]>
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> |
133ff0ea | 43 | #include <linux/hmm.h> |
1da177e4 | 44 | #include <linux/fs.h> |
615d6e87 DB |
45 | #include <linux/mm.h> |
46 | #include <linux/vmacache.h> | |
ab516013 | 47 | #include <linux/nsproxy.h> |
c59ede7b | 48 | #include <linux/capability.h> |
1da177e4 | 49 | #include <linux/cpu.h> |
b4f48b63 | 50 | #include <linux/cgroup.h> |
1da177e4 | 51 | #include <linux/security.h> |
a1e78772 | 52 | #include <linux/hugetlb.h> |
e2cfabdf | 53 | #include <linux/seccomp.h> |
1da177e4 LT |
54 | #include <linux/swap.h> |
55 | #include <linux/syscalls.h> | |
56 | #include <linux/jiffies.h> | |
57 | #include <linux/futex.h> | |
8141c7f3 | 58 | #include <linux/compat.h> |
207205a2 | 59 | #include <linux/kthread.h> |
7c3ab738 | 60 | #include <linux/task_io_accounting_ops.h> |
ab2af1f5 | 61 | #include <linux/rcupdate.h> |
1da177e4 LT |
62 | #include <linux/ptrace.h> |
63 | #include <linux/mount.h> | |
64 | #include <linux/audit.h> | |
78fb7466 | 65 | #include <linux/memcontrol.h> |
f201ae23 | 66 | #include <linux/ftrace.h> |
5e2bf014 | 67 | #include <linux/proc_fs.h> |
1da177e4 LT |
68 | #include <linux/profile.h> |
69 | #include <linux/rmap.h> | |
f8af4da3 | 70 | #include <linux/ksm.h> |
1da177e4 | 71 | #include <linux/acct.h> |
893e26e6 | 72 | #include <linux/userfaultfd_k.h> |
8f0ab514 | 73 | #include <linux/tsacct_kern.h> |
9f46080c | 74 | #include <linux/cn_proc.h> |
ba96a0c8 | 75 | #include <linux/freezer.h> |
ca74e92b | 76 | #include <linux/delayacct.h> |
ad4ecbcb | 77 | #include <linux/taskstats_kern.h> |
0a425405 | 78 | #include <linux/random.h> |
522ed776 | 79 | #include <linux/tty.h> |
fd0928df | 80 | #include <linux/blkdev.h> |
5ad4e53b | 81 | #include <linux/fs_struct.h> |
7c9f8861 | 82 | #include <linux/magic.h> |
cdd6c482 | 83 | #include <linux/perf_event.h> |
42c4ab41 | 84 | #include <linux/posix-timers.h> |
8e7cac79 | 85 | #include <linux/user-return-notifier.h> |
3d5992d2 | 86 | #include <linux/oom.h> |
ba76149f | 87 | #include <linux/khugepaged.h> |
d80e731e | 88 | #include <linux/signalfd.h> |
0326f5a9 | 89 | #include <linux/uprobes.h> |
a27bb332 | 90 | #include <linux/aio.h> |
52f5684c | 91 | #include <linux/compiler.h> |
16db3d3f | 92 | #include <linux/sysctl.h> |
5c9a8750 | 93 | #include <linux/kcov.h> |
d83a7cb3 | 94 | #include <linux/livepatch.h> |
48ac3c18 | 95 | #include <linux/thread_info.h> |
afaef01c | 96 | #include <linux/stackleak.h> |
1da177e4 LT |
97 | |
98 | #include <asm/pgtable.h> | |
99 | #include <asm/pgalloc.h> | |
7c0f6ba6 | 100 | #include <linux/uaccess.h> |
1da177e4 LT |
101 | #include <asm/mmu_context.h> |
102 | #include <asm/cacheflush.h> | |
103 | #include <asm/tlbflush.h> | |
104 | ||
ad8d75ff SR |
105 | #include <trace/events/sched.h> |
106 | ||
43d2b113 KH |
107 | #define CREATE_TRACE_POINTS |
108 | #include <trace/events/task.h> | |
2b612266 SH |
109 | #ifdef CONFIG_USER_NS |
110 | extern int unprivileged_userns_clone; | |
111 | #else | |
112 | #define unprivileged_userns_clone 0 | |
113 | #endif | |
43d2b113 | 114 | |
ac1b398d HS |
115 | /* |
116 | * Minimum number of threads to boot the kernel | |
117 | */ | |
118 | #define MIN_THREADS 20 | |
119 | ||
120 | /* | |
121 | * Maximum number of threads | |
122 | */ | |
123 | #define MAX_THREADS FUTEX_TID_MASK | |
124 | ||
1da177e4 LT |
125 | /* |
126 | * Protected counters by write_lock_irq(&tasklist_lock) | |
127 | */ | |
128 | unsigned long total_forks; /* Handle normal Linux uptimes. */ | |
fb0a685c | 129 | int nr_threads; /* The idle threads do not count.. */ |
1da177e4 | 130 | |
8856ae4d | 131 | static int max_threads; /* tunable limit on nr_threads */ |
1da177e4 LT |
132 | |
133 | DEFINE_PER_CPU(unsigned long, process_counts) = 0; | |
134 | ||
c59923a1 | 135 | __cacheline_aligned DEFINE_RWLOCK(tasklist_lock); /* outer */ |
db1466b3 PM |
136 | |
137 | #ifdef CONFIG_PROVE_RCU | |
138 | int lockdep_tasklist_lock_is_held(void) | |
139 | { | |
140 | return lockdep_is_held(&tasklist_lock); | |
141 | } | |
142 | EXPORT_SYMBOL_GPL(lockdep_tasklist_lock_is_held); | |
143 | #endif /* #ifdef CONFIG_PROVE_RCU */ | |
1da177e4 LT |
144 | |
145 | int nr_processes(void) | |
146 | { | |
147 | int cpu; | |
148 | int total = 0; | |
149 | ||
1d510750 | 150 | for_each_possible_cpu(cpu) |
1da177e4 LT |
151 | total += per_cpu(process_counts, cpu); |
152 | ||
153 | return total; | |
154 | } | |
155 | ||
f19b9f74 AM |
156 | void __weak arch_release_task_struct(struct task_struct *tsk) |
157 | { | |
158 | } | |
159 | ||
f5e10287 | 160 | #ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR |
e18b890b | 161 | static struct kmem_cache *task_struct_cachep; |
41101809 TG |
162 | |
163 | static inline struct task_struct *alloc_task_struct_node(int node) | |
164 | { | |
165 | return kmem_cache_alloc_node(task_struct_cachep, GFP_KERNEL, node); | |
166 | } | |
167 | ||
41101809 TG |
168 | static inline void free_task_struct(struct task_struct *tsk) |
169 | { | |
41101809 TG |
170 | kmem_cache_free(task_struct_cachep, tsk); |
171 | } | |
1da177e4 LT |
172 | #endif |
173 | ||
b235beea | 174 | #ifndef CONFIG_ARCH_THREAD_STACK_ALLOCATOR |
41101809 | 175 | |
0d15d74a TG |
176 | /* |
177 | * Allocate pages if THREAD_SIZE is >= PAGE_SIZE, otherwise use a | |
178 | * kmemcache based allocator. | |
179 | */ | |
ba14a194 | 180 | # if THREAD_SIZE >= PAGE_SIZE || defined(CONFIG_VMAP_STACK) |
ac496bf4 AL |
181 | |
182 | #ifdef CONFIG_VMAP_STACK | |
183 | /* | |
184 | * vmalloc() is a bit slow, and calling vfree() enough times will force a TLB | |
185 | * flush. Try to minimize the number of calls by caching stacks. | |
186 | */ | |
187 | #define NR_CACHED_STACKS 2 | |
188 | static DEFINE_PER_CPU(struct vm_struct *, cached_stacks[NR_CACHED_STACKS]); | |
19659c59 HR |
189 | |
190 | static int free_vm_stack_cache(unsigned int cpu) | |
191 | { | |
192 | struct vm_struct **cached_vm_stacks = per_cpu_ptr(cached_stacks, cpu); | |
193 | int i; | |
194 | ||
195 | for (i = 0; i < NR_CACHED_STACKS; i++) { | |
196 | struct vm_struct *vm_stack = cached_vm_stacks[i]; | |
197 | ||
198 | if (!vm_stack) | |
199 | continue; | |
200 | ||
201 | vfree(vm_stack->addr); | |
202 | cached_vm_stacks[i] = NULL; | |
203 | } | |
204 | ||
205 | return 0; | |
206 | } | |
ac496bf4 AL |
207 | #endif |
208 | ||
ba14a194 | 209 | static unsigned long *alloc_thread_stack_node(struct task_struct *tsk, int node) |
b69c49b7 | 210 | { |
ba14a194 | 211 | #ifdef CONFIG_VMAP_STACK |
ac496bf4 AL |
212 | void *stack; |
213 | int i; | |
214 | ||
ac496bf4 | 215 | for (i = 0; i < NR_CACHED_STACKS; i++) { |
112166f8 CL |
216 | struct vm_struct *s; |
217 | ||
218 | s = this_cpu_xchg(cached_stacks[i], NULL); | |
ac496bf4 AL |
219 | |
220 | if (!s) | |
221 | continue; | |
ac496bf4 | 222 | |
ca182551 KK |
223 | /* Clear stale pointers from reused stack. */ |
224 | memset(s->addr, 0, THREAD_SIZE); | |
e01e8063 | 225 | |
ac496bf4 | 226 | tsk->stack_vm_area = s; |
ba4a4574 | 227 | tsk->stack = s->addr; |
ac496bf4 AL |
228 | return s->addr; |
229 | } | |
ac496bf4 | 230 | |
9b6f7e16 RG |
231 | /* |
232 | * Allocated stacks are cached and later reused by new threads, | |
233 | * so memcg accounting is performed manually on assigning/releasing | |
234 | * stacks to tasks. Drop __GFP_ACCOUNT. | |
235 | */ | |
48ac3c18 | 236 | stack = __vmalloc_node_range(THREAD_SIZE, THREAD_ALIGN, |
ac496bf4 | 237 | VMALLOC_START, VMALLOC_END, |
9b6f7e16 | 238 | THREADINFO_GFP & ~__GFP_ACCOUNT, |
ac496bf4 AL |
239 | PAGE_KERNEL, |
240 | 0, node, __builtin_return_address(0)); | |
ba14a194 AL |
241 | |
242 | /* | |
243 | * We can't call find_vm_area() in interrupt context, and | |
244 | * free_thread_stack() can be called in interrupt context, | |
245 | * so cache the vm_struct. | |
246 | */ | |
5eed6f1d | 247 | if (stack) { |
ba14a194 | 248 | tsk->stack_vm_area = find_vm_area(stack); |
5eed6f1d RR |
249 | tsk->stack = stack; |
250 | } | |
ba14a194 AL |
251 | return stack; |
252 | #else | |
4949148a VD |
253 | struct page *page = alloc_pages_node(node, THREADINFO_GFP, |
254 | THREAD_SIZE_ORDER); | |
b6a84016 | 255 | |
1bf4580e AA |
256 | if (likely(page)) { |
257 | tsk->stack = page_address(page); | |
258 | return tsk->stack; | |
259 | } | |
260 | return NULL; | |
ba14a194 | 261 | #endif |
b69c49b7 FT |
262 | } |
263 | ||
ba14a194 | 264 | static inline void free_thread_stack(struct task_struct *tsk) |
b69c49b7 | 265 | { |
ac496bf4 | 266 | #ifdef CONFIG_VMAP_STACK |
9b6f7e16 RG |
267 | struct vm_struct *vm = task_stack_vm_area(tsk); |
268 | ||
269 | if (vm) { | |
ac496bf4 AL |
270 | int i; |
271 | ||
9b6f7e16 RG |
272 | for (i = 0; i < THREAD_SIZE / PAGE_SIZE; i++) { |
273 | mod_memcg_page_state(vm->pages[i], | |
274 | MEMCG_KERNEL_STACK_KB, | |
275 | -(int)(PAGE_SIZE / 1024)); | |
276 | ||
277 | memcg_kmem_uncharge(vm->pages[i], 0); | |
278 | } | |
279 | ||
ac496bf4 | 280 | for (i = 0; i < NR_CACHED_STACKS; i++) { |
112166f8 CL |
281 | if (this_cpu_cmpxchg(cached_stacks[i], |
282 | NULL, tsk->stack_vm_area) != NULL) | |
ac496bf4 AL |
283 | continue; |
284 | ||
ac496bf4 AL |
285 | return; |
286 | } | |
ac496bf4 | 287 | |
0f110a9b | 288 | vfree_atomic(tsk->stack); |
ac496bf4 AL |
289 | return; |
290 | } | |
291 | #endif | |
292 | ||
293 | __free_pages(virt_to_page(tsk->stack), THREAD_SIZE_ORDER); | |
b69c49b7 | 294 | } |
0d15d74a | 295 | # else |
b235beea | 296 | static struct kmem_cache *thread_stack_cache; |
0d15d74a | 297 | |
9521d399 | 298 | static unsigned long *alloc_thread_stack_node(struct task_struct *tsk, |
0d15d74a TG |
299 | int node) |
300 | { | |
5eed6f1d RR |
301 | unsigned long *stack; |
302 | stack = kmem_cache_alloc_node(thread_stack_cache, THREADINFO_GFP, node); | |
303 | tsk->stack = stack; | |
304 | return stack; | |
0d15d74a TG |
305 | } |
306 | ||
ba14a194 | 307 | static void free_thread_stack(struct task_struct *tsk) |
0d15d74a | 308 | { |
ba14a194 | 309 | kmem_cache_free(thread_stack_cache, tsk->stack); |
0d15d74a TG |
310 | } |
311 | ||
b235beea | 312 | void thread_stack_cache_init(void) |
0d15d74a | 313 | { |
f9d29946 DW |
314 | thread_stack_cache = kmem_cache_create_usercopy("thread_stack", |
315 | THREAD_SIZE, THREAD_SIZE, 0, 0, | |
316 | THREAD_SIZE, NULL); | |
b235beea | 317 | BUG_ON(thread_stack_cache == NULL); |
0d15d74a TG |
318 | } |
319 | # endif | |
b69c49b7 FT |
320 | #endif |
321 | ||
1da177e4 | 322 | /* SLAB cache for signal_struct structures (tsk->signal) */ |
e18b890b | 323 | static struct kmem_cache *signal_cachep; |
1da177e4 LT |
324 | |
325 | /* SLAB cache for sighand_struct structures (tsk->sighand) */ | |
e18b890b | 326 | struct kmem_cache *sighand_cachep; |
1da177e4 LT |
327 | |
328 | /* SLAB cache for files_struct structures (tsk->files) */ | |
e18b890b | 329 | struct kmem_cache *files_cachep; |
1da177e4 LT |
330 | |
331 | /* SLAB cache for fs_struct structures (tsk->fs) */ | |
e18b890b | 332 | struct kmem_cache *fs_cachep; |
1da177e4 LT |
333 | |
334 | /* SLAB cache for vm_area_struct structures */ | |
3928d4f5 | 335 | static struct kmem_cache *vm_area_cachep; |
1da177e4 LT |
336 | |
337 | /* SLAB cache for mm_struct structures (tsk->mm) */ | |
e18b890b | 338 | static struct kmem_cache *mm_cachep; |
1da177e4 | 339 | |
490fc053 | 340 | struct vm_area_struct *vm_area_alloc(struct mm_struct *mm) |
3928d4f5 | 341 | { |
a670468f | 342 | struct vm_area_struct *vma; |
490fc053 | 343 | |
a670468f | 344 | vma = kmem_cache_alloc(vm_area_cachep, GFP_KERNEL); |
027232da KS |
345 | if (vma) |
346 | vma_init(vma, mm); | |
490fc053 | 347 | return vma; |
3928d4f5 LT |
348 | } |
349 | ||
350 | struct vm_area_struct *vm_area_dup(struct vm_area_struct *orig) | |
351 | { | |
95faf699 LT |
352 | struct vm_area_struct *new = kmem_cache_alloc(vm_area_cachep, GFP_KERNEL); |
353 | ||
354 | if (new) { | |
355 | *new = *orig; | |
356 | INIT_LIST_HEAD(&new->anon_vma_chain); | |
357 | } | |
358 | return new; | |
3928d4f5 LT |
359 | } |
360 | ||
361 | void vm_area_free(struct vm_area_struct *vma) | |
362 | { | |
363 | kmem_cache_free(vm_area_cachep, vma); | |
364 | } | |
365 | ||
ba14a194 | 366 | static void account_kernel_stack(struct task_struct *tsk, int account) |
c6a7f572 | 367 | { |
ba14a194 AL |
368 | void *stack = task_stack_page(tsk); |
369 | struct vm_struct *vm = task_stack_vm_area(tsk); | |
370 | ||
371 | BUILD_BUG_ON(IS_ENABLED(CONFIG_VMAP_STACK) && PAGE_SIZE % 1024 != 0); | |
372 | ||
373 | if (vm) { | |
374 | int i; | |
375 | ||
376 | BUG_ON(vm->nr_pages != THREAD_SIZE / PAGE_SIZE); | |
377 | ||
378 | for (i = 0; i < THREAD_SIZE / PAGE_SIZE; i++) { | |
379 | mod_zone_page_state(page_zone(vm->pages[i]), | |
380 | NR_KERNEL_STACK_KB, | |
381 | PAGE_SIZE / 1024 * account); | |
382 | } | |
ba14a194 AL |
383 | } else { |
384 | /* | |
385 | * All stack pages are in the same zone and belong to the | |
386 | * same memcg. | |
387 | */ | |
388 | struct page *first_page = virt_to_page(stack); | |
389 | ||
390 | mod_zone_page_state(page_zone(first_page), NR_KERNEL_STACK_KB, | |
391 | THREAD_SIZE / 1024 * account); | |
392 | ||
ed52be7b JW |
393 | mod_memcg_page_state(first_page, MEMCG_KERNEL_STACK_KB, |
394 | account * (THREAD_SIZE / 1024)); | |
ba14a194 | 395 | } |
c6a7f572 KM |
396 | } |
397 | ||
9b6f7e16 RG |
398 | static int memcg_charge_kernel_stack(struct task_struct *tsk) |
399 | { | |
400 | #ifdef CONFIG_VMAP_STACK | |
401 | struct vm_struct *vm = task_stack_vm_area(tsk); | |
402 | int ret; | |
403 | ||
404 | if (vm) { | |
405 | int i; | |
406 | ||
407 | for (i = 0; i < THREAD_SIZE / PAGE_SIZE; i++) { | |
408 | /* | |
409 | * If memcg_kmem_charge() fails, page->mem_cgroup | |
410 | * pointer is NULL, and both memcg_kmem_uncharge() | |
411 | * and mod_memcg_page_state() in free_thread_stack() | |
412 | * will ignore this page. So it's safe. | |
413 | */ | |
414 | ret = memcg_kmem_charge(vm->pages[i], GFP_KERNEL, 0); | |
415 | if (ret) | |
416 | return ret; | |
417 | ||
418 | mod_memcg_page_state(vm->pages[i], | |
419 | MEMCG_KERNEL_STACK_KB, | |
420 | PAGE_SIZE / 1024); | |
421 | } | |
422 | } | |
423 | #endif | |
424 | return 0; | |
425 | } | |
426 | ||
68f24b08 | 427 | static void release_task_stack(struct task_struct *tsk) |
1da177e4 | 428 | { |
405c0759 AL |
429 | if (WARN_ON(tsk->state != TASK_DEAD)) |
430 | return; /* Better to leak the stack than to free prematurely */ | |
431 | ||
ba14a194 | 432 | account_kernel_stack(tsk, -1); |
ba14a194 | 433 | free_thread_stack(tsk); |
68f24b08 AL |
434 | tsk->stack = NULL; |
435 | #ifdef CONFIG_VMAP_STACK | |
436 | tsk->stack_vm_area = NULL; | |
437 | #endif | |
438 | } | |
439 | ||
440 | #ifdef CONFIG_THREAD_INFO_IN_TASK | |
441 | void put_task_stack(struct task_struct *tsk) | |
442 | { | |
f0b89d39 | 443 | if (refcount_dec_and_test(&tsk->stack_refcount)) |
68f24b08 AL |
444 | release_task_stack(tsk); |
445 | } | |
446 | #endif | |
447 | ||
448 | void free_task(struct task_struct *tsk) | |
449 | { | |
450 | #ifndef CONFIG_THREAD_INFO_IN_TASK | |
451 | /* | |
452 | * The task is finally done with both the stack and thread_info, | |
453 | * so free both. | |
454 | */ | |
455 | release_task_stack(tsk); | |
456 | #else | |
457 | /* | |
458 | * If the task had a separate stack allocation, it should be gone | |
459 | * by now. | |
460 | */ | |
f0b89d39 | 461 | WARN_ON_ONCE(refcount_read(&tsk->stack_refcount) != 0); |
68f24b08 | 462 | #endif |
23f78d4a | 463 | rt_mutex_debug_task_free(tsk); |
fb52607a | 464 | ftrace_graph_exit_task(tsk); |
e2cfabdf | 465 | put_seccomp_filter(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(); | |
484 | if (down_write_killable(&oldmm->mmap_sem)) { | |
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 | */ | |
493 | down_write_nested(&mm->mmap_sem, SINGLE_DEPTH_NESTING); | |
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) { | |
548 | /* VM_WIPEONFORK gets a clean slate in the child. */ | |
549 | tmp->anon_vma = NULL; | |
550 | if (anon_vma_prepare(tmp)) | |
551 | goto fail_nomem_anon_vma_fork; | |
552 | } else if (anon_vma_fork(tmp, mpnt)) | |
553 | goto fail_nomem_anon_vma_fork; | |
554 | tmp->vm_flags &= ~(VM_LOCKED | VM_LOCKONFAULT); | |
555 | tmp->vm_next = tmp->vm_prev = NULL; | |
556 | file = tmp->vm_file; | |
557 | if (file) { | |
558 | struct inode *inode = file_inode(file); | |
559 | struct address_space *mapping = file->f_mapping; | |
560 | ||
8994cce9 | 561 | vma_get_file(tmp); |
d70f2a14 AM |
562 | if (tmp->vm_flags & VM_DENYWRITE) |
563 | atomic_dec(&inode->i_writecount); | |
564 | i_mmap_lock_write(mapping); | |
565 | if (tmp->vm_flags & VM_SHARED) | |
566 | atomic_inc(&mapping->i_mmap_writable); | |
567 | flush_dcache_mmap_lock(mapping); | |
568 | /* insert tmp into the share list, just after mpnt */ | |
569 | vma_interval_tree_insert_after(tmp, mpnt, | |
570 | &mapping->i_mmap); | |
571 | flush_dcache_mmap_unlock(mapping); | |
572 | i_mmap_unlock_write(mapping); | |
573 | } | |
574 | ||
575 | /* | |
576 | * Clear hugetlb-related page reserves for children. This only | |
577 | * affects MAP_PRIVATE mappings. Faults generated by the child | |
578 | * are not guaranteed to succeed, even if read-only | |
579 | */ | |
580 | if (is_vm_hugetlb_page(tmp)) | |
581 | reset_vma_resv_huge_pages(tmp); | |
582 | ||
583 | /* | |
584 | * Link in the new vma and copy the page table entries. | |
585 | */ | |
586 | *pprev = tmp; | |
587 | pprev = &tmp->vm_next; | |
588 | tmp->vm_prev = prev; | |
589 | prev = tmp; | |
590 | ||
591 | __vma_link_rb(mm, tmp, rb_link, rb_parent); | |
592 | rb_link = &tmp->vm_rb.rb_right; | |
593 | rb_parent = &tmp->vm_rb; | |
594 | ||
595 | mm->map_count++; | |
596 | if (!(tmp->vm_flags & VM_WIPEONFORK)) | |
597 | retval = copy_page_range(mm, oldmm, mpnt); | |
598 | ||
599 | if (tmp->vm_ops && tmp->vm_ops->open) | |
600 | tmp->vm_ops->open(tmp); | |
601 | ||
602 | if (retval) | |
603 | goto out; | |
604 | } | |
605 | /* a new mm has just been created */ | |
1ed0cc5a | 606 | retval = arch_dup_mmap(oldmm, mm); |
d70f2a14 AM |
607 | out: |
608 | up_write(&mm->mmap_sem); | |
609 | flush_tlb_mm(oldmm); | |
610 | up_write(&oldmm->mmap_sem); | |
611 | dup_userfaultfd_complete(&uf); | |
612 | fail_uprobe_end: | |
613 | uprobe_end_dup_mmap(); | |
614 | return retval; | |
615 | fail_nomem_anon_vma_fork: | |
616 | mpol_put(vma_policy(tmp)); | |
617 | fail_nomem_policy: | |
3928d4f5 | 618 | vm_area_free(tmp); |
d70f2a14 AM |
619 | fail_nomem: |
620 | retval = -ENOMEM; | |
621 | vm_unacct_memory(charge); | |
622 | goto out; | |
623 | } | |
624 | ||
625 | static inline int mm_alloc_pgd(struct mm_struct *mm) | |
626 | { | |
627 | mm->pgd = pgd_alloc(mm); | |
628 | if (unlikely(!mm->pgd)) | |
629 | return -ENOMEM; | |
630 | return 0; | |
631 | } | |
632 | ||
633 | static inline void mm_free_pgd(struct mm_struct *mm) | |
634 | { | |
635 | pgd_free(mm, mm->pgd); | |
636 | } | |
637 | #else | |
638 | static int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm) | |
639 | { | |
640 | down_write(&oldmm->mmap_sem); | |
641 | RCU_INIT_POINTER(mm->exe_file, get_mm_exe_file(oldmm)); | |
642 | up_write(&oldmm->mmap_sem); | |
643 | return 0; | |
644 | } | |
645 | #define mm_alloc_pgd(mm) (0) | |
646 | #define mm_free_pgd(mm) | |
647 | #endif /* CONFIG_MMU */ | |
648 | ||
649 | static void check_mm(struct mm_struct *mm) | |
650 | { | |
651 | int i; | |
652 | ||
653 | for (i = 0; i < NR_MM_COUNTERS; i++) { | |
654 | long x = atomic_long_read(&mm->rss_stat.count[i]); | |
655 | ||
656 | if (unlikely(x)) | |
657 | printk(KERN_ALERT "BUG: Bad rss-counter state " | |
658 | "mm:%p idx:%d val:%ld\n", mm, i, x); | |
659 | } | |
660 | ||
661 | if (mm_pgtables_bytes(mm)) | |
662 | pr_alert("BUG: non-zero pgtables_bytes on freeing mm: %ld\n", | |
663 | mm_pgtables_bytes(mm)); | |
664 | ||
665 | #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS | |
666 | VM_BUG_ON_MM(mm->pmd_huge_pte, mm); | |
667 | #endif | |
668 | } | |
669 | ||
670 | #define allocate_mm() (kmem_cache_alloc(mm_cachep, GFP_KERNEL)) | |
671 | #define free_mm(mm) (kmem_cache_free(mm_cachep, (mm))) | |
672 | ||
673 | /* | |
674 | * Called when the last reference to the mm | |
675 | * is dropped: either by a lazy thread or by | |
676 | * mmput. Free the page directory and the mm. | |
677 | */ | |
d34bc48f | 678 | void __mmdrop(struct mm_struct *mm) |
d70f2a14 AM |
679 | { |
680 | BUG_ON(mm == &init_mm); | |
3eda69c9 MR |
681 | WARN_ON_ONCE(mm == current->mm); |
682 | WARN_ON_ONCE(mm == current->active_mm); | |
d70f2a14 AM |
683 | mm_free_pgd(mm); |
684 | destroy_context(mm); | |
d70f2a14 AM |
685 | mmu_notifier_mm_destroy(mm); |
686 | check_mm(mm); | |
687 | put_user_ns(mm->user_ns); | |
688 | free_mm(mm); | |
689 | } | |
d34bc48f | 690 | EXPORT_SYMBOL_GPL(__mmdrop); |
d70f2a14 AM |
691 | |
692 | static void mmdrop_async_fn(struct work_struct *work) | |
693 | { | |
694 | struct mm_struct *mm; | |
695 | ||
696 | mm = container_of(work, struct mm_struct, async_put_work); | |
697 | __mmdrop(mm); | |
698 | } | |
699 | ||
700 | static void mmdrop_async(struct mm_struct *mm) | |
701 | { | |
702 | if (unlikely(atomic_dec_and_test(&mm->mm_count))) { | |
703 | INIT_WORK(&mm->async_put_work, mmdrop_async_fn); | |
704 | schedule_work(&mm->async_put_work); | |
705 | } | |
706 | } | |
707 | ||
ea6d290c ON |
708 | static inline void free_signal_struct(struct signal_struct *sig) |
709 | { | |
97101eb4 | 710 | taskstats_tgid_free(sig); |
1c5354de | 711 | sched_autogroup_exit(sig); |
7283094e MH |
712 | /* |
713 | * __mmdrop is not safe to call from softirq context on x86 due to | |
714 | * pgd_dtor so postpone it to the async context | |
715 | */ | |
26db62f1 | 716 | if (sig->oom_mm) |
7283094e | 717 | mmdrop_async(sig->oom_mm); |
ea6d290c ON |
718 | kmem_cache_free(signal_cachep, sig); |
719 | } | |
720 | ||
721 | static inline void put_signal_struct(struct signal_struct *sig) | |
722 | { | |
60d4de3f | 723 | if (refcount_dec_and_test(&sig->sigcnt)) |
ea6d290c ON |
724 | free_signal_struct(sig); |
725 | } | |
726 | ||
158d9ebd | 727 | void __put_task_struct(struct task_struct *tsk) |
1da177e4 | 728 | { |
270f722d | 729 | WARN_ON(!tsk->exit_state); |
ec1d2819 | 730 | WARN_ON(refcount_read(&tsk->usage)); |
1da177e4 LT |
731 | WARN_ON(tsk == current); |
732 | ||
2e91fa7f | 733 | cgroup_free(tsk); |
16d51a59 | 734 | task_numa_free(tsk, true); |
1a2a4d06 | 735 | security_task_free(tsk); |
e0e81739 | 736 | exit_creds(tsk); |
35df17c5 | 737 | delayacct_tsk_free(tsk); |
ea6d290c | 738 | put_signal_struct(tsk->signal); |
1da177e4 LT |
739 | |
740 | if (!profile_handoff_task(tsk)) | |
741 | free_task(tsk); | |
742 | } | |
77c100c8 | 743 | EXPORT_SYMBOL_GPL(__put_task_struct); |
1da177e4 | 744 | |
6c0a9fa6 | 745 | void __init __weak arch_task_cache_init(void) { } |
61c4628b | 746 | |
ff691f6e HS |
747 | /* |
748 | * set_max_threads | |
749 | */ | |
16db3d3f | 750 | static void set_max_threads(unsigned int max_threads_suggested) |
ff691f6e | 751 | { |
ac1b398d | 752 | u64 threads; |
ca79b0c2 | 753 | unsigned long nr_pages = totalram_pages(); |
ff691f6e HS |
754 | |
755 | /* | |
ac1b398d HS |
756 | * The number of threads shall be limited such that the thread |
757 | * structures may only consume a small part of the available memory. | |
ff691f6e | 758 | */ |
3d6357de | 759 | if (fls64(nr_pages) + fls64(PAGE_SIZE) > 64) |
ac1b398d HS |
760 | threads = MAX_THREADS; |
761 | else | |
3d6357de | 762 | threads = div64_u64((u64) nr_pages * (u64) PAGE_SIZE, |
ac1b398d HS |
763 | (u64) THREAD_SIZE * 8UL); |
764 | ||
16db3d3f HS |
765 | if (threads > max_threads_suggested) |
766 | threads = max_threads_suggested; | |
767 | ||
ac1b398d | 768 | max_threads = clamp_t(u64, threads, MIN_THREADS, MAX_THREADS); |
ff691f6e HS |
769 | } |
770 | ||
5aaeb5c0 IM |
771 | #ifdef CONFIG_ARCH_WANTS_DYNAMIC_TASK_STRUCT |
772 | /* Initialized by the architecture: */ | |
773 | int arch_task_struct_size __read_mostly; | |
774 | #endif | |
0c8c0f03 | 775 | |
5905429a KC |
776 | static void task_struct_whitelist(unsigned long *offset, unsigned long *size) |
777 | { | |
778 | /* Fetch thread_struct whitelist for the architecture. */ | |
779 | arch_thread_struct_whitelist(offset, size); | |
780 | ||
781 | /* | |
782 | * Handle zero-sized whitelist or empty thread_struct, otherwise | |
783 | * adjust offset to position of thread_struct in task_struct. | |
784 | */ | |
785 | if (unlikely(*size == 0)) | |
786 | *offset = 0; | |
787 | else | |
788 | *offset += offsetof(struct task_struct, thread); | |
789 | } | |
790 | ||
ff691f6e | 791 | void __init fork_init(void) |
1da177e4 | 792 | { |
25f9c081 | 793 | int i; |
f5e10287 | 794 | #ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR |
1da177e4 | 795 | #ifndef ARCH_MIN_TASKALIGN |
e274795e | 796 | #define ARCH_MIN_TASKALIGN 0 |
1da177e4 | 797 | #endif |
95cb64c1 | 798 | int align = max_t(int, L1_CACHE_BYTES, ARCH_MIN_TASKALIGN); |
5905429a | 799 | unsigned long useroffset, usersize; |
e274795e | 800 | |
1da177e4 | 801 | /* create a slab on which task_structs can be allocated */ |
5905429a KC |
802 | task_struct_whitelist(&useroffset, &usersize); |
803 | task_struct_cachep = kmem_cache_create_usercopy("task_struct", | |
e274795e | 804 | arch_task_struct_size, align, |
5905429a KC |
805 | SLAB_PANIC|SLAB_ACCOUNT, |
806 | useroffset, usersize, NULL); | |
1da177e4 LT |
807 | #endif |
808 | ||
61c4628b SS |
809 | /* do the arch specific task caches init */ |
810 | arch_task_cache_init(); | |
811 | ||
16db3d3f | 812 | set_max_threads(MAX_THREADS); |
1da177e4 LT |
813 | |
814 | init_task.signal->rlim[RLIMIT_NPROC].rlim_cur = max_threads/2; | |
815 | init_task.signal->rlim[RLIMIT_NPROC].rlim_max = max_threads/2; | |
816 | init_task.signal->rlim[RLIMIT_SIGPENDING] = | |
817 | init_task.signal->rlim[RLIMIT_NPROC]; | |
b376c3e1 | 818 | |
25f9c081 EB |
819 | for (i = 0; i < UCOUNT_COUNTS; i++) { |
820 | init_user_ns.ucount_max[i] = max_threads/2; | |
821 | } | |
19659c59 HR |
822 | |
823 | #ifdef CONFIG_VMAP_STACK | |
824 | cpuhp_setup_state(CPUHP_BP_PREPARE_DYN, "fork:vm_stack_cache", | |
825 | NULL, free_vm_stack_cache); | |
826 | #endif | |
b09be676 BP |
827 | |
828 | lockdep_init_task(&init_task); | |
aad42dd4 | 829 | uprobes_init(); |
1da177e4 LT |
830 | } |
831 | ||
52f5684c | 832 | int __weak arch_dup_task_struct(struct task_struct *dst, |
61c4628b SS |
833 | struct task_struct *src) |
834 | { | |
835 | *dst = *src; | |
836 | return 0; | |
837 | } | |
838 | ||
d4311ff1 AT |
839 | void set_task_stack_end_magic(struct task_struct *tsk) |
840 | { | |
841 | unsigned long *stackend; | |
842 | ||
843 | stackend = end_of_stack(tsk); | |
844 | *stackend = STACK_END_MAGIC; /* for overflow detection */ | |
845 | } | |
846 | ||
725fc629 | 847 | static struct task_struct *dup_task_struct(struct task_struct *orig, int node) |
1da177e4 LT |
848 | { |
849 | struct task_struct *tsk; | |
b235beea | 850 | unsigned long *stack; |
0f4991e8 | 851 | struct vm_struct *stack_vm_area __maybe_unused; |
3e26c149 | 852 | int err; |
1da177e4 | 853 | |
725fc629 AK |
854 | if (node == NUMA_NO_NODE) |
855 | node = tsk_fork_get_node(orig); | |
504f52b5 | 856 | tsk = alloc_task_struct_node(node); |
1da177e4 LT |
857 | if (!tsk) |
858 | return NULL; | |
859 | ||
b235beea LT |
860 | stack = alloc_thread_stack_node(tsk, node); |
861 | if (!stack) | |
f19b9f74 | 862 | goto free_tsk; |
1da177e4 | 863 | |
9b6f7e16 RG |
864 | if (memcg_charge_kernel_stack(tsk)) |
865 | goto free_stack; | |
866 | ||
ba14a194 AL |
867 | stack_vm_area = task_stack_vm_area(tsk); |
868 | ||
fb0a685c | 869 | err = arch_dup_task_struct(tsk, orig); |
ba14a194 AL |
870 | |
871 | /* | |
872 | * arch_dup_task_struct() clobbers the stack-related fields. Make | |
873 | * sure they're properly initialized before using any stack-related | |
874 | * functions again. | |
875 | */ | |
876 | tsk->stack = stack; | |
877 | #ifdef CONFIG_VMAP_STACK | |
878 | tsk->stack_vm_area = stack_vm_area; | |
879 | #endif | |
68f24b08 | 880 | #ifdef CONFIG_THREAD_INFO_IN_TASK |
f0b89d39 | 881 | refcount_set(&tsk->stack_refcount, 1); |
68f24b08 | 882 | #endif |
ba14a194 | 883 | |
164c33c6 | 884 | if (err) |
b235beea | 885 | goto free_stack; |
164c33c6 | 886 | |
dbd95212 KC |
887 | #ifdef CONFIG_SECCOMP |
888 | /* | |
889 | * We must handle setting up seccomp filters once we're under | |
890 | * the sighand lock in case orig has changed between now and | |
891 | * then. Until then, filter must be NULL to avoid messing up | |
892 | * the usage counts on the error path calling free_task. | |
893 | */ | |
894 | tsk->seccomp.filter = NULL; | |
895 | #endif | |
87bec58a AM |
896 | |
897 | setup_thread_stack(tsk, orig); | |
8e7cac79 | 898 | clear_user_return_notifier(tsk); |
f26f9aff | 899 | clear_tsk_need_resched(tsk); |
d4311ff1 | 900 | set_task_stack_end_magic(tsk); |
1da177e4 | 901 | |
050e9baa | 902 | #ifdef CONFIG_STACKPROTECTOR |
7cd815bc | 903 | tsk->stack_canary = get_random_canary(); |
0a425405 | 904 | #endif |
3bd37062 SAS |
905 | if (orig->cpus_ptr == &orig->cpus_mask) |
906 | tsk->cpus_ptr = &tsk->cpus_mask; | |
0a425405 | 907 | |
fb0a685c DRO |
908 | /* |
909 | * One for us, one for whoever does the "release_task()" (usually | |
910 | * parent) | |
911 | */ | |
ec1d2819 | 912 | refcount_set(&tsk->usage, 2); |
6c5c9341 | 913 | #ifdef CONFIG_BLK_DEV_IO_TRACE |
2056a782 | 914 | tsk->btrace_seq = 0; |
6c5c9341 | 915 | #endif |
a0aa7f68 | 916 | tsk->splice_pipe = NULL; |
5640f768 | 917 | tsk->task_frag.page = NULL; |
093e5840 | 918 | tsk->wake_q.next = NULL; |
c6a7f572 | 919 | |
ba14a194 | 920 | account_kernel_stack(tsk, 1); |
c6a7f572 | 921 | |
5c9a8750 DV |
922 | kcov_task_init(tsk); |
923 | ||
e41d5818 DV |
924 | #ifdef CONFIG_FAULT_INJECTION |
925 | tsk->fail_nth = 0; | |
926 | #endif | |
927 | ||
2c323017 JB |
928 | #ifdef CONFIG_BLK_CGROUP |
929 | tsk->throttle_queue = NULL; | |
930 | tsk->use_memdelay = 0; | |
931 | #endif | |
932 | ||
d46eb14b SB |
933 | #ifdef CONFIG_MEMCG |
934 | tsk->active_memcg = NULL; | |
935 | #endif | |
1da177e4 | 936 | return tsk; |
61c4628b | 937 | |
b235beea | 938 | free_stack: |
ba14a194 | 939 | free_thread_stack(tsk); |
f19b9f74 | 940 | free_tsk: |
61c4628b SS |
941 | free_task_struct(tsk); |
942 | return NULL; | |
1da177e4 LT |
943 | } |
944 | ||
23ff4440 | 945 | __cacheline_aligned_in_smp DEFINE_SPINLOCK(mmlist_lock); |
1da177e4 | 946 | |
4cb0e11b HK |
947 | static unsigned long default_dump_filter = MMF_DUMP_FILTER_DEFAULT; |
948 | ||
949 | static int __init coredump_filter_setup(char *s) | |
950 | { | |
951 | default_dump_filter = | |
952 | (simple_strtoul(s, NULL, 0) << MMF_DUMP_FILTER_SHIFT) & | |
953 | MMF_DUMP_FILTER_MASK; | |
954 | return 1; | |
955 | } | |
956 | ||
957 | __setup("coredump_filter=", coredump_filter_setup); | |
958 | ||
1da177e4 LT |
959 | #include <linux/init_task.h> |
960 | ||
858f0993 AD |
961 | static void mm_init_aio(struct mm_struct *mm) |
962 | { | |
963 | #ifdef CONFIG_AIO | |
964 | spin_lock_init(&mm->ioctx_lock); | |
db446a08 | 965 | mm->ioctx_table = NULL; |
858f0993 AD |
966 | #endif |
967 | } | |
968 | ||
c3f3ce04 AA |
969 | static __always_inline void mm_clear_owner(struct mm_struct *mm, |
970 | struct task_struct *p) | |
971 | { | |
972 | #ifdef CONFIG_MEMCG | |
973 | if (mm->owner == p) | |
974 | WRITE_ONCE(mm->owner, NULL); | |
975 | #endif | |
976 | } | |
977 | ||
33144e84 VD |
978 | static void mm_init_owner(struct mm_struct *mm, struct task_struct *p) |
979 | { | |
980 | #ifdef CONFIG_MEMCG | |
981 | mm->owner = p; | |
982 | #endif | |
983 | } | |
984 | ||
355627f5 EB |
985 | static void mm_init_uprobes_state(struct mm_struct *mm) |
986 | { | |
987 | #ifdef CONFIG_UPROBES | |
988 | mm->uprobes_state.xol_area = NULL; | |
989 | #endif | |
990 | } | |
991 | ||
bfedb589 EB |
992 | static struct mm_struct *mm_init(struct mm_struct *mm, struct task_struct *p, |
993 | struct user_namespace *user_ns) | |
1da177e4 | 994 | { |
41f727fd VD |
995 | mm->mmap = NULL; |
996 | mm->mm_rb = RB_ROOT; | |
997 | mm->vmacache_seqnum = 0; | |
1da177e4 LT |
998 | atomic_set(&mm->mm_users, 1); |
999 | atomic_set(&mm->mm_count, 1); | |
1000 | init_rwsem(&mm->mmap_sem); | |
1001 | INIT_LIST_HEAD(&mm->mmlist); | |
999d9fc1 | 1002 | mm->core_state = NULL; |
af5b0f6a | 1003 | mm_pgtables_bytes_init(mm); |
41f727fd VD |
1004 | mm->map_count = 0; |
1005 | mm->locked_vm = 0; | |
70f8a3ca | 1006 | atomic64_set(&mm->pinned_vm, 0); |
d559db08 | 1007 | memset(&mm->rss_stat, 0, sizeof(mm->rss_stat)); |
1da177e4 | 1008 | spin_lock_init(&mm->page_table_lock); |
88aa7cc6 | 1009 | spin_lock_init(&mm->arg_lock); |
41f727fd | 1010 | mm_init_cpumask(mm); |
858f0993 | 1011 | mm_init_aio(mm); |
cf475ad2 | 1012 | mm_init_owner(mm, p); |
2b7e8665 | 1013 | RCU_INIT_POINTER(mm->exe_file, NULL); |
41f727fd | 1014 | mmu_notifier_mm_init(mm); |
133ff0ea | 1015 | hmm_mm_init(mm); |
16af97dc | 1016 | init_tlb_flush_pending(mm); |
41f727fd VD |
1017 | #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS |
1018 | mm->pmd_huge_pte = NULL; | |
1019 | #endif | |
355627f5 | 1020 | mm_init_uprobes_state(mm); |
1da177e4 | 1021 | |
a0715cc2 AT |
1022 | if (current->mm) { |
1023 | mm->flags = current->mm->flags & MMF_INIT_MASK; | |
1024 | mm->def_flags = current->mm->def_flags & VM_INIT_DEF_MASK; | |
1025 | } else { | |
1026 | mm->flags = default_dump_filter; | |
1da177e4 | 1027 | mm->def_flags = 0; |
a0715cc2 AT |
1028 | } |
1029 | ||
41f727fd VD |
1030 | if (mm_alloc_pgd(mm)) |
1031 | goto fail_nopgd; | |
1032 | ||
1033 | if (init_new_context(p, mm)) | |
1034 | goto fail_nocontext; | |
78fb7466 | 1035 | |
bfedb589 | 1036 | mm->user_ns = get_user_ns(user_ns); |
41f727fd VD |
1037 | return mm; |
1038 | ||
1039 | fail_nocontext: | |
1040 | mm_free_pgd(mm); | |
1041 | fail_nopgd: | |
1da177e4 LT |
1042 | free_mm(mm); |
1043 | return NULL; | |
1044 | } | |
1045 | ||
1046 | /* | |
1047 | * Allocate and initialize an mm_struct. | |
1048 | */ | |
fb0a685c | 1049 | struct mm_struct *mm_alloc(void) |
1da177e4 | 1050 | { |
fb0a685c | 1051 | struct mm_struct *mm; |
1da177e4 LT |
1052 | |
1053 | mm = allocate_mm(); | |
de03c72c KM |
1054 | if (!mm) |
1055 | return NULL; | |
1056 | ||
1057 | memset(mm, 0, sizeof(*mm)); | |
bfedb589 | 1058 | return mm_init(mm, current, current_user_ns()); |
1da177e4 LT |
1059 | } |
1060 | ||
ec8d7c14 MH |
1061 | static inline void __mmput(struct mm_struct *mm) |
1062 | { | |
1063 | VM_BUG_ON(atomic_read(&mm->mm_users)); | |
1064 | ||
1065 | uprobe_clear_state(mm); | |
1066 | exit_aio(mm); | |
1067 | ksm_exit(mm); | |
1068 | khugepaged_exit(mm); /* must run before exit_mmap */ | |
1069 | exit_mmap(mm); | |
6fcb52a5 | 1070 | mm_put_huge_zero_page(mm); |
ec8d7c14 MH |
1071 | set_mm_exe_file(mm, NULL); |
1072 | if (!list_empty(&mm->mmlist)) { | |
1073 | spin_lock(&mmlist_lock); | |
1074 | list_del(&mm->mmlist); | |
1075 | spin_unlock(&mmlist_lock); | |
1076 | } | |
1077 | if (mm->binfmt) | |
1078 | module_put(mm->binfmt->module); | |
1079 | mmdrop(mm); | |
1080 | } | |
1081 | ||
1da177e4 LT |
1082 | /* |
1083 | * Decrement the use count and release all resources for an mm. | |
1084 | */ | |
1085 | void mmput(struct mm_struct *mm) | |
1086 | { | |
0ae26f1b AM |
1087 | might_sleep(); |
1088 | ||
ec8d7c14 MH |
1089 | if (atomic_dec_and_test(&mm->mm_users)) |
1090 | __mmput(mm); | |
1091 | } | |
1092 | EXPORT_SYMBOL_GPL(mmput); | |
1093 | ||
a1b2289c SY |
1094 | #ifdef CONFIG_MMU |
1095 | static void mmput_async_fn(struct work_struct *work) | |
1096 | { | |
1097 | struct mm_struct *mm = container_of(work, struct mm_struct, | |
1098 | async_put_work); | |
1099 | ||
1100 | __mmput(mm); | |
1101 | } | |
1102 | ||
1103 | void mmput_async(struct mm_struct *mm) | |
1104 | { | |
1105 | if (atomic_dec_and_test(&mm->mm_users)) { | |
1106 | INIT_WORK(&mm->async_put_work, mmput_async_fn); | |
1107 | schedule_work(&mm->async_put_work); | |
1108 | } | |
1109 | } | |
1de76885 | 1110 | EXPORT_SYMBOL(mmput_async); |
a1b2289c SY |
1111 | #endif |
1112 | ||
90f31d0e KK |
1113 | /** |
1114 | * set_mm_exe_file - change a reference to the mm's executable file | |
1115 | * | |
1116 | * This changes mm's executable file (shown as symlink /proc/[pid]/exe). | |
1117 | * | |
6e399cd1 DB |
1118 | * Main users are mmput() and sys_execve(). Callers prevent concurrent |
1119 | * invocations: in mmput() nobody alive left, in execve task is single | |
1120 | * threaded. sys_prctl(PR_SET_MM_MAP/EXE_FILE) also needs to set the | |
1121 | * mm->exe_file, but does so without using set_mm_exe_file() in order | |
1122 | * to do avoid the need for any locks. | |
90f31d0e | 1123 | */ |
38646013 JS |
1124 | void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file) |
1125 | { | |
6e399cd1 DB |
1126 | struct file *old_exe_file; |
1127 | ||
1128 | /* | |
1129 | * It is safe to dereference the exe_file without RCU as | |
1130 | * this function is only called if nobody else can access | |
1131 | * this mm -- see comment above for justification. | |
1132 | */ | |
1133 | old_exe_file = rcu_dereference_raw(mm->exe_file); | |
90f31d0e | 1134 | |
38646013 JS |
1135 | if (new_exe_file) |
1136 | get_file(new_exe_file); | |
90f31d0e KK |
1137 | rcu_assign_pointer(mm->exe_file, new_exe_file); |
1138 | if (old_exe_file) | |
1139 | fput(old_exe_file); | |
38646013 JS |
1140 | } |
1141 | ||
90f31d0e KK |
1142 | /** |
1143 | * get_mm_exe_file - acquire a reference to the mm's executable file | |
1144 | * | |
1145 | * Returns %NULL if mm has no associated executable file. | |
1146 | * User must release file via fput(). | |
1147 | */ | |
38646013 JS |
1148 | struct file *get_mm_exe_file(struct mm_struct *mm) |
1149 | { | |
1150 | struct file *exe_file; | |
1151 | ||
90f31d0e KK |
1152 | rcu_read_lock(); |
1153 | exe_file = rcu_dereference(mm->exe_file); | |
1154 | if (exe_file && !get_file_rcu(exe_file)) | |
1155 | exe_file = NULL; | |
1156 | rcu_read_unlock(); | |
38646013 JS |
1157 | return exe_file; |
1158 | } | |
11163348 | 1159 | EXPORT_SYMBOL(get_mm_exe_file); |
38646013 | 1160 | |
cd81a917 MG |
1161 | /** |
1162 | * get_task_exe_file - acquire a reference to the task's executable file | |
1163 | * | |
1164 | * Returns %NULL if task's mm (if any) has no associated executable file or | |
1165 | * this is a kernel thread with borrowed mm (see the comment above get_task_mm). | |
1166 | * User must release file via fput(). | |
1167 | */ | |
1168 | struct file *get_task_exe_file(struct task_struct *task) | |
1169 | { | |
1170 | struct file *exe_file = NULL; | |
1171 | struct mm_struct *mm; | |
1172 | ||
1173 | task_lock(task); | |
1174 | mm = task->mm; | |
1175 | if (mm) { | |
1176 | if (!(task->flags & PF_KTHREAD)) | |
1177 | exe_file = get_mm_exe_file(mm); | |
1178 | } | |
1179 | task_unlock(task); | |
1180 | return exe_file; | |
1181 | } | |
1182 | EXPORT_SYMBOL(get_task_exe_file); | |
38646013 | 1183 | |
1da177e4 LT |
1184 | /** |
1185 | * get_task_mm - acquire a reference to the task's mm | |
1186 | * | |
246bb0b1 | 1187 | * Returns %NULL if the task has no mm. Checks PF_KTHREAD (meaning |
1da177e4 LT |
1188 | * this kernel workthread has transiently adopted a user mm with use_mm, |
1189 | * to do its AIO) is not set and if so returns a reference to it, after | |
1190 | * bumping up the use count. User must release the mm via mmput() | |
1191 | * after use. Typically used by /proc and ptrace. | |
1192 | */ | |
1193 | struct mm_struct *get_task_mm(struct task_struct *task) | |
1194 | { | |
1195 | struct mm_struct *mm; | |
1196 | ||
1197 | task_lock(task); | |
1198 | mm = task->mm; | |
1199 | if (mm) { | |
246bb0b1 | 1200 | if (task->flags & PF_KTHREAD) |
1da177e4 LT |
1201 | mm = NULL; |
1202 | else | |
3fce371b | 1203 | mmget(mm); |
1da177e4 LT |
1204 | } |
1205 | task_unlock(task); | |
1206 | return mm; | |
1207 | } | |
1208 | EXPORT_SYMBOL_GPL(get_task_mm); | |
1209 | ||
8cdb878d CY |
1210 | struct mm_struct *mm_access(struct task_struct *task, unsigned int mode) |
1211 | { | |
1212 | struct mm_struct *mm; | |
1213 | int err; | |
1214 | ||
1215 | err = mutex_lock_killable(&task->signal->cred_guard_mutex); | |
1216 | if (err) | |
1217 | return ERR_PTR(err); | |
1218 | ||
1219 | mm = get_task_mm(task); | |
1220 | if (mm && mm != current->mm && | |
1221 | !ptrace_may_access(task, mode)) { | |
1222 | mmput(mm); | |
1223 | mm = ERR_PTR(-EACCES); | |
1224 | } | |
1225 | mutex_unlock(&task->signal->cred_guard_mutex); | |
1226 | ||
1227 | return mm; | |
1228 | } | |
1229 | ||
57b59c4a | 1230 | static void complete_vfork_done(struct task_struct *tsk) |
c415c3b4 | 1231 | { |
d68b46fe | 1232 | struct completion *vfork; |
c415c3b4 | 1233 | |
d68b46fe ON |
1234 | task_lock(tsk); |
1235 | vfork = tsk->vfork_done; | |
1236 | if (likely(vfork)) { | |
1237 | tsk->vfork_done = NULL; | |
1238 | complete(vfork); | |
1239 | } | |
1240 | task_unlock(tsk); | |
1241 | } | |
1242 | ||
1243 | static int wait_for_vfork_done(struct task_struct *child, | |
1244 | struct completion *vfork) | |
1245 | { | |
1246 | int killed; | |
1247 | ||
1248 | freezer_do_not_count(); | |
76f969e8 | 1249 | cgroup_enter_frozen(); |
d68b46fe | 1250 | killed = wait_for_completion_killable(vfork); |
76f969e8 | 1251 | cgroup_leave_frozen(false); |
d68b46fe ON |
1252 | freezer_count(); |
1253 | ||
1254 | if (killed) { | |
1255 | task_lock(child); | |
1256 | child->vfork_done = NULL; | |
1257 | task_unlock(child); | |
1258 | } | |
1259 | ||
1260 | put_task_struct(child); | |
1261 | return killed; | |
c415c3b4 ON |
1262 | } |
1263 | ||
1da177e4 LT |
1264 | /* Please note the differences between mmput and mm_release. |
1265 | * mmput is called whenever we stop holding onto a mm_struct, | |
1266 | * error success whatever. | |
1267 | * | |
1268 | * mm_release is called after a mm_struct has been removed | |
1269 | * from the current process. | |
1270 | * | |
1271 | * This difference is important for error handling, when we | |
1272 | * only half set up a mm_struct for a new process and need to restore | |
1273 | * the old one. Because we mmput the new mm_struct before | |
1274 | * restoring the old one. . . | |
1275 | * Eric Biederman 10 January 1998 | |
1276 | */ | |
1277 | void mm_release(struct task_struct *tsk, struct mm_struct *mm) | |
1278 | { | |
8141c7f3 LT |
1279 | /* Get rid of any futexes when releasing the mm */ |
1280 | #ifdef CONFIG_FUTEX | |
fc6b177d | 1281 | if (unlikely(tsk->robust_list)) { |
8141c7f3 | 1282 | exit_robust_list(tsk); |
fc6b177d PZ |
1283 | tsk->robust_list = NULL; |
1284 | } | |
8141c7f3 | 1285 | #ifdef CONFIG_COMPAT |
fc6b177d | 1286 | if (unlikely(tsk->compat_robust_list)) { |
8141c7f3 | 1287 | compat_exit_robust_list(tsk); |
fc6b177d PZ |
1288 | tsk->compat_robust_list = NULL; |
1289 | } | |
8141c7f3 | 1290 | #endif |
322a2c10 TG |
1291 | if (unlikely(!list_empty(&tsk->pi_state_list))) |
1292 | exit_pi_state_list(tsk); | |
8141c7f3 LT |
1293 | #endif |
1294 | ||
0326f5a9 SD |
1295 | uprobe_free_utask(tsk); |
1296 | ||
1da177e4 LT |
1297 | /* Get rid of any cached register state */ |
1298 | deactivate_mm(tsk, mm); | |
1299 | ||
fec1d011 | 1300 | /* |
735f2770 MH |
1301 | * Signal userspace if we're not exiting with a core dump |
1302 | * because we want to leave the value intact for debugging | |
1303 | * purposes. | |
fec1d011 | 1304 | */ |
9c8a8228 | 1305 | if (tsk->clear_child_tid) { |
735f2770 | 1306 | if (!(tsk->signal->flags & SIGNAL_GROUP_COREDUMP) && |
9c8a8228 ED |
1307 | atomic_read(&mm->mm_users) > 1) { |
1308 | /* | |
1309 | * We don't check the error code - if userspace has | |
1310 | * not set up a proper pointer then tough luck. | |
1311 | */ | |
1312 | put_user(0, tsk->clear_child_tid); | |
2de0db99 DB |
1313 | do_futex(tsk->clear_child_tid, FUTEX_WAKE, |
1314 | 1, NULL, NULL, 0, 0); | |
9c8a8228 | 1315 | } |
1da177e4 | 1316 | tsk->clear_child_tid = NULL; |
1da177e4 | 1317 | } |
f7505d64 KK |
1318 | |
1319 | /* | |
1320 | * All done, finally we can wake up parent and return this mm to him. | |
1321 | * Also kthread_stop() uses this completion for synchronization. | |
1322 | */ | |
1323 | if (tsk->vfork_done) | |
1324 | complete_vfork_done(tsk); | |
1da177e4 LT |
1325 | } |
1326 | ||
13585fa0 NA |
1327 | /** |
1328 | * dup_mm() - duplicates an existing mm structure | |
1329 | * @tsk: the task_struct with which the new mm will be associated. | |
1330 | * @oldmm: the mm to duplicate. | |
1331 | * | |
1332 | * Allocates a new mm structure and duplicates the provided @oldmm structure | |
1333 | * content into it. | |
1334 | * | |
1335 | * Return: the duplicated mm or NULL on failure. | |
a0a7ec30 | 1336 | */ |
13585fa0 NA |
1337 | static struct mm_struct *dup_mm(struct task_struct *tsk, |
1338 | struct mm_struct *oldmm) | |
a0a7ec30 | 1339 | { |
13585fa0 | 1340 | struct mm_struct *mm; |
a0a7ec30 JD |
1341 | int err; |
1342 | ||
a0a7ec30 JD |
1343 | mm = allocate_mm(); |
1344 | if (!mm) | |
1345 | goto fail_nomem; | |
1346 | ||
1347 | memcpy(mm, oldmm, sizeof(*mm)); | |
1348 | ||
bfedb589 | 1349 | if (!mm_init(mm, tsk, mm->user_ns)) |
a0a7ec30 JD |
1350 | goto fail_nomem; |
1351 | ||
a0a7ec30 JD |
1352 | err = dup_mmap(mm, oldmm); |
1353 | if (err) | |
1354 | goto free_pt; | |
1355 | ||
1356 | mm->hiwater_rss = get_mm_rss(mm); | |
1357 | mm->hiwater_vm = mm->total_vm; | |
1358 | ||
801460d0 HS |
1359 | if (mm->binfmt && !try_module_get(mm->binfmt->module)) |
1360 | goto free_pt; | |
1361 | ||
a0a7ec30 JD |
1362 | return mm; |
1363 | ||
1364 | free_pt: | |
801460d0 HS |
1365 | /* don't put binfmt in mmput, we haven't got module yet */ |
1366 | mm->binfmt = NULL; | |
c3f3ce04 | 1367 | mm_init_owner(mm, NULL); |
a0a7ec30 JD |
1368 | mmput(mm); |
1369 | ||
1370 | fail_nomem: | |
1371 | return NULL; | |
a0a7ec30 JD |
1372 | } |
1373 | ||
fb0a685c | 1374 | static int copy_mm(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 | 1375 | { |
fb0a685c | 1376 | struct mm_struct *mm, *oldmm; |
1da177e4 LT |
1377 | int retval; |
1378 | ||
1379 | tsk->min_flt = tsk->maj_flt = 0; | |
1380 | tsk->nvcsw = tsk->nivcsw = 0; | |
17406b82 MSB |
1381 | #ifdef CONFIG_DETECT_HUNG_TASK |
1382 | tsk->last_switch_count = tsk->nvcsw + tsk->nivcsw; | |
a2e51445 | 1383 | tsk->last_switch_time = 0; |
17406b82 | 1384 | #endif |
1da177e4 LT |
1385 | |
1386 | tsk->mm = NULL; | |
1387 | tsk->active_mm = NULL; | |
1388 | ||
1389 | /* | |
1390 | * Are we cloning a kernel thread? | |
1391 | * | |
1392 | * We need to steal a active VM for that.. | |
1393 | */ | |
1394 | oldmm = current->mm; | |
1395 | if (!oldmm) | |
1396 | return 0; | |
1397 | ||
615d6e87 DB |
1398 | /* initialize the new vmacache entries */ |
1399 | vmacache_flush(tsk); | |
1400 | ||
1da177e4 | 1401 | if (clone_flags & CLONE_VM) { |
3fce371b | 1402 | mmget(oldmm); |
1da177e4 | 1403 | mm = oldmm; |
1da177e4 LT |
1404 | goto good_mm; |
1405 | } | |
1406 | ||
1407 | retval = -ENOMEM; | |
13585fa0 | 1408 | mm = dup_mm(tsk, current->mm); |
1da177e4 LT |
1409 | if (!mm) |
1410 | goto fail_nomem; | |
1411 | ||
1da177e4 LT |
1412 | good_mm: |
1413 | tsk->mm = mm; | |
1414 | tsk->active_mm = mm; | |
1415 | return 0; | |
1416 | ||
1da177e4 LT |
1417 | fail_nomem: |
1418 | return retval; | |
1da177e4 LT |
1419 | } |
1420 | ||
a39bc516 | 1421 | static int copy_fs(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 | 1422 | { |
498052bb | 1423 | struct fs_struct *fs = current->fs; |
1da177e4 | 1424 | if (clone_flags & CLONE_FS) { |
498052bb | 1425 | /* tsk->fs is already what we want */ |
2a4419b5 | 1426 | spin_lock(&fs->lock); |
498052bb | 1427 | if (fs->in_exec) { |
2a4419b5 | 1428 | spin_unlock(&fs->lock); |
498052bb AV |
1429 | return -EAGAIN; |
1430 | } | |
1431 | fs->users++; | |
2a4419b5 | 1432 | spin_unlock(&fs->lock); |
1da177e4 LT |
1433 | return 0; |
1434 | } | |
498052bb | 1435 | tsk->fs = copy_fs_struct(fs); |
1da177e4 LT |
1436 | if (!tsk->fs) |
1437 | return -ENOMEM; | |
1438 | return 0; | |
1439 | } | |
1440 | ||
fb0a685c | 1441 | static int copy_files(unsigned long clone_flags, struct task_struct *tsk) |
a016f338 JD |
1442 | { |
1443 | struct files_struct *oldf, *newf; | |
1444 | int error = 0; | |
1445 | ||
1446 | /* | |
1447 | * A background process may not have any files ... | |
1448 | */ | |
1449 | oldf = current->files; | |
1450 | if (!oldf) | |
1451 | goto out; | |
1452 | ||
1453 | if (clone_flags & CLONE_FILES) { | |
1454 | atomic_inc(&oldf->count); | |
1455 | goto out; | |
1456 | } | |
1457 | ||
a016f338 JD |
1458 | newf = dup_fd(oldf, &error); |
1459 | if (!newf) | |
1460 | goto out; | |
1461 | ||
1462 | tsk->files = newf; | |
1463 | error = 0; | |
1464 | out: | |
1465 | return error; | |
1466 | } | |
1467 | ||
fadad878 | 1468 | static int copy_io(unsigned long clone_flags, struct task_struct *tsk) |
fd0928df JA |
1469 | { |
1470 | #ifdef CONFIG_BLOCK | |
1471 | struct io_context *ioc = current->io_context; | |
6e736be7 | 1472 | struct io_context *new_ioc; |
fd0928df JA |
1473 | |
1474 | if (!ioc) | |
1475 | return 0; | |
fadad878 JA |
1476 | /* |
1477 | * Share io context with parent, if CLONE_IO is set | |
1478 | */ | |
1479 | if (clone_flags & CLONE_IO) { | |
3d48749d TH |
1480 | ioc_task_link(ioc); |
1481 | tsk->io_context = ioc; | |
fadad878 | 1482 | } else if (ioprio_valid(ioc->ioprio)) { |
6e736be7 TH |
1483 | new_ioc = get_task_io_context(tsk, GFP_KERNEL, NUMA_NO_NODE); |
1484 | if (unlikely(!new_ioc)) | |
fd0928df JA |
1485 | return -ENOMEM; |
1486 | ||
6e736be7 | 1487 | new_ioc->ioprio = ioc->ioprio; |
11a3122f | 1488 | put_io_context(new_ioc); |
fd0928df JA |
1489 | } |
1490 | #endif | |
1491 | return 0; | |
1492 | } | |
1493 | ||
a39bc516 | 1494 | static int copy_sighand(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 LT |
1495 | { |
1496 | struct sighand_struct *sig; | |
1497 | ||
60348802 | 1498 | if (clone_flags & CLONE_SIGHAND) { |
d036bda7 | 1499 | refcount_inc(¤t->sighand->count); |
1da177e4 LT |
1500 | return 0; |
1501 | } | |
1502 | sig = kmem_cache_alloc(sighand_cachep, GFP_KERNEL); | |
e56d0903 | 1503 | rcu_assign_pointer(tsk->sighand, sig); |
1da177e4 LT |
1504 | if (!sig) |
1505 | return -ENOMEM; | |
9d7fb042 | 1506 | |
d036bda7 | 1507 | refcount_set(&sig->count, 1); |
06e62a46 | 1508 | spin_lock_irq(¤t->sighand->siglock); |
1da177e4 | 1509 | memcpy(sig->action, current->sighand->action, sizeof(sig->action)); |
06e62a46 | 1510 | spin_unlock_irq(¤t->sighand->siglock); |
1da177e4 LT |
1511 | return 0; |
1512 | } | |
1513 | ||
a7e5328a | 1514 | void __cleanup_sighand(struct sighand_struct *sighand) |
c81addc9 | 1515 | { |
d036bda7 | 1516 | if (refcount_dec_and_test(&sighand->count)) { |
d80e731e | 1517 | signalfd_cleanup(sighand); |
392809b2 | 1518 | /* |
5f0d5a3a | 1519 | * sighand_cachep is SLAB_TYPESAFE_BY_RCU so we can free it |
392809b2 ON |
1520 | * without an RCU grace period, see __lock_task_sighand(). |
1521 | */ | |
c81addc9 | 1522 | kmem_cache_free(sighand_cachep, sighand); |
d80e731e | 1523 | } |
c81addc9 ON |
1524 | } |
1525 | ||
b18b6a9c | 1526 | #ifdef CONFIG_POSIX_TIMERS |
f06febc9 FM |
1527 | /* |
1528 | * Initialize POSIX timer handling for a thread group. | |
1529 | */ | |
1530 | static void posix_cpu_timers_init_group(struct signal_struct *sig) | |
1531 | { | |
78d7d407 JS |
1532 | unsigned long cpu_limit; |
1533 | ||
316c1608 | 1534 | cpu_limit = READ_ONCE(sig->rlim[RLIMIT_CPU].rlim_cur); |
78d7d407 | 1535 | if (cpu_limit != RLIM_INFINITY) { |
ebd7e7fc | 1536 | sig->cputime_expires.prof_exp = cpu_limit * NSEC_PER_SEC; |
d5c373eb | 1537 | sig->cputimer.running = true; |
6279a751 ON |
1538 | } |
1539 | ||
f06febc9 FM |
1540 | /* The timer lists. */ |
1541 | INIT_LIST_HEAD(&sig->cpu_timers[0]); | |
1542 | INIT_LIST_HEAD(&sig->cpu_timers[1]); | |
1543 | INIT_LIST_HEAD(&sig->cpu_timers[2]); | |
1544 | } | |
b18b6a9c NP |
1545 | #else |
1546 | static inline void posix_cpu_timers_init_group(struct signal_struct *sig) { } | |
1547 | #endif | |
f06febc9 | 1548 | |
a39bc516 | 1549 | static int copy_signal(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 LT |
1550 | { |
1551 | struct signal_struct *sig; | |
1da177e4 | 1552 | |
4ab6c083 | 1553 | if (clone_flags & CLONE_THREAD) |
490dea45 | 1554 | return 0; |
490dea45 | 1555 | |
a56704ef | 1556 | sig = kmem_cache_zalloc(signal_cachep, GFP_KERNEL); |
1da177e4 LT |
1557 | tsk->signal = sig; |
1558 | if (!sig) | |
1559 | return -ENOMEM; | |
1560 | ||
b3ac022c | 1561 | sig->nr_threads = 1; |
1da177e4 | 1562 | atomic_set(&sig->live, 1); |
60d4de3f | 1563 | refcount_set(&sig->sigcnt, 1); |
0c740d0a ON |
1564 | |
1565 | /* list_add(thread_node, thread_head) without INIT_LIST_HEAD() */ | |
1566 | sig->thread_head = (struct list_head)LIST_HEAD_INIT(tsk->thread_node); | |
1567 | tsk->thread_node = (struct list_head)LIST_HEAD_INIT(sig->thread_head); | |
1568 | ||
1da177e4 | 1569 | init_waitqueue_head(&sig->wait_chldexit); |
db51aecc | 1570 | sig->curr_target = tsk; |
1da177e4 | 1571 | init_sigpending(&sig->shared_pending); |
c3ad2c3b | 1572 | INIT_HLIST_HEAD(&sig->multiprocess); |
e78c3496 | 1573 | seqlock_init(&sig->stats_lock); |
9d7fb042 | 1574 | prev_cputime_init(&sig->prev_cputime); |
1da177e4 | 1575 | |
baa73d9e | 1576 | #ifdef CONFIG_POSIX_TIMERS |
b18b6a9c | 1577 | INIT_LIST_HEAD(&sig->posix_timers); |
c9cb2e3d | 1578 | hrtimer_init(&sig->real_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL); |
1da177e4 | 1579 | sig->real_timer.function = it_real_fn; |
baa73d9e | 1580 | #endif |
1da177e4 | 1581 | |
1da177e4 LT |
1582 | task_lock(current->group_leader); |
1583 | memcpy(sig->rlim, current->signal->rlim, sizeof sig->rlim); | |
1584 | task_unlock(current->group_leader); | |
1585 | ||
6279a751 ON |
1586 | posix_cpu_timers_init_group(sig); |
1587 | ||
522ed776 | 1588 | tty_audit_fork(sig); |
5091faa4 | 1589 | sched_autogroup_fork(sig); |
522ed776 | 1590 | |
a63d83f4 | 1591 | sig->oom_score_adj = current->signal->oom_score_adj; |
dabb16f6 | 1592 | sig->oom_score_adj_min = current->signal->oom_score_adj_min; |
28b83c51 | 1593 | |
9b1bf12d KM |
1594 | mutex_init(&sig->cred_guard_mutex); |
1595 | ||
1da177e4 LT |
1596 | return 0; |
1597 | } | |
1598 | ||
dbd95212 KC |
1599 | static void copy_seccomp(struct task_struct *p) |
1600 | { | |
1601 | #ifdef CONFIG_SECCOMP | |
1602 | /* | |
1603 | * Must be called with sighand->lock held, which is common to | |
1604 | * all threads in the group. Holding cred_guard_mutex is not | |
1605 | * needed because this new task is not yet running and cannot | |
1606 | * be racing exec. | |
1607 | */ | |
69f6a34b | 1608 | assert_spin_locked(¤t->sighand->siglock); |
dbd95212 KC |
1609 | |
1610 | /* Ref-count the new filter user, and assign it. */ | |
1611 | get_seccomp_filter(current); | |
1612 | p->seccomp = current->seccomp; | |
1613 | ||
1614 | /* | |
1615 | * Explicitly enable no_new_privs here in case it got set | |
1616 | * between the task_struct being duplicated and holding the | |
1617 | * sighand lock. The seccomp state and nnp must be in sync. | |
1618 | */ | |
1619 | if (task_no_new_privs(current)) | |
1620 | task_set_no_new_privs(p); | |
1621 | ||
1622 | /* | |
1623 | * If the parent gained a seccomp mode after copying thread | |
1624 | * flags and between before we held the sighand lock, we have | |
1625 | * to manually enable the seccomp thread flag here. | |
1626 | */ | |
1627 | if (p->seccomp.mode != SECCOMP_MODE_DISABLED) | |
1628 | set_tsk_thread_flag(p, TIF_SECCOMP); | |
1629 | #endif | |
1630 | } | |
1631 | ||
17da2bd9 | 1632 | SYSCALL_DEFINE1(set_tid_address, int __user *, tidptr) |
1da177e4 LT |
1633 | { |
1634 | current->clear_child_tid = tidptr; | |
1635 | ||
b488893a | 1636 | return task_pid_vnr(current); |
1da177e4 LT |
1637 | } |
1638 | ||
a39bc516 | 1639 | static void rt_mutex_init_task(struct task_struct *p) |
23f78d4a | 1640 | { |
1d615482 | 1641 | raw_spin_lock_init(&p->pi_lock); |
e29e175b | 1642 | #ifdef CONFIG_RT_MUTEXES |
a23ba907 | 1643 | p->pi_waiters = RB_ROOT_CACHED; |
e96a7705 | 1644 | p->pi_top_task = NULL; |
23f78d4a | 1645 | p->pi_blocked_on = NULL; |
23f78d4a IM |
1646 | #endif |
1647 | } | |
1648 | ||
b18b6a9c | 1649 | #ifdef CONFIG_POSIX_TIMERS |
f06febc9 FM |
1650 | /* |
1651 | * Initialize POSIX timer handling for a single task. | |
1652 | */ | |
1653 | static void posix_cpu_timers_init(struct task_struct *tsk) | |
1654 | { | |
64861634 MS |
1655 | tsk->cputime_expires.prof_exp = 0; |
1656 | tsk->cputime_expires.virt_exp = 0; | |
f06febc9 FM |
1657 | tsk->cputime_expires.sched_exp = 0; |
1658 | INIT_LIST_HEAD(&tsk->cpu_timers[0]); | |
1659 | INIT_LIST_HEAD(&tsk->cpu_timers[1]); | |
1660 | INIT_LIST_HEAD(&tsk->cpu_timers[2]); | |
1661 | } | |
b18b6a9c NP |
1662 | #else |
1663 | static inline void posix_cpu_timers_init(struct task_struct *tsk) { } | |
1664 | #endif | |
f06febc9 | 1665 | |
2c470475 EB |
1666 | static inline void init_task_pid_links(struct task_struct *task) |
1667 | { | |
1668 | enum pid_type type; | |
1669 | ||
1670 | for (type = PIDTYPE_PID; type < PIDTYPE_MAX; ++type) { | |
1671 | INIT_HLIST_NODE(&task->pid_links[type]); | |
1672 | } | |
1673 | } | |
1674 | ||
81907739 ON |
1675 | static inline void |
1676 | init_task_pid(struct task_struct *task, enum pid_type type, struct pid *pid) | |
1677 | { | |
2c470475 EB |
1678 | if (type == PIDTYPE_PID) |
1679 | task->thread_pid = pid; | |
1680 | else | |
1681 | task->signal->pids[type] = pid; | |
81907739 ON |
1682 | } |
1683 | ||
6bfbaa51 IM |
1684 | static inline void rcu_copy_process(struct task_struct *p) |
1685 | { | |
1686 | #ifdef CONFIG_PREEMPT_RCU | |
1687 | p->rcu_read_lock_nesting = 0; | |
1688 | p->rcu_read_unlock_special.s = 0; | |
1689 | p->rcu_blocked_node = NULL; | |
1690 | INIT_LIST_HEAD(&p->rcu_node_entry); | |
1691 | #endif /* #ifdef CONFIG_PREEMPT_RCU */ | |
1692 | #ifdef CONFIG_TASKS_RCU | |
1693 | p->rcu_tasks_holdout = false; | |
1694 | INIT_LIST_HEAD(&p->rcu_tasks_holdout_list); | |
1695 | p->rcu_tasks_idle_cpu = -1; | |
1696 | #endif /* #ifdef CONFIG_TASKS_RCU */ | |
1697 | } | |
1698 | ||
b3e58382 CB |
1699 | static int pidfd_release(struct inode *inode, struct file *file) |
1700 | { | |
1701 | struct pid *pid = file->private_data; | |
1702 | ||
1703 | file->private_data = NULL; | |
1704 | put_pid(pid); | |
1705 | return 0; | |
1706 | } | |
1707 | ||
1708 | #ifdef CONFIG_PROC_FS | |
1709 | static void pidfd_show_fdinfo(struct seq_file *m, struct file *f) | |
1710 | { | |
1711 | struct pid_namespace *ns = proc_pid_ns(file_inode(m->file)); | |
1712 | struct pid *pid = f->private_data; | |
1713 | ||
1714 | seq_put_decimal_ull(m, "Pid:\t", pid_nr_ns(pid, ns)); | |
1715 | seq_putc(m, '\n'); | |
1716 | } | |
1717 | #endif | |
1718 | ||
b53b0b9d JFG |
1719 | /* |
1720 | * Poll support for process exit notification. | |
1721 | */ | |
9d153525 | 1722 | static __poll_t pidfd_poll(struct file *file, struct poll_table_struct *pts) |
b53b0b9d JFG |
1723 | { |
1724 | struct task_struct *task; | |
1725 | struct pid *pid = file->private_data; | |
9d153525 | 1726 | __poll_t poll_flags = 0; |
b53b0b9d JFG |
1727 | |
1728 | poll_wait(file, &pid->wait_pidfd, pts); | |
1729 | ||
1730 | rcu_read_lock(); | |
1731 | task = pid_task(pid, PIDTYPE_PID); | |
1732 | /* | |
1733 | * Inform pollers only when the whole thread group exits. | |
1734 | * If the thread group leader exits before all other threads in the | |
1735 | * group, then poll(2) should block, similar to the wait(2) family. | |
1736 | */ | |
1737 | if (!task || (task->exit_state && thread_group_empty(task))) | |
9d153525 | 1738 | poll_flags = EPOLLIN | EPOLLRDNORM; |
b53b0b9d JFG |
1739 | rcu_read_unlock(); |
1740 | ||
1741 | return poll_flags; | |
1742 | } | |
1743 | ||
b3e58382 CB |
1744 | const struct file_operations pidfd_fops = { |
1745 | .release = pidfd_release, | |
b53b0b9d | 1746 | .poll = pidfd_poll, |
b3e58382 CB |
1747 | #ifdef CONFIG_PROC_FS |
1748 | .show_fdinfo = pidfd_show_fdinfo, | |
1749 | #endif | |
1750 | }; | |
1751 | ||
c3f3ce04 AA |
1752 | static void __delayed_free_task(struct rcu_head *rhp) |
1753 | { | |
1754 | struct task_struct *tsk = container_of(rhp, struct task_struct, rcu); | |
1755 | ||
1756 | free_task(tsk); | |
1757 | } | |
1758 | ||
1759 | static __always_inline void delayed_free_task(struct task_struct *tsk) | |
1760 | { | |
1761 | if (IS_ENABLED(CONFIG_MEMCG)) | |
1762 | call_rcu(&tsk->rcu, __delayed_free_task); | |
1763 | else | |
1764 | free_task(tsk); | |
1765 | } | |
1766 | ||
1da177e4 LT |
1767 | /* |
1768 | * This creates a new process as a copy of the old one, | |
1769 | * but does not actually start it yet. | |
1770 | * | |
1771 | * It copies the registers, and all the appropriate | |
1772 | * parts of the process environment (as per the clone | |
1773 | * flags). The actual kick-off is left to the caller. | |
1774 | */ | |
0766f788 | 1775 | static __latent_entropy struct task_struct *copy_process( |
09a05394 | 1776 | struct pid *pid, |
3033f14a | 1777 | int trace, |
7f192e3c CB |
1778 | int node, |
1779 | struct kernel_clone_args *args) | |
1da177e4 | 1780 | { |
b3e58382 | 1781 | int pidfd = -1, retval; |
a24efe62 | 1782 | struct task_struct *p; |
c3ad2c3b | 1783 | struct multiprocess_signals delayed; |
6fd2fe49 | 1784 | struct file *pidfile = NULL; |
7f192e3c | 1785 | u64 clone_flags = args->flags; |
1da177e4 | 1786 | |
667b6094 MPS |
1787 | /* |
1788 | * Don't allow sharing the root directory with processes in a different | |
1789 | * namespace | |
1790 | */ | |
1da177e4 LT |
1791 | if ((clone_flags & (CLONE_NEWNS|CLONE_FS)) == (CLONE_NEWNS|CLONE_FS)) |
1792 | return ERR_PTR(-EINVAL); | |
1793 | ||
e66eded8 EB |
1794 | if ((clone_flags & (CLONE_NEWUSER|CLONE_FS)) == (CLONE_NEWUSER|CLONE_FS)) |
1795 | return ERR_PTR(-EINVAL); | |
1796 | ||
2b612266 SH |
1797 | if ((clone_flags & CLONE_NEWUSER) && !unprivileged_userns_clone) |
1798 | if (!capable(CAP_SYS_ADMIN)) | |
1799 | return ERR_PTR(-EPERM); | |
1800 | ||
1da177e4 LT |
1801 | /* |
1802 | * Thread groups must share signals as well, and detached threads | |
1803 | * can only be started up within the thread group. | |
1804 | */ | |
1805 | if ((clone_flags & CLONE_THREAD) && !(clone_flags & CLONE_SIGHAND)) | |
1806 | return ERR_PTR(-EINVAL); | |
1807 | ||
1808 | /* | |
1809 | * Shared signal handlers imply shared VM. By way of the above, | |
1810 | * thread groups also imply shared VM. Blocking this case allows | |
1811 | * for various simplifications in other code. | |
1812 | */ | |
1813 | if ((clone_flags & CLONE_SIGHAND) && !(clone_flags & CLONE_VM)) | |
1814 | return ERR_PTR(-EINVAL); | |
1815 | ||
123be07b SB |
1816 | /* |
1817 | * Siblings of global init remain as zombies on exit since they are | |
1818 | * not reaped by their parent (swapper). To solve this and to avoid | |
1819 | * multi-rooted process trees, prevent global and container-inits | |
1820 | * from creating siblings. | |
1821 | */ | |
1822 | if ((clone_flags & CLONE_PARENT) && | |
1823 | current->signal->flags & SIGNAL_UNKILLABLE) | |
1824 | return ERR_PTR(-EINVAL); | |
1825 | ||
8382fcac | 1826 | /* |
40a0d32d | 1827 | * If the new process will be in a different pid or user namespace |
faf00da5 | 1828 | * do not allow it to share a thread group with the forking task. |
8382fcac | 1829 | */ |
faf00da5 | 1830 | if (clone_flags & CLONE_THREAD) { |
40a0d32d ON |
1831 | if ((clone_flags & (CLONE_NEWUSER | CLONE_NEWPID)) || |
1832 | (task_active_pid_ns(current) != | |
1833 | current->nsproxy->pid_ns_for_children)) | |
1834 | return ERR_PTR(-EINVAL); | |
1835 | } | |
8382fcac | 1836 | |
b3e58382 | 1837 | if (clone_flags & CLONE_PIDFD) { |
b3e58382 | 1838 | /* |
b3e58382 CB |
1839 | * - CLONE_DETACHED is blocked so that we can potentially |
1840 | * reuse it later for CLONE_PIDFD. | |
1841 | * - CLONE_THREAD is blocked until someone really needs it. | |
1842 | */ | |
7f192e3c | 1843 | if (clone_flags & (CLONE_DETACHED | CLONE_THREAD)) |
b3e58382 | 1844 | return ERR_PTR(-EINVAL); |
b3e58382 CB |
1845 | } |
1846 | ||
c3ad2c3b EB |
1847 | /* |
1848 | * Force any signals received before this point to be delivered | |
1849 | * before the fork happens. Collect up signals sent to multiple | |
1850 | * processes that happen during the fork and delay them so that | |
1851 | * they appear to happen after the fork. | |
1852 | */ | |
1853 | sigemptyset(&delayed.signal); | |
1854 | INIT_HLIST_NODE(&delayed.node); | |
1855 | ||
1856 | spin_lock_irq(¤t->sighand->siglock); | |
1857 | if (!(clone_flags & CLONE_THREAD)) | |
1858 | hlist_add_head(&delayed.node, ¤t->signal->multiprocess); | |
1859 | recalc_sigpending(); | |
1860 | spin_unlock_irq(¤t->sighand->siglock); | |
1861 | retval = -ERESTARTNOINTR; | |
1862 | if (signal_pending(current)) | |
1863 | goto fork_out; | |
1864 | ||
1da177e4 | 1865 | retval = -ENOMEM; |
725fc629 | 1866 | p = dup_task_struct(current, node); |
1da177e4 LT |
1867 | if (!p) |
1868 | goto fork_out; | |
1869 | ||
4d6501dc VN |
1870 | /* |
1871 | * This _must_ happen before we call free_task(), i.e. before we jump | |
1872 | * to any of the bad_fork_* labels. This is to avoid freeing | |
1873 | * p->set_child_tid which is (ab)used as a kthread's data pointer for | |
1874 | * kernel threads (PF_KTHREAD). | |
1875 | */ | |
7f192e3c | 1876 | p->set_child_tid = (clone_flags & CLONE_CHILD_SETTID) ? args->child_tid : NULL; |
4d6501dc VN |
1877 | /* |
1878 | * Clear TID on mm_release()? | |
1879 | */ | |
7f192e3c | 1880 | p->clear_child_tid = (clone_flags & CLONE_CHILD_CLEARTID) ? args->child_tid : NULL; |
4d6501dc | 1881 | |
f7e8b616 SR |
1882 | ftrace_graph_init_task(p); |
1883 | ||
bea493a0 PZ |
1884 | rt_mutex_init_task(p); |
1885 | ||
d12c1a37 | 1886 | #ifdef CONFIG_PROVE_LOCKING |
de30a2b3 IM |
1887 | DEBUG_LOCKS_WARN_ON(!p->hardirqs_enabled); |
1888 | DEBUG_LOCKS_WARN_ON(!p->softirqs_enabled); | |
1889 | #endif | |
1da177e4 | 1890 | retval = -EAGAIN; |
3b11a1de | 1891 | if (atomic_read(&p->real_cred->user->processes) >= |
78d7d407 | 1892 | task_rlimit(p, RLIMIT_NPROC)) { |
b57922b6 EP |
1893 | if (p->real_cred->user != INIT_USER && |
1894 | !capable(CAP_SYS_RESOURCE) && !capable(CAP_SYS_ADMIN)) | |
1da177e4 LT |
1895 | goto bad_fork_free; |
1896 | } | |
72fa5997 | 1897 | current->flags &= ~PF_NPROC_EXCEEDED; |
1da177e4 | 1898 | |
f1752eec DH |
1899 | retval = copy_creds(p, clone_flags); |
1900 | if (retval < 0) | |
1901 | goto bad_fork_free; | |
1da177e4 LT |
1902 | |
1903 | /* | |
1904 | * If multiple threads are within copy_process(), then this check | |
1905 | * triggers too late. This doesn't hurt, the check is only there | |
1906 | * to stop root fork bombs. | |
1907 | */ | |
04ec93fe | 1908 | retval = -EAGAIN; |
1da177e4 LT |
1909 | if (nr_threads >= max_threads) |
1910 | goto bad_fork_cleanup_count; | |
1911 | ||
ca74e92b | 1912 | delayacct_tsk_init(p); /* Must remain after dup_task_struct() */ |
c1de45ca | 1913 | p->flags &= ~(PF_SUPERPRIV | PF_WQ_WORKER | PF_IDLE); |
514ddb44 | 1914 | p->flags |= PF_FORKNOEXEC; |
1da177e4 LT |
1915 | INIT_LIST_HEAD(&p->children); |
1916 | INIT_LIST_HEAD(&p->sibling); | |
f41d911f | 1917 | rcu_copy_process(p); |
1da177e4 LT |
1918 | p->vfork_done = NULL; |
1919 | spin_lock_init(&p->alloc_lock); | |
1da177e4 | 1920 | |
1da177e4 LT |
1921 | init_sigpending(&p->pending); |
1922 | ||
64861634 | 1923 | p->utime = p->stime = p->gtime = 0; |
40565b5a | 1924 | #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME |
64861634 | 1925 | p->utimescaled = p->stimescaled = 0; |
40565b5a | 1926 | #endif |
9d7fb042 PZ |
1927 | prev_cputime_init(&p->prev_cputime); |
1928 | ||
6a61671b | 1929 | #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN |
bac5b6b6 FW |
1930 | seqcount_init(&p->vtime.seqcount); |
1931 | p->vtime.starttime = 0; | |
1932 | p->vtime.state = VTIME_INACTIVE; | |
6a61671b FW |
1933 | #endif |
1934 | ||
a3a2e76c KH |
1935 | #if defined(SPLIT_RSS_COUNTING) |
1936 | memset(&p->rss_stat, 0, sizeof(p->rss_stat)); | |
1937 | #endif | |
172ba844 | 1938 | |
6976675d AV |
1939 | p->default_timer_slack_ns = current->timer_slack_ns; |
1940 | ||
eb414681 JW |
1941 | #ifdef CONFIG_PSI |
1942 | p->psi_flags = 0; | |
1943 | #endif | |
1944 | ||
5995477a | 1945 | task_io_accounting_init(&p->ioac); |
1da177e4 LT |
1946 | acct_clear_integrals(p); |
1947 | ||
f06febc9 | 1948 | posix_cpu_timers_init(p); |
1da177e4 | 1949 | |
1da177e4 | 1950 | p->io_context = NULL; |
c0b0ae8a | 1951 | audit_set_context(p, NULL); |
b4f48b63 | 1952 | cgroup_fork(p); |
1da177e4 | 1953 | #ifdef CONFIG_NUMA |
846a16bf | 1954 | p->mempolicy = mpol_dup(p->mempolicy); |
fb0a685c DRO |
1955 | if (IS_ERR(p->mempolicy)) { |
1956 | retval = PTR_ERR(p->mempolicy); | |
1957 | p->mempolicy = NULL; | |
e8604cb4 | 1958 | goto bad_fork_cleanup_threadgroup_lock; |
fb0a685c | 1959 | } |
1da177e4 | 1960 | #endif |
778d3b0f MH |
1961 | #ifdef CONFIG_CPUSETS |
1962 | p->cpuset_mem_spread_rotor = NUMA_NO_NODE; | |
1963 | p->cpuset_slab_spread_rotor = NUMA_NO_NODE; | |
cc9a6c87 | 1964 | seqcount_init(&p->mems_allowed_seq); |
778d3b0f | 1965 | #endif |
de30a2b3 IM |
1966 | #ifdef CONFIG_TRACE_IRQFLAGS |
1967 | p->irq_events = 0; | |
1968 | p->hardirqs_enabled = 0; | |
1969 | p->hardirq_enable_ip = 0; | |
1970 | p->hardirq_enable_event = 0; | |
1971 | p->hardirq_disable_ip = _THIS_IP_; | |
1972 | p->hardirq_disable_event = 0; | |
1973 | p->softirqs_enabled = 1; | |
1974 | p->softirq_enable_ip = _THIS_IP_; | |
1975 | p->softirq_enable_event = 0; | |
1976 | p->softirq_disable_ip = 0; | |
1977 | p->softirq_disable_event = 0; | |
1978 | p->hardirq_context = 0; | |
1979 | p->softirq_context = 0; | |
1980 | #endif | |
8bcbde54 DH |
1981 | |
1982 | p->pagefault_disabled = 0; | |
1983 | ||
fbb9ce95 | 1984 | #ifdef CONFIG_LOCKDEP |
b09be676 | 1985 | lockdep_init_task(p); |
fbb9ce95 | 1986 | #endif |
1da177e4 | 1987 | |
408894ee IM |
1988 | #ifdef CONFIG_DEBUG_MUTEXES |
1989 | p->blocked_on = NULL; /* not blocked yet */ | |
1990 | #endif | |
cafe5635 KO |
1991 | #ifdef CONFIG_BCACHE |
1992 | p->sequential_io = 0; | |
1993 | p->sequential_io_avg = 0; | |
1994 | #endif | |
0f481406 | 1995 | |
3c90e6e9 | 1996 | /* Perform scheduler related setup. Assign this task to a CPU. */ |
aab03e05 DF |
1997 | retval = sched_fork(clone_flags, p); |
1998 | if (retval) | |
1999 | goto bad_fork_cleanup_policy; | |
6ab423e0 | 2000 | |
cdd6c482 | 2001 | retval = perf_event_init_task(p); |
6ab423e0 PZ |
2002 | if (retval) |
2003 | goto bad_fork_cleanup_policy; | |
fb0a685c DRO |
2004 | retval = audit_alloc(p); |
2005 | if (retval) | |
6c72e350 | 2006 | goto bad_fork_cleanup_perf; |
1da177e4 | 2007 | /* copy all the process information */ |
ab602f79 | 2008 | shm_init_task(p); |
e4e55b47 | 2009 | retval = security_task_alloc(p, clone_flags); |
fb0a685c | 2010 | if (retval) |
1da177e4 | 2011 | goto bad_fork_cleanup_audit; |
e4e55b47 TH |
2012 | retval = copy_semundo(clone_flags, p); |
2013 | if (retval) | |
2014 | goto bad_fork_cleanup_security; | |
fb0a685c DRO |
2015 | retval = copy_files(clone_flags, p); |
2016 | if (retval) | |
1da177e4 | 2017 | goto bad_fork_cleanup_semundo; |
fb0a685c DRO |
2018 | retval = copy_fs(clone_flags, p); |
2019 | if (retval) | |
1da177e4 | 2020 | goto bad_fork_cleanup_files; |
fb0a685c DRO |
2021 | retval = copy_sighand(clone_flags, p); |
2022 | if (retval) | |
1da177e4 | 2023 | goto bad_fork_cleanup_fs; |
fb0a685c DRO |
2024 | retval = copy_signal(clone_flags, p); |
2025 | if (retval) | |
1da177e4 | 2026 | goto bad_fork_cleanup_sighand; |
fb0a685c DRO |
2027 | retval = copy_mm(clone_flags, p); |
2028 | if (retval) | |
1da177e4 | 2029 | goto bad_fork_cleanup_signal; |
fb0a685c DRO |
2030 | retval = copy_namespaces(clone_flags, p); |
2031 | if (retval) | |
d84f4f99 | 2032 | goto bad_fork_cleanup_mm; |
fb0a685c DRO |
2033 | retval = copy_io(clone_flags, p); |
2034 | if (retval) | |
fd0928df | 2035 | goto bad_fork_cleanup_namespaces; |
7f192e3c CB |
2036 | retval = copy_thread_tls(clone_flags, args->stack, args->stack_size, p, |
2037 | args->tls); | |
1da177e4 | 2038 | if (retval) |
fd0928df | 2039 | goto bad_fork_cleanup_io; |
1da177e4 | 2040 | |
afaef01c AP |
2041 | stackleak_task_init(p); |
2042 | ||
425fb2b4 | 2043 | if (pid != &init_struct_pid) { |
c2b1df2e | 2044 | pid = alloc_pid(p->nsproxy->pid_ns_for_children); |
35f71bc0 MH |
2045 | if (IS_ERR(pid)) { |
2046 | retval = PTR_ERR(pid); | |
0740aa5f | 2047 | goto bad_fork_cleanup_thread; |
35f71bc0 | 2048 | } |
425fb2b4 PE |
2049 | } |
2050 | ||
b3e58382 CB |
2051 | /* |
2052 | * This has to happen after we've potentially unshared the file | |
2053 | * descriptor table (so that the pidfd doesn't leak into the child | |
2054 | * if the fd table isn't shared). | |
2055 | */ | |
2056 | if (clone_flags & CLONE_PIDFD) { | |
6fd2fe49 | 2057 | retval = get_unused_fd_flags(O_RDWR | O_CLOEXEC); |
b3e58382 CB |
2058 | if (retval < 0) |
2059 | goto bad_fork_free_pid; | |
2060 | ||
2061 | pidfd = retval; | |
6fd2fe49 AV |
2062 | |
2063 | pidfile = anon_inode_getfile("[pidfd]", &pidfd_fops, pid, | |
2064 | O_RDWR | O_CLOEXEC); | |
2065 | if (IS_ERR(pidfile)) { | |
2066 | put_unused_fd(pidfd); | |
28dd29c0 | 2067 | retval = PTR_ERR(pidfile); |
6fd2fe49 AV |
2068 | goto bad_fork_free_pid; |
2069 | } | |
2070 | get_pid(pid); /* held by pidfile now */ | |
2071 | ||
7f192e3c | 2072 | retval = put_user(pidfd, args->pidfd); |
b3e58382 CB |
2073 | if (retval) |
2074 | goto bad_fork_put_pidfd; | |
2075 | } | |
2076 | ||
73c10101 JA |
2077 | #ifdef CONFIG_BLOCK |
2078 | p->plug = NULL; | |
2079 | #endif | |
42b2dd0a | 2080 | #ifdef CONFIG_FUTEX |
8f17d3a5 IM |
2081 | p->robust_list = NULL; |
2082 | #ifdef CONFIG_COMPAT | |
2083 | p->compat_robust_list = NULL; | |
2084 | #endif | |
c87e2837 IM |
2085 | INIT_LIST_HEAD(&p->pi_state_list); |
2086 | p->pi_state_cache = NULL; | |
42b2dd0a | 2087 | #endif |
f9a3879a GM |
2088 | /* |
2089 | * sigaltstack should be cleared when sharing the same VM | |
2090 | */ | |
2091 | if ((clone_flags & (CLONE_VM|CLONE_VFORK)) == CLONE_VM) | |
2a742138 | 2092 | sas_ss_reset(p); |
f9a3879a | 2093 | |
1da177e4 | 2094 | /* |
6580807d ON |
2095 | * Syscall tracing and stepping should be turned off in the |
2096 | * child regardless of CLONE_PTRACE. | |
1da177e4 | 2097 | */ |
6580807d | 2098 | user_disable_single_step(p); |
1da177e4 | 2099 | clear_tsk_thread_flag(p, TIF_SYSCALL_TRACE); |
ed75e8d5 LV |
2100 | #ifdef TIF_SYSCALL_EMU |
2101 | clear_tsk_thread_flag(p, TIF_SYSCALL_EMU); | |
2102 | #endif | |
e02c9b0d | 2103 | clear_tsk_latency_tracing(p); |
1da177e4 | 2104 | |
1da177e4 | 2105 | /* ok, now we should be set up.. */ |
18c830df ON |
2106 | p->pid = pid_nr(pid); |
2107 | if (clone_flags & CLONE_THREAD) { | |
5f8aadd8 | 2108 | p->exit_signal = -1; |
18c830df ON |
2109 | p->group_leader = current->group_leader; |
2110 | p->tgid = current->tgid; | |
2111 | } else { | |
2112 | if (clone_flags & CLONE_PARENT) | |
2113 | p->exit_signal = current->group_leader->exit_signal; | |
2114 | else | |
7f192e3c | 2115 | p->exit_signal = args->exit_signal; |
18c830df ON |
2116 | p->group_leader = p; |
2117 | p->tgid = p->pid; | |
2118 | } | |
5f8aadd8 | 2119 | |
9d823e8f WF |
2120 | p->nr_dirtied = 0; |
2121 | p->nr_dirtied_pause = 128 >> (PAGE_SHIFT - 10); | |
83712358 | 2122 | p->dirty_paused_when = 0; |
9d823e8f | 2123 | |
bb8cbbfe | 2124 | p->pdeath_signal = 0; |
47e65328 | 2125 | INIT_LIST_HEAD(&p->thread_group); |
158e1645 | 2126 | p->task_works = NULL; |
1da177e4 | 2127 | |
780de9dd | 2128 | cgroup_threadgroup_change_begin(current); |
7e47682e AS |
2129 | /* |
2130 | * Ensure that the cgroup subsystem policies allow the new process to be | |
2131 | * forked. It should be noted the the new process's css_set can be changed | |
2132 | * between here and cgroup_post_fork() if an organisation operation is in | |
2133 | * progress. | |
2134 | */ | |
b53202e6 | 2135 | retval = cgroup_can_fork(p); |
7e47682e | 2136 | if (retval) |
c3b7112d | 2137 | goto bad_fork_cgroup_threadgroup_change_end; |
7e47682e | 2138 | |
7b558513 DH |
2139 | /* |
2140 | * From this point on we must avoid any synchronous user-space | |
2141 | * communication until we take the tasklist-lock. In particular, we do | |
2142 | * not want user-space to be able to predict the process start-time by | |
2143 | * stalling fork(2) after we recorded the start_time but before it is | |
2144 | * visible to the system. | |
2145 | */ | |
2146 | ||
2147 | p->start_time = ktime_get_ns(); | |
9285ec4c | 2148 | p->real_start_time = ktime_get_boottime_ns(); |
7b558513 | 2149 | |
18c830df ON |
2150 | /* |
2151 | * Make it visible to the rest of the system, but dont wake it up yet. | |
2152 | * Need tasklist lock for parent etc handling! | |
2153 | */ | |
1da177e4 LT |
2154 | write_lock_irq(&tasklist_lock); |
2155 | ||
1da177e4 | 2156 | /* CLONE_PARENT re-uses the old parent */ |
2d5516cb | 2157 | if (clone_flags & (CLONE_PARENT|CLONE_THREAD)) { |
1da177e4 | 2158 | p->real_parent = current->real_parent; |
2d5516cb ON |
2159 | p->parent_exec_id = current->parent_exec_id; |
2160 | } else { | |
1da177e4 | 2161 | p->real_parent = current; |
2d5516cb ON |
2162 | p->parent_exec_id = current->self_exec_id; |
2163 | } | |
1da177e4 | 2164 | |
d83a7cb3 JP |
2165 | klp_copy_process(p); |
2166 | ||
3f17da69 | 2167 | spin_lock(¤t->sighand->siglock); |
4a2c7a78 | 2168 | |
dbd95212 KC |
2169 | /* |
2170 | * Copy seccomp details explicitly here, in case they were changed | |
2171 | * before holding sighand lock. | |
2172 | */ | |
2173 | copy_seccomp(p); | |
2174 | ||
d7822b1e MD |
2175 | rseq_fork(p, clone_flags); |
2176 | ||
4ca1d3ee | 2177 | /* Don't start children in a dying pid namespace */ |
e8cfbc24 | 2178 | if (unlikely(!(ns_of_pid(pid)->pid_allocated & PIDNS_ADDING))) { |
3fd37226 KT |
2179 | retval = -ENOMEM; |
2180 | goto bad_fork_cancel_cgroup; | |
2181 | } | |
4a2c7a78 | 2182 | |
7673bf55 EB |
2183 | /* Let kill terminate clone/fork in the middle */ |
2184 | if (fatal_signal_pending(current)) { | |
2185 | retval = -EINTR; | |
2186 | goto bad_fork_cancel_cgroup; | |
2187 | } | |
2188 | ||
6fd2fe49 AV |
2189 | /* past the last point of failure */ |
2190 | if (pidfile) | |
2191 | fd_install(pidfd, pidfile); | |
4a2c7a78 | 2192 | |
2c470475 | 2193 | init_task_pid_links(p); |
73b9ebfe | 2194 | if (likely(p->pid)) { |
4b9d33e6 | 2195 | ptrace_init_task(p, (clone_flags & CLONE_PTRACE) || trace); |
73b9ebfe | 2196 | |
81907739 | 2197 | init_task_pid(p, PIDTYPE_PID, pid); |
73b9ebfe | 2198 | if (thread_group_leader(p)) { |
6883f81a | 2199 | init_task_pid(p, PIDTYPE_TGID, pid); |
81907739 ON |
2200 | init_task_pid(p, PIDTYPE_PGID, task_pgrp(current)); |
2201 | init_task_pid(p, PIDTYPE_SID, task_session(current)); | |
2202 | ||
1c4042c2 | 2203 | if (is_child_reaper(pid)) { |
17cf22c3 | 2204 | ns_of_pid(pid)->child_reaper = p; |
1c4042c2 EB |
2205 | p->signal->flags |= SIGNAL_UNKILLABLE; |
2206 | } | |
c3ad2c3b | 2207 | p->signal->shared_pending.signal = delayed.signal; |
9c9f4ded | 2208 | p->signal->tty = tty_kref_get(current->signal->tty); |
749860ce PT |
2209 | /* |
2210 | * Inherit has_child_subreaper flag under the same | |
2211 | * tasklist_lock with adding child to the process tree | |
2212 | * for propagate_has_child_subreaper optimization. | |
2213 | */ | |
2214 | p->signal->has_child_subreaper = p->real_parent->signal->has_child_subreaper || | |
2215 | p->real_parent->signal->is_child_subreaper; | |
9cd80bbb | 2216 | list_add_tail(&p->sibling, &p->real_parent->children); |
5e85d4ab | 2217 | list_add_tail_rcu(&p->tasks, &init_task.tasks); |
6883f81a | 2218 | attach_pid(p, PIDTYPE_TGID); |
81907739 ON |
2219 | attach_pid(p, PIDTYPE_PGID); |
2220 | attach_pid(p, PIDTYPE_SID); | |
909ea964 | 2221 | __this_cpu_inc(process_counts); |
80628ca0 ON |
2222 | } else { |
2223 | current->signal->nr_threads++; | |
2224 | atomic_inc(¤t->signal->live); | |
60d4de3f | 2225 | refcount_inc(¤t->signal->sigcnt); |
924de3b8 | 2226 | task_join_group_stop(p); |
80628ca0 ON |
2227 | list_add_tail_rcu(&p->thread_group, |
2228 | &p->group_leader->thread_group); | |
0c740d0a ON |
2229 | list_add_tail_rcu(&p->thread_node, |
2230 | &p->signal->thread_head); | |
73b9ebfe | 2231 | } |
81907739 | 2232 | attach_pid(p, PIDTYPE_PID); |
73b9ebfe | 2233 | nr_threads++; |
1da177e4 | 2234 | } |
1da177e4 | 2235 | total_forks++; |
c3ad2c3b | 2236 | hlist_del_init(&delayed.node); |
3f17da69 | 2237 | spin_unlock(¤t->sighand->siglock); |
4af4206b | 2238 | syscall_tracepoint_update(p); |
1da177e4 | 2239 | write_unlock_irq(&tasklist_lock); |
4af4206b | 2240 | |
c13cf856 | 2241 | proc_fork_connector(p); |
b53202e6 | 2242 | cgroup_post_fork(p); |
780de9dd | 2243 | cgroup_threadgroup_change_end(current); |
cdd6c482 | 2244 | perf_event_fork(p); |
43d2b113 KH |
2245 | |
2246 | trace_task_newtask(p, clone_flags); | |
3ab67966 | 2247 | uprobe_copy_process(p, clone_flags); |
43d2b113 | 2248 | |
1da177e4 LT |
2249 | return p; |
2250 | ||
7e47682e | 2251 | bad_fork_cancel_cgroup: |
3fd37226 KT |
2252 | spin_unlock(¤t->sighand->siglock); |
2253 | write_unlock_irq(&tasklist_lock); | |
b53202e6 | 2254 | cgroup_cancel_fork(p); |
c3b7112d CB |
2255 | bad_fork_cgroup_threadgroup_change_end: |
2256 | cgroup_threadgroup_change_end(current); | |
b3e58382 | 2257 | bad_fork_put_pidfd: |
6fd2fe49 AV |
2258 | if (clone_flags & CLONE_PIDFD) { |
2259 | fput(pidfile); | |
2260 | put_unused_fd(pidfd); | |
2261 | } | |
425fb2b4 PE |
2262 | bad_fork_free_pid: |
2263 | if (pid != &init_struct_pid) | |
2264 | free_pid(pid); | |
0740aa5f JS |
2265 | bad_fork_cleanup_thread: |
2266 | exit_thread(p); | |
fd0928df | 2267 | bad_fork_cleanup_io: |
b69f2292 LR |
2268 | if (p->io_context) |
2269 | exit_io_context(p); | |
ab516013 | 2270 | bad_fork_cleanup_namespaces: |
444f378b | 2271 | exit_task_namespaces(p); |
1da177e4 | 2272 | bad_fork_cleanup_mm: |
c3f3ce04 AA |
2273 | if (p->mm) { |
2274 | mm_clear_owner(p->mm, p); | |
1da177e4 | 2275 | mmput(p->mm); |
c3f3ce04 | 2276 | } |
1da177e4 | 2277 | bad_fork_cleanup_signal: |
4ab6c083 | 2278 | if (!(clone_flags & CLONE_THREAD)) |
1c5354de | 2279 | free_signal_struct(p->signal); |
1da177e4 | 2280 | bad_fork_cleanup_sighand: |
a7e5328a | 2281 | __cleanup_sighand(p->sighand); |
1da177e4 LT |
2282 | bad_fork_cleanup_fs: |
2283 | exit_fs(p); /* blocking */ | |
2284 | bad_fork_cleanup_files: | |
2285 | exit_files(p); /* blocking */ | |
2286 | bad_fork_cleanup_semundo: | |
2287 | exit_sem(p); | |
e4e55b47 TH |
2288 | bad_fork_cleanup_security: |
2289 | security_task_free(p); | |
1da177e4 LT |
2290 | bad_fork_cleanup_audit: |
2291 | audit_free(p); | |
6c72e350 | 2292 | bad_fork_cleanup_perf: |
cdd6c482 | 2293 | perf_event_free_task(p); |
6c72e350 | 2294 | bad_fork_cleanup_policy: |
b09be676 | 2295 | lockdep_free_task(p); |
1da177e4 | 2296 | #ifdef CONFIG_NUMA |
f0be3d32 | 2297 | mpol_put(p->mempolicy); |
e8604cb4 | 2298 | bad_fork_cleanup_threadgroup_lock: |
1da177e4 | 2299 | #endif |
35df17c5 | 2300 | delayacct_tsk_free(p); |
1da177e4 | 2301 | bad_fork_cleanup_count: |
d84f4f99 | 2302 | atomic_dec(&p->cred->user->processes); |
e0e81739 | 2303 | exit_creds(p); |
1da177e4 | 2304 | bad_fork_free: |
405c0759 | 2305 | p->state = TASK_DEAD; |
68f24b08 | 2306 | put_task_stack(p); |
c3f3ce04 | 2307 | delayed_free_task(p); |
fe7d37d1 | 2308 | fork_out: |
c3ad2c3b EB |
2309 | spin_lock_irq(¤t->sighand->siglock); |
2310 | hlist_del_init(&delayed.node); | |
2311 | spin_unlock_irq(¤t->sighand->siglock); | |
fe7d37d1 | 2312 | return ERR_PTR(retval); |
1da177e4 LT |
2313 | } |
2314 | ||
2c470475 | 2315 | static inline void init_idle_pids(struct task_struct *idle) |
f106eee1 ON |
2316 | { |
2317 | enum pid_type type; | |
2318 | ||
2319 | for (type = PIDTYPE_PID; type < PIDTYPE_MAX; ++type) { | |
2c470475 EB |
2320 | INIT_HLIST_NODE(&idle->pid_links[type]); /* not really needed */ |
2321 | init_task_pid(idle, type, &init_struct_pid); | |
f106eee1 ON |
2322 | } |
2323 | } | |
2324 | ||
0db0628d | 2325 | struct task_struct *fork_idle(int cpu) |
1da177e4 | 2326 | { |
36c8b586 | 2327 | struct task_struct *task; |
7f192e3c CB |
2328 | struct kernel_clone_args args = { |
2329 | .flags = CLONE_VM, | |
2330 | }; | |
2331 | ||
2332 | task = copy_process(&init_struct_pid, 0, cpu_to_node(cpu), &args); | |
f106eee1 | 2333 | if (!IS_ERR(task)) { |
2c470475 | 2334 | init_idle_pids(task); |
753ca4f3 | 2335 | init_idle(task, cpu); |
f106eee1 | 2336 | } |
73b9ebfe | 2337 | |
1da177e4 LT |
2338 | return task; |
2339 | } | |
2340 | ||
13585fa0 NA |
2341 | struct mm_struct *copy_init_mm(void) |
2342 | { | |
2343 | return dup_mm(NULL, &init_mm); | |
2344 | } | |
2345 | ||
1da177e4 LT |
2346 | /* |
2347 | * Ok, this is the main fork-routine. | |
2348 | * | |
2349 | * It copies the process, and if successful kick-starts | |
2350 | * it and waits for it to finish using the VM if required. | |
a0eb9abd ES |
2351 | * |
2352 | * args->exit_signal is expected to be checked for sanity by the caller. | |
1da177e4 | 2353 | */ |
7f192e3c | 2354 | long _do_fork(struct kernel_clone_args *args) |
1da177e4 | 2355 | { |
7f192e3c | 2356 | u64 clone_flags = args->flags; |
9f5325aa MPS |
2357 | struct completion vfork; |
2358 | struct pid *pid; | |
1da177e4 LT |
2359 | struct task_struct *p; |
2360 | int trace = 0; | |
92476d7f | 2361 | long nr; |
1da177e4 | 2362 | |
09a05394 | 2363 | /* |
4b9d33e6 TH |
2364 | * Determine whether and which event to report to ptracer. When |
2365 | * called from kernel_thread or CLONE_UNTRACED is explicitly | |
2366 | * requested, no event is reported; otherwise, report if the event | |
2367 | * for the type of forking is enabled. | |
09a05394 | 2368 | */ |
e80d6661 | 2369 | if (!(clone_flags & CLONE_UNTRACED)) { |
4b9d33e6 TH |
2370 | if (clone_flags & CLONE_VFORK) |
2371 | trace = PTRACE_EVENT_VFORK; | |
7f192e3c | 2372 | else if (args->exit_signal != SIGCHLD) |
4b9d33e6 TH |
2373 | trace = PTRACE_EVENT_CLONE; |
2374 | else | |
2375 | trace = PTRACE_EVENT_FORK; | |
2376 | ||
2377 | if (likely(!ptrace_event_enabled(current, trace))) | |
2378 | trace = 0; | |
2379 | } | |
1da177e4 | 2380 | |
7f192e3c | 2381 | p = copy_process(NULL, trace, NUMA_NO_NODE, args); |
38addce8 | 2382 | add_latent_entropy(); |
9f5325aa MPS |
2383 | |
2384 | if (IS_ERR(p)) | |
2385 | return PTR_ERR(p); | |
2386 | ||
1da177e4 LT |
2387 | /* |
2388 | * Do this prior waking up the new thread - the thread pointer | |
2389 | * might get invalid after that point, if the thread exits quickly. | |
2390 | */ | |
9f5325aa | 2391 | trace_sched_process_fork(current, p); |
0a16b607 | 2392 | |
9f5325aa MPS |
2393 | pid = get_task_pid(p, PIDTYPE_PID); |
2394 | nr = pid_vnr(pid); | |
30e49c26 | 2395 | |
9f5325aa | 2396 | if (clone_flags & CLONE_PARENT_SETTID) |
7f192e3c | 2397 | put_user(nr, args->parent_tid); |
a6f5e063 | 2398 | |
9f5325aa MPS |
2399 | if (clone_flags & CLONE_VFORK) { |
2400 | p->vfork_done = &vfork; | |
2401 | init_completion(&vfork); | |
2402 | get_task_struct(p); | |
2403 | } | |
1da177e4 | 2404 | |
9f5325aa | 2405 | wake_up_new_task(p); |
09a05394 | 2406 | |
9f5325aa MPS |
2407 | /* forking complete and child started to run, tell ptracer */ |
2408 | if (unlikely(trace)) | |
2409 | ptrace_event_pid(trace, pid); | |
4e52365f | 2410 | |
9f5325aa MPS |
2411 | if (clone_flags & CLONE_VFORK) { |
2412 | if (!wait_for_vfork_done(p, &vfork)) | |
2413 | ptrace_event_pid(PTRACE_EVENT_VFORK_DONE, pid); | |
1da177e4 | 2414 | } |
9f5325aa MPS |
2415 | |
2416 | put_pid(pid); | |
92476d7f | 2417 | return nr; |
1da177e4 LT |
2418 | } |
2419 | ||
028b6e8a DL |
2420 | bool legacy_clone_args_valid(const struct kernel_clone_args *kargs) |
2421 | { | |
2422 | /* clone(CLONE_PIDFD) uses parent_tidptr to return a pidfd */ | |
2423 | if ((kargs->flags & CLONE_PIDFD) && | |
2424 | (kargs->flags & CLONE_PARENT_SETTID)) | |
2425 | return false; | |
2426 | ||
2427 | return true; | |
2428 | } | |
2429 | ||
3033f14a JT |
2430 | #ifndef CONFIG_HAVE_COPY_THREAD_TLS |
2431 | /* For compatibility with architectures that call do_fork directly rather than | |
2432 | * using the syscall entry points below. */ | |
2433 | long do_fork(unsigned long clone_flags, | |
2434 | unsigned long stack_start, | |
2435 | unsigned long stack_size, | |
2436 | int __user *parent_tidptr, | |
2437 | int __user *child_tidptr) | |
2438 | { | |
7f192e3c CB |
2439 | struct kernel_clone_args args = { |
2440 | .flags = (clone_flags & ~CSIGNAL), | |
028b6e8a | 2441 | .pidfd = parent_tidptr, |
7f192e3c CB |
2442 | .child_tid = child_tidptr, |
2443 | .parent_tid = parent_tidptr, | |
2444 | .exit_signal = (clone_flags & CSIGNAL), | |
2445 | .stack = stack_start, | |
2446 | .stack_size = stack_size, | |
2447 | }; | |
2448 | ||
028b6e8a DL |
2449 | if (!legacy_clone_args_valid(&args)) |
2450 | return -EINVAL; | |
2451 | ||
7f192e3c | 2452 | return _do_fork(&args); |
3033f14a JT |
2453 | } |
2454 | #endif | |
2455 | ||
2aa3a7f8 AV |
2456 | /* |
2457 | * Create a kernel thread. | |
2458 | */ | |
2459 | pid_t kernel_thread(int (*fn)(void *), void *arg, unsigned long flags) | |
2460 | { | |
7f192e3c CB |
2461 | struct kernel_clone_args args = { |
2462 | .flags = ((flags | CLONE_VM | CLONE_UNTRACED) & ~CSIGNAL), | |
2463 | .exit_signal = (flags & CSIGNAL), | |
2464 | .stack = (unsigned long)fn, | |
2465 | .stack_size = (unsigned long)arg, | |
2466 | }; | |
2467 | ||
2468 | return _do_fork(&args); | |
2aa3a7f8 | 2469 | } |
2aa3a7f8 | 2470 | |
d2125043 AV |
2471 | #ifdef __ARCH_WANT_SYS_FORK |
2472 | SYSCALL_DEFINE0(fork) | |
2473 | { | |
2474 | #ifdef CONFIG_MMU | |
7f192e3c CB |
2475 | struct kernel_clone_args args = { |
2476 | .exit_signal = SIGCHLD, | |
2477 | }; | |
2478 | ||
2479 | return _do_fork(&args); | |
d2125043 AV |
2480 | #else |
2481 | /* can not support in nommu mode */ | |
5d59e182 | 2482 | return -EINVAL; |
d2125043 AV |
2483 | #endif |
2484 | } | |
2485 | #endif | |
2486 | ||
2487 | #ifdef __ARCH_WANT_SYS_VFORK | |
2488 | SYSCALL_DEFINE0(vfork) | |
2489 | { | |
7f192e3c CB |
2490 | struct kernel_clone_args args = { |
2491 | .flags = CLONE_VFORK | CLONE_VM, | |
2492 | .exit_signal = SIGCHLD, | |
2493 | }; | |
2494 | ||
2495 | return _do_fork(&args); | |
d2125043 AV |
2496 | } |
2497 | #endif | |
2498 | ||
2499 | #ifdef __ARCH_WANT_SYS_CLONE | |
2500 | #ifdef CONFIG_CLONE_BACKWARDS | |
2501 | SYSCALL_DEFINE5(clone, unsigned long, clone_flags, unsigned long, newsp, | |
2502 | int __user *, parent_tidptr, | |
3033f14a | 2503 | unsigned long, tls, |
d2125043 AV |
2504 | int __user *, child_tidptr) |
2505 | #elif defined(CONFIG_CLONE_BACKWARDS2) | |
2506 | SYSCALL_DEFINE5(clone, unsigned long, newsp, unsigned long, clone_flags, | |
2507 | int __user *, parent_tidptr, | |
2508 | int __user *, child_tidptr, | |
3033f14a | 2509 | unsigned long, tls) |
dfa9771a MS |
2510 | #elif defined(CONFIG_CLONE_BACKWARDS3) |
2511 | SYSCALL_DEFINE6(clone, unsigned long, clone_flags, unsigned long, newsp, | |
2512 | int, stack_size, | |
2513 | int __user *, parent_tidptr, | |
2514 | int __user *, child_tidptr, | |
3033f14a | 2515 | unsigned long, tls) |
d2125043 AV |
2516 | #else |
2517 | SYSCALL_DEFINE5(clone, unsigned long, clone_flags, unsigned long, newsp, | |
2518 | int __user *, parent_tidptr, | |
2519 | int __user *, child_tidptr, | |
3033f14a | 2520 | unsigned long, tls) |
d2125043 AV |
2521 | #endif |
2522 | { | |
7f192e3c CB |
2523 | struct kernel_clone_args args = { |
2524 | .flags = (clone_flags & ~CSIGNAL), | |
2525 | .pidfd = parent_tidptr, | |
2526 | .child_tid = child_tidptr, | |
2527 | .parent_tid = parent_tidptr, | |
2528 | .exit_signal = (clone_flags & CSIGNAL), | |
2529 | .stack = newsp, | |
2530 | .tls = tls, | |
2531 | }; | |
2532 | ||
028b6e8a | 2533 | if (!legacy_clone_args_valid(&args)) |
7f192e3c CB |
2534 | return -EINVAL; |
2535 | ||
2536 | return _do_fork(&args); | |
2537 | } | |
d68dbb0c | 2538 | #endif |
7f192e3c | 2539 | |
d68dbb0c | 2540 | #ifdef __ARCH_WANT_SYS_CLONE3 |
a47781f0 AA |
2541 | |
2542 | /* | |
2543 | * copy_thread implementations handle CLONE_SETTLS by reading the TLS value from | |
2544 | * the registers containing the syscall arguments for clone. This doesn't work | |
2545 | * with clone3 since the TLS value is passed in clone_args instead. | |
2546 | */ | |
2547 | #ifndef CONFIG_HAVE_COPY_THREAD_TLS | |
2548 | #error clone3 requires copy_thread_tls support in arch | |
2549 | #endif | |
2550 | ||
7f192e3c CB |
2551 | noinline static int copy_clone_args_from_user(struct kernel_clone_args *kargs, |
2552 | struct clone_args __user *uargs, | |
2553 | size_t size) | |
2554 | { | |
2555 | struct clone_args args; | |
2556 | ||
2557 | if (unlikely(size > PAGE_SIZE)) | |
2558 | return -E2BIG; | |
2559 | ||
2560 | if (unlikely(size < sizeof(struct clone_args))) | |
2561 | return -EINVAL; | |
2562 | ||
2563 | if (unlikely(!access_ok(uargs, size))) | |
2564 | return -EFAULT; | |
2565 | ||
2566 | if (size > sizeof(struct clone_args)) { | |
2567 | unsigned char __user *addr; | |
2568 | unsigned char __user *end; | |
2569 | unsigned char val; | |
2570 | ||
2571 | addr = (void __user *)uargs + sizeof(struct clone_args); | |
2572 | end = (void __user *)uargs + size; | |
2573 | ||
2574 | for (; addr < end; addr++) { | |
2575 | if (get_user(val, addr)) | |
2576 | return -EFAULT; | |
2577 | if (val) | |
2578 | return -E2BIG; | |
2579 | } | |
2580 | ||
2581 | size = sizeof(struct clone_args); | |
2582 | } | |
2583 | ||
2584 | if (copy_from_user(&args, uargs, size)) | |
2585 | return -EFAULT; | |
2586 | ||
a0eb9abd ES |
2587 | /* |
2588 | * Verify that higher 32bits of exit_signal are unset and that | |
2589 | * it is a valid signal | |
2590 | */ | |
2591 | if (unlikely((args.exit_signal & ~((u64)CSIGNAL)) || | |
2592 | !valid_signal(args.exit_signal))) | |
2593 | return -EINVAL; | |
2594 | ||
7f192e3c CB |
2595 | *kargs = (struct kernel_clone_args){ |
2596 | .flags = args.flags, | |
2597 | .pidfd = u64_to_user_ptr(args.pidfd), | |
2598 | .child_tid = u64_to_user_ptr(args.child_tid), | |
2599 | .parent_tid = u64_to_user_ptr(args.parent_tid), | |
2600 | .exit_signal = args.exit_signal, | |
2601 | .stack = args.stack, | |
2602 | .stack_size = args.stack_size, | |
2603 | .tls = args.tls, | |
2604 | }; | |
2605 | ||
2606 | return 0; | |
2607 | } | |
2608 | ||
f5316b81 CB |
2609 | /** |
2610 | * clone3_stack_valid - check and prepare stack | |
2611 | * @kargs: kernel clone args | |
2612 | * | |
2613 | * Verify that the stack arguments userspace gave us are sane. | |
2614 | * In addition, set the stack direction for userspace since it's easy for us to | |
2615 | * determine. | |
2616 | */ | |
2617 | static inline bool clone3_stack_valid(struct kernel_clone_args *kargs) | |
2618 | { | |
2619 | if (kargs->stack == 0) { | |
2620 | if (kargs->stack_size > 0) | |
2621 | return false; | |
2622 | } else { | |
2623 | if (kargs->stack_size == 0) | |
2624 | return false; | |
2625 | ||
2626 | if (!access_ok((void __user *)kargs->stack, kargs->stack_size)) | |
2627 | return false; | |
2628 | ||
2629 | #if !defined(CONFIG_STACK_GROWSUP) && !defined(CONFIG_IA64) | |
2630 | kargs->stack += kargs->stack_size; | |
2631 | #endif | |
2632 | } | |
2633 | ||
2634 | return true; | |
2635 | } | |
2636 | ||
2637 | static bool clone3_args_valid(struct kernel_clone_args *kargs) | |
7f192e3c CB |
2638 | { |
2639 | /* | |
2640 | * All lower bits of the flag word are taken. | |
2641 | * Verify that no other unknown flags are passed along. | |
2642 | */ | |
2643 | if (kargs->flags & ~CLONE_LEGACY_FLAGS) | |
2644 | return false; | |
2645 | ||
2646 | /* | |
2647 | * - make the CLONE_DETACHED bit reuseable for clone3 | |
2648 | * - make the CSIGNAL bits reuseable for clone3 | |
2649 | */ | |
2650 | if (kargs->flags & (CLONE_DETACHED | CSIGNAL)) | |
2651 | return false; | |
2652 | ||
2653 | if ((kargs->flags & (CLONE_THREAD | CLONE_PARENT)) && | |
2654 | kargs->exit_signal) | |
2655 | return false; | |
2656 | ||
f5316b81 CB |
2657 | if (!clone3_stack_valid(kargs)) |
2658 | return false; | |
2659 | ||
7f192e3c CB |
2660 | return true; |
2661 | } | |
2662 | ||
2663 | SYSCALL_DEFINE2(clone3, struct clone_args __user *, uargs, size_t, size) | |
2664 | { | |
2665 | int err; | |
2666 | ||
2667 | struct kernel_clone_args kargs; | |
2668 | ||
2669 | err = copy_clone_args_from_user(&kargs, uargs, size); | |
2670 | if (err) | |
2671 | return err; | |
2672 | ||
2673 | if (!clone3_args_valid(&kargs)) | |
2674 | return -EINVAL; | |
2675 | ||
2676 | return _do_fork(&kargs); | |
d2125043 AV |
2677 | } |
2678 | #endif | |
2679 | ||
0f1b92cb ON |
2680 | void walk_process_tree(struct task_struct *top, proc_visitor visitor, void *data) |
2681 | { | |
2682 | struct task_struct *leader, *parent, *child; | |
2683 | int res; | |
2684 | ||
2685 | read_lock(&tasklist_lock); | |
2686 | leader = top = top->group_leader; | |
2687 | down: | |
2688 | for_each_thread(leader, parent) { | |
2689 | list_for_each_entry(child, &parent->children, sibling) { | |
2690 | res = visitor(child, data); | |
2691 | if (res) { | |
2692 | if (res < 0) | |
2693 | goto out; | |
2694 | leader = child; | |
2695 | goto down; | |
2696 | } | |
2697 | up: | |
2698 | ; | |
2699 | } | |
2700 | } | |
2701 | ||
2702 | if (leader != top) { | |
2703 | child = leader; | |
2704 | parent = child->real_parent; | |
2705 | leader = parent->group_leader; | |
2706 | goto up; | |
2707 | } | |
2708 | out: | |
2709 | read_unlock(&tasklist_lock); | |
2710 | } | |
2711 | ||
5fd63b30 RT |
2712 | #ifndef ARCH_MIN_MMSTRUCT_ALIGN |
2713 | #define ARCH_MIN_MMSTRUCT_ALIGN 0 | |
2714 | #endif | |
2715 | ||
51cc5068 | 2716 | static void sighand_ctor(void *data) |
aa1757f9 ON |
2717 | { |
2718 | struct sighand_struct *sighand = data; | |
2719 | ||
a35afb83 | 2720 | spin_lock_init(&sighand->siglock); |
b8fceee1 | 2721 | init_waitqueue_head(&sighand->signalfd_wqh); |
aa1757f9 ON |
2722 | } |
2723 | ||
1da177e4 LT |
2724 | void __init proc_caches_init(void) |
2725 | { | |
c1a2f7f0 RR |
2726 | unsigned int mm_size; |
2727 | ||
1da177e4 LT |
2728 | sighand_cachep = kmem_cache_create("sighand_cache", |
2729 | sizeof(struct sighand_struct), 0, | |
5f0d5a3a | 2730 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_TYPESAFE_BY_RCU| |
75f296d9 | 2731 | SLAB_ACCOUNT, sighand_ctor); |
1da177e4 LT |
2732 | signal_cachep = kmem_cache_create("signal_cache", |
2733 | sizeof(struct signal_struct), 0, | |
75f296d9 | 2734 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_ACCOUNT, |
5d097056 | 2735 | NULL); |
20c2df83 | 2736 | files_cachep = kmem_cache_create("files_cache", |
1da177e4 | 2737 | sizeof(struct files_struct), 0, |
75f296d9 | 2738 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_ACCOUNT, |
5d097056 | 2739 | NULL); |
20c2df83 | 2740 | fs_cachep = kmem_cache_create("fs_cache", |
1da177e4 | 2741 | sizeof(struct fs_struct), 0, |
75f296d9 | 2742 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_ACCOUNT, |
5d097056 | 2743 | NULL); |
c1a2f7f0 | 2744 | |
6345d24d | 2745 | /* |
c1a2f7f0 RR |
2746 | * The mm_cpumask is located at the end of mm_struct, and is |
2747 | * dynamically sized based on the maximum CPU number this system | |
2748 | * can have, taking hotplug into account (nr_cpu_ids). | |
6345d24d | 2749 | */ |
c1a2f7f0 RR |
2750 | mm_size = sizeof(struct mm_struct) + cpumask_size(); |
2751 | ||
07dcd7fe | 2752 | mm_cachep = kmem_cache_create_usercopy("mm_struct", |
c1a2f7f0 | 2753 | mm_size, ARCH_MIN_MMSTRUCT_ALIGN, |
75f296d9 | 2754 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_ACCOUNT, |
07dcd7fe DW |
2755 | offsetof(struct mm_struct, saved_auxv), |
2756 | sizeof_field(struct mm_struct, saved_auxv), | |
5d097056 VD |
2757 | NULL); |
2758 | vm_area_cachep = KMEM_CACHE(vm_area_struct, SLAB_PANIC|SLAB_ACCOUNT); | |
8feae131 | 2759 | mmap_init(); |
66577193 | 2760 | nsproxy_cache_init(); |
1da177e4 | 2761 | } |
cf2e340f | 2762 | |
cf2e340f | 2763 | /* |
9bfb23fc | 2764 | * Check constraints on flags passed to the unshare system call. |
cf2e340f | 2765 | */ |
9bfb23fc | 2766 | static int check_unshare_flags(unsigned long unshare_flags) |
cf2e340f | 2767 | { |
9bfb23fc ON |
2768 | if (unshare_flags & ~(CLONE_THREAD|CLONE_FS|CLONE_NEWNS|CLONE_SIGHAND| |
2769 | CLONE_VM|CLONE_FILES|CLONE_SYSVSEM| | |
50804fe3 | 2770 | CLONE_NEWUTS|CLONE_NEWIPC|CLONE_NEWNET| |
a79a908f | 2771 | CLONE_NEWUSER|CLONE_NEWPID|CLONE_NEWCGROUP)) |
9bfb23fc | 2772 | return -EINVAL; |
cf2e340f | 2773 | /* |
12c641ab EB |
2774 | * Not implemented, but pretend it works if there is nothing |
2775 | * to unshare. Note that unsharing the address space or the | |
2776 | * signal handlers also need to unshare the signal queues (aka | |
2777 | * CLONE_THREAD). | |
cf2e340f | 2778 | */ |
9bfb23fc | 2779 | if (unshare_flags & (CLONE_THREAD | CLONE_SIGHAND | CLONE_VM)) { |
12c641ab EB |
2780 | if (!thread_group_empty(current)) |
2781 | return -EINVAL; | |
2782 | } | |
2783 | if (unshare_flags & (CLONE_SIGHAND | CLONE_VM)) { | |
d036bda7 | 2784 | if (refcount_read(¤t->sighand->count) > 1) |
12c641ab EB |
2785 | return -EINVAL; |
2786 | } | |
2787 | if (unshare_flags & CLONE_VM) { | |
2788 | if (!current_is_single_threaded()) | |
9bfb23fc ON |
2789 | return -EINVAL; |
2790 | } | |
cf2e340f JD |
2791 | |
2792 | return 0; | |
2793 | } | |
2794 | ||
2795 | /* | |
99d1419d | 2796 | * Unshare the filesystem structure if it is being shared |
cf2e340f JD |
2797 | */ |
2798 | static int unshare_fs(unsigned long unshare_flags, struct fs_struct **new_fsp) | |
2799 | { | |
2800 | struct fs_struct *fs = current->fs; | |
2801 | ||
498052bb AV |
2802 | if (!(unshare_flags & CLONE_FS) || !fs) |
2803 | return 0; | |
2804 | ||
2805 | /* don't need lock here; in the worst case we'll do useless copy */ | |
2806 | if (fs->users == 1) | |
2807 | return 0; | |
2808 | ||
2809 | *new_fsp = copy_fs_struct(fs); | |
2810 | if (!*new_fsp) | |
2811 | return -ENOMEM; | |
cf2e340f JD |
2812 | |
2813 | return 0; | |
2814 | } | |
2815 | ||
cf2e340f | 2816 | /* |
a016f338 | 2817 | * Unshare file descriptor table if it is being shared |
cf2e340f JD |
2818 | */ |
2819 | static int unshare_fd(unsigned long unshare_flags, struct files_struct **new_fdp) | |
2820 | { | |
2821 | struct files_struct *fd = current->files; | |
a016f338 | 2822 | int error = 0; |
cf2e340f JD |
2823 | |
2824 | if ((unshare_flags & CLONE_FILES) && | |
a016f338 JD |
2825 | (fd && atomic_read(&fd->count) > 1)) { |
2826 | *new_fdp = dup_fd(fd, &error); | |
2827 | if (!*new_fdp) | |
2828 | return error; | |
2829 | } | |
cf2e340f JD |
2830 | |
2831 | return 0; | |
2832 | } | |
2833 | ||
cf2e340f JD |
2834 | /* |
2835 | * unshare allows a process to 'unshare' part of the process | |
2836 | * context which was originally shared using clone. copy_* | |
2837 | * functions used by do_fork() cannot be used here directly | |
2838 | * because they modify an inactive task_struct that is being | |
2839 | * constructed. Here we are modifying the current, active, | |
2840 | * task_struct. | |
2841 | */ | |
9b32105e | 2842 | int ksys_unshare(unsigned long unshare_flags) |
cf2e340f | 2843 | { |
cf2e340f | 2844 | struct fs_struct *fs, *new_fs = NULL; |
cf2e340f | 2845 | struct files_struct *fd, *new_fd = NULL; |
b2e0d987 | 2846 | struct cred *new_cred = NULL; |
cf7b708c | 2847 | struct nsproxy *new_nsproxy = NULL; |
9edff4ab | 2848 | int do_sysvsem = 0; |
9bfb23fc | 2849 | int err; |
cf2e340f | 2850 | |
b2e0d987 | 2851 | /* |
faf00da5 EB |
2852 | * If unsharing a user namespace must also unshare the thread group |
2853 | * and unshare the filesystem root and working directories. | |
b2e0d987 EB |
2854 | */ |
2855 | if (unshare_flags & CLONE_NEWUSER) | |
e66eded8 | 2856 | unshare_flags |= CLONE_THREAD | CLONE_FS; |
50804fe3 EB |
2857 | /* |
2858 | * If unsharing vm, must also unshare signal handlers. | |
2859 | */ | |
2860 | if (unshare_flags & CLONE_VM) | |
2861 | unshare_flags |= CLONE_SIGHAND; | |
12c641ab EB |
2862 | /* |
2863 | * If unsharing a signal handlers, must also unshare the signal queues. | |
2864 | */ | |
2865 | if (unshare_flags & CLONE_SIGHAND) | |
2866 | unshare_flags |= CLONE_THREAD; | |
9bfb23fc ON |
2867 | /* |
2868 | * If unsharing namespace, must also unshare filesystem information. | |
2869 | */ | |
2870 | if (unshare_flags & CLONE_NEWNS) | |
2871 | unshare_flags |= CLONE_FS; | |
50804fe3 | 2872 | |
2b612266 SH |
2873 | if ((unshare_flags & CLONE_NEWUSER) && !unprivileged_userns_clone) { |
2874 | err = -EPERM; | |
2875 | if (!capable(CAP_SYS_ADMIN)) | |
2876 | goto bad_unshare_out; | |
2877 | } | |
2878 | ||
50804fe3 EB |
2879 | err = check_unshare_flags(unshare_flags); |
2880 | if (err) | |
2881 | goto bad_unshare_out; | |
6013f67f MS |
2882 | /* |
2883 | * CLONE_NEWIPC must also detach from the undolist: after switching | |
2884 | * to a new ipc namespace, the semaphore arrays from the old | |
2885 | * namespace are unreachable. | |
2886 | */ | |
2887 | if (unshare_flags & (CLONE_NEWIPC|CLONE_SYSVSEM)) | |
9edff4ab | 2888 | do_sysvsem = 1; |
fb0a685c DRO |
2889 | err = unshare_fs(unshare_flags, &new_fs); |
2890 | if (err) | |
9bfb23fc | 2891 | goto bad_unshare_out; |
fb0a685c DRO |
2892 | err = unshare_fd(unshare_flags, &new_fd); |
2893 | if (err) | |
9bfb23fc | 2894 | goto bad_unshare_cleanup_fs; |
b2e0d987 | 2895 | err = unshare_userns(unshare_flags, &new_cred); |
fb0a685c | 2896 | if (err) |
9edff4ab | 2897 | goto bad_unshare_cleanup_fd; |
b2e0d987 EB |
2898 | err = unshare_nsproxy_namespaces(unshare_flags, &new_nsproxy, |
2899 | new_cred, new_fs); | |
2900 | if (err) | |
2901 | goto bad_unshare_cleanup_cred; | |
c0b2fc31 | 2902 | |
b2e0d987 | 2903 | if (new_fs || new_fd || do_sysvsem || new_cred || new_nsproxy) { |
9edff4ab MS |
2904 | if (do_sysvsem) { |
2905 | /* | |
2906 | * CLONE_SYSVSEM is equivalent to sys_exit(). | |
2907 | */ | |
2908 | exit_sem(current); | |
2909 | } | |
ab602f79 JM |
2910 | if (unshare_flags & CLONE_NEWIPC) { |
2911 | /* Orphan segments in old ns (see sem above). */ | |
2912 | exit_shm(current); | |
2913 | shm_init_task(current); | |
2914 | } | |
ab516013 | 2915 | |
6f977e6b | 2916 | if (new_nsproxy) |
cf7b708c | 2917 | switch_task_namespaces(current, new_nsproxy); |
cf2e340f | 2918 | |
cf7b708c PE |
2919 | task_lock(current); |
2920 | ||
cf2e340f JD |
2921 | if (new_fs) { |
2922 | fs = current->fs; | |
2a4419b5 | 2923 | spin_lock(&fs->lock); |
cf2e340f | 2924 | current->fs = new_fs; |
498052bb AV |
2925 | if (--fs->users) |
2926 | new_fs = NULL; | |
2927 | else | |
2928 | new_fs = fs; | |
2a4419b5 | 2929 | spin_unlock(&fs->lock); |
cf2e340f JD |
2930 | } |
2931 | ||
cf2e340f JD |
2932 | if (new_fd) { |
2933 | fd = current->files; | |
2934 | current->files = new_fd; | |
2935 | new_fd = fd; | |
2936 | } | |
2937 | ||
2938 | task_unlock(current); | |
b2e0d987 EB |
2939 | |
2940 | if (new_cred) { | |
2941 | /* Install the new user namespace */ | |
2942 | commit_creds(new_cred); | |
2943 | new_cred = NULL; | |
2944 | } | |
cf2e340f JD |
2945 | } |
2946 | ||
e4222673 HB |
2947 | perf_event_namespaces(current); |
2948 | ||
b2e0d987 EB |
2949 | bad_unshare_cleanup_cred: |
2950 | if (new_cred) | |
2951 | put_cred(new_cred); | |
cf2e340f JD |
2952 | bad_unshare_cleanup_fd: |
2953 | if (new_fd) | |
2954 | put_files_struct(new_fd); | |
2955 | ||
cf2e340f JD |
2956 | bad_unshare_cleanup_fs: |
2957 | if (new_fs) | |
498052bb | 2958 | free_fs_struct(new_fs); |
cf2e340f | 2959 | |
cf2e340f JD |
2960 | bad_unshare_out: |
2961 | return err; | |
2962 | } | |
3b125388 | 2963 | |
9b32105e DB |
2964 | SYSCALL_DEFINE1(unshare, unsigned long, unshare_flags) |
2965 | { | |
2966 | return ksys_unshare(unshare_flags); | |
2967 | } | |
2968 | ||
3b125388 AV |
2969 | /* |
2970 | * Helper to unshare the files of the current task. | |
2971 | * We don't want to expose copy_files internals to | |
2972 | * the exec layer of the kernel. | |
2973 | */ | |
2974 | ||
2975 | int unshare_files(struct files_struct **displaced) | |
2976 | { | |
2977 | struct task_struct *task = current; | |
50704516 | 2978 | struct files_struct *copy = NULL; |
3b125388 AV |
2979 | int error; |
2980 | ||
2981 | error = unshare_fd(CLONE_FILES, ©); | |
2982 | if (error || !copy) { | |
2983 | *displaced = NULL; | |
2984 | return error; | |
2985 | } | |
2986 | *displaced = task->files; | |
2987 | task_lock(task); | |
2988 | task->files = copy; | |
2989 | task_unlock(task); | |
2990 | return 0; | |
2991 | } | |
16db3d3f HS |
2992 | |
2993 | int sysctl_max_threads(struct ctl_table *table, int write, | |
2994 | void __user *buffer, size_t *lenp, loff_t *ppos) | |
2995 | { | |
2996 | struct ctl_table t; | |
2997 | int ret; | |
2998 | int threads = max_threads; | |
8e597517 | 2999 | int min = 1; |
16db3d3f HS |
3000 | int max = MAX_THREADS; |
3001 | ||
3002 | t = *table; | |
3003 | t.data = &threads; | |
3004 | t.extra1 = &min; | |
3005 | t.extra2 = &max; | |
3006 | ||
3007 | ret = proc_dointvec_minmax(&t, write, buffer, lenp, ppos); | |
3008 | if (ret || !write) | |
3009 | return ret; | |
3010 | ||
8e597517 | 3011 | max_threads = threads; |
16db3d3f HS |
3012 | |
3013 | return 0; | |
3014 | } |