4 * Copyright (C) 1991, 1992 Linus Torvalds
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()'
14 #include <linux/slab.h>
15 #include <linux/init.h>
16 #include <linux/unistd.h>
17 #include <linux/module.h>
18 #include <linux/vmalloc.h>
19 #include <linux/completion.h>
20 #include <linux/personality.h>
21 #include <linux/mempolicy.h>
22 #include <linux/sem.h>
23 #include <linux/file.h>
24 #include <linux/fdtable.h>
25 #include <linux/iocontext.h>
26 #include <linux/key.h>
27 #include <linux/binfmts.h>
28 #include <linux/mman.h>
29 #include <linux/mmu_notifier.h>
31 #include <linux/nsproxy.h>
32 #include <linux/capability.h>
33 #include <linux/cpu.h>
34 #include <linux/cgroup.h>
35 #include <linux/security.h>
36 #include <linux/hugetlb.h>
37 #include <linux/swap.h>
38 #include <linux/syscalls.h>
39 #include <linux/jiffies.h>
40 #include <linux/futex.h>
41 #include <linux/compat.h>
42 #include <linux/kthread.h>
43 #include <linux/task_io_accounting_ops.h>
44 #include <linux/rcupdate.h>
45 #include <linux/ptrace.h>
46 #include <linux/mount.h>
47 #include <linux/audit.h>
48 #include <linux/memcontrol.h>
49 #include <linux/ftrace.h>
50 #include <linux/profile.h>
51 #include <linux/rmap.h>
52 #include <linux/ksm.h>
53 #include <linux/acct.h>
54 #include <linux/tsacct_kern.h>
55 #include <linux/cn_proc.h>
56 #include <linux/freezer.h>
57 #include <linux/delayacct.h>
58 #include <linux/taskstats_kern.h>
59 #include <linux/random.h>
60 #include <linux/tty.h>
61 #include <linux/blkdev.h>
62 #include <linux/fs_struct.h>
63 #include <linux/magic.h>
64 #include <linux/perf_event.h>
65 #include <linux/posix-timers.h>
66 #include <linux/user-return-notifier.h>
67 #include <linux/oom.h>
68 #include <linux/khugepaged.h>
70 #include <asm/pgtable.h>
71 #include <asm/pgalloc.h>
72 #include <asm/uaccess.h>
73 #include <asm/mmu_context.h>
74 #include <asm/cacheflush.h>
75 #include <asm/tlbflush.h>
77 #include <trace/events/sched.h>
80 * Protected counters by write_lock_irq(&tasklist_lock)
82 unsigned long total_forks; /* Handle normal Linux uptimes. */
83 int nr_threads; /* The idle threads do not count.. */
85 int max_threads; /* tunable limit on nr_threads */
87 DEFINE_PER_CPU(unsigned long, process_counts) = 0;
89 __cacheline_aligned DEFINE_RWLOCK(tasklist_lock); /* outer */
91 #ifdef CONFIG_PROVE_RCU
92 int lockdep_tasklist_lock_is_held(void)
94 return lockdep_is_held(&tasklist_lock);
96 EXPORT_SYMBOL_GPL(lockdep_tasklist_lock_is_held);
97 #endif /* #ifdef CONFIG_PROVE_RCU */
99 int nr_processes(void)
104 for_each_possible_cpu(cpu)
105 total += per_cpu(process_counts, cpu);
110 #ifndef __HAVE_ARCH_TASK_STRUCT_ALLOCATOR
111 # define alloc_task_struct_node(node) \
112 kmem_cache_alloc_node(task_struct_cachep, GFP_KERNEL, node)
113 # define free_task_struct(tsk) \
114 kmem_cache_free(task_struct_cachep, (tsk))
115 static struct kmem_cache *task_struct_cachep;
118 #ifndef __HAVE_ARCH_THREAD_INFO_ALLOCATOR
119 static struct thread_info *alloc_thread_info_node(struct task_struct *tsk,
122 #ifdef CONFIG_DEBUG_STACK_USAGE
123 gfp_t mask = GFP_KERNEL | __GFP_ZERO;
125 gfp_t mask = GFP_KERNEL;
127 struct page *page = alloc_pages_node(node, mask, THREAD_SIZE_ORDER);
129 return page ? page_address(page) : NULL;
132 static inline void free_thread_info(struct thread_info *ti)
134 free_pages((unsigned long)ti, THREAD_SIZE_ORDER);
138 /* SLAB cache for signal_struct structures (tsk->signal) */
139 static struct kmem_cache *signal_cachep;
141 /* SLAB cache for sighand_struct structures (tsk->sighand) */
142 struct kmem_cache *sighand_cachep;
144 /* SLAB cache for files_struct structures (tsk->files) */
145 struct kmem_cache *files_cachep;
147 /* SLAB cache for fs_struct structures (tsk->fs) */
148 struct kmem_cache *fs_cachep;
150 /* SLAB cache for vm_area_struct structures */
151 struct kmem_cache *vm_area_cachep;
153 /* SLAB cache for mm_struct structures (tsk->mm) */
154 static struct kmem_cache *mm_cachep;
156 static void account_kernel_stack(struct thread_info *ti, int account)
158 struct zone *zone = page_zone(virt_to_page(ti));
160 mod_zone_page_state(zone, NR_KERNEL_STACK, account);
163 void free_task(struct task_struct *tsk)
165 account_kernel_stack(tsk->stack, -1);
166 free_thread_info(tsk->stack);
167 rt_mutex_debug_task_free(tsk);
168 ftrace_graph_exit_task(tsk);
169 free_task_struct(tsk);
171 EXPORT_SYMBOL(free_task);
173 static inline void free_signal_struct(struct signal_struct *sig)
175 taskstats_tgid_free(sig);
176 sched_autogroup_exit(sig);
177 kmem_cache_free(signal_cachep, sig);
180 static inline void put_signal_struct(struct signal_struct *sig)
182 if (atomic_dec_and_test(&sig->sigcnt))
183 free_signal_struct(sig);
186 void __put_task_struct(struct task_struct *tsk)
188 WARN_ON(!tsk->exit_state);
189 WARN_ON(atomic_read(&tsk->usage));
190 WARN_ON(tsk == current);
193 delayacct_tsk_free(tsk);
194 put_signal_struct(tsk->signal);
196 if (!profile_handoff_task(tsk))
199 EXPORT_SYMBOL_GPL(__put_task_struct);
202 * macro override instead of weak attribute alias, to workaround
203 * gcc 4.1.0 and 4.1.1 bugs with weak attribute and empty functions.
205 #ifndef arch_task_cache_init
206 #define arch_task_cache_init()
209 void __init fork_init(unsigned long mempages)
211 #ifndef __HAVE_ARCH_TASK_STRUCT_ALLOCATOR
212 #ifndef ARCH_MIN_TASKALIGN
213 #define ARCH_MIN_TASKALIGN L1_CACHE_BYTES
215 /* create a slab on which task_structs can be allocated */
217 kmem_cache_create("task_struct", sizeof(struct task_struct),
218 ARCH_MIN_TASKALIGN, SLAB_PANIC | SLAB_NOTRACK, NULL);
221 /* do the arch specific task caches init */
222 arch_task_cache_init();
225 * The default maximum number of threads is set to a safe
226 * value: the thread structures can take up at most half
229 max_threads = mempages / (8 * THREAD_SIZE / PAGE_SIZE);
232 * we need to allow at least 20 threads to boot a system
234 if (max_threads < 20)
237 init_task.signal->rlim[RLIMIT_NPROC].rlim_cur = max_threads/2;
238 init_task.signal->rlim[RLIMIT_NPROC].rlim_max = max_threads/2;
239 init_task.signal->rlim[RLIMIT_SIGPENDING] =
240 init_task.signal->rlim[RLIMIT_NPROC];
243 int __attribute__((weak)) arch_dup_task_struct(struct task_struct *dst,
244 struct task_struct *src)
250 static struct task_struct *dup_task_struct(struct task_struct *orig)
252 struct task_struct *tsk;
253 struct thread_info *ti;
254 unsigned long *stackend;
255 int node = tsk_fork_get_node(orig);
258 prepare_to_copy(orig);
260 tsk = alloc_task_struct_node(node);
264 ti = alloc_thread_info_node(tsk, node);
266 free_task_struct(tsk);
270 err = arch_dup_task_struct(tsk, orig);
276 setup_thread_stack(tsk, orig);
277 clear_user_return_notifier(tsk);
278 clear_tsk_need_resched(tsk);
279 stackend = end_of_stack(tsk);
280 *stackend = STACK_END_MAGIC; /* for overflow detection */
282 #ifdef CONFIG_CC_STACKPROTECTOR
283 tsk->stack_canary = get_random_int();
287 * One for us, one for whoever does the "release_task()" (usually
290 atomic_set(&tsk->usage, 2);
291 #ifdef CONFIG_BLK_DEV_IO_TRACE
294 tsk->splice_pipe = NULL;
296 account_kernel_stack(ti, 1);
301 free_thread_info(ti);
302 free_task_struct(tsk);
307 static int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm)
309 struct vm_area_struct *mpnt, *tmp, *prev, **pprev;
310 struct rb_node **rb_link, *rb_parent;
312 unsigned long charge;
313 struct mempolicy *pol;
315 down_write(&oldmm->mmap_sem);
316 flush_cache_dup_mm(oldmm);
318 * Not linked in yet - no deadlock potential:
320 down_write_nested(&mm->mmap_sem, SINGLE_DEPTH_NESTING);
324 mm->mmap_cache = NULL;
325 mm->free_area_cache = oldmm->mmap_base;
326 mm->cached_hole_size = ~0UL;
328 cpumask_clear(mm_cpumask(mm));
330 rb_link = &mm->mm_rb.rb_node;
333 retval = ksm_fork(mm, oldmm);
336 retval = khugepaged_fork(mm, oldmm);
341 for (mpnt = oldmm->mmap; mpnt; mpnt = mpnt->vm_next) {
344 if (mpnt->vm_flags & VM_DONTCOPY) {
345 long pages = vma_pages(mpnt);
346 mm->total_vm -= pages;
347 vm_stat_account(mm, mpnt->vm_flags, mpnt->vm_file,
352 if (mpnt->vm_flags & VM_ACCOUNT) {
353 unsigned int len = (mpnt->vm_end - mpnt->vm_start) >> PAGE_SHIFT;
354 if (security_vm_enough_memory(len))
358 tmp = kmem_cache_alloc(vm_area_cachep, GFP_KERNEL);
362 INIT_LIST_HEAD(&tmp->anon_vma_chain);
363 pol = mpol_dup(vma_policy(mpnt));
364 retval = PTR_ERR(pol);
366 goto fail_nomem_policy;
367 vma_set_policy(tmp, pol);
369 if (anon_vma_fork(tmp, mpnt))
370 goto fail_nomem_anon_vma_fork;
371 tmp->vm_flags &= ~VM_LOCKED;
372 tmp->vm_next = tmp->vm_prev = NULL;
375 struct inode *inode = file->f_path.dentry->d_inode;
376 struct address_space *mapping = file->f_mapping;
379 if (tmp->vm_flags & VM_DENYWRITE)
380 atomic_dec(&inode->i_writecount);
381 mutex_lock(&mapping->i_mmap_mutex);
382 if (tmp->vm_flags & VM_SHARED)
383 mapping->i_mmap_writable++;
384 flush_dcache_mmap_lock(mapping);
385 /* insert tmp into the share list, just after mpnt */
386 vma_prio_tree_add(tmp, mpnt);
387 flush_dcache_mmap_unlock(mapping);
388 mutex_unlock(&mapping->i_mmap_mutex);
392 * Clear hugetlb-related page reserves for children. This only
393 * affects MAP_PRIVATE mappings. Faults generated by the child
394 * are not guaranteed to succeed, even if read-only
396 if (is_vm_hugetlb_page(tmp))
397 reset_vma_resv_huge_pages(tmp);
400 * Link in the new vma and copy the page table entries.
403 pprev = &tmp->vm_next;
407 __vma_link_rb(mm, tmp, rb_link, rb_parent);
408 rb_link = &tmp->vm_rb.rb_right;
409 rb_parent = &tmp->vm_rb;
412 retval = copy_page_range(mm, oldmm, mpnt);
414 if (tmp->vm_ops && tmp->vm_ops->open)
415 tmp->vm_ops->open(tmp);
420 /* a new mm has just been created */
421 arch_dup_mmap(oldmm, mm);
424 up_write(&mm->mmap_sem);
426 up_write(&oldmm->mmap_sem);
428 fail_nomem_anon_vma_fork:
431 kmem_cache_free(vm_area_cachep, tmp);
434 vm_unacct_memory(charge);
438 static inline int mm_alloc_pgd(struct mm_struct *mm)
440 mm->pgd = pgd_alloc(mm);
441 if (unlikely(!mm->pgd))
446 static inline void mm_free_pgd(struct mm_struct *mm)
448 pgd_free(mm, mm->pgd);
451 #define dup_mmap(mm, oldmm) (0)
452 #define mm_alloc_pgd(mm) (0)
453 #define mm_free_pgd(mm)
454 #endif /* CONFIG_MMU */
456 __cacheline_aligned_in_smp DEFINE_SPINLOCK(mmlist_lock);
458 #define allocate_mm() (kmem_cache_alloc(mm_cachep, GFP_KERNEL))
459 #define free_mm(mm) (kmem_cache_free(mm_cachep, (mm)))
461 static unsigned long default_dump_filter = MMF_DUMP_FILTER_DEFAULT;
463 static int __init coredump_filter_setup(char *s)
465 default_dump_filter =
466 (simple_strtoul(s, NULL, 0) << MMF_DUMP_FILTER_SHIFT) &
467 MMF_DUMP_FILTER_MASK;
471 __setup("coredump_filter=", coredump_filter_setup);
473 #include <linux/init_task.h>
475 static void mm_init_aio(struct mm_struct *mm)
478 spin_lock_init(&mm->ioctx_lock);
479 INIT_HLIST_HEAD(&mm->ioctx_list);
483 static struct mm_struct *mm_init(struct mm_struct *mm, struct task_struct *p)
485 atomic_set(&mm->mm_users, 1);
486 atomic_set(&mm->mm_count, 1);
487 init_rwsem(&mm->mmap_sem);
488 INIT_LIST_HEAD(&mm->mmlist);
489 mm->flags = (current->mm) ?
490 (current->mm->flags & MMF_INIT_MASK) : default_dump_filter;
491 mm->core_state = NULL;
493 memset(&mm->rss_stat, 0, sizeof(mm->rss_stat));
494 spin_lock_init(&mm->page_table_lock);
495 mm->free_area_cache = TASK_UNMAPPED_BASE;
496 mm->cached_hole_size = ~0UL;
498 mm_init_owner(mm, p);
500 if (likely(!mm_alloc_pgd(mm))) {
502 mmu_notifier_mm_init(mm);
511 * Allocate and initialize an mm_struct.
513 struct mm_struct *mm_alloc(void)
515 struct mm_struct *mm;
521 memset(mm, 0, sizeof(*mm));
523 return mm_init(mm, current);
527 * Called when the last reference to the mm
528 * is dropped: either by a lazy thread or by
529 * mmput. Free the page directory and the mm.
531 void __mmdrop(struct mm_struct *mm)
533 BUG_ON(mm == &init_mm);
536 mmu_notifier_mm_destroy(mm);
537 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
538 VM_BUG_ON(mm->pmd_huge_pte);
542 EXPORT_SYMBOL_GPL(__mmdrop);
545 * Decrement the use count and release all resources for an mm.
547 void mmput(struct mm_struct *mm)
551 if (atomic_dec_and_test(&mm->mm_users)) {
554 khugepaged_exit(mm); /* must run before exit_mmap */
556 set_mm_exe_file(mm, NULL);
557 if (!list_empty(&mm->mmlist)) {
558 spin_lock(&mmlist_lock);
559 list_del(&mm->mmlist);
560 spin_unlock(&mmlist_lock);
564 module_put(mm->binfmt->module);
568 EXPORT_SYMBOL_GPL(mmput);
571 * We added or removed a vma mapping the executable. The vmas are only mapped
572 * during exec and are not mapped with the mmap system call.
573 * Callers must hold down_write() on the mm's mmap_sem for these
575 void added_exe_file_vma(struct mm_struct *mm)
577 mm->num_exe_file_vmas++;
580 void removed_exe_file_vma(struct mm_struct *mm)
582 mm->num_exe_file_vmas--;
583 if ((mm->num_exe_file_vmas == 0) && mm->exe_file) {
590 void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file)
593 get_file(new_exe_file);
596 mm->exe_file = new_exe_file;
597 mm->num_exe_file_vmas = 0;
600 struct file *get_mm_exe_file(struct mm_struct *mm)
602 struct file *exe_file;
604 /* We need mmap_sem to protect against races with removal of
605 * VM_EXECUTABLE vmas */
606 down_read(&mm->mmap_sem);
607 exe_file = mm->exe_file;
610 up_read(&mm->mmap_sem);
614 static void dup_mm_exe_file(struct mm_struct *oldmm, struct mm_struct *newmm)
616 /* It's safe to write the exe_file pointer without exe_file_lock because
617 * this is called during fork when the task is not yet in /proc */
618 newmm->exe_file = get_mm_exe_file(oldmm);
622 * get_task_mm - acquire a reference to the task's mm
624 * Returns %NULL if the task has no mm. Checks PF_KTHREAD (meaning
625 * this kernel workthread has transiently adopted a user mm with use_mm,
626 * to do its AIO) is not set and if so returns a reference to it, after
627 * bumping up the use count. User must release the mm via mmput()
628 * after use. Typically used by /proc and ptrace.
630 struct mm_struct *get_task_mm(struct task_struct *task)
632 struct mm_struct *mm;
637 if (task->flags & PF_KTHREAD)
640 atomic_inc(&mm->mm_users);
645 EXPORT_SYMBOL_GPL(get_task_mm);
647 /* Please note the differences between mmput and mm_release.
648 * mmput is called whenever we stop holding onto a mm_struct,
649 * error success whatever.
651 * mm_release is called after a mm_struct has been removed
652 * from the current process.
654 * This difference is important for error handling, when we
655 * only half set up a mm_struct for a new process and need to restore
656 * the old one. Because we mmput the new mm_struct before
657 * restoring the old one. . .
658 * Eric Biederman 10 January 1998
660 void mm_release(struct task_struct *tsk, struct mm_struct *mm)
662 struct completion *vfork_done = tsk->vfork_done;
664 /* Get rid of any futexes when releasing the mm */
666 if (unlikely(tsk->robust_list)) {
667 exit_robust_list(tsk);
668 tsk->robust_list = NULL;
671 if (unlikely(tsk->compat_robust_list)) {
672 compat_exit_robust_list(tsk);
673 tsk->compat_robust_list = NULL;
676 if (unlikely(!list_empty(&tsk->pi_state_list)))
677 exit_pi_state_list(tsk);
680 /* Get rid of any cached register state */
681 deactivate_mm(tsk, mm);
683 /* notify parent sleeping on vfork() */
685 tsk->vfork_done = NULL;
686 complete(vfork_done);
690 * If we're exiting normally, clear a user-space tid field if
691 * requested. We leave this alone when dying by signal, to leave
692 * the value intact in a core dump, and to save the unnecessary
693 * trouble otherwise. Userland only wants this done for a sys_exit.
695 if (tsk->clear_child_tid) {
696 if (!(tsk->flags & PF_SIGNALED) &&
697 atomic_read(&mm->mm_users) > 1) {
699 * We don't check the error code - if userspace has
700 * not set up a proper pointer then tough luck.
702 put_user(0, tsk->clear_child_tid);
703 sys_futex(tsk->clear_child_tid, FUTEX_WAKE,
706 tsk->clear_child_tid = NULL;
711 * Allocate a new mm structure and copy contents from the
712 * mm structure of the passed in task structure.
714 struct mm_struct *dup_mm(struct task_struct *tsk)
716 struct mm_struct *mm, *oldmm = current->mm;
726 memcpy(mm, oldmm, sizeof(*mm));
729 /* Initializing for Swap token stuff */
730 mm->token_priority = 0;
731 mm->last_interval = 0;
733 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
734 mm->pmd_huge_pte = NULL;
737 if (!mm_init(mm, tsk))
740 if (init_new_context(tsk, mm))
743 dup_mm_exe_file(oldmm, mm);
745 err = dup_mmap(mm, oldmm);
749 mm->hiwater_rss = get_mm_rss(mm);
750 mm->hiwater_vm = mm->total_vm;
752 if (mm->binfmt && !try_module_get(mm->binfmt->module))
758 /* don't put binfmt in mmput, we haven't got module yet */
767 * If init_new_context() failed, we cannot use mmput() to free the mm
768 * because it calls destroy_context()
775 static int copy_mm(unsigned long clone_flags, struct task_struct *tsk)
777 struct mm_struct *mm, *oldmm;
780 tsk->min_flt = tsk->maj_flt = 0;
781 tsk->nvcsw = tsk->nivcsw = 0;
782 #ifdef CONFIG_DETECT_HUNG_TASK
783 tsk->last_switch_count = tsk->nvcsw + tsk->nivcsw;
787 tsk->active_mm = NULL;
790 * Are we cloning a kernel thread?
792 * We need to steal a active VM for that..
798 if (clone_flags & CLONE_VM) {
799 atomic_inc(&oldmm->mm_users);
810 /* Initializing for Swap token stuff */
811 mm->token_priority = 0;
812 mm->last_interval = 0;
822 static int copy_fs(unsigned long clone_flags, struct task_struct *tsk)
824 struct fs_struct *fs = current->fs;
825 if (clone_flags & CLONE_FS) {
826 /* tsk->fs is already what we want */
827 spin_lock(&fs->lock);
829 spin_unlock(&fs->lock);
833 spin_unlock(&fs->lock);
836 tsk->fs = copy_fs_struct(fs);
842 static int copy_files(unsigned long clone_flags, struct task_struct *tsk)
844 struct files_struct *oldf, *newf;
848 * A background process may not have any files ...
850 oldf = current->files;
854 if (clone_flags & CLONE_FILES) {
855 atomic_inc(&oldf->count);
859 newf = dup_fd(oldf, &error);
869 static int copy_io(unsigned long clone_flags, struct task_struct *tsk)
872 struct io_context *ioc = current->io_context;
877 * Share io context with parent, if CLONE_IO is set
879 if (clone_flags & CLONE_IO) {
880 tsk->io_context = ioc_task_link(ioc);
881 if (unlikely(!tsk->io_context))
883 } else if (ioprio_valid(ioc->ioprio)) {
884 tsk->io_context = alloc_io_context(GFP_KERNEL, -1);
885 if (unlikely(!tsk->io_context))
888 tsk->io_context->ioprio = ioc->ioprio;
894 static int copy_sighand(unsigned long clone_flags, struct task_struct *tsk)
896 struct sighand_struct *sig;
898 if (clone_flags & CLONE_SIGHAND) {
899 atomic_inc(¤t->sighand->count);
902 sig = kmem_cache_alloc(sighand_cachep, GFP_KERNEL);
903 rcu_assign_pointer(tsk->sighand, sig);
906 atomic_set(&sig->count, 1);
907 memcpy(sig->action, current->sighand->action, sizeof(sig->action));
911 void __cleanup_sighand(struct sighand_struct *sighand)
913 if (atomic_dec_and_test(&sighand->count))
914 kmem_cache_free(sighand_cachep, sighand);
919 * Initialize POSIX timer handling for a thread group.
921 static void posix_cpu_timers_init_group(struct signal_struct *sig)
923 unsigned long cpu_limit;
925 /* Thread group counters. */
926 thread_group_cputime_init(sig);
928 cpu_limit = ACCESS_ONCE(sig->rlim[RLIMIT_CPU].rlim_cur);
929 if (cpu_limit != RLIM_INFINITY) {
930 sig->cputime_expires.prof_exp = secs_to_cputime(cpu_limit);
931 sig->cputimer.running = 1;
934 /* The timer lists. */
935 INIT_LIST_HEAD(&sig->cpu_timers[0]);
936 INIT_LIST_HEAD(&sig->cpu_timers[1]);
937 INIT_LIST_HEAD(&sig->cpu_timers[2]);
940 static int copy_signal(unsigned long clone_flags, struct task_struct *tsk)
942 struct signal_struct *sig;
944 if (clone_flags & CLONE_THREAD)
947 sig = kmem_cache_zalloc(signal_cachep, GFP_KERNEL);
953 atomic_set(&sig->live, 1);
954 atomic_set(&sig->sigcnt, 1);
955 init_waitqueue_head(&sig->wait_chldexit);
956 if (clone_flags & CLONE_NEWPID)
957 sig->flags |= SIGNAL_UNKILLABLE;
958 sig->curr_target = tsk;
959 init_sigpending(&sig->shared_pending);
960 INIT_LIST_HEAD(&sig->posix_timers);
962 hrtimer_init(&sig->real_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
963 sig->real_timer.function = it_real_fn;
965 task_lock(current->group_leader);
966 memcpy(sig->rlim, current->signal->rlim, sizeof sig->rlim);
967 task_unlock(current->group_leader);
969 posix_cpu_timers_init_group(sig);
972 sched_autogroup_fork(sig);
974 #ifdef CONFIG_CGROUPS
975 init_rwsem(&sig->group_rwsem);
978 sig->oom_adj = current->signal->oom_adj;
979 sig->oom_score_adj = current->signal->oom_score_adj;
980 sig->oom_score_adj_min = current->signal->oom_score_adj_min;
982 mutex_init(&sig->cred_guard_mutex);
987 static void copy_flags(unsigned long clone_flags, struct task_struct *p)
989 unsigned long new_flags = p->flags;
991 new_flags &= ~(PF_SUPERPRIV | PF_WQ_WORKER);
992 new_flags |= PF_FORKNOEXEC;
993 new_flags |= PF_STARTING;
994 p->flags = new_flags;
997 SYSCALL_DEFINE1(set_tid_address, int __user *, tidptr)
999 current->clear_child_tid = tidptr;
1001 return task_pid_vnr(current);
1004 static void rt_mutex_init_task(struct task_struct *p)
1006 raw_spin_lock_init(&p->pi_lock);
1007 #ifdef CONFIG_RT_MUTEXES
1008 plist_head_init(&p->pi_waiters);
1009 p->pi_blocked_on = NULL;
1013 #ifdef CONFIG_MM_OWNER
1014 void mm_init_owner(struct mm_struct *mm, struct task_struct *p)
1018 #endif /* CONFIG_MM_OWNER */
1021 * Initialize POSIX timer handling for a single task.
1023 static void posix_cpu_timers_init(struct task_struct *tsk)
1025 tsk->cputime_expires.prof_exp = 0;
1026 tsk->cputime_expires.virt_exp = 0;
1027 tsk->cputime_expires.sched_exp = 0;
1028 INIT_LIST_HEAD(&tsk->cpu_timers[0]);
1029 INIT_LIST_HEAD(&tsk->cpu_timers[1]);
1030 INIT_LIST_HEAD(&tsk->cpu_timers[2]);
1034 * This creates a new process as a copy of the old one,
1035 * but does not actually start it yet.
1037 * It copies the registers, and all the appropriate
1038 * parts of the process environment (as per the clone
1039 * flags). The actual kick-off is left to the caller.
1041 static struct task_struct *copy_process(unsigned long clone_flags,
1042 unsigned long stack_start,
1043 struct pt_regs *regs,
1044 unsigned long stack_size,
1045 int __user *child_tidptr,
1050 struct task_struct *p;
1051 int cgroup_callbacks_done = 0;
1053 if ((clone_flags & (CLONE_NEWNS|CLONE_FS)) == (CLONE_NEWNS|CLONE_FS))
1054 return ERR_PTR(-EINVAL);
1057 * Thread groups must share signals as well, and detached threads
1058 * can only be started up within the thread group.
1060 if ((clone_flags & CLONE_THREAD) && !(clone_flags & CLONE_SIGHAND))
1061 return ERR_PTR(-EINVAL);
1064 * Shared signal handlers imply shared VM. By way of the above,
1065 * thread groups also imply shared VM. Blocking this case allows
1066 * for various simplifications in other code.
1068 if ((clone_flags & CLONE_SIGHAND) && !(clone_flags & CLONE_VM))
1069 return ERR_PTR(-EINVAL);
1072 * Siblings of global init remain as zombies on exit since they are
1073 * not reaped by their parent (swapper). To solve this and to avoid
1074 * multi-rooted process trees, prevent global and container-inits
1075 * from creating siblings.
1077 if ((clone_flags & CLONE_PARENT) &&
1078 current->signal->flags & SIGNAL_UNKILLABLE)
1079 return ERR_PTR(-EINVAL);
1081 retval = security_task_create(clone_flags);
1086 p = dup_task_struct(current);
1090 ftrace_graph_init_task(p);
1092 rt_mutex_init_task(p);
1094 #ifdef CONFIG_PROVE_LOCKING
1095 DEBUG_LOCKS_WARN_ON(!p->hardirqs_enabled);
1096 DEBUG_LOCKS_WARN_ON(!p->softirqs_enabled);
1099 if (atomic_read(&p->real_cred->user->processes) >=
1100 task_rlimit(p, RLIMIT_NPROC)) {
1101 if (!capable(CAP_SYS_ADMIN) && !capable(CAP_SYS_RESOURCE) &&
1102 p->real_cred->user != INIT_USER)
1105 current->flags &= ~PF_NPROC_EXCEEDED;
1107 retval = copy_creds(p, clone_flags);
1112 * If multiple threads are within copy_process(), then this check
1113 * triggers too late. This doesn't hurt, the check is only there
1114 * to stop root fork bombs.
1117 if (nr_threads >= max_threads)
1118 goto bad_fork_cleanup_count;
1120 if (!try_module_get(task_thread_info(p)->exec_domain->module))
1121 goto bad_fork_cleanup_count;
1124 delayacct_tsk_init(p); /* Must remain after dup_task_struct() */
1125 copy_flags(clone_flags, p);
1126 INIT_LIST_HEAD(&p->children);
1127 INIT_LIST_HEAD(&p->sibling);
1128 rcu_copy_process(p);
1129 p->vfork_done = NULL;
1130 spin_lock_init(&p->alloc_lock);
1132 init_sigpending(&p->pending);
1134 p->utime = p->stime = p->gtime = 0;
1135 p->utimescaled = p->stimescaled = 0;
1136 #ifndef CONFIG_VIRT_CPU_ACCOUNTING
1137 p->prev_utime = p->prev_stime = 0;
1139 #if defined(SPLIT_RSS_COUNTING)
1140 memset(&p->rss_stat, 0, sizeof(p->rss_stat));
1143 p->default_timer_slack_ns = current->timer_slack_ns;
1145 task_io_accounting_init(&p->ioac);
1146 acct_clear_integrals(p);
1148 posix_cpu_timers_init(p);
1150 do_posix_clock_monotonic_gettime(&p->start_time);
1151 p->real_start_time = p->start_time;
1152 monotonic_to_bootbased(&p->real_start_time);
1153 p->io_context = NULL;
1154 p->audit_context = NULL;
1155 if (clone_flags & CLONE_THREAD)
1156 threadgroup_change_begin(current);
1159 p->mempolicy = mpol_dup(p->mempolicy);
1160 if (IS_ERR(p->mempolicy)) {
1161 retval = PTR_ERR(p->mempolicy);
1162 p->mempolicy = NULL;
1163 goto bad_fork_cleanup_cgroup;
1165 mpol_fix_fork_child_flag(p);
1167 #ifdef CONFIG_CPUSETS
1168 p->cpuset_mem_spread_rotor = NUMA_NO_NODE;
1169 p->cpuset_slab_spread_rotor = NUMA_NO_NODE;
1171 #ifdef CONFIG_TRACE_IRQFLAGS
1173 #ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW
1174 p->hardirqs_enabled = 1;
1176 p->hardirqs_enabled = 0;
1178 p->hardirq_enable_ip = 0;
1179 p->hardirq_enable_event = 0;
1180 p->hardirq_disable_ip = _THIS_IP_;
1181 p->hardirq_disable_event = 0;
1182 p->softirqs_enabled = 1;
1183 p->softirq_enable_ip = _THIS_IP_;
1184 p->softirq_enable_event = 0;
1185 p->softirq_disable_ip = 0;
1186 p->softirq_disable_event = 0;
1187 p->hardirq_context = 0;
1188 p->softirq_context = 0;
1190 #ifdef CONFIG_LOCKDEP
1191 p->lockdep_depth = 0; /* no locks held yet */
1192 p->curr_chain_key = 0;
1193 p->lockdep_recursion = 0;
1196 #ifdef CONFIG_DEBUG_MUTEXES
1197 p->blocked_on = NULL; /* not blocked yet */
1199 #ifdef CONFIG_CGROUP_MEM_RES_CTLR
1200 p->memcg_batch.do_batch = 0;
1201 p->memcg_batch.memcg = NULL;
1204 /* Perform scheduler related setup. Assign this task to a CPU. */
1207 retval = perf_event_init_task(p);
1209 goto bad_fork_cleanup_policy;
1210 retval = audit_alloc(p);
1212 goto bad_fork_cleanup_policy;
1213 /* copy all the process information */
1214 retval = copy_semundo(clone_flags, p);
1216 goto bad_fork_cleanup_audit;
1217 retval = copy_files(clone_flags, p);
1219 goto bad_fork_cleanup_semundo;
1220 retval = copy_fs(clone_flags, p);
1222 goto bad_fork_cleanup_files;
1223 retval = copy_sighand(clone_flags, p);
1225 goto bad_fork_cleanup_fs;
1226 retval = copy_signal(clone_flags, p);
1228 goto bad_fork_cleanup_sighand;
1229 retval = copy_mm(clone_flags, p);
1231 goto bad_fork_cleanup_signal;
1232 retval = copy_namespaces(clone_flags, p);
1234 goto bad_fork_cleanup_mm;
1235 retval = copy_io(clone_flags, p);
1237 goto bad_fork_cleanup_namespaces;
1238 retval = copy_thread(clone_flags, stack_start, stack_size, p, regs);
1240 goto bad_fork_cleanup_io;
1242 if (pid != &init_struct_pid) {
1244 pid = alloc_pid(p->nsproxy->pid_ns);
1246 goto bad_fork_cleanup_io;
1249 p->pid = pid_nr(pid);
1251 if (clone_flags & CLONE_THREAD)
1252 p->tgid = current->tgid;
1254 p->set_child_tid = (clone_flags & CLONE_CHILD_SETTID) ? child_tidptr : NULL;
1256 * Clear TID on mm_release()?
1258 p->clear_child_tid = (clone_flags & CLONE_CHILD_CLEARTID) ? child_tidptr : NULL;
1263 p->robust_list = NULL;
1264 #ifdef CONFIG_COMPAT
1265 p->compat_robust_list = NULL;
1267 INIT_LIST_HEAD(&p->pi_state_list);
1268 p->pi_state_cache = NULL;
1271 * sigaltstack should be cleared when sharing the same VM
1273 if ((clone_flags & (CLONE_VM|CLONE_VFORK)) == CLONE_VM)
1274 p->sas_ss_sp = p->sas_ss_size = 0;
1277 * Syscall tracing and stepping should be turned off in the
1278 * child regardless of CLONE_PTRACE.
1280 user_disable_single_step(p);
1281 clear_tsk_thread_flag(p, TIF_SYSCALL_TRACE);
1282 #ifdef TIF_SYSCALL_EMU
1283 clear_tsk_thread_flag(p, TIF_SYSCALL_EMU);
1285 clear_all_latency_tracing(p);
1287 /* ok, now we should be set up.. */
1288 p->exit_signal = (clone_flags & CLONE_THREAD) ? -1 : (clone_flags & CSIGNAL);
1289 p->pdeath_signal = 0;
1293 p->nr_dirtied_pause = 128 >> (PAGE_SHIFT - 10);
1296 * Ok, make it visible to the rest of the system.
1297 * We dont wake it up yet.
1299 p->group_leader = p;
1300 INIT_LIST_HEAD(&p->thread_group);
1302 /* Now that the task is set up, run cgroup callbacks if
1303 * necessary. We need to run them before the task is visible
1304 * on the tasklist. */
1305 cgroup_fork_callbacks(p);
1306 cgroup_callbacks_done = 1;
1308 /* Need tasklist lock for parent etc handling! */
1309 write_lock_irq(&tasklist_lock);
1311 /* CLONE_PARENT re-uses the old parent */
1312 if (clone_flags & (CLONE_PARENT|CLONE_THREAD)) {
1313 p->real_parent = current->real_parent;
1314 p->parent_exec_id = current->parent_exec_id;
1316 p->real_parent = current;
1317 p->parent_exec_id = current->self_exec_id;
1320 spin_lock(¤t->sighand->siglock);
1323 * Process group and session signals need to be delivered to just the
1324 * parent before the fork or both the parent and the child after the
1325 * fork. Restart if a signal comes in before we add the new process to
1326 * it's process group.
1327 * A fatal signal pending means that current will exit, so the new
1328 * thread can't slip out of an OOM kill (or normal SIGKILL).
1330 recalc_sigpending();
1331 if (signal_pending(current)) {
1332 spin_unlock(¤t->sighand->siglock);
1333 write_unlock_irq(&tasklist_lock);
1334 retval = -ERESTARTNOINTR;
1335 goto bad_fork_free_pid;
1338 if (clone_flags & CLONE_THREAD) {
1339 current->signal->nr_threads++;
1340 atomic_inc(¤t->signal->live);
1341 atomic_inc(¤t->signal->sigcnt);
1342 p->group_leader = current->group_leader;
1343 list_add_tail_rcu(&p->thread_group, &p->group_leader->thread_group);
1346 if (likely(p->pid)) {
1347 ptrace_init_task(p, (clone_flags & CLONE_PTRACE) || trace);
1349 if (thread_group_leader(p)) {
1350 if (is_child_reaper(pid))
1351 p->nsproxy->pid_ns->child_reaper = p;
1353 p->signal->leader_pid = pid;
1354 p->signal->tty = tty_kref_get(current->signal->tty);
1355 attach_pid(p, PIDTYPE_PGID, task_pgrp(current));
1356 attach_pid(p, PIDTYPE_SID, task_session(current));
1357 list_add_tail(&p->sibling, &p->real_parent->children);
1358 list_add_tail_rcu(&p->tasks, &init_task.tasks);
1359 __this_cpu_inc(process_counts);
1361 attach_pid(p, PIDTYPE_PID, pid);
1366 spin_unlock(¤t->sighand->siglock);
1367 write_unlock_irq(&tasklist_lock);
1368 proc_fork_connector(p);
1369 cgroup_post_fork(p);
1370 if (clone_flags & CLONE_THREAD)
1371 threadgroup_change_end(current);
1376 if (pid != &init_struct_pid)
1378 bad_fork_cleanup_io:
1381 bad_fork_cleanup_namespaces:
1382 exit_task_namespaces(p);
1383 bad_fork_cleanup_mm:
1386 bad_fork_cleanup_signal:
1387 if (!(clone_flags & CLONE_THREAD))
1388 free_signal_struct(p->signal);
1389 bad_fork_cleanup_sighand:
1390 __cleanup_sighand(p->sighand);
1391 bad_fork_cleanup_fs:
1392 exit_fs(p); /* blocking */
1393 bad_fork_cleanup_files:
1394 exit_files(p); /* blocking */
1395 bad_fork_cleanup_semundo:
1397 bad_fork_cleanup_audit:
1399 bad_fork_cleanup_policy:
1400 perf_event_free_task(p);
1402 mpol_put(p->mempolicy);
1403 bad_fork_cleanup_cgroup:
1405 if (clone_flags & CLONE_THREAD)
1406 threadgroup_change_end(current);
1407 cgroup_exit(p, cgroup_callbacks_done);
1408 delayacct_tsk_free(p);
1409 module_put(task_thread_info(p)->exec_domain->module);
1410 bad_fork_cleanup_count:
1411 atomic_dec(&p->cred->user->processes);
1416 return ERR_PTR(retval);
1419 noinline struct pt_regs * __cpuinit __attribute__((weak)) idle_regs(struct pt_regs *regs)
1421 memset(regs, 0, sizeof(struct pt_regs));
1425 static inline void init_idle_pids(struct pid_link *links)
1429 for (type = PIDTYPE_PID; type < PIDTYPE_MAX; ++type) {
1430 INIT_HLIST_NODE(&links[type].node); /* not really needed */
1431 links[type].pid = &init_struct_pid;
1435 struct task_struct * __cpuinit fork_idle(int cpu)
1437 struct task_struct *task;
1438 struct pt_regs regs;
1440 task = copy_process(CLONE_VM, 0, idle_regs(®s), 0, NULL,
1441 &init_struct_pid, 0);
1442 if (!IS_ERR(task)) {
1443 init_idle_pids(task->pids);
1444 init_idle(task, cpu);
1451 * Ok, this is the main fork-routine.
1453 * It copies the process, and if successful kick-starts
1454 * it and waits for it to finish using the VM if required.
1456 long do_fork(unsigned long clone_flags,
1457 unsigned long stack_start,
1458 struct pt_regs *regs,
1459 unsigned long stack_size,
1460 int __user *parent_tidptr,
1461 int __user *child_tidptr)
1463 struct task_struct *p;
1468 * Do some preliminary argument and permissions checking before we
1469 * actually start allocating stuff
1471 if (clone_flags & CLONE_NEWUSER) {
1472 if (clone_flags & CLONE_THREAD)
1474 /* hopefully this check will go away when userns support is
1477 if (!capable(CAP_SYS_ADMIN) || !capable(CAP_SETUID) ||
1478 !capable(CAP_SETGID))
1483 * Determine whether and which event to report to ptracer. When
1484 * called from kernel_thread or CLONE_UNTRACED is explicitly
1485 * requested, no event is reported; otherwise, report if the event
1486 * for the type of forking is enabled.
1488 if (likely(user_mode(regs)) && !(clone_flags & CLONE_UNTRACED)) {
1489 if (clone_flags & CLONE_VFORK)
1490 trace = PTRACE_EVENT_VFORK;
1491 else if ((clone_flags & CSIGNAL) != SIGCHLD)
1492 trace = PTRACE_EVENT_CLONE;
1494 trace = PTRACE_EVENT_FORK;
1496 if (likely(!ptrace_event_enabled(current, trace)))
1500 p = copy_process(clone_flags, stack_start, regs, stack_size,
1501 child_tidptr, NULL, trace);
1503 * Do this prior waking up the new thread - the thread pointer
1504 * might get invalid after that point, if the thread exits quickly.
1507 struct completion vfork;
1509 trace_sched_process_fork(current, p);
1511 nr = task_pid_vnr(p);
1513 if (clone_flags & CLONE_PARENT_SETTID)
1514 put_user(nr, parent_tidptr);
1516 if (clone_flags & CLONE_VFORK) {
1517 p->vfork_done = &vfork;
1518 init_completion(&vfork);
1521 audit_finish_fork(p);
1524 * We set PF_STARTING at creation in case tracing wants to
1525 * use this to distinguish a fully live task from one that
1526 * hasn't finished SIGSTOP raising yet. Now we clear it
1527 * and set the child going.
1529 p->flags &= ~PF_STARTING;
1531 wake_up_new_task(p);
1533 /* forking complete and child started to run, tell ptracer */
1534 if (unlikely(trace))
1535 ptrace_event(trace, nr);
1537 if (clone_flags & CLONE_VFORK) {
1538 freezer_do_not_count();
1539 wait_for_completion(&vfork);
1541 ptrace_event(PTRACE_EVENT_VFORK_DONE, nr);
1549 #ifndef ARCH_MIN_MMSTRUCT_ALIGN
1550 #define ARCH_MIN_MMSTRUCT_ALIGN 0
1553 static void sighand_ctor(void *data)
1555 struct sighand_struct *sighand = data;
1557 spin_lock_init(&sighand->siglock);
1558 init_waitqueue_head(&sighand->signalfd_wqh);
1561 void __init proc_caches_init(void)
1563 sighand_cachep = kmem_cache_create("sighand_cache",
1564 sizeof(struct sighand_struct), 0,
1565 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_DESTROY_BY_RCU|
1566 SLAB_NOTRACK, sighand_ctor);
1567 signal_cachep = kmem_cache_create("signal_cache",
1568 sizeof(struct signal_struct), 0,
1569 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL);
1570 files_cachep = kmem_cache_create("files_cache",
1571 sizeof(struct files_struct), 0,
1572 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL);
1573 fs_cachep = kmem_cache_create("fs_cache",
1574 sizeof(struct fs_struct), 0,
1575 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL);
1577 * FIXME! The "sizeof(struct mm_struct)" currently includes the
1578 * whole struct cpumask for the OFFSTACK case. We could change
1579 * this to *only* allocate as much of it as required by the
1580 * maximum number of CPU's we can ever have. The cpumask_allocation
1581 * is at the end of the structure, exactly for that reason.
1583 mm_cachep = kmem_cache_create("mm_struct",
1584 sizeof(struct mm_struct), ARCH_MIN_MMSTRUCT_ALIGN,
1585 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL);
1586 vm_area_cachep = KMEM_CACHE(vm_area_struct, SLAB_PANIC);
1588 nsproxy_cache_init();
1592 * Check constraints on flags passed to the unshare system call.
1594 static int check_unshare_flags(unsigned long unshare_flags)
1596 if (unshare_flags & ~(CLONE_THREAD|CLONE_FS|CLONE_NEWNS|CLONE_SIGHAND|
1597 CLONE_VM|CLONE_FILES|CLONE_SYSVSEM|
1598 CLONE_NEWUTS|CLONE_NEWIPC|CLONE_NEWNET))
1601 * Not implemented, but pretend it works if there is nothing to
1602 * unshare. Note that unsharing CLONE_THREAD or CLONE_SIGHAND
1603 * needs to unshare vm.
1605 if (unshare_flags & (CLONE_THREAD | CLONE_SIGHAND | CLONE_VM)) {
1606 /* FIXME: get_task_mm() increments ->mm_users */
1607 if (atomic_read(¤t->mm->mm_users) > 1)
1615 * Unshare the filesystem structure if it is being shared
1617 static int unshare_fs(unsigned long unshare_flags, struct fs_struct **new_fsp)
1619 struct fs_struct *fs = current->fs;
1621 if (!(unshare_flags & CLONE_FS) || !fs)
1624 /* don't need lock here; in the worst case we'll do useless copy */
1628 *new_fsp = copy_fs_struct(fs);
1636 * Unshare file descriptor table if it is being shared
1638 static int unshare_fd(unsigned long unshare_flags, struct files_struct **new_fdp)
1640 struct files_struct *fd = current->files;
1643 if ((unshare_flags & CLONE_FILES) &&
1644 (fd && atomic_read(&fd->count) > 1)) {
1645 *new_fdp = dup_fd(fd, &error);
1654 * unshare allows a process to 'unshare' part of the process
1655 * context which was originally shared using clone. copy_*
1656 * functions used by do_fork() cannot be used here directly
1657 * because they modify an inactive task_struct that is being
1658 * constructed. Here we are modifying the current, active,
1661 SYSCALL_DEFINE1(unshare, unsigned long, unshare_flags)
1663 struct fs_struct *fs, *new_fs = NULL;
1664 struct files_struct *fd, *new_fd = NULL;
1665 struct nsproxy *new_nsproxy = NULL;
1669 err = check_unshare_flags(unshare_flags);
1671 goto bad_unshare_out;
1674 * If unsharing namespace, must also unshare filesystem information.
1676 if (unshare_flags & CLONE_NEWNS)
1677 unshare_flags |= CLONE_FS;
1679 * CLONE_NEWIPC must also detach from the undolist: after switching
1680 * to a new ipc namespace, the semaphore arrays from the old
1681 * namespace are unreachable.
1683 if (unshare_flags & (CLONE_NEWIPC|CLONE_SYSVSEM))
1685 err = unshare_fs(unshare_flags, &new_fs);
1687 goto bad_unshare_out;
1688 err = unshare_fd(unshare_flags, &new_fd);
1690 goto bad_unshare_cleanup_fs;
1691 err = unshare_nsproxy_namespaces(unshare_flags, &new_nsproxy, new_fs);
1693 goto bad_unshare_cleanup_fd;
1695 if (new_fs || new_fd || do_sysvsem || new_nsproxy) {
1698 * CLONE_SYSVSEM is equivalent to sys_exit().
1704 switch_task_namespaces(current, new_nsproxy);
1712 spin_lock(&fs->lock);
1713 current->fs = new_fs;
1718 spin_unlock(&fs->lock);
1722 fd = current->files;
1723 current->files = new_fd;
1727 task_unlock(current);
1731 put_nsproxy(new_nsproxy);
1733 bad_unshare_cleanup_fd:
1735 put_files_struct(new_fd);
1737 bad_unshare_cleanup_fs:
1739 free_fs_struct(new_fs);
1746 * Helper to unshare the files of the current task.
1747 * We don't want to expose copy_files internals to
1748 * the exec layer of the kernel.
1751 int unshare_files(struct files_struct **displaced)
1753 struct task_struct *task = current;
1754 struct files_struct *copy = NULL;
1757 error = unshare_fd(CLONE_FILES, ©);
1758 if (error || !copy) {
1762 *displaced = task->files;