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/tracehook.h>
41 #include <linux/futex.h>
42 #include <linux/compat.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/proc_fs.h>
62 #include <linux/blkdev.h>
63 #include <linux/fs_struct.h>
64 #include <linux/magic.h>
65 #include <linux/perf_event.h>
66 #include <linux/posix-timers.h>
67 #include <linux/user-return-notifier.h>
69 #include <asm/pgtable.h>
70 #include <asm/pgalloc.h>
71 #include <asm/uaccess.h>
72 #include <asm/mmu_context.h>
73 #include <asm/cacheflush.h>
74 #include <asm/tlbflush.h>
76 #include <trace/events/sched.h>
79 * Protected counters by write_lock_irq(&tasklist_lock)
81 unsigned long total_forks; /* Handle normal Linux uptimes. */
82 int nr_threads; /* The idle threads do not count.. */
84 int max_threads; /* tunable limit on nr_threads */
86 DEFINE_PER_CPU(unsigned long, process_counts) = 0;
88 __cacheline_aligned DEFINE_RWLOCK(tasklist_lock); /* outer */
89 EXPORT_SYMBOL_GPL(tasklist_lock);
91 int nr_processes(void)
96 for_each_possible_cpu(cpu)
97 total += per_cpu(process_counts, cpu);
102 #ifndef __HAVE_ARCH_TASK_STRUCT_ALLOCATOR
103 # define alloc_task_struct() kmem_cache_alloc(task_struct_cachep, GFP_KERNEL)
104 # define free_task_struct(tsk) kmem_cache_free(task_struct_cachep, (tsk))
105 static struct kmem_cache *task_struct_cachep;
108 #ifndef __HAVE_ARCH_THREAD_INFO_ALLOCATOR
109 static inline struct thread_info *alloc_thread_info(struct task_struct *tsk)
111 #ifdef CONFIG_DEBUG_STACK_USAGE
112 gfp_t mask = GFP_KERNEL | __GFP_ZERO;
114 gfp_t mask = GFP_KERNEL;
116 return (struct thread_info *)__get_free_pages(mask, THREAD_SIZE_ORDER);
119 static inline void free_thread_info(struct thread_info *ti)
121 free_pages((unsigned long)ti, THREAD_SIZE_ORDER);
125 /* SLAB cache for signal_struct structures (tsk->signal) */
126 static struct kmem_cache *signal_cachep;
128 /* SLAB cache for sighand_struct structures (tsk->sighand) */
129 struct kmem_cache *sighand_cachep;
131 /* SLAB cache for files_struct structures (tsk->files) */
132 struct kmem_cache *files_cachep;
134 /* SLAB cache for fs_struct structures (tsk->fs) */
135 struct kmem_cache *fs_cachep;
137 /* SLAB cache for vm_area_struct structures */
138 struct kmem_cache *vm_area_cachep;
140 /* SLAB cache for mm_struct structures (tsk->mm) */
141 static struct kmem_cache *mm_cachep;
143 static void account_kernel_stack(struct thread_info *ti, int account)
145 struct zone *zone = page_zone(virt_to_page(ti));
147 mod_zone_page_state(zone, NR_KERNEL_STACK, account);
150 void free_task(struct task_struct *tsk)
152 prop_local_destroy_single(&tsk->dirties);
153 account_kernel_stack(tsk->stack, -1);
154 free_thread_info(tsk->stack);
155 rt_mutex_debug_task_free(tsk);
156 ftrace_graph_exit_task(tsk);
157 free_task_struct(tsk);
159 EXPORT_SYMBOL(free_task);
161 void __put_task_struct(struct task_struct *tsk)
163 WARN_ON(!tsk->exit_state);
164 WARN_ON(atomic_read(&tsk->usage));
165 WARN_ON(tsk == current);
168 delayacct_tsk_free(tsk);
170 if (!profile_handoff_task(tsk))
175 * macro override instead of weak attribute alias, to workaround
176 * gcc 4.1.0 and 4.1.1 bugs with weak attribute and empty functions.
178 #ifndef arch_task_cache_init
179 #define arch_task_cache_init()
182 void __init fork_init(unsigned long mempages)
184 #ifndef __HAVE_ARCH_TASK_STRUCT_ALLOCATOR
185 #ifndef ARCH_MIN_TASKALIGN
186 #define ARCH_MIN_TASKALIGN L1_CACHE_BYTES
188 /* create a slab on which task_structs can be allocated */
190 kmem_cache_create("task_struct", sizeof(struct task_struct),
191 ARCH_MIN_TASKALIGN, SLAB_PANIC | SLAB_NOTRACK, NULL);
194 /* do the arch specific task caches init */
195 arch_task_cache_init();
198 * The default maximum number of threads is set to a safe
199 * value: the thread structures can take up at most half
202 max_threads = mempages / (8 * THREAD_SIZE / PAGE_SIZE);
205 * we need to allow at least 20 threads to boot a system
210 init_task.signal->rlim[RLIMIT_NPROC].rlim_cur = max_threads/2;
211 init_task.signal->rlim[RLIMIT_NPROC].rlim_max = max_threads/2;
212 init_task.signal->rlim[RLIMIT_SIGPENDING] =
213 init_task.signal->rlim[RLIMIT_NPROC];
216 int __attribute__((weak)) arch_dup_task_struct(struct task_struct *dst,
217 struct task_struct *src)
223 static struct task_struct *dup_task_struct(struct task_struct *orig)
225 struct task_struct *tsk;
226 struct thread_info *ti;
227 unsigned long *stackend;
231 prepare_to_copy(orig);
233 tsk = alloc_task_struct();
237 ti = alloc_thread_info(tsk);
239 free_task_struct(tsk);
243 err = arch_dup_task_struct(tsk, orig);
249 err = prop_local_init_single(&tsk->dirties);
253 setup_thread_stack(tsk, orig);
254 clear_user_return_notifier(tsk);
255 stackend = end_of_stack(tsk);
256 *stackend = STACK_END_MAGIC; /* for overflow detection */
258 #ifdef CONFIG_CC_STACKPROTECTOR
259 tsk->stack_canary = get_random_int();
262 /* One for us, one for whoever does the "release_task()" (usually parent) */
263 atomic_set(&tsk->usage,2);
264 atomic_set(&tsk->fs_excl, 0);
265 #ifdef CONFIG_BLK_DEV_IO_TRACE
268 tsk->splice_pipe = NULL;
270 account_kernel_stack(ti, 1);
275 free_thread_info(ti);
276 free_task_struct(tsk);
281 static int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm)
283 struct vm_area_struct *mpnt, *tmp, **pprev;
284 struct rb_node **rb_link, *rb_parent;
286 unsigned long charge;
287 struct mempolicy *pol;
289 down_write(&oldmm->mmap_sem);
290 flush_cache_dup_mm(oldmm);
292 * Not linked in yet - no deadlock potential:
294 down_write_nested(&mm->mmap_sem, SINGLE_DEPTH_NESTING);
298 mm->mmap_cache = NULL;
299 mm->free_area_cache = oldmm->mmap_base;
300 mm->cached_hole_size = ~0UL;
302 cpumask_clear(mm_cpumask(mm));
304 rb_link = &mm->mm_rb.rb_node;
307 retval = ksm_fork(mm, oldmm);
311 for (mpnt = oldmm->mmap; mpnt; mpnt = mpnt->vm_next) {
314 if (mpnt->vm_flags & VM_DONTCOPY) {
315 long pages = vma_pages(mpnt);
316 mm->total_vm -= pages;
317 vm_stat_account(mm, mpnt->vm_flags, mpnt->vm_file,
322 if (mpnt->vm_flags & VM_ACCOUNT) {
323 unsigned int len = (mpnt->vm_end - mpnt->vm_start) >> PAGE_SHIFT;
324 if (security_vm_enough_memory(len))
328 tmp = kmem_cache_alloc(vm_area_cachep, GFP_KERNEL);
332 pol = mpol_dup(vma_policy(mpnt));
333 retval = PTR_ERR(pol);
335 goto fail_nomem_policy;
336 vma_set_policy(tmp, pol);
337 tmp->vm_flags &= ~VM_LOCKED;
343 struct inode *inode = file->f_path.dentry->d_inode;
344 struct address_space *mapping = file->f_mapping;
347 if (tmp->vm_flags & VM_DENYWRITE)
348 atomic_dec(&inode->i_writecount);
349 spin_lock(&mapping->i_mmap_lock);
350 if (tmp->vm_flags & VM_SHARED)
351 mapping->i_mmap_writable++;
352 tmp->vm_truncate_count = mpnt->vm_truncate_count;
353 flush_dcache_mmap_lock(mapping);
354 /* insert tmp into the share list, just after mpnt */
355 vma_prio_tree_add(tmp, mpnt);
356 flush_dcache_mmap_unlock(mapping);
357 spin_unlock(&mapping->i_mmap_lock);
361 * Clear hugetlb-related page reserves for children. This only
362 * affects MAP_PRIVATE mappings. Faults generated by the child
363 * are not guaranteed to succeed, even if read-only
365 if (is_vm_hugetlb_page(tmp))
366 reset_vma_resv_huge_pages(tmp);
369 * Link in the new vma and copy the page table entries.
372 pprev = &tmp->vm_next;
374 __vma_link_rb(mm, tmp, rb_link, rb_parent);
375 rb_link = &tmp->vm_rb.rb_right;
376 rb_parent = &tmp->vm_rb;
379 retval = copy_page_range(mm, oldmm, mpnt);
381 if (tmp->vm_ops && tmp->vm_ops->open)
382 tmp->vm_ops->open(tmp);
387 /* a new mm has just been created */
388 arch_dup_mmap(oldmm, mm);
391 up_write(&mm->mmap_sem);
393 up_write(&oldmm->mmap_sem);
396 kmem_cache_free(vm_area_cachep, tmp);
399 vm_unacct_memory(charge);
403 static inline int mm_alloc_pgd(struct mm_struct * mm)
405 mm->pgd = pgd_alloc(mm);
406 if (unlikely(!mm->pgd))
411 static inline void mm_free_pgd(struct mm_struct * mm)
413 pgd_free(mm, mm->pgd);
416 #define dup_mmap(mm, oldmm) (0)
417 #define mm_alloc_pgd(mm) (0)
418 #define mm_free_pgd(mm)
419 #endif /* CONFIG_MMU */
421 __cacheline_aligned_in_smp DEFINE_SPINLOCK(mmlist_lock);
423 #define allocate_mm() (kmem_cache_alloc(mm_cachep, GFP_KERNEL))
424 #define free_mm(mm) (kmem_cache_free(mm_cachep, (mm)))
426 static unsigned long default_dump_filter = MMF_DUMP_FILTER_DEFAULT;
428 static int __init coredump_filter_setup(char *s)
430 default_dump_filter =
431 (simple_strtoul(s, NULL, 0) << MMF_DUMP_FILTER_SHIFT) &
432 MMF_DUMP_FILTER_MASK;
436 __setup("coredump_filter=", coredump_filter_setup);
438 #include <linux/init_task.h>
440 static void mm_init_aio(struct mm_struct *mm)
443 spin_lock_init(&mm->ioctx_lock);
444 INIT_HLIST_HEAD(&mm->ioctx_list);
448 static struct mm_struct * mm_init(struct mm_struct * mm, struct task_struct *p)
450 atomic_set(&mm->mm_users, 1);
451 atomic_set(&mm->mm_count, 1);
452 init_rwsem(&mm->mmap_sem);
453 INIT_LIST_HEAD(&mm->mmlist);
454 mm->flags = (current->mm) ?
455 (current->mm->flags & MMF_INIT_MASK) : default_dump_filter;
456 mm->core_state = NULL;
458 memset(&mm->rss_stat, 0, sizeof(mm->rss_stat));
459 spin_lock_init(&mm->page_table_lock);
460 mm->free_area_cache = TASK_UNMAPPED_BASE;
461 mm->cached_hole_size = ~0UL;
463 mm_init_owner(mm, p);
465 if (likely(!mm_alloc_pgd(mm))) {
467 mmu_notifier_mm_init(mm);
476 * Allocate and initialize an mm_struct.
478 struct mm_struct * mm_alloc(void)
480 struct mm_struct * mm;
484 memset(mm, 0, sizeof(*mm));
485 mm = mm_init(mm, current);
491 * Called when the last reference to the mm
492 * is dropped: either by a lazy thread or by
493 * mmput. Free the page directory and the mm.
495 void __mmdrop(struct mm_struct *mm)
497 BUG_ON(mm == &init_mm);
500 mmu_notifier_mm_destroy(mm);
503 EXPORT_SYMBOL_GPL(__mmdrop);
506 * Decrement the use count and release all resources for an mm.
508 void mmput(struct mm_struct *mm)
512 if (atomic_dec_and_test(&mm->mm_users)) {
516 set_mm_exe_file(mm, NULL);
517 if (!list_empty(&mm->mmlist)) {
518 spin_lock(&mmlist_lock);
519 list_del(&mm->mmlist);
520 spin_unlock(&mmlist_lock);
524 module_put(mm->binfmt->module);
528 EXPORT_SYMBOL_GPL(mmput);
531 * get_task_mm - acquire a reference to the task's mm
533 * Returns %NULL if the task has no mm. Checks PF_KTHREAD (meaning
534 * this kernel workthread has transiently adopted a user mm with use_mm,
535 * to do its AIO) is not set and if so returns a reference to it, after
536 * bumping up the use count. User must release the mm via mmput()
537 * after use. Typically used by /proc and ptrace.
539 struct mm_struct *get_task_mm(struct task_struct *task)
541 struct mm_struct *mm;
546 if (task->flags & PF_KTHREAD)
549 atomic_inc(&mm->mm_users);
554 EXPORT_SYMBOL_GPL(get_task_mm);
556 /* Please note the differences between mmput and mm_release.
557 * mmput is called whenever we stop holding onto a mm_struct,
558 * error success whatever.
560 * mm_release is called after a mm_struct has been removed
561 * from the current process.
563 * This difference is important for error handling, when we
564 * only half set up a mm_struct for a new process and need to restore
565 * the old one. Because we mmput the new mm_struct before
566 * restoring the old one. . .
567 * Eric Biederman 10 January 1998
569 void mm_release(struct task_struct *tsk, struct mm_struct *mm)
571 struct completion *vfork_done = tsk->vfork_done;
573 /* Get rid of any futexes when releasing the mm */
575 if (unlikely(tsk->robust_list)) {
576 exit_robust_list(tsk);
577 tsk->robust_list = NULL;
580 if (unlikely(tsk->compat_robust_list)) {
581 compat_exit_robust_list(tsk);
582 tsk->compat_robust_list = NULL;
585 if (unlikely(!list_empty(&tsk->pi_state_list)))
586 exit_pi_state_list(tsk);
589 /* Get rid of any cached register state */
590 deactivate_mm(tsk, mm);
592 /* notify parent sleeping on vfork() */
594 tsk->vfork_done = NULL;
595 complete(vfork_done);
599 * If we're exiting normally, clear a user-space tid field if
600 * requested. We leave this alone when dying by signal, to leave
601 * the value intact in a core dump, and to save the unnecessary
602 * trouble otherwise. Userland only wants this done for a sys_exit.
604 if (tsk->clear_child_tid) {
605 if (!(tsk->flags & PF_SIGNALED) &&
606 atomic_read(&mm->mm_users) > 1) {
608 * We don't check the error code - if userspace has
609 * not set up a proper pointer then tough luck.
611 put_user(0, tsk->clear_child_tid);
612 sys_futex(tsk->clear_child_tid, FUTEX_WAKE,
615 tsk->clear_child_tid = NULL;
620 * Allocate a new mm structure and copy contents from the
621 * mm structure of the passed in task structure.
623 struct mm_struct *dup_mm(struct task_struct *tsk)
625 struct mm_struct *mm, *oldmm = current->mm;
635 memcpy(mm, oldmm, sizeof(*mm));
637 /* Initializing for Swap token stuff */
638 mm->token_priority = 0;
639 mm->last_interval = 0;
641 if (!mm_init(mm, tsk))
644 if (init_new_context(tsk, mm))
647 dup_mm_exe_file(oldmm, mm);
649 err = dup_mmap(mm, oldmm);
653 mm->hiwater_rss = get_mm_rss(mm);
654 mm->hiwater_vm = mm->total_vm;
656 if (mm->binfmt && !try_module_get(mm->binfmt->module))
662 /* don't put binfmt in mmput, we haven't got module yet */
671 * If init_new_context() failed, we cannot use mmput() to free the mm
672 * because it calls destroy_context()
679 static int copy_mm(unsigned long clone_flags, struct task_struct * tsk)
681 struct mm_struct * mm, *oldmm;
684 tsk->min_flt = tsk->maj_flt = 0;
685 tsk->nvcsw = tsk->nivcsw = 0;
686 #ifdef CONFIG_DETECT_HUNG_TASK
687 tsk->last_switch_count = tsk->nvcsw + tsk->nivcsw;
691 tsk->active_mm = NULL;
694 * Are we cloning a kernel thread?
696 * We need to steal a active VM for that..
702 if (clone_flags & CLONE_VM) {
703 atomic_inc(&oldmm->mm_users);
714 /* Initializing for Swap token stuff */
715 mm->token_priority = 0;
716 mm->last_interval = 0;
726 static int copy_fs(unsigned long clone_flags, struct task_struct *tsk)
728 struct fs_struct *fs = current->fs;
729 if (clone_flags & CLONE_FS) {
730 /* tsk->fs is already what we want */
731 write_lock(&fs->lock);
733 write_unlock(&fs->lock);
737 write_unlock(&fs->lock);
740 tsk->fs = copy_fs_struct(fs);
746 static int copy_files(unsigned long clone_flags, struct task_struct * tsk)
748 struct files_struct *oldf, *newf;
752 * A background process may not have any files ...
754 oldf = current->files;
758 if (clone_flags & CLONE_FILES) {
759 atomic_inc(&oldf->count);
763 newf = dup_fd(oldf, &error);
773 static int copy_io(unsigned long clone_flags, struct task_struct *tsk)
776 struct io_context *ioc = current->io_context;
781 * Share io context with parent, if CLONE_IO is set
783 if (clone_flags & CLONE_IO) {
784 tsk->io_context = ioc_task_link(ioc);
785 if (unlikely(!tsk->io_context))
787 } else if (ioprio_valid(ioc->ioprio)) {
788 tsk->io_context = alloc_io_context(GFP_KERNEL, -1);
789 if (unlikely(!tsk->io_context))
792 tsk->io_context->ioprio = ioc->ioprio;
798 static int copy_sighand(unsigned long clone_flags, struct task_struct *tsk)
800 struct sighand_struct *sig;
802 if (clone_flags & CLONE_SIGHAND) {
803 atomic_inc(¤t->sighand->count);
806 sig = kmem_cache_alloc(sighand_cachep, GFP_KERNEL);
807 rcu_assign_pointer(tsk->sighand, sig);
810 atomic_set(&sig->count, 1);
811 memcpy(sig->action, current->sighand->action, sizeof(sig->action));
815 void __cleanup_sighand(struct sighand_struct *sighand)
817 if (atomic_dec_and_test(&sighand->count))
818 kmem_cache_free(sighand_cachep, sighand);
823 * Initialize POSIX timer handling for a thread group.
825 static void posix_cpu_timers_init_group(struct signal_struct *sig)
827 /* Thread group counters. */
828 thread_group_cputime_init(sig);
830 /* Expiration times and increments. */
831 sig->it[CPUCLOCK_PROF].expires = cputime_zero;
832 sig->it[CPUCLOCK_PROF].incr = cputime_zero;
833 sig->it[CPUCLOCK_VIRT].expires = cputime_zero;
834 sig->it[CPUCLOCK_VIRT].incr = cputime_zero;
836 /* Cached expiration times. */
837 sig->cputime_expires.prof_exp = cputime_zero;
838 sig->cputime_expires.virt_exp = cputime_zero;
839 sig->cputime_expires.sched_exp = 0;
841 if (sig->rlim[RLIMIT_CPU].rlim_cur != RLIM_INFINITY) {
842 sig->cputime_expires.prof_exp =
843 secs_to_cputime(sig->rlim[RLIMIT_CPU].rlim_cur);
844 sig->cputimer.running = 1;
847 /* The timer lists. */
848 INIT_LIST_HEAD(&sig->cpu_timers[0]);
849 INIT_LIST_HEAD(&sig->cpu_timers[1]);
850 INIT_LIST_HEAD(&sig->cpu_timers[2]);
853 static int copy_signal(unsigned long clone_flags, struct task_struct *tsk)
855 struct signal_struct *sig;
857 if (clone_flags & CLONE_THREAD)
860 sig = kmem_cache_alloc(signal_cachep, GFP_KERNEL);
865 atomic_set(&sig->count, 1);
866 atomic_set(&sig->live, 1);
867 init_waitqueue_head(&sig->wait_chldexit);
869 if (clone_flags & CLONE_NEWPID)
870 sig->flags |= SIGNAL_UNKILLABLE;
871 sig->group_exit_code = 0;
872 sig->group_exit_task = NULL;
873 sig->group_stop_count = 0;
874 sig->curr_target = tsk;
875 init_sigpending(&sig->shared_pending);
876 INIT_LIST_HEAD(&sig->posix_timers);
878 hrtimer_init(&sig->real_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
879 sig->it_real_incr.tv64 = 0;
880 sig->real_timer.function = it_real_fn;
882 sig->leader = 0; /* session leadership doesn't inherit */
883 sig->tty_old_pgrp = NULL;
886 sig->utime = sig->stime = sig->cutime = sig->cstime = cputime_zero;
887 sig->gtime = cputime_zero;
888 sig->cgtime = cputime_zero;
889 #ifndef CONFIG_VIRT_CPU_ACCOUNTING
890 sig->prev_utime = sig->prev_stime = cputime_zero;
892 sig->nvcsw = sig->nivcsw = sig->cnvcsw = sig->cnivcsw = 0;
893 sig->min_flt = sig->maj_flt = sig->cmin_flt = sig->cmaj_flt = 0;
894 sig->inblock = sig->oublock = sig->cinblock = sig->coublock = 0;
895 sig->maxrss = sig->cmaxrss = 0;
896 task_io_accounting_init(&sig->ioac);
897 sig->sum_sched_runtime = 0;
898 taskstats_tgid_init(sig);
900 task_lock(current->group_leader);
901 memcpy(sig->rlim, current->signal->rlim, sizeof sig->rlim);
902 task_unlock(current->group_leader);
904 posix_cpu_timers_init_group(sig);
906 acct_init_pacct(&sig->pacct);
910 sig->oom_adj = current->signal->oom_adj;
915 void __cleanup_signal(struct signal_struct *sig)
917 thread_group_cputime_free(sig);
918 tty_kref_put(sig->tty);
919 kmem_cache_free(signal_cachep, sig);
922 static void copy_flags(unsigned long clone_flags, struct task_struct *p)
924 unsigned long new_flags = p->flags;
926 new_flags &= ~PF_SUPERPRIV;
927 new_flags |= PF_FORKNOEXEC;
928 new_flags |= PF_STARTING;
929 p->flags = new_flags;
930 clear_freeze_flag(p);
933 SYSCALL_DEFINE1(set_tid_address, int __user *, tidptr)
935 current->clear_child_tid = tidptr;
937 return task_pid_vnr(current);
940 static void rt_mutex_init_task(struct task_struct *p)
942 raw_spin_lock_init(&p->pi_lock);
943 #ifdef CONFIG_RT_MUTEXES
944 plist_head_init_raw(&p->pi_waiters, &p->pi_lock);
945 p->pi_blocked_on = NULL;
949 #ifdef CONFIG_MM_OWNER
950 void mm_init_owner(struct mm_struct *mm, struct task_struct *p)
954 #endif /* CONFIG_MM_OWNER */
957 * Initialize POSIX timer handling for a single task.
959 static void posix_cpu_timers_init(struct task_struct *tsk)
961 tsk->cputime_expires.prof_exp = cputime_zero;
962 tsk->cputime_expires.virt_exp = cputime_zero;
963 tsk->cputime_expires.sched_exp = 0;
964 INIT_LIST_HEAD(&tsk->cpu_timers[0]);
965 INIT_LIST_HEAD(&tsk->cpu_timers[1]);
966 INIT_LIST_HEAD(&tsk->cpu_timers[2]);
970 * This creates a new process as a copy of the old one,
971 * but does not actually start it yet.
973 * It copies the registers, and all the appropriate
974 * parts of the process environment (as per the clone
975 * flags). The actual kick-off is left to the caller.
977 static struct task_struct *copy_process(unsigned long clone_flags,
978 unsigned long stack_start,
979 struct pt_regs *regs,
980 unsigned long stack_size,
981 int __user *child_tidptr,
986 struct task_struct *p;
987 int cgroup_callbacks_done = 0;
989 if ((clone_flags & (CLONE_NEWNS|CLONE_FS)) == (CLONE_NEWNS|CLONE_FS))
990 return ERR_PTR(-EINVAL);
993 * Thread groups must share signals as well, and detached threads
994 * can only be started up within the thread group.
996 if ((clone_flags & CLONE_THREAD) && !(clone_flags & CLONE_SIGHAND))
997 return ERR_PTR(-EINVAL);
1000 * Shared signal handlers imply shared VM. By way of the above,
1001 * thread groups also imply shared VM. Blocking this case allows
1002 * for various simplifications in other code.
1004 if ((clone_flags & CLONE_SIGHAND) && !(clone_flags & CLONE_VM))
1005 return ERR_PTR(-EINVAL);
1008 * Siblings of global init remain as zombies on exit since they are
1009 * not reaped by their parent (swapper). To solve this and to avoid
1010 * multi-rooted process trees, prevent global and container-inits
1011 * from creating siblings.
1013 if ((clone_flags & CLONE_PARENT) &&
1014 current->signal->flags & SIGNAL_UNKILLABLE)
1015 return ERR_PTR(-EINVAL);
1017 retval = security_task_create(clone_flags);
1022 p = dup_task_struct(current);
1026 ftrace_graph_init_task(p);
1028 rt_mutex_init_task(p);
1030 #ifdef CONFIG_PROVE_LOCKING
1031 DEBUG_LOCKS_WARN_ON(!p->hardirqs_enabled);
1032 DEBUG_LOCKS_WARN_ON(!p->softirqs_enabled);
1035 if (atomic_read(&p->real_cred->user->processes) >=
1036 p->signal->rlim[RLIMIT_NPROC].rlim_cur) {
1037 if (!capable(CAP_SYS_ADMIN) && !capable(CAP_SYS_RESOURCE) &&
1038 p->real_cred->user != INIT_USER)
1042 retval = copy_creds(p, clone_flags);
1047 * If multiple threads are within copy_process(), then this check
1048 * triggers too late. This doesn't hurt, the check is only there
1049 * to stop root fork bombs.
1052 if (nr_threads >= max_threads)
1053 goto bad_fork_cleanup_count;
1055 if (!try_module_get(task_thread_info(p)->exec_domain->module))
1056 goto bad_fork_cleanup_count;
1059 delayacct_tsk_init(p); /* Must remain after dup_task_struct() */
1060 copy_flags(clone_flags, p);
1061 INIT_LIST_HEAD(&p->children);
1062 INIT_LIST_HEAD(&p->sibling);
1063 rcu_copy_process(p);
1064 p->vfork_done = NULL;
1065 spin_lock_init(&p->alloc_lock);
1067 init_sigpending(&p->pending);
1069 p->utime = cputime_zero;
1070 p->stime = cputime_zero;
1071 p->gtime = cputime_zero;
1072 p->utimescaled = cputime_zero;
1073 p->stimescaled = cputime_zero;
1074 #ifndef CONFIG_VIRT_CPU_ACCOUNTING
1075 p->prev_utime = cputime_zero;
1076 p->prev_stime = cputime_zero;
1079 p->default_timer_slack_ns = current->timer_slack_ns;
1081 task_io_accounting_init(&p->ioac);
1082 acct_clear_integrals(p);
1084 posix_cpu_timers_init(p);
1086 p->lock_depth = -1; /* -1 = no lock */
1087 do_posix_clock_monotonic_gettime(&p->start_time);
1088 p->real_start_time = p->start_time;
1089 monotonic_to_bootbased(&p->real_start_time);
1090 p->io_context = NULL;
1091 p->audit_context = NULL;
1094 p->mempolicy = mpol_dup(p->mempolicy);
1095 if (IS_ERR(p->mempolicy)) {
1096 retval = PTR_ERR(p->mempolicy);
1097 p->mempolicy = NULL;
1098 goto bad_fork_cleanup_cgroup;
1100 mpol_fix_fork_child_flag(p);
1102 #ifdef CONFIG_TRACE_IRQFLAGS
1104 #ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW
1105 p->hardirqs_enabled = 1;
1107 p->hardirqs_enabled = 0;
1109 p->hardirq_enable_ip = 0;
1110 p->hardirq_enable_event = 0;
1111 p->hardirq_disable_ip = _THIS_IP_;
1112 p->hardirq_disable_event = 0;
1113 p->softirqs_enabled = 1;
1114 p->softirq_enable_ip = _THIS_IP_;
1115 p->softirq_enable_event = 0;
1116 p->softirq_disable_ip = 0;
1117 p->softirq_disable_event = 0;
1118 p->hardirq_context = 0;
1119 p->softirq_context = 0;
1121 #ifdef CONFIG_LOCKDEP
1122 p->lockdep_depth = 0; /* no locks held yet */
1123 p->curr_chain_key = 0;
1124 p->lockdep_recursion = 0;
1127 #ifdef CONFIG_DEBUG_MUTEXES
1128 p->blocked_on = NULL; /* not blocked yet */
1130 #ifdef CONFIG_CGROUP_MEM_RES_CTLR
1131 p->memcg_batch.do_batch = 0;
1132 p->memcg_batch.memcg = NULL;
1137 p->stack_start = stack_start;
1139 /* Perform scheduler related setup. Assign this task to a CPU. */
1140 sched_fork(p, clone_flags);
1142 retval = perf_event_init_task(p);
1144 goto bad_fork_cleanup_policy;
1146 if ((retval = audit_alloc(p)))
1147 goto bad_fork_cleanup_policy;
1148 /* copy all the process information */
1149 if ((retval = copy_semundo(clone_flags, p)))
1150 goto bad_fork_cleanup_audit;
1151 if ((retval = copy_files(clone_flags, p)))
1152 goto bad_fork_cleanup_semundo;
1153 if ((retval = copy_fs(clone_flags, p)))
1154 goto bad_fork_cleanup_files;
1155 if ((retval = copy_sighand(clone_flags, p)))
1156 goto bad_fork_cleanup_fs;
1157 if ((retval = copy_signal(clone_flags, p)))
1158 goto bad_fork_cleanup_sighand;
1159 if ((retval = copy_mm(clone_flags, p)))
1160 goto bad_fork_cleanup_signal;
1161 if ((retval = copy_namespaces(clone_flags, p)))
1162 goto bad_fork_cleanup_mm;
1163 if ((retval = copy_io(clone_flags, p)))
1164 goto bad_fork_cleanup_namespaces;
1165 retval = copy_thread(clone_flags, stack_start, stack_size, p, regs);
1167 goto bad_fork_cleanup_io;
1169 if (pid != &init_struct_pid) {
1171 pid = alloc_pid(p->nsproxy->pid_ns);
1173 goto bad_fork_cleanup_io;
1175 if (clone_flags & CLONE_NEWPID) {
1176 retval = pid_ns_prepare_proc(p->nsproxy->pid_ns);
1178 goto bad_fork_free_pid;
1182 p->pid = pid_nr(pid);
1184 if (clone_flags & CLONE_THREAD)
1185 p->tgid = current->tgid;
1187 if (current->nsproxy != p->nsproxy) {
1188 retval = ns_cgroup_clone(p, pid);
1190 goto bad_fork_free_pid;
1193 p->set_child_tid = (clone_flags & CLONE_CHILD_SETTID) ? child_tidptr : NULL;
1195 * Clear TID on mm_release()?
1197 p->clear_child_tid = (clone_flags & CLONE_CHILD_CLEARTID) ? child_tidptr: NULL;
1199 p->robust_list = NULL;
1200 #ifdef CONFIG_COMPAT
1201 p->compat_robust_list = NULL;
1203 INIT_LIST_HEAD(&p->pi_state_list);
1204 p->pi_state_cache = NULL;
1207 * sigaltstack should be cleared when sharing the same VM
1209 if ((clone_flags & (CLONE_VM|CLONE_VFORK)) == CLONE_VM)
1210 p->sas_ss_sp = p->sas_ss_size = 0;
1213 * Syscall tracing and stepping should be turned off in the
1214 * child regardless of CLONE_PTRACE.
1216 user_disable_single_step(p);
1217 clear_tsk_thread_flag(p, TIF_SYSCALL_TRACE);
1218 #ifdef TIF_SYSCALL_EMU
1219 clear_tsk_thread_flag(p, TIF_SYSCALL_EMU);
1221 clear_all_latency_tracing(p);
1223 /* ok, now we should be set up.. */
1224 p->exit_signal = (clone_flags & CLONE_THREAD) ? -1 : (clone_flags & CSIGNAL);
1225 p->pdeath_signal = 0;
1229 * Ok, make it visible to the rest of the system.
1230 * We dont wake it up yet.
1232 p->group_leader = p;
1233 INIT_LIST_HEAD(&p->thread_group);
1235 /* Now that the task is set up, run cgroup callbacks if
1236 * necessary. We need to run them before the task is visible
1237 * on the tasklist. */
1238 cgroup_fork_callbacks(p);
1239 cgroup_callbacks_done = 1;
1241 /* Need tasklist lock for parent etc handling! */
1242 write_lock_irq(&tasklist_lock);
1244 /* CLONE_PARENT re-uses the old parent */
1245 if (clone_flags & (CLONE_PARENT|CLONE_THREAD)) {
1246 p->real_parent = current->real_parent;
1247 p->parent_exec_id = current->parent_exec_id;
1249 p->real_parent = current;
1250 p->parent_exec_id = current->self_exec_id;
1253 spin_lock(¤t->sighand->siglock);
1256 * Process group and session signals need to be delivered to just the
1257 * parent before the fork or both the parent and the child after the
1258 * fork. Restart if a signal comes in before we add the new process to
1259 * it's process group.
1260 * A fatal signal pending means that current will exit, so the new
1261 * thread can't slip out of an OOM kill (or normal SIGKILL).
1263 recalc_sigpending();
1264 if (signal_pending(current)) {
1265 spin_unlock(¤t->sighand->siglock);
1266 write_unlock_irq(&tasklist_lock);
1267 retval = -ERESTARTNOINTR;
1268 goto bad_fork_free_pid;
1271 if (clone_flags & CLONE_THREAD) {
1272 atomic_inc(¤t->signal->count);
1273 atomic_inc(¤t->signal->live);
1274 p->group_leader = current->group_leader;
1275 list_add_tail_rcu(&p->thread_group, &p->group_leader->thread_group);
1278 if (likely(p->pid)) {
1279 tracehook_finish_clone(p, clone_flags, trace);
1281 if (thread_group_leader(p)) {
1282 if (clone_flags & CLONE_NEWPID)
1283 p->nsproxy->pid_ns->child_reaper = p;
1285 p->signal->leader_pid = pid;
1286 tty_kref_put(p->signal->tty);
1287 p->signal->tty = tty_kref_get(current->signal->tty);
1288 attach_pid(p, PIDTYPE_PGID, task_pgrp(current));
1289 attach_pid(p, PIDTYPE_SID, task_session(current));
1290 list_add_tail(&p->sibling, &p->real_parent->children);
1291 list_add_tail_rcu(&p->tasks, &init_task.tasks);
1292 __get_cpu_var(process_counts)++;
1294 attach_pid(p, PIDTYPE_PID, pid);
1299 spin_unlock(¤t->sighand->siglock);
1300 write_unlock_irq(&tasklist_lock);
1301 proc_fork_connector(p);
1302 cgroup_post_fork(p);
1307 if (pid != &init_struct_pid)
1309 bad_fork_cleanup_io:
1312 bad_fork_cleanup_namespaces:
1313 exit_task_namespaces(p);
1314 bad_fork_cleanup_mm:
1317 bad_fork_cleanup_signal:
1318 if (!(clone_flags & CLONE_THREAD))
1319 __cleanup_signal(p->signal);
1320 bad_fork_cleanup_sighand:
1321 __cleanup_sighand(p->sighand);
1322 bad_fork_cleanup_fs:
1323 exit_fs(p); /* blocking */
1324 bad_fork_cleanup_files:
1325 exit_files(p); /* blocking */
1326 bad_fork_cleanup_semundo:
1328 bad_fork_cleanup_audit:
1330 bad_fork_cleanup_policy:
1331 perf_event_free_task(p);
1333 mpol_put(p->mempolicy);
1334 bad_fork_cleanup_cgroup:
1336 cgroup_exit(p, cgroup_callbacks_done);
1337 delayacct_tsk_free(p);
1338 module_put(task_thread_info(p)->exec_domain->module);
1339 bad_fork_cleanup_count:
1340 atomic_dec(&p->cred->user->processes);
1345 return ERR_PTR(retval);
1348 noinline struct pt_regs * __cpuinit __attribute__((weak)) idle_regs(struct pt_regs *regs)
1350 memset(regs, 0, sizeof(struct pt_regs));
1354 struct task_struct * __cpuinit fork_idle(int cpu)
1356 struct task_struct *task;
1357 struct pt_regs regs;
1359 task = copy_process(CLONE_VM, 0, idle_regs(®s), 0, NULL,
1360 &init_struct_pid, 0);
1362 init_idle(task, cpu);
1368 * Ok, this is the main fork-routine.
1370 * It copies the process, and if successful kick-starts
1371 * it and waits for it to finish using the VM if required.
1373 long do_fork(unsigned long clone_flags,
1374 unsigned long stack_start,
1375 struct pt_regs *regs,
1376 unsigned long stack_size,
1377 int __user *parent_tidptr,
1378 int __user *child_tidptr)
1380 struct task_struct *p;
1385 * Do some preliminary argument and permissions checking before we
1386 * actually start allocating stuff
1388 if (clone_flags & CLONE_NEWUSER) {
1389 if (clone_flags & CLONE_THREAD)
1391 /* hopefully this check will go away when userns support is
1394 if (!capable(CAP_SYS_ADMIN) || !capable(CAP_SETUID) ||
1395 !capable(CAP_SETGID))
1400 * We hope to recycle these flags after 2.6.26
1402 if (unlikely(clone_flags & CLONE_STOPPED)) {
1403 static int __read_mostly count = 100;
1405 if (count > 0 && printk_ratelimit()) {
1406 char comm[TASK_COMM_LEN];
1409 printk(KERN_INFO "fork(): process `%s' used deprecated "
1410 "clone flags 0x%lx\n",
1411 get_task_comm(comm, current),
1412 clone_flags & CLONE_STOPPED);
1417 * When called from kernel_thread, don't do user tracing stuff.
1419 if (likely(user_mode(regs)))
1420 trace = tracehook_prepare_clone(clone_flags);
1422 p = copy_process(clone_flags, stack_start, regs, stack_size,
1423 child_tidptr, NULL, trace);
1425 * Do this prior waking up the new thread - the thread pointer
1426 * might get invalid after that point, if the thread exits quickly.
1429 struct completion vfork;
1431 trace_sched_process_fork(current, p);
1433 nr = task_pid_vnr(p);
1435 if (clone_flags & CLONE_PARENT_SETTID)
1436 put_user(nr, parent_tidptr);
1438 if (clone_flags & CLONE_VFORK) {
1439 p->vfork_done = &vfork;
1440 init_completion(&vfork);
1443 audit_finish_fork(p);
1444 tracehook_report_clone(regs, clone_flags, nr, p);
1447 * We set PF_STARTING at creation in case tracing wants to
1448 * use this to distinguish a fully live task from one that
1449 * hasn't gotten to tracehook_report_clone() yet. Now we
1450 * clear it and set the child going.
1452 p->flags &= ~PF_STARTING;
1454 if (unlikely(clone_flags & CLONE_STOPPED)) {
1456 * We'll start up with an immediate SIGSTOP.
1458 sigaddset(&p->pending.signal, SIGSTOP);
1459 set_tsk_thread_flag(p, TIF_SIGPENDING);
1460 __set_task_state(p, TASK_STOPPED);
1462 wake_up_new_task(p, clone_flags);
1465 tracehook_report_clone_complete(trace, regs,
1466 clone_flags, nr, p);
1468 if (clone_flags & CLONE_VFORK) {
1469 freezer_do_not_count();
1470 wait_for_completion(&vfork);
1472 tracehook_report_vfork_done(p, nr);
1480 #ifndef ARCH_MIN_MMSTRUCT_ALIGN
1481 #define ARCH_MIN_MMSTRUCT_ALIGN 0
1484 static void sighand_ctor(void *data)
1486 struct sighand_struct *sighand = data;
1488 spin_lock_init(&sighand->siglock);
1489 init_waitqueue_head(&sighand->signalfd_wqh);
1492 void __init proc_caches_init(void)
1494 sighand_cachep = kmem_cache_create("sighand_cache",
1495 sizeof(struct sighand_struct), 0,
1496 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_DESTROY_BY_RCU|
1497 SLAB_NOTRACK, sighand_ctor);
1498 signal_cachep = kmem_cache_create("signal_cache",
1499 sizeof(struct signal_struct), 0,
1500 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL);
1501 files_cachep = kmem_cache_create("files_cache",
1502 sizeof(struct files_struct), 0,
1503 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL);
1504 fs_cachep = kmem_cache_create("fs_cache",
1505 sizeof(struct fs_struct), 0,
1506 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL);
1507 mm_cachep = kmem_cache_create("mm_struct",
1508 sizeof(struct mm_struct), ARCH_MIN_MMSTRUCT_ALIGN,
1509 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL);
1510 vm_area_cachep = KMEM_CACHE(vm_area_struct, SLAB_PANIC);
1515 * Check constraints on flags passed to the unshare system call and
1516 * force unsharing of additional process context as appropriate.
1518 static void check_unshare_flags(unsigned long *flags_ptr)
1521 * If unsharing a thread from a thread group, must also
1524 if (*flags_ptr & CLONE_THREAD)
1525 *flags_ptr |= CLONE_VM;
1528 * If unsharing vm, must also unshare signal handlers.
1530 if (*flags_ptr & CLONE_VM)
1531 *flags_ptr |= CLONE_SIGHAND;
1534 * If unsharing signal handlers and the task was created
1535 * using CLONE_THREAD, then must unshare the thread
1537 if ((*flags_ptr & CLONE_SIGHAND) &&
1538 (atomic_read(¤t->signal->count) > 1))
1539 *flags_ptr |= CLONE_THREAD;
1542 * If unsharing namespace, must also unshare filesystem information.
1544 if (*flags_ptr & CLONE_NEWNS)
1545 *flags_ptr |= CLONE_FS;
1549 * Unsharing of tasks created with CLONE_THREAD is not supported yet
1551 static int unshare_thread(unsigned long unshare_flags)
1553 if (unshare_flags & CLONE_THREAD)
1560 * Unshare the filesystem structure if it is being shared
1562 static int unshare_fs(unsigned long unshare_flags, struct fs_struct **new_fsp)
1564 struct fs_struct *fs = current->fs;
1566 if (!(unshare_flags & CLONE_FS) || !fs)
1569 /* don't need lock here; in the worst case we'll do useless copy */
1573 *new_fsp = copy_fs_struct(fs);
1581 * Unsharing of sighand is not supported yet
1583 static int unshare_sighand(unsigned long unshare_flags, struct sighand_struct **new_sighp)
1585 struct sighand_struct *sigh = current->sighand;
1587 if ((unshare_flags & CLONE_SIGHAND) && atomic_read(&sigh->count) > 1)
1594 * Unshare vm if it is being shared
1596 static int unshare_vm(unsigned long unshare_flags, struct mm_struct **new_mmp)
1598 struct mm_struct *mm = current->mm;
1600 if ((unshare_flags & CLONE_VM) &&
1601 (mm && atomic_read(&mm->mm_users) > 1)) {
1609 * Unshare file descriptor table if it is being shared
1611 static int unshare_fd(unsigned long unshare_flags, struct files_struct **new_fdp)
1613 struct files_struct *fd = current->files;
1616 if ((unshare_flags & CLONE_FILES) &&
1617 (fd && atomic_read(&fd->count) > 1)) {
1618 *new_fdp = dup_fd(fd, &error);
1627 * unshare allows a process to 'unshare' part of the process
1628 * context which was originally shared using clone. copy_*
1629 * functions used by do_fork() cannot be used here directly
1630 * because they modify an inactive task_struct that is being
1631 * constructed. Here we are modifying the current, active,
1634 SYSCALL_DEFINE1(unshare, unsigned long, unshare_flags)
1637 struct fs_struct *fs, *new_fs = NULL;
1638 struct sighand_struct *new_sigh = NULL;
1639 struct mm_struct *mm, *new_mm = NULL, *active_mm = NULL;
1640 struct files_struct *fd, *new_fd = NULL;
1641 struct nsproxy *new_nsproxy = NULL;
1644 check_unshare_flags(&unshare_flags);
1646 /* Return -EINVAL for all unsupported flags */
1648 if (unshare_flags & ~(CLONE_THREAD|CLONE_FS|CLONE_NEWNS|CLONE_SIGHAND|
1649 CLONE_VM|CLONE_FILES|CLONE_SYSVSEM|
1650 CLONE_NEWUTS|CLONE_NEWIPC|CLONE_NEWNET))
1651 goto bad_unshare_out;
1654 * CLONE_NEWIPC must also detach from the undolist: after switching
1655 * to a new ipc namespace, the semaphore arrays from the old
1656 * namespace are unreachable.
1658 if (unshare_flags & (CLONE_NEWIPC|CLONE_SYSVSEM))
1660 if ((err = unshare_thread(unshare_flags)))
1661 goto bad_unshare_out;
1662 if ((err = unshare_fs(unshare_flags, &new_fs)))
1663 goto bad_unshare_cleanup_thread;
1664 if ((err = unshare_sighand(unshare_flags, &new_sigh)))
1665 goto bad_unshare_cleanup_fs;
1666 if ((err = unshare_vm(unshare_flags, &new_mm)))
1667 goto bad_unshare_cleanup_sigh;
1668 if ((err = unshare_fd(unshare_flags, &new_fd)))
1669 goto bad_unshare_cleanup_vm;
1670 if ((err = unshare_nsproxy_namespaces(unshare_flags, &new_nsproxy,
1672 goto bad_unshare_cleanup_fd;
1674 if (new_fs || new_mm || new_fd || do_sysvsem || new_nsproxy) {
1677 * CLONE_SYSVSEM is equivalent to sys_exit().
1683 switch_task_namespaces(current, new_nsproxy);
1691 write_lock(&fs->lock);
1692 current->fs = new_fs;
1697 write_unlock(&fs->lock);
1702 active_mm = current->active_mm;
1703 current->mm = new_mm;
1704 current->active_mm = new_mm;
1705 activate_mm(active_mm, new_mm);
1710 fd = current->files;
1711 current->files = new_fd;
1715 task_unlock(current);
1719 put_nsproxy(new_nsproxy);
1721 bad_unshare_cleanup_fd:
1723 put_files_struct(new_fd);
1725 bad_unshare_cleanup_vm:
1729 bad_unshare_cleanup_sigh:
1731 if (atomic_dec_and_test(&new_sigh->count))
1732 kmem_cache_free(sighand_cachep, new_sigh);
1734 bad_unshare_cleanup_fs:
1736 free_fs_struct(new_fs);
1738 bad_unshare_cleanup_thread:
1744 * Helper to unshare the files of the current task.
1745 * We don't want to expose copy_files internals to
1746 * the exec layer of the kernel.
1749 int unshare_files(struct files_struct **displaced)
1751 struct task_struct *task = current;
1752 struct files_struct *copy = NULL;
1755 error = unshare_fd(CLONE_FILES, ©);
1756 if (error || !copy) {
1760 *displaced = task->files;