Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * linux/kernel/fork.c | |
3 | * | |
4 | * Copyright (C) 1991, 1992 Linus Torvalds | |
5 | */ | |
6 | ||
7 | /* | |
8 | * 'fork.c' contains the help-routines for the 'fork' system call | |
9 | * (see also entry.S and others). | |
10 | * Fork is rather simple, once you get the hang of it, but the memory | |
11 | * management can be a bitch. See 'mm/memory.c': 'copy_page_range()' | |
12 | */ | |
13 | ||
1da177e4 LT |
14 | #include <linux/slab.h> |
15 | #include <linux/init.h> | |
16 | #include <linux/unistd.h> | |
1da177e4 LT |
17 | #include <linux/module.h> |
18 | #include <linux/vmalloc.h> | |
19 | #include <linux/completion.h> | |
1da177e4 LT |
20 | #include <linux/personality.h> |
21 | #include <linux/mempolicy.h> | |
22 | #include <linux/sem.h> | |
23 | #include <linux/file.h> | |
9f3acc31 | 24 | #include <linux/fdtable.h> |
da9cbc87 | 25 | #include <linux/iocontext.h> |
1da177e4 LT |
26 | #include <linux/key.h> |
27 | #include <linux/binfmts.h> | |
28 | #include <linux/mman.h> | |
cddb8a5c | 29 | #include <linux/mmu_notifier.h> |
1da177e4 | 30 | #include <linux/fs.h> |
615d6e87 DB |
31 | #include <linux/mm.h> |
32 | #include <linux/vmacache.h> | |
ab516013 | 33 | #include <linux/nsproxy.h> |
c59ede7b | 34 | #include <linux/capability.h> |
1da177e4 | 35 | #include <linux/cpu.h> |
b4f48b63 | 36 | #include <linux/cgroup.h> |
1da177e4 | 37 | #include <linux/security.h> |
a1e78772 | 38 | #include <linux/hugetlb.h> |
e2cfabdf | 39 | #include <linux/seccomp.h> |
1da177e4 LT |
40 | #include <linux/swap.h> |
41 | #include <linux/syscalls.h> | |
42 | #include <linux/jiffies.h> | |
43 | #include <linux/futex.h> | |
8141c7f3 | 44 | #include <linux/compat.h> |
207205a2 | 45 | #include <linux/kthread.h> |
7c3ab738 | 46 | #include <linux/task_io_accounting_ops.h> |
ab2af1f5 | 47 | #include <linux/rcupdate.h> |
1da177e4 LT |
48 | #include <linux/ptrace.h> |
49 | #include <linux/mount.h> | |
50 | #include <linux/audit.h> | |
78fb7466 | 51 | #include <linux/memcontrol.h> |
f201ae23 | 52 | #include <linux/ftrace.h> |
5e2bf014 | 53 | #include <linux/proc_fs.h> |
1da177e4 LT |
54 | #include <linux/profile.h> |
55 | #include <linux/rmap.h> | |
f8af4da3 | 56 | #include <linux/ksm.h> |
1da177e4 | 57 | #include <linux/acct.h> |
8f0ab514 | 58 | #include <linux/tsacct_kern.h> |
9f46080c | 59 | #include <linux/cn_proc.h> |
ba96a0c8 | 60 | #include <linux/freezer.h> |
ca74e92b | 61 | #include <linux/delayacct.h> |
ad4ecbcb | 62 | #include <linux/taskstats_kern.h> |
0a425405 | 63 | #include <linux/random.h> |
522ed776 | 64 | #include <linux/tty.h> |
fd0928df | 65 | #include <linux/blkdev.h> |
5ad4e53b | 66 | #include <linux/fs_struct.h> |
7c9f8861 | 67 | #include <linux/magic.h> |
cdd6c482 | 68 | #include <linux/perf_event.h> |
42c4ab41 | 69 | #include <linux/posix-timers.h> |
8e7cac79 | 70 | #include <linux/user-return-notifier.h> |
3d5992d2 | 71 | #include <linux/oom.h> |
ba76149f | 72 | #include <linux/khugepaged.h> |
d80e731e | 73 | #include <linux/signalfd.h> |
0326f5a9 | 74 | #include <linux/uprobes.h> |
a27bb332 | 75 | #include <linux/aio.h> |
52f5684c | 76 | #include <linux/compiler.h> |
16db3d3f | 77 | #include <linux/sysctl.h> |
1da177e4 LT |
78 | |
79 | #include <asm/pgtable.h> | |
80 | #include <asm/pgalloc.h> | |
81 | #include <asm/uaccess.h> | |
82 | #include <asm/mmu_context.h> | |
83 | #include <asm/cacheflush.h> | |
84 | #include <asm/tlbflush.h> | |
85 | ||
ad8d75ff SR |
86 | #include <trace/events/sched.h> |
87 | ||
43d2b113 KH |
88 | #define CREATE_TRACE_POINTS |
89 | #include <trace/events/task.h> | |
90 | ||
ac1b398d HS |
91 | /* |
92 | * Minimum number of threads to boot the kernel | |
93 | */ | |
94 | #define MIN_THREADS 20 | |
95 | ||
96 | /* | |
97 | * Maximum number of threads | |
98 | */ | |
99 | #define MAX_THREADS FUTEX_TID_MASK | |
100 | ||
1da177e4 LT |
101 | /* |
102 | * Protected counters by write_lock_irq(&tasklist_lock) | |
103 | */ | |
104 | unsigned long total_forks; /* Handle normal Linux uptimes. */ | |
fb0a685c | 105 | int nr_threads; /* The idle threads do not count.. */ |
1da177e4 LT |
106 | |
107 | int max_threads; /* tunable limit on nr_threads */ | |
108 | ||
109 | DEFINE_PER_CPU(unsigned long, process_counts) = 0; | |
110 | ||
c59923a1 | 111 | __cacheline_aligned DEFINE_RWLOCK(tasklist_lock); /* outer */ |
db1466b3 PM |
112 | |
113 | #ifdef CONFIG_PROVE_RCU | |
114 | int lockdep_tasklist_lock_is_held(void) | |
115 | { | |
116 | return lockdep_is_held(&tasklist_lock); | |
117 | } | |
118 | EXPORT_SYMBOL_GPL(lockdep_tasklist_lock_is_held); | |
119 | #endif /* #ifdef CONFIG_PROVE_RCU */ | |
1da177e4 LT |
120 | |
121 | int nr_processes(void) | |
122 | { | |
123 | int cpu; | |
124 | int total = 0; | |
125 | ||
1d510750 | 126 | for_each_possible_cpu(cpu) |
1da177e4 LT |
127 | total += per_cpu(process_counts, cpu); |
128 | ||
129 | return total; | |
130 | } | |
131 | ||
f19b9f74 AM |
132 | void __weak arch_release_task_struct(struct task_struct *tsk) |
133 | { | |
134 | } | |
135 | ||
f5e10287 | 136 | #ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR |
e18b890b | 137 | static struct kmem_cache *task_struct_cachep; |
41101809 TG |
138 | |
139 | static inline struct task_struct *alloc_task_struct_node(int node) | |
140 | { | |
141 | return kmem_cache_alloc_node(task_struct_cachep, GFP_KERNEL, node); | |
142 | } | |
143 | ||
41101809 TG |
144 | static inline void free_task_struct(struct task_struct *tsk) |
145 | { | |
41101809 TG |
146 | kmem_cache_free(task_struct_cachep, tsk); |
147 | } | |
1da177e4 LT |
148 | #endif |
149 | ||
f19b9f74 AM |
150 | void __weak arch_release_thread_info(struct thread_info *ti) |
151 | { | |
152 | } | |
153 | ||
f5e10287 | 154 | #ifndef CONFIG_ARCH_THREAD_INFO_ALLOCATOR |
41101809 | 155 | |
0d15d74a TG |
156 | /* |
157 | * Allocate pages if THREAD_SIZE is >= PAGE_SIZE, otherwise use a | |
158 | * kmemcache based allocator. | |
159 | */ | |
160 | # if THREAD_SIZE >= PAGE_SIZE | |
b6a84016 ED |
161 | static struct thread_info *alloc_thread_info_node(struct task_struct *tsk, |
162 | int node) | |
b69c49b7 | 163 | { |
52383431 VD |
164 | struct page *page = alloc_kmem_pages_node(node, THREADINFO_GFP, |
165 | THREAD_SIZE_ORDER); | |
b6a84016 ED |
166 | |
167 | return page ? page_address(page) : NULL; | |
b69c49b7 FT |
168 | } |
169 | ||
170 | static inline void free_thread_info(struct thread_info *ti) | |
171 | { | |
52383431 | 172 | free_kmem_pages((unsigned long)ti, THREAD_SIZE_ORDER); |
b69c49b7 | 173 | } |
0d15d74a TG |
174 | # else |
175 | static struct kmem_cache *thread_info_cache; | |
176 | ||
177 | static struct thread_info *alloc_thread_info_node(struct task_struct *tsk, | |
178 | int node) | |
179 | { | |
180 | return kmem_cache_alloc_node(thread_info_cache, THREADINFO_GFP, node); | |
181 | } | |
182 | ||
183 | static void free_thread_info(struct thread_info *ti) | |
184 | { | |
0d15d74a TG |
185 | kmem_cache_free(thread_info_cache, ti); |
186 | } | |
187 | ||
188 | void thread_info_cache_init(void) | |
189 | { | |
190 | thread_info_cache = kmem_cache_create("thread_info", THREAD_SIZE, | |
191 | THREAD_SIZE, 0, NULL); | |
192 | BUG_ON(thread_info_cache == NULL); | |
193 | } | |
194 | # endif | |
b69c49b7 FT |
195 | #endif |
196 | ||
1da177e4 | 197 | /* SLAB cache for signal_struct structures (tsk->signal) */ |
e18b890b | 198 | static struct kmem_cache *signal_cachep; |
1da177e4 LT |
199 | |
200 | /* SLAB cache for sighand_struct structures (tsk->sighand) */ | |
e18b890b | 201 | struct kmem_cache *sighand_cachep; |
1da177e4 LT |
202 | |
203 | /* SLAB cache for files_struct structures (tsk->files) */ | |
e18b890b | 204 | struct kmem_cache *files_cachep; |
1da177e4 LT |
205 | |
206 | /* SLAB cache for fs_struct structures (tsk->fs) */ | |
e18b890b | 207 | struct kmem_cache *fs_cachep; |
1da177e4 LT |
208 | |
209 | /* SLAB cache for vm_area_struct structures */ | |
e18b890b | 210 | struct kmem_cache *vm_area_cachep; |
1da177e4 LT |
211 | |
212 | /* SLAB cache for mm_struct structures (tsk->mm) */ | |
e18b890b | 213 | static struct kmem_cache *mm_cachep; |
1da177e4 | 214 | |
c6a7f572 KM |
215 | static void account_kernel_stack(struct thread_info *ti, int account) |
216 | { | |
217 | struct zone *zone = page_zone(virt_to_page(ti)); | |
218 | ||
219 | mod_zone_page_state(zone, NR_KERNEL_STACK, account); | |
220 | } | |
221 | ||
1da177e4 LT |
222 | void free_task(struct task_struct *tsk) |
223 | { | |
c6a7f572 | 224 | account_kernel_stack(tsk->stack, -1); |
f19b9f74 | 225 | arch_release_thread_info(tsk->stack); |
f7e4217b | 226 | free_thread_info(tsk->stack); |
23f78d4a | 227 | rt_mutex_debug_task_free(tsk); |
fb52607a | 228 | ftrace_graph_exit_task(tsk); |
e2cfabdf | 229 | put_seccomp_filter(tsk); |
f19b9f74 | 230 | arch_release_task_struct(tsk); |
1da177e4 LT |
231 | free_task_struct(tsk); |
232 | } | |
233 | EXPORT_SYMBOL(free_task); | |
234 | ||
ea6d290c ON |
235 | static inline void free_signal_struct(struct signal_struct *sig) |
236 | { | |
97101eb4 | 237 | taskstats_tgid_free(sig); |
1c5354de | 238 | sched_autogroup_exit(sig); |
ea6d290c ON |
239 | kmem_cache_free(signal_cachep, sig); |
240 | } | |
241 | ||
242 | static inline void put_signal_struct(struct signal_struct *sig) | |
243 | { | |
1c5354de | 244 | if (atomic_dec_and_test(&sig->sigcnt)) |
ea6d290c ON |
245 | free_signal_struct(sig); |
246 | } | |
247 | ||
158d9ebd | 248 | void __put_task_struct(struct task_struct *tsk) |
1da177e4 | 249 | { |
270f722d | 250 | WARN_ON(!tsk->exit_state); |
1da177e4 LT |
251 | WARN_ON(atomic_read(&tsk->usage)); |
252 | WARN_ON(tsk == current); | |
253 | ||
156654f4 | 254 | task_numa_free(tsk); |
1a2a4d06 | 255 | security_task_free(tsk); |
e0e81739 | 256 | exit_creds(tsk); |
35df17c5 | 257 | delayacct_tsk_free(tsk); |
ea6d290c | 258 | put_signal_struct(tsk->signal); |
1da177e4 LT |
259 | |
260 | if (!profile_handoff_task(tsk)) | |
261 | free_task(tsk); | |
262 | } | |
77c100c8 | 263 | EXPORT_SYMBOL_GPL(__put_task_struct); |
1da177e4 | 264 | |
6c0a9fa6 | 265 | void __init __weak arch_task_cache_init(void) { } |
61c4628b | 266 | |
ff691f6e HS |
267 | /* |
268 | * set_max_threads | |
269 | */ | |
16db3d3f | 270 | static void set_max_threads(unsigned int max_threads_suggested) |
ff691f6e | 271 | { |
ac1b398d | 272 | u64 threads; |
ff691f6e HS |
273 | |
274 | /* | |
ac1b398d HS |
275 | * The number of threads shall be limited such that the thread |
276 | * structures may only consume a small part of the available memory. | |
ff691f6e | 277 | */ |
ac1b398d HS |
278 | if (fls64(totalram_pages) + fls64(PAGE_SIZE) > 64) |
279 | threads = MAX_THREADS; | |
280 | else | |
281 | threads = div64_u64((u64) totalram_pages * (u64) PAGE_SIZE, | |
282 | (u64) THREAD_SIZE * 8UL); | |
283 | ||
16db3d3f HS |
284 | if (threads > max_threads_suggested) |
285 | threads = max_threads_suggested; | |
286 | ||
ac1b398d | 287 | max_threads = clamp_t(u64, threads, MIN_THREADS, MAX_THREADS); |
ff691f6e HS |
288 | } |
289 | ||
290 | void __init fork_init(void) | |
1da177e4 | 291 | { |
f5e10287 | 292 | #ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR |
1da177e4 LT |
293 | #ifndef ARCH_MIN_TASKALIGN |
294 | #define ARCH_MIN_TASKALIGN L1_CACHE_BYTES | |
295 | #endif | |
296 | /* create a slab on which task_structs can be allocated */ | |
297 | task_struct_cachep = | |
298 | kmem_cache_create("task_struct", sizeof(struct task_struct), | |
2dff4405 | 299 | ARCH_MIN_TASKALIGN, SLAB_PANIC | SLAB_NOTRACK, NULL); |
1da177e4 LT |
300 | #endif |
301 | ||
61c4628b SS |
302 | /* do the arch specific task caches init */ |
303 | arch_task_cache_init(); | |
304 | ||
16db3d3f | 305 | set_max_threads(MAX_THREADS); |
1da177e4 LT |
306 | |
307 | init_task.signal->rlim[RLIMIT_NPROC].rlim_cur = max_threads/2; | |
308 | init_task.signal->rlim[RLIMIT_NPROC].rlim_max = max_threads/2; | |
309 | init_task.signal->rlim[RLIMIT_SIGPENDING] = | |
310 | init_task.signal->rlim[RLIMIT_NPROC]; | |
311 | } | |
312 | ||
52f5684c | 313 | int __weak arch_dup_task_struct(struct task_struct *dst, |
61c4628b SS |
314 | struct task_struct *src) |
315 | { | |
316 | *dst = *src; | |
317 | return 0; | |
318 | } | |
319 | ||
d4311ff1 AT |
320 | void set_task_stack_end_magic(struct task_struct *tsk) |
321 | { | |
322 | unsigned long *stackend; | |
323 | ||
324 | stackend = end_of_stack(tsk); | |
325 | *stackend = STACK_END_MAGIC; /* for overflow detection */ | |
326 | } | |
327 | ||
1da177e4 LT |
328 | static struct task_struct *dup_task_struct(struct task_struct *orig) |
329 | { | |
330 | struct task_struct *tsk; | |
331 | struct thread_info *ti; | |
207205a2 | 332 | int node = tsk_fork_get_node(orig); |
3e26c149 | 333 | int err; |
1da177e4 | 334 | |
504f52b5 | 335 | tsk = alloc_task_struct_node(node); |
1da177e4 LT |
336 | if (!tsk) |
337 | return NULL; | |
338 | ||
b6a84016 | 339 | ti = alloc_thread_info_node(tsk, node); |
f19b9f74 AM |
340 | if (!ti) |
341 | goto free_tsk; | |
1da177e4 | 342 | |
fb0a685c | 343 | err = arch_dup_task_struct(tsk, orig); |
164c33c6 | 344 | if (err) |
f19b9f74 | 345 | goto free_ti; |
164c33c6 | 346 | |
87bec58a | 347 | tsk->stack = ti; |
dbd95212 KC |
348 | #ifdef CONFIG_SECCOMP |
349 | /* | |
350 | * We must handle setting up seccomp filters once we're under | |
351 | * the sighand lock in case orig has changed between now and | |
352 | * then. Until then, filter must be NULL to avoid messing up | |
353 | * the usage counts on the error path calling free_task. | |
354 | */ | |
355 | tsk->seccomp.filter = NULL; | |
356 | #endif | |
87bec58a AM |
357 | |
358 | setup_thread_stack(tsk, orig); | |
8e7cac79 | 359 | clear_user_return_notifier(tsk); |
f26f9aff | 360 | clear_tsk_need_resched(tsk); |
d4311ff1 | 361 | set_task_stack_end_magic(tsk); |
1da177e4 | 362 | |
0a425405 AV |
363 | #ifdef CONFIG_CC_STACKPROTECTOR |
364 | tsk->stack_canary = get_random_int(); | |
365 | #endif | |
366 | ||
fb0a685c DRO |
367 | /* |
368 | * One for us, one for whoever does the "release_task()" (usually | |
369 | * parent) | |
370 | */ | |
371 | atomic_set(&tsk->usage, 2); | |
6c5c9341 | 372 | #ifdef CONFIG_BLK_DEV_IO_TRACE |
2056a782 | 373 | tsk->btrace_seq = 0; |
6c5c9341 | 374 | #endif |
a0aa7f68 | 375 | tsk->splice_pipe = NULL; |
5640f768 | 376 | tsk->task_frag.page = NULL; |
c6a7f572 KM |
377 | |
378 | account_kernel_stack(ti, 1); | |
379 | ||
1da177e4 | 380 | return tsk; |
61c4628b | 381 | |
f19b9f74 | 382 | free_ti: |
61c4628b | 383 | free_thread_info(ti); |
f19b9f74 | 384 | free_tsk: |
61c4628b SS |
385 | free_task_struct(tsk); |
386 | return NULL; | |
1da177e4 LT |
387 | } |
388 | ||
389 | #ifdef CONFIG_MMU | |
a39bc516 | 390 | static int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm) |
1da177e4 | 391 | { |
297c5eee | 392 | struct vm_area_struct *mpnt, *tmp, *prev, **pprev; |
1da177e4 LT |
393 | struct rb_node **rb_link, *rb_parent; |
394 | int retval; | |
395 | unsigned long charge; | |
1da177e4 | 396 | |
32cdba1e | 397 | uprobe_start_dup_mmap(); |
1da177e4 | 398 | down_write(&oldmm->mmap_sem); |
ec8c0446 | 399 | flush_cache_dup_mm(oldmm); |
f8ac4ec9 | 400 | uprobe_dup_mmap(oldmm, mm); |
ad339451 IM |
401 | /* |
402 | * Not linked in yet - no deadlock potential: | |
403 | */ | |
404 | down_write_nested(&mm->mmap_sem, SINGLE_DEPTH_NESTING); | |
7ee78232 | 405 | |
90f31d0e KK |
406 | /* No ordering required: file already has been exposed. */ |
407 | RCU_INIT_POINTER(mm->exe_file, get_mm_exe_file(oldmm)); | |
408 | ||
4f7d4614 VD |
409 | mm->total_vm = oldmm->total_vm; |
410 | mm->shared_vm = oldmm->shared_vm; | |
411 | mm->exec_vm = oldmm->exec_vm; | |
412 | mm->stack_vm = oldmm->stack_vm; | |
413 | ||
1da177e4 LT |
414 | rb_link = &mm->mm_rb.rb_node; |
415 | rb_parent = NULL; | |
416 | pprev = &mm->mmap; | |
f8af4da3 | 417 | retval = ksm_fork(mm, oldmm); |
ba76149f AA |
418 | if (retval) |
419 | goto out; | |
420 | retval = khugepaged_fork(mm, oldmm); | |
f8af4da3 HD |
421 | if (retval) |
422 | goto out; | |
1da177e4 | 423 | |
297c5eee | 424 | prev = NULL; |
fd3e42fc | 425 | for (mpnt = oldmm->mmap; mpnt; mpnt = mpnt->vm_next) { |
1da177e4 LT |
426 | struct file *file; |
427 | ||
428 | if (mpnt->vm_flags & VM_DONTCOPY) { | |
ab50b8ed | 429 | vm_stat_account(mm, mpnt->vm_flags, mpnt->vm_file, |
44de9d0c | 430 | -vma_pages(mpnt)); |
1da177e4 LT |
431 | continue; |
432 | } | |
433 | charge = 0; | |
434 | if (mpnt->vm_flags & VM_ACCOUNT) { | |
b2412b7f HS |
435 | unsigned long len = vma_pages(mpnt); |
436 | ||
191c5424 | 437 | if (security_vm_enough_memory_mm(oldmm, len)) /* sic */ |
1da177e4 LT |
438 | goto fail_nomem; |
439 | charge = len; | |
440 | } | |
e94b1766 | 441 | tmp = kmem_cache_alloc(vm_area_cachep, GFP_KERNEL); |
1da177e4 LT |
442 | if (!tmp) |
443 | goto fail_nomem; | |
444 | *tmp = *mpnt; | |
5beb4930 | 445 | INIT_LIST_HEAD(&tmp->anon_vma_chain); |
ef0855d3 ON |
446 | retval = vma_dup_policy(mpnt, tmp); |
447 | if (retval) | |
1da177e4 | 448 | goto fail_nomem_policy; |
a247c3a9 | 449 | tmp->vm_mm = mm; |
5beb4930 RR |
450 | if (anon_vma_fork(tmp, mpnt)) |
451 | goto fail_nomem_anon_vma_fork; | |
1da177e4 | 452 | tmp->vm_flags &= ~VM_LOCKED; |
297c5eee | 453 | tmp->vm_next = tmp->vm_prev = NULL; |
1da177e4 LT |
454 | file = tmp->vm_file; |
455 | if (file) { | |
496ad9aa | 456 | struct inode *inode = file_inode(file); |
b88ed205 HD |
457 | struct address_space *mapping = file->f_mapping; |
458 | ||
1da177e4 LT |
459 | get_file(file); |
460 | if (tmp->vm_flags & VM_DENYWRITE) | |
461 | atomic_dec(&inode->i_writecount); | |
83cde9e8 | 462 | i_mmap_lock_write(mapping); |
b88ed205 | 463 | if (tmp->vm_flags & VM_SHARED) |
4bb5f5d9 | 464 | atomic_inc(&mapping->i_mmap_writable); |
b88ed205 HD |
465 | flush_dcache_mmap_lock(mapping); |
466 | /* insert tmp into the share list, just after mpnt */ | |
27ba0644 KS |
467 | vma_interval_tree_insert_after(tmp, mpnt, |
468 | &mapping->i_mmap); | |
b88ed205 | 469 | flush_dcache_mmap_unlock(mapping); |
83cde9e8 | 470 | i_mmap_unlock_write(mapping); |
1da177e4 LT |
471 | } |
472 | ||
a1e78772 MG |
473 | /* |
474 | * Clear hugetlb-related page reserves for children. This only | |
475 | * affects MAP_PRIVATE mappings. Faults generated by the child | |
476 | * are not guaranteed to succeed, even if read-only | |
477 | */ | |
478 | if (is_vm_hugetlb_page(tmp)) | |
479 | reset_vma_resv_huge_pages(tmp); | |
480 | ||
1da177e4 | 481 | /* |
7ee78232 | 482 | * Link in the new vma and copy the page table entries. |
1da177e4 | 483 | */ |
1da177e4 LT |
484 | *pprev = tmp; |
485 | pprev = &tmp->vm_next; | |
297c5eee LT |
486 | tmp->vm_prev = prev; |
487 | prev = tmp; | |
1da177e4 LT |
488 | |
489 | __vma_link_rb(mm, tmp, rb_link, rb_parent); | |
490 | rb_link = &tmp->vm_rb.rb_right; | |
491 | rb_parent = &tmp->vm_rb; | |
492 | ||
493 | mm->map_count++; | |
0b0db14c | 494 | retval = copy_page_range(mm, oldmm, mpnt); |
1da177e4 LT |
495 | |
496 | if (tmp->vm_ops && tmp->vm_ops->open) | |
497 | tmp->vm_ops->open(tmp); | |
498 | ||
499 | if (retval) | |
500 | goto out; | |
501 | } | |
d6dd61c8 JF |
502 | /* a new mm has just been created */ |
503 | arch_dup_mmap(oldmm, mm); | |
1da177e4 | 504 | retval = 0; |
1da177e4 | 505 | out: |
7ee78232 | 506 | up_write(&mm->mmap_sem); |
fd3e42fc | 507 | flush_tlb_mm(oldmm); |
1da177e4 | 508 | up_write(&oldmm->mmap_sem); |
32cdba1e | 509 | uprobe_end_dup_mmap(); |
1da177e4 | 510 | return retval; |
5beb4930 | 511 | fail_nomem_anon_vma_fork: |
ef0855d3 | 512 | mpol_put(vma_policy(tmp)); |
1da177e4 LT |
513 | fail_nomem_policy: |
514 | kmem_cache_free(vm_area_cachep, tmp); | |
515 | fail_nomem: | |
516 | retval = -ENOMEM; | |
517 | vm_unacct_memory(charge); | |
518 | goto out; | |
519 | } | |
520 | ||
fb0a685c | 521 | static inline int mm_alloc_pgd(struct mm_struct *mm) |
1da177e4 LT |
522 | { |
523 | mm->pgd = pgd_alloc(mm); | |
524 | if (unlikely(!mm->pgd)) | |
525 | return -ENOMEM; | |
526 | return 0; | |
527 | } | |
528 | ||
fb0a685c | 529 | static inline void mm_free_pgd(struct mm_struct *mm) |
1da177e4 | 530 | { |
5e541973 | 531 | pgd_free(mm, mm->pgd); |
1da177e4 LT |
532 | } |
533 | #else | |
90f31d0e KK |
534 | static int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm) |
535 | { | |
536 | down_write(&oldmm->mmap_sem); | |
537 | RCU_INIT_POINTER(mm->exe_file, get_mm_exe_file(oldmm)); | |
538 | up_write(&oldmm->mmap_sem); | |
539 | return 0; | |
540 | } | |
1da177e4 LT |
541 | #define mm_alloc_pgd(mm) (0) |
542 | #define mm_free_pgd(mm) | |
543 | #endif /* CONFIG_MMU */ | |
544 | ||
23ff4440 | 545 | __cacheline_aligned_in_smp DEFINE_SPINLOCK(mmlist_lock); |
1da177e4 | 546 | |
e94b1766 | 547 | #define allocate_mm() (kmem_cache_alloc(mm_cachep, GFP_KERNEL)) |
1da177e4 LT |
548 | #define free_mm(mm) (kmem_cache_free(mm_cachep, (mm))) |
549 | ||
4cb0e11b HK |
550 | static unsigned long default_dump_filter = MMF_DUMP_FILTER_DEFAULT; |
551 | ||
552 | static int __init coredump_filter_setup(char *s) | |
553 | { | |
554 | default_dump_filter = | |
555 | (simple_strtoul(s, NULL, 0) << MMF_DUMP_FILTER_SHIFT) & | |
556 | MMF_DUMP_FILTER_MASK; | |
557 | return 1; | |
558 | } | |
559 | ||
560 | __setup("coredump_filter=", coredump_filter_setup); | |
561 | ||
1da177e4 LT |
562 | #include <linux/init_task.h> |
563 | ||
858f0993 AD |
564 | static void mm_init_aio(struct mm_struct *mm) |
565 | { | |
566 | #ifdef CONFIG_AIO | |
567 | spin_lock_init(&mm->ioctx_lock); | |
db446a08 | 568 | mm->ioctx_table = NULL; |
858f0993 AD |
569 | #endif |
570 | } | |
571 | ||
33144e84 VD |
572 | static void mm_init_owner(struct mm_struct *mm, struct task_struct *p) |
573 | { | |
574 | #ifdef CONFIG_MEMCG | |
575 | mm->owner = p; | |
576 | #endif | |
577 | } | |
578 | ||
fb0a685c | 579 | static struct mm_struct *mm_init(struct mm_struct *mm, struct task_struct *p) |
1da177e4 | 580 | { |
41f727fd VD |
581 | mm->mmap = NULL; |
582 | mm->mm_rb = RB_ROOT; | |
583 | mm->vmacache_seqnum = 0; | |
1da177e4 LT |
584 | atomic_set(&mm->mm_users, 1); |
585 | atomic_set(&mm->mm_count, 1); | |
586 | init_rwsem(&mm->mmap_sem); | |
587 | INIT_LIST_HEAD(&mm->mmlist); | |
999d9fc1 | 588 | mm->core_state = NULL; |
e1f56c89 | 589 | atomic_long_set(&mm->nr_ptes, 0); |
2d2f5119 | 590 | mm_nr_pmds_init(mm); |
41f727fd VD |
591 | mm->map_count = 0; |
592 | mm->locked_vm = 0; | |
ce65cefa | 593 | mm->pinned_vm = 0; |
d559db08 | 594 | memset(&mm->rss_stat, 0, sizeof(mm->rss_stat)); |
1da177e4 | 595 | spin_lock_init(&mm->page_table_lock); |
41f727fd | 596 | mm_init_cpumask(mm); |
858f0993 | 597 | mm_init_aio(mm); |
cf475ad2 | 598 | mm_init_owner(mm, p); |
41f727fd | 599 | mmu_notifier_mm_init(mm); |
20841405 | 600 | clear_tlb_flush_pending(mm); |
41f727fd VD |
601 | #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS |
602 | mm->pmd_huge_pte = NULL; | |
603 | #endif | |
1da177e4 | 604 | |
a0715cc2 AT |
605 | if (current->mm) { |
606 | mm->flags = current->mm->flags & MMF_INIT_MASK; | |
607 | mm->def_flags = current->mm->def_flags & VM_INIT_DEF_MASK; | |
608 | } else { | |
609 | mm->flags = default_dump_filter; | |
1da177e4 | 610 | mm->def_flags = 0; |
a0715cc2 AT |
611 | } |
612 | ||
41f727fd VD |
613 | if (mm_alloc_pgd(mm)) |
614 | goto fail_nopgd; | |
615 | ||
616 | if (init_new_context(p, mm)) | |
617 | goto fail_nocontext; | |
78fb7466 | 618 | |
41f727fd VD |
619 | return mm; |
620 | ||
621 | fail_nocontext: | |
622 | mm_free_pgd(mm); | |
623 | fail_nopgd: | |
1da177e4 LT |
624 | free_mm(mm); |
625 | return NULL; | |
626 | } | |
627 | ||
c3f0327f KK |
628 | static void check_mm(struct mm_struct *mm) |
629 | { | |
630 | int i; | |
631 | ||
632 | for (i = 0; i < NR_MM_COUNTERS; i++) { | |
633 | long x = atomic_long_read(&mm->rss_stat.count[i]); | |
634 | ||
635 | if (unlikely(x)) | |
636 | printk(KERN_ALERT "BUG: Bad rss-counter state " | |
637 | "mm:%p idx:%d val:%ld\n", mm, i, x); | |
638 | } | |
b30fe6c7 KS |
639 | |
640 | if (atomic_long_read(&mm->nr_ptes)) | |
641 | pr_alert("BUG: non-zero nr_ptes on freeing mm: %ld\n", | |
642 | atomic_long_read(&mm->nr_ptes)); | |
643 | if (mm_nr_pmds(mm)) | |
644 | pr_alert("BUG: non-zero nr_pmds on freeing mm: %ld\n", | |
645 | mm_nr_pmds(mm)); | |
646 | ||
e009bb30 | 647 | #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS |
96dad67f | 648 | VM_BUG_ON_MM(mm->pmd_huge_pte, mm); |
c3f0327f KK |
649 | #endif |
650 | } | |
651 | ||
1da177e4 LT |
652 | /* |
653 | * Allocate and initialize an mm_struct. | |
654 | */ | |
fb0a685c | 655 | struct mm_struct *mm_alloc(void) |
1da177e4 | 656 | { |
fb0a685c | 657 | struct mm_struct *mm; |
1da177e4 LT |
658 | |
659 | mm = allocate_mm(); | |
de03c72c KM |
660 | if (!mm) |
661 | return NULL; | |
662 | ||
663 | memset(mm, 0, sizeof(*mm)); | |
6345d24d | 664 | return mm_init(mm, current); |
1da177e4 LT |
665 | } |
666 | ||
667 | /* | |
668 | * Called when the last reference to the mm | |
669 | * is dropped: either by a lazy thread or by | |
670 | * mmput. Free the page directory and the mm. | |
671 | */ | |
7ad5b3a5 | 672 | void __mmdrop(struct mm_struct *mm) |
1da177e4 LT |
673 | { |
674 | BUG_ON(mm == &init_mm); | |
675 | mm_free_pgd(mm); | |
676 | destroy_context(mm); | |
cddb8a5c | 677 | mmu_notifier_mm_destroy(mm); |
c3f0327f | 678 | check_mm(mm); |
1da177e4 LT |
679 | free_mm(mm); |
680 | } | |
6d4e4c4f | 681 | EXPORT_SYMBOL_GPL(__mmdrop); |
1da177e4 LT |
682 | |
683 | /* | |
684 | * Decrement the use count and release all resources for an mm. | |
685 | */ | |
686 | void mmput(struct mm_struct *mm) | |
687 | { | |
0ae26f1b AM |
688 | might_sleep(); |
689 | ||
1da177e4 | 690 | if (atomic_dec_and_test(&mm->mm_users)) { |
d4b3b638 | 691 | uprobe_clear_state(mm); |
1da177e4 | 692 | exit_aio(mm); |
1c2fb7a4 | 693 | ksm_exit(mm); |
ba76149f | 694 | khugepaged_exit(mm); /* must run before exit_mmap */ |
1da177e4 | 695 | exit_mmap(mm); |
925d1c40 | 696 | set_mm_exe_file(mm, NULL); |
1da177e4 LT |
697 | if (!list_empty(&mm->mmlist)) { |
698 | spin_lock(&mmlist_lock); | |
699 | list_del(&mm->mmlist); | |
700 | spin_unlock(&mmlist_lock); | |
701 | } | |
801460d0 HS |
702 | if (mm->binfmt) |
703 | module_put(mm->binfmt->module); | |
1da177e4 LT |
704 | mmdrop(mm); |
705 | } | |
706 | } | |
707 | EXPORT_SYMBOL_GPL(mmput); | |
708 | ||
90f31d0e KK |
709 | /** |
710 | * set_mm_exe_file - change a reference to the mm's executable file | |
711 | * | |
712 | * This changes mm's executable file (shown as symlink /proc/[pid]/exe). | |
713 | * | |
6e399cd1 DB |
714 | * Main users are mmput() and sys_execve(). Callers prevent concurrent |
715 | * invocations: in mmput() nobody alive left, in execve task is single | |
716 | * threaded. sys_prctl(PR_SET_MM_MAP/EXE_FILE) also needs to set the | |
717 | * mm->exe_file, but does so without using set_mm_exe_file() in order | |
718 | * to do avoid the need for any locks. | |
90f31d0e | 719 | */ |
38646013 JS |
720 | void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file) |
721 | { | |
6e399cd1 DB |
722 | struct file *old_exe_file; |
723 | ||
724 | /* | |
725 | * It is safe to dereference the exe_file without RCU as | |
726 | * this function is only called if nobody else can access | |
727 | * this mm -- see comment above for justification. | |
728 | */ | |
729 | old_exe_file = rcu_dereference_raw(mm->exe_file); | |
90f31d0e | 730 | |
38646013 JS |
731 | if (new_exe_file) |
732 | get_file(new_exe_file); | |
90f31d0e KK |
733 | rcu_assign_pointer(mm->exe_file, new_exe_file); |
734 | if (old_exe_file) | |
735 | fput(old_exe_file); | |
38646013 JS |
736 | } |
737 | ||
90f31d0e KK |
738 | /** |
739 | * get_mm_exe_file - acquire a reference to the mm's executable file | |
740 | * | |
741 | * Returns %NULL if mm has no associated executable file. | |
742 | * User must release file via fput(). | |
743 | */ | |
38646013 JS |
744 | struct file *get_mm_exe_file(struct mm_struct *mm) |
745 | { | |
746 | struct file *exe_file; | |
747 | ||
90f31d0e KK |
748 | rcu_read_lock(); |
749 | exe_file = rcu_dereference(mm->exe_file); | |
750 | if (exe_file && !get_file_rcu(exe_file)) | |
751 | exe_file = NULL; | |
752 | rcu_read_unlock(); | |
38646013 JS |
753 | return exe_file; |
754 | } | |
755 | ||
1da177e4 LT |
756 | /** |
757 | * get_task_mm - acquire a reference to the task's mm | |
758 | * | |
246bb0b1 | 759 | * Returns %NULL if the task has no mm. Checks PF_KTHREAD (meaning |
1da177e4 LT |
760 | * this kernel workthread has transiently adopted a user mm with use_mm, |
761 | * to do its AIO) is not set and if so returns a reference to it, after | |
762 | * bumping up the use count. User must release the mm via mmput() | |
763 | * after use. Typically used by /proc and ptrace. | |
764 | */ | |
765 | struct mm_struct *get_task_mm(struct task_struct *task) | |
766 | { | |
767 | struct mm_struct *mm; | |
768 | ||
769 | task_lock(task); | |
770 | mm = task->mm; | |
771 | if (mm) { | |
246bb0b1 | 772 | if (task->flags & PF_KTHREAD) |
1da177e4 LT |
773 | mm = NULL; |
774 | else | |
775 | atomic_inc(&mm->mm_users); | |
776 | } | |
777 | task_unlock(task); | |
778 | return mm; | |
779 | } | |
780 | EXPORT_SYMBOL_GPL(get_task_mm); | |
781 | ||
8cdb878d CY |
782 | struct mm_struct *mm_access(struct task_struct *task, unsigned int mode) |
783 | { | |
784 | struct mm_struct *mm; | |
785 | int err; | |
786 | ||
787 | err = mutex_lock_killable(&task->signal->cred_guard_mutex); | |
788 | if (err) | |
789 | return ERR_PTR(err); | |
790 | ||
791 | mm = get_task_mm(task); | |
792 | if (mm && mm != current->mm && | |
793 | !ptrace_may_access(task, mode)) { | |
794 | mmput(mm); | |
795 | mm = ERR_PTR(-EACCES); | |
796 | } | |
797 | mutex_unlock(&task->signal->cred_guard_mutex); | |
798 | ||
799 | return mm; | |
800 | } | |
801 | ||
57b59c4a | 802 | static void complete_vfork_done(struct task_struct *tsk) |
c415c3b4 | 803 | { |
d68b46fe | 804 | struct completion *vfork; |
c415c3b4 | 805 | |
d68b46fe ON |
806 | task_lock(tsk); |
807 | vfork = tsk->vfork_done; | |
808 | if (likely(vfork)) { | |
809 | tsk->vfork_done = NULL; | |
810 | complete(vfork); | |
811 | } | |
812 | task_unlock(tsk); | |
813 | } | |
814 | ||
815 | static int wait_for_vfork_done(struct task_struct *child, | |
816 | struct completion *vfork) | |
817 | { | |
818 | int killed; | |
819 | ||
820 | freezer_do_not_count(); | |
821 | killed = wait_for_completion_killable(vfork); | |
822 | freezer_count(); | |
823 | ||
824 | if (killed) { | |
825 | task_lock(child); | |
826 | child->vfork_done = NULL; | |
827 | task_unlock(child); | |
828 | } | |
829 | ||
830 | put_task_struct(child); | |
831 | return killed; | |
c415c3b4 ON |
832 | } |
833 | ||
1da177e4 LT |
834 | /* Please note the differences between mmput and mm_release. |
835 | * mmput is called whenever we stop holding onto a mm_struct, | |
836 | * error success whatever. | |
837 | * | |
838 | * mm_release is called after a mm_struct has been removed | |
839 | * from the current process. | |
840 | * | |
841 | * This difference is important for error handling, when we | |
842 | * only half set up a mm_struct for a new process and need to restore | |
843 | * the old one. Because we mmput the new mm_struct before | |
844 | * restoring the old one. . . | |
845 | * Eric Biederman 10 January 1998 | |
846 | */ | |
847 | void mm_release(struct task_struct *tsk, struct mm_struct *mm) | |
848 | { | |
8141c7f3 LT |
849 | /* Get rid of any futexes when releasing the mm */ |
850 | #ifdef CONFIG_FUTEX | |
fc6b177d | 851 | if (unlikely(tsk->robust_list)) { |
8141c7f3 | 852 | exit_robust_list(tsk); |
fc6b177d PZ |
853 | tsk->robust_list = NULL; |
854 | } | |
8141c7f3 | 855 | #ifdef CONFIG_COMPAT |
fc6b177d | 856 | if (unlikely(tsk->compat_robust_list)) { |
8141c7f3 | 857 | compat_exit_robust_list(tsk); |
fc6b177d PZ |
858 | tsk->compat_robust_list = NULL; |
859 | } | |
8141c7f3 | 860 | #endif |
322a2c10 TG |
861 | if (unlikely(!list_empty(&tsk->pi_state_list))) |
862 | exit_pi_state_list(tsk); | |
8141c7f3 LT |
863 | #endif |
864 | ||
0326f5a9 SD |
865 | uprobe_free_utask(tsk); |
866 | ||
1da177e4 LT |
867 | /* Get rid of any cached register state */ |
868 | deactivate_mm(tsk, mm); | |
869 | ||
fec1d011 RM |
870 | /* |
871 | * If we're exiting normally, clear a user-space tid field if | |
872 | * requested. We leave this alone when dying by signal, to leave | |
873 | * the value intact in a core dump, and to save the unnecessary | |
d68b46fe ON |
874 | * trouble, say, a killed vfork parent shouldn't touch this mm. |
875 | * Userland only wants this done for a sys_exit. | |
fec1d011 | 876 | */ |
9c8a8228 ED |
877 | if (tsk->clear_child_tid) { |
878 | if (!(tsk->flags & PF_SIGNALED) && | |
879 | atomic_read(&mm->mm_users) > 1) { | |
880 | /* | |
881 | * We don't check the error code - if userspace has | |
882 | * not set up a proper pointer then tough luck. | |
883 | */ | |
884 | put_user(0, tsk->clear_child_tid); | |
885 | sys_futex(tsk->clear_child_tid, FUTEX_WAKE, | |
886 | 1, NULL, NULL, 0); | |
887 | } | |
1da177e4 | 888 | tsk->clear_child_tid = NULL; |
1da177e4 | 889 | } |
f7505d64 KK |
890 | |
891 | /* | |
892 | * All done, finally we can wake up parent and return this mm to him. | |
893 | * Also kthread_stop() uses this completion for synchronization. | |
894 | */ | |
895 | if (tsk->vfork_done) | |
896 | complete_vfork_done(tsk); | |
1da177e4 LT |
897 | } |
898 | ||
a0a7ec30 JD |
899 | /* |
900 | * Allocate a new mm structure and copy contents from the | |
901 | * mm structure of the passed in task structure. | |
902 | */ | |
ff252c1f | 903 | static struct mm_struct *dup_mm(struct task_struct *tsk) |
a0a7ec30 JD |
904 | { |
905 | struct mm_struct *mm, *oldmm = current->mm; | |
906 | int err; | |
907 | ||
a0a7ec30 JD |
908 | mm = allocate_mm(); |
909 | if (!mm) | |
910 | goto fail_nomem; | |
911 | ||
912 | memcpy(mm, oldmm, sizeof(*mm)); | |
913 | ||
78fb7466 | 914 | if (!mm_init(mm, tsk)) |
a0a7ec30 JD |
915 | goto fail_nomem; |
916 | ||
a0a7ec30 JD |
917 | err = dup_mmap(mm, oldmm); |
918 | if (err) | |
919 | goto free_pt; | |
920 | ||
921 | mm->hiwater_rss = get_mm_rss(mm); | |
922 | mm->hiwater_vm = mm->total_vm; | |
923 | ||
801460d0 HS |
924 | if (mm->binfmt && !try_module_get(mm->binfmt->module)) |
925 | goto free_pt; | |
926 | ||
a0a7ec30 JD |
927 | return mm; |
928 | ||
929 | free_pt: | |
801460d0 HS |
930 | /* don't put binfmt in mmput, we haven't got module yet */ |
931 | mm->binfmt = NULL; | |
a0a7ec30 JD |
932 | mmput(mm); |
933 | ||
934 | fail_nomem: | |
935 | return NULL; | |
a0a7ec30 JD |
936 | } |
937 | ||
fb0a685c | 938 | static int copy_mm(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 | 939 | { |
fb0a685c | 940 | struct mm_struct *mm, *oldmm; |
1da177e4 LT |
941 | int retval; |
942 | ||
943 | tsk->min_flt = tsk->maj_flt = 0; | |
944 | tsk->nvcsw = tsk->nivcsw = 0; | |
17406b82 MSB |
945 | #ifdef CONFIG_DETECT_HUNG_TASK |
946 | tsk->last_switch_count = tsk->nvcsw + tsk->nivcsw; | |
947 | #endif | |
1da177e4 LT |
948 | |
949 | tsk->mm = NULL; | |
950 | tsk->active_mm = NULL; | |
951 | ||
952 | /* | |
953 | * Are we cloning a kernel thread? | |
954 | * | |
955 | * We need to steal a active VM for that.. | |
956 | */ | |
957 | oldmm = current->mm; | |
958 | if (!oldmm) | |
959 | return 0; | |
960 | ||
615d6e87 DB |
961 | /* initialize the new vmacache entries */ |
962 | vmacache_flush(tsk); | |
963 | ||
1da177e4 LT |
964 | if (clone_flags & CLONE_VM) { |
965 | atomic_inc(&oldmm->mm_users); | |
966 | mm = oldmm; | |
1da177e4 LT |
967 | goto good_mm; |
968 | } | |
969 | ||
970 | retval = -ENOMEM; | |
a0a7ec30 | 971 | mm = dup_mm(tsk); |
1da177e4 LT |
972 | if (!mm) |
973 | goto fail_nomem; | |
974 | ||
1da177e4 LT |
975 | good_mm: |
976 | tsk->mm = mm; | |
977 | tsk->active_mm = mm; | |
978 | return 0; | |
979 | ||
1da177e4 LT |
980 | fail_nomem: |
981 | return retval; | |
1da177e4 LT |
982 | } |
983 | ||
a39bc516 | 984 | static int copy_fs(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 | 985 | { |
498052bb | 986 | struct fs_struct *fs = current->fs; |
1da177e4 | 987 | if (clone_flags & CLONE_FS) { |
498052bb | 988 | /* tsk->fs is already what we want */ |
2a4419b5 | 989 | spin_lock(&fs->lock); |
498052bb | 990 | if (fs->in_exec) { |
2a4419b5 | 991 | spin_unlock(&fs->lock); |
498052bb AV |
992 | return -EAGAIN; |
993 | } | |
994 | fs->users++; | |
2a4419b5 | 995 | spin_unlock(&fs->lock); |
1da177e4 LT |
996 | return 0; |
997 | } | |
498052bb | 998 | tsk->fs = copy_fs_struct(fs); |
1da177e4 LT |
999 | if (!tsk->fs) |
1000 | return -ENOMEM; | |
1001 | return 0; | |
1002 | } | |
1003 | ||
fb0a685c | 1004 | static int copy_files(unsigned long clone_flags, struct task_struct *tsk) |
a016f338 JD |
1005 | { |
1006 | struct files_struct *oldf, *newf; | |
1007 | int error = 0; | |
1008 | ||
1009 | /* | |
1010 | * A background process may not have any files ... | |
1011 | */ | |
1012 | oldf = current->files; | |
1013 | if (!oldf) | |
1014 | goto out; | |
1015 | ||
1016 | if (clone_flags & CLONE_FILES) { | |
1017 | atomic_inc(&oldf->count); | |
1018 | goto out; | |
1019 | } | |
1020 | ||
a016f338 JD |
1021 | newf = dup_fd(oldf, &error); |
1022 | if (!newf) | |
1023 | goto out; | |
1024 | ||
1025 | tsk->files = newf; | |
1026 | error = 0; | |
1027 | out: | |
1028 | return error; | |
1029 | } | |
1030 | ||
fadad878 | 1031 | static int copy_io(unsigned long clone_flags, struct task_struct *tsk) |
fd0928df JA |
1032 | { |
1033 | #ifdef CONFIG_BLOCK | |
1034 | struct io_context *ioc = current->io_context; | |
6e736be7 | 1035 | struct io_context *new_ioc; |
fd0928df JA |
1036 | |
1037 | if (!ioc) | |
1038 | return 0; | |
fadad878 JA |
1039 | /* |
1040 | * Share io context with parent, if CLONE_IO is set | |
1041 | */ | |
1042 | if (clone_flags & CLONE_IO) { | |
3d48749d TH |
1043 | ioc_task_link(ioc); |
1044 | tsk->io_context = ioc; | |
fadad878 | 1045 | } else if (ioprio_valid(ioc->ioprio)) { |
6e736be7 TH |
1046 | new_ioc = get_task_io_context(tsk, GFP_KERNEL, NUMA_NO_NODE); |
1047 | if (unlikely(!new_ioc)) | |
fd0928df JA |
1048 | return -ENOMEM; |
1049 | ||
6e736be7 | 1050 | new_ioc->ioprio = ioc->ioprio; |
11a3122f | 1051 | put_io_context(new_ioc); |
fd0928df JA |
1052 | } |
1053 | #endif | |
1054 | return 0; | |
1055 | } | |
1056 | ||
a39bc516 | 1057 | static int copy_sighand(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 LT |
1058 | { |
1059 | struct sighand_struct *sig; | |
1060 | ||
60348802 | 1061 | if (clone_flags & CLONE_SIGHAND) { |
1da177e4 LT |
1062 | atomic_inc(¤t->sighand->count); |
1063 | return 0; | |
1064 | } | |
1065 | sig = kmem_cache_alloc(sighand_cachep, GFP_KERNEL); | |
e56d0903 | 1066 | rcu_assign_pointer(tsk->sighand, sig); |
1da177e4 LT |
1067 | if (!sig) |
1068 | return -ENOMEM; | |
1da177e4 LT |
1069 | atomic_set(&sig->count, 1); |
1070 | memcpy(sig->action, current->sighand->action, sizeof(sig->action)); | |
1071 | return 0; | |
1072 | } | |
1073 | ||
a7e5328a | 1074 | void __cleanup_sighand(struct sighand_struct *sighand) |
c81addc9 | 1075 | { |
d80e731e ON |
1076 | if (atomic_dec_and_test(&sighand->count)) { |
1077 | signalfd_cleanup(sighand); | |
392809b2 ON |
1078 | /* |
1079 | * sighand_cachep is SLAB_DESTROY_BY_RCU so we can free it | |
1080 | * without an RCU grace period, see __lock_task_sighand(). | |
1081 | */ | |
c81addc9 | 1082 | kmem_cache_free(sighand_cachep, sighand); |
d80e731e | 1083 | } |
c81addc9 ON |
1084 | } |
1085 | ||
f06febc9 FM |
1086 | /* |
1087 | * Initialize POSIX timer handling for a thread group. | |
1088 | */ | |
1089 | static void posix_cpu_timers_init_group(struct signal_struct *sig) | |
1090 | { | |
78d7d407 JS |
1091 | unsigned long cpu_limit; |
1092 | ||
f06febc9 FM |
1093 | /* Thread group counters. */ |
1094 | thread_group_cputime_init(sig); | |
1095 | ||
78d7d407 JS |
1096 | cpu_limit = ACCESS_ONCE(sig->rlim[RLIMIT_CPU].rlim_cur); |
1097 | if (cpu_limit != RLIM_INFINITY) { | |
1098 | sig->cputime_expires.prof_exp = secs_to_cputime(cpu_limit); | |
6279a751 ON |
1099 | sig->cputimer.running = 1; |
1100 | } | |
1101 | ||
f06febc9 FM |
1102 | /* The timer lists. */ |
1103 | INIT_LIST_HEAD(&sig->cpu_timers[0]); | |
1104 | INIT_LIST_HEAD(&sig->cpu_timers[1]); | |
1105 | INIT_LIST_HEAD(&sig->cpu_timers[2]); | |
1106 | } | |
1107 | ||
a39bc516 | 1108 | static int copy_signal(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 LT |
1109 | { |
1110 | struct signal_struct *sig; | |
1da177e4 | 1111 | |
4ab6c083 | 1112 | if (clone_flags & CLONE_THREAD) |
490dea45 | 1113 | return 0; |
490dea45 | 1114 | |
a56704ef | 1115 | sig = kmem_cache_zalloc(signal_cachep, GFP_KERNEL); |
1da177e4 LT |
1116 | tsk->signal = sig; |
1117 | if (!sig) | |
1118 | return -ENOMEM; | |
1119 | ||
b3ac022c | 1120 | sig->nr_threads = 1; |
1da177e4 | 1121 | atomic_set(&sig->live, 1); |
b3ac022c | 1122 | atomic_set(&sig->sigcnt, 1); |
0c740d0a ON |
1123 | |
1124 | /* list_add(thread_node, thread_head) without INIT_LIST_HEAD() */ | |
1125 | sig->thread_head = (struct list_head)LIST_HEAD_INIT(tsk->thread_node); | |
1126 | tsk->thread_node = (struct list_head)LIST_HEAD_INIT(sig->thread_head); | |
1127 | ||
1da177e4 | 1128 | init_waitqueue_head(&sig->wait_chldexit); |
db51aecc | 1129 | sig->curr_target = tsk; |
1da177e4 LT |
1130 | init_sigpending(&sig->shared_pending); |
1131 | INIT_LIST_HEAD(&sig->posix_timers); | |
e78c3496 | 1132 | seqlock_init(&sig->stats_lock); |
1da177e4 | 1133 | |
c9cb2e3d | 1134 | hrtimer_init(&sig->real_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL); |
1da177e4 | 1135 | sig->real_timer.function = it_real_fn; |
1da177e4 | 1136 | |
1da177e4 LT |
1137 | task_lock(current->group_leader); |
1138 | memcpy(sig->rlim, current->signal->rlim, sizeof sig->rlim); | |
1139 | task_unlock(current->group_leader); | |
1140 | ||
6279a751 ON |
1141 | posix_cpu_timers_init_group(sig); |
1142 | ||
522ed776 | 1143 | tty_audit_fork(sig); |
5091faa4 | 1144 | sched_autogroup_fork(sig); |
522ed776 | 1145 | |
4714d1d3 | 1146 | #ifdef CONFIG_CGROUPS |
257058ae | 1147 | init_rwsem(&sig->group_rwsem); |
4714d1d3 BB |
1148 | #endif |
1149 | ||
a63d83f4 | 1150 | sig->oom_score_adj = current->signal->oom_score_adj; |
dabb16f6 | 1151 | sig->oom_score_adj_min = current->signal->oom_score_adj_min; |
28b83c51 | 1152 | |
ebec18a6 LP |
1153 | sig->has_child_subreaper = current->signal->has_child_subreaper || |
1154 | current->signal->is_child_subreaper; | |
1155 | ||
9b1bf12d KM |
1156 | mutex_init(&sig->cred_guard_mutex); |
1157 | ||
1da177e4 LT |
1158 | return 0; |
1159 | } | |
1160 | ||
dbd95212 KC |
1161 | static void copy_seccomp(struct task_struct *p) |
1162 | { | |
1163 | #ifdef CONFIG_SECCOMP | |
1164 | /* | |
1165 | * Must be called with sighand->lock held, which is common to | |
1166 | * all threads in the group. Holding cred_guard_mutex is not | |
1167 | * needed because this new task is not yet running and cannot | |
1168 | * be racing exec. | |
1169 | */ | |
69f6a34b | 1170 | assert_spin_locked(¤t->sighand->siglock); |
dbd95212 KC |
1171 | |
1172 | /* Ref-count the new filter user, and assign it. */ | |
1173 | get_seccomp_filter(current); | |
1174 | p->seccomp = current->seccomp; | |
1175 | ||
1176 | /* | |
1177 | * Explicitly enable no_new_privs here in case it got set | |
1178 | * between the task_struct being duplicated and holding the | |
1179 | * sighand lock. The seccomp state and nnp must be in sync. | |
1180 | */ | |
1181 | if (task_no_new_privs(current)) | |
1182 | task_set_no_new_privs(p); | |
1183 | ||
1184 | /* | |
1185 | * If the parent gained a seccomp mode after copying thread | |
1186 | * flags and between before we held the sighand lock, we have | |
1187 | * to manually enable the seccomp thread flag here. | |
1188 | */ | |
1189 | if (p->seccomp.mode != SECCOMP_MODE_DISABLED) | |
1190 | set_tsk_thread_flag(p, TIF_SECCOMP); | |
1191 | #endif | |
1192 | } | |
1193 | ||
17da2bd9 | 1194 | SYSCALL_DEFINE1(set_tid_address, int __user *, tidptr) |
1da177e4 LT |
1195 | { |
1196 | current->clear_child_tid = tidptr; | |
1197 | ||
b488893a | 1198 | return task_pid_vnr(current); |
1da177e4 LT |
1199 | } |
1200 | ||
a39bc516 | 1201 | static void rt_mutex_init_task(struct task_struct *p) |
23f78d4a | 1202 | { |
1d615482 | 1203 | raw_spin_lock_init(&p->pi_lock); |
e29e175b | 1204 | #ifdef CONFIG_RT_MUTEXES |
fb00aca4 PZ |
1205 | p->pi_waiters = RB_ROOT; |
1206 | p->pi_waiters_leftmost = NULL; | |
23f78d4a | 1207 | p->pi_blocked_on = NULL; |
23f78d4a IM |
1208 | #endif |
1209 | } | |
1210 | ||
f06febc9 FM |
1211 | /* |
1212 | * Initialize POSIX timer handling for a single task. | |
1213 | */ | |
1214 | static void posix_cpu_timers_init(struct task_struct *tsk) | |
1215 | { | |
64861634 MS |
1216 | tsk->cputime_expires.prof_exp = 0; |
1217 | tsk->cputime_expires.virt_exp = 0; | |
f06febc9 FM |
1218 | tsk->cputime_expires.sched_exp = 0; |
1219 | INIT_LIST_HEAD(&tsk->cpu_timers[0]); | |
1220 | INIT_LIST_HEAD(&tsk->cpu_timers[1]); | |
1221 | INIT_LIST_HEAD(&tsk->cpu_timers[2]); | |
1222 | } | |
1223 | ||
81907739 ON |
1224 | static inline void |
1225 | init_task_pid(struct task_struct *task, enum pid_type type, struct pid *pid) | |
1226 | { | |
1227 | task->pids[type].pid = pid; | |
1228 | } | |
1229 | ||
1da177e4 LT |
1230 | /* |
1231 | * This creates a new process as a copy of the old one, | |
1232 | * but does not actually start it yet. | |
1233 | * | |
1234 | * It copies the registers, and all the appropriate | |
1235 | * parts of the process environment (as per the clone | |
1236 | * flags). The actual kick-off is left to the caller. | |
1237 | */ | |
36c8b586 IM |
1238 | static struct task_struct *copy_process(unsigned long clone_flags, |
1239 | unsigned long stack_start, | |
36c8b586 | 1240 | unsigned long stack_size, |
36c8b586 | 1241 | int __user *child_tidptr, |
09a05394 RM |
1242 | struct pid *pid, |
1243 | int trace) | |
1da177e4 LT |
1244 | { |
1245 | int retval; | |
a24efe62 | 1246 | struct task_struct *p; |
1da177e4 LT |
1247 | |
1248 | if ((clone_flags & (CLONE_NEWNS|CLONE_FS)) == (CLONE_NEWNS|CLONE_FS)) | |
1249 | return ERR_PTR(-EINVAL); | |
1250 | ||
e66eded8 EB |
1251 | if ((clone_flags & (CLONE_NEWUSER|CLONE_FS)) == (CLONE_NEWUSER|CLONE_FS)) |
1252 | return ERR_PTR(-EINVAL); | |
1253 | ||
1da177e4 LT |
1254 | /* |
1255 | * Thread groups must share signals as well, and detached threads | |
1256 | * can only be started up within the thread group. | |
1257 | */ | |
1258 | if ((clone_flags & CLONE_THREAD) && !(clone_flags & CLONE_SIGHAND)) | |
1259 | return ERR_PTR(-EINVAL); | |
1260 | ||
1261 | /* | |
1262 | * Shared signal handlers imply shared VM. By way of the above, | |
1263 | * thread groups also imply shared VM. Blocking this case allows | |
1264 | * for various simplifications in other code. | |
1265 | */ | |
1266 | if ((clone_flags & CLONE_SIGHAND) && !(clone_flags & CLONE_VM)) | |
1267 | return ERR_PTR(-EINVAL); | |
1268 | ||
123be07b SB |
1269 | /* |
1270 | * Siblings of global init remain as zombies on exit since they are | |
1271 | * not reaped by their parent (swapper). To solve this and to avoid | |
1272 | * multi-rooted process trees, prevent global and container-inits | |
1273 | * from creating siblings. | |
1274 | */ | |
1275 | if ((clone_flags & CLONE_PARENT) && | |
1276 | current->signal->flags & SIGNAL_UNKILLABLE) | |
1277 | return ERR_PTR(-EINVAL); | |
1278 | ||
8382fcac | 1279 | /* |
40a0d32d ON |
1280 | * If the new process will be in a different pid or user namespace |
1281 | * do not allow it to share a thread group or signal handlers or | |
1282 | * parent with the forking task. | |
8382fcac | 1283 | */ |
1f7f4dde | 1284 | if (clone_flags & CLONE_SIGHAND) { |
40a0d32d ON |
1285 | if ((clone_flags & (CLONE_NEWUSER | CLONE_NEWPID)) || |
1286 | (task_active_pid_ns(current) != | |
1287 | current->nsproxy->pid_ns_for_children)) | |
1288 | return ERR_PTR(-EINVAL); | |
1289 | } | |
8382fcac | 1290 | |
1da177e4 LT |
1291 | retval = security_task_create(clone_flags); |
1292 | if (retval) | |
1293 | goto fork_out; | |
1294 | ||
1295 | retval = -ENOMEM; | |
1296 | p = dup_task_struct(current); | |
1297 | if (!p) | |
1298 | goto fork_out; | |
1299 | ||
f7e8b616 SR |
1300 | ftrace_graph_init_task(p); |
1301 | ||
bea493a0 PZ |
1302 | rt_mutex_init_task(p); |
1303 | ||
d12c1a37 | 1304 | #ifdef CONFIG_PROVE_LOCKING |
de30a2b3 IM |
1305 | DEBUG_LOCKS_WARN_ON(!p->hardirqs_enabled); |
1306 | DEBUG_LOCKS_WARN_ON(!p->softirqs_enabled); | |
1307 | #endif | |
1da177e4 | 1308 | retval = -EAGAIN; |
3b11a1de | 1309 | if (atomic_read(&p->real_cred->user->processes) >= |
78d7d407 | 1310 | task_rlimit(p, RLIMIT_NPROC)) { |
b57922b6 EP |
1311 | if (p->real_cred->user != INIT_USER && |
1312 | !capable(CAP_SYS_RESOURCE) && !capable(CAP_SYS_ADMIN)) | |
1da177e4 LT |
1313 | goto bad_fork_free; |
1314 | } | |
72fa5997 | 1315 | current->flags &= ~PF_NPROC_EXCEEDED; |
1da177e4 | 1316 | |
f1752eec DH |
1317 | retval = copy_creds(p, clone_flags); |
1318 | if (retval < 0) | |
1319 | goto bad_fork_free; | |
1da177e4 LT |
1320 | |
1321 | /* | |
1322 | * If multiple threads are within copy_process(), then this check | |
1323 | * triggers too late. This doesn't hurt, the check is only there | |
1324 | * to stop root fork bombs. | |
1325 | */ | |
04ec93fe | 1326 | retval = -EAGAIN; |
1da177e4 LT |
1327 | if (nr_threads >= max_threads) |
1328 | goto bad_fork_cleanup_count; | |
1329 | ||
ca74e92b | 1330 | delayacct_tsk_init(p); /* Must remain after dup_task_struct() */ |
514ddb44 DR |
1331 | p->flags &= ~(PF_SUPERPRIV | PF_WQ_WORKER); |
1332 | p->flags |= PF_FORKNOEXEC; | |
1da177e4 LT |
1333 | INIT_LIST_HEAD(&p->children); |
1334 | INIT_LIST_HEAD(&p->sibling); | |
f41d911f | 1335 | rcu_copy_process(p); |
1da177e4 LT |
1336 | p->vfork_done = NULL; |
1337 | spin_lock_init(&p->alloc_lock); | |
1da177e4 | 1338 | |
1da177e4 LT |
1339 | init_sigpending(&p->pending); |
1340 | ||
64861634 MS |
1341 | p->utime = p->stime = p->gtime = 0; |
1342 | p->utimescaled = p->stimescaled = 0; | |
9fbc42ea | 1343 | #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE |
d37f761d | 1344 | p->prev_cputime.utime = p->prev_cputime.stime = 0; |
d99ca3b9 | 1345 | #endif |
6a61671b FW |
1346 | #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN |
1347 | seqlock_init(&p->vtime_seqlock); | |
1348 | p->vtime_snap = 0; | |
1349 | p->vtime_snap_whence = VTIME_SLEEPING; | |
1350 | #endif | |
1351 | ||
a3a2e76c KH |
1352 | #if defined(SPLIT_RSS_COUNTING) |
1353 | memset(&p->rss_stat, 0, sizeof(p->rss_stat)); | |
1354 | #endif | |
172ba844 | 1355 | |
6976675d AV |
1356 | p->default_timer_slack_ns = current->timer_slack_ns; |
1357 | ||
5995477a | 1358 | task_io_accounting_init(&p->ioac); |
1da177e4 LT |
1359 | acct_clear_integrals(p); |
1360 | ||
f06febc9 | 1361 | posix_cpu_timers_init(p); |
1da177e4 | 1362 | |
ccbf62d8 | 1363 | p->start_time = ktime_get_ns(); |
57e0be04 | 1364 | p->real_start_time = ktime_get_boot_ns(); |
1da177e4 | 1365 | p->io_context = NULL; |
1da177e4 | 1366 | p->audit_context = NULL; |
4714d1d3 | 1367 | if (clone_flags & CLONE_THREAD) |
257058ae | 1368 | threadgroup_change_begin(current); |
b4f48b63 | 1369 | cgroup_fork(p); |
1da177e4 | 1370 | #ifdef CONFIG_NUMA |
846a16bf | 1371 | p->mempolicy = mpol_dup(p->mempolicy); |
fb0a685c DRO |
1372 | if (IS_ERR(p->mempolicy)) { |
1373 | retval = PTR_ERR(p->mempolicy); | |
1374 | p->mempolicy = NULL; | |
e8604cb4 | 1375 | goto bad_fork_cleanup_threadgroup_lock; |
fb0a685c | 1376 | } |
1da177e4 | 1377 | #endif |
778d3b0f MH |
1378 | #ifdef CONFIG_CPUSETS |
1379 | p->cpuset_mem_spread_rotor = NUMA_NO_NODE; | |
1380 | p->cpuset_slab_spread_rotor = NUMA_NO_NODE; | |
cc9a6c87 | 1381 | seqcount_init(&p->mems_allowed_seq); |
778d3b0f | 1382 | #endif |
de30a2b3 IM |
1383 | #ifdef CONFIG_TRACE_IRQFLAGS |
1384 | p->irq_events = 0; | |
1385 | p->hardirqs_enabled = 0; | |
1386 | p->hardirq_enable_ip = 0; | |
1387 | p->hardirq_enable_event = 0; | |
1388 | p->hardirq_disable_ip = _THIS_IP_; | |
1389 | p->hardirq_disable_event = 0; | |
1390 | p->softirqs_enabled = 1; | |
1391 | p->softirq_enable_ip = _THIS_IP_; | |
1392 | p->softirq_enable_event = 0; | |
1393 | p->softirq_disable_ip = 0; | |
1394 | p->softirq_disable_event = 0; | |
1395 | p->hardirq_context = 0; | |
1396 | p->softirq_context = 0; | |
1397 | #endif | |
fbb9ce95 IM |
1398 | #ifdef CONFIG_LOCKDEP |
1399 | p->lockdep_depth = 0; /* no locks held yet */ | |
1400 | p->curr_chain_key = 0; | |
1401 | p->lockdep_recursion = 0; | |
1402 | #endif | |
1da177e4 | 1403 | |
408894ee IM |
1404 | #ifdef CONFIG_DEBUG_MUTEXES |
1405 | p->blocked_on = NULL; /* not blocked yet */ | |
1406 | #endif | |
cafe5635 KO |
1407 | #ifdef CONFIG_BCACHE |
1408 | p->sequential_io = 0; | |
1409 | p->sequential_io_avg = 0; | |
1410 | #endif | |
0f481406 | 1411 | |
3c90e6e9 | 1412 | /* Perform scheduler related setup. Assign this task to a CPU. */ |
aab03e05 DF |
1413 | retval = sched_fork(clone_flags, p); |
1414 | if (retval) | |
1415 | goto bad_fork_cleanup_policy; | |
6ab423e0 | 1416 | |
cdd6c482 | 1417 | retval = perf_event_init_task(p); |
6ab423e0 PZ |
1418 | if (retval) |
1419 | goto bad_fork_cleanup_policy; | |
fb0a685c DRO |
1420 | retval = audit_alloc(p); |
1421 | if (retval) | |
6c72e350 | 1422 | goto bad_fork_cleanup_perf; |
1da177e4 | 1423 | /* copy all the process information */ |
ab602f79 | 1424 | shm_init_task(p); |
fb0a685c DRO |
1425 | retval = copy_semundo(clone_flags, p); |
1426 | if (retval) | |
1da177e4 | 1427 | goto bad_fork_cleanup_audit; |
fb0a685c DRO |
1428 | retval = copy_files(clone_flags, p); |
1429 | if (retval) | |
1da177e4 | 1430 | goto bad_fork_cleanup_semundo; |
fb0a685c DRO |
1431 | retval = copy_fs(clone_flags, p); |
1432 | if (retval) | |
1da177e4 | 1433 | goto bad_fork_cleanup_files; |
fb0a685c DRO |
1434 | retval = copy_sighand(clone_flags, p); |
1435 | if (retval) | |
1da177e4 | 1436 | goto bad_fork_cleanup_fs; |
fb0a685c DRO |
1437 | retval = copy_signal(clone_flags, p); |
1438 | if (retval) | |
1da177e4 | 1439 | goto bad_fork_cleanup_sighand; |
fb0a685c DRO |
1440 | retval = copy_mm(clone_flags, p); |
1441 | if (retval) | |
1da177e4 | 1442 | goto bad_fork_cleanup_signal; |
fb0a685c DRO |
1443 | retval = copy_namespaces(clone_flags, p); |
1444 | if (retval) | |
d84f4f99 | 1445 | goto bad_fork_cleanup_mm; |
fb0a685c DRO |
1446 | retval = copy_io(clone_flags, p); |
1447 | if (retval) | |
fd0928df | 1448 | goto bad_fork_cleanup_namespaces; |
afa86fc4 | 1449 | retval = copy_thread(clone_flags, stack_start, stack_size, p); |
1da177e4 | 1450 | if (retval) |
fd0928df | 1451 | goto bad_fork_cleanup_io; |
1da177e4 | 1452 | |
425fb2b4 | 1453 | if (pid != &init_struct_pid) { |
c2b1df2e | 1454 | pid = alloc_pid(p->nsproxy->pid_ns_for_children); |
35f71bc0 MH |
1455 | if (IS_ERR(pid)) { |
1456 | retval = PTR_ERR(pid); | |
fd0928df | 1457 | goto bad_fork_cleanup_io; |
35f71bc0 | 1458 | } |
425fb2b4 PE |
1459 | } |
1460 | ||
1da177e4 LT |
1461 | p->set_child_tid = (clone_flags & CLONE_CHILD_SETTID) ? child_tidptr : NULL; |
1462 | /* | |
1463 | * Clear TID on mm_release()? | |
1464 | */ | |
fb0a685c | 1465 | p->clear_child_tid = (clone_flags & CLONE_CHILD_CLEARTID) ? child_tidptr : NULL; |
73c10101 JA |
1466 | #ifdef CONFIG_BLOCK |
1467 | p->plug = NULL; | |
1468 | #endif | |
42b2dd0a | 1469 | #ifdef CONFIG_FUTEX |
8f17d3a5 IM |
1470 | p->robust_list = NULL; |
1471 | #ifdef CONFIG_COMPAT | |
1472 | p->compat_robust_list = NULL; | |
1473 | #endif | |
c87e2837 IM |
1474 | INIT_LIST_HEAD(&p->pi_state_list); |
1475 | p->pi_state_cache = NULL; | |
42b2dd0a | 1476 | #endif |
f9a3879a GM |
1477 | /* |
1478 | * sigaltstack should be cleared when sharing the same VM | |
1479 | */ | |
1480 | if ((clone_flags & (CLONE_VM|CLONE_VFORK)) == CLONE_VM) | |
1481 | p->sas_ss_sp = p->sas_ss_size = 0; | |
1482 | ||
1da177e4 | 1483 | /* |
6580807d ON |
1484 | * Syscall tracing and stepping should be turned off in the |
1485 | * child regardless of CLONE_PTRACE. | |
1da177e4 | 1486 | */ |
6580807d | 1487 | user_disable_single_step(p); |
1da177e4 | 1488 | clear_tsk_thread_flag(p, TIF_SYSCALL_TRACE); |
ed75e8d5 LV |
1489 | #ifdef TIF_SYSCALL_EMU |
1490 | clear_tsk_thread_flag(p, TIF_SYSCALL_EMU); | |
1491 | #endif | |
9745512c | 1492 | clear_all_latency_tracing(p); |
1da177e4 | 1493 | |
1da177e4 | 1494 | /* ok, now we should be set up.. */ |
18c830df ON |
1495 | p->pid = pid_nr(pid); |
1496 | if (clone_flags & CLONE_THREAD) { | |
5f8aadd8 | 1497 | p->exit_signal = -1; |
18c830df ON |
1498 | p->group_leader = current->group_leader; |
1499 | p->tgid = current->tgid; | |
1500 | } else { | |
1501 | if (clone_flags & CLONE_PARENT) | |
1502 | p->exit_signal = current->group_leader->exit_signal; | |
1503 | else | |
1504 | p->exit_signal = (clone_flags & CSIGNAL); | |
1505 | p->group_leader = p; | |
1506 | p->tgid = p->pid; | |
1507 | } | |
5f8aadd8 | 1508 | |
9d823e8f WF |
1509 | p->nr_dirtied = 0; |
1510 | p->nr_dirtied_pause = 128 >> (PAGE_SHIFT - 10); | |
83712358 | 1511 | p->dirty_paused_when = 0; |
9d823e8f | 1512 | |
bb8cbbfe | 1513 | p->pdeath_signal = 0; |
47e65328 | 1514 | INIT_LIST_HEAD(&p->thread_group); |
158e1645 | 1515 | p->task_works = NULL; |
1da177e4 | 1516 | |
18c830df ON |
1517 | /* |
1518 | * Make it visible to the rest of the system, but dont wake it up yet. | |
1519 | * Need tasklist lock for parent etc handling! | |
1520 | */ | |
1da177e4 LT |
1521 | write_lock_irq(&tasklist_lock); |
1522 | ||
1da177e4 | 1523 | /* CLONE_PARENT re-uses the old parent */ |
2d5516cb | 1524 | if (clone_flags & (CLONE_PARENT|CLONE_THREAD)) { |
1da177e4 | 1525 | p->real_parent = current->real_parent; |
2d5516cb ON |
1526 | p->parent_exec_id = current->parent_exec_id; |
1527 | } else { | |
1da177e4 | 1528 | p->real_parent = current; |
2d5516cb ON |
1529 | p->parent_exec_id = current->self_exec_id; |
1530 | } | |
1da177e4 | 1531 | |
3f17da69 | 1532 | spin_lock(¤t->sighand->siglock); |
4a2c7a78 | 1533 | |
dbd95212 KC |
1534 | /* |
1535 | * Copy seccomp details explicitly here, in case they were changed | |
1536 | * before holding sighand lock. | |
1537 | */ | |
1538 | copy_seccomp(p); | |
1539 | ||
4a2c7a78 ON |
1540 | /* |
1541 | * Process group and session signals need to be delivered to just the | |
1542 | * parent before the fork or both the parent and the child after the | |
1543 | * fork. Restart if a signal comes in before we add the new process to | |
1544 | * it's process group. | |
1545 | * A fatal signal pending means that current will exit, so the new | |
1546 | * thread can't slip out of an OOM kill (or normal SIGKILL). | |
fb0a685c | 1547 | */ |
23ff4440 | 1548 | recalc_sigpending(); |
4a2c7a78 ON |
1549 | if (signal_pending(current)) { |
1550 | spin_unlock(¤t->sighand->siglock); | |
1551 | write_unlock_irq(&tasklist_lock); | |
1552 | retval = -ERESTARTNOINTR; | |
f7e8b616 | 1553 | goto bad_fork_free_pid; |
4a2c7a78 ON |
1554 | } |
1555 | ||
73b9ebfe | 1556 | if (likely(p->pid)) { |
4b9d33e6 | 1557 | ptrace_init_task(p, (clone_flags & CLONE_PTRACE) || trace); |
73b9ebfe | 1558 | |
81907739 | 1559 | init_task_pid(p, PIDTYPE_PID, pid); |
73b9ebfe | 1560 | if (thread_group_leader(p)) { |
81907739 ON |
1561 | init_task_pid(p, PIDTYPE_PGID, task_pgrp(current)); |
1562 | init_task_pid(p, PIDTYPE_SID, task_session(current)); | |
1563 | ||
1c4042c2 | 1564 | if (is_child_reaper(pid)) { |
17cf22c3 | 1565 | ns_of_pid(pid)->child_reaper = p; |
1c4042c2 EB |
1566 | p->signal->flags |= SIGNAL_UNKILLABLE; |
1567 | } | |
73b9ebfe | 1568 | |
fea9d175 | 1569 | p->signal->leader_pid = pid; |
9c9f4ded | 1570 | p->signal->tty = tty_kref_get(current->signal->tty); |
9cd80bbb | 1571 | list_add_tail(&p->sibling, &p->real_parent->children); |
5e85d4ab | 1572 | list_add_tail_rcu(&p->tasks, &init_task.tasks); |
81907739 ON |
1573 | attach_pid(p, PIDTYPE_PGID); |
1574 | attach_pid(p, PIDTYPE_SID); | |
909ea964 | 1575 | __this_cpu_inc(process_counts); |
80628ca0 ON |
1576 | } else { |
1577 | current->signal->nr_threads++; | |
1578 | atomic_inc(¤t->signal->live); | |
1579 | atomic_inc(¤t->signal->sigcnt); | |
80628ca0 ON |
1580 | list_add_tail_rcu(&p->thread_group, |
1581 | &p->group_leader->thread_group); | |
0c740d0a ON |
1582 | list_add_tail_rcu(&p->thread_node, |
1583 | &p->signal->thread_head); | |
73b9ebfe | 1584 | } |
81907739 | 1585 | attach_pid(p, PIDTYPE_PID); |
73b9ebfe | 1586 | nr_threads++; |
1da177e4 LT |
1587 | } |
1588 | ||
1da177e4 | 1589 | total_forks++; |
3f17da69 | 1590 | spin_unlock(¤t->sighand->siglock); |
4af4206b | 1591 | syscall_tracepoint_update(p); |
1da177e4 | 1592 | write_unlock_irq(&tasklist_lock); |
4af4206b | 1593 | |
c13cf856 | 1594 | proc_fork_connector(p); |
817929ec | 1595 | cgroup_post_fork(p); |
4714d1d3 | 1596 | if (clone_flags & CLONE_THREAD) |
257058ae | 1597 | threadgroup_change_end(current); |
cdd6c482 | 1598 | perf_event_fork(p); |
43d2b113 KH |
1599 | |
1600 | trace_task_newtask(p, clone_flags); | |
3ab67966 | 1601 | uprobe_copy_process(p, clone_flags); |
43d2b113 | 1602 | |
1da177e4 LT |
1603 | return p; |
1604 | ||
425fb2b4 PE |
1605 | bad_fork_free_pid: |
1606 | if (pid != &init_struct_pid) | |
1607 | free_pid(pid); | |
fd0928df | 1608 | bad_fork_cleanup_io: |
b69f2292 LR |
1609 | if (p->io_context) |
1610 | exit_io_context(p); | |
ab516013 | 1611 | bad_fork_cleanup_namespaces: |
444f378b | 1612 | exit_task_namespaces(p); |
1da177e4 | 1613 | bad_fork_cleanup_mm: |
c9f01245 | 1614 | if (p->mm) |
1da177e4 LT |
1615 | mmput(p->mm); |
1616 | bad_fork_cleanup_signal: | |
4ab6c083 | 1617 | if (!(clone_flags & CLONE_THREAD)) |
1c5354de | 1618 | free_signal_struct(p->signal); |
1da177e4 | 1619 | bad_fork_cleanup_sighand: |
a7e5328a | 1620 | __cleanup_sighand(p->sighand); |
1da177e4 LT |
1621 | bad_fork_cleanup_fs: |
1622 | exit_fs(p); /* blocking */ | |
1623 | bad_fork_cleanup_files: | |
1624 | exit_files(p); /* blocking */ | |
1625 | bad_fork_cleanup_semundo: | |
1626 | exit_sem(p); | |
1627 | bad_fork_cleanup_audit: | |
1628 | audit_free(p); | |
6c72e350 | 1629 | bad_fork_cleanup_perf: |
cdd6c482 | 1630 | perf_event_free_task(p); |
6c72e350 | 1631 | bad_fork_cleanup_policy: |
1da177e4 | 1632 | #ifdef CONFIG_NUMA |
f0be3d32 | 1633 | mpol_put(p->mempolicy); |
e8604cb4 | 1634 | bad_fork_cleanup_threadgroup_lock: |
1da177e4 | 1635 | #endif |
4714d1d3 | 1636 | if (clone_flags & CLONE_THREAD) |
257058ae | 1637 | threadgroup_change_end(current); |
35df17c5 | 1638 | delayacct_tsk_free(p); |
1da177e4 | 1639 | bad_fork_cleanup_count: |
d84f4f99 | 1640 | atomic_dec(&p->cred->user->processes); |
e0e81739 | 1641 | exit_creds(p); |
1da177e4 LT |
1642 | bad_fork_free: |
1643 | free_task(p); | |
fe7d37d1 ON |
1644 | fork_out: |
1645 | return ERR_PTR(retval); | |
1da177e4 LT |
1646 | } |
1647 | ||
f106eee1 ON |
1648 | static inline void init_idle_pids(struct pid_link *links) |
1649 | { | |
1650 | enum pid_type type; | |
1651 | ||
1652 | for (type = PIDTYPE_PID; type < PIDTYPE_MAX; ++type) { | |
1653 | INIT_HLIST_NODE(&links[type].node); /* not really needed */ | |
1654 | links[type].pid = &init_struct_pid; | |
1655 | } | |
1656 | } | |
1657 | ||
0db0628d | 1658 | struct task_struct *fork_idle(int cpu) |
1da177e4 | 1659 | { |
36c8b586 | 1660 | struct task_struct *task; |
62e791c1 | 1661 | task = copy_process(CLONE_VM, 0, 0, NULL, &init_struct_pid, 0); |
f106eee1 ON |
1662 | if (!IS_ERR(task)) { |
1663 | init_idle_pids(task->pids); | |
753ca4f3 | 1664 | init_idle(task, cpu); |
f106eee1 | 1665 | } |
73b9ebfe | 1666 | |
1da177e4 LT |
1667 | return task; |
1668 | } | |
1669 | ||
1da177e4 LT |
1670 | /* |
1671 | * Ok, this is the main fork-routine. | |
1672 | * | |
1673 | * It copies the process, and if successful kick-starts | |
1674 | * it and waits for it to finish using the VM if required. | |
1675 | */ | |
1676 | long do_fork(unsigned long clone_flags, | |
1677 | unsigned long stack_start, | |
1da177e4 LT |
1678 | unsigned long stack_size, |
1679 | int __user *parent_tidptr, | |
1680 | int __user *child_tidptr) | |
1681 | { | |
1682 | struct task_struct *p; | |
1683 | int trace = 0; | |
92476d7f | 1684 | long nr; |
1da177e4 | 1685 | |
09a05394 | 1686 | /* |
4b9d33e6 TH |
1687 | * Determine whether and which event to report to ptracer. When |
1688 | * called from kernel_thread or CLONE_UNTRACED is explicitly | |
1689 | * requested, no event is reported; otherwise, report if the event | |
1690 | * for the type of forking is enabled. | |
09a05394 | 1691 | */ |
e80d6661 | 1692 | if (!(clone_flags & CLONE_UNTRACED)) { |
4b9d33e6 TH |
1693 | if (clone_flags & CLONE_VFORK) |
1694 | trace = PTRACE_EVENT_VFORK; | |
1695 | else if ((clone_flags & CSIGNAL) != SIGCHLD) | |
1696 | trace = PTRACE_EVENT_CLONE; | |
1697 | else | |
1698 | trace = PTRACE_EVENT_FORK; | |
1699 | ||
1700 | if (likely(!ptrace_event_enabled(current, trace))) | |
1701 | trace = 0; | |
1702 | } | |
1da177e4 | 1703 | |
62e791c1 | 1704 | p = copy_process(clone_flags, stack_start, stack_size, |
09a05394 | 1705 | child_tidptr, NULL, trace); |
1da177e4 LT |
1706 | /* |
1707 | * Do this prior waking up the new thread - the thread pointer | |
1708 | * might get invalid after that point, if the thread exits quickly. | |
1709 | */ | |
1710 | if (!IS_ERR(p)) { | |
1711 | struct completion vfork; | |
4e52365f | 1712 | struct pid *pid; |
1da177e4 | 1713 | |
0a16b607 MD |
1714 | trace_sched_process_fork(current, p); |
1715 | ||
4e52365f MD |
1716 | pid = get_task_pid(p, PIDTYPE_PID); |
1717 | nr = pid_vnr(pid); | |
30e49c26 PE |
1718 | |
1719 | if (clone_flags & CLONE_PARENT_SETTID) | |
1720 | put_user(nr, parent_tidptr); | |
a6f5e063 | 1721 | |
1da177e4 LT |
1722 | if (clone_flags & CLONE_VFORK) { |
1723 | p->vfork_done = &vfork; | |
1724 | init_completion(&vfork); | |
d68b46fe | 1725 | get_task_struct(p); |
1da177e4 LT |
1726 | } |
1727 | ||
3e51e3ed | 1728 | wake_up_new_task(p); |
1da177e4 | 1729 | |
4b9d33e6 TH |
1730 | /* forking complete and child started to run, tell ptracer */ |
1731 | if (unlikely(trace)) | |
4e52365f | 1732 | ptrace_event_pid(trace, pid); |
09a05394 | 1733 | |
1da177e4 | 1734 | if (clone_flags & CLONE_VFORK) { |
d68b46fe | 1735 | if (!wait_for_vfork_done(p, &vfork)) |
4e52365f | 1736 | ptrace_event_pid(PTRACE_EVENT_VFORK_DONE, pid); |
1da177e4 | 1737 | } |
4e52365f MD |
1738 | |
1739 | put_pid(pid); | |
1da177e4 | 1740 | } else { |
92476d7f | 1741 | nr = PTR_ERR(p); |
1da177e4 | 1742 | } |
92476d7f | 1743 | return nr; |
1da177e4 LT |
1744 | } |
1745 | ||
2aa3a7f8 AV |
1746 | /* |
1747 | * Create a kernel thread. | |
1748 | */ | |
1749 | pid_t kernel_thread(int (*fn)(void *), void *arg, unsigned long flags) | |
1750 | { | |
e80d6661 | 1751 | return do_fork(flags|CLONE_VM|CLONE_UNTRACED, (unsigned long)fn, |
2aa3a7f8 AV |
1752 | (unsigned long)arg, NULL, NULL); |
1753 | } | |
2aa3a7f8 | 1754 | |
d2125043 AV |
1755 | #ifdef __ARCH_WANT_SYS_FORK |
1756 | SYSCALL_DEFINE0(fork) | |
1757 | { | |
1758 | #ifdef CONFIG_MMU | |
e80d6661 | 1759 | return do_fork(SIGCHLD, 0, 0, NULL, NULL); |
d2125043 AV |
1760 | #else |
1761 | /* can not support in nommu mode */ | |
5d59e182 | 1762 | return -EINVAL; |
d2125043 AV |
1763 | #endif |
1764 | } | |
1765 | #endif | |
1766 | ||
1767 | #ifdef __ARCH_WANT_SYS_VFORK | |
1768 | SYSCALL_DEFINE0(vfork) | |
1769 | { | |
5d59e182 | 1770 | return do_fork(CLONE_VFORK | CLONE_VM | SIGCHLD, 0, |
d2125043 AV |
1771 | 0, NULL, NULL); |
1772 | } | |
1773 | #endif | |
1774 | ||
1775 | #ifdef __ARCH_WANT_SYS_CLONE | |
1776 | #ifdef CONFIG_CLONE_BACKWARDS | |
1777 | SYSCALL_DEFINE5(clone, unsigned long, clone_flags, unsigned long, newsp, | |
1778 | int __user *, parent_tidptr, | |
1779 | int, tls_val, | |
1780 | int __user *, child_tidptr) | |
1781 | #elif defined(CONFIG_CLONE_BACKWARDS2) | |
1782 | SYSCALL_DEFINE5(clone, unsigned long, newsp, unsigned long, clone_flags, | |
1783 | int __user *, parent_tidptr, | |
1784 | int __user *, child_tidptr, | |
1785 | int, tls_val) | |
dfa9771a MS |
1786 | #elif defined(CONFIG_CLONE_BACKWARDS3) |
1787 | SYSCALL_DEFINE6(clone, unsigned long, clone_flags, unsigned long, newsp, | |
1788 | int, stack_size, | |
1789 | int __user *, parent_tidptr, | |
1790 | int __user *, child_tidptr, | |
1791 | int, tls_val) | |
d2125043 AV |
1792 | #else |
1793 | SYSCALL_DEFINE5(clone, unsigned long, clone_flags, unsigned long, newsp, | |
1794 | int __user *, parent_tidptr, | |
1795 | int __user *, child_tidptr, | |
1796 | int, tls_val) | |
1797 | #endif | |
1798 | { | |
2cf09666 | 1799 | return do_fork(clone_flags, newsp, 0, parent_tidptr, child_tidptr); |
d2125043 AV |
1800 | } |
1801 | #endif | |
1802 | ||
5fd63b30 RT |
1803 | #ifndef ARCH_MIN_MMSTRUCT_ALIGN |
1804 | #define ARCH_MIN_MMSTRUCT_ALIGN 0 | |
1805 | #endif | |
1806 | ||
51cc5068 | 1807 | static void sighand_ctor(void *data) |
aa1757f9 ON |
1808 | { |
1809 | struct sighand_struct *sighand = data; | |
1810 | ||
a35afb83 | 1811 | spin_lock_init(&sighand->siglock); |
b8fceee1 | 1812 | init_waitqueue_head(&sighand->signalfd_wqh); |
aa1757f9 ON |
1813 | } |
1814 | ||
1da177e4 LT |
1815 | void __init proc_caches_init(void) |
1816 | { | |
1817 | sighand_cachep = kmem_cache_create("sighand_cache", | |
1818 | sizeof(struct sighand_struct), 0, | |
2dff4405 VN |
1819 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_DESTROY_BY_RCU| |
1820 | SLAB_NOTRACK, sighand_ctor); | |
1da177e4 LT |
1821 | signal_cachep = kmem_cache_create("signal_cache", |
1822 | sizeof(struct signal_struct), 0, | |
2dff4405 | 1823 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL); |
20c2df83 | 1824 | files_cachep = kmem_cache_create("files_cache", |
1da177e4 | 1825 | sizeof(struct files_struct), 0, |
2dff4405 | 1826 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL); |
20c2df83 | 1827 | fs_cachep = kmem_cache_create("fs_cache", |
1da177e4 | 1828 | sizeof(struct fs_struct), 0, |
2dff4405 | 1829 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL); |
6345d24d LT |
1830 | /* |
1831 | * FIXME! The "sizeof(struct mm_struct)" currently includes the | |
1832 | * whole struct cpumask for the OFFSTACK case. We could change | |
1833 | * this to *only* allocate as much of it as required by the | |
1834 | * maximum number of CPU's we can ever have. The cpumask_allocation | |
1835 | * is at the end of the structure, exactly for that reason. | |
1836 | */ | |
1da177e4 | 1837 | mm_cachep = kmem_cache_create("mm_struct", |
5fd63b30 | 1838 | sizeof(struct mm_struct), ARCH_MIN_MMSTRUCT_ALIGN, |
2dff4405 | 1839 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL); |
33e5d769 | 1840 | vm_area_cachep = KMEM_CACHE(vm_area_struct, SLAB_PANIC); |
8feae131 | 1841 | mmap_init(); |
66577193 | 1842 | nsproxy_cache_init(); |
1da177e4 | 1843 | } |
cf2e340f | 1844 | |
cf2e340f | 1845 | /* |
9bfb23fc | 1846 | * Check constraints on flags passed to the unshare system call. |
cf2e340f | 1847 | */ |
9bfb23fc | 1848 | static int check_unshare_flags(unsigned long unshare_flags) |
cf2e340f | 1849 | { |
9bfb23fc ON |
1850 | if (unshare_flags & ~(CLONE_THREAD|CLONE_FS|CLONE_NEWNS|CLONE_SIGHAND| |
1851 | CLONE_VM|CLONE_FILES|CLONE_SYSVSEM| | |
50804fe3 | 1852 | CLONE_NEWUTS|CLONE_NEWIPC|CLONE_NEWNET| |
b2e0d987 | 1853 | CLONE_NEWUSER|CLONE_NEWPID)) |
9bfb23fc | 1854 | return -EINVAL; |
cf2e340f | 1855 | /* |
9bfb23fc ON |
1856 | * Not implemented, but pretend it works if there is nothing to |
1857 | * unshare. Note that unsharing CLONE_THREAD or CLONE_SIGHAND | |
1858 | * needs to unshare vm. | |
cf2e340f | 1859 | */ |
9bfb23fc ON |
1860 | if (unshare_flags & (CLONE_THREAD | CLONE_SIGHAND | CLONE_VM)) { |
1861 | /* FIXME: get_task_mm() increments ->mm_users */ | |
1862 | if (atomic_read(¤t->mm->mm_users) > 1) | |
1863 | return -EINVAL; | |
1864 | } | |
cf2e340f JD |
1865 | |
1866 | return 0; | |
1867 | } | |
1868 | ||
1869 | /* | |
99d1419d | 1870 | * Unshare the filesystem structure if it is being shared |
cf2e340f JD |
1871 | */ |
1872 | static int unshare_fs(unsigned long unshare_flags, struct fs_struct **new_fsp) | |
1873 | { | |
1874 | struct fs_struct *fs = current->fs; | |
1875 | ||
498052bb AV |
1876 | if (!(unshare_flags & CLONE_FS) || !fs) |
1877 | return 0; | |
1878 | ||
1879 | /* don't need lock here; in the worst case we'll do useless copy */ | |
1880 | if (fs->users == 1) | |
1881 | return 0; | |
1882 | ||
1883 | *new_fsp = copy_fs_struct(fs); | |
1884 | if (!*new_fsp) | |
1885 | return -ENOMEM; | |
cf2e340f JD |
1886 | |
1887 | return 0; | |
1888 | } | |
1889 | ||
cf2e340f | 1890 | /* |
a016f338 | 1891 | * Unshare file descriptor table if it is being shared |
cf2e340f JD |
1892 | */ |
1893 | static int unshare_fd(unsigned long unshare_flags, struct files_struct **new_fdp) | |
1894 | { | |
1895 | struct files_struct *fd = current->files; | |
a016f338 | 1896 | int error = 0; |
cf2e340f JD |
1897 | |
1898 | if ((unshare_flags & CLONE_FILES) && | |
a016f338 JD |
1899 | (fd && atomic_read(&fd->count) > 1)) { |
1900 | *new_fdp = dup_fd(fd, &error); | |
1901 | if (!*new_fdp) | |
1902 | return error; | |
1903 | } | |
cf2e340f JD |
1904 | |
1905 | return 0; | |
1906 | } | |
1907 | ||
cf2e340f JD |
1908 | /* |
1909 | * unshare allows a process to 'unshare' part of the process | |
1910 | * context which was originally shared using clone. copy_* | |
1911 | * functions used by do_fork() cannot be used here directly | |
1912 | * because they modify an inactive task_struct that is being | |
1913 | * constructed. Here we are modifying the current, active, | |
1914 | * task_struct. | |
1915 | */ | |
6559eed8 | 1916 | SYSCALL_DEFINE1(unshare, unsigned long, unshare_flags) |
cf2e340f | 1917 | { |
cf2e340f | 1918 | struct fs_struct *fs, *new_fs = NULL; |
cf2e340f | 1919 | struct files_struct *fd, *new_fd = NULL; |
b2e0d987 | 1920 | struct cred *new_cred = NULL; |
cf7b708c | 1921 | struct nsproxy *new_nsproxy = NULL; |
9edff4ab | 1922 | int do_sysvsem = 0; |
9bfb23fc | 1923 | int err; |
cf2e340f | 1924 | |
b2e0d987 EB |
1925 | /* |
1926 | * If unsharing a user namespace must also unshare the thread. | |
1927 | */ | |
1928 | if (unshare_flags & CLONE_NEWUSER) | |
e66eded8 | 1929 | unshare_flags |= CLONE_THREAD | CLONE_FS; |
50804fe3 EB |
1930 | /* |
1931 | * If unsharing a thread from a thread group, must also unshare vm. | |
1932 | */ | |
1933 | if (unshare_flags & CLONE_THREAD) | |
1934 | unshare_flags |= CLONE_VM; | |
1935 | /* | |
1936 | * If unsharing vm, must also unshare signal handlers. | |
1937 | */ | |
1938 | if (unshare_flags & CLONE_VM) | |
1939 | unshare_flags |= CLONE_SIGHAND; | |
9bfb23fc ON |
1940 | /* |
1941 | * If unsharing namespace, must also unshare filesystem information. | |
1942 | */ | |
1943 | if (unshare_flags & CLONE_NEWNS) | |
1944 | unshare_flags |= CLONE_FS; | |
50804fe3 EB |
1945 | |
1946 | err = check_unshare_flags(unshare_flags); | |
1947 | if (err) | |
1948 | goto bad_unshare_out; | |
6013f67f MS |
1949 | /* |
1950 | * CLONE_NEWIPC must also detach from the undolist: after switching | |
1951 | * to a new ipc namespace, the semaphore arrays from the old | |
1952 | * namespace are unreachable. | |
1953 | */ | |
1954 | if (unshare_flags & (CLONE_NEWIPC|CLONE_SYSVSEM)) | |
9edff4ab | 1955 | do_sysvsem = 1; |
fb0a685c DRO |
1956 | err = unshare_fs(unshare_flags, &new_fs); |
1957 | if (err) | |
9bfb23fc | 1958 | goto bad_unshare_out; |
fb0a685c DRO |
1959 | err = unshare_fd(unshare_flags, &new_fd); |
1960 | if (err) | |
9bfb23fc | 1961 | goto bad_unshare_cleanup_fs; |
b2e0d987 | 1962 | err = unshare_userns(unshare_flags, &new_cred); |
fb0a685c | 1963 | if (err) |
9edff4ab | 1964 | goto bad_unshare_cleanup_fd; |
b2e0d987 EB |
1965 | err = unshare_nsproxy_namespaces(unshare_flags, &new_nsproxy, |
1966 | new_cred, new_fs); | |
1967 | if (err) | |
1968 | goto bad_unshare_cleanup_cred; | |
c0b2fc31 | 1969 | |
b2e0d987 | 1970 | if (new_fs || new_fd || do_sysvsem || new_cred || new_nsproxy) { |
9edff4ab MS |
1971 | if (do_sysvsem) { |
1972 | /* | |
1973 | * CLONE_SYSVSEM is equivalent to sys_exit(). | |
1974 | */ | |
1975 | exit_sem(current); | |
1976 | } | |
ab602f79 JM |
1977 | if (unshare_flags & CLONE_NEWIPC) { |
1978 | /* Orphan segments in old ns (see sem above). */ | |
1979 | exit_shm(current); | |
1980 | shm_init_task(current); | |
1981 | } | |
ab516013 | 1982 | |
6f977e6b | 1983 | if (new_nsproxy) |
cf7b708c | 1984 | switch_task_namespaces(current, new_nsproxy); |
cf2e340f | 1985 | |
cf7b708c PE |
1986 | task_lock(current); |
1987 | ||
cf2e340f JD |
1988 | if (new_fs) { |
1989 | fs = current->fs; | |
2a4419b5 | 1990 | spin_lock(&fs->lock); |
cf2e340f | 1991 | current->fs = new_fs; |
498052bb AV |
1992 | if (--fs->users) |
1993 | new_fs = NULL; | |
1994 | else | |
1995 | new_fs = fs; | |
2a4419b5 | 1996 | spin_unlock(&fs->lock); |
cf2e340f JD |
1997 | } |
1998 | ||
cf2e340f JD |
1999 | if (new_fd) { |
2000 | fd = current->files; | |
2001 | current->files = new_fd; | |
2002 | new_fd = fd; | |
2003 | } | |
2004 | ||
2005 | task_unlock(current); | |
b2e0d987 EB |
2006 | |
2007 | if (new_cred) { | |
2008 | /* Install the new user namespace */ | |
2009 | commit_creds(new_cred); | |
2010 | new_cred = NULL; | |
2011 | } | |
cf2e340f JD |
2012 | } |
2013 | ||
b2e0d987 EB |
2014 | bad_unshare_cleanup_cred: |
2015 | if (new_cred) | |
2016 | put_cred(new_cred); | |
cf2e340f JD |
2017 | bad_unshare_cleanup_fd: |
2018 | if (new_fd) | |
2019 | put_files_struct(new_fd); | |
2020 | ||
cf2e340f JD |
2021 | bad_unshare_cleanup_fs: |
2022 | if (new_fs) | |
498052bb | 2023 | free_fs_struct(new_fs); |
cf2e340f | 2024 | |
cf2e340f JD |
2025 | bad_unshare_out: |
2026 | return err; | |
2027 | } | |
3b125388 AV |
2028 | |
2029 | /* | |
2030 | * Helper to unshare the files of the current task. | |
2031 | * We don't want to expose copy_files internals to | |
2032 | * the exec layer of the kernel. | |
2033 | */ | |
2034 | ||
2035 | int unshare_files(struct files_struct **displaced) | |
2036 | { | |
2037 | struct task_struct *task = current; | |
50704516 | 2038 | struct files_struct *copy = NULL; |
3b125388 AV |
2039 | int error; |
2040 | ||
2041 | error = unshare_fd(CLONE_FILES, ©); | |
2042 | if (error || !copy) { | |
2043 | *displaced = NULL; | |
2044 | return error; | |
2045 | } | |
2046 | *displaced = task->files; | |
2047 | task_lock(task); | |
2048 | task->files = copy; | |
2049 | task_unlock(task); | |
2050 | return 0; | |
2051 | } | |
16db3d3f HS |
2052 | |
2053 | int sysctl_max_threads(struct ctl_table *table, int write, | |
2054 | void __user *buffer, size_t *lenp, loff_t *ppos) | |
2055 | { | |
2056 | struct ctl_table t; | |
2057 | int ret; | |
2058 | int threads = max_threads; | |
2059 | int min = MIN_THREADS; | |
2060 | int max = MAX_THREADS; | |
2061 | ||
2062 | t = *table; | |
2063 | t.data = &threads; | |
2064 | t.extra1 = &min; | |
2065 | t.extra2 = &max; | |
2066 | ||
2067 | ret = proc_dointvec_minmax(&t, write, buffer, lenp, ppos); | |
2068 | if (ret || !write) | |
2069 | return ret; | |
2070 | ||
2071 | set_max_threads(threads); | |
2072 | ||
2073 | return 0; | |
2074 | } |