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 /* Notifier list called when a task struct is freed */
160 static ATOMIC_NOTIFIER_HEAD(task_free_notifier);
162 static void account_kernel_stack(struct thread_info *ti, int account)
164 struct zone *zone = page_zone(virt_to_page(ti));
166 mod_zone_page_state(zone, NR_KERNEL_STACK, account);
169 void free_task(struct task_struct *tsk)
171 prop_local_destroy_single(&tsk->dirties);
172 account_kernel_stack(tsk->stack, -1);
173 free_thread_info(tsk->stack);
174 rt_mutex_debug_task_free(tsk);
175 ftrace_graph_exit_task(tsk);
176 free_task_struct(tsk);
178 EXPORT_SYMBOL(free_task);
180 static inline void free_signal_struct(struct signal_struct *sig)
182 taskstats_tgid_free(sig);
183 sched_autogroup_exit(sig);
184 kmem_cache_free(signal_cachep, sig);
187 static inline void put_signal_struct(struct signal_struct *sig)
189 if (atomic_dec_and_test(&sig->sigcnt))
190 free_signal_struct(sig);
193 int task_free_register(struct notifier_block *n)
195 return atomic_notifier_chain_register(&task_free_notifier, n);
197 EXPORT_SYMBOL(task_free_register);
199 int task_free_unregister(struct notifier_block *n)
201 return atomic_notifier_chain_unregister(&task_free_notifier, n);
203 EXPORT_SYMBOL(task_free_unregister);
205 void __put_task_struct(struct task_struct *tsk)
207 WARN_ON(!tsk->exit_state);
208 WARN_ON(atomic_read(&tsk->usage));
209 WARN_ON(tsk == current);
212 delayacct_tsk_free(tsk);
213 put_signal_struct(tsk->signal);
215 atomic_notifier_call_chain(&task_free_notifier, 0, tsk);
216 if (!profile_handoff_task(tsk))
219 EXPORT_SYMBOL_GPL(__put_task_struct);
222 * macro override instead of weak attribute alias, to workaround
223 * gcc 4.1.0 and 4.1.1 bugs with weak attribute and empty functions.
225 #ifndef arch_task_cache_init
226 #define arch_task_cache_init()
229 void __init fork_init(unsigned long mempages)
231 #ifndef __HAVE_ARCH_TASK_STRUCT_ALLOCATOR
232 #ifndef ARCH_MIN_TASKALIGN
233 #define ARCH_MIN_TASKALIGN L1_CACHE_BYTES
235 /* create a slab on which task_structs can be allocated */
237 kmem_cache_create("task_struct", sizeof(struct task_struct),
238 ARCH_MIN_TASKALIGN, SLAB_PANIC | SLAB_NOTRACK, NULL);
241 /* do the arch specific task caches init */
242 arch_task_cache_init();
245 * The default maximum number of threads is set to a safe
246 * value: the thread structures can take up at most half
249 max_threads = mempages / (8 * THREAD_SIZE / PAGE_SIZE);
252 * we need to allow at least 20 threads to boot a system
257 init_task.signal->rlim[RLIMIT_NPROC].rlim_cur = max_threads/2;
258 init_task.signal->rlim[RLIMIT_NPROC].rlim_max = max_threads/2;
259 init_task.signal->rlim[RLIMIT_SIGPENDING] =
260 init_task.signal->rlim[RLIMIT_NPROC];
263 int __attribute__((weak)) arch_dup_task_struct(struct task_struct *dst,
264 struct task_struct *src)
270 static struct task_struct *dup_task_struct(struct task_struct *orig)
272 struct task_struct *tsk;
273 struct thread_info *ti;
274 unsigned long *stackend;
275 int node = tsk_fork_get_node(orig);
278 prepare_to_copy(orig);
280 tsk = alloc_task_struct_node(node);
284 ti = alloc_thread_info_node(tsk, node);
286 free_task_struct(tsk);
290 err = arch_dup_task_struct(tsk, orig);
296 err = prop_local_init_single(&tsk->dirties);
300 setup_thread_stack(tsk, orig);
301 clear_user_return_notifier(tsk);
302 clear_tsk_need_resched(tsk);
303 stackend = end_of_stack(tsk);
304 *stackend = STACK_END_MAGIC; /* for overflow detection */
306 #ifdef CONFIG_CC_STACKPROTECTOR
307 tsk->stack_canary = get_random_int();
310 /* One for us, one for whoever does the "release_task()" (usually parent) */
311 atomic_set(&tsk->usage,2);
312 atomic_set(&tsk->fs_excl, 0);
313 #ifdef CONFIG_BLK_DEV_IO_TRACE
316 tsk->splice_pipe = NULL;
318 account_kernel_stack(ti, 1);
323 free_thread_info(ti);
324 free_task_struct(tsk);
329 static int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm)
331 struct vm_area_struct *mpnt, *tmp, *prev, **pprev;
332 struct rb_node **rb_link, *rb_parent;
334 unsigned long charge;
335 struct mempolicy *pol;
337 down_write(&oldmm->mmap_sem);
338 flush_cache_dup_mm(oldmm);
340 * Not linked in yet - no deadlock potential:
342 down_write_nested(&mm->mmap_sem, SINGLE_DEPTH_NESTING);
346 mm->mmap_cache = NULL;
347 mm->free_area_cache = oldmm->mmap_base;
348 mm->cached_hole_size = ~0UL;
350 cpumask_clear(mm_cpumask(mm));
352 rb_link = &mm->mm_rb.rb_node;
355 retval = ksm_fork(mm, oldmm);
358 retval = khugepaged_fork(mm, oldmm);
363 for (mpnt = oldmm->mmap; mpnt; mpnt = mpnt->vm_next) {
366 if (mpnt->vm_flags & VM_DONTCOPY) {
367 long pages = vma_pages(mpnt);
368 mm->total_vm -= pages;
369 vm_stat_account(mm, mpnt->vm_flags, mpnt->vm_file,
374 if (mpnt->vm_flags & VM_ACCOUNT) {
375 unsigned int len = (mpnt->vm_end - mpnt->vm_start) >> PAGE_SHIFT;
376 if (security_vm_enough_memory(len))
380 tmp = kmem_cache_alloc(vm_area_cachep, GFP_KERNEL);
384 INIT_LIST_HEAD(&tmp->anon_vma_chain);
385 pol = mpol_dup(vma_policy(mpnt));
386 retval = PTR_ERR(pol);
388 goto fail_nomem_policy;
389 vma_set_policy(tmp, pol);
391 if (anon_vma_fork(tmp, mpnt))
392 goto fail_nomem_anon_vma_fork;
393 tmp->vm_flags &= ~VM_LOCKED;
394 tmp->vm_next = tmp->vm_prev = NULL;
397 struct inode *inode = file->f_path.dentry->d_inode;
398 struct address_space *mapping = file->f_mapping;
401 if (tmp->vm_flags & VM_DENYWRITE)
402 atomic_dec(&inode->i_writecount);
403 mutex_lock(&mapping->i_mmap_mutex);
404 if (tmp->vm_flags & VM_SHARED)
405 mapping->i_mmap_writable++;
406 flush_dcache_mmap_lock(mapping);
407 /* insert tmp into the share list, just after mpnt */
408 vma_prio_tree_add(tmp, mpnt);
409 flush_dcache_mmap_unlock(mapping);
410 mutex_unlock(&mapping->i_mmap_mutex);
414 * Clear hugetlb-related page reserves for children. This only
415 * affects MAP_PRIVATE mappings. Faults generated by the child
416 * are not guaranteed to succeed, even if read-only
418 if (is_vm_hugetlb_page(tmp))
419 reset_vma_resv_huge_pages(tmp);
422 * Link in the new vma and copy the page table entries.
425 pprev = &tmp->vm_next;
429 __vma_link_rb(mm, tmp, rb_link, rb_parent);
430 rb_link = &tmp->vm_rb.rb_right;
431 rb_parent = &tmp->vm_rb;
434 retval = copy_page_range(mm, oldmm, mpnt);
436 if (tmp->vm_ops && tmp->vm_ops->open)
437 tmp->vm_ops->open(tmp);
442 /* a new mm has just been created */
443 arch_dup_mmap(oldmm, mm);
446 up_write(&mm->mmap_sem);
448 up_write(&oldmm->mmap_sem);
450 fail_nomem_anon_vma_fork:
453 kmem_cache_free(vm_area_cachep, tmp);
456 vm_unacct_memory(charge);
460 static inline int mm_alloc_pgd(struct mm_struct * mm)
462 mm->pgd = pgd_alloc(mm);
463 if (unlikely(!mm->pgd))
468 static inline void mm_free_pgd(struct mm_struct * mm)
470 pgd_free(mm, mm->pgd);
473 #define dup_mmap(mm, oldmm) (0)
474 #define mm_alloc_pgd(mm) (0)
475 #define mm_free_pgd(mm)
476 #endif /* CONFIG_MMU */
478 __cacheline_aligned_in_smp DEFINE_SPINLOCK(mmlist_lock);
480 #define allocate_mm() (kmem_cache_alloc(mm_cachep, GFP_KERNEL))
481 #define free_mm(mm) (kmem_cache_free(mm_cachep, (mm)))
483 static unsigned long default_dump_filter = MMF_DUMP_FILTER_DEFAULT;
485 static int __init coredump_filter_setup(char *s)
487 default_dump_filter =
488 (simple_strtoul(s, NULL, 0) << MMF_DUMP_FILTER_SHIFT) &
489 MMF_DUMP_FILTER_MASK;
493 __setup("coredump_filter=", coredump_filter_setup);
495 #include <linux/init_task.h>
497 static void mm_init_aio(struct mm_struct *mm)
500 spin_lock_init(&mm->ioctx_lock);
501 INIT_HLIST_HEAD(&mm->ioctx_list);
505 static struct mm_struct * mm_init(struct mm_struct * mm, struct task_struct *p)
507 atomic_set(&mm->mm_users, 1);
508 atomic_set(&mm->mm_count, 1);
509 init_rwsem(&mm->mmap_sem);
510 INIT_LIST_HEAD(&mm->mmlist);
511 mm->flags = (current->mm) ?
512 (current->mm->flags & MMF_INIT_MASK) : default_dump_filter;
513 mm->core_state = NULL;
515 memset(&mm->rss_stat, 0, sizeof(mm->rss_stat));
516 spin_lock_init(&mm->page_table_lock);
517 mm->free_area_cache = TASK_UNMAPPED_BASE;
518 mm->cached_hole_size = ~0UL;
520 mm_init_owner(mm, p);
521 atomic_set(&mm->oom_disable_count, 0);
523 if (likely(!mm_alloc_pgd(mm))) {
525 mmu_notifier_mm_init(mm);
534 * Allocate and initialize an mm_struct.
536 struct mm_struct * mm_alloc(void)
538 struct mm_struct * mm;
544 memset(mm, 0, sizeof(*mm));
546 return mm_init(mm, current);
550 * Called when the last reference to the mm
551 * is dropped: either by a lazy thread or by
552 * mmput. Free the page directory and the mm.
554 void __mmdrop(struct mm_struct *mm)
556 BUG_ON(mm == &init_mm);
559 mmu_notifier_mm_destroy(mm);
560 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
561 VM_BUG_ON(mm->pmd_huge_pte);
565 EXPORT_SYMBOL_GPL(__mmdrop);
568 * Decrement the use count and release all resources for an mm.
570 void mmput(struct mm_struct *mm)
574 if (atomic_dec_and_test(&mm->mm_users)) {
577 khugepaged_exit(mm); /* must run before exit_mmap */
579 set_mm_exe_file(mm, NULL);
580 if (!list_empty(&mm->mmlist)) {
581 spin_lock(&mmlist_lock);
582 list_del(&mm->mmlist);
583 spin_unlock(&mmlist_lock);
587 module_put(mm->binfmt->module);
591 EXPORT_SYMBOL_GPL(mmput);
594 * We added or removed a vma mapping the executable. The vmas are only mapped
595 * during exec and are not mapped with the mmap system call.
596 * Callers must hold down_write() on the mm's mmap_sem for these
598 void added_exe_file_vma(struct mm_struct *mm)
600 mm->num_exe_file_vmas++;
603 void removed_exe_file_vma(struct mm_struct *mm)
605 mm->num_exe_file_vmas--;
606 if ((mm->num_exe_file_vmas == 0) && mm->exe_file){
613 void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file)
616 get_file(new_exe_file);
619 mm->exe_file = new_exe_file;
620 mm->num_exe_file_vmas = 0;
623 struct file *get_mm_exe_file(struct mm_struct *mm)
625 struct file *exe_file;
627 /* We need mmap_sem to protect against races with removal of
628 * VM_EXECUTABLE vmas */
629 down_read(&mm->mmap_sem);
630 exe_file = mm->exe_file;
633 up_read(&mm->mmap_sem);
637 static void dup_mm_exe_file(struct mm_struct *oldmm, struct mm_struct *newmm)
639 /* It's safe to write the exe_file pointer without exe_file_lock because
640 * this is called during fork when the task is not yet in /proc */
641 newmm->exe_file = get_mm_exe_file(oldmm);
645 * get_task_mm - acquire a reference to the task's mm
647 * Returns %NULL if the task has no mm. Checks PF_KTHREAD (meaning
648 * this kernel workthread has transiently adopted a user mm with use_mm,
649 * to do its AIO) is not set and if so returns a reference to it, after
650 * bumping up the use count. User must release the mm via mmput()
651 * after use. Typically used by /proc and ptrace.
653 struct mm_struct *get_task_mm(struct task_struct *task)
655 struct mm_struct *mm;
660 if (task->flags & PF_KTHREAD)
663 atomic_inc(&mm->mm_users);
668 EXPORT_SYMBOL_GPL(get_task_mm);
670 /* Please note the differences between mmput and mm_release.
671 * mmput is called whenever we stop holding onto a mm_struct,
672 * error success whatever.
674 * mm_release is called after a mm_struct has been removed
675 * from the current process.
677 * This difference is important for error handling, when we
678 * only half set up a mm_struct for a new process and need to restore
679 * the old one. Because we mmput the new mm_struct before
680 * restoring the old one. . .
681 * Eric Biederman 10 January 1998
683 void mm_release(struct task_struct *tsk, struct mm_struct *mm)
685 struct completion *vfork_done = tsk->vfork_done;
687 /* Get rid of any futexes when releasing the mm */
689 if (unlikely(tsk->robust_list)) {
690 exit_robust_list(tsk);
691 tsk->robust_list = NULL;
694 if (unlikely(tsk->compat_robust_list)) {
695 compat_exit_robust_list(tsk);
696 tsk->compat_robust_list = NULL;
699 if (unlikely(!list_empty(&tsk->pi_state_list)))
700 exit_pi_state_list(tsk);
703 /* Get rid of any cached register state */
704 deactivate_mm(tsk, mm);
706 /* notify parent sleeping on vfork() */
708 tsk->vfork_done = NULL;
709 complete(vfork_done);
713 * If we're exiting normally, clear a user-space tid field if
714 * requested. We leave this alone when dying by signal, to leave
715 * the value intact in a core dump, and to save the unnecessary
716 * trouble otherwise. Userland only wants this done for a sys_exit.
718 if (tsk->clear_child_tid) {
719 if (!(tsk->flags & PF_SIGNALED) &&
720 atomic_read(&mm->mm_users) > 1) {
722 * We don't check the error code - if userspace has
723 * not set up a proper pointer then tough luck.
725 put_user(0, tsk->clear_child_tid);
726 sys_futex(tsk->clear_child_tid, FUTEX_WAKE,
729 tsk->clear_child_tid = NULL;
734 * Allocate a new mm structure and copy contents from the
735 * mm structure of the passed in task structure.
737 struct mm_struct *dup_mm(struct task_struct *tsk)
739 struct mm_struct *mm, *oldmm = current->mm;
749 memcpy(mm, oldmm, sizeof(*mm));
752 /* Initializing for Swap token stuff */
753 mm->token_priority = 0;
754 mm->last_interval = 0;
756 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
757 mm->pmd_huge_pte = NULL;
760 if (!mm_init(mm, tsk))
763 if (init_new_context(tsk, mm))
766 dup_mm_exe_file(oldmm, mm);
768 err = dup_mmap(mm, oldmm);
772 mm->hiwater_rss = get_mm_rss(mm);
773 mm->hiwater_vm = mm->total_vm;
775 if (mm->binfmt && !try_module_get(mm->binfmt->module))
781 /* don't put binfmt in mmput, we haven't got module yet */
790 * If init_new_context() failed, we cannot use mmput() to free the mm
791 * because it calls destroy_context()
798 static int copy_mm(unsigned long clone_flags, struct task_struct * tsk)
800 struct mm_struct * mm, *oldmm;
803 tsk->min_flt = tsk->maj_flt = 0;
804 tsk->nvcsw = tsk->nivcsw = 0;
805 #ifdef CONFIG_DETECT_HUNG_TASK
806 tsk->last_switch_count = tsk->nvcsw + tsk->nivcsw;
810 tsk->active_mm = NULL;
813 * Are we cloning a kernel thread?
815 * We need to steal a active VM for that..
821 if (clone_flags & CLONE_VM) {
822 atomic_inc(&oldmm->mm_users);
833 /* Initializing for Swap token stuff */
834 mm->token_priority = 0;
835 mm->last_interval = 0;
836 if (tsk->signal->oom_score_adj == OOM_SCORE_ADJ_MIN)
837 atomic_inc(&mm->oom_disable_count);
847 static int copy_fs(unsigned long clone_flags, struct task_struct *tsk)
849 struct fs_struct *fs = current->fs;
850 if (clone_flags & CLONE_FS) {
851 /* tsk->fs is already what we want */
852 spin_lock(&fs->lock);
854 spin_unlock(&fs->lock);
858 spin_unlock(&fs->lock);
861 tsk->fs = copy_fs_struct(fs);
867 static int copy_files(unsigned long clone_flags, struct task_struct * tsk)
869 struct files_struct *oldf, *newf;
873 * A background process may not have any files ...
875 oldf = current->files;
879 if (clone_flags & CLONE_FILES) {
880 atomic_inc(&oldf->count);
884 newf = dup_fd(oldf, &error);
894 static int copy_io(unsigned long clone_flags, struct task_struct *tsk)
897 struct io_context *ioc = current->io_context;
902 * Share io context with parent, if CLONE_IO is set
904 if (clone_flags & CLONE_IO) {
905 tsk->io_context = ioc_task_link(ioc);
906 if (unlikely(!tsk->io_context))
908 } else if (ioprio_valid(ioc->ioprio)) {
909 tsk->io_context = alloc_io_context(GFP_KERNEL, -1);
910 if (unlikely(!tsk->io_context))
913 tsk->io_context->ioprio = ioc->ioprio;
919 static int copy_sighand(unsigned long clone_flags, struct task_struct *tsk)
921 struct sighand_struct *sig;
923 if (clone_flags & CLONE_SIGHAND) {
924 atomic_inc(¤t->sighand->count);
927 sig = kmem_cache_alloc(sighand_cachep, GFP_KERNEL);
928 rcu_assign_pointer(tsk->sighand, sig);
931 atomic_set(&sig->count, 1);
932 memcpy(sig->action, current->sighand->action, sizeof(sig->action));
936 void __cleanup_sighand(struct sighand_struct *sighand)
938 if (atomic_dec_and_test(&sighand->count)) {
939 signalfd_cleanup(sighand);
940 kmem_cache_free(sighand_cachep, sighand);
946 * Initialize POSIX timer handling for a thread group.
948 static void posix_cpu_timers_init_group(struct signal_struct *sig)
950 unsigned long cpu_limit;
952 /* Thread group counters. */
953 thread_group_cputime_init(sig);
955 cpu_limit = ACCESS_ONCE(sig->rlim[RLIMIT_CPU].rlim_cur);
956 if (cpu_limit != RLIM_INFINITY) {
957 sig->cputime_expires.prof_exp = secs_to_cputime(cpu_limit);
958 sig->cputimer.running = 1;
961 /* The timer lists. */
962 INIT_LIST_HEAD(&sig->cpu_timers[0]);
963 INIT_LIST_HEAD(&sig->cpu_timers[1]);
964 INIT_LIST_HEAD(&sig->cpu_timers[2]);
967 static int copy_signal(unsigned long clone_flags, struct task_struct *tsk)
969 struct signal_struct *sig;
971 if (clone_flags & CLONE_THREAD)
974 sig = kmem_cache_zalloc(signal_cachep, GFP_KERNEL);
980 atomic_set(&sig->live, 1);
981 atomic_set(&sig->sigcnt, 1);
982 init_waitqueue_head(&sig->wait_chldexit);
983 if (clone_flags & CLONE_NEWPID)
984 sig->flags |= SIGNAL_UNKILLABLE;
985 sig->curr_target = tsk;
986 init_sigpending(&sig->shared_pending);
987 INIT_LIST_HEAD(&sig->posix_timers);
989 hrtimer_init(&sig->real_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
990 sig->real_timer.function = it_real_fn;
992 task_lock(current->group_leader);
993 memcpy(sig->rlim, current->signal->rlim, sizeof sig->rlim);
994 task_unlock(current->group_leader);
996 posix_cpu_timers_init_group(sig);
999 sched_autogroup_fork(sig);
1001 #ifdef CONFIG_CGROUPS
1002 init_rwsem(&sig->threadgroup_fork_lock);
1004 #ifdef CONFIG_CPUSETS
1005 seqcount_init(&tsk->mems_allowed_seq);
1008 sig->oom_adj = current->signal->oom_adj;
1009 sig->oom_score_adj = current->signal->oom_score_adj;
1010 sig->oom_score_adj_min = current->signal->oom_score_adj_min;
1012 mutex_init(&sig->cred_guard_mutex);
1017 static void copy_flags(unsigned long clone_flags, struct task_struct *p)
1019 unsigned long new_flags = p->flags;
1021 new_flags &= ~(PF_SUPERPRIV | PF_WQ_WORKER);
1022 new_flags |= PF_FORKNOEXEC;
1023 new_flags |= PF_STARTING;
1024 p->flags = new_flags;
1025 clear_freeze_flag(p);
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 = cputime_zero;
1057 tsk->cputime_expires.virt_exp = cputime_zero;
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)
1137 retval = copy_creds(p, clone_flags);
1142 * If multiple threads are within copy_process(), then this check
1143 * triggers too late. This doesn't hurt, the check is only there
1144 * to stop root fork bombs.
1147 if (nr_threads >= max_threads)
1148 goto bad_fork_cleanup_count;
1150 if (!try_module_get(task_thread_info(p)->exec_domain->module))
1151 goto bad_fork_cleanup_count;
1154 delayacct_tsk_init(p); /* Must remain after dup_task_struct() */
1155 copy_flags(clone_flags, p);
1156 INIT_LIST_HEAD(&p->children);
1157 INIT_LIST_HEAD(&p->sibling);
1158 rcu_copy_process(p);
1159 p->vfork_done = NULL;
1160 spin_lock_init(&p->alloc_lock);
1162 init_sigpending(&p->pending);
1164 p->utime = cputime_zero;
1165 p->stime = cputime_zero;
1166 p->gtime = cputime_zero;
1167 p->utimescaled = cputime_zero;
1168 p->stimescaled = cputime_zero;
1169 #ifndef CONFIG_VIRT_CPU_ACCOUNTING
1170 p->prev_utime = cputime_zero;
1171 p->prev_stime = cputime_zero;
1173 #if defined(SPLIT_RSS_COUNTING)
1174 memset(&p->rss_stat, 0, sizeof(p->rss_stat));
1177 p->default_timer_slack_ns = current->timer_slack_ns;
1179 task_io_accounting_init(&p->ioac);
1180 acct_clear_integrals(p);
1182 posix_cpu_timers_init(p);
1184 do_posix_clock_monotonic_gettime(&p->start_time);
1185 p->real_start_time = p->start_time;
1186 monotonic_to_bootbased(&p->real_start_time);
1187 p->io_context = NULL;
1188 p->audit_context = NULL;
1189 if (clone_flags & CLONE_THREAD)
1190 threadgroup_fork_read_lock(current);
1193 p->mempolicy = mpol_dup(p->mempolicy);
1194 if (IS_ERR(p->mempolicy)) {
1195 retval = PTR_ERR(p->mempolicy);
1196 p->mempolicy = NULL;
1197 goto bad_fork_cleanup_cgroup;
1199 mpol_fix_fork_child_flag(p);
1201 #ifdef CONFIG_TRACE_IRQFLAGS
1203 #ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW
1204 p->hardirqs_enabled = 1;
1206 p->hardirqs_enabled = 0;
1208 p->hardirq_enable_ip = 0;
1209 p->hardirq_enable_event = 0;
1210 p->hardirq_disable_ip = _THIS_IP_;
1211 p->hardirq_disable_event = 0;
1212 p->softirqs_enabled = 1;
1213 p->softirq_enable_ip = _THIS_IP_;
1214 p->softirq_enable_event = 0;
1215 p->softirq_disable_ip = 0;
1216 p->softirq_disable_event = 0;
1217 p->hardirq_context = 0;
1218 p->softirq_context = 0;
1220 #ifdef CONFIG_LOCKDEP
1221 p->lockdep_depth = 0; /* no locks held yet */
1222 p->curr_chain_key = 0;
1223 p->lockdep_recursion = 0;
1226 #ifdef CONFIG_DEBUG_MUTEXES
1227 p->blocked_on = NULL; /* not blocked yet */
1229 #ifdef CONFIG_CGROUP_MEM_RES_CTLR
1230 p->memcg_batch.do_batch = 0;
1231 p->memcg_batch.memcg = NULL;
1234 /* Perform scheduler related setup. Assign this task to a CPU. */
1237 retval = perf_event_init_task(p);
1239 goto bad_fork_cleanup_policy;
1241 if ((retval = audit_alloc(p)))
1242 goto bad_fork_cleanup_policy;
1243 /* copy all the process information */
1244 if ((retval = copy_semundo(clone_flags, p)))
1245 goto bad_fork_cleanup_audit;
1246 if ((retval = copy_files(clone_flags, p)))
1247 goto bad_fork_cleanup_semundo;
1248 if ((retval = copy_fs(clone_flags, p)))
1249 goto bad_fork_cleanup_files;
1250 if ((retval = copy_sighand(clone_flags, p)))
1251 goto bad_fork_cleanup_fs;
1252 if ((retval = copy_signal(clone_flags, p)))
1253 goto bad_fork_cleanup_sighand;
1254 if ((retval = copy_mm(clone_flags, p)))
1255 goto bad_fork_cleanup_signal;
1256 if ((retval = copy_namespaces(clone_flags, p)))
1257 goto bad_fork_cleanup_mm;
1258 if ((retval = copy_io(clone_flags, p)))
1259 goto bad_fork_cleanup_namespaces;
1260 retval = copy_thread(clone_flags, stack_start, stack_size, p, regs);
1262 goto bad_fork_cleanup_io;
1264 if (pid != &init_struct_pid) {
1266 pid = alloc_pid(p->nsproxy->pid_ns);
1268 goto bad_fork_cleanup_io;
1271 p->pid = pid_nr(pid);
1273 if (clone_flags & CLONE_THREAD)
1274 p->tgid = current->tgid;
1276 p->set_child_tid = (clone_flags & CLONE_CHILD_SETTID) ? child_tidptr : NULL;
1278 * Clear TID on mm_release()?
1280 p->clear_child_tid = (clone_flags & CLONE_CHILD_CLEARTID) ? child_tidptr: NULL;
1285 p->robust_list = NULL;
1286 #ifdef CONFIG_COMPAT
1287 p->compat_robust_list = NULL;
1289 INIT_LIST_HEAD(&p->pi_state_list);
1290 p->pi_state_cache = NULL;
1293 * sigaltstack should be cleared when sharing the same VM
1295 if ((clone_flags & (CLONE_VM|CLONE_VFORK)) == CLONE_VM)
1296 p->sas_ss_sp = p->sas_ss_size = 0;
1299 * Syscall tracing and stepping should be turned off in the
1300 * child regardless of CLONE_PTRACE.
1302 user_disable_single_step(p);
1303 clear_tsk_thread_flag(p, TIF_SYSCALL_TRACE);
1304 #ifdef TIF_SYSCALL_EMU
1305 clear_tsk_thread_flag(p, TIF_SYSCALL_EMU);
1307 clear_all_latency_tracing(p);
1309 /* ok, now we should be set up.. */
1310 p->exit_signal = (clone_flags & CLONE_THREAD) ? -1 : (clone_flags & CSIGNAL);
1311 p->pdeath_signal = 0;
1315 * Ok, make it visible to the rest of the system.
1316 * We dont wake it up yet.
1318 p->group_leader = p;
1319 INIT_LIST_HEAD(&p->thread_group);
1321 /* Now that the task is set up, run cgroup callbacks if
1322 * necessary. We need to run them before the task is visible
1323 * on the tasklist. */
1324 cgroup_fork_callbacks(p);
1325 cgroup_callbacks_done = 1;
1327 /* Need tasklist lock for parent etc handling! */
1328 write_lock_irq(&tasklist_lock);
1330 /* CLONE_PARENT re-uses the old parent */
1331 if (clone_flags & (CLONE_PARENT|CLONE_THREAD)) {
1332 p->real_parent = current->real_parent;
1333 p->parent_exec_id = current->parent_exec_id;
1335 p->real_parent = current;
1336 p->parent_exec_id = current->self_exec_id;
1339 spin_lock(¤t->sighand->siglock);
1342 * Process group and session signals need to be delivered to just the
1343 * parent before the fork or both the parent and the child after the
1344 * fork. Restart if a signal comes in before we add the new process to
1345 * it's process group.
1346 * A fatal signal pending means that current will exit, so the new
1347 * thread can't slip out of an OOM kill (or normal SIGKILL).
1349 recalc_sigpending();
1350 if (signal_pending(current)) {
1351 spin_unlock(¤t->sighand->siglock);
1352 write_unlock_irq(&tasklist_lock);
1353 retval = -ERESTARTNOINTR;
1354 goto bad_fork_free_pid;
1357 if (clone_flags & CLONE_THREAD) {
1358 current->signal->nr_threads++;
1359 atomic_inc(¤t->signal->live);
1360 atomic_inc(¤t->signal->sigcnt);
1361 p->group_leader = current->group_leader;
1362 list_add_tail_rcu(&p->thread_group, &p->group_leader->thread_group);
1365 if (likely(p->pid)) {
1366 tracehook_finish_clone(p, clone_flags, trace);
1368 if (thread_group_leader(p)) {
1369 if (is_child_reaper(pid))
1370 p->nsproxy->pid_ns->child_reaper = p;
1372 p->signal->leader_pid = pid;
1373 p->signal->tty = tty_kref_get(current->signal->tty);
1374 attach_pid(p, PIDTYPE_PGID, task_pgrp(current));
1375 attach_pid(p, PIDTYPE_SID, task_session(current));
1376 list_add_tail(&p->sibling, &p->real_parent->children);
1377 list_add_tail_rcu(&p->tasks, &init_task.tasks);
1378 __this_cpu_inc(process_counts);
1380 attach_pid(p, PIDTYPE_PID, pid);
1385 spin_unlock(¤t->sighand->siglock);
1386 write_unlock_irq(&tasklist_lock);
1387 proc_fork_connector(p);
1388 cgroup_post_fork(p);
1389 if (clone_flags & CLONE_THREAD)
1390 threadgroup_fork_read_unlock(current);
1395 if (pid != &init_struct_pid)
1397 bad_fork_cleanup_io:
1400 bad_fork_cleanup_namespaces:
1401 if (unlikely(clone_flags & CLONE_NEWPID))
1402 pid_ns_release_proc(p->nsproxy->pid_ns);
1403 exit_task_namespaces(p);
1404 bad_fork_cleanup_mm:
1407 if (p->signal->oom_score_adj == OOM_SCORE_ADJ_MIN)
1408 atomic_dec(&p->mm->oom_disable_count);
1412 bad_fork_cleanup_signal:
1413 if (!(clone_flags & CLONE_THREAD))
1414 free_signal_struct(p->signal);
1415 bad_fork_cleanup_sighand:
1416 __cleanup_sighand(p->sighand);
1417 bad_fork_cleanup_fs:
1418 exit_fs(p); /* blocking */
1419 bad_fork_cleanup_files:
1420 exit_files(p); /* blocking */
1421 bad_fork_cleanup_semundo:
1423 bad_fork_cleanup_audit:
1425 bad_fork_cleanup_policy:
1426 perf_event_free_task(p);
1428 mpol_put(p->mempolicy);
1429 bad_fork_cleanup_cgroup:
1431 if (clone_flags & CLONE_THREAD)
1432 threadgroup_fork_read_unlock(current);
1433 cgroup_exit(p, cgroup_callbacks_done);
1434 delayacct_tsk_free(p);
1435 module_put(task_thread_info(p)->exec_domain->module);
1436 bad_fork_cleanup_count:
1437 atomic_dec(&p->cred->user->processes);
1442 return ERR_PTR(retval);
1445 noinline struct pt_regs * __cpuinit __attribute__((weak)) idle_regs(struct pt_regs *regs)
1447 memset(regs, 0, sizeof(struct pt_regs));
1451 static inline void init_idle_pids(struct pid_link *links)
1455 for (type = PIDTYPE_PID; type < PIDTYPE_MAX; ++type) {
1456 INIT_HLIST_NODE(&links[type].node); /* not really needed */
1457 links[type].pid = &init_struct_pid;
1461 struct task_struct * __cpuinit fork_idle(int cpu)
1463 struct task_struct *task;
1464 struct pt_regs regs;
1466 task = copy_process(CLONE_VM, 0, idle_regs(®s), 0, NULL,
1467 &init_struct_pid, 0);
1468 if (!IS_ERR(task)) {
1469 init_idle_pids(task->pids);
1470 init_idle(task, cpu);
1477 * Ok, this is the main fork-routine.
1479 * It copies the process, and if successful kick-starts
1480 * it and waits for it to finish using the VM if required.
1482 long do_fork(unsigned long clone_flags,
1483 unsigned long stack_start,
1484 struct pt_regs *regs,
1485 unsigned long stack_size,
1486 int __user *parent_tidptr,
1487 int __user *child_tidptr)
1489 struct task_struct *p;
1494 * Do some preliminary argument and permissions checking before we
1495 * actually start allocating stuff
1497 if (clone_flags & CLONE_NEWUSER) {
1498 if (clone_flags & CLONE_THREAD)
1500 /* hopefully this check will go away when userns support is
1503 if (!capable(CAP_SYS_ADMIN) || !capable(CAP_SETUID) ||
1504 !capable(CAP_SETGID))
1509 * When called from kernel_thread, don't do user tracing stuff.
1511 if (likely(user_mode(regs)))
1512 trace = tracehook_prepare_clone(clone_flags);
1514 p = copy_process(clone_flags, stack_start, regs, stack_size,
1515 child_tidptr, NULL, trace);
1517 * Do this prior waking up the new thread - the thread pointer
1518 * might get invalid after that point, if the thread exits quickly.
1521 struct completion vfork;
1523 trace_sched_process_fork(current, p);
1525 nr = task_pid_vnr(p);
1527 if (clone_flags & CLONE_PARENT_SETTID)
1528 put_user(nr, parent_tidptr);
1530 if (clone_flags & CLONE_VFORK) {
1531 p->vfork_done = &vfork;
1532 init_completion(&vfork);
1535 audit_finish_fork(p);
1536 tracehook_report_clone(regs, clone_flags, nr, p);
1539 * We set PF_STARTING at creation in case tracing wants to
1540 * use this to distinguish a fully live task from one that
1541 * hasn't gotten to tracehook_report_clone() yet. Now we
1542 * clear it and set the child going.
1544 p->flags &= ~PF_STARTING;
1546 wake_up_new_task(p);
1548 tracehook_report_clone_complete(trace, regs,
1549 clone_flags, nr, p);
1551 if (clone_flags & CLONE_VFORK) {
1552 freezer_do_not_count();
1553 wait_for_completion(&vfork);
1555 tracehook_report_vfork_done(p, nr);
1563 #ifndef ARCH_MIN_MMSTRUCT_ALIGN
1564 #define ARCH_MIN_MMSTRUCT_ALIGN 0
1567 static void sighand_ctor(void *data)
1569 struct sighand_struct *sighand = data;
1571 spin_lock_init(&sighand->siglock);
1572 init_waitqueue_head(&sighand->signalfd_wqh);
1575 void __init proc_caches_init(void)
1577 sighand_cachep = kmem_cache_create("sighand_cache",
1578 sizeof(struct sighand_struct), 0,
1579 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_DESTROY_BY_RCU|
1580 SLAB_NOTRACK, sighand_ctor);
1581 signal_cachep = kmem_cache_create("signal_cache",
1582 sizeof(struct signal_struct), 0,
1583 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL);
1584 files_cachep = kmem_cache_create("files_cache",
1585 sizeof(struct files_struct), 0,
1586 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL);
1587 fs_cachep = kmem_cache_create("fs_cache",
1588 sizeof(struct fs_struct), 0,
1589 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL);
1591 * FIXME! The "sizeof(struct mm_struct)" currently includes the
1592 * whole struct cpumask for the OFFSTACK case. We could change
1593 * this to *only* allocate as much of it as required by the
1594 * maximum number of CPU's we can ever have. The cpumask_allocation
1595 * is at the end of the structure, exactly for that reason.
1597 mm_cachep = kmem_cache_create("mm_struct",
1598 sizeof(struct mm_struct), ARCH_MIN_MMSTRUCT_ALIGN,
1599 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL);
1600 vm_area_cachep = KMEM_CACHE(vm_area_struct, SLAB_PANIC);
1605 * Check constraints on flags passed to the unshare system call.
1607 static int check_unshare_flags(unsigned long unshare_flags)
1609 if (unshare_flags & ~(CLONE_THREAD|CLONE_FS|CLONE_NEWNS|CLONE_SIGHAND|
1610 CLONE_VM|CLONE_FILES|CLONE_SYSVSEM|
1611 CLONE_NEWUTS|CLONE_NEWIPC|CLONE_NEWNET))
1614 * Not implemented, but pretend it works if there is nothing to
1615 * unshare. Note that unsharing CLONE_THREAD or CLONE_SIGHAND
1616 * needs to unshare vm.
1618 if (unshare_flags & (CLONE_THREAD | CLONE_SIGHAND | CLONE_VM)) {
1619 /* FIXME: get_task_mm() increments ->mm_users */
1620 if (atomic_read(¤t->mm->mm_users) > 1)
1628 * Unshare the filesystem structure if it is being shared
1630 static int unshare_fs(unsigned long unshare_flags, struct fs_struct **new_fsp)
1632 struct fs_struct *fs = current->fs;
1634 if (!(unshare_flags & CLONE_FS) || !fs)
1637 /* don't need lock here; in the worst case we'll do useless copy */
1641 *new_fsp = copy_fs_struct(fs);
1649 * Unshare file descriptor table if it is being shared
1651 static int unshare_fd(unsigned long unshare_flags, struct files_struct **new_fdp)
1653 struct files_struct *fd = current->files;
1656 if ((unshare_flags & CLONE_FILES) &&
1657 (fd && atomic_read(&fd->count) > 1)) {
1658 *new_fdp = dup_fd(fd, &error);
1667 * unshare allows a process to 'unshare' part of the process
1668 * context which was originally shared using clone. copy_*
1669 * functions used by do_fork() cannot be used here directly
1670 * because they modify an inactive task_struct that is being
1671 * constructed. Here we are modifying the current, active,
1674 SYSCALL_DEFINE1(unshare, unsigned long, unshare_flags)
1676 struct fs_struct *fs, *new_fs = NULL;
1677 struct files_struct *fd, *new_fd = NULL;
1678 struct nsproxy *new_nsproxy = NULL;
1682 err = check_unshare_flags(unshare_flags);
1684 goto bad_unshare_out;
1687 * If unsharing namespace, must also unshare filesystem information.
1689 if (unshare_flags & CLONE_NEWNS)
1690 unshare_flags |= CLONE_FS;
1692 * CLONE_NEWIPC must also detach from the undolist: after switching
1693 * to a new ipc namespace, the semaphore arrays from the old
1694 * namespace are unreachable.
1696 if (unshare_flags & (CLONE_NEWIPC|CLONE_SYSVSEM))
1698 if ((err = unshare_fs(unshare_flags, &new_fs)))
1699 goto bad_unshare_out;
1700 if ((err = unshare_fd(unshare_flags, &new_fd)))
1701 goto bad_unshare_cleanup_fs;
1702 if ((err = unshare_nsproxy_namespaces(unshare_flags, &new_nsproxy,
1704 goto bad_unshare_cleanup_fd;
1706 if (new_fs || new_fd || do_sysvsem || new_nsproxy) {
1709 * CLONE_SYSVSEM is equivalent to sys_exit().
1715 switch_task_namespaces(current, new_nsproxy);
1723 spin_lock(&fs->lock);
1724 current->fs = new_fs;
1729 spin_unlock(&fs->lock);
1733 fd = current->files;
1734 current->files = new_fd;
1738 task_unlock(current);
1742 put_nsproxy(new_nsproxy);
1744 bad_unshare_cleanup_fd:
1746 put_files_struct(new_fd);
1748 bad_unshare_cleanup_fs:
1750 free_fs_struct(new_fs);
1757 * Helper to unshare the files of the current task.
1758 * We don't want to expose copy_files internals to
1759 * the exec layer of the kernel.
1762 int unshare_files(struct files_struct **displaced)
1764 struct task_struct *task = current;
1765 struct files_struct *copy = NULL;
1768 error = unshare_fd(CLONE_FILES, ©);
1769 if (error || !copy) {
1773 *displaced = task->files;