]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * linux/kernel/fork.c | |
3 | * | |
4 | * Copyright (C) 1991, 1992 Linus Torvalds | |
5 | */ | |
6 | ||
7 | /* | |
8 | * 'fork.c' contains the help-routines for the 'fork' system call | |
9 | * (see also entry.S and others). | |
10 | * Fork is rather simple, once you get the hang of it, but the memory | |
11 | * management can be a bitch. See 'mm/memory.c': 'copy_page_range()' | |
12 | */ | |
13 | ||
1da177e4 LT |
14 | #include <linux/slab.h> |
15 | #include <linux/init.h> | |
16 | #include <linux/unistd.h> | |
1da177e4 LT |
17 | #include <linux/module.h> |
18 | #include <linux/vmalloc.h> | |
19 | #include <linux/completion.h> | |
1da177e4 LT |
20 | #include <linux/personality.h> |
21 | #include <linux/mempolicy.h> | |
22 | #include <linux/sem.h> | |
23 | #include <linux/file.h> | |
9f3acc31 | 24 | #include <linux/fdtable.h> |
da9cbc87 | 25 | #include <linux/iocontext.h> |
1da177e4 LT |
26 | #include <linux/key.h> |
27 | #include <linux/binfmts.h> | |
28 | #include <linux/mman.h> | |
cddb8a5c | 29 | #include <linux/mmu_notifier.h> |
1da177e4 | 30 | #include <linux/fs.h> |
615d6e87 DB |
31 | #include <linux/mm.h> |
32 | #include <linux/vmacache.h> | |
ab516013 | 33 | #include <linux/nsproxy.h> |
c59ede7b | 34 | #include <linux/capability.h> |
1da177e4 | 35 | #include <linux/cpu.h> |
b4f48b63 | 36 | #include <linux/cgroup.h> |
1da177e4 | 37 | #include <linux/security.h> |
a1e78772 | 38 | #include <linux/hugetlb.h> |
e2cfabdf | 39 | #include <linux/seccomp.h> |
1da177e4 LT |
40 | #include <linux/swap.h> |
41 | #include <linux/syscalls.h> | |
42 | #include <linux/jiffies.h> | |
43 | #include <linux/futex.h> | |
8141c7f3 | 44 | #include <linux/compat.h> |
207205a2 | 45 | #include <linux/kthread.h> |
7c3ab738 | 46 | #include <linux/task_io_accounting_ops.h> |
ab2af1f5 | 47 | #include <linux/rcupdate.h> |
1da177e4 LT |
48 | #include <linux/ptrace.h> |
49 | #include <linux/mount.h> | |
50 | #include <linux/audit.h> | |
78fb7466 | 51 | #include <linux/memcontrol.h> |
f201ae23 | 52 | #include <linux/ftrace.h> |
5e2bf014 | 53 | #include <linux/proc_fs.h> |
1da177e4 LT |
54 | #include <linux/profile.h> |
55 | #include <linux/rmap.h> | |
f8af4da3 | 56 | #include <linux/ksm.h> |
1da177e4 | 57 | #include <linux/acct.h> |
8f0ab514 | 58 | #include <linux/tsacct_kern.h> |
9f46080c | 59 | #include <linux/cn_proc.h> |
ba96a0c8 | 60 | #include <linux/freezer.h> |
ca74e92b | 61 | #include <linux/delayacct.h> |
ad4ecbcb | 62 | #include <linux/taskstats_kern.h> |
0a425405 | 63 | #include <linux/random.h> |
522ed776 | 64 | #include <linux/tty.h> |
fd0928df | 65 | #include <linux/blkdev.h> |
5ad4e53b | 66 | #include <linux/fs_struct.h> |
7c9f8861 | 67 | #include <linux/magic.h> |
cdd6c482 | 68 | #include <linux/perf_event.h> |
42c4ab41 | 69 | #include <linux/posix-timers.h> |
8e7cac79 | 70 | #include <linux/user-return-notifier.h> |
3d5992d2 | 71 | #include <linux/oom.h> |
ba76149f | 72 | #include <linux/khugepaged.h> |
d80e731e | 73 | #include <linux/signalfd.h> |
0326f5a9 | 74 | #include <linux/uprobes.h> |
a27bb332 | 75 | #include <linux/aio.h> |
52f5684c | 76 | #include <linux/compiler.h> |
16db3d3f | 77 | #include <linux/sysctl.h> |
5c9a8750 | 78 | #include <linux/kcov.h> |
1da177e4 LT |
79 | |
80 | #include <asm/pgtable.h> | |
81 | #include <asm/pgalloc.h> | |
82 | #include <asm/uaccess.h> | |
83 | #include <asm/mmu_context.h> | |
84 | #include <asm/cacheflush.h> | |
85 | #include <asm/tlbflush.h> | |
86 | ||
ad8d75ff SR |
87 | #include <trace/events/sched.h> |
88 | ||
43d2b113 KH |
89 | #define CREATE_TRACE_POINTS |
90 | #include <trace/events/task.h> | |
91 | ||
ac1b398d HS |
92 | /* |
93 | * Minimum number of threads to boot the kernel | |
94 | */ | |
95 | #define MIN_THREADS 20 | |
96 | ||
97 | /* | |
98 | * Maximum number of threads | |
99 | */ | |
100 | #define MAX_THREADS FUTEX_TID_MASK | |
101 | ||
1da177e4 LT |
102 | /* |
103 | * Protected counters by write_lock_irq(&tasklist_lock) | |
104 | */ | |
105 | unsigned long total_forks; /* Handle normal Linux uptimes. */ | |
fb0a685c | 106 | int nr_threads; /* The idle threads do not count.. */ |
1da177e4 LT |
107 | |
108 | int max_threads; /* tunable limit on nr_threads */ | |
109 | ||
110 | DEFINE_PER_CPU(unsigned long, process_counts) = 0; | |
111 | ||
c59923a1 | 112 | __cacheline_aligned DEFINE_RWLOCK(tasklist_lock); /* outer */ |
db1466b3 PM |
113 | |
114 | #ifdef CONFIG_PROVE_RCU | |
115 | int lockdep_tasklist_lock_is_held(void) | |
116 | { | |
117 | return lockdep_is_held(&tasklist_lock); | |
118 | } | |
119 | EXPORT_SYMBOL_GPL(lockdep_tasklist_lock_is_held); | |
120 | #endif /* #ifdef CONFIG_PROVE_RCU */ | |
1da177e4 LT |
121 | |
122 | int nr_processes(void) | |
123 | { | |
124 | int cpu; | |
125 | int total = 0; | |
126 | ||
1d510750 | 127 | for_each_possible_cpu(cpu) |
1da177e4 LT |
128 | total += per_cpu(process_counts, cpu); |
129 | ||
130 | return total; | |
131 | } | |
132 | ||
f19b9f74 AM |
133 | void __weak arch_release_task_struct(struct task_struct *tsk) |
134 | { | |
135 | } | |
136 | ||
f5e10287 | 137 | #ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR |
e18b890b | 138 | static struct kmem_cache *task_struct_cachep; |
41101809 TG |
139 | |
140 | static inline struct task_struct *alloc_task_struct_node(int node) | |
141 | { | |
142 | return kmem_cache_alloc_node(task_struct_cachep, GFP_KERNEL, node); | |
143 | } | |
144 | ||
41101809 TG |
145 | static inline void free_task_struct(struct task_struct *tsk) |
146 | { | |
41101809 TG |
147 | kmem_cache_free(task_struct_cachep, tsk); |
148 | } | |
1da177e4 LT |
149 | #endif |
150 | ||
f19b9f74 AM |
151 | void __weak arch_release_thread_info(struct thread_info *ti) |
152 | { | |
153 | } | |
154 | ||
f5e10287 | 155 | #ifndef CONFIG_ARCH_THREAD_INFO_ALLOCATOR |
41101809 | 156 | |
0d15d74a TG |
157 | /* |
158 | * Allocate pages if THREAD_SIZE is >= PAGE_SIZE, otherwise use a | |
159 | * kmemcache based allocator. | |
160 | */ | |
161 | # if THREAD_SIZE >= PAGE_SIZE | |
b6a84016 ED |
162 | static struct thread_info *alloc_thread_info_node(struct task_struct *tsk, |
163 | int node) | |
b69c49b7 | 164 | { |
52383431 VD |
165 | struct page *page = alloc_kmem_pages_node(node, THREADINFO_GFP, |
166 | THREAD_SIZE_ORDER); | |
b6a84016 | 167 | |
12580e4b VD |
168 | if (page) |
169 | memcg_kmem_update_page_stat(page, MEMCG_KERNEL_STACK, | |
170 | 1 << THREAD_SIZE_ORDER); | |
171 | ||
b6a84016 | 172 | return page ? page_address(page) : NULL; |
b69c49b7 FT |
173 | } |
174 | ||
175 | static inline void free_thread_info(struct thread_info *ti) | |
176 | { | |
12580e4b VD |
177 | struct page *page = virt_to_page(ti); |
178 | ||
179 | memcg_kmem_update_page_stat(page, MEMCG_KERNEL_STACK, | |
180 | -(1 << THREAD_SIZE_ORDER)); | |
181 | __free_kmem_pages(page, THREAD_SIZE_ORDER); | |
b69c49b7 | 182 | } |
0d15d74a TG |
183 | # else |
184 | static struct kmem_cache *thread_info_cache; | |
185 | ||
186 | static struct thread_info *alloc_thread_info_node(struct task_struct *tsk, | |
187 | int node) | |
188 | { | |
189 | return kmem_cache_alloc_node(thread_info_cache, THREADINFO_GFP, node); | |
190 | } | |
191 | ||
192 | static void free_thread_info(struct thread_info *ti) | |
193 | { | |
0d15d74a TG |
194 | kmem_cache_free(thread_info_cache, ti); |
195 | } | |
196 | ||
197 | void thread_info_cache_init(void) | |
198 | { | |
199 | thread_info_cache = kmem_cache_create("thread_info", THREAD_SIZE, | |
200 | THREAD_SIZE, 0, NULL); | |
201 | BUG_ON(thread_info_cache == NULL); | |
202 | } | |
203 | # endif | |
b69c49b7 FT |
204 | #endif |
205 | ||
1da177e4 | 206 | /* SLAB cache for signal_struct structures (tsk->signal) */ |
e18b890b | 207 | static struct kmem_cache *signal_cachep; |
1da177e4 LT |
208 | |
209 | /* SLAB cache for sighand_struct structures (tsk->sighand) */ | |
e18b890b | 210 | struct kmem_cache *sighand_cachep; |
1da177e4 LT |
211 | |
212 | /* SLAB cache for files_struct structures (tsk->files) */ | |
e18b890b | 213 | struct kmem_cache *files_cachep; |
1da177e4 LT |
214 | |
215 | /* SLAB cache for fs_struct structures (tsk->fs) */ | |
e18b890b | 216 | struct kmem_cache *fs_cachep; |
1da177e4 LT |
217 | |
218 | /* SLAB cache for vm_area_struct structures */ | |
e18b890b | 219 | struct kmem_cache *vm_area_cachep; |
1da177e4 LT |
220 | |
221 | /* SLAB cache for mm_struct structures (tsk->mm) */ | |
e18b890b | 222 | static struct kmem_cache *mm_cachep; |
1da177e4 | 223 | |
c6a7f572 KM |
224 | static void account_kernel_stack(struct thread_info *ti, int account) |
225 | { | |
226 | struct zone *zone = page_zone(virt_to_page(ti)); | |
227 | ||
228 | mod_zone_page_state(zone, NR_KERNEL_STACK, account); | |
229 | } | |
230 | ||
1da177e4 LT |
231 | void free_task(struct task_struct *tsk) |
232 | { | |
c6a7f572 | 233 | account_kernel_stack(tsk->stack, -1); |
f19b9f74 | 234 | arch_release_thread_info(tsk->stack); |
f7e4217b | 235 | free_thread_info(tsk->stack); |
23f78d4a | 236 | rt_mutex_debug_task_free(tsk); |
fb52607a | 237 | ftrace_graph_exit_task(tsk); |
e2cfabdf | 238 | put_seccomp_filter(tsk); |
f19b9f74 | 239 | arch_release_task_struct(tsk); |
1da177e4 LT |
240 | free_task_struct(tsk); |
241 | } | |
242 | EXPORT_SYMBOL(free_task); | |
243 | ||
ea6d290c ON |
244 | static inline void free_signal_struct(struct signal_struct *sig) |
245 | { | |
97101eb4 | 246 | taskstats_tgid_free(sig); |
1c5354de | 247 | sched_autogroup_exit(sig); |
ea6d290c ON |
248 | kmem_cache_free(signal_cachep, sig); |
249 | } | |
250 | ||
251 | static inline void put_signal_struct(struct signal_struct *sig) | |
252 | { | |
1c5354de | 253 | if (atomic_dec_and_test(&sig->sigcnt)) |
ea6d290c ON |
254 | free_signal_struct(sig); |
255 | } | |
256 | ||
158d9ebd | 257 | void __put_task_struct(struct task_struct *tsk) |
1da177e4 | 258 | { |
270f722d | 259 | WARN_ON(!tsk->exit_state); |
1da177e4 LT |
260 | WARN_ON(atomic_read(&tsk->usage)); |
261 | WARN_ON(tsk == current); | |
262 | ||
2e91fa7f | 263 | cgroup_free(tsk); |
156654f4 | 264 | task_numa_free(tsk); |
1a2a4d06 | 265 | security_task_free(tsk); |
e0e81739 | 266 | exit_creds(tsk); |
35df17c5 | 267 | delayacct_tsk_free(tsk); |
ea6d290c | 268 | put_signal_struct(tsk->signal); |
1da177e4 LT |
269 | |
270 | if (!profile_handoff_task(tsk)) | |
271 | free_task(tsk); | |
272 | } | |
77c100c8 | 273 | EXPORT_SYMBOL_GPL(__put_task_struct); |
1da177e4 | 274 | |
6c0a9fa6 | 275 | void __init __weak arch_task_cache_init(void) { } |
61c4628b | 276 | |
ff691f6e HS |
277 | /* |
278 | * set_max_threads | |
279 | */ | |
16db3d3f | 280 | static void set_max_threads(unsigned int max_threads_suggested) |
ff691f6e | 281 | { |
ac1b398d | 282 | u64 threads; |
ff691f6e HS |
283 | |
284 | /* | |
ac1b398d HS |
285 | * The number of threads shall be limited such that the thread |
286 | * structures may only consume a small part of the available memory. | |
ff691f6e | 287 | */ |
ac1b398d HS |
288 | if (fls64(totalram_pages) + fls64(PAGE_SIZE) > 64) |
289 | threads = MAX_THREADS; | |
290 | else | |
291 | threads = div64_u64((u64) totalram_pages * (u64) PAGE_SIZE, | |
292 | (u64) THREAD_SIZE * 8UL); | |
293 | ||
16db3d3f HS |
294 | if (threads > max_threads_suggested) |
295 | threads = max_threads_suggested; | |
296 | ||
ac1b398d | 297 | max_threads = clamp_t(u64, threads, MIN_THREADS, MAX_THREADS); |
ff691f6e HS |
298 | } |
299 | ||
5aaeb5c0 IM |
300 | #ifdef CONFIG_ARCH_WANTS_DYNAMIC_TASK_STRUCT |
301 | /* Initialized by the architecture: */ | |
302 | int arch_task_struct_size __read_mostly; | |
303 | #endif | |
0c8c0f03 | 304 | |
ff691f6e | 305 | void __init fork_init(void) |
1da177e4 | 306 | { |
f5e10287 | 307 | #ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR |
1da177e4 LT |
308 | #ifndef ARCH_MIN_TASKALIGN |
309 | #define ARCH_MIN_TASKALIGN L1_CACHE_BYTES | |
310 | #endif | |
311 | /* create a slab on which task_structs can be allocated */ | |
5d097056 VD |
312 | task_struct_cachep = kmem_cache_create("task_struct", |
313 | arch_task_struct_size, ARCH_MIN_TASKALIGN, | |
314 | SLAB_PANIC|SLAB_NOTRACK|SLAB_ACCOUNT, NULL); | |
1da177e4 LT |
315 | #endif |
316 | ||
61c4628b SS |
317 | /* do the arch specific task caches init */ |
318 | arch_task_cache_init(); | |
319 | ||
16db3d3f | 320 | set_max_threads(MAX_THREADS); |
1da177e4 LT |
321 | |
322 | init_task.signal->rlim[RLIMIT_NPROC].rlim_cur = max_threads/2; | |
323 | init_task.signal->rlim[RLIMIT_NPROC].rlim_max = max_threads/2; | |
324 | init_task.signal->rlim[RLIMIT_SIGPENDING] = | |
325 | init_task.signal->rlim[RLIMIT_NPROC]; | |
326 | } | |
327 | ||
52f5684c | 328 | int __weak arch_dup_task_struct(struct task_struct *dst, |
61c4628b SS |
329 | struct task_struct *src) |
330 | { | |
331 | *dst = *src; | |
332 | return 0; | |
333 | } | |
334 | ||
d4311ff1 AT |
335 | void set_task_stack_end_magic(struct task_struct *tsk) |
336 | { | |
337 | unsigned long *stackend; | |
338 | ||
339 | stackend = end_of_stack(tsk); | |
340 | *stackend = STACK_END_MAGIC; /* for overflow detection */ | |
341 | } | |
342 | ||
725fc629 | 343 | static struct task_struct *dup_task_struct(struct task_struct *orig, int node) |
1da177e4 LT |
344 | { |
345 | struct task_struct *tsk; | |
346 | struct thread_info *ti; | |
3e26c149 | 347 | int err; |
1da177e4 | 348 | |
725fc629 AK |
349 | if (node == NUMA_NO_NODE) |
350 | node = tsk_fork_get_node(orig); | |
504f52b5 | 351 | tsk = alloc_task_struct_node(node); |
1da177e4 LT |
352 | if (!tsk) |
353 | return NULL; | |
354 | ||
b6a84016 | 355 | ti = alloc_thread_info_node(tsk, node); |
f19b9f74 AM |
356 | if (!ti) |
357 | goto free_tsk; | |
1da177e4 | 358 | |
fb0a685c | 359 | err = arch_dup_task_struct(tsk, orig); |
164c33c6 | 360 | if (err) |
f19b9f74 | 361 | goto free_ti; |
164c33c6 | 362 | |
87bec58a | 363 | tsk->stack = ti; |
dbd95212 KC |
364 | #ifdef CONFIG_SECCOMP |
365 | /* | |
366 | * We must handle setting up seccomp filters once we're under | |
367 | * the sighand lock in case orig has changed between now and | |
368 | * then. Until then, filter must be NULL to avoid messing up | |
369 | * the usage counts on the error path calling free_task. | |
370 | */ | |
371 | tsk->seccomp.filter = NULL; | |
372 | #endif | |
87bec58a AM |
373 | |
374 | setup_thread_stack(tsk, orig); | |
8e7cac79 | 375 | clear_user_return_notifier(tsk); |
f26f9aff | 376 | clear_tsk_need_resched(tsk); |
d4311ff1 | 377 | set_task_stack_end_magic(tsk); |
1da177e4 | 378 | |
0a425405 AV |
379 | #ifdef CONFIG_CC_STACKPROTECTOR |
380 | tsk->stack_canary = get_random_int(); | |
381 | #endif | |
382 | ||
fb0a685c DRO |
383 | /* |
384 | * One for us, one for whoever does the "release_task()" (usually | |
385 | * parent) | |
386 | */ | |
387 | atomic_set(&tsk->usage, 2); | |
6c5c9341 | 388 | #ifdef CONFIG_BLK_DEV_IO_TRACE |
2056a782 | 389 | tsk->btrace_seq = 0; |
6c5c9341 | 390 | #endif |
a0aa7f68 | 391 | tsk->splice_pipe = NULL; |
5640f768 | 392 | tsk->task_frag.page = NULL; |
093e5840 | 393 | tsk->wake_q.next = NULL; |
c6a7f572 KM |
394 | |
395 | account_kernel_stack(ti, 1); | |
396 | ||
5c9a8750 DV |
397 | kcov_task_init(tsk); |
398 | ||
1da177e4 | 399 | return tsk; |
61c4628b | 400 | |
f19b9f74 | 401 | free_ti: |
61c4628b | 402 | free_thread_info(ti); |
f19b9f74 | 403 | free_tsk: |
61c4628b SS |
404 | free_task_struct(tsk); |
405 | return NULL; | |
1da177e4 LT |
406 | } |
407 | ||
408 | #ifdef CONFIG_MMU | |
a39bc516 | 409 | static int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm) |
1da177e4 | 410 | { |
297c5eee | 411 | struct vm_area_struct *mpnt, *tmp, *prev, **pprev; |
1da177e4 LT |
412 | struct rb_node **rb_link, *rb_parent; |
413 | int retval; | |
414 | unsigned long charge; | |
1da177e4 | 415 | |
32cdba1e | 416 | uprobe_start_dup_mmap(); |
7c051267 MH |
417 | if (down_write_killable(&oldmm->mmap_sem)) { |
418 | retval = -EINTR; | |
419 | goto fail_uprobe_end; | |
420 | } | |
ec8c0446 | 421 | flush_cache_dup_mm(oldmm); |
f8ac4ec9 | 422 | uprobe_dup_mmap(oldmm, mm); |
ad339451 IM |
423 | /* |
424 | * Not linked in yet - no deadlock potential: | |
425 | */ | |
426 | down_write_nested(&mm->mmap_sem, SINGLE_DEPTH_NESTING); | |
7ee78232 | 427 | |
90f31d0e KK |
428 | /* No ordering required: file already has been exposed. */ |
429 | RCU_INIT_POINTER(mm->exe_file, get_mm_exe_file(oldmm)); | |
430 | ||
4f7d4614 | 431 | mm->total_vm = oldmm->total_vm; |
84638335 | 432 | mm->data_vm = oldmm->data_vm; |
4f7d4614 VD |
433 | mm->exec_vm = oldmm->exec_vm; |
434 | mm->stack_vm = oldmm->stack_vm; | |
435 | ||
1da177e4 LT |
436 | rb_link = &mm->mm_rb.rb_node; |
437 | rb_parent = NULL; | |
438 | pprev = &mm->mmap; | |
f8af4da3 | 439 | retval = ksm_fork(mm, oldmm); |
ba76149f AA |
440 | if (retval) |
441 | goto out; | |
442 | retval = khugepaged_fork(mm, oldmm); | |
f8af4da3 HD |
443 | if (retval) |
444 | goto out; | |
1da177e4 | 445 | |
297c5eee | 446 | prev = NULL; |
fd3e42fc | 447 | for (mpnt = oldmm->mmap; mpnt; mpnt = mpnt->vm_next) { |
1da177e4 LT |
448 | struct file *file; |
449 | ||
450 | if (mpnt->vm_flags & VM_DONTCOPY) { | |
84638335 | 451 | vm_stat_account(mm, mpnt->vm_flags, -vma_pages(mpnt)); |
1da177e4 LT |
452 | continue; |
453 | } | |
454 | charge = 0; | |
455 | if (mpnt->vm_flags & VM_ACCOUNT) { | |
b2412b7f HS |
456 | unsigned long len = vma_pages(mpnt); |
457 | ||
191c5424 | 458 | if (security_vm_enough_memory_mm(oldmm, len)) /* sic */ |
1da177e4 LT |
459 | goto fail_nomem; |
460 | charge = len; | |
461 | } | |
e94b1766 | 462 | tmp = kmem_cache_alloc(vm_area_cachep, GFP_KERNEL); |
1da177e4 LT |
463 | if (!tmp) |
464 | goto fail_nomem; | |
465 | *tmp = *mpnt; | |
5beb4930 | 466 | INIT_LIST_HEAD(&tmp->anon_vma_chain); |
ef0855d3 ON |
467 | retval = vma_dup_policy(mpnt, tmp); |
468 | if (retval) | |
1da177e4 | 469 | goto fail_nomem_policy; |
a247c3a9 | 470 | tmp->vm_mm = mm; |
5beb4930 RR |
471 | if (anon_vma_fork(tmp, mpnt)) |
472 | goto fail_nomem_anon_vma_fork; | |
de60f5f1 EM |
473 | tmp->vm_flags &= |
474 | ~(VM_LOCKED|VM_LOCKONFAULT|VM_UFFD_MISSING|VM_UFFD_WP); | |
297c5eee | 475 | tmp->vm_next = tmp->vm_prev = NULL; |
745f234b | 476 | tmp->vm_userfaultfd_ctx = NULL_VM_UFFD_CTX; |
1da177e4 LT |
477 | file = tmp->vm_file; |
478 | if (file) { | |
496ad9aa | 479 | struct inode *inode = file_inode(file); |
b88ed205 HD |
480 | struct address_space *mapping = file->f_mapping; |
481 | ||
1da177e4 LT |
482 | get_file(file); |
483 | if (tmp->vm_flags & VM_DENYWRITE) | |
484 | atomic_dec(&inode->i_writecount); | |
83cde9e8 | 485 | i_mmap_lock_write(mapping); |
b88ed205 | 486 | if (tmp->vm_flags & VM_SHARED) |
4bb5f5d9 | 487 | atomic_inc(&mapping->i_mmap_writable); |
b88ed205 HD |
488 | flush_dcache_mmap_lock(mapping); |
489 | /* insert tmp into the share list, just after mpnt */ | |
27ba0644 KS |
490 | vma_interval_tree_insert_after(tmp, mpnt, |
491 | &mapping->i_mmap); | |
b88ed205 | 492 | flush_dcache_mmap_unlock(mapping); |
83cde9e8 | 493 | i_mmap_unlock_write(mapping); |
1da177e4 LT |
494 | } |
495 | ||
a1e78772 MG |
496 | /* |
497 | * Clear hugetlb-related page reserves for children. This only | |
498 | * affects MAP_PRIVATE mappings. Faults generated by the child | |
499 | * are not guaranteed to succeed, even if read-only | |
500 | */ | |
501 | if (is_vm_hugetlb_page(tmp)) | |
502 | reset_vma_resv_huge_pages(tmp); | |
503 | ||
1da177e4 | 504 | /* |
7ee78232 | 505 | * Link in the new vma and copy the page table entries. |
1da177e4 | 506 | */ |
1da177e4 LT |
507 | *pprev = tmp; |
508 | pprev = &tmp->vm_next; | |
297c5eee LT |
509 | tmp->vm_prev = prev; |
510 | prev = tmp; | |
1da177e4 LT |
511 | |
512 | __vma_link_rb(mm, tmp, rb_link, rb_parent); | |
513 | rb_link = &tmp->vm_rb.rb_right; | |
514 | rb_parent = &tmp->vm_rb; | |
515 | ||
516 | mm->map_count++; | |
0b0db14c | 517 | retval = copy_page_range(mm, oldmm, mpnt); |
1da177e4 LT |
518 | |
519 | if (tmp->vm_ops && tmp->vm_ops->open) | |
520 | tmp->vm_ops->open(tmp); | |
521 | ||
522 | if (retval) | |
523 | goto out; | |
524 | } | |
d6dd61c8 JF |
525 | /* a new mm has just been created */ |
526 | arch_dup_mmap(oldmm, mm); | |
1da177e4 | 527 | retval = 0; |
1da177e4 | 528 | out: |
7ee78232 | 529 | up_write(&mm->mmap_sem); |
fd3e42fc | 530 | flush_tlb_mm(oldmm); |
1da177e4 | 531 | up_write(&oldmm->mmap_sem); |
7c051267 | 532 | fail_uprobe_end: |
32cdba1e | 533 | uprobe_end_dup_mmap(); |
1da177e4 | 534 | return retval; |
5beb4930 | 535 | fail_nomem_anon_vma_fork: |
ef0855d3 | 536 | mpol_put(vma_policy(tmp)); |
1da177e4 LT |
537 | fail_nomem_policy: |
538 | kmem_cache_free(vm_area_cachep, tmp); | |
539 | fail_nomem: | |
540 | retval = -ENOMEM; | |
541 | vm_unacct_memory(charge); | |
542 | goto out; | |
543 | } | |
544 | ||
fb0a685c | 545 | static inline int mm_alloc_pgd(struct mm_struct *mm) |
1da177e4 LT |
546 | { |
547 | mm->pgd = pgd_alloc(mm); | |
548 | if (unlikely(!mm->pgd)) | |
549 | return -ENOMEM; | |
550 | return 0; | |
551 | } | |
552 | ||
fb0a685c | 553 | static inline void mm_free_pgd(struct mm_struct *mm) |
1da177e4 | 554 | { |
5e541973 | 555 | pgd_free(mm, mm->pgd); |
1da177e4 LT |
556 | } |
557 | #else | |
90f31d0e KK |
558 | static int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm) |
559 | { | |
560 | down_write(&oldmm->mmap_sem); | |
561 | RCU_INIT_POINTER(mm->exe_file, get_mm_exe_file(oldmm)); | |
562 | up_write(&oldmm->mmap_sem); | |
563 | return 0; | |
564 | } | |
1da177e4 LT |
565 | #define mm_alloc_pgd(mm) (0) |
566 | #define mm_free_pgd(mm) | |
567 | #endif /* CONFIG_MMU */ | |
568 | ||
23ff4440 | 569 | __cacheline_aligned_in_smp DEFINE_SPINLOCK(mmlist_lock); |
1da177e4 | 570 | |
e94b1766 | 571 | #define allocate_mm() (kmem_cache_alloc(mm_cachep, GFP_KERNEL)) |
1da177e4 LT |
572 | #define free_mm(mm) (kmem_cache_free(mm_cachep, (mm))) |
573 | ||
4cb0e11b HK |
574 | static unsigned long default_dump_filter = MMF_DUMP_FILTER_DEFAULT; |
575 | ||
576 | static int __init coredump_filter_setup(char *s) | |
577 | { | |
578 | default_dump_filter = | |
579 | (simple_strtoul(s, NULL, 0) << MMF_DUMP_FILTER_SHIFT) & | |
580 | MMF_DUMP_FILTER_MASK; | |
581 | return 1; | |
582 | } | |
583 | ||
584 | __setup("coredump_filter=", coredump_filter_setup); | |
585 | ||
1da177e4 LT |
586 | #include <linux/init_task.h> |
587 | ||
858f0993 AD |
588 | static void mm_init_aio(struct mm_struct *mm) |
589 | { | |
590 | #ifdef CONFIG_AIO | |
591 | spin_lock_init(&mm->ioctx_lock); | |
db446a08 | 592 | mm->ioctx_table = NULL; |
858f0993 AD |
593 | #endif |
594 | } | |
595 | ||
33144e84 VD |
596 | static void mm_init_owner(struct mm_struct *mm, struct task_struct *p) |
597 | { | |
598 | #ifdef CONFIG_MEMCG | |
599 | mm->owner = p; | |
600 | #endif | |
601 | } | |
602 | ||
fb0a685c | 603 | static struct mm_struct *mm_init(struct mm_struct *mm, struct task_struct *p) |
1da177e4 | 604 | { |
41f727fd VD |
605 | mm->mmap = NULL; |
606 | mm->mm_rb = RB_ROOT; | |
607 | mm->vmacache_seqnum = 0; | |
1da177e4 LT |
608 | atomic_set(&mm->mm_users, 1); |
609 | atomic_set(&mm->mm_count, 1); | |
610 | init_rwsem(&mm->mmap_sem); | |
611 | INIT_LIST_HEAD(&mm->mmlist); | |
999d9fc1 | 612 | mm->core_state = NULL; |
e1f56c89 | 613 | atomic_long_set(&mm->nr_ptes, 0); |
2d2f5119 | 614 | mm_nr_pmds_init(mm); |
41f727fd VD |
615 | mm->map_count = 0; |
616 | mm->locked_vm = 0; | |
ce65cefa | 617 | mm->pinned_vm = 0; |
d559db08 | 618 | memset(&mm->rss_stat, 0, sizeof(mm->rss_stat)); |
1da177e4 | 619 | spin_lock_init(&mm->page_table_lock); |
41f727fd | 620 | mm_init_cpumask(mm); |
858f0993 | 621 | mm_init_aio(mm); |
cf475ad2 | 622 | mm_init_owner(mm, p); |
41f727fd | 623 | mmu_notifier_mm_init(mm); |
20841405 | 624 | clear_tlb_flush_pending(mm); |
41f727fd VD |
625 | #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS |
626 | mm->pmd_huge_pte = NULL; | |
627 | #endif | |
1da177e4 | 628 | |
a0715cc2 AT |
629 | if (current->mm) { |
630 | mm->flags = current->mm->flags & MMF_INIT_MASK; | |
631 | mm->def_flags = current->mm->def_flags & VM_INIT_DEF_MASK; | |
632 | } else { | |
633 | mm->flags = default_dump_filter; | |
1da177e4 | 634 | mm->def_flags = 0; |
a0715cc2 AT |
635 | } |
636 | ||
41f727fd VD |
637 | if (mm_alloc_pgd(mm)) |
638 | goto fail_nopgd; | |
639 | ||
640 | if (init_new_context(p, mm)) | |
641 | goto fail_nocontext; | |
78fb7466 | 642 | |
41f727fd VD |
643 | return mm; |
644 | ||
645 | fail_nocontext: | |
646 | mm_free_pgd(mm); | |
647 | fail_nopgd: | |
1da177e4 LT |
648 | free_mm(mm); |
649 | return NULL; | |
650 | } | |
651 | ||
c3f0327f KK |
652 | static void check_mm(struct mm_struct *mm) |
653 | { | |
654 | int i; | |
655 | ||
656 | for (i = 0; i < NR_MM_COUNTERS; i++) { | |
657 | long x = atomic_long_read(&mm->rss_stat.count[i]); | |
658 | ||
659 | if (unlikely(x)) | |
660 | printk(KERN_ALERT "BUG: Bad rss-counter state " | |
661 | "mm:%p idx:%d val:%ld\n", mm, i, x); | |
662 | } | |
b30fe6c7 KS |
663 | |
664 | if (atomic_long_read(&mm->nr_ptes)) | |
665 | pr_alert("BUG: non-zero nr_ptes on freeing mm: %ld\n", | |
666 | atomic_long_read(&mm->nr_ptes)); | |
667 | if (mm_nr_pmds(mm)) | |
668 | pr_alert("BUG: non-zero nr_pmds on freeing mm: %ld\n", | |
669 | mm_nr_pmds(mm)); | |
670 | ||
e009bb30 | 671 | #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS |
96dad67f | 672 | VM_BUG_ON_MM(mm->pmd_huge_pte, mm); |
c3f0327f KK |
673 | #endif |
674 | } | |
675 | ||
1da177e4 LT |
676 | /* |
677 | * Allocate and initialize an mm_struct. | |
678 | */ | |
fb0a685c | 679 | struct mm_struct *mm_alloc(void) |
1da177e4 | 680 | { |
fb0a685c | 681 | struct mm_struct *mm; |
1da177e4 LT |
682 | |
683 | mm = allocate_mm(); | |
de03c72c KM |
684 | if (!mm) |
685 | return NULL; | |
686 | ||
687 | memset(mm, 0, sizeof(*mm)); | |
6345d24d | 688 | return mm_init(mm, current); |
1da177e4 LT |
689 | } |
690 | ||
691 | /* | |
692 | * Called when the last reference to the mm | |
693 | * is dropped: either by a lazy thread or by | |
694 | * mmput. Free the page directory and the mm. | |
695 | */ | |
7ad5b3a5 | 696 | void __mmdrop(struct mm_struct *mm) |
1da177e4 LT |
697 | { |
698 | BUG_ON(mm == &init_mm); | |
699 | mm_free_pgd(mm); | |
700 | destroy_context(mm); | |
cddb8a5c | 701 | mmu_notifier_mm_destroy(mm); |
c3f0327f | 702 | check_mm(mm); |
1da177e4 LT |
703 | free_mm(mm); |
704 | } | |
6d4e4c4f | 705 | EXPORT_SYMBOL_GPL(__mmdrop); |
1da177e4 | 706 | |
ec8d7c14 MH |
707 | static inline void __mmput(struct mm_struct *mm) |
708 | { | |
709 | VM_BUG_ON(atomic_read(&mm->mm_users)); | |
710 | ||
711 | uprobe_clear_state(mm); | |
712 | exit_aio(mm); | |
713 | ksm_exit(mm); | |
714 | khugepaged_exit(mm); /* must run before exit_mmap */ | |
715 | exit_mmap(mm); | |
716 | set_mm_exe_file(mm, NULL); | |
717 | if (!list_empty(&mm->mmlist)) { | |
718 | spin_lock(&mmlist_lock); | |
719 | list_del(&mm->mmlist); | |
720 | spin_unlock(&mmlist_lock); | |
721 | } | |
722 | if (mm->binfmt) | |
723 | module_put(mm->binfmt->module); | |
724 | mmdrop(mm); | |
725 | } | |
726 | ||
1da177e4 LT |
727 | /* |
728 | * Decrement the use count and release all resources for an mm. | |
729 | */ | |
730 | void mmput(struct mm_struct *mm) | |
731 | { | |
0ae26f1b AM |
732 | might_sleep(); |
733 | ||
ec8d7c14 MH |
734 | if (atomic_dec_and_test(&mm->mm_users)) |
735 | __mmput(mm); | |
736 | } | |
737 | EXPORT_SYMBOL_GPL(mmput); | |
738 | ||
7ef949d7 | 739 | #ifdef CONFIG_MMU |
ec8d7c14 MH |
740 | static void mmput_async_fn(struct work_struct *work) |
741 | { | |
742 | struct mm_struct *mm = container_of(work, struct mm_struct, async_put_work); | |
743 | __mmput(mm); | |
744 | } | |
745 | ||
746 | void mmput_async(struct mm_struct *mm) | |
747 | { | |
1da177e4 | 748 | if (atomic_dec_and_test(&mm->mm_users)) { |
ec8d7c14 MH |
749 | INIT_WORK(&mm->async_put_work, mmput_async_fn); |
750 | schedule_work(&mm->async_put_work); | |
1da177e4 LT |
751 | } |
752 | } | |
7ef949d7 | 753 | #endif |
1da177e4 | 754 | |
90f31d0e KK |
755 | /** |
756 | * set_mm_exe_file - change a reference to the mm's executable file | |
757 | * | |
758 | * This changes mm's executable file (shown as symlink /proc/[pid]/exe). | |
759 | * | |
6e399cd1 DB |
760 | * Main users are mmput() and sys_execve(). Callers prevent concurrent |
761 | * invocations: in mmput() nobody alive left, in execve task is single | |
762 | * threaded. sys_prctl(PR_SET_MM_MAP/EXE_FILE) also needs to set the | |
763 | * mm->exe_file, but does so without using set_mm_exe_file() in order | |
764 | * to do avoid the need for any locks. | |
90f31d0e | 765 | */ |
38646013 JS |
766 | void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file) |
767 | { | |
6e399cd1 DB |
768 | struct file *old_exe_file; |
769 | ||
770 | /* | |
771 | * It is safe to dereference the exe_file without RCU as | |
772 | * this function is only called if nobody else can access | |
773 | * this mm -- see comment above for justification. | |
774 | */ | |
775 | old_exe_file = rcu_dereference_raw(mm->exe_file); | |
90f31d0e | 776 | |
38646013 JS |
777 | if (new_exe_file) |
778 | get_file(new_exe_file); | |
90f31d0e KK |
779 | rcu_assign_pointer(mm->exe_file, new_exe_file); |
780 | if (old_exe_file) | |
781 | fput(old_exe_file); | |
38646013 JS |
782 | } |
783 | ||
90f31d0e KK |
784 | /** |
785 | * get_mm_exe_file - acquire a reference to the mm's executable file | |
786 | * | |
787 | * Returns %NULL if mm has no associated executable file. | |
788 | * User must release file via fput(). | |
789 | */ | |
38646013 JS |
790 | struct file *get_mm_exe_file(struct mm_struct *mm) |
791 | { | |
792 | struct file *exe_file; | |
793 | ||
90f31d0e KK |
794 | rcu_read_lock(); |
795 | exe_file = rcu_dereference(mm->exe_file); | |
796 | if (exe_file && !get_file_rcu(exe_file)) | |
797 | exe_file = NULL; | |
798 | rcu_read_unlock(); | |
38646013 JS |
799 | return exe_file; |
800 | } | |
11163348 | 801 | EXPORT_SYMBOL(get_mm_exe_file); |
38646013 | 802 | |
1da177e4 LT |
803 | /** |
804 | * get_task_mm - acquire a reference to the task's mm | |
805 | * | |
246bb0b1 | 806 | * Returns %NULL if the task has no mm. Checks PF_KTHREAD (meaning |
1da177e4 LT |
807 | * this kernel workthread has transiently adopted a user mm with use_mm, |
808 | * to do its AIO) is not set and if so returns a reference to it, after | |
809 | * bumping up the use count. User must release the mm via mmput() | |
810 | * after use. Typically used by /proc and ptrace. | |
811 | */ | |
812 | struct mm_struct *get_task_mm(struct task_struct *task) | |
813 | { | |
814 | struct mm_struct *mm; | |
815 | ||
816 | task_lock(task); | |
817 | mm = task->mm; | |
818 | if (mm) { | |
246bb0b1 | 819 | if (task->flags & PF_KTHREAD) |
1da177e4 LT |
820 | mm = NULL; |
821 | else | |
822 | atomic_inc(&mm->mm_users); | |
823 | } | |
824 | task_unlock(task); | |
825 | return mm; | |
826 | } | |
827 | EXPORT_SYMBOL_GPL(get_task_mm); | |
828 | ||
8cdb878d CY |
829 | struct mm_struct *mm_access(struct task_struct *task, unsigned int mode) |
830 | { | |
831 | struct mm_struct *mm; | |
832 | int err; | |
833 | ||
834 | err = mutex_lock_killable(&task->signal->cred_guard_mutex); | |
835 | if (err) | |
836 | return ERR_PTR(err); | |
837 | ||
838 | mm = get_task_mm(task); | |
839 | if (mm && mm != current->mm && | |
840 | !ptrace_may_access(task, mode)) { | |
841 | mmput(mm); | |
842 | mm = ERR_PTR(-EACCES); | |
843 | } | |
844 | mutex_unlock(&task->signal->cred_guard_mutex); | |
845 | ||
846 | return mm; | |
847 | } | |
848 | ||
57b59c4a | 849 | static void complete_vfork_done(struct task_struct *tsk) |
c415c3b4 | 850 | { |
d68b46fe | 851 | struct completion *vfork; |
c415c3b4 | 852 | |
d68b46fe ON |
853 | task_lock(tsk); |
854 | vfork = tsk->vfork_done; | |
855 | if (likely(vfork)) { | |
856 | tsk->vfork_done = NULL; | |
857 | complete(vfork); | |
858 | } | |
859 | task_unlock(tsk); | |
860 | } | |
861 | ||
862 | static int wait_for_vfork_done(struct task_struct *child, | |
863 | struct completion *vfork) | |
864 | { | |
865 | int killed; | |
866 | ||
867 | freezer_do_not_count(); | |
868 | killed = wait_for_completion_killable(vfork); | |
869 | freezer_count(); | |
870 | ||
871 | if (killed) { | |
872 | task_lock(child); | |
873 | child->vfork_done = NULL; | |
874 | task_unlock(child); | |
875 | } | |
876 | ||
877 | put_task_struct(child); | |
878 | return killed; | |
c415c3b4 ON |
879 | } |
880 | ||
1da177e4 LT |
881 | /* Please note the differences between mmput and mm_release. |
882 | * mmput is called whenever we stop holding onto a mm_struct, | |
883 | * error success whatever. | |
884 | * | |
885 | * mm_release is called after a mm_struct has been removed | |
886 | * from the current process. | |
887 | * | |
888 | * This difference is important for error handling, when we | |
889 | * only half set up a mm_struct for a new process and need to restore | |
890 | * the old one. Because we mmput the new mm_struct before | |
891 | * restoring the old one. . . | |
892 | * Eric Biederman 10 January 1998 | |
893 | */ | |
894 | void mm_release(struct task_struct *tsk, struct mm_struct *mm) | |
895 | { | |
8141c7f3 LT |
896 | /* Get rid of any futexes when releasing the mm */ |
897 | #ifdef CONFIG_FUTEX | |
fc6b177d | 898 | if (unlikely(tsk->robust_list)) { |
8141c7f3 | 899 | exit_robust_list(tsk); |
fc6b177d PZ |
900 | tsk->robust_list = NULL; |
901 | } | |
8141c7f3 | 902 | #ifdef CONFIG_COMPAT |
fc6b177d | 903 | if (unlikely(tsk->compat_robust_list)) { |
8141c7f3 | 904 | compat_exit_robust_list(tsk); |
fc6b177d PZ |
905 | tsk->compat_robust_list = NULL; |
906 | } | |
8141c7f3 | 907 | #endif |
322a2c10 TG |
908 | if (unlikely(!list_empty(&tsk->pi_state_list))) |
909 | exit_pi_state_list(tsk); | |
8141c7f3 LT |
910 | #endif |
911 | ||
0326f5a9 SD |
912 | uprobe_free_utask(tsk); |
913 | ||
1da177e4 LT |
914 | /* Get rid of any cached register state */ |
915 | deactivate_mm(tsk, mm); | |
916 | ||
fec1d011 RM |
917 | /* |
918 | * If we're exiting normally, clear a user-space tid field if | |
919 | * requested. We leave this alone when dying by signal, to leave | |
920 | * the value intact in a core dump, and to save the unnecessary | |
d68b46fe ON |
921 | * trouble, say, a killed vfork parent shouldn't touch this mm. |
922 | * Userland only wants this done for a sys_exit. | |
fec1d011 | 923 | */ |
9c8a8228 ED |
924 | if (tsk->clear_child_tid) { |
925 | if (!(tsk->flags & PF_SIGNALED) && | |
926 | atomic_read(&mm->mm_users) > 1) { | |
927 | /* | |
928 | * We don't check the error code - if userspace has | |
929 | * not set up a proper pointer then tough luck. | |
930 | */ | |
931 | put_user(0, tsk->clear_child_tid); | |
932 | sys_futex(tsk->clear_child_tid, FUTEX_WAKE, | |
933 | 1, NULL, NULL, 0); | |
934 | } | |
1da177e4 | 935 | tsk->clear_child_tid = NULL; |
1da177e4 | 936 | } |
f7505d64 KK |
937 | |
938 | /* | |
939 | * All done, finally we can wake up parent and return this mm to him. | |
940 | * Also kthread_stop() uses this completion for synchronization. | |
941 | */ | |
942 | if (tsk->vfork_done) | |
943 | complete_vfork_done(tsk); | |
1da177e4 LT |
944 | } |
945 | ||
a0a7ec30 JD |
946 | /* |
947 | * Allocate a new mm structure and copy contents from the | |
948 | * mm structure of the passed in task structure. | |
949 | */ | |
ff252c1f | 950 | static struct mm_struct *dup_mm(struct task_struct *tsk) |
a0a7ec30 JD |
951 | { |
952 | struct mm_struct *mm, *oldmm = current->mm; | |
953 | int err; | |
954 | ||
a0a7ec30 JD |
955 | mm = allocate_mm(); |
956 | if (!mm) | |
957 | goto fail_nomem; | |
958 | ||
959 | memcpy(mm, oldmm, sizeof(*mm)); | |
960 | ||
78fb7466 | 961 | if (!mm_init(mm, tsk)) |
a0a7ec30 JD |
962 | goto fail_nomem; |
963 | ||
a0a7ec30 JD |
964 | err = dup_mmap(mm, oldmm); |
965 | if (err) | |
966 | goto free_pt; | |
967 | ||
968 | mm->hiwater_rss = get_mm_rss(mm); | |
969 | mm->hiwater_vm = mm->total_vm; | |
970 | ||
801460d0 HS |
971 | if (mm->binfmt && !try_module_get(mm->binfmt->module)) |
972 | goto free_pt; | |
973 | ||
a0a7ec30 JD |
974 | return mm; |
975 | ||
976 | free_pt: | |
801460d0 HS |
977 | /* don't put binfmt in mmput, we haven't got module yet */ |
978 | mm->binfmt = NULL; | |
a0a7ec30 JD |
979 | mmput(mm); |
980 | ||
981 | fail_nomem: | |
982 | return NULL; | |
a0a7ec30 JD |
983 | } |
984 | ||
fb0a685c | 985 | static int copy_mm(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 | 986 | { |
fb0a685c | 987 | struct mm_struct *mm, *oldmm; |
1da177e4 LT |
988 | int retval; |
989 | ||
990 | tsk->min_flt = tsk->maj_flt = 0; | |
991 | tsk->nvcsw = tsk->nivcsw = 0; | |
17406b82 MSB |
992 | #ifdef CONFIG_DETECT_HUNG_TASK |
993 | tsk->last_switch_count = tsk->nvcsw + tsk->nivcsw; | |
994 | #endif | |
1da177e4 LT |
995 | |
996 | tsk->mm = NULL; | |
997 | tsk->active_mm = NULL; | |
998 | ||
999 | /* | |
1000 | * Are we cloning a kernel thread? | |
1001 | * | |
1002 | * We need to steal a active VM for that.. | |
1003 | */ | |
1004 | oldmm = current->mm; | |
1005 | if (!oldmm) | |
1006 | return 0; | |
1007 | ||
615d6e87 DB |
1008 | /* initialize the new vmacache entries */ |
1009 | vmacache_flush(tsk); | |
1010 | ||
1da177e4 LT |
1011 | if (clone_flags & CLONE_VM) { |
1012 | atomic_inc(&oldmm->mm_users); | |
1013 | mm = oldmm; | |
1da177e4 LT |
1014 | goto good_mm; |
1015 | } | |
1016 | ||
1017 | retval = -ENOMEM; | |
a0a7ec30 | 1018 | mm = dup_mm(tsk); |
1da177e4 LT |
1019 | if (!mm) |
1020 | goto fail_nomem; | |
1021 | ||
1da177e4 LT |
1022 | good_mm: |
1023 | tsk->mm = mm; | |
1024 | tsk->active_mm = mm; | |
1025 | return 0; | |
1026 | ||
1da177e4 LT |
1027 | fail_nomem: |
1028 | return retval; | |
1da177e4 LT |
1029 | } |
1030 | ||
a39bc516 | 1031 | static int copy_fs(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 | 1032 | { |
498052bb | 1033 | struct fs_struct *fs = current->fs; |
1da177e4 | 1034 | if (clone_flags & CLONE_FS) { |
498052bb | 1035 | /* tsk->fs is already what we want */ |
2a4419b5 | 1036 | spin_lock(&fs->lock); |
498052bb | 1037 | if (fs->in_exec) { |
2a4419b5 | 1038 | spin_unlock(&fs->lock); |
498052bb AV |
1039 | return -EAGAIN; |
1040 | } | |
1041 | fs->users++; | |
2a4419b5 | 1042 | spin_unlock(&fs->lock); |
1da177e4 LT |
1043 | return 0; |
1044 | } | |
498052bb | 1045 | tsk->fs = copy_fs_struct(fs); |
1da177e4 LT |
1046 | if (!tsk->fs) |
1047 | return -ENOMEM; | |
1048 | return 0; | |
1049 | } | |
1050 | ||
fb0a685c | 1051 | static int copy_files(unsigned long clone_flags, struct task_struct *tsk) |
a016f338 JD |
1052 | { |
1053 | struct files_struct *oldf, *newf; | |
1054 | int error = 0; | |
1055 | ||
1056 | /* | |
1057 | * A background process may not have any files ... | |
1058 | */ | |
1059 | oldf = current->files; | |
1060 | if (!oldf) | |
1061 | goto out; | |
1062 | ||
1063 | if (clone_flags & CLONE_FILES) { | |
1064 | atomic_inc(&oldf->count); | |
1065 | goto out; | |
1066 | } | |
1067 | ||
a016f338 JD |
1068 | newf = dup_fd(oldf, &error); |
1069 | if (!newf) | |
1070 | goto out; | |
1071 | ||
1072 | tsk->files = newf; | |
1073 | error = 0; | |
1074 | out: | |
1075 | return error; | |
1076 | } | |
1077 | ||
fadad878 | 1078 | static int copy_io(unsigned long clone_flags, struct task_struct *tsk) |
fd0928df JA |
1079 | { |
1080 | #ifdef CONFIG_BLOCK | |
1081 | struct io_context *ioc = current->io_context; | |
6e736be7 | 1082 | struct io_context *new_ioc; |
fd0928df JA |
1083 | |
1084 | if (!ioc) | |
1085 | return 0; | |
fadad878 JA |
1086 | /* |
1087 | * Share io context with parent, if CLONE_IO is set | |
1088 | */ | |
1089 | if (clone_flags & CLONE_IO) { | |
3d48749d TH |
1090 | ioc_task_link(ioc); |
1091 | tsk->io_context = ioc; | |
fadad878 | 1092 | } else if (ioprio_valid(ioc->ioprio)) { |
6e736be7 TH |
1093 | new_ioc = get_task_io_context(tsk, GFP_KERNEL, NUMA_NO_NODE); |
1094 | if (unlikely(!new_ioc)) | |
fd0928df JA |
1095 | return -ENOMEM; |
1096 | ||
6e736be7 | 1097 | new_ioc->ioprio = ioc->ioprio; |
11a3122f | 1098 | put_io_context(new_ioc); |
fd0928df JA |
1099 | } |
1100 | #endif | |
1101 | return 0; | |
1102 | } | |
1103 | ||
a39bc516 | 1104 | static int copy_sighand(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 LT |
1105 | { |
1106 | struct sighand_struct *sig; | |
1107 | ||
60348802 | 1108 | if (clone_flags & CLONE_SIGHAND) { |
1da177e4 LT |
1109 | atomic_inc(¤t->sighand->count); |
1110 | return 0; | |
1111 | } | |
1112 | sig = kmem_cache_alloc(sighand_cachep, GFP_KERNEL); | |
e56d0903 | 1113 | rcu_assign_pointer(tsk->sighand, sig); |
1da177e4 LT |
1114 | if (!sig) |
1115 | return -ENOMEM; | |
9d7fb042 | 1116 | |
1da177e4 LT |
1117 | atomic_set(&sig->count, 1); |
1118 | memcpy(sig->action, current->sighand->action, sizeof(sig->action)); | |
1119 | return 0; | |
1120 | } | |
1121 | ||
a7e5328a | 1122 | void __cleanup_sighand(struct sighand_struct *sighand) |
c81addc9 | 1123 | { |
d80e731e ON |
1124 | if (atomic_dec_and_test(&sighand->count)) { |
1125 | signalfd_cleanup(sighand); | |
392809b2 ON |
1126 | /* |
1127 | * sighand_cachep is SLAB_DESTROY_BY_RCU so we can free it | |
1128 | * without an RCU grace period, see __lock_task_sighand(). | |
1129 | */ | |
c81addc9 | 1130 | kmem_cache_free(sighand_cachep, sighand); |
d80e731e | 1131 | } |
c81addc9 ON |
1132 | } |
1133 | ||
f06febc9 FM |
1134 | /* |
1135 | * Initialize POSIX timer handling for a thread group. | |
1136 | */ | |
1137 | static void posix_cpu_timers_init_group(struct signal_struct *sig) | |
1138 | { | |
78d7d407 JS |
1139 | unsigned long cpu_limit; |
1140 | ||
316c1608 | 1141 | cpu_limit = READ_ONCE(sig->rlim[RLIMIT_CPU].rlim_cur); |
78d7d407 JS |
1142 | if (cpu_limit != RLIM_INFINITY) { |
1143 | sig->cputime_expires.prof_exp = secs_to_cputime(cpu_limit); | |
d5c373eb | 1144 | sig->cputimer.running = true; |
6279a751 ON |
1145 | } |
1146 | ||
f06febc9 FM |
1147 | /* The timer lists. */ |
1148 | INIT_LIST_HEAD(&sig->cpu_timers[0]); | |
1149 | INIT_LIST_HEAD(&sig->cpu_timers[1]); | |
1150 | INIT_LIST_HEAD(&sig->cpu_timers[2]); | |
1151 | } | |
1152 | ||
a39bc516 | 1153 | static int copy_signal(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 LT |
1154 | { |
1155 | struct signal_struct *sig; | |
1da177e4 | 1156 | |
4ab6c083 | 1157 | if (clone_flags & CLONE_THREAD) |
490dea45 | 1158 | return 0; |
490dea45 | 1159 | |
a56704ef | 1160 | sig = kmem_cache_zalloc(signal_cachep, GFP_KERNEL); |
1da177e4 LT |
1161 | tsk->signal = sig; |
1162 | if (!sig) | |
1163 | return -ENOMEM; | |
1164 | ||
b3ac022c | 1165 | sig->nr_threads = 1; |
1da177e4 | 1166 | atomic_set(&sig->live, 1); |
b3ac022c | 1167 | atomic_set(&sig->sigcnt, 1); |
0c740d0a ON |
1168 | |
1169 | /* list_add(thread_node, thread_head) without INIT_LIST_HEAD() */ | |
1170 | sig->thread_head = (struct list_head)LIST_HEAD_INIT(tsk->thread_node); | |
1171 | tsk->thread_node = (struct list_head)LIST_HEAD_INIT(sig->thread_head); | |
1172 | ||
1da177e4 | 1173 | init_waitqueue_head(&sig->wait_chldexit); |
db51aecc | 1174 | sig->curr_target = tsk; |
1da177e4 LT |
1175 | init_sigpending(&sig->shared_pending); |
1176 | INIT_LIST_HEAD(&sig->posix_timers); | |
e78c3496 | 1177 | seqlock_init(&sig->stats_lock); |
9d7fb042 | 1178 | prev_cputime_init(&sig->prev_cputime); |
1da177e4 | 1179 | |
c9cb2e3d | 1180 | hrtimer_init(&sig->real_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL); |
1da177e4 | 1181 | sig->real_timer.function = it_real_fn; |
1da177e4 | 1182 | |
1da177e4 LT |
1183 | task_lock(current->group_leader); |
1184 | memcpy(sig->rlim, current->signal->rlim, sizeof sig->rlim); | |
1185 | task_unlock(current->group_leader); | |
1186 | ||
6279a751 ON |
1187 | posix_cpu_timers_init_group(sig); |
1188 | ||
522ed776 | 1189 | tty_audit_fork(sig); |
5091faa4 | 1190 | sched_autogroup_fork(sig); |
522ed776 | 1191 | |
a63d83f4 | 1192 | sig->oom_score_adj = current->signal->oom_score_adj; |
dabb16f6 | 1193 | sig->oom_score_adj_min = current->signal->oom_score_adj_min; |
28b83c51 | 1194 | |
ebec18a6 LP |
1195 | sig->has_child_subreaper = current->signal->has_child_subreaper || |
1196 | current->signal->is_child_subreaper; | |
1197 | ||
9b1bf12d KM |
1198 | mutex_init(&sig->cred_guard_mutex); |
1199 | ||
1da177e4 LT |
1200 | return 0; |
1201 | } | |
1202 | ||
dbd95212 KC |
1203 | static void copy_seccomp(struct task_struct *p) |
1204 | { | |
1205 | #ifdef CONFIG_SECCOMP | |
1206 | /* | |
1207 | * Must be called with sighand->lock held, which is common to | |
1208 | * all threads in the group. Holding cred_guard_mutex is not | |
1209 | * needed because this new task is not yet running and cannot | |
1210 | * be racing exec. | |
1211 | */ | |
69f6a34b | 1212 | assert_spin_locked(¤t->sighand->siglock); |
dbd95212 KC |
1213 | |
1214 | /* Ref-count the new filter user, and assign it. */ | |
1215 | get_seccomp_filter(current); | |
1216 | p->seccomp = current->seccomp; | |
1217 | ||
1218 | /* | |
1219 | * Explicitly enable no_new_privs here in case it got set | |
1220 | * between the task_struct being duplicated and holding the | |
1221 | * sighand lock. The seccomp state and nnp must be in sync. | |
1222 | */ | |
1223 | if (task_no_new_privs(current)) | |
1224 | task_set_no_new_privs(p); | |
1225 | ||
1226 | /* | |
1227 | * If the parent gained a seccomp mode after copying thread | |
1228 | * flags and between before we held the sighand lock, we have | |
1229 | * to manually enable the seccomp thread flag here. | |
1230 | */ | |
1231 | if (p->seccomp.mode != SECCOMP_MODE_DISABLED) | |
1232 | set_tsk_thread_flag(p, TIF_SECCOMP); | |
1233 | #endif | |
1234 | } | |
1235 | ||
17da2bd9 | 1236 | SYSCALL_DEFINE1(set_tid_address, int __user *, tidptr) |
1da177e4 LT |
1237 | { |
1238 | current->clear_child_tid = tidptr; | |
1239 | ||
b488893a | 1240 | return task_pid_vnr(current); |
1da177e4 LT |
1241 | } |
1242 | ||
a39bc516 | 1243 | static void rt_mutex_init_task(struct task_struct *p) |
23f78d4a | 1244 | { |
1d615482 | 1245 | raw_spin_lock_init(&p->pi_lock); |
e29e175b | 1246 | #ifdef CONFIG_RT_MUTEXES |
fb00aca4 PZ |
1247 | p->pi_waiters = RB_ROOT; |
1248 | p->pi_waiters_leftmost = NULL; | |
23f78d4a | 1249 | p->pi_blocked_on = NULL; |
23f78d4a IM |
1250 | #endif |
1251 | } | |
1252 | ||
f06febc9 FM |
1253 | /* |
1254 | * Initialize POSIX timer handling for a single task. | |
1255 | */ | |
1256 | static void posix_cpu_timers_init(struct task_struct *tsk) | |
1257 | { | |
64861634 MS |
1258 | tsk->cputime_expires.prof_exp = 0; |
1259 | tsk->cputime_expires.virt_exp = 0; | |
f06febc9 FM |
1260 | tsk->cputime_expires.sched_exp = 0; |
1261 | INIT_LIST_HEAD(&tsk->cpu_timers[0]); | |
1262 | INIT_LIST_HEAD(&tsk->cpu_timers[1]); | |
1263 | INIT_LIST_HEAD(&tsk->cpu_timers[2]); | |
1264 | } | |
1265 | ||
81907739 ON |
1266 | static inline void |
1267 | init_task_pid(struct task_struct *task, enum pid_type type, struct pid *pid) | |
1268 | { | |
1269 | task->pids[type].pid = pid; | |
1270 | } | |
1271 | ||
1da177e4 LT |
1272 | /* |
1273 | * This creates a new process as a copy of the old one, | |
1274 | * but does not actually start it yet. | |
1275 | * | |
1276 | * It copies the registers, and all the appropriate | |
1277 | * parts of the process environment (as per the clone | |
1278 | * flags). The actual kick-off is left to the caller. | |
1279 | */ | |
36c8b586 IM |
1280 | static struct task_struct *copy_process(unsigned long clone_flags, |
1281 | unsigned long stack_start, | |
36c8b586 | 1282 | unsigned long stack_size, |
36c8b586 | 1283 | int __user *child_tidptr, |
09a05394 | 1284 | struct pid *pid, |
3033f14a | 1285 | int trace, |
725fc629 AK |
1286 | unsigned long tls, |
1287 | int node) | |
1da177e4 LT |
1288 | { |
1289 | int retval; | |
a24efe62 | 1290 | struct task_struct *p; |
1da177e4 LT |
1291 | |
1292 | if ((clone_flags & (CLONE_NEWNS|CLONE_FS)) == (CLONE_NEWNS|CLONE_FS)) | |
1293 | return ERR_PTR(-EINVAL); | |
1294 | ||
e66eded8 EB |
1295 | if ((clone_flags & (CLONE_NEWUSER|CLONE_FS)) == (CLONE_NEWUSER|CLONE_FS)) |
1296 | return ERR_PTR(-EINVAL); | |
1297 | ||
1da177e4 LT |
1298 | /* |
1299 | * Thread groups must share signals as well, and detached threads | |
1300 | * can only be started up within the thread group. | |
1301 | */ | |
1302 | if ((clone_flags & CLONE_THREAD) && !(clone_flags & CLONE_SIGHAND)) | |
1303 | return ERR_PTR(-EINVAL); | |
1304 | ||
1305 | /* | |
1306 | * Shared signal handlers imply shared VM. By way of the above, | |
1307 | * thread groups also imply shared VM. Blocking this case allows | |
1308 | * for various simplifications in other code. | |
1309 | */ | |
1310 | if ((clone_flags & CLONE_SIGHAND) && !(clone_flags & CLONE_VM)) | |
1311 | return ERR_PTR(-EINVAL); | |
1312 | ||
123be07b SB |
1313 | /* |
1314 | * Siblings of global init remain as zombies on exit since they are | |
1315 | * not reaped by their parent (swapper). To solve this and to avoid | |
1316 | * multi-rooted process trees, prevent global and container-inits | |
1317 | * from creating siblings. | |
1318 | */ | |
1319 | if ((clone_flags & CLONE_PARENT) && | |
1320 | current->signal->flags & SIGNAL_UNKILLABLE) | |
1321 | return ERR_PTR(-EINVAL); | |
1322 | ||
8382fcac | 1323 | /* |
40a0d32d | 1324 | * If the new process will be in a different pid or user namespace |
faf00da5 | 1325 | * do not allow it to share a thread group with the forking task. |
8382fcac | 1326 | */ |
faf00da5 | 1327 | if (clone_flags & CLONE_THREAD) { |
40a0d32d ON |
1328 | if ((clone_flags & (CLONE_NEWUSER | CLONE_NEWPID)) || |
1329 | (task_active_pid_ns(current) != | |
1330 | current->nsproxy->pid_ns_for_children)) | |
1331 | return ERR_PTR(-EINVAL); | |
1332 | } | |
8382fcac | 1333 | |
1da177e4 LT |
1334 | retval = security_task_create(clone_flags); |
1335 | if (retval) | |
1336 | goto fork_out; | |
1337 | ||
1338 | retval = -ENOMEM; | |
725fc629 | 1339 | p = dup_task_struct(current, node); |
1da177e4 LT |
1340 | if (!p) |
1341 | goto fork_out; | |
1342 | ||
f7e8b616 SR |
1343 | ftrace_graph_init_task(p); |
1344 | ||
bea493a0 PZ |
1345 | rt_mutex_init_task(p); |
1346 | ||
d12c1a37 | 1347 | #ifdef CONFIG_PROVE_LOCKING |
de30a2b3 IM |
1348 | DEBUG_LOCKS_WARN_ON(!p->hardirqs_enabled); |
1349 | DEBUG_LOCKS_WARN_ON(!p->softirqs_enabled); | |
1350 | #endif | |
1da177e4 | 1351 | retval = -EAGAIN; |
3b11a1de | 1352 | if (atomic_read(&p->real_cred->user->processes) >= |
78d7d407 | 1353 | task_rlimit(p, RLIMIT_NPROC)) { |
b57922b6 EP |
1354 | if (p->real_cred->user != INIT_USER && |
1355 | !capable(CAP_SYS_RESOURCE) && !capable(CAP_SYS_ADMIN)) | |
1da177e4 LT |
1356 | goto bad_fork_free; |
1357 | } | |
72fa5997 | 1358 | current->flags &= ~PF_NPROC_EXCEEDED; |
1da177e4 | 1359 | |
f1752eec DH |
1360 | retval = copy_creds(p, clone_flags); |
1361 | if (retval < 0) | |
1362 | goto bad_fork_free; | |
1da177e4 LT |
1363 | |
1364 | /* | |
1365 | * If multiple threads are within copy_process(), then this check | |
1366 | * triggers too late. This doesn't hurt, the check is only there | |
1367 | * to stop root fork bombs. | |
1368 | */ | |
04ec93fe | 1369 | retval = -EAGAIN; |
1da177e4 LT |
1370 | if (nr_threads >= max_threads) |
1371 | goto bad_fork_cleanup_count; | |
1372 | ||
ca74e92b | 1373 | delayacct_tsk_init(p); /* Must remain after dup_task_struct() */ |
514ddb44 DR |
1374 | p->flags &= ~(PF_SUPERPRIV | PF_WQ_WORKER); |
1375 | p->flags |= PF_FORKNOEXEC; | |
1da177e4 LT |
1376 | INIT_LIST_HEAD(&p->children); |
1377 | INIT_LIST_HEAD(&p->sibling); | |
f41d911f | 1378 | rcu_copy_process(p); |
1da177e4 LT |
1379 | p->vfork_done = NULL; |
1380 | spin_lock_init(&p->alloc_lock); | |
1da177e4 | 1381 | |
1da177e4 LT |
1382 | init_sigpending(&p->pending); |
1383 | ||
64861634 MS |
1384 | p->utime = p->stime = p->gtime = 0; |
1385 | p->utimescaled = p->stimescaled = 0; | |
9d7fb042 PZ |
1386 | prev_cputime_init(&p->prev_cputime); |
1387 | ||
6a61671b | 1388 | #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN |
b7ce2277 | 1389 | seqcount_init(&p->vtime_seqcount); |
6a61671b | 1390 | p->vtime_snap = 0; |
7098c1ea | 1391 | p->vtime_snap_whence = VTIME_INACTIVE; |
6a61671b FW |
1392 | #endif |
1393 | ||
a3a2e76c KH |
1394 | #if defined(SPLIT_RSS_COUNTING) |
1395 | memset(&p->rss_stat, 0, sizeof(p->rss_stat)); | |
1396 | #endif | |
172ba844 | 1397 | |
6976675d AV |
1398 | p->default_timer_slack_ns = current->timer_slack_ns; |
1399 | ||
5995477a | 1400 | task_io_accounting_init(&p->ioac); |
1da177e4 LT |
1401 | acct_clear_integrals(p); |
1402 | ||
f06febc9 | 1403 | posix_cpu_timers_init(p); |
1da177e4 | 1404 | |
ccbf62d8 | 1405 | p->start_time = ktime_get_ns(); |
57e0be04 | 1406 | p->real_start_time = ktime_get_boot_ns(); |
1da177e4 | 1407 | p->io_context = NULL; |
1da177e4 | 1408 | p->audit_context = NULL; |
c9e75f04 | 1409 | threadgroup_change_begin(current); |
b4f48b63 | 1410 | cgroup_fork(p); |
1da177e4 | 1411 | #ifdef CONFIG_NUMA |
846a16bf | 1412 | p->mempolicy = mpol_dup(p->mempolicy); |
fb0a685c DRO |
1413 | if (IS_ERR(p->mempolicy)) { |
1414 | retval = PTR_ERR(p->mempolicy); | |
1415 | p->mempolicy = NULL; | |
e8604cb4 | 1416 | goto bad_fork_cleanup_threadgroup_lock; |
fb0a685c | 1417 | } |
1da177e4 | 1418 | #endif |
778d3b0f MH |
1419 | #ifdef CONFIG_CPUSETS |
1420 | p->cpuset_mem_spread_rotor = NUMA_NO_NODE; | |
1421 | p->cpuset_slab_spread_rotor = NUMA_NO_NODE; | |
cc9a6c87 | 1422 | seqcount_init(&p->mems_allowed_seq); |
778d3b0f | 1423 | #endif |
de30a2b3 IM |
1424 | #ifdef CONFIG_TRACE_IRQFLAGS |
1425 | p->irq_events = 0; | |
1426 | p->hardirqs_enabled = 0; | |
1427 | p->hardirq_enable_ip = 0; | |
1428 | p->hardirq_enable_event = 0; | |
1429 | p->hardirq_disable_ip = _THIS_IP_; | |
1430 | p->hardirq_disable_event = 0; | |
1431 | p->softirqs_enabled = 1; | |
1432 | p->softirq_enable_ip = _THIS_IP_; | |
1433 | p->softirq_enable_event = 0; | |
1434 | p->softirq_disable_ip = 0; | |
1435 | p->softirq_disable_event = 0; | |
1436 | p->hardirq_context = 0; | |
1437 | p->softirq_context = 0; | |
1438 | #endif | |
8bcbde54 DH |
1439 | |
1440 | p->pagefault_disabled = 0; | |
1441 | ||
fbb9ce95 IM |
1442 | #ifdef CONFIG_LOCKDEP |
1443 | p->lockdep_depth = 0; /* no locks held yet */ | |
1444 | p->curr_chain_key = 0; | |
1445 | p->lockdep_recursion = 0; | |
1446 | #endif | |
1da177e4 | 1447 | |
408894ee IM |
1448 | #ifdef CONFIG_DEBUG_MUTEXES |
1449 | p->blocked_on = NULL; /* not blocked yet */ | |
1450 | #endif | |
cafe5635 KO |
1451 | #ifdef CONFIG_BCACHE |
1452 | p->sequential_io = 0; | |
1453 | p->sequential_io_avg = 0; | |
1454 | #endif | |
0f481406 | 1455 | |
3c90e6e9 | 1456 | /* Perform scheduler related setup. Assign this task to a CPU. */ |
aab03e05 DF |
1457 | retval = sched_fork(clone_flags, p); |
1458 | if (retval) | |
1459 | goto bad_fork_cleanup_policy; | |
6ab423e0 | 1460 | |
cdd6c482 | 1461 | retval = perf_event_init_task(p); |
6ab423e0 PZ |
1462 | if (retval) |
1463 | goto bad_fork_cleanup_policy; | |
fb0a685c DRO |
1464 | retval = audit_alloc(p); |
1465 | if (retval) | |
6c72e350 | 1466 | goto bad_fork_cleanup_perf; |
1da177e4 | 1467 | /* copy all the process information */ |
ab602f79 | 1468 | shm_init_task(p); |
fb0a685c DRO |
1469 | retval = copy_semundo(clone_flags, p); |
1470 | if (retval) | |
1da177e4 | 1471 | goto bad_fork_cleanup_audit; |
fb0a685c DRO |
1472 | retval = copy_files(clone_flags, p); |
1473 | if (retval) | |
1da177e4 | 1474 | goto bad_fork_cleanup_semundo; |
fb0a685c DRO |
1475 | retval = copy_fs(clone_flags, p); |
1476 | if (retval) | |
1da177e4 | 1477 | goto bad_fork_cleanup_files; |
fb0a685c DRO |
1478 | retval = copy_sighand(clone_flags, p); |
1479 | if (retval) | |
1da177e4 | 1480 | goto bad_fork_cleanup_fs; |
fb0a685c DRO |
1481 | retval = copy_signal(clone_flags, p); |
1482 | if (retval) | |
1da177e4 | 1483 | goto bad_fork_cleanup_sighand; |
fb0a685c DRO |
1484 | retval = copy_mm(clone_flags, p); |
1485 | if (retval) | |
1da177e4 | 1486 | goto bad_fork_cleanup_signal; |
fb0a685c DRO |
1487 | retval = copy_namespaces(clone_flags, p); |
1488 | if (retval) | |
d84f4f99 | 1489 | goto bad_fork_cleanup_mm; |
fb0a685c DRO |
1490 | retval = copy_io(clone_flags, p); |
1491 | if (retval) | |
fd0928df | 1492 | goto bad_fork_cleanup_namespaces; |
3033f14a | 1493 | retval = copy_thread_tls(clone_flags, stack_start, stack_size, p, tls); |
1da177e4 | 1494 | if (retval) |
fd0928df | 1495 | goto bad_fork_cleanup_io; |
1da177e4 | 1496 | |
425fb2b4 | 1497 | if (pid != &init_struct_pid) { |
c2b1df2e | 1498 | pid = alloc_pid(p->nsproxy->pid_ns_for_children); |
35f71bc0 MH |
1499 | if (IS_ERR(pid)) { |
1500 | retval = PTR_ERR(pid); | |
0740aa5f | 1501 | goto bad_fork_cleanup_thread; |
35f71bc0 | 1502 | } |
425fb2b4 PE |
1503 | } |
1504 | ||
1da177e4 LT |
1505 | p->set_child_tid = (clone_flags & CLONE_CHILD_SETTID) ? child_tidptr : NULL; |
1506 | /* | |
1507 | * Clear TID on mm_release()? | |
1508 | */ | |
fb0a685c | 1509 | p->clear_child_tid = (clone_flags & CLONE_CHILD_CLEARTID) ? child_tidptr : NULL; |
73c10101 JA |
1510 | #ifdef CONFIG_BLOCK |
1511 | p->plug = NULL; | |
1512 | #endif | |
42b2dd0a | 1513 | #ifdef CONFIG_FUTEX |
8f17d3a5 IM |
1514 | p->robust_list = NULL; |
1515 | #ifdef CONFIG_COMPAT | |
1516 | p->compat_robust_list = NULL; | |
1517 | #endif | |
c87e2837 IM |
1518 | INIT_LIST_HEAD(&p->pi_state_list); |
1519 | p->pi_state_cache = NULL; | |
42b2dd0a | 1520 | #endif |
f9a3879a GM |
1521 | /* |
1522 | * sigaltstack should be cleared when sharing the same VM | |
1523 | */ | |
1524 | if ((clone_flags & (CLONE_VM|CLONE_VFORK)) == CLONE_VM) | |
2a742138 | 1525 | sas_ss_reset(p); |
f9a3879a | 1526 | |
1da177e4 | 1527 | /* |
6580807d ON |
1528 | * Syscall tracing and stepping should be turned off in the |
1529 | * child regardless of CLONE_PTRACE. | |
1da177e4 | 1530 | */ |
6580807d | 1531 | user_disable_single_step(p); |
1da177e4 | 1532 | clear_tsk_thread_flag(p, TIF_SYSCALL_TRACE); |
ed75e8d5 LV |
1533 | #ifdef TIF_SYSCALL_EMU |
1534 | clear_tsk_thread_flag(p, TIF_SYSCALL_EMU); | |
1535 | #endif | |
9745512c | 1536 | clear_all_latency_tracing(p); |
1da177e4 | 1537 | |
1da177e4 | 1538 | /* ok, now we should be set up.. */ |
18c830df ON |
1539 | p->pid = pid_nr(pid); |
1540 | if (clone_flags & CLONE_THREAD) { | |
5f8aadd8 | 1541 | p->exit_signal = -1; |
18c830df ON |
1542 | p->group_leader = current->group_leader; |
1543 | p->tgid = current->tgid; | |
1544 | } else { | |
1545 | if (clone_flags & CLONE_PARENT) | |
1546 | p->exit_signal = current->group_leader->exit_signal; | |
1547 | else | |
1548 | p->exit_signal = (clone_flags & CSIGNAL); | |
1549 | p->group_leader = p; | |
1550 | p->tgid = p->pid; | |
1551 | } | |
5f8aadd8 | 1552 | |
9d823e8f WF |
1553 | p->nr_dirtied = 0; |
1554 | p->nr_dirtied_pause = 128 >> (PAGE_SHIFT - 10); | |
83712358 | 1555 | p->dirty_paused_when = 0; |
9d823e8f | 1556 | |
bb8cbbfe | 1557 | p->pdeath_signal = 0; |
47e65328 | 1558 | INIT_LIST_HEAD(&p->thread_group); |
158e1645 | 1559 | p->task_works = NULL; |
1da177e4 | 1560 | |
7e47682e AS |
1561 | /* |
1562 | * Ensure that the cgroup subsystem policies allow the new process to be | |
1563 | * forked. It should be noted the the new process's css_set can be changed | |
1564 | * between here and cgroup_post_fork() if an organisation operation is in | |
1565 | * progress. | |
1566 | */ | |
b53202e6 | 1567 | retval = cgroup_can_fork(p); |
7e47682e AS |
1568 | if (retval) |
1569 | goto bad_fork_free_pid; | |
1570 | ||
18c830df ON |
1571 | /* |
1572 | * Make it visible to the rest of the system, but dont wake it up yet. | |
1573 | * Need tasklist lock for parent etc handling! | |
1574 | */ | |
1da177e4 LT |
1575 | write_lock_irq(&tasklist_lock); |
1576 | ||
1da177e4 | 1577 | /* CLONE_PARENT re-uses the old parent */ |
2d5516cb | 1578 | if (clone_flags & (CLONE_PARENT|CLONE_THREAD)) { |
1da177e4 | 1579 | p->real_parent = current->real_parent; |
2d5516cb ON |
1580 | p->parent_exec_id = current->parent_exec_id; |
1581 | } else { | |
1da177e4 | 1582 | p->real_parent = current; |
2d5516cb ON |
1583 | p->parent_exec_id = current->self_exec_id; |
1584 | } | |
1da177e4 | 1585 | |
3f17da69 | 1586 | spin_lock(¤t->sighand->siglock); |
4a2c7a78 | 1587 | |
dbd95212 KC |
1588 | /* |
1589 | * Copy seccomp details explicitly here, in case they were changed | |
1590 | * before holding sighand lock. | |
1591 | */ | |
1592 | copy_seccomp(p); | |
1593 | ||
4a2c7a78 ON |
1594 | /* |
1595 | * Process group and session signals need to be delivered to just the | |
1596 | * parent before the fork or both the parent and the child after the | |
1597 | * fork. Restart if a signal comes in before we add the new process to | |
1598 | * it's process group. | |
1599 | * A fatal signal pending means that current will exit, so the new | |
1600 | * thread can't slip out of an OOM kill (or normal SIGKILL). | |
fb0a685c | 1601 | */ |
23ff4440 | 1602 | recalc_sigpending(); |
4a2c7a78 ON |
1603 | if (signal_pending(current)) { |
1604 | spin_unlock(¤t->sighand->siglock); | |
1605 | write_unlock_irq(&tasklist_lock); | |
1606 | retval = -ERESTARTNOINTR; | |
7e47682e | 1607 | goto bad_fork_cancel_cgroup; |
4a2c7a78 ON |
1608 | } |
1609 | ||
73b9ebfe | 1610 | if (likely(p->pid)) { |
4b9d33e6 | 1611 | ptrace_init_task(p, (clone_flags & CLONE_PTRACE) || trace); |
73b9ebfe | 1612 | |
81907739 | 1613 | init_task_pid(p, PIDTYPE_PID, pid); |
73b9ebfe | 1614 | if (thread_group_leader(p)) { |
81907739 ON |
1615 | init_task_pid(p, PIDTYPE_PGID, task_pgrp(current)); |
1616 | init_task_pid(p, PIDTYPE_SID, task_session(current)); | |
1617 | ||
1c4042c2 | 1618 | if (is_child_reaper(pid)) { |
17cf22c3 | 1619 | ns_of_pid(pid)->child_reaper = p; |
1c4042c2 EB |
1620 | p->signal->flags |= SIGNAL_UNKILLABLE; |
1621 | } | |
73b9ebfe | 1622 | |
fea9d175 | 1623 | p->signal->leader_pid = pid; |
9c9f4ded | 1624 | p->signal->tty = tty_kref_get(current->signal->tty); |
9cd80bbb | 1625 | list_add_tail(&p->sibling, &p->real_parent->children); |
5e85d4ab | 1626 | list_add_tail_rcu(&p->tasks, &init_task.tasks); |
81907739 ON |
1627 | attach_pid(p, PIDTYPE_PGID); |
1628 | attach_pid(p, PIDTYPE_SID); | |
909ea964 | 1629 | __this_cpu_inc(process_counts); |
80628ca0 ON |
1630 | } else { |
1631 | current->signal->nr_threads++; | |
1632 | atomic_inc(¤t->signal->live); | |
1633 | atomic_inc(¤t->signal->sigcnt); | |
80628ca0 ON |
1634 | list_add_tail_rcu(&p->thread_group, |
1635 | &p->group_leader->thread_group); | |
0c740d0a ON |
1636 | list_add_tail_rcu(&p->thread_node, |
1637 | &p->signal->thread_head); | |
73b9ebfe | 1638 | } |
81907739 | 1639 | attach_pid(p, PIDTYPE_PID); |
73b9ebfe | 1640 | nr_threads++; |
1da177e4 LT |
1641 | } |
1642 | ||
1da177e4 | 1643 | total_forks++; |
3f17da69 | 1644 | spin_unlock(¤t->sighand->siglock); |
4af4206b | 1645 | syscall_tracepoint_update(p); |
1da177e4 | 1646 | write_unlock_irq(&tasklist_lock); |
4af4206b | 1647 | |
c13cf856 | 1648 | proc_fork_connector(p); |
b53202e6 | 1649 | cgroup_post_fork(p); |
c9e75f04 | 1650 | threadgroup_change_end(current); |
cdd6c482 | 1651 | perf_event_fork(p); |
43d2b113 KH |
1652 | |
1653 | trace_task_newtask(p, clone_flags); | |
3ab67966 | 1654 | uprobe_copy_process(p, clone_flags); |
43d2b113 | 1655 | |
1da177e4 LT |
1656 | return p; |
1657 | ||
7e47682e | 1658 | bad_fork_cancel_cgroup: |
b53202e6 | 1659 | cgroup_cancel_fork(p); |
425fb2b4 PE |
1660 | bad_fork_free_pid: |
1661 | if (pid != &init_struct_pid) | |
1662 | free_pid(pid); | |
0740aa5f JS |
1663 | bad_fork_cleanup_thread: |
1664 | exit_thread(p); | |
fd0928df | 1665 | bad_fork_cleanup_io: |
b69f2292 LR |
1666 | if (p->io_context) |
1667 | exit_io_context(p); | |
ab516013 | 1668 | bad_fork_cleanup_namespaces: |
444f378b | 1669 | exit_task_namespaces(p); |
1da177e4 | 1670 | bad_fork_cleanup_mm: |
c9f01245 | 1671 | if (p->mm) |
1da177e4 LT |
1672 | mmput(p->mm); |
1673 | bad_fork_cleanup_signal: | |
4ab6c083 | 1674 | if (!(clone_flags & CLONE_THREAD)) |
1c5354de | 1675 | free_signal_struct(p->signal); |
1da177e4 | 1676 | bad_fork_cleanup_sighand: |
a7e5328a | 1677 | __cleanup_sighand(p->sighand); |
1da177e4 LT |
1678 | bad_fork_cleanup_fs: |
1679 | exit_fs(p); /* blocking */ | |
1680 | bad_fork_cleanup_files: | |
1681 | exit_files(p); /* blocking */ | |
1682 | bad_fork_cleanup_semundo: | |
1683 | exit_sem(p); | |
1684 | bad_fork_cleanup_audit: | |
1685 | audit_free(p); | |
6c72e350 | 1686 | bad_fork_cleanup_perf: |
cdd6c482 | 1687 | perf_event_free_task(p); |
6c72e350 | 1688 | bad_fork_cleanup_policy: |
1da177e4 | 1689 | #ifdef CONFIG_NUMA |
f0be3d32 | 1690 | mpol_put(p->mempolicy); |
e8604cb4 | 1691 | bad_fork_cleanup_threadgroup_lock: |
1da177e4 | 1692 | #endif |
c9e75f04 | 1693 | threadgroup_change_end(current); |
35df17c5 | 1694 | delayacct_tsk_free(p); |
1da177e4 | 1695 | bad_fork_cleanup_count: |
d84f4f99 | 1696 | atomic_dec(&p->cred->user->processes); |
e0e81739 | 1697 | exit_creds(p); |
1da177e4 LT |
1698 | bad_fork_free: |
1699 | free_task(p); | |
fe7d37d1 ON |
1700 | fork_out: |
1701 | return ERR_PTR(retval); | |
1da177e4 LT |
1702 | } |
1703 | ||
f106eee1 ON |
1704 | static inline void init_idle_pids(struct pid_link *links) |
1705 | { | |
1706 | enum pid_type type; | |
1707 | ||
1708 | for (type = PIDTYPE_PID; type < PIDTYPE_MAX; ++type) { | |
1709 | INIT_HLIST_NODE(&links[type].node); /* not really needed */ | |
1710 | links[type].pid = &init_struct_pid; | |
1711 | } | |
1712 | } | |
1713 | ||
0db0628d | 1714 | struct task_struct *fork_idle(int cpu) |
1da177e4 | 1715 | { |
36c8b586 | 1716 | struct task_struct *task; |
725fc629 AK |
1717 | task = copy_process(CLONE_VM, 0, 0, NULL, &init_struct_pid, 0, 0, |
1718 | cpu_to_node(cpu)); | |
f106eee1 ON |
1719 | if (!IS_ERR(task)) { |
1720 | init_idle_pids(task->pids); | |
753ca4f3 | 1721 | init_idle(task, cpu); |
f106eee1 | 1722 | } |
73b9ebfe | 1723 | |
1da177e4 LT |
1724 | return task; |
1725 | } | |
1726 | ||
1da177e4 LT |
1727 | /* |
1728 | * Ok, this is the main fork-routine. | |
1729 | * | |
1730 | * It copies the process, and if successful kick-starts | |
1731 | * it and waits for it to finish using the VM if required. | |
1732 | */ | |
3033f14a | 1733 | long _do_fork(unsigned long clone_flags, |
1da177e4 | 1734 | unsigned long stack_start, |
1da177e4 LT |
1735 | unsigned long stack_size, |
1736 | int __user *parent_tidptr, | |
3033f14a JT |
1737 | int __user *child_tidptr, |
1738 | unsigned long tls) | |
1da177e4 LT |
1739 | { |
1740 | struct task_struct *p; | |
1741 | int trace = 0; | |
92476d7f | 1742 | long nr; |
1da177e4 | 1743 | |
09a05394 | 1744 | /* |
4b9d33e6 TH |
1745 | * Determine whether and which event to report to ptracer. When |
1746 | * called from kernel_thread or CLONE_UNTRACED is explicitly | |
1747 | * requested, no event is reported; otherwise, report if the event | |
1748 | * for the type of forking is enabled. | |
09a05394 | 1749 | */ |
e80d6661 | 1750 | if (!(clone_flags & CLONE_UNTRACED)) { |
4b9d33e6 TH |
1751 | if (clone_flags & CLONE_VFORK) |
1752 | trace = PTRACE_EVENT_VFORK; | |
1753 | else if ((clone_flags & CSIGNAL) != SIGCHLD) | |
1754 | trace = PTRACE_EVENT_CLONE; | |
1755 | else | |
1756 | trace = PTRACE_EVENT_FORK; | |
1757 | ||
1758 | if (likely(!ptrace_event_enabled(current, trace))) | |
1759 | trace = 0; | |
1760 | } | |
1da177e4 | 1761 | |
62e791c1 | 1762 | p = copy_process(clone_flags, stack_start, stack_size, |
725fc629 | 1763 | child_tidptr, NULL, trace, tls, NUMA_NO_NODE); |
1da177e4 LT |
1764 | /* |
1765 | * Do this prior waking up the new thread - the thread pointer | |
1766 | * might get invalid after that point, if the thread exits quickly. | |
1767 | */ | |
1768 | if (!IS_ERR(p)) { | |
1769 | struct completion vfork; | |
4e52365f | 1770 | struct pid *pid; |
1da177e4 | 1771 | |
0a16b607 MD |
1772 | trace_sched_process_fork(current, p); |
1773 | ||
4e52365f MD |
1774 | pid = get_task_pid(p, PIDTYPE_PID); |
1775 | nr = pid_vnr(pid); | |
30e49c26 PE |
1776 | |
1777 | if (clone_flags & CLONE_PARENT_SETTID) | |
1778 | put_user(nr, parent_tidptr); | |
a6f5e063 | 1779 | |
1da177e4 LT |
1780 | if (clone_flags & CLONE_VFORK) { |
1781 | p->vfork_done = &vfork; | |
1782 | init_completion(&vfork); | |
d68b46fe | 1783 | get_task_struct(p); |
1da177e4 LT |
1784 | } |
1785 | ||
3e51e3ed | 1786 | wake_up_new_task(p); |
1da177e4 | 1787 | |
4b9d33e6 TH |
1788 | /* forking complete and child started to run, tell ptracer */ |
1789 | if (unlikely(trace)) | |
4e52365f | 1790 | ptrace_event_pid(trace, pid); |
09a05394 | 1791 | |
1da177e4 | 1792 | if (clone_flags & CLONE_VFORK) { |
d68b46fe | 1793 | if (!wait_for_vfork_done(p, &vfork)) |
4e52365f | 1794 | ptrace_event_pid(PTRACE_EVENT_VFORK_DONE, pid); |
1da177e4 | 1795 | } |
4e52365f MD |
1796 | |
1797 | put_pid(pid); | |
1da177e4 | 1798 | } else { |
92476d7f | 1799 | nr = PTR_ERR(p); |
1da177e4 | 1800 | } |
92476d7f | 1801 | return nr; |
1da177e4 LT |
1802 | } |
1803 | ||
3033f14a JT |
1804 | #ifndef CONFIG_HAVE_COPY_THREAD_TLS |
1805 | /* For compatibility with architectures that call do_fork directly rather than | |
1806 | * using the syscall entry points below. */ | |
1807 | long do_fork(unsigned long clone_flags, | |
1808 | unsigned long stack_start, | |
1809 | unsigned long stack_size, | |
1810 | int __user *parent_tidptr, | |
1811 | int __user *child_tidptr) | |
1812 | { | |
1813 | return _do_fork(clone_flags, stack_start, stack_size, | |
1814 | parent_tidptr, child_tidptr, 0); | |
1815 | } | |
1816 | #endif | |
1817 | ||
2aa3a7f8 AV |
1818 | /* |
1819 | * Create a kernel thread. | |
1820 | */ | |
1821 | pid_t kernel_thread(int (*fn)(void *), void *arg, unsigned long flags) | |
1822 | { | |
3033f14a JT |
1823 | return _do_fork(flags|CLONE_VM|CLONE_UNTRACED, (unsigned long)fn, |
1824 | (unsigned long)arg, NULL, NULL, 0); | |
2aa3a7f8 | 1825 | } |
2aa3a7f8 | 1826 | |
d2125043 AV |
1827 | #ifdef __ARCH_WANT_SYS_FORK |
1828 | SYSCALL_DEFINE0(fork) | |
1829 | { | |
1830 | #ifdef CONFIG_MMU | |
3033f14a | 1831 | return _do_fork(SIGCHLD, 0, 0, NULL, NULL, 0); |
d2125043 AV |
1832 | #else |
1833 | /* can not support in nommu mode */ | |
5d59e182 | 1834 | return -EINVAL; |
d2125043 AV |
1835 | #endif |
1836 | } | |
1837 | #endif | |
1838 | ||
1839 | #ifdef __ARCH_WANT_SYS_VFORK | |
1840 | SYSCALL_DEFINE0(vfork) | |
1841 | { | |
3033f14a JT |
1842 | return _do_fork(CLONE_VFORK | CLONE_VM | SIGCHLD, 0, |
1843 | 0, NULL, NULL, 0); | |
d2125043 AV |
1844 | } |
1845 | #endif | |
1846 | ||
1847 | #ifdef __ARCH_WANT_SYS_CLONE | |
1848 | #ifdef CONFIG_CLONE_BACKWARDS | |
1849 | SYSCALL_DEFINE5(clone, unsigned long, clone_flags, unsigned long, newsp, | |
1850 | int __user *, parent_tidptr, | |
3033f14a | 1851 | unsigned long, tls, |
d2125043 AV |
1852 | int __user *, child_tidptr) |
1853 | #elif defined(CONFIG_CLONE_BACKWARDS2) | |
1854 | SYSCALL_DEFINE5(clone, unsigned long, newsp, unsigned long, clone_flags, | |
1855 | int __user *, parent_tidptr, | |
1856 | int __user *, child_tidptr, | |
3033f14a | 1857 | unsigned long, tls) |
dfa9771a MS |
1858 | #elif defined(CONFIG_CLONE_BACKWARDS3) |
1859 | SYSCALL_DEFINE6(clone, unsigned long, clone_flags, unsigned long, newsp, | |
1860 | int, stack_size, | |
1861 | int __user *, parent_tidptr, | |
1862 | int __user *, child_tidptr, | |
3033f14a | 1863 | unsigned long, tls) |
d2125043 AV |
1864 | #else |
1865 | SYSCALL_DEFINE5(clone, unsigned long, clone_flags, unsigned long, newsp, | |
1866 | int __user *, parent_tidptr, | |
1867 | int __user *, child_tidptr, | |
3033f14a | 1868 | unsigned long, tls) |
d2125043 AV |
1869 | #endif |
1870 | { | |
3033f14a | 1871 | return _do_fork(clone_flags, newsp, 0, parent_tidptr, child_tidptr, tls); |
d2125043 AV |
1872 | } |
1873 | #endif | |
1874 | ||
5fd63b30 RT |
1875 | #ifndef ARCH_MIN_MMSTRUCT_ALIGN |
1876 | #define ARCH_MIN_MMSTRUCT_ALIGN 0 | |
1877 | #endif | |
1878 | ||
51cc5068 | 1879 | static void sighand_ctor(void *data) |
aa1757f9 ON |
1880 | { |
1881 | struct sighand_struct *sighand = data; | |
1882 | ||
a35afb83 | 1883 | spin_lock_init(&sighand->siglock); |
b8fceee1 | 1884 | init_waitqueue_head(&sighand->signalfd_wqh); |
aa1757f9 ON |
1885 | } |
1886 | ||
1da177e4 LT |
1887 | void __init proc_caches_init(void) |
1888 | { | |
1889 | sighand_cachep = kmem_cache_create("sighand_cache", | |
1890 | sizeof(struct sighand_struct), 0, | |
2dff4405 | 1891 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_DESTROY_BY_RCU| |
5d097056 | 1892 | SLAB_NOTRACK|SLAB_ACCOUNT, sighand_ctor); |
1da177e4 LT |
1893 | signal_cachep = kmem_cache_create("signal_cache", |
1894 | sizeof(struct signal_struct), 0, | |
5d097056 VD |
1895 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK|SLAB_ACCOUNT, |
1896 | NULL); | |
20c2df83 | 1897 | files_cachep = kmem_cache_create("files_cache", |
1da177e4 | 1898 | sizeof(struct files_struct), 0, |
5d097056 VD |
1899 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK|SLAB_ACCOUNT, |
1900 | NULL); | |
20c2df83 | 1901 | fs_cachep = kmem_cache_create("fs_cache", |
1da177e4 | 1902 | sizeof(struct fs_struct), 0, |
5d097056 VD |
1903 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK|SLAB_ACCOUNT, |
1904 | NULL); | |
6345d24d LT |
1905 | /* |
1906 | * FIXME! The "sizeof(struct mm_struct)" currently includes the | |
1907 | * whole struct cpumask for the OFFSTACK case. We could change | |
1908 | * this to *only* allocate as much of it as required by the | |
1909 | * maximum number of CPU's we can ever have. The cpumask_allocation | |
1910 | * is at the end of the structure, exactly for that reason. | |
1911 | */ | |
1da177e4 | 1912 | mm_cachep = kmem_cache_create("mm_struct", |
5fd63b30 | 1913 | sizeof(struct mm_struct), ARCH_MIN_MMSTRUCT_ALIGN, |
5d097056 VD |
1914 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK|SLAB_ACCOUNT, |
1915 | NULL); | |
1916 | vm_area_cachep = KMEM_CACHE(vm_area_struct, SLAB_PANIC|SLAB_ACCOUNT); | |
8feae131 | 1917 | mmap_init(); |
66577193 | 1918 | nsproxy_cache_init(); |
1da177e4 | 1919 | } |
cf2e340f | 1920 | |
cf2e340f | 1921 | /* |
9bfb23fc | 1922 | * Check constraints on flags passed to the unshare system call. |
cf2e340f | 1923 | */ |
9bfb23fc | 1924 | static int check_unshare_flags(unsigned long unshare_flags) |
cf2e340f | 1925 | { |
9bfb23fc ON |
1926 | if (unshare_flags & ~(CLONE_THREAD|CLONE_FS|CLONE_NEWNS|CLONE_SIGHAND| |
1927 | CLONE_VM|CLONE_FILES|CLONE_SYSVSEM| | |
50804fe3 | 1928 | CLONE_NEWUTS|CLONE_NEWIPC|CLONE_NEWNET| |
a79a908f | 1929 | CLONE_NEWUSER|CLONE_NEWPID|CLONE_NEWCGROUP)) |
9bfb23fc | 1930 | return -EINVAL; |
cf2e340f | 1931 | /* |
12c641ab EB |
1932 | * Not implemented, but pretend it works if there is nothing |
1933 | * to unshare. Note that unsharing the address space or the | |
1934 | * signal handlers also need to unshare the signal queues (aka | |
1935 | * CLONE_THREAD). | |
cf2e340f | 1936 | */ |
9bfb23fc | 1937 | if (unshare_flags & (CLONE_THREAD | CLONE_SIGHAND | CLONE_VM)) { |
12c641ab EB |
1938 | if (!thread_group_empty(current)) |
1939 | return -EINVAL; | |
1940 | } | |
1941 | if (unshare_flags & (CLONE_SIGHAND | CLONE_VM)) { | |
1942 | if (atomic_read(¤t->sighand->count) > 1) | |
1943 | return -EINVAL; | |
1944 | } | |
1945 | if (unshare_flags & CLONE_VM) { | |
1946 | if (!current_is_single_threaded()) | |
9bfb23fc ON |
1947 | return -EINVAL; |
1948 | } | |
cf2e340f JD |
1949 | |
1950 | return 0; | |
1951 | } | |
1952 | ||
1953 | /* | |
99d1419d | 1954 | * Unshare the filesystem structure if it is being shared |
cf2e340f JD |
1955 | */ |
1956 | static int unshare_fs(unsigned long unshare_flags, struct fs_struct **new_fsp) | |
1957 | { | |
1958 | struct fs_struct *fs = current->fs; | |
1959 | ||
498052bb AV |
1960 | if (!(unshare_flags & CLONE_FS) || !fs) |
1961 | return 0; | |
1962 | ||
1963 | /* don't need lock here; in the worst case we'll do useless copy */ | |
1964 | if (fs->users == 1) | |
1965 | return 0; | |
1966 | ||
1967 | *new_fsp = copy_fs_struct(fs); | |
1968 | if (!*new_fsp) | |
1969 | return -ENOMEM; | |
cf2e340f JD |
1970 | |
1971 | return 0; | |
1972 | } | |
1973 | ||
cf2e340f | 1974 | /* |
a016f338 | 1975 | * Unshare file descriptor table if it is being shared |
cf2e340f JD |
1976 | */ |
1977 | static int unshare_fd(unsigned long unshare_flags, struct files_struct **new_fdp) | |
1978 | { | |
1979 | struct files_struct *fd = current->files; | |
a016f338 | 1980 | int error = 0; |
cf2e340f JD |
1981 | |
1982 | if ((unshare_flags & CLONE_FILES) && | |
a016f338 JD |
1983 | (fd && atomic_read(&fd->count) > 1)) { |
1984 | *new_fdp = dup_fd(fd, &error); | |
1985 | if (!*new_fdp) | |
1986 | return error; | |
1987 | } | |
cf2e340f JD |
1988 | |
1989 | return 0; | |
1990 | } | |
1991 | ||
cf2e340f JD |
1992 | /* |
1993 | * unshare allows a process to 'unshare' part of the process | |
1994 | * context which was originally shared using clone. copy_* | |
1995 | * functions used by do_fork() cannot be used here directly | |
1996 | * because they modify an inactive task_struct that is being | |
1997 | * constructed. Here we are modifying the current, active, | |
1998 | * task_struct. | |
1999 | */ | |
6559eed8 | 2000 | SYSCALL_DEFINE1(unshare, unsigned long, unshare_flags) |
cf2e340f | 2001 | { |
cf2e340f | 2002 | struct fs_struct *fs, *new_fs = NULL; |
cf2e340f | 2003 | struct files_struct *fd, *new_fd = NULL; |
b2e0d987 | 2004 | struct cred *new_cred = NULL; |
cf7b708c | 2005 | struct nsproxy *new_nsproxy = NULL; |
9edff4ab | 2006 | int do_sysvsem = 0; |
9bfb23fc | 2007 | int err; |
cf2e340f | 2008 | |
b2e0d987 | 2009 | /* |
faf00da5 EB |
2010 | * If unsharing a user namespace must also unshare the thread group |
2011 | * and unshare the filesystem root and working directories. | |
b2e0d987 EB |
2012 | */ |
2013 | if (unshare_flags & CLONE_NEWUSER) | |
e66eded8 | 2014 | unshare_flags |= CLONE_THREAD | CLONE_FS; |
50804fe3 EB |
2015 | /* |
2016 | * If unsharing vm, must also unshare signal handlers. | |
2017 | */ | |
2018 | if (unshare_flags & CLONE_VM) | |
2019 | unshare_flags |= CLONE_SIGHAND; | |
12c641ab EB |
2020 | /* |
2021 | * If unsharing a signal handlers, must also unshare the signal queues. | |
2022 | */ | |
2023 | if (unshare_flags & CLONE_SIGHAND) | |
2024 | unshare_flags |= CLONE_THREAD; | |
9bfb23fc ON |
2025 | /* |
2026 | * If unsharing namespace, must also unshare filesystem information. | |
2027 | */ | |
2028 | if (unshare_flags & CLONE_NEWNS) | |
2029 | unshare_flags |= CLONE_FS; | |
50804fe3 EB |
2030 | |
2031 | err = check_unshare_flags(unshare_flags); | |
2032 | if (err) | |
2033 | goto bad_unshare_out; | |
6013f67f MS |
2034 | /* |
2035 | * CLONE_NEWIPC must also detach from the undolist: after switching | |
2036 | * to a new ipc namespace, the semaphore arrays from the old | |
2037 | * namespace are unreachable. | |
2038 | */ | |
2039 | if (unshare_flags & (CLONE_NEWIPC|CLONE_SYSVSEM)) | |
9edff4ab | 2040 | do_sysvsem = 1; |
fb0a685c DRO |
2041 | err = unshare_fs(unshare_flags, &new_fs); |
2042 | if (err) | |
9bfb23fc | 2043 | goto bad_unshare_out; |
fb0a685c DRO |
2044 | err = unshare_fd(unshare_flags, &new_fd); |
2045 | if (err) | |
9bfb23fc | 2046 | goto bad_unshare_cleanup_fs; |
b2e0d987 | 2047 | err = unshare_userns(unshare_flags, &new_cred); |
fb0a685c | 2048 | if (err) |
9edff4ab | 2049 | goto bad_unshare_cleanup_fd; |
b2e0d987 EB |
2050 | err = unshare_nsproxy_namespaces(unshare_flags, &new_nsproxy, |
2051 | new_cred, new_fs); | |
2052 | if (err) | |
2053 | goto bad_unshare_cleanup_cred; | |
c0b2fc31 | 2054 | |
b2e0d987 | 2055 | if (new_fs || new_fd || do_sysvsem || new_cred || new_nsproxy) { |
9edff4ab MS |
2056 | if (do_sysvsem) { |
2057 | /* | |
2058 | * CLONE_SYSVSEM is equivalent to sys_exit(). | |
2059 | */ | |
2060 | exit_sem(current); | |
2061 | } | |
ab602f79 JM |
2062 | if (unshare_flags & CLONE_NEWIPC) { |
2063 | /* Orphan segments in old ns (see sem above). */ | |
2064 | exit_shm(current); | |
2065 | shm_init_task(current); | |
2066 | } | |
ab516013 | 2067 | |
6f977e6b | 2068 | if (new_nsproxy) |
cf7b708c | 2069 | switch_task_namespaces(current, new_nsproxy); |
cf2e340f | 2070 | |
cf7b708c PE |
2071 | task_lock(current); |
2072 | ||
cf2e340f JD |
2073 | if (new_fs) { |
2074 | fs = current->fs; | |
2a4419b5 | 2075 | spin_lock(&fs->lock); |
cf2e340f | 2076 | current->fs = new_fs; |
498052bb AV |
2077 | if (--fs->users) |
2078 | new_fs = NULL; | |
2079 | else | |
2080 | new_fs = fs; | |
2a4419b5 | 2081 | spin_unlock(&fs->lock); |
cf2e340f JD |
2082 | } |
2083 | ||
cf2e340f JD |
2084 | if (new_fd) { |
2085 | fd = current->files; | |
2086 | current->files = new_fd; | |
2087 | new_fd = fd; | |
2088 | } | |
2089 | ||
2090 | task_unlock(current); | |
b2e0d987 EB |
2091 | |
2092 | if (new_cred) { | |
2093 | /* Install the new user namespace */ | |
2094 | commit_creds(new_cred); | |
2095 | new_cred = NULL; | |
2096 | } | |
cf2e340f JD |
2097 | } |
2098 | ||
b2e0d987 EB |
2099 | bad_unshare_cleanup_cred: |
2100 | if (new_cred) | |
2101 | put_cred(new_cred); | |
cf2e340f JD |
2102 | bad_unshare_cleanup_fd: |
2103 | if (new_fd) | |
2104 | put_files_struct(new_fd); | |
2105 | ||
cf2e340f JD |
2106 | bad_unshare_cleanup_fs: |
2107 | if (new_fs) | |
498052bb | 2108 | free_fs_struct(new_fs); |
cf2e340f | 2109 | |
cf2e340f JD |
2110 | bad_unshare_out: |
2111 | return err; | |
2112 | } | |
3b125388 AV |
2113 | |
2114 | /* | |
2115 | * Helper to unshare the files of the current task. | |
2116 | * We don't want to expose copy_files internals to | |
2117 | * the exec layer of the kernel. | |
2118 | */ | |
2119 | ||
2120 | int unshare_files(struct files_struct **displaced) | |
2121 | { | |
2122 | struct task_struct *task = current; | |
50704516 | 2123 | struct files_struct *copy = NULL; |
3b125388 AV |
2124 | int error; |
2125 | ||
2126 | error = unshare_fd(CLONE_FILES, ©); | |
2127 | if (error || !copy) { | |
2128 | *displaced = NULL; | |
2129 | return error; | |
2130 | } | |
2131 | *displaced = task->files; | |
2132 | task_lock(task); | |
2133 | task->files = copy; | |
2134 | task_unlock(task); | |
2135 | return 0; | |
2136 | } | |
16db3d3f HS |
2137 | |
2138 | int sysctl_max_threads(struct ctl_table *table, int write, | |
2139 | void __user *buffer, size_t *lenp, loff_t *ppos) | |
2140 | { | |
2141 | struct ctl_table t; | |
2142 | int ret; | |
2143 | int threads = max_threads; | |
2144 | int min = MIN_THREADS; | |
2145 | int max = MAX_THREADS; | |
2146 | ||
2147 | t = *table; | |
2148 | t.data = &threads; | |
2149 | t.extra1 = &min; | |
2150 | t.extra2 = &max; | |
2151 | ||
2152 | ret = proc_dointvec_minmax(&t, write, buffer, lenp, ppos); | |
2153 | if (ret || !write) | |
2154 | return ret; | |
2155 | ||
2156 | set_max_threads(threads); | |
2157 | ||
2158 | return 0; | |
2159 | } |