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/seccomp.h>
38 #include <linux/swap.h>
39 #include <linux/syscalls.h>
40 #include <linux/jiffies.h>
41 #include <linux/futex.h>
42 #include <linux/compat.h>
43 #include <linux/kthread.h>
44 #include <linux/task_io_accounting_ops.h>
45 #include <linux/rcupdate.h>
46 #include <linux/ptrace.h>
47 #include <linux/mount.h>
48 #include <linux/audit.h>
49 #include <linux/memcontrol.h>
50 #include <linux/ftrace.h>
51 #include <linux/profile.h>
52 #include <linux/rmap.h>
53 #include <linux/ksm.h>
54 #include <linux/acct.h>
55 #include <linux/tsacct_kern.h>
56 #include <linux/cn_proc.h>
57 #include <linux/freezer.h>
58 #include <linux/delayacct.h>
59 #include <linux/taskstats_kern.h>
60 #include <linux/random.h>
61 #include <linux/tty.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>
68 #include <linux/oom.h>
69 #include <linux/khugepaged.h>
70 #include <linux/signalfd.h>
72 #include <asm/pgtable.h>
73 #include <asm/pgalloc.h>
74 #include <asm/uaccess.h>
75 #include <asm/mmu_context.h>
76 #include <asm/cacheflush.h>
77 #include <asm/tlbflush.h>
79 #include <trace/events/sched.h>
82 * Protected counters by write_lock_irq(&tasklist_lock)
84 unsigned long total_forks; /* Handle normal Linux uptimes. */
85 int nr_threads; /* The idle threads do not count.. */
87 int max_threads; /* tunable limit on nr_threads */
89 DEFINE_PER_CPU(unsigned long, process_counts) = 0;
91 __cacheline_aligned DEFINE_RWLOCK(tasklist_lock); /* outer */
93 #ifdef CONFIG_PROVE_RCU
94 int lockdep_tasklist_lock_is_held(void)
96 return lockdep_is_held(&tasklist_lock);
98 EXPORT_SYMBOL_GPL(lockdep_tasklist_lock_is_held);
99 #endif /* #ifdef CONFIG_PROVE_RCU */
101 int nr_processes(void)
106 for_each_possible_cpu(cpu)
107 total += per_cpu(process_counts, cpu);
112 #ifndef __HAVE_ARCH_TASK_STRUCT_ALLOCATOR
113 # define alloc_task_struct_node(node) \
114 kmem_cache_alloc_node(task_struct_cachep, GFP_KERNEL, node)
115 # define free_task_struct(tsk) \
116 kmem_cache_free(task_struct_cachep, (tsk))
117 static struct kmem_cache *task_struct_cachep;
120 #ifndef __HAVE_ARCH_THREAD_INFO_ALLOCATOR
121 static struct thread_info *alloc_thread_info_node(struct task_struct *tsk,
124 #ifdef CONFIG_DEBUG_STACK_USAGE
125 gfp_t mask = GFP_KERNEL | __GFP_ZERO;
127 gfp_t mask = GFP_KERNEL;
129 struct page *page = alloc_pages_node(node, mask, THREAD_SIZE_ORDER);
131 return page ? page_address(page) : NULL;
134 static inline void free_thread_info(struct thread_info *ti)
136 free_pages((unsigned long)ti, THREAD_SIZE_ORDER);
140 /* SLAB cache for signal_struct structures (tsk->signal) */
141 static struct kmem_cache *signal_cachep;
143 /* SLAB cache for sighand_struct structures (tsk->sighand) */
144 struct kmem_cache *sighand_cachep;
146 /* SLAB cache for files_struct structures (tsk->files) */
147 struct kmem_cache *files_cachep;
149 /* SLAB cache for fs_struct structures (tsk->fs) */
150 struct kmem_cache *fs_cachep;
152 /* SLAB cache for vm_area_struct structures */
153 struct kmem_cache *vm_area_cachep;
155 /* SLAB cache for mm_struct structures (tsk->mm) */
156 static struct kmem_cache *mm_cachep;
158 static void account_kernel_stack(struct thread_info *ti, int account)
160 struct zone *zone = page_zone(virt_to_page(ti));
162 mod_zone_page_state(zone, NR_KERNEL_STACK, account);
165 void free_task(struct task_struct *tsk)
167 account_kernel_stack(tsk->stack, -1);
168 free_thread_info(tsk->stack);
169 rt_mutex_debug_task_free(tsk);
170 ftrace_graph_exit_task(tsk);
171 put_seccomp_filter(tsk);
172 free_task_struct(tsk);
174 EXPORT_SYMBOL(free_task);
176 static inline void free_signal_struct(struct signal_struct *sig)
178 taskstats_tgid_free(sig);
179 sched_autogroup_exit(sig);
180 kmem_cache_free(signal_cachep, sig);
183 static inline void put_signal_struct(struct signal_struct *sig)
185 if (atomic_dec_and_test(&sig->sigcnt))
186 free_signal_struct(sig);
189 void __put_task_struct(struct task_struct *tsk)
191 WARN_ON(!tsk->exit_state);
192 WARN_ON(atomic_read(&tsk->usage));
193 WARN_ON(tsk == current);
195 security_task_free(tsk);
197 delayacct_tsk_free(tsk);
198 put_signal_struct(tsk->signal);
200 if (!profile_handoff_task(tsk))
203 EXPORT_SYMBOL_GPL(__put_task_struct);
206 * macro override instead of weak attribute alias, to workaround
207 * gcc 4.1.0 and 4.1.1 bugs with weak attribute and empty functions.
209 #ifndef arch_task_cache_init
210 #define arch_task_cache_init()
213 void __init fork_init(unsigned long mempages)
215 #ifndef __HAVE_ARCH_TASK_STRUCT_ALLOCATOR
216 #ifndef ARCH_MIN_TASKALIGN
217 #define ARCH_MIN_TASKALIGN L1_CACHE_BYTES
219 /* create a slab on which task_structs can be allocated */
221 kmem_cache_create("task_struct", sizeof(struct task_struct),
222 ARCH_MIN_TASKALIGN, SLAB_PANIC | SLAB_NOTRACK, NULL);
225 /* do the arch specific task caches init */
226 arch_task_cache_init();
229 * The default maximum number of threads is set to a safe
230 * value: the thread structures can take up at most half
233 max_threads = mempages / (8 * THREAD_SIZE / PAGE_SIZE);
236 * we need to allow at least 20 threads to boot a system
238 if (max_threads < 20)
241 init_task.signal->rlim[RLIMIT_NPROC].rlim_cur = max_threads/2;
242 init_task.signal->rlim[RLIMIT_NPROC].rlim_max = max_threads/2;
243 init_task.signal->rlim[RLIMIT_SIGPENDING] =
244 init_task.signal->rlim[RLIMIT_NPROC];
247 int __attribute__((weak)) arch_dup_task_struct(struct task_struct *dst,
248 struct task_struct *src)
254 static struct task_struct *dup_task_struct(struct task_struct *orig)
256 struct task_struct *tsk;
257 struct thread_info *ti;
258 unsigned long *stackend;
259 int node = tsk_fork_get_node(orig);
262 prepare_to_copy(orig);
264 tsk = alloc_task_struct_node(node);
268 ti = alloc_thread_info_node(tsk, node);
270 free_task_struct(tsk);
274 err = arch_dup_task_struct(tsk, orig);
280 setup_thread_stack(tsk, orig);
281 clear_user_return_notifier(tsk);
282 clear_tsk_need_resched(tsk);
283 stackend = end_of_stack(tsk);
284 *stackend = STACK_END_MAGIC; /* for overflow detection */
286 #ifdef CONFIG_CC_STACKPROTECTOR
287 tsk->stack_canary = get_random_int();
291 * One for us, one for whoever does the "release_task()" (usually
294 atomic_set(&tsk->usage, 2);
295 #ifdef CONFIG_BLK_DEV_IO_TRACE
298 tsk->splice_pipe = NULL;
300 account_kernel_stack(ti, 1);
305 free_thread_info(ti);
306 free_task_struct(tsk);
311 static int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm)
313 struct vm_area_struct *mpnt, *tmp, *prev, **pprev;
314 struct rb_node **rb_link, *rb_parent;
316 unsigned long charge;
317 struct mempolicy *pol;
319 down_write(&oldmm->mmap_sem);
320 flush_cache_dup_mm(oldmm);
322 * Not linked in yet - no deadlock potential:
324 down_write_nested(&mm->mmap_sem, SINGLE_DEPTH_NESTING);
328 mm->mmap_cache = NULL;
329 mm->free_area_cache = oldmm->mmap_base;
330 mm->cached_hole_size = ~0UL;
332 cpumask_clear(mm_cpumask(mm));
334 rb_link = &mm->mm_rb.rb_node;
337 retval = ksm_fork(mm, oldmm);
340 retval = khugepaged_fork(mm, oldmm);
345 for (mpnt = oldmm->mmap; mpnt; mpnt = mpnt->vm_next) {
348 if (mpnt->vm_flags & VM_DONTCOPY) {
349 long pages = vma_pages(mpnt);
350 mm->total_vm -= pages;
351 vm_stat_account(mm, mpnt->vm_flags, mpnt->vm_file,
356 if (mpnt->vm_flags & VM_ACCOUNT) {
357 unsigned int len = (mpnt->vm_end - mpnt->vm_start) >> PAGE_SHIFT;
358 if (security_vm_enough_memory(len))
362 tmp = kmem_cache_alloc(vm_area_cachep, GFP_KERNEL);
366 INIT_LIST_HEAD(&tmp->anon_vma_chain);
367 pol = mpol_dup(vma_policy(mpnt));
368 retval = PTR_ERR(pol);
370 goto fail_nomem_policy;
371 vma_set_policy(tmp, pol);
373 if (anon_vma_fork(tmp, mpnt))
374 goto fail_nomem_anon_vma_fork;
375 tmp->vm_flags &= ~VM_LOCKED;
376 tmp->vm_next = tmp->vm_prev = NULL;
379 struct inode *inode = file->f_path.dentry->d_inode;
380 struct address_space *mapping = file->f_mapping;
383 if (tmp->vm_flags & VM_DENYWRITE)
384 atomic_dec(&inode->i_writecount);
385 mutex_lock(&mapping->i_mmap_mutex);
386 if (tmp->vm_flags & VM_SHARED)
387 mapping->i_mmap_writable++;
388 flush_dcache_mmap_lock(mapping);
389 /* insert tmp into the share list, just after mpnt */
390 vma_prio_tree_add(tmp, mpnt);
391 flush_dcache_mmap_unlock(mapping);
392 mutex_unlock(&mapping->i_mmap_mutex);
396 * Clear hugetlb-related page reserves for children. This only
397 * affects MAP_PRIVATE mappings. Faults generated by the child
398 * are not guaranteed to succeed, even if read-only
400 if (is_vm_hugetlb_page(tmp))
401 reset_vma_resv_huge_pages(tmp);
404 * Link in the new vma and copy the page table entries.
407 pprev = &tmp->vm_next;
411 __vma_link_rb(mm, tmp, rb_link, rb_parent);
412 rb_link = &tmp->vm_rb.rb_right;
413 rb_parent = &tmp->vm_rb;
416 retval = copy_page_range(mm, oldmm, mpnt);
418 if (tmp->vm_ops && tmp->vm_ops->open)
419 tmp->vm_ops->open(tmp);
424 /* a new mm has just been created */
425 arch_dup_mmap(oldmm, mm);
428 up_write(&mm->mmap_sem);
430 up_write(&oldmm->mmap_sem);
432 fail_nomem_anon_vma_fork:
435 kmem_cache_free(vm_area_cachep, tmp);
438 vm_unacct_memory(charge);
442 static inline int mm_alloc_pgd(struct mm_struct *mm)
444 mm->pgd = pgd_alloc(mm);
445 if (unlikely(!mm->pgd))
450 static inline void mm_free_pgd(struct mm_struct *mm)
452 pgd_free(mm, mm->pgd);
455 #define dup_mmap(mm, oldmm) (0)
456 #define mm_alloc_pgd(mm) (0)
457 #define mm_free_pgd(mm)
458 #endif /* CONFIG_MMU */
460 __cacheline_aligned_in_smp DEFINE_SPINLOCK(mmlist_lock);
462 #define allocate_mm() (kmem_cache_alloc(mm_cachep, GFP_KERNEL))
463 #define free_mm(mm) (kmem_cache_free(mm_cachep, (mm)))
465 static unsigned long default_dump_filter = MMF_DUMP_FILTER_DEFAULT;
467 static int __init coredump_filter_setup(char *s)
469 default_dump_filter =
470 (simple_strtoul(s, NULL, 0) << MMF_DUMP_FILTER_SHIFT) &
471 MMF_DUMP_FILTER_MASK;
475 __setup("coredump_filter=", coredump_filter_setup);
477 #include <linux/init_task.h>
479 static void mm_init_aio(struct mm_struct *mm)
482 spin_lock_init(&mm->ioctx_lock);
483 INIT_HLIST_HEAD(&mm->ioctx_list);
487 static struct mm_struct *mm_init(struct mm_struct *mm, struct task_struct *p)
489 atomic_set(&mm->mm_users, 1);
490 atomic_set(&mm->mm_count, 1);
491 init_rwsem(&mm->mmap_sem);
492 INIT_LIST_HEAD(&mm->mmlist);
493 mm->flags = (current->mm) ?
494 (current->mm->flags & MMF_INIT_MASK) : default_dump_filter;
495 mm->core_state = NULL;
497 memset(&mm->rss_stat, 0, sizeof(mm->rss_stat));
498 spin_lock_init(&mm->page_table_lock);
499 mm->free_area_cache = TASK_UNMAPPED_BASE;
500 mm->cached_hole_size = ~0UL;
502 mm_init_owner(mm, p);
504 if (likely(!mm_alloc_pgd(mm))) {
506 mmu_notifier_mm_init(mm);
515 * Allocate and initialize an mm_struct.
517 struct mm_struct *mm_alloc(void)
519 struct mm_struct *mm;
525 memset(mm, 0, sizeof(*mm));
527 return 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);
541 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
542 VM_BUG_ON(mm->pmd_huge_pte);
546 EXPORT_SYMBOL_GPL(__mmdrop);
549 * Decrement the use count and release all resources for an mm.
551 void mmput(struct mm_struct *mm)
555 if (atomic_dec_and_test(&mm->mm_users)) {
558 khugepaged_exit(mm); /* must run before exit_mmap */
560 set_mm_exe_file(mm, NULL);
561 if (!list_empty(&mm->mmlist)) {
562 spin_lock(&mmlist_lock);
563 list_del(&mm->mmlist);
564 spin_unlock(&mmlist_lock);
568 module_put(mm->binfmt->module);
572 EXPORT_SYMBOL_GPL(mmput);
575 * We added or removed a vma mapping the executable. The vmas are only mapped
576 * during exec and are not mapped with the mmap system call.
577 * Callers must hold down_write() on the mm's mmap_sem for these
579 void added_exe_file_vma(struct mm_struct *mm)
581 mm->num_exe_file_vmas++;
584 void removed_exe_file_vma(struct mm_struct *mm)
586 mm->num_exe_file_vmas--;
587 if ((mm->num_exe_file_vmas == 0) && mm->exe_file) {
594 void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file)
597 get_file(new_exe_file);
600 mm->exe_file = new_exe_file;
601 mm->num_exe_file_vmas = 0;
604 struct file *get_mm_exe_file(struct mm_struct *mm)
606 struct file *exe_file;
608 /* We need mmap_sem to protect against races with removal of
609 * VM_EXECUTABLE vmas */
610 down_read(&mm->mmap_sem);
611 exe_file = mm->exe_file;
614 up_read(&mm->mmap_sem);
618 static void dup_mm_exe_file(struct mm_struct *oldmm, struct mm_struct *newmm)
620 /* It's safe to write the exe_file pointer without exe_file_lock because
621 * this is called during fork when the task is not yet in /proc */
622 newmm->exe_file = get_mm_exe_file(oldmm);
626 * get_task_mm - acquire a reference to the task's mm
628 * Returns %NULL if the task has no mm. Checks PF_KTHREAD (meaning
629 * this kernel workthread has transiently adopted a user mm with use_mm,
630 * to do its AIO) is not set and if so returns a reference to it, after
631 * bumping up the use count. User must release the mm via mmput()
632 * after use. Typically used by /proc and ptrace.
634 struct mm_struct *get_task_mm(struct task_struct *task)
636 struct mm_struct *mm;
641 if (task->flags & PF_KTHREAD)
644 atomic_inc(&mm->mm_users);
649 EXPORT_SYMBOL_GPL(get_task_mm);
651 /* Please note the differences between mmput and mm_release.
652 * mmput is called whenever we stop holding onto a mm_struct,
653 * error success whatever.
655 * mm_release is called after a mm_struct has been removed
656 * from the current process.
658 * This difference is important for error handling, when we
659 * only half set up a mm_struct for a new process and need to restore
660 * the old one. Because we mmput the new mm_struct before
661 * restoring the old one. . .
662 * Eric Biederman 10 January 1998
664 void mm_release(struct task_struct *tsk, struct mm_struct *mm)
666 struct completion *vfork_done = tsk->vfork_done;
668 /* Get rid of any futexes when releasing the mm */
670 if (unlikely(tsk->robust_list)) {
671 exit_robust_list(tsk);
672 tsk->robust_list = NULL;
675 if (unlikely(tsk->compat_robust_list)) {
676 compat_exit_robust_list(tsk);
677 tsk->compat_robust_list = NULL;
680 if (unlikely(!list_empty(&tsk->pi_state_list)))
681 exit_pi_state_list(tsk);
684 /* Get rid of any cached register state */
685 deactivate_mm(tsk, mm);
687 /* notify parent sleeping on vfork() */
689 tsk->vfork_done = NULL;
690 complete(vfork_done);
694 * If we're exiting normally, clear a user-space tid field if
695 * requested. We leave this alone when dying by signal, to leave
696 * the value intact in a core dump, and to save the unnecessary
697 * trouble otherwise. Userland only wants this done for a sys_exit.
699 if (tsk->clear_child_tid) {
700 if (!(tsk->flags & PF_SIGNALED) &&
701 atomic_read(&mm->mm_users) > 1) {
703 * We don't check the error code - if userspace has
704 * not set up a proper pointer then tough luck.
706 put_user(0, tsk->clear_child_tid);
707 sys_futex(tsk->clear_child_tid, FUTEX_WAKE,
710 tsk->clear_child_tid = NULL;
715 * Allocate a new mm structure and copy contents from the
716 * mm structure of the passed in task structure.
718 struct mm_struct *dup_mm(struct task_struct *tsk)
720 struct mm_struct *mm, *oldmm = current->mm;
730 memcpy(mm, oldmm, sizeof(*mm));
733 /* Initializing for Swap token stuff */
734 mm->token_priority = 0;
735 mm->last_interval = 0;
737 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
738 mm->pmd_huge_pte = NULL;
741 if (!mm_init(mm, tsk))
744 if (init_new_context(tsk, mm))
747 dup_mm_exe_file(oldmm, mm);
749 err = dup_mmap(mm, oldmm);
753 mm->hiwater_rss = get_mm_rss(mm);
754 mm->hiwater_vm = mm->total_vm;
756 if (mm->binfmt && !try_module_get(mm->binfmt->module))
762 /* don't put binfmt in mmput, we haven't got module yet */
771 * If init_new_context() failed, we cannot use mmput() to free the mm
772 * because it calls destroy_context()
779 static int copy_mm(unsigned long clone_flags, struct task_struct *tsk)
781 struct mm_struct *mm, *oldmm;
784 tsk->min_flt = tsk->maj_flt = 0;
785 tsk->nvcsw = tsk->nivcsw = 0;
786 #ifdef CONFIG_DETECT_HUNG_TASK
787 tsk->last_switch_count = tsk->nvcsw + tsk->nivcsw;
791 tsk->active_mm = NULL;
794 * Are we cloning a kernel thread?
796 * We need to steal a active VM for that..
802 if (clone_flags & CLONE_VM) {
803 atomic_inc(&oldmm->mm_users);
814 /* Initializing for Swap token stuff */
815 mm->token_priority = 0;
816 mm->last_interval = 0;
826 static int copy_fs(unsigned long clone_flags, struct task_struct *tsk)
828 struct fs_struct *fs = current->fs;
829 if (clone_flags & CLONE_FS) {
830 /* tsk->fs is already what we want */
831 spin_lock(&fs->lock);
833 spin_unlock(&fs->lock);
837 spin_unlock(&fs->lock);
840 tsk->fs = copy_fs_struct(fs);
846 static int copy_files(unsigned long clone_flags, struct task_struct *tsk)
848 struct files_struct *oldf, *newf;
852 * A background process may not have any files ...
854 oldf = current->files;
858 if (clone_flags & CLONE_FILES) {
859 atomic_inc(&oldf->count);
863 newf = dup_fd(oldf, &error);
873 static int copy_io(unsigned long clone_flags, struct task_struct *tsk)
876 struct io_context *ioc = current->io_context;
881 * Share io context with parent, if CLONE_IO is set
883 if (clone_flags & CLONE_IO) {
884 tsk->io_context = ioc_task_link(ioc);
885 if (unlikely(!tsk->io_context))
887 } else if (ioprio_valid(ioc->ioprio)) {
888 tsk->io_context = alloc_io_context(GFP_KERNEL, -1);
889 if (unlikely(!tsk->io_context))
892 tsk->io_context->ioprio = ioc->ioprio;
898 static int copy_sighand(unsigned long clone_flags, struct task_struct *tsk)
900 struct sighand_struct *sig;
902 if (clone_flags & CLONE_SIGHAND) {
903 atomic_inc(¤t->sighand->count);
906 sig = kmem_cache_alloc(sighand_cachep, GFP_KERNEL);
907 rcu_assign_pointer(tsk->sighand, sig);
910 atomic_set(&sig->count, 1);
911 memcpy(sig->action, current->sighand->action, sizeof(sig->action));
915 void __cleanup_sighand(struct sighand_struct *sighand)
917 if (atomic_dec_and_test(&sighand->count)) {
918 signalfd_cleanup(sighand);
919 kmem_cache_free(sighand_cachep, sighand);
925 * Initialize POSIX timer handling for a thread group.
927 static void posix_cpu_timers_init_group(struct signal_struct *sig)
929 unsigned long cpu_limit;
931 /* Thread group counters. */
932 thread_group_cputime_init(sig);
934 cpu_limit = ACCESS_ONCE(sig->rlim[RLIMIT_CPU].rlim_cur);
935 if (cpu_limit != RLIM_INFINITY) {
936 sig->cputime_expires.prof_exp = secs_to_cputime(cpu_limit);
937 sig->cputimer.running = 1;
940 /* The timer lists. */
941 INIT_LIST_HEAD(&sig->cpu_timers[0]);
942 INIT_LIST_HEAD(&sig->cpu_timers[1]);
943 INIT_LIST_HEAD(&sig->cpu_timers[2]);
946 static int copy_signal(unsigned long clone_flags, struct task_struct *tsk)
948 struct signal_struct *sig;
950 if (clone_flags & CLONE_THREAD)
953 sig = kmem_cache_zalloc(signal_cachep, GFP_KERNEL);
959 atomic_set(&sig->live, 1);
960 atomic_set(&sig->sigcnt, 1);
961 init_waitqueue_head(&sig->wait_chldexit);
962 if (clone_flags & CLONE_NEWPID)
963 sig->flags |= SIGNAL_UNKILLABLE;
964 sig->curr_target = tsk;
965 init_sigpending(&sig->shared_pending);
966 INIT_LIST_HEAD(&sig->posix_timers);
968 hrtimer_init(&sig->real_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
969 sig->real_timer.function = it_real_fn;
971 task_lock(current->group_leader);
972 memcpy(sig->rlim, current->signal->rlim, sizeof sig->rlim);
973 task_unlock(current->group_leader);
975 posix_cpu_timers_init_group(sig);
978 sched_autogroup_fork(sig);
980 #ifdef CONFIG_CGROUPS
981 init_rwsem(&sig->threadgroup_fork_lock);
984 sig->oom_adj = current->signal->oom_adj;
985 sig->oom_score_adj = current->signal->oom_score_adj;
986 sig->oom_score_adj_min = current->signal->oom_score_adj_min;
988 mutex_init(&sig->cred_guard_mutex);
993 static void copy_flags(unsigned long clone_flags, struct task_struct *p)
995 unsigned long new_flags = p->flags;
997 new_flags &= ~(PF_SUPERPRIV | PF_WQ_WORKER);
998 new_flags |= PF_FORKNOEXEC;
999 new_flags |= PF_STARTING;
1000 p->flags = new_flags;
1001 clear_freeze_flag(p);
1004 SYSCALL_DEFINE1(set_tid_address, int __user *, tidptr)
1006 current->clear_child_tid = tidptr;
1008 return task_pid_vnr(current);
1011 static void rt_mutex_init_task(struct task_struct *p)
1013 raw_spin_lock_init(&p->pi_lock);
1014 #ifdef CONFIG_RT_MUTEXES
1015 plist_head_init(&p->pi_waiters);
1016 p->pi_blocked_on = NULL;
1020 #ifdef CONFIG_MM_OWNER
1021 void mm_init_owner(struct mm_struct *mm, struct task_struct *p)
1025 #endif /* CONFIG_MM_OWNER */
1028 * Initialize POSIX timer handling for a single task.
1030 static void posix_cpu_timers_init(struct task_struct *tsk)
1032 tsk->cputime_expires.prof_exp = cputime_zero;
1033 tsk->cputime_expires.virt_exp = cputime_zero;
1034 tsk->cputime_expires.sched_exp = 0;
1035 INIT_LIST_HEAD(&tsk->cpu_timers[0]);
1036 INIT_LIST_HEAD(&tsk->cpu_timers[1]);
1037 INIT_LIST_HEAD(&tsk->cpu_timers[2]);
1041 * This creates a new process as a copy of the old one,
1042 * but does not actually start it yet.
1044 * It copies the registers, and all the appropriate
1045 * parts of the process environment (as per the clone
1046 * flags). The actual kick-off is left to the caller.
1048 static struct task_struct *copy_process(unsigned long clone_flags,
1049 unsigned long stack_start,
1050 struct pt_regs *regs,
1051 unsigned long stack_size,
1052 int __user *child_tidptr,
1057 struct task_struct *p;
1058 int cgroup_callbacks_done = 0;
1060 if ((clone_flags & (CLONE_NEWNS|CLONE_FS)) == (CLONE_NEWNS|CLONE_FS))
1061 return ERR_PTR(-EINVAL);
1064 * Thread groups must share signals as well, and detached threads
1065 * can only be started up within the thread group.
1067 if ((clone_flags & CLONE_THREAD) && !(clone_flags & CLONE_SIGHAND))
1068 return ERR_PTR(-EINVAL);
1071 * Shared signal handlers imply shared VM. By way of the above,
1072 * thread groups also imply shared VM. Blocking this case allows
1073 * for various simplifications in other code.
1075 if ((clone_flags & CLONE_SIGHAND) && !(clone_flags & CLONE_VM))
1076 return ERR_PTR(-EINVAL);
1079 * Siblings of global init remain as zombies on exit since they are
1080 * not reaped by their parent (swapper). To solve this and to avoid
1081 * multi-rooted process trees, prevent global and container-inits
1082 * from creating siblings.
1084 if ((clone_flags & CLONE_PARENT) &&
1085 current->signal->flags & SIGNAL_UNKILLABLE)
1086 return ERR_PTR(-EINVAL);
1088 retval = security_task_create(clone_flags);
1093 p = dup_task_struct(current);
1097 ftrace_graph_init_task(p);
1098 get_seccomp_filter(p);
1100 rt_mutex_init_task(p);
1102 #ifdef CONFIG_PROVE_LOCKING
1103 DEBUG_LOCKS_WARN_ON(!p->hardirqs_enabled);
1104 DEBUG_LOCKS_WARN_ON(!p->softirqs_enabled);
1107 if (atomic_read(&p->real_cred->user->processes) >=
1108 task_rlimit(p, RLIMIT_NPROC)) {
1109 if (!capable(CAP_SYS_ADMIN) && !capable(CAP_SYS_RESOURCE) &&
1110 p->real_cred->user != INIT_USER)
1113 current->flags &= ~PF_NPROC_EXCEEDED;
1115 retval = copy_creds(p, clone_flags);
1120 * If multiple threads are within copy_process(), then this check
1121 * triggers too late. This doesn't hurt, the check is only there
1122 * to stop root fork bombs.
1125 if (nr_threads >= max_threads)
1126 goto bad_fork_cleanup_count;
1128 if (!try_module_get(task_thread_info(p)->exec_domain->module))
1129 goto bad_fork_cleanup_count;
1132 delayacct_tsk_init(p); /* Must remain after dup_task_struct() */
1133 copy_flags(clone_flags, p);
1134 INIT_LIST_HEAD(&p->children);
1135 INIT_LIST_HEAD(&p->sibling);
1136 rcu_copy_process(p);
1137 p->vfork_done = NULL;
1138 spin_lock_init(&p->alloc_lock);
1140 init_sigpending(&p->pending);
1142 p->utime = cputime_zero;
1143 p->stime = cputime_zero;
1144 p->gtime = cputime_zero;
1145 p->utimescaled = cputime_zero;
1146 p->stimescaled = cputime_zero;
1147 #ifndef CONFIG_VIRT_CPU_ACCOUNTING
1148 p->prev_utime = cputime_zero;
1149 p->prev_stime = cputime_zero;
1151 #if defined(SPLIT_RSS_COUNTING)
1152 memset(&p->rss_stat, 0, sizeof(p->rss_stat));
1155 p->default_timer_slack_ns = current->timer_slack_ns;
1157 task_io_accounting_init(&p->ioac);
1158 acct_clear_integrals(p);
1160 posix_cpu_timers_init(p);
1162 do_posix_clock_monotonic_gettime(&p->start_time);
1163 p->real_start_time = p->start_time;
1164 monotonic_to_bootbased(&p->real_start_time);
1165 p->io_context = NULL;
1166 p->audit_context = NULL;
1167 if (clone_flags & CLONE_THREAD)
1168 threadgroup_fork_read_lock(current);
1171 p->mempolicy = mpol_dup(p->mempolicy);
1172 if (IS_ERR(p->mempolicy)) {
1173 retval = PTR_ERR(p->mempolicy);
1174 p->mempolicy = NULL;
1175 goto bad_fork_cleanup_cgroup;
1177 mpol_fix_fork_child_flag(p);
1179 #ifdef CONFIG_CPUSETS
1180 p->cpuset_mem_spread_rotor = NUMA_NO_NODE;
1181 p->cpuset_slab_spread_rotor = NUMA_NO_NODE;
1183 #ifdef CONFIG_TRACE_IRQFLAGS
1185 #ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW
1186 p->hardirqs_enabled = 1;
1188 p->hardirqs_enabled = 0;
1190 p->hardirq_enable_ip = 0;
1191 p->hardirq_enable_event = 0;
1192 p->hardirq_disable_ip = _THIS_IP_;
1193 p->hardirq_disable_event = 0;
1194 p->softirqs_enabled = 1;
1195 p->softirq_enable_ip = _THIS_IP_;
1196 p->softirq_enable_event = 0;
1197 p->softirq_disable_ip = 0;
1198 p->softirq_disable_event = 0;
1199 p->hardirq_context = 0;
1200 p->softirq_context = 0;
1202 #ifdef CONFIG_LOCKDEP
1203 p->lockdep_depth = 0; /* no locks held yet */
1204 p->curr_chain_key = 0;
1205 p->lockdep_recursion = 0;
1208 #ifdef CONFIG_DEBUG_MUTEXES
1209 p->blocked_on = NULL; /* not blocked yet */
1211 #ifdef CONFIG_CGROUP_MEM_RES_CTLR
1212 p->memcg_batch.do_batch = 0;
1213 p->memcg_batch.memcg = NULL;
1216 /* Perform scheduler related setup. Assign this task to a CPU. */
1219 retval = perf_event_init_task(p);
1221 goto bad_fork_cleanup_policy;
1222 retval = audit_alloc(p);
1224 goto bad_fork_cleanup_policy;
1225 /* copy all the process information */
1226 retval = copy_semundo(clone_flags, p);
1228 goto bad_fork_cleanup_audit;
1229 retval = copy_files(clone_flags, p);
1231 goto bad_fork_cleanup_semundo;
1232 retval = copy_fs(clone_flags, p);
1234 goto bad_fork_cleanup_files;
1235 retval = copy_sighand(clone_flags, p);
1237 goto bad_fork_cleanup_fs;
1238 retval = copy_signal(clone_flags, p);
1240 goto bad_fork_cleanup_sighand;
1241 retval = copy_mm(clone_flags, p);
1243 goto bad_fork_cleanup_signal;
1244 retval = copy_namespaces(clone_flags, p);
1246 goto bad_fork_cleanup_mm;
1247 retval = copy_io(clone_flags, p);
1249 goto bad_fork_cleanup_namespaces;
1250 retval = copy_thread(clone_flags, stack_start, stack_size, p, regs);
1252 goto bad_fork_cleanup_io;
1254 if (pid != &init_struct_pid) {
1256 pid = alloc_pid(p->nsproxy->pid_ns);
1258 goto bad_fork_cleanup_io;
1261 p->pid = pid_nr(pid);
1263 if (clone_flags & CLONE_THREAD)
1264 p->tgid = current->tgid;
1266 p->set_child_tid = (clone_flags & CLONE_CHILD_SETTID) ? child_tidptr : NULL;
1268 * Clear TID on mm_release()?
1270 p->clear_child_tid = (clone_flags & CLONE_CHILD_CLEARTID) ? child_tidptr : NULL;
1275 p->robust_list = NULL;
1276 #ifdef CONFIG_COMPAT
1277 p->compat_robust_list = NULL;
1279 INIT_LIST_HEAD(&p->pi_state_list);
1280 p->pi_state_cache = NULL;
1283 * sigaltstack should be cleared when sharing the same VM
1285 if ((clone_flags & (CLONE_VM|CLONE_VFORK)) == CLONE_VM)
1286 p->sas_ss_sp = p->sas_ss_size = 0;
1289 * Syscall tracing and stepping should be turned off in the
1290 * child regardless of CLONE_PTRACE.
1292 user_disable_single_step(p);
1293 clear_tsk_thread_flag(p, TIF_SYSCALL_TRACE);
1294 #ifdef TIF_SYSCALL_EMU
1295 clear_tsk_thread_flag(p, TIF_SYSCALL_EMU);
1297 clear_all_latency_tracing(p);
1299 /* ok, now we should be set up.. */
1300 p->exit_signal = (clone_flags & CLONE_THREAD) ? -1 : (clone_flags & CSIGNAL);
1301 p->pdeath_signal = 0;
1305 p->nr_dirtied_pause = 128 >> (PAGE_SHIFT - 10);
1308 * Ok, make it visible to the rest of the system.
1309 * We dont wake it up yet.
1311 p->group_leader = p;
1312 INIT_LIST_HEAD(&p->thread_group);
1314 /* Now that the task is set up, run cgroup callbacks if
1315 * necessary. We need to run them before the task is visible
1316 * on the tasklist. */
1317 cgroup_fork_callbacks(p);
1318 cgroup_callbacks_done = 1;
1320 /* Need tasklist lock for parent etc handling! */
1321 write_lock_irq(&tasklist_lock);
1323 /* CLONE_PARENT re-uses the old parent */
1324 if (clone_flags & (CLONE_PARENT|CLONE_THREAD)) {
1325 p->real_parent = current->real_parent;
1326 p->parent_exec_id = current->parent_exec_id;
1328 p->real_parent = current;
1329 p->parent_exec_id = current->self_exec_id;
1332 spin_lock(¤t->sighand->siglock);
1335 * Process group and session signals need to be delivered to just the
1336 * parent before the fork or both the parent and the child after the
1337 * fork. Restart if a signal comes in before we add the new process to
1338 * it's process group.
1339 * A fatal signal pending means that current will exit, so the new
1340 * thread can't slip out of an OOM kill (or normal SIGKILL).
1342 recalc_sigpending();
1343 if (signal_pending(current)) {
1344 spin_unlock(¤t->sighand->siglock);
1345 write_unlock_irq(&tasklist_lock);
1346 retval = -ERESTARTNOINTR;
1347 goto bad_fork_free_pid;
1350 if (clone_flags & CLONE_THREAD) {
1351 current->signal->nr_threads++;
1352 atomic_inc(¤t->signal->live);
1353 atomic_inc(¤t->signal->sigcnt);
1354 p->group_leader = current->group_leader;
1355 list_add_tail_rcu(&p->thread_group, &p->group_leader->thread_group);
1358 if (likely(p->pid)) {
1359 ptrace_init_task(p, (clone_flags & CLONE_PTRACE) || trace);
1361 if (thread_group_leader(p)) {
1362 if (is_child_reaper(pid))
1363 p->nsproxy->pid_ns->child_reaper = p;
1365 p->signal->leader_pid = pid;
1366 p->signal->tty = tty_kref_get(current->signal->tty);
1367 attach_pid(p, PIDTYPE_PGID, task_pgrp(current));
1368 attach_pid(p, PIDTYPE_SID, task_session(current));
1369 list_add_tail(&p->sibling, &p->real_parent->children);
1370 list_add_tail_rcu(&p->tasks, &init_task.tasks);
1371 __this_cpu_inc(process_counts);
1373 attach_pid(p, PIDTYPE_PID, pid);
1378 spin_unlock(¤t->sighand->siglock);
1379 write_unlock_irq(&tasklist_lock);
1380 proc_fork_connector(p);
1381 cgroup_post_fork(p);
1382 if (clone_flags & CLONE_THREAD)
1383 threadgroup_fork_read_unlock(current);
1388 if (pid != &init_struct_pid)
1390 bad_fork_cleanup_io:
1393 bad_fork_cleanup_namespaces:
1394 exit_task_namespaces(p);
1395 bad_fork_cleanup_mm:
1398 bad_fork_cleanup_signal:
1399 if (!(clone_flags & CLONE_THREAD))
1400 free_signal_struct(p->signal);
1401 bad_fork_cleanup_sighand:
1402 __cleanup_sighand(p->sighand);
1403 bad_fork_cleanup_fs:
1404 exit_fs(p); /* blocking */
1405 bad_fork_cleanup_files:
1406 exit_files(p); /* blocking */
1407 bad_fork_cleanup_semundo:
1409 bad_fork_cleanup_audit:
1411 bad_fork_cleanup_policy:
1412 perf_event_free_task(p);
1414 mpol_put(p->mempolicy);
1415 bad_fork_cleanup_cgroup:
1417 if (clone_flags & CLONE_THREAD)
1418 threadgroup_fork_read_unlock(current);
1419 cgroup_exit(p, cgroup_callbacks_done);
1420 delayacct_tsk_free(p);
1421 module_put(task_thread_info(p)->exec_domain->module);
1422 bad_fork_cleanup_count:
1423 atomic_dec(&p->cred->user->processes);
1428 return ERR_PTR(retval);
1431 noinline struct pt_regs * __cpuinit __attribute__((weak)) idle_regs(struct pt_regs *regs)
1433 memset(regs, 0, sizeof(struct pt_regs));
1437 static inline void init_idle_pids(struct pid_link *links)
1441 for (type = PIDTYPE_PID; type < PIDTYPE_MAX; ++type) {
1442 INIT_HLIST_NODE(&links[type].node); /* not really needed */
1443 links[type].pid = &init_struct_pid;
1447 struct task_struct * __cpuinit fork_idle(int cpu)
1449 struct task_struct *task;
1450 struct pt_regs regs;
1452 task = copy_process(CLONE_VM, 0, idle_regs(®s), 0, NULL,
1453 &init_struct_pid, 0);
1454 if (!IS_ERR(task)) {
1455 init_idle_pids(task->pids);
1456 init_idle(task, cpu);
1463 * Ok, this is the main fork-routine.
1465 * It copies the process, and if successful kick-starts
1466 * it and waits for it to finish using the VM if required.
1468 long do_fork(unsigned long clone_flags,
1469 unsigned long stack_start,
1470 struct pt_regs *regs,
1471 unsigned long stack_size,
1472 int __user *parent_tidptr,
1473 int __user *child_tidptr)
1475 struct task_struct *p;
1480 * Do some preliminary argument and permissions checking before we
1481 * actually start allocating stuff
1483 if (clone_flags & CLONE_NEWUSER) {
1484 if (clone_flags & CLONE_THREAD)
1486 /* hopefully this check will go away when userns support is
1489 if (!capable(CAP_SYS_ADMIN) || !capable(CAP_SETUID) ||
1490 !capable(CAP_SETGID))
1495 * Determine whether and which event to report to ptracer. When
1496 * called from kernel_thread or CLONE_UNTRACED is explicitly
1497 * requested, no event is reported; otherwise, report if the event
1498 * for the type of forking is enabled.
1500 if (likely(user_mode(regs)) && !(clone_flags & CLONE_UNTRACED)) {
1501 if (clone_flags & CLONE_VFORK)
1502 trace = PTRACE_EVENT_VFORK;
1503 else if ((clone_flags & CSIGNAL) != SIGCHLD)
1504 trace = PTRACE_EVENT_CLONE;
1506 trace = PTRACE_EVENT_FORK;
1508 if (likely(!ptrace_event_enabled(current, trace)))
1512 p = copy_process(clone_flags, stack_start, regs, stack_size,
1513 child_tidptr, NULL, trace);
1515 * Do this prior waking up the new thread - the thread pointer
1516 * might get invalid after that point, if the thread exits quickly.
1519 struct completion vfork;
1521 trace_sched_process_fork(current, p);
1523 nr = task_pid_vnr(p);
1525 if (clone_flags & CLONE_PARENT_SETTID)
1526 put_user(nr, parent_tidptr);
1528 if (clone_flags & CLONE_VFORK) {
1529 p->vfork_done = &vfork;
1530 init_completion(&vfork);
1533 audit_finish_fork(p);
1536 * We set PF_STARTING at creation in case tracing wants to
1537 * use this to distinguish a fully live task from one that
1538 * hasn't finished SIGSTOP raising yet. Now we clear it
1539 * and set the child going.
1541 p->flags &= ~PF_STARTING;
1543 wake_up_new_task(p);
1545 /* forking complete and child started to run, tell ptracer */
1546 if (unlikely(trace))
1547 ptrace_event(trace, nr);
1549 if (clone_flags & CLONE_VFORK) {
1550 freezer_do_not_count();
1551 wait_for_completion(&vfork);
1553 ptrace_event(PTRACE_EVENT_VFORK_DONE, nr);
1561 #ifndef ARCH_MIN_MMSTRUCT_ALIGN
1562 #define ARCH_MIN_MMSTRUCT_ALIGN 0
1565 static void sighand_ctor(void *data)
1567 struct sighand_struct *sighand = data;
1569 spin_lock_init(&sighand->siglock);
1570 init_waitqueue_head(&sighand->signalfd_wqh);
1573 void __init proc_caches_init(void)
1575 sighand_cachep = kmem_cache_create("sighand_cache",
1576 sizeof(struct sighand_struct), 0,
1577 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_DESTROY_BY_RCU|
1578 SLAB_NOTRACK, sighand_ctor);
1579 signal_cachep = kmem_cache_create("signal_cache",
1580 sizeof(struct signal_struct), 0,
1581 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL);
1582 files_cachep = kmem_cache_create("files_cache",
1583 sizeof(struct files_struct), 0,
1584 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL);
1585 fs_cachep = kmem_cache_create("fs_cache",
1586 sizeof(struct fs_struct), 0,
1587 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL);
1589 * FIXME! The "sizeof(struct mm_struct)" currently includes the
1590 * whole struct cpumask for the OFFSTACK case. We could change
1591 * this to *only* allocate as much of it as required by the
1592 * maximum number of CPU's we can ever have. The cpumask_allocation
1593 * is at the end of the structure, exactly for that reason.
1595 mm_cachep = kmem_cache_create("mm_struct",
1596 sizeof(struct mm_struct), ARCH_MIN_MMSTRUCT_ALIGN,
1597 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL);
1598 vm_area_cachep = KMEM_CACHE(vm_area_struct, SLAB_PANIC);
1600 nsproxy_cache_init();
1604 * Check constraints on flags passed to the unshare system call.
1606 static int check_unshare_flags(unsigned long unshare_flags)
1608 if (unshare_flags & ~(CLONE_THREAD|CLONE_FS|CLONE_NEWNS|CLONE_SIGHAND|
1609 CLONE_VM|CLONE_FILES|CLONE_SYSVSEM|
1610 CLONE_NEWUTS|CLONE_NEWIPC|CLONE_NEWNET))
1613 * Not implemented, but pretend it works if there is nothing to
1614 * unshare. Note that unsharing CLONE_THREAD or CLONE_SIGHAND
1615 * needs to unshare vm.
1617 if (unshare_flags & (CLONE_THREAD | CLONE_SIGHAND | CLONE_VM)) {
1618 /* FIXME: get_task_mm() increments ->mm_users */
1619 if (atomic_read(¤t->mm->mm_users) > 1)
1627 * Unshare the filesystem structure if it is being shared
1629 static int unshare_fs(unsigned long unshare_flags, struct fs_struct **new_fsp)
1631 struct fs_struct *fs = current->fs;
1633 if (!(unshare_flags & CLONE_FS) || !fs)
1636 /* don't need lock here; in the worst case we'll do useless copy */
1640 *new_fsp = copy_fs_struct(fs);
1648 * Unshare file descriptor table if it is being shared
1650 static int unshare_fd(unsigned long unshare_flags, struct files_struct **new_fdp)
1652 struct files_struct *fd = current->files;
1655 if ((unshare_flags & CLONE_FILES) &&
1656 (fd && atomic_read(&fd->count) > 1)) {
1657 *new_fdp = dup_fd(fd, &error);
1666 * unshare allows a process to 'unshare' part of the process
1667 * context which was originally shared using clone. copy_*
1668 * functions used by do_fork() cannot be used here directly
1669 * because they modify an inactive task_struct that is being
1670 * constructed. Here we are modifying the current, active,
1673 SYSCALL_DEFINE1(unshare, unsigned long, unshare_flags)
1675 struct fs_struct *fs, *new_fs = NULL;
1676 struct files_struct *fd, *new_fd = NULL;
1677 struct nsproxy *new_nsproxy = NULL;
1681 err = check_unshare_flags(unshare_flags);
1683 goto bad_unshare_out;
1686 * If unsharing namespace, must also unshare filesystem information.
1688 if (unshare_flags & CLONE_NEWNS)
1689 unshare_flags |= CLONE_FS;
1691 * CLONE_NEWIPC must also detach from the undolist: after switching
1692 * to a new ipc namespace, the semaphore arrays from the old
1693 * namespace are unreachable.
1695 if (unshare_flags & (CLONE_NEWIPC|CLONE_SYSVSEM))
1697 err = unshare_fs(unshare_flags, &new_fs);
1699 goto bad_unshare_out;
1700 err = unshare_fd(unshare_flags, &new_fd);
1702 goto bad_unshare_cleanup_fs;
1703 err = unshare_nsproxy_namespaces(unshare_flags, &new_nsproxy, new_fs);
1705 goto bad_unshare_cleanup_fd;
1707 if (new_fs || new_fd || do_sysvsem || new_nsproxy) {
1710 * CLONE_SYSVSEM is equivalent to sys_exit().
1716 switch_task_namespaces(current, new_nsproxy);
1724 spin_lock(&fs->lock);
1725 current->fs = new_fs;
1730 spin_unlock(&fs->lock);
1734 fd = current->files;
1735 current->files = new_fd;
1739 task_unlock(current);
1743 put_nsproxy(new_nsproxy);
1745 bad_unshare_cleanup_fd:
1747 put_files_struct(new_fd);
1749 bad_unshare_cleanup_fs:
1751 free_fs_struct(new_fs);
1758 * Helper to unshare the files of the current task.
1759 * We don't want to expose copy_files internals to
1760 * the exec layer of the kernel.
1763 int unshare_files(struct files_struct **displaced)
1765 struct task_struct *task = current;
1766 struct files_struct *copy = NULL;
1769 error = unshare_fd(CLONE_FILES, ©);
1770 if (error || !copy) {
1774 *displaced = task->files;