4 * Copyright (C) 1991, 1992 Linus Torvalds
8 * 'fork.c' contains the help-routines for the 'fork' system call
9 * (see also entry.S and others).
10 * Fork is rather simple, once you get the hang of it, but the memory
11 * management can be a bitch. See 'mm/memory.c': 'copy_page_range()'
14 #include <linux/slab.h>
15 #include <linux/init.h>
16 #include <linux/unistd.h>
17 #include <linux/module.h>
18 #include <linux/vmalloc.h>
19 #include <linux/completion.h>
20 #include <linux/personality.h>
21 #include <linux/mempolicy.h>
22 #include <linux/sem.h>
23 #include <linux/file.h>
24 #include <linux/fdtable.h>
25 #include <linux/iocontext.h>
26 #include <linux/key.h>
27 #include <linux/binfmts.h>
28 #include <linux/mman.h>
29 #include <linux/mmu_notifier.h>
31 #include <linux/nsproxy.h>
32 #include <linux/capability.h>
33 #include <linux/cpu.h>
34 #include <linux/cgroup.h>
35 #include <linux/security.h>
36 #include <linux/hugetlb.h>
37 #include <linux/swap.h>
38 #include <linux/syscalls.h>
39 #include <linux/jiffies.h>
40 #include <linux/futex.h>
41 #include <linux/compat.h>
42 #include <linux/kthread.h>
43 #include <linux/task_io_accounting_ops.h>
44 #include <linux/rcupdate.h>
45 #include <linux/ptrace.h>
46 #include <linux/mount.h>
47 #include <linux/audit.h>
48 #include <linux/memcontrol.h>
49 #include <linux/ftrace.h>
50 #include <linux/profile.h>
51 #include <linux/rmap.h>
52 #include <linux/ksm.h>
53 #include <linux/acct.h>
54 #include <linux/tsacct_kern.h>
55 #include <linux/cn_proc.h>
56 #include <linux/freezer.h>
57 #include <linux/delayacct.h>
58 #include <linux/taskstats_kern.h>
59 #include <linux/random.h>
60 #include <linux/tty.h>
61 #include <linux/blkdev.h>
62 #include <linux/fs_struct.h>
63 #include <linux/magic.h>
64 #include <linux/perf_event.h>
65 #include <linux/posix-timers.h>
66 #include <linux/user-return-notifier.h>
67 #include <linux/oom.h>
68 #include <linux/khugepaged.h>
69 #include <linux/signalfd.h>
71 #include <asm/pgtable.h>
72 #include <asm/pgalloc.h>
73 #include <asm/uaccess.h>
74 #include <asm/mmu_context.h>
75 #include <asm/cacheflush.h>
76 #include <asm/tlbflush.h>
78 #include <trace/events/sched.h>
80 #define CREATE_TRACE_POINTS
81 #include <trace/events/task.h>
84 * Protected counters by write_lock_irq(&tasklist_lock)
86 unsigned long total_forks; /* Handle normal Linux uptimes. */
87 int nr_threads; /* The idle threads do not count.. */
89 int max_threads; /* tunable limit on nr_threads */
91 DEFINE_PER_CPU(unsigned long, process_counts) = 0;
93 __cacheline_aligned DEFINE_RWLOCK(tasklist_lock); /* outer */
95 #ifdef CONFIG_PROVE_RCU
96 int lockdep_tasklist_lock_is_held(void)
98 return lockdep_is_held(&tasklist_lock);
100 EXPORT_SYMBOL_GPL(lockdep_tasklist_lock_is_held);
101 #endif /* #ifdef CONFIG_PROVE_RCU */
103 int nr_processes(void)
108 for_each_possible_cpu(cpu)
109 total += per_cpu(process_counts, cpu);
114 #ifndef __HAVE_ARCH_TASK_STRUCT_ALLOCATOR
115 # define alloc_task_struct_node(node) \
116 kmem_cache_alloc_node(task_struct_cachep, GFP_KERNEL, node)
117 # define free_task_struct(tsk) \
118 kmem_cache_free(task_struct_cachep, (tsk))
119 static struct kmem_cache *task_struct_cachep;
122 #ifndef __HAVE_ARCH_THREAD_INFO_ALLOCATOR
123 static struct thread_info *alloc_thread_info_node(struct task_struct *tsk,
126 #ifdef CONFIG_DEBUG_STACK_USAGE
127 gfp_t mask = GFP_KERNEL | __GFP_ZERO;
129 gfp_t mask = GFP_KERNEL;
131 struct page *page = alloc_pages_node(node, mask, THREAD_SIZE_ORDER);
133 return page ? page_address(page) : NULL;
136 static inline void free_thread_info(struct thread_info *ti)
138 free_pages((unsigned long)ti, THREAD_SIZE_ORDER);
142 /* SLAB cache for signal_struct structures (tsk->signal) */
143 static struct kmem_cache *signal_cachep;
145 /* SLAB cache for sighand_struct structures (tsk->sighand) */
146 struct kmem_cache *sighand_cachep;
148 /* SLAB cache for files_struct structures (tsk->files) */
149 struct kmem_cache *files_cachep;
151 /* SLAB cache for fs_struct structures (tsk->fs) */
152 struct kmem_cache *fs_cachep;
154 /* SLAB cache for vm_area_struct structures */
155 struct kmem_cache *vm_area_cachep;
157 /* SLAB cache for mm_struct structures (tsk->mm) */
158 static struct kmem_cache *mm_cachep;
160 static void account_kernel_stack(struct thread_info *ti, int account)
162 struct zone *zone = page_zone(virt_to_page(ti));
164 mod_zone_page_state(zone, NR_KERNEL_STACK, account);
167 void free_task(struct task_struct *tsk)
169 account_kernel_stack(tsk->stack, -1);
170 free_thread_info(tsk->stack);
171 rt_mutex_debug_task_free(tsk);
172 ftrace_graph_exit_task(tsk);
173 free_task_struct(tsk);
175 EXPORT_SYMBOL(free_task);
177 static inline void free_signal_struct(struct signal_struct *sig)
179 taskstats_tgid_free(sig);
180 sched_autogroup_exit(sig);
181 kmem_cache_free(signal_cachep, sig);
184 static inline void put_signal_struct(struct signal_struct *sig)
186 if (atomic_dec_and_test(&sig->sigcnt))
187 free_signal_struct(sig);
190 void __put_task_struct(struct task_struct *tsk)
192 WARN_ON(!tsk->exit_state);
193 WARN_ON(atomic_read(&tsk->usage));
194 WARN_ON(tsk == current);
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 struct mm_struct *mm_access(struct task_struct *task, unsigned int mode)
653 struct mm_struct *mm;
656 err = mutex_lock_killable(&task->signal->cred_guard_mutex);
660 mm = get_task_mm(task);
661 if (mm && mm != current->mm &&
662 !ptrace_may_access(task, mode)) {
664 mm = ERR_PTR(-EACCES);
666 mutex_unlock(&task->signal->cred_guard_mutex);
671 void complete_vfork_done(struct task_struct *tsk)
673 struct completion *vfork_done = tsk->vfork_done;
675 tsk->vfork_done = NULL;
676 complete(vfork_done);
679 /* Please note the differences between mmput and mm_release.
680 * mmput is called whenever we stop holding onto a mm_struct,
681 * error success whatever.
683 * mm_release is called after a mm_struct has been removed
684 * from the current process.
686 * This difference is important for error handling, when we
687 * only half set up a mm_struct for a new process and need to restore
688 * the old one. Because we mmput the new mm_struct before
689 * restoring the old one. . .
690 * Eric Biederman 10 January 1998
692 void mm_release(struct task_struct *tsk, struct mm_struct *mm)
694 /* Get rid of any futexes when releasing the mm */
696 if (unlikely(tsk->robust_list)) {
697 exit_robust_list(tsk);
698 tsk->robust_list = NULL;
701 if (unlikely(tsk->compat_robust_list)) {
702 compat_exit_robust_list(tsk);
703 tsk->compat_robust_list = NULL;
706 if (unlikely(!list_empty(&tsk->pi_state_list)))
707 exit_pi_state_list(tsk);
710 /* Get rid of any cached register state */
711 deactivate_mm(tsk, mm);
714 complete_vfork_done(tsk);
717 * If we're exiting normally, clear a user-space tid field if
718 * requested. We leave this alone when dying by signal, to leave
719 * the value intact in a core dump, and to save the unnecessary
720 * trouble otherwise. Userland only wants this done for a sys_exit.
722 if (tsk->clear_child_tid) {
723 if (!(tsk->flags & PF_SIGNALED) &&
724 atomic_read(&mm->mm_users) > 1) {
726 * We don't check the error code - if userspace has
727 * not set up a proper pointer then tough luck.
729 put_user(0, tsk->clear_child_tid);
730 sys_futex(tsk->clear_child_tid, FUTEX_WAKE,
733 tsk->clear_child_tid = NULL;
738 * Allocate a new mm structure and copy contents from the
739 * mm structure of the passed in task structure.
741 struct mm_struct *dup_mm(struct task_struct *tsk)
743 struct mm_struct *mm, *oldmm = current->mm;
753 memcpy(mm, oldmm, sizeof(*mm));
756 /* Initializing for Swap token stuff */
757 mm->token_priority = 0;
758 mm->last_interval = 0;
760 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
761 mm->pmd_huge_pte = NULL;
764 if (!mm_init(mm, tsk))
767 if (init_new_context(tsk, mm))
770 dup_mm_exe_file(oldmm, mm);
772 err = dup_mmap(mm, oldmm);
776 mm->hiwater_rss = get_mm_rss(mm);
777 mm->hiwater_vm = mm->total_vm;
779 if (mm->binfmt && !try_module_get(mm->binfmt->module))
785 /* don't put binfmt in mmput, we haven't got module yet */
794 * If init_new_context() failed, we cannot use mmput() to free the mm
795 * because it calls destroy_context()
802 static int copy_mm(unsigned long clone_flags, struct task_struct *tsk)
804 struct mm_struct *mm, *oldmm;
807 tsk->min_flt = tsk->maj_flt = 0;
808 tsk->nvcsw = tsk->nivcsw = 0;
809 #ifdef CONFIG_DETECT_HUNG_TASK
810 tsk->last_switch_count = tsk->nvcsw + tsk->nivcsw;
814 tsk->active_mm = NULL;
817 * Are we cloning a kernel thread?
819 * We need to steal a active VM for that..
825 if (clone_flags & CLONE_VM) {
826 atomic_inc(&oldmm->mm_users);
837 /* Initializing for Swap token stuff */
838 mm->token_priority = 0;
839 mm->last_interval = 0;
849 static int copy_fs(unsigned long clone_flags, struct task_struct *tsk)
851 struct fs_struct *fs = current->fs;
852 if (clone_flags & CLONE_FS) {
853 /* tsk->fs is already what we want */
854 spin_lock(&fs->lock);
856 spin_unlock(&fs->lock);
860 spin_unlock(&fs->lock);
863 tsk->fs = copy_fs_struct(fs);
869 static int copy_files(unsigned long clone_flags, struct task_struct *tsk)
871 struct files_struct *oldf, *newf;
875 * A background process may not have any files ...
877 oldf = current->files;
881 if (clone_flags & CLONE_FILES) {
882 atomic_inc(&oldf->count);
886 newf = dup_fd(oldf, &error);
896 static int copy_io(unsigned long clone_flags, struct task_struct *tsk)
899 struct io_context *ioc = current->io_context;
900 struct io_context *new_ioc;
905 * Share io context with parent, if CLONE_IO is set
907 if (clone_flags & CLONE_IO) {
908 tsk->io_context = ioc_task_link(ioc);
909 if (unlikely(!tsk->io_context))
911 } else if (ioprio_valid(ioc->ioprio)) {
912 new_ioc = get_task_io_context(tsk, GFP_KERNEL, NUMA_NO_NODE);
913 if (unlikely(!new_ioc))
916 new_ioc->ioprio = ioc->ioprio;
917 put_io_context(new_ioc);
923 static int copy_sighand(unsigned long clone_flags, struct task_struct *tsk)
925 struct sighand_struct *sig;
927 if (clone_flags & CLONE_SIGHAND) {
928 atomic_inc(¤t->sighand->count);
931 sig = kmem_cache_alloc(sighand_cachep, GFP_KERNEL);
932 rcu_assign_pointer(tsk->sighand, sig);
935 atomic_set(&sig->count, 1);
936 memcpy(sig->action, current->sighand->action, sizeof(sig->action));
940 void __cleanup_sighand(struct sighand_struct *sighand)
942 if (atomic_dec_and_test(&sighand->count)) {
943 signalfd_cleanup(sighand);
944 kmem_cache_free(sighand_cachep, sighand);
950 * Initialize POSIX timer handling for a thread group.
952 static void posix_cpu_timers_init_group(struct signal_struct *sig)
954 unsigned long cpu_limit;
956 /* Thread group counters. */
957 thread_group_cputime_init(sig);
959 cpu_limit = ACCESS_ONCE(sig->rlim[RLIMIT_CPU].rlim_cur);
960 if (cpu_limit != RLIM_INFINITY) {
961 sig->cputime_expires.prof_exp = secs_to_cputime(cpu_limit);
962 sig->cputimer.running = 1;
965 /* The timer lists. */
966 INIT_LIST_HEAD(&sig->cpu_timers[0]);
967 INIT_LIST_HEAD(&sig->cpu_timers[1]);
968 INIT_LIST_HEAD(&sig->cpu_timers[2]);
971 static int copy_signal(unsigned long clone_flags, struct task_struct *tsk)
973 struct signal_struct *sig;
975 if (clone_flags & CLONE_THREAD)
978 sig = kmem_cache_zalloc(signal_cachep, GFP_KERNEL);
984 atomic_set(&sig->live, 1);
985 atomic_set(&sig->sigcnt, 1);
986 init_waitqueue_head(&sig->wait_chldexit);
987 if (clone_flags & CLONE_NEWPID)
988 sig->flags |= SIGNAL_UNKILLABLE;
989 sig->curr_target = tsk;
990 init_sigpending(&sig->shared_pending);
991 INIT_LIST_HEAD(&sig->posix_timers);
993 hrtimer_init(&sig->real_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
994 sig->real_timer.function = it_real_fn;
996 task_lock(current->group_leader);
997 memcpy(sig->rlim, current->signal->rlim, sizeof sig->rlim);
998 task_unlock(current->group_leader);
1000 posix_cpu_timers_init_group(sig);
1002 tty_audit_fork(sig);
1003 sched_autogroup_fork(sig);
1005 #ifdef CONFIG_CGROUPS
1006 init_rwsem(&sig->group_rwsem);
1009 sig->oom_adj = current->signal->oom_adj;
1010 sig->oom_score_adj = current->signal->oom_score_adj;
1011 sig->oom_score_adj_min = current->signal->oom_score_adj_min;
1013 mutex_init(&sig->cred_guard_mutex);
1018 static void copy_flags(unsigned long clone_flags, struct task_struct *p)
1020 unsigned long new_flags = p->flags;
1022 new_flags &= ~(PF_SUPERPRIV | PF_WQ_WORKER);
1023 new_flags |= PF_FORKNOEXEC;
1024 new_flags |= PF_STARTING;
1025 p->flags = new_flags;
1028 SYSCALL_DEFINE1(set_tid_address, int __user *, tidptr)
1030 current->clear_child_tid = tidptr;
1032 return task_pid_vnr(current);
1035 static void rt_mutex_init_task(struct task_struct *p)
1037 raw_spin_lock_init(&p->pi_lock);
1038 #ifdef CONFIG_RT_MUTEXES
1039 plist_head_init(&p->pi_waiters);
1040 p->pi_blocked_on = NULL;
1044 #ifdef CONFIG_MM_OWNER
1045 void mm_init_owner(struct mm_struct *mm, struct task_struct *p)
1049 #endif /* CONFIG_MM_OWNER */
1052 * Initialize POSIX timer handling for a single task.
1054 static void posix_cpu_timers_init(struct task_struct *tsk)
1056 tsk->cputime_expires.prof_exp = 0;
1057 tsk->cputime_expires.virt_exp = 0;
1058 tsk->cputime_expires.sched_exp = 0;
1059 INIT_LIST_HEAD(&tsk->cpu_timers[0]);
1060 INIT_LIST_HEAD(&tsk->cpu_timers[1]);
1061 INIT_LIST_HEAD(&tsk->cpu_timers[2]);
1065 * This creates a new process as a copy of the old one,
1066 * but does not actually start it yet.
1068 * It copies the registers, and all the appropriate
1069 * parts of the process environment (as per the clone
1070 * flags). The actual kick-off is left to the caller.
1072 static struct task_struct *copy_process(unsigned long clone_flags,
1073 unsigned long stack_start,
1074 struct pt_regs *regs,
1075 unsigned long stack_size,
1076 int __user *child_tidptr,
1081 struct task_struct *p;
1082 int cgroup_callbacks_done = 0;
1084 if ((clone_flags & (CLONE_NEWNS|CLONE_FS)) == (CLONE_NEWNS|CLONE_FS))
1085 return ERR_PTR(-EINVAL);
1088 * Thread groups must share signals as well, and detached threads
1089 * can only be started up within the thread group.
1091 if ((clone_flags & CLONE_THREAD) && !(clone_flags & CLONE_SIGHAND))
1092 return ERR_PTR(-EINVAL);
1095 * Shared signal handlers imply shared VM. By way of the above,
1096 * thread groups also imply shared VM. Blocking this case allows
1097 * for various simplifications in other code.
1099 if ((clone_flags & CLONE_SIGHAND) && !(clone_flags & CLONE_VM))
1100 return ERR_PTR(-EINVAL);
1103 * Siblings of global init remain as zombies on exit since they are
1104 * not reaped by their parent (swapper). To solve this and to avoid
1105 * multi-rooted process trees, prevent global and container-inits
1106 * from creating siblings.
1108 if ((clone_flags & CLONE_PARENT) &&
1109 current->signal->flags & SIGNAL_UNKILLABLE)
1110 return ERR_PTR(-EINVAL);
1112 retval = security_task_create(clone_flags);
1117 p = dup_task_struct(current);
1121 ftrace_graph_init_task(p);
1123 rt_mutex_init_task(p);
1125 #ifdef CONFIG_PROVE_LOCKING
1126 DEBUG_LOCKS_WARN_ON(!p->hardirqs_enabled);
1127 DEBUG_LOCKS_WARN_ON(!p->softirqs_enabled);
1130 if (atomic_read(&p->real_cred->user->processes) >=
1131 task_rlimit(p, RLIMIT_NPROC)) {
1132 if (!capable(CAP_SYS_ADMIN) && !capable(CAP_SYS_RESOURCE) &&
1133 p->real_cred->user != INIT_USER)
1136 current->flags &= ~PF_NPROC_EXCEEDED;
1138 retval = copy_creds(p, clone_flags);
1143 * If multiple threads are within copy_process(), then this check
1144 * triggers too late. This doesn't hurt, the check is only there
1145 * to stop root fork bombs.
1148 if (nr_threads >= max_threads)
1149 goto bad_fork_cleanup_count;
1151 if (!try_module_get(task_thread_info(p)->exec_domain->module))
1152 goto bad_fork_cleanup_count;
1155 delayacct_tsk_init(p); /* Must remain after dup_task_struct() */
1156 copy_flags(clone_flags, p);
1157 INIT_LIST_HEAD(&p->children);
1158 INIT_LIST_HEAD(&p->sibling);
1159 rcu_copy_process(p);
1160 p->vfork_done = NULL;
1161 spin_lock_init(&p->alloc_lock);
1163 init_sigpending(&p->pending);
1165 p->utime = p->stime = p->gtime = 0;
1166 p->utimescaled = p->stimescaled = 0;
1167 #ifndef CONFIG_VIRT_CPU_ACCOUNTING
1168 p->prev_utime = p->prev_stime = 0;
1170 #if defined(SPLIT_RSS_COUNTING)
1171 memset(&p->rss_stat, 0, sizeof(p->rss_stat));
1174 p->default_timer_slack_ns = current->timer_slack_ns;
1176 task_io_accounting_init(&p->ioac);
1177 acct_clear_integrals(p);
1179 posix_cpu_timers_init(p);
1181 do_posix_clock_monotonic_gettime(&p->start_time);
1182 p->real_start_time = p->start_time;
1183 monotonic_to_bootbased(&p->real_start_time);
1184 p->io_context = NULL;
1185 p->audit_context = NULL;
1186 if (clone_flags & CLONE_THREAD)
1187 threadgroup_change_begin(current);
1190 p->mempolicy = mpol_dup(p->mempolicy);
1191 if (IS_ERR(p->mempolicy)) {
1192 retval = PTR_ERR(p->mempolicy);
1193 p->mempolicy = NULL;
1194 goto bad_fork_cleanup_cgroup;
1196 mpol_fix_fork_child_flag(p);
1198 #ifdef CONFIG_CPUSETS
1199 p->cpuset_mem_spread_rotor = NUMA_NO_NODE;
1200 p->cpuset_slab_spread_rotor = NUMA_NO_NODE;
1202 #ifdef CONFIG_TRACE_IRQFLAGS
1204 #ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW
1205 p->hardirqs_enabled = 1;
1207 p->hardirqs_enabled = 0;
1209 p->hardirq_enable_ip = 0;
1210 p->hardirq_enable_event = 0;
1211 p->hardirq_disable_ip = _THIS_IP_;
1212 p->hardirq_disable_event = 0;
1213 p->softirqs_enabled = 1;
1214 p->softirq_enable_ip = _THIS_IP_;
1215 p->softirq_enable_event = 0;
1216 p->softirq_disable_ip = 0;
1217 p->softirq_disable_event = 0;
1218 p->hardirq_context = 0;
1219 p->softirq_context = 0;
1221 #ifdef CONFIG_LOCKDEP
1222 p->lockdep_depth = 0; /* no locks held yet */
1223 p->curr_chain_key = 0;
1224 p->lockdep_recursion = 0;
1227 #ifdef CONFIG_DEBUG_MUTEXES
1228 p->blocked_on = NULL; /* not blocked yet */
1230 #ifdef CONFIG_CGROUP_MEM_RES_CTLR
1231 p->memcg_batch.do_batch = 0;
1232 p->memcg_batch.memcg = NULL;
1235 /* Perform scheduler related setup. Assign this task to a CPU. */
1238 retval = perf_event_init_task(p);
1240 goto bad_fork_cleanup_policy;
1241 retval = audit_alloc(p);
1243 goto bad_fork_cleanup_policy;
1244 /* copy all the process information */
1245 retval = copy_semundo(clone_flags, p);
1247 goto bad_fork_cleanup_audit;
1248 retval = copy_files(clone_flags, p);
1250 goto bad_fork_cleanup_semundo;
1251 retval = copy_fs(clone_flags, p);
1253 goto bad_fork_cleanup_files;
1254 retval = copy_sighand(clone_flags, p);
1256 goto bad_fork_cleanup_fs;
1257 retval = copy_signal(clone_flags, p);
1259 goto bad_fork_cleanup_sighand;
1260 retval = copy_mm(clone_flags, p);
1262 goto bad_fork_cleanup_signal;
1263 retval = copy_namespaces(clone_flags, p);
1265 goto bad_fork_cleanup_mm;
1266 retval = copy_io(clone_flags, p);
1268 goto bad_fork_cleanup_namespaces;
1269 retval = copy_thread(clone_flags, stack_start, stack_size, p, regs);
1271 goto bad_fork_cleanup_io;
1273 if (pid != &init_struct_pid) {
1275 pid = alloc_pid(p->nsproxy->pid_ns);
1277 goto bad_fork_cleanup_io;
1280 p->pid = pid_nr(pid);
1282 if (clone_flags & CLONE_THREAD)
1283 p->tgid = current->tgid;
1285 p->set_child_tid = (clone_flags & CLONE_CHILD_SETTID) ? child_tidptr : NULL;
1287 * Clear TID on mm_release()?
1289 p->clear_child_tid = (clone_flags & CLONE_CHILD_CLEARTID) ? child_tidptr : NULL;
1294 p->robust_list = NULL;
1295 #ifdef CONFIG_COMPAT
1296 p->compat_robust_list = NULL;
1298 INIT_LIST_HEAD(&p->pi_state_list);
1299 p->pi_state_cache = NULL;
1302 * sigaltstack should be cleared when sharing the same VM
1304 if ((clone_flags & (CLONE_VM|CLONE_VFORK)) == CLONE_VM)
1305 p->sas_ss_sp = p->sas_ss_size = 0;
1308 * Syscall tracing and stepping should be turned off in the
1309 * child regardless of CLONE_PTRACE.
1311 user_disable_single_step(p);
1312 clear_tsk_thread_flag(p, TIF_SYSCALL_TRACE);
1313 #ifdef TIF_SYSCALL_EMU
1314 clear_tsk_thread_flag(p, TIF_SYSCALL_EMU);
1316 clear_all_latency_tracing(p);
1318 /* ok, now we should be set up.. */
1319 p->exit_signal = (clone_flags & CLONE_THREAD) ? -1 : (clone_flags & CSIGNAL);
1320 p->pdeath_signal = 0;
1324 p->nr_dirtied_pause = 128 >> (PAGE_SHIFT - 10);
1325 p->dirty_paused_when = 0;
1328 * Ok, make it visible to the rest of the system.
1329 * We dont wake it up yet.
1331 p->group_leader = p;
1332 INIT_LIST_HEAD(&p->thread_group);
1334 /* Now that the task is set up, run cgroup callbacks if
1335 * necessary. We need to run them before the task is visible
1336 * on the tasklist. */
1337 cgroup_fork_callbacks(p);
1338 cgroup_callbacks_done = 1;
1340 /* Need tasklist lock for parent etc handling! */
1341 write_lock_irq(&tasklist_lock);
1343 /* CLONE_PARENT re-uses the old parent */
1344 if (clone_flags & (CLONE_PARENT|CLONE_THREAD)) {
1345 p->real_parent = current->real_parent;
1346 p->parent_exec_id = current->parent_exec_id;
1348 p->real_parent = current;
1349 p->parent_exec_id = current->self_exec_id;
1352 spin_lock(¤t->sighand->siglock);
1355 * Process group and session signals need to be delivered to just the
1356 * parent before the fork or both the parent and the child after the
1357 * fork. Restart if a signal comes in before we add the new process to
1358 * it's process group.
1359 * A fatal signal pending means that current will exit, so the new
1360 * thread can't slip out of an OOM kill (or normal SIGKILL).
1362 recalc_sigpending();
1363 if (signal_pending(current)) {
1364 spin_unlock(¤t->sighand->siglock);
1365 write_unlock_irq(&tasklist_lock);
1366 retval = -ERESTARTNOINTR;
1367 goto bad_fork_free_pid;
1370 if (clone_flags & CLONE_THREAD) {
1371 current->signal->nr_threads++;
1372 atomic_inc(¤t->signal->live);
1373 atomic_inc(¤t->signal->sigcnt);
1374 p->group_leader = current->group_leader;
1375 list_add_tail_rcu(&p->thread_group, &p->group_leader->thread_group);
1378 if (likely(p->pid)) {
1379 ptrace_init_task(p, (clone_flags & CLONE_PTRACE) || trace);
1381 if (thread_group_leader(p)) {
1382 if (is_child_reaper(pid))
1383 p->nsproxy->pid_ns->child_reaper = p;
1385 p->signal->leader_pid = pid;
1386 p->signal->tty = tty_kref_get(current->signal->tty);
1387 attach_pid(p, PIDTYPE_PGID, task_pgrp(current));
1388 attach_pid(p, PIDTYPE_SID, task_session(current));
1389 list_add_tail(&p->sibling, &p->real_parent->children);
1390 list_add_tail_rcu(&p->tasks, &init_task.tasks);
1391 __this_cpu_inc(process_counts);
1393 attach_pid(p, PIDTYPE_PID, pid);
1398 spin_unlock(¤t->sighand->siglock);
1399 write_unlock_irq(&tasklist_lock);
1400 proc_fork_connector(p);
1401 cgroup_post_fork(p);
1402 if (clone_flags & CLONE_THREAD)
1403 threadgroup_change_end(current);
1406 trace_task_newtask(p, clone_flags);
1411 if (pid != &init_struct_pid)
1413 bad_fork_cleanup_io:
1416 bad_fork_cleanup_namespaces:
1417 exit_task_namespaces(p);
1418 bad_fork_cleanup_mm:
1421 bad_fork_cleanup_signal:
1422 if (!(clone_flags & CLONE_THREAD))
1423 free_signal_struct(p->signal);
1424 bad_fork_cleanup_sighand:
1425 __cleanup_sighand(p->sighand);
1426 bad_fork_cleanup_fs:
1427 exit_fs(p); /* blocking */
1428 bad_fork_cleanup_files:
1429 exit_files(p); /* blocking */
1430 bad_fork_cleanup_semundo:
1432 bad_fork_cleanup_audit:
1434 bad_fork_cleanup_policy:
1435 perf_event_free_task(p);
1437 mpol_put(p->mempolicy);
1438 bad_fork_cleanup_cgroup:
1440 if (clone_flags & CLONE_THREAD)
1441 threadgroup_change_end(current);
1442 cgroup_exit(p, cgroup_callbacks_done);
1443 delayacct_tsk_free(p);
1444 module_put(task_thread_info(p)->exec_domain->module);
1445 bad_fork_cleanup_count:
1446 atomic_dec(&p->cred->user->processes);
1451 return ERR_PTR(retval);
1454 noinline struct pt_regs * __cpuinit __attribute__((weak)) idle_regs(struct pt_regs *regs)
1456 memset(regs, 0, sizeof(struct pt_regs));
1460 static inline void init_idle_pids(struct pid_link *links)
1464 for (type = PIDTYPE_PID; type < PIDTYPE_MAX; ++type) {
1465 INIT_HLIST_NODE(&links[type].node); /* not really needed */
1466 links[type].pid = &init_struct_pid;
1470 struct task_struct * __cpuinit fork_idle(int cpu)
1472 struct task_struct *task;
1473 struct pt_regs regs;
1475 task = copy_process(CLONE_VM, 0, idle_regs(®s), 0, NULL,
1476 &init_struct_pid, 0);
1477 if (!IS_ERR(task)) {
1478 init_idle_pids(task->pids);
1479 init_idle(task, cpu);
1486 * Ok, this is the main fork-routine.
1488 * It copies the process, and if successful kick-starts
1489 * it and waits for it to finish using the VM if required.
1491 long do_fork(unsigned long clone_flags,
1492 unsigned long stack_start,
1493 struct pt_regs *regs,
1494 unsigned long stack_size,
1495 int __user *parent_tidptr,
1496 int __user *child_tidptr)
1498 struct task_struct *p;
1503 * Do some preliminary argument and permissions checking before we
1504 * actually start allocating stuff
1506 if (clone_flags & CLONE_NEWUSER) {
1507 if (clone_flags & CLONE_THREAD)
1509 /* hopefully this check will go away when userns support is
1512 if (!capable(CAP_SYS_ADMIN) || !capable(CAP_SETUID) ||
1513 !capable(CAP_SETGID))
1518 * Determine whether and which event to report to ptracer. When
1519 * called from kernel_thread or CLONE_UNTRACED is explicitly
1520 * requested, no event is reported; otherwise, report if the event
1521 * for the type of forking is enabled.
1523 if (likely(user_mode(regs)) && !(clone_flags & CLONE_UNTRACED)) {
1524 if (clone_flags & CLONE_VFORK)
1525 trace = PTRACE_EVENT_VFORK;
1526 else if ((clone_flags & CSIGNAL) != SIGCHLD)
1527 trace = PTRACE_EVENT_CLONE;
1529 trace = PTRACE_EVENT_FORK;
1531 if (likely(!ptrace_event_enabled(current, trace)))
1535 p = copy_process(clone_flags, stack_start, regs, stack_size,
1536 child_tidptr, NULL, trace);
1538 * Do this prior waking up the new thread - the thread pointer
1539 * might get invalid after that point, if the thread exits quickly.
1542 struct completion vfork;
1544 trace_sched_process_fork(current, p);
1546 nr = task_pid_vnr(p);
1548 if (clone_flags & CLONE_PARENT_SETTID)
1549 put_user(nr, parent_tidptr);
1551 if (clone_flags & CLONE_VFORK) {
1552 p->vfork_done = &vfork;
1553 init_completion(&vfork);
1557 * We set PF_STARTING at creation in case tracing wants to
1558 * use this to distinguish a fully live task from one that
1559 * hasn't finished SIGSTOP raising yet. Now we clear it
1560 * and set the child going.
1562 p->flags &= ~PF_STARTING;
1564 wake_up_new_task(p);
1566 /* forking complete and child started to run, tell ptracer */
1567 if (unlikely(trace))
1568 ptrace_event(trace, nr);
1570 if (clone_flags & CLONE_VFORK) {
1571 freezer_do_not_count();
1572 wait_for_completion(&vfork);
1574 ptrace_event(PTRACE_EVENT_VFORK_DONE, nr);
1582 #ifndef ARCH_MIN_MMSTRUCT_ALIGN
1583 #define ARCH_MIN_MMSTRUCT_ALIGN 0
1586 static void sighand_ctor(void *data)
1588 struct sighand_struct *sighand = data;
1590 spin_lock_init(&sighand->siglock);
1591 init_waitqueue_head(&sighand->signalfd_wqh);
1594 void __init proc_caches_init(void)
1596 sighand_cachep = kmem_cache_create("sighand_cache",
1597 sizeof(struct sighand_struct), 0,
1598 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_DESTROY_BY_RCU|
1599 SLAB_NOTRACK, sighand_ctor);
1600 signal_cachep = kmem_cache_create("signal_cache",
1601 sizeof(struct signal_struct), 0,
1602 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL);
1603 files_cachep = kmem_cache_create("files_cache",
1604 sizeof(struct files_struct), 0,
1605 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL);
1606 fs_cachep = kmem_cache_create("fs_cache",
1607 sizeof(struct fs_struct), 0,
1608 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL);
1610 * FIXME! The "sizeof(struct mm_struct)" currently includes the
1611 * whole struct cpumask for the OFFSTACK case. We could change
1612 * this to *only* allocate as much of it as required by the
1613 * maximum number of CPU's we can ever have. The cpumask_allocation
1614 * is at the end of the structure, exactly for that reason.
1616 mm_cachep = kmem_cache_create("mm_struct",
1617 sizeof(struct mm_struct), ARCH_MIN_MMSTRUCT_ALIGN,
1618 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL);
1619 vm_area_cachep = KMEM_CACHE(vm_area_struct, SLAB_PANIC);
1621 nsproxy_cache_init();
1625 * Check constraints on flags passed to the unshare system call.
1627 static int check_unshare_flags(unsigned long unshare_flags)
1629 if (unshare_flags & ~(CLONE_THREAD|CLONE_FS|CLONE_NEWNS|CLONE_SIGHAND|
1630 CLONE_VM|CLONE_FILES|CLONE_SYSVSEM|
1631 CLONE_NEWUTS|CLONE_NEWIPC|CLONE_NEWNET))
1634 * Not implemented, but pretend it works if there is nothing to
1635 * unshare. Note that unsharing CLONE_THREAD or CLONE_SIGHAND
1636 * needs to unshare vm.
1638 if (unshare_flags & (CLONE_THREAD | CLONE_SIGHAND | CLONE_VM)) {
1639 /* FIXME: get_task_mm() increments ->mm_users */
1640 if (atomic_read(¤t->mm->mm_users) > 1)
1648 * Unshare the filesystem structure if it is being shared
1650 static int unshare_fs(unsigned long unshare_flags, struct fs_struct **new_fsp)
1652 struct fs_struct *fs = current->fs;
1654 if (!(unshare_flags & CLONE_FS) || !fs)
1657 /* don't need lock here; in the worst case we'll do useless copy */
1661 *new_fsp = copy_fs_struct(fs);
1669 * Unshare file descriptor table if it is being shared
1671 static int unshare_fd(unsigned long unshare_flags, struct files_struct **new_fdp)
1673 struct files_struct *fd = current->files;
1676 if ((unshare_flags & CLONE_FILES) &&
1677 (fd && atomic_read(&fd->count) > 1)) {
1678 *new_fdp = dup_fd(fd, &error);
1687 * unshare allows a process to 'unshare' part of the process
1688 * context which was originally shared using clone. copy_*
1689 * functions used by do_fork() cannot be used here directly
1690 * because they modify an inactive task_struct that is being
1691 * constructed. Here we are modifying the current, active,
1694 SYSCALL_DEFINE1(unshare, unsigned long, unshare_flags)
1696 struct fs_struct *fs, *new_fs = NULL;
1697 struct files_struct *fd, *new_fd = NULL;
1698 struct nsproxy *new_nsproxy = NULL;
1702 err = check_unshare_flags(unshare_flags);
1704 goto bad_unshare_out;
1707 * If unsharing namespace, must also unshare filesystem information.
1709 if (unshare_flags & CLONE_NEWNS)
1710 unshare_flags |= CLONE_FS;
1712 * CLONE_NEWIPC must also detach from the undolist: after switching
1713 * to a new ipc namespace, the semaphore arrays from the old
1714 * namespace are unreachable.
1716 if (unshare_flags & (CLONE_NEWIPC|CLONE_SYSVSEM))
1718 err = unshare_fs(unshare_flags, &new_fs);
1720 goto bad_unshare_out;
1721 err = unshare_fd(unshare_flags, &new_fd);
1723 goto bad_unshare_cleanup_fs;
1724 err = unshare_nsproxy_namespaces(unshare_flags, &new_nsproxy, new_fs);
1726 goto bad_unshare_cleanup_fd;
1728 if (new_fs || new_fd || do_sysvsem || new_nsproxy) {
1731 * CLONE_SYSVSEM is equivalent to sys_exit().
1737 switch_task_namespaces(current, new_nsproxy);
1745 spin_lock(&fs->lock);
1746 current->fs = new_fs;
1751 spin_unlock(&fs->lock);
1755 fd = current->files;
1756 current->files = new_fd;
1760 task_unlock(current);
1764 put_nsproxy(new_nsproxy);
1766 bad_unshare_cleanup_fd:
1768 put_files_struct(new_fd);
1770 bad_unshare_cleanup_fs:
1772 free_fs_struct(new_fs);
1779 * Helper to unshare the files of the current task.
1780 * We don't want to expose copy_files internals to
1781 * the exec layer of the kernel.
1784 int unshare_files(struct files_struct **displaced)
1786 struct task_struct *task = current;
1787 struct files_struct *copy = NULL;
1790 error = unshare_fd(CLONE_FILES, ©);
1791 if (error || !copy) {
1795 *displaced = task->files;