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 #ifdef CONFIG_PROVE_RCU
91 int lockdep_tasklist_lock_is_held(void)
93 return lockdep_is_held(&tasklist_lock);
95 EXPORT_SYMBOL_GPL(lockdep_tasklist_lock_is_held);
96 #endif /* #ifdef CONFIG_PROVE_RCU */
98 int nr_processes(void)
103 for_each_possible_cpu(cpu)
104 total += per_cpu(process_counts, cpu);
109 #ifndef __HAVE_ARCH_TASK_STRUCT_ALLOCATOR
110 # define alloc_task_struct() kmem_cache_alloc(task_struct_cachep, GFP_KERNEL)
111 # define free_task_struct(tsk) kmem_cache_free(task_struct_cachep, (tsk))
112 static struct kmem_cache *task_struct_cachep;
115 #ifndef __HAVE_ARCH_THREAD_INFO_ALLOCATOR
116 static inline struct thread_info *alloc_thread_info(struct task_struct *tsk)
118 #ifdef CONFIG_DEBUG_STACK_USAGE
119 gfp_t mask = GFP_KERNEL | __GFP_ZERO;
121 gfp_t mask = GFP_KERNEL;
123 return (struct thread_info *)__get_free_pages(mask, THREAD_SIZE_ORDER);
126 static inline void free_thread_info(struct thread_info *ti)
128 free_pages((unsigned long)ti, THREAD_SIZE_ORDER);
132 /* SLAB cache for signal_struct structures (tsk->signal) */
133 static struct kmem_cache *signal_cachep;
135 /* SLAB cache for sighand_struct structures (tsk->sighand) */
136 struct kmem_cache *sighand_cachep;
138 /* SLAB cache for files_struct structures (tsk->files) */
139 struct kmem_cache *files_cachep;
141 /* SLAB cache for fs_struct structures (tsk->fs) */
142 struct kmem_cache *fs_cachep;
144 /* SLAB cache for vm_area_struct structures */
145 struct kmem_cache *vm_area_cachep;
147 /* SLAB cache for mm_struct structures (tsk->mm) */
148 static struct kmem_cache *mm_cachep;
150 /* Notifier list called when a task struct is freed */
151 static ATOMIC_NOTIFIER_HEAD(task_free_notifier);
153 static void account_kernel_stack(struct thread_info *ti, int account)
155 struct zone *zone = page_zone(virt_to_page(ti));
157 mod_zone_page_state(zone, NR_KERNEL_STACK, account);
160 void free_task(struct task_struct *tsk)
162 prop_local_destroy_single(&tsk->dirties);
163 account_kernel_stack(tsk->stack, -1);
164 free_thread_info(tsk->stack);
165 rt_mutex_debug_task_free(tsk);
166 ftrace_graph_exit_task(tsk);
167 free_task_struct(tsk);
169 EXPORT_SYMBOL(free_task);
171 static inline void free_signal_struct(struct signal_struct *sig)
173 taskstats_tgid_free(sig);
174 kmem_cache_free(signal_cachep, sig);
177 static inline void put_signal_struct(struct signal_struct *sig)
179 if (atomic_dec_and_test(&sig->sigcnt))
180 free_signal_struct(sig);
183 int task_free_register(struct notifier_block *n)
185 return atomic_notifier_chain_register(&task_free_notifier, n);
187 EXPORT_SYMBOL(task_free_register);
189 int task_free_unregister(struct notifier_block *n)
191 return atomic_notifier_chain_unregister(&task_free_notifier, n);
193 EXPORT_SYMBOL(task_free_unregister);
195 void __put_task_struct(struct task_struct *tsk)
197 WARN_ON(!tsk->exit_state);
198 WARN_ON(atomic_read(&tsk->usage));
199 WARN_ON(tsk == current);
202 delayacct_tsk_free(tsk);
203 put_signal_struct(tsk->signal);
205 atomic_notifier_call_chain(&task_free_notifier, 0, tsk);
206 if (!profile_handoff_task(tsk))
211 * macro override instead of weak attribute alias, to workaround
212 * gcc 4.1.0 and 4.1.1 bugs with weak attribute and empty functions.
214 #ifndef arch_task_cache_init
215 #define arch_task_cache_init()
218 void __init fork_init(unsigned long mempages)
220 #ifndef __HAVE_ARCH_TASK_STRUCT_ALLOCATOR
221 #ifndef ARCH_MIN_TASKALIGN
222 #define ARCH_MIN_TASKALIGN L1_CACHE_BYTES
224 /* create a slab on which task_structs can be allocated */
226 kmem_cache_create("task_struct", sizeof(struct task_struct),
227 ARCH_MIN_TASKALIGN, SLAB_PANIC | SLAB_NOTRACK, NULL);
230 /* do the arch specific task caches init */
231 arch_task_cache_init();
234 * The default maximum number of threads is set to a safe
235 * value: the thread structures can take up at most half
238 max_threads = mempages / (8 * THREAD_SIZE / PAGE_SIZE);
241 * we need to allow at least 20 threads to boot a system
246 init_task.signal->rlim[RLIMIT_NPROC].rlim_cur = max_threads/2;
247 init_task.signal->rlim[RLIMIT_NPROC].rlim_max = max_threads/2;
248 init_task.signal->rlim[RLIMIT_SIGPENDING] =
249 init_task.signal->rlim[RLIMIT_NPROC];
252 int __attribute__((weak)) arch_dup_task_struct(struct task_struct *dst,
253 struct task_struct *src)
259 static struct task_struct *dup_task_struct(struct task_struct *orig)
261 struct task_struct *tsk;
262 struct thread_info *ti;
263 unsigned long *stackend;
267 prepare_to_copy(orig);
269 tsk = alloc_task_struct();
273 ti = alloc_thread_info(tsk);
275 free_task_struct(tsk);
279 err = arch_dup_task_struct(tsk, orig);
285 err = prop_local_init_single(&tsk->dirties);
289 setup_thread_stack(tsk, orig);
290 clear_user_return_notifier(tsk);
291 stackend = end_of_stack(tsk);
292 *stackend = STACK_END_MAGIC; /* for overflow detection */
294 #ifdef CONFIG_CC_STACKPROTECTOR
295 tsk->stack_canary = get_random_int();
298 /* One for us, one for whoever does the "release_task()" (usually parent) */
299 atomic_set(&tsk->usage,2);
300 atomic_set(&tsk->fs_excl, 0);
301 #ifdef CONFIG_BLK_DEV_IO_TRACE
304 tsk->splice_pipe = NULL;
306 account_kernel_stack(ti, 1);
311 free_thread_info(ti);
312 free_task_struct(tsk);
317 static int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm)
319 struct vm_area_struct *mpnt, *tmp, **pprev;
320 struct rb_node **rb_link, *rb_parent;
322 unsigned long charge;
323 struct mempolicy *pol;
325 down_write(&oldmm->mmap_sem);
326 flush_cache_dup_mm(oldmm);
328 * Not linked in yet - no deadlock potential:
330 down_write_nested(&mm->mmap_sem, SINGLE_DEPTH_NESTING);
334 mm->mmap_cache = NULL;
335 mm->free_area_cache = oldmm->mmap_base;
336 mm->cached_hole_size = ~0UL;
338 cpumask_clear(mm_cpumask(mm));
340 rb_link = &mm->mm_rb.rb_node;
343 retval = ksm_fork(mm, oldmm);
347 for (mpnt = oldmm->mmap; mpnt; mpnt = mpnt->vm_next) {
350 if (mpnt->vm_flags & VM_DONTCOPY) {
351 long pages = vma_pages(mpnt);
352 mm->total_vm -= pages;
353 vm_stat_account(mm, mpnt->vm_flags, mpnt->vm_file,
358 if (mpnt->vm_flags & VM_ACCOUNT) {
359 unsigned int len = (mpnt->vm_end - mpnt->vm_start) >> PAGE_SHIFT;
360 if (security_vm_enough_memory(len))
364 tmp = kmem_cache_alloc(vm_area_cachep, GFP_KERNEL);
368 INIT_LIST_HEAD(&tmp->anon_vma_chain);
369 pol = mpol_dup(vma_policy(mpnt));
370 retval = PTR_ERR(pol);
372 goto fail_nomem_policy;
373 vma_set_policy(tmp, pol);
374 if (anon_vma_fork(tmp, mpnt))
375 goto fail_nomem_anon_vma_fork;
376 tmp->vm_flags &= ~VM_LOCKED;
381 struct inode *inode = file->f_path.dentry->d_inode;
382 struct address_space *mapping = file->f_mapping;
385 if (tmp->vm_flags & VM_DENYWRITE)
386 atomic_dec(&inode->i_writecount);
387 spin_lock(&mapping->i_mmap_lock);
388 if (tmp->vm_flags & VM_SHARED)
389 mapping->i_mmap_writable++;
390 tmp->vm_truncate_count = mpnt->vm_truncate_count;
391 flush_dcache_mmap_lock(mapping);
392 /* insert tmp into the share list, just after mpnt */
393 vma_prio_tree_add(tmp, mpnt);
394 flush_dcache_mmap_unlock(mapping);
395 spin_unlock(&mapping->i_mmap_lock);
399 * Clear hugetlb-related page reserves for children. This only
400 * affects MAP_PRIVATE mappings. Faults generated by the child
401 * are not guaranteed to succeed, even if read-only
403 if (is_vm_hugetlb_page(tmp))
404 reset_vma_resv_huge_pages(tmp);
407 * Link in the new vma and copy the page table entries.
410 pprev = &tmp->vm_next;
412 __vma_link_rb(mm, tmp, rb_link, rb_parent);
413 rb_link = &tmp->vm_rb.rb_right;
414 rb_parent = &tmp->vm_rb;
417 retval = copy_page_range(mm, oldmm, mpnt);
419 if (tmp->vm_ops && tmp->vm_ops->open)
420 tmp->vm_ops->open(tmp);
425 /* a new mm has just been created */
426 arch_dup_mmap(oldmm, mm);
429 up_write(&mm->mmap_sem);
431 up_write(&oldmm->mmap_sem);
433 fail_nomem_anon_vma_fork:
436 kmem_cache_free(vm_area_cachep, tmp);
439 vm_unacct_memory(charge);
443 static inline int mm_alloc_pgd(struct mm_struct * mm)
445 mm->pgd = pgd_alloc(mm);
446 if (unlikely(!mm->pgd))
451 static inline void mm_free_pgd(struct mm_struct * mm)
453 pgd_free(mm, mm->pgd);
456 #define dup_mmap(mm, oldmm) (0)
457 #define mm_alloc_pgd(mm) (0)
458 #define mm_free_pgd(mm)
459 #endif /* CONFIG_MMU */
461 __cacheline_aligned_in_smp DEFINE_SPINLOCK(mmlist_lock);
463 #define allocate_mm() (kmem_cache_alloc(mm_cachep, GFP_KERNEL))
464 #define free_mm(mm) (kmem_cache_free(mm_cachep, (mm)))
466 static unsigned long default_dump_filter = MMF_DUMP_FILTER_DEFAULT;
468 static int __init coredump_filter_setup(char *s)
470 default_dump_filter =
471 (simple_strtoul(s, NULL, 0) << MMF_DUMP_FILTER_SHIFT) &
472 MMF_DUMP_FILTER_MASK;
476 __setup("coredump_filter=", coredump_filter_setup);
478 #include <linux/init_task.h>
480 static void mm_init_aio(struct mm_struct *mm)
483 spin_lock_init(&mm->ioctx_lock);
484 INIT_HLIST_HEAD(&mm->ioctx_list);
488 static struct mm_struct * mm_init(struct mm_struct * mm, struct task_struct *p)
490 atomic_set(&mm->mm_users, 1);
491 atomic_set(&mm->mm_count, 1);
492 init_rwsem(&mm->mmap_sem);
493 INIT_LIST_HEAD(&mm->mmlist);
494 mm->flags = (current->mm) ?
495 (current->mm->flags & MMF_INIT_MASK) : default_dump_filter;
496 mm->core_state = NULL;
498 memset(&mm->rss_stat, 0, sizeof(mm->rss_stat));
499 spin_lock_init(&mm->page_table_lock);
500 mm->free_area_cache = TASK_UNMAPPED_BASE;
501 mm->cached_hole_size = ~0UL;
503 mm_init_owner(mm, p);
505 if (likely(!mm_alloc_pgd(mm))) {
507 mmu_notifier_mm_init(mm);
516 * Allocate and initialize an mm_struct.
518 struct mm_struct * mm_alloc(void)
520 struct mm_struct * mm;
524 memset(mm, 0, sizeof(*mm));
525 mm = mm_init(mm, current);
531 * Called when the last reference to the mm
532 * is dropped: either by a lazy thread or by
533 * mmput. Free the page directory and the mm.
535 void __mmdrop(struct mm_struct *mm)
537 BUG_ON(mm == &init_mm);
540 mmu_notifier_mm_destroy(mm);
543 EXPORT_SYMBOL_GPL(__mmdrop);
546 * Decrement the use count and release all resources for an mm.
548 void mmput(struct mm_struct *mm)
552 if (atomic_dec_and_test(&mm->mm_users)) {
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 * get_task_mm - acquire a reference to the task's mm
573 * Returns %NULL if the task has no mm. Checks PF_KTHREAD (meaning
574 * this kernel workthread has transiently adopted a user mm with use_mm,
575 * to do its AIO) is not set and if so returns a reference to it, after
576 * bumping up the use count. User must release the mm via mmput()
577 * after use. Typically used by /proc and ptrace.
579 struct mm_struct *get_task_mm(struct task_struct *task)
581 struct mm_struct *mm;
586 if (task->flags & PF_KTHREAD)
589 atomic_inc(&mm->mm_users);
594 EXPORT_SYMBOL_GPL(get_task_mm);
596 /* Please note the differences between mmput and mm_release.
597 * mmput is called whenever we stop holding onto a mm_struct,
598 * error success whatever.
600 * mm_release is called after a mm_struct has been removed
601 * from the current process.
603 * This difference is important for error handling, when we
604 * only half set up a mm_struct for a new process and need to restore
605 * the old one. Because we mmput the new mm_struct before
606 * restoring the old one. . .
607 * Eric Biederman 10 January 1998
609 void mm_release(struct task_struct *tsk, struct mm_struct *mm)
611 struct completion *vfork_done = tsk->vfork_done;
613 /* Get rid of any futexes when releasing the mm */
615 if (unlikely(tsk->robust_list)) {
616 exit_robust_list(tsk);
617 tsk->robust_list = NULL;
620 if (unlikely(tsk->compat_robust_list)) {
621 compat_exit_robust_list(tsk);
622 tsk->compat_robust_list = NULL;
625 if (unlikely(!list_empty(&tsk->pi_state_list)))
626 exit_pi_state_list(tsk);
629 /* Get rid of any cached register state */
630 deactivate_mm(tsk, mm);
632 /* notify parent sleeping on vfork() */
634 tsk->vfork_done = NULL;
635 complete(vfork_done);
639 * If we're exiting normally, clear a user-space tid field if
640 * requested. We leave this alone when dying by signal, to leave
641 * the value intact in a core dump, and to save the unnecessary
642 * trouble otherwise. Userland only wants this done for a sys_exit.
644 if (tsk->clear_child_tid) {
645 if (!(tsk->flags & PF_SIGNALED) &&
646 atomic_read(&mm->mm_users) > 1) {
648 * We don't check the error code - if userspace has
649 * not set up a proper pointer then tough luck.
651 put_user(0, tsk->clear_child_tid);
652 sys_futex(tsk->clear_child_tid, FUTEX_WAKE,
655 tsk->clear_child_tid = NULL;
660 * Allocate a new mm structure and copy contents from the
661 * mm structure of the passed in task structure.
663 struct mm_struct *dup_mm(struct task_struct *tsk)
665 struct mm_struct *mm, *oldmm = current->mm;
675 memcpy(mm, oldmm, sizeof(*mm));
677 /* Initializing for Swap token stuff */
678 mm->token_priority = 0;
679 mm->last_interval = 0;
681 if (!mm_init(mm, tsk))
684 if (init_new_context(tsk, mm))
687 dup_mm_exe_file(oldmm, mm);
689 err = dup_mmap(mm, oldmm);
693 mm->hiwater_rss = get_mm_rss(mm);
694 mm->hiwater_vm = mm->total_vm;
696 if (mm->binfmt && !try_module_get(mm->binfmt->module))
702 /* don't put binfmt in mmput, we haven't got module yet */
711 * If init_new_context() failed, we cannot use mmput() to free the mm
712 * because it calls destroy_context()
719 static int copy_mm(unsigned long clone_flags, struct task_struct * tsk)
721 struct mm_struct * mm, *oldmm;
724 tsk->min_flt = tsk->maj_flt = 0;
725 tsk->nvcsw = tsk->nivcsw = 0;
726 #ifdef CONFIG_DETECT_HUNG_TASK
727 tsk->last_switch_count = tsk->nvcsw + tsk->nivcsw;
731 tsk->active_mm = NULL;
734 * Are we cloning a kernel thread?
736 * We need to steal a active VM for that..
742 if (clone_flags & CLONE_VM) {
743 atomic_inc(&oldmm->mm_users);
754 /* Initializing for Swap token stuff */
755 mm->token_priority = 0;
756 mm->last_interval = 0;
766 static int copy_fs(unsigned long clone_flags, struct task_struct *tsk)
768 struct fs_struct *fs = current->fs;
769 if (clone_flags & CLONE_FS) {
770 /* tsk->fs is already what we want */
771 write_lock(&fs->lock);
773 write_unlock(&fs->lock);
777 write_unlock(&fs->lock);
780 tsk->fs = copy_fs_struct(fs);
786 static int copy_files(unsigned long clone_flags, struct task_struct * tsk)
788 struct files_struct *oldf, *newf;
792 * A background process may not have any files ...
794 oldf = current->files;
798 if (clone_flags & CLONE_FILES) {
799 atomic_inc(&oldf->count);
803 newf = dup_fd(oldf, &error);
813 static int copy_io(unsigned long clone_flags, struct task_struct *tsk)
816 struct io_context *ioc = current->io_context;
821 * Share io context with parent, if CLONE_IO is set
823 if (clone_flags & CLONE_IO) {
824 tsk->io_context = ioc_task_link(ioc);
825 if (unlikely(!tsk->io_context))
827 } else if (ioprio_valid(ioc->ioprio)) {
828 tsk->io_context = alloc_io_context(GFP_KERNEL, -1);
829 if (unlikely(!tsk->io_context))
832 tsk->io_context->ioprio = ioc->ioprio;
838 static int copy_sighand(unsigned long clone_flags, struct task_struct *tsk)
840 struct sighand_struct *sig;
842 if (clone_flags & CLONE_SIGHAND) {
843 atomic_inc(¤t->sighand->count);
846 sig = kmem_cache_alloc(sighand_cachep, GFP_KERNEL);
847 rcu_assign_pointer(tsk->sighand, sig);
850 atomic_set(&sig->count, 1);
851 memcpy(sig->action, current->sighand->action, sizeof(sig->action));
855 void __cleanup_sighand(struct sighand_struct *sighand)
857 if (atomic_dec_and_test(&sighand->count))
858 kmem_cache_free(sighand_cachep, sighand);
863 * Initialize POSIX timer handling for a thread group.
865 static void posix_cpu_timers_init_group(struct signal_struct *sig)
867 unsigned long cpu_limit;
869 /* Thread group counters. */
870 thread_group_cputime_init(sig);
872 cpu_limit = ACCESS_ONCE(sig->rlim[RLIMIT_CPU].rlim_cur);
873 if (cpu_limit != RLIM_INFINITY) {
874 sig->cputime_expires.prof_exp = secs_to_cputime(cpu_limit);
875 sig->cputimer.running = 1;
878 /* The timer lists. */
879 INIT_LIST_HEAD(&sig->cpu_timers[0]);
880 INIT_LIST_HEAD(&sig->cpu_timers[1]);
881 INIT_LIST_HEAD(&sig->cpu_timers[2]);
884 static int copy_signal(unsigned long clone_flags, struct task_struct *tsk)
886 struct signal_struct *sig;
888 if (clone_flags & CLONE_THREAD)
891 sig = kmem_cache_zalloc(signal_cachep, GFP_KERNEL);
897 atomic_set(&sig->live, 1);
898 atomic_set(&sig->sigcnt, 1);
899 init_waitqueue_head(&sig->wait_chldexit);
900 if (clone_flags & CLONE_NEWPID)
901 sig->flags |= SIGNAL_UNKILLABLE;
902 sig->curr_target = tsk;
903 init_sigpending(&sig->shared_pending);
904 INIT_LIST_HEAD(&sig->posix_timers);
906 hrtimer_init(&sig->real_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
907 sig->real_timer.function = it_real_fn;
909 task_lock(current->group_leader);
910 memcpy(sig->rlim, current->signal->rlim, sizeof sig->rlim);
911 task_unlock(current->group_leader);
913 posix_cpu_timers_init_group(sig);
917 sig->oom_adj = current->signal->oom_adj;
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 task_rlimit(p, RLIMIT_NPROC)) {
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;
1078 #if defined(SPLIT_RSS_COUNTING)
1079 memset(&p->rss_stat, 0, sizeof(p->rss_stat));
1082 p->default_timer_slack_ns = current->timer_slack_ns;
1084 task_io_accounting_init(&p->ioac);
1085 acct_clear_integrals(p);
1087 posix_cpu_timers_init(p);
1089 p->lock_depth = -1; /* -1 = no lock */
1090 do_posix_clock_monotonic_gettime(&p->start_time);
1091 p->real_start_time = p->start_time;
1092 monotonic_to_bootbased(&p->real_start_time);
1093 p->io_context = NULL;
1094 p->audit_context = NULL;
1097 p->mempolicy = mpol_dup(p->mempolicy);
1098 if (IS_ERR(p->mempolicy)) {
1099 retval = PTR_ERR(p->mempolicy);
1100 p->mempolicy = NULL;
1101 goto bad_fork_cleanup_cgroup;
1103 mpol_fix_fork_child_flag(p);
1105 #ifdef CONFIG_TRACE_IRQFLAGS
1107 #ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW
1108 p->hardirqs_enabled = 1;
1110 p->hardirqs_enabled = 0;
1112 p->hardirq_enable_ip = 0;
1113 p->hardirq_enable_event = 0;
1114 p->hardirq_disable_ip = _THIS_IP_;
1115 p->hardirq_disable_event = 0;
1116 p->softirqs_enabled = 1;
1117 p->softirq_enable_ip = _THIS_IP_;
1118 p->softirq_enable_event = 0;
1119 p->softirq_disable_ip = 0;
1120 p->softirq_disable_event = 0;
1121 p->hardirq_context = 0;
1122 p->softirq_context = 0;
1124 #ifdef CONFIG_LOCKDEP
1125 p->lockdep_depth = 0; /* no locks held yet */
1126 p->curr_chain_key = 0;
1127 p->lockdep_recursion = 0;
1130 #ifdef CONFIG_DEBUG_MUTEXES
1131 p->blocked_on = NULL; /* not blocked yet */
1133 #ifdef CONFIG_CGROUP_MEM_RES_CTLR
1134 p->memcg_batch.do_batch = 0;
1135 p->memcg_batch.memcg = NULL;
1138 /* Perform scheduler related setup. Assign this task to a CPU. */
1139 sched_fork(p, clone_flags);
1141 retval = perf_event_init_task(p);
1143 goto bad_fork_cleanup_policy;
1145 if ((retval = audit_alloc(p)))
1146 goto bad_fork_cleanup_policy;
1147 /* copy all the process information */
1148 if ((retval = copy_semundo(clone_flags, p)))
1149 goto bad_fork_cleanup_audit;
1150 if ((retval = copy_files(clone_flags, p)))
1151 goto bad_fork_cleanup_semundo;
1152 if ((retval = copy_fs(clone_flags, p)))
1153 goto bad_fork_cleanup_files;
1154 if ((retval = copy_sighand(clone_flags, p)))
1155 goto bad_fork_cleanup_fs;
1156 if ((retval = copy_signal(clone_flags, p)))
1157 goto bad_fork_cleanup_sighand;
1158 if ((retval = copy_mm(clone_flags, p)))
1159 goto bad_fork_cleanup_signal;
1160 if ((retval = copy_namespaces(clone_flags, p)))
1161 goto bad_fork_cleanup_mm;
1162 if ((retval = copy_io(clone_flags, p)))
1163 goto bad_fork_cleanup_namespaces;
1164 retval = copy_thread(clone_flags, stack_start, stack_size, p, regs);
1166 goto bad_fork_cleanup_io;
1168 if (pid != &init_struct_pid) {
1170 pid = alloc_pid(p->nsproxy->pid_ns);
1172 goto bad_fork_cleanup_io;
1174 if (clone_flags & CLONE_NEWPID) {
1175 retval = pid_ns_prepare_proc(p->nsproxy->pid_ns);
1177 goto bad_fork_free_pid;
1181 p->pid = pid_nr(pid);
1183 if (clone_flags & CLONE_THREAD)
1184 p->tgid = current->tgid;
1186 if (current->nsproxy != p->nsproxy) {
1187 retval = ns_cgroup_clone(p, pid);
1189 goto bad_fork_free_pid;
1192 p->set_child_tid = (clone_flags & CLONE_CHILD_SETTID) ? child_tidptr : NULL;
1194 * Clear TID on mm_release()?
1196 p->clear_child_tid = (clone_flags & CLONE_CHILD_CLEARTID) ? child_tidptr: NULL;
1198 p->robust_list = NULL;
1199 #ifdef CONFIG_COMPAT
1200 p->compat_robust_list = NULL;
1202 INIT_LIST_HEAD(&p->pi_state_list);
1203 p->pi_state_cache = NULL;
1206 * sigaltstack should be cleared when sharing the same VM
1208 if ((clone_flags & (CLONE_VM|CLONE_VFORK)) == CLONE_VM)
1209 p->sas_ss_sp = p->sas_ss_size = 0;
1212 * Syscall tracing and stepping should be turned off in the
1213 * child regardless of CLONE_PTRACE.
1215 user_disable_single_step(p);
1216 clear_tsk_thread_flag(p, TIF_SYSCALL_TRACE);
1217 #ifdef TIF_SYSCALL_EMU
1218 clear_tsk_thread_flag(p, TIF_SYSCALL_EMU);
1220 clear_all_latency_tracing(p);
1222 /* ok, now we should be set up.. */
1223 p->exit_signal = (clone_flags & CLONE_THREAD) ? -1 : (clone_flags & CSIGNAL);
1224 p->pdeath_signal = 0;
1228 * Ok, make it visible to the rest of the system.
1229 * We dont wake it up yet.
1231 p->group_leader = p;
1232 INIT_LIST_HEAD(&p->thread_group);
1234 /* Now that the task is set up, run cgroup callbacks if
1235 * necessary. We need to run them before the task is visible
1236 * on the tasklist. */
1237 cgroup_fork_callbacks(p);
1238 cgroup_callbacks_done = 1;
1240 /* Need tasklist lock for parent etc handling! */
1241 write_lock_irq(&tasklist_lock);
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 current->signal->nr_threads++;
1272 atomic_inc(¤t->signal->live);
1273 atomic_inc(¤t->signal->sigcnt);
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 p->signal->tty = tty_kref_get(current->signal->tty);
1287 attach_pid(p, PIDTYPE_PGID, task_pgrp(current));
1288 attach_pid(p, PIDTYPE_SID, task_session(current));
1289 list_add_tail(&p->sibling, &p->real_parent->children);
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:
1311 bad_fork_cleanup_namespaces:
1312 exit_task_namespaces(p);
1313 bad_fork_cleanup_mm:
1316 bad_fork_cleanup_signal:
1317 if (!(clone_flags & CLONE_THREAD))
1318 free_signal_struct(p->signal);
1319 bad_fork_cleanup_sighand:
1320 __cleanup_sighand(p->sighand);
1321 bad_fork_cleanup_fs:
1322 exit_fs(p); /* blocking */
1323 bad_fork_cleanup_files:
1324 exit_files(p); /* blocking */
1325 bad_fork_cleanup_semundo:
1327 bad_fork_cleanup_audit:
1329 bad_fork_cleanup_policy:
1330 perf_event_free_task(p);
1332 mpol_put(p->mempolicy);
1333 bad_fork_cleanup_cgroup:
1335 cgroup_exit(p, cgroup_callbacks_done);
1336 delayacct_tsk_free(p);
1337 module_put(task_thread_info(p)->exec_domain->module);
1338 bad_fork_cleanup_count:
1339 atomic_dec(&p->cred->user->processes);
1344 return ERR_PTR(retval);
1347 noinline struct pt_regs * __cpuinit __attribute__((weak)) idle_regs(struct pt_regs *regs)
1349 memset(regs, 0, sizeof(struct pt_regs));
1353 static inline void init_idle_pids(struct pid_link *links)
1357 for (type = PIDTYPE_PID; type < PIDTYPE_MAX; ++type) {
1358 INIT_HLIST_NODE(&links[type].node); /* not really needed */
1359 links[type].pid = &init_struct_pid;
1363 struct task_struct * __cpuinit fork_idle(int cpu)
1365 struct task_struct *task;
1366 struct pt_regs regs;
1368 task = copy_process(CLONE_VM, 0, idle_regs(®s), 0, NULL,
1369 &init_struct_pid, 0);
1370 if (!IS_ERR(task)) {
1371 init_idle_pids(task->pids);
1372 init_idle(task, cpu);
1379 * Ok, this is the main fork-routine.
1381 * It copies the process, and if successful kick-starts
1382 * it and waits for it to finish using the VM if required.
1384 long do_fork(unsigned long clone_flags,
1385 unsigned long stack_start,
1386 struct pt_regs *regs,
1387 unsigned long stack_size,
1388 int __user *parent_tidptr,
1389 int __user *child_tidptr)
1391 struct task_struct *p;
1396 * Do some preliminary argument and permissions checking before we
1397 * actually start allocating stuff
1399 if (clone_flags & CLONE_NEWUSER) {
1400 if (clone_flags & CLONE_THREAD)
1402 /* hopefully this check will go away when userns support is
1405 if (!capable(CAP_SYS_ADMIN) || !capable(CAP_SETUID) ||
1406 !capable(CAP_SETGID))
1411 * We hope to recycle these flags after 2.6.26
1413 if (unlikely(clone_flags & CLONE_STOPPED)) {
1414 static int __read_mostly count = 100;
1416 if (count > 0 && printk_ratelimit()) {
1417 char comm[TASK_COMM_LEN];
1420 printk(KERN_INFO "fork(): process `%s' used deprecated "
1421 "clone flags 0x%lx\n",
1422 get_task_comm(comm, current),
1423 clone_flags & CLONE_STOPPED);
1428 * When called from kernel_thread, don't do user tracing stuff.
1430 if (likely(user_mode(regs)))
1431 trace = tracehook_prepare_clone(clone_flags);
1433 p = copy_process(clone_flags, stack_start, regs, stack_size,
1434 child_tidptr, NULL, trace);
1436 * Do this prior waking up the new thread - the thread pointer
1437 * might get invalid after that point, if the thread exits quickly.
1440 struct completion vfork;
1442 trace_sched_process_fork(current, p);
1444 nr = task_pid_vnr(p);
1446 if (clone_flags & CLONE_PARENT_SETTID)
1447 put_user(nr, parent_tidptr);
1449 if (clone_flags & CLONE_VFORK) {
1450 p->vfork_done = &vfork;
1451 init_completion(&vfork);
1454 audit_finish_fork(p);
1455 tracehook_report_clone(regs, clone_flags, nr, p);
1458 * We set PF_STARTING at creation in case tracing wants to
1459 * use this to distinguish a fully live task from one that
1460 * hasn't gotten to tracehook_report_clone() yet. Now we
1461 * clear it and set the child going.
1463 p->flags &= ~PF_STARTING;
1465 if (unlikely(clone_flags & CLONE_STOPPED)) {
1467 * We'll start up with an immediate SIGSTOP.
1469 sigaddset(&p->pending.signal, SIGSTOP);
1470 set_tsk_thread_flag(p, TIF_SIGPENDING);
1471 __set_task_state(p, TASK_STOPPED);
1473 wake_up_new_task(p, clone_flags);
1476 tracehook_report_clone_complete(trace, regs,
1477 clone_flags, nr, p);
1479 if (clone_flags & CLONE_VFORK) {
1480 freezer_do_not_count();
1481 wait_for_completion(&vfork);
1483 tracehook_report_vfork_done(p, nr);
1491 #ifndef ARCH_MIN_MMSTRUCT_ALIGN
1492 #define ARCH_MIN_MMSTRUCT_ALIGN 0
1495 static void sighand_ctor(void *data)
1497 struct sighand_struct *sighand = data;
1499 spin_lock_init(&sighand->siglock);
1500 init_waitqueue_head(&sighand->signalfd_wqh);
1503 void __init proc_caches_init(void)
1505 sighand_cachep = kmem_cache_create("sighand_cache",
1506 sizeof(struct sighand_struct), 0,
1507 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_DESTROY_BY_RCU|
1508 SLAB_NOTRACK, sighand_ctor);
1509 signal_cachep = kmem_cache_create("signal_cache",
1510 sizeof(struct signal_struct), 0,
1511 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL);
1512 files_cachep = kmem_cache_create("files_cache",
1513 sizeof(struct files_struct), 0,
1514 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL);
1515 fs_cachep = kmem_cache_create("fs_cache",
1516 sizeof(struct fs_struct), 0,
1517 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL);
1518 mm_cachep = kmem_cache_create("mm_struct",
1519 sizeof(struct mm_struct), ARCH_MIN_MMSTRUCT_ALIGN,
1520 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL);
1521 vm_area_cachep = KMEM_CACHE(vm_area_struct, SLAB_PANIC);
1526 * Check constraints on flags passed to the unshare system call and
1527 * force unsharing of additional process context as appropriate.
1529 static void check_unshare_flags(unsigned long *flags_ptr)
1532 * If unsharing a thread from a thread group, must also
1535 if (*flags_ptr & CLONE_THREAD)
1536 *flags_ptr |= CLONE_VM;
1539 * If unsharing vm, must also unshare signal handlers.
1541 if (*flags_ptr & CLONE_VM)
1542 *flags_ptr |= CLONE_SIGHAND;
1545 * If unsharing namespace, must also unshare filesystem information.
1547 if (*flags_ptr & CLONE_NEWNS)
1548 *flags_ptr |= CLONE_FS;
1552 * Unsharing of tasks created with CLONE_THREAD is not supported yet
1554 static int unshare_thread(unsigned long unshare_flags)
1556 if (unshare_flags & CLONE_THREAD)
1563 * Unshare the filesystem structure if it is being shared
1565 static int unshare_fs(unsigned long unshare_flags, struct fs_struct **new_fsp)
1567 struct fs_struct *fs = current->fs;
1569 if (!(unshare_flags & CLONE_FS) || !fs)
1572 /* don't need lock here; in the worst case we'll do useless copy */
1576 *new_fsp = copy_fs_struct(fs);
1584 * Unsharing of sighand is not supported yet
1586 static int unshare_sighand(unsigned long unshare_flags, struct sighand_struct **new_sighp)
1588 struct sighand_struct *sigh = current->sighand;
1590 if ((unshare_flags & CLONE_SIGHAND) && atomic_read(&sigh->count) > 1)
1597 * Unshare vm if it is being shared
1599 static int unshare_vm(unsigned long unshare_flags, struct mm_struct **new_mmp)
1601 struct mm_struct *mm = current->mm;
1603 if ((unshare_flags & CLONE_VM) &&
1604 (mm && atomic_read(&mm->mm_users) > 1)) {
1612 * Unshare file descriptor table if it is being shared
1614 static int unshare_fd(unsigned long unshare_flags, struct files_struct **new_fdp)
1616 struct files_struct *fd = current->files;
1619 if ((unshare_flags & CLONE_FILES) &&
1620 (fd && atomic_read(&fd->count) > 1)) {
1621 *new_fdp = dup_fd(fd, &error);
1630 * unshare allows a process to 'unshare' part of the process
1631 * context which was originally shared using clone. copy_*
1632 * functions used by do_fork() cannot be used here directly
1633 * because they modify an inactive task_struct that is being
1634 * constructed. Here we are modifying the current, active,
1637 SYSCALL_DEFINE1(unshare, unsigned long, unshare_flags)
1640 struct fs_struct *fs, *new_fs = NULL;
1641 struct sighand_struct *new_sigh = NULL;
1642 struct mm_struct *mm, *new_mm = NULL, *active_mm = NULL;
1643 struct files_struct *fd, *new_fd = NULL;
1644 struct nsproxy *new_nsproxy = NULL;
1647 check_unshare_flags(&unshare_flags);
1649 /* Return -EINVAL for all unsupported flags */
1651 if (unshare_flags & ~(CLONE_THREAD|CLONE_FS|CLONE_NEWNS|CLONE_SIGHAND|
1652 CLONE_VM|CLONE_FILES|CLONE_SYSVSEM|
1653 CLONE_NEWUTS|CLONE_NEWIPC|CLONE_NEWNET))
1654 goto bad_unshare_out;
1657 * CLONE_NEWIPC must also detach from the undolist: after switching
1658 * to a new ipc namespace, the semaphore arrays from the old
1659 * namespace are unreachable.
1661 if (unshare_flags & (CLONE_NEWIPC|CLONE_SYSVSEM))
1663 if ((err = unshare_thread(unshare_flags)))
1664 goto bad_unshare_out;
1665 if ((err = unshare_fs(unshare_flags, &new_fs)))
1666 goto bad_unshare_cleanup_thread;
1667 if ((err = unshare_sighand(unshare_flags, &new_sigh)))
1668 goto bad_unshare_cleanup_fs;
1669 if ((err = unshare_vm(unshare_flags, &new_mm)))
1670 goto bad_unshare_cleanup_sigh;
1671 if ((err = unshare_fd(unshare_flags, &new_fd)))
1672 goto bad_unshare_cleanup_vm;
1673 if ((err = unshare_nsproxy_namespaces(unshare_flags, &new_nsproxy,
1675 goto bad_unshare_cleanup_fd;
1677 if (new_fs || new_mm || new_fd || do_sysvsem || new_nsproxy) {
1680 * CLONE_SYSVSEM is equivalent to sys_exit().
1686 switch_task_namespaces(current, new_nsproxy);
1694 write_lock(&fs->lock);
1695 current->fs = new_fs;
1700 write_unlock(&fs->lock);
1705 active_mm = current->active_mm;
1706 current->mm = new_mm;
1707 current->active_mm = new_mm;
1708 activate_mm(active_mm, new_mm);
1713 fd = current->files;
1714 current->files = new_fd;
1718 task_unlock(current);
1722 put_nsproxy(new_nsproxy);
1724 bad_unshare_cleanup_fd:
1726 put_files_struct(new_fd);
1728 bad_unshare_cleanup_vm:
1732 bad_unshare_cleanup_sigh:
1734 if (atomic_dec_and_test(&new_sigh->count))
1735 kmem_cache_free(sighand_cachep, new_sigh);
1737 bad_unshare_cleanup_fs:
1739 free_fs_struct(new_fs);
1741 bad_unshare_cleanup_thread:
1747 * Helper to unshare the files of the current task.
1748 * We don't want to expose copy_files internals to
1749 * the exec layer of the kernel.
1752 int unshare_files(struct files_struct **displaced)
1754 struct task_struct *task = current;
1755 struct files_struct *copy = NULL;
1758 error = unshare_fd(CLONE_FILES, ©);
1759 if (error || !copy) {
1763 *displaced = task->files;