]>
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> |
d08b9f0c | 97 | #include <linux/scs.h> |
0f212204 | 98 | #include <linux/io_uring.h> |
a10787e6 | 99 | #include <linux/bpf.h> |
1da177e4 | 100 | |
1da177e4 | 101 | #include <asm/pgalloc.h> |
7c0f6ba6 | 102 | #include <linux/uaccess.h> |
1da177e4 LT |
103 | #include <asm/mmu_context.h> |
104 | #include <asm/cacheflush.h> | |
105 | #include <asm/tlbflush.h> | |
106 | ||
ad8d75ff SR |
107 | #include <trace/events/sched.h> |
108 | ||
43d2b113 KH |
109 | #define CREATE_TRACE_POINTS |
110 | #include <trace/events/task.h> | |
111 | ||
ac1b398d HS |
112 | /* |
113 | * Minimum number of threads to boot the kernel | |
114 | */ | |
115 | #define MIN_THREADS 20 | |
116 | ||
117 | /* | |
118 | * Maximum number of threads | |
119 | */ | |
120 | #define MAX_THREADS FUTEX_TID_MASK | |
121 | ||
1da177e4 LT |
122 | /* |
123 | * Protected counters by write_lock_irq(&tasklist_lock) | |
124 | */ | |
125 | unsigned long total_forks; /* Handle normal Linux uptimes. */ | |
fb0a685c | 126 | int nr_threads; /* The idle threads do not count.. */ |
1da177e4 | 127 | |
8856ae4d | 128 | static int max_threads; /* tunable limit on nr_threads */ |
1da177e4 | 129 | |
8495f7e6 SPP |
130 | #define NAMED_ARRAY_INDEX(x) [x] = __stringify(x) |
131 | ||
132 | static const char * const resident_page_types[] = { | |
133 | NAMED_ARRAY_INDEX(MM_FILEPAGES), | |
134 | NAMED_ARRAY_INDEX(MM_ANONPAGES), | |
135 | NAMED_ARRAY_INDEX(MM_SWAPENTS), | |
136 | NAMED_ARRAY_INDEX(MM_SHMEMPAGES), | |
137 | }; | |
138 | ||
1da177e4 LT |
139 | DEFINE_PER_CPU(unsigned long, process_counts) = 0; |
140 | ||
c59923a1 | 141 | __cacheline_aligned DEFINE_RWLOCK(tasklist_lock); /* outer */ |
db1466b3 PM |
142 | |
143 | #ifdef CONFIG_PROVE_RCU | |
144 | int lockdep_tasklist_lock_is_held(void) | |
145 | { | |
146 | return lockdep_is_held(&tasklist_lock); | |
147 | } | |
148 | EXPORT_SYMBOL_GPL(lockdep_tasklist_lock_is_held); | |
149 | #endif /* #ifdef CONFIG_PROVE_RCU */ | |
1da177e4 LT |
150 | |
151 | int nr_processes(void) | |
152 | { | |
153 | int cpu; | |
154 | int total = 0; | |
155 | ||
1d510750 | 156 | for_each_possible_cpu(cpu) |
1da177e4 LT |
157 | total += per_cpu(process_counts, cpu); |
158 | ||
159 | return total; | |
160 | } | |
161 | ||
f19b9f74 AM |
162 | void __weak arch_release_task_struct(struct task_struct *tsk) |
163 | { | |
164 | } | |
165 | ||
f5e10287 | 166 | #ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR |
e18b890b | 167 | static struct kmem_cache *task_struct_cachep; |
41101809 TG |
168 | |
169 | static inline struct task_struct *alloc_task_struct_node(int node) | |
170 | { | |
171 | return kmem_cache_alloc_node(task_struct_cachep, GFP_KERNEL, node); | |
172 | } | |
173 | ||
41101809 TG |
174 | static inline void free_task_struct(struct task_struct *tsk) |
175 | { | |
41101809 TG |
176 | kmem_cache_free(task_struct_cachep, tsk); |
177 | } | |
1da177e4 LT |
178 | #endif |
179 | ||
b235beea | 180 | #ifndef CONFIG_ARCH_THREAD_STACK_ALLOCATOR |
41101809 | 181 | |
0d15d74a TG |
182 | /* |
183 | * Allocate pages if THREAD_SIZE is >= PAGE_SIZE, otherwise use a | |
184 | * kmemcache based allocator. | |
185 | */ | |
ba14a194 | 186 | # if THREAD_SIZE >= PAGE_SIZE || defined(CONFIG_VMAP_STACK) |
ac496bf4 AL |
187 | |
188 | #ifdef CONFIG_VMAP_STACK | |
189 | /* | |
190 | * vmalloc() is a bit slow, and calling vfree() enough times will force a TLB | |
191 | * flush. Try to minimize the number of calls by caching stacks. | |
192 | */ | |
193 | #define NR_CACHED_STACKS 2 | |
194 | static DEFINE_PER_CPU(struct vm_struct *, cached_stacks[NR_CACHED_STACKS]); | |
19659c59 HR |
195 | |
196 | static int free_vm_stack_cache(unsigned int cpu) | |
197 | { | |
198 | struct vm_struct **cached_vm_stacks = per_cpu_ptr(cached_stacks, cpu); | |
199 | int i; | |
200 | ||
201 | for (i = 0; i < NR_CACHED_STACKS; i++) { | |
202 | struct vm_struct *vm_stack = cached_vm_stacks[i]; | |
203 | ||
204 | if (!vm_stack) | |
205 | continue; | |
206 | ||
207 | vfree(vm_stack->addr); | |
208 | cached_vm_stacks[i] = NULL; | |
209 | } | |
210 | ||
211 | return 0; | |
212 | } | |
ac496bf4 AL |
213 | #endif |
214 | ||
ba14a194 | 215 | static unsigned long *alloc_thread_stack_node(struct task_struct *tsk, int node) |
b69c49b7 | 216 | { |
ba14a194 | 217 | #ifdef CONFIG_VMAP_STACK |
ac496bf4 AL |
218 | void *stack; |
219 | int i; | |
220 | ||
ac496bf4 | 221 | for (i = 0; i < NR_CACHED_STACKS; i++) { |
112166f8 CL |
222 | struct vm_struct *s; |
223 | ||
224 | s = this_cpu_xchg(cached_stacks[i], NULL); | |
ac496bf4 AL |
225 | |
226 | if (!s) | |
227 | continue; | |
ac496bf4 | 228 | |
cebd0eb2 AK |
229 | /* Mark stack accessible for KASAN. */ |
230 | kasan_unpoison_range(s->addr, THREAD_SIZE); | |
eafb149e | 231 | |
ca182551 KK |
232 | /* Clear stale pointers from reused stack. */ |
233 | memset(s->addr, 0, THREAD_SIZE); | |
e01e8063 | 234 | |
ac496bf4 | 235 | tsk->stack_vm_area = s; |
ba4a4574 | 236 | tsk->stack = s->addr; |
ac496bf4 AL |
237 | return s->addr; |
238 | } | |
ac496bf4 | 239 | |
9b6f7e16 RG |
240 | /* |
241 | * Allocated stacks are cached and later reused by new threads, | |
242 | * so memcg accounting is performed manually on assigning/releasing | |
243 | * stacks to tasks. Drop __GFP_ACCOUNT. | |
244 | */ | |
48ac3c18 | 245 | stack = __vmalloc_node_range(THREAD_SIZE, THREAD_ALIGN, |
ac496bf4 | 246 | VMALLOC_START, VMALLOC_END, |
9b6f7e16 | 247 | THREADINFO_GFP & ~__GFP_ACCOUNT, |
ac496bf4 AL |
248 | PAGE_KERNEL, |
249 | 0, node, __builtin_return_address(0)); | |
ba14a194 AL |
250 | |
251 | /* | |
252 | * We can't call find_vm_area() in interrupt context, and | |
253 | * free_thread_stack() can be called in interrupt context, | |
254 | * so cache the vm_struct. | |
255 | */ | |
5eed6f1d | 256 | if (stack) { |
ba14a194 | 257 | tsk->stack_vm_area = find_vm_area(stack); |
5eed6f1d RR |
258 | tsk->stack = stack; |
259 | } | |
ba14a194 AL |
260 | return stack; |
261 | #else | |
4949148a VD |
262 | struct page *page = alloc_pages_node(node, THREADINFO_GFP, |
263 | THREAD_SIZE_ORDER); | |
b6a84016 | 264 | |
1bf4580e | 265 | if (likely(page)) { |
8dcc1d34 | 266 | tsk->stack = kasan_reset_tag(page_address(page)); |
1bf4580e AA |
267 | return tsk->stack; |
268 | } | |
269 | return NULL; | |
ba14a194 | 270 | #endif |
b69c49b7 FT |
271 | } |
272 | ||
ba14a194 | 273 | static inline void free_thread_stack(struct task_struct *tsk) |
b69c49b7 | 274 | { |
ac496bf4 | 275 | #ifdef CONFIG_VMAP_STACK |
9b6f7e16 RG |
276 | struct vm_struct *vm = task_stack_vm_area(tsk); |
277 | ||
278 | if (vm) { | |
ac496bf4 AL |
279 | int i; |
280 | ||
991e7673 | 281 | for (i = 0; i < THREAD_SIZE / PAGE_SIZE; i++) |
f4b00eab | 282 | memcg_kmem_uncharge_page(vm->pages[i], 0); |
9b6f7e16 | 283 | |
ac496bf4 | 284 | for (i = 0; i < NR_CACHED_STACKS; i++) { |
112166f8 CL |
285 | if (this_cpu_cmpxchg(cached_stacks[i], |
286 | NULL, tsk->stack_vm_area) != NULL) | |
ac496bf4 AL |
287 | continue; |
288 | ||
ac496bf4 AL |
289 | return; |
290 | } | |
ac496bf4 | 291 | |
0f110a9b | 292 | vfree_atomic(tsk->stack); |
ac496bf4 AL |
293 | return; |
294 | } | |
295 | #endif | |
296 | ||
297 | __free_pages(virt_to_page(tsk->stack), THREAD_SIZE_ORDER); | |
b69c49b7 | 298 | } |
0d15d74a | 299 | # else |
b235beea | 300 | static struct kmem_cache *thread_stack_cache; |
0d15d74a | 301 | |
9521d399 | 302 | static unsigned long *alloc_thread_stack_node(struct task_struct *tsk, |
0d15d74a TG |
303 | int node) |
304 | { | |
5eed6f1d RR |
305 | unsigned long *stack; |
306 | stack = kmem_cache_alloc_node(thread_stack_cache, THREADINFO_GFP, node); | |
8dcc1d34 | 307 | stack = kasan_reset_tag(stack); |
5eed6f1d RR |
308 | tsk->stack = stack; |
309 | return stack; | |
0d15d74a TG |
310 | } |
311 | ||
ba14a194 | 312 | static void free_thread_stack(struct task_struct *tsk) |
0d15d74a | 313 | { |
ba14a194 | 314 | kmem_cache_free(thread_stack_cache, tsk->stack); |
0d15d74a TG |
315 | } |
316 | ||
b235beea | 317 | void thread_stack_cache_init(void) |
0d15d74a | 318 | { |
f9d29946 DW |
319 | thread_stack_cache = kmem_cache_create_usercopy("thread_stack", |
320 | THREAD_SIZE, THREAD_SIZE, 0, 0, | |
321 | THREAD_SIZE, NULL); | |
b235beea | 322 | BUG_ON(thread_stack_cache == NULL); |
0d15d74a TG |
323 | } |
324 | # endif | |
b69c49b7 FT |
325 | #endif |
326 | ||
1da177e4 | 327 | /* SLAB cache for signal_struct structures (tsk->signal) */ |
e18b890b | 328 | static struct kmem_cache *signal_cachep; |
1da177e4 LT |
329 | |
330 | /* SLAB cache for sighand_struct structures (tsk->sighand) */ | |
e18b890b | 331 | struct kmem_cache *sighand_cachep; |
1da177e4 LT |
332 | |
333 | /* SLAB cache for files_struct structures (tsk->files) */ | |
e18b890b | 334 | struct kmem_cache *files_cachep; |
1da177e4 LT |
335 | |
336 | /* SLAB cache for fs_struct structures (tsk->fs) */ | |
e18b890b | 337 | struct kmem_cache *fs_cachep; |
1da177e4 LT |
338 | |
339 | /* SLAB cache for vm_area_struct structures */ | |
3928d4f5 | 340 | static struct kmem_cache *vm_area_cachep; |
1da177e4 LT |
341 | |
342 | /* SLAB cache for mm_struct structures (tsk->mm) */ | |
e18b890b | 343 | static struct kmem_cache *mm_cachep; |
1da177e4 | 344 | |
490fc053 | 345 | struct vm_area_struct *vm_area_alloc(struct mm_struct *mm) |
3928d4f5 | 346 | { |
a670468f | 347 | struct vm_area_struct *vma; |
490fc053 | 348 | |
a670468f | 349 | vma = kmem_cache_alloc(vm_area_cachep, GFP_KERNEL); |
027232da KS |
350 | if (vma) |
351 | vma_init(vma, mm); | |
490fc053 | 352 | return vma; |
3928d4f5 LT |
353 | } |
354 | ||
355 | struct vm_area_struct *vm_area_dup(struct vm_area_struct *orig) | |
356 | { | |
95faf699 LT |
357 | struct vm_area_struct *new = kmem_cache_alloc(vm_area_cachep, GFP_KERNEL); |
358 | ||
359 | if (new) { | |
cda099b3 QC |
360 | ASSERT_EXCLUSIVE_WRITER(orig->vm_flags); |
361 | ASSERT_EXCLUSIVE_WRITER(orig->vm_file); | |
362 | /* | |
363 | * orig->shared.rb may be modified concurrently, but the clone | |
364 | * will be reinitialized. | |
365 | */ | |
366 | *new = data_race(*orig); | |
95faf699 | 367 | INIT_LIST_HEAD(&new->anon_vma_chain); |
e39a4b33 | 368 | new->vm_next = new->vm_prev = NULL; |
95faf699 LT |
369 | } |
370 | return new; | |
3928d4f5 LT |
371 | } |
372 | ||
373 | void vm_area_free(struct vm_area_struct *vma) | |
374 | { | |
375 | kmem_cache_free(vm_area_cachep, vma); | |
376 | } | |
377 | ||
ba14a194 | 378 | static void account_kernel_stack(struct task_struct *tsk, int account) |
c6a7f572 | 379 | { |
ba14a194 AL |
380 | void *stack = task_stack_page(tsk); |
381 | struct vm_struct *vm = task_stack_vm_area(tsk); | |
382 | ||
27faca83 MS |
383 | if (vm) { |
384 | int i; | |
ba14a194 | 385 | |
27faca83 MS |
386 | for (i = 0; i < THREAD_SIZE / PAGE_SIZE; i++) |
387 | mod_lruvec_page_state(vm->pages[i], NR_KERNEL_STACK_KB, | |
388 | account * (PAGE_SIZE / 1024)); | |
389 | } else { | |
390 | /* All stack pages are in the same node. */ | |
da3ceeff | 391 | mod_lruvec_kmem_state(stack, NR_KERNEL_STACK_KB, |
991e7673 | 392 | account * (THREAD_SIZE / 1024)); |
27faca83 | 393 | } |
c6a7f572 KM |
394 | } |
395 | ||
9b6f7e16 RG |
396 | static int memcg_charge_kernel_stack(struct task_struct *tsk) |
397 | { | |
398 | #ifdef CONFIG_VMAP_STACK | |
399 | struct vm_struct *vm = task_stack_vm_area(tsk); | |
400 | int ret; | |
401 | ||
991e7673 SB |
402 | BUILD_BUG_ON(IS_ENABLED(CONFIG_VMAP_STACK) && PAGE_SIZE % 1024 != 0); |
403 | ||
9b6f7e16 RG |
404 | if (vm) { |
405 | int i; | |
406 | ||
991e7673 SB |
407 | BUG_ON(vm->nr_pages != THREAD_SIZE / PAGE_SIZE); |
408 | ||
9b6f7e16 RG |
409 | for (i = 0; i < THREAD_SIZE / PAGE_SIZE; i++) { |
410 | /* | |
bcfe06bf RG |
411 | * If memcg_kmem_charge_page() fails, page's |
412 | * memory cgroup pointer is NULL, and | |
413 | * memcg_kmem_uncharge_page() in free_thread_stack() | |
414 | * will ignore this page. | |
9b6f7e16 | 415 | */ |
f4b00eab RG |
416 | ret = memcg_kmem_charge_page(vm->pages[i], GFP_KERNEL, |
417 | 0); | |
9b6f7e16 RG |
418 | if (ret) |
419 | return ret; | |
9b6f7e16 RG |
420 | } |
421 | } | |
422 | #endif | |
423 | return 0; | |
424 | } | |
425 | ||
68f24b08 | 426 | static void release_task_stack(struct task_struct *tsk) |
1da177e4 | 427 | { |
2f064a59 | 428 | if (WARN_ON(READ_ONCE(tsk->__state) != TASK_DEAD)) |
405c0759 AL |
429 | return; /* Better to leak the stack than to free prematurely */ |
430 | ||
ba14a194 | 431 | account_kernel_stack(tsk, -1); |
ba14a194 | 432 | free_thread_stack(tsk); |
68f24b08 AL |
433 | tsk->stack = NULL; |
434 | #ifdef CONFIG_VMAP_STACK | |
435 | tsk->stack_vm_area = NULL; | |
436 | #endif | |
437 | } | |
438 | ||
439 | #ifdef CONFIG_THREAD_INFO_IN_TASK | |
440 | void put_task_stack(struct task_struct *tsk) | |
441 | { | |
f0b89d39 | 442 | if (refcount_dec_and_test(&tsk->stack_refcount)) |
68f24b08 AL |
443 | release_task_stack(tsk); |
444 | } | |
445 | #endif | |
446 | ||
447 | void free_task(struct task_struct *tsk) | |
448 | { | |
d08b9f0c ST |
449 | scs_release(tsk); |
450 | ||
68f24b08 AL |
451 | #ifndef CONFIG_THREAD_INFO_IN_TASK |
452 | /* | |
453 | * The task is finally done with both the stack and thread_info, | |
454 | * so free both. | |
455 | */ | |
456 | release_task_stack(tsk); | |
457 | #else | |
458 | /* | |
459 | * If the task had a separate stack allocation, it should be gone | |
460 | * by now. | |
461 | */ | |
f0b89d39 | 462 | WARN_ON_ONCE(refcount_read(&tsk->stack_refcount) != 0); |
68f24b08 | 463 | #endif |
23f78d4a | 464 | rt_mutex_debug_task_free(tsk); |
fb52607a | 465 | ftrace_graph_exit_task(tsk); |
f19b9f74 | 466 | arch_release_task_struct(tsk); |
1da5c46f ON |
467 | if (tsk->flags & PF_KTHREAD) |
468 | free_kthread_struct(tsk); | |
1da177e4 LT |
469 | free_task_struct(tsk); |
470 | } | |
471 | EXPORT_SYMBOL(free_task); | |
472 | ||
d70f2a14 AM |
473 | #ifdef CONFIG_MMU |
474 | static __latent_entropy int dup_mmap(struct mm_struct *mm, | |
475 | struct mm_struct *oldmm) | |
476 | { | |
477 | struct vm_area_struct *mpnt, *tmp, *prev, **pprev; | |
478 | struct rb_node **rb_link, *rb_parent; | |
479 | int retval; | |
480 | unsigned long charge; | |
481 | LIST_HEAD(uf); | |
482 | ||
483 | uprobe_start_dup_mmap(); | |
d8ed45c5 | 484 | if (mmap_write_lock_killable(oldmm)) { |
d70f2a14 AM |
485 | retval = -EINTR; |
486 | goto fail_uprobe_end; | |
487 | } | |
488 | flush_cache_dup_mm(oldmm); | |
489 | uprobe_dup_mmap(oldmm, mm); | |
490 | /* | |
491 | * Not linked in yet - no deadlock potential: | |
492 | */ | |
aaa2cc56 | 493 | mmap_write_lock_nested(mm, SINGLE_DEPTH_NESTING); |
d70f2a14 AM |
494 | |
495 | /* No ordering required: file already has been exposed. */ | |
496 | RCU_INIT_POINTER(mm->exe_file, get_mm_exe_file(oldmm)); | |
497 | ||
498 | mm->total_vm = oldmm->total_vm; | |
499 | mm->data_vm = oldmm->data_vm; | |
500 | mm->exec_vm = oldmm->exec_vm; | |
501 | mm->stack_vm = oldmm->stack_vm; | |
502 | ||
503 | rb_link = &mm->mm_rb.rb_node; | |
504 | rb_parent = NULL; | |
505 | pprev = &mm->mmap; | |
506 | retval = ksm_fork(mm, oldmm); | |
507 | if (retval) | |
508 | goto out; | |
509 | retval = khugepaged_fork(mm, oldmm); | |
510 | if (retval) | |
511 | goto out; | |
512 | ||
513 | prev = NULL; | |
514 | for (mpnt = oldmm->mmap; mpnt; mpnt = mpnt->vm_next) { | |
515 | struct file *file; | |
516 | ||
517 | if (mpnt->vm_flags & VM_DONTCOPY) { | |
518 | vm_stat_account(mm, mpnt->vm_flags, -vma_pages(mpnt)); | |
519 | continue; | |
520 | } | |
521 | charge = 0; | |
655c79bb TH |
522 | /* |
523 | * Don't duplicate many vmas if we've been oom-killed (for | |
524 | * example) | |
525 | */ | |
526 | if (fatal_signal_pending(current)) { | |
527 | retval = -EINTR; | |
528 | goto out; | |
529 | } | |
d70f2a14 AM |
530 | if (mpnt->vm_flags & VM_ACCOUNT) { |
531 | unsigned long len = vma_pages(mpnt); | |
532 | ||
533 | if (security_vm_enough_memory_mm(oldmm, len)) /* sic */ | |
534 | goto fail_nomem; | |
535 | charge = len; | |
536 | } | |
3928d4f5 | 537 | tmp = vm_area_dup(mpnt); |
d70f2a14 AM |
538 | if (!tmp) |
539 | goto fail_nomem; | |
d70f2a14 AM |
540 | retval = vma_dup_policy(mpnt, tmp); |
541 | if (retval) | |
542 | goto fail_nomem_policy; | |
543 | tmp->vm_mm = mm; | |
544 | retval = dup_userfaultfd(tmp, &uf); | |
545 | if (retval) | |
546 | goto fail_nomem_anon_vma_fork; | |
547 | if (tmp->vm_flags & VM_WIPEONFORK) { | |
93949bb2 LX |
548 | /* |
549 | * VM_WIPEONFORK gets a clean slate in the child. | |
550 | * Don't prepare anon_vma until fault since we don't | |
551 | * copy page for current vma. | |
552 | */ | |
d70f2a14 | 553 | tmp->anon_vma = NULL; |
d70f2a14 AM |
554 | } else if (anon_vma_fork(tmp, mpnt)) |
555 | goto fail_nomem_anon_vma_fork; | |
556 | tmp->vm_flags &= ~(VM_LOCKED | VM_LOCKONFAULT); | |
d70f2a14 AM |
557 | file = tmp->vm_file; |
558 | if (file) { | |
559 | struct inode *inode = file_inode(file); | |
560 | struct address_space *mapping = file->f_mapping; | |
561 | ||
562 | get_file(file); | |
563 | if (tmp->vm_flags & VM_DENYWRITE) | |
73eb7f9a | 564 | put_write_access(inode); |
d70f2a14 AM |
565 | i_mmap_lock_write(mapping); |
566 | if (tmp->vm_flags & VM_SHARED) | |
cf508b58 | 567 | mapping_allow_writable(mapping); |
d70f2a14 AM |
568 | flush_dcache_mmap_lock(mapping); |
569 | /* insert tmp into the share list, just after mpnt */ | |
570 | vma_interval_tree_insert_after(tmp, mpnt, | |
571 | &mapping->i_mmap); | |
572 | flush_dcache_mmap_unlock(mapping); | |
573 | i_mmap_unlock_write(mapping); | |
574 | } | |
575 | ||
576 | /* | |
577 | * Clear hugetlb-related page reserves for children. This only | |
578 | * affects MAP_PRIVATE mappings. Faults generated by the child | |
579 | * are not guaranteed to succeed, even if read-only | |
580 | */ | |
581 | if (is_vm_hugetlb_page(tmp)) | |
582 | reset_vma_resv_huge_pages(tmp); | |
583 | ||
584 | /* | |
585 | * Link in the new vma and copy the page table entries. | |
586 | */ | |
587 | *pprev = tmp; | |
588 | pprev = &tmp->vm_next; | |
589 | tmp->vm_prev = prev; | |
590 | prev = tmp; | |
591 | ||
592 | __vma_link_rb(mm, tmp, rb_link, rb_parent); | |
593 | rb_link = &tmp->vm_rb.rb_right; | |
594 | rb_parent = &tmp->vm_rb; | |
595 | ||
596 | mm->map_count++; | |
597 | if (!(tmp->vm_flags & VM_WIPEONFORK)) | |
c78f4636 | 598 | retval = copy_page_range(tmp, mpnt); |
d70f2a14 AM |
599 | |
600 | if (tmp->vm_ops && tmp->vm_ops->open) | |
601 | tmp->vm_ops->open(tmp); | |
602 | ||
603 | if (retval) | |
604 | goto out; | |
605 | } | |
606 | /* a new mm has just been created */ | |
1ed0cc5a | 607 | retval = arch_dup_mmap(oldmm, mm); |
d70f2a14 | 608 | out: |
d8ed45c5 | 609 | mmap_write_unlock(mm); |
d70f2a14 | 610 | flush_tlb_mm(oldmm); |
d8ed45c5 | 611 | mmap_write_unlock(oldmm); |
d70f2a14 AM |
612 | dup_userfaultfd_complete(&uf); |
613 | fail_uprobe_end: | |
614 | uprobe_end_dup_mmap(); | |
615 | return retval; | |
616 | fail_nomem_anon_vma_fork: | |
617 | mpol_put(vma_policy(tmp)); | |
618 | fail_nomem_policy: | |
3928d4f5 | 619 | vm_area_free(tmp); |
d70f2a14 AM |
620 | fail_nomem: |
621 | retval = -ENOMEM; | |
622 | vm_unacct_memory(charge); | |
623 | goto out; | |
624 | } | |
625 | ||
626 | static inline int mm_alloc_pgd(struct mm_struct *mm) | |
627 | { | |
628 | mm->pgd = pgd_alloc(mm); | |
629 | if (unlikely(!mm->pgd)) | |
630 | return -ENOMEM; | |
631 | return 0; | |
632 | } | |
633 | ||
634 | static inline void mm_free_pgd(struct mm_struct *mm) | |
635 | { | |
636 | pgd_free(mm, mm->pgd); | |
637 | } | |
638 | #else | |
639 | static int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm) | |
640 | { | |
d8ed45c5 | 641 | mmap_write_lock(oldmm); |
d70f2a14 | 642 | RCU_INIT_POINTER(mm->exe_file, get_mm_exe_file(oldmm)); |
d8ed45c5 | 643 | mmap_write_unlock(oldmm); |
d70f2a14 AM |
644 | return 0; |
645 | } | |
646 | #define mm_alloc_pgd(mm) (0) | |
647 | #define mm_free_pgd(mm) | |
648 | #endif /* CONFIG_MMU */ | |
649 | ||
650 | static void check_mm(struct mm_struct *mm) | |
651 | { | |
652 | int i; | |
653 | ||
8495f7e6 SPP |
654 | BUILD_BUG_ON_MSG(ARRAY_SIZE(resident_page_types) != NR_MM_COUNTERS, |
655 | "Please make sure 'struct resident_page_types[]' is updated as well"); | |
656 | ||
d70f2a14 AM |
657 | for (i = 0; i < NR_MM_COUNTERS; i++) { |
658 | long x = atomic_long_read(&mm->rss_stat.count[i]); | |
659 | ||
660 | if (unlikely(x)) | |
8495f7e6 SPP |
661 | pr_alert("BUG: Bad rss-counter state mm:%p type:%s val:%ld\n", |
662 | mm, resident_page_types[i], x); | |
d70f2a14 AM |
663 | } |
664 | ||
665 | if (mm_pgtables_bytes(mm)) | |
666 | pr_alert("BUG: non-zero pgtables_bytes on freeing mm: %ld\n", | |
667 | mm_pgtables_bytes(mm)); | |
668 | ||
669 | #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS | |
670 | VM_BUG_ON_MM(mm->pmd_huge_pte, mm); | |
671 | #endif | |
672 | } | |
673 | ||
674 | #define allocate_mm() (kmem_cache_alloc(mm_cachep, GFP_KERNEL)) | |
675 | #define free_mm(mm) (kmem_cache_free(mm_cachep, (mm))) | |
676 | ||
677 | /* | |
678 | * Called when the last reference to the mm | |
679 | * is dropped: either by a lazy thread or by | |
680 | * mmput. Free the page directory and the mm. | |
681 | */ | |
d34bc48f | 682 | void __mmdrop(struct mm_struct *mm) |
d70f2a14 AM |
683 | { |
684 | BUG_ON(mm == &init_mm); | |
3eda69c9 MR |
685 | WARN_ON_ONCE(mm == current->mm); |
686 | WARN_ON_ONCE(mm == current->active_mm); | |
d70f2a14 AM |
687 | mm_free_pgd(mm); |
688 | destroy_context(mm); | |
984cfe4e | 689 | mmu_notifier_subscriptions_destroy(mm); |
d70f2a14 AM |
690 | check_mm(mm); |
691 | put_user_ns(mm->user_ns); | |
692 | free_mm(mm); | |
693 | } | |
d34bc48f | 694 | EXPORT_SYMBOL_GPL(__mmdrop); |
d70f2a14 AM |
695 | |
696 | static void mmdrop_async_fn(struct work_struct *work) | |
697 | { | |
698 | struct mm_struct *mm; | |
699 | ||
700 | mm = container_of(work, struct mm_struct, async_put_work); | |
701 | __mmdrop(mm); | |
702 | } | |
703 | ||
704 | static void mmdrop_async(struct mm_struct *mm) | |
705 | { | |
706 | if (unlikely(atomic_dec_and_test(&mm->mm_count))) { | |
707 | INIT_WORK(&mm->async_put_work, mmdrop_async_fn); | |
708 | schedule_work(&mm->async_put_work); | |
709 | } | |
710 | } | |
711 | ||
ea6d290c ON |
712 | static inline void free_signal_struct(struct signal_struct *sig) |
713 | { | |
97101eb4 | 714 | taskstats_tgid_free(sig); |
1c5354de | 715 | sched_autogroup_exit(sig); |
7283094e MH |
716 | /* |
717 | * __mmdrop is not safe to call from softirq context on x86 due to | |
718 | * pgd_dtor so postpone it to the async context | |
719 | */ | |
26db62f1 | 720 | if (sig->oom_mm) |
7283094e | 721 | mmdrop_async(sig->oom_mm); |
ea6d290c ON |
722 | kmem_cache_free(signal_cachep, sig); |
723 | } | |
724 | ||
725 | static inline void put_signal_struct(struct signal_struct *sig) | |
726 | { | |
60d4de3f | 727 | if (refcount_dec_and_test(&sig->sigcnt)) |
ea6d290c ON |
728 | free_signal_struct(sig); |
729 | } | |
730 | ||
158d9ebd | 731 | void __put_task_struct(struct task_struct *tsk) |
1da177e4 | 732 | { |
270f722d | 733 | WARN_ON(!tsk->exit_state); |
ec1d2819 | 734 | WARN_ON(refcount_read(&tsk->usage)); |
1da177e4 LT |
735 | WARN_ON(tsk == current); |
736 | ||
0f212204 | 737 | io_uring_free(tsk); |
2e91fa7f | 738 | cgroup_free(tsk); |
16d51a59 | 739 | task_numa_free(tsk, true); |
1a2a4d06 | 740 | security_task_free(tsk); |
a10787e6 | 741 | bpf_task_storage_free(tsk); |
e0e81739 | 742 | exit_creds(tsk); |
35df17c5 | 743 | delayacct_tsk_free(tsk); |
ea6d290c | 744 | put_signal_struct(tsk->signal); |
6e33cad0 | 745 | sched_core_free(tsk); |
1da177e4 LT |
746 | |
747 | if (!profile_handoff_task(tsk)) | |
748 | free_task(tsk); | |
749 | } | |
77c100c8 | 750 | EXPORT_SYMBOL_GPL(__put_task_struct); |
1da177e4 | 751 | |
6c0a9fa6 | 752 | void __init __weak arch_task_cache_init(void) { } |
61c4628b | 753 | |
ff691f6e HS |
754 | /* |
755 | * set_max_threads | |
756 | */ | |
16db3d3f | 757 | static void set_max_threads(unsigned int max_threads_suggested) |
ff691f6e | 758 | { |
ac1b398d | 759 | u64 threads; |
ca79b0c2 | 760 | unsigned long nr_pages = totalram_pages(); |
ff691f6e HS |
761 | |
762 | /* | |
ac1b398d HS |
763 | * The number of threads shall be limited such that the thread |
764 | * structures may only consume a small part of the available memory. | |
ff691f6e | 765 | */ |
3d6357de | 766 | if (fls64(nr_pages) + fls64(PAGE_SIZE) > 64) |
ac1b398d HS |
767 | threads = MAX_THREADS; |
768 | else | |
3d6357de | 769 | threads = div64_u64((u64) nr_pages * (u64) PAGE_SIZE, |
ac1b398d HS |
770 | (u64) THREAD_SIZE * 8UL); |
771 | ||
16db3d3f HS |
772 | if (threads > max_threads_suggested) |
773 | threads = max_threads_suggested; | |
774 | ||
ac1b398d | 775 | max_threads = clamp_t(u64, threads, MIN_THREADS, MAX_THREADS); |
ff691f6e HS |
776 | } |
777 | ||
5aaeb5c0 IM |
778 | #ifdef CONFIG_ARCH_WANTS_DYNAMIC_TASK_STRUCT |
779 | /* Initialized by the architecture: */ | |
780 | int arch_task_struct_size __read_mostly; | |
781 | #endif | |
0c8c0f03 | 782 | |
4189ff23 | 783 | #ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR |
5905429a KC |
784 | static void task_struct_whitelist(unsigned long *offset, unsigned long *size) |
785 | { | |
786 | /* Fetch thread_struct whitelist for the architecture. */ | |
787 | arch_thread_struct_whitelist(offset, size); | |
788 | ||
789 | /* | |
790 | * Handle zero-sized whitelist or empty thread_struct, otherwise | |
791 | * adjust offset to position of thread_struct in task_struct. | |
792 | */ | |
793 | if (unlikely(*size == 0)) | |
794 | *offset = 0; | |
795 | else | |
796 | *offset += offsetof(struct task_struct, thread); | |
797 | } | |
4189ff23 | 798 | #endif /* CONFIG_ARCH_TASK_STRUCT_ALLOCATOR */ |
5905429a | 799 | |
ff691f6e | 800 | void __init fork_init(void) |
1da177e4 | 801 | { |
25f9c081 | 802 | int i; |
f5e10287 | 803 | #ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR |
1da177e4 | 804 | #ifndef ARCH_MIN_TASKALIGN |
e274795e | 805 | #define ARCH_MIN_TASKALIGN 0 |
1da177e4 | 806 | #endif |
95cb64c1 | 807 | int align = max_t(int, L1_CACHE_BYTES, ARCH_MIN_TASKALIGN); |
5905429a | 808 | unsigned long useroffset, usersize; |
e274795e | 809 | |
1da177e4 | 810 | /* create a slab on which task_structs can be allocated */ |
5905429a KC |
811 | task_struct_whitelist(&useroffset, &usersize); |
812 | task_struct_cachep = kmem_cache_create_usercopy("task_struct", | |
e274795e | 813 | arch_task_struct_size, align, |
5905429a KC |
814 | SLAB_PANIC|SLAB_ACCOUNT, |
815 | useroffset, usersize, NULL); | |
1da177e4 LT |
816 | #endif |
817 | ||
61c4628b SS |
818 | /* do the arch specific task caches init */ |
819 | arch_task_cache_init(); | |
820 | ||
16db3d3f | 821 | set_max_threads(MAX_THREADS); |
1da177e4 LT |
822 | |
823 | init_task.signal->rlim[RLIMIT_NPROC].rlim_cur = max_threads/2; | |
824 | init_task.signal->rlim[RLIMIT_NPROC].rlim_max = max_threads/2; | |
825 | init_task.signal->rlim[RLIMIT_SIGPENDING] = | |
826 | init_task.signal->rlim[RLIMIT_NPROC]; | |
b376c3e1 | 827 | |
21d1c5e3 | 828 | for (i = 0; i < MAX_PER_NAMESPACE_UCOUNTS; i++) |
25f9c081 | 829 | init_user_ns.ucount_max[i] = max_threads/2; |
19659c59 | 830 | |
c1ada3dc AG |
831 | set_rlimit_ucount_max(&init_user_ns, UCOUNT_RLIMIT_NPROC, task_rlimit(&init_task, RLIMIT_NPROC)); |
832 | set_rlimit_ucount_max(&init_user_ns, UCOUNT_RLIMIT_MSGQUEUE, task_rlimit(&init_task, RLIMIT_MSGQUEUE)); | |
833 | set_rlimit_ucount_max(&init_user_ns, UCOUNT_RLIMIT_SIGPENDING, task_rlimit(&init_task, RLIMIT_SIGPENDING)); | |
834 | set_rlimit_ucount_max(&init_user_ns, UCOUNT_RLIMIT_MEMLOCK, task_rlimit(&init_task, RLIMIT_MEMLOCK)); | |
21d1c5e3 | 835 | |
19659c59 HR |
836 | #ifdef CONFIG_VMAP_STACK |
837 | cpuhp_setup_state(CPUHP_BP_PREPARE_DYN, "fork:vm_stack_cache", | |
838 | NULL, free_vm_stack_cache); | |
839 | #endif | |
b09be676 | 840 | |
d08b9f0c ST |
841 | scs_init(); |
842 | ||
b09be676 | 843 | lockdep_init_task(&init_task); |
aad42dd4 | 844 | uprobes_init(); |
1da177e4 LT |
845 | } |
846 | ||
52f5684c | 847 | int __weak arch_dup_task_struct(struct task_struct *dst, |
61c4628b SS |
848 | struct task_struct *src) |
849 | { | |
850 | *dst = *src; | |
851 | return 0; | |
852 | } | |
853 | ||
d4311ff1 AT |
854 | void set_task_stack_end_magic(struct task_struct *tsk) |
855 | { | |
856 | unsigned long *stackend; | |
857 | ||
858 | stackend = end_of_stack(tsk); | |
859 | *stackend = STACK_END_MAGIC; /* for overflow detection */ | |
860 | } | |
861 | ||
725fc629 | 862 | static struct task_struct *dup_task_struct(struct task_struct *orig, int node) |
1da177e4 LT |
863 | { |
864 | struct task_struct *tsk; | |
b235beea | 865 | unsigned long *stack; |
0f4991e8 | 866 | struct vm_struct *stack_vm_area __maybe_unused; |
3e26c149 | 867 | int err; |
1da177e4 | 868 | |
725fc629 AK |
869 | if (node == NUMA_NO_NODE) |
870 | node = tsk_fork_get_node(orig); | |
504f52b5 | 871 | tsk = alloc_task_struct_node(node); |
1da177e4 LT |
872 | if (!tsk) |
873 | return NULL; | |
874 | ||
b235beea LT |
875 | stack = alloc_thread_stack_node(tsk, node); |
876 | if (!stack) | |
f19b9f74 | 877 | goto free_tsk; |
1da177e4 | 878 | |
9b6f7e16 RG |
879 | if (memcg_charge_kernel_stack(tsk)) |
880 | goto free_stack; | |
881 | ||
ba14a194 AL |
882 | stack_vm_area = task_stack_vm_area(tsk); |
883 | ||
fb0a685c | 884 | err = arch_dup_task_struct(tsk, orig); |
ba14a194 AL |
885 | |
886 | /* | |
887 | * arch_dup_task_struct() clobbers the stack-related fields. Make | |
888 | * sure they're properly initialized before using any stack-related | |
889 | * functions again. | |
890 | */ | |
891 | tsk->stack = stack; | |
892 | #ifdef CONFIG_VMAP_STACK | |
893 | tsk->stack_vm_area = stack_vm_area; | |
894 | #endif | |
68f24b08 | 895 | #ifdef CONFIG_THREAD_INFO_IN_TASK |
f0b89d39 | 896 | refcount_set(&tsk->stack_refcount, 1); |
68f24b08 | 897 | #endif |
ba14a194 | 898 | |
164c33c6 | 899 | if (err) |
b235beea | 900 | goto free_stack; |
164c33c6 | 901 | |
d08b9f0c ST |
902 | err = scs_prepare(tsk, node); |
903 | if (err) | |
904 | goto free_stack; | |
905 | ||
dbd95212 KC |
906 | #ifdef CONFIG_SECCOMP |
907 | /* | |
908 | * We must handle setting up seccomp filters once we're under | |
909 | * the sighand lock in case orig has changed between now and | |
910 | * then. Until then, filter must be NULL to avoid messing up | |
911 | * the usage counts on the error path calling free_task. | |
912 | */ | |
913 | tsk->seccomp.filter = NULL; | |
914 | #endif | |
87bec58a AM |
915 | |
916 | setup_thread_stack(tsk, orig); | |
8e7cac79 | 917 | clear_user_return_notifier(tsk); |
f26f9aff | 918 | clear_tsk_need_resched(tsk); |
d4311ff1 | 919 | set_task_stack_end_magic(tsk); |
1446e1df | 920 | clear_syscall_work_syscall_user_dispatch(tsk); |
1da177e4 | 921 | |
050e9baa | 922 | #ifdef CONFIG_STACKPROTECTOR |
7cd815bc | 923 | tsk->stack_canary = get_random_canary(); |
0a425405 | 924 | #endif |
3bd37062 SAS |
925 | if (orig->cpus_ptr == &orig->cpus_mask) |
926 | tsk->cpus_ptr = &tsk->cpus_mask; | |
0a425405 | 927 | |
fb0a685c | 928 | /* |
0ff7b2cf EB |
929 | * One for the user space visible state that goes away when reaped. |
930 | * One for the scheduler. | |
fb0a685c | 931 | */ |
0ff7b2cf EB |
932 | refcount_set(&tsk->rcu_users, 2); |
933 | /* One for the rcu users */ | |
934 | refcount_set(&tsk->usage, 1); | |
6c5c9341 | 935 | #ifdef CONFIG_BLK_DEV_IO_TRACE |
2056a782 | 936 | tsk->btrace_seq = 0; |
6c5c9341 | 937 | #endif |
a0aa7f68 | 938 | tsk->splice_pipe = NULL; |
5640f768 | 939 | tsk->task_frag.page = NULL; |
093e5840 | 940 | tsk->wake_q.next = NULL; |
ff244303 | 941 | tsk->pf_io_worker = NULL; |
c6a7f572 | 942 | |
ba14a194 | 943 | account_kernel_stack(tsk, 1); |
c6a7f572 | 944 | |
5c9a8750 | 945 | kcov_task_init(tsk); |
5fbda3ec | 946 | kmap_local_fork(tsk); |
5c9a8750 | 947 | |
e41d5818 DV |
948 | #ifdef CONFIG_FAULT_INJECTION |
949 | tsk->fail_nth = 0; | |
950 | #endif | |
951 | ||
2c323017 JB |
952 | #ifdef CONFIG_BLK_CGROUP |
953 | tsk->throttle_queue = NULL; | |
954 | tsk->use_memdelay = 0; | |
955 | #endif | |
956 | ||
d46eb14b SB |
957 | #ifdef CONFIG_MEMCG |
958 | tsk->active_memcg = NULL; | |
959 | #endif | |
1da177e4 | 960 | return tsk; |
61c4628b | 961 | |
b235beea | 962 | free_stack: |
ba14a194 | 963 | free_thread_stack(tsk); |
f19b9f74 | 964 | free_tsk: |
61c4628b SS |
965 | free_task_struct(tsk); |
966 | return NULL; | |
1da177e4 LT |
967 | } |
968 | ||
23ff4440 | 969 | __cacheline_aligned_in_smp DEFINE_SPINLOCK(mmlist_lock); |
1da177e4 | 970 | |
4cb0e11b HK |
971 | static unsigned long default_dump_filter = MMF_DUMP_FILTER_DEFAULT; |
972 | ||
973 | static int __init coredump_filter_setup(char *s) | |
974 | { | |
975 | default_dump_filter = | |
976 | (simple_strtoul(s, NULL, 0) << MMF_DUMP_FILTER_SHIFT) & | |
977 | MMF_DUMP_FILTER_MASK; | |
978 | return 1; | |
979 | } | |
980 | ||
981 | __setup("coredump_filter=", coredump_filter_setup); | |
982 | ||
1da177e4 LT |
983 | #include <linux/init_task.h> |
984 | ||
858f0993 AD |
985 | static void mm_init_aio(struct mm_struct *mm) |
986 | { | |
987 | #ifdef CONFIG_AIO | |
988 | spin_lock_init(&mm->ioctx_lock); | |
db446a08 | 989 | mm->ioctx_table = NULL; |
858f0993 AD |
990 | #endif |
991 | } | |
992 | ||
c3f3ce04 AA |
993 | static __always_inline void mm_clear_owner(struct mm_struct *mm, |
994 | struct task_struct *p) | |
995 | { | |
996 | #ifdef CONFIG_MEMCG | |
997 | if (mm->owner == p) | |
998 | WRITE_ONCE(mm->owner, NULL); | |
999 | #endif | |
1000 | } | |
1001 | ||
33144e84 VD |
1002 | static void mm_init_owner(struct mm_struct *mm, struct task_struct *p) |
1003 | { | |
1004 | #ifdef CONFIG_MEMCG | |
1005 | mm->owner = p; | |
1006 | #endif | |
1007 | } | |
1008 | ||
82e69a12 FY |
1009 | static void mm_init_pasid(struct mm_struct *mm) |
1010 | { | |
1011 | #ifdef CONFIG_IOMMU_SUPPORT | |
1012 | mm->pasid = INIT_PASID; | |
1013 | #endif | |
1014 | } | |
1015 | ||
355627f5 EB |
1016 | static void mm_init_uprobes_state(struct mm_struct *mm) |
1017 | { | |
1018 | #ifdef CONFIG_UPROBES | |
1019 | mm->uprobes_state.xol_area = NULL; | |
1020 | #endif | |
1021 | } | |
1022 | ||
bfedb589 EB |
1023 | static struct mm_struct *mm_init(struct mm_struct *mm, struct task_struct *p, |
1024 | struct user_namespace *user_ns) | |
1da177e4 | 1025 | { |
41f727fd VD |
1026 | mm->mmap = NULL; |
1027 | mm->mm_rb = RB_ROOT; | |
1028 | mm->vmacache_seqnum = 0; | |
1da177e4 LT |
1029 | atomic_set(&mm->mm_users, 1); |
1030 | atomic_set(&mm->mm_count, 1); | |
57efa1fe | 1031 | seqcount_init(&mm->write_protect_seq); |
d8ed45c5 | 1032 | mmap_init_lock(mm); |
1da177e4 | 1033 | INIT_LIST_HEAD(&mm->mmlist); |
999d9fc1 | 1034 | mm->core_state = NULL; |
af5b0f6a | 1035 | mm_pgtables_bytes_init(mm); |
41f727fd VD |
1036 | mm->map_count = 0; |
1037 | mm->locked_vm = 0; | |
70f8a3ca | 1038 | atomic64_set(&mm->pinned_vm, 0); |
d559db08 | 1039 | memset(&mm->rss_stat, 0, sizeof(mm->rss_stat)); |
1da177e4 | 1040 | spin_lock_init(&mm->page_table_lock); |
88aa7cc6 | 1041 | spin_lock_init(&mm->arg_lock); |
41f727fd | 1042 | mm_init_cpumask(mm); |
858f0993 | 1043 | mm_init_aio(mm); |
cf475ad2 | 1044 | mm_init_owner(mm, p); |
82e69a12 | 1045 | mm_init_pasid(mm); |
2b7e8665 | 1046 | RCU_INIT_POINTER(mm->exe_file, NULL); |
984cfe4e | 1047 | mmu_notifier_subscriptions_init(mm); |
16af97dc | 1048 | init_tlb_flush_pending(mm); |
41f727fd VD |
1049 | #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS |
1050 | mm->pmd_huge_pte = NULL; | |
1051 | #endif | |
355627f5 | 1052 | mm_init_uprobes_state(mm); |
1da177e4 | 1053 | |
a0715cc2 AT |
1054 | if (current->mm) { |
1055 | mm->flags = current->mm->flags & MMF_INIT_MASK; | |
1056 | mm->def_flags = current->mm->def_flags & VM_INIT_DEF_MASK; | |
1057 | } else { | |
1058 | mm->flags = default_dump_filter; | |
1da177e4 | 1059 | mm->def_flags = 0; |
a0715cc2 AT |
1060 | } |
1061 | ||
41f727fd VD |
1062 | if (mm_alloc_pgd(mm)) |
1063 | goto fail_nopgd; | |
1064 | ||
1065 | if (init_new_context(p, mm)) | |
1066 | goto fail_nocontext; | |
78fb7466 | 1067 | |
bfedb589 | 1068 | mm->user_ns = get_user_ns(user_ns); |
41f727fd VD |
1069 | return mm; |
1070 | ||
1071 | fail_nocontext: | |
1072 | mm_free_pgd(mm); | |
1073 | fail_nopgd: | |
1da177e4 LT |
1074 | free_mm(mm); |
1075 | return NULL; | |
1076 | } | |
1077 | ||
1078 | /* | |
1079 | * Allocate and initialize an mm_struct. | |
1080 | */ | |
fb0a685c | 1081 | struct mm_struct *mm_alloc(void) |
1da177e4 | 1082 | { |
fb0a685c | 1083 | struct mm_struct *mm; |
1da177e4 LT |
1084 | |
1085 | mm = allocate_mm(); | |
de03c72c KM |
1086 | if (!mm) |
1087 | return NULL; | |
1088 | ||
1089 | memset(mm, 0, sizeof(*mm)); | |
bfedb589 | 1090 | return mm_init(mm, current, current_user_ns()); |
1da177e4 LT |
1091 | } |
1092 | ||
ec8d7c14 MH |
1093 | static inline void __mmput(struct mm_struct *mm) |
1094 | { | |
1095 | VM_BUG_ON(atomic_read(&mm->mm_users)); | |
1096 | ||
1097 | uprobe_clear_state(mm); | |
1098 | exit_aio(mm); | |
1099 | ksm_exit(mm); | |
1100 | khugepaged_exit(mm); /* must run before exit_mmap */ | |
1101 | exit_mmap(mm); | |
6fcb52a5 | 1102 | mm_put_huge_zero_page(mm); |
ec8d7c14 MH |
1103 | set_mm_exe_file(mm, NULL); |
1104 | if (!list_empty(&mm->mmlist)) { | |
1105 | spin_lock(&mmlist_lock); | |
1106 | list_del(&mm->mmlist); | |
1107 | spin_unlock(&mmlist_lock); | |
1108 | } | |
1109 | if (mm->binfmt) | |
1110 | module_put(mm->binfmt->module); | |
1111 | mmdrop(mm); | |
1112 | } | |
1113 | ||
1da177e4 LT |
1114 | /* |
1115 | * Decrement the use count and release all resources for an mm. | |
1116 | */ | |
1117 | void mmput(struct mm_struct *mm) | |
1118 | { | |
0ae26f1b AM |
1119 | might_sleep(); |
1120 | ||
ec8d7c14 MH |
1121 | if (atomic_dec_and_test(&mm->mm_users)) |
1122 | __mmput(mm); | |
1123 | } | |
1124 | EXPORT_SYMBOL_GPL(mmput); | |
1125 | ||
a1b2289c SY |
1126 | #ifdef CONFIG_MMU |
1127 | static void mmput_async_fn(struct work_struct *work) | |
1128 | { | |
1129 | struct mm_struct *mm = container_of(work, struct mm_struct, | |
1130 | async_put_work); | |
1131 | ||
1132 | __mmput(mm); | |
1133 | } | |
1134 | ||
1135 | void mmput_async(struct mm_struct *mm) | |
1136 | { | |
1137 | if (atomic_dec_and_test(&mm->mm_users)) { | |
1138 | INIT_WORK(&mm->async_put_work, mmput_async_fn); | |
1139 | schedule_work(&mm->async_put_work); | |
1140 | } | |
1141 | } | |
1142 | #endif | |
1143 | ||
90f31d0e KK |
1144 | /** |
1145 | * set_mm_exe_file - change a reference to the mm's executable file | |
1146 | * | |
1147 | * This changes mm's executable file (shown as symlink /proc/[pid]/exe). | |
1148 | * | |
6e399cd1 DB |
1149 | * Main users are mmput() and sys_execve(). Callers prevent concurrent |
1150 | * invocations: in mmput() nobody alive left, in execve task is single | |
1151 | * threaded. sys_prctl(PR_SET_MM_MAP/EXE_FILE) also needs to set the | |
1152 | * mm->exe_file, but does so without using set_mm_exe_file() in order | |
a8ca6b13 | 1153 | * to avoid the need for any locks. |
90f31d0e | 1154 | */ |
38646013 JS |
1155 | void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file) |
1156 | { | |
6e399cd1 DB |
1157 | struct file *old_exe_file; |
1158 | ||
1159 | /* | |
1160 | * It is safe to dereference the exe_file without RCU as | |
1161 | * this function is only called if nobody else can access | |
1162 | * this mm -- see comment above for justification. | |
1163 | */ | |
1164 | old_exe_file = rcu_dereference_raw(mm->exe_file); | |
90f31d0e | 1165 | |
38646013 JS |
1166 | if (new_exe_file) |
1167 | get_file(new_exe_file); | |
90f31d0e KK |
1168 | rcu_assign_pointer(mm->exe_file, new_exe_file); |
1169 | if (old_exe_file) | |
1170 | fput(old_exe_file); | |
38646013 JS |
1171 | } |
1172 | ||
90f31d0e KK |
1173 | /** |
1174 | * get_mm_exe_file - acquire a reference to the mm's executable file | |
1175 | * | |
1176 | * Returns %NULL if mm has no associated executable file. | |
1177 | * User must release file via fput(). | |
1178 | */ | |
38646013 JS |
1179 | struct file *get_mm_exe_file(struct mm_struct *mm) |
1180 | { | |
1181 | struct file *exe_file; | |
1182 | ||
90f31d0e KK |
1183 | rcu_read_lock(); |
1184 | exe_file = rcu_dereference(mm->exe_file); | |
1185 | if (exe_file && !get_file_rcu(exe_file)) | |
1186 | exe_file = NULL; | |
1187 | rcu_read_unlock(); | |
38646013 JS |
1188 | return exe_file; |
1189 | } | |
11163348 | 1190 | EXPORT_SYMBOL(get_mm_exe_file); |
38646013 | 1191 | |
cd81a917 MG |
1192 | /** |
1193 | * get_task_exe_file - acquire a reference to the task's executable file | |
1194 | * | |
1195 | * Returns %NULL if task's mm (if any) has no associated executable file or | |
1196 | * this is a kernel thread with borrowed mm (see the comment above get_task_mm). | |
1197 | * User must release file via fput(). | |
1198 | */ | |
1199 | struct file *get_task_exe_file(struct task_struct *task) | |
1200 | { | |
1201 | struct file *exe_file = NULL; | |
1202 | struct mm_struct *mm; | |
1203 | ||
1204 | task_lock(task); | |
1205 | mm = task->mm; | |
1206 | if (mm) { | |
1207 | if (!(task->flags & PF_KTHREAD)) | |
1208 | exe_file = get_mm_exe_file(mm); | |
1209 | } | |
1210 | task_unlock(task); | |
1211 | return exe_file; | |
1212 | } | |
1213 | EXPORT_SYMBOL(get_task_exe_file); | |
38646013 | 1214 | |
1da177e4 LT |
1215 | /** |
1216 | * get_task_mm - acquire a reference to the task's mm | |
1217 | * | |
246bb0b1 | 1218 | * Returns %NULL if the task has no mm. Checks PF_KTHREAD (meaning |
1da177e4 LT |
1219 | * this kernel workthread has transiently adopted a user mm with use_mm, |
1220 | * to do its AIO) is not set and if so returns a reference to it, after | |
1221 | * bumping up the use count. User must release the mm via mmput() | |
1222 | * after use. Typically used by /proc and ptrace. | |
1223 | */ | |
1224 | struct mm_struct *get_task_mm(struct task_struct *task) | |
1225 | { | |
1226 | struct mm_struct *mm; | |
1227 | ||
1228 | task_lock(task); | |
1229 | mm = task->mm; | |
1230 | if (mm) { | |
246bb0b1 | 1231 | if (task->flags & PF_KTHREAD) |
1da177e4 LT |
1232 | mm = NULL; |
1233 | else | |
3fce371b | 1234 | mmget(mm); |
1da177e4 LT |
1235 | } |
1236 | task_unlock(task); | |
1237 | return mm; | |
1238 | } | |
1239 | EXPORT_SYMBOL_GPL(get_task_mm); | |
1240 | ||
8cdb878d CY |
1241 | struct mm_struct *mm_access(struct task_struct *task, unsigned int mode) |
1242 | { | |
1243 | struct mm_struct *mm; | |
1244 | int err; | |
1245 | ||
f7cfd871 | 1246 | err = down_read_killable(&task->signal->exec_update_lock); |
8cdb878d CY |
1247 | if (err) |
1248 | return ERR_PTR(err); | |
1249 | ||
1250 | mm = get_task_mm(task); | |
1251 | if (mm && mm != current->mm && | |
1252 | !ptrace_may_access(task, mode)) { | |
1253 | mmput(mm); | |
1254 | mm = ERR_PTR(-EACCES); | |
1255 | } | |
f7cfd871 | 1256 | up_read(&task->signal->exec_update_lock); |
8cdb878d CY |
1257 | |
1258 | return mm; | |
1259 | } | |
1260 | ||
57b59c4a | 1261 | static void complete_vfork_done(struct task_struct *tsk) |
c415c3b4 | 1262 | { |
d68b46fe | 1263 | struct completion *vfork; |
c415c3b4 | 1264 | |
d68b46fe ON |
1265 | task_lock(tsk); |
1266 | vfork = tsk->vfork_done; | |
1267 | if (likely(vfork)) { | |
1268 | tsk->vfork_done = NULL; | |
1269 | complete(vfork); | |
1270 | } | |
1271 | task_unlock(tsk); | |
1272 | } | |
1273 | ||
1274 | static int wait_for_vfork_done(struct task_struct *child, | |
1275 | struct completion *vfork) | |
1276 | { | |
1277 | int killed; | |
1278 | ||
1279 | freezer_do_not_count(); | |
76f969e8 | 1280 | cgroup_enter_frozen(); |
d68b46fe | 1281 | killed = wait_for_completion_killable(vfork); |
76f969e8 | 1282 | cgroup_leave_frozen(false); |
d68b46fe ON |
1283 | freezer_count(); |
1284 | ||
1285 | if (killed) { | |
1286 | task_lock(child); | |
1287 | child->vfork_done = NULL; | |
1288 | task_unlock(child); | |
1289 | } | |
1290 | ||
1291 | put_task_struct(child); | |
1292 | return killed; | |
c415c3b4 ON |
1293 | } |
1294 | ||
1da177e4 LT |
1295 | /* Please note the differences between mmput and mm_release. |
1296 | * mmput is called whenever we stop holding onto a mm_struct, | |
1297 | * error success whatever. | |
1298 | * | |
1299 | * mm_release is called after a mm_struct has been removed | |
1300 | * from the current process. | |
1301 | * | |
1302 | * This difference is important for error handling, when we | |
1303 | * only half set up a mm_struct for a new process and need to restore | |
1304 | * the old one. Because we mmput the new mm_struct before | |
1305 | * restoring the old one. . . | |
1306 | * Eric Biederman 10 January 1998 | |
1307 | */ | |
4610ba7a | 1308 | static void mm_release(struct task_struct *tsk, struct mm_struct *mm) |
1da177e4 | 1309 | { |
0326f5a9 SD |
1310 | uprobe_free_utask(tsk); |
1311 | ||
1da177e4 LT |
1312 | /* Get rid of any cached register state */ |
1313 | deactivate_mm(tsk, mm); | |
1314 | ||
fec1d011 | 1315 | /* |
735f2770 MH |
1316 | * Signal userspace if we're not exiting with a core dump |
1317 | * because we want to leave the value intact for debugging | |
1318 | * purposes. | |
fec1d011 | 1319 | */ |
9c8a8228 | 1320 | if (tsk->clear_child_tid) { |
735f2770 | 1321 | if (!(tsk->signal->flags & SIGNAL_GROUP_COREDUMP) && |
9c8a8228 ED |
1322 | atomic_read(&mm->mm_users) > 1) { |
1323 | /* | |
1324 | * We don't check the error code - if userspace has | |
1325 | * not set up a proper pointer then tough luck. | |
1326 | */ | |
1327 | put_user(0, tsk->clear_child_tid); | |
2de0db99 DB |
1328 | do_futex(tsk->clear_child_tid, FUTEX_WAKE, |
1329 | 1, NULL, NULL, 0, 0); | |
9c8a8228 | 1330 | } |
1da177e4 | 1331 | tsk->clear_child_tid = NULL; |
1da177e4 | 1332 | } |
f7505d64 KK |
1333 | |
1334 | /* | |
1335 | * All done, finally we can wake up parent and return this mm to him. | |
1336 | * Also kthread_stop() uses this completion for synchronization. | |
1337 | */ | |
1338 | if (tsk->vfork_done) | |
1339 | complete_vfork_done(tsk); | |
1da177e4 LT |
1340 | } |
1341 | ||
4610ba7a TG |
1342 | void exit_mm_release(struct task_struct *tsk, struct mm_struct *mm) |
1343 | { | |
150d7158 | 1344 | futex_exit_release(tsk); |
4610ba7a TG |
1345 | mm_release(tsk, mm); |
1346 | } | |
1347 | ||
1348 | void exec_mm_release(struct task_struct *tsk, struct mm_struct *mm) | |
1349 | { | |
150d7158 | 1350 | futex_exec_release(tsk); |
4610ba7a TG |
1351 | mm_release(tsk, mm); |
1352 | } | |
1353 | ||
13585fa0 NA |
1354 | /** |
1355 | * dup_mm() - duplicates an existing mm structure | |
1356 | * @tsk: the task_struct with which the new mm will be associated. | |
1357 | * @oldmm: the mm to duplicate. | |
1358 | * | |
1359 | * Allocates a new mm structure and duplicates the provided @oldmm structure | |
1360 | * content into it. | |
1361 | * | |
1362 | * Return: the duplicated mm or NULL on failure. | |
a0a7ec30 | 1363 | */ |
13585fa0 NA |
1364 | static struct mm_struct *dup_mm(struct task_struct *tsk, |
1365 | struct mm_struct *oldmm) | |
a0a7ec30 | 1366 | { |
13585fa0 | 1367 | struct mm_struct *mm; |
a0a7ec30 JD |
1368 | int err; |
1369 | ||
a0a7ec30 JD |
1370 | mm = allocate_mm(); |
1371 | if (!mm) | |
1372 | goto fail_nomem; | |
1373 | ||
1374 | memcpy(mm, oldmm, sizeof(*mm)); | |
1375 | ||
bfedb589 | 1376 | if (!mm_init(mm, tsk, mm->user_ns)) |
a0a7ec30 JD |
1377 | goto fail_nomem; |
1378 | ||
a0a7ec30 JD |
1379 | err = dup_mmap(mm, oldmm); |
1380 | if (err) | |
1381 | goto free_pt; | |
1382 | ||
1383 | mm->hiwater_rss = get_mm_rss(mm); | |
1384 | mm->hiwater_vm = mm->total_vm; | |
1385 | ||
801460d0 HS |
1386 | if (mm->binfmt && !try_module_get(mm->binfmt->module)) |
1387 | goto free_pt; | |
1388 | ||
a0a7ec30 JD |
1389 | return mm; |
1390 | ||
1391 | free_pt: | |
801460d0 HS |
1392 | /* don't put binfmt in mmput, we haven't got module yet */ |
1393 | mm->binfmt = NULL; | |
c3f3ce04 | 1394 | mm_init_owner(mm, NULL); |
a0a7ec30 JD |
1395 | mmput(mm); |
1396 | ||
1397 | fail_nomem: | |
1398 | return NULL; | |
a0a7ec30 JD |
1399 | } |
1400 | ||
fb0a685c | 1401 | static int copy_mm(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 | 1402 | { |
fb0a685c | 1403 | struct mm_struct *mm, *oldmm; |
1da177e4 LT |
1404 | |
1405 | tsk->min_flt = tsk->maj_flt = 0; | |
1406 | tsk->nvcsw = tsk->nivcsw = 0; | |
17406b82 MSB |
1407 | #ifdef CONFIG_DETECT_HUNG_TASK |
1408 | tsk->last_switch_count = tsk->nvcsw + tsk->nivcsw; | |
a2e51445 | 1409 | tsk->last_switch_time = 0; |
17406b82 | 1410 | #endif |
1da177e4 LT |
1411 | |
1412 | tsk->mm = NULL; | |
1413 | tsk->active_mm = NULL; | |
1414 | ||
1415 | /* | |
1416 | * Are we cloning a kernel thread? | |
1417 | * | |
1418 | * We need to steal a active VM for that.. | |
1419 | */ | |
1420 | oldmm = current->mm; | |
1421 | if (!oldmm) | |
1422 | return 0; | |
1423 | ||
615d6e87 DB |
1424 | /* initialize the new vmacache entries */ |
1425 | vmacache_flush(tsk); | |
1426 | ||
1da177e4 | 1427 | if (clone_flags & CLONE_VM) { |
3fce371b | 1428 | mmget(oldmm); |
1da177e4 | 1429 | mm = oldmm; |
a6895399 REB |
1430 | } else { |
1431 | mm = dup_mm(tsk, current->mm); | |
1432 | if (!mm) | |
1433 | return -ENOMEM; | |
1da177e4 LT |
1434 | } |
1435 | ||
1da177e4 LT |
1436 | tsk->mm = mm; |
1437 | tsk->active_mm = mm; | |
1438 | return 0; | |
1da177e4 LT |
1439 | } |
1440 | ||
a39bc516 | 1441 | static int copy_fs(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 | 1442 | { |
498052bb | 1443 | struct fs_struct *fs = current->fs; |
1da177e4 | 1444 | if (clone_flags & CLONE_FS) { |
498052bb | 1445 | /* tsk->fs is already what we want */ |
2a4419b5 | 1446 | spin_lock(&fs->lock); |
498052bb | 1447 | if (fs->in_exec) { |
2a4419b5 | 1448 | spin_unlock(&fs->lock); |
498052bb AV |
1449 | return -EAGAIN; |
1450 | } | |
1451 | fs->users++; | |
2a4419b5 | 1452 | spin_unlock(&fs->lock); |
1da177e4 LT |
1453 | return 0; |
1454 | } | |
498052bb | 1455 | tsk->fs = copy_fs_struct(fs); |
1da177e4 LT |
1456 | if (!tsk->fs) |
1457 | return -ENOMEM; | |
1458 | return 0; | |
1459 | } | |
1460 | ||
fb0a685c | 1461 | static int copy_files(unsigned long clone_flags, struct task_struct *tsk) |
a016f338 JD |
1462 | { |
1463 | struct files_struct *oldf, *newf; | |
1464 | int error = 0; | |
1465 | ||
1466 | /* | |
1467 | * A background process may not have any files ... | |
1468 | */ | |
1469 | oldf = current->files; | |
1470 | if (!oldf) | |
1471 | goto out; | |
1472 | ||
1473 | if (clone_flags & CLONE_FILES) { | |
1474 | atomic_inc(&oldf->count); | |
1475 | goto out; | |
1476 | } | |
1477 | ||
60997c3d | 1478 | newf = dup_fd(oldf, NR_OPEN_MAX, &error); |
a016f338 JD |
1479 | if (!newf) |
1480 | goto out; | |
1481 | ||
1482 | tsk->files = newf; | |
1483 | error = 0; | |
1484 | out: | |
1485 | return error; | |
1486 | } | |
1487 | ||
fadad878 | 1488 | static int copy_io(unsigned long clone_flags, struct task_struct *tsk) |
fd0928df JA |
1489 | { |
1490 | #ifdef CONFIG_BLOCK | |
1491 | struct io_context *ioc = current->io_context; | |
6e736be7 | 1492 | struct io_context *new_ioc; |
fd0928df JA |
1493 | |
1494 | if (!ioc) | |
1495 | return 0; | |
fadad878 JA |
1496 | /* |
1497 | * Share io context with parent, if CLONE_IO is set | |
1498 | */ | |
1499 | if (clone_flags & CLONE_IO) { | |
3d48749d TH |
1500 | ioc_task_link(ioc); |
1501 | tsk->io_context = ioc; | |
fadad878 | 1502 | } else if (ioprio_valid(ioc->ioprio)) { |
6e736be7 TH |
1503 | new_ioc = get_task_io_context(tsk, GFP_KERNEL, NUMA_NO_NODE); |
1504 | if (unlikely(!new_ioc)) | |
fd0928df JA |
1505 | return -ENOMEM; |
1506 | ||
6e736be7 | 1507 | new_ioc->ioprio = ioc->ioprio; |
11a3122f | 1508 | put_io_context(new_ioc); |
fd0928df JA |
1509 | } |
1510 | #endif | |
1511 | return 0; | |
1512 | } | |
1513 | ||
a39bc516 | 1514 | static int copy_sighand(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 LT |
1515 | { |
1516 | struct sighand_struct *sig; | |
1517 | ||
60348802 | 1518 | if (clone_flags & CLONE_SIGHAND) { |
d036bda7 | 1519 | refcount_inc(¤t->sighand->count); |
1da177e4 LT |
1520 | return 0; |
1521 | } | |
1522 | sig = kmem_cache_alloc(sighand_cachep, GFP_KERNEL); | |
0c282b06 | 1523 | RCU_INIT_POINTER(tsk->sighand, sig); |
1da177e4 LT |
1524 | if (!sig) |
1525 | return -ENOMEM; | |
9d7fb042 | 1526 | |
d036bda7 | 1527 | refcount_set(&sig->count, 1); |
06e62a46 | 1528 | spin_lock_irq(¤t->sighand->siglock); |
1da177e4 | 1529 | memcpy(sig->action, current->sighand->action, sizeof(sig->action)); |
06e62a46 | 1530 | spin_unlock_irq(¤t->sighand->siglock); |
b612e5df CB |
1531 | |
1532 | /* Reset all signal handler not set to SIG_IGN to SIG_DFL. */ | |
1533 | if (clone_flags & CLONE_CLEAR_SIGHAND) | |
1534 | flush_signal_handlers(tsk, 0); | |
1535 | ||
1da177e4 LT |
1536 | return 0; |
1537 | } | |
1538 | ||
a7e5328a | 1539 | void __cleanup_sighand(struct sighand_struct *sighand) |
c81addc9 | 1540 | { |
d036bda7 | 1541 | if (refcount_dec_and_test(&sighand->count)) { |
d80e731e | 1542 | signalfd_cleanup(sighand); |
392809b2 | 1543 | /* |
5f0d5a3a | 1544 | * sighand_cachep is SLAB_TYPESAFE_BY_RCU so we can free it |
392809b2 ON |
1545 | * without an RCU grace period, see __lock_task_sighand(). |
1546 | */ | |
c81addc9 | 1547 | kmem_cache_free(sighand_cachep, sighand); |
d80e731e | 1548 | } |
c81addc9 ON |
1549 | } |
1550 | ||
f06febc9 FM |
1551 | /* |
1552 | * Initialize POSIX timer handling for a thread group. | |
1553 | */ | |
1554 | static void posix_cpu_timers_init_group(struct signal_struct *sig) | |
1555 | { | |
2b69942f | 1556 | struct posix_cputimers *pct = &sig->posix_cputimers; |
78d7d407 JS |
1557 | unsigned long cpu_limit; |
1558 | ||
316c1608 | 1559 | cpu_limit = READ_ONCE(sig->rlim[RLIMIT_CPU].rlim_cur); |
3a245c0f | 1560 | posix_cputimers_group_init(pct, cpu_limit); |
f06febc9 FM |
1561 | } |
1562 | ||
a39bc516 | 1563 | static int copy_signal(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 LT |
1564 | { |
1565 | struct signal_struct *sig; | |
1da177e4 | 1566 | |
4ab6c083 | 1567 | if (clone_flags & CLONE_THREAD) |
490dea45 | 1568 | return 0; |
490dea45 | 1569 | |
a56704ef | 1570 | sig = kmem_cache_zalloc(signal_cachep, GFP_KERNEL); |
1da177e4 LT |
1571 | tsk->signal = sig; |
1572 | if (!sig) | |
1573 | return -ENOMEM; | |
1574 | ||
b3ac022c | 1575 | sig->nr_threads = 1; |
1da177e4 | 1576 | atomic_set(&sig->live, 1); |
60d4de3f | 1577 | refcount_set(&sig->sigcnt, 1); |
0c740d0a ON |
1578 | |
1579 | /* list_add(thread_node, thread_head) without INIT_LIST_HEAD() */ | |
1580 | sig->thread_head = (struct list_head)LIST_HEAD_INIT(tsk->thread_node); | |
1581 | tsk->thread_node = (struct list_head)LIST_HEAD_INIT(sig->thread_head); | |
1582 | ||
1da177e4 | 1583 | init_waitqueue_head(&sig->wait_chldexit); |
db51aecc | 1584 | sig->curr_target = tsk; |
1da177e4 | 1585 | init_sigpending(&sig->shared_pending); |
c3ad2c3b | 1586 | INIT_HLIST_HEAD(&sig->multiprocess); |
e78c3496 | 1587 | seqlock_init(&sig->stats_lock); |
9d7fb042 | 1588 | prev_cputime_init(&sig->prev_cputime); |
1da177e4 | 1589 | |
baa73d9e | 1590 | #ifdef CONFIG_POSIX_TIMERS |
b18b6a9c | 1591 | INIT_LIST_HEAD(&sig->posix_timers); |
c9cb2e3d | 1592 | hrtimer_init(&sig->real_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL); |
1da177e4 | 1593 | sig->real_timer.function = it_real_fn; |
baa73d9e | 1594 | #endif |
1da177e4 | 1595 | |
1da177e4 LT |
1596 | task_lock(current->group_leader); |
1597 | memcpy(sig->rlim, current->signal->rlim, sizeof sig->rlim); | |
1598 | task_unlock(current->group_leader); | |
1599 | ||
6279a751 ON |
1600 | posix_cpu_timers_init_group(sig); |
1601 | ||
522ed776 | 1602 | tty_audit_fork(sig); |
5091faa4 | 1603 | sched_autogroup_fork(sig); |
522ed776 | 1604 | |
a63d83f4 | 1605 | sig->oom_score_adj = current->signal->oom_score_adj; |
dabb16f6 | 1606 | sig->oom_score_adj_min = current->signal->oom_score_adj_min; |
28b83c51 | 1607 | |
9b1bf12d | 1608 | mutex_init(&sig->cred_guard_mutex); |
f7cfd871 | 1609 | init_rwsem(&sig->exec_update_lock); |
9b1bf12d | 1610 | |
1da177e4 LT |
1611 | return 0; |
1612 | } | |
1613 | ||
dbd95212 KC |
1614 | static void copy_seccomp(struct task_struct *p) |
1615 | { | |
1616 | #ifdef CONFIG_SECCOMP | |
1617 | /* | |
1618 | * Must be called with sighand->lock held, which is common to | |
1619 | * all threads in the group. Holding cred_guard_mutex is not | |
1620 | * needed because this new task is not yet running and cannot | |
1621 | * be racing exec. | |
1622 | */ | |
69f6a34b | 1623 | assert_spin_locked(¤t->sighand->siglock); |
dbd95212 KC |
1624 | |
1625 | /* Ref-count the new filter user, and assign it. */ | |
1626 | get_seccomp_filter(current); | |
1627 | p->seccomp = current->seccomp; | |
1628 | ||
1629 | /* | |
1630 | * Explicitly enable no_new_privs here in case it got set | |
1631 | * between the task_struct being duplicated and holding the | |
1632 | * sighand lock. The seccomp state and nnp must be in sync. | |
1633 | */ | |
1634 | if (task_no_new_privs(current)) | |
1635 | task_set_no_new_privs(p); | |
1636 | ||
1637 | /* | |
1638 | * If the parent gained a seccomp mode after copying thread | |
1639 | * flags and between before we held the sighand lock, we have | |
1640 | * to manually enable the seccomp thread flag here. | |
1641 | */ | |
1642 | if (p->seccomp.mode != SECCOMP_MODE_DISABLED) | |
23d67a54 | 1643 | set_task_syscall_work(p, SECCOMP); |
dbd95212 KC |
1644 | #endif |
1645 | } | |
1646 | ||
17da2bd9 | 1647 | SYSCALL_DEFINE1(set_tid_address, int __user *, tidptr) |
1da177e4 LT |
1648 | { |
1649 | current->clear_child_tid = tidptr; | |
1650 | ||
b488893a | 1651 | return task_pid_vnr(current); |
1da177e4 LT |
1652 | } |
1653 | ||
a39bc516 | 1654 | static void rt_mutex_init_task(struct task_struct *p) |
23f78d4a | 1655 | { |
1d615482 | 1656 | raw_spin_lock_init(&p->pi_lock); |
e29e175b | 1657 | #ifdef CONFIG_RT_MUTEXES |
a23ba907 | 1658 | p->pi_waiters = RB_ROOT_CACHED; |
e96a7705 | 1659 | p->pi_top_task = NULL; |
23f78d4a | 1660 | p->pi_blocked_on = NULL; |
23f78d4a IM |
1661 | #endif |
1662 | } | |
1663 | ||
2c470475 EB |
1664 | static inline void init_task_pid_links(struct task_struct *task) |
1665 | { | |
1666 | enum pid_type type; | |
1667 | ||
96e1e984 | 1668 | for (type = PIDTYPE_PID; type < PIDTYPE_MAX; ++type) |
2c470475 | 1669 | INIT_HLIST_NODE(&task->pid_links[type]); |
2c470475 EB |
1670 | } |
1671 | ||
81907739 ON |
1672 | static inline void |
1673 | init_task_pid(struct task_struct *task, enum pid_type type, struct pid *pid) | |
1674 | { | |
2c470475 EB |
1675 | if (type == PIDTYPE_PID) |
1676 | task->thread_pid = pid; | |
1677 | else | |
1678 | task->signal->pids[type] = pid; | |
81907739 ON |
1679 | } |
1680 | ||
6bfbaa51 IM |
1681 | static inline void rcu_copy_process(struct task_struct *p) |
1682 | { | |
1683 | #ifdef CONFIG_PREEMPT_RCU | |
1684 | p->rcu_read_lock_nesting = 0; | |
1685 | p->rcu_read_unlock_special.s = 0; | |
1686 | p->rcu_blocked_node = NULL; | |
1687 | INIT_LIST_HEAD(&p->rcu_node_entry); | |
1688 | #endif /* #ifdef CONFIG_PREEMPT_RCU */ | |
1689 | #ifdef CONFIG_TASKS_RCU | |
1690 | p->rcu_tasks_holdout = false; | |
1691 | INIT_LIST_HEAD(&p->rcu_tasks_holdout_list); | |
1692 | p->rcu_tasks_idle_cpu = -1; | |
1693 | #endif /* #ifdef CONFIG_TASKS_RCU */ | |
d5f177d3 PM |
1694 | #ifdef CONFIG_TASKS_TRACE_RCU |
1695 | p->trc_reader_nesting = 0; | |
276c4104 | 1696 | p->trc_reader_special.s = 0; |
d5f177d3 PM |
1697 | INIT_LIST_HEAD(&p->trc_holdout_list); |
1698 | #endif /* #ifdef CONFIG_TASKS_TRACE_RCU */ | |
6bfbaa51 IM |
1699 | } |
1700 | ||
3695eae5 CB |
1701 | struct pid *pidfd_pid(const struct file *file) |
1702 | { | |
1703 | if (file->f_op == &pidfd_fops) | |
1704 | return file->private_data; | |
1705 | ||
1706 | return ERR_PTR(-EBADF); | |
1707 | } | |
1708 | ||
b3e58382 CB |
1709 | static int pidfd_release(struct inode *inode, struct file *file) |
1710 | { | |
1711 | struct pid *pid = file->private_data; | |
1712 | ||
1713 | file->private_data = NULL; | |
1714 | put_pid(pid); | |
1715 | return 0; | |
1716 | } | |
1717 | ||
1718 | #ifdef CONFIG_PROC_FS | |
15d42eb2 CK |
1719 | /** |
1720 | * pidfd_show_fdinfo - print information about a pidfd | |
1721 | * @m: proc fdinfo file | |
1722 | * @f: file referencing a pidfd | |
1723 | * | |
1724 | * Pid: | |
1725 | * This function will print the pid that a given pidfd refers to in the | |
1726 | * pid namespace of the procfs instance. | |
1727 | * If the pid namespace of the process is not a descendant of the pid | |
1728 | * namespace of the procfs instance 0 will be shown as its pid. This is | |
1729 | * similar to calling getppid() on a process whose parent is outside of | |
1730 | * its pid namespace. | |
1731 | * | |
1732 | * NSpid: | |
1733 | * If pid namespaces are supported then this function will also print | |
1734 | * the pid of a given pidfd refers to for all descendant pid namespaces | |
1735 | * starting from the current pid namespace of the instance, i.e. the | |
1736 | * Pid field and the first entry in the NSpid field will be identical. | |
1737 | * If the pid namespace of the process is not a descendant of the pid | |
1738 | * namespace of the procfs instance 0 will be shown as its first NSpid | |
1739 | * entry and no others will be shown. | |
1740 | * Note that this differs from the Pid and NSpid fields in | |
1741 | * /proc/<pid>/status where Pid and NSpid are always shown relative to | |
1742 | * the pid namespace of the procfs instance. The difference becomes | |
1743 | * obvious when sending around a pidfd between pid namespaces from a | |
a8ca6b13 | 1744 | * different branch of the tree, i.e. where no ancestral relation is |
15d42eb2 CK |
1745 | * present between the pid namespaces: |
1746 | * - create two new pid namespaces ns1 and ns2 in the initial pid | |
1747 | * namespace (also take care to create new mount namespaces in the | |
1748 | * new pid namespace and mount procfs) | |
1749 | * - create a process with a pidfd in ns1 | |
1750 | * - send pidfd from ns1 to ns2 | |
1751 | * - read /proc/self/fdinfo/<pidfd> and observe that both Pid and NSpid | |
1752 | * have exactly one entry, which is 0 | |
1753 | */ | |
b3e58382 CB |
1754 | static void pidfd_show_fdinfo(struct seq_file *m, struct file *f) |
1755 | { | |
b3e58382 | 1756 | struct pid *pid = f->private_data; |
3d6d8da4 CB |
1757 | struct pid_namespace *ns; |
1758 | pid_t nr = -1; | |
15d42eb2 | 1759 | |
3d6d8da4 | 1760 | if (likely(pid_has_task(pid, PIDTYPE_PID))) { |
9d78edea | 1761 | ns = proc_pid_ns(file_inode(m->file)->i_sb); |
3d6d8da4 CB |
1762 | nr = pid_nr_ns(pid, ns); |
1763 | } | |
1764 | ||
1765 | seq_put_decimal_ll(m, "Pid:\t", nr); | |
b3e58382 | 1766 | |
15d42eb2 | 1767 | #ifdef CONFIG_PID_NS |
3d6d8da4 CB |
1768 | seq_put_decimal_ll(m, "\nNSpid:\t", nr); |
1769 | if (nr > 0) { | |
15d42eb2 | 1770 | int i; |
b3e58382 | 1771 | |
15d42eb2 CK |
1772 | /* If nr is non-zero it means that 'pid' is valid and that |
1773 | * ns, i.e. the pid namespace associated with the procfs | |
1774 | * instance, is in the pid namespace hierarchy of pid. | |
1775 | * Start at one below the already printed level. | |
1776 | */ | |
1777 | for (i = ns->level + 1; i <= pid->level; i++) | |
3d6d8da4 | 1778 | seq_put_decimal_ll(m, "\t", pid->numbers[i].nr); |
15d42eb2 CK |
1779 | } |
1780 | #endif | |
b3e58382 CB |
1781 | seq_putc(m, '\n'); |
1782 | } | |
1783 | #endif | |
1784 | ||
b53b0b9d JFG |
1785 | /* |
1786 | * Poll support for process exit notification. | |
1787 | */ | |
9e77716a | 1788 | static __poll_t pidfd_poll(struct file *file, struct poll_table_struct *pts) |
b53b0b9d | 1789 | { |
b53b0b9d | 1790 | struct pid *pid = file->private_data; |
9e77716a | 1791 | __poll_t poll_flags = 0; |
b53b0b9d JFG |
1792 | |
1793 | poll_wait(file, &pid->wait_pidfd, pts); | |
1794 | ||
b53b0b9d JFG |
1795 | /* |
1796 | * Inform pollers only when the whole thread group exits. | |
1797 | * If the thread group leader exits before all other threads in the | |
1798 | * group, then poll(2) should block, similar to the wait(2) family. | |
1799 | */ | |
38fd525a | 1800 | if (thread_group_exited(pid)) |
9e77716a | 1801 | poll_flags = EPOLLIN | EPOLLRDNORM; |
b53b0b9d JFG |
1802 | |
1803 | return poll_flags; | |
1804 | } | |
1805 | ||
b3e58382 CB |
1806 | const struct file_operations pidfd_fops = { |
1807 | .release = pidfd_release, | |
b53b0b9d | 1808 | .poll = pidfd_poll, |
b3e58382 CB |
1809 | #ifdef CONFIG_PROC_FS |
1810 | .show_fdinfo = pidfd_show_fdinfo, | |
1811 | #endif | |
1812 | }; | |
1813 | ||
c3f3ce04 AA |
1814 | static void __delayed_free_task(struct rcu_head *rhp) |
1815 | { | |
1816 | struct task_struct *tsk = container_of(rhp, struct task_struct, rcu); | |
1817 | ||
1818 | free_task(tsk); | |
1819 | } | |
1820 | ||
1821 | static __always_inline void delayed_free_task(struct task_struct *tsk) | |
1822 | { | |
1823 | if (IS_ENABLED(CONFIG_MEMCG)) | |
1824 | call_rcu(&tsk->rcu, __delayed_free_task); | |
1825 | else | |
1826 | free_task(tsk); | |
1827 | } | |
1828 | ||
67197a4f SB |
1829 | static void copy_oom_score_adj(u64 clone_flags, struct task_struct *tsk) |
1830 | { | |
1831 | /* Skip if kernel thread */ | |
1832 | if (!tsk->mm) | |
1833 | return; | |
1834 | ||
1835 | /* Skip if spawning a thread or using vfork */ | |
1836 | if ((clone_flags & (CLONE_VM | CLONE_THREAD | CLONE_VFORK)) != CLONE_VM) | |
1837 | return; | |
1838 | ||
1839 | /* We need to synchronize with __set_oom_adj */ | |
1840 | mutex_lock(&oom_adj_mutex); | |
1841 | set_bit(MMF_MULTIPROCESS, &tsk->mm->flags); | |
1842 | /* Update the values in case they were changed after copy_signal */ | |
1843 | tsk->signal->oom_score_adj = current->signal->oom_score_adj; | |
1844 | tsk->signal->oom_score_adj_min = current->signal->oom_score_adj_min; | |
1845 | mutex_unlock(&oom_adj_mutex); | |
1846 | } | |
1847 | ||
1da177e4 LT |
1848 | /* |
1849 | * This creates a new process as a copy of the old one, | |
1850 | * but does not actually start it yet. | |
1851 | * | |
1852 | * It copies the registers, and all the appropriate | |
1853 | * parts of the process environment (as per the clone | |
1854 | * flags). The actual kick-off is left to the caller. | |
1855 | */ | |
0766f788 | 1856 | static __latent_entropy struct task_struct *copy_process( |
09a05394 | 1857 | struct pid *pid, |
3033f14a | 1858 | int trace, |
7f192e3c CB |
1859 | int node, |
1860 | struct kernel_clone_args *args) | |
1da177e4 | 1861 | { |
b3e58382 | 1862 | int pidfd = -1, retval; |
a24efe62 | 1863 | struct task_struct *p; |
c3ad2c3b | 1864 | struct multiprocess_signals delayed; |
6fd2fe49 | 1865 | struct file *pidfile = NULL; |
7f192e3c | 1866 | u64 clone_flags = args->flags; |
769071ac | 1867 | struct nsproxy *nsp = current->nsproxy; |
1da177e4 | 1868 | |
667b6094 MPS |
1869 | /* |
1870 | * Don't allow sharing the root directory with processes in a different | |
1871 | * namespace | |
1872 | */ | |
1da177e4 LT |
1873 | if ((clone_flags & (CLONE_NEWNS|CLONE_FS)) == (CLONE_NEWNS|CLONE_FS)) |
1874 | return ERR_PTR(-EINVAL); | |
1875 | ||
e66eded8 EB |
1876 | if ((clone_flags & (CLONE_NEWUSER|CLONE_FS)) == (CLONE_NEWUSER|CLONE_FS)) |
1877 | return ERR_PTR(-EINVAL); | |
1878 | ||
1da177e4 LT |
1879 | /* |
1880 | * Thread groups must share signals as well, and detached threads | |
1881 | * can only be started up within the thread group. | |
1882 | */ | |
1883 | if ((clone_flags & CLONE_THREAD) && !(clone_flags & CLONE_SIGHAND)) | |
1884 | return ERR_PTR(-EINVAL); | |
1885 | ||
1886 | /* | |
1887 | * Shared signal handlers imply shared VM. By way of the above, | |
1888 | * thread groups also imply shared VM. Blocking this case allows | |
1889 | * for various simplifications in other code. | |
1890 | */ | |
1891 | if ((clone_flags & CLONE_SIGHAND) && !(clone_flags & CLONE_VM)) | |
1892 | return ERR_PTR(-EINVAL); | |
1893 | ||
123be07b SB |
1894 | /* |
1895 | * Siblings of global init remain as zombies on exit since they are | |
1896 | * not reaped by their parent (swapper). To solve this and to avoid | |
1897 | * multi-rooted process trees, prevent global and container-inits | |
1898 | * from creating siblings. | |
1899 | */ | |
1900 | if ((clone_flags & CLONE_PARENT) && | |
1901 | current->signal->flags & SIGNAL_UNKILLABLE) | |
1902 | return ERR_PTR(-EINVAL); | |
1903 | ||
8382fcac | 1904 | /* |
40a0d32d | 1905 | * If the new process will be in a different pid or user namespace |
faf00da5 | 1906 | * do not allow it to share a thread group with the forking task. |
8382fcac | 1907 | */ |
faf00da5 | 1908 | if (clone_flags & CLONE_THREAD) { |
40a0d32d | 1909 | if ((clone_flags & (CLONE_NEWUSER | CLONE_NEWPID)) || |
769071ac AV |
1910 | (task_active_pid_ns(current) != nsp->pid_ns_for_children)) |
1911 | return ERR_PTR(-EINVAL); | |
1912 | } | |
1913 | ||
1914 | /* | |
1915 | * If the new process will be in a different time namespace | |
1916 | * do not allow it to share VM or a thread group with the forking task. | |
1917 | */ | |
1918 | if (clone_flags & (CLONE_THREAD | CLONE_VM)) { | |
1919 | if (nsp->time_ns != nsp->time_ns_for_children) | |
40a0d32d ON |
1920 | return ERR_PTR(-EINVAL); |
1921 | } | |
8382fcac | 1922 | |
b3e58382 | 1923 | if (clone_flags & CLONE_PIDFD) { |
b3e58382 | 1924 | /* |
b3e58382 CB |
1925 | * - CLONE_DETACHED is blocked so that we can potentially |
1926 | * reuse it later for CLONE_PIDFD. | |
1927 | * - CLONE_THREAD is blocked until someone really needs it. | |
1928 | */ | |
7f192e3c | 1929 | if (clone_flags & (CLONE_DETACHED | CLONE_THREAD)) |
b3e58382 | 1930 | return ERR_PTR(-EINVAL); |
b3e58382 CB |
1931 | } |
1932 | ||
c3ad2c3b EB |
1933 | /* |
1934 | * Force any signals received before this point to be delivered | |
1935 | * before the fork happens. Collect up signals sent to multiple | |
1936 | * processes that happen during the fork and delay them so that | |
1937 | * they appear to happen after the fork. | |
1938 | */ | |
1939 | sigemptyset(&delayed.signal); | |
1940 | INIT_HLIST_NODE(&delayed.node); | |
1941 | ||
1942 | spin_lock_irq(¤t->sighand->siglock); | |
1943 | if (!(clone_flags & CLONE_THREAD)) | |
1944 | hlist_add_head(&delayed.node, ¤t->signal->multiprocess); | |
1945 | recalc_sigpending(); | |
1946 | spin_unlock_irq(¤t->sighand->siglock); | |
1947 | retval = -ERESTARTNOINTR; | |
66ae0d1e | 1948 | if (task_sigpending(current)) |
c3ad2c3b EB |
1949 | goto fork_out; |
1950 | ||
1da177e4 | 1951 | retval = -ENOMEM; |
725fc629 | 1952 | p = dup_task_struct(current, node); |
1da177e4 LT |
1953 | if (!p) |
1954 | goto fork_out; | |
b16b3855 JA |
1955 | if (args->io_thread) { |
1956 | /* | |
1957 | * Mark us an IO worker, and block any signal that isn't | |
1958 | * fatal or STOP | |
1959 | */ | |
cc440e87 | 1960 | p->flags |= PF_IO_WORKER; |
b16b3855 JA |
1961 | siginitsetinv(&p->blocked, sigmask(SIGKILL)|sigmask(SIGSTOP)); |
1962 | } | |
1da177e4 | 1963 | |
4d6501dc VN |
1964 | /* |
1965 | * This _must_ happen before we call free_task(), i.e. before we jump | |
1966 | * to any of the bad_fork_* labels. This is to avoid freeing | |
1967 | * p->set_child_tid which is (ab)used as a kthread's data pointer for | |
1968 | * kernel threads (PF_KTHREAD). | |
1969 | */ | |
7f192e3c | 1970 | p->set_child_tid = (clone_flags & CLONE_CHILD_SETTID) ? args->child_tid : NULL; |
4d6501dc VN |
1971 | /* |
1972 | * Clear TID on mm_release()? | |
1973 | */ | |
7f192e3c | 1974 | p->clear_child_tid = (clone_flags & CLONE_CHILD_CLEARTID) ? args->child_tid : NULL; |
4d6501dc | 1975 | |
f7e8b616 SR |
1976 | ftrace_graph_init_task(p); |
1977 | ||
bea493a0 PZ |
1978 | rt_mutex_init_task(p); |
1979 | ||
a21ee605 | 1980 | lockdep_assert_irqs_enabled(); |
d12c1a37 | 1981 | #ifdef CONFIG_PROVE_LOCKING |
de30a2b3 IM |
1982 | DEBUG_LOCKS_WARN_ON(!p->softirqs_enabled); |
1983 | #endif | |
1da177e4 | 1984 | retval = -EAGAIN; |
21d1c5e3 | 1985 | if (is_ucounts_overlimit(task_ucounts(p), UCOUNT_RLIMIT_NPROC, rlimit(RLIMIT_NPROC))) { |
b57922b6 EP |
1986 | if (p->real_cred->user != INIT_USER && |
1987 | !capable(CAP_SYS_RESOURCE) && !capable(CAP_SYS_ADMIN)) | |
1da177e4 LT |
1988 | goto bad_fork_free; |
1989 | } | |
72fa5997 | 1990 | current->flags &= ~PF_NPROC_EXCEEDED; |
1da177e4 | 1991 | |
f1752eec DH |
1992 | retval = copy_creds(p, clone_flags); |
1993 | if (retval < 0) | |
1994 | goto bad_fork_free; | |
1da177e4 LT |
1995 | |
1996 | /* | |
1997 | * If multiple threads are within copy_process(), then this check | |
1998 | * triggers too late. This doesn't hurt, the check is only there | |
1999 | * to stop root fork bombs. | |
2000 | */ | |
04ec93fe | 2001 | retval = -EAGAIN; |
c17d1a3a | 2002 | if (data_race(nr_threads >= max_threads)) |
1da177e4 LT |
2003 | goto bad_fork_cleanup_count; |
2004 | ||
ca74e92b | 2005 | delayacct_tsk_init(p); /* Must remain after dup_task_struct() */ |
a8ea6fc9 | 2006 | p->flags &= ~(PF_SUPERPRIV | PF_WQ_WORKER | PF_IDLE | PF_NO_SETAFFINITY); |
514ddb44 | 2007 | p->flags |= PF_FORKNOEXEC; |
1da177e4 LT |
2008 | INIT_LIST_HEAD(&p->children); |
2009 | INIT_LIST_HEAD(&p->sibling); | |
f41d911f | 2010 | rcu_copy_process(p); |
1da177e4 LT |
2011 | p->vfork_done = NULL; |
2012 | spin_lock_init(&p->alloc_lock); | |
1da177e4 | 2013 | |
1da177e4 LT |
2014 | init_sigpending(&p->pending); |
2015 | ||
64861634 | 2016 | p->utime = p->stime = p->gtime = 0; |
40565b5a | 2017 | #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME |
64861634 | 2018 | p->utimescaled = p->stimescaled = 0; |
40565b5a | 2019 | #endif |
9d7fb042 PZ |
2020 | prev_cputime_init(&p->prev_cputime); |
2021 | ||
6a61671b | 2022 | #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN |
bac5b6b6 FW |
2023 | seqcount_init(&p->vtime.seqcount); |
2024 | p->vtime.starttime = 0; | |
2025 | p->vtime.state = VTIME_INACTIVE; | |
6a61671b FW |
2026 | #endif |
2027 | ||
0f212204 JA |
2028 | #ifdef CONFIG_IO_URING |
2029 | p->io_uring = NULL; | |
2030 | #endif | |
2031 | ||
a3a2e76c KH |
2032 | #if defined(SPLIT_RSS_COUNTING) |
2033 | memset(&p->rss_stat, 0, sizeof(p->rss_stat)); | |
2034 | #endif | |
172ba844 | 2035 | |
6976675d AV |
2036 | p->default_timer_slack_ns = current->timer_slack_ns; |
2037 | ||
eb414681 JW |
2038 | #ifdef CONFIG_PSI |
2039 | p->psi_flags = 0; | |
2040 | #endif | |
2041 | ||
5995477a | 2042 | task_io_accounting_init(&p->ioac); |
1da177e4 LT |
2043 | acct_clear_integrals(p); |
2044 | ||
3a245c0f | 2045 | posix_cputimers_init(&p->posix_cputimers); |
1da177e4 | 2046 | |
1da177e4 | 2047 | p->io_context = NULL; |
c0b0ae8a | 2048 | audit_set_context(p, NULL); |
b4f48b63 | 2049 | cgroup_fork(p); |
1da177e4 | 2050 | #ifdef CONFIG_NUMA |
846a16bf | 2051 | p->mempolicy = mpol_dup(p->mempolicy); |
fb0a685c DRO |
2052 | if (IS_ERR(p->mempolicy)) { |
2053 | retval = PTR_ERR(p->mempolicy); | |
2054 | p->mempolicy = NULL; | |
e8604cb4 | 2055 | goto bad_fork_cleanup_threadgroup_lock; |
fb0a685c | 2056 | } |
1da177e4 | 2057 | #endif |
778d3b0f MH |
2058 | #ifdef CONFIG_CPUSETS |
2059 | p->cpuset_mem_spread_rotor = NUMA_NO_NODE; | |
2060 | p->cpuset_slab_spread_rotor = NUMA_NO_NODE; | |
b7505861 | 2061 | seqcount_spinlock_init(&p->mems_allowed_seq, &p->alloc_lock); |
778d3b0f | 2062 | #endif |
de30a2b3 | 2063 | #ifdef CONFIG_TRACE_IRQFLAGS |
0584df9c ME |
2064 | memset(&p->irqtrace, 0, sizeof(p->irqtrace)); |
2065 | p->irqtrace.hardirq_disable_ip = _THIS_IP_; | |
2066 | p->irqtrace.softirq_enable_ip = _THIS_IP_; | |
2067 | p->softirqs_enabled = 1; | |
2068 | p->softirq_context = 0; | |
de30a2b3 | 2069 | #endif |
8bcbde54 DH |
2070 | |
2071 | p->pagefault_disabled = 0; | |
2072 | ||
fbb9ce95 | 2073 | #ifdef CONFIG_LOCKDEP |
b09be676 | 2074 | lockdep_init_task(p); |
fbb9ce95 | 2075 | #endif |
1da177e4 | 2076 | |
408894ee IM |
2077 | #ifdef CONFIG_DEBUG_MUTEXES |
2078 | p->blocked_on = NULL; /* not blocked yet */ | |
2079 | #endif | |
cafe5635 KO |
2080 | #ifdef CONFIG_BCACHE |
2081 | p->sequential_io = 0; | |
2082 | p->sequential_io_avg = 0; | |
2083 | #endif | |
a10787e6 SL |
2084 | #ifdef CONFIG_BPF_SYSCALL |
2085 | RCU_INIT_POINTER(p->bpf_storage, NULL); | |
2086 | #endif | |
0f481406 | 2087 | |
3c90e6e9 | 2088 | /* Perform scheduler related setup. Assign this task to a CPU. */ |
aab03e05 DF |
2089 | retval = sched_fork(clone_flags, p); |
2090 | if (retval) | |
2091 | goto bad_fork_cleanup_policy; | |
6ab423e0 | 2092 | |
2b26f0aa | 2093 | retval = perf_event_init_task(p, clone_flags); |
6ab423e0 PZ |
2094 | if (retval) |
2095 | goto bad_fork_cleanup_policy; | |
fb0a685c DRO |
2096 | retval = audit_alloc(p); |
2097 | if (retval) | |
6c72e350 | 2098 | goto bad_fork_cleanup_perf; |
1da177e4 | 2099 | /* copy all the process information */ |
ab602f79 | 2100 | shm_init_task(p); |
e4e55b47 | 2101 | retval = security_task_alloc(p, clone_flags); |
fb0a685c | 2102 | if (retval) |
1da177e4 | 2103 | goto bad_fork_cleanup_audit; |
e4e55b47 TH |
2104 | retval = copy_semundo(clone_flags, p); |
2105 | if (retval) | |
2106 | goto bad_fork_cleanup_security; | |
fb0a685c DRO |
2107 | retval = copy_files(clone_flags, p); |
2108 | if (retval) | |
1da177e4 | 2109 | goto bad_fork_cleanup_semundo; |
fb0a685c DRO |
2110 | retval = copy_fs(clone_flags, p); |
2111 | if (retval) | |
1da177e4 | 2112 | goto bad_fork_cleanup_files; |
fb0a685c DRO |
2113 | retval = copy_sighand(clone_flags, p); |
2114 | if (retval) | |
1da177e4 | 2115 | goto bad_fork_cleanup_fs; |
fb0a685c DRO |
2116 | retval = copy_signal(clone_flags, p); |
2117 | if (retval) | |
1da177e4 | 2118 | goto bad_fork_cleanup_sighand; |
fb0a685c DRO |
2119 | retval = copy_mm(clone_flags, p); |
2120 | if (retval) | |
1da177e4 | 2121 | goto bad_fork_cleanup_signal; |
fb0a685c DRO |
2122 | retval = copy_namespaces(clone_flags, p); |
2123 | if (retval) | |
d84f4f99 | 2124 | goto bad_fork_cleanup_mm; |
fb0a685c DRO |
2125 | retval = copy_io(clone_flags, p); |
2126 | if (retval) | |
fd0928df | 2127 | goto bad_fork_cleanup_namespaces; |
714acdbd | 2128 | retval = copy_thread(clone_flags, args->stack, args->stack_size, p, args->tls); |
1da177e4 | 2129 | if (retval) |
fd0928df | 2130 | goto bad_fork_cleanup_io; |
1da177e4 | 2131 | |
afaef01c AP |
2132 | stackleak_task_init(p); |
2133 | ||
425fb2b4 | 2134 | if (pid != &init_struct_pid) { |
49cb2fc4 AR |
2135 | pid = alloc_pid(p->nsproxy->pid_ns_for_children, args->set_tid, |
2136 | args->set_tid_size); | |
35f71bc0 MH |
2137 | if (IS_ERR(pid)) { |
2138 | retval = PTR_ERR(pid); | |
0740aa5f | 2139 | goto bad_fork_cleanup_thread; |
35f71bc0 | 2140 | } |
425fb2b4 PE |
2141 | } |
2142 | ||
b3e58382 CB |
2143 | /* |
2144 | * This has to happen after we've potentially unshared the file | |
2145 | * descriptor table (so that the pidfd doesn't leak into the child | |
2146 | * if the fd table isn't shared). | |
2147 | */ | |
2148 | if (clone_flags & CLONE_PIDFD) { | |
6fd2fe49 | 2149 | retval = get_unused_fd_flags(O_RDWR | O_CLOEXEC); |
b3e58382 CB |
2150 | if (retval < 0) |
2151 | goto bad_fork_free_pid; | |
2152 | ||
2153 | pidfd = retval; | |
6fd2fe49 AV |
2154 | |
2155 | pidfile = anon_inode_getfile("[pidfd]", &pidfd_fops, pid, | |
2156 | O_RDWR | O_CLOEXEC); | |
2157 | if (IS_ERR(pidfile)) { | |
2158 | put_unused_fd(pidfd); | |
28dd29c0 | 2159 | retval = PTR_ERR(pidfile); |
6fd2fe49 AV |
2160 | goto bad_fork_free_pid; |
2161 | } | |
2162 | get_pid(pid); /* held by pidfile now */ | |
2163 | ||
7f192e3c | 2164 | retval = put_user(pidfd, args->pidfd); |
b3e58382 CB |
2165 | if (retval) |
2166 | goto bad_fork_put_pidfd; | |
2167 | } | |
2168 | ||
73c10101 JA |
2169 | #ifdef CONFIG_BLOCK |
2170 | p->plug = NULL; | |
2171 | #endif | |
ba31c1a4 TG |
2172 | futex_init_task(p); |
2173 | ||
f9a3879a GM |
2174 | /* |
2175 | * sigaltstack should be cleared when sharing the same VM | |
2176 | */ | |
2177 | if ((clone_flags & (CLONE_VM|CLONE_VFORK)) == CLONE_VM) | |
2a742138 | 2178 | sas_ss_reset(p); |
f9a3879a | 2179 | |
1da177e4 | 2180 | /* |
6580807d ON |
2181 | * Syscall tracing and stepping should be turned off in the |
2182 | * child regardless of CLONE_PTRACE. | |
1da177e4 | 2183 | */ |
6580807d | 2184 | user_disable_single_step(p); |
64c19ba2 | 2185 | clear_task_syscall_work(p, SYSCALL_TRACE); |
64eb35f7 GKB |
2186 | #if defined(CONFIG_GENERIC_ENTRY) || defined(TIF_SYSCALL_EMU) |
2187 | clear_task_syscall_work(p, SYSCALL_EMU); | |
ed75e8d5 | 2188 | #endif |
e02c9b0d | 2189 | clear_tsk_latency_tracing(p); |
1da177e4 | 2190 | |
1da177e4 | 2191 | /* ok, now we should be set up.. */ |
18c830df ON |
2192 | p->pid = pid_nr(pid); |
2193 | if (clone_flags & CLONE_THREAD) { | |
18c830df ON |
2194 | p->group_leader = current->group_leader; |
2195 | p->tgid = current->tgid; | |
2196 | } else { | |
18c830df ON |
2197 | p->group_leader = p; |
2198 | p->tgid = p->pid; | |
2199 | } | |
5f8aadd8 | 2200 | |
9d823e8f WF |
2201 | p->nr_dirtied = 0; |
2202 | p->nr_dirtied_pause = 128 >> (PAGE_SHIFT - 10); | |
83712358 | 2203 | p->dirty_paused_when = 0; |
9d823e8f | 2204 | |
bb8cbbfe | 2205 | p->pdeath_signal = 0; |
47e65328 | 2206 | INIT_LIST_HEAD(&p->thread_group); |
158e1645 | 2207 | p->task_works = NULL; |
1da177e4 | 2208 | |
d741bf41 PZ |
2209 | #ifdef CONFIG_KRETPROBES |
2210 | p->kretprobe_instances.first = NULL; | |
2211 | #endif | |
2212 | ||
7e47682e AS |
2213 | /* |
2214 | * Ensure that the cgroup subsystem policies allow the new process to be | |
7b7b8a2c | 2215 | * forked. It should be noted that the new process's css_set can be changed |
7e47682e AS |
2216 | * between here and cgroup_post_fork() if an organisation operation is in |
2217 | * progress. | |
2218 | */ | |
ef2c41cf | 2219 | retval = cgroup_can_fork(p, args); |
7e47682e | 2220 | if (retval) |
5a5cf5cb | 2221 | goto bad_fork_put_pidfd; |
7e47682e | 2222 | |
7b558513 DH |
2223 | /* |
2224 | * From this point on we must avoid any synchronous user-space | |
2225 | * communication until we take the tasklist-lock. In particular, we do | |
2226 | * not want user-space to be able to predict the process start-time by | |
2227 | * stalling fork(2) after we recorded the start_time but before it is | |
2228 | * visible to the system. | |
2229 | */ | |
2230 | ||
2231 | p->start_time = ktime_get_ns(); | |
cf25e24d | 2232 | p->start_boottime = ktime_get_boottime_ns(); |
7b558513 | 2233 | |
18c830df ON |
2234 | /* |
2235 | * Make it visible to the rest of the system, but dont wake it up yet. | |
2236 | * Need tasklist lock for parent etc handling! | |
2237 | */ | |
1da177e4 LT |
2238 | write_lock_irq(&tasklist_lock); |
2239 | ||
1da177e4 | 2240 | /* CLONE_PARENT re-uses the old parent */ |
2d5516cb | 2241 | if (clone_flags & (CLONE_PARENT|CLONE_THREAD)) { |
1da177e4 | 2242 | p->real_parent = current->real_parent; |
2d5516cb | 2243 | p->parent_exec_id = current->parent_exec_id; |
b4e00444 EW |
2244 | if (clone_flags & CLONE_THREAD) |
2245 | p->exit_signal = -1; | |
2246 | else | |
2247 | p->exit_signal = current->group_leader->exit_signal; | |
2d5516cb | 2248 | } else { |
1da177e4 | 2249 | p->real_parent = current; |
2d5516cb | 2250 | p->parent_exec_id = current->self_exec_id; |
b4e00444 | 2251 | p->exit_signal = args->exit_signal; |
2d5516cb | 2252 | } |
1da177e4 | 2253 | |
d83a7cb3 JP |
2254 | klp_copy_process(p); |
2255 | ||
85dd3f61 PZ |
2256 | sched_core_fork(p); |
2257 | ||
3f17da69 | 2258 | spin_lock(¤t->sighand->siglock); |
4a2c7a78 | 2259 | |
dbd95212 KC |
2260 | /* |
2261 | * Copy seccomp details explicitly here, in case they were changed | |
2262 | * before holding sighand lock. | |
2263 | */ | |
2264 | copy_seccomp(p); | |
2265 | ||
d7822b1e MD |
2266 | rseq_fork(p, clone_flags); |
2267 | ||
4ca1d3ee | 2268 | /* Don't start children in a dying pid namespace */ |
e8cfbc24 | 2269 | if (unlikely(!(ns_of_pid(pid)->pid_allocated & PIDNS_ADDING))) { |
3fd37226 KT |
2270 | retval = -ENOMEM; |
2271 | goto bad_fork_cancel_cgroup; | |
2272 | } | |
4a2c7a78 | 2273 | |
7673bf55 EB |
2274 | /* Let kill terminate clone/fork in the middle */ |
2275 | if (fatal_signal_pending(current)) { | |
2276 | retval = -EINTR; | |
2277 | goto bad_fork_cancel_cgroup; | |
2278 | } | |
2279 | ||
6fd2fe49 AV |
2280 | /* past the last point of failure */ |
2281 | if (pidfile) | |
2282 | fd_install(pidfd, pidfile); | |
4a2c7a78 | 2283 | |
2c470475 | 2284 | init_task_pid_links(p); |
73b9ebfe | 2285 | if (likely(p->pid)) { |
4b9d33e6 | 2286 | ptrace_init_task(p, (clone_flags & CLONE_PTRACE) || trace); |
73b9ebfe | 2287 | |
81907739 | 2288 | init_task_pid(p, PIDTYPE_PID, pid); |
73b9ebfe | 2289 | if (thread_group_leader(p)) { |
6883f81a | 2290 | init_task_pid(p, PIDTYPE_TGID, pid); |
81907739 ON |
2291 | init_task_pid(p, PIDTYPE_PGID, task_pgrp(current)); |
2292 | init_task_pid(p, PIDTYPE_SID, task_session(current)); | |
2293 | ||
1c4042c2 | 2294 | if (is_child_reaper(pid)) { |
17cf22c3 | 2295 | ns_of_pid(pid)->child_reaper = p; |
1c4042c2 EB |
2296 | p->signal->flags |= SIGNAL_UNKILLABLE; |
2297 | } | |
c3ad2c3b | 2298 | p->signal->shared_pending.signal = delayed.signal; |
9c9f4ded | 2299 | p->signal->tty = tty_kref_get(current->signal->tty); |
749860ce PT |
2300 | /* |
2301 | * Inherit has_child_subreaper flag under the same | |
2302 | * tasklist_lock with adding child to the process tree | |
2303 | * for propagate_has_child_subreaper optimization. | |
2304 | */ | |
2305 | p->signal->has_child_subreaper = p->real_parent->signal->has_child_subreaper || | |
2306 | p->real_parent->signal->is_child_subreaper; | |
9cd80bbb | 2307 | list_add_tail(&p->sibling, &p->real_parent->children); |
5e85d4ab | 2308 | list_add_tail_rcu(&p->tasks, &init_task.tasks); |
6883f81a | 2309 | attach_pid(p, PIDTYPE_TGID); |
81907739 ON |
2310 | attach_pid(p, PIDTYPE_PGID); |
2311 | attach_pid(p, PIDTYPE_SID); | |
909ea964 | 2312 | __this_cpu_inc(process_counts); |
80628ca0 ON |
2313 | } else { |
2314 | current->signal->nr_threads++; | |
2315 | atomic_inc(¤t->signal->live); | |
60d4de3f | 2316 | refcount_inc(¤t->signal->sigcnt); |
924de3b8 | 2317 | task_join_group_stop(p); |
80628ca0 ON |
2318 | list_add_tail_rcu(&p->thread_group, |
2319 | &p->group_leader->thread_group); | |
0c740d0a ON |
2320 | list_add_tail_rcu(&p->thread_node, |
2321 | &p->signal->thread_head); | |
73b9ebfe | 2322 | } |
81907739 | 2323 | attach_pid(p, PIDTYPE_PID); |
73b9ebfe | 2324 | nr_threads++; |
1da177e4 | 2325 | } |
1da177e4 | 2326 | total_forks++; |
c3ad2c3b | 2327 | hlist_del_init(&delayed.node); |
3f17da69 | 2328 | spin_unlock(¤t->sighand->siglock); |
4af4206b | 2329 | syscall_tracepoint_update(p); |
1da177e4 | 2330 | write_unlock_irq(&tasklist_lock); |
4af4206b | 2331 | |
c13cf856 | 2332 | proc_fork_connector(p); |
13685c4a | 2333 | sched_post_fork(p); |
ef2c41cf | 2334 | cgroup_post_fork(p, args); |
cdd6c482 | 2335 | perf_event_fork(p); |
43d2b113 KH |
2336 | |
2337 | trace_task_newtask(p, clone_flags); | |
3ab67966 | 2338 | uprobe_copy_process(p, clone_flags); |
43d2b113 | 2339 | |
67197a4f SB |
2340 | copy_oom_score_adj(clone_flags, p); |
2341 | ||
1da177e4 LT |
2342 | return p; |
2343 | ||
7e47682e | 2344 | bad_fork_cancel_cgroup: |
85dd3f61 | 2345 | sched_core_free(p); |
3fd37226 KT |
2346 | spin_unlock(¤t->sighand->siglock); |
2347 | write_unlock_irq(&tasklist_lock); | |
ef2c41cf | 2348 | cgroup_cancel_fork(p, args); |
b3e58382 | 2349 | bad_fork_put_pidfd: |
6fd2fe49 AV |
2350 | if (clone_flags & CLONE_PIDFD) { |
2351 | fput(pidfile); | |
2352 | put_unused_fd(pidfd); | |
2353 | } | |
425fb2b4 PE |
2354 | bad_fork_free_pid: |
2355 | if (pid != &init_struct_pid) | |
2356 | free_pid(pid); | |
0740aa5f JS |
2357 | bad_fork_cleanup_thread: |
2358 | exit_thread(p); | |
fd0928df | 2359 | bad_fork_cleanup_io: |
b69f2292 LR |
2360 | if (p->io_context) |
2361 | exit_io_context(p); | |
ab516013 | 2362 | bad_fork_cleanup_namespaces: |
444f378b | 2363 | exit_task_namespaces(p); |
1da177e4 | 2364 | bad_fork_cleanup_mm: |
c3f3ce04 AA |
2365 | if (p->mm) { |
2366 | mm_clear_owner(p->mm, p); | |
1da177e4 | 2367 | mmput(p->mm); |
c3f3ce04 | 2368 | } |
1da177e4 | 2369 | bad_fork_cleanup_signal: |
4ab6c083 | 2370 | if (!(clone_flags & CLONE_THREAD)) |
1c5354de | 2371 | free_signal_struct(p->signal); |
1da177e4 | 2372 | bad_fork_cleanup_sighand: |
a7e5328a | 2373 | __cleanup_sighand(p->sighand); |
1da177e4 LT |
2374 | bad_fork_cleanup_fs: |
2375 | exit_fs(p); /* blocking */ | |
2376 | bad_fork_cleanup_files: | |
2377 | exit_files(p); /* blocking */ | |
2378 | bad_fork_cleanup_semundo: | |
2379 | exit_sem(p); | |
e4e55b47 TH |
2380 | bad_fork_cleanup_security: |
2381 | security_task_free(p); | |
1da177e4 LT |
2382 | bad_fork_cleanup_audit: |
2383 | audit_free(p); | |
6c72e350 | 2384 | bad_fork_cleanup_perf: |
cdd6c482 | 2385 | perf_event_free_task(p); |
6c72e350 | 2386 | bad_fork_cleanup_policy: |
b09be676 | 2387 | lockdep_free_task(p); |
1da177e4 | 2388 | #ifdef CONFIG_NUMA |
f0be3d32 | 2389 | mpol_put(p->mempolicy); |
e8604cb4 | 2390 | bad_fork_cleanup_threadgroup_lock: |
1da177e4 | 2391 | #endif |
35df17c5 | 2392 | delayacct_tsk_free(p); |
1da177e4 | 2393 | bad_fork_cleanup_count: |
21d1c5e3 | 2394 | dec_rlimit_ucounts(task_ucounts(p), UCOUNT_RLIMIT_NPROC, 1); |
e0e81739 | 2395 | exit_creds(p); |
1da177e4 | 2396 | bad_fork_free: |
2f064a59 | 2397 | WRITE_ONCE(p->__state, TASK_DEAD); |
68f24b08 | 2398 | put_task_stack(p); |
c3f3ce04 | 2399 | delayed_free_task(p); |
fe7d37d1 | 2400 | fork_out: |
c3ad2c3b EB |
2401 | spin_lock_irq(¤t->sighand->siglock); |
2402 | hlist_del_init(&delayed.node); | |
2403 | spin_unlock_irq(¤t->sighand->siglock); | |
fe7d37d1 | 2404 | return ERR_PTR(retval); |
1da177e4 LT |
2405 | } |
2406 | ||
2c470475 | 2407 | static inline void init_idle_pids(struct task_struct *idle) |
f106eee1 ON |
2408 | { |
2409 | enum pid_type type; | |
2410 | ||
2411 | for (type = PIDTYPE_PID; type < PIDTYPE_MAX; ++type) { | |
2c470475 EB |
2412 | INIT_HLIST_NODE(&idle->pid_links[type]); /* not really needed */ |
2413 | init_task_pid(idle, type, &init_struct_pid); | |
f106eee1 ON |
2414 | } |
2415 | } | |
2416 | ||
f1a0a376 | 2417 | struct task_struct * __init fork_idle(int cpu) |
1da177e4 | 2418 | { |
36c8b586 | 2419 | struct task_struct *task; |
7f192e3c CB |
2420 | struct kernel_clone_args args = { |
2421 | .flags = CLONE_VM, | |
2422 | }; | |
2423 | ||
2424 | task = copy_process(&init_struct_pid, 0, cpu_to_node(cpu), &args); | |
f106eee1 | 2425 | if (!IS_ERR(task)) { |
2c470475 | 2426 | init_idle_pids(task); |
753ca4f3 | 2427 | init_idle(task, cpu); |
f106eee1 | 2428 | } |
73b9ebfe | 2429 | |
1da177e4 LT |
2430 | return task; |
2431 | } | |
2432 | ||
13585fa0 NA |
2433 | struct mm_struct *copy_init_mm(void) |
2434 | { | |
2435 | return dup_mm(NULL, &init_mm); | |
2436 | } | |
2437 | ||
cc440e87 JA |
2438 | /* |
2439 | * This is like kernel_clone(), but shaved down and tailored to just | |
2440 | * creating io_uring workers. It returns a created task, or an error pointer. | |
2441 | * The returned task is inactive, and the caller must fire it up through | |
2442 | * wake_up_new_task(p). All signals are blocked in the created task. | |
2443 | */ | |
2444 | struct task_struct *create_io_thread(int (*fn)(void *), void *arg, int node) | |
2445 | { | |
2446 | unsigned long flags = CLONE_FS|CLONE_FILES|CLONE_SIGHAND|CLONE_THREAD| | |
2447 | CLONE_IO; | |
2448 | struct kernel_clone_args args = { | |
2449 | .flags = ((lower_32_bits(flags) | CLONE_VM | | |
2450 | CLONE_UNTRACED) & ~CSIGNAL), | |
2451 | .exit_signal = (lower_32_bits(flags) & CSIGNAL), | |
2452 | .stack = (unsigned long)fn, | |
2453 | .stack_size = (unsigned long)arg, | |
2454 | .io_thread = 1, | |
2455 | }; | |
cc440e87 | 2456 | |
b16b3855 | 2457 | return copy_process(NULL, 0, node, &args); |
cc440e87 JA |
2458 | } |
2459 | ||
1da177e4 LT |
2460 | /* |
2461 | * Ok, this is the main fork-routine. | |
2462 | * | |
2463 | * It copies the process, and if successful kick-starts | |
2464 | * it and waits for it to finish using the VM if required. | |
a0eb9abd ES |
2465 | * |
2466 | * args->exit_signal is expected to be checked for sanity by the caller. | |
1da177e4 | 2467 | */ |
cad6967a | 2468 | pid_t kernel_clone(struct kernel_clone_args *args) |
1da177e4 | 2469 | { |
7f192e3c | 2470 | u64 clone_flags = args->flags; |
9f5325aa MPS |
2471 | struct completion vfork; |
2472 | struct pid *pid; | |
1da177e4 LT |
2473 | struct task_struct *p; |
2474 | int trace = 0; | |
cad6967a | 2475 | pid_t nr; |
1da177e4 | 2476 | |
3af8588c CB |
2477 | /* |
2478 | * For legacy clone() calls, CLONE_PIDFD uses the parent_tid argument | |
2479 | * to return the pidfd. Hence, CLONE_PIDFD and CLONE_PARENT_SETTID are | |
2480 | * mutually exclusive. With clone3() CLONE_PIDFD has grown a separate | |
2481 | * field in struct clone_args and it still doesn't make sense to have | |
2482 | * them both point at the same memory location. Performing this check | |
2483 | * here has the advantage that we don't need to have a separate helper | |
2484 | * to check for legacy clone(). | |
2485 | */ | |
2486 | if ((args->flags & CLONE_PIDFD) && | |
2487 | (args->flags & CLONE_PARENT_SETTID) && | |
2488 | (args->pidfd == args->parent_tid)) | |
2489 | return -EINVAL; | |
2490 | ||
09a05394 | 2491 | /* |
4b9d33e6 TH |
2492 | * Determine whether and which event to report to ptracer. When |
2493 | * called from kernel_thread or CLONE_UNTRACED is explicitly | |
2494 | * requested, no event is reported; otherwise, report if the event | |
2495 | * for the type of forking is enabled. | |
09a05394 | 2496 | */ |
e80d6661 | 2497 | if (!(clone_flags & CLONE_UNTRACED)) { |
4b9d33e6 TH |
2498 | if (clone_flags & CLONE_VFORK) |
2499 | trace = PTRACE_EVENT_VFORK; | |
7f192e3c | 2500 | else if (args->exit_signal != SIGCHLD) |
4b9d33e6 TH |
2501 | trace = PTRACE_EVENT_CLONE; |
2502 | else | |
2503 | trace = PTRACE_EVENT_FORK; | |
2504 | ||
2505 | if (likely(!ptrace_event_enabled(current, trace))) | |
2506 | trace = 0; | |
2507 | } | |
1da177e4 | 2508 | |
7f192e3c | 2509 | p = copy_process(NULL, trace, NUMA_NO_NODE, args); |
38addce8 | 2510 | add_latent_entropy(); |
9f5325aa MPS |
2511 | |
2512 | if (IS_ERR(p)) | |
2513 | return PTR_ERR(p); | |
2514 | ||
1da177e4 LT |
2515 | /* |
2516 | * Do this prior waking up the new thread - the thread pointer | |
2517 | * might get invalid after that point, if the thread exits quickly. | |
2518 | */ | |
9f5325aa | 2519 | trace_sched_process_fork(current, p); |
0a16b607 | 2520 | |
9f5325aa MPS |
2521 | pid = get_task_pid(p, PIDTYPE_PID); |
2522 | nr = pid_vnr(pid); | |
30e49c26 | 2523 | |
9f5325aa | 2524 | if (clone_flags & CLONE_PARENT_SETTID) |
7f192e3c | 2525 | put_user(nr, args->parent_tid); |
a6f5e063 | 2526 | |
9f5325aa MPS |
2527 | if (clone_flags & CLONE_VFORK) { |
2528 | p->vfork_done = &vfork; | |
2529 | init_completion(&vfork); | |
2530 | get_task_struct(p); | |
2531 | } | |
1da177e4 | 2532 | |
9f5325aa | 2533 | wake_up_new_task(p); |
09a05394 | 2534 | |
9f5325aa MPS |
2535 | /* forking complete and child started to run, tell ptracer */ |
2536 | if (unlikely(trace)) | |
2537 | ptrace_event_pid(trace, pid); | |
4e52365f | 2538 | |
9f5325aa MPS |
2539 | if (clone_flags & CLONE_VFORK) { |
2540 | if (!wait_for_vfork_done(p, &vfork)) | |
2541 | ptrace_event_pid(PTRACE_EVENT_VFORK_DONE, pid); | |
1da177e4 | 2542 | } |
9f5325aa MPS |
2543 | |
2544 | put_pid(pid); | |
92476d7f | 2545 | return nr; |
1da177e4 LT |
2546 | } |
2547 | ||
2aa3a7f8 AV |
2548 | /* |
2549 | * Create a kernel thread. | |
2550 | */ | |
2551 | pid_t kernel_thread(int (*fn)(void *), void *arg, unsigned long flags) | |
2552 | { | |
7f192e3c | 2553 | struct kernel_clone_args args = { |
3f2c788a CB |
2554 | .flags = ((lower_32_bits(flags) | CLONE_VM | |
2555 | CLONE_UNTRACED) & ~CSIGNAL), | |
2556 | .exit_signal = (lower_32_bits(flags) & CSIGNAL), | |
7f192e3c CB |
2557 | .stack = (unsigned long)fn, |
2558 | .stack_size = (unsigned long)arg, | |
2559 | }; | |
2560 | ||
cad6967a | 2561 | return kernel_clone(&args); |
2aa3a7f8 | 2562 | } |
2aa3a7f8 | 2563 | |
d2125043 AV |
2564 | #ifdef __ARCH_WANT_SYS_FORK |
2565 | SYSCALL_DEFINE0(fork) | |
2566 | { | |
2567 | #ifdef CONFIG_MMU | |
7f192e3c CB |
2568 | struct kernel_clone_args args = { |
2569 | .exit_signal = SIGCHLD, | |
2570 | }; | |
2571 | ||
cad6967a | 2572 | return kernel_clone(&args); |
d2125043 AV |
2573 | #else |
2574 | /* can not support in nommu mode */ | |
5d59e182 | 2575 | return -EINVAL; |
d2125043 AV |
2576 | #endif |
2577 | } | |
2578 | #endif | |
2579 | ||
2580 | #ifdef __ARCH_WANT_SYS_VFORK | |
2581 | SYSCALL_DEFINE0(vfork) | |
2582 | { | |
7f192e3c CB |
2583 | struct kernel_clone_args args = { |
2584 | .flags = CLONE_VFORK | CLONE_VM, | |
2585 | .exit_signal = SIGCHLD, | |
2586 | }; | |
2587 | ||
cad6967a | 2588 | return kernel_clone(&args); |
d2125043 AV |
2589 | } |
2590 | #endif | |
2591 | ||
2592 | #ifdef __ARCH_WANT_SYS_CLONE | |
2593 | #ifdef CONFIG_CLONE_BACKWARDS | |
2594 | SYSCALL_DEFINE5(clone, unsigned long, clone_flags, unsigned long, newsp, | |
2595 | int __user *, parent_tidptr, | |
3033f14a | 2596 | unsigned long, tls, |
d2125043 AV |
2597 | int __user *, child_tidptr) |
2598 | #elif defined(CONFIG_CLONE_BACKWARDS2) | |
2599 | SYSCALL_DEFINE5(clone, unsigned long, newsp, unsigned long, clone_flags, | |
2600 | int __user *, parent_tidptr, | |
2601 | int __user *, child_tidptr, | |
3033f14a | 2602 | unsigned long, tls) |
dfa9771a MS |
2603 | #elif defined(CONFIG_CLONE_BACKWARDS3) |
2604 | SYSCALL_DEFINE6(clone, unsigned long, clone_flags, unsigned long, newsp, | |
2605 | int, stack_size, | |
2606 | int __user *, parent_tidptr, | |
2607 | int __user *, child_tidptr, | |
3033f14a | 2608 | unsigned long, tls) |
d2125043 AV |
2609 | #else |
2610 | SYSCALL_DEFINE5(clone, unsigned long, clone_flags, unsigned long, newsp, | |
2611 | int __user *, parent_tidptr, | |
2612 | int __user *, child_tidptr, | |
3033f14a | 2613 | unsigned long, tls) |
d2125043 AV |
2614 | #endif |
2615 | { | |
7f192e3c | 2616 | struct kernel_clone_args args = { |
3f2c788a | 2617 | .flags = (lower_32_bits(clone_flags) & ~CSIGNAL), |
7f192e3c CB |
2618 | .pidfd = parent_tidptr, |
2619 | .child_tid = child_tidptr, | |
2620 | .parent_tid = parent_tidptr, | |
3f2c788a | 2621 | .exit_signal = (lower_32_bits(clone_flags) & CSIGNAL), |
7f192e3c CB |
2622 | .stack = newsp, |
2623 | .tls = tls, | |
2624 | }; | |
2625 | ||
cad6967a | 2626 | return kernel_clone(&args); |
7f192e3c | 2627 | } |
d68dbb0c | 2628 | #endif |
7f192e3c | 2629 | |
d68dbb0c | 2630 | #ifdef __ARCH_WANT_SYS_CLONE3 |
dd499f7a | 2631 | |
7f192e3c CB |
2632 | noinline static int copy_clone_args_from_user(struct kernel_clone_args *kargs, |
2633 | struct clone_args __user *uargs, | |
f14c234b | 2634 | size_t usize) |
7f192e3c | 2635 | { |
f14c234b | 2636 | int err; |
7f192e3c | 2637 | struct clone_args args; |
49cb2fc4 | 2638 | pid_t *kset_tid = kargs->set_tid; |
7f192e3c | 2639 | |
a966dcfe ES |
2640 | BUILD_BUG_ON(offsetofend(struct clone_args, tls) != |
2641 | CLONE_ARGS_SIZE_VER0); | |
2642 | BUILD_BUG_ON(offsetofend(struct clone_args, set_tid_size) != | |
2643 | CLONE_ARGS_SIZE_VER1); | |
2644 | BUILD_BUG_ON(offsetofend(struct clone_args, cgroup) != | |
2645 | CLONE_ARGS_SIZE_VER2); | |
2646 | BUILD_BUG_ON(sizeof(struct clone_args) != CLONE_ARGS_SIZE_VER2); | |
2647 | ||
f14c234b | 2648 | if (unlikely(usize > PAGE_SIZE)) |
7f192e3c | 2649 | return -E2BIG; |
f14c234b | 2650 | if (unlikely(usize < CLONE_ARGS_SIZE_VER0)) |
7f192e3c CB |
2651 | return -EINVAL; |
2652 | ||
f14c234b AS |
2653 | err = copy_struct_from_user(&args, sizeof(args), uargs, usize); |
2654 | if (err) | |
2655 | return err; | |
7f192e3c | 2656 | |
49cb2fc4 AR |
2657 | if (unlikely(args.set_tid_size > MAX_PID_NS_LEVEL)) |
2658 | return -EINVAL; | |
2659 | ||
2660 | if (unlikely(!args.set_tid && args.set_tid_size > 0)) | |
2661 | return -EINVAL; | |
2662 | ||
2663 | if (unlikely(args.set_tid && args.set_tid_size == 0)) | |
2664 | return -EINVAL; | |
2665 | ||
a0eb9abd ES |
2666 | /* |
2667 | * Verify that higher 32bits of exit_signal are unset and that | |
2668 | * it is a valid signal | |
2669 | */ | |
2670 | if (unlikely((args.exit_signal & ~((u64)CSIGNAL)) || | |
2671 | !valid_signal(args.exit_signal))) | |
2672 | return -EINVAL; | |
2673 | ||
62173872 ES |
2674 | if ((args.flags & CLONE_INTO_CGROUP) && |
2675 | (args.cgroup > INT_MAX || usize < CLONE_ARGS_SIZE_VER2)) | |
ef2c41cf CB |
2676 | return -EINVAL; |
2677 | ||
7f192e3c CB |
2678 | *kargs = (struct kernel_clone_args){ |
2679 | .flags = args.flags, | |
2680 | .pidfd = u64_to_user_ptr(args.pidfd), | |
2681 | .child_tid = u64_to_user_ptr(args.child_tid), | |
2682 | .parent_tid = u64_to_user_ptr(args.parent_tid), | |
2683 | .exit_signal = args.exit_signal, | |
2684 | .stack = args.stack, | |
2685 | .stack_size = args.stack_size, | |
2686 | .tls = args.tls, | |
49cb2fc4 | 2687 | .set_tid_size = args.set_tid_size, |
ef2c41cf | 2688 | .cgroup = args.cgroup, |
7f192e3c CB |
2689 | }; |
2690 | ||
49cb2fc4 AR |
2691 | if (args.set_tid && |
2692 | copy_from_user(kset_tid, u64_to_user_ptr(args.set_tid), | |
2693 | (kargs->set_tid_size * sizeof(pid_t)))) | |
2694 | return -EFAULT; | |
2695 | ||
2696 | kargs->set_tid = kset_tid; | |
2697 | ||
7f192e3c CB |
2698 | return 0; |
2699 | } | |
2700 | ||
fa729c4d CB |
2701 | /** |
2702 | * clone3_stack_valid - check and prepare stack | |
2703 | * @kargs: kernel clone args | |
2704 | * | |
2705 | * Verify that the stack arguments userspace gave us are sane. | |
2706 | * In addition, set the stack direction for userspace since it's easy for us to | |
2707 | * determine. | |
2708 | */ | |
2709 | static inline bool clone3_stack_valid(struct kernel_clone_args *kargs) | |
2710 | { | |
2711 | if (kargs->stack == 0) { | |
2712 | if (kargs->stack_size > 0) | |
2713 | return false; | |
2714 | } else { | |
2715 | if (kargs->stack_size == 0) | |
2716 | return false; | |
2717 | ||
2718 | if (!access_ok((void __user *)kargs->stack, kargs->stack_size)) | |
2719 | return false; | |
2720 | ||
2721 | #if !defined(CONFIG_STACK_GROWSUP) && !defined(CONFIG_IA64) | |
2722 | kargs->stack += kargs->stack_size; | |
2723 | #endif | |
2724 | } | |
2725 | ||
2726 | return true; | |
2727 | } | |
2728 | ||
2729 | static bool clone3_args_valid(struct kernel_clone_args *kargs) | |
7f192e3c | 2730 | { |
b612e5df | 2731 | /* Verify that no unknown flags are passed along. */ |
ef2c41cf CB |
2732 | if (kargs->flags & |
2733 | ~(CLONE_LEGACY_FLAGS | CLONE_CLEAR_SIGHAND | CLONE_INTO_CGROUP)) | |
7f192e3c CB |
2734 | return false; |
2735 | ||
2736 | /* | |
a8ca6b13 XC |
2737 | * - make the CLONE_DETACHED bit reusable for clone3 |
2738 | * - make the CSIGNAL bits reusable for clone3 | |
7f192e3c CB |
2739 | */ |
2740 | if (kargs->flags & (CLONE_DETACHED | CSIGNAL)) | |
2741 | return false; | |
2742 | ||
b612e5df CB |
2743 | if ((kargs->flags & (CLONE_SIGHAND | CLONE_CLEAR_SIGHAND)) == |
2744 | (CLONE_SIGHAND | CLONE_CLEAR_SIGHAND)) | |
2745 | return false; | |
2746 | ||
7f192e3c CB |
2747 | if ((kargs->flags & (CLONE_THREAD | CLONE_PARENT)) && |
2748 | kargs->exit_signal) | |
2749 | return false; | |
2750 | ||
fa729c4d CB |
2751 | if (!clone3_stack_valid(kargs)) |
2752 | return false; | |
2753 | ||
7f192e3c CB |
2754 | return true; |
2755 | } | |
2756 | ||
501bd016 CB |
2757 | /** |
2758 | * clone3 - create a new process with specific properties | |
2759 | * @uargs: argument structure | |
2760 | * @size: size of @uargs | |
2761 | * | |
2762 | * clone3() is the extensible successor to clone()/clone2(). | |
2763 | * It takes a struct as argument that is versioned by its size. | |
2764 | * | |
2765 | * Return: On success, a positive PID for the child process. | |
2766 | * On error, a negative errno number. | |
2767 | */ | |
7f192e3c CB |
2768 | SYSCALL_DEFINE2(clone3, struct clone_args __user *, uargs, size_t, size) |
2769 | { | |
2770 | int err; | |
2771 | ||
2772 | struct kernel_clone_args kargs; | |
49cb2fc4 AR |
2773 | pid_t set_tid[MAX_PID_NS_LEVEL]; |
2774 | ||
2775 | kargs.set_tid = set_tid; | |
7f192e3c CB |
2776 | |
2777 | err = copy_clone_args_from_user(&kargs, uargs, size); | |
2778 | if (err) | |
2779 | return err; | |
2780 | ||
2781 | if (!clone3_args_valid(&kargs)) | |
2782 | return -EINVAL; | |
2783 | ||
cad6967a | 2784 | return kernel_clone(&kargs); |
d2125043 AV |
2785 | } |
2786 | #endif | |
2787 | ||
0f1b92cb ON |
2788 | void walk_process_tree(struct task_struct *top, proc_visitor visitor, void *data) |
2789 | { | |
2790 | struct task_struct *leader, *parent, *child; | |
2791 | int res; | |
2792 | ||
2793 | read_lock(&tasklist_lock); | |
2794 | leader = top = top->group_leader; | |
2795 | down: | |
2796 | for_each_thread(leader, parent) { | |
2797 | list_for_each_entry(child, &parent->children, sibling) { | |
2798 | res = visitor(child, data); | |
2799 | if (res) { | |
2800 | if (res < 0) | |
2801 | goto out; | |
2802 | leader = child; | |
2803 | goto down; | |
2804 | } | |
2805 | up: | |
2806 | ; | |
2807 | } | |
2808 | } | |
2809 | ||
2810 | if (leader != top) { | |
2811 | child = leader; | |
2812 | parent = child->real_parent; | |
2813 | leader = parent->group_leader; | |
2814 | goto up; | |
2815 | } | |
2816 | out: | |
2817 | read_unlock(&tasklist_lock); | |
2818 | } | |
2819 | ||
5fd63b30 RT |
2820 | #ifndef ARCH_MIN_MMSTRUCT_ALIGN |
2821 | #define ARCH_MIN_MMSTRUCT_ALIGN 0 | |
2822 | #endif | |
2823 | ||
51cc5068 | 2824 | static void sighand_ctor(void *data) |
aa1757f9 ON |
2825 | { |
2826 | struct sighand_struct *sighand = data; | |
2827 | ||
a35afb83 | 2828 | spin_lock_init(&sighand->siglock); |
b8fceee1 | 2829 | init_waitqueue_head(&sighand->signalfd_wqh); |
aa1757f9 ON |
2830 | } |
2831 | ||
1da177e4 LT |
2832 | void __init proc_caches_init(void) |
2833 | { | |
c1a2f7f0 RR |
2834 | unsigned int mm_size; |
2835 | ||
1da177e4 LT |
2836 | sighand_cachep = kmem_cache_create("sighand_cache", |
2837 | sizeof(struct sighand_struct), 0, | |
5f0d5a3a | 2838 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_TYPESAFE_BY_RCU| |
75f296d9 | 2839 | SLAB_ACCOUNT, sighand_ctor); |
1da177e4 LT |
2840 | signal_cachep = kmem_cache_create("signal_cache", |
2841 | sizeof(struct signal_struct), 0, | |
75f296d9 | 2842 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_ACCOUNT, |
5d097056 | 2843 | NULL); |
20c2df83 | 2844 | files_cachep = kmem_cache_create("files_cache", |
1da177e4 | 2845 | sizeof(struct files_struct), 0, |
75f296d9 | 2846 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_ACCOUNT, |
5d097056 | 2847 | NULL); |
20c2df83 | 2848 | fs_cachep = kmem_cache_create("fs_cache", |
1da177e4 | 2849 | sizeof(struct fs_struct), 0, |
75f296d9 | 2850 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_ACCOUNT, |
5d097056 | 2851 | NULL); |
c1a2f7f0 | 2852 | |
6345d24d | 2853 | /* |
c1a2f7f0 RR |
2854 | * The mm_cpumask is located at the end of mm_struct, and is |
2855 | * dynamically sized based on the maximum CPU number this system | |
2856 | * can have, taking hotplug into account (nr_cpu_ids). | |
6345d24d | 2857 | */ |
c1a2f7f0 RR |
2858 | mm_size = sizeof(struct mm_struct) + cpumask_size(); |
2859 | ||
07dcd7fe | 2860 | mm_cachep = kmem_cache_create_usercopy("mm_struct", |
c1a2f7f0 | 2861 | mm_size, ARCH_MIN_MMSTRUCT_ALIGN, |
75f296d9 | 2862 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_ACCOUNT, |
07dcd7fe DW |
2863 | offsetof(struct mm_struct, saved_auxv), |
2864 | sizeof_field(struct mm_struct, saved_auxv), | |
5d097056 VD |
2865 | NULL); |
2866 | vm_area_cachep = KMEM_CACHE(vm_area_struct, SLAB_PANIC|SLAB_ACCOUNT); | |
8feae131 | 2867 | mmap_init(); |
66577193 | 2868 | nsproxy_cache_init(); |
1da177e4 | 2869 | } |
cf2e340f | 2870 | |
cf2e340f | 2871 | /* |
9bfb23fc | 2872 | * Check constraints on flags passed to the unshare system call. |
cf2e340f | 2873 | */ |
9bfb23fc | 2874 | static int check_unshare_flags(unsigned long unshare_flags) |
cf2e340f | 2875 | { |
9bfb23fc ON |
2876 | if (unshare_flags & ~(CLONE_THREAD|CLONE_FS|CLONE_NEWNS|CLONE_SIGHAND| |
2877 | CLONE_VM|CLONE_FILES|CLONE_SYSVSEM| | |
50804fe3 | 2878 | CLONE_NEWUTS|CLONE_NEWIPC|CLONE_NEWNET| |
769071ac AV |
2879 | CLONE_NEWUSER|CLONE_NEWPID|CLONE_NEWCGROUP| |
2880 | CLONE_NEWTIME)) | |
9bfb23fc | 2881 | return -EINVAL; |
cf2e340f | 2882 | /* |
12c641ab EB |
2883 | * Not implemented, but pretend it works if there is nothing |
2884 | * to unshare. Note that unsharing the address space or the | |
2885 | * signal handlers also need to unshare the signal queues (aka | |
2886 | * CLONE_THREAD). | |
cf2e340f | 2887 | */ |
9bfb23fc | 2888 | if (unshare_flags & (CLONE_THREAD | CLONE_SIGHAND | CLONE_VM)) { |
12c641ab EB |
2889 | if (!thread_group_empty(current)) |
2890 | return -EINVAL; | |
2891 | } | |
2892 | if (unshare_flags & (CLONE_SIGHAND | CLONE_VM)) { | |
d036bda7 | 2893 | if (refcount_read(¤t->sighand->count) > 1) |
12c641ab EB |
2894 | return -EINVAL; |
2895 | } | |
2896 | if (unshare_flags & CLONE_VM) { | |
2897 | if (!current_is_single_threaded()) | |
9bfb23fc ON |
2898 | return -EINVAL; |
2899 | } | |
cf2e340f JD |
2900 | |
2901 | return 0; | |
2902 | } | |
2903 | ||
2904 | /* | |
99d1419d | 2905 | * Unshare the filesystem structure if it is being shared |
cf2e340f JD |
2906 | */ |
2907 | static int unshare_fs(unsigned long unshare_flags, struct fs_struct **new_fsp) | |
2908 | { | |
2909 | struct fs_struct *fs = current->fs; | |
2910 | ||
498052bb AV |
2911 | if (!(unshare_flags & CLONE_FS) || !fs) |
2912 | return 0; | |
2913 | ||
2914 | /* don't need lock here; in the worst case we'll do useless copy */ | |
2915 | if (fs->users == 1) | |
2916 | return 0; | |
2917 | ||
2918 | *new_fsp = copy_fs_struct(fs); | |
2919 | if (!*new_fsp) | |
2920 | return -ENOMEM; | |
cf2e340f JD |
2921 | |
2922 | return 0; | |
2923 | } | |
2924 | ||
cf2e340f | 2925 | /* |
a016f338 | 2926 | * Unshare file descriptor table if it is being shared |
cf2e340f | 2927 | */ |
60997c3d CB |
2928 | int unshare_fd(unsigned long unshare_flags, unsigned int max_fds, |
2929 | struct files_struct **new_fdp) | |
cf2e340f JD |
2930 | { |
2931 | struct files_struct *fd = current->files; | |
a016f338 | 2932 | int error = 0; |
cf2e340f JD |
2933 | |
2934 | if ((unshare_flags & CLONE_FILES) && | |
a016f338 | 2935 | (fd && atomic_read(&fd->count) > 1)) { |
60997c3d | 2936 | *new_fdp = dup_fd(fd, max_fds, &error); |
a016f338 JD |
2937 | if (!*new_fdp) |
2938 | return error; | |
2939 | } | |
cf2e340f JD |
2940 | |
2941 | return 0; | |
2942 | } | |
2943 | ||
cf2e340f JD |
2944 | /* |
2945 | * unshare allows a process to 'unshare' part of the process | |
2946 | * context which was originally shared using clone. copy_* | |
cad6967a | 2947 | * functions used by kernel_clone() cannot be used here directly |
cf2e340f JD |
2948 | * because they modify an inactive task_struct that is being |
2949 | * constructed. Here we are modifying the current, active, | |
2950 | * task_struct. | |
2951 | */ | |
9b32105e | 2952 | int ksys_unshare(unsigned long unshare_flags) |
cf2e340f | 2953 | { |
cf2e340f | 2954 | struct fs_struct *fs, *new_fs = NULL; |
cf2e340f | 2955 | struct files_struct *fd, *new_fd = NULL; |
b2e0d987 | 2956 | struct cred *new_cred = NULL; |
cf7b708c | 2957 | struct nsproxy *new_nsproxy = NULL; |
9edff4ab | 2958 | int do_sysvsem = 0; |
9bfb23fc | 2959 | int err; |
cf2e340f | 2960 | |
b2e0d987 | 2961 | /* |
faf00da5 EB |
2962 | * If unsharing a user namespace must also unshare the thread group |
2963 | * and unshare the filesystem root and working directories. | |
b2e0d987 EB |
2964 | */ |
2965 | if (unshare_flags & CLONE_NEWUSER) | |
e66eded8 | 2966 | unshare_flags |= CLONE_THREAD | CLONE_FS; |
50804fe3 EB |
2967 | /* |
2968 | * If unsharing vm, must also unshare signal handlers. | |
2969 | */ | |
2970 | if (unshare_flags & CLONE_VM) | |
2971 | unshare_flags |= CLONE_SIGHAND; | |
12c641ab EB |
2972 | /* |
2973 | * If unsharing a signal handlers, must also unshare the signal queues. | |
2974 | */ | |
2975 | if (unshare_flags & CLONE_SIGHAND) | |
2976 | unshare_flags |= CLONE_THREAD; | |
9bfb23fc ON |
2977 | /* |
2978 | * If unsharing namespace, must also unshare filesystem information. | |
2979 | */ | |
2980 | if (unshare_flags & CLONE_NEWNS) | |
2981 | unshare_flags |= CLONE_FS; | |
50804fe3 EB |
2982 | |
2983 | err = check_unshare_flags(unshare_flags); | |
2984 | if (err) | |
2985 | goto bad_unshare_out; | |
6013f67f MS |
2986 | /* |
2987 | * CLONE_NEWIPC must also detach from the undolist: after switching | |
2988 | * to a new ipc namespace, the semaphore arrays from the old | |
2989 | * namespace are unreachable. | |
2990 | */ | |
2991 | if (unshare_flags & (CLONE_NEWIPC|CLONE_SYSVSEM)) | |
9edff4ab | 2992 | do_sysvsem = 1; |
fb0a685c DRO |
2993 | err = unshare_fs(unshare_flags, &new_fs); |
2994 | if (err) | |
9bfb23fc | 2995 | goto bad_unshare_out; |
60997c3d | 2996 | err = unshare_fd(unshare_flags, NR_OPEN_MAX, &new_fd); |
fb0a685c | 2997 | if (err) |
9bfb23fc | 2998 | goto bad_unshare_cleanup_fs; |
b2e0d987 | 2999 | err = unshare_userns(unshare_flags, &new_cred); |
fb0a685c | 3000 | if (err) |
9edff4ab | 3001 | goto bad_unshare_cleanup_fd; |
b2e0d987 EB |
3002 | err = unshare_nsproxy_namespaces(unshare_flags, &new_nsproxy, |
3003 | new_cred, new_fs); | |
3004 | if (err) | |
3005 | goto bad_unshare_cleanup_cred; | |
c0b2fc31 | 3006 | |
905ae01c AG |
3007 | if (new_cred) { |
3008 | err = set_cred_ucounts(new_cred); | |
3009 | if (err) | |
3010 | goto bad_unshare_cleanup_cred; | |
3011 | } | |
3012 | ||
b2e0d987 | 3013 | if (new_fs || new_fd || do_sysvsem || new_cred || new_nsproxy) { |
9edff4ab MS |
3014 | if (do_sysvsem) { |
3015 | /* | |
3016 | * CLONE_SYSVSEM is equivalent to sys_exit(). | |
3017 | */ | |
3018 | exit_sem(current); | |
3019 | } | |
ab602f79 JM |
3020 | if (unshare_flags & CLONE_NEWIPC) { |
3021 | /* Orphan segments in old ns (see sem above). */ | |
3022 | exit_shm(current); | |
3023 | shm_init_task(current); | |
3024 | } | |
ab516013 | 3025 | |
6f977e6b | 3026 | if (new_nsproxy) |
cf7b708c | 3027 | switch_task_namespaces(current, new_nsproxy); |
cf2e340f | 3028 | |
cf7b708c PE |
3029 | task_lock(current); |
3030 | ||
cf2e340f JD |
3031 | if (new_fs) { |
3032 | fs = current->fs; | |
2a4419b5 | 3033 | spin_lock(&fs->lock); |
cf2e340f | 3034 | current->fs = new_fs; |
498052bb AV |
3035 | if (--fs->users) |
3036 | new_fs = NULL; | |
3037 | else | |
3038 | new_fs = fs; | |
2a4419b5 | 3039 | spin_unlock(&fs->lock); |
cf2e340f JD |
3040 | } |
3041 | ||
cf2e340f JD |
3042 | if (new_fd) { |
3043 | fd = current->files; | |
3044 | current->files = new_fd; | |
3045 | new_fd = fd; | |
3046 | } | |
3047 | ||
3048 | task_unlock(current); | |
b2e0d987 EB |
3049 | |
3050 | if (new_cred) { | |
3051 | /* Install the new user namespace */ | |
3052 | commit_creds(new_cred); | |
3053 | new_cred = NULL; | |
3054 | } | |
cf2e340f JD |
3055 | } |
3056 | ||
e4222673 HB |
3057 | perf_event_namespaces(current); |
3058 | ||
b2e0d987 EB |
3059 | bad_unshare_cleanup_cred: |
3060 | if (new_cred) | |
3061 | put_cred(new_cred); | |
cf2e340f JD |
3062 | bad_unshare_cleanup_fd: |
3063 | if (new_fd) | |
3064 | put_files_struct(new_fd); | |
3065 | ||
cf2e340f JD |
3066 | bad_unshare_cleanup_fs: |
3067 | if (new_fs) | |
498052bb | 3068 | free_fs_struct(new_fs); |
cf2e340f | 3069 | |
cf2e340f JD |
3070 | bad_unshare_out: |
3071 | return err; | |
3072 | } | |
3b125388 | 3073 | |
9b32105e DB |
3074 | SYSCALL_DEFINE1(unshare, unsigned long, unshare_flags) |
3075 | { | |
3076 | return ksys_unshare(unshare_flags); | |
3077 | } | |
3078 | ||
3b125388 AV |
3079 | /* |
3080 | * Helper to unshare the files of the current task. | |
3081 | * We don't want to expose copy_files internals to | |
3082 | * the exec layer of the kernel. | |
3083 | */ | |
3084 | ||
1f702603 | 3085 | int unshare_files(void) |
3b125388 AV |
3086 | { |
3087 | struct task_struct *task = current; | |
1f702603 | 3088 | struct files_struct *old, *copy = NULL; |
3b125388 AV |
3089 | int error; |
3090 | ||
60997c3d | 3091 | error = unshare_fd(CLONE_FILES, NR_OPEN_MAX, ©); |
1f702603 | 3092 | if (error || !copy) |
3b125388 | 3093 | return error; |
1f702603 EB |
3094 | |
3095 | old = task->files; | |
3b125388 AV |
3096 | task_lock(task); |
3097 | task->files = copy; | |
3098 | task_unlock(task); | |
1f702603 | 3099 | put_files_struct(old); |
3b125388 AV |
3100 | return 0; |
3101 | } | |
16db3d3f HS |
3102 | |
3103 | int sysctl_max_threads(struct ctl_table *table, int write, | |
b0daa2c7 | 3104 | void *buffer, size_t *lenp, loff_t *ppos) |
16db3d3f HS |
3105 | { |
3106 | struct ctl_table t; | |
3107 | int ret; | |
3108 | int threads = max_threads; | |
b0f53dbc | 3109 | int min = 1; |
16db3d3f HS |
3110 | int max = MAX_THREADS; |
3111 | ||
3112 | t = *table; | |
3113 | t.data = &threads; | |
3114 | t.extra1 = &min; | |
3115 | t.extra2 = &max; | |
3116 | ||
3117 | ret = proc_dointvec_minmax(&t, write, buffer, lenp, ppos); | |
3118 | if (ret || !write) | |
3119 | return ret; | |
3120 | ||
b0f53dbc | 3121 | max_threads = threads; |
16db3d3f HS |
3122 | |
3123 | return 0; | |
3124 | } |