]>
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 LT |
14 | #include <linux/slab.h> |
15 | #include <linux/init.h> | |
16 | #include <linux/unistd.h> | |
1da177e4 LT |
17 | #include <linux/module.h> |
18 | #include <linux/vmalloc.h> | |
19 | #include <linux/completion.h> | |
1da177e4 LT |
20 | #include <linux/personality.h> |
21 | #include <linux/mempolicy.h> | |
22 | #include <linux/sem.h> | |
23 | #include <linux/file.h> | |
9f3acc31 | 24 | #include <linux/fdtable.h> |
da9cbc87 | 25 | #include <linux/iocontext.h> |
1da177e4 LT |
26 | #include <linux/key.h> |
27 | #include <linux/binfmts.h> | |
28 | #include <linux/mman.h> | |
cddb8a5c | 29 | #include <linux/mmu_notifier.h> |
1da177e4 | 30 | #include <linux/fs.h> |
615d6e87 DB |
31 | #include <linux/mm.h> |
32 | #include <linux/vmacache.h> | |
ab516013 | 33 | #include <linux/nsproxy.h> |
c59ede7b | 34 | #include <linux/capability.h> |
1da177e4 | 35 | #include <linux/cpu.h> |
b4f48b63 | 36 | #include <linux/cgroup.h> |
1da177e4 | 37 | #include <linux/security.h> |
a1e78772 | 38 | #include <linux/hugetlb.h> |
e2cfabdf | 39 | #include <linux/seccomp.h> |
1da177e4 LT |
40 | #include <linux/swap.h> |
41 | #include <linux/syscalls.h> | |
42 | #include <linux/jiffies.h> | |
43 | #include <linux/futex.h> | |
8141c7f3 | 44 | #include <linux/compat.h> |
207205a2 | 45 | #include <linux/kthread.h> |
7c3ab738 | 46 | #include <linux/task_io_accounting_ops.h> |
ab2af1f5 | 47 | #include <linux/rcupdate.h> |
1da177e4 LT |
48 | #include <linux/ptrace.h> |
49 | #include <linux/mount.h> | |
50 | #include <linux/audit.h> | |
78fb7466 | 51 | #include <linux/memcontrol.h> |
f201ae23 | 52 | #include <linux/ftrace.h> |
5e2bf014 | 53 | #include <linux/proc_fs.h> |
1da177e4 LT |
54 | #include <linux/profile.h> |
55 | #include <linux/rmap.h> | |
f8af4da3 | 56 | #include <linux/ksm.h> |
1da177e4 | 57 | #include <linux/acct.h> |
8f0ab514 | 58 | #include <linux/tsacct_kern.h> |
9f46080c | 59 | #include <linux/cn_proc.h> |
ba96a0c8 | 60 | #include <linux/freezer.h> |
ca74e92b | 61 | #include <linux/delayacct.h> |
ad4ecbcb | 62 | #include <linux/taskstats_kern.h> |
0a425405 | 63 | #include <linux/random.h> |
522ed776 | 64 | #include <linux/tty.h> |
fd0928df | 65 | #include <linux/blkdev.h> |
5ad4e53b | 66 | #include <linux/fs_struct.h> |
7c9f8861 | 67 | #include <linux/magic.h> |
cdd6c482 | 68 | #include <linux/perf_event.h> |
42c4ab41 | 69 | #include <linux/posix-timers.h> |
8e7cac79 | 70 | #include <linux/user-return-notifier.h> |
3d5992d2 | 71 | #include <linux/oom.h> |
ba76149f | 72 | #include <linux/khugepaged.h> |
d80e731e | 73 | #include <linux/signalfd.h> |
0326f5a9 | 74 | #include <linux/uprobes.h> |
a27bb332 | 75 | #include <linux/aio.h> |
52f5684c | 76 | #include <linux/compiler.h> |
16db3d3f | 77 | #include <linux/sysctl.h> |
5c9a8750 | 78 | #include <linux/kcov.h> |
1da177e4 LT |
79 | |
80 | #include <asm/pgtable.h> | |
81 | #include <asm/pgalloc.h> | |
82 | #include <asm/uaccess.h> | |
83 | #include <asm/mmu_context.h> | |
84 | #include <asm/cacheflush.h> | |
85 | #include <asm/tlbflush.h> | |
86 | ||
ad8d75ff SR |
87 | #include <trace/events/sched.h> |
88 | ||
43d2b113 KH |
89 | #define CREATE_TRACE_POINTS |
90 | #include <trace/events/task.h> | |
91 | ||
ac1b398d HS |
92 | /* |
93 | * Minimum number of threads to boot the kernel | |
94 | */ | |
95 | #define MIN_THREADS 20 | |
96 | ||
97 | /* | |
98 | * Maximum number of threads | |
99 | */ | |
100 | #define MAX_THREADS FUTEX_TID_MASK | |
101 | ||
1da177e4 LT |
102 | /* |
103 | * Protected counters by write_lock_irq(&tasklist_lock) | |
104 | */ | |
105 | unsigned long total_forks; /* Handle normal Linux uptimes. */ | |
fb0a685c | 106 | int nr_threads; /* The idle threads do not count.. */ |
1da177e4 LT |
107 | |
108 | int max_threads; /* tunable limit on nr_threads */ | |
109 | ||
110 | DEFINE_PER_CPU(unsigned long, process_counts) = 0; | |
111 | ||
c59923a1 | 112 | __cacheline_aligned DEFINE_RWLOCK(tasklist_lock); /* outer */ |
db1466b3 PM |
113 | |
114 | #ifdef CONFIG_PROVE_RCU | |
115 | int lockdep_tasklist_lock_is_held(void) | |
116 | { | |
117 | return lockdep_is_held(&tasklist_lock); | |
118 | } | |
119 | EXPORT_SYMBOL_GPL(lockdep_tasklist_lock_is_held); | |
120 | #endif /* #ifdef CONFIG_PROVE_RCU */ | |
1da177e4 LT |
121 | |
122 | int nr_processes(void) | |
123 | { | |
124 | int cpu; | |
125 | int total = 0; | |
126 | ||
1d510750 | 127 | for_each_possible_cpu(cpu) |
1da177e4 LT |
128 | total += per_cpu(process_counts, cpu); |
129 | ||
130 | return total; | |
131 | } | |
132 | ||
f19b9f74 AM |
133 | void __weak arch_release_task_struct(struct task_struct *tsk) |
134 | { | |
135 | } | |
136 | ||
f5e10287 | 137 | #ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR |
e18b890b | 138 | static struct kmem_cache *task_struct_cachep; |
41101809 TG |
139 | |
140 | static inline struct task_struct *alloc_task_struct_node(int node) | |
141 | { | |
142 | return kmem_cache_alloc_node(task_struct_cachep, GFP_KERNEL, node); | |
143 | } | |
144 | ||
41101809 TG |
145 | static inline void free_task_struct(struct task_struct *tsk) |
146 | { | |
41101809 TG |
147 | kmem_cache_free(task_struct_cachep, tsk); |
148 | } | |
1da177e4 LT |
149 | #endif |
150 | ||
b235beea | 151 | void __weak arch_release_thread_stack(unsigned long *stack) |
f19b9f74 AM |
152 | { |
153 | } | |
154 | ||
b235beea | 155 | #ifndef CONFIG_ARCH_THREAD_STACK_ALLOCATOR |
41101809 | 156 | |
0d15d74a TG |
157 | /* |
158 | * Allocate pages if THREAD_SIZE is >= PAGE_SIZE, otherwise use a | |
159 | * kmemcache based allocator. | |
160 | */ | |
ba14a194 | 161 | # if THREAD_SIZE >= PAGE_SIZE || defined(CONFIG_VMAP_STACK) |
ac496bf4 AL |
162 | |
163 | #ifdef CONFIG_VMAP_STACK | |
164 | /* | |
165 | * vmalloc() is a bit slow, and calling vfree() enough times will force a TLB | |
166 | * flush. Try to minimize the number of calls by caching stacks. | |
167 | */ | |
168 | #define NR_CACHED_STACKS 2 | |
169 | static DEFINE_PER_CPU(struct vm_struct *, cached_stacks[NR_CACHED_STACKS]); | |
170 | #endif | |
171 | ||
ba14a194 | 172 | static unsigned long *alloc_thread_stack_node(struct task_struct *tsk, int node) |
b69c49b7 | 173 | { |
ba14a194 | 174 | #ifdef CONFIG_VMAP_STACK |
ac496bf4 AL |
175 | void *stack; |
176 | int i; | |
177 | ||
178 | local_irq_disable(); | |
179 | for (i = 0; i < NR_CACHED_STACKS; i++) { | |
180 | struct vm_struct *s = this_cpu_read(cached_stacks[i]); | |
181 | ||
182 | if (!s) | |
183 | continue; | |
184 | this_cpu_write(cached_stacks[i], NULL); | |
185 | ||
186 | tsk->stack_vm_area = s; | |
187 | local_irq_enable(); | |
188 | return s->addr; | |
189 | } | |
190 | local_irq_enable(); | |
191 | ||
192 | stack = __vmalloc_node_range(THREAD_SIZE, THREAD_SIZE, | |
193 | VMALLOC_START, VMALLOC_END, | |
194 | THREADINFO_GFP | __GFP_HIGHMEM, | |
195 | PAGE_KERNEL, | |
196 | 0, node, __builtin_return_address(0)); | |
ba14a194 AL |
197 | |
198 | /* | |
199 | * We can't call find_vm_area() in interrupt context, and | |
200 | * free_thread_stack() can be called in interrupt context, | |
201 | * so cache the vm_struct. | |
202 | */ | |
203 | if (stack) | |
204 | tsk->stack_vm_area = find_vm_area(stack); | |
205 | return stack; | |
206 | #else | |
4949148a VD |
207 | struct page *page = alloc_pages_node(node, THREADINFO_GFP, |
208 | THREAD_SIZE_ORDER); | |
b6a84016 ED |
209 | |
210 | return page ? page_address(page) : NULL; | |
ba14a194 | 211 | #endif |
b69c49b7 FT |
212 | } |
213 | ||
ba14a194 | 214 | static inline void free_thread_stack(struct task_struct *tsk) |
b69c49b7 | 215 | { |
ac496bf4 AL |
216 | #ifdef CONFIG_VMAP_STACK |
217 | if (task_stack_vm_area(tsk)) { | |
218 | unsigned long flags; | |
219 | int i; | |
220 | ||
221 | local_irq_save(flags); | |
222 | for (i = 0; i < NR_CACHED_STACKS; i++) { | |
223 | if (this_cpu_read(cached_stacks[i])) | |
224 | continue; | |
225 | ||
226 | this_cpu_write(cached_stacks[i], tsk->stack_vm_area); | |
227 | local_irq_restore(flags); | |
228 | return; | |
229 | } | |
230 | local_irq_restore(flags); | |
231 | ||
ba14a194 | 232 | vfree(tsk->stack); |
ac496bf4 AL |
233 | return; |
234 | } | |
235 | #endif | |
236 | ||
237 | __free_pages(virt_to_page(tsk->stack), THREAD_SIZE_ORDER); | |
b69c49b7 | 238 | } |
0d15d74a | 239 | # else |
b235beea | 240 | static struct kmem_cache *thread_stack_cache; |
0d15d74a | 241 | |
9521d399 | 242 | static unsigned long *alloc_thread_stack_node(struct task_struct *tsk, |
0d15d74a TG |
243 | int node) |
244 | { | |
b235beea | 245 | return kmem_cache_alloc_node(thread_stack_cache, THREADINFO_GFP, node); |
0d15d74a TG |
246 | } |
247 | ||
ba14a194 | 248 | static void free_thread_stack(struct task_struct *tsk) |
0d15d74a | 249 | { |
ba14a194 | 250 | kmem_cache_free(thread_stack_cache, tsk->stack); |
0d15d74a TG |
251 | } |
252 | ||
b235beea | 253 | void thread_stack_cache_init(void) |
0d15d74a | 254 | { |
b235beea | 255 | thread_stack_cache = kmem_cache_create("thread_stack", THREAD_SIZE, |
0d15d74a | 256 | THREAD_SIZE, 0, NULL); |
b235beea | 257 | BUG_ON(thread_stack_cache == NULL); |
0d15d74a TG |
258 | } |
259 | # endif | |
b69c49b7 FT |
260 | #endif |
261 | ||
1da177e4 | 262 | /* SLAB cache for signal_struct structures (tsk->signal) */ |
e18b890b | 263 | static struct kmem_cache *signal_cachep; |
1da177e4 LT |
264 | |
265 | /* SLAB cache for sighand_struct structures (tsk->sighand) */ | |
e18b890b | 266 | struct kmem_cache *sighand_cachep; |
1da177e4 LT |
267 | |
268 | /* SLAB cache for files_struct structures (tsk->files) */ | |
e18b890b | 269 | struct kmem_cache *files_cachep; |
1da177e4 LT |
270 | |
271 | /* SLAB cache for fs_struct structures (tsk->fs) */ | |
e18b890b | 272 | struct kmem_cache *fs_cachep; |
1da177e4 LT |
273 | |
274 | /* SLAB cache for vm_area_struct structures */ | |
e18b890b | 275 | struct kmem_cache *vm_area_cachep; |
1da177e4 LT |
276 | |
277 | /* SLAB cache for mm_struct structures (tsk->mm) */ | |
e18b890b | 278 | static struct kmem_cache *mm_cachep; |
1da177e4 | 279 | |
ba14a194 | 280 | static void account_kernel_stack(struct task_struct *tsk, int account) |
c6a7f572 | 281 | { |
ba14a194 AL |
282 | void *stack = task_stack_page(tsk); |
283 | struct vm_struct *vm = task_stack_vm_area(tsk); | |
284 | ||
285 | BUILD_BUG_ON(IS_ENABLED(CONFIG_VMAP_STACK) && PAGE_SIZE % 1024 != 0); | |
286 | ||
287 | if (vm) { | |
288 | int i; | |
289 | ||
290 | BUG_ON(vm->nr_pages != THREAD_SIZE / PAGE_SIZE); | |
291 | ||
292 | for (i = 0; i < THREAD_SIZE / PAGE_SIZE; i++) { | |
293 | mod_zone_page_state(page_zone(vm->pages[i]), | |
294 | NR_KERNEL_STACK_KB, | |
295 | PAGE_SIZE / 1024 * account); | |
296 | } | |
297 | ||
298 | /* All stack pages belong to the same memcg. */ | |
299 | memcg_kmem_update_page_stat(vm->pages[0], MEMCG_KERNEL_STACK_KB, | |
300 | account * (THREAD_SIZE / 1024)); | |
301 | } else { | |
302 | /* | |
303 | * All stack pages are in the same zone and belong to the | |
304 | * same memcg. | |
305 | */ | |
306 | struct page *first_page = virt_to_page(stack); | |
307 | ||
308 | mod_zone_page_state(page_zone(first_page), NR_KERNEL_STACK_KB, | |
309 | THREAD_SIZE / 1024 * account); | |
310 | ||
311 | memcg_kmem_update_page_stat(first_page, MEMCG_KERNEL_STACK_KB, | |
312 | account * (THREAD_SIZE / 1024)); | |
313 | } | |
c6a7f572 KM |
314 | } |
315 | ||
68f24b08 | 316 | static void release_task_stack(struct task_struct *tsk) |
1da177e4 | 317 | { |
ba14a194 | 318 | account_kernel_stack(tsk, -1); |
b235beea | 319 | arch_release_thread_stack(tsk->stack); |
ba14a194 | 320 | free_thread_stack(tsk); |
68f24b08 AL |
321 | tsk->stack = NULL; |
322 | #ifdef CONFIG_VMAP_STACK | |
323 | tsk->stack_vm_area = NULL; | |
324 | #endif | |
325 | } | |
326 | ||
327 | #ifdef CONFIG_THREAD_INFO_IN_TASK | |
328 | void put_task_stack(struct task_struct *tsk) | |
329 | { | |
330 | if (atomic_dec_and_test(&tsk->stack_refcount)) | |
331 | release_task_stack(tsk); | |
332 | } | |
333 | #endif | |
334 | ||
335 | void free_task(struct task_struct *tsk) | |
336 | { | |
337 | #ifndef CONFIG_THREAD_INFO_IN_TASK | |
338 | /* | |
339 | * The task is finally done with both the stack and thread_info, | |
340 | * so free both. | |
341 | */ | |
342 | release_task_stack(tsk); | |
343 | #else | |
344 | /* | |
345 | * If the task had a separate stack allocation, it should be gone | |
346 | * by now. | |
347 | */ | |
348 | WARN_ON_ONCE(atomic_read(&tsk->stack_refcount) != 0); | |
349 | #endif | |
23f78d4a | 350 | rt_mutex_debug_task_free(tsk); |
fb52607a | 351 | ftrace_graph_exit_task(tsk); |
e2cfabdf | 352 | put_seccomp_filter(tsk); |
f19b9f74 | 353 | arch_release_task_struct(tsk); |
1da177e4 LT |
354 | free_task_struct(tsk); |
355 | } | |
356 | EXPORT_SYMBOL(free_task); | |
357 | ||
ea6d290c ON |
358 | static inline void free_signal_struct(struct signal_struct *sig) |
359 | { | |
97101eb4 | 360 | taskstats_tgid_free(sig); |
1c5354de | 361 | sched_autogroup_exit(sig); |
7283094e MH |
362 | /* |
363 | * __mmdrop is not safe to call from softirq context on x86 due to | |
364 | * pgd_dtor so postpone it to the async context | |
365 | */ | |
26db62f1 | 366 | if (sig->oom_mm) |
7283094e | 367 | mmdrop_async(sig->oom_mm); |
ea6d290c ON |
368 | kmem_cache_free(signal_cachep, sig); |
369 | } | |
370 | ||
371 | static inline void put_signal_struct(struct signal_struct *sig) | |
372 | { | |
1c5354de | 373 | if (atomic_dec_and_test(&sig->sigcnt)) |
ea6d290c ON |
374 | free_signal_struct(sig); |
375 | } | |
376 | ||
158d9ebd | 377 | void __put_task_struct(struct task_struct *tsk) |
1da177e4 | 378 | { |
270f722d | 379 | WARN_ON(!tsk->exit_state); |
1da177e4 LT |
380 | WARN_ON(atomic_read(&tsk->usage)); |
381 | WARN_ON(tsk == current); | |
382 | ||
2e91fa7f | 383 | cgroup_free(tsk); |
156654f4 | 384 | task_numa_free(tsk); |
1a2a4d06 | 385 | security_task_free(tsk); |
e0e81739 | 386 | exit_creds(tsk); |
35df17c5 | 387 | delayacct_tsk_free(tsk); |
ea6d290c | 388 | put_signal_struct(tsk->signal); |
1da177e4 LT |
389 | |
390 | if (!profile_handoff_task(tsk)) | |
391 | free_task(tsk); | |
392 | } | |
77c100c8 | 393 | EXPORT_SYMBOL_GPL(__put_task_struct); |
1da177e4 | 394 | |
6c0a9fa6 | 395 | void __init __weak arch_task_cache_init(void) { } |
61c4628b | 396 | |
ff691f6e HS |
397 | /* |
398 | * set_max_threads | |
399 | */ | |
16db3d3f | 400 | static void set_max_threads(unsigned int max_threads_suggested) |
ff691f6e | 401 | { |
ac1b398d | 402 | u64 threads; |
ff691f6e HS |
403 | |
404 | /* | |
ac1b398d HS |
405 | * The number of threads shall be limited such that the thread |
406 | * structures may only consume a small part of the available memory. | |
ff691f6e | 407 | */ |
ac1b398d HS |
408 | if (fls64(totalram_pages) + fls64(PAGE_SIZE) > 64) |
409 | threads = MAX_THREADS; | |
410 | else | |
411 | threads = div64_u64((u64) totalram_pages * (u64) PAGE_SIZE, | |
412 | (u64) THREAD_SIZE * 8UL); | |
413 | ||
16db3d3f HS |
414 | if (threads > max_threads_suggested) |
415 | threads = max_threads_suggested; | |
416 | ||
ac1b398d | 417 | max_threads = clamp_t(u64, threads, MIN_THREADS, MAX_THREADS); |
ff691f6e HS |
418 | } |
419 | ||
5aaeb5c0 IM |
420 | #ifdef CONFIG_ARCH_WANTS_DYNAMIC_TASK_STRUCT |
421 | /* Initialized by the architecture: */ | |
422 | int arch_task_struct_size __read_mostly; | |
423 | #endif | |
0c8c0f03 | 424 | |
ff691f6e | 425 | void __init fork_init(void) |
1da177e4 | 426 | { |
25f9c081 | 427 | int i; |
f5e10287 | 428 | #ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR |
1da177e4 LT |
429 | #ifndef ARCH_MIN_TASKALIGN |
430 | #define ARCH_MIN_TASKALIGN L1_CACHE_BYTES | |
431 | #endif | |
432 | /* create a slab on which task_structs can be allocated */ | |
5d097056 VD |
433 | task_struct_cachep = kmem_cache_create("task_struct", |
434 | arch_task_struct_size, ARCH_MIN_TASKALIGN, | |
435 | SLAB_PANIC|SLAB_NOTRACK|SLAB_ACCOUNT, NULL); | |
1da177e4 LT |
436 | #endif |
437 | ||
61c4628b SS |
438 | /* do the arch specific task caches init */ |
439 | arch_task_cache_init(); | |
440 | ||
16db3d3f | 441 | set_max_threads(MAX_THREADS); |
1da177e4 LT |
442 | |
443 | init_task.signal->rlim[RLIMIT_NPROC].rlim_cur = max_threads/2; | |
444 | init_task.signal->rlim[RLIMIT_NPROC].rlim_max = max_threads/2; | |
445 | init_task.signal->rlim[RLIMIT_SIGPENDING] = | |
446 | init_task.signal->rlim[RLIMIT_NPROC]; | |
b376c3e1 | 447 | |
25f9c081 EB |
448 | for (i = 0; i < UCOUNT_COUNTS; i++) { |
449 | init_user_ns.ucount_max[i] = max_threads/2; | |
450 | } | |
1da177e4 LT |
451 | } |
452 | ||
52f5684c | 453 | int __weak arch_dup_task_struct(struct task_struct *dst, |
61c4628b SS |
454 | struct task_struct *src) |
455 | { | |
456 | *dst = *src; | |
457 | return 0; | |
458 | } | |
459 | ||
d4311ff1 AT |
460 | void set_task_stack_end_magic(struct task_struct *tsk) |
461 | { | |
462 | unsigned long *stackend; | |
463 | ||
464 | stackend = end_of_stack(tsk); | |
465 | *stackend = STACK_END_MAGIC; /* for overflow detection */ | |
466 | } | |
467 | ||
725fc629 | 468 | static struct task_struct *dup_task_struct(struct task_struct *orig, int node) |
1da177e4 LT |
469 | { |
470 | struct task_struct *tsk; | |
b235beea | 471 | unsigned long *stack; |
ba14a194 | 472 | struct vm_struct *stack_vm_area; |
3e26c149 | 473 | int err; |
1da177e4 | 474 | |
725fc629 AK |
475 | if (node == NUMA_NO_NODE) |
476 | node = tsk_fork_get_node(orig); | |
504f52b5 | 477 | tsk = alloc_task_struct_node(node); |
1da177e4 LT |
478 | if (!tsk) |
479 | return NULL; | |
480 | ||
b235beea LT |
481 | stack = alloc_thread_stack_node(tsk, node); |
482 | if (!stack) | |
f19b9f74 | 483 | goto free_tsk; |
1da177e4 | 484 | |
ba14a194 AL |
485 | stack_vm_area = task_stack_vm_area(tsk); |
486 | ||
fb0a685c | 487 | err = arch_dup_task_struct(tsk, orig); |
ba14a194 AL |
488 | |
489 | /* | |
490 | * arch_dup_task_struct() clobbers the stack-related fields. Make | |
491 | * sure they're properly initialized before using any stack-related | |
492 | * functions again. | |
493 | */ | |
494 | tsk->stack = stack; | |
495 | #ifdef CONFIG_VMAP_STACK | |
496 | tsk->stack_vm_area = stack_vm_area; | |
497 | #endif | |
68f24b08 AL |
498 | #ifdef CONFIG_THREAD_INFO_IN_TASK |
499 | atomic_set(&tsk->stack_refcount, 1); | |
500 | #endif | |
ba14a194 | 501 | |
164c33c6 | 502 | if (err) |
b235beea | 503 | goto free_stack; |
164c33c6 | 504 | |
dbd95212 KC |
505 | #ifdef CONFIG_SECCOMP |
506 | /* | |
507 | * We must handle setting up seccomp filters once we're under | |
508 | * the sighand lock in case orig has changed between now and | |
509 | * then. Until then, filter must be NULL to avoid messing up | |
510 | * the usage counts on the error path calling free_task. | |
511 | */ | |
512 | tsk->seccomp.filter = NULL; | |
513 | #endif | |
87bec58a AM |
514 | |
515 | setup_thread_stack(tsk, orig); | |
8e7cac79 | 516 | clear_user_return_notifier(tsk); |
f26f9aff | 517 | clear_tsk_need_resched(tsk); |
d4311ff1 | 518 | set_task_stack_end_magic(tsk); |
1da177e4 | 519 | |
0a425405 AV |
520 | #ifdef CONFIG_CC_STACKPROTECTOR |
521 | tsk->stack_canary = get_random_int(); | |
522 | #endif | |
523 | ||
fb0a685c DRO |
524 | /* |
525 | * One for us, one for whoever does the "release_task()" (usually | |
526 | * parent) | |
527 | */ | |
528 | atomic_set(&tsk->usage, 2); | |
6c5c9341 | 529 | #ifdef CONFIG_BLK_DEV_IO_TRACE |
2056a782 | 530 | tsk->btrace_seq = 0; |
6c5c9341 | 531 | #endif |
a0aa7f68 | 532 | tsk->splice_pipe = NULL; |
5640f768 | 533 | tsk->task_frag.page = NULL; |
093e5840 | 534 | tsk->wake_q.next = NULL; |
c6a7f572 | 535 | |
ba14a194 | 536 | account_kernel_stack(tsk, 1); |
c6a7f572 | 537 | |
5c9a8750 DV |
538 | kcov_task_init(tsk); |
539 | ||
1da177e4 | 540 | return tsk; |
61c4628b | 541 | |
b235beea | 542 | free_stack: |
ba14a194 | 543 | free_thread_stack(tsk); |
f19b9f74 | 544 | free_tsk: |
61c4628b SS |
545 | free_task_struct(tsk); |
546 | return NULL; | |
1da177e4 LT |
547 | } |
548 | ||
549 | #ifdef CONFIG_MMU | |
0766f788 ER |
550 | static __latent_entropy int dup_mmap(struct mm_struct *mm, |
551 | struct mm_struct *oldmm) | |
1da177e4 | 552 | { |
297c5eee | 553 | struct vm_area_struct *mpnt, *tmp, *prev, **pprev; |
1da177e4 LT |
554 | struct rb_node **rb_link, *rb_parent; |
555 | int retval; | |
556 | unsigned long charge; | |
1da177e4 | 557 | |
32cdba1e | 558 | uprobe_start_dup_mmap(); |
7c051267 MH |
559 | if (down_write_killable(&oldmm->mmap_sem)) { |
560 | retval = -EINTR; | |
561 | goto fail_uprobe_end; | |
562 | } | |
ec8c0446 | 563 | flush_cache_dup_mm(oldmm); |
f8ac4ec9 | 564 | uprobe_dup_mmap(oldmm, mm); |
ad339451 IM |
565 | /* |
566 | * Not linked in yet - no deadlock potential: | |
567 | */ | |
568 | down_write_nested(&mm->mmap_sem, SINGLE_DEPTH_NESTING); | |
7ee78232 | 569 | |
90f31d0e KK |
570 | /* No ordering required: file already has been exposed. */ |
571 | RCU_INIT_POINTER(mm->exe_file, get_mm_exe_file(oldmm)); | |
572 | ||
4f7d4614 | 573 | mm->total_vm = oldmm->total_vm; |
84638335 | 574 | mm->data_vm = oldmm->data_vm; |
4f7d4614 VD |
575 | mm->exec_vm = oldmm->exec_vm; |
576 | mm->stack_vm = oldmm->stack_vm; | |
577 | ||
1da177e4 LT |
578 | rb_link = &mm->mm_rb.rb_node; |
579 | rb_parent = NULL; | |
580 | pprev = &mm->mmap; | |
f8af4da3 | 581 | retval = ksm_fork(mm, oldmm); |
ba76149f AA |
582 | if (retval) |
583 | goto out; | |
584 | retval = khugepaged_fork(mm, oldmm); | |
f8af4da3 HD |
585 | if (retval) |
586 | goto out; | |
1da177e4 | 587 | |
297c5eee | 588 | prev = NULL; |
fd3e42fc | 589 | for (mpnt = oldmm->mmap; mpnt; mpnt = mpnt->vm_next) { |
1da177e4 LT |
590 | struct file *file; |
591 | ||
592 | if (mpnt->vm_flags & VM_DONTCOPY) { | |
84638335 | 593 | vm_stat_account(mm, mpnt->vm_flags, -vma_pages(mpnt)); |
1da177e4 LT |
594 | continue; |
595 | } | |
596 | charge = 0; | |
597 | if (mpnt->vm_flags & VM_ACCOUNT) { | |
b2412b7f HS |
598 | unsigned long len = vma_pages(mpnt); |
599 | ||
191c5424 | 600 | if (security_vm_enough_memory_mm(oldmm, len)) /* sic */ |
1da177e4 LT |
601 | goto fail_nomem; |
602 | charge = len; | |
603 | } | |
e94b1766 | 604 | tmp = kmem_cache_alloc(vm_area_cachep, GFP_KERNEL); |
1da177e4 LT |
605 | if (!tmp) |
606 | goto fail_nomem; | |
607 | *tmp = *mpnt; | |
5beb4930 | 608 | INIT_LIST_HEAD(&tmp->anon_vma_chain); |
ef0855d3 ON |
609 | retval = vma_dup_policy(mpnt, tmp); |
610 | if (retval) | |
1da177e4 | 611 | goto fail_nomem_policy; |
a247c3a9 | 612 | tmp->vm_mm = mm; |
5beb4930 RR |
613 | if (anon_vma_fork(tmp, mpnt)) |
614 | goto fail_nomem_anon_vma_fork; | |
de60f5f1 EM |
615 | tmp->vm_flags &= |
616 | ~(VM_LOCKED|VM_LOCKONFAULT|VM_UFFD_MISSING|VM_UFFD_WP); | |
297c5eee | 617 | tmp->vm_next = tmp->vm_prev = NULL; |
745f234b | 618 | tmp->vm_userfaultfd_ctx = NULL_VM_UFFD_CTX; |
1da177e4 LT |
619 | file = tmp->vm_file; |
620 | if (file) { | |
496ad9aa | 621 | struct inode *inode = file_inode(file); |
b88ed205 HD |
622 | struct address_space *mapping = file->f_mapping; |
623 | ||
1da177e4 LT |
624 | get_file(file); |
625 | if (tmp->vm_flags & VM_DENYWRITE) | |
626 | atomic_dec(&inode->i_writecount); | |
83cde9e8 | 627 | i_mmap_lock_write(mapping); |
b88ed205 | 628 | if (tmp->vm_flags & VM_SHARED) |
4bb5f5d9 | 629 | atomic_inc(&mapping->i_mmap_writable); |
b88ed205 HD |
630 | flush_dcache_mmap_lock(mapping); |
631 | /* insert tmp into the share list, just after mpnt */ | |
27ba0644 KS |
632 | vma_interval_tree_insert_after(tmp, mpnt, |
633 | &mapping->i_mmap); | |
b88ed205 | 634 | flush_dcache_mmap_unlock(mapping); |
83cde9e8 | 635 | i_mmap_unlock_write(mapping); |
1da177e4 LT |
636 | } |
637 | ||
a1e78772 MG |
638 | /* |
639 | * Clear hugetlb-related page reserves for children. This only | |
640 | * affects MAP_PRIVATE mappings. Faults generated by the child | |
641 | * are not guaranteed to succeed, even if read-only | |
642 | */ | |
643 | if (is_vm_hugetlb_page(tmp)) | |
644 | reset_vma_resv_huge_pages(tmp); | |
645 | ||
1da177e4 | 646 | /* |
7ee78232 | 647 | * Link in the new vma and copy the page table entries. |
1da177e4 | 648 | */ |
1da177e4 LT |
649 | *pprev = tmp; |
650 | pprev = &tmp->vm_next; | |
297c5eee LT |
651 | tmp->vm_prev = prev; |
652 | prev = tmp; | |
1da177e4 LT |
653 | |
654 | __vma_link_rb(mm, tmp, rb_link, rb_parent); | |
655 | rb_link = &tmp->vm_rb.rb_right; | |
656 | rb_parent = &tmp->vm_rb; | |
657 | ||
658 | mm->map_count++; | |
0b0db14c | 659 | retval = copy_page_range(mm, oldmm, mpnt); |
1da177e4 LT |
660 | |
661 | if (tmp->vm_ops && tmp->vm_ops->open) | |
662 | tmp->vm_ops->open(tmp); | |
663 | ||
664 | if (retval) | |
665 | goto out; | |
666 | } | |
d6dd61c8 JF |
667 | /* a new mm has just been created */ |
668 | arch_dup_mmap(oldmm, mm); | |
1da177e4 | 669 | retval = 0; |
1da177e4 | 670 | out: |
7ee78232 | 671 | up_write(&mm->mmap_sem); |
fd3e42fc | 672 | flush_tlb_mm(oldmm); |
1da177e4 | 673 | up_write(&oldmm->mmap_sem); |
7c051267 | 674 | fail_uprobe_end: |
32cdba1e | 675 | uprobe_end_dup_mmap(); |
1da177e4 | 676 | return retval; |
5beb4930 | 677 | fail_nomem_anon_vma_fork: |
ef0855d3 | 678 | mpol_put(vma_policy(tmp)); |
1da177e4 LT |
679 | fail_nomem_policy: |
680 | kmem_cache_free(vm_area_cachep, tmp); | |
681 | fail_nomem: | |
682 | retval = -ENOMEM; | |
683 | vm_unacct_memory(charge); | |
684 | goto out; | |
685 | } | |
686 | ||
fb0a685c | 687 | static inline int mm_alloc_pgd(struct mm_struct *mm) |
1da177e4 LT |
688 | { |
689 | mm->pgd = pgd_alloc(mm); | |
690 | if (unlikely(!mm->pgd)) | |
691 | return -ENOMEM; | |
692 | return 0; | |
693 | } | |
694 | ||
fb0a685c | 695 | static inline void mm_free_pgd(struct mm_struct *mm) |
1da177e4 | 696 | { |
5e541973 | 697 | pgd_free(mm, mm->pgd); |
1da177e4 LT |
698 | } |
699 | #else | |
90f31d0e KK |
700 | static int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm) |
701 | { | |
702 | down_write(&oldmm->mmap_sem); | |
703 | RCU_INIT_POINTER(mm->exe_file, get_mm_exe_file(oldmm)); | |
704 | up_write(&oldmm->mmap_sem); | |
705 | return 0; | |
706 | } | |
1da177e4 LT |
707 | #define mm_alloc_pgd(mm) (0) |
708 | #define mm_free_pgd(mm) | |
709 | #endif /* CONFIG_MMU */ | |
710 | ||
23ff4440 | 711 | __cacheline_aligned_in_smp DEFINE_SPINLOCK(mmlist_lock); |
1da177e4 | 712 | |
e94b1766 | 713 | #define allocate_mm() (kmem_cache_alloc(mm_cachep, GFP_KERNEL)) |
1da177e4 LT |
714 | #define free_mm(mm) (kmem_cache_free(mm_cachep, (mm))) |
715 | ||
4cb0e11b HK |
716 | static unsigned long default_dump_filter = MMF_DUMP_FILTER_DEFAULT; |
717 | ||
718 | static int __init coredump_filter_setup(char *s) | |
719 | { | |
720 | default_dump_filter = | |
721 | (simple_strtoul(s, NULL, 0) << MMF_DUMP_FILTER_SHIFT) & | |
722 | MMF_DUMP_FILTER_MASK; | |
723 | return 1; | |
724 | } | |
725 | ||
726 | __setup("coredump_filter=", coredump_filter_setup); | |
727 | ||
1da177e4 LT |
728 | #include <linux/init_task.h> |
729 | ||
858f0993 AD |
730 | static void mm_init_aio(struct mm_struct *mm) |
731 | { | |
732 | #ifdef CONFIG_AIO | |
733 | spin_lock_init(&mm->ioctx_lock); | |
db446a08 | 734 | mm->ioctx_table = NULL; |
858f0993 AD |
735 | #endif |
736 | } | |
737 | ||
33144e84 VD |
738 | static void mm_init_owner(struct mm_struct *mm, struct task_struct *p) |
739 | { | |
740 | #ifdef CONFIG_MEMCG | |
741 | mm->owner = p; | |
742 | #endif | |
743 | } | |
744 | ||
fb0a685c | 745 | static struct mm_struct *mm_init(struct mm_struct *mm, struct task_struct *p) |
1da177e4 | 746 | { |
41f727fd VD |
747 | mm->mmap = NULL; |
748 | mm->mm_rb = RB_ROOT; | |
749 | mm->vmacache_seqnum = 0; | |
1da177e4 LT |
750 | atomic_set(&mm->mm_users, 1); |
751 | atomic_set(&mm->mm_count, 1); | |
752 | init_rwsem(&mm->mmap_sem); | |
753 | INIT_LIST_HEAD(&mm->mmlist); | |
999d9fc1 | 754 | mm->core_state = NULL; |
e1f56c89 | 755 | atomic_long_set(&mm->nr_ptes, 0); |
2d2f5119 | 756 | mm_nr_pmds_init(mm); |
41f727fd VD |
757 | mm->map_count = 0; |
758 | mm->locked_vm = 0; | |
ce65cefa | 759 | mm->pinned_vm = 0; |
d559db08 | 760 | memset(&mm->rss_stat, 0, sizeof(mm->rss_stat)); |
1da177e4 | 761 | spin_lock_init(&mm->page_table_lock); |
41f727fd | 762 | mm_init_cpumask(mm); |
858f0993 | 763 | mm_init_aio(mm); |
cf475ad2 | 764 | mm_init_owner(mm, p); |
41f727fd | 765 | mmu_notifier_mm_init(mm); |
20841405 | 766 | clear_tlb_flush_pending(mm); |
41f727fd VD |
767 | #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS |
768 | mm->pmd_huge_pte = NULL; | |
769 | #endif | |
1da177e4 | 770 | |
a0715cc2 AT |
771 | if (current->mm) { |
772 | mm->flags = current->mm->flags & MMF_INIT_MASK; | |
773 | mm->def_flags = current->mm->def_flags & VM_INIT_DEF_MASK; | |
774 | } else { | |
775 | mm->flags = default_dump_filter; | |
1da177e4 | 776 | mm->def_flags = 0; |
a0715cc2 AT |
777 | } |
778 | ||
41f727fd VD |
779 | if (mm_alloc_pgd(mm)) |
780 | goto fail_nopgd; | |
781 | ||
782 | if (init_new_context(p, mm)) | |
783 | goto fail_nocontext; | |
78fb7466 | 784 | |
41f727fd VD |
785 | return mm; |
786 | ||
787 | fail_nocontext: | |
788 | mm_free_pgd(mm); | |
789 | fail_nopgd: | |
1da177e4 LT |
790 | free_mm(mm); |
791 | return NULL; | |
792 | } | |
793 | ||
c3f0327f KK |
794 | static void check_mm(struct mm_struct *mm) |
795 | { | |
796 | int i; | |
797 | ||
798 | for (i = 0; i < NR_MM_COUNTERS; i++) { | |
799 | long x = atomic_long_read(&mm->rss_stat.count[i]); | |
800 | ||
801 | if (unlikely(x)) | |
802 | printk(KERN_ALERT "BUG: Bad rss-counter state " | |
803 | "mm:%p idx:%d val:%ld\n", mm, i, x); | |
804 | } | |
b30fe6c7 KS |
805 | |
806 | if (atomic_long_read(&mm->nr_ptes)) | |
807 | pr_alert("BUG: non-zero nr_ptes on freeing mm: %ld\n", | |
808 | atomic_long_read(&mm->nr_ptes)); | |
809 | if (mm_nr_pmds(mm)) | |
810 | pr_alert("BUG: non-zero nr_pmds on freeing mm: %ld\n", | |
811 | mm_nr_pmds(mm)); | |
812 | ||
e009bb30 | 813 | #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS |
96dad67f | 814 | VM_BUG_ON_MM(mm->pmd_huge_pte, mm); |
c3f0327f KK |
815 | #endif |
816 | } | |
817 | ||
1da177e4 LT |
818 | /* |
819 | * Allocate and initialize an mm_struct. | |
820 | */ | |
fb0a685c | 821 | struct mm_struct *mm_alloc(void) |
1da177e4 | 822 | { |
fb0a685c | 823 | struct mm_struct *mm; |
1da177e4 LT |
824 | |
825 | mm = allocate_mm(); | |
de03c72c KM |
826 | if (!mm) |
827 | return NULL; | |
828 | ||
829 | memset(mm, 0, sizeof(*mm)); | |
6345d24d | 830 | return mm_init(mm, current); |
1da177e4 LT |
831 | } |
832 | ||
833 | /* | |
834 | * Called when the last reference to the mm | |
835 | * is dropped: either by a lazy thread or by | |
836 | * mmput. Free the page directory and the mm. | |
837 | */ | |
7ad5b3a5 | 838 | void __mmdrop(struct mm_struct *mm) |
1da177e4 LT |
839 | { |
840 | BUG_ON(mm == &init_mm); | |
841 | mm_free_pgd(mm); | |
842 | destroy_context(mm); | |
cddb8a5c | 843 | mmu_notifier_mm_destroy(mm); |
c3f0327f | 844 | check_mm(mm); |
1da177e4 LT |
845 | free_mm(mm); |
846 | } | |
6d4e4c4f | 847 | EXPORT_SYMBOL_GPL(__mmdrop); |
1da177e4 | 848 | |
ec8d7c14 MH |
849 | static inline void __mmput(struct mm_struct *mm) |
850 | { | |
851 | VM_BUG_ON(atomic_read(&mm->mm_users)); | |
852 | ||
853 | uprobe_clear_state(mm); | |
854 | exit_aio(mm); | |
855 | ksm_exit(mm); | |
856 | khugepaged_exit(mm); /* must run before exit_mmap */ | |
857 | exit_mmap(mm); | |
6fcb52a5 | 858 | mm_put_huge_zero_page(mm); |
ec8d7c14 MH |
859 | set_mm_exe_file(mm, NULL); |
860 | if (!list_empty(&mm->mmlist)) { | |
861 | spin_lock(&mmlist_lock); | |
862 | list_del(&mm->mmlist); | |
863 | spin_unlock(&mmlist_lock); | |
864 | } | |
865 | if (mm->binfmt) | |
866 | module_put(mm->binfmt->module); | |
862e3073 | 867 | set_bit(MMF_OOM_SKIP, &mm->flags); |
ec8d7c14 MH |
868 | mmdrop(mm); |
869 | } | |
870 | ||
1da177e4 LT |
871 | /* |
872 | * Decrement the use count and release all resources for an mm. | |
873 | */ | |
874 | void mmput(struct mm_struct *mm) | |
875 | { | |
0ae26f1b AM |
876 | might_sleep(); |
877 | ||
ec8d7c14 MH |
878 | if (atomic_dec_and_test(&mm->mm_users)) |
879 | __mmput(mm); | |
880 | } | |
881 | EXPORT_SYMBOL_GPL(mmput); | |
882 | ||
7ef949d7 | 883 | #ifdef CONFIG_MMU |
ec8d7c14 MH |
884 | static void mmput_async_fn(struct work_struct *work) |
885 | { | |
886 | struct mm_struct *mm = container_of(work, struct mm_struct, async_put_work); | |
887 | __mmput(mm); | |
888 | } | |
889 | ||
890 | void mmput_async(struct mm_struct *mm) | |
891 | { | |
1da177e4 | 892 | if (atomic_dec_and_test(&mm->mm_users)) { |
ec8d7c14 MH |
893 | INIT_WORK(&mm->async_put_work, mmput_async_fn); |
894 | schedule_work(&mm->async_put_work); | |
1da177e4 LT |
895 | } |
896 | } | |
7ef949d7 | 897 | #endif |
1da177e4 | 898 | |
90f31d0e KK |
899 | /** |
900 | * set_mm_exe_file - change a reference to the mm's executable file | |
901 | * | |
902 | * This changes mm's executable file (shown as symlink /proc/[pid]/exe). | |
903 | * | |
6e399cd1 DB |
904 | * Main users are mmput() and sys_execve(). Callers prevent concurrent |
905 | * invocations: in mmput() nobody alive left, in execve task is single | |
906 | * threaded. sys_prctl(PR_SET_MM_MAP/EXE_FILE) also needs to set the | |
907 | * mm->exe_file, but does so without using set_mm_exe_file() in order | |
908 | * to do avoid the need for any locks. | |
90f31d0e | 909 | */ |
38646013 JS |
910 | void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file) |
911 | { | |
6e399cd1 DB |
912 | struct file *old_exe_file; |
913 | ||
914 | /* | |
915 | * It is safe to dereference the exe_file without RCU as | |
916 | * this function is only called if nobody else can access | |
917 | * this mm -- see comment above for justification. | |
918 | */ | |
919 | old_exe_file = rcu_dereference_raw(mm->exe_file); | |
90f31d0e | 920 | |
38646013 JS |
921 | if (new_exe_file) |
922 | get_file(new_exe_file); | |
90f31d0e KK |
923 | rcu_assign_pointer(mm->exe_file, new_exe_file); |
924 | if (old_exe_file) | |
925 | fput(old_exe_file); | |
38646013 JS |
926 | } |
927 | ||
90f31d0e KK |
928 | /** |
929 | * get_mm_exe_file - acquire a reference to the mm's executable file | |
930 | * | |
931 | * Returns %NULL if mm has no associated executable file. | |
932 | * User must release file via fput(). | |
933 | */ | |
38646013 JS |
934 | struct file *get_mm_exe_file(struct mm_struct *mm) |
935 | { | |
936 | struct file *exe_file; | |
937 | ||
90f31d0e KK |
938 | rcu_read_lock(); |
939 | exe_file = rcu_dereference(mm->exe_file); | |
940 | if (exe_file && !get_file_rcu(exe_file)) | |
941 | exe_file = NULL; | |
942 | rcu_read_unlock(); | |
38646013 JS |
943 | return exe_file; |
944 | } | |
11163348 | 945 | EXPORT_SYMBOL(get_mm_exe_file); |
38646013 | 946 | |
cd81a917 MG |
947 | /** |
948 | * get_task_exe_file - acquire a reference to the task's executable file | |
949 | * | |
950 | * Returns %NULL if task's mm (if any) has no associated executable file or | |
951 | * this is a kernel thread with borrowed mm (see the comment above get_task_mm). | |
952 | * User must release file via fput(). | |
953 | */ | |
954 | struct file *get_task_exe_file(struct task_struct *task) | |
955 | { | |
956 | struct file *exe_file = NULL; | |
957 | struct mm_struct *mm; | |
958 | ||
959 | task_lock(task); | |
960 | mm = task->mm; | |
961 | if (mm) { | |
962 | if (!(task->flags & PF_KTHREAD)) | |
963 | exe_file = get_mm_exe_file(mm); | |
964 | } | |
965 | task_unlock(task); | |
966 | return exe_file; | |
967 | } | |
968 | EXPORT_SYMBOL(get_task_exe_file); | |
38646013 | 969 | |
1da177e4 LT |
970 | /** |
971 | * get_task_mm - acquire a reference to the task's mm | |
972 | * | |
246bb0b1 | 973 | * Returns %NULL if the task has no mm. Checks PF_KTHREAD (meaning |
1da177e4 LT |
974 | * this kernel workthread has transiently adopted a user mm with use_mm, |
975 | * to do its AIO) is not set and if so returns a reference to it, after | |
976 | * bumping up the use count. User must release the mm via mmput() | |
977 | * after use. Typically used by /proc and ptrace. | |
978 | */ | |
979 | struct mm_struct *get_task_mm(struct task_struct *task) | |
980 | { | |
981 | struct mm_struct *mm; | |
982 | ||
983 | task_lock(task); | |
984 | mm = task->mm; | |
985 | if (mm) { | |
246bb0b1 | 986 | if (task->flags & PF_KTHREAD) |
1da177e4 LT |
987 | mm = NULL; |
988 | else | |
989 | atomic_inc(&mm->mm_users); | |
990 | } | |
991 | task_unlock(task); | |
992 | return mm; | |
993 | } | |
994 | EXPORT_SYMBOL_GPL(get_task_mm); | |
995 | ||
8cdb878d CY |
996 | struct mm_struct *mm_access(struct task_struct *task, unsigned int mode) |
997 | { | |
998 | struct mm_struct *mm; | |
999 | int err; | |
1000 | ||
1001 | err = mutex_lock_killable(&task->signal->cred_guard_mutex); | |
1002 | if (err) | |
1003 | return ERR_PTR(err); | |
1004 | ||
1005 | mm = get_task_mm(task); | |
1006 | if (mm && mm != current->mm && | |
1007 | !ptrace_may_access(task, mode)) { | |
1008 | mmput(mm); | |
1009 | mm = ERR_PTR(-EACCES); | |
1010 | } | |
1011 | mutex_unlock(&task->signal->cred_guard_mutex); | |
1012 | ||
1013 | return mm; | |
1014 | } | |
1015 | ||
57b59c4a | 1016 | static void complete_vfork_done(struct task_struct *tsk) |
c415c3b4 | 1017 | { |
d68b46fe | 1018 | struct completion *vfork; |
c415c3b4 | 1019 | |
d68b46fe ON |
1020 | task_lock(tsk); |
1021 | vfork = tsk->vfork_done; | |
1022 | if (likely(vfork)) { | |
1023 | tsk->vfork_done = NULL; | |
1024 | complete(vfork); | |
1025 | } | |
1026 | task_unlock(tsk); | |
1027 | } | |
1028 | ||
1029 | static int wait_for_vfork_done(struct task_struct *child, | |
1030 | struct completion *vfork) | |
1031 | { | |
1032 | int killed; | |
1033 | ||
1034 | freezer_do_not_count(); | |
1035 | killed = wait_for_completion_killable(vfork); | |
1036 | freezer_count(); | |
1037 | ||
1038 | if (killed) { | |
1039 | task_lock(child); | |
1040 | child->vfork_done = NULL; | |
1041 | task_unlock(child); | |
1042 | } | |
1043 | ||
1044 | put_task_struct(child); | |
1045 | return killed; | |
c415c3b4 ON |
1046 | } |
1047 | ||
1da177e4 LT |
1048 | /* Please note the differences between mmput and mm_release. |
1049 | * mmput is called whenever we stop holding onto a mm_struct, | |
1050 | * error success whatever. | |
1051 | * | |
1052 | * mm_release is called after a mm_struct has been removed | |
1053 | * from the current process. | |
1054 | * | |
1055 | * This difference is important for error handling, when we | |
1056 | * only half set up a mm_struct for a new process and need to restore | |
1057 | * the old one. Because we mmput the new mm_struct before | |
1058 | * restoring the old one. . . | |
1059 | * Eric Biederman 10 January 1998 | |
1060 | */ | |
1061 | void mm_release(struct task_struct *tsk, struct mm_struct *mm) | |
1062 | { | |
8141c7f3 LT |
1063 | /* Get rid of any futexes when releasing the mm */ |
1064 | #ifdef CONFIG_FUTEX | |
fc6b177d | 1065 | if (unlikely(tsk->robust_list)) { |
8141c7f3 | 1066 | exit_robust_list(tsk); |
fc6b177d PZ |
1067 | tsk->robust_list = NULL; |
1068 | } | |
8141c7f3 | 1069 | #ifdef CONFIG_COMPAT |
fc6b177d | 1070 | if (unlikely(tsk->compat_robust_list)) { |
8141c7f3 | 1071 | compat_exit_robust_list(tsk); |
fc6b177d PZ |
1072 | tsk->compat_robust_list = NULL; |
1073 | } | |
8141c7f3 | 1074 | #endif |
322a2c10 TG |
1075 | if (unlikely(!list_empty(&tsk->pi_state_list))) |
1076 | exit_pi_state_list(tsk); | |
8141c7f3 LT |
1077 | #endif |
1078 | ||
0326f5a9 SD |
1079 | uprobe_free_utask(tsk); |
1080 | ||
1da177e4 LT |
1081 | /* Get rid of any cached register state */ |
1082 | deactivate_mm(tsk, mm); | |
1083 | ||
fec1d011 | 1084 | /* |
735f2770 MH |
1085 | * Signal userspace if we're not exiting with a core dump |
1086 | * because we want to leave the value intact for debugging | |
1087 | * purposes. | |
fec1d011 | 1088 | */ |
9c8a8228 | 1089 | if (tsk->clear_child_tid) { |
735f2770 | 1090 | if (!(tsk->signal->flags & SIGNAL_GROUP_COREDUMP) && |
9c8a8228 ED |
1091 | atomic_read(&mm->mm_users) > 1) { |
1092 | /* | |
1093 | * We don't check the error code - if userspace has | |
1094 | * not set up a proper pointer then tough luck. | |
1095 | */ | |
1096 | put_user(0, tsk->clear_child_tid); | |
1097 | sys_futex(tsk->clear_child_tid, FUTEX_WAKE, | |
1098 | 1, NULL, NULL, 0); | |
1099 | } | |
1da177e4 | 1100 | tsk->clear_child_tid = NULL; |
1da177e4 | 1101 | } |
f7505d64 KK |
1102 | |
1103 | /* | |
1104 | * All done, finally we can wake up parent and return this mm to him. | |
1105 | * Also kthread_stop() uses this completion for synchronization. | |
1106 | */ | |
1107 | if (tsk->vfork_done) | |
1108 | complete_vfork_done(tsk); | |
1da177e4 LT |
1109 | } |
1110 | ||
a0a7ec30 JD |
1111 | /* |
1112 | * Allocate a new mm structure and copy contents from the | |
1113 | * mm structure of the passed in task structure. | |
1114 | */ | |
ff252c1f | 1115 | static struct mm_struct *dup_mm(struct task_struct *tsk) |
a0a7ec30 JD |
1116 | { |
1117 | struct mm_struct *mm, *oldmm = current->mm; | |
1118 | int err; | |
1119 | ||
a0a7ec30 JD |
1120 | mm = allocate_mm(); |
1121 | if (!mm) | |
1122 | goto fail_nomem; | |
1123 | ||
1124 | memcpy(mm, oldmm, sizeof(*mm)); | |
1125 | ||
78fb7466 | 1126 | if (!mm_init(mm, tsk)) |
a0a7ec30 JD |
1127 | goto fail_nomem; |
1128 | ||
a0a7ec30 JD |
1129 | err = dup_mmap(mm, oldmm); |
1130 | if (err) | |
1131 | goto free_pt; | |
1132 | ||
1133 | mm->hiwater_rss = get_mm_rss(mm); | |
1134 | mm->hiwater_vm = mm->total_vm; | |
1135 | ||
801460d0 HS |
1136 | if (mm->binfmt && !try_module_get(mm->binfmt->module)) |
1137 | goto free_pt; | |
1138 | ||
a0a7ec30 JD |
1139 | return mm; |
1140 | ||
1141 | free_pt: | |
801460d0 HS |
1142 | /* don't put binfmt in mmput, we haven't got module yet */ |
1143 | mm->binfmt = NULL; | |
a0a7ec30 JD |
1144 | mmput(mm); |
1145 | ||
1146 | fail_nomem: | |
1147 | return NULL; | |
a0a7ec30 JD |
1148 | } |
1149 | ||
fb0a685c | 1150 | static int copy_mm(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 | 1151 | { |
fb0a685c | 1152 | struct mm_struct *mm, *oldmm; |
1da177e4 LT |
1153 | int retval; |
1154 | ||
1155 | tsk->min_flt = tsk->maj_flt = 0; | |
1156 | tsk->nvcsw = tsk->nivcsw = 0; | |
17406b82 MSB |
1157 | #ifdef CONFIG_DETECT_HUNG_TASK |
1158 | tsk->last_switch_count = tsk->nvcsw + tsk->nivcsw; | |
1159 | #endif | |
1da177e4 LT |
1160 | |
1161 | tsk->mm = NULL; | |
1162 | tsk->active_mm = NULL; | |
1163 | ||
1164 | /* | |
1165 | * Are we cloning a kernel thread? | |
1166 | * | |
1167 | * We need to steal a active VM for that.. | |
1168 | */ | |
1169 | oldmm = current->mm; | |
1170 | if (!oldmm) | |
1171 | return 0; | |
1172 | ||
615d6e87 DB |
1173 | /* initialize the new vmacache entries */ |
1174 | vmacache_flush(tsk); | |
1175 | ||
1da177e4 LT |
1176 | if (clone_flags & CLONE_VM) { |
1177 | atomic_inc(&oldmm->mm_users); | |
1178 | mm = oldmm; | |
1da177e4 LT |
1179 | goto good_mm; |
1180 | } | |
1181 | ||
1182 | retval = -ENOMEM; | |
a0a7ec30 | 1183 | mm = dup_mm(tsk); |
1da177e4 LT |
1184 | if (!mm) |
1185 | goto fail_nomem; | |
1186 | ||
1da177e4 LT |
1187 | good_mm: |
1188 | tsk->mm = mm; | |
1189 | tsk->active_mm = mm; | |
1190 | return 0; | |
1191 | ||
1da177e4 LT |
1192 | fail_nomem: |
1193 | return retval; | |
1da177e4 LT |
1194 | } |
1195 | ||
a39bc516 | 1196 | static int copy_fs(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 | 1197 | { |
498052bb | 1198 | struct fs_struct *fs = current->fs; |
1da177e4 | 1199 | if (clone_flags & CLONE_FS) { |
498052bb | 1200 | /* tsk->fs is already what we want */ |
2a4419b5 | 1201 | spin_lock(&fs->lock); |
498052bb | 1202 | if (fs->in_exec) { |
2a4419b5 | 1203 | spin_unlock(&fs->lock); |
498052bb AV |
1204 | return -EAGAIN; |
1205 | } | |
1206 | fs->users++; | |
2a4419b5 | 1207 | spin_unlock(&fs->lock); |
1da177e4 LT |
1208 | return 0; |
1209 | } | |
498052bb | 1210 | tsk->fs = copy_fs_struct(fs); |
1da177e4 LT |
1211 | if (!tsk->fs) |
1212 | return -ENOMEM; | |
1213 | return 0; | |
1214 | } | |
1215 | ||
fb0a685c | 1216 | static int copy_files(unsigned long clone_flags, struct task_struct *tsk) |
a016f338 JD |
1217 | { |
1218 | struct files_struct *oldf, *newf; | |
1219 | int error = 0; | |
1220 | ||
1221 | /* | |
1222 | * A background process may not have any files ... | |
1223 | */ | |
1224 | oldf = current->files; | |
1225 | if (!oldf) | |
1226 | goto out; | |
1227 | ||
1228 | if (clone_flags & CLONE_FILES) { | |
1229 | atomic_inc(&oldf->count); | |
1230 | goto out; | |
1231 | } | |
1232 | ||
a016f338 JD |
1233 | newf = dup_fd(oldf, &error); |
1234 | if (!newf) | |
1235 | goto out; | |
1236 | ||
1237 | tsk->files = newf; | |
1238 | error = 0; | |
1239 | out: | |
1240 | return error; | |
1241 | } | |
1242 | ||
fadad878 | 1243 | static int copy_io(unsigned long clone_flags, struct task_struct *tsk) |
fd0928df JA |
1244 | { |
1245 | #ifdef CONFIG_BLOCK | |
1246 | struct io_context *ioc = current->io_context; | |
6e736be7 | 1247 | struct io_context *new_ioc; |
fd0928df JA |
1248 | |
1249 | if (!ioc) | |
1250 | return 0; | |
fadad878 JA |
1251 | /* |
1252 | * Share io context with parent, if CLONE_IO is set | |
1253 | */ | |
1254 | if (clone_flags & CLONE_IO) { | |
3d48749d TH |
1255 | ioc_task_link(ioc); |
1256 | tsk->io_context = ioc; | |
fadad878 | 1257 | } else if (ioprio_valid(ioc->ioprio)) { |
6e736be7 TH |
1258 | new_ioc = get_task_io_context(tsk, GFP_KERNEL, NUMA_NO_NODE); |
1259 | if (unlikely(!new_ioc)) | |
fd0928df JA |
1260 | return -ENOMEM; |
1261 | ||
6e736be7 | 1262 | new_ioc->ioprio = ioc->ioprio; |
11a3122f | 1263 | put_io_context(new_ioc); |
fd0928df JA |
1264 | } |
1265 | #endif | |
1266 | return 0; | |
1267 | } | |
1268 | ||
a39bc516 | 1269 | static int copy_sighand(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 LT |
1270 | { |
1271 | struct sighand_struct *sig; | |
1272 | ||
60348802 | 1273 | if (clone_flags & CLONE_SIGHAND) { |
1da177e4 LT |
1274 | atomic_inc(¤t->sighand->count); |
1275 | return 0; | |
1276 | } | |
1277 | sig = kmem_cache_alloc(sighand_cachep, GFP_KERNEL); | |
e56d0903 | 1278 | rcu_assign_pointer(tsk->sighand, sig); |
1da177e4 LT |
1279 | if (!sig) |
1280 | return -ENOMEM; | |
9d7fb042 | 1281 | |
1da177e4 LT |
1282 | atomic_set(&sig->count, 1); |
1283 | memcpy(sig->action, current->sighand->action, sizeof(sig->action)); | |
1284 | return 0; | |
1285 | } | |
1286 | ||
a7e5328a | 1287 | void __cleanup_sighand(struct sighand_struct *sighand) |
c81addc9 | 1288 | { |
d80e731e ON |
1289 | if (atomic_dec_and_test(&sighand->count)) { |
1290 | signalfd_cleanup(sighand); | |
392809b2 ON |
1291 | /* |
1292 | * sighand_cachep is SLAB_DESTROY_BY_RCU so we can free it | |
1293 | * without an RCU grace period, see __lock_task_sighand(). | |
1294 | */ | |
c81addc9 | 1295 | kmem_cache_free(sighand_cachep, sighand); |
d80e731e | 1296 | } |
c81addc9 ON |
1297 | } |
1298 | ||
f06febc9 FM |
1299 | /* |
1300 | * Initialize POSIX timer handling for a thread group. | |
1301 | */ | |
1302 | static void posix_cpu_timers_init_group(struct signal_struct *sig) | |
1303 | { | |
78d7d407 JS |
1304 | unsigned long cpu_limit; |
1305 | ||
316c1608 | 1306 | cpu_limit = READ_ONCE(sig->rlim[RLIMIT_CPU].rlim_cur); |
78d7d407 JS |
1307 | if (cpu_limit != RLIM_INFINITY) { |
1308 | sig->cputime_expires.prof_exp = secs_to_cputime(cpu_limit); | |
d5c373eb | 1309 | sig->cputimer.running = true; |
6279a751 ON |
1310 | } |
1311 | ||
f06febc9 FM |
1312 | /* The timer lists. */ |
1313 | INIT_LIST_HEAD(&sig->cpu_timers[0]); | |
1314 | INIT_LIST_HEAD(&sig->cpu_timers[1]); | |
1315 | INIT_LIST_HEAD(&sig->cpu_timers[2]); | |
1316 | } | |
1317 | ||
a39bc516 | 1318 | static int copy_signal(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 LT |
1319 | { |
1320 | struct signal_struct *sig; | |
1da177e4 | 1321 | |
4ab6c083 | 1322 | if (clone_flags & CLONE_THREAD) |
490dea45 | 1323 | return 0; |
490dea45 | 1324 | |
a56704ef | 1325 | sig = kmem_cache_zalloc(signal_cachep, GFP_KERNEL); |
1da177e4 LT |
1326 | tsk->signal = sig; |
1327 | if (!sig) | |
1328 | return -ENOMEM; | |
1329 | ||
b3ac022c | 1330 | sig->nr_threads = 1; |
1da177e4 | 1331 | atomic_set(&sig->live, 1); |
b3ac022c | 1332 | atomic_set(&sig->sigcnt, 1); |
0c740d0a ON |
1333 | |
1334 | /* list_add(thread_node, thread_head) without INIT_LIST_HEAD() */ | |
1335 | sig->thread_head = (struct list_head)LIST_HEAD_INIT(tsk->thread_node); | |
1336 | tsk->thread_node = (struct list_head)LIST_HEAD_INIT(sig->thread_head); | |
1337 | ||
1da177e4 | 1338 | init_waitqueue_head(&sig->wait_chldexit); |
db51aecc | 1339 | sig->curr_target = tsk; |
1da177e4 LT |
1340 | init_sigpending(&sig->shared_pending); |
1341 | INIT_LIST_HEAD(&sig->posix_timers); | |
e78c3496 | 1342 | seqlock_init(&sig->stats_lock); |
9d7fb042 | 1343 | prev_cputime_init(&sig->prev_cputime); |
1da177e4 | 1344 | |
c9cb2e3d | 1345 | hrtimer_init(&sig->real_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL); |
1da177e4 | 1346 | sig->real_timer.function = it_real_fn; |
1da177e4 | 1347 | |
1da177e4 LT |
1348 | task_lock(current->group_leader); |
1349 | memcpy(sig->rlim, current->signal->rlim, sizeof sig->rlim); | |
1350 | task_unlock(current->group_leader); | |
1351 | ||
6279a751 ON |
1352 | posix_cpu_timers_init_group(sig); |
1353 | ||
522ed776 | 1354 | tty_audit_fork(sig); |
5091faa4 | 1355 | sched_autogroup_fork(sig); |
522ed776 | 1356 | |
a63d83f4 | 1357 | sig->oom_score_adj = current->signal->oom_score_adj; |
dabb16f6 | 1358 | sig->oom_score_adj_min = current->signal->oom_score_adj_min; |
28b83c51 | 1359 | |
ebec18a6 LP |
1360 | sig->has_child_subreaper = current->signal->has_child_subreaper || |
1361 | current->signal->is_child_subreaper; | |
1362 | ||
9b1bf12d KM |
1363 | mutex_init(&sig->cred_guard_mutex); |
1364 | ||
1da177e4 LT |
1365 | return 0; |
1366 | } | |
1367 | ||
dbd95212 KC |
1368 | static void copy_seccomp(struct task_struct *p) |
1369 | { | |
1370 | #ifdef CONFIG_SECCOMP | |
1371 | /* | |
1372 | * Must be called with sighand->lock held, which is common to | |
1373 | * all threads in the group. Holding cred_guard_mutex is not | |
1374 | * needed because this new task is not yet running and cannot | |
1375 | * be racing exec. | |
1376 | */ | |
69f6a34b | 1377 | assert_spin_locked(¤t->sighand->siglock); |
dbd95212 KC |
1378 | |
1379 | /* Ref-count the new filter user, and assign it. */ | |
1380 | get_seccomp_filter(current); | |
1381 | p->seccomp = current->seccomp; | |
1382 | ||
1383 | /* | |
1384 | * Explicitly enable no_new_privs here in case it got set | |
1385 | * between the task_struct being duplicated and holding the | |
1386 | * sighand lock. The seccomp state and nnp must be in sync. | |
1387 | */ | |
1388 | if (task_no_new_privs(current)) | |
1389 | task_set_no_new_privs(p); | |
1390 | ||
1391 | /* | |
1392 | * If the parent gained a seccomp mode after copying thread | |
1393 | * flags and between before we held the sighand lock, we have | |
1394 | * to manually enable the seccomp thread flag here. | |
1395 | */ | |
1396 | if (p->seccomp.mode != SECCOMP_MODE_DISABLED) | |
1397 | set_tsk_thread_flag(p, TIF_SECCOMP); | |
1398 | #endif | |
1399 | } | |
1400 | ||
17da2bd9 | 1401 | SYSCALL_DEFINE1(set_tid_address, int __user *, tidptr) |
1da177e4 LT |
1402 | { |
1403 | current->clear_child_tid = tidptr; | |
1404 | ||
b488893a | 1405 | return task_pid_vnr(current); |
1da177e4 LT |
1406 | } |
1407 | ||
a39bc516 | 1408 | static void rt_mutex_init_task(struct task_struct *p) |
23f78d4a | 1409 | { |
1d615482 | 1410 | raw_spin_lock_init(&p->pi_lock); |
e29e175b | 1411 | #ifdef CONFIG_RT_MUTEXES |
fb00aca4 PZ |
1412 | p->pi_waiters = RB_ROOT; |
1413 | p->pi_waiters_leftmost = NULL; | |
23f78d4a | 1414 | p->pi_blocked_on = NULL; |
23f78d4a IM |
1415 | #endif |
1416 | } | |
1417 | ||
f06febc9 FM |
1418 | /* |
1419 | * Initialize POSIX timer handling for a single task. | |
1420 | */ | |
1421 | static void posix_cpu_timers_init(struct task_struct *tsk) | |
1422 | { | |
64861634 MS |
1423 | tsk->cputime_expires.prof_exp = 0; |
1424 | tsk->cputime_expires.virt_exp = 0; | |
f06febc9 FM |
1425 | tsk->cputime_expires.sched_exp = 0; |
1426 | INIT_LIST_HEAD(&tsk->cpu_timers[0]); | |
1427 | INIT_LIST_HEAD(&tsk->cpu_timers[1]); | |
1428 | INIT_LIST_HEAD(&tsk->cpu_timers[2]); | |
1429 | } | |
1430 | ||
81907739 ON |
1431 | static inline void |
1432 | init_task_pid(struct task_struct *task, enum pid_type type, struct pid *pid) | |
1433 | { | |
1434 | task->pids[type].pid = pid; | |
1435 | } | |
1436 | ||
1da177e4 LT |
1437 | /* |
1438 | * This creates a new process as a copy of the old one, | |
1439 | * but does not actually start it yet. | |
1440 | * | |
1441 | * It copies the registers, and all the appropriate | |
1442 | * parts of the process environment (as per the clone | |
1443 | * flags). The actual kick-off is left to the caller. | |
1444 | */ | |
0766f788 ER |
1445 | static __latent_entropy struct task_struct *copy_process( |
1446 | unsigned long clone_flags, | |
36c8b586 | 1447 | unsigned long stack_start, |
36c8b586 | 1448 | unsigned long stack_size, |
36c8b586 | 1449 | int __user *child_tidptr, |
09a05394 | 1450 | struct pid *pid, |
3033f14a | 1451 | int trace, |
725fc629 AK |
1452 | unsigned long tls, |
1453 | int node) | |
1da177e4 LT |
1454 | { |
1455 | int retval; | |
a24efe62 | 1456 | struct task_struct *p; |
1da177e4 LT |
1457 | |
1458 | if ((clone_flags & (CLONE_NEWNS|CLONE_FS)) == (CLONE_NEWNS|CLONE_FS)) | |
1459 | return ERR_PTR(-EINVAL); | |
1460 | ||
e66eded8 EB |
1461 | if ((clone_flags & (CLONE_NEWUSER|CLONE_FS)) == (CLONE_NEWUSER|CLONE_FS)) |
1462 | return ERR_PTR(-EINVAL); | |
1463 | ||
1da177e4 LT |
1464 | /* |
1465 | * Thread groups must share signals as well, and detached threads | |
1466 | * can only be started up within the thread group. | |
1467 | */ | |
1468 | if ((clone_flags & CLONE_THREAD) && !(clone_flags & CLONE_SIGHAND)) | |
1469 | return ERR_PTR(-EINVAL); | |
1470 | ||
1471 | /* | |
1472 | * Shared signal handlers imply shared VM. By way of the above, | |
1473 | * thread groups also imply shared VM. Blocking this case allows | |
1474 | * for various simplifications in other code. | |
1475 | */ | |
1476 | if ((clone_flags & CLONE_SIGHAND) && !(clone_flags & CLONE_VM)) | |
1477 | return ERR_PTR(-EINVAL); | |
1478 | ||
123be07b SB |
1479 | /* |
1480 | * Siblings of global init remain as zombies on exit since they are | |
1481 | * not reaped by their parent (swapper). To solve this and to avoid | |
1482 | * multi-rooted process trees, prevent global and container-inits | |
1483 | * from creating siblings. | |
1484 | */ | |
1485 | if ((clone_flags & CLONE_PARENT) && | |
1486 | current->signal->flags & SIGNAL_UNKILLABLE) | |
1487 | return ERR_PTR(-EINVAL); | |
1488 | ||
8382fcac | 1489 | /* |
40a0d32d | 1490 | * If the new process will be in a different pid or user namespace |
faf00da5 | 1491 | * do not allow it to share a thread group with the forking task. |
8382fcac | 1492 | */ |
faf00da5 | 1493 | if (clone_flags & CLONE_THREAD) { |
40a0d32d ON |
1494 | if ((clone_flags & (CLONE_NEWUSER | CLONE_NEWPID)) || |
1495 | (task_active_pid_ns(current) != | |
1496 | current->nsproxy->pid_ns_for_children)) | |
1497 | return ERR_PTR(-EINVAL); | |
1498 | } | |
8382fcac | 1499 | |
1da177e4 LT |
1500 | retval = security_task_create(clone_flags); |
1501 | if (retval) | |
1502 | goto fork_out; | |
1503 | ||
1504 | retval = -ENOMEM; | |
725fc629 | 1505 | p = dup_task_struct(current, node); |
1da177e4 LT |
1506 | if (!p) |
1507 | goto fork_out; | |
1508 | ||
f7e8b616 SR |
1509 | ftrace_graph_init_task(p); |
1510 | ||
bea493a0 PZ |
1511 | rt_mutex_init_task(p); |
1512 | ||
d12c1a37 | 1513 | #ifdef CONFIG_PROVE_LOCKING |
de30a2b3 IM |
1514 | DEBUG_LOCKS_WARN_ON(!p->hardirqs_enabled); |
1515 | DEBUG_LOCKS_WARN_ON(!p->softirqs_enabled); | |
1516 | #endif | |
1da177e4 | 1517 | retval = -EAGAIN; |
3b11a1de | 1518 | if (atomic_read(&p->real_cred->user->processes) >= |
78d7d407 | 1519 | task_rlimit(p, RLIMIT_NPROC)) { |
b57922b6 EP |
1520 | if (p->real_cred->user != INIT_USER && |
1521 | !capable(CAP_SYS_RESOURCE) && !capable(CAP_SYS_ADMIN)) | |
1da177e4 LT |
1522 | goto bad_fork_free; |
1523 | } | |
72fa5997 | 1524 | current->flags &= ~PF_NPROC_EXCEEDED; |
1da177e4 | 1525 | |
f1752eec DH |
1526 | retval = copy_creds(p, clone_flags); |
1527 | if (retval < 0) | |
1528 | goto bad_fork_free; | |
1da177e4 LT |
1529 | |
1530 | /* | |
1531 | * If multiple threads are within copy_process(), then this check | |
1532 | * triggers too late. This doesn't hurt, the check is only there | |
1533 | * to stop root fork bombs. | |
1534 | */ | |
04ec93fe | 1535 | retval = -EAGAIN; |
1da177e4 LT |
1536 | if (nr_threads >= max_threads) |
1537 | goto bad_fork_cleanup_count; | |
1538 | ||
ca74e92b | 1539 | delayacct_tsk_init(p); /* Must remain after dup_task_struct() */ |
514ddb44 DR |
1540 | p->flags &= ~(PF_SUPERPRIV | PF_WQ_WORKER); |
1541 | p->flags |= PF_FORKNOEXEC; | |
1da177e4 LT |
1542 | INIT_LIST_HEAD(&p->children); |
1543 | INIT_LIST_HEAD(&p->sibling); | |
f41d911f | 1544 | rcu_copy_process(p); |
1da177e4 LT |
1545 | p->vfork_done = NULL; |
1546 | spin_lock_init(&p->alloc_lock); | |
1da177e4 | 1547 | |
1da177e4 LT |
1548 | init_sigpending(&p->pending); |
1549 | ||
64861634 MS |
1550 | p->utime = p->stime = p->gtime = 0; |
1551 | p->utimescaled = p->stimescaled = 0; | |
9d7fb042 PZ |
1552 | prev_cputime_init(&p->prev_cputime); |
1553 | ||
6a61671b | 1554 | #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN |
b7ce2277 | 1555 | seqcount_init(&p->vtime_seqcount); |
6a61671b | 1556 | p->vtime_snap = 0; |
7098c1ea | 1557 | p->vtime_snap_whence = VTIME_INACTIVE; |
6a61671b FW |
1558 | #endif |
1559 | ||
a3a2e76c KH |
1560 | #if defined(SPLIT_RSS_COUNTING) |
1561 | memset(&p->rss_stat, 0, sizeof(p->rss_stat)); | |
1562 | #endif | |
172ba844 | 1563 | |
6976675d AV |
1564 | p->default_timer_slack_ns = current->timer_slack_ns; |
1565 | ||
5995477a | 1566 | task_io_accounting_init(&p->ioac); |
1da177e4 LT |
1567 | acct_clear_integrals(p); |
1568 | ||
f06febc9 | 1569 | posix_cpu_timers_init(p); |
1da177e4 | 1570 | |
ccbf62d8 | 1571 | p->start_time = ktime_get_ns(); |
57e0be04 | 1572 | p->real_start_time = ktime_get_boot_ns(); |
1da177e4 | 1573 | p->io_context = NULL; |
1da177e4 | 1574 | p->audit_context = NULL; |
b4f48b63 | 1575 | cgroup_fork(p); |
1da177e4 | 1576 | #ifdef CONFIG_NUMA |
846a16bf | 1577 | p->mempolicy = mpol_dup(p->mempolicy); |
fb0a685c DRO |
1578 | if (IS_ERR(p->mempolicy)) { |
1579 | retval = PTR_ERR(p->mempolicy); | |
1580 | p->mempolicy = NULL; | |
e8604cb4 | 1581 | goto bad_fork_cleanup_threadgroup_lock; |
fb0a685c | 1582 | } |
1da177e4 | 1583 | #endif |
778d3b0f MH |
1584 | #ifdef CONFIG_CPUSETS |
1585 | p->cpuset_mem_spread_rotor = NUMA_NO_NODE; | |
1586 | p->cpuset_slab_spread_rotor = NUMA_NO_NODE; | |
cc9a6c87 | 1587 | seqcount_init(&p->mems_allowed_seq); |
778d3b0f | 1588 | #endif |
de30a2b3 IM |
1589 | #ifdef CONFIG_TRACE_IRQFLAGS |
1590 | p->irq_events = 0; | |
1591 | p->hardirqs_enabled = 0; | |
1592 | p->hardirq_enable_ip = 0; | |
1593 | p->hardirq_enable_event = 0; | |
1594 | p->hardirq_disable_ip = _THIS_IP_; | |
1595 | p->hardirq_disable_event = 0; | |
1596 | p->softirqs_enabled = 1; | |
1597 | p->softirq_enable_ip = _THIS_IP_; | |
1598 | p->softirq_enable_event = 0; | |
1599 | p->softirq_disable_ip = 0; | |
1600 | p->softirq_disable_event = 0; | |
1601 | p->hardirq_context = 0; | |
1602 | p->softirq_context = 0; | |
1603 | #endif | |
8bcbde54 DH |
1604 | |
1605 | p->pagefault_disabled = 0; | |
1606 | ||
fbb9ce95 IM |
1607 | #ifdef CONFIG_LOCKDEP |
1608 | p->lockdep_depth = 0; /* no locks held yet */ | |
1609 | p->curr_chain_key = 0; | |
1610 | p->lockdep_recursion = 0; | |
1611 | #endif | |
1da177e4 | 1612 | |
408894ee IM |
1613 | #ifdef CONFIG_DEBUG_MUTEXES |
1614 | p->blocked_on = NULL; /* not blocked yet */ | |
1615 | #endif | |
cafe5635 KO |
1616 | #ifdef CONFIG_BCACHE |
1617 | p->sequential_io = 0; | |
1618 | p->sequential_io_avg = 0; | |
1619 | #endif | |
0f481406 | 1620 | |
3c90e6e9 | 1621 | /* Perform scheduler related setup. Assign this task to a CPU. */ |
aab03e05 DF |
1622 | retval = sched_fork(clone_flags, p); |
1623 | if (retval) | |
1624 | goto bad_fork_cleanup_policy; | |
6ab423e0 | 1625 | |
cdd6c482 | 1626 | retval = perf_event_init_task(p); |
6ab423e0 PZ |
1627 | if (retval) |
1628 | goto bad_fork_cleanup_policy; | |
fb0a685c DRO |
1629 | retval = audit_alloc(p); |
1630 | if (retval) | |
6c72e350 | 1631 | goto bad_fork_cleanup_perf; |
1da177e4 | 1632 | /* copy all the process information */ |
ab602f79 | 1633 | shm_init_task(p); |
fb0a685c DRO |
1634 | retval = copy_semundo(clone_flags, p); |
1635 | if (retval) | |
1da177e4 | 1636 | goto bad_fork_cleanup_audit; |
fb0a685c DRO |
1637 | retval = copy_files(clone_flags, p); |
1638 | if (retval) | |
1da177e4 | 1639 | goto bad_fork_cleanup_semundo; |
fb0a685c DRO |
1640 | retval = copy_fs(clone_flags, p); |
1641 | if (retval) | |
1da177e4 | 1642 | goto bad_fork_cleanup_files; |
fb0a685c DRO |
1643 | retval = copy_sighand(clone_flags, p); |
1644 | if (retval) | |
1da177e4 | 1645 | goto bad_fork_cleanup_fs; |
fb0a685c DRO |
1646 | retval = copy_signal(clone_flags, p); |
1647 | if (retval) | |
1da177e4 | 1648 | goto bad_fork_cleanup_sighand; |
fb0a685c DRO |
1649 | retval = copy_mm(clone_flags, p); |
1650 | if (retval) | |
1da177e4 | 1651 | goto bad_fork_cleanup_signal; |
fb0a685c DRO |
1652 | retval = copy_namespaces(clone_flags, p); |
1653 | if (retval) | |
d84f4f99 | 1654 | goto bad_fork_cleanup_mm; |
fb0a685c DRO |
1655 | retval = copy_io(clone_flags, p); |
1656 | if (retval) | |
fd0928df | 1657 | goto bad_fork_cleanup_namespaces; |
3033f14a | 1658 | retval = copy_thread_tls(clone_flags, stack_start, stack_size, p, tls); |
1da177e4 | 1659 | if (retval) |
fd0928df | 1660 | goto bad_fork_cleanup_io; |
1da177e4 | 1661 | |
425fb2b4 | 1662 | if (pid != &init_struct_pid) { |
c2b1df2e | 1663 | pid = alloc_pid(p->nsproxy->pid_ns_for_children); |
35f71bc0 MH |
1664 | if (IS_ERR(pid)) { |
1665 | retval = PTR_ERR(pid); | |
0740aa5f | 1666 | goto bad_fork_cleanup_thread; |
35f71bc0 | 1667 | } |
425fb2b4 PE |
1668 | } |
1669 | ||
1da177e4 LT |
1670 | p->set_child_tid = (clone_flags & CLONE_CHILD_SETTID) ? child_tidptr : NULL; |
1671 | /* | |
1672 | * Clear TID on mm_release()? | |
1673 | */ | |
fb0a685c | 1674 | p->clear_child_tid = (clone_flags & CLONE_CHILD_CLEARTID) ? child_tidptr : NULL; |
73c10101 JA |
1675 | #ifdef CONFIG_BLOCK |
1676 | p->plug = NULL; | |
1677 | #endif | |
42b2dd0a | 1678 | #ifdef CONFIG_FUTEX |
8f17d3a5 IM |
1679 | p->robust_list = NULL; |
1680 | #ifdef CONFIG_COMPAT | |
1681 | p->compat_robust_list = NULL; | |
1682 | #endif | |
c87e2837 IM |
1683 | INIT_LIST_HEAD(&p->pi_state_list); |
1684 | p->pi_state_cache = NULL; | |
42b2dd0a | 1685 | #endif |
f9a3879a GM |
1686 | /* |
1687 | * sigaltstack should be cleared when sharing the same VM | |
1688 | */ | |
1689 | if ((clone_flags & (CLONE_VM|CLONE_VFORK)) == CLONE_VM) | |
2a742138 | 1690 | sas_ss_reset(p); |
f9a3879a | 1691 | |
1da177e4 | 1692 | /* |
6580807d ON |
1693 | * Syscall tracing and stepping should be turned off in the |
1694 | * child regardless of CLONE_PTRACE. | |
1da177e4 | 1695 | */ |
6580807d | 1696 | user_disable_single_step(p); |
1da177e4 | 1697 | clear_tsk_thread_flag(p, TIF_SYSCALL_TRACE); |
ed75e8d5 LV |
1698 | #ifdef TIF_SYSCALL_EMU |
1699 | clear_tsk_thread_flag(p, TIF_SYSCALL_EMU); | |
1700 | #endif | |
9745512c | 1701 | clear_all_latency_tracing(p); |
1da177e4 | 1702 | |
1da177e4 | 1703 | /* ok, now we should be set up.. */ |
18c830df ON |
1704 | p->pid = pid_nr(pid); |
1705 | if (clone_flags & CLONE_THREAD) { | |
5f8aadd8 | 1706 | p->exit_signal = -1; |
18c830df ON |
1707 | p->group_leader = current->group_leader; |
1708 | p->tgid = current->tgid; | |
1709 | } else { | |
1710 | if (clone_flags & CLONE_PARENT) | |
1711 | p->exit_signal = current->group_leader->exit_signal; | |
1712 | else | |
1713 | p->exit_signal = (clone_flags & CSIGNAL); | |
1714 | p->group_leader = p; | |
1715 | p->tgid = p->pid; | |
1716 | } | |
5f8aadd8 | 1717 | |
9d823e8f WF |
1718 | p->nr_dirtied = 0; |
1719 | p->nr_dirtied_pause = 128 >> (PAGE_SHIFT - 10); | |
83712358 | 1720 | p->dirty_paused_when = 0; |
9d823e8f | 1721 | |
bb8cbbfe | 1722 | p->pdeath_signal = 0; |
47e65328 | 1723 | INIT_LIST_HEAD(&p->thread_group); |
158e1645 | 1724 | p->task_works = NULL; |
1da177e4 | 1725 | |
568ac888 | 1726 | threadgroup_change_begin(current); |
7e47682e AS |
1727 | /* |
1728 | * Ensure that the cgroup subsystem policies allow the new process to be | |
1729 | * forked. It should be noted the the new process's css_set can be changed | |
1730 | * between here and cgroup_post_fork() if an organisation operation is in | |
1731 | * progress. | |
1732 | */ | |
b53202e6 | 1733 | retval = cgroup_can_fork(p); |
7e47682e AS |
1734 | if (retval) |
1735 | goto bad_fork_free_pid; | |
1736 | ||
18c830df ON |
1737 | /* |
1738 | * Make it visible to the rest of the system, but dont wake it up yet. | |
1739 | * Need tasklist lock for parent etc handling! | |
1740 | */ | |
1da177e4 LT |
1741 | write_lock_irq(&tasklist_lock); |
1742 | ||
1da177e4 | 1743 | /* CLONE_PARENT re-uses the old parent */ |
2d5516cb | 1744 | if (clone_flags & (CLONE_PARENT|CLONE_THREAD)) { |
1da177e4 | 1745 | p->real_parent = current->real_parent; |
2d5516cb ON |
1746 | p->parent_exec_id = current->parent_exec_id; |
1747 | } else { | |
1da177e4 | 1748 | p->real_parent = current; |
2d5516cb ON |
1749 | p->parent_exec_id = current->self_exec_id; |
1750 | } | |
1da177e4 | 1751 | |
3f17da69 | 1752 | spin_lock(¤t->sighand->siglock); |
4a2c7a78 | 1753 | |
dbd95212 KC |
1754 | /* |
1755 | * Copy seccomp details explicitly here, in case they were changed | |
1756 | * before holding sighand lock. | |
1757 | */ | |
1758 | copy_seccomp(p); | |
1759 | ||
4a2c7a78 ON |
1760 | /* |
1761 | * Process group and session signals need to be delivered to just the | |
1762 | * parent before the fork or both the parent and the child after the | |
1763 | * fork. Restart if a signal comes in before we add the new process to | |
1764 | * it's process group. | |
1765 | * A fatal signal pending means that current will exit, so the new | |
1766 | * thread can't slip out of an OOM kill (or normal SIGKILL). | |
fb0a685c | 1767 | */ |
23ff4440 | 1768 | recalc_sigpending(); |
4a2c7a78 ON |
1769 | if (signal_pending(current)) { |
1770 | spin_unlock(¤t->sighand->siglock); | |
1771 | write_unlock_irq(&tasklist_lock); | |
1772 | retval = -ERESTARTNOINTR; | |
7e47682e | 1773 | goto bad_fork_cancel_cgroup; |
4a2c7a78 ON |
1774 | } |
1775 | ||
73b9ebfe | 1776 | if (likely(p->pid)) { |
4b9d33e6 | 1777 | ptrace_init_task(p, (clone_flags & CLONE_PTRACE) || trace); |
73b9ebfe | 1778 | |
81907739 | 1779 | init_task_pid(p, PIDTYPE_PID, pid); |
73b9ebfe | 1780 | if (thread_group_leader(p)) { |
81907739 ON |
1781 | init_task_pid(p, PIDTYPE_PGID, task_pgrp(current)); |
1782 | init_task_pid(p, PIDTYPE_SID, task_session(current)); | |
1783 | ||
1c4042c2 | 1784 | if (is_child_reaper(pid)) { |
17cf22c3 | 1785 | ns_of_pid(pid)->child_reaper = p; |
1c4042c2 EB |
1786 | p->signal->flags |= SIGNAL_UNKILLABLE; |
1787 | } | |
73b9ebfe | 1788 | |
fea9d175 | 1789 | p->signal->leader_pid = pid; |
9c9f4ded | 1790 | p->signal->tty = tty_kref_get(current->signal->tty); |
9cd80bbb | 1791 | list_add_tail(&p->sibling, &p->real_parent->children); |
5e85d4ab | 1792 | list_add_tail_rcu(&p->tasks, &init_task.tasks); |
81907739 ON |
1793 | attach_pid(p, PIDTYPE_PGID); |
1794 | attach_pid(p, PIDTYPE_SID); | |
909ea964 | 1795 | __this_cpu_inc(process_counts); |
80628ca0 ON |
1796 | } else { |
1797 | current->signal->nr_threads++; | |
1798 | atomic_inc(¤t->signal->live); | |
1799 | atomic_inc(¤t->signal->sigcnt); | |
80628ca0 ON |
1800 | list_add_tail_rcu(&p->thread_group, |
1801 | &p->group_leader->thread_group); | |
0c740d0a ON |
1802 | list_add_tail_rcu(&p->thread_node, |
1803 | &p->signal->thread_head); | |
73b9ebfe | 1804 | } |
81907739 | 1805 | attach_pid(p, PIDTYPE_PID); |
73b9ebfe | 1806 | nr_threads++; |
1da177e4 LT |
1807 | } |
1808 | ||
1da177e4 | 1809 | total_forks++; |
3f17da69 | 1810 | spin_unlock(¤t->sighand->siglock); |
4af4206b | 1811 | syscall_tracepoint_update(p); |
1da177e4 | 1812 | write_unlock_irq(&tasklist_lock); |
4af4206b | 1813 | |
c13cf856 | 1814 | proc_fork_connector(p); |
b53202e6 | 1815 | cgroup_post_fork(p); |
c9e75f04 | 1816 | threadgroup_change_end(current); |
cdd6c482 | 1817 | perf_event_fork(p); |
43d2b113 KH |
1818 | |
1819 | trace_task_newtask(p, clone_flags); | |
3ab67966 | 1820 | uprobe_copy_process(p, clone_flags); |
43d2b113 | 1821 | |
1da177e4 LT |
1822 | return p; |
1823 | ||
7e47682e | 1824 | bad_fork_cancel_cgroup: |
b53202e6 | 1825 | cgroup_cancel_fork(p); |
425fb2b4 | 1826 | bad_fork_free_pid: |
568ac888 | 1827 | threadgroup_change_end(current); |
425fb2b4 PE |
1828 | if (pid != &init_struct_pid) |
1829 | free_pid(pid); | |
0740aa5f JS |
1830 | bad_fork_cleanup_thread: |
1831 | exit_thread(p); | |
fd0928df | 1832 | bad_fork_cleanup_io: |
b69f2292 LR |
1833 | if (p->io_context) |
1834 | exit_io_context(p); | |
ab516013 | 1835 | bad_fork_cleanup_namespaces: |
444f378b | 1836 | exit_task_namespaces(p); |
1da177e4 | 1837 | bad_fork_cleanup_mm: |
c9f01245 | 1838 | if (p->mm) |
1da177e4 LT |
1839 | mmput(p->mm); |
1840 | bad_fork_cleanup_signal: | |
4ab6c083 | 1841 | if (!(clone_flags & CLONE_THREAD)) |
1c5354de | 1842 | free_signal_struct(p->signal); |
1da177e4 | 1843 | bad_fork_cleanup_sighand: |
a7e5328a | 1844 | __cleanup_sighand(p->sighand); |
1da177e4 LT |
1845 | bad_fork_cleanup_fs: |
1846 | exit_fs(p); /* blocking */ | |
1847 | bad_fork_cleanup_files: | |
1848 | exit_files(p); /* blocking */ | |
1849 | bad_fork_cleanup_semundo: | |
1850 | exit_sem(p); | |
1851 | bad_fork_cleanup_audit: | |
1852 | audit_free(p); | |
6c72e350 | 1853 | bad_fork_cleanup_perf: |
cdd6c482 | 1854 | perf_event_free_task(p); |
6c72e350 | 1855 | bad_fork_cleanup_policy: |
1da177e4 | 1856 | #ifdef CONFIG_NUMA |
f0be3d32 | 1857 | mpol_put(p->mempolicy); |
e8604cb4 | 1858 | bad_fork_cleanup_threadgroup_lock: |
1da177e4 | 1859 | #endif |
35df17c5 | 1860 | delayacct_tsk_free(p); |
1da177e4 | 1861 | bad_fork_cleanup_count: |
d84f4f99 | 1862 | atomic_dec(&p->cred->user->processes); |
e0e81739 | 1863 | exit_creds(p); |
1da177e4 | 1864 | bad_fork_free: |
68f24b08 | 1865 | put_task_stack(p); |
1da177e4 | 1866 | free_task(p); |
fe7d37d1 ON |
1867 | fork_out: |
1868 | return ERR_PTR(retval); | |
1da177e4 LT |
1869 | } |
1870 | ||
f106eee1 ON |
1871 | static inline void init_idle_pids(struct pid_link *links) |
1872 | { | |
1873 | enum pid_type type; | |
1874 | ||
1875 | for (type = PIDTYPE_PID; type < PIDTYPE_MAX; ++type) { | |
1876 | INIT_HLIST_NODE(&links[type].node); /* not really needed */ | |
1877 | links[type].pid = &init_struct_pid; | |
1878 | } | |
1879 | } | |
1880 | ||
0db0628d | 1881 | struct task_struct *fork_idle(int cpu) |
1da177e4 | 1882 | { |
36c8b586 | 1883 | struct task_struct *task; |
725fc629 AK |
1884 | task = copy_process(CLONE_VM, 0, 0, NULL, &init_struct_pid, 0, 0, |
1885 | cpu_to_node(cpu)); | |
f106eee1 ON |
1886 | if (!IS_ERR(task)) { |
1887 | init_idle_pids(task->pids); | |
753ca4f3 | 1888 | init_idle(task, cpu); |
f106eee1 | 1889 | } |
73b9ebfe | 1890 | |
1da177e4 LT |
1891 | return task; |
1892 | } | |
1893 | ||
1da177e4 LT |
1894 | /* |
1895 | * Ok, this is the main fork-routine. | |
1896 | * | |
1897 | * It copies the process, and if successful kick-starts | |
1898 | * it and waits for it to finish using the VM if required. | |
1899 | */ | |
3033f14a | 1900 | long _do_fork(unsigned long clone_flags, |
1da177e4 | 1901 | unsigned long stack_start, |
1da177e4 LT |
1902 | unsigned long stack_size, |
1903 | int __user *parent_tidptr, | |
3033f14a JT |
1904 | int __user *child_tidptr, |
1905 | unsigned long tls) | |
1da177e4 LT |
1906 | { |
1907 | struct task_struct *p; | |
1908 | int trace = 0; | |
92476d7f | 1909 | long nr; |
1da177e4 | 1910 | |
09a05394 | 1911 | /* |
4b9d33e6 TH |
1912 | * Determine whether and which event to report to ptracer. When |
1913 | * called from kernel_thread or CLONE_UNTRACED is explicitly | |
1914 | * requested, no event is reported; otherwise, report if the event | |
1915 | * for the type of forking is enabled. | |
09a05394 | 1916 | */ |
e80d6661 | 1917 | if (!(clone_flags & CLONE_UNTRACED)) { |
4b9d33e6 TH |
1918 | if (clone_flags & CLONE_VFORK) |
1919 | trace = PTRACE_EVENT_VFORK; | |
1920 | else if ((clone_flags & CSIGNAL) != SIGCHLD) | |
1921 | trace = PTRACE_EVENT_CLONE; | |
1922 | else | |
1923 | trace = PTRACE_EVENT_FORK; | |
1924 | ||
1925 | if (likely(!ptrace_event_enabled(current, trace))) | |
1926 | trace = 0; | |
1927 | } | |
1da177e4 | 1928 | |
62e791c1 | 1929 | p = copy_process(clone_flags, stack_start, stack_size, |
725fc629 | 1930 | child_tidptr, NULL, trace, tls, NUMA_NO_NODE); |
38addce8 | 1931 | add_latent_entropy(); |
1da177e4 LT |
1932 | /* |
1933 | * Do this prior waking up the new thread - the thread pointer | |
1934 | * might get invalid after that point, if the thread exits quickly. | |
1935 | */ | |
1936 | if (!IS_ERR(p)) { | |
1937 | struct completion vfork; | |
4e52365f | 1938 | struct pid *pid; |
1da177e4 | 1939 | |
0a16b607 MD |
1940 | trace_sched_process_fork(current, p); |
1941 | ||
4e52365f MD |
1942 | pid = get_task_pid(p, PIDTYPE_PID); |
1943 | nr = pid_vnr(pid); | |
30e49c26 PE |
1944 | |
1945 | if (clone_flags & CLONE_PARENT_SETTID) | |
1946 | put_user(nr, parent_tidptr); | |
a6f5e063 | 1947 | |
1da177e4 LT |
1948 | if (clone_flags & CLONE_VFORK) { |
1949 | p->vfork_done = &vfork; | |
1950 | init_completion(&vfork); | |
d68b46fe | 1951 | get_task_struct(p); |
1da177e4 LT |
1952 | } |
1953 | ||
3e51e3ed | 1954 | wake_up_new_task(p); |
1da177e4 | 1955 | |
4b9d33e6 TH |
1956 | /* forking complete and child started to run, tell ptracer */ |
1957 | if (unlikely(trace)) | |
4e52365f | 1958 | ptrace_event_pid(trace, pid); |
09a05394 | 1959 | |
1da177e4 | 1960 | if (clone_flags & CLONE_VFORK) { |
d68b46fe | 1961 | if (!wait_for_vfork_done(p, &vfork)) |
4e52365f | 1962 | ptrace_event_pid(PTRACE_EVENT_VFORK_DONE, pid); |
1da177e4 | 1963 | } |
4e52365f MD |
1964 | |
1965 | put_pid(pid); | |
1da177e4 | 1966 | } else { |
92476d7f | 1967 | nr = PTR_ERR(p); |
1da177e4 | 1968 | } |
92476d7f | 1969 | return nr; |
1da177e4 LT |
1970 | } |
1971 | ||
3033f14a JT |
1972 | #ifndef CONFIG_HAVE_COPY_THREAD_TLS |
1973 | /* For compatibility with architectures that call do_fork directly rather than | |
1974 | * using the syscall entry points below. */ | |
1975 | long do_fork(unsigned long clone_flags, | |
1976 | unsigned long stack_start, | |
1977 | unsigned long stack_size, | |
1978 | int __user *parent_tidptr, | |
1979 | int __user *child_tidptr) | |
1980 | { | |
1981 | return _do_fork(clone_flags, stack_start, stack_size, | |
1982 | parent_tidptr, child_tidptr, 0); | |
1983 | } | |
1984 | #endif | |
1985 | ||
2aa3a7f8 AV |
1986 | /* |
1987 | * Create a kernel thread. | |
1988 | */ | |
1989 | pid_t kernel_thread(int (*fn)(void *), void *arg, unsigned long flags) | |
1990 | { | |
3033f14a JT |
1991 | return _do_fork(flags|CLONE_VM|CLONE_UNTRACED, (unsigned long)fn, |
1992 | (unsigned long)arg, NULL, NULL, 0); | |
2aa3a7f8 | 1993 | } |
2aa3a7f8 | 1994 | |
d2125043 AV |
1995 | #ifdef __ARCH_WANT_SYS_FORK |
1996 | SYSCALL_DEFINE0(fork) | |
1997 | { | |
1998 | #ifdef CONFIG_MMU | |
3033f14a | 1999 | return _do_fork(SIGCHLD, 0, 0, NULL, NULL, 0); |
d2125043 AV |
2000 | #else |
2001 | /* can not support in nommu mode */ | |
5d59e182 | 2002 | return -EINVAL; |
d2125043 AV |
2003 | #endif |
2004 | } | |
2005 | #endif | |
2006 | ||
2007 | #ifdef __ARCH_WANT_SYS_VFORK | |
2008 | SYSCALL_DEFINE0(vfork) | |
2009 | { | |
3033f14a JT |
2010 | return _do_fork(CLONE_VFORK | CLONE_VM | SIGCHLD, 0, |
2011 | 0, NULL, NULL, 0); | |
d2125043 AV |
2012 | } |
2013 | #endif | |
2014 | ||
2015 | #ifdef __ARCH_WANT_SYS_CLONE | |
2016 | #ifdef CONFIG_CLONE_BACKWARDS | |
2017 | SYSCALL_DEFINE5(clone, unsigned long, clone_flags, unsigned long, newsp, | |
2018 | int __user *, parent_tidptr, | |
3033f14a | 2019 | unsigned long, tls, |
d2125043 AV |
2020 | int __user *, child_tidptr) |
2021 | #elif defined(CONFIG_CLONE_BACKWARDS2) | |
2022 | SYSCALL_DEFINE5(clone, unsigned long, newsp, unsigned long, clone_flags, | |
2023 | int __user *, parent_tidptr, | |
2024 | int __user *, child_tidptr, | |
3033f14a | 2025 | unsigned long, tls) |
dfa9771a MS |
2026 | #elif defined(CONFIG_CLONE_BACKWARDS3) |
2027 | SYSCALL_DEFINE6(clone, unsigned long, clone_flags, unsigned long, newsp, | |
2028 | int, stack_size, | |
2029 | int __user *, parent_tidptr, | |
2030 | int __user *, child_tidptr, | |
3033f14a | 2031 | unsigned long, tls) |
d2125043 AV |
2032 | #else |
2033 | SYSCALL_DEFINE5(clone, unsigned long, clone_flags, unsigned long, newsp, | |
2034 | int __user *, parent_tidptr, | |
2035 | int __user *, child_tidptr, | |
3033f14a | 2036 | unsigned long, tls) |
d2125043 AV |
2037 | #endif |
2038 | { | |
3033f14a | 2039 | return _do_fork(clone_flags, newsp, 0, parent_tidptr, child_tidptr, tls); |
d2125043 AV |
2040 | } |
2041 | #endif | |
2042 | ||
5fd63b30 RT |
2043 | #ifndef ARCH_MIN_MMSTRUCT_ALIGN |
2044 | #define ARCH_MIN_MMSTRUCT_ALIGN 0 | |
2045 | #endif | |
2046 | ||
51cc5068 | 2047 | static void sighand_ctor(void *data) |
aa1757f9 ON |
2048 | { |
2049 | struct sighand_struct *sighand = data; | |
2050 | ||
a35afb83 | 2051 | spin_lock_init(&sighand->siglock); |
b8fceee1 | 2052 | init_waitqueue_head(&sighand->signalfd_wqh); |
aa1757f9 ON |
2053 | } |
2054 | ||
1da177e4 LT |
2055 | void __init proc_caches_init(void) |
2056 | { | |
2057 | sighand_cachep = kmem_cache_create("sighand_cache", | |
2058 | sizeof(struct sighand_struct), 0, | |
2dff4405 | 2059 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_DESTROY_BY_RCU| |
5d097056 | 2060 | SLAB_NOTRACK|SLAB_ACCOUNT, sighand_ctor); |
1da177e4 LT |
2061 | signal_cachep = kmem_cache_create("signal_cache", |
2062 | sizeof(struct signal_struct), 0, | |
5d097056 VD |
2063 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK|SLAB_ACCOUNT, |
2064 | NULL); | |
20c2df83 | 2065 | files_cachep = kmem_cache_create("files_cache", |
1da177e4 | 2066 | sizeof(struct files_struct), 0, |
5d097056 VD |
2067 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK|SLAB_ACCOUNT, |
2068 | NULL); | |
20c2df83 | 2069 | fs_cachep = kmem_cache_create("fs_cache", |
1da177e4 | 2070 | sizeof(struct fs_struct), 0, |
5d097056 VD |
2071 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK|SLAB_ACCOUNT, |
2072 | NULL); | |
6345d24d LT |
2073 | /* |
2074 | * FIXME! The "sizeof(struct mm_struct)" currently includes the | |
2075 | * whole struct cpumask for the OFFSTACK case. We could change | |
2076 | * this to *only* allocate as much of it as required by the | |
2077 | * maximum number of CPU's we can ever have. The cpumask_allocation | |
2078 | * is at the end of the structure, exactly for that reason. | |
2079 | */ | |
1da177e4 | 2080 | mm_cachep = kmem_cache_create("mm_struct", |
5fd63b30 | 2081 | sizeof(struct mm_struct), ARCH_MIN_MMSTRUCT_ALIGN, |
5d097056 VD |
2082 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK|SLAB_ACCOUNT, |
2083 | NULL); | |
2084 | vm_area_cachep = KMEM_CACHE(vm_area_struct, SLAB_PANIC|SLAB_ACCOUNT); | |
8feae131 | 2085 | mmap_init(); |
66577193 | 2086 | nsproxy_cache_init(); |
1da177e4 | 2087 | } |
cf2e340f | 2088 | |
cf2e340f | 2089 | /* |
9bfb23fc | 2090 | * Check constraints on flags passed to the unshare system call. |
cf2e340f | 2091 | */ |
9bfb23fc | 2092 | static int check_unshare_flags(unsigned long unshare_flags) |
cf2e340f | 2093 | { |
9bfb23fc ON |
2094 | if (unshare_flags & ~(CLONE_THREAD|CLONE_FS|CLONE_NEWNS|CLONE_SIGHAND| |
2095 | CLONE_VM|CLONE_FILES|CLONE_SYSVSEM| | |
50804fe3 | 2096 | CLONE_NEWUTS|CLONE_NEWIPC|CLONE_NEWNET| |
a79a908f | 2097 | CLONE_NEWUSER|CLONE_NEWPID|CLONE_NEWCGROUP)) |
9bfb23fc | 2098 | return -EINVAL; |
cf2e340f | 2099 | /* |
12c641ab EB |
2100 | * Not implemented, but pretend it works if there is nothing |
2101 | * to unshare. Note that unsharing the address space or the | |
2102 | * signal handlers also need to unshare the signal queues (aka | |
2103 | * CLONE_THREAD). | |
cf2e340f | 2104 | */ |
9bfb23fc | 2105 | if (unshare_flags & (CLONE_THREAD | CLONE_SIGHAND | CLONE_VM)) { |
12c641ab EB |
2106 | if (!thread_group_empty(current)) |
2107 | return -EINVAL; | |
2108 | } | |
2109 | if (unshare_flags & (CLONE_SIGHAND | CLONE_VM)) { | |
2110 | if (atomic_read(¤t->sighand->count) > 1) | |
2111 | return -EINVAL; | |
2112 | } | |
2113 | if (unshare_flags & CLONE_VM) { | |
2114 | if (!current_is_single_threaded()) | |
9bfb23fc ON |
2115 | return -EINVAL; |
2116 | } | |
cf2e340f JD |
2117 | |
2118 | return 0; | |
2119 | } | |
2120 | ||
2121 | /* | |
99d1419d | 2122 | * Unshare the filesystem structure if it is being shared |
cf2e340f JD |
2123 | */ |
2124 | static int unshare_fs(unsigned long unshare_flags, struct fs_struct **new_fsp) | |
2125 | { | |
2126 | struct fs_struct *fs = current->fs; | |
2127 | ||
498052bb AV |
2128 | if (!(unshare_flags & CLONE_FS) || !fs) |
2129 | return 0; | |
2130 | ||
2131 | /* don't need lock here; in the worst case we'll do useless copy */ | |
2132 | if (fs->users == 1) | |
2133 | return 0; | |
2134 | ||
2135 | *new_fsp = copy_fs_struct(fs); | |
2136 | if (!*new_fsp) | |
2137 | return -ENOMEM; | |
cf2e340f JD |
2138 | |
2139 | return 0; | |
2140 | } | |
2141 | ||
cf2e340f | 2142 | /* |
a016f338 | 2143 | * Unshare file descriptor table if it is being shared |
cf2e340f JD |
2144 | */ |
2145 | static int unshare_fd(unsigned long unshare_flags, struct files_struct **new_fdp) | |
2146 | { | |
2147 | struct files_struct *fd = current->files; | |
a016f338 | 2148 | int error = 0; |
cf2e340f JD |
2149 | |
2150 | if ((unshare_flags & CLONE_FILES) && | |
a016f338 JD |
2151 | (fd && atomic_read(&fd->count) > 1)) { |
2152 | *new_fdp = dup_fd(fd, &error); | |
2153 | if (!*new_fdp) | |
2154 | return error; | |
2155 | } | |
cf2e340f JD |
2156 | |
2157 | return 0; | |
2158 | } | |
2159 | ||
cf2e340f JD |
2160 | /* |
2161 | * unshare allows a process to 'unshare' part of the process | |
2162 | * context which was originally shared using clone. copy_* | |
2163 | * functions used by do_fork() cannot be used here directly | |
2164 | * because they modify an inactive task_struct that is being | |
2165 | * constructed. Here we are modifying the current, active, | |
2166 | * task_struct. | |
2167 | */ | |
6559eed8 | 2168 | SYSCALL_DEFINE1(unshare, unsigned long, unshare_flags) |
cf2e340f | 2169 | { |
cf2e340f | 2170 | struct fs_struct *fs, *new_fs = NULL; |
cf2e340f | 2171 | struct files_struct *fd, *new_fd = NULL; |
b2e0d987 | 2172 | struct cred *new_cred = NULL; |
cf7b708c | 2173 | struct nsproxy *new_nsproxy = NULL; |
9edff4ab | 2174 | int do_sysvsem = 0; |
9bfb23fc | 2175 | int err; |
cf2e340f | 2176 | |
b2e0d987 | 2177 | /* |
faf00da5 EB |
2178 | * If unsharing a user namespace must also unshare the thread group |
2179 | * and unshare the filesystem root and working directories. | |
b2e0d987 EB |
2180 | */ |
2181 | if (unshare_flags & CLONE_NEWUSER) | |
e66eded8 | 2182 | unshare_flags |= CLONE_THREAD | CLONE_FS; |
50804fe3 EB |
2183 | /* |
2184 | * If unsharing vm, must also unshare signal handlers. | |
2185 | */ | |
2186 | if (unshare_flags & CLONE_VM) | |
2187 | unshare_flags |= CLONE_SIGHAND; | |
12c641ab EB |
2188 | /* |
2189 | * If unsharing a signal handlers, must also unshare the signal queues. | |
2190 | */ | |
2191 | if (unshare_flags & CLONE_SIGHAND) | |
2192 | unshare_flags |= CLONE_THREAD; | |
9bfb23fc ON |
2193 | /* |
2194 | * If unsharing namespace, must also unshare filesystem information. | |
2195 | */ | |
2196 | if (unshare_flags & CLONE_NEWNS) | |
2197 | unshare_flags |= CLONE_FS; | |
50804fe3 EB |
2198 | |
2199 | err = check_unshare_flags(unshare_flags); | |
2200 | if (err) | |
2201 | goto bad_unshare_out; | |
6013f67f MS |
2202 | /* |
2203 | * CLONE_NEWIPC must also detach from the undolist: after switching | |
2204 | * to a new ipc namespace, the semaphore arrays from the old | |
2205 | * namespace are unreachable. | |
2206 | */ | |
2207 | if (unshare_flags & (CLONE_NEWIPC|CLONE_SYSVSEM)) | |
9edff4ab | 2208 | do_sysvsem = 1; |
fb0a685c DRO |
2209 | err = unshare_fs(unshare_flags, &new_fs); |
2210 | if (err) | |
9bfb23fc | 2211 | goto bad_unshare_out; |
fb0a685c DRO |
2212 | err = unshare_fd(unshare_flags, &new_fd); |
2213 | if (err) | |
9bfb23fc | 2214 | goto bad_unshare_cleanup_fs; |
b2e0d987 | 2215 | err = unshare_userns(unshare_flags, &new_cred); |
fb0a685c | 2216 | if (err) |
9edff4ab | 2217 | goto bad_unshare_cleanup_fd; |
b2e0d987 EB |
2218 | err = unshare_nsproxy_namespaces(unshare_flags, &new_nsproxy, |
2219 | new_cred, new_fs); | |
2220 | if (err) | |
2221 | goto bad_unshare_cleanup_cred; | |
c0b2fc31 | 2222 | |
b2e0d987 | 2223 | if (new_fs || new_fd || do_sysvsem || new_cred || new_nsproxy) { |
9edff4ab MS |
2224 | if (do_sysvsem) { |
2225 | /* | |
2226 | * CLONE_SYSVSEM is equivalent to sys_exit(). | |
2227 | */ | |
2228 | exit_sem(current); | |
2229 | } | |
ab602f79 JM |
2230 | if (unshare_flags & CLONE_NEWIPC) { |
2231 | /* Orphan segments in old ns (see sem above). */ | |
2232 | exit_shm(current); | |
2233 | shm_init_task(current); | |
2234 | } | |
ab516013 | 2235 | |
6f977e6b | 2236 | if (new_nsproxy) |
cf7b708c | 2237 | switch_task_namespaces(current, new_nsproxy); |
cf2e340f | 2238 | |
cf7b708c PE |
2239 | task_lock(current); |
2240 | ||
cf2e340f JD |
2241 | if (new_fs) { |
2242 | fs = current->fs; | |
2a4419b5 | 2243 | spin_lock(&fs->lock); |
cf2e340f | 2244 | current->fs = new_fs; |
498052bb AV |
2245 | if (--fs->users) |
2246 | new_fs = NULL; | |
2247 | else | |
2248 | new_fs = fs; | |
2a4419b5 | 2249 | spin_unlock(&fs->lock); |
cf2e340f JD |
2250 | } |
2251 | ||
cf2e340f JD |
2252 | if (new_fd) { |
2253 | fd = current->files; | |
2254 | current->files = new_fd; | |
2255 | new_fd = fd; | |
2256 | } | |
2257 | ||
2258 | task_unlock(current); | |
b2e0d987 EB |
2259 | |
2260 | if (new_cred) { | |
2261 | /* Install the new user namespace */ | |
2262 | commit_creds(new_cred); | |
2263 | new_cred = NULL; | |
2264 | } | |
cf2e340f JD |
2265 | } |
2266 | ||
b2e0d987 EB |
2267 | bad_unshare_cleanup_cred: |
2268 | if (new_cred) | |
2269 | put_cred(new_cred); | |
cf2e340f JD |
2270 | bad_unshare_cleanup_fd: |
2271 | if (new_fd) | |
2272 | put_files_struct(new_fd); | |
2273 | ||
cf2e340f JD |
2274 | bad_unshare_cleanup_fs: |
2275 | if (new_fs) | |
498052bb | 2276 | free_fs_struct(new_fs); |
cf2e340f | 2277 | |
cf2e340f JD |
2278 | bad_unshare_out: |
2279 | return err; | |
2280 | } | |
3b125388 AV |
2281 | |
2282 | /* | |
2283 | * Helper to unshare the files of the current task. | |
2284 | * We don't want to expose copy_files internals to | |
2285 | * the exec layer of the kernel. | |
2286 | */ | |
2287 | ||
2288 | int unshare_files(struct files_struct **displaced) | |
2289 | { | |
2290 | struct task_struct *task = current; | |
50704516 | 2291 | struct files_struct *copy = NULL; |
3b125388 AV |
2292 | int error; |
2293 | ||
2294 | error = unshare_fd(CLONE_FILES, ©); | |
2295 | if (error || !copy) { | |
2296 | *displaced = NULL; | |
2297 | return error; | |
2298 | } | |
2299 | *displaced = task->files; | |
2300 | task_lock(task); | |
2301 | task->files = copy; | |
2302 | task_unlock(task); | |
2303 | return 0; | |
2304 | } | |
16db3d3f HS |
2305 | |
2306 | int sysctl_max_threads(struct ctl_table *table, int write, | |
2307 | void __user *buffer, size_t *lenp, loff_t *ppos) | |
2308 | { | |
2309 | struct ctl_table t; | |
2310 | int ret; | |
2311 | int threads = max_threads; | |
2312 | int min = MIN_THREADS; | |
2313 | int max = MAX_THREADS; | |
2314 | ||
2315 | t = *table; | |
2316 | t.data = &threads; | |
2317 | t.extra1 = &min; | |
2318 | t.extra2 = &max; | |
2319 | ||
2320 | ret = proc_dointvec_minmax(&t, write, buffer, lenp, ppos); | |
2321 | if (ret || !write) | |
2322 | return ret; | |
2323 | ||
2324 | set_max_threads(threads); | |
2325 | ||
2326 | return 0; | |
2327 | } |