]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * linux/kernel/fork.c | |
3 | * | |
4 | * Copyright (C) 1991, 1992 Linus Torvalds | |
5 | */ | |
6 | ||
7 | /* | |
8 | * 'fork.c' contains the help-routines for the 'fork' system call | |
9 | * (see also entry.S and others). | |
10 | * Fork is rather simple, once you get the hang of it, but the memory | |
11 | * management can be a bitch. See 'mm/memory.c': 'copy_page_range()' | |
12 | */ | |
13 | ||
1da177e4 LT |
14 | #include <linux/slab.h> |
15 | #include <linux/init.h> | |
16 | #include <linux/unistd.h> | |
1da177e4 LT |
17 | #include <linux/module.h> |
18 | #include <linux/vmalloc.h> | |
19 | #include <linux/completion.h> | |
1da177e4 LT |
20 | #include <linux/personality.h> |
21 | #include <linux/mempolicy.h> | |
22 | #include <linux/sem.h> | |
23 | #include <linux/file.h> | |
9f3acc31 | 24 | #include <linux/fdtable.h> |
da9cbc87 | 25 | #include <linux/iocontext.h> |
1da177e4 LT |
26 | #include <linux/key.h> |
27 | #include <linux/binfmts.h> | |
28 | #include <linux/mman.h> | |
cddb8a5c | 29 | #include <linux/mmu_notifier.h> |
1da177e4 | 30 | #include <linux/fs.h> |
615d6e87 DB |
31 | #include <linux/mm.h> |
32 | #include <linux/vmacache.h> | |
ab516013 | 33 | #include <linux/nsproxy.h> |
c59ede7b | 34 | #include <linux/capability.h> |
1da177e4 | 35 | #include <linux/cpu.h> |
b4f48b63 | 36 | #include <linux/cgroup.h> |
1da177e4 | 37 | #include <linux/security.h> |
a1e78772 | 38 | #include <linux/hugetlb.h> |
e2cfabdf | 39 | #include <linux/seccomp.h> |
1da177e4 LT |
40 | #include <linux/swap.h> |
41 | #include <linux/syscalls.h> | |
42 | #include <linux/jiffies.h> | |
43 | #include <linux/futex.h> | |
8141c7f3 | 44 | #include <linux/compat.h> |
207205a2 | 45 | #include <linux/kthread.h> |
7c3ab738 | 46 | #include <linux/task_io_accounting_ops.h> |
ab2af1f5 | 47 | #include <linux/rcupdate.h> |
1da177e4 LT |
48 | #include <linux/ptrace.h> |
49 | #include <linux/mount.h> | |
50 | #include <linux/audit.h> | |
78fb7466 | 51 | #include <linux/memcontrol.h> |
f201ae23 | 52 | #include <linux/ftrace.h> |
5e2bf014 | 53 | #include <linux/proc_fs.h> |
1da177e4 LT |
54 | #include <linux/profile.h> |
55 | #include <linux/rmap.h> | |
f8af4da3 | 56 | #include <linux/ksm.h> |
1da177e4 | 57 | #include <linux/acct.h> |
8f0ab514 | 58 | #include <linux/tsacct_kern.h> |
9f46080c | 59 | #include <linux/cn_proc.h> |
ba96a0c8 | 60 | #include <linux/freezer.h> |
ca74e92b | 61 | #include <linux/delayacct.h> |
ad4ecbcb | 62 | #include <linux/taskstats_kern.h> |
0a425405 | 63 | #include <linux/random.h> |
522ed776 | 64 | #include <linux/tty.h> |
fd0928df | 65 | #include <linux/blkdev.h> |
5ad4e53b | 66 | #include <linux/fs_struct.h> |
7c9f8861 | 67 | #include <linux/magic.h> |
cdd6c482 | 68 | #include <linux/perf_event.h> |
42c4ab41 | 69 | #include <linux/posix-timers.h> |
8e7cac79 | 70 | #include <linux/user-return-notifier.h> |
3d5992d2 | 71 | #include <linux/oom.h> |
ba76149f | 72 | #include <linux/khugepaged.h> |
d80e731e | 73 | #include <linux/signalfd.h> |
0326f5a9 | 74 | #include <linux/uprobes.h> |
a27bb332 | 75 | #include <linux/aio.h> |
52f5684c | 76 | #include <linux/compiler.h> |
1da177e4 LT |
77 | |
78 | #include <asm/pgtable.h> | |
79 | #include <asm/pgalloc.h> | |
80 | #include <asm/uaccess.h> | |
81 | #include <asm/mmu_context.h> | |
82 | #include <asm/cacheflush.h> | |
83 | #include <asm/tlbflush.h> | |
84 | ||
ad8d75ff SR |
85 | #include <trace/events/sched.h> |
86 | ||
43d2b113 KH |
87 | #define CREATE_TRACE_POINTS |
88 | #include <trace/events/task.h> | |
89 | ||
1da177e4 LT |
90 | /* |
91 | * Protected counters by write_lock_irq(&tasklist_lock) | |
92 | */ | |
93 | unsigned long total_forks; /* Handle normal Linux uptimes. */ | |
fb0a685c | 94 | int nr_threads; /* The idle threads do not count.. */ |
1da177e4 LT |
95 | |
96 | int max_threads; /* tunable limit on nr_threads */ | |
97 | ||
98 | DEFINE_PER_CPU(unsigned long, process_counts) = 0; | |
99 | ||
c59923a1 | 100 | __cacheline_aligned DEFINE_RWLOCK(tasklist_lock); /* outer */ |
db1466b3 PM |
101 | |
102 | #ifdef CONFIG_PROVE_RCU | |
103 | int lockdep_tasklist_lock_is_held(void) | |
104 | { | |
105 | return lockdep_is_held(&tasklist_lock); | |
106 | } | |
107 | EXPORT_SYMBOL_GPL(lockdep_tasklist_lock_is_held); | |
108 | #endif /* #ifdef CONFIG_PROVE_RCU */ | |
1da177e4 LT |
109 | |
110 | int nr_processes(void) | |
111 | { | |
112 | int cpu; | |
113 | int total = 0; | |
114 | ||
1d510750 | 115 | for_each_possible_cpu(cpu) |
1da177e4 LT |
116 | total += per_cpu(process_counts, cpu); |
117 | ||
118 | return total; | |
119 | } | |
120 | ||
f19b9f74 AM |
121 | void __weak arch_release_task_struct(struct task_struct *tsk) |
122 | { | |
123 | } | |
124 | ||
f5e10287 | 125 | #ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR |
e18b890b | 126 | static struct kmem_cache *task_struct_cachep; |
41101809 TG |
127 | |
128 | static inline struct task_struct *alloc_task_struct_node(int node) | |
129 | { | |
130 | return kmem_cache_alloc_node(task_struct_cachep, GFP_KERNEL, node); | |
131 | } | |
132 | ||
41101809 TG |
133 | static inline void free_task_struct(struct task_struct *tsk) |
134 | { | |
41101809 TG |
135 | kmem_cache_free(task_struct_cachep, tsk); |
136 | } | |
1da177e4 LT |
137 | #endif |
138 | ||
f19b9f74 AM |
139 | void __weak arch_release_thread_info(struct thread_info *ti) |
140 | { | |
141 | } | |
142 | ||
f5e10287 | 143 | #ifndef CONFIG_ARCH_THREAD_INFO_ALLOCATOR |
41101809 | 144 | |
0d15d74a TG |
145 | /* |
146 | * Allocate pages if THREAD_SIZE is >= PAGE_SIZE, otherwise use a | |
147 | * kmemcache based allocator. | |
148 | */ | |
149 | # if THREAD_SIZE >= PAGE_SIZE | |
b6a84016 ED |
150 | static struct thread_info *alloc_thread_info_node(struct task_struct *tsk, |
151 | int node) | |
b69c49b7 | 152 | { |
52383431 VD |
153 | struct page *page = alloc_kmem_pages_node(node, THREADINFO_GFP, |
154 | THREAD_SIZE_ORDER); | |
b6a84016 ED |
155 | |
156 | return page ? page_address(page) : NULL; | |
b69c49b7 FT |
157 | } |
158 | ||
159 | static inline void free_thread_info(struct thread_info *ti) | |
160 | { | |
52383431 | 161 | free_kmem_pages((unsigned long)ti, THREAD_SIZE_ORDER); |
b69c49b7 | 162 | } |
0d15d74a TG |
163 | # else |
164 | static struct kmem_cache *thread_info_cache; | |
165 | ||
166 | static struct thread_info *alloc_thread_info_node(struct task_struct *tsk, | |
167 | int node) | |
168 | { | |
169 | return kmem_cache_alloc_node(thread_info_cache, THREADINFO_GFP, node); | |
170 | } | |
171 | ||
172 | static void free_thread_info(struct thread_info *ti) | |
173 | { | |
0d15d74a TG |
174 | kmem_cache_free(thread_info_cache, ti); |
175 | } | |
176 | ||
177 | void thread_info_cache_init(void) | |
178 | { | |
179 | thread_info_cache = kmem_cache_create("thread_info", THREAD_SIZE, | |
180 | THREAD_SIZE, 0, NULL); | |
181 | BUG_ON(thread_info_cache == NULL); | |
182 | } | |
183 | # endif | |
b69c49b7 FT |
184 | #endif |
185 | ||
1da177e4 | 186 | /* SLAB cache for signal_struct structures (tsk->signal) */ |
e18b890b | 187 | static struct kmem_cache *signal_cachep; |
1da177e4 LT |
188 | |
189 | /* SLAB cache for sighand_struct structures (tsk->sighand) */ | |
e18b890b | 190 | struct kmem_cache *sighand_cachep; |
1da177e4 LT |
191 | |
192 | /* SLAB cache for files_struct structures (tsk->files) */ | |
e18b890b | 193 | struct kmem_cache *files_cachep; |
1da177e4 LT |
194 | |
195 | /* SLAB cache for fs_struct structures (tsk->fs) */ | |
e18b890b | 196 | struct kmem_cache *fs_cachep; |
1da177e4 LT |
197 | |
198 | /* SLAB cache for vm_area_struct structures */ | |
e18b890b | 199 | struct kmem_cache *vm_area_cachep; |
1da177e4 LT |
200 | |
201 | /* SLAB cache for mm_struct structures (tsk->mm) */ | |
e18b890b | 202 | static struct kmem_cache *mm_cachep; |
1da177e4 | 203 | |
c6a7f572 KM |
204 | static void account_kernel_stack(struct thread_info *ti, int account) |
205 | { | |
206 | struct zone *zone = page_zone(virt_to_page(ti)); | |
207 | ||
208 | mod_zone_page_state(zone, NR_KERNEL_STACK, account); | |
209 | } | |
210 | ||
1da177e4 LT |
211 | void free_task(struct task_struct *tsk) |
212 | { | |
c6a7f572 | 213 | account_kernel_stack(tsk->stack, -1); |
f19b9f74 | 214 | arch_release_thread_info(tsk->stack); |
f7e4217b | 215 | free_thread_info(tsk->stack); |
23f78d4a | 216 | rt_mutex_debug_task_free(tsk); |
fb52607a | 217 | ftrace_graph_exit_task(tsk); |
e2cfabdf | 218 | put_seccomp_filter(tsk); |
f19b9f74 | 219 | arch_release_task_struct(tsk); |
1da177e4 LT |
220 | free_task_struct(tsk); |
221 | } | |
222 | EXPORT_SYMBOL(free_task); | |
223 | ||
ea6d290c ON |
224 | static inline void free_signal_struct(struct signal_struct *sig) |
225 | { | |
97101eb4 | 226 | taskstats_tgid_free(sig); |
1c5354de | 227 | sched_autogroup_exit(sig); |
ea6d290c ON |
228 | kmem_cache_free(signal_cachep, sig); |
229 | } | |
230 | ||
231 | static inline void put_signal_struct(struct signal_struct *sig) | |
232 | { | |
1c5354de | 233 | if (atomic_dec_and_test(&sig->sigcnt)) |
ea6d290c ON |
234 | free_signal_struct(sig); |
235 | } | |
236 | ||
158d9ebd | 237 | void __put_task_struct(struct task_struct *tsk) |
1da177e4 | 238 | { |
270f722d | 239 | WARN_ON(!tsk->exit_state); |
1da177e4 LT |
240 | WARN_ON(atomic_read(&tsk->usage)); |
241 | WARN_ON(tsk == current); | |
242 | ||
156654f4 | 243 | task_numa_free(tsk); |
1a2a4d06 | 244 | security_task_free(tsk); |
e0e81739 | 245 | exit_creds(tsk); |
35df17c5 | 246 | delayacct_tsk_free(tsk); |
ea6d290c | 247 | put_signal_struct(tsk->signal); |
1da177e4 LT |
248 | |
249 | if (!profile_handoff_task(tsk)) | |
250 | free_task(tsk); | |
251 | } | |
77c100c8 | 252 | EXPORT_SYMBOL_GPL(__put_task_struct); |
1da177e4 | 253 | |
6c0a9fa6 | 254 | void __init __weak arch_task_cache_init(void) { } |
61c4628b | 255 | |
1da177e4 LT |
256 | void __init fork_init(unsigned long mempages) |
257 | { | |
f5e10287 | 258 | #ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR |
1da177e4 LT |
259 | #ifndef ARCH_MIN_TASKALIGN |
260 | #define ARCH_MIN_TASKALIGN L1_CACHE_BYTES | |
261 | #endif | |
262 | /* create a slab on which task_structs can be allocated */ | |
263 | task_struct_cachep = | |
264 | kmem_cache_create("task_struct", sizeof(struct task_struct), | |
2dff4405 | 265 | ARCH_MIN_TASKALIGN, SLAB_PANIC | SLAB_NOTRACK, NULL); |
1da177e4 LT |
266 | #endif |
267 | ||
61c4628b SS |
268 | /* do the arch specific task caches init */ |
269 | arch_task_cache_init(); | |
270 | ||
1da177e4 LT |
271 | /* |
272 | * The default maximum number of threads is set to a safe | |
273 | * value: the thread structures can take up at most half | |
274 | * of memory. | |
275 | */ | |
276 | max_threads = mempages / (8 * THREAD_SIZE / PAGE_SIZE); | |
277 | ||
278 | /* | |
279 | * we need to allow at least 20 threads to boot a system | |
280 | */ | |
fb0a685c | 281 | if (max_threads < 20) |
1da177e4 LT |
282 | max_threads = 20; |
283 | ||
284 | init_task.signal->rlim[RLIMIT_NPROC].rlim_cur = max_threads/2; | |
285 | init_task.signal->rlim[RLIMIT_NPROC].rlim_max = max_threads/2; | |
286 | init_task.signal->rlim[RLIMIT_SIGPENDING] = | |
287 | init_task.signal->rlim[RLIMIT_NPROC]; | |
288 | } | |
289 | ||
52f5684c | 290 | int __weak arch_dup_task_struct(struct task_struct *dst, |
61c4628b SS |
291 | struct task_struct *src) |
292 | { | |
293 | *dst = *src; | |
294 | return 0; | |
295 | } | |
296 | ||
1da177e4 LT |
297 | static struct task_struct *dup_task_struct(struct task_struct *orig) |
298 | { | |
299 | struct task_struct *tsk; | |
300 | struct thread_info *ti; | |
7c9f8861 | 301 | unsigned long *stackend; |
207205a2 | 302 | int node = tsk_fork_get_node(orig); |
3e26c149 | 303 | int err; |
1da177e4 | 304 | |
504f52b5 | 305 | tsk = alloc_task_struct_node(node); |
1da177e4 LT |
306 | if (!tsk) |
307 | return NULL; | |
308 | ||
b6a84016 | 309 | ti = alloc_thread_info_node(tsk, node); |
f19b9f74 AM |
310 | if (!ti) |
311 | goto free_tsk; | |
1da177e4 | 312 | |
fb0a685c | 313 | err = arch_dup_task_struct(tsk, orig); |
164c33c6 | 314 | if (err) |
f19b9f74 | 315 | goto free_ti; |
164c33c6 | 316 | |
87bec58a | 317 | tsk->stack = ti; |
dbd95212 KC |
318 | #ifdef CONFIG_SECCOMP |
319 | /* | |
320 | * We must handle setting up seccomp filters once we're under | |
321 | * the sighand lock in case orig has changed between now and | |
322 | * then. Until then, filter must be NULL to avoid messing up | |
323 | * the usage counts on the error path calling free_task. | |
324 | */ | |
325 | tsk->seccomp.filter = NULL; | |
326 | #endif | |
87bec58a AM |
327 | |
328 | setup_thread_stack(tsk, orig); | |
8e7cac79 | 329 | clear_user_return_notifier(tsk); |
f26f9aff | 330 | clear_tsk_need_resched(tsk); |
7c9f8861 ES |
331 | stackend = end_of_stack(tsk); |
332 | *stackend = STACK_END_MAGIC; /* for overflow detection */ | |
1da177e4 | 333 | |
0a425405 AV |
334 | #ifdef CONFIG_CC_STACKPROTECTOR |
335 | tsk->stack_canary = get_random_int(); | |
336 | #endif | |
337 | ||
fb0a685c DRO |
338 | /* |
339 | * One for us, one for whoever does the "release_task()" (usually | |
340 | * parent) | |
341 | */ | |
342 | atomic_set(&tsk->usage, 2); | |
6c5c9341 | 343 | #ifdef CONFIG_BLK_DEV_IO_TRACE |
2056a782 | 344 | tsk->btrace_seq = 0; |
6c5c9341 | 345 | #endif |
a0aa7f68 | 346 | tsk->splice_pipe = NULL; |
5640f768 | 347 | tsk->task_frag.page = NULL; |
c6a7f572 KM |
348 | |
349 | account_kernel_stack(ti, 1); | |
350 | ||
1da177e4 | 351 | return tsk; |
61c4628b | 352 | |
f19b9f74 | 353 | free_ti: |
61c4628b | 354 | free_thread_info(ti); |
f19b9f74 | 355 | free_tsk: |
61c4628b SS |
356 | free_task_struct(tsk); |
357 | return NULL; | |
1da177e4 LT |
358 | } |
359 | ||
360 | #ifdef CONFIG_MMU | |
a39bc516 | 361 | static int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm) |
1da177e4 | 362 | { |
297c5eee | 363 | struct vm_area_struct *mpnt, *tmp, *prev, **pprev; |
1da177e4 LT |
364 | struct rb_node **rb_link, *rb_parent; |
365 | int retval; | |
366 | unsigned long charge; | |
1da177e4 | 367 | |
32cdba1e | 368 | uprobe_start_dup_mmap(); |
1da177e4 | 369 | down_write(&oldmm->mmap_sem); |
ec8c0446 | 370 | flush_cache_dup_mm(oldmm); |
f8ac4ec9 | 371 | uprobe_dup_mmap(oldmm, mm); |
ad339451 IM |
372 | /* |
373 | * Not linked in yet - no deadlock potential: | |
374 | */ | |
375 | down_write_nested(&mm->mmap_sem, SINGLE_DEPTH_NESTING); | |
7ee78232 | 376 | |
4f7d4614 VD |
377 | mm->total_vm = oldmm->total_vm; |
378 | mm->shared_vm = oldmm->shared_vm; | |
379 | mm->exec_vm = oldmm->exec_vm; | |
380 | mm->stack_vm = oldmm->stack_vm; | |
381 | ||
1da177e4 LT |
382 | rb_link = &mm->mm_rb.rb_node; |
383 | rb_parent = NULL; | |
384 | pprev = &mm->mmap; | |
f8af4da3 | 385 | retval = ksm_fork(mm, oldmm); |
ba76149f AA |
386 | if (retval) |
387 | goto out; | |
388 | retval = khugepaged_fork(mm, oldmm); | |
f8af4da3 HD |
389 | if (retval) |
390 | goto out; | |
1da177e4 | 391 | |
297c5eee | 392 | prev = NULL; |
fd3e42fc | 393 | for (mpnt = oldmm->mmap; mpnt; mpnt = mpnt->vm_next) { |
1da177e4 LT |
394 | struct file *file; |
395 | ||
396 | if (mpnt->vm_flags & VM_DONTCOPY) { | |
ab50b8ed | 397 | vm_stat_account(mm, mpnt->vm_flags, mpnt->vm_file, |
44de9d0c | 398 | -vma_pages(mpnt)); |
1da177e4 LT |
399 | continue; |
400 | } | |
401 | charge = 0; | |
402 | if (mpnt->vm_flags & VM_ACCOUNT) { | |
b2412b7f HS |
403 | unsigned long len = vma_pages(mpnt); |
404 | ||
191c5424 | 405 | if (security_vm_enough_memory_mm(oldmm, len)) /* sic */ |
1da177e4 LT |
406 | goto fail_nomem; |
407 | charge = len; | |
408 | } | |
e94b1766 | 409 | tmp = kmem_cache_alloc(vm_area_cachep, GFP_KERNEL); |
1da177e4 LT |
410 | if (!tmp) |
411 | goto fail_nomem; | |
412 | *tmp = *mpnt; | |
5beb4930 | 413 | INIT_LIST_HEAD(&tmp->anon_vma_chain); |
ef0855d3 ON |
414 | retval = vma_dup_policy(mpnt, tmp); |
415 | if (retval) | |
1da177e4 | 416 | goto fail_nomem_policy; |
a247c3a9 | 417 | tmp->vm_mm = mm; |
5beb4930 RR |
418 | if (anon_vma_fork(tmp, mpnt)) |
419 | goto fail_nomem_anon_vma_fork; | |
1da177e4 | 420 | tmp->vm_flags &= ~VM_LOCKED; |
297c5eee | 421 | tmp->vm_next = tmp->vm_prev = NULL; |
1da177e4 LT |
422 | file = tmp->vm_file; |
423 | if (file) { | |
496ad9aa | 424 | struct inode *inode = file_inode(file); |
b88ed205 HD |
425 | struct address_space *mapping = file->f_mapping; |
426 | ||
1da177e4 LT |
427 | get_file(file); |
428 | if (tmp->vm_flags & VM_DENYWRITE) | |
429 | atomic_dec(&inode->i_writecount); | |
3d48ae45 | 430 | mutex_lock(&mapping->i_mmap_mutex); |
b88ed205 | 431 | if (tmp->vm_flags & VM_SHARED) |
4bb5f5d9 | 432 | atomic_inc(&mapping->i_mmap_writable); |
b88ed205 HD |
433 | flush_dcache_mmap_lock(mapping); |
434 | /* insert tmp into the share list, just after mpnt */ | |
9826a516 ML |
435 | if (unlikely(tmp->vm_flags & VM_NONLINEAR)) |
436 | vma_nonlinear_insert(tmp, | |
437 | &mapping->i_mmap_nonlinear); | |
438 | else | |
439 | vma_interval_tree_insert_after(tmp, mpnt, | |
440 | &mapping->i_mmap); | |
b88ed205 | 441 | flush_dcache_mmap_unlock(mapping); |
3d48ae45 | 442 | mutex_unlock(&mapping->i_mmap_mutex); |
1da177e4 LT |
443 | } |
444 | ||
a1e78772 MG |
445 | /* |
446 | * Clear hugetlb-related page reserves for children. This only | |
447 | * affects MAP_PRIVATE mappings. Faults generated by the child | |
448 | * are not guaranteed to succeed, even if read-only | |
449 | */ | |
450 | if (is_vm_hugetlb_page(tmp)) | |
451 | reset_vma_resv_huge_pages(tmp); | |
452 | ||
1da177e4 | 453 | /* |
7ee78232 | 454 | * Link in the new vma and copy the page table entries. |
1da177e4 | 455 | */ |
1da177e4 LT |
456 | *pprev = tmp; |
457 | pprev = &tmp->vm_next; | |
297c5eee LT |
458 | tmp->vm_prev = prev; |
459 | prev = tmp; | |
1da177e4 LT |
460 | |
461 | __vma_link_rb(mm, tmp, rb_link, rb_parent); | |
462 | rb_link = &tmp->vm_rb.rb_right; | |
463 | rb_parent = &tmp->vm_rb; | |
464 | ||
465 | mm->map_count++; | |
0b0db14c | 466 | retval = copy_page_range(mm, oldmm, mpnt); |
1da177e4 LT |
467 | |
468 | if (tmp->vm_ops && tmp->vm_ops->open) | |
469 | tmp->vm_ops->open(tmp); | |
470 | ||
471 | if (retval) | |
472 | goto out; | |
473 | } | |
d6dd61c8 JF |
474 | /* a new mm has just been created */ |
475 | arch_dup_mmap(oldmm, mm); | |
1da177e4 | 476 | retval = 0; |
1da177e4 | 477 | out: |
7ee78232 | 478 | up_write(&mm->mmap_sem); |
fd3e42fc | 479 | flush_tlb_mm(oldmm); |
1da177e4 | 480 | up_write(&oldmm->mmap_sem); |
32cdba1e | 481 | uprobe_end_dup_mmap(); |
1da177e4 | 482 | return retval; |
5beb4930 | 483 | fail_nomem_anon_vma_fork: |
ef0855d3 | 484 | mpol_put(vma_policy(tmp)); |
1da177e4 LT |
485 | fail_nomem_policy: |
486 | kmem_cache_free(vm_area_cachep, tmp); | |
487 | fail_nomem: | |
488 | retval = -ENOMEM; | |
489 | vm_unacct_memory(charge); | |
490 | goto out; | |
491 | } | |
492 | ||
fb0a685c | 493 | static inline int mm_alloc_pgd(struct mm_struct *mm) |
1da177e4 LT |
494 | { |
495 | mm->pgd = pgd_alloc(mm); | |
496 | if (unlikely(!mm->pgd)) | |
497 | return -ENOMEM; | |
498 | return 0; | |
499 | } | |
500 | ||
fb0a685c | 501 | static inline void mm_free_pgd(struct mm_struct *mm) |
1da177e4 | 502 | { |
5e541973 | 503 | pgd_free(mm, mm->pgd); |
1da177e4 LT |
504 | } |
505 | #else | |
506 | #define dup_mmap(mm, oldmm) (0) | |
507 | #define mm_alloc_pgd(mm) (0) | |
508 | #define mm_free_pgd(mm) | |
509 | #endif /* CONFIG_MMU */ | |
510 | ||
23ff4440 | 511 | __cacheline_aligned_in_smp DEFINE_SPINLOCK(mmlist_lock); |
1da177e4 | 512 | |
e94b1766 | 513 | #define allocate_mm() (kmem_cache_alloc(mm_cachep, GFP_KERNEL)) |
1da177e4 LT |
514 | #define free_mm(mm) (kmem_cache_free(mm_cachep, (mm))) |
515 | ||
4cb0e11b HK |
516 | static unsigned long default_dump_filter = MMF_DUMP_FILTER_DEFAULT; |
517 | ||
518 | static int __init coredump_filter_setup(char *s) | |
519 | { | |
520 | default_dump_filter = | |
521 | (simple_strtoul(s, NULL, 0) << MMF_DUMP_FILTER_SHIFT) & | |
522 | MMF_DUMP_FILTER_MASK; | |
523 | return 1; | |
524 | } | |
525 | ||
526 | __setup("coredump_filter=", coredump_filter_setup); | |
527 | ||
1da177e4 LT |
528 | #include <linux/init_task.h> |
529 | ||
858f0993 AD |
530 | static void mm_init_aio(struct mm_struct *mm) |
531 | { | |
532 | #ifdef CONFIG_AIO | |
533 | spin_lock_init(&mm->ioctx_lock); | |
db446a08 | 534 | mm->ioctx_table = NULL; |
858f0993 AD |
535 | #endif |
536 | } | |
537 | ||
33144e84 VD |
538 | static void mm_init_owner(struct mm_struct *mm, struct task_struct *p) |
539 | { | |
540 | #ifdef CONFIG_MEMCG | |
541 | mm->owner = p; | |
542 | #endif | |
543 | } | |
544 | ||
fb0a685c | 545 | static struct mm_struct *mm_init(struct mm_struct *mm, struct task_struct *p) |
1da177e4 | 546 | { |
41f727fd VD |
547 | mm->mmap = NULL; |
548 | mm->mm_rb = RB_ROOT; | |
549 | mm->vmacache_seqnum = 0; | |
1da177e4 LT |
550 | atomic_set(&mm->mm_users, 1); |
551 | atomic_set(&mm->mm_count, 1); | |
552 | init_rwsem(&mm->mmap_sem); | |
553 | INIT_LIST_HEAD(&mm->mmlist); | |
999d9fc1 | 554 | mm->core_state = NULL; |
e1f56c89 | 555 | atomic_long_set(&mm->nr_ptes, 0); |
41f727fd VD |
556 | mm->map_count = 0; |
557 | mm->locked_vm = 0; | |
ce65cefa | 558 | mm->pinned_vm = 0; |
d559db08 | 559 | memset(&mm->rss_stat, 0, sizeof(mm->rss_stat)); |
1da177e4 | 560 | spin_lock_init(&mm->page_table_lock); |
41f727fd | 561 | mm_init_cpumask(mm); |
858f0993 | 562 | mm_init_aio(mm); |
cf475ad2 | 563 | mm_init_owner(mm, p); |
41f727fd | 564 | mmu_notifier_mm_init(mm); |
20841405 | 565 | clear_tlb_flush_pending(mm); |
41f727fd VD |
566 | #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS |
567 | mm->pmd_huge_pte = NULL; | |
568 | #endif | |
1da177e4 | 569 | |
a0715cc2 AT |
570 | if (current->mm) { |
571 | mm->flags = current->mm->flags & MMF_INIT_MASK; | |
572 | mm->def_flags = current->mm->def_flags & VM_INIT_DEF_MASK; | |
573 | } else { | |
574 | mm->flags = default_dump_filter; | |
1da177e4 | 575 | mm->def_flags = 0; |
a0715cc2 AT |
576 | } |
577 | ||
41f727fd VD |
578 | if (mm_alloc_pgd(mm)) |
579 | goto fail_nopgd; | |
580 | ||
581 | if (init_new_context(p, mm)) | |
582 | goto fail_nocontext; | |
78fb7466 | 583 | |
41f727fd VD |
584 | return mm; |
585 | ||
586 | fail_nocontext: | |
587 | mm_free_pgd(mm); | |
588 | fail_nopgd: | |
1da177e4 LT |
589 | free_mm(mm); |
590 | return NULL; | |
591 | } | |
592 | ||
c3f0327f KK |
593 | static void check_mm(struct mm_struct *mm) |
594 | { | |
595 | int i; | |
596 | ||
597 | for (i = 0; i < NR_MM_COUNTERS; i++) { | |
598 | long x = atomic_long_read(&mm->rss_stat.count[i]); | |
599 | ||
600 | if (unlikely(x)) | |
601 | printk(KERN_ALERT "BUG: Bad rss-counter state " | |
602 | "mm:%p idx:%d val:%ld\n", mm, i, x); | |
603 | } | |
e009bb30 | 604 | #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS |
96dad67f | 605 | VM_BUG_ON_MM(mm->pmd_huge_pte, mm); |
c3f0327f KK |
606 | #endif |
607 | } | |
608 | ||
1da177e4 LT |
609 | /* |
610 | * Allocate and initialize an mm_struct. | |
611 | */ | |
fb0a685c | 612 | struct mm_struct *mm_alloc(void) |
1da177e4 | 613 | { |
fb0a685c | 614 | struct mm_struct *mm; |
1da177e4 LT |
615 | |
616 | mm = allocate_mm(); | |
de03c72c KM |
617 | if (!mm) |
618 | return NULL; | |
619 | ||
620 | memset(mm, 0, sizeof(*mm)); | |
6345d24d | 621 | return mm_init(mm, current); |
1da177e4 LT |
622 | } |
623 | ||
624 | /* | |
625 | * Called when the last reference to the mm | |
626 | * is dropped: either by a lazy thread or by | |
627 | * mmput. Free the page directory and the mm. | |
628 | */ | |
7ad5b3a5 | 629 | void __mmdrop(struct mm_struct *mm) |
1da177e4 LT |
630 | { |
631 | BUG_ON(mm == &init_mm); | |
632 | mm_free_pgd(mm); | |
633 | destroy_context(mm); | |
cddb8a5c | 634 | mmu_notifier_mm_destroy(mm); |
c3f0327f | 635 | check_mm(mm); |
1da177e4 LT |
636 | free_mm(mm); |
637 | } | |
6d4e4c4f | 638 | EXPORT_SYMBOL_GPL(__mmdrop); |
1da177e4 LT |
639 | |
640 | /* | |
641 | * Decrement the use count and release all resources for an mm. | |
642 | */ | |
643 | void mmput(struct mm_struct *mm) | |
644 | { | |
0ae26f1b AM |
645 | might_sleep(); |
646 | ||
1da177e4 | 647 | if (atomic_dec_and_test(&mm->mm_users)) { |
d4b3b638 | 648 | uprobe_clear_state(mm); |
1da177e4 | 649 | exit_aio(mm); |
1c2fb7a4 | 650 | ksm_exit(mm); |
ba76149f | 651 | khugepaged_exit(mm); /* must run before exit_mmap */ |
1da177e4 | 652 | exit_mmap(mm); |
925d1c40 | 653 | set_mm_exe_file(mm, NULL); |
1da177e4 LT |
654 | if (!list_empty(&mm->mmlist)) { |
655 | spin_lock(&mmlist_lock); | |
656 | list_del(&mm->mmlist); | |
657 | spin_unlock(&mmlist_lock); | |
658 | } | |
801460d0 HS |
659 | if (mm->binfmt) |
660 | module_put(mm->binfmt->module); | |
1da177e4 LT |
661 | mmdrop(mm); |
662 | } | |
663 | } | |
664 | EXPORT_SYMBOL_GPL(mmput); | |
665 | ||
38646013 JS |
666 | void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file) |
667 | { | |
668 | if (new_exe_file) | |
669 | get_file(new_exe_file); | |
670 | if (mm->exe_file) | |
671 | fput(mm->exe_file); | |
672 | mm->exe_file = new_exe_file; | |
38646013 JS |
673 | } |
674 | ||
675 | struct file *get_mm_exe_file(struct mm_struct *mm) | |
676 | { | |
677 | struct file *exe_file; | |
678 | ||
2dd8ad81 | 679 | /* We need mmap_sem to protect against races with removal of exe_file */ |
38646013 JS |
680 | down_read(&mm->mmap_sem); |
681 | exe_file = mm->exe_file; | |
682 | if (exe_file) | |
683 | get_file(exe_file); | |
684 | up_read(&mm->mmap_sem); | |
685 | return exe_file; | |
686 | } | |
687 | ||
688 | static void dup_mm_exe_file(struct mm_struct *oldmm, struct mm_struct *newmm) | |
689 | { | |
690 | /* It's safe to write the exe_file pointer without exe_file_lock because | |
691 | * this is called during fork when the task is not yet in /proc */ | |
692 | newmm->exe_file = get_mm_exe_file(oldmm); | |
693 | } | |
694 | ||
1da177e4 LT |
695 | /** |
696 | * get_task_mm - acquire a reference to the task's mm | |
697 | * | |
246bb0b1 | 698 | * Returns %NULL if the task has no mm. Checks PF_KTHREAD (meaning |
1da177e4 LT |
699 | * this kernel workthread has transiently adopted a user mm with use_mm, |
700 | * to do its AIO) is not set and if so returns a reference to it, after | |
701 | * bumping up the use count. User must release the mm via mmput() | |
702 | * after use. Typically used by /proc and ptrace. | |
703 | */ | |
704 | struct mm_struct *get_task_mm(struct task_struct *task) | |
705 | { | |
706 | struct mm_struct *mm; | |
707 | ||
708 | task_lock(task); | |
709 | mm = task->mm; | |
710 | if (mm) { | |
246bb0b1 | 711 | if (task->flags & PF_KTHREAD) |
1da177e4 LT |
712 | mm = NULL; |
713 | else | |
714 | atomic_inc(&mm->mm_users); | |
715 | } | |
716 | task_unlock(task); | |
717 | return mm; | |
718 | } | |
719 | EXPORT_SYMBOL_GPL(get_task_mm); | |
720 | ||
8cdb878d CY |
721 | struct mm_struct *mm_access(struct task_struct *task, unsigned int mode) |
722 | { | |
723 | struct mm_struct *mm; | |
724 | int err; | |
725 | ||
726 | err = mutex_lock_killable(&task->signal->cred_guard_mutex); | |
727 | if (err) | |
728 | return ERR_PTR(err); | |
729 | ||
730 | mm = get_task_mm(task); | |
731 | if (mm && mm != current->mm && | |
732 | !ptrace_may_access(task, mode)) { | |
733 | mmput(mm); | |
734 | mm = ERR_PTR(-EACCES); | |
735 | } | |
736 | mutex_unlock(&task->signal->cred_guard_mutex); | |
737 | ||
738 | return mm; | |
739 | } | |
740 | ||
57b59c4a | 741 | static void complete_vfork_done(struct task_struct *tsk) |
c415c3b4 | 742 | { |
d68b46fe | 743 | struct completion *vfork; |
c415c3b4 | 744 | |
d68b46fe ON |
745 | task_lock(tsk); |
746 | vfork = tsk->vfork_done; | |
747 | if (likely(vfork)) { | |
748 | tsk->vfork_done = NULL; | |
749 | complete(vfork); | |
750 | } | |
751 | task_unlock(tsk); | |
752 | } | |
753 | ||
754 | static int wait_for_vfork_done(struct task_struct *child, | |
755 | struct completion *vfork) | |
756 | { | |
757 | int killed; | |
758 | ||
759 | freezer_do_not_count(); | |
760 | killed = wait_for_completion_killable(vfork); | |
761 | freezer_count(); | |
762 | ||
763 | if (killed) { | |
764 | task_lock(child); | |
765 | child->vfork_done = NULL; | |
766 | task_unlock(child); | |
767 | } | |
768 | ||
769 | put_task_struct(child); | |
770 | return killed; | |
c415c3b4 ON |
771 | } |
772 | ||
1da177e4 LT |
773 | /* Please note the differences between mmput and mm_release. |
774 | * mmput is called whenever we stop holding onto a mm_struct, | |
775 | * error success whatever. | |
776 | * | |
777 | * mm_release is called after a mm_struct has been removed | |
778 | * from the current process. | |
779 | * | |
780 | * This difference is important for error handling, when we | |
781 | * only half set up a mm_struct for a new process and need to restore | |
782 | * the old one. Because we mmput the new mm_struct before | |
783 | * restoring the old one. . . | |
784 | * Eric Biederman 10 January 1998 | |
785 | */ | |
786 | void mm_release(struct task_struct *tsk, struct mm_struct *mm) | |
787 | { | |
8141c7f3 LT |
788 | /* Get rid of any futexes when releasing the mm */ |
789 | #ifdef CONFIG_FUTEX | |
fc6b177d | 790 | if (unlikely(tsk->robust_list)) { |
8141c7f3 | 791 | exit_robust_list(tsk); |
fc6b177d PZ |
792 | tsk->robust_list = NULL; |
793 | } | |
8141c7f3 | 794 | #ifdef CONFIG_COMPAT |
fc6b177d | 795 | if (unlikely(tsk->compat_robust_list)) { |
8141c7f3 | 796 | compat_exit_robust_list(tsk); |
fc6b177d PZ |
797 | tsk->compat_robust_list = NULL; |
798 | } | |
8141c7f3 | 799 | #endif |
322a2c10 TG |
800 | if (unlikely(!list_empty(&tsk->pi_state_list))) |
801 | exit_pi_state_list(tsk); | |
8141c7f3 LT |
802 | #endif |
803 | ||
0326f5a9 SD |
804 | uprobe_free_utask(tsk); |
805 | ||
1da177e4 LT |
806 | /* Get rid of any cached register state */ |
807 | deactivate_mm(tsk, mm); | |
808 | ||
fec1d011 RM |
809 | /* |
810 | * If we're exiting normally, clear a user-space tid field if | |
811 | * requested. We leave this alone when dying by signal, to leave | |
812 | * the value intact in a core dump, and to save the unnecessary | |
d68b46fe ON |
813 | * trouble, say, a killed vfork parent shouldn't touch this mm. |
814 | * Userland only wants this done for a sys_exit. | |
fec1d011 | 815 | */ |
9c8a8228 ED |
816 | if (tsk->clear_child_tid) { |
817 | if (!(tsk->flags & PF_SIGNALED) && | |
818 | atomic_read(&mm->mm_users) > 1) { | |
819 | /* | |
820 | * We don't check the error code - if userspace has | |
821 | * not set up a proper pointer then tough luck. | |
822 | */ | |
823 | put_user(0, tsk->clear_child_tid); | |
824 | sys_futex(tsk->clear_child_tid, FUTEX_WAKE, | |
825 | 1, NULL, NULL, 0); | |
826 | } | |
1da177e4 | 827 | tsk->clear_child_tid = NULL; |
1da177e4 | 828 | } |
f7505d64 KK |
829 | |
830 | /* | |
831 | * All done, finally we can wake up parent and return this mm to him. | |
832 | * Also kthread_stop() uses this completion for synchronization. | |
833 | */ | |
834 | if (tsk->vfork_done) | |
835 | complete_vfork_done(tsk); | |
1da177e4 LT |
836 | } |
837 | ||
a0a7ec30 JD |
838 | /* |
839 | * Allocate a new mm structure and copy contents from the | |
840 | * mm structure of the passed in task structure. | |
841 | */ | |
ff252c1f | 842 | static struct mm_struct *dup_mm(struct task_struct *tsk) |
a0a7ec30 JD |
843 | { |
844 | struct mm_struct *mm, *oldmm = current->mm; | |
845 | int err; | |
846 | ||
a0a7ec30 JD |
847 | mm = allocate_mm(); |
848 | if (!mm) | |
849 | goto fail_nomem; | |
850 | ||
851 | memcpy(mm, oldmm, sizeof(*mm)); | |
852 | ||
78fb7466 | 853 | if (!mm_init(mm, tsk)) |
a0a7ec30 JD |
854 | goto fail_nomem; |
855 | ||
925d1c40 MH |
856 | dup_mm_exe_file(oldmm, mm); |
857 | ||
a0a7ec30 JD |
858 | err = dup_mmap(mm, oldmm); |
859 | if (err) | |
860 | goto free_pt; | |
861 | ||
862 | mm->hiwater_rss = get_mm_rss(mm); | |
863 | mm->hiwater_vm = mm->total_vm; | |
864 | ||
801460d0 HS |
865 | if (mm->binfmt && !try_module_get(mm->binfmt->module)) |
866 | goto free_pt; | |
867 | ||
a0a7ec30 JD |
868 | return mm; |
869 | ||
870 | free_pt: | |
801460d0 HS |
871 | /* don't put binfmt in mmput, we haven't got module yet */ |
872 | mm->binfmt = NULL; | |
a0a7ec30 JD |
873 | mmput(mm); |
874 | ||
875 | fail_nomem: | |
876 | return NULL; | |
a0a7ec30 JD |
877 | } |
878 | ||
fb0a685c | 879 | static int copy_mm(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 | 880 | { |
fb0a685c | 881 | struct mm_struct *mm, *oldmm; |
1da177e4 LT |
882 | int retval; |
883 | ||
884 | tsk->min_flt = tsk->maj_flt = 0; | |
885 | tsk->nvcsw = tsk->nivcsw = 0; | |
17406b82 MSB |
886 | #ifdef CONFIG_DETECT_HUNG_TASK |
887 | tsk->last_switch_count = tsk->nvcsw + tsk->nivcsw; | |
888 | #endif | |
1da177e4 LT |
889 | |
890 | tsk->mm = NULL; | |
891 | tsk->active_mm = NULL; | |
892 | ||
893 | /* | |
894 | * Are we cloning a kernel thread? | |
895 | * | |
896 | * We need to steal a active VM for that.. | |
897 | */ | |
898 | oldmm = current->mm; | |
899 | if (!oldmm) | |
900 | return 0; | |
901 | ||
615d6e87 DB |
902 | /* initialize the new vmacache entries */ |
903 | vmacache_flush(tsk); | |
904 | ||
1da177e4 LT |
905 | if (clone_flags & CLONE_VM) { |
906 | atomic_inc(&oldmm->mm_users); | |
907 | mm = oldmm; | |
1da177e4 LT |
908 | goto good_mm; |
909 | } | |
910 | ||
911 | retval = -ENOMEM; | |
a0a7ec30 | 912 | mm = dup_mm(tsk); |
1da177e4 LT |
913 | if (!mm) |
914 | goto fail_nomem; | |
915 | ||
1da177e4 LT |
916 | good_mm: |
917 | tsk->mm = mm; | |
918 | tsk->active_mm = mm; | |
919 | return 0; | |
920 | ||
1da177e4 LT |
921 | fail_nomem: |
922 | return retval; | |
1da177e4 LT |
923 | } |
924 | ||
a39bc516 | 925 | static int copy_fs(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 | 926 | { |
498052bb | 927 | struct fs_struct *fs = current->fs; |
1da177e4 | 928 | if (clone_flags & CLONE_FS) { |
498052bb | 929 | /* tsk->fs is already what we want */ |
2a4419b5 | 930 | spin_lock(&fs->lock); |
498052bb | 931 | if (fs->in_exec) { |
2a4419b5 | 932 | spin_unlock(&fs->lock); |
498052bb AV |
933 | return -EAGAIN; |
934 | } | |
935 | fs->users++; | |
2a4419b5 | 936 | spin_unlock(&fs->lock); |
1da177e4 LT |
937 | return 0; |
938 | } | |
498052bb | 939 | tsk->fs = copy_fs_struct(fs); |
1da177e4 LT |
940 | if (!tsk->fs) |
941 | return -ENOMEM; | |
942 | return 0; | |
943 | } | |
944 | ||
fb0a685c | 945 | static int copy_files(unsigned long clone_flags, struct task_struct *tsk) |
a016f338 JD |
946 | { |
947 | struct files_struct *oldf, *newf; | |
948 | int error = 0; | |
949 | ||
950 | /* | |
951 | * A background process may not have any files ... | |
952 | */ | |
953 | oldf = current->files; | |
954 | if (!oldf) | |
955 | goto out; | |
956 | ||
957 | if (clone_flags & CLONE_FILES) { | |
958 | atomic_inc(&oldf->count); | |
959 | goto out; | |
960 | } | |
961 | ||
a016f338 JD |
962 | newf = dup_fd(oldf, &error); |
963 | if (!newf) | |
964 | goto out; | |
965 | ||
966 | tsk->files = newf; | |
967 | error = 0; | |
968 | out: | |
969 | return error; | |
970 | } | |
971 | ||
fadad878 | 972 | static int copy_io(unsigned long clone_flags, struct task_struct *tsk) |
fd0928df JA |
973 | { |
974 | #ifdef CONFIG_BLOCK | |
975 | struct io_context *ioc = current->io_context; | |
6e736be7 | 976 | struct io_context *new_ioc; |
fd0928df JA |
977 | |
978 | if (!ioc) | |
979 | return 0; | |
fadad878 JA |
980 | /* |
981 | * Share io context with parent, if CLONE_IO is set | |
982 | */ | |
983 | if (clone_flags & CLONE_IO) { | |
3d48749d TH |
984 | ioc_task_link(ioc); |
985 | tsk->io_context = ioc; | |
fadad878 | 986 | } else if (ioprio_valid(ioc->ioprio)) { |
6e736be7 TH |
987 | new_ioc = get_task_io_context(tsk, GFP_KERNEL, NUMA_NO_NODE); |
988 | if (unlikely(!new_ioc)) | |
fd0928df JA |
989 | return -ENOMEM; |
990 | ||
6e736be7 | 991 | new_ioc->ioprio = ioc->ioprio; |
11a3122f | 992 | put_io_context(new_ioc); |
fd0928df JA |
993 | } |
994 | #endif | |
995 | return 0; | |
996 | } | |
997 | ||
a39bc516 | 998 | static int copy_sighand(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 LT |
999 | { |
1000 | struct sighand_struct *sig; | |
1001 | ||
60348802 | 1002 | if (clone_flags & CLONE_SIGHAND) { |
1da177e4 LT |
1003 | atomic_inc(¤t->sighand->count); |
1004 | return 0; | |
1005 | } | |
1006 | sig = kmem_cache_alloc(sighand_cachep, GFP_KERNEL); | |
e56d0903 | 1007 | rcu_assign_pointer(tsk->sighand, sig); |
1da177e4 LT |
1008 | if (!sig) |
1009 | return -ENOMEM; | |
1da177e4 LT |
1010 | atomic_set(&sig->count, 1); |
1011 | memcpy(sig->action, current->sighand->action, sizeof(sig->action)); | |
1012 | return 0; | |
1013 | } | |
1014 | ||
a7e5328a | 1015 | void __cleanup_sighand(struct sighand_struct *sighand) |
c81addc9 | 1016 | { |
d80e731e ON |
1017 | if (atomic_dec_and_test(&sighand->count)) { |
1018 | signalfd_cleanup(sighand); | |
c81addc9 | 1019 | kmem_cache_free(sighand_cachep, sighand); |
d80e731e | 1020 | } |
c81addc9 ON |
1021 | } |
1022 | ||
f06febc9 FM |
1023 | |
1024 | /* | |
1025 | * Initialize POSIX timer handling for a thread group. | |
1026 | */ | |
1027 | static void posix_cpu_timers_init_group(struct signal_struct *sig) | |
1028 | { | |
78d7d407 JS |
1029 | unsigned long cpu_limit; |
1030 | ||
f06febc9 FM |
1031 | /* Thread group counters. */ |
1032 | thread_group_cputime_init(sig); | |
1033 | ||
78d7d407 JS |
1034 | cpu_limit = ACCESS_ONCE(sig->rlim[RLIMIT_CPU].rlim_cur); |
1035 | if (cpu_limit != RLIM_INFINITY) { | |
1036 | sig->cputime_expires.prof_exp = secs_to_cputime(cpu_limit); | |
6279a751 ON |
1037 | sig->cputimer.running = 1; |
1038 | } | |
1039 | ||
f06febc9 FM |
1040 | /* The timer lists. */ |
1041 | INIT_LIST_HEAD(&sig->cpu_timers[0]); | |
1042 | INIT_LIST_HEAD(&sig->cpu_timers[1]); | |
1043 | INIT_LIST_HEAD(&sig->cpu_timers[2]); | |
1044 | } | |
1045 | ||
a39bc516 | 1046 | static int copy_signal(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 LT |
1047 | { |
1048 | struct signal_struct *sig; | |
1da177e4 | 1049 | |
4ab6c083 | 1050 | if (clone_flags & CLONE_THREAD) |
490dea45 | 1051 | return 0; |
490dea45 | 1052 | |
a56704ef | 1053 | sig = kmem_cache_zalloc(signal_cachep, GFP_KERNEL); |
1da177e4 LT |
1054 | tsk->signal = sig; |
1055 | if (!sig) | |
1056 | return -ENOMEM; | |
1057 | ||
b3ac022c | 1058 | sig->nr_threads = 1; |
1da177e4 | 1059 | atomic_set(&sig->live, 1); |
b3ac022c | 1060 | atomic_set(&sig->sigcnt, 1); |
0c740d0a ON |
1061 | |
1062 | /* list_add(thread_node, thread_head) without INIT_LIST_HEAD() */ | |
1063 | sig->thread_head = (struct list_head)LIST_HEAD_INIT(tsk->thread_node); | |
1064 | tsk->thread_node = (struct list_head)LIST_HEAD_INIT(sig->thread_head); | |
1065 | ||
1da177e4 | 1066 | init_waitqueue_head(&sig->wait_chldexit); |
db51aecc | 1067 | sig->curr_target = tsk; |
1da177e4 LT |
1068 | init_sigpending(&sig->shared_pending); |
1069 | INIT_LIST_HEAD(&sig->posix_timers); | |
1070 | ||
c9cb2e3d | 1071 | hrtimer_init(&sig->real_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL); |
1da177e4 | 1072 | sig->real_timer.function = it_real_fn; |
1da177e4 | 1073 | |
1da177e4 LT |
1074 | task_lock(current->group_leader); |
1075 | memcpy(sig->rlim, current->signal->rlim, sizeof sig->rlim); | |
1076 | task_unlock(current->group_leader); | |
1077 | ||
6279a751 ON |
1078 | posix_cpu_timers_init_group(sig); |
1079 | ||
522ed776 | 1080 | tty_audit_fork(sig); |
5091faa4 | 1081 | sched_autogroup_fork(sig); |
522ed776 | 1082 | |
4714d1d3 | 1083 | #ifdef CONFIG_CGROUPS |
257058ae | 1084 | init_rwsem(&sig->group_rwsem); |
4714d1d3 BB |
1085 | #endif |
1086 | ||
a63d83f4 | 1087 | sig->oom_score_adj = current->signal->oom_score_adj; |
dabb16f6 | 1088 | sig->oom_score_adj_min = current->signal->oom_score_adj_min; |
28b83c51 | 1089 | |
ebec18a6 LP |
1090 | sig->has_child_subreaper = current->signal->has_child_subreaper || |
1091 | current->signal->is_child_subreaper; | |
1092 | ||
9b1bf12d KM |
1093 | mutex_init(&sig->cred_guard_mutex); |
1094 | ||
1da177e4 LT |
1095 | return 0; |
1096 | } | |
1097 | ||
dbd95212 KC |
1098 | static void copy_seccomp(struct task_struct *p) |
1099 | { | |
1100 | #ifdef CONFIG_SECCOMP | |
1101 | /* | |
1102 | * Must be called with sighand->lock held, which is common to | |
1103 | * all threads in the group. Holding cred_guard_mutex is not | |
1104 | * needed because this new task is not yet running and cannot | |
1105 | * be racing exec. | |
1106 | */ | |
69f6a34b | 1107 | assert_spin_locked(¤t->sighand->siglock); |
dbd95212 KC |
1108 | |
1109 | /* Ref-count the new filter user, and assign it. */ | |
1110 | get_seccomp_filter(current); | |
1111 | p->seccomp = current->seccomp; | |
1112 | ||
1113 | /* | |
1114 | * Explicitly enable no_new_privs here in case it got set | |
1115 | * between the task_struct being duplicated and holding the | |
1116 | * sighand lock. The seccomp state and nnp must be in sync. | |
1117 | */ | |
1118 | if (task_no_new_privs(current)) | |
1119 | task_set_no_new_privs(p); | |
1120 | ||
1121 | /* | |
1122 | * If the parent gained a seccomp mode after copying thread | |
1123 | * flags and between before we held the sighand lock, we have | |
1124 | * to manually enable the seccomp thread flag here. | |
1125 | */ | |
1126 | if (p->seccomp.mode != SECCOMP_MODE_DISABLED) | |
1127 | set_tsk_thread_flag(p, TIF_SECCOMP); | |
1128 | #endif | |
1129 | } | |
1130 | ||
17da2bd9 | 1131 | SYSCALL_DEFINE1(set_tid_address, int __user *, tidptr) |
1da177e4 LT |
1132 | { |
1133 | current->clear_child_tid = tidptr; | |
1134 | ||
b488893a | 1135 | return task_pid_vnr(current); |
1da177e4 LT |
1136 | } |
1137 | ||
a39bc516 | 1138 | static void rt_mutex_init_task(struct task_struct *p) |
23f78d4a | 1139 | { |
1d615482 | 1140 | raw_spin_lock_init(&p->pi_lock); |
e29e175b | 1141 | #ifdef CONFIG_RT_MUTEXES |
fb00aca4 PZ |
1142 | p->pi_waiters = RB_ROOT; |
1143 | p->pi_waiters_leftmost = NULL; | |
23f78d4a | 1144 | p->pi_blocked_on = NULL; |
23f78d4a IM |
1145 | #endif |
1146 | } | |
1147 | ||
f06febc9 FM |
1148 | /* |
1149 | * Initialize POSIX timer handling for a single task. | |
1150 | */ | |
1151 | static void posix_cpu_timers_init(struct task_struct *tsk) | |
1152 | { | |
64861634 MS |
1153 | tsk->cputime_expires.prof_exp = 0; |
1154 | tsk->cputime_expires.virt_exp = 0; | |
f06febc9 FM |
1155 | tsk->cputime_expires.sched_exp = 0; |
1156 | INIT_LIST_HEAD(&tsk->cpu_timers[0]); | |
1157 | INIT_LIST_HEAD(&tsk->cpu_timers[1]); | |
1158 | INIT_LIST_HEAD(&tsk->cpu_timers[2]); | |
1159 | } | |
1160 | ||
81907739 ON |
1161 | static inline void |
1162 | init_task_pid(struct task_struct *task, enum pid_type type, struct pid *pid) | |
1163 | { | |
1164 | task->pids[type].pid = pid; | |
1165 | } | |
1166 | ||
1da177e4 LT |
1167 | /* |
1168 | * This creates a new process as a copy of the old one, | |
1169 | * but does not actually start it yet. | |
1170 | * | |
1171 | * It copies the registers, and all the appropriate | |
1172 | * parts of the process environment (as per the clone | |
1173 | * flags). The actual kick-off is left to the caller. | |
1174 | */ | |
36c8b586 IM |
1175 | static struct task_struct *copy_process(unsigned long clone_flags, |
1176 | unsigned long stack_start, | |
36c8b586 | 1177 | unsigned long stack_size, |
36c8b586 | 1178 | int __user *child_tidptr, |
09a05394 RM |
1179 | struct pid *pid, |
1180 | int trace) | |
1da177e4 LT |
1181 | { |
1182 | int retval; | |
a24efe62 | 1183 | struct task_struct *p; |
1da177e4 LT |
1184 | |
1185 | if ((clone_flags & (CLONE_NEWNS|CLONE_FS)) == (CLONE_NEWNS|CLONE_FS)) | |
1186 | return ERR_PTR(-EINVAL); | |
1187 | ||
e66eded8 EB |
1188 | if ((clone_flags & (CLONE_NEWUSER|CLONE_FS)) == (CLONE_NEWUSER|CLONE_FS)) |
1189 | return ERR_PTR(-EINVAL); | |
1190 | ||
1da177e4 LT |
1191 | /* |
1192 | * Thread groups must share signals as well, and detached threads | |
1193 | * can only be started up within the thread group. | |
1194 | */ | |
1195 | if ((clone_flags & CLONE_THREAD) && !(clone_flags & CLONE_SIGHAND)) | |
1196 | return ERR_PTR(-EINVAL); | |
1197 | ||
1198 | /* | |
1199 | * Shared signal handlers imply shared VM. By way of the above, | |
1200 | * thread groups also imply shared VM. Blocking this case allows | |
1201 | * for various simplifications in other code. | |
1202 | */ | |
1203 | if ((clone_flags & CLONE_SIGHAND) && !(clone_flags & CLONE_VM)) | |
1204 | return ERR_PTR(-EINVAL); | |
1205 | ||
123be07b SB |
1206 | /* |
1207 | * Siblings of global init remain as zombies on exit since they are | |
1208 | * not reaped by their parent (swapper). To solve this and to avoid | |
1209 | * multi-rooted process trees, prevent global and container-inits | |
1210 | * from creating siblings. | |
1211 | */ | |
1212 | if ((clone_flags & CLONE_PARENT) && | |
1213 | current->signal->flags & SIGNAL_UNKILLABLE) | |
1214 | return ERR_PTR(-EINVAL); | |
1215 | ||
8382fcac | 1216 | /* |
40a0d32d ON |
1217 | * If the new process will be in a different pid or user namespace |
1218 | * do not allow it to share a thread group or signal handlers or | |
1219 | * parent with the forking task. | |
8382fcac | 1220 | */ |
1f7f4dde | 1221 | if (clone_flags & CLONE_SIGHAND) { |
40a0d32d ON |
1222 | if ((clone_flags & (CLONE_NEWUSER | CLONE_NEWPID)) || |
1223 | (task_active_pid_ns(current) != | |
1224 | current->nsproxy->pid_ns_for_children)) | |
1225 | return ERR_PTR(-EINVAL); | |
1226 | } | |
8382fcac | 1227 | |
1da177e4 LT |
1228 | retval = security_task_create(clone_flags); |
1229 | if (retval) | |
1230 | goto fork_out; | |
1231 | ||
1232 | retval = -ENOMEM; | |
1233 | p = dup_task_struct(current); | |
1234 | if (!p) | |
1235 | goto fork_out; | |
1236 | ||
f7e8b616 SR |
1237 | ftrace_graph_init_task(p); |
1238 | ||
bea493a0 PZ |
1239 | rt_mutex_init_task(p); |
1240 | ||
d12c1a37 | 1241 | #ifdef CONFIG_PROVE_LOCKING |
de30a2b3 IM |
1242 | DEBUG_LOCKS_WARN_ON(!p->hardirqs_enabled); |
1243 | DEBUG_LOCKS_WARN_ON(!p->softirqs_enabled); | |
1244 | #endif | |
1da177e4 | 1245 | retval = -EAGAIN; |
3b11a1de | 1246 | if (atomic_read(&p->real_cred->user->processes) >= |
78d7d407 | 1247 | task_rlimit(p, RLIMIT_NPROC)) { |
b57922b6 EP |
1248 | if (p->real_cred->user != INIT_USER && |
1249 | !capable(CAP_SYS_RESOURCE) && !capable(CAP_SYS_ADMIN)) | |
1da177e4 LT |
1250 | goto bad_fork_free; |
1251 | } | |
72fa5997 | 1252 | current->flags &= ~PF_NPROC_EXCEEDED; |
1da177e4 | 1253 | |
f1752eec DH |
1254 | retval = copy_creds(p, clone_flags); |
1255 | if (retval < 0) | |
1256 | goto bad_fork_free; | |
1da177e4 LT |
1257 | |
1258 | /* | |
1259 | * If multiple threads are within copy_process(), then this check | |
1260 | * triggers too late. This doesn't hurt, the check is only there | |
1261 | * to stop root fork bombs. | |
1262 | */ | |
04ec93fe | 1263 | retval = -EAGAIN; |
1da177e4 LT |
1264 | if (nr_threads >= max_threads) |
1265 | goto bad_fork_cleanup_count; | |
1266 | ||
a1261f54 | 1267 | if (!try_module_get(task_thread_info(p)->exec_domain->module)) |
1da177e4 LT |
1268 | goto bad_fork_cleanup_count; |
1269 | ||
ca74e92b | 1270 | delayacct_tsk_init(p); /* Must remain after dup_task_struct() */ |
514ddb44 DR |
1271 | p->flags &= ~(PF_SUPERPRIV | PF_WQ_WORKER); |
1272 | p->flags |= PF_FORKNOEXEC; | |
1da177e4 LT |
1273 | INIT_LIST_HEAD(&p->children); |
1274 | INIT_LIST_HEAD(&p->sibling); | |
f41d911f | 1275 | rcu_copy_process(p); |
1da177e4 LT |
1276 | p->vfork_done = NULL; |
1277 | spin_lock_init(&p->alloc_lock); | |
1da177e4 | 1278 | |
1da177e4 LT |
1279 | init_sigpending(&p->pending); |
1280 | ||
64861634 MS |
1281 | p->utime = p->stime = p->gtime = 0; |
1282 | p->utimescaled = p->stimescaled = 0; | |
9fbc42ea | 1283 | #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE |
d37f761d | 1284 | p->prev_cputime.utime = p->prev_cputime.stime = 0; |
d99ca3b9 | 1285 | #endif |
6a61671b FW |
1286 | #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN |
1287 | seqlock_init(&p->vtime_seqlock); | |
1288 | p->vtime_snap = 0; | |
1289 | p->vtime_snap_whence = VTIME_SLEEPING; | |
1290 | #endif | |
1291 | ||
a3a2e76c KH |
1292 | #if defined(SPLIT_RSS_COUNTING) |
1293 | memset(&p->rss_stat, 0, sizeof(p->rss_stat)); | |
1294 | #endif | |
172ba844 | 1295 | |
6976675d AV |
1296 | p->default_timer_slack_ns = current->timer_slack_ns; |
1297 | ||
5995477a | 1298 | task_io_accounting_init(&p->ioac); |
1da177e4 LT |
1299 | acct_clear_integrals(p); |
1300 | ||
f06febc9 | 1301 | posix_cpu_timers_init(p); |
1da177e4 | 1302 | |
ccbf62d8 | 1303 | p->start_time = ktime_get_ns(); |
57e0be04 | 1304 | p->real_start_time = ktime_get_boot_ns(); |
1da177e4 | 1305 | p->io_context = NULL; |
1da177e4 | 1306 | p->audit_context = NULL; |
4714d1d3 | 1307 | if (clone_flags & CLONE_THREAD) |
257058ae | 1308 | threadgroup_change_begin(current); |
b4f48b63 | 1309 | cgroup_fork(p); |
1da177e4 | 1310 | #ifdef CONFIG_NUMA |
846a16bf | 1311 | p->mempolicy = mpol_dup(p->mempolicy); |
fb0a685c DRO |
1312 | if (IS_ERR(p->mempolicy)) { |
1313 | retval = PTR_ERR(p->mempolicy); | |
1314 | p->mempolicy = NULL; | |
e8604cb4 | 1315 | goto bad_fork_cleanup_threadgroup_lock; |
fb0a685c | 1316 | } |
1da177e4 | 1317 | #endif |
778d3b0f MH |
1318 | #ifdef CONFIG_CPUSETS |
1319 | p->cpuset_mem_spread_rotor = NUMA_NO_NODE; | |
1320 | p->cpuset_slab_spread_rotor = NUMA_NO_NODE; | |
cc9a6c87 | 1321 | seqcount_init(&p->mems_allowed_seq); |
778d3b0f | 1322 | #endif |
de30a2b3 IM |
1323 | #ifdef CONFIG_TRACE_IRQFLAGS |
1324 | p->irq_events = 0; | |
1325 | p->hardirqs_enabled = 0; | |
1326 | p->hardirq_enable_ip = 0; | |
1327 | p->hardirq_enable_event = 0; | |
1328 | p->hardirq_disable_ip = _THIS_IP_; | |
1329 | p->hardirq_disable_event = 0; | |
1330 | p->softirqs_enabled = 1; | |
1331 | p->softirq_enable_ip = _THIS_IP_; | |
1332 | p->softirq_enable_event = 0; | |
1333 | p->softirq_disable_ip = 0; | |
1334 | p->softirq_disable_event = 0; | |
1335 | p->hardirq_context = 0; | |
1336 | p->softirq_context = 0; | |
1337 | #endif | |
fbb9ce95 IM |
1338 | #ifdef CONFIG_LOCKDEP |
1339 | p->lockdep_depth = 0; /* no locks held yet */ | |
1340 | p->curr_chain_key = 0; | |
1341 | p->lockdep_recursion = 0; | |
1342 | #endif | |
1da177e4 | 1343 | |
408894ee IM |
1344 | #ifdef CONFIG_DEBUG_MUTEXES |
1345 | p->blocked_on = NULL; /* not blocked yet */ | |
1346 | #endif | |
cafe5635 KO |
1347 | #ifdef CONFIG_BCACHE |
1348 | p->sequential_io = 0; | |
1349 | p->sequential_io_avg = 0; | |
1350 | #endif | |
0f481406 | 1351 | |
3c90e6e9 | 1352 | /* Perform scheduler related setup. Assign this task to a CPU. */ |
aab03e05 DF |
1353 | retval = sched_fork(clone_flags, p); |
1354 | if (retval) | |
1355 | goto bad_fork_cleanup_policy; | |
6ab423e0 | 1356 | |
cdd6c482 | 1357 | retval = perf_event_init_task(p); |
6ab423e0 PZ |
1358 | if (retval) |
1359 | goto bad_fork_cleanup_policy; | |
fb0a685c DRO |
1360 | retval = audit_alloc(p); |
1361 | if (retval) | |
6c72e350 | 1362 | goto bad_fork_cleanup_perf; |
1da177e4 | 1363 | /* copy all the process information */ |
ab602f79 | 1364 | shm_init_task(p); |
fb0a685c DRO |
1365 | retval = copy_semundo(clone_flags, p); |
1366 | if (retval) | |
1da177e4 | 1367 | goto bad_fork_cleanup_audit; |
fb0a685c DRO |
1368 | retval = copy_files(clone_flags, p); |
1369 | if (retval) | |
1da177e4 | 1370 | goto bad_fork_cleanup_semundo; |
fb0a685c DRO |
1371 | retval = copy_fs(clone_flags, p); |
1372 | if (retval) | |
1da177e4 | 1373 | goto bad_fork_cleanup_files; |
fb0a685c DRO |
1374 | retval = copy_sighand(clone_flags, p); |
1375 | if (retval) | |
1da177e4 | 1376 | goto bad_fork_cleanup_fs; |
fb0a685c DRO |
1377 | retval = copy_signal(clone_flags, p); |
1378 | if (retval) | |
1da177e4 | 1379 | goto bad_fork_cleanup_sighand; |
fb0a685c DRO |
1380 | retval = copy_mm(clone_flags, p); |
1381 | if (retval) | |
1da177e4 | 1382 | goto bad_fork_cleanup_signal; |
fb0a685c DRO |
1383 | retval = copy_namespaces(clone_flags, p); |
1384 | if (retval) | |
d84f4f99 | 1385 | goto bad_fork_cleanup_mm; |
fb0a685c DRO |
1386 | retval = copy_io(clone_flags, p); |
1387 | if (retval) | |
fd0928df | 1388 | goto bad_fork_cleanup_namespaces; |
afa86fc4 | 1389 | retval = copy_thread(clone_flags, stack_start, stack_size, p); |
1da177e4 | 1390 | if (retval) |
fd0928df | 1391 | goto bad_fork_cleanup_io; |
1da177e4 | 1392 | |
425fb2b4 PE |
1393 | if (pid != &init_struct_pid) { |
1394 | retval = -ENOMEM; | |
c2b1df2e | 1395 | pid = alloc_pid(p->nsproxy->pid_ns_for_children); |
425fb2b4 | 1396 | if (!pid) |
fd0928df | 1397 | goto bad_fork_cleanup_io; |
425fb2b4 PE |
1398 | } |
1399 | ||
1da177e4 LT |
1400 | p->set_child_tid = (clone_flags & CLONE_CHILD_SETTID) ? child_tidptr : NULL; |
1401 | /* | |
1402 | * Clear TID on mm_release()? | |
1403 | */ | |
fb0a685c | 1404 | p->clear_child_tid = (clone_flags & CLONE_CHILD_CLEARTID) ? child_tidptr : NULL; |
73c10101 JA |
1405 | #ifdef CONFIG_BLOCK |
1406 | p->plug = NULL; | |
1407 | #endif | |
42b2dd0a | 1408 | #ifdef CONFIG_FUTEX |
8f17d3a5 IM |
1409 | p->robust_list = NULL; |
1410 | #ifdef CONFIG_COMPAT | |
1411 | p->compat_robust_list = NULL; | |
1412 | #endif | |
c87e2837 IM |
1413 | INIT_LIST_HEAD(&p->pi_state_list); |
1414 | p->pi_state_cache = NULL; | |
42b2dd0a | 1415 | #endif |
f9a3879a GM |
1416 | /* |
1417 | * sigaltstack should be cleared when sharing the same VM | |
1418 | */ | |
1419 | if ((clone_flags & (CLONE_VM|CLONE_VFORK)) == CLONE_VM) | |
1420 | p->sas_ss_sp = p->sas_ss_size = 0; | |
1421 | ||
1da177e4 | 1422 | /* |
6580807d ON |
1423 | * Syscall tracing and stepping should be turned off in the |
1424 | * child regardless of CLONE_PTRACE. | |
1da177e4 | 1425 | */ |
6580807d | 1426 | user_disable_single_step(p); |
1da177e4 | 1427 | clear_tsk_thread_flag(p, TIF_SYSCALL_TRACE); |
ed75e8d5 LV |
1428 | #ifdef TIF_SYSCALL_EMU |
1429 | clear_tsk_thread_flag(p, TIF_SYSCALL_EMU); | |
1430 | #endif | |
9745512c | 1431 | clear_all_latency_tracing(p); |
1da177e4 | 1432 | |
1da177e4 | 1433 | /* ok, now we should be set up.. */ |
18c830df ON |
1434 | p->pid = pid_nr(pid); |
1435 | if (clone_flags & CLONE_THREAD) { | |
5f8aadd8 | 1436 | p->exit_signal = -1; |
18c830df ON |
1437 | p->group_leader = current->group_leader; |
1438 | p->tgid = current->tgid; | |
1439 | } else { | |
1440 | if (clone_flags & CLONE_PARENT) | |
1441 | p->exit_signal = current->group_leader->exit_signal; | |
1442 | else | |
1443 | p->exit_signal = (clone_flags & CSIGNAL); | |
1444 | p->group_leader = p; | |
1445 | p->tgid = p->pid; | |
1446 | } | |
5f8aadd8 | 1447 | |
9d823e8f WF |
1448 | p->nr_dirtied = 0; |
1449 | p->nr_dirtied_pause = 128 >> (PAGE_SHIFT - 10); | |
83712358 | 1450 | p->dirty_paused_when = 0; |
9d823e8f | 1451 | |
bb8cbbfe | 1452 | p->pdeath_signal = 0; |
47e65328 | 1453 | INIT_LIST_HEAD(&p->thread_group); |
158e1645 | 1454 | p->task_works = NULL; |
1da177e4 | 1455 | |
18c830df ON |
1456 | /* |
1457 | * Make it visible to the rest of the system, but dont wake it up yet. | |
1458 | * Need tasklist lock for parent etc handling! | |
1459 | */ | |
1da177e4 LT |
1460 | write_lock_irq(&tasklist_lock); |
1461 | ||
1da177e4 | 1462 | /* CLONE_PARENT re-uses the old parent */ |
2d5516cb | 1463 | if (clone_flags & (CLONE_PARENT|CLONE_THREAD)) { |
1da177e4 | 1464 | p->real_parent = current->real_parent; |
2d5516cb ON |
1465 | p->parent_exec_id = current->parent_exec_id; |
1466 | } else { | |
1da177e4 | 1467 | p->real_parent = current; |
2d5516cb ON |
1468 | p->parent_exec_id = current->self_exec_id; |
1469 | } | |
1da177e4 | 1470 | |
3f17da69 | 1471 | spin_lock(¤t->sighand->siglock); |
4a2c7a78 | 1472 | |
dbd95212 KC |
1473 | /* |
1474 | * Copy seccomp details explicitly here, in case they were changed | |
1475 | * before holding sighand lock. | |
1476 | */ | |
1477 | copy_seccomp(p); | |
1478 | ||
4a2c7a78 ON |
1479 | /* |
1480 | * Process group and session signals need to be delivered to just the | |
1481 | * parent before the fork or both the parent and the child after the | |
1482 | * fork. Restart if a signal comes in before we add the new process to | |
1483 | * it's process group. | |
1484 | * A fatal signal pending means that current will exit, so the new | |
1485 | * thread can't slip out of an OOM kill (or normal SIGKILL). | |
fb0a685c | 1486 | */ |
23ff4440 | 1487 | recalc_sigpending(); |
4a2c7a78 ON |
1488 | if (signal_pending(current)) { |
1489 | spin_unlock(¤t->sighand->siglock); | |
1490 | write_unlock_irq(&tasklist_lock); | |
1491 | retval = -ERESTARTNOINTR; | |
f7e8b616 | 1492 | goto bad_fork_free_pid; |
4a2c7a78 ON |
1493 | } |
1494 | ||
73b9ebfe | 1495 | if (likely(p->pid)) { |
4b9d33e6 | 1496 | ptrace_init_task(p, (clone_flags & CLONE_PTRACE) || trace); |
73b9ebfe | 1497 | |
81907739 | 1498 | init_task_pid(p, PIDTYPE_PID, pid); |
73b9ebfe | 1499 | if (thread_group_leader(p)) { |
81907739 ON |
1500 | init_task_pid(p, PIDTYPE_PGID, task_pgrp(current)); |
1501 | init_task_pid(p, PIDTYPE_SID, task_session(current)); | |
1502 | ||
1c4042c2 | 1503 | if (is_child_reaper(pid)) { |
17cf22c3 | 1504 | ns_of_pid(pid)->child_reaper = p; |
1c4042c2 EB |
1505 | p->signal->flags |= SIGNAL_UNKILLABLE; |
1506 | } | |
73b9ebfe | 1507 | |
fea9d175 | 1508 | p->signal->leader_pid = pid; |
9c9f4ded | 1509 | p->signal->tty = tty_kref_get(current->signal->tty); |
9cd80bbb | 1510 | list_add_tail(&p->sibling, &p->real_parent->children); |
5e85d4ab | 1511 | list_add_tail_rcu(&p->tasks, &init_task.tasks); |
81907739 ON |
1512 | attach_pid(p, PIDTYPE_PGID); |
1513 | attach_pid(p, PIDTYPE_SID); | |
909ea964 | 1514 | __this_cpu_inc(process_counts); |
80628ca0 ON |
1515 | } else { |
1516 | current->signal->nr_threads++; | |
1517 | atomic_inc(¤t->signal->live); | |
1518 | atomic_inc(¤t->signal->sigcnt); | |
80628ca0 ON |
1519 | list_add_tail_rcu(&p->thread_group, |
1520 | &p->group_leader->thread_group); | |
0c740d0a ON |
1521 | list_add_tail_rcu(&p->thread_node, |
1522 | &p->signal->thread_head); | |
73b9ebfe | 1523 | } |
81907739 | 1524 | attach_pid(p, PIDTYPE_PID); |
73b9ebfe | 1525 | nr_threads++; |
1da177e4 LT |
1526 | } |
1527 | ||
1da177e4 | 1528 | total_forks++; |
3f17da69 | 1529 | spin_unlock(¤t->sighand->siglock); |
4af4206b | 1530 | syscall_tracepoint_update(p); |
1da177e4 | 1531 | write_unlock_irq(&tasklist_lock); |
4af4206b | 1532 | |
c13cf856 | 1533 | proc_fork_connector(p); |
817929ec | 1534 | cgroup_post_fork(p); |
4714d1d3 | 1535 | if (clone_flags & CLONE_THREAD) |
257058ae | 1536 | threadgroup_change_end(current); |
cdd6c482 | 1537 | perf_event_fork(p); |
43d2b113 KH |
1538 | |
1539 | trace_task_newtask(p, clone_flags); | |
3ab67966 | 1540 | uprobe_copy_process(p, clone_flags); |
43d2b113 | 1541 | |
1da177e4 LT |
1542 | return p; |
1543 | ||
425fb2b4 PE |
1544 | bad_fork_free_pid: |
1545 | if (pid != &init_struct_pid) | |
1546 | free_pid(pid); | |
fd0928df | 1547 | bad_fork_cleanup_io: |
b69f2292 LR |
1548 | if (p->io_context) |
1549 | exit_io_context(p); | |
ab516013 | 1550 | bad_fork_cleanup_namespaces: |
444f378b | 1551 | exit_task_namespaces(p); |
1da177e4 | 1552 | bad_fork_cleanup_mm: |
c9f01245 | 1553 | if (p->mm) |
1da177e4 LT |
1554 | mmput(p->mm); |
1555 | bad_fork_cleanup_signal: | |
4ab6c083 | 1556 | if (!(clone_flags & CLONE_THREAD)) |
1c5354de | 1557 | free_signal_struct(p->signal); |
1da177e4 | 1558 | bad_fork_cleanup_sighand: |
a7e5328a | 1559 | __cleanup_sighand(p->sighand); |
1da177e4 LT |
1560 | bad_fork_cleanup_fs: |
1561 | exit_fs(p); /* blocking */ | |
1562 | bad_fork_cleanup_files: | |
1563 | exit_files(p); /* blocking */ | |
1564 | bad_fork_cleanup_semundo: | |
1565 | exit_sem(p); | |
1566 | bad_fork_cleanup_audit: | |
1567 | audit_free(p); | |
6c72e350 | 1568 | bad_fork_cleanup_perf: |
cdd6c482 | 1569 | perf_event_free_task(p); |
6c72e350 | 1570 | bad_fork_cleanup_policy: |
1da177e4 | 1571 | #ifdef CONFIG_NUMA |
f0be3d32 | 1572 | mpol_put(p->mempolicy); |
e8604cb4 | 1573 | bad_fork_cleanup_threadgroup_lock: |
1da177e4 | 1574 | #endif |
4714d1d3 | 1575 | if (clone_flags & CLONE_THREAD) |
257058ae | 1576 | threadgroup_change_end(current); |
35df17c5 | 1577 | delayacct_tsk_free(p); |
a1261f54 | 1578 | module_put(task_thread_info(p)->exec_domain->module); |
1da177e4 | 1579 | bad_fork_cleanup_count: |
d84f4f99 | 1580 | atomic_dec(&p->cred->user->processes); |
e0e81739 | 1581 | exit_creds(p); |
1da177e4 LT |
1582 | bad_fork_free: |
1583 | free_task(p); | |
fe7d37d1 ON |
1584 | fork_out: |
1585 | return ERR_PTR(retval); | |
1da177e4 LT |
1586 | } |
1587 | ||
f106eee1 ON |
1588 | static inline void init_idle_pids(struct pid_link *links) |
1589 | { | |
1590 | enum pid_type type; | |
1591 | ||
1592 | for (type = PIDTYPE_PID; type < PIDTYPE_MAX; ++type) { | |
1593 | INIT_HLIST_NODE(&links[type].node); /* not really needed */ | |
1594 | links[type].pid = &init_struct_pid; | |
1595 | } | |
1596 | } | |
1597 | ||
0db0628d | 1598 | struct task_struct *fork_idle(int cpu) |
1da177e4 | 1599 | { |
36c8b586 | 1600 | struct task_struct *task; |
62e791c1 | 1601 | task = copy_process(CLONE_VM, 0, 0, NULL, &init_struct_pid, 0); |
f106eee1 ON |
1602 | if (!IS_ERR(task)) { |
1603 | init_idle_pids(task->pids); | |
753ca4f3 | 1604 | init_idle(task, cpu); |
f106eee1 | 1605 | } |
73b9ebfe | 1606 | |
1da177e4 LT |
1607 | return task; |
1608 | } | |
1609 | ||
1da177e4 LT |
1610 | /* |
1611 | * Ok, this is the main fork-routine. | |
1612 | * | |
1613 | * It copies the process, and if successful kick-starts | |
1614 | * it and waits for it to finish using the VM if required. | |
1615 | */ | |
1616 | long do_fork(unsigned long clone_flags, | |
1617 | unsigned long stack_start, | |
1da177e4 LT |
1618 | unsigned long stack_size, |
1619 | int __user *parent_tidptr, | |
1620 | int __user *child_tidptr) | |
1621 | { | |
1622 | struct task_struct *p; | |
1623 | int trace = 0; | |
92476d7f | 1624 | long nr; |
1da177e4 | 1625 | |
09a05394 | 1626 | /* |
4b9d33e6 TH |
1627 | * Determine whether and which event to report to ptracer. When |
1628 | * called from kernel_thread or CLONE_UNTRACED is explicitly | |
1629 | * requested, no event is reported; otherwise, report if the event | |
1630 | * for the type of forking is enabled. | |
09a05394 | 1631 | */ |
e80d6661 | 1632 | if (!(clone_flags & CLONE_UNTRACED)) { |
4b9d33e6 TH |
1633 | if (clone_flags & CLONE_VFORK) |
1634 | trace = PTRACE_EVENT_VFORK; | |
1635 | else if ((clone_flags & CSIGNAL) != SIGCHLD) | |
1636 | trace = PTRACE_EVENT_CLONE; | |
1637 | else | |
1638 | trace = PTRACE_EVENT_FORK; | |
1639 | ||
1640 | if (likely(!ptrace_event_enabled(current, trace))) | |
1641 | trace = 0; | |
1642 | } | |
1da177e4 | 1643 | |
62e791c1 | 1644 | p = copy_process(clone_flags, stack_start, stack_size, |
09a05394 | 1645 | child_tidptr, NULL, trace); |
1da177e4 LT |
1646 | /* |
1647 | * Do this prior waking up the new thread - the thread pointer | |
1648 | * might get invalid after that point, if the thread exits quickly. | |
1649 | */ | |
1650 | if (!IS_ERR(p)) { | |
1651 | struct completion vfork; | |
4e52365f | 1652 | struct pid *pid; |
1da177e4 | 1653 | |
0a16b607 MD |
1654 | trace_sched_process_fork(current, p); |
1655 | ||
4e52365f MD |
1656 | pid = get_task_pid(p, PIDTYPE_PID); |
1657 | nr = pid_vnr(pid); | |
30e49c26 PE |
1658 | |
1659 | if (clone_flags & CLONE_PARENT_SETTID) | |
1660 | put_user(nr, parent_tidptr); | |
a6f5e063 | 1661 | |
1da177e4 LT |
1662 | if (clone_flags & CLONE_VFORK) { |
1663 | p->vfork_done = &vfork; | |
1664 | init_completion(&vfork); | |
d68b46fe | 1665 | get_task_struct(p); |
1da177e4 LT |
1666 | } |
1667 | ||
3e51e3ed | 1668 | wake_up_new_task(p); |
1da177e4 | 1669 | |
4b9d33e6 TH |
1670 | /* forking complete and child started to run, tell ptracer */ |
1671 | if (unlikely(trace)) | |
4e52365f | 1672 | ptrace_event_pid(trace, pid); |
09a05394 | 1673 | |
1da177e4 | 1674 | if (clone_flags & CLONE_VFORK) { |
d68b46fe | 1675 | if (!wait_for_vfork_done(p, &vfork)) |
4e52365f | 1676 | ptrace_event_pid(PTRACE_EVENT_VFORK_DONE, pid); |
1da177e4 | 1677 | } |
4e52365f MD |
1678 | |
1679 | put_pid(pid); | |
1da177e4 | 1680 | } else { |
92476d7f | 1681 | nr = PTR_ERR(p); |
1da177e4 | 1682 | } |
92476d7f | 1683 | return nr; |
1da177e4 LT |
1684 | } |
1685 | ||
2aa3a7f8 AV |
1686 | /* |
1687 | * Create a kernel thread. | |
1688 | */ | |
1689 | pid_t kernel_thread(int (*fn)(void *), void *arg, unsigned long flags) | |
1690 | { | |
e80d6661 | 1691 | return do_fork(flags|CLONE_VM|CLONE_UNTRACED, (unsigned long)fn, |
2aa3a7f8 AV |
1692 | (unsigned long)arg, NULL, NULL); |
1693 | } | |
2aa3a7f8 | 1694 | |
d2125043 AV |
1695 | #ifdef __ARCH_WANT_SYS_FORK |
1696 | SYSCALL_DEFINE0(fork) | |
1697 | { | |
1698 | #ifdef CONFIG_MMU | |
e80d6661 | 1699 | return do_fork(SIGCHLD, 0, 0, NULL, NULL); |
d2125043 AV |
1700 | #else |
1701 | /* can not support in nommu mode */ | |
5d59e182 | 1702 | return -EINVAL; |
d2125043 AV |
1703 | #endif |
1704 | } | |
1705 | #endif | |
1706 | ||
1707 | #ifdef __ARCH_WANT_SYS_VFORK | |
1708 | SYSCALL_DEFINE0(vfork) | |
1709 | { | |
5d59e182 | 1710 | return do_fork(CLONE_VFORK | CLONE_VM | SIGCHLD, 0, |
d2125043 AV |
1711 | 0, NULL, NULL); |
1712 | } | |
1713 | #endif | |
1714 | ||
1715 | #ifdef __ARCH_WANT_SYS_CLONE | |
1716 | #ifdef CONFIG_CLONE_BACKWARDS | |
1717 | SYSCALL_DEFINE5(clone, unsigned long, clone_flags, unsigned long, newsp, | |
1718 | int __user *, parent_tidptr, | |
1719 | int, tls_val, | |
1720 | int __user *, child_tidptr) | |
1721 | #elif defined(CONFIG_CLONE_BACKWARDS2) | |
1722 | SYSCALL_DEFINE5(clone, unsigned long, newsp, unsigned long, clone_flags, | |
1723 | int __user *, parent_tidptr, | |
1724 | int __user *, child_tidptr, | |
1725 | int, tls_val) | |
dfa9771a MS |
1726 | #elif defined(CONFIG_CLONE_BACKWARDS3) |
1727 | SYSCALL_DEFINE6(clone, unsigned long, clone_flags, unsigned long, newsp, | |
1728 | int, stack_size, | |
1729 | int __user *, parent_tidptr, | |
1730 | int __user *, child_tidptr, | |
1731 | int, tls_val) | |
d2125043 AV |
1732 | #else |
1733 | SYSCALL_DEFINE5(clone, unsigned long, clone_flags, unsigned long, newsp, | |
1734 | int __user *, parent_tidptr, | |
1735 | int __user *, child_tidptr, | |
1736 | int, tls_val) | |
1737 | #endif | |
1738 | { | |
2cf09666 | 1739 | return do_fork(clone_flags, newsp, 0, parent_tidptr, child_tidptr); |
d2125043 AV |
1740 | } |
1741 | #endif | |
1742 | ||
5fd63b30 RT |
1743 | #ifndef ARCH_MIN_MMSTRUCT_ALIGN |
1744 | #define ARCH_MIN_MMSTRUCT_ALIGN 0 | |
1745 | #endif | |
1746 | ||
51cc5068 | 1747 | static void sighand_ctor(void *data) |
aa1757f9 ON |
1748 | { |
1749 | struct sighand_struct *sighand = data; | |
1750 | ||
a35afb83 | 1751 | spin_lock_init(&sighand->siglock); |
b8fceee1 | 1752 | init_waitqueue_head(&sighand->signalfd_wqh); |
aa1757f9 ON |
1753 | } |
1754 | ||
1da177e4 LT |
1755 | void __init proc_caches_init(void) |
1756 | { | |
1757 | sighand_cachep = kmem_cache_create("sighand_cache", | |
1758 | sizeof(struct sighand_struct), 0, | |
2dff4405 VN |
1759 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_DESTROY_BY_RCU| |
1760 | SLAB_NOTRACK, sighand_ctor); | |
1da177e4 LT |
1761 | signal_cachep = kmem_cache_create("signal_cache", |
1762 | sizeof(struct signal_struct), 0, | |
2dff4405 | 1763 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL); |
20c2df83 | 1764 | files_cachep = kmem_cache_create("files_cache", |
1da177e4 | 1765 | sizeof(struct files_struct), 0, |
2dff4405 | 1766 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL); |
20c2df83 | 1767 | fs_cachep = kmem_cache_create("fs_cache", |
1da177e4 | 1768 | sizeof(struct fs_struct), 0, |
2dff4405 | 1769 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL); |
6345d24d LT |
1770 | /* |
1771 | * FIXME! The "sizeof(struct mm_struct)" currently includes the | |
1772 | * whole struct cpumask for the OFFSTACK case. We could change | |
1773 | * this to *only* allocate as much of it as required by the | |
1774 | * maximum number of CPU's we can ever have. The cpumask_allocation | |
1775 | * is at the end of the structure, exactly for that reason. | |
1776 | */ | |
1da177e4 | 1777 | mm_cachep = kmem_cache_create("mm_struct", |
5fd63b30 | 1778 | sizeof(struct mm_struct), ARCH_MIN_MMSTRUCT_ALIGN, |
2dff4405 | 1779 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL); |
33e5d769 | 1780 | vm_area_cachep = KMEM_CACHE(vm_area_struct, SLAB_PANIC); |
8feae131 | 1781 | mmap_init(); |
66577193 | 1782 | nsproxy_cache_init(); |
1da177e4 | 1783 | } |
cf2e340f | 1784 | |
cf2e340f | 1785 | /* |
9bfb23fc | 1786 | * Check constraints on flags passed to the unshare system call. |
cf2e340f | 1787 | */ |
9bfb23fc | 1788 | static int check_unshare_flags(unsigned long unshare_flags) |
cf2e340f | 1789 | { |
9bfb23fc ON |
1790 | if (unshare_flags & ~(CLONE_THREAD|CLONE_FS|CLONE_NEWNS|CLONE_SIGHAND| |
1791 | CLONE_VM|CLONE_FILES|CLONE_SYSVSEM| | |
50804fe3 | 1792 | CLONE_NEWUTS|CLONE_NEWIPC|CLONE_NEWNET| |
b2e0d987 | 1793 | CLONE_NEWUSER|CLONE_NEWPID)) |
9bfb23fc | 1794 | return -EINVAL; |
cf2e340f | 1795 | /* |
9bfb23fc ON |
1796 | * Not implemented, but pretend it works if there is nothing to |
1797 | * unshare. Note that unsharing CLONE_THREAD or CLONE_SIGHAND | |
1798 | * needs to unshare vm. | |
cf2e340f | 1799 | */ |
9bfb23fc ON |
1800 | if (unshare_flags & (CLONE_THREAD | CLONE_SIGHAND | CLONE_VM)) { |
1801 | /* FIXME: get_task_mm() increments ->mm_users */ | |
1802 | if (atomic_read(¤t->mm->mm_users) > 1) | |
1803 | return -EINVAL; | |
1804 | } | |
cf2e340f JD |
1805 | |
1806 | return 0; | |
1807 | } | |
1808 | ||
1809 | /* | |
99d1419d | 1810 | * Unshare the filesystem structure if it is being shared |
cf2e340f JD |
1811 | */ |
1812 | static int unshare_fs(unsigned long unshare_flags, struct fs_struct **new_fsp) | |
1813 | { | |
1814 | struct fs_struct *fs = current->fs; | |
1815 | ||
498052bb AV |
1816 | if (!(unshare_flags & CLONE_FS) || !fs) |
1817 | return 0; | |
1818 | ||
1819 | /* don't need lock here; in the worst case we'll do useless copy */ | |
1820 | if (fs->users == 1) | |
1821 | return 0; | |
1822 | ||
1823 | *new_fsp = copy_fs_struct(fs); | |
1824 | if (!*new_fsp) | |
1825 | return -ENOMEM; | |
cf2e340f JD |
1826 | |
1827 | return 0; | |
1828 | } | |
1829 | ||
cf2e340f | 1830 | /* |
a016f338 | 1831 | * Unshare file descriptor table if it is being shared |
cf2e340f JD |
1832 | */ |
1833 | static int unshare_fd(unsigned long unshare_flags, struct files_struct **new_fdp) | |
1834 | { | |
1835 | struct files_struct *fd = current->files; | |
a016f338 | 1836 | int error = 0; |
cf2e340f JD |
1837 | |
1838 | if ((unshare_flags & CLONE_FILES) && | |
a016f338 JD |
1839 | (fd && atomic_read(&fd->count) > 1)) { |
1840 | *new_fdp = dup_fd(fd, &error); | |
1841 | if (!*new_fdp) | |
1842 | return error; | |
1843 | } | |
cf2e340f JD |
1844 | |
1845 | return 0; | |
1846 | } | |
1847 | ||
cf2e340f JD |
1848 | /* |
1849 | * unshare allows a process to 'unshare' part of the process | |
1850 | * context which was originally shared using clone. copy_* | |
1851 | * functions used by do_fork() cannot be used here directly | |
1852 | * because they modify an inactive task_struct that is being | |
1853 | * constructed. Here we are modifying the current, active, | |
1854 | * task_struct. | |
1855 | */ | |
6559eed8 | 1856 | SYSCALL_DEFINE1(unshare, unsigned long, unshare_flags) |
cf2e340f | 1857 | { |
cf2e340f | 1858 | struct fs_struct *fs, *new_fs = NULL; |
cf2e340f | 1859 | struct files_struct *fd, *new_fd = NULL; |
b2e0d987 | 1860 | struct cred *new_cred = NULL; |
cf7b708c | 1861 | struct nsproxy *new_nsproxy = NULL; |
9edff4ab | 1862 | int do_sysvsem = 0; |
9bfb23fc | 1863 | int err; |
cf2e340f | 1864 | |
b2e0d987 EB |
1865 | /* |
1866 | * If unsharing a user namespace must also unshare the thread. | |
1867 | */ | |
1868 | if (unshare_flags & CLONE_NEWUSER) | |
e66eded8 | 1869 | unshare_flags |= CLONE_THREAD | CLONE_FS; |
50804fe3 EB |
1870 | /* |
1871 | * If unsharing a thread from a thread group, must also unshare vm. | |
1872 | */ | |
1873 | if (unshare_flags & CLONE_THREAD) | |
1874 | unshare_flags |= CLONE_VM; | |
1875 | /* | |
1876 | * If unsharing vm, must also unshare signal handlers. | |
1877 | */ | |
1878 | if (unshare_flags & CLONE_VM) | |
1879 | unshare_flags |= CLONE_SIGHAND; | |
9bfb23fc ON |
1880 | /* |
1881 | * If unsharing namespace, must also unshare filesystem information. | |
1882 | */ | |
1883 | if (unshare_flags & CLONE_NEWNS) | |
1884 | unshare_flags |= CLONE_FS; | |
50804fe3 EB |
1885 | |
1886 | err = check_unshare_flags(unshare_flags); | |
1887 | if (err) | |
1888 | goto bad_unshare_out; | |
6013f67f MS |
1889 | /* |
1890 | * CLONE_NEWIPC must also detach from the undolist: after switching | |
1891 | * to a new ipc namespace, the semaphore arrays from the old | |
1892 | * namespace are unreachable. | |
1893 | */ | |
1894 | if (unshare_flags & (CLONE_NEWIPC|CLONE_SYSVSEM)) | |
9edff4ab | 1895 | do_sysvsem = 1; |
fb0a685c DRO |
1896 | err = unshare_fs(unshare_flags, &new_fs); |
1897 | if (err) | |
9bfb23fc | 1898 | goto bad_unshare_out; |
fb0a685c DRO |
1899 | err = unshare_fd(unshare_flags, &new_fd); |
1900 | if (err) | |
9bfb23fc | 1901 | goto bad_unshare_cleanup_fs; |
b2e0d987 | 1902 | err = unshare_userns(unshare_flags, &new_cred); |
fb0a685c | 1903 | if (err) |
9edff4ab | 1904 | goto bad_unshare_cleanup_fd; |
b2e0d987 EB |
1905 | err = unshare_nsproxy_namespaces(unshare_flags, &new_nsproxy, |
1906 | new_cred, new_fs); | |
1907 | if (err) | |
1908 | goto bad_unshare_cleanup_cred; | |
c0b2fc31 | 1909 | |
b2e0d987 | 1910 | if (new_fs || new_fd || do_sysvsem || new_cred || new_nsproxy) { |
9edff4ab MS |
1911 | if (do_sysvsem) { |
1912 | /* | |
1913 | * CLONE_SYSVSEM is equivalent to sys_exit(). | |
1914 | */ | |
1915 | exit_sem(current); | |
1916 | } | |
ab602f79 JM |
1917 | if (unshare_flags & CLONE_NEWIPC) { |
1918 | /* Orphan segments in old ns (see sem above). */ | |
1919 | exit_shm(current); | |
1920 | shm_init_task(current); | |
1921 | } | |
ab516013 | 1922 | |
6f977e6b | 1923 | if (new_nsproxy) |
cf7b708c | 1924 | switch_task_namespaces(current, new_nsproxy); |
cf2e340f | 1925 | |
cf7b708c PE |
1926 | task_lock(current); |
1927 | ||
cf2e340f JD |
1928 | if (new_fs) { |
1929 | fs = current->fs; | |
2a4419b5 | 1930 | spin_lock(&fs->lock); |
cf2e340f | 1931 | current->fs = new_fs; |
498052bb AV |
1932 | if (--fs->users) |
1933 | new_fs = NULL; | |
1934 | else | |
1935 | new_fs = fs; | |
2a4419b5 | 1936 | spin_unlock(&fs->lock); |
cf2e340f JD |
1937 | } |
1938 | ||
cf2e340f JD |
1939 | if (new_fd) { |
1940 | fd = current->files; | |
1941 | current->files = new_fd; | |
1942 | new_fd = fd; | |
1943 | } | |
1944 | ||
1945 | task_unlock(current); | |
b2e0d987 EB |
1946 | |
1947 | if (new_cred) { | |
1948 | /* Install the new user namespace */ | |
1949 | commit_creds(new_cred); | |
1950 | new_cred = NULL; | |
1951 | } | |
cf2e340f JD |
1952 | } |
1953 | ||
b2e0d987 EB |
1954 | bad_unshare_cleanup_cred: |
1955 | if (new_cred) | |
1956 | put_cred(new_cred); | |
cf2e340f JD |
1957 | bad_unshare_cleanup_fd: |
1958 | if (new_fd) | |
1959 | put_files_struct(new_fd); | |
1960 | ||
cf2e340f JD |
1961 | bad_unshare_cleanup_fs: |
1962 | if (new_fs) | |
498052bb | 1963 | free_fs_struct(new_fs); |
cf2e340f | 1964 | |
cf2e340f JD |
1965 | bad_unshare_out: |
1966 | return err; | |
1967 | } | |
3b125388 AV |
1968 | |
1969 | /* | |
1970 | * Helper to unshare the files of the current task. | |
1971 | * We don't want to expose copy_files internals to | |
1972 | * the exec layer of the kernel. | |
1973 | */ | |
1974 | ||
1975 | int unshare_files(struct files_struct **displaced) | |
1976 | { | |
1977 | struct task_struct *task = current; | |
50704516 | 1978 | struct files_struct *copy = NULL; |
3b125388 AV |
1979 | int error; |
1980 | ||
1981 | error = unshare_fd(CLONE_FILES, ©); | |
1982 | if (error || !copy) { | |
1983 | *displaced = NULL; | |
1984 | return error; | |
1985 | } | |
1986 | *displaced = task->files; | |
1987 | task_lock(task); | |
1988 | task->files = copy; | |
1989 | task_unlock(task); | |
1990 | return 0; | |
1991 | } |