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/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/proc_fs.h>
52 #include <linux/profile.h>
53 #include <linux/rmap.h>
54 #include <linux/ksm.h>
55 #include <linux/acct.h>
56 #include <linux/tsacct_kern.h>
57 #include <linux/cn_proc.h>
58 #include <linux/freezer.h>
59 #include <linux/delayacct.h>
60 #include <linux/taskstats_kern.h>
61 #include <linux/random.h>
62 #include <linux/tty.h>
63 #include <linux/blkdev.h>
64 #include <linux/fs_struct.h>
65 #include <linux/magic.h>
66 #include <linux/perf_event.h>
67 #include <linux/posix-timers.h>
68 #include <linux/user-return-notifier.h>
69 #include <linux/oom.h>
70 #include <linux/khugepaged.h>
71 #include <linux/signalfd.h>
73 #include <asm/pgtable.h>
74 #include <asm/pgalloc.h>
75 #include <asm/uaccess.h>
76 #include <asm/mmu_context.h>
77 #include <asm/cacheflush.h>
78 #include <asm/tlbflush.h>
80 #include <trace/events/sched.h>
83 * Protected counters by write_lock_irq(&tasklist_lock)
85 unsigned long total_forks; /* Handle normal Linux uptimes. */
86 int nr_threads; /* The idle threads do not count.. */
88 int max_threads; /* tunable limit on nr_threads */
90 DEFINE_PER_CPU(unsigned long, process_counts) = 0;
92 __cacheline_aligned DEFINE_RWLOCK(tasklist_lock); /* outer */
94 #ifdef CONFIG_PROVE_RCU
95 int lockdep_tasklist_lock_is_held(void)
97 return lockdep_is_held(&tasklist_lock);
99 EXPORT_SYMBOL_GPL(lockdep_tasklist_lock_is_held);
100 #endif /* #ifdef CONFIG_PROVE_RCU */
102 int nr_processes(void)
107 for_each_possible_cpu(cpu)
108 total += per_cpu(process_counts, cpu);
113 #ifndef __HAVE_ARCH_TASK_STRUCT_ALLOCATOR
114 # define alloc_task_struct_node(node) \
115 kmem_cache_alloc_node(task_struct_cachep, GFP_KERNEL, node)
116 # define free_task_struct(tsk) \
117 kmem_cache_free(task_struct_cachep, (tsk))
118 static struct kmem_cache *task_struct_cachep;
121 #ifndef __HAVE_ARCH_THREAD_INFO_ALLOCATOR
122 static struct thread_info *alloc_thread_info_node(struct task_struct *tsk,
125 #ifdef CONFIG_DEBUG_STACK_USAGE
126 gfp_t mask = GFP_KERNEL | __GFP_ZERO;
128 gfp_t mask = GFP_KERNEL;
130 struct page *page = alloc_pages_node(node, mask, THREAD_SIZE_ORDER);
132 return page ? page_address(page) : NULL;
135 static inline void free_thread_info(struct thread_info *ti)
137 free_pages((unsigned long)ti, THREAD_SIZE_ORDER);
141 /* SLAB cache for signal_struct structures (tsk->signal) */
142 static struct kmem_cache *signal_cachep;
144 /* SLAB cache for sighand_struct structures (tsk->sighand) */
145 struct kmem_cache *sighand_cachep;
147 /* SLAB cache for files_struct structures (tsk->files) */
148 struct kmem_cache *files_cachep;
150 /* SLAB cache for fs_struct structures (tsk->fs) */
151 struct kmem_cache *fs_cachep;
153 /* SLAB cache for vm_area_struct structures */
154 struct kmem_cache *vm_area_cachep;
156 /* SLAB cache for mm_struct structures (tsk->mm) */
157 static struct kmem_cache *mm_cachep;
159 static void account_kernel_stack(struct thread_info *ti, int account)
161 struct zone *zone = page_zone(virt_to_page(ti));
163 mod_zone_page_state(zone, NR_KERNEL_STACK, account);
166 void free_task(struct task_struct *tsk)
168 prop_local_destroy_single(&tsk->dirties);
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
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 err = prop_local_init_single(&tsk->dirties);
284 setup_thread_stack(tsk, orig);
285 clear_user_return_notifier(tsk);
286 clear_tsk_need_resched(tsk);
287 stackend = end_of_stack(tsk);
288 *stackend = STACK_END_MAGIC; /* for overflow detection */
290 #ifdef CONFIG_CC_STACKPROTECTOR
291 tsk->stack_canary = get_random_int();
294 /* One for us, one for whoever does the "release_task()" (usually parent) */
295 atomic_set(&tsk->usage,2);
296 atomic_set(&tsk->fs_excl, 0);
297 #ifdef CONFIG_BLK_DEV_IO_TRACE
300 tsk->splice_pipe = NULL;
302 account_kernel_stack(ti, 1);
307 free_thread_info(ti);
308 free_task_struct(tsk);
313 static int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm)
315 struct vm_area_struct *mpnt, *tmp, *prev, **pprev;
316 struct rb_node **rb_link, *rb_parent;
318 unsigned long charge;
319 struct mempolicy *pol;
321 down_write(&oldmm->mmap_sem);
322 flush_cache_dup_mm(oldmm);
324 * Not linked in yet - no deadlock potential:
326 down_write_nested(&mm->mmap_sem, SINGLE_DEPTH_NESTING);
330 mm->mmap_cache = NULL;
331 mm->free_area_cache = oldmm->mmap_base;
332 mm->cached_hole_size = ~0UL;
334 cpumask_clear(mm_cpumask(mm));
336 rb_link = &mm->mm_rb.rb_node;
339 retval = ksm_fork(mm, oldmm);
342 retval = khugepaged_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);
375 if (anon_vma_fork(tmp, mpnt))
376 goto fail_nomem_anon_vma_fork;
377 tmp->vm_flags &= ~VM_LOCKED;
378 tmp->vm_next = tmp->vm_prev = NULL;
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 mutex_lock(&mapping->i_mmap_mutex);
388 if (tmp->vm_flags & VM_SHARED)
389 mapping->i_mmap_writable++;
390 flush_dcache_mmap_lock(mapping);
391 /* insert tmp into the share list, just after mpnt */
392 vma_prio_tree_add(tmp, mpnt);
393 flush_dcache_mmap_unlock(mapping);
394 mutex_unlock(&mapping->i_mmap_mutex);
398 * Clear hugetlb-related page reserves for children. This only
399 * affects MAP_PRIVATE mappings. Faults generated by the child
400 * are not guaranteed to succeed, even if read-only
402 if (is_vm_hugetlb_page(tmp))
403 reset_vma_resv_huge_pages(tmp);
406 * Link in the new vma and copy the page table entries.
409 pprev = &tmp->vm_next;
413 __vma_link_rb(mm, tmp, rb_link, rb_parent);
414 rb_link = &tmp->vm_rb.rb_right;
415 rb_parent = &tmp->vm_rb;
418 retval = copy_page_range(mm, oldmm, mpnt);
420 if (tmp->vm_ops && tmp->vm_ops->open)
421 tmp->vm_ops->open(tmp);
426 /* a new mm has just been created */
427 arch_dup_mmap(oldmm, mm);
430 up_write(&mm->mmap_sem);
432 up_write(&oldmm->mmap_sem);
434 fail_nomem_anon_vma_fork:
437 kmem_cache_free(vm_area_cachep, tmp);
440 vm_unacct_memory(charge);
444 static inline int mm_alloc_pgd(struct mm_struct * mm)
446 mm->pgd = pgd_alloc(mm);
447 if (unlikely(!mm->pgd))
452 static inline void mm_free_pgd(struct mm_struct * mm)
454 pgd_free(mm, mm->pgd);
457 #define dup_mmap(mm, oldmm) (0)
458 #define mm_alloc_pgd(mm) (0)
459 #define mm_free_pgd(mm)
460 #endif /* CONFIG_MMU */
462 __cacheline_aligned_in_smp DEFINE_SPINLOCK(mmlist_lock);
464 #define allocate_mm() (kmem_cache_alloc(mm_cachep, GFP_KERNEL))
465 #define free_mm(mm) (kmem_cache_free(mm_cachep, (mm)))
467 static unsigned long default_dump_filter = MMF_DUMP_FILTER_DEFAULT;
469 static int __init coredump_filter_setup(char *s)
471 default_dump_filter =
472 (simple_strtoul(s, NULL, 0) << MMF_DUMP_FILTER_SHIFT) &
473 MMF_DUMP_FILTER_MASK;
477 __setup("coredump_filter=", coredump_filter_setup);
479 #include <linux/init_task.h>
481 static void mm_init_aio(struct mm_struct *mm)
484 spin_lock_init(&mm->ioctx_lock);
485 INIT_HLIST_HEAD(&mm->ioctx_list);
489 static struct mm_struct * mm_init(struct mm_struct * mm, struct task_struct *p)
491 atomic_set(&mm->mm_users, 1);
492 atomic_set(&mm->mm_count, 1);
493 init_rwsem(&mm->mmap_sem);
494 INIT_LIST_HEAD(&mm->mmlist);
495 mm->flags = (current->mm) ?
496 (current->mm->flags & MMF_INIT_MASK) : default_dump_filter;
497 mm->core_state = NULL;
499 memset(&mm->rss_stat, 0, sizeof(mm->rss_stat));
500 spin_lock_init(&mm->page_table_lock);
501 mm->free_area_cache = TASK_UNMAPPED_BASE;
502 mm->cached_hole_size = ~0UL;
504 mm_init_owner(mm, p);
505 atomic_set(&mm->oom_disable_count, 0);
507 if (likely(!mm_alloc_pgd(mm))) {
509 mmu_notifier_mm_init(mm);
518 * Allocate and initialize an mm_struct.
520 struct mm_struct * mm_alloc(void)
522 struct mm_struct * mm;
528 memset(mm, 0, sizeof(*mm));
530 return mm_init(mm, current);
534 * Called when the last reference to the mm
535 * is dropped: either by a lazy thread or by
536 * mmput. Free the page directory and the mm.
538 void __mmdrop(struct mm_struct *mm)
540 BUG_ON(mm == &init_mm);
543 mmu_notifier_mm_destroy(mm);
544 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
545 VM_BUG_ON(mm->pmd_huge_pte);
549 EXPORT_SYMBOL_GPL(__mmdrop);
552 * Decrement the use count and release all resources for an mm.
554 void mmput(struct mm_struct *mm)
558 if (atomic_dec_and_test(&mm->mm_users)) {
561 khugepaged_exit(mm); /* must run before exit_mmap */
563 set_mm_exe_file(mm, NULL);
564 if (!list_empty(&mm->mmlist)) {
565 spin_lock(&mmlist_lock);
566 list_del(&mm->mmlist);
567 spin_unlock(&mmlist_lock);
571 module_put(mm->binfmt->module);
575 EXPORT_SYMBOL_GPL(mmput);
578 * We added or removed a vma mapping the executable. The vmas are only mapped
579 * during exec and are not mapped with the mmap system call.
580 * Callers must hold down_write() on the mm's mmap_sem for these
582 void added_exe_file_vma(struct mm_struct *mm)
584 mm->num_exe_file_vmas++;
587 void removed_exe_file_vma(struct mm_struct *mm)
589 mm->num_exe_file_vmas--;
590 if ((mm->num_exe_file_vmas == 0) && mm->exe_file){
597 void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file)
600 get_file(new_exe_file);
603 mm->exe_file = new_exe_file;
604 mm->num_exe_file_vmas = 0;
607 struct file *get_mm_exe_file(struct mm_struct *mm)
609 struct file *exe_file;
611 /* We need mmap_sem to protect against races with removal of
612 * VM_EXECUTABLE vmas */
613 down_read(&mm->mmap_sem);
614 exe_file = mm->exe_file;
617 up_read(&mm->mmap_sem);
621 static void dup_mm_exe_file(struct mm_struct *oldmm, struct mm_struct *newmm)
623 /* It's safe to write the exe_file pointer without exe_file_lock because
624 * this is called during fork when the task is not yet in /proc */
625 newmm->exe_file = get_mm_exe_file(oldmm);
629 * get_task_mm - acquire a reference to the task's mm
631 * Returns %NULL if the task has no mm. Checks PF_KTHREAD (meaning
632 * this kernel workthread has transiently adopted a user mm with use_mm,
633 * to do its AIO) is not set and if so returns a reference to it, after
634 * bumping up the use count. User must release the mm via mmput()
635 * after use. Typically used by /proc and ptrace.
637 struct mm_struct *get_task_mm(struct task_struct *task)
639 struct mm_struct *mm;
644 if (task->flags & PF_KTHREAD)
647 atomic_inc(&mm->mm_users);
652 EXPORT_SYMBOL_GPL(get_task_mm);
654 /* Please note the differences between mmput and mm_release.
655 * mmput is called whenever we stop holding onto a mm_struct,
656 * error success whatever.
658 * mm_release is called after a mm_struct has been removed
659 * from the current process.
661 * This difference is important for error handling, when we
662 * only half set up a mm_struct for a new process and need to restore
663 * the old one. Because we mmput the new mm_struct before
664 * restoring the old one. . .
665 * Eric Biederman 10 January 1998
667 void mm_release(struct task_struct *tsk, struct mm_struct *mm)
669 struct completion *vfork_done = tsk->vfork_done;
671 /* Get rid of any futexes when releasing the mm */
673 if (unlikely(tsk->robust_list)) {
674 exit_robust_list(tsk);
675 tsk->robust_list = NULL;
678 if (unlikely(tsk->compat_robust_list)) {
679 compat_exit_robust_list(tsk);
680 tsk->compat_robust_list = NULL;
683 if (unlikely(!list_empty(&tsk->pi_state_list)))
684 exit_pi_state_list(tsk);
687 /* Get rid of any cached register state */
688 deactivate_mm(tsk, mm);
690 /* notify parent sleeping on vfork() */
692 tsk->vfork_done = NULL;
693 complete(vfork_done);
697 * If we're exiting normally, clear a user-space tid field if
698 * requested. We leave this alone when dying by signal, to leave
699 * the value intact in a core dump, and to save the unnecessary
700 * trouble otherwise. Userland only wants this done for a sys_exit.
702 if (tsk->clear_child_tid) {
703 if (!(tsk->flags & PF_SIGNALED) &&
704 atomic_read(&mm->mm_users) > 1) {
706 * We don't check the error code - if userspace has
707 * not set up a proper pointer then tough luck.
709 put_user(0, tsk->clear_child_tid);
710 sys_futex(tsk->clear_child_tid, FUTEX_WAKE,
713 tsk->clear_child_tid = NULL;
718 * Allocate a new mm structure and copy contents from the
719 * mm structure of the passed in task structure.
721 struct mm_struct *dup_mm(struct task_struct *tsk)
723 struct mm_struct *mm, *oldmm = current->mm;
733 memcpy(mm, oldmm, sizeof(*mm));
736 /* Initializing for Swap token stuff */
737 mm->token_priority = 0;
738 mm->last_interval = 0;
740 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
741 mm->pmd_huge_pte = NULL;
744 if (!mm_init(mm, tsk))
747 if (init_new_context(tsk, mm))
750 dup_mm_exe_file(oldmm, mm);
752 err = dup_mmap(mm, oldmm);
756 mm->hiwater_rss = get_mm_rss(mm);
757 mm->hiwater_vm = mm->total_vm;
759 if (mm->binfmt && !try_module_get(mm->binfmt->module))
765 /* don't put binfmt in mmput, we haven't got module yet */
774 * If init_new_context() failed, we cannot use mmput() to free the mm
775 * because it calls destroy_context()
782 static int copy_mm(unsigned long clone_flags, struct task_struct * tsk)
784 struct mm_struct * mm, *oldmm;
787 tsk->min_flt = tsk->maj_flt = 0;
788 tsk->nvcsw = tsk->nivcsw = 0;
789 #ifdef CONFIG_DETECT_HUNG_TASK
790 tsk->last_switch_count = tsk->nvcsw + tsk->nivcsw;
794 tsk->active_mm = NULL;
797 * Are we cloning a kernel thread?
799 * We need to steal a active VM for that..
805 if (clone_flags & CLONE_VM) {
806 atomic_inc(&oldmm->mm_users);
817 /* Initializing for Swap token stuff */
818 mm->token_priority = 0;
819 mm->last_interval = 0;
820 if (tsk->signal->oom_score_adj == OOM_SCORE_ADJ_MIN)
821 atomic_inc(&mm->oom_disable_count);
831 static int copy_fs(unsigned long clone_flags, struct task_struct *tsk)
833 struct fs_struct *fs = current->fs;
834 if (clone_flags & CLONE_FS) {
835 /* tsk->fs is already what we want */
836 spin_lock(&fs->lock);
838 spin_unlock(&fs->lock);
842 spin_unlock(&fs->lock);
845 tsk->fs = copy_fs_struct(fs);
851 static int copy_files(unsigned long clone_flags, struct task_struct * tsk)
853 struct files_struct *oldf, *newf;
857 * A background process may not have any files ...
859 oldf = current->files;
863 if (clone_flags & CLONE_FILES) {
864 atomic_inc(&oldf->count);
868 newf = dup_fd(oldf, &error);
878 static int copy_io(unsigned long clone_flags, struct task_struct *tsk)
881 struct io_context *ioc = current->io_context;
886 * Share io context with parent, if CLONE_IO is set
888 if (clone_flags & CLONE_IO) {
889 tsk->io_context = ioc_task_link(ioc);
890 if (unlikely(!tsk->io_context))
892 } else if (ioprio_valid(ioc->ioprio)) {
893 tsk->io_context = alloc_io_context(GFP_KERNEL, -1);
894 if (unlikely(!tsk->io_context))
897 tsk->io_context->ioprio = ioc->ioprio;
903 static int copy_sighand(unsigned long clone_flags, struct task_struct *tsk)
905 struct sighand_struct *sig;
907 if (clone_flags & CLONE_SIGHAND) {
908 atomic_inc(¤t->sighand->count);
911 sig = kmem_cache_alloc(sighand_cachep, GFP_KERNEL);
912 rcu_assign_pointer(tsk->sighand, sig);
915 atomic_set(&sig->count, 1);
916 memcpy(sig->action, current->sighand->action, sizeof(sig->action));
920 void __cleanup_sighand(struct sighand_struct *sighand)
922 if (atomic_dec_and_test(&sighand->count)) {
923 signalfd_cleanup(sighand);
924 kmem_cache_free(sighand_cachep, sighand);
930 * Initialize POSIX timer handling for a thread group.
932 static void posix_cpu_timers_init_group(struct signal_struct *sig)
934 unsigned long cpu_limit;
936 /* Thread group counters. */
937 thread_group_cputime_init(sig);
939 cpu_limit = ACCESS_ONCE(sig->rlim[RLIMIT_CPU].rlim_cur);
940 if (cpu_limit != RLIM_INFINITY) {
941 sig->cputime_expires.prof_exp = secs_to_cputime(cpu_limit);
942 sig->cputimer.running = 1;
945 /* The timer lists. */
946 INIT_LIST_HEAD(&sig->cpu_timers[0]);
947 INIT_LIST_HEAD(&sig->cpu_timers[1]);
948 INIT_LIST_HEAD(&sig->cpu_timers[2]);
951 static int copy_signal(unsigned long clone_flags, struct task_struct *tsk)
953 struct signal_struct *sig;
955 if (clone_flags & CLONE_THREAD)
958 sig = kmem_cache_zalloc(signal_cachep, GFP_KERNEL);
964 atomic_set(&sig->live, 1);
965 atomic_set(&sig->sigcnt, 1);
966 init_waitqueue_head(&sig->wait_chldexit);
967 if (clone_flags & CLONE_NEWPID)
968 sig->flags |= SIGNAL_UNKILLABLE;
969 sig->curr_target = tsk;
970 init_sigpending(&sig->shared_pending);
971 INIT_LIST_HEAD(&sig->posix_timers);
973 hrtimer_init(&sig->real_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
974 sig->real_timer.function = it_real_fn;
976 task_lock(current->group_leader);
977 memcpy(sig->rlim, current->signal->rlim, sizeof sig->rlim);
978 task_unlock(current->group_leader);
980 posix_cpu_timers_init_group(sig);
983 sched_autogroup_fork(sig);
985 #ifdef CONFIG_CGROUPS
986 init_rwsem(&sig->threadgroup_fork_lock);
989 sig->oom_adj = current->signal->oom_adj;
990 sig->oom_score_adj = current->signal->oom_score_adj;
991 sig->oom_score_adj_min = current->signal->oom_score_adj_min;
993 mutex_init(&sig->cred_guard_mutex);
998 static void copy_flags(unsigned long clone_flags, struct task_struct *p)
1000 unsigned long new_flags = p->flags;
1002 new_flags &= ~(PF_SUPERPRIV | PF_WQ_WORKER);
1003 new_flags |= PF_FORKNOEXEC;
1004 new_flags |= PF_STARTING;
1005 p->flags = new_flags;
1006 clear_freeze_flag(p);
1009 SYSCALL_DEFINE1(set_tid_address, int __user *, tidptr)
1011 current->clear_child_tid = tidptr;
1013 return task_pid_vnr(current);
1016 static void rt_mutex_init_task(struct task_struct *p)
1018 raw_spin_lock_init(&p->pi_lock);
1019 #ifdef CONFIG_RT_MUTEXES
1020 plist_head_init_raw(&p->pi_waiters, &p->pi_lock);
1021 p->pi_blocked_on = NULL;
1025 #ifdef CONFIG_MM_OWNER
1026 void mm_init_owner(struct mm_struct *mm, struct task_struct *p)
1030 #endif /* CONFIG_MM_OWNER */
1033 * Initialize POSIX timer handling for a single task.
1035 static void posix_cpu_timers_init(struct task_struct *tsk)
1037 tsk->cputime_expires.prof_exp = cputime_zero;
1038 tsk->cputime_expires.virt_exp = cputime_zero;
1039 tsk->cputime_expires.sched_exp = 0;
1040 INIT_LIST_HEAD(&tsk->cpu_timers[0]);
1041 INIT_LIST_HEAD(&tsk->cpu_timers[1]);
1042 INIT_LIST_HEAD(&tsk->cpu_timers[2]);
1046 * This creates a new process as a copy of the old one,
1047 * but does not actually start it yet.
1049 * It copies the registers, and all the appropriate
1050 * parts of the process environment (as per the clone
1051 * flags). The actual kick-off is left to the caller.
1053 static struct task_struct *copy_process(unsigned long clone_flags,
1054 unsigned long stack_start,
1055 struct pt_regs *regs,
1056 unsigned long stack_size,
1057 int __user *child_tidptr,
1062 struct task_struct *p;
1063 int cgroup_callbacks_done = 0;
1065 if ((clone_flags & (CLONE_NEWNS|CLONE_FS)) == (CLONE_NEWNS|CLONE_FS))
1066 return ERR_PTR(-EINVAL);
1069 * Thread groups must share signals as well, and detached threads
1070 * can only be started up within the thread group.
1072 if ((clone_flags & CLONE_THREAD) && !(clone_flags & CLONE_SIGHAND))
1073 return ERR_PTR(-EINVAL);
1076 * Shared signal handlers imply shared VM. By way of the above,
1077 * thread groups also imply shared VM. Blocking this case allows
1078 * for various simplifications in other code.
1080 if ((clone_flags & CLONE_SIGHAND) && !(clone_flags & CLONE_VM))
1081 return ERR_PTR(-EINVAL);
1084 * Siblings of global init remain as zombies on exit since they are
1085 * not reaped by their parent (swapper). To solve this and to avoid
1086 * multi-rooted process trees, prevent global and container-inits
1087 * from creating siblings.
1089 if ((clone_flags & CLONE_PARENT) &&
1090 current->signal->flags & SIGNAL_UNKILLABLE)
1091 return ERR_PTR(-EINVAL);
1093 retval = security_task_create(clone_flags);
1098 p = dup_task_struct(current);
1102 ftrace_graph_init_task(p);
1104 rt_mutex_init_task(p);
1106 #ifdef CONFIG_PROVE_LOCKING
1107 DEBUG_LOCKS_WARN_ON(!p->hardirqs_enabled);
1108 DEBUG_LOCKS_WARN_ON(!p->softirqs_enabled);
1111 if (atomic_read(&p->real_cred->user->processes) >=
1112 task_rlimit(p, RLIMIT_NPROC)) {
1113 if (!capable(CAP_SYS_ADMIN) && !capable(CAP_SYS_RESOURCE) &&
1114 p->real_cred->user != INIT_USER)
1118 retval = copy_creds(p, clone_flags);
1123 * If multiple threads are within copy_process(), then this check
1124 * triggers too late. This doesn't hurt, the check is only there
1125 * to stop root fork bombs.
1128 if (nr_threads >= max_threads)
1129 goto bad_fork_cleanup_count;
1131 if (!try_module_get(task_thread_info(p)->exec_domain->module))
1132 goto bad_fork_cleanup_count;
1135 delayacct_tsk_init(p); /* Must remain after dup_task_struct() */
1136 copy_flags(clone_flags, p);
1137 INIT_LIST_HEAD(&p->children);
1138 INIT_LIST_HEAD(&p->sibling);
1139 rcu_copy_process(p);
1140 p->vfork_done = NULL;
1141 spin_lock_init(&p->alloc_lock);
1143 init_sigpending(&p->pending);
1145 p->utime = cputime_zero;
1146 p->stime = cputime_zero;
1147 p->gtime = cputime_zero;
1148 p->utimescaled = cputime_zero;
1149 p->stimescaled = cputime_zero;
1150 #ifndef CONFIG_VIRT_CPU_ACCOUNTING
1151 p->prev_utime = cputime_zero;
1152 p->prev_stime = cputime_zero;
1154 #if defined(SPLIT_RSS_COUNTING)
1155 memset(&p->rss_stat, 0, sizeof(p->rss_stat));
1158 p->default_timer_slack_ns = current->timer_slack_ns;
1160 task_io_accounting_init(&p->ioac);
1161 acct_clear_integrals(p);
1163 posix_cpu_timers_init(p);
1165 do_posix_clock_monotonic_gettime(&p->start_time);
1166 p->real_start_time = p->start_time;
1167 monotonic_to_bootbased(&p->real_start_time);
1168 p->io_context = NULL;
1169 p->audit_context = NULL;
1170 if (clone_flags & CLONE_THREAD)
1171 threadgroup_fork_read_lock(current);
1174 p->mempolicy = mpol_dup(p->mempolicy);
1175 if (IS_ERR(p->mempolicy)) {
1176 retval = PTR_ERR(p->mempolicy);
1177 p->mempolicy = NULL;
1178 goto bad_fork_cleanup_cgroup;
1180 mpol_fix_fork_child_flag(p);
1182 #ifdef CONFIG_TRACE_IRQFLAGS
1184 #ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW
1185 p->hardirqs_enabled = 1;
1187 p->hardirqs_enabled = 0;
1189 p->hardirq_enable_ip = 0;
1190 p->hardirq_enable_event = 0;
1191 p->hardirq_disable_ip = _THIS_IP_;
1192 p->hardirq_disable_event = 0;
1193 p->softirqs_enabled = 1;
1194 p->softirq_enable_ip = _THIS_IP_;
1195 p->softirq_enable_event = 0;
1196 p->softirq_disable_ip = 0;
1197 p->softirq_disable_event = 0;
1198 p->hardirq_context = 0;
1199 p->softirq_context = 0;
1201 #ifdef CONFIG_LOCKDEP
1202 p->lockdep_depth = 0; /* no locks held yet */
1203 p->curr_chain_key = 0;
1204 p->lockdep_recursion = 0;
1207 #ifdef CONFIG_DEBUG_MUTEXES
1208 p->blocked_on = NULL; /* not blocked yet */
1210 #ifdef CONFIG_CGROUP_MEM_RES_CTLR
1211 p->memcg_batch.do_batch = 0;
1212 p->memcg_batch.memcg = NULL;
1215 /* Perform scheduler related setup. Assign this task to a CPU. */
1218 retval = perf_event_init_task(p);
1220 goto bad_fork_cleanup_policy;
1222 if ((retval = audit_alloc(p)))
1223 goto bad_fork_cleanup_policy;
1224 /* copy all the process information */
1225 if ((retval = copy_semundo(clone_flags, p)))
1226 goto bad_fork_cleanup_audit;
1227 if ((retval = copy_files(clone_flags, p)))
1228 goto bad_fork_cleanup_semundo;
1229 if ((retval = copy_fs(clone_flags, p)))
1230 goto bad_fork_cleanup_files;
1231 if ((retval = copy_sighand(clone_flags, p)))
1232 goto bad_fork_cleanup_fs;
1233 if ((retval = copy_signal(clone_flags, p)))
1234 goto bad_fork_cleanup_sighand;
1235 if ((retval = copy_mm(clone_flags, p)))
1236 goto bad_fork_cleanup_signal;
1237 if ((retval = copy_namespaces(clone_flags, p)))
1238 goto bad_fork_cleanup_mm;
1239 if ((retval = copy_io(clone_flags, p)))
1240 goto bad_fork_cleanup_namespaces;
1241 retval = copy_thread(clone_flags, stack_start, stack_size, p, regs);
1243 goto bad_fork_cleanup_io;
1245 if (pid != &init_struct_pid) {
1247 pid = alloc_pid(p->nsproxy->pid_ns);
1249 goto bad_fork_cleanup_io;
1252 p->pid = pid_nr(pid);
1254 if (clone_flags & CLONE_THREAD)
1255 p->tgid = current->tgid;
1257 p->set_child_tid = (clone_flags & CLONE_CHILD_SETTID) ? child_tidptr : NULL;
1259 * Clear TID on mm_release()?
1261 p->clear_child_tid = (clone_flags & CLONE_CHILD_CLEARTID) ? child_tidptr: NULL;
1266 p->robust_list = NULL;
1267 #ifdef CONFIG_COMPAT
1268 p->compat_robust_list = NULL;
1270 INIT_LIST_HEAD(&p->pi_state_list);
1271 p->pi_state_cache = NULL;
1274 * sigaltstack should be cleared when sharing the same VM
1276 if ((clone_flags & (CLONE_VM|CLONE_VFORK)) == CLONE_VM)
1277 p->sas_ss_sp = p->sas_ss_size = 0;
1280 * Syscall tracing and stepping should be turned off in the
1281 * child regardless of CLONE_PTRACE.
1283 user_disable_single_step(p);
1284 clear_tsk_thread_flag(p, TIF_SYSCALL_TRACE);
1285 #ifdef TIF_SYSCALL_EMU
1286 clear_tsk_thread_flag(p, TIF_SYSCALL_EMU);
1288 clear_all_latency_tracing(p);
1290 /* ok, now we should be set up.. */
1291 p->exit_signal = (clone_flags & CLONE_THREAD) ? -1 : (clone_flags & CSIGNAL);
1292 p->pdeath_signal = 0;
1296 * Ok, make it visible to the rest of the system.
1297 * We dont wake it up yet.
1299 p->group_leader = p;
1300 INIT_LIST_HEAD(&p->thread_group);
1302 /* Now that the task is set up, run cgroup callbacks if
1303 * necessary. We need to run them before the task is visible
1304 * on the tasklist. */
1305 cgroup_fork_callbacks(p);
1306 cgroup_callbacks_done = 1;
1308 /* Need tasklist lock for parent etc handling! */
1309 write_lock_irq(&tasklist_lock);
1311 /* CLONE_PARENT re-uses the old parent */
1312 if (clone_flags & (CLONE_PARENT|CLONE_THREAD)) {
1313 p->real_parent = current->real_parent;
1314 p->parent_exec_id = current->parent_exec_id;
1316 p->real_parent = current;
1317 p->parent_exec_id = current->self_exec_id;
1320 spin_lock(¤t->sighand->siglock);
1323 * Process group and session signals need to be delivered to just the
1324 * parent before the fork or both the parent and the child after the
1325 * fork. Restart if a signal comes in before we add the new process to
1326 * it's process group.
1327 * A fatal signal pending means that current will exit, so the new
1328 * thread can't slip out of an OOM kill (or normal SIGKILL).
1330 recalc_sigpending();
1331 if (signal_pending(current)) {
1332 spin_unlock(¤t->sighand->siglock);
1333 write_unlock_irq(&tasklist_lock);
1334 retval = -ERESTARTNOINTR;
1335 goto bad_fork_free_pid;
1338 if (clone_flags & CLONE_THREAD) {
1339 current->signal->nr_threads++;
1340 atomic_inc(¤t->signal->live);
1341 atomic_inc(¤t->signal->sigcnt);
1342 p->group_leader = current->group_leader;
1343 list_add_tail_rcu(&p->thread_group, &p->group_leader->thread_group);
1346 if (likely(p->pid)) {
1347 tracehook_finish_clone(p, clone_flags, trace);
1349 if (thread_group_leader(p)) {
1350 if (is_child_reaper(pid))
1351 p->nsproxy->pid_ns->child_reaper = p;
1353 p->signal->leader_pid = pid;
1354 p->signal->tty = tty_kref_get(current->signal->tty);
1355 attach_pid(p, PIDTYPE_PGID, task_pgrp(current));
1356 attach_pid(p, PIDTYPE_SID, task_session(current));
1357 list_add_tail(&p->sibling, &p->real_parent->children);
1358 list_add_tail_rcu(&p->tasks, &init_task.tasks);
1359 __this_cpu_inc(process_counts);
1361 attach_pid(p, PIDTYPE_PID, pid);
1366 spin_unlock(¤t->sighand->siglock);
1367 write_unlock_irq(&tasklist_lock);
1368 proc_fork_connector(p);
1369 cgroup_post_fork(p);
1370 if (clone_flags & CLONE_THREAD)
1371 threadgroup_fork_read_unlock(current);
1376 if (pid != &init_struct_pid)
1378 bad_fork_cleanup_io:
1381 bad_fork_cleanup_namespaces:
1382 if (unlikely(clone_flags & CLONE_NEWPID))
1383 pid_ns_release_proc(p->nsproxy->pid_ns);
1384 exit_task_namespaces(p);
1385 bad_fork_cleanup_mm:
1388 if (p->signal->oom_score_adj == OOM_SCORE_ADJ_MIN)
1389 atomic_dec(&p->mm->oom_disable_count);
1393 bad_fork_cleanup_signal:
1394 if (!(clone_flags & CLONE_THREAD))
1395 free_signal_struct(p->signal);
1396 bad_fork_cleanup_sighand:
1397 __cleanup_sighand(p->sighand);
1398 bad_fork_cleanup_fs:
1399 exit_fs(p); /* blocking */
1400 bad_fork_cleanup_files:
1401 exit_files(p); /* blocking */
1402 bad_fork_cleanup_semundo:
1404 bad_fork_cleanup_audit:
1406 bad_fork_cleanup_policy:
1407 perf_event_free_task(p);
1409 mpol_put(p->mempolicy);
1410 bad_fork_cleanup_cgroup:
1412 if (clone_flags & CLONE_THREAD)
1413 threadgroup_fork_read_unlock(current);
1414 cgroup_exit(p, cgroup_callbacks_done);
1415 delayacct_tsk_free(p);
1416 module_put(task_thread_info(p)->exec_domain->module);
1417 bad_fork_cleanup_count:
1418 atomic_dec(&p->cred->user->processes);
1423 return ERR_PTR(retval);
1426 noinline struct pt_regs * __cpuinit __attribute__((weak)) idle_regs(struct pt_regs *regs)
1428 memset(regs, 0, sizeof(struct pt_regs));
1432 static inline void init_idle_pids(struct pid_link *links)
1436 for (type = PIDTYPE_PID; type < PIDTYPE_MAX; ++type) {
1437 INIT_HLIST_NODE(&links[type].node); /* not really needed */
1438 links[type].pid = &init_struct_pid;
1442 struct task_struct * __cpuinit fork_idle(int cpu)
1444 struct task_struct *task;
1445 struct pt_regs regs;
1447 task = copy_process(CLONE_VM, 0, idle_regs(®s), 0, NULL,
1448 &init_struct_pid, 0);
1449 if (!IS_ERR(task)) {
1450 init_idle_pids(task->pids);
1451 init_idle(task, cpu);
1458 * Ok, this is the main fork-routine.
1460 * It copies the process, and if successful kick-starts
1461 * it and waits for it to finish using the VM if required.
1463 long do_fork(unsigned long clone_flags,
1464 unsigned long stack_start,
1465 struct pt_regs *regs,
1466 unsigned long stack_size,
1467 int __user *parent_tidptr,
1468 int __user *child_tidptr)
1470 struct task_struct *p;
1475 * Do some preliminary argument and permissions checking before we
1476 * actually start allocating stuff
1478 if (clone_flags & CLONE_NEWUSER) {
1479 if (clone_flags & CLONE_THREAD)
1481 /* hopefully this check will go away when userns support is
1484 if (!capable(CAP_SYS_ADMIN) || !capable(CAP_SETUID) ||
1485 !capable(CAP_SETGID))
1490 * When called from kernel_thread, don't do user tracing stuff.
1492 if (likely(user_mode(regs)))
1493 trace = tracehook_prepare_clone(clone_flags);
1495 p = copy_process(clone_flags, stack_start, regs, stack_size,
1496 child_tidptr, NULL, trace);
1498 * Do this prior waking up the new thread - the thread pointer
1499 * might get invalid after that point, if the thread exits quickly.
1502 struct completion vfork;
1504 trace_sched_process_fork(current, p);
1506 nr = task_pid_vnr(p);
1508 if (clone_flags & CLONE_PARENT_SETTID)
1509 put_user(nr, parent_tidptr);
1511 if (clone_flags & CLONE_VFORK) {
1512 p->vfork_done = &vfork;
1513 init_completion(&vfork);
1516 audit_finish_fork(p);
1517 tracehook_report_clone(regs, clone_flags, nr, p);
1520 * We set PF_STARTING at creation in case tracing wants to
1521 * use this to distinguish a fully live task from one that
1522 * hasn't gotten to tracehook_report_clone() yet. Now we
1523 * clear it and set the child going.
1525 p->flags &= ~PF_STARTING;
1527 wake_up_new_task(p);
1529 tracehook_report_clone_complete(trace, regs,
1530 clone_flags, nr, p);
1532 if (clone_flags & CLONE_VFORK) {
1533 freezer_do_not_count();
1534 wait_for_completion(&vfork);
1536 tracehook_report_vfork_done(p, nr);
1544 #ifndef ARCH_MIN_MMSTRUCT_ALIGN
1545 #define ARCH_MIN_MMSTRUCT_ALIGN 0
1548 static void sighand_ctor(void *data)
1550 struct sighand_struct *sighand = data;
1552 spin_lock_init(&sighand->siglock);
1553 init_waitqueue_head(&sighand->signalfd_wqh);
1556 void __init proc_caches_init(void)
1558 sighand_cachep = kmem_cache_create("sighand_cache",
1559 sizeof(struct sighand_struct), 0,
1560 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_DESTROY_BY_RCU|
1561 SLAB_NOTRACK, sighand_ctor);
1562 signal_cachep = kmem_cache_create("signal_cache",
1563 sizeof(struct signal_struct), 0,
1564 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL);
1565 files_cachep = kmem_cache_create("files_cache",
1566 sizeof(struct files_struct), 0,
1567 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL);
1568 fs_cachep = kmem_cache_create("fs_cache",
1569 sizeof(struct fs_struct), 0,
1570 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL);
1572 * FIXME! The "sizeof(struct mm_struct)" currently includes the
1573 * whole struct cpumask for the OFFSTACK case. We could change
1574 * this to *only* allocate as much of it as required by the
1575 * maximum number of CPU's we can ever have. The cpumask_allocation
1576 * is at the end of the structure, exactly for that reason.
1578 mm_cachep = kmem_cache_create("mm_struct",
1579 sizeof(struct mm_struct), ARCH_MIN_MMSTRUCT_ALIGN,
1580 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL);
1581 vm_area_cachep = KMEM_CACHE(vm_area_struct, SLAB_PANIC);
1586 * Check constraints on flags passed to the unshare system call.
1588 static int check_unshare_flags(unsigned long unshare_flags)
1590 if (unshare_flags & ~(CLONE_THREAD|CLONE_FS|CLONE_NEWNS|CLONE_SIGHAND|
1591 CLONE_VM|CLONE_FILES|CLONE_SYSVSEM|
1592 CLONE_NEWUTS|CLONE_NEWIPC|CLONE_NEWNET))
1595 * Not implemented, but pretend it works if there is nothing to
1596 * unshare. Note that unsharing CLONE_THREAD or CLONE_SIGHAND
1597 * needs to unshare vm.
1599 if (unshare_flags & (CLONE_THREAD | CLONE_SIGHAND | CLONE_VM)) {
1600 /* FIXME: get_task_mm() increments ->mm_users */
1601 if (atomic_read(¤t->mm->mm_users) > 1)
1609 * Unshare the filesystem structure if it is being shared
1611 static int unshare_fs(unsigned long unshare_flags, struct fs_struct **new_fsp)
1613 struct fs_struct *fs = current->fs;
1615 if (!(unshare_flags & CLONE_FS) || !fs)
1618 /* don't need lock here; in the worst case we'll do useless copy */
1622 *new_fsp = copy_fs_struct(fs);
1630 * Unshare file descriptor table if it is being shared
1632 static int unshare_fd(unsigned long unshare_flags, struct files_struct **new_fdp)
1634 struct files_struct *fd = current->files;
1637 if ((unshare_flags & CLONE_FILES) &&
1638 (fd && atomic_read(&fd->count) > 1)) {
1639 *new_fdp = dup_fd(fd, &error);
1648 * unshare allows a process to 'unshare' part of the process
1649 * context which was originally shared using clone. copy_*
1650 * functions used by do_fork() cannot be used here directly
1651 * because they modify an inactive task_struct that is being
1652 * constructed. Here we are modifying the current, active,
1655 SYSCALL_DEFINE1(unshare, unsigned long, unshare_flags)
1657 struct fs_struct *fs, *new_fs = NULL;
1658 struct files_struct *fd, *new_fd = NULL;
1659 struct nsproxy *new_nsproxy = NULL;
1663 err = check_unshare_flags(unshare_flags);
1665 goto bad_unshare_out;
1668 * If unsharing namespace, must also unshare filesystem information.
1670 if (unshare_flags & CLONE_NEWNS)
1671 unshare_flags |= CLONE_FS;
1673 * CLONE_NEWIPC must also detach from the undolist: after switching
1674 * to a new ipc namespace, the semaphore arrays from the old
1675 * namespace are unreachable.
1677 if (unshare_flags & (CLONE_NEWIPC|CLONE_SYSVSEM))
1679 if ((err = unshare_fs(unshare_flags, &new_fs)))
1680 goto bad_unshare_out;
1681 if ((err = unshare_fd(unshare_flags, &new_fd)))
1682 goto bad_unshare_cleanup_fs;
1683 if ((err = unshare_nsproxy_namespaces(unshare_flags, &new_nsproxy,
1685 goto bad_unshare_cleanup_fd;
1687 if (new_fs || new_fd || do_sysvsem || new_nsproxy) {
1690 * CLONE_SYSVSEM is equivalent to sys_exit().
1696 switch_task_namespaces(current, new_nsproxy);
1704 spin_lock(&fs->lock);
1705 current->fs = new_fs;
1710 spin_unlock(&fs->lock);
1714 fd = current->files;
1715 current->files = new_fd;
1719 task_unlock(current);
1723 put_nsproxy(new_nsproxy);
1725 bad_unshare_cleanup_fd:
1727 put_files_struct(new_fd);
1729 bad_unshare_cleanup_fs:
1731 free_fs_struct(new_fs);
1738 * Helper to unshare the files of the current task.
1739 * We don't want to expose copy_files internals to
1740 * the exec layer of the kernel.
1743 int unshare_files(struct files_struct **displaced)
1745 struct task_struct *task = current;
1746 struct files_struct *copy = NULL;
1749 error = unshare_fd(CLONE_FILES, ©);
1750 if (error || !copy) {
1754 *displaced = task->files;