]> git.proxmox.com Git - mirror_ubuntu-focal-kernel.git/blame - kernel/fork.c
kernel, oom: fix potential pgd_lock deadlock from __mmdrop
[mirror_ubuntu-focal-kernel.git] / kernel / fork.c
CommitLineData
1da177e4
LT
1/*
2 * linux/kernel/fork.c
3 *
4 * Copyright (C) 1991, 1992 Linus Torvalds
5 */
6
7/*
8 * 'fork.c' contains the help-routines for the 'fork' system call
9 * (see also entry.S and others).
10 * Fork is rather simple, once you get the hang of it, but the memory
11 * management can be a bitch. See 'mm/memory.c': 'copy_page_range()'
12 */
13
1da177e4
LT
14#include <linux/slab.h>
15#include <linux/init.h>
16#include <linux/unistd.h>
1da177e4
LT
17#include <linux/module.h>
18#include <linux/vmalloc.h>
19#include <linux/completion.h>
1da177e4
LT
20#include <linux/personality.h>
21#include <linux/mempolicy.h>
22#include <linux/sem.h>
23#include <linux/file.h>
9f3acc31 24#include <linux/fdtable.h>
da9cbc87 25#include <linux/iocontext.h>
1da177e4
LT
26#include <linux/key.h>
27#include <linux/binfmts.h>
28#include <linux/mman.h>
cddb8a5c 29#include <linux/mmu_notifier.h>
1da177e4 30#include <linux/fs.h>
615d6e87
DB
31#include <linux/mm.h>
32#include <linux/vmacache.h>
ab516013 33#include <linux/nsproxy.h>
c59ede7b 34#include <linux/capability.h>
1da177e4 35#include <linux/cpu.h>
b4f48b63 36#include <linux/cgroup.h>
1da177e4 37#include <linux/security.h>
a1e78772 38#include <linux/hugetlb.h>
e2cfabdf 39#include <linux/seccomp.h>
1da177e4
LT
40#include <linux/swap.h>
41#include <linux/syscalls.h>
42#include <linux/jiffies.h>
43#include <linux/futex.h>
8141c7f3 44#include <linux/compat.h>
207205a2 45#include <linux/kthread.h>
7c3ab738 46#include <linux/task_io_accounting_ops.h>
ab2af1f5 47#include <linux/rcupdate.h>
1da177e4
LT
48#include <linux/ptrace.h>
49#include <linux/mount.h>
50#include <linux/audit.h>
78fb7466 51#include <linux/memcontrol.h>
f201ae23 52#include <linux/ftrace.h>
5e2bf014 53#include <linux/proc_fs.h>
1da177e4
LT
54#include <linux/profile.h>
55#include <linux/rmap.h>
f8af4da3 56#include <linux/ksm.h>
1da177e4 57#include <linux/acct.h>
8f0ab514 58#include <linux/tsacct_kern.h>
9f46080c 59#include <linux/cn_proc.h>
ba96a0c8 60#include <linux/freezer.h>
ca74e92b 61#include <linux/delayacct.h>
ad4ecbcb 62#include <linux/taskstats_kern.h>
0a425405 63#include <linux/random.h>
522ed776 64#include <linux/tty.h>
fd0928df 65#include <linux/blkdev.h>
5ad4e53b 66#include <linux/fs_struct.h>
7c9f8861 67#include <linux/magic.h>
cdd6c482 68#include <linux/perf_event.h>
42c4ab41 69#include <linux/posix-timers.h>
8e7cac79 70#include <linux/user-return-notifier.h>
3d5992d2 71#include <linux/oom.h>
ba76149f 72#include <linux/khugepaged.h>
d80e731e 73#include <linux/signalfd.h>
0326f5a9 74#include <linux/uprobes.h>
a27bb332 75#include <linux/aio.h>
52f5684c 76#include <linux/compiler.h>
16db3d3f 77#include <linux/sysctl.h>
5c9a8750 78#include <linux/kcov.h>
1da177e4
LT
79
80#include <asm/pgtable.h>
81#include <asm/pgalloc.h>
82#include <asm/uaccess.h>
83#include <asm/mmu_context.h>
84#include <asm/cacheflush.h>
85#include <asm/tlbflush.h>
86
ad8d75ff
SR
87#include <trace/events/sched.h>
88
43d2b113
KH
89#define CREATE_TRACE_POINTS
90#include <trace/events/task.h>
91
ac1b398d
HS
92/*
93 * Minimum number of threads to boot the kernel
94 */
95#define MIN_THREADS 20
96
97/*
98 * Maximum number of threads
99 */
100#define MAX_THREADS FUTEX_TID_MASK
101
1da177e4
LT
102/*
103 * Protected counters by write_lock_irq(&tasklist_lock)
104 */
105unsigned long total_forks; /* Handle normal Linux uptimes. */
fb0a685c 106int nr_threads; /* The idle threads do not count.. */
1da177e4
LT
107
108int max_threads; /* tunable limit on nr_threads */
109
110DEFINE_PER_CPU(unsigned long, process_counts) = 0;
111
c59923a1 112__cacheline_aligned DEFINE_RWLOCK(tasklist_lock); /* outer */
db1466b3
PM
113
114#ifdef CONFIG_PROVE_RCU
115int lockdep_tasklist_lock_is_held(void)
116{
117 return lockdep_is_held(&tasklist_lock);
118}
119EXPORT_SYMBOL_GPL(lockdep_tasklist_lock_is_held);
120#endif /* #ifdef CONFIG_PROVE_RCU */
1da177e4
LT
121
122int nr_processes(void)
123{
124 int cpu;
125 int total = 0;
126
1d510750 127 for_each_possible_cpu(cpu)
1da177e4
LT
128 total += per_cpu(process_counts, cpu);
129
130 return total;
131}
132
f19b9f74
AM
133void __weak arch_release_task_struct(struct task_struct *tsk)
134{
135}
136
f5e10287 137#ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR
e18b890b 138static struct kmem_cache *task_struct_cachep;
41101809
TG
139
140static inline struct task_struct *alloc_task_struct_node(int node)
141{
142 return kmem_cache_alloc_node(task_struct_cachep, GFP_KERNEL, node);
143}
144
41101809
TG
145static inline void free_task_struct(struct task_struct *tsk)
146{
41101809
TG
147 kmem_cache_free(task_struct_cachep, tsk);
148}
1da177e4
LT
149#endif
150
b235beea 151void __weak arch_release_thread_stack(unsigned long *stack)
f19b9f74
AM
152{
153}
154
b235beea 155#ifndef CONFIG_ARCH_THREAD_STACK_ALLOCATOR
41101809 156
0d15d74a
TG
157/*
158 * Allocate pages if THREAD_SIZE is >= PAGE_SIZE, otherwise use a
159 * kmemcache based allocator.
160 */
ba14a194 161# if THREAD_SIZE >= PAGE_SIZE || defined(CONFIG_VMAP_STACK)
ac496bf4
AL
162
163#ifdef CONFIG_VMAP_STACK
164/*
165 * vmalloc() is a bit slow, and calling vfree() enough times will force a TLB
166 * flush. Try to minimize the number of calls by caching stacks.
167 */
168#define NR_CACHED_STACKS 2
169static DEFINE_PER_CPU(struct vm_struct *, cached_stacks[NR_CACHED_STACKS]);
170#endif
171
ba14a194 172static unsigned long *alloc_thread_stack_node(struct task_struct *tsk, int node)
b69c49b7 173{
ba14a194 174#ifdef CONFIG_VMAP_STACK
ac496bf4
AL
175 void *stack;
176 int i;
177
178 local_irq_disable();
179 for (i = 0; i < NR_CACHED_STACKS; i++) {
180 struct vm_struct *s = this_cpu_read(cached_stacks[i]);
181
182 if (!s)
183 continue;
184 this_cpu_write(cached_stacks[i], NULL);
185
186 tsk->stack_vm_area = s;
187 local_irq_enable();
188 return s->addr;
189 }
190 local_irq_enable();
191
192 stack = __vmalloc_node_range(THREAD_SIZE, THREAD_SIZE,
193 VMALLOC_START, VMALLOC_END,
194 THREADINFO_GFP | __GFP_HIGHMEM,
195 PAGE_KERNEL,
196 0, node, __builtin_return_address(0));
ba14a194
AL
197
198 /*
199 * We can't call find_vm_area() in interrupt context, and
200 * free_thread_stack() can be called in interrupt context,
201 * so cache the vm_struct.
202 */
203 if (stack)
204 tsk->stack_vm_area = find_vm_area(stack);
205 return stack;
206#else
4949148a
VD
207 struct page *page = alloc_pages_node(node, THREADINFO_GFP,
208 THREAD_SIZE_ORDER);
b6a84016
ED
209
210 return page ? page_address(page) : NULL;
ba14a194 211#endif
b69c49b7
FT
212}
213
ba14a194 214static inline void free_thread_stack(struct task_struct *tsk)
b69c49b7 215{
ac496bf4
AL
216#ifdef CONFIG_VMAP_STACK
217 if (task_stack_vm_area(tsk)) {
218 unsigned long flags;
219 int i;
220
221 local_irq_save(flags);
222 for (i = 0; i < NR_CACHED_STACKS; i++) {
223 if (this_cpu_read(cached_stacks[i]))
224 continue;
225
226 this_cpu_write(cached_stacks[i], tsk->stack_vm_area);
227 local_irq_restore(flags);
228 return;
229 }
230 local_irq_restore(flags);
231
ba14a194 232 vfree(tsk->stack);
ac496bf4
AL
233 return;
234 }
235#endif
236
237 __free_pages(virt_to_page(tsk->stack), THREAD_SIZE_ORDER);
b69c49b7 238}
0d15d74a 239# else
b235beea 240static struct kmem_cache *thread_stack_cache;
0d15d74a 241
9521d399 242static unsigned long *alloc_thread_stack_node(struct task_struct *tsk,
0d15d74a
TG
243 int node)
244{
b235beea 245 return kmem_cache_alloc_node(thread_stack_cache, THREADINFO_GFP, node);
0d15d74a
TG
246}
247
ba14a194 248static void free_thread_stack(struct task_struct *tsk)
0d15d74a 249{
ba14a194 250 kmem_cache_free(thread_stack_cache, tsk->stack);
0d15d74a
TG
251}
252
b235beea 253void thread_stack_cache_init(void)
0d15d74a 254{
b235beea 255 thread_stack_cache = kmem_cache_create("thread_stack", THREAD_SIZE,
0d15d74a 256 THREAD_SIZE, 0, NULL);
b235beea 257 BUG_ON(thread_stack_cache == NULL);
0d15d74a
TG
258}
259# endif
b69c49b7
FT
260#endif
261
1da177e4 262/* SLAB cache for signal_struct structures (tsk->signal) */
e18b890b 263static struct kmem_cache *signal_cachep;
1da177e4
LT
264
265/* SLAB cache for sighand_struct structures (tsk->sighand) */
e18b890b 266struct kmem_cache *sighand_cachep;
1da177e4
LT
267
268/* SLAB cache for files_struct structures (tsk->files) */
e18b890b 269struct kmem_cache *files_cachep;
1da177e4
LT
270
271/* SLAB cache for fs_struct structures (tsk->fs) */
e18b890b 272struct kmem_cache *fs_cachep;
1da177e4
LT
273
274/* SLAB cache for vm_area_struct structures */
e18b890b 275struct kmem_cache *vm_area_cachep;
1da177e4
LT
276
277/* SLAB cache for mm_struct structures (tsk->mm) */
e18b890b 278static struct kmem_cache *mm_cachep;
1da177e4 279
ba14a194 280static void account_kernel_stack(struct task_struct *tsk, int account)
c6a7f572 281{
ba14a194
AL
282 void *stack = task_stack_page(tsk);
283 struct vm_struct *vm = task_stack_vm_area(tsk);
284
285 BUILD_BUG_ON(IS_ENABLED(CONFIG_VMAP_STACK) && PAGE_SIZE % 1024 != 0);
286
287 if (vm) {
288 int i;
289
290 BUG_ON(vm->nr_pages != THREAD_SIZE / PAGE_SIZE);
291
292 for (i = 0; i < THREAD_SIZE / PAGE_SIZE; i++) {
293 mod_zone_page_state(page_zone(vm->pages[i]),
294 NR_KERNEL_STACK_KB,
295 PAGE_SIZE / 1024 * account);
296 }
297
298 /* All stack pages belong to the same memcg. */
299 memcg_kmem_update_page_stat(vm->pages[0], MEMCG_KERNEL_STACK_KB,
300 account * (THREAD_SIZE / 1024));
301 } else {
302 /*
303 * All stack pages are in the same zone and belong to the
304 * same memcg.
305 */
306 struct page *first_page = virt_to_page(stack);
307
308 mod_zone_page_state(page_zone(first_page), NR_KERNEL_STACK_KB,
309 THREAD_SIZE / 1024 * account);
310
311 memcg_kmem_update_page_stat(first_page, MEMCG_KERNEL_STACK_KB,
312 account * (THREAD_SIZE / 1024));
313 }
c6a7f572
KM
314}
315
68f24b08 316static void release_task_stack(struct task_struct *tsk)
1da177e4 317{
ba14a194 318 account_kernel_stack(tsk, -1);
b235beea 319 arch_release_thread_stack(tsk->stack);
ba14a194 320 free_thread_stack(tsk);
68f24b08
AL
321 tsk->stack = NULL;
322#ifdef CONFIG_VMAP_STACK
323 tsk->stack_vm_area = NULL;
324#endif
325}
326
327#ifdef CONFIG_THREAD_INFO_IN_TASK
328void put_task_stack(struct task_struct *tsk)
329{
330 if (atomic_dec_and_test(&tsk->stack_refcount))
331 release_task_stack(tsk);
332}
333#endif
334
335void free_task(struct task_struct *tsk)
336{
337#ifndef CONFIG_THREAD_INFO_IN_TASK
338 /*
339 * The task is finally done with both the stack and thread_info,
340 * so free both.
341 */
342 release_task_stack(tsk);
343#else
344 /*
345 * If the task had a separate stack allocation, it should be gone
346 * by now.
347 */
348 WARN_ON_ONCE(atomic_read(&tsk->stack_refcount) != 0);
349#endif
23f78d4a 350 rt_mutex_debug_task_free(tsk);
fb52607a 351 ftrace_graph_exit_task(tsk);
e2cfabdf 352 put_seccomp_filter(tsk);
f19b9f74 353 arch_release_task_struct(tsk);
1da177e4
LT
354 free_task_struct(tsk);
355}
356EXPORT_SYMBOL(free_task);
357
ea6d290c
ON
358static inline void free_signal_struct(struct signal_struct *sig)
359{
97101eb4 360 taskstats_tgid_free(sig);
1c5354de 361 sched_autogroup_exit(sig);
7283094e
MH
362 /*
363 * __mmdrop is not safe to call from softirq context on x86 due to
364 * pgd_dtor so postpone it to the async context
365 */
26db62f1 366 if (sig->oom_mm)
7283094e 367 mmdrop_async(sig->oom_mm);
ea6d290c
ON
368 kmem_cache_free(signal_cachep, sig);
369}
370
371static inline void put_signal_struct(struct signal_struct *sig)
372{
1c5354de 373 if (atomic_dec_and_test(&sig->sigcnt))
ea6d290c
ON
374 free_signal_struct(sig);
375}
376
158d9ebd 377void __put_task_struct(struct task_struct *tsk)
1da177e4 378{
270f722d 379 WARN_ON(!tsk->exit_state);
1da177e4
LT
380 WARN_ON(atomic_read(&tsk->usage));
381 WARN_ON(tsk == current);
382
2e91fa7f 383 cgroup_free(tsk);
156654f4 384 task_numa_free(tsk);
1a2a4d06 385 security_task_free(tsk);
e0e81739 386 exit_creds(tsk);
35df17c5 387 delayacct_tsk_free(tsk);
ea6d290c 388 put_signal_struct(tsk->signal);
1da177e4
LT
389
390 if (!profile_handoff_task(tsk))
391 free_task(tsk);
392}
77c100c8 393EXPORT_SYMBOL_GPL(__put_task_struct);
1da177e4 394
6c0a9fa6 395void __init __weak arch_task_cache_init(void) { }
61c4628b 396
ff691f6e
HS
397/*
398 * set_max_threads
399 */
16db3d3f 400static void set_max_threads(unsigned int max_threads_suggested)
ff691f6e 401{
ac1b398d 402 u64 threads;
ff691f6e
HS
403
404 /*
ac1b398d
HS
405 * The number of threads shall be limited such that the thread
406 * structures may only consume a small part of the available memory.
ff691f6e 407 */
ac1b398d
HS
408 if (fls64(totalram_pages) + fls64(PAGE_SIZE) > 64)
409 threads = MAX_THREADS;
410 else
411 threads = div64_u64((u64) totalram_pages * (u64) PAGE_SIZE,
412 (u64) THREAD_SIZE * 8UL);
413
16db3d3f
HS
414 if (threads > max_threads_suggested)
415 threads = max_threads_suggested;
416
ac1b398d 417 max_threads = clamp_t(u64, threads, MIN_THREADS, MAX_THREADS);
ff691f6e
HS
418}
419
5aaeb5c0
IM
420#ifdef CONFIG_ARCH_WANTS_DYNAMIC_TASK_STRUCT
421/* Initialized by the architecture: */
422int arch_task_struct_size __read_mostly;
423#endif
0c8c0f03 424
ff691f6e 425void __init fork_init(void)
1da177e4 426{
25f9c081 427 int i;
f5e10287 428#ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR
1da177e4
LT
429#ifndef ARCH_MIN_TASKALIGN
430#define ARCH_MIN_TASKALIGN L1_CACHE_BYTES
431#endif
432 /* create a slab on which task_structs can be allocated */
5d097056
VD
433 task_struct_cachep = kmem_cache_create("task_struct",
434 arch_task_struct_size, ARCH_MIN_TASKALIGN,
435 SLAB_PANIC|SLAB_NOTRACK|SLAB_ACCOUNT, NULL);
1da177e4
LT
436#endif
437
61c4628b
SS
438 /* do the arch specific task caches init */
439 arch_task_cache_init();
440
16db3d3f 441 set_max_threads(MAX_THREADS);
1da177e4
LT
442
443 init_task.signal->rlim[RLIMIT_NPROC].rlim_cur = max_threads/2;
444 init_task.signal->rlim[RLIMIT_NPROC].rlim_max = max_threads/2;
445 init_task.signal->rlim[RLIMIT_SIGPENDING] =
446 init_task.signal->rlim[RLIMIT_NPROC];
b376c3e1 447
25f9c081
EB
448 for (i = 0; i < UCOUNT_COUNTS; i++) {
449 init_user_ns.ucount_max[i] = max_threads/2;
450 }
1da177e4
LT
451}
452
52f5684c 453int __weak arch_dup_task_struct(struct task_struct *dst,
61c4628b
SS
454 struct task_struct *src)
455{
456 *dst = *src;
457 return 0;
458}
459
d4311ff1
AT
460void set_task_stack_end_magic(struct task_struct *tsk)
461{
462 unsigned long *stackend;
463
464 stackend = end_of_stack(tsk);
465 *stackend = STACK_END_MAGIC; /* for overflow detection */
466}
467
725fc629 468static struct task_struct *dup_task_struct(struct task_struct *orig, int node)
1da177e4
LT
469{
470 struct task_struct *tsk;
b235beea 471 unsigned long *stack;
ba14a194 472 struct vm_struct *stack_vm_area;
3e26c149 473 int err;
1da177e4 474
725fc629
AK
475 if (node == NUMA_NO_NODE)
476 node = tsk_fork_get_node(orig);
504f52b5 477 tsk = alloc_task_struct_node(node);
1da177e4
LT
478 if (!tsk)
479 return NULL;
480
b235beea
LT
481 stack = alloc_thread_stack_node(tsk, node);
482 if (!stack)
f19b9f74 483 goto free_tsk;
1da177e4 484
ba14a194
AL
485 stack_vm_area = task_stack_vm_area(tsk);
486
fb0a685c 487 err = arch_dup_task_struct(tsk, orig);
ba14a194
AL
488
489 /*
490 * arch_dup_task_struct() clobbers the stack-related fields. Make
491 * sure they're properly initialized before using any stack-related
492 * functions again.
493 */
494 tsk->stack = stack;
495#ifdef CONFIG_VMAP_STACK
496 tsk->stack_vm_area = stack_vm_area;
497#endif
68f24b08
AL
498#ifdef CONFIG_THREAD_INFO_IN_TASK
499 atomic_set(&tsk->stack_refcount, 1);
500#endif
ba14a194 501
164c33c6 502 if (err)
b235beea 503 goto free_stack;
164c33c6 504
dbd95212
KC
505#ifdef CONFIG_SECCOMP
506 /*
507 * We must handle setting up seccomp filters once we're under
508 * the sighand lock in case orig has changed between now and
509 * then. Until then, filter must be NULL to avoid messing up
510 * the usage counts on the error path calling free_task.
511 */
512 tsk->seccomp.filter = NULL;
513#endif
87bec58a
AM
514
515 setup_thread_stack(tsk, orig);
8e7cac79 516 clear_user_return_notifier(tsk);
f26f9aff 517 clear_tsk_need_resched(tsk);
d4311ff1 518 set_task_stack_end_magic(tsk);
1da177e4 519
0a425405
AV
520#ifdef CONFIG_CC_STACKPROTECTOR
521 tsk->stack_canary = get_random_int();
522#endif
523
fb0a685c
DRO
524 /*
525 * One for us, one for whoever does the "release_task()" (usually
526 * parent)
527 */
528 atomic_set(&tsk->usage, 2);
6c5c9341 529#ifdef CONFIG_BLK_DEV_IO_TRACE
2056a782 530 tsk->btrace_seq = 0;
6c5c9341 531#endif
a0aa7f68 532 tsk->splice_pipe = NULL;
5640f768 533 tsk->task_frag.page = NULL;
093e5840 534 tsk->wake_q.next = NULL;
c6a7f572 535
ba14a194 536 account_kernel_stack(tsk, 1);
c6a7f572 537
5c9a8750
DV
538 kcov_task_init(tsk);
539
1da177e4 540 return tsk;
61c4628b 541
b235beea 542free_stack:
ba14a194 543 free_thread_stack(tsk);
f19b9f74 544free_tsk:
61c4628b
SS
545 free_task_struct(tsk);
546 return NULL;
1da177e4
LT
547}
548
549#ifdef CONFIG_MMU
a39bc516 550static int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm)
1da177e4 551{
297c5eee 552 struct vm_area_struct *mpnt, *tmp, *prev, **pprev;
1da177e4
LT
553 struct rb_node **rb_link, *rb_parent;
554 int retval;
555 unsigned long charge;
1da177e4 556
32cdba1e 557 uprobe_start_dup_mmap();
7c051267
MH
558 if (down_write_killable(&oldmm->mmap_sem)) {
559 retval = -EINTR;
560 goto fail_uprobe_end;
561 }
ec8c0446 562 flush_cache_dup_mm(oldmm);
f8ac4ec9 563 uprobe_dup_mmap(oldmm, mm);
ad339451
IM
564 /*
565 * Not linked in yet - no deadlock potential:
566 */
567 down_write_nested(&mm->mmap_sem, SINGLE_DEPTH_NESTING);
7ee78232 568
90f31d0e
KK
569 /* No ordering required: file already has been exposed. */
570 RCU_INIT_POINTER(mm->exe_file, get_mm_exe_file(oldmm));
571
4f7d4614 572 mm->total_vm = oldmm->total_vm;
84638335 573 mm->data_vm = oldmm->data_vm;
4f7d4614
VD
574 mm->exec_vm = oldmm->exec_vm;
575 mm->stack_vm = oldmm->stack_vm;
576
1da177e4
LT
577 rb_link = &mm->mm_rb.rb_node;
578 rb_parent = NULL;
579 pprev = &mm->mmap;
f8af4da3 580 retval = ksm_fork(mm, oldmm);
ba76149f
AA
581 if (retval)
582 goto out;
583 retval = khugepaged_fork(mm, oldmm);
f8af4da3
HD
584 if (retval)
585 goto out;
1da177e4 586
297c5eee 587 prev = NULL;
fd3e42fc 588 for (mpnt = oldmm->mmap; mpnt; mpnt = mpnt->vm_next) {
1da177e4
LT
589 struct file *file;
590
591 if (mpnt->vm_flags & VM_DONTCOPY) {
84638335 592 vm_stat_account(mm, mpnt->vm_flags, -vma_pages(mpnt));
1da177e4
LT
593 continue;
594 }
595 charge = 0;
596 if (mpnt->vm_flags & VM_ACCOUNT) {
b2412b7f
HS
597 unsigned long len = vma_pages(mpnt);
598
191c5424 599 if (security_vm_enough_memory_mm(oldmm, len)) /* sic */
1da177e4
LT
600 goto fail_nomem;
601 charge = len;
602 }
e94b1766 603 tmp = kmem_cache_alloc(vm_area_cachep, GFP_KERNEL);
1da177e4
LT
604 if (!tmp)
605 goto fail_nomem;
606 *tmp = *mpnt;
5beb4930 607 INIT_LIST_HEAD(&tmp->anon_vma_chain);
ef0855d3
ON
608 retval = vma_dup_policy(mpnt, tmp);
609 if (retval)
1da177e4 610 goto fail_nomem_policy;
a247c3a9 611 tmp->vm_mm = mm;
5beb4930
RR
612 if (anon_vma_fork(tmp, mpnt))
613 goto fail_nomem_anon_vma_fork;
de60f5f1
EM
614 tmp->vm_flags &=
615 ~(VM_LOCKED|VM_LOCKONFAULT|VM_UFFD_MISSING|VM_UFFD_WP);
297c5eee 616 tmp->vm_next = tmp->vm_prev = NULL;
745f234b 617 tmp->vm_userfaultfd_ctx = NULL_VM_UFFD_CTX;
1da177e4
LT
618 file = tmp->vm_file;
619 if (file) {
496ad9aa 620 struct inode *inode = file_inode(file);
b88ed205
HD
621 struct address_space *mapping = file->f_mapping;
622
1da177e4
LT
623 get_file(file);
624 if (tmp->vm_flags & VM_DENYWRITE)
625 atomic_dec(&inode->i_writecount);
83cde9e8 626 i_mmap_lock_write(mapping);
b88ed205 627 if (tmp->vm_flags & VM_SHARED)
4bb5f5d9 628 atomic_inc(&mapping->i_mmap_writable);
b88ed205
HD
629 flush_dcache_mmap_lock(mapping);
630 /* insert tmp into the share list, just after mpnt */
27ba0644
KS
631 vma_interval_tree_insert_after(tmp, mpnt,
632 &mapping->i_mmap);
b88ed205 633 flush_dcache_mmap_unlock(mapping);
83cde9e8 634 i_mmap_unlock_write(mapping);
1da177e4
LT
635 }
636
a1e78772
MG
637 /*
638 * Clear hugetlb-related page reserves for children. This only
639 * affects MAP_PRIVATE mappings. Faults generated by the child
640 * are not guaranteed to succeed, even if read-only
641 */
642 if (is_vm_hugetlb_page(tmp))
643 reset_vma_resv_huge_pages(tmp);
644
1da177e4 645 /*
7ee78232 646 * Link in the new vma and copy the page table entries.
1da177e4 647 */
1da177e4
LT
648 *pprev = tmp;
649 pprev = &tmp->vm_next;
297c5eee
LT
650 tmp->vm_prev = prev;
651 prev = tmp;
1da177e4
LT
652
653 __vma_link_rb(mm, tmp, rb_link, rb_parent);
654 rb_link = &tmp->vm_rb.rb_right;
655 rb_parent = &tmp->vm_rb;
656
657 mm->map_count++;
0b0db14c 658 retval = copy_page_range(mm, oldmm, mpnt);
1da177e4
LT
659
660 if (tmp->vm_ops && tmp->vm_ops->open)
661 tmp->vm_ops->open(tmp);
662
663 if (retval)
664 goto out;
665 }
d6dd61c8
JF
666 /* a new mm has just been created */
667 arch_dup_mmap(oldmm, mm);
1da177e4 668 retval = 0;
1da177e4 669out:
7ee78232 670 up_write(&mm->mmap_sem);
fd3e42fc 671 flush_tlb_mm(oldmm);
1da177e4 672 up_write(&oldmm->mmap_sem);
7c051267 673fail_uprobe_end:
32cdba1e 674 uprobe_end_dup_mmap();
1da177e4 675 return retval;
5beb4930 676fail_nomem_anon_vma_fork:
ef0855d3 677 mpol_put(vma_policy(tmp));
1da177e4
LT
678fail_nomem_policy:
679 kmem_cache_free(vm_area_cachep, tmp);
680fail_nomem:
681 retval = -ENOMEM;
682 vm_unacct_memory(charge);
683 goto out;
684}
685
fb0a685c 686static inline int mm_alloc_pgd(struct mm_struct *mm)
1da177e4
LT
687{
688 mm->pgd = pgd_alloc(mm);
689 if (unlikely(!mm->pgd))
690 return -ENOMEM;
691 return 0;
692}
693
fb0a685c 694static inline void mm_free_pgd(struct mm_struct *mm)
1da177e4 695{
5e541973 696 pgd_free(mm, mm->pgd);
1da177e4
LT
697}
698#else
90f31d0e
KK
699static int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm)
700{
701 down_write(&oldmm->mmap_sem);
702 RCU_INIT_POINTER(mm->exe_file, get_mm_exe_file(oldmm));
703 up_write(&oldmm->mmap_sem);
704 return 0;
705}
1da177e4
LT
706#define mm_alloc_pgd(mm) (0)
707#define mm_free_pgd(mm)
708#endif /* CONFIG_MMU */
709
23ff4440 710__cacheline_aligned_in_smp DEFINE_SPINLOCK(mmlist_lock);
1da177e4 711
e94b1766 712#define allocate_mm() (kmem_cache_alloc(mm_cachep, GFP_KERNEL))
1da177e4
LT
713#define free_mm(mm) (kmem_cache_free(mm_cachep, (mm)))
714
4cb0e11b
HK
715static unsigned long default_dump_filter = MMF_DUMP_FILTER_DEFAULT;
716
717static int __init coredump_filter_setup(char *s)
718{
719 default_dump_filter =
720 (simple_strtoul(s, NULL, 0) << MMF_DUMP_FILTER_SHIFT) &
721 MMF_DUMP_FILTER_MASK;
722 return 1;
723}
724
725__setup("coredump_filter=", coredump_filter_setup);
726
1da177e4
LT
727#include <linux/init_task.h>
728
858f0993
AD
729static void mm_init_aio(struct mm_struct *mm)
730{
731#ifdef CONFIG_AIO
732 spin_lock_init(&mm->ioctx_lock);
db446a08 733 mm->ioctx_table = NULL;
858f0993
AD
734#endif
735}
736
33144e84
VD
737static void mm_init_owner(struct mm_struct *mm, struct task_struct *p)
738{
739#ifdef CONFIG_MEMCG
740 mm->owner = p;
741#endif
742}
743
fb0a685c 744static struct mm_struct *mm_init(struct mm_struct *mm, struct task_struct *p)
1da177e4 745{
41f727fd
VD
746 mm->mmap = NULL;
747 mm->mm_rb = RB_ROOT;
748 mm->vmacache_seqnum = 0;
1da177e4
LT
749 atomic_set(&mm->mm_users, 1);
750 atomic_set(&mm->mm_count, 1);
751 init_rwsem(&mm->mmap_sem);
752 INIT_LIST_HEAD(&mm->mmlist);
999d9fc1 753 mm->core_state = NULL;
e1f56c89 754 atomic_long_set(&mm->nr_ptes, 0);
2d2f5119 755 mm_nr_pmds_init(mm);
41f727fd
VD
756 mm->map_count = 0;
757 mm->locked_vm = 0;
ce65cefa 758 mm->pinned_vm = 0;
d559db08 759 memset(&mm->rss_stat, 0, sizeof(mm->rss_stat));
1da177e4 760 spin_lock_init(&mm->page_table_lock);
41f727fd 761 mm_init_cpumask(mm);
858f0993 762 mm_init_aio(mm);
cf475ad2 763 mm_init_owner(mm, p);
41f727fd 764 mmu_notifier_mm_init(mm);
20841405 765 clear_tlb_flush_pending(mm);
41f727fd
VD
766#if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS
767 mm->pmd_huge_pte = NULL;
768#endif
1da177e4 769
a0715cc2
AT
770 if (current->mm) {
771 mm->flags = current->mm->flags & MMF_INIT_MASK;
772 mm->def_flags = current->mm->def_flags & VM_INIT_DEF_MASK;
773 } else {
774 mm->flags = default_dump_filter;
1da177e4 775 mm->def_flags = 0;
a0715cc2
AT
776 }
777
41f727fd
VD
778 if (mm_alloc_pgd(mm))
779 goto fail_nopgd;
780
781 if (init_new_context(p, mm))
782 goto fail_nocontext;
78fb7466 783
41f727fd
VD
784 return mm;
785
786fail_nocontext:
787 mm_free_pgd(mm);
788fail_nopgd:
1da177e4
LT
789 free_mm(mm);
790 return NULL;
791}
792
c3f0327f
KK
793static void check_mm(struct mm_struct *mm)
794{
795 int i;
796
797 for (i = 0; i < NR_MM_COUNTERS; i++) {
798 long x = atomic_long_read(&mm->rss_stat.count[i]);
799
800 if (unlikely(x))
801 printk(KERN_ALERT "BUG: Bad rss-counter state "
802 "mm:%p idx:%d val:%ld\n", mm, i, x);
803 }
b30fe6c7
KS
804
805 if (atomic_long_read(&mm->nr_ptes))
806 pr_alert("BUG: non-zero nr_ptes on freeing mm: %ld\n",
807 atomic_long_read(&mm->nr_ptes));
808 if (mm_nr_pmds(mm))
809 pr_alert("BUG: non-zero nr_pmds on freeing mm: %ld\n",
810 mm_nr_pmds(mm));
811
e009bb30 812#if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS
96dad67f 813 VM_BUG_ON_MM(mm->pmd_huge_pte, mm);
c3f0327f
KK
814#endif
815}
816
1da177e4
LT
817/*
818 * Allocate and initialize an mm_struct.
819 */
fb0a685c 820struct mm_struct *mm_alloc(void)
1da177e4 821{
fb0a685c 822 struct mm_struct *mm;
1da177e4
LT
823
824 mm = allocate_mm();
de03c72c
KM
825 if (!mm)
826 return NULL;
827
828 memset(mm, 0, sizeof(*mm));
6345d24d 829 return mm_init(mm, current);
1da177e4
LT
830}
831
832/*
833 * Called when the last reference to the mm
834 * is dropped: either by a lazy thread or by
835 * mmput. Free the page directory and the mm.
836 */
7ad5b3a5 837void __mmdrop(struct mm_struct *mm)
1da177e4
LT
838{
839 BUG_ON(mm == &init_mm);
840 mm_free_pgd(mm);
841 destroy_context(mm);
cddb8a5c 842 mmu_notifier_mm_destroy(mm);
c3f0327f 843 check_mm(mm);
1da177e4
LT
844 free_mm(mm);
845}
6d4e4c4f 846EXPORT_SYMBOL_GPL(__mmdrop);
1da177e4 847
ec8d7c14
MH
848static inline void __mmput(struct mm_struct *mm)
849{
850 VM_BUG_ON(atomic_read(&mm->mm_users));
851
852 uprobe_clear_state(mm);
853 exit_aio(mm);
854 ksm_exit(mm);
855 khugepaged_exit(mm); /* must run before exit_mmap */
856 exit_mmap(mm);
857 set_mm_exe_file(mm, NULL);
858 if (!list_empty(&mm->mmlist)) {
859 spin_lock(&mmlist_lock);
860 list_del(&mm->mmlist);
861 spin_unlock(&mmlist_lock);
862 }
863 if (mm->binfmt)
864 module_put(mm->binfmt->module);
865 mmdrop(mm);
866}
867
1da177e4
LT
868/*
869 * Decrement the use count and release all resources for an mm.
870 */
871void mmput(struct mm_struct *mm)
872{
0ae26f1b
AM
873 might_sleep();
874
ec8d7c14
MH
875 if (atomic_dec_and_test(&mm->mm_users))
876 __mmput(mm);
877}
878EXPORT_SYMBOL_GPL(mmput);
879
7ef949d7 880#ifdef CONFIG_MMU
ec8d7c14
MH
881static void mmput_async_fn(struct work_struct *work)
882{
883 struct mm_struct *mm = container_of(work, struct mm_struct, async_put_work);
884 __mmput(mm);
885}
886
887void mmput_async(struct mm_struct *mm)
888{
1da177e4 889 if (atomic_dec_and_test(&mm->mm_users)) {
ec8d7c14
MH
890 INIT_WORK(&mm->async_put_work, mmput_async_fn);
891 schedule_work(&mm->async_put_work);
1da177e4
LT
892 }
893}
7ef949d7 894#endif
1da177e4 895
90f31d0e
KK
896/**
897 * set_mm_exe_file - change a reference to the mm's executable file
898 *
899 * This changes mm's executable file (shown as symlink /proc/[pid]/exe).
900 *
6e399cd1
DB
901 * Main users are mmput() and sys_execve(). Callers prevent concurrent
902 * invocations: in mmput() nobody alive left, in execve task is single
903 * threaded. sys_prctl(PR_SET_MM_MAP/EXE_FILE) also needs to set the
904 * mm->exe_file, but does so without using set_mm_exe_file() in order
905 * to do avoid the need for any locks.
90f31d0e 906 */
38646013
JS
907void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file)
908{
6e399cd1
DB
909 struct file *old_exe_file;
910
911 /*
912 * It is safe to dereference the exe_file without RCU as
913 * this function is only called if nobody else can access
914 * this mm -- see comment above for justification.
915 */
916 old_exe_file = rcu_dereference_raw(mm->exe_file);
90f31d0e 917
38646013
JS
918 if (new_exe_file)
919 get_file(new_exe_file);
90f31d0e
KK
920 rcu_assign_pointer(mm->exe_file, new_exe_file);
921 if (old_exe_file)
922 fput(old_exe_file);
38646013
JS
923}
924
90f31d0e
KK
925/**
926 * get_mm_exe_file - acquire a reference to the mm's executable file
927 *
928 * Returns %NULL if mm has no associated executable file.
929 * User must release file via fput().
930 */
38646013
JS
931struct file *get_mm_exe_file(struct mm_struct *mm)
932{
933 struct file *exe_file;
934
90f31d0e
KK
935 rcu_read_lock();
936 exe_file = rcu_dereference(mm->exe_file);
937 if (exe_file && !get_file_rcu(exe_file))
938 exe_file = NULL;
939 rcu_read_unlock();
38646013
JS
940 return exe_file;
941}
11163348 942EXPORT_SYMBOL(get_mm_exe_file);
38646013 943
cd81a917
MG
944/**
945 * get_task_exe_file - acquire a reference to the task's executable file
946 *
947 * Returns %NULL if task's mm (if any) has no associated executable file or
948 * this is a kernel thread with borrowed mm (see the comment above get_task_mm).
949 * User must release file via fput().
950 */
951struct file *get_task_exe_file(struct task_struct *task)
952{
953 struct file *exe_file = NULL;
954 struct mm_struct *mm;
955
956 task_lock(task);
957 mm = task->mm;
958 if (mm) {
959 if (!(task->flags & PF_KTHREAD))
960 exe_file = get_mm_exe_file(mm);
961 }
962 task_unlock(task);
963 return exe_file;
964}
965EXPORT_SYMBOL(get_task_exe_file);
38646013 966
1da177e4
LT
967/**
968 * get_task_mm - acquire a reference to the task's mm
969 *
246bb0b1 970 * Returns %NULL if the task has no mm. Checks PF_KTHREAD (meaning
1da177e4
LT
971 * this kernel workthread has transiently adopted a user mm with use_mm,
972 * to do its AIO) is not set and if so returns a reference to it, after
973 * bumping up the use count. User must release the mm via mmput()
974 * after use. Typically used by /proc and ptrace.
975 */
976struct mm_struct *get_task_mm(struct task_struct *task)
977{
978 struct mm_struct *mm;
979
980 task_lock(task);
981 mm = task->mm;
982 if (mm) {
246bb0b1 983 if (task->flags & PF_KTHREAD)
1da177e4
LT
984 mm = NULL;
985 else
986 atomic_inc(&mm->mm_users);
987 }
988 task_unlock(task);
989 return mm;
990}
991EXPORT_SYMBOL_GPL(get_task_mm);
992
8cdb878d
CY
993struct mm_struct *mm_access(struct task_struct *task, unsigned int mode)
994{
995 struct mm_struct *mm;
996 int err;
997
998 err = mutex_lock_killable(&task->signal->cred_guard_mutex);
999 if (err)
1000 return ERR_PTR(err);
1001
1002 mm = get_task_mm(task);
1003 if (mm && mm != current->mm &&
1004 !ptrace_may_access(task, mode)) {
1005 mmput(mm);
1006 mm = ERR_PTR(-EACCES);
1007 }
1008 mutex_unlock(&task->signal->cred_guard_mutex);
1009
1010 return mm;
1011}
1012
57b59c4a 1013static void complete_vfork_done(struct task_struct *tsk)
c415c3b4 1014{
d68b46fe 1015 struct completion *vfork;
c415c3b4 1016
d68b46fe
ON
1017 task_lock(tsk);
1018 vfork = tsk->vfork_done;
1019 if (likely(vfork)) {
1020 tsk->vfork_done = NULL;
1021 complete(vfork);
1022 }
1023 task_unlock(tsk);
1024}
1025
1026static int wait_for_vfork_done(struct task_struct *child,
1027 struct completion *vfork)
1028{
1029 int killed;
1030
1031 freezer_do_not_count();
1032 killed = wait_for_completion_killable(vfork);
1033 freezer_count();
1034
1035 if (killed) {
1036 task_lock(child);
1037 child->vfork_done = NULL;
1038 task_unlock(child);
1039 }
1040
1041 put_task_struct(child);
1042 return killed;
c415c3b4
ON
1043}
1044
1da177e4
LT
1045/* Please note the differences between mmput and mm_release.
1046 * mmput is called whenever we stop holding onto a mm_struct,
1047 * error success whatever.
1048 *
1049 * mm_release is called after a mm_struct has been removed
1050 * from the current process.
1051 *
1052 * This difference is important for error handling, when we
1053 * only half set up a mm_struct for a new process and need to restore
1054 * the old one. Because we mmput the new mm_struct before
1055 * restoring the old one. . .
1056 * Eric Biederman 10 January 1998
1057 */
1058void mm_release(struct task_struct *tsk, struct mm_struct *mm)
1059{
8141c7f3
LT
1060 /* Get rid of any futexes when releasing the mm */
1061#ifdef CONFIG_FUTEX
fc6b177d 1062 if (unlikely(tsk->robust_list)) {
8141c7f3 1063 exit_robust_list(tsk);
fc6b177d
PZ
1064 tsk->robust_list = NULL;
1065 }
8141c7f3 1066#ifdef CONFIG_COMPAT
fc6b177d 1067 if (unlikely(tsk->compat_robust_list)) {
8141c7f3 1068 compat_exit_robust_list(tsk);
fc6b177d
PZ
1069 tsk->compat_robust_list = NULL;
1070 }
8141c7f3 1071#endif
322a2c10
TG
1072 if (unlikely(!list_empty(&tsk->pi_state_list)))
1073 exit_pi_state_list(tsk);
8141c7f3
LT
1074#endif
1075
0326f5a9
SD
1076 uprobe_free_utask(tsk);
1077
1da177e4
LT
1078 /* Get rid of any cached register state */
1079 deactivate_mm(tsk, mm);
1080
fec1d011 1081 /*
735f2770
MH
1082 * Signal userspace if we're not exiting with a core dump
1083 * because we want to leave the value intact for debugging
1084 * purposes.
fec1d011 1085 */
9c8a8228 1086 if (tsk->clear_child_tid) {
735f2770 1087 if (!(tsk->signal->flags & SIGNAL_GROUP_COREDUMP) &&
9c8a8228
ED
1088 atomic_read(&mm->mm_users) > 1) {
1089 /*
1090 * We don't check the error code - if userspace has
1091 * not set up a proper pointer then tough luck.
1092 */
1093 put_user(0, tsk->clear_child_tid);
1094 sys_futex(tsk->clear_child_tid, FUTEX_WAKE,
1095 1, NULL, NULL, 0);
1096 }
1da177e4 1097 tsk->clear_child_tid = NULL;
1da177e4 1098 }
f7505d64
KK
1099
1100 /*
1101 * All done, finally we can wake up parent and return this mm to him.
1102 * Also kthread_stop() uses this completion for synchronization.
1103 */
1104 if (tsk->vfork_done)
1105 complete_vfork_done(tsk);
1da177e4
LT
1106}
1107
a0a7ec30
JD
1108/*
1109 * Allocate a new mm structure and copy contents from the
1110 * mm structure of the passed in task structure.
1111 */
ff252c1f 1112static struct mm_struct *dup_mm(struct task_struct *tsk)
a0a7ec30
JD
1113{
1114 struct mm_struct *mm, *oldmm = current->mm;
1115 int err;
1116
a0a7ec30
JD
1117 mm = allocate_mm();
1118 if (!mm)
1119 goto fail_nomem;
1120
1121 memcpy(mm, oldmm, sizeof(*mm));
1122
78fb7466 1123 if (!mm_init(mm, tsk))
a0a7ec30
JD
1124 goto fail_nomem;
1125
a0a7ec30
JD
1126 err = dup_mmap(mm, oldmm);
1127 if (err)
1128 goto free_pt;
1129
1130 mm->hiwater_rss = get_mm_rss(mm);
1131 mm->hiwater_vm = mm->total_vm;
1132
801460d0
HS
1133 if (mm->binfmt && !try_module_get(mm->binfmt->module))
1134 goto free_pt;
1135
a0a7ec30
JD
1136 return mm;
1137
1138free_pt:
801460d0
HS
1139 /* don't put binfmt in mmput, we haven't got module yet */
1140 mm->binfmt = NULL;
a0a7ec30
JD
1141 mmput(mm);
1142
1143fail_nomem:
1144 return NULL;
a0a7ec30
JD
1145}
1146
fb0a685c 1147static int copy_mm(unsigned long clone_flags, struct task_struct *tsk)
1da177e4 1148{
fb0a685c 1149 struct mm_struct *mm, *oldmm;
1da177e4
LT
1150 int retval;
1151
1152 tsk->min_flt = tsk->maj_flt = 0;
1153 tsk->nvcsw = tsk->nivcsw = 0;
17406b82
MSB
1154#ifdef CONFIG_DETECT_HUNG_TASK
1155 tsk->last_switch_count = tsk->nvcsw + tsk->nivcsw;
1156#endif
1da177e4
LT
1157
1158 tsk->mm = NULL;
1159 tsk->active_mm = NULL;
1160
1161 /*
1162 * Are we cloning a kernel thread?
1163 *
1164 * We need to steal a active VM for that..
1165 */
1166 oldmm = current->mm;
1167 if (!oldmm)
1168 return 0;
1169
615d6e87
DB
1170 /* initialize the new vmacache entries */
1171 vmacache_flush(tsk);
1172
1da177e4
LT
1173 if (clone_flags & CLONE_VM) {
1174 atomic_inc(&oldmm->mm_users);
1175 mm = oldmm;
1da177e4
LT
1176 goto good_mm;
1177 }
1178
1179 retval = -ENOMEM;
a0a7ec30 1180 mm = dup_mm(tsk);
1da177e4
LT
1181 if (!mm)
1182 goto fail_nomem;
1183
1da177e4
LT
1184good_mm:
1185 tsk->mm = mm;
1186 tsk->active_mm = mm;
1187 return 0;
1188
1da177e4
LT
1189fail_nomem:
1190 return retval;
1da177e4
LT
1191}
1192
a39bc516 1193static int copy_fs(unsigned long clone_flags, struct task_struct *tsk)
1da177e4 1194{
498052bb 1195 struct fs_struct *fs = current->fs;
1da177e4 1196 if (clone_flags & CLONE_FS) {
498052bb 1197 /* tsk->fs is already what we want */
2a4419b5 1198 spin_lock(&fs->lock);
498052bb 1199 if (fs->in_exec) {
2a4419b5 1200 spin_unlock(&fs->lock);
498052bb
AV
1201 return -EAGAIN;
1202 }
1203 fs->users++;
2a4419b5 1204 spin_unlock(&fs->lock);
1da177e4
LT
1205 return 0;
1206 }
498052bb 1207 tsk->fs = copy_fs_struct(fs);
1da177e4
LT
1208 if (!tsk->fs)
1209 return -ENOMEM;
1210 return 0;
1211}
1212
fb0a685c 1213static int copy_files(unsigned long clone_flags, struct task_struct *tsk)
a016f338
JD
1214{
1215 struct files_struct *oldf, *newf;
1216 int error = 0;
1217
1218 /*
1219 * A background process may not have any files ...
1220 */
1221 oldf = current->files;
1222 if (!oldf)
1223 goto out;
1224
1225 if (clone_flags & CLONE_FILES) {
1226 atomic_inc(&oldf->count);
1227 goto out;
1228 }
1229
a016f338
JD
1230 newf = dup_fd(oldf, &error);
1231 if (!newf)
1232 goto out;
1233
1234 tsk->files = newf;
1235 error = 0;
1236out:
1237 return error;
1238}
1239
fadad878 1240static int copy_io(unsigned long clone_flags, struct task_struct *tsk)
fd0928df
JA
1241{
1242#ifdef CONFIG_BLOCK
1243 struct io_context *ioc = current->io_context;
6e736be7 1244 struct io_context *new_ioc;
fd0928df
JA
1245
1246 if (!ioc)
1247 return 0;
fadad878
JA
1248 /*
1249 * Share io context with parent, if CLONE_IO is set
1250 */
1251 if (clone_flags & CLONE_IO) {
3d48749d
TH
1252 ioc_task_link(ioc);
1253 tsk->io_context = ioc;
fadad878 1254 } else if (ioprio_valid(ioc->ioprio)) {
6e736be7
TH
1255 new_ioc = get_task_io_context(tsk, GFP_KERNEL, NUMA_NO_NODE);
1256 if (unlikely(!new_ioc))
fd0928df
JA
1257 return -ENOMEM;
1258
6e736be7 1259 new_ioc->ioprio = ioc->ioprio;
11a3122f 1260 put_io_context(new_ioc);
fd0928df
JA
1261 }
1262#endif
1263 return 0;
1264}
1265
a39bc516 1266static int copy_sighand(unsigned long clone_flags, struct task_struct *tsk)
1da177e4
LT
1267{
1268 struct sighand_struct *sig;
1269
60348802 1270 if (clone_flags & CLONE_SIGHAND) {
1da177e4
LT
1271 atomic_inc(&current->sighand->count);
1272 return 0;
1273 }
1274 sig = kmem_cache_alloc(sighand_cachep, GFP_KERNEL);
e56d0903 1275 rcu_assign_pointer(tsk->sighand, sig);
1da177e4
LT
1276 if (!sig)
1277 return -ENOMEM;
9d7fb042 1278
1da177e4
LT
1279 atomic_set(&sig->count, 1);
1280 memcpy(sig->action, current->sighand->action, sizeof(sig->action));
1281 return 0;
1282}
1283
a7e5328a 1284void __cleanup_sighand(struct sighand_struct *sighand)
c81addc9 1285{
d80e731e
ON
1286 if (atomic_dec_and_test(&sighand->count)) {
1287 signalfd_cleanup(sighand);
392809b2
ON
1288 /*
1289 * sighand_cachep is SLAB_DESTROY_BY_RCU so we can free it
1290 * without an RCU grace period, see __lock_task_sighand().
1291 */
c81addc9 1292 kmem_cache_free(sighand_cachep, sighand);
d80e731e 1293 }
c81addc9
ON
1294}
1295
f06febc9
FM
1296/*
1297 * Initialize POSIX timer handling for a thread group.
1298 */
1299static void posix_cpu_timers_init_group(struct signal_struct *sig)
1300{
78d7d407
JS
1301 unsigned long cpu_limit;
1302
316c1608 1303 cpu_limit = READ_ONCE(sig->rlim[RLIMIT_CPU].rlim_cur);
78d7d407
JS
1304 if (cpu_limit != RLIM_INFINITY) {
1305 sig->cputime_expires.prof_exp = secs_to_cputime(cpu_limit);
d5c373eb 1306 sig->cputimer.running = true;
6279a751
ON
1307 }
1308
f06febc9
FM
1309 /* The timer lists. */
1310 INIT_LIST_HEAD(&sig->cpu_timers[0]);
1311 INIT_LIST_HEAD(&sig->cpu_timers[1]);
1312 INIT_LIST_HEAD(&sig->cpu_timers[2]);
1313}
1314
a39bc516 1315static int copy_signal(unsigned long clone_flags, struct task_struct *tsk)
1da177e4
LT
1316{
1317 struct signal_struct *sig;
1da177e4 1318
4ab6c083 1319 if (clone_flags & CLONE_THREAD)
490dea45 1320 return 0;
490dea45 1321
a56704ef 1322 sig = kmem_cache_zalloc(signal_cachep, GFP_KERNEL);
1da177e4
LT
1323 tsk->signal = sig;
1324 if (!sig)
1325 return -ENOMEM;
1326
b3ac022c 1327 sig->nr_threads = 1;
1da177e4 1328 atomic_set(&sig->live, 1);
b3ac022c 1329 atomic_set(&sig->sigcnt, 1);
0c740d0a
ON
1330
1331 /* list_add(thread_node, thread_head) without INIT_LIST_HEAD() */
1332 sig->thread_head = (struct list_head)LIST_HEAD_INIT(tsk->thread_node);
1333 tsk->thread_node = (struct list_head)LIST_HEAD_INIT(sig->thread_head);
1334
1da177e4 1335 init_waitqueue_head(&sig->wait_chldexit);
db51aecc 1336 sig->curr_target = tsk;
1da177e4
LT
1337 init_sigpending(&sig->shared_pending);
1338 INIT_LIST_HEAD(&sig->posix_timers);
e78c3496 1339 seqlock_init(&sig->stats_lock);
9d7fb042 1340 prev_cputime_init(&sig->prev_cputime);
1da177e4 1341
c9cb2e3d 1342 hrtimer_init(&sig->real_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
1da177e4 1343 sig->real_timer.function = it_real_fn;
1da177e4 1344
1da177e4
LT
1345 task_lock(current->group_leader);
1346 memcpy(sig->rlim, current->signal->rlim, sizeof sig->rlim);
1347 task_unlock(current->group_leader);
1348
6279a751
ON
1349 posix_cpu_timers_init_group(sig);
1350
522ed776 1351 tty_audit_fork(sig);
5091faa4 1352 sched_autogroup_fork(sig);
522ed776 1353
a63d83f4 1354 sig->oom_score_adj = current->signal->oom_score_adj;
dabb16f6 1355 sig->oom_score_adj_min = current->signal->oom_score_adj_min;
28b83c51 1356
ebec18a6
LP
1357 sig->has_child_subreaper = current->signal->has_child_subreaper ||
1358 current->signal->is_child_subreaper;
1359
9b1bf12d
KM
1360 mutex_init(&sig->cred_guard_mutex);
1361
1da177e4
LT
1362 return 0;
1363}
1364
dbd95212
KC
1365static void copy_seccomp(struct task_struct *p)
1366{
1367#ifdef CONFIG_SECCOMP
1368 /*
1369 * Must be called with sighand->lock held, which is common to
1370 * all threads in the group. Holding cred_guard_mutex is not
1371 * needed because this new task is not yet running and cannot
1372 * be racing exec.
1373 */
69f6a34b 1374 assert_spin_locked(&current->sighand->siglock);
dbd95212
KC
1375
1376 /* Ref-count the new filter user, and assign it. */
1377 get_seccomp_filter(current);
1378 p->seccomp = current->seccomp;
1379
1380 /*
1381 * Explicitly enable no_new_privs here in case it got set
1382 * between the task_struct being duplicated and holding the
1383 * sighand lock. The seccomp state and nnp must be in sync.
1384 */
1385 if (task_no_new_privs(current))
1386 task_set_no_new_privs(p);
1387
1388 /*
1389 * If the parent gained a seccomp mode after copying thread
1390 * flags and between before we held the sighand lock, we have
1391 * to manually enable the seccomp thread flag here.
1392 */
1393 if (p->seccomp.mode != SECCOMP_MODE_DISABLED)
1394 set_tsk_thread_flag(p, TIF_SECCOMP);
1395#endif
1396}
1397
17da2bd9 1398SYSCALL_DEFINE1(set_tid_address, int __user *, tidptr)
1da177e4
LT
1399{
1400 current->clear_child_tid = tidptr;
1401
b488893a 1402 return task_pid_vnr(current);
1da177e4
LT
1403}
1404
a39bc516 1405static void rt_mutex_init_task(struct task_struct *p)
23f78d4a 1406{
1d615482 1407 raw_spin_lock_init(&p->pi_lock);
e29e175b 1408#ifdef CONFIG_RT_MUTEXES
fb00aca4
PZ
1409 p->pi_waiters = RB_ROOT;
1410 p->pi_waiters_leftmost = NULL;
23f78d4a 1411 p->pi_blocked_on = NULL;
23f78d4a
IM
1412#endif
1413}
1414
f06febc9
FM
1415/*
1416 * Initialize POSIX timer handling for a single task.
1417 */
1418static void posix_cpu_timers_init(struct task_struct *tsk)
1419{
64861634
MS
1420 tsk->cputime_expires.prof_exp = 0;
1421 tsk->cputime_expires.virt_exp = 0;
f06febc9
FM
1422 tsk->cputime_expires.sched_exp = 0;
1423 INIT_LIST_HEAD(&tsk->cpu_timers[0]);
1424 INIT_LIST_HEAD(&tsk->cpu_timers[1]);
1425 INIT_LIST_HEAD(&tsk->cpu_timers[2]);
1426}
1427
81907739
ON
1428static inline void
1429init_task_pid(struct task_struct *task, enum pid_type type, struct pid *pid)
1430{
1431 task->pids[type].pid = pid;
1432}
1433
1da177e4
LT
1434/*
1435 * This creates a new process as a copy of the old one,
1436 * but does not actually start it yet.
1437 *
1438 * It copies the registers, and all the appropriate
1439 * parts of the process environment (as per the clone
1440 * flags). The actual kick-off is left to the caller.
1441 */
36c8b586
IM
1442static struct task_struct *copy_process(unsigned long clone_flags,
1443 unsigned long stack_start,
36c8b586 1444 unsigned long stack_size,
36c8b586 1445 int __user *child_tidptr,
09a05394 1446 struct pid *pid,
3033f14a 1447 int trace,
725fc629
AK
1448 unsigned long tls,
1449 int node)
1da177e4
LT
1450{
1451 int retval;
a24efe62 1452 struct task_struct *p;
1da177e4
LT
1453
1454 if ((clone_flags & (CLONE_NEWNS|CLONE_FS)) == (CLONE_NEWNS|CLONE_FS))
1455 return ERR_PTR(-EINVAL);
1456
e66eded8
EB
1457 if ((clone_flags & (CLONE_NEWUSER|CLONE_FS)) == (CLONE_NEWUSER|CLONE_FS))
1458 return ERR_PTR(-EINVAL);
1459
1da177e4
LT
1460 /*
1461 * Thread groups must share signals as well, and detached threads
1462 * can only be started up within the thread group.
1463 */
1464 if ((clone_flags & CLONE_THREAD) && !(clone_flags & CLONE_SIGHAND))
1465 return ERR_PTR(-EINVAL);
1466
1467 /*
1468 * Shared signal handlers imply shared VM. By way of the above,
1469 * thread groups also imply shared VM. Blocking this case allows
1470 * for various simplifications in other code.
1471 */
1472 if ((clone_flags & CLONE_SIGHAND) && !(clone_flags & CLONE_VM))
1473 return ERR_PTR(-EINVAL);
1474
123be07b
SB
1475 /*
1476 * Siblings of global init remain as zombies on exit since they are
1477 * not reaped by their parent (swapper). To solve this and to avoid
1478 * multi-rooted process trees, prevent global and container-inits
1479 * from creating siblings.
1480 */
1481 if ((clone_flags & CLONE_PARENT) &&
1482 current->signal->flags & SIGNAL_UNKILLABLE)
1483 return ERR_PTR(-EINVAL);
1484
8382fcac 1485 /*
40a0d32d 1486 * If the new process will be in a different pid or user namespace
faf00da5 1487 * do not allow it to share a thread group with the forking task.
8382fcac 1488 */
faf00da5 1489 if (clone_flags & CLONE_THREAD) {
40a0d32d
ON
1490 if ((clone_flags & (CLONE_NEWUSER | CLONE_NEWPID)) ||
1491 (task_active_pid_ns(current) !=
1492 current->nsproxy->pid_ns_for_children))
1493 return ERR_PTR(-EINVAL);
1494 }
8382fcac 1495
1da177e4
LT
1496 retval = security_task_create(clone_flags);
1497 if (retval)
1498 goto fork_out;
1499
1500 retval = -ENOMEM;
725fc629 1501 p = dup_task_struct(current, node);
1da177e4
LT
1502 if (!p)
1503 goto fork_out;
1504
f7e8b616
SR
1505 ftrace_graph_init_task(p);
1506
bea493a0
PZ
1507 rt_mutex_init_task(p);
1508
d12c1a37 1509#ifdef CONFIG_PROVE_LOCKING
de30a2b3
IM
1510 DEBUG_LOCKS_WARN_ON(!p->hardirqs_enabled);
1511 DEBUG_LOCKS_WARN_ON(!p->softirqs_enabled);
1512#endif
1da177e4 1513 retval = -EAGAIN;
3b11a1de 1514 if (atomic_read(&p->real_cred->user->processes) >=
78d7d407 1515 task_rlimit(p, RLIMIT_NPROC)) {
b57922b6
EP
1516 if (p->real_cred->user != INIT_USER &&
1517 !capable(CAP_SYS_RESOURCE) && !capable(CAP_SYS_ADMIN))
1da177e4
LT
1518 goto bad_fork_free;
1519 }
72fa5997 1520 current->flags &= ~PF_NPROC_EXCEEDED;
1da177e4 1521
f1752eec
DH
1522 retval = copy_creds(p, clone_flags);
1523 if (retval < 0)
1524 goto bad_fork_free;
1da177e4
LT
1525
1526 /*
1527 * If multiple threads are within copy_process(), then this check
1528 * triggers too late. This doesn't hurt, the check is only there
1529 * to stop root fork bombs.
1530 */
04ec93fe 1531 retval = -EAGAIN;
1da177e4
LT
1532 if (nr_threads >= max_threads)
1533 goto bad_fork_cleanup_count;
1534
ca74e92b 1535 delayacct_tsk_init(p); /* Must remain after dup_task_struct() */
514ddb44
DR
1536 p->flags &= ~(PF_SUPERPRIV | PF_WQ_WORKER);
1537 p->flags |= PF_FORKNOEXEC;
1da177e4
LT
1538 INIT_LIST_HEAD(&p->children);
1539 INIT_LIST_HEAD(&p->sibling);
f41d911f 1540 rcu_copy_process(p);
1da177e4
LT
1541 p->vfork_done = NULL;
1542 spin_lock_init(&p->alloc_lock);
1da177e4 1543
1da177e4
LT
1544 init_sigpending(&p->pending);
1545
64861634
MS
1546 p->utime = p->stime = p->gtime = 0;
1547 p->utimescaled = p->stimescaled = 0;
9d7fb042
PZ
1548 prev_cputime_init(&p->prev_cputime);
1549
6a61671b 1550#ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
b7ce2277 1551 seqcount_init(&p->vtime_seqcount);
6a61671b 1552 p->vtime_snap = 0;
7098c1ea 1553 p->vtime_snap_whence = VTIME_INACTIVE;
6a61671b
FW
1554#endif
1555
a3a2e76c
KH
1556#if defined(SPLIT_RSS_COUNTING)
1557 memset(&p->rss_stat, 0, sizeof(p->rss_stat));
1558#endif
172ba844 1559
6976675d
AV
1560 p->default_timer_slack_ns = current->timer_slack_ns;
1561
5995477a 1562 task_io_accounting_init(&p->ioac);
1da177e4
LT
1563 acct_clear_integrals(p);
1564
f06febc9 1565 posix_cpu_timers_init(p);
1da177e4 1566
ccbf62d8 1567 p->start_time = ktime_get_ns();
57e0be04 1568 p->real_start_time = ktime_get_boot_ns();
1da177e4 1569 p->io_context = NULL;
1da177e4 1570 p->audit_context = NULL;
b4f48b63 1571 cgroup_fork(p);
1da177e4 1572#ifdef CONFIG_NUMA
846a16bf 1573 p->mempolicy = mpol_dup(p->mempolicy);
fb0a685c
DRO
1574 if (IS_ERR(p->mempolicy)) {
1575 retval = PTR_ERR(p->mempolicy);
1576 p->mempolicy = NULL;
e8604cb4 1577 goto bad_fork_cleanup_threadgroup_lock;
fb0a685c 1578 }
1da177e4 1579#endif
778d3b0f
MH
1580#ifdef CONFIG_CPUSETS
1581 p->cpuset_mem_spread_rotor = NUMA_NO_NODE;
1582 p->cpuset_slab_spread_rotor = NUMA_NO_NODE;
cc9a6c87 1583 seqcount_init(&p->mems_allowed_seq);
778d3b0f 1584#endif
de30a2b3
IM
1585#ifdef CONFIG_TRACE_IRQFLAGS
1586 p->irq_events = 0;
1587 p->hardirqs_enabled = 0;
1588 p->hardirq_enable_ip = 0;
1589 p->hardirq_enable_event = 0;
1590 p->hardirq_disable_ip = _THIS_IP_;
1591 p->hardirq_disable_event = 0;
1592 p->softirqs_enabled = 1;
1593 p->softirq_enable_ip = _THIS_IP_;
1594 p->softirq_enable_event = 0;
1595 p->softirq_disable_ip = 0;
1596 p->softirq_disable_event = 0;
1597 p->hardirq_context = 0;
1598 p->softirq_context = 0;
1599#endif
8bcbde54
DH
1600
1601 p->pagefault_disabled = 0;
1602
fbb9ce95
IM
1603#ifdef CONFIG_LOCKDEP
1604 p->lockdep_depth = 0; /* no locks held yet */
1605 p->curr_chain_key = 0;
1606 p->lockdep_recursion = 0;
1607#endif
1da177e4 1608
408894ee
IM
1609#ifdef CONFIG_DEBUG_MUTEXES
1610 p->blocked_on = NULL; /* not blocked yet */
1611#endif
cafe5635
KO
1612#ifdef CONFIG_BCACHE
1613 p->sequential_io = 0;
1614 p->sequential_io_avg = 0;
1615#endif
0f481406 1616
3c90e6e9 1617 /* Perform scheduler related setup. Assign this task to a CPU. */
aab03e05
DF
1618 retval = sched_fork(clone_flags, p);
1619 if (retval)
1620 goto bad_fork_cleanup_policy;
6ab423e0 1621
cdd6c482 1622 retval = perf_event_init_task(p);
6ab423e0
PZ
1623 if (retval)
1624 goto bad_fork_cleanup_policy;
fb0a685c
DRO
1625 retval = audit_alloc(p);
1626 if (retval)
6c72e350 1627 goto bad_fork_cleanup_perf;
1da177e4 1628 /* copy all the process information */
ab602f79 1629 shm_init_task(p);
fb0a685c
DRO
1630 retval = copy_semundo(clone_flags, p);
1631 if (retval)
1da177e4 1632 goto bad_fork_cleanup_audit;
fb0a685c
DRO
1633 retval = copy_files(clone_flags, p);
1634 if (retval)
1da177e4 1635 goto bad_fork_cleanup_semundo;
fb0a685c
DRO
1636 retval = copy_fs(clone_flags, p);
1637 if (retval)
1da177e4 1638 goto bad_fork_cleanup_files;
fb0a685c
DRO
1639 retval = copy_sighand(clone_flags, p);
1640 if (retval)
1da177e4 1641 goto bad_fork_cleanup_fs;
fb0a685c
DRO
1642 retval = copy_signal(clone_flags, p);
1643 if (retval)
1da177e4 1644 goto bad_fork_cleanup_sighand;
fb0a685c
DRO
1645 retval = copy_mm(clone_flags, p);
1646 if (retval)
1da177e4 1647 goto bad_fork_cleanup_signal;
fb0a685c
DRO
1648 retval = copy_namespaces(clone_flags, p);
1649 if (retval)
d84f4f99 1650 goto bad_fork_cleanup_mm;
fb0a685c
DRO
1651 retval = copy_io(clone_flags, p);
1652 if (retval)
fd0928df 1653 goto bad_fork_cleanup_namespaces;
3033f14a 1654 retval = copy_thread_tls(clone_flags, stack_start, stack_size, p, tls);
1da177e4 1655 if (retval)
fd0928df 1656 goto bad_fork_cleanup_io;
1da177e4 1657
425fb2b4 1658 if (pid != &init_struct_pid) {
c2b1df2e 1659 pid = alloc_pid(p->nsproxy->pid_ns_for_children);
35f71bc0
MH
1660 if (IS_ERR(pid)) {
1661 retval = PTR_ERR(pid);
0740aa5f 1662 goto bad_fork_cleanup_thread;
35f71bc0 1663 }
425fb2b4
PE
1664 }
1665
1da177e4
LT
1666 p->set_child_tid = (clone_flags & CLONE_CHILD_SETTID) ? child_tidptr : NULL;
1667 /*
1668 * Clear TID on mm_release()?
1669 */
fb0a685c 1670 p->clear_child_tid = (clone_flags & CLONE_CHILD_CLEARTID) ? child_tidptr : NULL;
73c10101
JA
1671#ifdef CONFIG_BLOCK
1672 p->plug = NULL;
1673#endif
42b2dd0a 1674#ifdef CONFIG_FUTEX
8f17d3a5
IM
1675 p->robust_list = NULL;
1676#ifdef CONFIG_COMPAT
1677 p->compat_robust_list = NULL;
1678#endif
c87e2837
IM
1679 INIT_LIST_HEAD(&p->pi_state_list);
1680 p->pi_state_cache = NULL;
42b2dd0a 1681#endif
f9a3879a
GM
1682 /*
1683 * sigaltstack should be cleared when sharing the same VM
1684 */
1685 if ((clone_flags & (CLONE_VM|CLONE_VFORK)) == CLONE_VM)
2a742138 1686 sas_ss_reset(p);
f9a3879a 1687
1da177e4 1688 /*
6580807d
ON
1689 * Syscall tracing and stepping should be turned off in the
1690 * child regardless of CLONE_PTRACE.
1da177e4 1691 */
6580807d 1692 user_disable_single_step(p);
1da177e4 1693 clear_tsk_thread_flag(p, TIF_SYSCALL_TRACE);
ed75e8d5
LV
1694#ifdef TIF_SYSCALL_EMU
1695 clear_tsk_thread_flag(p, TIF_SYSCALL_EMU);
1696#endif
9745512c 1697 clear_all_latency_tracing(p);
1da177e4 1698
1da177e4 1699 /* ok, now we should be set up.. */
18c830df
ON
1700 p->pid = pid_nr(pid);
1701 if (clone_flags & CLONE_THREAD) {
5f8aadd8 1702 p->exit_signal = -1;
18c830df
ON
1703 p->group_leader = current->group_leader;
1704 p->tgid = current->tgid;
1705 } else {
1706 if (clone_flags & CLONE_PARENT)
1707 p->exit_signal = current->group_leader->exit_signal;
1708 else
1709 p->exit_signal = (clone_flags & CSIGNAL);
1710 p->group_leader = p;
1711 p->tgid = p->pid;
1712 }
5f8aadd8 1713
9d823e8f
WF
1714 p->nr_dirtied = 0;
1715 p->nr_dirtied_pause = 128 >> (PAGE_SHIFT - 10);
83712358 1716 p->dirty_paused_when = 0;
9d823e8f 1717
bb8cbbfe 1718 p->pdeath_signal = 0;
47e65328 1719 INIT_LIST_HEAD(&p->thread_group);
158e1645 1720 p->task_works = NULL;
1da177e4 1721
568ac888 1722 threadgroup_change_begin(current);
7e47682e
AS
1723 /*
1724 * Ensure that the cgroup subsystem policies allow the new process to be
1725 * forked. It should be noted the the new process's css_set can be changed
1726 * between here and cgroup_post_fork() if an organisation operation is in
1727 * progress.
1728 */
b53202e6 1729 retval = cgroup_can_fork(p);
7e47682e
AS
1730 if (retval)
1731 goto bad_fork_free_pid;
1732
18c830df
ON
1733 /*
1734 * Make it visible to the rest of the system, but dont wake it up yet.
1735 * Need tasklist lock for parent etc handling!
1736 */
1da177e4
LT
1737 write_lock_irq(&tasklist_lock);
1738
1da177e4 1739 /* CLONE_PARENT re-uses the old parent */
2d5516cb 1740 if (clone_flags & (CLONE_PARENT|CLONE_THREAD)) {
1da177e4 1741 p->real_parent = current->real_parent;
2d5516cb
ON
1742 p->parent_exec_id = current->parent_exec_id;
1743 } else {
1da177e4 1744 p->real_parent = current;
2d5516cb
ON
1745 p->parent_exec_id = current->self_exec_id;
1746 }
1da177e4 1747
3f17da69 1748 spin_lock(&current->sighand->siglock);
4a2c7a78 1749
dbd95212
KC
1750 /*
1751 * Copy seccomp details explicitly here, in case they were changed
1752 * before holding sighand lock.
1753 */
1754 copy_seccomp(p);
1755
4a2c7a78
ON
1756 /*
1757 * Process group and session signals need to be delivered to just the
1758 * parent before the fork or both the parent and the child after the
1759 * fork. Restart if a signal comes in before we add the new process to
1760 * it's process group.
1761 * A fatal signal pending means that current will exit, so the new
1762 * thread can't slip out of an OOM kill (or normal SIGKILL).
fb0a685c 1763 */
23ff4440 1764 recalc_sigpending();
4a2c7a78
ON
1765 if (signal_pending(current)) {
1766 spin_unlock(&current->sighand->siglock);
1767 write_unlock_irq(&tasklist_lock);
1768 retval = -ERESTARTNOINTR;
7e47682e 1769 goto bad_fork_cancel_cgroup;
4a2c7a78
ON
1770 }
1771
73b9ebfe 1772 if (likely(p->pid)) {
4b9d33e6 1773 ptrace_init_task(p, (clone_flags & CLONE_PTRACE) || trace);
73b9ebfe 1774
81907739 1775 init_task_pid(p, PIDTYPE_PID, pid);
73b9ebfe 1776 if (thread_group_leader(p)) {
81907739
ON
1777 init_task_pid(p, PIDTYPE_PGID, task_pgrp(current));
1778 init_task_pid(p, PIDTYPE_SID, task_session(current));
1779
1c4042c2 1780 if (is_child_reaper(pid)) {
17cf22c3 1781 ns_of_pid(pid)->child_reaper = p;
1c4042c2
EB
1782 p->signal->flags |= SIGNAL_UNKILLABLE;
1783 }
73b9ebfe 1784
fea9d175 1785 p->signal->leader_pid = pid;
9c9f4ded 1786 p->signal->tty = tty_kref_get(current->signal->tty);
9cd80bbb 1787 list_add_tail(&p->sibling, &p->real_parent->children);
5e85d4ab 1788 list_add_tail_rcu(&p->tasks, &init_task.tasks);
81907739
ON
1789 attach_pid(p, PIDTYPE_PGID);
1790 attach_pid(p, PIDTYPE_SID);
909ea964 1791 __this_cpu_inc(process_counts);
80628ca0
ON
1792 } else {
1793 current->signal->nr_threads++;
1794 atomic_inc(&current->signal->live);
1795 atomic_inc(&current->signal->sigcnt);
80628ca0
ON
1796 list_add_tail_rcu(&p->thread_group,
1797 &p->group_leader->thread_group);
0c740d0a
ON
1798 list_add_tail_rcu(&p->thread_node,
1799 &p->signal->thread_head);
73b9ebfe 1800 }
81907739 1801 attach_pid(p, PIDTYPE_PID);
73b9ebfe 1802 nr_threads++;
1da177e4
LT
1803 }
1804
1da177e4 1805 total_forks++;
3f17da69 1806 spin_unlock(&current->sighand->siglock);
4af4206b 1807 syscall_tracepoint_update(p);
1da177e4 1808 write_unlock_irq(&tasklist_lock);
4af4206b 1809
c13cf856 1810 proc_fork_connector(p);
b53202e6 1811 cgroup_post_fork(p);
c9e75f04 1812 threadgroup_change_end(current);
cdd6c482 1813 perf_event_fork(p);
43d2b113
KH
1814
1815 trace_task_newtask(p, clone_flags);
3ab67966 1816 uprobe_copy_process(p, clone_flags);
43d2b113 1817
1da177e4
LT
1818 return p;
1819
7e47682e 1820bad_fork_cancel_cgroup:
b53202e6 1821 cgroup_cancel_fork(p);
425fb2b4 1822bad_fork_free_pid:
568ac888 1823 threadgroup_change_end(current);
425fb2b4
PE
1824 if (pid != &init_struct_pid)
1825 free_pid(pid);
0740aa5f
JS
1826bad_fork_cleanup_thread:
1827 exit_thread(p);
fd0928df 1828bad_fork_cleanup_io:
b69f2292
LR
1829 if (p->io_context)
1830 exit_io_context(p);
ab516013 1831bad_fork_cleanup_namespaces:
444f378b 1832 exit_task_namespaces(p);
1da177e4 1833bad_fork_cleanup_mm:
c9f01245 1834 if (p->mm)
1da177e4
LT
1835 mmput(p->mm);
1836bad_fork_cleanup_signal:
4ab6c083 1837 if (!(clone_flags & CLONE_THREAD))
1c5354de 1838 free_signal_struct(p->signal);
1da177e4 1839bad_fork_cleanup_sighand:
a7e5328a 1840 __cleanup_sighand(p->sighand);
1da177e4
LT
1841bad_fork_cleanup_fs:
1842 exit_fs(p); /* blocking */
1843bad_fork_cleanup_files:
1844 exit_files(p); /* blocking */
1845bad_fork_cleanup_semundo:
1846 exit_sem(p);
1847bad_fork_cleanup_audit:
1848 audit_free(p);
6c72e350 1849bad_fork_cleanup_perf:
cdd6c482 1850 perf_event_free_task(p);
6c72e350 1851bad_fork_cleanup_policy:
1da177e4 1852#ifdef CONFIG_NUMA
f0be3d32 1853 mpol_put(p->mempolicy);
e8604cb4 1854bad_fork_cleanup_threadgroup_lock:
1da177e4 1855#endif
35df17c5 1856 delayacct_tsk_free(p);
1da177e4 1857bad_fork_cleanup_count:
d84f4f99 1858 atomic_dec(&p->cred->user->processes);
e0e81739 1859 exit_creds(p);
1da177e4 1860bad_fork_free:
68f24b08 1861 put_task_stack(p);
1da177e4 1862 free_task(p);
fe7d37d1
ON
1863fork_out:
1864 return ERR_PTR(retval);
1da177e4
LT
1865}
1866
f106eee1
ON
1867static inline void init_idle_pids(struct pid_link *links)
1868{
1869 enum pid_type type;
1870
1871 for (type = PIDTYPE_PID; type < PIDTYPE_MAX; ++type) {
1872 INIT_HLIST_NODE(&links[type].node); /* not really needed */
1873 links[type].pid = &init_struct_pid;
1874 }
1875}
1876
0db0628d 1877struct task_struct *fork_idle(int cpu)
1da177e4 1878{
36c8b586 1879 struct task_struct *task;
725fc629
AK
1880 task = copy_process(CLONE_VM, 0, 0, NULL, &init_struct_pid, 0, 0,
1881 cpu_to_node(cpu));
f106eee1
ON
1882 if (!IS_ERR(task)) {
1883 init_idle_pids(task->pids);
753ca4f3 1884 init_idle(task, cpu);
f106eee1 1885 }
73b9ebfe 1886
1da177e4
LT
1887 return task;
1888}
1889
1da177e4
LT
1890/*
1891 * Ok, this is the main fork-routine.
1892 *
1893 * It copies the process, and if successful kick-starts
1894 * it and waits for it to finish using the VM if required.
1895 */
3033f14a 1896long _do_fork(unsigned long clone_flags,
1da177e4 1897 unsigned long stack_start,
1da177e4
LT
1898 unsigned long stack_size,
1899 int __user *parent_tidptr,
3033f14a
JT
1900 int __user *child_tidptr,
1901 unsigned long tls)
1da177e4
LT
1902{
1903 struct task_struct *p;
1904 int trace = 0;
92476d7f 1905 long nr;
1da177e4 1906
09a05394 1907 /*
4b9d33e6
TH
1908 * Determine whether and which event to report to ptracer. When
1909 * called from kernel_thread or CLONE_UNTRACED is explicitly
1910 * requested, no event is reported; otherwise, report if the event
1911 * for the type of forking is enabled.
09a05394 1912 */
e80d6661 1913 if (!(clone_flags & CLONE_UNTRACED)) {
4b9d33e6
TH
1914 if (clone_flags & CLONE_VFORK)
1915 trace = PTRACE_EVENT_VFORK;
1916 else if ((clone_flags & CSIGNAL) != SIGCHLD)
1917 trace = PTRACE_EVENT_CLONE;
1918 else
1919 trace = PTRACE_EVENT_FORK;
1920
1921 if (likely(!ptrace_event_enabled(current, trace)))
1922 trace = 0;
1923 }
1da177e4 1924
62e791c1 1925 p = copy_process(clone_flags, stack_start, stack_size,
725fc629 1926 child_tidptr, NULL, trace, tls, NUMA_NO_NODE);
1da177e4
LT
1927 /*
1928 * Do this prior waking up the new thread - the thread pointer
1929 * might get invalid after that point, if the thread exits quickly.
1930 */
1931 if (!IS_ERR(p)) {
1932 struct completion vfork;
4e52365f 1933 struct pid *pid;
1da177e4 1934
0a16b607
MD
1935 trace_sched_process_fork(current, p);
1936
4e52365f
MD
1937 pid = get_task_pid(p, PIDTYPE_PID);
1938 nr = pid_vnr(pid);
30e49c26
PE
1939
1940 if (clone_flags & CLONE_PARENT_SETTID)
1941 put_user(nr, parent_tidptr);
a6f5e063 1942
1da177e4
LT
1943 if (clone_flags & CLONE_VFORK) {
1944 p->vfork_done = &vfork;
1945 init_completion(&vfork);
d68b46fe 1946 get_task_struct(p);
1da177e4
LT
1947 }
1948
3e51e3ed 1949 wake_up_new_task(p);
1da177e4 1950
4b9d33e6
TH
1951 /* forking complete and child started to run, tell ptracer */
1952 if (unlikely(trace))
4e52365f 1953 ptrace_event_pid(trace, pid);
09a05394 1954
1da177e4 1955 if (clone_flags & CLONE_VFORK) {
d68b46fe 1956 if (!wait_for_vfork_done(p, &vfork))
4e52365f 1957 ptrace_event_pid(PTRACE_EVENT_VFORK_DONE, pid);
1da177e4 1958 }
4e52365f
MD
1959
1960 put_pid(pid);
1da177e4 1961 } else {
92476d7f 1962 nr = PTR_ERR(p);
1da177e4 1963 }
92476d7f 1964 return nr;
1da177e4
LT
1965}
1966
3033f14a
JT
1967#ifndef CONFIG_HAVE_COPY_THREAD_TLS
1968/* For compatibility with architectures that call do_fork directly rather than
1969 * using the syscall entry points below. */
1970long do_fork(unsigned long clone_flags,
1971 unsigned long stack_start,
1972 unsigned long stack_size,
1973 int __user *parent_tidptr,
1974 int __user *child_tidptr)
1975{
1976 return _do_fork(clone_flags, stack_start, stack_size,
1977 parent_tidptr, child_tidptr, 0);
1978}
1979#endif
1980
2aa3a7f8
AV
1981/*
1982 * Create a kernel thread.
1983 */
1984pid_t kernel_thread(int (*fn)(void *), void *arg, unsigned long flags)
1985{
3033f14a
JT
1986 return _do_fork(flags|CLONE_VM|CLONE_UNTRACED, (unsigned long)fn,
1987 (unsigned long)arg, NULL, NULL, 0);
2aa3a7f8 1988}
2aa3a7f8 1989
d2125043
AV
1990#ifdef __ARCH_WANT_SYS_FORK
1991SYSCALL_DEFINE0(fork)
1992{
1993#ifdef CONFIG_MMU
3033f14a 1994 return _do_fork(SIGCHLD, 0, 0, NULL, NULL, 0);
d2125043
AV
1995#else
1996 /* can not support in nommu mode */
5d59e182 1997 return -EINVAL;
d2125043
AV
1998#endif
1999}
2000#endif
2001
2002#ifdef __ARCH_WANT_SYS_VFORK
2003SYSCALL_DEFINE0(vfork)
2004{
3033f14a
JT
2005 return _do_fork(CLONE_VFORK | CLONE_VM | SIGCHLD, 0,
2006 0, NULL, NULL, 0);
d2125043
AV
2007}
2008#endif
2009
2010#ifdef __ARCH_WANT_SYS_CLONE
2011#ifdef CONFIG_CLONE_BACKWARDS
2012SYSCALL_DEFINE5(clone, unsigned long, clone_flags, unsigned long, newsp,
2013 int __user *, parent_tidptr,
3033f14a 2014 unsigned long, tls,
d2125043
AV
2015 int __user *, child_tidptr)
2016#elif defined(CONFIG_CLONE_BACKWARDS2)
2017SYSCALL_DEFINE5(clone, unsigned long, newsp, unsigned long, clone_flags,
2018 int __user *, parent_tidptr,
2019 int __user *, child_tidptr,
3033f14a 2020 unsigned long, tls)
dfa9771a
MS
2021#elif defined(CONFIG_CLONE_BACKWARDS3)
2022SYSCALL_DEFINE6(clone, unsigned long, clone_flags, unsigned long, newsp,
2023 int, stack_size,
2024 int __user *, parent_tidptr,
2025 int __user *, child_tidptr,
3033f14a 2026 unsigned long, tls)
d2125043
AV
2027#else
2028SYSCALL_DEFINE5(clone, unsigned long, clone_flags, unsigned long, newsp,
2029 int __user *, parent_tidptr,
2030 int __user *, child_tidptr,
3033f14a 2031 unsigned long, tls)
d2125043
AV
2032#endif
2033{
3033f14a 2034 return _do_fork(clone_flags, newsp, 0, parent_tidptr, child_tidptr, tls);
d2125043
AV
2035}
2036#endif
2037
5fd63b30
RT
2038#ifndef ARCH_MIN_MMSTRUCT_ALIGN
2039#define ARCH_MIN_MMSTRUCT_ALIGN 0
2040#endif
2041
51cc5068 2042static void sighand_ctor(void *data)
aa1757f9
ON
2043{
2044 struct sighand_struct *sighand = data;
2045
a35afb83 2046 spin_lock_init(&sighand->siglock);
b8fceee1 2047 init_waitqueue_head(&sighand->signalfd_wqh);
aa1757f9
ON
2048}
2049
1da177e4
LT
2050void __init proc_caches_init(void)
2051{
2052 sighand_cachep = kmem_cache_create("sighand_cache",
2053 sizeof(struct sighand_struct), 0,
2dff4405 2054 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_DESTROY_BY_RCU|
5d097056 2055 SLAB_NOTRACK|SLAB_ACCOUNT, sighand_ctor);
1da177e4
LT
2056 signal_cachep = kmem_cache_create("signal_cache",
2057 sizeof(struct signal_struct), 0,
5d097056
VD
2058 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK|SLAB_ACCOUNT,
2059 NULL);
20c2df83 2060 files_cachep = kmem_cache_create("files_cache",
1da177e4 2061 sizeof(struct files_struct), 0,
5d097056
VD
2062 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK|SLAB_ACCOUNT,
2063 NULL);
20c2df83 2064 fs_cachep = kmem_cache_create("fs_cache",
1da177e4 2065 sizeof(struct fs_struct), 0,
5d097056
VD
2066 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK|SLAB_ACCOUNT,
2067 NULL);
6345d24d
LT
2068 /*
2069 * FIXME! The "sizeof(struct mm_struct)" currently includes the
2070 * whole struct cpumask for the OFFSTACK case. We could change
2071 * this to *only* allocate as much of it as required by the
2072 * maximum number of CPU's we can ever have. The cpumask_allocation
2073 * is at the end of the structure, exactly for that reason.
2074 */
1da177e4 2075 mm_cachep = kmem_cache_create("mm_struct",
5fd63b30 2076 sizeof(struct mm_struct), ARCH_MIN_MMSTRUCT_ALIGN,
5d097056
VD
2077 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK|SLAB_ACCOUNT,
2078 NULL);
2079 vm_area_cachep = KMEM_CACHE(vm_area_struct, SLAB_PANIC|SLAB_ACCOUNT);
8feae131 2080 mmap_init();
66577193 2081 nsproxy_cache_init();
1da177e4 2082}
cf2e340f 2083
cf2e340f 2084/*
9bfb23fc 2085 * Check constraints on flags passed to the unshare system call.
cf2e340f 2086 */
9bfb23fc 2087static int check_unshare_flags(unsigned long unshare_flags)
cf2e340f 2088{
9bfb23fc
ON
2089 if (unshare_flags & ~(CLONE_THREAD|CLONE_FS|CLONE_NEWNS|CLONE_SIGHAND|
2090 CLONE_VM|CLONE_FILES|CLONE_SYSVSEM|
50804fe3 2091 CLONE_NEWUTS|CLONE_NEWIPC|CLONE_NEWNET|
a79a908f 2092 CLONE_NEWUSER|CLONE_NEWPID|CLONE_NEWCGROUP))
9bfb23fc 2093 return -EINVAL;
cf2e340f 2094 /*
12c641ab
EB
2095 * Not implemented, but pretend it works if there is nothing
2096 * to unshare. Note that unsharing the address space or the
2097 * signal handlers also need to unshare the signal queues (aka
2098 * CLONE_THREAD).
cf2e340f 2099 */
9bfb23fc 2100 if (unshare_flags & (CLONE_THREAD | CLONE_SIGHAND | CLONE_VM)) {
12c641ab
EB
2101 if (!thread_group_empty(current))
2102 return -EINVAL;
2103 }
2104 if (unshare_flags & (CLONE_SIGHAND | CLONE_VM)) {
2105 if (atomic_read(&current->sighand->count) > 1)
2106 return -EINVAL;
2107 }
2108 if (unshare_flags & CLONE_VM) {
2109 if (!current_is_single_threaded())
9bfb23fc
ON
2110 return -EINVAL;
2111 }
cf2e340f
JD
2112
2113 return 0;
2114}
2115
2116/*
99d1419d 2117 * Unshare the filesystem structure if it is being shared
cf2e340f
JD
2118 */
2119static int unshare_fs(unsigned long unshare_flags, struct fs_struct **new_fsp)
2120{
2121 struct fs_struct *fs = current->fs;
2122
498052bb
AV
2123 if (!(unshare_flags & CLONE_FS) || !fs)
2124 return 0;
2125
2126 /* don't need lock here; in the worst case we'll do useless copy */
2127 if (fs->users == 1)
2128 return 0;
2129
2130 *new_fsp = copy_fs_struct(fs);
2131 if (!*new_fsp)
2132 return -ENOMEM;
cf2e340f
JD
2133
2134 return 0;
2135}
2136
cf2e340f 2137/*
a016f338 2138 * Unshare file descriptor table if it is being shared
cf2e340f
JD
2139 */
2140static int unshare_fd(unsigned long unshare_flags, struct files_struct **new_fdp)
2141{
2142 struct files_struct *fd = current->files;
a016f338 2143 int error = 0;
cf2e340f
JD
2144
2145 if ((unshare_flags & CLONE_FILES) &&
a016f338
JD
2146 (fd && atomic_read(&fd->count) > 1)) {
2147 *new_fdp = dup_fd(fd, &error);
2148 if (!*new_fdp)
2149 return error;
2150 }
cf2e340f
JD
2151
2152 return 0;
2153}
2154
cf2e340f
JD
2155/*
2156 * unshare allows a process to 'unshare' part of the process
2157 * context which was originally shared using clone. copy_*
2158 * functions used by do_fork() cannot be used here directly
2159 * because they modify an inactive task_struct that is being
2160 * constructed. Here we are modifying the current, active,
2161 * task_struct.
2162 */
6559eed8 2163SYSCALL_DEFINE1(unshare, unsigned long, unshare_flags)
cf2e340f 2164{
cf2e340f 2165 struct fs_struct *fs, *new_fs = NULL;
cf2e340f 2166 struct files_struct *fd, *new_fd = NULL;
b2e0d987 2167 struct cred *new_cred = NULL;
cf7b708c 2168 struct nsproxy *new_nsproxy = NULL;
9edff4ab 2169 int do_sysvsem = 0;
9bfb23fc 2170 int err;
cf2e340f 2171
b2e0d987 2172 /*
faf00da5
EB
2173 * If unsharing a user namespace must also unshare the thread group
2174 * and unshare the filesystem root and working directories.
b2e0d987
EB
2175 */
2176 if (unshare_flags & CLONE_NEWUSER)
e66eded8 2177 unshare_flags |= CLONE_THREAD | CLONE_FS;
50804fe3
EB
2178 /*
2179 * If unsharing vm, must also unshare signal handlers.
2180 */
2181 if (unshare_flags & CLONE_VM)
2182 unshare_flags |= CLONE_SIGHAND;
12c641ab
EB
2183 /*
2184 * If unsharing a signal handlers, must also unshare the signal queues.
2185 */
2186 if (unshare_flags & CLONE_SIGHAND)
2187 unshare_flags |= CLONE_THREAD;
9bfb23fc
ON
2188 /*
2189 * If unsharing namespace, must also unshare filesystem information.
2190 */
2191 if (unshare_flags & CLONE_NEWNS)
2192 unshare_flags |= CLONE_FS;
50804fe3
EB
2193
2194 err = check_unshare_flags(unshare_flags);
2195 if (err)
2196 goto bad_unshare_out;
6013f67f
MS
2197 /*
2198 * CLONE_NEWIPC must also detach from the undolist: after switching
2199 * to a new ipc namespace, the semaphore arrays from the old
2200 * namespace are unreachable.
2201 */
2202 if (unshare_flags & (CLONE_NEWIPC|CLONE_SYSVSEM))
9edff4ab 2203 do_sysvsem = 1;
fb0a685c
DRO
2204 err = unshare_fs(unshare_flags, &new_fs);
2205 if (err)
9bfb23fc 2206 goto bad_unshare_out;
fb0a685c
DRO
2207 err = unshare_fd(unshare_flags, &new_fd);
2208 if (err)
9bfb23fc 2209 goto bad_unshare_cleanup_fs;
b2e0d987 2210 err = unshare_userns(unshare_flags, &new_cred);
fb0a685c 2211 if (err)
9edff4ab 2212 goto bad_unshare_cleanup_fd;
b2e0d987
EB
2213 err = unshare_nsproxy_namespaces(unshare_flags, &new_nsproxy,
2214 new_cred, new_fs);
2215 if (err)
2216 goto bad_unshare_cleanup_cred;
c0b2fc31 2217
b2e0d987 2218 if (new_fs || new_fd || do_sysvsem || new_cred || new_nsproxy) {
9edff4ab
MS
2219 if (do_sysvsem) {
2220 /*
2221 * CLONE_SYSVSEM is equivalent to sys_exit().
2222 */
2223 exit_sem(current);
2224 }
ab602f79
JM
2225 if (unshare_flags & CLONE_NEWIPC) {
2226 /* Orphan segments in old ns (see sem above). */
2227 exit_shm(current);
2228 shm_init_task(current);
2229 }
ab516013 2230
6f977e6b 2231 if (new_nsproxy)
cf7b708c 2232 switch_task_namespaces(current, new_nsproxy);
cf2e340f 2233
cf7b708c
PE
2234 task_lock(current);
2235
cf2e340f
JD
2236 if (new_fs) {
2237 fs = current->fs;
2a4419b5 2238 spin_lock(&fs->lock);
cf2e340f 2239 current->fs = new_fs;
498052bb
AV
2240 if (--fs->users)
2241 new_fs = NULL;
2242 else
2243 new_fs = fs;
2a4419b5 2244 spin_unlock(&fs->lock);
cf2e340f
JD
2245 }
2246
cf2e340f
JD
2247 if (new_fd) {
2248 fd = current->files;
2249 current->files = new_fd;
2250 new_fd = fd;
2251 }
2252
2253 task_unlock(current);
b2e0d987
EB
2254
2255 if (new_cred) {
2256 /* Install the new user namespace */
2257 commit_creds(new_cred);
2258 new_cred = NULL;
2259 }
cf2e340f
JD
2260 }
2261
b2e0d987
EB
2262bad_unshare_cleanup_cred:
2263 if (new_cred)
2264 put_cred(new_cred);
cf2e340f
JD
2265bad_unshare_cleanup_fd:
2266 if (new_fd)
2267 put_files_struct(new_fd);
2268
cf2e340f
JD
2269bad_unshare_cleanup_fs:
2270 if (new_fs)
498052bb 2271 free_fs_struct(new_fs);
cf2e340f 2272
cf2e340f
JD
2273bad_unshare_out:
2274 return err;
2275}
3b125388
AV
2276
2277/*
2278 * Helper to unshare the files of the current task.
2279 * We don't want to expose copy_files internals to
2280 * the exec layer of the kernel.
2281 */
2282
2283int unshare_files(struct files_struct **displaced)
2284{
2285 struct task_struct *task = current;
50704516 2286 struct files_struct *copy = NULL;
3b125388
AV
2287 int error;
2288
2289 error = unshare_fd(CLONE_FILES, &copy);
2290 if (error || !copy) {
2291 *displaced = NULL;
2292 return error;
2293 }
2294 *displaced = task->files;
2295 task_lock(task);
2296 task->files = copy;
2297 task_unlock(task);
2298 return 0;
2299}
16db3d3f
HS
2300
2301int sysctl_max_threads(struct ctl_table *table, int write,
2302 void __user *buffer, size_t *lenp, loff_t *ppos)
2303{
2304 struct ctl_table t;
2305 int ret;
2306 int threads = max_threads;
2307 int min = MIN_THREADS;
2308 int max = MAX_THREADS;
2309
2310 t = *table;
2311 t.data = &threads;
2312 t.extra1 = &min;
2313 t.extra2 = &max;
2314
2315 ret = proc_dointvec_minmax(&t, write, buffer, lenp, ppos);
2316 if (ret || !write)
2317 return ret;
2318
2319 set_max_threads(threads);
2320
2321 return 0;
2322}