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 LT |
39 | #include <linux/key.h> |
40 | #include <linux/binfmts.h> | |
41 | #include <linux/mman.h> | |
cddb8a5c | 42 | #include <linux/mmu_notifier.h> |
1da177e4 | 43 | #include <linux/fs.h> |
615d6e87 DB |
44 | #include <linux/mm.h> |
45 | #include <linux/vmacache.h> | |
ab516013 | 46 | #include <linux/nsproxy.h> |
c59ede7b | 47 | #include <linux/capability.h> |
1da177e4 | 48 | #include <linux/cpu.h> |
b4f48b63 | 49 | #include <linux/cgroup.h> |
1da177e4 | 50 | #include <linux/security.h> |
a1e78772 | 51 | #include <linux/hugetlb.h> |
e2cfabdf | 52 | #include <linux/seccomp.h> |
1da177e4 LT |
53 | #include <linux/swap.h> |
54 | #include <linux/syscalls.h> | |
55 | #include <linux/jiffies.h> | |
56 | #include <linux/futex.h> | |
8141c7f3 | 57 | #include <linux/compat.h> |
207205a2 | 58 | #include <linux/kthread.h> |
7c3ab738 | 59 | #include <linux/task_io_accounting_ops.h> |
ab2af1f5 | 60 | #include <linux/rcupdate.h> |
1da177e4 LT |
61 | #include <linux/ptrace.h> |
62 | #include <linux/mount.h> | |
63 | #include <linux/audit.h> | |
78fb7466 | 64 | #include <linux/memcontrol.h> |
f201ae23 | 65 | #include <linux/ftrace.h> |
5e2bf014 | 66 | #include <linux/proc_fs.h> |
1da177e4 LT |
67 | #include <linux/profile.h> |
68 | #include <linux/rmap.h> | |
f8af4da3 | 69 | #include <linux/ksm.h> |
1da177e4 | 70 | #include <linux/acct.h> |
893e26e6 | 71 | #include <linux/userfaultfd_k.h> |
8f0ab514 | 72 | #include <linux/tsacct_kern.h> |
9f46080c | 73 | #include <linux/cn_proc.h> |
ba96a0c8 | 74 | #include <linux/freezer.h> |
ca74e92b | 75 | #include <linux/delayacct.h> |
ad4ecbcb | 76 | #include <linux/taskstats_kern.h> |
0a425405 | 77 | #include <linux/random.h> |
522ed776 | 78 | #include <linux/tty.h> |
fd0928df | 79 | #include <linux/blkdev.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> |
1da177e4 LT |
97 | |
98 | #include <asm/pgtable.h> | |
99 | #include <asm/pgalloc.h> | |
7c0f6ba6 | 100 | #include <linux/uaccess.h> |
1da177e4 LT |
101 | #include <asm/mmu_context.h> |
102 | #include <asm/cacheflush.h> | |
103 | #include <asm/tlbflush.h> | |
104 | ||
ad8d75ff SR |
105 | #include <trace/events/sched.h> |
106 | ||
43d2b113 KH |
107 | #define CREATE_TRACE_POINTS |
108 | #include <trace/events/task.h> | |
109 | ||
ac1b398d HS |
110 | /* |
111 | * Minimum number of threads to boot the kernel | |
112 | */ | |
113 | #define MIN_THREADS 20 | |
114 | ||
115 | /* | |
116 | * Maximum number of threads | |
117 | */ | |
118 | #define MAX_THREADS FUTEX_TID_MASK | |
119 | ||
1da177e4 LT |
120 | /* |
121 | * Protected counters by write_lock_irq(&tasklist_lock) | |
122 | */ | |
123 | unsigned long total_forks; /* Handle normal Linux uptimes. */ | |
fb0a685c | 124 | int nr_threads; /* The idle threads do not count.. */ |
1da177e4 | 125 | |
8856ae4d | 126 | static int max_threads; /* tunable limit on nr_threads */ |
1da177e4 | 127 | |
8495f7e6 SPP |
128 | #define NAMED_ARRAY_INDEX(x) [x] = __stringify(x) |
129 | ||
130 | static const char * const resident_page_types[] = { | |
131 | NAMED_ARRAY_INDEX(MM_FILEPAGES), | |
132 | NAMED_ARRAY_INDEX(MM_ANONPAGES), | |
133 | NAMED_ARRAY_INDEX(MM_SWAPENTS), | |
134 | NAMED_ARRAY_INDEX(MM_SHMEMPAGES), | |
135 | }; | |
136 | ||
1da177e4 LT |
137 | DEFINE_PER_CPU(unsigned long, process_counts) = 0; |
138 | ||
c59923a1 | 139 | __cacheline_aligned DEFINE_RWLOCK(tasklist_lock); /* outer */ |
db1466b3 PM |
140 | |
141 | #ifdef CONFIG_PROVE_RCU | |
142 | int lockdep_tasklist_lock_is_held(void) | |
143 | { | |
144 | return lockdep_is_held(&tasklist_lock); | |
145 | } | |
146 | EXPORT_SYMBOL_GPL(lockdep_tasklist_lock_is_held); | |
147 | #endif /* #ifdef CONFIG_PROVE_RCU */ | |
1da177e4 LT |
148 | |
149 | int nr_processes(void) | |
150 | { | |
151 | int cpu; | |
152 | int total = 0; | |
153 | ||
1d510750 | 154 | for_each_possible_cpu(cpu) |
1da177e4 LT |
155 | total += per_cpu(process_counts, cpu); |
156 | ||
157 | return total; | |
158 | } | |
159 | ||
f19b9f74 AM |
160 | void __weak arch_release_task_struct(struct task_struct *tsk) |
161 | { | |
162 | } | |
163 | ||
f5e10287 | 164 | #ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR |
e18b890b | 165 | static struct kmem_cache *task_struct_cachep; |
41101809 TG |
166 | |
167 | static inline struct task_struct *alloc_task_struct_node(int node) | |
168 | { | |
169 | return kmem_cache_alloc_node(task_struct_cachep, GFP_KERNEL, node); | |
170 | } | |
171 | ||
41101809 TG |
172 | static inline void free_task_struct(struct task_struct *tsk) |
173 | { | |
41101809 TG |
174 | kmem_cache_free(task_struct_cachep, tsk); |
175 | } | |
1da177e4 LT |
176 | #endif |
177 | ||
b235beea | 178 | #ifndef CONFIG_ARCH_THREAD_STACK_ALLOCATOR |
41101809 | 179 | |
0d15d74a TG |
180 | /* |
181 | * Allocate pages if THREAD_SIZE is >= PAGE_SIZE, otherwise use a | |
182 | * kmemcache based allocator. | |
183 | */ | |
ba14a194 | 184 | # if THREAD_SIZE >= PAGE_SIZE || defined(CONFIG_VMAP_STACK) |
ac496bf4 AL |
185 | |
186 | #ifdef CONFIG_VMAP_STACK | |
187 | /* | |
188 | * vmalloc() is a bit slow, and calling vfree() enough times will force a TLB | |
189 | * flush. Try to minimize the number of calls by caching stacks. | |
190 | */ | |
191 | #define NR_CACHED_STACKS 2 | |
192 | static DEFINE_PER_CPU(struct vm_struct *, cached_stacks[NR_CACHED_STACKS]); | |
19659c59 HR |
193 | |
194 | static int free_vm_stack_cache(unsigned int cpu) | |
195 | { | |
196 | struct vm_struct **cached_vm_stacks = per_cpu_ptr(cached_stacks, cpu); | |
197 | int i; | |
198 | ||
199 | for (i = 0; i < NR_CACHED_STACKS; i++) { | |
200 | struct vm_struct *vm_stack = cached_vm_stacks[i]; | |
201 | ||
202 | if (!vm_stack) | |
203 | continue; | |
204 | ||
205 | vfree(vm_stack->addr); | |
206 | cached_vm_stacks[i] = NULL; | |
207 | } | |
208 | ||
209 | return 0; | |
210 | } | |
ac496bf4 AL |
211 | #endif |
212 | ||
ba14a194 | 213 | static unsigned long *alloc_thread_stack_node(struct task_struct *tsk, int node) |
b69c49b7 | 214 | { |
ba14a194 | 215 | #ifdef CONFIG_VMAP_STACK |
ac496bf4 AL |
216 | void *stack; |
217 | int i; | |
218 | ||
ac496bf4 | 219 | for (i = 0; i < NR_CACHED_STACKS; i++) { |
112166f8 CL |
220 | struct vm_struct *s; |
221 | ||
222 | s = this_cpu_xchg(cached_stacks[i], NULL); | |
ac496bf4 AL |
223 | |
224 | if (!s) | |
225 | continue; | |
ac496bf4 | 226 | |
eafb149e DA |
227 | /* Clear the KASAN shadow of the stack. */ |
228 | kasan_unpoison_shadow(s->addr, THREAD_SIZE); | |
229 | ||
ca182551 KK |
230 | /* Clear stale pointers from reused stack. */ |
231 | memset(s->addr, 0, THREAD_SIZE); | |
e01e8063 | 232 | |
ac496bf4 | 233 | tsk->stack_vm_area = s; |
ba4a4574 | 234 | tsk->stack = s->addr; |
ac496bf4 AL |
235 | return s->addr; |
236 | } | |
ac496bf4 | 237 | |
9b6f7e16 RG |
238 | /* |
239 | * Allocated stacks are cached and later reused by new threads, | |
240 | * so memcg accounting is performed manually on assigning/releasing | |
241 | * stacks to tasks. Drop __GFP_ACCOUNT. | |
242 | */ | |
48ac3c18 | 243 | stack = __vmalloc_node_range(THREAD_SIZE, THREAD_ALIGN, |
ac496bf4 | 244 | VMALLOC_START, VMALLOC_END, |
9b6f7e16 | 245 | THREADINFO_GFP & ~__GFP_ACCOUNT, |
ac496bf4 AL |
246 | PAGE_KERNEL, |
247 | 0, node, __builtin_return_address(0)); | |
ba14a194 AL |
248 | |
249 | /* | |
250 | * We can't call find_vm_area() in interrupt context, and | |
251 | * free_thread_stack() can be called in interrupt context, | |
252 | * so cache the vm_struct. | |
253 | */ | |
5eed6f1d | 254 | if (stack) { |
ba14a194 | 255 | tsk->stack_vm_area = find_vm_area(stack); |
5eed6f1d RR |
256 | tsk->stack = stack; |
257 | } | |
ba14a194 AL |
258 | return stack; |
259 | #else | |
4949148a VD |
260 | struct page *page = alloc_pages_node(node, THREADINFO_GFP, |
261 | THREAD_SIZE_ORDER); | |
b6a84016 | 262 | |
1bf4580e AA |
263 | if (likely(page)) { |
264 | tsk->stack = page_address(page); | |
265 | return tsk->stack; | |
266 | } | |
267 | return NULL; | |
ba14a194 | 268 | #endif |
b69c49b7 FT |
269 | } |
270 | ||
ba14a194 | 271 | static inline void free_thread_stack(struct task_struct *tsk) |
b69c49b7 | 272 | { |
ac496bf4 | 273 | #ifdef CONFIG_VMAP_STACK |
9b6f7e16 RG |
274 | struct vm_struct *vm = task_stack_vm_area(tsk); |
275 | ||
276 | if (vm) { | |
ac496bf4 AL |
277 | int i; |
278 | ||
9b6f7e16 RG |
279 | for (i = 0; i < THREAD_SIZE / PAGE_SIZE; i++) { |
280 | mod_memcg_page_state(vm->pages[i], | |
281 | MEMCG_KERNEL_STACK_KB, | |
282 | -(int)(PAGE_SIZE / 1024)); | |
283 | ||
284 | memcg_kmem_uncharge(vm->pages[i], 0); | |
285 | } | |
286 | ||
ac496bf4 | 287 | for (i = 0; i < NR_CACHED_STACKS; i++) { |
112166f8 CL |
288 | if (this_cpu_cmpxchg(cached_stacks[i], |
289 | NULL, tsk->stack_vm_area) != NULL) | |
ac496bf4 AL |
290 | continue; |
291 | ||
ac496bf4 AL |
292 | return; |
293 | } | |
ac496bf4 | 294 | |
0f110a9b | 295 | vfree_atomic(tsk->stack); |
ac496bf4 AL |
296 | return; |
297 | } | |
298 | #endif | |
299 | ||
300 | __free_pages(virt_to_page(tsk->stack), THREAD_SIZE_ORDER); | |
b69c49b7 | 301 | } |
0d15d74a | 302 | # else |
b235beea | 303 | static struct kmem_cache *thread_stack_cache; |
0d15d74a | 304 | |
9521d399 | 305 | static unsigned long *alloc_thread_stack_node(struct task_struct *tsk, |
0d15d74a TG |
306 | int node) |
307 | { | |
5eed6f1d RR |
308 | unsigned long *stack; |
309 | stack = kmem_cache_alloc_node(thread_stack_cache, THREADINFO_GFP, node); | |
310 | tsk->stack = stack; | |
311 | return stack; | |
0d15d74a TG |
312 | } |
313 | ||
ba14a194 | 314 | static void free_thread_stack(struct task_struct *tsk) |
0d15d74a | 315 | { |
ba14a194 | 316 | kmem_cache_free(thread_stack_cache, tsk->stack); |
0d15d74a TG |
317 | } |
318 | ||
b235beea | 319 | void thread_stack_cache_init(void) |
0d15d74a | 320 | { |
f9d29946 DW |
321 | thread_stack_cache = kmem_cache_create_usercopy("thread_stack", |
322 | THREAD_SIZE, THREAD_SIZE, 0, 0, | |
323 | THREAD_SIZE, NULL); | |
b235beea | 324 | BUG_ON(thread_stack_cache == NULL); |
0d15d74a TG |
325 | } |
326 | # endif | |
b69c49b7 FT |
327 | #endif |
328 | ||
1da177e4 | 329 | /* SLAB cache for signal_struct structures (tsk->signal) */ |
e18b890b | 330 | static struct kmem_cache *signal_cachep; |
1da177e4 LT |
331 | |
332 | /* SLAB cache for sighand_struct structures (tsk->sighand) */ | |
e18b890b | 333 | struct kmem_cache *sighand_cachep; |
1da177e4 LT |
334 | |
335 | /* SLAB cache for files_struct structures (tsk->files) */ | |
e18b890b | 336 | struct kmem_cache *files_cachep; |
1da177e4 LT |
337 | |
338 | /* SLAB cache for fs_struct structures (tsk->fs) */ | |
e18b890b | 339 | struct kmem_cache *fs_cachep; |
1da177e4 LT |
340 | |
341 | /* SLAB cache for vm_area_struct structures */ | |
3928d4f5 | 342 | static struct kmem_cache *vm_area_cachep; |
1da177e4 LT |
343 | |
344 | /* SLAB cache for mm_struct structures (tsk->mm) */ | |
e18b890b | 345 | static struct kmem_cache *mm_cachep; |
1da177e4 | 346 | |
490fc053 | 347 | struct vm_area_struct *vm_area_alloc(struct mm_struct *mm) |
3928d4f5 | 348 | { |
a670468f | 349 | struct vm_area_struct *vma; |
490fc053 | 350 | |
a670468f | 351 | vma = kmem_cache_alloc(vm_area_cachep, GFP_KERNEL); |
027232da KS |
352 | if (vma) |
353 | vma_init(vma, mm); | |
490fc053 | 354 | return vma; |
3928d4f5 LT |
355 | } |
356 | ||
357 | struct vm_area_struct *vm_area_dup(struct vm_area_struct *orig) | |
358 | { | |
95faf699 LT |
359 | struct vm_area_struct *new = kmem_cache_alloc(vm_area_cachep, GFP_KERNEL); |
360 | ||
361 | if (new) { | |
362 | *new = *orig; | |
363 | INIT_LIST_HEAD(&new->anon_vma_chain); | |
364 | } | |
365 | return new; | |
3928d4f5 LT |
366 | } |
367 | ||
368 | void vm_area_free(struct vm_area_struct *vma) | |
369 | { | |
370 | kmem_cache_free(vm_area_cachep, vma); | |
371 | } | |
372 | ||
ba14a194 | 373 | static void account_kernel_stack(struct task_struct *tsk, int account) |
c6a7f572 | 374 | { |
ba14a194 AL |
375 | void *stack = task_stack_page(tsk); |
376 | struct vm_struct *vm = task_stack_vm_area(tsk); | |
377 | ||
378 | BUILD_BUG_ON(IS_ENABLED(CONFIG_VMAP_STACK) && PAGE_SIZE % 1024 != 0); | |
379 | ||
380 | if (vm) { | |
381 | int i; | |
382 | ||
383 | BUG_ON(vm->nr_pages != THREAD_SIZE / PAGE_SIZE); | |
384 | ||
385 | for (i = 0; i < THREAD_SIZE / PAGE_SIZE; i++) { | |
386 | mod_zone_page_state(page_zone(vm->pages[i]), | |
387 | NR_KERNEL_STACK_KB, | |
388 | PAGE_SIZE / 1024 * account); | |
389 | } | |
ba14a194 AL |
390 | } else { |
391 | /* | |
392 | * All stack pages are in the same zone and belong to the | |
393 | * same memcg. | |
394 | */ | |
395 | struct page *first_page = virt_to_page(stack); | |
396 | ||
397 | mod_zone_page_state(page_zone(first_page), NR_KERNEL_STACK_KB, | |
398 | THREAD_SIZE / 1024 * account); | |
399 | ||
ed52be7b JW |
400 | mod_memcg_page_state(first_page, MEMCG_KERNEL_STACK_KB, |
401 | account * (THREAD_SIZE / 1024)); | |
ba14a194 | 402 | } |
c6a7f572 KM |
403 | } |
404 | ||
9b6f7e16 RG |
405 | static int memcg_charge_kernel_stack(struct task_struct *tsk) |
406 | { | |
407 | #ifdef CONFIG_VMAP_STACK | |
408 | struct vm_struct *vm = task_stack_vm_area(tsk); | |
409 | int ret; | |
410 | ||
411 | if (vm) { | |
412 | int i; | |
413 | ||
414 | for (i = 0; i < THREAD_SIZE / PAGE_SIZE; i++) { | |
415 | /* | |
416 | * If memcg_kmem_charge() fails, page->mem_cgroup | |
417 | * pointer is NULL, and both memcg_kmem_uncharge() | |
418 | * and mod_memcg_page_state() in free_thread_stack() | |
419 | * will ignore this page. So it's safe. | |
420 | */ | |
421 | ret = memcg_kmem_charge(vm->pages[i], GFP_KERNEL, 0); | |
422 | if (ret) | |
423 | return ret; | |
424 | ||
425 | mod_memcg_page_state(vm->pages[i], | |
426 | MEMCG_KERNEL_STACK_KB, | |
427 | PAGE_SIZE / 1024); | |
428 | } | |
429 | } | |
430 | #endif | |
431 | return 0; | |
432 | } | |
433 | ||
68f24b08 | 434 | static void release_task_stack(struct task_struct *tsk) |
1da177e4 | 435 | { |
405c0759 AL |
436 | if (WARN_ON(tsk->state != TASK_DEAD)) |
437 | return; /* Better to leak the stack than to free prematurely */ | |
438 | ||
ba14a194 | 439 | account_kernel_stack(tsk, -1); |
ba14a194 | 440 | free_thread_stack(tsk); |
68f24b08 AL |
441 | tsk->stack = NULL; |
442 | #ifdef CONFIG_VMAP_STACK | |
443 | tsk->stack_vm_area = NULL; | |
444 | #endif | |
445 | } | |
446 | ||
447 | #ifdef CONFIG_THREAD_INFO_IN_TASK | |
448 | void put_task_stack(struct task_struct *tsk) | |
449 | { | |
f0b89d39 | 450 | if (refcount_dec_and_test(&tsk->stack_refcount)) |
68f24b08 AL |
451 | release_task_stack(tsk); |
452 | } | |
453 | #endif | |
454 | ||
455 | void free_task(struct task_struct *tsk) | |
456 | { | |
457 | #ifndef CONFIG_THREAD_INFO_IN_TASK | |
458 | /* | |
459 | * The task is finally done with both the stack and thread_info, | |
460 | * so free both. | |
461 | */ | |
462 | release_task_stack(tsk); | |
463 | #else | |
464 | /* | |
465 | * If the task had a separate stack allocation, it should be gone | |
466 | * by now. | |
467 | */ | |
f0b89d39 | 468 | WARN_ON_ONCE(refcount_read(&tsk->stack_refcount) != 0); |
68f24b08 | 469 | #endif |
23f78d4a | 470 | rt_mutex_debug_task_free(tsk); |
fb52607a | 471 | ftrace_graph_exit_task(tsk); |
e2cfabdf | 472 | put_seccomp_filter(tsk); |
f19b9f74 | 473 | arch_release_task_struct(tsk); |
1da5c46f ON |
474 | if (tsk->flags & PF_KTHREAD) |
475 | free_kthread_struct(tsk); | |
1da177e4 LT |
476 | free_task_struct(tsk); |
477 | } | |
478 | EXPORT_SYMBOL(free_task); | |
479 | ||
d70f2a14 AM |
480 | #ifdef CONFIG_MMU |
481 | static __latent_entropy int dup_mmap(struct mm_struct *mm, | |
482 | struct mm_struct *oldmm) | |
483 | { | |
484 | struct vm_area_struct *mpnt, *tmp, *prev, **pprev; | |
485 | struct rb_node **rb_link, *rb_parent; | |
486 | int retval; | |
487 | unsigned long charge; | |
488 | LIST_HEAD(uf); | |
489 | ||
490 | uprobe_start_dup_mmap(); | |
491 | if (down_write_killable(&oldmm->mmap_sem)) { | |
492 | retval = -EINTR; | |
493 | goto fail_uprobe_end; | |
494 | } | |
495 | flush_cache_dup_mm(oldmm); | |
496 | uprobe_dup_mmap(oldmm, mm); | |
497 | /* | |
498 | * Not linked in yet - no deadlock potential: | |
499 | */ | |
500 | down_write_nested(&mm->mmap_sem, SINGLE_DEPTH_NESTING); | |
501 | ||
502 | /* No ordering required: file already has been exposed. */ | |
503 | RCU_INIT_POINTER(mm->exe_file, get_mm_exe_file(oldmm)); | |
504 | ||
505 | mm->total_vm = oldmm->total_vm; | |
506 | mm->data_vm = oldmm->data_vm; | |
507 | mm->exec_vm = oldmm->exec_vm; | |
508 | mm->stack_vm = oldmm->stack_vm; | |
509 | ||
510 | rb_link = &mm->mm_rb.rb_node; | |
511 | rb_parent = NULL; | |
512 | pprev = &mm->mmap; | |
513 | retval = ksm_fork(mm, oldmm); | |
514 | if (retval) | |
515 | goto out; | |
516 | retval = khugepaged_fork(mm, oldmm); | |
517 | if (retval) | |
518 | goto out; | |
519 | ||
520 | prev = NULL; | |
521 | for (mpnt = oldmm->mmap; mpnt; mpnt = mpnt->vm_next) { | |
522 | struct file *file; | |
523 | ||
524 | if (mpnt->vm_flags & VM_DONTCOPY) { | |
525 | vm_stat_account(mm, mpnt->vm_flags, -vma_pages(mpnt)); | |
526 | continue; | |
527 | } | |
528 | charge = 0; | |
655c79bb TH |
529 | /* |
530 | * Don't duplicate many vmas if we've been oom-killed (for | |
531 | * example) | |
532 | */ | |
533 | if (fatal_signal_pending(current)) { | |
534 | retval = -EINTR; | |
535 | goto out; | |
536 | } | |
d70f2a14 AM |
537 | if (mpnt->vm_flags & VM_ACCOUNT) { |
538 | unsigned long len = vma_pages(mpnt); | |
539 | ||
540 | if (security_vm_enough_memory_mm(oldmm, len)) /* sic */ | |
541 | goto fail_nomem; | |
542 | charge = len; | |
543 | } | |
3928d4f5 | 544 | tmp = vm_area_dup(mpnt); |
d70f2a14 AM |
545 | if (!tmp) |
546 | goto fail_nomem; | |
d70f2a14 AM |
547 | retval = vma_dup_policy(mpnt, tmp); |
548 | if (retval) | |
549 | goto fail_nomem_policy; | |
550 | tmp->vm_mm = mm; | |
551 | retval = dup_userfaultfd(tmp, &uf); | |
552 | if (retval) | |
553 | goto fail_nomem_anon_vma_fork; | |
554 | if (tmp->vm_flags & VM_WIPEONFORK) { | |
555 | /* VM_WIPEONFORK gets a clean slate in the child. */ | |
556 | tmp->anon_vma = NULL; | |
557 | if (anon_vma_prepare(tmp)) | |
558 | goto fail_nomem_anon_vma_fork; | |
559 | } else if (anon_vma_fork(tmp, mpnt)) | |
560 | goto fail_nomem_anon_vma_fork; | |
561 | tmp->vm_flags &= ~(VM_LOCKED | VM_LOCKONFAULT); | |
562 | tmp->vm_next = tmp->vm_prev = NULL; | |
563 | file = tmp->vm_file; | |
564 | if (file) { | |
565 | struct inode *inode = file_inode(file); | |
566 | struct address_space *mapping = file->f_mapping; | |
567 | ||
568 | get_file(file); | |
569 | if (tmp->vm_flags & VM_DENYWRITE) | |
570 | atomic_dec(&inode->i_writecount); | |
571 | i_mmap_lock_write(mapping); | |
572 | if (tmp->vm_flags & VM_SHARED) | |
573 | atomic_inc(&mapping->i_mmap_writable); | |
574 | flush_dcache_mmap_lock(mapping); | |
575 | /* insert tmp into the share list, just after mpnt */ | |
576 | vma_interval_tree_insert_after(tmp, mpnt, | |
577 | &mapping->i_mmap); | |
578 | flush_dcache_mmap_unlock(mapping); | |
579 | i_mmap_unlock_write(mapping); | |
580 | } | |
581 | ||
582 | /* | |
583 | * Clear hugetlb-related page reserves for children. This only | |
584 | * affects MAP_PRIVATE mappings. Faults generated by the child | |
585 | * are not guaranteed to succeed, even if read-only | |
586 | */ | |
587 | if (is_vm_hugetlb_page(tmp)) | |
588 | reset_vma_resv_huge_pages(tmp); | |
589 | ||
590 | /* | |
591 | * Link in the new vma and copy the page table entries. | |
592 | */ | |
593 | *pprev = tmp; | |
594 | pprev = &tmp->vm_next; | |
595 | tmp->vm_prev = prev; | |
596 | prev = tmp; | |
597 | ||
598 | __vma_link_rb(mm, tmp, rb_link, rb_parent); | |
599 | rb_link = &tmp->vm_rb.rb_right; | |
600 | rb_parent = &tmp->vm_rb; | |
601 | ||
602 | mm->map_count++; | |
603 | if (!(tmp->vm_flags & VM_WIPEONFORK)) | |
604 | retval = copy_page_range(mm, oldmm, mpnt); | |
605 | ||
606 | if (tmp->vm_ops && tmp->vm_ops->open) | |
607 | tmp->vm_ops->open(tmp); | |
608 | ||
609 | if (retval) | |
610 | goto out; | |
611 | } | |
612 | /* a new mm has just been created */ | |
1ed0cc5a | 613 | retval = arch_dup_mmap(oldmm, mm); |
d70f2a14 AM |
614 | out: |
615 | up_write(&mm->mmap_sem); | |
616 | flush_tlb_mm(oldmm); | |
617 | up_write(&oldmm->mmap_sem); | |
618 | dup_userfaultfd_complete(&uf); | |
619 | fail_uprobe_end: | |
620 | uprobe_end_dup_mmap(); | |
621 | return retval; | |
622 | fail_nomem_anon_vma_fork: | |
623 | mpol_put(vma_policy(tmp)); | |
624 | fail_nomem_policy: | |
3928d4f5 | 625 | vm_area_free(tmp); |
d70f2a14 AM |
626 | fail_nomem: |
627 | retval = -ENOMEM; | |
628 | vm_unacct_memory(charge); | |
629 | goto out; | |
630 | } | |
631 | ||
632 | static inline int mm_alloc_pgd(struct mm_struct *mm) | |
633 | { | |
634 | mm->pgd = pgd_alloc(mm); | |
635 | if (unlikely(!mm->pgd)) | |
636 | return -ENOMEM; | |
637 | return 0; | |
638 | } | |
639 | ||
640 | static inline void mm_free_pgd(struct mm_struct *mm) | |
641 | { | |
642 | pgd_free(mm, mm->pgd); | |
643 | } | |
644 | #else | |
645 | static int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm) | |
646 | { | |
647 | down_write(&oldmm->mmap_sem); | |
648 | RCU_INIT_POINTER(mm->exe_file, get_mm_exe_file(oldmm)); | |
649 | up_write(&oldmm->mmap_sem); | |
650 | return 0; | |
651 | } | |
652 | #define mm_alloc_pgd(mm) (0) | |
653 | #define mm_free_pgd(mm) | |
654 | #endif /* CONFIG_MMU */ | |
655 | ||
656 | static void check_mm(struct mm_struct *mm) | |
657 | { | |
658 | int i; | |
659 | ||
8495f7e6 SPP |
660 | BUILD_BUG_ON_MSG(ARRAY_SIZE(resident_page_types) != NR_MM_COUNTERS, |
661 | "Please make sure 'struct resident_page_types[]' is updated as well"); | |
662 | ||
d70f2a14 AM |
663 | for (i = 0; i < NR_MM_COUNTERS; i++) { |
664 | long x = atomic_long_read(&mm->rss_stat.count[i]); | |
665 | ||
666 | if (unlikely(x)) | |
8495f7e6 SPP |
667 | pr_alert("BUG: Bad rss-counter state mm:%p type:%s val:%ld\n", |
668 | mm, resident_page_types[i], x); | |
d70f2a14 AM |
669 | } |
670 | ||
671 | if (mm_pgtables_bytes(mm)) | |
672 | pr_alert("BUG: non-zero pgtables_bytes on freeing mm: %ld\n", | |
673 | mm_pgtables_bytes(mm)); | |
674 | ||
675 | #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS | |
676 | VM_BUG_ON_MM(mm->pmd_huge_pte, mm); | |
677 | #endif | |
678 | } | |
679 | ||
680 | #define allocate_mm() (kmem_cache_alloc(mm_cachep, GFP_KERNEL)) | |
681 | #define free_mm(mm) (kmem_cache_free(mm_cachep, (mm))) | |
682 | ||
683 | /* | |
684 | * Called when the last reference to the mm | |
685 | * is dropped: either by a lazy thread or by | |
686 | * mmput. Free the page directory and the mm. | |
687 | */ | |
d34bc48f | 688 | void __mmdrop(struct mm_struct *mm) |
d70f2a14 AM |
689 | { |
690 | BUG_ON(mm == &init_mm); | |
3eda69c9 MR |
691 | WARN_ON_ONCE(mm == current->mm); |
692 | WARN_ON_ONCE(mm == current->active_mm); | |
d70f2a14 AM |
693 | mm_free_pgd(mm); |
694 | destroy_context(mm); | |
984cfe4e | 695 | mmu_notifier_subscriptions_destroy(mm); |
d70f2a14 AM |
696 | check_mm(mm); |
697 | put_user_ns(mm->user_ns); | |
698 | free_mm(mm); | |
699 | } | |
d34bc48f | 700 | EXPORT_SYMBOL_GPL(__mmdrop); |
d70f2a14 AM |
701 | |
702 | static void mmdrop_async_fn(struct work_struct *work) | |
703 | { | |
704 | struct mm_struct *mm; | |
705 | ||
706 | mm = container_of(work, struct mm_struct, async_put_work); | |
707 | __mmdrop(mm); | |
708 | } | |
709 | ||
710 | static void mmdrop_async(struct mm_struct *mm) | |
711 | { | |
712 | if (unlikely(atomic_dec_and_test(&mm->mm_count))) { | |
713 | INIT_WORK(&mm->async_put_work, mmdrop_async_fn); | |
714 | schedule_work(&mm->async_put_work); | |
715 | } | |
716 | } | |
717 | ||
ea6d290c ON |
718 | static inline void free_signal_struct(struct signal_struct *sig) |
719 | { | |
97101eb4 | 720 | taskstats_tgid_free(sig); |
1c5354de | 721 | sched_autogroup_exit(sig); |
7283094e MH |
722 | /* |
723 | * __mmdrop is not safe to call from softirq context on x86 due to | |
724 | * pgd_dtor so postpone it to the async context | |
725 | */ | |
26db62f1 | 726 | if (sig->oom_mm) |
7283094e | 727 | mmdrop_async(sig->oom_mm); |
ea6d290c ON |
728 | kmem_cache_free(signal_cachep, sig); |
729 | } | |
730 | ||
731 | static inline void put_signal_struct(struct signal_struct *sig) | |
732 | { | |
60d4de3f | 733 | if (refcount_dec_and_test(&sig->sigcnt)) |
ea6d290c ON |
734 | free_signal_struct(sig); |
735 | } | |
736 | ||
158d9ebd | 737 | void __put_task_struct(struct task_struct *tsk) |
1da177e4 | 738 | { |
270f722d | 739 | WARN_ON(!tsk->exit_state); |
ec1d2819 | 740 | WARN_ON(refcount_read(&tsk->usage)); |
1da177e4 LT |
741 | WARN_ON(tsk == current); |
742 | ||
2e91fa7f | 743 | cgroup_free(tsk); |
16d51a59 | 744 | task_numa_free(tsk, true); |
1a2a4d06 | 745 | security_task_free(tsk); |
e0e81739 | 746 | exit_creds(tsk); |
35df17c5 | 747 | delayacct_tsk_free(tsk); |
ea6d290c | 748 | put_signal_struct(tsk->signal); |
1da177e4 LT |
749 | |
750 | if (!profile_handoff_task(tsk)) | |
751 | free_task(tsk); | |
752 | } | |
77c100c8 | 753 | EXPORT_SYMBOL_GPL(__put_task_struct); |
1da177e4 | 754 | |
6c0a9fa6 | 755 | void __init __weak arch_task_cache_init(void) { } |
61c4628b | 756 | |
ff691f6e HS |
757 | /* |
758 | * set_max_threads | |
759 | */ | |
16db3d3f | 760 | static void set_max_threads(unsigned int max_threads_suggested) |
ff691f6e | 761 | { |
ac1b398d | 762 | u64 threads; |
ca79b0c2 | 763 | unsigned long nr_pages = totalram_pages(); |
ff691f6e HS |
764 | |
765 | /* | |
ac1b398d HS |
766 | * The number of threads shall be limited such that the thread |
767 | * structures may only consume a small part of the available memory. | |
ff691f6e | 768 | */ |
3d6357de | 769 | if (fls64(nr_pages) + fls64(PAGE_SIZE) > 64) |
ac1b398d HS |
770 | threads = MAX_THREADS; |
771 | else | |
3d6357de | 772 | threads = div64_u64((u64) nr_pages * (u64) PAGE_SIZE, |
ac1b398d HS |
773 | (u64) THREAD_SIZE * 8UL); |
774 | ||
16db3d3f HS |
775 | if (threads > max_threads_suggested) |
776 | threads = max_threads_suggested; | |
777 | ||
ac1b398d | 778 | max_threads = clamp_t(u64, threads, MIN_THREADS, MAX_THREADS); |
ff691f6e HS |
779 | } |
780 | ||
5aaeb5c0 IM |
781 | #ifdef CONFIG_ARCH_WANTS_DYNAMIC_TASK_STRUCT |
782 | /* Initialized by the architecture: */ | |
783 | int arch_task_struct_size __read_mostly; | |
784 | #endif | |
0c8c0f03 | 785 | |
4189ff23 | 786 | #ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR |
5905429a KC |
787 | static void task_struct_whitelist(unsigned long *offset, unsigned long *size) |
788 | { | |
789 | /* Fetch thread_struct whitelist for the architecture. */ | |
790 | arch_thread_struct_whitelist(offset, size); | |
791 | ||
792 | /* | |
793 | * Handle zero-sized whitelist or empty thread_struct, otherwise | |
794 | * adjust offset to position of thread_struct in task_struct. | |
795 | */ | |
796 | if (unlikely(*size == 0)) | |
797 | *offset = 0; | |
798 | else | |
799 | *offset += offsetof(struct task_struct, thread); | |
800 | } | |
4189ff23 | 801 | #endif /* CONFIG_ARCH_TASK_STRUCT_ALLOCATOR */ |
5905429a | 802 | |
ff691f6e | 803 | void __init fork_init(void) |
1da177e4 | 804 | { |
25f9c081 | 805 | int i; |
f5e10287 | 806 | #ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR |
1da177e4 | 807 | #ifndef ARCH_MIN_TASKALIGN |
e274795e | 808 | #define ARCH_MIN_TASKALIGN 0 |
1da177e4 | 809 | #endif |
95cb64c1 | 810 | int align = max_t(int, L1_CACHE_BYTES, ARCH_MIN_TASKALIGN); |
5905429a | 811 | unsigned long useroffset, usersize; |
e274795e | 812 | |
1da177e4 | 813 | /* create a slab on which task_structs can be allocated */ |
5905429a KC |
814 | task_struct_whitelist(&useroffset, &usersize); |
815 | task_struct_cachep = kmem_cache_create_usercopy("task_struct", | |
e274795e | 816 | arch_task_struct_size, align, |
5905429a KC |
817 | SLAB_PANIC|SLAB_ACCOUNT, |
818 | useroffset, usersize, NULL); | |
1da177e4 LT |
819 | #endif |
820 | ||
61c4628b SS |
821 | /* do the arch specific task caches init */ |
822 | arch_task_cache_init(); | |
823 | ||
16db3d3f | 824 | set_max_threads(MAX_THREADS); |
1da177e4 LT |
825 | |
826 | init_task.signal->rlim[RLIMIT_NPROC].rlim_cur = max_threads/2; | |
827 | init_task.signal->rlim[RLIMIT_NPROC].rlim_max = max_threads/2; | |
828 | init_task.signal->rlim[RLIMIT_SIGPENDING] = | |
829 | init_task.signal->rlim[RLIMIT_NPROC]; | |
b376c3e1 | 830 | |
25f9c081 EB |
831 | for (i = 0; i < UCOUNT_COUNTS; i++) { |
832 | init_user_ns.ucount_max[i] = max_threads/2; | |
833 | } | |
19659c59 HR |
834 | |
835 | #ifdef CONFIG_VMAP_STACK | |
836 | cpuhp_setup_state(CPUHP_BP_PREPARE_DYN, "fork:vm_stack_cache", | |
837 | NULL, free_vm_stack_cache); | |
838 | #endif | |
b09be676 BP |
839 | |
840 | lockdep_init_task(&init_task); | |
aad42dd4 | 841 | uprobes_init(); |
1da177e4 LT |
842 | } |
843 | ||
52f5684c | 844 | int __weak arch_dup_task_struct(struct task_struct *dst, |
61c4628b SS |
845 | struct task_struct *src) |
846 | { | |
847 | *dst = *src; | |
848 | return 0; | |
849 | } | |
850 | ||
d4311ff1 AT |
851 | void set_task_stack_end_magic(struct task_struct *tsk) |
852 | { | |
853 | unsigned long *stackend; | |
854 | ||
855 | stackend = end_of_stack(tsk); | |
856 | *stackend = STACK_END_MAGIC; /* for overflow detection */ | |
857 | } | |
858 | ||
725fc629 | 859 | static struct task_struct *dup_task_struct(struct task_struct *orig, int node) |
1da177e4 LT |
860 | { |
861 | struct task_struct *tsk; | |
b235beea | 862 | unsigned long *stack; |
0f4991e8 | 863 | struct vm_struct *stack_vm_area __maybe_unused; |
3e26c149 | 864 | int err; |
1da177e4 | 865 | |
725fc629 AK |
866 | if (node == NUMA_NO_NODE) |
867 | node = tsk_fork_get_node(orig); | |
504f52b5 | 868 | tsk = alloc_task_struct_node(node); |
1da177e4 LT |
869 | if (!tsk) |
870 | return NULL; | |
871 | ||
b235beea LT |
872 | stack = alloc_thread_stack_node(tsk, node); |
873 | if (!stack) | |
f19b9f74 | 874 | goto free_tsk; |
1da177e4 | 875 | |
9b6f7e16 RG |
876 | if (memcg_charge_kernel_stack(tsk)) |
877 | goto free_stack; | |
878 | ||
ba14a194 AL |
879 | stack_vm_area = task_stack_vm_area(tsk); |
880 | ||
fb0a685c | 881 | err = arch_dup_task_struct(tsk, orig); |
ba14a194 AL |
882 | |
883 | /* | |
884 | * arch_dup_task_struct() clobbers the stack-related fields. Make | |
885 | * sure they're properly initialized before using any stack-related | |
886 | * functions again. | |
887 | */ | |
888 | tsk->stack = stack; | |
889 | #ifdef CONFIG_VMAP_STACK | |
890 | tsk->stack_vm_area = stack_vm_area; | |
891 | #endif | |
68f24b08 | 892 | #ifdef CONFIG_THREAD_INFO_IN_TASK |
f0b89d39 | 893 | refcount_set(&tsk->stack_refcount, 1); |
68f24b08 | 894 | #endif |
ba14a194 | 895 | |
164c33c6 | 896 | if (err) |
b235beea | 897 | goto free_stack; |
164c33c6 | 898 | |
dbd95212 KC |
899 | #ifdef CONFIG_SECCOMP |
900 | /* | |
901 | * We must handle setting up seccomp filters once we're under | |
902 | * the sighand lock in case orig has changed between now and | |
903 | * then. Until then, filter must be NULL to avoid messing up | |
904 | * the usage counts on the error path calling free_task. | |
905 | */ | |
906 | tsk->seccomp.filter = NULL; | |
907 | #endif | |
87bec58a AM |
908 | |
909 | setup_thread_stack(tsk, orig); | |
8e7cac79 | 910 | clear_user_return_notifier(tsk); |
f26f9aff | 911 | clear_tsk_need_resched(tsk); |
d4311ff1 | 912 | set_task_stack_end_magic(tsk); |
1da177e4 | 913 | |
050e9baa | 914 | #ifdef CONFIG_STACKPROTECTOR |
7cd815bc | 915 | tsk->stack_canary = get_random_canary(); |
0a425405 | 916 | #endif |
3bd37062 SAS |
917 | if (orig->cpus_ptr == &orig->cpus_mask) |
918 | tsk->cpus_ptr = &tsk->cpus_mask; | |
0a425405 | 919 | |
fb0a685c | 920 | /* |
0ff7b2cf EB |
921 | * One for the user space visible state that goes away when reaped. |
922 | * One for the scheduler. | |
fb0a685c | 923 | */ |
0ff7b2cf EB |
924 | refcount_set(&tsk->rcu_users, 2); |
925 | /* One for the rcu users */ | |
926 | refcount_set(&tsk->usage, 1); | |
6c5c9341 | 927 | #ifdef CONFIG_BLK_DEV_IO_TRACE |
2056a782 | 928 | tsk->btrace_seq = 0; |
6c5c9341 | 929 | #endif |
a0aa7f68 | 930 | tsk->splice_pipe = NULL; |
5640f768 | 931 | tsk->task_frag.page = NULL; |
093e5840 | 932 | tsk->wake_q.next = NULL; |
c6a7f572 | 933 | |
ba14a194 | 934 | account_kernel_stack(tsk, 1); |
c6a7f572 | 935 | |
5c9a8750 DV |
936 | kcov_task_init(tsk); |
937 | ||
e41d5818 DV |
938 | #ifdef CONFIG_FAULT_INJECTION |
939 | tsk->fail_nth = 0; | |
940 | #endif | |
941 | ||
2c323017 JB |
942 | #ifdef CONFIG_BLK_CGROUP |
943 | tsk->throttle_queue = NULL; | |
944 | tsk->use_memdelay = 0; | |
945 | #endif | |
946 | ||
d46eb14b SB |
947 | #ifdef CONFIG_MEMCG |
948 | tsk->active_memcg = NULL; | |
949 | #endif | |
1da177e4 | 950 | return tsk; |
61c4628b | 951 | |
b235beea | 952 | free_stack: |
ba14a194 | 953 | free_thread_stack(tsk); |
f19b9f74 | 954 | free_tsk: |
61c4628b SS |
955 | free_task_struct(tsk); |
956 | return NULL; | |
1da177e4 LT |
957 | } |
958 | ||
23ff4440 | 959 | __cacheline_aligned_in_smp DEFINE_SPINLOCK(mmlist_lock); |
1da177e4 | 960 | |
4cb0e11b HK |
961 | static unsigned long default_dump_filter = MMF_DUMP_FILTER_DEFAULT; |
962 | ||
963 | static int __init coredump_filter_setup(char *s) | |
964 | { | |
965 | default_dump_filter = | |
966 | (simple_strtoul(s, NULL, 0) << MMF_DUMP_FILTER_SHIFT) & | |
967 | MMF_DUMP_FILTER_MASK; | |
968 | return 1; | |
969 | } | |
970 | ||
971 | __setup("coredump_filter=", coredump_filter_setup); | |
972 | ||
1da177e4 LT |
973 | #include <linux/init_task.h> |
974 | ||
858f0993 AD |
975 | static void mm_init_aio(struct mm_struct *mm) |
976 | { | |
977 | #ifdef CONFIG_AIO | |
978 | spin_lock_init(&mm->ioctx_lock); | |
db446a08 | 979 | mm->ioctx_table = NULL; |
858f0993 AD |
980 | #endif |
981 | } | |
982 | ||
c3f3ce04 AA |
983 | static __always_inline void mm_clear_owner(struct mm_struct *mm, |
984 | struct task_struct *p) | |
985 | { | |
986 | #ifdef CONFIG_MEMCG | |
987 | if (mm->owner == p) | |
988 | WRITE_ONCE(mm->owner, NULL); | |
989 | #endif | |
990 | } | |
991 | ||
33144e84 VD |
992 | static void mm_init_owner(struct mm_struct *mm, struct task_struct *p) |
993 | { | |
994 | #ifdef CONFIG_MEMCG | |
995 | mm->owner = p; | |
996 | #endif | |
997 | } | |
998 | ||
355627f5 EB |
999 | static void mm_init_uprobes_state(struct mm_struct *mm) |
1000 | { | |
1001 | #ifdef CONFIG_UPROBES | |
1002 | mm->uprobes_state.xol_area = NULL; | |
1003 | #endif | |
1004 | } | |
1005 | ||
bfedb589 EB |
1006 | static struct mm_struct *mm_init(struct mm_struct *mm, struct task_struct *p, |
1007 | struct user_namespace *user_ns) | |
1da177e4 | 1008 | { |
41f727fd VD |
1009 | mm->mmap = NULL; |
1010 | mm->mm_rb = RB_ROOT; | |
1011 | mm->vmacache_seqnum = 0; | |
1da177e4 LT |
1012 | atomic_set(&mm->mm_users, 1); |
1013 | atomic_set(&mm->mm_count, 1); | |
1014 | init_rwsem(&mm->mmap_sem); | |
1015 | INIT_LIST_HEAD(&mm->mmlist); | |
999d9fc1 | 1016 | mm->core_state = NULL; |
af5b0f6a | 1017 | mm_pgtables_bytes_init(mm); |
41f727fd VD |
1018 | mm->map_count = 0; |
1019 | mm->locked_vm = 0; | |
70f8a3ca | 1020 | atomic64_set(&mm->pinned_vm, 0); |
d559db08 | 1021 | memset(&mm->rss_stat, 0, sizeof(mm->rss_stat)); |
1da177e4 | 1022 | spin_lock_init(&mm->page_table_lock); |
88aa7cc6 | 1023 | spin_lock_init(&mm->arg_lock); |
41f727fd | 1024 | mm_init_cpumask(mm); |
858f0993 | 1025 | mm_init_aio(mm); |
cf475ad2 | 1026 | mm_init_owner(mm, p); |
2b7e8665 | 1027 | RCU_INIT_POINTER(mm->exe_file, NULL); |
984cfe4e | 1028 | mmu_notifier_subscriptions_init(mm); |
16af97dc | 1029 | init_tlb_flush_pending(mm); |
41f727fd VD |
1030 | #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS |
1031 | mm->pmd_huge_pte = NULL; | |
1032 | #endif | |
355627f5 | 1033 | mm_init_uprobes_state(mm); |
1da177e4 | 1034 | |
a0715cc2 AT |
1035 | if (current->mm) { |
1036 | mm->flags = current->mm->flags & MMF_INIT_MASK; | |
1037 | mm->def_flags = current->mm->def_flags & VM_INIT_DEF_MASK; | |
1038 | } else { | |
1039 | mm->flags = default_dump_filter; | |
1da177e4 | 1040 | mm->def_flags = 0; |
a0715cc2 AT |
1041 | } |
1042 | ||
41f727fd VD |
1043 | if (mm_alloc_pgd(mm)) |
1044 | goto fail_nopgd; | |
1045 | ||
1046 | if (init_new_context(p, mm)) | |
1047 | goto fail_nocontext; | |
78fb7466 | 1048 | |
bfedb589 | 1049 | mm->user_ns = get_user_ns(user_ns); |
41f727fd VD |
1050 | return mm; |
1051 | ||
1052 | fail_nocontext: | |
1053 | mm_free_pgd(mm); | |
1054 | fail_nopgd: | |
1da177e4 LT |
1055 | free_mm(mm); |
1056 | return NULL; | |
1057 | } | |
1058 | ||
1059 | /* | |
1060 | * Allocate and initialize an mm_struct. | |
1061 | */ | |
fb0a685c | 1062 | struct mm_struct *mm_alloc(void) |
1da177e4 | 1063 | { |
fb0a685c | 1064 | struct mm_struct *mm; |
1da177e4 LT |
1065 | |
1066 | mm = allocate_mm(); | |
de03c72c KM |
1067 | if (!mm) |
1068 | return NULL; | |
1069 | ||
1070 | memset(mm, 0, sizeof(*mm)); | |
bfedb589 | 1071 | return mm_init(mm, current, current_user_ns()); |
1da177e4 LT |
1072 | } |
1073 | ||
ec8d7c14 MH |
1074 | static inline void __mmput(struct mm_struct *mm) |
1075 | { | |
1076 | VM_BUG_ON(atomic_read(&mm->mm_users)); | |
1077 | ||
1078 | uprobe_clear_state(mm); | |
1079 | exit_aio(mm); | |
1080 | ksm_exit(mm); | |
1081 | khugepaged_exit(mm); /* must run before exit_mmap */ | |
1082 | exit_mmap(mm); | |
6fcb52a5 | 1083 | mm_put_huge_zero_page(mm); |
ec8d7c14 MH |
1084 | set_mm_exe_file(mm, NULL); |
1085 | if (!list_empty(&mm->mmlist)) { | |
1086 | spin_lock(&mmlist_lock); | |
1087 | list_del(&mm->mmlist); | |
1088 | spin_unlock(&mmlist_lock); | |
1089 | } | |
1090 | if (mm->binfmt) | |
1091 | module_put(mm->binfmt->module); | |
1092 | mmdrop(mm); | |
1093 | } | |
1094 | ||
1da177e4 LT |
1095 | /* |
1096 | * Decrement the use count and release all resources for an mm. | |
1097 | */ | |
1098 | void mmput(struct mm_struct *mm) | |
1099 | { | |
0ae26f1b AM |
1100 | might_sleep(); |
1101 | ||
ec8d7c14 MH |
1102 | if (atomic_dec_and_test(&mm->mm_users)) |
1103 | __mmput(mm); | |
1104 | } | |
1105 | EXPORT_SYMBOL_GPL(mmput); | |
1106 | ||
a1b2289c SY |
1107 | #ifdef CONFIG_MMU |
1108 | static void mmput_async_fn(struct work_struct *work) | |
1109 | { | |
1110 | struct mm_struct *mm = container_of(work, struct mm_struct, | |
1111 | async_put_work); | |
1112 | ||
1113 | __mmput(mm); | |
1114 | } | |
1115 | ||
1116 | void mmput_async(struct mm_struct *mm) | |
1117 | { | |
1118 | if (atomic_dec_and_test(&mm->mm_users)) { | |
1119 | INIT_WORK(&mm->async_put_work, mmput_async_fn); | |
1120 | schedule_work(&mm->async_put_work); | |
1121 | } | |
1122 | } | |
1123 | #endif | |
1124 | ||
90f31d0e KK |
1125 | /** |
1126 | * set_mm_exe_file - change a reference to the mm's executable file | |
1127 | * | |
1128 | * This changes mm's executable file (shown as symlink /proc/[pid]/exe). | |
1129 | * | |
6e399cd1 DB |
1130 | * Main users are mmput() and sys_execve(). Callers prevent concurrent |
1131 | * invocations: in mmput() nobody alive left, in execve task is single | |
1132 | * threaded. sys_prctl(PR_SET_MM_MAP/EXE_FILE) also needs to set the | |
1133 | * mm->exe_file, but does so without using set_mm_exe_file() in order | |
1134 | * to do avoid the need for any locks. | |
90f31d0e | 1135 | */ |
38646013 JS |
1136 | void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file) |
1137 | { | |
6e399cd1 DB |
1138 | struct file *old_exe_file; |
1139 | ||
1140 | /* | |
1141 | * It is safe to dereference the exe_file without RCU as | |
1142 | * this function is only called if nobody else can access | |
1143 | * this mm -- see comment above for justification. | |
1144 | */ | |
1145 | old_exe_file = rcu_dereference_raw(mm->exe_file); | |
90f31d0e | 1146 | |
38646013 JS |
1147 | if (new_exe_file) |
1148 | get_file(new_exe_file); | |
90f31d0e KK |
1149 | rcu_assign_pointer(mm->exe_file, new_exe_file); |
1150 | if (old_exe_file) | |
1151 | fput(old_exe_file); | |
38646013 JS |
1152 | } |
1153 | ||
90f31d0e KK |
1154 | /** |
1155 | * get_mm_exe_file - acquire a reference to the mm's executable file | |
1156 | * | |
1157 | * Returns %NULL if mm has no associated executable file. | |
1158 | * User must release file via fput(). | |
1159 | */ | |
38646013 JS |
1160 | struct file *get_mm_exe_file(struct mm_struct *mm) |
1161 | { | |
1162 | struct file *exe_file; | |
1163 | ||
90f31d0e KK |
1164 | rcu_read_lock(); |
1165 | exe_file = rcu_dereference(mm->exe_file); | |
1166 | if (exe_file && !get_file_rcu(exe_file)) | |
1167 | exe_file = NULL; | |
1168 | rcu_read_unlock(); | |
38646013 JS |
1169 | return exe_file; |
1170 | } | |
11163348 | 1171 | EXPORT_SYMBOL(get_mm_exe_file); |
38646013 | 1172 | |
cd81a917 MG |
1173 | /** |
1174 | * get_task_exe_file - acquire a reference to the task's executable file | |
1175 | * | |
1176 | * Returns %NULL if task's mm (if any) has no associated executable file or | |
1177 | * this is a kernel thread with borrowed mm (see the comment above get_task_mm). | |
1178 | * User must release file via fput(). | |
1179 | */ | |
1180 | struct file *get_task_exe_file(struct task_struct *task) | |
1181 | { | |
1182 | struct file *exe_file = NULL; | |
1183 | struct mm_struct *mm; | |
1184 | ||
1185 | task_lock(task); | |
1186 | mm = task->mm; | |
1187 | if (mm) { | |
1188 | if (!(task->flags & PF_KTHREAD)) | |
1189 | exe_file = get_mm_exe_file(mm); | |
1190 | } | |
1191 | task_unlock(task); | |
1192 | return exe_file; | |
1193 | } | |
1194 | EXPORT_SYMBOL(get_task_exe_file); | |
38646013 | 1195 | |
1da177e4 LT |
1196 | /** |
1197 | * get_task_mm - acquire a reference to the task's mm | |
1198 | * | |
246bb0b1 | 1199 | * Returns %NULL if the task has no mm. Checks PF_KTHREAD (meaning |
1da177e4 LT |
1200 | * this kernel workthread has transiently adopted a user mm with use_mm, |
1201 | * to do its AIO) is not set and if so returns a reference to it, after | |
1202 | * bumping up the use count. User must release the mm via mmput() | |
1203 | * after use. Typically used by /proc and ptrace. | |
1204 | */ | |
1205 | struct mm_struct *get_task_mm(struct task_struct *task) | |
1206 | { | |
1207 | struct mm_struct *mm; | |
1208 | ||
1209 | task_lock(task); | |
1210 | mm = task->mm; | |
1211 | if (mm) { | |
246bb0b1 | 1212 | if (task->flags & PF_KTHREAD) |
1da177e4 LT |
1213 | mm = NULL; |
1214 | else | |
3fce371b | 1215 | mmget(mm); |
1da177e4 LT |
1216 | } |
1217 | task_unlock(task); | |
1218 | return mm; | |
1219 | } | |
1220 | EXPORT_SYMBOL_GPL(get_task_mm); | |
1221 | ||
8cdb878d CY |
1222 | struct mm_struct *mm_access(struct task_struct *task, unsigned int mode) |
1223 | { | |
1224 | struct mm_struct *mm; | |
1225 | int err; | |
1226 | ||
1227 | err = mutex_lock_killable(&task->signal->cred_guard_mutex); | |
1228 | if (err) | |
1229 | return ERR_PTR(err); | |
1230 | ||
1231 | mm = get_task_mm(task); | |
1232 | if (mm && mm != current->mm && | |
1233 | !ptrace_may_access(task, mode)) { | |
1234 | mmput(mm); | |
1235 | mm = ERR_PTR(-EACCES); | |
1236 | } | |
1237 | mutex_unlock(&task->signal->cred_guard_mutex); | |
1238 | ||
1239 | return mm; | |
1240 | } | |
1241 | ||
57b59c4a | 1242 | static void complete_vfork_done(struct task_struct *tsk) |
c415c3b4 | 1243 | { |
d68b46fe | 1244 | struct completion *vfork; |
c415c3b4 | 1245 | |
d68b46fe ON |
1246 | task_lock(tsk); |
1247 | vfork = tsk->vfork_done; | |
1248 | if (likely(vfork)) { | |
1249 | tsk->vfork_done = NULL; | |
1250 | complete(vfork); | |
1251 | } | |
1252 | task_unlock(tsk); | |
1253 | } | |
1254 | ||
1255 | static int wait_for_vfork_done(struct task_struct *child, | |
1256 | struct completion *vfork) | |
1257 | { | |
1258 | int killed; | |
1259 | ||
1260 | freezer_do_not_count(); | |
76f969e8 | 1261 | cgroup_enter_frozen(); |
d68b46fe | 1262 | killed = wait_for_completion_killable(vfork); |
76f969e8 | 1263 | cgroup_leave_frozen(false); |
d68b46fe ON |
1264 | freezer_count(); |
1265 | ||
1266 | if (killed) { | |
1267 | task_lock(child); | |
1268 | child->vfork_done = NULL; | |
1269 | task_unlock(child); | |
1270 | } | |
1271 | ||
1272 | put_task_struct(child); | |
1273 | return killed; | |
c415c3b4 ON |
1274 | } |
1275 | ||
1da177e4 LT |
1276 | /* Please note the differences between mmput and mm_release. |
1277 | * mmput is called whenever we stop holding onto a mm_struct, | |
1278 | * error success whatever. | |
1279 | * | |
1280 | * mm_release is called after a mm_struct has been removed | |
1281 | * from the current process. | |
1282 | * | |
1283 | * This difference is important for error handling, when we | |
1284 | * only half set up a mm_struct for a new process and need to restore | |
1285 | * the old one. Because we mmput the new mm_struct before | |
1286 | * restoring the old one. . . | |
1287 | * Eric Biederman 10 January 1998 | |
1288 | */ | |
4610ba7a | 1289 | static void mm_release(struct task_struct *tsk, struct mm_struct *mm) |
1da177e4 | 1290 | { |
0326f5a9 SD |
1291 | uprobe_free_utask(tsk); |
1292 | ||
1da177e4 LT |
1293 | /* Get rid of any cached register state */ |
1294 | deactivate_mm(tsk, mm); | |
1295 | ||
fec1d011 | 1296 | /* |
735f2770 MH |
1297 | * Signal userspace if we're not exiting with a core dump |
1298 | * because we want to leave the value intact for debugging | |
1299 | * purposes. | |
fec1d011 | 1300 | */ |
9c8a8228 | 1301 | if (tsk->clear_child_tid) { |
735f2770 | 1302 | if (!(tsk->signal->flags & SIGNAL_GROUP_COREDUMP) && |
9c8a8228 ED |
1303 | atomic_read(&mm->mm_users) > 1) { |
1304 | /* | |
1305 | * We don't check the error code - if userspace has | |
1306 | * not set up a proper pointer then tough luck. | |
1307 | */ | |
1308 | put_user(0, tsk->clear_child_tid); | |
2de0db99 DB |
1309 | do_futex(tsk->clear_child_tid, FUTEX_WAKE, |
1310 | 1, NULL, NULL, 0, 0); | |
9c8a8228 | 1311 | } |
1da177e4 | 1312 | tsk->clear_child_tid = NULL; |
1da177e4 | 1313 | } |
f7505d64 KK |
1314 | |
1315 | /* | |
1316 | * All done, finally we can wake up parent and return this mm to him. | |
1317 | * Also kthread_stop() uses this completion for synchronization. | |
1318 | */ | |
1319 | if (tsk->vfork_done) | |
1320 | complete_vfork_done(tsk); | |
1da177e4 LT |
1321 | } |
1322 | ||
4610ba7a TG |
1323 | void exit_mm_release(struct task_struct *tsk, struct mm_struct *mm) |
1324 | { | |
150d7158 | 1325 | futex_exit_release(tsk); |
4610ba7a TG |
1326 | mm_release(tsk, mm); |
1327 | } | |
1328 | ||
1329 | void exec_mm_release(struct task_struct *tsk, struct mm_struct *mm) | |
1330 | { | |
150d7158 | 1331 | futex_exec_release(tsk); |
4610ba7a TG |
1332 | mm_release(tsk, mm); |
1333 | } | |
1334 | ||
13585fa0 NA |
1335 | /** |
1336 | * dup_mm() - duplicates an existing mm structure | |
1337 | * @tsk: the task_struct with which the new mm will be associated. | |
1338 | * @oldmm: the mm to duplicate. | |
1339 | * | |
1340 | * Allocates a new mm structure and duplicates the provided @oldmm structure | |
1341 | * content into it. | |
1342 | * | |
1343 | * Return: the duplicated mm or NULL on failure. | |
a0a7ec30 | 1344 | */ |
13585fa0 NA |
1345 | static struct mm_struct *dup_mm(struct task_struct *tsk, |
1346 | struct mm_struct *oldmm) | |
a0a7ec30 | 1347 | { |
13585fa0 | 1348 | struct mm_struct *mm; |
a0a7ec30 JD |
1349 | int err; |
1350 | ||
a0a7ec30 JD |
1351 | mm = allocate_mm(); |
1352 | if (!mm) | |
1353 | goto fail_nomem; | |
1354 | ||
1355 | memcpy(mm, oldmm, sizeof(*mm)); | |
1356 | ||
bfedb589 | 1357 | if (!mm_init(mm, tsk, mm->user_ns)) |
a0a7ec30 JD |
1358 | goto fail_nomem; |
1359 | ||
a0a7ec30 JD |
1360 | err = dup_mmap(mm, oldmm); |
1361 | if (err) | |
1362 | goto free_pt; | |
1363 | ||
1364 | mm->hiwater_rss = get_mm_rss(mm); | |
1365 | mm->hiwater_vm = mm->total_vm; | |
1366 | ||
801460d0 HS |
1367 | if (mm->binfmt && !try_module_get(mm->binfmt->module)) |
1368 | goto free_pt; | |
1369 | ||
a0a7ec30 JD |
1370 | return mm; |
1371 | ||
1372 | free_pt: | |
801460d0 HS |
1373 | /* don't put binfmt in mmput, we haven't got module yet */ |
1374 | mm->binfmt = NULL; | |
c3f3ce04 | 1375 | mm_init_owner(mm, NULL); |
a0a7ec30 JD |
1376 | mmput(mm); |
1377 | ||
1378 | fail_nomem: | |
1379 | return NULL; | |
a0a7ec30 JD |
1380 | } |
1381 | ||
fb0a685c | 1382 | static int copy_mm(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 | 1383 | { |
fb0a685c | 1384 | struct mm_struct *mm, *oldmm; |
1da177e4 LT |
1385 | int retval; |
1386 | ||
1387 | tsk->min_flt = tsk->maj_flt = 0; | |
1388 | tsk->nvcsw = tsk->nivcsw = 0; | |
17406b82 MSB |
1389 | #ifdef CONFIG_DETECT_HUNG_TASK |
1390 | tsk->last_switch_count = tsk->nvcsw + tsk->nivcsw; | |
a2e51445 | 1391 | tsk->last_switch_time = 0; |
17406b82 | 1392 | #endif |
1da177e4 LT |
1393 | |
1394 | tsk->mm = NULL; | |
1395 | tsk->active_mm = NULL; | |
1396 | ||
1397 | /* | |
1398 | * Are we cloning a kernel thread? | |
1399 | * | |
1400 | * We need to steal a active VM for that.. | |
1401 | */ | |
1402 | oldmm = current->mm; | |
1403 | if (!oldmm) | |
1404 | return 0; | |
1405 | ||
615d6e87 DB |
1406 | /* initialize the new vmacache entries */ |
1407 | vmacache_flush(tsk); | |
1408 | ||
1da177e4 | 1409 | if (clone_flags & CLONE_VM) { |
3fce371b | 1410 | mmget(oldmm); |
1da177e4 | 1411 | mm = oldmm; |
1da177e4 LT |
1412 | goto good_mm; |
1413 | } | |
1414 | ||
1415 | retval = -ENOMEM; | |
13585fa0 | 1416 | mm = dup_mm(tsk, current->mm); |
1da177e4 LT |
1417 | if (!mm) |
1418 | goto fail_nomem; | |
1419 | ||
1da177e4 LT |
1420 | good_mm: |
1421 | tsk->mm = mm; | |
1422 | tsk->active_mm = mm; | |
1423 | return 0; | |
1424 | ||
1da177e4 LT |
1425 | fail_nomem: |
1426 | return retval; | |
1da177e4 LT |
1427 | } |
1428 | ||
a39bc516 | 1429 | static int copy_fs(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 | 1430 | { |
498052bb | 1431 | struct fs_struct *fs = current->fs; |
1da177e4 | 1432 | if (clone_flags & CLONE_FS) { |
498052bb | 1433 | /* tsk->fs is already what we want */ |
2a4419b5 | 1434 | spin_lock(&fs->lock); |
498052bb | 1435 | if (fs->in_exec) { |
2a4419b5 | 1436 | spin_unlock(&fs->lock); |
498052bb AV |
1437 | return -EAGAIN; |
1438 | } | |
1439 | fs->users++; | |
2a4419b5 | 1440 | spin_unlock(&fs->lock); |
1da177e4 LT |
1441 | return 0; |
1442 | } | |
498052bb | 1443 | tsk->fs = copy_fs_struct(fs); |
1da177e4 LT |
1444 | if (!tsk->fs) |
1445 | return -ENOMEM; | |
1446 | return 0; | |
1447 | } | |
1448 | ||
fb0a685c | 1449 | static int copy_files(unsigned long clone_flags, struct task_struct *tsk) |
a016f338 JD |
1450 | { |
1451 | struct files_struct *oldf, *newf; | |
1452 | int error = 0; | |
1453 | ||
1454 | /* | |
1455 | * A background process may not have any files ... | |
1456 | */ | |
1457 | oldf = current->files; | |
1458 | if (!oldf) | |
1459 | goto out; | |
1460 | ||
1461 | if (clone_flags & CLONE_FILES) { | |
1462 | atomic_inc(&oldf->count); | |
1463 | goto out; | |
1464 | } | |
1465 | ||
a016f338 JD |
1466 | newf = dup_fd(oldf, &error); |
1467 | if (!newf) | |
1468 | goto out; | |
1469 | ||
1470 | tsk->files = newf; | |
1471 | error = 0; | |
1472 | out: | |
1473 | return error; | |
1474 | } | |
1475 | ||
fadad878 | 1476 | static int copy_io(unsigned long clone_flags, struct task_struct *tsk) |
fd0928df JA |
1477 | { |
1478 | #ifdef CONFIG_BLOCK | |
1479 | struct io_context *ioc = current->io_context; | |
6e736be7 | 1480 | struct io_context *new_ioc; |
fd0928df JA |
1481 | |
1482 | if (!ioc) | |
1483 | return 0; | |
fadad878 JA |
1484 | /* |
1485 | * Share io context with parent, if CLONE_IO is set | |
1486 | */ | |
1487 | if (clone_flags & CLONE_IO) { | |
3d48749d TH |
1488 | ioc_task_link(ioc); |
1489 | tsk->io_context = ioc; | |
fadad878 | 1490 | } else if (ioprio_valid(ioc->ioprio)) { |
6e736be7 TH |
1491 | new_ioc = get_task_io_context(tsk, GFP_KERNEL, NUMA_NO_NODE); |
1492 | if (unlikely(!new_ioc)) | |
fd0928df JA |
1493 | return -ENOMEM; |
1494 | ||
6e736be7 | 1495 | new_ioc->ioprio = ioc->ioprio; |
11a3122f | 1496 | put_io_context(new_ioc); |
fd0928df JA |
1497 | } |
1498 | #endif | |
1499 | return 0; | |
1500 | } | |
1501 | ||
a39bc516 | 1502 | static int copy_sighand(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 LT |
1503 | { |
1504 | struct sighand_struct *sig; | |
1505 | ||
60348802 | 1506 | if (clone_flags & CLONE_SIGHAND) { |
d036bda7 | 1507 | refcount_inc(¤t->sighand->count); |
1da177e4 LT |
1508 | return 0; |
1509 | } | |
1510 | sig = kmem_cache_alloc(sighand_cachep, GFP_KERNEL); | |
e56d0903 | 1511 | rcu_assign_pointer(tsk->sighand, sig); |
1da177e4 LT |
1512 | if (!sig) |
1513 | return -ENOMEM; | |
9d7fb042 | 1514 | |
d036bda7 | 1515 | refcount_set(&sig->count, 1); |
06e62a46 | 1516 | spin_lock_irq(¤t->sighand->siglock); |
1da177e4 | 1517 | memcpy(sig->action, current->sighand->action, sizeof(sig->action)); |
06e62a46 | 1518 | spin_unlock_irq(¤t->sighand->siglock); |
b612e5df CB |
1519 | |
1520 | /* Reset all signal handler not set to SIG_IGN to SIG_DFL. */ | |
1521 | if (clone_flags & CLONE_CLEAR_SIGHAND) | |
1522 | flush_signal_handlers(tsk, 0); | |
1523 | ||
1da177e4 LT |
1524 | return 0; |
1525 | } | |
1526 | ||
a7e5328a | 1527 | void __cleanup_sighand(struct sighand_struct *sighand) |
c81addc9 | 1528 | { |
d036bda7 | 1529 | if (refcount_dec_and_test(&sighand->count)) { |
d80e731e | 1530 | signalfd_cleanup(sighand); |
392809b2 | 1531 | /* |
5f0d5a3a | 1532 | * sighand_cachep is SLAB_TYPESAFE_BY_RCU so we can free it |
392809b2 ON |
1533 | * without an RCU grace period, see __lock_task_sighand(). |
1534 | */ | |
c81addc9 | 1535 | kmem_cache_free(sighand_cachep, sighand); |
d80e731e | 1536 | } |
c81addc9 ON |
1537 | } |
1538 | ||
f06febc9 FM |
1539 | /* |
1540 | * Initialize POSIX timer handling for a thread group. | |
1541 | */ | |
1542 | static void posix_cpu_timers_init_group(struct signal_struct *sig) | |
1543 | { | |
2b69942f | 1544 | struct posix_cputimers *pct = &sig->posix_cputimers; |
78d7d407 JS |
1545 | unsigned long cpu_limit; |
1546 | ||
316c1608 | 1547 | cpu_limit = READ_ONCE(sig->rlim[RLIMIT_CPU].rlim_cur); |
3a245c0f | 1548 | posix_cputimers_group_init(pct, cpu_limit); |
f06febc9 FM |
1549 | } |
1550 | ||
a39bc516 | 1551 | static int copy_signal(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 LT |
1552 | { |
1553 | struct signal_struct *sig; | |
1da177e4 | 1554 | |
4ab6c083 | 1555 | if (clone_flags & CLONE_THREAD) |
490dea45 | 1556 | return 0; |
490dea45 | 1557 | |
a56704ef | 1558 | sig = kmem_cache_zalloc(signal_cachep, GFP_KERNEL); |
1da177e4 LT |
1559 | tsk->signal = sig; |
1560 | if (!sig) | |
1561 | return -ENOMEM; | |
1562 | ||
b3ac022c | 1563 | sig->nr_threads = 1; |
1da177e4 | 1564 | atomic_set(&sig->live, 1); |
60d4de3f | 1565 | refcount_set(&sig->sigcnt, 1); |
0c740d0a ON |
1566 | |
1567 | /* list_add(thread_node, thread_head) without INIT_LIST_HEAD() */ | |
1568 | sig->thread_head = (struct list_head)LIST_HEAD_INIT(tsk->thread_node); | |
1569 | tsk->thread_node = (struct list_head)LIST_HEAD_INIT(sig->thread_head); | |
1570 | ||
1da177e4 | 1571 | init_waitqueue_head(&sig->wait_chldexit); |
db51aecc | 1572 | sig->curr_target = tsk; |
1da177e4 | 1573 | init_sigpending(&sig->shared_pending); |
c3ad2c3b | 1574 | INIT_HLIST_HEAD(&sig->multiprocess); |
e78c3496 | 1575 | seqlock_init(&sig->stats_lock); |
9d7fb042 | 1576 | prev_cputime_init(&sig->prev_cputime); |
1da177e4 | 1577 | |
baa73d9e | 1578 | #ifdef CONFIG_POSIX_TIMERS |
b18b6a9c | 1579 | INIT_LIST_HEAD(&sig->posix_timers); |
c9cb2e3d | 1580 | hrtimer_init(&sig->real_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL); |
1da177e4 | 1581 | sig->real_timer.function = it_real_fn; |
baa73d9e | 1582 | #endif |
1da177e4 | 1583 | |
1da177e4 LT |
1584 | task_lock(current->group_leader); |
1585 | memcpy(sig->rlim, current->signal->rlim, sizeof sig->rlim); | |
1586 | task_unlock(current->group_leader); | |
1587 | ||
6279a751 ON |
1588 | posix_cpu_timers_init_group(sig); |
1589 | ||
522ed776 | 1590 | tty_audit_fork(sig); |
5091faa4 | 1591 | sched_autogroup_fork(sig); |
522ed776 | 1592 | |
a63d83f4 | 1593 | sig->oom_score_adj = current->signal->oom_score_adj; |
dabb16f6 | 1594 | sig->oom_score_adj_min = current->signal->oom_score_adj_min; |
28b83c51 | 1595 | |
9b1bf12d | 1596 | mutex_init(&sig->cred_guard_mutex); |
eea96732 | 1597 | mutex_init(&sig->exec_update_mutex); |
9b1bf12d | 1598 | |
1da177e4 LT |
1599 | return 0; |
1600 | } | |
1601 | ||
dbd95212 KC |
1602 | static void copy_seccomp(struct task_struct *p) |
1603 | { | |
1604 | #ifdef CONFIG_SECCOMP | |
1605 | /* | |
1606 | * Must be called with sighand->lock held, which is common to | |
1607 | * all threads in the group. Holding cred_guard_mutex is not | |
1608 | * needed because this new task is not yet running and cannot | |
1609 | * be racing exec. | |
1610 | */ | |
69f6a34b | 1611 | assert_spin_locked(¤t->sighand->siglock); |
dbd95212 KC |
1612 | |
1613 | /* Ref-count the new filter user, and assign it. */ | |
1614 | get_seccomp_filter(current); | |
1615 | p->seccomp = current->seccomp; | |
1616 | ||
1617 | /* | |
1618 | * Explicitly enable no_new_privs here in case it got set | |
1619 | * between the task_struct being duplicated and holding the | |
1620 | * sighand lock. The seccomp state and nnp must be in sync. | |
1621 | */ | |
1622 | if (task_no_new_privs(current)) | |
1623 | task_set_no_new_privs(p); | |
1624 | ||
1625 | /* | |
1626 | * If the parent gained a seccomp mode after copying thread | |
1627 | * flags and between before we held the sighand lock, we have | |
1628 | * to manually enable the seccomp thread flag here. | |
1629 | */ | |
1630 | if (p->seccomp.mode != SECCOMP_MODE_DISABLED) | |
1631 | set_tsk_thread_flag(p, TIF_SECCOMP); | |
1632 | #endif | |
1633 | } | |
1634 | ||
17da2bd9 | 1635 | SYSCALL_DEFINE1(set_tid_address, int __user *, tidptr) |
1da177e4 LT |
1636 | { |
1637 | current->clear_child_tid = tidptr; | |
1638 | ||
b488893a | 1639 | return task_pid_vnr(current); |
1da177e4 LT |
1640 | } |
1641 | ||
a39bc516 | 1642 | static void rt_mutex_init_task(struct task_struct *p) |
23f78d4a | 1643 | { |
1d615482 | 1644 | raw_spin_lock_init(&p->pi_lock); |
e29e175b | 1645 | #ifdef CONFIG_RT_MUTEXES |
a23ba907 | 1646 | p->pi_waiters = RB_ROOT_CACHED; |
e96a7705 | 1647 | p->pi_top_task = NULL; |
23f78d4a | 1648 | p->pi_blocked_on = NULL; |
23f78d4a IM |
1649 | #endif |
1650 | } | |
1651 | ||
2c470475 EB |
1652 | static inline void init_task_pid_links(struct task_struct *task) |
1653 | { | |
1654 | enum pid_type type; | |
1655 | ||
1656 | for (type = PIDTYPE_PID; type < PIDTYPE_MAX; ++type) { | |
1657 | INIT_HLIST_NODE(&task->pid_links[type]); | |
1658 | } | |
1659 | } | |
1660 | ||
81907739 ON |
1661 | static inline void |
1662 | init_task_pid(struct task_struct *task, enum pid_type type, struct pid *pid) | |
1663 | { | |
2c470475 EB |
1664 | if (type == PIDTYPE_PID) |
1665 | task->thread_pid = pid; | |
1666 | else | |
1667 | task->signal->pids[type] = pid; | |
81907739 ON |
1668 | } |
1669 | ||
6bfbaa51 IM |
1670 | static inline void rcu_copy_process(struct task_struct *p) |
1671 | { | |
1672 | #ifdef CONFIG_PREEMPT_RCU | |
1673 | p->rcu_read_lock_nesting = 0; | |
1674 | p->rcu_read_unlock_special.s = 0; | |
1675 | p->rcu_blocked_node = NULL; | |
1676 | INIT_LIST_HEAD(&p->rcu_node_entry); | |
1677 | #endif /* #ifdef CONFIG_PREEMPT_RCU */ | |
1678 | #ifdef CONFIG_TASKS_RCU | |
1679 | p->rcu_tasks_holdout = false; | |
1680 | INIT_LIST_HEAD(&p->rcu_tasks_holdout_list); | |
1681 | p->rcu_tasks_idle_cpu = -1; | |
1682 | #endif /* #ifdef CONFIG_TASKS_RCU */ | |
1683 | } | |
1684 | ||
3695eae5 CB |
1685 | struct pid *pidfd_pid(const struct file *file) |
1686 | { | |
1687 | if (file->f_op == &pidfd_fops) | |
1688 | return file->private_data; | |
1689 | ||
1690 | return ERR_PTR(-EBADF); | |
1691 | } | |
1692 | ||
b3e58382 CB |
1693 | static int pidfd_release(struct inode *inode, struct file *file) |
1694 | { | |
1695 | struct pid *pid = file->private_data; | |
1696 | ||
1697 | file->private_data = NULL; | |
1698 | put_pid(pid); | |
1699 | return 0; | |
1700 | } | |
1701 | ||
1702 | #ifdef CONFIG_PROC_FS | |
15d42eb2 CK |
1703 | /** |
1704 | * pidfd_show_fdinfo - print information about a pidfd | |
1705 | * @m: proc fdinfo file | |
1706 | * @f: file referencing a pidfd | |
1707 | * | |
1708 | * Pid: | |
1709 | * This function will print the pid that a given pidfd refers to in the | |
1710 | * pid namespace of the procfs instance. | |
1711 | * If the pid namespace of the process is not a descendant of the pid | |
1712 | * namespace of the procfs instance 0 will be shown as its pid. This is | |
1713 | * similar to calling getppid() on a process whose parent is outside of | |
1714 | * its pid namespace. | |
1715 | * | |
1716 | * NSpid: | |
1717 | * If pid namespaces are supported then this function will also print | |
1718 | * the pid of a given pidfd refers to for all descendant pid namespaces | |
1719 | * starting from the current pid namespace of the instance, i.e. the | |
1720 | * Pid field and the first entry in the NSpid field will be identical. | |
1721 | * If the pid namespace of the process is not a descendant of the pid | |
1722 | * namespace of the procfs instance 0 will be shown as its first NSpid | |
1723 | * entry and no others will be shown. | |
1724 | * Note that this differs from the Pid and NSpid fields in | |
1725 | * /proc/<pid>/status where Pid and NSpid are always shown relative to | |
1726 | * the pid namespace of the procfs instance. The difference becomes | |
1727 | * obvious when sending around a pidfd between pid namespaces from a | |
1728 | * different branch of the tree, i.e. where no ancestoral relation is | |
1729 | * present between the pid namespaces: | |
1730 | * - create two new pid namespaces ns1 and ns2 in the initial pid | |
1731 | * namespace (also take care to create new mount namespaces in the | |
1732 | * new pid namespace and mount procfs) | |
1733 | * - create a process with a pidfd in ns1 | |
1734 | * - send pidfd from ns1 to ns2 | |
1735 | * - read /proc/self/fdinfo/<pidfd> and observe that both Pid and NSpid | |
1736 | * have exactly one entry, which is 0 | |
1737 | */ | |
b3e58382 CB |
1738 | static void pidfd_show_fdinfo(struct seq_file *m, struct file *f) |
1739 | { | |
b3e58382 | 1740 | struct pid *pid = f->private_data; |
3d6d8da4 CB |
1741 | struct pid_namespace *ns; |
1742 | pid_t nr = -1; | |
15d42eb2 | 1743 | |
3d6d8da4 CB |
1744 | if (likely(pid_has_task(pid, PIDTYPE_PID))) { |
1745 | ns = proc_pid_ns(file_inode(m->file)); | |
1746 | nr = pid_nr_ns(pid, ns); | |
1747 | } | |
1748 | ||
1749 | seq_put_decimal_ll(m, "Pid:\t", nr); | |
b3e58382 | 1750 | |
15d42eb2 | 1751 | #ifdef CONFIG_PID_NS |
3d6d8da4 CB |
1752 | seq_put_decimal_ll(m, "\nNSpid:\t", nr); |
1753 | if (nr > 0) { | |
15d42eb2 | 1754 | int i; |
b3e58382 | 1755 | |
15d42eb2 CK |
1756 | /* If nr is non-zero it means that 'pid' is valid and that |
1757 | * ns, i.e. the pid namespace associated with the procfs | |
1758 | * instance, is in the pid namespace hierarchy of pid. | |
1759 | * Start at one below the already printed level. | |
1760 | */ | |
1761 | for (i = ns->level + 1; i <= pid->level; i++) | |
3d6d8da4 | 1762 | seq_put_decimal_ll(m, "\t", pid->numbers[i].nr); |
15d42eb2 CK |
1763 | } |
1764 | #endif | |
b3e58382 CB |
1765 | seq_putc(m, '\n'); |
1766 | } | |
1767 | #endif | |
1768 | ||
b53b0b9d JFG |
1769 | /* |
1770 | * Poll support for process exit notification. | |
1771 | */ | |
9e77716a | 1772 | static __poll_t pidfd_poll(struct file *file, struct poll_table_struct *pts) |
b53b0b9d JFG |
1773 | { |
1774 | struct task_struct *task; | |
1775 | struct pid *pid = file->private_data; | |
9e77716a | 1776 | __poll_t poll_flags = 0; |
b53b0b9d JFG |
1777 | |
1778 | poll_wait(file, &pid->wait_pidfd, pts); | |
1779 | ||
1780 | rcu_read_lock(); | |
1781 | task = pid_task(pid, PIDTYPE_PID); | |
1782 | /* | |
1783 | * Inform pollers only when the whole thread group exits. | |
1784 | * If the thread group leader exits before all other threads in the | |
1785 | * group, then poll(2) should block, similar to the wait(2) family. | |
1786 | */ | |
1787 | if (!task || (task->exit_state && thread_group_empty(task))) | |
9e77716a | 1788 | poll_flags = EPOLLIN | EPOLLRDNORM; |
b53b0b9d JFG |
1789 | rcu_read_unlock(); |
1790 | ||
1791 | return poll_flags; | |
1792 | } | |
1793 | ||
b3e58382 CB |
1794 | const struct file_operations pidfd_fops = { |
1795 | .release = pidfd_release, | |
b53b0b9d | 1796 | .poll = pidfd_poll, |
b3e58382 CB |
1797 | #ifdef CONFIG_PROC_FS |
1798 | .show_fdinfo = pidfd_show_fdinfo, | |
1799 | #endif | |
1800 | }; | |
1801 | ||
c3f3ce04 AA |
1802 | static void __delayed_free_task(struct rcu_head *rhp) |
1803 | { | |
1804 | struct task_struct *tsk = container_of(rhp, struct task_struct, rcu); | |
1805 | ||
1806 | free_task(tsk); | |
1807 | } | |
1808 | ||
1809 | static __always_inline void delayed_free_task(struct task_struct *tsk) | |
1810 | { | |
1811 | if (IS_ENABLED(CONFIG_MEMCG)) | |
1812 | call_rcu(&tsk->rcu, __delayed_free_task); | |
1813 | else | |
1814 | free_task(tsk); | |
1815 | } | |
1816 | ||
1da177e4 LT |
1817 | /* |
1818 | * This creates a new process as a copy of the old one, | |
1819 | * but does not actually start it yet. | |
1820 | * | |
1821 | * It copies the registers, and all the appropriate | |
1822 | * parts of the process environment (as per the clone | |
1823 | * flags). The actual kick-off is left to the caller. | |
1824 | */ | |
0766f788 | 1825 | static __latent_entropy struct task_struct *copy_process( |
09a05394 | 1826 | struct pid *pid, |
3033f14a | 1827 | int trace, |
7f192e3c CB |
1828 | int node, |
1829 | struct kernel_clone_args *args) | |
1da177e4 | 1830 | { |
b3e58382 | 1831 | int pidfd = -1, retval; |
a24efe62 | 1832 | struct task_struct *p; |
c3ad2c3b | 1833 | struct multiprocess_signals delayed; |
6fd2fe49 | 1834 | struct file *pidfile = NULL; |
7f192e3c | 1835 | u64 clone_flags = args->flags; |
769071ac | 1836 | struct nsproxy *nsp = current->nsproxy; |
1da177e4 | 1837 | |
667b6094 MPS |
1838 | /* |
1839 | * Don't allow sharing the root directory with processes in a different | |
1840 | * namespace | |
1841 | */ | |
1da177e4 LT |
1842 | if ((clone_flags & (CLONE_NEWNS|CLONE_FS)) == (CLONE_NEWNS|CLONE_FS)) |
1843 | return ERR_PTR(-EINVAL); | |
1844 | ||
e66eded8 EB |
1845 | if ((clone_flags & (CLONE_NEWUSER|CLONE_FS)) == (CLONE_NEWUSER|CLONE_FS)) |
1846 | return ERR_PTR(-EINVAL); | |
1847 | ||
1da177e4 LT |
1848 | /* |
1849 | * Thread groups must share signals as well, and detached threads | |
1850 | * can only be started up within the thread group. | |
1851 | */ | |
1852 | if ((clone_flags & CLONE_THREAD) && !(clone_flags & CLONE_SIGHAND)) | |
1853 | return ERR_PTR(-EINVAL); | |
1854 | ||
1855 | /* | |
1856 | * Shared signal handlers imply shared VM. By way of the above, | |
1857 | * thread groups also imply shared VM. Blocking this case allows | |
1858 | * for various simplifications in other code. | |
1859 | */ | |
1860 | if ((clone_flags & CLONE_SIGHAND) && !(clone_flags & CLONE_VM)) | |
1861 | return ERR_PTR(-EINVAL); | |
1862 | ||
123be07b SB |
1863 | /* |
1864 | * Siblings of global init remain as zombies on exit since they are | |
1865 | * not reaped by their parent (swapper). To solve this and to avoid | |
1866 | * multi-rooted process trees, prevent global and container-inits | |
1867 | * from creating siblings. | |
1868 | */ | |
1869 | if ((clone_flags & CLONE_PARENT) && | |
1870 | current->signal->flags & SIGNAL_UNKILLABLE) | |
1871 | return ERR_PTR(-EINVAL); | |
1872 | ||
8382fcac | 1873 | /* |
40a0d32d | 1874 | * If the new process will be in a different pid or user namespace |
faf00da5 | 1875 | * do not allow it to share a thread group with the forking task. |
8382fcac | 1876 | */ |
faf00da5 | 1877 | if (clone_flags & CLONE_THREAD) { |
40a0d32d | 1878 | if ((clone_flags & (CLONE_NEWUSER | CLONE_NEWPID)) || |
769071ac AV |
1879 | (task_active_pid_ns(current) != nsp->pid_ns_for_children)) |
1880 | return ERR_PTR(-EINVAL); | |
1881 | } | |
1882 | ||
1883 | /* | |
1884 | * If the new process will be in a different time namespace | |
1885 | * do not allow it to share VM or a thread group with the forking task. | |
1886 | */ | |
1887 | if (clone_flags & (CLONE_THREAD | CLONE_VM)) { | |
1888 | if (nsp->time_ns != nsp->time_ns_for_children) | |
40a0d32d ON |
1889 | return ERR_PTR(-EINVAL); |
1890 | } | |
8382fcac | 1891 | |
b3e58382 | 1892 | if (clone_flags & CLONE_PIDFD) { |
b3e58382 | 1893 | /* |
b3e58382 CB |
1894 | * - CLONE_DETACHED is blocked so that we can potentially |
1895 | * reuse it later for CLONE_PIDFD. | |
1896 | * - CLONE_THREAD is blocked until someone really needs it. | |
1897 | */ | |
7f192e3c | 1898 | if (clone_flags & (CLONE_DETACHED | CLONE_THREAD)) |
b3e58382 | 1899 | return ERR_PTR(-EINVAL); |
b3e58382 CB |
1900 | } |
1901 | ||
c3ad2c3b EB |
1902 | /* |
1903 | * Force any signals received before this point to be delivered | |
1904 | * before the fork happens. Collect up signals sent to multiple | |
1905 | * processes that happen during the fork and delay them so that | |
1906 | * they appear to happen after the fork. | |
1907 | */ | |
1908 | sigemptyset(&delayed.signal); | |
1909 | INIT_HLIST_NODE(&delayed.node); | |
1910 | ||
1911 | spin_lock_irq(¤t->sighand->siglock); | |
1912 | if (!(clone_flags & CLONE_THREAD)) | |
1913 | hlist_add_head(&delayed.node, ¤t->signal->multiprocess); | |
1914 | recalc_sigpending(); | |
1915 | spin_unlock_irq(¤t->sighand->siglock); | |
1916 | retval = -ERESTARTNOINTR; | |
1917 | if (signal_pending(current)) | |
1918 | goto fork_out; | |
1919 | ||
1da177e4 | 1920 | retval = -ENOMEM; |
725fc629 | 1921 | p = dup_task_struct(current, node); |
1da177e4 LT |
1922 | if (!p) |
1923 | goto fork_out; | |
1924 | ||
4d6501dc VN |
1925 | /* |
1926 | * This _must_ happen before we call free_task(), i.e. before we jump | |
1927 | * to any of the bad_fork_* labels. This is to avoid freeing | |
1928 | * p->set_child_tid which is (ab)used as a kthread's data pointer for | |
1929 | * kernel threads (PF_KTHREAD). | |
1930 | */ | |
7f192e3c | 1931 | p->set_child_tid = (clone_flags & CLONE_CHILD_SETTID) ? args->child_tid : NULL; |
4d6501dc VN |
1932 | /* |
1933 | * Clear TID on mm_release()? | |
1934 | */ | |
7f192e3c | 1935 | p->clear_child_tid = (clone_flags & CLONE_CHILD_CLEARTID) ? args->child_tid : NULL; |
4d6501dc | 1936 | |
f7e8b616 SR |
1937 | ftrace_graph_init_task(p); |
1938 | ||
bea493a0 PZ |
1939 | rt_mutex_init_task(p); |
1940 | ||
d12c1a37 | 1941 | #ifdef CONFIG_PROVE_LOCKING |
de30a2b3 IM |
1942 | DEBUG_LOCKS_WARN_ON(!p->hardirqs_enabled); |
1943 | DEBUG_LOCKS_WARN_ON(!p->softirqs_enabled); | |
1944 | #endif | |
1da177e4 | 1945 | retval = -EAGAIN; |
3b11a1de | 1946 | if (atomic_read(&p->real_cred->user->processes) >= |
78d7d407 | 1947 | task_rlimit(p, RLIMIT_NPROC)) { |
b57922b6 EP |
1948 | if (p->real_cred->user != INIT_USER && |
1949 | !capable(CAP_SYS_RESOURCE) && !capable(CAP_SYS_ADMIN)) | |
1da177e4 LT |
1950 | goto bad_fork_free; |
1951 | } | |
72fa5997 | 1952 | current->flags &= ~PF_NPROC_EXCEEDED; |
1da177e4 | 1953 | |
f1752eec DH |
1954 | retval = copy_creds(p, clone_flags); |
1955 | if (retval < 0) | |
1956 | goto bad_fork_free; | |
1da177e4 LT |
1957 | |
1958 | /* | |
1959 | * If multiple threads are within copy_process(), then this check | |
1960 | * triggers too late. This doesn't hurt, the check is only there | |
1961 | * to stop root fork bombs. | |
1962 | */ | |
04ec93fe | 1963 | retval = -EAGAIN; |
1da177e4 LT |
1964 | if (nr_threads >= max_threads) |
1965 | goto bad_fork_cleanup_count; | |
1966 | ||
ca74e92b | 1967 | delayacct_tsk_init(p); /* Must remain after dup_task_struct() */ |
c1de45ca | 1968 | p->flags &= ~(PF_SUPERPRIV | PF_WQ_WORKER | PF_IDLE); |
514ddb44 | 1969 | p->flags |= PF_FORKNOEXEC; |
1da177e4 LT |
1970 | INIT_LIST_HEAD(&p->children); |
1971 | INIT_LIST_HEAD(&p->sibling); | |
f41d911f | 1972 | rcu_copy_process(p); |
1da177e4 LT |
1973 | p->vfork_done = NULL; |
1974 | spin_lock_init(&p->alloc_lock); | |
1da177e4 | 1975 | |
1da177e4 LT |
1976 | init_sigpending(&p->pending); |
1977 | ||
64861634 | 1978 | p->utime = p->stime = p->gtime = 0; |
40565b5a | 1979 | #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME |
64861634 | 1980 | p->utimescaled = p->stimescaled = 0; |
40565b5a | 1981 | #endif |
9d7fb042 PZ |
1982 | prev_cputime_init(&p->prev_cputime); |
1983 | ||
6a61671b | 1984 | #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN |
bac5b6b6 FW |
1985 | seqcount_init(&p->vtime.seqcount); |
1986 | p->vtime.starttime = 0; | |
1987 | p->vtime.state = VTIME_INACTIVE; | |
6a61671b FW |
1988 | #endif |
1989 | ||
a3a2e76c KH |
1990 | #if defined(SPLIT_RSS_COUNTING) |
1991 | memset(&p->rss_stat, 0, sizeof(p->rss_stat)); | |
1992 | #endif | |
172ba844 | 1993 | |
6976675d AV |
1994 | p->default_timer_slack_ns = current->timer_slack_ns; |
1995 | ||
eb414681 JW |
1996 | #ifdef CONFIG_PSI |
1997 | p->psi_flags = 0; | |
1998 | #endif | |
1999 | ||
5995477a | 2000 | task_io_accounting_init(&p->ioac); |
1da177e4 LT |
2001 | acct_clear_integrals(p); |
2002 | ||
3a245c0f | 2003 | posix_cputimers_init(&p->posix_cputimers); |
1da177e4 | 2004 | |
1da177e4 | 2005 | p->io_context = NULL; |
c0b0ae8a | 2006 | audit_set_context(p, NULL); |
b4f48b63 | 2007 | cgroup_fork(p); |
1da177e4 | 2008 | #ifdef CONFIG_NUMA |
846a16bf | 2009 | p->mempolicy = mpol_dup(p->mempolicy); |
fb0a685c DRO |
2010 | if (IS_ERR(p->mempolicy)) { |
2011 | retval = PTR_ERR(p->mempolicy); | |
2012 | p->mempolicy = NULL; | |
e8604cb4 | 2013 | goto bad_fork_cleanup_threadgroup_lock; |
fb0a685c | 2014 | } |
1da177e4 | 2015 | #endif |
778d3b0f MH |
2016 | #ifdef CONFIG_CPUSETS |
2017 | p->cpuset_mem_spread_rotor = NUMA_NO_NODE; | |
2018 | p->cpuset_slab_spread_rotor = NUMA_NO_NODE; | |
cc9a6c87 | 2019 | seqcount_init(&p->mems_allowed_seq); |
778d3b0f | 2020 | #endif |
de30a2b3 IM |
2021 | #ifdef CONFIG_TRACE_IRQFLAGS |
2022 | p->irq_events = 0; | |
2023 | p->hardirqs_enabled = 0; | |
2024 | p->hardirq_enable_ip = 0; | |
2025 | p->hardirq_enable_event = 0; | |
2026 | p->hardirq_disable_ip = _THIS_IP_; | |
2027 | p->hardirq_disable_event = 0; | |
2028 | p->softirqs_enabled = 1; | |
2029 | p->softirq_enable_ip = _THIS_IP_; | |
2030 | p->softirq_enable_event = 0; | |
2031 | p->softirq_disable_ip = 0; | |
2032 | p->softirq_disable_event = 0; | |
2033 | p->hardirq_context = 0; | |
2034 | p->softirq_context = 0; | |
2035 | #endif | |
8bcbde54 DH |
2036 | |
2037 | p->pagefault_disabled = 0; | |
2038 | ||
fbb9ce95 | 2039 | #ifdef CONFIG_LOCKDEP |
b09be676 | 2040 | lockdep_init_task(p); |
fbb9ce95 | 2041 | #endif |
1da177e4 | 2042 | |
408894ee IM |
2043 | #ifdef CONFIG_DEBUG_MUTEXES |
2044 | p->blocked_on = NULL; /* not blocked yet */ | |
2045 | #endif | |
cafe5635 KO |
2046 | #ifdef CONFIG_BCACHE |
2047 | p->sequential_io = 0; | |
2048 | p->sequential_io_avg = 0; | |
2049 | #endif | |
0f481406 | 2050 | |
3c90e6e9 | 2051 | /* Perform scheduler related setup. Assign this task to a CPU. */ |
aab03e05 DF |
2052 | retval = sched_fork(clone_flags, p); |
2053 | if (retval) | |
2054 | goto bad_fork_cleanup_policy; | |
6ab423e0 | 2055 | |
cdd6c482 | 2056 | retval = perf_event_init_task(p); |
6ab423e0 PZ |
2057 | if (retval) |
2058 | goto bad_fork_cleanup_policy; | |
fb0a685c DRO |
2059 | retval = audit_alloc(p); |
2060 | if (retval) | |
6c72e350 | 2061 | goto bad_fork_cleanup_perf; |
1da177e4 | 2062 | /* copy all the process information */ |
ab602f79 | 2063 | shm_init_task(p); |
e4e55b47 | 2064 | retval = security_task_alloc(p, clone_flags); |
fb0a685c | 2065 | if (retval) |
1da177e4 | 2066 | goto bad_fork_cleanup_audit; |
e4e55b47 TH |
2067 | retval = copy_semundo(clone_flags, p); |
2068 | if (retval) | |
2069 | goto bad_fork_cleanup_security; | |
fb0a685c DRO |
2070 | retval = copy_files(clone_flags, p); |
2071 | if (retval) | |
1da177e4 | 2072 | goto bad_fork_cleanup_semundo; |
fb0a685c DRO |
2073 | retval = copy_fs(clone_flags, p); |
2074 | if (retval) | |
1da177e4 | 2075 | goto bad_fork_cleanup_files; |
fb0a685c DRO |
2076 | retval = copy_sighand(clone_flags, p); |
2077 | if (retval) | |
1da177e4 | 2078 | goto bad_fork_cleanup_fs; |
fb0a685c DRO |
2079 | retval = copy_signal(clone_flags, p); |
2080 | if (retval) | |
1da177e4 | 2081 | goto bad_fork_cleanup_sighand; |
fb0a685c DRO |
2082 | retval = copy_mm(clone_flags, p); |
2083 | if (retval) | |
1da177e4 | 2084 | goto bad_fork_cleanup_signal; |
fb0a685c DRO |
2085 | retval = copy_namespaces(clone_flags, p); |
2086 | if (retval) | |
d84f4f99 | 2087 | goto bad_fork_cleanup_mm; |
fb0a685c DRO |
2088 | retval = copy_io(clone_flags, p); |
2089 | if (retval) | |
fd0928df | 2090 | goto bad_fork_cleanup_namespaces; |
7f192e3c CB |
2091 | retval = copy_thread_tls(clone_flags, args->stack, args->stack_size, p, |
2092 | args->tls); | |
1da177e4 | 2093 | if (retval) |
fd0928df | 2094 | goto bad_fork_cleanup_io; |
1da177e4 | 2095 | |
afaef01c AP |
2096 | stackleak_task_init(p); |
2097 | ||
425fb2b4 | 2098 | if (pid != &init_struct_pid) { |
49cb2fc4 AR |
2099 | pid = alloc_pid(p->nsproxy->pid_ns_for_children, args->set_tid, |
2100 | args->set_tid_size); | |
35f71bc0 MH |
2101 | if (IS_ERR(pid)) { |
2102 | retval = PTR_ERR(pid); | |
0740aa5f | 2103 | goto bad_fork_cleanup_thread; |
35f71bc0 | 2104 | } |
425fb2b4 PE |
2105 | } |
2106 | ||
b3e58382 CB |
2107 | /* |
2108 | * This has to happen after we've potentially unshared the file | |
2109 | * descriptor table (so that the pidfd doesn't leak into the child | |
2110 | * if the fd table isn't shared). | |
2111 | */ | |
2112 | if (clone_flags & CLONE_PIDFD) { | |
6fd2fe49 | 2113 | retval = get_unused_fd_flags(O_RDWR | O_CLOEXEC); |
b3e58382 CB |
2114 | if (retval < 0) |
2115 | goto bad_fork_free_pid; | |
2116 | ||
2117 | pidfd = retval; | |
6fd2fe49 AV |
2118 | |
2119 | pidfile = anon_inode_getfile("[pidfd]", &pidfd_fops, pid, | |
2120 | O_RDWR | O_CLOEXEC); | |
2121 | if (IS_ERR(pidfile)) { | |
2122 | put_unused_fd(pidfd); | |
28dd29c0 | 2123 | retval = PTR_ERR(pidfile); |
6fd2fe49 AV |
2124 | goto bad_fork_free_pid; |
2125 | } | |
2126 | get_pid(pid); /* held by pidfile now */ | |
2127 | ||
7f192e3c | 2128 | retval = put_user(pidfd, args->pidfd); |
b3e58382 CB |
2129 | if (retval) |
2130 | goto bad_fork_put_pidfd; | |
2131 | } | |
2132 | ||
73c10101 JA |
2133 | #ifdef CONFIG_BLOCK |
2134 | p->plug = NULL; | |
2135 | #endif | |
ba31c1a4 TG |
2136 | futex_init_task(p); |
2137 | ||
f9a3879a GM |
2138 | /* |
2139 | * sigaltstack should be cleared when sharing the same VM | |
2140 | */ | |
2141 | if ((clone_flags & (CLONE_VM|CLONE_VFORK)) == CLONE_VM) | |
2a742138 | 2142 | sas_ss_reset(p); |
f9a3879a | 2143 | |
1da177e4 | 2144 | /* |
6580807d ON |
2145 | * Syscall tracing and stepping should be turned off in the |
2146 | * child regardless of CLONE_PTRACE. | |
1da177e4 | 2147 | */ |
6580807d | 2148 | user_disable_single_step(p); |
1da177e4 | 2149 | clear_tsk_thread_flag(p, TIF_SYSCALL_TRACE); |
ed75e8d5 LV |
2150 | #ifdef TIF_SYSCALL_EMU |
2151 | clear_tsk_thread_flag(p, TIF_SYSCALL_EMU); | |
2152 | #endif | |
e02c9b0d | 2153 | clear_tsk_latency_tracing(p); |
1da177e4 | 2154 | |
1da177e4 | 2155 | /* ok, now we should be set up.. */ |
18c830df ON |
2156 | p->pid = pid_nr(pid); |
2157 | if (clone_flags & CLONE_THREAD) { | |
5f8aadd8 | 2158 | p->exit_signal = -1; |
18c830df ON |
2159 | p->group_leader = current->group_leader; |
2160 | p->tgid = current->tgid; | |
2161 | } else { | |
2162 | if (clone_flags & CLONE_PARENT) | |
2163 | p->exit_signal = current->group_leader->exit_signal; | |
2164 | else | |
7f192e3c | 2165 | p->exit_signal = args->exit_signal; |
18c830df ON |
2166 | p->group_leader = p; |
2167 | p->tgid = p->pid; | |
2168 | } | |
5f8aadd8 | 2169 | |
9d823e8f WF |
2170 | p->nr_dirtied = 0; |
2171 | p->nr_dirtied_pause = 128 >> (PAGE_SHIFT - 10); | |
83712358 | 2172 | p->dirty_paused_when = 0; |
9d823e8f | 2173 | |
bb8cbbfe | 2174 | p->pdeath_signal = 0; |
47e65328 | 2175 | INIT_LIST_HEAD(&p->thread_group); |
158e1645 | 2176 | p->task_works = NULL; |
1da177e4 | 2177 | |
780de9dd | 2178 | cgroup_threadgroup_change_begin(current); |
7e47682e AS |
2179 | /* |
2180 | * Ensure that the cgroup subsystem policies allow the new process to be | |
2181 | * forked. It should be noted the the new process's css_set can be changed | |
2182 | * between here and cgroup_post_fork() if an organisation operation is in | |
2183 | * progress. | |
2184 | */ | |
b53202e6 | 2185 | retval = cgroup_can_fork(p); |
7e47682e | 2186 | if (retval) |
c3b7112d | 2187 | goto bad_fork_cgroup_threadgroup_change_end; |
7e47682e | 2188 | |
7b558513 DH |
2189 | /* |
2190 | * From this point on we must avoid any synchronous user-space | |
2191 | * communication until we take the tasklist-lock. In particular, we do | |
2192 | * not want user-space to be able to predict the process start-time by | |
2193 | * stalling fork(2) after we recorded the start_time but before it is | |
2194 | * visible to the system. | |
2195 | */ | |
2196 | ||
2197 | p->start_time = ktime_get_ns(); | |
cf25e24d | 2198 | p->start_boottime = ktime_get_boottime_ns(); |
7b558513 | 2199 | |
18c830df ON |
2200 | /* |
2201 | * Make it visible to the rest of the system, but dont wake it up yet. | |
2202 | * Need tasklist lock for parent etc handling! | |
2203 | */ | |
1da177e4 LT |
2204 | write_lock_irq(&tasklist_lock); |
2205 | ||
1da177e4 | 2206 | /* CLONE_PARENT re-uses the old parent */ |
2d5516cb | 2207 | if (clone_flags & (CLONE_PARENT|CLONE_THREAD)) { |
1da177e4 | 2208 | p->real_parent = current->real_parent; |
2d5516cb ON |
2209 | p->parent_exec_id = current->parent_exec_id; |
2210 | } else { | |
1da177e4 | 2211 | p->real_parent = current; |
2d5516cb ON |
2212 | p->parent_exec_id = current->self_exec_id; |
2213 | } | |
1da177e4 | 2214 | |
d83a7cb3 JP |
2215 | klp_copy_process(p); |
2216 | ||
3f17da69 | 2217 | spin_lock(¤t->sighand->siglock); |
4a2c7a78 | 2218 | |
dbd95212 KC |
2219 | /* |
2220 | * Copy seccomp details explicitly here, in case they were changed | |
2221 | * before holding sighand lock. | |
2222 | */ | |
2223 | copy_seccomp(p); | |
2224 | ||
d7822b1e MD |
2225 | rseq_fork(p, clone_flags); |
2226 | ||
4ca1d3ee | 2227 | /* Don't start children in a dying pid namespace */ |
e8cfbc24 | 2228 | if (unlikely(!(ns_of_pid(pid)->pid_allocated & PIDNS_ADDING))) { |
3fd37226 KT |
2229 | retval = -ENOMEM; |
2230 | goto bad_fork_cancel_cgroup; | |
2231 | } | |
4a2c7a78 | 2232 | |
7673bf55 EB |
2233 | /* Let kill terminate clone/fork in the middle */ |
2234 | if (fatal_signal_pending(current)) { | |
2235 | retval = -EINTR; | |
2236 | goto bad_fork_cancel_cgroup; | |
2237 | } | |
2238 | ||
6fd2fe49 AV |
2239 | /* past the last point of failure */ |
2240 | if (pidfile) | |
2241 | fd_install(pidfd, pidfile); | |
4a2c7a78 | 2242 | |
2c470475 | 2243 | init_task_pid_links(p); |
73b9ebfe | 2244 | if (likely(p->pid)) { |
4b9d33e6 | 2245 | ptrace_init_task(p, (clone_flags & CLONE_PTRACE) || trace); |
73b9ebfe | 2246 | |
81907739 | 2247 | init_task_pid(p, PIDTYPE_PID, pid); |
73b9ebfe | 2248 | if (thread_group_leader(p)) { |
6883f81a | 2249 | init_task_pid(p, PIDTYPE_TGID, pid); |
81907739 ON |
2250 | init_task_pid(p, PIDTYPE_PGID, task_pgrp(current)); |
2251 | init_task_pid(p, PIDTYPE_SID, task_session(current)); | |
2252 | ||
1c4042c2 | 2253 | if (is_child_reaper(pid)) { |
17cf22c3 | 2254 | ns_of_pid(pid)->child_reaper = p; |
1c4042c2 EB |
2255 | p->signal->flags |= SIGNAL_UNKILLABLE; |
2256 | } | |
c3ad2c3b | 2257 | p->signal->shared_pending.signal = delayed.signal; |
9c9f4ded | 2258 | p->signal->tty = tty_kref_get(current->signal->tty); |
749860ce PT |
2259 | /* |
2260 | * Inherit has_child_subreaper flag under the same | |
2261 | * tasklist_lock with adding child to the process tree | |
2262 | * for propagate_has_child_subreaper optimization. | |
2263 | */ | |
2264 | p->signal->has_child_subreaper = p->real_parent->signal->has_child_subreaper || | |
2265 | p->real_parent->signal->is_child_subreaper; | |
9cd80bbb | 2266 | list_add_tail(&p->sibling, &p->real_parent->children); |
5e85d4ab | 2267 | list_add_tail_rcu(&p->tasks, &init_task.tasks); |
6883f81a | 2268 | attach_pid(p, PIDTYPE_TGID); |
81907739 ON |
2269 | attach_pid(p, PIDTYPE_PGID); |
2270 | attach_pid(p, PIDTYPE_SID); | |
909ea964 | 2271 | __this_cpu_inc(process_counts); |
80628ca0 ON |
2272 | } else { |
2273 | current->signal->nr_threads++; | |
2274 | atomic_inc(¤t->signal->live); | |
60d4de3f | 2275 | refcount_inc(¤t->signal->sigcnt); |
924de3b8 | 2276 | task_join_group_stop(p); |
80628ca0 ON |
2277 | list_add_tail_rcu(&p->thread_group, |
2278 | &p->group_leader->thread_group); | |
0c740d0a ON |
2279 | list_add_tail_rcu(&p->thread_node, |
2280 | &p->signal->thread_head); | |
73b9ebfe | 2281 | } |
81907739 | 2282 | attach_pid(p, PIDTYPE_PID); |
73b9ebfe | 2283 | nr_threads++; |
1da177e4 | 2284 | } |
1da177e4 | 2285 | total_forks++; |
c3ad2c3b | 2286 | hlist_del_init(&delayed.node); |
3f17da69 | 2287 | spin_unlock(¤t->sighand->siglock); |
4af4206b | 2288 | syscall_tracepoint_update(p); |
1da177e4 | 2289 | write_unlock_irq(&tasklist_lock); |
4af4206b | 2290 | |
c13cf856 | 2291 | proc_fork_connector(p); |
b53202e6 | 2292 | cgroup_post_fork(p); |
780de9dd | 2293 | cgroup_threadgroup_change_end(current); |
cdd6c482 | 2294 | perf_event_fork(p); |
43d2b113 KH |
2295 | |
2296 | trace_task_newtask(p, clone_flags); | |
3ab67966 | 2297 | uprobe_copy_process(p, clone_flags); |
43d2b113 | 2298 | |
1da177e4 LT |
2299 | return p; |
2300 | ||
7e47682e | 2301 | bad_fork_cancel_cgroup: |
3fd37226 KT |
2302 | spin_unlock(¤t->sighand->siglock); |
2303 | write_unlock_irq(&tasklist_lock); | |
b53202e6 | 2304 | cgroup_cancel_fork(p); |
c3b7112d CB |
2305 | bad_fork_cgroup_threadgroup_change_end: |
2306 | cgroup_threadgroup_change_end(current); | |
b3e58382 | 2307 | bad_fork_put_pidfd: |
6fd2fe49 AV |
2308 | if (clone_flags & CLONE_PIDFD) { |
2309 | fput(pidfile); | |
2310 | put_unused_fd(pidfd); | |
2311 | } | |
425fb2b4 PE |
2312 | bad_fork_free_pid: |
2313 | if (pid != &init_struct_pid) | |
2314 | free_pid(pid); | |
0740aa5f JS |
2315 | bad_fork_cleanup_thread: |
2316 | exit_thread(p); | |
fd0928df | 2317 | bad_fork_cleanup_io: |
b69f2292 LR |
2318 | if (p->io_context) |
2319 | exit_io_context(p); | |
ab516013 | 2320 | bad_fork_cleanup_namespaces: |
444f378b | 2321 | exit_task_namespaces(p); |
1da177e4 | 2322 | bad_fork_cleanup_mm: |
c3f3ce04 AA |
2323 | if (p->mm) { |
2324 | mm_clear_owner(p->mm, p); | |
1da177e4 | 2325 | mmput(p->mm); |
c3f3ce04 | 2326 | } |
1da177e4 | 2327 | bad_fork_cleanup_signal: |
4ab6c083 | 2328 | if (!(clone_flags & CLONE_THREAD)) |
1c5354de | 2329 | free_signal_struct(p->signal); |
1da177e4 | 2330 | bad_fork_cleanup_sighand: |
a7e5328a | 2331 | __cleanup_sighand(p->sighand); |
1da177e4 LT |
2332 | bad_fork_cleanup_fs: |
2333 | exit_fs(p); /* blocking */ | |
2334 | bad_fork_cleanup_files: | |
2335 | exit_files(p); /* blocking */ | |
2336 | bad_fork_cleanup_semundo: | |
2337 | exit_sem(p); | |
e4e55b47 TH |
2338 | bad_fork_cleanup_security: |
2339 | security_task_free(p); | |
1da177e4 LT |
2340 | bad_fork_cleanup_audit: |
2341 | audit_free(p); | |
6c72e350 | 2342 | bad_fork_cleanup_perf: |
cdd6c482 | 2343 | perf_event_free_task(p); |
6c72e350 | 2344 | bad_fork_cleanup_policy: |
b09be676 | 2345 | lockdep_free_task(p); |
1da177e4 | 2346 | #ifdef CONFIG_NUMA |
f0be3d32 | 2347 | mpol_put(p->mempolicy); |
e8604cb4 | 2348 | bad_fork_cleanup_threadgroup_lock: |
1da177e4 | 2349 | #endif |
35df17c5 | 2350 | delayacct_tsk_free(p); |
1da177e4 | 2351 | bad_fork_cleanup_count: |
d84f4f99 | 2352 | atomic_dec(&p->cred->user->processes); |
e0e81739 | 2353 | exit_creds(p); |
1da177e4 | 2354 | bad_fork_free: |
405c0759 | 2355 | p->state = TASK_DEAD; |
68f24b08 | 2356 | put_task_stack(p); |
c3f3ce04 | 2357 | delayed_free_task(p); |
fe7d37d1 | 2358 | fork_out: |
c3ad2c3b EB |
2359 | spin_lock_irq(¤t->sighand->siglock); |
2360 | hlist_del_init(&delayed.node); | |
2361 | spin_unlock_irq(¤t->sighand->siglock); | |
fe7d37d1 | 2362 | return ERR_PTR(retval); |
1da177e4 LT |
2363 | } |
2364 | ||
2c470475 | 2365 | static inline void init_idle_pids(struct task_struct *idle) |
f106eee1 ON |
2366 | { |
2367 | enum pid_type type; | |
2368 | ||
2369 | for (type = PIDTYPE_PID; type < PIDTYPE_MAX; ++type) { | |
2c470475 EB |
2370 | INIT_HLIST_NODE(&idle->pid_links[type]); /* not really needed */ |
2371 | init_task_pid(idle, type, &init_struct_pid); | |
f106eee1 ON |
2372 | } |
2373 | } | |
2374 | ||
0db0628d | 2375 | struct task_struct *fork_idle(int cpu) |
1da177e4 | 2376 | { |
36c8b586 | 2377 | struct task_struct *task; |
7f192e3c CB |
2378 | struct kernel_clone_args args = { |
2379 | .flags = CLONE_VM, | |
2380 | }; | |
2381 | ||
2382 | task = copy_process(&init_struct_pid, 0, cpu_to_node(cpu), &args); | |
f106eee1 | 2383 | if (!IS_ERR(task)) { |
2c470475 | 2384 | init_idle_pids(task); |
753ca4f3 | 2385 | init_idle(task, cpu); |
f106eee1 | 2386 | } |
73b9ebfe | 2387 | |
1da177e4 LT |
2388 | return task; |
2389 | } | |
2390 | ||
13585fa0 NA |
2391 | struct mm_struct *copy_init_mm(void) |
2392 | { | |
2393 | return dup_mm(NULL, &init_mm); | |
2394 | } | |
2395 | ||
1da177e4 LT |
2396 | /* |
2397 | * Ok, this is the main fork-routine. | |
2398 | * | |
2399 | * It copies the process, and if successful kick-starts | |
2400 | * it and waits for it to finish using the VM if required. | |
a0eb9abd ES |
2401 | * |
2402 | * args->exit_signal is expected to be checked for sanity by the caller. | |
1da177e4 | 2403 | */ |
7f192e3c | 2404 | long _do_fork(struct kernel_clone_args *args) |
1da177e4 | 2405 | { |
7f192e3c | 2406 | u64 clone_flags = args->flags; |
9f5325aa MPS |
2407 | struct completion vfork; |
2408 | struct pid *pid; | |
1da177e4 LT |
2409 | struct task_struct *p; |
2410 | int trace = 0; | |
92476d7f | 2411 | long nr; |
1da177e4 | 2412 | |
09a05394 | 2413 | /* |
4b9d33e6 TH |
2414 | * Determine whether and which event to report to ptracer. When |
2415 | * called from kernel_thread or CLONE_UNTRACED is explicitly | |
2416 | * requested, no event is reported; otherwise, report if the event | |
2417 | * for the type of forking is enabled. | |
09a05394 | 2418 | */ |
e80d6661 | 2419 | if (!(clone_flags & CLONE_UNTRACED)) { |
4b9d33e6 TH |
2420 | if (clone_flags & CLONE_VFORK) |
2421 | trace = PTRACE_EVENT_VFORK; | |
7f192e3c | 2422 | else if (args->exit_signal != SIGCHLD) |
4b9d33e6 TH |
2423 | trace = PTRACE_EVENT_CLONE; |
2424 | else | |
2425 | trace = PTRACE_EVENT_FORK; | |
2426 | ||
2427 | if (likely(!ptrace_event_enabled(current, trace))) | |
2428 | trace = 0; | |
2429 | } | |
1da177e4 | 2430 | |
7f192e3c | 2431 | p = copy_process(NULL, trace, NUMA_NO_NODE, args); |
38addce8 | 2432 | add_latent_entropy(); |
9f5325aa MPS |
2433 | |
2434 | if (IS_ERR(p)) | |
2435 | return PTR_ERR(p); | |
2436 | ||
1da177e4 LT |
2437 | /* |
2438 | * Do this prior waking up the new thread - the thread pointer | |
2439 | * might get invalid after that point, if the thread exits quickly. | |
2440 | */ | |
9f5325aa | 2441 | trace_sched_process_fork(current, p); |
0a16b607 | 2442 | |
9f5325aa MPS |
2443 | pid = get_task_pid(p, PIDTYPE_PID); |
2444 | nr = pid_vnr(pid); | |
30e49c26 | 2445 | |
9f5325aa | 2446 | if (clone_flags & CLONE_PARENT_SETTID) |
7f192e3c | 2447 | put_user(nr, args->parent_tid); |
a6f5e063 | 2448 | |
9f5325aa MPS |
2449 | if (clone_flags & CLONE_VFORK) { |
2450 | p->vfork_done = &vfork; | |
2451 | init_completion(&vfork); | |
2452 | get_task_struct(p); | |
2453 | } | |
1da177e4 | 2454 | |
9f5325aa | 2455 | wake_up_new_task(p); |
09a05394 | 2456 | |
9f5325aa MPS |
2457 | /* forking complete and child started to run, tell ptracer */ |
2458 | if (unlikely(trace)) | |
2459 | ptrace_event_pid(trace, pid); | |
4e52365f | 2460 | |
9f5325aa MPS |
2461 | if (clone_flags & CLONE_VFORK) { |
2462 | if (!wait_for_vfork_done(p, &vfork)) | |
2463 | ptrace_event_pid(PTRACE_EVENT_VFORK_DONE, pid); | |
1da177e4 | 2464 | } |
9f5325aa MPS |
2465 | |
2466 | put_pid(pid); | |
92476d7f | 2467 | return nr; |
1da177e4 LT |
2468 | } |
2469 | ||
028b6e8a DL |
2470 | bool legacy_clone_args_valid(const struct kernel_clone_args *kargs) |
2471 | { | |
2472 | /* clone(CLONE_PIDFD) uses parent_tidptr to return a pidfd */ | |
2473 | if ((kargs->flags & CLONE_PIDFD) && | |
2474 | (kargs->flags & CLONE_PARENT_SETTID)) | |
2475 | return false; | |
2476 | ||
2477 | return true; | |
2478 | } | |
2479 | ||
3033f14a JT |
2480 | #ifndef CONFIG_HAVE_COPY_THREAD_TLS |
2481 | /* For compatibility with architectures that call do_fork directly rather than | |
2482 | * using the syscall entry points below. */ | |
2483 | long do_fork(unsigned long clone_flags, | |
2484 | unsigned long stack_start, | |
2485 | unsigned long stack_size, | |
2486 | int __user *parent_tidptr, | |
2487 | int __user *child_tidptr) | |
2488 | { | |
7f192e3c CB |
2489 | struct kernel_clone_args args = { |
2490 | .flags = (clone_flags & ~CSIGNAL), | |
028b6e8a | 2491 | .pidfd = parent_tidptr, |
7f192e3c CB |
2492 | .child_tid = child_tidptr, |
2493 | .parent_tid = parent_tidptr, | |
2494 | .exit_signal = (clone_flags & CSIGNAL), | |
2495 | .stack = stack_start, | |
2496 | .stack_size = stack_size, | |
2497 | }; | |
2498 | ||
028b6e8a DL |
2499 | if (!legacy_clone_args_valid(&args)) |
2500 | return -EINVAL; | |
2501 | ||
7f192e3c | 2502 | return _do_fork(&args); |
3033f14a JT |
2503 | } |
2504 | #endif | |
2505 | ||
2aa3a7f8 AV |
2506 | /* |
2507 | * Create a kernel thread. | |
2508 | */ | |
2509 | pid_t kernel_thread(int (*fn)(void *), void *arg, unsigned long flags) | |
2510 | { | |
7f192e3c CB |
2511 | struct kernel_clone_args args = { |
2512 | .flags = ((flags | CLONE_VM | CLONE_UNTRACED) & ~CSIGNAL), | |
2513 | .exit_signal = (flags & CSIGNAL), | |
2514 | .stack = (unsigned long)fn, | |
2515 | .stack_size = (unsigned long)arg, | |
2516 | }; | |
2517 | ||
2518 | return _do_fork(&args); | |
2aa3a7f8 | 2519 | } |
2aa3a7f8 | 2520 | |
d2125043 AV |
2521 | #ifdef __ARCH_WANT_SYS_FORK |
2522 | SYSCALL_DEFINE0(fork) | |
2523 | { | |
2524 | #ifdef CONFIG_MMU | |
7f192e3c CB |
2525 | struct kernel_clone_args args = { |
2526 | .exit_signal = SIGCHLD, | |
2527 | }; | |
2528 | ||
2529 | return _do_fork(&args); | |
d2125043 AV |
2530 | #else |
2531 | /* can not support in nommu mode */ | |
5d59e182 | 2532 | return -EINVAL; |
d2125043 AV |
2533 | #endif |
2534 | } | |
2535 | #endif | |
2536 | ||
2537 | #ifdef __ARCH_WANT_SYS_VFORK | |
2538 | SYSCALL_DEFINE0(vfork) | |
2539 | { | |
7f192e3c CB |
2540 | struct kernel_clone_args args = { |
2541 | .flags = CLONE_VFORK | CLONE_VM, | |
2542 | .exit_signal = SIGCHLD, | |
2543 | }; | |
2544 | ||
2545 | return _do_fork(&args); | |
d2125043 AV |
2546 | } |
2547 | #endif | |
2548 | ||
2549 | #ifdef __ARCH_WANT_SYS_CLONE | |
2550 | #ifdef CONFIG_CLONE_BACKWARDS | |
2551 | SYSCALL_DEFINE5(clone, unsigned long, clone_flags, unsigned long, newsp, | |
2552 | int __user *, parent_tidptr, | |
3033f14a | 2553 | unsigned long, tls, |
d2125043 AV |
2554 | int __user *, child_tidptr) |
2555 | #elif defined(CONFIG_CLONE_BACKWARDS2) | |
2556 | SYSCALL_DEFINE5(clone, unsigned long, newsp, unsigned long, clone_flags, | |
2557 | int __user *, parent_tidptr, | |
2558 | int __user *, child_tidptr, | |
3033f14a | 2559 | unsigned long, tls) |
dfa9771a MS |
2560 | #elif defined(CONFIG_CLONE_BACKWARDS3) |
2561 | SYSCALL_DEFINE6(clone, unsigned long, clone_flags, unsigned long, newsp, | |
2562 | int, stack_size, | |
2563 | int __user *, parent_tidptr, | |
2564 | int __user *, child_tidptr, | |
3033f14a | 2565 | unsigned long, tls) |
d2125043 AV |
2566 | #else |
2567 | SYSCALL_DEFINE5(clone, unsigned long, clone_flags, unsigned long, newsp, | |
2568 | int __user *, parent_tidptr, | |
2569 | int __user *, child_tidptr, | |
3033f14a | 2570 | unsigned long, tls) |
d2125043 AV |
2571 | #endif |
2572 | { | |
7f192e3c CB |
2573 | struct kernel_clone_args args = { |
2574 | .flags = (clone_flags & ~CSIGNAL), | |
2575 | .pidfd = parent_tidptr, | |
2576 | .child_tid = child_tidptr, | |
2577 | .parent_tid = parent_tidptr, | |
2578 | .exit_signal = (clone_flags & CSIGNAL), | |
2579 | .stack = newsp, | |
2580 | .tls = tls, | |
2581 | }; | |
2582 | ||
028b6e8a | 2583 | if (!legacy_clone_args_valid(&args)) |
7f192e3c CB |
2584 | return -EINVAL; |
2585 | ||
2586 | return _do_fork(&args); | |
2587 | } | |
d68dbb0c | 2588 | #endif |
7f192e3c | 2589 | |
d68dbb0c | 2590 | #ifdef __ARCH_WANT_SYS_CLONE3 |
dd499f7a AA |
2591 | |
2592 | /* | |
2593 | * copy_thread implementations handle CLONE_SETTLS by reading the TLS value from | |
2594 | * the registers containing the syscall arguments for clone. This doesn't work | |
2595 | * with clone3 since the TLS value is passed in clone_args instead. | |
2596 | */ | |
2597 | #ifndef CONFIG_HAVE_COPY_THREAD_TLS | |
2598 | #error clone3 requires copy_thread_tls support in arch | |
2599 | #endif | |
2600 | ||
7f192e3c CB |
2601 | noinline static int copy_clone_args_from_user(struct kernel_clone_args *kargs, |
2602 | struct clone_args __user *uargs, | |
f14c234b | 2603 | size_t usize) |
7f192e3c | 2604 | { |
f14c234b | 2605 | int err; |
7f192e3c | 2606 | struct clone_args args; |
49cb2fc4 | 2607 | pid_t *kset_tid = kargs->set_tid; |
7f192e3c | 2608 | |
f14c234b | 2609 | if (unlikely(usize > PAGE_SIZE)) |
7f192e3c | 2610 | return -E2BIG; |
f14c234b | 2611 | if (unlikely(usize < CLONE_ARGS_SIZE_VER0)) |
7f192e3c CB |
2612 | return -EINVAL; |
2613 | ||
f14c234b AS |
2614 | err = copy_struct_from_user(&args, sizeof(args), uargs, usize); |
2615 | if (err) | |
2616 | return err; | |
7f192e3c | 2617 | |
49cb2fc4 AR |
2618 | if (unlikely(args.set_tid_size > MAX_PID_NS_LEVEL)) |
2619 | return -EINVAL; | |
2620 | ||
2621 | if (unlikely(!args.set_tid && args.set_tid_size > 0)) | |
2622 | return -EINVAL; | |
2623 | ||
2624 | if (unlikely(args.set_tid && args.set_tid_size == 0)) | |
2625 | return -EINVAL; | |
2626 | ||
a0eb9abd ES |
2627 | /* |
2628 | * Verify that higher 32bits of exit_signal are unset and that | |
2629 | * it is a valid signal | |
2630 | */ | |
2631 | if (unlikely((args.exit_signal & ~((u64)CSIGNAL)) || | |
2632 | !valid_signal(args.exit_signal))) | |
2633 | return -EINVAL; | |
2634 | ||
7f192e3c CB |
2635 | *kargs = (struct kernel_clone_args){ |
2636 | .flags = args.flags, | |
2637 | .pidfd = u64_to_user_ptr(args.pidfd), | |
2638 | .child_tid = u64_to_user_ptr(args.child_tid), | |
2639 | .parent_tid = u64_to_user_ptr(args.parent_tid), | |
2640 | .exit_signal = args.exit_signal, | |
2641 | .stack = args.stack, | |
2642 | .stack_size = args.stack_size, | |
2643 | .tls = args.tls, | |
49cb2fc4 | 2644 | .set_tid_size = args.set_tid_size, |
7f192e3c CB |
2645 | }; |
2646 | ||
49cb2fc4 AR |
2647 | if (args.set_tid && |
2648 | copy_from_user(kset_tid, u64_to_user_ptr(args.set_tid), | |
2649 | (kargs->set_tid_size * sizeof(pid_t)))) | |
2650 | return -EFAULT; | |
2651 | ||
2652 | kargs->set_tid = kset_tid; | |
2653 | ||
7f192e3c CB |
2654 | return 0; |
2655 | } | |
2656 | ||
fa729c4d CB |
2657 | /** |
2658 | * clone3_stack_valid - check and prepare stack | |
2659 | * @kargs: kernel clone args | |
2660 | * | |
2661 | * Verify that the stack arguments userspace gave us are sane. | |
2662 | * In addition, set the stack direction for userspace since it's easy for us to | |
2663 | * determine. | |
2664 | */ | |
2665 | static inline bool clone3_stack_valid(struct kernel_clone_args *kargs) | |
2666 | { | |
2667 | if (kargs->stack == 0) { | |
2668 | if (kargs->stack_size > 0) | |
2669 | return false; | |
2670 | } else { | |
2671 | if (kargs->stack_size == 0) | |
2672 | return false; | |
2673 | ||
2674 | if (!access_ok((void __user *)kargs->stack, kargs->stack_size)) | |
2675 | return false; | |
2676 | ||
2677 | #if !defined(CONFIG_STACK_GROWSUP) && !defined(CONFIG_IA64) | |
2678 | kargs->stack += kargs->stack_size; | |
2679 | #endif | |
2680 | } | |
2681 | ||
2682 | return true; | |
2683 | } | |
2684 | ||
2685 | static bool clone3_args_valid(struct kernel_clone_args *kargs) | |
7f192e3c | 2686 | { |
b612e5df CB |
2687 | /* Verify that no unknown flags are passed along. */ |
2688 | if (kargs->flags & ~(CLONE_LEGACY_FLAGS | CLONE_CLEAR_SIGHAND)) | |
7f192e3c CB |
2689 | return false; |
2690 | ||
2691 | /* | |
2692 | * - make the CLONE_DETACHED bit reuseable for clone3 | |
2693 | * - make the CSIGNAL bits reuseable for clone3 | |
2694 | */ | |
2695 | if (kargs->flags & (CLONE_DETACHED | CSIGNAL)) | |
2696 | return false; | |
2697 | ||
b612e5df CB |
2698 | if ((kargs->flags & (CLONE_SIGHAND | CLONE_CLEAR_SIGHAND)) == |
2699 | (CLONE_SIGHAND | CLONE_CLEAR_SIGHAND)) | |
2700 | return false; | |
2701 | ||
7f192e3c CB |
2702 | if ((kargs->flags & (CLONE_THREAD | CLONE_PARENT)) && |
2703 | kargs->exit_signal) | |
2704 | return false; | |
2705 | ||
fa729c4d CB |
2706 | if (!clone3_stack_valid(kargs)) |
2707 | return false; | |
2708 | ||
7f192e3c CB |
2709 | return true; |
2710 | } | |
2711 | ||
501bd016 CB |
2712 | /** |
2713 | * clone3 - create a new process with specific properties | |
2714 | * @uargs: argument structure | |
2715 | * @size: size of @uargs | |
2716 | * | |
2717 | * clone3() is the extensible successor to clone()/clone2(). | |
2718 | * It takes a struct as argument that is versioned by its size. | |
2719 | * | |
2720 | * Return: On success, a positive PID for the child process. | |
2721 | * On error, a negative errno number. | |
2722 | */ | |
7f192e3c CB |
2723 | SYSCALL_DEFINE2(clone3, struct clone_args __user *, uargs, size_t, size) |
2724 | { | |
2725 | int err; | |
2726 | ||
2727 | struct kernel_clone_args kargs; | |
49cb2fc4 AR |
2728 | pid_t set_tid[MAX_PID_NS_LEVEL]; |
2729 | ||
2730 | kargs.set_tid = set_tid; | |
7f192e3c CB |
2731 | |
2732 | err = copy_clone_args_from_user(&kargs, uargs, size); | |
2733 | if (err) | |
2734 | return err; | |
2735 | ||
2736 | if (!clone3_args_valid(&kargs)) | |
2737 | return -EINVAL; | |
2738 | ||
2739 | return _do_fork(&kargs); | |
d2125043 AV |
2740 | } |
2741 | #endif | |
2742 | ||
0f1b92cb ON |
2743 | void walk_process_tree(struct task_struct *top, proc_visitor visitor, void *data) |
2744 | { | |
2745 | struct task_struct *leader, *parent, *child; | |
2746 | int res; | |
2747 | ||
2748 | read_lock(&tasklist_lock); | |
2749 | leader = top = top->group_leader; | |
2750 | down: | |
2751 | for_each_thread(leader, parent) { | |
2752 | list_for_each_entry(child, &parent->children, sibling) { | |
2753 | res = visitor(child, data); | |
2754 | if (res) { | |
2755 | if (res < 0) | |
2756 | goto out; | |
2757 | leader = child; | |
2758 | goto down; | |
2759 | } | |
2760 | up: | |
2761 | ; | |
2762 | } | |
2763 | } | |
2764 | ||
2765 | if (leader != top) { | |
2766 | child = leader; | |
2767 | parent = child->real_parent; | |
2768 | leader = parent->group_leader; | |
2769 | goto up; | |
2770 | } | |
2771 | out: | |
2772 | read_unlock(&tasklist_lock); | |
2773 | } | |
2774 | ||
5fd63b30 RT |
2775 | #ifndef ARCH_MIN_MMSTRUCT_ALIGN |
2776 | #define ARCH_MIN_MMSTRUCT_ALIGN 0 | |
2777 | #endif | |
2778 | ||
51cc5068 | 2779 | static void sighand_ctor(void *data) |
aa1757f9 ON |
2780 | { |
2781 | struct sighand_struct *sighand = data; | |
2782 | ||
a35afb83 | 2783 | spin_lock_init(&sighand->siglock); |
b8fceee1 | 2784 | init_waitqueue_head(&sighand->signalfd_wqh); |
aa1757f9 ON |
2785 | } |
2786 | ||
1da177e4 LT |
2787 | void __init proc_caches_init(void) |
2788 | { | |
c1a2f7f0 RR |
2789 | unsigned int mm_size; |
2790 | ||
1da177e4 LT |
2791 | sighand_cachep = kmem_cache_create("sighand_cache", |
2792 | sizeof(struct sighand_struct), 0, | |
5f0d5a3a | 2793 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_TYPESAFE_BY_RCU| |
75f296d9 | 2794 | SLAB_ACCOUNT, sighand_ctor); |
1da177e4 LT |
2795 | signal_cachep = kmem_cache_create("signal_cache", |
2796 | sizeof(struct signal_struct), 0, | |
75f296d9 | 2797 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_ACCOUNT, |
5d097056 | 2798 | NULL); |
20c2df83 | 2799 | files_cachep = kmem_cache_create("files_cache", |
1da177e4 | 2800 | sizeof(struct files_struct), 0, |
75f296d9 | 2801 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_ACCOUNT, |
5d097056 | 2802 | NULL); |
20c2df83 | 2803 | fs_cachep = kmem_cache_create("fs_cache", |
1da177e4 | 2804 | sizeof(struct fs_struct), 0, |
75f296d9 | 2805 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_ACCOUNT, |
5d097056 | 2806 | NULL); |
c1a2f7f0 | 2807 | |
6345d24d | 2808 | /* |
c1a2f7f0 RR |
2809 | * The mm_cpumask is located at the end of mm_struct, and is |
2810 | * dynamically sized based on the maximum CPU number this system | |
2811 | * can have, taking hotplug into account (nr_cpu_ids). | |
6345d24d | 2812 | */ |
c1a2f7f0 RR |
2813 | mm_size = sizeof(struct mm_struct) + cpumask_size(); |
2814 | ||
07dcd7fe | 2815 | mm_cachep = kmem_cache_create_usercopy("mm_struct", |
c1a2f7f0 | 2816 | mm_size, ARCH_MIN_MMSTRUCT_ALIGN, |
75f296d9 | 2817 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_ACCOUNT, |
07dcd7fe DW |
2818 | offsetof(struct mm_struct, saved_auxv), |
2819 | sizeof_field(struct mm_struct, saved_auxv), | |
5d097056 VD |
2820 | NULL); |
2821 | vm_area_cachep = KMEM_CACHE(vm_area_struct, SLAB_PANIC|SLAB_ACCOUNT); | |
8feae131 | 2822 | mmap_init(); |
66577193 | 2823 | nsproxy_cache_init(); |
1da177e4 | 2824 | } |
cf2e340f | 2825 | |
cf2e340f | 2826 | /* |
9bfb23fc | 2827 | * Check constraints on flags passed to the unshare system call. |
cf2e340f | 2828 | */ |
9bfb23fc | 2829 | static int check_unshare_flags(unsigned long unshare_flags) |
cf2e340f | 2830 | { |
9bfb23fc ON |
2831 | if (unshare_flags & ~(CLONE_THREAD|CLONE_FS|CLONE_NEWNS|CLONE_SIGHAND| |
2832 | CLONE_VM|CLONE_FILES|CLONE_SYSVSEM| | |
50804fe3 | 2833 | CLONE_NEWUTS|CLONE_NEWIPC|CLONE_NEWNET| |
769071ac AV |
2834 | CLONE_NEWUSER|CLONE_NEWPID|CLONE_NEWCGROUP| |
2835 | CLONE_NEWTIME)) | |
9bfb23fc | 2836 | return -EINVAL; |
cf2e340f | 2837 | /* |
12c641ab EB |
2838 | * Not implemented, but pretend it works if there is nothing |
2839 | * to unshare. Note that unsharing the address space or the | |
2840 | * signal handlers also need to unshare the signal queues (aka | |
2841 | * CLONE_THREAD). | |
cf2e340f | 2842 | */ |
9bfb23fc | 2843 | if (unshare_flags & (CLONE_THREAD | CLONE_SIGHAND | CLONE_VM)) { |
12c641ab EB |
2844 | if (!thread_group_empty(current)) |
2845 | return -EINVAL; | |
2846 | } | |
2847 | if (unshare_flags & (CLONE_SIGHAND | CLONE_VM)) { | |
d036bda7 | 2848 | if (refcount_read(¤t->sighand->count) > 1) |
12c641ab EB |
2849 | return -EINVAL; |
2850 | } | |
2851 | if (unshare_flags & CLONE_VM) { | |
2852 | if (!current_is_single_threaded()) | |
9bfb23fc ON |
2853 | return -EINVAL; |
2854 | } | |
cf2e340f JD |
2855 | |
2856 | return 0; | |
2857 | } | |
2858 | ||
2859 | /* | |
99d1419d | 2860 | * Unshare the filesystem structure if it is being shared |
cf2e340f JD |
2861 | */ |
2862 | static int unshare_fs(unsigned long unshare_flags, struct fs_struct **new_fsp) | |
2863 | { | |
2864 | struct fs_struct *fs = current->fs; | |
2865 | ||
498052bb AV |
2866 | if (!(unshare_flags & CLONE_FS) || !fs) |
2867 | return 0; | |
2868 | ||
2869 | /* don't need lock here; in the worst case we'll do useless copy */ | |
2870 | if (fs->users == 1) | |
2871 | return 0; | |
2872 | ||
2873 | *new_fsp = copy_fs_struct(fs); | |
2874 | if (!*new_fsp) | |
2875 | return -ENOMEM; | |
cf2e340f JD |
2876 | |
2877 | return 0; | |
2878 | } | |
2879 | ||
cf2e340f | 2880 | /* |
a016f338 | 2881 | * Unshare file descriptor table if it is being shared |
cf2e340f JD |
2882 | */ |
2883 | static int unshare_fd(unsigned long unshare_flags, struct files_struct **new_fdp) | |
2884 | { | |
2885 | struct files_struct *fd = current->files; | |
a016f338 | 2886 | int error = 0; |
cf2e340f JD |
2887 | |
2888 | if ((unshare_flags & CLONE_FILES) && | |
a016f338 JD |
2889 | (fd && atomic_read(&fd->count) > 1)) { |
2890 | *new_fdp = dup_fd(fd, &error); | |
2891 | if (!*new_fdp) | |
2892 | return error; | |
2893 | } | |
cf2e340f JD |
2894 | |
2895 | return 0; | |
2896 | } | |
2897 | ||
cf2e340f JD |
2898 | /* |
2899 | * unshare allows a process to 'unshare' part of the process | |
2900 | * context which was originally shared using clone. copy_* | |
2901 | * functions used by do_fork() cannot be used here directly | |
2902 | * because they modify an inactive task_struct that is being | |
2903 | * constructed. Here we are modifying the current, active, | |
2904 | * task_struct. | |
2905 | */ | |
9b32105e | 2906 | int ksys_unshare(unsigned long unshare_flags) |
cf2e340f | 2907 | { |
cf2e340f | 2908 | struct fs_struct *fs, *new_fs = NULL; |
cf2e340f | 2909 | struct files_struct *fd, *new_fd = NULL; |
b2e0d987 | 2910 | struct cred *new_cred = NULL; |
cf7b708c | 2911 | struct nsproxy *new_nsproxy = NULL; |
9edff4ab | 2912 | int do_sysvsem = 0; |
9bfb23fc | 2913 | int err; |
cf2e340f | 2914 | |
b2e0d987 | 2915 | /* |
faf00da5 EB |
2916 | * If unsharing a user namespace must also unshare the thread group |
2917 | * and unshare the filesystem root and working directories. | |
b2e0d987 EB |
2918 | */ |
2919 | if (unshare_flags & CLONE_NEWUSER) | |
e66eded8 | 2920 | unshare_flags |= CLONE_THREAD | CLONE_FS; |
50804fe3 EB |
2921 | /* |
2922 | * If unsharing vm, must also unshare signal handlers. | |
2923 | */ | |
2924 | if (unshare_flags & CLONE_VM) | |
2925 | unshare_flags |= CLONE_SIGHAND; | |
12c641ab EB |
2926 | /* |
2927 | * If unsharing a signal handlers, must also unshare the signal queues. | |
2928 | */ | |
2929 | if (unshare_flags & CLONE_SIGHAND) | |
2930 | unshare_flags |= CLONE_THREAD; | |
9bfb23fc ON |
2931 | /* |
2932 | * If unsharing namespace, must also unshare filesystem information. | |
2933 | */ | |
2934 | if (unshare_flags & CLONE_NEWNS) | |
2935 | unshare_flags |= CLONE_FS; | |
50804fe3 EB |
2936 | |
2937 | err = check_unshare_flags(unshare_flags); | |
2938 | if (err) | |
2939 | goto bad_unshare_out; | |
6013f67f MS |
2940 | /* |
2941 | * CLONE_NEWIPC must also detach from the undolist: after switching | |
2942 | * to a new ipc namespace, the semaphore arrays from the old | |
2943 | * namespace are unreachable. | |
2944 | */ | |
2945 | if (unshare_flags & (CLONE_NEWIPC|CLONE_SYSVSEM)) | |
9edff4ab | 2946 | do_sysvsem = 1; |
fb0a685c DRO |
2947 | err = unshare_fs(unshare_flags, &new_fs); |
2948 | if (err) | |
9bfb23fc | 2949 | goto bad_unshare_out; |
fb0a685c DRO |
2950 | err = unshare_fd(unshare_flags, &new_fd); |
2951 | if (err) | |
9bfb23fc | 2952 | goto bad_unshare_cleanup_fs; |
b2e0d987 | 2953 | err = unshare_userns(unshare_flags, &new_cred); |
fb0a685c | 2954 | if (err) |
9edff4ab | 2955 | goto bad_unshare_cleanup_fd; |
b2e0d987 EB |
2956 | err = unshare_nsproxy_namespaces(unshare_flags, &new_nsproxy, |
2957 | new_cred, new_fs); | |
2958 | if (err) | |
2959 | goto bad_unshare_cleanup_cred; | |
c0b2fc31 | 2960 | |
b2e0d987 | 2961 | if (new_fs || new_fd || do_sysvsem || new_cred || new_nsproxy) { |
9edff4ab MS |
2962 | if (do_sysvsem) { |
2963 | /* | |
2964 | * CLONE_SYSVSEM is equivalent to sys_exit(). | |
2965 | */ | |
2966 | exit_sem(current); | |
2967 | } | |
ab602f79 JM |
2968 | if (unshare_flags & CLONE_NEWIPC) { |
2969 | /* Orphan segments in old ns (see sem above). */ | |
2970 | exit_shm(current); | |
2971 | shm_init_task(current); | |
2972 | } | |
ab516013 | 2973 | |
6f977e6b | 2974 | if (new_nsproxy) |
cf7b708c | 2975 | switch_task_namespaces(current, new_nsproxy); |
cf2e340f | 2976 | |
cf7b708c PE |
2977 | task_lock(current); |
2978 | ||
cf2e340f JD |
2979 | if (new_fs) { |
2980 | fs = current->fs; | |
2a4419b5 | 2981 | spin_lock(&fs->lock); |
cf2e340f | 2982 | current->fs = new_fs; |
498052bb AV |
2983 | if (--fs->users) |
2984 | new_fs = NULL; | |
2985 | else | |
2986 | new_fs = fs; | |
2a4419b5 | 2987 | spin_unlock(&fs->lock); |
cf2e340f JD |
2988 | } |
2989 | ||
cf2e340f JD |
2990 | if (new_fd) { |
2991 | fd = current->files; | |
2992 | current->files = new_fd; | |
2993 | new_fd = fd; | |
2994 | } | |
2995 | ||
2996 | task_unlock(current); | |
b2e0d987 EB |
2997 | |
2998 | if (new_cred) { | |
2999 | /* Install the new user namespace */ | |
3000 | commit_creds(new_cred); | |
3001 | new_cred = NULL; | |
3002 | } | |
cf2e340f JD |
3003 | } |
3004 | ||
e4222673 HB |
3005 | perf_event_namespaces(current); |
3006 | ||
b2e0d987 EB |
3007 | bad_unshare_cleanup_cred: |
3008 | if (new_cred) | |
3009 | put_cred(new_cred); | |
cf2e340f JD |
3010 | bad_unshare_cleanup_fd: |
3011 | if (new_fd) | |
3012 | put_files_struct(new_fd); | |
3013 | ||
cf2e340f JD |
3014 | bad_unshare_cleanup_fs: |
3015 | if (new_fs) | |
498052bb | 3016 | free_fs_struct(new_fs); |
cf2e340f | 3017 | |
cf2e340f JD |
3018 | bad_unshare_out: |
3019 | return err; | |
3020 | } | |
3b125388 | 3021 | |
9b32105e DB |
3022 | SYSCALL_DEFINE1(unshare, unsigned long, unshare_flags) |
3023 | { | |
3024 | return ksys_unshare(unshare_flags); | |
3025 | } | |
3026 | ||
3b125388 AV |
3027 | /* |
3028 | * Helper to unshare the files of the current task. | |
3029 | * We don't want to expose copy_files internals to | |
3030 | * the exec layer of the kernel. | |
3031 | */ | |
3032 | ||
3033 | int unshare_files(struct files_struct **displaced) | |
3034 | { | |
3035 | struct task_struct *task = current; | |
50704516 | 3036 | struct files_struct *copy = NULL; |
3b125388 AV |
3037 | int error; |
3038 | ||
3039 | error = unshare_fd(CLONE_FILES, ©); | |
3040 | if (error || !copy) { | |
3041 | *displaced = NULL; | |
3042 | return error; | |
3043 | } | |
3044 | *displaced = task->files; | |
3045 | task_lock(task); | |
3046 | task->files = copy; | |
3047 | task_unlock(task); | |
3048 | return 0; | |
3049 | } | |
16db3d3f HS |
3050 | |
3051 | int sysctl_max_threads(struct ctl_table *table, int write, | |
3052 | void __user *buffer, size_t *lenp, loff_t *ppos) | |
3053 | { | |
3054 | struct ctl_table t; | |
3055 | int ret; | |
3056 | int threads = max_threads; | |
b0f53dbc | 3057 | int min = 1; |
16db3d3f HS |
3058 | int max = MAX_THREADS; |
3059 | ||
3060 | t = *table; | |
3061 | t.data = &threads; | |
3062 | t.extra1 = &min; | |
3063 | t.extra2 = &max; | |
3064 | ||
3065 | ret = proc_dointvec_minmax(&t, write, buffer, lenp, ppos); | |
3066 | if (ret || !write) | |
3067 | return ret; | |
3068 | ||
b0f53dbc | 3069 | max_threads = threads; |
16db3d3f HS |
3070 | |
3071 | return 0; | |
3072 | } |