child_tid = NULL; |
1da177e4 | 1638 | } |
f7505d64 KK |
1639 | |
1640 | /* | |
1641 | * All done, finally we can wake up parent and return this mm to him. | |
1642 | * Also kthread_stop() uses this completion for synchronization. | |
1643 | */ | |
1644 | if (tsk->vfork_done) | |
1645 | complete_vfork_done(tsk); | |
1da177e4 LT |
1646 | } |
1647 | ||
4610ba7a TG |
1648 | void exit_mm_release(struct task_struct *tsk, struct mm_struct *mm) |
1649 | { | |
150d7158 | 1650 | futex_exit_release(tsk); |
4610ba7a TG |
1651 | mm_release(tsk, mm); |
1652 | } | |
1653 | ||
1654 | void exec_mm_release(struct task_struct *tsk, struct mm_struct *mm) | |
1655 | { | |
150d7158 | 1656 | futex_exec_release(tsk); |
4610ba7a TG |
1657 | mm_release(tsk, mm); |
1658 | } | |
1659 | ||
13585fa0 NA |
1660 | /** |
1661 | * dup_mm() - duplicates an existing mm structure | |
1662 | * @tsk: the task_struct with which the new mm will be associated. | |
1663 | * @oldmm: the mm to duplicate. | |
1664 | * | |
1665 | * Allocates a new mm structure and duplicates the provided @oldmm structure | |
1666 | * content into it. | |
1667 | * | |
1668 | * Return: the duplicated mm or NULL on failure. | |
a0a7ec30 | 1669 | */ |
13585fa0 NA |
1670 | static struct mm_struct *dup_mm(struct task_struct *tsk, |
1671 | struct mm_struct *oldmm) | |
a0a7ec30 | 1672 | { |
13585fa0 | 1673 | struct mm_struct *mm; |
a0a7ec30 JD |
1674 | int err; |
1675 | ||
a0a7ec30 JD |
1676 | mm = allocate_mm(); |
1677 | if (!mm) | |
1678 | goto fail_nomem; | |
1679 | ||
1680 | memcpy(mm, oldmm, sizeof(*mm)); | |
1681 | ||
bfedb589 | 1682 | if (!mm_init(mm, tsk, mm->user_ns)) |
a0a7ec30 JD |
1683 | goto fail_nomem; |
1684 | ||
a0a7ec30 JD |
1685 | err = dup_mmap(mm, oldmm); |
1686 | if (err) | |
1687 | goto free_pt; | |
1688 | ||
1689 | mm->hiwater_rss = get_mm_rss(mm); | |
1690 | mm->hiwater_vm = mm->total_vm; | |
1691 | ||
801460d0 HS |
1692 | if (mm->binfmt && !try_module_get(mm->binfmt->module)) |
1693 | goto free_pt; | |
1694 | ||
a0a7ec30 JD |
1695 | return mm; |
1696 | ||
1697 | free_pt: | |
801460d0 HS |
1698 | /* don't put binfmt in mmput, we haven't got module yet */ |
1699 | mm->binfmt = NULL; | |
c3f3ce04 | 1700 | mm_init_owner(mm, NULL); |
a0a7ec30 JD |
1701 | mmput(mm); |
1702 | ||
1703 | fail_nomem: | |
1704 | return NULL; | |
a0a7ec30 JD |
1705 | } |
1706 | ||
fb0a685c | 1707 | static int copy_mm(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 | 1708 | { |
fb0a685c | 1709 | struct mm_struct *mm, *oldmm; |
1da177e4 LT |
1710 | |
1711 | tsk->min_flt = tsk->maj_flt = 0; | |
1712 | tsk->nvcsw = tsk->nivcsw = 0; | |
17406b82 MSB |
1713 | #ifdef CONFIG_DETECT_HUNG_TASK |
1714 | tsk->last_switch_count = tsk->nvcsw + tsk->nivcsw; | |
a2e51445 | 1715 | tsk->last_switch_time = 0; |
17406b82 | 1716 | #endif |
1da177e4 LT |
1717 | |
1718 | tsk->mm = NULL; | |
1719 | tsk->active_mm = NULL; | |
1720 | ||
1721 | /* | |
1722 | * Are we cloning a kernel thread? | |
1723 | * | |
1724 | * We need to steal a active VM for that.. | |
1725 | */ | |
1726 | oldmm = current->mm; | |
1727 | if (!oldmm) | |
1728 | return 0; | |
1729 | ||
1730 | if (clone_flags & CLONE_VM) { | |
3fce371b | 1731 | mmget(oldmm); |
1da177e4 | 1732 | mm = oldmm; |
a6895399 REB |
1733 | } else { |
1734 | mm = dup_mm(tsk, current->mm); | |
1735 | if (!mm) | |
1736 | return -ENOMEM; | |
1da177e4 LT |
1737 | } |
1738 | ||
1da177e4 LT |
1739 | tsk->mm = mm; |
1740 | tsk->active_mm = mm; | |
af7f588d | 1741 | sched_mm_cid_fork(tsk); |
1da177e4 | 1742 | return 0; |
1da177e4 LT |
1743 | } |
1744 | ||
a39bc516 | 1745 | static int copy_fs(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 | 1746 | { |
498052bb | 1747 | struct fs_struct *fs = current->fs; |
1da177e4 | 1748 | if (clone_flags & CLONE_FS) { |
498052bb | 1749 | /* tsk->fs is already what we want */ |
2a4419b5 | 1750 | spin_lock(&fs->lock); |
90383cc0 | 1751 | /* "users" and "in_exec" locked for check_unsafe_exec() */ |
498052bb | 1752 | if (fs->in_exec) { |
2a4419b5 | 1753 | spin_unlock(&fs->lock); |
498052bb AV |
1754 | return -EAGAIN; |
1755 | } | |
1756 | fs->users++; | |
2a4419b5 | 1757 | spin_unlock(&fs->lock); |
1da177e4 LT |
1758 | return 0; |
1759 | } | |
498052bb | 1760 | tsk->fs = copy_fs_struct(fs); |
1da177e4 LT |
1761 | if (!tsk->fs) |
1762 | return -ENOMEM; | |
1763 | return 0; | |
1764 | } | |
1765 | ||
11f3f500 MC |
1766 | static int copy_files(unsigned long clone_flags, struct task_struct *tsk, |
1767 | int no_files) | |
a016f338 JD |
1768 | { |
1769 | struct files_struct *oldf, *newf; | |
1770 | int error = 0; | |
1771 | ||
1772 | /* | |
1773 | * A background process may not have any files ... | |
1774 | */ | |
1775 | oldf = current->files; | |
1776 | if (!oldf) | |
1777 | goto out; | |
1778 | ||
11f3f500 MC |
1779 | if (no_files) { |
1780 | tsk->files = NULL; | |
1781 | goto out; | |
1782 | } | |
1783 | ||
a016f338 JD |
1784 | if (clone_flags & CLONE_FILES) { |
1785 | atomic_inc(&oldf->count); | |
1786 | goto out; | |
1787 | } | |
1788 | ||
60997c3d | 1789 | newf = dup_fd(oldf, NR_OPEN_MAX, &error); |
a016f338 JD |
1790 | if (!newf) |
1791 | goto out; | |
1792 | ||
1793 | tsk->files = newf; | |
1794 | error = 0; | |
1795 | out: | |
1796 | return error; | |
1797 | } | |
1798 | ||
a39bc516 | 1799 | static int copy_sighand(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 LT |
1800 | { |
1801 | struct sighand_struct *sig; | |
1802 | ||
60348802 | 1803 | if (clone_flags & CLONE_SIGHAND) { |
d036bda7 | 1804 | refcount_inc(¤t->sighand->count); |
1da177e4 LT |
1805 | return 0; |
1806 | } | |
1807 | sig = kmem_cache_alloc(sighand_cachep, GFP_KERNEL); | |
0c282b06 | 1808 | RCU_INIT_POINTER(tsk->sighand, sig); |
1da177e4 LT |
1809 | if (!sig) |
1810 | return -ENOMEM; | |
9d7fb042 | 1811 | |
d036bda7 | 1812 | refcount_set(&sig->count, 1); |
06e62a46 | 1813 | spin_lock_irq(¤t->sighand->siglock); |
1da177e4 | 1814 | memcpy(sig->action, current->sighand->action, sizeof(sig->action)); |
06e62a46 | 1815 | spin_unlock_irq(¤t->sighand->siglock); |
b612e5df CB |
1816 | |
1817 | /* Reset all signal handler not set to SIG_IGN to SIG_DFL. */ | |
1818 | if (clone_flags & CLONE_CLEAR_SIGHAND) | |
1819 | flush_signal_handlers(tsk, 0); | |
1820 | ||
1da177e4 LT |
1821 | return 0; |
1822 | } | |
1823 | ||
a7e5328a | 1824 | void __cleanup_sighand(struct sighand_struct *sighand) |
c81addc9 | 1825 | { |
d036bda7 | 1826 | if (refcount_dec_and_test(&sighand->count)) { |
d80e731e | 1827 | signalfd_cleanup(sighand); |
392809b2 | 1828 | /* |
5f0d5a3a | 1829 | * sighand_cachep is SLAB_TYPESAFE_BY_RCU so we can free it |
392809b2 ON |
1830 | * without an RCU grace period, see __lock_task_sighand(). |
1831 | */ | |
c81addc9 | 1832 | kmem_cache_free(sighand_cachep, sighand); |
d80e731e | 1833 | } |
c81addc9 ON |
1834 | } |
1835 | ||
f06febc9 FM |
1836 | /* |
1837 | * Initialize POSIX timer handling for a thread group. | |
1838 | */ | |
1839 | static void posix_cpu_timers_init_group(struct signal_struct *sig) | |
1840 | { | |
2b69942f | 1841 | struct posix_cputimers *pct = &sig->posix_cputimers; |
78d7d407 JS |
1842 | unsigned long cpu_limit; |
1843 | ||
316c1608 | 1844 | cpu_limit = READ_ONCE(sig->rlim[RLIMIT_CPU].rlim_cur); |
3a245c0f | 1845 | posix_cputimers_group_init(pct, cpu_limit); |
f06febc9 FM |
1846 | } |
1847 | ||
a39bc516 | 1848 | static int copy_signal(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 LT |
1849 | { |
1850 | struct signal_struct *sig; | |
1da177e4 | 1851 | |
4ab6c083 | 1852 | if (clone_flags & CLONE_THREAD) |
490dea45 | 1853 | return 0; |
490dea45 | 1854 | |
a56704ef | 1855 | sig = kmem_cache_zalloc(signal_cachep, GFP_KERNEL); |
1da177e4 LT |
1856 | tsk->signal = sig; |
1857 | if (!sig) | |
1858 | return -ENOMEM; | |
1859 | ||
b3ac022c | 1860 | sig->nr_threads = 1; |
d80f7d7b | 1861 | sig->quick_threads = 1; |
1da177e4 | 1862 | atomic_set(&sig->live, 1); |
60d4de3f | 1863 | refcount_set(&sig->sigcnt, 1); |
0c740d0a ON |
1864 | |
1865 | /* list_add(thread_node, thread_head) without INIT_LIST_HEAD() */ | |
1866 | sig->thread_head = (struct list_head)LIST_HEAD_INIT(tsk->thread_node); | |
1867 | tsk->thread_node = (struct list_head)LIST_HEAD_INIT(sig->thread_head); | |
1868 | ||
1da177e4 | 1869 | init_waitqueue_head(&sig->wait_chldexit); |
db51aecc | 1870 | sig->curr_target = tsk; |
1da177e4 | 1871 | init_sigpending(&sig->shared_pending); |
c3ad2c3b | 1872 | INIT_HLIST_HEAD(&sig->multiprocess); |
e78c3496 | 1873 | seqlock_init(&sig->stats_lock); |
9d7fb042 | 1874 | prev_cputime_init(&sig->prev_cputime); |
1da177e4 | 1875 | |
baa73d9e | 1876 | #ifdef CONFIG_POSIX_TIMERS |
b18b6a9c | 1877 | INIT_LIST_HEAD(&sig->posix_timers); |
c9cb2e3d | 1878 | hrtimer_init(&sig->real_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL); |
1da177e4 | 1879 | sig->real_timer.function = it_real_fn; |
baa73d9e | 1880 | #endif |
1da177e4 | 1881 | |
1da177e4 LT |
1882 | task_lock(current->group_leader); |
1883 | memcpy(sig->rlim, current->signal->rlim, sizeof sig->rlim); | |
1884 | task_unlock(current->group_leader); | |
1885 | ||
6279a751 ON |
1886 | posix_cpu_timers_init_group(sig); |
1887 | ||
522ed776 | 1888 | tty_audit_fork(sig); |
5091faa4 | 1889 | sched_autogroup_fork(sig); |
522ed776 | 1890 | |
a63d83f4 | 1891 | sig->oom_score_adj = current->signal->oom_score_adj; |
dabb16f6 | 1892 | sig->oom_score_adj_min = current->signal->oom_score_adj_min; |
28b83c51 | 1893 | |
9b1bf12d | 1894 | mutex_init(&sig->cred_guard_mutex); |
f7cfd871 | 1895 | init_rwsem(&sig->exec_update_lock); |
9b1bf12d | 1896 | |
1da177e4 LT |
1897 | return 0; |
1898 | } | |
1899 | ||
dbd95212 KC |
1900 | static void copy_seccomp(struct task_struct *p) |
1901 | { | |
1902 | #ifdef CONFIG_SECCOMP | |
1903 | /* | |
1904 | * Must be called with sighand->lock held, which is common to | |
1905 | * all threads in the group. Holding cred_guard_mutex is not | |
1906 | * needed because this new task is not yet running and cannot | |
1907 | * be racing exec. | |
1908 | */ | |
69f6a34b | 1909 | assert_spin_locked(¤t->sighand->siglock); |
dbd95212 KC |
1910 | |
1911 | /* Ref-count the new filter user, and assign it. */ | |
1912 | get_seccomp_filter(current); | |
1913 | p->seccomp = current->seccomp; | |
1914 | ||
1915 | /* | |
1916 | * Explicitly enable no_new_privs here in case it got set | |
1917 | * between the task_struct being duplicated and holding the | |
1918 | * sighand lock. The seccomp state and nnp must be in sync. | |
1919 | */ | |
1920 | if (task_no_new_privs(current)) | |
1921 | task_set_no_new_privs(p); | |
1922 | ||
1923 | /* | |
1924 | * If the parent gained a seccomp mode after copying thread | |
1925 | * flags and between before we held the sighand lock, we have | |
1926 | * to manually enable the seccomp thread flag here. | |
1927 | */ | |
1928 | if (p->seccomp.mode != SECCOMP_MODE_DISABLED) | |
23d67a54 | 1929 | set_task_syscall_work(p, SECCOMP); |
dbd95212 KC |
1930 | #endif |
1931 | } | |
1932 | ||
17da2bd9 | 1933 | SYSCALL_DEFINE1(set_tid_address, int __user *, tidptr) |
1da177e4 LT |
1934 | { |
1935 | current->clear_child_tid = tidptr; | |
1936 | ||
b488893a | 1937 | return task_pid_vnr(current); |
1da177e4 LT |
1938 | } |
1939 | ||
a39bc516 | 1940 | static void rt_mutex_init_task(struct task_struct *p) |
23f78d4a | 1941 | { |
1d615482 | 1942 | raw_spin_lock_init(&p->pi_lock); |
e29e175b | 1943 | #ifdef CONFIG_RT_MUTEXES |
a23ba907 | 1944 | p->pi_waiters = RB_ROOT_CACHED; |
e96a7705 | 1945 | p->pi_top_task = NULL; |
23f78d4a | 1946 | p->pi_blocked_on = NULL; |
23f78d4a IM |
1947 | #endif |
1948 | } | |
1949 | ||
2c470475 EB |
1950 | static inline void init_task_pid_links(struct task_struct *task) |
1951 | { | |
1952 | enum pid_type type; | |
1953 | ||
96e1e984 | 1954 | for (type = PIDTYPE_PID; type < PIDTYPE_MAX; ++type) |
2c470475 | 1955 | INIT_HLIST_NODE(&task->pid_links[type]); |
2c470475 EB |
1956 | } |
1957 | ||
81907739 ON |
1958 | static inline void |
1959 | init_task_pid(struct task_struct *task, enum pid_type type, struct pid *pid) | |
1960 | { | |
2c470475 EB |
1961 | if (type == PIDTYPE_PID) |
1962 | task->thread_pid = pid; | |
1963 | else | |
1964 | task->signal->pids[type] = pid; | |
81907739 ON |
1965 | } |
1966 | ||
6bfbaa51 IM |
1967 | static inline void rcu_copy_process(struct task_struct *p) |
1968 | { | |
1969 | #ifdef CONFIG_PREEMPT_RCU | |
1970 | p->rcu_read_lock_nesting = 0; | |
1971 | p->rcu_read_unlock_special.s = 0; | |
1972 | p->rcu_blocked_node = NULL; | |
1973 | INIT_LIST_HEAD(&p->rcu_node_entry); | |
1974 | #endif /* #ifdef CONFIG_PREEMPT_RCU */ | |
1975 | #ifdef CONFIG_TASKS_RCU | |
1976 | p->rcu_tasks_holdout = false; | |
1977 | INIT_LIST_HEAD(&p->rcu_tasks_holdout_list); | |
1978 | p->rcu_tasks_idle_cpu = -1; | |
1979 | #endif /* #ifdef CONFIG_TASKS_RCU */ | |
d5f177d3 PM |
1980 | #ifdef CONFIG_TASKS_TRACE_RCU |
1981 | p->trc_reader_nesting = 0; | |
276c4104 | 1982 | p->trc_reader_special.s = 0; |
d5f177d3 | 1983 | INIT_LIST_HEAD(&p->trc_holdout_list); |
434c9eef | 1984 | INIT_LIST_HEAD(&p->trc_blkd_node); |
d5f177d3 | 1985 | #endif /* #ifdef CONFIG_TASKS_TRACE_RCU */ |
6bfbaa51 IM |
1986 | } |
1987 | ||
3695eae5 CB |
1988 | struct pid *pidfd_pid(const struct file *file) |
1989 | { | |
1990 | if (file->f_op == &pidfd_fops) | |
1991 | return file->private_data; | |
1992 | ||
1993 | return ERR_PTR(-EBADF); | |
1994 | } | |
1995 | ||
b3e58382 CB |
1996 | static int pidfd_release(struct inode *inode, struct file *file) |
1997 | { | |
1998 | struct pid *pid = file->private_data; | |
1999 | ||
2000 | file->private_data = NULL; | |
2001 | put_pid(pid); | |
2002 | return 0; | |
2003 | } | |
2004 | ||
2005 | #ifdef CONFIG_PROC_FS | |
15d42eb2 CK |
2006 | /** |
2007 | * pidfd_show_fdinfo - print information about a pidfd | |
2008 | * @m: proc fdinfo file | |
2009 | * @f: file referencing a pidfd | |
2010 | * | |
2011 | * Pid: | |
2012 | * This function will print the pid that a given pidfd refers to in the | |
2013 | * pid namespace of the procfs instance. | |
2014 | * If the pid namespace of the process is not a descendant of the pid | |
2015 | * namespace of the procfs instance 0 will be shown as its pid. This is | |
2016 | * similar to calling getppid() on a process whose parent is outside of | |
2017 | * its pid namespace. | |
2018 | * | |
2019 | * NSpid: | |
2020 | * If pid namespaces are supported then this function will also print | |
2021 | * the pid of a given pidfd refers to for all descendant pid namespaces | |
2022 | * starting from the current pid namespace of the instance, i.e. the | |
2023 | * Pid field and the first entry in the NSpid field will be identical. | |
2024 | * If the pid namespace of the process is not a descendant of the pid | |
2025 | * namespace of the procfs instance 0 will be shown as its first NSpid | |
2026 | * entry and no others will be shown. | |
2027 | * Note that this differs from the Pid and NSpid fields in | |
2028 | * /proc/<pid>/status where Pid and NSpid are always shown relative to | |
2029 | * the pid namespace of the procfs instance. The difference becomes | |
2030 | * obvious when sending around a pidfd between pid namespaces from a | |
a8ca6b13 | 2031 | * different branch of the tree, i.e. where no ancestral relation is |
15d42eb2 CK |
2032 | * present between the pid namespaces: |
2033 | * - create two new pid namespaces ns1 and ns2 in the initial pid | |
2034 | * namespace (also take care to create new mount namespaces in the | |
2035 | * new pid namespace and mount procfs) | |
2036 | * - create a process with a pidfd in ns1 | |
2037 | * - send pidfd from ns1 to ns2 | |
2038 | * - read /proc/self/fdinfo/<pidfd> and observe that both Pid and NSpid | |
2039 | * have exactly one entry, which is 0 | |
2040 | */ | |
b3e58382 CB |
2041 | static void pidfd_show_fdinfo(struct seq_file *m, struct file *f) |
2042 | { | |
b3e58382 | 2043 | struct pid *pid = f->private_data; |
3d6d8da4 CB |
2044 | struct pid_namespace *ns; |
2045 | pid_t nr = -1; | |
15d42eb2 | 2046 | |
3d6d8da4 | 2047 | if (likely(pid_has_task(pid, PIDTYPE_PID))) { |
9d78edea | 2048 | ns = proc_pid_ns(file_inode(m->file)->i_sb); |
3d6d8da4 CB |
2049 | nr = pid_nr_ns(pid, ns); |
2050 | } | |
2051 | ||
2052 | seq_put_decimal_ll(m, "Pid:\t", nr); | |
b3e58382 | 2053 | |
15d42eb2 | 2054 | #ifdef CONFIG_PID_NS |
3d6d8da4 CB |
2055 | seq_put_decimal_ll(m, "\nNSpid:\t", nr); |
2056 | if (nr > 0) { | |
15d42eb2 | 2057 | int i; |
b3e58382 | 2058 | |
15d42eb2 CK |
2059 | /* If nr is non-zero it means that 'pid' is valid and that |
2060 | * ns, i.e. the pid namespace associated with the procfs | |
2061 | * instance, is in the pid namespace hierarchy of pid. | |
2062 | * Start at one below the already printed level. | |
2063 | */ | |
2064 | for (i = ns->level + 1; i <= pid->level; i++) | |
3d6d8da4 | 2065 | seq_put_decimal_ll(m, "\t", pid->numbers[i].nr); |
15d42eb2 CK |
2066 | } |
2067 | #endif | |
b3e58382 CB |
2068 | seq_putc(m, '\n'); |
2069 | } | |
2070 | #endif | |
2071 | ||
b53b0b9d JFG |
2072 | /* |
2073 | * Poll support for process exit notification. | |
2074 | */ | |
9e77716a | 2075 | static __poll_t pidfd_poll(struct file *file, struct poll_table_struct *pts) |
b53b0b9d | 2076 | { |
b53b0b9d | 2077 | struct pid *pid = file->private_data; |
9e77716a | 2078 | __poll_t poll_flags = 0; |
b53b0b9d JFG |
2079 | |
2080 | poll_wait(file, &pid->wait_pidfd, pts); | |
2081 | ||
b53b0b9d JFG |
2082 | /* |
2083 | * Inform pollers only when the whole thread group exits. | |
2084 | * If the thread group leader exits before all other threads in the | |
2085 | * group, then poll(2) should block, similar to the wait(2) family. | |
2086 | */ | |
38fd525a | 2087 | if (thread_group_exited(pid)) |
9e77716a | 2088 | poll_flags = EPOLLIN | EPOLLRDNORM; |
b53b0b9d JFG |
2089 | |
2090 | return poll_flags; | |
2091 | } | |
2092 | ||
b3e58382 CB |
2093 | const struct file_operations pidfd_fops = { |
2094 | .release = pidfd_release, | |
b53b0b9d | 2095 | .poll = pidfd_poll, |
b3e58382 CB |
2096 | #ifdef CONFIG_PROC_FS |
2097 | .show_fdinfo = pidfd_show_fdinfo, | |
2098 | #endif | |
2099 | }; | |
2100 | ||
6ae930d9 CB |
2101 | /** |
2102 | * __pidfd_prepare - allocate a new pidfd_file and reserve a pidfd | |
2103 | * @pid: the struct pid for which to create a pidfd | |
2104 | * @flags: flags of the new @pidfd | |
ff0712ea | 2105 | * @ret: Where to return the file for the pidfd. |
6ae930d9 CB |
2106 | * |
2107 | * Allocate a new file that stashes @pid and reserve a new pidfd number in the | |
2108 | * caller's file descriptor table. The pidfd is reserved but not installed yet. | |
ff0712ea | 2109 | * |
6ae930d9 CB |
2110 | * The helper doesn't perform checks on @pid which makes it useful for pidfds |
2111 | * created via CLONE_PIDFD where @pid has no task attached when the pidfd and | |
2112 | * pidfd file are prepared. | |
2113 | * | |
2114 | * If this function returns successfully the caller is responsible to either | |
2115 | * call fd_install() passing the returned pidfd and pidfd file as arguments in | |
2116 | * order to install the pidfd into its file descriptor table or they must use | |
2117 | * put_unused_fd() and fput() on the returned pidfd and pidfd file | |
2118 | * respectively. | |
2119 | * | |
2120 | * This function is useful when a pidfd must already be reserved but there | |
2121 | * might still be points of failure afterwards and the caller wants to ensure | |
2122 | * that no pidfd is leaked into its file descriptor table. | |
2123 | * | |
2124 | * Return: On success, a reserved pidfd is returned from the function and a new | |
2125 | * pidfd file is returned in the last argument to the function. On | |
2126 | * error, a negative error code is returned from the function and the | |
2127 | * last argument remains unchanged. | |
2128 | */ | |
2129 | static int __pidfd_prepare(struct pid *pid, unsigned int flags, struct file **ret) | |
2130 | { | |
2131 | int pidfd; | |
2132 | struct file *pidfd_file; | |
2133 | ||
2134 | if (flags & ~(O_NONBLOCK | O_RDWR | O_CLOEXEC)) | |
2135 | return -EINVAL; | |
2136 | ||
2137 | pidfd = get_unused_fd_flags(O_RDWR | O_CLOEXEC); | |
2138 | if (pidfd < 0) | |
2139 | return pidfd; | |
2140 | ||
2141 | pidfd_file = anon_inode_getfile("[pidfd]", &pidfd_fops, pid, | |
2142 | flags | O_RDWR | O_CLOEXEC); | |
2143 | if (IS_ERR(pidfd_file)) { | |
2144 | put_unused_fd(pidfd); | |
2145 | return PTR_ERR(pidfd_file); | |
2146 | } | |
2147 | get_pid(pid); /* held by pidfd_file now */ | |
2148 | *ret = pidfd_file; | |
2149 | return pidfd; | |
2150 | } | |
2151 | ||
2152 | /** | |
2153 | * pidfd_prepare - allocate a new pidfd_file and reserve a pidfd | |
2154 | * @pid: the struct pid for which to create a pidfd | |
2155 | * @flags: flags of the new @pidfd | |
ff0712ea | 2156 | * @ret: Where to return the pidfd. |
6ae930d9 CB |
2157 | * |
2158 | * Allocate a new file that stashes @pid and reserve a new pidfd number in the | |
2159 | * caller's file descriptor table. The pidfd is reserved but not installed yet. | |
2160 | * | |
2161 | * The helper verifies that @pid is used as a thread group leader. | |
2162 | * | |
2163 | * If this function returns successfully the caller is responsible to either | |
2164 | * call fd_install() passing the returned pidfd and pidfd file as arguments in | |
2165 | * order to install the pidfd into its file descriptor table or they must use | |
2166 | * put_unused_fd() and fput() on the returned pidfd and pidfd file | |
2167 | * respectively. | |
2168 | * | |
2169 | * This function is useful when a pidfd must already be reserved but there | |
2170 | * might still be points of failure afterwards and the caller wants to ensure | |
2171 | * that no pidfd is leaked into its file descriptor table. | |
2172 | * | |
2173 | * Return: On success, a reserved pidfd is returned from the function and a new | |
2174 | * pidfd file is returned in the last argument to the function. On | |
2175 | * error, a negative error code is returned from the function and the | |
2176 | * last argument remains unchanged. | |
2177 | */ | |
2178 | int pidfd_prepare(struct pid *pid, unsigned int flags, struct file **ret) | |
2179 | { | |
2180 | if (!pid || !pid_has_task(pid, PIDTYPE_TGID)) | |
2181 | return -EINVAL; | |
2182 | ||
2183 | return __pidfd_prepare(pid, flags, ret); | |
2184 | } | |
2185 | ||
c3f3ce04 AA |
2186 | static void __delayed_free_task(struct rcu_head *rhp) |
2187 | { | |
2188 | struct task_struct *tsk = container_of(rhp, struct task_struct, rcu); | |
2189 | ||
2190 | free_task(tsk); | |
2191 | } | |
2192 | ||
2193 | static __always_inline void delayed_free_task(struct task_struct *tsk) | |
2194 | { | |
2195 | if (IS_ENABLED(CONFIG_MEMCG)) | |
2196 | call_rcu(&tsk->rcu, __delayed_free_task); | |
2197 | else | |
2198 | free_task(tsk); | |
2199 | } | |
2200 | ||
67197a4f SB |
2201 | static void copy_oom_score_adj(u64 clone_flags, struct task_struct *tsk) |
2202 | { | |
2203 | /* Skip if kernel thread */ | |
2204 | if (!tsk->mm) | |
2205 | return; | |
2206 | ||
2207 | /* Skip if spawning a thread or using vfork */ | |
2208 | if ((clone_flags & (CLONE_VM | CLONE_THREAD | CLONE_VFORK)) != CLONE_VM) | |
2209 | return; | |
2210 | ||
2211 | /* We need to synchronize with __set_oom_adj */ | |
2212 | mutex_lock(&oom_adj_mutex); | |
2213 | set_bit(MMF_MULTIPROCESS, &tsk->mm->flags); | |
2214 | /* Update the values in case they were changed after copy_signal */ | |
2215 | tsk->signal->oom_score_adj = current->signal->oom_score_adj; | |
2216 | tsk->signal->oom_score_adj_min = current->signal->oom_score_adj_min; | |
2217 | mutex_unlock(&oom_adj_mutex); | |
2218 | } | |
2219 | ||
79257534 DBO |
2220 | #ifdef CONFIG_RV |
2221 | static void rv_task_fork(struct task_struct *p) | |
2222 | { | |
2223 | int i; | |
2224 | ||
2225 | for (i = 0; i < RV_PER_TASK_MONITORS; i++) | |
2226 | p->rv[i].da_mon.monitoring = false; | |
2227 | } | |
2228 | #else | |
2229 | #define rv_task_fork(p) do {} while (0) | |
2230 | #endif | |
2231 | ||
1da177e4 LT |
2232 | /* |
2233 | * This creates a new process as a copy of the old one, | |
2234 | * but does not actually start it yet. | |
2235 | * | |
2236 | * It copies the registers, and all the appropriate | |
2237 | * parts of the process environment (as per the clone | |
2238 | * flags). The actual kick-off is left to the caller. | |
2239 | */ | |
89c8e98d | 2240 | __latent_entropy struct task_struct *copy_process( |
09a05394 | 2241 | struct pid *pid, |
3033f14a | 2242 | int trace, |
7f192e3c CB |
2243 | int node, |
2244 | struct kernel_clone_args *args) | |
1da177e4 | 2245 | { |
b3e58382 | 2246 | int pidfd = -1, retval; |
a24efe62 | 2247 | struct task_struct *p; |
c3ad2c3b | 2248 | struct multiprocess_signals delayed; |
6fd2fe49 | 2249 | struct file *pidfile = NULL; |
c5febea0 | 2250 | const u64 clone_flags = args->flags; |
769071ac | 2251 | struct nsproxy *nsp = current->nsproxy; |
1da177e4 | 2252 | |
667b6094 MPS |
2253 | /* |
2254 | * Don't allow sharing the root directory with processes in a different | |
2255 | * namespace | |
2256 | */ | |
1da177e4 LT |
2257 | if ((clone_flags & (CLONE_NEWNS|CLONE_FS)) == (CLONE_NEWNS|CLONE_FS)) |
2258 | return ERR_PTR(-EINVAL); | |
2259 | ||
e66eded8 EB |
2260 | if ((clone_flags & (CLONE_NEWUSER|CLONE_FS)) == (CLONE_NEWUSER|CLONE_FS)) |
2261 | return ERR_PTR(-EINVAL); | |
2262 | ||
1da177e4 LT |
2263 | /* |
2264 | * Thread groups must share signals as well, and detached threads | |
2265 | * can only be started up within the thread group. | |
2266 | */ | |
2267 | if ((clone_flags & CLONE_THREAD) && !(clone_flags & CLONE_SIGHAND)) | |
2268 | return ERR_PTR(-EINVAL); | |
2269 | ||
2270 | /* | |
2271 | * Shared signal handlers imply shared VM. By way of the above, | |
2272 | * thread groups also imply shared VM. Blocking this case allows | |
2273 | * for various simplifications in other code. | |
2274 | */ | |
2275 | if ((clone_flags & CLONE_SIGHAND) && !(clone_flags & CLONE_VM)) | |
2276 | return ERR_PTR(-EINVAL); | |
2277 | ||
123be07b SB |
2278 | /* |
2279 | * Siblings of global init remain as zombies on exit since they are | |
2280 | * not reaped by their parent (swapper). To solve this and to avoid | |
2281 | * multi-rooted process trees, prevent global and container-inits | |
2282 | * from creating siblings. | |
2283 | */ | |
2284 | if ((clone_flags & CLONE_PARENT) && | |
2285 | current->signal->flags & SIGNAL_UNKILLABLE) | |
2286 | return ERR_PTR(-EINVAL); | |
2287 | ||
8382fcac | 2288 | /* |
40a0d32d | 2289 | * If the new process will be in a different pid or user namespace |
faf00da5 | 2290 | * do not allow it to share a thread group with the forking task. |
8382fcac | 2291 | */ |
faf00da5 | 2292 | if (clone_flags & CLONE_THREAD) { |
40a0d32d | 2293 | if ((clone_flags & (CLONE_NEWUSER | CLONE_NEWPID)) || |
769071ac AV |
2294 | (task_active_pid_ns(current) != nsp->pid_ns_for_children)) |
2295 | return ERR_PTR(-EINVAL); | |
2296 | } | |
2297 | ||
b3e58382 | 2298 | if (clone_flags & CLONE_PIDFD) { |
b3e58382 | 2299 | /* |
b3e58382 CB |
2300 | * - CLONE_DETACHED is blocked so that we can potentially |
2301 | * reuse it later for CLONE_PIDFD. | |
2302 | * - CLONE_THREAD is blocked until someone really needs it. | |
2303 | */ | |
7f192e3c | 2304 | if (clone_flags & (CLONE_DETACHED | CLONE_THREAD)) |
b3e58382 | 2305 | return ERR_PTR(-EINVAL); |
b3e58382 CB |
2306 | } |
2307 | ||
c3ad2c3b EB |
2308 | /* |
2309 | * Force any signals received before this point to be delivered | |
2310 | * before the fork happens. Collect up signals sent to multiple | |
2311 | * processes that happen during the fork and delay them so that | |
2312 | * they appear to happen after the fork. | |
2313 | */ | |
2314 | sigemptyset(&delayed.signal); | |
2315 | INIT_HLIST_NODE(&delayed.node); | |
2316 | ||
2317 | spin_lock_irq(¤t->sighand->siglock); | |
2318 | if (!(clone_flags & CLONE_THREAD)) | |
2319 | hlist_add_head(&delayed.node, ¤t->signal->multiprocess); | |
2320 | recalc_sigpending(); | |
2321 | spin_unlock_irq(¤t->sighand->siglock); | |
2322 | retval = -ERESTARTNOINTR; | |
66ae0d1e | 2323 | if (task_sigpending(current)) |
c3ad2c3b EB |
2324 | goto fork_out; |
2325 | ||
1da177e4 | 2326 | retval = -ENOMEM; |
725fc629 | 2327 | p = dup_task_struct(current, node); |
1da177e4 LT |
2328 | if (!p) |
2329 | goto fork_out; | |
753550eb EB |
2330 | p->flags &= ~PF_KTHREAD; |
2331 | if (args->kthread) | |
2332 | p->flags |= PF_KTHREAD; | |
f9010dbd | 2333 | if (args->user_worker) { |
b16b3855 | 2334 | /* |
f9010dbd | 2335 | * Mark us a user worker, and block any signal that isn't |
b16b3855 JA |
2336 | * fatal or STOP |
2337 | */ | |
f9010dbd | 2338 | p->flags |= PF_USER_WORKER; |
b16b3855 JA |
2339 | siginitsetinv(&p->blocked, sigmask(SIGKILL)|sigmask(SIGSTOP)); |
2340 | } | |
f9010dbd MC |
2341 | if (args->io_thread) |
2342 | p->flags |= PF_IO_WORKER; | |
1da177e4 | 2343 | |
cf587db2 MC |
2344 | if (args->name) |
2345 | strscpy_pad(p->comm, args->name, sizeof(p->comm)); | |
2346 | ||
7f192e3c | 2347 | p->set_child_tid = (clone_flags & CLONE_CHILD_SETTID) ? args->child_tid : NULL; |
4d6501dc VN |
2348 | /* |
2349 | * Clear TID on mm_release()? | |
2350 | */ | |
7f192e3c | 2351 | p->clear_child_tid = (clone_flags & CLONE_CHILD_CLEARTID) ? args->child_tid : NULL; |
4d6501dc | 2352 | |
f7e8b616 SR |
2353 | ftrace_graph_init_task(p); |
2354 | ||
bea493a0 PZ |
2355 | rt_mutex_init_task(p); |
2356 | ||
a21ee605 | 2357 | lockdep_assert_irqs_enabled(); |
d12c1a37 | 2358 | #ifdef CONFIG_PROVE_LOCKING |
de30a2b3 IM |
2359 | DEBUG_LOCKS_WARN_ON(!p->softirqs_enabled); |
2360 | #endif | |
8f2f9c4d EB |
2361 | retval = copy_creds(p, clone_flags); |
2362 | if (retval < 0) | |
2363 | goto bad_fork_free; | |
2364 | ||
1da177e4 | 2365 | retval = -EAGAIN; |
de399236 | 2366 | if (is_rlimit_overlimit(task_ucounts(p), UCOUNT_RLIMIT_NPROC, rlimit(RLIMIT_NPROC))) { |
b57922b6 EP |
2367 | if (p->real_cred->user != INIT_USER && |
2368 | !capable(CAP_SYS_RESOURCE) && !capable(CAP_SYS_ADMIN)) | |
8f2f9c4d | 2369 | goto bad_fork_cleanup_count; |
1da177e4 | 2370 | } |
72fa5997 | 2371 | current->flags &= ~PF_NPROC_EXCEEDED; |
1da177e4 | 2372 | |
1da177e4 LT |
2373 | /* |
2374 | * If multiple threads are within copy_process(), then this check | |
2375 | * triggers too late. This doesn't hurt, the check is only there | |
2376 | * to stop root fork bombs. | |
2377 | */ | |
04ec93fe | 2378 | retval = -EAGAIN; |
c17d1a3a | 2379 | if (data_race(nr_threads >= max_threads)) |
1da177e4 LT |
2380 | goto bad_fork_cleanup_count; |
2381 | ||
ca74e92b | 2382 | delayacct_tsk_init(p); /* Must remain after dup_task_struct() */ |
a8ea6fc9 | 2383 | p->flags &= ~(PF_SUPERPRIV | PF_WQ_WORKER | PF_IDLE | PF_NO_SETAFFINITY); |
514ddb44 | 2384 | p->flags |= PF_FORKNOEXEC; |
1da177e4 LT |
2385 | INIT_LIST_HEAD(&p->children); |
2386 | INIT_LIST_HEAD(&p->sibling); | |
f41d911f | 2387 | rcu_copy_process(p); |
1da177e4 LT |
2388 | p->vfork_done = NULL; |
2389 | spin_lock_init(&p->alloc_lock); | |
1da177e4 | 2390 | |
1da177e4 LT |
2391 | init_sigpending(&p->pending); |
2392 | ||
64861634 | 2393 | p->utime = p->stime = p->gtime = 0; |
40565b5a | 2394 | #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME |
64861634 | 2395 | p->utimescaled = p->stimescaled = 0; |
40565b5a | 2396 | #endif |
9d7fb042 PZ |
2397 | prev_cputime_init(&p->prev_cputime); |
2398 | ||
6a61671b | 2399 | #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN |
bac5b6b6 FW |
2400 | seqcount_init(&p->vtime.seqcount); |
2401 | p->vtime.starttime = 0; | |
2402 | p->vtime.state = VTIME_INACTIVE; | |
6a61671b FW |
2403 | #endif |
2404 | ||
0f212204 JA |
2405 | #ifdef CONFIG_IO_URING |
2406 | p->io_uring = NULL; | |
2407 | #endif | |
2408 | ||
6976675d AV |
2409 | p->default_timer_slack_ns = current->timer_slack_ns; |
2410 | ||
eb414681 JW |
2411 | #ifdef CONFIG_PSI |
2412 | p->psi_flags = 0; | |
2413 | #endif | |
2414 | ||
5995477a | 2415 | task_io_accounting_init(&p->ioac); |
1da177e4 LT |
2416 | acct_clear_integrals(p); |
2417 | ||
3a245c0f | 2418 | posix_cputimers_init(&p->posix_cputimers); |
1da177e4 | 2419 | |
1da177e4 | 2420 | p->io_context = NULL; |
c0b0ae8a | 2421 | audit_set_context(p, NULL); |
b4f48b63 | 2422 | cgroup_fork(p); |
343f4c49 | 2423 | if (args->kthread) { |
40966e31 | 2424 | if (!set_kthread_struct(p)) |
ff8288ff | 2425 | goto bad_fork_cleanup_delayacct; |
40966e31 | 2426 | } |
1da177e4 | 2427 | #ifdef CONFIG_NUMA |
846a16bf | 2428 | p->mempolicy = mpol_dup(p->mempolicy); |
fb0a685c DRO |
2429 | if (IS_ERR(p->mempolicy)) { |
2430 | retval = PTR_ERR(p->mempolicy); | |
2431 | p->mempolicy = NULL; | |
ff8288ff | 2432 | goto bad_fork_cleanup_delayacct; |
fb0a685c | 2433 | } |
1da177e4 | 2434 | #endif |
778d3b0f MH |
2435 | #ifdef CONFIG_CPUSETS |
2436 | p->cpuset_mem_spread_rotor = NUMA_NO_NODE; | |
2437 | p->cpuset_slab_spread_rotor = NUMA_NO_NODE; | |
b7505861 | 2438 | seqcount_spinlock_init(&p->mems_allowed_seq, &p->alloc_lock); |
778d3b0f | 2439 | #endif |
de30a2b3 | 2440 | #ifdef CONFIG_TRACE_IRQFLAGS |
0584df9c ME |
2441 | memset(&p->irqtrace, 0, sizeof(p->irqtrace)); |
2442 | p->irqtrace.hardirq_disable_ip = _THIS_IP_; | |
2443 | p->irqtrace.softirq_enable_ip = _THIS_IP_; | |
2444 | p->softirqs_enabled = 1; | |
2445 | p->softirq_context = 0; | |
de30a2b3 | 2446 | #endif |
8bcbde54 DH |
2447 | |
2448 | p->pagefault_disabled = 0; | |
2449 | ||
fbb9ce95 | 2450 | #ifdef CONFIG_LOCKDEP |
b09be676 | 2451 | lockdep_init_task(p); |
fbb9ce95 | 2452 | #endif |
1da177e4 | 2453 | |
408894ee IM |
2454 | #ifdef CONFIG_DEBUG_MUTEXES |
2455 | p->blocked_on = NULL; /* not blocked yet */ | |
2456 | #endif | |
cafe5635 KO |
2457 | #ifdef CONFIG_BCACHE |
2458 | p->sequential_io = 0; | |
2459 | p->sequential_io_avg = 0; | |
2460 | #endif | |
a10787e6 SL |
2461 | #ifdef CONFIG_BPF_SYSCALL |
2462 | RCU_INIT_POINTER(p->bpf_storage, NULL); | |
c7603cfa | 2463 | p->bpf_ctx = NULL; |
a10787e6 | 2464 | #endif |
0f481406 | 2465 | |
3c90e6e9 | 2466 | /* Perform scheduler related setup. Assign this task to a CPU. */ |
aab03e05 DF |
2467 | retval = sched_fork(clone_flags, p); |
2468 | if (retval) | |
2469 | goto bad_fork_cleanup_policy; | |
6ab423e0 | 2470 | |
2b26f0aa | 2471 | retval = perf_event_init_task(p, clone_flags); |
6ab423e0 PZ |
2472 | if (retval) |
2473 | goto bad_fork_cleanup_policy; | |
fb0a685c DRO |
2474 | retval = audit_alloc(p); |
2475 | if (retval) | |
6c72e350 | 2476 | goto bad_fork_cleanup_perf; |
1da177e4 | 2477 | /* copy all the process information */ |
ab602f79 | 2478 | shm_init_task(p); |
e4e55b47 | 2479 | retval = security_task_alloc(p, clone_flags); |
fb0a685c | 2480 | if (retval) |
1da177e4 | 2481 | goto bad_fork_cleanup_audit; |
e4e55b47 TH |
2482 | retval = copy_semundo(clone_flags, p); |
2483 | if (retval) | |
2484 | goto bad_fork_cleanup_security; | |
11f3f500 | 2485 | retval = copy_files(clone_flags, p, args->no_files); |
fb0a685c | 2486 | if (retval) |
1da177e4 | 2487 | goto bad_fork_cleanup_semundo; |
fb0a685c DRO |
2488 | retval = copy_fs(clone_flags, p); |
2489 | if (retval) | |
1da177e4 | 2490 | goto bad_fork_cleanup_files; |
fb0a685c DRO |
2491 | retval = copy_sighand(clone_flags, p); |
2492 | if (retval) | |
1da177e4 | 2493 | goto bad_fork_cleanup_fs; |
fb0a685c DRO |
2494 | retval = copy_signal(clone_flags, p); |
2495 | if (retval) | |
1da177e4 | 2496 | goto bad_fork_cleanup_sighand; |
fb0a685c DRO |
2497 | retval = copy_mm(clone_flags, p); |
2498 | if (retval) | |
1da177e4 | 2499 | goto bad_fork_cleanup_signal; |
fb0a685c DRO |
2500 | retval = copy_namespaces(clone_flags, p); |
2501 | if (retval) | |
d84f4f99 | 2502 | goto bad_fork_cleanup_mm; |
fb0a685c DRO |
2503 | retval = copy_io(clone_flags, p); |
2504 | if (retval) | |
fd0928df | 2505 | goto bad_fork_cleanup_namespaces; |
c5febea0 | 2506 | retval = copy_thread(p, args); |
1da177e4 | 2507 | if (retval) |
fd0928df | 2508 | goto bad_fork_cleanup_io; |
1da177e4 | 2509 | |
afaef01c AP |
2510 | stackleak_task_init(p); |
2511 | ||
425fb2b4 | 2512 | if (pid != &init_struct_pid) { |
49cb2fc4 AR |
2513 | pid = alloc_pid(p->nsproxy->pid_ns_for_children, args->set_tid, |
2514 | args->set_tid_size); | |
35f71bc0 MH |
2515 | if (IS_ERR(pid)) { |
2516 | retval = PTR_ERR(pid); | |
0740aa5f | 2517 | goto bad_fork_cleanup_thread; |
35f71bc0 | 2518 | } |
425fb2b4 PE |
2519 | } |
2520 | ||
b3e58382 CB |
2521 | /* |
2522 | * This has to happen after we've potentially unshared the file | |
2523 | * descriptor table (so that the pidfd doesn't leak into the child | |
2524 | * if the fd table isn't shared). | |
2525 | */ | |
2526 | if (clone_flags & CLONE_PIDFD) { | |
ca7707f5 CB |
2527 | /* Note that no task has been attached to @pid yet. */ |
2528 | retval = __pidfd_prepare(pid, O_RDWR | O_CLOEXEC, &pidfile); | |
b3e58382 CB |
2529 | if (retval < 0) |
2530 | goto bad_fork_free_pid; | |
b3e58382 | 2531 | pidfd = retval; |
6fd2fe49 | 2532 | |
7f192e3c | 2533 | retval = put_user(pidfd, args->pidfd); |
b3e58382 CB |
2534 | if (retval) |
2535 | goto bad_fork_put_pidfd; | |
2536 | } | |
2537 | ||
73c10101 JA |
2538 | #ifdef CONFIG_BLOCK |
2539 | p->plug = NULL; | |
2540 | #endif | |
ba31c1a4 TG |
2541 | futex_init_task(p); |
2542 | ||
f9a3879a GM |
2543 | /* |
2544 | * sigaltstack should be cleared when sharing the same VM | |
2545 | */ | |
2546 | if ((clone_flags & (CLONE_VM|CLONE_VFORK)) == CLONE_VM) | |
2a742138 | 2547 | sas_ss_reset(p); |
f9a3879a | 2548 | |
1da177e4 | 2549 | /* |
6580807d ON |
2550 | * Syscall tracing and stepping should be turned off in the |
2551 | * child regardless of CLONE_PTRACE. | |
1da177e4 | 2552 | */ |
6580807d | 2553 | user_disable_single_step(p); |
64c19ba2 | 2554 | clear_task_syscall_work(p, SYSCALL_TRACE); |
64eb35f7 GKB |
2555 | #if defined(CONFIG_GENERIC_ENTRY) || defined(TIF_SYSCALL_EMU) |
2556 | clear_task_syscall_work(p, SYSCALL_EMU); | |
ed75e8d5 | 2557 | #endif |
e02c9b0d | 2558 | clear_tsk_latency_tracing(p); |
1da177e4 | 2559 | |
1da177e4 | 2560 | /* ok, now we should be set up.. */ |
18c830df ON |
2561 | p->pid = pid_nr(pid); |
2562 | if (clone_flags & CLONE_THREAD) { | |
18c830df ON |
2563 | p->group_leader = current->group_leader; |
2564 | p->tgid = current->tgid; | |
2565 | } else { | |
18c830df ON |
2566 | p->group_leader = p; |
2567 | p->tgid = p->pid; | |
2568 | } | |
5f8aadd8 | 2569 | |
9d823e8f WF |
2570 | p->nr_dirtied = 0; |
2571 | p->nr_dirtied_pause = 128 >> (PAGE_SHIFT - 10); | |
83712358 | 2572 | p->dirty_paused_when = 0; |
9d823e8f | 2573 | |
bb8cbbfe | 2574 | p->pdeath_signal = 0; |
158e1645 | 2575 | p->task_works = NULL; |
ca7752ca | 2576 | clear_posix_cputimers_work(p); |
1da177e4 | 2577 | |
d741bf41 PZ |
2578 | #ifdef CONFIG_KRETPROBES |
2579 | p->kretprobe_instances.first = NULL; | |
2580 | #endif | |
54ecbe6f MH |
2581 | #ifdef CONFIG_RETHOOK |
2582 | p->rethooks.first = NULL; | |
2583 | #endif | |
d741bf41 | 2584 | |
7e47682e AS |
2585 | /* |
2586 | * Ensure that the cgroup subsystem policies allow the new process to be | |
7b7b8a2c | 2587 | * forked. It should be noted that the new process's css_set can be changed |
7e47682e AS |
2588 | * between here and cgroup_post_fork() if an organisation operation is in |
2589 | * progress. | |
2590 | */ | |
ef2c41cf | 2591 | retval = cgroup_can_fork(p, args); |
7e47682e | 2592 | if (retval) |
5a5cf5cb | 2593 | goto bad_fork_put_pidfd; |
7e47682e | 2594 | |
b1e82065 PZ |
2595 | /* |
2596 | * Now that the cgroups are pinned, re-clone the parent cgroup and put | |
2597 | * the new task on the correct runqueue. All this *before* the task | |
2598 | * becomes visible. | |
2599 | * | |
2600 | * This isn't part of ->can_fork() because while the re-cloning is | |
2601 | * cgroup specific, it unconditionally needs to place the task on a | |
2602 | * runqueue. | |
2603 | */ | |
2604 | sched_cgroup_fork(p, args); | |
2605 | ||
7b558513 DH |
2606 | /* |
2607 | * From this point on we must avoid any synchronous user-space | |
2608 | * communication until we take the tasklist-lock. In particular, we do | |
2609 | * not want user-space to be able to predict the process start-time by | |
2610 | * stalling fork(2) after we recorded the start_time but before it is | |
2611 | * visible to the system. | |
2612 | */ | |
2613 | ||
2614 | p->start_time = ktime_get_ns(); | |
cf25e24d | 2615 | p->start_boottime = ktime_get_boottime_ns(); |
7b558513 | 2616 | |
18c830df ON |
2617 | /* |
2618 | * Make it visible to the rest of the system, but dont wake it up yet. | |
2619 | * Need tasklist lock for parent etc handling! | |
2620 | */ | |
1da177e4 LT |
2621 | write_lock_irq(&tasklist_lock); |
2622 | ||
1da177e4 | 2623 | /* CLONE_PARENT re-uses the old parent */ |
2d5516cb | 2624 | if (clone_flags & (CLONE_PARENT|CLONE_THREAD)) { |
1da177e4 | 2625 | p->real_parent = current->real_parent; |
2d5516cb | 2626 | p->parent_exec_id = current->parent_exec_id; |
b4e00444 EW |
2627 | if (clone_flags & CLONE_THREAD) |
2628 | p->exit_signal = -1; | |
2629 | else | |
2630 | p->exit_signal = current->group_leader->exit_signal; | |
2d5516cb | 2631 | } else { |
1da177e4 | 2632 | p->real_parent = current; |
2d5516cb | 2633 | p->parent_exec_id = current->self_exec_id; |
b4e00444 | 2634 | p->exit_signal = args->exit_signal; |
2d5516cb | 2635 | } |
1da177e4 | 2636 | |
d83a7cb3 JP |
2637 | klp_copy_process(p); |
2638 | ||
85dd3f61 PZ |
2639 | sched_core_fork(p); |
2640 | ||
3f17da69 | 2641 | spin_lock(¤t->sighand->siglock); |
4a2c7a78 | 2642 | |
79257534 DBO |
2643 | rv_task_fork(p); |
2644 | ||
d7822b1e MD |
2645 | rseq_fork(p, clone_flags); |
2646 | ||
4ca1d3ee | 2647 | /* Don't start children in a dying pid namespace */ |
e8cfbc24 | 2648 | if (unlikely(!(ns_of_pid(pid)->pid_allocated & PIDNS_ADDING))) { |
3fd37226 KT |
2649 | retval = -ENOMEM; |
2650 | goto bad_fork_cancel_cgroup; | |
2651 | } | |
4a2c7a78 | 2652 | |
7673bf55 EB |
2653 | /* Let kill terminate clone/fork in the middle */ |
2654 | if (fatal_signal_pending(current)) { | |
2655 | retval = -EINTR; | |
2656 | goto bad_fork_cancel_cgroup; | |
2657 | } | |
2658 | ||
a1140cb2 KI |
2659 | /* No more failure paths after this point. */ |
2660 | ||
2661 | /* | |
2662 | * Copy seccomp details explicitly here, in case they were changed | |
2663 | * before holding sighand lock. | |
2664 | */ | |
2665 | copy_seccomp(p); | |
2666 | ||
2c470475 | 2667 | init_task_pid_links(p); |
73b9ebfe | 2668 | if (likely(p->pid)) { |
4b9d33e6 | 2669 | ptrace_init_task(p, (clone_flags & CLONE_PTRACE) || trace); |
73b9ebfe | 2670 | |
81907739 | 2671 | init_task_pid(p, PIDTYPE_PID, pid); |
73b9ebfe | 2672 | if (thread_group_leader(p)) { |
6883f81a | 2673 | init_task_pid(p, PIDTYPE_TGID, pid); |
81907739 ON |
2674 | init_task_pid(p, PIDTYPE_PGID, task_pgrp(current)); |
2675 | init_task_pid(p, PIDTYPE_SID, task_session(current)); | |
2676 | ||
1c4042c2 | 2677 | if (is_child_reaper(pid)) { |
17cf22c3 | 2678 | ns_of_pid(pid)->child_reaper = p; |
1c4042c2 EB |
2679 | p->signal->flags |= SIGNAL_UNKILLABLE; |
2680 | } | |
c3ad2c3b | 2681 | p->signal->shared_pending.signal = delayed.signal; |
9c9f4ded | 2682 | p->signal->tty = tty_kref_get(current->signal->tty); |
749860ce PT |
2683 | /* |
2684 | * Inherit has_child_subreaper flag under the same | |
2685 | * tasklist_lock with adding child to the process tree | |
2686 | * for propagate_has_child_subreaper optimization. | |
2687 | */ | |
2688 | p->signal->has_child_subreaper = p->real_parent->signal->has_child_subreaper || | |
2689 | p->real_parent->signal->is_child_subreaper; | |
9cd80bbb | 2690 | list_add_tail(&p->sibling, &p->real_parent->children); |
5e85d4ab | 2691 | list_add_tail_rcu(&p->tasks, &init_task.tasks); |
6883f81a | 2692 | attach_pid(p, PIDTYPE_TGID); |
81907739 ON |
2693 | attach_pid(p, PIDTYPE_PGID); |
2694 | attach_pid(p, PIDTYPE_SID); | |
909ea964 | 2695 | __this_cpu_inc(process_counts); |
80628ca0 ON |
2696 | } else { |
2697 | current->signal->nr_threads++; | |
d80f7d7b | 2698 | current->signal->quick_threads++; |
80628ca0 | 2699 | atomic_inc(¤t->signal->live); |
60d4de3f | 2700 | refcount_inc(¤t->signal->sigcnt); |
924de3b8 | 2701 | task_join_group_stop(p); |
0c740d0a ON |
2702 | list_add_tail_rcu(&p->thread_node, |
2703 | &p->signal->thread_head); | |
73b9ebfe | 2704 | } |
81907739 | 2705 | attach_pid(p, PIDTYPE_PID); |
73b9ebfe | 2706 | nr_threads++; |
1da177e4 | 2707 | } |
1da177e4 | 2708 | total_forks++; |
c3ad2c3b | 2709 | hlist_del_init(&delayed.node); |
3f17da69 | 2710 | spin_unlock(¤t->sighand->siglock); |
4af4206b | 2711 | syscall_tracepoint_update(p); |
1da177e4 | 2712 | write_unlock_irq(&tasklist_lock); |
4af4206b | 2713 | |
ddc204b5 WL |
2714 | if (pidfile) |
2715 | fd_install(pidfd, pidfile); | |
2716 | ||
c13cf856 | 2717 | proc_fork_connector(p); |
b1e82065 | 2718 | sched_post_fork(p); |
ef2c41cf | 2719 | cgroup_post_fork(p, args); |
cdd6c482 | 2720 | perf_event_fork(p); |
43d2b113 KH |
2721 | |
2722 | trace_task_newtask(p, clone_flags); | |
3ab67966 | 2723 | uprobe_copy_process(p, clone_flags); |
fd593511 | 2724 | user_events_fork(p, clone_flags); |
43d2b113 | 2725 | |
67197a4f SB |
2726 | copy_oom_score_adj(clone_flags, p); |
2727 | ||
1da177e4 LT |
2728 | return p; |
2729 | ||
7e47682e | 2730 | bad_fork_cancel_cgroup: |
85dd3f61 | 2731 | sched_core_free(p); |
3fd37226 KT |
2732 | spin_unlock(¤t->sighand->siglock); |
2733 | write_unlock_irq(&tasklist_lock); | |
ef2c41cf | 2734 | cgroup_cancel_fork(p, args); |
b3e58382 | 2735 | bad_fork_put_pidfd: |
6fd2fe49 AV |
2736 | if (clone_flags & CLONE_PIDFD) { |
2737 | fput(pidfile); | |
2738 | put_unused_fd(pidfd); | |
2739 | } | |
425fb2b4 PE |
2740 | bad_fork_free_pid: |
2741 | if (pid != &init_struct_pid) | |
2742 | free_pid(pid); | |
0740aa5f JS |
2743 | bad_fork_cleanup_thread: |
2744 | exit_thread(p); | |
fd0928df | 2745 | bad_fork_cleanup_io: |
b69f2292 LR |
2746 | if (p->io_context) |
2747 | exit_io_context(p); | |
ab516013 | 2748 | bad_fork_cleanup_namespaces: |
444f378b | 2749 | exit_task_namespaces(p); |
1da177e4 | 2750 | bad_fork_cleanup_mm: |
c3f3ce04 AA |
2751 | if (p->mm) { |
2752 | mm_clear_owner(p->mm, p); | |
1da177e4 | 2753 | mmput(p->mm); |
c3f3ce04 | 2754 | } |
1da177e4 | 2755 | bad_fork_cleanup_signal: |
4ab6c083 | 2756 | if (!(clone_flags & CLONE_THREAD)) |
1c5354de | 2757 | free_signal_struct(p->signal); |
1da177e4 | 2758 | bad_fork_cleanup_sighand: |
a7e5328a | 2759 | __cleanup_sighand(p->sighand); |
1da177e4 LT |
2760 | bad_fork_cleanup_fs: |
2761 | exit_fs(p); /* blocking */ | |
2762 | bad_fork_cleanup_files: | |
2763 | exit_files(p); /* blocking */ | |
2764 | bad_fork_cleanup_semundo: | |
2765 | exit_sem(p); | |
e4e55b47 TH |
2766 | bad_fork_cleanup_security: |
2767 | security_task_free(p); | |
1da177e4 LT |
2768 | bad_fork_cleanup_audit: |
2769 | audit_free(p); | |
6c72e350 | 2770 | bad_fork_cleanup_perf: |
cdd6c482 | 2771 | perf_event_free_task(p); |
6c72e350 | 2772 | bad_fork_cleanup_policy: |
b09be676 | 2773 | lockdep_free_task(p); |
1da177e4 | 2774 | #ifdef CONFIG_NUMA |
f0be3d32 | 2775 | mpol_put(p->mempolicy); |
1da177e4 | 2776 | #endif |
ff8288ff | 2777 | bad_fork_cleanup_delayacct: |
35df17c5 | 2778 | delayacct_tsk_free(p); |
1da177e4 | 2779 | bad_fork_cleanup_count: |
21d1c5e3 | 2780 | dec_rlimit_ucounts(task_ucounts(p), UCOUNT_RLIMIT_NPROC, 1); |
e0e81739 | 2781 | exit_creds(p); |
1da177e4 | 2782 | bad_fork_free: |
2f064a59 | 2783 | WRITE_ONCE(p->__state, TASK_DEAD); |
1a03d3f1 | 2784 | exit_task_stack_account(p); |
68f24b08 | 2785 | put_task_stack(p); |
c3f3ce04 | 2786 | delayed_free_task(p); |
fe7d37d1 | 2787 | fork_out: |
c3ad2c3b EB |
2788 | spin_lock_irq(¤t->sighand->siglock); |
2789 | hlist_del_init(&delayed.node); | |
2790 | spin_unlock_irq(¤t->sighand->siglock); | |
fe7d37d1 | 2791 | return ERR_PTR(retval); |
1da177e4 LT |
2792 | } |
2793 | ||
2c470475 | 2794 | static inline void init_idle_pids(struct task_struct *idle) |
f106eee1 ON |
2795 | { |
2796 | enum pid_type type; | |
2797 | ||
2798 | for (type = PIDTYPE_PID; type < PIDTYPE_MAX; ++type) { | |
2c470475 EB |
2799 | INIT_HLIST_NODE(&idle->pid_links[type]); /* not really needed */ |
2800 | init_task_pid(idle, type, &init_struct_pid); | |
f106eee1 ON |
2801 | } |
2802 | } | |
2803 | ||
36cb0e1c EB |
2804 | static int idle_dummy(void *dummy) |
2805 | { | |
2806 | /* This function is never called */ | |
2807 | return 0; | |
2808 | } | |
2809 | ||
f1a0a376 | 2810 | struct task_struct * __init fork_idle(int cpu) |
1da177e4 | 2811 | { |
36c8b586 | 2812 | struct task_struct *task; |
7f192e3c | 2813 | struct kernel_clone_args args = { |
343f4c49 | 2814 | .flags = CLONE_VM, |
5bd2e97c EB |
2815 | .fn = &idle_dummy, |
2816 | .fn_arg = NULL, | |
343f4c49 | 2817 | .kthread = 1, |
36cb0e1c | 2818 | .idle = 1, |
7f192e3c CB |
2819 | }; |
2820 | ||
2821 | task = copy_process(&init_struct_pid, 0, cpu_to_node(cpu), &args); | |
f106eee1 | 2822 | if (!IS_ERR(task)) { |
2c470475 | 2823 | init_idle_pids(task); |
753ca4f3 | 2824 | init_idle(task, cpu); |
f106eee1 | 2825 | } |
73b9ebfe | 2826 | |
1da177e4 LT |
2827 | return task; |
2828 | } | |
2829 | ||
cc440e87 JA |
2830 | /* |
2831 | * This is like kernel_clone(), but shaved down and tailored to just | |
2832 | * creating io_uring workers. It returns a created task, or an error pointer. | |
2833 | * The returned task is inactive, and the caller must fire it up through | |
2834 | * wake_up_new_task(p). All signals are blocked in the created task. | |
2835 | */ | |
2836 | struct task_struct *create_io_thread(int (*fn)(void *), void *arg, int node) | |
2837 | { | |
2838 | unsigned long flags = CLONE_FS|CLONE_FILES|CLONE_SIGHAND|CLONE_THREAD| | |
2839 | CLONE_IO; | |
2840 | struct kernel_clone_args args = { | |
2841 | .flags = ((lower_32_bits(flags) | CLONE_VM | | |
2842 | CLONE_UNTRACED) & ~CSIGNAL), | |
2843 | .exit_signal = (lower_32_bits(flags) & CSIGNAL), | |
5bd2e97c EB |
2844 | .fn = fn, |
2845 | .fn_arg = arg, | |
cc440e87 | 2846 | .io_thread = 1, |
54e6842d | 2847 | .user_worker = 1, |
cc440e87 | 2848 | }; |
cc440e87 | 2849 | |
b16b3855 | 2850 | return copy_process(NULL, 0, node, &args); |
cc440e87 JA |
2851 | } |
2852 | ||
1da177e4 LT |
2853 | /* |
2854 | * Ok, this is the main fork-routine. | |
2855 | * | |
2856 | * It copies the process, and if successful kick-starts | |
2857 | * it and waits for it to finish using the VM if required. | |
a0eb9abd ES |
2858 | * |
2859 | * args->exit_signal is expected to be checked for sanity by the caller. | |
1da177e4 | 2860 | */ |
cad6967a | 2861 | pid_t kernel_clone(struct kernel_clone_args *args) |
1da177e4 | 2862 | { |
7f192e3c | 2863 | u64 clone_flags = args->flags; |
9f5325aa MPS |
2864 | struct completion vfork; |
2865 | struct pid *pid; | |
1da177e4 LT |
2866 | struct task_struct *p; |
2867 | int trace = 0; | |
cad6967a | 2868 | pid_t nr; |
1da177e4 | 2869 | |
3af8588c CB |
2870 | /* |
2871 | * For legacy clone() calls, CLONE_PIDFD uses the parent_tid argument | |
2872 | * to return the pidfd. Hence, CLONE_PIDFD and CLONE_PARENT_SETTID are | |
2873 | * mutually exclusive. With clone3() CLONE_PIDFD has grown a separate | |
2874 | * field in struct clone_args and it still doesn't make sense to have | |
2875 | * them both point at the same memory location. Performing this check | |
2876 | * here has the advantage that we don't need to have a separate helper | |
2877 | * to check for legacy clone(). | |
2878 | */ | |
2879 | if ((args->flags & CLONE_PIDFD) && | |
2880 | (args->flags & CLONE_PARENT_SETTID) && | |
2881 | (args->pidfd == args->parent_tid)) | |
2882 | return -EINVAL; | |
2883 | ||
09a05394 | 2884 | /* |
4b9d33e6 TH |
2885 | * Determine whether and which event to report to ptracer. When |
2886 | * called from kernel_thread or CLONE_UNTRACED is explicitly | |
2887 | * requested, no event is reported; otherwise, report if the event | |
2888 | * for the type of forking is enabled. | |
09a05394 | 2889 | */ |
e80d6661 | 2890 | if (!(clone_flags & CLONE_UNTRACED)) { |
4b9d33e6 TH |
2891 | if (clone_flags & CLONE_VFORK) |
2892 | trace = PTRACE_EVENT_VFORK; | |
7f192e3c | 2893 | else if (args->exit_signal != SIGCHLD) |
4b9d33e6 TH |
2894 | trace = PTRACE_EVENT_CLONE; |
2895 | else | |
2896 | trace = PTRACE_EVENT_FORK; | |
2897 | ||
2898 | if (likely(!ptrace_event_enabled(current, trace))) | |
2899 | trace = 0; | |
2900 | } | |
1da177e4 | 2901 | |
7f192e3c | 2902 | p = copy_process(NULL, trace, NUMA_NO_NODE, args); |
38addce8 | 2903 | add_latent_entropy(); |
9f5325aa MPS |
2904 | |
2905 | if (IS_ERR(p)) | |
2906 | return PTR_ERR(p); | |
2907 | ||
1da177e4 LT |
2908 | /* |
2909 | * Do this prior waking up the new thread - the thread pointer | |
2910 | * might get invalid after that point, if the thread exits quickly. | |
2911 | */ | |
9f5325aa | 2912 | trace_sched_process_fork(current, p); |
0a16b607 | 2913 | |
9f5325aa MPS |
2914 | pid = get_task_pid(p, PIDTYPE_PID); |
2915 | nr = pid_vnr(pid); | |
30e49c26 | 2916 | |
9f5325aa | 2917 | if (clone_flags & CLONE_PARENT_SETTID) |
7f192e3c | 2918 | put_user(nr, args->parent_tid); |
a6f5e063 | 2919 | |
9f5325aa MPS |
2920 | if (clone_flags & CLONE_VFORK) { |
2921 | p->vfork_done = &vfork; | |
2922 | init_completion(&vfork); | |
2923 | get_task_struct(p); | |
2924 | } | |
1da177e4 | 2925 | |
61dd3f24 | 2926 | if (IS_ENABLED(CONFIG_LRU_GEN_WALKS_MMU) && !(clone_flags & CLONE_VM)) { |
bd74fdae YZ |
2927 | /* lock the task to synchronize with memcg migration */ |
2928 | task_lock(p); | |
2929 | lru_gen_add_mm(p->mm); | |
2930 | task_unlock(p); | |
2931 | } | |
2932 | ||
9f5325aa | 2933 | wake_up_new_task(p); |
09a05394 | 2934 | |
9f5325aa MPS |
2935 | /* forking complete and child started to run, tell ptracer */ |
2936 | if (unlikely(trace)) | |
2937 | ptrace_event_pid(trace, pid); | |
4e52365f | 2938 | |
9f5325aa MPS |
2939 | if (clone_flags & CLONE_VFORK) { |
2940 | if (!wait_for_vfork_done(p, &vfork)) | |
2941 | ptrace_event_pid(PTRACE_EVENT_VFORK_DONE, pid); | |
1da177e4 | 2942 | } |
9f5325aa MPS |
2943 | |
2944 | put_pid(pid); | |
92476d7f | 2945 | return nr; |
1da177e4 LT |
2946 | } |
2947 | ||
2aa3a7f8 AV |
2948 | /* |
2949 | * Create a kernel thread. | |
2950 | */ | |
cf587db2 MC |
2951 | pid_t kernel_thread(int (*fn)(void *), void *arg, const char *name, |
2952 | unsigned long flags) | |
2aa3a7f8 | 2953 | { |
7f192e3c | 2954 | struct kernel_clone_args args = { |
3f2c788a CB |
2955 | .flags = ((lower_32_bits(flags) | CLONE_VM | |
2956 | CLONE_UNTRACED) & ~CSIGNAL), | |
2957 | .exit_signal = (lower_32_bits(flags) & CSIGNAL), | |
5bd2e97c EB |
2958 | .fn = fn, |
2959 | .fn_arg = arg, | |
cf587db2 | 2960 | .name = name, |
343f4c49 EB |
2961 | .kthread = 1, |
2962 | }; | |
2963 | ||
2964 | return kernel_clone(&args); | |
2965 | } | |
2966 | ||
2967 | /* | |
2968 | * Create a user mode thread. | |
2969 | */ | |
2970 | pid_t user_mode_thread(int (*fn)(void *), void *arg, unsigned long flags) | |
2aa3a7f8 | 2971 | { |
7f192e3c | 2972 | struct kernel_clone_args args = { |
3f2c788a CB |
2973 | .flags = ((lower_32_bits(flags) | CLONE_VM | |
2974 | CLONE_UNTRACED) & ~CSIGNAL), | |
2975 | .exit_signal = (lower_32_bits(flags) & CSIGNAL), | |
5bd2e97c EB |
2976 | .fn = fn, |
2977 | .fn_arg = arg, | |
7f192e3c CB |
2978 | }; |
2979 | ||
cad6967a | 2980 | return kernel_clone(&args); |
2aa3a7f8 | 2981 | } |
2aa3a7f8 | 2982 | |
d2125043 AV |
2983 | #ifdef __ARCH_WANT_SYS_FORK |
2984 | SYSCALL_DEFINE0(fork) | |
2985 | { | |
2986 | #ifdef CONFIG_MMU | |
7f192e3c CB |
2987 | struct kernel_clone_args args = { |
2988 | .exit_signal = SIGCHLD, | |
2989 | }; | |
2990 | ||
cad6967a | 2991 | return kernel_clone(&args); |
d2125043 AV |
2992 | #else |
2993 | /* can not support in nommu mode */ | |
5d59e182 | 2994 | return -EINVAL; |
d2125043 AV |
2995 | #endif |
2996 | } | |
2997 | #endif | |
2998 | ||
2999 | #ifdef __ARCH_WANT_SYS_VFORK | |
3000 | SYSCALL_DEFINE0(vfork) | |
3001 | { | |
7f192e3c CB |
3002 | struct kernel_clone_args args = { |
3003 | .flags = CLONE_VFORK | CLONE_VM, | |
3004 | .exit_signal = SIGCHLD, | |
3005 | }; | |
3006 | ||
cad6967a | 3007 | return kernel_clone(&args); |
d2125043 AV |
3008 | } |
3009 | #endif | |
3010 | ||
3011 | #ifdef __ARCH_WANT_SYS_CLONE | |
3012 | #ifdef CONFIG_CLONE_BACKWARDS | |
3013 | SYSCALL_DEFINE5(clone, unsigned long, clone_flags, unsigned long, newsp, | |
3014 | int __user *, parent_tidptr, | |
3033f14a | 3015 | unsigned long, tls, |
d2125043 AV |
3016 | int __user *, child_tidptr) |
3017 | #elif defined(CONFIG_CLONE_BACKWARDS2) | |
3018 | SYSCALL_DEFINE5(clone, unsigned long, newsp, unsigned long, clone_flags, | |
3019 | int __user *, parent_tidptr, | |
3020 | int __user *, child_tidptr, | |
3033f14a | 3021 | unsigned long, tls) |
dfa9771a MS |
3022 | #elif defined(CONFIG_CLONE_BACKWARDS3) |
3023 | SYSCALL_DEFINE6(clone, unsigned long, clone_flags, unsigned long, newsp, | |
3024 | int, stack_size, | |
3025 | int __user *, parent_tidptr, | |
3026 | int __user *, child_tidptr, | |
3033f14a | 3027 | unsigned long, tls) |
d2125043 AV |
3028 | #else |
3029 | SYSCALL_DEFINE5(clone, unsigned long, clone_flags, unsigned long, newsp, | |
3030 | int __user *, parent_tidptr, | |
3031 | int __user *, child_tidptr, | |
3033f14a | 3032 | unsigned long, tls) |
d2125043 AV |
3033 | #endif |
3034 | { | |
7f192e3c | 3035 | struct kernel_clone_args args = { |
3f2c788a | 3036 | .flags = (lower_32_bits(clone_flags) & ~CSIGNAL), |
7f192e3c CB |
3037 | .pidfd = parent_tidptr, |
3038 | .child_tid = child_tidptr, | |
3039 | .parent_tid = parent_tidptr, | |
3f2c788a | 3040 | .exit_signal = (lower_32_bits(clone_flags) & CSIGNAL), |
7f192e3c CB |
3041 | .stack = newsp, |
3042 | .tls = tls, | |
3043 | }; | |
3044 | ||
cad6967a | 3045 | return kernel_clone(&args); |
7f192e3c | 3046 | } |
d68dbb0c | 3047 | #endif |
7f192e3c | 3048 | |
d68dbb0c | 3049 | #ifdef __ARCH_WANT_SYS_CLONE3 |
dd499f7a | 3050 | |
7f192e3c CB |
3051 | noinline static int copy_clone_args_from_user(struct kernel_clone_args *kargs, |
3052 | struct clone_args __user *uargs, | |
f14c234b | 3053 | size_t usize) |
7f192e3c | 3054 | { |
f14c234b | 3055 | int err; |
7f192e3c | 3056 | struct clone_args args; |
49cb2fc4 | 3057 | pid_t *kset_tid = kargs->set_tid; |
7f192e3c | 3058 | |
a966dcfe ES |
3059 | BUILD_BUG_ON(offsetofend(struct clone_args, tls) != |
3060 | CLONE_ARGS_SIZE_VER0); | |
3061 | BUILD_BUG_ON(offsetofend(struct clone_args, set_tid_size) != | |
3062 | CLONE_ARGS_SIZE_VER1); | |
3063 | BUILD_BUG_ON(offsetofend(struct clone_args, cgroup) != | |
3064 | CLONE_ARGS_SIZE_VER2); | |
3065 | BUILD_BUG_ON(sizeof(struct clone_args) != CLONE_ARGS_SIZE_VER2); | |
3066 | ||
f14c234b | 3067 | if (unlikely(usize > PAGE_SIZE)) |
7f192e3c | 3068 | return -E2BIG; |
f14c234b | 3069 | if (unlikely(usize < CLONE_ARGS_SIZE_VER0)) |
7f192e3c CB |
3070 | return -EINVAL; |
3071 | ||
f14c234b AS |
3072 | err = copy_struct_from_user(&args, sizeof(args), uargs, usize); |
3073 | if (err) | |
3074 | return err; | |
7f192e3c | 3075 | |
49cb2fc4 AR |
3076 | if (unlikely(args.set_tid_size > MAX_PID_NS_LEVEL)) |
3077 | return -EINVAL; | |
3078 | ||
3079 | if (unlikely(!args.set_tid && args.set_tid_size > 0)) | |
3080 | return -EINVAL; | |
3081 | ||
3082 | if (unlikely(args.set_tid && args.set_tid_size == 0)) | |
3083 | return -EINVAL; | |
3084 | ||
a0eb9abd ES |
3085 | /* |
3086 | * Verify that higher 32bits of exit_signal are unset and that | |
3087 | * it is a valid signal | |
3088 | */ | |
3089 | if (unlikely((args.exit_signal & ~((u64)CSIGNAL)) || | |
3090 | !valid_signal(args.exit_signal))) | |
3091 | return -EINVAL; | |
3092 | ||
62173872 ES |
3093 | if ((args.flags & CLONE_INTO_CGROUP) && |
3094 | (args.cgroup > INT_MAX || usize < CLONE_ARGS_SIZE_VER2)) | |
ef2c41cf CB |
3095 | return -EINVAL; |
3096 | ||
7f192e3c CB |
3097 | *kargs = (struct kernel_clone_args){ |
3098 | .flags = args.flags, | |
3099 | .pidfd = u64_to_user_ptr(args.pidfd), | |
3100 | .child_tid = u64_to_user_ptr(args.child_tid), | |
3101 | .parent_tid = u64_to_user_ptr(args.parent_tid), | |
3102 | .exit_signal = args.exit_signal, | |
3103 | .stack = args.stack, | |
3104 | .stack_size = args.stack_size, | |
3105 | .tls = args.tls, | |
49cb2fc4 | 3106 | .set_tid_size = args.set_tid_size, |
ef2c41cf | 3107 | .cgroup = args.cgroup, |
7f192e3c CB |
3108 | }; |
3109 | ||
49cb2fc4 AR |
3110 | if (args.set_tid && |
3111 | copy_from_user(kset_tid, u64_to_user_ptr(args.set_tid), | |
3112 | (kargs->set_tid_size * sizeof(pid_t)))) | |
3113 | return -EFAULT; | |
3114 | ||
3115 | kargs->set_tid = kset_tid; | |
3116 | ||
7f192e3c CB |
3117 | return 0; |
3118 | } | |
3119 | ||
fa729c4d CB |
3120 | /** |
3121 | * clone3_stack_valid - check and prepare stack | |
3122 | * @kargs: kernel clone args | |
3123 | * | |
3124 | * Verify that the stack arguments userspace gave us are sane. | |
3125 | * In addition, set the stack direction for userspace since it's easy for us to | |
3126 | * determine. | |
3127 | */ | |
3128 | static inline bool clone3_stack_valid(struct kernel_clone_args *kargs) | |
3129 | { | |
3130 | if (kargs->stack == 0) { | |
3131 | if (kargs->stack_size > 0) | |
3132 | return false; | |
3133 | } else { | |
3134 | if (kargs->stack_size == 0) | |
3135 | return false; | |
3136 | ||
3137 | if (!access_ok((void __user *)kargs->stack, kargs->stack_size)) | |
3138 | return false; | |
3139 | ||
cf8e8658 | 3140 | #if !defined(CONFIG_STACK_GROWSUP) |
fa729c4d CB |
3141 | kargs->stack += kargs->stack_size; |
3142 | #endif | |
3143 | } | |
3144 | ||
3145 | return true; | |
3146 | } | |
3147 | ||
3148 | static bool clone3_args_valid(struct kernel_clone_args *kargs) | |
7f192e3c | 3149 | { |
b612e5df | 3150 | /* Verify that no unknown flags are passed along. */ |
ef2c41cf CB |
3151 | if (kargs->flags & |
3152 | ~(CLONE_LEGACY_FLAGS | CLONE_CLEAR_SIGHAND | CLONE_INTO_CGROUP)) | |
7f192e3c CB |
3153 | return false; |
3154 | ||
3155 | /* | |
a8ca6b13 XC |
3156 | * - make the CLONE_DETACHED bit reusable for clone3 |
3157 | * - make the CSIGNAL bits reusable for clone3 | |
7f192e3c | 3158 | */ |
a402f1e3 | 3159 | if (kargs->flags & (CLONE_DETACHED | (CSIGNAL & (~CLONE_NEWTIME)))) |
7f192e3c CB |
3160 | return false; |
3161 | ||
b612e5df CB |
3162 | if ((kargs->flags & (CLONE_SIGHAND | CLONE_CLEAR_SIGHAND)) == |
3163 | (CLONE_SIGHAND | CLONE_CLEAR_SIGHAND)) | |
3164 | return false; | |
3165 | ||
7f192e3c CB |
3166 | if ((kargs->flags & (CLONE_THREAD | CLONE_PARENT)) && |
3167 | kargs->exit_signal) | |
3168 | return false; | |
3169 | ||
fa729c4d CB |
3170 | if (!clone3_stack_valid(kargs)) |
3171 | return false; | |
3172 | ||
7f192e3c CB |
3173 | return true; |
3174 | } | |
3175 | ||
501bd016 | 3176 | /** |
ff0712ea | 3177 | * sys_clone3 - create a new process with specific properties |
501bd016 CB |
3178 | * @uargs: argument structure |
3179 | * @size: size of @uargs | |
3180 | * | |
3181 | * clone3() is the extensible successor to clone()/clone2(). | |
3182 | * It takes a struct as argument that is versioned by its size. | |
3183 | * | |
3184 | * Return: On success, a positive PID for the child process. | |
3185 | * On error, a negative errno number. | |
3186 | */ | |
7f192e3c CB |
3187 | SYSCALL_DEFINE2(clone3, struct clone_args __user *, uargs, size_t, size) |
3188 | { | |
3189 | int err; | |
3190 | ||
3191 | struct kernel_clone_args kargs; | |
49cb2fc4 AR |
3192 | pid_t set_tid[MAX_PID_NS_LEVEL]; |
3193 | ||
3194 | kargs.set_tid = set_tid; | |
7f192e3c CB |
3195 | |
3196 | err = copy_clone_args_from_user(&kargs, uargs, size); | |
3197 | if (err) | |
3198 | return err; | |
3199 | ||
3200 | if (!clone3_args_valid(&kargs)) | |
3201 | return -EINVAL; | |
3202 | ||
cad6967a | 3203 | return kernel_clone(&kargs); |
d2125043 AV |
3204 | } |
3205 | #endif | |
3206 | ||
0f1b92cb ON |
3207 | void walk_process_tree(struct task_struct *top, proc_visitor visitor, void *data) |
3208 | { | |
3209 | struct task_struct *leader, *parent, *child; | |
3210 | int res; | |
3211 | ||
3212 | read_lock(&tasklist_lock); | |
3213 | leader = top = top->group_leader; | |
3214 | down: | |
3215 | for_each_thread(leader, parent) { | |
3216 | list_for_each_entry(child, &parent->children, sibling) { | |
3217 | res = visitor(child, data); | |
3218 | if (res) { | |
3219 | if (res < 0) | |
3220 | goto out; | |
3221 | leader = child; | |
3222 | goto down; | |
3223 | } | |
3224 | up: | |
3225 | ; | |
3226 | } | |
3227 | } | |
3228 | ||
3229 | if (leader != top) { | |
3230 | child = leader; | |
3231 | parent = child->real_parent; | |
3232 | leader = parent->group_leader; | |
3233 | goto up; | |
3234 | } | |
3235 | out: | |
3236 | read_unlock(&tasklist_lock); | |
3237 | } | |
3238 | ||
5fd63b30 RT |
3239 | #ifndef ARCH_MIN_MMSTRUCT_ALIGN |
3240 | #define ARCH_MIN_MMSTRUCT_ALIGN 0 | |
3241 | #endif | |
3242 | ||
51cc5068 | 3243 | static void sighand_ctor(void *data) |
aa1757f9 ON |
3244 | { |
3245 | struct sighand_struct *sighand = data; | |
3246 | ||
a35afb83 | 3247 | spin_lock_init(&sighand->siglock); |
b8fceee1 | 3248 | init_waitqueue_head(&sighand->signalfd_wqh); |
aa1757f9 ON |
3249 | } |
3250 | ||
af806027 | 3251 | void __init mm_cache_init(void) |
1da177e4 | 3252 | { |
c1a2f7f0 RR |
3253 | unsigned int mm_size; |
3254 | ||
af806027 PZ |
3255 | /* |
3256 | * The mm_cpumask is located at the end of mm_struct, and is | |
3257 | * dynamically sized based on the maximum CPU number this system | |
3258 | * can have, taking hotplug into account (nr_cpu_ids). | |
3259 | */ | |
af7f588d | 3260 | mm_size = sizeof(struct mm_struct) + cpumask_size() + mm_cid_size(); |
af806027 PZ |
3261 | |
3262 | mm_cachep = kmem_cache_create_usercopy("mm_struct", | |
3263 | mm_size, ARCH_MIN_MMSTRUCT_ALIGN, | |
3264 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_ACCOUNT, | |
3265 | offsetof(struct mm_struct, saved_auxv), | |
3266 | sizeof_field(struct mm_struct, saved_auxv), | |
3267 | NULL); | |
3268 | } | |
3269 | ||
3270 | void __init proc_caches_init(void) | |
3271 | { | |
1da177e4 LT |
3272 | sighand_cachep = kmem_cache_create("sighand_cache", |
3273 | sizeof(struct sighand_struct), 0, | |
5f0d5a3a | 3274 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_TYPESAFE_BY_RCU| |
75f296d9 | 3275 | SLAB_ACCOUNT, sighand_ctor); |
1da177e4 LT |
3276 | signal_cachep = kmem_cache_create("signal_cache", |
3277 | sizeof(struct signal_struct), 0, | |
75f296d9 | 3278 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_ACCOUNT, |
5d097056 | 3279 | NULL); |
20c2df83 | 3280 | files_cachep = kmem_cache_create("files_cache", |
1da177e4 | 3281 | sizeof(struct files_struct), 0, |
75f296d9 | 3282 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_ACCOUNT, |
5d097056 | 3283 | NULL); |
20c2df83 | 3284 | fs_cachep = kmem_cache_create("fs_cache", |
1da177e4 | 3285 | sizeof(struct fs_struct), 0, |
75f296d9 | 3286 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_ACCOUNT, |
5d097056 | 3287 | NULL); |
c1a2f7f0 | 3288 | |
5d097056 | 3289 | vm_area_cachep = KMEM_CACHE(vm_area_struct, SLAB_PANIC|SLAB_ACCOUNT); |
c7f8f31c SB |
3290 | #ifdef CONFIG_PER_VMA_LOCK |
3291 | vma_lock_cachep = KMEM_CACHE(vma_lock, SLAB_PANIC|SLAB_ACCOUNT); | |
3292 | #endif | |
8feae131 | 3293 | mmap_init(); |
66577193 | 3294 | nsproxy_cache_init(); |
1da177e4 | 3295 | } |
cf2e340f | 3296 | |
cf2e340f | 3297 | /* |
9bfb23fc | 3298 | * Check constraints on flags passed to the unshare system call. |
cf2e340f | 3299 | */ |
9bfb23fc | 3300 | static int check_unshare_flags(unsigned long unshare_flags) |
cf2e340f | 3301 | { |
9bfb23fc ON |
3302 | if (unshare_flags & ~(CLONE_THREAD|CLONE_FS|CLONE_NEWNS|CLONE_SIGHAND| |
3303 | CLONE_VM|CLONE_FILES|CLONE_SYSVSEM| | |
50804fe3 | 3304 | CLONE_NEWUTS|CLONE_NEWIPC|CLONE_NEWNET| |
769071ac AV |
3305 | CLONE_NEWUSER|CLONE_NEWPID|CLONE_NEWCGROUP| |
3306 | CLONE_NEWTIME)) | |
9bfb23fc | 3307 | return -EINVAL; |
cf2e340f | 3308 | /* |
12c641ab EB |
3309 | * Not implemented, but pretend it works if there is nothing |
3310 | * to unshare. Note that unsharing the address space or the | |
3311 | * signal handlers also need to unshare the signal queues (aka | |
3312 | * CLONE_THREAD). | |
cf2e340f | 3313 | */ |
9bfb23fc | 3314 | if (unshare_flags & (CLONE_THREAD | CLONE_SIGHAND | CLONE_VM)) { |
12c641ab EB |
3315 | if (!thread_group_empty(current)) |
3316 | return -EINVAL; | |
3317 | } | |
3318 | if (unshare_flags & (CLONE_SIGHAND | CLONE_VM)) { | |
d036bda7 | 3319 | if (refcount_read(¤t->sighand->count) > 1) |
12c641ab EB |
3320 | return -EINVAL; |
3321 | } | |
3322 | if (unshare_flags & CLONE_VM) { | |
3323 | if (!current_is_single_threaded()) | |
9bfb23fc ON |
3324 | return -EINVAL; |
3325 | } | |
cf2e340f JD |
3326 | |
3327 | return 0; | |
3328 | } | |
3329 | ||
3330 | /* | |
99d1419d | 3331 | * Unshare the filesystem structure if it is being shared |
cf2e340f JD |
3332 | */ |
3333 | static int unshare_fs(unsigned long unshare_flags, struct fs_struct **new_fsp) | |
3334 | { | |
3335 | struct fs_struct *fs = current->fs; | |
3336 | ||
498052bb AV |
3337 | if (!(unshare_flags & CLONE_FS) || !fs) |
3338 | return 0; | |
3339 | ||
3340 | /* don't need lock here; in the worst case we'll do useless copy */ | |
3341 | if (fs->users == 1) | |
3342 | return 0; | |
3343 | ||
3344 | *new_fsp = copy_fs_struct(fs); | |
3345 | if (!*new_fsp) | |
3346 | return -ENOMEM; | |
cf2e340f JD |
3347 | |
3348 | return 0; | |
3349 | } | |
3350 | ||
cf2e340f | 3351 | /* |
a016f338 | 3352 | * Unshare file descriptor table if it is being shared |
cf2e340f | 3353 | */ |
60997c3d CB |
3354 | int unshare_fd(unsigned long unshare_flags, unsigned int max_fds, |
3355 | struct files_struct **new_fdp) | |
cf2e340f JD |
3356 | { |
3357 | struct files_struct *fd = current->files; | |
a016f338 | 3358 | int error = 0; |
cf2e340f JD |
3359 | |
3360 | if ((unshare_flags & CLONE_FILES) && | |
a016f338 | 3361 | (fd && atomic_read(&fd->count) > 1)) { |
60997c3d | 3362 | *new_fdp = dup_fd(fd, max_fds, &error); |
a016f338 JD |
3363 | if (!*new_fdp) |
3364 | return error; | |
3365 | } | |
cf2e340f JD |
3366 | |
3367 | return 0; | |
3368 | } | |
3369 | ||
cf2e340f JD |
3370 | /* |
3371 | * unshare allows a process to 'unshare' part of the process | |
3372 | * context which was originally shared using clone. copy_* | |
cad6967a | 3373 | * functions used by kernel_clone() cannot be used here directly |
cf2e340f JD |
3374 | * because they modify an inactive task_struct that is being |
3375 | * constructed. Here we are modifying the current, active, | |
3376 | * task_struct. | |
3377 | */ | |
9b32105e | 3378 | int ksys_unshare(unsigned long unshare_flags) |
cf2e340f | 3379 | { |
cf2e340f | 3380 | struct fs_struct *fs, *new_fs = NULL; |
ba1f70dd | 3381 | struct files_struct *new_fd = NULL; |
b2e0d987 | 3382 | struct cred *new_cred = NULL; |
cf7b708c | 3383 | struct nsproxy *new_nsproxy = NULL; |
9edff4ab | 3384 | int do_sysvsem = 0; |
9bfb23fc | 3385 | int err; |
cf2e340f | 3386 | |
b2e0d987 | 3387 | /* |
faf00da5 EB |
3388 | * If unsharing a user namespace must also unshare the thread group |
3389 | * and unshare the filesystem root and working directories. | |
b2e0d987 EB |
3390 | */ |
3391 | if (unshare_flags & CLONE_NEWUSER) | |
e66eded8 | 3392 | unshare_flags |= CLONE_THREAD | CLONE_FS; |
50804fe3 EB |
3393 | /* |
3394 | * If unsharing vm, must also unshare signal handlers. | |
3395 | */ | |
3396 | if (unshare_flags & CLONE_VM) | |
3397 | unshare_flags |= CLONE_SIGHAND; | |
12c641ab EB |
3398 | /* |
3399 | * If unsharing a signal handlers, must also unshare the signal queues. | |
3400 | */ | |
3401 | if (unshare_flags & CLONE_SIGHAND) | |
3402 | unshare_flags |= CLONE_THREAD; | |
9bfb23fc ON |
3403 | /* |
3404 | * If unsharing namespace, must also unshare filesystem information. | |
3405 | */ | |
3406 | if (unshare_flags & CLONE_NEWNS) | |
3407 | unshare_flags |= CLONE_FS; | |
50804fe3 EB |
3408 | |
3409 | err = check_unshare_flags(unshare_flags); | |
3410 | if (err) | |
3411 | goto bad_unshare_out; | |
6013f67f MS |
3412 | /* |
3413 | * CLONE_NEWIPC must also detach from the undolist: after switching | |
3414 | * to a new ipc namespace, the semaphore arrays from the old | |
3415 | * namespace are unreachable. | |
3416 | */ | |
3417 | if (unshare_flags & (CLONE_NEWIPC|CLONE_SYSVSEM)) | |
9edff4ab | 3418 | do_sysvsem = 1; |
fb0a685c DRO |
3419 | err = unshare_fs(unshare_flags, &new_fs); |
3420 | if (err) | |
9bfb23fc | 3421 | goto bad_unshare_out; |
60997c3d | 3422 | err = unshare_fd(unshare_flags, NR_OPEN_MAX, &new_fd); |
fb0a685c | 3423 | if (err) |
9bfb23fc | 3424 | goto bad_unshare_cleanup_fs; |
b2e0d987 | 3425 | err = unshare_userns(unshare_flags, &new_cred); |
fb0a685c | 3426 | if (err) |
9edff4ab | 3427 | goto bad_unshare_cleanup_fd; |
b2e0d987 EB |
3428 | err = unshare_nsproxy_namespaces(unshare_flags, &new_nsproxy, |
3429 | new_cred, new_fs); | |
3430 | if (err) | |
3431 | goto bad_unshare_cleanup_cred; | |
c0b2fc31 | 3432 | |
905ae01c AG |
3433 | if (new_cred) { |
3434 | err = set_cred_ucounts(new_cred); | |
3435 | if (err) | |
3436 | goto bad_unshare_cleanup_cred; | |
3437 | } | |
3438 | ||
b2e0d987 | 3439 | if (new_fs || new_fd || do_sysvsem || new_cred || new_nsproxy) { |
9edff4ab MS |
3440 | if (do_sysvsem) { |
3441 | /* | |
3442 | * CLONE_SYSVSEM is equivalent to sys_exit(). | |
3443 | */ | |
3444 | exit_sem(current); | |
3445 | } | |
ab602f79 JM |
3446 | if (unshare_flags & CLONE_NEWIPC) { |
3447 | /* Orphan segments in old ns (see sem above). */ | |
3448 | exit_shm(current); | |
3449 | shm_init_task(current); | |
3450 | } | |
ab516013 | 3451 | |
6f977e6b | 3452 | if (new_nsproxy) |
cf7b708c | 3453 | switch_task_namespaces(current, new_nsproxy); |
cf2e340f | 3454 | |
cf7b708c PE |
3455 | task_lock(current); |
3456 | ||
cf2e340f JD |
3457 | if (new_fs) { |
3458 | fs = current->fs; | |
2a4419b5 | 3459 | spin_lock(&fs->lock); |
cf2e340f | 3460 | current->fs = new_fs; |
498052bb AV |
3461 | if (--fs->users) |
3462 | new_fs = NULL; | |
3463 | else | |
3464 | new_fs = fs; | |
2a4419b5 | 3465 | spin_unlock(&fs->lock); |
cf2e340f JD |
3466 | } |
3467 | ||
ba1f70dd RX |
3468 | if (new_fd) |
3469 | swap(current->files, new_fd); | |
cf2e340f JD |
3470 | |
3471 | task_unlock(current); | |
b2e0d987 EB |
3472 | |
3473 | if (new_cred) { | |
3474 | /* Install the new user namespace */ | |
3475 | commit_creds(new_cred); | |
3476 | new_cred = NULL; | |
3477 | } | |
cf2e340f JD |
3478 | } |
3479 | ||
e4222673 HB |
3480 | perf_event_namespaces(current); |
3481 | ||
b2e0d987 EB |
3482 | bad_unshare_cleanup_cred: |
3483 | if (new_cred) | |
3484 | put_cred(new_cred); | |
cf2e340f JD |
3485 | bad_unshare_cleanup_fd: |
3486 | if (new_fd) | |
3487 | put_files_struct(new_fd); | |
3488 | ||
cf2e340f JD |
3489 | bad_unshare_cleanup_fs: |
3490 | if (new_fs) | |
498052bb | 3491 | free_fs_struct(new_fs); |
cf2e340f | 3492 | |
cf2e340f JD |
3493 | bad_unshare_out: |
3494 | return err; | |
3495 | } | |
3b125388 | 3496 | |
9b32105e DB |
3497 | SYSCALL_DEFINE1(unshare, unsigned long, unshare_flags) |
3498 | { | |
3499 | return ksys_unshare(unshare_flags); | |
3500 | } | |
3501 | ||
3b125388 AV |
3502 | /* |
3503 | * Helper to unshare the files of the current task. | |
3504 | * We don't want to expose copy_files internals to | |
3505 | * the exec layer of the kernel. | |
3506 | */ | |
3507 | ||
1f702603 | 3508 | int unshare_files(void) |
3b125388 AV |
3509 | { |
3510 | struct task_struct *task = current; | |
1f702603 | 3511 | struct files_struct *old, *copy = NULL; |
3b125388 AV |
3512 | int error; |
3513 | ||
60997c3d | 3514 | error = unshare_fd(CLONE_FILES, NR_OPEN_MAX, ©); |
1f702603 | 3515 | if (error || !copy) |
3b125388 | 3516 | return error; |
1f702603 EB |
3517 | |
3518 | old = task->files; | |
3b125388 AV |
3519 | task_lock(task); |
3520 | task->files = copy; | |
3521 | task_unlock(task); | |
1f702603 | 3522 | put_files_struct(old); |
3b125388 AV |
3523 | return 0; |
3524 | } | |
16db3d3f HS |
3525 | |
3526 | int sysctl_max_threads(struct ctl_table *table, int write, | |
b0daa2c7 | 3527 | void *buffer, size_t *lenp, loff_t *ppos) |
16db3d3f HS |
3528 | { |
3529 | struct ctl_table t; | |
3530 | int ret; | |
3531 | int threads = max_threads; | |
b0f53dbc | 3532 | int min = 1; |
16db3d3f HS |
3533 | int max = MAX_THREADS; |
3534 | ||
3535 | t = *table; | |
3536 | t.data = &threads; | |
3537 | t.extra1 = &min; | |
3538 | t.extra2 = &max; | |
3539 | ||
3540 | ret = proc_dointvec_minmax(&t, write, buffer, lenp, ppos); | |
3541 | if (ret || !write) | |
3542 | return ret; | |
3543 | ||
b0f53dbc | 3544 | max_threads = threads; |
16db3d3f HS |
3545 | |
3546 | return 0; | |
3547 | } |