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> |
1da177e4 LT |
77 | |
78 | #include <asm/pgtable.h> | |
79 | #include <asm/pgalloc.h> | |
80 | #include <asm/uaccess.h> | |
81 | #include <asm/mmu_context.h> | |
82 | #include <asm/cacheflush.h> | |
83 | #include <asm/tlbflush.h> | |
84 | ||
ad8d75ff SR |
85 | #include <trace/events/sched.h> |
86 | ||
43d2b113 KH |
87 | #define CREATE_TRACE_POINTS |
88 | #include <trace/events/task.h> | |
89 | ||
ac1b398d HS |
90 | /* |
91 | * Minimum number of threads to boot the kernel | |
92 | */ | |
93 | #define MIN_THREADS 20 | |
94 | ||
95 | /* | |
96 | * Maximum number of threads | |
97 | */ | |
98 | #define MAX_THREADS FUTEX_TID_MASK | |
99 | ||
1da177e4 LT |
100 | /* |
101 | * Protected counters by write_lock_irq(&tasklist_lock) | |
102 | */ | |
103 | unsigned long total_forks; /* Handle normal Linux uptimes. */ | |
fb0a685c | 104 | int nr_threads; /* The idle threads do not count.. */ |
1da177e4 LT |
105 | |
106 | int max_threads; /* tunable limit on nr_threads */ | |
107 | ||
108 | DEFINE_PER_CPU(unsigned long, process_counts) = 0; | |
109 | ||
c59923a1 | 110 | __cacheline_aligned DEFINE_RWLOCK(tasklist_lock); /* outer */ |
db1466b3 PM |
111 | |
112 | #ifdef CONFIG_PROVE_RCU | |
113 | int lockdep_tasklist_lock_is_held(void) | |
114 | { | |
115 | return lockdep_is_held(&tasklist_lock); | |
116 | } | |
117 | EXPORT_SYMBOL_GPL(lockdep_tasklist_lock_is_held); | |
118 | #endif /* #ifdef CONFIG_PROVE_RCU */ | |
1da177e4 LT |
119 | |
120 | int nr_processes(void) | |
121 | { | |
122 | int cpu; | |
123 | int total = 0; | |
124 | ||
1d510750 | 125 | for_each_possible_cpu(cpu) |
1da177e4 LT |
126 | total += per_cpu(process_counts, cpu); |
127 | ||
128 | return total; | |
129 | } | |
130 | ||
f19b9f74 AM |
131 | void __weak arch_release_task_struct(struct task_struct *tsk) |
132 | { | |
133 | } | |
134 | ||
f5e10287 | 135 | #ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR |
e18b890b | 136 | static struct kmem_cache *task_struct_cachep; |
41101809 TG |
137 | |
138 | static inline struct task_struct *alloc_task_struct_node(int node) | |
139 | { | |
140 | return kmem_cache_alloc_node(task_struct_cachep, GFP_KERNEL, node); | |
141 | } | |
142 | ||
41101809 TG |
143 | static inline void free_task_struct(struct task_struct *tsk) |
144 | { | |
41101809 TG |
145 | kmem_cache_free(task_struct_cachep, tsk); |
146 | } | |
1da177e4 LT |
147 | #endif |
148 | ||
f19b9f74 AM |
149 | void __weak arch_release_thread_info(struct thread_info *ti) |
150 | { | |
151 | } | |
152 | ||
f5e10287 | 153 | #ifndef CONFIG_ARCH_THREAD_INFO_ALLOCATOR |
41101809 | 154 | |
0d15d74a TG |
155 | /* |
156 | * Allocate pages if THREAD_SIZE is >= PAGE_SIZE, otherwise use a | |
157 | * kmemcache based allocator. | |
158 | */ | |
159 | # if THREAD_SIZE >= PAGE_SIZE | |
b6a84016 ED |
160 | static struct thread_info *alloc_thread_info_node(struct task_struct *tsk, |
161 | int node) | |
b69c49b7 | 162 | { |
52383431 VD |
163 | struct page *page = alloc_kmem_pages_node(node, THREADINFO_GFP, |
164 | THREAD_SIZE_ORDER); | |
b6a84016 ED |
165 | |
166 | return page ? page_address(page) : NULL; | |
b69c49b7 FT |
167 | } |
168 | ||
169 | static inline void free_thread_info(struct thread_info *ti) | |
170 | { | |
52383431 | 171 | free_kmem_pages((unsigned long)ti, THREAD_SIZE_ORDER); |
b69c49b7 | 172 | } |
0d15d74a TG |
173 | # else |
174 | static struct kmem_cache *thread_info_cache; | |
175 | ||
176 | static struct thread_info *alloc_thread_info_node(struct task_struct *tsk, | |
177 | int node) | |
178 | { | |
179 | return kmem_cache_alloc_node(thread_info_cache, THREADINFO_GFP, node); | |
180 | } | |
181 | ||
182 | static void free_thread_info(struct thread_info *ti) | |
183 | { | |
0d15d74a TG |
184 | kmem_cache_free(thread_info_cache, ti); |
185 | } | |
186 | ||
187 | void thread_info_cache_init(void) | |
188 | { | |
189 | thread_info_cache = kmem_cache_create("thread_info", THREAD_SIZE, | |
190 | THREAD_SIZE, 0, NULL); | |
191 | BUG_ON(thread_info_cache == NULL); | |
192 | } | |
193 | # endif | |
b69c49b7 FT |
194 | #endif |
195 | ||
1da177e4 | 196 | /* SLAB cache for signal_struct structures (tsk->signal) */ |
e18b890b | 197 | static struct kmem_cache *signal_cachep; |
1da177e4 LT |
198 | |
199 | /* SLAB cache for sighand_struct structures (tsk->sighand) */ | |
e18b890b | 200 | struct kmem_cache *sighand_cachep; |
1da177e4 LT |
201 | |
202 | /* SLAB cache for files_struct structures (tsk->files) */ | |
e18b890b | 203 | struct kmem_cache *files_cachep; |
1da177e4 LT |
204 | |
205 | /* SLAB cache for fs_struct structures (tsk->fs) */ | |
e18b890b | 206 | struct kmem_cache *fs_cachep; |
1da177e4 LT |
207 | |
208 | /* SLAB cache for vm_area_struct structures */ | |
e18b890b | 209 | struct kmem_cache *vm_area_cachep; |
1da177e4 LT |
210 | |
211 | /* SLAB cache for mm_struct structures (tsk->mm) */ | |
e18b890b | 212 | static struct kmem_cache *mm_cachep; |
1da177e4 | 213 | |
c6a7f572 KM |
214 | static void account_kernel_stack(struct thread_info *ti, int account) |
215 | { | |
216 | struct zone *zone = page_zone(virt_to_page(ti)); | |
217 | ||
218 | mod_zone_page_state(zone, NR_KERNEL_STACK, account); | |
219 | } | |
220 | ||
1da177e4 LT |
221 | void free_task(struct task_struct *tsk) |
222 | { | |
c6a7f572 | 223 | account_kernel_stack(tsk->stack, -1); |
f19b9f74 | 224 | arch_release_thread_info(tsk->stack); |
f7e4217b | 225 | free_thread_info(tsk->stack); |
23f78d4a | 226 | rt_mutex_debug_task_free(tsk); |
fb52607a | 227 | ftrace_graph_exit_task(tsk); |
e2cfabdf | 228 | put_seccomp_filter(tsk); |
f19b9f74 | 229 | arch_release_task_struct(tsk); |
1da177e4 LT |
230 | free_task_struct(tsk); |
231 | } | |
232 | EXPORT_SYMBOL(free_task); | |
233 | ||
ea6d290c ON |
234 | static inline void free_signal_struct(struct signal_struct *sig) |
235 | { | |
97101eb4 | 236 | taskstats_tgid_free(sig); |
1c5354de | 237 | sched_autogroup_exit(sig); |
ea6d290c ON |
238 | kmem_cache_free(signal_cachep, sig); |
239 | } | |
240 | ||
241 | static inline void put_signal_struct(struct signal_struct *sig) | |
242 | { | |
1c5354de | 243 | if (atomic_dec_and_test(&sig->sigcnt)) |
ea6d290c ON |
244 | free_signal_struct(sig); |
245 | } | |
246 | ||
158d9ebd | 247 | void __put_task_struct(struct task_struct *tsk) |
1da177e4 | 248 | { |
270f722d | 249 | WARN_ON(!tsk->exit_state); |
1da177e4 LT |
250 | WARN_ON(atomic_read(&tsk->usage)); |
251 | WARN_ON(tsk == current); | |
252 | ||
156654f4 | 253 | task_numa_free(tsk); |
1a2a4d06 | 254 | security_task_free(tsk); |
e0e81739 | 255 | exit_creds(tsk); |
35df17c5 | 256 | delayacct_tsk_free(tsk); |
ea6d290c | 257 | put_signal_struct(tsk->signal); |
1da177e4 LT |
258 | |
259 | if (!profile_handoff_task(tsk)) | |
260 | free_task(tsk); | |
261 | } | |
77c100c8 | 262 | EXPORT_SYMBOL_GPL(__put_task_struct); |
1da177e4 | 263 | |
6c0a9fa6 | 264 | void __init __weak arch_task_cache_init(void) { } |
61c4628b | 265 | |
ff691f6e HS |
266 | /* |
267 | * set_max_threads | |
268 | */ | |
269 | static void set_max_threads(void) | |
270 | { | |
ac1b398d | 271 | u64 threads; |
ff691f6e HS |
272 | |
273 | /* | |
ac1b398d HS |
274 | * The number of threads shall be limited such that the thread |
275 | * structures may only consume a small part of the available memory. | |
ff691f6e | 276 | */ |
ac1b398d HS |
277 | if (fls64(totalram_pages) + fls64(PAGE_SIZE) > 64) |
278 | threads = MAX_THREADS; | |
279 | else | |
280 | threads = div64_u64((u64) totalram_pages * (u64) PAGE_SIZE, | |
281 | (u64) THREAD_SIZE * 8UL); | |
282 | ||
283 | max_threads = clamp_t(u64, threads, MIN_THREADS, MAX_THREADS); | |
ff691f6e HS |
284 | } |
285 | ||
286 | void __init fork_init(void) | |
1da177e4 | 287 | { |
f5e10287 | 288 | #ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR |
1da177e4 LT |
289 | #ifndef ARCH_MIN_TASKALIGN |
290 | #define ARCH_MIN_TASKALIGN L1_CACHE_BYTES | |
291 | #endif | |
292 | /* create a slab on which task_structs can be allocated */ | |
293 | task_struct_cachep = | |
294 | kmem_cache_create("task_struct", sizeof(struct task_struct), | |
2dff4405 | 295 | ARCH_MIN_TASKALIGN, SLAB_PANIC | SLAB_NOTRACK, NULL); |
1da177e4 LT |
296 | #endif |
297 | ||
61c4628b SS |
298 | /* do the arch specific task caches init */ |
299 | arch_task_cache_init(); | |
300 | ||
ff691f6e | 301 | set_max_threads(); |
1da177e4 LT |
302 | |
303 | init_task.signal->rlim[RLIMIT_NPROC].rlim_cur = max_threads/2; | |
304 | init_task.signal->rlim[RLIMIT_NPROC].rlim_max = max_threads/2; | |
305 | init_task.signal->rlim[RLIMIT_SIGPENDING] = | |
306 | init_task.signal->rlim[RLIMIT_NPROC]; | |
307 | } | |
308 | ||
52f5684c | 309 | int __weak arch_dup_task_struct(struct task_struct *dst, |
61c4628b SS |
310 | struct task_struct *src) |
311 | { | |
312 | *dst = *src; | |
313 | return 0; | |
314 | } | |
315 | ||
d4311ff1 AT |
316 | void set_task_stack_end_magic(struct task_struct *tsk) |
317 | { | |
318 | unsigned long *stackend; | |
319 | ||
320 | stackend = end_of_stack(tsk); | |
321 | *stackend = STACK_END_MAGIC; /* for overflow detection */ | |
322 | } | |
323 | ||
1da177e4 LT |
324 | static struct task_struct *dup_task_struct(struct task_struct *orig) |
325 | { | |
326 | struct task_struct *tsk; | |
327 | struct thread_info *ti; | |
207205a2 | 328 | int node = tsk_fork_get_node(orig); |
3e26c149 | 329 | int err; |
1da177e4 | 330 | |
504f52b5 | 331 | tsk = alloc_task_struct_node(node); |
1da177e4 LT |
332 | if (!tsk) |
333 | return NULL; | |
334 | ||
b6a84016 | 335 | ti = alloc_thread_info_node(tsk, node); |
f19b9f74 AM |
336 | if (!ti) |
337 | goto free_tsk; | |
1da177e4 | 338 | |
fb0a685c | 339 | err = arch_dup_task_struct(tsk, orig); |
164c33c6 | 340 | if (err) |
f19b9f74 | 341 | goto free_ti; |
164c33c6 | 342 | |
87bec58a | 343 | tsk->stack = ti; |
dbd95212 KC |
344 | #ifdef CONFIG_SECCOMP |
345 | /* | |
346 | * We must handle setting up seccomp filters once we're under | |
347 | * the sighand lock in case orig has changed between now and | |
348 | * then. Until then, filter must be NULL to avoid messing up | |
349 | * the usage counts on the error path calling free_task. | |
350 | */ | |
351 | tsk->seccomp.filter = NULL; | |
352 | #endif | |
87bec58a AM |
353 | |
354 | setup_thread_stack(tsk, orig); | |
8e7cac79 | 355 | clear_user_return_notifier(tsk); |
f26f9aff | 356 | clear_tsk_need_resched(tsk); |
d4311ff1 | 357 | set_task_stack_end_magic(tsk); |
1da177e4 | 358 | |
0a425405 AV |
359 | #ifdef CONFIG_CC_STACKPROTECTOR |
360 | tsk->stack_canary = get_random_int(); | |
361 | #endif | |
362 | ||
fb0a685c DRO |
363 | /* |
364 | * One for us, one for whoever does the "release_task()" (usually | |
365 | * parent) | |
366 | */ | |
367 | atomic_set(&tsk->usage, 2); | |
6c5c9341 | 368 | #ifdef CONFIG_BLK_DEV_IO_TRACE |
2056a782 | 369 | tsk->btrace_seq = 0; |
6c5c9341 | 370 | #endif |
a0aa7f68 | 371 | tsk->splice_pipe = NULL; |
5640f768 | 372 | tsk->task_frag.page = NULL; |
c6a7f572 KM |
373 | |
374 | account_kernel_stack(ti, 1); | |
375 | ||
1da177e4 | 376 | return tsk; |
61c4628b | 377 | |
f19b9f74 | 378 | free_ti: |
61c4628b | 379 | free_thread_info(ti); |
f19b9f74 | 380 | free_tsk: |
61c4628b SS |
381 | free_task_struct(tsk); |
382 | return NULL; | |
1da177e4 LT |
383 | } |
384 | ||
385 | #ifdef CONFIG_MMU | |
a39bc516 | 386 | static int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm) |
1da177e4 | 387 | { |
297c5eee | 388 | struct vm_area_struct *mpnt, *tmp, *prev, **pprev; |
1da177e4 LT |
389 | struct rb_node **rb_link, *rb_parent; |
390 | int retval; | |
391 | unsigned long charge; | |
1da177e4 | 392 | |
32cdba1e | 393 | uprobe_start_dup_mmap(); |
1da177e4 | 394 | down_write(&oldmm->mmap_sem); |
ec8c0446 | 395 | flush_cache_dup_mm(oldmm); |
f8ac4ec9 | 396 | uprobe_dup_mmap(oldmm, mm); |
ad339451 IM |
397 | /* |
398 | * Not linked in yet - no deadlock potential: | |
399 | */ | |
400 | down_write_nested(&mm->mmap_sem, SINGLE_DEPTH_NESTING); | |
7ee78232 | 401 | |
4f7d4614 VD |
402 | mm->total_vm = oldmm->total_vm; |
403 | mm->shared_vm = oldmm->shared_vm; | |
404 | mm->exec_vm = oldmm->exec_vm; | |
405 | mm->stack_vm = oldmm->stack_vm; | |
406 | ||
1da177e4 LT |
407 | rb_link = &mm->mm_rb.rb_node; |
408 | rb_parent = NULL; | |
409 | pprev = &mm->mmap; | |
f8af4da3 | 410 | retval = ksm_fork(mm, oldmm); |
ba76149f AA |
411 | if (retval) |
412 | goto out; | |
413 | retval = khugepaged_fork(mm, oldmm); | |
f8af4da3 HD |
414 | if (retval) |
415 | goto out; | |
1da177e4 | 416 | |
297c5eee | 417 | prev = NULL; |
fd3e42fc | 418 | for (mpnt = oldmm->mmap; mpnt; mpnt = mpnt->vm_next) { |
1da177e4 LT |
419 | struct file *file; |
420 | ||
421 | if (mpnt->vm_flags & VM_DONTCOPY) { | |
ab50b8ed | 422 | vm_stat_account(mm, mpnt->vm_flags, mpnt->vm_file, |
44de9d0c | 423 | -vma_pages(mpnt)); |
1da177e4 LT |
424 | continue; |
425 | } | |
426 | charge = 0; | |
427 | if (mpnt->vm_flags & VM_ACCOUNT) { | |
b2412b7f HS |
428 | unsigned long len = vma_pages(mpnt); |
429 | ||
191c5424 | 430 | if (security_vm_enough_memory_mm(oldmm, len)) /* sic */ |
1da177e4 LT |
431 | goto fail_nomem; |
432 | charge = len; | |
433 | } | |
e94b1766 | 434 | tmp = kmem_cache_alloc(vm_area_cachep, GFP_KERNEL); |
1da177e4 LT |
435 | if (!tmp) |
436 | goto fail_nomem; | |
437 | *tmp = *mpnt; | |
5beb4930 | 438 | INIT_LIST_HEAD(&tmp->anon_vma_chain); |
ef0855d3 ON |
439 | retval = vma_dup_policy(mpnt, tmp); |
440 | if (retval) | |
1da177e4 | 441 | goto fail_nomem_policy; |
a247c3a9 | 442 | tmp->vm_mm = mm; |
5beb4930 RR |
443 | if (anon_vma_fork(tmp, mpnt)) |
444 | goto fail_nomem_anon_vma_fork; | |
1da177e4 | 445 | tmp->vm_flags &= ~VM_LOCKED; |
297c5eee | 446 | tmp->vm_next = tmp->vm_prev = NULL; |
1da177e4 LT |
447 | file = tmp->vm_file; |
448 | if (file) { | |
496ad9aa | 449 | struct inode *inode = file_inode(file); |
b88ed205 HD |
450 | struct address_space *mapping = file->f_mapping; |
451 | ||
1da177e4 LT |
452 | get_file(file); |
453 | if (tmp->vm_flags & VM_DENYWRITE) | |
454 | atomic_dec(&inode->i_writecount); | |
83cde9e8 | 455 | i_mmap_lock_write(mapping); |
b88ed205 | 456 | if (tmp->vm_flags & VM_SHARED) |
4bb5f5d9 | 457 | atomic_inc(&mapping->i_mmap_writable); |
b88ed205 HD |
458 | flush_dcache_mmap_lock(mapping); |
459 | /* insert tmp into the share list, just after mpnt */ | |
27ba0644 KS |
460 | vma_interval_tree_insert_after(tmp, mpnt, |
461 | &mapping->i_mmap); | |
b88ed205 | 462 | flush_dcache_mmap_unlock(mapping); |
83cde9e8 | 463 | i_mmap_unlock_write(mapping); |
1da177e4 LT |
464 | } |
465 | ||
a1e78772 MG |
466 | /* |
467 | * Clear hugetlb-related page reserves for children. This only | |
468 | * affects MAP_PRIVATE mappings. Faults generated by the child | |
469 | * are not guaranteed to succeed, even if read-only | |
470 | */ | |
471 | if (is_vm_hugetlb_page(tmp)) | |
472 | reset_vma_resv_huge_pages(tmp); | |
473 | ||
1da177e4 | 474 | /* |
7ee78232 | 475 | * Link in the new vma and copy the page table entries. |
1da177e4 | 476 | */ |
1da177e4 LT |
477 | *pprev = tmp; |
478 | pprev = &tmp->vm_next; | |
297c5eee LT |
479 | tmp->vm_prev = prev; |
480 | prev = tmp; | |
1da177e4 LT |
481 | |
482 | __vma_link_rb(mm, tmp, rb_link, rb_parent); | |
483 | rb_link = &tmp->vm_rb.rb_right; | |
484 | rb_parent = &tmp->vm_rb; | |
485 | ||
486 | mm->map_count++; | |
0b0db14c | 487 | retval = copy_page_range(mm, oldmm, mpnt); |
1da177e4 LT |
488 | |
489 | if (tmp->vm_ops && tmp->vm_ops->open) | |
490 | tmp->vm_ops->open(tmp); | |
491 | ||
492 | if (retval) | |
493 | goto out; | |
494 | } | |
d6dd61c8 JF |
495 | /* a new mm has just been created */ |
496 | arch_dup_mmap(oldmm, mm); | |
1da177e4 | 497 | retval = 0; |
1da177e4 | 498 | out: |
7ee78232 | 499 | up_write(&mm->mmap_sem); |
fd3e42fc | 500 | flush_tlb_mm(oldmm); |
1da177e4 | 501 | up_write(&oldmm->mmap_sem); |
32cdba1e | 502 | uprobe_end_dup_mmap(); |
1da177e4 | 503 | return retval; |
5beb4930 | 504 | fail_nomem_anon_vma_fork: |
ef0855d3 | 505 | mpol_put(vma_policy(tmp)); |
1da177e4 LT |
506 | fail_nomem_policy: |
507 | kmem_cache_free(vm_area_cachep, tmp); | |
508 | fail_nomem: | |
509 | retval = -ENOMEM; | |
510 | vm_unacct_memory(charge); | |
511 | goto out; | |
512 | } | |
513 | ||
fb0a685c | 514 | static inline int mm_alloc_pgd(struct mm_struct *mm) |
1da177e4 LT |
515 | { |
516 | mm->pgd = pgd_alloc(mm); | |
517 | if (unlikely(!mm->pgd)) | |
518 | return -ENOMEM; | |
519 | return 0; | |
520 | } | |
521 | ||
fb0a685c | 522 | static inline void mm_free_pgd(struct mm_struct *mm) |
1da177e4 | 523 | { |
5e541973 | 524 | pgd_free(mm, mm->pgd); |
1da177e4 LT |
525 | } |
526 | #else | |
527 | #define dup_mmap(mm, oldmm) (0) | |
528 | #define mm_alloc_pgd(mm) (0) | |
529 | #define mm_free_pgd(mm) | |
530 | #endif /* CONFIG_MMU */ | |
531 | ||
23ff4440 | 532 | __cacheline_aligned_in_smp DEFINE_SPINLOCK(mmlist_lock); |
1da177e4 | 533 | |
e94b1766 | 534 | #define allocate_mm() (kmem_cache_alloc(mm_cachep, GFP_KERNEL)) |
1da177e4 LT |
535 | #define free_mm(mm) (kmem_cache_free(mm_cachep, (mm))) |
536 | ||
4cb0e11b HK |
537 | static unsigned long default_dump_filter = MMF_DUMP_FILTER_DEFAULT; |
538 | ||
539 | static int __init coredump_filter_setup(char *s) | |
540 | { | |
541 | default_dump_filter = | |
542 | (simple_strtoul(s, NULL, 0) << MMF_DUMP_FILTER_SHIFT) & | |
543 | MMF_DUMP_FILTER_MASK; | |
544 | return 1; | |
545 | } | |
546 | ||
547 | __setup("coredump_filter=", coredump_filter_setup); | |
548 | ||
1da177e4 LT |
549 | #include <linux/init_task.h> |
550 | ||
858f0993 AD |
551 | static void mm_init_aio(struct mm_struct *mm) |
552 | { | |
553 | #ifdef CONFIG_AIO | |
554 | spin_lock_init(&mm->ioctx_lock); | |
db446a08 | 555 | mm->ioctx_table = NULL; |
858f0993 AD |
556 | #endif |
557 | } | |
558 | ||
33144e84 VD |
559 | static void mm_init_owner(struct mm_struct *mm, struct task_struct *p) |
560 | { | |
561 | #ifdef CONFIG_MEMCG | |
562 | mm->owner = p; | |
563 | #endif | |
564 | } | |
565 | ||
fb0a685c | 566 | static struct mm_struct *mm_init(struct mm_struct *mm, struct task_struct *p) |
1da177e4 | 567 | { |
41f727fd VD |
568 | mm->mmap = NULL; |
569 | mm->mm_rb = RB_ROOT; | |
570 | mm->vmacache_seqnum = 0; | |
1da177e4 LT |
571 | atomic_set(&mm->mm_users, 1); |
572 | atomic_set(&mm->mm_count, 1); | |
573 | init_rwsem(&mm->mmap_sem); | |
574 | INIT_LIST_HEAD(&mm->mmlist); | |
999d9fc1 | 575 | mm->core_state = NULL; |
e1f56c89 | 576 | atomic_long_set(&mm->nr_ptes, 0); |
2d2f5119 | 577 | mm_nr_pmds_init(mm); |
41f727fd VD |
578 | mm->map_count = 0; |
579 | mm->locked_vm = 0; | |
ce65cefa | 580 | mm->pinned_vm = 0; |
d559db08 | 581 | memset(&mm->rss_stat, 0, sizeof(mm->rss_stat)); |
1da177e4 | 582 | spin_lock_init(&mm->page_table_lock); |
41f727fd | 583 | mm_init_cpumask(mm); |
858f0993 | 584 | mm_init_aio(mm); |
cf475ad2 | 585 | mm_init_owner(mm, p); |
41f727fd | 586 | mmu_notifier_mm_init(mm); |
20841405 | 587 | clear_tlb_flush_pending(mm); |
41f727fd VD |
588 | #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS |
589 | mm->pmd_huge_pte = NULL; | |
590 | #endif | |
1da177e4 | 591 | |
a0715cc2 AT |
592 | if (current->mm) { |
593 | mm->flags = current->mm->flags & MMF_INIT_MASK; | |
594 | mm->def_flags = current->mm->def_flags & VM_INIT_DEF_MASK; | |
595 | } else { | |
596 | mm->flags = default_dump_filter; | |
1da177e4 | 597 | mm->def_flags = 0; |
a0715cc2 AT |
598 | } |
599 | ||
41f727fd VD |
600 | if (mm_alloc_pgd(mm)) |
601 | goto fail_nopgd; | |
602 | ||
603 | if (init_new_context(p, mm)) | |
604 | goto fail_nocontext; | |
78fb7466 | 605 | |
41f727fd VD |
606 | return mm; |
607 | ||
608 | fail_nocontext: | |
609 | mm_free_pgd(mm); | |
610 | fail_nopgd: | |
1da177e4 LT |
611 | free_mm(mm); |
612 | return NULL; | |
613 | } | |
614 | ||
c3f0327f KK |
615 | static void check_mm(struct mm_struct *mm) |
616 | { | |
617 | int i; | |
618 | ||
619 | for (i = 0; i < NR_MM_COUNTERS; i++) { | |
620 | long x = atomic_long_read(&mm->rss_stat.count[i]); | |
621 | ||
622 | if (unlikely(x)) | |
623 | printk(KERN_ALERT "BUG: Bad rss-counter state " | |
624 | "mm:%p idx:%d val:%ld\n", mm, i, x); | |
625 | } | |
b30fe6c7 KS |
626 | |
627 | if (atomic_long_read(&mm->nr_ptes)) | |
628 | pr_alert("BUG: non-zero nr_ptes on freeing mm: %ld\n", | |
629 | atomic_long_read(&mm->nr_ptes)); | |
630 | if (mm_nr_pmds(mm)) | |
631 | pr_alert("BUG: non-zero nr_pmds on freeing mm: %ld\n", | |
632 | mm_nr_pmds(mm)); | |
633 | ||
e009bb30 | 634 | #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS |
96dad67f | 635 | VM_BUG_ON_MM(mm->pmd_huge_pte, mm); |
c3f0327f KK |
636 | #endif |
637 | } | |
638 | ||
1da177e4 LT |
639 | /* |
640 | * Allocate and initialize an mm_struct. | |
641 | */ | |
fb0a685c | 642 | struct mm_struct *mm_alloc(void) |
1da177e4 | 643 | { |
fb0a685c | 644 | struct mm_struct *mm; |
1da177e4 LT |
645 | |
646 | mm = allocate_mm(); | |
de03c72c KM |
647 | if (!mm) |
648 | return NULL; | |
649 | ||
650 | memset(mm, 0, sizeof(*mm)); | |
6345d24d | 651 | return mm_init(mm, current); |
1da177e4 LT |
652 | } |
653 | ||
654 | /* | |
655 | * Called when the last reference to the mm | |
656 | * is dropped: either by a lazy thread or by | |
657 | * mmput. Free the page directory and the mm. | |
658 | */ | |
7ad5b3a5 | 659 | void __mmdrop(struct mm_struct *mm) |
1da177e4 LT |
660 | { |
661 | BUG_ON(mm == &init_mm); | |
662 | mm_free_pgd(mm); | |
663 | destroy_context(mm); | |
cddb8a5c | 664 | mmu_notifier_mm_destroy(mm); |
c3f0327f | 665 | check_mm(mm); |
1da177e4 LT |
666 | free_mm(mm); |
667 | } | |
6d4e4c4f | 668 | EXPORT_SYMBOL_GPL(__mmdrop); |
1da177e4 LT |
669 | |
670 | /* | |
671 | * Decrement the use count and release all resources for an mm. | |
672 | */ | |
673 | void mmput(struct mm_struct *mm) | |
674 | { | |
0ae26f1b AM |
675 | might_sleep(); |
676 | ||
1da177e4 | 677 | if (atomic_dec_and_test(&mm->mm_users)) { |
d4b3b638 | 678 | uprobe_clear_state(mm); |
1da177e4 | 679 | exit_aio(mm); |
1c2fb7a4 | 680 | ksm_exit(mm); |
ba76149f | 681 | khugepaged_exit(mm); /* must run before exit_mmap */ |
1da177e4 | 682 | exit_mmap(mm); |
925d1c40 | 683 | set_mm_exe_file(mm, NULL); |
1da177e4 LT |
684 | if (!list_empty(&mm->mmlist)) { |
685 | spin_lock(&mmlist_lock); | |
686 | list_del(&mm->mmlist); | |
687 | spin_unlock(&mmlist_lock); | |
688 | } | |
801460d0 HS |
689 | if (mm->binfmt) |
690 | module_put(mm->binfmt->module); | |
1da177e4 LT |
691 | mmdrop(mm); |
692 | } | |
693 | } | |
694 | EXPORT_SYMBOL_GPL(mmput); | |
695 | ||
38646013 JS |
696 | void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file) |
697 | { | |
698 | if (new_exe_file) | |
699 | get_file(new_exe_file); | |
700 | if (mm->exe_file) | |
701 | fput(mm->exe_file); | |
702 | mm->exe_file = new_exe_file; | |
38646013 JS |
703 | } |
704 | ||
705 | struct file *get_mm_exe_file(struct mm_struct *mm) | |
706 | { | |
707 | struct file *exe_file; | |
708 | ||
2dd8ad81 | 709 | /* We need mmap_sem to protect against races with removal of exe_file */ |
38646013 JS |
710 | down_read(&mm->mmap_sem); |
711 | exe_file = mm->exe_file; | |
712 | if (exe_file) | |
713 | get_file(exe_file); | |
714 | up_read(&mm->mmap_sem); | |
715 | return exe_file; | |
716 | } | |
717 | ||
718 | static void dup_mm_exe_file(struct mm_struct *oldmm, struct mm_struct *newmm) | |
719 | { | |
720 | /* It's safe to write the exe_file pointer without exe_file_lock because | |
721 | * this is called during fork when the task is not yet in /proc */ | |
722 | newmm->exe_file = get_mm_exe_file(oldmm); | |
723 | } | |
724 | ||
1da177e4 LT |
725 | /** |
726 | * get_task_mm - acquire a reference to the task's mm | |
727 | * | |
246bb0b1 | 728 | * Returns %NULL if the task has no mm. Checks PF_KTHREAD (meaning |
1da177e4 LT |
729 | * this kernel workthread has transiently adopted a user mm with use_mm, |
730 | * to do its AIO) is not set and if so returns a reference to it, after | |
731 | * bumping up the use count. User must release the mm via mmput() | |
732 | * after use. Typically used by /proc and ptrace. | |
733 | */ | |
734 | struct mm_struct *get_task_mm(struct task_struct *task) | |
735 | { | |
736 | struct mm_struct *mm; | |
737 | ||
738 | task_lock(task); | |
739 | mm = task->mm; | |
740 | if (mm) { | |
246bb0b1 | 741 | if (task->flags & PF_KTHREAD) |
1da177e4 LT |
742 | mm = NULL; |
743 | else | |
744 | atomic_inc(&mm->mm_users); | |
745 | } | |
746 | task_unlock(task); | |
747 | return mm; | |
748 | } | |
749 | EXPORT_SYMBOL_GPL(get_task_mm); | |
750 | ||
8cdb878d CY |
751 | struct mm_struct *mm_access(struct task_struct *task, unsigned int mode) |
752 | { | |
753 | struct mm_struct *mm; | |
754 | int err; | |
755 | ||
756 | err = mutex_lock_killable(&task->signal->cred_guard_mutex); | |
757 | if (err) | |
758 | return ERR_PTR(err); | |
759 | ||
760 | mm = get_task_mm(task); | |
761 | if (mm && mm != current->mm && | |
762 | !ptrace_may_access(task, mode)) { | |
763 | mmput(mm); | |
764 | mm = ERR_PTR(-EACCES); | |
765 | } | |
766 | mutex_unlock(&task->signal->cred_guard_mutex); | |
767 | ||
768 | return mm; | |
769 | } | |
770 | ||
57b59c4a | 771 | static void complete_vfork_done(struct task_struct *tsk) |
c415c3b4 | 772 | { |
d68b46fe | 773 | struct completion *vfork; |
c415c3b4 | 774 | |
d68b46fe ON |
775 | task_lock(tsk); |
776 | vfork = tsk->vfork_done; | |
777 | if (likely(vfork)) { | |
778 | tsk->vfork_done = NULL; | |
779 | complete(vfork); | |
780 | } | |
781 | task_unlock(tsk); | |
782 | } | |
783 | ||
784 | static int wait_for_vfork_done(struct task_struct *child, | |
785 | struct completion *vfork) | |
786 | { | |
787 | int killed; | |
788 | ||
789 | freezer_do_not_count(); | |
790 | killed = wait_for_completion_killable(vfork); | |
791 | freezer_count(); | |
792 | ||
793 | if (killed) { | |
794 | task_lock(child); | |
795 | child->vfork_done = NULL; | |
796 | task_unlock(child); | |
797 | } | |
798 | ||
799 | put_task_struct(child); | |
800 | return killed; | |
c415c3b4 ON |
801 | } |
802 | ||
1da177e4 LT |
803 | /* Please note the differences between mmput and mm_release. |
804 | * mmput is called whenever we stop holding onto a mm_struct, | |
805 | * error success whatever. | |
806 | * | |
807 | * mm_release is called after a mm_struct has been removed | |
808 | * from the current process. | |
809 | * | |
810 | * This difference is important for error handling, when we | |
811 | * only half set up a mm_struct for a new process and need to restore | |
812 | * the old one. Because we mmput the new mm_struct before | |
813 | * restoring the old one. . . | |
814 | * Eric Biederman 10 January 1998 | |
815 | */ | |
816 | void mm_release(struct task_struct *tsk, struct mm_struct *mm) | |
817 | { | |
8141c7f3 LT |
818 | /* Get rid of any futexes when releasing the mm */ |
819 | #ifdef CONFIG_FUTEX | |
fc6b177d | 820 | if (unlikely(tsk->robust_list)) { |
8141c7f3 | 821 | exit_robust_list(tsk); |
fc6b177d PZ |
822 | tsk->robust_list = NULL; |
823 | } | |
8141c7f3 | 824 | #ifdef CONFIG_COMPAT |
fc6b177d | 825 | if (unlikely(tsk->compat_robust_list)) { |
8141c7f3 | 826 | compat_exit_robust_list(tsk); |
fc6b177d PZ |
827 | tsk->compat_robust_list = NULL; |
828 | } | |
8141c7f3 | 829 | #endif |
322a2c10 TG |
830 | if (unlikely(!list_empty(&tsk->pi_state_list))) |
831 | exit_pi_state_list(tsk); | |
8141c7f3 LT |
832 | #endif |
833 | ||
0326f5a9 SD |
834 | uprobe_free_utask(tsk); |
835 | ||
1da177e4 LT |
836 | /* Get rid of any cached register state */ |
837 | deactivate_mm(tsk, mm); | |
838 | ||
fec1d011 RM |
839 | /* |
840 | * If we're exiting normally, clear a user-space tid field if | |
841 | * requested. We leave this alone when dying by signal, to leave | |
842 | * the value intact in a core dump, and to save the unnecessary | |
d68b46fe ON |
843 | * trouble, say, a killed vfork parent shouldn't touch this mm. |
844 | * Userland only wants this done for a sys_exit. | |
fec1d011 | 845 | */ |
9c8a8228 ED |
846 | if (tsk->clear_child_tid) { |
847 | if (!(tsk->flags & PF_SIGNALED) && | |
848 | atomic_read(&mm->mm_users) > 1) { | |
849 | /* | |
850 | * We don't check the error code - if userspace has | |
851 | * not set up a proper pointer then tough luck. | |
852 | */ | |
853 | put_user(0, tsk->clear_child_tid); | |
854 | sys_futex(tsk->clear_child_tid, FUTEX_WAKE, | |
855 | 1, NULL, NULL, 0); | |
856 | } | |
1da177e4 | 857 | tsk->clear_child_tid = NULL; |
1da177e4 | 858 | } |
f7505d64 KK |
859 | |
860 | /* | |
861 | * All done, finally we can wake up parent and return this mm to him. | |
862 | * Also kthread_stop() uses this completion for synchronization. | |
863 | */ | |
864 | if (tsk->vfork_done) | |
865 | complete_vfork_done(tsk); | |
1da177e4 LT |
866 | } |
867 | ||
a0a7ec30 JD |
868 | /* |
869 | * Allocate a new mm structure and copy contents from the | |
870 | * mm structure of the passed in task structure. | |
871 | */ | |
ff252c1f | 872 | static struct mm_struct *dup_mm(struct task_struct *tsk) |
a0a7ec30 JD |
873 | { |
874 | struct mm_struct *mm, *oldmm = current->mm; | |
875 | int err; | |
876 | ||
a0a7ec30 JD |
877 | mm = allocate_mm(); |
878 | if (!mm) | |
879 | goto fail_nomem; | |
880 | ||
881 | memcpy(mm, oldmm, sizeof(*mm)); | |
882 | ||
78fb7466 | 883 | if (!mm_init(mm, tsk)) |
a0a7ec30 JD |
884 | goto fail_nomem; |
885 | ||
925d1c40 MH |
886 | dup_mm_exe_file(oldmm, mm); |
887 | ||
a0a7ec30 JD |
888 | err = dup_mmap(mm, oldmm); |
889 | if (err) | |
890 | goto free_pt; | |
891 | ||
892 | mm->hiwater_rss = get_mm_rss(mm); | |
893 | mm->hiwater_vm = mm->total_vm; | |
894 | ||
801460d0 HS |
895 | if (mm->binfmt && !try_module_get(mm->binfmt->module)) |
896 | goto free_pt; | |
897 | ||
a0a7ec30 JD |
898 | return mm; |
899 | ||
900 | free_pt: | |
801460d0 HS |
901 | /* don't put binfmt in mmput, we haven't got module yet */ |
902 | mm->binfmt = NULL; | |
a0a7ec30 JD |
903 | mmput(mm); |
904 | ||
905 | fail_nomem: | |
906 | return NULL; | |
a0a7ec30 JD |
907 | } |
908 | ||
fb0a685c | 909 | static int copy_mm(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 | 910 | { |
fb0a685c | 911 | struct mm_struct *mm, *oldmm; |
1da177e4 LT |
912 | int retval; |
913 | ||
914 | tsk->min_flt = tsk->maj_flt = 0; | |
915 | tsk->nvcsw = tsk->nivcsw = 0; | |
17406b82 MSB |
916 | #ifdef CONFIG_DETECT_HUNG_TASK |
917 | tsk->last_switch_count = tsk->nvcsw + tsk->nivcsw; | |
918 | #endif | |
1da177e4 LT |
919 | |
920 | tsk->mm = NULL; | |
921 | tsk->active_mm = NULL; | |
922 | ||
923 | /* | |
924 | * Are we cloning a kernel thread? | |
925 | * | |
926 | * We need to steal a active VM for that.. | |
927 | */ | |
928 | oldmm = current->mm; | |
929 | if (!oldmm) | |
930 | return 0; | |
931 | ||
615d6e87 DB |
932 | /* initialize the new vmacache entries */ |
933 | vmacache_flush(tsk); | |
934 | ||
1da177e4 LT |
935 | if (clone_flags & CLONE_VM) { |
936 | atomic_inc(&oldmm->mm_users); | |
937 | mm = oldmm; | |
1da177e4 LT |
938 | goto good_mm; |
939 | } | |
940 | ||
941 | retval = -ENOMEM; | |
a0a7ec30 | 942 | mm = dup_mm(tsk); |
1da177e4 LT |
943 | if (!mm) |
944 | goto fail_nomem; | |
945 | ||
1da177e4 LT |
946 | good_mm: |
947 | tsk->mm = mm; | |
948 | tsk->active_mm = mm; | |
949 | return 0; | |
950 | ||
1da177e4 LT |
951 | fail_nomem: |
952 | return retval; | |
1da177e4 LT |
953 | } |
954 | ||
a39bc516 | 955 | static int copy_fs(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 | 956 | { |
498052bb | 957 | struct fs_struct *fs = current->fs; |
1da177e4 | 958 | if (clone_flags & CLONE_FS) { |
498052bb | 959 | /* tsk->fs is already what we want */ |
2a4419b5 | 960 | spin_lock(&fs->lock); |
498052bb | 961 | if (fs->in_exec) { |
2a4419b5 | 962 | spin_unlock(&fs->lock); |
498052bb AV |
963 | return -EAGAIN; |
964 | } | |
965 | fs->users++; | |
2a4419b5 | 966 | spin_unlock(&fs->lock); |
1da177e4 LT |
967 | return 0; |
968 | } | |
498052bb | 969 | tsk->fs = copy_fs_struct(fs); |
1da177e4 LT |
970 | if (!tsk->fs) |
971 | return -ENOMEM; | |
972 | return 0; | |
973 | } | |
974 | ||
fb0a685c | 975 | static int copy_files(unsigned long clone_flags, struct task_struct *tsk) |
a016f338 JD |
976 | { |
977 | struct files_struct *oldf, *newf; | |
978 | int error = 0; | |
979 | ||
980 | /* | |
981 | * A background process may not have any files ... | |
982 | */ | |
983 | oldf = current->files; | |
984 | if (!oldf) | |
985 | goto out; | |
986 | ||
987 | if (clone_flags & CLONE_FILES) { | |
988 | atomic_inc(&oldf->count); | |
989 | goto out; | |
990 | } | |
991 | ||
a016f338 JD |
992 | newf = dup_fd(oldf, &error); |
993 | if (!newf) | |
994 | goto out; | |
995 | ||
996 | tsk->files = newf; | |
997 | error = 0; | |
998 | out: | |
999 | return error; | |
1000 | } | |
1001 | ||
fadad878 | 1002 | static int copy_io(unsigned long clone_flags, struct task_struct *tsk) |
fd0928df JA |
1003 | { |
1004 | #ifdef CONFIG_BLOCK | |
1005 | struct io_context *ioc = current->io_context; | |
6e736be7 | 1006 | struct io_context *new_ioc; |
fd0928df JA |
1007 | |
1008 | if (!ioc) | |
1009 | return 0; | |
fadad878 JA |
1010 | /* |
1011 | * Share io context with parent, if CLONE_IO is set | |
1012 | */ | |
1013 | if (clone_flags & CLONE_IO) { | |
3d48749d TH |
1014 | ioc_task_link(ioc); |
1015 | tsk->io_context = ioc; | |
fadad878 | 1016 | } else if (ioprio_valid(ioc->ioprio)) { |
6e736be7 TH |
1017 | new_ioc = get_task_io_context(tsk, GFP_KERNEL, NUMA_NO_NODE); |
1018 | if (unlikely(!new_ioc)) | |
fd0928df JA |
1019 | return -ENOMEM; |
1020 | ||
6e736be7 | 1021 | new_ioc->ioprio = ioc->ioprio; |
11a3122f | 1022 | put_io_context(new_ioc); |
fd0928df JA |
1023 | } |
1024 | #endif | |
1025 | return 0; | |
1026 | } | |
1027 | ||
a39bc516 | 1028 | static int copy_sighand(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 LT |
1029 | { |
1030 | struct sighand_struct *sig; | |
1031 | ||
60348802 | 1032 | if (clone_flags & CLONE_SIGHAND) { |
1da177e4 LT |
1033 | atomic_inc(¤t->sighand->count); |
1034 | return 0; | |
1035 | } | |
1036 | sig = kmem_cache_alloc(sighand_cachep, GFP_KERNEL); | |
e56d0903 | 1037 | rcu_assign_pointer(tsk->sighand, sig); |
1da177e4 LT |
1038 | if (!sig) |
1039 | return -ENOMEM; | |
1da177e4 LT |
1040 | atomic_set(&sig->count, 1); |
1041 | memcpy(sig->action, current->sighand->action, sizeof(sig->action)); | |
1042 | return 0; | |
1043 | } | |
1044 | ||
a7e5328a | 1045 | void __cleanup_sighand(struct sighand_struct *sighand) |
c81addc9 | 1046 | { |
d80e731e ON |
1047 | if (atomic_dec_and_test(&sighand->count)) { |
1048 | signalfd_cleanup(sighand); | |
392809b2 ON |
1049 | /* |
1050 | * sighand_cachep is SLAB_DESTROY_BY_RCU so we can free it | |
1051 | * without an RCU grace period, see __lock_task_sighand(). | |
1052 | */ | |
c81addc9 | 1053 | kmem_cache_free(sighand_cachep, sighand); |
d80e731e | 1054 | } |
c81addc9 ON |
1055 | } |
1056 | ||
f06febc9 FM |
1057 | /* |
1058 | * Initialize POSIX timer handling for a thread group. | |
1059 | */ | |
1060 | static void posix_cpu_timers_init_group(struct signal_struct *sig) | |
1061 | { | |
78d7d407 JS |
1062 | unsigned long cpu_limit; |
1063 | ||
f06febc9 FM |
1064 | /* Thread group counters. */ |
1065 | thread_group_cputime_init(sig); | |
1066 | ||
78d7d407 JS |
1067 | cpu_limit = ACCESS_ONCE(sig->rlim[RLIMIT_CPU].rlim_cur); |
1068 | if (cpu_limit != RLIM_INFINITY) { | |
1069 | sig->cputime_expires.prof_exp = secs_to_cputime(cpu_limit); | |
6279a751 ON |
1070 | sig->cputimer.running = 1; |
1071 | } | |
1072 | ||
f06febc9 FM |
1073 | /* The timer lists. */ |
1074 | INIT_LIST_HEAD(&sig->cpu_timers[0]); | |
1075 | INIT_LIST_HEAD(&sig->cpu_timers[1]); | |
1076 | INIT_LIST_HEAD(&sig->cpu_timers[2]); | |
1077 | } | |
1078 | ||
a39bc516 | 1079 | static int copy_signal(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 LT |
1080 | { |
1081 | struct signal_struct *sig; | |
1da177e4 | 1082 | |
4ab6c083 | 1083 | if (clone_flags & CLONE_THREAD) |
490dea45 | 1084 | return 0; |
490dea45 | 1085 | |
a56704ef | 1086 | sig = kmem_cache_zalloc(signal_cachep, GFP_KERNEL); |
1da177e4 LT |
1087 | tsk->signal = sig; |
1088 | if (!sig) | |
1089 | return -ENOMEM; | |
1090 | ||
b3ac022c | 1091 | sig->nr_threads = 1; |
1da177e4 | 1092 | atomic_set(&sig->live, 1); |
b3ac022c | 1093 | atomic_set(&sig->sigcnt, 1); |
0c740d0a ON |
1094 | |
1095 | /* list_add(thread_node, thread_head) without INIT_LIST_HEAD() */ | |
1096 | sig->thread_head = (struct list_head)LIST_HEAD_INIT(tsk->thread_node); | |
1097 | tsk->thread_node = (struct list_head)LIST_HEAD_INIT(sig->thread_head); | |
1098 | ||
1da177e4 | 1099 | init_waitqueue_head(&sig->wait_chldexit); |
db51aecc | 1100 | sig->curr_target = tsk; |
1da177e4 LT |
1101 | init_sigpending(&sig->shared_pending); |
1102 | INIT_LIST_HEAD(&sig->posix_timers); | |
e78c3496 | 1103 | seqlock_init(&sig->stats_lock); |
1da177e4 | 1104 | |
c9cb2e3d | 1105 | hrtimer_init(&sig->real_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL); |
1da177e4 | 1106 | sig->real_timer.function = it_real_fn; |
1da177e4 | 1107 | |
1da177e4 LT |
1108 | task_lock(current->group_leader); |
1109 | memcpy(sig->rlim, current->signal->rlim, sizeof sig->rlim); | |
1110 | task_unlock(current->group_leader); | |
1111 | ||
6279a751 ON |
1112 | posix_cpu_timers_init_group(sig); |
1113 | ||
522ed776 | 1114 | tty_audit_fork(sig); |
5091faa4 | 1115 | sched_autogroup_fork(sig); |
522ed776 | 1116 | |
4714d1d3 | 1117 | #ifdef CONFIG_CGROUPS |
257058ae | 1118 | init_rwsem(&sig->group_rwsem); |
4714d1d3 BB |
1119 | #endif |
1120 | ||
a63d83f4 | 1121 | sig->oom_score_adj = current->signal->oom_score_adj; |
dabb16f6 | 1122 | sig->oom_score_adj_min = current->signal->oom_score_adj_min; |
28b83c51 | 1123 | |
ebec18a6 LP |
1124 | sig->has_child_subreaper = current->signal->has_child_subreaper || |
1125 | current->signal->is_child_subreaper; | |
1126 | ||
9b1bf12d KM |
1127 | mutex_init(&sig->cred_guard_mutex); |
1128 | ||
1da177e4 LT |
1129 | return 0; |
1130 | } | |
1131 | ||
dbd95212 KC |
1132 | static void copy_seccomp(struct task_struct *p) |
1133 | { | |
1134 | #ifdef CONFIG_SECCOMP | |
1135 | /* | |
1136 | * Must be called with sighand->lock held, which is common to | |
1137 | * all threads in the group. Holding cred_guard_mutex is not | |
1138 | * needed because this new task is not yet running and cannot | |
1139 | * be racing exec. | |
1140 | */ | |
69f6a34b | 1141 | assert_spin_locked(¤t->sighand->siglock); |
dbd95212 KC |
1142 | |
1143 | /* Ref-count the new filter user, and assign it. */ | |
1144 | get_seccomp_filter(current); | |
1145 | p->seccomp = current->seccomp; | |
1146 | ||
1147 | /* | |
1148 | * Explicitly enable no_new_privs here in case it got set | |
1149 | * between the task_struct being duplicated and holding the | |
1150 | * sighand lock. The seccomp state and nnp must be in sync. | |
1151 | */ | |
1152 | if (task_no_new_privs(current)) | |
1153 | task_set_no_new_privs(p); | |
1154 | ||
1155 | /* | |
1156 | * If the parent gained a seccomp mode after copying thread | |
1157 | * flags and between before we held the sighand lock, we have | |
1158 | * to manually enable the seccomp thread flag here. | |
1159 | */ | |
1160 | if (p->seccomp.mode != SECCOMP_MODE_DISABLED) | |
1161 | set_tsk_thread_flag(p, TIF_SECCOMP); | |
1162 | #endif | |
1163 | } | |
1164 | ||
17da2bd9 | 1165 | SYSCALL_DEFINE1(set_tid_address, int __user *, tidptr) |
1da177e4 LT |
1166 | { |
1167 | current->clear_child_tid = tidptr; | |
1168 | ||
b488893a | 1169 | return task_pid_vnr(current); |
1da177e4 LT |
1170 | } |
1171 | ||
a39bc516 | 1172 | static void rt_mutex_init_task(struct task_struct *p) |
23f78d4a | 1173 | { |
1d615482 | 1174 | raw_spin_lock_init(&p->pi_lock); |
e29e175b | 1175 | #ifdef CONFIG_RT_MUTEXES |
fb00aca4 PZ |
1176 | p->pi_waiters = RB_ROOT; |
1177 | p->pi_waiters_leftmost = NULL; | |
23f78d4a | 1178 | p->pi_blocked_on = NULL; |
23f78d4a IM |
1179 | #endif |
1180 | } | |
1181 | ||
f06febc9 FM |
1182 | /* |
1183 | * Initialize POSIX timer handling for a single task. | |
1184 | */ | |
1185 | static void posix_cpu_timers_init(struct task_struct *tsk) | |
1186 | { | |
64861634 MS |
1187 | tsk->cputime_expires.prof_exp = 0; |
1188 | tsk->cputime_expires.virt_exp = 0; | |
f06febc9 FM |
1189 | tsk->cputime_expires.sched_exp = 0; |
1190 | INIT_LIST_HEAD(&tsk->cpu_timers[0]); | |
1191 | INIT_LIST_HEAD(&tsk->cpu_timers[1]); | |
1192 | INIT_LIST_HEAD(&tsk->cpu_timers[2]); | |
1193 | } | |
1194 | ||
81907739 ON |
1195 | static inline void |
1196 | init_task_pid(struct task_struct *task, enum pid_type type, struct pid *pid) | |
1197 | { | |
1198 | task->pids[type].pid = pid; | |
1199 | } | |
1200 | ||
1da177e4 LT |
1201 | /* |
1202 | * This creates a new process as a copy of the old one, | |
1203 | * but does not actually start it yet. | |
1204 | * | |
1205 | * It copies the registers, and all the appropriate | |
1206 | * parts of the process environment (as per the clone | |
1207 | * flags). The actual kick-off is left to the caller. | |
1208 | */ | |
36c8b586 IM |
1209 | static struct task_struct *copy_process(unsigned long clone_flags, |
1210 | unsigned long stack_start, | |
36c8b586 | 1211 | unsigned long stack_size, |
36c8b586 | 1212 | int __user *child_tidptr, |
09a05394 RM |
1213 | struct pid *pid, |
1214 | int trace) | |
1da177e4 LT |
1215 | { |
1216 | int retval; | |
a24efe62 | 1217 | struct task_struct *p; |
1da177e4 LT |
1218 | |
1219 | if ((clone_flags & (CLONE_NEWNS|CLONE_FS)) == (CLONE_NEWNS|CLONE_FS)) | |
1220 | return ERR_PTR(-EINVAL); | |
1221 | ||
e66eded8 EB |
1222 | if ((clone_flags & (CLONE_NEWUSER|CLONE_FS)) == (CLONE_NEWUSER|CLONE_FS)) |
1223 | return ERR_PTR(-EINVAL); | |
1224 | ||
1da177e4 LT |
1225 | /* |
1226 | * Thread groups must share signals as well, and detached threads | |
1227 | * can only be started up within the thread group. | |
1228 | */ | |
1229 | if ((clone_flags & CLONE_THREAD) && !(clone_flags & CLONE_SIGHAND)) | |
1230 | return ERR_PTR(-EINVAL); | |
1231 | ||
1232 | /* | |
1233 | * Shared signal handlers imply shared VM. By way of the above, | |
1234 | * thread groups also imply shared VM. Blocking this case allows | |
1235 | * for various simplifications in other code. | |
1236 | */ | |
1237 | if ((clone_flags & CLONE_SIGHAND) && !(clone_flags & CLONE_VM)) | |
1238 | return ERR_PTR(-EINVAL); | |
1239 | ||
123be07b SB |
1240 | /* |
1241 | * Siblings of global init remain as zombies on exit since they are | |
1242 | * not reaped by their parent (swapper). To solve this and to avoid | |
1243 | * multi-rooted process trees, prevent global and container-inits | |
1244 | * from creating siblings. | |
1245 | */ | |
1246 | if ((clone_flags & CLONE_PARENT) && | |
1247 | current->signal->flags & SIGNAL_UNKILLABLE) | |
1248 | return ERR_PTR(-EINVAL); | |
1249 | ||
8382fcac | 1250 | /* |
40a0d32d ON |
1251 | * If the new process will be in a different pid or user namespace |
1252 | * do not allow it to share a thread group or signal handlers or | |
1253 | * parent with the forking task. | |
8382fcac | 1254 | */ |
1f7f4dde | 1255 | if (clone_flags & CLONE_SIGHAND) { |
40a0d32d ON |
1256 | if ((clone_flags & (CLONE_NEWUSER | CLONE_NEWPID)) || |
1257 | (task_active_pid_ns(current) != | |
1258 | current->nsproxy->pid_ns_for_children)) | |
1259 | return ERR_PTR(-EINVAL); | |
1260 | } | |
8382fcac | 1261 | |
1da177e4 LT |
1262 | retval = security_task_create(clone_flags); |
1263 | if (retval) | |
1264 | goto fork_out; | |
1265 | ||
1266 | retval = -ENOMEM; | |
1267 | p = dup_task_struct(current); | |
1268 | if (!p) | |
1269 | goto fork_out; | |
1270 | ||
f7e8b616 SR |
1271 | ftrace_graph_init_task(p); |
1272 | ||
bea493a0 PZ |
1273 | rt_mutex_init_task(p); |
1274 | ||
d12c1a37 | 1275 | #ifdef CONFIG_PROVE_LOCKING |
de30a2b3 IM |
1276 | DEBUG_LOCKS_WARN_ON(!p->hardirqs_enabled); |
1277 | DEBUG_LOCKS_WARN_ON(!p->softirqs_enabled); | |
1278 | #endif | |
1da177e4 | 1279 | retval = -EAGAIN; |
3b11a1de | 1280 | if (atomic_read(&p->real_cred->user->processes) >= |
78d7d407 | 1281 | task_rlimit(p, RLIMIT_NPROC)) { |
b57922b6 EP |
1282 | if (p->real_cred->user != INIT_USER && |
1283 | !capable(CAP_SYS_RESOURCE) && !capable(CAP_SYS_ADMIN)) | |
1da177e4 LT |
1284 | goto bad_fork_free; |
1285 | } | |
72fa5997 | 1286 | current->flags &= ~PF_NPROC_EXCEEDED; |
1da177e4 | 1287 | |
f1752eec DH |
1288 | retval = copy_creds(p, clone_flags); |
1289 | if (retval < 0) | |
1290 | goto bad_fork_free; | |
1da177e4 LT |
1291 | |
1292 | /* | |
1293 | * If multiple threads are within copy_process(), then this check | |
1294 | * triggers too late. This doesn't hurt, the check is only there | |
1295 | * to stop root fork bombs. | |
1296 | */ | |
04ec93fe | 1297 | retval = -EAGAIN; |
1da177e4 LT |
1298 | if (nr_threads >= max_threads) |
1299 | goto bad_fork_cleanup_count; | |
1300 | ||
ca74e92b | 1301 | delayacct_tsk_init(p); /* Must remain after dup_task_struct() */ |
514ddb44 DR |
1302 | p->flags &= ~(PF_SUPERPRIV | PF_WQ_WORKER); |
1303 | p->flags |= PF_FORKNOEXEC; | |
1da177e4 LT |
1304 | INIT_LIST_HEAD(&p->children); |
1305 | INIT_LIST_HEAD(&p->sibling); | |
f41d911f | 1306 | rcu_copy_process(p); |
1da177e4 LT |
1307 | p->vfork_done = NULL; |
1308 | spin_lock_init(&p->alloc_lock); | |
1da177e4 | 1309 | |
1da177e4 LT |
1310 | init_sigpending(&p->pending); |
1311 | ||
64861634 MS |
1312 | p->utime = p->stime = p->gtime = 0; |
1313 | p->utimescaled = p->stimescaled = 0; | |
9fbc42ea | 1314 | #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE |
d37f761d | 1315 | p->prev_cputime.utime = p->prev_cputime.stime = 0; |
d99ca3b9 | 1316 | #endif |
6a61671b FW |
1317 | #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN |
1318 | seqlock_init(&p->vtime_seqlock); | |
1319 | p->vtime_snap = 0; | |
1320 | p->vtime_snap_whence = VTIME_SLEEPING; | |
1321 | #endif | |
1322 | ||
a3a2e76c KH |
1323 | #if defined(SPLIT_RSS_COUNTING) |
1324 | memset(&p->rss_stat, 0, sizeof(p->rss_stat)); | |
1325 | #endif | |
172ba844 | 1326 | |
6976675d AV |
1327 | p->default_timer_slack_ns = current->timer_slack_ns; |
1328 | ||
5995477a | 1329 | task_io_accounting_init(&p->ioac); |
1da177e4 LT |
1330 | acct_clear_integrals(p); |
1331 | ||
f06febc9 | 1332 | posix_cpu_timers_init(p); |
1da177e4 | 1333 | |
ccbf62d8 | 1334 | p->start_time = ktime_get_ns(); |
57e0be04 | 1335 | p->real_start_time = ktime_get_boot_ns(); |
1da177e4 | 1336 | p->io_context = NULL; |
1da177e4 | 1337 | p->audit_context = NULL; |
4714d1d3 | 1338 | if (clone_flags & CLONE_THREAD) |
257058ae | 1339 | threadgroup_change_begin(current); |
b4f48b63 | 1340 | cgroup_fork(p); |
1da177e4 | 1341 | #ifdef CONFIG_NUMA |
846a16bf | 1342 | p->mempolicy = mpol_dup(p->mempolicy); |
fb0a685c DRO |
1343 | if (IS_ERR(p->mempolicy)) { |
1344 | retval = PTR_ERR(p->mempolicy); | |
1345 | p->mempolicy = NULL; | |
e8604cb4 | 1346 | goto bad_fork_cleanup_threadgroup_lock; |
fb0a685c | 1347 | } |
1da177e4 | 1348 | #endif |
778d3b0f MH |
1349 | #ifdef CONFIG_CPUSETS |
1350 | p->cpuset_mem_spread_rotor = NUMA_NO_NODE; | |
1351 | p->cpuset_slab_spread_rotor = NUMA_NO_NODE; | |
cc9a6c87 | 1352 | seqcount_init(&p->mems_allowed_seq); |
778d3b0f | 1353 | #endif |
de30a2b3 IM |
1354 | #ifdef CONFIG_TRACE_IRQFLAGS |
1355 | p->irq_events = 0; | |
1356 | p->hardirqs_enabled = 0; | |
1357 | p->hardirq_enable_ip = 0; | |
1358 | p->hardirq_enable_event = 0; | |
1359 | p->hardirq_disable_ip = _THIS_IP_; | |
1360 | p->hardirq_disable_event = 0; | |
1361 | p->softirqs_enabled = 1; | |
1362 | p->softirq_enable_ip = _THIS_IP_; | |
1363 | p->softirq_enable_event = 0; | |
1364 | p->softirq_disable_ip = 0; | |
1365 | p->softirq_disable_event = 0; | |
1366 | p->hardirq_context = 0; | |
1367 | p->softirq_context = 0; | |
1368 | #endif | |
fbb9ce95 IM |
1369 | #ifdef CONFIG_LOCKDEP |
1370 | p->lockdep_depth = 0; /* no locks held yet */ | |
1371 | p->curr_chain_key = 0; | |
1372 | p->lockdep_recursion = 0; | |
1373 | #endif | |
1da177e4 | 1374 | |
408894ee IM |
1375 | #ifdef CONFIG_DEBUG_MUTEXES |
1376 | p->blocked_on = NULL; /* not blocked yet */ | |
1377 | #endif | |
cafe5635 KO |
1378 | #ifdef CONFIG_BCACHE |
1379 | p->sequential_io = 0; | |
1380 | p->sequential_io_avg = 0; | |
1381 | #endif | |
0f481406 | 1382 | |
3c90e6e9 | 1383 | /* Perform scheduler related setup. Assign this task to a CPU. */ |
aab03e05 DF |
1384 | retval = sched_fork(clone_flags, p); |
1385 | if (retval) | |
1386 | goto bad_fork_cleanup_policy; | |
6ab423e0 | 1387 | |
cdd6c482 | 1388 | retval = perf_event_init_task(p); |
6ab423e0 PZ |
1389 | if (retval) |
1390 | goto bad_fork_cleanup_policy; | |
fb0a685c DRO |
1391 | retval = audit_alloc(p); |
1392 | if (retval) | |
6c72e350 | 1393 | goto bad_fork_cleanup_perf; |
1da177e4 | 1394 | /* copy all the process information */ |
ab602f79 | 1395 | shm_init_task(p); |
fb0a685c DRO |
1396 | retval = copy_semundo(clone_flags, p); |
1397 | if (retval) | |
1da177e4 | 1398 | goto bad_fork_cleanup_audit; |
fb0a685c DRO |
1399 | retval = copy_files(clone_flags, p); |
1400 | if (retval) | |
1da177e4 | 1401 | goto bad_fork_cleanup_semundo; |
fb0a685c DRO |
1402 | retval = copy_fs(clone_flags, p); |
1403 | if (retval) | |
1da177e4 | 1404 | goto bad_fork_cleanup_files; |
fb0a685c DRO |
1405 | retval = copy_sighand(clone_flags, p); |
1406 | if (retval) | |
1da177e4 | 1407 | goto bad_fork_cleanup_fs; |
fb0a685c DRO |
1408 | retval = copy_signal(clone_flags, p); |
1409 | if (retval) | |
1da177e4 | 1410 | goto bad_fork_cleanup_sighand; |
fb0a685c DRO |
1411 | retval = copy_mm(clone_flags, p); |
1412 | if (retval) | |
1da177e4 | 1413 | goto bad_fork_cleanup_signal; |
fb0a685c DRO |
1414 | retval = copy_namespaces(clone_flags, p); |
1415 | if (retval) | |
d84f4f99 | 1416 | goto bad_fork_cleanup_mm; |
fb0a685c DRO |
1417 | retval = copy_io(clone_flags, p); |
1418 | if (retval) | |
fd0928df | 1419 | goto bad_fork_cleanup_namespaces; |
afa86fc4 | 1420 | retval = copy_thread(clone_flags, stack_start, stack_size, p); |
1da177e4 | 1421 | if (retval) |
fd0928df | 1422 | goto bad_fork_cleanup_io; |
1da177e4 | 1423 | |
425fb2b4 | 1424 | if (pid != &init_struct_pid) { |
c2b1df2e | 1425 | pid = alloc_pid(p->nsproxy->pid_ns_for_children); |
35f71bc0 MH |
1426 | if (IS_ERR(pid)) { |
1427 | retval = PTR_ERR(pid); | |
fd0928df | 1428 | goto bad_fork_cleanup_io; |
35f71bc0 | 1429 | } |
425fb2b4 PE |
1430 | } |
1431 | ||
1da177e4 LT |
1432 | p->set_child_tid = (clone_flags & CLONE_CHILD_SETTID) ? child_tidptr : NULL; |
1433 | /* | |
1434 | * Clear TID on mm_release()? | |
1435 | */ | |
fb0a685c | 1436 | p->clear_child_tid = (clone_flags & CLONE_CHILD_CLEARTID) ? child_tidptr : NULL; |
73c10101 JA |
1437 | #ifdef CONFIG_BLOCK |
1438 | p->plug = NULL; | |
1439 | #endif | |
42b2dd0a | 1440 | #ifdef CONFIG_FUTEX |
8f17d3a5 IM |
1441 | p->robust_list = NULL; |
1442 | #ifdef CONFIG_COMPAT | |
1443 | p->compat_robust_list = NULL; | |
1444 | #endif | |
c87e2837 IM |
1445 | INIT_LIST_HEAD(&p->pi_state_list); |
1446 | p->pi_state_cache = NULL; | |
42b2dd0a | 1447 | #endif |
f9a3879a GM |
1448 | /* |
1449 | * sigaltstack should be cleared when sharing the same VM | |
1450 | */ | |
1451 | if ((clone_flags & (CLONE_VM|CLONE_VFORK)) == CLONE_VM) | |
1452 | p->sas_ss_sp = p->sas_ss_size = 0; | |
1453 | ||
1da177e4 | 1454 | /* |
6580807d ON |
1455 | * Syscall tracing and stepping should be turned off in the |
1456 | * child regardless of CLONE_PTRACE. | |
1da177e4 | 1457 | */ |
6580807d | 1458 | user_disable_single_step(p); |
1da177e4 | 1459 | clear_tsk_thread_flag(p, TIF_SYSCALL_TRACE); |
ed75e8d5 LV |
1460 | #ifdef TIF_SYSCALL_EMU |
1461 | clear_tsk_thread_flag(p, TIF_SYSCALL_EMU); | |
1462 | #endif | |
9745512c | 1463 | clear_all_latency_tracing(p); |
1da177e4 | 1464 | |
1da177e4 | 1465 | /* ok, now we should be set up.. */ |
18c830df ON |
1466 | p->pid = pid_nr(pid); |
1467 | if (clone_flags & CLONE_THREAD) { | |
5f8aadd8 | 1468 | p->exit_signal = -1; |
18c830df ON |
1469 | p->group_leader = current->group_leader; |
1470 | p->tgid = current->tgid; | |
1471 | } else { | |
1472 | if (clone_flags & CLONE_PARENT) | |
1473 | p->exit_signal = current->group_leader->exit_signal; | |
1474 | else | |
1475 | p->exit_signal = (clone_flags & CSIGNAL); | |
1476 | p->group_leader = p; | |
1477 | p->tgid = p->pid; | |
1478 | } | |
5f8aadd8 | 1479 | |
9d823e8f WF |
1480 | p->nr_dirtied = 0; |
1481 | p->nr_dirtied_pause = 128 >> (PAGE_SHIFT - 10); | |
83712358 | 1482 | p->dirty_paused_when = 0; |
9d823e8f | 1483 | |
bb8cbbfe | 1484 | p->pdeath_signal = 0; |
47e65328 | 1485 | INIT_LIST_HEAD(&p->thread_group); |
158e1645 | 1486 | p->task_works = NULL; |
1da177e4 | 1487 | |
18c830df ON |
1488 | /* |
1489 | * Make it visible to the rest of the system, but dont wake it up yet. | |
1490 | * Need tasklist lock for parent etc handling! | |
1491 | */ | |
1da177e4 LT |
1492 | write_lock_irq(&tasklist_lock); |
1493 | ||
1da177e4 | 1494 | /* CLONE_PARENT re-uses the old parent */ |
2d5516cb | 1495 | if (clone_flags & (CLONE_PARENT|CLONE_THREAD)) { |
1da177e4 | 1496 | p->real_parent = current->real_parent; |
2d5516cb ON |
1497 | p->parent_exec_id = current->parent_exec_id; |
1498 | } else { | |
1da177e4 | 1499 | p->real_parent = current; |
2d5516cb ON |
1500 | p->parent_exec_id = current->self_exec_id; |
1501 | } | |
1da177e4 | 1502 | |
3f17da69 | 1503 | spin_lock(¤t->sighand->siglock); |
4a2c7a78 | 1504 | |
dbd95212 KC |
1505 | /* |
1506 | * Copy seccomp details explicitly here, in case they were changed | |
1507 | * before holding sighand lock. | |
1508 | */ | |
1509 | copy_seccomp(p); | |
1510 | ||
4a2c7a78 ON |
1511 | /* |
1512 | * Process group and session signals need to be delivered to just the | |
1513 | * parent before the fork or both the parent and the child after the | |
1514 | * fork. Restart if a signal comes in before we add the new process to | |
1515 | * it's process group. | |
1516 | * A fatal signal pending means that current will exit, so the new | |
1517 | * thread can't slip out of an OOM kill (or normal SIGKILL). | |
fb0a685c | 1518 | */ |
23ff4440 | 1519 | recalc_sigpending(); |
4a2c7a78 ON |
1520 | if (signal_pending(current)) { |
1521 | spin_unlock(¤t->sighand->siglock); | |
1522 | write_unlock_irq(&tasklist_lock); | |
1523 | retval = -ERESTARTNOINTR; | |
f7e8b616 | 1524 | goto bad_fork_free_pid; |
4a2c7a78 ON |
1525 | } |
1526 | ||
73b9ebfe | 1527 | if (likely(p->pid)) { |
4b9d33e6 | 1528 | ptrace_init_task(p, (clone_flags & CLONE_PTRACE) || trace); |
73b9ebfe | 1529 | |
81907739 | 1530 | init_task_pid(p, PIDTYPE_PID, pid); |
73b9ebfe | 1531 | if (thread_group_leader(p)) { |
81907739 ON |
1532 | init_task_pid(p, PIDTYPE_PGID, task_pgrp(current)); |
1533 | init_task_pid(p, PIDTYPE_SID, task_session(current)); | |
1534 | ||
1c4042c2 | 1535 | if (is_child_reaper(pid)) { |
17cf22c3 | 1536 | ns_of_pid(pid)->child_reaper = p; |
1c4042c2 EB |
1537 | p->signal->flags |= SIGNAL_UNKILLABLE; |
1538 | } | |
73b9ebfe | 1539 | |
fea9d175 | 1540 | p->signal->leader_pid = pid; |
9c9f4ded | 1541 | p->signal->tty = tty_kref_get(current->signal->tty); |
9cd80bbb | 1542 | list_add_tail(&p->sibling, &p->real_parent->children); |
5e85d4ab | 1543 | list_add_tail_rcu(&p->tasks, &init_task.tasks); |
81907739 ON |
1544 | attach_pid(p, PIDTYPE_PGID); |
1545 | attach_pid(p, PIDTYPE_SID); | |
909ea964 | 1546 | __this_cpu_inc(process_counts); |
80628ca0 ON |
1547 | } else { |
1548 | current->signal->nr_threads++; | |
1549 | atomic_inc(¤t->signal->live); | |
1550 | atomic_inc(¤t->signal->sigcnt); | |
80628ca0 ON |
1551 | list_add_tail_rcu(&p->thread_group, |
1552 | &p->group_leader->thread_group); | |
0c740d0a ON |
1553 | list_add_tail_rcu(&p->thread_node, |
1554 | &p->signal->thread_head); | |
73b9ebfe | 1555 | } |
81907739 | 1556 | attach_pid(p, PIDTYPE_PID); |
73b9ebfe | 1557 | nr_threads++; |
1da177e4 LT |
1558 | } |
1559 | ||
1da177e4 | 1560 | total_forks++; |
3f17da69 | 1561 | spin_unlock(¤t->sighand->siglock); |
4af4206b | 1562 | syscall_tracepoint_update(p); |
1da177e4 | 1563 | write_unlock_irq(&tasklist_lock); |
4af4206b | 1564 | |
c13cf856 | 1565 | proc_fork_connector(p); |
817929ec | 1566 | cgroup_post_fork(p); |
4714d1d3 | 1567 | if (clone_flags & CLONE_THREAD) |
257058ae | 1568 | threadgroup_change_end(current); |
cdd6c482 | 1569 | perf_event_fork(p); |
43d2b113 KH |
1570 | |
1571 | trace_task_newtask(p, clone_flags); | |
3ab67966 | 1572 | uprobe_copy_process(p, clone_flags); |
43d2b113 | 1573 | |
1da177e4 LT |
1574 | return p; |
1575 | ||
425fb2b4 PE |
1576 | bad_fork_free_pid: |
1577 | if (pid != &init_struct_pid) | |
1578 | free_pid(pid); | |
fd0928df | 1579 | bad_fork_cleanup_io: |
b69f2292 LR |
1580 | if (p->io_context) |
1581 | exit_io_context(p); | |
ab516013 | 1582 | bad_fork_cleanup_namespaces: |
444f378b | 1583 | exit_task_namespaces(p); |
1da177e4 | 1584 | bad_fork_cleanup_mm: |
c9f01245 | 1585 | if (p->mm) |
1da177e4 LT |
1586 | mmput(p->mm); |
1587 | bad_fork_cleanup_signal: | |
4ab6c083 | 1588 | if (!(clone_flags & CLONE_THREAD)) |
1c5354de | 1589 | free_signal_struct(p->signal); |
1da177e4 | 1590 | bad_fork_cleanup_sighand: |
a7e5328a | 1591 | __cleanup_sighand(p->sighand); |
1da177e4 LT |
1592 | bad_fork_cleanup_fs: |
1593 | exit_fs(p); /* blocking */ | |
1594 | bad_fork_cleanup_files: | |
1595 | exit_files(p); /* blocking */ | |
1596 | bad_fork_cleanup_semundo: | |
1597 | exit_sem(p); | |
1598 | bad_fork_cleanup_audit: | |
1599 | audit_free(p); | |
6c72e350 | 1600 | bad_fork_cleanup_perf: |
cdd6c482 | 1601 | perf_event_free_task(p); |
6c72e350 | 1602 | bad_fork_cleanup_policy: |
1da177e4 | 1603 | #ifdef CONFIG_NUMA |
f0be3d32 | 1604 | mpol_put(p->mempolicy); |
e8604cb4 | 1605 | bad_fork_cleanup_threadgroup_lock: |
1da177e4 | 1606 | #endif |
4714d1d3 | 1607 | if (clone_flags & CLONE_THREAD) |
257058ae | 1608 | threadgroup_change_end(current); |
35df17c5 | 1609 | delayacct_tsk_free(p); |
1da177e4 | 1610 | bad_fork_cleanup_count: |
d84f4f99 | 1611 | atomic_dec(&p->cred->user->processes); |
e0e81739 | 1612 | exit_creds(p); |
1da177e4 LT |
1613 | bad_fork_free: |
1614 | free_task(p); | |
fe7d37d1 ON |
1615 | fork_out: |
1616 | return ERR_PTR(retval); | |
1da177e4 LT |
1617 | } |
1618 | ||
f106eee1 ON |
1619 | static inline void init_idle_pids(struct pid_link *links) |
1620 | { | |
1621 | enum pid_type type; | |
1622 | ||
1623 | for (type = PIDTYPE_PID; type < PIDTYPE_MAX; ++type) { | |
1624 | INIT_HLIST_NODE(&links[type].node); /* not really needed */ | |
1625 | links[type].pid = &init_struct_pid; | |
1626 | } | |
1627 | } | |
1628 | ||
0db0628d | 1629 | struct task_struct *fork_idle(int cpu) |
1da177e4 | 1630 | { |
36c8b586 | 1631 | struct task_struct *task; |
62e791c1 | 1632 | task = copy_process(CLONE_VM, 0, 0, NULL, &init_struct_pid, 0); |
f106eee1 ON |
1633 | if (!IS_ERR(task)) { |
1634 | init_idle_pids(task->pids); | |
753ca4f3 | 1635 | init_idle(task, cpu); |
f106eee1 | 1636 | } |
73b9ebfe | 1637 | |
1da177e4 LT |
1638 | return task; |
1639 | } | |
1640 | ||
1da177e4 LT |
1641 | /* |
1642 | * Ok, this is the main fork-routine. | |
1643 | * | |
1644 | * It copies the process, and if successful kick-starts | |
1645 | * it and waits for it to finish using the VM if required. | |
1646 | */ | |
1647 | long do_fork(unsigned long clone_flags, | |
1648 | unsigned long stack_start, | |
1da177e4 LT |
1649 | unsigned long stack_size, |
1650 | int __user *parent_tidptr, | |
1651 | int __user *child_tidptr) | |
1652 | { | |
1653 | struct task_struct *p; | |
1654 | int trace = 0; | |
92476d7f | 1655 | long nr; |
1da177e4 | 1656 | |
09a05394 | 1657 | /* |
4b9d33e6 TH |
1658 | * Determine whether and which event to report to ptracer. When |
1659 | * called from kernel_thread or CLONE_UNTRACED is explicitly | |
1660 | * requested, no event is reported; otherwise, report if the event | |
1661 | * for the type of forking is enabled. | |
09a05394 | 1662 | */ |
e80d6661 | 1663 | if (!(clone_flags & CLONE_UNTRACED)) { |
4b9d33e6 TH |
1664 | if (clone_flags & CLONE_VFORK) |
1665 | trace = PTRACE_EVENT_VFORK; | |
1666 | else if ((clone_flags & CSIGNAL) != SIGCHLD) | |
1667 | trace = PTRACE_EVENT_CLONE; | |
1668 | else | |
1669 | trace = PTRACE_EVENT_FORK; | |
1670 | ||
1671 | if (likely(!ptrace_event_enabled(current, trace))) | |
1672 | trace = 0; | |
1673 | } | |
1da177e4 | 1674 | |
62e791c1 | 1675 | p = copy_process(clone_flags, stack_start, stack_size, |
09a05394 | 1676 | child_tidptr, NULL, trace); |
1da177e4 LT |
1677 | /* |
1678 | * Do this prior waking up the new thread - the thread pointer | |
1679 | * might get invalid after that point, if the thread exits quickly. | |
1680 | */ | |
1681 | if (!IS_ERR(p)) { | |
1682 | struct completion vfork; | |
4e52365f | 1683 | struct pid *pid; |
1da177e4 | 1684 | |
0a16b607 MD |
1685 | trace_sched_process_fork(current, p); |
1686 | ||
4e52365f MD |
1687 | pid = get_task_pid(p, PIDTYPE_PID); |
1688 | nr = pid_vnr(pid); | |
30e49c26 PE |
1689 | |
1690 | if (clone_flags & CLONE_PARENT_SETTID) | |
1691 | put_user(nr, parent_tidptr); | |
a6f5e063 | 1692 | |
1da177e4 LT |
1693 | if (clone_flags & CLONE_VFORK) { |
1694 | p->vfork_done = &vfork; | |
1695 | init_completion(&vfork); | |
d68b46fe | 1696 | get_task_struct(p); |
1da177e4 LT |
1697 | } |
1698 | ||
3e51e3ed | 1699 | wake_up_new_task(p); |
1da177e4 | 1700 | |
4b9d33e6 TH |
1701 | /* forking complete and child started to run, tell ptracer */ |
1702 | if (unlikely(trace)) | |
4e52365f | 1703 | ptrace_event_pid(trace, pid); |
09a05394 | 1704 | |
1da177e4 | 1705 | if (clone_flags & CLONE_VFORK) { |
d68b46fe | 1706 | if (!wait_for_vfork_done(p, &vfork)) |
4e52365f | 1707 | ptrace_event_pid(PTRACE_EVENT_VFORK_DONE, pid); |
1da177e4 | 1708 | } |
4e52365f MD |
1709 | |
1710 | put_pid(pid); | |
1da177e4 | 1711 | } else { |
92476d7f | 1712 | nr = PTR_ERR(p); |
1da177e4 | 1713 | } |
92476d7f | 1714 | return nr; |
1da177e4 LT |
1715 | } |
1716 | ||
2aa3a7f8 AV |
1717 | /* |
1718 | * Create a kernel thread. | |
1719 | */ | |
1720 | pid_t kernel_thread(int (*fn)(void *), void *arg, unsigned long flags) | |
1721 | { | |
e80d6661 | 1722 | return do_fork(flags|CLONE_VM|CLONE_UNTRACED, (unsigned long)fn, |
2aa3a7f8 AV |
1723 | (unsigned long)arg, NULL, NULL); |
1724 | } | |
2aa3a7f8 | 1725 | |
d2125043 AV |
1726 | #ifdef __ARCH_WANT_SYS_FORK |
1727 | SYSCALL_DEFINE0(fork) | |
1728 | { | |
1729 | #ifdef CONFIG_MMU | |
e80d6661 | 1730 | return do_fork(SIGCHLD, 0, 0, NULL, NULL); |
d2125043 AV |
1731 | #else |
1732 | /* can not support in nommu mode */ | |
5d59e182 | 1733 | return -EINVAL; |
d2125043 AV |
1734 | #endif |
1735 | } | |
1736 | #endif | |
1737 | ||
1738 | #ifdef __ARCH_WANT_SYS_VFORK | |
1739 | SYSCALL_DEFINE0(vfork) | |
1740 | { | |
5d59e182 | 1741 | return do_fork(CLONE_VFORK | CLONE_VM | SIGCHLD, 0, |
d2125043 AV |
1742 | 0, NULL, NULL); |
1743 | } | |
1744 | #endif | |
1745 | ||
1746 | #ifdef __ARCH_WANT_SYS_CLONE | |
1747 | #ifdef CONFIG_CLONE_BACKWARDS | |
1748 | SYSCALL_DEFINE5(clone, unsigned long, clone_flags, unsigned long, newsp, | |
1749 | int __user *, parent_tidptr, | |
1750 | int, tls_val, | |
1751 | int __user *, child_tidptr) | |
1752 | #elif defined(CONFIG_CLONE_BACKWARDS2) | |
1753 | SYSCALL_DEFINE5(clone, unsigned long, newsp, unsigned long, clone_flags, | |
1754 | int __user *, parent_tidptr, | |
1755 | int __user *, child_tidptr, | |
1756 | int, tls_val) | |
dfa9771a MS |
1757 | #elif defined(CONFIG_CLONE_BACKWARDS3) |
1758 | SYSCALL_DEFINE6(clone, unsigned long, clone_flags, unsigned long, newsp, | |
1759 | int, stack_size, | |
1760 | int __user *, parent_tidptr, | |
1761 | int __user *, child_tidptr, | |
1762 | int, tls_val) | |
d2125043 AV |
1763 | #else |
1764 | SYSCALL_DEFINE5(clone, unsigned long, clone_flags, unsigned long, newsp, | |
1765 | int __user *, parent_tidptr, | |
1766 | int __user *, child_tidptr, | |
1767 | int, tls_val) | |
1768 | #endif | |
1769 | { | |
2cf09666 | 1770 | return do_fork(clone_flags, newsp, 0, parent_tidptr, child_tidptr); |
d2125043 AV |
1771 | } |
1772 | #endif | |
1773 | ||
5fd63b30 RT |
1774 | #ifndef ARCH_MIN_MMSTRUCT_ALIGN |
1775 | #define ARCH_MIN_MMSTRUCT_ALIGN 0 | |
1776 | #endif | |
1777 | ||
51cc5068 | 1778 | static void sighand_ctor(void *data) |
aa1757f9 ON |
1779 | { |
1780 | struct sighand_struct *sighand = data; | |
1781 | ||
a35afb83 | 1782 | spin_lock_init(&sighand->siglock); |
b8fceee1 | 1783 | init_waitqueue_head(&sighand->signalfd_wqh); |
aa1757f9 ON |
1784 | } |
1785 | ||
1da177e4 LT |
1786 | void __init proc_caches_init(void) |
1787 | { | |
1788 | sighand_cachep = kmem_cache_create("sighand_cache", | |
1789 | sizeof(struct sighand_struct), 0, | |
2dff4405 VN |
1790 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_DESTROY_BY_RCU| |
1791 | SLAB_NOTRACK, sighand_ctor); | |
1da177e4 LT |
1792 | signal_cachep = kmem_cache_create("signal_cache", |
1793 | sizeof(struct signal_struct), 0, | |
2dff4405 | 1794 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL); |
20c2df83 | 1795 | files_cachep = kmem_cache_create("files_cache", |
1da177e4 | 1796 | sizeof(struct files_struct), 0, |
2dff4405 | 1797 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL); |
20c2df83 | 1798 | fs_cachep = kmem_cache_create("fs_cache", |
1da177e4 | 1799 | sizeof(struct fs_struct), 0, |
2dff4405 | 1800 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL); |
6345d24d LT |
1801 | /* |
1802 | * FIXME! The "sizeof(struct mm_struct)" currently includes the | |
1803 | * whole struct cpumask for the OFFSTACK case. We could change | |
1804 | * this to *only* allocate as much of it as required by the | |
1805 | * maximum number of CPU's we can ever have. The cpumask_allocation | |
1806 | * is at the end of the structure, exactly for that reason. | |
1807 | */ | |
1da177e4 | 1808 | mm_cachep = kmem_cache_create("mm_struct", |
5fd63b30 | 1809 | sizeof(struct mm_struct), ARCH_MIN_MMSTRUCT_ALIGN, |
2dff4405 | 1810 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL); |
33e5d769 | 1811 | vm_area_cachep = KMEM_CACHE(vm_area_struct, SLAB_PANIC); |
8feae131 | 1812 | mmap_init(); |
66577193 | 1813 | nsproxy_cache_init(); |
1da177e4 | 1814 | } |
cf2e340f | 1815 | |
cf2e340f | 1816 | /* |
9bfb23fc | 1817 | * Check constraints on flags passed to the unshare system call. |
cf2e340f | 1818 | */ |
9bfb23fc | 1819 | static int check_unshare_flags(unsigned long unshare_flags) |
cf2e340f | 1820 | { |
9bfb23fc ON |
1821 | if (unshare_flags & ~(CLONE_THREAD|CLONE_FS|CLONE_NEWNS|CLONE_SIGHAND| |
1822 | CLONE_VM|CLONE_FILES|CLONE_SYSVSEM| | |
50804fe3 | 1823 | CLONE_NEWUTS|CLONE_NEWIPC|CLONE_NEWNET| |
b2e0d987 | 1824 | CLONE_NEWUSER|CLONE_NEWPID)) |
9bfb23fc | 1825 | return -EINVAL; |
cf2e340f | 1826 | /* |
9bfb23fc ON |
1827 | * Not implemented, but pretend it works if there is nothing to |
1828 | * unshare. Note that unsharing CLONE_THREAD or CLONE_SIGHAND | |
1829 | * needs to unshare vm. | |
cf2e340f | 1830 | */ |
9bfb23fc ON |
1831 | if (unshare_flags & (CLONE_THREAD | CLONE_SIGHAND | CLONE_VM)) { |
1832 | /* FIXME: get_task_mm() increments ->mm_users */ | |
1833 | if (atomic_read(¤t->mm->mm_users) > 1) | |
1834 | return -EINVAL; | |
1835 | } | |
cf2e340f JD |
1836 | |
1837 | return 0; | |
1838 | } | |
1839 | ||
1840 | /* | |
99d1419d | 1841 | * Unshare the filesystem structure if it is being shared |
cf2e340f JD |
1842 | */ |
1843 | static int unshare_fs(unsigned long unshare_flags, struct fs_struct **new_fsp) | |
1844 | { | |
1845 | struct fs_struct *fs = current->fs; | |
1846 | ||
498052bb AV |
1847 | if (!(unshare_flags & CLONE_FS) || !fs) |
1848 | return 0; | |
1849 | ||
1850 | /* don't need lock here; in the worst case we'll do useless copy */ | |
1851 | if (fs->users == 1) | |
1852 | return 0; | |
1853 | ||
1854 | *new_fsp = copy_fs_struct(fs); | |
1855 | if (!*new_fsp) | |
1856 | return -ENOMEM; | |
cf2e340f JD |
1857 | |
1858 | return 0; | |
1859 | } | |
1860 | ||
cf2e340f | 1861 | /* |
a016f338 | 1862 | * Unshare file descriptor table if it is being shared |
cf2e340f JD |
1863 | */ |
1864 | static int unshare_fd(unsigned long unshare_flags, struct files_struct **new_fdp) | |
1865 | { | |
1866 | struct files_struct *fd = current->files; | |
a016f338 | 1867 | int error = 0; |
cf2e340f JD |
1868 | |
1869 | if ((unshare_flags & CLONE_FILES) && | |
a016f338 JD |
1870 | (fd && atomic_read(&fd->count) > 1)) { |
1871 | *new_fdp = dup_fd(fd, &error); | |
1872 | if (!*new_fdp) | |
1873 | return error; | |
1874 | } | |
cf2e340f JD |
1875 | |
1876 | return 0; | |
1877 | } | |
1878 | ||
cf2e340f JD |
1879 | /* |
1880 | * unshare allows a process to 'unshare' part of the process | |
1881 | * context which was originally shared using clone. copy_* | |
1882 | * functions used by do_fork() cannot be used here directly | |
1883 | * because they modify an inactive task_struct that is being | |
1884 | * constructed. Here we are modifying the current, active, | |
1885 | * task_struct. | |
1886 | */ | |
6559eed8 | 1887 | SYSCALL_DEFINE1(unshare, unsigned long, unshare_flags) |
cf2e340f | 1888 | { |
cf2e340f | 1889 | struct fs_struct *fs, *new_fs = NULL; |
cf2e340f | 1890 | struct files_struct *fd, *new_fd = NULL; |
b2e0d987 | 1891 | struct cred *new_cred = NULL; |
cf7b708c | 1892 | struct nsproxy *new_nsproxy = NULL; |
9edff4ab | 1893 | int do_sysvsem = 0; |
9bfb23fc | 1894 | int err; |
cf2e340f | 1895 | |
b2e0d987 EB |
1896 | /* |
1897 | * If unsharing a user namespace must also unshare the thread. | |
1898 | */ | |
1899 | if (unshare_flags & CLONE_NEWUSER) | |
e66eded8 | 1900 | unshare_flags |= CLONE_THREAD | CLONE_FS; |
50804fe3 EB |
1901 | /* |
1902 | * If unsharing a thread from a thread group, must also unshare vm. | |
1903 | */ | |
1904 | if (unshare_flags & CLONE_THREAD) | |
1905 | unshare_flags |= CLONE_VM; | |
1906 | /* | |
1907 | * If unsharing vm, must also unshare signal handlers. | |
1908 | */ | |
1909 | if (unshare_flags & CLONE_VM) | |
1910 | unshare_flags |= CLONE_SIGHAND; | |
9bfb23fc ON |
1911 | /* |
1912 | * If unsharing namespace, must also unshare filesystem information. | |
1913 | */ | |
1914 | if (unshare_flags & CLONE_NEWNS) | |
1915 | unshare_flags |= CLONE_FS; | |
50804fe3 EB |
1916 | |
1917 | err = check_unshare_flags(unshare_flags); | |
1918 | if (err) | |
1919 | goto bad_unshare_out; | |
6013f67f MS |
1920 | /* |
1921 | * CLONE_NEWIPC must also detach from the undolist: after switching | |
1922 | * to a new ipc namespace, the semaphore arrays from the old | |
1923 | * namespace are unreachable. | |
1924 | */ | |
1925 | if (unshare_flags & (CLONE_NEWIPC|CLONE_SYSVSEM)) | |
9edff4ab | 1926 | do_sysvsem = 1; |
fb0a685c DRO |
1927 | err = unshare_fs(unshare_flags, &new_fs); |
1928 | if (err) | |
9bfb23fc | 1929 | goto bad_unshare_out; |
fb0a685c DRO |
1930 | err = unshare_fd(unshare_flags, &new_fd); |
1931 | if (err) | |
9bfb23fc | 1932 | goto bad_unshare_cleanup_fs; |
b2e0d987 | 1933 | err = unshare_userns(unshare_flags, &new_cred); |
fb0a685c | 1934 | if (err) |
9edff4ab | 1935 | goto bad_unshare_cleanup_fd; |
b2e0d987 EB |
1936 | err = unshare_nsproxy_namespaces(unshare_flags, &new_nsproxy, |
1937 | new_cred, new_fs); | |
1938 | if (err) | |
1939 | goto bad_unshare_cleanup_cred; | |
c0b2fc31 | 1940 | |
b2e0d987 | 1941 | if (new_fs || new_fd || do_sysvsem || new_cred || new_nsproxy) { |
9edff4ab MS |
1942 | if (do_sysvsem) { |
1943 | /* | |
1944 | * CLONE_SYSVSEM is equivalent to sys_exit(). | |
1945 | */ | |
1946 | exit_sem(current); | |
1947 | } | |
ab602f79 JM |
1948 | if (unshare_flags & CLONE_NEWIPC) { |
1949 | /* Orphan segments in old ns (see sem above). */ | |
1950 | exit_shm(current); | |
1951 | shm_init_task(current); | |
1952 | } | |
ab516013 | 1953 | |
6f977e6b | 1954 | if (new_nsproxy) |
cf7b708c | 1955 | switch_task_namespaces(current, new_nsproxy); |
cf2e340f | 1956 | |
cf7b708c PE |
1957 | task_lock(current); |
1958 | ||
cf2e340f JD |
1959 | if (new_fs) { |
1960 | fs = current->fs; | |
2a4419b5 | 1961 | spin_lock(&fs->lock); |
cf2e340f | 1962 | current->fs = new_fs; |
498052bb AV |
1963 | if (--fs->users) |
1964 | new_fs = NULL; | |
1965 | else | |
1966 | new_fs = fs; | |
2a4419b5 | 1967 | spin_unlock(&fs->lock); |
cf2e340f JD |
1968 | } |
1969 | ||
cf2e340f JD |
1970 | if (new_fd) { |
1971 | fd = current->files; | |
1972 | current->files = new_fd; | |
1973 | new_fd = fd; | |
1974 | } | |
1975 | ||
1976 | task_unlock(current); | |
b2e0d987 EB |
1977 | |
1978 | if (new_cred) { | |
1979 | /* Install the new user namespace */ | |
1980 | commit_creds(new_cred); | |
1981 | new_cred = NULL; | |
1982 | } | |
cf2e340f JD |
1983 | } |
1984 | ||
b2e0d987 EB |
1985 | bad_unshare_cleanup_cred: |
1986 | if (new_cred) | |
1987 | put_cred(new_cred); | |
cf2e340f JD |
1988 | bad_unshare_cleanup_fd: |
1989 | if (new_fd) | |
1990 | put_files_struct(new_fd); | |
1991 | ||
cf2e340f JD |
1992 | bad_unshare_cleanup_fs: |
1993 | if (new_fs) | |
498052bb | 1994 | free_fs_struct(new_fs); |
cf2e340f | 1995 | |
cf2e340f JD |
1996 | bad_unshare_out: |
1997 | return err; | |
1998 | } | |
3b125388 AV |
1999 | |
2000 | /* | |
2001 | * Helper to unshare the files of the current task. | |
2002 | * We don't want to expose copy_files internals to | |
2003 | * the exec layer of the kernel. | |
2004 | */ | |
2005 | ||
2006 | int unshare_files(struct files_struct **displaced) | |
2007 | { | |
2008 | struct task_struct *task = current; | |
50704516 | 2009 | struct files_struct *copy = NULL; |
3b125388 AV |
2010 | int error; |
2011 | ||
2012 | error = unshare_fd(CLONE_FILES, ©); | |
2013 | if (error || !copy) { | |
2014 | *displaced = NULL; | |
2015 | return error; | |
2016 | } | |
2017 | *displaced = task->files; | |
2018 | task_lock(task); | |
2019 | task->files = copy; | |
2020 | task_unlock(task); | |
2021 | return 0; | |
2022 | } |