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> |
1da177e4 LT |
78 | |
79 | #include <asm/pgtable.h> | |
80 | #include <asm/pgalloc.h> | |
81 | #include <asm/uaccess.h> | |
82 | #include <asm/mmu_context.h> | |
83 | #include <asm/cacheflush.h> | |
84 | #include <asm/tlbflush.h> | |
85 | ||
ad8d75ff SR |
86 | #include <trace/events/sched.h> |
87 | ||
43d2b113 KH |
88 | #define CREATE_TRACE_POINTS |
89 | #include <trace/events/task.h> | |
90 | ||
ac1b398d HS |
91 | /* |
92 | * Minimum number of threads to boot the kernel | |
93 | */ | |
94 | #define MIN_THREADS 20 | |
95 | ||
96 | /* | |
97 | * Maximum number of threads | |
98 | */ | |
99 | #define MAX_THREADS FUTEX_TID_MASK | |
100 | ||
1da177e4 LT |
101 | /* |
102 | * Protected counters by write_lock_irq(&tasklist_lock) | |
103 | */ | |
104 | unsigned long total_forks; /* Handle normal Linux uptimes. */ | |
fb0a685c | 105 | int nr_threads; /* The idle threads do not count.. */ |
1da177e4 LT |
106 | |
107 | int max_threads; /* tunable limit on nr_threads */ | |
108 | ||
109 | DEFINE_PER_CPU(unsigned long, process_counts) = 0; | |
110 | ||
c59923a1 | 111 | __cacheline_aligned DEFINE_RWLOCK(tasklist_lock); /* outer */ |
db1466b3 PM |
112 | |
113 | #ifdef CONFIG_PROVE_RCU | |
114 | int lockdep_tasklist_lock_is_held(void) | |
115 | { | |
116 | return lockdep_is_held(&tasklist_lock); | |
117 | } | |
118 | EXPORT_SYMBOL_GPL(lockdep_tasklist_lock_is_held); | |
119 | #endif /* #ifdef CONFIG_PROVE_RCU */ | |
1da177e4 LT |
120 | |
121 | int nr_processes(void) | |
122 | { | |
123 | int cpu; | |
124 | int total = 0; | |
125 | ||
1d510750 | 126 | for_each_possible_cpu(cpu) |
1da177e4 LT |
127 | total += per_cpu(process_counts, cpu); |
128 | ||
129 | return total; | |
130 | } | |
131 | ||
f19b9f74 AM |
132 | void __weak arch_release_task_struct(struct task_struct *tsk) |
133 | { | |
134 | } | |
135 | ||
f5e10287 | 136 | #ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR |
e18b890b | 137 | static struct kmem_cache *task_struct_cachep; |
41101809 TG |
138 | |
139 | static inline struct task_struct *alloc_task_struct_node(int node) | |
140 | { | |
141 | return kmem_cache_alloc_node(task_struct_cachep, GFP_KERNEL, node); | |
142 | } | |
143 | ||
41101809 TG |
144 | static inline void free_task_struct(struct task_struct *tsk) |
145 | { | |
41101809 TG |
146 | kmem_cache_free(task_struct_cachep, tsk); |
147 | } | |
1da177e4 LT |
148 | #endif |
149 | ||
f19b9f74 AM |
150 | void __weak arch_release_thread_info(struct thread_info *ti) |
151 | { | |
152 | } | |
153 | ||
f5e10287 | 154 | #ifndef CONFIG_ARCH_THREAD_INFO_ALLOCATOR |
41101809 | 155 | |
0d15d74a TG |
156 | /* |
157 | * Allocate pages if THREAD_SIZE is >= PAGE_SIZE, otherwise use a | |
158 | * kmemcache based allocator. | |
159 | */ | |
160 | # if THREAD_SIZE >= PAGE_SIZE | |
b6a84016 ED |
161 | static struct thread_info *alloc_thread_info_node(struct task_struct *tsk, |
162 | int node) | |
b69c49b7 | 163 | { |
52383431 VD |
164 | struct page *page = alloc_kmem_pages_node(node, THREADINFO_GFP, |
165 | THREAD_SIZE_ORDER); | |
b6a84016 ED |
166 | |
167 | return page ? page_address(page) : NULL; | |
b69c49b7 FT |
168 | } |
169 | ||
170 | static inline void free_thread_info(struct thread_info *ti) | |
171 | { | |
52383431 | 172 | free_kmem_pages((unsigned long)ti, THREAD_SIZE_ORDER); |
b69c49b7 | 173 | } |
0d15d74a TG |
174 | # else |
175 | static struct kmem_cache *thread_info_cache; | |
176 | ||
177 | static struct thread_info *alloc_thread_info_node(struct task_struct *tsk, | |
178 | int node) | |
179 | { | |
180 | return kmem_cache_alloc_node(thread_info_cache, THREADINFO_GFP, node); | |
181 | } | |
182 | ||
183 | static void free_thread_info(struct thread_info *ti) | |
184 | { | |
0d15d74a TG |
185 | kmem_cache_free(thread_info_cache, ti); |
186 | } | |
187 | ||
188 | void thread_info_cache_init(void) | |
189 | { | |
190 | thread_info_cache = kmem_cache_create("thread_info", THREAD_SIZE, | |
191 | THREAD_SIZE, 0, NULL); | |
192 | BUG_ON(thread_info_cache == NULL); | |
193 | } | |
194 | # endif | |
b69c49b7 FT |
195 | #endif |
196 | ||
1da177e4 | 197 | /* SLAB cache for signal_struct structures (tsk->signal) */ |
e18b890b | 198 | static struct kmem_cache *signal_cachep; |
1da177e4 LT |
199 | |
200 | /* SLAB cache for sighand_struct structures (tsk->sighand) */ | |
e18b890b | 201 | struct kmem_cache *sighand_cachep; |
1da177e4 LT |
202 | |
203 | /* SLAB cache for files_struct structures (tsk->files) */ | |
e18b890b | 204 | struct kmem_cache *files_cachep; |
1da177e4 LT |
205 | |
206 | /* SLAB cache for fs_struct structures (tsk->fs) */ | |
e18b890b | 207 | struct kmem_cache *fs_cachep; |
1da177e4 LT |
208 | |
209 | /* SLAB cache for vm_area_struct structures */ | |
e18b890b | 210 | struct kmem_cache *vm_area_cachep; |
1da177e4 LT |
211 | |
212 | /* SLAB cache for mm_struct structures (tsk->mm) */ | |
e18b890b | 213 | static struct kmem_cache *mm_cachep; |
1da177e4 | 214 | |
c6a7f572 KM |
215 | static void account_kernel_stack(struct thread_info *ti, int account) |
216 | { | |
217 | struct zone *zone = page_zone(virt_to_page(ti)); | |
218 | ||
219 | mod_zone_page_state(zone, NR_KERNEL_STACK, account); | |
220 | } | |
221 | ||
1da177e4 LT |
222 | void free_task(struct task_struct *tsk) |
223 | { | |
c6a7f572 | 224 | account_kernel_stack(tsk->stack, -1); |
f19b9f74 | 225 | arch_release_thread_info(tsk->stack); |
f7e4217b | 226 | free_thread_info(tsk->stack); |
23f78d4a | 227 | rt_mutex_debug_task_free(tsk); |
fb52607a | 228 | ftrace_graph_exit_task(tsk); |
e2cfabdf | 229 | put_seccomp_filter(tsk); |
f19b9f74 | 230 | arch_release_task_struct(tsk); |
1da177e4 LT |
231 | free_task_struct(tsk); |
232 | } | |
233 | EXPORT_SYMBOL(free_task); | |
234 | ||
ea6d290c ON |
235 | static inline void free_signal_struct(struct signal_struct *sig) |
236 | { | |
97101eb4 | 237 | taskstats_tgid_free(sig); |
1c5354de | 238 | sched_autogroup_exit(sig); |
ea6d290c ON |
239 | kmem_cache_free(signal_cachep, sig); |
240 | } | |
241 | ||
242 | static inline void put_signal_struct(struct signal_struct *sig) | |
243 | { | |
1c5354de | 244 | if (atomic_dec_and_test(&sig->sigcnt)) |
ea6d290c ON |
245 | free_signal_struct(sig); |
246 | } | |
247 | ||
158d9ebd | 248 | void __put_task_struct(struct task_struct *tsk) |
1da177e4 | 249 | { |
270f722d | 250 | WARN_ON(!tsk->exit_state); |
1da177e4 LT |
251 | WARN_ON(atomic_read(&tsk->usage)); |
252 | WARN_ON(tsk == current); | |
253 | ||
2e91fa7f | 254 | cgroup_free(tsk); |
156654f4 | 255 | task_numa_free(tsk); |
1a2a4d06 | 256 | security_task_free(tsk); |
e0e81739 | 257 | exit_creds(tsk); |
35df17c5 | 258 | delayacct_tsk_free(tsk); |
ea6d290c | 259 | put_signal_struct(tsk->signal); |
1da177e4 LT |
260 | |
261 | if (!profile_handoff_task(tsk)) | |
262 | free_task(tsk); | |
263 | } | |
77c100c8 | 264 | EXPORT_SYMBOL_GPL(__put_task_struct); |
1da177e4 | 265 | |
6c0a9fa6 | 266 | void __init __weak arch_task_cache_init(void) { } |
61c4628b | 267 | |
ff691f6e HS |
268 | /* |
269 | * set_max_threads | |
270 | */ | |
16db3d3f | 271 | static void set_max_threads(unsigned int max_threads_suggested) |
ff691f6e | 272 | { |
ac1b398d | 273 | u64 threads; |
ff691f6e HS |
274 | |
275 | /* | |
ac1b398d HS |
276 | * The number of threads shall be limited such that the thread |
277 | * structures may only consume a small part of the available memory. | |
ff691f6e | 278 | */ |
ac1b398d HS |
279 | if (fls64(totalram_pages) + fls64(PAGE_SIZE) > 64) |
280 | threads = MAX_THREADS; | |
281 | else | |
282 | threads = div64_u64((u64) totalram_pages * (u64) PAGE_SIZE, | |
283 | (u64) THREAD_SIZE * 8UL); | |
284 | ||
16db3d3f HS |
285 | if (threads > max_threads_suggested) |
286 | threads = max_threads_suggested; | |
287 | ||
ac1b398d | 288 | max_threads = clamp_t(u64, threads, MIN_THREADS, MAX_THREADS); |
ff691f6e HS |
289 | } |
290 | ||
5aaeb5c0 IM |
291 | #ifdef CONFIG_ARCH_WANTS_DYNAMIC_TASK_STRUCT |
292 | /* Initialized by the architecture: */ | |
293 | int arch_task_struct_size __read_mostly; | |
294 | #endif | |
0c8c0f03 | 295 | |
ff691f6e | 296 | void __init fork_init(void) |
1da177e4 | 297 | { |
f5e10287 | 298 | #ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR |
1da177e4 LT |
299 | #ifndef ARCH_MIN_TASKALIGN |
300 | #define ARCH_MIN_TASKALIGN L1_CACHE_BYTES | |
301 | #endif | |
302 | /* create a slab on which task_structs can be allocated */ | |
5d097056 VD |
303 | task_struct_cachep = kmem_cache_create("task_struct", |
304 | arch_task_struct_size, ARCH_MIN_TASKALIGN, | |
305 | SLAB_PANIC|SLAB_NOTRACK|SLAB_ACCOUNT, NULL); | |
1da177e4 LT |
306 | #endif |
307 | ||
61c4628b SS |
308 | /* do the arch specific task caches init */ |
309 | arch_task_cache_init(); | |
310 | ||
16db3d3f | 311 | set_max_threads(MAX_THREADS); |
1da177e4 LT |
312 | |
313 | init_task.signal->rlim[RLIMIT_NPROC].rlim_cur = max_threads/2; | |
314 | init_task.signal->rlim[RLIMIT_NPROC].rlim_max = max_threads/2; | |
315 | init_task.signal->rlim[RLIMIT_SIGPENDING] = | |
316 | init_task.signal->rlim[RLIMIT_NPROC]; | |
317 | } | |
318 | ||
52f5684c | 319 | int __weak arch_dup_task_struct(struct task_struct *dst, |
61c4628b SS |
320 | struct task_struct *src) |
321 | { | |
322 | *dst = *src; | |
323 | return 0; | |
324 | } | |
325 | ||
d4311ff1 AT |
326 | void set_task_stack_end_magic(struct task_struct *tsk) |
327 | { | |
328 | unsigned long *stackend; | |
329 | ||
330 | stackend = end_of_stack(tsk); | |
331 | *stackend = STACK_END_MAGIC; /* for overflow detection */ | |
332 | } | |
333 | ||
1da177e4 LT |
334 | static struct task_struct *dup_task_struct(struct task_struct *orig) |
335 | { | |
336 | struct task_struct *tsk; | |
337 | struct thread_info *ti; | |
207205a2 | 338 | int node = tsk_fork_get_node(orig); |
3e26c149 | 339 | int err; |
1da177e4 | 340 | |
504f52b5 | 341 | tsk = alloc_task_struct_node(node); |
1da177e4 LT |
342 | if (!tsk) |
343 | return NULL; | |
344 | ||
b6a84016 | 345 | ti = alloc_thread_info_node(tsk, node); |
f19b9f74 AM |
346 | if (!ti) |
347 | goto free_tsk; | |
1da177e4 | 348 | |
fb0a685c | 349 | err = arch_dup_task_struct(tsk, orig); |
164c33c6 | 350 | if (err) |
f19b9f74 | 351 | goto free_ti; |
164c33c6 | 352 | |
87bec58a | 353 | tsk->stack = ti; |
dbd95212 KC |
354 | #ifdef CONFIG_SECCOMP |
355 | /* | |
356 | * We must handle setting up seccomp filters once we're under | |
357 | * the sighand lock in case orig has changed between now and | |
358 | * then. Until then, filter must be NULL to avoid messing up | |
359 | * the usage counts on the error path calling free_task. | |
360 | */ | |
361 | tsk->seccomp.filter = NULL; | |
362 | #endif | |
87bec58a AM |
363 | |
364 | setup_thread_stack(tsk, orig); | |
8e7cac79 | 365 | clear_user_return_notifier(tsk); |
f26f9aff | 366 | clear_tsk_need_resched(tsk); |
d4311ff1 | 367 | set_task_stack_end_magic(tsk); |
1da177e4 | 368 | |
0a425405 AV |
369 | #ifdef CONFIG_CC_STACKPROTECTOR |
370 | tsk->stack_canary = get_random_int(); | |
371 | #endif | |
372 | ||
fb0a685c DRO |
373 | /* |
374 | * One for us, one for whoever does the "release_task()" (usually | |
375 | * parent) | |
376 | */ | |
377 | atomic_set(&tsk->usage, 2); | |
6c5c9341 | 378 | #ifdef CONFIG_BLK_DEV_IO_TRACE |
2056a782 | 379 | tsk->btrace_seq = 0; |
6c5c9341 | 380 | #endif |
a0aa7f68 | 381 | tsk->splice_pipe = NULL; |
5640f768 | 382 | tsk->task_frag.page = NULL; |
093e5840 | 383 | tsk->wake_q.next = NULL; |
c6a7f572 KM |
384 | |
385 | account_kernel_stack(ti, 1); | |
386 | ||
1da177e4 | 387 | return tsk; |
61c4628b | 388 | |
f19b9f74 | 389 | free_ti: |
61c4628b | 390 | free_thread_info(ti); |
f19b9f74 | 391 | free_tsk: |
61c4628b SS |
392 | free_task_struct(tsk); |
393 | return NULL; | |
1da177e4 LT |
394 | } |
395 | ||
396 | #ifdef CONFIG_MMU | |
a39bc516 | 397 | static int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm) |
1da177e4 | 398 | { |
297c5eee | 399 | struct vm_area_struct *mpnt, *tmp, *prev, **pprev; |
1da177e4 LT |
400 | struct rb_node **rb_link, *rb_parent; |
401 | int retval; | |
402 | unsigned long charge; | |
1da177e4 | 403 | |
32cdba1e | 404 | uprobe_start_dup_mmap(); |
1da177e4 | 405 | down_write(&oldmm->mmap_sem); |
ec8c0446 | 406 | flush_cache_dup_mm(oldmm); |
f8ac4ec9 | 407 | uprobe_dup_mmap(oldmm, mm); |
ad339451 IM |
408 | /* |
409 | * Not linked in yet - no deadlock potential: | |
410 | */ | |
411 | down_write_nested(&mm->mmap_sem, SINGLE_DEPTH_NESTING); | |
7ee78232 | 412 | |
90f31d0e KK |
413 | /* No ordering required: file already has been exposed. */ |
414 | RCU_INIT_POINTER(mm->exe_file, get_mm_exe_file(oldmm)); | |
415 | ||
4f7d4614 | 416 | mm->total_vm = oldmm->total_vm; |
84638335 | 417 | mm->data_vm = oldmm->data_vm; |
4f7d4614 VD |
418 | mm->exec_vm = oldmm->exec_vm; |
419 | mm->stack_vm = oldmm->stack_vm; | |
420 | ||
1da177e4 LT |
421 | rb_link = &mm->mm_rb.rb_node; |
422 | rb_parent = NULL; | |
423 | pprev = &mm->mmap; | |
f8af4da3 | 424 | retval = ksm_fork(mm, oldmm); |
ba76149f AA |
425 | if (retval) |
426 | goto out; | |
427 | retval = khugepaged_fork(mm, oldmm); | |
f8af4da3 HD |
428 | if (retval) |
429 | goto out; | |
1da177e4 | 430 | |
297c5eee | 431 | prev = NULL; |
fd3e42fc | 432 | for (mpnt = oldmm->mmap; mpnt; mpnt = mpnt->vm_next) { |
1da177e4 LT |
433 | struct file *file; |
434 | ||
435 | if (mpnt->vm_flags & VM_DONTCOPY) { | |
84638335 | 436 | vm_stat_account(mm, mpnt->vm_flags, -vma_pages(mpnt)); |
1da177e4 LT |
437 | continue; |
438 | } | |
439 | charge = 0; | |
440 | if (mpnt->vm_flags & VM_ACCOUNT) { | |
b2412b7f HS |
441 | unsigned long len = vma_pages(mpnt); |
442 | ||
191c5424 | 443 | if (security_vm_enough_memory_mm(oldmm, len)) /* sic */ |
1da177e4 LT |
444 | goto fail_nomem; |
445 | charge = len; | |
446 | } | |
e94b1766 | 447 | tmp = kmem_cache_alloc(vm_area_cachep, GFP_KERNEL); |
1da177e4 LT |
448 | if (!tmp) |
449 | goto fail_nomem; | |
450 | *tmp = *mpnt; | |
5beb4930 | 451 | INIT_LIST_HEAD(&tmp->anon_vma_chain); |
ef0855d3 ON |
452 | retval = vma_dup_policy(mpnt, tmp); |
453 | if (retval) | |
1da177e4 | 454 | goto fail_nomem_policy; |
a247c3a9 | 455 | tmp->vm_mm = mm; |
5beb4930 RR |
456 | if (anon_vma_fork(tmp, mpnt)) |
457 | goto fail_nomem_anon_vma_fork; | |
de60f5f1 EM |
458 | tmp->vm_flags &= |
459 | ~(VM_LOCKED|VM_LOCKONFAULT|VM_UFFD_MISSING|VM_UFFD_WP); | |
297c5eee | 460 | tmp->vm_next = tmp->vm_prev = NULL; |
745f234b | 461 | tmp->vm_userfaultfd_ctx = NULL_VM_UFFD_CTX; |
1da177e4 LT |
462 | file = tmp->vm_file; |
463 | if (file) { | |
496ad9aa | 464 | struct inode *inode = file_inode(file); |
b88ed205 HD |
465 | struct address_space *mapping = file->f_mapping; |
466 | ||
1da177e4 LT |
467 | get_file(file); |
468 | if (tmp->vm_flags & VM_DENYWRITE) | |
469 | atomic_dec(&inode->i_writecount); | |
83cde9e8 | 470 | i_mmap_lock_write(mapping); |
b88ed205 | 471 | if (tmp->vm_flags & VM_SHARED) |
4bb5f5d9 | 472 | atomic_inc(&mapping->i_mmap_writable); |
b88ed205 HD |
473 | flush_dcache_mmap_lock(mapping); |
474 | /* insert tmp into the share list, just after mpnt */ | |
27ba0644 KS |
475 | vma_interval_tree_insert_after(tmp, mpnt, |
476 | &mapping->i_mmap); | |
b88ed205 | 477 | flush_dcache_mmap_unlock(mapping); |
83cde9e8 | 478 | i_mmap_unlock_write(mapping); |
1da177e4 LT |
479 | } |
480 | ||
a1e78772 MG |
481 | /* |
482 | * Clear hugetlb-related page reserves for children. This only | |
483 | * affects MAP_PRIVATE mappings. Faults generated by the child | |
484 | * are not guaranteed to succeed, even if read-only | |
485 | */ | |
486 | if (is_vm_hugetlb_page(tmp)) | |
487 | reset_vma_resv_huge_pages(tmp); | |
488 | ||
1da177e4 | 489 | /* |
7ee78232 | 490 | * Link in the new vma and copy the page table entries. |
1da177e4 | 491 | */ |
1da177e4 LT |
492 | *pprev = tmp; |
493 | pprev = &tmp->vm_next; | |
297c5eee LT |
494 | tmp->vm_prev = prev; |
495 | prev = tmp; | |
1da177e4 LT |
496 | |
497 | __vma_link_rb(mm, tmp, rb_link, rb_parent); | |
498 | rb_link = &tmp->vm_rb.rb_right; | |
499 | rb_parent = &tmp->vm_rb; | |
500 | ||
501 | mm->map_count++; | |
0b0db14c | 502 | retval = copy_page_range(mm, oldmm, mpnt); |
1da177e4 LT |
503 | |
504 | if (tmp->vm_ops && tmp->vm_ops->open) | |
505 | tmp->vm_ops->open(tmp); | |
506 | ||
507 | if (retval) | |
508 | goto out; | |
509 | } | |
d6dd61c8 JF |
510 | /* a new mm has just been created */ |
511 | arch_dup_mmap(oldmm, mm); | |
1da177e4 | 512 | retval = 0; |
1da177e4 | 513 | out: |
7ee78232 | 514 | up_write(&mm->mmap_sem); |
fd3e42fc | 515 | flush_tlb_mm(oldmm); |
1da177e4 | 516 | up_write(&oldmm->mmap_sem); |
32cdba1e | 517 | uprobe_end_dup_mmap(); |
1da177e4 | 518 | return retval; |
5beb4930 | 519 | fail_nomem_anon_vma_fork: |
ef0855d3 | 520 | mpol_put(vma_policy(tmp)); |
1da177e4 LT |
521 | fail_nomem_policy: |
522 | kmem_cache_free(vm_area_cachep, tmp); | |
523 | fail_nomem: | |
524 | retval = -ENOMEM; | |
525 | vm_unacct_memory(charge); | |
526 | goto out; | |
527 | } | |
528 | ||
fb0a685c | 529 | static inline int mm_alloc_pgd(struct mm_struct *mm) |
1da177e4 LT |
530 | { |
531 | mm->pgd = pgd_alloc(mm); | |
532 | if (unlikely(!mm->pgd)) | |
533 | return -ENOMEM; | |
534 | return 0; | |
535 | } | |
536 | ||
fb0a685c | 537 | static inline void mm_free_pgd(struct mm_struct *mm) |
1da177e4 | 538 | { |
5e541973 | 539 | pgd_free(mm, mm->pgd); |
1da177e4 LT |
540 | } |
541 | #else | |
90f31d0e KK |
542 | static int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm) |
543 | { | |
544 | down_write(&oldmm->mmap_sem); | |
545 | RCU_INIT_POINTER(mm->exe_file, get_mm_exe_file(oldmm)); | |
546 | up_write(&oldmm->mmap_sem); | |
547 | return 0; | |
548 | } | |
1da177e4 LT |
549 | #define mm_alloc_pgd(mm) (0) |
550 | #define mm_free_pgd(mm) | |
551 | #endif /* CONFIG_MMU */ | |
552 | ||
23ff4440 | 553 | __cacheline_aligned_in_smp DEFINE_SPINLOCK(mmlist_lock); |
1da177e4 | 554 | |
e94b1766 | 555 | #define allocate_mm() (kmem_cache_alloc(mm_cachep, GFP_KERNEL)) |
1da177e4 LT |
556 | #define free_mm(mm) (kmem_cache_free(mm_cachep, (mm))) |
557 | ||
4cb0e11b HK |
558 | static unsigned long default_dump_filter = MMF_DUMP_FILTER_DEFAULT; |
559 | ||
560 | static int __init coredump_filter_setup(char *s) | |
561 | { | |
562 | default_dump_filter = | |
563 | (simple_strtoul(s, NULL, 0) << MMF_DUMP_FILTER_SHIFT) & | |
564 | MMF_DUMP_FILTER_MASK; | |
565 | return 1; | |
566 | } | |
567 | ||
568 | __setup("coredump_filter=", coredump_filter_setup); | |
569 | ||
1da177e4 LT |
570 | #include <linux/init_task.h> |
571 | ||
858f0993 AD |
572 | static void mm_init_aio(struct mm_struct *mm) |
573 | { | |
574 | #ifdef CONFIG_AIO | |
575 | spin_lock_init(&mm->ioctx_lock); | |
db446a08 | 576 | mm->ioctx_table = NULL; |
858f0993 AD |
577 | #endif |
578 | } | |
579 | ||
33144e84 VD |
580 | static void mm_init_owner(struct mm_struct *mm, struct task_struct *p) |
581 | { | |
582 | #ifdef CONFIG_MEMCG | |
583 | mm->owner = p; | |
584 | #endif | |
585 | } | |
586 | ||
fb0a685c | 587 | static struct mm_struct *mm_init(struct mm_struct *mm, struct task_struct *p) |
1da177e4 | 588 | { |
41f727fd VD |
589 | mm->mmap = NULL; |
590 | mm->mm_rb = RB_ROOT; | |
591 | mm->vmacache_seqnum = 0; | |
1da177e4 LT |
592 | atomic_set(&mm->mm_users, 1); |
593 | atomic_set(&mm->mm_count, 1); | |
594 | init_rwsem(&mm->mmap_sem); | |
595 | INIT_LIST_HEAD(&mm->mmlist); | |
999d9fc1 | 596 | mm->core_state = NULL; |
e1f56c89 | 597 | atomic_long_set(&mm->nr_ptes, 0); |
2d2f5119 | 598 | mm_nr_pmds_init(mm); |
41f727fd VD |
599 | mm->map_count = 0; |
600 | mm->locked_vm = 0; | |
ce65cefa | 601 | mm->pinned_vm = 0; |
d559db08 | 602 | memset(&mm->rss_stat, 0, sizeof(mm->rss_stat)); |
1da177e4 | 603 | spin_lock_init(&mm->page_table_lock); |
41f727fd | 604 | mm_init_cpumask(mm); |
858f0993 | 605 | mm_init_aio(mm); |
cf475ad2 | 606 | mm_init_owner(mm, p); |
41f727fd | 607 | mmu_notifier_mm_init(mm); |
20841405 | 608 | clear_tlb_flush_pending(mm); |
41f727fd VD |
609 | #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS |
610 | mm->pmd_huge_pte = NULL; | |
611 | #endif | |
1da177e4 | 612 | |
a0715cc2 AT |
613 | if (current->mm) { |
614 | mm->flags = current->mm->flags & MMF_INIT_MASK; | |
615 | mm->def_flags = current->mm->def_flags & VM_INIT_DEF_MASK; | |
616 | } else { | |
617 | mm->flags = default_dump_filter; | |
1da177e4 | 618 | mm->def_flags = 0; |
a0715cc2 AT |
619 | } |
620 | ||
41f727fd VD |
621 | if (mm_alloc_pgd(mm)) |
622 | goto fail_nopgd; | |
623 | ||
624 | if (init_new_context(p, mm)) | |
625 | goto fail_nocontext; | |
78fb7466 | 626 | |
41f727fd VD |
627 | return mm; |
628 | ||
629 | fail_nocontext: | |
630 | mm_free_pgd(mm); | |
631 | fail_nopgd: | |
1da177e4 LT |
632 | free_mm(mm); |
633 | return NULL; | |
634 | } | |
635 | ||
c3f0327f KK |
636 | static void check_mm(struct mm_struct *mm) |
637 | { | |
638 | int i; | |
639 | ||
640 | for (i = 0; i < NR_MM_COUNTERS; i++) { | |
641 | long x = atomic_long_read(&mm->rss_stat.count[i]); | |
642 | ||
643 | if (unlikely(x)) | |
644 | printk(KERN_ALERT "BUG: Bad rss-counter state " | |
645 | "mm:%p idx:%d val:%ld\n", mm, i, x); | |
646 | } | |
b30fe6c7 KS |
647 | |
648 | if (atomic_long_read(&mm->nr_ptes)) | |
649 | pr_alert("BUG: non-zero nr_ptes on freeing mm: %ld\n", | |
650 | atomic_long_read(&mm->nr_ptes)); | |
651 | if (mm_nr_pmds(mm)) | |
652 | pr_alert("BUG: non-zero nr_pmds on freeing mm: %ld\n", | |
653 | mm_nr_pmds(mm)); | |
654 | ||
e009bb30 | 655 | #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS |
96dad67f | 656 | VM_BUG_ON_MM(mm->pmd_huge_pte, mm); |
c3f0327f KK |
657 | #endif |
658 | } | |
659 | ||
1da177e4 LT |
660 | /* |
661 | * Allocate and initialize an mm_struct. | |
662 | */ | |
fb0a685c | 663 | struct mm_struct *mm_alloc(void) |
1da177e4 | 664 | { |
fb0a685c | 665 | struct mm_struct *mm; |
1da177e4 LT |
666 | |
667 | mm = allocate_mm(); | |
de03c72c KM |
668 | if (!mm) |
669 | return NULL; | |
670 | ||
671 | memset(mm, 0, sizeof(*mm)); | |
6345d24d | 672 | return mm_init(mm, current); |
1da177e4 LT |
673 | } |
674 | ||
675 | /* | |
676 | * Called when the last reference to the mm | |
677 | * is dropped: either by a lazy thread or by | |
678 | * mmput. Free the page directory and the mm. | |
679 | */ | |
7ad5b3a5 | 680 | void __mmdrop(struct mm_struct *mm) |
1da177e4 LT |
681 | { |
682 | BUG_ON(mm == &init_mm); | |
683 | mm_free_pgd(mm); | |
684 | destroy_context(mm); | |
cddb8a5c | 685 | mmu_notifier_mm_destroy(mm); |
c3f0327f | 686 | check_mm(mm); |
1da177e4 LT |
687 | free_mm(mm); |
688 | } | |
6d4e4c4f | 689 | EXPORT_SYMBOL_GPL(__mmdrop); |
1da177e4 LT |
690 | |
691 | /* | |
692 | * Decrement the use count and release all resources for an mm. | |
693 | */ | |
694 | void mmput(struct mm_struct *mm) | |
695 | { | |
0ae26f1b AM |
696 | might_sleep(); |
697 | ||
1da177e4 | 698 | if (atomic_dec_and_test(&mm->mm_users)) { |
d4b3b638 | 699 | uprobe_clear_state(mm); |
1da177e4 | 700 | exit_aio(mm); |
1c2fb7a4 | 701 | ksm_exit(mm); |
ba76149f | 702 | khugepaged_exit(mm); /* must run before exit_mmap */ |
1da177e4 | 703 | exit_mmap(mm); |
925d1c40 | 704 | set_mm_exe_file(mm, NULL); |
1da177e4 LT |
705 | if (!list_empty(&mm->mmlist)) { |
706 | spin_lock(&mmlist_lock); | |
707 | list_del(&mm->mmlist); | |
708 | spin_unlock(&mmlist_lock); | |
709 | } | |
801460d0 HS |
710 | if (mm->binfmt) |
711 | module_put(mm->binfmt->module); | |
1da177e4 LT |
712 | mmdrop(mm); |
713 | } | |
714 | } | |
715 | EXPORT_SYMBOL_GPL(mmput); | |
716 | ||
90f31d0e KK |
717 | /** |
718 | * set_mm_exe_file - change a reference to the mm's executable file | |
719 | * | |
720 | * This changes mm's executable file (shown as symlink /proc/[pid]/exe). | |
721 | * | |
6e399cd1 DB |
722 | * Main users are mmput() and sys_execve(). Callers prevent concurrent |
723 | * invocations: in mmput() nobody alive left, in execve task is single | |
724 | * threaded. sys_prctl(PR_SET_MM_MAP/EXE_FILE) also needs to set the | |
725 | * mm->exe_file, but does so without using set_mm_exe_file() in order | |
726 | * to do avoid the need for any locks. | |
90f31d0e | 727 | */ |
38646013 JS |
728 | void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file) |
729 | { | |
6e399cd1 DB |
730 | struct file *old_exe_file; |
731 | ||
732 | /* | |
733 | * It is safe to dereference the exe_file without RCU as | |
734 | * this function is only called if nobody else can access | |
735 | * this mm -- see comment above for justification. | |
736 | */ | |
737 | old_exe_file = rcu_dereference_raw(mm->exe_file); | |
90f31d0e | 738 | |
38646013 JS |
739 | if (new_exe_file) |
740 | get_file(new_exe_file); | |
90f31d0e KK |
741 | rcu_assign_pointer(mm->exe_file, new_exe_file); |
742 | if (old_exe_file) | |
743 | fput(old_exe_file); | |
38646013 JS |
744 | } |
745 | ||
90f31d0e KK |
746 | /** |
747 | * get_mm_exe_file - acquire a reference to the mm's executable file | |
748 | * | |
749 | * Returns %NULL if mm has no associated executable file. | |
750 | * User must release file via fput(). | |
751 | */ | |
38646013 JS |
752 | struct file *get_mm_exe_file(struct mm_struct *mm) |
753 | { | |
754 | struct file *exe_file; | |
755 | ||
90f31d0e KK |
756 | rcu_read_lock(); |
757 | exe_file = rcu_dereference(mm->exe_file); | |
758 | if (exe_file && !get_file_rcu(exe_file)) | |
759 | exe_file = NULL; | |
760 | rcu_read_unlock(); | |
38646013 JS |
761 | return exe_file; |
762 | } | |
11163348 | 763 | EXPORT_SYMBOL(get_mm_exe_file); |
38646013 | 764 | |
1da177e4 LT |
765 | /** |
766 | * get_task_mm - acquire a reference to the task's mm | |
767 | * | |
246bb0b1 | 768 | * Returns %NULL if the task has no mm. Checks PF_KTHREAD (meaning |
1da177e4 LT |
769 | * this kernel workthread has transiently adopted a user mm with use_mm, |
770 | * to do its AIO) is not set and if so returns a reference to it, after | |
771 | * bumping up the use count. User must release the mm via mmput() | |
772 | * after use. Typically used by /proc and ptrace. | |
773 | */ | |
774 | struct mm_struct *get_task_mm(struct task_struct *task) | |
775 | { | |
776 | struct mm_struct *mm; | |
777 | ||
778 | task_lock(task); | |
779 | mm = task->mm; | |
780 | if (mm) { | |
246bb0b1 | 781 | if (task->flags & PF_KTHREAD) |
1da177e4 LT |
782 | mm = NULL; |
783 | else | |
784 | atomic_inc(&mm->mm_users); | |
785 | } | |
786 | task_unlock(task); | |
787 | return mm; | |
788 | } | |
789 | EXPORT_SYMBOL_GPL(get_task_mm); | |
790 | ||
8cdb878d CY |
791 | struct mm_struct *mm_access(struct task_struct *task, unsigned int mode) |
792 | { | |
793 | struct mm_struct *mm; | |
794 | int err; | |
795 | ||
796 | err = mutex_lock_killable(&task->signal->cred_guard_mutex); | |
797 | if (err) | |
798 | return ERR_PTR(err); | |
799 | ||
800 | mm = get_task_mm(task); | |
801 | if (mm && mm != current->mm && | |
802 | !ptrace_may_access(task, mode)) { | |
803 | mmput(mm); | |
804 | mm = ERR_PTR(-EACCES); | |
805 | } | |
806 | mutex_unlock(&task->signal->cred_guard_mutex); | |
807 | ||
808 | return mm; | |
809 | } | |
810 | ||
57b59c4a | 811 | static void complete_vfork_done(struct task_struct *tsk) |
c415c3b4 | 812 | { |
d68b46fe | 813 | struct completion *vfork; |
c415c3b4 | 814 | |
d68b46fe ON |
815 | task_lock(tsk); |
816 | vfork = tsk->vfork_done; | |
817 | if (likely(vfork)) { | |
818 | tsk->vfork_done = NULL; | |
819 | complete(vfork); | |
820 | } | |
821 | task_unlock(tsk); | |
822 | } | |
823 | ||
824 | static int wait_for_vfork_done(struct task_struct *child, | |
825 | struct completion *vfork) | |
826 | { | |
827 | int killed; | |
828 | ||
829 | freezer_do_not_count(); | |
830 | killed = wait_for_completion_killable(vfork); | |
831 | freezer_count(); | |
832 | ||
833 | if (killed) { | |
834 | task_lock(child); | |
835 | child->vfork_done = NULL; | |
836 | task_unlock(child); | |
837 | } | |
838 | ||
839 | put_task_struct(child); | |
840 | return killed; | |
c415c3b4 ON |
841 | } |
842 | ||
1da177e4 LT |
843 | /* Please note the differences between mmput and mm_release. |
844 | * mmput is called whenever we stop holding onto a mm_struct, | |
845 | * error success whatever. | |
846 | * | |
847 | * mm_release is called after a mm_struct has been removed | |
848 | * from the current process. | |
849 | * | |
850 | * This difference is important for error handling, when we | |
851 | * only half set up a mm_struct for a new process and need to restore | |
852 | * the old one. Because we mmput the new mm_struct before | |
853 | * restoring the old one. . . | |
854 | * Eric Biederman 10 January 1998 | |
855 | */ | |
856 | void mm_release(struct task_struct *tsk, struct mm_struct *mm) | |
857 | { | |
8141c7f3 LT |
858 | /* Get rid of any futexes when releasing the mm */ |
859 | #ifdef CONFIG_FUTEX | |
fc6b177d | 860 | if (unlikely(tsk->robust_list)) { |
8141c7f3 | 861 | exit_robust_list(tsk); |
fc6b177d PZ |
862 | tsk->robust_list = NULL; |
863 | } | |
8141c7f3 | 864 | #ifdef CONFIG_COMPAT |
fc6b177d | 865 | if (unlikely(tsk->compat_robust_list)) { |
8141c7f3 | 866 | compat_exit_robust_list(tsk); |
fc6b177d PZ |
867 | tsk->compat_robust_list = NULL; |
868 | } | |
8141c7f3 | 869 | #endif |
322a2c10 TG |
870 | if (unlikely(!list_empty(&tsk->pi_state_list))) |
871 | exit_pi_state_list(tsk); | |
8141c7f3 LT |
872 | #endif |
873 | ||
0326f5a9 SD |
874 | uprobe_free_utask(tsk); |
875 | ||
1da177e4 LT |
876 | /* Get rid of any cached register state */ |
877 | deactivate_mm(tsk, mm); | |
878 | ||
fec1d011 RM |
879 | /* |
880 | * If we're exiting normally, clear a user-space tid field if | |
881 | * requested. We leave this alone when dying by signal, to leave | |
882 | * the value intact in a core dump, and to save the unnecessary | |
d68b46fe ON |
883 | * trouble, say, a killed vfork parent shouldn't touch this mm. |
884 | * Userland only wants this done for a sys_exit. | |
fec1d011 | 885 | */ |
9c8a8228 ED |
886 | if (tsk->clear_child_tid) { |
887 | if (!(tsk->flags & PF_SIGNALED) && | |
888 | atomic_read(&mm->mm_users) > 1) { | |
889 | /* | |
890 | * We don't check the error code - if userspace has | |
891 | * not set up a proper pointer then tough luck. | |
892 | */ | |
893 | put_user(0, tsk->clear_child_tid); | |
894 | sys_futex(tsk->clear_child_tid, FUTEX_WAKE, | |
895 | 1, NULL, NULL, 0); | |
896 | } | |
1da177e4 | 897 | tsk->clear_child_tid = NULL; |
1da177e4 | 898 | } |
f7505d64 KK |
899 | |
900 | /* | |
901 | * All done, finally we can wake up parent and return this mm to him. | |
902 | * Also kthread_stop() uses this completion for synchronization. | |
903 | */ | |
904 | if (tsk->vfork_done) | |
905 | complete_vfork_done(tsk); | |
1da177e4 LT |
906 | } |
907 | ||
a0a7ec30 JD |
908 | /* |
909 | * Allocate a new mm structure and copy contents from the | |
910 | * mm structure of the passed in task structure. | |
911 | */ | |
ff252c1f | 912 | static struct mm_struct *dup_mm(struct task_struct *tsk) |
a0a7ec30 JD |
913 | { |
914 | struct mm_struct *mm, *oldmm = current->mm; | |
915 | int err; | |
916 | ||
a0a7ec30 JD |
917 | mm = allocate_mm(); |
918 | if (!mm) | |
919 | goto fail_nomem; | |
920 | ||
921 | memcpy(mm, oldmm, sizeof(*mm)); | |
922 | ||
78fb7466 | 923 | if (!mm_init(mm, tsk)) |
a0a7ec30 JD |
924 | goto fail_nomem; |
925 | ||
a0a7ec30 JD |
926 | err = dup_mmap(mm, oldmm); |
927 | if (err) | |
928 | goto free_pt; | |
929 | ||
930 | mm->hiwater_rss = get_mm_rss(mm); | |
931 | mm->hiwater_vm = mm->total_vm; | |
932 | ||
801460d0 HS |
933 | if (mm->binfmt && !try_module_get(mm->binfmt->module)) |
934 | goto free_pt; | |
935 | ||
a0a7ec30 JD |
936 | return mm; |
937 | ||
938 | free_pt: | |
801460d0 HS |
939 | /* don't put binfmt in mmput, we haven't got module yet */ |
940 | mm->binfmt = NULL; | |
a0a7ec30 JD |
941 | mmput(mm); |
942 | ||
943 | fail_nomem: | |
944 | return NULL; | |
a0a7ec30 JD |
945 | } |
946 | ||
fb0a685c | 947 | static int copy_mm(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 | 948 | { |
fb0a685c | 949 | struct mm_struct *mm, *oldmm; |
1da177e4 LT |
950 | int retval; |
951 | ||
952 | tsk->min_flt = tsk->maj_flt = 0; | |
953 | tsk->nvcsw = tsk->nivcsw = 0; | |
17406b82 MSB |
954 | #ifdef CONFIG_DETECT_HUNG_TASK |
955 | tsk->last_switch_count = tsk->nvcsw + tsk->nivcsw; | |
956 | #endif | |
1da177e4 LT |
957 | |
958 | tsk->mm = NULL; | |
959 | tsk->active_mm = NULL; | |
960 | ||
961 | /* | |
962 | * Are we cloning a kernel thread? | |
963 | * | |
964 | * We need to steal a active VM for that.. | |
965 | */ | |
966 | oldmm = current->mm; | |
967 | if (!oldmm) | |
968 | return 0; | |
969 | ||
615d6e87 DB |
970 | /* initialize the new vmacache entries */ |
971 | vmacache_flush(tsk); | |
972 | ||
1da177e4 LT |
973 | if (clone_flags & CLONE_VM) { |
974 | atomic_inc(&oldmm->mm_users); | |
975 | mm = oldmm; | |
1da177e4 LT |
976 | goto good_mm; |
977 | } | |
978 | ||
979 | retval = -ENOMEM; | |
a0a7ec30 | 980 | mm = dup_mm(tsk); |
1da177e4 LT |
981 | if (!mm) |
982 | goto fail_nomem; | |
983 | ||
1da177e4 LT |
984 | good_mm: |
985 | tsk->mm = mm; | |
986 | tsk->active_mm = mm; | |
987 | return 0; | |
988 | ||
1da177e4 LT |
989 | fail_nomem: |
990 | return retval; | |
1da177e4 LT |
991 | } |
992 | ||
a39bc516 | 993 | static int copy_fs(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 | 994 | { |
498052bb | 995 | struct fs_struct *fs = current->fs; |
1da177e4 | 996 | if (clone_flags & CLONE_FS) { |
498052bb | 997 | /* tsk->fs is already what we want */ |
2a4419b5 | 998 | spin_lock(&fs->lock); |
498052bb | 999 | if (fs->in_exec) { |
2a4419b5 | 1000 | spin_unlock(&fs->lock); |
498052bb AV |
1001 | return -EAGAIN; |
1002 | } | |
1003 | fs->users++; | |
2a4419b5 | 1004 | spin_unlock(&fs->lock); |
1da177e4 LT |
1005 | return 0; |
1006 | } | |
498052bb | 1007 | tsk->fs = copy_fs_struct(fs); |
1da177e4 LT |
1008 | if (!tsk->fs) |
1009 | return -ENOMEM; | |
1010 | return 0; | |
1011 | } | |
1012 | ||
fb0a685c | 1013 | static int copy_files(unsigned long clone_flags, struct task_struct *tsk) |
a016f338 JD |
1014 | { |
1015 | struct files_struct *oldf, *newf; | |
1016 | int error = 0; | |
1017 | ||
1018 | /* | |
1019 | * A background process may not have any files ... | |
1020 | */ | |
1021 | oldf = current->files; | |
1022 | if (!oldf) | |
1023 | goto out; | |
1024 | ||
1025 | if (clone_flags & CLONE_FILES) { | |
1026 | atomic_inc(&oldf->count); | |
1027 | goto out; | |
1028 | } | |
1029 | ||
a016f338 JD |
1030 | newf = dup_fd(oldf, &error); |
1031 | if (!newf) | |
1032 | goto out; | |
1033 | ||
1034 | tsk->files = newf; | |
1035 | error = 0; | |
1036 | out: | |
1037 | return error; | |
1038 | } | |
1039 | ||
fadad878 | 1040 | static int copy_io(unsigned long clone_flags, struct task_struct *tsk) |
fd0928df JA |
1041 | { |
1042 | #ifdef CONFIG_BLOCK | |
1043 | struct io_context *ioc = current->io_context; | |
6e736be7 | 1044 | struct io_context *new_ioc; |
fd0928df JA |
1045 | |
1046 | if (!ioc) | |
1047 | return 0; | |
fadad878 JA |
1048 | /* |
1049 | * Share io context with parent, if CLONE_IO is set | |
1050 | */ | |
1051 | if (clone_flags & CLONE_IO) { | |
3d48749d TH |
1052 | ioc_task_link(ioc); |
1053 | tsk->io_context = ioc; | |
fadad878 | 1054 | } else if (ioprio_valid(ioc->ioprio)) { |
6e736be7 TH |
1055 | new_ioc = get_task_io_context(tsk, GFP_KERNEL, NUMA_NO_NODE); |
1056 | if (unlikely(!new_ioc)) | |
fd0928df JA |
1057 | return -ENOMEM; |
1058 | ||
6e736be7 | 1059 | new_ioc->ioprio = ioc->ioprio; |
11a3122f | 1060 | put_io_context(new_ioc); |
fd0928df JA |
1061 | } |
1062 | #endif | |
1063 | return 0; | |
1064 | } | |
1065 | ||
a39bc516 | 1066 | static int copy_sighand(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 LT |
1067 | { |
1068 | struct sighand_struct *sig; | |
1069 | ||
60348802 | 1070 | if (clone_flags & CLONE_SIGHAND) { |
1da177e4 LT |
1071 | atomic_inc(¤t->sighand->count); |
1072 | return 0; | |
1073 | } | |
1074 | sig = kmem_cache_alloc(sighand_cachep, GFP_KERNEL); | |
e56d0903 | 1075 | rcu_assign_pointer(tsk->sighand, sig); |
1da177e4 LT |
1076 | if (!sig) |
1077 | return -ENOMEM; | |
9d7fb042 | 1078 | |
1da177e4 LT |
1079 | atomic_set(&sig->count, 1); |
1080 | memcpy(sig->action, current->sighand->action, sizeof(sig->action)); | |
1081 | return 0; | |
1082 | } | |
1083 | ||
a7e5328a | 1084 | void __cleanup_sighand(struct sighand_struct *sighand) |
c81addc9 | 1085 | { |
d80e731e ON |
1086 | if (atomic_dec_and_test(&sighand->count)) { |
1087 | signalfd_cleanup(sighand); | |
392809b2 ON |
1088 | /* |
1089 | * sighand_cachep is SLAB_DESTROY_BY_RCU so we can free it | |
1090 | * without an RCU grace period, see __lock_task_sighand(). | |
1091 | */ | |
c81addc9 | 1092 | kmem_cache_free(sighand_cachep, sighand); |
d80e731e | 1093 | } |
c81addc9 ON |
1094 | } |
1095 | ||
f06febc9 FM |
1096 | /* |
1097 | * Initialize POSIX timer handling for a thread group. | |
1098 | */ | |
1099 | static void posix_cpu_timers_init_group(struct signal_struct *sig) | |
1100 | { | |
78d7d407 JS |
1101 | unsigned long cpu_limit; |
1102 | ||
316c1608 | 1103 | cpu_limit = READ_ONCE(sig->rlim[RLIMIT_CPU].rlim_cur); |
78d7d407 JS |
1104 | if (cpu_limit != RLIM_INFINITY) { |
1105 | sig->cputime_expires.prof_exp = secs_to_cputime(cpu_limit); | |
d5c373eb | 1106 | sig->cputimer.running = true; |
6279a751 ON |
1107 | } |
1108 | ||
f06febc9 FM |
1109 | /* The timer lists. */ |
1110 | INIT_LIST_HEAD(&sig->cpu_timers[0]); | |
1111 | INIT_LIST_HEAD(&sig->cpu_timers[1]); | |
1112 | INIT_LIST_HEAD(&sig->cpu_timers[2]); | |
1113 | } | |
1114 | ||
a39bc516 | 1115 | static int copy_signal(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 LT |
1116 | { |
1117 | struct signal_struct *sig; | |
1da177e4 | 1118 | |
4ab6c083 | 1119 | if (clone_flags & CLONE_THREAD) |
490dea45 | 1120 | return 0; |
490dea45 | 1121 | |
a56704ef | 1122 | sig = kmem_cache_zalloc(signal_cachep, GFP_KERNEL); |
1da177e4 LT |
1123 | tsk->signal = sig; |
1124 | if (!sig) | |
1125 | return -ENOMEM; | |
1126 | ||
b3ac022c | 1127 | sig->nr_threads = 1; |
1da177e4 | 1128 | atomic_set(&sig->live, 1); |
b3ac022c | 1129 | atomic_set(&sig->sigcnt, 1); |
0c740d0a ON |
1130 | |
1131 | /* list_add(thread_node, thread_head) without INIT_LIST_HEAD() */ | |
1132 | sig->thread_head = (struct list_head)LIST_HEAD_INIT(tsk->thread_node); | |
1133 | tsk->thread_node = (struct list_head)LIST_HEAD_INIT(sig->thread_head); | |
1134 | ||
1da177e4 | 1135 | init_waitqueue_head(&sig->wait_chldexit); |
db51aecc | 1136 | sig->curr_target = tsk; |
1da177e4 LT |
1137 | init_sigpending(&sig->shared_pending); |
1138 | INIT_LIST_HEAD(&sig->posix_timers); | |
e78c3496 | 1139 | seqlock_init(&sig->stats_lock); |
9d7fb042 | 1140 | prev_cputime_init(&sig->prev_cputime); |
1da177e4 | 1141 | |
c9cb2e3d | 1142 | hrtimer_init(&sig->real_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL); |
1da177e4 | 1143 | sig->real_timer.function = it_real_fn; |
1da177e4 | 1144 | |
1da177e4 LT |
1145 | task_lock(current->group_leader); |
1146 | memcpy(sig->rlim, current->signal->rlim, sizeof sig->rlim); | |
1147 | task_unlock(current->group_leader); | |
1148 | ||
6279a751 ON |
1149 | posix_cpu_timers_init_group(sig); |
1150 | ||
522ed776 | 1151 | tty_audit_fork(sig); |
5091faa4 | 1152 | sched_autogroup_fork(sig); |
522ed776 | 1153 | |
a63d83f4 | 1154 | sig->oom_score_adj = current->signal->oom_score_adj; |
dabb16f6 | 1155 | sig->oom_score_adj_min = current->signal->oom_score_adj_min; |
28b83c51 | 1156 | |
ebec18a6 LP |
1157 | sig->has_child_subreaper = current->signal->has_child_subreaper || |
1158 | current->signal->is_child_subreaper; | |
1159 | ||
9b1bf12d KM |
1160 | mutex_init(&sig->cred_guard_mutex); |
1161 | ||
1da177e4 LT |
1162 | return 0; |
1163 | } | |
1164 | ||
dbd95212 KC |
1165 | static void copy_seccomp(struct task_struct *p) |
1166 | { | |
1167 | #ifdef CONFIG_SECCOMP | |
1168 | /* | |
1169 | * Must be called with sighand->lock held, which is common to | |
1170 | * all threads in the group. Holding cred_guard_mutex is not | |
1171 | * needed because this new task is not yet running and cannot | |
1172 | * be racing exec. | |
1173 | */ | |
69f6a34b | 1174 | assert_spin_locked(¤t->sighand->siglock); |
dbd95212 KC |
1175 | |
1176 | /* Ref-count the new filter user, and assign it. */ | |
1177 | get_seccomp_filter(current); | |
1178 | p->seccomp = current->seccomp; | |
1179 | ||
1180 | /* | |
1181 | * Explicitly enable no_new_privs here in case it got set | |
1182 | * between the task_struct being duplicated and holding the | |
1183 | * sighand lock. The seccomp state and nnp must be in sync. | |
1184 | */ | |
1185 | if (task_no_new_privs(current)) | |
1186 | task_set_no_new_privs(p); | |
1187 | ||
1188 | /* | |
1189 | * If the parent gained a seccomp mode after copying thread | |
1190 | * flags and between before we held the sighand lock, we have | |
1191 | * to manually enable the seccomp thread flag here. | |
1192 | */ | |
1193 | if (p->seccomp.mode != SECCOMP_MODE_DISABLED) | |
1194 | set_tsk_thread_flag(p, TIF_SECCOMP); | |
1195 | #endif | |
1196 | } | |
1197 | ||
17da2bd9 | 1198 | SYSCALL_DEFINE1(set_tid_address, int __user *, tidptr) |
1da177e4 LT |
1199 | { |
1200 | current->clear_child_tid = tidptr; | |
1201 | ||
b488893a | 1202 | return task_pid_vnr(current); |
1da177e4 LT |
1203 | } |
1204 | ||
a39bc516 | 1205 | static void rt_mutex_init_task(struct task_struct *p) |
23f78d4a | 1206 | { |
1d615482 | 1207 | raw_spin_lock_init(&p->pi_lock); |
e29e175b | 1208 | #ifdef CONFIG_RT_MUTEXES |
fb00aca4 PZ |
1209 | p->pi_waiters = RB_ROOT; |
1210 | p->pi_waiters_leftmost = NULL; | |
23f78d4a | 1211 | p->pi_blocked_on = NULL; |
23f78d4a IM |
1212 | #endif |
1213 | } | |
1214 | ||
f06febc9 FM |
1215 | /* |
1216 | * Initialize POSIX timer handling for a single task. | |
1217 | */ | |
1218 | static void posix_cpu_timers_init(struct task_struct *tsk) | |
1219 | { | |
64861634 MS |
1220 | tsk->cputime_expires.prof_exp = 0; |
1221 | tsk->cputime_expires.virt_exp = 0; | |
f06febc9 FM |
1222 | tsk->cputime_expires.sched_exp = 0; |
1223 | INIT_LIST_HEAD(&tsk->cpu_timers[0]); | |
1224 | INIT_LIST_HEAD(&tsk->cpu_timers[1]); | |
1225 | INIT_LIST_HEAD(&tsk->cpu_timers[2]); | |
1226 | } | |
1227 | ||
81907739 ON |
1228 | static inline void |
1229 | init_task_pid(struct task_struct *task, enum pid_type type, struct pid *pid) | |
1230 | { | |
1231 | task->pids[type].pid = pid; | |
1232 | } | |
1233 | ||
1da177e4 LT |
1234 | /* |
1235 | * This creates a new process as a copy of the old one, | |
1236 | * but does not actually start it yet. | |
1237 | * | |
1238 | * It copies the registers, and all the appropriate | |
1239 | * parts of the process environment (as per the clone | |
1240 | * flags). The actual kick-off is left to the caller. | |
1241 | */ | |
36c8b586 IM |
1242 | static struct task_struct *copy_process(unsigned long clone_flags, |
1243 | unsigned long stack_start, | |
36c8b586 | 1244 | unsigned long stack_size, |
36c8b586 | 1245 | int __user *child_tidptr, |
09a05394 | 1246 | struct pid *pid, |
3033f14a JT |
1247 | int trace, |
1248 | unsigned long tls) | |
1da177e4 LT |
1249 | { |
1250 | int retval; | |
a24efe62 | 1251 | struct task_struct *p; |
1da177e4 LT |
1252 | |
1253 | if ((clone_flags & (CLONE_NEWNS|CLONE_FS)) == (CLONE_NEWNS|CLONE_FS)) | |
1254 | return ERR_PTR(-EINVAL); | |
1255 | ||
e66eded8 EB |
1256 | if ((clone_flags & (CLONE_NEWUSER|CLONE_FS)) == (CLONE_NEWUSER|CLONE_FS)) |
1257 | return ERR_PTR(-EINVAL); | |
1258 | ||
1da177e4 LT |
1259 | /* |
1260 | * Thread groups must share signals as well, and detached threads | |
1261 | * can only be started up within the thread group. | |
1262 | */ | |
1263 | if ((clone_flags & CLONE_THREAD) && !(clone_flags & CLONE_SIGHAND)) | |
1264 | return ERR_PTR(-EINVAL); | |
1265 | ||
1266 | /* | |
1267 | * Shared signal handlers imply shared VM. By way of the above, | |
1268 | * thread groups also imply shared VM. Blocking this case allows | |
1269 | * for various simplifications in other code. | |
1270 | */ | |
1271 | if ((clone_flags & CLONE_SIGHAND) && !(clone_flags & CLONE_VM)) | |
1272 | return ERR_PTR(-EINVAL); | |
1273 | ||
123be07b SB |
1274 | /* |
1275 | * Siblings of global init remain as zombies on exit since they are | |
1276 | * not reaped by their parent (swapper). To solve this and to avoid | |
1277 | * multi-rooted process trees, prevent global and container-inits | |
1278 | * from creating siblings. | |
1279 | */ | |
1280 | if ((clone_flags & CLONE_PARENT) && | |
1281 | current->signal->flags & SIGNAL_UNKILLABLE) | |
1282 | return ERR_PTR(-EINVAL); | |
1283 | ||
8382fcac | 1284 | /* |
40a0d32d | 1285 | * If the new process will be in a different pid or user namespace |
faf00da5 | 1286 | * do not allow it to share a thread group with the forking task. |
8382fcac | 1287 | */ |
faf00da5 | 1288 | if (clone_flags & CLONE_THREAD) { |
40a0d32d ON |
1289 | if ((clone_flags & (CLONE_NEWUSER | CLONE_NEWPID)) || |
1290 | (task_active_pid_ns(current) != | |
1291 | current->nsproxy->pid_ns_for_children)) | |
1292 | return ERR_PTR(-EINVAL); | |
1293 | } | |
8382fcac | 1294 | |
1da177e4 LT |
1295 | retval = security_task_create(clone_flags); |
1296 | if (retval) | |
1297 | goto fork_out; | |
1298 | ||
1299 | retval = -ENOMEM; | |
1300 | p = dup_task_struct(current); | |
1301 | if (!p) | |
1302 | goto fork_out; | |
1303 | ||
f7e8b616 SR |
1304 | ftrace_graph_init_task(p); |
1305 | ||
bea493a0 PZ |
1306 | rt_mutex_init_task(p); |
1307 | ||
d12c1a37 | 1308 | #ifdef CONFIG_PROVE_LOCKING |
de30a2b3 IM |
1309 | DEBUG_LOCKS_WARN_ON(!p->hardirqs_enabled); |
1310 | DEBUG_LOCKS_WARN_ON(!p->softirqs_enabled); | |
1311 | #endif | |
1da177e4 | 1312 | retval = -EAGAIN; |
3b11a1de | 1313 | if (atomic_read(&p->real_cred->user->processes) >= |
78d7d407 | 1314 | task_rlimit(p, RLIMIT_NPROC)) { |
b57922b6 EP |
1315 | if (p->real_cred->user != INIT_USER && |
1316 | !capable(CAP_SYS_RESOURCE) && !capable(CAP_SYS_ADMIN)) | |
1da177e4 LT |
1317 | goto bad_fork_free; |
1318 | } | |
72fa5997 | 1319 | current->flags &= ~PF_NPROC_EXCEEDED; |
1da177e4 | 1320 | |
f1752eec DH |
1321 | retval = copy_creds(p, clone_flags); |
1322 | if (retval < 0) | |
1323 | goto bad_fork_free; | |
1da177e4 LT |
1324 | |
1325 | /* | |
1326 | * If multiple threads are within copy_process(), then this check | |
1327 | * triggers too late. This doesn't hurt, the check is only there | |
1328 | * to stop root fork bombs. | |
1329 | */ | |
04ec93fe | 1330 | retval = -EAGAIN; |
1da177e4 LT |
1331 | if (nr_threads >= max_threads) |
1332 | goto bad_fork_cleanup_count; | |
1333 | ||
ca74e92b | 1334 | delayacct_tsk_init(p); /* Must remain after dup_task_struct() */ |
514ddb44 DR |
1335 | p->flags &= ~(PF_SUPERPRIV | PF_WQ_WORKER); |
1336 | p->flags |= PF_FORKNOEXEC; | |
1da177e4 LT |
1337 | INIT_LIST_HEAD(&p->children); |
1338 | INIT_LIST_HEAD(&p->sibling); | |
f41d911f | 1339 | rcu_copy_process(p); |
1da177e4 LT |
1340 | p->vfork_done = NULL; |
1341 | spin_lock_init(&p->alloc_lock); | |
1da177e4 | 1342 | |
1da177e4 LT |
1343 | init_sigpending(&p->pending); |
1344 | ||
64861634 MS |
1345 | p->utime = p->stime = p->gtime = 0; |
1346 | p->utimescaled = p->stimescaled = 0; | |
9d7fb042 PZ |
1347 | prev_cputime_init(&p->prev_cputime); |
1348 | ||
6a61671b | 1349 | #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN |
b7ce2277 | 1350 | seqcount_init(&p->vtime_seqcount); |
6a61671b | 1351 | p->vtime_snap = 0; |
7098c1ea | 1352 | p->vtime_snap_whence = VTIME_INACTIVE; |
6a61671b FW |
1353 | #endif |
1354 | ||
a3a2e76c KH |
1355 | #if defined(SPLIT_RSS_COUNTING) |
1356 | memset(&p->rss_stat, 0, sizeof(p->rss_stat)); | |
1357 | #endif | |
172ba844 | 1358 | |
6976675d AV |
1359 | p->default_timer_slack_ns = current->timer_slack_ns; |
1360 | ||
5995477a | 1361 | task_io_accounting_init(&p->ioac); |
1da177e4 LT |
1362 | acct_clear_integrals(p); |
1363 | ||
f06febc9 | 1364 | posix_cpu_timers_init(p); |
1da177e4 | 1365 | |
ccbf62d8 | 1366 | p->start_time = ktime_get_ns(); |
57e0be04 | 1367 | p->real_start_time = ktime_get_boot_ns(); |
1da177e4 | 1368 | p->io_context = NULL; |
1da177e4 | 1369 | p->audit_context = NULL; |
c9e75f04 | 1370 | threadgroup_change_begin(current); |
b4f48b63 | 1371 | cgroup_fork(p); |
1da177e4 | 1372 | #ifdef CONFIG_NUMA |
846a16bf | 1373 | p->mempolicy = mpol_dup(p->mempolicy); |
fb0a685c DRO |
1374 | if (IS_ERR(p->mempolicy)) { |
1375 | retval = PTR_ERR(p->mempolicy); | |
1376 | p->mempolicy = NULL; | |
e8604cb4 | 1377 | goto bad_fork_cleanup_threadgroup_lock; |
fb0a685c | 1378 | } |
1da177e4 | 1379 | #endif |
778d3b0f MH |
1380 | #ifdef CONFIG_CPUSETS |
1381 | p->cpuset_mem_spread_rotor = NUMA_NO_NODE; | |
1382 | p->cpuset_slab_spread_rotor = NUMA_NO_NODE; | |
cc9a6c87 | 1383 | seqcount_init(&p->mems_allowed_seq); |
778d3b0f | 1384 | #endif |
de30a2b3 IM |
1385 | #ifdef CONFIG_TRACE_IRQFLAGS |
1386 | p->irq_events = 0; | |
1387 | p->hardirqs_enabled = 0; | |
1388 | p->hardirq_enable_ip = 0; | |
1389 | p->hardirq_enable_event = 0; | |
1390 | p->hardirq_disable_ip = _THIS_IP_; | |
1391 | p->hardirq_disable_event = 0; | |
1392 | p->softirqs_enabled = 1; | |
1393 | p->softirq_enable_ip = _THIS_IP_; | |
1394 | p->softirq_enable_event = 0; | |
1395 | p->softirq_disable_ip = 0; | |
1396 | p->softirq_disable_event = 0; | |
1397 | p->hardirq_context = 0; | |
1398 | p->softirq_context = 0; | |
1399 | #endif | |
8bcbde54 DH |
1400 | |
1401 | p->pagefault_disabled = 0; | |
1402 | ||
fbb9ce95 IM |
1403 | #ifdef CONFIG_LOCKDEP |
1404 | p->lockdep_depth = 0; /* no locks held yet */ | |
1405 | p->curr_chain_key = 0; | |
1406 | p->lockdep_recursion = 0; | |
1407 | #endif | |
1da177e4 | 1408 | |
408894ee IM |
1409 | #ifdef CONFIG_DEBUG_MUTEXES |
1410 | p->blocked_on = NULL; /* not blocked yet */ | |
1411 | #endif | |
cafe5635 KO |
1412 | #ifdef CONFIG_BCACHE |
1413 | p->sequential_io = 0; | |
1414 | p->sequential_io_avg = 0; | |
1415 | #endif | |
0f481406 | 1416 | |
3c90e6e9 | 1417 | /* Perform scheduler related setup. Assign this task to a CPU. */ |
aab03e05 DF |
1418 | retval = sched_fork(clone_flags, p); |
1419 | if (retval) | |
1420 | goto bad_fork_cleanup_policy; | |
6ab423e0 | 1421 | |
cdd6c482 | 1422 | retval = perf_event_init_task(p); |
6ab423e0 PZ |
1423 | if (retval) |
1424 | goto bad_fork_cleanup_policy; | |
fb0a685c DRO |
1425 | retval = audit_alloc(p); |
1426 | if (retval) | |
6c72e350 | 1427 | goto bad_fork_cleanup_perf; |
1da177e4 | 1428 | /* copy all the process information */ |
ab602f79 | 1429 | shm_init_task(p); |
fb0a685c DRO |
1430 | retval = copy_semundo(clone_flags, p); |
1431 | if (retval) | |
1da177e4 | 1432 | goto bad_fork_cleanup_audit; |
fb0a685c DRO |
1433 | retval = copy_files(clone_flags, p); |
1434 | if (retval) | |
1da177e4 | 1435 | goto bad_fork_cleanup_semundo; |
fb0a685c DRO |
1436 | retval = copy_fs(clone_flags, p); |
1437 | if (retval) | |
1da177e4 | 1438 | goto bad_fork_cleanup_files; |
fb0a685c DRO |
1439 | retval = copy_sighand(clone_flags, p); |
1440 | if (retval) | |
1da177e4 | 1441 | goto bad_fork_cleanup_fs; |
fb0a685c DRO |
1442 | retval = copy_signal(clone_flags, p); |
1443 | if (retval) | |
1da177e4 | 1444 | goto bad_fork_cleanup_sighand; |
fb0a685c DRO |
1445 | retval = copy_mm(clone_flags, p); |
1446 | if (retval) | |
1da177e4 | 1447 | goto bad_fork_cleanup_signal; |
fb0a685c DRO |
1448 | retval = copy_namespaces(clone_flags, p); |
1449 | if (retval) | |
d84f4f99 | 1450 | goto bad_fork_cleanup_mm; |
fb0a685c DRO |
1451 | retval = copy_io(clone_flags, p); |
1452 | if (retval) | |
fd0928df | 1453 | goto bad_fork_cleanup_namespaces; |
3033f14a | 1454 | retval = copy_thread_tls(clone_flags, stack_start, stack_size, p, tls); |
1da177e4 | 1455 | if (retval) |
fd0928df | 1456 | goto bad_fork_cleanup_io; |
1da177e4 | 1457 | |
425fb2b4 | 1458 | if (pid != &init_struct_pid) { |
c2b1df2e | 1459 | pid = alloc_pid(p->nsproxy->pid_ns_for_children); |
35f71bc0 MH |
1460 | if (IS_ERR(pid)) { |
1461 | retval = PTR_ERR(pid); | |
fd0928df | 1462 | goto bad_fork_cleanup_io; |
35f71bc0 | 1463 | } |
425fb2b4 PE |
1464 | } |
1465 | ||
1da177e4 LT |
1466 | p->set_child_tid = (clone_flags & CLONE_CHILD_SETTID) ? child_tidptr : NULL; |
1467 | /* | |
1468 | * Clear TID on mm_release()? | |
1469 | */ | |
fb0a685c | 1470 | p->clear_child_tid = (clone_flags & CLONE_CHILD_CLEARTID) ? child_tidptr : NULL; |
73c10101 JA |
1471 | #ifdef CONFIG_BLOCK |
1472 | p->plug = NULL; | |
1473 | #endif | |
42b2dd0a | 1474 | #ifdef CONFIG_FUTEX |
8f17d3a5 IM |
1475 | p->robust_list = NULL; |
1476 | #ifdef CONFIG_COMPAT | |
1477 | p->compat_robust_list = NULL; | |
1478 | #endif | |
c87e2837 IM |
1479 | INIT_LIST_HEAD(&p->pi_state_list); |
1480 | p->pi_state_cache = NULL; | |
42b2dd0a | 1481 | #endif |
f9a3879a GM |
1482 | /* |
1483 | * sigaltstack should be cleared when sharing the same VM | |
1484 | */ | |
1485 | if ((clone_flags & (CLONE_VM|CLONE_VFORK)) == CLONE_VM) | |
1486 | p->sas_ss_sp = p->sas_ss_size = 0; | |
1487 | ||
1da177e4 | 1488 | /* |
6580807d ON |
1489 | * Syscall tracing and stepping should be turned off in the |
1490 | * child regardless of CLONE_PTRACE. | |
1da177e4 | 1491 | */ |
6580807d | 1492 | user_disable_single_step(p); |
1da177e4 | 1493 | clear_tsk_thread_flag(p, TIF_SYSCALL_TRACE); |
ed75e8d5 LV |
1494 | #ifdef TIF_SYSCALL_EMU |
1495 | clear_tsk_thread_flag(p, TIF_SYSCALL_EMU); | |
1496 | #endif | |
9745512c | 1497 | clear_all_latency_tracing(p); |
1da177e4 | 1498 | |
1da177e4 | 1499 | /* ok, now we should be set up.. */ |
18c830df ON |
1500 | p->pid = pid_nr(pid); |
1501 | if (clone_flags & CLONE_THREAD) { | |
5f8aadd8 | 1502 | p->exit_signal = -1; |
18c830df ON |
1503 | p->group_leader = current->group_leader; |
1504 | p->tgid = current->tgid; | |
1505 | } else { | |
1506 | if (clone_flags & CLONE_PARENT) | |
1507 | p->exit_signal = current->group_leader->exit_signal; | |
1508 | else | |
1509 | p->exit_signal = (clone_flags & CSIGNAL); | |
1510 | p->group_leader = p; | |
1511 | p->tgid = p->pid; | |
1512 | } | |
5f8aadd8 | 1513 | |
9d823e8f WF |
1514 | p->nr_dirtied = 0; |
1515 | p->nr_dirtied_pause = 128 >> (PAGE_SHIFT - 10); | |
83712358 | 1516 | p->dirty_paused_when = 0; |
9d823e8f | 1517 | |
bb8cbbfe | 1518 | p->pdeath_signal = 0; |
47e65328 | 1519 | INIT_LIST_HEAD(&p->thread_group); |
158e1645 | 1520 | p->task_works = NULL; |
1da177e4 | 1521 | |
7e47682e AS |
1522 | /* |
1523 | * Ensure that the cgroup subsystem policies allow the new process to be | |
1524 | * forked. It should be noted the the new process's css_set can be changed | |
1525 | * between here and cgroup_post_fork() if an organisation operation is in | |
1526 | * progress. | |
1527 | */ | |
b53202e6 | 1528 | retval = cgroup_can_fork(p); |
7e47682e AS |
1529 | if (retval) |
1530 | goto bad_fork_free_pid; | |
1531 | ||
18c830df ON |
1532 | /* |
1533 | * Make it visible to the rest of the system, but dont wake it up yet. | |
1534 | * Need tasklist lock for parent etc handling! | |
1535 | */ | |
1da177e4 LT |
1536 | write_lock_irq(&tasklist_lock); |
1537 | ||
1da177e4 | 1538 | /* CLONE_PARENT re-uses the old parent */ |
2d5516cb | 1539 | if (clone_flags & (CLONE_PARENT|CLONE_THREAD)) { |
1da177e4 | 1540 | p->real_parent = current->real_parent; |
2d5516cb ON |
1541 | p->parent_exec_id = current->parent_exec_id; |
1542 | } else { | |
1da177e4 | 1543 | p->real_parent = current; |
2d5516cb ON |
1544 | p->parent_exec_id = current->self_exec_id; |
1545 | } | |
1da177e4 | 1546 | |
3f17da69 | 1547 | spin_lock(¤t->sighand->siglock); |
4a2c7a78 | 1548 | |
dbd95212 KC |
1549 | /* |
1550 | * Copy seccomp details explicitly here, in case they were changed | |
1551 | * before holding sighand lock. | |
1552 | */ | |
1553 | copy_seccomp(p); | |
1554 | ||
4a2c7a78 ON |
1555 | /* |
1556 | * Process group and session signals need to be delivered to just the | |
1557 | * parent before the fork or both the parent and the child after the | |
1558 | * fork. Restart if a signal comes in before we add the new process to | |
1559 | * it's process group. | |
1560 | * A fatal signal pending means that current will exit, so the new | |
1561 | * thread can't slip out of an OOM kill (or normal SIGKILL). | |
fb0a685c | 1562 | */ |
23ff4440 | 1563 | recalc_sigpending(); |
4a2c7a78 ON |
1564 | if (signal_pending(current)) { |
1565 | spin_unlock(¤t->sighand->siglock); | |
1566 | write_unlock_irq(&tasklist_lock); | |
1567 | retval = -ERESTARTNOINTR; | |
7e47682e | 1568 | goto bad_fork_cancel_cgroup; |
4a2c7a78 ON |
1569 | } |
1570 | ||
73b9ebfe | 1571 | if (likely(p->pid)) { |
4b9d33e6 | 1572 | ptrace_init_task(p, (clone_flags & CLONE_PTRACE) || trace); |
73b9ebfe | 1573 | |
81907739 | 1574 | init_task_pid(p, PIDTYPE_PID, pid); |
73b9ebfe | 1575 | if (thread_group_leader(p)) { |
81907739 ON |
1576 | init_task_pid(p, PIDTYPE_PGID, task_pgrp(current)); |
1577 | init_task_pid(p, PIDTYPE_SID, task_session(current)); | |
1578 | ||
1c4042c2 | 1579 | if (is_child_reaper(pid)) { |
17cf22c3 | 1580 | ns_of_pid(pid)->child_reaper = p; |
1c4042c2 EB |
1581 | p->signal->flags |= SIGNAL_UNKILLABLE; |
1582 | } | |
73b9ebfe | 1583 | |
fea9d175 | 1584 | p->signal->leader_pid = pid; |
9c9f4ded | 1585 | p->signal->tty = tty_kref_get(current->signal->tty); |
9cd80bbb | 1586 | list_add_tail(&p->sibling, &p->real_parent->children); |
5e85d4ab | 1587 | list_add_tail_rcu(&p->tasks, &init_task.tasks); |
81907739 ON |
1588 | attach_pid(p, PIDTYPE_PGID); |
1589 | attach_pid(p, PIDTYPE_SID); | |
909ea964 | 1590 | __this_cpu_inc(process_counts); |
80628ca0 ON |
1591 | } else { |
1592 | current->signal->nr_threads++; | |
1593 | atomic_inc(¤t->signal->live); | |
1594 | atomic_inc(¤t->signal->sigcnt); | |
80628ca0 ON |
1595 | list_add_tail_rcu(&p->thread_group, |
1596 | &p->group_leader->thread_group); | |
0c740d0a ON |
1597 | list_add_tail_rcu(&p->thread_node, |
1598 | &p->signal->thread_head); | |
73b9ebfe | 1599 | } |
81907739 | 1600 | attach_pid(p, PIDTYPE_PID); |
73b9ebfe | 1601 | nr_threads++; |
1da177e4 LT |
1602 | } |
1603 | ||
1da177e4 | 1604 | total_forks++; |
3f17da69 | 1605 | spin_unlock(¤t->sighand->siglock); |
4af4206b | 1606 | syscall_tracepoint_update(p); |
1da177e4 | 1607 | write_unlock_irq(&tasklist_lock); |
4af4206b | 1608 | |
c13cf856 | 1609 | proc_fork_connector(p); |
b53202e6 | 1610 | cgroup_post_fork(p); |
c9e75f04 | 1611 | threadgroup_change_end(current); |
cdd6c482 | 1612 | perf_event_fork(p); |
43d2b113 KH |
1613 | |
1614 | trace_task_newtask(p, clone_flags); | |
3ab67966 | 1615 | uprobe_copy_process(p, clone_flags); |
43d2b113 | 1616 | |
1da177e4 LT |
1617 | return p; |
1618 | ||
7e47682e | 1619 | bad_fork_cancel_cgroup: |
b53202e6 | 1620 | cgroup_cancel_fork(p); |
425fb2b4 PE |
1621 | bad_fork_free_pid: |
1622 | if (pid != &init_struct_pid) | |
1623 | free_pid(pid); | |
fd0928df | 1624 | bad_fork_cleanup_io: |
b69f2292 LR |
1625 | if (p->io_context) |
1626 | exit_io_context(p); | |
ab516013 | 1627 | bad_fork_cleanup_namespaces: |
444f378b | 1628 | exit_task_namespaces(p); |
1da177e4 | 1629 | bad_fork_cleanup_mm: |
c9f01245 | 1630 | if (p->mm) |
1da177e4 LT |
1631 | mmput(p->mm); |
1632 | bad_fork_cleanup_signal: | |
4ab6c083 | 1633 | if (!(clone_flags & CLONE_THREAD)) |
1c5354de | 1634 | free_signal_struct(p->signal); |
1da177e4 | 1635 | bad_fork_cleanup_sighand: |
a7e5328a | 1636 | __cleanup_sighand(p->sighand); |
1da177e4 LT |
1637 | bad_fork_cleanup_fs: |
1638 | exit_fs(p); /* blocking */ | |
1639 | bad_fork_cleanup_files: | |
1640 | exit_files(p); /* blocking */ | |
1641 | bad_fork_cleanup_semundo: | |
1642 | exit_sem(p); | |
1643 | bad_fork_cleanup_audit: | |
1644 | audit_free(p); | |
6c72e350 | 1645 | bad_fork_cleanup_perf: |
cdd6c482 | 1646 | perf_event_free_task(p); |
6c72e350 | 1647 | bad_fork_cleanup_policy: |
1da177e4 | 1648 | #ifdef CONFIG_NUMA |
f0be3d32 | 1649 | mpol_put(p->mempolicy); |
e8604cb4 | 1650 | bad_fork_cleanup_threadgroup_lock: |
1da177e4 | 1651 | #endif |
c9e75f04 | 1652 | threadgroup_change_end(current); |
35df17c5 | 1653 | delayacct_tsk_free(p); |
1da177e4 | 1654 | bad_fork_cleanup_count: |
d84f4f99 | 1655 | atomic_dec(&p->cred->user->processes); |
e0e81739 | 1656 | exit_creds(p); |
1da177e4 LT |
1657 | bad_fork_free: |
1658 | free_task(p); | |
fe7d37d1 ON |
1659 | fork_out: |
1660 | return ERR_PTR(retval); | |
1da177e4 LT |
1661 | } |
1662 | ||
f106eee1 ON |
1663 | static inline void init_idle_pids(struct pid_link *links) |
1664 | { | |
1665 | enum pid_type type; | |
1666 | ||
1667 | for (type = PIDTYPE_PID; type < PIDTYPE_MAX; ++type) { | |
1668 | INIT_HLIST_NODE(&links[type].node); /* not really needed */ | |
1669 | links[type].pid = &init_struct_pid; | |
1670 | } | |
1671 | } | |
1672 | ||
0db0628d | 1673 | struct task_struct *fork_idle(int cpu) |
1da177e4 | 1674 | { |
36c8b586 | 1675 | struct task_struct *task; |
3033f14a | 1676 | task = copy_process(CLONE_VM, 0, 0, NULL, &init_struct_pid, 0, 0); |
f106eee1 ON |
1677 | if (!IS_ERR(task)) { |
1678 | init_idle_pids(task->pids); | |
753ca4f3 | 1679 | init_idle(task, cpu); |
f106eee1 | 1680 | } |
73b9ebfe | 1681 | |
1da177e4 LT |
1682 | return task; |
1683 | } | |
1684 | ||
1da177e4 LT |
1685 | /* |
1686 | * Ok, this is the main fork-routine. | |
1687 | * | |
1688 | * It copies the process, and if successful kick-starts | |
1689 | * it and waits for it to finish using the VM if required. | |
1690 | */ | |
3033f14a | 1691 | long _do_fork(unsigned long clone_flags, |
1da177e4 | 1692 | unsigned long stack_start, |
1da177e4 LT |
1693 | unsigned long stack_size, |
1694 | int __user *parent_tidptr, | |
3033f14a JT |
1695 | int __user *child_tidptr, |
1696 | unsigned long tls) | |
1da177e4 LT |
1697 | { |
1698 | struct task_struct *p; | |
1699 | int trace = 0; | |
92476d7f | 1700 | long nr; |
1da177e4 | 1701 | |
09a05394 | 1702 | /* |
4b9d33e6 TH |
1703 | * Determine whether and which event to report to ptracer. When |
1704 | * called from kernel_thread or CLONE_UNTRACED is explicitly | |
1705 | * requested, no event is reported; otherwise, report if the event | |
1706 | * for the type of forking is enabled. | |
09a05394 | 1707 | */ |
e80d6661 | 1708 | if (!(clone_flags & CLONE_UNTRACED)) { |
4b9d33e6 TH |
1709 | if (clone_flags & CLONE_VFORK) |
1710 | trace = PTRACE_EVENT_VFORK; | |
1711 | else if ((clone_flags & CSIGNAL) != SIGCHLD) | |
1712 | trace = PTRACE_EVENT_CLONE; | |
1713 | else | |
1714 | trace = PTRACE_EVENT_FORK; | |
1715 | ||
1716 | if (likely(!ptrace_event_enabled(current, trace))) | |
1717 | trace = 0; | |
1718 | } | |
1da177e4 | 1719 | |
62e791c1 | 1720 | p = copy_process(clone_flags, stack_start, stack_size, |
3033f14a | 1721 | child_tidptr, NULL, trace, tls); |
1da177e4 LT |
1722 | /* |
1723 | * Do this prior waking up the new thread - the thread pointer | |
1724 | * might get invalid after that point, if the thread exits quickly. | |
1725 | */ | |
1726 | if (!IS_ERR(p)) { | |
1727 | struct completion vfork; | |
4e52365f | 1728 | struct pid *pid; |
1da177e4 | 1729 | |
0a16b607 MD |
1730 | trace_sched_process_fork(current, p); |
1731 | ||
4e52365f MD |
1732 | pid = get_task_pid(p, PIDTYPE_PID); |
1733 | nr = pid_vnr(pid); | |
30e49c26 PE |
1734 | |
1735 | if (clone_flags & CLONE_PARENT_SETTID) | |
1736 | put_user(nr, parent_tidptr); | |
a6f5e063 | 1737 | |
1da177e4 LT |
1738 | if (clone_flags & CLONE_VFORK) { |
1739 | p->vfork_done = &vfork; | |
1740 | init_completion(&vfork); | |
d68b46fe | 1741 | get_task_struct(p); |
1da177e4 LT |
1742 | } |
1743 | ||
3e51e3ed | 1744 | wake_up_new_task(p); |
1da177e4 | 1745 | |
4b9d33e6 TH |
1746 | /* forking complete and child started to run, tell ptracer */ |
1747 | if (unlikely(trace)) | |
4e52365f | 1748 | ptrace_event_pid(trace, pid); |
09a05394 | 1749 | |
1da177e4 | 1750 | if (clone_flags & CLONE_VFORK) { |
d68b46fe | 1751 | if (!wait_for_vfork_done(p, &vfork)) |
4e52365f | 1752 | ptrace_event_pid(PTRACE_EVENT_VFORK_DONE, pid); |
1da177e4 | 1753 | } |
4e52365f MD |
1754 | |
1755 | put_pid(pid); | |
1da177e4 | 1756 | } else { |
92476d7f | 1757 | nr = PTR_ERR(p); |
1da177e4 | 1758 | } |
92476d7f | 1759 | return nr; |
1da177e4 LT |
1760 | } |
1761 | ||
3033f14a JT |
1762 | #ifndef CONFIG_HAVE_COPY_THREAD_TLS |
1763 | /* For compatibility with architectures that call do_fork directly rather than | |
1764 | * using the syscall entry points below. */ | |
1765 | long do_fork(unsigned long clone_flags, | |
1766 | unsigned long stack_start, | |
1767 | unsigned long stack_size, | |
1768 | int __user *parent_tidptr, | |
1769 | int __user *child_tidptr) | |
1770 | { | |
1771 | return _do_fork(clone_flags, stack_start, stack_size, | |
1772 | parent_tidptr, child_tidptr, 0); | |
1773 | } | |
1774 | #endif | |
1775 | ||
2aa3a7f8 AV |
1776 | /* |
1777 | * Create a kernel thread. | |
1778 | */ | |
1779 | pid_t kernel_thread(int (*fn)(void *), void *arg, unsigned long flags) | |
1780 | { | |
3033f14a JT |
1781 | return _do_fork(flags|CLONE_VM|CLONE_UNTRACED, (unsigned long)fn, |
1782 | (unsigned long)arg, NULL, NULL, 0); | |
2aa3a7f8 | 1783 | } |
2aa3a7f8 | 1784 | |
d2125043 AV |
1785 | #ifdef __ARCH_WANT_SYS_FORK |
1786 | SYSCALL_DEFINE0(fork) | |
1787 | { | |
1788 | #ifdef CONFIG_MMU | |
3033f14a | 1789 | return _do_fork(SIGCHLD, 0, 0, NULL, NULL, 0); |
d2125043 AV |
1790 | #else |
1791 | /* can not support in nommu mode */ | |
5d59e182 | 1792 | return -EINVAL; |
d2125043 AV |
1793 | #endif |
1794 | } | |
1795 | #endif | |
1796 | ||
1797 | #ifdef __ARCH_WANT_SYS_VFORK | |
1798 | SYSCALL_DEFINE0(vfork) | |
1799 | { | |
3033f14a JT |
1800 | return _do_fork(CLONE_VFORK | CLONE_VM | SIGCHLD, 0, |
1801 | 0, NULL, NULL, 0); | |
d2125043 AV |
1802 | } |
1803 | #endif | |
1804 | ||
1805 | #ifdef __ARCH_WANT_SYS_CLONE | |
1806 | #ifdef CONFIG_CLONE_BACKWARDS | |
1807 | SYSCALL_DEFINE5(clone, unsigned long, clone_flags, unsigned long, newsp, | |
1808 | int __user *, parent_tidptr, | |
3033f14a | 1809 | unsigned long, tls, |
d2125043 AV |
1810 | int __user *, child_tidptr) |
1811 | #elif defined(CONFIG_CLONE_BACKWARDS2) | |
1812 | SYSCALL_DEFINE5(clone, unsigned long, newsp, unsigned long, clone_flags, | |
1813 | int __user *, parent_tidptr, | |
1814 | int __user *, child_tidptr, | |
3033f14a | 1815 | unsigned long, tls) |
dfa9771a MS |
1816 | #elif defined(CONFIG_CLONE_BACKWARDS3) |
1817 | SYSCALL_DEFINE6(clone, unsigned long, clone_flags, unsigned long, newsp, | |
1818 | int, stack_size, | |
1819 | int __user *, parent_tidptr, | |
1820 | int __user *, child_tidptr, | |
3033f14a | 1821 | unsigned long, tls) |
d2125043 AV |
1822 | #else |
1823 | SYSCALL_DEFINE5(clone, unsigned long, clone_flags, unsigned long, newsp, | |
1824 | int __user *, parent_tidptr, | |
1825 | int __user *, child_tidptr, | |
3033f14a | 1826 | unsigned long, tls) |
d2125043 AV |
1827 | #endif |
1828 | { | |
3033f14a | 1829 | return _do_fork(clone_flags, newsp, 0, parent_tidptr, child_tidptr, tls); |
d2125043 AV |
1830 | } |
1831 | #endif | |
1832 | ||
5fd63b30 RT |
1833 | #ifndef ARCH_MIN_MMSTRUCT_ALIGN |
1834 | #define ARCH_MIN_MMSTRUCT_ALIGN 0 | |
1835 | #endif | |
1836 | ||
51cc5068 | 1837 | static void sighand_ctor(void *data) |
aa1757f9 ON |
1838 | { |
1839 | struct sighand_struct *sighand = data; | |
1840 | ||
a35afb83 | 1841 | spin_lock_init(&sighand->siglock); |
b8fceee1 | 1842 | init_waitqueue_head(&sighand->signalfd_wqh); |
aa1757f9 ON |
1843 | } |
1844 | ||
1da177e4 LT |
1845 | void __init proc_caches_init(void) |
1846 | { | |
1847 | sighand_cachep = kmem_cache_create("sighand_cache", | |
1848 | sizeof(struct sighand_struct), 0, | |
2dff4405 | 1849 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_DESTROY_BY_RCU| |
5d097056 | 1850 | SLAB_NOTRACK|SLAB_ACCOUNT, sighand_ctor); |
1da177e4 LT |
1851 | signal_cachep = kmem_cache_create("signal_cache", |
1852 | sizeof(struct signal_struct), 0, | |
5d097056 VD |
1853 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK|SLAB_ACCOUNT, |
1854 | NULL); | |
20c2df83 | 1855 | files_cachep = kmem_cache_create("files_cache", |
1da177e4 | 1856 | sizeof(struct files_struct), 0, |
5d097056 VD |
1857 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK|SLAB_ACCOUNT, |
1858 | NULL); | |
20c2df83 | 1859 | fs_cachep = kmem_cache_create("fs_cache", |
1da177e4 | 1860 | sizeof(struct fs_struct), 0, |
5d097056 VD |
1861 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK|SLAB_ACCOUNT, |
1862 | NULL); | |
6345d24d LT |
1863 | /* |
1864 | * FIXME! The "sizeof(struct mm_struct)" currently includes the | |
1865 | * whole struct cpumask for the OFFSTACK case. We could change | |
1866 | * this to *only* allocate as much of it as required by the | |
1867 | * maximum number of CPU's we can ever have. The cpumask_allocation | |
1868 | * is at the end of the structure, exactly for that reason. | |
1869 | */ | |
1da177e4 | 1870 | mm_cachep = kmem_cache_create("mm_struct", |
5fd63b30 | 1871 | sizeof(struct mm_struct), ARCH_MIN_MMSTRUCT_ALIGN, |
5d097056 VD |
1872 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK|SLAB_ACCOUNT, |
1873 | NULL); | |
1874 | vm_area_cachep = KMEM_CACHE(vm_area_struct, SLAB_PANIC|SLAB_ACCOUNT); | |
8feae131 | 1875 | mmap_init(); |
66577193 | 1876 | nsproxy_cache_init(); |
1da177e4 | 1877 | } |
cf2e340f | 1878 | |
cf2e340f | 1879 | /* |
9bfb23fc | 1880 | * Check constraints on flags passed to the unshare system call. |
cf2e340f | 1881 | */ |
9bfb23fc | 1882 | static int check_unshare_flags(unsigned long unshare_flags) |
cf2e340f | 1883 | { |
9bfb23fc ON |
1884 | if (unshare_flags & ~(CLONE_THREAD|CLONE_FS|CLONE_NEWNS|CLONE_SIGHAND| |
1885 | CLONE_VM|CLONE_FILES|CLONE_SYSVSEM| | |
50804fe3 | 1886 | CLONE_NEWUTS|CLONE_NEWIPC|CLONE_NEWNET| |
b2e0d987 | 1887 | CLONE_NEWUSER|CLONE_NEWPID)) |
9bfb23fc | 1888 | return -EINVAL; |
cf2e340f | 1889 | /* |
12c641ab EB |
1890 | * Not implemented, but pretend it works if there is nothing |
1891 | * to unshare. Note that unsharing the address space or the | |
1892 | * signal handlers also need to unshare the signal queues (aka | |
1893 | * CLONE_THREAD). | |
cf2e340f | 1894 | */ |
9bfb23fc | 1895 | if (unshare_flags & (CLONE_THREAD | CLONE_SIGHAND | CLONE_VM)) { |
12c641ab EB |
1896 | if (!thread_group_empty(current)) |
1897 | return -EINVAL; | |
1898 | } | |
1899 | if (unshare_flags & (CLONE_SIGHAND | CLONE_VM)) { | |
1900 | if (atomic_read(¤t->sighand->count) > 1) | |
1901 | return -EINVAL; | |
1902 | } | |
1903 | if (unshare_flags & CLONE_VM) { | |
1904 | if (!current_is_single_threaded()) | |
9bfb23fc ON |
1905 | return -EINVAL; |
1906 | } | |
cf2e340f JD |
1907 | |
1908 | return 0; | |
1909 | } | |
1910 | ||
1911 | /* | |
99d1419d | 1912 | * Unshare the filesystem structure if it is being shared |
cf2e340f JD |
1913 | */ |
1914 | static int unshare_fs(unsigned long unshare_flags, struct fs_struct **new_fsp) | |
1915 | { | |
1916 | struct fs_struct *fs = current->fs; | |
1917 | ||
498052bb AV |
1918 | if (!(unshare_flags & CLONE_FS) || !fs) |
1919 | return 0; | |
1920 | ||
1921 | /* don't need lock here; in the worst case we'll do useless copy */ | |
1922 | if (fs->users == 1) | |
1923 | return 0; | |
1924 | ||
1925 | *new_fsp = copy_fs_struct(fs); | |
1926 | if (!*new_fsp) | |
1927 | return -ENOMEM; | |
cf2e340f JD |
1928 | |
1929 | return 0; | |
1930 | } | |
1931 | ||
cf2e340f | 1932 | /* |
a016f338 | 1933 | * Unshare file descriptor table if it is being shared |
cf2e340f JD |
1934 | */ |
1935 | static int unshare_fd(unsigned long unshare_flags, struct files_struct **new_fdp) | |
1936 | { | |
1937 | struct files_struct *fd = current->files; | |
a016f338 | 1938 | int error = 0; |
cf2e340f JD |
1939 | |
1940 | if ((unshare_flags & CLONE_FILES) && | |
a016f338 JD |
1941 | (fd && atomic_read(&fd->count) > 1)) { |
1942 | *new_fdp = dup_fd(fd, &error); | |
1943 | if (!*new_fdp) | |
1944 | return error; | |
1945 | } | |
cf2e340f JD |
1946 | |
1947 | return 0; | |
1948 | } | |
1949 | ||
cf2e340f JD |
1950 | /* |
1951 | * unshare allows a process to 'unshare' part of the process | |
1952 | * context which was originally shared using clone. copy_* | |
1953 | * functions used by do_fork() cannot be used here directly | |
1954 | * because they modify an inactive task_struct that is being | |
1955 | * constructed. Here we are modifying the current, active, | |
1956 | * task_struct. | |
1957 | */ | |
6559eed8 | 1958 | SYSCALL_DEFINE1(unshare, unsigned long, unshare_flags) |
cf2e340f | 1959 | { |
cf2e340f | 1960 | struct fs_struct *fs, *new_fs = NULL; |
cf2e340f | 1961 | struct files_struct *fd, *new_fd = NULL; |
b2e0d987 | 1962 | struct cred *new_cred = NULL; |
cf7b708c | 1963 | struct nsproxy *new_nsproxy = NULL; |
9edff4ab | 1964 | int do_sysvsem = 0; |
9bfb23fc | 1965 | int err; |
cf2e340f | 1966 | |
b2e0d987 | 1967 | /* |
faf00da5 EB |
1968 | * If unsharing a user namespace must also unshare the thread group |
1969 | * and unshare the filesystem root and working directories. | |
b2e0d987 EB |
1970 | */ |
1971 | if (unshare_flags & CLONE_NEWUSER) | |
e66eded8 | 1972 | unshare_flags |= CLONE_THREAD | CLONE_FS; |
50804fe3 EB |
1973 | /* |
1974 | * If unsharing vm, must also unshare signal handlers. | |
1975 | */ | |
1976 | if (unshare_flags & CLONE_VM) | |
1977 | unshare_flags |= CLONE_SIGHAND; | |
12c641ab EB |
1978 | /* |
1979 | * If unsharing a signal handlers, must also unshare the signal queues. | |
1980 | */ | |
1981 | if (unshare_flags & CLONE_SIGHAND) | |
1982 | unshare_flags |= CLONE_THREAD; | |
9bfb23fc ON |
1983 | /* |
1984 | * If unsharing namespace, must also unshare filesystem information. | |
1985 | */ | |
1986 | if (unshare_flags & CLONE_NEWNS) | |
1987 | unshare_flags |= CLONE_FS; | |
50804fe3 EB |
1988 | |
1989 | err = check_unshare_flags(unshare_flags); | |
1990 | if (err) | |
1991 | goto bad_unshare_out; | |
6013f67f MS |
1992 | /* |
1993 | * CLONE_NEWIPC must also detach from the undolist: after switching | |
1994 | * to a new ipc namespace, the semaphore arrays from the old | |
1995 | * namespace are unreachable. | |
1996 | */ | |
1997 | if (unshare_flags & (CLONE_NEWIPC|CLONE_SYSVSEM)) | |
9edff4ab | 1998 | do_sysvsem = 1; |
fb0a685c DRO |
1999 | err = unshare_fs(unshare_flags, &new_fs); |
2000 | if (err) | |
9bfb23fc | 2001 | goto bad_unshare_out; |
fb0a685c DRO |
2002 | err = unshare_fd(unshare_flags, &new_fd); |
2003 | if (err) | |
9bfb23fc | 2004 | goto bad_unshare_cleanup_fs; |
b2e0d987 | 2005 | err = unshare_userns(unshare_flags, &new_cred); |
fb0a685c | 2006 | if (err) |
9edff4ab | 2007 | goto bad_unshare_cleanup_fd; |
b2e0d987 EB |
2008 | err = unshare_nsproxy_namespaces(unshare_flags, &new_nsproxy, |
2009 | new_cred, new_fs); | |
2010 | if (err) | |
2011 | goto bad_unshare_cleanup_cred; | |
c0b2fc31 | 2012 | |
b2e0d987 | 2013 | if (new_fs || new_fd || do_sysvsem || new_cred || new_nsproxy) { |
9edff4ab MS |
2014 | if (do_sysvsem) { |
2015 | /* | |
2016 | * CLONE_SYSVSEM is equivalent to sys_exit(). | |
2017 | */ | |
2018 | exit_sem(current); | |
2019 | } | |
ab602f79 JM |
2020 | if (unshare_flags & CLONE_NEWIPC) { |
2021 | /* Orphan segments in old ns (see sem above). */ | |
2022 | exit_shm(current); | |
2023 | shm_init_task(current); | |
2024 | } | |
ab516013 | 2025 | |
6f977e6b | 2026 | if (new_nsproxy) |
cf7b708c | 2027 | switch_task_namespaces(current, new_nsproxy); |
cf2e340f | 2028 | |
cf7b708c PE |
2029 | task_lock(current); |
2030 | ||
cf2e340f JD |
2031 | if (new_fs) { |
2032 | fs = current->fs; | |
2a4419b5 | 2033 | spin_lock(&fs->lock); |
cf2e340f | 2034 | current->fs = new_fs; |
498052bb AV |
2035 | if (--fs->users) |
2036 | new_fs = NULL; | |
2037 | else | |
2038 | new_fs = fs; | |
2a4419b5 | 2039 | spin_unlock(&fs->lock); |
cf2e340f JD |
2040 | } |
2041 | ||
cf2e340f JD |
2042 | if (new_fd) { |
2043 | fd = current->files; | |
2044 | current->files = new_fd; | |
2045 | new_fd = fd; | |
2046 | } | |
2047 | ||
2048 | task_unlock(current); | |
b2e0d987 EB |
2049 | |
2050 | if (new_cred) { | |
2051 | /* Install the new user namespace */ | |
2052 | commit_creds(new_cred); | |
2053 | new_cred = NULL; | |
2054 | } | |
cf2e340f JD |
2055 | } |
2056 | ||
b2e0d987 EB |
2057 | bad_unshare_cleanup_cred: |
2058 | if (new_cred) | |
2059 | put_cred(new_cred); | |
cf2e340f JD |
2060 | bad_unshare_cleanup_fd: |
2061 | if (new_fd) | |
2062 | put_files_struct(new_fd); | |
2063 | ||
cf2e340f JD |
2064 | bad_unshare_cleanup_fs: |
2065 | if (new_fs) | |
498052bb | 2066 | free_fs_struct(new_fs); |
cf2e340f | 2067 | |
cf2e340f JD |
2068 | bad_unshare_out: |
2069 | return err; | |
2070 | } | |
3b125388 AV |
2071 | |
2072 | /* | |
2073 | * Helper to unshare the files of the current task. | |
2074 | * We don't want to expose copy_files internals to | |
2075 | * the exec layer of the kernel. | |
2076 | */ | |
2077 | ||
2078 | int unshare_files(struct files_struct **displaced) | |
2079 | { | |
2080 | struct task_struct *task = current; | |
50704516 | 2081 | struct files_struct *copy = NULL; |
3b125388 AV |
2082 | int error; |
2083 | ||
2084 | error = unshare_fd(CLONE_FILES, ©); | |
2085 | if (error || !copy) { | |
2086 | *displaced = NULL; | |
2087 | return error; | |
2088 | } | |
2089 | *displaced = task->files; | |
2090 | task_lock(task); | |
2091 | task->files = copy; | |
2092 | task_unlock(task); | |
2093 | return 0; | |
2094 | } | |
16db3d3f HS |
2095 | |
2096 | int sysctl_max_threads(struct ctl_table *table, int write, | |
2097 | void __user *buffer, size_t *lenp, loff_t *ppos) | |
2098 | { | |
2099 | struct ctl_table t; | |
2100 | int ret; | |
2101 | int threads = max_threads; | |
2102 | int min = MIN_THREADS; | |
2103 | int max = MAX_THREADS; | |
2104 | ||
2105 | t = *table; | |
2106 | t.data = &threads; | |
2107 | t.extra1 = &min; | |
2108 | t.extra2 = &max; | |
2109 | ||
2110 | ret = proc_dointvec_minmax(&t, write, buffer, lenp, ppos); | |
2111 | if (ret || !write) | |
2112 | return ret; | |
2113 | ||
2114 | set_max_threads(threads); | |
2115 | ||
2116 | return 0; | |
2117 | } |