]>
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> |
1da177e4 LT |
76 | |
77 | #include <asm/pgtable.h> | |
78 | #include <asm/pgalloc.h> | |
79 | #include <asm/uaccess.h> | |
80 | #include <asm/mmu_context.h> | |
81 | #include <asm/cacheflush.h> | |
82 | #include <asm/tlbflush.h> | |
83 | ||
ad8d75ff SR |
84 | #include <trace/events/sched.h> |
85 | ||
43d2b113 KH |
86 | #define CREATE_TRACE_POINTS |
87 | #include <trace/events/task.h> | |
88 | ||
1da177e4 LT |
89 | /* |
90 | * Protected counters by write_lock_irq(&tasklist_lock) | |
91 | */ | |
92 | unsigned long total_forks; /* Handle normal Linux uptimes. */ | |
fb0a685c | 93 | int nr_threads; /* The idle threads do not count.. */ |
1da177e4 LT |
94 | |
95 | int max_threads; /* tunable limit on nr_threads */ | |
96 | ||
97 | DEFINE_PER_CPU(unsigned long, process_counts) = 0; | |
98 | ||
c59923a1 | 99 | __cacheline_aligned DEFINE_RWLOCK(tasklist_lock); /* outer */ |
db1466b3 PM |
100 | |
101 | #ifdef CONFIG_PROVE_RCU | |
102 | int lockdep_tasklist_lock_is_held(void) | |
103 | { | |
104 | return lockdep_is_held(&tasklist_lock); | |
105 | } | |
106 | EXPORT_SYMBOL_GPL(lockdep_tasklist_lock_is_held); | |
107 | #endif /* #ifdef CONFIG_PROVE_RCU */ | |
1da177e4 LT |
108 | |
109 | int nr_processes(void) | |
110 | { | |
111 | int cpu; | |
112 | int total = 0; | |
113 | ||
1d510750 | 114 | for_each_possible_cpu(cpu) |
1da177e4 LT |
115 | total += per_cpu(process_counts, cpu); |
116 | ||
117 | return total; | |
118 | } | |
119 | ||
f19b9f74 AM |
120 | void __weak arch_release_task_struct(struct task_struct *tsk) |
121 | { | |
122 | } | |
123 | ||
f5e10287 | 124 | #ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR |
e18b890b | 125 | static struct kmem_cache *task_struct_cachep; |
41101809 TG |
126 | |
127 | static inline struct task_struct *alloc_task_struct_node(int node) | |
128 | { | |
129 | return kmem_cache_alloc_node(task_struct_cachep, GFP_KERNEL, node); | |
130 | } | |
131 | ||
41101809 TG |
132 | static inline void free_task_struct(struct task_struct *tsk) |
133 | { | |
41101809 TG |
134 | kmem_cache_free(task_struct_cachep, tsk); |
135 | } | |
1da177e4 LT |
136 | #endif |
137 | ||
f19b9f74 AM |
138 | void __weak arch_release_thread_info(struct thread_info *ti) |
139 | { | |
140 | } | |
141 | ||
f5e10287 | 142 | #ifndef CONFIG_ARCH_THREAD_INFO_ALLOCATOR |
41101809 | 143 | |
0d15d74a TG |
144 | /* |
145 | * Allocate pages if THREAD_SIZE is >= PAGE_SIZE, otherwise use a | |
146 | * kmemcache based allocator. | |
147 | */ | |
148 | # if THREAD_SIZE >= PAGE_SIZE | |
b6a84016 ED |
149 | static struct thread_info *alloc_thread_info_node(struct task_struct *tsk, |
150 | int node) | |
b69c49b7 | 151 | { |
2ad306b1 | 152 | struct page *page = alloc_pages_node(node, THREADINFO_GFP_ACCOUNTED, |
2889f608 | 153 | THREAD_SIZE_ORDER); |
b6a84016 ED |
154 | |
155 | return page ? page_address(page) : NULL; | |
b69c49b7 FT |
156 | } |
157 | ||
158 | static inline void free_thread_info(struct thread_info *ti) | |
159 | { | |
2ad306b1 | 160 | free_memcg_kmem_pages((unsigned long)ti, THREAD_SIZE_ORDER); |
b69c49b7 | 161 | } |
0d15d74a TG |
162 | # else |
163 | static struct kmem_cache *thread_info_cache; | |
164 | ||
165 | static struct thread_info *alloc_thread_info_node(struct task_struct *tsk, | |
166 | int node) | |
167 | { | |
168 | return kmem_cache_alloc_node(thread_info_cache, THREADINFO_GFP, node); | |
169 | } | |
170 | ||
171 | static void free_thread_info(struct thread_info *ti) | |
172 | { | |
0d15d74a TG |
173 | kmem_cache_free(thread_info_cache, ti); |
174 | } | |
175 | ||
176 | void thread_info_cache_init(void) | |
177 | { | |
178 | thread_info_cache = kmem_cache_create("thread_info", THREAD_SIZE, | |
179 | THREAD_SIZE, 0, NULL); | |
180 | BUG_ON(thread_info_cache == NULL); | |
181 | } | |
182 | # endif | |
b69c49b7 FT |
183 | #endif |
184 | ||
1da177e4 | 185 | /* SLAB cache for signal_struct structures (tsk->signal) */ |
e18b890b | 186 | static struct kmem_cache *signal_cachep; |
1da177e4 LT |
187 | |
188 | /* SLAB cache for sighand_struct structures (tsk->sighand) */ | |
e18b890b | 189 | struct kmem_cache *sighand_cachep; |
1da177e4 LT |
190 | |
191 | /* SLAB cache for files_struct structures (tsk->files) */ | |
e18b890b | 192 | struct kmem_cache *files_cachep; |
1da177e4 LT |
193 | |
194 | /* SLAB cache for fs_struct structures (tsk->fs) */ | |
e18b890b | 195 | struct kmem_cache *fs_cachep; |
1da177e4 LT |
196 | |
197 | /* SLAB cache for vm_area_struct structures */ | |
e18b890b | 198 | struct kmem_cache *vm_area_cachep; |
1da177e4 LT |
199 | |
200 | /* SLAB cache for mm_struct structures (tsk->mm) */ | |
e18b890b | 201 | static struct kmem_cache *mm_cachep; |
1da177e4 | 202 | |
c6a7f572 KM |
203 | static void account_kernel_stack(struct thread_info *ti, int account) |
204 | { | |
205 | struct zone *zone = page_zone(virt_to_page(ti)); | |
206 | ||
207 | mod_zone_page_state(zone, NR_KERNEL_STACK, account); | |
208 | } | |
209 | ||
1da177e4 LT |
210 | void free_task(struct task_struct *tsk) |
211 | { | |
c6a7f572 | 212 | account_kernel_stack(tsk->stack, -1); |
f19b9f74 | 213 | arch_release_thread_info(tsk->stack); |
f7e4217b | 214 | free_thread_info(tsk->stack); |
23f78d4a | 215 | rt_mutex_debug_task_free(tsk); |
fb52607a | 216 | ftrace_graph_exit_task(tsk); |
e2cfabdf | 217 | put_seccomp_filter(tsk); |
f19b9f74 | 218 | arch_release_task_struct(tsk); |
1da177e4 LT |
219 | free_task_struct(tsk); |
220 | } | |
221 | EXPORT_SYMBOL(free_task); | |
222 | ||
ea6d290c ON |
223 | static inline void free_signal_struct(struct signal_struct *sig) |
224 | { | |
97101eb4 | 225 | taskstats_tgid_free(sig); |
1c5354de | 226 | sched_autogroup_exit(sig); |
ea6d290c ON |
227 | kmem_cache_free(signal_cachep, sig); |
228 | } | |
229 | ||
230 | static inline void put_signal_struct(struct signal_struct *sig) | |
231 | { | |
1c5354de | 232 | if (atomic_dec_and_test(&sig->sigcnt)) |
ea6d290c ON |
233 | free_signal_struct(sig); |
234 | } | |
235 | ||
158d9ebd | 236 | void __put_task_struct(struct task_struct *tsk) |
1da177e4 | 237 | { |
270f722d | 238 | WARN_ON(!tsk->exit_state); |
1da177e4 LT |
239 | WARN_ON(atomic_read(&tsk->usage)); |
240 | WARN_ON(tsk == current); | |
241 | ||
156654f4 | 242 | task_numa_free(tsk); |
1a2a4d06 | 243 | security_task_free(tsk); |
e0e81739 | 244 | exit_creds(tsk); |
35df17c5 | 245 | delayacct_tsk_free(tsk); |
ea6d290c | 246 | put_signal_struct(tsk->signal); |
1da177e4 LT |
247 | |
248 | if (!profile_handoff_task(tsk)) | |
249 | free_task(tsk); | |
250 | } | |
77c100c8 | 251 | EXPORT_SYMBOL_GPL(__put_task_struct); |
1da177e4 | 252 | |
6c0a9fa6 | 253 | void __init __weak arch_task_cache_init(void) { } |
61c4628b | 254 | |
1da177e4 LT |
255 | void __init fork_init(unsigned long mempages) |
256 | { | |
f5e10287 | 257 | #ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR |
1da177e4 LT |
258 | #ifndef ARCH_MIN_TASKALIGN |
259 | #define ARCH_MIN_TASKALIGN L1_CACHE_BYTES | |
260 | #endif | |
261 | /* create a slab on which task_structs can be allocated */ | |
262 | task_struct_cachep = | |
263 | kmem_cache_create("task_struct", sizeof(struct task_struct), | |
2dff4405 | 264 | ARCH_MIN_TASKALIGN, SLAB_PANIC | SLAB_NOTRACK, NULL); |
1da177e4 LT |
265 | #endif |
266 | ||
61c4628b SS |
267 | /* do the arch specific task caches init */ |
268 | arch_task_cache_init(); | |
269 | ||
1da177e4 LT |
270 | /* |
271 | * The default maximum number of threads is set to a safe | |
272 | * value: the thread structures can take up at most half | |
273 | * of memory. | |
274 | */ | |
275 | max_threads = mempages / (8 * THREAD_SIZE / PAGE_SIZE); | |
276 | ||
277 | /* | |
278 | * we need to allow at least 20 threads to boot a system | |
279 | */ | |
fb0a685c | 280 | if (max_threads < 20) |
1da177e4 LT |
281 | max_threads = 20; |
282 | ||
283 | init_task.signal->rlim[RLIMIT_NPROC].rlim_cur = max_threads/2; | |
284 | init_task.signal->rlim[RLIMIT_NPROC].rlim_max = max_threads/2; | |
285 | init_task.signal->rlim[RLIMIT_SIGPENDING] = | |
286 | init_task.signal->rlim[RLIMIT_NPROC]; | |
287 | } | |
288 | ||
61c4628b SS |
289 | int __attribute__((weak)) arch_dup_task_struct(struct task_struct *dst, |
290 | struct task_struct *src) | |
291 | { | |
292 | *dst = *src; | |
293 | return 0; | |
294 | } | |
295 | ||
1da177e4 LT |
296 | static struct task_struct *dup_task_struct(struct task_struct *orig) |
297 | { | |
298 | struct task_struct *tsk; | |
299 | struct thread_info *ti; | |
7c9f8861 | 300 | unsigned long *stackend; |
207205a2 | 301 | int node = tsk_fork_get_node(orig); |
3e26c149 | 302 | int err; |
1da177e4 | 303 | |
504f52b5 | 304 | tsk = alloc_task_struct_node(node); |
1da177e4 LT |
305 | if (!tsk) |
306 | return NULL; | |
307 | ||
b6a84016 | 308 | ti = alloc_thread_info_node(tsk, node); |
f19b9f74 AM |
309 | if (!ti) |
310 | goto free_tsk; | |
1da177e4 | 311 | |
fb0a685c | 312 | err = arch_dup_task_struct(tsk, orig); |
164c33c6 | 313 | if (err) |
f19b9f74 | 314 | goto free_ti; |
164c33c6 | 315 | |
87bec58a AM |
316 | tsk->stack = ti; |
317 | ||
318 | setup_thread_stack(tsk, orig); | |
8e7cac79 | 319 | clear_user_return_notifier(tsk); |
f26f9aff | 320 | clear_tsk_need_resched(tsk); |
7c9f8861 ES |
321 | stackend = end_of_stack(tsk); |
322 | *stackend = STACK_END_MAGIC; /* for overflow detection */ | |
1da177e4 | 323 | |
0a425405 AV |
324 | #ifdef CONFIG_CC_STACKPROTECTOR |
325 | tsk->stack_canary = get_random_int(); | |
326 | #endif | |
327 | ||
fb0a685c DRO |
328 | /* |
329 | * One for us, one for whoever does the "release_task()" (usually | |
330 | * parent) | |
331 | */ | |
332 | atomic_set(&tsk->usage, 2); | |
6c5c9341 | 333 | #ifdef CONFIG_BLK_DEV_IO_TRACE |
2056a782 | 334 | tsk->btrace_seq = 0; |
6c5c9341 | 335 | #endif |
a0aa7f68 | 336 | tsk->splice_pipe = NULL; |
5640f768 | 337 | tsk->task_frag.page = NULL; |
c6a7f572 KM |
338 | |
339 | account_kernel_stack(ti, 1); | |
340 | ||
1da177e4 | 341 | return tsk; |
61c4628b | 342 | |
f19b9f74 | 343 | free_ti: |
61c4628b | 344 | free_thread_info(ti); |
f19b9f74 | 345 | free_tsk: |
61c4628b SS |
346 | free_task_struct(tsk); |
347 | return NULL; | |
1da177e4 LT |
348 | } |
349 | ||
350 | #ifdef CONFIG_MMU | |
a39bc516 | 351 | static int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm) |
1da177e4 | 352 | { |
297c5eee | 353 | struct vm_area_struct *mpnt, *tmp, *prev, **pprev; |
1da177e4 LT |
354 | struct rb_node **rb_link, *rb_parent; |
355 | int retval; | |
356 | unsigned long charge; | |
1da177e4 | 357 | |
32cdba1e | 358 | uprobe_start_dup_mmap(); |
1da177e4 | 359 | down_write(&oldmm->mmap_sem); |
ec8c0446 | 360 | flush_cache_dup_mm(oldmm); |
f8ac4ec9 | 361 | uprobe_dup_mmap(oldmm, mm); |
ad339451 IM |
362 | /* |
363 | * Not linked in yet - no deadlock potential: | |
364 | */ | |
365 | down_write_nested(&mm->mmap_sem, SINGLE_DEPTH_NESTING); | |
7ee78232 | 366 | |
1da177e4 LT |
367 | mm->locked_vm = 0; |
368 | mm->mmap = NULL; | |
615d6e87 | 369 | mm->vmacache_seqnum = 0; |
1da177e4 | 370 | mm->map_count = 0; |
94894244 | 371 | cpumask_clear(mm_cpumask(mm)); |
1da177e4 LT |
372 | mm->mm_rb = RB_ROOT; |
373 | rb_link = &mm->mm_rb.rb_node; | |
374 | rb_parent = NULL; | |
375 | pprev = &mm->mmap; | |
f8af4da3 | 376 | retval = ksm_fork(mm, oldmm); |
ba76149f AA |
377 | if (retval) |
378 | goto out; | |
379 | retval = khugepaged_fork(mm, oldmm); | |
f8af4da3 HD |
380 | if (retval) |
381 | goto out; | |
1da177e4 | 382 | |
297c5eee | 383 | prev = NULL; |
fd3e42fc | 384 | for (mpnt = oldmm->mmap; mpnt; mpnt = mpnt->vm_next) { |
1da177e4 LT |
385 | struct file *file; |
386 | ||
387 | if (mpnt->vm_flags & VM_DONTCOPY) { | |
ab50b8ed | 388 | vm_stat_account(mm, mpnt->vm_flags, mpnt->vm_file, |
44de9d0c | 389 | -vma_pages(mpnt)); |
1da177e4 LT |
390 | continue; |
391 | } | |
392 | charge = 0; | |
393 | if (mpnt->vm_flags & VM_ACCOUNT) { | |
b2412b7f HS |
394 | unsigned long len = vma_pages(mpnt); |
395 | ||
191c5424 | 396 | if (security_vm_enough_memory_mm(oldmm, len)) /* sic */ |
1da177e4 LT |
397 | goto fail_nomem; |
398 | charge = len; | |
399 | } | |
e94b1766 | 400 | tmp = kmem_cache_alloc(vm_area_cachep, GFP_KERNEL); |
1da177e4 LT |
401 | if (!tmp) |
402 | goto fail_nomem; | |
403 | *tmp = *mpnt; | |
5beb4930 | 404 | INIT_LIST_HEAD(&tmp->anon_vma_chain); |
ef0855d3 ON |
405 | retval = vma_dup_policy(mpnt, tmp); |
406 | if (retval) | |
1da177e4 | 407 | goto fail_nomem_policy; |
a247c3a9 | 408 | tmp->vm_mm = mm; |
5beb4930 RR |
409 | if (anon_vma_fork(tmp, mpnt)) |
410 | goto fail_nomem_anon_vma_fork; | |
1da177e4 | 411 | tmp->vm_flags &= ~VM_LOCKED; |
297c5eee | 412 | tmp->vm_next = tmp->vm_prev = NULL; |
1da177e4 LT |
413 | file = tmp->vm_file; |
414 | if (file) { | |
496ad9aa | 415 | struct inode *inode = file_inode(file); |
b88ed205 HD |
416 | struct address_space *mapping = file->f_mapping; |
417 | ||
1da177e4 LT |
418 | get_file(file); |
419 | if (tmp->vm_flags & VM_DENYWRITE) | |
420 | atomic_dec(&inode->i_writecount); | |
3d48ae45 | 421 | mutex_lock(&mapping->i_mmap_mutex); |
b88ed205 HD |
422 | if (tmp->vm_flags & VM_SHARED) |
423 | mapping->i_mmap_writable++; | |
b88ed205 HD |
424 | flush_dcache_mmap_lock(mapping); |
425 | /* insert tmp into the share list, just after mpnt */ | |
9826a516 ML |
426 | if (unlikely(tmp->vm_flags & VM_NONLINEAR)) |
427 | vma_nonlinear_insert(tmp, | |
428 | &mapping->i_mmap_nonlinear); | |
429 | else | |
430 | vma_interval_tree_insert_after(tmp, mpnt, | |
431 | &mapping->i_mmap); | |
b88ed205 | 432 | flush_dcache_mmap_unlock(mapping); |
3d48ae45 | 433 | mutex_unlock(&mapping->i_mmap_mutex); |
1da177e4 LT |
434 | } |
435 | ||
a1e78772 MG |
436 | /* |
437 | * Clear hugetlb-related page reserves for children. This only | |
438 | * affects MAP_PRIVATE mappings. Faults generated by the child | |
439 | * are not guaranteed to succeed, even if read-only | |
440 | */ | |
441 | if (is_vm_hugetlb_page(tmp)) | |
442 | reset_vma_resv_huge_pages(tmp); | |
443 | ||
1da177e4 | 444 | /* |
7ee78232 | 445 | * Link in the new vma and copy the page table entries. |
1da177e4 | 446 | */ |
1da177e4 LT |
447 | *pprev = tmp; |
448 | pprev = &tmp->vm_next; | |
297c5eee LT |
449 | tmp->vm_prev = prev; |
450 | prev = tmp; | |
1da177e4 LT |
451 | |
452 | __vma_link_rb(mm, tmp, rb_link, rb_parent); | |
453 | rb_link = &tmp->vm_rb.rb_right; | |
454 | rb_parent = &tmp->vm_rb; | |
455 | ||
456 | mm->map_count++; | |
0b0db14c | 457 | retval = copy_page_range(mm, oldmm, mpnt); |
1da177e4 LT |
458 | |
459 | if (tmp->vm_ops && tmp->vm_ops->open) | |
460 | tmp->vm_ops->open(tmp); | |
461 | ||
462 | if (retval) | |
463 | goto out; | |
464 | } | |
d6dd61c8 JF |
465 | /* a new mm has just been created */ |
466 | arch_dup_mmap(oldmm, mm); | |
1da177e4 | 467 | retval = 0; |
1da177e4 | 468 | out: |
7ee78232 | 469 | up_write(&mm->mmap_sem); |
fd3e42fc | 470 | flush_tlb_mm(oldmm); |
1da177e4 | 471 | up_write(&oldmm->mmap_sem); |
32cdba1e | 472 | uprobe_end_dup_mmap(); |
1da177e4 | 473 | return retval; |
5beb4930 | 474 | fail_nomem_anon_vma_fork: |
ef0855d3 | 475 | mpol_put(vma_policy(tmp)); |
1da177e4 LT |
476 | fail_nomem_policy: |
477 | kmem_cache_free(vm_area_cachep, tmp); | |
478 | fail_nomem: | |
479 | retval = -ENOMEM; | |
480 | vm_unacct_memory(charge); | |
481 | goto out; | |
482 | } | |
483 | ||
fb0a685c | 484 | static inline int mm_alloc_pgd(struct mm_struct *mm) |
1da177e4 LT |
485 | { |
486 | mm->pgd = pgd_alloc(mm); | |
487 | if (unlikely(!mm->pgd)) | |
488 | return -ENOMEM; | |
489 | return 0; | |
490 | } | |
491 | ||
fb0a685c | 492 | static inline void mm_free_pgd(struct mm_struct *mm) |
1da177e4 | 493 | { |
5e541973 | 494 | pgd_free(mm, mm->pgd); |
1da177e4 LT |
495 | } |
496 | #else | |
497 | #define dup_mmap(mm, oldmm) (0) | |
498 | #define mm_alloc_pgd(mm) (0) | |
499 | #define mm_free_pgd(mm) | |
500 | #endif /* CONFIG_MMU */ | |
501 | ||
23ff4440 | 502 | __cacheline_aligned_in_smp DEFINE_SPINLOCK(mmlist_lock); |
1da177e4 | 503 | |
e94b1766 | 504 | #define allocate_mm() (kmem_cache_alloc(mm_cachep, GFP_KERNEL)) |
1da177e4 LT |
505 | #define free_mm(mm) (kmem_cache_free(mm_cachep, (mm))) |
506 | ||
4cb0e11b HK |
507 | static unsigned long default_dump_filter = MMF_DUMP_FILTER_DEFAULT; |
508 | ||
509 | static int __init coredump_filter_setup(char *s) | |
510 | { | |
511 | default_dump_filter = | |
512 | (simple_strtoul(s, NULL, 0) << MMF_DUMP_FILTER_SHIFT) & | |
513 | MMF_DUMP_FILTER_MASK; | |
514 | return 1; | |
515 | } | |
516 | ||
517 | __setup("coredump_filter=", coredump_filter_setup); | |
518 | ||
1da177e4 LT |
519 | #include <linux/init_task.h> |
520 | ||
858f0993 AD |
521 | static void mm_init_aio(struct mm_struct *mm) |
522 | { | |
523 | #ifdef CONFIG_AIO | |
524 | spin_lock_init(&mm->ioctx_lock); | |
db446a08 | 525 | mm->ioctx_table = NULL; |
858f0993 AD |
526 | #endif |
527 | } | |
528 | ||
fb0a685c | 529 | static struct mm_struct *mm_init(struct mm_struct *mm, struct task_struct *p) |
1da177e4 LT |
530 | { |
531 | atomic_set(&mm->mm_users, 1); | |
532 | atomic_set(&mm->mm_count, 1); | |
533 | init_rwsem(&mm->mmap_sem); | |
534 | INIT_LIST_HEAD(&mm->mmlist); | |
999d9fc1 | 535 | mm->core_state = NULL; |
e1f56c89 | 536 | atomic_long_set(&mm->nr_ptes, 0); |
d559db08 | 537 | memset(&mm->rss_stat, 0, sizeof(mm->rss_stat)); |
1da177e4 | 538 | spin_lock_init(&mm->page_table_lock); |
858f0993 | 539 | mm_init_aio(mm); |
cf475ad2 | 540 | mm_init_owner(mm, p); |
20841405 | 541 | clear_tlb_flush_pending(mm); |
1da177e4 | 542 | |
a0715cc2 AT |
543 | if (current->mm) { |
544 | mm->flags = current->mm->flags & MMF_INIT_MASK; | |
545 | mm->def_flags = current->mm->def_flags & VM_INIT_DEF_MASK; | |
546 | } else { | |
547 | mm->flags = default_dump_filter; | |
1da177e4 | 548 | mm->def_flags = 0; |
a0715cc2 AT |
549 | } |
550 | ||
551 | if (likely(!mm_alloc_pgd(mm))) { | |
cddb8a5c | 552 | mmu_notifier_mm_init(mm); |
1da177e4 LT |
553 | return mm; |
554 | } | |
78fb7466 | 555 | |
1da177e4 LT |
556 | free_mm(mm); |
557 | return NULL; | |
558 | } | |
559 | ||
c3f0327f KK |
560 | static void check_mm(struct mm_struct *mm) |
561 | { | |
562 | int i; | |
563 | ||
564 | for (i = 0; i < NR_MM_COUNTERS; i++) { | |
565 | long x = atomic_long_read(&mm->rss_stat.count[i]); | |
566 | ||
567 | if (unlikely(x)) | |
568 | printk(KERN_ALERT "BUG: Bad rss-counter state " | |
569 | "mm:%p idx:%d val:%ld\n", mm, i, x); | |
570 | } | |
571 | ||
e009bb30 | 572 | #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS |
c3f0327f KK |
573 | VM_BUG_ON(mm->pmd_huge_pte); |
574 | #endif | |
575 | } | |
576 | ||
1da177e4 LT |
577 | /* |
578 | * Allocate and initialize an mm_struct. | |
579 | */ | |
fb0a685c | 580 | struct mm_struct *mm_alloc(void) |
1da177e4 | 581 | { |
fb0a685c | 582 | struct mm_struct *mm; |
1da177e4 LT |
583 | |
584 | mm = allocate_mm(); | |
de03c72c KM |
585 | if (!mm) |
586 | return NULL; | |
587 | ||
588 | memset(mm, 0, sizeof(*mm)); | |
6345d24d LT |
589 | mm_init_cpumask(mm); |
590 | return mm_init(mm, current); | |
1da177e4 LT |
591 | } |
592 | ||
593 | /* | |
594 | * Called when the last reference to the mm | |
595 | * is dropped: either by a lazy thread or by | |
596 | * mmput. Free the page directory and the mm. | |
597 | */ | |
7ad5b3a5 | 598 | void __mmdrop(struct mm_struct *mm) |
1da177e4 LT |
599 | { |
600 | BUG_ON(mm == &init_mm); | |
601 | mm_free_pgd(mm); | |
602 | destroy_context(mm); | |
cddb8a5c | 603 | mmu_notifier_mm_destroy(mm); |
c3f0327f | 604 | check_mm(mm); |
1da177e4 LT |
605 | free_mm(mm); |
606 | } | |
6d4e4c4f | 607 | EXPORT_SYMBOL_GPL(__mmdrop); |
1da177e4 LT |
608 | |
609 | /* | |
610 | * Decrement the use count and release all resources for an mm. | |
611 | */ | |
612 | void mmput(struct mm_struct *mm) | |
613 | { | |
0ae26f1b AM |
614 | might_sleep(); |
615 | ||
1da177e4 | 616 | if (atomic_dec_and_test(&mm->mm_users)) { |
d4b3b638 | 617 | uprobe_clear_state(mm); |
1da177e4 | 618 | exit_aio(mm); |
1c2fb7a4 | 619 | ksm_exit(mm); |
ba76149f | 620 | khugepaged_exit(mm); /* must run before exit_mmap */ |
1da177e4 | 621 | exit_mmap(mm); |
925d1c40 | 622 | set_mm_exe_file(mm, NULL); |
1da177e4 LT |
623 | if (!list_empty(&mm->mmlist)) { |
624 | spin_lock(&mmlist_lock); | |
625 | list_del(&mm->mmlist); | |
626 | spin_unlock(&mmlist_lock); | |
627 | } | |
801460d0 HS |
628 | if (mm->binfmt) |
629 | module_put(mm->binfmt->module); | |
1da177e4 LT |
630 | mmdrop(mm); |
631 | } | |
632 | } | |
633 | EXPORT_SYMBOL_GPL(mmput); | |
634 | ||
38646013 JS |
635 | void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file) |
636 | { | |
637 | if (new_exe_file) | |
638 | get_file(new_exe_file); | |
639 | if (mm->exe_file) | |
640 | fput(mm->exe_file); | |
641 | mm->exe_file = new_exe_file; | |
38646013 JS |
642 | } |
643 | ||
644 | struct file *get_mm_exe_file(struct mm_struct *mm) | |
645 | { | |
646 | struct file *exe_file; | |
647 | ||
2dd8ad81 | 648 | /* We need mmap_sem to protect against races with removal of exe_file */ |
38646013 JS |
649 | down_read(&mm->mmap_sem); |
650 | exe_file = mm->exe_file; | |
651 | if (exe_file) | |
652 | get_file(exe_file); | |
653 | up_read(&mm->mmap_sem); | |
654 | return exe_file; | |
655 | } | |
656 | ||
657 | static void dup_mm_exe_file(struct mm_struct *oldmm, struct mm_struct *newmm) | |
658 | { | |
659 | /* It's safe to write the exe_file pointer without exe_file_lock because | |
660 | * this is called during fork when the task is not yet in /proc */ | |
661 | newmm->exe_file = get_mm_exe_file(oldmm); | |
662 | } | |
663 | ||
1da177e4 LT |
664 | /** |
665 | * get_task_mm - acquire a reference to the task's mm | |
666 | * | |
246bb0b1 | 667 | * Returns %NULL if the task has no mm. Checks PF_KTHREAD (meaning |
1da177e4 LT |
668 | * this kernel workthread has transiently adopted a user mm with use_mm, |
669 | * to do its AIO) is not set and if so returns a reference to it, after | |
670 | * bumping up the use count. User must release the mm via mmput() | |
671 | * after use. Typically used by /proc and ptrace. | |
672 | */ | |
673 | struct mm_struct *get_task_mm(struct task_struct *task) | |
674 | { | |
675 | struct mm_struct *mm; | |
676 | ||
677 | task_lock(task); | |
678 | mm = task->mm; | |
679 | if (mm) { | |
246bb0b1 | 680 | if (task->flags & PF_KTHREAD) |
1da177e4 LT |
681 | mm = NULL; |
682 | else | |
683 | atomic_inc(&mm->mm_users); | |
684 | } | |
685 | task_unlock(task); | |
686 | return mm; | |
687 | } | |
688 | EXPORT_SYMBOL_GPL(get_task_mm); | |
689 | ||
8cdb878d CY |
690 | struct mm_struct *mm_access(struct task_struct *task, unsigned int mode) |
691 | { | |
692 | struct mm_struct *mm; | |
693 | int err; | |
694 | ||
695 | err = mutex_lock_killable(&task->signal->cred_guard_mutex); | |
696 | if (err) | |
697 | return ERR_PTR(err); | |
698 | ||
699 | mm = get_task_mm(task); | |
700 | if (mm && mm != current->mm && | |
701 | !ptrace_may_access(task, mode)) { | |
702 | mmput(mm); | |
703 | mm = ERR_PTR(-EACCES); | |
704 | } | |
705 | mutex_unlock(&task->signal->cred_guard_mutex); | |
706 | ||
707 | return mm; | |
708 | } | |
709 | ||
57b59c4a | 710 | static void complete_vfork_done(struct task_struct *tsk) |
c415c3b4 | 711 | { |
d68b46fe | 712 | struct completion *vfork; |
c415c3b4 | 713 | |
d68b46fe ON |
714 | task_lock(tsk); |
715 | vfork = tsk->vfork_done; | |
716 | if (likely(vfork)) { | |
717 | tsk->vfork_done = NULL; | |
718 | complete(vfork); | |
719 | } | |
720 | task_unlock(tsk); | |
721 | } | |
722 | ||
723 | static int wait_for_vfork_done(struct task_struct *child, | |
724 | struct completion *vfork) | |
725 | { | |
726 | int killed; | |
727 | ||
728 | freezer_do_not_count(); | |
729 | killed = wait_for_completion_killable(vfork); | |
730 | freezer_count(); | |
731 | ||
732 | if (killed) { | |
733 | task_lock(child); | |
734 | child->vfork_done = NULL; | |
735 | task_unlock(child); | |
736 | } | |
737 | ||
738 | put_task_struct(child); | |
739 | return killed; | |
c415c3b4 ON |
740 | } |
741 | ||
1da177e4 LT |
742 | /* Please note the differences between mmput and mm_release. |
743 | * mmput is called whenever we stop holding onto a mm_struct, | |
744 | * error success whatever. | |
745 | * | |
746 | * mm_release is called after a mm_struct has been removed | |
747 | * from the current process. | |
748 | * | |
749 | * This difference is important for error handling, when we | |
750 | * only half set up a mm_struct for a new process and need to restore | |
751 | * the old one. Because we mmput the new mm_struct before | |
752 | * restoring the old one. . . | |
753 | * Eric Biederman 10 January 1998 | |
754 | */ | |
755 | void mm_release(struct task_struct *tsk, struct mm_struct *mm) | |
756 | { | |
8141c7f3 LT |
757 | /* Get rid of any futexes when releasing the mm */ |
758 | #ifdef CONFIG_FUTEX | |
fc6b177d | 759 | if (unlikely(tsk->robust_list)) { |
8141c7f3 | 760 | exit_robust_list(tsk); |
fc6b177d PZ |
761 | tsk->robust_list = NULL; |
762 | } | |
8141c7f3 | 763 | #ifdef CONFIG_COMPAT |
fc6b177d | 764 | if (unlikely(tsk->compat_robust_list)) { |
8141c7f3 | 765 | compat_exit_robust_list(tsk); |
fc6b177d PZ |
766 | tsk->compat_robust_list = NULL; |
767 | } | |
8141c7f3 | 768 | #endif |
322a2c10 TG |
769 | if (unlikely(!list_empty(&tsk->pi_state_list))) |
770 | exit_pi_state_list(tsk); | |
8141c7f3 LT |
771 | #endif |
772 | ||
0326f5a9 SD |
773 | uprobe_free_utask(tsk); |
774 | ||
1da177e4 LT |
775 | /* Get rid of any cached register state */ |
776 | deactivate_mm(tsk, mm); | |
777 | ||
fec1d011 RM |
778 | /* |
779 | * If we're exiting normally, clear a user-space tid field if | |
780 | * requested. We leave this alone when dying by signal, to leave | |
781 | * the value intact in a core dump, and to save the unnecessary | |
d68b46fe ON |
782 | * trouble, say, a killed vfork parent shouldn't touch this mm. |
783 | * Userland only wants this done for a sys_exit. | |
fec1d011 | 784 | */ |
9c8a8228 ED |
785 | if (tsk->clear_child_tid) { |
786 | if (!(tsk->flags & PF_SIGNALED) && | |
787 | atomic_read(&mm->mm_users) > 1) { | |
788 | /* | |
789 | * We don't check the error code - if userspace has | |
790 | * not set up a proper pointer then tough luck. | |
791 | */ | |
792 | put_user(0, tsk->clear_child_tid); | |
793 | sys_futex(tsk->clear_child_tid, FUTEX_WAKE, | |
794 | 1, NULL, NULL, 0); | |
795 | } | |
1da177e4 | 796 | tsk->clear_child_tid = NULL; |
1da177e4 | 797 | } |
f7505d64 KK |
798 | |
799 | /* | |
800 | * All done, finally we can wake up parent and return this mm to him. | |
801 | * Also kthread_stop() uses this completion for synchronization. | |
802 | */ | |
803 | if (tsk->vfork_done) | |
804 | complete_vfork_done(tsk); | |
1da177e4 LT |
805 | } |
806 | ||
a0a7ec30 JD |
807 | /* |
808 | * Allocate a new mm structure and copy contents from the | |
809 | * mm structure of the passed in task structure. | |
810 | */ | |
ff252c1f | 811 | static struct mm_struct *dup_mm(struct task_struct *tsk) |
a0a7ec30 JD |
812 | { |
813 | struct mm_struct *mm, *oldmm = current->mm; | |
814 | int err; | |
815 | ||
a0a7ec30 JD |
816 | mm = allocate_mm(); |
817 | if (!mm) | |
818 | goto fail_nomem; | |
819 | ||
820 | memcpy(mm, oldmm, sizeof(*mm)); | |
6345d24d | 821 | mm_init_cpumask(mm); |
a0a7ec30 | 822 | |
e009bb30 | 823 | #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS |
e7a00c45 AA |
824 | mm->pmd_huge_pte = NULL; |
825 | #endif | |
78fb7466 | 826 | if (!mm_init(mm, tsk)) |
a0a7ec30 JD |
827 | goto fail_nomem; |
828 | ||
829 | if (init_new_context(tsk, mm)) | |
830 | goto fail_nocontext; | |
831 | ||
925d1c40 MH |
832 | dup_mm_exe_file(oldmm, mm); |
833 | ||
a0a7ec30 JD |
834 | err = dup_mmap(mm, oldmm); |
835 | if (err) | |
836 | goto free_pt; | |
837 | ||
838 | mm->hiwater_rss = get_mm_rss(mm); | |
839 | mm->hiwater_vm = mm->total_vm; | |
840 | ||
801460d0 HS |
841 | if (mm->binfmt && !try_module_get(mm->binfmt->module)) |
842 | goto free_pt; | |
843 | ||
a0a7ec30 JD |
844 | return mm; |
845 | ||
846 | free_pt: | |
801460d0 HS |
847 | /* don't put binfmt in mmput, we haven't got module yet */ |
848 | mm->binfmt = NULL; | |
a0a7ec30 JD |
849 | mmput(mm); |
850 | ||
851 | fail_nomem: | |
852 | return NULL; | |
853 | ||
854 | fail_nocontext: | |
855 | /* | |
856 | * If init_new_context() failed, we cannot use mmput() to free the mm | |
857 | * because it calls destroy_context() | |
858 | */ | |
859 | mm_free_pgd(mm); | |
860 | free_mm(mm); | |
861 | return NULL; | |
862 | } | |
863 | ||
fb0a685c | 864 | static int copy_mm(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 | 865 | { |
fb0a685c | 866 | struct mm_struct *mm, *oldmm; |
1da177e4 LT |
867 | int retval; |
868 | ||
869 | tsk->min_flt = tsk->maj_flt = 0; | |
870 | tsk->nvcsw = tsk->nivcsw = 0; | |
17406b82 MSB |
871 | #ifdef CONFIG_DETECT_HUNG_TASK |
872 | tsk->last_switch_count = tsk->nvcsw + tsk->nivcsw; | |
873 | #endif | |
1da177e4 LT |
874 | |
875 | tsk->mm = NULL; | |
876 | tsk->active_mm = NULL; | |
877 | ||
878 | /* | |
879 | * Are we cloning a kernel thread? | |
880 | * | |
881 | * We need to steal a active VM for that.. | |
882 | */ | |
883 | oldmm = current->mm; | |
884 | if (!oldmm) | |
885 | return 0; | |
886 | ||
615d6e87 DB |
887 | /* initialize the new vmacache entries */ |
888 | vmacache_flush(tsk); | |
889 | ||
1da177e4 LT |
890 | if (clone_flags & CLONE_VM) { |
891 | atomic_inc(&oldmm->mm_users); | |
892 | mm = oldmm; | |
1da177e4 LT |
893 | goto good_mm; |
894 | } | |
895 | ||
896 | retval = -ENOMEM; | |
a0a7ec30 | 897 | mm = dup_mm(tsk); |
1da177e4 LT |
898 | if (!mm) |
899 | goto fail_nomem; | |
900 | ||
1da177e4 LT |
901 | good_mm: |
902 | tsk->mm = mm; | |
903 | tsk->active_mm = mm; | |
904 | return 0; | |
905 | ||
1da177e4 LT |
906 | fail_nomem: |
907 | return retval; | |
1da177e4 LT |
908 | } |
909 | ||
a39bc516 | 910 | static int copy_fs(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 | 911 | { |
498052bb | 912 | struct fs_struct *fs = current->fs; |
1da177e4 | 913 | if (clone_flags & CLONE_FS) { |
498052bb | 914 | /* tsk->fs is already what we want */ |
2a4419b5 | 915 | spin_lock(&fs->lock); |
498052bb | 916 | if (fs->in_exec) { |
2a4419b5 | 917 | spin_unlock(&fs->lock); |
498052bb AV |
918 | return -EAGAIN; |
919 | } | |
920 | fs->users++; | |
2a4419b5 | 921 | spin_unlock(&fs->lock); |
1da177e4 LT |
922 | return 0; |
923 | } | |
498052bb | 924 | tsk->fs = copy_fs_struct(fs); |
1da177e4 LT |
925 | if (!tsk->fs) |
926 | return -ENOMEM; | |
927 | return 0; | |
928 | } | |
929 | ||
fb0a685c | 930 | static int copy_files(unsigned long clone_flags, struct task_struct *tsk) |
a016f338 JD |
931 | { |
932 | struct files_struct *oldf, *newf; | |
933 | int error = 0; | |
934 | ||
935 | /* | |
936 | * A background process may not have any files ... | |
937 | */ | |
938 | oldf = current->files; | |
939 | if (!oldf) | |
940 | goto out; | |
941 | ||
942 | if (clone_flags & CLONE_FILES) { | |
943 | atomic_inc(&oldf->count); | |
944 | goto out; | |
945 | } | |
946 | ||
a016f338 JD |
947 | newf = dup_fd(oldf, &error); |
948 | if (!newf) | |
949 | goto out; | |
950 | ||
951 | tsk->files = newf; | |
952 | error = 0; | |
953 | out: | |
954 | return error; | |
955 | } | |
956 | ||
fadad878 | 957 | static int copy_io(unsigned long clone_flags, struct task_struct *tsk) |
fd0928df JA |
958 | { |
959 | #ifdef CONFIG_BLOCK | |
960 | struct io_context *ioc = current->io_context; | |
6e736be7 | 961 | struct io_context *new_ioc; |
fd0928df JA |
962 | |
963 | if (!ioc) | |
964 | return 0; | |
fadad878 JA |
965 | /* |
966 | * Share io context with parent, if CLONE_IO is set | |
967 | */ | |
968 | if (clone_flags & CLONE_IO) { | |
3d48749d TH |
969 | ioc_task_link(ioc); |
970 | tsk->io_context = ioc; | |
fadad878 | 971 | } else if (ioprio_valid(ioc->ioprio)) { |
6e736be7 TH |
972 | new_ioc = get_task_io_context(tsk, GFP_KERNEL, NUMA_NO_NODE); |
973 | if (unlikely(!new_ioc)) | |
fd0928df JA |
974 | return -ENOMEM; |
975 | ||
6e736be7 | 976 | new_ioc->ioprio = ioc->ioprio; |
11a3122f | 977 | put_io_context(new_ioc); |
fd0928df JA |
978 | } |
979 | #endif | |
980 | return 0; | |
981 | } | |
982 | ||
a39bc516 | 983 | static int copy_sighand(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 LT |
984 | { |
985 | struct sighand_struct *sig; | |
986 | ||
60348802 | 987 | if (clone_flags & CLONE_SIGHAND) { |
1da177e4 LT |
988 | atomic_inc(¤t->sighand->count); |
989 | return 0; | |
990 | } | |
991 | sig = kmem_cache_alloc(sighand_cachep, GFP_KERNEL); | |
e56d0903 | 992 | rcu_assign_pointer(tsk->sighand, sig); |
1da177e4 LT |
993 | if (!sig) |
994 | return -ENOMEM; | |
1da177e4 LT |
995 | atomic_set(&sig->count, 1); |
996 | memcpy(sig->action, current->sighand->action, sizeof(sig->action)); | |
997 | return 0; | |
998 | } | |
999 | ||
a7e5328a | 1000 | void __cleanup_sighand(struct sighand_struct *sighand) |
c81addc9 | 1001 | { |
d80e731e ON |
1002 | if (atomic_dec_and_test(&sighand->count)) { |
1003 | signalfd_cleanup(sighand); | |
c81addc9 | 1004 | kmem_cache_free(sighand_cachep, sighand); |
d80e731e | 1005 | } |
c81addc9 ON |
1006 | } |
1007 | ||
f06febc9 FM |
1008 | |
1009 | /* | |
1010 | * Initialize POSIX timer handling for a thread group. | |
1011 | */ | |
1012 | static void posix_cpu_timers_init_group(struct signal_struct *sig) | |
1013 | { | |
78d7d407 JS |
1014 | unsigned long cpu_limit; |
1015 | ||
f06febc9 FM |
1016 | /* Thread group counters. */ |
1017 | thread_group_cputime_init(sig); | |
1018 | ||
78d7d407 JS |
1019 | cpu_limit = ACCESS_ONCE(sig->rlim[RLIMIT_CPU].rlim_cur); |
1020 | if (cpu_limit != RLIM_INFINITY) { | |
1021 | sig->cputime_expires.prof_exp = secs_to_cputime(cpu_limit); | |
6279a751 ON |
1022 | sig->cputimer.running = 1; |
1023 | } | |
1024 | ||
f06febc9 FM |
1025 | /* The timer lists. */ |
1026 | INIT_LIST_HEAD(&sig->cpu_timers[0]); | |
1027 | INIT_LIST_HEAD(&sig->cpu_timers[1]); | |
1028 | INIT_LIST_HEAD(&sig->cpu_timers[2]); | |
1029 | } | |
1030 | ||
a39bc516 | 1031 | static int copy_signal(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 LT |
1032 | { |
1033 | struct signal_struct *sig; | |
1da177e4 | 1034 | |
4ab6c083 | 1035 | if (clone_flags & CLONE_THREAD) |
490dea45 | 1036 | return 0; |
490dea45 | 1037 | |
a56704ef | 1038 | sig = kmem_cache_zalloc(signal_cachep, GFP_KERNEL); |
1da177e4 LT |
1039 | tsk->signal = sig; |
1040 | if (!sig) | |
1041 | return -ENOMEM; | |
1042 | ||
b3ac022c | 1043 | sig->nr_threads = 1; |
1da177e4 | 1044 | atomic_set(&sig->live, 1); |
b3ac022c | 1045 | atomic_set(&sig->sigcnt, 1); |
0c740d0a ON |
1046 | |
1047 | /* list_add(thread_node, thread_head) without INIT_LIST_HEAD() */ | |
1048 | sig->thread_head = (struct list_head)LIST_HEAD_INIT(tsk->thread_node); | |
1049 | tsk->thread_node = (struct list_head)LIST_HEAD_INIT(sig->thread_head); | |
1050 | ||
1da177e4 | 1051 | init_waitqueue_head(&sig->wait_chldexit); |
db51aecc | 1052 | sig->curr_target = tsk; |
1da177e4 LT |
1053 | init_sigpending(&sig->shared_pending); |
1054 | INIT_LIST_HEAD(&sig->posix_timers); | |
1055 | ||
c9cb2e3d | 1056 | hrtimer_init(&sig->real_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL); |
1da177e4 | 1057 | sig->real_timer.function = it_real_fn; |
1da177e4 | 1058 | |
1da177e4 LT |
1059 | task_lock(current->group_leader); |
1060 | memcpy(sig->rlim, current->signal->rlim, sizeof sig->rlim); | |
1061 | task_unlock(current->group_leader); | |
1062 | ||
6279a751 ON |
1063 | posix_cpu_timers_init_group(sig); |
1064 | ||
522ed776 | 1065 | tty_audit_fork(sig); |
5091faa4 | 1066 | sched_autogroup_fork(sig); |
522ed776 | 1067 | |
4714d1d3 | 1068 | #ifdef CONFIG_CGROUPS |
257058ae | 1069 | init_rwsem(&sig->group_rwsem); |
4714d1d3 BB |
1070 | #endif |
1071 | ||
a63d83f4 | 1072 | sig->oom_score_adj = current->signal->oom_score_adj; |
dabb16f6 | 1073 | sig->oom_score_adj_min = current->signal->oom_score_adj_min; |
28b83c51 | 1074 | |
ebec18a6 LP |
1075 | sig->has_child_subreaper = current->signal->has_child_subreaper || |
1076 | current->signal->is_child_subreaper; | |
1077 | ||
9b1bf12d KM |
1078 | mutex_init(&sig->cred_guard_mutex); |
1079 | ||
1da177e4 LT |
1080 | return 0; |
1081 | } | |
1082 | ||
a39bc516 | 1083 | static void copy_flags(unsigned long clone_flags, struct task_struct *p) |
1da177e4 LT |
1084 | { |
1085 | unsigned long new_flags = p->flags; | |
1086 | ||
21aa9af0 | 1087 | new_flags &= ~(PF_SUPERPRIV | PF_WQ_WORKER); |
1da177e4 | 1088 | new_flags |= PF_FORKNOEXEC; |
1da177e4 LT |
1089 | p->flags = new_flags; |
1090 | } | |
1091 | ||
17da2bd9 | 1092 | SYSCALL_DEFINE1(set_tid_address, int __user *, tidptr) |
1da177e4 LT |
1093 | { |
1094 | current->clear_child_tid = tidptr; | |
1095 | ||
b488893a | 1096 | return task_pid_vnr(current); |
1da177e4 LT |
1097 | } |
1098 | ||
a39bc516 | 1099 | static void rt_mutex_init_task(struct task_struct *p) |
23f78d4a | 1100 | { |
1d615482 | 1101 | raw_spin_lock_init(&p->pi_lock); |
e29e175b | 1102 | #ifdef CONFIG_RT_MUTEXES |
fb00aca4 PZ |
1103 | p->pi_waiters = RB_ROOT; |
1104 | p->pi_waiters_leftmost = NULL; | |
23f78d4a | 1105 | p->pi_blocked_on = NULL; |
2d3d891d | 1106 | p->pi_top_task = NULL; |
23f78d4a IM |
1107 | #endif |
1108 | } | |
1109 | ||
cf475ad2 BS |
1110 | #ifdef CONFIG_MM_OWNER |
1111 | void mm_init_owner(struct mm_struct *mm, struct task_struct *p) | |
1112 | { | |
1113 | mm->owner = p; | |
1114 | } | |
1115 | #endif /* CONFIG_MM_OWNER */ | |
1116 | ||
f06febc9 FM |
1117 | /* |
1118 | * Initialize POSIX timer handling for a single task. | |
1119 | */ | |
1120 | static void posix_cpu_timers_init(struct task_struct *tsk) | |
1121 | { | |
64861634 MS |
1122 | tsk->cputime_expires.prof_exp = 0; |
1123 | tsk->cputime_expires.virt_exp = 0; | |
f06febc9 FM |
1124 | tsk->cputime_expires.sched_exp = 0; |
1125 | INIT_LIST_HEAD(&tsk->cpu_timers[0]); | |
1126 | INIT_LIST_HEAD(&tsk->cpu_timers[1]); | |
1127 | INIT_LIST_HEAD(&tsk->cpu_timers[2]); | |
1128 | } | |
1129 | ||
81907739 ON |
1130 | static inline void |
1131 | init_task_pid(struct task_struct *task, enum pid_type type, struct pid *pid) | |
1132 | { | |
1133 | task->pids[type].pid = pid; | |
1134 | } | |
1135 | ||
1da177e4 LT |
1136 | /* |
1137 | * This creates a new process as a copy of the old one, | |
1138 | * but does not actually start it yet. | |
1139 | * | |
1140 | * It copies the registers, and all the appropriate | |
1141 | * parts of the process environment (as per the clone | |
1142 | * flags). The actual kick-off is left to the caller. | |
1143 | */ | |
36c8b586 IM |
1144 | static struct task_struct *copy_process(unsigned long clone_flags, |
1145 | unsigned long stack_start, | |
36c8b586 | 1146 | unsigned long stack_size, |
36c8b586 | 1147 | int __user *child_tidptr, |
09a05394 RM |
1148 | struct pid *pid, |
1149 | int trace) | |
1da177e4 LT |
1150 | { |
1151 | int retval; | |
a24efe62 | 1152 | struct task_struct *p; |
1da177e4 LT |
1153 | |
1154 | if ((clone_flags & (CLONE_NEWNS|CLONE_FS)) == (CLONE_NEWNS|CLONE_FS)) | |
1155 | return ERR_PTR(-EINVAL); | |
1156 | ||
e66eded8 EB |
1157 | if ((clone_flags & (CLONE_NEWUSER|CLONE_FS)) == (CLONE_NEWUSER|CLONE_FS)) |
1158 | return ERR_PTR(-EINVAL); | |
1159 | ||
1da177e4 LT |
1160 | /* |
1161 | * Thread groups must share signals as well, and detached threads | |
1162 | * can only be started up within the thread group. | |
1163 | */ | |
1164 | if ((clone_flags & CLONE_THREAD) && !(clone_flags & CLONE_SIGHAND)) | |
1165 | return ERR_PTR(-EINVAL); | |
1166 | ||
1167 | /* | |
1168 | * Shared signal handlers imply shared VM. By way of the above, | |
1169 | * thread groups also imply shared VM. Blocking this case allows | |
1170 | * for various simplifications in other code. | |
1171 | */ | |
1172 | if ((clone_flags & CLONE_SIGHAND) && !(clone_flags & CLONE_VM)) | |
1173 | return ERR_PTR(-EINVAL); | |
1174 | ||
123be07b SB |
1175 | /* |
1176 | * Siblings of global init remain as zombies on exit since they are | |
1177 | * not reaped by their parent (swapper). To solve this and to avoid | |
1178 | * multi-rooted process trees, prevent global and container-inits | |
1179 | * from creating siblings. | |
1180 | */ | |
1181 | if ((clone_flags & CLONE_PARENT) && | |
1182 | current->signal->flags & SIGNAL_UNKILLABLE) | |
1183 | return ERR_PTR(-EINVAL); | |
1184 | ||
8382fcac | 1185 | /* |
40a0d32d ON |
1186 | * If the new process will be in a different pid or user namespace |
1187 | * do not allow it to share a thread group or signal handlers or | |
1188 | * parent with the forking task. | |
8382fcac | 1189 | */ |
1f7f4dde | 1190 | if (clone_flags & CLONE_SIGHAND) { |
40a0d32d ON |
1191 | if ((clone_flags & (CLONE_NEWUSER | CLONE_NEWPID)) || |
1192 | (task_active_pid_ns(current) != | |
1193 | current->nsproxy->pid_ns_for_children)) | |
1194 | return ERR_PTR(-EINVAL); | |
1195 | } | |
8382fcac | 1196 | |
1da177e4 LT |
1197 | retval = security_task_create(clone_flags); |
1198 | if (retval) | |
1199 | goto fork_out; | |
1200 | ||
1201 | retval = -ENOMEM; | |
1202 | p = dup_task_struct(current); | |
1203 | if (!p) | |
1204 | goto fork_out; | |
1205 | ||
f7e8b616 | 1206 | ftrace_graph_init_task(p); |
e2cfabdf | 1207 | get_seccomp_filter(p); |
f7e8b616 | 1208 | |
bea493a0 PZ |
1209 | rt_mutex_init_task(p); |
1210 | ||
d12c1a37 | 1211 | #ifdef CONFIG_PROVE_LOCKING |
de30a2b3 IM |
1212 | DEBUG_LOCKS_WARN_ON(!p->hardirqs_enabled); |
1213 | DEBUG_LOCKS_WARN_ON(!p->softirqs_enabled); | |
1214 | #endif | |
1da177e4 | 1215 | retval = -EAGAIN; |
3b11a1de | 1216 | if (atomic_read(&p->real_cred->user->processes) >= |
78d7d407 | 1217 | task_rlimit(p, RLIMIT_NPROC)) { |
b57922b6 EP |
1218 | if (p->real_cred->user != INIT_USER && |
1219 | !capable(CAP_SYS_RESOURCE) && !capable(CAP_SYS_ADMIN)) | |
1da177e4 LT |
1220 | goto bad_fork_free; |
1221 | } | |
72fa5997 | 1222 | current->flags &= ~PF_NPROC_EXCEEDED; |
1da177e4 | 1223 | |
f1752eec DH |
1224 | retval = copy_creds(p, clone_flags); |
1225 | if (retval < 0) | |
1226 | goto bad_fork_free; | |
1da177e4 LT |
1227 | |
1228 | /* | |
1229 | * If multiple threads are within copy_process(), then this check | |
1230 | * triggers too late. This doesn't hurt, the check is only there | |
1231 | * to stop root fork bombs. | |
1232 | */ | |
04ec93fe | 1233 | retval = -EAGAIN; |
1da177e4 LT |
1234 | if (nr_threads >= max_threads) |
1235 | goto bad_fork_cleanup_count; | |
1236 | ||
a1261f54 | 1237 | if (!try_module_get(task_thread_info(p)->exec_domain->module)) |
1da177e4 LT |
1238 | goto bad_fork_cleanup_count; |
1239 | ||
ca74e92b | 1240 | delayacct_tsk_init(p); /* Must remain after dup_task_struct() */ |
1da177e4 | 1241 | copy_flags(clone_flags, p); |
1da177e4 LT |
1242 | INIT_LIST_HEAD(&p->children); |
1243 | INIT_LIST_HEAD(&p->sibling); | |
f41d911f | 1244 | rcu_copy_process(p); |
1da177e4 LT |
1245 | p->vfork_done = NULL; |
1246 | spin_lock_init(&p->alloc_lock); | |
1da177e4 | 1247 | |
1da177e4 LT |
1248 | init_sigpending(&p->pending); |
1249 | ||
64861634 MS |
1250 | p->utime = p->stime = p->gtime = 0; |
1251 | p->utimescaled = p->stimescaled = 0; | |
9fbc42ea | 1252 | #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE |
d37f761d | 1253 | p->prev_cputime.utime = p->prev_cputime.stime = 0; |
d99ca3b9 | 1254 | #endif |
6a61671b FW |
1255 | #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN |
1256 | seqlock_init(&p->vtime_seqlock); | |
1257 | p->vtime_snap = 0; | |
1258 | p->vtime_snap_whence = VTIME_SLEEPING; | |
1259 | #endif | |
1260 | ||
a3a2e76c KH |
1261 | #if defined(SPLIT_RSS_COUNTING) |
1262 | memset(&p->rss_stat, 0, sizeof(p->rss_stat)); | |
1263 | #endif | |
172ba844 | 1264 | |
6976675d AV |
1265 | p->default_timer_slack_ns = current->timer_slack_ns; |
1266 | ||
5995477a | 1267 | task_io_accounting_init(&p->ioac); |
1da177e4 LT |
1268 | acct_clear_integrals(p); |
1269 | ||
f06febc9 | 1270 | posix_cpu_timers_init(p); |
1da177e4 | 1271 | |
1da177e4 | 1272 | do_posix_clock_monotonic_gettime(&p->start_time); |
924b42d5 TJ |
1273 | p->real_start_time = p->start_time; |
1274 | monotonic_to_bootbased(&p->real_start_time); | |
1da177e4 | 1275 | p->io_context = NULL; |
1da177e4 | 1276 | p->audit_context = NULL; |
4714d1d3 | 1277 | if (clone_flags & CLONE_THREAD) |
257058ae | 1278 | threadgroup_change_begin(current); |
b4f48b63 | 1279 | cgroup_fork(p); |
1da177e4 | 1280 | #ifdef CONFIG_NUMA |
846a16bf | 1281 | p->mempolicy = mpol_dup(p->mempolicy); |
fb0a685c DRO |
1282 | if (IS_ERR(p->mempolicy)) { |
1283 | retval = PTR_ERR(p->mempolicy); | |
1284 | p->mempolicy = NULL; | |
e8604cb4 | 1285 | goto bad_fork_cleanup_threadgroup_lock; |
fb0a685c | 1286 | } |
c61afb18 | 1287 | mpol_fix_fork_child_flag(p); |
1da177e4 | 1288 | #endif |
778d3b0f MH |
1289 | #ifdef CONFIG_CPUSETS |
1290 | p->cpuset_mem_spread_rotor = NUMA_NO_NODE; | |
1291 | p->cpuset_slab_spread_rotor = NUMA_NO_NODE; | |
cc9a6c87 | 1292 | seqcount_init(&p->mems_allowed_seq); |
778d3b0f | 1293 | #endif |
de30a2b3 IM |
1294 | #ifdef CONFIG_TRACE_IRQFLAGS |
1295 | p->irq_events = 0; | |
1296 | p->hardirqs_enabled = 0; | |
1297 | p->hardirq_enable_ip = 0; | |
1298 | p->hardirq_enable_event = 0; | |
1299 | p->hardirq_disable_ip = _THIS_IP_; | |
1300 | p->hardirq_disable_event = 0; | |
1301 | p->softirqs_enabled = 1; | |
1302 | p->softirq_enable_ip = _THIS_IP_; | |
1303 | p->softirq_enable_event = 0; | |
1304 | p->softirq_disable_ip = 0; | |
1305 | p->softirq_disable_event = 0; | |
1306 | p->hardirq_context = 0; | |
1307 | p->softirq_context = 0; | |
1308 | #endif | |
fbb9ce95 IM |
1309 | #ifdef CONFIG_LOCKDEP |
1310 | p->lockdep_depth = 0; /* no locks held yet */ | |
1311 | p->curr_chain_key = 0; | |
1312 | p->lockdep_recursion = 0; | |
1313 | #endif | |
1da177e4 | 1314 | |
408894ee IM |
1315 | #ifdef CONFIG_DEBUG_MUTEXES |
1316 | p->blocked_on = NULL; /* not blocked yet */ | |
1317 | #endif | |
c255a458 | 1318 | #ifdef CONFIG_MEMCG |
569b846d KH |
1319 | p->memcg_batch.do_batch = 0; |
1320 | p->memcg_batch.memcg = NULL; | |
1321 | #endif | |
cafe5635 KO |
1322 | #ifdef CONFIG_BCACHE |
1323 | p->sequential_io = 0; | |
1324 | p->sequential_io_avg = 0; | |
1325 | #endif | |
0f481406 | 1326 | |
3c90e6e9 | 1327 | /* Perform scheduler related setup. Assign this task to a CPU. */ |
aab03e05 DF |
1328 | retval = sched_fork(clone_flags, p); |
1329 | if (retval) | |
1330 | goto bad_fork_cleanup_policy; | |
6ab423e0 | 1331 | |
cdd6c482 | 1332 | retval = perf_event_init_task(p); |
6ab423e0 PZ |
1333 | if (retval) |
1334 | goto bad_fork_cleanup_policy; | |
fb0a685c DRO |
1335 | retval = audit_alloc(p); |
1336 | if (retval) | |
f1752eec | 1337 | goto bad_fork_cleanup_policy; |
1da177e4 | 1338 | /* copy all the process information */ |
fb0a685c DRO |
1339 | retval = copy_semundo(clone_flags, p); |
1340 | if (retval) | |
1da177e4 | 1341 | goto bad_fork_cleanup_audit; |
fb0a685c DRO |
1342 | retval = copy_files(clone_flags, p); |
1343 | if (retval) | |
1da177e4 | 1344 | goto bad_fork_cleanup_semundo; |
fb0a685c DRO |
1345 | retval = copy_fs(clone_flags, p); |
1346 | if (retval) | |
1da177e4 | 1347 | goto bad_fork_cleanup_files; |
fb0a685c DRO |
1348 | retval = copy_sighand(clone_flags, p); |
1349 | if (retval) | |
1da177e4 | 1350 | goto bad_fork_cleanup_fs; |
fb0a685c DRO |
1351 | retval = copy_signal(clone_flags, p); |
1352 | if (retval) | |
1da177e4 | 1353 | goto bad_fork_cleanup_sighand; |
fb0a685c DRO |
1354 | retval = copy_mm(clone_flags, p); |
1355 | if (retval) | |
1da177e4 | 1356 | goto bad_fork_cleanup_signal; |
fb0a685c DRO |
1357 | retval = copy_namespaces(clone_flags, p); |
1358 | if (retval) | |
d84f4f99 | 1359 | goto bad_fork_cleanup_mm; |
fb0a685c DRO |
1360 | retval = copy_io(clone_flags, p); |
1361 | if (retval) | |
fd0928df | 1362 | goto bad_fork_cleanup_namespaces; |
afa86fc4 | 1363 | retval = copy_thread(clone_flags, stack_start, stack_size, p); |
1da177e4 | 1364 | if (retval) |
fd0928df | 1365 | goto bad_fork_cleanup_io; |
1da177e4 | 1366 | |
425fb2b4 PE |
1367 | if (pid != &init_struct_pid) { |
1368 | retval = -ENOMEM; | |
c2b1df2e | 1369 | pid = alloc_pid(p->nsproxy->pid_ns_for_children); |
425fb2b4 | 1370 | if (!pid) |
fd0928df | 1371 | goto bad_fork_cleanup_io; |
425fb2b4 PE |
1372 | } |
1373 | ||
1da177e4 LT |
1374 | p->set_child_tid = (clone_flags & CLONE_CHILD_SETTID) ? child_tidptr : NULL; |
1375 | /* | |
1376 | * Clear TID on mm_release()? | |
1377 | */ | |
fb0a685c | 1378 | p->clear_child_tid = (clone_flags & CLONE_CHILD_CLEARTID) ? child_tidptr : NULL; |
73c10101 JA |
1379 | #ifdef CONFIG_BLOCK |
1380 | p->plug = NULL; | |
1381 | #endif | |
42b2dd0a | 1382 | #ifdef CONFIG_FUTEX |
8f17d3a5 IM |
1383 | p->robust_list = NULL; |
1384 | #ifdef CONFIG_COMPAT | |
1385 | p->compat_robust_list = NULL; | |
1386 | #endif | |
c87e2837 IM |
1387 | INIT_LIST_HEAD(&p->pi_state_list); |
1388 | p->pi_state_cache = NULL; | |
42b2dd0a | 1389 | #endif |
f9a3879a GM |
1390 | /* |
1391 | * sigaltstack should be cleared when sharing the same VM | |
1392 | */ | |
1393 | if ((clone_flags & (CLONE_VM|CLONE_VFORK)) == CLONE_VM) | |
1394 | p->sas_ss_sp = p->sas_ss_size = 0; | |
1395 | ||
1da177e4 | 1396 | /* |
6580807d ON |
1397 | * Syscall tracing and stepping should be turned off in the |
1398 | * child regardless of CLONE_PTRACE. | |
1da177e4 | 1399 | */ |
6580807d | 1400 | user_disable_single_step(p); |
1da177e4 | 1401 | clear_tsk_thread_flag(p, TIF_SYSCALL_TRACE); |
ed75e8d5 LV |
1402 | #ifdef TIF_SYSCALL_EMU |
1403 | clear_tsk_thread_flag(p, TIF_SYSCALL_EMU); | |
1404 | #endif | |
9745512c | 1405 | clear_all_latency_tracing(p); |
1da177e4 | 1406 | |
1da177e4 | 1407 | /* ok, now we should be set up.. */ |
18c830df ON |
1408 | p->pid = pid_nr(pid); |
1409 | if (clone_flags & CLONE_THREAD) { | |
5f8aadd8 | 1410 | p->exit_signal = -1; |
18c830df ON |
1411 | p->group_leader = current->group_leader; |
1412 | p->tgid = current->tgid; | |
1413 | } else { | |
1414 | if (clone_flags & CLONE_PARENT) | |
1415 | p->exit_signal = current->group_leader->exit_signal; | |
1416 | else | |
1417 | p->exit_signal = (clone_flags & CSIGNAL); | |
1418 | p->group_leader = p; | |
1419 | p->tgid = p->pid; | |
1420 | } | |
5f8aadd8 | 1421 | |
9d823e8f WF |
1422 | p->nr_dirtied = 0; |
1423 | p->nr_dirtied_pause = 128 >> (PAGE_SHIFT - 10); | |
83712358 | 1424 | p->dirty_paused_when = 0; |
9d823e8f | 1425 | |
bb8cbbfe | 1426 | p->pdeath_signal = 0; |
47e65328 | 1427 | INIT_LIST_HEAD(&p->thread_group); |
158e1645 | 1428 | p->task_works = NULL; |
1da177e4 | 1429 | |
18c830df ON |
1430 | /* |
1431 | * Make it visible to the rest of the system, but dont wake it up yet. | |
1432 | * Need tasklist lock for parent etc handling! | |
1433 | */ | |
1da177e4 LT |
1434 | write_lock_irq(&tasklist_lock); |
1435 | ||
1da177e4 | 1436 | /* CLONE_PARENT re-uses the old parent */ |
2d5516cb | 1437 | if (clone_flags & (CLONE_PARENT|CLONE_THREAD)) { |
1da177e4 | 1438 | p->real_parent = current->real_parent; |
2d5516cb ON |
1439 | p->parent_exec_id = current->parent_exec_id; |
1440 | } else { | |
1da177e4 | 1441 | p->real_parent = current; |
2d5516cb ON |
1442 | p->parent_exec_id = current->self_exec_id; |
1443 | } | |
1da177e4 | 1444 | |
3f17da69 | 1445 | spin_lock(¤t->sighand->siglock); |
4a2c7a78 ON |
1446 | |
1447 | /* | |
1448 | * Process group and session signals need to be delivered to just the | |
1449 | * parent before the fork or both the parent and the child after the | |
1450 | * fork. Restart if a signal comes in before we add the new process to | |
1451 | * it's process group. | |
1452 | * A fatal signal pending means that current will exit, so the new | |
1453 | * thread can't slip out of an OOM kill (or normal SIGKILL). | |
fb0a685c | 1454 | */ |
23ff4440 | 1455 | recalc_sigpending(); |
4a2c7a78 ON |
1456 | if (signal_pending(current)) { |
1457 | spin_unlock(¤t->sighand->siglock); | |
1458 | write_unlock_irq(&tasklist_lock); | |
1459 | retval = -ERESTARTNOINTR; | |
f7e8b616 | 1460 | goto bad_fork_free_pid; |
4a2c7a78 ON |
1461 | } |
1462 | ||
73b9ebfe | 1463 | if (likely(p->pid)) { |
4b9d33e6 | 1464 | ptrace_init_task(p, (clone_flags & CLONE_PTRACE) || trace); |
73b9ebfe | 1465 | |
81907739 | 1466 | init_task_pid(p, PIDTYPE_PID, pid); |
73b9ebfe | 1467 | if (thread_group_leader(p)) { |
81907739 ON |
1468 | init_task_pid(p, PIDTYPE_PGID, task_pgrp(current)); |
1469 | init_task_pid(p, PIDTYPE_SID, task_session(current)); | |
1470 | ||
1c4042c2 | 1471 | if (is_child_reaper(pid)) { |
17cf22c3 | 1472 | ns_of_pid(pid)->child_reaper = p; |
1c4042c2 EB |
1473 | p->signal->flags |= SIGNAL_UNKILLABLE; |
1474 | } | |
73b9ebfe | 1475 | |
fea9d175 | 1476 | p->signal->leader_pid = pid; |
9c9f4ded | 1477 | p->signal->tty = tty_kref_get(current->signal->tty); |
9cd80bbb | 1478 | list_add_tail(&p->sibling, &p->real_parent->children); |
5e85d4ab | 1479 | list_add_tail_rcu(&p->tasks, &init_task.tasks); |
81907739 ON |
1480 | attach_pid(p, PIDTYPE_PGID); |
1481 | attach_pid(p, PIDTYPE_SID); | |
909ea964 | 1482 | __this_cpu_inc(process_counts); |
80628ca0 ON |
1483 | } else { |
1484 | current->signal->nr_threads++; | |
1485 | atomic_inc(¤t->signal->live); | |
1486 | atomic_inc(¤t->signal->sigcnt); | |
80628ca0 ON |
1487 | list_add_tail_rcu(&p->thread_group, |
1488 | &p->group_leader->thread_group); | |
0c740d0a ON |
1489 | list_add_tail_rcu(&p->thread_node, |
1490 | &p->signal->thread_head); | |
73b9ebfe | 1491 | } |
81907739 | 1492 | attach_pid(p, PIDTYPE_PID); |
73b9ebfe | 1493 | nr_threads++; |
1da177e4 LT |
1494 | } |
1495 | ||
1da177e4 | 1496 | total_forks++; |
3f17da69 | 1497 | spin_unlock(¤t->sighand->siglock); |
1da177e4 | 1498 | write_unlock_irq(&tasklist_lock); |
c13cf856 | 1499 | proc_fork_connector(p); |
817929ec | 1500 | cgroup_post_fork(p); |
4714d1d3 | 1501 | if (clone_flags & CLONE_THREAD) |
257058ae | 1502 | threadgroup_change_end(current); |
cdd6c482 | 1503 | perf_event_fork(p); |
43d2b113 KH |
1504 | |
1505 | trace_task_newtask(p, clone_flags); | |
3ab67966 | 1506 | uprobe_copy_process(p, clone_flags); |
43d2b113 | 1507 | |
1da177e4 LT |
1508 | return p; |
1509 | ||
425fb2b4 PE |
1510 | bad_fork_free_pid: |
1511 | if (pid != &init_struct_pid) | |
1512 | free_pid(pid); | |
fd0928df | 1513 | bad_fork_cleanup_io: |
b69f2292 LR |
1514 | if (p->io_context) |
1515 | exit_io_context(p); | |
ab516013 | 1516 | bad_fork_cleanup_namespaces: |
444f378b | 1517 | exit_task_namespaces(p); |
1da177e4 | 1518 | bad_fork_cleanup_mm: |
c9f01245 | 1519 | if (p->mm) |
1da177e4 LT |
1520 | mmput(p->mm); |
1521 | bad_fork_cleanup_signal: | |
4ab6c083 | 1522 | if (!(clone_flags & CLONE_THREAD)) |
1c5354de | 1523 | free_signal_struct(p->signal); |
1da177e4 | 1524 | bad_fork_cleanup_sighand: |
a7e5328a | 1525 | __cleanup_sighand(p->sighand); |
1da177e4 LT |
1526 | bad_fork_cleanup_fs: |
1527 | exit_fs(p); /* blocking */ | |
1528 | bad_fork_cleanup_files: | |
1529 | exit_files(p); /* blocking */ | |
1530 | bad_fork_cleanup_semundo: | |
1531 | exit_sem(p); | |
1532 | bad_fork_cleanup_audit: | |
1533 | audit_free(p); | |
1da177e4 | 1534 | bad_fork_cleanup_policy: |
cdd6c482 | 1535 | perf_event_free_task(p); |
1da177e4 | 1536 | #ifdef CONFIG_NUMA |
f0be3d32 | 1537 | mpol_put(p->mempolicy); |
e8604cb4 | 1538 | bad_fork_cleanup_threadgroup_lock: |
1da177e4 | 1539 | #endif |
4714d1d3 | 1540 | if (clone_flags & CLONE_THREAD) |
257058ae | 1541 | threadgroup_change_end(current); |
35df17c5 | 1542 | delayacct_tsk_free(p); |
a1261f54 | 1543 | module_put(task_thread_info(p)->exec_domain->module); |
1da177e4 | 1544 | bad_fork_cleanup_count: |
d84f4f99 | 1545 | atomic_dec(&p->cred->user->processes); |
e0e81739 | 1546 | exit_creds(p); |
1da177e4 LT |
1547 | bad_fork_free: |
1548 | free_task(p); | |
fe7d37d1 ON |
1549 | fork_out: |
1550 | return ERR_PTR(retval); | |
1da177e4 LT |
1551 | } |
1552 | ||
f106eee1 ON |
1553 | static inline void init_idle_pids(struct pid_link *links) |
1554 | { | |
1555 | enum pid_type type; | |
1556 | ||
1557 | for (type = PIDTYPE_PID; type < PIDTYPE_MAX; ++type) { | |
1558 | INIT_HLIST_NODE(&links[type].node); /* not really needed */ | |
1559 | links[type].pid = &init_struct_pid; | |
1560 | } | |
1561 | } | |
1562 | ||
0db0628d | 1563 | struct task_struct *fork_idle(int cpu) |
1da177e4 | 1564 | { |
36c8b586 | 1565 | struct task_struct *task; |
62e791c1 | 1566 | task = copy_process(CLONE_VM, 0, 0, NULL, &init_struct_pid, 0); |
f106eee1 ON |
1567 | if (!IS_ERR(task)) { |
1568 | init_idle_pids(task->pids); | |
753ca4f3 | 1569 | init_idle(task, cpu); |
f106eee1 | 1570 | } |
73b9ebfe | 1571 | |
1da177e4 LT |
1572 | return task; |
1573 | } | |
1574 | ||
1da177e4 LT |
1575 | /* |
1576 | * Ok, this is the main fork-routine. | |
1577 | * | |
1578 | * It copies the process, and if successful kick-starts | |
1579 | * it and waits for it to finish using the VM if required. | |
1580 | */ | |
1581 | long do_fork(unsigned long clone_flags, | |
1582 | unsigned long stack_start, | |
1da177e4 LT |
1583 | unsigned long stack_size, |
1584 | int __user *parent_tidptr, | |
1585 | int __user *child_tidptr) | |
1586 | { | |
1587 | struct task_struct *p; | |
1588 | int trace = 0; | |
92476d7f | 1589 | long nr; |
1da177e4 | 1590 | |
09a05394 | 1591 | /* |
4b9d33e6 TH |
1592 | * Determine whether and which event to report to ptracer. When |
1593 | * called from kernel_thread or CLONE_UNTRACED is explicitly | |
1594 | * requested, no event is reported; otherwise, report if the event | |
1595 | * for the type of forking is enabled. | |
09a05394 | 1596 | */ |
e80d6661 | 1597 | if (!(clone_flags & CLONE_UNTRACED)) { |
4b9d33e6 TH |
1598 | if (clone_flags & CLONE_VFORK) |
1599 | trace = PTRACE_EVENT_VFORK; | |
1600 | else if ((clone_flags & CSIGNAL) != SIGCHLD) | |
1601 | trace = PTRACE_EVENT_CLONE; | |
1602 | else | |
1603 | trace = PTRACE_EVENT_FORK; | |
1604 | ||
1605 | if (likely(!ptrace_event_enabled(current, trace))) | |
1606 | trace = 0; | |
1607 | } | |
1da177e4 | 1608 | |
62e791c1 | 1609 | p = copy_process(clone_flags, stack_start, stack_size, |
09a05394 | 1610 | child_tidptr, NULL, trace); |
1da177e4 LT |
1611 | /* |
1612 | * Do this prior waking up the new thread - the thread pointer | |
1613 | * might get invalid after that point, if the thread exits quickly. | |
1614 | */ | |
1615 | if (!IS_ERR(p)) { | |
1616 | struct completion vfork; | |
1617 | ||
0a16b607 MD |
1618 | trace_sched_process_fork(current, p); |
1619 | ||
6c5f3e7b | 1620 | nr = task_pid_vnr(p); |
30e49c26 PE |
1621 | |
1622 | if (clone_flags & CLONE_PARENT_SETTID) | |
1623 | put_user(nr, parent_tidptr); | |
a6f5e063 | 1624 | |
1da177e4 LT |
1625 | if (clone_flags & CLONE_VFORK) { |
1626 | p->vfork_done = &vfork; | |
1627 | init_completion(&vfork); | |
d68b46fe | 1628 | get_task_struct(p); |
1da177e4 LT |
1629 | } |
1630 | ||
3e51e3ed | 1631 | wake_up_new_task(p); |
1da177e4 | 1632 | |
4b9d33e6 TH |
1633 | /* forking complete and child started to run, tell ptracer */ |
1634 | if (unlikely(trace)) | |
1635 | ptrace_event(trace, nr); | |
09a05394 | 1636 | |
1da177e4 | 1637 | if (clone_flags & CLONE_VFORK) { |
d68b46fe ON |
1638 | if (!wait_for_vfork_done(p, &vfork)) |
1639 | ptrace_event(PTRACE_EVENT_VFORK_DONE, nr); | |
1da177e4 LT |
1640 | } |
1641 | } else { | |
92476d7f | 1642 | nr = PTR_ERR(p); |
1da177e4 | 1643 | } |
92476d7f | 1644 | return nr; |
1da177e4 LT |
1645 | } |
1646 | ||
2aa3a7f8 AV |
1647 | /* |
1648 | * Create a kernel thread. | |
1649 | */ | |
1650 | pid_t kernel_thread(int (*fn)(void *), void *arg, unsigned long flags) | |
1651 | { | |
e80d6661 | 1652 | return do_fork(flags|CLONE_VM|CLONE_UNTRACED, (unsigned long)fn, |
2aa3a7f8 AV |
1653 | (unsigned long)arg, NULL, NULL); |
1654 | } | |
2aa3a7f8 | 1655 | |
d2125043 AV |
1656 | #ifdef __ARCH_WANT_SYS_FORK |
1657 | SYSCALL_DEFINE0(fork) | |
1658 | { | |
1659 | #ifdef CONFIG_MMU | |
e80d6661 | 1660 | return do_fork(SIGCHLD, 0, 0, NULL, NULL); |
d2125043 AV |
1661 | #else |
1662 | /* can not support in nommu mode */ | |
5d59e182 | 1663 | return -EINVAL; |
d2125043 AV |
1664 | #endif |
1665 | } | |
1666 | #endif | |
1667 | ||
1668 | #ifdef __ARCH_WANT_SYS_VFORK | |
1669 | SYSCALL_DEFINE0(vfork) | |
1670 | { | |
5d59e182 | 1671 | return do_fork(CLONE_VFORK | CLONE_VM | SIGCHLD, 0, |
d2125043 AV |
1672 | 0, NULL, NULL); |
1673 | } | |
1674 | #endif | |
1675 | ||
1676 | #ifdef __ARCH_WANT_SYS_CLONE | |
1677 | #ifdef CONFIG_CLONE_BACKWARDS | |
1678 | SYSCALL_DEFINE5(clone, unsigned long, clone_flags, unsigned long, newsp, | |
1679 | int __user *, parent_tidptr, | |
1680 | int, tls_val, | |
1681 | int __user *, child_tidptr) | |
1682 | #elif defined(CONFIG_CLONE_BACKWARDS2) | |
1683 | SYSCALL_DEFINE5(clone, unsigned long, newsp, unsigned long, clone_flags, | |
1684 | int __user *, parent_tidptr, | |
1685 | int __user *, child_tidptr, | |
1686 | int, tls_val) | |
dfa9771a MS |
1687 | #elif defined(CONFIG_CLONE_BACKWARDS3) |
1688 | SYSCALL_DEFINE6(clone, unsigned long, clone_flags, unsigned long, newsp, | |
1689 | int, stack_size, | |
1690 | int __user *, parent_tidptr, | |
1691 | int __user *, child_tidptr, | |
1692 | int, tls_val) | |
d2125043 AV |
1693 | #else |
1694 | SYSCALL_DEFINE5(clone, unsigned long, clone_flags, unsigned long, newsp, | |
1695 | int __user *, parent_tidptr, | |
1696 | int __user *, child_tidptr, | |
1697 | int, tls_val) | |
1698 | #endif | |
1699 | { | |
2cf09666 | 1700 | return do_fork(clone_flags, newsp, 0, parent_tidptr, child_tidptr); |
d2125043 AV |
1701 | } |
1702 | #endif | |
1703 | ||
5fd63b30 RT |
1704 | #ifndef ARCH_MIN_MMSTRUCT_ALIGN |
1705 | #define ARCH_MIN_MMSTRUCT_ALIGN 0 | |
1706 | #endif | |
1707 | ||
51cc5068 | 1708 | static void sighand_ctor(void *data) |
aa1757f9 ON |
1709 | { |
1710 | struct sighand_struct *sighand = data; | |
1711 | ||
a35afb83 | 1712 | spin_lock_init(&sighand->siglock); |
b8fceee1 | 1713 | init_waitqueue_head(&sighand->signalfd_wqh); |
aa1757f9 ON |
1714 | } |
1715 | ||
1da177e4 LT |
1716 | void __init proc_caches_init(void) |
1717 | { | |
1718 | sighand_cachep = kmem_cache_create("sighand_cache", | |
1719 | sizeof(struct sighand_struct), 0, | |
2dff4405 VN |
1720 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_DESTROY_BY_RCU| |
1721 | SLAB_NOTRACK, sighand_ctor); | |
1da177e4 LT |
1722 | signal_cachep = kmem_cache_create("signal_cache", |
1723 | sizeof(struct signal_struct), 0, | |
2dff4405 | 1724 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL); |
20c2df83 | 1725 | files_cachep = kmem_cache_create("files_cache", |
1da177e4 | 1726 | sizeof(struct files_struct), 0, |
2dff4405 | 1727 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL); |
20c2df83 | 1728 | fs_cachep = kmem_cache_create("fs_cache", |
1da177e4 | 1729 | sizeof(struct fs_struct), 0, |
2dff4405 | 1730 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL); |
6345d24d LT |
1731 | /* |
1732 | * FIXME! The "sizeof(struct mm_struct)" currently includes the | |
1733 | * whole struct cpumask for the OFFSTACK case. We could change | |
1734 | * this to *only* allocate as much of it as required by the | |
1735 | * maximum number of CPU's we can ever have. The cpumask_allocation | |
1736 | * is at the end of the structure, exactly for that reason. | |
1737 | */ | |
1da177e4 | 1738 | mm_cachep = kmem_cache_create("mm_struct", |
5fd63b30 | 1739 | sizeof(struct mm_struct), ARCH_MIN_MMSTRUCT_ALIGN, |
2dff4405 | 1740 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL); |
33e5d769 | 1741 | vm_area_cachep = KMEM_CACHE(vm_area_struct, SLAB_PANIC); |
8feae131 | 1742 | mmap_init(); |
66577193 | 1743 | nsproxy_cache_init(); |
1da177e4 | 1744 | } |
cf2e340f | 1745 | |
cf2e340f | 1746 | /* |
9bfb23fc | 1747 | * Check constraints on flags passed to the unshare system call. |
cf2e340f | 1748 | */ |
9bfb23fc | 1749 | static int check_unshare_flags(unsigned long unshare_flags) |
cf2e340f | 1750 | { |
9bfb23fc ON |
1751 | if (unshare_flags & ~(CLONE_THREAD|CLONE_FS|CLONE_NEWNS|CLONE_SIGHAND| |
1752 | CLONE_VM|CLONE_FILES|CLONE_SYSVSEM| | |
50804fe3 | 1753 | CLONE_NEWUTS|CLONE_NEWIPC|CLONE_NEWNET| |
b2e0d987 | 1754 | CLONE_NEWUSER|CLONE_NEWPID)) |
9bfb23fc | 1755 | return -EINVAL; |
cf2e340f | 1756 | /* |
9bfb23fc ON |
1757 | * Not implemented, but pretend it works if there is nothing to |
1758 | * unshare. Note that unsharing CLONE_THREAD or CLONE_SIGHAND | |
1759 | * needs to unshare vm. | |
cf2e340f | 1760 | */ |
9bfb23fc ON |
1761 | if (unshare_flags & (CLONE_THREAD | CLONE_SIGHAND | CLONE_VM)) { |
1762 | /* FIXME: get_task_mm() increments ->mm_users */ | |
1763 | if (atomic_read(¤t->mm->mm_users) > 1) | |
1764 | return -EINVAL; | |
1765 | } | |
cf2e340f JD |
1766 | |
1767 | return 0; | |
1768 | } | |
1769 | ||
1770 | /* | |
99d1419d | 1771 | * Unshare the filesystem structure if it is being shared |
cf2e340f JD |
1772 | */ |
1773 | static int unshare_fs(unsigned long unshare_flags, struct fs_struct **new_fsp) | |
1774 | { | |
1775 | struct fs_struct *fs = current->fs; | |
1776 | ||
498052bb AV |
1777 | if (!(unshare_flags & CLONE_FS) || !fs) |
1778 | return 0; | |
1779 | ||
1780 | /* don't need lock here; in the worst case we'll do useless copy */ | |
1781 | if (fs->users == 1) | |
1782 | return 0; | |
1783 | ||
1784 | *new_fsp = copy_fs_struct(fs); | |
1785 | if (!*new_fsp) | |
1786 | return -ENOMEM; | |
cf2e340f JD |
1787 | |
1788 | return 0; | |
1789 | } | |
1790 | ||
cf2e340f | 1791 | /* |
a016f338 | 1792 | * Unshare file descriptor table if it is being shared |
cf2e340f JD |
1793 | */ |
1794 | static int unshare_fd(unsigned long unshare_flags, struct files_struct **new_fdp) | |
1795 | { | |
1796 | struct files_struct *fd = current->files; | |
a016f338 | 1797 | int error = 0; |
cf2e340f JD |
1798 | |
1799 | if ((unshare_flags & CLONE_FILES) && | |
a016f338 JD |
1800 | (fd && atomic_read(&fd->count) > 1)) { |
1801 | *new_fdp = dup_fd(fd, &error); | |
1802 | if (!*new_fdp) | |
1803 | return error; | |
1804 | } | |
cf2e340f JD |
1805 | |
1806 | return 0; | |
1807 | } | |
1808 | ||
cf2e340f JD |
1809 | /* |
1810 | * unshare allows a process to 'unshare' part of the process | |
1811 | * context which was originally shared using clone. copy_* | |
1812 | * functions used by do_fork() cannot be used here directly | |
1813 | * because they modify an inactive task_struct that is being | |
1814 | * constructed. Here we are modifying the current, active, | |
1815 | * task_struct. | |
1816 | */ | |
6559eed8 | 1817 | SYSCALL_DEFINE1(unshare, unsigned long, unshare_flags) |
cf2e340f | 1818 | { |
cf2e340f | 1819 | struct fs_struct *fs, *new_fs = NULL; |
cf2e340f | 1820 | struct files_struct *fd, *new_fd = NULL; |
b2e0d987 | 1821 | struct cred *new_cred = NULL; |
cf7b708c | 1822 | struct nsproxy *new_nsproxy = NULL; |
9edff4ab | 1823 | int do_sysvsem = 0; |
9bfb23fc | 1824 | int err; |
cf2e340f | 1825 | |
b2e0d987 EB |
1826 | /* |
1827 | * If unsharing a user namespace must also unshare the thread. | |
1828 | */ | |
1829 | if (unshare_flags & CLONE_NEWUSER) | |
e66eded8 | 1830 | unshare_flags |= CLONE_THREAD | CLONE_FS; |
50804fe3 EB |
1831 | /* |
1832 | * If unsharing a thread from a thread group, must also unshare vm. | |
1833 | */ | |
1834 | if (unshare_flags & CLONE_THREAD) | |
1835 | unshare_flags |= CLONE_VM; | |
1836 | /* | |
1837 | * If unsharing vm, must also unshare signal handlers. | |
1838 | */ | |
1839 | if (unshare_flags & CLONE_VM) | |
1840 | unshare_flags |= CLONE_SIGHAND; | |
9bfb23fc ON |
1841 | /* |
1842 | * If unsharing namespace, must also unshare filesystem information. | |
1843 | */ | |
1844 | if (unshare_flags & CLONE_NEWNS) | |
1845 | unshare_flags |= CLONE_FS; | |
50804fe3 EB |
1846 | |
1847 | err = check_unshare_flags(unshare_flags); | |
1848 | if (err) | |
1849 | goto bad_unshare_out; | |
6013f67f MS |
1850 | /* |
1851 | * CLONE_NEWIPC must also detach from the undolist: after switching | |
1852 | * to a new ipc namespace, the semaphore arrays from the old | |
1853 | * namespace are unreachable. | |
1854 | */ | |
1855 | if (unshare_flags & (CLONE_NEWIPC|CLONE_SYSVSEM)) | |
9edff4ab | 1856 | do_sysvsem = 1; |
fb0a685c DRO |
1857 | err = unshare_fs(unshare_flags, &new_fs); |
1858 | if (err) | |
9bfb23fc | 1859 | goto bad_unshare_out; |
fb0a685c DRO |
1860 | err = unshare_fd(unshare_flags, &new_fd); |
1861 | if (err) | |
9bfb23fc | 1862 | goto bad_unshare_cleanup_fs; |
b2e0d987 | 1863 | err = unshare_userns(unshare_flags, &new_cred); |
fb0a685c | 1864 | if (err) |
9edff4ab | 1865 | goto bad_unshare_cleanup_fd; |
b2e0d987 EB |
1866 | err = unshare_nsproxy_namespaces(unshare_flags, &new_nsproxy, |
1867 | new_cred, new_fs); | |
1868 | if (err) | |
1869 | goto bad_unshare_cleanup_cred; | |
c0b2fc31 | 1870 | |
b2e0d987 | 1871 | if (new_fs || new_fd || do_sysvsem || new_cred || new_nsproxy) { |
9edff4ab MS |
1872 | if (do_sysvsem) { |
1873 | /* | |
1874 | * CLONE_SYSVSEM is equivalent to sys_exit(). | |
1875 | */ | |
1876 | exit_sem(current); | |
1877 | } | |
ab516013 | 1878 | |
6f977e6b | 1879 | if (new_nsproxy) |
cf7b708c | 1880 | switch_task_namespaces(current, new_nsproxy); |
cf2e340f | 1881 | |
cf7b708c PE |
1882 | task_lock(current); |
1883 | ||
cf2e340f JD |
1884 | if (new_fs) { |
1885 | fs = current->fs; | |
2a4419b5 | 1886 | spin_lock(&fs->lock); |
cf2e340f | 1887 | current->fs = new_fs; |
498052bb AV |
1888 | if (--fs->users) |
1889 | new_fs = NULL; | |
1890 | else | |
1891 | new_fs = fs; | |
2a4419b5 | 1892 | spin_unlock(&fs->lock); |
cf2e340f JD |
1893 | } |
1894 | ||
cf2e340f JD |
1895 | if (new_fd) { |
1896 | fd = current->files; | |
1897 | current->files = new_fd; | |
1898 | new_fd = fd; | |
1899 | } | |
1900 | ||
1901 | task_unlock(current); | |
b2e0d987 EB |
1902 | |
1903 | if (new_cred) { | |
1904 | /* Install the new user namespace */ | |
1905 | commit_creds(new_cred); | |
1906 | new_cred = NULL; | |
1907 | } | |
cf2e340f JD |
1908 | } |
1909 | ||
b2e0d987 EB |
1910 | bad_unshare_cleanup_cred: |
1911 | if (new_cred) | |
1912 | put_cred(new_cred); | |
cf2e340f JD |
1913 | bad_unshare_cleanup_fd: |
1914 | if (new_fd) | |
1915 | put_files_struct(new_fd); | |
1916 | ||
cf2e340f JD |
1917 | bad_unshare_cleanup_fs: |
1918 | if (new_fs) | |
498052bb | 1919 | free_fs_struct(new_fs); |
cf2e340f | 1920 | |
cf2e340f JD |
1921 | bad_unshare_out: |
1922 | return err; | |
1923 | } | |
3b125388 AV |
1924 | |
1925 | /* | |
1926 | * Helper to unshare the files of the current task. | |
1927 | * We don't want to expose copy_files internals to | |
1928 | * the exec layer of the kernel. | |
1929 | */ | |
1930 | ||
1931 | int unshare_files(struct files_struct **displaced) | |
1932 | { | |
1933 | struct task_struct *task = current; | |
50704516 | 1934 | struct files_struct *copy = NULL; |
3b125388 AV |
1935 | int error; |
1936 | ||
1937 | error = unshare_fd(CLONE_FILES, ©); | |
1938 | if (error || !copy) { | |
1939 | *displaced = NULL; | |
1940 | return error; | |
1941 | } | |
1942 | *displaced = task->files; | |
1943 | task_lock(task); | |
1944 | task->files = copy; | |
1945 | task_unlock(task); | |
1946 | return 0; | |
1947 | } |