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 */
90 int nr_processes(void)
95 for_each_online_cpu(cpu)
96 total += per_cpu(process_counts, cpu);
101 #ifndef __HAVE_ARCH_TASK_STRUCT_ALLOCATOR
102 # define alloc_task_struct() kmem_cache_alloc(task_struct_cachep, GFP_KERNEL)
103 # define free_task_struct(tsk) kmem_cache_free(task_struct_cachep, (tsk))
104 static struct kmem_cache *task_struct_cachep;
107 #ifndef __HAVE_ARCH_THREAD_INFO_ALLOCATOR
108 static inline struct thread_info *alloc_thread_info(struct task_struct *tsk)
110 #ifdef CONFIG_DEBUG_STACK_USAGE
111 gfp_t mask = GFP_KERNEL | __GFP_ZERO;
113 gfp_t mask = GFP_KERNEL;
115 return (struct thread_info *)__get_free_pages(mask, THREAD_SIZE_ORDER);
118 static inline void free_thread_info(struct thread_info *ti)
120 free_pages((unsigned long)ti, THREAD_SIZE_ORDER);
124 /* SLAB cache for signal_struct structures (tsk->signal) */
125 static struct kmem_cache *signal_cachep;
127 /* SLAB cache for sighand_struct structures (tsk->sighand) */
128 struct kmem_cache *sighand_cachep;
130 /* SLAB cache for files_struct structures (tsk->files) */
131 struct kmem_cache *files_cachep;
133 /* SLAB cache for fs_struct structures (tsk->fs) */
134 struct kmem_cache *fs_cachep;
136 /* SLAB cache for vm_area_struct structures */
137 struct kmem_cache *vm_area_cachep;
139 /* SLAB cache for mm_struct structures (tsk->mm) */
140 static struct kmem_cache *mm_cachep;
142 static void account_kernel_stack(struct thread_info *ti, int account)
144 struct zone *zone = page_zone(virt_to_page(ti));
146 mod_zone_page_state(zone, NR_KERNEL_STACK, account);
149 void free_task(struct task_struct *tsk)
151 prop_local_destroy_single(&tsk->dirties);
152 account_kernel_stack(tsk->stack, -1);
153 free_thread_info(tsk->stack);
154 rt_mutex_debug_task_free(tsk);
155 ftrace_graph_exit_task(tsk);
156 free_task_struct(tsk);
158 EXPORT_SYMBOL(free_task);
160 void __put_task_struct(struct task_struct *tsk)
162 WARN_ON(!tsk->exit_state);
163 WARN_ON(atomic_read(&tsk->usage));
164 WARN_ON(tsk == current);
167 delayacct_tsk_free(tsk);
169 if (!profile_handoff_task(tsk))
174 * macro override instead of weak attribute alias, to workaround
175 * gcc 4.1.0 and 4.1.1 bugs with weak attribute and empty functions.
177 #ifndef arch_task_cache_init
178 #define arch_task_cache_init()
181 void __init fork_init(unsigned long mempages)
183 #ifndef __HAVE_ARCH_TASK_STRUCT_ALLOCATOR
184 #ifndef ARCH_MIN_TASKALIGN
185 #define ARCH_MIN_TASKALIGN L1_CACHE_BYTES
187 /* create a slab on which task_structs can be allocated */
189 kmem_cache_create("task_struct", sizeof(struct task_struct),
190 ARCH_MIN_TASKALIGN, SLAB_PANIC | SLAB_NOTRACK, NULL);
193 /* do the arch specific task caches init */
194 arch_task_cache_init();
197 * The default maximum number of threads is set to a safe
198 * value: the thread structures can take up at most half
201 max_threads = mempages / (8 * THREAD_SIZE / PAGE_SIZE);
204 * we need to allow at least 20 threads to boot a system
209 init_task.signal->rlim[RLIMIT_NPROC].rlim_cur = max_threads/2;
210 init_task.signal->rlim[RLIMIT_NPROC].rlim_max = max_threads/2;
211 init_task.signal->rlim[RLIMIT_SIGPENDING] =
212 init_task.signal->rlim[RLIMIT_NPROC];
215 int __attribute__((weak)) arch_dup_task_struct(struct task_struct *dst,
216 struct task_struct *src)
222 static struct task_struct *dup_task_struct(struct task_struct *orig)
224 struct task_struct *tsk;
225 struct thread_info *ti;
226 unsigned long *stackend;
230 prepare_to_copy(orig);
232 tsk = alloc_task_struct();
236 ti = alloc_thread_info(tsk);
238 free_task_struct(tsk);
242 err = arch_dup_task_struct(tsk, orig);
248 err = prop_local_init_single(&tsk->dirties);
252 setup_thread_stack(tsk, orig);
253 clear_user_return_notifier(tsk);
254 stackend = end_of_stack(tsk);
255 *stackend = STACK_END_MAGIC; /* for overflow detection */
257 #ifdef CONFIG_CC_STACKPROTECTOR
258 tsk->stack_canary = get_random_int();
261 /* One for us, one for whoever does the "release_task()" (usually parent) */
262 atomic_set(&tsk->usage,2);
263 atomic_set(&tsk->fs_excl, 0);
264 #ifdef CONFIG_BLK_DEV_IO_TRACE
267 tsk->splice_pipe = NULL;
269 account_kernel_stack(ti, 1);
274 free_thread_info(ti);
275 free_task_struct(tsk);
280 static int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm)
282 struct vm_area_struct *mpnt, *tmp, **pprev;
283 struct rb_node **rb_link, *rb_parent;
285 unsigned long charge;
286 struct mempolicy *pol;
288 down_write(&oldmm->mmap_sem);
289 flush_cache_dup_mm(oldmm);
291 * Not linked in yet - no deadlock potential:
293 down_write_nested(&mm->mmap_sem, SINGLE_DEPTH_NESTING);
297 mm->mmap_cache = NULL;
298 mm->free_area_cache = oldmm->mmap_base;
299 mm->cached_hole_size = ~0UL;
301 cpumask_clear(mm_cpumask(mm));
303 rb_link = &mm->mm_rb.rb_node;
306 retval = ksm_fork(mm, oldmm);
310 for (mpnt = oldmm->mmap; mpnt; mpnt = mpnt->vm_next) {
313 if (mpnt->vm_flags & VM_DONTCOPY) {
314 long pages = vma_pages(mpnt);
315 mm->total_vm -= pages;
316 vm_stat_account(mm, mpnt->vm_flags, mpnt->vm_file,
321 if (mpnt->vm_flags & VM_ACCOUNT) {
322 unsigned int len = (mpnt->vm_end - mpnt->vm_start) >> PAGE_SHIFT;
323 if (security_vm_enough_memory(len))
327 tmp = kmem_cache_alloc(vm_area_cachep, GFP_KERNEL);
331 pol = mpol_dup(vma_policy(mpnt));
332 retval = PTR_ERR(pol);
334 goto fail_nomem_policy;
335 vma_set_policy(tmp, pol);
336 tmp->vm_flags &= ~VM_LOCKED;
342 struct inode *inode = file->f_path.dentry->d_inode;
343 struct address_space *mapping = file->f_mapping;
346 if (tmp->vm_flags & VM_DENYWRITE)
347 atomic_dec(&inode->i_writecount);
348 spin_lock(&mapping->i_mmap_lock);
349 if (tmp->vm_flags & VM_SHARED)
350 mapping->i_mmap_writable++;
351 tmp->vm_truncate_count = mpnt->vm_truncate_count;
352 flush_dcache_mmap_lock(mapping);
353 /* insert tmp into the share list, just after mpnt */
354 vma_prio_tree_add(tmp, mpnt);
355 flush_dcache_mmap_unlock(mapping);
356 spin_unlock(&mapping->i_mmap_lock);
360 * Clear hugetlb-related page reserves for children. This only
361 * affects MAP_PRIVATE mappings. Faults generated by the child
362 * are not guaranteed to succeed, even if read-only
364 if (is_vm_hugetlb_page(tmp))
365 reset_vma_resv_huge_pages(tmp);
368 * Link in the new vma and copy the page table entries.
371 pprev = &tmp->vm_next;
373 __vma_link_rb(mm, tmp, rb_link, rb_parent);
374 rb_link = &tmp->vm_rb.rb_right;
375 rb_parent = &tmp->vm_rb;
378 retval = copy_page_range(mm, oldmm, mpnt);
380 if (tmp->vm_ops && tmp->vm_ops->open)
381 tmp->vm_ops->open(tmp);
386 /* a new mm has just been created */
387 arch_dup_mmap(oldmm, mm);
390 up_write(&mm->mmap_sem);
392 up_write(&oldmm->mmap_sem);
395 kmem_cache_free(vm_area_cachep, tmp);
398 vm_unacct_memory(charge);
402 static inline int mm_alloc_pgd(struct mm_struct * mm)
404 mm->pgd = pgd_alloc(mm);
405 if (unlikely(!mm->pgd))
410 static inline void mm_free_pgd(struct mm_struct * mm)
412 pgd_free(mm, mm->pgd);
415 #define dup_mmap(mm, oldmm) (0)
416 #define mm_alloc_pgd(mm) (0)
417 #define mm_free_pgd(mm)
418 #endif /* CONFIG_MMU */
420 __cacheline_aligned_in_smp DEFINE_SPINLOCK(mmlist_lock);
422 #define allocate_mm() (kmem_cache_alloc(mm_cachep, GFP_KERNEL))
423 #define free_mm(mm) (kmem_cache_free(mm_cachep, (mm)))
425 static unsigned long default_dump_filter = MMF_DUMP_FILTER_DEFAULT;
427 static int __init coredump_filter_setup(char *s)
429 default_dump_filter =
430 (simple_strtoul(s, NULL, 0) << MMF_DUMP_FILTER_SHIFT) &
431 MMF_DUMP_FILTER_MASK;
435 __setup("coredump_filter=", coredump_filter_setup);
437 #include <linux/init_task.h>
439 static void mm_init_aio(struct mm_struct *mm)
442 spin_lock_init(&mm->ioctx_lock);
443 INIT_HLIST_HEAD(&mm->ioctx_list);
447 static struct mm_struct * mm_init(struct mm_struct * mm, struct task_struct *p)
449 atomic_set(&mm->mm_users, 1);
450 atomic_set(&mm->mm_count, 1);
451 init_rwsem(&mm->mmap_sem);
452 INIT_LIST_HEAD(&mm->mmlist);
453 mm->flags = (current->mm) ?
454 (current->mm->flags & MMF_INIT_MASK) : default_dump_filter;
455 mm->core_state = NULL;
457 set_mm_counter(mm, file_rss, 0);
458 set_mm_counter(mm, anon_rss, 0);
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);
578 if (unlikely(tsk->compat_robust_list))
579 compat_exit_robust_list(tsk);
583 /* Get rid of any cached register state */
584 deactivate_mm(tsk, mm);
586 /* notify parent sleeping on vfork() */
588 tsk->vfork_done = NULL;
589 complete(vfork_done);
593 * If we're exiting normally, clear a user-space tid field if
594 * requested. We leave this alone when dying by signal, to leave
595 * the value intact in a core dump, and to save the unnecessary
596 * trouble otherwise. Userland only wants this done for a sys_exit.
598 if (tsk->clear_child_tid) {
599 if (!(tsk->flags & PF_SIGNALED) &&
600 atomic_read(&mm->mm_users) > 1) {
602 * We don't check the error code - if userspace has
603 * not set up a proper pointer then tough luck.
605 put_user(0, tsk->clear_child_tid);
606 sys_futex(tsk->clear_child_tid, FUTEX_WAKE,
609 tsk->clear_child_tid = NULL;
614 * Allocate a new mm structure and copy contents from the
615 * mm structure of the passed in task structure.
617 struct mm_struct *dup_mm(struct task_struct *tsk)
619 struct mm_struct *mm, *oldmm = current->mm;
629 memcpy(mm, oldmm, sizeof(*mm));
631 /* Initializing for Swap token stuff */
632 mm->token_priority = 0;
633 mm->last_interval = 0;
635 if (!mm_init(mm, tsk))
638 if (init_new_context(tsk, mm))
641 dup_mm_exe_file(oldmm, mm);
643 err = dup_mmap(mm, oldmm);
647 mm->hiwater_rss = get_mm_rss(mm);
648 mm->hiwater_vm = mm->total_vm;
650 if (mm->binfmt && !try_module_get(mm->binfmt->module))
656 /* don't put binfmt in mmput, we haven't got module yet */
665 * If init_new_context() failed, we cannot use mmput() to free the mm
666 * because it calls destroy_context()
673 static int copy_mm(unsigned long clone_flags, struct task_struct * tsk)
675 struct mm_struct * mm, *oldmm;
678 tsk->min_flt = tsk->maj_flt = 0;
679 tsk->nvcsw = tsk->nivcsw = 0;
680 #ifdef CONFIG_DETECT_HUNG_TASK
681 tsk->last_switch_count = tsk->nvcsw + tsk->nivcsw;
685 tsk->active_mm = NULL;
688 * Are we cloning a kernel thread?
690 * We need to steal a active VM for that..
696 if (clone_flags & CLONE_VM) {
697 atomic_inc(&oldmm->mm_users);
708 /* Initializing for Swap token stuff */
709 mm->token_priority = 0;
710 mm->last_interval = 0;
720 static int copy_fs(unsigned long clone_flags, struct task_struct *tsk)
722 struct fs_struct *fs = current->fs;
723 if (clone_flags & CLONE_FS) {
724 /* tsk->fs is already what we want */
725 write_lock(&fs->lock);
727 write_unlock(&fs->lock);
731 write_unlock(&fs->lock);
734 tsk->fs = copy_fs_struct(fs);
740 static int copy_files(unsigned long clone_flags, struct task_struct * tsk)
742 struct files_struct *oldf, *newf;
746 * A background process may not have any files ...
748 oldf = current->files;
752 if (clone_flags & CLONE_FILES) {
753 atomic_inc(&oldf->count);
757 newf = dup_fd(oldf, &error);
767 static int copy_io(unsigned long clone_flags, struct task_struct *tsk)
770 struct io_context *ioc = current->io_context;
775 * Share io context with parent, if CLONE_IO is set
777 if (clone_flags & CLONE_IO) {
778 tsk->io_context = ioc_task_link(ioc);
779 if (unlikely(!tsk->io_context))
781 } else if (ioprio_valid(ioc->ioprio)) {
782 tsk->io_context = alloc_io_context(GFP_KERNEL, -1);
783 if (unlikely(!tsk->io_context))
786 tsk->io_context->ioprio = ioc->ioprio;
792 static int copy_sighand(unsigned long clone_flags, struct task_struct *tsk)
794 struct sighand_struct *sig;
796 if (clone_flags & CLONE_SIGHAND) {
797 atomic_inc(¤t->sighand->count);
800 sig = kmem_cache_alloc(sighand_cachep, GFP_KERNEL);
801 rcu_assign_pointer(tsk->sighand, sig);
804 atomic_set(&sig->count, 1);
805 memcpy(sig->action, current->sighand->action, sizeof(sig->action));
809 void __cleanup_sighand(struct sighand_struct *sighand)
811 if (atomic_dec_and_test(&sighand->count))
812 kmem_cache_free(sighand_cachep, sighand);
817 * Initialize POSIX timer handling for a thread group.
819 static void posix_cpu_timers_init_group(struct signal_struct *sig)
821 /* Thread group counters. */
822 thread_group_cputime_init(sig);
824 /* Expiration times and increments. */
825 sig->it[CPUCLOCK_PROF].expires = cputime_zero;
826 sig->it[CPUCLOCK_PROF].incr = cputime_zero;
827 sig->it[CPUCLOCK_VIRT].expires = cputime_zero;
828 sig->it[CPUCLOCK_VIRT].incr = cputime_zero;
830 /* Cached expiration times. */
831 sig->cputime_expires.prof_exp = cputime_zero;
832 sig->cputime_expires.virt_exp = cputime_zero;
833 sig->cputime_expires.sched_exp = 0;
835 if (sig->rlim[RLIMIT_CPU].rlim_cur != RLIM_INFINITY) {
836 sig->cputime_expires.prof_exp =
837 secs_to_cputime(sig->rlim[RLIMIT_CPU].rlim_cur);
838 sig->cputimer.running = 1;
841 /* The timer lists. */
842 INIT_LIST_HEAD(&sig->cpu_timers[0]);
843 INIT_LIST_HEAD(&sig->cpu_timers[1]);
844 INIT_LIST_HEAD(&sig->cpu_timers[2]);
847 static int copy_signal(unsigned long clone_flags, struct task_struct *tsk)
849 struct signal_struct *sig;
851 if (clone_flags & CLONE_THREAD)
854 sig = kmem_cache_alloc(signal_cachep, GFP_KERNEL);
859 atomic_set(&sig->count, 1);
860 atomic_set(&sig->live, 1);
861 init_waitqueue_head(&sig->wait_chldexit);
863 if (clone_flags & CLONE_NEWPID)
864 sig->flags |= SIGNAL_UNKILLABLE;
865 sig->group_exit_code = 0;
866 sig->group_exit_task = NULL;
867 sig->group_stop_count = 0;
868 sig->curr_target = tsk;
869 init_sigpending(&sig->shared_pending);
870 INIT_LIST_HEAD(&sig->posix_timers);
872 hrtimer_init(&sig->real_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
873 sig->it_real_incr.tv64 = 0;
874 sig->real_timer.function = it_real_fn;
876 sig->leader = 0; /* session leadership doesn't inherit */
877 sig->tty_old_pgrp = NULL;
880 sig->utime = sig->stime = sig->cutime = sig->cstime = cputime_zero;
881 sig->gtime = cputime_zero;
882 sig->cgtime = cputime_zero;
883 sig->nvcsw = sig->nivcsw = sig->cnvcsw = sig->cnivcsw = 0;
884 sig->min_flt = sig->maj_flt = sig->cmin_flt = sig->cmaj_flt = 0;
885 sig->inblock = sig->oublock = sig->cinblock = sig->coublock = 0;
886 sig->maxrss = sig->cmaxrss = 0;
887 task_io_accounting_init(&sig->ioac);
888 sig->sum_sched_runtime = 0;
889 taskstats_tgid_init(sig);
891 task_lock(current->group_leader);
892 memcpy(sig->rlim, current->signal->rlim, sizeof sig->rlim);
893 task_unlock(current->group_leader);
895 posix_cpu_timers_init_group(sig);
897 acct_init_pacct(&sig->pacct);
901 sig->oom_adj = current->signal->oom_adj;
906 void __cleanup_signal(struct signal_struct *sig)
908 thread_group_cputime_free(sig);
909 tty_kref_put(sig->tty);
910 kmem_cache_free(signal_cachep, sig);
913 static void copy_flags(unsigned long clone_flags, struct task_struct *p)
915 unsigned long new_flags = p->flags;
917 new_flags &= ~PF_SUPERPRIV;
918 new_flags |= PF_FORKNOEXEC;
919 new_flags |= PF_STARTING;
920 p->flags = new_flags;
921 clear_freeze_flag(p);
924 SYSCALL_DEFINE1(set_tid_address, int __user *, tidptr)
926 current->clear_child_tid = tidptr;
928 return task_pid_vnr(current);
931 static void rt_mutex_init_task(struct task_struct *p)
933 spin_lock_init(&p->pi_lock);
934 #ifdef CONFIG_RT_MUTEXES
935 plist_head_init(&p->pi_waiters, &p->pi_lock);
936 p->pi_blocked_on = NULL;
940 #ifdef CONFIG_MM_OWNER
941 void mm_init_owner(struct mm_struct *mm, struct task_struct *p)
945 #endif /* CONFIG_MM_OWNER */
948 * Initialize POSIX timer handling for a single task.
950 static void posix_cpu_timers_init(struct task_struct *tsk)
952 tsk->cputime_expires.prof_exp = cputime_zero;
953 tsk->cputime_expires.virt_exp = cputime_zero;
954 tsk->cputime_expires.sched_exp = 0;
955 INIT_LIST_HEAD(&tsk->cpu_timers[0]);
956 INIT_LIST_HEAD(&tsk->cpu_timers[1]);
957 INIT_LIST_HEAD(&tsk->cpu_timers[2]);
961 * This creates a new process as a copy of the old one,
962 * but does not actually start it yet.
964 * It copies the registers, and all the appropriate
965 * parts of the process environment (as per the clone
966 * flags). The actual kick-off is left to the caller.
968 static struct task_struct *copy_process(unsigned long clone_flags,
969 unsigned long stack_start,
970 struct pt_regs *regs,
971 unsigned long stack_size,
972 int __user *child_tidptr,
977 struct task_struct *p;
978 int cgroup_callbacks_done = 0;
980 if ((clone_flags & (CLONE_NEWNS|CLONE_FS)) == (CLONE_NEWNS|CLONE_FS))
981 return ERR_PTR(-EINVAL);
984 * Thread groups must share signals as well, and detached threads
985 * can only be started up within the thread group.
987 if ((clone_flags & CLONE_THREAD) && !(clone_flags & CLONE_SIGHAND))
988 return ERR_PTR(-EINVAL);
991 * Shared signal handlers imply shared VM. By way of the above,
992 * thread groups also imply shared VM. Blocking this case allows
993 * for various simplifications in other code.
995 if ((clone_flags & CLONE_SIGHAND) && !(clone_flags & CLONE_VM))
996 return ERR_PTR(-EINVAL);
999 * Siblings of global init remain as zombies on exit since they are
1000 * not reaped by their parent (swapper). To solve this and to avoid
1001 * multi-rooted process trees, prevent global and container-inits
1002 * from creating siblings.
1004 if ((clone_flags & CLONE_PARENT) &&
1005 current->signal->flags & SIGNAL_UNKILLABLE)
1006 return ERR_PTR(-EINVAL);
1008 retval = security_task_create(clone_flags);
1013 p = dup_task_struct(current);
1017 ftrace_graph_init_task(p);
1019 rt_mutex_init_task(p);
1021 #ifdef CONFIG_PROVE_LOCKING
1022 DEBUG_LOCKS_WARN_ON(!p->hardirqs_enabled);
1023 DEBUG_LOCKS_WARN_ON(!p->softirqs_enabled);
1026 if (atomic_read(&p->real_cred->user->processes) >=
1027 p->signal->rlim[RLIMIT_NPROC].rlim_cur) {
1028 if (!capable(CAP_SYS_ADMIN) && !capable(CAP_SYS_RESOURCE) &&
1029 p->real_cred->user != INIT_USER)
1033 retval = copy_creds(p, clone_flags);
1038 * If multiple threads are within copy_process(), then this check
1039 * triggers too late. This doesn't hurt, the check is only there
1040 * to stop root fork bombs.
1043 if (nr_threads >= max_threads)
1044 goto bad_fork_cleanup_count;
1046 if (!try_module_get(task_thread_info(p)->exec_domain->module))
1047 goto bad_fork_cleanup_count;
1050 delayacct_tsk_init(p); /* Must remain after dup_task_struct() */
1051 copy_flags(clone_flags, p);
1052 INIT_LIST_HEAD(&p->children);
1053 INIT_LIST_HEAD(&p->sibling);
1054 rcu_copy_process(p);
1055 p->vfork_done = NULL;
1056 spin_lock_init(&p->alloc_lock);
1058 init_sigpending(&p->pending);
1060 p->utime = cputime_zero;
1061 p->stime = cputime_zero;
1062 p->gtime = cputime_zero;
1063 p->utimescaled = cputime_zero;
1064 p->stimescaled = cputime_zero;
1065 p->prev_utime = cputime_zero;
1066 p->prev_stime = cputime_zero;
1068 p->default_timer_slack_ns = current->timer_slack_ns;
1070 task_io_accounting_init(&p->ioac);
1071 acct_clear_integrals(p);
1073 posix_cpu_timers_init(p);
1075 p->lock_depth = -1; /* -1 = no lock */
1076 do_posix_clock_monotonic_gettime(&p->start_time);
1077 p->real_start_time = p->start_time;
1078 monotonic_to_bootbased(&p->real_start_time);
1079 p->io_context = NULL;
1080 p->audit_context = NULL;
1083 p->mempolicy = mpol_dup(p->mempolicy);
1084 if (IS_ERR(p->mempolicy)) {
1085 retval = PTR_ERR(p->mempolicy);
1086 p->mempolicy = NULL;
1087 goto bad_fork_cleanup_cgroup;
1089 mpol_fix_fork_child_flag(p);
1091 #ifdef CONFIG_TRACE_IRQFLAGS
1093 #ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW
1094 p->hardirqs_enabled = 1;
1096 p->hardirqs_enabled = 0;
1098 p->hardirq_enable_ip = 0;
1099 p->hardirq_enable_event = 0;
1100 p->hardirq_disable_ip = _THIS_IP_;
1101 p->hardirq_disable_event = 0;
1102 p->softirqs_enabled = 1;
1103 p->softirq_enable_ip = _THIS_IP_;
1104 p->softirq_enable_event = 0;
1105 p->softirq_disable_ip = 0;
1106 p->softirq_disable_event = 0;
1107 p->hardirq_context = 0;
1108 p->softirq_context = 0;
1110 #ifdef CONFIG_LOCKDEP
1111 p->lockdep_depth = 0; /* no locks held yet */
1112 p->curr_chain_key = 0;
1113 p->lockdep_recursion = 0;
1116 #ifdef CONFIG_DEBUG_MUTEXES
1117 p->blocked_on = NULL; /* not blocked yet */
1122 p->stack_start = stack_start;
1124 /* Perform scheduler related setup. Assign this task to a CPU. */
1125 sched_fork(p, clone_flags);
1127 retval = perf_event_init_task(p);
1129 goto bad_fork_cleanup_policy;
1131 if ((retval = audit_alloc(p)))
1132 goto bad_fork_cleanup_policy;
1133 /* copy all the process information */
1134 if ((retval = copy_semundo(clone_flags, p)))
1135 goto bad_fork_cleanup_audit;
1136 if ((retval = copy_files(clone_flags, p)))
1137 goto bad_fork_cleanup_semundo;
1138 if ((retval = copy_fs(clone_flags, p)))
1139 goto bad_fork_cleanup_files;
1140 if ((retval = copy_sighand(clone_flags, p)))
1141 goto bad_fork_cleanup_fs;
1142 if ((retval = copy_signal(clone_flags, p)))
1143 goto bad_fork_cleanup_sighand;
1144 if ((retval = copy_mm(clone_flags, p)))
1145 goto bad_fork_cleanup_signal;
1146 if ((retval = copy_namespaces(clone_flags, p)))
1147 goto bad_fork_cleanup_mm;
1148 if ((retval = copy_io(clone_flags, p)))
1149 goto bad_fork_cleanup_namespaces;
1150 retval = copy_thread(clone_flags, stack_start, stack_size, p, regs);
1152 goto bad_fork_cleanup_io;
1154 if (pid != &init_struct_pid) {
1156 pid = alloc_pid(p->nsproxy->pid_ns);
1158 goto bad_fork_cleanup_io;
1160 if (clone_flags & CLONE_NEWPID) {
1161 retval = pid_ns_prepare_proc(p->nsproxy->pid_ns);
1163 goto bad_fork_free_pid;
1167 p->pid = pid_nr(pid);
1169 if (clone_flags & CLONE_THREAD)
1170 p->tgid = current->tgid;
1172 if (current->nsproxy != p->nsproxy) {
1173 retval = ns_cgroup_clone(p, pid);
1175 goto bad_fork_free_pid;
1178 p->set_child_tid = (clone_flags & CLONE_CHILD_SETTID) ? child_tidptr : NULL;
1180 * Clear TID on mm_release()?
1182 p->clear_child_tid = (clone_flags & CLONE_CHILD_CLEARTID) ? child_tidptr: NULL;
1184 p->robust_list = NULL;
1185 #ifdef CONFIG_COMPAT
1186 p->compat_robust_list = NULL;
1188 INIT_LIST_HEAD(&p->pi_state_list);
1189 p->pi_state_cache = NULL;
1192 * sigaltstack should be cleared when sharing the same VM
1194 if ((clone_flags & (CLONE_VM|CLONE_VFORK)) == CLONE_VM)
1195 p->sas_ss_sp = p->sas_ss_size = 0;
1198 * Syscall tracing should be turned off in the child regardless
1201 clear_tsk_thread_flag(p, TIF_SYSCALL_TRACE);
1202 #ifdef TIF_SYSCALL_EMU
1203 clear_tsk_thread_flag(p, TIF_SYSCALL_EMU);
1205 clear_all_latency_tracing(p);
1207 /* ok, now we should be set up.. */
1208 p->exit_signal = (clone_flags & CLONE_THREAD) ? -1 : (clone_flags & CSIGNAL);
1209 p->pdeath_signal = 0;
1213 * Ok, make it visible to the rest of the system.
1214 * We dont wake it up yet.
1216 p->group_leader = p;
1217 INIT_LIST_HEAD(&p->thread_group);
1219 /* Now that the task is set up, run cgroup callbacks if
1220 * necessary. We need to run them before the task is visible
1221 * on the tasklist. */
1222 cgroup_fork_callbacks(p);
1223 cgroup_callbacks_done = 1;
1225 /* Need tasklist lock for parent etc handling! */
1226 write_lock_irq(&tasklist_lock);
1229 * The task hasn't been attached yet, so its cpus_allowed mask will
1230 * not be changed, nor will its assigned CPU.
1232 * The cpus_allowed mask of the parent may have changed after it was
1233 * copied first time - so re-copy it here, then check the child's CPU
1234 * to ensure it is on a valid CPU (and if not, just force it back to
1235 * parent's CPU). This avoids alot of nasty races.
1237 p->cpus_allowed = current->cpus_allowed;
1238 p->rt.nr_cpus_allowed = current->rt.nr_cpus_allowed;
1239 if (unlikely(!cpu_isset(task_cpu(p), p->cpus_allowed) ||
1240 !cpu_online(task_cpu(p))))
1241 set_task_cpu(p, smp_processor_id());
1243 /* CLONE_PARENT re-uses the old parent */
1244 if (clone_flags & (CLONE_PARENT|CLONE_THREAD)) {
1245 p->real_parent = current->real_parent;
1246 p->parent_exec_id = current->parent_exec_id;
1248 p->real_parent = current;
1249 p->parent_exec_id = current->self_exec_id;
1252 spin_lock(¤t->sighand->siglock);
1255 * Process group and session signals need to be delivered to just the
1256 * parent before the fork or both the parent and the child after the
1257 * fork. Restart if a signal comes in before we add the new process to
1258 * it's process group.
1259 * A fatal signal pending means that current will exit, so the new
1260 * thread can't slip out of an OOM kill (or normal SIGKILL).
1262 recalc_sigpending();
1263 if (signal_pending(current)) {
1264 spin_unlock(¤t->sighand->siglock);
1265 write_unlock_irq(&tasklist_lock);
1266 retval = -ERESTARTNOINTR;
1267 goto bad_fork_free_pid;
1270 if (clone_flags & CLONE_THREAD) {
1271 atomic_inc(¤t->signal->count);
1272 atomic_inc(¤t->signal->live);
1273 p->group_leader = current->group_leader;
1274 list_add_tail_rcu(&p->thread_group, &p->group_leader->thread_group);
1277 if (likely(p->pid)) {
1278 list_add_tail(&p->sibling, &p->real_parent->children);
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_rcu(&p->tasks, &init_task.tasks);
1291 __get_cpu_var(process_counts)++;
1293 attach_pid(p, PIDTYPE_PID, pid);
1298 spin_unlock(¤t->sighand->siglock);
1299 write_unlock_irq(&tasklist_lock);
1300 proc_fork_connector(p);
1301 cgroup_post_fork(p);
1306 if (pid != &init_struct_pid)
1308 bad_fork_cleanup_io:
1309 put_io_context(p->io_context);
1310 bad_fork_cleanup_namespaces:
1311 exit_task_namespaces(p);
1312 bad_fork_cleanup_mm:
1315 bad_fork_cleanup_signal:
1316 if (!(clone_flags & CLONE_THREAD))
1317 __cleanup_signal(p->signal);
1318 bad_fork_cleanup_sighand:
1319 __cleanup_sighand(p->sighand);
1320 bad_fork_cleanup_fs:
1321 exit_fs(p); /* blocking */
1322 bad_fork_cleanup_files:
1323 exit_files(p); /* blocking */
1324 bad_fork_cleanup_semundo:
1326 bad_fork_cleanup_audit:
1328 bad_fork_cleanup_policy:
1329 perf_event_free_task(p);
1331 mpol_put(p->mempolicy);
1332 bad_fork_cleanup_cgroup:
1334 cgroup_exit(p, cgroup_callbacks_done);
1335 delayacct_tsk_free(p);
1336 module_put(task_thread_info(p)->exec_domain->module);
1337 bad_fork_cleanup_count:
1338 atomic_dec(&p->cred->user->processes);
1343 return ERR_PTR(retval);
1346 noinline struct pt_regs * __cpuinit __attribute__((weak)) idle_regs(struct pt_regs *regs)
1348 memset(regs, 0, sizeof(struct pt_regs));
1352 struct task_struct * __cpuinit fork_idle(int cpu)
1354 struct task_struct *task;
1355 struct pt_regs regs;
1357 task = copy_process(CLONE_VM, 0, idle_regs(®s), 0, NULL,
1358 &init_struct_pid, 0);
1360 init_idle(task, cpu);
1366 * Ok, this is the main fork-routine.
1368 * It copies the process, and if successful kick-starts
1369 * it and waits for it to finish using the VM if required.
1371 long do_fork(unsigned long clone_flags,
1372 unsigned long stack_start,
1373 struct pt_regs *regs,
1374 unsigned long stack_size,
1375 int __user *parent_tidptr,
1376 int __user *child_tidptr)
1378 struct task_struct *p;
1383 * Do some preliminary argument and permissions checking before we
1384 * actually start allocating stuff
1386 if (clone_flags & CLONE_NEWUSER) {
1387 if (clone_flags & CLONE_THREAD)
1389 /* hopefully this check will go away when userns support is
1392 if (!capable(CAP_SYS_ADMIN) || !capable(CAP_SETUID) ||
1393 !capable(CAP_SETGID))
1398 * We hope to recycle these flags after 2.6.26
1400 if (unlikely(clone_flags & CLONE_STOPPED)) {
1401 static int __read_mostly count = 100;
1403 if (count > 0 && printk_ratelimit()) {
1404 char comm[TASK_COMM_LEN];
1407 printk(KERN_INFO "fork(): process `%s' used deprecated "
1408 "clone flags 0x%lx\n",
1409 get_task_comm(comm, current),
1410 clone_flags & CLONE_STOPPED);
1415 * When called from kernel_thread, don't do user tracing stuff.
1417 if (likely(user_mode(regs)))
1418 trace = tracehook_prepare_clone(clone_flags);
1420 p = copy_process(clone_flags, stack_start, regs, stack_size,
1421 child_tidptr, NULL, trace);
1423 * Do this prior waking up the new thread - the thread pointer
1424 * might get invalid after that point, if the thread exits quickly.
1427 struct completion vfork;
1429 trace_sched_process_fork(current, p);
1431 nr = task_pid_vnr(p);
1433 if (clone_flags & CLONE_PARENT_SETTID)
1434 put_user(nr, parent_tidptr);
1436 if (clone_flags & CLONE_VFORK) {
1437 p->vfork_done = &vfork;
1438 init_completion(&vfork);
1441 audit_finish_fork(p);
1442 tracehook_report_clone(regs, clone_flags, nr, p);
1445 * We set PF_STARTING at creation in case tracing wants to
1446 * use this to distinguish a fully live task from one that
1447 * hasn't gotten to tracehook_report_clone() yet. Now we
1448 * clear it and set the child going.
1450 p->flags &= ~PF_STARTING;
1452 if (unlikely(clone_flags & CLONE_STOPPED)) {
1454 * We'll start up with an immediate SIGSTOP.
1456 sigaddset(&p->pending.signal, SIGSTOP);
1457 set_tsk_thread_flag(p, TIF_SIGPENDING);
1458 __set_task_state(p, TASK_STOPPED);
1460 wake_up_new_task(p, clone_flags);
1463 tracehook_report_clone_complete(trace, regs,
1464 clone_flags, nr, p);
1466 if (clone_flags & CLONE_VFORK) {
1467 freezer_do_not_count();
1468 wait_for_completion(&vfork);
1470 tracehook_report_vfork_done(p, nr);
1478 #ifndef ARCH_MIN_MMSTRUCT_ALIGN
1479 #define ARCH_MIN_MMSTRUCT_ALIGN 0
1482 static void sighand_ctor(void *data)
1484 struct sighand_struct *sighand = data;
1486 spin_lock_init(&sighand->siglock);
1487 init_waitqueue_head(&sighand->signalfd_wqh);
1490 void __init proc_caches_init(void)
1492 sighand_cachep = kmem_cache_create("sighand_cache",
1493 sizeof(struct sighand_struct), 0,
1494 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_DESTROY_BY_RCU|
1495 SLAB_NOTRACK, sighand_ctor);
1496 signal_cachep = kmem_cache_create("signal_cache",
1497 sizeof(struct signal_struct), 0,
1498 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL);
1499 files_cachep = kmem_cache_create("files_cache",
1500 sizeof(struct files_struct), 0,
1501 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL);
1502 fs_cachep = kmem_cache_create("fs_cache",
1503 sizeof(struct fs_struct), 0,
1504 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL);
1505 mm_cachep = kmem_cache_create("mm_struct",
1506 sizeof(struct mm_struct), ARCH_MIN_MMSTRUCT_ALIGN,
1507 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL);
1508 vm_area_cachep = KMEM_CACHE(vm_area_struct, SLAB_PANIC);
1513 * Check constraints on flags passed to the unshare system call and
1514 * force unsharing of additional process context as appropriate.
1516 static void check_unshare_flags(unsigned long *flags_ptr)
1519 * If unsharing a thread from a thread group, must also
1522 if (*flags_ptr & CLONE_THREAD)
1523 *flags_ptr |= CLONE_VM;
1526 * If unsharing vm, must also unshare signal handlers.
1528 if (*flags_ptr & CLONE_VM)
1529 *flags_ptr |= CLONE_SIGHAND;
1532 * If unsharing signal handlers and the task was created
1533 * using CLONE_THREAD, then must unshare the thread
1535 if ((*flags_ptr & CLONE_SIGHAND) &&
1536 (atomic_read(¤t->signal->count) > 1))
1537 *flags_ptr |= CLONE_THREAD;
1540 * If unsharing namespace, must also unshare filesystem information.
1542 if (*flags_ptr & CLONE_NEWNS)
1543 *flags_ptr |= CLONE_FS;
1547 * Unsharing of tasks created with CLONE_THREAD is not supported yet
1549 static int unshare_thread(unsigned long unshare_flags)
1551 if (unshare_flags & CLONE_THREAD)
1558 * Unshare the filesystem structure if it is being shared
1560 static int unshare_fs(unsigned long unshare_flags, struct fs_struct **new_fsp)
1562 struct fs_struct *fs = current->fs;
1564 if (!(unshare_flags & CLONE_FS) || !fs)
1567 /* don't need lock here; in the worst case we'll do useless copy */
1571 *new_fsp = copy_fs_struct(fs);
1579 * Unsharing of sighand is not supported yet
1581 static int unshare_sighand(unsigned long unshare_flags, struct sighand_struct **new_sighp)
1583 struct sighand_struct *sigh = current->sighand;
1585 if ((unshare_flags & CLONE_SIGHAND) && atomic_read(&sigh->count) > 1)
1592 * Unshare vm if it is being shared
1594 static int unshare_vm(unsigned long unshare_flags, struct mm_struct **new_mmp)
1596 struct mm_struct *mm = current->mm;
1598 if ((unshare_flags & CLONE_VM) &&
1599 (mm && atomic_read(&mm->mm_users) > 1)) {
1607 * Unshare file descriptor table if it is being shared
1609 static int unshare_fd(unsigned long unshare_flags, struct files_struct **new_fdp)
1611 struct files_struct *fd = current->files;
1614 if ((unshare_flags & CLONE_FILES) &&
1615 (fd && atomic_read(&fd->count) > 1)) {
1616 *new_fdp = dup_fd(fd, &error);
1625 * unshare allows a process to 'unshare' part of the process
1626 * context which was originally shared using clone. copy_*
1627 * functions used by do_fork() cannot be used here directly
1628 * because they modify an inactive task_struct that is being
1629 * constructed. Here we are modifying the current, active,
1632 SYSCALL_DEFINE1(unshare, unsigned long, unshare_flags)
1635 struct fs_struct *fs, *new_fs = NULL;
1636 struct sighand_struct *new_sigh = NULL;
1637 struct mm_struct *mm, *new_mm = NULL, *active_mm = NULL;
1638 struct files_struct *fd, *new_fd = NULL;
1639 struct nsproxy *new_nsproxy = NULL;
1642 check_unshare_flags(&unshare_flags);
1644 /* Return -EINVAL for all unsupported flags */
1646 if (unshare_flags & ~(CLONE_THREAD|CLONE_FS|CLONE_NEWNS|CLONE_SIGHAND|
1647 CLONE_VM|CLONE_FILES|CLONE_SYSVSEM|
1648 CLONE_NEWUTS|CLONE_NEWIPC|CLONE_NEWNET))
1649 goto bad_unshare_out;
1652 * CLONE_NEWIPC must also detach from the undolist: after switching
1653 * to a new ipc namespace, the semaphore arrays from the old
1654 * namespace are unreachable.
1656 if (unshare_flags & (CLONE_NEWIPC|CLONE_SYSVSEM))
1658 if ((err = unshare_thread(unshare_flags)))
1659 goto bad_unshare_out;
1660 if ((err = unshare_fs(unshare_flags, &new_fs)))
1661 goto bad_unshare_cleanup_thread;
1662 if ((err = unshare_sighand(unshare_flags, &new_sigh)))
1663 goto bad_unshare_cleanup_fs;
1664 if ((err = unshare_vm(unshare_flags, &new_mm)))
1665 goto bad_unshare_cleanup_sigh;
1666 if ((err = unshare_fd(unshare_flags, &new_fd)))
1667 goto bad_unshare_cleanup_vm;
1668 if ((err = unshare_nsproxy_namespaces(unshare_flags, &new_nsproxy,
1670 goto bad_unshare_cleanup_fd;
1672 if (new_fs || new_mm || new_fd || do_sysvsem || new_nsproxy) {
1675 * CLONE_SYSVSEM is equivalent to sys_exit().
1681 switch_task_namespaces(current, new_nsproxy);
1689 write_lock(&fs->lock);
1690 current->fs = new_fs;
1695 write_unlock(&fs->lock);
1700 active_mm = current->active_mm;
1701 current->mm = new_mm;
1702 current->active_mm = new_mm;
1703 activate_mm(active_mm, new_mm);
1708 fd = current->files;
1709 current->files = new_fd;
1713 task_unlock(current);
1717 put_nsproxy(new_nsproxy);
1719 bad_unshare_cleanup_fd:
1721 put_files_struct(new_fd);
1723 bad_unshare_cleanup_vm:
1727 bad_unshare_cleanup_sigh:
1729 if (atomic_dec_and_test(&new_sigh->count))
1730 kmem_cache_free(sighand_cachep, new_sigh);
1732 bad_unshare_cleanup_fs:
1734 free_fs_struct(new_fs);
1736 bad_unshare_cleanup_thread:
1742 * Helper to unshare the files of the current task.
1743 * We don't want to expose copy_files internals to
1744 * the exec layer of the kernel.
1747 int unshare_files(struct files_struct **displaced)
1749 struct task_struct *task = current;
1750 struct files_struct *copy = NULL;
1753 error = unshare_fd(CLONE_FILES, ©);
1754 if (error || !copy) {
1758 *displaced = task->files;