]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * linux/kernel/fork.c | |
3 | * | |
4 | * Copyright (C) 1991, 1992 Linus Torvalds | |
5 | */ | |
6 | ||
7 | /* | |
8 | * 'fork.c' contains the help-routines for the 'fork' system call | |
9 | * (see also entry.S and others). | |
10 | * Fork is rather simple, once you get the hang of it, but the memory | |
11 | * management can be a bitch. See 'mm/memory.c': 'copy_page_range()' | |
12 | */ | |
13 | ||
1da177e4 | 14 | #include <linux/slab.h> |
4eb5aaa3 | 15 | #include <linux/sched/autogroup.h> |
6e84f315 | 16 | #include <linux/sched/mm.h> |
f7ccbae4 | 17 | #include <linux/sched/coredump.h> |
8703e8a4 | 18 | #include <linux/sched/user.h> |
6a3827d7 | 19 | #include <linux/sched/numa_balancing.h> |
03441a34 | 20 | #include <linux/sched/stat.h> |
29930025 | 21 | #include <linux/sched/task.h> |
68db0cf1 | 22 | #include <linux/sched/task_stack.h> |
32ef5517 | 23 | #include <linux/sched/cputime.h> |
037741a6 | 24 | #include <linux/rtmutex.h> |
1da177e4 LT |
25 | #include <linux/init.h> |
26 | #include <linux/unistd.h> | |
1da177e4 LT |
27 | #include <linux/module.h> |
28 | #include <linux/vmalloc.h> | |
29 | #include <linux/completion.h> | |
1da177e4 LT |
30 | #include <linux/personality.h> |
31 | #include <linux/mempolicy.h> | |
32 | #include <linux/sem.h> | |
33 | #include <linux/file.h> | |
9f3acc31 | 34 | #include <linux/fdtable.h> |
da9cbc87 | 35 | #include <linux/iocontext.h> |
1da177e4 LT |
36 | #include <linux/key.h> |
37 | #include <linux/binfmts.h> | |
38 | #include <linux/mman.h> | |
cddb8a5c | 39 | #include <linux/mmu_notifier.h> |
133ff0ea | 40 | #include <linux/hmm.h> |
1da177e4 | 41 | #include <linux/fs.h> |
615d6e87 DB |
42 | #include <linux/mm.h> |
43 | #include <linux/vmacache.h> | |
ab516013 | 44 | #include <linux/nsproxy.h> |
c59ede7b | 45 | #include <linux/capability.h> |
1da177e4 | 46 | #include <linux/cpu.h> |
b4f48b63 | 47 | #include <linux/cgroup.h> |
1da177e4 | 48 | #include <linux/security.h> |
a1e78772 | 49 | #include <linux/hugetlb.h> |
e2cfabdf | 50 | #include <linux/seccomp.h> |
1da177e4 LT |
51 | #include <linux/swap.h> |
52 | #include <linux/syscalls.h> | |
53 | #include <linux/jiffies.h> | |
54 | #include <linux/futex.h> | |
8141c7f3 | 55 | #include <linux/compat.h> |
207205a2 | 56 | #include <linux/kthread.h> |
7c3ab738 | 57 | #include <linux/task_io_accounting_ops.h> |
ab2af1f5 | 58 | #include <linux/rcupdate.h> |
1da177e4 LT |
59 | #include <linux/ptrace.h> |
60 | #include <linux/mount.h> | |
61 | #include <linux/audit.h> | |
78fb7466 | 62 | #include <linux/memcontrol.h> |
f201ae23 | 63 | #include <linux/ftrace.h> |
5e2bf014 | 64 | #include <linux/proc_fs.h> |
1da177e4 LT |
65 | #include <linux/profile.h> |
66 | #include <linux/rmap.h> | |
f8af4da3 | 67 | #include <linux/ksm.h> |
1da177e4 | 68 | #include <linux/acct.h> |
893e26e6 | 69 | #include <linux/userfaultfd_k.h> |
8f0ab514 | 70 | #include <linux/tsacct_kern.h> |
9f46080c | 71 | #include <linux/cn_proc.h> |
ba96a0c8 | 72 | #include <linux/freezer.h> |
ca74e92b | 73 | #include <linux/delayacct.h> |
ad4ecbcb | 74 | #include <linux/taskstats_kern.h> |
0a425405 | 75 | #include <linux/random.h> |
522ed776 | 76 | #include <linux/tty.h> |
fd0928df | 77 | #include <linux/blkdev.h> |
5ad4e53b | 78 | #include <linux/fs_struct.h> |
7c9f8861 | 79 | #include <linux/magic.h> |
cdd6c482 | 80 | #include <linux/perf_event.h> |
42c4ab41 | 81 | #include <linux/posix-timers.h> |
8e7cac79 | 82 | #include <linux/user-return-notifier.h> |
3d5992d2 | 83 | #include <linux/oom.h> |
ba76149f | 84 | #include <linux/khugepaged.h> |
d80e731e | 85 | #include <linux/signalfd.h> |
0326f5a9 | 86 | #include <linux/uprobes.h> |
a27bb332 | 87 | #include <linux/aio.h> |
52f5684c | 88 | #include <linux/compiler.h> |
16db3d3f | 89 | #include <linux/sysctl.h> |
5c9a8750 | 90 | #include <linux/kcov.h> |
d83a7cb3 | 91 | #include <linux/livepatch.h> |
48ac3c18 | 92 | #include <linux/thread_info.h> |
1da177e4 LT |
93 | |
94 | #include <asm/pgtable.h> | |
95 | #include <asm/pgalloc.h> | |
7c0f6ba6 | 96 | #include <linux/uaccess.h> |
1da177e4 LT |
97 | #include <asm/mmu_context.h> |
98 | #include <asm/cacheflush.h> | |
99 | #include <asm/tlbflush.h> | |
100 | ||
ad8d75ff SR |
101 | #include <trace/events/sched.h> |
102 | ||
43d2b113 KH |
103 | #define CREATE_TRACE_POINTS |
104 | #include <trace/events/task.h> | |
680f7875 SH |
105 | #ifdef CONFIG_USER_NS |
106 | extern int unprivileged_userns_clone; | |
107 | #else | |
108 | #define unprivileged_userns_clone 0 | |
109 | #endif | |
43d2b113 | 110 | |
ac1b398d HS |
111 | /* |
112 | * Minimum number of threads to boot the kernel | |
113 | */ | |
114 | #define MIN_THREADS 20 | |
115 | ||
116 | /* | |
117 | * Maximum number of threads | |
118 | */ | |
119 | #define MAX_THREADS FUTEX_TID_MASK | |
120 | ||
1da177e4 LT |
121 | /* |
122 | * Protected counters by write_lock_irq(&tasklist_lock) | |
123 | */ | |
124 | unsigned long total_forks; /* Handle normal Linux uptimes. */ | |
fb0a685c | 125 | int nr_threads; /* The idle threads do not count.. */ |
1da177e4 LT |
126 | |
127 | int max_threads; /* tunable limit on nr_threads */ | |
128 | ||
129 | DEFINE_PER_CPU(unsigned long, process_counts) = 0; | |
130 | ||
c59923a1 | 131 | __cacheline_aligned DEFINE_RWLOCK(tasklist_lock); /* outer */ |
db1466b3 PM |
132 | |
133 | #ifdef CONFIG_PROVE_RCU | |
134 | int lockdep_tasklist_lock_is_held(void) | |
135 | { | |
136 | return lockdep_is_held(&tasklist_lock); | |
137 | } | |
138 | EXPORT_SYMBOL_GPL(lockdep_tasklist_lock_is_held); | |
139 | #endif /* #ifdef CONFIG_PROVE_RCU */ | |
1da177e4 LT |
140 | |
141 | int nr_processes(void) | |
142 | { | |
143 | int cpu; | |
144 | int total = 0; | |
145 | ||
1d510750 | 146 | for_each_possible_cpu(cpu) |
1da177e4 LT |
147 | total += per_cpu(process_counts, cpu); |
148 | ||
149 | return total; | |
150 | } | |
151 | ||
f19b9f74 AM |
152 | void __weak arch_release_task_struct(struct task_struct *tsk) |
153 | { | |
154 | } | |
155 | ||
f5e10287 | 156 | #ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR |
e18b890b | 157 | static struct kmem_cache *task_struct_cachep; |
41101809 TG |
158 | |
159 | static inline struct task_struct *alloc_task_struct_node(int node) | |
160 | { | |
161 | return kmem_cache_alloc_node(task_struct_cachep, GFP_KERNEL, node); | |
162 | } | |
163 | ||
41101809 TG |
164 | static inline void free_task_struct(struct task_struct *tsk) |
165 | { | |
41101809 TG |
166 | kmem_cache_free(task_struct_cachep, tsk); |
167 | } | |
1da177e4 LT |
168 | #endif |
169 | ||
b235beea | 170 | #ifndef CONFIG_ARCH_THREAD_STACK_ALLOCATOR |
41101809 | 171 | |
0d15d74a TG |
172 | /* |
173 | * Allocate pages if THREAD_SIZE is >= PAGE_SIZE, otherwise use a | |
174 | * kmemcache based allocator. | |
175 | */ | |
ba14a194 | 176 | # if THREAD_SIZE >= PAGE_SIZE || defined(CONFIG_VMAP_STACK) |
ac496bf4 AL |
177 | |
178 | #ifdef CONFIG_VMAP_STACK | |
179 | /* | |
180 | * vmalloc() is a bit slow, and calling vfree() enough times will force a TLB | |
181 | * flush. Try to minimize the number of calls by caching stacks. | |
182 | */ | |
183 | #define NR_CACHED_STACKS 2 | |
184 | static DEFINE_PER_CPU(struct vm_struct *, cached_stacks[NR_CACHED_STACKS]); | |
19659c59 HR |
185 | |
186 | static int free_vm_stack_cache(unsigned int cpu) | |
187 | { | |
188 | struct vm_struct **cached_vm_stacks = per_cpu_ptr(cached_stacks, cpu); | |
189 | int i; | |
190 | ||
191 | for (i = 0; i < NR_CACHED_STACKS; i++) { | |
192 | struct vm_struct *vm_stack = cached_vm_stacks[i]; | |
193 | ||
194 | if (!vm_stack) | |
195 | continue; | |
196 | ||
197 | vfree(vm_stack->addr); | |
198 | cached_vm_stacks[i] = NULL; | |
199 | } | |
200 | ||
201 | return 0; | |
202 | } | |
ac496bf4 AL |
203 | #endif |
204 | ||
ba14a194 | 205 | static unsigned long *alloc_thread_stack_node(struct task_struct *tsk, int node) |
b69c49b7 | 206 | { |
ba14a194 | 207 | #ifdef CONFIG_VMAP_STACK |
ac496bf4 AL |
208 | void *stack; |
209 | int i; | |
210 | ||
ac496bf4 | 211 | for (i = 0; i < NR_CACHED_STACKS; i++) { |
112166f8 CL |
212 | struct vm_struct *s; |
213 | ||
214 | s = this_cpu_xchg(cached_stacks[i], NULL); | |
ac496bf4 AL |
215 | |
216 | if (!s) | |
217 | continue; | |
ac496bf4 | 218 | |
ca182551 KK |
219 | /* Clear stale pointers from reused stack. */ |
220 | memset(s->addr, 0, THREAD_SIZE); | |
c3169857 | 221 | |
ac496bf4 | 222 | tsk->stack_vm_area = s; |
0145a582 | 223 | tsk->stack = s->addr; |
ac496bf4 AL |
224 | return s->addr; |
225 | } | |
ac496bf4 | 226 | |
48ac3c18 | 227 | stack = __vmalloc_node_range(THREAD_SIZE, THREAD_ALIGN, |
ac496bf4 | 228 | VMALLOC_START, VMALLOC_END, |
19809c2d | 229 | THREADINFO_GFP, |
ac496bf4 AL |
230 | PAGE_KERNEL, |
231 | 0, node, __builtin_return_address(0)); | |
ba14a194 AL |
232 | |
233 | /* | |
234 | * We can't call find_vm_area() in interrupt context, and | |
235 | * free_thread_stack() can be called in interrupt context, | |
236 | * so cache the vm_struct. | |
237 | */ | |
03fb3ef2 | 238 | if (stack) { |
ba14a194 | 239 | tsk->stack_vm_area = find_vm_area(stack); |
03fb3ef2 RR |
240 | tsk->stack = stack; |
241 | } | |
ba14a194 AL |
242 | return stack; |
243 | #else | |
4949148a VD |
244 | struct page *page = alloc_pages_node(node, THREADINFO_GFP, |
245 | THREAD_SIZE_ORDER); | |
b6a84016 | 246 | |
1ffd8a9a AA |
247 | if (likely(page)) { |
248 | tsk->stack = page_address(page); | |
249 | return tsk->stack; | |
250 | } | |
251 | return NULL; | |
ba14a194 | 252 | #endif |
b69c49b7 FT |
253 | } |
254 | ||
ba14a194 | 255 | static inline void free_thread_stack(struct task_struct *tsk) |
b69c49b7 | 256 | { |
ac496bf4 AL |
257 | #ifdef CONFIG_VMAP_STACK |
258 | if (task_stack_vm_area(tsk)) { | |
ac496bf4 AL |
259 | int i; |
260 | ||
ac496bf4 | 261 | for (i = 0; i < NR_CACHED_STACKS; i++) { |
112166f8 CL |
262 | if (this_cpu_cmpxchg(cached_stacks[i], |
263 | NULL, tsk->stack_vm_area) != NULL) | |
ac496bf4 AL |
264 | continue; |
265 | ||
ac496bf4 AL |
266 | return; |
267 | } | |
ac496bf4 | 268 | |
0f110a9b | 269 | vfree_atomic(tsk->stack); |
ac496bf4 AL |
270 | return; |
271 | } | |
272 | #endif | |
273 | ||
274 | __free_pages(virt_to_page(tsk->stack), THREAD_SIZE_ORDER); | |
b69c49b7 | 275 | } |
0d15d74a | 276 | # else |
b235beea | 277 | static struct kmem_cache *thread_stack_cache; |
0d15d74a | 278 | |
9521d399 | 279 | static unsigned long *alloc_thread_stack_node(struct task_struct *tsk, |
0d15d74a TG |
280 | int node) |
281 | { | |
03fb3ef2 RR |
282 | unsigned long *stack; |
283 | stack = kmem_cache_alloc_node(thread_stack_cache, THREADINFO_GFP, node); | |
284 | tsk->stack = stack; | |
285 | return stack; | |
0d15d74a TG |
286 | } |
287 | ||
ba14a194 | 288 | static void free_thread_stack(struct task_struct *tsk) |
0d15d74a | 289 | { |
ba14a194 | 290 | kmem_cache_free(thread_stack_cache, tsk->stack); |
0d15d74a TG |
291 | } |
292 | ||
b235beea | 293 | void thread_stack_cache_init(void) |
0d15d74a | 294 | { |
b235beea | 295 | thread_stack_cache = kmem_cache_create("thread_stack", THREAD_SIZE, |
0d15d74a | 296 | THREAD_SIZE, 0, NULL); |
b235beea | 297 | BUG_ON(thread_stack_cache == NULL); |
0d15d74a TG |
298 | } |
299 | # endif | |
b69c49b7 FT |
300 | #endif |
301 | ||
1da177e4 | 302 | /* SLAB cache for signal_struct structures (tsk->signal) */ |
e18b890b | 303 | static struct kmem_cache *signal_cachep; |
1da177e4 LT |
304 | |
305 | /* SLAB cache for sighand_struct structures (tsk->sighand) */ | |
e18b890b | 306 | struct kmem_cache *sighand_cachep; |
1da177e4 LT |
307 | |
308 | /* SLAB cache for files_struct structures (tsk->files) */ | |
e18b890b | 309 | struct kmem_cache *files_cachep; |
1da177e4 LT |
310 | |
311 | /* SLAB cache for fs_struct structures (tsk->fs) */ | |
e18b890b | 312 | struct kmem_cache *fs_cachep; |
1da177e4 LT |
313 | |
314 | /* SLAB cache for vm_area_struct structures */ | |
e88c5cf9 | 315 | static struct kmem_cache *vm_area_cachep; |
1da177e4 LT |
316 | |
317 | /* SLAB cache for mm_struct structures (tsk->mm) */ | |
e18b890b | 318 | static struct kmem_cache *mm_cachep; |
1da177e4 | 319 | |
0d2c7e02 | 320 | struct vm_area_struct *vm_area_alloc(struct mm_struct *mm) |
e88c5cf9 | 321 | { |
0d2c7e02 LT |
322 | struct vm_area_struct *vma = kmem_cache_zalloc(vm_area_cachep, GFP_KERNEL); |
323 | ||
324 | if (vma) { | |
325 | vma->vm_mm = mm; | |
326 | INIT_LIST_HEAD(&vma->anon_vma_chain); | |
327 | } | |
328 | return vma; | |
e88c5cf9 LT |
329 | } |
330 | ||
331 | struct vm_area_struct *vm_area_dup(struct vm_area_struct *orig) | |
332 | { | |
ee8cad6e LT |
333 | struct vm_area_struct *new = kmem_cache_alloc(vm_area_cachep, GFP_KERNEL); |
334 | ||
335 | if (new) { | |
336 | *new = *orig; | |
337 | INIT_LIST_HEAD(&new->anon_vma_chain); | |
338 | } | |
339 | return new; | |
e88c5cf9 LT |
340 | } |
341 | ||
342 | void vm_area_free(struct vm_area_struct *vma) | |
343 | { | |
344 | kmem_cache_free(vm_area_cachep, vma); | |
345 | } | |
346 | ||
ba14a194 | 347 | static void account_kernel_stack(struct task_struct *tsk, int account) |
c6a7f572 | 348 | { |
ba14a194 AL |
349 | void *stack = task_stack_page(tsk); |
350 | struct vm_struct *vm = task_stack_vm_area(tsk); | |
351 | ||
352 | BUILD_BUG_ON(IS_ENABLED(CONFIG_VMAP_STACK) && PAGE_SIZE % 1024 != 0); | |
353 | ||
354 | if (vm) { | |
355 | int i; | |
356 | ||
357 | BUG_ON(vm->nr_pages != THREAD_SIZE / PAGE_SIZE); | |
358 | ||
359 | for (i = 0; i < THREAD_SIZE / PAGE_SIZE; i++) { | |
360 | mod_zone_page_state(page_zone(vm->pages[i]), | |
361 | NR_KERNEL_STACK_KB, | |
362 | PAGE_SIZE / 1024 * account); | |
363 | } | |
364 | ||
365 | /* All stack pages belong to the same memcg. */ | |
ed52be7b JW |
366 | mod_memcg_page_state(vm->pages[0], MEMCG_KERNEL_STACK_KB, |
367 | account * (THREAD_SIZE / 1024)); | |
ba14a194 AL |
368 | } else { |
369 | /* | |
370 | * All stack pages are in the same zone and belong to the | |
371 | * same memcg. | |
372 | */ | |
373 | struct page *first_page = virt_to_page(stack); | |
374 | ||
375 | mod_zone_page_state(page_zone(first_page), NR_KERNEL_STACK_KB, | |
376 | THREAD_SIZE / 1024 * account); | |
377 | ||
ed52be7b JW |
378 | mod_memcg_page_state(first_page, MEMCG_KERNEL_STACK_KB, |
379 | account * (THREAD_SIZE / 1024)); | |
ba14a194 | 380 | } |
c6a7f572 KM |
381 | } |
382 | ||
68f24b08 | 383 | static void release_task_stack(struct task_struct *tsk) |
1da177e4 | 384 | { |
405c0759 AL |
385 | if (WARN_ON(tsk->state != TASK_DEAD)) |
386 | return; /* Better to leak the stack than to free prematurely */ | |
387 | ||
ba14a194 | 388 | account_kernel_stack(tsk, -1); |
ba14a194 | 389 | free_thread_stack(tsk); |
68f24b08 AL |
390 | tsk->stack = NULL; |
391 | #ifdef CONFIG_VMAP_STACK | |
392 | tsk->stack_vm_area = NULL; | |
393 | #endif | |
394 | } | |
395 | ||
396 | #ifdef CONFIG_THREAD_INFO_IN_TASK | |
397 | void put_task_stack(struct task_struct *tsk) | |
398 | { | |
399 | if (atomic_dec_and_test(&tsk->stack_refcount)) | |
400 | release_task_stack(tsk); | |
401 | } | |
402 | #endif | |
403 | ||
404 | void free_task(struct task_struct *tsk) | |
405 | { | |
406 | #ifndef CONFIG_THREAD_INFO_IN_TASK | |
407 | /* | |
408 | * The task is finally done with both the stack and thread_info, | |
409 | * so free both. | |
410 | */ | |
411 | release_task_stack(tsk); | |
412 | #else | |
413 | /* | |
414 | * If the task had a separate stack allocation, it should be gone | |
415 | * by now. | |
416 | */ | |
417 | WARN_ON_ONCE(atomic_read(&tsk->stack_refcount) != 0); | |
418 | #endif | |
23f78d4a | 419 | rt_mutex_debug_task_free(tsk); |
fb52607a | 420 | ftrace_graph_exit_task(tsk); |
e2cfabdf | 421 | put_seccomp_filter(tsk); |
f19b9f74 | 422 | arch_release_task_struct(tsk); |
1da5c46f ON |
423 | if (tsk->flags & PF_KTHREAD) |
424 | free_kthread_struct(tsk); | |
1da177e4 LT |
425 | free_task_struct(tsk); |
426 | } | |
427 | EXPORT_SYMBOL(free_task); | |
428 | ||
ea6d290c ON |
429 | static inline void free_signal_struct(struct signal_struct *sig) |
430 | { | |
97101eb4 | 431 | taskstats_tgid_free(sig); |
1c5354de | 432 | sched_autogroup_exit(sig); |
7283094e MH |
433 | /* |
434 | * __mmdrop is not safe to call from softirq context on x86 due to | |
435 | * pgd_dtor so postpone it to the async context | |
436 | */ | |
26db62f1 | 437 | if (sig->oom_mm) |
7283094e | 438 | mmdrop_async(sig->oom_mm); |
ea6d290c ON |
439 | kmem_cache_free(signal_cachep, sig); |
440 | } | |
441 | ||
442 | static inline void put_signal_struct(struct signal_struct *sig) | |
443 | { | |
1c5354de | 444 | if (atomic_dec_and_test(&sig->sigcnt)) |
ea6d290c ON |
445 | free_signal_struct(sig); |
446 | } | |
447 | ||
158d9ebd | 448 | void __put_task_struct(struct task_struct *tsk) |
1da177e4 | 449 | { |
270f722d | 450 | WARN_ON(!tsk->exit_state); |
1da177e4 LT |
451 | WARN_ON(atomic_read(&tsk->usage)); |
452 | WARN_ON(tsk == current); | |
453 | ||
2e91fa7f | 454 | cgroup_free(tsk); |
30197fa5 | 455 | task_numa_free(tsk, true); |
1a2a4d06 | 456 | security_task_free(tsk); |
e0e81739 | 457 | exit_creds(tsk); |
35df17c5 | 458 | delayacct_tsk_free(tsk); |
ea6d290c | 459 | put_signal_struct(tsk->signal); |
1da177e4 LT |
460 | |
461 | if (!profile_handoff_task(tsk)) | |
462 | free_task(tsk); | |
463 | } | |
77c100c8 | 464 | EXPORT_SYMBOL_GPL(__put_task_struct); |
1da177e4 | 465 | |
6c0a9fa6 | 466 | void __init __weak arch_task_cache_init(void) { } |
61c4628b | 467 | |
ff691f6e HS |
468 | /* |
469 | * set_max_threads | |
470 | */ | |
16db3d3f | 471 | static void set_max_threads(unsigned int max_threads_suggested) |
ff691f6e | 472 | { |
ac1b398d | 473 | u64 threads; |
ff691f6e HS |
474 | |
475 | /* | |
ac1b398d HS |
476 | * The number of threads shall be limited such that the thread |
477 | * structures may only consume a small part of the available memory. | |
ff691f6e | 478 | */ |
ac1b398d HS |
479 | if (fls64(totalram_pages) + fls64(PAGE_SIZE) > 64) |
480 | threads = MAX_THREADS; | |
481 | else | |
482 | threads = div64_u64((u64) totalram_pages * (u64) PAGE_SIZE, | |
483 | (u64) THREAD_SIZE * 8UL); | |
484 | ||
16db3d3f HS |
485 | if (threads > max_threads_suggested) |
486 | threads = max_threads_suggested; | |
487 | ||
ac1b398d | 488 | max_threads = clamp_t(u64, threads, MIN_THREADS, MAX_THREADS); |
ff691f6e HS |
489 | } |
490 | ||
5aaeb5c0 IM |
491 | #ifdef CONFIG_ARCH_WANTS_DYNAMIC_TASK_STRUCT |
492 | /* Initialized by the architecture: */ | |
493 | int arch_task_struct_size __read_mostly; | |
494 | #endif | |
0c8c0f03 | 495 | |
ff691f6e | 496 | void __init fork_init(void) |
1da177e4 | 497 | { |
25f9c081 | 498 | int i; |
f5e10287 | 499 | #ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR |
1da177e4 | 500 | #ifndef ARCH_MIN_TASKALIGN |
e274795e | 501 | #define ARCH_MIN_TASKALIGN 0 |
1da177e4 | 502 | #endif |
95cb64c1 | 503 | int align = max_t(int, L1_CACHE_BYTES, ARCH_MIN_TASKALIGN); |
e274795e | 504 | |
1da177e4 | 505 | /* create a slab on which task_structs can be allocated */ |
5d097056 | 506 | task_struct_cachep = kmem_cache_create("task_struct", |
e274795e | 507 | arch_task_struct_size, align, |
75f296d9 | 508 | SLAB_PANIC|SLAB_ACCOUNT, NULL); |
1da177e4 LT |
509 | #endif |
510 | ||
61c4628b SS |
511 | /* do the arch specific task caches init */ |
512 | arch_task_cache_init(); | |
513 | ||
16db3d3f | 514 | set_max_threads(MAX_THREADS); |
1da177e4 LT |
515 | |
516 | init_task.signal->rlim[RLIMIT_NPROC].rlim_cur = max_threads/2; | |
517 | init_task.signal->rlim[RLIMIT_NPROC].rlim_max = max_threads/2; | |
518 | init_task.signal->rlim[RLIMIT_SIGPENDING] = | |
519 | init_task.signal->rlim[RLIMIT_NPROC]; | |
b376c3e1 | 520 | |
25f9c081 EB |
521 | for (i = 0; i < UCOUNT_COUNTS; i++) { |
522 | init_user_ns.ucount_max[i] = max_threads/2; | |
523 | } | |
19659c59 HR |
524 | |
525 | #ifdef CONFIG_VMAP_STACK | |
526 | cpuhp_setup_state(CPUHP_BP_PREPARE_DYN, "fork:vm_stack_cache", | |
527 | NULL, free_vm_stack_cache); | |
528 | #endif | |
b09be676 BP |
529 | |
530 | lockdep_init_task(&init_task); | |
1da177e4 LT |
531 | } |
532 | ||
52f5684c | 533 | int __weak arch_dup_task_struct(struct task_struct *dst, |
61c4628b SS |
534 | struct task_struct *src) |
535 | { | |
536 | *dst = *src; | |
537 | return 0; | |
538 | } | |
539 | ||
d4311ff1 AT |
540 | void set_task_stack_end_magic(struct task_struct *tsk) |
541 | { | |
542 | unsigned long *stackend; | |
543 | ||
544 | stackend = end_of_stack(tsk); | |
545 | *stackend = STACK_END_MAGIC; /* for overflow detection */ | |
546 | } | |
547 | ||
725fc629 | 548 | static struct task_struct *dup_task_struct(struct task_struct *orig, int node) |
1da177e4 LT |
549 | { |
550 | struct task_struct *tsk; | |
b235beea | 551 | unsigned long *stack; |
ba14a194 | 552 | struct vm_struct *stack_vm_area; |
3e26c149 | 553 | int err; |
1da177e4 | 554 | |
725fc629 AK |
555 | if (node == NUMA_NO_NODE) |
556 | node = tsk_fork_get_node(orig); | |
504f52b5 | 557 | tsk = alloc_task_struct_node(node); |
1da177e4 LT |
558 | if (!tsk) |
559 | return NULL; | |
560 | ||
b235beea LT |
561 | stack = alloc_thread_stack_node(tsk, node); |
562 | if (!stack) | |
f19b9f74 | 563 | goto free_tsk; |
1da177e4 | 564 | |
ba14a194 AL |
565 | stack_vm_area = task_stack_vm_area(tsk); |
566 | ||
fb0a685c | 567 | err = arch_dup_task_struct(tsk, orig); |
ba14a194 AL |
568 | |
569 | /* | |
570 | * arch_dup_task_struct() clobbers the stack-related fields. Make | |
571 | * sure they're properly initialized before using any stack-related | |
572 | * functions again. | |
573 | */ | |
574 | tsk->stack = stack; | |
575 | #ifdef CONFIG_VMAP_STACK | |
576 | tsk->stack_vm_area = stack_vm_area; | |
577 | #endif | |
68f24b08 AL |
578 | #ifdef CONFIG_THREAD_INFO_IN_TASK |
579 | atomic_set(&tsk->stack_refcount, 1); | |
580 | #endif | |
ba14a194 | 581 | |
164c33c6 | 582 | if (err) |
b235beea | 583 | goto free_stack; |
164c33c6 | 584 | |
dbd95212 KC |
585 | #ifdef CONFIG_SECCOMP |
586 | /* | |
587 | * We must handle setting up seccomp filters once we're under | |
588 | * the sighand lock in case orig has changed between now and | |
589 | * then. Until then, filter must be NULL to avoid messing up | |
590 | * the usage counts on the error path calling free_task. | |
591 | */ | |
592 | tsk->seccomp.filter = NULL; | |
593 | #endif | |
87bec58a AM |
594 | |
595 | setup_thread_stack(tsk, orig); | |
8e7cac79 | 596 | clear_user_return_notifier(tsk); |
f26f9aff | 597 | clear_tsk_need_resched(tsk); |
d4311ff1 | 598 | set_task_stack_end_magic(tsk); |
1da177e4 | 599 | |
0a425405 | 600 | #ifdef CONFIG_CC_STACKPROTECTOR |
7cd815bc | 601 | tsk->stack_canary = get_random_canary(); |
0a425405 AV |
602 | #endif |
603 | ||
fb0a685c DRO |
604 | /* |
605 | * One for us, one for whoever does the "release_task()" (usually | |
606 | * parent) | |
607 | */ | |
608 | atomic_set(&tsk->usage, 2); | |
6c5c9341 | 609 | #ifdef CONFIG_BLK_DEV_IO_TRACE |
2056a782 | 610 | tsk->btrace_seq = 0; |
6c5c9341 | 611 | #endif |
a0aa7f68 | 612 | tsk->splice_pipe = NULL; |
5640f768 | 613 | tsk->task_frag.page = NULL; |
093e5840 | 614 | tsk->wake_q.next = NULL; |
c6a7f572 | 615 | |
ba14a194 | 616 | account_kernel_stack(tsk, 1); |
c6a7f572 | 617 | |
5c9a8750 DV |
618 | kcov_task_init(tsk); |
619 | ||
e41d5818 DV |
620 | #ifdef CONFIG_FAULT_INJECTION |
621 | tsk->fail_nth = 0; | |
622 | #endif | |
623 | ||
1da177e4 | 624 | return tsk; |
61c4628b | 625 | |
b235beea | 626 | free_stack: |
ba14a194 | 627 | free_thread_stack(tsk); |
f19b9f74 | 628 | free_tsk: |
61c4628b SS |
629 | free_task_struct(tsk); |
630 | return NULL; | |
1da177e4 LT |
631 | } |
632 | ||
633 | #ifdef CONFIG_MMU | |
0766f788 ER |
634 | static __latent_entropy int dup_mmap(struct mm_struct *mm, |
635 | struct mm_struct *oldmm) | |
1da177e4 | 636 | { |
297c5eee | 637 | struct vm_area_struct *mpnt, *tmp, *prev, **pprev; |
1da177e4 LT |
638 | struct rb_node **rb_link, *rb_parent; |
639 | int retval; | |
640 | unsigned long charge; | |
893e26e6 | 641 | LIST_HEAD(uf); |
1da177e4 | 642 | |
32cdba1e | 643 | uprobe_start_dup_mmap(); |
7c051267 MH |
644 | if (down_write_killable(&oldmm->mmap_sem)) { |
645 | retval = -EINTR; | |
646 | goto fail_uprobe_end; | |
647 | } | |
ec8c0446 | 648 | flush_cache_dup_mm(oldmm); |
f8ac4ec9 | 649 | uprobe_dup_mmap(oldmm, mm); |
ad339451 IM |
650 | /* |
651 | * Not linked in yet - no deadlock potential: | |
652 | */ | |
653 | down_write_nested(&mm->mmap_sem, SINGLE_DEPTH_NESTING); | |
7ee78232 | 654 | |
90f31d0e KK |
655 | /* No ordering required: file already has been exposed. */ |
656 | RCU_INIT_POINTER(mm->exe_file, get_mm_exe_file(oldmm)); | |
657 | ||
4f7d4614 | 658 | mm->total_vm = oldmm->total_vm; |
84638335 | 659 | mm->data_vm = oldmm->data_vm; |
4f7d4614 VD |
660 | mm->exec_vm = oldmm->exec_vm; |
661 | mm->stack_vm = oldmm->stack_vm; | |
662 | ||
1da177e4 LT |
663 | rb_link = &mm->mm_rb.rb_node; |
664 | rb_parent = NULL; | |
665 | pprev = &mm->mmap; | |
f8af4da3 | 666 | retval = ksm_fork(mm, oldmm); |
ba76149f AA |
667 | if (retval) |
668 | goto out; | |
669 | retval = khugepaged_fork(mm, oldmm); | |
f8af4da3 HD |
670 | if (retval) |
671 | goto out; | |
1da177e4 | 672 | |
297c5eee | 673 | prev = NULL; |
fd3e42fc | 674 | for (mpnt = oldmm->mmap; mpnt; mpnt = mpnt->vm_next) { |
1da177e4 LT |
675 | struct file *file; |
676 | ||
677 | if (mpnt->vm_flags & VM_DONTCOPY) { | |
84638335 | 678 | vm_stat_account(mm, mpnt->vm_flags, -vma_pages(mpnt)); |
1da177e4 LT |
679 | continue; |
680 | } | |
681 | charge = 0; | |
32f0a536 TH |
682 | /* |
683 | * Don't duplicate many vmas if we've been oom-killed (for | |
684 | * example) | |
685 | */ | |
686 | if (fatal_signal_pending(current)) { | |
687 | retval = -EINTR; | |
688 | goto out; | |
689 | } | |
1da177e4 | 690 | if (mpnt->vm_flags & VM_ACCOUNT) { |
b2412b7f HS |
691 | unsigned long len = vma_pages(mpnt); |
692 | ||
191c5424 | 693 | if (security_vm_enough_memory_mm(oldmm, len)) /* sic */ |
1da177e4 LT |
694 | goto fail_nomem; |
695 | charge = len; | |
696 | } | |
e88c5cf9 | 697 | tmp = vm_area_dup(mpnt); |
1da177e4 LT |
698 | if (!tmp) |
699 | goto fail_nomem; | |
ef0855d3 ON |
700 | retval = vma_dup_policy(mpnt, tmp); |
701 | if (retval) | |
1da177e4 | 702 | goto fail_nomem_policy; |
a247c3a9 | 703 | tmp->vm_mm = mm; |
893e26e6 PE |
704 | retval = dup_userfaultfd(tmp, &uf); |
705 | if (retval) | |
706 | goto fail_nomem_anon_vma_fork; | |
d2cd9ede RR |
707 | if (tmp->vm_flags & VM_WIPEONFORK) { |
708 | /* VM_WIPEONFORK gets a clean slate in the child. */ | |
709 | tmp->anon_vma = NULL; | |
710 | if (anon_vma_prepare(tmp)) | |
711 | goto fail_nomem_anon_vma_fork; | |
712 | } else if (anon_vma_fork(tmp, mpnt)) | |
5beb4930 | 713 | goto fail_nomem_anon_vma_fork; |
893e26e6 | 714 | tmp->vm_flags &= ~(VM_LOCKED | VM_LOCKONFAULT); |
297c5eee | 715 | tmp->vm_next = tmp->vm_prev = NULL; |
1da177e4 LT |
716 | file = tmp->vm_file; |
717 | if (file) { | |
496ad9aa | 718 | struct inode *inode = file_inode(file); |
b88ed205 HD |
719 | struct address_space *mapping = file->f_mapping; |
720 | ||
c088e31d | 721 | vma_get_file(tmp); |
1da177e4 LT |
722 | if (tmp->vm_flags & VM_DENYWRITE) |
723 | atomic_dec(&inode->i_writecount); | |
83cde9e8 | 724 | i_mmap_lock_write(mapping); |
b88ed205 | 725 | if (tmp->vm_flags & VM_SHARED) |
4bb5f5d9 | 726 | atomic_inc(&mapping->i_mmap_writable); |
b88ed205 HD |
727 | flush_dcache_mmap_lock(mapping); |
728 | /* insert tmp into the share list, just after mpnt */ | |
27ba0644 KS |
729 | vma_interval_tree_insert_after(tmp, mpnt, |
730 | &mapping->i_mmap); | |
b88ed205 | 731 | flush_dcache_mmap_unlock(mapping); |
83cde9e8 | 732 | i_mmap_unlock_write(mapping); |
1da177e4 LT |
733 | } |
734 | ||
a1e78772 MG |
735 | /* |
736 | * Clear hugetlb-related page reserves for children. This only | |
737 | * affects MAP_PRIVATE mappings. Faults generated by the child | |
738 | * are not guaranteed to succeed, even if read-only | |
739 | */ | |
740 | if (is_vm_hugetlb_page(tmp)) | |
741 | reset_vma_resv_huge_pages(tmp); | |
742 | ||
1da177e4 | 743 | /* |
7ee78232 | 744 | * Link in the new vma and copy the page table entries. |
1da177e4 | 745 | */ |
1da177e4 LT |
746 | *pprev = tmp; |
747 | pprev = &tmp->vm_next; | |
297c5eee LT |
748 | tmp->vm_prev = prev; |
749 | prev = tmp; | |
1da177e4 LT |
750 | |
751 | __vma_link_rb(mm, tmp, rb_link, rb_parent); | |
752 | rb_link = &tmp->vm_rb.rb_right; | |
753 | rb_parent = &tmp->vm_rb; | |
754 | ||
755 | mm->map_count++; | |
d2cd9ede RR |
756 | if (!(tmp->vm_flags & VM_WIPEONFORK)) |
757 | retval = copy_page_range(mm, oldmm, mpnt); | |
1da177e4 LT |
758 | |
759 | if (tmp->vm_ops && tmp->vm_ops->open) | |
760 | tmp->vm_ops->open(tmp); | |
761 | ||
762 | if (retval) | |
763 | goto out; | |
764 | } | |
d6dd61c8 | 765 | /* a new mm has just been created */ |
c10e83f5 | 766 | retval = arch_dup_mmap(oldmm, mm); |
1da177e4 | 767 | out: |
7ee78232 | 768 | up_write(&mm->mmap_sem); |
fd3e42fc | 769 | flush_tlb_mm(oldmm); |
1da177e4 | 770 | up_write(&oldmm->mmap_sem); |
893e26e6 | 771 | dup_userfaultfd_complete(&uf); |
7c051267 | 772 | fail_uprobe_end: |
32cdba1e | 773 | uprobe_end_dup_mmap(); |
1da177e4 | 774 | return retval; |
5beb4930 | 775 | fail_nomem_anon_vma_fork: |
ef0855d3 | 776 | mpol_put(vma_policy(tmp)); |
1da177e4 | 777 | fail_nomem_policy: |
e88c5cf9 | 778 | vm_area_free(tmp); |
1da177e4 LT |
779 | fail_nomem: |
780 | retval = -ENOMEM; | |
781 | vm_unacct_memory(charge); | |
782 | goto out; | |
783 | } | |
784 | ||
fb0a685c | 785 | static inline int mm_alloc_pgd(struct mm_struct *mm) |
1da177e4 LT |
786 | { |
787 | mm->pgd = pgd_alloc(mm); | |
788 | if (unlikely(!mm->pgd)) | |
789 | return -ENOMEM; | |
790 | return 0; | |
791 | } | |
792 | ||
fb0a685c | 793 | static inline void mm_free_pgd(struct mm_struct *mm) |
1da177e4 | 794 | { |
5e541973 | 795 | pgd_free(mm, mm->pgd); |
1da177e4 LT |
796 | } |
797 | #else | |
90f31d0e KK |
798 | static int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm) |
799 | { | |
800 | down_write(&oldmm->mmap_sem); | |
801 | RCU_INIT_POINTER(mm->exe_file, get_mm_exe_file(oldmm)); | |
802 | up_write(&oldmm->mmap_sem); | |
803 | return 0; | |
804 | } | |
1da177e4 LT |
805 | #define mm_alloc_pgd(mm) (0) |
806 | #define mm_free_pgd(mm) | |
807 | #endif /* CONFIG_MMU */ | |
808 | ||
23ff4440 | 809 | __cacheline_aligned_in_smp DEFINE_SPINLOCK(mmlist_lock); |
1da177e4 | 810 | |
e94b1766 | 811 | #define allocate_mm() (kmem_cache_alloc(mm_cachep, GFP_KERNEL)) |
1da177e4 LT |
812 | #define free_mm(mm) (kmem_cache_free(mm_cachep, (mm))) |
813 | ||
4cb0e11b HK |
814 | static unsigned long default_dump_filter = MMF_DUMP_FILTER_DEFAULT; |
815 | ||
816 | static int __init coredump_filter_setup(char *s) | |
817 | { | |
818 | default_dump_filter = | |
819 | (simple_strtoul(s, NULL, 0) << MMF_DUMP_FILTER_SHIFT) & | |
820 | MMF_DUMP_FILTER_MASK; | |
821 | return 1; | |
822 | } | |
823 | ||
824 | __setup("coredump_filter=", coredump_filter_setup); | |
825 | ||
1da177e4 LT |
826 | #include <linux/init_task.h> |
827 | ||
858f0993 AD |
828 | static void mm_init_aio(struct mm_struct *mm) |
829 | { | |
830 | #ifdef CONFIG_AIO | |
831 | spin_lock_init(&mm->ioctx_lock); | |
db446a08 | 832 | mm->ioctx_table = NULL; |
858f0993 AD |
833 | #endif |
834 | } | |
835 | ||
343ef896 AA |
836 | static __always_inline void mm_clear_owner(struct mm_struct *mm, |
837 | struct task_struct *p) | |
838 | { | |
839 | #ifdef CONFIG_MEMCG | |
840 | if (mm->owner == p) | |
841 | WRITE_ONCE(mm->owner, NULL); | |
842 | #endif | |
843 | } | |
844 | ||
33144e84 VD |
845 | static void mm_init_owner(struct mm_struct *mm, struct task_struct *p) |
846 | { | |
847 | #ifdef CONFIG_MEMCG | |
848 | mm->owner = p; | |
849 | #endif | |
850 | } | |
851 | ||
355627f5 EB |
852 | static void mm_init_uprobes_state(struct mm_struct *mm) |
853 | { | |
854 | #ifdef CONFIG_UPROBES | |
855 | mm->uprobes_state.xol_area = NULL; | |
856 | #endif | |
857 | } | |
858 | ||
bfedb589 EB |
859 | static struct mm_struct *mm_init(struct mm_struct *mm, struct task_struct *p, |
860 | struct user_namespace *user_ns) | |
1da177e4 | 861 | { |
41f727fd VD |
862 | mm->mmap = NULL; |
863 | mm->mm_rb = RB_ROOT; | |
864 | mm->vmacache_seqnum = 0; | |
1da177e4 LT |
865 | atomic_set(&mm->mm_users, 1); |
866 | atomic_set(&mm->mm_count, 1); | |
867 | init_rwsem(&mm->mmap_sem); | |
868 | INIT_LIST_HEAD(&mm->mmlist); | |
999d9fc1 | 869 | mm->core_state = NULL; |
af5b0f6a | 870 | mm_pgtables_bytes_init(mm); |
41f727fd VD |
871 | mm->map_count = 0; |
872 | mm->locked_vm = 0; | |
ce65cefa | 873 | mm->pinned_vm = 0; |
d559db08 | 874 | memset(&mm->rss_stat, 0, sizeof(mm->rss_stat)); |
1da177e4 | 875 | spin_lock_init(&mm->page_table_lock); |
41f727fd | 876 | mm_init_cpumask(mm); |
858f0993 | 877 | mm_init_aio(mm); |
cf475ad2 | 878 | mm_init_owner(mm, p); |
2b7e8665 | 879 | RCU_INIT_POINTER(mm->exe_file, NULL); |
41f727fd | 880 | mmu_notifier_mm_init(mm); |
133ff0ea | 881 | hmm_mm_init(mm); |
16af97dc | 882 | init_tlb_flush_pending(mm); |
41f727fd VD |
883 | #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS |
884 | mm->pmd_huge_pte = NULL; | |
885 | #endif | |
355627f5 | 886 | mm_init_uprobes_state(mm); |
1da177e4 | 887 | |
a0715cc2 AT |
888 | if (current->mm) { |
889 | mm->flags = current->mm->flags & MMF_INIT_MASK; | |
890 | mm->def_flags = current->mm->def_flags & VM_INIT_DEF_MASK; | |
891 | } else { | |
892 | mm->flags = default_dump_filter; | |
1da177e4 | 893 | mm->def_flags = 0; |
a0715cc2 AT |
894 | } |
895 | ||
41f727fd VD |
896 | if (mm_alloc_pgd(mm)) |
897 | goto fail_nopgd; | |
898 | ||
899 | if (init_new_context(p, mm)) | |
900 | goto fail_nocontext; | |
78fb7466 | 901 | |
bfedb589 | 902 | mm->user_ns = get_user_ns(user_ns); |
41f727fd VD |
903 | return mm; |
904 | ||
905 | fail_nocontext: | |
906 | mm_free_pgd(mm); | |
907 | fail_nopgd: | |
1da177e4 LT |
908 | free_mm(mm); |
909 | return NULL; | |
910 | } | |
911 | ||
c3f0327f KK |
912 | static void check_mm(struct mm_struct *mm) |
913 | { | |
914 | int i; | |
915 | ||
916 | for (i = 0; i < NR_MM_COUNTERS; i++) { | |
917 | long x = atomic_long_read(&mm->rss_stat.count[i]); | |
918 | ||
919 | if (unlikely(x)) | |
920 | printk(KERN_ALERT "BUG: Bad rss-counter state " | |
921 | "mm:%p idx:%d val:%ld\n", mm, i, x); | |
922 | } | |
b30fe6c7 | 923 | |
af5b0f6a KS |
924 | if (mm_pgtables_bytes(mm)) |
925 | pr_alert("BUG: non-zero pgtables_bytes on freeing mm: %ld\n", | |
926 | mm_pgtables_bytes(mm)); | |
b30fe6c7 | 927 | |
e009bb30 | 928 | #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS |
96dad67f | 929 | VM_BUG_ON_MM(mm->pmd_huge_pte, mm); |
c3f0327f KK |
930 | #endif |
931 | } | |
932 | ||
1da177e4 LT |
933 | /* |
934 | * Allocate and initialize an mm_struct. | |
935 | */ | |
fb0a685c | 936 | struct mm_struct *mm_alloc(void) |
1da177e4 | 937 | { |
fb0a685c | 938 | struct mm_struct *mm; |
1da177e4 LT |
939 | |
940 | mm = allocate_mm(); | |
de03c72c KM |
941 | if (!mm) |
942 | return NULL; | |
943 | ||
944 | memset(mm, 0, sizeof(*mm)); | |
bfedb589 | 945 | return mm_init(mm, current, current_user_ns()); |
1da177e4 LT |
946 | } |
947 | ||
948 | /* | |
949 | * Called when the last reference to the mm | |
950 | * is dropped: either by a lazy thread or by | |
951 | * mmput. Free the page directory and the mm. | |
952 | */ | |
7ad5b3a5 | 953 | void __mmdrop(struct mm_struct *mm) |
1da177e4 LT |
954 | { |
955 | BUG_ON(mm == &init_mm); | |
956 | mm_free_pgd(mm); | |
957 | destroy_context(mm); | |
133ff0ea | 958 | hmm_mm_destroy(mm); |
cddb8a5c | 959 | mmu_notifier_mm_destroy(mm); |
c3f0327f | 960 | check_mm(mm); |
bfedb589 | 961 | put_user_ns(mm->user_ns); |
1da177e4 LT |
962 | free_mm(mm); |
963 | } | |
6d4e4c4f | 964 | EXPORT_SYMBOL_GPL(__mmdrop); |
1da177e4 | 965 | |
ec8d7c14 MH |
966 | static inline void __mmput(struct mm_struct *mm) |
967 | { | |
968 | VM_BUG_ON(atomic_read(&mm->mm_users)); | |
969 | ||
970 | uprobe_clear_state(mm); | |
971 | exit_aio(mm); | |
972 | ksm_exit(mm); | |
973 | khugepaged_exit(mm); /* must run before exit_mmap */ | |
974 | exit_mmap(mm); | |
6fcb52a5 | 975 | mm_put_huge_zero_page(mm); |
ec8d7c14 MH |
976 | set_mm_exe_file(mm, NULL); |
977 | if (!list_empty(&mm->mmlist)) { | |
978 | spin_lock(&mmlist_lock); | |
979 | list_del(&mm->mmlist); | |
980 | spin_unlock(&mmlist_lock); | |
981 | } | |
982 | if (mm->binfmt) | |
983 | module_put(mm->binfmt->module); | |
984 | mmdrop(mm); | |
985 | } | |
986 | ||
1da177e4 LT |
987 | /* |
988 | * Decrement the use count and release all resources for an mm. | |
989 | */ | |
990 | void mmput(struct mm_struct *mm) | |
991 | { | |
0ae26f1b AM |
992 | might_sleep(); |
993 | ||
ec8d7c14 MH |
994 | if (atomic_dec_and_test(&mm->mm_users)) |
995 | __mmput(mm); | |
996 | } | |
997 | EXPORT_SYMBOL_GPL(mmput); | |
998 | ||
a1b2289c SY |
999 | #ifdef CONFIG_MMU |
1000 | static void mmput_async_fn(struct work_struct *work) | |
1001 | { | |
1002 | struct mm_struct *mm = container_of(work, struct mm_struct, | |
1003 | async_put_work); | |
1004 | ||
1005 | __mmput(mm); | |
1006 | } | |
1007 | ||
1008 | void mmput_async(struct mm_struct *mm) | |
1009 | { | |
1010 | if (atomic_dec_and_test(&mm->mm_users)) { | |
1011 | INIT_WORK(&mm->async_put_work, mmput_async_fn); | |
1012 | schedule_work(&mm->async_put_work); | |
1013 | } | |
1014 | } | |
1015 | #endif | |
1016 | ||
90f31d0e KK |
1017 | /** |
1018 | * set_mm_exe_file - change a reference to the mm's executable file | |
1019 | * | |
1020 | * This changes mm's executable file (shown as symlink /proc/[pid]/exe). | |
1021 | * | |
6e399cd1 DB |
1022 | * Main users are mmput() and sys_execve(). Callers prevent concurrent |
1023 | * invocations: in mmput() nobody alive left, in execve task is single | |
1024 | * threaded. sys_prctl(PR_SET_MM_MAP/EXE_FILE) also needs to set the | |
1025 | * mm->exe_file, but does so without using set_mm_exe_file() in order | |
1026 | * to do avoid the need for any locks. | |
90f31d0e | 1027 | */ |
38646013 JS |
1028 | void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file) |
1029 | { | |
6e399cd1 DB |
1030 | struct file *old_exe_file; |
1031 | ||
1032 | /* | |
1033 | * It is safe to dereference the exe_file without RCU as | |
1034 | * this function is only called if nobody else can access | |
1035 | * this mm -- see comment above for justification. | |
1036 | */ | |
1037 | old_exe_file = rcu_dereference_raw(mm->exe_file); | |
90f31d0e | 1038 | |
38646013 JS |
1039 | if (new_exe_file) |
1040 | get_file(new_exe_file); | |
90f31d0e KK |
1041 | rcu_assign_pointer(mm->exe_file, new_exe_file); |
1042 | if (old_exe_file) | |
1043 | fput(old_exe_file); | |
38646013 JS |
1044 | } |
1045 | ||
90f31d0e KK |
1046 | /** |
1047 | * get_mm_exe_file - acquire a reference to the mm's executable file | |
1048 | * | |
1049 | * Returns %NULL if mm has no associated executable file. | |
1050 | * User must release file via fput(). | |
1051 | */ | |
38646013 JS |
1052 | struct file *get_mm_exe_file(struct mm_struct *mm) |
1053 | { | |
1054 | struct file *exe_file; | |
1055 | ||
90f31d0e KK |
1056 | rcu_read_lock(); |
1057 | exe_file = rcu_dereference(mm->exe_file); | |
1058 | if (exe_file && !get_file_rcu(exe_file)) | |
1059 | exe_file = NULL; | |
1060 | rcu_read_unlock(); | |
38646013 JS |
1061 | return exe_file; |
1062 | } | |
11163348 | 1063 | EXPORT_SYMBOL(get_mm_exe_file); |
38646013 | 1064 | |
cd81a917 MG |
1065 | /** |
1066 | * get_task_exe_file - acquire a reference to the task's executable file | |
1067 | * | |
1068 | * Returns %NULL if task's mm (if any) has no associated executable file or | |
1069 | * this is a kernel thread with borrowed mm (see the comment above get_task_mm). | |
1070 | * User must release file via fput(). | |
1071 | */ | |
1072 | struct file *get_task_exe_file(struct task_struct *task) | |
1073 | { | |
1074 | struct file *exe_file = NULL; | |
1075 | struct mm_struct *mm; | |
1076 | ||
1077 | task_lock(task); | |
1078 | mm = task->mm; | |
1079 | if (mm) { | |
1080 | if (!(task->flags & PF_KTHREAD)) | |
1081 | exe_file = get_mm_exe_file(mm); | |
1082 | } | |
1083 | task_unlock(task); | |
1084 | return exe_file; | |
1085 | } | |
1086 | EXPORT_SYMBOL(get_task_exe_file); | |
38646013 | 1087 | |
1da177e4 LT |
1088 | /** |
1089 | * get_task_mm - acquire a reference to the task's mm | |
1090 | * | |
246bb0b1 | 1091 | * Returns %NULL if the task has no mm. Checks PF_KTHREAD (meaning |
1da177e4 LT |
1092 | * this kernel workthread has transiently adopted a user mm with use_mm, |
1093 | * to do its AIO) is not set and if so returns a reference to it, after | |
1094 | * bumping up the use count. User must release the mm via mmput() | |
1095 | * after use. Typically used by /proc and ptrace. | |
1096 | */ | |
1097 | struct mm_struct *get_task_mm(struct task_struct *task) | |
1098 | { | |
1099 | struct mm_struct *mm; | |
1100 | ||
1101 | task_lock(task); | |
1102 | mm = task->mm; | |
1103 | if (mm) { | |
246bb0b1 | 1104 | if (task->flags & PF_KTHREAD) |
1da177e4 LT |
1105 | mm = NULL; |
1106 | else | |
3fce371b | 1107 | mmget(mm); |
1da177e4 LT |
1108 | } |
1109 | task_unlock(task); | |
1110 | return mm; | |
1111 | } | |
1112 | EXPORT_SYMBOL_GPL(get_task_mm); | |
1113 | ||
8cdb878d CY |
1114 | struct mm_struct *mm_access(struct task_struct *task, unsigned int mode) |
1115 | { | |
1116 | struct mm_struct *mm; | |
1117 | int err; | |
1118 | ||
1119 | err = mutex_lock_killable(&task->signal->cred_guard_mutex); | |
1120 | if (err) | |
1121 | return ERR_PTR(err); | |
1122 | ||
1123 | mm = get_task_mm(task); | |
1124 | if (mm && mm != current->mm && | |
1125 | !ptrace_may_access(task, mode)) { | |
1126 | mmput(mm); | |
1127 | mm = ERR_PTR(-EACCES); | |
1128 | } | |
1129 | mutex_unlock(&task->signal->cred_guard_mutex); | |
1130 | ||
1131 | return mm; | |
1132 | } | |
1133 | ||
57b59c4a | 1134 | static void complete_vfork_done(struct task_struct *tsk) |
c415c3b4 | 1135 | { |
d68b46fe | 1136 | struct completion *vfork; |
c415c3b4 | 1137 | |
d68b46fe ON |
1138 | task_lock(tsk); |
1139 | vfork = tsk->vfork_done; | |
1140 | if (likely(vfork)) { | |
1141 | tsk->vfork_done = NULL; | |
1142 | complete(vfork); | |
1143 | } | |
1144 | task_unlock(tsk); | |
1145 | } | |
1146 | ||
1147 | static int wait_for_vfork_done(struct task_struct *child, | |
1148 | struct completion *vfork) | |
1149 | { | |
1150 | int killed; | |
1151 | ||
1152 | freezer_do_not_count(); | |
1153 | killed = wait_for_completion_killable(vfork); | |
1154 | freezer_count(); | |
1155 | ||
1156 | if (killed) { | |
1157 | task_lock(child); | |
1158 | child->vfork_done = NULL; | |
1159 | task_unlock(child); | |
1160 | } | |
1161 | ||
1162 | put_task_struct(child); | |
1163 | return killed; | |
c415c3b4 ON |
1164 | } |
1165 | ||
1da177e4 LT |
1166 | /* Please note the differences between mmput and mm_release. |
1167 | * mmput is called whenever we stop holding onto a mm_struct, | |
1168 | * error success whatever. | |
1169 | * | |
1170 | * mm_release is called after a mm_struct has been removed | |
1171 | * from the current process. | |
1172 | * | |
1173 | * This difference is important for error handling, when we | |
1174 | * only half set up a mm_struct for a new process and need to restore | |
1175 | * the old one. Because we mmput the new mm_struct before | |
1176 | * restoring the old one. . . | |
1177 | * Eric Biederman 10 January 1998 | |
1178 | */ | |
0c10e4b6 | 1179 | static void mm_release(struct task_struct *tsk, struct mm_struct *mm) |
1da177e4 | 1180 | { |
0326f5a9 SD |
1181 | uprobe_free_utask(tsk); |
1182 | ||
1da177e4 LT |
1183 | /* Get rid of any cached register state */ |
1184 | deactivate_mm(tsk, mm); | |
1185 | ||
fec1d011 | 1186 | /* |
735f2770 MH |
1187 | * Signal userspace if we're not exiting with a core dump |
1188 | * because we want to leave the value intact for debugging | |
1189 | * purposes. | |
fec1d011 | 1190 | */ |
9c8a8228 | 1191 | if (tsk->clear_child_tid) { |
735f2770 | 1192 | if (!(tsk->signal->flags & SIGNAL_GROUP_COREDUMP) && |
9c8a8228 ED |
1193 | atomic_read(&mm->mm_users) > 1) { |
1194 | /* | |
1195 | * We don't check the error code - if userspace has | |
1196 | * not set up a proper pointer then tough luck. | |
1197 | */ | |
1198 | put_user(0, tsk->clear_child_tid); | |
1199 | sys_futex(tsk->clear_child_tid, FUTEX_WAKE, | |
1200 | 1, NULL, NULL, 0); | |
1201 | } | |
1da177e4 | 1202 | tsk->clear_child_tid = NULL; |
1da177e4 | 1203 | } |
f7505d64 KK |
1204 | |
1205 | /* | |
1206 | * All done, finally we can wake up parent and return this mm to him. | |
1207 | * Also kthread_stop() uses this completion for synchronization. | |
1208 | */ | |
1209 | if (tsk->vfork_done) | |
1210 | complete_vfork_done(tsk); | |
1da177e4 LT |
1211 | } |
1212 | ||
0c10e4b6 TG |
1213 | void exit_mm_release(struct task_struct *tsk, struct mm_struct *mm) |
1214 | { | |
b19fd552 | 1215 | futex_exit_release(tsk); |
0c10e4b6 TG |
1216 | mm_release(tsk, mm); |
1217 | } | |
1218 | ||
1219 | void exec_mm_release(struct task_struct *tsk, struct mm_struct *mm) | |
1220 | { | |
b19fd552 | 1221 | futex_exec_release(tsk); |
0c10e4b6 TG |
1222 | mm_release(tsk, mm); |
1223 | } | |
1224 | ||
a0a7ec30 JD |
1225 | /* |
1226 | * Allocate a new mm structure and copy contents from the | |
1227 | * mm structure of the passed in task structure. | |
1228 | */ | |
ff252c1f | 1229 | static struct mm_struct *dup_mm(struct task_struct *tsk) |
a0a7ec30 JD |
1230 | { |
1231 | struct mm_struct *mm, *oldmm = current->mm; | |
1232 | int err; | |
1233 | ||
a0a7ec30 JD |
1234 | mm = allocate_mm(); |
1235 | if (!mm) | |
1236 | goto fail_nomem; | |
1237 | ||
1238 | memcpy(mm, oldmm, sizeof(*mm)); | |
1239 | ||
bfedb589 | 1240 | if (!mm_init(mm, tsk, mm->user_ns)) |
a0a7ec30 JD |
1241 | goto fail_nomem; |
1242 | ||
a0a7ec30 JD |
1243 | err = dup_mmap(mm, oldmm); |
1244 | if (err) | |
1245 | goto free_pt; | |
1246 | ||
1247 | mm->hiwater_rss = get_mm_rss(mm); | |
1248 | mm->hiwater_vm = mm->total_vm; | |
1249 | ||
801460d0 HS |
1250 | if (mm->binfmt && !try_module_get(mm->binfmt->module)) |
1251 | goto free_pt; | |
1252 | ||
a0a7ec30 JD |
1253 | return mm; |
1254 | ||
1255 | free_pt: | |
801460d0 HS |
1256 | /* don't put binfmt in mmput, we haven't got module yet */ |
1257 | mm->binfmt = NULL; | |
343ef896 | 1258 | mm_init_owner(mm, NULL); |
a0a7ec30 JD |
1259 | mmput(mm); |
1260 | ||
1261 | fail_nomem: | |
1262 | return NULL; | |
a0a7ec30 JD |
1263 | } |
1264 | ||
fb0a685c | 1265 | static int copy_mm(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 | 1266 | { |
fb0a685c | 1267 | struct mm_struct *mm, *oldmm; |
1da177e4 LT |
1268 | int retval; |
1269 | ||
1270 | tsk->min_flt = tsk->maj_flt = 0; | |
1271 | tsk->nvcsw = tsk->nivcsw = 0; | |
17406b82 MSB |
1272 | #ifdef CONFIG_DETECT_HUNG_TASK |
1273 | tsk->last_switch_count = tsk->nvcsw + tsk->nivcsw; | |
1274 | #endif | |
1da177e4 LT |
1275 | |
1276 | tsk->mm = NULL; | |
1277 | tsk->active_mm = NULL; | |
1278 | ||
1279 | /* | |
1280 | * Are we cloning a kernel thread? | |
1281 | * | |
1282 | * We need to steal a active VM for that.. | |
1283 | */ | |
1284 | oldmm = current->mm; | |
1285 | if (!oldmm) | |
1286 | return 0; | |
1287 | ||
615d6e87 DB |
1288 | /* initialize the new vmacache entries */ |
1289 | vmacache_flush(tsk); | |
1290 | ||
1da177e4 | 1291 | if (clone_flags & CLONE_VM) { |
3fce371b | 1292 | mmget(oldmm); |
1da177e4 | 1293 | mm = oldmm; |
1da177e4 LT |
1294 | goto good_mm; |
1295 | } | |
1296 | ||
1297 | retval = -ENOMEM; | |
a0a7ec30 | 1298 | mm = dup_mm(tsk); |
1da177e4 LT |
1299 | if (!mm) |
1300 | goto fail_nomem; | |
1301 | ||
1da177e4 LT |
1302 | good_mm: |
1303 | tsk->mm = mm; | |
1304 | tsk->active_mm = mm; | |
1305 | return 0; | |
1306 | ||
1da177e4 LT |
1307 | fail_nomem: |
1308 | return retval; | |
1da177e4 LT |
1309 | } |
1310 | ||
a39bc516 | 1311 | static int copy_fs(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 | 1312 | { |
498052bb | 1313 | struct fs_struct *fs = current->fs; |
1da177e4 | 1314 | if (clone_flags & CLONE_FS) { |
498052bb | 1315 | /* tsk->fs is already what we want */ |
2a4419b5 | 1316 | spin_lock(&fs->lock); |
498052bb | 1317 | if (fs->in_exec) { |
2a4419b5 | 1318 | spin_unlock(&fs->lock); |
498052bb AV |
1319 | return -EAGAIN; |
1320 | } | |
1321 | fs->users++; | |
2a4419b5 | 1322 | spin_unlock(&fs->lock); |
1da177e4 LT |
1323 | return 0; |
1324 | } | |
498052bb | 1325 | tsk->fs = copy_fs_struct(fs); |
1da177e4 LT |
1326 | if (!tsk->fs) |
1327 | return -ENOMEM; | |
1328 | return 0; | |
1329 | } | |
1330 | ||
fb0a685c | 1331 | static int copy_files(unsigned long clone_flags, struct task_struct *tsk) |
a016f338 JD |
1332 | { |
1333 | struct files_struct *oldf, *newf; | |
1334 | int error = 0; | |
1335 | ||
1336 | /* | |
1337 | * A background process may not have any files ... | |
1338 | */ | |
1339 | oldf = current->files; | |
1340 | if (!oldf) | |
1341 | goto out; | |
1342 | ||
1343 | if (clone_flags & CLONE_FILES) { | |
1344 | atomic_inc(&oldf->count); | |
1345 | goto out; | |
1346 | } | |
1347 | ||
a016f338 JD |
1348 | newf = dup_fd(oldf, &error); |
1349 | if (!newf) | |
1350 | goto out; | |
1351 | ||
1352 | tsk->files = newf; | |
1353 | error = 0; | |
1354 | out: | |
1355 | return error; | |
1356 | } | |
1357 | ||
fadad878 | 1358 | static int copy_io(unsigned long clone_flags, struct task_struct *tsk) |
fd0928df JA |
1359 | { |
1360 | #ifdef CONFIG_BLOCK | |
1361 | struct io_context *ioc = current->io_context; | |
6e736be7 | 1362 | struct io_context *new_ioc; |
fd0928df JA |
1363 | |
1364 | if (!ioc) | |
1365 | return 0; | |
fadad878 JA |
1366 | /* |
1367 | * Share io context with parent, if CLONE_IO is set | |
1368 | */ | |
1369 | if (clone_flags & CLONE_IO) { | |
3d48749d TH |
1370 | ioc_task_link(ioc); |
1371 | tsk->io_context = ioc; | |
fadad878 | 1372 | } else if (ioprio_valid(ioc->ioprio)) { |
6e736be7 TH |
1373 | new_ioc = get_task_io_context(tsk, GFP_KERNEL, NUMA_NO_NODE); |
1374 | if (unlikely(!new_ioc)) | |
fd0928df JA |
1375 | return -ENOMEM; |
1376 | ||
6e736be7 | 1377 | new_ioc->ioprio = ioc->ioprio; |
11a3122f | 1378 | put_io_context(new_ioc); |
fd0928df JA |
1379 | } |
1380 | #endif | |
1381 | return 0; | |
1382 | } | |
1383 | ||
a39bc516 | 1384 | static int copy_sighand(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 LT |
1385 | { |
1386 | struct sighand_struct *sig; | |
1387 | ||
60348802 | 1388 | if (clone_flags & CLONE_SIGHAND) { |
1da177e4 LT |
1389 | atomic_inc(¤t->sighand->count); |
1390 | return 0; | |
1391 | } | |
1392 | sig = kmem_cache_alloc(sighand_cachep, GFP_KERNEL); | |
e56d0903 | 1393 | rcu_assign_pointer(tsk->sighand, sig); |
1da177e4 LT |
1394 | if (!sig) |
1395 | return -ENOMEM; | |
9d7fb042 | 1396 | |
1da177e4 | 1397 | atomic_set(&sig->count, 1); |
9aa84979 | 1398 | spin_lock_irq(¤t->sighand->siglock); |
1da177e4 | 1399 | memcpy(sig->action, current->sighand->action, sizeof(sig->action)); |
9aa84979 | 1400 | spin_unlock_irq(¤t->sighand->siglock); |
1da177e4 LT |
1401 | return 0; |
1402 | } | |
1403 | ||
a7e5328a | 1404 | void __cleanup_sighand(struct sighand_struct *sighand) |
c81addc9 | 1405 | { |
d80e731e ON |
1406 | if (atomic_dec_and_test(&sighand->count)) { |
1407 | signalfd_cleanup(sighand); | |
392809b2 | 1408 | /* |
5f0d5a3a | 1409 | * sighand_cachep is SLAB_TYPESAFE_BY_RCU so we can free it |
392809b2 ON |
1410 | * without an RCU grace period, see __lock_task_sighand(). |
1411 | */ | |
c81addc9 | 1412 | kmem_cache_free(sighand_cachep, sighand); |
d80e731e | 1413 | } |
c81addc9 ON |
1414 | } |
1415 | ||
b18b6a9c | 1416 | #ifdef CONFIG_POSIX_TIMERS |
f06febc9 FM |
1417 | /* |
1418 | * Initialize POSIX timer handling for a thread group. | |
1419 | */ | |
1420 | static void posix_cpu_timers_init_group(struct signal_struct *sig) | |
1421 | { | |
78d7d407 JS |
1422 | unsigned long cpu_limit; |
1423 | ||
316c1608 | 1424 | cpu_limit = READ_ONCE(sig->rlim[RLIMIT_CPU].rlim_cur); |
78d7d407 | 1425 | if (cpu_limit != RLIM_INFINITY) { |
ebd7e7fc | 1426 | sig->cputime_expires.prof_exp = cpu_limit * NSEC_PER_SEC; |
d5c373eb | 1427 | sig->cputimer.running = true; |
6279a751 ON |
1428 | } |
1429 | ||
f06febc9 FM |
1430 | /* The timer lists. */ |
1431 | INIT_LIST_HEAD(&sig->cpu_timers[0]); | |
1432 | INIT_LIST_HEAD(&sig->cpu_timers[1]); | |
1433 | INIT_LIST_HEAD(&sig->cpu_timers[2]); | |
1434 | } | |
b18b6a9c NP |
1435 | #else |
1436 | static inline void posix_cpu_timers_init_group(struct signal_struct *sig) { } | |
1437 | #endif | |
f06febc9 | 1438 | |
a39bc516 | 1439 | static int copy_signal(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 LT |
1440 | { |
1441 | struct signal_struct *sig; | |
1da177e4 | 1442 | |
4ab6c083 | 1443 | if (clone_flags & CLONE_THREAD) |
490dea45 | 1444 | return 0; |
490dea45 | 1445 | |
a56704ef | 1446 | sig = kmem_cache_zalloc(signal_cachep, GFP_KERNEL); |
1da177e4 LT |
1447 | tsk->signal = sig; |
1448 | if (!sig) | |
1449 | return -ENOMEM; | |
1450 | ||
b3ac022c | 1451 | sig->nr_threads = 1; |
1da177e4 | 1452 | atomic_set(&sig->live, 1); |
b3ac022c | 1453 | atomic_set(&sig->sigcnt, 1); |
0c740d0a ON |
1454 | |
1455 | /* list_add(thread_node, thread_head) without INIT_LIST_HEAD() */ | |
1456 | sig->thread_head = (struct list_head)LIST_HEAD_INIT(tsk->thread_node); | |
1457 | tsk->thread_node = (struct list_head)LIST_HEAD_INIT(sig->thread_head); | |
1458 | ||
1da177e4 | 1459 | init_waitqueue_head(&sig->wait_chldexit); |
db51aecc | 1460 | sig->curr_target = tsk; |
1da177e4 | 1461 | init_sigpending(&sig->shared_pending); |
e78c3496 | 1462 | seqlock_init(&sig->stats_lock); |
9d7fb042 | 1463 | prev_cputime_init(&sig->prev_cputime); |
1da177e4 | 1464 | |
baa73d9e | 1465 | #ifdef CONFIG_POSIX_TIMERS |
b18b6a9c | 1466 | INIT_LIST_HEAD(&sig->posix_timers); |
c9cb2e3d | 1467 | hrtimer_init(&sig->real_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL); |
1da177e4 | 1468 | sig->real_timer.function = it_real_fn; |
baa73d9e | 1469 | #endif |
1da177e4 | 1470 | |
1da177e4 LT |
1471 | task_lock(current->group_leader); |
1472 | memcpy(sig->rlim, current->signal->rlim, sizeof sig->rlim); | |
1473 | task_unlock(current->group_leader); | |
1474 | ||
6279a751 ON |
1475 | posix_cpu_timers_init_group(sig); |
1476 | ||
522ed776 | 1477 | tty_audit_fork(sig); |
5091faa4 | 1478 | sched_autogroup_fork(sig); |
522ed776 | 1479 | |
a63d83f4 | 1480 | sig->oom_score_adj = current->signal->oom_score_adj; |
dabb16f6 | 1481 | sig->oom_score_adj_min = current->signal->oom_score_adj_min; |
28b83c51 | 1482 | |
9b1bf12d KM |
1483 | mutex_init(&sig->cred_guard_mutex); |
1484 | ||
1da177e4 LT |
1485 | return 0; |
1486 | } | |
1487 | ||
dbd95212 KC |
1488 | static void copy_seccomp(struct task_struct *p) |
1489 | { | |
1490 | #ifdef CONFIG_SECCOMP | |
1491 | /* | |
1492 | * Must be called with sighand->lock held, which is common to | |
1493 | * all threads in the group. Holding cred_guard_mutex is not | |
1494 | * needed because this new task is not yet running and cannot | |
1495 | * be racing exec. | |
1496 | */ | |
69f6a34b | 1497 | assert_spin_locked(¤t->sighand->siglock); |
dbd95212 KC |
1498 | |
1499 | /* Ref-count the new filter user, and assign it. */ | |
1500 | get_seccomp_filter(current); | |
1501 | p->seccomp = current->seccomp; | |
1502 | ||
1503 | /* | |
1504 | * Explicitly enable no_new_privs here in case it got set | |
1505 | * between the task_struct being duplicated and holding the | |
1506 | * sighand lock. The seccomp state and nnp must be in sync. | |
1507 | */ | |
1508 | if (task_no_new_privs(current)) | |
1509 | task_set_no_new_privs(p); | |
1510 | ||
1511 | /* | |
1512 | * If the parent gained a seccomp mode after copying thread | |
1513 | * flags and between before we held the sighand lock, we have | |
1514 | * to manually enable the seccomp thread flag here. | |
1515 | */ | |
1516 | if (p->seccomp.mode != SECCOMP_MODE_DISABLED) | |
1517 | set_tsk_thread_flag(p, TIF_SECCOMP); | |
1518 | #endif | |
1519 | } | |
1520 | ||
17da2bd9 | 1521 | SYSCALL_DEFINE1(set_tid_address, int __user *, tidptr) |
1da177e4 LT |
1522 | { |
1523 | current->clear_child_tid = tidptr; | |
1524 | ||
b488893a | 1525 | return task_pid_vnr(current); |
1da177e4 LT |
1526 | } |
1527 | ||
a39bc516 | 1528 | static void rt_mutex_init_task(struct task_struct *p) |
23f78d4a | 1529 | { |
1d615482 | 1530 | raw_spin_lock_init(&p->pi_lock); |
e29e175b | 1531 | #ifdef CONFIG_RT_MUTEXES |
a23ba907 | 1532 | p->pi_waiters = RB_ROOT_CACHED; |
e96a7705 | 1533 | p->pi_top_task = NULL; |
23f78d4a | 1534 | p->pi_blocked_on = NULL; |
23f78d4a IM |
1535 | #endif |
1536 | } | |
1537 | ||
b18b6a9c | 1538 | #ifdef CONFIG_POSIX_TIMERS |
f06febc9 FM |
1539 | /* |
1540 | * Initialize POSIX timer handling for a single task. | |
1541 | */ | |
1542 | static void posix_cpu_timers_init(struct task_struct *tsk) | |
1543 | { | |
64861634 MS |
1544 | tsk->cputime_expires.prof_exp = 0; |
1545 | tsk->cputime_expires.virt_exp = 0; | |
f06febc9 FM |
1546 | tsk->cputime_expires.sched_exp = 0; |
1547 | INIT_LIST_HEAD(&tsk->cpu_timers[0]); | |
1548 | INIT_LIST_HEAD(&tsk->cpu_timers[1]); | |
1549 | INIT_LIST_HEAD(&tsk->cpu_timers[2]); | |
1550 | } | |
b18b6a9c NP |
1551 | #else |
1552 | static inline void posix_cpu_timers_init(struct task_struct *tsk) { } | |
1553 | #endif | |
f06febc9 | 1554 | |
81907739 ON |
1555 | static inline void |
1556 | init_task_pid(struct task_struct *task, enum pid_type type, struct pid *pid) | |
1557 | { | |
1558 | task->pids[type].pid = pid; | |
1559 | } | |
1560 | ||
6bfbaa51 IM |
1561 | static inline void rcu_copy_process(struct task_struct *p) |
1562 | { | |
1563 | #ifdef CONFIG_PREEMPT_RCU | |
1564 | p->rcu_read_lock_nesting = 0; | |
1565 | p->rcu_read_unlock_special.s = 0; | |
1566 | p->rcu_blocked_node = NULL; | |
1567 | INIT_LIST_HEAD(&p->rcu_node_entry); | |
1568 | #endif /* #ifdef CONFIG_PREEMPT_RCU */ | |
1569 | #ifdef CONFIG_TASKS_RCU | |
1570 | p->rcu_tasks_holdout = false; | |
1571 | INIT_LIST_HEAD(&p->rcu_tasks_holdout_list); | |
1572 | p->rcu_tasks_idle_cpu = -1; | |
1573 | #endif /* #ifdef CONFIG_TASKS_RCU */ | |
1574 | } | |
1575 | ||
343ef896 AA |
1576 | static void __delayed_free_task(struct rcu_head *rhp) |
1577 | { | |
1578 | struct task_struct *tsk = container_of(rhp, struct task_struct, rcu); | |
1579 | ||
1580 | free_task(tsk); | |
1581 | } | |
1582 | ||
1583 | static __always_inline void delayed_free_task(struct task_struct *tsk) | |
1584 | { | |
1585 | if (IS_ENABLED(CONFIG_MEMCG)) | |
1586 | call_rcu(&tsk->rcu, __delayed_free_task); | |
1587 | else | |
1588 | free_task(tsk); | |
1589 | } | |
1590 | ||
1da177e4 LT |
1591 | /* |
1592 | * This creates a new process as a copy of the old one, | |
1593 | * but does not actually start it yet. | |
1594 | * | |
1595 | * It copies the registers, and all the appropriate | |
1596 | * parts of the process environment (as per the clone | |
1597 | * flags). The actual kick-off is left to the caller. | |
1598 | */ | |
0766f788 ER |
1599 | static __latent_entropy struct task_struct *copy_process( |
1600 | unsigned long clone_flags, | |
36c8b586 | 1601 | unsigned long stack_start, |
36c8b586 | 1602 | unsigned long stack_size, |
36c8b586 | 1603 | int __user *child_tidptr, |
09a05394 | 1604 | struct pid *pid, |
3033f14a | 1605 | int trace, |
725fc629 AK |
1606 | unsigned long tls, |
1607 | int node) | |
1da177e4 LT |
1608 | { |
1609 | int retval; | |
a24efe62 | 1610 | struct task_struct *p; |
1da177e4 LT |
1611 | |
1612 | if ((clone_flags & (CLONE_NEWNS|CLONE_FS)) == (CLONE_NEWNS|CLONE_FS)) | |
1613 | return ERR_PTR(-EINVAL); | |
1614 | ||
e66eded8 EB |
1615 | if ((clone_flags & (CLONE_NEWUSER|CLONE_FS)) == (CLONE_NEWUSER|CLONE_FS)) |
1616 | return ERR_PTR(-EINVAL); | |
1617 | ||
680f7875 SH |
1618 | if ((clone_flags & CLONE_NEWUSER) && !unprivileged_userns_clone) |
1619 | if (!capable(CAP_SYS_ADMIN)) | |
1620 | return ERR_PTR(-EPERM); | |
1621 | ||
1da177e4 LT |
1622 | /* |
1623 | * Thread groups must share signals as well, and detached threads | |
1624 | * can only be started up within the thread group. | |
1625 | */ | |
1626 | if ((clone_flags & CLONE_THREAD) && !(clone_flags & CLONE_SIGHAND)) | |
1627 | return ERR_PTR(-EINVAL); | |
1628 | ||
1629 | /* | |
1630 | * Shared signal handlers imply shared VM. By way of the above, | |
1631 | * thread groups also imply shared VM. Blocking this case allows | |
1632 | * for various simplifications in other code. | |
1633 | */ | |
1634 | if ((clone_flags & CLONE_SIGHAND) && !(clone_flags & CLONE_VM)) | |
1635 | return ERR_PTR(-EINVAL); | |
1636 | ||
123be07b SB |
1637 | /* |
1638 | * Siblings of global init remain as zombies on exit since they are | |
1639 | * not reaped by their parent (swapper). To solve this and to avoid | |
1640 | * multi-rooted process trees, prevent global and container-inits | |
1641 | * from creating siblings. | |
1642 | */ | |
1643 | if ((clone_flags & CLONE_PARENT) && | |
1644 | current->signal->flags & SIGNAL_UNKILLABLE) | |
1645 | return ERR_PTR(-EINVAL); | |
1646 | ||
8382fcac | 1647 | /* |
40a0d32d | 1648 | * If the new process will be in a different pid or user namespace |
faf00da5 | 1649 | * do not allow it to share a thread group with the forking task. |
8382fcac | 1650 | */ |
faf00da5 | 1651 | if (clone_flags & CLONE_THREAD) { |
40a0d32d ON |
1652 | if ((clone_flags & (CLONE_NEWUSER | CLONE_NEWPID)) || |
1653 | (task_active_pid_ns(current) != | |
1654 | current->nsproxy->pid_ns_for_children)) | |
1655 | return ERR_PTR(-EINVAL); | |
1656 | } | |
8382fcac | 1657 | |
1da177e4 | 1658 | retval = -ENOMEM; |
725fc629 | 1659 | p = dup_task_struct(current, node); |
1da177e4 LT |
1660 | if (!p) |
1661 | goto fork_out; | |
1662 | ||
4d6501dc VN |
1663 | /* |
1664 | * This _must_ happen before we call free_task(), i.e. before we jump | |
1665 | * to any of the bad_fork_* labels. This is to avoid freeing | |
1666 | * p->set_child_tid which is (ab)used as a kthread's data pointer for | |
1667 | * kernel threads (PF_KTHREAD). | |
1668 | */ | |
1669 | p->set_child_tid = (clone_flags & CLONE_CHILD_SETTID) ? child_tidptr : NULL; | |
1670 | /* | |
1671 | * Clear TID on mm_release()? | |
1672 | */ | |
1673 | p->clear_child_tid = (clone_flags & CLONE_CHILD_CLEARTID) ? child_tidptr : NULL; | |
1674 | ||
f7e8b616 SR |
1675 | ftrace_graph_init_task(p); |
1676 | ||
bea493a0 PZ |
1677 | rt_mutex_init_task(p); |
1678 | ||
d12c1a37 | 1679 | #ifdef CONFIG_PROVE_LOCKING |
de30a2b3 IM |
1680 | DEBUG_LOCKS_WARN_ON(!p->hardirqs_enabled); |
1681 | DEBUG_LOCKS_WARN_ON(!p->softirqs_enabled); | |
1682 | #endif | |
1da177e4 | 1683 | retval = -EAGAIN; |
3b11a1de | 1684 | if (atomic_read(&p->real_cred->user->processes) >= |
78d7d407 | 1685 | task_rlimit(p, RLIMIT_NPROC)) { |
b57922b6 EP |
1686 | if (p->real_cred->user != INIT_USER && |
1687 | !capable(CAP_SYS_RESOURCE) && !capable(CAP_SYS_ADMIN)) | |
1da177e4 LT |
1688 | goto bad_fork_free; |
1689 | } | |
72fa5997 | 1690 | current->flags &= ~PF_NPROC_EXCEEDED; |
1da177e4 | 1691 | |
f1752eec DH |
1692 | retval = copy_creds(p, clone_flags); |
1693 | if (retval < 0) | |
1694 | goto bad_fork_free; | |
1da177e4 LT |
1695 | |
1696 | /* | |
1697 | * If multiple threads are within copy_process(), then this check | |
1698 | * triggers too late. This doesn't hurt, the check is only there | |
1699 | * to stop root fork bombs. | |
1700 | */ | |
04ec93fe | 1701 | retval = -EAGAIN; |
1da177e4 LT |
1702 | if (nr_threads >= max_threads) |
1703 | goto bad_fork_cleanup_count; | |
1704 | ||
ca74e92b | 1705 | delayacct_tsk_init(p); /* Must remain after dup_task_struct() */ |
c1de45ca | 1706 | p->flags &= ~(PF_SUPERPRIV | PF_WQ_WORKER | PF_IDLE); |
514ddb44 | 1707 | p->flags |= PF_FORKNOEXEC; |
1da177e4 LT |
1708 | INIT_LIST_HEAD(&p->children); |
1709 | INIT_LIST_HEAD(&p->sibling); | |
f41d911f | 1710 | rcu_copy_process(p); |
1da177e4 LT |
1711 | p->vfork_done = NULL; |
1712 | spin_lock_init(&p->alloc_lock); | |
1da177e4 | 1713 | |
1da177e4 LT |
1714 | init_sigpending(&p->pending); |
1715 | ||
64861634 | 1716 | p->utime = p->stime = p->gtime = 0; |
40565b5a | 1717 | #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME |
64861634 | 1718 | p->utimescaled = p->stimescaled = 0; |
40565b5a | 1719 | #endif |
9d7fb042 PZ |
1720 | prev_cputime_init(&p->prev_cputime); |
1721 | ||
6a61671b | 1722 | #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN |
bac5b6b6 FW |
1723 | seqcount_init(&p->vtime.seqcount); |
1724 | p->vtime.starttime = 0; | |
1725 | p->vtime.state = VTIME_INACTIVE; | |
6a61671b FW |
1726 | #endif |
1727 | ||
a3a2e76c KH |
1728 | #if defined(SPLIT_RSS_COUNTING) |
1729 | memset(&p->rss_stat, 0, sizeof(p->rss_stat)); | |
1730 | #endif | |
172ba844 | 1731 | |
6976675d AV |
1732 | p->default_timer_slack_ns = current->timer_slack_ns; |
1733 | ||
5995477a | 1734 | task_io_accounting_init(&p->ioac); |
1da177e4 LT |
1735 | acct_clear_integrals(p); |
1736 | ||
f06febc9 | 1737 | posix_cpu_timers_init(p); |
1da177e4 | 1738 | |
1da177e4 | 1739 | p->io_context = NULL; |
1da177e4 | 1740 | p->audit_context = NULL; |
b4f48b63 | 1741 | cgroup_fork(p); |
1da177e4 | 1742 | #ifdef CONFIG_NUMA |
846a16bf | 1743 | p->mempolicy = mpol_dup(p->mempolicy); |
fb0a685c DRO |
1744 | if (IS_ERR(p->mempolicy)) { |
1745 | retval = PTR_ERR(p->mempolicy); | |
1746 | p->mempolicy = NULL; | |
e8604cb4 | 1747 | goto bad_fork_cleanup_threadgroup_lock; |
fb0a685c | 1748 | } |
1da177e4 | 1749 | #endif |
778d3b0f MH |
1750 | #ifdef CONFIG_CPUSETS |
1751 | p->cpuset_mem_spread_rotor = NUMA_NO_NODE; | |
1752 | p->cpuset_slab_spread_rotor = NUMA_NO_NODE; | |
cc9a6c87 | 1753 | seqcount_init(&p->mems_allowed_seq); |
778d3b0f | 1754 | #endif |
de30a2b3 IM |
1755 | #ifdef CONFIG_TRACE_IRQFLAGS |
1756 | p->irq_events = 0; | |
1757 | p->hardirqs_enabled = 0; | |
1758 | p->hardirq_enable_ip = 0; | |
1759 | p->hardirq_enable_event = 0; | |
1760 | p->hardirq_disable_ip = _THIS_IP_; | |
1761 | p->hardirq_disable_event = 0; | |
1762 | p->softirqs_enabled = 1; | |
1763 | p->softirq_enable_ip = _THIS_IP_; | |
1764 | p->softirq_enable_event = 0; | |
1765 | p->softirq_disable_ip = 0; | |
1766 | p->softirq_disable_event = 0; | |
1767 | p->hardirq_context = 0; | |
1768 | p->softirq_context = 0; | |
1769 | #endif | |
8bcbde54 DH |
1770 | |
1771 | p->pagefault_disabled = 0; | |
1772 | ||
fbb9ce95 IM |
1773 | #ifdef CONFIG_LOCKDEP |
1774 | p->lockdep_depth = 0; /* no locks held yet */ | |
1775 | p->curr_chain_key = 0; | |
1776 | p->lockdep_recursion = 0; | |
b09be676 | 1777 | lockdep_init_task(p); |
fbb9ce95 | 1778 | #endif |
1da177e4 | 1779 | |
408894ee IM |
1780 | #ifdef CONFIG_DEBUG_MUTEXES |
1781 | p->blocked_on = NULL; /* not blocked yet */ | |
1782 | #endif | |
cafe5635 KO |
1783 | #ifdef CONFIG_BCACHE |
1784 | p->sequential_io = 0; | |
1785 | p->sequential_io_avg = 0; | |
1786 | #endif | |
09227ab6 JJ |
1787 | #ifdef CONFIG_SECURITY |
1788 | p->security = NULL; | |
1789 | #endif | |
0f481406 | 1790 | |
3c90e6e9 | 1791 | /* Perform scheduler related setup. Assign this task to a CPU. */ |
aab03e05 DF |
1792 | retval = sched_fork(clone_flags, p); |
1793 | if (retval) | |
1794 | goto bad_fork_cleanup_policy; | |
6ab423e0 | 1795 | |
cdd6c482 | 1796 | retval = perf_event_init_task(p); |
6ab423e0 PZ |
1797 | if (retval) |
1798 | goto bad_fork_cleanup_policy; | |
fb0a685c DRO |
1799 | retval = audit_alloc(p); |
1800 | if (retval) | |
6c72e350 | 1801 | goto bad_fork_cleanup_perf; |
1da177e4 | 1802 | /* copy all the process information */ |
ab602f79 | 1803 | shm_init_task(p); |
e4e55b47 | 1804 | retval = security_task_alloc(p, clone_flags); |
fb0a685c | 1805 | if (retval) |
1da177e4 | 1806 | goto bad_fork_cleanup_audit; |
e4e55b47 TH |
1807 | retval = copy_semundo(clone_flags, p); |
1808 | if (retval) | |
1809 | goto bad_fork_cleanup_security; | |
fb0a685c DRO |
1810 | retval = copy_files(clone_flags, p); |
1811 | if (retval) | |
1da177e4 | 1812 | goto bad_fork_cleanup_semundo; |
fb0a685c DRO |
1813 | retval = copy_fs(clone_flags, p); |
1814 | if (retval) | |
1da177e4 | 1815 | goto bad_fork_cleanup_files; |
fb0a685c DRO |
1816 | retval = copy_sighand(clone_flags, p); |
1817 | if (retval) | |
1da177e4 | 1818 | goto bad_fork_cleanup_fs; |
fb0a685c DRO |
1819 | retval = copy_signal(clone_flags, p); |
1820 | if (retval) | |
1da177e4 | 1821 | goto bad_fork_cleanup_sighand; |
fb0a685c DRO |
1822 | retval = copy_mm(clone_flags, p); |
1823 | if (retval) | |
1da177e4 | 1824 | goto bad_fork_cleanup_signal; |
fb0a685c DRO |
1825 | retval = copy_namespaces(clone_flags, p); |
1826 | if (retval) | |
d84f4f99 | 1827 | goto bad_fork_cleanup_mm; |
fb0a685c DRO |
1828 | retval = copy_io(clone_flags, p); |
1829 | if (retval) | |
fd0928df | 1830 | goto bad_fork_cleanup_namespaces; |
3033f14a | 1831 | retval = copy_thread_tls(clone_flags, stack_start, stack_size, p, tls); |
1da177e4 | 1832 | if (retval) |
fd0928df | 1833 | goto bad_fork_cleanup_io; |
1da177e4 | 1834 | |
425fb2b4 | 1835 | if (pid != &init_struct_pid) { |
c2b1df2e | 1836 | pid = alloc_pid(p->nsproxy->pid_ns_for_children); |
35f71bc0 MH |
1837 | if (IS_ERR(pid)) { |
1838 | retval = PTR_ERR(pid); | |
0740aa5f | 1839 | goto bad_fork_cleanup_thread; |
35f71bc0 | 1840 | } |
425fb2b4 PE |
1841 | } |
1842 | ||
73c10101 JA |
1843 | #ifdef CONFIG_BLOCK |
1844 | p->plug = NULL; | |
1845 | #endif | |
e891c8f7 TG |
1846 | futex_init_task(p); |
1847 | ||
f9a3879a GM |
1848 | /* |
1849 | * sigaltstack should be cleared when sharing the same VM | |
1850 | */ | |
1851 | if ((clone_flags & (CLONE_VM|CLONE_VFORK)) == CLONE_VM) | |
2a742138 | 1852 | sas_ss_reset(p); |
f9a3879a | 1853 | |
1da177e4 | 1854 | /* |
6580807d ON |
1855 | * Syscall tracing and stepping should be turned off in the |
1856 | * child regardless of CLONE_PTRACE. | |
1da177e4 | 1857 | */ |
6580807d | 1858 | user_disable_single_step(p); |
1da177e4 | 1859 | clear_tsk_thread_flag(p, TIF_SYSCALL_TRACE); |
ed75e8d5 LV |
1860 | #ifdef TIF_SYSCALL_EMU |
1861 | clear_tsk_thread_flag(p, TIF_SYSCALL_EMU); | |
1862 | #endif | |
9745512c | 1863 | clear_all_latency_tracing(p); |
1da177e4 | 1864 | |
1da177e4 | 1865 | /* ok, now we should be set up.. */ |
18c830df ON |
1866 | p->pid = pid_nr(pid); |
1867 | if (clone_flags & CLONE_THREAD) { | |
5f8aadd8 | 1868 | p->exit_signal = -1; |
18c830df ON |
1869 | p->group_leader = current->group_leader; |
1870 | p->tgid = current->tgid; | |
1871 | } else { | |
1872 | if (clone_flags & CLONE_PARENT) | |
1873 | p->exit_signal = current->group_leader->exit_signal; | |
1874 | else | |
1875 | p->exit_signal = (clone_flags & CSIGNAL); | |
1876 | p->group_leader = p; | |
1877 | p->tgid = p->pid; | |
1878 | } | |
5f8aadd8 | 1879 | |
9d823e8f WF |
1880 | p->nr_dirtied = 0; |
1881 | p->nr_dirtied_pause = 128 >> (PAGE_SHIFT - 10); | |
83712358 | 1882 | p->dirty_paused_when = 0; |
9d823e8f | 1883 | |
bb8cbbfe | 1884 | p->pdeath_signal = 0; |
47e65328 | 1885 | INIT_LIST_HEAD(&p->thread_group); |
158e1645 | 1886 | p->task_works = NULL; |
1da177e4 | 1887 | |
780de9dd | 1888 | cgroup_threadgroup_change_begin(current); |
7e47682e AS |
1889 | /* |
1890 | * Ensure that the cgroup subsystem policies allow the new process to be | |
1891 | * forked. It should be noted the the new process's css_set can be changed | |
1892 | * between here and cgroup_post_fork() if an organisation operation is in | |
1893 | * progress. | |
1894 | */ | |
b53202e6 | 1895 | retval = cgroup_can_fork(p); |
7e47682e AS |
1896 | if (retval) |
1897 | goto bad_fork_free_pid; | |
1898 | ||
e78a7b49 DH |
1899 | /* |
1900 | * From this point on we must avoid any synchronous user-space | |
1901 | * communication until we take the tasklist-lock. In particular, we do | |
1902 | * not want user-space to be able to predict the process start-time by | |
1903 | * stalling fork(2) after we recorded the start_time but before it is | |
1904 | * visible to the system. | |
1905 | */ | |
1906 | ||
1907 | p->start_time = ktime_get_ns(); | |
1908 | p->real_start_time = ktime_get_boot_ns(); | |
1909 | ||
18c830df ON |
1910 | /* |
1911 | * Make it visible to the rest of the system, but dont wake it up yet. | |
1912 | * Need tasklist lock for parent etc handling! | |
1913 | */ | |
1da177e4 LT |
1914 | write_lock_irq(&tasklist_lock); |
1915 | ||
1da177e4 | 1916 | /* CLONE_PARENT re-uses the old parent */ |
2d5516cb | 1917 | if (clone_flags & (CLONE_PARENT|CLONE_THREAD)) { |
1da177e4 | 1918 | p->real_parent = current->real_parent; |
2d5516cb ON |
1919 | p->parent_exec_id = current->parent_exec_id; |
1920 | } else { | |
1da177e4 | 1921 | p->real_parent = current; |
2d5516cb ON |
1922 | p->parent_exec_id = current->self_exec_id; |
1923 | } | |
1da177e4 | 1924 | |
d83a7cb3 JP |
1925 | klp_copy_process(p); |
1926 | ||
3f17da69 | 1927 | spin_lock(¤t->sighand->siglock); |
4a2c7a78 | 1928 | |
dbd95212 KC |
1929 | /* |
1930 | * Copy seccomp details explicitly here, in case they were changed | |
1931 | * before holding sighand lock. | |
1932 | */ | |
1933 | copy_seccomp(p); | |
1934 | ||
4a2c7a78 ON |
1935 | /* |
1936 | * Process group and session signals need to be delivered to just the | |
1937 | * parent before the fork or both the parent and the child after the | |
1938 | * fork. Restart if a signal comes in before we add the new process to | |
1939 | * it's process group. | |
1940 | * A fatal signal pending means that current will exit, so the new | |
1941 | * thread can't slip out of an OOM kill (or normal SIGKILL). | |
fb0a685c | 1942 | */ |
23ff4440 | 1943 | recalc_sigpending(); |
4a2c7a78 | 1944 | if (signal_pending(current)) { |
4a2c7a78 | 1945 | retval = -ERESTARTNOINTR; |
7e47682e | 1946 | goto bad_fork_cancel_cgroup; |
4a2c7a78 | 1947 | } |
e8cfbc24 | 1948 | if (unlikely(!(ns_of_pid(pid)->pid_allocated & PIDNS_ADDING))) { |
3fd37226 KT |
1949 | retval = -ENOMEM; |
1950 | goto bad_fork_cancel_cgroup; | |
1951 | } | |
4a2c7a78 | 1952 | |
73b9ebfe | 1953 | if (likely(p->pid)) { |
4b9d33e6 | 1954 | ptrace_init_task(p, (clone_flags & CLONE_PTRACE) || trace); |
73b9ebfe | 1955 | |
81907739 | 1956 | init_task_pid(p, PIDTYPE_PID, pid); |
73b9ebfe | 1957 | if (thread_group_leader(p)) { |
81907739 ON |
1958 | init_task_pid(p, PIDTYPE_PGID, task_pgrp(current)); |
1959 | init_task_pid(p, PIDTYPE_SID, task_session(current)); | |
1960 | ||
1c4042c2 | 1961 | if (is_child_reaper(pid)) { |
17cf22c3 | 1962 | ns_of_pid(pid)->child_reaper = p; |
1c4042c2 EB |
1963 | p->signal->flags |= SIGNAL_UNKILLABLE; |
1964 | } | |
73b9ebfe | 1965 | |
fea9d175 | 1966 | p->signal->leader_pid = pid; |
9c9f4ded | 1967 | p->signal->tty = tty_kref_get(current->signal->tty); |
749860ce PT |
1968 | /* |
1969 | * Inherit has_child_subreaper flag under the same | |
1970 | * tasklist_lock with adding child to the process tree | |
1971 | * for propagate_has_child_subreaper optimization. | |
1972 | */ | |
1973 | p->signal->has_child_subreaper = p->real_parent->signal->has_child_subreaper || | |
1974 | p->real_parent->signal->is_child_subreaper; | |
9cd80bbb | 1975 | list_add_tail(&p->sibling, &p->real_parent->children); |
5e85d4ab | 1976 | list_add_tail_rcu(&p->tasks, &init_task.tasks); |
81907739 ON |
1977 | attach_pid(p, PIDTYPE_PGID); |
1978 | attach_pid(p, PIDTYPE_SID); | |
909ea964 | 1979 | __this_cpu_inc(process_counts); |
80628ca0 ON |
1980 | } else { |
1981 | current->signal->nr_threads++; | |
1982 | atomic_inc(¤t->signal->live); | |
1983 | atomic_inc(¤t->signal->sigcnt); | |
80628ca0 ON |
1984 | list_add_tail_rcu(&p->thread_group, |
1985 | &p->group_leader->thread_group); | |
0c740d0a ON |
1986 | list_add_tail_rcu(&p->thread_node, |
1987 | &p->signal->thread_head); | |
73b9ebfe | 1988 | } |
81907739 | 1989 | attach_pid(p, PIDTYPE_PID); |
73b9ebfe | 1990 | nr_threads++; |
1da177e4 LT |
1991 | } |
1992 | ||
1da177e4 | 1993 | total_forks++; |
3f17da69 | 1994 | spin_unlock(¤t->sighand->siglock); |
4af4206b | 1995 | syscall_tracepoint_update(p); |
1da177e4 | 1996 | write_unlock_irq(&tasklist_lock); |
4af4206b | 1997 | |
c13cf856 | 1998 | proc_fork_connector(p); |
b53202e6 | 1999 | cgroup_post_fork(p); |
780de9dd | 2000 | cgroup_threadgroup_change_end(current); |
cdd6c482 | 2001 | perf_event_fork(p); |
43d2b113 KH |
2002 | |
2003 | trace_task_newtask(p, clone_flags); | |
3ab67966 | 2004 | uprobe_copy_process(p, clone_flags); |
43d2b113 | 2005 | |
1da177e4 LT |
2006 | return p; |
2007 | ||
7e47682e | 2008 | bad_fork_cancel_cgroup: |
3fd37226 KT |
2009 | spin_unlock(¤t->sighand->siglock); |
2010 | write_unlock_irq(&tasklist_lock); | |
b53202e6 | 2011 | cgroup_cancel_fork(p); |
425fb2b4 | 2012 | bad_fork_free_pid: |
780de9dd | 2013 | cgroup_threadgroup_change_end(current); |
425fb2b4 PE |
2014 | if (pid != &init_struct_pid) |
2015 | free_pid(pid); | |
0740aa5f JS |
2016 | bad_fork_cleanup_thread: |
2017 | exit_thread(p); | |
fd0928df | 2018 | bad_fork_cleanup_io: |
b69f2292 LR |
2019 | if (p->io_context) |
2020 | exit_io_context(p); | |
ab516013 | 2021 | bad_fork_cleanup_namespaces: |
444f378b | 2022 | exit_task_namespaces(p); |
1da177e4 | 2023 | bad_fork_cleanup_mm: |
343ef896 AA |
2024 | if (p->mm) { |
2025 | mm_clear_owner(p->mm, p); | |
1da177e4 | 2026 | mmput(p->mm); |
343ef896 | 2027 | } |
1da177e4 | 2028 | bad_fork_cleanup_signal: |
4ab6c083 | 2029 | if (!(clone_flags & CLONE_THREAD)) |
1c5354de | 2030 | free_signal_struct(p->signal); |
1da177e4 | 2031 | bad_fork_cleanup_sighand: |
a7e5328a | 2032 | __cleanup_sighand(p->sighand); |
1da177e4 LT |
2033 | bad_fork_cleanup_fs: |
2034 | exit_fs(p); /* blocking */ | |
2035 | bad_fork_cleanup_files: | |
2036 | exit_files(p); /* blocking */ | |
2037 | bad_fork_cleanup_semundo: | |
2038 | exit_sem(p); | |
e4e55b47 TH |
2039 | bad_fork_cleanup_security: |
2040 | security_task_free(p); | |
1da177e4 LT |
2041 | bad_fork_cleanup_audit: |
2042 | audit_free(p); | |
6c72e350 | 2043 | bad_fork_cleanup_perf: |
cdd6c482 | 2044 | perf_event_free_task(p); |
6c72e350 | 2045 | bad_fork_cleanup_policy: |
b09be676 | 2046 | lockdep_free_task(p); |
1da177e4 | 2047 | #ifdef CONFIG_NUMA |
f0be3d32 | 2048 | mpol_put(p->mempolicy); |
e8604cb4 | 2049 | bad_fork_cleanup_threadgroup_lock: |
1da177e4 | 2050 | #endif |
35df17c5 | 2051 | delayacct_tsk_free(p); |
1da177e4 | 2052 | bad_fork_cleanup_count: |
d84f4f99 | 2053 | atomic_dec(&p->cred->user->processes); |
e0e81739 | 2054 | exit_creds(p); |
1da177e4 | 2055 | bad_fork_free: |
405c0759 | 2056 | p->state = TASK_DEAD; |
68f24b08 | 2057 | put_task_stack(p); |
343ef896 | 2058 | delayed_free_task(p); |
fe7d37d1 ON |
2059 | fork_out: |
2060 | return ERR_PTR(retval); | |
1da177e4 LT |
2061 | } |
2062 | ||
f106eee1 ON |
2063 | static inline void init_idle_pids(struct pid_link *links) |
2064 | { | |
2065 | enum pid_type type; | |
2066 | ||
2067 | for (type = PIDTYPE_PID; type < PIDTYPE_MAX; ++type) { | |
2068 | INIT_HLIST_NODE(&links[type].node); /* not really needed */ | |
2069 | links[type].pid = &init_struct_pid; | |
2070 | } | |
2071 | } | |
2072 | ||
0db0628d | 2073 | struct task_struct *fork_idle(int cpu) |
1da177e4 | 2074 | { |
36c8b586 | 2075 | struct task_struct *task; |
725fc629 AK |
2076 | task = copy_process(CLONE_VM, 0, 0, NULL, &init_struct_pid, 0, 0, |
2077 | cpu_to_node(cpu)); | |
f106eee1 ON |
2078 | if (!IS_ERR(task)) { |
2079 | init_idle_pids(task->pids); | |
753ca4f3 | 2080 | init_idle(task, cpu); |
f106eee1 | 2081 | } |
73b9ebfe | 2082 | |
1da177e4 LT |
2083 | return task; |
2084 | } | |
2085 | ||
1da177e4 LT |
2086 | /* |
2087 | * Ok, this is the main fork-routine. | |
2088 | * | |
2089 | * It copies the process, and if successful kick-starts | |
2090 | * it and waits for it to finish using the VM if required. | |
2091 | */ | |
3033f14a | 2092 | long _do_fork(unsigned long clone_flags, |
1da177e4 | 2093 | unsigned long stack_start, |
1da177e4 LT |
2094 | unsigned long stack_size, |
2095 | int __user *parent_tidptr, | |
3033f14a JT |
2096 | int __user *child_tidptr, |
2097 | unsigned long tls) | |
1da177e4 LT |
2098 | { |
2099 | struct task_struct *p; | |
2100 | int trace = 0; | |
92476d7f | 2101 | long nr; |
1da177e4 | 2102 | |
09a05394 | 2103 | /* |
4b9d33e6 TH |
2104 | * Determine whether and which event to report to ptracer. When |
2105 | * called from kernel_thread or CLONE_UNTRACED is explicitly | |
2106 | * requested, no event is reported; otherwise, report if the event | |
2107 | * for the type of forking is enabled. | |
09a05394 | 2108 | */ |
e80d6661 | 2109 | if (!(clone_flags & CLONE_UNTRACED)) { |
4b9d33e6 TH |
2110 | if (clone_flags & CLONE_VFORK) |
2111 | trace = PTRACE_EVENT_VFORK; | |
2112 | else if ((clone_flags & CSIGNAL) != SIGCHLD) | |
2113 | trace = PTRACE_EVENT_CLONE; | |
2114 | else | |
2115 | trace = PTRACE_EVENT_FORK; | |
2116 | ||
2117 | if (likely(!ptrace_event_enabled(current, trace))) | |
2118 | trace = 0; | |
2119 | } | |
1da177e4 | 2120 | |
62e791c1 | 2121 | p = copy_process(clone_flags, stack_start, stack_size, |
725fc629 | 2122 | child_tidptr, NULL, trace, tls, NUMA_NO_NODE); |
38addce8 | 2123 | add_latent_entropy(); |
1da177e4 LT |
2124 | /* |
2125 | * Do this prior waking up the new thread - the thread pointer | |
2126 | * might get invalid after that point, if the thread exits quickly. | |
2127 | */ | |
2128 | if (!IS_ERR(p)) { | |
2129 | struct completion vfork; | |
4e52365f | 2130 | struct pid *pid; |
1da177e4 | 2131 | |
0a16b607 MD |
2132 | trace_sched_process_fork(current, p); |
2133 | ||
4e52365f MD |
2134 | pid = get_task_pid(p, PIDTYPE_PID); |
2135 | nr = pid_vnr(pid); | |
30e49c26 PE |
2136 | |
2137 | if (clone_flags & CLONE_PARENT_SETTID) | |
2138 | put_user(nr, parent_tidptr); | |
a6f5e063 | 2139 | |
1da177e4 LT |
2140 | if (clone_flags & CLONE_VFORK) { |
2141 | p->vfork_done = &vfork; | |
2142 | init_completion(&vfork); | |
d68b46fe | 2143 | get_task_struct(p); |
1da177e4 LT |
2144 | } |
2145 | ||
3e51e3ed | 2146 | wake_up_new_task(p); |
1da177e4 | 2147 | |
4b9d33e6 TH |
2148 | /* forking complete and child started to run, tell ptracer */ |
2149 | if (unlikely(trace)) | |
4e52365f | 2150 | ptrace_event_pid(trace, pid); |
09a05394 | 2151 | |
1da177e4 | 2152 | if (clone_flags & CLONE_VFORK) { |
d68b46fe | 2153 | if (!wait_for_vfork_done(p, &vfork)) |
4e52365f | 2154 | ptrace_event_pid(PTRACE_EVENT_VFORK_DONE, pid); |
1da177e4 | 2155 | } |
4e52365f MD |
2156 | |
2157 | put_pid(pid); | |
1da177e4 | 2158 | } else { |
92476d7f | 2159 | nr = PTR_ERR(p); |
1da177e4 | 2160 | } |
92476d7f | 2161 | return nr; |
1da177e4 LT |
2162 | } |
2163 | ||
3033f14a JT |
2164 | #ifndef CONFIG_HAVE_COPY_THREAD_TLS |
2165 | /* For compatibility with architectures that call do_fork directly rather than | |
2166 | * using the syscall entry points below. */ | |
2167 | long do_fork(unsigned long clone_flags, | |
2168 | unsigned long stack_start, | |
2169 | unsigned long stack_size, | |
2170 | int __user *parent_tidptr, | |
2171 | int __user *child_tidptr) | |
2172 | { | |
2173 | return _do_fork(clone_flags, stack_start, stack_size, | |
2174 | parent_tidptr, child_tidptr, 0); | |
2175 | } | |
2176 | #endif | |
2177 | ||
2aa3a7f8 AV |
2178 | /* |
2179 | * Create a kernel thread. | |
2180 | */ | |
2181 | pid_t kernel_thread(int (*fn)(void *), void *arg, unsigned long flags) | |
2182 | { | |
3033f14a JT |
2183 | return _do_fork(flags|CLONE_VM|CLONE_UNTRACED, (unsigned long)fn, |
2184 | (unsigned long)arg, NULL, NULL, 0); | |
2aa3a7f8 | 2185 | } |
2aa3a7f8 | 2186 | |
d2125043 AV |
2187 | #ifdef __ARCH_WANT_SYS_FORK |
2188 | SYSCALL_DEFINE0(fork) | |
2189 | { | |
2190 | #ifdef CONFIG_MMU | |
3033f14a | 2191 | return _do_fork(SIGCHLD, 0, 0, NULL, NULL, 0); |
d2125043 AV |
2192 | #else |
2193 | /* can not support in nommu mode */ | |
5d59e182 | 2194 | return -EINVAL; |
d2125043 AV |
2195 | #endif |
2196 | } | |
2197 | #endif | |
2198 | ||
2199 | #ifdef __ARCH_WANT_SYS_VFORK | |
2200 | SYSCALL_DEFINE0(vfork) | |
2201 | { | |
3033f14a JT |
2202 | return _do_fork(CLONE_VFORK | CLONE_VM | SIGCHLD, 0, |
2203 | 0, NULL, NULL, 0); | |
d2125043 AV |
2204 | } |
2205 | #endif | |
2206 | ||
2207 | #ifdef __ARCH_WANT_SYS_CLONE | |
2208 | #ifdef CONFIG_CLONE_BACKWARDS | |
2209 | SYSCALL_DEFINE5(clone, unsigned long, clone_flags, unsigned long, newsp, | |
2210 | int __user *, parent_tidptr, | |
3033f14a | 2211 | unsigned long, tls, |
d2125043 AV |
2212 | int __user *, child_tidptr) |
2213 | #elif defined(CONFIG_CLONE_BACKWARDS2) | |
2214 | SYSCALL_DEFINE5(clone, unsigned long, newsp, unsigned long, clone_flags, | |
2215 | int __user *, parent_tidptr, | |
2216 | int __user *, child_tidptr, | |
3033f14a | 2217 | unsigned long, tls) |
dfa9771a MS |
2218 | #elif defined(CONFIG_CLONE_BACKWARDS3) |
2219 | SYSCALL_DEFINE6(clone, unsigned long, clone_flags, unsigned long, newsp, | |
2220 | int, stack_size, | |
2221 | int __user *, parent_tidptr, | |
2222 | int __user *, child_tidptr, | |
3033f14a | 2223 | unsigned long, tls) |
d2125043 AV |
2224 | #else |
2225 | SYSCALL_DEFINE5(clone, unsigned long, clone_flags, unsigned long, newsp, | |
2226 | int __user *, parent_tidptr, | |
2227 | int __user *, child_tidptr, | |
3033f14a | 2228 | unsigned long, tls) |
d2125043 AV |
2229 | #endif |
2230 | { | |
3033f14a | 2231 | return _do_fork(clone_flags, newsp, 0, parent_tidptr, child_tidptr, tls); |
d2125043 AV |
2232 | } |
2233 | #endif | |
2234 | ||
0f1b92cb ON |
2235 | void walk_process_tree(struct task_struct *top, proc_visitor visitor, void *data) |
2236 | { | |
2237 | struct task_struct *leader, *parent, *child; | |
2238 | int res; | |
2239 | ||
2240 | read_lock(&tasklist_lock); | |
2241 | leader = top = top->group_leader; | |
2242 | down: | |
2243 | for_each_thread(leader, parent) { | |
2244 | list_for_each_entry(child, &parent->children, sibling) { | |
2245 | res = visitor(child, data); | |
2246 | if (res) { | |
2247 | if (res < 0) | |
2248 | goto out; | |
2249 | leader = child; | |
2250 | goto down; | |
2251 | } | |
2252 | up: | |
2253 | ; | |
2254 | } | |
2255 | } | |
2256 | ||
2257 | if (leader != top) { | |
2258 | child = leader; | |
2259 | parent = child->real_parent; | |
2260 | leader = parent->group_leader; | |
2261 | goto up; | |
2262 | } | |
2263 | out: | |
2264 | read_unlock(&tasklist_lock); | |
2265 | } | |
2266 | ||
5fd63b30 RT |
2267 | #ifndef ARCH_MIN_MMSTRUCT_ALIGN |
2268 | #define ARCH_MIN_MMSTRUCT_ALIGN 0 | |
2269 | #endif | |
2270 | ||
51cc5068 | 2271 | static void sighand_ctor(void *data) |
aa1757f9 ON |
2272 | { |
2273 | struct sighand_struct *sighand = data; | |
2274 | ||
a35afb83 | 2275 | spin_lock_init(&sighand->siglock); |
b8fceee1 | 2276 | init_waitqueue_head(&sighand->signalfd_wqh); |
aa1757f9 ON |
2277 | } |
2278 | ||
1da177e4 LT |
2279 | void __init proc_caches_init(void) |
2280 | { | |
2281 | sighand_cachep = kmem_cache_create("sighand_cache", | |
2282 | sizeof(struct sighand_struct), 0, | |
5f0d5a3a | 2283 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_TYPESAFE_BY_RCU| |
75f296d9 | 2284 | SLAB_ACCOUNT, sighand_ctor); |
1da177e4 LT |
2285 | signal_cachep = kmem_cache_create("signal_cache", |
2286 | sizeof(struct signal_struct), 0, | |
75f296d9 | 2287 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_ACCOUNT, |
5d097056 | 2288 | NULL); |
20c2df83 | 2289 | files_cachep = kmem_cache_create("files_cache", |
1da177e4 | 2290 | sizeof(struct files_struct), 0, |
75f296d9 | 2291 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_ACCOUNT, |
5d097056 | 2292 | NULL); |
20c2df83 | 2293 | fs_cachep = kmem_cache_create("fs_cache", |
1da177e4 | 2294 | sizeof(struct fs_struct), 0, |
75f296d9 | 2295 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_ACCOUNT, |
5d097056 | 2296 | NULL); |
6345d24d LT |
2297 | /* |
2298 | * FIXME! The "sizeof(struct mm_struct)" currently includes the | |
2299 | * whole struct cpumask for the OFFSTACK case. We could change | |
2300 | * this to *only* allocate as much of it as required by the | |
2301 | * maximum number of CPU's we can ever have. The cpumask_allocation | |
2302 | * is at the end of the structure, exactly for that reason. | |
2303 | */ | |
1da177e4 | 2304 | mm_cachep = kmem_cache_create("mm_struct", |
5fd63b30 | 2305 | sizeof(struct mm_struct), ARCH_MIN_MMSTRUCT_ALIGN, |
75f296d9 | 2306 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_ACCOUNT, |
5d097056 VD |
2307 | NULL); |
2308 | vm_area_cachep = KMEM_CACHE(vm_area_struct, SLAB_PANIC|SLAB_ACCOUNT); | |
8feae131 | 2309 | mmap_init(); |
66577193 | 2310 | nsproxy_cache_init(); |
1da177e4 | 2311 | } |
cf2e340f | 2312 | |
cf2e340f | 2313 | /* |
9bfb23fc | 2314 | * Check constraints on flags passed to the unshare system call. |
cf2e340f | 2315 | */ |
9bfb23fc | 2316 | static int check_unshare_flags(unsigned long unshare_flags) |
cf2e340f | 2317 | { |
9bfb23fc ON |
2318 | if (unshare_flags & ~(CLONE_THREAD|CLONE_FS|CLONE_NEWNS|CLONE_SIGHAND| |
2319 | CLONE_VM|CLONE_FILES|CLONE_SYSVSEM| | |
50804fe3 | 2320 | CLONE_NEWUTS|CLONE_NEWIPC|CLONE_NEWNET| |
a79a908f | 2321 | CLONE_NEWUSER|CLONE_NEWPID|CLONE_NEWCGROUP)) |
9bfb23fc | 2322 | return -EINVAL; |
cf2e340f | 2323 | /* |
12c641ab EB |
2324 | * Not implemented, but pretend it works if there is nothing |
2325 | * to unshare. Note that unsharing the address space or the | |
2326 | * signal handlers also need to unshare the signal queues (aka | |
2327 | * CLONE_THREAD). | |
cf2e340f | 2328 | */ |
9bfb23fc | 2329 | if (unshare_flags & (CLONE_THREAD | CLONE_SIGHAND | CLONE_VM)) { |
12c641ab EB |
2330 | if (!thread_group_empty(current)) |
2331 | return -EINVAL; | |
2332 | } | |
2333 | if (unshare_flags & (CLONE_SIGHAND | CLONE_VM)) { | |
2334 | if (atomic_read(¤t->sighand->count) > 1) | |
2335 | return -EINVAL; | |
2336 | } | |
2337 | if (unshare_flags & CLONE_VM) { | |
2338 | if (!current_is_single_threaded()) | |
9bfb23fc ON |
2339 | return -EINVAL; |
2340 | } | |
cf2e340f JD |
2341 | |
2342 | return 0; | |
2343 | } | |
2344 | ||
2345 | /* | |
99d1419d | 2346 | * Unshare the filesystem structure if it is being shared |
cf2e340f JD |
2347 | */ |
2348 | static int unshare_fs(unsigned long unshare_flags, struct fs_struct **new_fsp) | |
2349 | { | |
2350 | struct fs_struct *fs = current->fs; | |
2351 | ||
498052bb AV |
2352 | if (!(unshare_flags & CLONE_FS) || !fs) |
2353 | return 0; | |
2354 | ||
2355 | /* don't need lock here; in the worst case we'll do useless copy */ | |
2356 | if (fs->users == 1) | |
2357 | return 0; | |
2358 | ||
2359 | *new_fsp = copy_fs_struct(fs); | |
2360 | if (!*new_fsp) | |
2361 | return -ENOMEM; | |
cf2e340f JD |
2362 | |
2363 | return 0; | |
2364 | } | |
2365 | ||
cf2e340f | 2366 | /* |
a016f338 | 2367 | * Unshare file descriptor table if it is being shared |
cf2e340f JD |
2368 | */ |
2369 | static int unshare_fd(unsigned long unshare_flags, struct files_struct **new_fdp) | |
2370 | { | |
2371 | struct files_struct *fd = current->files; | |
a016f338 | 2372 | int error = 0; |
cf2e340f JD |
2373 | |
2374 | if ((unshare_flags & CLONE_FILES) && | |
a016f338 JD |
2375 | (fd && atomic_read(&fd->count) > 1)) { |
2376 | *new_fdp = dup_fd(fd, &error); | |
2377 | if (!*new_fdp) | |
2378 | return error; | |
2379 | } | |
cf2e340f JD |
2380 | |
2381 | return 0; | |
2382 | } | |
2383 | ||
cf2e340f JD |
2384 | /* |
2385 | * unshare allows a process to 'unshare' part of the process | |
2386 | * context which was originally shared using clone. copy_* | |
2387 | * functions used by do_fork() cannot be used here directly | |
2388 | * because they modify an inactive task_struct that is being | |
2389 | * constructed. Here we are modifying the current, active, | |
2390 | * task_struct. | |
2391 | */ | |
6559eed8 | 2392 | SYSCALL_DEFINE1(unshare, unsigned long, unshare_flags) |
cf2e340f | 2393 | { |
cf2e340f | 2394 | struct fs_struct *fs, *new_fs = NULL; |
cf2e340f | 2395 | struct files_struct *fd, *new_fd = NULL; |
b2e0d987 | 2396 | struct cred *new_cred = NULL; |
cf7b708c | 2397 | struct nsproxy *new_nsproxy = NULL; |
9edff4ab | 2398 | int do_sysvsem = 0; |
9bfb23fc | 2399 | int err; |
cf2e340f | 2400 | |
b2e0d987 | 2401 | /* |
faf00da5 EB |
2402 | * If unsharing a user namespace must also unshare the thread group |
2403 | * and unshare the filesystem root and working directories. | |
b2e0d987 EB |
2404 | */ |
2405 | if (unshare_flags & CLONE_NEWUSER) | |
e66eded8 | 2406 | unshare_flags |= CLONE_THREAD | CLONE_FS; |
50804fe3 EB |
2407 | /* |
2408 | * If unsharing vm, must also unshare signal handlers. | |
2409 | */ | |
2410 | if (unshare_flags & CLONE_VM) | |
2411 | unshare_flags |= CLONE_SIGHAND; | |
12c641ab EB |
2412 | /* |
2413 | * If unsharing a signal handlers, must also unshare the signal queues. | |
2414 | */ | |
2415 | if (unshare_flags & CLONE_SIGHAND) | |
2416 | unshare_flags |= CLONE_THREAD; | |
9bfb23fc ON |
2417 | /* |
2418 | * If unsharing namespace, must also unshare filesystem information. | |
2419 | */ | |
2420 | if (unshare_flags & CLONE_NEWNS) | |
2421 | unshare_flags |= CLONE_FS; | |
50804fe3 | 2422 | |
680f7875 SH |
2423 | if ((unshare_flags & CLONE_NEWUSER) && !unprivileged_userns_clone) { |
2424 | err = -EPERM; | |
2425 | if (!capable(CAP_SYS_ADMIN)) | |
2426 | goto bad_unshare_out; | |
2427 | } | |
2428 | ||
50804fe3 EB |
2429 | err = check_unshare_flags(unshare_flags); |
2430 | if (err) | |
2431 | goto bad_unshare_out; | |
6013f67f MS |
2432 | /* |
2433 | * CLONE_NEWIPC must also detach from the undolist: after switching | |
2434 | * to a new ipc namespace, the semaphore arrays from the old | |
2435 | * namespace are unreachable. | |
2436 | */ | |
2437 | if (unshare_flags & (CLONE_NEWIPC|CLONE_SYSVSEM)) | |
9edff4ab | 2438 | do_sysvsem = 1; |
fb0a685c DRO |
2439 | err = unshare_fs(unshare_flags, &new_fs); |
2440 | if (err) | |
9bfb23fc | 2441 | goto bad_unshare_out; |
fb0a685c DRO |
2442 | err = unshare_fd(unshare_flags, &new_fd); |
2443 | if (err) | |
9bfb23fc | 2444 | goto bad_unshare_cleanup_fs; |
b2e0d987 | 2445 | err = unshare_userns(unshare_flags, &new_cred); |
fb0a685c | 2446 | if (err) |
9edff4ab | 2447 | goto bad_unshare_cleanup_fd; |
b2e0d987 EB |
2448 | err = unshare_nsproxy_namespaces(unshare_flags, &new_nsproxy, |
2449 | new_cred, new_fs); | |
2450 | if (err) | |
2451 | goto bad_unshare_cleanup_cred; | |
c0b2fc31 | 2452 | |
b2e0d987 | 2453 | if (new_fs || new_fd || do_sysvsem || new_cred || new_nsproxy) { |
9edff4ab MS |
2454 | if (do_sysvsem) { |
2455 | /* | |
2456 | * CLONE_SYSVSEM is equivalent to sys_exit(). | |
2457 | */ | |
2458 | exit_sem(current); | |
2459 | } | |
ab602f79 JM |
2460 | if (unshare_flags & CLONE_NEWIPC) { |
2461 | /* Orphan segments in old ns (see sem above). */ | |
2462 | exit_shm(current); | |
2463 | shm_init_task(current); | |
2464 | } | |
ab516013 | 2465 | |
6f977e6b | 2466 | if (new_nsproxy) |
cf7b708c | 2467 | switch_task_namespaces(current, new_nsproxy); |
cf2e340f | 2468 | |
cf7b708c PE |
2469 | task_lock(current); |
2470 | ||
cf2e340f JD |
2471 | if (new_fs) { |
2472 | fs = current->fs; | |
2a4419b5 | 2473 | spin_lock(&fs->lock); |
cf2e340f | 2474 | current->fs = new_fs; |
498052bb AV |
2475 | if (--fs->users) |
2476 | new_fs = NULL; | |
2477 | else | |
2478 | new_fs = fs; | |
2a4419b5 | 2479 | spin_unlock(&fs->lock); |
cf2e340f JD |
2480 | } |
2481 | ||
cf2e340f JD |
2482 | if (new_fd) { |
2483 | fd = current->files; | |
2484 | current->files = new_fd; | |
2485 | new_fd = fd; | |
2486 | } | |
2487 | ||
2488 | task_unlock(current); | |
b2e0d987 EB |
2489 | |
2490 | if (new_cred) { | |
2491 | /* Install the new user namespace */ | |
2492 | commit_creds(new_cred); | |
2493 | new_cred = NULL; | |
2494 | } | |
cf2e340f JD |
2495 | } |
2496 | ||
e4222673 HB |
2497 | perf_event_namespaces(current); |
2498 | ||
b2e0d987 EB |
2499 | bad_unshare_cleanup_cred: |
2500 | if (new_cred) | |
2501 | put_cred(new_cred); | |
cf2e340f JD |
2502 | bad_unshare_cleanup_fd: |
2503 | if (new_fd) | |
2504 | put_files_struct(new_fd); | |
2505 | ||
cf2e340f JD |
2506 | bad_unshare_cleanup_fs: |
2507 | if (new_fs) | |
498052bb | 2508 | free_fs_struct(new_fs); |
cf2e340f | 2509 | |
cf2e340f JD |
2510 | bad_unshare_out: |
2511 | return err; | |
2512 | } | |
3b125388 AV |
2513 | |
2514 | /* | |
2515 | * Helper to unshare the files of the current task. | |
2516 | * We don't want to expose copy_files internals to | |
2517 | * the exec layer of the kernel. | |
2518 | */ | |
2519 | ||
2520 | int unshare_files(struct files_struct **displaced) | |
2521 | { | |
2522 | struct task_struct *task = current; | |
50704516 | 2523 | struct files_struct *copy = NULL; |
3b125388 AV |
2524 | int error; |
2525 | ||
2526 | error = unshare_fd(CLONE_FILES, ©); | |
2527 | if (error || !copy) { | |
2528 | *displaced = NULL; | |
2529 | return error; | |
2530 | } | |
2531 | *displaced = task->files; | |
2532 | task_lock(task); | |
2533 | task->files = copy; | |
2534 | task_unlock(task); | |
2535 | return 0; | |
2536 | } | |
16db3d3f HS |
2537 | |
2538 | int sysctl_max_threads(struct ctl_table *table, int write, | |
2539 | void __user *buffer, size_t *lenp, loff_t *ppos) | |
2540 | { | |
2541 | struct ctl_table t; | |
2542 | int ret; | |
2543 | int threads = max_threads; | |
eb3794c5 | 2544 | int min = 1; |
16db3d3f HS |
2545 | int max = MAX_THREADS; |
2546 | ||
2547 | t = *table; | |
2548 | t.data = &threads; | |
2549 | t.extra1 = &min; | |
2550 | t.extra2 = &max; | |
2551 | ||
2552 | ret = proc_dointvec_minmax(&t, write, buffer, lenp, ppos); | |
2553 | if (ret || !write) | |
2554 | return ret; | |
2555 | ||
eb3794c5 | 2556 | max_threads = threads; |
16db3d3f HS |
2557 | |
2558 | return 0; | |
2559 | } |